From a0651db490328a972185e44ff637970b3456406b Mon Sep 17 00:00:00 2001
From: Younan Zhang <zyn7109@gmail.com>
Date: Wed, 10 Apr 2024 19:23:32 +0800
Subject: [PATCH 001/886] [clang][Sema] Avoid guessing unexpanded packs' size
 in getFullyPackExpandedSize (#87768)

There has been an optimization for `SizeOfPackExprs` since c5452ed9, in
which
we overlooked a case where the template arguments were not yet
formed into a `PackExpansionType` at the token annotation stage. This
led to a problem in that a template involving such expressions may
lose its nature of being dependent, causing some false-positive
diagnostics.

Fixes https://github.com/llvm/llvm-project/issues/84220
---
 clang/docs/ReleaseNotes.rst                 |  1 +
 clang/lib/Sema/SemaTemplateVariadic.cpp     | 11 ++++++++++
 clang/test/SemaTemplate/alias-templates.cpp | 23 +++++++++++++++++++++
 3 files changed, 35 insertions(+)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index f96cebbde3d82..6bff80ed4d210 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -526,6 +526,7 @@ Bug Fixes to C++ Support
 - Fix crash when inheriting from a cv-qualified type. Fixes:
   (`#35603 <https://github.com/llvm/llvm-project/issues/35603>`_)
 - Fix a crash when the using enum declaration uses an anonymous enumeration. Fixes (#GH86790).
+- Handled an edge case in ``getFullyPackExpandedSize`` so that we now avoid a false-positive diagnostic. (#GH84220)
 - Clang now correctly tracks type dependence of by-value captures in lambdas with an explicit
   object parameter.
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
diff --git a/clang/lib/Sema/SemaTemplateVariadic.cpp b/clang/lib/Sema/SemaTemplateVariadic.cpp
index 903fbfd18e779..4909414c0c78d 100644
--- a/clang/lib/Sema/SemaTemplateVariadic.cpp
+++ b/clang/lib/Sema/SemaTemplateVariadic.cpp
@@ -1243,6 +1243,17 @@ std::optional<unsigned> Sema::getFullyPackExpandedSize(TemplateArgument Arg) {
     // expanded this pack expansion into the enclosing pack if we could.
     if (Elem.isPackExpansion())
       return std::nullopt;
+    // Don't guess the size of unexpanded packs. The pack within a template
+    // argument may have yet to be of a PackExpansion type before we see the
+    // ellipsis in the annotation stage.
+    //
+    // This doesn't mean we would invalidate the optimization: Arg can be an
+    // unexpanded pack regardless of Elem's dependence. For instance,
+    // A TemplateArgument that contains either a SubstTemplateTypeParmPackType
+    // or SubstNonTypeTemplateParmPackExpr is always considered Unexpanded, but
+    // the underlying TemplateArgument thereof may not.
+    if (Elem.containsUnexpandedParameterPack())
+      return std::nullopt;
   }
   return Pack.pack_size();
 }
diff --git a/clang/test/SemaTemplate/alias-templates.cpp b/clang/test/SemaTemplate/alias-templates.cpp
index 8d7cc6118610a..ab5cad72faf1b 100644
--- a/clang/test/SemaTemplate/alias-templates.cpp
+++ b/clang/test/SemaTemplate/alias-templates.cpp
@@ -236,6 +236,29 @@ namespace PR14858 {
   void test_q(int (&a)[5]) { Q<B, B, B>().f<B, B>(&a); }
 }
 
+namespace PR84220 {
+
+template <class...> class list {};
+
+template <int> struct foo_impl {
+  template <class> using f = int;
+};
+
+template <class... xs>
+using foo = typename foo_impl<sizeof...(xs)>::template f<xs...>;
+
+// We call getFullyPackExpandedSize at the annotation stage
+// before parsing the ellipsis next to the foo<xs>. This happens before
+// a PackExpansionType is formed for foo<xs>.
+// getFullyPackExpandedSize shouldn't determine the value here. Otherwise,
+// foo_impl<sizeof...(xs)> would lose its dependency despite the template
+// arguments being unsubstituted.
+template <class... xs> using test = list<foo<xs>...>;
+
+test<int> a;
+
+}
+
 namespace redecl {
   template<typename> using A = int;
   template<typename = void> using A = int;

From 8d206f51497fdf1ceebd6430b2f7d31ef735d0dc Mon Sep 17 00:00:00 2001
From: Edwin Vane <revane@google.com>
Date: Wed, 10 Apr 2024 07:40:35 -0400
Subject: [PATCH 002/886] [clang-tidy] Allow renaming macro arguments (#87792)

Although the identifier-naming.cpp lit test expected macro arguments not
to be renamed, the code seemed to already allow it. The code was simply
not being exercised because a SourceManager argument wasn't being
provided. With this change, renaming of macro arguments that expand to
renamable decls is permitted.
---
 .../utils/RenamerClangTidyCheck.cpp           |  2 +-
 clang-tools-extra/docs/ReleaseNotes.rst       |  2 +-
 .../readability/identifier-naming.cpp         | 23 +++++++++++++++----
 3 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
index 69b7d40ef628d..ad8048e2a92b7 100644
--- a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
+++ b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
@@ -489,7 +489,7 @@ void RenamerClangTidyCheck::checkNamedDecl(const NamedDecl *Decl,
   }
 
   Failure.Info = std::move(Info);
-  addUsage(Decl, Range);
+  addUsage(Decl, Range, &SourceMgr);
 }
 
 void RenamerClangTidyCheck::check(const MatchFinder::MatchResult &Result) {
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index a7193e90c38da..b66be44e9f8a6 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -268,7 +268,7 @@ Changes in existing checks
   <clang-tidy/checks/readability/identifier-naming>` check in `GetConfigPerFile`
   mode by resolving symbolic links to header files. Fixed handling of Hungarian
   Prefix when configured to `LowerCase`. Added support for renaming designated
-  initializers.
+  initializers. Added support for renaming macro arguments.
 
 - Improved :doc:`readability-implicit-bool-conversion
   <clang-tidy/checks/readability/implicit-bool-conversion>` check to provide
diff --git a/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming.cpp b/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming.cpp
index 57ef4aae5ddb7..99149fe86acee 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/readability/identifier-naming.cpp
@@ -108,10 +108,12 @@ USER_NS::object g_s2;
 // NO warnings or fixes expected as USER_NS and object are declared in a header file
 
 SYSTEM_MACRO(var1);
-// NO warnings or fixes expected as var1 is from macro expansion
+// CHECK-MESSAGES: :[[@LINE-1]]:14: warning: invalid case style for global variable 'var1' [readability-identifier-naming]
+// CHECK-FIXES: {{^}}SYSTEM_MACRO(g_var1);
 
 USER_MACRO(var2);
-// NO warnings or fixes expected as var2 is declared in a macro expansion
+// CHECK-MESSAGES: :[[@LINE-1]]:12: warning: invalid case style for global variable 'var2' [readability-identifier-naming]
+// CHECK-FIXES: {{^}}USER_MACRO(g_var2);
 
 #define BLA int FOO_bar
 BLA;
@@ -602,9 +604,20 @@ static void static_Function() {
 // CHECK-FIXES: {{^}}#define MY_TEST_MACRO(X) X()
 
 void MY_TEST_Macro(function) {}
-// CHECK-FIXES: {{^}}void MY_TEST_MACRO(function) {}
-}
-}
+// CHECK-MESSAGES: :[[@LINE-1]]:20: warning: invalid case style for global function 'function' [readability-identifier-naming]
+// CHECK-FIXES: {{^}}void MY_TEST_MACRO(Function) {}
+
+#define MY_CAT_IMPL(l, r) l ## r
+#define MY_CAT(l, r) MY_CAT_IMPL(l, r)
+#define MY_MACRO2(foo) int MY_CAT(awesome_, MY_CAT(foo, __COUNTER__)) = 0
+#define MY_MACRO3(foo) int MY_CAT(awesome_, foo) = 0
+MY_MACRO2(myglob);
+MY_MACRO3(myglob);
+// No suggestions should occur even though the resulting decl of awesome_myglob#
+// or awesome_myglob are not entirely within a macro argument.
+
+} // namespace InlineNamespace
+} // namespace FOO_NS
 
 template <typename t_t> struct a {
 // CHECK-MESSAGES: :[[@LINE-1]]:32: warning: invalid case style for struct 'a'

From f2ade91a9fe7c222ea919748d30b74397911ecc8 Mon Sep 17 00:00:00 2001
From: Jeff Niu <jeff@modular.com>
Date: Wed, 10 Apr 2024 14:11:45 +0200
Subject: [PATCH 003/886] [mlir] Optimize getting properties on concrete ops
 (#88259)

This makes retrieving properties on concrete operations faster by
removing a branch when it is known that the operation must have
properties.
---
 mlir/include/mlir/IR/OpDefinition.h | 2 +-
 mlir/include/mlir/IR/Operation.h    | 9 +++++++--
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/mlir/include/mlir/IR/OpDefinition.h b/mlir/include/mlir/IR/OpDefinition.h
index c177ae3594d11..2d1dee2303e8f 100644
--- a/mlir/include/mlir/IR/OpDefinition.h
+++ b/mlir/include/mlir/IR/OpDefinition.h
@@ -1965,7 +1965,7 @@ class Op : public OpState, public Traits<ConcreteType>... {
     if constexpr (!hasProperties())
       return getEmptyProperties();
     return *getOperation()
-                ->getPropertiesStorage()
+                ->getPropertiesStorageUnsafe()
                 .template as<InferredProperties<T> *>();
   }
 
diff --git a/mlir/include/mlir/IR/Operation.h b/mlir/include/mlir/IR/Operation.h
index 3ffd3517fe5a6..c52a6fcac10c1 100644
--- a/mlir/include/mlir/IR/Operation.h
+++ b/mlir/include/mlir/IR/Operation.h
@@ -895,8 +895,7 @@ class alignas(8) Operation final
   /// Returns the properties storage.
   OpaqueProperties getPropertiesStorage() {
     if (propertiesStorageSize)
-      return {
-          reinterpret_cast<void *>(getTrailingObjects<detail::OpProperties>())};
+      return getPropertiesStorageUnsafe();
     return {nullptr};
   }
   OpaqueProperties getPropertiesStorage() const {
@@ -905,6 +904,12 @@ class alignas(8) Operation final
           getTrailingObjects<detail::OpProperties>()))};
     return {nullptr};
   }
+  /// Returns the properties storage without checking whether properties are
+  /// present.
+  OpaqueProperties getPropertiesStorageUnsafe() {
+    return {
+        reinterpret_cast<void *>(getTrailingObjects<detail::OpProperties>())};
+  }
 
   /// Return the properties converted to an attribute.
   /// This is expensive, and mostly useful when dealing with unregistered

From 94ed57dab64ccb248a342a91957f390209c5c7ce Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Wed, 10 Apr 2024 13:30:29 +0100
Subject: [PATCH 004/886] [PhaseOrdering] Add test for #85551.

Add test for missed hoisting of checks from std::span
https://github.com/llvm/llvm-project/issues/85551
---
 .../AArch64/hoist-runtime-checks.ll           | 143 ++++++++++++++++++
 1 file changed, 143 insertions(+)

diff --git a/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll b/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
index c6c9a52167d54..a140e17a0dd15 100644
--- a/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
+++ b/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
@@ -91,8 +91,151 @@ for.end:                                          ; preds = %for.cond.cleanup
   ret i32 %9
 }
 
+%"class.std::__1::span" = type { ptr, i64 }
+%"class.std::__1::__wrap_iter" = type { ptr }
+
+define dso_local noundef i32 @sum_prefix_with_sum(ptr %s.coerce0, i64 %s.coerce1, i64 noundef %n) {
+; CHECK-LABEL: define dso_local noundef i32 @sum_prefix_with_sum(
+; CHECK-SAME: ptr nocapture readonly [[S_COERCE0:%.*]], i64 [[S_COERCE1:%.*]], i64 noundef [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP5_NOT:%.*]] = icmp eq i64 [[N]], 0
+; CHECK-NEXT:    br i1 [[CMP5_NOT]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY_PREHEADER:%.*]]
+; CHECK:       for.body.preheader:
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
+; CHECK-NEXT:    [[DOTNOT_NOT:%.*]] = icmp ult i64 [[TMP0]], [[S_COERCE1]]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    [[RET_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[SPAN_CHECKED_ACCESS_EXIT:%.*]] ]
+; CHECK-NEXT:    ret i32 [[RET_0_LCSSA]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I_07:%.*]] = phi i64 [ [[INC:%.*]], [[SPAN_CHECKED_ACCESS_EXIT]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; CHECK-NEXT:    [[RET_06:%.*]] = phi i32 [ [[ADD]], [[SPAN_CHECKED_ACCESS_EXIT]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
+; CHECK-NEXT:    br i1 [[DOTNOT_NOT]], label [[SPAN_CHECKED_ACCESS_EXIT]], label [[COND_FALSE_I:%.*]], !prof [[PROF0:![0-9]+]]
+; CHECK:       cond.false.i:
+; CHECK-NEXT:    tail call void @llvm.trap()
+; CHECK-NEXT:    unreachable
+; CHECK:       span_checked_access.exit:
+; CHECK-NEXT:    [[ARRAYIDX_I:%.*]] = getelementptr inbounds i32, ptr [[S_COERCE0]], i64 [[I_07]]
+; CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX_I]], align 4
+; CHECK-NEXT:    [[ADD]] = add nsw i32 [[TMP7]], [[RET_06]]
+; CHECK-NEXT:    [[INC]] = add nuw i64 [[I_07]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]]
+;
+entry:
+  %s = alloca %"class.std::__1::span", align 8
+  %n.addr = alloca i64, align 8
+  %ret = alloca i32, align 4
+  %i = alloca i64, align 8
+  %0 = getelementptr inbounds { ptr, i64 }, ptr %s, i32 0, i32 0
+  store ptr %s.coerce0, ptr %0, align 8
+  %1 = getelementptr inbounds { ptr, i64 }, ptr %s, i32 0, i32 1
+  store i64 %s.coerce1, ptr %1, align 8
+  store i64 %n, ptr %n.addr, align 8
+  call void @llvm.lifetime.start.p0(i64 4, ptr %ret) #7
+  store i32 0, ptr %ret, align 4
+  call void @llvm.lifetime.start.p0(i64 8, ptr %i) #7
+  store i64 0, ptr %i, align 8
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %2 = load i64, ptr %i, align 8
+  %3 = load i64, ptr %n.addr, align 8
+  %cmp = icmp ult i64 %2, %3
+  br i1 %cmp, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 8, ptr %i) #7
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %4 = load i64, ptr %i, align 8
+  %call = call noundef nonnull align 4 dereferenceable(4) ptr @span_checked_access(ptr noundef nonnull align 8 dereferenceable(16) %s, i64 noundef %4) #7
+  %5 = load i32, ptr %call, align 4
+  %6 = load i32, ptr %ret, align 4
+  %add = add nsw i32 %6, %5
+  store i32 %add, ptr %ret, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body
+  %7 = load i64, ptr %i, align 8
+  %inc = add i64 %7, 1
+  store i64 %inc, ptr %i, align 8
+  br label %for.cond
+
+for.end:                                          ; preds = %for.cond.cleanup
+  %8 = load i32, ptr %ret, align 4
+  call void @llvm.lifetime.end.p0(i64 4, ptr %ret)
+  ret i32 %8
+}
+
+define hidden noundef nonnull align 4 dereferenceable(4) ptr @span_checked_access(ptr noundef nonnull align 8 dereferenceable(16) %this, i64 noundef %__idx) {
+; CHECK-LABEL: define hidden noundef nonnull align 4 dereferenceable(4) ptr @span_checked_access(
+; CHECK-SAME: ptr nocapture noundef nonnull readonly align 8 dereferenceable(16) [[THIS:%.*]], i64 noundef [[__IDX:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[__SIZE__I:%.*]] = getelementptr inbounds i8, ptr [[THIS]], i64 8
+; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[__SIZE__I]], align 8
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP0]], [[__IDX]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[COND_END:%.*]], label [[COND_FALSE:%.*]], !prof [[PROF0]]
+; CHECK:       cond.false:
+; CHECK-NEXT:    tail call void @llvm.trap()
+; CHECK-NEXT:    unreachable
+; CHECK:       cond.end:
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[THIS]], align 8
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[__IDX]]
+; CHECK-NEXT:    ret ptr [[ARRAYIDX]]
+;
+entry:
+  %this.addr = alloca ptr, align 8
+  %__idx.addr = alloca i64, align 8
+  store ptr %this, ptr %this.addr, align 8
+  store i64 %__idx, ptr %__idx.addr, align 8
+  %this1 = load ptr, ptr %this.addr, align 8
+  %0 = load i64, ptr %__idx.addr, align 8
+  %call = call noundef i64 @span_access(ptr noundef nonnull align 8 dereferenceable(16) %this1)
+  %cmp = icmp ult i64 %0, %call
+  %conv = zext i1 %cmp to i64
+  %expval = call i64 @llvm.expect.i64(i64 %conv, i64 1)
+  %tobool = icmp ne i64 %expval, 0
+  br i1 %tobool, label %cond.true, label %cond.false
+
+cond.true:                                        ; preds = %entry
+  br label %cond.end
+
+cond.false:                                       ; preds = %entry
+  call void @llvm.trap()
+  br label %cond.end
+
+cond.end:                                         ; preds = %cond.false, %cond.true
+  %__data_ = getelementptr inbounds %"class.std::__1::span", ptr %this1, i32 0, i32 0
+  %1 = load ptr, ptr %__data_, align 8
+  %2 = load i64, ptr %__idx.addr, align 8
+  %arrayidx = getelementptr inbounds i32, ptr %1, i64 %2
+  ret ptr %arrayidx
+}
+
+define hidden noundef i64 @span_access(ptr noundef nonnull align 8 dereferenceable(16) %this) {
+; CHECK-LABEL: define hidden noundef i64 @span_access(
+; CHECK-SAME: ptr nocapture noundef nonnull readonly align 8 dereferenceable(16) [[THIS:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[__SIZE_:%.*]] = getelementptr inbounds i8, ptr [[THIS]], i64 8
+; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[__SIZE_]], align 8
+; CHECK-NEXT:    ret i64 [[TMP0]]
+;
+entry:
+  %this.addr = alloca ptr, align 8
+  store ptr %this, ptr %this.addr, align 8
+  %this1 = load ptr, ptr %this.addr, align 8
+  %__size_ = getelementptr inbounds %"class.std::__1::span", ptr %this1, i32 0, i32 1
+  %0 = load i64, ptr %__size_, align 8
+  ret i64 %0
+}
+
 declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture)
 
 declare void @llvm.trap()
 
 declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture)
+;.
+; CHECK: [[PROF0]] = !{!"branch_weights", i32 2000, i32 1}
+;.

From a2bdbc6f0da2a9d0cecb23c64bd20423b3fd0340 Mon Sep 17 00:00:00 2001
From: Jacek Caban <jacek@codeweavers.com>
Date: Wed, 10 Apr 2024 14:37:18 +0200
Subject: [PATCH 005/886] [LLD][COFF] Check machine types in
 ICF::equalsConstant. (#88140)

Avoid replacing replacing a chunk with one from a different type. It's
mostly a concern for ARM64X, where we don't want to merge aarch64 and
arm64ec chunks, but it may also in theory happen between arm64ec and
 x86_64 chunks.
---
 lld/COFF/ICF.cpp           |  2 +-
 lld/test/COFF/arm64x-icf.s | 37 +++++++++++++++++++++++++++++++++++++
 2 files changed, 38 insertions(+), 1 deletion(-)
 create mode 100644 lld/test/COFF/arm64x-icf.s

diff --git a/lld/COFF/ICF.cpp b/lld/COFF/ICF.cpp
index 013ffcfb3d5d1..b899a25324239 100644
--- a/lld/COFF/ICF.cpp
+++ b/lld/COFF/ICF.cpp
@@ -178,7 +178,7 @@ bool ICF::equalsConstant(const SectionChunk *a, const SectionChunk *b) {
          a->getSectionName() == b->getSectionName() &&
          a->header->SizeOfRawData == b->header->SizeOfRawData &&
          a->checksum == b->checksum && a->getContents() == b->getContents() &&
-         assocEquals(a, b);
+         a->getMachine() == b->getMachine() && assocEquals(a, b);
 }
 
 // Compare "moving" part of two sections, namely relocation targets.
diff --git a/lld/test/COFF/arm64x-icf.s b/lld/test/COFF/arm64x-icf.s
new file mode 100644
index 0000000000000..c8df21d3e4969
--- /dev/null
+++ b/lld/test/COFF/arm64x-icf.s
@@ -0,0 +1,37 @@
+// REQUIRES: aarch64
+// RUN: split-file %s %t.dir && cd %t.dir
+
+// RUN: llvm-mc -filetype=obj -triple=arm64ec-windows func-arm64ec.s -o func-arm64ec.obj
+// RUN: llvm-mc -filetype=obj -triple=aarch64-windows func-arm64.s -o func-arm64.obj
+// RUN: lld-link -machine:arm64x -dll -noentry -out:out.dll func-arm64ec.obj func-arm64.obj
+// RUN: llvm-objdump -d out.dll | FileCheck %s
+
+// CHECK:      0000000180001000 <.text>:
+// CHECK-NEXT: 180001000: 52800020     mov     w0, #0x1                // =1
+// CHECK-NEXT: 180001004: d65f03c0     ret
+// CHECK-NEXT:                 ...
+// CHECK-NEXT: 180002000: 52800020     mov     w0, #0x1                // =1
+// CHECK-NEXT: 180002004: d65f03c0     ret
+
+
+#--- func-arm64.s
+        .section .text,"xr",discard,func
+        .globl func
+        .p2align 2
+func:
+        mov w0, #1
+        ret
+
+        .data
+        .rva func
+
+#--- func-arm64ec.s
+        .section .text,"xr",discard,"#func"
+        .globl "#func"
+        .p2align 2
+"#func":
+        mov w0, #1
+        ret
+
+        .data
+        .rva "#func"

From b47e439559ad03a1b32614f573aad66f145a634d Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Wed, 10 Apr 2024 08:38:49 -0400
Subject: [PATCH 006/886] Revert "[Clang][Sema] Fix crash when 'this' is used
 in a dependent class scope function template specialization that instantiates
 to a static member function" (#88264)

Reverts llvm/llvm-project#87541
---
 clang/docs/ReleaseNotes.rst                   |  2 -
 clang/include/clang/Sema/Sema.h               |  8 +---
 clang/lib/Sema/SemaExpr.cpp                   |  5 +--
 clang/lib/Sema/SemaExprCXX.cpp                | 44 ++++++-------------
 clang/lib/Sema/SemaExprMember.cpp             | 42 +++++-------------
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |  8 ----
 clang/lib/Sema/TreeTransform.h                |  7 ++-
 ...ms-function-specialization-class-scope.cpp | 44 ++-----------------
 8 files changed, 36 insertions(+), 124 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 6bff80ed4d210..f5359afe1f099 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -520,8 +520,6 @@ Bug Fixes to C++ Support
 - Fix an issue caused by not handling invalid cases when substituting into the parameter mapping of a constraint. Fixes (#GH86757).
 - Fixed a bug that prevented member function templates of class templates declared with a deduced return type
   from being explicitly specialized for a given implicit instantiation of the class template.
-- Fixed a crash when ``this`` is used in a dependent class scope function template specialization
-  that instantiates to a static member function.
 
 - Fix crash when inheriting from a cv-qualified type. Fixes:
   (`#35603 <https://github.com/llvm/llvm-project/issues/35603>`_)
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index f311f9f374345..9769d36900664 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -5439,8 +5439,7 @@ class Sema final : public SemaBase {
 
   ExprResult BuildDeclarationNameExpr(const CXXScopeSpec &SS, LookupResult &R,
                                       bool NeedsADL,
-                                      bool AcceptInvalidDecl = false,
-                                      bool NeedUnresolved = false);
+                                      bool AcceptInvalidDecl = false);
   ExprResult BuildDeclarationNameExpr(
       const CXXScopeSpec &SS, const DeclarationNameInfo &NameInfo, NamedDecl *D,
       NamedDecl *FoundD = nullptr,
@@ -6592,10 +6591,7 @@ class Sema final : public SemaBase {
                             SourceLocation RParenLoc);
 
   //// ActOnCXXThis -  Parse 'this' pointer.
-  ExprResult ActOnCXXThis(SourceLocation Loc);
-
-  /// Check whether the type of 'this' is valid in the current context.
-  bool CheckCXXThisType(SourceLocation Loc, QualType Type);
+  ExprResult ActOnCXXThis(SourceLocation loc);
 
   /// Build a CXXThisExpr and mark it referenced in the current context.
   Expr *BuildCXXThisExpr(SourceLocation Loc, QualType Type, bool IsImplicit);
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 45acbf197ea6b..594c11788f4e7 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -3442,11 +3442,10 @@ static bool ShouldLookupResultBeMultiVersionOverload(const LookupResult &R) {
 
 ExprResult Sema::BuildDeclarationNameExpr(const CXXScopeSpec &SS,
                                           LookupResult &R, bool NeedsADL,
-                                          bool AcceptInvalidDecl,
-                                          bool NeedUnresolved) {
+                                          bool AcceptInvalidDecl) {
   // If this is a single, fully-resolved result and we don't need ADL,
   // just build an ordinary singleton decl ref.
-  if (!NeedUnresolved && !NeedsADL && R.isSingleResult() &&
+  if (!NeedsADL && R.isSingleResult() &&
       !R.getAsSingle<FunctionTemplateDecl>() &&
       !ShouldLookupResultBeMultiVersionOverload(R))
     return BuildDeclarationNameExpr(SS, R.getLookupNameInfo(), R.getFoundDecl(),
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 9822477260e59..7b9b8f149d9ed 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -1415,42 +1415,26 @@ bool Sema::CheckCXXThisCapture(SourceLocation Loc, const bool Explicit,
 }
 
 ExprResult Sema::ActOnCXXThis(SourceLocation Loc) {
-  // C++20 [expr.prim.this]p1:
-  //   The keyword this names a pointer to the object for which an
-  //   implicit object member function is invoked or a non-static
-  //   data member's initializer is evaluated.
+  /// C++ 9.3.2: In the body of a non-static member function, the keyword this
+  /// is a non-lvalue expression whose value is the address of the object for
+  /// which the function is called.
   QualType ThisTy = getCurrentThisType();
 
-  if (CheckCXXThisType(Loc, ThisTy))
-    return ExprError();
+  if (ThisTy.isNull()) {
+    DeclContext *DC = getFunctionLevelDeclContext();
 
-  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
-}
+    if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
+        Method && Method->isExplicitObjectMemberFunction()) {
+      return Diag(Loc, diag::err_invalid_this_use) << 1;
+    }
 
-bool Sema::CheckCXXThisType(SourceLocation Loc, QualType Type) {
-  if (!Type.isNull())
-    return false;
+    if (isLambdaCallWithExplicitObjectParameter(CurContext))
+      return Diag(Loc, diag::err_invalid_this_use) << 1;
 
-  // C++20 [expr.prim.this]p3:
-  //   If a declaration declares a member function or member function template
-  //   of a class X, the expression this is a prvalue of type
-  //   "pointer to cv-qualifier-seq X" wherever X is the current class between
-  //   the optional cv-qualifier-seq and the end of the function-definition,
-  //   member-declarator, or declarator. It shall not appear within the
-  //   declaration of either a static member function or an explicit object
-  //   member function of the current class (although its type and value
-  //   category are defined within such member functions as they are within
-  //   an implicit object member function).
-  DeclContext *DC = getFunctionLevelDeclContext();
-  if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
-      Method && Method->isExplicitObjectMemberFunction()) {
-    Diag(Loc, diag::err_invalid_this_use) << 1;
-  } else if (isLambdaCallWithExplicitObjectParameter(CurContext)) {
-    Diag(Loc, diag::err_invalid_this_use) << 1;
-  } else {
-    Diag(Loc, diag::err_invalid_this_use) << 0;
+    return Diag(Loc, diag::err_invalid_this_use) << 0;
   }
-  return true;
+
+  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
 }
 
 Expr *Sema::BuildCXXThisExpr(SourceLocation Loc, QualType Type,
diff --git a/clang/lib/Sema/SemaExprMember.cpp b/clang/lib/Sema/SemaExprMember.cpp
index 8cd2288d279cc..32998ae60eafe 100644
--- a/clang/lib/Sema/SemaExprMember.cpp
+++ b/clang/lib/Sema/SemaExprMember.cpp
@@ -61,10 +61,6 @@ enum IMAKind {
   /// The reference is a contextually-permitted abstract member reference.
   IMA_Abstract,
 
-  /// Whether the context is static is dependent on the enclosing template (i.e.
-  /// in a dependent class scope explicit specialization).
-  IMA_Dependent,
-
   /// The reference may be to an unresolved using declaration and the
   /// context is not an instance method.
   IMA_Unresolved_StaticOrExplicitContext,
@@ -95,18 +91,10 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
 
   DeclContext *DC = SemaRef.getFunctionLevelDeclContext();
 
-  bool couldInstantiateToStatic = false;
-  bool isStaticOrExplicitContext = SemaRef.CXXThisTypeOverride.isNull();
-
-  if (auto *MD = dyn_cast<CXXMethodDecl>(DC)) {
-    if (MD->isImplicitObjectMemberFunction()) {
-      isStaticOrExplicitContext = false;
-      // A dependent class scope function template explicit specialization
-      // that is neither declared 'static' nor with an explicit object
-      // parameter could instantiate to a static or non-static member function.
-      couldInstantiateToStatic = MD->getDependentSpecializationInfo();
-    }
-  }
+  bool isStaticOrExplicitContext =
+      SemaRef.CXXThisTypeOverride.isNull() &&
+      (!isa<CXXMethodDecl>(DC) || cast<CXXMethodDecl>(DC)->isStatic() ||
+       cast<CXXMethodDecl>(DC)->isExplicitObjectMemberFunction());
 
   if (R.isUnresolvableResult())
     return isStaticOrExplicitContext ? IMA_Unresolved_StaticOrExplicitContext
@@ -135,9 +123,6 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
   if (Classes.empty())
     return IMA_Static;
 
-  if (couldInstantiateToStatic)
-    return IMA_Dependent;
-
   // C++11 [expr.prim.general]p12:
   //   An id-expression that denotes a non-static data member or non-static
   //   member function of a class can only be used:
@@ -283,30 +268,27 @@ ExprResult Sema::BuildPossibleImplicitMemberExpr(
     const CXXScopeSpec &SS, SourceLocation TemplateKWLoc, LookupResult &R,
     const TemplateArgumentListInfo *TemplateArgs, const Scope *S,
     UnresolvedLookupExpr *AsULE) {
-  switch (IMAKind Classification = ClassifyImplicitMemberAccess(*this, R)) {
+  switch (ClassifyImplicitMemberAccess(*this, R)) {
   case IMA_Instance:
+    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, true, S);
+
   case IMA_Mixed:
   case IMA_Mixed_Unrelated:
   case IMA_Unresolved:
-    return BuildImplicitMemberExpr(
-        SS, TemplateKWLoc, R, TemplateArgs,
-        /*IsKnownInstance=*/Classification == IMA_Instance, S);
+    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, false,
+                                   S);
+
   case IMA_Field_Uneval_Context:
     Diag(R.getNameLoc(), diag::warn_cxx98_compat_non_static_member_use)
       << R.getLookupNameInfo().getName();
     [[fallthrough]];
   case IMA_Static:
   case IMA_Abstract:
-  case IMA_Dependent:
   case IMA_Mixed_StaticOrExplicitContext:
   case IMA_Unresolved_StaticOrExplicitContext:
     if (TemplateArgs || TemplateKWLoc.isValid())
-      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, /*RequiresADL=*/false,
-                                 TemplateArgs);
-    return AsULE ? AsULE
-                 : BuildDeclarationNameExpr(
-                       SS, R, /*NeedsADL=*/false, /*AcceptInvalidDecl=*/false,
-                       /*NeedUnresolved=*/Classification == IMA_Dependent);
+      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, false, TemplateArgs);
+    return AsULE ? AsULE : BuildDeclarationNameExpr(SS, R, false);
 
   case IMA_Error_StaticOrExplicitContext:
   case IMA_Error_Unrelated:
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 8248b10814fea..127a432367b95 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5093,14 +5093,6 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
   EnterExpressionEvaluationContext EvalContext(
       *this, Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 
-  Qualifiers ThisTypeQuals;
-  CXXRecordDecl *ThisContext = nullptr;
-  if (CXXMethodDecl *Method = dyn_cast<CXXMethodDecl>(Function)) {
-    ThisContext = Method->getParent();
-    ThisTypeQuals = Method->getMethodQualifiers();
-  }
-  CXXThisScopeRAII ThisScope(*this, ThisContext, ThisTypeQuals);
-
   // Introduce a new scope where local variable instantiations will be
   // recorded, unless we're actually a member function within a local
   // class, in which case we need to merge our results with the parent
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 13d7b00430d52..d4d2fa61d65ea 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -3307,13 +3307,12 @@ class TreeTransform {
 
   /// Build a new C++ "this" expression.
   ///
-  /// By default, performs semantic analysis to build a new "this" expression.
-  /// Subclasses may override this routine to provide different behavior.
+  /// By default, builds a new "this" expression without performing any
+  /// semantic analysis. Subclasses may override this routine to provide
+  /// different behavior.
   ExprResult RebuildCXXThisExpr(SourceLocation ThisLoc,
                                 QualType ThisType,
                                 bool isImplicit) {
-    if (getSema().CheckCXXThisType(ThisLoc, ThisType))
-      return ExprError();
     return getSema().BuildCXXThisExpr(ThisLoc, ThisType, isImplicit);
   }
 
diff --git a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
index 6977623a0816e..dcab9bfaeabcb 100644
--- a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
+++ b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
@@ -1,6 +1,7 @@
-// RUN: %clang_cc1 -fms-extensions -fsyntax-only -Wno-unused-value -verify %s
-// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -Wno-unused-value -verify %s
+// RUN: %clang_cc1 -fms-extensions -fsyntax-only -verify %s
+// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -verify %s
 
+// expected-no-diagnostics
 class A {
 public:
   template<class U> A(U p) {}
@@ -75,42 +76,3 @@ struct S {
   int f<0>(int);
 };
 }
-
-namespace UsesThis {
-  template<typename T>
-  struct A {
-    int x;
-
-    template<typename U>
-    static void f();
-
-    template<>
-    void f<int>() {
-      this->x; // expected-error {{invalid use of 'this' outside of a non-static member function}}
-      x; // expected-error {{invalid use of member 'x' in static member function}}
-      A::x; // expected-error {{invalid use of member 'x' in static member function}}
-      +x; // expected-error {{invalid use of member 'x' in static member function}}
-      +A::x; // expected-error {{invalid use of member 'x' in static member function}}
-    }
-
-    template<typename U>
-    void g();
-
-    template<>
-    void g<int>() {
-      this->x;
-      x;
-      A::x;
-      +x;
-      +A::x;
-    }
-
-    template<typename U>
-    static auto h() -> A*;
-
-    template<>
-    auto h<int>() -> decltype(this); // expected-error {{'this' cannot be used in a static member function declaration}}
-  };
-
-  template struct A<int>; // expected-note 2{{in instantiation of}}
-}

From 1ca01958310f2956abd72ece1652c3218bcf27e1 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Wed, 10 Apr 2024 08:41:22 -0400
Subject: [PATCH 007/886] [Clang][AST][NFC] Fix printing of dependent
 PackIndexTypes (#88146)

Dependent `PackIndexType`s currently print the memory address of the
index `Expr*` rather than pretty printing the expression. This patch
fixes that.
---
 clang/lib/AST/TypePrinter.cpp | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/clang/lib/AST/TypePrinter.cpp b/clang/lib/AST/TypePrinter.cpp
index 075c8aba11fcb..9602f448e9427 100644
--- a/clang/lib/AST/TypePrinter.cpp
+++ b/clang/lib/AST/TypePrinter.cpp
@@ -1213,10 +1213,13 @@ void TypePrinter::printDecltypeBefore(const DecltypeType *T, raw_ostream &OS) {
 
 void TypePrinter::printPackIndexingBefore(const PackIndexingType *T,
                                           raw_ostream &OS) {
-  if (T->hasSelectedType())
+  if (T->hasSelectedType()) {
     OS << T->getSelectedType();
-  else
-    OS << T->getPattern() << "...[" << T->getIndexExpr() << "]";
+  } else {
+    OS << T->getPattern() << "...[";
+    T->getIndexExpr()->printPretty(OS, nullptr, Policy);
+    OS << "]";
+  }
   spaceBeforePlaceHolder(OS);
 }
 

From 49ef12a08c4c7d7ae4765929e72fe2320a12b08c Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Wed, 10 Apr 2024 14:55:56 +0200
Subject: [PATCH 008/886] Fix complex log1p accuracy with large abs values.
 (#88260)

This ports https://github.com/openxla/xla/pull/10503 by @pearu. The new
implementation matches mpmath's results for most inputs, see caveats in
the linked pull request. In addition to the filecheck test here, the
accuracy was tested with XLA's complex_unary_op_test and its MLIR
emitters.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 50 ++++++++++---------
 .../convert-to-standard.mlir                  | 48 +++++++++++-------
 2 files changed, 57 insertions(+), 41 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 9c3c4d96a301e..0aa1de5fa5d9a 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -570,37 +570,39 @@ struct Log1pOpConversion : public OpConversionPattern<complex::Log1pOp> {
                   ConversionPatternRewriter &rewriter) const override {
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
-    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
     mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
-    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
-    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
+    Value real = b.create<complex::ReOp>(adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(adaptor.getComplex());
 
     Value half = b.create<arith::ConstantOp>(elementType,
                                              b.getFloatAttr(elementType, 0.5));
     Value one = b.create<arith::ConstantOp>(elementType,
                                             b.getFloatAttr(elementType, 1));
-    Value two = b.create<arith::ConstantOp>(elementType,
-                                            b.getFloatAttr(elementType, 2));
-
-    // log1p(a+bi) = .5*log((a+1)^2+b^2) + i*atan2(b, a + 1)
-    // log((a+1)+bi) = .5*log(a*a + 2*a + 1 + b*b) + i*atan2(b, a+1)
-    // log((a+1)+bi) = .5*log1p(a*a + 2*a + b*b) + i*atan2(b, a+1)
-    Value sumSq = b.create<arith::MulFOp>(real, real, fmf.getValue());
-    sumSq = b.create<arith::AddFOp>(
-        sumSq, b.create<arith::MulFOp>(real, two, fmf.getValue()),
-        fmf.getValue());
-    sumSq = b.create<arith::AddFOp>(
-        sumSq, b.create<arith::MulFOp>(imag, imag, fmf.getValue()),
-        fmf.getValue());
-    Value logSumSq =
-        b.create<math::Log1pOp>(elementType, sumSq, fmf.getValue());
-    Value resultReal = b.create<arith::MulFOp>(logSumSq, half, fmf.getValue());
-
-    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf.getValue());
-
-    Value resultImag =
-        b.create<math::Atan2Op>(elementType, imag, realPlusOne, fmf.getValue());
+    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf);
+    Value absRealPlusOne = b.create<math::AbsFOp>(realPlusOne, fmf);
+    Value absImag = b.create<math::AbsFOp>(imag, fmf);
+
+    Value maxAbs = b.create<arith::MaximumFOp>(absRealPlusOne, absImag, fmf);
+    Value minAbs = b.create<arith::MinimumFOp>(absRealPlusOne, absImag, fmf);
+
+    Value maxAbsOfRealPlusOneAndImagMinusOne = b.create<arith::SelectOp>(
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, realPlusOne, absImag,
+                                fmf),
+        real, b.create<arith::SubFOp>(maxAbs, one, fmf));
+    Value minMaxRatio = b.create<arith::DivFOp>(minAbs, maxAbs, fmf);
+    Value logOfMaxAbsOfRealPlusOneAndImag =
+        b.create<math::Log1pOp>(maxAbsOfRealPlusOneAndImagMinusOne, fmf);
+    Value logOfSqrtPart = b.create<math::Log1pOp>(
+        b.create<arith::MulFOp>(minMaxRatio, minMaxRatio, fmf), fmf);
+    Value r = b.create<arith::AddFOp>(
+        b.create<arith::MulFOp>(half, logOfSqrtPart, fmf),
+        logOfMaxAbsOfRealPlusOneAndImag, fmf);
+    Value resultReal = b.create<arith::SelectOp>(
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, r, r, fmf), minAbs,
+        r);
+    Value resultImag = b.create<math::Atan2Op>(imag, realPlusOne, fmf);
     rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
                                                    resultImag);
     return success();
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index f5d9499eadda4..43918904a09f4 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -300,15 +300,22 @@ func.func @complex_log1p(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] : f32
-// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] : f32
-// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] : f32
-// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] : f32
-// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] : f32
-// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] : f32
 // CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] : f32
+// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %cst_0 : f32
+// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
+// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] : f32
+// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] : f32
+// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] : f32
+// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] : f32
+// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
@@ -963,15 +970,22 @@ func.func @complex_log1p_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] fastmath<nnan,contract> : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %cst_0 fastmath<nnan,contract>  : f32
+// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
+// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract>  : f32
+// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] fastmath<nnan,contract>  : f32
+// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] fastmath<nnan,contract>  : f32
+// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] fastmath<nnan,contract>  : f32
+// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] fastmath<nnan,contract> : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>

From 54a9f0007cb4f19d2e9df30405c5027229f5def0 Mon Sep 17 00:00:00 2001
From: annamthomas <anna@azul.com>
Date: Wed, 10 Apr 2024 09:02:23 -0400
Subject: [PATCH 009/886] [SCEV] Fix BinomialCoefficient Iteration to fit in W
 bits (#88010)

BinomialCoefficient computes the value of W-bit IV at iteration It of a loop. When W is 1, we can call multiplicative inverse on 0 which triggers an assert since 1b76120.

Since the arithmetic is supposed to wrap if It or K does not fit in W bits, do the truncation into W bits after we do the shift.

 Fixes #87798
---
 llvm/lib/Analysis/ScalarEvolution.cpp         |  6 +-
 llvm/test/Analysis/ScalarEvolution/pr87798.ll | 68 +++++++++++++++++++
 2 files changed, 70 insertions(+), 4 deletions(-)
 create mode 100644 llvm/test/Analysis/ScalarEvolution/pr87798.ll

diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index e030b9fc7dac4..9fcce797f5597 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -928,11 +928,9 @@ static const SCEV *BinomialCoefficient(const SCEV *It, unsigned K,
   APInt OddFactorial(W, 1);
   unsigned T = 1;
   for (unsigned i = 3; i <= K; ++i) {
-    APInt Mult(W, i);
-    unsigned TwoFactors = Mult.countr_zero();
+    unsigned TwoFactors = countr_zero(i);
     T += TwoFactors;
-    Mult.lshrInPlace(TwoFactors);
-    OddFactorial *= Mult;
+    OddFactorial *= (i >> TwoFactors);
   }
 
   // We need at least W + T bits for the multiplication step
diff --git a/llvm/test/Analysis/ScalarEvolution/pr87798.ll b/llvm/test/Analysis/ScalarEvolution/pr87798.ll
new file mode 100644
index 0000000000000..acd445993e47b
--- /dev/null
+++ b/llvm/test/Analysis/ScalarEvolution/pr87798.ll
@@ -0,0 +1,68 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -disable-output -passes='print<scalar-evolution>' -verify-scev < %s 2>&1 | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2"
+target triple = "x86_64-unknown-linux-gnu"
+
+; print<scalar-evolution> is used to compute SCEVs for all values in the
+; function.
+; We should not crash on multiplicative inverse called within SCEV's binomial
+; coefficient function.
+
+define i32 @pr87798() {
+; CHECK-LABEL: 'pr87798'
+; CHECK-NEXT:  Classifying expressions for: @pr87798
+; CHECK-NEXT:    %phi = phi i32 [ 0, %bb ], [ %add4, %bb1 ]
+; CHECK-NEXT:    --> {0,+,0,+,0,+,2,+,3}<%bb1> U: full-set S: full-set Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %phi2 = phi i32 [ 0, %bb ], [ %add, %bb1 ]
+; CHECK-NEXT:    --> {0,+,0,+,1}<%bb1> U: full-set S: full-set Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %phi3 = phi i32 [ 0, %bb ], [ %add5, %bb1 ]
+; CHECK-NEXT:    --> {0,+,1}<nuw><nsw><%bb1> U: [0,1) S: [0,1) Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %add = add i32 %phi2, %phi3
+; CHECK-NEXT:    --> {0,+,1,+,1}<%bb1> U: full-set S: full-set Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %mul = mul i32 %phi2, %phi3
+; CHECK-NEXT:    --> {0,+,0,+,2,+,3}<%bb1> U: full-set S: full-set Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %add4 = add i32 %mul, %phi
+; CHECK-NEXT:    --> {0,+,0,+,2,+,5,+,3}<%bb1> U: full-set S: full-set Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %and = and i32 %phi, 1
+; CHECK-NEXT:    --> (zext i1 {false,+,false,+,false,+,false,+,true}<%bb1> to i32) U: [0,2) S: [0,2) Exits: 0 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %add5 = add i32 %phi3, 1
+; CHECK-NEXT:    --> {1,+,1}<nuw><nsw><%bb1> U: [1,2) S: [1,2) Exits: 1 LoopDispositions: { %bb1: Computable }
+; CHECK-NEXT:    %phi9 = phi i32 [ %and, %bb1 ]
+; CHECK-NEXT:    --> (zext i1 {false,+,false,+,false,+,false,+,true}<%bb1> to i32) U: [0,2) S: [0,2) --> 0 U: [0,1) S: [0,1)
+; CHECK-NEXT:    %zext = zext i32 %phi9 to i64
+; CHECK-NEXT:    --> poison U: full-set S: full-set
+; CHECK-NEXT:  Determining loop execution counts for: @pr87798
+; CHECK-NEXT:  Loop %loop: <multiple exits> Unpredictable backedge-taken count.
+; CHECK-NEXT:  Loop %loop: Unpredictable constant max backedge-taken count.
+; CHECK-NEXT:  Loop %loop: Unpredictable symbolic max backedge-taken count.
+; CHECK-NEXT:  Loop %bb1: backedge-taken count is i1 false
+; CHECK-NEXT:  Loop %bb1: constant max backedge-taken count is i1 false
+; CHECK-NEXT:  Loop %bb1: symbolic max backedge-taken count is i1 false
+; CHECK-NEXT:  Loop %bb1: Trip multiple is 1
+;
+bb:
+  br label %bb1
+
+bb1:                                              ; preds = %bb1, %bb
+  %phi = phi i32 [ 0, %bb ], [ %add4, %bb1 ]
+  %phi2 = phi i32 [ 0, %bb ], [ %add, %bb1 ]
+  %phi3 = phi i32 [ 0, %bb ], [ %add5, %bb1 ]
+  %add = add i32 %phi2, %phi3
+  %mul = mul i32 %phi2, %phi3
+  %add4 = add i32 %mul, %phi
+  %and = and i32 %phi, 1
+  %add5 = add i32 %phi3, 1
+  br i1 true, label %preheader, label %bb1
+
+preheader:                                              ; preds = %bb1
+  %phi9 = phi i32 [ %and, %bb1 ]
+  br label %loop
+
+loop:                                              ; preds = %preheader, %loop
+  br label %loop
+
+bb7:                                              ; No predecessors!
+  %zext = zext i32 %phi9 to i64
+  ret i32 0
+}

From 938a73422e0b964eba16f272acdfae1d0281772c Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 4 Apr 2024 13:30:57 -0700
Subject: [PATCH 010/886] [SLP][NFC]Walk over entries, not single values.

Better to walk over SLP nodes rather than single values. Matching
a value to a node is not a 1-to-1 relation, one value may be part of
several nodes and compiler may get wrong node, when trying to map it.
Currently there are no such issues detected, but they may appear in
future.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 339 +++++++++---------
 1 file changed, 167 insertions(+), 172 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index c3dcf73b0b762..22ef9b5fb994e 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -2325,19 +2325,17 @@ class BoUpSLP {
   ~BoUpSLP();
 
 private:
-  /// Determine if a vectorized value \p V in can be demoted to
-  /// a smaller type with a truncation. We collect the values that will be
-  /// demoted in ToDemote and additional roots that require investigating in
-  /// Roots.
-  /// \param DemotedConsts list of Instruction/OperandIndex pairs that are
-  /// constant and to be demoted. Required to correctly identify constant nodes
-  /// to be demoted.
-  bool collectValuesToDemote(
-      Value *V, bool IsProfitableToDemoteRoot, unsigned &BitWidth,
-      SmallVectorImpl<Value *> &ToDemote,
-      DenseMap<Instruction *, SmallVector<unsigned>> &DemotedConsts,
-      DenseSet<Value *> &Visited, unsigned &MaxDepthLevel,
-      bool &IsProfitableToDemote, bool IsTruncRoot) const;
+  /// Determine if a node \p E in can be demoted to a smaller type with a
+  /// truncation. We collect the entries that will be demoted in ToDemote.
+  /// \param E Node for analysis
+  /// \param ToDemote indices of the nodes to be demoted.
+  bool collectValuesToDemote(const TreeEntry &E, bool IsProfitableToDemoteRoot,
+                             unsigned &BitWidth,
+                             SmallVectorImpl<unsigned> &ToDemote,
+                             DenseSet<const TreeEntry *> &Visited,
+                             unsigned &MaxDepthLevel,
+                             bool &IsProfitableToDemote,
+                             bool IsTruncRoot) const;
 
   /// Check if the operands on the edges \p Edges of the \p UserTE allows
   /// reordering (i.e. the operands can be reordered because they have only one
@@ -14126,20 +14124,17 @@ unsigned BoUpSLP::getVectorElementSize(Value *V) {
   return Width;
 }
 
-// Determine if a value V in a vectorizable expression Expr can be demoted to a
-// smaller type with a truncation. We collect the values that will be demoted
-// in ToDemote and additional roots that require investigating in Roots.
 bool BoUpSLP::collectValuesToDemote(
-    Value *V, bool IsProfitableToDemoteRoot, unsigned &BitWidth,
-    SmallVectorImpl<Value *> &ToDemote,
-    DenseMap<Instruction *, SmallVector<unsigned>> &DemotedConsts,
-    DenseSet<Value *> &Visited, unsigned &MaxDepthLevel,
-    bool &IsProfitableToDemote, bool IsTruncRoot) const {
+    const TreeEntry &E, bool IsProfitableToDemoteRoot, unsigned &BitWidth,
+    SmallVectorImpl<unsigned> &ToDemote, DenseSet<const TreeEntry *> &Visited,
+    unsigned &MaxDepthLevel, bool &IsProfitableToDemote,
+    bool IsTruncRoot) const {
   // We can always demote constants.
-  if (isa<Constant>(V))
+  if (all_of(E.Scalars, IsaPred<Constant>))
     return true;
 
-  if (DL->getTypeSizeInBits(V->getType()) == BitWidth) {
+  unsigned OrigBitWidth = DL->getTypeSizeInBits(E.Scalars.front()->getType());
+  if (OrigBitWidth == BitWidth) {
     MaxDepthLevel = 1;
     return true;
   }
@@ -14150,7 +14145,6 @@ bool BoUpSLP::collectValuesToDemote(
   auto IsPotentiallyTruncated = [&](Value *V, unsigned &BitWidth) -> bool {
     if (MultiNodeScalars.contains(V))
       return false;
-    uint32_t OrigBitWidth = DL->getTypeSizeInBits(V->getType());
     if (OrigBitWidth > BitWidth) {
       APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
       if (MaskedValueIsZero(V, Mask, SimplifyQuery(*DL)))
@@ -14168,47 +14162,50 @@ bool BoUpSLP::collectValuesToDemote(
     BitWidth = std::max(BitWidth, BitWidth1);
     return BitWidth > 0 && OrigBitWidth >= (BitWidth * 2);
   };
-  auto FinalAnalysis = [&](const TreeEntry *ITE = nullptr) {
+  using namespace std::placeholders;
+  auto FinalAnalysis = [&]() {
     if (!IsProfitableToDemote)
       return false;
-    return (ITE && ITE->UserTreeIndices.size() > 1) ||
-           IsPotentiallyTruncated(V, BitWidth);
+    bool Res = all_of(
+        E.Scalars, std::bind(IsPotentiallyTruncated, _1, std::ref(BitWidth)));
+    // Gather demoted constant operands.
+    if (Res && E.State == TreeEntry::NeedToGather &&
+        all_of(E.Scalars, IsaPred<Constant>))
+      ToDemote.push_back(E.Idx);
+    return Res;
   };
   // TODO: improve handling of gathered values and others.
-  auto *I = dyn_cast<Instruction>(V);
-  const TreeEntry *ITE = I ? getTreeEntry(I) : nullptr;
-  if (!ITE || !Visited.insert(I).second || MultiNodeScalars.contains(I) ||
-      all_of(I->users(), [&](User *U) {
-        return isa<InsertElementInst>(U) && !getTreeEntry(U);
+  if (E.State == TreeEntry::NeedToGather || !Visited.insert(&E).second ||
+      any_of(E.Scalars, [&](Value *V) {
+        return all_of(V->users(), [&](User *U) {
+          return isa<InsertElementInst>(U) && !getTreeEntry(U);
+        });
       }))
     return FinalAnalysis();
 
-  if (!all_of(I->users(),
-              [=](User *U) {
-                return getTreeEntry(U) ||
-                       (UserIgnoreList && UserIgnoreList->contains(U)) ||
-                       (U->getType()->isSized() &&
-                        !U->getType()->isScalableTy() &&
-                        DL->getTypeSizeInBits(U->getType()) <= BitWidth);
-              }) &&
-      !IsPotentiallyTruncated(I, BitWidth))
+  if (any_of(E.Scalars, [&](Value *V) {
+        return !all_of(V->users(), [=](User *U) {
+          return getTreeEntry(U) ||
+                 (UserIgnoreList && UserIgnoreList->contains(U)) ||
+                 (U->getType()->isSized() && !U->getType()->isScalableTy() &&
+                  DL->getTypeSizeInBits(U->getType()) <= BitWidth);
+        }) && !IsPotentiallyTruncated(V, BitWidth);
+      }))
     return false;
 
-  unsigned Start = 0;
-  unsigned End = I->getNumOperands();
-
-  auto ProcessOperands = [&](ArrayRef<Value *> Operands, bool &NeedToExit) {
+  auto ProcessOperands = [&](ArrayRef<const TreeEntry *> Operands,
+                             bool &NeedToExit) {
     NeedToExit = false;
     unsigned InitLevel = MaxDepthLevel;
-    for (Value *IncValue : Operands) {
+    for (const TreeEntry *Op : Operands) {
       unsigned Level = InitLevel;
-      if (!collectValuesToDemote(IncValue, IsProfitableToDemoteRoot, BitWidth,
-                                 ToDemote, DemotedConsts, Visited, Level,
-                                 IsProfitableToDemote, IsTruncRoot)) {
+      if (!collectValuesToDemote(*Op, IsProfitableToDemoteRoot, BitWidth,
+                                 ToDemote, Visited, Level, IsProfitableToDemote,
+                                 IsTruncRoot)) {
         if (!IsProfitableToDemote)
           return false;
         NeedToExit = true;
-        if (!FinalAnalysis(ITE))
+        if (!FinalAnalysis())
           return false;
         continue;
       }
@@ -14220,7 +14217,6 @@ bool BoUpSLP::collectValuesToDemote(
       [&](function_ref<bool(unsigned, unsigned)> Checker, bool &NeedToExit) {
         // Try all bitwidth < OrigBitWidth.
         NeedToExit = false;
-        uint32_t OrigBitWidth = DL->getTypeSizeInBits(I->getType());
         unsigned BestFailBitwidth = 0;
         for (; BitWidth < OrigBitWidth; BitWidth *= 2) {
           if (Checker(BitWidth, OrigBitWidth))
@@ -14241,18 +14237,20 @@ bool BoUpSLP::collectValuesToDemote(
         return false;
       };
   auto TryProcessInstruction =
-      [&](Instruction *I, const TreeEntry &ITE, unsigned &BitWidth,
-          ArrayRef<Value *> Operands = std::nullopt,
+      [&](unsigned &BitWidth,
+          ArrayRef<const TreeEntry *> Operands = std::nullopt,
           function_ref<bool(unsigned, unsigned)> Checker = {}) {
         if (Operands.empty()) {
           if (!IsTruncRoot)
             MaxDepthLevel = 1;
-          (void)IsPotentiallyTruncated(V, BitWidth);
+          (void)for_each(E.Scalars, std::bind(IsPotentiallyTruncated, _1,
+                                              std::ref(BitWidth)));
         } else {
           // Several vectorized uses? Check if we can truncate it, otherwise -
           // exit.
-          if (ITE.UserTreeIndices.size() > 1 &&
-              !IsPotentiallyTruncated(I, BitWidth))
+          if (E.UserTreeIndices.size() > 1 &&
+              !all_of(E.Scalars, std::bind(IsPotentiallyTruncated, _1,
+                                           std::ref(BitWidth))))
             return false;
           bool NeedToExit = false;
           if (Checker && !AttemptCheckBitwidth(Checker, NeedToExit))
@@ -14266,26 +14264,22 @@ bool BoUpSLP::collectValuesToDemote(
         }
 
         ++MaxDepthLevel;
-        // Gather demoted constant operands.
-        for (unsigned Idx : seq<unsigned>(Start, End))
-          if (isa<Constant>(I->getOperand(Idx)))
-            DemotedConsts.try_emplace(I).first->getSecond().push_back(Idx);
-        // Record the value that we can demote.
-        ToDemote.push_back(V);
+        // Record the entry that we can demote.
+        ToDemote.push_back(E.Idx);
         return IsProfitableToDemote;
       };
-  switch (I->getOpcode()) {
+  switch (E.getOpcode()) {
 
   // We can always demote truncations and extensions. Since truncations can
   // seed additional demotion, we save the truncated value.
   case Instruction::Trunc:
     if (IsProfitableToDemoteRoot)
       IsProfitableToDemote = true;
-    return TryProcessInstruction(I, *ITE, BitWidth);
+    return TryProcessInstruction(BitWidth);
   case Instruction::ZExt:
   case Instruction::SExt:
     IsProfitableToDemote = true;
-    return TryProcessInstruction(I, *ITE, BitWidth);
+    return TryProcessInstruction(BitWidth);
 
   // We can demote certain binary operations if we can demote both of their
   // operands.
@@ -14295,112 +14289,128 @@ bool BoUpSLP::collectValuesToDemote(
   case Instruction::And:
   case Instruction::Or:
   case Instruction::Xor: {
-    return TryProcessInstruction(I, *ITE, BitWidth,
-                                 {I->getOperand(0), I->getOperand(1)});
+    return TryProcessInstruction(
+        BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)});
   }
   case Instruction::Shl: {
     // If we are truncating the result of this SHL, and if it's a shift of an
     // inrange amount, we can always perform a SHL in a smaller type.
     auto ShlChecker = [&](unsigned BitWidth, unsigned) {
-      KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
-      return AmtKnownBits.getMaxValue().ult(BitWidth);
+      return all_of(E.Scalars, [&](Value *V) {
+        auto *I = cast<Instruction>(V);
+        KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
+        return AmtKnownBits.getMaxValue().ult(BitWidth);
+      });
     };
     return TryProcessInstruction(
-        I, *ITE, BitWidth, {I->getOperand(0), I->getOperand(1)}, ShlChecker);
+        BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)}, ShlChecker);
   }
   case Instruction::LShr: {
     // If this is a truncate of a logical shr, we can truncate it to a smaller
     // lshr iff we know that the bits we would otherwise be shifting in are
     // already zeros.
     auto LShrChecker = [&](unsigned BitWidth, unsigned OrigBitWidth) {
-      KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
-      APInt ShiftedBits = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
-      return AmtKnownBits.getMaxValue().ult(BitWidth) &&
-             MaskedValueIsZero(I->getOperand(0), ShiftedBits,
-                               SimplifyQuery(*DL));
+      return all_of(E.Scalars, [&](Value *V) {
+        auto *I = cast<Instruction>(V);
+        KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
+        APInt ShiftedBits = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
+        return AmtKnownBits.getMaxValue().ult(BitWidth) &&
+               MaskedValueIsZero(I->getOperand(0), ShiftedBits,
+                                 SimplifyQuery(*DL));
+      });
     };
     return TryProcessInstruction(
-        I, *ITE, BitWidth, {I->getOperand(0), I->getOperand(1)}, LShrChecker);
+        BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)},
+        LShrChecker);
   }
   case Instruction::AShr: {
     // If this is a truncate of an arithmetic shr, we can truncate it to a
     // smaller ashr iff we know that all the bits from the sign bit of the
     // original type and the sign bit of the truncate type are similar.
     auto AShrChecker = [&](unsigned BitWidth, unsigned OrigBitWidth) {
-      KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
-      unsigned ShiftedBits = OrigBitWidth - BitWidth;
-      return AmtKnownBits.getMaxValue().ult(BitWidth) &&
-             ShiftedBits <
-                 ComputeNumSignBits(I->getOperand(0), *DL, 0, AC, nullptr, DT);
+      return all_of(E.Scalars, [&](Value *V) {
+        auto *I = cast<Instruction>(V);
+        KnownBits AmtKnownBits = computeKnownBits(I->getOperand(1), *DL);
+        unsigned ShiftedBits = OrigBitWidth - BitWidth;
+        return AmtKnownBits.getMaxValue().ult(BitWidth) &&
+               ShiftedBits < ComputeNumSignBits(I->getOperand(0), *DL, 0, AC,
+                                                nullptr, DT);
+      });
     };
     return TryProcessInstruction(
-        I, *ITE, BitWidth, {I->getOperand(0), I->getOperand(1)}, AShrChecker);
+        BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)},
+        AShrChecker);
   }
   case Instruction::UDiv:
   case Instruction::URem: {
     // UDiv and URem can be truncated if all the truncated bits are zero.
     auto Checker = [&](unsigned BitWidth, unsigned OrigBitWidth) {
       assert(BitWidth <= OrigBitWidth && "Unexpected bitwidths!");
-      APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
-      return MaskedValueIsZero(I->getOperand(0), Mask, SimplifyQuery(*DL)) &&
-             MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL));
+      return all_of(E.Scalars, [&](Value *V) {
+        auto *I = cast<Instruction>(V);
+        APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
+        return MaskedValueIsZero(I->getOperand(0), Mask, SimplifyQuery(*DL)) &&
+               MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL));
+      });
     };
-    return TryProcessInstruction(I, *ITE, BitWidth,
-                                 {I->getOperand(0), I->getOperand(1)}, Checker);
+    return TryProcessInstruction(
+        BitWidth, {getOperandEntry(&E, 0), getOperandEntry(&E, 1)}, Checker);
   }
 
   // We can demote selects if we can demote their true and false values.
   case Instruction::Select: {
-    Start = 1;
-    auto *SI = cast<SelectInst>(I);
-    return TryProcessInstruction(I, *ITE, BitWidth,
-                                 {SI->getTrueValue(), SI->getFalseValue()});
+    return TryProcessInstruction(
+        BitWidth, {getOperandEntry(&E, 1), getOperandEntry(&E, 2)});
   }
 
   // We can demote phis if we can demote all their incoming operands. Note that
   // we don't need to worry about cycles since we ensure single use above.
   case Instruction::PHI: {
-    PHINode *PN = cast<PHINode>(I);
-    SmallVector<Value *> Ops(PN->incoming_values().begin(),
-                             PN->incoming_values().end());
-    return TryProcessInstruction(I, *ITE, BitWidth, Ops);
+    const unsigned NumOps = E.getNumOperands();
+    SmallVector<const TreeEntry *> Ops(NumOps);
+    transform(seq<unsigned>(0, NumOps), Ops.begin(),
+              std::bind(&BoUpSLP::getOperandEntry, this, &E, _1));
+
+    return TryProcessInstruction(BitWidth, Ops);
   }
 
   case Instruction::Call: {
-    auto *IC = dyn_cast<IntrinsicInst>(I);
+    auto *IC = dyn_cast<IntrinsicInst>(E.getMainOp());
     if (!IC)
       break;
     Intrinsic::ID ID = getVectorIntrinsicIDForCall(IC, TLI);
     if (ID != Intrinsic::abs && ID != Intrinsic::smin &&
         ID != Intrinsic::smax && ID != Intrinsic::umin && ID != Intrinsic::umax)
       break;
-    SmallVector<Value *> Operands(1, I->getOperand(0));
+    SmallVector<const TreeEntry *, 2> Operands(1, getOperandEntry(&E, 0));
     function_ref<bool(unsigned, unsigned)> CallChecker;
     auto CompChecker = [&](unsigned BitWidth, unsigned OrigBitWidth) {
       assert(BitWidth <= OrigBitWidth && "Unexpected bitwidths!");
-      if (ID == Intrinsic::umin || ID == Intrinsic::umax) {
-        APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
-        return MaskedValueIsZero(I->getOperand(0), Mask, SimplifyQuery(*DL)) &&
-               MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL));
-      }
-      assert((ID == Intrinsic::smin || ID == Intrinsic::smax) &&
-             "Expected min/max intrinsics only.");
-      unsigned SignBits = OrigBitWidth - BitWidth;
-      return SignBits <= ComputeNumSignBits(I->getOperand(0), *DL, 0, AC,
-                                            nullptr, DT) &&
-             SignBits <=
-                 ComputeNumSignBits(I->getOperand(1), *DL, 0, AC, nullptr, DT);
+      return all_of(E.Scalars, [&](Value *V) {
+        auto *I = cast<Instruction>(V);
+        if (ID == Intrinsic::umin || ID == Intrinsic::umax) {
+          APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth);
+          return MaskedValueIsZero(I->getOperand(0), Mask,
+                                   SimplifyQuery(*DL)) &&
+                 MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL));
+        }
+        assert((ID == Intrinsic::smin || ID == Intrinsic::smax) &&
+               "Expected min/max intrinsics only.");
+        unsigned SignBits = OrigBitWidth - BitWidth;
+        return SignBits <= ComputeNumSignBits(I->getOperand(0), *DL, 0, AC,
+                                              nullptr, DT) &&
+               SignBits <= ComputeNumSignBits(I->getOperand(1), *DL, 0, AC,
+                                              nullptr, DT);
+      });
     };
-    End = 1;
     if (ID != Intrinsic::abs) {
-      Operands.push_back(I->getOperand(1));
-      End = 2;
+      Operands.push_back(getOperandEntry(&E, 1));
       CallChecker = CompChecker;
     }
     InstructionCost BestCost =
         std::numeric_limits<InstructionCost::CostType>::max();
     unsigned BestBitWidth = BitWidth;
-    unsigned VF = ITE->Scalars.size();
+    unsigned VF = E.Scalars.size();
     // Choose the best bitwidth based on cost estimations.
     auto Checker = [&](unsigned BitWidth, unsigned) {
       unsigned MinBW = PowerOf2Ceil(BitWidth);
@@ -14419,7 +14429,7 @@ bool BoUpSLP::collectValuesToDemote(
     [[maybe_unused]] bool NeedToExit;
     (void)AttemptCheckBitwidth(Checker, NeedToExit);
     BitWidth = BestBitWidth;
-    return TryProcessInstruction(I, *ITE, BitWidth, Operands, CallChecker);
+    return TryProcessInstruction(BitWidth, Operands, CallChecker);
   }
 
   // Otherwise, conservatively give up.
@@ -14473,26 +14483,27 @@ void BoUpSLP::computeMinimumValueSizes() {
     ++NodeIdx;
   }
 
-  // Analyzed in reduction already and not profitable - exit.
+  // Analyzed the reduction already and not profitable - exit.
   if (AnalyzedMinBWVals.contains(VectorizableTree[NodeIdx]->Scalars.front()))
     return;
 
-  SmallVector<Value *> ToDemote;
-  DenseMap<Instruction *, SmallVector<unsigned>> DemotedConsts;
-  auto ComputeMaxBitWidth = [&](ArrayRef<Value *> TreeRoot, unsigned VF,
-                                bool IsTopRoot, bool IsProfitableToDemoteRoot,
-                                unsigned Opcode, unsigned Limit,
-                                bool IsTruncRoot, bool IsSignedCmp) {
+  SmallVector<unsigned> ToDemote;
+  auto ComputeMaxBitWidth = [&](const TreeEntry &E, bool IsTopRoot,
+                                bool IsProfitableToDemoteRoot, unsigned Opcode,
+                                unsigned Limit, bool IsTruncRoot,
+                                bool IsSignedCmp) {
     ToDemote.clear();
-    auto *TreeRootIT = dyn_cast<IntegerType>(TreeRoot[0]->getType());
+    unsigned VF = E.getVectorFactor();
+    auto *TreeRootIT = dyn_cast<IntegerType>(E.Scalars.front()->getType());
     if (!TreeRootIT || !Opcode)
       return 0u;
 
-    if (AnalyzedMinBWVals.contains(TreeRoot.front()))
+    if (any_of(E.Scalars,
+               [&](Value *V) { return AnalyzedMinBWVals.contains(V); }))
       return 0u;
 
-    unsigned NumParts = TTI->getNumberOfParts(
-        FixedVectorType::get(TreeRoot.front()->getType(), VF));
+    unsigned NumParts =
+        TTI->getNumberOfParts(FixedVectorType::get(TreeRootIT, VF));
 
     // The maximum bit width required to represent all the values that can be
     // demoted without loss of precision. It would be safe to truncate the roots
@@ -14505,14 +14516,14 @@ void BoUpSLP::computeMinimumValueSizes() {
     // True.
     // Determine if the sign bit of all the roots is known to be zero. If not,
     // IsKnownPositive is set to False.
-    bool IsKnownPositive = !IsSignedCmp && all_of(TreeRoot, [&](Value *R) {
+    bool IsKnownPositive = !IsSignedCmp && all_of(E.Scalars, [&](Value *R) {
       KnownBits Known = computeKnownBits(R, *DL);
       return Known.isNonNegative();
     });
 
     // We first check if all the bits of the roots are demanded. If they're not,
     // we can truncate the roots to this narrower type.
-    for (auto *Root : TreeRoot) {
+    for (Value *Root : E.Scalars) {
       unsigned NumSignBits = ComputeNumSignBits(Root, *DL, 0, AC, nullptr, DT);
       TypeSize NumTypeBits = DL->getTypeSizeInBits(Root->getType());
       unsigned BitWidth1 = NumTypeBits - NumSignBits;
@@ -14557,23 +14568,22 @@ void BoUpSLP::computeMinimumValueSizes() {
     // Conservatively determine if we can actually truncate the roots of the
     // expression. Collect the values that can be demoted in ToDemote and
     // additional roots that require investigating in Roots.
-    for (auto *Root : TreeRoot) {
-      DenseSet<Value *> Visited;
-      unsigned MaxDepthLevel = IsTruncRoot ? Limit : 1;
-      bool NeedToDemote = IsProfitableToDemote;
-
-      if (!collectValuesToDemote(Root, IsProfitableToDemoteRoot, MaxBitWidth,
-                                 ToDemote, DemotedConsts, Visited,
-                                 MaxDepthLevel, NeedToDemote, IsTruncRoot) ||
-          (MaxDepthLevel <= Limit &&
-           !(((Opcode == Instruction::SExt || Opcode == Instruction::ZExt) &&
-              (!IsTopRoot || !(IsStoreOrInsertElt || UserIgnoreList) ||
-               DL->getTypeSizeInBits(Root->getType()) /
-                       DL->getTypeSizeInBits(
-                           cast<Instruction>(Root)->getOperand(0)->getType()) >
-                   2)))))
-        return 0u;
-    }
+    DenseSet<const TreeEntry *> Visited;
+    unsigned MaxDepthLevel = IsTruncRoot ? Limit : 1;
+    bool NeedToDemote = IsProfitableToDemote;
+
+    if (!collectValuesToDemote(E, IsProfitableToDemoteRoot, MaxBitWidth,
+                               ToDemote, Visited, MaxDepthLevel, NeedToDemote,
+                               IsTruncRoot) ||
+        (MaxDepthLevel <= Limit &&
+         !(((Opcode == Instruction::SExt || Opcode == Instruction::ZExt) &&
+            (!IsTopRoot || !(IsStoreOrInsertElt || UserIgnoreList) ||
+             DL->getTypeSizeInBits(TreeRootIT) /
+                     DL->getTypeSizeInBits(cast<Instruction>(E.Scalars.front())
+                                               ->getOperand(0)
+                                               ->getType()) >
+                 2)))))
+      return 0u;
     // Round MaxBitWidth up to the next power-of-two.
     MaxBitWidth = bit_ceil(MaxBitWidth);
 
@@ -14624,8 +14634,8 @@ void BoUpSLP::computeMinimumValueSizes() {
                 VectorizableTree.front()->Scalars.front()->getType()))
       Limit = 3;
     unsigned MaxBitWidth = ComputeMaxBitWidth(
-        TreeRoot, VectorizableTree[NodeIdx]->getVectorFactor(), IsTopRoot,
-        IsProfitableToDemoteRoot, Opcode, Limit, IsTruncRoot, IsSignedCmp);
+        *VectorizableTree[NodeIdx].get(), IsTopRoot, IsProfitableToDemoteRoot,
+        Opcode, Limit, IsTruncRoot, IsSignedCmp);
     if (ReductionBitWidth != 0 && (IsTopRoot || !RootDemotes.empty())) {
       if (MaxBitWidth != 0 && ReductionBitWidth < MaxBitWidth)
         ReductionBitWidth = bit_ceil(MaxBitWidth);
@@ -14634,13 +14644,15 @@ void BoUpSLP::computeMinimumValueSizes() {
     }
 
     for (unsigned Idx : RootDemotes) {
-      Value *V = VectorizableTree[Idx]->Scalars.front();
-      uint32_t OrigBitWidth = DL->getTypeSizeInBits(V->getType());
-      if (OrigBitWidth > MaxBitWidth) {
-      APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, MaxBitWidth);
-      if (MaskedValueIsZero(V, Mask, SimplifyQuery(*DL)))
-        ToDemote.push_back(V);
-      }
+      if (all_of(VectorizableTree[Idx]->Scalars, [&](Value *V) {
+            uint32_t OrigBitWidth = DL->getTypeSizeInBits(V->getType());
+            if (OrigBitWidth > MaxBitWidth) {
+              APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, MaxBitWidth);
+              return MaskedValueIsZero(V, Mask, SimplifyQuery(*DL));
+            }
+            return false;
+          }))
+        ToDemote.push_back(Idx);
     }
     RootDemotes.clear();
     IsTopRoot = false;
@@ -14687,9 +14699,8 @@ void BoUpSLP::computeMinimumValueSizes() {
 
     // Finally, map the values we can demote to the maximum bit with we
     // computed.
-    for (Value *Scalar : ToDemote) {
-      TreeEntry *TE = getTreeEntry(Scalar);
-      assert(TE && "Expected vectorized scalar.");
+    for (unsigned Idx : ToDemote) {
+      TreeEntry *TE = VectorizableTree[Idx].get();
       if (MinBWs.contains(TE))
         continue;
       bool IsSigned = TE->getOpcode() == Instruction::SExt ||
@@ -14697,22 +14708,6 @@ void BoUpSLP::computeMinimumValueSizes() {
                         return !isKnownNonNegative(R, SimplifyQuery(*DL));
                       });
       MinBWs.try_emplace(TE, MaxBitWidth, IsSigned);
-      const auto *I = cast<Instruction>(Scalar);
-      auto DCIt = DemotedConsts.find(I);
-      if (DCIt != DemotedConsts.end()) {
-        for (unsigned Idx : DCIt->getSecond()) {
-          // Check that all instructions operands are demoted.
-          const TreeEntry *CTE = getOperandEntry(TE, Idx);
-          if (all_of(TE->Scalars,
-                     [&](Value *V) {
-                       auto SIt = DemotedConsts.find(cast<Instruction>(V));
-                       return SIt != DemotedConsts.end() &&
-                              is_contained(SIt->getSecond(), Idx);
-                     }) ||
-              all_of(CTE->Scalars, IsaPred<Constant>))
-            MinBWs.try_emplace(CTE, MaxBitWidth, IsSigned);
-        }
-      }
     }
   }
 }

From 50d368aee981738cd05f3d16f5d1cfc122c9b0ab Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Wed, 10 Apr 2024 09:07:43 -0500
Subject: [PATCH 011/886] [LinkerWrapper] Relax ordering of static libraries
 for offloading (#87532)

Summary:
The linker wrapper attempts to maintain consistent semantics with
existing host invocations. Static libraries by default only extract if
there are non-weak symbols that remain undefined. However, we have
situations between linkers that put different meanings on ordering. The
ld.bfd linker requires static libraries to be defined after the symbols,
while `ld.lld` relaxes this rule. The linker wrapper went with the
former as it's the easier solution, however this has caused a lot of
issues as I've had to explain this rule to several people, it also make
it difficult to include things like `libc` in the OpenMP runtime because
it would sometimes be linked before or after.

This patch reworks the logic to more or less perform the following logic
for static libraries.

  1. Split library / object inputs.
  2. Include every object input and record its undefined symbols
  3. Repeatedly try to extract static libraries to resolve these
     symbols. If a file is extracted we need to check every library
     again to resolve any new undefined symbols.

This allows the following to work and will cause fewer issues when
replacing HIP, which does `--whole-archive` so it's very likely the old
logic will regress.
```console
$ clang -lfoo main.c -fopenmp --offload-arch=native
```
---
 clang/test/Driver/linker-wrapper-libs.c       |   4 +
 .../ClangLinkerWrapper.cpp                    | 121 ++++++++++++------
 2 files changed, 83 insertions(+), 42 deletions(-)

diff --git a/clang/test/Driver/linker-wrapper-libs.c b/clang/test/Driver/linker-wrapper-libs.c
index 9a78200d7d3cf..119e306857187 100644
--- a/clang/test/Driver/linker-wrapper-libs.c
+++ b/clang/test/Driver/linker-wrapper-libs.c
@@ -44,6 +44,8 @@ int bar() { return weak; }
 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.o -fembed-offload-object=%t.out
 // RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
 // RUN:   --linker-path=/usr/bin/ld %t.o %t.a -o a.out 2>&1 \
+// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
+// RUN:   --linker-path=/usr/bin/ld %t.a %t.o -o a.out 2>&1 \
 // RUN: | FileCheck %s --check-prefix=LIBRARY-RESOLVES
 
 // LIBRARY-RESOLVES: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx1030 {{.*}}.o {{.*}}.o
@@ -66,6 +68,8 @@ int bar() { return weak; }
 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.o -fembed-offload-object=%t.out
 // RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
 // RUN:   --linker-path=/usr/bin/ld %t.o %t.a -o a.out 2>&1 \
+// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
+// RUN:   --linker-path=/usr/bin/ld %t.a %t.o -o a.out 2>&1 \
 // RUN: | FileCheck %s --check-prefix=LIBRARY-GLOBAL
 
 // LIBRARY-GLOBAL: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx1030 {{.*}}.o {{.*}}.o
diff --git a/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp b/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
index 73e695a67093e..a1879fc7712dc 100644
--- a/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
+++ b/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
@@ -1448,9 +1448,9 @@ getDeviceInput(const ArgList &Args) {
   StringSaver Saver(Alloc);
 
   // Try to extract device code from the linker input files.
-  DenseMap<OffloadFile::TargetID, SmallVector<OffloadFile>> InputFiles;
-  DenseMap<OffloadFile::TargetID, DenseMap<StringRef, Symbol>> Syms;
   bool WholeArchive = Args.hasArg(OPT_wholearchive_flag) ? true : false;
+  SmallVector<OffloadFile> ObjectFilesToExtract;
+  SmallVector<OffloadFile> ArchiveFilesToExtract;
   for (const opt::Arg *Arg : Args.filtered(
            OPT_INPUT, OPT_library, OPT_whole_archive, OPT_no_whole_archive)) {
     if (Arg->getOption().matches(OPT_whole_archive) ||
@@ -1486,50 +1486,87 @@ getDeviceInput(const ArgList &Args) {
     if (Error Err = extractOffloadBinaries(Buffer, Binaries))
       return std::move(Err);
 
-    // We only extract archive members that are needed.
-    bool IsArchive = identify_magic(Buffer.getBuffer()) == file_magic::archive;
-    bool Extracted = true;
-    while (Extracted) {
-      Extracted = false;
-      for (OffloadFile &Binary : Binaries) {
-        // If the binary was previously extracted it will be set to null.
-        if (!Binary.getBinary())
+    for (auto &OffloadFile : Binaries) {
+      if (identify_magic(Buffer.getBuffer()) == file_magic::archive &&
+          !WholeArchive)
+        ArchiveFilesToExtract.emplace_back(std::move(OffloadFile));
+      else
+        ObjectFilesToExtract.emplace_back(std::move(OffloadFile));
+    }
+  }
+
+  // Link all standard input files and update the list of symbols.
+  DenseMap<OffloadFile::TargetID, SmallVector<OffloadFile>> InputFiles;
+  DenseMap<OffloadFile::TargetID, DenseMap<StringRef, Symbol>> Syms;
+  for (OffloadFile &Binary : ObjectFilesToExtract) {
+    if (!Binary.getBinary())
+      continue;
+
+    SmallVector<OffloadFile::TargetID> CompatibleTargets = {Binary};
+    for (const auto &[ID, Input] : InputFiles)
+      if (object::areTargetsCompatible(Binary, ID))
+        CompatibleTargets.emplace_back(ID);
+
+    for (const auto &[Index, ID] : llvm::enumerate(CompatibleTargets)) {
+      Expected<bool> ExtractOrErr = getSymbols(
+          Binary.getBinary()->getImage(), Binary.getBinary()->getOffloadKind(),
+          /*IsArchive=*/false, Saver, Syms[ID]);
+      if (!ExtractOrErr)
+        return ExtractOrErr.takeError();
+
+      // If another target needs this binary it must be copied instead.
+      if (Index == CompatibleTargets.size() - 1)
+        InputFiles[ID].emplace_back(std::move(Binary));
+      else
+        InputFiles[ID].emplace_back(Binary.copy());
+    }
+  }
+
+  // Archive members only extract if they define needed symbols. We do this
+  // after every regular input file so that libraries may be included out of
+  // order. This follows 'ld.lld' semantics which are more lenient.
+  bool Extracted = true;
+  while (Extracted) {
+    Extracted = false;
+    for (OffloadFile &Binary : ArchiveFilesToExtract) {
+      // If the binary was previously extracted it will be set to null.
+      if (!Binary.getBinary())
+        continue;
+
+      SmallVector<OffloadFile::TargetID> CompatibleTargets = {Binary};
+      for (const auto &[ID, Input] : InputFiles)
+        if (object::areTargetsCompatible(Binary, ID))
+          CompatibleTargets.emplace_back(ID);
+
+      for (const auto &[Index, ID] : llvm::enumerate(CompatibleTargets)) {
+        // Only extract an if we have an an object matching this target.
+        if (!InputFiles.count(ID))
           continue;
 
-        SmallVector<OffloadFile::TargetID> CompatibleTargets = {Binary};
-        for (const auto &[ID, Input] : InputFiles)
-          if (object::areTargetsCompatible(Binary, ID))
-            CompatibleTargets.emplace_back(ID);
-
-        for (const auto &[Index, ID] : llvm::enumerate(CompatibleTargets)) {
-          // Only extract an if we have an an object matching this target.
-          if (IsArchive && !WholeArchive && !InputFiles.count(ID))
-            continue;
-
-          Expected<bool> ExtractOrErr = getSymbols(
-              Binary.getBinary()->getImage(),
-              Binary.getBinary()->getOffloadKind(), IsArchive, Saver, Syms[ID]);
-          if (!ExtractOrErr)
-            return ExtractOrErr.takeError();
-
-          Extracted = !WholeArchive && *ExtractOrErr;
-
-          // Skip including the file if it is an archive that does not resolve
-          // any symbols.
-          if (IsArchive && !WholeArchive && !Extracted)
-            continue;
-
-          // If another target needs this binary it must be copied instead.
-          if (Index == CompatibleTargets.size() - 1)
-            InputFiles[ID].emplace_back(std::move(Binary));
-          else
-            InputFiles[ID].emplace_back(Binary.copy());
-        }
+        Expected<bool> ExtractOrErr =
+            getSymbols(Binary.getBinary()->getImage(),
+                       Binary.getBinary()->getOffloadKind(), /*IsArchive=*/true,
+                       Saver, Syms[ID]);
+        if (!ExtractOrErr)
+          return ExtractOrErr.takeError();
 
-        // If we extracted any files we need to check all the symbols again.
-        if (Extracted)
-          break;
+        Extracted = *ExtractOrErr;
+
+        // Skip including the file if it is an archive that does not resolve
+        // any symbols.
+        if (!Extracted)
+          continue;
+
+        // If another target needs this binary it must be copied instead.
+        if (Index == CompatibleTargets.size() - 1)
+          InputFiles[ID].emplace_back(std::move(Binary));
+        else
+          InputFiles[ID].emplace_back(Binary.copy());
       }
+
+      // If we extracted any files we need to check all the symbols again.
+      if (Extracted)
+        break;
     }
   }
 

From 6b35cbee3f577d9ee55f7277affa0fe194859b25 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Wed, 10 Apr 2024 18:09:39 +0400
Subject: [PATCH 012/886] [clang] Introduce `SemaSYCL` (#88086)

This patch moves SYCL-related `Sema` functions into new `SemaSYCL`
class, following the recent example of OpenACC and HLSL. This is a part
of the effort to split `Sema`. Additional context can be found in
#82217, #84184, #87634.
---
 clang/include/clang/Sema/Sema.h     | 55 ++++--------------------
 clang/include/clang/Sema/SemaSYCL.h | 65 +++++++++++++++++++++++++++++
 clang/lib/Parse/ParseExpr.cpp       |  5 ++-
 clang/lib/Sema/Sema.cpp             |  6 ++-
 clang/lib/Sema/SemaExpr.cpp         | 22 ----------
 clang/lib/Sema/SemaSYCL.cpp         | 52 +++++++++++++++++------
 clang/lib/Sema/TreeTransform.h      |  4 +-
 7 files changed, 121 insertions(+), 88 deletions(-)
 create mode 100644 clang/include/clang/Sema/SemaSYCL.h

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 9769d36900664..e3e255a0dd76f 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -184,6 +184,7 @@ class PseudoObjectExpr;
 class QualType;
 class SemaHLSL;
 class SemaOpenACC;
+class SemaSYCL;
 class StandardConversionSequence;
 class Stmt;
 class StringLiteral;
@@ -467,7 +468,6 @@ class Sema final : public SemaBase {
   // 37. Name Lookup for RISC-V Vector Intrinsic (SemaRISCVVectorLookup.cpp)
   // 38. CUDA (SemaCUDA.cpp)
   // 39. OpenMP Directives and Clauses (SemaOpenMP.cpp)
-  // 40. SYCL Constructs (SemaSYCL.cpp)
 
   /// \name Semantic Analysis
   /// Implementations are in Sema.cpp
@@ -974,6 +974,11 @@ class Sema final : public SemaBase {
     return *OpenACCPtr;
   }
 
+  SemaSYCL &SYCL() {
+    assert(SYCLPtr);
+    return *SYCLPtr;
+  }
+
 protected:
   friend class Parser;
   friend class InitializationSequence;
@@ -1006,6 +1011,7 @@ class Sema final : public SemaBase {
 
   std::unique_ptr<SemaHLSL> HLSLPtr;
   std::unique_ptr<SemaOpenACC> OpenACCPtr;
+  std::unique_ptr<SemaSYCL> SYCLPtr;
 
   ///@}
 
@@ -5455,15 +5461,6 @@ class Sema final : public SemaBase {
   ExprResult ActOnPredefinedExpr(SourceLocation Loc, tok::TokenKind Kind);
   ExprResult ActOnIntegerConstant(SourceLocation Loc, uint64_t Val);
 
-  ExprResult BuildSYCLUniqueStableNameExpr(SourceLocation OpLoc,
-                                           SourceLocation LParen,
-                                           SourceLocation RParen,
-                                           TypeSourceInfo *TSI);
-  ExprResult ActOnSYCLUniqueStableNameExpr(SourceLocation OpLoc,
-                                           SourceLocation LParen,
-                                           SourceLocation RParen,
-                                           ParsedType ParsedTy);
-
   bool CheckLoopHintExpr(Expr *E, SourceLocation Loc);
 
   ExprResult ActOnNumericConstant(const Token &Tok, Scope *UDLScope = nullptr);
@@ -14516,44 +14513,6 @@ class Sema final : public SemaBase {
                         OpenMPDirectiveKind CancelRegion);
 
   ///@}
-
-  //
-  //
-  // -------------------------------------------------------------------------
-  //
-  //
-
-  /// \name SYCL Constructs
-  /// Implementations are in SemaSYCL.cpp
-  ///@{
-
-public:
-  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
-  /// context is "used as device code".
-  ///
-  /// - If CurLexicalContext is a kernel function or it is known that the
-  ///   function will be emitted for the device, emits the diagnostics
-  ///   immediately.
-  /// - If CurLexicalContext is a function and we are compiling
-  ///   for the device, but we don't know that this function will be codegen'ed
-  ///   for devive yet, creates a diagnostic which is emitted if and when we
-  ///   realize that the function will be codegen'ed.
-  ///
-  /// Example usage:
-  ///
-  /// Diagnose __float128 type usage only from SYCL device code if the current
-  /// target doesn't support it
-  /// if (!S.Context.getTargetInfo().hasFloat128Type() &&
-  ///     S.getLangOpts().SYCLIsDevice)
-  ///   SYCLDiagIfDeviceCode(Loc, diag::err_type_unsupported) << "__float128";
-  SemaDiagnosticBuilder SYCLDiagIfDeviceCode(SourceLocation Loc,
-                                             unsigned DiagID);
-
-  void deepTypeCheckForSYCLDevice(SourceLocation UsedAt,
-                                  llvm::DenseSet<QualType> Visited,
-                                  ValueDecl *DeclToCheck);
-
-  ///@}
 };
 
 DeductionFailureInfo
diff --git a/clang/include/clang/Sema/SemaSYCL.h b/clang/include/clang/Sema/SemaSYCL.h
new file mode 100644
index 0000000000000..f0dcb92ee9ab3
--- /dev/null
+++ b/clang/include/clang/Sema/SemaSYCL.h
@@ -0,0 +1,65 @@
+//===----- SemaSYCL.h ------- Semantic Analysis for SYCL constructs -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file declares semantic analysis for SYCL constructs.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_SEMASYCL_H
+#define LLVM_CLANG_SEMA_SEMASYCL_H
+
+#include "clang/AST/Decl.h"
+#include "clang/AST/Type.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Sema/Ownership.h"
+#include "clang/Sema/SemaBase.h"
+#include "llvm/ADT/DenseSet.h"
+
+namespace clang {
+
+class SemaSYCL : public SemaBase {
+public:
+  SemaSYCL(Sema &S);
+
+  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
+  /// context is "used as device code".
+  ///
+  /// - If CurLexicalContext is a kernel function or it is known that the
+  ///   function will be emitted for the device, emits the diagnostics
+  ///   immediately.
+  /// - If CurLexicalContext is a function and we are compiling
+  ///   for the device, but we don't know yet that this function will be
+  ///   codegen'ed for the devive, creates a diagnostic which is emitted if and
+  ///   when we realize that the function will be codegen'ed.
+  ///
+  /// Example usage:
+  ///
+  /// Diagnose __float128 type usage only from SYCL device code if the current
+  /// target doesn't support it
+  /// if (!S.Context.getTargetInfo().hasFloat128Type() &&
+  ///     S.getLangOpts().SYCLIsDevice)
+  ///   DiagIfDeviceCode(Loc, diag::err_type_unsupported) << "__float128";
+  SemaDiagnosticBuilder DiagIfDeviceCode(SourceLocation Loc, unsigned DiagID);
+
+  void deepTypeCheckForDevice(SourceLocation UsedAt,
+                              llvm::DenseSet<QualType> Visited,
+                              ValueDecl *DeclToCheck);
+
+  ExprResult BuildUniqueStableNameExpr(SourceLocation OpLoc,
+                                       SourceLocation LParen,
+                                       SourceLocation RParen,
+                                       TypeSourceInfo *TSI);
+  ExprResult ActOnUniqueStableNameExpr(SourceLocation OpLoc,
+                                       SourceLocation LParen,
+                                       SourceLocation RParen,
+                                       ParsedType ParsedTy);
+};
+
+} // namespace clang
+
+#endif // LLVM_CLANG_SEMA_SEMASYCL_H
diff --git a/clang/lib/Parse/ParseExpr.cpp b/clang/lib/Parse/ParseExpr.cpp
index ae23cb432c439..d08e675604d19 100644
--- a/clang/lib/Parse/ParseExpr.cpp
+++ b/clang/lib/Parse/ParseExpr.cpp
@@ -30,6 +30,7 @@
 #include "clang/Sema/EnterExpressionEvaluationContext.h"
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaSYCL.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "llvm/ADT/SmallVector.h"
 #include <optional>
@@ -2490,8 +2491,8 @@ ExprResult Parser::ParseSYCLUniqueStableNameExpression() {
   if (T.consumeClose())
     return ExprError();
 
-  return Actions.ActOnSYCLUniqueStableNameExpr(OpLoc, T.getOpenLocation(),
-                                               T.getCloseLocation(), Ty.get());
+  return Actions.SYCL().ActOnUniqueStableNameExpr(
+      OpLoc, T.getOpenLocation(), T.getCloseLocation(), Ty.get());
 }
 
 /// Parse a sizeof or alignof expression.
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index 04eadb5f3b8ae..801b03a63dbc8 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -45,6 +45,7 @@
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/SemaOpenACC.h"
+#include "clang/Sema/SemaSYCL.h"
 #include "clang/Sema/TemplateDeduction.h"
 #include "clang/Sema/TemplateInstCallback.h"
 #include "clang/Sema/TypoCorrection.h"
@@ -201,6 +202,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       CurScope(nullptr), Ident_super(nullptr),
       HLSLPtr(std::make_unique<SemaHLSL>(*this)),
       OpenACCPtr(std::make_unique<SemaOpenACC>(*this)),
+      SYCLPtr(std::make_unique<SemaSYCL>(*this)),
       MSPointerToMemberRepresentationMethod(
           LangOpts.getMSPointerToMemberRepresentationMethod()),
       MSStructPragmaOn(false), VtorDispStack(LangOpts.getVtorDispMode()),
@@ -1903,7 +1905,7 @@ Sema::targetDiag(SourceLocation Loc, unsigned DiagID, const FunctionDecl *FD) {
                                       : CUDADiagIfHostCode(Loc, DiagID);
 
   if (getLangOpts().SYCLIsDevice)
-    return SYCLDiagIfDeviceCode(Loc, DiagID);
+    return SYCL().DiagIfDeviceCode(Loc, DiagID);
 
   return SemaDiagnosticBuilder(SemaDiagnosticBuilder::K_Immediate, Loc, DiagID,
                                FD, *this);
@@ -1919,7 +1921,7 @@ void Sema::checkTypeSupport(QualType Ty, SourceLocation Loc, ValueDecl *D) {
   // constant byte size like zero length arrays. So, do a deep check for SYCL.
   if (D && LangOpts.SYCLIsDevice) {
     llvm::DenseSet<QualType> Visited;
-    deepTypeCheckForSYCLDevice(Loc, Visited, D);
+    SYCL().deepTypeCheckForDevice(Loc, Visited, D);
   }
 
   Decl *C = cast<Decl>(getCurLexicalContext());
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 594c11788f4e7..4d4ef9b16381b 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -3794,28 +3794,6 @@ ExprResult Sema::BuildPredefinedExpr(SourceLocation Loc,
                                 SL);
 }
 
-ExprResult Sema::BuildSYCLUniqueStableNameExpr(SourceLocation OpLoc,
-                                               SourceLocation LParen,
-                                               SourceLocation RParen,
-                                               TypeSourceInfo *TSI) {
-  return SYCLUniqueStableNameExpr::Create(Context, OpLoc, LParen, RParen, TSI);
-}
-
-ExprResult Sema::ActOnSYCLUniqueStableNameExpr(SourceLocation OpLoc,
-                                               SourceLocation LParen,
-                                               SourceLocation RParen,
-                                               ParsedType ParsedTy) {
-  TypeSourceInfo *TSI = nullptr;
-  QualType Ty = GetTypeFromParser(ParsedTy, &TSI);
-
-  if (Ty.isNull())
-    return ExprError();
-  if (!TSI)
-    TSI = Context.getTrivialTypeSourceInfo(Ty, LParen);
-
-  return BuildSYCLUniqueStableNameExpr(OpLoc, LParen, RParen, TSI);
-}
-
 ExprResult Sema::ActOnPredefinedExpr(SourceLocation Loc, tok::TokenKind Kind) {
   return BuildPredefinedExpr(Loc, getPredefinedExprKind(Kind));
 }
diff --git a/clang/lib/Sema/SemaSYCL.cpp b/clang/lib/Sema/SemaSYCL.cpp
index 18ebaa13346a4..18f6d8f030473 100644
--- a/clang/lib/Sema/SemaSYCL.cpp
+++ b/clang/lib/Sema/SemaSYCL.cpp
@@ -8,6 +8,7 @@
 // This implements Semantic Analysis for SYCL constructs.
 //===----------------------------------------------------------------------===//
 
+#include "clang/Sema/SemaSYCL.h"
 #include "clang/AST/Mangle.h"
 #include "clang/Sema/Sema.h"
 #include "clang/Sema/SemaDiagnostic.h"
@@ -18,28 +19,30 @@ using namespace clang;
 // SYCL device specific diagnostics implementation
 // -----------------------------------------------------------------------------
 
-Sema::SemaDiagnosticBuilder Sema::SYCLDiagIfDeviceCode(SourceLocation Loc,
+SemaSYCL::SemaSYCL(Sema &S) : SemaBase(S) {}
+
+Sema::SemaDiagnosticBuilder SemaSYCL::DiagIfDeviceCode(SourceLocation Loc,
                                                        unsigned DiagID) {
   assert(getLangOpts().SYCLIsDevice &&
          "Should only be called during SYCL compilation");
-  FunctionDecl *FD = dyn_cast<FunctionDecl>(getCurLexicalContext());
+  FunctionDecl *FD = dyn_cast<FunctionDecl>(SemaRef.getCurLexicalContext());
   SemaDiagnosticBuilder::Kind DiagKind = [this, FD] {
     if (!FD)
       return SemaDiagnosticBuilder::K_Nop;
-    if (getEmissionStatus(FD) == Sema::FunctionEmissionStatus::Emitted)
+    if (SemaRef.getEmissionStatus(FD) == Sema::FunctionEmissionStatus::Emitted)
       return SemaDiagnosticBuilder::K_ImmediateWithCallStack;
     return SemaDiagnosticBuilder::K_Deferred;
   }();
-  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, FD, *this);
+  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, FD, SemaRef);
 }
 
-static bool isZeroSizedArray(Sema &SemaRef, QualType Ty) {
-  if (const auto *CAT = SemaRef.getASTContext().getAsConstantArrayType(Ty))
+static bool isZeroSizedArray(SemaSYCL &S, QualType Ty) {
+  if (const auto *CAT = S.getASTContext().getAsConstantArrayType(Ty))
     return CAT->isZeroSize();
   return false;
 }
 
-void Sema::deepTypeCheckForSYCLDevice(SourceLocation UsedAt,
+void SemaSYCL::deepTypeCheckForDevice(SourceLocation UsedAt,
                                       llvm::DenseSet<QualType> Visited,
                                       ValueDecl *DeclToCheck) {
   assert(getLangOpts().SYCLIsDevice &&
@@ -51,18 +54,18 @@ void Sema::deepTypeCheckForSYCLDevice(SourceLocation UsedAt,
   auto Check = [&](QualType TypeToCheck, const ValueDecl *D) {
     bool ErrorFound = false;
     if (isZeroSizedArray(*this, TypeToCheck)) {
-      SYCLDiagIfDeviceCode(UsedAt, diag::err_typecheck_zero_array_size) << 1;
+      DiagIfDeviceCode(UsedAt, diag::err_typecheck_zero_array_size) << 1;
       ErrorFound = true;
     }
     // Checks for other types can also be done here.
     if (ErrorFound) {
       if (NeedToEmitNotes) {
         if (auto *FD = dyn_cast<FieldDecl>(D))
-          SYCLDiagIfDeviceCode(FD->getLocation(),
-                               diag::note_illegal_field_declared_here)
+          DiagIfDeviceCode(FD->getLocation(),
+                           diag::note_illegal_field_declared_here)
               << FD->getType()->isPointerType() << FD->getType();
         else
-          SYCLDiagIfDeviceCode(D->getLocation(), diag::note_declared_at);
+          DiagIfDeviceCode(D->getLocation(), diag::note_declared_at);
       }
     }
 
@@ -93,8 +96,8 @@ void Sema::deepTypeCheckForSYCLDevice(SourceLocation UsedAt,
     auto EmitHistory = [&]() {
       // The first element is always nullptr.
       for (uint64_t Index = 1; Index < History.size(); ++Index) {
-        SYCLDiagIfDeviceCode(History[Index]->getLocation(),
-                             diag::note_within_field_of_type)
+        DiagIfDeviceCode(History[Index]->getLocation(),
+                         diag::note_within_field_of_type)
             << History[Index]->getType();
       }
     };
@@ -130,3 +133,26 @@ void Sema::deepTypeCheckForSYCLDevice(SourceLocation UsedAt,
     }
   } while (!StackForRecursion.empty());
 }
+
+ExprResult SemaSYCL::BuildUniqueStableNameExpr(SourceLocation OpLoc,
+                                               SourceLocation LParen,
+                                               SourceLocation RParen,
+                                               TypeSourceInfo *TSI) {
+  return SYCLUniqueStableNameExpr::Create(getASTContext(), OpLoc, LParen,
+                                          RParen, TSI);
+}
+
+ExprResult SemaSYCL::ActOnUniqueStableNameExpr(SourceLocation OpLoc,
+                                               SourceLocation LParen,
+                                               SourceLocation RParen,
+                                               ParsedType ParsedTy) {
+  TypeSourceInfo *TSI = nullptr;
+  QualType Ty = SemaRef.GetTypeFromParser(ParsedTy, &TSI);
+
+  if (Ty.isNull())
+    return ExprError();
+  if (!TSI)
+    TSI = getASTContext().getTrivialTypeSourceInfo(Ty, LParen);
+
+  return BuildUniqueStableNameExpr(OpLoc, LParen, RParen, TSI);
+}
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index d4d2fa61d65ea..79d60588ae536 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -40,6 +40,7 @@
 #include "clang/Sema/SemaDiagnostic.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/SemaOpenACC.h"
+#include "clang/Sema/SemaSYCL.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/Support/ErrorHandling.h"
 #include <algorithm>
@@ -2632,7 +2633,8 @@ class TreeTransform {
                                              SourceLocation LParen,
                                              SourceLocation RParen,
                                              TypeSourceInfo *TSI) {
-    return getSema().BuildSYCLUniqueStableNameExpr(OpLoc, LParen, RParen, TSI);
+    return getSema().SYCL().BuildUniqueStableNameExpr(OpLoc, LParen, RParen,
+                                                      TSI);
   }
 
   /// Build a new predefined expression.

From 0c7b92a42a36563dfd28e3a828e87f4f3a6e4311 Mon Sep 17 00:00:00 2001
From: Erich Keane <ekeane@nvidia.com>
Date: Wed, 10 Apr 2024 07:10:24 -0700
Subject: [PATCH 013/886] [OpenACC] Implement Default clause for Compute
 Constructs (#88135)

As a followup to my previous commits, this is an implementation of a
single clause, in this case the 'default' clause. This implements all
semantic analysis for it on compute clauses, and continues to leave it
rejected for all others (some as 'doesnt appertain', others as 'not
implemented' as appropriate).

This also implements and tests the TreeTransform as requested in the
previous patch.
---
 clang/include/clang/AST/OpenACCClause.h       | 38 ++++++++++
 .../clang/Basic/DiagnosticSemaKinds.td        |  4 +
 clang/include/clang/Basic/OpenACCKinds.h      | 23 ++++++
 clang/include/clang/Sema/SemaOpenACC.h        | 19 ++++-
 clang/lib/AST/OpenACCClause.cpp               | 19 +++++
 clang/lib/AST/StmtProfile.cpp                 |  5 ++
 clang/lib/AST/TextNodeDumper.cpp              | 11 +++
 clang/lib/Parse/ParseOpenACC.cpp              |  8 +-
 clang/lib/Sema/SemaOpenACC.cpp                | 65 ++++++++++++++--
 clang/lib/Sema/TreeTransform.h                | 42 ++++++++++-
 clang/lib/Serialization/ASTReader.cpp         |  7 +-
 clang/lib/Serialization/ASTWriter.cpp         |  7 +-
 clang/test/ParserOpenACC/parse-clauses.c      | 20 ++---
 .../SemaOpenACC/compute-construct-ast.cpp     | 14 +++-
 .../compute-construct-clause-ast.cpp          | 74 +++++++++++++++++++
 .../compute-construct-default-clause.c        | 55 ++++++++++++++
 .../compute-construct-default-clause.cpp      | 39 ++++++++++
 17 files changed, 417 insertions(+), 33 deletions(-)
 create mode 100644 clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
 create mode 100644 clang/test/SemaOpenACC/compute-construct-default-clause.c
 create mode 100644 clang/test/SemaOpenACC/compute-construct-default-clause.cpp

diff --git a/clang/include/clang/AST/OpenACCClause.h b/clang/include/clang/AST/OpenACCClause.h
index 06a0098bbda4c..27e4e1a12c983 100644
--- a/clang/include/clang/AST/OpenACCClause.h
+++ b/clang/include/clang/AST/OpenACCClause.h
@@ -51,6 +51,36 @@ class OpenACCClauseWithParams : public OpenACCClause {
   SourceLocation getLParenLoc() const { return LParenLoc; }
 };
 
+/// A 'default' clause, has the optional 'none' or 'present' argument.
+class OpenACCDefaultClause : public OpenACCClauseWithParams {
+  friend class ASTReaderStmt;
+  friend class ASTWriterStmt;
+
+  OpenACCDefaultClauseKind DefaultClauseKind;
+
+protected:
+  OpenACCDefaultClause(OpenACCDefaultClauseKind K, SourceLocation BeginLoc,
+                       SourceLocation LParenLoc, SourceLocation EndLoc)
+      : OpenACCClauseWithParams(OpenACCClauseKind::Default, BeginLoc, LParenLoc,
+                                EndLoc),
+        DefaultClauseKind(K) {
+    assert((DefaultClauseKind == OpenACCDefaultClauseKind::None ||
+            DefaultClauseKind == OpenACCDefaultClauseKind::Present) &&
+           "Invalid Clause Kind");
+  }
+
+public:
+  OpenACCDefaultClauseKind getDefaultClauseKind() const {
+    return DefaultClauseKind;
+  }
+
+  static OpenACCDefaultClause *Create(const ASTContext &C,
+                                      OpenACCDefaultClauseKind K,
+                                      SourceLocation BeginLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+};
+
 template <class Impl> class OpenACCClauseVisitor {
   Impl &getDerived() { return static_cast<Impl &>(*this); }
 
@@ -66,6 +96,8 @@ template <class Impl> class OpenACCClauseVisitor {
 
     switch (C->getClauseKind()) {
     case OpenACCClauseKind::Default:
+      VisitOpenACCDefaultClause(*cast<OpenACCDefaultClause>(C));
+      return;
     case OpenACCClauseKind::Finalize:
     case OpenACCClauseKind::IfPresent:
     case OpenACCClauseKind::Seq:
@@ -112,6 +144,10 @@ template <class Impl> class OpenACCClauseVisitor {
     }
     llvm_unreachable("Invalid Clause kind");
   }
+
+  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause) {
+    return getDerived().VisitOpenACCDefaultClause(Clause);
+  }
 };
 
 class OpenACCClausePrinter final
@@ -128,6 +164,8 @@ class OpenACCClausePrinter final
     }
   }
   OpenACCClausePrinter(raw_ostream &OS) : OS(OS) {}
+
+  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause);
 };
 
 } // namespace clang
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 64c58ab36338b..059a8f58da5db 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -12265,6 +12265,10 @@ def err_acc_construct_appertainment
             "be used in a statement context">;
 def err_acc_clause_appertainment
     : Error<"OpenACC '%1' clause is not valid on '%0' directive">;
+def err_acc_duplicate_clause_disallowed
+    : Error<"OpenACC '%1' clause cannot appear more than once on a '%0' "
+            "directive">;
+def note_acc_previous_clause_here : Note<"previous clause is here">;
 def err_acc_branch_in_out_compute_construct
     : Error<"invalid %select{branch|return|throw}0 %select{out of|into}1 "
             "OpenACC Compute Construct">;
diff --git a/clang/include/clang/Basic/OpenACCKinds.h b/clang/include/clang/Basic/OpenACCKinds.h
index 95fc35a5bedb7..e191e9e0a5a15 100644
--- a/clang/include/clang/Basic/OpenACCKinds.h
+++ b/clang/include/clang/Basic/OpenACCKinds.h
@@ -419,6 +419,29 @@ enum class OpenACCDefaultClauseKind {
   Invalid,
 };
 
+template <typename StreamTy>
+inline StreamTy &printOpenACCDefaultClauseKind(StreamTy &Out,
+                                               OpenACCDefaultClauseKind K) {
+  switch (K) {
+  case OpenACCDefaultClauseKind::None:
+    return Out << "none";
+  case OpenACCDefaultClauseKind::Present:
+    return Out << "present";
+  case OpenACCDefaultClauseKind::Invalid:
+    return Out << "<invalid>";
+  }
+}
+
+inline const StreamingDiagnostic &operator<<(const StreamingDiagnostic &Out,
+                                             OpenACCDefaultClauseKind K) {
+  return printOpenACCDefaultClauseKind(Out, K);
+}
+
+inline llvm::raw_ostream &operator<<(llvm::raw_ostream &Out,
+                                     OpenACCDefaultClauseKind K) {
+  return printOpenACCDefaultClauseKind(Out, K);
+}
+
 enum class OpenACCReductionOperator {
   /// '+'.
   Addition,
diff --git a/clang/include/clang/Sema/SemaOpenACC.h b/clang/include/clang/Sema/SemaOpenACC.h
index 45929e4a9db3f..27aaee164a288 100644
--- a/clang/include/clang/Sema/SemaOpenACC.h
+++ b/clang/include/clang/Sema/SemaOpenACC.h
@@ -19,6 +19,7 @@
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Sema/Ownership.h"
 #include "clang/Sema/SemaBase.h"
+#include <variant>
 
 namespace clang {
 class OpenACCClause;
@@ -35,7 +36,11 @@ class SemaOpenACC : public SemaBase {
     SourceRange ClauseRange;
     SourceLocation LParenLoc;
 
-    // TODO OpenACC: Add variant here to store details of individual clauses.
+    struct DefaultDetails {
+      OpenACCDefaultClauseKind DefaultClauseKind;
+    };
+
+    std::variant<DefaultDetails> Details;
 
   public:
     OpenACCParsedClause(OpenACCDirectiveKind DirKind,
@@ -52,8 +57,20 @@ class SemaOpenACC : public SemaBase {
 
     SourceLocation getEndLoc() const { return ClauseRange.getEnd(); }
 
+    OpenACCDefaultClauseKind getDefaultClauseKind() const {
+      assert(ClauseKind == OpenACCClauseKind::Default &&
+             "Parsed clause is not a default clause");
+      return std::get<DefaultDetails>(Details).DefaultClauseKind;
+    }
+
     void setLParenLoc(SourceLocation EndLoc) { LParenLoc = EndLoc; }
     void setEndLoc(SourceLocation EndLoc) { ClauseRange.setEnd(EndLoc); }
+
+    void setDefaultDetails(OpenACCDefaultClauseKind DefKind) {
+      assert(ClauseKind == OpenACCClauseKind::Default &&
+             "Parsed clause is not a default clause");
+      Details = DefaultDetails{DefKind};
+    }
   };
 
   SemaOpenACC(Sema &S);
diff --git a/clang/lib/AST/OpenACCClause.cpp b/clang/lib/AST/OpenACCClause.cpp
index e1db872f25c32..c83128b60e3ac 100644
--- a/clang/lib/AST/OpenACCClause.cpp
+++ b/clang/lib/AST/OpenACCClause.cpp
@@ -15,3 +15,22 @@
 #include "clang/AST/ASTContext.h"
 
 using namespace clang;
+
+OpenACCDefaultClause *OpenACCDefaultClause::Create(const ASTContext &C,
+                                                   OpenACCDefaultClauseKind K,
+                                                   SourceLocation BeginLoc,
+                                                   SourceLocation LParenLoc,
+                                                   SourceLocation EndLoc) {
+  void *Mem =
+      C.Allocate(sizeof(OpenACCDefaultClause), alignof(OpenACCDefaultClause));
+
+  return new (Mem) OpenACCDefaultClause(K, BeginLoc, LParenLoc, EndLoc);
+}
+
+//===----------------------------------------------------------------------===//
+//  OpenACC clauses printing methods
+//===----------------------------------------------------------------------===//
+void OpenACCClausePrinter::VisitOpenACCDefaultClause(
+    const OpenACCDefaultClause &C) {
+  OS << "default(" << C.getDefaultClauseKind() << ")";
+}
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index bec8bc71f5554..be3dd4b673cf9 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2456,7 +2456,12 @@ class OpenACCClauseProfiler
       Visit(Clause);
     }
   }
+  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause);
 };
+
+/// Nothing to do here, there are no sub-statements.
+void OpenACCClauseProfiler::VisitOpenACCDefaultClause(
+    const OpenACCDefaultClause &Clause) {}
 } // namespace
 
 void StmtProfiler::VisitOpenACCComputeConstruct(
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 431f5d8bdb2b5..085a7f51ce99a 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -390,6 +390,17 @@ void TextNodeDumper::Visit(const OpenACCClause *C) {
   {
     ColorScope Color(OS, ShowColors, AttrColor);
     OS << C->getClauseKind();
+
+    // Handle clauses with parens for types that have no children, likely
+    // because there is no sub expression.
+    switch (C->getClauseKind()) {
+    case OpenACCClauseKind::Default:
+      OS << '(' << cast<OpenACCDefaultClause>(C)->getDefaultClauseKind() << ')';
+      break;
+    default:
+      // Nothing to do here.
+      break;
+    }
   }
   dumpPointer(C);
   dumpSourceRange(SourceRange(C->getBeginLoc(), C->getEndLoc()));
diff --git a/clang/lib/Parse/ParseOpenACC.cpp b/clang/lib/Parse/ParseOpenACC.cpp
index f434e1542c801..59a4a5f534676 100644
--- a/clang/lib/Parse/ParseOpenACC.cpp
+++ b/clang/lib/Parse/ParseOpenACC.cpp
@@ -831,9 +831,13 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
 
       ConsumeToken();
 
-      if (getOpenACCDefaultClauseKind(DefKindTok) ==
-          OpenACCDefaultClauseKind::Invalid)
+      OpenACCDefaultClauseKind DefKind =
+          getOpenACCDefaultClauseKind(DefKindTok);
+
+      if (DefKind == OpenACCDefaultClauseKind::Invalid)
         Diag(DefKindTok, diag::err_acc_invalid_default_clause_kind);
+      else
+        ParsedClause.setDefaultDetails(DefKind);
 
       break;
     }
diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
index 2ba1e49b5739d..b6afb80b873e2 100644
--- a/clang/lib/Sema/SemaOpenACC.cpp
+++ b/clang/lib/Sema/SemaOpenACC.cpp
@@ -39,11 +39,27 @@ bool diagnoseConstructAppertainment(SemaOpenACC &S, OpenACCDirectiveKind K,
 
 bool doesClauseApplyToDirective(OpenACCDirectiveKind DirectiveKind,
                                 OpenACCClauseKind ClauseKind) {
-  // FIXME: For each clause as we implement them, we can add the
-  // 'legalization' list here.
-
-  // Do nothing so we can go to the 'unimplemented' diagnostic instead.
-  return true;
+  switch (ClauseKind) {
+    // FIXME: For each clause as we implement them, we can add the
+    // 'legalization' list here.
+  case OpenACCClauseKind::Default:
+    switch (DirectiveKind) {
+    case OpenACCDirectiveKind::Parallel:
+    case OpenACCDirectiveKind::Serial:
+    case OpenACCDirectiveKind::Kernels:
+    case OpenACCDirectiveKind::ParallelLoop:
+    case OpenACCDirectiveKind::SerialLoop:
+    case OpenACCDirectiveKind::KernelsLoop:
+    case OpenACCDirectiveKind::Data:
+      return true;
+    default:
+      return false;
+    }
+  default:
+    // Do nothing so we can go to the 'unimplemented' diagnostic instead.
+    return true;
+  }
+  llvm_unreachable("Invalid clause kind");
 }
 } // namespace
 
@@ -63,8 +79,43 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     return nullptr;
   }
 
-  // TODO OpenACC: Switch over the clauses we implement here and 'create'
-  // them.
+  switch (Clause.getClauseKind()) {
+  case OpenACCClauseKind::Default: {
+    // Restrictions only properly implemented on 'compute' constructs, and
+    // 'compute' constructs are the only construct that can do anything with
+    // this yet, so skip/treat as unimplemented in this case.
+    if (Clause.getDirectiveKind() != OpenACCDirectiveKind::Parallel &&
+        Clause.getDirectiveKind() != OpenACCDirectiveKind::Serial &&
+        Clause.getDirectiveKind() != OpenACCDirectiveKind::Kernels)
+      break;
+
+    // Don't add an invalid clause to the AST.
+    if (Clause.getDefaultClauseKind() == OpenACCDefaultClauseKind::Invalid)
+      return nullptr;
+
+    // OpenACC 3.3, Section 2.5.4:
+    // At most one 'default' clause may appear, and it must have a value of
+    // either 'none' or 'present'.
+    // Second half of the sentence is diagnosed during parsing.
+    auto Itr = llvm::find_if(ExistingClauses, [](const OpenACCClause *C) {
+      return C->getClauseKind() == OpenACCClauseKind::Default;
+    });
+
+    if (Itr != ExistingClauses.end()) {
+      SemaRef.Diag(Clause.getBeginLoc(),
+                   diag::err_acc_duplicate_clause_disallowed)
+          << Clause.getDirectiveKind() << Clause.getClauseKind();
+      SemaRef.Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+      return nullptr;
+    }
+
+    return OpenACCDefaultClause::Create(
+        getASTContext(), Clause.getDefaultClauseKind(), Clause.getBeginLoc(),
+        Clause.getLParenLoc(), Clause.getEndLoc());
+  }
+  default:
+    break;
+  }
 
   Diag(Clause.getBeginLoc(), diag::warn_acc_clause_unimplemented)
       << Clause.getClauseKind();
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 79d60588ae536..33a9356e82f40 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -4035,6 +4035,11 @@ class TreeTransform {
   llvm::SmallVector<OpenACCClause *>
   TransformOpenACCClauseList(OpenACCDirectiveKind DirKind,
                              ArrayRef<const OpenACCClause *> OldClauses);
+
+  OpenACCClause *
+  TransformOpenACCClause(ArrayRef<const OpenACCClause *> ExistingClauses,
+                         OpenACCDirectiveKind DirKind,
+                         const OpenACCClause *OldClause);
 };
 
 template <typename Derived>
@@ -11075,13 +11080,44 @@ OMPClause *TreeTransform<Derived>::TransformOMPXBareClause(OMPXBareClause *C) {
 //===----------------------------------------------------------------------===//
 // OpenACC transformation
 //===----------------------------------------------------------------------===//
+template <typename Derived>
+OpenACCClause *TreeTransform<Derived>::TransformOpenACCClause(
+    ArrayRef<const OpenACCClause *> ExistingClauses,
+    OpenACCDirectiveKind DirKind, const OpenACCClause *OldClause) {
+
+  SemaOpenACC::OpenACCParsedClause ParsedClause(
+      DirKind, OldClause->getClauseKind(), OldClause->getBeginLoc());
+  ParsedClause.setEndLoc(OldClause->getEndLoc());
+
+  if (const auto *WithParms = dyn_cast<OpenACCClauseWithParams>(OldClause))
+    ParsedClause.setLParenLoc(WithParms->getLParenLoc());
+
+  switch (OldClause->getClauseKind()) {
+  case OpenACCClauseKind::Default:
+    // There is nothing to do here as nothing dependent can appear in this
+    // clause. So just set the values so Sema can set the right value.
+    ParsedClause.setDefaultDetails(
+        cast<OpenACCDefaultClause>(OldClause)->getDefaultClauseKind());
+    break;
+  default:
+    assert(false && "Unhandled OpenACC clause in TreeTransform");
+    return nullptr;
+  }
+
+  return getSema().OpenACC().ActOnClause(ExistingClauses, ParsedClause);
+}
+
 template <typename Derived>
 llvm::SmallVector<OpenACCClause *>
 TreeTransform<Derived>::TransformOpenACCClauseList(
     OpenACCDirectiveKind DirKind, ArrayRef<const OpenACCClause *> OldClauses) {
-  // TODO OpenACC: Ensure we loop through the list and transform the individual
-  // clauses.
-  return {};
+  llvm::SmallVector<OpenACCClause *> TransformedClauses;
+  for (const auto *Clause : OldClauses) {
+    if (OpenACCClause *TransformedClause = getDerived().TransformOpenACCClause(
+            TransformedClauses, DirKind, Clause))
+      TransformedClauses.push_back(TransformedClause);
+  }
+  return TransformedClauses;
 }
 
 template <typename Derived>
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index fa5bb9f2d5435..679302e7a838f 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -11763,7 +11763,12 @@ OpenACCClause *ASTRecordReader::readOpenACCClause() {
   [[maybe_unused]] SourceLocation EndLoc = readSourceLocation();
 
   switch (ClauseKind) {
-  case OpenACCClauseKind::Default:
+  case OpenACCClauseKind::Default: {
+    SourceLocation LParenLoc = readSourceLocation();
+    OpenACCDefaultClauseKind DCK = readEnum<OpenACCDefaultClauseKind>();
+    return OpenACCDefaultClause::Create(getContext(), DCK, BeginLoc, LParenLoc,
+                                        EndLoc);
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index baf03f69d7306..4cd74b1ba9d72 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -7406,7 +7406,12 @@ void ASTRecordWriter::writeOpenACCClause(const OpenACCClause *C) {
   writeSourceLocation(C->getEndLoc());
 
   switch (C->getClauseKind()) {
-  case OpenACCClauseKind::Default:
+  case OpenACCClauseKind::Default: {
+    const auto *DC = cast<OpenACCDefaultClause>(C);
+    writeSourceLocation(DC->getLParenLoc());
+    writeEnum(DC->getDefaultClauseKind());
+    return;
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
diff --git a/clang/test/ParserOpenACC/parse-clauses.c b/clang/test/ParserOpenACC/parse-clauses.c
index b58b332ad3245..b363a0cb1362b 100644
--- a/clang/test/ParserOpenACC/parse-clauses.c
+++ b/clang/test/ParserOpenACC/parse-clauses.c
@@ -173,45 +173,37 @@ void DefaultClause() {
 #pragma acc serial default), seq
   for(;;){}
 
-  // expected-error@+2{{expected identifier}}
-  // expected-warning@+1{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+1{{expected identifier}}
 #pragma acc serial default()
   for(;;){}
 
-  // expected-error@+3{{expected identifier}}
-  // expected-warning@+2{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+2{{expected identifier}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial default() seq
   for(;;){}
 
-  // expected-error@+3{{expected identifier}}
-  // expected-warning@+2{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+2{{expected identifier}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial default(), seq
   for(;;){}
 
-  // expected-error@+2{{invalid value for 'default' clause; expected 'present' or 'none'}}
-  // expected-warning@+1{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+1{{invalid value for 'default' clause; expected 'present' or 'none'}}
 #pragma acc serial default(invalid)
   for(;;){}
 
-  // expected-error@+3{{invalid value for 'default' clause; expected 'present' or 'none'}}
-  // expected-warning@+2{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+2{{invalid value for 'default' clause; expected 'present' or 'none'}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial default(auto) seq
   for(;;){}
 
-  // expected-error@+3{{invalid value for 'default' clause; expected 'present' or 'none'}}
-  // expected-warning@+2{{OpenACC clause 'default' not yet implemented, clause ignored}}
+  // expected-error@+2{{invalid value for 'default' clause; expected 'present' or 'none'}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial default(invalid), seq
   for(;;){}
 
-  // expected-warning@+1{{OpenACC clause 'default' not yet implemented, clause ignored}}
 #pragma acc serial default(none)
   for(;;){}
 
-  // expected-warning@+2{{OpenACC clause 'default' not yet implemented, clause ignored}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial default(present), seq
   for(;;){}
diff --git a/clang/test/SemaOpenACC/compute-construct-ast.cpp b/clang/test/SemaOpenACC/compute-construct-ast.cpp
index 55c080838a188..e632522f877b5 100644
--- a/clang/test/SemaOpenACC/compute-construct-ast.cpp
+++ b/clang/test/SemaOpenACC/compute-construct-ast.cpp
@@ -12,14 +12,16 @@ void NormalFunc() {
   // CHECK-LABEL: NormalFunc
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}parallel
+  // CHECK-NEXT: default(none)
   // CHECK-NEXT: CompoundStmt
-#pragma acc parallel
+#pragma acc parallel default(none)
   {
 #pragma acc parallel
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}parallel
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}parallel
+  // CHECK-NEXT: default(present)
   // CHECK-NEXT: CompoundStmt
-#pragma acc parallel
+#pragma acc parallel default(present)
     {}
   }
   // FIXME: Add a test once we have clauses for this.
@@ -50,12 +52,12 @@ void NormalFunc() {
 
 template<typename T>
 void TemplFunc() {
-#pragma acc parallel
+#pragma acc parallel default(none)
   {
     typename T::type I;
   }
 
-#pragma acc serial
+#pragma acc serial default(present)
   {
     typename T::type I;
   }
@@ -72,10 +74,12 @@ void TemplFunc() {
   // CHECK-NEXT: FunctionDecl
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}parallel
+  // CHECK-NEXT: default(none)
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: DeclStmt
   // CHECK-NEXT: VarDecl{{.*}} I 'typename T::type'
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}serial
+  // CHECK-NEXT: default(present)
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: DeclStmt
   // CHECK-NEXT: VarDecl{{.*}} I 'typename T::type'
@@ -91,10 +95,12 @@ void TemplFunc() {
   // CHECK-NEXT: CXXRecord
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}parallel
+  // CHECK-NEXT: default(none)
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: DeclStmt
   // CHECK-NEXT: VarDecl{{.*}} I 'typename S::type':'int'
   // CHECK-NEXT: OpenACCComputeConstruct {{.*}}serial
+  // CHECK-NEXT: default(present)
   // CHECK-NEXT: CompoundStmt
   // CHECK-NEXT: DeclStmt
   // CHECK-NEXT: VarDecl{{.*}} I 'typename S::type':'int'
diff --git a/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
new file mode 100644
index 0000000000000..bd80103445028
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
@@ -0,0 +1,74 @@
+// RUN: %clang_cc1 %s -fopenacc -ast-dump | FileCheck %s
+
+// Test this with PCH.
+// RUN: %clang_cc1 %s -fopenacc -emit-pch -o %t %s
+// RUN: %clang_cc1 %s -fopenacc -include-pch %t -ast-dump-all | FileCheck %s
+
+#ifndef PCH_HELPER
+#define PCH_HELPER
+void NormalFunc() {
+  // CHECK: FunctionDecl{{.*}}NormalFunc
+  // CHECK-NEXT: CompoundStmt
+#pragma acc parallel default(none)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: default(none)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc serial default(present)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: default(present)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+}
+
+template<typename T>
+void TemplFunc() {
+  // CHECK: FunctionTemplateDecl{{.*}}TemplFunc
+  // CHECK-NEXT: TemplateTypeParmDecl
+
+  // Match the prototype:
+  // CHECK-NEXT: FunctionDecl{{.*}}TemplFunc
+  // CHECK-NEXT: CompoundStmt
+
+#pragma acc kernels default(none)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: default(none)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc parallel default(present)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: default(present)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // Match the instantiation:
+  // CHECK: FunctionDecl{{.*}}TemplFunc{{.*}}implicit_instantiation
+  // CHECK-NEXT: TemplateArgument type 'int'
+  // CHECK-NEXT: BuiltinType
+  // CHECK-NEXT: CompoundStmt
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: default(none)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: default(present)
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+}
+
+void Instantiate() {
+  TemplFunc<int>();
+}
+#endif
diff --git a/clang/test/SemaOpenACC/compute-construct-default-clause.c b/clang/test/SemaOpenACC/compute-construct-default-clause.c
new file mode 100644
index 0000000000000..b1235fcca1f6a
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-default-clause.c
@@ -0,0 +1,55 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+void SingleOnly() {
+  #pragma acc parallel default(none)
+  while(0);
+
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+2{{OpenACC 'default' clause cannot appear more than once on a 'serial' directive}}
+  // expected-note@+1{{previous clause is here}}
+  #pragma acc serial default(present) seq default(none)
+  while(0);
+
+  // expected-warning@+5{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+4{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+2{{OpenACC 'default' clause cannot appear more than once on a 'kernels' directive}}
+  // expected-note@+1{{previous clause is here}}
+  #pragma acc kernels seq default(present) seq default(none) seq
+  while(0);
+
+  // expected-warning@+6{{OpenACC construct 'parallel loop' not yet implemented}}
+  // expected-warning@+5{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+4{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+2{{OpenACC clause 'default' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'default' not yet implemented}}
+  #pragma acc parallel loop seq default(present) seq default(none) seq
+  while(0);
+
+  // expected-warning@+3{{OpenACC construct 'serial loop' not yet implemented}}
+  // expected-warning@+2{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+1{{expected '('}}
+  #pragma acc serial loop seq default seq default(none) seq
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'kernels loop' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'default' not yet implemented}}
+  #pragma acc kernels loop default(none)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'data' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'default' not yet implemented}}
+  #pragma acc data default(none)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'loop' not yet implemented}}
+  // expected-error@+1{{OpenACC 'default' clause is not valid on 'loop' directive}}
+  #pragma acc loop default(none)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'wait' not yet implemented}}
+  // expected-error@+1{{OpenACC 'default' clause is not valid on 'wait' directive}}
+  #pragma acc wait default(none)
+  while(0);
+}
diff --git a/clang/test/SemaOpenACC/compute-construct-default-clause.cpp b/clang/test/SemaOpenACC/compute-construct-default-clause.cpp
new file mode 100644
index 0000000000000..2c3e711ffd085
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-default-clause.cpp
@@ -0,0 +1,39 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+template<typename T>
+void SingleOnly() {
+  #pragma acc parallel default(none)
+  while(false);
+
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+2{{OpenACC 'default' clause cannot appear more than once on a 'parallel' directive}}
+  // expected-note@+1{{previous clause is here}}
+  #pragma acc parallel default(present) seq default(none)
+  while(false);
+
+  // expected-warning@+5{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+4{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+2{{OpenACC 'default' clause cannot appear more than once on a 'serial' directive}}
+  // expected-note@+1{{previous clause is here}}
+  #pragma acc serial seq default(present) seq default(none) seq
+  while(false);
+
+  // expected-warning@+5{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+4{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+2{{OpenACC 'default' clause cannot appear more than once on a 'kernels' directive}}
+  // expected-note@+1{{previous clause is here}}
+  #pragma acc kernels seq default(present) seq default(none) seq
+  while(false);
+
+  // expected-warning@+3{{OpenACC clause 'seq' not yet implemented}}
+  // expected-warning@+2{{OpenACC clause 'seq' not yet implemented}}
+  // expected-error@+1{{expected '('}}
+  #pragma acc parallel seq default(none) seq default seq
+  while(false);
+}
+
+void Instantiate() {
+  SingleOnly<int>();
+}

From 5ae9ffbd18fd93edbbc8efebe140aeb24cd763c2 Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Wed, 10 Apr 2024 07:16:52 -0700
Subject: [PATCH 014/886] [RISCV] Address review comment from 88062

As pointed out by Fraser, KillSrcReg is always false at this point in
code, and having the inconcistency on whether we check the flag between
the if and else blocks is confusing.
---
 llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp b/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
index 46e79272d60eb..84af6eec40ee6 100644
--- a/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
@@ -210,8 +210,7 @@ void RISCVRegisterInfo::adjustReg(MachineBasicBlock &MBB,
       unsigned Opc = NumOfVReg == 2 ? RISCV::SH1ADD :
         (NumOfVReg == 4 ? RISCV::SH2ADD : RISCV::SH3ADD);
       BuildMI(MBB, II, DL, TII->get(Opc), DestReg)
-          .addReg(ScratchReg, RegState::Kill)
-          .addReg(SrcReg, getKillRegState(KillSrcReg))
+          .addReg(ScratchReg, RegState::Kill).addReg(SrcReg)
           .setMIFlag(Flag);
     } else {
       TII->mulImm(MF, MBB, II, DL, ScratchReg, NumOfVReg, Flag);

From a8f9f85ab0114deb0f6adae2b578bc39c62c19b3 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Wed, 10 Apr 2024 10:00:48 -0500
Subject: [PATCH 015/886] [Libomptarget][NFC] Fix unused variable warnings

Summary:
This patch fixes a few warnings that would show up while building.
---
 openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp | 4 ++--
 openmp/libomptarget/src/interface.cpp                  | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
index a0fdde951b74a..00650b801b420 100644
--- a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
+++ b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
@@ -1885,8 +1885,8 @@ struct AMDGPUDeviceTy : public GenericDeviceTy, AMDGenericDeviceTy {
     // Get the frequency of the steady clock. If the attribute is missing
     // assume running on an older libhsa and default to 0, omp_get_wtime
     // will be inaccurate but otherwise programs can still run.
-    if (auto Err = getDeviceAttrRaw(HSA_AMD_AGENT_INFO_TIMESTAMP_FREQUENCY,
-                                    ClockFrequency))
+    if (getDeviceAttrRaw(HSA_AMD_AGENT_INFO_TIMESTAMP_FREQUENCY,
+                         ClockFrequency) != HSA_STATUS_SUCCESS)
       ClockFrequency = 0;
 
     // Load the grid values dependending on the wavefront.
diff --git a/openmp/libomptarget/src/interface.cpp b/openmp/libomptarget/src/interface.cpp
index b562ba8818c39..557703632c625 100644
--- a/openmp/libomptarget/src/interface.cpp
+++ b/openmp/libomptarget/src/interface.cpp
@@ -495,7 +495,7 @@ EXTERN void __tgt_target_nowait_query(void **AsyncHandle) {
   if (QueryCounter.isAboveThreshold())
     AsyncInfo->SyncType = AsyncInfoTy::SyncTy::BLOCKING;
 
-  if (const int Rc = AsyncInfo->synchronize())
+  if (AsyncInfo->synchronize())
     FATAL_MESSAGE0(1, "Error while querying the async queue for completion.\n");
   // If there are device operations still pending, return immediately without
   // deallocating the handle and increase the current thread query count.

From 2bf48892ab0ce5d53126c7b114070bba18521501 Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu" <yaxun.liu@amd.com>
Date: Wed, 10 Apr 2024 11:16:00 -0400
Subject: [PATCH 016/886] [HIP] document difference with CUDA (#86838)

---
 clang/docs/HIPSupport.rst | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/clang/docs/HIPSupport.rst b/clang/docs/HIPSupport.rst
index 543c82cf90244..5ba84c2f67055 100644
--- a/clang/docs/HIPSupport.rst
+++ b/clang/docs/HIPSupport.rst
@@ -208,6 +208,20 @@ Host Code Compilation
 - These relocatable objects are then linked together.
 - Host code within a TU can call host functions and launch kernels from another TU.
 
+Syntax Difference with CUDA
+===========================
+
+Clang's front end, used for both CUDA and HIP programming models, shares the same parsing and semantic analysis mechanisms. This includes the resolution of overloads concerning device and host functions. While there exists a comprehensive documentation on the syntax differences between Clang and NVCC for CUDA at `Dialect Differences Between Clang and NVCC <https://llvm.org/docs/CompileCudaWithLLVM.html#dialect-differences-between-clang-and-nvcc>`_, it is important to note that these differences also apply to HIP code compilation.
+
+Predefined Macros for Differentiation
+-------------------------------------
+
+To facilitate differentiation between HIP and CUDA code, as well as between device and host compilations within HIP, Clang defines specific macros:
+
+- ``__HIP__`` : This macro is defined only when compiling HIP code. It can be used to conditionally compile code specific to HIP, enabling developers to write portable code that can be compiled for both CUDA and HIP.
+
+- ``__HIP_DEVICE_COMPILE__`` : Defined exclusively during HIP device compilation, this macro allows for conditional compilation of device-specific code. It provides a mechanism to segregate device and host code, ensuring that each can be optimized for their respective execution environments.
+
 Function Pointers Support
 =========================
 

From 6ca5a410d26262f06f954e91200eefe0cbfb7fb8 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Wed, 10 Apr 2024 08:16:43 -0700
Subject: [PATCH 017/886] [SLP]Fix PR87358: broken module, Instruction does not
 dominate all uses.

If the first node is a gather node with extractelement instructions,
still need to put the vector value after all instructions, not after the
very first one.
---
 .../lib/Transforms/Vectorize/SLPVectorizer.cpp | 10 +++++++---
 .../X86/extractlements-gathered-first-node.ll  | 18 ++++++++++++++++++
 2 files changed, 25 insertions(+), 3 deletions(-)
 create mode 100644 llvm/test/Transforms/SLPVectorizer/X86/extractlements-gathered-first-node.ll

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 22ef9b5fb994e..6b758f63a7961 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -10736,9 +10736,13 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
                 [](Value *V) {
                   return !isa<GetElementPtrInst>(V) && isa<Instruction>(V);
                 })) ||
-        all_of(E->Scalars, [](Value *V) {
-          return !isVectorLikeInstWithConstOps(V) && isUsedOutsideBlock(V);
-        }))
+        all_of(E->Scalars,
+               [](Value *V) {
+                 return !isVectorLikeInstWithConstOps(V) &&
+                        isUsedOutsideBlock(V);
+               }) ||
+        (E->State == TreeEntry::NeedToGather && E->Idx == 0 &&
+         all_of(E->Scalars, IsaPred<ExtractElementInst, UndefValue>)))
       Res.second = FindLastInst();
     else
       Res.second = FindFirstInst();
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/extractlements-gathered-first-node.ll b/llvm/test/Transforms/SLPVectorizer/X86/extractlements-gathered-first-node.ll
new file mode 100644
index 0000000000000..57fa83b1ccdd6
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/extractlements-gathered-first-node.ll
@@ -0,0 +1,18 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S --passes=slp-vectorizer -slp-threshold=-99999 -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+
+define void @test() {
+; CHECK-LABEL: define void @test() {
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <4 x i32> zeroinitializer, i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <2 x i32> zeroinitializer, i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i32> <i32 0, i32 undef>, i32 [[TMP1]], i32 1
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp ult i32 [[TMP0]], [[TMP1]]
+; CHECK-NEXT:    ret void
+;
+bb:
+  %0 = extractelement <4 x i32> zeroinitializer, i32 0
+  %1 = extractelement <2 x i32> zeroinitializer, i32 0
+  %icmp = icmp ult i32 %0, %1
+  ret void
+}

From 7f1b9adfc8d86c77ee87a268b3d30e0eda8ed493 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 10 Apr 2024 08:39:56 -0700
Subject: [PATCH 018/886] [RISCV] Add MachineCombiner to fold (sh3add Z, (add
 X, (slli Y, 6))) -> (sh3add (sh3add Y, Z), X). (#87884)

This improves a pattern that occurs in 531.deepsjeng_r. Reducing the
dynamic instruction count by 0.5%.

This may be possible to improve in SelectionDAG, but given the special
cases around shXadd formation, it's not obvious it can be done in a
robust way without adding multiple special cases.

I've used a GEP with 2 indices because that mostly closely resembles the
motivating case. Most of the test cases are the simplest GEP case. One
test has a logical right shift on an index which is closer to the
deepsjeng code. This requires special handling in isel to reverse a
DAGCombiner canonicalization that turns a pair of shifts into (srl (and
X, C1), C2).
---
 .../llvm/CodeGen/MachineCombinerPattern.h     |   2 +
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      | 151 ++++++++++++++++++
 llvm/test/CodeGen/RISCV/rv64zba.ll            |  40 ++---
 3 files changed, 169 insertions(+), 24 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/MachineCombinerPattern.h b/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
index 89eed7463bd78..41b73eaae0298 100644
--- a/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
+++ b/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
@@ -175,6 +175,8 @@ enum class MachineCombinerPattern {
   FMADD_XA,
   FMSUB,
   FNMSUB,
+  SHXADD_ADD_SLLI_OP1,
+  SHXADD_ADD_SLLI_OP2,
 
   // X86 VNNI
   DPWSSD,
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index be63bc936ae8a..6b75efe684d91 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -1775,6 +1775,86 @@ static bool getFPPatterns(MachineInstr &Root,
   return getFPFusedMultiplyPatterns(Root, Patterns, DoRegPressureReduce);
 }
 
+/// Utility routine that checks if \param MO is defined by an
+/// \param CombineOpc instruction in the basic block \param MBB
+static const MachineInstr *canCombine(const MachineBasicBlock &MBB,
+                                      const MachineOperand &MO,
+                                      unsigned CombineOpc) {
+  const MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
+  const MachineInstr *MI = nullptr;
+
+  if (MO.isReg() && MO.getReg().isVirtual())
+    MI = MRI.getUniqueVRegDef(MO.getReg());
+  // And it needs to be in the trace (otherwise, it won't have a depth).
+  if (!MI || MI->getParent() != &MBB || MI->getOpcode() != CombineOpc)
+    return nullptr;
+  // Must only used by the user we combine with.
+  if (!MRI.hasOneNonDBGUse(MI->getOperand(0).getReg()))
+    return nullptr;
+
+  return MI;
+}
+
+/// Utility routine that checks if \param MO is defined by a SLLI in \param
+/// MBB that can be combined by splitting across 2 SHXADD instructions. The
+/// first SHXADD shift amount is given by \param OuterShiftAmt.
+static bool canCombineShiftIntoShXAdd(const MachineBasicBlock &MBB,
+                                      const MachineOperand &MO,
+                                      unsigned OuterShiftAmt) {
+  const MachineInstr *ShiftMI = canCombine(MBB, MO, RISCV::SLLI);
+  if (!ShiftMI)
+    return false;
+
+  unsigned InnerShiftAmt = ShiftMI->getOperand(2).getImm();
+  if (InnerShiftAmt < OuterShiftAmt || (InnerShiftAmt - OuterShiftAmt) > 3)
+    return false;
+
+  return true;
+}
+
+// Returns the shift amount from a SHXADD instruction. Returns 0 if the
+// instruction is not a SHXADD.
+static unsigned getSHXADDShiftAmount(unsigned Opc) {
+  switch (Opc) {
+  default:
+    return 0;
+  case RISCV::SH1ADD:
+    return 1;
+  case RISCV::SH2ADD:
+    return 2;
+  case RISCV::SH3ADD:
+    return 3;
+  }
+}
+
+// Look for opportunities to combine (sh3add Z, (add X, (slli Y, 5))) into
+// (sh3add (sh2add Y, Z), X).
+static bool
+getSHXADDPatterns(const MachineInstr &Root,
+                  SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+  unsigned ShiftAmt = getSHXADDShiftAmount(Root.getOpcode());
+  if (!ShiftAmt)
+    return false;
+
+  const MachineBasicBlock &MBB = *Root.getParent();
+
+  const MachineInstr *AddMI = canCombine(MBB, Root.getOperand(2), RISCV::ADD);
+  if (!AddMI)
+    return false;
+
+  bool Found = false;
+  if (canCombineShiftIntoShXAdd(MBB, AddMI->getOperand(1), ShiftAmt)) {
+    Patterns.push_back(MachineCombinerPattern::SHXADD_ADD_SLLI_OP1);
+    Found = true;
+  }
+  if (canCombineShiftIntoShXAdd(MBB, AddMI->getOperand(2), ShiftAmt)) {
+    Patterns.push_back(MachineCombinerPattern::SHXADD_ADD_SLLI_OP2);
+    Found = true;
+  }
+
+  return Found;
+}
+
 bool RISCVInstrInfo::getMachineCombinerPatterns(
     MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
     bool DoRegPressureReduce) const {
@@ -1782,6 +1862,9 @@ bool RISCVInstrInfo::getMachineCombinerPatterns(
   if (getFPPatterns(Root, Patterns, DoRegPressureReduce))
     return true;
 
+  if (getSHXADDPatterns(Root, Patterns))
+    return true;
+
   return TargetInstrInfo::getMachineCombinerPatterns(Root, Patterns,
                                                      DoRegPressureReduce);
 }
@@ -1864,6 +1947,68 @@ static void combineFPFusedMultiply(MachineInstr &Root, MachineInstr &Prev,
   DelInstrs.push_back(&Root);
 }
 
+// Combine patterns like (sh3add Z, (add X, (slli Y, 5))) to
+// (sh3add (sh2add Y, Z), X) if the shift amount can be split across two
+// shXadd instructions. The outer shXadd keeps its original opcode.
+static void
+genShXAddAddShift(MachineInstr &Root, unsigned AddOpIdx,
+                  SmallVectorImpl<MachineInstr *> &InsInstrs,
+                  SmallVectorImpl<MachineInstr *> &DelInstrs,
+                  DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) {
+  MachineFunction *MF = Root.getMF();
+  MachineRegisterInfo &MRI = MF->getRegInfo();
+  const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
+
+  unsigned OuterShiftAmt = getSHXADDShiftAmount(Root.getOpcode());
+  assert(OuterShiftAmt != 0 && "Unexpected opcode");
+
+  MachineInstr *AddMI = MRI.getUniqueVRegDef(Root.getOperand(2).getReg());
+  MachineInstr *ShiftMI =
+      MRI.getUniqueVRegDef(AddMI->getOperand(AddOpIdx).getReg());
+
+  unsigned InnerShiftAmt = ShiftMI->getOperand(2).getImm();
+  assert(InnerShiftAmt > OuterShiftAmt && "Unexpected shift amount");
+
+  unsigned InnerOpc;
+  switch (InnerShiftAmt - OuterShiftAmt) {
+  default:
+    llvm_unreachable("Unexpected shift amount");
+  case 0:
+    InnerOpc = RISCV::ADD;
+    break;
+  case 1:
+    InnerOpc = RISCV::SH1ADD;
+    break;
+  case 2:
+    InnerOpc = RISCV::SH2ADD;
+    break;
+  case 3:
+    InnerOpc = RISCV::SH3ADD;
+    break;
+  }
+
+  const MachineOperand &X = AddMI->getOperand(3 - AddOpIdx);
+  const MachineOperand &Y = ShiftMI->getOperand(1);
+  const MachineOperand &Z = Root.getOperand(1);
+
+  Register NewVR = MRI.createVirtualRegister(&RISCV::GPRRegClass);
+
+  auto MIB1 = BuildMI(*MF, MIMetadata(Root), TII->get(InnerOpc), NewVR)
+                  .addReg(Y.getReg(), getKillRegState(Y.isKill()))
+                  .addReg(Z.getReg(), getKillRegState(Z.isKill()));
+  auto MIB2 = BuildMI(*MF, MIMetadata(Root), TII->get(Root.getOpcode()),
+                      Root.getOperand(0).getReg())
+                  .addReg(NewVR, RegState::Kill)
+                  .addReg(X.getReg(), getKillRegState(X.isKill()));
+
+  InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
+  InsInstrs.push_back(MIB1);
+  InsInstrs.push_back(MIB2);
+  DelInstrs.push_back(ShiftMI);
+  DelInstrs.push_back(AddMI);
+  DelInstrs.push_back(&Root);
+}
+
 void RISCVInstrInfo::genAlternativeCodeSequence(
     MachineInstr &Root, MachineCombinerPattern Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
@@ -1887,6 +2032,12 @@ void RISCVInstrInfo::genAlternativeCodeSequence(
     combineFPFusedMultiply(Root, Prev, Pattern, InsInstrs, DelInstrs);
     return;
   }
+  case MachineCombinerPattern::SHXADD_ADD_SLLI_OP1:
+    genShXAddAddShift(Root, 1, InsInstrs, DelInstrs, InstrIdxForVirtReg);
+    return;
+  case MachineCombinerPattern::SHXADD_ADD_SLLI_OP2:
+    genShXAddAddShift(Root, 2, InsInstrs, DelInstrs, InstrIdxForVirtReg);
+    return;
   }
 }
 
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index 7e32253c8653f..067addc819f7e 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -1404,9 +1404,8 @@ define i64 @sh6_sh3_add2(i64 noundef %x, i64 noundef %y, i64 noundef %z) {
 ;
 ; RV64ZBA-LABEL: sh6_sh3_add2:
 ; RV64ZBA:       # %bb.0: # %entry
-; RV64ZBA-NEXT:    slli a1, a1, 6
-; RV64ZBA-NEXT:    add a0, a1, a0
-; RV64ZBA-NEXT:    sh3add a0, a2, a0
+; RV64ZBA-NEXT:    sh3add a1, a1, a2
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
 ; RV64ZBA-NEXT:    ret
 entry:
   %shl = shl i64 %z, 3
@@ -2111,9 +2110,8 @@ define i64 @array_index_sh1_sh3(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh1_sh3:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 4
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh3add a0, a2, a0
+; RV64ZBA-NEXT:    sh1add a1, a1, a2
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
 ; RV64ZBA-NEXT:    ld a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [2 x i64], ptr %p, i64 %idx1, i64 %idx2
@@ -2174,9 +2172,8 @@ define i32 @array_index_sh2_sh2(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh2_sh2:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 4
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh2add a0, a2, a0
+; RV64ZBA-NEXT:    sh2add a1, a1, a2
+; RV64ZBA-NEXT:    sh2add a0, a1, a0
 ; RV64ZBA-NEXT:    lw a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [4 x i32], ptr %p, i64 %idx1, i64 %idx2
@@ -2196,9 +2193,8 @@ define i64 @array_index_sh2_sh3(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh2_sh3:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 5
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh3add a0, a2, a0
+; RV64ZBA-NEXT:    sh2add a1, a1, a2
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
 ; RV64ZBA-NEXT:    ld a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [4 x i64], ptr %p, i64 %idx1, i64 %idx2
@@ -2238,9 +2234,8 @@ define i16 @array_index_sh3_sh1(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh3_sh1:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 4
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh1add a0, a2, a0
+; RV64ZBA-NEXT:    sh3add a1, a1, a2
+; RV64ZBA-NEXT:    sh1add a0, a1, a0
 ; RV64ZBA-NEXT:    lh a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [8 x i16], ptr %p, i64 %idx1, i64 %idx2
@@ -2260,9 +2255,8 @@ define i32 @array_index_sh3_sh2(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh3_sh2:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 5
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh2add a0, a2, a0
+; RV64ZBA-NEXT:    sh3add a1, a1, a2
+; RV64ZBA-NEXT:    sh2add a0, a1, a0
 ; RV64ZBA-NEXT:    lw a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [8 x i32], ptr %p, i64 %idx1, i64 %idx2
@@ -2282,9 +2276,8 @@ define i64 @array_index_sh3_sh3(ptr %p, i64 %idx1, i64 %idx2) {
 ;
 ; RV64ZBA-LABEL: array_index_sh3_sh3:
 ; RV64ZBA:       # %bb.0:
-; RV64ZBA-NEXT:    slli a1, a1, 6
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh3add a0, a2, a0
+; RV64ZBA-NEXT:    sh3add a1, a1, a2
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
 ; RV64ZBA-NEXT:    ld a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %a = getelementptr inbounds [8 x i64], ptr %p, i64 %idx1, i64 %idx2
@@ -2308,9 +2301,8 @@ define i64 @array_index_lshr_sh3_sh3(ptr %p, i64 %idx1, i64 %idx2) {
 ; RV64ZBA-LABEL: array_index_lshr_sh3_sh3:
 ; RV64ZBA:       # %bb.0:
 ; RV64ZBA-NEXT:    srli a1, a1, 58
-; RV64ZBA-NEXT:    slli a1, a1, 6
-; RV64ZBA-NEXT:    add a0, a0, a1
-; RV64ZBA-NEXT:    sh3add a0, a2, a0
+; RV64ZBA-NEXT:    sh3add a1, a1, a2
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
 ; RV64ZBA-NEXT:    ld a0, 0(a0)
 ; RV64ZBA-NEXT:    ret
   %shr = lshr i64 %idx1, 58

From f9f4aba547f50e6dcb2d9345b51fe4883bb64d8d Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 9 Apr 2024 12:45:05 -0500
Subject: [PATCH 019/886] [InstCombine] Add tests for non-zero/knownbits of
 `vector_reduce_{s,u}{min,max}`; NFC

---
 .../vector-reduce-min-max-known.ll            | 295 ++++++++++++++++++
 1 file changed, 295 insertions(+)
 create mode 100644 llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll

diff --git a/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
new file mode 100644
index 0000000000000..a02ebcca8090a
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
@@ -0,0 +1,295 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+
+define i1 @vec_reduce_umax_non_zero(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_non_zero(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 0, i8 1, i8 0, i8 0>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 0, i8 1, i8 0, i8 0>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_umax_non_zero_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_non_zero_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nsw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nsw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_umin_non_zero(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_non_zero(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_umin_non_zero_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_non_zero_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 0, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 0, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smax_non_zero0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smax_non_zero0(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smax_non_zero1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smax_non_zero1(
+; CHECK-NEXT:    [[X0:%.*]] = and <4 x i8> [[XX:%.*]], <i8 127, i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[X0]], <i8 1, i8 0, i8 0, i8 0>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <4 x i8> %xx, <i8 127, i8 255, i8 255, i8 255>
+  %x = or <4 x i8> %x0, <i8 1, i8 0, i8 0, i8 0>
+  %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smax_non_zero_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smax_non_zero_fail(
+; CHECK-NEXT:    [[X0:%.*]] = and <4 x i8> [[XX:%.*]], <i8 127, i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[X0]], <i8 1, i8 0, i8 0, i8 0>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <4 x i8> %xx, <i8 127, i8 255, i8 255, i8 255>
+  %x = add nuw <4 x i8> %x0, <i8 1, i8 0, i8 0, i8 0>
+  %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smin_non_zero0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smin_non_zero0(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smin_non_zero1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smin_non_zero1(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 0, i8 0, i8 0, i8 -128>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 0, i8 0, i8 0, i8 128>
+  %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @vec_reduce_smin_non_zero_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smin_non_zero_fail(
+; CHECK-NEXT:    [[X0:%.*]] = or <4 x i8> [[XX:%.*]], <i8 0, i8 0, i8 0, i8 -128>
+; CHECK-NEXT:    [[X:%.*]] = add <4 x i8> [[X0]], <i8 0, i8 0, i8 0, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = or <4 x i8> %xx, <i8 0, i8 0, i8 0, i8 128>
+  %x = add <4 x i8> %x0, <i8 0, i8 0, i8 0, i8 1>
+  %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i8 @vec_reduce_umax_known0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_known0(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umax_known1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_known1(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 -128>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], -128
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 128>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = and i8 %v, 128
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umax_known_fail0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_known_fail0(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 -128>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 128>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umax_known_fail1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umax_known_fail1(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 2, i8 4, i8 8>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 2, i8 4, i8 8>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umin_known0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_known0(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umin_known1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_known1(
+; CHECK-NEXT:    [[X:%.*]] = and <4 x i8> [[XX:%.*]], <i8 127, i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], -128
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = and <4 x i8> %xx, <i8 127, i8 255, i8 255, i8 255>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = and i8 %v, 128
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umin_known_fail0(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_known_fail0(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 0, i8 0, i8 0>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x0 = and <4 x i8> %xx, <i8 127, i8 255, i8 255, i8 255>
+  %x = or <4 x i8> %xx, <i8 1, i8 0, i8 0, i8 0>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_umin_known_fail1(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_umin_known_fail1(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 2, i8 4, i8 8>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 1, i8 2, i8 4, i8 8>
+  %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @vec_reduce_smax_known(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smax_known(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 4, i8 4, i8 4, i8 5>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 4
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 4, i8 4, i8 4, i8 5>
+  %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
+  %r = and i8 %v, 4
+  ret i8 %r
+}
+
+define i8 @vec_reduce_smax_known_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smax_known_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 4, i8 4, i8 8, i8 5>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 4
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 4, i8 4, i8 8, i8 5>
+  %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
+  %r = and i8 %v, 4
+  ret i8 %r
+}
+
+define i8 @vec_reduce_smin_known(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smin_known(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 8, i8 24, i8 56, i8 9>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 8
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 8, i8 24, i8 56, i8 9>
+  %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)
+  %r = and i8 %v, 8
+  ret i8 %r
+}
+
+define i8 @vec_reduce_smin_known_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @vec_reduce_smin_known_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 8, i8 23, i8 56, i8 9>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 8
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <4 x i8> %xx, <i8 8, i8 23, i8 56, i8 9>
+  %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)
+  %r = and i8 %v, 8
+  ret i8 %r
+}

From 77d668451ad2e6370eb595c171779429e9becdf2 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 9 Apr 2024 11:58:38 -0500
Subject: [PATCH 020/886] [ValueTracking] Add support for
 `vector_reduce_{s,u}{min,max}` in `isKnownNonZero`

Previously missing, proofs for all implementations:
https://alive2.llvm.org/ce/z/G8wpmG
---
 llvm/lib/Analysis/ValueTracking.cpp               |  6 ++++++
 .../InstCombine/vector-reduce-min-max-known.ll    | 15 +++------------
 2 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index ca48cfe773815..869a94d81f4df 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2824,6 +2824,12 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
         return isNonZeroAdd(DemandedElts, Depth, Q, BitWidth,
                             II->getArgOperand(0), II->getArgOperand(1),
                             /*NSW=*/true, /* NUW=*/false);
+        // umin/smin/smax/smin of all non-zero elements is always non-zero.
+      case Intrinsic::vector_reduce_umax:
+      case Intrinsic::vector_reduce_umin:
+      case Intrinsic::vector_reduce_smax:
+      case Intrinsic::vector_reduce_smin:
+        return isKnownNonZero(II->getArgOperand(0), Depth, Q);
       case Intrinsic::umax:
       case Intrinsic::uadd_sat:
         return isKnownNonZero(II->getArgOperand(1), DemandedElts, Depth, Q) ||
diff --git a/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
index a02ebcca8090a..29c08b17ef885 100644
--- a/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
+++ b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
@@ -29,10 +29,7 @@ define i1 @vec_reduce_umax_non_zero_fail(<4 x i8> %xx) {
 
 define i1 @vec_reduce_umin_non_zero(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_umin_non_zero(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
@@ -55,10 +52,7 @@ define i1 @vec_reduce_umin_non_zero_fail(<4 x i8> %xx) {
 
 define i1 @vec_reduce_smax_non_zero0(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_smax_non_zero0(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
@@ -98,10 +92,7 @@ define i1 @vec_reduce_smax_non_zero_fail(<4 x i8> %xx) {
 
 define i1 @vec_reduce_smin_non_zero0(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_smin_non_zero0(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)

From 41c52217b003ce9435ae534251b0d0d035495262 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 9 Apr 2024 11:58:48 -0500
Subject: [PATCH 021/886] [ValueTracking] Add support for
 `vector_reduce_{s,u}{min,max}` in `computeKnownBits`

Previously missing. We compute by just applying the reduce function on
the knownbits of each element.

Closes #88169
---
 llvm/lib/Analysis/ValueTracking.cpp           |  8 ++++++++
 .../vector-reduce-min-max-known.ll            | 20 ++++---------------
 2 files changed, 12 insertions(+), 16 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 869a94d81f4df..4120876889dec 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1621,6 +1621,14 @@ static void computeKnownBitsFromOperator(const Operator *I,
         computeKnownBits(I->getOperand(1), Known2, Depth + 1, Q);
         Known = KnownBits::ssub_sat(Known, Known2);
         break;
+        // for min/max reduce, any bit common to each element in the input vec
+        // is set in the output.
+      case Intrinsic::vector_reduce_umax:
+      case Intrinsic::vector_reduce_umin:
+      case Intrinsic::vector_reduce_smax:
+      case Intrinsic::vector_reduce_smin:
+        computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
+        break;
       case Intrinsic::umin:
         computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
         computeKnownBits(I->getOperand(1), Known2, Depth + 1, Q);
diff --git a/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
index 29c08b17ef885..65d0008353262 100644
--- a/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
+++ b/llvm/test/Transforms/InstCombine/vector-reduce-min-max-known.ll
@@ -130,10 +130,7 @@ define i1 @vec_reduce_smin_non_zero_fail(<4 x i8> %xx) {
 
 define i8 @vec_reduce_umax_known0(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_umax_known0(
-; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 1
 ;
   %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.umax(<4 x i8> %x)
@@ -182,10 +179,7 @@ define i8 @vec_reduce_umax_known_fail1(<4 x i8> %xx) {
 
 define i8 @vec_reduce_umin_known0(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_umin_known0(
-; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 1
 ;
   %x = or <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.umin(<4 x i8> %x)
@@ -235,10 +229,7 @@ define i8 @vec_reduce_umin_known_fail1(<4 x i8> %xx) {
 
 define i8 @vec_reduce_smax_known(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_smax_known(
-; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 4, i8 4, i8 4, i8 5>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 4
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 4
 ;
   %x = or <4 x i8> %xx, <i8 4, i8 4, i8 4, i8 5>
   %v = call i8 @llvm.vector.reduce.smax(<4 x i8> %x)
@@ -261,10 +252,7 @@ define i8 @vec_reduce_smax_known_fail(<4 x i8> %xx) {
 
 define i8 @vec_reduce_smin_known(<4 x i8> %xx) {
 ; CHECK-LABEL: @vec_reduce_smin_known(
-; CHECK-NEXT:    [[X:%.*]] = or <4 x i8> [[XX:%.*]], <i8 8, i8 24, i8 56, i8 9>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 8
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 8
 ;
   %x = or <4 x i8> %xx, <i8 8, i8 24, i8 56, i8 9>
   %v = call i8 @llvm.vector.reduce.smin(<4 x i8> %x)

From a02b3c01820090d4208146b51372587251fdce61 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 21:36:35 -0500
Subject: [PATCH 022/886] [ValueTracking] Add tests for overflow detection
 functions is `isKnownNonZero`; NFC

---
 .../test/Transforms/InstCombine/known-bits.ll | 381 ++++++++++++++++++
 1 file changed, 381 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index 769f7661fc8dc..ddd5970ccabdb 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -630,5 +630,386 @@ define i8 @or_ne_bits_must_be_unset2_fail(i8 %x, i8 %y) {
   ret i8 %r
 }
 
+declare void @use.i1(i1)
+declare void @use.i8(i8)
+
+declare void @use.2xi1(<2 x i1>)
+
+define i1 @extract_value_uadd(<2 x i8> %xx, <2 x i8> %yy) {
+; CHECK-LABEL: @extract_value_uadd(
+; CHECK-NEXT:    [[X0:%.*]] = and <2 x i8> [[XX:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[Y0:%.*]] = and <2 x i8> [[YY:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
+; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
+; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <2 x i8> %xx, <i8 63, i8 255>
+  %y0 = and <2 x i8> %yy, <i8 63, i8 255>
+  %x = add nuw <2 x i8> %x0, <i8 1, i8 0>
+  %y = add nuw <2 x i8> %y0, <i8 1, i8 0>
+
+  %add_uov = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow(<2 x i8> %x, <2 x i8> %y)
+  %add = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 0
+  %uov = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 1
+  call void @use.2xi1(<2 x i1> %uov)
+  %add_ele = extractelement <2 x i8> %add, i32 0
+  %r = icmp eq i8 %add_ele, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_uadd2(<2 x i8> %xx, <2 x i8> %yy) {
+; CHECK-LABEL: @extract_value_uadd2(
+; CHECK-NEXT:    [[X0:%.*]] = and <2 x i8> [[XX:%.*]], <i8 -1, i8 63>
+; CHECK-NEXT:    [[Y0:%.*]] = and <2 x i8> [[YY:%.*]], <i8 -1, i8 63>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 0, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 0, i8 1>
+; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
+; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
+; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 1
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <2 x i8> %xx, <i8 255, i8 63>
+  %y0 = and <2 x i8> %yy, <i8 255, i8 63>
+  %x = add nuw <2 x i8> %x0, <i8 0, i8 1>
+  %y = add nuw <2 x i8> %y0, <i8 0, i8 1>
+
+  %add_uov = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow(<2 x i8> %x, <2 x i8> %y)
+  %add = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 0
+  %uov = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 1
+  call void @use.2xi1(<2 x i1> %uov)
+  %add_ele = extractelement <2 x i8> %add, i32 1
+  %r = icmp eq i8 %add_ele, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_uadd_fail(<2 x i8> %xx, <2 x i8> %yy) {
+; CHECK-LABEL: @extract_value_uadd_fail(
+; CHECK-NEXT:    [[X0:%.*]] = and <2 x i8> [[XX:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[Y0:%.*]] = and <2 x i8> [[YY:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
+; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
+; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 1
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <2 x i8> %xx, <i8 63, i8 255>
+  %y0 = and <2 x i8> %yy, <i8 63, i8 255>
+  %x = add nuw <2 x i8> %x0, <i8 1, i8 0>
+  %y = add nuw <2 x i8> %y0, <i8 1, i8 0>
+
+  %add_uov = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow(<2 x i8> %x, <2 x i8> %y)
+  %add = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 0
+  %uov = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 1
+  call void @use.2xi1(<2 x i1> %uov)
+  %add_ele = extractelement <2 x i8> %add, i32 1
+  %r = icmp eq i8 %add_ele, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_uadd_fail2(<2 x i8> %xx, <2 x i8> %yy, i32 %idx) {
+; CHECK-LABEL: @extract_value_uadd_fail2(
+; CHECK-NEXT:    [[X0:%.*]] = and <2 x i8> [[XX:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[Y0:%.*]] = and <2 x i8> [[YY:%.*]], <i8 63, i8 -1>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 1, i8 0>
+; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
+; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
+; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i32 [[IDX:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <2 x i8> %xx, <i8 63, i8 255>
+  %y0 = and <2 x i8> %yy, <i8 63, i8 255>
+  %x = add nuw <2 x i8> %x0, <i8 1, i8 0>
+  %y = add nuw <2 x i8> %y0, <i8 1, i8 0>
+
+  %add_uov = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow(<2 x i8> %x, <2 x i8> %y)
+  %add = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 0
+  %uov = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 1
+  call void @use.2xi1(<2 x i1> %uov)
+  %add_ele = extractelement <2 x i8> %add, i32 %idx
+  %r = icmp eq i8 %add_ele, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_uadd_fail3(<2 x i8> %xx, <2 x i8> %yy) {
+; CHECK-LABEL: @extract_value_uadd_fail3(
+; CHECK-NEXT:    [[X0:%.*]] = and <2 x i8> [[XX:%.*]], <i8 127, i8 127>
+; CHECK-NEXT:    [[Y0:%.*]] = and <2 x i8> [[YY:%.*]], <i8 127, i8 127>
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 1, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 1, i8 1>
+; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
+; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
+; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and <2 x i8> %xx, <i8 127, i8 127>
+  %y0 = and <2 x i8> %yy, <i8 127, i8 127>
+  %x = add nuw <2 x i8> %x0, <i8 1, i8 1>
+  %y = add nuw <2 x i8> %y0, <i8 1, i8 1>
+
+  %add_uov = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow(<2 x i8> %x, <2 x i8> %y)
+  %add = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 0
+  %uov = extractvalue { <2 x i8>, <2 x i1> } %add_uov, 1
+  call void @use.2xi1(<2 x i1> %uov)
+  %add_ele = extractelement <2 x i8> %add, i32 0
+  %r = icmp eq i8 %add_ele, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_sadd(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_sadd(
+; CHECK-NEXT:    [[X:%.*]] = add nuw i8 [[XX:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[X_LEMMA:%.*]] = icmp sgt i8 [[X]], -1
+; CHECK-NEXT:    [[Y_LEMMA:%.*]] = icmp sgt i8 [[Y]], -1
+; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LEMMA]])
+; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LEMMA]])
+; CHECK-NEXT:    [[ADD_SOV:%.*]] = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw i8 %xx, 1
+  %y = add nuw i8 %yy, 1
+  %x_lemma = icmp ult i8 %x, 128
+  %y_lemma = icmp ult i8 %y, 128
+  call void @llvm.assume(i1 %x_lemma)
+  call void @llvm.assume(i1 %y_lemma)
+
+  %add_sov = call { i8, i1 } @llvm.sadd.with.overflow(i8 %x, i8 %y)
+  %add = extractvalue { i8, i1 } %add_sov, 0
+  %sov = extractvalue { i8, i1 } %add_sov, 1
+  call void @use.i1(i1 %sov)
+  %r = icmp eq i8 %add, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_sadd_fail(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_sadd_fail(
+; CHECK-NEXT:    [[X:%.*]] = add i8 [[XX:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[ADD_SOV:%.*]] = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add i8 %xx, 1
+  %y = add i8 %yy, 1
+
+  %add_sov = call { i8, i1 } @llvm.sadd.with.overflow(i8 %x, i8 %y)
+  %add = extractvalue { i8, i1 } %add_sov, 0
+  %sov = extractvalue { i8, i1 } %add_sov, 1
+  call void @use.i1(i1 %sov)
+  %r = icmp eq i8 %add, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_usub(i8 %x, i8 %zz) {
+; CHECK-LABEL: @extract_value_usub(
+; CHECK-NEXT:    [[Z:%.*]] = add nuw i8 [[ZZ:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[Z]], [[X:%.*]]
+; CHECK-NEXT:    [[SUB_UOV:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[SUB:%.*]] = extractvalue { i8, i1 } [[SUB_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[SUB_UOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %z = add nuw i8 %zz, 1
+  %y = add i8 %x, %z
+
+  %sub_uov = call { i8, i1 } @llvm.usub.with.overflow(i8 %x, i8 %y)
+  %sub = extractvalue { i8, i1 } %sub_uov, 0
+  %uov = extractvalue { i8, i1 } %sub_uov, 1
+  call void @use.i1(i1 %uov)
+  call void @use.i8(i8 %sub)
+  %r = icmp eq i8 %sub, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_usub_fail(i8 %x, i8 %z) {
+; CHECK-LABEL: @extract_value_usub_fail(
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[X:%.*]], [[Z:%.*]]
+; CHECK-NEXT:    [[SUB_UOV:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[SUB:%.*]] = extractvalue { i8, i1 } [[SUB_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[SUB_UOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %y = add i8 %x, %z
+  %sub_uov = call { i8, i1 } @llvm.usub.with.overflow(i8 %x, i8 %y)
+  %sub = extractvalue { i8, i1 } %sub_uov, 0
+  %uov = extractvalue { i8, i1 } %sub_uov, 1
+  call void @use.i1(i1 %uov)
+  call void @use.i8(i8 %sub)
+  %r = icmp eq i8 %sub, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_ssub(i8 %x, i8 %zz) {
+; CHECK-LABEL: @extract_value_ssub(
+; CHECK-NEXT:    [[Z:%.*]] = add nuw i8 [[ZZ:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[Z]], [[X:%.*]]
+; CHECK-NEXT:    [[SUB_SOV:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[Y]], i8 [[X]])
+; CHECK-NEXT:    [[SUB:%.*]] = extractvalue { i8, i1 } [[SUB_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[SUB_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %z = add nuw i8 %zz, 1
+  %y = add i8 %x, %z
+
+  %sub_sov = call { i8, i1 } @llvm.ssub.with.overflow(i8 %y, i8 %x)
+  %sub = extractvalue { i8, i1 } %sub_sov, 0
+  %sov = extractvalue { i8, i1 } %sub_sov, 1
+  call void @use.i1(i1 %sov)
+  call void @use.i8(i8 %sub)
+  %r = icmp eq i8 %sub, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_ssub_fail(i8 %x) {
+; CHECK-LABEL: @extract_value_ssub_fail(
+; CHECK-NEXT:    [[SUB_SOV:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 10, i8 [[X:%.*]])
+; CHECK-NEXT:    [[SUB:%.*]] = extractvalue { i8, i1 } [[SUB_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[SUB_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %sub_sov = call { i8, i1 } @llvm.ssub.with.overflow(i8 10, i8 %x)
+  %sub = extractvalue { i8, i1 } %sub_sov, 0
+  %sov = extractvalue { i8, i1 } %sub_sov, 1
+  call void @use.i1(i1 %sov)
+  call void @use.i8(i8 %sub)
+  %r = icmp eq i8 %sub, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_umul(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_umul(
+; CHECK-NEXT:    [[X:%.*]] = or i8 [[XX:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[MUL_UOV:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[MUL:%.*]] = extractvalue { i8, i1 } [[MUL_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[MUL_UOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = or i8 %xx, 1
+  %y = add nuw i8 %yy, 1
+
+  %mul_uov = call { i8, i1 } @llvm.umul.with.overflow(i8 %x, i8 %y)
+  %mul = extractvalue { i8, i1 } %mul_uov, 0
+  %uov = extractvalue { i8, i1 } %mul_uov, 1
+  call void @use.i1(i1 %uov)
+  call void @use.i8(i8 %mul)
+  %r = icmp eq i8 %mul, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_umul_fail(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_umul_fail(
+; CHECK-NEXT:    [[X:%.*]] = or i8 [[XX:%.*]], 2
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[MUL_UOV:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[X]], i8 [[Y]])
+; CHECK-NEXT:    [[MUL:%.*]] = extractvalue { i8, i1 } [[MUL_UOV]], 0
+; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[MUL_UOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = or i8 %xx, 2
+  %y = add nuw i8 %yy, 1
+
+  %mul_uov = call { i8, i1 } @llvm.umul.with.overflow(i8 %x, i8 %y)
+  %mul = extractvalue { i8, i1 } %mul_uov, 0
+  %uov = extractvalue { i8, i1 } %mul_uov, 1
+  call void @use.i1(i1 %uov)
+  call void @use.i8(i8 %mul)
+  %r = icmp eq i8 %mul, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_smul(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_smul(
+; CHECK-NEXT:    [[X:%.*]] = or i8 [[XX:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[MUL_SOV:%.*]] = call { i8, i1 } @llvm.smul.with.overflow.i8(i8 [[Y]], i8 [[X]])
+; CHECK-NEXT:    [[MUL:%.*]] = extractvalue { i8, i1 } [[MUL_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[MUL_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = or i8 %xx, 1
+  %y = add nuw i8 %yy, 1
+
+  %mul_sov = call { i8, i1 } @llvm.smul.with.overflow(i8 %y, i8 %x)
+  %mul = extractvalue { i8, i1 } %mul_sov, 0
+  %sov = extractvalue { i8, i1 } %mul_sov, 1
+  call void @use.i1(i1 %sov)
+  call void @use.i8(i8 %mul)
+  %r = icmp eq i8 %mul, 0
+  ret i1 %r
+}
+
+define i1 @extract_value_smul_fail(i8 %xx, i8 %yy) {
+; CHECK-LABEL: @extract_value_smul_fail(
+; CHECK-NEXT:    [[X:%.*]] = or i8 [[XX:%.*]], 1
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[MUL_SOV:%.*]] = call { i8, i1 } @llvm.smul.with.overflow.i8(i8 [[Y]], i8 [[X]])
+; CHECK-NEXT:    [[MUL:%.*]] = extractvalue { i8, i1 } [[MUL_SOV]], 0
+; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[MUL_SOV]], 1
+; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
+; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = or i8 %xx, 1
+  %y = add i8 %yy, 1
+
+  %mul_sov = call { i8, i1 } @llvm.smul.with.overflow(i8 %y, i8 %x)
+  %mul = extractvalue { i8, i1 } %mul_sov, 0
+  %sov = extractvalue { i8, i1 } %mul_sov, 1
+  call void @use.i1(i1 %sov)
+  call void @use.i8(i8 %mul)
+  %r = icmp eq i8 %mul, 0
+  ret i1 %r
+}
+
 declare void @use(i1)
 declare void @sink(i8)

From f0a487d7e2085e21f3691393070f54110d889fb6 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 15:33:18 -0500
Subject: [PATCH 023/886] [ValueTracking] Split `isNonZero(mul)` logic to a
 helper; NFC

---
 llvm/lib/Analysis/ValueTracking.cpp | 57 ++++++++++++++++-------------
 1 file changed, 31 insertions(+), 26 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 4120876889dec..2c9ea8aa38512 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2471,6 +2471,34 @@ static bool isNonZeroSub(const APInt &DemandedElts, unsigned Depth,
   return ::isKnownNonEqual(X, Y, Depth, Q);
 }
 
+static bool isNonZeroMul(const APInt &DemandedElts, unsigned Depth,
+                         const SimplifyQuery &Q, unsigned BitWidth, Value *X,
+                         Value *Y, bool NSW, bool NUW) {
+  // If X and Y are non-zero then so is X * Y as long as the multiplication
+  // does not overflow.
+  if (NSW || NUW)
+    return isKnownNonZero(X, DemandedElts, Depth, Q) &&
+           isKnownNonZero(Y, DemandedElts, Depth, Q);
+
+  // If either X or Y is odd, then if the other is non-zero the result can't
+  // be zero.
+  KnownBits XKnown = computeKnownBits(X, DemandedElts, Depth, Q);
+  if (XKnown.One[0])
+    return isKnownNonZero(Y, DemandedElts, Depth, Q);
+
+  KnownBits YKnown = computeKnownBits(Y, DemandedElts, Depth, Q);
+  if (YKnown.One[0])
+    return XKnown.isNonZero() || isKnownNonZero(X, DemandedElts, Depth, Q);
+
+  // If there exists any subset of X (sX) and subset of Y (sY) s.t sX * sY is
+  // non-zero, then X * Y is non-zero. We can find sX and sY by just taking
+  // the lowest known One of X and Y. If they are non-zero, the result
+  // must be non-zero. We can check if LSB(X) * LSB(Y) != 0 by doing
+  // X.CountLeadingZeros + Y.CountLeadingZeros < BitWidth.
+  return (XKnown.countMaxTrailingZeros() + YKnown.countMaxTrailingZeros()) <
+         BitWidth;
+}
+
 static bool isNonZeroShift(const Operator *I, const APInt &DemandedElts,
                            unsigned Depth, const SimplifyQuery &Q,
                            const KnownBits &KnownVal) {
@@ -2666,33 +2694,10 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
                         Q.IIQ.hasNoUnsignedWrap(BO));
   }
   case Instruction::Mul: {
-    // If X and Y are non-zero then so is X * Y as long as the multiplication
-    // does not overflow.
     const OverflowingBinaryOperator *BO = cast<OverflowingBinaryOperator>(I);
-    if (Q.IIQ.hasNoSignedWrap(BO) || Q.IIQ.hasNoUnsignedWrap(BO))
-      return isKnownNonZero(I->getOperand(0), DemandedElts, Depth, Q) &&
-             isKnownNonZero(I->getOperand(1), DemandedElts, Depth, Q);
-
-    // If either X or Y is odd, then if the other is non-zero the result can't
-    // be zero.
-    KnownBits XKnown =
-        computeKnownBits(I->getOperand(0), DemandedElts, Depth, Q);
-    if (XKnown.One[0])
-      return isKnownNonZero(I->getOperand(1), DemandedElts, Depth, Q);
-
-    KnownBits YKnown =
-        computeKnownBits(I->getOperand(1), DemandedElts, Depth, Q);
-    if (YKnown.One[0])
-      return XKnown.isNonZero() ||
-             isKnownNonZero(I->getOperand(0), DemandedElts, Depth, Q);
-
-    // If there exists any subset of X (sX) and subset of Y (sY) s.t sX * sY is
-    // non-zero, then X * Y is non-zero. We can find sX and sY by just taking
-    // the lowest known One of X and Y. If they are non-zero, the result
-    // must be non-zero. We can check if LSB(X) * LSB(Y) != 0 by doing
-    // X.CountLeadingZeros + Y.CountLeadingZeros < BitWidth.
-    return (XKnown.countMaxTrailingZeros() + YKnown.countMaxTrailingZeros()) <
-           BitWidth;
+    return isNonZeroMul(DemandedElts, Depth, Q, BitWidth, I->getOperand(0),
+                        I->getOperand(1), Q.IIQ.hasNoSignedWrap(BO),
+                        Q.IIQ.hasNoUnsignedWrap(BO));
   }
   case Instruction::Select: {
     // (C ? X : Y) != 0 if X != 0 and Y != 0.

From 37ca6fa1e26e86c85c544023b18695be420e80dd Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 15:35:17 -0500
Subject: [PATCH 024/886] [ValueTracking] Add support for overflow detection
 functions is `isKnownNonZero`

Adds support for: `{s,u}{add,sub,mul}.with.overflow`

The logic is identical to the the non-overflow binops, we where just
missing the cases.

Closes #87701
---
 llvm/lib/Analysis/ValueTracking.cpp           | 23 ++++++++++++++++
 .../test/Transforms/InstCombine/known-bits.ll | 26 +++++--------------
 2 files changed, 30 insertions(+), 19 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 2c9ea8aa38512..b32dc493ace91 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2798,6 +2798,29 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     // handled in isKnownNonZero.
     return false;
   }
+  case Instruction::ExtractValue: {
+    const WithOverflowInst *WO;
+    if (match(I, m_ExtractValue<0>(m_WithOverflowInst(WO)))) {
+      switch (WO->getBinaryOp()) {
+      default:
+        break;
+      case Instruction::Add:
+        return isNonZeroAdd(DemandedElts, Depth, Q, BitWidth,
+                            WO->getArgOperand(0), WO->getArgOperand(1),
+                            /*NSW=*/false,
+                            /*NUW=*/false);
+      case Instruction::Sub:
+        return isNonZeroSub(DemandedElts, Depth, Q, BitWidth,
+                            WO->getArgOperand(0), WO->getArgOperand(1));
+      case Instruction::Mul:
+        return isNonZeroMul(DemandedElts, Depth, Q, BitWidth,
+                            WO->getArgOperand(0), WO->getArgOperand(1),
+                            /*NSW=*/false, /*NUW=*/false);
+        break;
+      }
+    }
+    break;
+  }
   case Instruction::Call:
   case Instruction::Invoke: {
     const auto *Call = cast<CallBase>(I);
diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index ddd5970ccabdb..e27e4a3eddfbb 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -642,12 +642,9 @@ define i1 @extract_value_uadd(<2 x i8> %xx, <2 x i8> %yy) {
 ; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 1, i8 0>
 ; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 1, i8 0>
 ; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
-; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
 ; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
 ; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
-; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 0
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x0 = and <2 x i8> %xx, <i8 63, i8 255>
   %y0 = and <2 x i8> %yy, <i8 63, i8 255>
@@ -670,12 +667,9 @@ define i1 @extract_value_uadd2(<2 x i8> %xx, <2 x i8> %yy) {
 ; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[X0]], <i8 0, i8 1>
 ; CHECK-NEXT:    [[Y:%.*]] = add nuw <2 x i8> [[Y0]], <i8 0, i8 1>
 ; CHECK-NEXT:    [[ADD_UOV:%.*]] = call { <2 x i8>, <2 x i1> } @llvm.uadd.with.overflow.v2i8(<2 x i8> [[X]], <2 x i8> [[Y]])
-; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 0
 ; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { <2 x i8>, <2 x i1> } [[ADD_UOV]], 1
 ; CHECK-NEXT:    call void @use.2xi1(<2 x i1> [[UOV]])
-; CHECK-NEXT:    [[ADD_ELE:%.*]] = extractelement <2 x i8> [[ADD]], i64 1
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD_ELE]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x0 = and <2 x i8> %xx, <i8 255, i8 63>
   %y0 = and <2 x i8> %yy, <i8 255, i8 63>
@@ -784,11 +778,9 @@ define i1 @extract_value_sadd(i8 %xx, i8 %yy) {
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LEMMA]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LEMMA]])
 ; CHECK-NEXT:    [[ADD_SOV:%.*]] = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 [[X]], i8 [[Y]])
-; CHECK-NEXT:    [[ADD:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 0
 ; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[ADD_SOV]], 1
 ; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[ADD]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = add nuw i8 %xx, 1
   %y = add nuw i8 %yy, 1
@@ -836,8 +828,7 @@ define i1 @extract_value_usub(i8 %x, i8 %zz) {
 ; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[SUB_UOV]], 1
 ; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
 ; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %z = add nuw i8 %zz, 1
   %y = add i8 %x, %z
@@ -881,8 +872,7 @@ define i1 @extract_value_ssub(i8 %x, i8 %zz) {
 ; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[SUB_SOV]], 1
 ; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
 ; CHECK-NEXT:    call void @use.i8(i8 [[SUB]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[SUB]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %z = add nuw i8 %zz, 1
   %y = add i8 %x, %z
@@ -924,8 +914,7 @@ define i1 @extract_value_umul(i8 %xx, i8 %yy) {
 ; CHECK-NEXT:    [[UOV:%.*]] = extractvalue { i8, i1 } [[MUL_UOV]], 1
 ; CHECK-NEXT:    call void @use.i1(i1 [[UOV]])
 ; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = or i8 %xx, 1
   %y = add nuw i8 %yy, 1
@@ -972,8 +961,7 @@ define i1 @extract_value_smul(i8 %xx, i8 %yy) {
 ; CHECK-NEXT:    [[SOV:%.*]] = extractvalue { i8, i1 } [[MUL_SOV]], 1
 ; CHECK-NEXT:    call void @use.i1(i1 [[SOV]])
 ; CHECK-NEXT:    call void @use.i8(i8 [[MUL]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[MUL]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = or i8 %xx, 1
   %y = add nuw i8 %yy, 1

From 2ff82c2c6490a1478e4311f60f1ce80af0957403 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 9 Apr 2024 12:36:03 -0500
Subject: [PATCH 025/886] [ValueTracking] Add tests for improving
 `isKnownNonZero` of `smax`; NFC

---
 .../Transforms/InstSimplify/known-non-zero.ll     | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index b647f11af4461..6ebc4e0f31a9c 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -166,3 +166,18 @@ A:
 B:
   ret i1 0
 }
+
+define i1 @smax_non_zero(i8 %xx, i8 %y) {
+; CHECK-LABEL: @smax_non_zero(
+; CHECK-NEXT:    [[X0:%.*]] = and i8 [[XX:%.*]], 63
+; CHECK-NEXT:    [[X:%.*]] = add i8 [[X0]], 1
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.smax.i8(i8 [[X]], i8 [[Y:%.*]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x0 = and i8 %xx, 63
+  %x = add i8 %x0, 1
+  %v = call i8 @llvm.smax.i8(i8 %x, i8 %y)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}

From f1ee458ddb45c9887b3df583ce9a4ba12aae8b3b Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 9 Apr 2024 11:58:03 -0500
Subject: [PATCH 026/886] [ValueTracking] improve `isKnownNonZero` precision
 for `smax`

Instead of relying on known-bits for strictly positive, use the
`isKnownPositive` API. This will use `isKnownNonZero` which is more
accurate.

Closes #88170
---
 llvm/lib/Analysis/ValueTracking.cpp           | 41 ++++++++++++++-----
 .../Transforms/InstSimplify/known-non-zero.ll |  6 +--
 2 files changed, 32 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index b32dc493ace91..5ef1969893b42 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2870,23 +2870,44 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       case Intrinsic::uadd_sat:
         return isKnownNonZero(II->getArgOperand(1), DemandedElts, Depth, Q) ||
                isKnownNonZero(II->getArgOperand(0), DemandedElts, Depth, Q);
-      case Intrinsic::smin:
       case Intrinsic::smax: {
-        auto KnownOpImpliesNonZero = [&](const KnownBits &K) {
-          return II->getIntrinsicID() == Intrinsic::smin
-                     ? K.isNegative()
-                     : K.isStrictlyPositive();
+        // If either arg is strictly positive the result is non-zero. Otherwise
+        // the result is non-zero if both ops are non-zero.
+        auto IsNonZero = [&](Value *Op, std::optional<bool> &OpNonZero,
+                             const KnownBits &OpKnown) {
+          if (!OpNonZero.has_value())
+            OpNonZero = OpKnown.isNonZero() ||
+                        isKnownNonZero(Op, DemandedElts, Depth, Q);
+          return *OpNonZero;
         };
-        KnownBits XKnown =
+        // Avoid re-computing isKnownNonZero.
+        std::optional<bool> Op0NonZero, Op1NonZero;
+        KnownBits Op1Known =
+            computeKnownBits(II->getArgOperand(1), DemandedElts, Depth, Q);
+        if (Op1Known.isNonNegative() &&
+            IsNonZero(II->getArgOperand(1), Op1NonZero, Op1Known))
+          return true;
+        KnownBits Op0Known =
             computeKnownBits(II->getArgOperand(0), DemandedElts, Depth, Q);
-        if (KnownOpImpliesNonZero(XKnown))
+        if (Op0Known.isNonNegative() &&
+            IsNonZero(II->getArgOperand(0), Op0NonZero, Op0Known))
           return true;
-        KnownBits YKnown =
+        return IsNonZero(II->getArgOperand(1), Op1NonZero, Op1Known) &&
+               IsNonZero(II->getArgOperand(0), Op0NonZero, Op0Known);
+      }
+      case Intrinsic::smin: {
+        // If either arg is negative the result is non-zero. Otherwise
+        // the result is non-zero if both ops are non-zero.
+        KnownBits Op1Known =
             computeKnownBits(II->getArgOperand(1), DemandedElts, Depth, Q);
-        if (KnownOpImpliesNonZero(YKnown))
+        if (Op1Known.isNegative())
+          return true;
+        KnownBits Op0Known =
+            computeKnownBits(II->getArgOperand(0), DemandedElts, Depth, Q);
+        if (Op0Known.isNegative())
           return true;
 
-        if (XKnown.isNonZero() && YKnown.isNonZero())
+        if (Op1Known.isNonZero() && Op0Known.isNonZero())
           return true;
       }
         [[fallthrough]];
diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index 6ebc4e0f31a9c..51f80f62c2f34 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -169,11 +169,7 @@ B:
 
 define i1 @smax_non_zero(i8 %xx, i8 %y) {
 ; CHECK-LABEL: @smax_non_zero(
-; CHECK-NEXT:    [[X0:%.*]] = and i8 [[XX:%.*]], 63
-; CHECK-NEXT:    [[X:%.*]] = add i8 [[X0]], 1
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.smax.i8(i8 [[X]], i8 [[Y:%.*]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x0 = and i8 %xx, 63
   %x = add i8 %x0, 1

From 7d60232b38b66138dae1b31027d73ee5b9df5c58 Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek <Krzysztof.Parzyszek@amd.com>
Date: Wed, 10 Apr 2024 10:41:20 -0500
Subject: [PATCH 027/886] [flang][Frontend] Implement printing defined macros
 via -dM (#87627)

This should work the same way as in clang.
---
 clang/docs/tools/clang-formatted-files.txt    |  4 +-
 clang/include/clang/Driver/Options.td         |  2 +-
 clang/lib/Driver/ToolChains/Flang.cpp         |  5 +-
 .../flang/Frontend/PreprocessorOptions.h      |  3 +
 flang/include/flang/Parser/parsing.h          |  3 +
 .../flang}/Parser/preprocessor.h              | 15 +++-
 .../flang}/Parser/token-sequence.h            |  4 +-
 flang/lib/Frontend/CompilerInvocation.cpp     |  1 +
 flang/lib/Frontend/FrontendActions.cpp        |  4 +-
 flang/lib/Parser/parsing.cpp                  | 17 +++--
 flang/lib/Parser/preprocessor.cpp             | 68 +++++++++++++++++--
 flang/lib/Parser/prescan.cpp                  |  4 +-
 flang/lib/Parser/prescan.h                    |  2 +-
 flang/lib/Parser/token-sequence.cpp           |  3 +-
 flang/test/Driver/driver-help-hidden.f90      |  1 +
 flang/test/Driver/driver-help.f90             |  2 +
 flang/test/Preprocessing/show-macros1.F90     | 14 ++++
 flang/test/Preprocessing/show-macros2.F90     |  6 ++
 flang/test/Preprocessing/show-macros3.F90     |  9 +++
 19 files changed, 139 insertions(+), 28 deletions(-)
 rename flang/{lib => include/flang}/Parser/preprocessor.h (88%)
 rename flang/{lib => include/flang}/Parser/token-sequence.h (97%)
 create mode 100644 flang/test/Preprocessing/show-macros1.F90
 create mode 100644 flang/test/Preprocessing/show-macros2.F90
 create mode 100644 flang/test/Preprocessing/show-macros3.F90

diff --git a/clang/docs/tools/clang-formatted-files.txt b/clang/docs/tools/clang-formatted-files.txt
index 70687c23b15e6..8fd4fed25a32a 100644
--- a/clang/docs/tools/clang-formatted-files.txt
+++ b/clang/docs/tools/clang-formatted-files.txt
@@ -2147,8 +2147,10 @@ flang/include/flang/Parser/message.h
 flang/include/flang/Parser/parse-state.h
 flang/include/flang/Parser/parse-tree-visitor.h
 flang/include/flang/Parser/parsing.h
+flang/include/flang/Parser/preprocessor.h
 flang/include/flang/Parser/provenance.h
 flang/include/flang/Parser/source.h
+flang/include/flang/Parser/token-sequence.h
 flang/include/flang/Parser/tools.h
 flang/include/flang/Parser/unparse.h
 flang/include/flang/Parser/user-state.h
@@ -2319,7 +2321,6 @@ flang/lib/Parser/openmp-parsers.cpp
 flang/lib/Parser/parse-tree.cpp
 flang/lib/Parser/parsing.cpp
 flang/lib/Parser/preprocessor.cpp
-flang/lib/Parser/preprocessor.h
 flang/lib/Parser/prescan.cpp
 flang/lib/Parser/prescan.h
 flang/lib/Parser/program-parsers.cpp
@@ -2328,7 +2329,6 @@ flang/lib/Parser/source.cpp
 flang/lib/Parser/stmt-parser.h
 flang/lib/Parser/token-parsers.h
 flang/lib/Parser/token-sequence.cpp
-flang/lib/Parser/token-sequence.h
 flang/lib/Parser/tools.cpp
 flang/lib/Parser/type-parser-implementation.h
 flang/lib/Parser/type-parsers.h
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index f745e573eb268..0a74e6c75f95b 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -1449,7 +1449,7 @@ def dD : Flag<["-"], "dD">, Group<d_Group>, Visibility<[ClangOption, CC1Option]>
 def dI : Flag<["-"], "dI">, Group<d_Group>, Visibility<[ClangOption, CC1Option]>,
   HelpText<"Print include directives in -E mode in addition to normal output">,
   MarshallingInfoFlag<PreprocessorOutputOpts<"ShowIncludeDirectives">>;
-def dM : Flag<["-"], "dM">, Group<d_Group>, Visibility<[ClangOption, CC1Option]>,
+def dM : Flag<["-"], "dM">, Group<d_Group>, Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
   HelpText<"Print macro definitions in -E mode instead of normal output">;
 def dead__strip : Flag<["-"], "dead_strip">;
 def dependency_file : Separate<["-"], "dependency-file">,
diff --git a/clang/lib/Driver/ToolChains/Flang.cpp b/clang/lib/Driver/ToolChains/Flang.cpp
index 2c83f70eb7887..9699443603d36 100644
--- a/clang/lib/Driver/ToolChains/Flang.cpp
+++ b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -679,7 +679,10 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
   CmdArgs.push_back(Args.MakeArgString(TripleStr));
 
   if (isa<PreprocessJobAction>(JA)) {
-      CmdArgs.push_back("-E");
+    CmdArgs.push_back("-E");
+    if (Args.getLastArg(options::OPT_dM)) {
+      CmdArgs.push_back("-dM");
+    }
   } else if (isa<CompileJobAction>(JA) || isa<BackendJobAction>(JA)) {
     if (JA.getType() == types::TY_Nothing) {
       CmdArgs.push_back("-fsyntax-only");
diff --git a/flang/include/flang/Frontend/PreprocessorOptions.h b/flang/include/flang/Frontend/PreprocessorOptions.h
index b2e9ac0e963b7..13a91ee9a184f 100644
--- a/flang/include/flang/Frontend/PreprocessorOptions.h
+++ b/flang/include/flang/Frontend/PreprocessorOptions.h
@@ -56,6 +56,9 @@ struct PreprocessorOptions {
   // -fno-reformat: Emit cooked character stream as -E output
   bool noReformat{false};
 
+  // -dM: Show macro definitions with -dM -E
+  bool showMacros{false};
+
   void addMacroDef(llvm::StringRef name) {
     macros.emplace_back(std::string(name), false);
   }
diff --git a/flang/include/flang/Parser/parsing.h b/flang/include/flang/Parser/parsing.h
index e80d8f724ac8f..4d329c189cb80 100644
--- a/flang/include/flang/Parser/parsing.h
+++ b/flang/include/flang/Parser/parsing.h
@@ -15,6 +15,7 @@
 #include "parse-tree.h"
 #include "provenance.h"
 #include "flang/Common/Fortran-features.h"
+#include "flang/Parser/preprocessor.h"
 #include "llvm/Support/raw_ostream.h"
 #include <optional>
 #include <string>
@@ -59,6 +60,7 @@ class Parsing {
   const SourceFile *Prescan(const std::string &path, Options);
   void EmitPreprocessedSource(
       llvm::raw_ostream &, bool lineDirectives = true) const;
+  void EmitPreprocessorMacros(llvm::raw_ostream &) const;
   void DumpCookedChars(llvm::raw_ostream &) const;
   void DumpProvenance(llvm::raw_ostream &) const;
   void DumpParsingLog(llvm::raw_ostream &) const;
@@ -83,6 +85,7 @@ class Parsing {
   const char *finalRestingPlace_{nullptr};
   std::optional<Program> parseTree_;
   ParsingLog log_;
+  Preprocessor preprocessor_{allCooked_.allSources()};
 };
 } // namespace Fortran::parser
 #endif // FORTRAN_PARSER_PARSING_H_
diff --git a/flang/lib/Parser/preprocessor.h b/flang/include/flang/Parser/preprocessor.h
similarity index 88%
rename from flang/lib/Parser/preprocessor.h
rename to flang/include/flang/Parser/preprocessor.h
index b61f1577727be..630d5273d427c 100644
--- a/flang/lib/Parser/preprocessor.h
+++ b/flang/include/flang/Parser/preprocessor.h
@@ -15,9 +15,10 @@
 // performed, so that special compiler command options &/or source file name
 // extensions for preprocessing will not be necessary.
 
-#include "token-sequence.h"
 #include "flang/Parser/char-block.h"
 #include "flang/Parser/provenance.h"
+#include "flang/Parser/token-sequence.h"
+#include "llvm/Support/raw_ostream.h"
 #include <cstddef>
 #include <list>
 #include <stack>
@@ -39,7 +40,7 @@ class Definition {
   Definition(const std::string &predefined, AllSources &);
 
   bool isFunctionLike() const { return isFunctionLike_; }
-  std::size_t argumentCount() const { return argumentCount_; }
+  std::size_t argumentCount() const { return argNames_.size(); }
   bool isVariadic() const { return isVariadic_; }
   bool isDisabled() const { return isDisabled_; }
   bool isPredefined() const { return isPredefined_; }
@@ -49,15 +50,21 @@ class Definition {
 
   TokenSequence Apply(const std::vector<TokenSequence> &args, Prescanner &);
 
+  void Print(llvm::raw_ostream &out, const char *macroName = "") const;
+
 private:
   static TokenSequence Tokenize(const std::vector<std::string> &argNames,
       const TokenSequence &token, std::size_t firstToken, std::size_t tokens);
+  // For a given token, return the index of the argument to which the token
+  // corresponds, or `argumentCount` if the token does not correspond to any
+  // argument.
+  std::size_t GetArgumentIndex(const CharBlock &token) const;
 
   bool isFunctionLike_{false};
-  std::size_t argumentCount_{0};
   bool isVariadic_{false};
   bool isDisabled_{false};
   bool isPredefined_{false};
+  std::vector<std::string> argNames_;
   TokenSequence replacement_;
 };
 
@@ -89,6 +96,8 @@ class Preprocessor {
   // Implements a preprocessor directive.
   void Directive(const TokenSequence &, Prescanner &);
 
+  void PrintMacros(llvm::raw_ostream &out) const;
+
 private:
   enum class IsElseActive { No, Yes };
   enum class CanDeadElseAppear { No, Yes };
diff --git a/flang/lib/Parser/token-sequence.h b/flang/include/flang/Parser/token-sequence.h
similarity index 97%
rename from flang/lib/Parser/token-sequence.h
rename to flang/include/flang/Parser/token-sequence.h
index 3df403d41e636..849240d8ec62c 100644
--- a/flang/lib/Parser/token-sequence.h
+++ b/flang/include/flang/Parser/token-sequence.h
@@ -42,8 +42,8 @@ class TokenSequence {
   }
   TokenSequence(TokenSequence &&that)
       : start_{std::move(that.start_)}, nextStart_{that.nextStart_},
-        char_{std::move(that.char_)}, provenances_{
-                                          std::move(that.provenances_)} {}
+        char_{std::move(that.char_)},
+        provenances_{std::move(that.provenances_)} {}
   TokenSequence(const std::string &s, Provenance p) { Put(s, p); }
 
   TokenSequence &operator=(const TokenSequence &that) {
diff --git a/flang/lib/Frontend/CompilerInvocation.cpp b/flang/lib/Frontend/CompilerInvocation.cpp
index c830c7af2462c..8ce6ab7baf481 100644
--- a/flang/lib/Frontend/CompilerInvocation.cpp
+++ b/flang/lib/Frontend/CompilerInvocation.cpp
@@ -772,6 +772,7 @@ static void parsePreprocessorArgs(Fortran::frontend::PreprocessorOptions &opts,
 
   opts.noReformat = args.hasArg(clang::driver::options::OPT_fno_reformat);
   opts.noLineDirectives = args.hasArg(clang::driver::options::OPT_P);
+  opts.showMacros = args.hasArg(clang::driver::options::OPT_dM);
 }
 
 /// Parses all semantic related arguments and populates the variables
diff --git a/flang/lib/Frontend/FrontendActions.cpp b/flang/lib/Frontend/FrontendActions.cpp
index 849b3c8e4dc02..8f251997ed401 100644
--- a/flang/lib/Frontend/FrontendActions.cpp
+++ b/flang/lib/Frontend/FrontendActions.cpp
@@ -399,7 +399,9 @@ void PrintPreprocessedAction::executeAction() {
 
   // Format or dump the prescanner's output
   CompilerInstance &ci = this->getInstance();
-  if (ci.getInvocation().getPreprocessorOpts().noReformat) {
+  if (ci.getInvocation().getPreprocessorOpts().showMacros) {
+    ci.getParsing().EmitPreprocessorMacros(outForPP);
+  } else if (ci.getInvocation().getPreprocessorOpts().noReformat) {
     ci.getParsing().DumpCookedChars(outForPP);
   } else {
     ci.getParsing().EmitPreprocessedSource(
diff --git a/flang/lib/Parser/parsing.cpp b/flang/lib/Parser/parsing.cpp
index a55d33bf6b91d..43a898ff120c5 100644
--- a/flang/lib/Parser/parsing.cpp
+++ b/flang/lib/Parser/parsing.cpp
@@ -7,10 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "flang/Parser/parsing.h"
-#include "preprocessor.h"
 #include "prescan.h"
 #include "type-parsers.h"
 #include "flang/Parser/message.h"
+#include "flang/Parser/preprocessor.h"
 #include "flang/Parser/provenance.h"
 #include "flang/Parser/source.h"
 #include "llvm/Support/raw_ostream.h"
@@ -60,20 +60,19 @@ const SourceFile *Parsing::Prescan(const std::string &path, Options options) {
     }
   }
 
-  Preprocessor preprocessor{allSources};
   if (!options.predefinitions.empty()) {
-    preprocessor.DefineStandardMacros();
+    preprocessor_.DefineStandardMacros();
     for (const auto &predef : options.predefinitions) {
       if (predef.second) {
-        preprocessor.Define(predef.first, *predef.second);
+        preprocessor_.Define(predef.first, *predef.second);
       } else {
-        preprocessor.Undefine(predef.first);
+        preprocessor_.Undefine(predef.first);
       }
     }
   }
   currentCooked_ = &allCooked_.NewCookedSource();
   Prescanner prescanner{
-      messages_, *currentCooked_, preprocessor, options.features};
+      messages_, *currentCooked_, preprocessor_, options.features};
   prescanner.set_fixedForm(options.isFixedForm)
       .set_fixedFormColumnLimit(options.fixedFormColumns)
       .AddCompilerDirectiveSentinel("dir$");
@@ -87,7 +86,7 @@ const SourceFile *Parsing::Prescan(const std::string &path, Options options) {
   if (options.features.IsEnabled(LanguageFeature::CUDA)) {
     prescanner.AddCompilerDirectiveSentinel("$cuf");
     prescanner.AddCompilerDirectiveSentinel("@cuf");
-    preprocessor.Define("_CUDA", "1");
+    preprocessor_.Define("_CUDA", "1");
   }
   ProvenanceRange range{allSources.AddIncludedFile(
       *sourceFile, ProvenanceRange{}, options.isModuleFile)};
@@ -107,6 +106,10 @@ const SourceFile *Parsing::Prescan(const std::string &path, Options options) {
   return sourceFile;
 }
 
+void Parsing::EmitPreprocessorMacros(llvm::raw_ostream &out) const {
+  preprocessor_.PrintMacros(out);
+}
+
 void Parsing::EmitPreprocessedSource(
     llvm::raw_ostream &out, bool lineDirectives) const {
   const std::string *sourcePath{nullptr};
diff --git a/flang/lib/Parser/preprocessor.cpp b/flang/lib/Parser/preprocessor.cpp
index 515b8f62daf9a..2fba28b0c0c7d 100644
--- a/flang/lib/Parser/preprocessor.cpp
+++ b/flang/lib/Parser/preprocessor.cpp
@@ -6,7 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "preprocessor.h"
+#include "flang/Parser/preprocessor.h"
+
 #include "prescan.h"
 #include "flang/Common/idioms.h"
 #include "flang/Parser/characters.h"
@@ -21,6 +22,7 @@
 #include <optional>
 #include <set>
 #include <utility>
+#include <vector>
 
 namespace Fortran::parser {
 
@@ -31,8 +33,7 @@ Definition::Definition(
 Definition::Definition(const std::vector<std::string> &argNames,
     const TokenSequence &repl, std::size_t firstToken, std::size_t tokens,
     bool isVariadic)
-    : isFunctionLike_{true},
-      argumentCount_(argNames.size()), isVariadic_{isVariadic},
+    : isFunctionLike_{true}, isVariadic_{isVariadic}, argNames_{argNames},
       replacement_{Tokenize(argNames, repl, firstToken, tokens)} {}
 
 Definition::Definition(const std::string &predefined, AllSources &sources)
@@ -46,6 +47,37 @@ bool Definition::set_isDisabled(bool disable) {
   return was;
 }
 
+void Definition::Print(llvm::raw_ostream &out, const char *macroName) const {
+  if (!isFunctionLike_) {
+    // If it's not a function-like macro, then just print the replacement.
+    out << ' ' << replacement_.ToString();
+    return;
+  }
+
+  size_t argCount{argumentCount()};
+
+  out << '(';
+  for (size_t i{0}; i != argCount; ++i) {
+    if (i != 0) {
+      out << ", ";
+    }
+    out << argNames_[i];
+  }
+  if (isVariadic_) {
+    out << ", ...";
+  }
+  out << ") ";
+
+  for (size_t i{0}, e{replacement_.SizeInTokens()}; i != e; ++i) {
+    std::string tok{replacement_.TokenAt(i).ToString()};
+    if (size_t idx{GetArgumentIndex(tok)}; idx < argCount) {
+      out << argNames_[idx];
+    } else {
+      out << tok;
+    }
+  }
+}
+
 static bool IsLegalIdentifierStart(const CharBlock &cpl) {
   return cpl.size() > 0 && IsLegalIdentifierStart(cpl[0]);
 }
@@ -73,6 +105,13 @@ TokenSequence Definition::Tokenize(const std::vector<std::string> &argNames,
   return result;
 }
 
+std::size_t Definition::GetArgumentIndex(const CharBlock &token) const {
+  if (token.size() >= 2 && token[0] == '~') {
+    return static_cast<size_t>(token[1] - 'A');
+  }
+  return argumentCount();
+}
+
 static TokenSequence Stringify(
     const TokenSequence &tokens, AllSources &allSources) {
   TokenSequence result;
@@ -159,7 +198,7 @@ TokenSequence Definition::Apply(
       continue;
     }
     if (bytes == 2 && token[0] == '~') { // argument substitution
-      std::size_t index = token[1] - 'A';
+      std::size_t index{GetArgumentIndex(token)};
       if (index >= args.size()) {
         continue;
       }
@@ -202,8 +241,8 @@ TokenSequence Definition::Apply(
       Provenance commaProvenance{
           prescanner.preprocessor().allSources().CompilerInsertionProvenance(
               ',')};
-      for (std::size_t k{argumentCount_}; k < args.size(); ++k) {
-        if (k > argumentCount_) {
+      for (std::size_t k{argumentCount()}; k < args.size(); ++k) {
+        if (k > argumentCount()) {
           result.Put(","s, commaProvenance);
         }
         result.Put(args[k]);
@@ -212,7 +251,7 @@ TokenSequence Definition::Apply(
         j + 2 < tokens && replacement_.TokenAt(j + 1).OnlyNonBlank() == '(' &&
         parenthesesNesting == 0) {
       parenthesesNesting = 1;
-      skipping = args.size() == argumentCount_;
+      skipping = args.size() == argumentCount();
       ++j;
     } else {
       if (parenthesesNesting > 0) {
@@ -713,6 +752,21 @@ void Preprocessor::Directive(const TokenSequence &dir, Prescanner &prescanner) {
   }
 }
 
+void Preprocessor::PrintMacros(llvm::raw_ostream &out) const {
+  // std::set is ordered. Use that to print the macros in an
+  // alphabetical order.
+  std::set<std::string> macroNames;
+  for (const auto &[name, _] : definitions_) {
+    macroNames.insert(name.ToString());
+  }
+
+  for (const std::string &name : macroNames) {
+    out << "#define " << name;
+    definitions_.at(name).Print(out, name.c_str());
+    out << '\n';
+  }
+}
+
 CharBlock Preprocessor::SaveTokenAsName(const CharBlock &t) {
   names_.push_back(t.ToString());
   return {names_.back().data(), names_.back().size()};
diff --git a/flang/lib/Parser/prescan.cpp b/flang/lib/Parser/prescan.cpp
index e9b23172ed2e2..96db3955299f3 100644
--- a/flang/lib/Parser/prescan.cpp
+++ b/flang/lib/Parser/prescan.cpp
@@ -7,12 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "prescan.h"
-#include "preprocessor.h"
-#include "token-sequence.h"
 #include "flang/Common/idioms.h"
 #include "flang/Parser/characters.h"
 #include "flang/Parser/message.h"
+#include "flang/Parser/preprocessor.h"
 #include "flang/Parser/source.h"
+#include "flang/Parser/token-sequence.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cstddef>
 #include <cstring>
diff --git a/flang/lib/Parser/prescan.h b/flang/lib/Parser/prescan.h
index 7442b5d226335..581980001bcc2 100644
--- a/flang/lib/Parser/prescan.h
+++ b/flang/lib/Parser/prescan.h
@@ -16,11 +16,11 @@
 // fixed form character literals on truncated card images, file
 // inclusion, and driving the Fortran source preprocessor.
 
-#include "token-sequence.h"
 #include "flang/Common/Fortran-features.h"
 #include "flang/Parser/characters.h"
 #include "flang/Parser/message.h"
 #include "flang/Parser/provenance.h"
+#include "flang/Parser/token-sequence.h"
 #include <bitset>
 #include <optional>
 #include <string>
diff --git a/flang/lib/Parser/token-sequence.cpp b/flang/lib/Parser/token-sequence.cpp
index 799d13a423660..d0254ecd5aaef 100644
--- a/flang/lib/Parser/token-sequence.cpp
+++ b/flang/lib/Parser/token-sequence.cpp
@@ -6,7 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "token-sequence.h"
+#include "flang/Parser/token-sequence.h"
+
 #include "prescan.h"
 #include "flang/Parser/characters.h"
 #include "flang/Parser/message.h"
diff --git a/flang/test/Driver/driver-help-hidden.f90 b/flang/test/Driver/driver-help-hidden.f90
index 48f48f5384fdc..35b188ad3a9ec 100644
--- a/flang/test/Driver/driver-help-hidden.f90
+++ b/flang/test/Driver/driver-help-hidden.f90
@@ -21,6 +21,7 @@
 ! CHECK-NEXT: -ccc-print-phases       Dump list of actions to perform
 ! CHECK-NEXT: -cpp                    Enable predefined and command line preprocessor macros
 ! CHECK-NEXT: -c                      Only run preprocess, compile, and assemble steps
+! CHECK-NEXT: -dM                     Print macro definitions in -E mode instead of normal output
 ! CHECK-NEXT: -dumpmachine            Display the compiler's target processor
 ! CHECK-NEXT: -dumpversion            Display the version of the compiler
 ! CHECK-NEXT: -D <macro>=<value>      Define <macro> to <value> (or 1 if <value> omitted)
diff --git a/flang/test/Driver/driver-help.f90 b/flang/test/Driver/driver-help.f90
index 38f74395a678a..d4dab55f40e8b 100644
--- a/flang/test/Driver/driver-help.f90
+++ b/flang/test/Driver/driver-help.f90
@@ -17,6 +17,7 @@
 ! HELP-NEXT: -###                    Print (but do not run) the commands to run for this compilation
 ! HELP-NEXT: -cpp                    Enable predefined and command line preprocessor macros
 ! HELP-NEXT: -c                      Only run preprocess, compile, and assemble steps
+! HELP-NEXT: -dM                     Print macro definitions in -E mode instead of normal output
 ! HELP-NEXT: -dumpmachine            Display the compiler's target processor
 ! HELP-NEXT: -dumpversion            Display the version of the compiler
 ! HELP-NEXT: -D <macro>=<value>      Define <macro> to <value> (or 1 if <value> omitted)
@@ -155,6 +156,7 @@
 ! HELP-FC1-NEXT:OPTIONS:
 ! HELP-FC1-NEXT: -cpp                    Enable predefined and command line preprocessor macros
 ! HELP-FC1-NEXT: --dependent-lib=<value> Add dependent library
+! HELP-FC1-NEXT: -dM                     Print macro definitions in -E mode instead of normal output
 ! HELP-FC1-NEXT: -D <macro>=<value>      Define <macro> to <value> (or 1 if <value> omitted)
 ! HELP-FC1-NEXT: -emit-fir               Build the parse tree, then lower it to FIR
 ! HELP-FC1-NEXT: -emit-hlfir             Build the parse tree, then lower it to HLFIR
diff --git a/flang/test/Preprocessing/show-macros1.F90 b/flang/test/Preprocessing/show-macros1.F90
new file mode 100644
index 0000000000000..8e3d59a7849f7
--- /dev/null
+++ b/flang/test/Preprocessing/show-macros1.F90
@@ -0,0 +1,14 @@
+! RUN: %flang -dM -E -o - %s | FileCheck %s
+
+! Check the default macros. Omit certain ones such as __LINE__
+! or __FILE__, or target-specific ones, like __x86_64__.
+
+! Macros are printed in the alphabetical order.
+
+! CHECK: #define __DATE__
+! CHECK: #define __TIME__
+! CHECK: #define __flang__
+! CHECK: #define __flang_major__
+! CHECK: #define __flang_minor__
+! CHECK: #define __flang_patchlevel__
+
diff --git a/flang/test/Preprocessing/show-macros2.F90 b/flang/test/Preprocessing/show-macros2.F90
new file mode 100644
index 0000000000000..baf52ba8161f1
--- /dev/null
+++ b/flang/test/Preprocessing/show-macros2.F90
@@ -0,0 +1,6 @@
+! RUN: %flang -DFOO -DBAR=FOO -dM -E -o - %s | FileCheck %s
+
+! Check command line definitions
+
+! CHECK: #define BAR FOO
+! CHECK: #define FOO 1
diff --git a/flang/test/Preprocessing/show-macros3.F90 b/flang/test/Preprocessing/show-macros3.F90
new file mode 100644
index 0000000000000..951a1ec5ba16f
--- /dev/null
+++ b/flang/test/Preprocessing/show-macros3.F90
@@ -0,0 +1,9 @@
+! RUN: %flang -dM -E -o - %s | FileCheck %s
+
+! Variadic macro
+#define FOO1(X, Y, ...) bar(bar(X, Y), __VA_ARGS__)
+! CHECK: #define FOO1(X, Y, ...) bar(bar(X, Y), __VA_ARGS__)
+
+! Macro with an unused parameter
+#define FOO2(X, Y, Z) (X + Z)
+! CHECK: #define FOO2(X, Y, Z) (X + Z)

From 52aaa8a87960a7d342c5e6b7d5af82c76c8cc45d Mon Sep 17 00:00:00 2001
From: Jordan Rupprecht <rupprecht@google.com>
Date: Wed, 10 Apr 2024 10:46:53 -0500
Subject: [PATCH 028/886] [clang][test] Avoid writing to a potentially
 write-protected dir (#88258)

This test just checks for the stdout/stderr of clang, but it
incidentally tries to write to `a.out` in the current directory, which
may be write protected. Typically one would write `clang -o %t.o` for a
writeable dir, but since we only care about stdout/stderr, throw away
the object file and just write to /dev/null instead.
---
 clang/test/Driver/lld-repro.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/clang/test/Driver/lld-repro.c b/clang/test/Driver/lld-repro.c
index 9457dd334b5b9..61904c0e6df30 100644
--- a/clang/test/Driver/lld-repro.c
+++ b/clang/test/Driver/lld-repro.c
@@ -4,12 +4,12 @@
 // RUN: echo "-nostartfiles -nostdlib -fuse-ld=lld -gen-reproducer=error -fcrash-diagnostics-dir=%t" \
 // RUN:   | sed -e 's/\\/\\\\/g' > %t.rsp
 
-// RUN: not %clang %s @%t.rsp -fcrash-diagnostics=all 2>&1 \
+// RUN: not %clang %s @%t.rsp -fcrash-diagnostics=all -o /dev/null 2>&1 \
 // RUN:   | FileCheck %s
 
 // Test that the reproducer can still be created even when the input source cannot be preprocessed
 // again, like when reading from stdin.
-// RUN: not %clang -x c - @%t.rsp -fcrash-diagnostics=all 2>&1 < %s \
+// RUN: not %clang -x c - @%t.rsp -fcrash-diagnostics=all -o /dev/null 2>&1 < %s \
 // RUN:   | FileCheck %s
 
 // check that we still get lld's output
@@ -20,9 +20,9 @@
 // CHECK-NEXT: note: diagnostic msg:
 // CHECK: ********************
 
-// RUN: not %clang %s @%t.rsp -fcrash-diagnostics=compiler 2>&1 \
+// RUN: not %clang %s @%t.rsp -fcrash-diagnostics=compiler -o /dev/null 2>&1 \
 // RUN:   | FileCheck %s --check-prefix=NO-LINKER
-// RUN: not %clang %s @%t.rsp 2>&1 \
+// RUN: not %clang %s @%t.rsp -o /dev/null 2>&1 \
 // RUN:   | FileCheck %s --check-prefix=NO-LINKER
 
 // NO-LINKER-NOT: Preprocessed source(s) and associated run script(s) are located at:

From 0ad663ead1242e908a8c5005f35e72747d136a3b Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Wed, 10 Apr 2024 17:51:02 +0200
Subject: [PATCH 029/886] [libc++] Removes Clang-16 support. (#87810)

With the release of Clang-18 we no longer officially support Clang-16.
---
 libcxx/include/__algorithm/simd_utils.h                       | 2 +-
 libcxx/include/__config                                       | 4 ++--
 libcxx/test/libcxx/gdb/gdb_pretty_printer_test.sh.cpp         | 2 +-
 .../range.utility/range.utility.conv/to_deduction.pass.cpp    | 2 +-
 .../format/format.arguments/format.arg/visit.pass.cpp         | 2 +-
 .../format.arguments/format.arg/visit.return_type.pass.cpp    | 2 +-
 .../format.arg/visit_format_arg.deprecated.verify.cpp         | 2 +-
 .../variant/variant.visit.member/robust_against_adl.pass.cpp  | 2 +-
 .../std/utilities/variant/variant.visit.member/visit.pass.cpp | 2 +-
 .../variant/variant.visit.member/visit_return_type.pass.cpp   | 2 +-
 10 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/libcxx/include/__algorithm/simd_utils.h b/libcxx/include/__algorithm/simd_utils.h
index 989a1957987e1..8d540ae2cce88 100644
--- a/libcxx/include/__algorithm/simd_utils.h
+++ b/libcxx/include/__algorithm/simd_utils.h
@@ -27,7 +27,7 @@ _LIBCPP_PUSH_MACROS
 #include <__undef_macros>
 
 // TODO: Find out how altivec changes things and allow vectorizations there too.
-#if _LIBCPP_STD_VER >= 14 && defined(_LIBCPP_CLANG_VER) && _LIBCPP_CLANG_VER >= 1700 && !defined(__ALTIVEC__)
+#if _LIBCPP_STD_VER >= 14 && defined(_LIBCPP_CLANG_VER) && !defined(__ALTIVEC__)
 #  define _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS 1
 #else
 #  define _LIBCPP_HAS_ALGORITHM_VECTOR_UTILS 0
diff --git a/libcxx/include/__config b/libcxx/include/__config
index 8550b1da4a278..d98b54926bbe8 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -44,8 +44,8 @@
 // Warn if a compiler version is used that is not supported anymore
 // LLVM RELEASE Update the minimum compiler versions
 #  if defined(_LIBCPP_CLANG_VER)
-#    if _LIBCPP_CLANG_VER < 1600
-#      warning "Libc++ only supports Clang 16 and later"
+#    if _LIBCPP_CLANG_VER < 1700
+#      warning "Libc++ only supports Clang 17 and later"
 #    endif
 #  elif defined(_LIBCPP_APPLE_CLANG_VER)
 #    if _LIBCPP_APPLE_CLANG_VER < 1500
diff --git a/libcxx/test/libcxx/gdb/gdb_pretty_printer_test.sh.cpp b/libcxx/test/libcxx/gdb/gdb_pretty_printer_test.sh.cpp
index 479c1b93fcab9..640365889efae 100644
--- a/libcxx/test/libcxx/gdb/gdb_pretty_printer_test.sh.cpp
+++ b/libcxx/test/libcxx/gdb/gdb_pretty_printer_test.sh.cpp
@@ -12,7 +12,7 @@
 // UNSUPPORTED: c++03
 
 // TODO: Investigate these failures which break the CI.
-// UNSUPPORTED: clang-16, clang-17, clang-18, clang-19
+// UNSUPPORTED: clang-17, clang-18, clang-19
 
 // TODO: Investigate this failure on GCC 13 (in Ubuntu Jammy)
 // UNSUPPORTED: gcc-13
diff --git a/libcxx/test/std/ranges/range.utility/range.utility.conv/to_deduction.pass.cpp b/libcxx/test/std/ranges/range.utility/range.utility.conv/to_deduction.pass.cpp
index f84cedbc122a1..58307bd88d0fe 100644
--- a/libcxx/test/std/ranges/range.utility/range.utility.conv/to_deduction.pass.cpp
+++ b/libcxx/test/std/ranges/range.utility/range.utility.conv/to_deduction.pass.cpp
@@ -9,7 +9,7 @@
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20
 // There is a bug in older versions of Clang that causes trouble with constraints in classes like
 // `ContainerWithDirectCtr`.
-// XFAIL: clang-16, apple-clang-15
+// XFAIL: apple-clang-15
 
 // template<template<class...> class C, input_range R, class... Args>
 //   constexpr auto to(R&& r, Args&&... args);  // Since C++23
diff --git a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.pass.cpp b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.pass.cpp
index 994ccc70a38da..284b03c32cd09 100644
--- a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.pass.cpp
+++ b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.pass.cpp
@@ -8,7 +8,7 @@
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // UNSUPPORTED: GCC-ALWAYS_INLINE-FIXME
 // The tested functionality needs deducing this.
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <format>
diff --git a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.return_type.pass.cpp b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.return_type.pass.cpp
index b2c40d1604547..4c60cb0e9ae1d 100644
--- a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.return_type.pass.cpp
+++ b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit.return_type.pass.cpp
@@ -8,7 +8,7 @@
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // UNSUPPORTED: GCC-ALWAYS_INLINE-FIXME
 // The tested functionality needs deducing this.
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <format>
diff --git a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit_format_arg.deprecated.verify.cpp b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit_format_arg.deprecated.verify.cpp
index acd9228369e60..6a3896c8965c7 100644
--- a/libcxx/test/std/utilities/format/format.arguments/format.arg/visit_format_arg.deprecated.verify.cpp
+++ b/libcxx/test/std/utilities/format/format.arguments/format.arg/visit_format_arg.deprecated.verify.cpp
@@ -7,7 +7,7 @@
 
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // UNSUPPORTED: GCC-ALWAYS_INLINE-FIXME
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <format>
diff --git a/libcxx/test/std/utilities/variant/variant.visit.member/robust_against_adl.pass.cpp b/libcxx/test/std/utilities/variant/variant.visit.member/robust_against_adl.pass.cpp
index c54f2b722d46a..bea6d949924bd 100644
--- a/libcxx/test/std/utilities/variant/variant.visit.member/robust_against_adl.pass.cpp
+++ b/libcxx/test/std/utilities/variant/variant.visit.member/robust_against_adl.pass.cpp
@@ -8,7 +8,7 @@
 
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // The tested functionality needs deducing this.
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <variant>
diff --git a/libcxx/test/std/utilities/variant/variant.visit.member/visit.pass.cpp b/libcxx/test/std/utilities/variant/variant.visit.member/visit.pass.cpp
index d0c909985bbb3..857e85d00857a 100644
--- a/libcxx/test/std/utilities/variant/variant.visit.member/visit.pass.cpp
+++ b/libcxx/test/std/utilities/variant/variant.visit.member/visit.pass.cpp
@@ -8,7 +8,7 @@
 
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // The tested functionality needs deducing this.
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <variant>
diff --git a/libcxx/test/std/utilities/variant/variant.visit.member/visit_return_type.pass.cpp b/libcxx/test/std/utilities/variant/variant.visit.member/visit_return_type.pass.cpp
index 3312197d8df9c..2c1cbb06e7067 100644
--- a/libcxx/test/std/utilities/variant/variant.visit.member/visit_return_type.pass.cpp
+++ b/libcxx/test/std/utilities/variant/variant.visit.member/visit_return_type.pass.cpp
@@ -8,7 +8,7 @@
 
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23
 // The tested functionality needs deducing this.
-// UNSUPPORTED: clang-16 || clang-17
+// UNSUPPORTED: clang-17
 // XFAIL: apple-clang
 
 // <variant>

From fc3dff9b4637bb5960fe70add90cd27e6842d58b Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Wed, 10 Apr 2024 09:08:01 -0700
Subject: [PATCH 030/886] [clang][modules] Stop eagerly reading files with
 diagnostic pragmas (#87442)

This makes it so that the importer doesn't need to stat all input files
of a module that contain diagnostic pragmas, reducing file system
traffic.
---
 clang/lib/Serialization/ASTReader.cpp         |  2 --
 clang/test/Modules/home-is-cwd-search-paths.c | 34 +++++++++++++++++++
 2 files changed, 34 insertions(+), 2 deletions(-)
 create mode 100644 clang/test/Modules/home-is-cwd-search-paths.c

diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 679302e7a838f..2e73a0a714010 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -6626,8 +6626,6 @@ void ASTReader::ReadPragmaDiagnosticMappings(DiagnosticsEngine &Diag) {
              "Invalid data, missing pragma diagnostic states");
       FileID FID = ReadFileID(F, Record, Idx);
       assert(FID.isValid() && "invalid FileID for transition");
-      // FIXME: Remove this once we don't need the side-effects.
-      (void)SourceMgr.getSLocEntryOrNull(FID);
       unsigned Transitions = Record[Idx++];
 
       // Note that we don't need to set up Parent/ParentOffset here, because
diff --git a/clang/test/Modules/home-is-cwd-search-paths.c b/clang/test/Modules/home-is-cwd-search-paths.c
new file mode 100644
index 0000000000000..0b8954e691bc0
--- /dev/null
+++ b/clang/test/Modules/home-is-cwd-search-paths.c
@@ -0,0 +1,34 @@
+// This test demonstrates how -fmodule-map-file-home-is-cwd with -fmodules-embed-all-files
+// extend the importer search paths by relying on the side effects of pragma diagnostic
+// mappings deserialization.
+
+// RUN: rm -rf %t
+// RUN: split-file %s %t
+
+//--- dir1/a.modulemap
+module a { header "a.h" }
+//--- dir1/a.h
+#include "search.h"
+// The first compilation is configured such that -I search does contain the search.h header.
+//--- dir1/search/search.h
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wparentheses"
+#pragma clang diagnostic pop
+// RUN: cd %t/dir1 && %clang_cc1 -fmodules -I search \
+// RUN:   -emit-module -fmodule-name=a a.modulemap -o %t/a.pcm \
+// RUN:   -fmodules-embed-all-files -fmodule-map-file-home-is-cwd
+
+//--- dir2/b.modulemap
+module b { header "b.h" }
+//--- dir2/b.h
+#include "search.h" // expected-error{{'search.h' file not found}}
+// The second compilation is configured such that -I search is an empty directory.
+// However, since b.pcm simply embeds the headers as "search/search.h", this compilation
+// ends up seeing it too. This relies solely on ASTReader::ReadPragmaDiagnosticMappings()
+// eagerly reading the corresponding INPUT_FILE record before header search happens.
+// Removing the eager deserialization makes this header invisible and so does removing
+// the pragma directives.
+// RUN: mkdir %t/dir2/search
+// RUN: cd %t/dir2 && %clang_cc1 -fmodules -I search \
+// RUN:   -emit-module -fmodule-name=b b.modulemap -o %t/b.pcm \
+// RUN:   -fmodule-file=%t/a.pcm -verify

From 51786eb5bfc30e7eff998323a9ce433ec4620383 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Wed, 10 Apr 2024 09:08:40 -0700
Subject: [PATCH 031/886] [clang][modules] Only compute affecting module maps
 with implicit search (#87849)

When writing out a PCM, we compute the set of module maps that did
affect the compilation and we strip the rest to make the output
independent of them. The most common way to read a module map that is
not affecting is with implicit module map search. The other option is to
pass a bunch of unnecessary `-fmodule-map-file=<path>` arguments on the
command-line, in which case the client should probably not give those to
Clang anyway.

This makes serialization of explicit modules faster, mostly due to
reduced file system traffic.
---
 clang/lib/Serialization/ASTWriter.cpp | 21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 4cd74b1ba9d72..d2afe378bb0c3 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -163,8 +163,13 @@ static TypeCode getTypeCodeForTypeClass(Type::TypeClass id) {
 
 namespace {
 
-std::set<const FileEntry *> GetAffectingModuleMaps(const Preprocessor &PP,
-                                                   Module *RootModule) {
+std::optional<std::set<const FileEntry *>>
+GetAffectingModuleMaps(const Preprocessor &PP, Module *RootModule) {
+  // Without implicit module map search, there's no good reason to know about
+  // any module maps that are not affecting.
+  if (!PP.getHeaderSearchInfo().getHeaderSearchOpts().ImplicitModuleMaps)
+    return std::nullopt;
+
   SmallVector<const Module *> ModulesToProcess{RootModule};
 
   const HeaderSearch &HS = PP.getHeaderSearchInfo();
@@ -4735,8 +4740,16 @@ void ASTWriter::computeNonAffectingInputFiles() {
     if (!Cache->OrigEntry)
       continue;
 
-    if (!isModuleMap(File.getFileCharacteristic()) ||
-        llvm::is_contained(AffectingModuleMaps, *Cache->OrigEntry))
+    // Don't prune anything other than module maps.
+    if (!isModuleMap(File.getFileCharacteristic()))
+      continue;
+
+    // Don't prune module maps if all are guaranteed to be affecting.
+    if (!AffectingModuleMaps)
+      continue;
+
+    // Don't prune module maps that are affecting.
+    if (llvm::is_contained(*AffectingModuleMaps, *Cache->OrigEntry))
       continue;
 
     IsSLocAffecting[I] = false;

From 323d3ab2574ba9d371926bb1b5c67dbe7b2b4ec3 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 10 Apr 2024 09:08:50 -0700
Subject: [PATCH 032/886] [RISCV] Optimize undef Even vector in
 getWideningInterleave. (#88221)

We recently optimized the code when the Odd vector was undef to fix a
poison bug.

There are additional optimizations we can do if the even vector is
undef. With Zvbb, we can use a single vwsll. Without Zvbb, we can use a
vzext.vf2 and a vsll.
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 13 ++++++++++--
 .../CodeGen/RISCV/rvv/vector-interleave.ll    | 20 +++++++++++++++++++
 2 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 6e97575c167cd..944d8b6de895d 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -4638,8 +4638,17 @@ static SDValue getWideningInterleave(SDValue EvenV, SDValue OddV,
                                            Subtarget.getXLenVT()));
     Interleaved = DAG.getNode(RISCVISD::VWSLL_VL, DL, WideContainerVT, OddV,
                               OffsetVec, Passthru, Mask, VL);
-    Interleaved = DAG.getNode(RISCVISD::VWADDU_W_VL, DL, WideContainerVT,
-                              Interleaved, EvenV, Passthru, Mask, VL);
+    if (!EvenV.isUndef())
+      Interleaved = DAG.getNode(RISCVISD::VWADDU_W_VL, DL, WideContainerVT,
+                                Interleaved, EvenV, Passthru, Mask, VL);
+  } else if (EvenV.isUndef()) {
+    Interleaved =
+        DAG.getNode(RISCVISD::VZEXT_VL, DL, WideContainerVT, OddV, Mask, VL);
+
+    SDValue OffsetVec =
+        DAG.getConstant(VecVT.getScalarSizeInBits(), DL, WideContainerVT);
+    Interleaved = DAG.getNode(RISCVISD::SHL_VL, DL, WideContainerVT,
+                              Interleaved, OffsetVec, Passthru, Mask, VL);
   } else {
     // FIXME: We should freeze the odd vector here. We already handled the case
     // of provably undef/poison above.
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll b/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
index 0992c9fe495f4..4b6ad0f27214d 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
@@ -674,6 +674,26 @@ define <vscale x 8 x i32> @vector_interleave_nxv8i32_nxv4i32_poison(<vscale x 4
   ret <vscale x 8 x i32> %res
 }
 
+define <vscale x 8 x i32> @vector_interleave_nxv8i32_nxv4i32_poison2(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: vector_interleave_nxv8i32_nxv4i32_poison2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vzext.vf2 v12, v8
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsll.vx v8, v12, a0
+; CHECK-NEXT:    ret
+;
+; ZVBB-LABEL: vector_interleave_nxv8i32_nxv4i32_poison2:
+; ZVBB:       # %bb.0:
+; ZVBB-NEXT:    li a0, 32
+; ZVBB-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; ZVBB-NEXT:    vwsll.vx v12, v8, a0
+; ZVBB-NEXT:    vmv4r.v v8, v12
+; ZVBB-NEXT:    ret
+  %res = call <vscale x 8 x i32> @llvm.experimental.vector.interleave2.nxv8i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a)
+  ret <vscale x 8 x i32> %res
+}
+
 declare <vscale x 64 x half> @llvm.experimental.vector.interleave2.nxv64f16(<vscale x 32 x half>, <vscale x 32 x half>)
 declare <vscale x 32 x float> @llvm.experimental.vector.interleave2.nxv32f32(<vscale x 16 x float>, <vscale x 16 x float>)
 declare <vscale x 16 x double> @llvm.experimental.vector.interleave2.nxv16f64(<vscale x 8 x double>, <vscale x 8 x double>)

From e72c949c15208ba3dd53a9cebfee02734965a678 Mon Sep 17 00:00:00 2001
From: Evgenii Stepanov <eugeni.stepanov@gmail.com>
Date: Wed, 10 Apr 2024 09:12:25 -0700
Subject: [PATCH 033/886] [msan] Overflow intrinsics. (#88210)

---
 .../Instrumentation/MemorySanitizer.cpp       |  24 ++++
 .../MemorySanitizer/overflow.ll               | 103 ++++++------------
 2 files changed, 59 insertions(+), 68 deletions(-)

diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index 46b9181c8922e..ee3531bbd68df 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -3715,8 +3715,32 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
     setOrigin(&I, getOrigin(&I, 0));
   }
 
+  void handleArithmeticWithOverflow(IntrinsicInst &I) {
+    IRBuilder<> IRB(&I);
+    Value *Shadow0 = getShadow(&I, 0);
+    Value *Shadow1 = getShadow(&I, 1);
+    Value *ShadowElt0 = IRB.CreateOr(Shadow0, Shadow1);
+    Value *ShadowElt1 =
+        IRB.CreateICmpNE(ShadowElt0, getCleanShadow(ShadowElt0));
+
+    Value *Shadow = PoisonValue::get(getShadowTy(&I));
+    Shadow = IRB.CreateInsertValue(Shadow, ShadowElt0, 0);
+    Shadow = IRB.CreateInsertValue(Shadow, ShadowElt1, 1);
+
+    setShadow(&I, Shadow);
+    setOriginForNaryOp(I);
+  }
+
   void visitIntrinsicInst(IntrinsicInst &I) {
     switch (I.getIntrinsicID()) {
+    case Intrinsic::uadd_with_overflow:
+    case Intrinsic::sadd_with_overflow:
+    case Intrinsic::usub_with_overflow:
+    case Intrinsic::ssub_with_overflow:
+    case Intrinsic::umul_with_overflow:
+    case Intrinsic::smul_with_overflow:
+      handleArithmeticWithOverflow(I);
+      break;
     case Intrinsic::abs:
       handleAbsIntrinsic(I);
       break;
diff --git a/llvm/test/Instrumentation/MemorySanitizer/overflow.ll b/llvm/test/Instrumentation/MemorySanitizer/overflow.ll
index b1304faec3df0..0cfae0008263f 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/overflow.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/overflow.ll
@@ -10,16 +10,12 @@ define {i64, i1} @test_sadd_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0:![0-9]+]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4:[0-9]+]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
@@ -32,16 +28,12 @@ define {i64, i1} @test_uadd_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
@@ -54,16 +46,12 @@ define {i64, i1} @test_smul_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
@@ -75,16 +63,12 @@ define {i64, i1} @test_umul_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
@@ -96,16 +80,12 @@ define {i64, i1} @test_ssub_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
@@ -117,16 +97,12 @@ define {i64, i1} @test_usub_with_overflow(i64 %a, i64 %b) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i64 [[TMP2]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP3:%.*]], label [[TMP4:%.*]], !prof [[PROF0]]
-; CHECK:       3:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       4:
+; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { i64, i1 } poison, i64 [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { i64, i1 } [[TMP5]], i1 [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A]], i64 [[B]])
-; CHECK-NEXT:    store { i64, i1 } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { i64, i1 } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { i64, i1 } [[RES]]
 ;
   %res = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
@@ -139,18 +115,12 @@ define {<4 x i32>, <4 x i1>} @test_sadd_with_overflow_vec(<4 x i32> %a, <4 x i32
 ; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, ptr @__msan_param_tls, align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
 ; CHECK-NEXT:    call void @llvm.donothing()
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast <4 x i32> [[TMP1]] to i128
-; CHECK-NEXT:    [[_MSCMP:%.*]] = icmp ne i128 [[TMP3]], 0
-; CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP2]] to i128
-; CHECK-NEXT:    [[_MSCMP1:%.*]] = icmp ne i128 [[TMP4]], 0
-; CHECK-NEXT:    [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
-; CHECK-NEXT:    br i1 [[_MSOR]], label [[TMP5:%.*]], label [[TMP6:%.*]], !prof [[PROF0]]
-; CHECK:       5:
-; CHECK-NEXT:    call void @__msan_warning_noreturn() #[[ATTR4]]
-; CHECK-NEXT:    unreachable
-; CHECK:       6:
+; CHECK-NEXT:    [[TMP3:%.*]] = or <4 x i32> [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[TMP3]], zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = insertvalue { <4 x i32>, <4 x i1> } poison, <4 x i32> [[TMP3]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertvalue { <4 x i32>, <4 x i1> } [[TMP5]], <4 x i1> [[TMP4]], 1
 ; CHECK-NEXT:    [[RES:%.*]] = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> [[A]], <4 x i32> [[B]])
-; CHECK-NEXT:    store { <4 x i32>, <4 x i1> } zeroinitializer, ptr @__msan_retval_tls, align 8
+; CHECK-NEXT:    store { <4 x i32>, <4 x i1> } [[TMP6]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret { <4 x i32>, <4 x i1> } [[RES]]
 ;
   %res = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> %a, <4 x i32> %b)
@@ -158,6 +128,3 @@ define {<4 x i32>, <4 x i1>} @test_sadd_with_overflow_vec(<4 x i32> %a, <4 x i32
 }
 
 attributes #0 = { sanitize_memory }
-;.
-; CHECK: [[PROF0]] = !{!"branch_weights", i32 1, i32 1000}
-;.

From 43b2b2ebce635bec1e3c060092ea75db858ee3fd Mon Sep 17 00:00:00 2001
From: Mehdi Amini <joker.eph@gmail.com>
Date: Wed, 10 Apr 2024 18:25:16 +0200
Subject: [PATCH 034/886] Revert "Fix complex log1p accuracy with large abs
 values." (#88290)

Reverts llvm/llvm-project#88260

The test fails on the GCC7 buildbot.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 50 +++++++++----------
 .../convert-to-standard.mlir                  | 48 +++++++-----------
 2 files changed, 41 insertions(+), 57 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 0aa1de5fa5d9a..9c3c4d96a301e 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -570,39 +570,37 @@ struct Log1pOpConversion : public OpConversionPattern<complex::Log1pOp> {
                   ConversionPatternRewriter &rewriter) const override {
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
-    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
+    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
     mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
-    Value real = b.create<complex::ReOp>(adaptor.getComplex());
-    Value imag = b.create<complex::ImOp>(adaptor.getComplex());
+    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
 
     Value half = b.create<arith::ConstantOp>(elementType,
                                              b.getFloatAttr(elementType, 0.5));
     Value one = b.create<arith::ConstantOp>(elementType,
                                             b.getFloatAttr(elementType, 1));
-    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf);
-    Value absRealPlusOne = b.create<math::AbsFOp>(realPlusOne, fmf);
-    Value absImag = b.create<math::AbsFOp>(imag, fmf);
-
-    Value maxAbs = b.create<arith::MaximumFOp>(absRealPlusOne, absImag, fmf);
-    Value minAbs = b.create<arith::MinimumFOp>(absRealPlusOne, absImag, fmf);
-
-    Value maxAbsOfRealPlusOneAndImagMinusOne = b.create<arith::SelectOp>(
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, realPlusOne, absImag,
-                                fmf),
-        real, b.create<arith::SubFOp>(maxAbs, one, fmf));
-    Value minMaxRatio = b.create<arith::DivFOp>(minAbs, maxAbs, fmf);
-    Value logOfMaxAbsOfRealPlusOneAndImag =
-        b.create<math::Log1pOp>(maxAbsOfRealPlusOneAndImagMinusOne, fmf);
-    Value logOfSqrtPart = b.create<math::Log1pOp>(
-        b.create<arith::MulFOp>(minMaxRatio, minMaxRatio, fmf), fmf);
-    Value r = b.create<arith::AddFOp>(
-        b.create<arith::MulFOp>(half, logOfSqrtPart, fmf),
-        logOfMaxAbsOfRealPlusOneAndImag, fmf);
-    Value resultReal = b.create<arith::SelectOp>(
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, r, r, fmf), minAbs,
-        r);
-    Value resultImag = b.create<math::Atan2Op>(imag, realPlusOne, fmf);
+    Value two = b.create<arith::ConstantOp>(elementType,
+                                            b.getFloatAttr(elementType, 2));
+
+    // log1p(a+bi) = .5*log((a+1)^2+b^2) + i*atan2(b, a + 1)
+    // log((a+1)+bi) = .5*log(a*a + 2*a + 1 + b*b) + i*atan2(b, a+1)
+    // log((a+1)+bi) = .5*log1p(a*a + 2*a + b*b) + i*atan2(b, a+1)
+    Value sumSq = b.create<arith::MulFOp>(real, real, fmf.getValue());
+    sumSq = b.create<arith::AddFOp>(
+        sumSq, b.create<arith::MulFOp>(real, two, fmf.getValue()),
+        fmf.getValue());
+    sumSq = b.create<arith::AddFOp>(
+        sumSq, b.create<arith::MulFOp>(imag, imag, fmf.getValue()),
+        fmf.getValue());
+    Value logSumSq =
+        b.create<math::Log1pOp>(elementType, sumSq, fmf.getValue());
+    Value resultReal = b.create<arith::MulFOp>(logSumSq, half, fmf.getValue());
+
+    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf.getValue());
+
+    Value resultImag =
+        b.create<math::Atan2Op>(elementType, imag, realPlusOne, fmf.getValue());
     rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
                                                    resultImag);
     return success();
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 43918904a09f4..f5d9499eadda4 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -300,22 +300,15 @@ func.func @complex_log1p(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
+// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] : f32
+// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] : f32
+// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] : f32
+// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] : f32
+// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] : f32
+// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] : f32
 // CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] : f32
-// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] : f32
-// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
-// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
-// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
-// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
-// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %cst_0 : f32
-// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
-// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
-// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] : f32
-// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] : f32
-// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] : f32
-// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] : f32
-// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] : f32
-// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
@@ -970,22 +963,15 @@ func.func @complex_log1p_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %cst_0 fastmath<nnan,contract>  : f32
-// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
-// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract>  : f32
-// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] fastmath<nnan,contract> : f32
-// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] fastmath<nnan,contract>  : f32
-// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] fastmath<nnan,contract>  : f32
-// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] fastmath<nnan,contract>  : f32
-// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] fastmath<nnan,contract> : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
+// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
+// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract> : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] fastmath<nnan,contract> : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>

From 48c5c70fdd3bec2929e2e903e3bf4494a65f7a92 Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Wed, 10 Apr 2024 09:23:23 -0700
Subject: [PATCH 035/886] [NFC] Update SemaRef.Diag to just Diag in OpenACC
 implementation

I missed these two in my last patch as the two patches crossed in
review, so correct this now.
---
 clang/lib/Sema/SemaOpenACC.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
index b6afb80b873e2..a6f4453e525d0 100644
--- a/clang/lib/Sema/SemaOpenACC.cpp
+++ b/clang/lib/Sema/SemaOpenACC.cpp
@@ -102,10 +102,10 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     });
 
     if (Itr != ExistingClauses.end()) {
-      SemaRef.Diag(Clause.getBeginLoc(),
+      Diag(Clause.getBeginLoc(),
                    diag::err_acc_duplicate_clause_disallowed)
           << Clause.getDirectiveKind() << Clause.getClauseKind();
-      SemaRef.Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+      Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
       return nullptr;
     }
 

From 3d468566eb395995ac54fcf90d3afb9b9f822eb3 Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Wed, 10 Apr 2024 09:37:01 -0700
Subject: [PATCH 036/886] [NFC] Remove unneeded 'maybe_unused' attributes

This was added while we only had a partial implementation of clauses, so
we don't need these anymore.
---
 clang/lib/Serialization/ASTReader.cpp | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 2e73a0a714010..0ca7f6600eee3 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -11754,11 +11754,8 @@ void ASTRecordReader::readOMPChildren(OMPChildren *Data) {
 
 OpenACCClause *ASTRecordReader::readOpenACCClause() {
   OpenACCClauseKind ClauseKind = readEnum<OpenACCClauseKind>();
-  // TODO OpenACC: We don't have these used anywhere, but eventually we should
-  // be constructing the Clauses with them, so these attributes can go away at
-  // that point.
-  [[maybe_unused]] SourceLocation BeginLoc = readSourceLocation();
-  [[maybe_unused]] SourceLocation EndLoc = readSourceLocation();
+  SourceLocation BeginLoc = readSourceLocation();
+  SourceLocation EndLoc = readSourceLocation();
 
   switch (ClauseKind) {
   case OpenACCClauseKind::Default: {

From f388a3a446ef2566d73b6a73ba300738f8c2c002 Mon Sep 17 00:00:00 2001
From: Aart Bik <ajcbik@google.com>
Date: Wed, 10 Apr 2024 09:42:12 -0700
Subject: [PATCH 037/886] [mlir][sparse] update doc and examples of the
 [dis]assemble operations (#88213)

The doc and examples of the [dis]assemble operations did not reflect all
the recent changes on order of the operands. Also clarified some of the
text.
---
 .../SparseTensor/IR/SparseTensorOps.td        | 89 +++++++++----------
 mlir/test/Dialect/SparseTensor/invalid.mlir   |  6 +-
 mlir/test/Dialect/SparseTensor/roundtrip.mlir | 14 +--
 .../Dialect/SparseTensor/CPU/sparse_pack.mlir | 10 ++-
 4 files changed, 61 insertions(+), 58 deletions(-)

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
index 5df8a176459b7..0cfc64f9988a0 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
@@ -58,37 +58,35 @@ def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]>,
     Arguments<(ins Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$levels,
                    TensorOf<[AnyType]>:$values)>,
     Results<(outs AnySparseTensor: $result)> {
-  let summary = "Returns a sparse tensor assembled from the given values and levels";
+  let summary = "Returns a sparse tensor assembled from the given levels and values";
 
   let description = [{
-    Assembles the values and per-level coordinate or postion arrays into a sparse tensor.
-    The order and types of provided levels must be consistent with the actual storage
-    layout of the returned sparse tensor described below.
+    Assembles the per-level position and coordinate arrays together with
+    the values arrays into a sparse tensor. The order and types of the
+    provided levels must be consistent with the actual storage layout of
+    the returned sparse tensor described below.
 
-    - `values : tensor<? x V>`
-      supplies the value for each stored element in the sparse tensor.
     - `levels: [tensor<? x iType>, ...]`
-      each supplies the sparse tensor coordinates scheme in the sparse tensor for
-      the corresponding level as specifed by `sparse_tensor::StorageLayout`.
-
-    This operation can be used to assemble a sparse tensor from external
-    sources; e.g., when passing two numpy arrays from Python.
-
-    Disclaimer: This is the user's responsibility to provide input that can be
-    correctly interpreted by the sparsifier, which does not perform
-    any sanity test during runtime to verify data integrity.
+      supplies the sparse tensor position and coordinate arrays
+      of the sparse tensor for the corresponding level as specifed by
+      `sparse_tensor::StorageLayout`.
+    - `values : tensor<? x V>`
+      supplies the values array for the stored elements in the sparse tensor.
 
-    TODO: The returned tensor is allowed (in principle) to have non-identity
-    dimOrdering/higherOrdering mappings.  However, the current implementation
-    does not yet support them.
+    This operation can be used to assemble a sparse tensor from an
+    external source; e.g., by passing numpy arrays from Python. It
+    is the user's responsibility to provide input that can be correctly
+    interpreted by the sparsifier, which does not perform any sanity
+    test to verify data integrity.
 
     Example:
 
     ```mlir
-    %values      = arith.constant dense<[ 1.1,   2.2,   3.3 ]> : tensor<3xf64>
-    %coordinates = arith.constant dense<[[0,0], [1,2], [1,3]]> : tensor<3x2xindex>
-    %st = sparse_tensor.assemble %values, %coordinates
-        : tensor<3xf64>, tensor<3x2xindex> to tensor<3x4xf64, #COO>
+    %pos    = arith.constant dense<[0, 3]>                : tensor<2xindex>
+    %index  = arith.constant dense<[[0,0], [1,2], [1,3]]> : tensor<3x2xindex>
+    %values = arith.constant dense<[ 1.1,   2.2,   3.3 ]> : tensor<3xf64>
+    %s = sparse_tensor.assemble (%pos, %index), %values
+       : (tensor<2xindex>, tensor<3x2xindex>), tensor<3xf64> to tensor<3x4xf64, #COO>
     // yields COO format |1.1, 0.0, 0.0, 0.0|
     //     of 3x4 matrix |0.0, 0.0, 2.2, 3.3|
     //                   |0.0, 0.0, 0.0, 0.0|
@@ -96,8 +94,8 @@ def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]>,
   }];
 
   let assemblyFormat =
-    "` ` `(` $levels `)` `,` $values attr-dict"
-    " `:` `(` type($levels) `)` `,` type($values) `to` type($result)";
+    "` ` `(` $levels       `)` `,` $values attr-dict `:`"
+    "    `(` type($levels) `)` `,` type($values) `to` type($result)";
 
   let hasVerifier = 1;
 }
@@ -110,21 +108,20 @@ def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVaria
                   TensorOf<[AnyType]>:$ret_values,
                   Variadic<AnyIndexingScalarLike>:$lvl_lens,
                   AnyIndexingScalarLike:$val_len)> {
-  let summary = "Returns the (values, coordinates) pair disassembled from the input tensor";
+  let summary = "Copies the levels and values of the given sparse tensor";
 
   let description = [{
     The disassemble operation is the inverse of `sparse_tensor::assemble`.
-    It returns the values and per-level position and coordinate array to the
-    user from the sparse tensor along with the actual length of the memory used
-    in each returned buffer. This operation can be used for returning an
-    disassembled MLIR sparse tensor to frontend; e.g., returning two numpy arrays
-    to Python.
-
-    Disclaimer: This is the user's responsibility to allocate large enough buffers
-    to hold the sparse tensor. The sparsifier simply copies each fields
-    of the sparse tensor into the user-supplied buffer without bound checking.
+    It copies the per-level position and coordinate arrays together with
+    the values array of the given sparse tensor into the user-supplied buffers
+    along with the actual length of the memory used in each returned buffer.
 
-    TODO: the current implementation does not yet support non-identity mappings.
+    This operation can be used for returning a disassembled MLIR sparse tensor;
+    e.g., copying the sparse tensor contents into pre-allocated numpy arrays
+    back to Python. It is the user's responsibility to allocate large enough
+    buffers of the appropriate types to hold the sparse tensor contents.
+    The sparsifier simply copies all fields of the sparse tensor into the
+    user-supplied buffers without any sanity test to verify data integrity.
 
     Example:
 
@@ -132,26 +129,26 @@ def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVaria
     // input COO format |1.1, 0.0, 0.0, 0.0|
     //    of 3x4 matrix |0.0, 0.0, 2.2, 3.3|
     //                  |0.0, 0.0, 0.0, 0.0|
-    %v, %p, %c, %v_len, %p_len, %c_len =
-        sparse_tensor.disassemble %sp : tensor<3x4xf64, #COO>
-          out_lvls(%op, %oi) : tensor<2xindex>, tensor<3x2xindex>,
-          out_vals(%od) : tensor<3xf64> ->
-          tensor<3xf64>, (tensor<2xindex>, tensor<3x2xindex>), index, (index, index)
-    // %v = arith.constant dense<[ 1.1,   2.2,   3.3 ]> : tensor<3xf64>
+    %p, %c, %v, %p_len, %c_len, %v_len =
+      sparse_tensor.disassemble %s : tensor<3x4xf64, #COO>
+         out_lvls(%op, %oi : tensor<2xindex>, tensor<3x2xindex>)
+         out_vals(%od : tensor<3xf64>) ->
+           (tensor<2xindex>, tensor<3x2xindex>), tensor<3xf64>, (index, index), index
     // %p = arith.constant dense<[ 0,              3 ]> : tensor<2xindex>
     // %c = arith.constant dense<[[0,0], [1,2], [1,3]]> : tensor<3x2xindex>
-    // %v_len = 3
+    // %v = arith.constant dense<[ 1.1,   2.2,   3.3 ]> : tensor<3xf64>
     // %p_len = 2
     // %c_len = 6 (3x2)
+    // %v_len = 3
     ```
   }];
 
   let assemblyFormat =
-    "$tensor `:` type($tensor) "
+    "$tensor attr-dict `:` type($tensor)"
     "`out_lvls` `(` $out_levels `:` type($out_levels) `)` "
-    "`out_vals` `(` $out_values `:` type($out_values) `)` attr-dict"
-    "`->` `(` type($ret_levels) `)` `,` type($ret_values) `,` "
-    "`(` type($lvl_lens) `)` `,` type($val_len)";
+    "`out_vals` `(` $out_values `:` type($out_values) `)` `->`"
+    "`(` type($ret_levels) `)` `,` type($ret_values) `,` "
+    "`(` type($lvl_lens)   `)` `,` type($val_len)";
 
   let hasVerifier = 1;
 }
diff --git a/mlir/test/Dialect/SparseTensor/invalid.mlir b/mlir/test/Dialect/SparseTensor/invalid.mlir
index 18851f29d8eaa..7f5c05190fc9a 100644
--- a/mlir/test/Dialect/SparseTensor/invalid.mlir
+++ b/mlir/test/Dialect/SparseTensor/invalid.mlir
@@ -60,7 +60,7 @@ func.func @invalid_pack_mis_position(%values: tensor<6xf64>, %coordinates: tenso
 
 func.func @invalid_unpack_type(%sp: tensor<100xf32, #SparseVector>, %values: tensor<6xf64>, %pos: tensor<2xi32>, %coordinates: tensor<6x1xi32>) {
   // expected-error@+1 {{input/output element-types don't match}}
-  %rv, %rp, %rc, %vl, %pl, %cl = sparse_tensor.disassemble %sp : tensor<100xf32, #SparseVector>
+  %rp, %rc, %rv, %pl, %cl, %vl = sparse_tensor.disassemble %sp : tensor<100xf32, #SparseVector>
                   out_lvls(%pos, %coordinates : tensor<2xi32>, tensor<6x1xi32>)
                   out_vals(%values : tensor<6xf64>)
                   -> (tensor<2xi32>, tensor<6x1xi32>), tensor<6xf64>, (index, index), index
@@ -73,7 +73,7 @@ func.func @invalid_unpack_type(%sp: tensor<100xf32, #SparseVector>, %values: ten
 
 func.func @invalid_unpack_type(%sp: tensor<100x2xf64, #SparseVector>, %values: tensor<6xf64>, %pos: tensor<2xi32>, %coordinates: tensor<6x3xi32>) {
   // expected-error@+1 {{input/output trailing COO level-ranks don't match}}
-  %rv, %rp, %rc, %vl, %pl, %cl = sparse_tensor.disassemble %sp : tensor<100x2xf64, #SparseVector>
+  %rp, %rc, %rv, %pl, %cl, %vl = sparse_tensor.disassemble %sp : tensor<100x2xf64, #SparseVector>
                   out_lvls(%pos, %coordinates : tensor<2xi32>, tensor<6x3xi32> )
                   out_vals(%values : tensor<6xf64>)
                   -> (tensor<2xi32>, tensor<6x3xi32>), tensor<6xf64>, (index, index), index
@@ -86,7 +86,7 @@ func.func @invalid_unpack_type(%sp: tensor<100x2xf64, #SparseVector>, %values: t
 
 func.func @invalid_unpack_mis_position(%sp: tensor<2x100xf64, #CSR>, %values: tensor<6xf64>, %coordinates: tensor<6xi32>) {
   // expected-error@+1 {{inconsistent number of fields between input/output}}
-  %rv, %rc, %vl, %pl = sparse_tensor.disassemble %sp : tensor<2x100xf64, #CSR>
+  %rc, %rv, %cl, %vl = sparse_tensor.disassemble %sp : tensor<2x100xf64, #CSR>
              out_lvls(%coordinates : tensor<6xi32>)
              out_vals(%values : tensor<6xf64>)
              -> (tensor<6xi32>), tensor<6xf64>, (index), index
diff --git a/mlir/test/Dialect/SparseTensor/roundtrip.mlir b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
index a47a3d5119f96..12f69c1d37b9c 100644
--- a/mlir/test/Dialect/SparseTensor/roundtrip.mlir
+++ b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
@@ -33,21 +33,21 @@ func.func @sparse_pack(%pos: tensor<2xi32>, %index: tensor<6x1xi32>, %data: tens
 #SparseVector = #sparse_tensor.encoding<{map = (d0) -> (d0 : compressed), crdWidth=32}>
 // CHECK-LABEL: func @sparse_unpack(
 //  CHECK-SAME: %[[T:.*]]: tensor<100xf64, #
-//  CHECK-SAME: %[[OD:.*]]: tensor<6xf64>
-//  CHECK-SAME: %[[OP:.*]]: tensor<2xindex>
-//  CHECK-SAME: %[[OI:.*]]: tensor<6x1xi32>
+//  CHECK-SAME: %[[OP:.*]]: tensor<2xindex>,
+//  CHECK-SAME: %[[OI:.*]]: tensor<6x1xi32>,
+//  CHECK-SAME: %[[OD:.*]]: tensor<6xf64>)
 //       CHECK: %[[P:.*]]:2, %[[D:.*]], %[[PL:.*]]:2, %[[DL:.*]] = sparse_tensor.disassemble %[[T]]
 //       CHECK: return %[[P]]#0, %[[P]]#1, %[[D]]
 func.func @sparse_unpack(%sp : tensor<100xf64, #SparseVector>,
-                         %od : tensor<6xf64>,
                          %op : tensor<2xindex>,
-                         %oi : tensor<6x1xi32>)
+                         %oi : tensor<6x1xi32>,
+                         %od : tensor<6xf64>)
                        -> (tensor<2xindex>, tensor<6x1xi32>, tensor<6xf64>) {
-  %rp, %ri, %rd, %vl, %pl, %cl = sparse_tensor.disassemble %sp : tensor<100xf64, #SparseVector>
+  %rp, %ri, %d, %rpl, %ril, %dl = sparse_tensor.disassemble %sp : tensor<100xf64, #SparseVector>
                   out_lvls(%op, %oi : tensor<2xindex>, tensor<6x1xi32>)
                   out_vals(%od : tensor<6xf64>)
                   -> (tensor<2xindex>, tensor<6x1xi32>), tensor<6xf64>, (index, index), index
-  return %rp, %ri, %rd : tensor<2xindex>, tensor<6x1xi32>, tensor<6xf64>
+  return %rp, %ri, %d : tensor<2xindex>, tensor<6x1xi32>, tensor<6xf64>
 }
 
 // -----
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_pack.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_pack.mlir
index 7ecccad212cdb..5415625ff05d6 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_pack.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_pack.mlir
@@ -231,7 +231,7 @@ module {
     %od = tensor.empty() : tensor<3xf64>
     %op = tensor.empty() : tensor<2xi32>
     %oi = tensor.empty() : tensor<3x2xi32>
-    %p, %i, %d, %dl, %pl, %il = sparse_tensor.disassemble %s5 : tensor<10x10xf64, #SortedCOOI32>
+    %p, %i, %d, %pl, %il, %dl = sparse_tensor.disassemble %s5 : tensor<10x10xf64, #SortedCOOI32>
                  out_lvls(%op, %oi : tensor<2xi32>, tensor<3x2xi32>)
                  out_vals(%od : tensor<3xf64>)
                  -> (tensor<2xi32>, tensor<3x2xi32>), tensor<3xf64>, (i32, i64), index
@@ -244,10 +244,13 @@ module {
     %vi = vector.transfer_read %i[%c0, %c0], %i0 : tensor<3x2xi32>, vector<3x2xi32>
     vector.print %vi : vector<3x2xi32>
 
+    // CHECK-NEXT: 3
+    vector.print %dl : index
+
     %d_csr = tensor.empty() : tensor<4xf64>
     %p_csr = tensor.empty() : tensor<3xi32>
     %i_csr = tensor.empty() : tensor<3xi32>
-    %rp_csr, %ri_csr, %rd_csr, %ld_csr, %lp_csr, %li_csr = sparse_tensor.disassemble %csr : tensor<2x2xf64, #CSR>
+    %rp_csr, %ri_csr, %rd_csr, %lp_csr, %li_csr, %ld_csr = sparse_tensor.disassemble %csr : tensor<2x2xf64, #CSR>
                  out_lvls(%p_csr, %i_csr : tensor<3xi32>, tensor<3xi32>)
                  out_vals(%d_csr : tensor<4xf64>)
                  -> (tensor<3xi32>, tensor<3xi32>), tensor<4xf64>, (i32, i64), index
@@ -256,6 +259,9 @@ module {
     %vd_csr = vector.transfer_read %rd_csr[%c0], %f0 : tensor<4xf64>, vector<3xf64>
     vector.print %vd_csr : vector<3xf64>
 
+    // CHECK-NEXT: 3
+    vector.print %ld_csr : index
+
     %bod = tensor.empty() : tensor<6xf64>
     %bop = tensor.empty() : tensor<4xindex>
     %boi = tensor.empty() : tensor<6x2xindex>

From 798e04f93769318db857b27f51020e7115e00301 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 10 Apr 2024 17:50:13 +0100
Subject: [PATCH 038/886] Fix MSVC "not all control paths return a value"
 warning. NFC.

---
 clang/include/clang/Basic/OpenACCKinds.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/clang/include/clang/Basic/OpenACCKinds.h b/clang/include/clang/Basic/OpenACCKinds.h
index e191e9e0a5a15..3414df9999170 100644
--- a/clang/include/clang/Basic/OpenACCKinds.h
+++ b/clang/include/clang/Basic/OpenACCKinds.h
@@ -430,6 +430,7 @@ inline StreamTy &printOpenACCDefaultClauseKind(StreamTy &Out,
   case OpenACCDefaultClauseKind::Invalid:
     return Out << "<invalid>";
   }
+  llvm_unreachable("Unknown OpenACCDefaultClauseKind enum");
 }
 
 inline const StreamingDiagnostic &operator<<(const StreamingDiagnostic &Out,

From 335d5d5f47b883055e676ffe5f981469a5f5f4f6 Mon Sep 17 00:00:00 2001
From: Vyacheslav Levytskyy <vyacheslav.levytskyy@intel.com>
Date: Wed, 10 Apr 2024 19:04:31 +0200
Subject: [PATCH 039/886] [SPIRV] Tweak parsing of base type name in builtins
 (#88255)

This PR is a small improvement of parsing of base type name in builtins,
allowing to understand `unsigned ...` types. The test case that fails
without the fix is attached.
---
 llvm/lib/Target/SPIRV/SPIRVUtils.cpp           | 16 ++++++++++++----
 llvm/test/CodeGen/SPIRV/SampledImageRetType.ll |  9 +++++++++
 2 files changed, 21 insertions(+), 4 deletions(-)

diff --git a/llvm/lib/Target/SPIRV/SPIRVUtils.cpp b/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
index c87c1293c622f..299a4341193bf 100644
--- a/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
@@ -374,13 +374,21 @@ Type *parseBasicTypeName(StringRef TypeName, LLVMContext &Ctx) {
     return Type::getVoidTy(Ctx);
   else if (TypeName.consume_front("bool"))
     return Type::getIntNTy(Ctx, 1);
-  else if (TypeName.consume_front("char") || TypeName.consume_front("uchar"))
+  else if (TypeName.consume_front("char") ||
+           TypeName.consume_front("unsigned char") ||
+           TypeName.consume_front("uchar"))
     return Type::getInt8Ty(Ctx);
-  else if (TypeName.consume_front("short") || TypeName.consume_front("ushort"))
+  else if (TypeName.consume_front("short") ||
+           TypeName.consume_front("unsigned short") ||
+           TypeName.consume_front("ushort"))
     return Type::getInt16Ty(Ctx);
-  else if (TypeName.consume_front("int") || TypeName.consume_front("uint"))
+  else if (TypeName.consume_front("int") ||
+           TypeName.consume_front("unsigned int") ||
+           TypeName.consume_front("uint"))
     return Type::getInt32Ty(Ctx);
-  else if (TypeName.consume_front("long") || TypeName.consume_front("ulong"))
+  else if (TypeName.consume_front("long") ||
+           TypeName.consume_front("unsigned long") ||
+           TypeName.consume_front("ulong"))
     return Type::getInt64Ty(Ctx);
   else if (TypeName.consume_front("half"))
     return Type::getHalfTy(Ctx);
diff --git a/llvm/test/CodeGen/SPIRV/SampledImageRetType.ll b/llvm/test/CodeGen/SPIRV/SampledImageRetType.ll
index 1aa3af83bcd20..7af5876a023e5 100644
--- a/llvm/test/CodeGen/SPIRV/SampledImageRetType.ll
+++ b/llvm/test/CodeGen/SPIRV/SampledImageRetType.ll
@@ -8,6 +8,8 @@ declare dso_local spir_func ptr addrspace(4) @_Z20__spirv_SampledImageI14ocl_ima
 
 declare dso_local spir_func <4 x float> @_Z30__spirv_ImageSampleExplicitLodIPvDv4_fiET0_T_T1_if(ptr addrspace(4) %0, i32 %1, i32 %2, float %3) local_unnamed_addr
 
+declare dso_local spir_func <4 x i32> @_Z30__spirv_ImageSampleExplicitLodI32__spirv_SampledImage__image1d_roDv4_jfET0_T_T1_if(target("spirv.SampledImage", void, 0, 0, 0, 0, 0, 0, 0) %0, float %1, i32  %2, float %3) local_unnamed_addr
+
 @__spirv_BuiltInGlobalInvocationId = external dso_local local_unnamed_addr addrspace(2) constant <3 x i64>, align 32
 
 define weak_odr dso_local spir_kernel void @_ZTS17image_kernel_readILi1EE(target("spirv.Image", void, 0, 0, 0, 0, 0, 0, 0), target("spirv.Sampler")) {
@@ -25,3 +27,10 @@ define weak_odr dso_local spir_kernel void @_ZTS17image_kernel_readILi1EE(target
 
   ret void
 }
+
+define weak_odr dso_local spir_kernel void @foo_lod(target("spirv.SampledImage", void, 0, 0, 0, 0, 0, 0, 0) %_arg) {
+  %lod = call spir_func <4 x i32> @_Z30__spirv_ImageSampleExplicitLodI32__spirv_SampledImage__image1d_roDv4_jfET0_T_T1_if(target("spirv.SampledImage", void, 0, 0, 0, 0, 0, 0, 0) %_arg, float 0x3FE7FFEB00000000, i32 2, float 0.000000e+00)
+; CHECK: %[[#sampled_image_lod:]] = OpFunctionParameter %[[#sampled_image_t]]
+; CHECK: %[[#]] = OpImageSampleExplicitLod %[[#]] %[[#sampled_image_lod]] %[[#]] {{.*}} %[[#]]
+  ret void
+}

From 4dcf33b6c2806216dfe8c5e1e3582a45516dbc69 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Wed, 10 Apr 2024 18:13:57 +0100
Subject: [PATCH 040/886] [AArch64] Cleanup and GISel coverage for lrint tests.
 NFC

---
 llvm/test/CodeGen/AArch64/lrint-conv.ll   | 66 ++++++++++++++---------
 llvm/test/CodeGen/AArch64/vector-lrint.ll | 39 ++++++++++----
 2 files changed, 70 insertions(+), 35 deletions(-)

diff --git a/llvm/test/CodeGen/AArch64/lrint-conv.ll b/llvm/test/CodeGen/AArch64/lrint-conv.ll
index 80f1e8b8fc18e..b61d6f04b400e 100644
--- a/llvm/test/CodeGen/AArch64/lrint-conv.ll
+++ b/llvm/test/CodeGen/AArch64/lrint-conv.ll
@@ -1,64 +1,78 @@
-; RUN: llc < %s -mtriple=aarch64 -mattr=+neon | FileCheck %s
-; RUN: llc < %s -global-isel -global-isel-abort=2 -pass-remarks-missed=gisel* -mtriple=aarch64 | FileCheck %s --check-prefixes=FALLBACK,CHECK
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc < %s -mtriple=aarch64 | FileCheck %s
+; RUN: llc < %s -mtriple=aarch64 -global-isel -global-isel-abort=2 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
+
+; CHECK-GI:       warning: Instruction selection used fallback path for testmswl
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for testmsll
 
-; CHECK-LABEL: testmsws:
-; CHECK:       frintx  [[REG:s[0-9]]], s0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK:       ret
-; FALLBACK-NOT: remark{{.*}}testmsws
 define i32 @testmsws(float %x) {
+; CHECK-LABEL: testmsws:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx s0, s0
+; CHECK-NEXT:    fcvtzs x0, s0
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f32(float %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsxs:
-; CHECK:       frintx  [[REG:s[0-9]]], s0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK-NEXT:  ret
-; FALLBACK-NOT: remark{{.*}}testmsxs
 define i64 @testmsxs(float %x) {
+; CHECK-LABEL: testmsxs:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx s0, s0
+; CHECK-NEXT:    fcvtzs x0, s0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f32(float %x)
   ret i64 %0
 }
 
-; CHECK-LABEL: testmswd:
-; CHECK:       frintx  [[REG:d[0-9]]], d0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK:       ret
-; FALLBACK-NOT: remark{{.*}}testmswd
 define i32 @testmswd(double %x) {
+; CHECK-LABEL: testmswd:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx d0, d0
+; CHECK-NEXT:    fcvtzs x0, d0
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f64(double %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsxd:
-; CHECK:       frintx  [[REG:d[0-9]]], d0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK-NEXT:  ret
-; FALLBACK-NOT: remark{{.*}}testmsxd
 define i64 @testmsxd(double %x) {
+; CHECK-LABEL: testmsxd:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx d0, d0
+; CHECK-NEXT:    fcvtzs x0, d0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f64(double %x)
   ret i64 %0
 }
 
-; CHECK-LABEL: testmswl:
-; CHECK:       bl      lrintl
 define dso_local i32 @testmswl(fp128 %x) {
+; CHECK-LABEL: testmswl:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -16
+; CHECK-NEXT:    bl lrintl
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f128(fp128 %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsll:
-; CHECK:       b       lrintl
 define dso_local i64 @testmsll(fp128 %x) {
+; CHECK-LABEL: testmsll:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    b lrintl
 entry:
   %0 = tail call i64 @llvm.lrint.i64.f128(fp128 %x)
   ret i64 %0
diff --git a/llvm/test/CodeGen/AArch64/vector-lrint.ll b/llvm/test/CodeGen/AArch64/vector-lrint.ll
index 9c46cf69cb0bf..b7fcd11ba8d16 100644
--- a/llvm/test/CodeGen/AArch64/vector-lrint.ll
+++ b/llvm/test/CodeGen/AArch64/vector-lrint.ll
@@ -1,6 +1,20 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=aarch64 -mattr=+neon | FileCheck %s
-; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=aarch64 -mattr=+neon | FileCheck %s
+; RUN: llc < %s -mtriple=aarch64 | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -mtriple=aarch64 -global-isel -global-isel-abort=2 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
+
+; CHECK-GI:       warning: Instruction selection used fallback path for lrint_v1f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v2f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v4f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v8f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v16i64_v16f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v32i64_v32f16
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v2f32
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v4f32
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v8f32
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v16i64_v16f32
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v2f64
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v4f64
+; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v8f64
 
 define <1 x i64> @lrint_v1f16(<1 x half> %x) {
 ; CHECK-LABEL: lrint_v1f16:
@@ -372,13 +386,20 @@ define <32 x i64> @lrint_v32i64_v32f16(<32 x half> %x) {
 declare <32 x i64> @llvm.lrint.v32i64.v32f16(<32 x half>)
 
 define <1 x i64> @lrint_v1f32(<1 x float> %x) {
-; CHECK-LABEL: lrint_v1f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    frintx s0, s0
-; CHECK-NEXT:    fcvtzs x8, s0
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: lrint_v1f32:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 def $q0
+; CHECK-SD-NEXT:    frintx s0, s0
+; CHECK-SD-NEXT:    fcvtzs x8, s0
+; CHECK-SD-NEXT:    fmov d0, x8
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: lrint_v1f32:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    frintx s0, s0
+; CHECK-GI-NEXT:    fcvtzs x8, s0
+; CHECK-GI-NEXT:    fmov d0, x8
+; CHECK-GI-NEXT:    ret
   %a = call <1 x i64> @llvm.lrint.v1i64.v1f32(<1 x float> %x)
   ret <1 x i64> %a
 }

From 04bf1a4090c535e3a1033ab9a8ef92068166461f Mon Sep 17 00:00:00 2001
From: Kojo Acquah <KoolJBlack@users.noreply.github.com>
Date: Wed, 10 Apr 2024 10:18:47 -0700
Subject: [PATCH 041/886] Update `LowerContractionToSMMLAPattern` to ingnore
 matvec (#88288)

Patterns in `LowerContractionToSMMLAPattern` are designed to handle
vector-to-matrix multiplication but not matrix-to-vector. This leads to
the following error when processing `rhs` with rank < 2:

```
iree-compile: /usr/local/google/home/kooljblack/code/iree-build/llvm-project/tools/mlir/include/mlir/IR/BuiltinTypeInterfaces.h.inc:268: int64_t mlir::detail::ShapedTypeTrait<mlir::VectorType>::getDimSize(unsigned int) const [ConcreteType = mlir::VectorType]: Assertion `idx < getRank() && "invalid index for shaped type"' failed.
```

Updates to explicitly check the rhs rank and fail cases that cannot
process.
---
 .../Transforms/LowerContractionToSMMLAPattern.cpp     |  3 +++
 mlir/test/Dialect/ArmNeon/lower-to-arm-neon.mlir      | 11 +++++++++++
 2 files changed, 14 insertions(+)

diff --git a/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractionToSMMLAPattern.cpp b/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractionToSMMLAPattern.cpp
index 13740225749e4..3ae894692089b 100644
--- a/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractionToSMMLAPattern.cpp
+++ b/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractionToSMMLAPattern.cpp
@@ -54,6 +54,9 @@ class LowerContractionToSMMLAPattern
     // Note: RHS is not transposed.
     mlir::VectorType lhsType = op.getLhsType();
     mlir::VectorType rhsType = op.getRhsType();
+    // Avoid 0-D vectors and 1-D rhs:
+    if (!lhsType.hasRank() || !rhsType.hasRank() || rhsType.getRank() < 2)
+      return failure();
     auto dimM = lhsType.getRank() == 1 ? 1 : lhsType.getDimSize(0);
     auto dimN = rhsType.getDimSize(0);
     auto dimK = rhsType.getDimSize(1);
diff --git a/mlir/test/Dialect/ArmNeon/lower-to-arm-neon.mlir b/mlir/test/Dialect/ArmNeon/lower-to-arm-neon.mlir
index 46c4026d13b66..c276a5b0c2a14 100644
--- a/mlir/test/Dialect/ArmNeon/lower-to-arm-neon.mlir
+++ b/mlir/test/Dialect/ArmNeon/lower-to-arm-neon.mlir
@@ -258,3 +258,14 @@ func.func @test_lower_vector_arm_neon_vecmat_unroll_leading_dim(%lhs: vector<1x8
   %res = vector.contract {indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d2)>, affine_map<(d0, d1, d2) -> (d1, d2)>, affine_map<(d0, d1, d2) -> (d0, d1)>], iterator_types = ["parallel", "parallel", "reduction"], kind = #vector.kind<add>} %lhs_extsi, %rhs_extsi, %acc : vector<1x8xi32>, vector<8x8xi32> into vector<1x8xi32>
   return %res : vector<1x8xi32>
 }
+
+// -----
+
+// CHECK-LABEL: func.func @test_lower_vector_arm_neon_matvec
+// CHECK-NOT: arm_neon.intr.smmla
+func.func @test_lower_vector_arm_neon_matvec(%lhs: vector<8x8xi8>, %rhs: vector<8xi8>, %acc : vector<8xi32>) -> vector<8xi32> {
+  %rhs_extsi= arith.extsi %rhs : vector<8xi8> to vector<8xi32>
+  %lhs_extsi = arith.extsi %lhs : vector<8x8xi8> to vector<8x8xi32>
+  %res = vector.contract {indexing_maps = [affine_map<(d0, d1) -> (d0, d1)>,affine_map<(d0, d1) -> (d1)>, affine_map<(d0, d1) -> (d0)>], iterator_types = ["parallel", "reduction"], kind = #vector.kind<add>} %lhs_extsi, %rhs_extsi, %acc : vector<8x8xi32>, vector<8xi32> into vector<8xi32>
+  return %res : vector<8xi32>
+}

From c54afe5c33ca6159841d909fb8fe20e5d4e0069b Mon Sep 17 00:00:00 2001
From: higher-performance
 <113926381+higher-performance@users.noreply.github.com>
Date: Wed, 10 Apr 2024 13:24:19 -0400
Subject: [PATCH 042/886] Fix quadratic slowdown in AST matcher parent map
 generation (#87824)

Avoids the need to linearly re-scan all seen parent nodes to check for
duplicates, which previously caused a slowdown for ancestry checks in
Clang AST matchers.

Fixes: #86881
---
 clang/docs/ReleaseNotes.rst        |  3 +++
 clang/lib/AST/ParentMapContext.cpp | 25 ++++++++++++++++++++++---
 2 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index f5359afe1f099..c4a4893aec5cd 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -646,6 +646,9 @@ Fixed Point Support in Clang
 AST Matchers
 ------------
 
+- Fixes a long-standing performance issue in parent map generation for
+  ancestry-based matchers such as ``hasParent`` and ``hasAncestor``, making
+  them significantly faster.
 - ``isInStdNamespace`` now supports Decl declared with ``extern "C++"``.
 - Add ``isExplicitObjectMemberFunction``.
 - Fixed ``forEachArgumentWithParam`` and ``forEachArgumentWithParamType`` to
diff --git a/clang/lib/AST/ParentMapContext.cpp b/clang/lib/AST/ParentMapContext.cpp
index 21cfd5b1de6e9..9723c0cfa83bb 100644
--- a/clang/lib/AST/ParentMapContext.cpp
+++ b/clang/lib/AST/ParentMapContext.cpp
@@ -61,7 +61,26 @@ class ParentMapContext::ParentMap {
   template <typename, typename...> friend struct ::MatchParents;
 
   /// Contains parents of a node.
-  using ParentVector = llvm::SmallVector<DynTypedNode, 2>;
+  class ParentVector {
+  public:
+    ParentVector() = default;
+    explicit ParentVector(size_t N, const DynTypedNode &Value) {
+      Items.reserve(N);
+      for (; N > 0; --N)
+        push_back(Value);
+    }
+    bool contains(const DynTypedNode &Value) {
+      return Seen.contains(Value);
+    }
+    void push_back(const DynTypedNode &Value) {
+      if (!Value.getMemoizationData() || Seen.insert(Value).second)
+        Items.push_back(Value);
+    }
+    llvm::ArrayRef<DynTypedNode> view() const { return Items; }
+  private:
+    llvm::SmallVector<DynTypedNode, 2> Items;
+    llvm::SmallDenseSet<DynTypedNode, 2> Seen;
+  };
 
   /// Maps from a node to its parents. This is used for nodes that have
   /// pointer identity only, which are more common and we can save space by
@@ -99,7 +118,7 @@ class ParentMapContext::ParentMap {
       return llvm::ArrayRef<DynTypedNode>();
     }
     if (const auto *V = I->second.template dyn_cast<ParentVector *>()) {
-      return llvm::ArrayRef(*V);
+      return V->view();
     }
     return getSingleDynTypedNodeFromParentMap(I->second);
   }
@@ -252,7 +271,7 @@ class ParentMapContext::ParentMap {
       const auto *S = It->second.dyn_cast<const Stmt *>();
       if (!S) {
         if (auto *Vec = It->second.dyn_cast<ParentVector *>())
-          return llvm::ArrayRef(*Vec);
+          return Vec->view();
         return getSingleDynTypedNodeFromParentMap(It->second);
       }
       const auto *P = dyn_cast<Expr>(S);

From f27f3697108470c3e995cf3cb454641c22ec1fa9 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 10 Apr 2024 10:28:54 -0700
Subject: [PATCH 043/886] [RISCV] Remove interrupt handler special case from
 RISCVFrameLowering::determineCalleeSaves. (#88069)

This code was trying to save temporary argument registers in interrupt
handler functions that contain calls. With the exception that all FP
registers are saved including the normally callee saved registers.

If all of the callees use an FP ABI and the interrupt handler doesn't
touch the normally callee saved FP registers, we don't need to save
them.

It doesn't appear that we need to special case functions with calls. The
normal callee saved register handling will already check each of the calls
and consider a register clobbered if the call doesn't explicitly say it is preserved.

All of the test changes are from the removal of the FP callee saved
registers. There are tests for interrupt handlers with F and D extension
that use ilp32 or lp64 ABIs that are not affected by this change. They
still save the FP callee saved registers as they should.

gcc appears to have a bug where the D extension being enabled with the
ilp32f or lp64f ABI does not save the FP callee saved regs. The callee
would only save/restore the lower 32 bits and clobber the upper bits.
LLVM saves the FP callee saved regs in this case and there is an
unchanged test for it.

The unnecessary save/restore was raised in this thread
https://discourse.llvm.org/t/has-bugs-when-optimizing-save-restore-csrs-by-changing-csr-xlen-f32-interrupt/78200/1
---
 llvm/lib/Target/RISCV/RISCVFrameLowering.cpp  |   40 -
 .../CodeGen/RISCV/interrupt-attr-nocall.ll    |  318 ++---
 llvm/test/CodeGen/RISCV/interrupt-attr.ll     | 1272 +++++++----------
 3 files changed, 675 insertions(+), 955 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
index 39075c81b2921..71672ed7b4ae7 100644
--- a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
@@ -1001,46 +1001,6 @@ void RISCVFrameLowering::determineCalleeSaves(MachineFunction &MF,
   // Mark BP as used if function has dedicated base pointer.
   if (hasBP(MF))
     SavedRegs.set(RISCVABI::getBPReg());
-
-  // If interrupt is enabled and there are calls in the handler,
-  // unconditionally save all Caller-saved registers and
-  // all FP registers, regardless whether they are used.
-  MachineFrameInfo &MFI = MF.getFrameInfo();
-  auto &Subtarget = MF.getSubtarget<RISCVSubtarget>();
-
-  if (MF.getFunction().hasFnAttribute("interrupt") && MFI.hasCalls()) {
-
-    static const MCPhysReg CSRegs[] = { RISCV::X1,      /* ra */
-      RISCV::X5, RISCV::X6, RISCV::X7,                  /* t0-t2 */
-      RISCV::X10, RISCV::X11,                           /* a0-a1, a2-a7 */
-      RISCV::X12, RISCV::X13, RISCV::X14, RISCV::X15, RISCV::X16, RISCV::X17,
-      RISCV::X28, RISCV::X29, RISCV::X30, RISCV::X31 /* t3-t6 */
-    };
-
-    for (auto Reg : CSRegs)
-      SavedRegs.set(Reg);
-
-    // According to psABI, if ilp32e/lp64e ABIs are used with an ISA that
-    // has any of the registers x16-x31 and f0-f31, then these registers are
-    // considered temporaries, so we should also save x16-x31 here.
-    if (STI.getTargetABI() == RISCVABI::ABI_ILP32E ||
-        STI.getTargetABI() == RISCVABI::ABI_LP64E) {
-      for (MCPhysReg Reg = RISCV::X16; Reg <= RISCV::X31; Reg++)
-        SavedRegs.set(Reg);
-    }
-
-    if (Subtarget.hasStdExtF()) {
-
-      // If interrupt is enabled, this list contains all FP registers.
-      const MCPhysReg * Regs = MF.getRegInfo().getCalleeSavedRegs();
-
-      for (unsigned i = 0; Regs[i]; ++i)
-        if (RISCV::FPR16RegClass.contains(Regs[i]) ||
-            RISCV::FPR32RegClass.contains(Regs[i]) ||
-            RISCV::FPR64RegClass.contains(Regs[i]))
-          SavedRegs.set(Regs[i]);
-    }
-  }
 }
 
 std::pair<int64_t, Align>
diff --git a/llvm/test/CodeGen/RISCV/interrupt-attr-nocall.ll b/llvm/test/CodeGen/RISCV/interrupt-attr-nocall.ll
index 263743d39a8e6..fa6ac96b57b1e 100644
--- a/llvm/test/CodeGen/RISCV/interrupt-attr-nocall.ll
+++ b/llvm/test/CodeGen/RISCV/interrupt-attr-nocall.ll
@@ -412,51 +412,39 @@ define void @foo_double() nounwind #0 {
 ;
 ; CHECK-RV32IF-LABEL: foo_double:
 ; CHECK-RV32IF:       # %bb.0:
-; CHECK-RV32IF-NEXT:    addi sp, sp, -192
-; CHECK-RV32IF-NEXT:    sw ra, 188(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t0, 184(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t1, 180(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t2, 176(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a0, 172(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a1, 168(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a2, 164(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a3, 160(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a4, 156(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a5, 152(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a6, 148(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a7, 144(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t3, 140(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t4, 136(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t5, 132(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t6, 128(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft0, 124(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft1, 120(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft2, 116(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft3, 112(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft4, 108(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft5, 104(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft6, 100(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft7, 96(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs0, 92(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs1, 88(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa0, 84(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa1, 80(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa2, 76(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa3, 72(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa4, 68(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa5, 64(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa6, 60(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa7, 56(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs2, 52(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs3, 48(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs4, 44(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs5, 40(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs6, 36(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs7, 32(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs8, 28(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs9, 24(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs10, 20(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs11, 16(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    addi sp, sp, -144
+; CHECK-RV32IF-NEXT:    sw ra, 140(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t0, 136(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t1, 132(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t2, 128(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a0, 124(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a1, 120(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a2, 116(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a3, 112(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a4, 108(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a5, 104(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a6, 100(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a7, 96(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t3, 92(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t4, 88(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t5, 84(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t6, 80(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft0, 76(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft1, 72(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft2, 68(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft3, 64(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft4, 60(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft5, 56(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft6, 52(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft7, 48(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa0, 44(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa1, 40(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa2, 36(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa3, 32(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa4, 28(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa5, 24(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa6, 20(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa7, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft8, 12(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft9, 8(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft10, 4(sp) # 4-byte Folded Spill
@@ -471,55 +459,43 @@ define void @foo_double() nounwind #0 {
 ; CHECK-RV32IF-NEXT:    lui a2, %hi(g)
 ; CHECK-RV32IF-NEXT:    sw a1, %lo(g+4)(a2)
 ; CHECK-RV32IF-NEXT:    sw a0, %lo(g)(a2)
-; CHECK-RV32IF-NEXT:    lw ra, 188(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t0, 184(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t1, 180(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t2, 176(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a0, 172(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a1, 168(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a2, 164(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a3, 160(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a4, 156(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a5, 152(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a6, 148(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a7, 144(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t3, 140(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t4, 136(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t5, 132(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t6, 128(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft0, 124(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft1, 120(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft2, 116(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft3, 112(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft4, 108(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft5, 104(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft6, 100(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft7, 96(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs0, 92(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs1, 88(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa0, 84(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa1, 80(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa2, 76(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa3, 72(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa4, 68(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa5, 64(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa6, 60(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa7, 56(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs2, 52(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs3, 48(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs4, 44(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs5, 40(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs6, 36(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs7, 32(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs8, 28(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs9, 24(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs10, 20(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs11, 16(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw ra, 140(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t0, 136(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t1, 132(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t2, 128(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a0, 124(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a1, 120(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a2, 116(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a3, 112(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a4, 108(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a5, 104(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a6, 100(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a7, 96(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t3, 92(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t4, 88(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t5, 84(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t6, 80(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft0, 76(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft1, 72(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft2, 68(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft3, 64(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft4, 60(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft5, 56(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft6, 52(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft7, 48(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa0, 44(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa1, 40(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa2, 36(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa3, 32(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa4, 28(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa5, 24(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa6, 20(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa7, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft8, 12(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft9, 8(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft10, 4(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft11, 0(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    addi sp, sp, 192
+; CHECK-RV32IF-NEXT:    addi sp, sp, 144
 ; CHECK-RV32IF-NEXT:    mret
 ;
 ; CHECK-RV32IFD-LABEL: foo_double:
@@ -604,57 +580,45 @@ define void @foo_fp_double() nounwind #1 {
 ;
 ; CHECK-RV32IF-LABEL: foo_fp_double:
 ; CHECK-RV32IF:       # %bb.0:
-; CHECK-RV32IF-NEXT:    addi sp, sp, -208
-; CHECK-RV32IF-NEXT:    sw ra, 204(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t0, 200(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t1, 196(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t2, 192(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw s0, 188(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a0, 184(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a1, 180(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a2, 176(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a3, 172(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a4, 168(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a5, 164(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a6, 160(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw a7, 156(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t3, 152(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t4, 148(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t5, 144(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    sw t6, 140(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft0, 136(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft1, 132(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft2, 128(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft3, 124(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft4, 120(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft5, 116(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft6, 112(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw ft7, 108(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs0, 104(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs1, 100(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa0, 96(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa1, 92(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa2, 88(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa3, 84(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa4, 80(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa5, 76(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa6, 72(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fa7, 68(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs2, 64(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs3, 60(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs4, 56(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs5, 52(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs6, 48(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs7, 44(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs8, 40(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs9, 36(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs10, 32(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    fsw fs11, 28(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    addi sp, sp, -160
+; CHECK-RV32IF-NEXT:    sw ra, 156(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t0, 152(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t1, 148(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t2, 144(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw s0, 140(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a0, 136(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a1, 132(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a2, 128(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a3, 124(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a4, 120(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a5, 116(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a6, 112(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw a7, 108(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t3, 104(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t4, 100(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t5, 96(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    sw t6, 92(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft0, 88(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft1, 84(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft2, 80(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft3, 76(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft4, 72(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft5, 68(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft6, 64(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw ft7, 60(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa0, 56(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa1, 52(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa2, 48(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa3, 44(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa4, 40(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa5, 36(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa6, 32(sp) # 4-byte Folded Spill
+; CHECK-RV32IF-NEXT:    fsw fa7, 28(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft8, 24(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft9, 20(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft10, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV32IF-NEXT:    fsw ft11, 12(sp) # 4-byte Folded Spill
-; CHECK-RV32IF-NEXT:    addi s0, sp, 208
+; CHECK-RV32IF-NEXT:    addi s0, sp, 160
 ; CHECK-RV32IF-NEXT:    lui a1, %hi(h)
 ; CHECK-RV32IF-NEXT:    lw a0, %lo(h)(a1)
 ; CHECK-RV32IF-NEXT:    lw a1, %lo(h+4)(a1)
@@ -665,56 +629,44 @@ define void @foo_fp_double() nounwind #1 {
 ; CHECK-RV32IF-NEXT:    lui a2, %hi(g)
 ; CHECK-RV32IF-NEXT:    sw a1, %lo(g+4)(a2)
 ; CHECK-RV32IF-NEXT:    sw a0, %lo(g)(a2)
-; CHECK-RV32IF-NEXT:    lw ra, 204(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t0, 200(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t1, 196(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t2, 192(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw s0, 188(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a0, 184(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a1, 180(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a2, 176(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a3, 172(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a4, 168(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a5, 164(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a6, 160(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw a7, 156(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t3, 152(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t4, 148(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t5, 144(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    lw t6, 140(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft0, 136(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft1, 132(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft2, 128(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft3, 124(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft4, 120(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft5, 116(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft6, 112(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw ft7, 108(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs0, 104(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs1, 100(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa0, 96(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa1, 92(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa2, 88(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa3, 84(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa4, 80(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa5, 76(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa6, 72(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fa7, 68(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs2, 64(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs3, 60(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs4, 56(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs5, 52(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs6, 48(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs7, 44(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs8, 40(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs9, 36(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs10, 32(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    flw fs11, 28(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw ra, 156(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t0, 152(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t1, 148(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t2, 144(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw s0, 140(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a0, 136(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a1, 132(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a2, 128(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a3, 124(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a4, 120(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a5, 116(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a6, 112(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw a7, 108(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t3, 104(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t4, 100(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t5, 96(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    lw t6, 92(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft0, 88(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft1, 84(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft2, 80(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft3, 76(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft4, 72(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft5, 68(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft6, 64(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw ft7, 60(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa0, 56(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa1, 52(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa2, 48(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa3, 44(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa4, 40(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa5, 36(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa6, 32(sp) # 4-byte Folded Reload
+; CHECK-RV32IF-NEXT:    flw fa7, 28(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft8, 24(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft9, 20(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft10, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV32IF-NEXT:    flw ft11, 12(sp) # 4-byte Folded Reload
-; CHECK-RV32IF-NEXT:    addi sp, sp, 208
+; CHECK-RV32IF-NEXT:    addi sp, sp, 160
 ; CHECK-RV32IF-NEXT:    mret
 ;
 ; CHECK-RV32IFD-LABEL: foo_fp_double:
diff --git a/llvm/test/CodeGen/RISCV/interrupt-attr.ll b/llvm/test/CodeGen/RISCV/interrupt-attr.ll
index 50c789f8f86dc..739c9d8d0b0ac 100644
--- a/llvm/test/CodeGen/RISCV/interrupt-attr.ll
+++ b/llvm/test/CodeGen/RISCV/interrupt-attr.ll
@@ -115,208 +115,160 @@ define void @foo_with_call() #1 {
 ;
 ; CHECK-RV32-F-LABEL: foo_with_call:
 ; CHECK-RV32-F:       # %bb.0:
-; CHECK-RV32-F-NEXT:    addi sp, sp, -192
-; CHECK-RV32-F-NEXT:    sw ra, 188(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t0, 184(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t1, 180(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t2, 176(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a0, 172(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a1, 168(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a2, 164(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a3, 160(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a4, 156(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a5, 152(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a6, 148(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a7, 144(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t3, 140(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t4, 136(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t5, 132(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t6, 128(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft0, 124(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft1, 120(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft2, 116(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft3, 112(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft4, 108(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft5, 104(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft6, 100(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft7, 96(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs0, 92(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs1, 88(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa0, 84(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa1, 80(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa2, 76(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa3, 72(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa4, 68(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa5, 64(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa6, 60(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa7, 56(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs2, 52(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs3, 48(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs4, 44(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs5, 40(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs6, 36(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs7, 32(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs8, 28(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs9, 24(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs10, 20(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs11, 16(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    addi sp, sp, -144
+; CHECK-RV32-F-NEXT:    sw ra, 140(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t0, 136(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t1, 132(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t2, 128(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a0, 124(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a1, 120(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a2, 116(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a3, 112(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a4, 108(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a5, 104(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a6, 100(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a7, 96(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t3, 92(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t4, 88(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t5, 84(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t6, 80(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft0, 76(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft1, 72(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft2, 68(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft3, 64(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft4, 60(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft5, 56(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft6, 52(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft7, 48(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa0, 44(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa1, 40(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa2, 36(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa3, 32(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa4, 28(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa5, 24(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa6, 20(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa7, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft8, 12(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft9, 8(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft10, 4(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft11, 0(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    call otherfoo
-; CHECK-RV32-F-NEXT:    lw ra, 188(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t0, 184(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t1, 180(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t2, 176(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a0, 172(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a1, 168(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a2, 164(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a3, 160(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a4, 156(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a5, 152(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a6, 148(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a7, 144(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t3, 140(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t4, 136(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t5, 132(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t6, 128(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft0, 124(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft1, 120(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft2, 116(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft3, 112(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft4, 108(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft5, 104(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft6, 100(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft7, 96(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs0, 92(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs1, 88(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa0, 84(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa1, 80(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa2, 76(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa3, 72(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa4, 68(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa5, 64(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa6, 60(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa7, 56(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs2, 52(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs3, 48(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs4, 44(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs5, 40(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs6, 36(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs7, 32(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs8, 28(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs9, 24(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs10, 20(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs11, 16(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw ra, 140(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t0, 136(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t1, 132(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t2, 128(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a0, 124(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a1, 120(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a2, 116(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a3, 112(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a4, 108(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a5, 104(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a6, 100(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a7, 96(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t3, 92(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t4, 88(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t5, 84(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t6, 80(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft0, 76(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft1, 72(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft2, 68(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft3, 64(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft4, 60(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft5, 56(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft6, 52(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft7, 48(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa0, 44(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa1, 40(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa2, 36(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa3, 32(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa4, 28(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa5, 24(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa6, 20(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa7, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft8, 12(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft9, 8(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft10, 4(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft11, 0(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    addi sp, sp, 192
+; CHECK-RV32-F-NEXT:    addi sp, sp, 144
 ; CHECK-RV32-F-NEXT:    mret
 ;
 ; CHECK-RV32-FD-LABEL: foo_with_call:
 ; CHECK-RV32-FD:       # %bb.0:
-; CHECK-RV32-FD-NEXT:    addi sp, sp, -320
-; CHECK-RV32-FD-NEXT:    sw ra, 316(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t0, 312(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t1, 308(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t2, 304(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a0, 300(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a1, 296(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a2, 292(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a3, 288(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a4, 284(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a5, 280(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a6, 276(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a7, 272(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t3, 268(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t4, 264(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t5, 260(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t6, 256(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft0, 248(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft1, 240(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft2, 232(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft3, 224(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft4, 216(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft5, 208(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft6, 200(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft7, 192(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs0, 184(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs1, 176(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa0, 168(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa1, 160(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa2, 152(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa3, 144(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa4, 136(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa5, 128(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa6, 120(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa7, 112(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs2, 104(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs3, 96(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs4, 88(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs5, 80(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs6, 72(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs7, 64(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs8, 56(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs9, 48(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs10, 40(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs11, 32(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    addi sp, sp, -224
+; CHECK-RV32-FD-NEXT:    sw ra, 220(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t0, 216(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t1, 212(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t2, 208(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a0, 204(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a1, 200(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a2, 196(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a3, 192(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a4, 188(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a5, 184(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a6, 180(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a7, 176(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t3, 172(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t4, 168(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t5, 164(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t6, 160(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft0, 152(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft1, 144(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft2, 136(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft3, 128(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft4, 120(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft5, 112(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft6, 104(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft7, 96(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa0, 88(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa1, 80(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa2, 72(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa3, 64(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa4, 56(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa5, 48(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa6, 40(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa7, 32(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft8, 24(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft9, 16(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft10, 8(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft11, 0(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    call otherfoo
-; CHECK-RV32-FD-NEXT:    lw ra, 316(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t0, 312(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t1, 308(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t2, 304(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a0, 300(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a1, 296(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a2, 292(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a3, 288(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a4, 284(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a5, 280(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a6, 276(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a7, 272(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t3, 268(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t4, 264(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t5, 260(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t6, 256(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft0, 248(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft1, 240(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft2, 232(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft3, 224(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft4, 216(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft5, 208(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft6, 200(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft7, 192(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs0, 184(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs1, 176(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa0, 168(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa1, 160(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa2, 152(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa3, 144(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa4, 136(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa5, 128(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa6, 120(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa7, 112(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs2, 104(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs3, 96(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs4, 88(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs5, 80(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs6, 72(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs7, 64(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs8, 56(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs9, 48(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs10, 40(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs11, 32(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw ra, 220(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t0, 216(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t1, 212(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t2, 208(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a0, 204(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a1, 200(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a2, 196(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a3, 192(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a4, 188(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a5, 184(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a6, 180(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a7, 176(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t3, 172(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t4, 168(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t5, 164(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t6, 160(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft0, 152(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft1, 144(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft2, 136(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft3, 128(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft4, 120(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft5, 112(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft6, 104(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft7, 96(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa0, 88(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa1, 80(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa2, 72(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa3, 64(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa4, 56(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa5, 48(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa6, 40(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa7, 32(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft8, 24(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft9, 16(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft10, 8(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft11, 0(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    addi sp, sp, 320
+; CHECK-RV32-FD-NEXT:    addi sp, sp, 224
 ; CHECK-RV32-FD-NEXT:    mret
 ;
 ; CHECK-RV32-F-ILP3-LABEL: foo_with_call:
@@ -846,208 +798,160 @@ define void @foo_with_call() #1 {
 ;
 ; CHECK-RV64-F-LABEL: foo_with_call:
 ; CHECK-RV64-F:       # %bb.0:
-; CHECK-RV64-F-NEXT:    addi sp, sp, -256
-; CHECK-RV64-F-NEXT:    sd ra, 248(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t0, 240(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t1, 232(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t2, 224(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a0, 216(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a1, 208(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a2, 200(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a3, 192(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a4, 184(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a5, 176(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a6, 168(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a7, 160(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t3, 152(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t4, 144(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t5, 136(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t6, 128(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft0, 124(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft1, 120(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft2, 116(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft3, 112(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft4, 108(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft5, 104(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft6, 100(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft7, 96(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs0, 92(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs1, 88(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa0, 84(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa1, 80(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa2, 76(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa3, 72(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa4, 68(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa5, 64(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa6, 60(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa7, 56(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs2, 52(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs3, 48(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs4, 44(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs5, 40(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs6, 36(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs7, 32(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs8, 28(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs9, 24(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs10, 20(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs11, 16(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    addi sp, sp, -208
+; CHECK-RV64-F-NEXT:    sd ra, 200(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t0, 192(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t1, 184(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t2, 176(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a0, 168(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a1, 160(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a2, 152(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a3, 144(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a4, 136(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a5, 128(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a6, 120(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a7, 112(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t3, 104(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t4, 96(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t5, 88(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t6, 80(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft0, 76(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft1, 72(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft2, 68(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft3, 64(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft4, 60(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft5, 56(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft6, 52(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft7, 48(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa0, 44(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa1, 40(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa2, 36(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa3, 32(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa4, 28(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa5, 24(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa6, 20(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa7, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft8, 12(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft9, 8(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft10, 4(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft11, 0(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    call otherfoo
-; CHECK-RV64-F-NEXT:    ld ra, 248(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t0, 240(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t1, 232(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t2, 224(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a0, 216(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a1, 208(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a2, 200(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a3, 192(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a4, 184(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a5, 176(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a6, 168(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a7, 160(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t3, 152(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t4, 144(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t5, 136(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t6, 128(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft0, 124(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft1, 120(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft2, 116(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft3, 112(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft4, 108(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft5, 104(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft6, 100(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft7, 96(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs0, 92(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs1, 88(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa0, 84(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa1, 80(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa2, 76(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa3, 72(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa4, 68(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa5, 64(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa6, 60(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa7, 56(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs2, 52(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs3, 48(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs4, 44(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs5, 40(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs6, 36(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs7, 32(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs8, 28(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs9, 24(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs10, 20(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs11, 16(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld ra, 200(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t0, 192(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t1, 184(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t2, 176(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a0, 168(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a1, 160(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a2, 152(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a3, 144(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a4, 136(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a5, 128(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a6, 120(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a7, 112(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t3, 104(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t4, 96(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t5, 88(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t6, 80(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft0, 76(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft1, 72(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft2, 68(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft3, 64(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft4, 60(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft5, 56(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft6, 52(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft7, 48(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa0, 44(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa1, 40(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa2, 36(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa3, 32(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa4, 28(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa5, 24(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa6, 20(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa7, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft8, 12(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft9, 8(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft10, 4(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft11, 0(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    addi sp, sp, 256
+; CHECK-RV64-F-NEXT:    addi sp, sp, 208
 ; CHECK-RV64-F-NEXT:    mret
 ;
 ; CHECK-RV64-FD-LABEL: foo_with_call:
 ; CHECK-RV64-FD:       # %bb.0:
-; CHECK-RV64-FD-NEXT:    addi sp, sp, -384
-; CHECK-RV64-FD-NEXT:    sd ra, 376(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t0, 368(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t1, 360(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t2, 352(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a0, 344(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a1, 336(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a2, 328(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a3, 320(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a4, 312(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a5, 304(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a6, 296(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a7, 288(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t3, 280(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t4, 272(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t5, 264(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t6, 256(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft0, 248(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft1, 240(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft2, 232(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft3, 224(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft4, 216(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft5, 208(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft6, 200(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft7, 192(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs0, 184(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs1, 176(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa0, 168(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa1, 160(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa2, 152(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa3, 144(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa4, 136(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa5, 128(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa6, 120(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa7, 112(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs2, 104(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs3, 96(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs4, 88(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs5, 80(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs6, 72(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs7, 64(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs8, 56(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs9, 48(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs10, 40(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs11, 32(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    addi sp, sp, -288
+; CHECK-RV64-FD-NEXT:    sd ra, 280(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t0, 272(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t1, 264(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t2, 256(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a0, 248(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a1, 240(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a2, 232(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a3, 224(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a4, 216(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a5, 208(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a6, 200(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a7, 192(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t3, 184(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t4, 176(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t5, 168(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t6, 160(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft0, 152(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft1, 144(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft2, 136(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft3, 128(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft4, 120(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft5, 112(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft6, 104(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft7, 96(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa0, 88(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa1, 80(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa2, 72(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa3, 64(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa4, 56(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa5, 48(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa6, 40(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa7, 32(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft8, 24(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft9, 16(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft10, 8(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft11, 0(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    call otherfoo
-; CHECK-RV64-FD-NEXT:    ld ra, 376(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t0, 368(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t1, 360(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t2, 352(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a0, 344(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a1, 336(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a2, 328(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a3, 320(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a4, 312(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a5, 304(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a6, 296(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a7, 288(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t3, 280(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t4, 272(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t5, 264(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t6, 256(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft0, 248(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft1, 240(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft2, 232(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft3, 224(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft4, 216(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft5, 208(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft6, 200(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft7, 192(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs0, 184(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs1, 176(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa0, 168(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa1, 160(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa2, 152(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa3, 144(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa4, 136(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa5, 128(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa6, 120(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa7, 112(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs2, 104(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs3, 96(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs4, 88(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs5, 80(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs6, 72(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs7, 64(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs8, 56(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs9, 48(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs10, 40(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs11, 32(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld ra, 280(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t0, 272(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t1, 264(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t2, 256(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a0, 248(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a1, 240(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a2, 232(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a3, 224(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a4, 216(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a5, 208(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a6, 200(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a7, 192(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t3, 184(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t4, 176(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t5, 168(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t6, 160(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft0, 152(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft1, 144(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft2, 136(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft3, 128(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft4, 120(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft5, 112(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft6, 104(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft7, 96(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa0, 88(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa1, 80(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa2, 72(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa3, 64(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa4, 56(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa5, 48(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa6, 40(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa7, 32(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft8, 24(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft9, 16(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft10, 8(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft11, 0(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    addi sp, sp, 384
+; CHECK-RV64-FD-NEXT:    addi sp, sp, 288
 ; CHECK-RV64-FD-NEXT:    mret
 ;
 ; CHECK-RV64-F-LP64-LABEL: foo_with_call:
@@ -1711,214 +1615,166 @@ define void @foo_fp_with_call() #2 {
 ;
 ; CHECK-RV32-F-LABEL: foo_fp_with_call:
 ; CHECK-RV32-F:       # %bb.0:
-; CHECK-RV32-F-NEXT:    addi sp, sp, -208
-; CHECK-RV32-F-NEXT:    sw ra, 204(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t0, 200(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t1, 196(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t2, 192(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw s0, 188(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a0, 184(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a1, 180(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a2, 176(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a3, 172(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a4, 168(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a5, 164(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a6, 160(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw a7, 156(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t3, 152(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t4, 148(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t5, 144(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    sw t6, 140(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft0, 136(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft1, 132(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft2, 128(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft3, 124(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft4, 120(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft5, 116(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft6, 112(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw ft7, 108(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs0, 104(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs1, 100(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa0, 96(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa1, 92(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa2, 88(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa3, 84(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa4, 80(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa5, 76(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa6, 72(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fa7, 68(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs2, 64(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs3, 60(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs4, 56(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs5, 52(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs6, 48(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs7, 44(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs8, 40(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs9, 36(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs10, 32(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    fsw fs11, 28(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    addi sp, sp, -160
+; CHECK-RV32-F-NEXT:    sw ra, 156(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t0, 152(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t1, 148(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t2, 144(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw s0, 140(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a0, 136(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a1, 132(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a2, 128(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a3, 124(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a4, 120(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a5, 116(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a6, 112(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw a7, 108(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t3, 104(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t4, 100(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t5, 96(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    sw t6, 92(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft0, 88(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft1, 84(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft2, 80(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft3, 76(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft4, 72(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft5, 68(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft6, 64(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw ft7, 60(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa0, 56(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa1, 52(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa2, 48(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa3, 44(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa4, 40(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa5, 36(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa6, 32(sp) # 4-byte Folded Spill
+; CHECK-RV32-F-NEXT:    fsw fa7, 28(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft8, 24(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft9, 20(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft10, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV32-F-NEXT:    fsw ft11, 12(sp) # 4-byte Folded Spill
-; CHECK-RV32-F-NEXT:    addi s0, sp, 208
+; CHECK-RV32-F-NEXT:    addi s0, sp, 160
 ; CHECK-RV32-F-NEXT:    call otherfoo
-; CHECK-RV32-F-NEXT:    lw ra, 204(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t0, 200(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t1, 196(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t2, 192(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw s0, 188(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a0, 184(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a1, 180(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a2, 176(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a3, 172(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a4, 168(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a5, 164(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a6, 160(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw a7, 156(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t3, 152(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t4, 148(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t5, 144(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    lw t6, 140(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft0, 136(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft1, 132(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft2, 128(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft3, 124(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft4, 120(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft5, 116(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft6, 112(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw ft7, 108(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs0, 104(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs1, 100(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa0, 96(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa1, 92(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa2, 88(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa3, 84(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa4, 80(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa5, 76(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa6, 72(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fa7, 68(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs2, 64(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs3, 60(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs4, 56(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs5, 52(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs6, 48(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs7, 44(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs8, 40(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs9, 36(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs10, 32(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    flw fs11, 28(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw ra, 156(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t0, 152(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t1, 148(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t2, 144(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw s0, 140(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a0, 136(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a1, 132(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a2, 128(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a3, 124(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a4, 120(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a5, 116(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a6, 112(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw a7, 108(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t3, 104(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t4, 100(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t5, 96(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    lw t6, 92(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft0, 88(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft1, 84(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft2, 80(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft3, 76(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft4, 72(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft5, 68(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft6, 64(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw ft7, 60(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa0, 56(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa1, 52(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa2, 48(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa3, 44(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa4, 40(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa5, 36(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa6, 32(sp) # 4-byte Folded Reload
+; CHECK-RV32-F-NEXT:    flw fa7, 28(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft8, 24(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft9, 20(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft10, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV32-F-NEXT:    flw ft11, 12(sp) # 4-byte Folded Reload
-; CHECK-RV32-F-NEXT:    addi sp, sp, 208
+; CHECK-RV32-F-NEXT:    addi sp, sp, 160
 ; CHECK-RV32-F-NEXT:    mret
 ;
 ; CHECK-RV32-FD-LABEL: foo_fp_with_call:
 ; CHECK-RV32-FD:       # %bb.0:
-; CHECK-RV32-FD-NEXT:    addi sp, sp, -336
-; CHECK-RV32-FD-NEXT:    sw ra, 332(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t0, 328(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t1, 324(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t2, 320(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw s0, 316(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a0, 312(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a1, 308(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a2, 304(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a3, 300(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a4, 296(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a5, 292(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a6, 288(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw a7, 284(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t3, 280(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t4, 276(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t5, 272(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    sw t6, 268(sp) # 4-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft0, 256(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft1, 248(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft2, 240(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft3, 232(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft4, 224(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft5, 216(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft6, 208(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd ft7, 200(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs0, 192(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs1, 184(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa0, 176(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa1, 168(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa2, 160(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa3, 152(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa4, 144(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa5, 136(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa6, 128(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fa7, 120(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs2, 112(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs3, 104(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs4, 96(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs5, 88(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs6, 80(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs7, 72(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs8, 64(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs9, 56(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs10, 48(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    fsd fs11, 40(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    addi sp, sp, -240
+; CHECK-RV32-FD-NEXT:    sw ra, 236(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t0, 232(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t1, 228(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t2, 224(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw s0, 220(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a0, 216(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a1, 212(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a2, 208(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a3, 204(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a4, 200(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a5, 196(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a6, 192(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw a7, 188(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t3, 184(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t4, 180(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t5, 176(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    sw t6, 172(sp) # 4-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft0, 160(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft1, 152(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft2, 144(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft3, 136(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft4, 128(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft5, 120(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft6, 112(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd ft7, 104(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa0, 96(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa1, 88(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa2, 80(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa3, 72(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa4, 64(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa5, 56(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa6, 48(sp) # 8-byte Folded Spill
+; CHECK-RV32-FD-NEXT:    fsd fa7, 40(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft8, 32(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft9, 24(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft10, 16(sp) # 8-byte Folded Spill
 ; CHECK-RV32-FD-NEXT:    fsd ft11, 8(sp) # 8-byte Folded Spill
-; CHECK-RV32-FD-NEXT:    addi s0, sp, 336
+; CHECK-RV32-FD-NEXT:    addi s0, sp, 240
 ; CHECK-RV32-FD-NEXT:    call otherfoo
-; CHECK-RV32-FD-NEXT:    lw ra, 332(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t0, 328(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t1, 324(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t2, 320(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw s0, 316(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a0, 312(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a1, 308(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a2, 304(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a3, 300(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a4, 296(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a5, 292(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a6, 288(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw a7, 284(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t3, 280(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t4, 276(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t5, 272(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    lw t6, 268(sp) # 4-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft0, 256(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft1, 248(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft2, 240(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft3, 232(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft4, 224(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft5, 216(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft6, 208(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld ft7, 200(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs0, 192(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs1, 184(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa0, 176(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa1, 168(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa2, 160(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa3, 152(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa4, 144(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa5, 136(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa6, 128(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fa7, 120(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs2, 112(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs3, 104(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs4, 96(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs5, 88(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs6, 80(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs7, 72(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs8, 64(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs9, 56(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs10, 48(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    fld fs11, 40(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw ra, 236(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t0, 232(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t1, 228(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t2, 224(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw s0, 220(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a0, 216(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a1, 212(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a2, 208(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a3, 204(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a4, 200(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a5, 196(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a6, 192(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw a7, 188(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t3, 184(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t4, 180(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t5, 176(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    lw t6, 172(sp) # 4-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft0, 160(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft1, 152(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft2, 144(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft3, 136(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft4, 128(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft5, 120(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft6, 112(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld ft7, 104(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa0, 96(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa1, 88(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa2, 80(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa3, 72(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa4, 64(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa5, 56(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa6, 48(sp) # 8-byte Folded Reload
+; CHECK-RV32-FD-NEXT:    fld fa7, 40(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft8, 32(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft9, 24(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft10, 16(sp) # 8-byte Folded Reload
 ; CHECK-RV32-FD-NEXT:    fld ft11, 8(sp) # 8-byte Folded Reload
-; CHECK-RV32-FD-NEXT:    addi sp, sp, 336
+; CHECK-RV32-FD-NEXT:    addi sp, sp, 240
 ; CHECK-RV32-FD-NEXT:    mret
 ;
 ; CHECK-RV32-F-ILP3-LABEL: foo_fp_with_call:
@@ -2469,214 +2325,166 @@ define void @foo_fp_with_call() #2 {
 ;
 ; CHECK-RV64-F-LABEL: foo_fp_with_call:
 ; CHECK-RV64-F:       # %bb.0:
-; CHECK-RV64-F-NEXT:    addi sp, sp, -272
-; CHECK-RV64-F-NEXT:    sd ra, 264(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t0, 256(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t1, 248(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t2, 240(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd s0, 232(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a0, 224(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a1, 216(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a2, 208(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a3, 200(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a4, 192(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a5, 184(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a6, 176(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd a7, 168(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t3, 160(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t4, 152(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t5, 144(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    sd t6, 136(sp) # 8-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft0, 132(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft1, 128(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft2, 124(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft3, 120(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft4, 116(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft5, 112(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft6, 108(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw ft7, 104(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs0, 100(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs1, 96(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa0, 92(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa1, 88(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa2, 84(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa3, 80(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa4, 76(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa5, 72(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa6, 68(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fa7, 64(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs2, 60(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs3, 56(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs4, 52(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs5, 48(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs6, 44(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs7, 40(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs8, 36(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs9, 32(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs10, 28(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    fsw fs11, 24(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    addi sp, sp, -224
+; CHECK-RV64-F-NEXT:    sd ra, 216(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t0, 208(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t1, 200(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t2, 192(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd s0, 184(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a0, 176(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a1, 168(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a2, 160(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a3, 152(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a4, 144(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a5, 136(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a6, 128(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd a7, 120(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t3, 112(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t4, 104(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t5, 96(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    sd t6, 88(sp) # 8-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft0, 84(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft1, 80(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft2, 76(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft3, 72(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft4, 68(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft5, 64(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft6, 60(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw ft7, 56(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa0, 52(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa1, 48(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa2, 44(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa3, 40(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa4, 36(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa5, 32(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa6, 28(sp) # 4-byte Folded Spill
+; CHECK-RV64-F-NEXT:    fsw fa7, 24(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft8, 20(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft9, 16(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft10, 12(sp) # 4-byte Folded Spill
 ; CHECK-RV64-F-NEXT:    fsw ft11, 8(sp) # 4-byte Folded Spill
-; CHECK-RV64-F-NEXT:    addi s0, sp, 272
+; CHECK-RV64-F-NEXT:    addi s0, sp, 224
 ; CHECK-RV64-F-NEXT:    call otherfoo
-; CHECK-RV64-F-NEXT:    ld ra, 264(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t0, 256(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t1, 248(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t2, 240(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld s0, 232(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a0, 224(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a1, 216(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a2, 208(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a3, 200(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a4, 192(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a5, 184(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a6, 176(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld a7, 168(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t3, 160(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t4, 152(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t5, 144(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    ld t6, 136(sp) # 8-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft0, 132(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft1, 128(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft2, 124(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft3, 120(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft4, 116(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft5, 112(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft6, 108(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw ft7, 104(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs0, 100(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs1, 96(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa0, 92(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa1, 88(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa2, 84(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa3, 80(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa4, 76(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa5, 72(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa6, 68(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fa7, 64(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs2, 60(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs3, 56(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs4, 52(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs5, 48(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs6, 44(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs7, 40(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs8, 36(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs9, 32(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs10, 28(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    flw fs11, 24(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld ra, 216(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t0, 208(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t1, 200(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t2, 192(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld s0, 184(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a0, 176(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a1, 168(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a2, 160(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a3, 152(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a4, 144(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a5, 136(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a6, 128(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld a7, 120(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t3, 112(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t4, 104(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t5, 96(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    ld t6, 88(sp) # 8-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft0, 84(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft1, 80(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft2, 76(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft3, 72(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft4, 68(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft5, 64(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft6, 60(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw ft7, 56(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa0, 52(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa1, 48(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa2, 44(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa3, 40(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa4, 36(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa5, 32(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa6, 28(sp) # 4-byte Folded Reload
+; CHECK-RV64-F-NEXT:    flw fa7, 24(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft8, 20(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft9, 16(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft10, 12(sp) # 4-byte Folded Reload
 ; CHECK-RV64-F-NEXT:    flw ft11, 8(sp) # 4-byte Folded Reload
-; CHECK-RV64-F-NEXT:    addi sp, sp, 272
+; CHECK-RV64-F-NEXT:    addi sp, sp, 224
 ; CHECK-RV64-F-NEXT:    mret
 ;
 ; CHECK-RV64-FD-LABEL: foo_fp_with_call:
 ; CHECK-RV64-FD:       # %bb.0:
-; CHECK-RV64-FD-NEXT:    addi sp, sp, -400
-; CHECK-RV64-FD-NEXT:    sd ra, 392(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t0, 384(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t1, 376(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t2, 368(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd s0, 360(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a0, 352(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a1, 344(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a2, 336(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a3, 328(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a4, 320(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a5, 312(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a6, 304(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd a7, 296(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t3, 288(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t4, 280(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t5, 272(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    sd t6, 264(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft0, 256(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft1, 248(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft2, 240(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft3, 232(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft4, 224(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft5, 216(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft6, 208(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd ft7, 200(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs0, 192(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs1, 184(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa0, 176(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa1, 168(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa2, 160(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa3, 152(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa4, 144(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa5, 136(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa6, 128(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fa7, 120(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs2, 112(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs3, 104(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs4, 96(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs5, 88(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs6, 80(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs7, 72(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs8, 64(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs9, 56(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs10, 48(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    fsd fs11, 40(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    addi sp, sp, -304
+; CHECK-RV64-FD-NEXT:    sd ra, 296(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t0, 288(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t1, 280(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t2, 272(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd s0, 264(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a0, 256(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a1, 248(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a2, 240(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a3, 232(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a4, 224(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a5, 216(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a6, 208(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd a7, 200(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t3, 192(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t4, 184(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t5, 176(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    sd t6, 168(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft0, 160(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft1, 152(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft2, 144(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft3, 136(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft4, 128(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft5, 120(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft6, 112(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd ft7, 104(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa0, 96(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa1, 88(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa2, 80(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa3, 72(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa4, 64(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa5, 56(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa6, 48(sp) # 8-byte Folded Spill
+; CHECK-RV64-FD-NEXT:    fsd fa7, 40(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft8, 32(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft9, 24(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft10, 16(sp) # 8-byte Folded Spill
 ; CHECK-RV64-FD-NEXT:    fsd ft11, 8(sp) # 8-byte Folded Spill
-; CHECK-RV64-FD-NEXT:    addi s0, sp, 400
+; CHECK-RV64-FD-NEXT:    addi s0, sp, 304
 ; CHECK-RV64-FD-NEXT:    call otherfoo
-; CHECK-RV64-FD-NEXT:    ld ra, 392(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t0, 384(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t1, 376(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t2, 368(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld s0, 360(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a0, 352(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a1, 344(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a2, 336(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a3, 328(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a4, 320(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a5, 312(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a6, 304(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld a7, 296(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t3, 288(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t4, 280(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t5, 272(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    ld t6, 264(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft0, 256(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft1, 248(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft2, 240(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft3, 232(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft4, 224(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft5, 216(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft6, 208(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld ft7, 200(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs0, 192(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs1, 184(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa0, 176(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa1, 168(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa2, 160(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa3, 152(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa4, 144(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa5, 136(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa6, 128(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fa7, 120(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs2, 112(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs3, 104(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs4, 96(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs5, 88(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs6, 80(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs7, 72(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs8, 64(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs9, 56(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs10, 48(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    fld fs11, 40(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld ra, 296(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t0, 288(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t1, 280(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t2, 272(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld s0, 264(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a0, 256(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a1, 248(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a2, 240(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a3, 232(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a4, 224(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a5, 216(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a6, 208(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld a7, 200(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t3, 192(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t4, 184(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t5, 176(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    ld t6, 168(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft0, 160(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft1, 152(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft2, 144(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft3, 136(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft4, 128(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft5, 120(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft6, 112(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld ft7, 104(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa0, 96(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa1, 88(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa2, 80(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa3, 72(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa4, 64(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa5, 56(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa6, 48(sp) # 8-byte Folded Reload
+; CHECK-RV64-FD-NEXT:    fld fa7, 40(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft8, 32(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft9, 24(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft10, 16(sp) # 8-byte Folded Reload
 ; CHECK-RV64-FD-NEXT:    fld ft11, 8(sp) # 8-byte Folded Reload
-; CHECK-RV64-FD-NEXT:    addi sp, sp, 400
+; CHECK-RV64-FD-NEXT:    addi sp, sp, 304
 ; CHECK-RV64-FD-NEXT:    mret
 ;
 ; CHECK-RV64-F-LP64-LABEL: foo_fp_with_call:

From 86842e1f724fba5abae50ce438553895e69b8141 Mon Sep 17 00:00:00 2001
From: Jun Wang <jwang86@yahoo.com>
Date: Wed, 10 Apr 2024 10:47:04 -0700
Subject: [PATCH 044/886] [AMDGPU] New clang option for emitting a waitcnt
 instruction after each memory instruction (#79236)

This patch introduces a new command-line option for clang, namely,
amdgpu-precise-mem-op (or precise-memory in the backend). When this option is specified, a waitcnt
instruction is generated after each memory load/store instruction. The
counter values are always 0, but which counters are involved depends on
the memory instruction.

---------

Co-authored-by: Jun Wang <jun.wang7@amd.com>
---
 clang/include/clang/Driver/Options.td         |    3 +
 clang/lib/Driver/ToolChains/AMDGPU.cpp        |    4 +
 clang/test/Driver/amdgpu-features.c           |    6 +
 llvm/lib/Target/AMDGPU/AMDGPU.td              |    4 +
 llvm/lib/Target/AMDGPU/GCNSubtarget.h         |    3 +
 llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp   |    8 +
 .../insert_waitcnt_for_precise_memory.ll      | 1658 +++++++++++++++++
 7 files changed, 1686 insertions(+)
 create mode 100644 llvm/test/CodeGen/AMDGPU/insert_waitcnt_for_precise_memory.ll

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 0a74e6c75f95b..7ac36222644aa 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -4912,6 +4912,9 @@ defm tgsplit : SimpleMFlag<"tgsplit", "Enable", "Disable",
 defm wavefrontsize64 : SimpleMFlag<"wavefrontsize64",
   "Specify wavefront size 64", "Specify wavefront size 32",
   " mode (AMDGPU only)">;
+defm amdgpu_precise_memory_op
+    : SimpleMFlag<"amdgpu-precise-memory-op", "Enable", "Disable",
+                  " precise memory mode (AMDGPU only)">;
 
 defm unsafe_fp_atomics : BoolMOption<"unsafe-fp-atomics",
   TargetOpts<"AllowAMDGPUUnsafeFPAtomics">, DefaultFalse,
diff --git a/clang/lib/Driver/ToolChains/AMDGPU.cpp b/clang/lib/Driver/ToolChains/AMDGPU.cpp
index e122379e860e2..4e6362a0f4063 100644
--- a/clang/lib/Driver/ToolChains/AMDGPU.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPU.cpp
@@ -670,6 +670,10 @@ void amdgpu::getAMDGPUTargetFeatures(const Driver &D,
                    options::OPT_mno_wavefrontsize64, false))
     Features.push_back("+wavefrontsize64");
 
+  if (Args.hasFlag(options::OPT_mamdgpu_precise_memory_op,
+                   options::OPT_mno_amdgpu_precise_memory_op, false))
+    Features.push_back("+precise-memory");
+
   handleTargetFeaturesGroup(D, Triple, Args, Features,
                             options::OPT_m_amdgpu_Features_Group);
 }
diff --git a/clang/test/Driver/amdgpu-features.c b/clang/test/Driver/amdgpu-features.c
index a516bc6b7ff20..864744db203e9 100644
--- a/clang/test/Driver/amdgpu-features.c
+++ b/clang/test/Driver/amdgpu-features.c
@@ -32,3 +32,9 @@
 
 // RUN: %clang -### -target amdgcn -mcpu=gfx1010 -mno-cumode %s 2>&1 | FileCheck --check-prefix=NO-CUMODE %s
 // NO-CUMODE: "-target-feature" "-cumode"
+
+// RUN: %clang -### -target amdgcn -mcpu=gfx1010 -mamdgpu-precise-memory-op %s 2>&1 | FileCheck --check-prefix=PREC-MEM %s
+// PREC-MEM: "-target-feature" "+precise-memory"
+
+// RUN: %clang -### -target amdgcn -mcpu=gfx1010 -mno-amdgpu-precise-memory-op %s 2>&1 | FileCheck --check-prefix=NO-PREC-MEM %s
+// NO-PREC-MEM-NOT: {{".*precise-memory"}}
diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
index 37dcfef3b2a3d..9b09550159993 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -168,6 +168,10 @@ def FeatureCuMode : SubtargetFeature<"cumode",
   "Enable CU wavefront execution mode"
 >;
 
+def FeaturePreciseMemory
+    : SubtargetFeature<"precise-memory", "EnablePreciseMemory",
+                       "true", "Enable precise memory mode">;
+
 def FeatureSGPRInitBug : SubtargetFeature<"sgpr-init-bug",
   "SGPRInitBug",
   "true",
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index e24a18a2842f6..04ff53a6647bd 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -87,6 +87,7 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   bool EnableTgSplit = false;
   bool EnableCuMode = false;
   bool TrapHandler = false;
+  bool EnablePreciseMemory = false;
 
   // Used as options.
   bool EnableLoadStoreOpt = false;
@@ -599,6 +600,8 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
     return EnableCuMode;
   }
 
+  bool isPreciseMemoryEnabled() const { return EnablePreciseMemory; }
+
   bool hasFlatAddressSpace() const {
     return FlatAddressSpace;
   }
diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index bb499c5c8c578..556ec3e231ff1 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -2305,6 +2305,14 @@ bool SIInsertWaitcnts::insertWaitcntInBlock(MachineFunction &MF,
     }
 #endif
 
+    if (ST->isPreciseMemoryEnabled() && Inst.mayLoadOrStore()) {
+      AMDGPU::Waitcnt Wait = WCG->getAllZeroWaitcnt(
+          Inst.mayStore() && !SIInstrInfo::isAtomicRet(Inst));
+      ScoreBrackets.simplifyWaitcnt(Wait);
+      Modified |= generateWaitcnt(Wait, std::next(Inst.getIterator()), Block,
+                                  ScoreBrackets, /*OldWaitcntInstr=*/nullptr);
+    }
+
     LLVM_DEBUG({
       Inst.print(dbgs());
       ScoreBrackets.dump();
diff --git a/llvm/test/CodeGen/AMDGPU/insert_waitcnt_for_precise_memory.ll b/llvm/test/CodeGen/AMDGPU/insert_waitcnt_for_precise_memory.ll
new file mode 100644
index 0000000000000..df03e89370377
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/insert_waitcnt_for_precise_memory.ll
@@ -0,0 +1,1658 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -mattr=+precise-memory < %s | FileCheck %s -check-prefixes=GFX9
+; RUN: llc -mtriple=amdgcn -mcpu=gfx90a -mattr=+precise-memory < %s | FileCheck %s -check-prefixes=GFX90A
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+precise-memory < %s | FileCheck %s -check-prefixes=GFX10
+; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -mattr=+enable-flat-scratch,+precise-memory < %s | FileCheck --check-prefixes=GFX9-FLATSCR %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+precise-memory < %s | FileCheck %s -check-prefixes=GFX11
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -mattr=+precise-memory < %s | FileCheck %s -check-prefixes=GFX12
+
+; from atomicrmw-expand.ll
+; covers flat_load, flat_atomic (atomic with return)
+;
+define void @syncscope_workgroup_nortn(ptr %addr, float %val) {
+; GFX9-LABEL: syncscope_workgroup_nortn:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_load_dword v4, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b64 s[4:5], 0
+; GFX9-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX9-NEXT:    v_add_f32_e32 v3, v4, v2
+; GFX9-NEXT:    flat_atomic_cmpswap v3, v[0:1], v[3:4] glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
+; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; GFX9-NEXT:    v_mov_b32_e32 v4, v3
+; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX90A-LABEL: syncscope_workgroup_nortn:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    flat_load_dword v5, v[0:1]
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    s_mov_b64 s[4:5], 0
+; GFX90A-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX90A-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX90A-NEXT:    v_add_f32_e32 v4, v5, v2
+; GFX90A-NEXT:    flat_atomic_cmpswap v3, v[0:1], v[4:5] glc
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
+; GFX90A-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; GFX90A-NEXT:    v_mov_b32_e32 v5, v3
+; GFX90A-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; GFX90A-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX90A-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX90A-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX90A-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: syncscope_workgroup_nortn:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_load_dword v4, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_mov_b32 s4, 0
+; GFX10-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX10-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX10-NEXT:    v_add_f32_e32 v3, v4, v2
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    flat_atomic_cmpswap v3, v[0:1], v[3:4] glc
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    buffer_gl0_inv
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
+; GFX10-NEXT:    v_mov_b32_e32 v4, v3
+; GFX10-NEXT:    s_or_b32 s4, vcc_lo, s4
+; GFX10-NEXT:    s_andn2_b32 exec_lo, exec_lo, s4
+; GFX10-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX10-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX10-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-FLATSCR-LABEL: syncscope_workgroup_nortn:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    flat_load_dword v4, v[0:1]
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_mov_b64 s[0:1], 0
+; GFX9-FLATSCR-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX9-FLATSCR-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX9-FLATSCR-NEXT:    v_add_f32_e32 v3, v4, v2
+; GFX9-FLATSCR-NEXT:    flat_atomic_cmpswap v3, v[0:1], v[3:4] glc
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
+; GFX9-FLATSCR-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v4, v3
+; GFX9-FLATSCR-NEXT:    s_andn2_b64 exec, exec, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX9-FLATSCR-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX9-FLATSCR-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: syncscope_workgroup_nortn:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    flat_load_b32 v4, v[0:1]
+; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s0, 0
+; GFX11-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX11-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_add_f32_e32 v3, v4, v2
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] glc
+; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    buffer_gl0_inv
+; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
+; GFX11-NEXT:    v_mov_b32_e32 v4, v3
+; GFX11-NEXT:    s_or_b32 s0, vcc_lo, s0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
+; GFX11-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX11-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s0
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: syncscope_workgroup_nortn:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    flat_load_b32 v4, v[0:1]
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_mov_b32 s0, 0
+; GFX12-NEXT:  .LBB0_1: ; %atomicrmw.start
+; GFX12-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_add_f32_e32 v3, v4, v2
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    flat_atomic_cmpswap_b32 v3, v[0:1], v[3:4] th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    global_inv scope:SCOPE_SE
+; GFX12-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
+; GFX12-NEXT:    v_mov_b32_e32 v4, v3
+; GFX12-NEXT:    s_or_b32 s0, vcc_lo, s0
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
+; GFX12-NEXT:    s_cbranch_execnz .LBB0_1
+; GFX12-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX12-NEXT:    s_or_b32 exec_lo, exec_lo, s0
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %res = atomicrmw fadd ptr %addr, float %val syncscope("workgroup") seq_cst
+  ret void
+}
+
+; from atomicrmw-nand.ll
+; covers global_atomic (atomic with return), global_load
+;
+define i32 @atomic_nand_i32_global(ptr addrspace(1) %ptr) nounwind {
+; GFX9-LABEL: atomic_nand_i32_global:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v2, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_mov_b64 s[4:5], 0
+; GFX9-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX9-NEXT:    v_mov_b32_e32 v3, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v3
+; GFX9-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX9-NEXT:    global_atomic_cmpswap v2, v[0:1], v[2:3], off glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_wbinvl1_vol
+; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
+; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX9-NEXT:    v_mov_b32_e32 v0, v2
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX90A-LABEL: atomic_nand_i32_global:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    global_load_dword v2, v[0:1], off
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_mov_b64 s[4:5], 0
+; GFX90A-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX90A-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v2
+; GFX90A-NEXT:    v_not_b32_e32 v2, v3
+; GFX90A-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX90A-NEXT:    buffer_wbl2
+; GFX90A-NEXT:    global_atomic_cmpswap v2, v[0:1], v[2:3], off glc
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_invl2
+; GFX90A-NEXT:    buffer_wbinvl1_vol
+; GFX90A-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
+; GFX90A-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; GFX90A-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; GFX90A-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX90A-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX90A-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, v2
+; GFX90A-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: atomic_nand_i32_global:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    global_load_dword v2, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_mov_b32 s4, 0
+; GFX10-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX10-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX10-NEXT:    v_mov_b32_e32 v3, v2
+; GFX10-NEXT:    v_not_b32_e32 v2, v3
+; GFX10-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_atomic_cmpswap v2, v[0:1], v[2:3], off glc
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    buffer_gl1_inv
+; GFX10-NEXT:    buffer_gl0_inv
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
+; GFX10-NEXT:    s_or_b32 s4, vcc_lo, s4
+; GFX10-NEXT:    s_andn2_b32 exec_lo, exec_lo, s4
+; GFX10-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX10-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX10-NEXT:    s_or_b32 exec_lo, exec_lo, s4
+; GFX10-NEXT:    v_mov_b32_e32 v0, v2
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-FLATSCR-LABEL: atomic_nand_i32_global:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    global_load_dword v2, v[0:1], off
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_mov_b64 s[0:1], 0
+; GFX9-FLATSCR-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX9-FLATSCR-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v3, v2
+; GFX9-FLATSCR-NEXT:    v_not_b32_e32 v2, v3
+; GFX9-FLATSCR-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX9-FLATSCR-NEXT:    global_atomic_cmpswap v2, v[0:1], v[2:3], off glc
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    buffer_wbinvl1_vol
+; GFX9-FLATSCR-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
+; GFX9-FLATSCR-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_andn2_b64 exec, exec, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX9-FLATSCR-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX9-FLATSCR-NEXT:    s_or_b64 exec, exec, s[0:1]
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, v2
+; GFX9-FLATSCR-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: atomic_nand_i32_global:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    global_load_b32 v2, v[0:1], off
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    s_mov_b32 s0, 0
+; GFX11-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX11-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX11-NEXT:    v_mov_b32_e32 v3, v2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_not_b32_e32 v2, v3
+; GFX11-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], v[2:3], off glc
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    buffer_gl1_inv
+; GFX11-NEXT:    buffer_gl0_inv
+; GFX11-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
+; GFX11-NEXT:    s_or_b32 s0, vcc_lo, s0
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
+; GFX11-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX11-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s0
+; GFX11-NEXT:    v_mov_b32_e32 v0, v2
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: atomic_nand_i32_global:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    global_load_b32 v2, v[0:1], off
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    s_mov_b32 s0, 0
+; GFX12-NEXT:  .LBB1_1: ; %atomicrmw.start
+; GFX12-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-NEXT:    v_mov_b32_e32 v3, v2
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT:    v_not_b32_e32 v2, v3
+; GFX12-NEXT:    v_or_b32_e32 v2, -5, v2
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], v[2:3], off th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    global_inv scope:SCOPE_SYS
+; GFX12-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
+; GFX12-NEXT:    s_or_b32 s0, vcc_lo, s0
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
+; GFX12-NEXT:    s_cbranch_execnz .LBB1_1
+; GFX12-NEXT:  ; %bb.2: ; %atomicrmw.end
+; GFX12-NEXT:    s_or_b32 exec_lo, exec_lo, s0
+; GFX12-NEXT:    v_mov_b32_e32 v0, v2
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %result = atomicrmw nand ptr addrspace(1) %ptr, i32 4 seq_cst
+  ret i32 %result
+}
+
+; from call-argument-types.ll
+; covers scratch_load, scratch_store, buffer_load, buffer_store
+;
+declare hidden void @byval_align16_f64_arg(<32 x i32>, ptr addrspace(5) byval(double) align 16)
+define void @tail_call_byval_align16(<32 x i32> %val, double %tmp)  {
+; GFX9-LABEL: tail_call_byval_align16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:28
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_load_dword v33, off, s[0:3], s32
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_getpc_b64 s[16:17]
+; GFX9-NEXT:    s_add_u32 s16, s16, byval_align16_f64_arg@rel32@lo+4
+; GFX9-NEXT:    s_addc_u32 s17, s17, byval_align16_f64_arg@rel32@hi+12
+; GFX9-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:20
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:24
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:16
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_store_dword v33, off, s[0:3], s32
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[16:17]
+;
+; GFX90A-LABEL: tail_call_byval_align16:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:28
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_load_dword v33, off, s[0:3], s32 offset:24
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_load_dword v34, off, s[0:3], s32
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_getpc_b64 s[16:17]
+; GFX90A-NEXT:    s_add_u32 s16, s16, byval_align16_f64_arg@rel32@lo+4
+; GFX90A-NEXT:    s_addc_u32 s17, s17, byval_align16_f64_arg@rel32@hi+12
+; GFX90A-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:20
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_store_dword v33, off, s[0:3], s32 offset:16
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_store_dword v34, off, s[0:3], s32
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_setpc_b64 s[16:17]
+;
+; GFX10-LABEL: tail_call_byval_align16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:28
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    buffer_load_dword v33, off, s[0:3], s32 offset:24
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    buffer_load_dword v34, off, s[0:3], s32
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_getpc_b64 s[16:17]
+; GFX10-NEXT:    s_add_u32 s16, s16, byval_align16_f64_arg@rel32@lo+4
+; GFX10-NEXT:    s_addc_u32 s17, s17, byval_align16_f64_arg@rel32@hi+12
+; GFX10-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:20
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    buffer_store_dword v33, off, s[0:3], s32 offset:16
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    buffer_store_dword v34, off, s[0:3], s32
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_setpc_b64 s[16:17]
+;
+; GFX9-FLATSCR-LABEL: tail_call_byval_align16:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    scratch_load_dword v32, off, s32
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_getpc_b64 s[0:1]
+; GFX9-FLATSCR-NEXT:    s_add_u32 s0, s0, byval_align16_f64_arg@rel32@lo+4
+; GFX9-FLATSCR-NEXT:    s_addc_u32 s1, s1, byval_align16_f64_arg@rel32@hi+12
+; GFX9-FLATSCR-NEXT:    scratch_store_dword off, v32, s32
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    scratch_load_dwordx2 v[32:33], off, s32 offset:24
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    scratch_store_dwordx2 off, v[32:33], s32 offset:16
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_setpc_b64 s[0:1]
+;
+; GFX11-LABEL: tail_call_byval_align16:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    scratch_load_b32 v32, off, s32
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    s_getpc_b64 s[0:1]
+; GFX11-NEXT:    s_add_u32 s0, s0, byval_align16_f64_arg@rel32@lo+4
+; GFX11-NEXT:    s_addc_u32 s1, s1, byval_align16_f64_arg@rel32@hi+12
+; GFX11-NEXT:    scratch_store_b32 off, v32, s32
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    scratch_load_b64 v[32:33], off, s32 offset:24
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    scratch_store_b64 off, v[32:33], s32 offset:16
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_setpc_b64 s[0:1]
+;
+; GFX12-LABEL: tail_call_byval_align16:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    scratch_load_b32 v32, off, s32
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    s_getpc_b64 s[0:1]
+; GFX12-NEXT:    s_sext_i32_i16 s1, s1
+; GFX12-NEXT:    s_add_co_u32 s0, s0, byval_align16_f64_arg@rel32@lo+8
+; GFX12-NEXT:    s_add_co_ci_u32 s1, s1, byval_align16_f64_arg@rel32@hi+16
+; GFX12-NEXT:    scratch_store_b32 off, v32, s32
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    scratch_load_b64 v[32:33], off, s32 offset:24
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    scratch_store_b64 off, v[32:33], s32 offset:16
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    s_setpc_b64 s[0:1]
+entry:
+  %alloca = alloca double, align 8, addrspace(5)
+  tail call void @byval_align16_f64_arg(<32 x i32> %val, ptr addrspace(5) byval(double) align 16 %alloca)
+  ret void
+}
+
+; from udiv.ll
+; covers s_load
+;
+define amdgpu_kernel void @udiv_i32(ptr addrspace(1) %out, i32 %x, i32 %y) {
+; GFX9-LABEL: udiv_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    v_cvt_f32_u32_e32 v0, s3
+; GFX9-NEXT:    s_sub_i32 s4, 0, s3
+; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX9-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX9-NEXT:    s_mul_i32 s4, s4, s5
+; GFX9-NEXT:    s_mul_hi_u32 s4, s5, s4
+; GFX9-NEXT:    s_add_i32 s5, s5, s4
+; GFX9-NEXT:    s_mul_hi_u32 s4, s2, s5
+; GFX9-NEXT:    s_mul_i32 s5, s4, s3
+; GFX9-NEXT:    s_sub_i32 s2, s2, s5
+; GFX9-NEXT:    s_add_i32 s6, s4, 1
+; GFX9-NEXT:    s_sub_i32 s5, s2, s3
+; GFX9-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX9-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX9-NEXT:    s_cselect_b32 s2, s5, s2
+; GFX9-NEXT:    s_add_i32 s5, s4, 1
+; GFX9-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX9-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX9-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: udiv_i32:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_cvt_f32_u32_e32 v0, s3
+; GFX90A-NEXT:    s_sub_i32 s4, 0, s3
+; GFX90A-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX90A-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; GFX90A-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX90A-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX90A-NEXT:    s_mul_i32 s4, s4, s5
+; GFX90A-NEXT:    s_mul_hi_u32 s4, s5, s4
+; GFX90A-NEXT:    s_add_i32 s5, s5, s4
+; GFX90A-NEXT:    s_mul_hi_u32 s4, s2, s5
+; GFX90A-NEXT:    s_mul_i32 s5, s4, s3
+; GFX90A-NEXT:    s_sub_i32 s2, s2, s5
+; GFX90A-NEXT:    s_add_i32 s6, s4, 1
+; GFX90A-NEXT:    s_sub_i32 s5, s2, s3
+; GFX90A-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX90A-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX90A-NEXT:    s_cselect_b32 s2, s5, s2
+; GFX90A-NEXT:    s_add_i32 s5, s4, 1
+; GFX90A-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX90A-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s2
+; GFX90A-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: udiv_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_cvt_f32_u32_e32 v0, s3
+; GFX10-NEXT:    s_sub_i32 s5, 0, s3
+; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_mul_i32 s5, s5, s4
+; GFX10-NEXT:    s_mul_hi_u32 s5, s4, s5
+; GFX10-NEXT:    s_add_i32 s4, s4, s5
+; GFX10-NEXT:    s_mul_hi_u32 s4, s2, s4
+; GFX10-NEXT:    s_mul_i32 s5, s4, s3
+; GFX10-NEXT:    s_sub_i32 s2, s2, s5
+; GFX10-NEXT:    s_add_i32 s5, s4, 1
+; GFX10-NEXT:    s_sub_i32 s6, s2, s3
+; GFX10-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX10-NEXT:    s_cselect_b32 s4, s5, s4
+; GFX10-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX10-NEXT:    s_add_i32 s5, s4, 1
+; GFX10-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX10-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX10-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: udiv_i32:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-FLATSCR-NEXT:    v_cvt_f32_u32_e32 v0, s3
+; GFX9-FLATSCR-NEXT:    s_sub_i32 s4, 0, s3
+; GFX9-FLATSCR-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX9-FLATSCR-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; GFX9-FLATSCR-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX9-FLATSCR-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX9-FLATSCR-NEXT:    s_mul_i32 s4, s4, s5
+; GFX9-FLATSCR-NEXT:    s_mul_hi_u32 s4, s5, s4
+; GFX9-FLATSCR-NEXT:    s_add_i32 s5, s5, s4
+; GFX9-FLATSCR-NEXT:    s_mul_hi_u32 s4, s2, s5
+; GFX9-FLATSCR-NEXT:    s_mul_i32 s5, s4, s3
+; GFX9-FLATSCR-NEXT:    s_sub_i32 s2, s2, s5
+; GFX9-FLATSCR-NEXT:    s_add_i32 s6, s4, 1
+; GFX9-FLATSCR-NEXT:    s_sub_i32 s5, s2, s3
+; GFX9-FLATSCR-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX9-FLATSCR-NEXT:    s_cselect_b32 s4, s6, s4
+; GFX9-FLATSCR-NEXT:    s_cselect_b32 s2, s5, s2
+; GFX9-FLATSCR-NEXT:    s_add_i32 s5, s4, 1
+; GFX9-FLATSCR-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX9-FLATSCR-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, s2
+; GFX9-FLATSCR-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: udiv_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_cvt_f32_u32_e32 v0, s3
+; GFX11-NEXT:    s_sub_i32 s5, 0, s3
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; GFX11-NEXT:    s_waitcnt_depctr 0xfff
+; GFX11-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; GFX11-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-NEXT:    s_mul_i32 s5, s5, s4
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_mul_hi_u32 s5, s4, s5
+; GFX11-NEXT:    s_add_i32 s4, s4, s5
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_mul_hi_u32 s4, s2, s4
+; GFX11-NEXT:    s_mul_i32 s5, s4, s3
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    s_sub_i32 s2, s2, s5
+; GFX11-NEXT:    s_add_i32 s5, s4, 1
+; GFX11-NEXT:    s_sub_i32 s6, s2, s3
+; GFX11-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX11-NEXT:    s_cselect_b32 s4, s5, s4
+; GFX11-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX11-NEXT:    s_add_i32 s5, s4, 1
+; GFX11-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX11-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: udiv_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_cvt_f32_u32 s4, s3
+; GFX12-NEXT:    s_sub_co_i32 s5, 0, s3
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_2) | instskip(NEXT) | instid1(TRANS32_DEP_1)
+; GFX12-NEXT:    v_rcp_iflag_f32_e32 v0, s4
+; GFX12-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(SALU_CYCLE_3)
+; GFX12-NEXT:    s_mul_f32 s4, s4, 0x4f7ffffe
+; GFX12-NEXT:    s_cvt_u32_f32 s4, s4
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_3) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_mul_i32 s5, s5, s4
+; GFX12-NEXT:    s_mul_hi_u32 s5, s4, s5
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_add_co_i32 s4, s4, s5
+; GFX12-NEXT:    s_mul_hi_u32 s4, s2, s4
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_mul_i32 s5, s4, s3
+; GFX12-NEXT:    s_sub_co_i32 s2, s2, s5
+; GFX12-NEXT:    s_add_co_i32 s5, s4, 1
+; GFX12-NEXT:    s_sub_co_i32 s6, s2, s3
+; GFX12-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX12-NEXT:    s_cselect_b32 s4, s5, s4
+; GFX12-NEXT:    s_cselect_b32 s2, s6, s2
+; GFX12-NEXT:    s_add_co_i32 s5, s4, 1
+; GFX12-NEXT:    s_cmp_ge_u32 s2, s3
+; GFX12-NEXT:    s_cselect_b32 s2, s5, s4
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX12-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2
+; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    s_endpgm
+  %r = udiv i32 %x, %y
+  store i32 %r, ptr addrspace(1) %out
+  ret void
+}
+
+declare float @llvm.amdgcn.s.buffer.load.f32(<4 x i32>, i32, i32)
+
+; from smrd.ll
+; covers s_buffer_load
+;
+define amdgpu_ps float @smrd_sgpr_offset(<4 x i32> inreg %desc, i32 inreg %offset) #0 {
+; GFX9-LABEL: smrd_sgpr_offset:
+; GFX9:       ; %bb.0: ; %main_body
+; GFX9-NEXT:    s_buffer_load_dword s0, s[0:3], s4 offset:0x0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX90A-LABEL: smrd_sgpr_offset:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    s_buffer_load_dword s0, s[0:3], s4 offset:0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s0
+; GFX90A-NEXT:    ; return to shader part epilog
+;
+; GFX10-LABEL: smrd_sgpr_offset:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    s_buffer_load_dword s0, s[0:3], s4 offset:0x0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    ; return to shader part epilog
+;
+; GFX9-FLATSCR-LABEL: smrd_sgpr_offset:
+; GFX9-FLATSCR:       ; %bb.0: ; %main_body
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s11, s5
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s10, s4
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s9, s3
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s8, s2
+; GFX9-FLATSCR-NEXT:    s_buffer_load_dword s0, s[8:11], s6 offset:0x0
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-FLATSCR-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: smrd_sgpr_offset:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    s_buffer_load_b32 s0, s[0:3], s4 offset:0x0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_mov_b32_e32 v0, s0
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX12-LABEL: smrd_sgpr_offset:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    s_buffer_load_b32 s0, s[0:3], s4 offset:0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_mov_b32_e32 v0, s0
+; GFX12-NEXT:    ; return to shader part epilog
+main_body:
+  %r = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %desc, i32 %offset, i32 0)
+  ret float %r
+}
+
+; from atomic_load_add.ll
+; covers s_load, ds_add (atomic without return)
+;
+define amdgpu_kernel void @atomic_add_local(ptr addrspace(3) %local) {
+; GFX9-LABEL: atomic_add_local:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_mov_b64 s[2:3], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-NEXT:    s_cbranch_execz .LBB5_2
+; GFX9-NEXT:  ; %bb.1:
+; GFX9-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_bcnt1_i32_b64 s1, s[2:3]
+; GFX9-NEXT:    s_mul_i32 s1, s1, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    ds_add_u32 v0, v1
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:  .LBB5_2:
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: atomic_add_local:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_mov_b64 s[2:3], exec
+; GFX90A-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX90A-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX90A-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX90A-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX90A-NEXT:    s_cbranch_execz .LBB5_2
+; GFX90A-NEXT:  ; %bb.1:
+; GFX90A-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_bcnt1_i32_b64 s1, s[2:3]
+; GFX90A-NEXT:    s_mul_i32 s1, s1, 5
+; GFX90A-NEXT:    v_mov_b32_e32 v1, s1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s0
+; GFX90A-NEXT:    ds_add_u32 v0, v1
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:  .LBB5_2:
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: atomic_add_local:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_mov_b32 s2, exec_lo
+; GFX10-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX10-NEXT:    s_and_saveexec_b32 s3, vcc_lo
+; GFX10-NEXT:    s_cbranch_execz .LBB5_2
+; GFX10-NEXT:  ; %bb.1:
+; GFX10-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_bcnt1_i32_b32 s1, s2
+; GFX10-NEXT:    s_mul_i32 s1, s1, 5
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    ds_add_u32 v0, v1
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    buffer_gl0_inv
+; GFX10-NEXT:  .LBB5_2:
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: atomic_add_local:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_mov_b64 s[2:3], exec
+; GFX9-FLATSCR-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX9-FLATSCR-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
+; GFX9-FLATSCR-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-FLATSCR-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX9-FLATSCR-NEXT:    s_cbranch_execz .LBB5_2
+; GFX9-FLATSCR-NEXT:  ; %bb.1:
+; GFX9-FLATSCR-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_bcnt1_i32_b64 s1, s[2:3]
+; GFX9-FLATSCR-NEXT:    s_mul_i32 s1, s1, 5
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-FLATSCR-NEXT:    ds_add_u32 v0, v1
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:  .LBB5_2:
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: atomic_add_local:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX11-NEXT:    s_cbranch_execz .LBB5_2
+; GFX11-NEXT:  ; %bb.1:
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_bcnt1_i32_b32 s1, s2
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_mul_i32 s1, s1, 5
+; GFX11-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v0, s0
+; GFX11-NEXT:    ds_add_u32 v0, v1
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    buffer_gl0_inv
+; GFX11-NEXT:  .LBB5_2:
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: atomic_add_local:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_mov_b32 s2, exec_lo
+; GFX12-NEXT:    s_mov_b32 s3, exec_lo
+; GFX12-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX12-NEXT:    s_cbranch_execz .LBB5_2
+; GFX12-NEXT:  ; %bb.1:
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_bcnt1_i32_b32 s1, s2
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_mul_i32 s1, s1, 5
+; GFX12-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v0, s0
+; GFX12-NEXT:    ds_add_u32 v0, v1
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    global_inv scope:SCOPE_SE
+; GFX12-NEXT:  .LBB5_2:
+; GFX12-NEXT:    s_endpgm
+   %unused = atomicrmw volatile add ptr addrspace(3) %local, i32 5 seq_cst
+   ret void
+}
+
+; from flat_atomics_i32_system.ll
+; covers flat_atomic_swap (atomic without return)
+;
+define void @flat_atomic_xchg_i32_noret(ptr %ptr, i32 %in) {
+; GFX9-LABEL: flat_atomic_xchg_i32_noret:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_atomic_swap v[0:1], v2
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    buffer_wbinvl1_vol
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX90A-LABEL: flat_atomic_xchg_i32_noret:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    buffer_wbl2
+; GFX90A-NEXT:    flat_atomic_swap v[0:1], v2
+; GFX90A-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX90A-NEXT:    buffer_invl2
+; GFX90A-NEXT:    buffer_wbinvl1_vol
+; GFX90A-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_atomic_xchg_i32_noret:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    flat_atomic_swap v[0:1], v2
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    buffer_gl1_inv
+; GFX10-NEXT:    buffer_gl0_inv
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-FLATSCR-LABEL: flat_atomic_xchg_i32_noret:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    flat_atomic_swap v[0:1], v2
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    buffer_wbinvl1_vol
+; GFX9-FLATSCR-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: flat_atomic_xchg_i32_noret:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    buffer_gl1_inv
+; GFX11-NEXT:    buffer_gl0_inv
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: flat_atomic_xchg_i32_noret:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX12-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX12-NEXT:    global_inv scope:SCOPE_SYS
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %tmp0 = atomicrmw xchg ptr %ptr, i32 %in seq_cst
+  ret void
+}
+
+; from atomic_load_add.ll
+; covers s_load, ds_add_rtn (atomic with return)
+;
+define amdgpu_kernel void @atomic_add_ret_local(ptr addrspace(1) %out, ptr addrspace(3) %local) {
+; GFX9-LABEL: atomic_add_ret_local:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-NEXT:    ; implicit-def: $vgpr1
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX9-NEXT:    s_cbranch_execz .LBB7_2
+; GFX9-NEXT:  ; %bb.1:
+; GFX9-NEXT:    s_load_dword s6, s[0:1], 0x2c
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-NEXT:    v_mov_b32_e32 v1, s6
+; GFX9-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:  .LBB7_2:
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: atomic_add_ret_local:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_mov_b64 s[4:5], exec
+; GFX90A-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX90A-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX90A-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX90A-NEXT:    ; implicit-def: $vgpr1
+; GFX90A-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX90A-NEXT:    s_cbranch_execz .LBB7_2
+; GFX90A-NEXT:  ; %bb.1:
+; GFX90A-NEXT:    s_load_dword s6, s[0:1], 0x2c
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX90A-NEXT:    s_mul_i32 s4, s4, 5
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s4
+; GFX90A-NEXT:    v_mov_b32_e32 v1, s6
+; GFX90A-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:  .LBB7_2:
+; GFX90A-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX90A-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: atomic_add_ret_local:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_mov_b32 s3, exec_lo
+; GFX10-NEXT:    ; implicit-def: $vgpr1
+; GFX10-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX10-NEXT:    s_and_saveexec_b32 s2, vcc_lo
+; GFX10-NEXT:    s_cbranch_execz .LBB7_2
+; GFX10-NEXT:  ; %bb.1:
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10-NEXT:    v_mov_b32_e32 v2, s3
+; GFX10-NEXT:    v_mov_b32_e32 v1, s4
+; GFX10-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    buffer_gl0_inv
+; GFX10-NEXT:  .LBB7_2:
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX10-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: atomic_add_ret_local:
+; GFX9-FLATSCR:       ; %bb.0:
+; GFX9-FLATSCR-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-FLATSCR-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-FLATSCR-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX9-FLATSCR-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-FLATSCR-NEXT:    ; implicit-def: $vgpr1
+; GFX9-FLATSCR-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX9-FLATSCR-NEXT:    s_cbranch_execz .LBB7_2
+; GFX9-FLATSCR-NEXT:  ; %bb.1:
+; GFX9-FLATSCR-NEXT:    s_load_dword s6, s[0:1], 0x2c
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-FLATSCR-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v2, s4
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, s6
+; GFX9-FLATSCR-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:  .LBB7_2:
+; GFX9-FLATSCR-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-FLATSCR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-FLATSCR-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX9-FLATSCR-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: atomic_add_ret_local:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX11-NEXT:    ; implicit-def: $vgpr1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX11-NEXT:    s_cbranch_execz .LBB7_2
+; GFX11-NEXT:  ; %bb.1:
+; GFX11-NEXT:    s_load_b32 s4, s[0:1], 0x2c
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11-NEXT:    v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v1, s4
+; GFX11-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    buffer_gl0_inv
+; GFX11-NEXT:  .LBB7_2:
+; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: atomic_add_ret_local:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_mov_b32 s3, exec_lo
+; GFX12-NEXT:    s_mov_b32 s2, exec_lo
+; GFX12-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX12-NEXT:    ; implicit-def: $vgpr1
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX12-NEXT:    s_cbranch_execz .LBB7_2
+; GFX12-NEXT:  ; %bb.1:
+; GFX12-NEXT:    s_load_b32 s4, s[0:1], 0x2c
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_mul_i32 s3, s3, 5
+; GFX12-NEXT:    v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v1, s4
+; GFX12-NEXT:    ds_add_rtn_u32 v1, v1, v2
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    global_inv scope:SCOPE_SE
+; GFX12-NEXT:  .LBB7_2:
+; GFX12-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    s_endpgm
+  %val = atomicrmw volatile add ptr addrspace(3) %local, i32 5 seq_cst
+  store i32 %val, ptr addrspace(1) %out
+  ret void
+}
+
+declare i32 @llvm.amdgcn.raw.ptr.buffer.atomic.add(i32, ptr addrspace(8), i32, i32, i32 immarg)
+
+; from atomic_optimizations_buffer.ll
+; covers buffer_atomic (atomic with return)
+;
+define amdgpu_kernel void @add_i32_constant(ptr addrspace(1) %out, ptr addrspace(8) %inout) {
+; GFX9-LABEL: add_i32_constant:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-NEXT:    ; implicit-def: $vgpr1
+; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX9-NEXT:    s_cbranch_execz .LBB8_2
+; GFX9-NEXT:  ; %bb.1:
+; GFX9-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:  .LBB8_2:
+; GFX9-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX9-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: add_i32_constant:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    s_mov_b64 s[4:5], exec
+; GFX90A-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX90A-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX90A-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX90A-NEXT:    ; implicit-def: $vgpr1
+; GFX90A-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX90A-NEXT:    s_cbranch_execz .LBB8_2
+; GFX90A-NEXT:  ; %bb.1:
+; GFX90A-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX90A-NEXT:    s_mul_i32 s4, s4, 5
+; GFX90A-NEXT:    v_mov_b32_e32 v1, s4
+; GFX90A-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:  .LBB8_2:
+; GFX90A-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX90A-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: add_i32_constant:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_mov_b32 s3, exec_lo
+; GFX10-NEXT:    ; implicit-def: $vgpr1
+; GFX10-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX10-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX10-NEXT:    s_and_saveexec_b32 s2, vcc_lo
+; GFX10-NEXT:    s_cbranch_execz .LBB8_2
+; GFX10-NEXT:  ; %bb.1:
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x34
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX10-NEXT:    s_mul_i32 s3, s3, 5
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    buffer_atomic_add v1, off, s[4:7], 0 glc
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:  .LBB8_2:
+; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX10-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX10-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: add_i32_constant:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    s_mov_b64 s[4:5], exec
+; GFX9-FLATSCR-NEXT:    v_mbcnt_lo_u32_b32 v0, s4, 0
+; GFX9-FLATSCR-NEXT:    v_mbcnt_hi_u32_b32 v0, s5, v0
+; GFX9-FLATSCR-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
+; GFX9-FLATSCR-NEXT:    ; implicit-def: $vgpr1
+; GFX9-FLATSCR-NEXT:    s_and_saveexec_b64 s[2:3], vcc
+; GFX9-FLATSCR-NEXT:    s_cbranch_execz .LBB8_2
+; GFX9-FLATSCR-NEXT:  ; %bb.1:
+; GFX9-FLATSCR-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x34
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_bcnt1_i32_b64 s4, s[4:5]
+; GFX9-FLATSCR-NEXT:    s_mul_i32 s4, s4, 5
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, s4
+; GFX9-FLATSCR-NEXT:    buffer_atomic_add v1, off, s[8:11], 0 glc
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:  .LBB8_2:
+; GFX9-FLATSCR-NEXT:    s_or_b64 exec, exec, s[2:3]
+; GFX9-FLATSCR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-FLATSCR-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX9-FLATSCR-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: add_i32_constant:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_mov_b32 s3, exec_lo
+; GFX11-NEXT:    s_mov_b32 s2, exec_lo
+; GFX11-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX11-NEXT:    ; implicit-def: $vgpr1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX11-NEXT:    s_cbranch_execz .LBB8_2
+; GFX11-NEXT:  ; %bb.1:
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    s_mul_i32 s3, s3, 5
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    buffer_atomic_add_u32 v1, off, s[4:7], 0 glc
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:  .LBB8_2:
+; GFX11-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: add_i32_constant:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_mov_b32 s3, exec_lo
+; GFX12-NEXT:    s_mov_b32 s2, exec_lo
+; GFX12-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
+; GFX12-NEXT:    ; implicit-def: $vgpr1
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_cmpx_eq_u32_e32 0, v0
+; GFX12-NEXT:    s_cbranch_execz .LBB8_2
+; GFX12-NEXT:  ; %bb.1:
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_bcnt1_i32_b32 s3, s3
+; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    s_mul_i32 s3, s3, 5
+; GFX12-NEXT:    v_mov_b32_e32 v1, s3
+; GFX12-NEXT:    buffer_atomic_add_u32 v1, off, s[4:7], null th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:  .LBB8_2:
+; GFX12-NEXT:    s_or_b32 exec_lo, exec_lo, s2
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_mad_u32_u24 v0, v0, 5, s2
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    s_endpgm
+entry:
+  %old = call i32 @llvm.amdgcn.raw.ptr.buffer.atomic.add(i32 5, ptr addrspace(8) %inout, i32 0, i32 0, i32 0)
+  store i32 %old, ptr addrspace(1) %out
+  ret void
+}
+
+declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32, i16, <8 x i32>, i32, i32)
+
+; from llvm.amdgcn.image.load.a16.ll
+; covers image_load
+;
+define amdgpu_ps <4 x float> @load.f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
+; GFX9-LABEL: load.f32.1d:
+; GFX9:       ; %bb.0: ; %main_body
+; GFX9-NEXT:    image_load v0, v0, s[0:7] dmask:0x1 unorm a16
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX90A-LABEL: load.f32.1d:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    image_load v0, v0, s[0:7] dmask:0x1 unorm a16
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    ; return to shader part epilog
+;
+; GFX10-LABEL: load.f32.1d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_load v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    ; return to shader part epilog
+;
+; GFX9-FLATSCR-LABEL: load.f32.1d:
+; GFX9-FLATSCR:       ; %bb.0: ; %main_body
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s11, s9
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s10, s8
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s9, s7
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s8, s6
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s7, s5
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s6, s4
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s5, s3
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s4, s2
+; GFX9-FLATSCR-NEXT:    image_load v0, v0, s[4:11] dmask:0x1 unorm a16
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: load.f32.1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_load v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX12-LABEL: load.f32.1d:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    image_load v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D a16
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    ; return to shader part epilog
+main_body:
+  %x = extractelement <2 x i16> %coords, i32 0
+  %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 1, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
+  ret <4 x float> %v
+}
+
+declare void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float>, i32, i16, <8 x i32>, i32, i32)
+
+; from llvm.amdgcn.image.store.a16.ll
+; covers image_store
+;
+define amdgpu_ps void @store_f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <4 x float> %val) {
+; GFX9-LABEL: store_f32_1d:
+; GFX9:       ; %bb.0: ; %main_body
+; GFX9-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 unorm a16
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: store_f32_1d:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    v_mov_b32_e32 v5, v4
+; GFX90A-NEXT:    v_mov_b32_e32 v4, v3
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, v1
+; GFX90A-NEXT:    image_store v[2:5], v0, s[0:7] dmask:0x1 unorm a16
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: store_f32_1d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: store_f32_1d:
+; GFX9-FLATSCR:       ; %bb.0: ; %main_body
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s11, s9
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s10, s8
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s9, s7
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s8, s6
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s7, s5
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s6, s4
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s5, s3
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s4, s2
+; GFX9-FLATSCR-NEXT:    image_store v[1:4], v0, s[4:11] dmask:0x1 unorm a16
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f32_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: store_f32_1d:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D a16
+; GFX12-NEXT:    s_wait_storecnt 0x0
+; GFX12-NEXT:    s_endpgm
+
+main_body:
+  %x = extractelement <2 x i16> %coords, i32 0
+  call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %val, i32 1, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
+  ret void
+}
+
+declare i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32, i32, <8 x i32>, i32, i32)
+
+; from llvm.amdgcn.image.atomic.dim.ll
+; covers image_atomic (atomic with return)
+;
+define amdgpu_ps float @atomic_swap_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
+; GFX9-LABEL: atomic_swap_1d:
+; GFX9:       ; %bb.0: ; %main_body
+; GFX9-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 unorm glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX90A-LABEL: atomic_swap_1d:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    v_mov_b32_e32 v2, v1
+; GFX90A-NEXT:    image_atomic_swap v0, v2, s[0:7] dmask:0x1 unorm glc
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    ; return to shader part epilog
+;
+; GFX10-LABEL: atomic_swap_1d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    ; return to shader part epilog
+;
+; GFX9-FLATSCR-LABEL: atomic_swap_1d:
+; GFX9-FLATSCR:       ; %bb.0: ; %main_body
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s11, s9
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s10, s8
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s9, s7
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s8, s6
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s7, s5
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s6, s4
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s5, s3
+; GFX9-FLATSCR-NEXT:    s_mov_b32 s4, s2
+; GFX9-FLATSCR-NEXT:    image_atomic_swap v0, v1, s[4:11] dmask:0x1 unorm glc
+; GFX9-FLATSCR-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-FLATSCR-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: atomic_swap_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX12-LABEL: atomic_swap_1d:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    ; return to shader part epilog
+main_body:
+  %v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
+  %out = bitcast i32 %v to float
+  ret float %out
+}
+
+; from lds-bounds.ll
+; covers ds_write_b64 (atomic without return)
+@compute_lds = external addrspace(3) global [512 x i32], align 16
+;
+define amdgpu_cs void @store_aligned(ptr addrspace(3) %ptr) #0 {
+; GFX9-LABEL: store_aligned:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    v_mov_b32_e32 v1, 42
+; GFX9-NEXT:    v_mov_b32_e32 v2, 43
+; GFX9-NEXT:    ds_write_b64 v0, v[1:2]
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: store_aligned:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 42
+; GFX90A-NEXT:    v_mov_b32_e32 v3, 43
+; GFX90A-NEXT:    ds_write_b64 v0, v[2:3]
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: store_aligned:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    v_mov_b32_e32 v1, 42
+; GFX10-NEXT:    v_mov_b32_e32 v2, 43
+; GFX10-NEXT:    ds_write_b64 v0, v[1:2]
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: store_aligned:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, 42
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v2, 43
+; GFX9-FLATSCR-NEXT:    ds_write_b64 v0, v[1:2]
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_aligned:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    v_dual_mov_b32 v1, 42 :: v_dual_mov_b32 v2, 43
+; GFX11-NEXT:    ds_store_b64 v0, v[1:2]
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: store_aligned:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    v_dual_mov_b32 v1, 42 :: v_dual_mov_b32 v2, 43
+; GFX12-NEXT:    ds_store_b64 v0, v[1:2]
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    s_endpgm
+entry:
+  %ptr.gep.1 = getelementptr i32, ptr addrspace(3) %ptr, i32 1
+
+  store i32 42, ptr addrspace(3) %ptr, align 8
+  store i32 43, ptr addrspace(3) %ptr.gep.1
+  ret void
+}
+
+
+; from lds-bounds.ll
+; covers ds_read_b64
+;
+define amdgpu_cs <2 x float> @load_aligned(ptr addrspace(3) %ptr) #0 {
+; GFX9-LABEL: load_aligned:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    ds_read_b64 v[0:1], v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX90A-LABEL: load_aligned:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    ds_read_b64 v[0:1], v0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    ; return to shader part epilog
+;
+; GFX10-LABEL: load_aligned:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    ds_read_b64 v[0:1], v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    ; return to shader part epilog
+;
+; GFX9-FLATSCR-LABEL: load_aligned:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    ds_read_b64 v[0:1], v0
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: load_aligned:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    ds_load_b64 v[0:1], v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX12-LABEL: load_aligned:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    ds_load_b64 v[0:1], v0
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    ; return to shader part epilog
+entry:
+  %ptr.gep.1 = getelementptr i32, ptr addrspace(3) %ptr, i32 1
+
+  %v.0 = load i32, ptr addrspace(3) %ptr, align 8
+  %v.1 = load i32, ptr addrspace(3) %ptr.gep.1
+
+  %r.0 = insertelement <2 x i32> poison, i32 %v.0, i32 0
+  %r.1 = insertelement <2 x i32> %r.0, i32 %v.1, i32 1
+  %bc = bitcast <2 x i32> %r.1 to <2 x float>
+  ret <2 x float> %bc
+}
+
+; from lds-bounds.ll
+; covers ds_write2_b32
+;
+define amdgpu_cs void @store_global_const_idx() #0 {
+; GFX9-LABEL: store_global_const_idx:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX9-NEXT:    v_mov_b32_e32 v1, 42
+; GFX9-NEXT:    v_mov_b32_e32 v2, 43
+; GFX9-NEXT:    ds_write2_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: store_global_const_idx:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 42
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 43
+; GFX90A-NEXT:    ds_write2_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX10-LABEL: store_global_const_idx:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX10-NEXT:    v_mov_b32_e32 v1, 42
+; GFX10-NEXT:    v_mov_b32_e32 v2, 43
+; GFX10-NEXT:    ds_write2_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_endpgm
+;
+; GFX9-FLATSCR-LABEL: store_global_const_idx:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v1, 42
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v2, 43
+; GFX9-FLATSCR-NEXT:    ds_write2_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_global_const_idx:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    v_dual_mov_b32 v0, compute_lds@abs32@lo :: v_dual_mov_b32 v1, 42
+; GFX11-NEXT:    v_mov_b32_e32 v2, 43
+; GFX11-NEXT:    ds_store_2addr_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: store_global_const_idx:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    v_dual_mov_b32 v0, compute_lds@abs32@lo :: v_dual_mov_b32 v1, 42
+; GFX12-NEXT:    v_mov_b32_e32 v2, 43
+; GFX12-NEXT:    ds_store_2addr_b32 v0, v1, v2 offset0:3 offset1:4
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    s_endpgm
+entry:
+  %ptr.a = getelementptr [512 x i32], ptr addrspace(3) @compute_lds, i32 0, i32 3
+  %ptr.b = getelementptr [512 x i32], ptr addrspace(3) @compute_lds, i32 0, i32 4
+
+  store i32 42, ptr addrspace(3) %ptr.a
+  store i32 43, ptr addrspace(3) %ptr.b
+  ret void
+}
+
+; from lds-bounds.ll
+; covers ds_read2_b32
+;
+define amdgpu_cs <2 x float> @load_global_const_idx() #0 {
+; GFX9-LABEL: load_global_const_idx:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX9-NEXT:    ds_read2_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    ; return to shader part epilog
+;
+; GFX90A-LABEL: load_global_const_idx:
+; GFX90A:       ; %bb.0: ; %entry
+; GFX90A-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX90A-NEXT:    ds_read2_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    ; return to shader part epilog
+;
+; GFX10-LABEL: load_global_const_idx:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX10-NEXT:    ds_read2_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    ; return to shader part epilog
+;
+; GFX9-FLATSCR-LABEL: load_global_const_idx:
+; GFX9-FLATSCR:       ; %bb.0: ; %entry
+; GFX9-FLATSCR-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX9-FLATSCR-NEXT:    ds_read2_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX9-FLATSCR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-FLATSCR-NEXT:    ; return to shader part epilog
+;
+; GFX11-LABEL: load_global_const_idx:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    ; return to shader part epilog
+;
+; GFX12-LABEL: load_global_const_idx:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    v_mov_b32_e32 v0, compute_lds@abs32@lo
+; GFX12-NEXT:    ds_load_2addr_b32 v[0:1], v0 offset0:3 offset1:4
+; GFX12-NEXT:    s_wait_dscnt 0x0
+; GFX12-NEXT:    ; return to shader part epilog
+entry:
+  %ptr.a = getelementptr [512 x i32], ptr addrspace(3) @compute_lds, i32 0, i32 3
+  %ptr.b = getelementptr [512 x i32], ptr addrspace(3) @compute_lds, i32 0, i32 4
+
+  %v.0 = load i32, ptr addrspace(3) %ptr.a
+  %v.1 = load i32, ptr addrspace(3) %ptr.b
+
+  %r.0 = insertelement <2 x i32> poison, i32 %v.0, i32 0
+  %r.1 = insertelement <2 x i32> %r.0, i32 %v.1, i32 1
+  %bc = bitcast <2 x i32> %r.1 to <2 x float>
+  ret <2 x float> %bc
+}
+

From 4d80dff819d1164775d0d55fc68bffedb90ba53c Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Wed, 10 Apr 2024 13:56:18 -0400
Subject: [PATCH 045/886] int -> uintptr_t to silence diagnostics

'int' may not be sufficiently large to store a pointer representation
anyway, so this is also a correctness fix.
---
 clang/lib/AST/Interp/FunctionPointer.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/lib/AST/Interp/FunctionPointer.h b/clang/lib/AST/Interp/FunctionPointer.h
index e7fad8161fd9c..f61f9ded0bf00 100644
--- a/clang/lib/AST/Interp/FunctionPointer.h
+++ b/clang/lib/AST/Interp/FunctionPointer.h
@@ -24,7 +24,7 @@ class FunctionPointer final {
 public:
   // FIXME: We might want to track the fact that the Function pointer
   // has been created from an integer and is most likely garbage anyway.
-  FunctionPointer(int IntVal = 0, const Descriptor *Desc = nullptr)
+  FunctionPointer(uintptr_t IntVal = 0, const Descriptor *Desc = nullptr)
       : Func(reinterpret_cast<const Function *>(IntVal)) {}
 
   FunctionPointer(const Function *Func) : Func(Func) { assert(Func); }

From 21009f466ece9f21b18e1bb03bd74b566188bae5 Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Wed, 10 Apr 2024 20:03:35 +0200
Subject: [PATCH 046/886] [clang][dataflow] Propagate locations from result
 objects to initializers. (#87320)

Previously, we were propagating storage locations the other way around,
i.e.
from initializers to result objects, using `RecordValue::getLoc()`. This
gave
the wrong behavior in some cases -- see the newly added or fixed tests
in this
patch.

In addition, this patch now unblocks removing the `RecordValue` class
entirely,
as we no longer need `RecordValue::getLoc()`.

With this patch, the test `TransferTest.DifferentReferenceLocInJoin`
started to
fail because the framework now always uses the same storge location for
a
`MaterializeTemporaryExpr`, meaning that the code under test no longer
set up
the desired state where a variable of reference type is mapped to two
different
storage locations in environments being joined. Rather than trying to
modify
this test to set up the test condition again, I have chosen to replace
the test
with an equivalent test in DataflowEnvironmentTest.cpp that sets up the
test
condition directly; because this test is more direct, it will also be
less
brittle in the face of future changes.
---
 .../FlowSensitive/DataflowEnvironment.h       |  64 ++-
 .../FlowSensitive/DataflowEnvironment.cpp     | 405 +++++++++++++-----
 clang/lib/Analysis/FlowSensitive/Transfer.cpp | 176 ++++----
 .../TypeErasedDataflowAnalysis.cpp            |  13 +-
 .../FlowSensitive/DataflowEnvironmentTest.cpp |  43 ++
 .../Analysis/FlowSensitive/TransferTest.cpp   | 172 +++++---
 6 files changed, 590 insertions(+), 283 deletions(-)

diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
index 9a65f76cdf56b..706664d7db1c2 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
@@ -30,6 +30,7 @@
 #include "llvm/ADT/MapVector.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/ErrorHandling.h"
+#include <memory>
 #include <type_traits>
 #include <utility>
 
@@ -344,17 +345,6 @@ class Environment {
   /// location of the result object to pass in `this`, even though prvalues are
   /// otherwise not associated with storage locations.
   ///
-  /// FIXME: Currently, this simply returns a stable storage location for `E`,
-  /// but this doesn't do the right thing in scenarios like the following:
-  /// ```
-  /// MyClass c = some_condition()? MyClass(foo) : MyClass(bar);
-  /// ```
-  /// Here, `MyClass(foo)` and `MyClass(bar)` will have two different storage
-  /// locations, when in fact their storage locations should be the same.
-  /// Eventually, we want to propagate storage locations from result objects
-  /// down to the prvalues that initialize them, similar to the way that this is
-  /// done in Clang's CodeGen.
-  ///
   /// Requirements:
   ///  `E` must be a prvalue of record type.
   RecordStorageLocation &
@@ -462,7 +452,13 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values.
-  void initializeFieldsWithValues(RecordStorageLocation &Loc);
+  /// If `Type` is provided, initializes only those fields that are modeled for
+  /// `Type`; this is intended for use in cases where `Loc` is a derived type
+  /// and we only want to initialize the fields of a base type.
+  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type);
+  void initializeFieldsWithValues(RecordStorageLocation &Loc) {
+    initializeFieldsWithValues(Loc, Loc.getType());
+  }
 
   /// Assigns `Val` as the value of `Loc` in the environment.
   void setValue(const StorageLocation &Loc, Value &Val);
@@ -653,6 +649,9 @@ class Environment {
   LLVM_DUMP_METHOD void dump(raw_ostream &OS) const;
 
 private:
+  using PrValueToResultObject =
+      llvm::DenseMap<const Expr *, RecordStorageLocation *>;
+
   // The copy-constructor is for use in fork() only.
   Environment(const Environment &) = default;
 
@@ -682,8 +681,10 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values (controlled by `Visited`,
-  /// `Depth`, and `CreatedValuesCount`).
-  void initializeFieldsWithValues(RecordStorageLocation &Loc,
+  /// `Depth`, and `CreatedValuesCount`). If `Type` is different from
+  /// `Loc.getType()`, initializes only those fields that are modeled for
+  /// `Type`.
+  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type,
                                   llvm::DenseSet<QualType> &Visited, int Depth,
                                   int &CreatedValuesCount);
 
@@ -702,22 +703,45 @@ class Environment {
   /// and functions referenced in `FuncDecl`. `FuncDecl` must have a body.
   void initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl);
 
+  static PrValueToResultObject
+  buildResultObjectMap(DataflowAnalysisContext *DACtx,
+                       const FunctionDecl *FuncDecl,
+                       RecordStorageLocation *ThisPointeeLoc,
+                       RecordStorageLocation *LocForRecordReturnVal);
+
   // `DACtx` is not null and not owned by this object.
   DataflowAnalysisContext *DACtx;
 
-  // FIXME: move the fields `CallStack`, `ReturnVal`, `ReturnLoc` and
-  // `ThisPointeeLoc` into a separate call-context object, shared between
-  // environments in the same call.
+  // FIXME: move the fields `CallStack`, `ResultObjectMap`, `ReturnVal`,
+  // `ReturnLoc` and `ThisPointeeLoc` into a separate call-context object,
+  // shared between environments in the same call.
   // https://github.com/llvm/llvm-project/issues/59005
 
   // `DeclContext` of the block being analysed if provided.
   std::vector<const DeclContext *> CallStack;
 
-  // Value returned by the function (if it has non-reference return type).
+  // Maps from prvalues of record type to their result objects. Shared between
+  // all environments for the same function.
+  // FIXME: It's somewhat unsatisfactory that we have to use a `shared_ptr`
+  // here, though the cost is acceptable: The overhead of a `shared_ptr` is
+  // incurred when it is copied, and this happens only relatively rarely (when
+  // we fork the environment). The need for a `shared_ptr` will go away once we
+  // introduce a shared call-context object (see above).
+  std::shared_ptr<PrValueToResultObject> ResultObjectMap;
+
+  // The following three member variables handle various different types of
+  // return values.
+  // - If the return type is not a reference and not a record: Value returned
+  //   by the function.
   Value *ReturnVal = nullptr;
-  // Storage location of the reference returned by the function (if it has
-  // reference return type).
+  // - If the return type is a reference: Storage location of the reference
+  //   returned by the function.
   StorageLocation *ReturnLoc = nullptr;
+  // - If the return type is a record or the function being analyzed is a
+  //   constructor: Storage location into which the return value should be
+  //   constructed.
+  RecordStorageLocation *LocForRecordReturnVal = nullptr;
+
   // The storage location of the `this` pointee. Should only be null if the
   // function being analyzed is only a function and not a method.
   RecordStorageLocation *ThisPointeeLoc = nullptr;
diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index 1bfa7ebcfd50c..6c796b4ad923e 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -15,6 +15,7 @@
 #include "clang/Analysis/FlowSensitive/DataflowEnvironment.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/DeclCXX.h"
+#include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/AST/Type.h"
 #include "clang/Analysis/FlowSensitive/DataflowLattice.h"
 #include "clang/Analysis/FlowSensitive/Value.h"
@@ -26,6 +27,8 @@
 #include <cassert>
 #include <utility>
 
+#define DEBUG_TYPE "dataflow"
+
 namespace clang {
 namespace dataflow {
 
@@ -354,6 +357,8 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   for (auto *Child : S.children())
     if (Child != nullptr)
       getFieldsGlobalsAndFuncs(*Child, Fields, Vars, Funcs);
+  if (const auto *DefaultArg = dyn_cast<CXXDefaultArgExpr>(&S))
+    getFieldsGlobalsAndFuncs(*DefaultArg->getExpr(), Fields, Vars, Funcs);
   if (const auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(&S))
     getFieldsGlobalsAndFuncs(*DefaultInit->getExpr(), Fields, Vars, Funcs);
 
@@ -386,6 +391,186 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   }
 }
 
+namespace {
+
+// Visitor that builds a map from record prvalues to result objects.
+// This traverses the body of the function to be analyzed; for each result
+// object that it encounters, it propagates the storage location of the result
+// object to all record prvalues that can initialize it.
+class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
+public:
+  // `ResultObjectMap` will be filled with a map from record prvalues to result
+  // object. If the function being analyzed returns a record by value,
+  // `LocForRecordReturnVal` is the location to which this record should be
+  // written; otherwise, it is null.
+  explicit ResultObjectVisitor(
+      llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap,
+      RecordStorageLocation *LocForRecordReturnVal,
+      DataflowAnalysisContext &DACtx)
+      : ResultObjectMap(ResultObjectMap),
+        LocForRecordReturnVal(LocForRecordReturnVal), DACtx(DACtx) {}
+
+  bool shouldVisitImplicitCode() { return true; }
+
+  bool shouldVisitLambdaBody() const { return false; }
+
+  // Traverse all member and base initializers of `Ctor`. This function is not
+  // called by `RecursiveASTVisitor`; it should be called manually if we are
+  // analyzing a constructor. `ThisPointeeLoc` is the storage location that
+  // `this` points to.
+  void TraverseConstructorInits(const CXXConstructorDecl *Ctor,
+                                RecordStorageLocation *ThisPointeeLoc) {
+    assert(ThisPointeeLoc != nullptr);
+    for (const CXXCtorInitializer *Init : Ctor->inits()) {
+      Expr *InitExpr = Init->getInit();
+      if (FieldDecl *Field = Init->getMember();
+          Field != nullptr && Field->getType()->isRecordType()) {
+        PropagateResultObject(InitExpr, cast<RecordStorageLocation>(
+                                            ThisPointeeLoc->getChild(*Field)));
+      } else if (Init->getBaseClass()) {
+        PropagateResultObject(InitExpr, ThisPointeeLoc);
+      }
+
+      // Ensure that any result objects within `InitExpr` (e.g. temporaries)
+      // are also propagated to the prvalues that initialize them.
+      TraverseStmt(InitExpr);
+
+      // If this is a `CXXDefaultInitExpr`, also propagate any result objects
+      // within the default expression.
+      if (auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(InitExpr))
+        TraverseStmt(DefaultInit->getExpr());
+    }
+  }
+
+  bool TraverseBindingDecl(BindingDecl *BD) {
+    // `RecursiveASTVisitor` doesn't traverse holding variables for
+    // `BindingDecl`s by itself, so we need to tell it to.
+    if (VarDecl *HoldingVar = BD->getHoldingVar())
+      TraverseDecl(HoldingVar);
+    return RecursiveASTVisitor<ResultObjectVisitor>::TraverseBindingDecl(BD);
+  }
+
+  bool VisitVarDecl(VarDecl *VD) {
+    if (VD->getType()->isRecordType() && VD->hasInit())
+      PropagateResultObject(
+          VD->getInit(),
+          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*VD)));
+    return true;
+  }
+
+  bool VisitMaterializeTemporaryExpr(MaterializeTemporaryExpr *MTE) {
+    if (MTE->getType()->isRecordType())
+      PropagateResultObject(
+          MTE->getSubExpr(),
+          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*MTE)));
+    return true;
+  }
+
+  bool VisitReturnStmt(ReturnStmt *Return) {
+    Expr *RetValue = Return->getRetValue();
+    if (RetValue != nullptr && RetValue->getType()->isRecordType() &&
+        RetValue->isPRValue())
+      PropagateResultObject(RetValue, LocForRecordReturnVal);
+    return true;
+  }
+
+  bool VisitExpr(Expr *E) {
+    // Clang's AST can have record-type prvalues without a result object -- for
+    // example as full-expressions contained in a compound statement or as
+    // arguments of call expressions. We notice this if we get here and a
+    // storage location has not yet been associated with `E`. In this case,
+    // treat this as if it was a `MaterializeTemporaryExpr`.
+    if (E->isPRValue() && E->getType()->isRecordType() &&
+        !ResultObjectMap.contains(E))
+      PropagateResultObject(
+          E, &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*E)));
+    return true;
+  }
+
+  // Assigns `Loc` as the result object location of `E`, then propagates the
+  // location to all lower-level prvalues that initialize the same object as
+  // `E` (or one of its base classes or member variables).
+  void PropagateResultObject(Expr *E, RecordStorageLocation *Loc) {
+    if (!E->isPRValue() || !E->getType()->isRecordType()) {
+      assert(false);
+      // Ensure we don't propagate the result object if we hit this in a
+      // release build.
+      return;
+    }
+
+    ResultObjectMap[E] = Loc;
+
+    // The following AST node kinds are "original initializers": They are the
+    // lowest-level AST node that initializes a given object, and nothing
+    // below them can initialize the same object (or part of it).
+    if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
+        isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
+        isa<CXXStdInitializerListExpr>(E)) {
+      return;
+    }
+
+    if (auto *InitList = dyn_cast<InitListExpr>(E)) {
+      if (!InitList->isSemanticForm())
+        return;
+      if (InitList->isTransparent()) {
+        PropagateResultObject(InitList->getInit(0), Loc);
+        return;
+      }
+
+      RecordInitListHelper InitListHelper(InitList);
+
+      for (auto [Base, Init] : InitListHelper.base_inits()) {
+        assert(Base->getType().getCanonicalType() ==
+               Init->getType().getCanonicalType());
+
+        // Storage location for the base class is the same as that of the
+        // derived class because we "flatten" the object hierarchy and put all
+        // fields in `RecordStorageLocation` of the derived class.
+        PropagateResultObject(Init, Loc);
+      }
+
+      for (auto [Field, Init] : InitListHelper.field_inits()) {
+        // Fields of non-record type are handled in
+        // `TransferVisitor::VisitInitListExpr()`.
+        if (!Field->getType()->isRecordType())
+          continue;
+        PropagateResultObject(
+            Init, cast<RecordStorageLocation>(Loc->getChild(*Field)));
+      }
+      return;
+    }
+
+    if (auto *Op = dyn_cast<BinaryOperator>(E); Op && Op->isCommaOp()) {
+      PropagateResultObject(Op->getRHS(), Loc);
+      return;
+    }
+
+    if (auto *Cond = dyn_cast<AbstractConditionalOperator>(E)) {
+      PropagateResultObject(Cond->getTrueExpr(), Loc);
+      PropagateResultObject(Cond->getFalseExpr(), Loc);
+      return;
+    }
+
+    // All other expression nodes that propagate a record prvalue should have
+    // exactly one child.
+    SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
+    LLVM_DEBUG({
+      if (Children.size() != 1)
+        E->dump();
+    });
+    assert(Children.size() == 1);
+    for (Stmt *S : Children)
+      PropagateResultObject(cast<Expr>(S), Loc);
+  }
+
+private:
+  llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap;
+  RecordStorageLocation *LocForRecordReturnVal;
+  DataflowAnalysisContext &DACtx;
+};
+
+} // namespace
+
 Environment::Environment(DataflowAnalysisContext &DACtx)
     : DACtx(&DACtx),
       FlowConditionToken(DACtx.arena().makeFlowConditionToken()) {}
@@ -401,17 +586,23 @@ void Environment::initialize() {
   if (DeclCtx == nullptr)
     return;
 
-  if (const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx)) {
-    assert(FuncDecl->doesThisDeclarationHaveABody());
+  const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx);
+  if (FuncDecl == nullptr)
+    return;
+
+  assert(FuncDecl->doesThisDeclarationHaveABody());
 
-    initFieldsGlobalsAndFuncs(FuncDecl);
+  initFieldsGlobalsAndFuncs(FuncDecl);
 
-    for (const auto *ParamDecl : FuncDecl->parameters()) {
-      assert(ParamDecl != nullptr);
-      setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
-    }
+  for (const auto *ParamDecl : FuncDecl->parameters()) {
+    assert(ParamDecl != nullptr);
+    setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
   }
 
+  if (FuncDecl->getReturnType()->isRecordType())
+    LocForRecordReturnVal = &cast<RecordStorageLocation>(
+        createStorageLocation(FuncDecl->getReturnType()));
+
   if (const auto *MethodDecl = dyn_cast<CXXMethodDecl>(DeclCtx)) {
     auto *Parent = MethodDecl->getParent();
     assert(Parent != nullptr);
@@ -444,6 +635,12 @@ void Environment::initialize() {
         initializeFieldsWithValues(ThisLoc);
     }
   }
+
+  // We do this below the handling of `CXXMethodDecl` above so that we can
+  // be sure that the storage location for `this` has been set.
+  ResultObjectMap = std::make_shared<PrValueToResultObject>(
+      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
+                           LocForRecordReturnVal));
 }
 
 // FIXME: Add support for resetting globals after function calls to enable
@@ -484,13 +681,18 @@ void Environment::initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl) {
     if (getStorageLocation(*D) != nullptr)
       continue;
 
-    setStorageLocation(*D, createObject(*D));
+    // We don't run transfer functions on the initializers of global variables,
+    // so they won't be associated with a value or storage location. We
+    // therefore intentionally don't pass an initializer to `createObject()`;
+    // in particular, this ensures that `createObject()` will initialize the
+    // fields of record-type variables with values.
+    setStorageLocation(*D, createObject(*D, nullptr));
   }
 
   for (const FunctionDecl *FD : Funcs) {
     if (getStorageLocation(*FD) != nullptr)
       continue;
-    auto &Loc = createStorageLocation(FD->getType());
+    auto &Loc = createStorageLocation(*FD);
     setStorageLocation(*FD, Loc);
   }
 }
@@ -519,6 +721,9 @@ Environment Environment::pushCall(const CallExpr *Call) const {
     }
   }
 
+  if (Call->getType()->isRecordType() && Call->isPRValue())
+    Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
+
   Env.pushCallInternal(Call->getDirectCallee(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
 
@@ -529,6 +734,7 @@ Environment Environment::pushCall(const CXXConstructExpr *Call) const {
   Environment Env(*this);
 
   Env.ThisPointeeLoc = &Env.getResultObjectLocation(*Call);
+  Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
 
   Env.pushCallInternal(Call->getConstructor(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
@@ -557,6 +763,10 @@ void Environment::pushCallInternal(const FunctionDecl *FuncDecl,
     const VarDecl *Param = *ParamIt;
     setStorageLocation(*Param, createObject(*Param, Args[ArgIndex]));
   }
+
+  ResultObjectMap = std::make_shared<PrValueToResultObject>(
+      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
+                           LocForRecordReturnVal));
 }
 
 void Environment::popCall(const CallExpr *Call, const Environment &CalleeEnv) {
@@ -600,6 +810,9 @@ bool Environment::equivalentTo(const Environment &Other,
   if (ReturnLoc != Other.ReturnLoc)
     return false;
 
+  if (LocForRecordReturnVal != Other.LocForRecordReturnVal)
+    return false;
+
   if (ThisPointeeLoc != Other.ThisPointeeLoc)
     return false;
 
@@ -623,8 +836,10 @@ LatticeEffect Environment::widen(const Environment &PrevEnv,
   assert(DACtx == PrevEnv.DACtx);
   assert(ReturnVal == PrevEnv.ReturnVal);
   assert(ReturnLoc == PrevEnv.ReturnLoc);
+  assert(LocForRecordReturnVal == PrevEnv.LocForRecordReturnVal);
   assert(ThisPointeeLoc == PrevEnv.ThisPointeeLoc);
   assert(CallStack == PrevEnv.CallStack);
+  assert(ResultObjectMap == PrevEnv.ResultObjectMap);
 
   auto Effect = LatticeEffect::Unchanged;
 
@@ -656,12 +871,16 @@ Environment Environment::join(const Environment &EnvA, const Environment &EnvB,
                               Environment::ValueModel &Model,
                               ExprJoinBehavior ExprBehavior) {
   assert(EnvA.DACtx == EnvB.DACtx);
+  assert(EnvA.LocForRecordReturnVal == EnvB.LocForRecordReturnVal);
   assert(EnvA.ThisPointeeLoc == EnvB.ThisPointeeLoc);
   assert(EnvA.CallStack == EnvB.CallStack);
+  assert(EnvA.ResultObjectMap == EnvB.ResultObjectMap);
 
   Environment JoinedEnv(*EnvA.DACtx);
 
   JoinedEnv.CallStack = EnvA.CallStack;
+  JoinedEnv.ResultObjectMap = EnvA.ResultObjectMap;
+  JoinedEnv.LocForRecordReturnVal = EnvA.LocForRecordReturnVal;
   JoinedEnv.ThisPointeeLoc = EnvA.ThisPointeeLoc;
 
   if (EnvA.ReturnVal == nullptr || EnvB.ReturnVal == nullptr) {
@@ -730,6 +949,12 @@ StorageLocation &Environment::createStorageLocation(const Expr &E) {
 
 void Environment::setStorageLocation(const ValueDecl &D, StorageLocation &Loc) {
   assert(!DeclToLoc.contains(&D));
+  // The only kinds of declarations that may have a "variable" storage location
+  // are declarations of reference type and `BindingDecl`. For all other
+  // declaration, the storage location should be the stable storage location
+  // returned by `createStorageLocation()`.
+  assert(D.getType()->isReferenceType() || isa<BindingDecl>(D) ||
+         &Loc == &createStorageLocation(D));
   DeclToLoc[&D] = &Loc;
 }
 
@@ -791,50 +1016,29 @@ Environment::getResultObjectLocation(const Expr &RecordPRValue) const {
   assert(RecordPRValue.getType()->isRecordType());
   assert(RecordPRValue.isPRValue());
 
-  // Returns a storage location that we can use if assertions fail.
-  auto FallbackForAssertFailure =
-      [this, &RecordPRValue]() -> RecordStorageLocation & {
+  assert(ResultObjectMap != nullptr);
+  RecordStorageLocation *Loc = ResultObjectMap->lookup(&RecordPRValue);
+  assert(Loc != nullptr);
+  // In release builds, use the "stable" storage location if the map lookup
+  // failed.
+  if (Loc == nullptr)
     return cast<RecordStorageLocation>(
         DACtx->getStableStorageLocation(RecordPRValue));
-  };
-
-  if (isOriginalRecordConstructor(RecordPRValue)) {
-    auto *Val = cast_or_null<RecordValue>(getValue(RecordPRValue));
-    // The builtin transfer function should have created a `RecordValue` for all
-    // original record constructors.
-    assert(Val);
-    if (!Val)
-      return FallbackForAssertFailure();
-    return Val->getLoc();
-  }
-
-  if (auto *Op = dyn_cast<BinaryOperator>(&RecordPRValue);
-      Op && Op->isCommaOp()) {
-    return getResultObjectLocation(*Op->getRHS());
-  }
-
-  // All other expression nodes that propagate a record prvalue should have
-  // exactly one child.
-  llvm::SmallVector<const Stmt *> children(RecordPRValue.child_begin(),
-                                           RecordPRValue.child_end());
-  assert(children.size() == 1);
-  if (children.empty())
-    return FallbackForAssertFailure();
-
-  return getResultObjectLocation(*cast<Expr>(children[0]));
+  return *Loc;
 }
 
 PointerValue &Environment::getOrCreateNullPointerValue(QualType PointeeType) {
   return DACtx->getOrCreateNullPointerValue(PointeeType);
 }
 
-void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc) {
+void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
+                                             QualType Type) {
   llvm::DenseSet<QualType> Visited;
   int CreatedValuesCount = 0;
-  initializeFieldsWithValues(Loc, Visited, 0, CreatedValuesCount);
+  initializeFieldsWithValues(Loc, Type, Visited, 0, CreatedValuesCount);
   if (CreatedValuesCount > MaxCompositeValueSize) {
-    llvm::errs() << "Attempting to initialize a huge value of type: "
-                 << Loc.getType() << '\n';
+    llvm::errs() << "Attempting to initialize a huge value of type: " << Type
+                 << '\n';
   }
 }
 
@@ -848,8 +1052,7 @@ void Environment::setValue(const Expr &E, Value &Val) {
   const Expr &CanonE = ignoreCFGOmittedNodes(E);
 
   if (auto *RecordVal = dyn_cast<RecordValue>(&Val)) {
-    assert(isOriginalRecordConstructor(CanonE) ||
-           &RecordVal->getLoc() == &getResultObjectLocation(CanonE));
+    assert(&RecordVal->getLoc() == &getResultObjectLocation(CanonE));
     (void)RecordVal;
   }
 
@@ -928,7 +1131,8 @@ Value *Environment::createValueUnlessSelfReferential(
   if (Type->isRecordType()) {
     CreatedValuesCount++;
     auto &Loc = cast<RecordStorageLocation>(createStorageLocation(Type));
-    initializeFieldsWithValues(Loc, Visited, Depth, CreatedValuesCount);
+    initializeFieldsWithValues(Loc, Loc.getType(), Visited, Depth,
+                               CreatedValuesCount);
 
     return &refreshRecordValue(Loc, *this);
   }
@@ -960,6 +1164,7 @@ Environment::createLocAndMaybeValue(QualType Ty,
 }
 
 void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
+                                             QualType Type,
                                              llvm::DenseSet<QualType> &Visited,
                                              int Depth,
                                              int &CreatedValuesCount) {
@@ -967,8 +1172,8 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     if (FieldType->isRecordType()) {
       auto &FieldRecordLoc = cast<RecordStorageLocation>(FieldLoc);
       setValue(FieldRecordLoc, create<RecordValue>(FieldRecordLoc));
-      initializeFieldsWithValues(FieldRecordLoc, Visited, Depth + 1,
-                                 CreatedValuesCount);
+      initializeFieldsWithValues(FieldRecordLoc, FieldRecordLoc.getType(),
+                                 Visited, Depth + 1, CreatedValuesCount);
     } else {
       if (!Visited.insert(FieldType.getCanonicalType()).second)
         return;
@@ -979,7 +1184,7 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     }
   };
 
-  for (const auto &[Field, FieldLoc] : Loc.children()) {
+  for (const FieldDecl *Field : DACtx->getModeledFields(Type)) {
     assert(Field != nullptr);
     QualType FieldType = Field->getType();
 
@@ -988,14 +1193,12 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
                    &createLocAndMaybeValue(FieldType, Visited, Depth + 1,
                                            CreatedValuesCount));
     } else {
+      StorageLocation *FieldLoc = Loc.getChild(*Field);
       assert(FieldLoc != nullptr);
       initField(FieldType, *FieldLoc);
     }
   }
-  for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
-    assert(FieldLoc != nullptr);
-    QualType FieldType = FieldLoc->getType();
-
+  for (const auto &[FieldName, FieldType] : DACtx->getSyntheticFields(Type)) {
     // Synthetic fields cannot have reference type, so we don't need to deal
     // with this case.
     assert(!FieldType->isReferenceType());
@@ -1022,38 +1225,36 @@ StorageLocation &Environment::createObjectInternal(const ValueDecl *D,
     return createObjectInternal(D, Ty.getNonReferenceType(), nullptr);
   }
 
-  Value *Val = nullptr;
-  if (InitExpr) {
-    // In the (few) cases where an expression is intentionally
-    // "uninterpreted", `InitExpr` is not associated with a value.  There are
-    // two ways to handle this situation: propagate the status, so that
-    // uninterpreted initializers result in uninterpreted variables, or
-    // provide a default value. We choose the latter so that later refinements
-    // of the variable can be used for reasoning about the surrounding code.
-    // For this reason, we let this case be handled by the `createValue()`
-    // call below.
-    //
-    // FIXME. If and when we interpret all language cases, change this to
-    // assert that `InitExpr` is interpreted, rather than supplying a
-    // default value (assuming we don't update the environment API to return
-    // references).
-    Val = getValue(*InitExpr);
-
-    if (!Val && isa<ImplicitValueInitExpr>(InitExpr) &&
-        InitExpr->getType()->isPointerType())
-      Val = &getOrCreateNullPointerValue(InitExpr->getType()->getPointeeType());
-  }
-  if (!Val)
-    Val = createValue(Ty);
-
-  if (Ty->isRecordType())
-    return cast<RecordValue>(Val)->getLoc();
-
   StorageLocation &Loc =
       D ? createStorageLocation(*D) : createStorageLocation(Ty);
 
-  if (Val)
-    setValue(Loc, *Val);
+  if (Ty->isRecordType()) {
+    auto &RecordLoc = cast<RecordStorageLocation>(Loc);
+    if (!InitExpr)
+      initializeFieldsWithValues(RecordLoc);
+    refreshRecordValue(RecordLoc, *this);
+  } else {
+    Value *Val = nullptr;
+    if (InitExpr)
+      // In the (few) cases where an expression is intentionally
+      // "uninterpreted", `InitExpr` is not associated with a value.  There are
+      // two ways to handle this situation: propagate the status, so that
+      // uninterpreted initializers result in uninterpreted variables, or
+      // provide a default value. We choose the latter so that later refinements
+      // of the variable can be used for reasoning about the surrounding code.
+      // For this reason, we let this case be handled by the `createValue()`
+      // call below.
+      //
+      // FIXME. If and when we interpret all language cases, change this to
+      // assert that `InitExpr` is interpreted, rather than supplying a
+      // default value (assuming we don't update the environment API to return
+      // references).
+      Val = getValue(*InitExpr);
+    if (!Val)
+      Val = createValue(Ty);
+    if (Val)
+      setValue(Loc, *Val);
+  }
 
   return Loc;
 }
@@ -1072,6 +1273,8 @@ bool Environment::allows(const Formula &F) const {
 
 void Environment::dump(raw_ostream &OS) const {
   llvm::DenseMap<const StorageLocation *, std::string> LocToName;
+  if (LocForRecordReturnVal != nullptr)
+    LocToName[LocForRecordReturnVal] = "(returned record)";
   if (ThisPointeeLoc != nullptr)
     LocToName[ThisPointeeLoc] = "this";
 
@@ -1102,6 +1305,9 @@ void Environment::dump(raw_ostream &OS) const {
       if (auto Iter = LocToName.find(ReturnLoc); Iter != LocToName.end())
         OS << " (" << Iter->second << ")";
       OS << "\n";
+    } else if (Func->getReturnType()->isRecordType() ||
+               isa<CXXConstructorDecl>(Func)) {
+      OS << "LocForRecordReturnVal: " << LocForRecordReturnVal << "\n";
     } else if (!Func->getReturnType()->isVoidType()) {
       if (ReturnVal == nullptr)
         OS << "ReturnVal: nullptr\n";
@@ -1122,6 +1328,22 @@ void Environment::dump() const {
   dump(llvm::dbgs());
 }
 
+Environment::PrValueToResultObject Environment::buildResultObjectMap(
+    DataflowAnalysisContext *DACtx, const FunctionDecl *FuncDecl,
+    RecordStorageLocation *ThisPointeeLoc,
+    RecordStorageLocation *LocForRecordReturnVal) {
+  assert(FuncDecl->doesThisDeclarationHaveABody());
+
+  PrValueToResultObject Map;
+
+  ResultObjectVisitor Visitor(Map, LocForRecordReturnVal, *DACtx);
+  if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(FuncDecl))
+    Visitor.TraverseConstructorInits(Ctor, ThisPointeeLoc);
+  Visitor.TraverseStmt(FuncDecl->getBody());
+
+  return Map;
+}
+
 RecordStorageLocation *getImplicitObjectLocation(const CXXMemberCallExpr &MCE,
                                                  const Environment &Env) {
   Expr *ImplicitObject = MCE.getImplicitObjectArgument();
@@ -1216,24 +1438,11 @@ RecordValue &refreshRecordValue(RecordStorageLocation &Loc, Environment &Env) {
 RecordValue &refreshRecordValue(const Expr &Expr, Environment &Env) {
   assert(Expr.getType()->isRecordType());
 
-  if (Expr.isPRValue()) {
-    if (auto *ExistingVal = Env.get<RecordValue>(Expr)) {
-      auto &NewVal = Env.create<RecordValue>(ExistingVal->getLoc());
-      Env.setValue(Expr, NewVal);
-      Env.setValue(NewVal.getLoc(), NewVal);
-      return NewVal;
-    }
+  if (Expr.isPRValue())
+    refreshRecordValue(Env.getResultObjectLocation(Expr), Env);
 
-    auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
-    Env.setValue(Expr, NewVal);
-    return NewVal;
-  }
-
-  if (auto *Loc = Env.get<RecordStorageLocation>(Expr)) {
-    auto &NewVal = Env.create<RecordValue>(*Loc);
-    Env.setValue(*Loc, NewVal);
-    return NewVal;
-  }
+  if (auto *Loc = Env.get<RecordStorageLocation>(Expr))
+    refreshRecordValue(*Loc, Env);
 
   auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
   Env.setStorageLocation(Expr, NewVal.getLoc());
diff --git a/clang/lib/Analysis/FlowSensitive/Transfer.cpp b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
index 0a2e8368d541d..88a9c0eccbebc 100644
--- a/clang/lib/Analysis/FlowSensitive/Transfer.cpp
+++ b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
@@ -460,11 +460,9 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // So make sure we have a value if we didn't propagate one above.
     if (S->isPRValue() && S->getType()->isRecordType()) {
       if (Env.getValue(*S) == nullptr) {
-        Value *Val = Env.createValue(S->getType());
-        // We're guaranteed to always be able to create a value for record
-        // types.
-        assert(Val != nullptr);
-        Env.setValue(*S, *Val);
+        auto &Loc = Env.getResultObjectLocation(*S);
+        Env.initializeFieldsWithValues(Loc);
+        refreshRecordValue(Loc, Env);
       }
     }
   }
@@ -472,6 +470,13 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
   void VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *S) {
     const Expr *InitExpr = S->getExpr();
     assert(InitExpr != nullptr);
+
+    // If this is a prvalue of record type, the handler for `*InitExpr` (if one
+    // exists) will initialize the result object; there is no value to propgate
+    // here.
+    if (S->getType()->isRecordType() && S->isPRValue())
+      return;
+
     propagateValueOrStorageLocation(*InitExpr, *S, Env);
   }
 
@@ -479,6 +484,17 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     const CXXConstructorDecl *ConstructorDecl = S->getConstructor();
     assert(ConstructorDecl != nullptr);
 
+    // `CXXConstructExpr` can have array type if default-initializing an array
+    // of records. We don't handle this specifically beyond potentially inlining
+    // the call.
+    if (!S->getType()->isRecordType()) {
+      transferInlineCall(S, ConstructorDecl);
+      return;
+    }
+
+    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+    Env.setValue(*S, refreshRecordValue(Loc, Env));
+
     if (ConstructorDecl->isCopyOrMoveConstructor()) {
       // It is permissible for a copy/move constructor to have additional
       // parameters as long as they have default arguments defined for them.
@@ -491,24 +507,14 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (ArgLoc == nullptr)
         return;
 
-      if (S->isElidable()) {
-        if (Value *Val = Env.getValue(*ArgLoc))
-          Env.setValue(*S, *Val);
-      } else {
-        auto &Val = *cast<RecordValue>(Env.createValue(S->getType()));
-        Env.setValue(*S, Val);
-        copyRecord(*ArgLoc, Val.getLoc(), Env);
-      }
+      // Even if the copy/move constructor call is elidable, we choose to copy
+      // the record in all cases (which isn't wrong, just potentially not
+      // optimal).
+      copyRecord(*ArgLoc, Loc, Env);
       return;
     }
 
-    // `CXXConstructExpr` can have array type if default-initializing an array
-    // of records, and we currently can't create values for arrays. So check if
-    // we've got a record type.
-    if (S->getType()->isRecordType()) {
-      auto &InitialVal = *cast<RecordValue>(Env.createValue(S->getType()));
-      Env.setValue(*S, InitialVal);
-    }
+    Env.initializeFieldsWithValues(Loc, S->getType());
 
     transferInlineCall(S, ConstructorDecl);
   }
@@ -551,19 +557,15 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (S->isGLValue()) {
         Env.setStorageLocation(*S, *LocDst);
       } else if (S->getType()->isRecordType()) {
-        // Make sure that we have a `RecordValue` for this expression so that
-        // `Environment::getResultObjectLocation()` is able to return a location
-        // for it.
-        if (Env.getValue(*S) == nullptr)
-          refreshRecordValue(*S, Env);
+        // Assume that the assignment returns the assigned value.
+        copyRecord(*LocDst, Env.getResultObjectLocation(*S), Env);
       }
 
       return;
     }
 
-    // CXXOperatorCallExpr can be prvalues. Call `VisitCallExpr`() to create
-    // a `RecordValue` for them so that `Environment::getResultObjectLocation()`
-    // can return a value.
+    // `CXXOperatorCallExpr` can be a prvalue. Call `VisitCallExpr`() to
+    // initialize the prvalue's fields with values.
     VisitCallExpr(S);
   }
 
@@ -580,11 +582,6 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     }
   }
 
-  void VisitCXXTemporaryObjectExpr(const CXXTemporaryObjectExpr *S) {
-    if (Value *Val = Env.createValue(S->getType()))
-      Env.setValue(*S, *Val);
-  }
-
   void VisitCallExpr(const CallExpr *S) {
     // Of clang's builtins, only `__builtin_expect` is handled explicitly, since
     // others (like trap, debugtrap, and unreachable) are handled by CFG
@@ -612,13 +609,14 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     } else if (const FunctionDecl *F = S->getDirectCallee()) {
       transferInlineCall(S, F);
 
-      // If this call produces a prvalue of record type, make sure that we have
-      // a `RecordValue` for it. This is required so that
-      // `Environment::getResultObjectLocation()` is able to return a location
-      // for this `CallExpr`.
+      // If this call produces a prvalue of record type, initialize its fields
+      // with values.
       if (S->getType()->isRecordType() && S->isPRValue())
-        if (Env.getValue(*S) == nullptr)
-          refreshRecordValue(*S, Env);
+        if (Env.getValue(*S) == nullptr) {
+          RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+          Env.initializeFieldsWithValues(Loc);
+          Env.setValue(*S, refreshRecordValue(Loc, Env));
+        }
     }
   }
 
@@ -666,8 +664,10 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // `getLogicOperatorSubExprValue()`.
     if (S->isGLValue())
       Env.setStorageLocation(*S, Env.createObject(S->getType()));
-    else if (Value *Val = Env.createValue(S->getType()))
-      Env.setValue(*S, *Val);
+    else if (!S->getType()->isRecordType()) {
+      if (Value *Val = Env.createValue(S->getType()))
+        Env.setValue(*S, *Val);
+    }
   }
 
   void VisitInitListExpr(const InitListExpr *S) {
@@ -688,71 +688,51 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       return;
     }
 
-    llvm::DenseMap<const ValueDecl *, StorageLocation *> FieldLocs;
-    RecordInitListHelper InitListHelper(S);
+    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+    Env.setValue(*S, refreshRecordValue(Loc, Env));
 
-    for (auto [Base, Init] : InitListHelper.base_inits()) {
-      assert(Base->getType().getCanonicalType() ==
-             Init->getType().getCanonicalType());
-      auto *BaseVal = Env.get<RecordValue>(*Init);
-      if (!BaseVal)
-        BaseVal = cast<RecordValue>(Env.createValue(Init->getType()));
-      // Take ownership of the fields of the `RecordValue` for the base class
-      // and incorporate them into the "flattened" set of fields for the
-      // derived class.
-      auto Children = BaseVal->getLoc().children();
-      FieldLocs.insert(Children.begin(), Children.end());
-    }
+    // Initialization of base classes and fields of record type happens when we
+    // visit the nested `CXXConstructExpr` or `InitListExpr` for that base class
+    // or field. We therefore only need to deal with fields of non-record type
+    // here.
 
-    for (auto [Field, Init] : InitListHelper.field_inits()) {
-      assert(
-          // The types are same, or
-          Field->getType().getCanonicalType().getUnqualifiedType() ==
-              Init->getType().getCanonicalType().getUnqualifiedType() ||
-          // The field's type is T&, and initializer is T
-          (Field->getType()->isReferenceType() &&
-           Field->getType().getCanonicalType()->getPointeeType() ==
-               Init->getType().getCanonicalType()));
-      auto& Loc = Env.createObject(Field->getType(), Init);
-      FieldLocs.insert({Field, &Loc});
-    }
+    RecordInitListHelper InitListHelper(S);
 
-    // In the case of a union, we don't in general have initializers for all
-    // of the fields. Create storage locations for the remaining fields (but
-    // don't associate them with values).
-    if (Type->isUnionType()) {
-      for (const FieldDecl *Field :
-           Env.getDataflowAnalysisContext().getModeledFields(Type)) {
-        if (auto [it, inserted] = FieldLocs.insert({Field, nullptr}); inserted)
-          it->second = &Env.createStorageLocation(Field->getType());
+    for (auto [Field, Init] : InitListHelper.field_inits()) {
+      if (Field->getType()->isRecordType())
+        continue;
+      if (Field->getType()->isReferenceType()) {
+        assert(Field->getType().getCanonicalType()->getPointeeType() ==
+               Init->getType().getCanonicalType());
+        Loc.setChild(*Field, &Env.createObject(Field->getType(), Init));
+        continue;
       }
+      assert(Field->getType().getCanonicalType().getUnqualifiedType() ==
+             Init->getType().getCanonicalType().getUnqualifiedType());
+      StorageLocation *FieldLoc = Loc.getChild(*Field);
+      // Locations for non-reference fields must always be non-null.
+      assert(FieldLoc != nullptr);
+      Value *Val = Env.getValue(*Init);
+      if (Val == nullptr && isa<ImplicitValueInitExpr>(Init) &&
+          Init->getType()->isPointerType())
+        Val =
+            &Env.getOrCreateNullPointerValue(Init->getType()->getPointeeType());
+      if (Val == nullptr)
+        Val = Env.createValue(Field->getType());
+      if (Val != nullptr)
+        Env.setValue(*FieldLoc, *Val);
     }
 
-    // Check that we satisfy the invariant that a `RecordStorageLoation`
-    // contains exactly the set of modeled fields for that type.
-    // `ModeledFields` includes fields from all the bases, but only the
-    // modeled ones. However, if a class type is initialized with an
-    // `InitListExpr`, all fields in the class, including those from base
-    // classes, are included in the set of modeled fields. The code above
-    // should therefore populate exactly the modeled fields.
-    assert(containsSameFields(
-        Env.getDataflowAnalysisContext().getModeledFields(Type), FieldLocs));
-
-    RecordStorageLocation::SyntheticFieldMap SyntheticFieldLocs;
-    for (const auto &Entry :
-         Env.getDataflowAnalysisContext().getSyntheticFields(Type)) {
-      SyntheticFieldLocs.insert(
-          {Entry.getKey(), &Env.createObject(Entry.getValue())});
+    for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
+      QualType FieldType = FieldLoc->getType();
+      if (FieldType->isRecordType()) {
+        Env.initializeFieldsWithValues(*cast<RecordStorageLocation>(FieldLoc));
+      } else {
+        if (Value *Val = Env.createValue(FieldType))
+          Env.setValue(*FieldLoc, *Val);
+      }
     }
 
-    auto &Loc = Env.getDataflowAnalysisContext().createRecordStorageLocation(
-        Type, std::move(FieldLocs), std::move(SyntheticFieldLocs));
-    RecordValue &RecordVal = Env.create<RecordValue>(Loc);
-
-    Env.setValue(Loc, RecordVal);
-
-    Env.setValue(*S, RecordVal);
-
     // FIXME: Implement array initialization.
   }
 
diff --git a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
index 595f70f819ddb..1b73c5d683016 100644
--- a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
+++ b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
@@ -369,17 +369,10 @@ builtinTransferInitializer(const CFGInitializer &Elt,
     ParentLoc->setChild(*Member, InitExprLoc);
   } else if (auto *InitExprVal = Env.getValue(*InitExpr)) {
     assert(MemberLoc != nullptr);
-    if (Member->getType()->isRecordType()) {
-      auto *InitValStruct = cast<RecordValue>(InitExprVal);
-      // FIXME: Rather than performing a copy here, we should really be
-      // initializing the field in place. This would require us to propagate the
-      // storage location of the field to the AST node that creates the
-      // `RecordValue`.
-      copyRecord(InitValStruct->getLoc(),
-                 *cast<RecordStorageLocation>(MemberLoc), Env);
-    } else {
+    // Record-type initializers construct themselves directly into the result
+    // object, so there is no need to handle them here.
+    if (!Member->getType()->isRecordType())
       Env.setValue(*MemberLoc, *InitExprVal);
-    }
   }
 }
 
diff --git a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
index 465a8e21690c4..cc20623f881ff 100644
--- a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
@@ -24,6 +24,7 @@ namespace {
 
 using namespace clang;
 using namespace dataflow;
+using ::clang::dataflow::test::findValueDecl;
 using ::clang::dataflow::test::getFieldValue;
 using ::testing::Contains;
 using ::testing::IsNull;
@@ -199,6 +200,48 @@ TEST_F(EnvironmentTest, JoinRecords) {
   }
 }
 
+TEST_F(EnvironmentTest, DifferentReferenceLocInJoin) {
+  // This tests the case where the storage location for a reference-type
+  // variable is different for two states being joined. We used to believe this
+  // could not happen and therefore had an assertion disallowing this; this test
+  // exists to demonstrate that we can handle this condition without a failing
+  // assertion. See also the discussion here:
+  // https://discourse.llvm.org/t/70086/6
+
+  using namespace ast_matchers;
+
+  std::string Code = R"cc(
+    void f(int &ref) {}
+  )cc";
+
+  auto Unit =
+      tooling::buildASTFromCodeWithArgs(Code, {"-fsyntax-only", "-std=c++11"});
+  auto &Context = Unit->getASTContext();
+
+  ASSERT_EQ(Context.getDiagnostics().getClient()->getNumErrors(), 0U);
+
+  const ValueDecl *Ref = findValueDecl(Context, "ref");
+
+  Environment Env1(DAContext);
+  StorageLocation &Loc1 = Env1.createStorageLocation(Context.IntTy);
+  Env1.setStorageLocation(*Ref, Loc1);
+
+  Environment Env2(DAContext);
+  StorageLocation &Loc2 = Env2.createStorageLocation(Context.IntTy);
+  Env2.setStorageLocation(*Ref, Loc2);
+
+  EXPECT_NE(&Loc1, &Loc2);
+
+  Environment::ValueModel Model;
+  Environment EnvJoined =
+      Environment::join(Env1, Env2, Model, Environment::DiscardExprState);
+
+  // Joining environments with different storage locations for the same
+  // declaration results in the declaration being removed from the joined
+  // environment.
+  EXPECT_EQ(EnvJoined.getStorageLocation(*Ref), nullptr);
+}
+
 TEST_F(EnvironmentTest, InitGlobalVarsFun) {
   using namespace ast_matchers;
 
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index ca055a462a286..00dafb2988c69 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -1582,10 +1582,9 @@ TEST(TransferTest, FieldsDontHaveValuesInConstructorWithBaseClass) {
       [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
          ASTContext &ASTCtx) {
         const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-        // FIXME: The field of the base class should already have been
-        // initialized with a value by the base constructor. This test documents
-        // the current buggy behavior.
-        EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
+        // The field of the base class should already have been initialized with
+        // a value by the base constructor.
+        EXPECT_NE(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
                                 ASTCtx, Env),
                   nullptr);
         EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "Val",
@@ -2998,8 +2997,12 @@ TEST(TransferTest, ResultObjectLocation) {
 
 TEST(TransferTest, ResultObjectLocationForDefaultArgExpr) {
   std::string Code = R"(
-    struct S {};
-    void funcWithDefaultArg(S s = S());
+    struct Inner {};
+    struct Outer {
+        Inner I = {};
+    };
+
+    void funcWithDefaultArg(Outer O = {});
     void target() {
       funcWithDefaultArg();
       // [[p]]
@@ -3058,13 +3061,7 @@ TEST(TransferTest, ResultObjectLocationForDefaultInitExpr) {
 
         RecordStorageLocation &Loc = Env.getResultObjectLocation(*DefaultInit);
 
-        // FIXME: The result object location for the `CXXDefaultInitExpr` should
-        // be the location of the member variable being initialized, but we
-        // don't do this correctly yet; see also comments in
-        // `builtinTransferInitializer()`.
-        // For the time being, we just document the current erroneous behavior
-        // here (this should be `EXPECT_EQ` when the behavior is fixed).
-        EXPECT_NE(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
+        EXPECT_EQ(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
       });
 }
 
@@ -3101,6 +3098,79 @@ TEST(TransferTest, ResultObjectLocationForCXXOperatorCallExpr) {
       });
 }
 
+TEST(TransferTest, ResultObjectLocationForStdInitializerListExpr) {
+  std::string Code = R"(
+    namespace std {
+    template <typename T>
+    struct initializer_list {};
+    } // namespace std
+
+    void target() {
+      std::initializer_list<int> list = {1};
+      // [[p]]
+    }
+  )";
+
+  using ast_matchers::cxxStdInitializerListExpr;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *StdInitList = selectFirst<CXXStdInitializerListExpr>(
+            "std_init_list",
+            match(cxxStdInitializerListExpr().bind("std_init_list"), ASTCtx));
+        ASSERT_NE(StdInitList, nullptr);
+
+        EXPECT_EQ(&Env.getResultObjectLocation(*StdInitList),
+                  &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "list"));
+      });
+}
+
+TEST(TransferTest, ResultObjectLocationPropagatesThroughConditionalOperator) {
+  std::string Code = R"(
+    struct A {
+      A(int);
+    };
+
+    void target(bool b) {
+      A a = b ? A(0) : A(1);
+      (void)0; // [[p]]
+    }
+  )";
+  using ast_matchers::cxxConstructExpr;
+  using ast_matchers::equals;
+  using ast_matchers::hasArgument;
+  using ast_matchers::integerLiteral;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  using ast_matchers::traverse;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *ConstructExpr0 = selectFirst<CXXConstructExpr>(
+            "construct",
+            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(0))))
+                      .bind("construct"),
+                  ASTCtx));
+        auto *ConstructExpr1 = selectFirst<CXXConstructExpr>(
+            "construct",
+            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(1))))
+                      .bind("construct"),
+                  ASTCtx));
+
+        auto &ALoc = getLocForDecl<StorageLocation>(ASTCtx, Env, "a");
+        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr0), &ALoc);
+        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr1), &ALoc);
+      });
+}
+
 TEST(TransferTest, StaticCast) {
   std::string Code = R"(
     void target(int Foo) {
@@ -5886,6 +5956,38 @@ TEST(TransferTest, ContextSensitiveReturnRecord) {
       {BuiltinOptions{ContextSensitiveOptions{}}});
 }
 
+TEST(TransferTest, ContextSensitiveReturnSelfReferentialRecord) {
+  std::string Code = R"(
+    struct S {
+      S() { self = this; }
+      S *self;
+    };
+
+    S makeS() {
+      // RVO guarantees that this will be constructed directly into `MyS`.
+      return S();
+    }
+
+    void target() {
+      S MyS = makeS();
+      // [[p]]
+    }
+  )";
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto &MySLoc = getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "MyS");
+
+        auto *SelfVal =
+            cast<PointerValue>(getFieldValue(&MySLoc, "self", ASTCtx, Env));
+        EXPECT_EQ(&SelfVal->getPointeeLoc(), &MySLoc);
+      },
+      {BuiltinOptions{ContextSensitiveOptions{}}});
+}
+
 TEST(TransferTest, ContextSensitiveMethodLiteral) {
   std::string Code = R"(
     class MyClass {
@@ -6830,50 +6932,6 @@ TEST(TransferTest, LambdaCaptureThis) {
       });
 }
 
-TEST(TransferTest, DifferentReferenceLocInJoin) {
-  // This test triggers a case where the storage location for a reference-type
-  // variable is different for two states being joined. We used to believe this
-  // could not happen and therefore had an assertion disallowing this; this test
-  // exists to demonstrate that we can handle this condition without a failing
-  // assertion. See also the discussion here:
-  // https://discourse.llvm.org/t/70086/6
-  std::string Code = R"(
-    namespace std {
-      template <class T> struct initializer_list {
-        const T* begin();
-        const T* end();
-      };
-    }
-
-    void target(char* p, char* end) {
-      while (p != end) {
-        if (*p == ' ') {
-          p++;
-          continue;
-        }
-
-        auto && range = {1, 2};
-        for (auto b = range.begin(), e = range.end(); b != e; ++b) {
-        }
-        (void)0;
-        // [[p]]
-      }
-    }
-  )";
-  runDataflow(
-      Code,
-      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
-         ASTContext &ASTCtx) {
-        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-
-        // Joining environments with different storage locations for the same
-        // declaration results in the declaration being removed from the joined
-        // environment.
-        const ValueDecl *VD = findValueDecl(ASTCtx, "range");
-        ASSERT_EQ(Env.getStorageLocation(*VD), nullptr);
-      });
-}
-
 // This test verifies correct modeling of a relational dependency that goes
 // through unmodeled functions (the simple `cond()` in this case).
 TEST(TransferTest, ConditionalRelation) {

From b9a3551c905573df456ee52fa1051e49fa956c65 Mon Sep 17 00:00:00 2001
From: "Kevin P. Neal" <kevin.neal@sas.com>
Date: Wed, 10 Apr 2024 11:28:43 -0400
Subject: [PATCH 047/886] [FPEnv][BitcodeReader] Correct strictfp test.

Correct a strictfp test to follow the rules documented in the LangRef:
https://llvm.org/docs/LangRef.html#constrained-floating-point-intrinsics

This test needed the strictfp attribute added to a function definition.

Test changes verified with D146845.
---
 llvm/unittests/Bitcode/BitReaderTest.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/unittests/Bitcode/BitReaderTest.cpp b/llvm/unittests/Bitcode/BitReaderTest.cpp
index 3e449f9057789..22cc5e7492803 100644
--- a/llvm/unittests/Bitcode/BitReaderTest.cpp
+++ b/llvm/unittests/Bitcode/BitReaderTest.cpp
@@ -160,7 +160,7 @@ TEST(BitReaderTest, MaterializeConstrainedFPStrictFP) {
   LLVMContext Context;
   std::unique_ptr<Module> M = getLazyModuleFromAssembly(
       Context, Mem,
-      "define double @foo(double %a) {\n"
+      "define double @foo(double %a) strictfp {\n"
       "  %result = call double @llvm.experimental.constrained.sqrt.f64(double "
       "%a, metadata !\"round.tonearest\", metadata !\"fpexcept.strict\") "
       "strictfp\n"

From c1d3f39ae98535777c957aab3611d2abc97b2815 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 21:36:58 -0500
Subject: [PATCH 048/886] [ValueTracking] Add tests for `shufflevector` in
 `isKnownNonZero`

---
 .../Transforms/InstSimplify/known-non-zero.ll | 132 ++++++++++++++++++
 1 file changed, 132 insertions(+)

diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index 51f80f62c2f34..9486708369b77 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -177,3 +177,135 @@ define i1 @smax_non_zero(i8 %xx, i8 %y) {
   %r = icmp eq i8 %v, 0
   ret i1 %r
 }
+
+define <4 x i1> @shuf_nonzero_both(<4 x i8> %xx, <4 x i8> %yy) {
+; CHECK-LABEL: @shuf_nonzero_both(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <4 x i8> [[YY:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> [[Y]], <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %y = add nuw <4 x i8> %yy, <i8 1, i8 1, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> %y, <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_both_fail(<4 x i8> %xx, <4 x i8> %yy) {
+; CHECK-LABEL: @shuf_nonzero_both_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw <4 x i8> [[YY:%.*]], <i8 1, i8 1, i8 1, i8 0>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> [[Y]], <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %y = add nuw <4 x i8> %yy, <i8 1, i8 1, i8 1, i8 0>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> %y, <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_both_fail2(<4 x i8> %xx, <4 x i8> %yy) {
+; CHECK-LABEL: @shuf_nonzero_both_fail2(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add <4 x i8> [[YY:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> [[Y]], <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+  %y = add <4 x i8> %yy, <i8 1, i8 1, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> %y, <4 x i32> <i32 0, i32 4, i32 7, i32 2>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_lhs(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_lhs(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> poison, <4 x i32> <i32 2, i32 2, i32 0, i32 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 2, i32 0, i32 1>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_lhs2(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_lhs2(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 0>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 0>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 1>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_lhs2_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_lhs2_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 0>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> poison, <4 x i32> <i32 2, i32 0, i32 3, i32 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 0>
+
+  %shuf = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 0, i32 3, i32 1>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_rhs(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_rhs(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> poison, <4 x i8> [[X]], <4 x i32> <i32 6, i32 7, i32 5, i32 4>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> poison, <4 x i8> %x, <4 x i32> <i32 6, i32 7, i32 5, i32 4>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_rhs2(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_rhs2(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 0, i8 0, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> poison, <4 x i8> [[X]], <4 x i32> <i32 6, i32 7, i32 7, i32 6>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 0, i8 0, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> poison, <4 x i8> %x, <4 x i32> <i32 6, i32 7, i32 7, i32 6>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}
+
+define <4 x i1> @shuf_nonzero_rhs2_fail(<4 x i8> %xx) {
+; CHECK-LABEL: @shuf_nonzero_rhs2_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 0, i8 0, i8 1, i8 1>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> poison, <4 x i8> [[X]], <4 x i32> <i32 6, i32 7, i32 5, i32 6>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %x = add nuw <4 x i8> %xx, <i8 0, i8 0, i8 1, i8 1>
+
+  %shuf = shufflevector <4 x i8> poison, <4 x i8> %x, <4 x i32> <i32 6, i32 7, i32 5, i32 6>
+  %r = icmp eq <4 x i8> %shuf, zeroinitializer
+  ret <4 x i1> %r
+}

From 87528bfefbb50ed6560b9b8482fc7c9f86ca34cd Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 15:33:55 -0500
Subject: [PATCH 049/886] [ValueTracking] Add support for `shufflevector` in
 `isKnownNonZero`

Shuffles don't modify the data, so if all elements that end up in the
destination are non-zero the result is non-zero.

Closes #87702
---
 llvm/lib/Analysis/ValueTracking.cpp           | 15 +++++++++++
 .../Transforms/InstSimplify/known-non-zero.ll | 26 ++++---------------
 2 files changed, 20 insertions(+), 21 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 5ef1969893b42..b3029f440ca2a 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2777,6 +2777,21 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       }
     }
     break;
+  case Instruction::ShuffleVector: {
+    auto *Shuf = dyn_cast<ShuffleVectorInst>(I);
+    if (!Shuf)
+      break;
+    APInt DemandedLHS, DemandedRHS;
+    // For undef elements, we don't know anything about the common state of
+    // the shuffle result.
+    if (!getShuffleDemandedElts(Shuf, DemandedElts, DemandedLHS, DemandedRHS))
+      break;
+    // If demanded elements for both vecs are non-zero, the shuffle is non-zero.
+    return (DemandedRHS.isZero() ||
+            isKnownNonZero(Shuf->getOperand(1), DemandedRHS, Depth, Q)) &&
+           (DemandedLHS.isZero() ||
+            isKnownNonZero(Shuf->getOperand(0), DemandedLHS, Depth, Q));
+  }
   case Instruction::Freeze:
     return isKnownNonZero(I->getOperand(0), Depth, Q) &&
            isGuaranteedNotToBePoison(I->getOperand(0), Q.AC, Q.CxtI, Q.DT,
diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index 9486708369b77..c443dc68d0c13 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -180,11 +180,7 @@ define i1 @smax_non_zero(i8 %xx, i8 %y) {
 
 define <4 x i1> @shuf_nonzero_both(<4 x i8> %xx, <4 x i8> %yy) {
 ; CHECK-LABEL: @shuf_nonzero_both(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[Y:%.*]] = add nuw <4 x i8> [[YY:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> [[Y]], <4 x i32> <i32 0, i32 4, i32 7, i32 2>
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
-; CHECK-NEXT:    ret <4 x i1> [[R]]
+; CHECK-NEXT:    ret <4 x i1> zeroinitializer
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
   %y = add nuw <4 x i8> %yy, <i8 1, i8 1, i8 1, i8 1>
@@ -228,10 +224,7 @@ define <4 x i1> @shuf_nonzero_both_fail2(<4 x i8> %xx, <4 x i8> %yy) {
 
 define <4 x i1> @shuf_nonzero_lhs(<4 x i8> %xx) {
 ; CHECK-LABEL: @shuf_nonzero_lhs(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> poison, <4 x i32> <i32 2, i32 2, i32 0, i32 1>
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
-; CHECK-NEXT:    ret <4 x i1> [[R]]
+; CHECK-NEXT:    ret <4 x i1> zeroinitializer
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
 
@@ -242,10 +235,7 @@ define <4 x i1> @shuf_nonzero_lhs(<4 x i8> %xx) {
 
 define <4 x i1> @shuf_nonzero_lhs2(<4 x i8> %xx) {
 ; CHECK-LABEL: @shuf_nonzero_lhs2(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 0>
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> [[X]], <4 x i8> poison, <4 x i32> <i32 2, i32 0, i32 1, i32 1>
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
-; CHECK-NEXT:    ret <4 x i1> [[R]]
+; CHECK-NEXT:    ret <4 x i1> zeroinitializer
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 0>
 
@@ -270,10 +260,7 @@ define <4 x i1> @shuf_nonzero_lhs2_fail(<4 x i8> %xx) {
 
 define <4 x i1> @shuf_nonzero_rhs(<4 x i8> %xx) {
 ; CHECK-LABEL: @shuf_nonzero_rhs(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 1, i8 1, i8 1, i8 1>
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> poison, <4 x i8> [[X]], <4 x i32> <i32 6, i32 7, i32 5, i32 4>
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
-; CHECK-NEXT:    ret <4 x i1> [[R]]
+; CHECK-NEXT:    ret <4 x i1> zeroinitializer
 ;
   %x = add nuw <4 x i8> %xx, <i8 1, i8 1, i8 1, i8 1>
 
@@ -284,10 +271,7 @@ define <4 x i1> @shuf_nonzero_rhs(<4 x i8> %xx) {
 
 define <4 x i1> @shuf_nonzero_rhs2(<4 x i8> %xx) {
 ; CHECK-LABEL: @shuf_nonzero_rhs2(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <4 x i8> [[XX:%.*]], <i8 0, i8 0, i8 1, i8 1>
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i8> poison, <4 x i8> [[X]], <4 x i32> <i32 6, i32 7, i32 7, i32 6>
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <4 x i8> [[SHUF]], zeroinitializer
-; CHECK-NEXT:    ret <4 x i1> [[R]]
+; CHECK-NEXT:    ret <4 x i1> zeroinitializer
 ;
   %x = add nuw <4 x i8> %xx, <i8 0, i8 0, i8 1, i8 1>
 

From 8a28b9b8ec1686426a4b43c8431570eaa1da77d9 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 21:37:14 -0500
Subject: [PATCH 050/886] [ValueTracking] Add tests for `insertelement` in
 `isKnownNonZero`; NFC

---
 .../Transforms/InstSimplify/known-non-zero.ll | 96 +++++++++++++++++++
 1 file changed, 96 insertions(+)

diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index c443dc68d0c13..1417e86ee678c 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -293,3 +293,99 @@ define <4 x i1> @shuf_nonzero_rhs2_fail(<4 x i8> %xx) {
   %r = icmp eq <4 x i8> %shuf, zeroinitializer
   ret <4 x i1> %r
 }
+
+define <2 x i1> @insert_nonzero0(<2 x i8> %xx, i8 %yy) {
+; CHECK-LABEL: @insert_nonzero0(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 1
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 1, i8 0>
+  %y = add nuw i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 1
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @insert_nonzero1(<2 x i8> %xx, i8 %yy) {
+; CHECK-LABEL: @insert_nonzero1(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 0, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 0
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 0, i8 1>
+  %y = add nuw i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 0
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @insert_nonzero_fail(<2 x i8> %xx, i8 %yy) {
+; CHECK-LABEL: @insert_nonzero_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 0
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 1, i8 0>
+  %y = add nuw i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 0
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @insert_nonzero_fail2(<2 x i8> %xx, i8 %yy) {
+; CHECK-LABEL: @insert_nonzero_fail2(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 0, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 0
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 0, i8 1>
+  %y = add i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 0
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @insert_nonzero_any_idx(<2 x i8> %xx, i8 %yy, i32 %idx) {
+; CHECK-LABEL: @insert_nonzero_any_idx(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 1>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 [[IDX:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 1, i8 1>
+  %y = add nuw i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 %idx
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @insert_nonzero_any_idx_fail(<2 x i8> %xx, i8 %yy, i32 %idx) {
+; CHECK-LABEL: @insert_nonzero_any_idx_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 0>
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 [[IDX:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 1, i8 0>
+  %y = add nuw i8 %yy, 1
+
+  %ins = insertelement <2 x i8> %x, i8 %y, i32 %idx
+  %r = icmp eq <2 x i8> %ins, zeroinitializer
+  ret <2 x i1> %r
+}

From 9c545a14c09051b011358854655c1f466d656e79 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 15:34:28 -0500
Subject: [PATCH 051/886] [ValueTracking] Add support for `insertelement` in
 `isKnownNonZero`

Inserts don't modify the data, so if all elements that end up in the
destination are non-zero the result is non-zero.

Closes #87703
---
 llvm/lib/Analysis/ValueTracking.cpp           | 23 +++++++++++++++++++
 .../Transforms/InstSimplify/known-non-zero.ll | 18 +++------------
 2 files changed, 26 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index b3029f440ca2a..9f16eaf9e0990 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2763,6 +2763,29 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       return isKnownNonZero(U.get(), DemandedElts, NewDepth, RecQ);
     });
   }
+  case Instruction::InsertElement: {
+    if (isa<ScalableVectorType>(I->getType()))
+      break;
+
+    const Value *Vec = I->getOperand(0);
+    const Value *Elt = I->getOperand(1);
+    auto *CIdx = dyn_cast<ConstantInt>(I->getOperand(2));
+
+    unsigned NumElts = DemandedElts.getBitWidth();
+    APInt DemandedVecElts = DemandedElts;
+    bool SkipElt = false;
+    // If we know the index we are inserting too, clear it from Vec check.
+    if (CIdx && CIdx->getValue().ult(NumElts)) {
+      DemandedVecElts.clearBit(CIdx->getZExtValue());
+      SkipElt = !DemandedElts[CIdx->getZExtValue()];
+    }
+
+    // Result is zero if Elt is non-zero and rest of the demanded elts in Vec
+    // are non-zero.
+    return (SkipElt || isKnownNonZero(Elt, Depth, Q)) &&
+           (DemandedVecElts.isZero() ||
+            isKnownNonZero(Vec, DemandedVecElts, Depth, Q));
+  }
   case Instruction::ExtractElement:
     if (const auto *EEI = dyn_cast<ExtractElementInst>(I)) {
       const Value *Vec = EEI->getVectorOperand();
diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index 1417e86ee678c..d9b8f5eed3239 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -296,11 +296,7 @@ define <4 x i1> @shuf_nonzero_rhs2_fail(<4 x i8> %xx) {
 
 define <2 x i1> @insert_nonzero0(<2 x i8> %xx, i8 %yy) {
 ; CHECK-LABEL: @insert_nonzero0(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 0>
-; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 1
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
-; CHECK-NEXT:    ret <2 x i1> [[R]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %x = add nuw <2 x i8> %xx, <i8 1, i8 0>
   %y = add nuw i8 %yy, 1
@@ -312,11 +308,7 @@ define <2 x i1> @insert_nonzero0(<2 x i8> %xx, i8 %yy) {
 
 define <2 x i1> @insert_nonzero1(<2 x i8> %xx, i8 %yy) {
 ; CHECK-LABEL: @insert_nonzero1(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 0, i8 1>
-; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 0
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
-; CHECK-NEXT:    ret <2 x i1> [[R]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %x = add nuw <2 x i8> %xx, <i8 0, i8 1>
   %y = add nuw i8 %yy, 1
@@ -360,11 +352,7 @@ define <2 x i1> @insert_nonzero_fail2(<2 x i8> %xx, i8 %yy) {
 
 define <2 x i1> @insert_nonzero_any_idx(<2 x i8> %xx, i8 %yy, i32 %idx) {
 ; CHECK-LABEL: @insert_nonzero_any_idx(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 1>
-; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
-; CHECK-NEXT:    [[INS:%.*]] = insertelement <2 x i8> [[X]], i8 [[Y]], i32 [[IDX:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[INS]], zeroinitializer
-; CHECK-NEXT:    ret <2 x i1> [[R]]
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %x = add nuw <2 x i8> %xx, <i8 1, i8 1>
   %y = add nuw i8 %yy, 1

From 195d278d502308655edb1e9ff1c6f0c9256d0d15 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 21:41:47 -0500
Subject: [PATCH 052/886] [ValueTracking] Add tests for `xor`/`disjoint or` in
 `getInvertibleOperands`; NFC

---
 llvm/test/Transforms/InstSimplify/icmp.ll | 94 ++++++++++++++++++++++-
 1 file changed, 93 insertions(+), 1 deletion(-)

diff --git a/llvm/test/Transforms/InstSimplify/icmp.ll b/llvm/test/Transforms/InstSimplify/icmp.ll
index 3109768bdfe00..a66f7cb879ef9 100644
--- a/llvm/test/Transforms/InstSimplify/icmp.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp.ll
@@ -270,7 +270,7 @@ define i1 @load_ptr(ptr %p) {
 
 define i1 @load_ptr_null_valid(ptr %p) null_pointer_is_valid {
 ; CHECK-LABEL: @load_ptr_null_valid(
-; CHECK-NEXT:    [[LOAD_P:%.*]] = load ptr, ptr [[P:%.*]], align 8, !dereferenceable !0
+; CHECK-NEXT:    [[LOAD_P:%.*]] = load ptr, ptr [[P:%.*]], align 8, !dereferenceable [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne ptr [[LOAD_P]], null
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -278,3 +278,95 @@ define i1 @load_ptr_null_valid(ptr %p) null_pointer_is_valid {
   %r = icmp ne ptr %load_p, null
   ret i1 %r
 }
+
+define i1 @non_eq_disjoint_or_common_op(i8 %x, i8 %y, i8 %ww, i8 %a) {
+; CHECK-LABEL: @non_eq_disjoint_or_common_op(
+; CHECK-NEXT:    [[W:%.*]] = add nuw i8 [[WW:%.*]], 1
+; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
+; CHECK-NEXT:    [[XY:%.*]] = or disjoint i8 [[X:%.*]], [[Y]]
+; CHECK-NEXT:    [[XZ:%.*]] = or disjoint i8 [[X]], [[Z]]
+; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
+; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %w = add nuw i8 %ww, 1
+  %z = add i8 %y, %w
+
+  %xy = or disjoint i8 %x, %y
+  %xz = or disjoint i8 %x, %z
+
+  %axy = add i8 %a, %xy
+  %axz = add i8 %a, %xz
+  %r = icmp eq i8 %axy, %axz
+  ret i1 %r
+}
+
+define i1 @non_eq_disjoint_or_common_op_fail(i8 %x, i8 %y, i8 %ww, i8 %a) {
+; CHECK-LABEL: @non_eq_disjoint_or_common_op_fail(
+; CHECK-NEXT:    [[W:%.*]] = add nuw i8 [[WW:%.*]], 1
+; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
+; CHECK-NEXT:    [[XY:%.*]] = or i8 [[X:%.*]], [[Y]]
+; CHECK-NEXT:    [[XZ:%.*]] = or disjoint i8 [[X]], [[Z]]
+; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
+; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %w = add nuw i8 %ww, 1
+  %z = add i8 %y, %w
+
+  %xy = or i8 %x, %y
+  %xz = or disjoint i8 %x, %z
+
+  %axy = add i8 %a, %xy
+  %axz = add i8 %a, %xz
+  %r = icmp eq i8 %axy, %axz
+  ret i1 %r
+}
+
+define i1 @non_eq_xor_common_op(i8 %x, i8 %y, i8 %ww, i8 %a) {
+; CHECK-LABEL: @non_eq_xor_common_op(
+; CHECK-NEXT:    [[W:%.*]] = add nuw i8 [[WW:%.*]], 1
+; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
+; CHECK-NEXT:    [[XY:%.*]] = xor i8 [[Y]], [[X:%.*]]
+; CHECK-NEXT:    [[XZ:%.*]] = xor i8 [[X]], [[Z]]
+; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
+; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %w = add nuw i8 %ww, 1
+  %z = add i8 %y, %w
+
+  %xy = xor i8 %y, %x
+  %xz = xor i8 %x, %z
+
+  %axy = add i8 %a, %xy
+  %axz = add i8 %a, %xz
+  %r = icmp eq i8 %axy, %axz
+  ret i1 %r
+}
+
+define i1 @non_eq_xor_common_op_fail(i8 %x, i8 %y, i8 %ww, i8 %a) {
+; CHECK-LABEL: @non_eq_xor_common_op_fail(
+; CHECK-NEXT:    [[W:%.*]] = add nsw i8 [[WW:%.*]], 1
+; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
+; CHECK-NEXT:    [[XY:%.*]] = xor i8 [[Y]], [[X:%.*]]
+; CHECK-NEXT:    [[XZ:%.*]] = xor i8 [[X]], [[Z]]
+; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
+; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %w = add nsw i8 %ww, 1
+  %z = add i8 %y, %w
+
+  %xy = xor i8 %y, %x
+  %xz = xor i8 %x, %z
+
+  %axy = add i8 %a, %xy
+  %axz = add i8 %a, %xz
+  %r = icmp eq i8 %axy, %axz
+  ret i1 %r
+}

From 0c57a2e4b4e5a6e5dda78a313fc8d8e3c91797f5 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 17:40:03 -0500
Subject: [PATCH 053/886] [ValueTracking] Add support for `xor`/`disjoint or`
 in `getInvertibleOperands`

This strengthens our `isKnownNonEqual` logic with some fairly
trivial cases.

Proofs: https://alive2.llvm.org/ce/z/4pxRTj

Closes #87705
---
 llvm/lib/Analysis/ValueTracking.cpp       | 15 ++++++++++++++-
 llvm/test/Transforms/InstSimplify/icmp.ll | 18 ++----------------
 2 files changed, 16 insertions(+), 17 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 9f16eaf9e0990..f3ea73b2f0ec4 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -3106,7 +3106,20 @@ getInvertibleOperands(const Operator *Op1,
   switch (Op1->getOpcode()) {
   default:
     break;
-  case Instruction::Add:
+  case Instruction::Or:
+    if (!cast<PossiblyDisjointInst>(Op1)->isDisjoint() ||
+        !cast<PossiblyDisjointInst>(Op2)->isDisjoint())
+      break;
+    [[fallthrough]];
+  case Instruction::Xor:
+  case Instruction::Add: {
+    Value *Other;
+    if (match(Op2, m_c_BinOp(m_Specific(Op1->getOperand(0)), m_Value(Other))))
+      return std::make_pair(Op1->getOperand(1), Other);
+    if (match(Op2, m_c_BinOp(m_Specific(Op1->getOperand(1)), m_Value(Other))))
+      return std::make_pair(Op1->getOperand(0), Other);
+    break;
+  }
   case Instruction::Sub:
     if (Op1->getOperand(0) == Op2->getOperand(0))
       return getOperands(1);
diff --git a/llvm/test/Transforms/InstSimplify/icmp.ll b/llvm/test/Transforms/InstSimplify/icmp.ll
index a66f7cb879ef9..d179909811549 100644
--- a/llvm/test/Transforms/InstSimplify/icmp.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp.ll
@@ -281,14 +281,7 @@ define i1 @load_ptr_null_valid(ptr %p) null_pointer_is_valid {
 
 define i1 @non_eq_disjoint_or_common_op(i8 %x, i8 %y, i8 %ww, i8 %a) {
 ; CHECK-LABEL: @non_eq_disjoint_or_common_op(
-; CHECK-NEXT:    [[W:%.*]] = add nuw i8 [[WW:%.*]], 1
-; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
-; CHECK-NEXT:    [[XY:%.*]] = or disjoint i8 [[X:%.*]], [[Y]]
-; CHECK-NEXT:    [[XZ:%.*]] = or disjoint i8 [[X]], [[Z]]
-; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
-; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %w = add nuw i8 %ww, 1
   %z = add i8 %y, %w
@@ -327,14 +320,7 @@ define i1 @non_eq_disjoint_or_common_op_fail(i8 %x, i8 %y, i8 %ww, i8 %a) {
 
 define i1 @non_eq_xor_common_op(i8 %x, i8 %y, i8 %ww, i8 %a) {
 ; CHECK-LABEL: @non_eq_xor_common_op(
-; CHECK-NEXT:    [[W:%.*]] = add nuw i8 [[WW:%.*]], 1
-; CHECK-NEXT:    [[Z:%.*]] = add i8 [[Y:%.*]], [[W]]
-; CHECK-NEXT:    [[XY:%.*]] = xor i8 [[Y]], [[X:%.*]]
-; CHECK-NEXT:    [[XZ:%.*]] = xor i8 [[X]], [[Z]]
-; CHECK-NEXT:    [[AXY:%.*]] = add i8 [[A:%.*]], [[XY]]
-; CHECK-NEXT:    [[AXZ:%.*]] = add i8 [[A]], [[XZ]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[AXY]], [[AXZ]]
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %w = add nuw i8 %ww, 1
   %z = add i8 %y, %w

From 2646790155f73d6cfb28ec0ee472056740e4658e Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 21:41:53 -0500
Subject: [PATCH 054/886] [ValueTracking] Add tests for `xor`/`disjoint or` in
 `isKnownNonZero`; NFC

---
 llvm/test/Transforms/InstSimplify/icmp.ll | 68 +++++++++++++++++++++++
 1 file changed, 68 insertions(+)

diff --git a/llvm/test/Transforms/InstSimplify/icmp.ll b/llvm/test/Transforms/InstSimplify/icmp.ll
index d179909811549..530dc16144eba 100644
--- a/llvm/test/Transforms/InstSimplify/icmp.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp.ll
@@ -356,3 +356,71 @@ define i1 @non_eq_xor_common_op_fail(i8 %x, i8 %y, i8 %ww, i8 %a) {
   %r = icmp eq i8 %axy, %axz
   ret i1 %r
 }
+
+define i1 @non_eq_disjoint_or(i8 %x, i8 %yy, i8 %w) {
+; CHECK-LABEL: @non_eq_disjoint_or(
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
+; CHECK-NEXT:    [[VAL:%.*]] = or disjoint i8 [[Y]], [[W]]
+; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %y = add nuw i8 %yy, 1
+  %lhs = add i8 %x, %w
+  %val = or disjoint i8 %y, %w
+  %rhs = add i8 %x, %val
+  %r = icmp eq i8 %lhs, %rhs
+  ret i1 %r
+}
+
+define i1 @non_eq_or_fail(i8 %x, i8 %yy, i8 %w) {
+; CHECK-LABEL: @non_eq_or_fail(
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
+; CHECK-NEXT:    [[VAL:%.*]] = or i8 [[Y]], [[W]]
+; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %y = add nuw i8 %yy, 1
+  %lhs = add i8 %x, %w
+  %val = or i8 %y, %w
+  %rhs = add i8 %x, %val
+  %r = icmp eq i8 %lhs, %rhs
+  ret i1 %r
+}
+
+define i1 @non_eq_xor(i8 %x, i8 %yy, i8 %w) {
+; CHECK-LABEL: @non_eq_xor(
+; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
+; CHECK-NEXT:    [[VAL:%.*]] = xor i8 [[Y]], [[W]]
+; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %y = add nuw i8 %yy, 1
+  %lhs = add i8 %x, %w
+  %val = xor i8 %y, %w
+  %rhs = add i8 %x, %val
+  %r = icmp eq i8 %lhs, %rhs
+  ret i1 %r
+}
+
+define i1 @non_eq_xor_fail(i8 %x, i8 %yy, i8 %w) {
+; CHECK-LABEL: @non_eq_xor_fail(
+; CHECK-NEXT:    [[Y:%.*]] = add nsw i8 [[YY:%.*]], 1
+; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
+; CHECK-NEXT:    [[VAL:%.*]] = xor i8 [[Y]], [[W]]
+; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %y = add nsw i8 %yy, 1
+  %lhs = add i8 %x, %w
+  %val = xor i8 %y, %w
+  %rhs = add i8 %x, %val
+  %r = icmp eq i8 %lhs, %rhs
+  ret i1 %r
+}

From 81cdd35c0c8db22bfdd1f06cb2118d17fd99fc07 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 3 Apr 2024 17:41:45 -0500
Subject: [PATCH 055/886] [ValueTracking] Add support for `xor`/`disjoint or`
 in `isKnownNonZero`

Handles cases like `X ^ Y == X` / `X disjoint| Y == X`.

Both of these cases have identical logic to the existing `add` case,
so just converting the `add` code to a more general helper.

Proofs: https://alive2.llvm.org/ce/z/Htm7pe

Closes #87706
---
 llvm/lib/Analysis/ValueTracking.cpp       | 40 +++++++++++++++--------
 llvm/test/Transforms/InstSimplify/icmp.ll | 14 ++------
 2 files changed, 29 insertions(+), 25 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index f3ea73b2f0ec4..3a10de72a2756 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -3207,20 +3207,33 @@ getInvertibleOperands(const Operator *Op1,
   return std::nullopt;
 }
 
-/// Return true if V2 == V1 + X, where X is known non-zero.
-static bool isAddOfNonZero(const Value *V1, const Value *V2, unsigned Depth,
-                           const SimplifyQuery &Q) {
+/// Return true if V1 == (binop V2, X), where X is known non-zero.
+/// Only handle a small subset of binops where (binop V2, X) with non-zero X
+/// implies V2 != V1.
+static bool isModifyingBinopOfNonZero(const Value *V1, const Value *V2,
+                                      unsigned Depth, const SimplifyQuery &Q) {
   const BinaryOperator *BO = dyn_cast<BinaryOperator>(V1);
-  if (!BO || BO->getOpcode() != Instruction::Add)
+  if (!BO)
     return false;
-  Value *Op = nullptr;
-  if (V2 == BO->getOperand(0))
-    Op = BO->getOperand(1);
-  else if (V2 == BO->getOperand(1))
-    Op = BO->getOperand(0);
-  else
-    return false;
-  return isKnownNonZero(Op, Depth + 1, Q);
+  switch (BO->getOpcode()) {
+  default:
+    break;
+  case Instruction::Or:
+    if (!cast<PossiblyDisjointInst>(V1)->isDisjoint())
+      break;
+    [[fallthrough]];
+  case Instruction::Xor:
+  case Instruction::Add:
+    Value *Op = nullptr;
+    if (V2 == BO->getOperand(0))
+      Op = BO->getOperand(1);
+    else if (V2 == BO->getOperand(1))
+      Op = BO->getOperand(0);
+    else
+      return false;
+    return isKnownNonZero(Op, Depth + 1, Q);
+  }
+  return false;
 }
 
 /// Return true if V2 == V1 * C, where V1 is known non-zero, C is not 0/1 and
@@ -3380,7 +3393,8 @@ static bool isKnownNonEqual(const Value *V1, const Value *V2, unsigned Depth,
     };
   }
 
-  if (isAddOfNonZero(V1, V2, Depth, Q) || isAddOfNonZero(V2, V1, Depth, Q))
+  if (isModifyingBinopOfNonZero(V1, V2, Depth, Q) ||
+      isModifyingBinopOfNonZero(V2, V1, Depth, Q))
     return true;
 
   if (isNonEqualMul(V1, V2, Depth, Q) || isNonEqualMul(V2, V1, Depth, Q))
diff --git a/llvm/test/Transforms/InstSimplify/icmp.ll b/llvm/test/Transforms/InstSimplify/icmp.ll
index 530dc16144eba..c94922197096f 100644
--- a/llvm/test/Transforms/InstSimplify/icmp.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp.ll
@@ -359,12 +359,7 @@ define i1 @non_eq_xor_common_op_fail(i8 %x, i8 %y, i8 %ww, i8 %a) {
 
 define i1 @non_eq_disjoint_or(i8 %x, i8 %yy, i8 %w) {
 ; CHECK-LABEL: @non_eq_disjoint_or(
-; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
-; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
-; CHECK-NEXT:    [[VAL:%.*]] = or disjoint i8 [[Y]], [[W]]
-; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %y = add nuw i8 %yy, 1
   %lhs = add i8 %x, %w
@@ -393,12 +388,7 @@ define i1 @non_eq_or_fail(i8 %x, i8 %yy, i8 %w) {
 
 define i1 @non_eq_xor(i8 %x, i8 %yy, i8 %w) {
 ; CHECK-LABEL: @non_eq_xor(
-; CHECK-NEXT:    [[Y:%.*]] = add nuw i8 [[YY:%.*]], 1
-; CHECK-NEXT:    [[LHS:%.*]] = add i8 [[X:%.*]], [[W:%.*]]
-; CHECK-NEXT:    [[VAL:%.*]] = xor i8 [[Y]], [[W]]
-; CHECK-NEXT:    [[RHS:%.*]] = add i8 [[X]], [[VAL]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[LHS]], [[RHS]]
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %y = add nuw i8 %yy, 1
   %lhs = add i8 %x, %w

From 2b00a73f62605fcaeaedd358ba8b55fad06571aa Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Wed, 10 Apr 2024 14:33:56 -0400
Subject: [PATCH 056/886] [SLP]Buildvector for alternate instructions with
 non-profitable gather operands.

If the operands of the potentially alternate node are going to produce
buildvector sequences, which result in more instructions, than the
original code, then suhinstructions should be vectorized as alternate
node, better to end up with the buildvector node.

Left column - experimental, Right - reference.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/loop_unroll.test   413680.00   416272.00  0.6%
                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12351788.00 12354844.00  0.0%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   664901.00   664949.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   664901.00   664949.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/511.povray_r/511.povray_r.test  1171371.00  1171355.00 -0.0%
                                                                                         test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test  1036396.00  1036284.00 -0.0%
                                                                         test-suite :: MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg.test   111280.00   111248.00 -0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1392113.00  1391361.00 -0.1%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1392113.00  1391361.00 -0.1%
                                                                        test-suite :: MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc.test   281676.00   281452.00 -0.1%
                                                                                    test-suite :: MultiSource/Benchmarks/VersaBench/ecbdes/ecbdes.test     3025.00     3019.00 -0.2%
                                                                                test-suite :: MultiSource/Benchmarks/Prolangs-C/plot2fig/plot2fig.test     6351.00     6335.00 -0.3%

Metric: SLP.NumVectorInstructions

Program                                                                                                                                                SLP.NumVectorInstructions
                                                                                                                                                       results                   results0 diff
                                                                                    test-suite :: MultiSource/Benchmarks/VersaBench/ecbdes/ecbdes.test    15.00                     16.00   6.7%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test  1703.00                   1707.00   0.2%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test  1703.00                   1707.00   0.2%
                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 26241.00                  26239.00  -0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 11761.00                  11754.00  -0.1%
                                                                        test-suite :: MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc.test   824.00                    822.00  -0.2%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  5668.00                   5654.00  -0.2%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  5668.00                   5654.00  -0.2%
                                                                                     test-suite :: External/SPEC/CINT2017rate/502.gcc_r/502.gcc_r.test   792.00                    790.00  -0.3%
                                                                                    test-suite :: External/SPEC/CINT2017speed/602.gcc_s/602.gcc_s.test   792.00                    790.00  -0.3%
                                                                                       test-suite :: MultiSource/Benchmarks/FreeBench/pifft/pifft.test  1389.00                   1384.00  -0.4%
                                                                                         test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test   596.00                    590.00  -1.0%
                                                                                test-suite :: MultiSource/Benchmarks/Prolangs-C/plot2fig/plot2fig.test     6.00                      5.00 -16.7%

Metric: exec_time

Program                                                                                                                                                exec_time
                                                                                                                                                       results   results0  diff
                                                                               test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test     99.14    100.00    0.9%

Other changes are not significant (less than 0.1% percent with exectime
less 5 secs).

SingleSource/Benchmarks/Adobe-C++/loop_unroll - same small patterns
remain scalar, smaller code.
External/SPEC/CFP2017rate/526.blender_r/526.blender_r - many small
changes, some extra stores gets vectorized.
External/SPEC/CINT2017speed/625.x264_s/625.x264_s
External/SPEC/CINT2017rate/525.x264_r/525.x264_r
x264 has one change in a loop body, in function ssim_end4, some code
remain scalar, resulting in less code size.
External/SPEC/CFP2017rate/511.povray_r/511.povray_r - some extra code
gets vectorized, looks like some other patterns were matched.
MultiSource/Benchmarks/7zip/7zip-benchmark - extra stores were
vectorized (looks like the graphs become profitable)
MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg - small
changes in vectorized code (some small part remain scalar).
External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r
External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s
Many changes cause by the fact that the code of one function becomes
smaller (onvertLCHabToRGB) and this functions gets inlined after that.
MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc - some small
changes here and there, some extra code is vectorized, some remain
scalar (2 x vectors)
MultiSource/Benchmarks/VersaBench/ecbdes/ecbdes - emits 2 scalars
+ 2 insertelems instead of insert, broadcast, alt code (3 instructions,
  total 5 insts)
MultiSource/Benchmarks/Prolangs-C/plot2fig/plot2fig - small graph
becomes profitable and gets vectorized.
External/SPEC/CINT2017rate/502.gcc_r/502.gcc_r
External/SPEC/CINT2017speed/602.gcc_s/602.gcc_s
Some small graph becomes profitable and gets vectorized.
MultiSource/Benchmarks/FreeBench/pifft/pifft - no changes in final code.

Reviewers: RKSimon, dtcxzyw

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/84978
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 128 ++++++++++++++++++
 .../AArch64/extractelements-to-shuffle.ll     |  40 +++---
 .../X86/ext-int-reduced-not-operand.ll        |   9 +-
 .../X86/gather-move-out-of-loop.ll            |  10 +-
 ...gathered-delayed-nodes-with-reused-user.ll |  18 ++-
 .../non-scheduled-inst-reused-as-last-inst.ll |  12 +-
 .../X86/reorder_with_external_users.ll        |  16 +--
 .../SLPVectorizer/alternate-non-profitable.ll |  36 +++--
 8 files changed, 192 insertions(+), 77 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 6b758f63a7961..4ac719af60295 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -2995,6 +2995,15 @@ class BoUpSLP {
     return ScalarToTreeEntry.lookup(V);
   }
 
+  /// Check that the operand node of alternate node does not generate
+  /// buildvector sequence. If it is, then probably not worth it to build
+  /// alternate shuffle, if number of buildvector operands + alternate
+  /// instruction > than the number of buildvector instructions.
+  /// \param S the instructions state of the analyzed values.
+  /// \param VL list of the instructions with alternate opcodes.
+  bool areAltOperandsProfitable(const InstructionsState &S,
+                                ArrayRef<Value *> VL) const;
+
   /// Checks if the specified list of the instructions/values can be vectorized
   /// and fills required data before actual scheduling of the instructions.
   TreeEntry::EntryState getScalarsVectorizationState(
@@ -5777,6 +5786,117 @@ static bool isAlternateInstruction(const Instruction *I,
                                    const Instruction *AltOp,
                                    const TargetLibraryInfo &TLI);
 
+bool BoUpSLP::areAltOperandsProfitable(const InstructionsState &S,
+                                       ArrayRef<Value *> VL) const {
+  unsigned Opcode0 = S.getOpcode();
+  unsigned Opcode1 = S.getAltOpcode();
+  // The opcode mask selects between the two opcodes.
+  SmallBitVector OpcodeMask(VL.size(), false);
+  for (unsigned Lane : seq<unsigned>(0, VL.size()))
+    if (cast<Instruction>(VL[Lane])->getOpcode() == Opcode1)
+      OpcodeMask.set(Lane);
+  // If this pattern is supported by the target then consider it profitable.
+  if (TTI->isLegalAltInstr(FixedVectorType::get(S.MainOp->getType(), VL.size()),
+                           Opcode0, Opcode1, OpcodeMask))
+    return true;
+  SmallVector<ValueList> Operands;
+  for (unsigned I : seq<unsigned>(0, S.MainOp->getNumOperands())) {
+    Operands.emplace_back();
+    // Prepare the operand vector.
+    for (Value *V : VL)
+      Operands.back().push_back(cast<Instruction>(V)->getOperand(I));
+  }
+  if (Operands.size() == 2) {
+    // Try find best operands candidates.
+    for (unsigned I : seq<unsigned>(0, VL.size() - 1)) {
+      SmallVector<std::pair<Value *, Value *>> Candidates(3);
+      Candidates[0] = std::make_pair(Operands[0][I], Operands[0][I + 1]);
+      Candidates[1] = std::make_pair(Operands[0][I], Operands[1][I + 1]);
+      Candidates[2] = std::make_pair(Operands[1][I], Operands[0][I + 1]);
+      std::optional<int> Res = findBestRootPair(Candidates);
+      switch (Res.value_or(0)) {
+      case 0:
+        break;
+      case 1:
+        std::swap(Operands[0][I + 1], Operands[1][I + 1]);
+        break;
+      case 2:
+        std::swap(Operands[0][I], Operands[1][I]);
+        break;
+      default:
+        llvm_unreachable("Unexpected index.");
+      }
+    }
+  }
+  DenseSet<unsigned> UniqueOpcodes;
+  constexpr unsigned NumAltInsts = 3; // main + alt + shuffle.
+  unsigned NonInstCnt = 0;
+  // Estimate number of instructions, required for the vectorized node and for
+  // the buildvector node.
+  unsigned UndefCnt = 0;
+  // Count the number of extra shuffles, required for vector nodes.
+  unsigned ExtraShuffleInsts = 0;
+  // Check that operands do not contain same values and create either perfect
+  // diamond match or shuffled match.
+  if (Operands.size() == 2) {
+    // Do not count same operands twice.
+    if (Operands.front() == Operands.back()) {
+      Operands.erase(Operands.begin());
+    } else if (!allConstant(Operands.front()) &&
+               all_of(Operands.front(), [&](Value *V) {
+                 return is_contained(Operands.back(), V);
+               })) {
+      Operands.erase(Operands.begin());
+      ++ExtraShuffleInsts;
+    }
+  }
+  const Loop *L = LI->getLoopFor(S.MainOp->getParent());
+  // Vectorize node, if:
+  // 1. at least single operand is constant or splat.
+  // 2. Operands have many loop invariants (the instructions are not loop
+  // invariants).
+  // 3. At least single unique operands is supposed to vectorized.
+  return none_of(Operands,
+                 [&](ArrayRef<Value *> Op) {
+                   if (allConstant(Op) ||
+                       (!isSplat(Op) && allSameBlock(Op) && allSameType(Op) &&
+                        getSameOpcode(Op, *TLI).MainOp))
+                     return false;
+                   DenseMap<Value *, unsigned> Uniques;
+                   for (Value *V : Op) {
+                     if (isa<Constant, ExtractElementInst>(V) ||
+                         getTreeEntry(V) || (L && L->isLoopInvariant(V))) {
+                       if (isa<UndefValue>(V))
+                         ++UndefCnt;
+                       continue;
+                     }
+                     auto Res = Uniques.try_emplace(V, 0);
+                     // Found first duplicate - need to add shuffle.
+                     if (!Res.second && Res.first->second == 1)
+                       ++ExtraShuffleInsts;
+                     ++Res.first->getSecond();
+                     if (auto *I = dyn_cast<Instruction>(V))
+                       UniqueOpcodes.insert(I->getOpcode());
+                     else if (Res.second)
+                       ++NonInstCnt;
+                   }
+                   return none_of(Uniques, [&](const auto &P) {
+                     return P.first->hasNUsesOrMore(P.second + 1) &&
+                            none_of(P.first->users(), [&](User *U) {
+                              return getTreeEntry(U) || Uniques.contains(U);
+                            });
+                   });
+                 }) ||
+         // Do not vectorize node, if estimated number of vector instructions is
+         // more than estimated number of buildvector instructions. Number of
+         // vector operands is number of vector instructions + number of vector
+         // instructions for operands (buildvectors). Number of buildvector
+         // instructions is just number_of_operands * number_of_scalars.
+         (UndefCnt < (VL.size() - 1) * S.MainOp->getNumOperands() &&
+          (UniqueOpcodes.size() + NonInstCnt + ExtraShuffleInsts +
+           NumAltInsts) < S.MainOp->getNumOperands() * VL.size());
+}
+
 BoUpSLP::TreeEntry::EntryState BoUpSLP::getScalarsVectorizationState(
     InstructionsState &S, ArrayRef<Value *> VL, bool IsScatterVectorizeUserTE,
     OrdersType &CurrentOrder, SmallVectorImpl<Value *> &PointerOps) const {
@@ -6074,6 +6194,14 @@ BoUpSLP::TreeEntry::EntryState BoUpSLP::getScalarsVectorizationState(
       LLVM_DEBUG(dbgs() << "SLP: ShuffleVector are not vectorized.\n");
       return TreeEntry::NeedToGather;
     }
+    if (!areAltOperandsProfitable(S, VL)) {
+      LLVM_DEBUG(
+          dbgs()
+          << "SLP: ShuffleVector not vectorized, operands are buildvector and "
+             "the whole alt sequence is not profitable.\n");
+      return TreeEntry::NeedToGather;
+    }
+
     return TreeEntry::Vectorize;
   }
   default:
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/extractelements-to-shuffle.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/extractelements-to-shuffle.ll
index d2711d0546c0a..283cc07dfb9b9 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/extractelements-to-shuffle.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/extractelements-to-shuffle.ll
@@ -104,16 +104,16 @@ define void @dist_vec(ptr nocapture noundef readonly %pA, ptr nocapture noundef
 ; CHECK-NEXT:    [[AND95:%.*]] = and i32 [[B_0278]], 1
 ; CHECK-NEXT:    [[SHR96]] = lshr i32 [[A_0279]], 1
 ; CHECK-NEXT:    [[SHR97]] = lshr i32 [[B_0278]], 1
-; CHECK-NEXT:    [[TMP22:%.*]] = insertelement <2 x i32> poison, i32 [[AND94]], i32 0
-; CHECK-NEXT:    [[TMP23:%.*]] = shufflevector <2 x i32> [[TMP22]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq <2 x i32> [[TMP23]], zeroinitializer
-; CHECK-NEXT:    [[TMP25:%.*]] = icmp ne <2 x i32> [[TMP23]], zeroinitializer
-; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <2 x i1> [[TMP24]], <2 x i1> [[TMP25]], <4 x i32> <i32 0, i32 3, i32 0, i32 3>
-; CHECK-NEXT:    [[TMP27:%.*]] = insertelement <2 x i32> poison, i32 [[AND95]], i32 0
-; CHECK-NEXT:    [[TMP28:%.*]] = shufflevector <2 x i32> [[TMP27]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP29:%.*]] = icmp ne <2 x i32> [[TMP28]], zeroinitializer
-; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq <2 x i32> [[TMP28]], zeroinitializer
-; CHECK-NEXT:    [[TMP31:%.*]] = shufflevector <2 x i1> [[TMP29]], <2 x i1> [[TMP30]], <4 x i32> <i32 0, i32 3, i32 3, i32 0>
+; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[AND94]], 0
+; CHECK-NEXT:    [[TOBOOL98:%.*]] = icmp ne i32 [[AND95]], 0
+; CHECK-NEXT:    [[TOBOOL100:%.*]] = icmp eq i32 [[AND94]], 0
+; CHECK-NEXT:    [[TOBOOL103:%.*]] = icmp eq i32 [[AND95]], 0
+; CHECK-NEXT:    [[TMP22:%.*]] = insertelement <4 x i1> poison, i1 [[TOBOOL100]], i32 0
+; CHECK-NEXT:    [[TMP23:%.*]] = insertelement <4 x i1> [[TMP22]], i1 [[TOBOOL]], i32 1
+; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <4 x i1> [[TMP23]], <4 x i1> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
+; CHECK-NEXT:    [[TMP25:%.*]] = insertelement <4 x i1> poison, i1 [[TOBOOL98]], i32 0
+; CHECK-NEXT:    [[TMP27:%.*]] = insertelement <4 x i1> [[TMP25]], i1 [[TOBOOL103]], i32 1
+; CHECK-NEXT:    [[TMP31:%.*]] = shufflevector <4 x i1> [[TMP27]], <4 x i1> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP32:%.*]] = select <4 x i1> [[TMP26]], <4 x i1> [[TMP31]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP33:%.*]] = zext <4 x i1> [[TMP32]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP34]] = add <4 x i32> [[TMP21]], [[TMP33]]
@@ -149,16 +149,16 @@ define void @dist_vec(ptr nocapture noundef readonly %pA, ptr nocapture noundef
 ; CHECK-NEXT:    [[AND134:%.*]] = and i32 [[B_1300]], 1
 ; CHECK-NEXT:    [[SHR135]] = lshr i32 [[A_1301]], 1
 ; CHECK-NEXT:    [[SHR136]] = lshr i32 [[B_1300]], 1
-; CHECK-NEXT:    [[TMP39:%.*]] = insertelement <2 x i32> poison, i32 [[AND133]], i32 0
-; CHECK-NEXT:    [[TMP40:%.*]] = shufflevector <2 x i32> [[TMP39]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP41:%.*]] = icmp eq <2 x i32> [[TMP40]], zeroinitializer
-; CHECK-NEXT:    [[TMP42:%.*]] = icmp ne <2 x i32> [[TMP40]], zeroinitializer
-; CHECK-NEXT:    [[TMP43:%.*]] = shufflevector <2 x i1> [[TMP41]], <2 x i1> [[TMP42]], <4 x i32> <i32 0, i32 3, i32 0, i32 3>
-; CHECK-NEXT:    [[TMP44:%.*]] = insertelement <2 x i32> poison, i32 [[AND134]], i32 0
-; CHECK-NEXT:    [[TMP45:%.*]] = shufflevector <2 x i32> [[TMP44]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP46:%.*]] = icmp ne <2 x i32> [[TMP45]], zeroinitializer
-; CHECK-NEXT:    [[TMP47:%.*]] = icmp eq <2 x i32> [[TMP45]], zeroinitializer
-; CHECK-NEXT:    [[TMP48:%.*]] = shufflevector <2 x i1> [[TMP46]], <2 x i1> [[TMP47]], <4 x i32> <i32 0, i32 3, i32 3, i32 0>
+; CHECK-NEXT:    [[TOBOOL137:%.*]] = icmp ne i32 [[AND133]], 0
+; CHECK-NEXT:    [[TOBOOL139:%.*]] = icmp ne i32 [[AND134]], 0
+; CHECK-NEXT:    [[TOBOOL144:%.*]] = icmp eq i32 [[AND133]], 0
+; CHECK-NEXT:    [[TOBOOL147:%.*]] = icmp eq i32 [[AND134]], 0
+; CHECK-NEXT:    [[TMP40:%.*]] = insertelement <4 x i1> poison, i1 [[TOBOOL144]], i32 0
+; CHECK-NEXT:    [[TMP41:%.*]] = insertelement <4 x i1> [[TMP40]], i1 [[TOBOOL137]], i32 1
+; CHECK-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i1> [[TMP41]], <4 x i1> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
+; CHECK-NEXT:    [[TMP42:%.*]] = insertelement <4 x i1> poison, i1 [[TOBOOL139]], i32 0
+; CHECK-NEXT:    [[TMP39:%.*]] = insertelement <4 x i1> [[TMP42]], i1 [[TOBOOL147]], i32 1
+; CHECK-NEXT:    [[TMP48:%.*]] = shufflevector <4 x i1> [[TMP39]], <4 x i1> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP49:%.*]] = select <4 x i1> [[TMP43]], <4 x i1> [[TMP48]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP50:%.*]] = zext <4 x i1> [[TMP49]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP51]] = add <4 x i32> [[TMP38]], [[TMP50]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll b/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
index 05534fa961ee4..b76e26e0fd571 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
@@ -9,13 +9,8 @@ define i64 @wombat() {
 ; CHECK-NEXT:    br label [[BB2]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ 0, [[BB1:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[PHI]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[TMP0]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i1>
-; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i1> [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP3]] to i64
-; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP2]], i32 1
-; CHECK-NEXT:    [[TMP6:%.*]] = zext i1 [[TMP5]] to i64
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[PHI]] to i64
+; CHECK-NEXT:    [[TMP6:%.*]] = sext i32 [[PHI]] to i64
 ; CHECK-NEXT:    [[OR:%.*]] = or i64 [[TMP4]], [[TMP6]]
 ; CHECK-NEXT:    ret i64 [[OR]]
 ;
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll b/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
index 78fc3a60f0514..3c3dea3f1ea88 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
@@ -4,14 +4,12 @@
 define void @test(i16 %0) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  for.body92.preheader:
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i16> <i16 0, i16 poison>, i16 [[TMP0:%.*]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i32>
-; CHECK-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i32>
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> poison, <4 x i32> <i32 0, i32 poison, i32 1, i32 poison>
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> <i32 poison, i32 0, i32 poison, i32 0>, <4 x i32> [[TMP5]], <4 x i32> <i32 4, i32 1, i32 6, i32 3>
 ; CHECK-NEXT:    br label [[FOR_BODY92:%.*]]
 ; CHECK:       for.body92:
+; CHECK-NEXT:    [[CONV177_I:%.*]] = sext i16 0 to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i16 [[TMP0:%.*]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 poison, i32 0>, i32 [[CONV177_I]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP1]], i32 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> zeroinitializer, [[TMP6]]
 ; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr undef, align 8
 ; CHECK-NEXT:    br label [[FOR_BODY92]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll b/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
index 16ede231c200e..19a8aa9b61815 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
@@ -6,21 +6,19 @@ define i64 @foo() {
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    br label [[BB3:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x i64> [ [[TMP5:%.*]], [[BB3]] ]
+; CHECK-NEXT:    [[PHI:%.*]] = phi i64 [ [[ADD:%.*]], [[BB3]] ]
+; CHECK-NEXT:    [[PHI2:%.*]] = phi i64 [ [[TMP9:%.*]], [[BB3]] ]
 ; CHECK-NEXT:    ret i64 0
 ; CHECK:       bb3:
 ; CHECK-NEXT:    [[PHI5:%.*]] = phi i64 [ 0, [[BB:%.*]] ], [ 0, [[BB3]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x i64> [ zeroinitializer, [[BB]] ], [ [[TMP7:%.*]], [[BB3]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[PHI5]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = add <2 x i64> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = or <2 x i64> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    [[TMP5]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP4]], <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 poison, i64 0>, <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT:    [[TMP7]] = add <2 x i64> [[TMP6]], [[TMP2]]
-; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP7]], i32 1
-; CHECK-NEXT:    [[GETELEMENTPTR:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[TMP8]]
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i64> [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i64> [[TMP1]], i32 1
+; CHECK-NEXT:    [[ADD]] = add i64 [[TMP3]], [[TMP2]]
+; CHECK-NEXT:    [[GETELEMENTPTR:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 0
+; CHECK-NEXT:    [[TMP9]] = or i64 [[PHI5]], 0
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[TMP9]], 0
+; CHECK-NEXT:    [[TMP7]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[ADD]], i32 0
 ; CHECK-NEXT:    br i1 false, label [[BB3]], label [[BB1:%.*]]
 ;
 bb:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll b/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
index 3a9eca2bf2e6b..59cd1c0ccddf8 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
@@ -4,22 +4,22 @@
 define void @foo() {
 ; CHECK-LABEL: define void @foo() {
 ; CHECK-NEXT:  bb:
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 0, i32 0
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x i32> [ zeroinitializer, [[BB:%.*]] ], [ [[TMP6:%.*]], [[BB4:%.*]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = shl <2 x i32> [[TMP1]], [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = or <2 x i32> [[TMP1]], [[TMP0]]
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 0, i32 3>
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> [[TMP1]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 0
+; CHECK-NEXT:    [[SHL:%.*]] = shl i32 [[TMP2]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x i32> [[TMP1]], i32 [[SHL]], i32 0
 ; CHECK-NEXT:    [[TMP6]] = or <2 x i32> [[TMP5]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i32> [[TMP6]], i32 0
 ; CHECK-NEXT:    [[CALL:%.*]] = call i64 null(i32 [[TMP7]])
 ; CHECK-NEXT:    br label [[BB4]]
 ; CHECK:       bb4:
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i32> [[TMP6]], i32 1
 ; CHECK-NEXT:    br i1 false, label [[BB5:%.*]], label [[BB1]]
 ; CHECK:       bb5:
-; CHECK-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ [[TMP4]], [[BB4]] ]
+; CHECK-NEXT:    [[PHI6:%.*]] = phi i32 [ [[SHL]], [[BB4]] ]
+; CHECK-NEXT:    [[PHI7:%.*]] = phi i32 [ [[TMP8]], [[BB4]] ]
 ; CHECK-NEXT:    ret void
 ;
 bb:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/reorder_with_external_users.ll b/llvm/test/Transforms/SLPVectorizer/X86/reorder_with_external_users.ll
index 09b3d25fd6dc0..93258f2975f34 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/reorder_with_external_users.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/reorder_with_external_users.ll
@@ -112,10 +112,10 @@ define void @addsub_and_external_users(ptr %A, ptr %ptr) {
 ; CHECK-NEXT:  bb1:
 ; CHECK-NEXT:    [[LD:%.*]] = load double, ptr undef, align 8
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double> poison, double [[LD]], i32 0
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x double> [[TMP0]], <2 x double> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP1:%.*]] = fsub <2 x double> [[SHUFFLE]], <double 1.100000e+00, double 1.200000e+00>
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x double> [[SHUFFLE]], <double 1.100000e+00, double 1.200000e+00>
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP2]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x double> [[TMP0]], <2 x double> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = fsub <2 x double> [[TMP1]], <double 1.100000e+00, double 1.200000e+00>
+; CHECK-NEXT:    [[TMP6:%.*]] = fadd <2 x double> [[TMP1]], <double 1.100000e+00, double 1.200000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> [[TMP6]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP4:%.*]] = fdiv <2 x double> [[TMP3]], <double 2.100000e+00, double 2.200000e+00>
 ; CHECK-NEXT:    [[TMP5:%.*]] = fmul <2 x double> [[TMP4]], <double 3.100000e+00, double 3.200000e+00>
 ; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> poison, <2 x i32> <i32 1, i32 0>
@@ -159,10 +159,10 @@ define void @subadd_and_external_users(ptr %A, ptr %ptr) {
 ; CHECK-NEXT:  bb1:
 ; CHECK-NEXT:    [[LD:%.*]] = load double, ptr undef, align 8
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double> poison, double [[LD]], i32 0
-; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x double> [[TMP0]], <2 x double> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x double> [[SHUFFLE]], <double 1.200000e+00, double 1.100000e+00>
-; CHECK-NEXT:    [[TMP2:%.*]] = fsub <2 x double> [[SHUFFLE]], <double 1.200000e+00, double 1.100000e+00>
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> [[TMP2]], <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x double> [[TMP0]], <2 x double> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = fsub <2 x double> [[TMP1]], <double 1.200000e+00, double 1.100000e+00>
+; CHECK-NEXT:    [[TMP6:%.*]] = fadd <2 x double> [[TMP1]], <double 1.200000e+00, double 1.100000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> [[TMP6]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP4:%.*]] = fdiv <2 x double> [[TMP3]], <double 2.200000e+00, double 2.100000e+00>
 ; CHECK-NEXT:    [[TMP5:%.*]] = fmul <2 x double> [[TMP4]], <double 3.200000e+00, double 3.100000e+00>
 ; CHECK-NEXT:    store <2 x double> [[TMP5]], ptr [[A:%.*]], align 8
diff --git a/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll b/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
index c6e2cf5543e12..287b623f63690 100644
--- a/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
+++ b/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
@@ -33,11 +33,10 @@ define <2 x float> @replace_through_casts(i16 %inp) {
 ; CHECK-LABEL: define <2 x float> @replace_through_casts(
 ; CHECK-SAME: i16 [[INP:%.*]]) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i16 [[INP]], -10
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i16> poison, i16 [[INP]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i16> [[TMP1]], i16 [[ADD]], i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = uitofp <2 x i16> [[TMP2]] to <2 x float>
-; CHECK-NEXT:    [[TMP4:%.*]] = sitofp <2 x i16> [[TMP2]] to <2 x float>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP4]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = uitofp i16 [[INP]] to float
+; CHECK-NEXT:    [[TMP2:%.*]] = sitofp i16 [[ADD]] to float
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = insertelement <2 x float> [[TMP3]], float [[TMP2]], i64 1
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %add = add nsw i16 %inp, -10
@@ -118,11 +117,10 @@ define <2 x i32> @replace_through_int_casts(i16 %inp, <2 x i16> %dead) {
 ; CHECK-LABEL: define <2 x i32> @replace_through_int_casts(
 ; CHECK-SAME: i16 [[INP:%.*]], <2 x i16> [[DEAD:%.*]]) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i16 [[INP]], -10
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i16> poison, i16 [[INP]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i16> [[TMP1]], i16 [[ADD]], i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP2]] to <2 x i32>
-; CHECK-NEXT:    [[TMP4:%.*]] = sext <2 x i16> [[TMP2]] to <2 x i32>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i16 [[INP]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = sext i16 [[ADD]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> poison, i32 [[TMP1]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = insertelement <2 x i32> [[TMP3]], i32 [[TMP2]], i64 1
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %add = add nsw i16 %inp, -10
@@ -136,11 +134,10 @@ define <2 x i32> @replace_through_int_casts(i16 %inp, <2 x i16> %dead) {
 define <2 x i32> @replace_through_int_casts_ele0_only(i16 %inp, <2 x i16> %dead) {
 ; CHECK-LABEL: define <2 x i32> @replace_through_int_casts_ele0_only(
 ; CHECK-SAME: i16 [[INP:%.*]], <2 x i16> [[DEAD:%.*]]) {
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i16> poison, i16 [[INP]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i16> [[TMP1]], <2 x i16> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP2]] to <2 x i32>
-; CHECK-NEXT:    [[TMP4:%.*]] = sext <2 x i16> [[TMP2]] to <2 x i32>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP4]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = sext i16 [[INP]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i16 [[INP]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = insertelement <2 x i32> [[TMP3]], i32 [[TMP1]], i64 1
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %2 = sext i16 %inp to i32
@@ -174,11 +171,10 @@ define <2 x i8> @replace_through_binop_preserve_flags(i8 %inp, <2 x i8> %d, <2 x
 ; CHECK-LABEL: define <2 x i8> @replace_through_binop_preserve_flags(
 ; CHECK-SAME: i8 [[INP:%.*]], <2 x i8> [[D:%.*]], <2 x i8> [[ANY:%.*]]) {
 ; CHECK-NEXT:    [[ADD:%.*]] = xor i8 [[INP]], 5
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i8> poison, i8 [[INP]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x i8> [[TMP1]], i8 [[ADD]], i32 1
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i8> [[TMP2]], <i8 123, i8 1>
-; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i8> [[TMP2]], <i8 123, i8 1>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> [[TMP4]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = xor i8 [[INP]], 123
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw i8 [[ADD]], 1
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i8> poison, i8 [[TMP1]], i64 0
+; CHECK-NEXT:    [[R:%.*]] = insertelement <2 x i8> [[TMP3]], i8 [[TMP2]], i64 1
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %add = xor i8 %inp, 5

From 0a1317564a6b437760d96f0a227a3c910875428d Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Wed, 10 Apr 2024 20:34:58 +0200
Subject: [PATCH 057/886] [libc++] Adds a global private constructor tag.
 (#87920)

This removes the similar tags used in the chrono tzdb implementation.

Fixes: https://github.com/llvm/llvm-project/issues/85432
---
 libcxx/include/CMakeLists.txt                 |  1 +
 libcxx/include/__chrono/leap_second.h         |  4 +--
 libcxx/include/__chrono/time_zone_link.h      |  4 +--
 libcxx/include/__locale                       | 12 ++++----
 libcxx/include/__stop_token/stop_callback.h   |  9 +++---
 .../__utility/private_constructor_tag.h       | 28 +++++++++++++++++++
 libcxx/include/module.modulemap               | 21 +++++++-------
 libcxx/src/CMakeLists.txt                     |  2 --
 libcxx/src/include/tzdb/leap_second_private.h | 27 ------------------
 .../src/include/tzdb/time_zone_link_private.h | 27 ------------------
 libcxx/src/locale.cpp                         |  2 +-
 libcxx/src/tzdb.cpp                           |  6 ++--
 .../private_constructor_tag.compile.pass.cpp  | 19 +++++++++++++
 .../time.zone.leap/assign.copy.pass.cpp       |  2 --
 .../time.zone.leap/cons.copy.pass.cpp         |  2 --
 .../time.zone.leap/members/date.pass.cpp      |  2 --
 .../time.zone.leap/members/value.pass.cpp     |  2 --
 .../nonmembers/comparison.pass.cpp            |  2 --
 libcxx/test/support/test_chrono_leap_second.h |  9 ++----
 libcxx/utils/generate_iwyu_mapping.py         |  1 +
 20 files changed, 79 insertions(+), 103 deletions(-)
 create mode 100644 libcxx/include/__utility/private_constructor_tag.h
 delete mode 100644 libcxx/src/include/tzdb/leap_second_private.h
 delete mode 100644 libcxx/src/include/tzdb/time_zone_link_private.h
 create mode 100644 libcxx/test/libcxx/utilities/utility/private_constructor_tag.compile.pass.cpp

diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index a4a58a787ee9a..d4e8c196a9a88 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -863,6 +863,7 @@ set(files
   __utility/pair.h
   __utility/piecewise_construct.h
   __utility/priority_tag.h
+  __utility/private_constructor_tag.h
   __utility/rel_ops.h
   __utility/small_buffer.h
   __utility/swap.h
diff --git a/libcxx/include/__chrono/leap_second.h b/libcxx/include/__chrono/leap_second.h
index 4e67cc2d65277..557abc15ff184 100644
--- a/libcxx/include/__chrono/leap_second.h
+++ b/libcxx/include/__chrono/leap_second.h
@@ -22,6 +22,7 @@
 #  include <__compare/ordering.h>
 #  include <__compare/three_way_comparable.h>
 #  include <__config>
+#  include <__utility/private_constructor_tag.h>
 
 #  if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #    pragma GCC system_header
@@ -35,9 +36,8 @@ namespace chrono {
 
 class leap_second {
 public:
-  struct __constructor_tag;
   [[nodiscard]]
-  _LIBCPP_HIDE_FROM_ABI explicit constexpr leap_second(__constructor_tag&&, sys_seconds __date, seconds __value)
+  _LIBCPP_HIDE_FROM_ABI explicit constexpr leap_second(__private_constructor_tag, sys_seconds __date, seconds __value)
       : __date_(__date), __value_(__value) {}
 
   _LIBCPP_HIDE_FROM_ABI leap_second(const leap_second&)            = default;
diff --git a/libcxx/include/__chrono/time_zone_link.h b/libcxx/include/__chrono/time_zone_link.h
index 17e915d2677a8..c76ddeff9f966 100644
--- a/libcxx/include/__chrono/time_zone_link.h
+++ b/libcxx/include/__chrono/time_zone_link.h
@@ -18,6 +18,7 @@
 
 #  include <__compare/strong_order.h>
 #  include <__config>
+#  include <__utility/private_constructor_tag.h>
 #  include <string>
 #  include <string_view>
 
@@ -37,9 +38,8 @@ namespace chrono {
 
 class time_zone_link {
 public:
-  struct __constructor_tag;
   _LIBCPP_NODISCARD_EXT
-  _LIBCPP_HIDE_FROM_ABI explicit time_zone_link(__constructor_tag&&, string_view __name, string_view __target)
+  _LIBCPP_HIDE_FROM_ABI explicit time_zone_link(__private_constructor_tag, string_view __name, string_view __target)
       : __name_{__name}, __target_{__target} {}
 
   _LIBCPP_HIDE_FROM_ABI time_zone_link(time_zone_link&&)            = default;
diff --git a/libcxx/include/__locale b/libcxx/include/__locale
index fab87f0d6a279..36ac099d650e4 100644
--- a/libcxx/include/__locale
+++ b/libcxx/include/__locale
@@ -16,6 +16,7 @@
 #include <__mutex/once_flag.h>
 #include <__type_traits/make_unsigned.h>
 #include <__utility/no_destroy.h>
+#include <__utility/private_constructor_tag.h>
 #include <cctype>
 #include <clocale>
 #include <cstdint>
@@ -97,8 +98,7 @@ private:
 
   template <class>
   friend struct __no_destroy;
-  struct __private_tag {};
-  _LIBCPP_HIDE_FROM_ABI explicit locale(__private_tag, __imp* __loc) : __locale_(__loc) {}
+  _LIBCPP_HIDE_FROM_ABI explicit locale(__private_constructor_tag, __imp* __loc) : __locale_(__loc) {}
 
   void __install_ctor(const locale&, facet*, long);
   static locale& __global();
@@ -1248,10 +1248,10 @@ extern template class _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char, char
 #ifndef _LIBCPP_HAS_NO_WIDE_CHARACTERS
 extern template class _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<wchar_t, char, mbstate_t>;
 #endif
-extern template class _LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS
-    codecvt_byname<char16_t, char, mbstate_t>; // deprecated in C++20
-extern template class _LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS
-    codecvt_byname<char32_t, char, mbstate_t>; // deprecated in C++20
+extern template class _LIBCPP_DEPRECATED_IN_CXX20
+_LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char16_t, char, mbstate_t>; // deprecated in C++20
+extern template class _LIBCPP_DEPRECATED_IN_CXX20
+_LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char32_t, char, mbstate_t>; // deprecated in C++20
 #ifndef _LIBCPP_HAS_NO_CHAR8_T
 extern template class _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char16_t, char8_t, mbstate_t>; // C++20
 extern template class _LIBCPP_EXTERN_TEMPLATE_TYPE_VIS codecvt_byname<char32_t, char8_t, mbstate_t>; // C++20
diff --git a/libcxx/include/__stop_token/stop_callback.h b/libcxx/include/__stop_token/stop_callback.h
index 9e5b0338d4667..7b526820f98a3 100644
--- a/libcxx/include/__stop_token/stop_callback.h
+++ b/libcxx/include/__stop_token/stop_callback.h
@@ -21,6 +21,7 @@
 #include <__type_traits/is_nothrow_constructible.h>
 #include <__utility/forward.h>
 #include <__utility/move.h>
+#include <__utility/private_constructor_tag.h>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
@@ -49,13 +50,13 @@ class _LIBCPP_AVAILABILITY_SYNC stop_callback : private __stop_callback_base {
     requires constructible_from<_Callback, _Cb>
   _LIBCPP_HIDE_FROM_ABI explicit stop_callback(const stop_token& __st,
                                                _Cb&& __cb) noexcept(is_nothrow_constructible_v<_Callback, _Cb>)
-      : stop_callback(__private_tag{}, __st.__state_, std::forward<_Cb>(__cb)) {}
+      : stop_callback(__private_constructor_tag{}, __st.__state_, std::forward<_Cb>(__cb)) {}
 
   template <class _Cb>
     requires constructible_from<_Callback, _Cb>
   _LIBCPP_HIDE_FROM_ABI explicit stop_callback(stop_token&& __st,
                                                _Cb&& __cb) noexcept(is_nothrow_constructible_v<_Callback, _Cb>)
-      : stop_callback(__private_tag{}, std::move(__st.__state_), std::forward<_Cb>(__cb)) {}
+      : stop_callback(__private_constructor_tag{}, std::move(__st.__state_), std::forward<_Cb>(__cb)) {}
 
   _LIBCPP_HIDE_FROM_ABI ~stop_callback() {
     if (__state_) {
@@ -74,10 +75,8 @@ class _LIBCPP_AVAILABILITY_SYNC stop_callback : private __stop_callback_base {
 
   friend __stop_callback_base;
 
-  struct __private_tag {};
-
   template <class _StatePtr, class _Cb>
-  _LIBCPP_HIDE_FROM_ABI explicit stop_callback(__private_tag, _StatePtr&& __state, _Cb&& __cb) noexcept(
+  _LIBCPP_HIDE_FROM_ABI explicit stop_callback(__private_constructor_tag, _StatePtr&& __state, _Cb&& __cb) noexcept(
       is_nothrow_constructible_v<_Callback, _Cb>)
       : __stop_callback_base([](__stop_callback_base* __cb_base) noexcept {
           // stop callback is supposed to only be called once
diff --git a/libcxx/include/__utility/private_constructor_tag.h b/libcxx/include/__utility/private_constructor_tag.h
new file mode 100644
index 0000000000000..462cab48c9edd
--- /dev/null
+++ b/libcxx/include/__utility/private_constructor_tag.h
@@ -0,0 +1,28 @@
+// -*- C++ -*-
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP__UTILITY_PRIVATE_CONSTRUCTOR_TAG_H
+#define _LIBCPP__UTILITY_PRIVATE_CONSTRUCTOR_TAG_H
+
+#include <__config>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+
+// This tag allows defining non-standard exposition-only constructors while
+// preventing users from being able to use them, since this reserved-name tag
+// needs to be used.
+struct __private_constructor_tag {};
+
+_LIBCPP_END_NAMESPACE_STD
+
+#endif // _LIBCPP__UTILITY_PRIVATE_CONSTRUCTOR_TAG_H
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index 011a4818ab9d2..9c0f0ddd20c96 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -2090,18 +2090,19 @@ module std_private_utility_pair                   [system] {
   export std_private_type_traits_is_nothrow_move_assignable
   export std_private_utility_pair_fwd
 }
-module std_private_utility_pair_fwd               [system] { header "__fwd/pair.h" }
-module std_private_utility_piecewise_construct    [system] { header "__utility/piecewise_construct.h" }
-module std_private_utility_priority_tag           [system] { header "__utility/priority_tag.h" }
-module std_private_utility_rel_ops                [system] { header "__utility/rel_ops.h" }
-module std_private_utility_small_buffer           [system] { header "__utility/small_buffer.h" }
-module std_private_utility_swap                   [system] {
+module std_private_utility_pair_fwd                [system] { header "__fwd/pair.h" }
+module std_private_utility_piecewise_construct     [system] { header "__utility/piecewise_construct.h" }
+module std_private_utility_priority_tag            [system] { header "__utility/priority_tag.h" }
+module std_private_utility_private_constructor_tag [system] { header "__utility/private_constructor_tag.h" }
+module std_private_utility_rel_ops                 [system] { header "__utility/rel_ops.h" }
+module std_private_utility_small_buffer            [system] { header "__utility/small_buffer.h" }
+module std_private_utility_swap                    [system] {
   header "__utility/swap.h"
   export std_private_type_traits_is_swappable
 }
-module std_private_utility_to_underlying          [system] { header "__utility/to_underlying.h" }
-module std_private_utility_unreachable            [system] { header "__utility/unreachable.h" }
+module std_private_utility_to_underlying           [system] { header "__utility/to_underlying.h" }
+module std_private_utility_unreachable             [system] { header "__utility/unreachable.h" }
 
-module std_private_variant_monostate [system] { header "__variant/monostate.h" }
+module std_private_variant_monostate               [system] { header "__variant/monostate.h" }
 
-module std_private_vector_fwd [system] { header "__fwd/vector.h" }
+module std_private_vector_fwd                      [system] { header "__fwd/vector.h" }
diff --git a/libcxx/src/CMakeLists.txt b/libcxx/src/CMakeLists.txt
index 16ccb80ba3326..208500ec14fcd 100644
--- a/libcxx/src/CMakeLists.txt
+++ b/libcxx/src/CMakeLists.txt
@@ -334,8 +334,6 @@ endif()
 
 if (LIBCXX_ENABLE_LOCALIZATION AND LIBCXX_ENABLE_FILESYSTEM AND LIBCXX_ENABLE_TIME_ZONE_DATABASE)
   list(APPEND LIBCXX_EXPERIMENTAL_SOURCES
-    include/tzdb/leap_second_private.h
-    include/tzdb/time_zone_link_private.h
     include/tzdb/time_zone_private.h
     include/tzdb/types_private.h
     include/tzdb/tzdb_list_private.h
diff --git a/libcxx/src/include/tzdb/leap_second_private.h b/libcxx/src/include/tzdb/leap_second_private.h
deleted file mode 100644
index 7a811ab197594..0000000000000
--- a/libcxx/src/include/tzdb/leap_second_private.h
+++ /dev/null
@@ -1,27 +0,0 @@
-// -*- C++ -*-
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// For information see https://libcxx.llvm.org/DesignDocs/TimeZone.html
-
-#ifndef _LIBCPP_SRC_INCLUDE_TZDB_LEAP_SECOND_PRIVATE_H
-#define _LIBCPP_SRC_INCLUDE_TZDB_LEAP_SECOND_PRIVATE_H
-
-#include <chrono>
-
-_LIBCPP_BEGIN_NAMESPACE_STD
-
-namespace chrono {
-
-struct leap_second::__constructor_tag {};
-
-} // namespace chrono
-
-_LIBCPP_END_NAMESPACE_STD
-
-#endif // _LIBCPP_SRC_INCLUDE_TZDB_LEAP_SECOND_PRIVATE_H
diff --git a/libcxx/src/include/tzdb/time_zone_link_private.h b/libcxx/src/include/tzdb/time_zone_link_private.h
deleted file mode 100644
index 139237625274d..0000000000000
--- a/libcxx/src/include/tzdb/time_zone_link_private.h
+++ /dev/null
@@ -1,27 +0,0 @@
-// -*- C++ -*-
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// For information see https://libcxx.llvm.org/DesignDocs/TimeZone.html
-
-#ifndef _LIBCPP_SRC_INCLUDE_TZDB_TIME_ZONE_LINK_PRIVATE_H
-#define _LIBCPP_SRC_INCLUDE_TZDB_TIME_ZONE_LINK_PRIVATE_H
-
-#include <chrono>
-
-_LIBCPP_BEGIN_NAMESPACE_STD
-
-namespace chrono {
-
-struct time_zone_link::__constructor_tag {};
-
-} // namespace chrono
-
-_LIBCPP_END_NAMESPACE_STD
-
-#endif // _LIBCPP_SRC_INCLUDE_TZDB_TIME_ZONE_LINK_PRIVATE_H
diff --git a/libcxx/src/locale.cpp b/libcxx/src/locale.cpp
index 7fdd5be181ed9..1ca88e30f63ac 100644
--- a/libcxx/src/locale.cpp
+++ b/libcxx/src/locale.cpp
@@ -497,7 +497,7 @@ constinit __no_destroy<locale::__imp>
     locale::__imp::classic_locale_imp_(__uninitialized_tag{}); // initialized below in classic()
 
 const locale& locale::classic() {
-  static const __no_destroy<locale> classic_locale(__private_tag{}, [] {
+  static const __no_destroy<locale> classic_locale(__private_constructor_tag{}, [] {
     // executed exactly once on first initialization of `classic_locale`
     locale::__imp::classic_locale_imp_.__emplace(1u);
     return &locale::__imp::classic_locale_imp_.__get();
diff --git a/libcxx/src/tzdb.cpp b/libcxx/src/tzdb.cpp
index 8909ecd026add..d521d810523ea 100644
--- a/libcxx/src/tzdb.cpp
+++ b/libcxx/src/tzdb.cpp
@@ -15,8 +15,6 @@
 #include <stdexcept>
 #include <string>
 
-#include "include/tzdb/leap_second_private.h"
-#include "include/tzdb/time_zone_link_private.h"
 #include "include/tzdb/time_zone_private.h"
 #include "include/tzdb/types_private.h"
 #include "include/tzdb/tzdb_list_private.h"
@@ -582,7 +580,7 @@ static void __parse_link(tzdb& __tzdb, istream& __input) {
   string __name = chrono::__parse_string(__input);
   chrono::__skip_line(__input);
 
-  __tzdb.links.emplace_back(time_zone_link::__constructor_tag{}, std::move(__name), std::move(__target));
+  __tzdb.links.emplace_back(std::__private_constructor_tag{}, std::move(__name), std::move(__target));
 }
 
 static void __parse_tzdata(tzdb& __db, __tz::__rules_storage_type& __rules, istream& __input) {
@@ -649,7 +647,7 @@ static void __parse_leap_seconds(vector<leap_second>& __leap_seconds, istream&&
     seconds __value{chrono::__parse_integral(__input, false)};
     chrono::__skip_line(__input);
 
-    __leap_seconds.emplace_back(leap_second::__constructor_tag{}, __date, __value);
+    __leap_seconds.emplace_back(std::__private_constructor_tag{}, __date, __value);
   }
 }
 
diff --git a/libcxx/test/libcxx/utilities/utility/private_constructor_tag.compile.pass.cpp b/libcxx/test/libcxx/utilities/utility/private_constructor_tag.compile.pass.cpp
new file mode 100644
index 0000000000000..097e05f29cebd
--- /dev/null
+++ b/libcxx/test/libcxx/utilities/utility/private_constructor_tag.compile.pass.cpp
@@ -0,0 +1,19 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// struct __private_constructor_tag{};
+
+// The private constructor tag is intended to be a trivial type that can easily
+// be used to mark a constructor exposition-only.
+//
+// Tests whether the type is trivial.
+
+#include <__utility/private_constructor_tag.h>
+#include <type_traits>
+
+static_assert(std::is_trivial<std::__private_constructor_tag>::value, "");
diff --git a/libcxx/test/std/time/time.zone/time.zone.leap/assign.copy.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.leap/assign.copy.pass.cpp
index 4d91e73f38e41..6918ed6be5c14 100644
--- a/libcxx/test/std/time/time.zone/time.zone.leap/assign.copy.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.leap/assign.copy.pass.cpp
@@ -27,8 +27,6 @@
 #include <type_traits>
 #include <cassert>
 
-// Add the include path required by test_chrono_leap_second.h when using libc++.
-// ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
 #include "test_chrono_leap_second.h"
 
 constexpr bool test() {
diff --git a/libcxx/test/std/time/time.zone/time.zone.leap/cons.copy.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.leap/cons.copy.pass.cpp
index e2419b7d1f09d..3dad08968d12a 100644
--- a/libcxx/test/std/time/time.zone/time.zone.leap/cons.copy.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.leap/cons.copy.pass.cpp
@@ -25,8 +25,6 @@
 #include <concepts>
 #include <cassert>
 
-// Add the include path required by test_chrono_leap_second.h when using libc++.
-// ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
 #include "test_chrono_leap_second.h"
 
 constexpr bool test() {
diff --git a/libcxx/test/std/time/time.zone/time.zone.leap/members/date.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.leap/members/date.pass.cpp
index 23f95eccfdecd..6f9fe1c47d351 100644
--- a/libcxx/test/std/time/time.zone/time.zone.leap/members/date.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.leap/members/date.pass.cpp
@@ -23,8 +23,6 @@
 
 #include "test_macros.h"
 
-// Add the include path required by test_chrono_leap_second.h when using libc++.
-// ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
 #include "test_chrono_leap_second.h"
 
 constexpr void test(const std::chrono::leap_second leap_second, std::chrono::sys_seconds expected) {
diff --git a/libcxx/test/std/time/time.zone/time.zone.leap/members/value.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.leap/members/value.pass.cpp
index 844c74d002ac5..652e51ef0bf10 100644
--- a/libcxx/test/std/time/time.zone/time.zone.leap/members/value.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.leap/members/value.pass.cpp
@@ -23,8 +23,6 @@
 
 #include "test_macros.h"
 
-// Add the include path required by test_chrono_leap_second.h when using libc++.
-// ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
 #include "test_chrono_leap_second.h"
 
 constexpr void test(const std::chrono::leap_second leap_second, std::chrono::seconds expected) {
diff --git a/libcxx/test/std/time/time.zone/time.zone.leap/nonmembers/comparison.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.leap/nonmembers/comparison.pass.cpp
index ac8b780af854d..bf6855ea63dfc 100644
--- a/libcxx/test/std/time/time.zone/time.zone.leap/nonmembers/comparison.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.leap/nonmembers/comparison.pass.cpp
@@ -50,8 +50,6 @@
 #include "test_macros.h"
 #include "test_comparisons.h"
 
-// Add the include path required by test_chrono_leap_second.h when using libc++.
-// ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
 #include "test_chrono_leap_second.h"
 
 constexpr void test_comparison(const std::chrono::leap_second lhs, const std::chrono::leap_second rhs) {
diff --git a/libcxx/test/support/test_chrono_leap_second.h b/libcxx/test/support/test_chrono_leap_second.h
index 485f68d91b1a1..be5ce760bfe98 100644
--- a/libcxx/test/support/test_chrono_leap_second.h
+++ b/libcxx/test/support/test_chrono_leap_second.h
@@ -32,16 +32,11 @@
 
 #ifdef _LIBCPP_VERSION
 
-// In order to find this include the calling test needs to provide this path in
-// the search path. Typically this looks like:
-//   ADDITIONAL_COMPILE_FLAGS(stdlib=libc++): -I %{libcxx-dir}/src/include
-// where the number of `../` sequences depends on the subdirectory level of the
-// test.
-#  include "tzdb/leap_second_private.h" // Header in the dylib
+#  include <__utility/private_constructor_tag.h>
 
 inline constexpr std::chrono::leap_second
 test_leap_second_create(const std::chrono::sys_seconds& date, const std::chrono::seconds& value) {
-  return std::chrono::leap_second{std::chrono::leap_second::__constructor_tag{}, date, value};
+  return std::chrono::leap_second{std::__private_constructor_tag{}, date, value};
 }
 
 #else // _LIBCPP_VERSION
diff --git a/libcxx/utils/generate_iwyu_mapping.py b/libcxx/utils/generate_iwyu_mapping.py
index 8ab7b86299edc..2265438ab49cc 100644
--- a/libcxx/utils/generate_iwyu_mapping.py
+++ b/libcxx/utils/generate_iwyu_mapping.py
@@ -11,6 +11,7 @@ def IWYU_mapping(header: str) -> typing.Optional[typing.List[str]]:
         "__debug_utils/.+",
         "__fwd/get[.]h",
         "__support/.+",
+        "__utility/private_constructor_tag.h",
     ]
     if any(re.match(pattern, header) for pattern in ignore):
         return None

From f81879c0f70ee5a1cf1d5b716dfd49d1a271cc2d Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Wed, 10 Apr 2024 13:35:04 -0500
Subject: [PATCH 058/886] [Libomptarget] Add RPC-based printf implementation
 for OpenMP #85638

Summary:
Relanding after reverting, only applies to AMDGPU for now.

This patch adds an implementation of printf that's provided by the GPU
C library runtime. This pritnf currently implemented using the same
wrapper handling that OpenMP sets up. This will be removed once we have
proper varargs support.

This printf differs from the one CUDA offers in that it is synchronous
and uses a finite size. Additionally we support pretty much every
format specifier except the %n option.

Depends on #85331
---
 openmp/libomptarget/DeviceRTL/CMakeLists.txt |  5 +++++
 openmp/libomptarget/DeviceRTL/src/LibC.cpp   | 15 ++++++++++++++-
 2 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/openmp/libomptarget/DeviceRTL/CMakeLists.txt b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
index 2509f1276ccee..2e7f28df24d64 100644
--- a/openmp/libomptarget/DeviceRTL/CMakeLists.txt
+++ b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
@@ -122,6 +122,11 @@ set(clang_opt_flags -O3 -mllvm -openmp-opt-disable -DSHARED_SCRATCHPAD_SIZE=512
 set(link_opt_flags  -O3        -openmp-opt-disable -attributor-enable=module -vectorize-slp=false )
 set(link_export_flag -passes=internalize -internalize-public-api-file=${source_directory}/exports)
 
+# If the user built with the GPU C library enabled we will use that instead.
+if(${LIBOMPTARGET_GPU_LIBC_SUPPORT})
+  list(APPEND clang_opt_flags -DOMPTARGET_HAS_LIBC)
+endif()
+
 # Prepend -I to each list element
 set (LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL "${LIBOMPTARGET_LLVM_INCLUDE_DIRS}")
 list(TRANSFORM LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL PREPEND "-I")
diff --git a/openmp/libomptarget/DeviceRTL/src/LibC.cpp b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
index af675b97256f6..e587c3057f5ba 100644
--- a/openmp/libomptarget/DeviceRTL/src/LibC.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
@@ -25,13 +25,26 @@ int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t) {
 } // namespace impl
 #pragma omp end declare variant
 
-// We do not have a vprintf implementation for AMD GPU yet so we use a stub.
 #pragma omp begin declare variant match(device = {arch(amdgcn)})
+
+#ifdef OMPTARGET_HAS_LIBC
+// TODO: Remove this handling once we have varargs support.
+extern "C" struct FILE *stdout;
+extern "C" int32_t rpc_fprintf(FILE *, const char *, void *, uint64_t);
+
+namespace impl {
+int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t Size) {
+  return rpc_fprintf(stdout, Format, Arguments, Size);
+}
+} // namespace impl
+#else
+// We do not have a vprintf implementation for AMD GPU so we use a stub.
 namespace impl {
 int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t) {
   return -1;
 }
 } // namespace impl
+#endif
 #pragma omp end declare variant
 
 extern "C" {

From fad14707b73d6387e6276507e1c5726e67f08cd6 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Wed, 10 Apr 2024 14:07:18 -0500
Subject: [PATCH 059/886] [libc] Add note to use `LIBC_GPU_BUILD=ON` as another
 form

Summary:
This is a shorthand to enable GPU support so it should be listed in the
docs.
---
 libc/docs/gpu/building.rst | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/libc/docs/gpu/building.rst b/libc/docs/gpu/building.rst
index 6d94134a407d3..d3e64c6d42431 100644
--- a/libc/docs/gpu/building.rst
+++ b/libc/docs/gpu/building.rst
@@ -33,7 +33,8 @@ The simplest way to build the GPU libc is to use the existing LLVM runtimes
 support. This will automatically handle bootstrapping an up-to-date ``clang``
 compiler and using it to build the C library. The following CMake invocation
 will instruct it to build the ``libc`` runtime targeting both AMD and NVIDIA
-GPUs.
+GPUs. The ``LIBC_GPU_BUILD`` option can also be enabled to add the relevant
+arguments automatically.
 
 .. code-block:: sh
 
@@ -234,6 +235,10 @@ standard runtime build.
   This flag controls whether or not the libc build will generate its own
   headers. This must always be on when targeting the GPU.
 
+**LIBC_GPU_BUILD**:BOOL
+  Shorthand for enabling GPU support. Equivalent to enabling support for both
+  AMDGPU and NVPTX builds for ``libc``.
+
 **LIBC_GPU_TEST_ARCHITECTURE**:STRING
   Sets the architecture used to build the GPU tests for, such as ``gfx90a`` or
   ``sm_80`` for AMD and NVIDIA GPUs respectively. The default behavior is to

From ca6b8469c16edfe1713e9050dca3cd68bd585410 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Wed, 10 Apr 2024 12:20:28 -0700
Subject: [PATCH 060/886] [ELF] Avoid unneeded config->isLE and
 config->wordsize. NFC

---
 lld/ELF/SyntheticSections.cpp | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/lld/ELF/SyntheticSections.cpp b/lld/ELF/SyntheticSections.cpp
index 550659464a440..d8791e83dc9e5 100644
--- a/lld/ELF/SyntheticSections.cpp
+++ b/lld/ELF/SyntheticSections.cpp
@@ -2769,7 +2769,8 @@ readPubNamesAndTypes(const LLDDwarfObj<ELFT> &obj,
 
   SmallVector<GdbIndexSection::NameAttrEntry, 0> ret;
   for (const LLDDWARFSection *pub : {&pubNames, &pubTypes}) {
-    DWARFDataExtractor data(obj, *pub, config->isLE, config->wordsize);
+    DWARFDataExtractor data(obj, *pub, ELFT::Endianness == endianness::little,
+                            ELFT::Is64Bits ? 8 : 4);
     DWARFDebugPubTable table;
     table.extract(data, /*GnuStyle=*/true, [&](Error e) {
       warn(toString(pub->sec) + ": " + toString(std::move(e)));
@@ -3744,8 +3745,9 @@ template <typename ELFT> void elf::writeEhdr(uint8_t *buf, Partition &part) {
   memcpy(buf, "\177ELF", 4);
 
   auto *eHdr = reinterpret_cast<typename ELFT::Ehdr *>(buf);
-  eHdr->e_ident[EI_CLASS] = config->is64 ? ELFCLASS64 : ELFCLASS32;
-  eHdr->e_ident[EI_DATA] = config->isLE ? ELFDATA2LSB : ELFDATA2MSB;
+  eHdr->e_ident[EI_CLASS] = ELFT::Is64Bits ? ELFCLASS64 : ELFCLASS32;
+  eHdr->e_ident[EI_DATA] =
+      ELFT::Endianness == endianness::little ? ELFDATA2LSB : ELFDATA2MSB;
   eHdr->e_ident[EI_VERSION] = EV_CURRENT;
   eHdr->e_ident[EI_OSABI] = config->osabi;
   eHdr->e_ident[EI_ABIVERSION] = getAbiVersion();

From e3ef4612c18845876cda9a13c3435e102f74a3aa Mon Sep 17 00:00:00 2001
From: shamithoke <152091883+shamithoke@users.noreply.github.com>
Date: Thu, 11 Apr 2024 00:52:44 +0530
Subject: [PATCH 061/886] Perform bitreverse using AVX512 GFNI for i32 and i64.
 (#81764)

Currently, the lowering operation for bitreverse using Intel AVX512 GFNI only supports byte vectors

Extend the operation to i32 and i64.

---------

Co-authored-by: shami <shami_thoke@yahoo.com>
---
 llvm/lib/Target/X86/X86ISelLowering.cpp    |  22 ++
 llvm/test/CodeGen/X86/bitreverse.ll        | 281 ++++-----------------
 llvm/test/CodeGen/X86/vector-bitreverse.ll |  46 +---
 3 files changed, 84 insertions(+), 265 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 010f9c30ab403..52be35aafb0f5 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1496,6 +1496,11 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::TRUNCATE,          MVT::v32i32, Custom);
     setOperationAction(ISD::TRUNCATE,          MVT::v32i64, Custom);
 
+    if (Subtarget.hasGFNI()) {
+      setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
+      setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
+    }
+
     for (auto VT : { MVT::v32i8, MVT::v16i16, MVT::v8i32, MVT::v4i64 }) {
       setOperationAction(ISD::SETCC,           VT, Custom);
       setOperationAction(ISD::CTPOP,           VT, Custom);
@@ -31332,6 +31337,23 @@ static SDValue LowerBITREVERSE(SDValue Op, const X86Subtarget &Subtarget,
   if (VT.is256BitVector() && !Subtarget.hasInt256())
     return splitVectorIntUnary(Op, DAG, DL);
 
+  // Lower i32/i64 to GFNI as vXi8 BITREVERSE + BSWAP
+  if (!VT.isVector()) {
+
+    assert((VT.getScalarType() == MVT::i32) ||
+           (VT.getScalarType() == MVT::i64));
+
+    MVT VecVT = MVT::getVectorVT(VT, 128 / VT.getSizeInBits());
+    SDValue Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VecVT, In);
+    Res = DAG.getNode(ISD::BITREVERSE, DL, MVT::v16i8,
+                      DAG.getBitcast(MVT::v16i8, Res));
+    Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT,
+                      DAG.getBitcast(VecVT, Res), DAG.getIntPtrConstant(0, DL));
+    return DAG.getNode(ISD::BSWAP, DL, VT, Res);
+  }
+
+  assert(VT.isVector() && VT.getSizeInBits() >= 128);
+
   // Lower vXi16/vXi32/vXi64 as BSWAP + vXi8 BITREVERSE.
   if (VT.getScalarType() != MVT::i8) {
     MVT ByteVT = MVT::getVectorVT(MVT::i8, VT.getSizeInBits() / 8);
diff --git a/llvm/test/CodeGen/X86/bitreverse.ll b/llvm/test/CodeGen/X86/bitreverse.ll
index 26b1d64874e59..704563ab1bbf7 100644
--- a/llvm/test/CodeGen/X86/bitreverse.ll
+++ b/llvm/test/CodeGen/X86/bitreverse.ll
@@ -172,26 +172,10 @@ define i64 @test_bitreverse_i64(i64 %a) nounwind {
 ;
 ; GFNI-LABEL: test_bitreverse_i64:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    bswapq %rdi
-; GFNI-NEXT:    movq %rdi, %rax
-; GFNI-NEXT:    shrq $4, %rax
-; GFNI-NEXT:    movabsq $1085102592571150095, %rcx # imm = 0xF0F0F0F0F0F0F0F
-; GFNI-NEXT:    andq %rcx, %rax
-; GFNI-NEXT:    andq %rcx, %rdi
-; GFNI-NEXT:    shlq $4, %rdi
-; GFNI-NEXT:    orq %rax, %rdi
-; GFNI-NEXT:    movabsq $3689348814741910323, %rax # imm = 0x3333333333333333
-; GFNI-NEXT:    movq %rdi, %rcx
-; GFNI-NEXT:    andq %rax, %rcx
-; GFNI-NEXT:    shrq $2, %rdi
-; GFNI-NEXT:    andq %rax, %rdi
-; GFNI-NEXT:    leaq (%rdi,%rcx,4), %rax
-; GFNI-NEXT:    movabsq $6148914691236517205, %rcx # imm = 0x5555555555555555
-; GFNI-NEXT:    movq %rax, %rdx
-; GFNI-NEXT:    andq %rcx, %rdx
-; GFNI-NEXT:    shrq %rax
-; GFNI-NEXT:    andq %rcx, %rax
-; GFNI-NEXT:    leaq (%rax,%rdx,2), %rax
+; GFNI-NEXT:    vmovq %rdi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovq %xmm0, %rax
+; GFNI-NEXT:    bswapq %rax
 ; GFNI-NEXT:    retq
   %b = call i64 @llvm.bitreverse.i64(i64 %a)
   ret i64 %b
@@ -253,24 +237,10 @@ define i32 @test_bitreverse_i32(i32 %a) nounwind {
 ;
 ; GFNI-LABEL: test_bitreverse_i32:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNI-NEXT:    bswapl %edi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
-; GFNI-NEXT:    shll $4, %eax
-; GFNI-NEXT:    shrl $4, %edi
-; GFNI-NEXT:    andl $252645135, %edi # imm = 0xF0F0F0F
-; GFNI-NEXT:    orl %eax, %edi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $858993459, %eax # imm = 0x33333333
-; GFNI-NEXT:    shrl $2, %edi
-; GFNI-NEXT:    andl $858993459, %edi # imm = 0x33333333
-; GFNI-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNI-NEXT:    movl %eax, %ecx
-; GFNI-NEXT:    andl $1431655765, %ecx # imm = 0x55555555
-; GFNI-NEXT:    shrl %eax
-; GFNI-NEXT:    andl $1431655765, %eax # imm = 0x55555555
-; GFNI-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNI-NEXT:    vmovd %edi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovd %xmm0, %eax
+; GFNI-NEXT:    bswapl %eax
 ; GFNI-NEXT:    retq
   %b = call i32 @llvm.bitreverse.i32(i32 %a)
   ret i32 %b
@@ -335,24 +305,10 @@ define i24 @test_bitreverse_i24(i24 %a) nounwind {
 ;
 ; GFNI-LABEL: test_bitreverse_i24:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNI-NEXT:    bswapl %edi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
-; GFNI-NEXT:    shll $4, %eax
-; GFNI-NEXT:    shrl $4, %edi
-; GFNI-NEXT:    andl $252645135, %edi # imm = 0xF0F0F0F
-; GFNI-NEXT:    orl %eax, %edi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $858993459, %eax # imm = 0x33333333
-; GFNI-NEXT:    shrl $2, %edi
-; GFNI-NEXT:    andl $858993459, %edi # imm = 0x33333333
-; GFNI-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNI-NEXT:    movl %eax, %ecx
-; GFNI-NEXT:    andl $1431655680, %ecx # imm = 0x55555500
-; GFNI-NEXT:    shrl %eax
-; GFNI-NEXT:    andl $1431655680, %eax # imm = 0x55555500
-; GFNI-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNI-NEXT:    vmovd %edi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovd %xmm0, %eax
+; GFNI-NEXT:    bswapl %eax
 ; GFNI-NEXT:    shrl $8, %eax
 ; GFNI-NEXT:    retq
   %b = call i24 @llvm.bitreverse.i24(i24 %a)
@@ -1412,196 +1368,67 @@ define i528 @large_promotion(i528 %A) nounwind {
 ;
 ; GFNI-LABEL: large_promotion:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    pushq %r15
 ; GFNI-NEXT:    pushq %r14
-; GFNI-NEXT:    pushq %r13
-; GFNI-NEXT:    pushq %r12
 ; GFNI-NEXT:    pushq %rbx
 ; GFNI-NEXT:    movq %rdi, %rax
-; GFNI-NEXT:    movq {{[0-9]+}}(%rsp), %r12
-; GFNI-NEXT:    movq {{[0-9]+}}(%rsp), %r15
-; GFNI-NEXT:    movq {{[0-9]+}}(%rsp), %rbx
-; GFNI-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
+; GFNI-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [9241421688590303745,9241421688590303745]
+; GFNI-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %r10
+; GFNI-NEXT:    bswapq %r10
+; GFNI-NEXT:    vmovq %r9, %xmm1
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %rdi
 ; GFNI-NEXT:    bswapq %rdi
-; GFNI-NEXT:    movq %rdi, %r10
-; GFNI-NEXT:    shrq $4, %r10
-; GFNI-NEXT:    movabsq $1085102592571150095, %r11 # imm = 0xF0F0F0F0F0F0F0F
-; GFNI-NEXT:    andq %r11, %r10
-; GFNI-NEXT:    andq %r11, %rdi
-; GFNI-NEXT:    shlq $4, %rdi
-; GFNI-NEXT:    orq %r10, %rdi
-; GFNI-NEXT:    movabsq $3689348814741910323, %r10 # imm = 0x3333333333333333
-; GFNI-NEXT:    movq %rdi, %r14
-; GFNI-NEXT:    andq %r10, %r14
-; GFNI-NEXT:    shrq $2, %rdi
-; GFNI-NEXT:    andq %r10, %rdi
-; GFNI-NEXT:    leaq (%rdi,%r14,4), %rdi
-; GFNI-NEXT:    movabsq $6148820866244280320, %r14 # imm = 0x5555000000000000
-; GFNI-NEXT:    movq %rdi, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %rdi
-; GFNI-NEXT:    andq %r14, %rdi
-; GFNI-NEXT:    leaq (%rdi,%r13,2), %rdi
-; GFNI-NEXT:    bswapq %rbx
-; GFNI-NEXT:    movq %rbx, %r14
-; GFNI-NEXT:    shrq $4, %r14
-; GFNI-NEXT:    andq %r11, %r14
-; GFNI-NEXT:    andq %r11, %rbx
-; GFNI-NEXT:    shlq $4, %rbx
-; GFNI-NEXT:    orq %r14, %rbx
-; GFNI-NEXT:    movq %rbx, %r14
-; GFNI-NEXT:    andq %r10, %r14
-; GFNI-NEXT:    shrq $2, %rbx
-; GFNI-NEXT:    andq %r10, %rbx
-; GFNI-NEXT:    leaq (%rbx,%r14,4), %rbx
-; GFNI-NEXT:    movabsq $6148914691236517205, %r14 # imm = 0x5555555555555555
-; GFNI-NEXT:    movq %rbx, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %rbx
-; GFNI-NEXT:    andq %r14, %rbx
-; GFNI-NEXT:    leaq (%rbx,%r13,2), %rbx
-; GFNI-NEXT:    shrdq $48, %rbx, %rdi
-; GFNI-NEXT:    bswapq %r15
-; GFNI-NEXT:    movq %r15, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %r15
-; GFNI-NEXT:    shlq $4, %r15
-; GFNI-NEXT:    orq %r13, %r15
-; GFNI-NEXT:    movq %r15, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %r15
-; GFNI-NEXT:    andq %r10, %r15
-; GFNI-NEXT:    leaq (%r15,%r13,4), %r15
-; GFNI-NEXT:    movq %r15, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %r15
-; GFNI-NEXT:    andq %r14, %r15
-; GFNI-NEXT:    leaq (%r15,%r13,2), %r15
-; GFNI-NEXT:    shrdq $48, %r15, %rbx
-; GFNI-NEXT:    bswapq %r12
-; GFNI-NEXT:    movq %r12, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %r12
-; GFNI-NEXT:    shlq $4, %r12
-; GFNI-NEXT:    orq %r13, %r12
-; GFNI-NEXT:    movq %r12, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %r12
-; GFNI-NEXT:    andq %r10, %r12
-; GFNI-NEXT:    leaq (%r12,%r13,4), %r12
-; GFNI-NEXT:    movq %r12, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %r12
-; GFNI-NEXT:    andq %r14, %r12
-; GFNI-NEXT:    leaq (%r12,%r13,2), %r12
-; GFNI-NEXT:    shrdq $48, %r12, %r15
-; GFNI-NEXT:    bswapq %r9
-; GFNI-NEXT:    movq %r9, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %r9
-; GFNI-NEXT:    shlq $4, %r9
-; GFNI-NEXT:    orq %r13, %r9
-; GFNI-NEXT:    movq %r9, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %r9
-; GFNI-NEXT:    andq %r10, %r9
-; GFNI-NEXT:    leaq (%r9,%r13,4), %r9
-; GFNI-NEXT:    movq %r9, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %r9
-; GFNI-NEXT:    andq %r14, %r9
-; GFNI-NEXT:    leaq (%r9,%r13,2), %r9
-; GFNI-NEXT:    shrdq $48, %r9, %r12
+; GFNI-NEXT:    vmovq %r8, %xmm1
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %r8
 ; GFNI-NEXT:    bswapq %r8
-; GFNI-NEXT:    movq %r8, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %r8
-; GFNI-NEXT:    shlq $4, %r8
-; GFNI-NEXT:    orq %r13, %r8
-; GFNI-NEXT:    movq %r8, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %r8
-; GFNI-NEXT:    andq %r10, %r8
-; GFNI-NEXT:    leaq (%r8,%r13,4), %r8
-; GFNI-NEXT:    movq %r8, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %r8
-; GFNI-NEXT:    andq %r14, %r8
-; GFNI-NEXT:    leaq (%r8,%r13,2), %r8
-; GFNI-NEXT:    shrdq $48, %r8, %r9
+; GFNI-NEXT:    movq %r8, %r9
+; GFNI-NEXT:    shldq $16, %rdi, %r9
+; GFNI-NEXT:    shldq $16, %r10, %rdi
+; GFNI-NEXT:    vmovq %rcx, %xmm1
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %rcx
 ; GFNI-NEXT:    bswapq %rcx
-; GFNI-NEXT:    movq %rcx, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %rcx
-; GFNI-NEXT:    shlq $4, %rcx
-; GFNI-NEXT:    orq %r13, %rcx
-; GFNI-NEXT:    movq %rcx, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %rcx
-; GFNI-NEXT:    andq %r10, %rcx
-; GFNI-NEXT:    leaq (%rcx,%r13,4), %rcx
-; GFNI-NEXT:    movq %rcx, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %rcx
-; GFNI-NEXT:    andq %r14, %rcx
-; GFNI-NEXT:    leaq (%rcx,%r13,2), %rcx
 ; GFNI-NEXT:    shrdq $48, %rcx, %r8
+; GFNI-NEXT:    vmovq %rdx, %xmm1
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %rdx
 ; GFNI-NEXT:    bswapq %rdx
-; GFNI-NEXT:    movq %rdx, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %rdx
-; GFNI-NEXT:    shlq $4, %rdx
-; GFNI-NEXT:    orq %r13, %rdx
-; GFNI-NEXT:    movq %rdx, %r13
-; GFNI-NEXT:    andq %r10, %r13
-; GFNI-NEXT:    shrq $2, %rdx
-; GFNI-NEXT:    andq %r10, %rdx
-; GFNI-NEXT:    leaq (%rdx,%r13,4), %rdx
-; GFNI-NEXT:    movq %rdx, %r13
-; GFNI-NEXT:    andq %r14, %r13
-; GFNI-NEXT:    shrq %rdx
-; GFNI-NEXT:    andq %r14, %rdx
-; GFNI-NEXT:    leaq (%rdx,%r13,2), %rdx
 ; GFNI-NEXT:    shrdq $48, %rdx, %rcx
+; GFNI-NEXT:    vmovq %rsi, %xmm1
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %rsi
 ; GFNI-NEXT:    bswapq %rsi
-; GFNI-NEXT:    movq %rsi, %r13
-; GFNI-NEXT:    shrq $4, %r13
-; GFNI-NEXT:    andq %r11, %r13
-; GFNI-NEXT:    andq %r11, %rsi
-; GFNI-NEXT:    shlq $4, %rsi
-; GFNI-NEXT:    orq %r13, %rsi
-; GFNI-NEXT:    movq %rsi, %r11
-; GFNI-NEXT:    andq %r10, %r11
-; GFNI-NEXT:    shrq $2, %rsi
-; GFNI-NEXT:    andq %r10, %rsi
-; GFNI-NEXT:    leaq (%rsi,%r11,4), %rsi
-; GFNI-NEXT:    movq %rsi, %r10
-; GFNI-NEXT:    andq %r14, %r10
-; GFNI-NEXT:    shrq %rsi
-; GFNI-NEXT:    andq %r14, %rsi
-; GFNI-NEXT:    leaq (%rsi,%r10,2), %rsi
 ; GFNI-NEXT:    shrdq $48, %rsi, %rdx
+; GFNI-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %r11
+; GFNI-NEXT:    bswapq %r11
+; GFNI-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm1
+; GFNI-NEXT:    vmovq %xmm1, %rbx
+; GFNI-NEXT:    bswapq %rbx
+; GFNI-NEXT:    shrdq $48, %rbx, %r11
+; GFNI-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; GFNI-NEXT:    vgf2p8affineqb $0, %xmm0, %xmm1, %xmm0
+; GFNI-NEXT:    vmovq %xmm0, %r14
+; GFNI-NEXT:    bswapq %r14
+; GFNI-NEXT:    shrdq $48, %r14, %rbx
+; GFNI-NEXT:    shrdq $48, %r10, %r14
 ; GFNI-NEXT:    shrq $48, %rsi
+; GFNI-NEXT:    movq %r14, 16(%rax)
+; GFNI-NEXT:    movq %rbx, 8(%rax)
+; GFNI-NEXT:    movq %r11, (%rax)
 ; GFNI-NEXT:    movq %rdx, 56(%rax)
 ; GFNI-NEXT:    movq %rcx, 48(%rax)
 ; GFNI-NEXT:    movq %r8, 40(%rax)
 ; GFNI-NEXT:    movq %r9, 32(%rax)
-; GFNI-NEXT:    movq %r12, 24(%rax)
-; GFNI-NEXT:    movq %r15, 16(%rax)
-; GFNI-NEXT:    movq %rbx, 8(%rax)
-; GFNI-NEXT:    movq %rdi, (%rax)
+; GFNI-NEXT:    movq %rdi, 24(%rax)
 ; GFNI-NEXT:    movw %si, 64(%rax)
 ; GFNI-NEXT:    popq %rbx
-; GFNI-NEXT:    popq %r12
-; GFNI-NEXT:    popq %r13
 ; GFNI-NEXT:    popq %r14
-; GFNI-NEXT:    popq %r15
 ; GFNI-NEXT:    retq
   %Z = call i528 @llvm.bitreverse.i528(i528 %A)
   ret i528 %Z
diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index d3f357cd17952..1c5326d35bb00 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -276,24 +276,10 @@ define i32 @test_bitreverse_i32(i32 %a) nounwind {
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i32:
 ; GFNIAVX:       # %bb.0:
-; GFNIAVX-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNIAVX-NEXT:    bswapl %edi
-; GFNIAVX-NEXT:    movl %edi, %eax
-; GFNIAVX-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
-; GFNIAVX-NEXT:    shll $4, %eax
-; GFNIAVX-NEXT:    shrl $4, %edi
-; GFNIAVX-NEXT:    andl $252645135, %edi # imm = 0xF0F0F0F
-; GFNIAVX-NEXT:    orl %eax, %edi
-; GFNIAVX-NEXT:    movl %edi, %eax
-; GFNIAVX-NEXT:    andl $858993459, %eax # imm = 0x33333333
-; GFNIAVX-NEXT:    shrl $2, %edi
-; GFNIAVX-NEXT:    andl $858993459, %edi # imm = 0x33333333
-; GFNIAVX-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNIAVX-NEXT:    movl %eax, %ecx
-; GFNIAVX-NEXT:    andl $1431655765, %ecx # imm = 0x55555555
-; GFNIAVX-NEXT:    shrl %eax
-; GFNIAVX-NEXT:    andl $1431655765, %eax # imm = 0x55555555
-; GFNIAVX-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNIAVX-NEXT:    vmovd %edi, %xmm0
+; GFNIAVX-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; GFNIAVX-NEXT:    vmovd %xmm0, %eax
+; GFNIAVX-NEXT:    bswapl %eax
 ; GFNIAVX-NEXT:    retq
   %b = call i32 @llvm.bitreverse.i32(i32 %a)
   ret i32 %b
@@ -381,26 +367,10 @@ define i64 @test_bitreverse_i64(i64 %a) nounwind {
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i64:
 ; GFNIAVX:       # %bb.0:
-; GFNIAVX-NEXT:    bswapq %rdi
-; GFNIAVX-NEXT:    movq %rdi, %rax
-; GFNIAVX-NEXT:    shrq $4, %rax
-; GFNIAVX-NEXT:    movabsq $1085102592571150095, %rcx # imm = 0xF0F0F0F0F0F0F0F
-; GFNIAVX-NEXT:    andq %rcx, %rax
-; GFNIAVX-NEXT:    andq %rcx, %rdi
-; GFNIAVX-NEXT:    shlq $4, %rdi
-; GFNIAVX-NEXT:    orq %rax, %rdi
-; GFNIAVX-NEXT:    movabsq $3689348814741910323, %rax # imm = 0x3333333333333333
-; GFNIAVX-NEXT:    movq %rdi, %rcx
-; GFNIAVX-NEXT:    andq %rax, %rcx
-; GFNIAVX-NEXT:    shrq $2, %rdi
-; GFNIAVX-NEXT:    andq %rax, %rdi
-; GFNIAVX-NEXT:    leaq (%rdi,%rcx,4), %rax
-; GFNIAVX-NEXT:    movabsq $6148914691236517205, %rcx # imm = 0x5555555555555555
-; GFNIAVX-NEXT:    movq %rax, %rdx
-; GFNIAVX-NEXT:    andq %rcx, %rdx
-; GFNIAVX-NEXT:    shrq %rax
-; GFNIAVX-NEXT:    andq %rcx, %rax
-; GFNIAVX-NEXT:    leaq (%rax,%rdx,2), %rax
+; GFNIAVX-NEXT:    vmovq %rdi, %xmm0
+; GFNIAVX-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; GFNIAVX-NEXT:    vmovq %xmm0, %rax
+; GFNIAVX-NEXT:    bswapq %rax
 ; GFNIAVX-NEXT:    retq
   %b = call i64 @llvm.bitreverse.i64(i64 %a)
   ret i64 %b

From 7549b45825a05fc24fcdbacf006461165aa042cb Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Wed, 10 Apr 2024 21:27:10 +0200
Subject: [PATCH 062/886] Revert "[clang][dataflow] Propagate locations from
 result objects to initializers." (#88315)

Reverts llvm/llvm-project#87320

This is causing buildbots to fail because
`isOriginalRecordConstructor()` is now unused.
---
 .../FlowSensitive/DataflowEnvironment.h       |  64 +--
 .../FlowSensitive/DataflowEnvironment.cpp     | 405 +++++-------------
 clang/lib/Analysis/FlowSensitive/Transfer.cpp | 176 ++++----
 .../TypeErasedDataflowAnalysis.cpp            |  13 +-
 .../FlowSensitive/DataflowEnvironmentTest.cpp |  43 --
 .../Analysis/FlowSensitive/TransferTest.cpp   | 172 +++-----
 6 files changed, 283 insertions(+), 590 deletions(-)

diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
index 706664d7db1c2..9a65f76cdf56b 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
@@ -30,7 +30,6 @@
 #include "llvm/ADT/MapVector.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/ErrorHandling.h"
-#include <memory>
 #include <type_traits>
 #include <utility>
 
@@ -345,6 +344,17 @@ class Environment {
   /// location of the result object to pass in `this`, even though prvalues are
   /// otherwise not associated with storage locations.
   ///
+  /// FIXME: Currently, this simply returns a stable storage location for `E`,
+  /// but this doesn't do the right thing in scenarios like the following:
+  /// ```
+  /// MyClass c = some_condition()? MyClass(foo) : MyClass(bar);
+  /// ```
+  /// Here, `MyClass(foo)` and `MyClass(bar)` will have two different storage
+  /// locations, when in fact their storage locations should be the same.
+  /// Eventually, we want to propagate storage locations from result objects
+  /// down to the prvalues that initialize them, similar to the way that this is
+  /// done in Clang's CodeGen.
+  ///
   /// Requirements:
   ///  `E` must be a prvalue of record type.
   RecordStorageLocation &
@@ -452,13 +462,7 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values.
-  /// If `Type` is provided, initializes only those fields that are modeled for
-  /// `Type`; this is intended for use in cases where `Loc` is a derived type
-  /// and we only want to initialize the fields of a base type.
-  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type);
-  void initializeFieldsWithValues(RecordStorageLocation &Loc) {
-    initializeFieldsWithValues(Loc, Loc.getType());
-  }
+  void initializeFieldsWithValues(RecordStorageLocation &Loc);
 
   /// Assigns `Val` as the value of `Loc` in the environment.
   void setValue(const StorageLocation &Loc, Value &Val);
@@ -649,9 +653,6 @@ class Environment {
   LLVM_DUMP_METHOD void dump(raw_ostream &OS) const;
 
 private:
-  using PrValueToResultObject =
-      llvm::DenseMap<const Expr *, RecordStorageLocation *>;
-
   // The copy-constructor is for use in fork() only.
   Environment(const Environment &) = default;
 
@@ -681,10 +682,8 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values (controlled by `Visited`,
-  /// `Depth`, and `CreatedValuesCount`). If `Type` is different from
-  /// `Loc.getType()`, initializes only those fields that are modeled for
-  /// `Type`.
-  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type,
+  /// `Depth`, and `CreatedValuesCount`).
+  void initializeFieldsWithValues(RecordStorageLocation &Loc,
                                   llvm::DenseSet<QualType> &Visited, int Depth,
                                   int &CreatedValuesCount);
 
@@ -703,45 +702,22 @@ class Environment {
   /// and functions referenced in `FuncDecl`. `FuncDecl` must have a body.
   void initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl);
 
-  static PrValueToResultObject
-  buildResultObjectMap(DataflowAnalysisContext *DACtx,
-                       const FunctionDecl *FuncDecl,
-                       RecordStorageLocation *ThisPointeeLoc,
-                       RecordStorageLocation *LocForRecordReturnVal);
-
   // `DACtx` is not null and not owned by this object.
   DataflowAnalysisContext *DACtx;
 
-  // FIXME: move the fields `CallStack`, `ResultObjectMap`, `ReturnVal`,
-  // `ReturnLoc` and `ThisPointeeLoc` into a separate call-context object,
-  // shared between environments in the same call.
+  // FIXME: move the fields `CallStack`, `ReturnVal`, `ReturnLoc` and
+  // `ThisPointeeLoc` into a separate call-context object, shared between
+  // environments in the same call.
   // https://github.com/llvm/llvm-project/issues/59005
 
   // `DeclContext` of the block being analysed if provided.
   std::vector<const DeclContext *> CallStack;
 
-  // Maps from prvalues of record type to their result objects. Shared between
-  // all environments for the same function.
-  // FIXME: It's somewhat unsatisfactory that we have to use a `shared_ptr`
-  // here, though the cost is acceptable: The overhead of a `shared_ptr` is
-  // incurred when it is copied, and this happens only relatively rarely (when
-  // we fork the environment). The need for a `shared_ptr` will go away once we
-  // introduce a shared call-context object (see above).
-  std::shared_ptr<PrValueToResultObject> ResultObjectMap;
-
-  // The following three member variables handle various different types of
-  // return values.
-  // - If the return type is not a reference and not a record: Value returned
-  //   by the function.
+  // Value returned by the function (if it has non-reference return type).
   Value *ReturnVal = nullptr;
-  // - If the return type is a reference: Storage location of the reference
-  //   returned by the function.
+  // Storage location of the reference returned by the function (if it has
+  // reference return type).
   StorageLocation *ReturnLoc = nullptr;
-  // - If the return type is a record or the function being analyzed is a
-  //   constructor: Storage location into which the return value should be
-  //   constructed.
-  RecordStorageLocation *LocForRecordReturnVal = nullptr;
-
   // The storage location of the `this` pointee. Should only be null if the
   // function being analyzed is only a function and not a method.
   RecordStorageLocation *ThisPointeeLoc = nullptr;
diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index 6c796b4ad923e..1bfa7ebcfd50c 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -15,7 +15,6 @@
 #include "clang/Analysis/FlowSensitive/DataflowEnvironment.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/DeclCXX.h"
-#include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/AST/Type.h"
 #include "clang/Analysis/FlowSensitive/DataflowLattice.h"
 #include "clang/Analysis/FlowSensitive/Value.h"
@@ -27,8 +26,6 @@
 #include <cassert>
 #include <utility>
 
-#define DEBUG_TYPE "dataflow"
-
 namespace clang {
 namespace dataflow {
 
@@ -357,8 +354,6 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   for (auto *Child : S.children())
     if (Child != nullptr)
       getFieldsGlobalsAndFuncs(*Child, Fields, Vars, Funcs);
-  if (const auto *DefaultArg = dyn_cast<CXXDefaultArgExpr>(&S))
-    getFieldsGlobalsAndFuncs(*DefaultArg->getExpr(), Fields, Vars, Funcs);
   if (const auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(&S))
     getFieldsGlobalsAndFuncs(*DefaultInit->getExpr(), Fields, Vars, Funcs);
 
@@ -391,186 +386,6 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   }
 }
 
-namespace {
-
-// Visitor that builds a map from record prvalues to result objects.
-// This traverses the body of the function to be analyzed; for each result
-// object that it encounters, it propagates the storage location of the result
-// object to all record prvalues that can initialize it.
-class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
-public:
-  // `ResultObjectMap` will be filled with a map from record prvalues to result
-  // object. If the function being analyzed returns a record by value,
-  // `LocForRecordReturnVal` is the location to which this record should be
-  // written; otherwise, it is null.
-  explicit ResultObjectVisitor(
-      llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap,
-      RecordStorageLocation *LocForRecordReturnVal,
-      DataflowAnalysisContext &DACtx)
-      : ResultObjectMap(ResultObjectMap),
-        LocForRecordReturnVal(LocForRecordReturnVal), DACtx(DACtx) {}
-
-  bool shouldVisitImplicitCode() { return true; }
-
-  bool shouldVisitLambdaBody() const { return false; }
-
-  // Traverse all member and base initializers of `Ctor`. This function is not
-  // called by `RecursiveASTVisitor`; it should be called manually if we are
-  // analyzing a constructor. `ThisPointeeLoc` is the storage location that
-  // `this` points to.
-  void TraverseConstructorInits(const CXXConstructorDecl *Ctor,
-                                RecordStorageLocation *ThisPointeeLoc) {
-    assert(ThisPointeeLoc != nullptr);
-    for (const CXXCtorInitializer *Init : Ctor->inits()) {
-      Expr *InitExpr = Init->getInit();
-      if (FieldDecl *Field = Init->getMember();
-          Field != nullptr && Field->getType()->isRecordType()) {
-        PropagateResultObject(InitExpr, cast<RecordStorageLocation>(
-                                            ThisPointeeLoc->getChild(*Field)));
-      } else if (Init->getBaseClass()) {
-        PropagateResultObject(InitExpr, ThisPointeeLoc);
-      }
-
-      // Ensure that any result objects within `InitExpr` (e.g. temporaries)
-      // are also propagated to the prvalues that initialize them.
-      TraverseStmt(InitExpr);
-
-      // If this is a `CXXDefaultInitExpr`, also propagate any result objects
-      // within the default expression.
-      if (auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(InitExpr))
-        TraverseStmt(DefaultInit->getExpr());
-    }
-  }
-
-  bool TraverseBindingDecl(BindingDecl *BD) {
-    // `RecursiveASTVisitor` doesn't traverse holding variables for
-    // `BindingDecl`s by itself, so we need to tell it to.
-    if (VarDecl *HoldingVar = BD->getHoldingVar())
-      TraverseDecl(HoldingVar);
-    return RecursiveASTVisitor<ResultObjectVisitor>::TraverseBindingDecl(BD);
-  }
-
-  bool VisitVarDecl(VarDecl *VD) {
-    if (VD->getType()->isRecordType() && VD->hasInit())
-      PropagateResultObject(
-          VD->getInit(),
-          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*VD)));
-    return true;
-  }
-
-  bool VisitMaterializeTemporaryExpr(MaterializeTemporaryExpr *MTE) {
-    if (MTE->getType()->isRecordType())
-      PropagateResultObject(
-          MTE->getSubExpr(),
-          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*MTE)));
-    return true;
-  }
-
-  bool VisitReturnStmt(ReturnStmt *Return) {
-    Expr *RetValue = Return->getRetValue();
-    if (RetValue != nullptr && RetValue->getType()->isRecordType() &&
-        RetValue->isPRValue())
-      PropagateResultObject(RetValue, LocForRecordReturnVal);
-    return true;
-  }
-
-  bool VisitExpr(Expr *E) {
-    // Clang's AST can have record-type prvalues without a result object -- for
-    // example as full-expressions contained in a compound statement or as
-    // arguments of call expressions. We notice this if we get here and a
-    // storage location has not yet been associated with `E`. In this case,
-    // treat this as if it was a `MaterializeTemporaryExpr`.
-    if (E->isPRValue() && E->getType()->isRecordType() &&
-        !ResultObjectMap.contains(E))
-      PropagateResultObject(
-          E, &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*E)));
-    return true;
-  }
-
-  // Assigns `Loc` as the result object location of `E`, then propagates the
-  // location to all lower-level prvalues that initialize the same object as
-  // `E` (or one of its base classes or member variables).
-  void PropagateResultObject(Expr *E, RecordStorageLocation *Loc) {
-    if (!E->isPRValue() || !E->getType()->isRecordType()) {
-      assert(false);
-      // Ensure we don't propagate the result object if we hit this in a
-      // release build.
-      return;
-    }
-
-    ResultObjectMap[E] = Loc;
-
-    // The following AST node kinds are "original initializers": They are the
-    // lowest-level AST node that initializes a given object, and nothing
-    // below them can initialize the same object (or part of it).
-    if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
-        isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
-        isa<CXXStdInitializerListExpr>(E)) {
-      return;
-    }
-
-    if (auto *InitList = dyn_cast<InitListExpr>(E)) {
-      if (!InitList->isSemanticForm())
-        return;
-      if (InitList->isTransparent()) {
-        PropagateResultObject(InitList->getInit(0), Loc);
-        return;
-      }
-
-      RecordInitListHelper InitListHelper(InitList);
-
-      for (auto [Base, Init] : InitListHelper.base_inits()) {
-        assert(Base->getType().getCanonicalType() ==
-               Init->getType().getCanonicalType());
-
-        // Storage location for the base class is the same as that of the
-        // derived class because we "flatten" the object hierarchy and put all
-        // fields in `RecordStorageLocation` of the derived class.
-        PropagateResultObject(Init, Loc);
-      }
-
-      for (auto [Field, Init] : InitListHelper.field_inits()) {
-        // Fields of non-record type are handled in
-        // `TransferVisitor::VisitInitListExpr()`.
-        if (!Field->getType()->isRecordType())
-          continue;
-        PropagateResultObject(
-            Init, cast<RecordStorageLocation>(Loc->getChild(*Field)));
-      }
-      return;
-    }
-
-    if (auto *Op = dyn_cast<BinaryOperator>(E); Op && Op->isCommaOp()) {
-      PropagateResultObject(Op->getRHS(), Loc);
-      return;
-    }
-
-    if (auto *Cond = dyn_cast<AbstractConditionalOperator>(E)) {
-      PropagateResultObject(Cond->getTrueExpr(), Loc);
-      PropagateResultObject(Cond->getFalseExpr(), Loc);
-      return;
-    }
-
-    // All other expression nodes that propagate a record prvalue should have
-    // exactly one child.
-    SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
-    LLVM_DEBUG({
-      if (Children.size() != 1)
-        E->dump();
-    });
-    assert(Children.size() == 1);
-    for (Stmt *S : Children)
-      PropagateResultObject(cast<Expr>(S), Loc);
-  }
-
-private:
-  llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap;
-  RecordStorageLocation *LocForRecordReturnVal;
-  DataflowAnalysisContext &DACtx;
-};
-
-} // namespace
-
 Environment::Environment(DataflowAnalysisContext &DACtx)
     : DACtx(&DACtx),
       FlowConditionToken(DACtx.arena().makeFlowConditionToken()) {}
@@ -586,23 +401,17 @@ void Environment::initialize() {
   if (DeclCtx == nullptr)
     return;
 
-  const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx);
-  if (FuncDecl == nullptr)
-    return;
-
-  assert(FuncDecl->doesThisDeclarationHaveABody());
+  if (const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx)) {
+    assert(FuncDecl->doesThisDeclarationHaveABody());
 
-  initFieldsGlobalsAndFuncs(FuncDecl);
+    initFieldsGlobalsAndFuncs(FuncDecl);
 
-  for (const auto *ParamDecl : FuncDecl->parameters()) {
-    assert(ParamDecl != nullptr);
-    setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
+    for (const auto *ParamDecl : FuncDecl->parameters()) {
+      assert(ParamDecl != nullptr);
+      setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
+    }
   }
 
-  if (FuncDecl->getReturnType()->isRecordType())
-    LocForRecordReturnVal = &cast<RecordStorageLocation>(
-        createStorageLocation(FuncDecl->getReturnType()));
-
   if (const auto *MethodDecl = dyn_cast<CXXMethodDecl>(DeclCtx)) {
     auto *Parent = MethodDecl->getParent();
     assert(Parent != nullptr);
@@ -635,12 +444,6 @@ void Environment::initialize() {
         initializeFieldsWithValues(ThisLoc);
     }
   }
-
-  // We do this below the handling of `CXXMethodDecl` above so that we can
-  // be sure that the storage location for `this` has been set.
-  ResultObjectMap = std::make_shared<PrValueToResultObject>(
-      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
-                           LocForRecordReturnVal));
 }
 
 // FIXME: Add support for resetting globals after function calls to enable
@@ -681,18 +484,13 @@ void Environment::initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl) {
     if (getStorageLocation(*D) != nullptr)
       continue;
 
-    // We don't run transfer functions on the initializers of global variables,
-    // so they won't be associated with a value or storage location. We
-    // therefore intentionally don't pass an initializer to `createObject()`;
-    // in particular, this ensures that `createObject()` will initialize the
-    // fields of record-type variables with values.
-    setStorageLocation(*D, createObject(*D, nullptr));
+    setStorageLocation(*D, createObject(*D));
   }
 
   for (const FunctionDecl *FD : Funcs) {
     if (getStorageLocation(*FD) != nullptr)
       continue;
-    auto &Loc = createStorageLocation(*FD);
+    auto &Loc = createStorageLocation(FD->getType());
     setStorageLocation(*FD, Loc);
   }
 }
@@ -721,9 +519,6 @@ Environment Environment::pushCall(const CallExpr *Call) const {
     }
   }
 
-  if (Call->getType()->isRecordType() && Call->isPRValue())
-    Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
-
   Env.pushCallInternal(Call->getDirectCallee(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
 
@@ -734,7 +529,6 @@ Environment Environment::pushCall(const CXXConstructExpr *Call) const {
   Environment Env(*this);
 
   Env.ThisPointeeLoc = &Env.getResultObjectLocation(*Call);
-  Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
 
   Env.pushCallInternal(Call->getConstructor(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
@@ -763,10 +557,6 @@ void Environment::pushCallInternal(const FunctionDecl *FuncDecl,
     const VarDecl *Param = *ParamIt;
     setStorageLocation(*Param, createObject(*Param, Args[ArgIndex]));
   }
-
-  ResultObjectMap = std::make_shared<PrValueToResultObject>(
-      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
-                           LocForRecordReturnVal));
 }
 
 void Environment::popCall(const CallExpr *Call, const Environment &CalleeEnv) {
@@ -810,9 +600,6 @@ bool Environment::equivalentTo(const Environment &Other,
   if (ReturnLoc != Other.ReturnLoc)
     return false;
 
-  if (LocForRecordReturnVal != Other.LocForRecordReturnVal)
-    return false;
-
   if (ThisPointeeLoc != Other.ThisPointeeLoc)
     return false;
 
@@ -836,10 +623,8 @@ LatticeEffect Environment::widen(const Environment &PrevEnv,
   assert(DACtx == PrevEnv.DACtx);
   assert(ReturnVal == PrevEnv.ReturnVal);
   assert(ReturnLoc == PrevEnv.ReturnLoc);
-  assert(LocForRecordReturnVal == PrevEnv.LocForRecordReturnVal);
   assert(ThisPointeeLoc == PrevEnv.ThisPointeeLoc);
   assert(CallStack == PrevEnv.CallStack);
-  assert(ResultObjectMap == PrevEnv.ResultObjectMap);
 
   auto Effect = LatticeEffect::Unchanged;
 
@@ -871,16 +656,12 @@ Environment Environment::join(const Environment &EnvA, const Environment &EnvB,
                               Environment::ValueModel &Model,
                               ExprJoinBehavior ExprBehavior) {
   assert(EnvA.DACtx == EnvB.DACtx);
-  assert(EnvA.LocForRecordReturnVal == EnvB.LocForRecordReturnVal);
   assert(EnvA.ThisPointeeLoc == EnvB.ThisPointeeLoc);
   assert(EnvA.CallStack == EnvB.CallStack);
-  assert(EnvA.ResultObjectMap == EnvB.ResultObjectMap);
 
   Environment JoinedEnv(*EnvA.DACtx);
 
   JoinedEnv.CallStack = EnvA.CallStack;
-  JoinedEnv.ResultObjectMap = EnvA.ResultObjectMap;
-  JoinedEnv.LocForRecordReturnVal = EnvA.LocForRecordReturnVal;
   JoinedEnv.ThisPointeeLoc = EnvA.ThisPointeeLoc;
 
   if (EnvA.ReturnVal == nullptr || EnvB.ReturnVal == nullptr) {
@@ -949,12 +730,6 @@ StorageLocation &Environment::createStorageLocation(const Expr &E) {
 
 void Environment::setStorageLocation(const ValueDecl &D, StorageLocation &Loc) {
   assert(!DeclToLoc.contains(&D));
-  // The only kinds of declarations that may have a "variable" storage location
-  // are declarations of reference type and `BindingDecl`. For all other
-  // declaration, the storage location should be the stable storage location
-  // returned by `createStorageLocation()`.
-  assert(D.getType()->isReferenceType() || isa<BindingDecl>(D) ||
-         &Loc == &createStorageLocation(D));
   DeclToLoc[&D] = &Loc;
 }
 
@@ -1016,29 +791,50 @@ Environment::getResultObjectLocation(const Expr &RecordPRValue) const {
   assert(RecordPRValue.getType()->isRecordType());
   assert(RecordPRValue.isPRValue());
 
-  assert(ResultObjectMap != nullptr);
-  RecordStorageLocation *Loc = ResultObjectMap->lookup(&RecordPRValue);
-  assert(Loc != nullptr);
-  // In release builds, use the "stable" storage location if the map lookup
-  // failed.
-  if (Loc == nullptr)
+  // Returns a storage location that we can use if assertions fail.
+  auto FallbackForAssertFailure =
+      [this, &RecordPRValue]() -> RecordStorageLocation & {
     return cast<RecordStorageLocation>(
         DACtx->getStableStorageLocation(RecordPRValue));
-  return *Loc;
+  };
+
+  if (isOriginalRecordConstructor(RecordPRValue)) {
+    auto *Val = cast_or_null<RecordValue>(getValue(RecordPRValue));
+    // The builtin transfer function should have created a `RecordValue` for all
+    // original record constructors.
+    assert(Val);
+    if (!Val)
+      return FallbackForAssertFailure();
+    return Val->getLoc();
+  }
+
+  if (auto *Op = dyn_cast<BinaryOperator>(&RecordPRValue);
+      Op && Op->isCommaOp()) {
+    return getResultObjectLocation(*Op->getRHS());
+  }
+
+  // All other expression nodes that propagate a record prvalue should have
+  // exactly one child.
+  llvm::SmallVector<const Stmt *> children(RecordPRValue.child_begin(),
+                                           RecordPRValue.child_end());
+  assert(children.size() == 1);
+  if (children.empty())
+    return FallbackForAssertFailure();
+
+  return getResultObjectLocation(*cast<Expr>(children[0]));
 }
 
 PointerValue &Environment::getOrCreateNullPointerValue(QualType PointeeType) {
   return DACtx->getOrCreateNullPointerValue(PointeeType);
 }
 
-void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
-                                             QualType Type) {
+void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc) {
   llvm::DenseSet<QualType> Visited;
   int CreatedValuesCount = 0;
-  initializeFieldsWithValues(Loc, Type, Visited, 0, CreatedValuesCount);
+  initializeFieldsWithValues(Loc, Visited, 0, CreatedValuesCount);
   if (CreatedValuesCount > MaxCompositeValueSize) {
-    llvm::errs() << "Attempting to initialize a huge value of type: " << Type
-                 << '\n';
+    llvm::errs() << "Attempting to initialize a huge value of type: "
+                 << Loc.getType() << '\n';
   }
 }
 
@@ -1052,7 +848,8 @@ void Environment::setValue(const Expr &E, Value &Val) {
   const Expr &CanonE = ignoreCFGOmittedNodes(E);
 
   if (auto *RecordVal = dyn_cast<RecordValue>(&Val)) {
-    assert(&RecordVal->getLoc() == &getResultObjectLocation(CanonE));
+    assert(isOriginalRecordConstructor(CanonE) ||
+           &RecordVal->getLoc() == &getResultObjectLocation(CanonE));
     (void)RecordVal;
   }
 
@@ -1131,8 +928,7 @@ Value *Environment::createValueUnlessSelfReferential(
   if (Type->isRecordType()) {
     CreatedValuesCount++;
     auto &Loc = cast<RecordStorageLocation>(createStorageLocation(Type));
-    initializeFieldsWithValues(Loc, Loc.getType(), Visited, Depth,
-                               CreatedValuesCount);
+    initializeFieldsWithValues(Loc, Visited, Depth, CreatedValuesCount);
 
     return &refreshRecordValue(Loc, *this);
   }
@@ -1164,7 +960,6 @@ Environment::createLocAndMaybeValue(QualType Ty,
 }
 
 void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
-                                             QualType Type,
                                              llvm::DenseSet<QualType> &Visited,
                                              int Depth,
                                              int &CreatedValuesCount) {
@@ -1172,8 +967,8 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     if (FieldType->isRecordType()) {
       auto &FieldRecordLoc = cast<RecordStorageLocation>(FieldLoc);
       setValue(FieldRecordLoc, create<RecordValue>(FieldRecordLoc));
-      initializeFieldsWithValues(FieldRecordLoc, FieldRecordLoc.getType(),
-                                 Visited, Depth + 1, CreatedValuesCount);
+      initializeFieldsWithValues(FieldRecordLoc, Visited, Depth + 1,
+                                 CreatedValuesCount);
     } else {
       if (!Visited.insert(FieldType.getCanonicalType()).second)
         return;
@@ -1184,7 +979,7 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     }
   };
 
-  for (const FieldDecl *Field : DACtx->getModeledFields(Type)) {
+  for (const auto &[Field, FieldLoc] : Loc.children()) {
     assert(Field != nullptr);
     QualType FieldType = Field->getType();
 
@@ -1193,12 +988,14 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
                    &createLocAndMaybeValue(FieldType, Visited, Depth + 1,
                                            CreatedValuesCount));
     } else {
-      StorageLocation *FieldLoc = Loc.getChild(*Field);
       assert(FieldLoc != nullptr);
       initField(FieldType, *FieldLoc);
     }
   }
-  for (const auto &[FieldName, FieldType] : DACtx->getSyntheticFields(Type)) {
+  for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
+    assert(FieldLoc != nullptr);
+    QualType FieldType = FieldLoc->getType();
+
     // Synthetic fields cannot have reference type, so we don't need to deal
     // with this case.
     assert(!FieldType->isReferenceType());
@@ -1225,36 +1022,38 @@ StorageLocation &Environment::createObjectInternal(const ValueDecl *D,
     return createObjectInternal(D, Ty.getNonReferenceType(), nullptr);
   }
 
+  Value *Val = nullptr;
+  if (InitExpr) {
+    // In the (few) cases where an expression is intentionally
+    // "uninterpreted", `InitExpr` is not associated with a value.  There are
+    // two ways to handle this situation: propagate the status, so that
+    // uninterpreted initializers result in uninterpreted variables, or
+    // provide a default value. We choose the latter so that later refinements
+    // of the variable can be used for reasoning about the surrounding code.
+    // For this reason, we let this case be handled by the `createValue()`
+    // call below.
+    //
+    // FIXME. If and when we interpret all language cases, change this to
+    // assert that `InitExpr` is interpreted, rather than supplying a
+    // default value (assuming we don't update the environment API to return
+    // references).
+    Val = getValue(*InitExpr);
+
+    if (!Val && isa<ImplicitValueInitExpr>(InitExpr) &&
+        InitExpr->getType()->isPointerType())
+      Val = &getOrCreateNullPointerValue(InitExpr->getType()->getPointeeType());
+  }
+  if (!Val)
+    Val = createValue(Ty);
+
+  if (Ty->isRecordType())
+    return cast<RecordValue>(Val)->getLoc();
+
   StorageLocation &Loc =
       D ? createStorageLocation(*D) : createStorageLocation(Ty);
 
-  if (Ty->isRecordType()) {
-    auto &RecordLoc = cast<RecordStorageLocation>(Loc);
-    if (!InitExpr)
-      initializeFieldsWithValues(RecordLoc);
-    refreshRecordValue(RecordLoc, *this);
-  } else {
-    Value *Val = nullptr;
-    if (InitExpr)
-      // In the (few) cases where an expression is intentionally
-      // "uninterpreted", `InitExpr` is not associated with a value.  There are
-      // two ways to handle this situation: propagate the status, so that
-      // uninterpreted initializers result in uninterpreted variables, or
-      // provide a default value. We choose the latter so that later refinements
-      // of the variable can be used for reasoning about the surrounding code.
-      // For this reason, we let this case be handled by the `createValue()`
-      // call below.
-      //
-      // FIXME. If and when we interpret all language cases, change this to
-      // assert that `InitExpr` is interpreted, rather than supplying a
-      // default value (assuming we don't update the environment API to return
-      // references).
-      Val = getValue(*InitExpr);
-    if (!Val)
-      Val = createValue(Ty);
-    if (Val)
-      setValue(Loc, *Val);
-  }
+  if (Val)
+    setValue(Loc, *Val);
 
   return Loc;
 }
@@ -1273,8 +1072,6 @@ bool Environment::allows(const Formula &F) const {
 
 void Environment::dump(raw_ostream &OS) const {
   llvm::DenseMap<const StorageLocation *, std::string> LocToName;
-  if (LocForRecordReturnVal != nullptr)
-    LocToName[LocForRecordReturnVal] = "(returned record)";
   if (ThisPointeeLoc != nullptr)
     LocToName[ThisPointeeLoc] = "this";
 
@@ -1305,9 +1102,6 @@ void Environment::dump(raw_ostream &OS) const {
       if (auto Iter = LocToName.find(ReturnLoc); Iter != LocToName.end())
         OS << " (" << Iter->second << ")";
       OS << "\n";
-    } else if (Func->getReturnType()->isRecordType() ||
-               isa<CXXConstructorDecl>(Func)) {
-      OS << "LocForRecordReturnVal: " << LocForRecordReturnVal << "\n";
     } else if (!Func->getReturnType()->isVoidType()) {
       if (ReturnVal == nullptr)
         OS << "ReturnVal: nullptr\n";
@@ -1328,22 +1122,6 @@ void Environment::dump() const {
   dump(llvm::dbgs());
 }
 
-Environment::PrValueToResultObject Environment::buildResultObjectMap(
-    DataflowAnalysisContext *DACtx, const FunctionDecl *FuncDecl,
-    RecordStorageLocation *ThisPointeeLoc,
-    RecordStorageLocation *LocForRecordReturnVal) {
-  assert(FuncDecl->doesThisDeclarationHaveABody());
-
-  PrValueToResultObject Map;
-
-  ResultObjectVisitor Visitor(Map, LocForRecordReturnVal, *DACtx);
-  if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(FuncDecl))
-    Visitor.TraverseConstructorInits(Ctor, ThisPointeeLoc);
-  Visitor.TraverseStmt(FuncDecl->getBody());
-
-  return Map;
-}
-
 RecordStorageLocation *getImplicitObjectLocation(const CXXMemberCallExpr &MCE,
                                                  const Environment &Env) {
   Expr *ImplicitObject = MCE.getImplicitObjectArgument();
@@ -1438,11 +1216,24 @@ RecordValue &refreshRecordValue(RecordStorageLocation &Loc, Environment &Env) {
 RecordValue &refreshRecordValue(const Expr &Expr, Environment &Env) {
   assert(Expr.getType()->isRecordType());
 
-  if (Expr.isPRValue())
-    refreshRecordValue(Env.getResultObjectLocation(Expr), Env);
+  if (Expr.isPRValue()) {
+    if (auto *ExistingVal = Env.get<RecordValue>(Expr)) {
+      auto &NewVal = Env.create<RecordValue>(ExistingVal->getLoc());
+      Env.setValue(Expr, NewVal);
+      Env.setValue(NewVal.getLoc(), NewVal);
+      return NewVal;
+    }
 
-  if (auto *Loc = Env.get<RecordStorageLocation>(Expr))
-    refreshRecordValue(*Loc, Env);
+    auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
+    Env.setValue(Expr, NewVal);
+    return NewVal;
+  }
+
+  if (auto *Loc = Env.get<RecordStorageLocation>(Expr)) {
+    auto &NewVal = Env.create<RecordValue>(*Loc);
+    Env.setValue(*Loc, NewVal);
+    return NewVal;
+  }
 
   auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
   Env.setStorageLocation(Expr, NewVal.getLoc());
diff --git a/clang/lib/Analysis/FlowSensitive/Transfer.cpp b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
index 88a9c0eccbebc..0a2e8368d541d 100644
--- a/clang/lib/Analysis/FlowSensitive/Transfer.cpp
+++ b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
@@ -460,9 +460,11 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // So make sure we have a value if we didn't propagate one above.
     if (S->isPRValue() && S->getType()->isRecordType()) {
       if (Env.getValue(*S) == nullptr) {
-        auto &Loc = Env.getResultObjectLocation(*S);
-        Env.initializeFieldsWithValues(Loc);
-        refreshRecordValue(Loc, Env);
+        Value *Val = Env.createValue(S->getType());
+        // We're guaranteed to always be able to create a value for record
+        // types.
+        assert(Val != nullptr);
+        Env.setValue(*S, *Val);
       }
     }
   }
@@ -470,13 +472,6 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
   void VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *S) {
     const Expr *InitExpr = S->getExpr();
     assert(InitExpr != nullptr);
-
-    // If this is a prvalue of record type, the handler for `*InitExpr` (if one
-    // exists) will initialize the result object; there is no value to propgate
-    // here.
-    if (S->getType()->isRecordType() && S->isPRValue())
-      return;
-
     propagateValueOrStorageLocation(*InitExpr, *S, Env);
   }
 
@@ -484,17 +479,6 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     const CXXConstructorDecl *ConstructorDecl = S->getConstructor();
     assert(ConstructorDecl != nullptr);
 
-    // `CXXConstructExpr` can have array type if default-initializing an array
-    // of records. We don't handle this specifically beyond potentially inlining
-    // the call.
-    if (!S->getType()->isRecordType()) {
-      transferInlineCall(S, ConstructorDecl);
-      return;
-    }
-
-    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
-    Env.setValue(*S, refreshRecordValue(Loc, Env));
-
     if (ConstructorDecl->isCopyOrMoveConstructor()) {
       // It is permissible for a copy/move constructor to have additional
       // parameters as long as they have default arguments defined for them.
@@ -507,14 +491,24 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (ArgLoc == nullptr)
         return;
 
-      // Even if the copy/move constructor call is elidable, we choose to copy
-      // the record in all cases (which isn't wrong, just potentially not
-      // optimal).
-      copyRecord(*ArgLoc, Loc, Env);
+      if (S->isElidable()) {
+        if (Value *Val = Env.getValue(*ArgLoc))
+          Env.setValue(*S, *Val);
+      } else {
+        auto &Val = *cast<RecordValue>(Env.createValue(S->getType()));
+        Env.setValue(*S, Val);
+        copyRecord(*ArgLoc, Val.getLoc(), Env);
+      }
       return;
     }
 
-    Env.initializeFieldsWithValues(Loc, S->getType());
+    // `CXXConstructExpr` can have array type if default-initializing an array
+    // of records, and we currently can't create values for arrays. So check if
+    // we've got a record type.
+    if (S->getType()->isRecordType()) {
+      auto &InitialVal = *cast<RecordValue>(Env.createValue(S->getType()));
+      Env.setValue(*S, InitialVal);
+    }
 
     transferInlineCall(S, ConstructorDecl);
   }
@@ -557,15 +551,19 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (S->isGLValue()) {
         Env.setStorageLocation(*S, *LocDst);
       } else if (S->getType()->isRecordType()) {
-        // Assume that the assignment returns the assigned value.
-        copyRecord(*LocDst, Env.getResultObjectLocation(*S), Env);
+        // Make sure that we have a `RecordValue` for this expression so that
+        // `Environment::getResultObjectLocation()` is able to return a location
+        // for it.
+        if (Env.getValue(*S) == nullptr)
+          refreshRecordValue(*S, Env);
       }
 
       return;
     }
 
-    // `CXXOperatorCallExpr` can be a prvalue. Call `VisitCallExpr`() to
-    // initialize the prvalue's fields with values.
+    // CXXOperatorCallExpr can be prvalues. Call `VisitCallExpr`() to create
+    // a `RecordValue` for them so that `Environment::getResultObjectLocation()`
+    // can return a value.
     VisitCallExpr(S);
   }
 
@@ -582,6 +580,11 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     }
   }
 
+  void VisitCXXTemporaryObjectExpr(const CXXTemporaryObjectExpr *S) {
+    if (Value *Val = Env.createValue(S->getType()))
+      Env.setValue(*S, *Val);
+  }
+
   void VisitCallExpr(const CallExpr *S) {
     // Of clang's builtins, only `__builtin_expect` is handled explicitly, since
     // others (like trap, debugtrap, and unreachable) are handled by CFG
@@ -609,14 +612,13 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     } else if (const FunctionDecl *F = S->getDirectCallee()) {
       transferInlineCall(S, F);
 
-      // If this call produces a prvalue of record type, initialize its fields
-      // with values.
+      // If this call produces a prvalue of record type, make sure that we have
+      // a `RecordValue` for it. This is required so that
+      // `Environment::getResultObjectLocation()` is able to return a location
+      // for this `CallExpr`.
       if (S->getType()->isRecordType() && S->isPRValue())
-        if (Env.getValue(*S) == nullptr) {
-          RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
-          Env.initializeFieldsWithValues(Loc);
-          Env.setValue(*S, refreshRecordValue(Loc, Env));
-        }
+        if (Env.getValue(*S) == nullptr)
+          refreshRecordValue(*S, Env);
     }
   }
 
@@ -664,10 +666,8 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // `getLogicOperatorSubExprValue()`.
     if (S->isGLValue())
       Env.setStorageLocation(*S, Env.createObject(S->getType()));
-    else if (!S->getType()->isRecordType()) {
-      if (Value *Val = Env.createValue(S->getType()))
-        Env.setValue(*S, *Val);
-    }
+    else if (Value *Val = Env.createValue(S->getType()))
+      Env.setValue(*S, *Val);
   }
 
   void VisitInitListExpr(const InitListExpr *S) {
@@ -688,51 +688,71 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       return;
     }
 
-    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
-    Env.setValue(*S, refreshRecordValue(Loc, Env));
-
-    // Initialization of base classes and fields of record type happens when we
-    // visit the nested `CXXConstructExpr` or `InitListExpr` for that base class
-    // or field. We therefore only need to deal with fields of non-record type
-    // here.
-
+    llvm::DenseMap<const ValueDecl *, StorageLocation *> FieldLocs;
     RecordInitListHelper InitListHelper(S);
 
+    for (auto [Base, Init] : InitListHelper.base_inits()) {
+      assert(Base->getType().getCanonicalType() ==
+             Init->getType().getCanonicalType());
+      auto *BaseVal = Env.get<RecordValue>(*Init);
+      if (!BaseVal)
+        BaseVal = cast<RecordValue>(Env.createValue(Init->getType()));
+      // Take ownership of the fields of the `RecordValue` for the base class
+      // and incorporate them into the "flattened" set of fields for the
+      // derived class.
+      auto Children = BaseVal->getLoc().children();
+      FieldLocs.insert(Children.begin(), Children.end());
+    }
+
     for (auto [Field, Init] : InitListHelper.field_inits()) {
-      if (Field->getType()->isRecordType())
-        continue;
-      if (Field->getType()->isReferenceType()) {
-        assert(Field->getType().getCanonicalType()->getPointeeType() ==
-               Init->getType().getCanonicalType());
-        Loc.setChild(*Field, &Env.createObject(Field->getType(), Init));
-        continue;
-      }
-      assert(Field->getType().getCanonicalType().getUnqualifiedType() ==
-             Init->getType().getCanonicalType().getUnqualifiedType());
-      StorageLocation *FieldLoc = Loc.getChild(*Field);
-      // Locations for non-reference fields must always be non-null.
-      assert(FieldLoc != nullptr);
-      Value *Val = Env.getValue(*Init);
-      if (Val == nullptr && isa<ImplicitValueInitExpr>(Init) &&
-          Init->getType()->isPointerType())
-        Val =
-            &Env.getOrCreateNullPointerValue(Init->getType()->getPointeeType());
-      if (Val == nullptr)
-        Val = Env.createValue(Field->getType());
-      if (Val != nullptr)
-        Env.setValue(*FieldLoc, *Val);
+      assert(
+          // The types are same, or
+          Field->getType().getCanonicalType().getUnqualifiedType() ==
+              Init->getType().getCanonicalType().getUnqualifiedType() ||
+          // The field's type is T&, and initializer is T
+          (Field->getType()->isReferenceType() &&
+           Field->getType().getCanonicalType()->getPointeeType() ==
+               Init->getType().getCanonicalType()));
+      auto& Loc = Env.createObject(Field->getType(), Init);
+      FieldLocs.insert({Field, &Loc});
     }
 
-    for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
-      QualType FieldType = FieldLoc->getType();
-      if (FieldType->isRecordType()) {
-        Env.initializeFieldsWithValues(*cast<RecordStorageLocation>(FieldLoc));
-      } else {
-        if (Value *Val = Env.createValue(FieldType))
-          Env.setValue(*FieldLoc, *Val);
+    // In the case of a union, we don't in general have initializers for all
+    // of the fields. Create storage locations for the remaining fields (but
+    // don't associate them with values).
+    if (Type->isUnionType()) {
+      for (const FieldDecl *Field :
+           Env.getDataflowAnalysisContext().getModeledFields(Type)) {
+        if (auto [it, inserted] = FieldLocs.insert({Field, nullptr}); inserted)
+          it->second = &Env.createStorageLocation(Field->getType());
       }
     }
 
+    // Check that we satisfy the invariant that a `RecordStorageLoation`
+    // contains exactly the set of modeled fields for that type.
+    // `ModeledFields` includes fields from all the bases, but only the
+    // modeled ones. However, if a class type is initialized with an
+    // `InitListExpr`, all fields in the class, including those from base
+    // classes, are included in the set of modeled fields. The code above
+    // should therefore populate exactly the modeled fields.
+    assert(containsSameFields(
+        Env.getDataflowAnalysisContext().getModeledFields(Type), FieldLocs));
+
+    RecordStorageLocation::SyntheticFieldMap SyntheticFieldLocs;
+    for (const auto &Entry :
+         Env.getDataflowAnalysisContext().getSyntheticFields(Type)) {
+      SyntheticFieldLocs.insert(
+          {Entry.getKey(), &Env.createObject(Entry.getValue())});
+    }
+
+    auto &Loc = Env.getDataflowAnalysisContext().createRecordStorageLocation(
+        Type, std::move(FieldLocs), std::move(SyntheticFieldLocs));
+    RecordValue &RecordVal = Env.create<RecordValue>(Loc);
+
+    Env.setValue(Loc, RecordVal);
+
+    Env.setValue(*S, RecordVal);
+
     // FIXME: Implement array initialization.
   }
 
diff --git a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
index 1b73c5d683016..595f70f819ddb 100644
--- a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
+++ b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
@@ -369,10 +369,17 @@ builtinTransferInitializer(const CFGInitializer &Elt,
     ParentLoc->setChild(*Member, InitExprLoc);
   } else if (auto *InitExprVal = Env.getValue(*InitExpr)) {
     assert(MemberLoc != nullptr);
-    // Record-type initializers construct themselves directly into the result
-    // object, so there is no need to handle them here.
-    if (!Member->getType()->isRecordType())
+    if (Member->getType()->isRecordType()) {
+      auto *InitValStruct = cast<RecordValue>(InitExprVal);
+      // FIXME: Rather than performing a copy here, we should really be
+      // initializing the field in place. This would require us to propagate the
+      // storage location of the field to the AST node that creates the
+      // `RecordValue`.
+      copyRecord(InitValStruct->getLoc(),
+                 *cast<RecordStorageLocation>(MemberLoc), Env);
+    } else {
       Env.setValue(*MemberLoc, *InitExprVal);
+    }
   }
 }
 
diff --git a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
index cc20623f881ff..465a8e21690c4 100644
--- a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
@@ -24,7 +24,6 @@ namespace {
 
 using namespace clang;
 using namespace dataflow;
-using ::clang::dataflow::test::findValueDecl;
 using ::clang::dataflow::test::getFieldValue;
 using ::testing::Contains;
 using ::testing::IsNull;
@@ -200,48 +199,6 @@ TEST_F(EnvironmentTest, JoinRecords) {
   }
 }
 
-TEST_F(EnvironmentTest, DifferentReferenceLocInJoin) {
-  // This tests the case where the storage location for a reference-type
-  // variable is different for two states being joined. We used to believe this
-  // could not happen and therefore had an assertion disallowing this; this test
-  // exists to demonstrate that we can handle this condition without a failing
-  // assertion. See also the discussion here:
-  // https://discourse.llvm.org/t/70086/6
-
-  using namespace ast_matchers;
-
-  std::string Code = R"cc(
-    void f(int &ref) {}
-  )cc";
-
-  auto Unit =
-      tooling::buildASTFromCodeWithArgs(Code, {"-fsyntax-only", "-std=c++11"});
-  auto &Context = Unit->getASTContext();
-
-  ASSERT_EQ(Context.getDiagnostics().getClient()->getNumErrors(), 0U);
-
-  const ValueDecl *Ref = findValueDecl(Context, "ref");
-
-  Environment Env1(DAContext);
-  StorageLocation &Loc1 = Env1.createStorageLocation(Context.IntTy);
-  Env1.setStorageLocation(*Ref, Loc1);
-
-  Environment Env2(DAContext);
-  StorageLocation &Loc2 = Env2.createStorageLocation(Context.IntTy);
-  Env2.setStorageLocation(*Ref, Loc2);
-
-  EXPECT_NE(&Loc1, &Loc2);
-
-  Environment::ValueModel Model;
-  Environment EnvJoined =
-      Environment::join(Env1, Env2, Model, Environment::DiscardExprState);
-
-  // Joining environments with different storage locations for the same
-  // declaration results in the declaration being removed from the joined
-  // environment.
-  EXPECT_EQ(EnvJoined.getStorageLocation(*Ref), nullptr);
-}
-
 TEST_F(EnvironmentTest, InitGlobalVarsFun) {
   using namespace ast_matchers;
 
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index 00dafb2988c69..ca055a462a286 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -1582,9 +1582,10 @@ TEST(TransferTest, FieldsDontHaveValuesInConstructorWithBaseClass) {
       [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
          ASTContext &ASTCtx) {
         const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-        // The field of the base class should already have been initialized with
-        // a value by the base constructor.
-        EXPECT_NE(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
+        // FIXME: The field of the base class should already have been
+        // initialized with a value by the base constructor. This test documents
+        // the current buggy behavior.
+        EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
                                 ASTCtx, Env),
                   nullptr);
         EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "Val",
@@ -2997,12 +2998,8 @@ TEST(TransferTest, ResultObjectLocation) {
 
 TEST(TransferTest, ResultObjectLocationForDefaultArgExpr) {
   std::string Code = R"(
-    struct Inner {};
-    struct Outer {
-        Inner I = {};
-    };
-
-    void funcWithDefaultArg(Outer O = {});
+    struct S {};
+    void funcWithDefaultArg(S s = S());
     void target() {
       funcWithDefaultArg();
       // [[p]]
@@ -3061,7 +3058,13 @@ TEST(TransferTest, ResultObjectLocationForDefaultInitExpr) {
 
         RecordStorageLocation &Loc = Env.getResultObjectLocation(*DefaultInit);
 
-        EXPECT_EQ(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
+        // FIXME: The result object location for the `CXXDefaultInitExpr` should
+        // be the location of the member variable being initialized, but we
+        // don't do this correctly yet; see also comments in
+        // `builtinTransferInitializer()`.
+        // For the time being, we just document the current erroneous behavior
+        // here (this should be `EXPECT_EQ` when the behavior is fixed).
+        EXPECT_NE(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
       });
 }
 
@@ -3098,79 +3101,6 @@ TEST(TransferTest, ResultObjectLocationForCXXOperatorCallExpr) {
       });
 }
 
-TEST(TransferTest, ResultObjectLocationForStdInitializerListExpr) {
-  std::string Code = R"(
-    namespace std {
-    template <typename T>
-    struct initializer_list {};
-    } // namespace std
-
-    void target() {
-      std::initializer_list<int> list = {1};
-      // [[p]]
-    }
-  )";
-
-  using ast_matchers::cxxStdInitializerListExpr;
-  using ast_matchers::match;
-  using ast_matchers::selectFirst;
-  runDataflow(
-      Code,
-      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
-         ASTContext &ASTCtx) {
-        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-
-        auto *StdInitList = selectFirst<CXXStdInitializerListExpr>(
-            "std_init_list",
-            match(cxxStdInitializerListExpr().bind("std_init_list"), ASTCtx));
-        ASSERT_NE(StdInitList, nullptr);
-
-        EXPECT_EQ(&Env.getResultObjectLocation(*StdInitList),
-                  &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "list"));
-      });
-}
-
-TEST(TransferTest, ResultObjectLocationPropagatesThroughConditionalOperator) {
-  std::string Code = R"(
-    struct A {
-      A(int);
-    };
-
-    void target(bool b) {
-      A a = b ? A(0) : A(1);
-      (void)0; // [[p]]
-    }
-  )";
-  using ast_matchers::cxxConstructExpr;
-  using ast_matchers::equals;
-  using ast_matchers::hasArgument;
-  using ast_matchers::integerLiteral;
-  using ast_matchers::match;
-  using ast_matchers::selectFirst;
-  using ast_matchers::traverse;
-  runDataflow(
-      Code,
-      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
-         ASTContext &ASTCtx) {
-        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-
-        auto *ConstructExpr0 = selectFirst<CXXConstructExpr>(
-            "construct",
-            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(0))))
-                      .bind("construct"),
-                  ASTCtx));
-        auto *ConstructExpr1 = selectFirst<CXXConstructExpr>(
-            "construct",
-            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(1))))
-                      .bind("construct"),
-                  ASTCtx));
-
-        auto &ALoc = getLocForDecl<StorageLocation>(ASTCtx, Env, "a");
-        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr0), &ALoc);
-        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr1), &ALoc);
-      });
-}
-
 TEST(TransferTest, StaticCast) {
   std::string Code = R"(
     void target(int Foo) {
@@ -5956,38 +5886,6 @@ TEST(TransferTest, ContextSensitiveReturnRecord) {
       {BuiltinOptions{ContextSensitiveOptions{}}});
 }
 
-TEST(TransferTest, ContextSensitiveReturnSelfReferentialRecord) {
-  std::string Code = R"(
-    struct S {
-      S() { self = this; }
-      S *self;
-    };
-
-    S makeS() {
-      // RVO guarantees that this will be constructed directly into `MyS`.
-      return S();
-    }
-
-    void target() {
-      S MyS = makeS();
-      // [[p]]
-    }
-  )";
-  runDataflow(
-      Code,
-      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
-         ASTContext &ASTCtx) {
-        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-
-        auto &MySLoc = getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "MyS");
-
-        auto *SelfVal =
-            cast<PointerValue>(getFieldValue(&MySLoc, "self", ASTCtx, Env));
-        EXPECT_EQ(&SelfVal->getPointeeLoc(), &MySLoc);
-      },
-      {BuiltinOptions{ContextSensitiveOptions{}}});
-}
-
 TEST(TransferTest, ContextSensitiveMethodLiteral) {
   std::string Code = R"(
     class MyClass {
@@ -6932,6 +6830,50 @@ TEST(TransferTest, LambdaCaptureThis) {
       });
 }
 
+TEST(TransferTest, DifferentReferenceLocInJoin) {
+  // This test triggers a case where the storage location for a reference-type
+  // variable is different for two states being joined. We used to believe this
+  // could not happen and therefore had an assertion disallowing this; this test
+  // exists to demonstrate that we can handle this condition without a failing
+  // assertion. See also the discussion here:
+  // https://discourse.llvm.org/t/70086/6
+  std::string Code = R"(
+    namespace std {
+      template <class T> struct initializer_list {
+        const T* begin();
+        const T* end();
+      };
+    }
+
+    void target(char* p, char* end) {
+      while (p != end) {
+        if (*p == ' ') {
+          p++;
+          continue;
+        }
+
+        auto && range = {1, 2};
+        for (auto b = range.begin(), e = range.end(); b != e; ++b) {
+        }
+        (void)0;
+        // [[p]]
+      }
+    }
+  )";
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        // Joining environments with different storage locations for the same
+        // declaration results in the declaration being removed from the joined
+        // environment.
+        const ValueDecl *VD = findValueDecl(ASTCtx, "range");
+        ASSERT_EQ(Env.getStorageLocation(*VD), nullptr);
+      });
+}
+
 // This test verifies correct modeling of a relational dependency that goes
 // through unmodeled functions (the simple `cond()` in this case).
 TEST(TransferTest, ConditionalRelation) {

From a6d1366b736cad85b3bb9fbdda340e07488d6cde Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Wed, 10 Apr 2024 12:41:26 -0700
Subject: [PATCH 063/886] [NFC] Remove a pair of incorrect comments from
 ParseOpenACC

We attempt to continue parsing, but the comment says the opposite.  Just
remove the inaccurate comments in this patch.
---
 clang/lib/Parse/ParseOpenACC.cpp | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/clang/lib/Parse/ParseOpenACC.cpp b/clang/lib/Parse/ParseOpenACC.cpp
index 59a4a5f534676..b487a1968d1ec 100644
--- a/clang/lib/Parse/ParseOpenACC.cpp
+++ b/clang/lib/Parse/ParseOpenACC.cpp
@@ -843,8 +843,7 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
     }
     case OpenACCClauseKind::If: {
       ExprResult CondExpr = ParseOpenACCConditionalExpr(*this);
-      // An invalid expression can be just about anything, so just give up on
-      // this clause list.
+
       if (CondExpr.isInvalid()) {
         Parens.skipToEnd();
         return OpenACCCanContinue();
@@ -966,8 +965,7 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
       case OpenACCClauseKind::Self: {
         assert(DirKind != OpenACCDirectiveKind::Update);
         ExprResult CondExpr = ParseOpenACCConditionalExpr(*this);
-        // An invalid expression can be just about anything, so just give up on
-        // this clause list.
+
         if (CondExpr.isInvalid()) {
           Parens.skipToEnd();
           return OpenACCCanContinue();

From b3792ae42a4adda5cb51d53f3d6a4b9b025b11fd Mon Sep 17 00:00:00 2001
From: Xing Xue <xingxue@outlook.com>
Date: Wed, 10 Apr 2024 16:06:31 -0400
Subject: [PATCH 064/886] [OpenMP][AIX] Fix test config for AIX (#88272)

This patch fixes the test config so that it works for
`tasking/omp50_taskdep_depobj.c` which uses different flags to test with
compiler's `omp.h`.
* set test environment variable `OBJECT_MODE` to `64` if it is set
explicitly to `64` in the AIX environment. `OBJECT_MODE` is default to
`32` and is recognized by AIX compilers and toolchain. In this way, we
don't need to set `-m64` for all compiler flags for 64-bit mode
* add option `-Wl,-bmaxdata` to 32-bit `test_openmp_flags` used by
`tasking/omp50_taskdep_depobj.c`
---
 openmp/runtime/test/lit.cfg | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/openmp/runtime/test/lit.cfg b/openmp/runtime/test/lit.cfg
index e27e52bb4289b..e8f7f3470580e 100644
--- a/openmp/runtime/test/lit.cfg
+++ b/openmp/runtime/test/lit.cfg
@@ -112,13 +112,15 @@ if config.operating_system == 'AIX':
     config.available_features.add("aix")
     object_mode = os.environ.get('OBJECT_MODE', '32')
     if object_mode == '64':
-      config.test_flags += " -m64"
+      # Set OBJECT_MODE to 64 for LIT test if it is explicitly set.
+      config.environment['OBJECT_MODE'] = os.environ['OBJECT_MODE']
     elif object_mode == '32':
       # Set user data area to 2GB since the default size 256MB in 32-bit mode
       # is not sufficient to run LIT tests on systems that have a lot of
       # CPUs when creating one worker thread for each CPU and each worker
       # thread uses 4MB stack size.
       config.test_flags += " -Wl,-bmaxdata:0x80000000"
+      config.test_openmp_flags += " -Wl,-bmaxdata:0x80000000"
 
 if 'Linux' in config.operating_system:
     config.available_features.add("linux")

From a12836647e08c4ad203b9834ac55892fa0b9f2d3 Mon Sep 17 00:00:00 2001
From: David Pagan <dave.pagan@amd.com>
Date: Wed, 10 Apr 2024 13:09:17 -0700
Subject: [PATCH 065/886] [OpenMP][CodeGen] Improved codegen for combined loop
 directives (#87278)

IR for 'target teams loop' is now dependent on suitability of associated
loop-nest.

If a loop-nest:

- does not contain a function call, or
- the -fopenmp-assume-no-nested-parallelism has been specified,
- or the call is to an OpenMP API AND
- does not contain nested loop bind(parallel) directives

then it can be emitted as 'target teams distribute parallel for', which
is the current default. Otherwise, it is emitted as 'target teams
distribute'.

Added debug output indicating how 'target teams loop' was emitted. Flag
is -mllvm -debug-only=target-teams-loop-codegen

Added LIT tests explicitly verifying 'target teams loop' emitted as a
parallel loop and a distribute loop.

Updated other 'loop' related tests as needed to reflect change in IR.
- These updates account for most of the changed files and
additions/deletions.
---
 clang/include/clang/AST/StmtOpenMP.h          |   11 +-
 clang/lib/AST/StmtOpenMP.cpp                  |    3 +-
 clang/lib/CodeGen/CGOpenMPRuntime.cpp         |   15 +-
 clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp      |    7 +-
 clang/lib/CodeGen/CGStmtOpenMP.cpp            |   87 +-
 clang/lib/Sema/SemaOpenMP.cpp                 |   86 +-
 clang/lib/Serialization/ASTReaderStmt.cpp     |    1 +
 clang/lib/Serialization/ASTWriterStmt.cpp     |    1 +
 ...vptx_target_teams_generic_loop_codegen.cpp |   48 +-
 ...eams_generic_loop_generic_mode_codegen.cpp |  397 +-
 .../target_teams_generic_loop_codegen.cpp     | 1132 +----
 ...ams_generic_loop_codegen_as_distribute.cpp |  587 +++
 ...s_generic_loop_codegen_as_parallel_for.cpp | 3998 +++++++++++++++++
 .../target_teams_generic_loop_if_codegen.cpp  |  710 +--
 ...get_teams_generic_loop_private_codegen.cpp | 1438 +-----
 .../OpenMP/teams_generic_loop_codegen-1.cpp   | 1360 +-----
 .../OpenMP/teams_generic_loop_codegen.cpp     |  630 +--
 .../teams_generic_loop_collapse_codegen.cpp   |  808 +---
 .../teams_generic_loop_private_codegen.cpp    |  685 +--
 .../teams_generic_loop_reduction_codegen.cpp  |  785 +---
 20 files changed, 5886 insertions(+), 6903 deletions(-)
 create mode 100644 clang/test/OpenMP/target_teams_generic_loop_codegen_as_distribute.cpp
 create mode 100644 clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp

diff --git a/clang/include/clang/AST/StmtOpenMP.h b/clang/include/clang/AST/StmtOpenMP.h
index 3cb3c1014d73b..f735fa5643aec 100644
--- a/clang/include/clang/AST/StmtOpenMP.h
+++ b/clang/include/clang/AST/StmtOpenMP.h
@@ -6109,6 +6109,8 @@ class OMPTeamsGenericLoopDirective final : public OMPLoopDirective {
 class OMPTargetTeamsGenericLoopDirective final : public OMPLoopDirective {
   friend class ASTStmtReader;
   friend class OMPExecutableDirective;
+  /// true if loop directive's associated loop can be a parallel for.
+  bool CanBeParallelFor = false;
   /// Build directive with the given start and end location.
   ///
   /// \param StartLoc Starting location of the directive kind.
@@ -6131,6 +6133,9 @@ class OMPTargetTeamsGenericLoopDirective final : public OMPLoopDirective {
                          llvm::omp::OMPD_target_teams_loop, SourceLocation(),
                          SourceLocation(), CollapsedNum) {}
 
+  /// Set whether associated loop can be a parallel for.
+  void setCanBeParallelFor(bool ParFor) { CanBeParallelFor = ParFor; }
+
 public:
   /// Creates directive with a list of \p Clauses.
   ///
@@ -6145,7 +6150,7 @@ class OMPTargetTeamsGenericLoopDirective final : public OMPLoopDirective {
   static OMPTargetTeamsGenericLoopDirective *
   Create(const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
          unsigned CollapsedNum, ArrayRef<OMPClause *> Clauses,
-         Stmt *AssociatedStmt, const HelperExprs &Exprs);
+         Stmt *AssociatedStmt, const HelperExprs &Exprs, bool CanBeParallelFor);
 
   /// Creates an empty directive with the place
   /// for \a NumClauses clauses.
@@ -6159,6 +6164,10 @@ class OMPTargetTeamsGenericLoopDirective final : public OMPLoopDirective {
                                                          unsigned CollapsedNum,
                                                          EmptyShell);
 
+  /// Return true if current loop directive's associated loop can be a
+  /// parallel for.
+  bool canBeParallelFor() const { return CanBeParallelFor; }
+
   static bool classof(const Stmt *T) {
     return T->getStmtClass() == OMPTargetTeamsGenericLoopDirectiveClass;
   }
diff --git a/clang/lib/AST/StmtOpenMP.cpp b/clang/lib/AST/StmtOpenMP.cpp
index 426b35848cb5c..d8519b2071e6d 100644
--- a/clang/lib/AST/StmtOpenMP.cpp
+++ b/clang/lib/AST/StmtOpenMP.cpp
@@ -2431,7 +2431,7 @@ OMPTeamsGenericLoopDirective::CreateEmpty(const ASTContext &C,
 OMPTargetTeamsGenericLoopDirective *OMPTargetTeamsGenericLoopDirective::Create(
     const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
     unsigned CollapsedNum, ArrayRef<OMPClause *> Clauses, Stmt *AssociatedStmt,
-    const HelperExprs &Exprs) {
+    const HelperExprs &Exprs, bool CanBeParallelFor) {
   auto *Dir = createDirective<OMPTargetTeamsGenericLoopDirective>(
       C, Clauses, AssociatedStmt,
       numLoopChildren(CollapsedNum, OMPD_target_teams_loop), StartLoc, EndLoc,
@@ -2473,6 +2473,7 @@ OMPTargetTeamsGenericLoopDirective *OMPTargetTeamsGenericLoopDirective::Create(
   Dir->setCombinedNextUpperBound(Exprs.DistCombinedFields.NUB);
   Dir->setCombinedDistCond(Exprs.DistCombinedFields.DistCond);
   Dir->setCombinedParForInDistCond(Exprs.DistCombinedFields.ParForInDistCond);
+  Dir->setCanBeParallelFor(CanBeParallelFor);
   return Dir;
 }
 
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 8eb10584699fa..2ae11e129c75e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -2656,11 +2656,12 @@ void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
   // Call __kmpc_for_static_fini(ident_t *loc, kmp_int32 tid);
   llvm::Value *Args[] = {
       emitUpdateLocation(CGF, Loc,
-                         isOpenMPDistributeDirective(DKind)
+                         isOpenMPDistributeDirective(DKind) ||
+                                 (DKind == OMPD_target_teams_loop)
                              ? OMP_IDENT_WORK_DISTRIBUTE
-                             : isOpenMPLoopDirective(DKind)
-                                   ? OMP_IDENT_WORK_LOOP
-                                   : OMP_IDENT_WORK_SECTIONS),
+                         : isOpenMPLoopDirective(DKind)
+                             ? OMP_IDENT_WORK_LOOP
+                             : OMP_IDENT_WORK_SECTIONS),
       getThreadID(CGF, Loc)};
   auto DL = ApplyDebugLocation::CreateDefaultArtificial(CGF, Loc);
   if (isOpenMPDistributeDirective(DKind) &&
@@ -8885,7 +8886,8 @@ getNestedDistributeDirective(ASTContext &Ctx, const OMPExecutableDirective &D) {
     OpenMPDirectiveKind DKind = NestedDir->getDirectiveKind();
     switch (D.getDirectiveKind()) {
     case OMPD_target:
-      // For now, just treat 'target teams loop' as if it's distributed.
+      // For now, treat 'target' with nested 'teams loop' as if it's
+      // distributed (target teams distribute).
       if (isOpenMPDistributeDirective(DKind) || DKind == OMPD_teams_loop)
         return NestedDir;
       if (DKind == OMPD_teams) {
@@ -9369,7 +9371,8 @@ llvm::Value *CGOpenMPRuntime::emitTargetNumIterationsCall(
         SizeEmitter) {
   OpenMPDirectiveKind Kind = D.getDirectiveKind();
   const OMPExecutableDirective *TD = &D;
-  // Get nested teams distribute kind directive, if any.
+  // Get nested teams distribute kind directive, if any. For now, treat
+  // 'target_teams_loop' as if it's really a target_teams_distribute.
   if ((!isOpenMPDistributeDirective(Kind) || !isOpenMPTeamsDirective(Kind)) &&
       Kind != OMPD_target_teams_loop)
     TD = getNestedDistributeDirective(CGM.getContext(), D);
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 5baac8f0e3e26..59ba03c6b8625 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -646,7 +646,6 @@ static bool supportsSPMDExecutionMode(ASTContext &Ctx,
   case OMPD_target:
   case OMPD_target_teams:
     return hasNestedSPMDDirective(Ctx, D);
-  case OMPD_target_teams_loop:
   case OMPD_target_parallel_loop:
   case OMPD_target_parallel:
   case OMPD_target_parallel_for:
@@ -658,6 +657,12 @@ static bool supportsSPMDExecutionMode(ASTContext &Ctx,
     return true;
   case OMPD_target_teams_distribute:
     return false;
+  case OMPD_target_teams_loop:
+    // Whether this is true or not depends on how the directive will
+    // eventually be emitted.
+    if (auto *TTLD = dyn_cast<OMPTargetTeamsGenericLoopDirective>(&D))
+      return TTLD->canBeParallelFor();
+    return false;
   case OMPD_parallel:
   case OMPD_for:
   case OMPD_parallel_for:
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index e6d504bcdeca5..3bf99366b69ce 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -24,6 +24,7 @@
 #include "clang/AST/StmtVisitor.h"
 #include "clang/Basic/OpenMPKinds.h"
 #include "clang/Basic/PrettyStackTrace.h"
+#include "clang/Basic/SourceManager.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/BinaryFormat/Dwarf.h"
 #include "llvm/Frontend/OpenMP/OMPConstants.h"
@@ -34,11 +35,14 @@
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Metadata.h"
 #include "llvm/Support/AtomicOrdering.h"
+#include "llvm/Support/Debug.h"
 #include <optional>
 using namespace clang;
 using namespace CodeGen;
 using namespace llvm::omp;
 
+#define TTL_CODEGEN_TYPE "target-teams-loop-codegen"
+
 static const VarDecl *getBaseDecl(const Expr *Ref);
 
 namespace {
@@ -1432,9 +1436,12 @@ void CodeGenFunction::EmitOMPReductionClauseFinal(
           *this, D.getBeginLoc(),
           isOpenMPWorksharingDirective(D.getDirectiveKind()));
     }
+    bool TeamsLoopCanBeParallel = false;
+    if (auto *TTLD = dyn_cast<OMPTargetTeamsGenericLoopDirective>(&D))
+      TeamsLoopCanBeParallel = TTLD->canBeParallelFor();
     bool WithNowait = D.getSingleClause<OMPNowaitClause>() ||
                       isOpenMPParallelDirective(D.getDirectiveKind()) ||
-                      ReductionKind == OMPD_simd;
+                      TeamsLoopCanBeParallel || ReductionKind == OMPD_simd;
     bool SimpleReduction = ReductionKind == OMPD_simd;
     // Emit nowait reduction if nowait clause is present or directive is a
     // parallel directive (it always has implicit barrier).
@@ -7928,11 +7935,9 @@ void CodeGenFunction::EmitOMPParallelGenericLoopDirective(
 void CodeGenFunction::EmitOMPTeamsGenericLoopDirective(
     const OMPTeamsGenericLoopDirective &S) {
   // To be consistent with current behavior of 'target teams loop', emit
-  // 'teams loop' as if its constituent constructs are 'distribute,
-  // 'parallel, and 'for'.
+  // 'teams loop' as if its constituent constructs are 'teams' and 'distribute'.
   auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
-    CGF.EmitOMPDistributeLoop(S, emitInnerParallelForWhenCombined,
-                              S.getDistInc());
+    CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
   };
 
   // Emit teams region as a standalone region.
@@ -7946,15 +7951,33 @@ void CodeGenFunction::EmitOMPTeamsGenericLoopDirective(
                                                     CodeGenDistribute);
     CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
   };
-  emitCommonOMPTeamsDirective(*this, S, OMPD_distribute_parallel_for, CodeGen);
+  emitCommonOMPTeamsDirective(*this, S, OMPD_distribute, CodeGen);
   emitPostUpdateForReductionClause(*this, S,
                                    [](CodeGenFunction &) { return nullptr; });
 }
 
-static void
-emitTargetTeamsGenericLoopRegion(CodeGenFunction &CGF,
-                                 const OMPTargetTeamsGenericLoopDirective &S,
-                                 PrePostActionTy &Action) {
+static void emitTargetTeamsLoopCodegenStatus(CodeGenFunction &CGF,
+                                             std::string StatusMsg,
+                                             const OMPExecutableDirective &D) {
+#ifndef NDEBUG
+  bool IsDevice = CGF.CGM.getLangOpts().OpenMPIsTargetDevice;
+  if (IsDevice)
+    StatusMsg += ": DEVICE";
+  else
+    StatusMsg += ": HOST";
+  SourceLocation L = D.getBeginLoc();
+  auto &SM = CGF.getContext().getSourceManager();
+  PresumedLoc PLoc = SM.getPresumedLoc(L);
+  const char *FileName = PLoc.isValid() ? PLoc.getFilename() : nullptr;
+  unsigned LineNo =
+      PLoc.isValid() ? PLoc.getLine() : SM.getExpansionLineNumber(L);
+  llvm::dbgs() << StatusMsg << ": " << FileName << ": " << LineNo << "\n";
+#endif
+}
+
+static void emitTargetTeamsGenericLoopRegionAsParallel(
+    CodeGenFunction &CGF, PrePostActionTy &Action,
+    const OMPTargetTeamsGenericLoopDirective &S) {
   Action.Enter(CGF);
   // Emit 'teams loop' as if its constituent constructs are 'distribute,
   // 'parallel, and 'for'.
@@ -7974,19 +7997,50 @@ emitTargetTeamsGenericLoopRegion(CodeGenFunction &CGF,
         CGF, OMPD_distribute, CodeGenDistribute, /*HasCancel=*/false);
     CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
   };
-
+  DEBUG_WITH_TYPE(TTL_CODEGEN_TYPE,
+                  emitTargetTeamsLoopCodegenStatus(
+                      CGF, TTL_CODEGEN_TYPE " as parallel for", S));
   emitCommonOMPTeamsDirective(CGF, S, OMPD_distribute_parallel_for,
                               CodeGenTeams);
   emitPostUpdateForReductionClause(CGF, S,
                                    [](CodeGenFunction &) { return nullptr; });
 }
 
-/// Emit combined directive 'target teams loop' as if its constituent
-/// constructs are 'target', 'teams', 'distribute', 'parallel', and 'for'.
+static void emitTargetTeamsGenericLoopRegionAsDistribute(
+    CodeGenFunction &CGF, PrePostActionTy &Action,
+    const OMPTargetTeamsGenericLoopDirective &S) {
+  Action.Enter(CGF);
+  // Emit 'teams loop' as if its constituent construct is 'distribute'.
+  auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+    CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+  };
+
+  // Emit teams region as a standalone region.
+  auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+                                            PrePostActionTy &Action) {
+    Action.Enter(CGF);
+    CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
+    CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+    (void)PrivateScope.Privatize();
+    CGF.CGM.getOpenMPRuntime().emitInlinedDirective(
+        CGF, OMPD_distribute, CodeGenDistribute, /*HasCancel=*/false);
+    CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+  };
+  DEBUG_WITH_TYPE(TTL_CODEGEN_TYPE,
+                  emitTargetTeamsLoopCodegenStatus(
+                      CGF, TTL_CODEGEN_TYPE " as distribute", S));
+  emitCommonOMPTeamsDirective(CGF, S, OMPD_distribute, CodeGen);
+  emitPostUpdateForReductionClause(CGF, S,
+                                   [](CodeGenFunction &) { return nullptr; });
+}
+
 void CodeGenFunction::EmitOMPTargetTeamsGenericLoopDirective(
     const OMPTargetTeamsGenericLoopDirective &S) {
   auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
-    emitTargetTeamsGenericLoopRegion(CGF, S, Action);
+    if (S.canBeParallelFor())
+      emitTargetTeamsGenericLoopRegionAsParallel(CGF, Action, S);
+    else
+      emitTargetTeamsGenericLoopRegionAsDistribute(CGF, Action, S);
   };
   emitCommonOMPTargetDirective(*this, S, CodeGen);
 }
@@ -7996,7 +8050,10 @@ void CodeGenFunction::EmitOMPTargetTeamsGenericLoopDeviceFunction(
     const OMPTargetTeamsGenericLoopDirective &S) {
   // Emit SPMD target parallel loop region as a standalone region.
   auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
-    emitTargetTeamsGenericLoopRegion(CGF, S, Action);
+    if (S.canBeParallelFor())
+      emitTargetTeamsGenericLoopRegionAsParallel(CGF, Action, S);
+    else
+      emitTargetTeamsGenericLoopRegionAsDistribute(CGF, Action, S);
   };
   llvm::Function *Fn;
   llvm::Constant *Addr;
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index 0ba54a3a9cae3..c814535ad6bdb 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -4478,6 +4478,8 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
                              Params);
     break;
   }
+  // For 'target teams loop', collect all captured regions so codegen can
+  // later decide the best IR to emit given the associated loop-nest.
   case OMPD_target_teams_loop:
   case OMPD_target_teams_distribute_parallel_for:
   case OMPD_target_teams_distribute_parallel_for_simd: {
@@ -6135,6 +6137,79 @@ processImplicitMapsWithDefaultMappers(Sema &S, DSAStackTy *Stack,
   }
 }
 
+namespace {
+/// A 'teams loop' with a nested 'loop bind(parallel)' or generic function
+/// call in the associated loop-nest cannot be a 'parallel for'.
+class TeamsLoopChecker final : public ConstStmtVisitor<TeamsLoopChecker> {
+  Sema &SemaRef;
+
+public:
+  bool teamsLoopCanBeParallelFor() const { return TeamsLoopCanBeParallelFor; }
+
+  // Is there a nested OpenMP loop bind(parallel)
+  void VisitOMPExecutableDirective(const OMPExecutableDirective *D) {
+    if (D->getDirectiveKind() == llvm::omp::Directive::OMPD_loop) {
+      if (const auto *C = D->getSingleClause<OMPBindClause>())
+        if (C->getBindKind() == OMPC_BIND_parallel) {
+          TeamsLoopCanBeParallelFor = false;
+          // No need to continue visiting any more
+          return;
+        }
+    }
+    for (const Stmt *Child : D->children())
+      if (Child)
+        Visit(Child);
+  }
+
+  void VisitCallExpr(const CallExpr *C) {
+    // Function calls inhibit parallel loop translation of 'target teams loop'
+    // unless the assume-no-nested-parallelism flag has been specified.
+    // OpenMP API runtime library calls do not inhibit parallel loop
+    // translation, regardless of the assume-no-nested-parallelism.
+    if (C) {
+      bool IsOpenMPAPI = false;
+      auto *FD = dyn_cast_or_null<FunctionDecl>(C->getCalleeDecl());
+      if (FD) {
+        std::string Name = FD->getNameInfo().getAsString();
+        IsOpenMPAPI = Name.find("omp_") == 0;
+      }
+      TeamsLoopCanBeParallelFor =
+          IsOpenMPAPI || SemaRef.getLangOpts().OpenMPNoNestedParallelism;
+      if (!TeamsLoopCanBeParallelFor)
+        return;
+    }
+    for (const Stmt *Child : C->children())
+      if (Child)
+        Visit(Child);
+  }
+
+  void VisitCapturedStmt(const CapturedStmt *S) {
+    if (!S)
+      return;
+    Visit(S->getCapturedDecl()->getBody());
+  }
+
+  void VisitStmt(const Stmt *S) {
+    if (!S)
+      return;
+    for (const Stmt *Child : S->children())
+      if (Child)
+        Visit(Child);
+  }
+  explicit TeamsLoopChecker(Sema &SemaRef)
+      : SemaRef(SemaRef), TeamsLoopCanBeParallelFor(true) {}
+
+private:
+  bool TeamsLoopCanBeParallelFor;
+};
+} // namespace
+
+static bool teamsLoopCanBeParallelFor(Stmt *AStmt, Sema &SemaRef) {
+  TeamsLoopChecker Checker(SemaRef);
+  Checker.Visit(AStmt);
+  return Checker.teamsLoopCanBeParallelFor();
+}
+
 bool Sema::mapLoopConstruct(llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
                             ArrayRef<OMPClause *> Clauses,
                             OpenMPBindClauseKind &BindKind,
@@ -10895,7 +10970,8 @@ StmtResult Sema::ActOnOpenMPTargetTeamsGenericLoopDirective(
   setFunctionHasBranchProtectedScope();
 
   return OMPTargetTeamsGenericLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      teamsLoopCanBeParallelFor(AStmt, *this));
 }
 
 StmtResult Sema::ActOnOpenMPParallelGenericLoopDirective(
@@ -15645,6 +15721,12 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
       if (NameModifier == OMPD_unknown || NameModifier == OMPD_parallel)
         CaptureRegion = OMPD_target;
       break;
+    case OMPD_teams_loop:
+    case OMPD_target_teams_loop:
+      // For [target] teams loop, assume capture region is 'teams' so it's
+      // available for codegen later to use if/when necessary.
+      CaptureRegion = OMPD_teams;
+      break;
     case OMPD_target_teams_distribute_parallel_for_simd:
       if (OpenMPVersion >= 50 &&
           (NameModifier == OMPD_unknown || NameModifier == OMPD_simd)) {
@@ -15652,7 +15734,6 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
         break;
       }
       [[fallthrough]];
-    case OMPD_target_teams_loop:
     case OMPD_target_teams_distribute_parallel_for:
       // If this clause applies to the nested 'parallel' region, capture within
       // the 'teams' region, otherwise do not capture.
@@ -15775,7 +15856,6 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_loop:
-    case OMPD_teams_loop:
     case OMPD_teams:
     case OMPD_tile:
     case OMPD_unroll:
diff --git a/clang/lib/Serialization/ASTReaderStmt.cpp b/clang/lib/Serialization/ASTReaderStmt.cpp
index 3ddb15b10f48b..ca0460800898b 100644
--- a/clang/lib/Serialization/ASTReaderStmt.cpp
+++ b/clang/lib/Serialization/ASTReaderStmt.cpp
@@ -2769,6 +2769,7 @@ void ASTStmtReader::VisitOMPTeamsGenericLoopDirective(
 void ASTStmtReader::VisitOMPTargetTeamsGenericLoopDirective(
     OMPTargetTeamsGenericLoopDirective *D) {
   VisitOMPLoopDirective(D);
+  D->setCanBeParallelFor(Record.readBool());
 }
 
 void ASTStmtReader::VisitOMPParallelGenericLoopDirective(
diff --git a/clang/lib/Serialization/ASTWriterStmt.cpp b/clang/lib/Serialization/ASTWriterStmt.cpp
index ddee5db2b69ef..e3816181e2b2b 100644
--- a/clang/lib/Serialization/ASTWriterStmt.cpp
+++ b/clang/lib/Serialization/ASTWriterStmt.cpp
@@ -2821,6 +2821,7 @@ void ASTStmtWriter::VisitOMPTeamsGenericLoopDirective(
 void ASTStmtWriter::VisitOMPTargetTeamsGenericLoopDirective(
     OMPTargetTeamsGenericLoopDirective *D) {
   VisitOMPLoopDirective(D);
+  Record.writeBool(D->canBeParallelFor());
   Code = serialization::STMT_OMP_TARGET_TEAMS_GENERIC_LOOP_DIRECTIVE;
 }
 
diff --git a/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
index 5226b7498e4ce..c89c6eb65706a 100644
--- a/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_generic_loop_codegen.cpp
@@ -320,23 +320,31 @@ int bar(int n){
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33
-// CHECK1-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR4:[0-9]+]] {
+// CHECK1-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_kernel_environment, ptr [[DYN_PTR]])
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK1-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK1:       user_code.entry:
 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
+// CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]]) #[[ATTR2]]
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]], i64 [[TMP4]]) #[[ATTR2]]
 // CHECK1-NEXT:    call void @__kmpc_target_deinit()
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
@@ -344,11 +352,12 @@ int bar(int n){
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
@@ -360,6 +369,7 @@ int bar(int n){
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
@@ -1566,23 +1576,31 @@ int bar(int n){
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33
-// CHECK2-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR4:[0-9]+]] {
+// CHECK2-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK2-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 8
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_kernel_environment, ptr [[DYN_PTR]])
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK2-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK2:       user_code.entry:
 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK2-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
+// CHECK2-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK2-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK2-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]]) #[[ATTR2]]
+// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]], i64 [[TMP4]]) #[[ATTR2]]
 // CHECK2-NEXT:    call void @__kmpc_target_deinit()
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
@@ -1590,11 +1608,12 @@ int bar(int n){
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined
-// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
+// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
@@ -1606,6 +1625,7 @@ int bar(int n){
 // CHECK2-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK2-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 8
 // CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK2-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
@@ -2801,23 +2821,31 @@ int bar(int n){
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33
-// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR4:[0-9]+]] {
+// CHECK3-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_kernel_environment, ptr [[DYN_PTR]])
 // CHECK3-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
 // CHECK3-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
 // CHECK3:       user_code.entry:
 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
+// CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP_]], align 4
-// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]]) #[[ATTR2]]
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], ptr [[TMP0]], i32 [[TMP4]]) #[[ATTR2]]
 // CHECK3-NEXT:    call void @__kmpc_target_deinit()
 // CHECK3-NEXT:    ret void
 // CHECK3:       worker.exit:
@@ -2825,11 +2853,12 @@ int bar(int n){
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l33_omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR1]] {
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
@@ -2841,6 +2870,7 @@ int bar(int n){
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[B_ADDR]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
diff --git a/clang/test/OpenMP/nvptx_target_teams_generic_loop_generic_mode_codegen.cpp b/clang/test/OpenMP/nvptx_target_teams_generic_loop_generic_mode_codegen.cpp
index ca2670f0cd643..f0effa760dcdb 100644
--- a/clang/test/OpenMP/nvptx_target_teams_generic_loop_generic_mode_codegen.cpp
+++ b/clang/test/OpenMP/nvptx_target_teams_generic_loop_generic_mode_codegen.cpp
@@ -30,17 +30,20 @@ int main(int argc, char **argv) {
 
 #endif
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24
-// CHECK1-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK1-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[ARGC_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[ARGC]], ptr [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_kernel_environment, ptr [[DYN_PTR]])
 // CHECK1-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
@@ -50,9 +53,14 @@ int main(int argc, char **argv) {
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP3]], ptr [[ARGC_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[ARGC_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP5]] to i1
+// CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP_]], align 4
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i64 [[TMP4]], ptr [[TMP0]]) #[[ATTR2:[0-9]+]]
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i64 [[TMP4]], ptr [[TMP0]], i64 [[TMP6]]) #[[ATTR2:[0-9]+]]
 // CHECK1-NEXT:    call void @__kmpc_target_deinit()
 // CHECK1-NEXT:    ret void
 // CHECK1:       worker.exit:
@@ -60,56 +68,55 @@ int main(int argc, char **argv) {
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[ARGC_CASTED:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8
+// CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[ARGC]], ptr [[ARGC_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK1-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK1:       omp.precond.then:
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK1-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
+// CHECK1-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
@@ -122,177 +129,54 @@ int main(int argc, char **argv) {
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK1-NEXT:    store i32 [[TMP18]], ptr [[ARGC_CASTED]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[ARGC_CASTED]], align 8
-// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 0
-// CHECK1-NEXT:    [[TMP21:%.*]] = inttoptr i64 [[TMP15]] to ptr
-// CHECK1-NEXT:    store ptr [[TMP21]], ptr [[TMP20]], align 8
-// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 1
-// CHECK1-NEXT:    [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to ptr
-// CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP22]], align 8
-// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 2
-// CHECK1-NEXT:    [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to ptr
-// CHECK1-NEXT:    store ptr [[TMP25]], ptr [[TMP24]], align 8
-// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i64 0, i64 3
-// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[TMP26]], align 8
-// CHECK1-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
-// CHECK1-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP28]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i64 4)
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
-// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
-// CHECK1-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP35]], [[TMP36]]
-// CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
-// CHECK1:       cond.true10:
-// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    br label [[COND_END12:%.*]]
-// CHECK1:       cond.false11:
-// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END12]]
-// CHECK1:       cond.end12:
-// CHECK1-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP37]], [[COND_TRUE10]] ], [ [[TMP38]], [[COND_FALSE11]] ]
-// CHECK1-NEXT:    store i32 [[COND13]], ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP39]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP40:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, ptr [[TMP40]], align 4
-// CHECK1-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP41]])
-// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK1:       omp.precond.end:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined_omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[ARGC]], ptr [[ARGC_ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
-// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
-// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK1:       omp.precond.then:
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP5]] to i32
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP6]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP8]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[CONV5]], [[TMP11]]
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
-// CHECK1-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[I4]]) #[[ATTR5:[0-9]+]]
-// CHECK1-NEXT:    [[CALL7:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[TMP0]]) #[[ATTR5]]
+// CHECK1-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[I4]]) #[[ATTR4:[0-9]+]]
+// CHECK1-NEXT:    [[CALL7:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[TMP0]]) #[[ATTR4]]
 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[CALL]], [[CALL7]]
-// CHECK1-NEXT:    [[CALL9:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[ARGC_ADDR]]) #[[ATTR5]]
+// CHECK1-NEXT:    [[CALL9:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[ARGC_ADDR]]) #[[ATTR4]]
 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[ADD8]], [[CALL9]]
 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[TMP0]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP15]], 1
 // CHECK1-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP16]])
+// CHECK1-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
+// CHECK1-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24
-// CHECK2-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK2-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[ARGC_CASTED:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_kernel_environment, ptr [[DYN_PTR]])
 // CHECK2-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
@@ -302,9 +186,14 @@ int main(int argc, char **argv) {
 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP3]], ptr [[ARGC_CASTED]], align 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[ARGC_CASTED]], align 4
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK2-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP5]] to i1
+// CHECK2-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK2-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
 // CHECK2-NEXT:    store i32 0, ptr [[DOTZERO_ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP_]], align 4
-// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP4]], ptr [[TMP0]]) #[[ATTR2:[0-9]+]]
+// CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]], i32 [[TMP4]], ptr [[TMP0]], i32 [[TMP6]]) #[[ATTR2:[0-9]+]]
 // CHECK2-NEXT:    call void @__kmpc_target_deinit()
 // CHECK2-NEXT:    ret void
 // CHECK2:       worker.exit:
@@ -312,56 +201,55 @@ int main(int argc, char **argv) {
 //
 //
 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined
-// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
 // CHECK2-NEXT:  entry:
 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[ARGC_CASTED:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 4
+// CHECK2-NEXT:    [[I4:%.*]] = alloca i32, align 4
 // CHECK2-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK2-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 // CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK2-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK2-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 // CHECK2:       omp.precond.then:
 // CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK2-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
 // CHECK2-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK2-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[NVPTX_NUM_THREADS]])
+// CHECK2-NEXT:    call void @__kmpc_distribute_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
-// CHECK2-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP7]], [[TMP8]]
+// CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK2:       cond.true:
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 // CHECK2-NEXT:    br label [[COND_END:%.*]]
 // CHECK2:       cond.false:
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
@@ -374,155 +262,34 @@ int main(int argc, char **argv) {
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK2:       omp.inner.for.cond:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], 1
-// CHECK2-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP12]], [[ADD]]
-// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP16]], ptr [[ARGC_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, ptr [[ARGC_CASTED]], align 4
-// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 0
-// CHECK2-NEXT:    [[TMP19:%.*]] = inttoptr i32 [[TMP14]] to ptr
-// CHECK2-NEXT:    store ptr [[TMP19]], ptr [[TMP18]], align 4
-// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 1
-// CHECK2-NEXT:    [[TMP21:%.*]] = inttoptr i32 [[TMP15]] to ptr
-// CHECK2-NEXT:    store ptr [[TMP21]], ptr [[TMP20]], align 4
-// CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 2
-// CHECK2-NEXT:    [[TMP23:%.*]] = inttoptr i32 [[TMP17]] to ptr
-// CHECK2-NEXT:    store ptr [[TMP23]], ptr [[TMP22]], align 4
-// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS]], i32 0, i32 3
-// CHECK2-NEXT:    store ptr [[TMP0]], ptr [[TMP24]], align 4
-// CHECK2-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
-// CHECK2-NEXT:    call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP26]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS]], i32 4)
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
-// CHECK2-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP29]], [[TMP30]]
-// CHECK2-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
-// CHECK2-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP33]], [[TMP34]]
-// CHECK2-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
-// CHECK2:       cond.true10:
-// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    br label [[COND_END12:%.*]]
-// CHECK2:       cond.false11:
-// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    br label [[COND_END12]]
-// CHECK2:       cond.end12:
-// CHECK2-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP35]], [[COND_TRUE10]] ], [ [[TMP36]], [[COND_FALSE11]] ]
-// CHECK2-NEXT:    store i32 [[COND13]], ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP37]], ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK2:       omp.inner.for.end:
-// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK2:       omp.loop.exit:
-// CHECK2-NEXT:    [[TMP38:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, ptr [[TMP38]], align 4
-// CHECK2-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP39]])
-// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK2:       omp.precond.end:
-// CHECK2-NEXT:    ret void
-//
-//
-// CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l24_omp_outlined_omp_outlined
-// CHECK2-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[ARGC:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1]] {
-// CHECK2-NEXT:  entry:
-// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK2-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
-// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK2-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP3]]
-// CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK2:       omp.precond.then:
-// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK2-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK2-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
-// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP8]], i32 33, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK2-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK2:       omp.inner.for.cond:
-// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK2-NEXT:    [[CMP4:%.*]] = icmp ule i32 [[TMP10]], [[TMP11]]
-// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK2:       omp.inner.for.body:
-// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK2-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
-// CHECK2-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[I3]]) #[[ATTR5:[0-9]+]]
-// CHECK2-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[TMP0]]) #[[ATTR5]]
-// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CALL]], [[CALL5]]
-// CHECK2-NEXT:    [[CALL7:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[ARGC_ADDR]]) #[[ATTR5]]
-// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[ADD6]], [[CALL7]]
-// CHECK2-NEXT:    store i32 [[ADD8]], ptr [[TMP0]], align 4
+// CHECK2-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
+// CHECK2-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[I4]]) #[[ATTR4:[0-9]+]]
+// CHECK2-NEXT:    [[CALL7:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[TMP0]]) #[[ATTR4]]
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[CALL]], [[CALL7]]
+// CHECK2-NEXT:    [[CALL9:%.*]] = call noundef i32 @_Z3fooPi(ptr noundef [[ARGC_ADDR]]) #[[ATTR4]]
+// CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[ADD8]], [[CALL9]]
+// CHECK2-NEXT:    store i32 [[ADD10]], ptr [[TMP0]], align 4
 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK2:       omp.body.continue:
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK2:       omp.inner.for.inc:
-// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
-// CHECK2-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK2-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK2:       omp.inner.for.end:
 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK2:       omp.loop.exit:
-// CHECK2-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP16]])
+// CHECK2-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
+// CHECK2-NEXT:    call void @__kmpc_distribute_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK2:       omp.precond.end:
 // CHECK2-NEXT:    ret void
diff --git a/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp b/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
index 22cf534bf0ba2..3f752ac663f41 100644
--- a/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
+++ b/clang/test/OpenMP/target_teams_generic_loop_codegen.cpp
@@ -28,1118 +28,6 @@ int foo() {
   return 0;
 }
 #endif
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@_Z3foov
-// IR-PCH-HOST-SAME: () #[[ATTR0:[0-9]+]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[I:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-PCH-HOST-NEXT:    [[J_CASTED:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[TMP0:%.*]] = load i32, ptr [[J]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP0]], ptr [[J_CASTED]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP1:%.*]] = load i64, ptr [[J_CASTED]], align 8
-// IR-PCH-HOST-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22(i64 [[TMP1]], ptr [[SUM]]) #[[ATTR2:[0-9]+]]
-// IR-PCH-HOST-NEXT:    ret i32 0
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22
-// IR-PCH-HOST-SAME: (i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1:[0-9]+]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[J_CASTED:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    store i64 [[J]], ptr [[J_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP1:%.*]] = load i32, ptr [[J_ADDR]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP1]], ptr [[J_CASTED]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP2:%.*]] = load i64, ptr [[J_CASTED]], align 8
-// IR-PCH-HOST-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 2, ptr @.omp_outlined., i64 [[TMP2]], ptr [[TMP0]])
-// IR-PCH-HOST-NEXT:    ret void
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@.omp_outlined.
-// IR-PCH-HOST-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[SUM1:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-PCH-HOST-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J3:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[I:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J4:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J_CASTED:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store i64 [[J]], ptr [[J_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i32 0, i32 0, i32 0
-// IR-PCH-HOST-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP1]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// IR-PCH-HOST:       omp.arrayinit.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// IR-PCH-HOST:       omp.arrayinit.done:
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// IR-PCH-HOST-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// IR-PCH-HOST-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// IR-PCH-HOST-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// IR-PCH-HOST:       cond.true:
-// IR-PCH-HOST-NEXT:    br label [[COND_END:%.*]]
-// IR-PCH-HOST:       cond.false:
-// IR-PCH-HOST-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    br label [[COND_END]]
-// IR-PCH-HOST:       cond.end:
-// IR-PCH-HOST-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// IR-PCH-HOST-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// IR-PCH-HOST:       omp.inner.for.cond:
-// IR-PCH-HOST-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// IR-PCH-HOST-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// IR-PCH-HOST:       omp.inner.for.body:
-// IR-PCH-HOST-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// IR-PCH-HOST-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
-// IR-PCH-HOST-NEXT:    [[TMP13:%.*]] = load i32, ptr [[J3]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP13]], ptr [[J_CASTED]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP14:%.*]] = load i64, ptr [[J_CASTED]], align 8
-// IR-PCH-HOST-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 4, ptr @.omp_outlined..1, i64 [[TMP10]], i64 [[TMP12]], i64 [[TMP14]], ptr [[SUM1]])
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// IR-PCH-HOST:       omp.inner.for.inc:
-// IR-PCH-HOST-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// IR-PCH-HOST-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// IR-PCH-HOST:       omp.inner.for.end:
-// IR-PCH-HOST-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// IR-PCH-HOST:       omp.loop.exit:
-// IR-PCH-HOST-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
-// IR-PCH-HOST-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP18]])
-// IR-PCH-HOST-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
-// IR-PCH-HOST-NEXT:    br i1 [[TMP20]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// IR-PCH-HOST:       .omp.lastprivate.then:
-// IR-PCH-HOST-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP21:%.*]] = load i32, ptr [[J3]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP21]], ptr [[J_ADDR]], align 4
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// IR-PCH-HOST:       .omp.lastprivate.done:
-// IR-PCH-HOST-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    store ptr [[SUM1]], ptr [[TMP22]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3:[0-9]+]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @.omp.reduction.reduction_func.2, ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// IR-PCH-HOST-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// IR-PCH-HOST-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// IR-PCH-HOST-NEXT:    ]
-// IR-PCH-HOST:       .omp.reduction.case1:
-// IR-PCH-HOST-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP0]], [[TMP26]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD7]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP26]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH-HOST:       omp.arraycpy.done10:
-// IR-PCH-HOST-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH-HOST:       .omp.reduction.case2:
-// IR-PCH-HOST-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq ptr [[TMP0]], [[TMP29]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body12:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-PCH-HOST-NEXT:    [[TMP30:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP31:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 [[TMP30]] monotonic, align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP29]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
-// IR-PCH-HOST:       omp.arraycpy.done18:
-// IR-PCH-HOST-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH-HOST:       .omp.reduction.default:
-// IR-PCH-HOST-NEXT:    ret void
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@.omp_outlined..1
-// IR-PCH-HOST-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-HOST-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J3:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-PCH-HOST-NEXT:    [[I:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[J5:%.*]] = alloca i32, align 4
-// IR-PCH-HOST-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store i64 [[J]], ptr [[J_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// IR-PCH-HOST-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
-// IR-PCH-HOST-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-HOST-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i32 0, i32 0, i32 0
-// IR-PCH-HOST-NEXT:    [[TMP3:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// IR-PCH-HOST:       omp.arrayinit.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// IR-PCH-HOST-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// IR-PCH-HOST:       omp.arrayinit.done:
-// IR-PCH-HOST-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
-// IR-PCH-HOST-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP5]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// IR-PCH-HOST-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 99
-// IR-PCH-HOST-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// IR-PCH-HOST:       cond.true:
-// IR-PCH-HOST-NEXT:    br label [[COND_END:%.*]]
-// IR-PCH-HOST:       cond.false:
-// IR-PCH-HOST-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-HOST-NEXT:    br label [[COND_END]]
-// IR-PCH-HOST:       cond.end:
-// IR-PCH-HOST-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
-// IR-PCH-HOST-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV]], align 4
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// IR-PCH-HOST:       omp.inner.for.cond:
-// IR-PCH-HOST-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
-// IR-PCH-HOST-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
-// IR-PCH-HOST-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// IR-PCH-HOST:       omp.inner.for.body:
-// IR-PCH-HOST-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP11]], 10
-// IR-PCH-HOST-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
-// IR-PCH-HOST-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP13]], 10
-// IR-PCH-HOST-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
-// IR-PCH-HOST-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP12]], [[MUL8]]
-// IR-PCH-HOST-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
-// IR-PCH-HOST-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD10]], ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
-// IR-PCH-HOST-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i64 0, i64 [[IDXPROM]]
-// IR-PCH-HOST-NEXT:    [[TMP16:%.*]] = load i32, ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP16]] to i64
-// IR-PCH-HOST-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
-// IR-PCH-HOST-NEXT:    [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP17]], [[TMP14]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// IR-PCH-HOST:       omp.body.continue:
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// IR-PCH-HOST:       omp.inner.for.inc:
-// IR-PCH-HOST-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP18]], 1
-// IR-PCH-HOST-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-HOST-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
-// IR-PCH-HOST:       omp.inner.for.end:
-// IR-PCH-HOST-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// IR-PCH-HOST:       omp.loop.exit:
-// IR-PCH-HOST-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
-// IR-PCH-HOST-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP20]])
-// IR-PCH-HOST-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    store ptr [[SUM4]], ptr [[TMP21]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// IR-PCH-HOST-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// IR-PCH-HOST-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// IR-PCH-HOST-NEXT:    ]
-// IR-PCH-HOST:       .omp.reduction.case1:
-// IR-PCH-HOST-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP0]], [[TMP25]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP25]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH-HOST:       omp.arraycpy.done19:
-// IR-PCH-HOST-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH-HOST:       .omp.reduction.case2:
-// IR-PCH-HOST-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY20:%.*]] = icmp eq ptr [[TMP0]], [[TMP28]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY20]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY21:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body21:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST22:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST23:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-PCH-HOST-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP30:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 [[TMP29]] monotonic, align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP28]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
-// IR-PCH-HOST:       omp.arraycpy.done27:
-// IR-PCH-HOST-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH-HOST:       .omp.reduction.default:
-// IR-PCH-HOST-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
-// IR-PCH-HOST-NEXT:    br i1 [[TMP32]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// IR-PCH-HOST:       .omp.lastprivate.then:
-// IR-PCH-HOST-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP33:%.*]] = load i32, ptr [[J3]], align 4
-// IR-PCH-HOST-NEXT:    store i32 [[TMP33]], ptr [[J_ADDR]], align 4
-// IR-PCH-HOST-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// IR-PCH-HOST:       .omp.lastprivate.done:
-// IR-PCH-HOST-NEXT:    ret void
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
-// IR-PCH-HOST-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH-HOST:       omp.arraycpy.done2:
-// IR-PCH-HOST-NEXT:    ret void
-// IR-PCH-HOST-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
-// IR-PCH-HOST-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
-// IR-PCH-HOST-NEXT:  entry:
-// IR-PCH-HOST-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// IR-PCH-HOST-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// IR-PCH-HOST-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// IR-PCH-HOST-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// IR-PCH-HOST-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH-HOST:       omp.arraycpy.body:
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-HOST-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// IR-PCH-HOST-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-HOST-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
-// IR-PCH-HOST-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH-HOST:       omp.arraycpy.done2:
-// IR-PCH-HOST-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l23
-// CHECK-SAME: (i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR0:[0-9]+]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[J_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
-// CHECK-NEXT:    [[SUM_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM_ADDR]] to ptr
-// CHECK-NEXT:    [[J_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_CASTED]] to ptr
-// CHECK-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
-// CHECK-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
-// CHECK-NEXT:    store i64 [[J]], ptr [[J_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr), i8 2, i1 false)
-// CHECK-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP1]], -1
-// CHECK-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
-// CHECK:       user_code.entry:
-// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
-// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[J_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP3]], ptr [[J_CASTED_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr [[J_CASTED_ASCAST]], align 8
-// CHECK-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP2]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
-// CHECK-NEXT:    call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP4]], ptr [[TMP0]]) #[[ATTR2:[0-9]+]]
-// CHECK-NEXT:    call void @__kmpc_target_deinit(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i8 2)
-// CHECK-NEXT:    ret void
-// CHECK:       worker.exit:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@__omp_outlined__
-// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1:[0-9]+]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[SUM1:%.*]] = alloca [10 x [10 x i32]], align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J3:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J4:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [4 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
-// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
-// CHECK-NEXT:    [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
-// CHECK-NEXT:    [[SUM_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM_ADDR]] to ptr
-// CHECK-NEXT:    [[SUM1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM1]] to ptr
-// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
-// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
-// CHECK-NEXT:    [[TMP2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP2]] to ptr
-// CHECK-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
-// CHECK-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
-// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
-// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
-// CHECK-NEXT:    [[J3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J3]] to ptr
-// CHECK-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
-// CHECK-NEXT:    [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
-// CHECK-NEXT:    [[J_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_CASTED]] to ptr
-// CHECK-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
-// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i64 [[J]], ptr [[J_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1_ASCAST]], i32 0, i32 0, i32 0
-// CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// CHECK-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP1]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// CHECK:       omp.arrayinit.body:
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// CHECK-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// CHECK:       omp.arrayinit.done:
-// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
-// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
-// CHECK-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
-// CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP3]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK:       cond.true:
-// CHECK-NEXT:    br label [[COND_END:%.*]]
-// CHECK:       cond.false:
-// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    br label [[COND_END]]
-// CHECK:       cond.end:
-// CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK:       omp.inner.for.cond:
-// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[TMP7]], 100
-// CHECK-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK:       omp.inner.for.body:
-// CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J3_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP12]], ptr [[J_CASTED_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP13:%.*]] = load i64, ptr [[J_CASTED_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP15:%.*]] = inttoptr i64 [[TMP9]] to ptr
-// CHECK-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 8
-// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP17:%.*]] = inttoptr i64 [[TMP11]] to ptr
-// CHECK-NEXT:    store ptr [[TMP17]], ptr [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP19:%.*]] = inttoptr i64 [[TMP13]] to ptr
-// CHECK-NEXT:    store ptr [[TMP19]], ptr [[TMP18]], align 8
-// CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
-// CHECK-NEXT:    store ptr [[SUM1_ASCAST]], ptr [[TMP20]], align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP22]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__.1, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 4)
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK:       omp.inner.for.inc:
-// CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
-// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
-// CHECK-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
-// CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
-// CHECK-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
-// CHECK-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
-// CHECK-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[CMP8:%.*]] = icmp sgt i32 [[TMP29]], 99
-// CHECK-NEXT:    br i1 [[CMP8]], label [[COND_TRUE9:%.*]], label [[COND_FALSE10:%.*]]
-// CHECK:       cond.true9:
-// CHECK-NEXT:    br label [[COND_END11:%.*]]
-// CHECK:       cond.false10:
-// CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    br label [[COND_END11]]
-// CHECK:       cond.end11:
-// CHECK-NEXT:    [[COND12:%.*]] = phi i32 [ 99, [[COND_TRUE9]] ], [ [[TMP30]], [[COND_FALSE10]] ]
-// CHECK-NEXT:    store i32 [[COND12]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP31]], ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK:       omp.inner.for.end:
-// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK:       omp.loop.exit:
-// CHECK-NEXT:    [[TMP32:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP33:%.*]] = load i32, ptr [[TMP32]], align 4
-// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3:[0-9]+]] to ptr), i32 [[TMP33]])
-// CHECK-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
-// CHECK-NEXT:    br i1 [[TMP35]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// CHECK:       .omp.lastprivate.then:
-// CHECK-NEXT:    store i32 10, ptr [[J3_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP36:%.*]] = load i32, ptr [[J3_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP36]], ptr [[J_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// CHECK:       .omp.lastprivate.done:
-// CHECK-NEXT:    [[TMP37:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP38:%.*]] = load i32, ptr [[TMP37]], align 4
-// CHECK-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    store ptr [[SUM1_ASCAST]], ptr [[TMP39]], align 8
-// CHECK-NEXT:    [[TMP40:%.*]] = load ptr, ptr addrspace(1) @"_openmp_teams_reductions_buffer_$_$ptr", align 8
-// CHECK-NEXT:    [[TMP41:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP38]], ptr [[TMP40]], i32 1024, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.3, ptr @_omp_reduction_inter_warp_copy_func.4, ptr @_omp_reduction_list_to_global_copy_func, ptr @_omp_reduction_list_to_global_reduce_func, ptr @_omp_reduction_global_to_list_copy_func, ptr @_omp_reduction_global_to_list_reduce_func)
-// CHECK-NEXT:    [[TMP42:%.*]] = icmp eq i32 [[TMP41]], 1
-// CHECK-NEXT:    br i1 [[TMP42]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
-// CHECK:       .omp.reduction.then:
-// CHECK-NEXT:    [[TMP43:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// CHECK-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP0]], [[TMP43]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// CHECK:       omp.arraycpy.body:
-// CHECK-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1_ASCAST]], [[DOTOMP_REDUCTION_THEN]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST13:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_THEN]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// CHECK-NEXT:    [[TMP44:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST13]], align 4
-// CHECK-NEXT:    [[TMP45:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// CHECK-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP44]], [[TMP45]]
-// CHECK-NEXT:    store i32 [[ADD14]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST13]], align 4
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST13]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP43]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
-// CHECK:       omp.arraycpy.done17:
-// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
-// CHECK:       .omp.reduction.done:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@__omp_outlined__.1
-// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[J_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
-// CHECK-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J3:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 4, addrspace(5)
-// CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[J5:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
-// CHECK-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
-// CHECK-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
-// CHECK-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
-// CHECK-NEXT:    [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
-// CHECK-NEXT:    [[SUM_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM_ADDR]] to ptr
-// CHECK-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
-// CHECK-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
-// CHECK-NEXT:    [[TMP1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP1]] to ptr
-// CHECK-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
-// CHECK-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
-// CHECK-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
-// CHECK-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
-// CHECK-NEXT:    [[J3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J3]] to ptr
-// CHECK-NEXT:    [[SUM4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM4]] to ptr
-// CHECK-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
-// CHECK-NEXT:    [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
-// CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i64 [[J]], ptr [[J_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SUM_ADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 99, ptr [[DOTOMP_UB_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
-// CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
-// CHECK-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4_ASCAST]], i32 0, i32 0, i32 0
-// CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// CHECK-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// CHECK:       omp.arrayinit.body:
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// CHECK-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// CHECK:       omp.arrayinit.done:
-// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
-// CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP5]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
-// CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV_ASCAST]], align 4
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK:       omp.inner.for.cond:
-// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
-// CHECK-NEXT:    [[CONV6:%.*]] = sext i32 [[TMP7]] to i64
-// CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[CMP:%.*]] = icmp ule i64 [[CONV6]], [[TMP8]]
-// CHECK-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK:       omp.inner.for.body:
-// CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP9]], 10
-// CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
-// CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK-NEXT:    store i32 [[ADD]], ptr [[I_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP11]], 10
-// CHECK-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
-// CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP10]], [[MUL8]]
-// CHECK-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
-// CHECK-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
-// CHECK-NEXT:    store i32 [[ADD10]], ptr [[J3_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4_ASCAST]], i64 0, i64 [[IDXPROM]]
-// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J3_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP14]] to i64
-// CHECK-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
-// CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP15]], [[TMP12]]
-// CHECK-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// CHECK:       omp.body.continue:
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK:       omp.inner.for.inc:
-// CHECK-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV_ASCAST]], align 4, !llvm.access.group [[ACC_GRP7]]
-// CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
-// CHECK:       omp.inner.for.end:
-// CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK:       omp.loop.exit:
-// CHECK-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
-// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP19]])
-// CHECK-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    store ptr [[SUM4_ASCAST]], ptr [[TMP22]], align 8
-// CHECK-NEXT:    [[TMP23:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP21]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func, ptr @_omp_reduction_inter_warp_copy_func)
-// CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i32 [[TMP23]], 1
-// CHECK-NEXT:    br i1 [[TMP24]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
-// CHECK:       .omp.reduction.then:
-// CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
-// CHECK-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP0]], [[TMP25]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// CHECK:       omp.arraycpy.body:
-// CHECK-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4_ASCAST]], [[DOTOMP_REDUCTION_THEN]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP0]], [[DOTOMP_REDUCTION_THEN]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// CHECK-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// CHECK-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
-// CHECK-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// CHECK-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP25]]
-// CHECK-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
-// CHECK:       omp.arraycpy.done19:
-// CHECK-NEXT:    br label [[DOTOMP_REDUCTION_DONE]]
-// CHECK:       .omp.reduction.done:
-// CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
-// CHECK-NEXT:    br i1 [[TMP29]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// CHECK:       .omp.lastprivate.then:
-// CHECK-NEXT:    store i32 10, ptr [[J3_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[J3_ASCAST]], align 4
-// CHECK-NEXT:    store i32 [[TMP30]], ptr [[J_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// CHECK:       .omp.lastprivate.done:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR3:[0-9]+]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca [10 x [10 x i32]], align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
-// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
-// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr [10 x [10 x i32]], ptr [[TMP9]], i64 1
-// CHECK-NEXT:    br label [[DOTSHUFFLE_PRE_COND:%.*]]
-// CHECK:       .shuffle.pre_cond:
-// CHECK-NEXT:    [[TMP12:%.*]] = phi ptr [ [[TMP9]], [[ENTRY:%.*]] ], [ [[TMP23:%.*]], [[DOTSHUFFLE_THEN:%.*]] ]
-// CHECK-NEXT:    [[TMP13:%.*]] = phi ptr [ [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], [[ENTRY]] ], [ [[TMP24:%.*]], [[DOTSHUFFLE_THEN]] ]
-// CHECK-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[TMP11]] to i64
-// CHECK-NEXT:    [[TMP15:%.*]] = ptrtoint ptr [[TMP12]] to i64
-// CHECK-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
-// CHECK-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp sgt i64 [[TMP17]], 7
-// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTSHUFFLE_THEN]], label [[DOTSHUFFLE_EXIT:%.*]]
-// CHECK:       .shuffle.then:
-// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr [[TMP12]], align 4
-// CHECK-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_get_warp_size()
-// CHECK-NEXT:    [[TMP21:%.*]] = trunc i32 [[TMP20]] to i16
-// CHECK-NEXT:    [[TMP22:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP19]], i16 [[TMP6]], i16 [[TMP21]])
-// CHECK-NEXT:    store i64 [[TMP22]], ptr [[TMP13]], align 4
-// CHECK-NEXT:    [[TMP23]] = getelementptr i64, ptr [[TMP12]], i64 1
-// CHECK-NEXT:    [[TMP24]] = getelementptr i64, ptr [[TMP13]], i64 1
-// CHECK-NEXT:    br label [[DOTSHUFFLE_PRE_COND]]
-// CHECK:       .shuffle.exit:
-// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i16 [[TMP7]], 0
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP7]], 1
-// CHECK-NEXT:    [[TMP27:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
-// CHECK-NEXT:    [[TMP28:%.*]] = and i1 [[TMP26]], [[TMP27]]
-// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP7]], 2
-// CHECK-NEXT:    [[TMP30:%.*]] = and i16 [[TMP5]], 1
-// CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i16 [[TMP30]], 0
-// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP29]], [[TMP31]]
-// CHECK-NEXT:    [[TMP33:%.*]] = icmp sgt i16 [[TMP6]], 0
-// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
-// CHECK-NEXT:    [[TMP35:%.*]] = or i1 [[TMP25]], [[TMP28]]
-// CHECK-NEXT:    [[TMP36:%.*]] = or i1 [[TMP35]], [[TMP34]]
-// CHECK-NEXT:    br i1 [[TMP36]], label [[THEN:%.*]], label [[ELSE:%.*]]
-// CHECK:       then:
-// CHECK-NEXT:    call void @"_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
-// CHECK-NEXT:    br label [[IFCONT:%.*]]
-// CHECK:       else:
-// CHECK-NEXT:    br label [[IFCONT]]
-// CHECK:       ifcont:
-// CHECK-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP7]], 1
-// CHECK-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
-// CHECK-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
-// CHECK-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
-// CHECK:       then4:
-// CHECK-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP41:%.*]] = load ptr, ptr [[TMP40]], align 8
-// CHECK-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP43:%.*]] = load ptr, ptr [[TMP42]], align 8
-// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP43]], ptr align 4 [[TMP41]], i64 400, i1 false)
-// CHECK-NEXT:    br label [[IFCONT6:%.*]]
-// CHECK:       else5:
-// CHECK-NEXT:    br label [[IFCONT6]]
-// CHECK:       ifcont6:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP4]], 63
-// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP5]], 6
-// CHECK-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[PRECOND:%.*]]
-// CHECK:       precond:
-// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP8:%.*]] = icmp ult i32 [[TMP7]], 100
-// CHECK-NEXT:    br i1 [[TMP8]], label [[BODY:%.*]], label [[EXIT:%.*]]
-// CHECK:       body:
-// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4:[0-9]+]] to ptr), i32 [[TMP2]])
-// CHECK-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
-// CHECK-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
-// CHECK:       then:
-// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP6]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[TMP9]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 [[TMP7]]
-// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
-// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK-NEXT:    store volatile i32 [[TMP13]], ptr addrspace(3) [[TMP12]], align 4
-// CHECK-NEXT:    br label [[IFCONT:%.*]]
-// CHECK:       else:
-// CHECK-NEXT:    br label [[IFCONT]]
-// CHECK:       ifcont:
-// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[TMP2]])
-// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP3]], [[TMP14]]
-// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN2:%.*]], label [[ELSE3:%.*]]
-// CHECK:       then2:
-// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP3]]
-// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP6]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i32, ptr [[TMP17]], i32 [[TMP7]]
-// CHECK-NEXT:    [[TMP19:%.*]] = load volatile i32, ptr addrspace(3) [[TMP15]], align 4
-// CHECK-NEXT:    store i32 [[TMP19]], ptr [[TMP18]], align 4
-// CHECK-NEXT:    br label [[IFCONT4:%.*]]
-// CHECK:       else3:
-// CHECK-NEXT:    br label [[IFCONT4]]
-// CHECK:       ifcont4:
-// CHECK-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP7]], 1
-// CHECK-NEXT:    store i32 [[TMP20]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[PRECOND]]
-// CHECK:       exit:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.3
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca [10 x [10 x i32]], align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
-// CHECK-NEXT:    store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
-// CHECK-NEXT:    store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
-// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr [10 x [10 x i32]], ptr [[TMP9]], i64 1
-// CHECK-NEXT:    br label [[DOTSHUFFLE_PRE_COND:%.*]]
-// CHECK:       .shuffle.pre_cond:
-// CHECK-NEXT:    [[TMP12:%.*]] = phi ptr [ [[TMP9]], [[ENTRY:%.*]] ], [ [[TMP23:%.*]], [[DOTSHUFFLE_THEN:%.*]] ]
-// CHECK-NEXT:    [[TMP13:%.*]] = phi ptr [ [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], [[ENTRY]] ], [ [[TMP24:%.*]], [[DOTSHUFFLE_THEN]] ]
-// CHECK-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[TMP11]] to i64
-// CHECK-NEXT:    [[TMP15:%.*]] = ptrtoint ptr [[TMP12]] to i64
-// CHECK-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
-// CHECK-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp sgt i64 [[TMP17]], 7
-// CHECK-NEXT:    br i1 [[TMP18]], label [[DOTSHUFFLE_THEN]], label [[DOTSHUFFLE_EXIT:%.*]]
-// CHECK:       .shuffle.then:
-// CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr [[TMP12]], align 4
-// CHECK-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_get_warp_size()
-// CHECK-NEXT:    [[TMP21:%.*]] = trunc i32 [[TMP20]] to i16
-// CHECK-NEXT:    [[TMP22:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP19]], i16 [[TMP6]], i16 [[TMP21]])
-// CHECK-NEXT:    store i64 [[TMP22]], ptr [[TMP13]], align 4
-// CHECK-NEXT:    [[TMP23]] = getelementptr i64, ptr [[TMP12]], i64 1
-// CHECK-NEXT:    [[TMP24]] = getelementptr i64, ptr [[TMP13]], i64 1
-// CHECK-NEXT:    br label [[DOTSHUFFLE_PRE_COND]]
-// CHECK:       .shuffle.exit:
-// CHECK-NEXT:    store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i16 [[TMP7]], 0
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i16 [[TMP7]], 1
-// CHECK-NEXT:    [[TMP27:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
-// CHECK-NEXT:    [[TMP28:%.*]] = and i1 [[TMP26]], [[TMP27]]
-// CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i16 [[TMP7]], 2
-// CHECK-NEXT:    [[TMP30:%.*]] = and i16 [[TMP5]], 1
-// CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i16 [[TMP30]], 0
-// CHECK-NEXT:    [[TMP32:%.*]] = and i1 [[TMP29]], [[TMP31]]
-// CHECK-NEXT:    [[TMP33:%.*]] = icmp sgt i16 [[TMP6]], 0
-// CHECK-NEXT:    [[TMP34:%.*]] = and i1 [[TMP32]], [[TMP33]]
-// CHECK-NEXT:    [[TMP35:%.*]] = or i1 [[TMP25]], [[TMP28]]
-// CHECK-NEXT:    [[TMP36:%.*]] = or i1 [[TMP35]], [[TMP34]]
-// CHECK-NEXT:    br i1 [[TMP36]], label [[THEN:%.*]], label [[ELSE:%.*]]
-// CHECK:       then:
-// CHECK-NEXT:    call void @"_omp$reduction$reduction_func.2"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
-// CHECK-NEXT:    br label [[IFCONT:%.*]]
-// CHECK:       else:
-// CHECK-NEXT:    br label [[IFCONT]]
-// CHECK:       ifcont:
-// CHECK-NEXT:    [[TMP37:%.*]] = icmp eq i16 [[TMP7]], 1
-// CHECK-NEXT:    [[TMP38:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
-// CHECK-NEXT:    [[TMP39:%.*]] = and i1 [[TMP37]], [[TMP38]]
-// CHECK-NEXT:    br i1 [[TMP39]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
-// CHECK:       then4:
-// CHECK-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP41:%.*]] = load ptr, ptr [[TMP40]], align 8
-// CHECK-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP43:%.*]] = load ptr, ptr [[TMP42]], align 8
-// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP43]], ptr align 4 [[TMP41]], i64 400, i1 false)
-// CHECK-NEXT:    br label [[IFCONT6:%.*]]
-// CHECK:       else5:
-// CHECK-NEXT:    br label [[IFCONT6]]
-// CHECK:       ifcont6:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.4
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP4]], 63
-// CHECK-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
-// CHECK-NEXT:    [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP5]], 6
-// CHECK-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[PRECOND:%.*]]
-// CHECK:       precond:
-// CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP8:%.*]] = icmp ult i32 [[TMP7]], 100
-// CHECK-NEXT:    br i1 [[TMP8]], label [[BODY:%.*]], label [[EXIT:%.*]]
-// CHECK:       body:
-// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[TMP2]])
-// CHECK-NEXT:    [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
-// CHECK-NEXT:    br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
-// CHECK:       then:
-// CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP6]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[TMP9]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 [[TMP7]]
-// CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
-// CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK-NEXT:    store volatile i32 [[TMP13]], ptr addrspace(3) [[TMP12]], align 4
-// CHECK-NEXT:    br label [[IFCONT:%.*]]
-// CHECK:       else:
-// CHECK-NEXT:    br label [[IFCONT]]
-// CHECK:       ifcont:
-// CHECK-NEXT:    call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[TMP2]])
-// CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP3]], [[TMP14]]
-// CHECK-NEXT:    br i1 [[IS_ACTIVE_THREAD]], label [[THEN2:%.*]], label [[ELSE3:%.*]]
-// CHECK:       then2:
-// CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP3]]
-// CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP6]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i32, ptr [[TMP17]], i32 [[TMP7]]
-// CHECK-NEXT:    [[TMP19:%.*]] = load volatile i32, ptr addrspace(3) [[TMP15]], align 4
-// CHECK-NEXT:    store i32 [[TMP19]], ptr [[TMP18]], align 4
-// CHECK-NEXT:    br label [[IFCONT4:%.*]]
-// CHECK:       else3:
-// CHECK-NEXT:    br label [[IFCONT4]]
-// CHECK:       ifcont4:
-// CHECK-NEXT:    [[TMP20:%.*]] = add nsw i32 [[TMP7]], 1
-// CHECK-NEXT:    store i32 [[TMP20]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
-// CHECK-NEXT:    br label [[PRECOND]]
-// CHECK:       exit:
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// CHECK-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0
-// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP5]]
-// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[TMP8]], ptr align 4 [[TMP7]], i64 400, i1 false)
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 0, i32 0
-// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP4]]
-// CHECK-NEXT:    store ptr [[TMP6]], ptr [[TMP5]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    call void @"_omp$reduction$reduction_func.2"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP7]]) #[[ATTR2]]
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// CHECK-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0
-// CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP5]]
-// CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP7]], ptr align 128 [[TMP8]], i64 400, i1 false)
-// CHECK-NEXT:    ret void
-// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func
-// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR3]] {
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
-// CHECK-NEXT:    [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
-// CHECK-NEXT:    [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
-// CHECK-NEXT:    [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
-// CHECK-NEXT:    [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
-// CHECK-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
-// CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
-// CHECK-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 0, i32 0
-// CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP4]]
-// CHECK-NEXT:    store ptr [[TMP6]], ptr [[TMP5]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
-// CHECK-NEXT:    call void @"_omp$reduction$reduction_func.2"(ptr [[TMP7]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
-// CHECK-NEXT:    ret void
 // IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22
 // IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR0:[0-9]+]] {
 // IR-GPU-NEXT:  entry:
@@ -2015,7 +903,7 @@ int foo() {
 // IR-NEXT:    store ptr [[SUM1]], ptr [[TMP22]], align 8
 // IR-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
-// IR-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3:[0-9]+]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
@@ -2036,7 +924,7 @@ int foo() {
 // IR-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP26]]
 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
 // IR:       omp.arraycpy.done10:
-// IR-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.case2:
 // IR-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
@@ -2052,7 +940,6 @@ int foo() {
 // IR-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP29]]
 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
 // IR:       omp.arraycpy.done18:
-// IR-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.default:
 // IR-NEXT:    ret void
@@ -2171,7 +1058,7 @@ int foo() {
 // IR-NEXT:    store ptr [[SUM4]], ptr [[TMP21]], align 8
 // IR-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// IR-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
@@ -2192,7 +1079,7 @@ int foo() {
 // IR-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP25]]
 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
 // IR:       omp.arraycpy.done19:
-// IR-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.case2:
 // IR-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
@@ -2208,7 +1095,6 @@ int foo() {
 // IR-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP28]]
 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
 // IR:       omp.arraycpy.done27:
-// IR-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.default:
 // IR-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
@@ -2412,7 +1298,7 @@ int foo() {
 // IR-PCH-NEXT:    store ptr [[SUM1]], ptr [[TMP22]], align 8
 // IR-PCH-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
-// IR-PCH-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3:[0-9]+]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-PCH-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-PCH-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
@@ -2433,7 +1319,7 @@ int foo() {
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP26]]
 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
 // IR-PCH:       omp.arraycpy.done10:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.case2:
 // IR-PCH-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
@@ -2449,7 +1335,6 @@ int foo() {
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP29]]
 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
 // IR-PCH:       omp.arraycpy.done18:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.default:
 // IR-PCH-NEXT:    ret void
@@ -2568,7 +1453,7 @@ int foo() {
 // IR-PCH-NEXT:    store ptr [[SUM4]], ptr [[TMP21]], align 8
 // IR-PCH-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// IR-PCH-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB3]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3foov_l22.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-PCH-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-PCH-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
@@ -2589,7 +1474,7 @@ int foo() {
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP25]]
 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
 // IR-PCH:       omp.arraycpy.done19:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.case2:
 // IR-PCH-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP0]], i64 100
@@ -2605,7 +1490,6 @@ int foo() {
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP28]]
 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
 // IR-PCH:       omp.arraycpy.done27:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.default:
 // IR-PCH-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
diff --git a/clang/test/OpenMP/target_teams_generic_loop_codegen_as_distribute.cpp b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_distribute.cpp
new file mode 100644
index 0000000000000..f3bbbc6229abd
--- /dev/null
+++ b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_distribute.cpp
@@ -0,0 +1,587 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ --version 2
+// REQUIRES: amdgpu-registered-target
+
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=IR-GPU
+
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-PCH
+extern int foo(int i);
+
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+int N = 100000;
+int main()
+{
+  int i;
+  int a[N];
+  int b[N];
+
+  // Presence of call. Cannot use 'parallel for', must use 'distribute' when
+  // assume-no-neseted-parallelism isn't specified.
+  #pragma omp target teams loop
+  for (i=0; i < N; i++) {
+    for (int j=0; j < N; j++) {
+      a[i] = b[i] * N + foo(j);
+    }
+  }
+  return 0;
+}
+#endif
+// IR-GPU-LABEL: define weak_odr protected amdgpu_kernel void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27
+// IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR0:[0-9]+]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
+// IR-GPU-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27_kernel_environment to ptr), ptr [[DYN_PTR]])
+// IR-GPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
+// IR-GPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// IR-GPU:       user_code.entry:
+// IR-GPU-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
+// IR-GPU-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27_omp_outlined(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2:[0-9]+]]
+// IR-GPU-NEXT:    call void @__kmpc_target_deinit()
+// IR-GPU-NEXT:    ret void
+// IR-GPU:       worker.exit:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1:[0-9]+]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I5:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_3]] to ptr
+// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[I5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I5]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-GPU-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-GPU-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-GPU:       cond.true:
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END:%.*]]
+// IR-GPU:       cond.false:
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END]]
+// IR-GPU:       cond.end:
+// IR-GPU-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-GPU-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-GPU-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-GPU-NEXT:    store i32 [[ADD]], ptr [[I5_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[FOR_COND:%.*]]
+// IR-GPU:       for.cond:
+// IR-GPU-NEXT:    [[TMP18:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp slt i32 [[TMP18]], [[TMP19]]
+// IR-GPU-NEXT:    br i1 [[CMP8]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR-GPU:       for.body:
+// IR-GPU-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I5_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP21]], [[TMP22]]
+// IR-GPU-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP23]]) #[[ATTR4:[0-9]+]]
+// IR-GPU-NEXT:    [[ADD10:%.*]] = add nsw i32 [[MUL9]], [[CALL]]
+// IR-GPU-NEXT:    [[TMP24:%.*]] = load i32, ptr [[I5_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP24]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM11]]
+// IR-GPU-NEXT:    store i32 [[ADD10]], ptr [[ARRAYIDX12]], align 4
+// IR-GPU-NEXT:    br label [[FOR_INC:%.*]]
+// IR-GPU:       for.inc:
+// IR-GPU-NEXT:    [[TMP25:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP25]], 1
+// IR-GPU-NEXT:    store i32 [[INC]], ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
+// IR-GPU:       for.end:
+// IR-GPU-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-GPU:       omp.body.continue:
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP26]], 1
+// IR-GPU-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP28]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-LABEL: define dso_local noundef i32 @main
+// IR-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP7]])
+// IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-NEXT:    ret i32 [[TMP8]]
+//
+//
+// IR-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27
+// IR-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[I5]], align 4
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    br label [[FOR_COND:%.*]]
+// IR:       for.cond:
+// IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    [[CMP8:%.*]] = icmp slt i32 [[TMP18]], [[TMP19]]
+// IR-NEXT:    br i1 [[CMP8]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR:       for.body:
+// IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I5]], align 4
+// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
+// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP21]], [[TMP22]]
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP23]])
+// IR-NEXT:    [[ADD10:%.*]] = add nsw i32 [[MUL9]], [[CALL]]
+// IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[I5]], align 4
+// IR-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP24]] to i64
+// IR-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM11]]
+// IR-NEXT:    store i32 [[ADD10]], ptr [[ARRAYIDX12]], align 4
+// IR-NEXT:    br label [[FOR_INC:%.*]]
+// IR:       for.inc:
+// IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP25]], 1
+// IR-NEXT:    store i32 [[INC]], ptr [[J]], align 4
+// IR-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
+// IR:       for.end:
+// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR:       omp.body.continue:
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP26]], 1
+// IR-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define dso_local noundef i32 @main
+// IR-PCH-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-PCH-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-PCH-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-PCH-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-PCH-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-PCH-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-PCH-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-PCH-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-PCH-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP7]])
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    ret i32 [[TMP8]]
+//
+//
+// IR-PCH-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27
+// IR-PCH-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l27.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-PCH-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-PCH-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I5]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    br label [[FOR_COND:%.*]]
+// IR-PCH:       for.cond:
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    [[CMP8:%.*]] = icmp slt i32 [[TMP18]], [[TMP19]]
+// IR-PCH-NEXT:    br i1 [[CMP8]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR-PCH:       for.body:
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I5]], align 4
+// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP21]], [[TMP22]]
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP23]])
+// IR-PCH-NEXT:    [[ADD10:%.*]] = add nsw i32 [[MUL9]], [[CALL]]
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[I5]], align 4
+// IR-PCH-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP24]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM11]]
+// IR-PCH-NEXT:    store i32 [[ADD10]], ptr [[ARRAYIDX12]], align 4
+// IR-PCH-NEXT:    br label [[FOR_INC:%.*]]
+// IR-PCH:       for.inc:
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP25]], 1
+// IR-PCH-NEXT:    store i32 [[INC]], ptr [[J]], align 4
+// IR-PCH-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
+// IR-PCH:       for.end:
+// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH:       omp.body.continue:
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP26]], 1
+// IR-PCH-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
diff --git a/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp
new file mode 100644
index 0000000000000..7c7cdc53fa2d2
--- /dev/null
+++ b/clang/test/OpenMP/target_teams_generic_loop_codegen_as_parallel_for.cpp
@@ -0,0 +1,3998 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
+// REQUIRES: amdgpu-registered-target
+
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=IR-GPU
+
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-PCH
+
+
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -fopenmp-assume-no-nested-parallelism -DNESTED -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -fopenmp-is-target-device -fopenmp-assume-no-nested-parallelism -DNESTED -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=IR-GPU-NESTED
+
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-assume-no-nested-parallelism -DNESTED -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-NESTED
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-assume-no-nested-parallelism -DNESTED -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -include-pch %t -fopenmp-assume-no-nested-parallelism -DNESTED -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-PCH-NESTED
+
+// expected-no-diagnostics
+
+#ifndef NESTED
+extern int omp_get_num_teams(void);
+#endif
+
+#ifndef HEADER
+#define HEADER
+extern int foo(int i);
+
+int N = 100000;
+int main()
+{
+  int a[N];
+  int b[N];
+
+#ifndef NESTED
+  // Should be transformed into 'target teams distribute parallel for'
+  #pragma omp target teams loop
+  for (int j = 0; j != N; j++)
+    a[j]=b[j];
+
+  // Should be transformed into 'target teams distribute parallel for'
+  #pragma omp target teams loop collapse(2)
+  for (int i = 0; i < N; i++) {
+    for (int j = 0; j < N; j++) {
+      a[i] = b[i] * N + j;
+    }
+  }
+
+  int nt = 0;
+  // Should be transformed into 'target teams distribute parallel for'
+  #pragma omp target teams loop num_teams(32)
+  for (int i=0; i < N; i++) {
+    if (!nt) nt = omp_get_num_teams();
+    for (int j=0; j < N; j++)
+      a[j] = b[j] * N + nt;
+  }
+#else
+  // Should be transformed into 'target teams distribute parallel for'
+  // even with function call because of assume-no-nested-parallelism.
+  #pragma omp target teams loop collapse(2)
+  for (int i = 0; i < N; i++) {
+    for (int j = 0; j < N; j++) {
+      a[i] = b[i] * N + foo(j);
+    }
+  }
+#endif
+  return 0;
+}
+#endif
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
+// IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR0:[0-9]+]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
+// IR-GPU-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41_kernel_environment to ptr), ptr [[DYN_PTR]])
+// IR-GPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
+// IR-GPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// IR-GPU:       user_code.entry:
+// IR-GPU-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
+// IR-GPU-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41_omp_outlined(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2:[0-9]+]]
+// IR-GPU-NEXT:    call void @__kmpc_target_deinit()
+// IR-GPU-NEXT:    ret void
+// IR-GPU:       worker.exit:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1:[0-9]+]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J5:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [7 x ptr], align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_3]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-GPU-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP9]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-GPU-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-GPU:       cond.true:
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END:%.*]]
+// IR-GPU:       cond.false:
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END]]
+// IR-GPU:       cond.end:
+// IR-GPU-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-GPU-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// IR-GPU-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
+// IR-GPU-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
+// IR-GPU-NEXT:    [[TMP24:%.*]] = inttoptr i64 [[TMP18]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP24]], ptr [[TMP23]], align 8
+// IR-GPU-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
+// IR-GPU-NEXT:    [[TMP26:%.*]] = inttoptr i64 [[TMP20]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP26]], ptr [[TMP25]], align 8
+// IR-GPU-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
+// IR-GPU-NEXT:    [[TMP28:%.*]] = inttoptr i64 [[TMP22]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP28]], ptr [[TMP27]], align 8
+// IR-GPU-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
+// IR-GPU-NEXT:    [[TMP30:%.*]] = inttoptr i64 [[TMP0]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP30]], ptr [[TMP29]], align 8
+// IR-GPU-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
+// IR-GPU-NEXT:    store ptr [[TMP1]], ptr [[TMP31]], align 8
+// IR-GPU-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
+// IR-GPU-NEXT:    [[TMP33:%.*]] = inttoptr i64 [[TMP2]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP33]], ptr [[TMP32]], align 8
+// IR-GPU-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 6
+// IR-GPU-NEXT:    store ptr [[TMP3]], ptr [[TMP34]], align 8
+// IR-GPU-NEXT:    [[TMP35:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP36:%.*]] = load i32, ptr [[TMP35]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP36]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 7)
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP37:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP38:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP37]], [[TMP38]]
+// IR-GPU-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP39:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP40:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP39]], [[TMP40]]
+// IR-GPU-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP41:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP42:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP41]], [[TMP42]]
+// IR-GPU-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP44:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], [[TMP44]]
+// IR-GPU-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
+// IR-GPU:       cond.true12:
+// IR-GPU-NEXT:    [[TMP45:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END14:%.*]]
+// IR-GPU:       cond.false13:
+// IR-GPU-NEXT:    [[TMP46:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END14]]
+// IR-GPU:       cond.end14:
+// IR-GPU-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP45]], [[COND_TRUE12]] ], [ [[TMP46]], [[COND_FALSE13]] ]
+// IR-GPU-NEXT:    store i32 [[COND15]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP47:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP47]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP48:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP49:%.*]] = load i32, ptr [[TMP48]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP49]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41_omp_outlined_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J6:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_3]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[J6_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J6]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-GPU-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-GPU-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3:[0-9]+]] to ptr), i32 [[TMP11]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp ule i64 [[CONV7]], [[TMP14]]
+// IR-GPU-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-GPU-NEXT:    store i32 [[ADD]], ptr [[J6_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[J6_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-GPU-NEXT:    [[TMP18:%.*]] = load i32, ptr [[J6_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP18]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM9]]
+// IR-GPU-NEXT:    store i32 [[TMP17]], ptr [[ARRAYIDX10]], align 4
+// IR-GPU-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-GPU:       omp.body.continue:
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// IR-GPU-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP22]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46
+// IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR0]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
+// IR-GPU-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_kernel_environment to ptr), ptr [[DYN_PTR]])
+// IR-GPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
+// IR-GPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// IR-GPU:       user_code.entry:
+// IR-GPU-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
+// IR-GPU-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2]]
+// IR-GPU-NEXT:    call void @__kmpc_target_deinit()
+// IR-GPU-NEXT:    ret void
+// IR-GPU:       worker.exit:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I11:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J12:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [7 x ptr], align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[TMP3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP3]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_4]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_5]] to ptr
+// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[I11_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I11]] to ptr
+// IR-GPU-NEXT:    [[J12_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J12]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-GPU-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-GPU-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-GPU-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-GPU-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       land.lhs.true:
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-GPU-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
+// IR-GPU-NEXT:    [[CONV13:%.*]] = zext i32 [[NVPTX_NUM_THREADS]] to i64
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_init_8(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP12]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i64 1, i64 [[CONV13]])
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CMP14:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-GPU-NEXT:    br i1 [[CMP14]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-GPU:       cond.true:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[COND_END:%.*]]
+// IR-GPU:       cond.false:
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[COND_END]]
+// IR-GPU:       cond.end:
+// IR-GPU-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-GPU-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP19]], 1
+// IR-GPU-NEXT:    [[CMP15:%.*]] = icmp slt i64 [[TMP18]], [[ADD]]
+// IR-GPU-NEXT:    br i1 [[CMP15]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
+// IR-GPU-NEXT:    [[TMP25:%.*]] = inttoptr i64 [[TMP20]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP25]], ptr [[TMP24]], align 8
+// IR-GPU-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
+// IR-GPU-NEXT:    [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP27]], ptr [[TMP26]], align 8
+// IR-GPU-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
+// IR-GPU-NEXT:    [[TMP29:%.*]] = inttoptr i64 [[TMP23]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP29]], ptr [[TMP28]], align 8
+// IR-GPU-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
+// IR-GPU-NEXT:    [[TMP31:%.*]] = inttoptr i64 [[TMP0]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP31]], ptr [[TMP30]], align 8
+// IR-GPU-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
+// IR-GPU-NEXT:    store ptr [[TMP1]], ptr [[TMP32]], align 8
+// IR-GPU-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
+// IR-GPU-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP2]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP34]], ptr [[TMP33]], align 8
+// IR-GPU-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 6
+// IR-GPU-NEXT:    store ptr [[TMP3]], ptr [[TMP35]], align 8
+// IR-GPU-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP36]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP37]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 7)
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP38:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP39:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    [[ADD16:%.*]] = add nsw i64 [[TMP38]], [[TMP39]]
+// IR-GPU-NEXT:    store i64 [[ADD16]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP40:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP41:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP40]], [[TMP41]]
+// IR-GPU-NEXT:    store i64 [[ADD17]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP42:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP43:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP42]], [[TMP43]]
+// IR-GPU-NEXT:    store i64 [[ADD18]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP44:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP45:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CMP19:%.*]] = icmp sgt i64 [[TMP44]], [[TMP45]]
+// IR-GPU-NEXT:    br i1 [[CMP19]], label [[COND_TRUE20:%.*]], label [[COND_FALSE21:%.*]]
+// IR-GPU:       cond.true20:
+// IR-GPU-NEXT:    [[TMP46:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[COND_END22:%.*]]
+// IR-GPU:       cond.false21:
+// IR-GPU-NEXT:    [[TMP47:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[COND_END22]]
+// IR-GPU:       cond.end22:
+// IR-GPU-NEXT:    [[COND23:%.*]] = phi i64 [ [[TMP46]], [[COND_TRUE20]] ], [ [[TMP47]], [[COND_FALSE21]] ]
+// IR-GPU-NEXT:    store i64 [[COND23]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP48:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP48]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP49:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP50:%.*]] = load i32, ptr [[TMP49]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP50]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I11:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J12:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[TMP3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP3]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_4]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_5]] to ptr
+// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[I11_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I11]] to ptr
+// IR-GPU-NEXT:    [[J12_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J12]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-GPU-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-GPU-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-GPU-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-GPU-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       land.lhs.true:
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-GPU-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i64 0, ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_init_8(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP14]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i64 1, i64 1)
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CMP13:%.*]] = icmp ule i64 [[TMP16]], [[TMP17]]
+// IR-GPU-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP19]], 0
+// IR-GPU-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
+// IR-GPU-NEXT:    [[MUL16:%.*]] = mul nsw i32 1, [[DIV15]]
+// IR-GPU-NEXT:    [[CONV17:%.*]] = sext i32 [[MUL16]] to i64
+// IR-GPU-NEXT:    [[DIV18:%.*]] = sdiv i64 [[TMP18]], [[CONV17]]
+// IR-GPU-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[DIV18]], 1
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL19]]
+// IR-GPU-NEXT:    [[CONV20:%.*]] = trunc i64 [[ADD]] to i32
+// IR-GPU-NEXT:    store i32 [[CONV20]], ptr [[I11_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP22]], 0
+// IR-GPU-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
+// IR-GPU-NEXT:    [[MUL23:%.*]] = mul nsw i32 1, [[DIV22]]
+// IR-GPU-NEXT:    [[CONV24:%.*]] = sext i32 [[MUL23]] to i64
+// IR-GPU-NEXT:    [[DIV25:%.*]] = sdiv i64 [[TMP21]], [[CONV24]]
+// IR-GPU-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-GPU-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
+// IR-GPU-NEXT:    [[MUL28:%.*]] = mul nsw i32 1, [[DIV27]]
+// IR-GPU-NEXT:    [[CONV29:%.*]] = sext i32 [[MUL28]] to i64
+// IR-GPU-NEXT:    [[MUL30:%.*]] = mul nsw i64 [[DIV25]], [[CONV29]]
+// IR-GPU-NEXT:    [[SUB31:%.*]] = sub nsw i64 [[TMP20]], [[MUL30]]
+// IR-GPU-NEXT:    [[MUL32:%.*]] = mul nsw i64 [[SUB31]], 1
+// IR-GPU-NEXT:    [[ADD33:%.*]] = add nsw i64 0, [[MUL32]]
+// IR-GPU-NEXT:    [[CONV34:%.*]] = trunc i64 [[ADD33]] to i32
+// IR-GPU-NEXT:    store i32 [[CONV34]], ptr [[J12_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP24:%.*]] = load i32, ptr [[I11_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-GPU-NEXT:    [[TMP25:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-GPU-NEXT:    [[TMP26:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[TMP25]], [[TMP26]]
+// IR-GPU-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J12_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD36:%.*]] = add nsw i32 [[MUL35]], [[TMP27]]
+// IR-GPU-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM37:%.*]] = sext i32 [[TMP28]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM37]]
+// IR-GPU-NEXT:    store i32 [[ADD36]], ptr [[ARRAYIDX38]], align 4
+// IR-GPU-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-GPU:       omp.body.continue:
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP29:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP29]], [[TMP30]]
+// IR-GPU-NEXT:    store i64 [[ADD39]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP32]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55
+// IR-GPU-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR4:[0-9]+]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[NT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[NT_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[NT_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[NT_CASTED]] to ptr
+// IR-GPU-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
+// IR-GPU-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[NT]], ptr [[NT_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55_kernel_environment to ptr), ptr [[DYN_PTR]])
+// IR-GPU-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
+// IR-GPU-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// IR-GPU:       user_code.entry:
+// IR-GPU-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load i32, ptr [[NT_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP8]], ptr [[NT_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i64, ptr [[NT_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP5]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
+// IR-GPU-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55_omp_outlined(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2]]
+// IR-GPU-NEXT:    call void @__kmpc_target_deinit()
+// IR-GPU-NEXT:    ret void
+// IR-GPU:       worker.exit:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I5:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [8 x ptr], align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[NT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[NT_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_3]] to ptr
+// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[I5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I5]] to ptr
+// IR-GPU-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NEXT:    [[NT_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[NT_CASTED]] to ptr
+// IR-GPU-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[NT]], ptr [[NT_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-GPU-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP9]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-GPU-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-GPU:       cond.true:
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END:%.*]]
+// IR-GPU:       cond.false:
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END]]
+// IR-GPU:       cond.end:
+// IR-GPU-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-GPU-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
+// IR-GPU-NEXT:    [[CMP7:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
+// IR-GPU-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP23:%.*]] = load i32, ptr [[NT_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP23]], ptr [[NT_CASTED_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP24:%.*]] = load i64, ptr [[NT_CASTED_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
+// IR-GPU-NEXT:    [[TMP26:%.*]] = inttoptr i64 [[TMP18]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP26]], ptr [[TMP25]], align 8
+// IR-GPU-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
+// IR-GPU-NEXT:    [[TMP28:%.*]] = inttoptr i64 [[TMP20]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP28]], ptr [[TMP27]], align 8
+// IR-GPU-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
+// IR-GPU-NEXT:    [[TMP30:%.*]] = inttoptr i64 [[TMP22]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP30]], ptr [[TMP29]], align 8
+// IR-GPU-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
+// IR-GPU-NEXT:    [[TMP32:%.*]] = inttoptr i64 [[TMP24]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP32]], ptr [[TMP31]], align 8
+// IR-GPU-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
+// IR-GPU-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP0]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP34]], ptr [[TMP33]], align 8
+// IR-GPU-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
+// IR-GPU-NEXT:    store ptr [[TMP1]], ptr [[TMP35]], align 8
+// IR-GPU-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 6
+// IR-GPU-NEXT:    [[TMP37:%.*]] = inttoptr i64 [[TMP2]] to ptr
+// IR-GPU-NEXT:    store ptr [[TMP37]], ptr [[TMP36]], align 8
+// IR-GPU-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [8 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 7
+// IR-GPU-NEXT:    store ptr [[TMP3]], ptr [[TMP38]], align 8
+// IR-GPU-NEXT:    [[TMP39:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP40:%.*]] = load i32, ptr [[TMP39]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP40]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 8)
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP41:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP42:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP41]], [[TMP42]]
+// IR-GPU-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP44:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP43]], [[TMP44]]
+// IR-GPU-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP45:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP46:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP45]], [[TMP46]]
+// IR-GPU-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP47:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP48:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP47]], [[TMP48]]
+// IR-GPU-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
+// IR-GPU:       cond.true12:
+// IR-GPU-NEXT:    [[TMP49:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END14:%.*]]
+// IR-GPU:       cond.false13:
+// IR-GPU-NEXT:    [[TMP50:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[COND_END14]]
+// IR-GPU:       cond.end14:
+// IR-GPU-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP49]], [[COND_TRUE12]] ], [ [[TMP50]], [[COND_FALSE13]] ]
+// IR-GPU-NEXT:    store i32 [[COND15]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP51:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP51]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP52:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP53:%.*]] = load i32, ptr [[TMP52]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP53]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-GPU-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55_omp_outlined_omp_outlined
+// IR-GPU-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NEXT:  entry:
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[I6:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
+// IR-GPU-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NEXT:    [[NT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[NT_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NEXT:    [[DOTCAPTURE_EXPR_3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_3]] to ptr
+// IR-GPU-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NEXT:    [[I6_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I6]] to ptr
+// IR-GPU-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[NT]], ptr [[NT_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-GPU-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-GPU-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-GPU-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU:       omp.precond.then:
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-GPU-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-GPU-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP11]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
+// IR-GPU-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
+// IR-GPU-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU:       omp.inner.for.cond:
+// IR-GPU-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP13]] to i64
+// IR-GPU-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[CMP8:%.*]] = icmp ule i64 [[CONV7]], [[TMP14]]
+// IR-GPU-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU:       omp.inner.for.body:
+// IR-GPU-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// IR-GPU-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-GPU-NEXT:    store i32 [[ADD]], ptr [[I6_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP16:%.*]] = load i32, ptr [[NT_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
+// IR-GPU-NEXT:    br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
+// IR-GPU:       if.then:
+// IR-GPU-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z17omp_get_num_teamsv() #[[ATTR6:[0-9]+]]
+// IR-GPU-NEXT:    store i32 [[CALL]], ptr [[NT_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[IF_END]]
+// IR-GPU:       if.end:
+// IR-GPU-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[FOR_COND:%.*]]
+// IR-GPU:       for.cond:
+// IR-GPU-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP18:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP17]], [[TMP18]]
+// IR-GPU-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR-GPU:       for.body:
+// IR-GPU-NEXT:    [[TMP19:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-GPU-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-GPU-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP20]], [[TMP21]]
+// IR-GPU-NEXT:    [[TMP22:%.*]] = load i32, ptr [[NT_ADDR_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD11:%.*]] = add nsw i32 [[MUL10]], [[TMP22]]
+// IR-GPU-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP23]] to i64
+// IR-GPU-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM12]]
+// IR-GPU-NEXT:    store i32 [[ADD11]], ptr [[ARRAYIDX13]], align 4
+// IR-GPU-NEXT:    br label [[FOR_INC:%.*]]
+// IR-GPU:       for.inc:
+// IR-GPU-NEXT:    [[TMP24:%.*]] = load i32, ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP24]], 1
+// IR-GPU-NEXT:    store i32 [[INC]], ptr [[J_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
+// IR-GPU:       for.end:
+// IR-GPU-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-GPU:       omp.body.continue:
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU:       omp.inner.for.inc:
+// IR-GPU-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
+// IR-GPU-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
+// IR-GPU-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV_ASCAST]], align 4
+// IR-GPU-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU:       omp.inner.for.end:
+// IR-GPU-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU:       omp.loop.exit:
+// IR-GPU-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-GPU-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP28]])
+// IR-GPU-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU:       omp.precond.end:
+// IR-GPU-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@main
+// IR-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_CASTED2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[N_CASTED2]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i64, ptr [[N_CASTED2]], align 8
+// IR-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46(i64 [[TMP8]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3]]
+// IR-NEXT:    store i32 0, ptr [[NT]], align 4
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr @N, align 4
+// IR-NEXT:    store i32 [[TMP9]], ptr [[N_CASTED3]], align 4
+// IR-NEXT:    [[TMP10:%.*]] = load i64, ptr [[N_CASTED3]], align 8
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[NT]], align 4
+// IR-NEXT:    store i32 [[TMP11]], ptr [[NT_CASTED]], align 4
+// IR-NEXT:    [[TMP12:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55(i64 [[TMP10]], i64 [[TMP12]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3]]
+// IR-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP13]])
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-NEXT:    ret i32 [[TMP14]]
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
+// IR-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J5:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J6:%.*]] = alloca i32, align 4
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP11]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP12]], [[TMP13]]
+// IR-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
+// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 [[TMP16]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// IR-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[J6]], align 4
+// IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J6]], align 4
+// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
+// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J6]], align 4
+// IR-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP22]] to i64
+// IR-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM9]]
+// IR-NEXT:    store i32 [[TMP21]], ptr [[ARRAYIDX10]], align 4
+// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR:       omp.body.continue:
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP23]], 1
+// IR-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP25]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46
+// IR-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       land.lhs.true:
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1]], i32 [[TMP12]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP18]], [[TMP19]]
+// IR-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined.omp_outlined, i64 [[TMP20]], i64 [[TMP21]], i64 [[TMP23]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP24]], [[TMP25]]
+// IR-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       land.lhs.true:
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// IR-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB]], align 8
+// IR-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2]], i32 [[TMP14]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP15]], [[TMP16]]
+// IR-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
+// IR-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// IR-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP20]], [[TMP21]]
+// IR-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// IR-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
+// IR-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
+// IR-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP22]], [[CONV18]]
+// IR-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
+// IR-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
+// IR-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
+// IR-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
+// IR-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP26]], 0
+// IR-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
+// IR-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
+// IR-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
+// IR-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP25]], [[CONV25]]
+// IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP27]], 0
+// IR-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
+// IR-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
+// IR-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
+// IR-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
+// IR-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP24]], [[MUL31]]
+// IR-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
+// IR-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
+// IR-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
+// IR-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
+// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11]], align 4
+// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP28]] to i64
+// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[TMP29]], [[TMP30]]
+// IR-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
+// IR-NEXT:    [[ADD37:%.*]] = add nsw i32 [[MUL36]], [[TMP31]]
+// IR-NEXT:    [[TMP32:%.*]] = load i32, ptr [[I11]], align 4
+// IR-NEXT:    [[IDXPROM38:%.*]] = sext i32 [[TMP32]] to i64
+// IR-NEXT:    [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM38]]
+// IR-NEXT:    store i32 [[ADD37]], ptr [[ARRAYIDX39]], align 4
+// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR:       omp.body.continue:
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    [[ADD40:%.*]] = add nsw i64 [[TMP33]], 1
+// IR-NEXT:    store i64 [[ADD40]], ptr [[DOTOMP_IV]], align 8
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP35]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55
+// IR-SAME: (i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 32, i32 0)
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[NT_CASTED]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined, i64 [[TMP6]], i64 [[TMP8]], i64 [[TMP1]], ptr [[TMP2]], i64 [[TMP3]], ptr [[TMP4]])
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP23]], ptr [[NT_CASTED]], align 4
+// IR-NEXT:    [[TMP24:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP24]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined
+// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NEXT:  entry:
+// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I6:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR:       omp.precond.then:
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP11]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP12]], [[TMP13]]
+// IR-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR:       cond.true:
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    br label [[COND_END:%.*]]
+// IR:       cond.false:
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    br label [[COND_END]]
+// IR:       cond.end:
+// IR-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
+// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 [[TMP16]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR:       omp.inner.for.cond:
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// IR-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR:       omp.inner.for.body:
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[I6]], align 4
+// IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP20]], 0
+// IR-NEXT:    br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
+// IR:       if.then:
+// IR-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z17omp_get_num_teamsv()
+// IR-NEXT:    store i32 [[CALL]], ptr [[NT_ADDR]], align 4
+// IR-NEXT:    br label [[IF_END]]
+// IR:       if.end:
+// IR-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NEXT:    br label [[FOR_COND:%.*]]
+// IR:       for.cond:
+// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// IR-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR:       for.body:
+// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
+// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP24]], [[TMP25]]
+// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-NEXT:    [[ADD11:%.*]] = add nsw i32 [[MUL10]], [[TMP26]]
+// IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// IR-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM12]]
+// IR-NEXT:    store i32 [[ADD11]], ptr [[ARRAYIDX13]], align 4
+// IR-NEXT:    br label [[FOR_INC:%.*]]
+// IR:       for.inc:
+// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[J]], align 4
+// IR-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP28]], 1
+// IR-NEXT:    store i32 [[INC]], ptr [[J]], align 4
+// IR-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
+// IR:       for.end:
+// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR:       omp.body.continue:
+// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR:       omp.inner.for.inc:
+// IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
+// IR-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR:       omp.inner.for.end:
+// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR:       omp.loop.exit:
+// IR-NEXT:    [[TMP30:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP31:%.*]] = load i32, ptr [[TMP30]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP31]])
+// IR-NEXT:    br label [[OMP_PRECOND_END]]
+// IR:       omp.precond.end:
+// IR-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@main
+// IR-PCH-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_CASTED2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-PCH-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-PCH-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-PCH-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-PCH-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-PCH-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-PCH-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[N_CASTED2]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i64, ptr [[N_CASTED2]], align 8
+// IR-PCH-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46(i64 [[TMP8]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3]]
+// IR-PCH-NEXT:    store i32 0, ptr [[NT]], align 4
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NEXT:    store i32 [[TMP9]], ptr [[N_CASTED3]], align 4
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i64, ptr [[N_CASTED3]], align 8
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[NT]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP11]], ptr [[NT_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-PCH-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55(i64 [[TMP10]], i64 [[TMP12]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3]]
+// IR-PCH-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-PCH-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP13]])
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-PCH-NEXT:    ret i32 [[TMP14]]
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
+// IR-PCH-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J5:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-PCH-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-PCH-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J6:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-PCH-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP11]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP12]], [[TMP13]]
+// IR-PCH-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP16]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// IR-PCH-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[J6]], align 4
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J6]], align 4
+// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J6]], align 4
+// IR-PCH-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP22]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM9]]
+// IR-PCH-NEXT:    store i32 [[TMP21]], ptr [[ARRAYIDX10]], align 4
+// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH:       omp.body.continue:
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP23]], 1
+// IR-PCH-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP25]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46
+// IR-PCH-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-PCH-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-PCH-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-PCH-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-PCH-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       land.lhs.true:
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-PCH-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1]], i32 [[TMP12]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-PCH-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP18]], [[TMP19]]
+// IR-PCH-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined.omp_outlined, i64 [[TMP20]], i64 [[TMP21]], i64 [[TMP23]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP24]], [[TMP25]]
+// IR-PCH-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46.omp_outlined.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-PCH-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-PCH-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-PCH-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-PCH-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       land.lhs.true:
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-PCH-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2]], i32 [[TMP14]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP15]], [[TMP16]]
+// IR-PCH-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP20]], [[TMP21]]
+// IR-PCH-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-PCH-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// IR-PCH-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
+// IR-PCH-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
+// IR-PCH-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP22]], [[CONV18]]
+// IR-PCH-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
+// IR-PCH-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
+// IR-PCH-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP26]], 0
+// IR-PCH-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
+// IR-PCH-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
+// IR-PCH-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
+// IR-PCH-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP25]], [[CONV25]]
+// IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP27]], 0
+// IR-PCH-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
+// IR-PCH-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
+// IR-PCH-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
+// IR-PCH-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
+// IR-PCH-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP24]], [[MUL31]]
+// IR-PCH-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
+// IR-PCH-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
+// IR-PCH-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
+// IR-PCH-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
+// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11]], align 4
+// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP28]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-PCH-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[TMP29]], [[TMP30]]
+// IR-PCH-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
+// IR-PCH-NEXT:    [[ADD37:%.*]] = add nsw i32 [[MUL36]], [[TMP31]]
+// IR-PCH-NEXT:    [[TMP32:%.*]] = load i32, ptr [[I11]], align 4
+// IR-PCH-NEXT:    [[IDXPROM38:%.*]] = sext i32 [[TMP32]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM38]]
+// IR-PCH-NEXT:    store i32 [[ADD37]], ptr [[ARRAYIDX39]], align 4
+// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH:       omp.body.continue:
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    [[ADD40:%.*]] = add nsw i64 [[TMP33]], 1
+// IR-PCH-NEXT:    store i64 [[ADD40]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP35]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55
+// IR-PCH-SAME: (i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 32, i32 0)
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[NT_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined, i64 [[TMP6]], i64 [[TMP8]], i64 [[TMP1]], ptr [[TMP2]], i64 [[TMP3]], ptr [[TMP4]])
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I5:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
+// IR-PCH-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
+// IR-PCH-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// IR-PCH-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// IR-PCH-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP23]], ptr [[NT_CASTED]], align 4
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i64, ptr [[NT_CASTED]], align 8
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 8, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP24]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-PCH-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined
+// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[NT:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NEXT:  entry:
+// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[NT_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[I6:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[NT]], ptr [[NT_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
+// IR-PCH-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH:       omp.precond.then:
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
+// IR-PCH-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NEXT:    [[CONV5:%.*]] = trunc i64 [[TMP9]] to i32
+// IR-PCH-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[CONV5]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP11]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[TMP12]], [[TMP13]]
+// IR-PCH-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH:       cond.true:
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// IR-PCH-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH:       cond.false:
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    br label [[COND_END]]
+// IR-PCH:       cond.end:
+// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
+// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP16]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH:       omp.inner.for.cond:
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// IR-PCH-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
+// IR-PCH-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH:       omp.inner.for.body:
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I6]], align 4
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-PCH-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP20]], 0
+// IR-PCH-NEXT:    br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
+// IR-PCH:       if.then:
+// IR-PCH-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z17omp_get_num_teamsv()
+// IR-PCH-NEXT:    store i32 [[CALL]], ptr [[NT_ADDR]], align 4
+// IR-PCH-NEXT:    br label [[IF_END]]
+// IR-PCH:       if.end:
+// IR-PCH-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NEXT:    br label [[FOR_COND:%.*]]
+// IR-PCH:       for.cond:
+// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP21]], [[TMP22]]
+// IR-PCH-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// IR-PCH:       for.body:
+// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP23]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP24]], [[TMP25]]
+// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[NT_ADDR]], align 4
+// IR-PCH-NEXT:    [[ADD11:%.*]] = add nsw i32 [[MUL10]], [[TMP26]]
+// IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[IDXPROM12:%.*]] = sext i32 [[TMP27]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM12]]
+// IR-PCH-NEXT:    store i32 [[ADD11]], ptr [[ARRAYIDX13]], align 4
+// IR-PCH-NEXT:    br label [[FOR_INC:%.*]]
+// IR-PCH:       for.inc:
+// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[J]], align 4
+// IR-PCH-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP28]], 1
+// IR-PCH-NEXT:    store i32 [[INC]], ptr [[J]], align 4
+// IR-PCH-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
+// IR-PCH:       for.end:
+// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH:       omp.body.continue:
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH:       omp.inner.for.inc:
+// IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
+// IR-PCH-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH:       omp.inner.for.end:
+// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH:       omp.loop.exit:
+// IR-PCH-NEXT:    [[TMP30:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP31:%.*]] = load i32, ptr [[TMP30]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP31]])
+// IR-PCH-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH:       omp.precond.end:
+// IR-PCH-NEXT:    ret void
+//
+//
+// IR-GPU-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64
+// IR-GPU-NESTED-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR0:[0-9]+]] {
+// IR-GPU-NESTED-NEXT:  entry:
+// IR-GPU-NESTED-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NESTED-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTTHREADID_TEMP__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64_kernel_environment to ptr), ptr [[DYN_PTR]])
+// IR-GPU-NESTED-NEXT:    [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
+// IR-GPU-NESTED-NEXT:    br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
+// IR-GPU-NESTED:       user_code.entry:
+// IR-GPU-NESTED-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
+// IR-GPU-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTTHREADID_TEMP__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64_omp_outlined(ptr [[DOTTHREADID_TEMP__ASCAST]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2:[0-9]+]]
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_target_deinit()
+// IR-GPU-NESTED-NEXT:    ret void
+// IR-GPU-NESTED:       worker.exit:
+// IR-GPU-NESTED-NEXT:    ret void
+//
+//
+// IR-GPU-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64_omp_outlined
+// IR-GPU-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1:[0-9]+]] {
+// IR-GPU-NESTED-NEXT:  entry:
+// IR-GPU-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[CAPTURED_VARS_ADDRS:%.*]] = alloca [7 x ptr], align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NESTED-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NESTED-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NESTED-NEXT:    [[TMP3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP3]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_4]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_5]] to ptr
+// IR-GPU-NESTED-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NESTED-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NESTED-NEXT:    [[I11_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I11]] to ptr
+// IR-GPU-NESTED-NEXT:    [[J12_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J12]] to ptr
+// IR-GPU-NESTED-NEXT:    [[N_CASTED_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_CASTED]] to ptr
+// IR-GPU-NESTED-NEXT:    [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-GPU-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-GPU-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-GPU-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-GPU-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-GPU-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU-NESTED:       land.lhs.true:
+// IR-GPU-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-GPU-NESTED:       omp.precond.then:
+// IR-GPU-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
+// IR-GPU-NESTED-NEXT:    [[CONV13:%.*]] = zext i32 [[NVPTX_NUM_THREADS]] to i64
+// IR-GPU-NESTED-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_distribute_static_init_8(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP12]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i64 1, i64 [[CONV13]])
+// IR-GPU-NESTED-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[CMP14:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP14]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-GPU-NESTED:       cond.true:
+// IR-GPU-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[COND_END:%.*]]
+// IR-GPU-NESTED:       cond.false:
+// IR-GPU-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[COND_END]]
+// IR-GPU-NESTED:       cond.end:
+// IR-GPU-NESTED-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-GPU-NESTED-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.cond:
+// IR-GPU-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP19]], 1
+// IR-GPU-NESTED-NEXT:    [[CMP15:%.*]] = icmp slt i64 [[TMP18]], [[ADD]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP15]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.body:
+// IR-GPU-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
+// IR-GPU-NESTED-NEXT:    [[TMP25:%.*]] = inttoptr i64 [[TMP20]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP25]], ptr [[TMP24]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
+// IR-GPU-NESTED-NEXT:    [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP27]], ptr [[TMP26]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
+// IR-GPU-NESTED-NEXT:    [[TMP29:%.*]] = inttoptr i64 [[TMP23]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP29]], ptr [[TMP28]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
+// IR-GPU-NESTED-NEXT:    [[TMP31:%.*]] = inttoptr i64 [[TMP0]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP31]], ptr [[TMP30]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP1]], ptr [[TMP32]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
+// IR-GPU-NESTED-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP2]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP34]], ptr [[TMP33]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 6
+// IR-GPU-NESTED-NEXT:    store ptr [[TMP3]], ptr [[TMP35]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP36]], align 4
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP37]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 7)
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.inc:
+// IR-GPU-NESTED-NEXT:    [[TMP38:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP39:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[ADD16:%.*]] = add nsw i64 [[TMP38]], [[TMP39]]
+// IR-GPU-NESTED-NEXT:    store i64 [[ADD16]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP40:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP41:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP40]], [[TMP41]]
+// IR-GPU-NESTED-NEXT:    store i64 [[ADD17]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP42:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP43:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP42]], [[TMP43]]
+// IR-GPU-NESTED-NEXT:    store i64 [[ADD18]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP44:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP45:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[CMP19:%.*]] = icmp sgt i64 [[TMP44]], [[TMP45]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP19]], label [[COND_TRUE20:%.*]], label [[COND_FALSE21:%.*]]
+// IR-GPU-NESTED:       cond.true20:
+// IR-GPU-NESTED-NEXT:    [[TMP46:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[COND_END22:%.*]]
+// IR-GPU-NESTED:       cond.false21:
+// IR-GPU-NESTED-NEXT:    [[TMP47:%.*]] = load i64, ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[COND_END22]]
+// IR-GPU-NESTED:       cond.end22:
+// IR-GPU-NESTED-NEXT:    [[COND23:%.*]] = phi i64 [ [[TMP46]], [[COND_TRUE20]] ], [ [[TMP47]], [[COND_FALSE21]] ]
+// IR-GPU-NESTED-NEXT:    store i64 [[COND23]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP48:%.*]] = load i64, ptr [[DOTOMP_COMB_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP48]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU-NESTED:       omp.inner.for.end:
+// IR-GPU-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU-NESTED:       omp.loop.exit:
+// IR-GPU-NESTED-NEXT:    [[TMP49:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP50:%.*]] = load i32, ptr [[TMP49]], align 4
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP50]])
+// IR-GPU-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU-NESTED:       omp.precond.end:
+// IR-GPU-NESTED-NEXT:    ret void
+//
+//
+// IR-GPU-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64_omp_outlined_omp_outlined
+// IR-GPU-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR1]] {
+// IR-GPU-NESTED-NEXT:  entry:
+// IR-GPU-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4, addrspace(5)
+// IR-GPU-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[VLA_ADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR2]] to ptr
+// IR-GPU-NESTED-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
+// IR-GPU-NESTED-NEXT:    [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
+// IR-GPU-NESTED-NEXT:    [[TMP3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[_TMP3]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_4]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_5]] to ptr
+// IR-GPU-NESTED-NEXT:    [[I_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I]] to ptr
+// IR-GPU-NESTED-NEXT:    [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
+// IR-GPU-NESTED-NEXT:    [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
+// IR-GPU-NESTED-NEXT:    [[I11_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I11]] to ptr
+// IR-GPU-NESTED-NEXT:    [[J12_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J12]] to ptr
+// IR-GPU-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-GPU-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-GPU-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-GPU-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-GPU-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-GPU-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-GPU-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[I_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[J_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-GPU-NESTED:       land.lhs.true:
+// IR-GPU-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-GPU-NESTED:       omp.precond.then:
+// IR-GPU-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_for_static_init_8(ptr addrspacecast (ptr addrspace(1) @[[GLOB3:[0-9]+]] to ptr), i32 [[TMP14]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i64 1, i64 1)
+// IR-GPU-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_LB_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    store i64 [[TMP15]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.cond:
+// IR-GPU-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[CMP13:%.*]] = icmp ule i64 [[TMP16]], [[TMP17]]
+// IR-GPU-NESTED-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.body:
+// IR-GPU-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB14:%.*]] = sub nsw i32 [[TMP19]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
+// IR-GPU-NESTED-NEXT:    [[MUL16:%.*]] = mul nsw i32 1, [[DIV15]]
+// IR-GPU-NESTED-NEXT:    [[CONV17:%.*]] = sext i32 [[MUL16]] to i64
+// IR-GPU-NESTED-NEXT:    [[DIV18:%.*]] = sdiv i64 [[TMP18]], [[CONV17]]
+// IR-GPU-NESTED-NEXT:    [[MUL19:%.*]] = mul nsw i64 [[DIV18]], 1
+// IR-GPU-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL19]]
+// IR-GPU-NESTED-NEXT:    [[CONV20:%.*]] = trunc i64 [[ADD]] to i32
+// IR-GPU-NESTED-NEXT:    store i32 [[CONV20]], ptr [[I11_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB21:%.*]] = sub nsw i32 [[TMP22]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV22:%.*]] = sdiv i32 [[SUB21]], 1
+// IR-GPU-NESTED-NEXT:    [[MUL23:%.*]] = mul nsw i32 1, [[DIV22]]
+// IR-GPU-NESTED-NEXT:    [[CONV24:%.*]] = sext i32 [[MUL23]] to i64
+// IR-GPU-NESTED-NEXT:    [[DIV25:%.*]] = sdiv i64 [[TMP21]], [[CONV24]]
+// IR-GPU-NESTED-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-GPU-NESTED-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
+// IR-GPU-NESTED-NEXT:    [[MUL28:%.*]] = mul nsw i32 1, [[DIV27]]
+// IR-GPU-NESTED-NEXT:    [[CONV29:%.*]] = sext i32 [[MUL28]] to i64
+// IR-GPU-NESTED-NEXT:    [[MUL30:%.*]] = mul nsw i64 [[DIV25]], [[CONV29]]
+// IR-GPU-NESTED-NEXT:    [[SUB31:%.*]] = sub nsw i64 [[TMP20]], [[MUL30]]
+// IR-GPU-NESTED-NEXT:    [[MUL32:%.*]] = mul nsw i64 [[SUB31]], 1
+// IR-GPU-NESTED-NEXT:    [[ADD33:%.*]] = add nsw i64 0, [[MUL32]]
+// IR-GPU-NESTED-NEXT:    [[CONV34:%.*]] = trunc i64 [[ADD33]] to i32
+// IR-GPU-NESTED-NEXT:    store i32 [[CONV34]], ptr [[J12_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP24:%.*]] = load i32, ptr [[I11_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
+// IR-GPU-NESTED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-GPU-NESTED-NEXT:    [[TMP25:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-GPU-NESTED-NEXT:    [[TMP26:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[TMP25]], [[TMP26]]
+// IR-GPU-NESTED-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J12_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP27]]) #[[ATTR5:[0-9]+]]
+// IR-GPU-NESTED-NEXT:    [[ADD36:%.*]] = add nsw i32 [[MUL35]], [[CALL]]
+// IR-GPU-NESTED-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11_ASCAST]], align 4
+// IR-GPU-NESTED-NEXT:    [[IDXPROM37:%.*]] = sext i32 [[TMP28]] to i64
+// IR-GPU-NESTED-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM37]]
+// IR-GPU-NESTED-NEXT:    store i32 [[ADD36]], ptr [[ARRAYIDX38]], align 4
+// IR-GPU-NESTED-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-GPU-NESTED:       omp.body.continue:
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-GPU-NESTED:       omp.inner.for.inc:
+// IR-GPU-NESTED-NEXT:    [[TMP29:%.*]] = load i64, ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_STRIDE_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP29]], [[TMP30]]
+// IR-GPU-NESTED-NEXT:    store i64 [[ADD39]], ptr [[DOTOMP_IV_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-GPU-NESTED:       omp.inner.for.end:
+// IR-GPU-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-GPU-NESTED:       omp.loop.exit:
+// IR-GPU-NESTED-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
+// IR-GPU-NESTED-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
+// IR-GPU-NESTED-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP32]])
+// IR-GPU-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-GPU-NESTED:       omp.precond.end:
+// IR-GPU-NESTED-NEXT:    ret void
+//
+//
+// IR-NESTED-LABEL: define {{[^@]+}}@main
+// IR-NESTED-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-NESTED-NEXT:  entry:
+// IR-NESTED-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NESTED-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-NESTED-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-NESTED-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-NESTED-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-NESTED-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-NESTED-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-NESTED-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-NESTED-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-NESTED-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-NESTED-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-NESTED-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-NESTED-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NESTED-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-NESTED-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-NESTED-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-NESTED-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP7]])
+// IR-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-NESTED-NEXT:    ret i32 [[TMP8]]
+//
+//
+// IR-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64
+// IR-NESTED-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-NESTED-NEXT:  entry:
+// IR-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-NESTED-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NESTED-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NESTED-NEXT:    ret void
+//
+//
+// IR-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined
+// IR-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NESTED-NEXT:  entry:
+// IR-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NESTED-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-NESTED:       land.lhs.true:
+// IR-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-NESTED:       omp.precond.then:
+// IR-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NESTED-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-NESTED-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-NESTED-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-NESTED-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-NESTED:       cond.true:
+// IR-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    br label [[COND_END:%.*]]
+// IR-NESTED:       cond.false:
+// IR-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    br label [[COND_END]]
+// IR-NESTED:       cond.end:
+// IR-NESTED-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-NESTED-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-NESTED:       omp.inner.for.cond:
+// IR-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP18]], [[TMP19]]
+// IR-NESTED-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-NESTED:       omp.inner.for.body:
+// IR-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED]], align 4
+// IR-NESTED-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-NESTED-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined.omp_outlined, i64 [[TMP20]], i64 [[TMP21]], i64 [[TMP23]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-NESTED:       omp.inner.for.inc:
+// IR-NESTED-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP24]], [[TMP25]]
+// IR-NESTED-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-NESTED:       omp.inner.for.end:
+// IR-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-NESTED:       omp.loop.exit:
+// IR-NESTED-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
+// IR-NESTED-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
+// IR-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-NESTED:       omp.precond.end:
+// IR-NESTED-NEXT:    ret void
+//
+//
+// IR-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined.omp_outlined
+// IR-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-NESTED-NEXT:  entry:
+// IR-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-NESTED-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-NESTED:       land.lhs.true:
+// IR-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-NESTED:       omp.precond.then:
+// IR-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// IR-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NESTED-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-NESTED-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP14]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP15]], [[TMP16]]
+// IR-NESTED-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-NESTED:       cond.true:
+// IR-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-NESTED-NEXT:    br label [[COND_END:%.*]]
+// IR-NESTED:       cond.false:
+// IR-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    br label [[COND_END]]
+// IR-NESTED:       cond.end:
+// IR-NESTED-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
+// IR-NESTED-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// IR-NESTED-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-NESTED:       omp.inner.for.cond:
+// IR-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-NESTED-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP20]], [[TMP21]]
+// IR-NESTED-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-NESTED:       omp.inner.for.body:
+// IR-NESTED-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-NESTED-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// IR-NESTED-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
+// IR-NESTED-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
+// IR-NESTED-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP22]], [[CONV18]]
+// IR-NESTED-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
+// IR-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
+// IR-NESTED-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
+// IR-NESTED-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
+// IR-NESTED-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP26]], 0
+// IR-NESTED-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
+// IR-NESTED-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
+// IR-NESTED-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
+// IR-NESTED-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP25]], [[CONV25]]
+// IR-NESTED-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-NESTED-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP27]], 0
+// IR-NESTED-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
+// IR-NESTED-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
+// IR-NESTED-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
+// IR-NESTED-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
+// IR-NESTED-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP24]], [[MUL31]]
+// IR-NESTED-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
+// IR-NESTED-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
+// IR-NESTED-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
+// IR-NESTED-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
+// IR-NESTED-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11]], align 4
+// IR-NESTED-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP28]] to i64
+// IR-NESTED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-NESTED-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-NESTED-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-NESTED-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[TMP29]], [[TMP30]]
+// IR-NESTED-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
+// IR-NESTED-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP31]])
+// IR-NESTED-NEXT:    [[ADD37:%.*]] = add nsw i32 [[MUL36]], [[CALL]]
+// IR-NESTED-NEXT:    [[TMP32:%.*]] = load i32, ptr [[I11]], align 4
+// IR-NESTED-NEXT:    [[IDXPROM38:%.*]] = sext i32 [[TMP32]] to i64
+// IR-NESTED-NEXT:    [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM38]]
+// IR-NESTED-NEXT:    store i32 [[ADD37]], ptr [[ARRAYIDX39]], align 4
+// IR-NESTED-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-NESTED:       omp.body.continue:
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-NESTED:       omp.inner.for.inc:
+// IR-NESTED-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    [[ADD40:%.*]] = add nsw i64 [[TMP33]], 1
+// IR-NESTED-NEXT:    store i64 [[ADD40]], ptr [[DOTOMP_IV]], align 8
+// IR-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-NESTED:       omp.inner.for.end:
+// IR-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-NESTED:       omp.loop.exit:
+// IR-NESTED-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NESTED-NEXT:    [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
+// IR-NESTED-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP35]])
+// IR-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-NESTED:       omp.precond.end:
+// IR-NESTED-NEXT:    ret void
+//
+//
+// IR-PCH-NESTED-LABEL: define {{[^@]+}}@main
+// IR-PCH-NESTED-SAME: () #[[ATTR0:[0-9]+]] {
+// IR-PCH-NESTED-NEXT:  entry:
+// IR-PCH-NESTED-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP0:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NESTED-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
+// IR-PCH-NESTED-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
+// IR-PCH-NESTED-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
+// IR-PCH-NESTED-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 16
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP3:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NESTED-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
+// IR-PCH-NESTED-NEXT:    [[VLA1:%.*]] = alloca i32, i64 [[TMP4]], align 16
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP4]], ptr [[__VLA_EXPR1]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr @N, align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NESTED-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i64 [[TMP6]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP4]], ptr [[VLA1]]) #[[ATTR3:[0-9]+]]
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[RETVAL]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
+// IR-PCH-NESTED-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP7]])
+// IR-PCH-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[RETVAL]], align 4
+// IR-PCH-NESTED-NEXT:    ret i32 [[TMP8]]
+//
+//
+// IR-PCH-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64
+// IR-PCH-NESTED-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2:[0-9]+]] {
+// IR-PCH-NESTED-NEXT:  entry:
+// IR-PCH-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP5:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NESTED-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined, i64 [[TMP5]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NESTED-NEXT:    ret void
+//
+//
+// IR-PCH-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined
+// IR-PCH-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NESTED-NEXT:  entry:
+// IR-PCH-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-PCH-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-PCH-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-PCH-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-PCH-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-PCH-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH-NESTED:       land.lhs.true:
+// IR-PCH-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-PCH-NESTED:       omp.precond.then:
+// IR-PCH-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+// IR-PCH-NESTED-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP12]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-PCH-NESTED-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP13]], [[TMP14]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH-NESTED:       cond.true:
+// IR-PCH-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH-NESTED:       cond.false:
+// IR-PCH-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[COND_END]]
+// IR-PCH-NESTED:       cond.end:
+// IR-PCH-NESTED-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP15]], [[COND_TRUE]] ], [ [[TMP16]], [[COND_FALSE]] ]
+// IR-PCH-NESTED-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP17]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.cond:
+// IR-PCH-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP18]], [[TMP19]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.body:
+// IR-PCH-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP22:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP22]], ptr [[N_CASTED]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP23:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// IR-PCH-NESTED-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined.omp_outlined, i64 [[TMP20]], i64 [[TMP21]], i64 [[TMP23]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]])
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.inc:
+// IR-PCH-NESTED-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP24]], [[TMP25]]
+// IR-PCH-NESTED-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH-NESTED:       omp.inner.for.end:
+// IR-PCH-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH-NESTED:       omp.loop.exit:
+// IR-PCH-NESTED-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
+// IR-PCH-NESTED-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
+// IR-PCH-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH-NESTED:       omp.precond.end:
+// IR-PCH-NESTED-NEXT:    ret void
+//
+//
+// IR-PCH-NESTED-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.omp_outlined.omp_outlined
+// IR-PCH-NESTED-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[B:%.*]]) #[[ATTR2]] {
+// IR-PCH-NESTED-NEXT:  entry:
+// IR-PCH-NESTED-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[B_ADDR:%.*]] = alloca ptr, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[I:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[J:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// IR-PCH-NESTED-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[I11:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    [[J12:%.*]] = alloca i32, align 4
+// IR-PCH-NESTED-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    store ptr [[B]], ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[B_ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// IR-PCH-NESTED-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
+// IR-PCH-NESTED-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP7]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// IR-PCH-NESTED-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
+// IR-PCH-NESTED-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
+// IR-PCH-NESTED-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
+// IR-PCH-NESTED-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[I]], align 4
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[J]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// IR-PCH-NESTED-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP8]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// IR-PCH-NESTED:       land.lhs.true:
+// IR-PCH-NESTED-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP9]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// IR-PCH-NESTED:       omp.precond.then:
+// IR-PCH-NESTED-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// IR-PCH-NESTED-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// IR-PCH-NESTED-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP14]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-PCH-NESTED-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP15]], [[TMP16]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// IR-PCH-NESTED:       cond.true:
+// IR-PCH-NESTED-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[COND_END:%.*]]
+// IR-PCH-NESTED:       cond.false:
+// IR-PCH-NESTED-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[COND_END]]
+// IR-PCH-NESTED:       cond.end:
+// IR-PCH-NESTED-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP17]], [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
+// IR-PCH-NESTED-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// IR-PCH-NESTED-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.cond:
+// IR-PCH-NESTED-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// IR-PCH-NESTED-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP20]], [[TMP21]]
+// IR-PCH-NESTED-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.body:
+// IR-PCH-NESTED-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP23]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// IR-PCH-NESTED-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
+// IR-PCH-NESTED-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
+// IR-PCH-NESTED-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP22]], [[CONV18]]
+// IR-PCH-NESTED-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
+// IR-PCH-NESTED-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
+// IR-PCH-NESTED-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
+// IR-PCH-NESTED-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP26]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
+// IR-PCH-NESTED-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
+// IR-PCH-NESTED-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
+// IR-PCH-NESTED-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP25]], [[CONV25]]
+// IR-PCH-NESTED-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// IR-PCH-NESTED-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP27]], 0
+// IR-PCH-NESTED-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
+// IR-PCH-NESTED-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
+// IR-PCH-NESTED-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
+// IR-PCH-NESTED-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
+// IR-PCH-NESTED-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP24]], [[MUL31]]
+// IR-PCH-NESTED-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
+// IR-PCH-NESTED-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
+// IR-PCH-NESTED-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
+// IR-PCH-NESTED-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP28:%.*]] = load i32, ptr [[I11]], align 4
+// IR-PCH-NESTED-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP28]] to i64
+// IR-PCH-NESTED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]]
+// IR-PCH-NESTED-NEXT:    [[TMP29:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+// IR-PCH-NESTED-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// IR-PCH-NESTED-NEXT:    [[MUL36:%.*]] = mul nsw i32 [[TMP29]], [[TMP30]]
+// IR-PCH-NESTED-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
+// IR-PCH-NESTED-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP31]])
+// IR-PCH-NESTED-NEXT:    [[ADD37:%.*]] = add nsw i32 [[MUL36]], [[CALL]]
+// IR-PCH-NESTED-NEXT:    [[TMP32:%.*]] = load i32, ptr [[I11]], align 4
+// IR-PCH-NESTED-NEXT:    [[IDXPROM38:%.*]] = sext i32 [[TMP32]] to i64
+// IR-PCH-NESTED-NEXT:    [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM38]]
+// IR-PCH-NESTED-NEXT:    store i32 [[ADD37]], ptr [[ARRAYIDX39]], align 4
+// IR-PCH-NESTED-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH-NESTED:       omp.body.continue:
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// IR-PCH-NESTED:       omp.inner.for.inc:
+// IR-PCH-NESTED-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    [[ADD40:%.*]] = add nsw i64 [[TMP33]], 1
+// IR-PCH-NESTED-NEXT:    store i64 [[ADD40]], ptr [[DOTOMP_IV]], align 8
+// IR-PCH-NESTED-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// IR-PCH-NESTED:       omp.inner.for.end:
+// IR-PCH-NESTED-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// IR-PCH-NESTED:       omp.loop.exit:
+// IR-PCH-NESTED-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NESTED-NEXT:    [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
+// IR-PCH-NESTED-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP35]])
+// IR-PCH-NESTED-NEXT:    br label [[OMP_PRECOND_END]]
+// IR-PCH-NESTED:       omp.precond.end:
+// IR-PCH-NESTED-NEXT:    ret void
+//
diff --git a/clang/test/OpenMP/target_teams_generic_loop_if_codegen.cpp b/clang/test/OpenMP/target_teams_generic_loop_if_codegen.cpp
index 1edcbfe2d7779..e1a6aad65b796 100644
--- a/clang/test/OpenMP/target_teams_generic_loop_if_codegen.cpp
+++ b/clang/test/OpenMP/target_teams_generic_loop_if_codegen.cpp
@@ -332,78 +332,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l51.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l51.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z9gtid_testv()
@@ -411,14 +341,14 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -586,78 +516,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l76.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l76.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z3fn4v()
@@ -665,14 +525,14 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -695,7 +555,6 @@ int main() {
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
@@ -725,82 +584,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTBOUND_ZERO_ADDR]], align 4
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined(ptr [[TMP11]], ptr [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
-// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z3fn5v()
@@ -808,14 +593,14 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -847,7 +632,6 @@ int main() {
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
@@ -878,30 +662,18 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
-// CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP11]] to i1
-// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
-// CHECK1:       omp_if.then:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
-// CHECK1:       omp_if.else:
-// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTBOUND_ZERO_ADDR]], align 4
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.omp_outlined.omp_outlined(ptr [[TMP12]], ptr [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
-// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    br label [[OMP_IF_END]]
-// CHECK1:       omp_if.end:
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK1-NEXT:    call void @_Z3fn6v()
+// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
@@ -910,85 +682,19 @@ int main() {
 // CHECK1-NEXT:    ret void
 //
 //
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
+// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
+// CHECK1-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    call void @_Z3fn6v()
-// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// CHECK1:       omp.body.continue:
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
-// CHECK1-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
-// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 // CHECK1-NEXT:    store i32 [[ARG]], ptr [[ARG_ADDR]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
@@ -1026,44 +732,61 @@ int main() {
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l64() #[[ATTR2]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARG_ADDR]], align 4
 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
-// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
+// CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP16]] to i1
+// CHECK1-NEXT:    [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8
+// CHECK1-NEXT:    store i8 [[FROMBOOL2]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
+// CHECK1-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP18]] to i1
+// CHECK1-NEXT:    br i1 [[TOBOOL3]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
 // CHECK1:       omp_if.then:
-// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
-// CHECK1-NEXT:    store i32 3, ptr [[TMP16]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
-// CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
-// CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
-// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
-// CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
-// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
-// CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
-// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
+// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP17]], ptr [[TMP19]], align 8
+// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP17]], ptr [[TMP20]], align 8
+// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
-// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
-// CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
-// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
-// CHECK1-NEXT:    store ptr null, ptr [[TMP23]], align 8
-// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
-// CHECK1-NEXT:    store i64 100, ptr [[TMP24]], align 8
-// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
-// CHECK1-NEXT:    store i64 0, ptr [[TMP25]], align 8
-// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
-// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
-// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
-// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP27]], align 4
-// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
-// CHECK1-NEXT:    store i32 0, ptr [[TMP28]], align 4
-// CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.region_id, ptr [[KERNEL_ARGS2]])
-// CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
-// CHECK1-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
-// CHECK1:       omp_offload.failed3:
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68() #[[ATTR2]]
-// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
-// CHECK1:       omp_offload.cont4:
+// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 3, ptr [[TMP24]], align 4
+// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 1, ptr [[TMP25]], align 4
+// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP22]], ptr [[TMP26]], align 8
+// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
+// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP28]], align 8
+// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP29]], align 8
+// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP30]], align 8
+// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP31]], align 8
+// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 100, ptr [[TMP32]], align 8
+// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP33]], align 8
+// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
+// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
+// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 0, ptr [[TMP36]], align 4
+// CHECK1-NEXT:    [[TMP37:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.region_id, ptr [[KERNEL_ARGS5]])
+// CHECK1-NEXT:    [[TMP38:%.*]] = icmp ne i32 [[TMP37]], 0
+// CHECK1-NEXT:    br i1 [[TMP38]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
+// CHECK1:       omp_offload.failed6:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68(i64 [[TMP17]]) #[[ATTR2]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
+// CHECK1:       omp_offload.cont7:
 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
 // CHECK1:       omp_if.else:
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68() #[[ATTR2]]
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68(i64 [[TMP17]]) #[[ATTR2]]
 // CHECK1-NEXT:    br label [[OMP_IF_END]]
 // CHECK1:       omp_if.end:
 // CHECK1-NEXT:    ret i32 0
@@ -1117,78 +840,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l60.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l60.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z3fn1v()
@@ -1196,14 +849,14 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -1226,7 +879,6 @@ int main() {
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
@@ -1256,82 +908,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void @__kmpc_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTBOUND_ZERO_ADDR]], align 4
-// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l64.omp_outlined.omp_outlined(ptr [[TMP11]], ptr [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
-// CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(ptr @[[GLOB3]], i32 [[TMP1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l64.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z3fn2v()
@@ -1339,29 +917,38 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68
-// CHECK1-SAME: () #[[ATTR1]] {
+// CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.omp_outlined)
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
+// CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP0]] to i1
+// CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
+// CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.omp_outlined, i64 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
@@ -1371,6 +958,7 @@ int main() {
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
@@ -1398,78 +986,8 @@ int main() {
 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l68.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK1-NEXT:    call void @_Z3fn3v()
@@ -1477,13 +995,13 @@ int main() {
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
diff --git a/clang/test/OpenMP/target_teams_generic_loop_private_codegen.cpp b/clang/test/OpenMP/target_teams_generic_loop_private_codegen.cpp
index 99416f76e409c..9b3d77f5b0adc 100644
--- a/clang/test/OpenMP/target_teams_generic_loop_private_codegen.cpp
+++ b/clang/test/OpenMP/target_teams_generic_loop_private_codegen.cpp
@@ -326,7 +326,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -340,7 +340,7 @@ int main() {
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124
 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
 // CHECK1-NEXT:    ret void
 //
 //
@@ -403,149 +403,48 @@ int main() {
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
-// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done3:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK1-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK1-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK1:       arrayctor.loop:
-// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK1:       arrayctor.cont:
-// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK1:       omp.inner.for.cond.cleanup:
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP12]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM3]]
-// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[VAR]], i64 4, i1 false)
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[SIVAR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP10]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM2]]
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX3]], ptr align 4 [[VAR]], i64 4, i1 false)
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
+// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[SIVAR]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
+// CHECK1-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
+// CHECK1-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN6]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done8:
+// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
+// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE7:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK1:       arraydestroy.done7:
 // CHECK1-NEXT:    ret void
 //
 //
@@ -596,7 +495,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -645,7 +544,7 @@ int main() {
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80
 // CHECK1-SAME: () #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
 // CHECK1-NEXT:    ret void
 //
 //
@@ -711,149 +610,45 @@ int main() {
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
-// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done5:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK1-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK1-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK1:       arrayctor.loop:
-// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK1:       arrayctor.cont:
-// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK1-NEXT:    store ptr [[VAR]], ptr [[_TMP3]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK1:       omp.inner.for.cond.cleanup:
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP13]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM5]]
-// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX6]], ptr align 4 [[TMP12]], i64 4, i1 false)
+// CHECK1-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM4:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM4]]
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX5]], ptr align 4 [[TMP10]], i64 4, i1 false)
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN8:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN8]], i64 2
+// CHECK1-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN7]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN8]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE9:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done9:
+// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
+// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK1:       arraydestroy.done8:
 // CHECK1-NEXT:    ret void
 //
 //
@@ -1059,7 +854,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -1073,7 +868,7 @@ int main() {
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124
 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1136,138 +931,41 @@ int main() {
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK3:       arraydestroy.done3:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK3-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK3-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK3-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK3:       arrayctor.loop:
-// CHECK3-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK3-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK3-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK3-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK3-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK3:       arrayctor.cont:
-// CHECK3-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK3:       omp.inner.for.cond.cleanup:
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK3-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP12]]
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP10]]
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX2]], ptr align 4 [[VAR]], i32 4, i1 false)
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[SIVAR]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
+// CHECK3-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN5:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
+// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN5]]
@@ -1323,7 +1021,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -1372,7 +1070,7 @@ int main() {
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80
 // CHECK3-SAME: () #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1438,138 +1136,38 @@ int main() {
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK3:       arraydestroy.done5:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK3-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK3-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK3:       arrayctor.loop:
-// CHECK3-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK3-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK3-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK3-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK3-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK3:       arrayctor.cont:
-// CHECK3-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK3-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK3:       omp.inner.for.cond.cleanup:
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK3-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP2]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP13]]
-// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP12]], i32 4, i1 false)
+// CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i32 4, i1 false)
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], 1
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
+// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
@@ -1760,7 +1358,7 @@ int main() {
 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104
 // CHECK5-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK5-NEXT:  entry:
-// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined)
+// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined)
 // CHECK5-NEXT:    ret void
 //
 //
@@ -1781,6 +1379,7 @@ int main() {
 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK5-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
@@ -1812,17 +1411,29 @@ int main() {
 // CHECK5-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK5-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
-// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
+// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
+// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK5-NEXT:    store i32 1, ptr [[G]], align 4
+// CHECK5-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK5-NEXT:    store volatile i32 1, ptr [[TMP8]], align 4
+// CHECK5-NEXT:    store i32 2, ptr [[SIVAR]], align 4
+// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
+// CHECK5-NEXT:    store ptr [[G]], ptr [[TMP9]], align 8
+// CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
+// CHECK5-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK5-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
+// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
+// CHECK5-NEXT:    store ptr [[SIVAR]], ptr [[TMP12]], align 8
+// CHECK5-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
+// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
-// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK5-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
@@ -1831,98 +1442,8 @@ int main() {
 // CHECK5-NEXT:    ret void
 //
 //
-// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined.omp_outlined
-// CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK5-NEXT:  entry:
-// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[G:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[G1:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK5-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
-// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK5-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK5-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK5-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK5-NEXT:    store ptr [[G1]], ptr [[_TMP3]], align 8
-// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK5:       cond.true:
-// CHECK5-NEXT:    br label [[COND_END:%.*]]
-// CHECK5:       cond.false:
-// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    br label [[COND_END]]
-// CHECK5:       cond.end:
-// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK5:       omp.inner.for.cond:
-// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK5-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK5-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK5:       omp.inner.for.body:
-// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
-// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK5-NEXT:    store i32 1, ptr [[G]], align 4
-// CHECK5-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK5-NEXT:    store volatile i32 1, ptr [[TMP10]], align 4
-// CHECK5-NEXT:    store i32 2, ptr [[SIVAR]], align 4
-// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
-// CHECK5-NEXT:    store ptr [[G]], ptr [[TMP11]], align 8
-// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK5-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK5-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
-// CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
-// CHECK5-NEXT:    store ptr [[SIVAR]], ptr [[TMP14]], align 8
-// CHECK5-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
-// CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// CHECK5:       omp.body.continue:
-// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK5:       omp.inner.for.inc:
-// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK5-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
-// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK5:       omp.inner.for.end:
-// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK5:       omp.loop.exit:
-// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
-// CHECK5-NEXT:    ret void
-//
-//
-// CHECK5-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_target_teams_generic_loop_private_codegen.cpp
-// CHECK5-SAME: () #[[ATTR0]] {
+// CHECK5-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_target_teams_generic_loop_private_codegen.cpp
+// CHECK5-SAME: () #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    call void @__cxx_global_var_init()
 // CHECK5-NEXT:    call void @__cxx_global_var_init.1()
@@ -1935,7 +1456,7 @@ int main() {
 // CHECK13-NEXT:  entry:
 // CHECK13-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK13-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
-// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
+// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
 // CHECK13-NEXT:    ret void
 //
 //
@@ -1998,35 +1519,48 @@ int main() {
 // CHECK13:       omp.inner.for.cond.cleanup:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
-// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
+// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
+// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
+// CHECK13-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK13-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP10]] to i64
+// CHECK13-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM2]]
+// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX3]], ptr align 4 [[VAR]], i64 4, i1 false)
+// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK13-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
+// CHECK13-NEXT:    store i32 [[ADD4]], ptr [[SIVAR]], align 4
+// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
-// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK13-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK13:       omp.loop.exit:
-// CHECK13-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
-// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
+// CHECK13-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
 // CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5:[0-9]+]]
-// CHECK13-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i64 2
+// CHECK13-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
+// CHECK13-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN6]], i64 2
 // CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK13:       arraydestroy.body:
-// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK13:       arraydestroy.done3:
+// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
+// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE7:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK13:       arraydestroy.done7:
 // CHECK13-NEXT:    ret void
 //
 //
@@ -2040,120 +1574,6 @@ int main() {
 // CHECK13-NEXT:    ret void
 //
 //
-// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined
-// CHECK13-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR0]] {
-// CHECK13-NEXT:  entry:
-// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK13-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK13-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK13-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK13-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK13-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK13:       arrayctor.loop:
-// CHECK13-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK13-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]]) #[[ATTR4]]
-// CHECK13-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK13-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK13-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK13:       arrayctor.cont:
-// CHECK13-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
-// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK13:       cond.true:
-// CHECK13-NEXT:    br label [[COND_END:%.*]]
-// CHECK13:       cond.false:
-// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    br label [[COND_END]]
-// CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK13:       omp.inner.for.cond:
-// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK13:       omp.inner.for.cond.cleanup:
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK13:       omp.inner.for.body:
-// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
-// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
-// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK13-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK13-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP12]] to i64
-// CHECK13-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM3]]
-// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[VAR]], i64 4, i1 false)
-// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK13-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
-// CHECK13-NEXT:    store i32 [[ADD5]], ptr [[SIVAR]], align 4
-// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// CHECK13:       omp.body.continue:
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK13:       omp.inner.for.inc:
-// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK13-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK13:       omp.inner.for.end:
-// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK13:       omp.loop.exit:
-// CHECK13-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
-// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
-// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK13:       arraydestroy.body:
-// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK13-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
-// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK13:       arraydestroy.done8:
-// CHECK13-NEXT:    ret void
-//
-//
 // CHECK13-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK13-NEXT:  entry:
@@ -2169,7 +1589,7 @@ int main() {
 // CHECK13-NEXT:  entry:
 // CHECK13-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK13-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
-// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
+// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
 // CHECK13-NEXT:    ret void
 //
 //
@@ -2235,17 +1655,27 @@ int main() {
 // CHECK13:       omp.inner.for.cond.cleanup:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK13:       omp.inner.for.body:
-// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
+// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
+// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
+// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
+// CHECK13-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK13-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK13-NEXT:    [[IDXPROM4:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK13-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM4]]
+// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX5]], ptr align 4 [[TMP10]], i64 4, i1 false)
+// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK13:       omp.body.continue:
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK13:       omp.inner.for.inc:
-// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK13-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK13:       omp.inner.for.end:
 // CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
@@ -2254,16 +1684,16 @@ int main() {
 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
 // CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i64 2
+// CHECK13-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
+// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN7]], i64 2
 // CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK13:       arraydestroy.body:
 // CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK13:       arraydestroy.done5:
+// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
+// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK13:       arraydestroy.done8:
 // CHECK13-NEXT:    ret void
 //
 //
@@ -2277,120 +1707,6 @@ int main() {
 // CHECK13-NEXT:    ret void
 //
 //
-// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined
-// CHECK13-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR0]] {
-// CHECK13-NEXT:  entry:
-// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK13-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK13-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK13-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK13-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK13-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK13-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK13:       arrayctor.loop:
-// CHECK13-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK13-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]]) #[[ATTR4]]
-// CHECK13-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK13-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK13-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK13:       arrayctor.cont:
-// CHECK13-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
-// CHECK13-NEXT:    store ptr [[VAR]], ptr [[_TMP3]], align 8
-// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK13:       cond.true:
-// CHECK13-NEXT:    br label [[COND_END:%.*]]
-// CHECK13:       cond.false:
-// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    br label [[COND_END]]
-// CHECK13:       cond.end:
-// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK13:       omp.inner.for.cond:
-// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK13-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK13-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK13:       omp.inner.for.cond.cleanup:
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK13:       omp.inner.for.body:
-// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
-// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK13-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
-// CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK13-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK13-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK13-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP13]] to i64
-// CHECK13-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM5]]
-// CHECK13-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX6]], ptr align 4 [[TMP12]], i64 4, i1 false)
-// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// CHECK13:       omp.body.continue:
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK13:       omp.inner.for.inc:
-// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], 1
-// CHECK13-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
-// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK13:       omp.inner.for.end:
-// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK13:       omp.loop.exit:
-// CHECK13-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
-// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAY_BEGIN8:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK13-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN8]], i64 2
-// CHECK13-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK13:       arraydestroy.body:
-// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK13-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK13-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK13-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN8]]
-// CHECK13-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE9:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK13:       arraydestroy.done9:
-// CHECK13-NEXT:    ret void
-//
-//
 // CHECK13-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK13-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK13-NEXT:  entry:
@@ -2449,7 +1765,7 @@ int main() {
 // CHECK15-NEXT:  entry:
 // CHECK15-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK15-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
-// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
+// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined)
 // CHECK15-NEXT:    ret void
 //
 //
@@ -2512,148 +1828,41 @@ int main() {
 // CHECK15:       omp.inner.for.cond.cleanup:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK15:       omp.inner.for.body:
-// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK15:       omp.inner.for.inc:
-// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK15-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK15:       omp.inner.for.end:
-// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK15:       omp.loop.exit:
-// CHECK15-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5:[0-9]+]]
-// CHECK15-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i32 2
-// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK15:       arraydestroy.body:
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK15:       arraydestroy.done3:
-// CHECK15-NEXT:    ret void
-//
-//
-// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
-// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
-// CHECK15-NEXT:  entry:
-// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
-// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK15-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
-// CHECK15-NEXT:    ret void
-//
-//
-// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l124.omp_outlined.omp_outlined
-// CHECK15-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR0]] {
-// CHECK15-NEXT:  entry:
-// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK15-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK15-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK15-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK15-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK15-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK15:       arrayctor.loop:
-// CHECK15-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK15-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]]) #[[ATTR4]]
-// CHECK15-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK15-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK15-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK15:       arrayctor.cont:
-// CHECK15-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
-// CHECK15-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK15:       cond.true:
-// CHECK15-NEXT:    br label [[COND_END:%.*]]
-// CHECK15:       cond.false:
-// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    br label [[COND_END]]
-// CHECK15:       cond.end:
-// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK15-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK15:       omp.inner.for.cond:
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK15-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK15:       omp.inner.for.cond.cleanup:
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK15:       omp.inner.for.body:
-// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK15-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK15-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK15-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP12]]
+// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK15-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK15-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP10]]
 // CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX2]], ptr align 4 [[VAR]], i32 4, i1 false)
-// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
+// CHECK15-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK15-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
 // CHECK15-NEXT:    store i32 [[ADD3]], ptr [[SIVAR]], align 4
 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK15:       omp.body.continue:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK15:       omp.inner.for.inc:
-// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK15-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK15-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK15:       omp.inner.for.end:
 // CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK15:       omp.loop.exit:
-// CHECK15-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
-// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
+// CHECK15-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK15-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
+// CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5:[0-9]+]]
 // CHECK15-NEXT:    [[ARRAY_BEGIN5:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
+// CHECK15-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
 // CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK15:       arraydestroy.body:
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK15-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
 // CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN5]]
@@ -2662,6 +1871,16 @@ int main() {
 // CHECK15-NEXT:    ret void
 //
 //
+// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
+// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
+// CHECK15-NEXT:  entry:
+// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK15-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
+// CHECK15-NEXT:    ret void
+//
+//
 // CHECK15-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK15-NEXT:  entry:
@@ -2677,7 +1896,7 @@ int main() {
 // CHECK15-NEXT:  entry:
 // CHECK15-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK15-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
-// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
+// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined)
 // CHECK15-NEXT:    ret void
 //
 //
@@ -2743,148 +1962,38 @@ int main() {
 // CHECK15:       omp.inner.for.cond.cleanup:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK15:       omp.inner.for.body:
-// CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK15:       omp.inner.for.inc:
-// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK15-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK15:       omp.inner.for.end:
-// CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK15:       omp.loop.exit:
-// CHECK15-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
-// CHECK15-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i32 2
-// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK15:       arraydestroy.body:
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
-// CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK15:       arraydestroy.done5:
-// CHECK15-NEXT:    ret void
-//
-//
-// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
-// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
-// CHECK15-NEXT:  entry:
-// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
-// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK15-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
-// CHECK15-NEXT:    ret void
-//
-//
-// CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l80.omp_outlined.omp_outlined
-// CHECK15-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR0]] {
-// CHECK15-NEXT:  entry:
-// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK15-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK15-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK15-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 4
-// CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK15-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
-// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK15-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK15-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK15-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK15-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK15:       arrayctor.loop:
-// CHECK15-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK15-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]]) #[[ATTR4]]
-// CHECK15-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK15-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK15-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK15:       arrayctor.cont:
-// CHECK15-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR4]]
-// CHECK15-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 4
-// CHECK15-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK15-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK15-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK15:       cond.true:
-// CHECK15-NEXT:    br label [[COND_END:%.*]]
-// CHECK15:       cond.false:
-// CHECK15-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    br label [[COND_END]]
-// CHECK15:       cond.end:
-// CHECK15-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK15-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK15-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK15:       omp.inner.for.cond:
 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK15-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK15-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK15:       omp.inner.for.cond.cleanup:
-// CHECK15-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK15:       omp.inner.for.body:
-// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK15-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK15-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK15-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK15-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK15-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 4
 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK15-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK15-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP2]], align 4
-// CHECK15-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK15-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP13]]
-// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP12]], i32 4, i1 false)
+// CHECK15-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
+// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i32 4, i1 false)
 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK15:       omp.body.continue:
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK15:       omp.inner.for.inc:
-// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK15-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], 1
+// CHECK15-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK15-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK15-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK15:       omp.inner.for.end:
 // CHECK15-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK15:       omp.loop.exit:
-// CHECK15-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK15-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
+// CHECK15-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK15-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// CHECK15-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR5]]
 // CHECK15-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK15-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
+// CHECK15-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
 // CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK15:       arraydestroy.body:
-// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK15-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
 // CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
@@ -2893,6 +2002,16 @@ int main() {
 // CHECK15-NEXT:    ret void
 //
 //
+// CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
+// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
+// CHECK15-NEXT:  entry:
+// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK15-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
+// CHECK15-NEXT:    ret void
+//
+//
 // CHECK15-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK15-NEXT:  entry:
@@ -2951,7 +2070,7 @@ int main() {
 // CHECK17-NEXT:  entry:
 // CHECK17-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK17-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
-// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined)
+// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined)
 // CHECK17-NEXT:    ret void
 //
 //
@@ -2972,6 +2091,7 @@ int main() {
 // CHECK17-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
 // CHECK17-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK17-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
 // CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK17-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
@@ -3003,111 +2123,33 @@ int main() {
 // CHECK17-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK17-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK17:       omp.inner.for.body:
-// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK17-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK17:       omp.inner.for.inc:
-// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK17-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK17:       omp.inner.for.end:
-// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK17:       omp.loop.exit:
-// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK17-NEXT:    ret void
-//
-//
-// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l104.omp_outlined.omp_outlined
-// CHECK17-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR0]] {
-// CHECK17-NEXT:  entry:
-// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[G:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[G1:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
-// CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK17-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK17-NEXT:    store ptr [[G1]], ptr [[_TMP3]], align 8
-// CHECK17-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK17:       cond.true:
-// CHECK17-NEXT:    br label [[COND_END:%.*]]
-// CHECK17:       cond.false:
-// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    br label [[COND_END]]
-// CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK17:       omp.inner.for.cond:
 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK17-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK17:       omp.inner.for.body:
-// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK17-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK17-NEXT:    store i32 1, ptr [[G]], align 4
-// CHECK17-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK17-NEXT:    store volatile i32 1, ptr [[TMP10]], align 4
+// CHECK17-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK17-NEXT:    store volatile i32 1, ptr [[TMP8]], align 4
 // CHECK17-NEXT:    store i32 2, ptr [[SIVAR]], align 4
-// CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 0
-// CHECK17-NEXT:    store ptr [[G]], ptr [[TMP11]], align 8
-// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK17-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK17-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
-// CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 2
-// CHECK17-NEXT:    store ptr [[SIVAR]], ptr [[TMP14]], align 8
+// CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 0
+// CHECK17-NEXT:    store ptr [[G]], ptr [[TMP9]], align 8
+// CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 1
+// CHECK17-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK17-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
+// CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 2
+// CHECK17-NEXT:    store ptr [[SIVAR]], ptr [[TMP12]], align 8
 // CHECK17-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR3:[0-9]+]]
 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK17:       omp.body.continue:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK17:       omp.inner.for.inc:
-// CHECK17-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK17-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK17-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK17:       omp.inner.for.end:
 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK17:       omp.loop.exit:
-// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK17-NEXT:    ret void
 //
diff --git a/clang/test/OpenMP/teams_generic_loop_codegen-1.cpp b/clang/test/OpenMP/teams_generic_loop_codegen-1.cpp
index d4a98f07fe24d..23c5c9db9c700 100644
--- a/clang/test/OpenMP/teams_generic_loop_codegen-1.cpp
+++ b/clang/test/OpenMP/teams_generic_loop_codegen-1.cpp
@@ -278,7 +278,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP41]], align 4
-// CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 // CHECK1-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -338,7 +338,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP70]], align 4
-// CHECK1-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
+// CHECK1-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
 // CHECK1-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
 // CHECK1-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 // CHECK1:       omp_offload.failed16:
@@ -356,7 +356,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[TH_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
 // CHECK1-NEXT:    store i64 [[TE]], ptr [[TE_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[TH]], ptr [[TH_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
@@ -364,8 +364,8 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
-// CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
+// CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -434,126 +434,28 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined, i64 [[TMP16]], i64 [[TMP18]], ptr [[TMP0]], ptr [[TMP1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
-// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK1:       omp.precond.end:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
-// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
-// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK1:       omp.precond.then:
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP6]] to i32
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
+// CHECK1-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
@@ -567,7 +469,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -636,126 +538,28 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined, i64 [[TMP16]], i64 [[TMP18]], ptr [[TMP0]], ptr [[TMP1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
-// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK1:       omp.precond.end:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
-// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
-// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK1:       omp.precond.then:
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP6]] to i32
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
-// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
+// CHECK1-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK1:       omp.precond.end:
 // CHECK1-NEXT:    ret void
@@ -865,7 +669,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP41]], align 4
-// CHECK3-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 // CHECK3-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -925,7 +729,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
 // CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP70]], align 4
-// CHECK3-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
+// CHECK3-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
 // CHECK3-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
 // CHECK3-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 // CHECK3:       omp_offload.failed16:
@@ -943,7 +747,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[TH_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
 // CHECK3-NEXT:    store i32 [[TE]], ptr [[TE_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TH]], ptr [[TH_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
@@ -951,8 +755,8 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
-// CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
+// CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1021,121 +825,27 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined, i32 [[TMP15]], i32 [[TMP16]], ptr [[TMP0]], ptr [[TMP1]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
-// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK3:       omp.precond.end:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
-// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
-// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK3:       omp.precond.then:
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I3]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP18]]
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP16]]
 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP19]], 1
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
+// CHECK3-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
@@ -1149,7 +859,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1218,121 +928,27 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined, i32 [[TMP15]], i32 [[TMP16]], ptr [[TMP0]], ptr [[TMP1]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
-// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK3:       omp.precond.end:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
-// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
-// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK3:       omp.precond.then:
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
-// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I3]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP18]]
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP16]]
 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP19]], 1
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
+// CHECK3-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK3:       omp.precond.end:
 // CHECK3-NEXT:    ret void
@@ -1424,7 +1040,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK9-NEXT:    store i32 0, ptr [[TMP35]], align 4
-// CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
+// CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
 // CHECK9-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK9-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK9:       omp_offload.failed:
@@ -1449,7 +1065,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
 // CHECK9-NEXT:    ret void
 //
 //
@@ -1521,129 +1137,28 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK9-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK9-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined, i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK9:       omp.inner.for.end:
-// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
-// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK9:       omp.precond.end:
-// CHECK9-NEXT:    ret void
-//
-//
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined
-// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
-// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
-// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK9:       omp.precond.then:
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
-// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
-// CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK9:       cond.true:
-// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK9-NEXT:    br label [[COND_END:%.*]]
-// CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    br label [[COND_END]]
-// CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
-// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
-// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
+// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
+// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
-// CHECK9-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
+// CHECK9-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
@@ -1735,7 +1250,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK11-NEXT:    store i32 0, ptr [[TMP35]], align 4
-// CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
+// CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
 // CHECK11-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK11-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK11:       omp_offload.failed:
@@ -1760,7 +1275,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
 // CHECK11-NEXT:    ret void
 //
 //
@@ -1832,124 +1347,27 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
-// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined, i32 [[TMP16]], i32 [[TMP17]], ptr [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK11:       omp.inner.for.end:
-// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP21]])
-// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK11:       omp.precond.end:
-// CHECK11-NEXT:    ret void
-//
-//
-// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined
-// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK11-NEXT:  entry:
-// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
-// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
-// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK11:       omp.precond.then:
-// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK11-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
-// CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK11:       cond.true:
-// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK11-NEXT:    br label [[COND_END:%.*]]
-// CHECK11:       cond.false:
-// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    br label [[COND_END]]
-// CHECK11:       cond.end:
-// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
-// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK11:       omp.inner.for.body:
-// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
+// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
-// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
-// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP19]]
+// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP17]]
 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
+// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
+// CHECK11-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
@@ -2009,7 +1427,7 @@ int main (int argc, char **argv) {
 // CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK17-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK17-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
+// CHECK17-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
 // CHECK17-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK17-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK17:       omp_offload.failed:
@@ -2028,7 +1446,7 @@ int main (int argc, char **argv) {
 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
-// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
+// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
 // CHECK17-NEXT:    ret void
 //
 //
@@ -2041,135 +1459,62 @@ int main (int argc, char **argv) {
 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
-// CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
-// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK17-NEXT:    store i32 122, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK17-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
-// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
-// CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK17:       cond.true:
-// CHECK17-NEXT:    br label [[COND_END:%.*]]
-// CHECK17:       cond.false:
-// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    br label [[COND_END]]
-// CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK17-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK17:       omp.inner.for.cond:
-// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
-// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK17:       omp.inner.for.body:
-// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK17:       omp.inner.for.inc:
-// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK17-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK17:       omp.inner.for.end:
-// CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK17:       omp.loop.exit:
-// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK17-NEXT:    ret void
-//
-//
-// CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined
-// CHECK17-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
-// CHECK17-NEXT:  entry:
-// CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
-// CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
-// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK17-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK17-NEXT:    store i32 122, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK17-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
+// CHECK17-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK17-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
+// CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK17-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK17:       cond.true:
 // CHECK17-NEXT:    br label [[COND_END:%.*]]
 // CHECK17:       cond.false:
-// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK17-NEXT:    br label [[COND_END]]
 // CHECK17:       cond.end:
-// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK17-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
+// CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK17-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK17:       omp.inner.for.cond:
-// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
+// CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK17:       omp.inner.for.body:
-// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK17-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
-// CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK17-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 // CHECK17-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK17:       omp.body.continue:
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK17:       omp.inner.for.inc:
-// CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
-// CHECK17-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
+// CHECK17-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK17:       omp.inner.for.end:
 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK17:       omp.loop.exit:
-// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK17-NEXT:    ret void
 //
 //
@@ -2227,7 +1572,7 @@ int main (int argc, char **argv) {
 // CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK19-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK19-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
+// CHECK19-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
 // CHECK19-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK19-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK19:       omp_offload.failed:
@@ -2246,7 +1591,7 @@ int main (int argc, char **argv) {
 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
+// CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
 // CHECK19-NEXT:    ret void
 //
 //
@@ -2294,95 +1639,26 @@ int main (int argc, char **argv) {
 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK19:       omp.inner.for.body:
-// CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
-// CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK19:       omp.inner.for.inc:
-// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK19-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK19:       omp.inner.for.end:
-// CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK19:       omp.loop.exit:
-// CHECK19-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK19-NEXT:    ret void
-//
-//
-// CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined
-// CHECK19-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
-// CHECK19-NEXT:  entry:
-// CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
-// CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK19-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK19-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
-// CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK19-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK19-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK19-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK19-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK19-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK19-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK19-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
-// CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK19:       cond.true:
-// CHECK19-NEXT:    br label [[COND_END:%.*]]
-// CHECK19:       cond.false:
-// CHECK19-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    br label [[COND_END]]
-// CHECK19:       cond.end:
-// CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK19-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK19-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK19:       omp.inner.for.cond:
 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK19:       omp.inner.for.body:
-// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK19-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
-// CHECK19-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP11]]
+// CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP9]]
 // CHECK19-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK19:       omp.body.continue:
 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK19:       omp.inner.for.inc:
-// CHECK19-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK19-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK19:       omp.inner.for.end:
 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK19:       omp.loop.exit:
-// CHECK19-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK19-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK19-NEXT:    ret void
 //
 //
@@ -2478,7 +1754,7 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
 // CHECK25-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK25-NEXT:    store i32 0, ptr [[TMP35]], align 4
-// CHECK25-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
+// CHECK25-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
 // CHECK25-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK25-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK25:       omp_offload.failed:
@@ -2505,7 +1781,7 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK25-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 // CHECK25-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
+// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
 // CHECK25-NEXT:    ret void
 //
 //
@@ -2577,129 +1853,28 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK25-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK25:       omp.inner.for.body:
-// CHECK25-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK25-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
-// CHECK25-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK25-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
-// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined, i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK25:       omp.inner.for.inc:
-// CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
-// CHECK25-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK25:       omp.inner.for.end:
-// CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK25:       omp.loop.exit:
-// CHECK25-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
-// CHECK25-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK25:       omp.precond.end:
-// CHECK25-NEXT:    ret void
-//
-//
-// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined
-// CHECK25-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK25-NEXT:  entry:
-// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK25-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
-// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[I4:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK25-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK25-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK25-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
-// CHECK25-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
-// CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
-// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
-// CHECK25-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK25-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK25-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
-// CHECK25-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK25-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK25-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK25-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK25-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
-// CHECK25-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK25:       omp.precond.then:
-// CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK25-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK25-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
-// CHECK25-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK25-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
-// CHECK25-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK25-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
-// CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK25-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK25-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
-// CHECK25-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK25:       cond.true:
-// CHECK25-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK25-NEXT:    br label [[COND_END:%.*]]
-// CHECK25:       cond.false:
-// CHECK25-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    br label [[COND_END]]
-// CHECK25:       cond.end:
-// CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
-// CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK25:       omp.inner.for.cond:
 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK25-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK25:       omp.inner.for.body:
-// CHECK25-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
+// CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// CHECK25-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
-// CHECK25-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
-// CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
+// CHECK25-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
+// CHECK25-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
 // CHECK25-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK25:       omp.body.continue:
 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK25:       omp.inner.for.inc:
-// CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
-// CHECK25-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK25-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK25-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK25-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK25:       omp.inner.for.end:
 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK25:       omp.loop.exit:
-// CHECK25-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
+// CHECK25-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK25-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK25:       omp.precond.end:
 // CHECK25-NEXT:    ret void
@@ -2778,7 +1953,7 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    store [3 x i32] [[TMP18]], ptr [[TMP30]], align 4
 // CHECK25-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK25-NEXT:    store i32 0, ptr [[TMP31]], align 4
-// CHECK25-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
+// CHECK25-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
 // CHECK25-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
 // CHECK25-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK25:       omp_offload.failed:
@@ -2794,15 +1969,15 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    [[TE_ADDR:%.*]] = alloca i64, align 8
 // CHECK25-NEXT:    [[TH_ADDR:%.*]] = alloca i64, align 8
 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
 // CHECK25-NEXT:    store i64 [[TE]], ptr [[TE_ADDR]], align 8
 // CHECK25-NEXT:    store i64 [[TH]], ptr [[TH_ADDR]], align 8
 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK25-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
-// CHECK25-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
-// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
+// CHECK25-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
+// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
 // CHECK25-NEXT:    ret void
 //
 //
@@ -2850,99 +2025,26 @@ int main (int argc, char **argv) {
 // CHECK25-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK25-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK25:       omp.inner.for.body:
-// CHECK25-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK25-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK25-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK25:       omp.inner.for.inc:
-// CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK25-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK25:       omp.inner.for.end:
-// CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK25:       omp.loop.exit:
-// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK25-NEXT:    ret void
-//
-//
-// CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined
-// CHECK25-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
-// CHECK25-NEXT:  entry:
-// CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK25-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK25-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK25-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK25-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK25-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK25-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK25-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK25-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
-// CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK25:       cond.true:
-// CHECK25-NEXT:    br label [[COND_END:%.*]]
-// CHECK25:       cond.false:
-// CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    br label [[COND_END]]
-// CHECK25:       cond.end:
-// CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK25-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK25:       omp.inner.for.cond:
 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK25-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK25-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK25:       omp.inner.for.body:
-// CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK25-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK25-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK25-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 // CHECK25-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK25:       omp.body.continue:
 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK25:       omp.inner.for.inc:
-// CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
-// CHECK25-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
+// CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK25-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
+// CHECK25-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK25:       omp.inner.for.end:
 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK25:       omp.loop.exit:
-// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK25-NEXT:    ret void
 //
 //
@@ -3038,7 +2140,7 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
 // CHECK27-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK27-NEXT:    store i32 0, ptr [[TMP35]], align 4
-// CHECK27-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
+// CHECK27-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
 // CHECK27-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
 // CHECK27-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK27:       omp_offload.failed:
@@ -3065,7 +2167,7 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
 // CHECK27-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
+// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
 // CHECK27-NEXT:    ret void
 //
 //
@@ -3137,124 +2239,27 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
 // CHECK27-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK27:       omp.inner.for.body:
-// CHECK27-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK27-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined, i32 [[TMP16]], i32 [[TMP17]], ptr [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK27:       omp.inner.for.inc:
-// CHECK27-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
-// CHECK27-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK27:       omp.inner.for.end:
-// CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK27:       omp.loop.exit:
-// CHECK27-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP21]])
-// CHECK27-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK27:       omp.precond.end:
-// CHECK27-NEXT:    ret void
-//
-//
-// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined
-// CHECK27-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK27-NEXT:  entry:
-// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[I3:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK27-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
-// CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
-// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
-// CHECK27-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK27-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK27-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
-// CHECK27-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK27-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
-// CHECK27-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK27-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK27-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
-// CHECK27-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK27:       omp.precond.then:
-// CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK27-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK27-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
-// CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK27-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
-// CHECK27-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK27:       cond.true:
-// CHECK27-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
-// CHECK27-NEXT:    br label [[COND_END:%.*]]
-// CHECK27:       cond.false:
-// CHECK27-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    br label [[COND_END]]
-// CHECK27:       cond.end:
-// CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
-// CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK27:       omp.inner.for.cond:
 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
-// CHECK27-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK27:       omp.inner.for.body:
-// CHECK27-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
+// CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
-// CHECK27-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
-// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP19]]
+// CHECK27-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
+// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP17]]
 // CHECK27-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK27:       omp.body.continue:
 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK27:       omp.inner.for.inc:
-// CHECK27-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
+// CHECK27-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
 // CHECK27-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK27:       omp.inner.for.end:
 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK27:       omp.loop.exit:
-// CHECK27-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
-// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
+// CHECK27-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK27-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK27-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK27:       omp.precond.end:
 // CHECK27-NEXT:    ret void
@@ -3333,7 +2338,7 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    store [3 x i32] [[TMP18]], ptr [[TMP30]], align 4
 // CHECK27-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK27-NEXT:    store i32 0, ptr [[TMP31]], align 4
-// CHECK27-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
+// CHECK27-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
 // CHECK27-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
 // CHECK27-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK27:       omp_offload.failed:
@@ -3349,15 +2354,15 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    [[TE_ADDR:%.*]] = alloca i32, align 4
 // CHECK27-NEXT:    [[TH_ADDR:%.*]] = alloca i32, align 4
 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
 // CHECK27-NEXT:    store i32 [[TE]], ptr [[TE_ADDR]], align 4
 // CHECK27-NEXT:    store i32 [[TH]], ptr [[TH_ADDR]], align 4
 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK27-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
-// CHECK27-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
-// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
+// CHECK27-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
+// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
 // CHECK27-NEXT:    ret void
 //
 //
@@ -3405,93 +2410,24 @@ int main (int argc, char **argv) {
 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK27:       omp.inner.for.body:
-// CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK27:       omp.inner.for.inc:
-// CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK27-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK27:       omp.inner.for.end:
-// CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK27:       omp.loop.exit:
-// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK27-NEXT:    ret void
-//
-//
-// CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined
-// CHECK27-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
-// CHECK27-NEXT:  entry:
-// CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK27-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK27-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
-// CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK27:       cond.true:
-// CHECK27-NEXT:    br label [[COND_END:%.*]]
-// CHECK27:       cond.false:
-// CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    br label [[COND_END]]
-// CHECK27:       cond.end:
-// CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK27-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK27:       omp.inner.for.cond:
 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK27:       omp.inner.for.body:
-// CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP11]]
+// CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP9]]
 // CHECK27-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK27:       omp.body.continue:
 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK27:       omp.inner.for.inc:
-// CHECK27-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 // CHECK27-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK27:       omp.inner.for.end:
 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK27:       omp.loop.exit:
-// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK27-NEXT:    ret void
 //
diff --git a/clang/test/OpenMP/teams_generic_loop_codegen.cpp b/clang/test/OpenMP/teams_generic_loop_codegen.cpp
index 2499fbb6811c9..85dcae26970bc 100644
--- a/clang/test/OpenMP/teams_generic_loop_codegen.cpp
+++ b/clang/test/OpenMP/teams_generic_loop_codegen.cpp
@@ -29,7 +29,7 @@ int foo() {
 // IR-NEXT:    [[I:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[J:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
+// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
 // IR-NEXT:    ret i32 0
 //
 //
@@ -96,277 +96,100 @@ int foo() {
 // IR-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // IR-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR:       omp.inner.for.body:
-// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// IR-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
-// IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 4, ptr @_Z3foov.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]], ptr [[J3]], ptr [[SUM1]])
+// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 10
+// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[DIV6:%.*]] = sdiv i32 [[TMP12]], 10
+// IR-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[DIV6]], 10
+// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL7]]
+// IR-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[SUB]], 1
+// IR-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
+// IR-NEXT:    store i32 [[ADD9]], ptr [[J3]], align 4
+// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
+// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
+// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
+// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i64 0, i64 [[IDXPROM]]
+// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[J3]], align 4
+// IR-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP15]] to i64
+// IR-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM10]]
+// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX11]], align 4
+// IR-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP16]], [[TMP13]]
+// IR-NEXT:    store i32 [[ADD12]], ptr [[ARRAYIDX11]], align 4
+// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR:       omp.body.continue:
 // IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // IR:       omp.inner.for.inc:
-// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
-// IR-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP17]], 1
+// IR-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
 // IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // IR:       omp.inner.for.end:
 // IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // IR:       omp.loop.exit:
-// IR-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP17]])
-// IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
-// IR-NEXT:    br i1 [[TMP19]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
+// IR-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
+// IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
+// IR-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 // IR:       .omp.lastprivate.then:
 // IR-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J3]], align 4
-// IR-NEXT:    store i32 [[TMP20]], ptr [[TMP0]], align 4
+// IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J3]], align 4
+// IR-NEXT:    store i32 [[TMP22]], ptr [[TMP0]], align 4
 // IR-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // IR:       .omp.lastprivate.done:
-// IR-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-NEXT:    store ptr [[SUM1]], ptr [[TMP21]], align 8
-// IR-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// IR-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// IR-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// IR-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// IR-NEXT:    store ptr [[SUM1]], ptr [[TMP23]], align 8
+// IR-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
+// IR-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // IR-NEXT:    ]
 // IR:       .omp.reduction.case1:
-// IR-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP25]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
+// IR-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
+// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP27]]
+// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // IR:       omp.arraycpy.body:
 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
-// IR-NEXT:    store i32 [[ADD7]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
+// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
+// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
+// IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
+// IR-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
+// IR-NEXT:    store i32 [[ADD15]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
+// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP25]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
-// IR:       omp.arraycpy.done10:
-// IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
-// IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR:       .omp.reduction.case2:
-// IR-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq ptr [[TMP1]], [[TMP28]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
-// IR:       omp.arraycpy.body12:
-// IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], align 4
-// IR-NEXT:    [[TMP30:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 [[TMP29]] monotonic, align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP28]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
+// IR-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP27]]
+// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
 // IR:       omp.arraycpy.done18:
-// IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR:       .omp.reduction.default:
-// IR-NEXT:    ret void
-//
-//
-// IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined
-// IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
-// IR-NEXT:  entry:
-// IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// IR-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// IR-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
-// IR-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
-// IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[J3:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[J5:%.*]] = alloca i32, align 4
-// IR-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// IR-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
-// IR-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
-// IR-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
-// IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
-// IR-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// IR-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
-// IR-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
-// IR-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// IR-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i32 0, i32 0, i32 0
-// IR-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// IR-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP4]]
-// IR-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// IR:       omp.arrayinit.body:
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// IR-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// IR:       omp.arrayinit.done:
-// IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
-// IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 99
-// IR-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// IR:       cond.true:
-// IR-NEXT:    br label [[COND_END:%.*]]
-// IR:       cond.false:
-// IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-NEXT:    br label [[COND_END]]
-// IR:       cond.end:
-// IR-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
-// IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// IR-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
-// IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// IR:       omp.inner.for.cond:
-// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
-// IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// IR-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// IR:       omp.inner.for.body:
-// IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP12]], 10
-// IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
-// IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// IR-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP14]], 10
-// IR-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
-// IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP13]], [[MUL8]]
-// IR-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
-// IR-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
-// IR-NEXT:    store i32 [[ADD10]], ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
-// IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i64 0, i64 [[IDXPROM]]
-// IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP17]] to i64
-// IR-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
-// IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP18]], [[TMP15]]
-// IR-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// IR:       omp.body.continue:
-// IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// IR:       omp.inner.for.inc:
-// IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP19]], 1
-// IR-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
-// IR:       omp.inner.for.end:
-// IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// IR:       omp.loop.exit:
-// IR-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
-// IR-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-NEXT:    store ptr [[SUM4]], ptr [[TMP22]], align 8
-// IR-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
-// IR-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// IR-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// IR-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// IR-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// IR-NEXT:    ]
-// IR:       .omp.reduction.case1:
-// IR-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP26]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR:       omp.arraycpy.body:
-// IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
-// IR-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP26]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
-// IR:       omp.arraycpy.done19:
-// IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
+// IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP25]], ptr @.gomp_critical_user_.reduction.var)
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.case2:
-// IR-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY20:%.*]] = icmp eq ptr [[TMP1]], [[TMP29]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY20]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY21:%.*]]
-// IR:       omp.arraycpy.body21:
-// IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST22:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST23:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-NEXT:    [[TMP30:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], align 4
-// IR-NEXT:    [[TMP31:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 [[TMP30]] monotonic, align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP29]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
-// IR:       omp.arraycpy.done27:
+// IR-NEXT:    [[TMP30:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
+// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq ptr [[TMP1]], [[TMP30]]
+// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
+// IR:       omp.arraycpy.body20:
+// IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
+// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
+// IR-NEXT:    [[TMP31:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 4
+// IR-NEXT:    [[TMP32:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 [[TMP31]] monotonic, align 4
+// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
+// IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
+// IR-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP30]]
+// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
+// IR:       omp.arraycpy.done26:
 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR:       .omp.reduction.default:
-// IR-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
-// IR-NEXT:    br i1 [[TMP33]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// IR:       .omp.lastprivate.then:
-// IR-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-NEXT:    [[TMP34:%.*]] = load i32, ptr [[J3]], align 4
-// IR-NEXT:    store i32 [[TMP34]], ptr [[TMP0]], align 4
-// IR-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// IR:       .omp.lastprivate.done:
-// IR-NEXT:    ret void
-//
-//
-// IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
-// IR-NEXT:  entry:
-// IR-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// IR-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// IR-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// IR-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// IR-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// IR-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// IR-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// IR-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
-// IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR:       omp.arraycpy.body:
-// IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// IR-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
-// IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
-// IR:       omp.arraycpy.done2:
 // IR-NEXT:    ret void
 //
 //
 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
-// IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
+// IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 // IR-NEXT:  entry:
 // IR-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // IR-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
@@ -402,7 +225,7 @@ int foo() {
 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 // IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
 // IR-PCH-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
+// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
 // IR-PCH-NEXT:    ret i32 0
 //
 //
@@ -469,277 +292,100 @@ int foo() {
 // IR-PCH-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 // IR-PCH-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // IR-PCH:       omp.inner.for.body:
-// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// IR-PCH-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// IR-PCH-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
-// IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 4, ptr @_Z3foov.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]], ptr [[J3]], ptr [[SUM1]])
+// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 10
+// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
+// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[DIV6:%.*]] = sdiv i32 [[TMP12]], 10
+// IR-PCH-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[DIV6]], 10
+// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL7]]
+// IR-PCH-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[SUB]], 1
+// IR-PCH-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
+// IR-PCH-NEXT:    store i32 [[ADD9]], ptr [[J3]], align 4
+// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
+// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
+// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i64 0, i64 [[IDXPROM]]
+// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[J3]], align 4
+// IR-PCH-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP15]] to i64
+// IR-PCH-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM10]]
+// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX11]], align 4
+// IR-PCH-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP16]], [[TMP13]]
+// IR-PCH-NEXT:    store i32 [[ADD12]], ptr [[ARRAYIDX11]], align 4
+// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// IR-PCH:       omp.body.continue:
 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // IR-PCH:       omp.inner.for.inc:
-// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
-// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// IR-PCH-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP17]], 1
+// IR-PCH-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // IR-PCH:       omp.inner.for.end:
 // IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // IR-PCH:       omp.loop.exit:
-// IR-PCH-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP17]])
-// IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
-// IR-PCH-NEXT:    br i1 [[TMP19]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
+// IR-PCH-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
+// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
+// IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// IR-PCH-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
+// IR-PCH-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 // IR-PCH:       .omp.lastprivate.then:
 // IR-PCH-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J3]], align 4
-// IR-PCH-NEXT:    store i32 [[TMP20]], ptr [[TMP0]], align 4
+// IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J3]], align 4
+// IR-PCH-NEXT:    store i32 [[TMP22]], ptr [[TMP0]], align 4
 // IR-PCH-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 // IR-PCH:       .omp.lastprivate.done:
-// IR-PCH-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-PCH-NEXT:    store ptr [[SUM1]], ptr [[TMP21]], align 8
-// IR-PCH-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// IR-PCH-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// IR-PCH-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// IR-PCH-NEXT:    store ptr [[SUM1]], ptr [[TMP23]], align 8
+// IR-PCH-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
+// IR-PCH-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // IR-PCH-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // IR-PCH-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // IR-PCH-NEXT:    ]
 // IR-PCH:       .omp.reduction.case1:
-// IR-PCH-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP25]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
+// IR-PCH-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP27]]
+// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // IR-PCH:       omp.arraycpy.body:
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
-// IR-PCH-NEXT:    store i32 [[ADD7]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
+// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
+// IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
+// IR-PCH-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
+// IR-PCH-NEXT:    store i32 [[ADD15]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP25]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH:       omp.arraycpy.done10:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH:       .omp.reduction.case2:
-// IR-PCH-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq ptr [[TMP1]], [[TMP28]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
-// IR-PCH:       omp.arraycpy.body12:
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
-// IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], align 4
-// IR-PCH-NEXT:    [[TMP30:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 [[TMP29]] monotonic, align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP28]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP27]]
+// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
 // IR-PCH:       omp.arraycpy.done18:
-// IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// IR-PCH:       .omp.reduction.default:
-// IR-PCH-NEXT:    ret void
-//
-//
-// IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined
-// IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
-// IR-PCH-NEXT:  entry:
-// IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// IR-PCH-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[J3:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 16
-// IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[J5:%.*]] = alloca i32, align 4
-// IR-PCH-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-PCH-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
-// IR-PCH-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
-// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// IR-PCH-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// IR-PCH-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
-// IR-PCH-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// IR-PCH-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
-// IR-PCH-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// IR-PCH-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i32 0, i32 0, i32 0
-// IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP4]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
-// IR-PCH:       omp.arrayinit.body:
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
-// IR-PCH-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
-// IR-PCH:       omp.arrayinit.done:
-// IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
-// IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP6]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 99
-// IR-PCH-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// IR-PCH:       cond.true:
-// IR-PCH-NEXT:    br label [[COND_END:%.*]]
-// IR-PCH:       cond.false:
-// IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// IR-PCH-NEXT:    br label [[COND_END]]
-// IR-PCH:       cond.end:
-// IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
-// IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// IR-PCH-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
-// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// IR-PCH:       omp.inner.for.cond:
-// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
-// IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
-// IR-PCH-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// IR-PCH:       omp.inner.for.body:
-// IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP12]], 10
-// IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
-// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
-// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP14]], 10
-// IR-PCH-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
-// IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP13]], [[MUL8]]
-// IR-PCH-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
-// IR-PCH-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
-// IR-PCH-NEXT:    store i32 [[ADD10]], ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
-// IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i64 0, i64 [[IDXPROM]]
-// IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP17]] to i64
-// IR-PCH-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
-// IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP18]], [[TMP15]]
-// IR-PCH-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
-// IR-PCH:       omp.body.continue:
-// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// IR-PCH:       omp.inner.for.inc:
-// IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP19]], 1
-// IR-PCH-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
-// IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
-// IR-PCH:       omp.inner.for.end:
-// IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// IR-PCH:       omp.loop.exit:
-// IR-PCH-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
-// IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
-// IR-PCH-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// IR-PCH-NEXT:    store ptr [[SUM4]], ptr [[TMP22]], align 8
-// IR-PCH-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
-// IR-PCH-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// IR-PCH-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// IR-PCH-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// IR-PCH-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// IR-PCH-NEXT:    ]
-// IR-PCH:       .omp.reduction.case1:
-// IR-PCH-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP26]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH:       omp.arraycpy.body:
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
-// IR-PCH-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP26]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH:       omp.arraycpy.done19:
-// IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
+// IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP25]], ptr @.gomp_critical_user_.reduction.var)
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.case2:
-// IR-PCH-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY20:%.*]] = icmp eq ptr [[TMP1]], [[TMP29]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY20]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY21:%.*]]
-// IR-PCH:       omp.arraycpy.body21:
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST22:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST23:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
-// IR-PCH-NEXT:    [[TMP30:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], align 4
-// IR-PCH-NEXT:    [[TMP31:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 [[TMP30]] monotonic, align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP29]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
-// IR-PCH:       omp.arraycpy.done27:
+// IR-PCH-NEXT:    [[TMP30:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq ptr [[TMP1]], [[TMP30]]
+// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
+// IR-PCH:       omp.arraycpy.body20:
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
+// IR-PCH-NEXT:    [[TMP31:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 4
+// IR-PCH-NEXT:    [[TMP32:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 [[TMP31]] monotonic, align 4
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
+// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP30]]
+// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
+// IR-PCH:       omp.arraycpy.done26:
 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // IR-PCH:       .omp.reduction.default:
-// IR-PCH-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
-// IR-PCH-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
-// IR-PCH-NEXT:    br i1 [[TMP33]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
-// IR-PCH:       .omp.lastprivate.then:
-// IR-PCH-NEXT:    store i32 10, ptr [[J3]], align 4
-// IR-PCH-NEXT:    [[TMP34:%.*]] = load i32, ptr [[J3]], align 4
-// IR-PCH-NEXT:    store i32 [[TMP34]], ptr [[TMP0]], align 4
-// IR-PCH-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
-// IR-PCH:       .omp.lastprivate.done:
-// IR-PCH-NEXT:    ret void
-//
-//
-// IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
-// IR-PCH-NEXT:  entry:
-// IR-PCH-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// IR-PCH-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// IR-PCH-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// IR-PCH-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// IR-PCH-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// IR-PCH-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
-// IR-PCH:       omp.arraycpy.body:
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
-// IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// IR-PCH-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
-// IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
-// IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
-// IR-PCH:       omp.arraycpy.done2:
 // IR-PCH-NEXT:    ret void
 //
 //
 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
-// IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
+// IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 // IR-PCH-NEXT:  entry:
 // IR-PCH-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // IR-PCH-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
diff --git a/clang/test/OpenMP/teams_generic_loop_collapse_codegen.cpp b/clang/test/OpenMP/teams_generic_loop_collapse_codegen.cpp
index c0c04986f147e..901b6552a22b6 100644
--- a/clang/test/OpenMP/teams_generic_loop_collapse_codegen.cpp
+++ b/clang/test/OpenMP/teams_generic_loop_collapse_codegen.cpp
@@ -157,7 +157,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -177,7 +177,7 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -227,114 +227,39 @@ int main (int argc, char **argv) {
 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 456
-// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 456
-// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
-// CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
+// CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
+// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
-// CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
-// CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
+// CHECK1-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
+// CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX8]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK1-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -393,7 +318,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -413,7 +338,7 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -463,108 +388,37 @@ int main (int argc, char **argv) {
 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
+// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 456
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
-// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i32 0, i32 [[TMP13]]
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i32 0, i32 [[TMP11]]
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP12]]
 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -693,7 +547,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK9-NEXT:    store i32 0, ptr [[TMP48]], align 4
-// CHECK9-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
+// CHECK9-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
 // CHECK9-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 // CHECK9-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK9:       omp_offload.failed:
@@ -725,7 +579,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
 // CHECK9-NEXT:    ret void
 //
 //
@@ -820,178 +674,58 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
-// CHECK9-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined, i64 [[TMP21]], i64 [[TMP22]], ptr [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], i64 [[TMP3]], ptr [[TMP4]])
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP23]], [[TMP24]]
-// CHECK9-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK9:       omp.inner.for.end:
-// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
-// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK9:       omp.precond.end:
-// CHECK9-NEXT:    ret void
-//
-//
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined
-// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
-// CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
-// CHECK9-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
-// CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
-// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
-// CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
-// CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
-// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
-// CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
-// CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
-// CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK9:       land.lhs.true:
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
-// CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
-// CHECK9:       omp.precond.then:
-// CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK9-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_LB]], align 8
-// CHECK9-NEXT:    store i64 [[TMP13]], ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
-// CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP16]], [[TMP17]]
-// CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK9:       cond.true:
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK9-NEXT:    br label [[COND_END:%.*]]
-// CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    br label [[COND_END]]
-// CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP18]], [[COND_TRUE]] ], [ [[TMP19]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
-// CHECK9-NEXT:    store i64 [[TMP20]], ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP21]], [[TMP22]]
-// CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP24]], 0
+// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK9-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP22]], 0
 // CHECK9-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
 // CHECK9-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
 // CHECK9-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
-// CHECK9-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP23]], [[CONV18]]
+// CHECK9-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP21]], [[CONV18]]
 // CHECK9-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
 // CHECK9-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
 // CHECK9-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
-// CHECK9-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP27]], 0
+// CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK9-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP25]], 0
 // CHECK9-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
 // CHECK9-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
 // CHECK9-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
-// CHECK9-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP26]], [[CONV25]]
-// CHECK9-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP28]], 0
+// CHECK9-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP24]], [[CONV25]]
+// CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP26]], 0
 // CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
 // CHECK9-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
 // CHECK9-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
 // CHECK9-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
-// CHECK9-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP25]], [[MUL31]]
+// CHECK9-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP23]], [[MUL31]]
 // CHECK9-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
 // CHECK9-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
 // CHECK9-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
 // CHECK9-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
-// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, ptr [[I11]], align 4
-// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP29]] to i64
-// CHECK9-NEXT:    [[TMP30:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP3]]
-// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 [[TMP30]]
-// CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
-// CHECK9-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
+// CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I11]], align 4
+// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP27]] to i64
+// CHECK9-NEXT:    [[TMP28:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP3]]
+// CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 [[TMP28]]
+// CHECK9-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J12]], align 4
+// CHECK9-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP29]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i64 [[IDXPROM36]]
 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX37]], align 4
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK9-NEXT:    [[ADD38:%.*]] = add nsw i64 [[TMP32]], 1
+// CHECK9-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK9-NEXT:    [[ADD38:%.*]] = add nsw i64 [[TMP30]], 1
 // CHECK9-NEXT:    store i64 [[ADD38]], ptr [[DOTOMP_IV]], align 8
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP34:%.*]] = load i32, ptr [[TMP33]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP34]])
+// CHECK9-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP32]])
 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK9:       omp.precond.end:
 // CHECK9-NEXT:    ret void
@@ -1043,7 +777,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
+// CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK9:       omp_offload.failed:
@@ -1059,7 +793,7 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
 // CHECK9-NEXT:    ret void
 //
 //
@@ -1109,113 +843,38 @@ int main (int argc, char **argv) {
 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK9:       omp.inner.for.end:
-// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK9-NEXT:    ret void
-//
-//
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined
-// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
-// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK9:       cond.true:
-// CHECK9-NEXT:    br label [[COND_END:%.*]]
-// CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    br label [[COND_END]]
-// CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
+// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 2
-// CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 2
-// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
-// CHECK9-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
-// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
-// CHECK9-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
-// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
+// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
+// CHECK9-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
+// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
+// CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
+// CHECK9-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
+// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
-// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
-// CHECK9-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
-// CHECK9-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
-// CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
+// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
+// CHECK9-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
+// CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
+// CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX8]], align 4
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK9-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK9-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK9-NEXT:    ret void
 //
 //
@@ -1343,7 +1002,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK11-NEXT:    store i32 0, ptr [[TMP47]], align 4
-// CHECK11-NEXT:    [[TMP48:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
+// CHECK11-NEXT:    [[TMP48:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
 // CHECK11-NEXT:    [[TMP49:%.*]] = icmp ne i32 [[TMP48]], 0
 // CHECK11-NEXT:    br i1 [[TMP49]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK11:       omp_offload.failed:
@@ -1375,7 +1034,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
 // CHECK11-NEXT:    ret void
 //
 //
@@ -1470,180 +1129,56 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
 // CHECK11-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
-// CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
-// CHECK11-NEXT:    [[TMP22:%.*]] = trunc i64 [[TMP21]] to i32
-// CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
-// CHECK11-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined, i32 [[TMP22]], i32 [[TMP24]], ptr [[TMP0]], ptr [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], ptr [[TMP4]])
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
-// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP25]], [[TMP26]]
-// CHECK11-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK11:       omp.inner.for.end:
-// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
-// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
-// CHECK11:       omp.precond.end:
-// CHECK11-NEXT:    ret void
-//
-//
-// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined
-// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
-// CHECK11-NEXT:  entry:
-// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
-// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
-// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
-// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
-// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
-// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[I13:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[J14:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
-// CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
-// CHECK11-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
-// CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
-// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
-// CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
-// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
-// CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
-// CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
-// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
-// CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
-// CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
-// CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
-// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
-// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
-// CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
-// CHECK11:       land.lhs.true:
-// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
-// CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
-// CHECK11:       omp.precond.then:
-// CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
-// CHECK11-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK11-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    [[CONV11:%.*]] = zext i32 [[TMP12]] to i64
-// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    [[CONV12:%.*]] = zext i32 [[TMP13]] to i64
-// CHECK11-NEXT:    store i64 [[CONV11]], ptr [[DOTOMP_LB]], align 8
-// CHECK11-NEXT:    store i64 [[CONV12]], ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
-// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK11-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
-// CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK11-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP16]], [[TMP17]]
-// CHECK11-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK11:       cond.true:
-// CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
-// CHECK11-NEXT:    br label [[COND_END:%.*]]
-// CHECK11:       cond.false:
-// CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    br label [[COND_END]]
-// CHECK11:       cond.end:
-// CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP18]], [[COND_TRUE]] ], [ [[TMP19]], [[COND_FALSE]] ]
-// CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
-// CHECK11-NEXT:    store i64 [[TMP20]], ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
-// CHECK11-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP21]], [[TMP22]]
-// CHECK11-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK11:       omp.inner.for.body:
+// CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK11-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP22]], 0
+// CHECK11-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
+// CHECK11-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
+// CHECK11-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
+// CHECK11-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP21]], [[CONV18]]
+// CHECK11-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
+// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
+// CHECK11-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
+// CHECK11-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP24]], 0
-// CHECK11-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
-// CHECK11-NEXT:    [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
-// CHECK11-NEXT:    [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
-// CHECK11-NEXT:    [[DIV21:%.*]] = sdiv i64 [[TMP23]], [[CONV20]]
-// CHECK11-NEXT:    [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
-// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
-// CHECK11-NEXT:    [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
-// CHECK11-NEXT:    store i32 [[CONV23]], ptr [[I13]], align 4
-// CHECK11-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP27]], 0
-// CHECK11-NEXT:    [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
-// CHECK11-NEXT:    [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
-// CHECK11-NEXT:    [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
-// CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i64 [[TMP26]], [[CONV27]]
-// CHECK11-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
-// CHECK11-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP28]], 0
-// CHECK11-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
-// CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
-// CHECK11-NEXT:    [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
-// CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
-// CHECK11-NEXT:    [[SUB34:%.*]] = sub nsw i64 [[TMP25]], [[MUL33]]
-// CHECK11-NEXT:    [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
-// CHECK11-NEXT:    [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
-// CHECK11-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
-// CHECK11-NEXT:    store i32 [[CONV37]], ptr [[J14]], align 4
-// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, ptr [[I13]], align 4
-// CHECK11-NEXT:    [[TMP30:%.*]] = mul nsw i32 [[TMP29]], [[TMP3]]
-// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 [[TMP30]]
-// CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J14]], align 4
-// CHECK11-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 [[TMP31]]
-// CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX38]], align 4
+// CHECK11-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK11-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP25]], 0
+// CHECK11-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
+// CHECK11-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
+// CHECK11-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
+// CHECK11-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP24]], [[CONV25]]
+// CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK11-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP26]], 0
+// CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
+// CHECK11-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
+// CHECK11-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
+// CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
+// CHECK11-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP23]], [[MUL31]]
+// CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
+// CHECK11-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
+// CHECK11-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
+// CHECK11-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
+// CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I11]], align 4
+// CHECK11-NEXT:    [[TMP28:%.*]] = mul nsw i32 [[TMP27]], [[TMP3]]
+// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 [[TMP28]]
+// CHECK11-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J12]], align 4
+// CHECK11-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 [[TMP29]]
+// CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX36]], align 4
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
-// CHECK11-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP32]], 1
-// CHECK11-NEXT:    store i64 [[ADD39]], ptr [[DOTOMP_IV]], align 8
+// CHECK11-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK11-NEXT:    [[ADD37:%.*]] = add nsw i64 [[TMP30]], 1
+// CHECK11-NEXT:    store i64 [[ADD37]], ptr [[DOTOMP_IV]], align 8
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP34:%.*]] = load i32, ptr [[TMP33]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP34]])
+// CHECK11-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP32]])
 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK11:       omp.precond.end:
 // CHECK11-NEXT:    ret void
@@ -1695,7 +1230,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
-// CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
+// CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK11:       omp_offload.failed:
@@ -1711,7 +1246,7 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
 // CHECK11-NEXT:    ret void
 //
 //
@@ -1761,106 +1296,35 @@ int main (int argc, char **argv) {
 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK11:       omp.inner.for.body:
-// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK11:       omp.inner.for.end:
-// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK11-NEXT:    ret void
-//
-//
-// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined
-// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
-// CHECK11-NEXT:  entry:
-// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
-// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
-// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
-// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
-// CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK11:       cond.true:
-// CHECK11-NEXT:    br label [[COND_END:%.*]]
-// CHECK11:       cond.false:
-// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    br label [[COND_END]]
-// CHECK11:       cond.end:
-// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK11:       omp.inner.for.cond:
 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK11:       omp.inner.for.body:
-// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
+// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 2
+// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
 // CHECK11-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
-// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
+// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
 // CHECK11-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
-// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i32 0, i32 [[TMP13]]
-// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
-// CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
+// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i32 0, i32 [[TMP11]]
+// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
+// CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP12]]
 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK11:       omp.body.continue:
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK11:       omp.inner.for.inc:
-// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK11:       omp.inner.for.end:
 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK11:       omp.loop.exit:
-// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
+// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 // CHECK11-NEXT:    ret void
 //
diff --git a/clang/test/OpenMP/teams_generic_loop_private_codegen.cpp b/clang/test/OpenMP/teams_generic_loop_private_codegen.cpp
index 303bce8c648c2..e955db129d1da 100644
--- a/clang/test/OpenMP/teams_generic_loop_private_codegen.cpp
+++ b/clang/test/OpenMP/teams_generic_loop_private_codegen.cpp
@@ -288,7 +288,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -302,7 +302,7 @@ int main() {
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96
 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined)
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined)
 // CHECK1-NEXT:    ret void
 //
 //
@@ -365,149 +365,48 @@ int main() {
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
-// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done3:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK1-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK1-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK1:       arrayctor.loop:
-// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK1:       arrayctor.cont:
-// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK1:       omp.inner.for.cond.cleanup:
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP12]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM3]]
-// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[VAR]], i64 4, i1 false)
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[SIVAR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP10]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i64 0, i64 [[IDXPROM2]]
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX3]], ptr align 4 [[VAR]], i64 4, i1 false)
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
+// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[SIVAR]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
+// CHECK1-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
+// CHECK1-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN6]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done8:
+// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
+// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE7:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK1:       arraydestroy.done7:
 // CHECK1-NEXT:    ret void
 //
 //
@@ -558,7 +457,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -607,7 +506,7 @@ int main() {
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56
 // CHECK1-SAME: () #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined)
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined)
 // CHECK1-NEXT:    ret void
 //
 //
@@ -673,149 +572,45 @@ int main() {
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
-// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
-// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done5:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK1-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK1-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
-// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK1:       arrayctor.loop:
-// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i64 1
-// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK1:       arrayctor.cont:
-// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK1-NEXT:    store ptr [[VAR]], ptr [[_TMP3]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK1:       omp.inner.for.cond.cleanup:
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
-// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP13]] to i64
-// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM5]]
-// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX6]], ptr align 4 [[TMP12]], i64 4, i1 false)
+// CHECK1-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[IDXPROM4:%.*]] = sext i32 [[TMP11]] to i64
+// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM4]]
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX5]], ptr align 4 [[TMP10]], i64 4, i1 false)
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP14]], 1
-// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
+// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAY_BEGIN8:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN8]], i64 2
+// CHECK1-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN7]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
-// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN8]]
-// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE9:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK1:       arraydestroy.done9:
+// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
+// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
+// CHECK1:       arraydestroy.done8:
 // CHECK1-NEXT:    ret void
 //
 //
@@ -1021,7 +816,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -1035,7 +830,7 @@ int main() {
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96
 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined)
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined)
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1098,138 +893,41 @@ int main() {
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAY_BEGIN2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN2]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN2]]
-// CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE3:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK3:       arraydestroy.done3:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK3-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
-// CHECK3-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S:%.*]], align 4
-// CHECK3-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK3:       arrayctor.loop:
-// CHECK3-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK3-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK3-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK3-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK3-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK3:       arrayctor.cont:
-// CHECK3-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK3:       omp.inner.for.cond.cleanup:
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK3-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP12]]
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 [[TMP10]]
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX2]], ptr align 4 [[VAR]], i32 4, i1 false)
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[SIVAR]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], [[TMP13]]
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR]], align 4
+// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[SIVAR]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP17]])
+// CHECK3-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP15]])
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN5:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
+// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN5]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP18]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP16]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN5]]
@@ -1285,7 +983,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -1334,7 +1032,7 @@ int main() {
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56
 // CHECK3-SAME: () #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined)
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined)
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1400,138 +1098,38 @@ int main() {
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined, i32 [[TMP7]], i32 [[TMP8]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP12]])
-// CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAY_BEGIN4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN4]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
-// CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
-// CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
-// CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN4]]
-// CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
-// CHECK3:       arraydestroy.done5:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
-// CHECK3-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
-// CHECK3-NEXT:    [[VAR:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
-// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr undef, ptr [[_TMP1]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
-// CHECK3-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
-// CHECK3:       arrayctor.loop:
-// CHECK3-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
-// CHECK3-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
-// CHECK3-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYCTOR_CUR]], i32 1
-// CHECK3-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
-// CHECK3-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
-// CHECK3:       arrayctor.cont:
-// CHECK3-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]])
-// CHECK3-NEXT:    store ptr [[VAR]], ptr [[_TMP2]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
-// CHECK3:       omp.inner.for.cond.cleanup:
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP2]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP11]]
-// CHECK3-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[_TMP2]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP13]]
-// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP12]], i32 4, i1 false)
+// CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i32 4, i1 false)
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], 1
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP16]])
+// CHECK3-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP14]])
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN6:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
+// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN6]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
-// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP17]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
+// CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP15]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN6]]
@@ -1726,7 +1324,7 @@ int main() {
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined)
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined)
 // CHECK9-NEXT:    ret void
 //
 //
@@ -1747,6 +1345,7 @@ int main() {
 // CHECK9-NEXT:    [[_TMP2:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
@@ -1778,112 +1377,34 @@ int main() {
 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK9:       omp.inner.for.end:
-// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
-// CHECK9-NEXT:    ret void
-//
-//
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined.omp_outlined
-// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR4]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[G:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[G1:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[_TMP3:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[SIVAR:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
-// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store ptr undef, ptr [[_TMP1]], align 8
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    store ptr [[G1]], ptr [[_TMP3]], align 8
-// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
-// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK9:       cond.true:
-// CHECK9-NEXT:    br label [[COND_END:%.*]]
-// CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    br label [[COND_END]]
-// CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
-// CHECK9-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[G]], align 4
-// CHECK9-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK9-NEXT:    store volatile i32 1, ptr [[TMP10]], align 4
+// CHECK9-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK9-NEXT:    store volatile i32 1, ptr [[TMP8]], align 4
 // CHECK9-NEXT:    store i32 2, ptr [[SIVAR]], align 4
-// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
-// CHECK9-NEXT:    store ptr [[G]], ptr [[TMP11]], align 8
-// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
-// CHECK9-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[_TMP3]], align 8
-// CHECK9-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
-// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
-// CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[TMP14]], align 8
+// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
+// CHECK9-NEXT:    store ptr [[G]], ptr [[TMP9]], align 8
+// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
+// CHECK9-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[_TMP2]], align 8
+// CHECK9-NEXT:    store ptr [[TMP11]], ptr [[TMP10]], align 8
+// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
+// CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[TMP12]], align 8
 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], 1
-// CHECK9-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK9-NEXT:    ret void
 //
 //
diff --git a/clang/test/OpenMP/teams_generic_loop_reduction_codegen.cpp b/clang/test/OpenMP/teams_generic_loop_reduction_codegen.cpp
index d80c003c01099..f0067cb93d9b0 100644
--- a/clang/test/OpenMP/teams_generic_loop_reduction_codegen.cpp
+++ b/clang/test/OpenMP/teams_generic_loop_reduction_codegen.cpp
@@ -141,7 +141,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP19]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK1-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -157,7 +157,7 @@ int main() {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined, ptr [[SIVAR_ADDR]])
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined, ptr [[SIVAR_ADDR]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -208,165 +208,50 @@ int main() {
 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[SIVAR1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 8
-// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// CHECK1-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// CHECK1-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// CHECK1-NEXT:    ]
-// CHECK1:       .omp.reduction.case1:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK1:       .omp.reduction.case2:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
-// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK1:       .omp.reduction.default:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[SIVAR2:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[SIVAR2]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
-// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[SIVAR2]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
+// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1
-// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
+// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
-// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK1-NEXT:    store ptr [[SIVAR2]], ptr [[TMP14]], align 8
-// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
+// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK1-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK1-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK1-NEXT:    ]
 // CHECK1:       .omp.reduction.case1:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], ptr @.gomp_critical_user_.reduction.var)
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
+// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
+// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       .omp.reduction.case2:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       .omp.reduction.default:
 // CHECK1-NEXT:    ret void
 //
 //
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
-// CHECK1-NEXT:    ret void
-//
-//
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func
-// CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
+// CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
@@ -435,7 +320,7 @@ int main() {
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP19]], align 4
-// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK1-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
@@ -450,7 +335,7 @@ int main() {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[T_VAR_ADDR]])
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[T_VAR_ADDR]])
 // CHECK1-NEXT:    ret void
 //
 //
@@ -501,163 +386,48 @@ int main() {
 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[T_VAR1]])
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK1:       omp.inner.for.end:
-// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 8
-// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// CHECK1-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// CHECK1-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// CHECK1-NEXT:    ]
-// CHECK1:       .omp.reduction.case1:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK1:       .omp.reduction.case2:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
-// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK1:       .omp.reduction.default:
-// CHECK1-NEXT:    ret void
-//
-//
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined
-// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[T_VAR2:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
-// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK1-NEXT:    store i32 0, ptr [[T_VAR2]], align 4
-// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1
-// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK1:       cond.true:
-// CHECK1-NEXT:    br label [[COND_END:%.*]]
-// CHECK1:       cond.false:
-// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    br label [[COND_END]]
-// CHECK1:       cond.end:
-// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK1:       omp.inner.for.body:
-// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[T_VAR2]], align 4
-// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
-// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[T_VAR2]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
+// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[T_VAR1]], align 4
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
-// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1
-// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
+// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
-// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
-// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK1-NEXT:    store ptr [[T_VAR2]], ptr [[TMP14]], align 8
-// CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
+// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP12]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK1-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK1-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK1-NEXT:    ]
 // CHECK1:       .omp.reduction.case1:
-// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR2]], align 4
-// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[TMP0]], align 4
-// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], ptr @.gomp_critical_user_.reduction.var)
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
+// CHECK1-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
+// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       .omp.reduction.case2:
-// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR2]], align 4
-// CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK1:       .omp.reduction.default:
 // CHECK1-NEXT:    ret void
 //
 //
-// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
-// CHECK1-NEXT:  entry:
-// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
-// CHECK1-NEXT:    ret void
-//
-//
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func
 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
 // CHECK1-NEXT:  entry:
@@ -726,7 +496,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP19]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK3-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -742,7 +512,7 @@ int main() {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined, ptr [[SIVAR_ADDR]])
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined, ptr [[SIVAR_ADDR]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -793,161 +563,50 @@ int main() {
 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[SIVAR1]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
-// CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// CHECK3-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// CHECK3-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// CHECK3-NEXT:    ]
-// CHECK3:       .omp.reduction.case1:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
-// CHECK3-NEXT:    store i32 [[ADD3]], ptr [[TMP0]], align 4
-// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK3:       .omp.reduction.case2:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
-// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK3:       .omp.reduction.default:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
-// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
-// CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
+// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 4
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK3-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK3-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK3-NEXT:    ]
 // CHECK3:       .omp.reduction.case1:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
-// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], ptr @.gomp_critical_user_.reduction.var)
+// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK3:       .omp.reduction.case2:
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK3:       .omp.reduction.default:
 // CHECK3-NEXT:    ret void
 //
 //
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
-// CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
-// CHECK3-NEXT:    ret void
-//
-//
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp.reduction.reduction_func
-// CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
+// CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
@@ -1016,7 +675,7 @@ int main() {
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP19]], align 4
-// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB4]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK3-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
@@ -1031,7 +690,7 @@ int main() {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[T_VAR_ADDR]])
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[T_VAR_ADDR]])
 // CHECK3-NEXT:    ret void
 //
 //
@@ -1082,159 +741,48 @@ int main() {
 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[T_VAR1]])
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK3:       omp.inner.for.end:
-// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
-// CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP12]], align 4
-// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// CHECK3-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// CHECK3-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// CHECK3-NEXT:    ]
-// CHECK3:       .omp.reduction.case1:
-// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
-// CHECK3-NEXT:    store i32 [[ADD3]], ptr [[TMP0]], align 4
-// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK3:       .omp.reduction.case2:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
-// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK3:       .omp.reduction.default:
-// CHECK3-NEXT:    ret void
-//
-//
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined
-// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
-// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
-// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
-// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK3-NEXT:    store i32 0, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1
-// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK3:       cond.true:
-// CHECK3-NEXT:    br label [[COND_END:%.*]]
-// CHECK3:       cond.false:
-// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    br label [[COND_END]]
-// CHECK3:       cond.end:
-// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK3:       omp.inner.for.body:
-// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[T_VAR1]], align 4
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
-// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
-// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
-// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
-// CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 4
-// CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
+// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP12]], align 4
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK3-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK3-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK3-NEXT:    ]
 // CHECK3:       .omp.reduction.case1:
-// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
-// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], ptr @.gomp_critical_user_.reduction.var)
+// CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK3:       .omp.reduction.case2:
-// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
-// CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR1]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK3:       .omp.reduction.default:
 // CHECK3-NEXT:    ret void
 //
 //
-// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
-// CHECK3-NEXT:  entry:
-// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
-// CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
-// CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
-// CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
-// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
-// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
-// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
-// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
-// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
-// CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
-// CHECK3-NEXT:    ret void
-//
-//
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func
 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
 // CHECK3-NEXT:  entry:
@@ -1270,7 +818,7 @@ int main() {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined, ptr [[SIVAR_ADDR]])
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined, ptr [[SIVAR_ADDR]])
 // CHECK9-NEXT:    ret void
 //
 //
@@ -1288,6 +836,7 @@ int main() {
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
@@ -1321,169 +870,53 @@ int main() {
 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
-// CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
-// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[SIVAR1]])
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
-// CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
-// CHECK9:       omp.inner.for.end:
-// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
-// CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
-// CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 8
-// CHECK9-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK9-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
-// CHECK9-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
-// CHECK9-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
-// CHECK9-NEXT:    ]
-// CHECK9:       .omp.reduction.case1:
-// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
-// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[TMP0]], align 4
-// CHECK9-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
-// CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK9:       .omp.reduction.case2:
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR1]], align 4
-// CHECK9-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
-// CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
-// CHECK9:       .omp.reduction.default:
-// CHECK9-NEXT:    ret void
-//
-//
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp_outlined
-// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
-// CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[SIVAR2:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
-// CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
-// CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
-// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
-// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
-// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
-// CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
-// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
-// CHECK9-NEXT:    store i32 0, ptr [[SIVAR2]], align 4
-// CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
-// CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
-// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
-// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1
-// CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
-// CHECK9:       cond.true:
-// CHECK9-NEXT:    br label [[COND_END:%.*]]
-// CHECK9:       cond.false:
-// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    br label [[COND_END]]
-// CHECK9:       cond.end:
-// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
-// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
-// CHECK9-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
-// CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
-// CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
-// CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
-// CHECK9:       omp.inner.for.body:
-// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
-// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
-// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]]
-// CHECK9-NEXT:    store i32 [[ADD4]], ptr [[SIVAR2]], align 4
-// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
-// CHECK9-NEXT:    store ptr [[SIVAR2]], ptr [[TMP13]], align 8
+// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
+// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
+// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
+// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
+// CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP11]], align 8
 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
-// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
-// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], 1
-// CHECK9-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
-// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
-// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
-// CHECK9-NEXT:    store ptr [[SIVAR2]], ptr [[TMP15]], align 8
-// CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
-// CHECK9-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
+// CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP13]], align 8
+// CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
+// CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 // CHECK9-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 // CHECK9-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 // CHECK9-NEXT:    ]
 // CHECK9:       .omp.reduction.case1:
-// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP0]], align 4
-// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
-// CHECK9-NEXT:    store i32 [[ADD6]], ptr [[TMP0]], align 4
-// CHECK9-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP4]], ptr @.gomp_critical_user_.reduction.var)
+// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
+// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
+// CHECK9-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
+// CHECK9-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK9:       .omp.reduction.case2:
-// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR2]], align 4
-// CHECK9-NEXT:    [[TMP20:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP19]] monotonic, align 4
+// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
+// CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 // CHECK9:       .omp.reduction.default:
 // CHECK9-NEXT:    ret void
 //
 //
-// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp_outlined.omp.reduction.reduction_func
-// CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
-// CHECK9-NEXT:  entry:
-// CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
-// CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
-// CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
-// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
-// CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
-// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
-// CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
-// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
-// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
-// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
-// CHECK9-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
-// CHECK9-NEXT:    ret void
-//
-//
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45.omp_outlined.omp.reduction.reduction_func
-// CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
+// CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8

From d347235bddbeba2a72d94ebe9d8f98dc675c3776 Mon Sep 17 00:00:00 2001
From: Christopher Di Bella <cjdb@google.com>
Date: Wed, 10 Apr 2024 13:15:22 -0700
Subject: [PATCH 066/886] [Flang] responds to Clang Tidy feedback (#87847)

Line 267: performance-unnecessary-copy-initialization
Line 592: readability-container-size-empty
---
 clang/lib/Driver/ToolChains/Flang.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/lib/Driver/ToolChains/Flang.cpp b/clang/lib/Driver/ToolChains/Flang.cpp
index 9699443603d36..b00068c8098b9 100644
--- a/clang/lib/Driver/ToolChains/Flang.cpp
+++ b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -264,7 +264,7 @@ static void addVSDefines(const ToolChain &TC, const ArgList &Args,
   CmdArgs.push_back(Args.MakeArgString("-D_MSC_FULL_VER=" + Twine(ver)));
   CmdArgs.push_back(Args.MakeArgString("-D_WIN32"));
 
-  llvm::Triple triple = TC.getTriple();
+  const llvm::Triple &triple = TC.getTriple();
   if (triple.isAArch64()) {
     CmdArgs.push_back("-D_M_ARM64=1");
   } else if (triple.isX86() && triple.isArch32Bit()) {
@@ -589,7 +589,7 @@ static void addFloatingPointOptions(const Driver &D, const ArgList &Args,
 
   if (!HonorINFs && !HonorNaNs && AssociativeMath && ReciprocalMath &&
       ApproxFunc && !SignedZeros &&
-      (FPContract == "fast" || FPContract == "")) {
+      (FPContract == "fast" || FPContract.empty())) {
     CmdArgs.push_back("-ffast-math");
     return;
   }

From 05093e243859a371f96ffa1c320a4b51579c3da7 Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <1802579+farzonl@users.noreply.github.com>
Date: Wed, 10 Apr 2024 16:27:44 -0400
Subject: [PATCH 067/886] [Spirv][HLSL] Add OpAll lowering and float vec
 support (#87952)

The main point of this change was to add support for HLSL's all
intrinsic.
In the process of doing that I found a few issues around creating an
`OpConstantComposite` via `buildZerosVal`.

First the current code didn't support floats so the process of adding
`buildZerosValF` meant I needed a
float version of `getOrCreateIntConstVector`. After doing so I renamed
both versions to `getOrCreateConstVector`. That meant I needed to create
a float type version of `getOrCreateIntCompositeOrNull`. Luckily the
type information was low for this function so was able to split it out
into a helpwe and rename `getOrCreateIntCompositeOrNull` to
`getOrCreateCompositeOrNull` With the exception of type handling
differences of the code and Null vs 0 Constant Op codes these functions
should be identical.

To handle scalar floats I could not use `buildConstantFP` like this PR
did:
https://github.com/llvm/llvm-project/commit/0a2aaab5aba46#diff-733a189c5a8c3211f3a04fd6e719952a3fa231eadd8a7f11e6ecf1e584d57411R1603
because that would create too many superfluous registers (that causes
problems in the validator), I had to create a float version of
`getOrCreateConstInt` which I called `getOrCreateConstFP`.
similar problems with doing it like this:
https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/SPIRV/SPIRVBuiltins.cpp#L1540.

`buildZerosValF` also has a use of a function `getZeroFP`. This is
because half, float, and double scalar values of 0 would collide in
`SPIRVDuplicatesTracker<Constant> CT` if you use `APFloat(0.0f)`.

`getORCreateConstFP` needed its own version of `getOrCreateConstIntReg`
which I called `getOrCreateConstFloatReg` The one difference in this
function is `getOrCreateConstFloatReg` returns a bit width so we don't
have to call `getScalarOrVectorBitWidth` twice ie when it is used again
in `getOrCreateConstFP` for `OpConstantF` `addNumImm`.

`getOrCreateConstFloatReg` needed an `assignFloatTypeToVReg` helper
which called a `getOrCreateSPIRVFloatType` helper. There was no
equivalent IntegerType::get for floats so I handled this with a switch
statement on bit widths to get the right LLVM float type.

Finally, there is the use of `bool ZeroAsNull = STI.isOpenCLEnv();` This
is partly a cosmetic change. When Zeros are treated as nulls, we don't
create `OpConstantComposite` vectors which is something we do in the
DXCs SPIRV backend. The DXC SPIRV backend also does not use
`OpConstantNull`. Finally, I needed a means to test the behavior of the
OpConstantNull and `OpConstantComposite` changes and this was one way I
could do that via the same tests.
---
 llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp | 218 +++++++++++++++---
 llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h   |  42 +++-
 .../Target/SPIRV/SPIRVInstructionSelector.cpp |  98 +++++++-
 .../test/CodeGen/SPIRV/hlsl-intrinsics/all.ll | 187 +++++++++++++++
 4 files changed, 506 insertions(+), 39 deletions(-)
 create mode 100644 llvm/test/CodeGen/SPIRV/hlsl-intrinsics/all.ll

diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
index 9592f3e81b402..70197e948c658 100644
--- a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
@@ -20,7 +20,12 @@
 #include "SPIRVSubtarget.h"
 #include "SPIRVTargetMachine.h"
 #include "SPIRVUtils.h"
+#include "llvm/ADT/APInt.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/Type.h"
 #include "llvm/IR/TypedPointerType.h"
+#include "llvm/Support/Casting.h"
+#include <cassert>
 
 using namespace llvm;
 SPIRVGlobalRegistry::SPIRVGlobalRegistry(unsigned PointerSize)
@@ -35,6 +40,15 @@ SPIRVType *SPIRVGlobalRegistry::assignIntTypeToVReg(unsigned BitWidth,
   return SpirvType;
 }
 
+SPIRVType *
+SPIRVGlobalRegistry::assignFloatTypeToVReg(unsigned BitWidth, Register VReg,
+                                           MachineInstr &I,
+                                           const SPIRVInstrInfo &TII) {
+  SPIRVType *SpirvType = getOrCreateSPIRVFloatType(BitWidth, I, TII);
+  assignSPIRVTypeToVReg(SpirvType, VReg, *CurMF);
+  return SpirvType;
+}
+
 SPIRVType *SPIRVGlobalRegistry::assignVectTypeToVReg(
     SPIRVType *BaseType, unsigned NumElements, Register VReg, MachineInstr &I,
     const SPIRVInstrInfo &TII) {
@@ -151,6 +165,8 @@ SPIRVGlobalRegistry::getOrCreateConstIntReg(uint64_t Val, SPIRVType *SpvType,
   Register Res = DT.find(CI, CurMF);
   if (!Res.isValid()) {
     unsigned BitWidth = SpvType ? getScalarOrVectorBitWidth(SpvType) : 32;
+    // TODO: handle cases where the type is not 32bit wide
+    // TODO: https://github.com/llvm/llvm-project/issues/88129
     LLT LLTy = LLT::scalar(32);
     Res = CurMF->getRegInfo().createGenericVirtualRegister(LLTy);
     CurMF->getRegInfo().setRegClass(Res, &SPIRV::IDRegClass);
@@ -164,9 +180,83 @@ SPIRVGlobalRegistry::getOrCreateConstIntReg(uint64_t Val, SPIRVType *SpvType,
   return std::make_tuple(Res, CI, NewInstr);
 }
 
+std::tuple<Register, ConstantFP *, bool, unsigned>
+SPIRVGlobalRegistry::getOrCreateConstFloatReg(APFloat Val, SPIRVType *SpvType,
+                                              MachineIRBuilder *MIRBuilder,
+                                              MachineInstr *I,
+                                              const SPIRVInstrInfo *TII) {
+  const Type *LLVMFloatTy;
+  LLVMContext &Ctx = CurMF->getFunction().getContext();
+  unsigned BitWidth = 32;
+  if (SpvType)
+    LLVMFloatTy = getTypeForSPIRVType(SpvType);
+  else {
+    LLVMFloatTy = Type::getFloatTy(Ctx);
+    if (MIRBuilder)
+      SpvType = getOrCreateSPIRVType(LLVMFloatTy, *MIRBuilder);
+  }
+  bool NewInstr = false;
+  // Find a constant in DT or build a new one.
+  auto *const CI = ConstantFP::get(Ctx, Val);
+  Register Res = DT.find(CI, CurMF);
+  if (!Res.isValid()) {
+    if (SpvType)
+      BitWidth = getScalarOrVectorBitWidth(SpvType);
+    // TODO: handle cases where the type is not 32bit wide
+    // TODO: https://github.com/llvm/llvm-project/issues/88129
+    LLT LLTy = LLT::scalar(32);
+    Res = CurMF->getRegInfo().createGenericVirtualRegister(LLTy);
+    CurMF->getRegInfo().setRegClass(Res, &SPIRV::IDRegClass);
+    if (MIRBuilder)
+      assignTypeToVReg(LLVMFloatTy, Res, *MIRBuilder);
+    else
+      assignFloatTypeToVReg(BitWidth, Res, *I, *TII);
+    DT.add(CI, CurMF, Res);
+    NewInstr = true;
+  }
+  return std::make_tuple(Res, CI, NewInstr, BitWidth);
+}
+
+Register SPIRVGlobalRegistry::getOrCreateConstFP(APFloat Val, MachineInstr &I,
+                                                 SPIRVType *SpvType,
+                                                 const SPIRVInstrInfo &TII,
+                                                 bool ZeroAsNull) {
+  assert(SpvType);
+  ConstantFP *CI;
+  Register Res;
+  bool New;
+  unsigned BitWidth;
+  std::tie(Res, CI, New, BitWidth) =
+      getOrCreateConstFloatReg(Val, SpvType, nullptr, &I, &TII);
+  // If we have found Res register which is defined by the passed G_CONSTANT
+  // machine instruction, a new constant instruction should be created.
+  if (!New && (!I.getOperand(0).isReg() || Res != I.getOperand(0).getReg()))
+    return Res;
+  MachineInstrBuilder MIB;
+  MachineBasicBlock &BB = *I.getParent();
+  // In OpenCL OpConstantNull - Scalar floating point: +0.0 (all bits 0)
+  if (Val.isPosZero() && ZeroAsNull) {
+    MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpConstantNull))
+              .addDef(Res)
+              .addUse(getSPIRVTypeID(SpvType));
+  } else {
+    MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpConstantF))
+              .addDef(Res)
+              .addUse(getSPIRVTypeID(SpvType));
+    addNumImm(
+        APInt(BitWidth, CI->getValueAPF().bitcastToAPInt().getZExtValue()),
+        MIB);
+  }
+  const auto &ST = CurMF->getSubtarget();
+  constrainSelectedInstRegOperands(*MIB, *ST.getInstrInfo(),
+                                   *ST.getRegisterInfo(), *ST.getRegBankInfo());
+  return Res;
+}
+
 Register SPIRVGlobalRegistry::getOrCreateConstInt(uint64_t Val, MachineInstr &I,
                                                   SPIRVType *SpvType,
-                                                  const SPIRVInstrInfo &TII) {
+                                                  const SPIRVInstrInfo &TII,
+                                                  bool ZeroAsNull) {
   assert(SpvType);
   ConstantInt *CI;
   Register Res;
@@ -179,7 +269,7 @@ Register SPIRVGlobalRegistry::getOrCreateConstInt(uint64_t Val, MachineInstr &I,
     return Res;
   MachineInstrBuilder MIB;
   MachineBasicBlock &BB = *I.getParent();
-  if (Val) {
+  if (Val || !ZeroAsNull) {
     MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpConstantI))
               .addDef(Res)
               .addUse(getSPIRVTypeID(SpvType));
@@ -270,21 +360,46 @@ Register SPIRVGlobalRegistry::buildConstantFP(APFloat Val,
   return Res;
 }
 
-Register SPIRVGlobalRegistry::getOrCreateIntCompositeOrNull(
-    uint64_t Val, MachineInstr &I, SPIRVType *SpvType,
+Register SPIRVGlobalRegistry::getOrCreateBaseRegister(Constant *Val,
+                                                      MachineInstr &I,
+                                                      SPIRVType *SpvType,
+                                                      const SPIRVInstrInfo &TII,
+                                                      unsigned BitWidth) {
+  SPIRVType *Type = SpvType;
+  if (SpvType->getOpcode() == SPIRV::OpTypeVector ||
+      SpvType->getOpcode() == SPIRV::OpTypeArray) {
+    auto EleTypeReg = SpvType->getOperand(1).getReg();
+    Type = getSPIRVTypeForVReg(EleTypeReg);
+  }
+  if (Type->getOpcode() == SPIRV::OpTypeFloat) {
+    SPIRVType *SpvBaseType = getOrCreateSPIRVFloatType(BitWidth, I, TII);
+    return getOrCreateConstFP(dyn_cast<ConstantFP>(Val)->getValue(), I,
+                              SpvBaseType, TII);
+  }
+  assert(Type->getOpcode() == SPIRV::OpTypeInt);
+  SPIRVType *SpvBaseType = getOrCreateSPIRVIntegerType(BitWidth, I, TII);
+  return getOrCreateConstInt(Val->getUniqueInteger().getSExtValue(), I,
+                             SpvBaseType, TII);
+}
+
+Register SPIRVGlobalRegistry::getOrCreateCompositeOrNull(
+    Constant *Val, MachineInstr &I, SPIRVType *SpvType,
     const SPIRVInstrInfo &TII, Constant *CA, unsigned BitWidth,
-    unsigned ElemCnt) {
+    unsigned ElemCnt, bool ZeroAsNull) {
   // Find a constant vector in DT or build a new one.
   Register Res = DT.find(CA, CurMF);
+  // If no values are attached, the composite is null constant.
+  bool IsNull = Val->isNullValue() && ZeroAsNull;
   if (!Res.isValid()) {
-    SPIRVType *SpvBaseType = getOrCreateSPIRVIntegerType(BitWidth, I, TII);
     // SpvScalConst should be created before SpvVecConst to avoid undefined ID
     // error on validation.
     // TODO: can moved below once sorting of types/consts/defs is implemented.
     Register SpvScalConst;
-    if (Val)
-      SpvScalConst = getOrCreateConstInt(Val, I, SpvBaseType, TII);
-    // TODO: maybe use bitwidth of base type.
+    if (!IsNull)
+      SpvScalConst = getOrCreateBaseRegister(Val, I, SpvType, TII, BitWidth);
+
+    // TODO: handle cases where the type is not 32bit wide
+    // TODO: https://github.com/llvm/llvm-project/issues/88129
     LLT LLTy = LLT::scalar(32);
     Register SpvVecConst =
         CurMF->getRegInfo().createGenericVirtualRegister(LLTy);
@@ -293,7 +408,7 @@ Register SPIRVGlobalRegistry::getOrCreateIntCompositeOrNull(
     DT.add(CA, CurMF, SpvVecConst);
     MachineInstrBuilder MIB;
     MachineBasicBlock &BB = *I.getParent();
-    if (Val) {
+    if (!IsNull) {
       MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpConstantComposite))
                 .addDef(SpvVecConst)
                 .addUse(getSPIRVTypeID(SpvType));
@@ -313,20 +428,42 @@ Register SPIRVGlobalRegistry::getOrCreateIntCompositeOrNull(
   return Res;
 }
 
-Register
-SPIRVGlobalRegistry::getOrCreateConsIntVector(uint64_t Val, MachineInstr &I,
-                                              SPIRVType *SpvType,
-                                              const SPIRVInstrInfo &TII) {
+Register SPIRVGlobalRegistry::getOrCreateConstVector(uint64_t Val,
+                                                     MachineInstr &I,
+                                                     SPIRVType *SpvType,
+                                                     const SPIRVInstrInfo &TII,
+                                                     bool ZeroAsNull) {
   const Type *LLVMTy = getTypeForSPIRVType(SpvType);
   assert(LLVMTy->isVectorTy());
   const FixedVectorType *LLVMVecTy = cast<FixedVectorType>(LLVMTy);
   Type *LLVMBaseTy = LLVMVecTy->getElementType();
-  const auto ConstInt = ConstantInt::get(LLVMBaseTy, Val);
-  auto ConstVec =
-      ConstantVector::getSplat(LLVMVecTy->getElementCount(), ConstInt);
+  assert(LLVMBaseTy->isIntegerTy());
+  auto *ConstVal = ConstantInt::get(LLVMBaseTy, Val);
+  auto *ConstVec =
+      ConstantVector::getSplat(LLVMVecTy->getElementCount(), ConstVal);
   unsigned BW = getScalarOrVectorBitWidth(SpvType);
-  return getOrCreateIntCompositeOrNull(Val, I, SpvType, TII, ConstVec, BW,
-                                       SpvType->getOperand(2).getImm());
+  return getOrCreateCompositeOrNull(ConstVal, I, SpvType, TII, ConstVec, BW,
+                                    SpvType->getOperand(2).getImm(),
+                                    ZeroAsNull);
+}
+
+Register SPIRVGlobalRegistry::getOrCreateConstVector(APFloat Val,
+                                                     MachineInstr &I,
+                                                     SPIRVType *SpvType,
+                                                     const SPIRVInstrInfo &TII,
+                                                     bool ZeroAsNull) {
+  const Type *LLVMTy = getTypeForSPIRVType(SpvType);
+  assert(LLVMTy->isVectorTy());
+  const FixedVectorType *LLVMVecTy = cast<FixedVectorType>(LLVMTy);
+  Type *LLVMBaseTy = LLVMVecTy->getElementType();
+  assert(LLVMBaseTy->isFloatingPointTy());
+  auto *ConstVal = ConstantFP::get(LLVMBaseTy, Val);
+  auto *ConstVec =
+      ConstantVector::getSplat(LLVMVecTy->getElementCount(), ConstVal);
+  unsigned BW = getScalarOrVectorBitWidth(SpvType);
+  return getOrCreateCompositeOrNull(ConstVal, I, SpvType, TII, ConstVec, BW,
+                                    SpvType->getOperand(2).getImm(),
+                                    ZeroAsNull);
 }
 
 Register
@@ -337,13 +474,13 @@ SPIRVGlobalRegistry::getOrCreateConsIntArray(uint64_t Val, MachineInstr &I,
   assert(LLVMTy->isArrayTy());
   const ArrayType *LLVMArrTy = cast<ArrayType>(LLVMTy);
   Type *LLVMBaseTy = LLVMArrTy->getElementType();
-  const auto ConstInt = ConstantInt::get(LLVMBaseTy, Val);
-  auto ConstArr =
+  auto *ConstInt = ConstantInt::get(LLVMBaseTy, Val);
+  auto *ConstArr =
       ConstantArray::get(const_cast<ArrayType *>(LLVMArrTy), {ConstInt});
   SPIRVType *SpvBaseTy = getSPIRVTypeForVReg(SpvType->getOperand(1).getReg());
   unsigned BW = getScalarOrVectorBitWidth(SpvBaseTy);
-  return getOrCreateIntCompositeOrNull(Val, I, SpvType, TII, ConstArr, BW,
-                                       LLVMArrTy->getNumElements());
+  return getOrCreateCompositeOrNull(ConstInt, I, SpvType, TII, ConstArr, BW,
+                                    LLVMArrTy->getNumElements());
 }
 
 Register SPIRVGlobalRegistry::getOrCreateIntCompositeOrNull(
@@ -1093,14 +1230,16 @@ SPIRVType *SPIRVGlobalRegistry::finishCreatingSPIRVType(const Type *LLVMTy,
   return SpirvType;
 }
 
-SPIRVType *SPIRVGlobalRegistry::getOrCreateSPIRVIntegerType(
-    unsigned BitWidth, MachineInstr &I, const SPIRVInstrInfo &TII) {
-  Type *LLVMTy = IntegerType::get(CurMF->getFunction().getContext(), BitWidth);
+SPIRVType *SPIRVGlobalRegistry::getOrCreateSPIRVType(unsigned BitWidth,
+                                                     MachineInstr &I,
+                                                     const SPIRVInstrInfo &TII,
+                                                     unsigned SPIRVOPcode,
+                                                     Type *LLVMTy) {
   Register Reg = DT.find(LLVMTy, CurMF);
   if (Reg.isValid())
     return getSPIRVTypeForVReg(Reg);
   MachineBasicBlock &BB = *I.getParent();
-  auto MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpTypeInt))
+  auto MIB = BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRVOPcode))
                  .addDef(createTypeVReg(CurMF->getRegInfo()))
                  .addImm(BitWidth)
                  .addImm(0);
@@ -1108,6 +1247,31 @@ SPIRVType *SPIRVGlobalRegistry::getOrCreateSPIRVIntegerType(
   return finishCreatingSPIRVType(LLVMTy, MIB);
 }
 
+SPIRVType *SPIRVGlobalRegistry::getOrCreateSPIRVIntegerType(
+    unsigned BitWidth, MachineInstr &I, const SPIRVInstrInfo &TII) {
+  Type *LLVMTy = IntegerType::get(CurMF->getFunction().getContext(), BitWidth);
+  return getOrCreateSPIRVType(BitWidth, I, TII, SPIRV::OpTypeInt, LLVMTy);
+}
+SPIRVType *SPIRVGlobalRegistry::getOrCreateSPIRVFloatType(
+    unsigned BitWidth, MachineInstr &I, const SPIRVInstrInfo &TII) {
+  LLVMContext &Ctx = CurMF->getFunction().getContext();
+  Type *LLVMTy;
+  switch (BitWidth) {
+  case 16:
+    LLVMTy = Type::getHalfTy(Ctx);
+    break;
+  case 32:
+    LLVMTy = Type::getFloatTy(Ctx);
+    break;
+  case 64:
+    LLVMTy = Type::getDoubleTy(Ctx);
+    break;
+  default:
+    llvm_unreachable("Bit width is of unexpected size.");
+  }
+  return getOrCreateSPIRVType(BitWidth, I, TII, SPIRV::OpTypeFloat, LLVMTy);
+}
+
 SPIRVType *
 SPIRVGlobalRegistry::getOrCreateSPIRVBoolType(MachineIRBuilder &MIRBuilder) {
   return getOrCreateSPIRVType(
diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
index 37f575e884ef4..2e3e69456ac26 100644
--- a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
+++ b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
@@ -20,6 +20,7 @@
 #include "SPIRVDuplicatesTracker.h"
 #include "SPIRVInstrInfo.h"
 #include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h"
+#include "llvm/IR/Constant.h"
 
 namespace llvm {
 using SPIRVType = const MachineInstr;
@@ -234,6 +235,8 @@ class SPIRVGlobalRegistry {
                               bool EmitIR = true);
   SPIRVType *assignIntTypeToVReg(unsigned BitWidth, Register VReg,
                                  MachineInstr &I, const SPIRVInstrInfo &TII);
+  SPIRVType *assignFloatTypeToVReg(unsigned BitWidth, Register VReg,
+                                   MachineInstr &I, const SPIRVInstrInfo &TII);
   SPIRVType *assignVectTypeToVReg(SPIRVType *BaseType, unsigned NumElements,
                                   Register VReg, MachineInstr &I,
                                   const SPIRVInstrInfo &TII);
@@ -372,12 +375,20 @@ class SPIRVGlobalRegistry {
   std::tuple<Register, ConstantInt *, bool> getOrCreateConstIntReg(
       uint64_t Val, SPIRVType *SpvType, MachineIRBuilder *MIRBuilder,
       MachineInstr *I = nullptr, const SPIRVInstrInfo *TII = nullptr);
+  std::tuple<Register, ConstantFP *, bool, unsigned> getOrCreateConstFloatReg(
+      APFloat Val, SPIRVType *SpvType, MachineIRBuilder *MIRBuilder,
+      MachineInstr *I = nullptr, const SPIRVInstrInfo *TII = nullptr);
   SPIRVType *finishCreatingSPIRVType(const Type *LLVMTy, SPIRVType *SpirvType);
-  Register getOrCreateIntCompositeOrNull(uint64_t Val, MachineInstr &I,
-                                         SPIRVType *SpvType,
-                                         const SPIRVInstrInfo &TII,
-                                         Constant *CA, unsigned BitWidth,
-                                         unsigned ElemCnt);
+  Register getOrCreateBaseRegister(Constant *Val, MachineInstr &I,
+                                   SPIRVType *SpvType,
+                                   const SPIRVInstrInfo &TII,
+                                   unsigned BitWidth);
+  Register getOrCreateCompositeOrNull(Constant *Val, MachineInstr &I,
+                                      SPIRVType *SpvType,
+                                      const SPIRVInstrInfo &TII, Constant *CA,
+                                      unsigned BitWidth, unsigned ElemCnt,
+                                      bool ZeroAsNull = true);
+
   Register getOrCreateIntCompositeOrNull(uint64_t Val,
                                          MachineIRBuilder &MIRBuilder,
                                          SPIRVType *SpvType, bool EmitIR,
@@ -388,12 +399,20 @@ class SPIRVGlobalRegistry {
   Register buildConstantInt(uint64_t Val, MachineIRBuilder &MIRBuilder,
                             SPIRVType *SpvType = nullptr, bool EmitIR = true);
   Register getOrCreateConstInt(uint64_t Val, MachineInstr &I,
-                               SPIRVType *SpvType, const SPIRVInstrInfo &TII);
+                               SPIRVType *SpvType, const SPIRVInstrInfo &TII,
+                               bool ZeroAsNull = true);
+  Register getOrCreateConstFP(APFloat Val, MachineInstr &I, SPIRVType *SpvType,
+                              const SPIRVInstrInfo &TII,
+                              bool ZeroAsNull = true);
   Register buildConstantFP(APFloat Val, MachineIRBuilder &MIRBuilder,
                            SPIRVType *SpvType = nullptr);
-  Register getOrCreateConsIntVector(uint64_t Val, MachineInstr &I,
-                                    SPIRVType *SpvType,
-                                    const SPIRVInstrInfo &TII);
+
+  Register getOrCreateConstVector(uint64_t Val, MachineInstr &I,
+                                  SPIRVType *SpvType, const SPIRVInstrInfo &TII,
+                                  bool ZeroAsNull = true);
+  Register getOrCreateConstVector(APFloat Val, MachineInstr &I,
+                                  SPIRVType *SpvType, const SPIRVInstrInfo &TII,
+                                  bool ZeroAsNull = true);
   Register getOrCreateConsIntArray(uint64_t Val, MachineInstr &I,
                                    SPIRVType *SpvType,
                                    const SPIRVInstrInfo &TII);
@@ -423,6 +442,11 @@ class SPIRVGlobalRegistry {
                                          MachineIRBuilder &MIRBuilder);
   SPIRVType *getOrCreateSPIRVIntegerType(unsigned BitWidth, MachineInstr &I,
                                          const SPIRVInstrInfo &TII);
+  SPIRVType *getOrCreateSPIRVType(unsigned BitWidth, MachineInstr &I,
+                                  const SPIRVInstrInfo &TII,
+                                  unsigned SPIRVOPcode, Type *LLVMTy);
+  SPIRVType *getOrCreateSPIRVFloatType(unsigned BitWidth, MachineInstr &I,
+                                       const SPIRVInstrInfo &TII);
   SPIRVType *getOrCreateSPIRVBoolType(MachineIRBuilder &MIRBuilder);
   SPIRVType *getOrCreateSPIRVBoolType(MachineInstr &I,
                                       const SPIRVInstrInfo &TII);
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index 45a70da7f8690..c1c0fc4b7dd48 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -28,6 +28,7 @@
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineModuleInfoImpls.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/TargetOpcodes.h"
 #include "llvm/IR/IntrinsicsSPIRV.h"
 #include "llvm/Support/Debug.h"
 
@@ -144,6 +145,9 @@ class SPIRVInstructionSelector : public InstructionSelector {
   bool selectAddrSpaceCast(Register ResVReg, const SPIRVType *ResType,
                            MachineInstr &I) const;
 
+  bool selectAll(Register ResVReg, const SPIRVType *ResType,
+                 MachineInstr &I) const;
+
   bool selectBitreverse(Register ResVReg, const SPIRVType *ResType,
                         MachineInstr &I) const;
 
@@ -229,6 +233,7 @@ class SPIRVInstructionSelector : public InstructionSelector {
                             const SPIRVType *ResType = nullptr) const;
 
   Register buildZerosVal(const SPIRVType *ResType, MachineInstr &I) const;
+  Register buildZerosValF(const SPIRVType *ResType, MachineInstr &I) const;
   Register buildOnesVal(bool AllOnes, const SPIRVType *ResType,
                         MachineInstr &I) const;
 
@@ -1155,6 +1160,65 @@ static unsigned getBoolCmpOpcode(unsigned PredNum) {
   }
 }
 
+bool SPIRVInstructionSelector::selectAll(Register ResVReg,
+                                         const SPIRVType *ResType,
+                                         MachineInstr &I) const {
+  assert(I.getNumOperands() == 3);
+  assert(I.getOperand(2).isReg());
+  MachineBasicBlock &BB = *I.getParent();
+  Register InputRegister = I.getOperand(2).getReg();
+  SPIRVType *InputType = GR.getSPIRVTypeForVReg(InputRegister);
+
+  if (!InputType)
+    report_fatal_error("Input Type could not be determined.");
+
+  bool IsBoolTy = GR.isScalarOrVectorOfType(InputRegister, SPIRV::OpTypeBool);
+  bool IsVectorTy = InputType->getOpcode() == SPIRV::OpTypeVector;
+  if (IsBoolTy && !IsVectorTy) {
+    assert(ResVReg == I.getOperand(0).getReg());
+    return BuildMI(*I.getParent(), I, I.getDebugLoc(),
+                   TII.get(TargetOpcode::COPY))
+        .addDef(ResVReg)
+        .addUse(InputRegister)
+        .constrainAllUses(TII, TRI, RBI);
+  }
+
+  bool IsFloatTy = GR.isScalarOrVectorOfType(InputRegister, SPIRV::OpTypeFloat);
+  unsigned SpirvNotEqualId =
+      IsFloatTy ? SPIRV::OpFOrdNotEqual : SPIRV::OpINotEqual;
+  SPIRVType *SpvBoolScalarTy = GR.getOrCreateSPIRVBoolType(I, TII);
+  SPIRVType *SpvBoolTy = SpvBoolScalarTy;
+  Register NotEqualReg = ResVReg;
+
+  if (IsVectorTy) {
+    NotEqualReg = IsBoolTy ? InputRegister
+                           : MRI->createVirtualRegister(&SPIRV::IDRegClass);
+    const unsigned NumElts = InputType->getOperand(2).getImm();
+    SpvBoolTy = GR.getOrCreateSPIRVVectorType(SpvBoolTy, NumElts, I, TII);
+  }
+
+  if (!IsBoolTy) {
+    Register ConstZeroReg =
+        IsFloatTy ? buildZerosValF(InputType, I) : buildZerosVal(InputType, I);
+
+    BuildMI(BB, I, I.getDebugLoc(), TII.get(SpirvNotEqualId))
+        .addDef(NotEqualReg)
+        .addUse(GR.getSPIRVTypeID(SpvBoolTy))
+        .addUse(InputRegister)
+        .addUse(ConstZeroReg)
+        .constrainAllUses(TII, TRI, RBI);
+  }
+
+  if (!IsVectorTy)
+    return true;
+
+  return BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpAll))
+      .addDef(ResVReg)
+      .addUse(GR.getSPIRVTypeID(SpvBoolScalarTy))
+      .addUse(NotEqualReg)
+      .constrainAllUses(TII, TRI, RBI);
+}
+
 bool SPIRVInstructionSelector::selectBitreverse(Register ResVReg,
                                                 const SPIRVType *ResType,
                                                 MachineInstr &I) const {
@@ -1391,9 +1455,35 @@ bool SPIRVInstructionSelector::selectFCmp(Register ResVReg,
 
 Register SPIRVInstructionSelector::buildZerosVal(const SPIRVType *ResType,
                                                  MachineInstr &I) const {
+  // OpenCL uses nulls for Zero. In HLSL we don't use null constants.
+  bool ZeroAsNull = STI.isOpenCLEnv();
+  if (ResType->getOpcode() == SPIRV::OpTypeVector)
+    return GR.getOrCreateConstVector(0UL, I, ResType, TII, ZeroAsNull);
+  return GR.getOrCreateConstInt(0, I, ResType, TII, ZeroAsNull);
+}
+
+static APFloat getZeroFP(const Type *LLVMFloatTy) {
+  if (!LLVMFloatTy)
+    return APFloat::getZero(APFloat::IEEEsingle());
+  switch (LLVMFloatTy->getScalarType()->getTypeID()) {
+  case Type::HalfTyID:
+    return APFloat::getZero(APFloat::IEEEhalf());
+  default:
+  case Type::FloatTyID:
+    return APFloat::getZero(APFloat::IEEEsingle());
+  case Type::DoubleTyID:
+    return APFloat::getZero(APFloat::IEEEdouble());
+  }
+}
+
+Register SPIRVInstructionSelector::buildZerosValF(const SPIRVType *ResType,
+                                                  MachineInstr &I) const {
+  // OpenCL uses nulls for Zero. In HLSL we don't use null constants.
+  bool ZeroAsNull = STI.isOpenCLEnv();
+  APFloat VZero = getZeroFP(GR.getTypeForSPIRVType(ResType));
   if (ResType->getOpcode() == SPIRV::OpTypeVector)
-    return GR.getOrCreateConsIntVector(0, I, ResType, TII);
-  return GR.getOrCreateConstInt(0, I, ResType, TII);
+    return GR.getOrCreateConstVector(VZero, I, ResType, TII, ZeroAsNull);
+  return GR.getOrCreateConstFP(VZero, I, ResType, TII, ZeroAsNull);
 }
 
 Register SPIRVInstructionSelector::buildOnesVal(bool AllOnes,
@@ -1403,7 +1493,7 @@ Register SPIRVInstructionSelector::buildOnesVal(bool AllOnes,
   APInt One =
       AllOnes ? APInt::getAllOnes(BitWidth) : APInt::getOneBitSet(BitWidth, 0);
   if (ResType->getOpcode() == SPIRV::OpTypeVector)
-    return GR.getOrCreateConsIntVector(One.getZExtValue(), I, ResType, TII);
+    return GR.getOrCreateConstVector(One.getZExtValue(), I, ResType, TII);
   return GR.getOrCreateConstInt(One.getZExtValue(), I, ResType, TII);
 }
 
@@ -1785,6 +1875,8 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
     break;
   case Intrinsic::spv_thread_id:
     return selectSpvThreadId(ResVReg, ResType, I);
+  case Intrinsic::spv_all:
+    return selectAll(ResVReg, ResType, I);
   case Intrinsic::spv_lifetime_start:
   case Intrinsic::spv_lifetime_end: {
     unsigned Op = IID == Intrinsic::spv_lifetime_start ? SPIRV::OpLifetimeStart
diff --git a/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/all.ll b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/all.ll
new file mode 100644
index 0000000000000..ef8d463cbd815
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/all.ll
@@ -0,0 +1,187 @@
+; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-HLSL
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-OCL
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; Make sure spirv operation function calls for all are generated.
+
+; CHECK-HLSL-DAG: OpMemoryModel Logical GLSL450
+; CHECK-OCL-DAG: OpMemoryModel Physical32 OpenCL
+; CHECK-DAG: OpName %[[#all_bool_arg:]] "a"
+; CHECK-DAG: %[[#int_64:]] = OpTypeInt 64 0
+; CHECK-DAG: %[[#bool:]] = OpTypeBool
+; CHECK-DAG: %[[#int_32:]] = OpTypeInt 32 0
+; CHECK-DAG: %[[#int_16:]] = OpTypeInt 16 0
+; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
+; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
+; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
+; CHECK-DAG: %[[#vec4_bool:]] = OpTypeVector %[[#bool]] 4
+; CHECK-DAG: %[[#vec4_16:]] = OpTypeVector %[[#int_16]] 4
+; CHECK-DAG: %[[#vec4_32:]] = OpTypeVector %[[#int_32]] 4
+; CHECK-DAG: %[[#vec4_64:]] = OpTypeVector %[[#int_64]] 4
+; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
+; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
+; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
+
+; CHECK-HLSL-DAG: %[[#const_i64_0:]] = OpConstant %[[#int_64]] 0
+; CHECK-HLSL-DAG: %[[#const_i32_0:]] = OpConstant %[[#int_32]] 0
+; CHECK-HLSL-DAG: %[[#const_i16_0:]] = OpConstant %[[#int_16]] 0
+; CHECK-HLSL-DAG: %[[#const_f64_0:]] = OpConstant %[[#float_64]] 0
+; CHECK-HLSL-DAG: %[[#const_f32_0:]] = OpConstant %[[#float_32:]] 0
+; CHECK-HLSL-DAG: %[[#const_f16_0:]] = OpConstant %[[#float_16:]] 0
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i16:]] = OpConstantComposite %[[#vec4_16:]] %[[#const_i16_0:]] %[[#const_i16_0:]] %[[#const_i16_0:]] %[[#const_i16_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i32:]] = OpConstantComposite %[[#vec4_32:]] %[[#const_i32_0:]] %[[#const_i32_0:]] %[[#const_i32_0:]] %[[#const_i32_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i64:]] = OpConstantComposite %[[#vec4_64:]] %[[#const_i64_0:]] %[[#const_i64_0:]] %[[#const_i64_0:]] %[[#const_i64_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f16:]] = OpConstantComposite %[[#vec4_float_16:]] %[[#const_f16_0:]] %[[#const_f16_0:]] %[[#const_f16_0:]] %[[#const_f16_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f32:]] = OpConstantComposite %[[#vec4_float_32:]] %[[#const_f32_0:]] %[[#const_f32_0:]] %[[#const_f32_0:]] %[[#const_f32_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f64:]] = OpConstantComposite %[[#vec4_float_64:]] %[[#const_f64_0:]] %[[#const_f64_0:]] %[[#const_f64_0:]] %[[#const_f64_0:]]
+
+; CHECK-OCL-DAG: %[[#const_i64_0:]] = OpConstantNull %[[#int_64]]
+; CHECK-OCL-DAG: %[[#const_i32_0:]] = OpConstantNull %[[#int_32]]
+; CHECK-OCL-DAG: %[[#const_i16_0:]] = OpConstantNull %[[#int_16]]
+; CHECK-OCL-DAG: %[[#const_f64_0:]] = OpConstantNull %[[#float_64]] 
+; CHECK-OCL-DAG: %[[#const_f32_0:]] = OpConstantNull %[[#float_32:]]
+; CHECK-OCL-DAG: %[[#const_f16_0:]] = OpConstantNull %[[#float_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i16:]] = OpConstantNull %[[#vec4_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i32:]] = OpConstantNull %[[#vec4_32:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i64:]] = OpConstantNull %[[#vec4_64:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f16:]] = OpConstantNull %[[#vec4_float_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f32:]] = OpConstantNull %[[#vec4_float_32:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f64:]] = OpConstantNull %[[#vec4_float_64:]]
+
+define noundef i1 @all_int64_t(i64 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i64_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.i64(i64 %p0)
+  ret i1 %hlsl.all
+}
+
+
+define noundef i1 @all_int(i32 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i32_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.i32(i32 %p0)
+  ret i1 %hlsl.all
+}
+
+
+define noundef i1 @all_int16_t(i16 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i16_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.i16(i16 %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_double(double noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f64_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.f64(double %p0)
+  ret i1 %hlsl.all
+}
+
+
+define noundef i1 @all_float(float noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f32_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.f32(float %p0)
+  ret i1 %hlsl.all
+}
+
+
+define noundef i1 @all_half(half noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f16_0:]]
+  %hlsl.all = call i1 @llvm.spv.all.f16(half %p0)
+  ret i1 %hlsl.all
+}
+
+
+define noundef i1 @all_bool4(<4 x i1> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpAll %[[#vec4_bool:]] %[[#arg0:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4i1(<4 x i1> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_short4(<4 x i16> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#shortVecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i16:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool:]] %[[#shortVecNotEq:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4i16(<4 x i16> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_int4(<4 x i32> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#i32VecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i32:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool:]] %[[#i32VecNotEq:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4i32(<4 x i32> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_int64_t4(<4 x i64> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#i64VecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i64:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool:]] %[[#i64VecNotEq]]
+  %hlsl.all = call i1 @llvm.spv.all.v4i64(<4 x i64> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_half4(<4 x half> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f16VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f16:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool]] %[[#f16VecNotEq:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4f16(<4 x half> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_float4(<4 x float> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f32VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f32:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool:]] %[[#f32VecNotEq:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4f32(<4 x float> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_double4(<4 x double> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f64VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f64:]]
+  ; CHECK: %[[#]] = OpAll %[[#bool:]] %[[#f64VecNotEq:]]
+  %hlsl.all = call i1 @llvm.spv.all.v4f64(<4 x double> %p0)
+  ret i1 %hlsl.all
+}
+
+define noundef i1 @all_bool(i1 noundef %a) {
+entry:
+  ; CHECK: %[[#all_bool_arg:]] = OpFunctionParameter %[[#bool:]]
+  ; CHECK: OpReturnValue %[[#all_bool_arg:]]
+  %hlsl.all = call i1 @llvm.spv.all.i1(i1 %a)
+  ret i1 %hlsl.all
+}
+
+declare i1 @llvm.spv.all.v4f16(<4 x half>)
+declare i1 @llvm.spv.all.v4f32(<4 x float>)
+declare i1 @llvm.spv.all.v4f64(<4 x double>)
+declare i1 @llvm.spv.all.v4i1(<4 x i1>)
+declare i1 @llvm.spv.all.v4i16(<4 x i16>)
+declare i1 @llvm.spv.all.v4i32(<4 x i32>)
+declare i1 @llvm.spv.all.v4i64(<4 x i64>)
+declare i1 @llvm.spv.all.i1(i1)
+declare i1 @llvm.spv.all.i16(i16)
+declare i1 @llvm.spv.all.i32(i32)
+declare i1 @llvm.spv.all.i64(i64)
+declare i1 @llvm.spv.all.f16(half)
+declare i1 @llvm.spv.all.f32(float)
+declare i1 @llvm.spv.all.f64(double)

From c258f573981336cd9f87f89e59c6c2117e5d44ec Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Wed, 10 Apr 2024 13:42:51 -0700
Subject: [PATCH 068/886] [ELF] Move createSyntheticSections from Writer.cpp to
 SyntheticSections.cpp. NFC

SyntheticSections.cpp is more appropriate. This change enables
elimination of many explicit template instantiations.

Due to `make<SymbolTableSection<ELFT>>(*strtab)` in Arch/ARM.cpp,
we do not remove explicit template instantiations for SymbolTableSection.
---
 lld/ELF/SyntheticSections.cpp | 367 +++++++++++++++++++++++++++++-----
 lld/ELF/SyntheticSections.h   |   4 +
 lld/ELF/Writer.cpp            | 292 ---------------------------
 lld/ELF/Writer.h              |   3 -
 4 files changed, 317 insertions(+), 349 deletions(-)

diff --git a/lld/ELF/SyntheticSections.cpp b/lld/ELF/SyntheticSections.cpp
index d8791e83dc9e5..0d7f393a9f3f4 100644
--- a/lld/ELF/SyntheticSections.cpp
+++ b/lld/ELF/SyntheticSections.cpp
@@ -3870,6 +3870,27 @@ void InStruct::reset() {
   symTabShndx.reset();
 }
 
+static bool needsInterpSection() {
+  return !config->relocatable && !config->shared &&
+         !config->dynamicLinker.empty() && script->needsInterpSection();
+}
+
+bool elf::hasMemtag() {
+  return config->emachine == EM_AARCH64 &&
+         config->androidMemtagMode != ELF::NT_MEMTAG_LEVEL_NONE;
+}
+
+// Fully static executables don't support MTE globals at this point in time, as
+// we currently rely on:
+//   - A dynamic loader to process relocations, and
+//   - Dynamic entries.
+// This restriction could be removed in future by re-using some of the ideas
+// that ifuncs use in fully static executables.
+bool elf::canHaveMemtagGlobals() {
+  return hasMemtag() &&
+         (config->relocatable || config->shared || needsInterpSection());
+}
+
 constexpr char kMemtagAndroidNoteName[] = "Android";
 void MemtagAndroidNote::writeTo(uint8_t *buf) {
   static_assert(
@@ -3985,31 +4006,304 @@ size_t MemtagGlobalDescriptors::getSize() const {
   return createMemtagGlobalDescriptors(symbols);
 }
 
+static OutputSection *findSection(StringRef name) {
+  for (SectionCommand *cmd : script->sectionCommands)
+    if (auto *osd = dyn_cast<OutputDesc>(cmd))
+      if (osd->osec.name == name)
+        return &osd->osec;
+  return nullptr;
+}
+
+static Defined *addOptionalRegular(StringRef name, SectionBase *sec,
+                                   uint64_t val, uint8_t stOther = STV_HIDDEN) {
+  Symbol *s = symtab.find(name);
+  if (!s || s->isDefined() || s->isCommon())
+    return nullptr;
+
+  s->resolve(Defined{ctx.internalFile, StringRef(), STB_GLOBAL, stOther,
+                     STT_NOTYPE, val,
+                     /*size=*/0, sec});
+  s->isUsedInRegularObj = true;
+  return cast<Defined>(s);
+}
+
+template <class ELFT> void elf::createSyntheticSections() {
+  // Initialize all pointers with NULL. This is needed because
+  // you can call lld::elf::main more than once as a library.
+  Out::tlsPhdr = nullptr;
+  Out::preinitArray = nullptr;
+  Out::initArray = nullptr;
+  Out::finiArray = nullptr;
+
+  // Add the .interp section first because it is not a SyntheticSection.
+  // The removeUnusedSyntheticSections() function relies on the
+  // SyntheticSections coming last.
+  if (needsInterpSection()) {
+    for (size_t i = 1; i <= partitions.size(); ++i) {
+      InputSection *sec = createInterpSection();
+      sec->partition = i;
+      ctx.inputSections.push_back(sec);
+    }
+  }
+
+  auto add = [](SyntheticSection &sec) { ctx.inputSections.push_back(&sec); };
+
+  in.shStrTab = std::make_unique<StringTableSection>(".shstrtab", false);
+
+  Out::programHeaders = make<OutputSection>("", 0, SHF_ALLOC);
+  Out::programHeaders->addralign = config->wordsize;
+
+  if (config->strip != StripPolicy::All) {
+    in.strTab = std::make_unique<StringTableSection>(".strtab", false);
+    in.symTab = std::make_unique<SymbolTableSection<ELFT>>(*in.strTab);
+    in.symTabShndx = std::make_unique<SymtabShndxSection>();
+  }
+
+  in.bss = std::make_unique<BssSection>(".bss", 0, 1);
+  add(*in.bss);
+
+  // If there is a SECTIONS command and a .data.rel.ro section name use name
+  // .data.rel.ro.bss so that we match in the .data.rel.ro output section.
+  // This makes sure our relro is contiguous.
+  bool hasDataRelRo = script->hasSectionsCommand && findSection(".data.rel.ro");
+  in.bssRelRo = std::make_unique<BssSection>(
+      hasDataRelRo ? ".data.rel.ro.bss" : ".bss.rel.ro", 0, 1);
+  add(*in.bssRelRo);
+
+  // Add MIPS-specific sections.
+  if (config->emachine == EM_MIPS) {
+    if (!config->shared && config->hasDynSymTab) {
+      in.mipsRldMap = std::make_unique<MipsRldMapSection>();
+      add(*in.mipsRldMap);
+    }
+    if ((in.mipsAbiFlags = MipsAbiFlagsSection<ELFT>::create()))
+      add(*in.mipsAbiFlags);
+    if ((in.mipsOptions = MipsOptionsSection<ELFT>::create()))
+      add(*in.mipsOptions);
+    if ((in.mipsReginfo = MipsReginfoSection<ELFT>::create()))
+      add(*in.mipsReginfo);
+  }
+
+  StringRef relaDynName = config->isRela ? ".rela.dyn" : ".rel.dyn";
+
+  const unsigned threadCount = config->threadCount;
+  for (Partition &part : partitions) {
+    auto add = [&](SyntheticSection &sec) {
+      sec.partition = part.getNumber();
+      ctx.inputSections.push_back(&sec);
+    };
+
+    if (!part.name.empty()) {
+      part.elfHeader = std::make_unique<PartitionElfHeaderSection<ELFT>>();
+      part.elfHeader->name = part.name;
+      add(*part.elfHeader);
+
+      part.programHeaders =
+          std::make_unique<PartitionProgramHeadersSection<ELFT>>();
+      add(*part.programHeaders);
+    }
+
+    if (config->buildId != BuildIdKind::None) {
+      part.buildId = std::make_unique<BuildIdSection>();
+      add(*part.buildId);
+    }
+
+    part.dynStrTab = std::make_unique<StringTableSection>(".dynstr", true);
+    part.dynSymTab =
+        std::make_unique<SymbolTableSection<ELFT>>(*part.dynStrTab);
+    part.dynamic = std::make_unique<DynamicSection<ELFT>>();
+
+    if (hasMemtag()) {
+      part.memtagAndroidNote = std::make_unique<MemtagAndroidNote>();
+      add(*part.memtagAndroidNote);
+      if (canHaveMemtagGlobals()) {
+        part.memtagGlobalDescriptors =
+            std::make_unique<MemtagGlobalDescriptors>();
+        add(*part.memtagGlobalDescriptors);
+      }
+    }
+
+    if (config->androidPackDynRelocs)
+      part.relaDyn = std::make_unique<AndroidPackedRelocationSection<ELFT>>(
+          relaDynName, threadCount);
+    else
+      part.relaDyn = std::make_unique<RelocationSection<ELFT>>(
+          relaDynName, config->zCombreloc, threadCount);
+
+    if (config->hasDynSymTab) {
+      add(*part.dynSymTab);
+
+      part.verSym = std::make_unique<VersionTableSection>();
+      add(*part.verSym);
+
+      if (!namedVersionDefs().empty()) {
+        part.verDef = std::make_unique<VersionDefinitionSection>();
+        add(*part.verDef);
+      }
+
+      part.verNeed = std::make_unique<VersionNeedSection<ELFT>>();
+      add(*part.verNeed);
+
+      if (config->gnuHash) {
+        part.gnuHashTab = std::make_unique<GnuHashTableSection>();
+        add(*part.gnuHashTab);
+      }
+
+      if (config->sysvHash) {
+        part.hashTab = std::make_unique<HashTableSection>();
+        add(*part.hashTab);
+      }
+
+      add(*part.dynamic);
+      add(*part.dynStrTab);
+    }
+    add(*part.relaDyn);
+
+    if (config->relrPackDynRelocs) {
+      part.relrDyn = std::make_unique<RelrSection<ELFT>>(threadCount);
+      add(*part.relrDyn);
+    }
+
+    if (!config->relocatable) {
+      if (config->ehFrameHdr) {
+        part.ehFrameHdr = std::make_unique<EhFrameHeader>();
+        add(*part.ehFrameHdr);
+      }
+      part.ehFrame = std::make_unique<EhFrameSection>();
+      add(*part.ehFrame);
+
+      if (config->emachine == EM_ARM) {
+        // This section replaces all the individual .ARM.exidx InputSections.
+        part.armExidx = std::make_unique<ARMExidxSyntheticSection>();
+        add(*part.armExidx);
+      }
+    }
+
+    if (!config->packageMetadata.empty()) {
+      part.packageMetadataNote = std::make_unique<PackageMetadataNote>();
+      add(*part.packageMetadataNote);
+    }
+  }
+
+  if (partitions.size() != 1) {
+    // Create the partition end marker. This needs to be in partition number 255
+    // so that it is sorted after all other partitions. It also has other
+    // special handling (see createPhdrs() and combineEhSections()).
+    in.partEnd =
+        std::make_unique<BssSection>(".part.end", config->maxPageSize, 1);
+    in.partEnd->partition = 255;
+    add(*in.partEnd);
+
+    in.partIndex = std::make_unique<PartitionIndexSection>();
+    addOptionalRegular("__part_index_begin", in.partIndex.get(), 0);
+    addOptionalRegular("__part_index_end", in.partIndex.get(),
+                       in.partIndex->getSize());
+    add(*in.partIndex);
+  }
+
+  // Add .got. MIPS' .got is so different from the other archs,
+  // it has its own class.
+  if (config->emachine == EM_MIPS) {
+    in.mipsGot = std::make_unique<MipsGotSection>();
+    add(*in.mipsGot);
+  } else {
+    in.got = std::make_unique<GotSection>();
+    add(*in.got);
+  }
+
+  if (config->emachine == EM_PPC) {
+    in.ppc32Got2 = std::make_unique<PPC32Got2Section>();
+    add(*in.ppc32Got2);
+  }
+
+  if (config->emachine == EM_PPC64) {
+    in.ppc64LongBranchTarget = std::make_unique<PPC64LongBranchTargetSection>();
+    add(*in.ppc64LongBranchTarget);
+  }
+
+  in.gotPlt = std::make_unique<GotPltSection>();
+  add(*in.gotPlt);
+  in.igotPlt = std::make_unique<IgotPltSection>();
+  add(*in.igotPlt);
+  // Add .relro_padding if DATA_SEGMENT_RELRO_END is used; otherwise, add the
+  // section in the absence of PHDRS/SECTIONS commands.
+  if (config->zRelro &&
+      ((script->phdrsCommands.empty() && !script->hasSectionsCommand) ||
+       script->seenRelroEnd)) {
+    in.relroPadding = std::make_unique<RelroPaddingSection>();
+    add(*in.relroPadding);
+  }
+
+  if (config->emachine == EM_ARM) {
+    in.armCmseSGSection = std::make_unique<ArmCmseSGSection>();
+    add(*in.armCmseSGSection);
+  }
+
+  // _GLOBAL_OFFSET_TABLE_ is defined relative to either .got.plt or .got. Treat
+  // it as a relocation and ensure the referenced section is created.
+  if (ElfSym::globalOffsetTable && config->emachine != EM_MIPS) {
+    if (target->gotBaseSymInGotPlt)
+      in.gotPlt->hasGotPltOffRel = true;
+    else
+      in.got->hasGotOffRel = true;
+  }
+
+  // We always need to add rel[a].plt to output if it has entries.
+  // Even for static linking it can contain R_[*]_IRELATIVE relocations.
+  in.relaPlt = std::make_unique<RelocationSection<ELFT>>(
+      config->isRela ? ".rela.plt" : ".rel.plt", /*sort=*/false,
+      /*threadCount=*/1);
+  add(*in.relaPlt);
+
+  if ((config->emachine == EM_386 || config->emachine == EM_X86_64) &&
+      (config->andFeatures & GNU_PROPERTY_X86_FEATURE_1_IBT)) {
+    in.ibtPlt = std::make_unique<IBTPltSection>();
+    add(*in.ibtPlt);
+  }
+
+  if (config->emachine == EM_PPC)
+    in.plt = std::make_unique<PPC32GlinkSection>();
+  else
+    in.plt = std::make_unique<PltSection>();
+  add(*in.plt);
+  in.iplt = std::make_unique<IpltSection>();
+  add(*in.iplt);
+
+  if (config->andFeatures || !ctx.aarch64PauthAbiCoreInfo.empty())
+    add(*make<GnuPropertySection>());
+
+  if (config->gdbIndex) {
+    in.gdbIndex = GdbIndexSection::create<ELFT>();
+    add(*in.gdbIndex);
+  }
+
+  // .note.GNU-stack is always added when we are creating a re-linkable
+  // object file. Other linkers are using the presence of this marker
+  // section to control the executable-ness of the stack area, but that
+  // is irrelevant these days. Stack area should always be non-executable
+  // by default. So we emit this section unconditionally.
+  if (config->relocatable)
+    add(*make<GnuStackSection>());
+
+  if (in.symTab)
+    add(*in.symTab);
+  if (in.symTabShndx)
+    add(*in.symTabShndx);
+  add(*in.shStrTab);
+  if (in.strTab)
+    add(*in.strTab);
+}
+
 InStruct elf::in;
 
 std::vector<Partition> elf::partitions;
 Partition *elf::mainPart;
 
-template std::unique_ptr<GdbIndexSection> GdbIndexSection::create<ELF32LE>();
-template std::unique_ptr<GdbIndexSection> GdbIndexSection::create<ELF32BE>();
-template std::unique_ptr<GdbIndexSection> GdbIndexSection::create<ELF64LE>();
-template std::unique_ptr<GdbIndexSection> GdbIndexSection::create<ELF64BE>();
-
 template void elf::splitSections<ELF32LE>();
 template void elf::splitSections<ELF32BE>();
 template void elf::splitSections<ELF64LE>();
 template void elf::splitSections<ELF64BE>();
 
-template class elf::MipsAbiFlagsSection<ELF32LE>;
-template class elf::MipsAbiFlagsSection<ELF32BE>;
-template class elf::MipsAbiFlagsSection<ELF64LE>;
-template class elf::MipsAbiFlagsSection<ELF64BE>;
-
-template class elf::MipsOptionsSection<ELF32LE>;
-template class elf::MipsOptionsSection<ELF32BE>;
-template class elf::MipsOptionsSection<ELF64LE>;
-template class elf::MipsOptionsSection<ELF64BE>;
-
 template void EhFrameSection::iterateFDEWithLSDA<ELF32LE>(
     function_ref<void(InputSection &)>);
 template void EhFrameSection::iterateFDEWithLSDA<ELF32BE>(
@@ -4019,41 +4313,11 @@ template void EhFrameSection::iterateFDEWithLSDA<ELF64LE>(
 template void EhFrameSection::iterateFDEWithLSDA<ELF64BE>(
     function_ref<void(InputSection &)>);
 
-template class elf::MipsReginfoSection<ELF32LE>;
-template class elf::MipsReginfoSection<ELF32BE>;
-template class elf::MipsReginfoSection<ELF64LE>;
-template class elf::MipsReginfoSection<ELF64BE>;
-
-template class elf::DynamicSection<ELF32LE>;
-template class elf::DynamicSection<ELF32BE>;
-template class elf::DynamicSection<ELF64LE>;
-template class elf::DynamicSection<ELF64BE>;
-
-template class elf::RelocationSection<ELF32LE>;
-template class elf::RelocationSection<ELF32BE>;
-template class elf::RelocationSection<ELF64LE>;
-template class elf::RelocationSection<ELF64BE>;
-
-template class elf::AndroidPackedRelocationSection<ELF32LE>;
-template class elf::AndroidPackedRelocationSection<ELF32BE>;
-template class elf::AndroidPackedRelocationSection<ELF64LE>;
-template class elf::AndroidPackedRelocationSection<ELF64BE>;
-
-template class elf::RelrSection<ELF32LE>;
-template class elf::RelrSection<ELF32BE>;
-template class elf::RelrSection<ELF64LE>;
-template class elf::RelrSection<ELF64BE>;
-
 template class elf::SymbolTableSection<ELF32LE>;
 template class elf::SymbolTableSection<ELF32BE>;
 template class elf::SymbolTableSection<ELF64LE>;
 template class elf::SymbolTableSection<ELF64BE>;
 
-template class elf::VersionNeedSection<ELF32LE>;
-template class elf::VersionNeedSection<ELF32BE>;
-template class elf::VersionNeedSection<ELF64LE>;
-template class elf::VersionNeedSection<ELF64BE>;
-
 template void elf::writeEhdr<ELF32LE>(uint8_t *Buf, Partition &Part);
 template void elf::writeEhdr<ELF32BE>(uint8_t *Buf, Partition &Part);
 template void elf::writeEhdr<ELF64LE>(uint8_t *Buf, Partition &Part);
@@ -4064,12 +4328,7 @@ template void elf::writePhdrs<ELF32BE>(uint8_t *Buf, Partition &Part);
 template void elf::writePhdrs<ELF64LE>(uint8_t *Buf, Partition &Part);
 template void elf::writePhdrs<ELF64BE>(uint8_t *Buf, Partition &Part);
 
-template class elf::PartitionElfHeaderSection<ELF32LE>;
-template class elf::PartitionElfHeaderSection<ELF32BE>;
-template class elf::PartitionElfHeaderSection<ELF64LE>;
-template class elf::PartitionElfHeaderSection<ELF64BE>;
-
-template class elf::PartitionProgramHeadersSection<ELF32LE>;
-template class elf::PartitionProgramHeadersSection<ELF32BE>;
-template class elf::PartitionProgramHeadersSection<ELF64LE>;
-template class elf::PartitionProgramHeadersSection<ELF64BE>;
+template void elf::createSyntheticSections<ELF32LE>();
+template void elf::createSyntheticSections<ELF32BE>();
+template void elf::createSyntheticSections<ELF64LE>();
+template void elf::createSyntheticSections<ELF64BE>();
diff --git a/lld/ELF/SyntheticSections.h b/lld/ELF/SyntheticSections.h
index 68b4cdb1dde04..759b78668f546 100644
--- a/lld/ELF/SyntheticSections.h
+++ b/lld/ELF/SyntheticSections.h
@@ -1284,11 +1284,15 @@ class MemtagGlobalDescriptors final : public SyntheticSection {
   SmallVector<const Symbol *, 0> symbols;
 };
 
+template <class ELFT> void createSyntheticSections();
 InputSection *createInterpSection();
 MergeInputSection *createCommentSection();
 template <class ELFT> void splitSections();
 void combineEhSections();
 
+bool hasMemtag();
+bool canHaveMemtagGlobals();
+
 template <typename ELFT> void writeEhdr(uint8_t *buf, Partition &part);
 template <typename ELFT> void writePhdrs(uint8_t *buf, Partition &part);
 
diff --git a/lld/ELF/Writer.cpp b/lld/ELF/Writer.cpp
index 021b9bb0d5e22..d2a9e872dab91 100644
--- a/lld/ELF/Writer.cpp
+++ b/lld/ELF/Writer.cpp
@@ -91,11 +91,6 @@ template <class ELFT> class Writer {
 };
 } // anonymous namespace
 
-static bool needsInterpSection() {
-  return !config->relocatable && !config->shared &&
-         !config->dynamicLinker.empty() && script->needsInterpSection();
-}
-
 template <class ELFT> void elf::writeResult() {
   Writer<ELFT>().run();
 }
@@ -297,22 +292,6 @@ static void demoteSymbolsAndComputeIsPreemptible() {
   }
 }
 
-bool elf::hasMemtag() {
-  return config->emachine == EM_AARCH64 &&
-         config->androidMemtagMode != ELF::NT_MEMTAG_LEVEL_NONE;
-}
-
-// Fully static executables don't support MTE globals at this point in time, as
-// we currently rely on:
-//   - A dynamic loader to process relocations, and
-//   - Dynamic entries.
-// This restriction could be removed in future by re-using some of the ideas
-// that ifuncs use in fully static executables.
-bool elf::canHaveMemtagGlobals() {
-  return hasMemtag() &&
-         (config->relocatable || config->shared || needsInterpSection());
-}
-
 static OutputSection *findSection(StringRef name, unsigned partition = 1) {
   for (SectionCommand *cmd : script->sectionCommands)
     if (auto *osd = dyn_cast<OutputDesc>(cmd))
@@ -321,272 +300,6 @@ static OutputSection *findSection(StringRef name, unsigned partition = 1) {
   return nullptr;
 }
 
-template <class ELFT> void elf::createSyntheticSections() {
-  // Initialize all pointers with NULL. This is needed because
-  // you can call lld::elf::main more than once as a library.
-  Out::tlsPhdr = nullptr;
-  Out::preinitArray = nullptr;
-  Out::initArray = nullptr;
-  Out::finiArray = nullptr;
-
-  // Add the .interp section first because it is not a SyntheticSection.
-  // The removeUnusedSyntheticSections() function relies on the
-  // SyntheticSections coming last.
-  if (needsInterpSection()) {
-    for (size_t i = 1; i <= partitions.size(); ++i) {
-      InputSection *sec = createInterpSection();
-      sec->partition = i;
-      ctx.inputSections.push_back(sec);
-    }
-  }
-
-  auto add = [](SyntheticSection &sec) { ctx.inputSections.push_back(&sec); };
-
-  in.shStrTab = std::make_unique<StringTableSection>(".shstrtab", false);
-
-  Out::programHeaders = make<OutputSection>("", 0, SHF_ALLOC);
-  Out::programHeaders->addralign = config->wordsize;
-
-  if (config->strip != StripPolicy::All) {
-    in.strTab = std::make_unique<StringTableSection>(".strtab", false);
-    in.symTab = std::make_unique<SymbolTableSection<ELFT>>(*in.strTab);
-    in.symTabShndx = std::make_unique<SymtabShndxSection>();
-  }
-
-  in.bss = std::make_unique<BssSection>(".bss", 0, 1);
-  add(*in.bss);
-
-  // If there is a SECTIONS command and a .data.rel.ro section name use name
-  // .data.rel.ro.bss so that we match in the .data.rel.ro output section.
-  // This makes sure our relro is contiguous.
-  bool hasDataRelRo = script->hasSectionsCommand && findSection(".data.rel.ro");
-  in.bssRelRo = std::make_unique<BssSection>(
-      hasDataRelRo ? ".data.rel.ro.bss" : ".bss.rel.ro", 0, 1);
-  add(*in.bssRelRo);
-
-  // Add MIPS-specific sections.
-  if (config->emachine == EM_MIPS) {
-    if (!config->shared && config->hasDynSymTab) {
-      in.mipsRldMap = std::make_unique<MipsRldMapSection>();
-      add(*in.mipsRldMap);
-    }
-    if ((in.mipsAbiFlags = MipsAbiFlagsSection<ELFT>::create()))
-      add(*in.mipsAbiFlags);
-    if ((in.mipsOptions = MipsOptionsSection<ELFT>::create()))
-      add(*in.mipsOptions);
-    if ((in.mipsReginfo = MipsReginfoSection<ELFT>::create()))
-      add(*in.mipsReginfo);
-  }
-
-  StringRef relaDynName = config->isRela ? ".rela.dyn" : ".rel.dyn";
-
-  const unsigned threadCount = config->threadCount;
-  for (Partition &part : partitions) {
-    auto add = [&](SyntheticSection &sec) {
-      sec.partition = part.getNumber();
-      ctx.inputSections.push_back(&sec);
-    };
-
-    if (!part.name.empty()) {
-      part.elfHeader = std::make_unique<PartitionElfHeaderSection<ELFT>>();
-      part.elfHeader->name = part.name;
-      add(*part.elfHeader);
-
-      part.programHeaders =
-          std::make_unique<PartitionProgramHeadersSection<ELFT>>();
-      add(*part.programHeaders);
-    }
-
-    if (config->buildId != BuildIdKind::None) {
-      part.buildId = std::make_unique<BuildIdSection>();
-      add(*part.buildId);
-    }
-
-    part.dynStrTab = std::make_unique<StringTableSection>(".dynstr", true);
-    part.dynSymTab =
-        std::make_unique<SymbolTableSection<ELFT>>(*part.dynStrTab);
-    part.dynamic = std::make_unique<DynamicSection<ELFT>>();
-
-    if (hasMemtag()) {
-      part.memtagAndroidNote = std::make_unique<MemtagAndroidNote>();
-      add(*part.memtagAndroidNote);
-      if (canHaveMemtagGlobals()) {
-        part.memtagGlobalDescriptors =
-            std::make_unique<MemtagGlobalDescriptors>();
-        add(*part.memtagGlobalDescriptors);
-      }
-    }
-
-    if (config->androidPackDynRelocs)
-      part.relaDyn = std::make_unique<AndroidPackedRelocationSection<ELFT>>(
-          relaDynName, threadCount);
-    else
-      part.relaDyn = std::make_unique<RelocationSection<ELFT>>(
-          relaDynName, config->zCombreloc, threadCount);
-
-    if (config->hasDynSymTab) {
-      add(*part.dynSymTab);
-
-      part.verSym = std::make_unique<VersionTableSection>();
-      add(*part.verSym);
-
-      if (!namedVersionDefs().empty()) {
-        part.verDef = std::make_unique<VersionDefinitionSection>();
-        add(*part.verDef);
-      }
-
-      part.verNeed = std::make_unique<VersionNeedSection<ELFT>>();
-      add(*part.verNeed);
-
-      if (config->gnuHash) {
-        part.gnuHashTab = std::make_unique<GnuHashTableSection>();
-        add(*part.gnuHashTab);
-      }
-
-      if (config->sysvHash) {
-        part.hashTab = std::make_unique<HashTableSection>();
-        add(*part.hashTab);
-      }
-
-      add(*part.dynamic);
-      add(*part.dynStrTab);
-    }
-    add(*part.relaDyn);
-
-    if (config->relrPackDynRelocs) {
-      part.relrDyn = std::make_unique<RelrSection<ELFT>>(threadCount);
-      add(*part.relrDyn);
-    }
-
-    if (!config->relocatable) {
-      if (config->ehFrameHdr) {
-        part.ehFrameHdr = std::make_unique<EhFrameHeader>();
-        add(*part.ehFrameHdr);
-      }
-      part.ehFrame = std::make_unique<EhFrameSection>();
-      add(*part.ehFrame);
-
-      if (config->emachine == EM_ARM) {
-        // This section replaces all the individual .ARM.exidx InputSections.
-        part.armExidx = std::make_unique<ARMExidxSyntheticSection>();
-        add(*part.armExidx);
-      }
-    }
-
-    if (!config->packageMetadata.empty()) {
-      part.packageMetadataNote = std::make_unique<PackageMetadataNote>();
-      add(*part.packageMetadataNote);
-    }
-  }
-
-  if (partitions.size() != 1) {
-    // Create the partition end marker. This needs to be in partition number 255
-    // so that it is sorted after all other partitions. It also has other
-    // special handling (see createPhdrs() and combineEhSections()).
-    in.partEnd =
-        std::make_unique<BssSection>(".part.end", config->maxPageSize, 1);
-    in.partEnd->partition = 255;
-    add(*in.partEnd);
-
-    in.partIndex = std::make_unique<PartitionIndexSection>();
-    addOptionalRegular("__part_index_begin", in.partIndex.get(), 0);
-    addOptionalRegular("__part_index_end", in.partIndex.get(),
-                       in.partIndex->getSize());
-    add(*in.partIndex);
-  }
-
-  // Add .got. MIPS' .got is so different from the other archs,
-  // it has its own class.
-  if (config->emachine == EM_MIPS) {
-    in.mipsGot = std::make_unique<MipsGotSection>();
-    add(*in.mipsGot);
-  } else {
-    in.got = std::make_unique<GotSection>();
-    add(*in.got);
-  }
-
-  if (config->emachine == EM_PPC) {
-    in.ppc32Got2 = std::make_unique<PPC32Got2Section>();
-    add(*in.ppc32Got2);
-  }
-
-  if (config->emachine == EM_PPC64) {
-    in.ppc64LongBranchTarget = std::make_unique<PPC64LongBranchTargetSection>();
-    add(*in.ppc64LongBranchTarget);
-  }
-
-  in.gotPlt = std::make_unique<GotPltSection>();
-  add(*in.gotPlt);
-  in.igotPlt = std::make_unique<IgotPltSection>();
-  add(*in.igotPlt);
-  // Add .relro_padding if DATA_SEGMENT_RELRO_END is used; otherwise, add the
-  // section in the absence of PHDRS/SECTIONS commands.
-  if (config->zRelro && ((script->phdrsCommands.empty() &&
-        !script->hasSectionsCommand) || script->seenRelroEnd)) {
-    in.relroPadding = std::make_unique<RelroPaddingSection>();
-    add(*in.relroPadding);
-  }
-
-  if (config->emachine == EM_ARM) {
-    in.armCmseSGSection = std::make_unique<ArmCmseSGSection>();
-    add(*in.armCmseSGSection);
-  }
-
-  // _GLOBAL_OFFSET_TABLE_ is defined relative to either .got.plt or .got. Treat
-  // it as a relocation and ensure the referenced section is created.
-  if (ElfSym::globalOffsetTable && config->emachine != EM_MIPS) {
-    if (target->gotBaseSymInGotPlt)
-      in.gotPlt->hasGotPltOffRel = true;
-    else
-      in.got->hasGotOffRel = true;
-  }
-
-  // We always need to add rel[a].plt to output if it has entries.
-  // Even for static linking it can contain R_[*]_IRELATIVE relocations.
-  in.relaPlt = std::make_unique<RelocationSection<ELFT>>(
-      config->isRela ? ".rela.plt" : ".rel.plt", /*sort=*/false,
-      /*threadCount=*/1);
-  add(*in.relaPlt);
-
-  if ((config->emachine == EM_386 || config->emachine == EM_X86_64) &&
-      (config->andFeatures & GNU_PROPERTY_X86_FEATURE_1_IBT)) {
-    in.ibtPlt = std::make_unique<IBTPltSection>();
-    add(*in.ibtPlt);
-  }
-
-  if (config->emachine == EM_PPC)
-    in.plt = std::make_unique<PPC32GlinkSection>();
-  else
-    in.plt = std::make_unique<PltSection>();
-  add(*in.plt);
-  in.iplt = std::make_unique<IpltSection>();
-  add(*in.iplt);
-
-  if (config->andFeatures || !ctx.aarch64PauthAbiCoreInfo.empty())
-    add(*make<GnuPropertySection>());
-
-  if (config->gdbIndex) {
-    in.gdbIndex = GdbIndexSection::create<ELFT>();
-    add(*in.gdbIndex);
-  }
-
-  // .note.GNU-stack is always added when we are creating a re-linkable
-  // object file. Other linkers are using the presence of this marker
-  // section to control the executable-ness of the stack area, but that
-  // is irrelevant these days. Stack area should always be non-executable
-  // by default. So we emit this section unconditionally.
-  if (config->relocatable)
-    add(*make<GnuStackSection>());
-
-  if (in.symTab)
-    add(*in.symTab);
-  if (in.symTabShndx)
-    add(*in.symTabShndx);
-  add(*in.shStrTab);
-  if (in.strTab)
-    add(*in.strTab);
-}
-
 // The main function of the writer.
 template <class ELFT> void Writer<ELFT>::run() {
   // Now that we have a complete set of output sections. This function
@@ -3114,11 +2827,6 @@ template <class ELFT> void Writer<ELFT>::writeBuildId() {
     part.buildId->writeBuildId(output);
 }
 
-template void elf::createSyntheticSections<ELF32LE>();
-template void elf::createSyntheticSections<ELF32BE>();
-template void elf::createSyntheticSections<ELF64LE>();
-template void elf::createSyntheticSections<ELF64BE>();
-
 template void elf::writeResult<ELF32LE>();
 template void elf::writeResult<ELF32BE>();
 template void elf::writeResult<ELF64LE>();
diff --git a/lld/ELF/Writer.h b/lld/ELF/Writer.h
index aac8176d90989..7aa06dbcb131a 100644
--- a/lld/ELF/Writer.h
+++ b/lld/ELF/Writer.h
@@ -17,7 +17,6 @@ namespace lld::elf {
 class InputFile;
 class OutputSection;
 void copySectionsIntoPartitions();
-template <class ELFT> void createSyntheticSections();
 template <class ELFT> void writeResult();
 
 // This describes a program header entry.
@@ -57,8 +56,6 @@ bool isMipsN32Abi(const InputFile *f);
 bool isMicroMips();
 bool isMipsR6();
 
-bool hasMemtag();
-bool canHaveMemtagGlobals();
 } // namespace lld::elf
 
 #endif

From 8cfa72ade9f2f7df81a008efea84f833b73494b9 Mon Sep 17 00:00:00 2001
From: Nick Desaulniers <ndesaulniers@google.com>
Date: Wed, 10 Apr 2024 13:51:23 -0700
Subject: [PATCH 069/886] [libc] fix typo in hdr/CMakeLists

Fixes #87896
---
 libc/hdr/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libc/hdr/CMakeLists.txt b/libc/hdr/CMakeLists.txt
index 4ca7db5e98d60..5a1acd9d17ab4 100644
--- a/libc/hdr/CMakeLists.txt
+++ b/libc/hdr/CMakeLists.txt
@@ -38,5 +38,5 @@ add_proxy_header_library(
     fenv_macros.h
   FULL_BUILD_DEPENDS
     libc.include.llvm-libc-macros.fenv_macros
-    libc.incude.fenv
+    libc.include.fenv
 )

From fb771fe315654231f613a5501ebd538f036c78b6 Mon Sep 17 00:00:00 2001
From: Jeff Niu <jeff@modular.com>
Date: Wed, 10 Apr 2024 23:34:48 +0200
Subject: [PATCH 070/886] [mlir] Slightly optimize bytecode op numbering
 (#88310)

If the bytecode encoding supports properties, then the dictionary
attribute is always the raw dictionary attribute of the operation,
regardless of what it contains. Otherwise, get the dictionary attribute
from the op: if the op does not have properties, then it returns the raw
dictionary, otherwise it returns the combined inherent and discardable
attributes.
---
 mlir/lib/Bytecode/Writer/IRNumbering.cpp | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/mlir/lib/Bytecode/Writer/IRNumbering.cpp b/mlir/lib/Bytecode/Writer/IRNumbering.cpp
index f36c9ef060b6d..d2144dd7f3348 100644
--- a/mlir/lib/Bytecode/Writer/IRNumbering.cpp
+++ b/mlir/lib/Bytecode/Writer/IRNumbering.cpp
@@ -424,22 +424,22 @@ void IRNumberingState::number(Operation &op) {
     number(result.getType());
   }
 
-  // Only number the operation's dictionary if it isn't empty.
-  DictionaryAttr dictAttr = op.getDiscardableAttrDictionary();
   // Prior to a version with native property encoding, or when properties are
   // not used, we need to number also the merged dictionary containing both the
   // inherent and discardable attribute.
-  if (config.getDesiredBytecodeVersion() <
-          bytecode::kNativePropertiesEncoding ||
-      !op.getPropertiesStorage()) {
+  DictionaryAttr dictAttr;
+  if (config.getDesiredBytecodeVersion() >= bytecode::kNativePropertiesEncoding)
+    dictAttr = op.getRawDictionaryAttrs();
+  else
     dictAttr = op.getAttrDictionary();
-  }
+  // Only number the operation's dictionary if it isn't empty.
   if (!dictAttr.empty())
     number(dictAttr);
 
   // Visit the operation properties (if any) to make sure referenced attributes
   // are numbered.
-  if (config.getDesiredBytecodeVersion() >= bytecode::kNativePropertiesEncoding &&
+  if (config.getDesiredBytecodeVersion() >=
+          bytecode::kNativePropertiesEncoding &&
       op.getPropertiesStorageSize()) {
     if (op.isRegistered()) {
       // Operation that have properties *must* implement this interface.

From af7c196fb8d10f58a704b5a8d142feacf2f0236d Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Wed, 10 Apr 2024 14:45:49 -0700
Subject: [PATCH 071/886] [lldb][sbdebugger] Move SBDebugger Broadcast bit enum
 into lldb-enumerations.h (#87409)

When the `eBroadcastBitProgressCategory` bit was originally added to
Debugger.h and SBDebugger.h, each corresponding bit was added in order
of the other bits that were previously there. Since `Debugger.h` has an
enum bit that `SBDebugger.h` does not, this meant that their offsets did
not match.

Instead of trying to keep the bit offsets in sync between the two, it's
preferable to just move SBDebugger's enum into the main enumerations
header and use the bits from there. This also requires that API tests using the bits from SBDebugger update their usage.
---
 lldb/include/lldb/API/SBDebugger.h                        | 7 -------
 lldb/include/lldb/lldb-enumerations.h                     | 8 ++++++++
 .../diagnostic_reporting/TestDiagnosticReporting.py       | 2 +-
 .../progress_reporting/TestProgressReporting.py           | 2 +-
 .../clang_modules/TestClangModuleBuildProgress.py         | 2 +-
 lldb/test/API/macosx/rosetta/TestRosetta.py               | 2 +-
 6 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/lldb/include/lldb/API/SBDebugger.h b/lldb/include/lldb/API/SBDebugger.h
index 62b2f91f5076d..cf5409a12a056 100644
--- a/lldb/include/lldb/API/SBDebugger.h
+++ b/lldb/include/lldb/API/SBDebugger.h
@@ -42,13 +42,6 @@ class LLDB_API SBInputReader {
 
 class LLDB_API SBDebugger {
 public:
-  FLAGS_ANONYMOUS_ENUM(){
-      eBroadcastBitProgress = (1 << 0),
-      eBroadcastBitWarning = (1 << 1),
-      eBroadcastBitError = (1 << 2),
-      eBroadcastBitProgressCategory = (1 << 3),
-  };
-
   SBDebugger();
 
   SBDebugger(const lldb::SBDebugger &rhs);
diff --git a/lldb/include/lldb/lldb-enumerations.h b/lldb/include/lldb/lldb-enumerations.h
index 646f7bfda9847..f3b07ea6d2039 100644
--- a/lldb/include/lldb/lldb-enumerations.h
+++ b/lldb/include/lldb/lldb-enumerations.h
@@ -1339,6 +1339,14 @@ enum AddressMaskRange {
   eAddressMaskRangeAll = eAddressMaskRangeAny,
 };
 
+/// Used by the debugger to indicate which events are being broadcasted.
+enum DebuggerBroadcastBit {
+  eBroadcastBitProgress = (1 << 0),
+  eBroadcastBitWarning = (1 << 1),
+  eBroadcastBitError = (1 << 2),
+  eBroadcastBitProgressCategory = (1 << 3),
+};
+
 } // namespace lldb
 
 #endif // LLDB_LLDB_ENUMERATIONS_H
diff --git a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
index 36a3be695628f..6353e3e8cbedb 100644
--- a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
+++ b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
@@ -15,7 +15,7 @@ def setUp(self):
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
             self.broadcaster,
-            lldb.SBDebugger.eBroadcastBitWarning | lldb.SBDebugger.eBroadcastBitError,
+            lldb.eBroadcastBitWarning | lldb.eBroadcastBitError,
         )
 
     def test_dwarf_symbol_loading_diagnostic_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
index 9af53845ca1b7..98988d7624da3 100644
--- a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
+++ b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
@@ -13,7 +13,7 @@ def setUp(self):
         TestBase.setUp(self)
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
-            self.broadcaster, lldb.SBDebugger.eBroadcastBitProgress
+            self.broadcaster, lldb.eBroadcastBitProgress
         )
 
     def test_dwarf_symbol_loading_progress_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
index 228f676aedf6a..33c7c269c081e 100644
--- a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
+++ b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
@@ -34,7 +34,7 @@ def test_clang_module_build_progress_report(self):
         # other unrelated progress events.
         broadcaster = self.dbg.GetBroadcaster()
         listener = lldbutil.start_listening_from(
-            broadcaster, lldb.SBDebugger.eBroadcastBitProgress
+            broadcaster, lldb.eBroadcastBitProgress
         )
 
         # Trigger module builds.
diff --git a/lldb/test/API/macosx/rosetta/TestRosetta.py b/lldb/test/API/macosx/rosetta/TestRosetta.py
index ce40de475ef16..669db95a1624c 100644
--- a/lldb/test/API/macosx/rosetta/TestRosetta.py
+++ b/lldb/test/API/macosx/rosetta/TestRosetta.py
@@ -49,7 +49,7 @@ def test_rosetta(self):
         if rosetta_debugserver_installed():
             broadcaster = self.dbg.GetBroadcaster()
             listener = lldbutil.start_listening_from(
-                broadcaster, lldb.SBDebugger.eBroadcastBitWarning
+                broadcaster, lldb.eBroadcastBitWarning
             )
 
             target, process, thread, bkpt = lldbutil.run_to_source_breakpoint(

From 2fdfea088c8d78119b74116b94bc6729ce0e3efe Mon Sep 17 00:00:00 2001
From: Stanislav Mekhanoshin <rampitec@users.noreply.github.com>
Date: Wed, 10 Apr 2024 14:50:54 -0700
Subject: [PATCH 072/886] [AMDGPU] Add v2i32 to the VS_64 types. NFCI. (#88318)

I am trying to use VOP3Inst with intrinsic taking v2i32 operand and it
fails to create patterm without it.
---
 llvm/lib/Target/AMDGPU/SIInstructions.td | 8 ++++----
 llvm/lib/Target/AMDGPU/SIRegisterInfo.td | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index 04f3a2f576053..d6d49889656bb 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -2087,7 +2087,7 @@ def : GCNPat <
 def : GCNPat <
   (DivergentUnaryFrag<fneg> (v2f32 VReg_64:$src)),
   (V_PK_ADD_F32 11 /* OP_SEL_1 | NEG_LO | HEG_HI */, VReg_64:$src,
-                11 /* OP_SEL_1 | NEG_LO | HEG_HI */, 0,
+                11 /* OP_SEL_1 | NEG_LO | HEG_HI */, (i64 0),
                 0, 0, 0, 0, 0)
 > {
   let SubtargetPredicate = HasPackedFP32Ops;
@@ -2999,7 +2999,7 @@ def : GCNPat<
 let SubtargetPredicate = HasPackedFP32Ops in {
 def : GCNPat<
   (fcanonicalize (v2f32 (VOP3PMods v2f32:$src, i32:$src_mods))),
-  (V_PK_MUL_F32 0, CONST.FP32_ONE, $src_mods, $src)
+  (V_PK_MUL_F32 0, (i64 CONST.FP32_ONE), $src_mods, $src)
 >;
 }
 
@@ -3007,7 +3007,7 @@ def : GCNPat<
 let SubtargetPredicate = isNotGFX12Plus in {
 def : GCNPat<
   (fcanonicalize (f64 (VOP3Mods f64:$src, i32:$src_mods))),
-  (V_MUL_F64_e64  0, CONST.FP64_ONE, $src_mods, $src)
+  (V_MUL_F64_e64  0, (i64 CONST.FP64_ONE), $src_mods, $src)
 >;
 }
 } // End AddedComplexity = -5
@@ -3369,7 +3369,7 @@ def : GCNPat <
              SRCMODS.NONE,
              (V_FRACT_F64_e64 $mods, $x),
              SRCMODS.NONE,
-             (V_MOV_B64_PSEUDO 0x3fefffffffffffff)),
+             (V_MOV_B64_PSEUDO (i64 0x3fefffffffffffff))),
          $x,
          (V_CMP_CLASS_F64_e64 SRCMODS.NONE, $x, (i32 3 /*NaN*/))))
 >;
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
index cb6591bf62449..01ed565bb756d 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
@@ -1046,7 +1046,7 @@ def VS_32_Lo128 : SIRegisterClass<"AMDGPU", [i32, f32, i16, f16, bf16, v2i16, v2
   let HasSGPR = 1;
 }
 
-def VS_64 : SIRegisterClass<"AMDGPU", [i64, f64, v2f32], 32, (add VReg_64, SReg_64)> {
+def VS_64 : SIRegisterClass<"AMDGPU", VReg_64.RegTypes, 32, (add VReg_64, SReg_64)> {
   let isAllocatable = 0;
   let HasVGPR = 1;
   let HasSGPR = 1;

From 9f6d08f2566a26144ea1753f80aebb1f2ecfdc63 Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Wed, 10 Apr 2024 14:54:30 -0700
Subject: [PATCH 073/886] Revert "[lldb][sbdebugger] Move SBDebugger Broadcast
 bit enum into lldb-enumerations.h" (#88324)

Reverts llvm/llvm-project#87409 due a missed update to the broadcast bit
causing a build failure on the x86_64 Debian buildbot.
---
 lldb/include/lldb/API/SBDebugger.h                        | 7 +++++++
 lldb/include/lldb/lldb-enumerations.h                     | 8 --------
 .../diagnostic_reporting/TestDiagnosticReporting.py       | 2 +-
 .../progress_reporting/TestProgressReporting.py           | 2 +-
 .../clang_modules/TestClangModuleBuildProgress.py         | 2 +-
 lldb/test/API/macosx/rosetta/TestRosetta.py               | 2 +-
 6 files changed, 11 insertions(+), 12 deletions(-)

diff --git a/lldb/include/lldb/API/SBDebugger.h b/lldb/include/lldb/API/SBDebugger.h
index cf5409a12a056..62b2f91f5076d 100644
--- a/lldb/include/lldb/API/SBDebugger.h
+++ b/lldb/include/lldb/API/SBDebugger.h
@@ -42,6 +42,13 @@ class LLDB_API SBInputReader {
 
 class LLDB_API SBDebugger {
 public:
+  FLAGS_ANONYMOUS_ENUM(){
+      eBroadcastBitProgress = (1 << 0),
+      eBroadcastBitWarning = (1 << 1),
+      eBroadcastBitError = (1 << 2),
+      eBroadcastBitProgressCategory = (1 << 3),
+  };
+
   SBDebugger();
 
   SBDebugger(const lldb::SBDebugger &rhs);
diff --git a/lldb/include/lldb/lldb-enumerations.h b/lldb/include/lldb/lldb-enumerations.h
index f3b07ea6d2039..646f7bfda9847 100644
--- a/lldb/include/lldb/lldb-enumerations.h
+++ b/lldb/include/lldb/lldb-enumerations.h
@@ -1339,14 +1339,6 @@ enum AddressMaskRange {
   eAddressMaskRangeAll = eAddressMaskRangeAny,
 };
 
-/// Used by the debugger to indicate which events are being broadcasted.
-enum DebuggerBroadcastBit {
-  eBroadcastBitProgress = (1 << 0),
-  eBroadcastBitWarning = (1 << 1),
-  eBroadcastBitError = (1 << 2),
-  eBroadcastBitProgressCategory = (1 << 3),
-};
-
 } // namespace lldb
 
 #endif // LLDB_LLDB_ENUMERATIONS_H
diff --git a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
index 6353e3e8cbedb..36a3be695628f 100644
--- a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
+++ b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
@@ -15,7 +15,7 @@ def setUp(self):
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
             self.broadcaster,
-            lldb.eBroadcastBitWarning | lldb.eBroadcastBitError,
+            lldb.SBDebugger.eBroadcastBitWarning | lldb.SBDebugger.eBroadcastBitError,
         )
 
     def test_dwarf_symbol_loading_diagnostic_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
index 98988d7624da3..9af53845ca1b7 100644
--- a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
+++ b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
@@ -13,7 +13,7 @@ def setUp(self):
         TestBase.setUp(self)
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
-            self.broadcaster, lldb.eBroadcastBitProgress
+            self.broadcaster, lldb.SBDebugger.eBroadcastBitProgress
         )
 
     def test_dwarf_symbol_loading_progress_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
index 33c7c269c081e..228f676aedf6a 100644
--- a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
+++ b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
@@ -34,7 +34,7 @@ def test_clang_module_build_progress_report(self):
         # other unrelated progress events.
         broadcaster = self.dbg.GetBroadcaster()
         listener = lldbutil.start_listening_from(
-            broadcaster, lldb.eBroadcastBitProgress
+            broadcaster, lldb.SBDebugger.eBroadcastBitProgress
         )
 
         # Trigger module builds.
diff --git a/lldb/test/API/macosx/rosetta/TestRosetta.py b/lldb/test/API/macosx/rosetta/TestRosetta.py
index 669db95a1624c..ce40de475ef16 100644
--- a/lldb/test/API/macosx/rosetta/TestRosetta.py
+++ b/lldb/test/API/macosx/rosetta/TestRosetta.py
@@ -49,7 +49,7 @@ def test_rosetta(self):
         if rosetta_debugserver_installed():
             broadcaster = self.dbg.GetBroadcaster()
             listener = lldbutil.start_listening_from(
-                broadcaster, lldb.eBroadcastBitWarning
+                broadcaster, lldb.SBDebugger.eBroadcastBitWarning
             )
 
             target, process, thread, bkpt = lldbutil.run_to_source_breakpoint(

From d8f1e5d2894f7f4edc2e85e63def456c7f430f34 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 10 Apr 2024 15:07:16 -0700
Subject: [PATCH 074/886] [APInt] Remove accumulator initialization from
 tcMultiply and tcFullMultiply. NFCI (#88202)

The tcMultiplyPart routine has a flag that says whether to add to the
accumulator or overwrite it. By using the overwrite mode on the first
iteration we don't need to initialize the accumulator to zero.

Note, the initialization in tcFullMultiply was only initializing the
first rhsParts of dst. tcMultiplyPart always overwrites the rhsParts+1
part that just contains the last carry. The first write to each part of
dst past rhsParts is a carry write so that's how the upper part of dst
is initialized.
---
 llvm/lib/Support/APInt.cpp | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index 8825025ec3213..18feca4c05533 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -2585,11 +2585,13 @@ int APInt::tcMultiply(WordType *dst, const WordType *lhs,
   assert(dst != lhs && dst != rhs);
 
   int overflow = 0;
-  tcSet(dst, 0, parts);
 
-  for (unsigned i = 0; i < parts; i++)
-    overflow |= tcMultiplyPart(&dst[i], lhs, rhs[i], 0, parts,
-                               parts - i, true);
+  for (unsigned i = 0; i < parts; i++) {
+    // Don't accumulate on the first iteration so we don't need to initalize
+    // dst to 0.
+    overflow |=
+        tcMultiplyPart(&dst[i], lhs, rhs[i], 0, parts, parts - i, i != 0);
+  }
 
   return overflow;
 }
@@ -2605,10 +2607,11 @@ void APInt::tcFullMultiply(WordType *dst, const WordType *lhs,
 
   assert(dst != lhs && dst != rhs);
 
-  tcSet(dst, 0, rhsParts);
-
-  for (unsigned i = 0; i < lhsParts; i++)
-    tcMultiplyPart(&dst[i], rhs, lhs[i], 0, rhsParts, rhsParts + 1, true);
+  for (unsigned i = 0; i < lhsParts; i++) {
+    // Don't accumulate on the first iteration so we don't need to initalize
+    // dst to 0.
+    tcMultiplyPart(&dst[i], rhs, lhs[i], 0, rhsParts, rhsParts + 1, i != 0);
+  }
 }
 
 // If RHS is zero LHS and REMAINDER are left unchanged, return one.

From a9d4ddd98a0bc495126027122fdca751b6841ceb Mon Sep 17 00:00:00 2001
From: Oskar Wirga <oskar.wirga@gmail.com>
Date: Wed, 10 Apr 2024 15:37:27 -0700
Subject: [PATCH 075/886] [MergeFuncs/CFI] Ensure all type metadata is
 propogated for CFI (#88218)

I noticed that we weren't propagating ALL type metadata that was
attached to CFI functions:

# BEFORE

```
; Function Attrs: minsize nounwind optsize ssp uwtable(sync)
define internal void @foo(ptr nocapture noundef readonly %0) #0 !dbg !62311 !type !34028 !type !34029 !type !34030
... fn merging
; Function Attrs: minsize nounwind optsize ssp uwtable(sync)
define internal void @foo(ptr nocapture noundef readonly %0) #0 !type !34028
```

# AFTER

```
; Function Attrs: minsize nounwind optsize ssp uwtable(sync)
define internal void @foo(ptr nocapture noundef readonly %0) #0 !dbg !62311 !type !34028 !type !34029 !type !34030
... fn merging
; Function Attrs: minsize nounwind optsize ssp uwtable(sync)
define internal void @foo(ptr nocapture noundef readonly %0) #0 !type !type !34028 !type !34029 !type !34030
```

This patch makes sure that the entire vector of metadata is copied over.
---
 llvm/lib/Transforms/IPO/MergeFunctions.cpp    | 12 +++++----
 .../Transforms/MergeFunc/cfi-thunk-merging.ll | 26 +++++++++----------
 2 files changed, 20 insertions(+), 18 deletions(-)

diff --git a/llvm/lib/Transforms/IPO/MergeFunctions.cpp b/llvm/lib/Transforms/IPO/MergeFunctions.cpp
index 05a3b169aaaf4..b50a700e09038 100644
--- a/llvm/lib/Transforms/IPO/MergeFunctions.cpp
+++ b/llvm/lib/Transforms/IPO/MergeFunctions.cpp
@@ -712,11 +712,13 @@ static bool canCreateThunkFor(Function *F) {
   return true;
 }
 
-/// Copy metadata from one function to another.
-static void copyMetadataIfPresent(Function *From, Function *To, StringRef Key) {
-  if (MDNode *MD = From->getMetadata(Key)) {
-    To->setMetadata(Key, MD);
-  }
+/// Copy all metadata of a specific kind from one function to another.
+static void copyMetadataIfPresent(Function *From, Function *To,
+                                  StringRef Kind) {
+  SmallVector<MDNode *, 4> MDs;
+  From->getMetadata(Kind, MDs);
+  for (MDNode *MD : MDs)
+    To->addMetadata(Kind, *MD);
 }
 
 // Replace G with a simple tail call to bitcast(F). Also (unless
diff --git a/llvm/test/Transforms/MergeFunc/cfi-thunk-merging.ll b/llvm/test/Transforms/MergeFunc/cfi-thunk-merging.ll
index d35d777282730..562cc1a973d81 100644
--- a/llvm/test/Transforms/MergeFunc/cfi-thunk-merging.ll
+++ b/llvm/test/Transforms/MergeFunc/cfi-thunk-merging.ll
@@ -98,7 +98,7 @@ attributes #3 = { noreturn nounwind }
 !4 = !{i64 0, !"_ZTSFiiE.generalized"}
 !5 = !{}
 ; CHECK-LABEL: define dso_local i32 @f
-; CHECK-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0:[0-9]+]] !type !2 !type !3 {
+; CHECK-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0:[0-9]+]] !type [[META2:![0-9]+]] !type [[META3:![0-9]+]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[A:%.*]] = alloca i32, align 4
@@ -119,7 +119,7 @@ attributes #3 = { noreturn nounwind }
 ;
 ;
 ; CHECK-LABEL: define dso_local i32 @g
-; CHECK-SAME: (i32 noundef [[B:%.*]]) #[[ATTR0]] !type !2 !type !3 {
+; CHECK-SAME: (i32 noundef [[B:%.*]]) #[[ATTR0]] !type [[META2]] !type [[META3]] {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 ; CHECK-NEXT:    [[FP:%.*]] = alloca ptr, align 8
@@ -130,11 +130,11 @@ attributes #3 = { noreturn nounwind }
 ; CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], ptr @f, ptr @f_thunk
 ; CHECK-NEXT:    store ptr [[COND]], ptr [[FP]], align 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[FP]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = call i1 @llvm.type.test(ptr [[TMP2]], metadata !"_ZTSFiiE"), !nosanitize !4
-; CHECK-NEXT:    br i1 [[TMP3]], label [[CONT:%.*]], label [[TRAP:%.*]], !nosanitize !4
+; CHECK-NEXT:    [[TMP3:%.*]] = call i1 @llvm.type.test(ptr [[TMP2]], metadata !"_ZTSFiiE"), !nosanitize [[META4:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP3]], label [[CONT:%.*]], label [[TRAP:%.*]], !nosanitize [[META4]]
 ; CHECK:       trap:
-; CHECK-NEXT:    call void @llvm.ubsantrap(i8 2) #[[ATTR3:[0-9]+]], !nosanitize !4
-; CHECK-NEXT:    unreachable, !nosanitize !4
+; CHECK-NEXT:    call void @llvm.ubsantrap(i8 2) #[[ATTR3:[0-9]+]], !nosanitize [[META4]]
+; CHECK-NEXT:    unreachable, !nosanitize [[META4]]
 ; CHECK:       cont:
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[B_ADDR]], align 4
 ; CHECK-NEXT:    [[CALL:%.*]] = call i32 [[TMP2]](i32 noundef [[TMP4]])
@@ -142,13 +142,13 @@ attributes #3 = { noreturn nounwind }
 ;
 ;
 ; CHECK-LABEL: define dso_local i32 @f_thunk
-; CHECK-SAME: (i32 noundef [[TMP0:%.*]]) #[[ATTR0]] !type !2 {
+; CHECK-SAME: (i32 noundef [[TMP0:%.*]]) #[[ATTR0]] !type [[META2]] !type [[META3]] {
 ; CHECK-NEXT:    [[TMP2:%.*]] = tail call i32 @f(i32 noundef [[TMP0]]) #[[ATTR0]]
 ; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
 ;
 ; LOWERTYPETESTS-LABEL: define dso_local i32 @f
-; LOWERTYPETESTS-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0:[0-9]+]] !type !2 !type !3 {
+; LOWERTYPETESTS-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0:[0-9]+]] !type [[META2:![0-9]+]] !type [[META3:![0-9]+]] {
 ; LOWERTYPETESTS-NEXT:  entry:
 ; LOWERTYPETESTS-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
 ; LOWERTYPETESTS-NEXT:    [[A:%.*]] = alloca i32, align 4
@@ -169,7 +169,7 @@ attributes #3 = { noreturn nounwind }
 ;
 ;
 ; LOWERTYPETESTS-LABEL: define dso_local i32 @g
-; LOWERTYPETESTS-SAME: (i32 noundef [[B:%.*]]) #[[ATTR0]] !type !2 !type !3 {
+; LOWERTYPETESTS-SAME: (i32 noundef [[B:%.*]]) #[[ATTR0]] !type [[META2]] !type [[META3]] {
 ; LOWERTYPETESTS-NEXT:  entry:
 ; LOWERTYPETESTS-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
 ; LOWERTYPETESTS-NEXT:    [[FP:%.*]] = alloca ptr, align 8
@@ -186,10 +186,10 @@ attributes #3 = { noreturn nounwind }
 ; LOWERTYPETESTS-NEXT:    [[TMP6:%.*]] = shl i64 [[TMP4]], 61
 ; LOWERTYPETESTS-NEXT:    [[TMP7:%.*]] = or i64 [[TMP5]], [[TMP6]]
 ; LOWERTYPETESTS-NEXT:    [[TMP8:%.*]] = icmp ule i64 [[TMP7]], 1
-; LOWERTYPETESTS-NEXT:    br i1 [[TMP8]], label [[CONT:%.*]], label [[TRAP:%.*]], !nosanitize !4
+; LOWERTYPETESTS-NEXT:    br i1 [[TMP8]], label [[CONT:%.*]], label [[TRAP:%.*]], !nosanitize [[META4:![0-9]+]]
 ; LOWERTYPETESTS:       trap:
-; LOWERTYPETESTS-NEXT:    call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !nosanitize !4
-; LOWERTYPETESTS-NEXT:    unreachable, !nosanitize !4
+; LOWERTYPETESTS-NEXT:    call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !nosanitize [[META4]]
+; LOWERTYPETESTS-NEXT:    unreachable, !nosanitize [[META4]]
 ; LOWERTYPETESTS:       cont:
 ; LOWERTYPETESTS-NEXT:    [[TMP9:%.*]] = load i32, ptr [[B_ADDR]], align 4
 ; LOWERTYPETESTS-NEXT:    [[CALL:%.*]] = call i32 [[TMP2]](i32 noundef [[TMP9]])
@@ -197,7 +197,7 @@ attributes #3 = { noreturn nounwind }
 ;
 ;
 ; LOWERTYPETESTS-LABEL: define dso_local i32 @f_thunk
-; LOWERTYPETESTS-SAME: (i32 noundef [[TMP0:%.*]]) #[[ATTR0]] !type !2 {
+; LOWERTYPETESTS-SAME: (i32 noundef [[TMP0:%.*]]) #[[ATTR0]] !type [[META2]] !type [[META3]] {
 ; LOWERTYPETESTS-NEXT:    [[TMP2:%.*]] = tail call i32 @f(i32 noundef [[TMP0]]) #[[ATTR0]]
 ; LOWERTYPETESTS-NEXT:    ret i32 [[TMP2]]
 ;

From 8136ac1c42dcfdd070f0bcba0f06424093df22db Mon Sep 17 00:00:00 2001
From: Daniel Chen <cdchen@ca.ibm.com>
Date: Wed, 10 Apr 2024 19:22:38 -0400
Subject: [PATCH 076/886] [Flang] Define c_int_fast16_t and c_int_fast32_t for
 PowerPC. (#88292)

On Linux, PowerPC defines `int_fast16_t` and `int_fast32_t` as `long`.
Need to update the corresponding type, `c_int_fast16_t` and
`c_int_fast32_t` in `iso_c_binding` module so they are interoparable.
---
 flang/lib/Frontend/CompilerInvocation.cpp       | 17 +++++++++--------
 flang/module/iso_c_binding.f90                  |  8 ++++++++
 .../test/Driver/predefined-macros-powerpc2.f90  | 13 +++++++++++++
 3 files changed, 30 insertions(+), 8 deletions(-)
 create mode 100644 flang/test/Driver/predefined-macros-powerpc2.f90

diff --git a/flang/lib/Frontend/CompilerInvocation.cpp b/flang/lib/Frontend/CompilerInvocation.cpp
index 8ce6ab7baf481..e432c5a302754 100644
--- a/flang/lib/Frontend/CompilerInvocation.cpp
+++ b/flang/lib/Frontend/CompilerInvocation.cpp
@@ -1333,6 +1333,15 @@ void CompilerInvocation::setDefaultPredefinitions() {
   }
 
   llvm::Triple targetTriple{llvm::Triple(this->targetOpts.triple)};
+  if (targetTriple.isPPC()) {
+    // '__powerpc__' is a generic macro for any PowerPC cases. e.g. Max integer
+    // size.
+    fortranOptions.predefinitions.emplace_back("__powerpc__", "1");
+  }
+  if (targetTriple.isOSLinux()) {
+    fortranOptions.predefinitions.emplace_back("__linux__", "1");
+  }
+
   switch (targetTriple.getArch()) {
   default:
     break;
@@ -1340,14 +1349,6 @@ void CompilerInvocation::setDefaultPredefinitions() {
     fortranOptions.predefinitions.emplace_back("__x86_64__", "1");
     fortranOptions.predefinitions.emplace_back("__x86_64", "1");
     break;
-  case llvm::Triple::ArchType::ppc:
-  case llvm::Triple::ArchType::ppcle:
-  case llvm::Triple::ArchType::ppc64:
-  case llvm::Triple::ArchType::ppc64le:
-    // '__powerpc__' is a generic macro for any PowerPC cases. e.g. Max integer
-    // size.
-    fortranOptions.predefinitions.emplace_back("__powerpc__", "1");
-    break;
   }
 }
 
diff --git a/flang/module/iso_c_binding.f90 b/flang/module/iso_c_binding.f90
index 1661fd5a6dcf6..eb0f8f2ef59ad 100644
--- a/flang/module/iso_c_binding.f90
+++ b/flang/module/iso_c_binding.f90
@@ -58,9 +58,17 @@ module iso_c_binding
     c_int_least8_t = c_int8_t, &
     c_int_fast8_t = c_int8_t, &
     c_int_least16_t = c_int16_t, &
+#if defined(__linux__) && defined(__powerpc__)
+    c_int_fast16_t = c_long, &
+#else
     c_int_fast16_t = c_int16_t, &
+#endif
     c_int_least32_t = c_int32_t, &
+#if defined(__linux__) && defined(__powerpc__)
+    c_int_fast32_t = c_long, &
+#else
     c_int_fast32_t = c_int32_t, &
+#endif
     c_int_least64_t = c_int64_t, &
     c_int_fast64_t = c_int64_t, &
     c_int_least128_t = c_int128_t, &
diff --git a/flang/test/Driver/predefined-macros-powerpc2.f90 b/flang/test/Driver/predefined-macros-powerpc2.f90
new file mode 100644
index 0000000000000..6e10235e21f86
--- /dev/null
+++ b/flang/test/Driver/predefined-macros-powerpc2.f90
@@ -0,0 +1,13 @@
+! Test predefined macro for PowerPC architecture
+
+! RUN: %flang_fc1 -triple ppc64le-unknown-linux -cpp -E %s | FileCheck %s
+! REQUIRES: target=powerpc{{.*}}
+
+! CHECK: integer :: var1 = 1
+! CHECK: integer :: var2 = 1
+
+#if defined(__linux__) && defined(__powerpc__)
+  integer :: var1 = __powerpc__
+  integer :: var2 = __linux__
+#endif
+end program

From acb7ddc5cf2f23416f65dcdc6c7fd08850ad961d Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze@braunis.de>
Date: Wed, 10 Apr 2024 16:24:02 -0700
Subject: [PATCH 077/886] [WebAssembly] Remove threadlocal.address when
 disabling TLS (#88209)

Remove `llvm.threadlocal.address` intrinsic usage when disabling TLS.
This fixes errors revealed by the stricter IR verification introduced in
PR #87841.
---
 .../WebAssembly/WebAssemblyTargetMachine.cpp  | 11 ++++++++
 .../WebAssembly/tls-general-dynamic.ll        | 26 +++++++++++++------
 .../CodeGen/WebAssembly/tls-local-exec.ll     | 17 ++++++++----
 3 files changed, 41 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
index 70685b2e3bb2d..769ee765e1907 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
@@ -291,6 +291,17 @@ class CoalesceFeaturesAndStripAtomics final : public ModulePass {
     bool Stripped = false;
     for (auto &GV : M.globals()) {
       if (GV.isThreadLocal()) {
+        // replace `@llvm.threadlocal.address.pX(GV)` with `GV`.
+        for (Use &U : make_early_inc_range(GV.uses())) {
+          if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(U.getUser())) {
+            if (II->getIntrinsicID() == Intrinsic::threadlocal_address &&
+                II->getArgOperand(0) == &GV) {
+              II->replaceAllUsesWith(&GV);
+              II->eraseFromParent();
+            }
+          }
+        }
+
         Stripped = true;
         GV.setThreadLocal(false);
       }
diff --git a/llvm/test/CodeGen/WebAssembly/tls-general-dynamic.ll b/llvm/test/CodeGen/WebAssembly/tls-general-dynamic.ll
index 46ab62dfaaa23..006b73922d2b9 100644
--- a/llvm/test/CodeGen/WebAssembly/tls-general-dynamic.ll
+++ b/llvm/test/CodeGen/WebAssembly/tls-general-dynamic.ll
@@ -14,7 +14,9 @@ define i32 @address_of_tls() {
 
   ; NO-TLS-NEXT: i32.const tls
   ; NO-TLS-NEXT: return
-  ret i32 ptrtoint(ptr @tls to i32)
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  %r = ptrtoint ptr %p to i32
+  ret i32 %r
 }
 
 ; CHECK-LABEL: address_of_tls_external:
@@ -25,7 +27,9 @@ define i32 @address_of_tls_external() {
 
   ; NO-TLS-NEXT: i32.const tls_external
   ; NO-TLS-NEXT: return
-  ret i32 ptrtoint(ptr @tls_external to i32)
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls_external)
+  %r = ptrtoint ptr %p to i32
+  ret i32 %r
 }
 
 ; CHECK-LABEL: ptr_to_tls:
@@ -38,7 +42,8 @@ define ptr @ptr_to_tls() {
 
   ; NO-TLS-NEXT: i32.const tls
   ; NO-TLS-NEXT: return
-  ret ptr @tls
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  ret ptr %p
 }
 
 ; CHECK-LABEL: ptr_to_tls_external:
@@ -49,7 +54,8 @@ define ptr @ptr_to_tls_external() {
 
   ; NO-TLS-NEXT: i32.const tls_external
   ; NO-TLS-NEXT: return
-  ret ptr @tls_external
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls_external)
+  ret ptr %p
 }
 
 ; CHECK-LABEL: tls_load:
@@ -64,7 +70,8 @@ define i32 @tls_load() {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.load tls
   ; NO-TLS-NEXT: return
-  %tmp = load i32, ptr @tls, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  %tmp = load i32, ptr %p, align 4
   ret i32 %tmp
 }
 
@@ -78,7 +85,8 @@ define i32 @tls_load_external() {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.load tls_external
   ; NO-TLS-NEXT: return
-  %tmp = load i32, ptr @tls_external, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls_external)
+  %tmp = load i32, ptr %p, align 4
   ret i32 %tmp
 }
 
@@ -94,7 +102,8 @@ define void @tls_store(i32 %x) {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.store tls
   ; NO-TLS-NEXT: return
-  store i32 %x, ptr @tls, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  store i32 %x, ptr %p, align 4
   ret void
 }
 
@@ -108,7 +117,8 @@ define void @tls_store_external(i32 %x) {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.store tls_external
   ; NO-TLS-NEXT: return
-  store i32 %x, ptr @tls_external, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls_external)
+  store i32 %x, ptr %p, align 4
   ret void
 }
 
diff --git a/llvm/test/CodeGen/WebAssembly/tls-local-exec.ll b/llvm/test/CodeGen/WebAssembly/tls-local-exec.ll
index 3aa044c34789e..dc0d40c7973ad 100644
--- a/llvm/test/CodeGen/WebAssembly/tls-local-exec.ll
+++ b/llvm/test/CodeGen/WebAssembly/tls-local-exec.ll
@@ -20,7 +20,9 @@ define i32 @address_of_tls() {
 
   ; NO-TLS-NEXT: i32.const tls
   ; NO-TLS-NEXT: return
-  ret i32 ptrtoint(ptr @tls to i32)
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  %r = ptrtoint ptr %p to i32
+  ret i32 %r
 }
 
 ; CHECK-LABEL: address_of_tls_external:
@@ -33,7 +35,9 @@ define i32 @address_of_tls_external() {
 
   ; NO-TLS-NEXT: i32.const tls_external
   ; NO-TLS-NEXT: return
-  ret i32 ptrtoint(ptr @tls_external to i32)
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls_external)
+  %r = ptrtoint ptr %p to i32
+  ret i32 %r
 }
 
 ; CHECK-LABEL: ptr_to_tls:
@@ -46,7 +50,8 @@ define ptr @ptr_to_tls() {
 
   ; NO-TLS-NEXT: i32.const tls
   ; NO-TLS-NEXT: return
-  ret ptr @tls
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  ret ptr %p
 }
 
 ; CHECK-LABEL: tls_load:
@@ -61,7 +66,8 @@ define i32 @tls_load() {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.load tls
   ; NO-TLS-NEXT: return
-  %tmp = load i32, ptr @tls, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  %tmp = load i32, ptr %p, align 4
   ret i32 %tmp
 }
 
@@ -77,7 +83,8 @@ define void @tls_store(i32 %x) {
   ; NO-TLS-NEXT: i32.const 0
   ; NO-TLS-NEXT: i32.store tls
   ; NO-TLS-NEXT: return
-  store i32 %x, ptr @tls, align 4
+  %p = call ptr @llvm.threadlocal.address.p0(ptr @tls)
+  store i32 %x, ptr %p, align 4
   ret void
 }
 

From d927d1867fa760836538beef2c4531c1a0b04e24 Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Wed, 10 Apr 2024 16:30:42 -0700
Subject: [PATCH 078/886] [UBSAN] Emit optimization remarks (#88304)

---
 llvm/lib/IR/DiagnosticInfo.cpp                |  6 ++-
 .../Instrumentation/LowerAllowCheckPass.cpp   | 46 +++++++++++++++++--
 .../lower-builtin-allow-check-remarks.ll      | 24 ++++++++++
 3 files changed, 70 insertions(+), 6 deletions(-)
 create mode 100644 llvm/test/Transforms/lower-builtin-allow-check-remarks.ll

diff --git a/llvm/lib/IR/DiagnosticInfo.cpp b/llvm/lib/IR/DiagnosticInfo.cpp
index 342c4cbbc39d6..31971b179fb4b 100644
--- a/llvm/lib/IR/DiagnosticInfo.cpp
+++ b/llvm/lib/IR/DiagnosticInfo.cpp
@@ -179,8 +179,12 @@ DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key,
   else if (isa<Constant>(V)) {
     raw_string_ostream OS(Val);
     V->printAsOperand(OS, /*PrintType=*/false);
-  } else if (auto *I = dyn_cast<Instruction>(V))
+  } else if (auto *I = dyn_cast<Instruction>(V)) {
     Val = I->getOpcodeName();
+  } else if (auto *MD = dyn_cast<MetadataAsValue>(V)) {
+    if (auto *S = dyn_cast<MDString>(MD->getMetadata()))
+      Val = S->getString();
+  }
 }
 
 DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, const Type *T)
diff --git a/llvm/lib/Transforms/Instrumentation/LowerAllowCheckPass.cpp b/llvm/lib/Transforms/Instrumentation/LowerAllowCheckPass.cpp
index cdc8318f088c2..465fa41b6c663 100644
--- a/llvm/lib/Transforms/Instrumentation/LowerAllowCheckPass.cpp
+++ b/llvm/lib/Transforms/Instrumentation/LowerAllowCheckPass.cpp
@@ -10,11 +10,16 @@
 
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/Analysis/ProfileSummaryInfo.h"
 #include "llvm/IR/Constant.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Intrinsics.h"
+#include "llvm/IR/Metadata.h"
 #include "llvm/Support/RandomNumberGenerator.h"
 #include <memory>
 #include <random>
@@ -35,13 +40,41 @@ static cl::opt<float>
 STATISTIC(NumChecksTotal, "Number of checks");
 STATISTIC(NumChecksRemoved, "Number of removed checks");
 
+struct RemarkInfo {
+  ore::NV Kind;
+  ore::NV F;
+  ore::NV BB;
+  explicit RemarkInfo(IntrinsicInst *II)
+      : Kind("Kind", II->getArgOperand(0)),
+        F("Function", II->getParent()->getParent()),
+        BB("Block", II->getParent()->getName()) {}
+};
+
+static void emitRemark(IntrinsicInst *II, OptimizationRemarkEmitter &ORE,
+                       bool Removed) {
+  if (Removed) {
+    ORE.emit([&]() {
+      RemarkInfo Info(II);
+      return OptimizationRemark(DEBUG_TYPE, "Removed", II)
+             << "Removed check: Kind=" << Info.Kind << " F=" << Info.F
+             << " BB=" << Info.BB;
+    });
+  } else {
+    ORE.emit([&]() {
+      RemarkInfo Info(II);
+      return OptimizationRemarkMissed(DEBUG_TYPE, "Allowed", II)
+             << "Allowed check: Kind=" << Info.Kind << " F=" << Info.F
+             << " BB=" << Info.BB;
+    });
+  }
+}
+
 static bool removeUbsanTraps(Function &F, const BlockFrequencyInfo &BFI,
-                             const ProfileSummaryInfo *PSI) {
+                             const ProfileSummaryInfo *PSI,
+                             OptimizationRemarkEmitter &ORE) {
   SmallVector<std::pair<IntrinsicInst *, bool>, 16> ReplaceWithValue;
   std::unique_ptr<RandomNumberGenerator> Rng;
 
-  // TODO:
-  // https://github.com/llvm/llvm-project/pull/84858#discussion_r1520603139
   auto ShouldRemove = [&](bool IsHot) {
     if (!RandomRate.getNumOccurrences())
       return IsHot;
@@ -75,6 +108,7 @@ static bool removeUbsanTraps(Function &F, const BlockFrequencyInfo &BFI,
         });
         if (ToRemove)
           ++NumChecksRemoved;
+        emitRemark(II, ORE, ToRemove);
         break;
       }
       default:
@@ -99,9 +133,11 @@ PreservedAnalyses LowerAllowCheckPass::run(Function &F,
   ProfileSummaryInfo *PSI =
       MAMProxy.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
   BlockFrequencyInfo &BFI = AM.getResult<BlockFrequencyAnalysis>(F);
+  OptimizationRemarkEmitter &ORE =
+      AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
 
-  return removeUbsanTraps(F, BFI, PSI) ? PreservedAnalyses::none()
-                                       : PreservedAnalyses::all();
+  return removeUbsanTraps(F, BFI, PSI, ORE) ? PreservedAnalyses::none()
+                                            : PreservedAnalyses::all();
 }
 
 bool LowerAllowCheckPass::IsRequested() {
diff --git a/llvm/test/Transforms/lower-builtin-allow-check-remarks.ll b/llvm/test/Transforms/lower-builtin-allow-check-remarks.ll
new file mode 100644
index 0000000000000..3422ab1b56d32
--- /dev/null
+++ b/llvm/test/Transforms/lower-builtin-allow-check-remarks.ll
@@ -0,0 +1,24 @@
+; RUN: opt < %s -passes='require<profile-summary>,function(lower-allow-check)' -lower-allow-check-random-rate=1 -pass-remarks=lower-allow-check -pass-remarks-missed=lower-allow-check -S 2>&1 | FileCheck %s
+; RUN: opt < %s -passes='require<profile-summary>,function(lower-allow-check)' -lower-allow-check-random-rate=0 -pass-remarks=lower-allow-check -pass-remarks-missed=lower-allow-check -S 2>&1 | FileCheck %s --check-prefixes=REMOVE
+
+; CHECK: remark: <unknown>:0:0: Allowed check: Kind=test_check F=test_runtime BB=entry1
+; CHECK: remark: <unknown>:0:0: Allowed check: Kind=7 F=test_ubsan BB=entry2
+
+; REMOVE: remark: <unknown>:0:0: Removed check: Kind=test_check F=test_runtime BB=entry1
+; REMOVE: remark: <unknown>:0:0: Removed check: Kind=7 F=test_ubsan BB=entry2
+
+target triple = "x86_64-pc-linux-gnu"
+
+define i1 @test_runtime() local_unnamed_addr {
+entry1:
+  %allow = call i1 @llvm.allow.runtime.check(metadata !"test_check")
+  ret i1 %allow
+}
+
+declare i1 @llvm.allow.runtime.check(metadata) nounwind
+
+define i1 @test_ubsan() local_unnamed_addr {
+entry2:
+  %allow = call i1 @llvm.allow.ubsan.check(i8 7)
+  ret i1 %allow
+}

From 6ef4450705473e5cccb025219e8980999f456b71 Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Thu, 11 Apr 2024 07:37:12 +0800
Subject: [PATCH 079/886] [clang] Fix -Wunused-function in CGStmtOpenMP.cpp
 (NFC)

llvm-project/clang/lib/CodeGen/CGStmtOpenMP.cpp:7959:13:
error: unused function 'emitTargetTeamsLoopCodegenStatus' [-Werror,-Wunused-function]
static void emitTargetTeamsLoopCodegenStatus(CodeGenFunction &CGF,
            ^
1 error generated.
---
 clang/lib/CodeGen/CGStmtOpenMP.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 3bf99366b69ce..a0a8a07c76ba1 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -7956,10 +7956,10 @@ void CodeGenFunction::EmitOMPTeamsGenericLoopDirective(
                                    [](CodeGenFunction &) { return nullptr; });
 }
 
+#ifndef NDEBUG
 static void emitTargetTeamsLoopCodegenStatus(CodeGenFunction &CGF,
                                              std::string StatusMsg,
                                              const OMPExecutableDirective &D) {
-#ifndef NDEBUG
   bool IsDevice = CGF.CGM.getLangOpts().OpenMPIsTargetDevice;
   if (IsDevice)
     StatusMsg += ": DEVICE";
@@ -7972,8 +7972,8 @@ static void emitTargetTeamsLoopCodegenStatus(CodeGenFunction &CGF,
   unsigned LineNo =
       PLoc.isValid() ? PLoc.getLine() : SM.getExpansionLineNumber(L);
   llvm::dbgs() << StatusMsg << ": " << FileName << ": " << LineNo << "\n";
-#endif
 }
+#endif
 
 static void emitTargetTeamsGenericLoopRegionAsParallel(
     CodeGenFunction &CGF, PrePostActionTy &Action,

From 19e516fbed809af094ce195a6a5baa2e1f30f3cd Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 10 Apr 2024 23:35:52 +0000
Subject: [PATCH 080/886] [gn build] Port 1fda1776e32b

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 66e8084d5808a..53dc5b92c2d98 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -361,6 +361,7 @@ if (current_toolchain == default_toolchain) {
       "__chrono/parser_std_format_spec.h",
       "__chrono/statically_widen.h",
       "__chrono/steady_clock.h",
+      "__chrono/sys_info.h",
       "__chrono/system_clock.h",
       "__chrono/time_point.h",
       "__chrono/time_zone.h",

From 402706668362fee8f9a9d29fb6d4628df4d4fc42 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 10 Apr 2024 23:35:53 +0000
Subject: [PATCH 081/886] [gn build] Port 59e66c515a47

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 53dc5b92c2d98..c3eda6878090f 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -466,6 +466,7 @@ if (current_toolchain == default_toolchain) {
       "__format/formatter_pointer.h",
       "__format/formatter_string.h",
       "__format/formatter_tuple.h",
+      "__format/indic_conjunct_break_table.h",
       "__format/parser_std_format_spec.h",
       "__format/range_default_formatter.h",
       "__format/range_formatter.h",

From 233edab8765686bd44611f9f7319d3ffbc12fbab Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 10 Apr 2024 23:35:54 +0000
Subject: [PATCH 082/886] [gn build] Port 5d7d6ad663f8

---
 llvm/utils/gn/secondary/clang/unittests/AST/Interp/BUILD.gn | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/llvm/utils/gn/secondary/clang/unittests/AST/Interp/BUILD.gn b/llvm/utils/gn/secondary/clang/unittests/AST/Interp/BUILD.gn
index dd47a3d2d3459..fcdb9a5b1aeb2 100644
--- a/llvm/utils/gn/secondary/clang/unittests/AST/Interp/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/unittests/AST/Interp/BUILD.gn
@@ -9,5 +9,8 @@ unittest("InterpTests") {
     "//clang/lib/Testing",
     "//clang/lib/Tooling",
   ]
-  sources = [ "Descriptor.cpp" ]
+  sources = [
+    "Descriptor.cpp",
+    "toAPValue.cpp",
+  ]
 }

From 9786a3b4cf9d050a6f87358e3295da3d32fade5c Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 10 Apr 2024 23:36:17 +0000
Subject: [PATCH 083/886] [gn build] Port 0a1317564a6b

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 llvm/utils/gn/secondary/libcxx/src/BUILD.gn     | 2 --
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index c3eda6878090f..4270bae57ff2c 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -932,6 +932,7 @@ if (current_toolchain == default_toolchain) {
       "__utility/pair.h",
       "__utility/piecewise_construct.h",
       "__utility/priority_tag.h",
+      "__utility/private_constructor_tag.h",
       "__utility/rel_ops.h",
       "__utility/small_buffer.h",
       "__utility/swap.h",
diff --git a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
index 90f6f5d0f1458..5da8db4574a0c 100644
--- a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
@@ -315,8 +315,6 @@ if (libcxx_enable_experimental) {
     sources = [ "experimental/keep.cpp" ]
     if (libcxx_enable_filesystem && libcxx_enable_time_zone_database) {
       sources += [
-        "include/tzdb/leap_second_private.h",
-        "include/tzdb/time_zone_link_private.h",
         "include/tzdb/time_zone_private.h",
         "include/tzdb/types_private.h",
         "include/tzdb/tzdb_list_private.h",

From be10070f91b86a6f126d2451852242bfcb2cd366 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 10 Apr 2024 23:41:51 +0000
Subject: [PATCH 084/886] Revert "[Driver] Ensure ToolChain::LibraryPaths is
 not empty for non-Darwin"

This reverts commit ccdebbae4d77d3efc236af92c22941de5d437e01.

Causes test failures in the presence of Android runtime libraries in resource-dir.
See comments on https://github.com/llvm/llvm-project/pull/87866.
---
 clang/lib/Driver/ToolChain.cpp                |  8 +----
 clang/test/Driver/arm-compiler-rt.c           | 14 ++++-----
 clang/test/Driver/cl-link.c                   | 16 +++++-----
 clang/test/Driver/compiler-rt-unwind.c        |  6 ++--
 clang/test/Driver/coverage-ld.c               |  8 ++---
 clang/test/Driver/instrprof-ld.c              | 16 +++++-----
 clang/test/Driver/linux-ld.c                  |  6 ++--
 clang/test/Driver/mingw-sanitizers.c          | 16 +++++-----
 clang/test/Driver/msp430-toolchain.c          |  4 +--
 .../Driver/print-libgcc-file-name-clangrt.c   | 12 ++++----
 clang/test/Driver/print-runtime-dir.c         |  6 ++++
 clang/test/Driver/riscv32-toolchain-extra.c   |  6 ++--
 clang/test/Driver/riscv32-toolchain.c         |  6 ++--
 clang/test/Driver/riscv64-toolchain-extra.c   |  6 ++--
 clang/test/Driver/riscv64-toolchain.c         |  6 ++--
 clang/test/Driver/sanitizer-ld.c              | 30 +++++++++----------
 clang/test/Driver/wasm-toolchain.c            | 18 +++++------
 clang/test/Driver/wasm-toolchain.cpp          | 16 +++++-----
 clang/test/Driver/windows-cross.c             | 18 +++++------
 clang/test/Driver/zos-ld.c                    | 12 ++++----
 .../test/Driver/msvc-dependent-lib-flags.f90  |  8 ++---
 21 files changed, 119 insertions(+), 119 deletions(-)

diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 237092ed07e5d..03450fc0f57b9 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -796,13 +796,7 @@ ToolChain::getTargetSubDirPath(StringRef BaseDir) const {
 std::optional<std::string> ToolChain::getRuntimePath() const {
   SmallString<128> P(D.ResourceDir);
   llvm::sys::path::append(P, "lib");
-  if (auto Ret = getTargetSubDirPath(P))
-    return Ret;
-  // Darwin does not use per-target runtime directory.
-  if (Triple.isOSDarwin())
-    return {};
-  llvm::sys::path::append(P, Triple.str());
-  return std::string(P);
+  return getTargetSubDirPath(P);
 }
 
 std::optional<std::string> ToolChain::getStdlibPath() const {
diff --git a/clang/test/Driver/arm-compiler-rt.c b/clang/test/Driver/arm-compiler-rt.c
index cb6c29f48a781..5e9e528400d08 100644
--- a/clang/test/Driver/arm-compiler-rt.c
+++ b/clang/test/Driver/arm-compiler-rt.c
@@ -10,47 +10,47 @@
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABI
-// ARM-GNUEABI: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-GNUEABI: "{{.*[/\\]}}libclang_rt.builtins-arm.a"
 
 // RUN: %clang -target arm-linux-gnueabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=hard -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABI-ABI
-// ARM-GNUEABI-ABI: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-GNUEABI-ABI: "{{.*[/\\]}}libclang_rt.builtins-armhf.a"
 
 // RUN: %clang -target arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABIHF
-// ARM-GNUEABIHF: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-GNUEABIHF: "{{.*[/\\]}}libclang_rt.builtins-armhf.a"
 
 // RUN: %clang -target arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=soft -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABIHF-ABI
-// ARM-GNUEABIHF-ABI: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-GNUEABIHF-ABI: "{{.*[/\\]}}libclang_rt.builtins-arm.a"
 
 // RUN: %clang -target arm-windows-itanium \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-WINDOWS
-// ARM-WINDOWS: "{{.*[/\\]}}clang_rt.builtins.lib"
+// ARM-WINDOWS: "{{.*[/\\]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -target arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-ANDROID
-// ARM-ANDROID: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-ANDROID: "{{.*[/\\]}}libclang_rt.builtins-arm-android.a"
 
 // RUN: not %clang --target=arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=hard -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-ANDROIDHF
-// ARM-ANDROIDHF: "{{.*[/\\]}}libclang_rt.builtins.a"
+// ARM-ANDROIDHF: "{{.*[/\\]}}libclang_rt.builtins-armhf-android.a"
 
diff --git a/clang/test/Driver/cl-link.c b/clang/test/Driver/cl-link.c
index ffd0b5ac4bade..444f0c01b3f99 100644
--- a/clang/test/Driver/cl-link.c
+++ b/clang/test/Driver/cl-link.c
@@ -13,20 +13,20 @@
 // ASAN: link.exe
 // ASAN: "-debug"
 // ASAN: "-incremental:no"
-// ASAN: "{{[^"]*}}clang_rt.asan.lib"
-// ASAN: "-wholearchive:{{.*}}clang_rt.asan.lib"
-// ASAN: "{{[^"]*}}clang_rt.asan_cxx.lib"
-// ASAN: "-wholearchive:{{.*}}clang_rt.asan_cxx.lib"
+// ASAN: "{{[^"]*}}clang_rt.asan-i386.lib"
+// ASAN: "-wholearchive:{{.*}}clang_rt.asan-i386.lib"
+// ASAN: "{{[^"]*}}clang_rt.asan_cxx-i386.lib"
+// ASAN: "-wholearchive:{{.*}}clang_rt.asan_cxx-i386.lib"
 // ASAN: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl -m32 -arch:IA32 --target=i386-pc-win32 /MD /Tc%s -fuse-ld=link -### -fsanitize=address 2>&1 | FileCheck --check-prefix=ASAN-MD %s
 // ASAN-MD: link.exe
 // ASAN-MD: "-debug"
 // ASAN-MD: "-incremental:no"
-// ASAN-MD: "{{.*}}clang_rt.asan_dynamic.lib"
-// ASAN-MD: "{{[^"]*}}clang_rt.asan_dynamic_runtime_thunk.lib"
+// ASAN-MD: "{{.*}}clang_rt.asan_dynamic-i386.lib"
+// ASAN-MD: "{{[^"]*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib"
 // ASAN-MD: "-include:___asan_seh_interceptor"
-// ASAN-MD: "-wholearchive:{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib"
+// ASAN-MD: "-wholearchive:{{.*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib"
 // ASAN-MD: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl /LD -fuse-ld=link -### /Tc%s 2>&1 | FileCheck --check-prefix=DLL %s
@@ -40,7 +40,7 @@
 // ASAN-DLL: "-dll"
 // ASAN-DLL: "-debug"
 // ASAN-DLL: "-incremental:no"
-// ASAN-DLL: "{{.*}}clang_rt.asan_dll_thunk.lib"
+// ASAN-DLL: "{{.*}}clang_rt.asan_dll_thunk-i386.lib"
 // ASAN-DLL: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl /Zi /Tc%s -fuse-ld=link -### 2>&1 | FileCheck --check-prefix=DEBUG %s
diff --git a/clang/test/Driver/compiler-rt-unwind.c b/clang/test/Driver/compiler-rt-unwind.c
index c5040d7fd900b..7f4e3f22ab19a 100644
--- a/clang/test/Driver/compiler-rt-unwind.c
+++ b/clang/test/Driver/compiler-rt-unwind.c
@@ -98,14 +98,14 @@
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:     -shared-libgcc \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
+// MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
 // MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT-SAME: "-l:libunwind.dll.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:     -static-libgcc \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
+// MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
 // MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT-SAME: "-l:libunwind.a"
 //
 // RUN: %clang -### %s 2>&1 \
@@ -114,5 +114,5 @@
 // RUN: %clangxx -### %s 2>&1 \
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
+// MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
 // MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT-SAME: "-lunwind"
diff --git a/clang/test/Driver/coverage-ld.c b/clang/test/Driver/coverage-ld.c
index be1d8320ab8be..acb08eb5db59a 100644
--- a/clang/test/Driver/coverage-ld.c
+++ b/clang/test/Driver/coverage-ld.c
@@ -33,7 +33,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64 %s
 //
 // CHECK-FREEBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-netbsd --coverage -fuse-ld=ld \
@@ -42,7 +42,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64 %s
 
 // CHECK-NETBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-openbsd --coverage -fuse-ld=ld \
@@ -51,7 +51,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64 %s
 
 // CHECK-OPENBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi --coverage -fuse-ld=ld \
@@ -60,4 +60,4 @@
 // RUN:   | FileCheck --check-prefix=CHECK-ANDROID-ARM %s
 //
 // CHECK-ANDROID-ARM: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-ANDROID-ARM: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}arm-unknown-linux-android{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-ANDROID-ARM: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}linux{{/|\\\\}}libclang_rt.profile-arm-android.a"
diff --git a/clang/test/Driver/instrprof-ld.c b/clang/test/Driver/instrprof-ld.c
index a96bba4a1e763..674580b349d42 100644
--- a/clang/test/Driver/instrprof-ld.c
+++ b/clang/test/Driver/instrprof-ld.c
@@ -34,7 +34,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64 %s
 //
 // CHECK-FREEBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-netbsd -fprofile-instr-generate -fuse-ld=ld \
@@ -43,7 +43,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64 %s
 
 // CHECK-NETBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-openbsd -fprofile-instr-generate -fuse-ld=ld \
@@ -52,7 +52,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64 %s
 
 // CHECK-OPENBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -72,7 +72,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-LINUX-X86-64-SHARED %s
 //
 // CHECK-LINUX-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-LINUX-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{.*}}x86_64-unknown-linux{{.*}}libclang_rt.profile.a" {{.*}} "-lc"
+// CHECK-LINUX-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{.*}}linux{{.*}}libclang_rt.profile.a" {{.*}} "-lc"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -82,7 +82,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64-SHARED %s
 //
 // CHECK-FREEBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-FREEBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -92,7 +92,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64-SHARED %s
 
 // CHECK-NETBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-NETBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -102,7 +102,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64-SHARED %s
 
 // CHECK-OPENBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-OPENBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-apple-darwin14 -fprofile-instr-generate -fuse-ld=ld \
@@ -174,7 +174,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-MINGW-X86-64 %s
 //
 // CHECK-MINGW-X86-64: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-MINGW-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-windows-gnu{{/|\\\\}}libclang_rt.profile.a"
+// CHECK-MINGW-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}windows{{/|\\\\}}libclang_rt.profile-x86_64.a"
 
 // Test instrumented profiling dependent-lib flags
 //
diff --git a/clang/test/Driver/linux-ld.c b/clang/test/Driver/linux-ld.c
index e5c5563673858..4020b138dc8fd 100644
--- a/clang/test/Driver/linux-ld.c
+++ b/clang/test/Driver/linux-ld.c
@@ -99,9 +99,9 @@
 // CHECK-LD-RT-ANDROID: "--eh-frame-hdr"
 // CHECK-LD-RT-ANDROID: "-m" "armelf_linux_eabi"
 // CHECK-LD-RT-ANDROID: "-dynamic-linker"
-// CHECK-LD-RT-ANDROID: libclang_rt.builtins.a"
+// CHECK-LD-RT-ANDROID: libclang_rt.builtins-arm-android.a"
 // CHECK-LD-RT-ANDROID: "-lc"
-// CHECK-LD-RT-ANDROID: libclang_rt.builtins.a"
+// CHECK-LD-RT-ANDROID: libclang_rt.builtins-arm-android.a"
 //
 // RUN: %clang -### %s -no-pie 2>&1 \
 // RUN:     --target=x86_64-unknown-linux -rtlib=platform --unwindlib=platform \
@@ -264,7 +264,7 @@
 // RUN:     --sysroot=%S/Inputs/basic_linux_tree \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANG-ANDROID-STATIC %s
 // CHECK-CLANG-ANDROID-STATIC: "{{.*}}ld{{(.exe)?}}" "--sysroot=[[SYSROOT:[^"]+]]"
-// CHECK-CLANG-ANDROID-STATIC: "--start-group" "{{[^"]*}}{{/|\\\\}}libclang_rt.builtins.a" "-l:libunwind.a" "-lc" "--end-group"
+// CHECK-CLANG-ANDROID-STATIC: "--start-group" "{{[^"]*}}{{/|\\\\}}libclang_rt.builtins-aarch64-android.a" "-l:libunwind.a" "-lc" "--end-group"
 //
 // RUN: %clang -### %s 2>&1      \
 // RUN:     --target=x86_64-unknown-linux -rtlib=platform --unwindlib=platform \
diff --git a/clang/test/Driver/mingw-sanitizers.c b/clang/test/Driver/mingw-sanitizers.c
index 2325f8f0f1f23..d165648a8fdf6 100644
--- a/clang/test/Driver/mingw-sanitizers.c
+++ b/clang/test/Driver/mingw-sanitizers.c
@@ -4,17 +4,17 @@
 //
 // ASAN-ALL-NOT:"-l{{[^"]+"]}}"
 // ASAN-ALL-NOT:"[[INPUT]]"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic-i386.dll.a"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic-x86_64.dll.a"
 // ASAN-ALL:    "-lcomponent"
 // ASAN-ALL:    "[[INPUT]]"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic-i386.dll.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-i386.a"
 // ASAN-I686:   "--require-defined" "___asan_seh_interceptor"
-// ASAN-I686:   "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a" "--no-whole-archive"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a"
+// ASAN-I686:   "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-i386.a" "--no-whole-archive"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic-x86_64.dll.a"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-x86_64.a"
 // ASAN-X86_64: "--require-defined" "__asan_seh_interceptor"
-// ASAN-X86_64: "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a" "--no-whole-archive"
+// ASAN-X86_64: "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-x86_64.a" "--no-whole-archive"
 
 // RUN: %clang -target x86_64-windows-gnu %s -### -fsanitize=vptr
diff --git a/clang/test/Driver/msp430-toolchain.c b/clang/test/Driver/msp430-toolchain.c
index 3c3042b482ef2..ef6780c38f2ee 100644
--- a/clang/test/Driver/msp430-toolchain.c
+++ b/clang/test/Driver/msp430-toolchain.c
@@ -103,8 +103,8 @@
 // LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtbegin_no_eh.o"
 // LIBS-COMPILER-RT-POS: "-L{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430"
 // LIBS-COMPILER-RT-POS: "-L{{.*}}/Inputs/basic_msp430_tree{{/|\\\\}}msp430-elf{{/|\\\\}}lib/430"
-// LIBS-COMPILER-RT-POS: "{{[^"]*}}libclang_rt.builtins.a" "--start-group" "-lmul_none" "-lc" "{{[^"]*}}libclang_rt.builtins.a" "-lcrt" "-lnosys" "--end-group" "{{[^"]*}}libclang_rt.builtins.a"
-// LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtend_no_eh.o" "{{[^"]*}}libclang_rt.builtins.a"
+// LIBS-COMPILER-RT-POS: "{{[^"]*}}libclang_rt.builtins-msp430.a" "--start-group" "-lmul_none" "-lc" "{{[^"]*}}libclang_rt.builtins-msp430.a" "-lcrt" "-lnosys" "--end-group" "{{[^"]*}}libclang_rt.builtins-msp430.a"
+// LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtend_no_eh.o" "{{[^"]*}}libclang_rt.builtins-msp430.a"
 // LIBS-COMPILER-RT-NEG-NOT: crtbegin.o
 // LIBS-COMPILER-RT-NEG-NOT: -lssp_nonshared
 // LIBS-COMPILER-RT-NEG-NOT: -lssp
diff --git a/clang/test/Driver/print-libgcc-file-name-clangrt.c b/clang/test/Driver/print-libgcc-file-name-clangrt.c
index a902eedc85209..ed740e0d2917d 100644
--- a/clang/test/Driver/print-libgcc-file-name-clangrt.c
+++ b/clang/test/Driver/print-libgcc-file-name-clangrt.c
@@ -5,14 +5,14 @@
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-X8664 %s
-// CHECK-CLANGRT-X8664: libclang_rt.builtins.a
+// CHECK-CLANGRT-X8664: libclang_rt.builtins-x86_64.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=i386-pc-linux \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-I386 %s
-// CHECK-CLANGRT-I386: libclang_rt.builtins.a
+// CHECK-CLANGRT-I386: libclang_rt.builtins-i386.a
 
 // Check whether alternate arch values map to the correct library.
 //
@@ -27,28 +27,28 @@
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM %s
-// CHECK-CLANGRT-ARM: libclang_rt.builtins.a
+// CHECK-CLANGRT-ARM: libclang_rt.builtins-arm.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM-ANDROID %s
-// CHECK-CLANGRT-ARM-ANDROID: libclang_rt.builtins.a
+// CHECK-CLANGRT-ARM-ANDROID: libclang_rt.builtins-arm-android.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARMHF %s
-// CHECK-CLANGRT-ARMHF: libclang_rt.builtins.a
+// CHECK-CLANGRT-ARMHF: libclang_rt.builtins-armhf.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-gnueabi -mfloat-abi=hard \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM-ABI %s
-// CHECK-CLANGRT-ARM-ABI: libclang_rt.builtins.a
+// CHECK-CLANGRT-ARM-ABI: libclang_rt.builtins-armhf.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=armv7m-none-eabi \
diff --git a/clang/test/Driver/print-runtime-dir.c b/clang/test/Driver/print-runtime-dir.c
index ac1ff7e634b81..550ffef1aaf6e 100644
--- a/clang/test/Driver/print-runtime-dir.c
+++ b/clang/test/Driver/print-runtime-dir.c
@@ -1,3 +1,9 @@
+// Default directory layout
+// RUN: %clang -print-runtime-dir --target=x86_64-pc-windows-msvc \
+// RUN:   -resource-dir=%S/Inputs/resource_dir \
+// RUN:   | FileCheck --check-prefix=PRINT-RUNTIME-DIR -DFILE=%S/Inputs/resource_dir %s
+// PRINT-RUNTIME-DIR: [[FILE]]{{/|\\}}lib{{/|\\}}windows
+
 // Per-target directory layout
 // RUN: %clang -print-runtime-dir --target=x86_64-pc-windows-msvc \
 // RUN:   -resource-dir=%S/Inputs/resource_dir_with_per_target_subdir \
diff --git a/clang/test/Driver/riscv32-toolchain-extra.c b/clang/test/Driver/riscv32-toolchain-extra.c
index aab6b36f3cfca..2d38aa3b545fe 100644
--- a/clang/test/Driver/riscv32-toolchain-extra.c
+++ b/clang/test/Driver/riscv32-toolchain-extra.c
@@ -29,8 +29,8 @@
 // C-RV32-BAREMETAL-ILP32-NOGCC: "-internal-isystem" "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/include"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/riscv32-unknown-elf-ld"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/lib/crt0.o"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/clang_rt.crtbegin.o"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/clang_rt.crtbegin-riscv32.o"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/lib"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "--start-group" "-lc" "-lgloss" "--end-group"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/libclang_rt.builtins.a"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/clang_rt.crtend.o"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/libclang_rt.builtins-riscv32.a"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/clang_rt.crtend-riscv32.o"
diff --git a/clang/test/Driver/riscv32-toolchain.c b/clang/test/Driver/riscv32-toolchain.c
index 322a6ca2840fb..bb2533cdf1bce 100644
--- a/clang/test/Driver/riscv32-toolchain.c
+++ b/clang/test/Driver/riscv32-toolchain.c
@@ -195,9 +195,9 @@
 // RUN:   --target=riscv32-unknown-elf --rtlib=compiler-rt --unwindlib=compiler-rt 2>&1 \
 // RUN:   | FileCheck -check-prefix=C-RV32-RTLIB-COMPILERRT-ILP32 %s
 // C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}crt0.o"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtbegin.o"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins.a"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtend.o"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtbegin-riscv32.o"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins-riscv32.a"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtend-riscv32.o"
 
 // RUN: %clang -### %s --target=riscv32 \
 // RUN:   --gcc-toolchain=%S/Inputs/basic_riscv32_tree --sysroot= \
diff --git a/clang/test/Driver/riscv64-toolchain-extra.c b/clang/test/Driver/riscv64-toolchain-extra.c
index d8d9b58441676..a6ec9b16cc5ca 100644
--- a/clang/test/Driver/riscv64-toolchain-extra.c
+++ b/clang/test/Driver/riscv64-toolchain-extra.c
@@ -29,8 +29,8 @@
 // C-RV64-BAREMETAL-LP64-NOGCC: "-internal-isystem" "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/include"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/riscv64-unknown-elf-ld"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/lib/crt0.o"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/clang_rt.crtbegin.o"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/clang_rt.crtbegin-riscv64.o"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/lib"
 // C-RV64-BAREMETAL-LP64-NOGCC: "--start-group" "-lc" "-lgloss" "--end-group"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/libclang_rt.builtins.a"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/clang_rt.crtend.o"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/libclang_rt.builtins-riscv64.a"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/clang_rt.crtend-riscv64.o"
diff --git a/clang/test/Driver/riscv64-toolchain.c b/clang/test/Driver/riscv64-toolchain.c
index b3216de307540..381ee58c470c5 100644
--- a/clang/test/Driver/riscv64-toolchain.c
+++ b/clang/test/Driver/riscv64-toolchain.c
@@ -151,9 +151,9 @@
 // RUN:   --target=riscv64-unknown-elf --rtlib=compiler-rt --unwindlib=compiler-rt 2>&1 \
 // RUN:   | FileCheck -check-prefix=C-RV64-RTLIB-COMPILERRT-LP64 %s
 // C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}crt0.o"
-// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtbegin.o"
-// C-RV64-RTLIB-COMPILERRT-LP64: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins.a"
-// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtend.o"
+// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtbegin-riscv64.o"
+// C-RV64-RTLIB-COMPILERRT-LP64: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins-riscv64.a"
+// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtend-riscv64.o"
 
 // RUN: %clang -### %s --target=riscv64 \
 // RUN:   --gcc-toolchain=%S/Inputs/basic_riscv64_tree --sysroot= \
diff --git a/clang/test/Driver/sanitizer-ld.c b/clang/test/Driver/sanitizer-ld.c
index 1d52fc1260959..53e536d772924 100644
--- a/clang/test/Driver/sanitizer-ld.c
+++ b/clang/test/Driver/sanitizer-ld.c
@@ -111,7 +111,7 @@
 // CHECK-ASAN-FREEBSD: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-FREEBSD-NOT: "-lc"
 // CHECK-ASAN-FREEBSD-NOT: libclang_rt.asan_cxx
-// CHECK-ASAN-FREEBSD: freebsd{{/|\\+}}libclang_rt.asan.a"
+// CHECK-ASAN-FREEBSD: freebsd{{/|\\+}}libclang_rt.asan-i386.a"
 // CHECK-ASAN-FREEBSD-NOT: libclang_rt.asan_cxx
 // CHECK-ASAN-FREEBSD-NOT: "--dynamic-list"
 // CHECK-ASAN-FREEBSD: "--export-dynamic"
@@ -135,8 +135,8 @@
 //
 // CHECK-ASAN-LINUX-CXX: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-LINUX-CXX-NOT: "-lc"
-// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan.a" "--no-whole-archive"
-// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan_cxx.a" "--no-whole-archive"
+// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan-i386.a" "--no-whole-archive"
+// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan_cxx-i386.a" "--no-whole-archive"
 // CHECK-ASAN-LINUX-CXX-NOT: "--dynamic-list"
 // CHECK-ASAN-LINUX-CXX: "--export-dynamic"
 // CHECK-ASAN-LINUX-CXX: stdc++
@@ -163,7 +163,7 @@
 //
 // CHECK-ASAN-ARM: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-ARM-NOT: "-lc"
-// CHECK-ASAN-ARM: libclang_rt.asan.a"
+// CHECK-ASAN-ARM: libclang_rt.asan-arm.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=armv7l-linux-gnueabi -fuse-ld=ld -fsanitize=address \
@@ -172,7 +172,7 @@
 //
 // CHECK-ASAN-ARMv7: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-ARMv7-NOT: "-lc"
-// CHECK-ASAN-ARMv7: libclang_rt.asan.a"
+// CHECK-ASAN-ARMv7: libclang_rt.asan-arm.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=address \
@@ -184,7 +184,7 @@
 // CHECK-ASAN-ANDROID-NOT: "-lc"
 // CHECK-ASAN-ANDROID-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-NOT: "-lresolv"
-// CHECK-ASAN-ANDROID: libclang_rt.asan.so"
+// CHECK-ASAN-ANDROID: libclang_rt.asan-arm-android.so"
 // CHECK-ASAN-ANDROID-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-NOT: "-lresolv"
 
@@ -195,7 +195,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID-STATICLIBASAN %s
 //
 // CHECK-ASAN-ANDROID-STATICLIBASAN: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-ASAN-ANDROID-STATICLIBASAN: libclang_rt.asan.a"
+// CHECK-ASAN-ANDROID-STATICLIBASAN: libclang_rt.asan-arm-android.a"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lrt"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lresolv"
@@ -210,7 +210,7 @@
 // CHECK-UBSAN-ANDROID-NOT: "-lc"
 // CHECK-UBSAN-ANDROID-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-NOT: "-lresolv"
-// CHECK-UBSAN-ANDROID: libclang_rt.ubsan_standalone.so"
+// CHECK-UBSAN-ANDROID: libclang_rt.ubsan_standalone-arm-android.so"
 // CHECK-UBSAN-ANDROID-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-NOT: "-lresolv"
 
@@ -221,7 +221,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-UBSAN-ANDROID-STATICLIBASAN %s
 //
 // CHECK-UBSAN-ANDROID-STATICLIBASAN: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-UBSAN-ANDROID-STATICLIBASAN: libclang_rt.ubsan_standalone.a"
+// CHECK-UBSAN-ANDROID-STATICLIBASAN: libclang_rt.ubsan_standalone-arm-android.a"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lrt"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lresolv"
@@ -237,7 +237,7 @@
 // CHECK-ASAN-ANDROID-X86-NOT: "-lc"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lresolv"
-// CHECK-ASAN-ANDROID-X86: libclang_rt.asan.so"
+// CHECK-ASAN-ANDROID-X86: libclang_rt.asan-i686-android.so"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lresolv"
 //
@@ -257,7 +257,7 @@
 //
 // CHECK-ASAN-ANDROID-SHARED: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lc"
-// CHECK-ASAN-ANDROID-SHARED: libclang_rt.asan.so"
+// CHECK-ASAN-ANDROID-SHARED: libclang_rt.asan-arm-android.so"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lresolv"
 
@@ -347,7 +347,7 @@
 // CHECK-UBSAN-LINUX: "{{.*}}ld{{(.exe)?}}"
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.asan
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.ubsan_standalone_cxx
-// CHECK-UBSAN-LINUX: "--whole-archive" "{{.*}}libclang_rt.ubsan_standalone.a" "--no-whole-archive"
+// CHECK-UBSAN-LINUX: "--whole-archive" "{{.*}}libclang_rt.ubsan_standalone-x32.a" "--no-whole-archive"
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.asan
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.ubsan_standalone_cxx
 // CHECK-UBSAN-LINUX-NOT: "-lstdc++"
@@ -678,7 +678,7 @@
 // RUN:     --sysroot=%S/Inputs/basic_android_tree \
 // RUN:   | FileCheck --check-prefix=CHECK-CFI-CROSS-DSO-DIAG-ANDROID %s
 // CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{.*}}ld{{(.exe)?}}"
-// CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{[^"]*}}libclang_rt.ubsan_standalone.so"
+// CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{[^"]*}}libclang_rt.ubsan_standalone-aarch64-android.so"
 // CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "--export-dynamic-symbol=__cfi_check"
 
 // RUN: %clangxx -fsanitize=address -### %s 2>&1 \
@@ -929,7 +929,7 @@
 // CHECK-SCUDO-ANDROID: "-pie"
 // CHECK-SCUDO-ANDROID-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-NOT: "-lresolv"
-// CHECK-SCUDO-ANDROID: libclang_rt.scudo_standalone.so"
+// CHECK-SCUDO-ANDROID: libclang_rt.scudo_standalone-arm-android.so"
 // CHECK-SCUDO-ANDROID-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-NOT: "-lresolv"
 
@@ -940,7 +940,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-SCUDO-ANDROID-STATIC %s
 // CHECK-SCUDO-ANDROID-STATIC: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
 // CHECK-SCUDO-ANDROID-STATIC: "-pie"
-// CHECK-SCUDO-ANDROID-STATIC: "--whole-archive" "{{.*}}libclang_rt.scudo_standalone.a" "--no-whole-archive"
+// CHECK-SCUDO-ANDROID-STATIC: "--whole-archive" "{{.*}}libclang_rt.scudo_standalone-arm-android.a" "--no-whole-archive"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lstdc++"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lrt"
diff --git a/clang/test/Driver/wasm-toolchain.c b/clang/test/Driver/wasm-toolchain.c
index dabf0ac2433bb..88590a3ba4c45 100644
--- a/clang/test/Driver/wasm-toolchain.c
+++ b/clang/test/Driver/wasm-toolchain.c
@@ -17,42 +17,42 @@
 // RUN: %clang -### --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK %s
 // LINK: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C link command-line with optimization with unknown OS.
 
 // RUN: %clang -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT %s
 // LINK_OPT: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C link command-line with known OS.
 
 // RUN: %clang -### --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN %s
 // LINK_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // -shared should be passed through to `wasm-ld` and include crt1-reactor.o with a known OS.
 
 // RUN: %clang -### -shared -mexec-model=reactor --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN_SHARED %s
 // LINK_KNOWN_SHARED: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN_SHARED: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_KNOWN_SHARED: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // -shared should be passed through to `wasm-ld` and include crt1-reactor.o with an unknown OS.
 
 // RUN: %clang -### -shared -mexec-model=reactor --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_UNKNOWN_SHARED %s
 // LINK_UNKNOWN_SHARED: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_UNKNOWN_SHARED: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_UNKNOWN_SHARED: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C link command-line with optimization with known OS.
 
 // RUN: %clang -### -O2 --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN %s
 // LINK_OPT_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C compile command-line with known OS.
 
@@ -180,12 +180,12 @@
 // RUN: %clang -### %s --target=wasm32-unknown-unknown --sysroot=%s/no-sysroot-there -mexec-model=command 2>&1 \
 // RUN:   | FileCheck -check-prefix=CHECK-COMMAND %s
 // CHECK-COMMAND: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// CHECK-COMMAND: wasm-ld{{.*}}" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// CHECK-COMMAND: wasm-ld{{.*}}" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // RUN: %clang -### %s --target=wasm32-unknown-unknown --sysroot=%s/no-sysroot-there -mexec-model=reactor 2>&1 \
 // RUN:   | FileCheck -check-prefix=CHECK-REACTOR %s
 // CHECK-REACTOR: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// CHECK-REACTOR: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// CHECK-REACTOR: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // -fPIC implies +mutable-globals
 
@@ -204,7 +204,7 @@
 // RUN: %clang -### -O2 --target=wasm32-wasip2 %s --sysroot /foo 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_WASIP2 %s
 // LINK_WASIP2: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_WASIP2: wasm-component-ld{{.*}}" "-L/foo/lib/wasm32-wasip2" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_WASIP2: wasm-component-ld{{.*}}" "-L/foo/lib/wasm32-wasip2" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // Test that on `wasm32-wasip2` the `wasm-component-ld` programs is told where
 // to find `wasm-ld` by default.
diff --git a/clang/test/Driver/wasm-toolchain.cpp b/clang/test/Driver/wasm-toolchain.cpp
index ba1c55b33edca..4af011097021f 100644
--- a/clang/test/Driver/wasm-toolchain.cpp
+++ b/clang/test/Driver/wasm-toolchain.cpp
@@ -17,48 +17,48 @@
 // RUN: %clangxx -### --target=wasm32-unknown-unknown --sysroot=/foo --stdlib=libc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK %s
 // LINK: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // RUN: %clangxx -### --target=wasm32-unknown-unknown --sysroot=/foo --stdlib=libstdc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_STDCXX %s
 // LINK_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C++ link command-line with optimization with unknown OS.
 
 // RUN: %clangxx -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s --stdlib=libc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT %s
 // LINK_OPT: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // RUN: %clangxx -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s --stdlib=libstdc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_STDCXX %s
 // LINK_OPT_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C++ link command-line with known OS.
 
 // RUN: %clangxx -### --target=wasm32-wasi --sysroot=/foo --stdlib=libc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN %s
 // LINK_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // RUN: %clangxx -### --target=wasm32-wasi --sysroot=/foo --stdlib=libstdc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN_STDCXX %s
 // LINK_KNOWN_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C++ link command-line with optimization with known OS.
 
 // RUN: %clangxx -### -O2 --target=wasm32-wasi --sysroot=/foo %s --stdlib=libc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN %s
 // LINK_OPT_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // RUN: %clangxx -### -O2 --target=wasm32-wasi --sysroot=/foo %s --stdlib=libstdc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN_STDCXX %s
 // LINK_OPT_KNOWN_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
+// LINK_OPT_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
 
 // A basic C++ compile command-line with known OS.
 
diff --git a/clang/test/Driver/windows-cross.c b/clang/test/Driver/windows-cross.c
index f6e831f00e13a..75490b992d78d 100644
--- a/clang/test/Driver/windows-cross.c
+++ b/clang/test/Driver/windows-cross.c
@@ -11,32 +11,32 @@
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %s/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libstdc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-RTLIB
 
-// CHECK-RTLIB: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-RTLIB: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-C-LIBCXX
 
-// CHECK-C-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-C-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clangxx -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-LIBCXX
 
-// CHECK-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lc++" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lc++" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-SHARED
 
-// CHECK-SHARED: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-SHARED: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -static -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-SHARED-STATIC
 
-// CHECK-SHARED-STATIC: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bstatic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-SHARED-STATIC: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bstatic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %s/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -nostartfiles -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-NOSTARTFILES
 
-// CHECK-NOSTARTFILES: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
+// CHECK-NOSTARTFILES: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -nostartfiles -nodefaultlibs -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-STANDALONE
@@ -52,19 +52,19 @@
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS
 
 // CHECK-SANITIZE-ADDRESS: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS: "{{.*}}clang_rt.asan_dll_thunk.lib"
+// CHECK-SANITIZE-ADDRESS: "{{.*}}clang_rt.asan_dll_thunk-arm.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -o test.exe -fsanitize=address -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS-EXE
 
 // CHECK-SANITIZE-ADDRESS-EXE: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS-EXE: "{{.*}}clang_rt.asan_dynamic.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib" "--undefined" "__asan_seh_interceptor"
+// CHECK-SANITIZE-ADDRESS-EXE: "{{.*}}clang_rt.asan_dynamic-arm.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk-arm.lib" "--undefined" "__asan_seh_interceptor"
 
 // RUN: %clang -### -target i686-windows-itanium -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -o test.exe -fsanitize=address -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS-EXE-X86
 
 // CHECK-SANITIZE-ADDRESS-EXE-X86: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS-EXE-X86: "{{.*}}clang_rt.asan_dynamic.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib" "--undefined" "___asan_seh_interceptor"
+// CHECK-SANITIZE-ADDRESS-EXE-X86: "{{.*}}clang_rt.asan_dynamic-i386.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib" "--undefined" "___asan_seh_interceptor"
 
 // RUN: not %clang -### --target=armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -shared -o shared.dll -fsanitize=tsan -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-TSAN
diff --git a/clang/test/Driver/zos-ld.c b/clang/test/Driver/zos-ld.c
index 87d169936e129..4d4decdd0e65b 100644
--- a/clang/test/Driver/zos-ld.c
+++ b/clang/test/Driver/zos-ld.c
@@ -14,7 +14,7 @@
 // C-LD-SAME: "-S" "//'SYS1.CSSLIB'"
 // C-LD-SAME: "//'CEE.SCEELIB(CELQS001)'"
 // C-LD-SAME: "//'CEE.SCEELIB(CELQS003)'"
-// C-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// C-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
 
 // 2. General C link for dll
 // RUN: %clang -### --shared --target=s390x-ibm-zos %s 2>&1 \
@@ -30,7 +30,7 @@
 // C-LD-DLL-SAME: "-S" "//'SYS1.CSSLIB'"
 // C-LD-DLL-SAME: "//'CEE.SCEELIB(CELQS001)'"
 // C-LD-DLL-SAME: "//'CEE.SCEELIB(CELQS003)'"
-// C-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// C-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
 
 // 3. General C++ link for executable
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -52,7 +52,7 @@
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQCXA)'"
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQXLA)'"
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQUNW)'"
-// CXX-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// CXX-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
 
 // 4. General C++ link for dll
 // RUN: %clangxx -### --shared --target=s390x-ibm-zos %s 2>&1 \
@@ -74,7 +74,7 @@
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQCXA)'"
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQXLA)'"
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQUNW)'"
-// CXX-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// CXX-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
 
 // 5. C++ link for executable w/ -mzos-hlq-le=, -mzos-hlq-csslib=
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -97,7 +97,7 @@
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQCXA)'"
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQXLA)'"
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQUNW)'"
-// CXX-LD5-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// CXX-LD5-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
 
 // 6. C++ link for executable w/ -mzos-hlq-clang=
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -120,4 +120,4 @@
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQCXA)'"
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQXLA)'"
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQUNW)'"
-// CXX-LD6-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
+// CXX-LD6-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
diff --git a/flang/test/Driver/msvc-dependent-lib-flags.f90 b/flang/test/Driver/msvc-dependent-lib-flags.f90
index 643dbe9e949cb..7c1f962e339f9 100644
--- a/flang/test/Driver/msvc-dependent-lib-flags.f90
+++ b/flang/test/Driver/msvc-dependent-lib-flags.f90
@@ -4,7 +4,7 @@
 ! RUN: %flang -### --target=aarch64-windows-msvc -fms-runtime-lib=dll_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL-DEBUG
 
 ! MSVC: -fc1
-! MSVC-SAME: --dependent-lib=clang_rt.builtins.lib
+! MSVC-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
 ! MSVC-SAME: -D_MT
 ! MSVC-SAME: --dependent-lib=libcmt
 ! MSVC-SAME: --dependent-lib=Fortran_main.static.lib
@@ -12,7 +12,7 @@
 ! MSVC-SAME: --dependent-lib=FortranDecimal.static.lib
 
 ! MSVC-DEBUG: -fc1
-! MSVC-DEBUG-SAME: --dependent-lib=clang_rt.builtins.lib
+! MSVC-DEBUG-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
 ! MSVC-DEBUG-SAME: -D_MT
 ! MSVC-DEBUG-SAME: -D_DEBUG
 ! MSVC-DEBUG-SAME: --dependent-lib=libcmtd
@@ -21,7 +21,7 @@
 ! MSVC-DEBUG-SAME: --dependent-lib=FortranDecimal.static_dbg.lib
 
 ! MSVC-DLL: -fc1
-! MSVC-DLL-SAME: --dependent-lib=clang_rt.builtins.lib
+! MSVC-DLL-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
 ! MSVC-DLL-SAME: -D_MT
 ! MSVC-DLL-SAME: -D_DLL
 ! MSVC-DLL-SAME: --dependent-lib=msvcrt
@@ -30,7 +30,7 @@
 ! MSVC-DLL-SAME: --dependent-lib=FortranDecimal.dynamic.lib
 
 ! MSVC-DLL-DEBUG: -fc1
-! MSVC-DLL-DEBUG-SAME: --dependent-lib=clang_rt.builtins.lib
+! MSVC-DLL-DEBUG-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
 ! MSVC-DLL-DEBUG-SAME: -D_MT
 ! MSVC-DLL-DEBUG-SAME: -D_DEBUG
 ! MSVC-DLL-DEBUG-SAME: -D_DLL

From fca51911d4668b3a6b79eb956327eb81fad3f40c Mon Sep 17 00:00:00 2001
From: Bill Wendling <5993918+bwendling@users.noreply.github.com>
Date: Thu, 11 Apr 2024 00:33:40 +0000
Subject: [PATCH 085/886] [NFC][Clang] Improve const correctness for
 IdentifierInfo (#79365)

The IdentifierInfo isn't typically modified. Use 'const' wherever
possible.
---
 clang/include/clang/AST/ASTContext.h          |   4 +-
 clang/include/clang/AST/Decl.h                |  37 +++---
 clang/include/clang/AST/DeclObjC.h            |  89 +++++++-------
 clang/include/clang/AST/DeclTemplate.h        |   8 +-
 clang/include/clang/AST/ExprCXX.h             |  10 +-
 clang/include/clang/AST/ExternalASTSource.h   |   2 +-
 clang/include/clang/AST/NestedNameSpecifier.h |   6 +-
 clang/include/clang/Analysis/SelectorExtras.h |   4 +-
 clang/include/clang/Basic/IdentifierTable.h   |  25 ++--
 .../clang/Lex/ExternalPreprocessorSource.h    |   2 +-
 clang/include/clang/Lex/MacroInfo.h           |   8 +-
 clang/include/clang/Lex/Preprocessor.h        |   9 +-
 clang/include/clang/Parse/Parser.h            |  16 +--
 .../include/clang/Sema/CodeCompleteConsumer.h |   9 +-
 clang/include/clang/Sema/DeclSpec.h           |  23 ++--
 clang/include/clang/Sema/ParsedTemplate.h     |   7 +-
 clang/include/clang/Sema/Sema.h               |  89 +++++++-------
 clang/include/clang/Serialization/ASTReader.h |  10 +-
 clang/lib/ARCMigrate/ObjCMT.cpp               |   7 +-
 clang/lib/ARCMigrate/TransAPIUses.cpp         |   2 +-
 clang/lib/AST/ASTContext.cpp                  |  11 +-
 clang/lib/AST/ASTImporter.cpp                 |   6 +-
 clang/lib/AST/Decl.cpp                        |  18 +--
 clang/lib/AST/DeclObjC.cpp                    |  94 ++++++---------
 clang/lib/AST/DeclTemplate.cpp                |  14 +--
 clang/lib/AST/NSAPI.cpp                       | 104 ++++++----------
 clang/lib/AST/NestedNameSpecifier.cpp         |  18 +--
 clang/lib/AST/SelectorLocationsKind.cpp       |   4 +-
 clang/lib/AST/StmtPrinter.cpp                 |   4 +-
 clang/lib/AST/StmtProfile.cpp                 |   6 +-
 clang/lib/Analysis/ObjCNoReturn.cpp           |   5 +-
 clang/lib/Basic/IdentifierTable.cpp           |  16 +--
 clang/lib/CodeGen/CGBlocks.cpp                |   4 +-
 clang/lib/CodeGen/CGCUDANV.cpp                |   2 +-
 clang/lib/CodeGen/CGDecl.cpp                  |   4 +-
 clang/lib/CodeGen/CGObjC.cpp                  |  23 ++--
 clang/lib/CodeGen/CGObjCMac.cpp               |  13 +-
 clang/lib/CodeGen/CodeGenFunction.cpp         |   2 +-
 clang/lib/CodeGen/CodeGenModule.cpp           |   6 +-
 .../Frontend/Rewrite/RewriteModernObjC.cpp    |   2 +-
 clang/lib/Lex/HeaderSearch.cpp                |   3 +-
 clang/lib/Lex/MacroInfo.cpp                   |   2 +-
 clang/lib/Lex/PPLexerChange.cpp               |   9 +-
 clang/lib/Lex/PPMacroExpansion.cpp            |   4 +-
 clang/lib/Lex/Preprocessor.cpp                |   2 +-
 clang/lib/Parse/ParseDecl.cpp                 |   2 +-
 clang/lib/Parse/ParseDeclCXX.cpp              |   2 +-
 clang/lib/Parse/ParseExprCXX.cpp              |   5 +-
 clang/lib/Parse/ParseObjc.cpp                 |  12 +-
 clang/lib/Parse/ParseTemplate.cpp             |   4 +-
 clang/lib/Sema/CodeCompleteConsumer.cpp       |   3 +-
 clang/lib/Sema/Sema.cpp                       |   5 +-
 clang/lib/Sema/SemaCodeComplete.cpp           | 112 +++++++++---------
 clang/lib/Sema/SemaDecl.cpp                   |  21 ++--
 clang/lib/Sema/SemaDeclCXX.cpp                |   9 +-
 clang/lib/Sema/SemaDeclObjC.cpp               |  23 ++--
 clang/lib/Sema/SemaExprCXX.cpp                |  24 ++--
 clang/lib/Sema/SemaExprObjC.cpp               |  17 ++-
 clang/lib/Sema/SemaObjCProperty.cpp           |   4 +-
 clang/lib/Sema/SemaOpenMP.cpp                 |   2 +-
 clang/lib/Sema/SemaPseudoObject.cpp           |  48 ++++----
 clang/lib/Sema/SemaStmt.cpp                   |   8 +-
 clang/lib/Sema/SemaTemplate.cpp               |  35 +++---
 clang/lib/Serialization/ASTCommon.cpp         |   2 +-
 clang/lib/Serialization/ASTReader.cpp         |  14 +--
 clang/lib/Serialization/ASTWriter.cpp         |  18 +--
 .../Checkers/CheckObjCDealloc.cpp             |   4 +-
 .../Checkers/LocalizationChecker.cpp          | 108 ++++++++---------
 .../Checkers/NullabilityChecker.cpp           |   3 +-
 .../Checkers/ObjCMissingSuperCallChecker.cpp  |   2 +-
 .../Checkers/ObjCSuperDeallocChecker.cpp      |   4 +-
 clang/tools/libclang/CIndexCodeCompletion.cpp |  10 +-
 .../Plugins/ExpressionParser/Clang/ASTUtils.h |   2 +-
 .../ExpressionParser/Clang/ClangASTSource.cpp |   7 +-
 .../AppleObjCRuntime/AppleObjCDeclVendor.cpp  |   4 +-
 .../TypeSystem/Clang/TypeSystemClang.cpp      |  11 +-
 76 files changed, 602 insertions(+), 666 deletions(-)

diff --git a/clang/include/clang/AST/ASTContext.h b/clang/include/clang/AST/ASTContext.h
index 08f71051e6cbf..28f8d67811f0a 100644
--- a/clang/include/clang/AST/ASTContext.h
+++ b/clang/include/clang/AST/ASTContext.h
@@ -3411,13 +3411,13 @@ const StreamingDiagnostic &operator<<(const StreamingDiagnostic &DB,
 
 /// Utility function for constructing a nullary selector.
 inline Selector GetNullarySelector(StringRef name, ASTContext &Ctx) {
-  IdentifierInfo* II = &Ctx.Idents.get(name);
+  const IdentifierInfo *II = &Ctx.Idents.get(name);
   return Ctx.Selectors.getSelector(0, &II);
 }
 
 /// Utility function for constructing an unary selector.
 inline Selector GetUnarySelector(StringRef name, ASTContext &Ctx) {
-  IdentifierInfo* II = &Ctx.Idents.get(name);
+  const IdentifierInfo *II = &Ctx.Idents.get(name);
   return Ctx.Selectors.getSelector(1, &II);
 }
 
diff --git a/clang/include/clang/AST/Decl.h b/clang/include/clang/AST/Decl.h
index 5f1f83bb00282..ed6790acdfc7c 100644
--- a/clang/include/clang/AST/Decl.h
+++ b/clang/include/clang/AST/Decl.h
@@ -1731,7 +1731,7 @@ class ImplicitParamDecl : public VarDecl {
   static ImplicitParamDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
   ImplicitParamDecl(ASTContext &C, DeclContext *DC, SourceLocation IdLoc,
-                    IdentifierInfo *Id, QualType Type,
+                    const IdentifierInfo *Id, QualType Type,
                     ImplicitParamKind ParamKind)
       : VarDecl(ImplicitParam, C, DC, IdLoc, IdLoc, Id, Type,
                 /*TInfo=*/nullptr, SC_None) {
@@ -1765,7 +1765,7 @@ class ParmVarDecl : public VarDecl {
 
 protected:
   ParmVarDecl(Kind DK, ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-              SourceLocation IdLoc, IdentifierInfo *Id, QualType T,
+              SourceLocation IdLoc, const IdentifierInfo *Id, QualType T,
               TypeSourceInfo *TInfo, StorageClass S, Expr *DefArg)
       : VarDecl(DK, C, DC, StartLoc, IdLoc, Id, T, TInfo, S) {
     assert(ParmVarDeclBits.HasInheritedDefaultArg == false);
@@ -1777,10 +1777,10 @@ class ParmVarDecl : public VarDecl {
 
 public:
   static ParmVarDecl *Create(ASTContext &C, DeclContext *DC,
-                             SourceLocation StartLoc,
-                             SourceLocation IdLoc, IdentifierInfo *Id,
-                             QualType T, TypeSourceInfo *TInfo,
-                             StorageClass S, Expr *DefArg);
+                             SourceLocation StartLoc, SourceLocation IdLoc,
+                             const IdentifierInfo *Id, QualType T,
+                             TypeSourceInfo *TInfo, StorageClass S,
+                             Expr *DefArg);
 
   static ParmVarDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
@@ -3095,7 +3095,7 @@ class FieldDecl : public DeclaratorDecl, public Mergeable<FieldDecl> {
 
 protected:
   FieldDecl(Kind DK, DeclContext *DC, SourceLocation StartLoc,
-            SourceLocation IdLoc, IdentifierInfo *Id, QualType T,
+            SourceLocation IdLoc, const IdentifierInfo *Id, QualType T,
             TypeSourceInfo *TInfo, Expr *BW, bool Mutable,
             InClassInitStyle InitStyle)
       : DeclaratorDecl(DK, DC, IdLoc, Id, T, TInfo, StartLoc), BitField(false),
@@ -3111,7 +3111,7 @@ class FieldDecl : public DeclaratorDecl, public Mergeable<FieldDecl> {
 
   static FieldDecl *Create(const ASTContext &C, DeclContext *DC,
                            SourceLocation StartLoc, SourceLocation IdLoc,
-                           IdentifierInfo *Id, QualType T,
+                           const IdentifierInfo *Id, QualType T,
                            TypeSourceInfo *TInfo, Expr *BW, bool Mutable,
                            InClassInitStyle InitStyle);
 
@@ -3332,8 +3332,9 @@ class IndirectFieldDecl : public ValueDecl,
   friend class ASTDeclReader;
 
   static IndirectFieldDecl *Create(ASTContext &C, DeclContext *DC,
-                                   SourceLocation L, IdentifierInfo *Id,
-                                   QualType T, llvm::MutableArrayRef<NamedDecl *> CH);
+                                   SourceLocation L, const IdentifierInfo *Id,
+                                   QualType T,
+                                   llvm::MutableArrayRef<NamedDecl *> CH);
 
   static IndirectFieldDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
@@ -3381,9 +3382,9 @@ class TypeDecl : public NamedDecl {
   void anchor() override;
 
 protected:
-  TypeDecl(Kind DK, DeclContext *DC, SourceLocation L, IdentifierInfo *Id,
+  TypeDecl(Kind DK, DeclContext *DC, SourceLocation L, const IdentifierInfo *Id,
            SourceLocation StartL = SourceLocation())
-    : NamedDecl(DK, DC, L, Id), LocStart(StartL) {}
+      : NamedDecl(DK, DC, L, Id), LocStart(StartL) {}
 
 public:
   // Low-level accessor. If you just want the type defined by this node,
@@ -3425,7 +3426,7 @@ class TypedefNameDecl : public TypeDecl, public Redeclarable<TypedefNameDecl> {
 protected:
   TypedefNameDecl(Kind DK, ASTContext &C, DeclContext *DC,
                   SourceLocation StartLoc, SourceLocation IdLoc,
-                  IdentifierInfo *Id, TypeSourceInfo *TInfo)
+                  const IdentifierInfo *Id, TypeSourceInfo *TInfo)
       : TypeDecl(DK, DC, IdLoc, Id, StartLoc), redeclarable_base(C),
         MaybeModedTInfo(TInfo, 0) {}
 
@@ -3512,13 +3513,14 @@ class TypedefNameDecl : public TypeDecl, public Redeclarable<TypedefNameDecl> {
 /// type specifier.
 class TypedefDecl : public TypedefNameDecl {
   TypedefDecl(ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-              SourceLocation IdLoc, IdentifierInfo *Id, TypeSourceInfo *TInfo)
+              SourceLocation IdLoc, const IdentifierInfo *Id,
+              TypeSourceInfo *TInfo)
       : TypedefNameDecl(Typedef, C, DC, StartLoc, IdLoc, Id, TInfo) {}
 
 public:
   static TypedefDecl *Create(ASTContext &C, DeclContext *DC,
                              SourceLocation StartLoc, SourceLocation IdLoc,
-                             IdentifierInfo *Id, TypeSourceInfo *TInfo);
+                             const IdentifierInfo *Id, TypeSourceInfo *TInfo);
   static TypedefDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
   SourceRange getSourceRange() const override LLVM_READONLY;
@@ -3535,14 +3537,15 @@ class TypeAliasDecl : public TypedefNameDecl {
   TypeAliasTemplateDecl *Template;
 
   TypeAliasDecl(ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-                SourceLocation IdLoc, IdentifierInfo *Id, TypeSourceInfo *TInfo)
+                SourceLocation IdLoc, const IdentifierInfo *Id,
+                TypeSourceInfo *TInfo)
       : TypedefNameDecl(TypeAlias, C, DC, StartLoc, IdLoc, Id, TInfo),
         Template(nullptr) {}
 
 public:
   static TypeAliasDecl *Create(ASTContext &C, DeclContext *DC,
                                SourceLocation StartLoc, SourceLocation IdLoc,
-                               IdentifierInfo *Id, TypeSourceInfo *TInfo);
+                               const IdentifierInfo *Id, TypeSourceInfo *TInfo);
   static TypeAliasDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
   SourceRange getSourceRange() const override LLVM_READONLY;
diff --git a/clang/include/clang/AST/DeclObjC.h b/clang/include/clang/AST/DeclObjC.h
index f8f894b4b10d1..b8d17dd06d155 100644
--- a/clang/include/clang/AST/DeclObjC.h
+++ b/clang/include/clang/AST/DeclObjC.h
@@ -772,7 +772,7 @@ class ObjCPropertyDecl : public NamedDecl {
   // Synthesize ivar for this property
   ObjCIvarDecl *PropertyIvarDecl = nullptr;
 
-  ObjCPropertyDecl(DeclContext *DC, SourceLocation L, IdentifierInfo *Id,
+  ObjCPropertyDecl(DeclContext *DC, SourceLocation L, const IdentifierInfo *Id,
                    SourceLocation AtLocation, SourceLocation LParenLocation,
                    QualType T, TypeSourceInfo *TSI, PropertyControl propControl)
       : NamedDecl(ObjCProperty, DC, L, Id), AtLoc(AtLocation),
@@ -782,10 +782,12 @@ class ObjCPropertyDecl : public NamedDecl {
         PropertyImplementation(propControl) {}
 
 public:
-  static ObjCPropertyDecl *
-  Create(ASTContext &C, DeclContext *DC, SourceLocation L, IdentifierInfo *Id,
-         SourceLocation AtLocation, SourceLocation LParenLocation, QualType T,
-         TypeSourceInfo *TSI, PropertyControl propControl = None);
+  static ObjCPropertyDecl *Create(ASTContext &C, DeclContext *DC,
+                                  SourceLocation L, const IdentifierInfo *Id,
+                                  SourceLocation AtLocation,
+                                  SourceLocation LParenLocation, QualType T,
+                                  TypeSourceInfo *TSI,
+                                  PropertyControl propControl = None);
 
   static ObjCPropertyDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
@@ -952,7 +954,7 @@ class ObjCContainerDecl : public NamedDecl, public DeclContext {
   void anchor() override;
 
 public:
-  ObjCContainerDecl(Kind DK, DeclContext *DC, IdentifierInfo *Id,
+  ObjCContainerDecl(Kind DK, DeclContext *DC, const IdentifierInfo *Id,
                     SourceLocation nameLoc, SourceLocation atStartLoc);
 
   // Iterator access to instance/class properties.
@@ -1240,7 +1242,7 @@ class ObjCInterfaceDecl : public ObjCContainerDecl
   llvm::PointerIntPair<DefinitionData *, 1, bool> Data;
 
   ObjCInterfaceDecl(const ASTContext &C, DeclContext *DC, SourceLocation AtLoc,
-                    IdentifierInfo *Id, ObjCTypeParamList *typeParamList,
+                    const IdentifierInfo *Id, ObjCTypeParamList *typeParamList,
                     SourceLocation CLoc, ObjCInterfaceDecl *PrevDecl,
                     bool IsInternal);
 
@@ -1271,13 +1273,11 @@ class ObjCInterfaceDecl : public ObjCContainerDecl
   }
 
 public:
-  static ObjCInterfaceDecl *Create(const ASTContext &C, DeclContext *DC,
-                                   SourceLocation atLoc,
-                                   IdentifierInfo *Id,
-                                   ObjCTypeParamList *typeParamList,
-                                   ObjCInterfaceDecl *PrevDecl,
-                                   SourceLocation ClassLoc = SourceLocation(),
-                                   bool isInternal = false);
+  static ObjCInterfaceDecl *
+  Create(const ASTContext &C, DeclContext *DC, SourceLocation atLoc,
+         const IdentifierInfo *Id, ObjCTypeParamList *typeParamList,
+         ObjCInterfaceDecl *PrevDecl,
+         SourceLocation ClassLoc = SourceLocation(), bool isInternal = false);
 
   static ObjCInterfaceDecl *CreateDeserialized(const ASTContext &C, unsigned ID);
 
@@ -1338,7 +1338,8 @@ class ObjCInterfaceDecl : public ObjCContainerDecl
   ObjCImplementationDecl *getImplementation() const;
   void setImplementation(ObjCImplementationDecl *ImplD);
 
-  ObjCCategoryDecl *FindCategoryDeclaration(IdentifierInfo *CategoryId) const;
+  ObjCCategoryDecl *
+  FindCategoryDeclaration(const IdentifierInfo *CategoryId) const;
 
   // Get the local instance/class method declared in a category.
   ObjCMethodDecl *getCategoryInstanceMethod(Selector Sel) const;
@@ -1794,9 +1795,9 @@ class ObjCInterfaceDecl : public ObjCContainerDecl
     data().CategoryList = category;
   }
 
-  ObjCPropertyDecl
-    *FindPropertyVisibleInPrimaryClass(IdentifierInfo *PropertyId,
-                                       ObjCPropertyQueryKind QueryKind) const;
+  ObjCPropertyDecl *
+  FindPropertyVisibleInPrimaryClass(const IdentifierInfo *PropertyId,
+                                    ObjCPropertyQueryKind QueryKind) const;
 
   void collectPropertiesToImplement(PropertyMap &PM) const override;
 
@@ -1954,8 +1955,8 @@ class ObjCIvarDecl : public FieldDecl {
 
 private:
   ObjCIvarDecl(ObjCContainerDecl *DC, SourceLocation StartLoc,
-               SourceLocation IdLoc, IdentifierInfo *Id,
-               QualType T, TypeSourceInfo *TInfo, AccessControl ac, Expr *BW,
+               SourceLocation IdLoc, const IdentifierInfo *Id, QualType T,
+               TypeSourceInfo *TInfo, AccessControl ac, Expr *BW,
                bool synthesized)
       : FieldDecl(ObjCIvar, DC, StartLoc, IdLoc, Id, T, TInfo, BW,
                   /*Mutable=*/false, /*HasInit=*/ICIS_NoInit),
@@ -1964,10 +1965,9 @@ class ObjCIvarDecl : public FieldDecl {
 public:
   static ObjCIvarDecl *Create(ASTContext &C, ObjCContainerDecl *DC,
                               SourceLocation StartLoc, SourceLocation IdLoc,
-                              IdentifierInfo *Id, QualType T,
-                              TypeSourceInfo *TInfo,
-                              AccessControl ac, Expr *BW = nullptr,
-                              bool synthesized=false);
+                              const IdentifierInfo *Id, QualType T,
+                              TypeSourceInfo *TInfo, AccessControl ac,
+                              Expr *BW = nullptr, bool synthesized = false);
 
   static ObjCIvarDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
@@ -2343,7 +2343,7 @@ class ObjCCategoryDecl : public ObjCContainerDecl {
 
   ObjCCategoryDecl(DeclContext *DC, SourceLocation AtLoc,
                    SourceLocation ClassNameLoc, SourceLocation CategoryNameLoc,
-                   IdentifierInfo *Id, ObjCInterfaceDecl *IDecl,
+                   const IdentifierInfo *Id, ObjCInterfaceDecl *IDecl,
                    ObjCTypeParamList *typeParamList,
                    SourceLocation IvarLBraceLoc = SourceLocation(),
                    SourceLocation IvarRBraceLoc = SourceLocation());
@@ -2354,15 +2354,13 @@ class ObjCCategoryDecl : public ObjCContainerDecl {
   friend class ASTDeclReader;
   friend class ASTDeclWriter;
 
-  static ObjCCategoryDecl *Create(ASTContext &C, DeclContext *DC,
-                                  SourceLocation AtLoc,
-                                  SourceLocation ClassNameLoc,
-                                  SourceLocation CategoryNameLoc,
-                                  IdentifierInfo *Id,
-                                  ObjCInterfaceDecl *IDecl,
-                                  ObjCTypeParamList *typeParamList,
-                                  SourceLocation IvarLBraceLoc=SourceLocation(),
-                                  SourceLocation IvarRBraceLoc=SourceLocation());
+  static ObjCCategoryDecl *
+  Create(ASTContext &C, DeclContext *DC, SourceLocation AtLoc,
+         SourceLocation ClassNameLoc, SourceLocation CategoryNameLoc,
+         const IdentifierInfo *Id, ObjCInterfaceDecl *IDecl,
+         ObjCTypeParamList *typeParamList,
+         SourceLocation IvarLBraceLoc = SourceLocation(),
+         SourceLocation IvarRBraceLoc = SourceLocation());
   static ObjCCategoryDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
   ObjCInterfaceDecl *getClassInterface() { return ClassInterface; }
@@ -2472,10 +2470,9 @@ class ObjCImplDecl : public ObjCContainerDecl {
   void anchor() override;
 
 protected:
-  ObjCImplDecl(Kind DK, DeclContext *DC,
-               ObjCInterfaceDecl *classInterface,
-               IdentifierInfo *Id,
-               SourceLocation nameLoc, SourceLocation atStartLoc)
+  ObjCImplDecl(Kind DK, DeclContext *DC, ObjCInterfaceDecl *classInterface,
+               const IdentifierInfo *Id, SourceLocation nameLoc,
+               SourceLocation atStartLoc)
       : ObjCContainerDecl(DK, DC, Id, nameLoc, atStartLoc),
         ClassInterface(classInterface) {}
 
@@ -2543,12 +2540,12 @@ class ObjCCategoryImplDecl : public ObjCImplDecl {
   // Category name location
   SourceLocation CategoryNameLoc;
 
-  ObjCCategoryImplDecl(DeclContext *DC, IdentifierInfo *Id,
+  ObjCCategoryImplDecl(DeclContext *DC, const IdentifierInfo *Id,
                        ObjCInterfaceDecl *classInterface,
                        SourceLocation nameLoc, SourceLocation atStartLoc,
                        SourceLocation CategoryNameLoc)
-      : ObjCImplDecl(ObjCCategoryImpl, DC, classInterface, Id,
-                     nameLoc, atStartLoc),
+      : ObjCImplDecl(ObjCCategoryImpl, DC, classInterface, Id, nameLoc,
+                     atStartLoc),
         CategoryNameLoc(CategoryNameLoc) {}
 
   void anchor() override;
@@ -2557,12 +2554,10 @@ class ObjCCategoryImplDecl : public ObjCImplDecl {
   friend class ASTDeclReader;
   friend class ASTDeclWriter;
 
-  static ObjCCategoryImplDecl *Create(ASTContext &C, DeclContext *DC,
-                                      IdentifierInfo *Id,
-                                      ObjCInterfaceDecl *classInterface,
-                                      SourceLocation nameLoc,
-                                      SourceLocation atStartLoc,
-                                      SourceLocation CategoryNameLoc);
+  static ObjCCategoryImplDecl *
+  Create(ASTContext &C, DeclContext *DC, const IdentifierInfo *Id,
+         ObjCInterfaceDecl *classInterface, SourceLocation nameLoc,
+         SourceLocation atStartLoc, SourceLocation CategoryNameLoc);
   static ObjCCategoryImplDecl *CreateDeserialized(ASTContext &C, unsigned ID);
 
   ObjCCategoryDecl *getCategoryDecl() const;
diff --git a/clang/include/clang/AST/DeclTemplate.h b/clang/include/clang/AST/DeclTemplate.h
index e3b6a7efb1127..cb598cb81840d 100644
--- a/clang/include/clang/AST/DeclTemplate.h
+++ b/clang/include/clang/AST/DeclTemplate.h
@@ -1389,14 +1389,14 @@ class NonTypeTemplateParmDecl final
 
   NonTypeTemplateParmDecl(DeclContext *DC, SourceLocation StartLoc,
                           SourceLocation IdLoc, unsigned D, unsigned P,
-                          IdentifierInfo *Id, QualType T,
+                          const IdentifierInfo *Id, QualType T,
                           bool ParameterPack, TypeSourceInfo *TInfo)
       : DeclaratorDecl(NonTypeTemplateParm, DC, IdLoc, Id, T, TInfo, StartLoc),
         TemplateParmPosition(D, P), ParameterPack(ParameterPack) {}
 
   NonTypeTemplateParmDecl(DeclContext *DC, SourceLocation StartLoc,
                           SourceLocation IdLoc, unsigned D, unsigned P,
-                          IdentifierInfo *Id, QualType T,
+                          const IdentifierInfo *Id, QualType T,
                           TypeSourceInfo *TInfo,
                           ArrayRef<QualType> ExpandedTypes,
                           ArrayRef<TypeSourceInfo *> ExpandedTInfos);
@@ -1404,12 +1404,12 @@ class NonTypeTemplateParmDecl final
 public:
   static NonTypeTemplateParmDecl *
   Create(const ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-         SourceLocation IdLoc, unsigned D, unsigned P, IdentifierInfo *Id,
+         SourceLocation IdLoc, unsigned D, unsigned P, const IdentifierInfo *Id,
          QualType T, bool ParameterPack, TypeSourceInfo *TInfo);
 
   static NonTypeTemplateParmDecl *
   Create(const ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-         SourceLocation IdLoc, unsigned D, unsigned P, IdentifierInfo *Id,
+         SourceLocation IdLoc, unsigned D, unsigned P, const IdentifierInfo *Id,
          QualType T, TypeSourceInfo *TInfo, ArrayRef<QualType> ExpandedTypes,
          ArrayRef<TypeSourceInfo *> ExpandedTInfos);
 
diff --git a/clang/include/clang/AST/ExprCXX.h b/clang/include/clang/AST/ExprCXX.h
index 7eb99a75e1fc9..d28e5c3a78ee4 100644
--- a/clang/include/clang/AST/ExprCXX.h
+++ b/clang/include/clang/AST/ExprCXX.h
@@ -2559,7 +2559,7 @@ class CXXDeleteExpr : public Expr {
 class PseudoDestructorTypeStorage {
   /// Either the type source information or the name of the type, if
   /// it couldn't be resolved due to type-dependence.
-  llvm::PointerUnion<TypeSourceInfo *, IdentifierInfo *> Type;
+  llvm::PointerUnion<TypeSourceInfo *, const IdentifierInfo *> Type;
 
   /// The starting source location of the pseudo-destructor type.
   SourceLocation Location;
@@ -2567,7 +2567,7 @@ class PseudoDestructorTypeStorage {
 public:
   PseudoDestructorTypeStorage() = default;
 
-  PseudoDestructorTypeStorage(IdentifierInfo *II, SourceLocation Loc)
+  PseudoDestructorTypeStorage(const IdentifierInfo *II, SourceLocation Loc)
       : Type(II), Location(Loc) {}
 
   PseudoDestructorTypeStorage(TypeSourceInfo *Info);
@@ -2576,8 +2576,8 @@ class PseudoDestructorTypeStorage {
     return Type.dyn_cast<TypeSourceInfo *>();
   }
 
-  IdentifierInfo *getIdentifier() const {
-    return Type.dyn_cast<IdentifierInfo *>();
+  const IdentifierInfo *getIdentifier() const {
+    return Type.dyn_cast<const IdentifierInfo *>();
   }
 
   SourceLocation getLocation() const { return Location; }
@@ -2708,7 +2708,7 @@ class CXXPseudoDestructorExpr : public Expr {
   /// In a dependent pseudo-destructor expression for which we do not
   /// have full type information on the destroyed type, provides the name
   /// of the destroyed type.
-  IdentifierInfo *getDestroyedTypeIdentifier() const {
+  const IdentifierInfo *getDestroyedTypeIdentifier() const {
     return DestroyedType.getIdentifier();
   }
 
diff --git a/clang/include/clang/AST/ExternalASTSource.h b/clang/include/clang/AST/ExternalASTSource.h
index 8e573965b0a33..230c83943c222 100644
--- a/clang/include/clang/AST/ExternalASTSource.h
+++ b/clang/include/clang/AST/ExternalASTSource.h
@@ -138,7 +138,7 @@ class ExternalASTSource : public RefCountedBase<ExternalASTSource> {
   virtual CXXBaseSpecifier *GetExternalCXXBaseSpecifiers(uint64_t Offset);
 
   /// Update an out-of-date identifier.
-  virtual void updateOutOfDateIdentifier(IdentifierInfo &II) {}
+  virtual void updateOutOfDateIdentifier(const IdentifierInfo &II) {}
 
   /// Find all declarations with the given name in the given context,
   /// and add them to the context by calling SetExternalVisibleDeclsForName
diff --git a/clang/include/clang/AST/NestedNameSpecifier.h b/clang/include/clang/AST/NestedNameSpecifier.h
index 3b6cf97211850..7b0c21b9e7cfb 100644
--- a/clang/include/clang/AST/NestedNameSpecifier.h
+++ b/clang/include/clang/AST/NestedNameSpecifier.h
@@ -124,7 +124,7 @@ class NestedNameSpecifier : public llvm::FoldingSetNode {
   /// cannot be resolved.
   static NestedNameSpecifier *Create(const ASTContext &Context,
                                      NestedNameSpecifier *Prefix,
-                                     IdentifierInfo *II);
+                                     const IdentifierInfo *II);
 
   /// Builds a nested name specifier that names a namespace.
   static NestedNameSpecifier *Create(const ASTContext &Context,
@@ -134,7 +134,7 @@ class NestedNameSpecifier : public llvm::FoldingSetNode {
   /// Builds a nested name specifier that names a namespace alias.
   static NestedNameSpecifier *Create(const ASTContext &Context,
                                      NestedNameSpecifier *Prefix,
-                                     NamespaceAliasDecl *Alias);
+                                     const NamespaceAliasDecl *Alias);
 
   /// Builds a nested name specifier that names a type.
   static NestedNameSpecifier *Create(const ASTContext &Context,
@@ -148,7 +148,7 @@ class NestedNameSpecifier : public llvm::FoldingSetNode {
   /// nested name specifier, e.g., in "x->Base::f", the "x" has a dependent
   /// type.
   static NestedNameSpecifier *Create(const ASTContext &Context,
-                                     IdentifierInfo *II);
+                                     const IdentifierInfo *II);
 
   /// Returns the nested name specifier representing the global
   /// scope.
diff --git a/clang/include/clang/Analysis/SelectorExtras.h b/clang/include/clang/Analysis/SelectorExtras.h
index 1e1daf5706bbf..ac2c2519beae3 100644
--- a/clang/include/clang/Analysis/SelectorExtras.h
+++ b/clang/include/clang/Analysis/SelectorExtras.h
@@ -15,10 +15,10 @@ namespace clang {
 
 template <typename... IdentifierInfos>
 static inline Selector getKeywordSelector(ASTContext &Ctx,
-                                          IdentifierInfos *... IIs) {
+                                          const IdentifierInfos *...IIs) {
   static_assert(sizeof...(IdentifierInfos) > 0,
                 "keyword selectors must have at least one argument");
-  SmallVector<IdentifierInfo *, 10> II({&Ctx.Idents.get(IIs)...});
+  SmallVector<const IdentifierInfo *, 10> II({&Ctx.Idents.get(IIs)...});
 
   return Ctx.Selectors.getSelector(II.size(), &II[0]);
 }
diff --git a/clang/include/clang/Basic/IdentifierTable.h b/clang/include/clang/Basic/IdentifierTable.h
index a091639bfa254..a893e6f4d3d39 100644
--- a/clang/include/clang/Basic/IdentifierTable.h
+++ b/clang/include/clang/Basic/IdentifierTable.h
@@ -913,12 +913,13 @@ class alignas(IdentifierInfoAlignment) MultiKeywordSelector
 
 public:
   // Constructor for keyword selectors.
-  MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV)
+  MultiKeywordSelector(unsigned nKeys, const IdentifierInfo **IIV)
       : DeclarationNameExtra(nKeys) {
     assert((nKeys > 1) && "not a multi-keyword selector");
 
     // Fill in the trailing keyword array.
-    IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this + 1);
+    const IdentifierInfo **KeyInfo =
+        reinterpret_cast<const IdentifierInfo **>(this + 1);
     for (unsigned i = 0; i != nKeys; ++i)
       KeyInfo[i] = IIV[i];
   }
@@ -928,7 +929,7 @@ class alignas(IdentifierInfoAlignment) MultiKeywordSelector
 
   using DeclarationNameExtra::getNumArgs;
 
-  using keyword_iterator = IdentifierInfo *const *;
+  using keyword_iterator = const IdentifierInfo *const *;
 
   keyword_iterator keyword_begin() const {
     return reinterpret_cast<keyword_iterator>(this + 1);
@@ -938,7 +939,7 @@ class alignas(IdentifierInfoAlignment) MultiKeywordSelector
     return keyword_begin() + getNumArgs();
   }
 
-  IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
+  const IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
     assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
     return keyword_begin()[i];
   }
@@ -991,10 +992,10 @@ class Selector {
   /// Do not reorder or add any arguments to this template
   /// without thoroughly understanding how tightly coupled these classes are.
   llvm::PointerIntPair<
-      llvm::PointerUnion<IdentifierInfo *, MultiKeywordSelector *>, 2>
+      llvm::PointerUnion<const IdentifierInfo *, MultiKeywordSelector *>, 2>
       InfoPtr;
 
-  Selector(IdentifierInfo *II, unsigned nArgs) {
+  Selector(const IdentifierInfo *II, unsigned nArgs) {
     assert(nArgs < 2 && "nArgs not equal to 0/1");
     InfoPtr.setPointerAndInt(II, nArgs + 1);
   }
@@ -1006,8 +1007,8 @@ class Selector {
     InfoPtr.setPointerAndInt(SI, MultiArg & 0b11);
   }
 
-  IdentifierInfo *getAsIdentifierInfo() const {
-    return InfoPtr.getPointer().dyn_cast<IdentifierInfo *>();
+  const IdentifierInfo *getAsIdentifierInfo() const {
+    return InfoPtr.getPointer().dyn_cast<const IdentifierInfo *>();
   }
 
   MultiKeywordSelector *getMultiKeywordSelector() const {
@@ -1075,7 +1076,7 @@ class Selector {
   ///
   /// \returns the uniqued identifier for this slot, or NULL if this slot has
   /// no corresponding identifier.
-  IdentifierInfo *getIdentifierInfoForSlot(unsigned argIndex) const;
+  const IdentifierInfo *getIdentifierInfoForSlot(unsigned argIndex) const;
 
   /// Retrieve the name at a given position in the selector.
   ///
@@ -1132,13 +1133,13 @@ class SelectorTable {
   ///
   /// \p NumArgs indicates whether this is a no argument selector "foo", a
   /// single argument selector "foo:" or multi-argument "foo:bar:".
-  Selector getSelector(unsigned NumArgs, IdentifierInfo **IIV);
+  Selector getSelector(unsigned NumArgs, const IdentifierInfo **IIV);
 
-  Selector getUnarySelector(IdentifierInfo *ID) {
+  Selector getUnarySelector(const IdentifierInfo *ID) {
     return Selector(ID, 1);
   }
 
-  Selector getNullarySelector(IdentifierInfo *ID) {
+  Selector getNullarySelector(const IdentifierInfo *ID) {
     return Selector(ID, 0);
   }
 
diff --git a/clang/include/clang/Lex/ExternalPreprocessorSource.h b/clang/include/clang/Lex/ExternalPreprocessorSource.h
index 685941b66bd8b..6775841860373 100644
--- a/clang/include/clang/Lex/ExternalPreprocessorSource.h
+++ b/clang/include/clang/Lex/ExternalPreprocessorSource.h
@@ -31,7 +31,7 @@ class ExternalPreprocessorSource {
   virtual void ReadDefinedMacros() = 0;
 
   /// Update an out-of-date identifier.
-  virtual void updateOutOfDateIdentifier(IdentifierInfo &II) = 0;
+  virtual void updateOutOfDateIdentifier(const IdentifierInfo &II) = 0;
 
   /// Return the identifier associated with the given ID number.
   ///
diff --git a/clang/include/clang/Lex/MacroInfo.h b/clang/include/clang/Lex/MacroInfo.h
index 1237fc62eb6cf..19a706216d509 100644
--- a/clang/include/clang/Lex/MacroInfo.h
+++ b/clang/include/clang/Lex/MacroInfo.h
@@ -515,7 +515,7 @@ class ModuleMacro : public llvm::FoldingSetNode {
   friend class Preprocessor;
 
   /// The name defined by the macro.
-  IdentifierInfo *II;
+  const IdentifierInfo *II;
 
   /// The body of the #define, or nullptr if this is a #undef.
   MacroInfo *Macro;
@@ -529,7 +529,7 @@ class ModuleMacro : public llvm::FoldingSetNode {
   /// The number of modules whose macros are directly overridden by this one.
   unsigned NumOverrides;
 
-  ModuleMacro(Module *OwningModule, IdentifierInfo *II, MacroInfo *Macro,
+  ModuleMacro(Module *OwningModule, const IdentifierInfo *II, MacroInfo *Macro,
               ArrayRef<ModuleMacro *> Overrides)
       : II(II), Macro(Macro), OwningModule(OwningModule),
         NumOverrides(Overrides.size()) {
@@ -539,7 +539,7 @@ class ModuleMacro : public llvm::FoldingSetNode {
 
 public:
   static ModuleMacro *create(Preprocessor &PP, Module *OwningModule,
-                             IdentifierInfo *II, MacroInfo *Macro,
+                             const IdentifierInfo *II, MacroInfo *Macro,
                              ArrayRef<ModuleMacro *> Overrides);
 
   void Profile(llvm::FoldingSetNodeID &ID) const {
@@ -553,7 +553,7 @@ class ModuleMacro : public llvm::FoldingSetNode {
   }
 
   /// Get the name of the macro.
-  IdentifierInfo *getName() const { return II; }
+  const IdentifierInfo *getName() const { return II; }
 
   /// Get the ID of the module that exports this macro.
   Module *getOwningModule() const { return OwningModule; }
diff --git a/clang/include/clang/Lex/Preprocessor.h b/clang/include/clang/Lex/Preprocessor.h
index 0836b7d439bb0..e89b4a2c5230e 100644
--- a/clang/include/clang/Lex/Preprocessor.h
+++ b/clang/include/clang/Lex/Preprocessor.h
@@ -836,7 +836,7 @@ class Preprocessor {
     ModuleMacroInfo *getModuleInfo(Preprocessor &PP,
                                    const IdentifierInfo *II) const {
       if (II->isOutOfDate())
-        PP.updateOutOfDateIdentifier(const_cast<IdentifierInfo&>(*II));
+        PP.updateOutOfDateIdentifier(*II);
       // FIXME: Find a spare bit on IdentifierInfo and store a
       //        HasModuleMacros flag.
       if (!II->hasMacroDefinition() ||
@@ -1162,7 +1162,7 @@ class Preprocessor {
   /// skipped.
   llvm::DenseMap<const char *, unsigned> RecordedSkippedRanges;
 
-  void updateOutOfDateIdentifier(IdentifierInfo &II) const;
+  void updateOutOfDateIdentifier(const IdentifierInfo &II) const;
 
 public:
   Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
@@ -1432,14 +1432,15 @@ class Preprocessor {
                                MacroDirective *MD);
 
   /// Register an exported macro for a module and identifier.
-  ModuleMacro *addModuleMacro(Module *Mod, IdentifierInfo *II, MacroInfo *Macro,
+  ModuleMacro *addModuleMacro(Module *Mod, const IdentifierInfo *II,
+                              MacroInfo *Macro,
                               ArrayRef<ModuleMacro *> Overrides, bool &IsNew);
   ModuleMacro *getModuleMacro(Module *Mod, const IdentifierInfo *II);
 
   /// Get the list of leaf (non-overridden) module macros for a name.
   ArrayRef<ModuleMacro*> getLeafModuleMacros(const IdentifierInfo *II) const {
     if (II->isOutOfDate())
-      updateOutOfDateIdentifier(const_cast<IdentifierInfo&>(*II));
+      updateOutOfDateIdentifier(*II);
     auto I = LeafModuleMacros.find(II);
     if (I != LeafModuleMacros.end())
       return I->second;
diff --git a/clang/include/clang/Parse/Parser.h b/clang/include/clang/Parse/Parser.h
index 8bc929b1dfe4b..3a055c10ffb38 100644
--- a/clang/include/clang/Parse/Parser.h
+++ b/clang/include/clang/Parse/Parser.h
@@ -329,7 +329,7 @@ class Parser : public CodeCompletionHandler {
   };
 
   /// Identifiers which have been declared within a tentative parse.
-  SmallVector<IdentifierInfo *, 8> TentativelyDeclaredIdentifiers;
+  SmallVector<const IdentifierInfo *, 8> TentativelyDeclaredIdentifiers;
 
   /// Tracker for '<' tokens that might have been intended to be treated as an
   /// angle bracket instead of a less-than comparison.
@@ -1927,15 +1927,11 @@ class Parser : public CodeCompletionHandler {
                                   bool EnteringContext, IdentifierInfo &II,
                                   CXXScopeSpec &SS);
 
-  bool ParseOptionalCXXScopeSpecifier(CXXScopeSpec &SS,
-                                      ParsedType ObjectType,
-                                      bool ObjectHasErrors,
-                                      bool EnteringContext,
-                                      bool *MayBePseudoDestructor = nullptr,
-                                      bool IsTypename = false,
-                                      IdentifierInfo **LastII = nullptr,
-                                      bool OnlyNamespace = false,
-                                      bool InUsingDeclaration = false);
+  bool ParseOptionalCXXScopeSpecifier(
+      CXXScopeSpec &SS, ParsedType ObjectType, bool ObjectHasErrors,
+      bool EnteringContext, bool *MayBePseudoDestructor = nullptr,
+      bool IsTypename = false, const IdentifierInfo **LastII = nullptr,
+      bool OnlyNamespace = false, bool InUsingDeclaration = false);
 
   //===--------------------------------------------------------------------===//
   // C++11 5.1.2: Lambda expressions
diff --git a/clang/include/clang/Sema/CodeCompleteConsumer.h b/clang/include/clang/Sema/CodeCompleteConsumer.h
index a2028e40f83d5..0924dc27af82b 100644
--- a/clang/include/clang/Sema/CodeCompleteConsumer.h
+++ b/clang/include/clang/Sema/CodeCompleteConsumer.h
@@ -362,7 +362,7 @@ class CodeCompletionContext {
   QualType BaseType;
 
   /// The identifiers for Objective-C selector parts.
-  ArrayRef<IdentifierInfo *> SelIdents;
+  ArrayRef<const IdentifierInfo *> SelIdents;
 
   /// The scope specifier that comes before the completion token e.g.
   /// "a::b::"
@@ -378,8 +378,9 @@ class CodeCompletionContext {
       : CCKind(CCKind), IsUsingDeclaration(false), SelIdents(std::nullopt) {}
 
   /// Construct a new code-completion context of the given kind.
-  CodeCompletionContext(Kind CCKind, QualType T,
-                        ArrayRef<IdentifierInfo *> SelIdents = std::nullopt)
+  CodeCompletionContext(
+      Kind CCKind, QualType T,
+      ArrayRef<const IdentifierInfo *> SelIdents = std::nullopt)
       : CCKind(CCKind), IsUsingDeclaration(false), SelIdents(SelIdents) {
     if (CCKind == CCC_DotMemberAccess || CCKind == CCC_ArrowMemberAccess ||
         CCKind == CCC_ObjCPropertyAccess || CCKind == CCC_ObjCClassMessage ||
@@ -406,7 +407,7 @@ class CodeCompletionContext {
   QualType getBaseType() const { return BaseType; }
 
   /// Retrieve the Objective-C selector identifiers.
-  ArrayRef<IdentifierInfo *> getSelIdents() const { return SelIdents; }
+  ArrayRef<const IdentifierInfo *> getSelIdents() const { return SelIdents; }
 
   /// Determines whether we want C++ constructors as results within this
   /// context.
diff --git a/clang/include/clang/Sema/DeclSpec.h b/clang/include/clang/Sema/DeclSpec.h
index a176159707486..c9eecdafe62c7 100644
--- a/clang/include/clang/Sema/DeclSpec.h
+++ b/clang/include/clang/Sema/DeclSpec.h
@@ -1049,7 +1049,7 @@ class UnqualifiedId {
   union {
     /// When Kind == IK_Identifier, the parsed identifier, or when
     /// Kind == IK_UserLiteralId, the identifier suffix.
-    IdentifierInfo *Identifier;
+    const IdentifierInfo *Identifier;
 
     /// When Kind == IK_OperatorFunctionId, the overloaded operator
     /// that we parsed.
@@ -1111,7 +1111,7 @@ class UnqualifiedId {
   /// \param IdLoc the location of the parsed identifier.
   void setIdentifier(const IdentifierInfo *Id, SourceLocation IdLoc) {
     Kind = UnqualifiedIdKind::IK_Identifier;
-    Identifier = const_cast<IdentifierInfo *>(Id);
+    Identifier = Id;
     StartLocation = EndLocation = IdLoc;
   }
 
@@ -1154,9 +1154,9 @@ class UnqualifiedId {
   ///
   /// \param IdLoc the location of the identifier.
   void setLiteralOperatorId(const IdentifierInfo *Id, SourceLocation OpLoc,
-                              SourceLocation IdLoc) {
+                            SourceLocation IdLoc) {
     Kind = UnqualifiedIdKind::IK_LiteralOperatorId;
-    Identifier = const_cast<IdentifierInfo *>(Id);
+    Identifier = Id;
     StartLocation = OpLoc;
     EndLocation = IdLoc;
   }
@@ -1225,7 +1225,7 @@ class UnqualifiedId {
   /// \param Id the identifier.
   void setImplicitSelfParam(const IdentifierInfo *Id) {
     Kind = UnqualifiedIdKind::IK_ImplicitSelfParam;
-    Identifier = const_cast<IdentifierInfo *>(Id);
+    Identifier = Id;
     StartLocation = EndLocation = SourceLocation();
   }
 
@@ -1327,7 +1327,7 @@ struct DeclaratorChunk {
   /// Parameter type lists will have type info (if the actions module provides
   /// it), but may have null identifier info: e.g. for 'void foo(int X, int)'.
   struct ParamInfo {
-    IdentifierInfo *Ident;
+    const IdentifierInfo *Ident;
     SourceLocation IdentLoc;
     Decl *Param;
 
@@ -1339,11 +1339,10 @@ struct DeclaratorChunk {
     std::unique_ptr<CachedTokens> DefaultArgTokens;
 
     ParamInfo() = default;
-    ParamInfo(IdentifierInfo *ident, SourceLocation iloc,
-              Decl *param,
+    ParamInfo(const IdentifierInfo *ident, SourceLocation iloc, Decl *param,
               std::unique_ptr<CachedTokens> DefArgTokens = nullptr)
-      : Ident(ident), IdentLoc(iloc), Param(param),
-        DefaultArgTokens(std::move(DefArgTokens)) {}
+        : Ident(ident), IdentLoc(iloc), Param(param),
+          DefaultArgTokens(std::move(DefArgTokens)) {}
   };
 
   struct TypeAndRange {
@@ -2326,7 +2325,7 @@ class Declarator {
     return BindingGroup.isSet();
   }
 
-  IdentifierInfo *getIdentifier() const {
+  const IdentifierInfo *getIdentifier() const {
     if (Name.getKind() == UnqualifiedIdKind::IK_Identifier)
       return Name.Identifier;
 
@@ -2335,7 +2334,7 @@ class Declarator {
   SourceLocation getIdentifierLoc() const { return Name.StartLocation; }
 
   /// Set the name of this declarator to be the given identifier.
-  void SetIdentifier(IdentifierInfo *Id, SourceLocation IdLoc) {
+  void SetIdentifier(const IdentifierInfo *Id, SourceLocation IdLoc) {
     Name.setIdentifier(Id, IdLoc);
   }
 
diff --git a/clang/include/clang/Sema/ParsedTemplate.h b/clang/include/clang/Sema/ParsedTemplate.h
index 65182d57246ae..ac4dbbf294caf 100644
--- a/clang/include/clang/Sema/ParsedTemplate.h
+++ b/clang/include/clang/Sema/ParsedTemplate.h
@@ -159,7 +159,7 @@ namespace clang {
     SourceLocation TemplateNameLoc;
 
     /// FIXME: Temporarily stores the name of a specialization
-    IdentifierInfo *Name;
+    const IdentifierInfo *Name;
 
     /// FIXME: Temporarily stores the overloaded operator kind.
     OverloadedOperatorKind Operator;
@@ -197,7 +197,7 @@ namespace clang {
     /// appends it to List.
     static TemplateIdAnnotation *
     Create(SourceLocation TemplateKWLoc, SourceLocation TemplateNameLoc,
-           IdentifierInfo *Name, OverloadedOperatorKind OperatorKind,
+           const IdentifierInfo *Name, OverloadedOperatorKind OperatorKind,
            ParsedTemplateTy OpaqueTemplateName, TemplateNameKind TemplateKind,
            SourceLocation LAngleLoc, SourceLocation RAngleLoc,
            ArrayRef<ParsedTemplateArgument> TemplateArgs, bool ArgsInvalid,
@@ -236,7 +236,8 @@ namespace clang {
     TemplateIdAnnotation(const TemplateIdAnnotation &) = delete;
 
     TemplateIdAnnotation(SourceLocation TemplateKWLoc,
-                         SourceLocation TemplateNameLoc, IdentifierInfo *Name,
+                         SourceLocation TemplateNameLoc,
+                         const IdentifierInfo *Name,
                          OverloadedOperatorKind OperatorKind,
                          ParsedTemplateTy OpaqueTemplateName,
                          TemplateNameKind TemplateKind,
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index e3e255a0dd76f..0ee4f3c8e127f 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -568,7 +568,7 @@ class Sema final : public SemaBase {
 
   /// Invent a new identifier for parameters of abbreviated templates.
   IdentifierInfo *
-  InventAbbreviatedTemplateParameterTypeName(IdentifierInfo *ParamName,
+  InventAbbreviatedTemplateParameterTypeName(const IdentifierInfo *ParamName,
                                              unsigned Index);
 
   void emitAndClearUnusedLocalTypedefWarnings();
@@ -2958,9 +2958,9 @@ class Sema final : public SemaBase {
                                                   SourceLocation NameLoc,
                                                   TypeSourceInfo *TSInfo);
   ParmVarDecl *CheckParameter(DeclContext *DC, SourceLocation StartLoc,
-                              SourceLocation NameLoc, IdentifierInfo *Name,
-                              QualType T, TypeSourceInfo *TSInfo,
-                              StorageClass SC);
+                              SourceLocation NameLoc,
+                              const IdentifierInfo *Name, QualType T,
+                              TypeSourceInfo *TSInfo, StorageClass SC);
 
   // Contexts where using non-trivial C union types can be disallowed. This is
   // passed to err_non_trivial_c_union_in_invalid_context.
@@ -3365,7 +3365,7 @@ class Sema final : public SemaBase {
   /// variable.
   void DiagnoseUnusedButSetDecl(const VarDecl *VD, DiagReceiverTy DiagReceiver);
 
-  ObjCInterfaceDecl *getObjCInterfaceDecl(IdentifierInfo *&Id,
+  ObjCInterfaceDecl *getObjCInterfaceDecl(const IdentifierInfo *&Id,
                                           SourceLocation IdLoc,
                                           bool TypoCorrection = false);
 
@@ -3442,8 +3442,9 @@ class Sema final : public SemaBase {
   /// VerifyBitField - verifies that a bit field expression is an ICE and has
   /// the correct width, and that the field type is valid.
   /// Returns false on success.
-  ExprResult VerifyBitField(SourceLocation FieldLoc, IdentifierInfo *FieldName,
-                            QualType FieldTy, bool IsMsStruct, Expr *BitWidth);
+  ExprResult VerifyBitField(SourceLocation FieldLoc,
+                            const IdentifierInfo *FieldName, QualType FieldTy,
+                            bool IsMsStruct, Expr *BitWidth);
 
   /// IsValueInFlagEnum - Determine if a value is allowed as part of a flag
   /// enum. If AllowMask is true, then we also allow the complement of a valid
@@ -4638,7 +4639,8 @@ class Sema final : public SemaBase {
 
   VarDecl *BuildExceptionDeclaration(Scope *S, TypeSourceInfo *TInfo,
                                      SourceLocation StartLoc,
-                                     SourceLocation IdLoc, IdentifierInfo *Id);
+                                     SourceLocation IdLoc,
+                                     const IdentifierInfo *Id);
 
   Decl *ActOnExceptionDeclarator(Scope *S, Declarator &D);
 
@@ -6555,12 +6557,12 @@ class Sema final : public SemaBase {
 
   ParsedType getInheritingConstructorName(CXXScopeSpec &SS,
                                           SourceLocation NameLoc,
-                                          IdentifierInfo &Name);
+                                          const IdentifierInfo &Name);
 
-  ParsedType getConstructorName(IdentifierInfo &II, SourceLocation NameLoc,
-                                Scope *S, CXXScopeSpec &SS,
-                                bool EnteringContext);
-  ParsedType getDestructorName(IdentifierInfo &II, SourceLocation NameLoc,
+  ParsedType getConstructorName(const IdentifierInfo &II,
+                                SourceLocation NameLoc, Scope *S,
+                                CXXScopeSpec &SS, bool EnteringContext);
+  ParsedType getDestructorName(const IdentifierInfo &II, SourceLocation NameLoc,
                                Scope *S, CXXScopeSpec &SS,
                                ParsedType ObjectType, bool EnteringContext);
 
@@ -6960,7 +6962,7 @@ class Sema final : public SemaBase {
   concepts::Requirement *ActOnTypeRequirement(SourceLocation TypenameKWLoc,
                                               CXXScopeSpec &SS,
                                               SourceLocation NameLoc,
-                                              IdentifierInfo *TypeName,
+                                              const IdentifierInfo *TypeName,
                                               TemplateIdAnnotation *TemplateId);
   concepts::Requirement *ActOnCompoundRequirement(Expr *E,
                                                   SourceLocation NoexceptLoc);
@@ -9116,7 +9118,7 @@ class Sema final : public SemaBase {
 
   TypeResult
   ActOnTemplateIdType(Scope *S, CXXScopeSpec &SS, SourceLocation TemplateKWLoc,
-                      TemplateTy Template, IdentifierInfo *TemplateII,
+                      TemplateTy Template, const IdentifierInfo *TemplateII,
                       SourceLocation TemplateIILoc, SourceLocation LAngleLoc,
                       ASTTemplateArgsPtr TemplateArgs, SourceLocation RAngleLoc,
                       bool IsCtorOrDtorName = false, bool IsClassName = false,
@@ -9457,7 +9459,7 @@ class Sema final : public SemaBase {
   TypeResult
   ActOnTypenameType(Scope *S, SourceLocation TypenameLoc,
                     const CXXScopeSpec &SS, SourceLocation TemplateLoc,
-                    TemplateTy TemplateName, IdentifierInfo *TemplateII,
+                    TemplateTy TemplateName, const IdentifierInfo *TemplateII,
                     SourceLocation TemplateIILoc, SourceLocation LAngleLoc,
                     ASTTemplateArgsPtr TemplateArgs, SourceLocation RAngleLoc);
 
@@ -9535,14 +9537,15 @@ class Sema final : public SemaBase {
 
   Decl *ActOnConceptDefinition(Scope *S,
                                MultiTemplateParamsArg TemplateParameterLists,
-                               IdentifierInfo *Name, SourceLocation NameLoc,
-                               Expr *ConstraintExpr);
+                               const IdentifierInfo *Name,
+                               SourceLocation NameLoc, Expr *ConstraintExpr);
 
   void CheckConceptRedefinition(ConceptDecl *NewDecl, LookupResult &Previous,
                                 bool &AddToScope);
 
   TypeResult ActOnDependentTag(Scope *S, unsigned TagSpec, TagUseKind TUK,
-                               const CXXScopeSpec &SS, IdentifierInfo *Name,
+                               const CXXScopeSpec &SS,
+                               const IdentifierInfo *Name,
                                SourceLocation TagLoc, SourceLocation NameLoc);
 
   void MarkAsLateParsedTemplate(FunctionDecl *FD, Decl *FnD,
@@ -11988,22 +11991,22 @@ class Sema final : public SemaBase {
       SkipBodyInfo *SkipBody);
 
   ObjCCategoryDecl *ActOnStartCategoryInterface(
-      SourceLocation AtInterfaceLoc, IdentifierInfo *ClassName,
+      SourceLocation AtInterfaceLoc, const IdentifierInfo *ClassName,
       SourceLocation ClassLoc, ObjCTypeParamList *typeParamList,
-      IdentifierInfo *CategoryName, SourceLocation CategoryLoc,
+      const IdentifierInfo *CategoryName, SourceLocation CategoryLoc,
       Decl *const *ProtoRefs, unsigned NumProtoRefs,
       const SourceLocation *ProtoLocs, SourceLocation EndProtoLoc,
       const ParsedAttributesView &AttrList);
 
   ObjCImplementationDecl *ActOnStartClassImplementation(
-      SourceLocation AtClassImplLoc, IdentifierInfo *ClassName,
-      SourceLocation ClassLoc, IdentifierInfo *SuperClassname,
+      SourceLocation AtClassImplLoc, const IdentifierInfo *ClassName,
+      SourceLocation ClassLoc, const IdentifierInfo *SuperClassname,
       SourceLocation SuperClassLoc, const ParsedAttributesView &AttrList);
 
   ObjCCategoryImplDecl *ActOnStartCategoryImplementation(
-      SourceLocation AtCatImplLoc, IdentifierInfo *ClassName,
-      SourceLocation ClassLoc, IdentifierInfo *CatName, SourceLocation CatLoc,
-      const ParsedAttributesView &AttrList);
+      SourceLocation AtCatImplLoc, const IdentifierInfo *ClassName,
+      SourceLocation ClassLoc, const IdentifierInfo *CatName,
+      SourceLocation CatLoc, const ParsedAttributesView &AttrList);
 
   DeclGroupPtrTy ActOnFinishObjCImplementation(Decl *ObjCImpDecl,
                                                ArrayRef<Decl *> Decls);
@@ -12186,11 +12189,13 @@ class Sema final : public SemaBase {
   bool CheckObjCDeclScope(Decl *D);
 
   void ActOnDefs(Scope *S, Decl *TagD, SourceLocation DeclStart,
-                 IdentifierInfo *ClassName, SmallVectorImpl<Decl *> &Decls);
+                 const IdentifierInfo *ClassName,
+                 SmallVectorImpl<Decl *> &Decls);
 
   VarDecl *BuildObjCExceptionDecl(TypeSourceInfo *TInfo, QualType ExceptionType,
                                   SourceLocation StartLoc, SourceLocation IdLoc,
-                                  IdentifierInfo *Id, bool Invalid = false);
+                                  const IdentifierInfo *Id,
+                                  bool Invalid = false);
 
   Decl *ActOnObjCExceptionDecl(Scope *S, Declarator &D);
 
@@ -12307,8 +12312,8 @@ class Sema final : public SemaBase {
                                        SourceLocation SuperLoc,
                                        QualType SuperType, bool Super);
 
-  ExprResult ActOnClassPropertyRefExpr(IdentifierInfo &receiverName,
-                                       IdentifierInfo &propertyName,
+  ExprResult ActOnClassPropertyRefExpr(const IdentifierInfo &receiverName,
+                                       const IdentifierInfo &propertyName,
                                        SourceLocation receiverNameLoc,
                                        SourceLocation propertyNameLoc);
 
@@ -12783,18 +12788,18 @@ class Sema final : public SemaBase {
                                    bool IsParameter);
   void CodeCompleteObjCMessageReceiver(Scope *S);
   void CodeCompleteObjCSuperMessage(Scope *S, SourceLocation SuperLoc,
-                                    ArrayRef<IdentifierInfo *> SelIdents,
+                                    ArrayRef<const IdentifierInfo *> SelIdents,
                                     bool AtArgumentExpression);
   void CodeCompleteObjCClassMessage(Scope *S, ParsedType Receiver,
-                                    ArrayRef<IdentifierInfo *> SelIdents,
+                                    ArrayRef<const IdentifierInfo *> SelIdents,
                                     bool AtArgumentExpression,
                                     bool IsSuper = false);
-  void CodeCompleteObjCInstanceMessage(Scope *S, Expr *Receiver,
-                                       ArrayRef<IdentifierInfo *> SelIdents,
-                                       bool AtArgumentExpression,
-                                       ObjCInterfaceDecl *Super = nullptr);
+  void CodeCompleteObjCInstanceMessage(
+      Scope *S, Expr *Receiver, ArrayRef<const IdentifierInfo *> SelIdents,
+      bool AtArgumentExpression, ObjCInterfaceDecl *Super = nullptr);
   void CodeCompleteObjCForCollection(Scope *S, DeclGroupPtrTy IterationVar);
-  void CodeCompleteObjCSelector(Scope *S, ArrayRef<IdentifierInfo *> SelIdents);
+  void CodeCompleteObjCSelector(Scope *S,
+                                ArrayRef<const IdentifierInfo *> SelIdents);
   void
   CodeCompleteObjCProtocolReferences(ArrayRef<IdentifierLocPair> Protocols);
   void CodeCompleteObjCProtocolDecl(Scope *S);
@@ -12814,11 +12819,11 @@ class Sema final : public SemaBase {
   void CodeCompleteObjCMethodDecl(Scope *S,
                                   std::optional<bool> IsInstanceMethod,
                                   ParsedType ReturnType);
-  void CodeCompleteObjCMethodDeclSelector(Scope *S, bool IsInstanceMethod,
-                                          bool AtParameterName,
-                                          ParsedType ReturnType,
-                                          ArrayRef<IdentifierInfo *> SelIdents);
-  void CodeCompleteObjCClassPropertyRefExpr(Scope *S, IdentifierInfo &ClassName,
+  void CodeCompleteObjCMethodDeclSelector(
+      Scope *S, bool IsInstanceMethod, bool AtParameterName,
+      ParsedType ReturnType, ArrayRef<const IdentifierInfo *> SelIdents);
+  void CodeCompleteObjCClassPropertyRefExpr(Scope *S,
+                                            const IdentifierInfo &ClassName,
                                             SourceLocation ClassNameLoc,
                                             bool IsBaseExprStatement);
   void CodeCompletePreprocessorDirective(bool InConditional);
diff --git a/clang/include/clang/Serialization/ASTReader.h b/clang/include/clang/Serialization/ASTReader.h
index 1911252b34cd1..5fd55a519c6b0 100644
--- a/clang/include/clang/Serialization/ASTReader.h
+++ b/clang/include/clang/Serialization/ASTReader.h
@@ -1082,12 +1082,12 @@ class ASTReader
 
   /// The set of lookup results that we have faked in order to support
   /// merging of partially deserialized decls but that we have not yet removed.
-  llvm::SmallMapVector<IdentifierInfo *, SmallVector<NamedDecl*, 2>, 16>
-    PendingFakeLookupResults;
+  llvm::SmallMapVector<const IdentifierInfo *, SmallVector<NamedDecl *, 2>, 16>
+      PendingFakeLookupResults;
 
   /// The generation number of each identifier, which keeps track of
   /// the last time we loaded information about this identifier.
-  llvm::DenseMap<IdentifierInfo *, unsigned> IdentifierGeneration;
+  llvm::DenseMap<const IdentifierInfo *, unsigned> IdentifierGeneration;
 
   /// Contains declarations and definitions that could be
   /// "interesting" to the ASTConsumer, when we get that AST consumer.
@@ -2330,10 +2330,10 @@ class ASTReader
   void ReadDefinedMacros() override;
 
   /// Update an out-of-date identifier.
-  void updateOutOfDateIdentifier(IdentifierInfo &II) override;
+  void updateOutOfDateIdentifier(const IdentifierInfo &II) override;
 
   /// Note that this identifier is up-to-date.
-  void markIdentifierUpToDate(IdentifierInfo *II);
+  void markIdentifierUpToDate(const IdentifierInfo *II);
 
   /// Load all external visible decls in the given DeclContext.
   void completeVisibleDeclsMap(const DeclContext *DC) override;
diff --git a/clang/lib/ARCMigrate/ObjCMT.cpp b/clang/lib/ARCMigrate/ObjCMT.cpp
index 0786c81516b2d..b9dcfb8951b3e 100644
--- a/clang/lib/ARCMigrate/ObjCMT.cpp
+++ b/clang/lib/ARCMigrate/ObjCMT.cpp
@@ -1144,7 +1144,7 @@ static bool IsValidIdentifier(ASTContext &Ctx,
     return false;
   std::string NameString = Name;
   NameString[0] = toLowercase(NameString[0]);
-  IdentifierInfo *II = &Ctx.Idents.get(NameString);
+  const IdentifierInfo *II = &Ctx.Idents.get(NameString);
   return II->getTokenID() ==  tok::identifier;
 }
 
@@ -1166,7 +1166,7 @@ bool ObjCMigrateASTConsumer::migrateProperty(ASTContext &Ctx,
   if (OIT_Family != OIT_None)
     return false;
 
-  IdentifierInfo *getterName = GetterSelector.getIdentifierInfoForSlot(0);
+  const IdentifierInfo *getterName = GetterSelector.getIdentifierInfoForSlot(0);
   Selector SetterSelector =
   SelectorTable::constructSetterSelector(PP.getIdentifierTable(),
                                          PP.getSelectorTable(),
@@ -1311,7 +1311,8 @@ void ObjCMigrateASTConsumer::migrateFactoryMethod(ASTContext &Ctx,
   std::string StringLoweredClassName = LoweredClassName.lower();
   LoweredClassName = StringLoweredClassName;
 
-  IdentifierInfo *MethodIdName = OM->getSelector().getIdentifierInfoForSlot(0);
+  const IdentifierInfo *MethodIdName =
+      OM->getSelector().getIdentifierInfoForSlot(0);
   // Handle method with no name at its first selector slot; e.g. + (id):(int)x.
   if (!MethodIdName)
     return;
diff --git a/clang/lib/ARCMigrate/TransAPIUses.cpp b/clang/lib/ARCMigrate/TransAPIUses.cpp
index 638850dcf9ecc..8f5d4f4bde06c 100644
--- a/clang/lib/ARCMigrate/TransAPIUses.cpp
+++ b/clang/lib/ARCMigrate/TransAPIUses.cpp
@@ -41,7 +41,7 @@ class APIChecker : public RecursiveASTVisitor<APIChecker> {
     getReturnValueSel = sels.getUnarySelector(&ids.get("getReturnValue"));
     setReturnValueSel = sels.getUnarySelector(&ids.get("setReturnValue"));
 
-    IdentifierInfo *selIds[2];
+    const IdentifierInfo *selIds[2];
     selIds[0] = &ids.get("getArgument");
     selIds[1] = &ids.get("atIndex");
     getArgumentSel = sels.getSelector(2, selIds);
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index f7f55dc4e7a9f..2fa6aedca4c6a 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -6929,16 +6929,13 @@ ASTContext::getCanonicalNestedNameSpecifier(NestedNameSpecifier *NNS) const {
     //   typedef typename T::type T1;
     //   typedef typename T1::type T2;
     if (const auto *DNT = T->getAs<DependentNameType>())
-      return NestedNameSpecifier::Create(
-          *this, DNT->getQualifier(),
-          const_cast<IdentifierInfo *>(DNT->getIdentifier()));
+      return NestedNameSpecifier::Create(*this, DNT->getQualifier(),
+                                         DNT->getIdentifier());
     if (const auto *DTST = T->getAs<DependentTemplateSpecializationType>())
-      return NestedNameSpecifier::Create(*this, DTST->getQualifier(), true,
-                                         const_cast<Type *>(T));
+      return NestedNameSpecifier::Create(*this, DTST->getQualifier(), true, T);
 
     // TODO: Set 'Template' parameter to true for other template types.
-    return NestedNameSpecifier::Create(*this, nullptr, false,
-                                       const_cast<Type *>(T));
+    return NestedNameSpecifier::Create(*this, nullptr, false, T);
   }
 
   case NestedNameSpecifier::Global:
diff --git a/clang/lib/AST/ASTImporter.cpp b/clang/lib/AST/ASTImporter.cpp
index 45d4c9600537b..d5ec5ee409156 100644
--- a/clang/lib/AST/ASTImporter.cpp
+++ b/clang/lib/AST/ASTImporter.cpp
@@ -8383,8 +8383,8 @@ ASTNodeImporter::VisitCXXPseudoDestructorExpr(CXXPseudoDestructorExpr *E) {
     return std::move(Err);
 
   PseudoDestructorTypeStorage Storage;
-  if (IdentifierInfo *FromII = E->getDestroyedTypeIdentifier()) {
-    IdentifierInfo *ToII = Importer.Import(FromII);
+  if (const IdentifierInfo *FromII = E->getDestroyedTypeIdentifier()) {
+    const IdentifierInfo *ToII = Importer.Import(FromII);
     ExpectedSLoc ToDestroyedTypeLocOrErr = import(E->getDestroyedTypeLoc());
     if (!ToDestroyedTypeLocOrErr)
       return ToDestroyedTypeLocOrErr.takeError();
@@ -10194,7 +10194,7 @@ Expected<Selector> ASTImporter::Import(Selector FromSel) {
   if (FromSel.isNull())
     return Selector{};
 
-  SmallVector<IdentifierInfo *, 4> Idents;
+  SmallVector<const IdentifierInfo *, 4> Idents;
   Idents.push_back(Import(FromSel.getIdentifierInfoForSlot(0)));
   for (unsigned I = 1, N = FromSel.getNumArgs(); I < N; ++I)
     Idents.push_back(Import(FromSel.getIdentifierInfoForSlot(I)));
diff --git a/clang/lib/AST/Decl.cpp b/clang/lib/AST/Decl.cpp
index 131f82985e903..60e0a3aecf6c8 100644
--- a/clang/lib/AST/Decl.cpp
+++ b/clang/lib/AST/Decl.cpp
@@ -2913,10 +2913,10 @@ VarDecl::setInstantiationOfStaticDataMember(VarDecl *VD,
 //===----------------------------------------------------------------------===//
 
 ParmVarDecl *ParmVarDecl::Create(ASTContext &C, DeclContext *DC,
-                                 SourceLocation StartLoc,
-                                 SourceLocation IdLoc, IdentifierInfo *Id,
-                                 QualType T, TypeSourceInfo *TInfo,
-                                 StorageClass S, Expr *DefArg) {
+                                 SourceLocation StartLoc, SourceLocation IdLoc,
+                                 const IdentifierInfo *Id, QualType T,
+                                 TypeSourceInfo *TInfo, StorageClass S,
+                                 Expr *DefArg) {
   return new (C, DC) ParmVarDecl(ParmVar, C, DC, StartLoc, IdLoc, Id, T, TInfo,
                                  S, DefArg);
 }
@@ -4511,7 +4511,7 @@ unsigned FunctionDecl::getODRHash() {
 
 FieldDecl *FieldDecl::Create(const ASTContext &C, DeclContext *DC,
                              SourceLocation StartLoc, SourceLocation IdLoc,
-                             IdentifierInfo *Id, QualType T,
+                             const IdentifierInfo *Id, QualType T,
                              TypeSourceInfo *TInfo, Expr *BW, bool Mutable,
                              InClassInitStyle InitStyle) {
   return new (C, DC) FieldDecl(Decl::Field, DC, StartLoc, IdLoc, Id, T, TInfo,
@@ -5438,7 +5438,7 @@ IndirectFieldDecl::IndirectFieldDecl(ASTContext &C, DeclContext *DC,
 
 IndirectFieldDecl *
 IndirectFieldDecl::Create(ASTContext &C, DeclContext *DC, SourceLocation L,
-                          IdentifierInfo *Id, QualType T,
+                          const IdentifierInfo *Id, QualType T,
                           llvm::MutableArrayRef<NamedDecl *> CH) {
   return new (C, DC) IndirectFieldDecl(C, DC, L, Id, T, CH);
 }
@@ -5461,7 +5461,8 @@ void TypeDecl::anchor() {}
 
 TypedefDecl *TypedefDecl::Create(ASTContext &C, DeclContext *DC,
                                  SourceLocation StartLoc, SourceLocation IdLoc,
-                                 IdentifierInfo *Id, TypeSourceInfo *TInfo) {
+                                 const IdentifierInfo *Id,
+                                 TypeSourceInfo *TInfo) {
   return new (C, DC) TypedefDecl(C, DC, StartLoc, IdLoc, Id, TInfo);
 }
 
@@ -5511,7 +5512,8 @@ TypedefDecl *TypedefDecl::CreateDeserialized(ASTContext &C, unsigned ID) {
 
 TypeAliasDecl *TypeAliasDecl::Create(ASTContext &C, DeclContext *DC,
                                      SourceLocation StartLoc,
-                                     SourceLocation IdLoc, IdentifierInfo *Id,
+                                     SourceLocation IdLoc,
+                                     const IdentifierInfo *Id,
                                      TypeSourceInfo *TInfo) {
   return new (C, DC) TypeAliasDecl(C, DC, StartLoc, IdLoc, Id, TInfo);
 }
diff --git a/clang/lib/AST/DeclObjC.cpp b/clang/lib/AST/DeclObjC.cpp
index 962f503306a0f..32c14938cd588 100644
--- a/clang/lib/AST/DeclObjC.cpp
+++ b/clang/lib/AST/DeclObjC.cpp
@@ -66,7 +66,8 @@ void ObjCProtocolList::set(ObjCProtocolDecl* const* InList, unsigned Elts,
 //===----------------------------------------------------------------------===//
 
 ObjCContainerDecl::ObjCContainerDecl(Kind DK, DeclContext *DC,
-                                     IdentifierInfo *Id, SourceLocation nameLoc,
+                                     const IdentifierInfo *Id,
+                                     SourceLocation nameLoc,
                                      SourceLocation atStartLoc)
     : NamedDecl(DK, DC, nameLoc, Id), DeclContext(DK) {
   setAtStartLoc(atStartLoc);
@@ -378,10 +379,8 @@ SourceLocation ObjCInterfaceDecl::getSuperClassLoc() const {
 /// FindPropertyVisibleInPrimaryClass - Finds declaration of the property
 /// with name 'PropertyId' in the primary class; including those in protocols
 /// (direct or indirect) used by the primary class.
-ObjCPropertyDecl *
-ObjCInterfaceDecl::FindPropertyVisibleInPrimaryClass(
-                       IdentifierInfo *PropertyId,
-                       ObjCPropertyQueryKind QueryKind) const {
+ObjCPropertyDecl *ObjCInterfaceDecl::FindPropertyVisibleInPrimaryClass(
+    const IdentifierInfo *PropertyId, ObjCPropertyQueryKind QueryKind) const {
   // FIXME: Should make sure no callers ever do this.
   if (!hasDefinition())
     return nullptr;
@@ -1539,14 +1538,10 @@ void ObjCTypeParamList::gatherDefaultTypeArgs(
 // ObjCInterfaceDecl
 //===----------------------------------------------------------------------===//
 
-ObjCInterfaceDecl *ObjCInterfaceDecl::Create(const ASTContext &C,
-                                             DeclContext *DC,
-                                             SourceLocation atLoc,
-                                             IdentifierInfo *Id,
-                                             ObjCTypeParamList *typeParamList,
-                                             ObjCInterfaceDecl *PrevDecl,
-                                             SourceLocation ClassLoc,
-                                             bool isInternal){
+ObjCInterfaceDecl *ObjCInterfaceDecl::Create(
+    const ASTContext &C, DeclContext *DC, SourceLocation atLoc,
+    const IdentifierInfo *Id, ObjCTypeParamList *typeParamList,
+    ObjCInterfaceDecl *PrevDecl, SourceLocation ClassLoc, bool isInternal) {
   auto *Result = new (C, DC)
       ObjCInterfaceDecl(C, DC, atLoc, Id, typeParamList, ClassLoc, PrevDecl,
                         isInternal);
@@ -1564,12 +1559,10 @@ ObjCInterfaceDecl *ObjCInterfaceDecl::CreateDeserialized(const ASTContext &C,
   return Result;
 }
 
-ObjCInterfaceDecl::ObjCInterfaceDecl(const ASTContext &C, DeclContext *DC,
-                                     SourceLocation AtLoc, IdentifierInfo *Id,
-                                     ObjCTypeParamList *typeParamList,
-                                     SourceLocation CLoc,
-                                     ObjCInterfaceDecl *PrevDecl,
-                                     bool IsInternal)
+ObjCInterfaceDecl::ObjCInterfaceDecl(
+    const ASTContext &C, DeclContext *DC, SourceLocation AtLoc,
+    const IdentifierInfo *Id, ObjCTypeParamList *typeParamList,
+    SourceLocation CLoc, ObjCInterfaceDecl *PrevDecl, bool IsInternal)
     : ObjCContainerDecl(ObjCInterface, DC, Id, CLoc, AtLoc),
       redeclarable_base(C) {
   setPreviousDecl(PrevDecl);
@@ -1751,8 +1744,8 @@ ObjCIvarDecl *ObjCInterfaceDecl::all_declared_ivar_begin() {
 /// categories for this class and returns it. Name of the category is passed
 /// in 'CategoryId'. If category not found, return 0;
 ///
-ObjCCategoryDecl *
-ObjCInterfaceDecl::FindCategoryDeclaration(IdentifierInfo *CategoryId) const {
+ObjCCategoryDecl *ObjCInterfaceDecl::FindCategoryDeclaration(
+    const IdentifierInfo *CategoryId) const {
   // FIXME: Should make sure no callers ever do this.
   if (!hasDefinition())
     return nullptr;
@@ -1838,10 +1831,10 @@ void ObjCIvarDecl::anchor() {}
 
 ObjCIvarDecl *ObjCIvarDecl::Create(ASTContext &C, ObjCContainerDecl *DC,
                                    SourceLocation StartLoc,
-                                   SourceLocation IdLoc, IdentifierInfo *Id,
-                                   QualType T, TypeSourceInfo *TInfo,
-                                   AccessControl ac, Expr *BW,
-                                   bool synthesized) {
+                                   SourceLocation IdLoc,
+                                   const IdentifierInfo *Id, QualType T,
+                                   TypeSourceInfo *TInfo, AccessControl ac,
+                                   Expr *BW, bool synthesized) {
   if (DC) {
     // Ivar's can only appear in interfaces, implementations (via synthesized
     // properties), and class extensions (via direct declaration, or synthesized
@@ -2120,28 +2113,23 @@ void ObjCProtocolDecl::setHasODRHash(bool HasHash) {
 
 void ObjCCategoryDecl::anchor() {}
 
-ObjCCategoryDecl::ObjCCategoryDecl(DeclContext *DC, SourceLocation AtLoc,
-                                   SourceLocation ClassNameLoc,
-                                   SourceLocation CategoryNameLoc,
-                                   IdentifierInfo *Id, ObjCInterfaceDecl *IDecl,
-                                   ObjCTypeParamList *typeParamList,
-                                   SourceLocation IvarLBraceLoc,
-                                   SourceLocation IvarRBraceLoc)
+ObjCCategoryDecl::ObjCCategoryDecl(
+    DeclContext *DC, SourceLocation AtLoc, SourceLocation ClassNameLoc,
+    SourceLocation CategoryNameLoc, const IdentifierInfo *Id,
+    ObjCInterfaceDecl *IDecl, ObjCTypeParamList *typeParamList,
+    SourceLocation IvarLBraceLoc, SourceLocation IvarRBraceLoc)
     : ObjCContainerDecl(ObjCCategory, DC, Id, ClassNameLoc, AtLoc),
       ClassInterface(IDecl), CategoryNameLoc(CategoryNameLoc),
       IvarLBraceLoc(IvarLBraceLoc), IvarRBraceLoc(IvarRBraceLoc) {
   setTypeParamList(typeParamList);
 }
 
-ObjCCategoryDecl *ObjCCategoryDecl::Create(ASTContext &C, DeclContext *DC,
-                                           SourceLocation AtLoc,
-                                           SourceLocation ClassNameLoc,
-                                           SourceLocation CategoryNameLoc,
-                                           IdentifierInfo *Id,
-                                           ObjCInterfaceDecl *IDecl,
-                                           ObjCTypeParamList *typeParamList,
-                                           SourceLocation IvarLBraceLoc,
-                                           SourceLocation IvarRBraceLoc) {
+ObjCCategoryDecl *ObjCCategoryDecl::Create(
+    ASTContext &C, DeclContext *DC, SourceLocation AtLoc,
+    SourceLocation ClassNameLoc, SourceLocation CategoryNameLoc,
+    const IdentifierInfo *Id, ObjCInterfaceDecl *IDecl,
+    ObjCTypeParamList *typeParamList, SourceLocation IvarLBraceLoc,
+    SourceLocation IvarRBraceLoc) {
   auto *CatDecl =
       new (C, DC) ObjCCategoryDecl(DC, AtLoc, ClassNameLoc, CategoryNameLoc, Id,
                                    IDecl, typeParamList, IvarLBraceLoc,
@@ -2190,13 +2178,10 @@ void ObjCCategoryDecl::setTypeParamList(ObjCTypeParamList *TPL) {
 
 void ObjCCategoryImplDecl::anchor() {}
 
-ObjCCategoryImplDecl *
-ObjCCategoryImplDecl::Create(ASTContext &C, DeclContext *DC,
-                             IdentifierInfo *Id,
-                             ObjCInterfaceDecl *ClassInterface,
-                             SourceLocation nameLoc,
-                             SourceLocation atStartLoc,
-                             SourceLocation CategoryNameLoc) {
+ObjCCategoryImplDecl *ObjCCategoryImplDecl::Create(
+    ASTContext &C, DeclContext *DC, const IdentifierInfo *Id,
+    ObjCInterfaceDecl *ClassInterface, SourceLocation nameLoc,
+    SourceLocation atStartLoc, SourceLocation CategoryNameLoc) {
   if (ClassInterface && ClassInterface->hasDefinition())
     ClassInterface = ClassInterface->getDefinition();
   return new (C, DC) ObjCCategoryImplDecl(DC, Id, ClassInterface, nameLoc,
@@ -2365,14 +2350,11 @@ ObjCCompatibleAliasDecl::CreateDeserialized(ASTContext &C, unsigned ID) {
 
 void ObjCPropertyDecl::anchor() {}
 
-ObjCPropertyDecl *ObjCPropertyDecl::Create(ASTContext &C, DeclContext *DC,
-                                           SourceLocation L,
-                                           IdentifierInfo *Id,
-                                           SourceLocation AtLoc,
-                                           SourceLocation LParenLoc,
-                                           QualType T,
-                                           TypeSourceInfo *TSI,
-                                           PropertyControl propControl) {
+ObjCPropertyDecl *
+ObjCPropertyDecl::Create(ASTContext &C, DeclContext *DC, SourceLocation L,
+                         const IdentifierInfo *Id, SourceLocation AtLoc,
+                         SourceLocation LParenLoc, QualType T,
+                         TypeSourceInfo *TSI, PropertyControl propControl) {
   return new (C, DC) ObjCPropertyDecl(DC, L, Id, AtLoc, LParenLoc, T, TSI,
                                       propControl);
 }
diff --git a/clang/lib/AST/DeclTemplate.cpp b/clang/lib/AST/DeclTemplate.cpp
index 3c217d6a6a5ae..571ed81a42e40 100644
--- a/clang/lib/AST/DeclTemplate.cpp
+++ b/clang/lib/AST/DeclTemplate.cpp
@@ -715,7 +715,7 @@ void TemplateTypeParmDecl::setTypeConstraint(
 
 NonTypeTemplateParmDecl::NonTypeTemplateParmDecl(
     DeclContext *DC, SourceLocation StartLoc, SourceLocation IdLoc, unsigned D,
-    unsigned P, IdentifierInfo *Id, QualType T, TypeSourceInfo *TInfo,
+    unsigned P, const IdentifierInfo *Id, QualType T, TypeSourceInfo *TInfo,
     ArrayRef<QualType> ExpandedTypes, ArrayRef<TypeSourceInfo *> ExpandedTInfos)
     : DeclaratorDecl(NonTypeTemplateParm, DC, IdLoc, Id, T, TInfo, StartLoc),
       TemplateParmPosition(D, P), ParameterPack(true),
@@ -730,12 +730,10 @@ NonTypeTemplateParmDecl::NonTypeTemplateParmDecl(
   }
 }
 
-NonTypeTemplateParmDecl *
-NonTypeTemplateParmDecl::Create(const ASTContext &C, DeclContext *DC,
-                                SourceLocation StartLoc, SourceLocation IdLoc,
-                                unsigned D, unsigned P, IdentifierInfo *Id,
-                                QualType T, bool ParameterPack,
-                                TypeSourceInfo *TInfo) {
+NonTypeTemplateParmDecl *NonTypeTemplateParmDecl::Create(
+    const ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
+    SourceLocation IdLoc, unsigned D, unsigned P, const IdentifierInfo *Id,
+    QualType T, bool ParameterPack, TypeSourceInfo *TInfo) {
   AutoType *AT =
       C.getLangOpts().CPlusPlus20 ? T->getContainedAutoType() : nullptr;
   return new (C, DC,
@@ -748,7 +746,7 @@ NonTypeTemplateParmDecl::Create(const ASTContext &C, DeclContext *DC,
 
 NonTypeTemplateParmDecl *NonTypeTemplateParmDecl::Create(
     const ASTContext &C, DeclContext *DC, SourceLocation StartLoc,
-    SourceLocation IdLoc, unsigned D, unsigned P, IdentifierInfo *Id,
+    SourceLocation IdLoc, unsigned D, unsigned P, const IdentifierInfo *Id,
     QualType T, TypeSourceInfo *TInfo, ArrayRef<QualType> ExpandedTypes,
     ArrayRef<TypeSourceInfo *> ExpandedTInfos) {
   AutoType *AT = TInfo->getType()->getContainedAutoType();
diff --git a/clang/lib/AST/NSAPI.cpp b/clang/lib/AST/NSAPI.cpp
index 86dee540e9e29..ecc56c13fb757 100644
--- a/clang/lib/AST/NSAPI.cpp
+++ b/clang/lib/AST/NSAPI.cpp
@@ -56,10 +56,8 @@ Selector NSAPI::getNSStringSelector(NSStringMethodKind MK) const {
                                        &Ctx.Idents.get("initWithUTF8String"));
       break;
     case NSStr_stringWithCStringEncoding: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("stringWithCString"),
-        &Ctx.Idents.get("encoding")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("stringWithCString"),
+                                           &Ctx.Idents.get("encoding")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -93,10 +91,8 @@ Selector NSAPI::getNSArraySelector(NSArrayMethodKind MK) const {
       Sel = Ctx.Selectors.getUnarySelector(&Ctx.Idents.get("arrayWithObjects"));
       break;
     case NSArr_arrayWithObjectsCount: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("arrayWithObjects"),
-        &Ctx.Idents.get("count")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("arrayWithObjects"),
+                                           &Ctx.Idents.get("count")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -110,10 +106,9 @@ Selector NSAPI::getNSArraySelector(NSArrayMethodKind MK) const {
       Sel = Ctx.Selectors.getUnarySelector(&Ctx.Idents.get("objectAtIndex"));
       break;
     case NSMutableArr_replaceObjectAtIndex: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("replaceObjectAtIndex"),
-        &Ctx.Idents.get("withObject")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("replaceObjectAtIndex"),
+          &Ctx.Idents.get("withObject")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -121,18 +116,14 @@ Selector NSAPI::getNSArraySelector(NSArrayMethodKind MK) const {
       Sel = Ctx.Selectors.getUnarySelector(&Ctx.Idents.get("addObject"));
       break;
     case NSMutableArr_insertObjectAtIndex: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("insertObject"),
-        &Ctx.Idents.get("atIndex")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("insertObject"),
+                                           &Ctx.Idents.get("atIndex")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSMutableArr_setObjectAtIndexedSubscript: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setObject"),
-        &Ctx.Idents.get("atIndexedSubscript")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("setObject"), &Ctx.Idents.get("atIndexedSubscript")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -167,27 +158,21 @@ Selector NSAPI::getNSDictionarySelector(
                                    &Ctx.Idents.get("dictionaryWithDictionary"));
       break;
     case NSDict_dictionaryWithObjectForKey: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("dictionaryWithObject"),
-        &Ctx.Idents.get("forKey")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("dictionaryWithObject"), &Ctx.Idents.get("forKey")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSDict_dictionaryWithObjectsForKeys: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("dictionaryWithObjects"),
-        &Ctx.Idents.get("forKeys")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("dictionaryWithObjects"), &Ctx.Idents.get("forKeys")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSDict_dictionaryWithObjectsForKeysCount: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("dictionaryWithObjects"),
-        &Ctx.Idents.get("forKeys"),
-        &Ctx.Idents.get("count")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("dictionaryWithObjects"), &Ctx.Idents.get("forKeys"),
+          &Ctx.Idents.get("count")};
       Sel = Ctx.Selectors.getSelector(3, KeyIdents);
       break;
     }
@@ -204,10 +189,8 @@ Selector NSAPI::getNSDictionarySelector(
                                      &Ctx.Idents.get("initWithObjectsAndKeys"));
       break;
     case NSDict_initWithObjectsForKeys: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("initWithObjects"),
-        &Ctx.Idents.get("forKeys")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("initWithObjects"),
+                                           &Ctx.Idents.get("forKeys")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -215,26 +198,20 @@ Selector NSAPI::getNSDictionarySelector(
       Sel = Ctx.Selectors.getUnarySelector(&Ctx.Idents.get("objectForKey"));
       break;
     case NSMutableDict_setObjectForKey: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setObject"),
-        &Ctx.Idents.get("forKey")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("setObject"),
+                                           &Ctx.Idents.get("forKey")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSMutableDict_setObjectForKeyedSubscript: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setObject"),
-        &Ctx.Idents.get("forKeyedSubscript")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("setObject"), &Ctx.Idents.get("forKeyedSubscript")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSMutableDict_setValueForKey: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setValue"),
-        &Ctx.Idents.get("forKey")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("setValue"),
+                                           &Ctx.Idents.get("forKey")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -264,34 +241,27 @@ Selector NSAPI::getNSSetSelector(NSSetMethodKind MK) const {
       Sel = Ctx.Selectors.getUnarySelector(&Ctx.Idents.get("addObject"));
       break;
     case NSOrderedSet_insertObjectAtIndex: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("insertObject"),
-        &Ctx.Idents.get("atIndex")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("insertObject"),
+                                           &Ctx.Idents.get("atIndex")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSOrderedSet_setObjectAtIndex: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setObject"),
-        &Ctx.Idents.get("atIndex")
-      };
+      const IdentifierInfo *KeyIdents[] = {&Ctx.Idents.get("setObject"),
+                                           &Ctx.Idents.get("atIndex")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSOrderedSet_setObjectAtIndexedSubscript: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("setObject"),
-        &Ctx.Idents.get("atIndexedSubscript")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("setObject"), &Ctx.Idents.get("atIndexedSubscript")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
     case NSOrderedSet_replaceObjectAtIndexWithObject: {
-      IdentifierInfo *KeyIdents[] = {
-        &Ctx.Idents.get("replaceObjectAtIndex"),
-        &Ctx.Idents.get("withObject")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &Ctx.Idents.get("replaceObjectAtIndex"),
+          &Ctx.Idents.get("withObject")};
       Sel = Ctx.Selectors.getSelector(2, KeyIdents);
       break;
     }
@@ -606,7 +576,7 @@ bool NSAPI::isObjCEnumerator(const Expr *E,
 Selector NSAPI::getOrInitSelector(ArrayRef<StringRef> Ids,
                                   Selector &Sel) const {
   if (Sel.isNull()) {
-    SmallVector<IdentifierInfo *, 4> Idents;
+    SmallVector<const IdentifierInfo *, 4> Idents;
     for (ArrayRef<StringRef>::const_iterator
            I = Ids.begin(), E = Ids.end(); I != E; ++I)
       Idents.push_back(&Ctx.Idents.get(*I));
@@ -617,7 +587,7 @@ Selector NSAPI::getOrInitSelector(ArrayRef<StringRef> Ids,
 
 Selector NSAPI::getOrInitNullarySelector(StringRef Id, Selector &Sel) const {
   if (Sel.isNull()) {
-    IdentifierInfo *Ident = &Ctx.Idents.get(Id);
+    const IdentifierInfo *Ident = &Ctx.Idents.get(Id);
     Sel = Ctx.Selectors.getSelector(0, &Ident);
   }
   return Sel;
diff --git a/clang/lib/AST/NestedNameSpecifier.cpp b/clang/lib/AST/NestedNameSpecifier.cpp
index 36f2c47b30005..785c46e86a77c 100644
--- a/clang/lib/AST/NestedNameSpecifier.cpp
+++ b/clang/lib/AST/NestedNameSpecifier.cpp
@@ -55,16 +55,16 @@ NestedNameSpecifier::FindOrInsert(const ASTContext &Context,
   return NNS;
 }
 
-NestedNameSpecifier *
-NestedNameSpecifier::Create(const ASTContext &Context,
-                            NestedNameSpecifier *Prefix, IdentifierInfo *II) {
+NestedNameSpecifier *NestedNameSpecifier::Create(const ASTContext &Context,
+                                                 NestedNameSpecifier *Prefix,
+                                                 const IdentifierInfo *II) {
   assert(II && "Identifier cannot be NULL");
   assert((!Prefix || Prefix->isDependent()) && "Prefix must be dependent");
 
   NestedNameSpecifier Mockup;
   Mockup.Prefix.setPointer(Prefix);
   Mockup.Prefix.setInt(StoredIdentifier);
-  Mockup.Specifier = II;
+  Mockup.Specifier = const_cast<IdentifierInfo *>(II);
   return FindOrInsert(Context, Mockup);
 }
 
@@ -87,7 +87,7 @@ NestedNameSpecifier::Create(const ASTContext &Context,
 NestedNameSpecifier *
 NestedNameSpecifier::Create(const ASTContext &Context,
                             NestedNameSpecifier *Prefix,
-                            NamespaceAliasDecl *Alias) {
+                            const NamespaceAliasDecl *Alias) {
   assert(Alias && "Namespace alias cannot be NULL");
   assert((!Prefix ||
           (Prefix->getAsType() == nullptr &&
@@ -96,7 +96,7 @@ NestedNameSpecifier::Create(const ASTContext &Context,
   NestedNameSpecifier Mockup;
   Mockup.Prefix.setPointer(Prefix);
   Mockup.Prefix.setInt(StoredDecl);
-  Mockup.Specifier = Alias;
+  Mockup.Specifier = const_cast<NamespaceAliasDecl *>(Alias);
   return FindOrInsert(Context, Mockup);
 }
 
@@ -112,13 +112,13 @@ NestedNameSpecifier::Create(const ASTContext &Context,
   return FindOrInsert(Context, Mockup);
 }
 
-NestedNameSpecifier *
-NestedNameSpecifier::Create(const ASTContext &Context, IdentifierInfo *II) {
+NestedNameSpecifier *NestedNameSpecifier::Create(const ASTContext &Context,
+                                                 const IdentifierInfo *II) {
   assert(II && "Identifier cannot be NULL");
   NestedNameSpecifier Mockup;
   Mockup.Prefix.setPointer(nullptr);
   Mockup.Prefix.setInt(StoredIdentifier);
-  Mockup.Specifier = II;
+  Mockup.Specifier = const_cast<IdentifierInfo *>(II);
   return FindOrInsert(Context, Mockup);
 }
 
diff --git a/clang/lib/AST/SelectorLocationsKind.cpp b/clang/lib/AST/SelectorLocationsKind.cpp
index 2c34c9c60c2b2..ebe6324f904c7 100644
--- a/clang/lib/AST/SelectorLocationsKind.cpp
+++ b/clang/lib/AST/SelectorLocationsKind.cpp
@@ -26,7 +26,7 @@ static SourceLocation getStandardSelLoc(unsigned Index,
     assert(Index == 0);
     if (EndLoc.isInvalid())
       return SourceLocation();
-    IdentifierInfo *II = Sel.getIdentifierInfoForSlot(0);
+    const IdentifierInfo *II = Sel.getIdentifierInfoForSlot(0);
     unsigned Len = II ? II->getLength() : 0;
     return EndLoc.getLocWithOffset(-Len);
   }
@@ -34,7 +34,7 @@ static SourceLocation getStandardSelLoc(unsigned Index,
   assert(Index < NumSelArgs);
   if (ArgLoc.isInvalid())
     return SourceLocation();
-  IdentifierInfo *II = Sel.getIdentifierInfoForSlot(Index);
+  const IdentifierInfo *II = Sel.getIdentifierInfoForSlot(Index);
   unsigned Len = /* selector id */ (II ? II->getLength() : 0) + /* ':' */ 1;
   if (WithArgSpace)
     ++Len;
diff --git a/clang/lib/AST/StmtPrinter.cpp b/clang/lib/AST/StmtPrinter.cpp
index 2ba93d17f2675..5855ab3141edc 100644
--- a/clang/lib/AST/StmtPrinter.cpp
+++ b/clang/lib/AST/StmtPrinter.cpp
@@ -1447,7 +1447,7 @@ void StmtPrinter::VisitOffsetOfExpr(OffsetOfExpr *Node) {
       continue;
 
     // Field or identifier node.
-    IdentifierInfo *Id = ON.getFieldName();
+    const IdentifierInfo *Id = ON.getFieldName();
     if (!Id)
       continue;
 
@@ -2348,7 +2348,7 @@ void StmtPrinter::VisitCXXPseudoDestructorExpr(CXXPseudoDestructorExpr *E) {
     E->getQualifier()->print(OS, Policy);
   OS << "~";
 
-  if (IdentifierInfo *II = E->getDestroyedTypeIdentifier())
+  if (const IdentifierInfo *II = E->getDestroyedTypeIdentifier())
     OS << II->getName();
   else
     E->getDestroyedType().print(OS, Policy);
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index be3dd4b673cf9..01e1d1cc8289b 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -61,7 +61,7 @@ namespace {
     virtual void VisitName(DeclarationName Name, bool TreatAsDecl = false) = 0;
 
     /// Visit identifiers that are not in Decl's or Type's.
-    virtual void VisitIdentifierInfo(IdentifierInfo *II) = 0;
+    virtual void VisitIdentifierInfo(const IdentifierInfo *II) = 0;
 
     /// Visit a nested-name-specifier that occurs within an expression
     /// or statement.
@@ -163,7 +163,7 @@ namespace {
       ID.AddPointer(Name.getAsOpaquePtr());
     }
 
-    void VisitIdentifierInfo(IdentifierInfo *II) override {
+    void VisitIdentifierInfo(const IdentifierInfo *II) override {
       ID.AddPointer(II);
     }
 
@@ -211,7 +211,7 @@ namespace {
       }
       Hash.AddDeclarationName(Name, TreatAsDecl);
     }
-    void VisitIdentifierInfo(IdentifierInfo *II) override {
+    void VisitIdentifierInfo(const IdentifierInfo *II) override {
       ID.AddBoolean(II);
       if (II) {
         Hash.AddIdentifierInfo(II);
diff --git a/clang/lib/Analysis/ObjCNoReturn.cpp b/clang/lib/Analysis/ObjCNoReturn.cpp
index 9d7c365c3b992..9e651c29e085d 100644
--- a/clang/lib/Analysis/ObjCNoReturn.cpp
+++ b/clang/lib/Analysis/ObjCNoReturn.cpp
@@ -17,7 +17,8 @@
 
 using namespace clang;
 
-static bool isSubclass(const ObjCInterfaceDecl *Class, IdentifierInfo *II) {
+static bool isSubclass(const ObjCInterfaceDecl *Class,
+                       const IdentifierInfo *II) {
   if (!Class)
     return false;
   if (Class->getIdentifier() == II)
@@ -30,7 +31,7 @@ ObjCNoReturn::ObjCNoReturn(ASTContext &C)
     NSExceptionII(&C.Idents.get("NSException"))
 {
   // Generate selectors.
-  SmallVector<IdentifierInfo*, 3> II;
+  SmallVector<const IdentifierInfo *, 3> II;
 
   // raise:format:
   II.push_back(&C.Idents.get("raise"));
diff --git a/clang/lib/Basic/IdentifierTable.cpp b/clang/lib/Basic/IdentifierTable.cpp
index a9b07aca65c05..feea84544d62f 100644
--- a/clang/lib/Basic/IdentifierTable.cpp
+++ b/clang/lib/Basic/IdentifierTable.cpp
@@ -541,7 +541,8 @@ unsigned Selector::getNumArgs() const {
   return SI->getNumArgs();
 }
 
-IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
+const IdentifierInfo *
+Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
   if (getIdentifierInfoFlag() < MultiArg) {
     assert(argIndex == 0 && "illegal keyword index");
     return getAsIdentifierInfo();
@@ -553,7 +554,7 @@ IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
 }
 
 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
-  IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
+  const IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
   return II ? II->getName() : StringRef();
 }
 
@@ -574,7 +575,7 @@ std::string Selector::getAsString() const {
     return "<null selector>";
 
   if (getIdentifierInfoFlag() < MultiArg) {
-    IdentifierInfo *II = getAsIdentifierInfo();
+    const IdentifierInfo *II = getAsIdentifierInfo();
 
     if (getNumArgs() == 0) {
       assert(II && "If the number of arguments is 0 then II is guaranteed to "
@@ -608,7 +609,7 @@ static bool startsWithWord(StringRef name, StringRef word) {
 }
 
 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
-  IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
+  const IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
   if (!first) return OMF_None;
 
   StringRef name = first->getName();
@@ -655,7 +656,7 @@ ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
 }
 
 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
-  IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
+  const IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
   if (!first) return OIT_None;
 
   StringRef name = first->getName();
@@ -683,7 +684,7 @@ ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
 }
 
 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
-  IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
+  const IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
   if (!first) return SFF_None;
 
   StringRef name = first->getName();
@@ -750,7 +751,8 @@ size_t SelectorTable::getTotalMemory() const {
   return SelTabImpl.Allocator.getTotalMemory();
 }
 
-Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
+Selector SelectorTable::getSelector(unsigned nKeys,
+                                    const IdentifierInfo **IIV) {
   if (nKeys < 2)
     return Selector(IIV[0], nKeys);
 
diff --git a/clang/lib/CodeGen/CGBlocks.cpp b/clang/lib/CodeGen/CGBlocks.cpp
index 47f063b5501cc..2742c39965b2c 100644
--- a/clang/lib/CodeGen/CGBlocks.cpp
+++ b/clang/lib/CodeGen/CGBlocks.cpp
@@ -1447,7 +1447,7 @@ llvm::Function *CodeGenFunction::GenerateBlockFunction(
     selfTy = getContext().getPointerType(getContext().getAddrSpaceQualType(
         getContext().VoidTy, LangAS::opencl_generic));
 
-  IdentifierInfo *II = &CGM.getContext().Idents.get(".block_descriptor");
+  const IdentifierInfo *II = &CGM.getContext().Idents.get(".block_descriptor");
 
   ImplicitParamDecl SelfDecl(getContext(), const_cast<BlockDecl *>(blockDecl),
                              SourceLocation(), II, selfTy,
@@ -2791,7 +2791,7 @@ static void configureBlocksRuntimeObject(CodeGenModule &CGM,
   auto *GV = cast<llvm::GlobalValue>(C->stripPointerCasts());
 
   if (CGM.getTarget().getTriple().isOSBinFormatCOFF()) {
-    IdentifierInfo &II = CGM.getContext().Idents.get(C->getName());
+    const IdentifierInfo &II = CGM.getContext().Idents.get(C->getName());
     TranslationUnitDecl *TUDecl = CGM.getContext().getTranslationUnitDecl();
     DeclContext *DC = TranslationUnitDecl::castToDeclContext(TUDecl);
 
diff --git a/clang/lib/CodeGen/CGCUDANV.cpp b/clang/lib/CodeGen/CGCUDANV.cpp
index 0cb5b06a519c0..370642cb3d536 100644
--- a/clang/lib/CodeGen/CGCUDANV.cpp
+++ b/clang/lib/CodeGen/CGCUDANV.cpp
@@ -361,7 +361,7 @@ void CGNVCUDARuntime::emitDeviceStubBodyNew(CodeGenFunction &CGF,
       KernelLaunchAPI = KernelLaunchAPI + "_ptsz";
   }
   auto LaunchKernelName = addPrefixToName(KernelLaunchAPI);
-  IdentifierInfo &cudaLaunchKernelII =
+  const IdentifierInfo &cudaLaunchKernelII =
       CGM.getContext().Idents.get(LaunchKernelName);
   FunctionDecl *cudaLaunchKernelFD = nullptr;
   for (auto *Result : DC->lookup(&cudaLaunchKernelII)) {
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 4f4013292b1fc..8bdafa7c569b0 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1384,7 +1384,7 @@ void CodeGenFunction::EmitAndRegisterVariableArrayDimensions(
   // For each dimension stores its QualType and corresponding
   // size-expression Value.
   SmallVector<CodeGenFunction::VlaSizePair, 4> Dimensions;
-  SmallVector<IdentifierInfo *, 4> VLAExprNames;
+  SmallVector<const IdentifierInfo *, 4> VLAExprNames;
 
   // Break down the array into individual dimensions.
   QualType Type1D = D.getType();
@@ -1421,7 +1421,7 @@ void CodeGenFunction::EmitAndRegisterVariableArrayDimensions(
       MD = llvm::ConstantAsMetadata::get(C);
     else {
       // Create an artificial VarDecl to generate debug info for.
-      IdentifierInfo *NameIdent = VLAExprNames[NameIdx++];
+      const IdentifierInfo *NameIdent = VLAExprNames[NameIdx++];
       auto QT = getContext().getIntTypeForBitwidth(
           SizeTy->getScalarSizeInBits(), false);
       auto *ArtificialDecl = VarDecl::Create(
diff --git a/clang/lib/CodeGen/CGObjC.cpp b/clang/lib/CodeGen/CGObjC.cpp
index c7f497a7c8451..ee571995ce4c3 100644
--- a/clang/lib/CodeGen/CGObjC.cpp
+++ b/clang/lib/CodeGen/CGObjC.cpp
@@ -1789,11 +1789,10 @@ void CodeGenFunction::EmitObjCForCollectionStmt(const ObjCForCollectionStmt &S){
   static const unsigned NumItems = 16;
 
   // Fetch the countByEnumeratingWithState:objects:count: selector.
-  IdentifierInfo *II[] = {
-    &CGM.getContext().Idents.get("countByEnumeratingWithState"),
-    &CGM.getContext().Idents.get("objects"),
-    &CGM.getContext().Idents.get("count")
-  };
+  const IdentifierInfo *II[] = {
+      &CGM.getContext().Idents.get("countByEnumeratingWithState"),
+      &CGM.getContext().Idents.get("objects"),
+      &CGM.getContext().Idents.get("count")};
   Selector FastEnumSel =
       CGM.getContext().Selectors.getSelector(std::size(II), &II[0]);
 
@@ -2720,7 +2719,7 @@ llvm::Value *CodeGenFunction::EmitObjCMRRAutoreleasePoolPush() {
   CGObjCRuntime &Runtime = CGM.getObjCRuntime();
   llvm::Value *Receiver = Runtime.EmitNSAutoreleasePoolClassRef(*this);
   // [NSAutoreleasePool alloc]
-  IdentifierInfo *II = &CGM.getContext().Idents.get("alloc");
+  const IdentifierInfo *II = &CGM.getContext().Idents.get("alloc");
   Selector AllocSel = getContext().Selectors.getSelector(0, &II);
   CallArgList Args;
   RValue AllocRV =
@@ -2767,7 +2766,7 @@ llvm::Value *CodeGenFunction::EmitObjCAllocInit(llvm::Value *value,
 /// Produce the code to do a primitive release.
 /// [tmp drain];
 void CodeGenFunction::EmitObjCMRRAutoreleasePoolPop(llvm::Value *Arg) {
-  IdentifierInfo *II = &CGM.getContext().Idents.get("drain");
+  const IdentifierInfo *II = &CGM.getContext().Idents.get("drain");
   Selector DrainSel = getContext().Selectors.getSelector(0, &II);
   CallArgList Args;
   CGM.getObjCRuntime().GenerateMessageSend(*this, ReturnValueSlot(),
@@ -3715,8 +3714,8 @@ CodeGenFunction::GenerateObjCAtomicSetterCopyHelperFunction(
   if ((HelperFn = CGM.getAtomicSetterHelperFnMap(Ty)))
     return HelperFn;
 
-  IdentifierInfo *II
-    = &CGM.getContext().Idents.get("__assign_helper_atomic_property_");
+  const IdentifierInfo *II =
+      &CGM.getContext().Idents.get("__assign_helper_atomic_property_");
 
   QualType ReturnTy = C.VoidTy;
   QualType DestTy = C.getPointerType(Ty);
@@ -3813,7 +3812,7 @@ llvm::Constant *CodeGenFunction::GenerateObjCAtomicGetterCopyHelperFunction(
   if ((HelperFn = CGM.getAtomicGetterHelperFnMap(Ty)))
     return HelperFn;
 
-  IdentifierInfo *II =
+  const IdentifierInfo *II =
       &CGM.getContext().Idents.get("__copy_helper_atomic_property_");
 
   QualType ReturnTy = C.VoidTy;
@@ -3907,10 +3906,10 @@ llvm::Constant *CodeGenFunction::GenerateObjCAtomicGetterCopyHelperFunction(
 llvm::Value *
 CodeGenFunction::EmitBlockCopyAndAutorelease(llvm::Value *Block, QualType Ty) {
   // Get selectors for retain/autorelease.
-  IdentifierInfo *CopyID = &getContext().Idents.get("copy");
+  const IdentifierInfo *CopyID = &getContext().Idents.get("copy");
   Selector CopySelector =
       getContext().Selectors.getNullarySelector(CopyID);
-  IdentifierInfo *AutoreleaseID = &getContext().Idents.get("autorelease");
+  const IdentifierInfo *AutoreleaseID = &getContext().Idents.get("autorelease");
   Selector AutoreleaseSelector =
       getContext().Selectors.getNullarySelector(AutoreleaseID);
 
diff --git a/clang/lib/CodeGen/CGObjCMac.cpp b/clang/lib/CodeGen/CGObjCMac.cpp
index 8a599c10e1caf..042cd5d46da4b 100644
--- a/clang/lib/CodeGen/CGObjCMac.cpp
+++ b/clang/lib/CodeGen/CGObjCMac.cpp
@@ -1555,12 +1555,12 @@ class CGObjCNonFragileABIMac : public CGObjCCommonMac {
 
   // Shamelessly stolen from Analysis/CFRefCount.cpp
   Selector GetNullarySelector(const char* name) const {
-    IdentifierInfo* II = &CGM.getContext().Idents.get(name);
+    const IdentifierInfo *II = &CGM.getContext().Idents.get(name);
     return CGM.getContext().Selectors.getSelector(0, &II);
   }
 
   Selector GetUnarySelector(const char* name) const {
-    IdentifierInfo* II = &CGM.getContext().Idents.get(name);
+    const IdentifierInfo *II = &CGM.getContext().Idents.get(name);
     return CGM.getContext().Selectors.getSelector(1, &II);
   }
 
@@ -6268,11 +6268,10 @@ bool CGObjCNonFragileABIMac::isVTableDispatchedSelector(Selector Sel) {
       VTableDispatchMethods.insert(GetUnarySelector("addObject"));
 
       // "countByEnumeratingWithState:objects:count"
-      IdentifierInfo *KeyIdents[] = {
-        &CGM.getContext().Idents.get("countByEnumeratingWithState"),
-        &CGM.getContext().Idents.get("objects"),
-        &CGM.getContext().Idents.get("count")
-      };
+      const IdentifierInfo *KeyIdents[] = {
+          &CGM.getContext().Idents.get("countByEnumeratingWithState"),
+          &CGM.getContext().Idents.get("objects"),
+          &CGM.getContext().Idents.get("count")};
       VTableDispatchMethods.insert(
         CGM.getContext().Selectors.getSelector(3, KeyIdents));
     }
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index a2d746bb8f4f9..87766a758311d 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -828,7 +828,7 @@ void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
   // .cxx_destruct, __destroy_helper_block_ and all of their calees at run time.
   if (SanOpts.has(SanitizerKind::Thread)) {
     if (const auto *OMD = dyn_cast_or_null<ObjCMethodDecl>(D)) {
-      IdentifierInfo *II = OMD->getSelector().getIdentifierInfoForSlot(0);
+      const IdentifierInfo *II = OMD->getSelector().getIdentifierInfoForSlot(0);
       if (OMD->getMethodFamily() == OMF_dealloc ||
           OMD->getMethodFamily() == OMF_initialize ||
           (OMD->getSelector().isUnarySelector() && II->isStr(".cxx_destruct"))) {
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 75519be8bba05..b15031dca4686 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -6626,7 +6626,7 @@ static bool AllTrivialInitializers(CodeGenModule &CGM,
 void CodeGenModule::EmitObjCIvarInitializations(ObjCImplementationDecl *D) {
   // We might need a .cxx_destruct even if we don't have any ivar initializers.
   if (needsDestructMethod(D)) {
-    IdentifierInfo *II = &getContext().Idents.get(".cxx_destruct");
+    const IdentifierInfo *II = &getContext().Idents.get(".cxx_destruct");
     Selector cxxSelector = getContext().Selectors.getSelector(0, &II);
     ObjCMethodDecl *DTORMethod = ObjCMethodDecl::Create(
         getContext(), D->getLocation(), D->getLocation(), cxxSelector,
@@ -6646,7 +6646,7 @@ void CodeGenModule::EmitObjCIvarInitializations(ObjCImplementationDecl *D) {
       AllTrivialInitializers(*this, D))
     return;
 
-  IdentifierInfo *II = &getContext().Idents.get(".cxx_construct");
+  const IdentifierInfo *II = &getContext().Idents.get(".cxx_construct");
   Selector cxxSelector = getContext().Selectors.getSelector(0, &II);
   // The constructor returns 'self'.
   ObjCMethodDecl *CTORMethod = ObjCMethodDecl::Create(
@@ -7214,7 +7214,7 @@ void CodeGenModule::EmitStaticExternCAliases() {
   if (!getTargetCodeGenInfo().shouldEmitStaticExternCAliases())
     return;
   for (auto &I : StaticExternCValues) {
-    IdentifierInfo *Name = I.first;
+    const IdentifierInfo *Name = I.first;
     llvm::GlobalValue *Val = I.second;
 
     // If Val is null, that implies there were multiple declarations that each
diff --git a/clang/lib/Frontend/Rewrite/RewriteModernObjC.cpp b/clang/lib/Frontend/Rewrite/RewriteModernObjC.cpp
index 1f40db785981d..6ae955a2380b7 100644
--- a/clang/lib/Frontend/Rewrite/RewriteModernObjC.cpp
+++ b/clang/lib/Frontend/Rewrite/RewriteModernObjC.cpp
@@ -592,7 +592,7 @@ namespace {
     }
 
     bool ImplementationIsNonLazy(const ObjCImplDecl *OD) const {
-      IdentifierInfo* II = &Context->Idents.get("load");
+      const IdentifierInfo *II = &Context->Idents.get("load");
       Selector LoadSel = Context->Selectors.getSelector(0, &II);
       return OD->getClassMethod(LoadSel) != nullptr;
     }
diff --git a/clang/lib/Lex/HeaderSearch.cpp b/clang/lib/Lex/HeaderSearch.cpp
index 7dffcf0e941e0..f0750e5336b6a 100644
--- a/clang/lib/Lex/HeaderSearch.cpp
+++ b/clang/lib/Lex/HeaderSearch.cpp
@@ -64,8 +64,7 @@ HeaderFileInfo::getControllingMacro(ExternalPreprocessorSource *External) {
     if (ControllingMacro->isOutOfDate()) {
       assert(External && "We must have an external source if we have a "
                          "controlling macro that is out of date.");
-      External->updateOutOfDateIdentifier(
-          *const_cast<IdentifierInfo *>(ControllingMacro));
+      External->updateOutOfDateIdentifier(*ControllingMacro);
     }
     return ControllingMacro;
   }
diff --git a/clang/lib/Lex/MacroInfo.cpp b/clang/lib/Lex/MacroInfo.cpp
index 39bb0f44eff25..dfdf463665f3c 100644
--- a/clang/lib/Lex/MacroInfo.cpp
+++ b/clang/lib/Lex/MacroInfo.cpp
@@ -257,7 +257,7 @@ LLVM_DUMP_METHOD void MacroDirective::dump() const {
 }
 
 ModuleMacro *ModuleMacro::create(Preprocessor &PP, Module *OwningModule,
-                                 IdentifierInfo *II, MacroInfo *Macro,
+                                 const IdentifierInfo *II, MacroInfo *Macro,
                                  ArrayRef<ModuleMacro *> Overrides) {
   void *Mem = PP.getPreprocessorAllocator().Allocate(
       sizeof(ModuleMacro) + sizeof(ModuleMacro *) * Overrides.size(),
diff --git a/clang/lib/Lex/PPLexerChange.cpp b/clang/lib/Lex/PPLexerChange.cpp
index 3b1b6df1dbae4..2ca2122ac7109 100644
--- a/clang/lib/Lex/PPLexerChange.cpp
+++ b/clang/lib/Lex/PPLexerChange.cpp
@@ -368,8 +368,7 @@ bool Preprocessor::HandleEndOfFile(Token &Result, bool isEndOfMacro) {
       // Okay, this has a controlling macro, remember in HeaderFileInfo.
       if (OptionalFileEntryRef FE = CurPPLexer->getFileEntry()) {
         HeaderInfo.SetFileControllingMacro(*FE, ControllingMacro);
-        if (MacroInfo *MI =
-              getMacroInfo(const_cast<IdentifierInfo*>(ControllingMacro)))
+        if (MacroInfo *MI = getMacroInfo(ControllingMacro))
           MI->setUsedForHeaderGuard(true);
         if (const IdentifierInfo *DefinedMacro =
               CurPPLexer->MIOpt.GetDefinedMacro()) {
@@ -805,7 +804,7 @@ Module *Preprocessor::LeaveSubmodule(bool ForPragma) {
   llvm::SmallPtrSet<const IdentifierInfo*, 8> VisitedMacros;
   for (unsigned I = Info.OuterPendingModuleMacroNames;
        I != PendingModuleMacroNames.size(); ++I) {
-    auto *II = const_cast<IdentifierInfo*>(PendingModuleMacroNames[I]);
+    const auto *II = PendingModuleMacroNames[I];
     if (!VisitedMacros.insert(II).second)
       continue;
 
@@ -855,8 +854,8 @@ Module *Preprocessor::LeaveSubmodule(bool ForPragma) {
         // Don't bother creating a module macro if it would represent a #undef
         // that doesn't override anything.
         if (Def || !Macro.getOverriddenMacros().empty())
-          addModuleMacro(LeavingMod, II, Def,
-                         Macro.getOverriddenMacros(), IsNew);
+          addModuleMacro(LeavingMod, II, Def, Macro.getOverriddenMacros(),
+                         IsNew);
 
         if (!getLangOpts().ModulesLocalVisibility) {
           // This macro is exposed to the rest of this compilation as a
diff --git a/clang/lib/Lex/PPMacroExpansion.cpp b/clang/lib/Lex/PPMacroExpansion.cpp
index 516269c0c6013..a5f22f01682d2 100644
--- a/clang/lib/Lex/PPMacroExpansion.cpp
+++ b/clang/lib/Lex/PPMacroExpansion.cpp
@@ -129,7 +129,7 @@ void Preprocessor::setLoadedMacroDirective(IdentifierInfo *II,
     II->setHasMacroDefinition(false);
 }
 
-ModuleMacro *Preprocessor::addModuleMacro(Module *Mod, IdentifierInfo *II,
+ModuleMacro *Preprocessor::addModuleMacro(Module *Mod, const IdentifierInfo *II,
                                           MacroInfo *Macro,
                                           ArrayRef<ModuleMacro *> Overrides,
                                           bool &New) {
@@ -162,7 +162,7 @@ ModuleMacro *Preprocessor::addModuleMacro(Module *Mod, IdentifierInfo *II,
   // The new macro is always a leaf macro.
   LeafMacros.push_back(MM);
   // The identifier now has defined macros (that may or may not be visible).
-  II->setHasMacroDefinition(true);
+  const_cast<IdentifierInfo *>(II)->setHasMacroDefinition(true);
 
   New = true;
   return MM;
diff --git a/clang/lib/Lex/Preprocessor.cpp b/clang/lib/Lex/Preprocessor.cpp
index 031ed1e16bb8f..0b70192743a39 100644
--- a/clang/lib/Lex/Preprocessor.cpp
+++ b/clang/lib/Lex/Preprocessor.cpp
@@ -759,7 +759,7 @@ void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
     Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
 }
 
-void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
+void Preprocessor::updateOutOfDateIdentifier(const IdentifierInfo &II) const {
   assert(II.isOutOfDate() && "not out of date");
   getExternalSource()->updateOutOfDateIdentifier(II);
 }
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index 0aa14b0510746..583232f2d610d 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -7700,7 +7700,7 @@ void Parser::ParseParameterDeclarationClause(
     }
 
     // Remember this parsed parameter in ParamInfo.
-    IdentifierInfo *ParmII = ParmDeclarator.getIdentifier();
+    const IdentifierInfo *ParmII = ParmDeclarator.getIdentifier();
 
     // DefArgToks is used when the parsing of default arguments needs
     // to be delayed.
diff --git a/clang/lib/Parse/ParseDeclCXX.cpp b/clang/lib/Parse/ParseDeclCXX.cpp
index 861a25dc5103c..477d81cdc2c23 100644
--- a/clang/lib/Parse/ParseDeclCXX.cpp
+++ b/clang/lib/Parse/ParseDeclCXX.cpp
@@ -616,7 +616,7 @@ bool Parser::ParseUsingDeclarator(DeclaratorContext Context,
   }
 
   // Parse nested-name-specifier.
-  IdentifierInfo *LastII = nullptr;
+  const IdentifierInfo *LastII = nullptr;
   if (ParseOptionalCXXScopeSpecifier(D.SS, /*ObjectType=*/nullptr,
                                      /*ObjectHasErrors=*/false,
                                      /*EnteringContext=*/false,
diff --git a/clang/lib/Parse/ParseExprCXX.cpp b/clang/lib/Parse/ParseExprCXX.cpp
index 73c85c585baae..43d6105dcf31c 100644
--- a/clang/lib/Parse/ParseExprCXX.cpp
+++ b/clang/lib/Parse/ParseExprCXX.cpp
@@ -157,7 +157,8 @@ void Parser::CheckForTemplateAndDigraph(Token &Next, ParsedType ObjectType,
 bool Parser::ParseOptionalCXXScopeSpecifier(
     CXXScopeSpec &SS, ParsedType ObjectType, bool ObjectHadErrors,
     bool EnteringContext, bool *MayBePseudoDestructor, bool IsTypename,
-    IdentifierInfo **LastII, bool OnlyNamespace, bool InUsingDeclaration) {
+    const IdentifierInfo **LastII, bool OnlyNamespace,
+    bool InUsingDeclaration) {
   assert(getLangOpts().CPlusPlus &&
          "Call sites of this function should be guarded by checking for C++");
 
@@ -2626,7 +2627,7 @@ bool Parser::ParseUnqualifiedIdTemplateId(
     // UnqualifiedId.
 
     // FIXME: Store name for literal operator too.
-    IdentifierInfo *TemplateII =
+    const IdentifierInfo *TemplateII =
         Id.getKind() == UnqualifiedIdKind::IK_Identifier ? Id.Identifier
                                                          : nullptr;
     OverloadedOperatorKind OpKind =
diff --git a/clang/lib/Parse/ParseObjc.cpp b/clang/lib/Parse/ParseObjc.cpp
index 88bab0eb27a3e..887d7a36cee7e 100644
--- a/clang/lib/Parse/ParseObjc.cpp
+++ b/clang/lib/Parse/ParseObjc.cpp
@@ -799,11 +799,11 @@ void Parser::ParseObjCInterfaceDeclList(tok::ObjCKeywordKind contextKey,
                                              addedToDeclSpec);
 
         // Install the property declarator into interfaceDecl.
-        IdentifierInfo *SelName =
+        const IdentifierInfo *SelName =
             OCDS.getGetterName() ? OCDS.getGetterName() : FD.D.getIdentifier();
 
         Selector GetterSel = PP.getSelectorTable().getNullarySelector(SelName);
-        IdentifierInfo *SetterName = OCDS.getSetterName();
+        const IdentifierInfo *SetterName = OCDS.getSetterName();
         Selector SetterSel;
         if (SetterName)
           SetterSel = PP.getSelectorTable().getSelector(1, &SetterName);
@@ -1445,7 +1445,7 @@ Decl *Parser::ParseObjCMethodDecl(SourceLocation mLoc,
     return Result;
   }
 
-  SmallVector<IdentifierInfo *, 12> KeyIdents;
+  SmallVector<const IdentifierInfo *, 12> KeyIdents;
   SmallVector<SourceLocation, 12> KeyLocs;
   SmallVector<Sema::ObjCArgInfo, 12> ArgInfos;
   ParseScope PrototypeScope(this, Scope::FunctionPrototypeScope |
@@ -1541,7 +1541,7 @@ Decl *Parser::ParseObjCMethodDecl(SourceLocation mLoc,
     Declarator ParmDecl(DS, ParsedAttributesView::none(),
                         DeclaratorContext::Prototype);
     ParseDeclarator(ParmDecl);
-    IdentifierInfo *ParmII = ParmDecl.getIdentifier();
+    const IdentifierInfo *ParmII = ParmDecl.getIdentifier();
     Decl *Param = Actions.ActOnParamDeclarator(getCurScope(), ParmDecl);
     CParamInfo.push_back(DeclaratorChunk::ParamInfo(ParmII,
                                                     ParmDecl.getIdentifierLoc(),
@@ -3242,7 +3242,7 @@ Parser::ParseObjCMessageExpressionBody(SourceLocation LBracLoc,
   SourceLocation Loc;
   IdentifierInfo *selIdent = ParseObjCSelectorPiece(Loc);
 
-  SmallVector<IdentifierInfo *, 12> KeyIdents;
+  SmallVector<const IdentifierInfo *, 12> KeyIdents;
   SmallVector<SourceLocation, 12> KeyLocs;
   ExprVector KeyExprs;
 
@@ -3642,7 +3642,7 @@ ExprResult Parser::ParseObjCSelectorExpression(SourceLocation AtLoc) {
   if (Tok.isNot(tok::l_paren))
     return ExprError(Diag(Tok, diag::err_expected_lparen_after) << "@selector");
 
-  SmallVector<IdentifierInfo *, 12> KeyIdents;
+  SmallVector<const IdentifierInfo *, 12> KeyIdents;
   SourceLocation sLoc;
 
   BalancedDelimiterTracker T(*this, tok::l_paren);
diff --git a/clang/lib/Parse/ParseTemplate.cpp b/clang/lib/Parse/ParseTemplate.cpp
index d4897f8f66072..03257500426e5 100644
--- a/clang/lib/Parse/ParseTemplate.cpp
+++ b/clang/lib/Parse/ParseTemplate.cpp
@@ -313,7 +313,7 @@ Parser::ParseConceptDefinition(const ParsedTemplateInfo &TemplateInfo,
     return nullptr;
   }
 
-  IdentifierInfo *Id = Result.Identifier;
+  const IdentifierInfo *Id = Result.Identifier;
   SourceLocation IdLoc = Result.getBeginLoc();
 
   DiagnoseAndSkipCXX11Attributes();
@@ -1289,7 +1289,7 @@ bool Parser::AnnotateTemplateIdToken(TemplateTy Template, TemplateNameKind TNK,
     // later.
     Tok.setKind(tok::annot_template_id);
 
-    IdentifierInfo *TemplateII =
+    const IdentifierInfo *TemplateII =
         TemplateName.getKind() == UnqualifiedIdKind::IK_Identifier
             ? TemplateName.Identifier
             : nullptr;
diff --git a/clang/lib/Sema/CodeCompleteConsumer.cpp b/clang/lib/Sema/CodeCompleteConsumer.cpp
index 350bd78b57107..91713d71786ee 100644
--- a/clang/lib/Sema/CodeCompleteConsumer.cpp
+++ b/clang/lib/Sema/CodeCompleteConsumer.cpp
@@ -854,7 +854,8 @@ StringRef CodeCompletionResult::getOrderedName(std::string &Saved) const {
   if (IdentifierInfo *Id = Name.getAsIdentifierInfo())
     return Id->getName();
   if (Name.isObjCZeroArgSelector())
-    if (IdentifierInfo *Id = Name.getObjCSelector().getIdentifierInfoForSlot(0))
+    if (const IdentifierInfo *Id =
+            Name.getObjCSelector().getIdentifierInfoForSlot(0))
       return Id->getName();
 
   Saved = Name.getAsString();
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index 801b03a63dbc8..a2ea66f339c8e 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -92,9 +92,8 @@ DarwinSDKInfo *Sema::getDarwinSDKInfoForAvailabilityChecking() {
   return nullptr;
 }
 
-IdentifierInfo *
-Sema::InventAbbreviatedTemplateParameterTypeName(IdentifierInfo *ParamName,
-                                                 unsigned int Index) {
+IdentifierInfo *Sema::InventAbbreviatedTemplateParameterTypeName(
+    const IdentifierInfo *ParamName, unsigned int Index) {
   std::string InventedName;
   llvm::raw_string_ostream OS(InventedName);
 
diff --git a/clang/lib/Sema/SemaCodeComplete.cpp b/clang/lib/Sema/SemaCodeComplete.cpp
index 83ebcaf9e765a..c335017f243eb 100644
--- a/clang/lib/Sema/SemaCodeComplete.cpp
+++ b/clang/lib/Sema/SemaCodeComplete.cpp
@@ -3691,7 +3691,7 @@ CodeCompletionString *CodeCompletionResult::createCodeCompletionStringForDecl(
         std::string Keyword;
         if (Idx > StartParameter)
           Result.AddChunk(CodeCompletionString::CK_HorizontalSpace);
-        if (IdentifierInfo *II = Sel.getIdentifierInfoForSlot(Idx))
+        if (const IdentifierInfo *II = Sel.getIdentifierInfoForSlot(Idx))
           Keyword += II->getName();
         Keyword += ":";
         if (Idx < StartParameter || AllParametersAreInformative)
@@ -3720,7 +3720,7 @@ CodeCompletionString *CodeCompletionResult::createCodeCompletionStringForDecl(
         Arg = "(" + formatObjCParamQualifiers((*P)->getObjCDeclQualifier(),
                                               ParamType);
         Arg += ParamType.getAsString(Policy) + ")";
-        if (IdentifierInfo *II = (*P)->getIdentifier())
+        if (const IdentifierInfo *II = (*P)->getIdentifier())
           if (DeclaringEntity || AllParametersAreInformative)
             Arg += II->getName();
       }
@@ -4500,11 +4500,11 @@ void Sema::CodeCompleteOrdinaryName(Scope *S,
                             Results.data(), Results.size());
 }
 
-static void AddClassMessageCompletions(Sema &SemaRef, Scope *S,
-                                       ParsedType Receiver,
-                                       ArrayRef<IdentifierInfo *> SelIdents,
-                                       bool AtArgumentExpression, bool IsSuper,
-                                       ResultBuilder &Results);
+static void
+AddClassMessageCompletions(Sema &SemaRef, Scope *S, ParsedType Receiver,
+                           ArrayRef<const IdentifierInfo *> SelIdents,
+                           bool AtArgumentExpression, bool IsSuper,
+                           ResultBuilder &Results);
 
 void Sema::CodeCompleteDeclSpec(Scope *S, DeclSpec &DS,
                                 bool AllowNonIdentifiers,
@@ -4928,7 +4928,7 @@ void Sema::CodeCompletePostfixExpression(Scope *S, ExprResult E,
 
 /// The set of properties that have already been added, referenced by
 /// property name.
-typedef llvm::SmallPtrSet<IdentifierInfo *, 16> AddedPropertiesSet;
+typedef llvm::SmallPtrSet<const IdentifierInfo *, 16> AddedPropertiesSet;
 
 /// Retrieve the container definition, if any?
 static ObjCContainerDecl *getContainerDef(ObjCContainerDecl *Container) {
@@ -5090,7 +5090,7 @@ AddObjCProperties(const CodeCompletionContext &CCContext,
     PrintingPolicy Policy = getCompletionPrintingPolicy(Results.getSema());
     // Adds a method result
     const auto AddMethod = [&](const ObjCMethodDecl *M) {
-      IdentifierInfo *Name = M->getSelector().getIdentifierInfoForSlot(0);
+      const IdentifierInfo *Name = M->getSelector().getIdentifierInfoForSlot(0);
       if (!Name)
         return;
       if (!AddedProperties.insert(Name).second)
@@ -5859,10 +5859,10 @@ void Sema::CodeCompleteMemberReferenceExpr(Scope *S, Expr *Base,
 }
 
 void Sema::CodeCompleteObjCClassPropertyRefExpr(Scope *S,
-                                                IdentifierInfo &ClassName,
+                                                const IdentifierInfo &ClassName,
                                                 SourceLocation ClassNameLoc,
                                                 bool IsBaseExprStatement) {
-  IdentifierInfo *ClassNamePtr = &ClassName;
+  const IdentifierInfo *ClassNamePtr = &ClassName;
   ObjCInterfaceDecl *IFace = getObjCInterfaceDecl(ClassNamePtr, ClassNameLoc);
   if (!IFace)
     return;
@@ -7527,7 +7527,7 @@ enum ObjCMethodKind {
 };
 
 static bool isAcceptableObjCSelector(Selector Sel, ObjCMethodKind WantKind,
-                                     ArrayRef<IdentifierInfo *> SelIdents,
+                                     ArrayRef<const IdentifierInfo *> SelIdents,
                                      bool AllowSameLength = true) {
   unsigned NumSelIdents = SelIdents.size();
   if (NumSelIdents > Sel.getNumArgs())
@@ -7554,7 +7554,7 @@ static bool isAcceptableObjCSelector(Selector Sel, ObjCMethodKind WantKind,
 
 static bool isAcceptableObjCMethod(ObjCMethodDecl *Method,
                                    ObjCMethodKind WantKind,
-                                   ArrayRef<IdentifierInfo *> SelIdents,
+                                   ArrayRef<const IdentifierInfo *> SelIdents,
                                    bool AllowSameLength = true) {
   return isAcceptableObjCSelector(Method->getSelector(), WantKind, SelIdents,
                                   AllowSameLength);
@@ -7586,7 +7586,7 @@ typedef llvm::SmallPtrSet<Selector, 16> VisitedSelectorSet;
 /// \param Results the structure into which we'll add results.
 static void AddObjCMethods(ObjCContainerDecl *Container,
                            bool WantInstanceMethods, ObjCMethodKind WantKind,
-                           ArrayRef<IdentifierInfo *> SelIdents,
+                           ArrayRef<const IdentifierInfo *> SelIdents,
                            DeclContext *CurContext,
                            VisitedSelectorSet &Selectors, bool AllowSameLength,
                            ResultBuilder &Results, bool InOriginalClass = true,
@@ -7819,7 +7819,7 @@ static ObjCInterfaceDecl *GetAssumedMessageSendExprType(Expr *E) {
   if (Sel.isNull())
     return nullptr;
 
-  IdentifierInfo *Id = Sel.getIdentifierInfoForSlot(0);
+  const IdentifierInfo *Id = Sel.getIdentifierInfoForSlot(0);
   if (!Id)
     return nullptr;
 
@@ -7895,7 +7895,7 @@ static ObjCInterfaceDecl *GetAssumedMessageSendExprType(Expr *E) {
 /// this "super" completion. If NULL, no completion was added.
 static ObjCMethodDecl *
 AddSuperSendCompletion(Sema &S, bool NeedSuperKeyword,
-                       ArrayRef<IdentifierInfo *> SelIdents,
+                       ArrayRef<const IdentifierInfo *> SelIdents,
                        ResultBuilder &Results) {
   ObjCMethodDecl *CurMethod = S.getCurMethodDecl();
   if (!CurMethod)
@@ -8032,9 +8032,9 @@ void Sema::CodeCompleteObjCMessageReceiver(Scope *S) {
                             Results.data(), Results.size());
 }
 
-void Sema::CodeCompleteObjCSuperMessage(Scope *S, SourceLocation SuperLoc,
-                                        ArrayRef<IdentifierInfo *> SelIdents,
-                                        bool AtArgumentExpression) {
+void Sema::CodeCompleteObjCSuperMessage(
+    Scope *S, SourceLocation SuperLoc,
+    ArrayRef<const IdentifierInfo *> SelIdents, bool AtArgumentExpression) {
   ObjCInterfaceDecl *CDecl = nullptr;
   if (ObjCMethodDecl *CurMethod = getCurMethodDecl()) {
     // Figure out which interface we're in.
@@ -8059,7 +8059,7 @@ void Sema::CodeCompleteObjCSuperMessage(Scope *S, SourceLocation SuperLoc,
   } else {
     // "super" may be the name of a type or variable. Figure out which
     // it is.
-    IdentifierInfo *Super = getSuperIdentifier();
+    const IdentifierInfo *Super = getSuperIdentifier();
     NamedDecl *ND = LookupSingleName(S, Super, SuperLoc, LookupOrdinaryName);
     if ((CDecl = dyn_cast_or_null<ObjCInterfaceDecl>(ND))) {
       // "super" names an interface. Use it.
@@ -8127,11 +8127,11 @@ static QualType getPreferredArgumentTypeForMessageSend(ResultBuilder &Results,
   return PreferredType;
 }
 
-static void AddClassMessageCompletions(Sema &SemaRef, Scope *S,
-                                       ParsedType Receiver,
-                                       ArrayRef<IdentifierInfo *> SelIdents,
-                                       bool AtArgumentExpression, bool IsSuper,
-                                       ResultBuilder &Results) {
+static void
+AddClassMessageCompletions(Sema &SemaRef, Scope *S, ParsedType Receiver,
+                           ArrayRef<const IdentifierInfo *> SelIdents,
+                           bool AtArgumentExpression, bool IsSuper,
+                           ResultBuilder &Results) {
   typedef CodeCompletionResult Result;
   ObjCInterfaceDecl *CDecl = nullptr;
 
@@ -8202,10 +8202,9 @@ static void AddClassMessageCompletions(Sema &SemaRef, Scope *S,
   Results.ExitScope();
 }
 
-void Sema::CodeCompleteObjCClassMessage(Scope *S, ParsedType Receiver,
-                                        ArrayRef<IdentifierInfo *> SelIdents,
-                                        bool AtArgumentExpression,
-                                        bool IsSuper) {
+void Sema::CodeCompleteObjCClassMessage(
+    Scope *S, ParsedType Receiver, ArrayRef<const IdentifierInfo *> SelIdents,
+    bool AtArgumentExpression, bool IsSuper) {
 
   QualType T = this->GetTypeFromParser(Receiver);
 
@@ -8237,10 +8236,9 @@ void Sema::CodeCompleteObjCClassMessage(Scope *S, ParsedType Receiver,
                             Results.data(), Results.size());
 }
 
-void Sema::CodeCompleteObjCInstanceMessage(Scope *S, Expr *Receiver,
-                                           ArrayRef<IdentifierInfo *> SelIdents,
-                                           bool AtArgumentExpression,
-                                           ObjCInterfaceDecl *Super) {
+void Sema::CodeCompleteObjCInstanceMessage(
+    Scope *S, Expr *Receiver, ArrayRef<const IdentifierInfo *> SelIdents,
+    bool AtArgumentExpression, ObjCInterfaceDecl *Super) {
   typedef CodeCompletionResult Result;
 
   Expr *RecExpr = static_cast<Expr *>(Receiver);
@@ -8410,8 +8408,8 @@ void Sema::CodeCompleteObjCForCollection(Scope *S,
   CodeCompleteExpression(S, Data);
 }
 
-void Sema::CodeCompleteObjCSelector(Scope *S,
-                                    ArrayRef<IdentifierInfo *> SelIdents) {
+void Sema::CodeCompleteObjCSelector(
+    Scope *S, ArrayRef<const IdentifierInfo *> SelIdents) {
   // If we have an external source, load the entire class method
   // pool from the AST file.
   if (ExternalSource) {
@@ -9166,8 +9164,8 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // Add -(void)getKey:(type **)buffer range:(NSRange)inRange
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("get") + UpperKey).str();
-    IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
-                                      &Context.Idents.get("range")};
+    const IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
+                                            &Context.Idents.get("range")};
 
     if (KnownSelectors.insert(Selectors.getSelector(2, SelectorIds)).second) {
       if (ReturnType.isNull()) {
@@ -9198,8 +9196,8 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // - (void)insertObject:(type *)object inKeyAtIndex:(NSUInteger)index
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("in") + UpperKey + "AtIndex").str();
-    IdentifierInfo *SelectorIds[2] = {&Context.Idents.get("insertObject"),
-                                      &Context.Idents.get(SelectorName)};
+    const IdentifierInfo *SelectorIds[2] = {&Context.Idents.get("insertObject"),
+                                            &Context.Idents.get(SelectorName)};
 
     if (KnownSelectors.insert(Selectors.getSelector(2, SelectorIds)).second) {
       if (ReturnType.isNull()) {
@@ -9228,8 +9226,8 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // - (void)insertKey:(NSArray *)array atIndexes:(NSIndexSet *)indexes
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("insert") + UpperKey).str();
-    IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
-                                      &Context.Idents.get("atIndexes")};
+    const IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
+                                            &Context.Idents.get("atIndexes")};
 
     if (KnownSelectors.insert(Selectors.getSelector(2, SelectorIds)).second) {
       if (ReturnType.isNull()) {
@@ -9258,7 +9256,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName =
         (Twine("removeObjectFrom") + UpperKey + "AtIndex").str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9279,7 +9277,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // -(void)removeKeyAtIndexes:(NSIndexSet *)indexes
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("remove") + UpperKey + "AtIndexes").str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9301,8 +9299,8 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName =
         (Twine("replaceObjectIn") + UpperKey + "AtIndex").str();
-    IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
-                                      &Context.Idents.get("withObject")};
+    const IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName),
+                                            &Context.Idents.get("withObject")};
 
     if (KnownSelectors.insert(Selectors.getSelector(2, SelectorIds)).second) {
       if (ReturnType.isNull()) {
@@ -9332,8 +9330,8 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
     std::string SelectorName1 =
         (Twine("replace") + UpperKey + "AtIndexes").str();
     std::string SelectorName2 = (Twine("with") + UpperKey).str();
-    IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName1),
-                                      &Context.Idents.get(SelectorName2)};
+    const IdentifierInfo *SelectorIds[2] = {&Context.Idents.get(SelectorName1),
+                                            &Context.Idents.get(SelectorName2)};
 
     if (KnownSelectors.insert(Selectors.getSelector(2, SelectorIds)).second) {
       if (ReturnType.isNull()) {
@@ -9368,7 +9366,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
                 ->getInterfaceDecl()
                 ->getName() == "NSEnumerator"))) {
     std::string SelectorName = (Twine("enumeratorOf") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getNullarySelector(SelectorId))
             .second) {
       if (ReturnType.isNull()) {
@@ -9387,7 +9385,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   if (IsInstanceMethod &&
       (ReturnType.isNull() || ReturnType->isObjCObjectPointerType())) {
     std::string SelectorName = (Twine("memberOf") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9417,7 +9415,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName =
         (Twine("add") + UpperKey + Twine("Object")).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9439,7 +9437,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // - (void)addKey:(NSSet *)objects
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("add") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9461,7 +9459,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName =
         (Twine("remove") + UpperKey + Twine("Object")).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9483,7 +9481,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // - (void)removeKey:(NSSet *)objects
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("remove") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9504,7 +9502,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
   // - (void)intersectKey:(NSSet *)objects
   if (IsInstanceMethod && ReturnTypeMatchesVoid) {
     std::string SelectorName = (Twine("intersect") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getUnarySelector(SelectorId)).second) {
       if (ReturnType.isNull()) {
         Builder.AddChunk(CodeCompletionString::CK_LeftParen);
@@ -9533,7 +9531,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
                 ->getName() == "NSSet"))) {
     std::string SelectorName =
         (Twine("keyPathsForValuesAffecting") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getNullarySelector(SelectorId))
             .second) {
       if (ReturnType.isNull()) {
@@ -9554,7 +9552,7 @@ static void AddObjCKeyValueCompletions(ObjCPropertyDecl *Property,
        ReturnType->isBooleanType())) {
     std::string SelectorName =
         (Twine("automaticallyNotifiesObserversOf") + UpperKey).str();
-    IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
+    const IdentifierInfo *SelectorId = &Context.Idents.get(SelectorName);
     if (KnownSelectors.insert(Selectors.getNullarySelector(SelectorId))
             .second) {
       if (ReturnType.isNull()) {
@@ -9749,7 +9747,7 @@ void Sema::CodeCompleteObjCMethodDecl(Scope *S,
 
 void Sema::CodeCompleteObjCMethodDeclSelector(
     Scope *S, bool IsInstanceMethod, bool AtParameterName, ParsedType ReturnTy,
-    ArrayRef<IdentifierInfo *> SelIdents) {
+    ArrayRef<const IdentifierInfo *> SelIdents) {
   // If we have an external source, load the entire class method
   // pool from the AST file.
   if (ExternalSource) {
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 8472aaeb6bad9..5a23179dfbbf4 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -2318,7 +2318,7 @@ void Sema::ActOnPopScope(SourceLocation Loc, Scope *S) {
 ///
 /// \returns The declaration of the named Objective-C class, or NULL if the
 /// class could not be found.
-ObjCInterfaceDecl *Sema::getObjCInterfaceDecl(IdentifierInfo *&Id,
+ObjCInterfaceDecl *Sema::getObjCInterfaceDecl(const IdentifierInfo *&Id,
                                               SourceLocation IdLoc,
                                               bool DoTypoCorrection) {
   // The third "scope" argument is 0 since we aren't enabling lazy built-in
@@ -15307,7 +15307,7 @@ Decl *Sema::ActOnParamDeclarator(Scope *S, Declarator &D,
   QualType parmDeclType = TInfo->getType();
 
   // Check for redeclaration of parameters, e.g. int foo(int x, int x);
-  IdentifierInfo *II = D.getIdentifier();
+  const IdentifierInfo *II = D.getIdentifier();
   if (II) {
     LookupResult R(*this, II, D.getIdentifierLoc(), LookupOrdinaryName,
                    ForVisibleRedeclaration);
@@ -15459,9 +15459,9 @@ QualType Sema::AdjustParameterTypeForObjCAutoRefCount(QualType T,
 }
 
 ParmVarDecl *Sema::CheckParameter(DeclContext *DC, SourceLocation StartLoc,
-                                  SourceLocation NameLoc, IdentifierInfo *Name,
-                                  QualType T, TypeSourceInfo *TSInfo,
-                                  StorageClass SC) {
+                                  SourceLocation NameLoc,
+                                  const IdentifierInfo *Name, QualType T,
+                                  TypeSourceInfo *TSInfo, StorageClass SC) {
   // In ARC, infer a lifetime qualifier for appropriate parameter types.
   if (getLangOpts().ObjCAutoRefCount &&
       T.getObjCLifetime() == Qualifiers::OCL_None &&
@@ -18551,8 +18551,9 @@ void Sema::ActOnTagDefinitionError(Scope *S, Decl *TagD) {
 
 // Note that FieldName may be null for anonymous bitfields.
 ExprResult Sema::VerifyBitField(SourceLocation FieldLoc,
-                                IdentifierInfo *FieldName, QualType FieldTy,
-                                bool IsMsStruct, Expr *BitWidth) {
+                                const IdentifierInfo *FieldName,
+                                QualType FieldTy, bool IsMsStruct,
+                                Expr *BitWidth) {
   assert(BitWidth);
   if (BitWidth->containsErrors())
     return ExprError();
@@ -18661,7 +18662,7 @@ FieldDecl *Sema::HandleField(Scope *S, RecordDecl *Record,
     return nullptr;
   }
 
-  IdentifierInfo *II = D.getIdentifier();
+  const IdentifierInfo *II = D.getIdentifier();
   SourceLocation Loc = DeclStart;
   if (II) Loc = D.getIdentifierLoc();
 
@@ -18762,7 +18763,7 @@ FieldDecl *Sema::CheckFieldDecl(DeclarationName Name, QualType T,
                                 SourceLocation TSSL,
                                 AccessSpecifier AS, NamedDecl *PrevDecl,
                                 Declarator *D) {
-  IdentifierInfo *II = Name.getAsIdentifierInfo();
+  const IdentifierInfo *II = Name.getAsIdentifierInfo();
   bool InvalidDecl = false;
   if (D) InvalidDecl = D->isInvalidType();
 
@@ -19022,7 +19023,7 @@ TranslateIvarVisibility(tok::ObjCKeywordKind ivarVisibility) {
 Decl *Sema::ActOnIvar(Scope *S, SourceLocation DeclStart, Declarator &D,
                       Expr *BitWidth, tok::ObjCKeywordKind Visibility) {
 
-  IdentifierInfo *II = D.getIdentifier();
+  const IdentifierInfo *II = D.getIdentifier();
   SourceLocation Loc = DeclStart;
   if (II) Loc = D.getIdentifierLoc();
 
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 068a2e4f04fa5..858951580ea45 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -16913,11 +16913,10 @@ Decl *Sema::ActOnEmptyDeclaration(Scope *S,
 /// Perform semantic analysis for the variable declaration that
 /// occurs within a C++ catch clause, returning the newly-created
 /// variable.
-VarDecl *Sema::BuildExceptionDeclaration(Scope *S,
-                                         TypeSourceInfo *TInfo,
+VarDecl *Sema::BuildExceptionDeclaration(Scope *S, TypeSourceInfo *TInfo,
                                          SourceLocation StartLoc,
                                          SourceLocation Loc,
-                                         IdentifierInfo *Name) {
+                                         const IdentifierInfo *Name) {
   bool Invalid = false;
   QualType ExDeclType = TInfo->getType();
 
@@ -17062,7 +17061,7 @@ Decl *Sema::ActOnExceptionDeclarator(Scope *S, Declarator &D) {
     Invalid = true;
   }
 
-  IdentifierInfo *II = D.getIdentifier();
+  const IdentifierInfo *II = D.getIdentifier();
   if (NamedDecl *PrevDecl = LookupSingleName(S, II, D.getIdentifierLoc(),
                                              LookupOrdinaryName,
                                              ForVisibleRedeclaration)) {
@@ -19158,7 +19157,7 @@ MSPropertyDecl *Sema::HandleMSProperty(Scope *S, RecordDecl *Record,
                                        InClassInitStyle InitStyle,
                                        AccessSpecifier AS,
                                        const ParsedAttr &MSPropertyAttr) {
-  IdentifierInfo *II = D.getIdentifier();
+  const IdentifierInfo *II = D.getIdentifier();
   if (!II) {
     Diag(DeclStart, diag::err_anonymous_property);
     return nullptr;
diff --git a/clang/lib/Sema/SemaDeclObjC.cpp b/clang/lib/Sema/SemaDeclObjC.cpp
index 94a245f0f905f..74d6f0700b0e4 100644
--- a/clang/lib/Sema/SemaDeclObjC.cpp
+++ b/clang/lib/Sema/SemaDeclObjC.cpp
@@ -1818,9 +1818,9 @@ Sema::ActOnForwardProtocolDeclaration(SourceLocation AtProtocolLoc,
 }
 
 ObjCCategoryDecl *Sema::ActOnStartCategoryInterface(
-    SourceLocation AtInterfaceLoc, IdentifierInfo *ClassName,
+    SourceLocation AtInterfaceLoc, const IdentifierInfo *ClassName,
     SourceLocation ClassLoc, ObjCTypeParamList *typeParamList,
-    IdentifierInfo *CategoryName, SourceLocation CategoryLoc,
+    const IdentifierInfo *CategoryName, SourceLocation CategoryLoc,
     Decl *const *ProtoRefs, unsigned NumProtoRefs,
     const SourceLocation *ProtoLocs, SourceLocation EndProtoLoc,
     const ParsedAttributesView &AttrList) {
@@ -1916,9 +1916,9 @@ ObjCCategoryDecl *Sema::ActOnStartCategoryInterface(
 /// category implementation declaration and build an ObjCCategoryImplDecl
 /// object.
 ObjCCategoryImplDecl *Sema::ActOnStartCategoryImplementation(
-    SourceLocation AtCatImplLoc, IdentifierInfo *ClassName,
-    SourceLocation ClassLoc, IdentifierInfo *CatName, SourceLocation CatLoc,
-    const ParsedAttributesView &Attrs) {
+    SourceLocation AtCatImplLoc, const IdentifierInfo *ClassName,
+    SourceLocation ClassLoc, const IdentifierInfo *CatName,
+    SourceLocation CatLoc, const ParsedAttributesView &Attrs) {
   ObjCInterfaceDecl *IDecl = getObjCInterfaceDecl(ClassName, ClassLoc, true);
   ObjCCategoryDecl *CatIDecl = nullptr;
   if (IDecl && IDecl->hasDefinition()) {
@@ -1982,8 +1982,8 @@ ObjCCategoryImplDecl *Sema::ActOnStartCategoryImplementation(
 }
 
 ObjCImplementationDecl *Sema::ActOnStartClassImplementation(
-    SourceLocation AtClassImplLoc, IdentifierInfo *ClassName,
-    SourceLocation ClassLoc, IdentifierInfo *SuperClassname,
+    SourceLocation AtClassImplLoc, const IdentifierInfo *ClassName,
+    SourceLocation ClassLoc, const IdentifierInfo *SuperClassname,
     SourceLocation SuperClassLoc, const ParsedAttributesView &Attrs) {
   ObjCInterfaceDecl *IDecl = nullptr;
   // Check for another declaration kind with the same name.
@@ -2751,7 +2751,7 @@ static void CheckProtocolMethodDefs(
     // implemented in the class, we should not issue "Method definition not
     // found" warnings.
     // FIXME: Use a general GetUnarySelector method for this.
-    IdentifierInfo* II = &S.Context.Idents.get("forwardInvocation");
+    const IdentifierInfo *II = &S.Context.Idents.get("forwardInvocation");
     Selector fISelector = S.Context.Selectors.getSelector(1, &II);
     if (InsMap.count(fISelector))
       // Is IDecl derived from 'NSProxy'? If so, no instance methods
@@ -5105,8 +5105,8 @@ bool Sema::CheckObjCDeclScope(Decl *D) {
 /// Called whenever \@defs(ClassName) is encountered in the source.  Inserts the
 /// instance variables of ClassName into Decls.
 void Sema::ActOnDefs(Scope *S, Decl *TagD, SourceLocation DeclStart,
-                     IdentifierInfo *ClassName,
-                     SmallVectorImpl<Decl*> &Decls) {
+                     const IdentifierInfo *ClassName,
+                     SmallVectorImpl<Decl *> &Decls) {
   // Check that ClassName is a valid class
   ObjCInterfaceDecl *Class = getObjCInterfaceDecl(ClassName, DeclStart);
   if (!Class) {
@@ -5148,8 +5148,7 @@ void Sema::ActOnDefs(Scope *S, Decl *TagD, SourceLocation DeclStart,
 VarDecl *Sema::BuildObjCExceptionDecl(TypeSourceInfo *TInfo, QualType T,
                                       SourceLocation StartLoc,
                                       SourceLocation IdLoc,
-                                      IdentifierInfo *Id,
-                                      bool Invalid) {
+                                      const IdentifierInfo *Id, bool Invalid) {
   // ISO/IEC TR 18037 S6.7.3: "The type of an object with automatic storage
   // duration shall not be qualified by an address-space qualifier."
   // Since all parameters have automatic store duration, they can not have
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 7b9b8f149d9ed..12f42f66e5e21 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -57,7 +57,7 @@ using namespace sema;
 /// name of the corresponding type.
 ParsedType Sema::getInheritingConstructorName(CXXScopeSpec &SS,
                                               SourceLocation NameLoc,
-                                              IdentifierInfo &Name) {
+                                              const IdentifierInfo &Name) {
   NestedNameSpecifier *NNS = SS.getScopeRep();
 
   // Convert the nested-name-specifier into a type.
@@ -89,10 +89,9 @@ ParsedType Sema::getInheritingConstructorName(CXXScopeSpec &SS,
                           Context.getTrivialTypeSourceInfo(Type, NameLoc));
 }
 
-ParsedType Sema::getConstructorName(IdentifierInfo &II,
-                                    SourceLocation NameLoc,
-                                    Scope *S, CXXScopeSpec &SS,
-                                    bool EnteringContext) {
+ParsedType Sema::getConstructorName(const IdentifierInfo &II,
+                                    SourceLocation NameLoc, Scope *S,
+                                    CXXScopeSpec &SS, bool EnteringContext) {
   CXXRecordDecl *CurClass = getCurrentClass(S, &SS);
   assert(CurClass && &II == CurClass->getIdentifier() &&
          "not a constructor name");
@@ -140,9 +139,9 @@ ParsedType Sema::getConstructorName(IdentifierInfo &II,
   return ParsedType::make(T);
 }
 
-ParsedType Sema::getDestructorName(IdentifierInfo &II, SourceLocation NameLoc,
-                                   Scope *S, CXXScopeSpec &SS,
-                                   ParsedType ObjectTypePtr,
+ParsedType Sema::getDestructorName(const IdentifierInfo &II,
+                                   SourceLocation NameLoc, Scope *S,
+                                   CXXScopeSpec &SS, ParsedType ObjectTypePtr,
                                    bool EnteringContext) {
   // Determine where to perform name lookup.
 
@@ -500,7 +499,7 @@ bool Sema::checkLiteralOperatorId(const CXXScopeSpec &SS,
     //
     // double operator""_Bq(long double);  // OK: not a reserved identifier
     // double operator"" _Bq(long double); // ill-formed, no diagnostic required
-    IdentifierInfo *II = Name.Identifier;
+    const IdentifierInfo *II = Name.Identifier;
     ReservedIdentifierStatus Status = II->isReserved(PP.getLangOpts());
     SourceLocation Loc = Name.getEndLoc();
     if (!PP.getSourceManager().isInSystemHeader(Loc)) {
@@ -9178,10 +9177,9 @@ concepts::Requirement *Sema::ActOnSimpleRequirement(Expr *E) {
                               /*ReturnTypeRequirement=*/{});
 }
 
-concepts::Requirement *
-Sema::ActOnTypeRequirement(SourceLocation TypenameKWLoc, CXXScopeSpec &SS,
-                           SourceLocation NameLoc, IdentifierInfo *TypeName,
-                           TemplateIdAnnotation *TemplateId) {
+concepts::Requirement *Sema::ActOnTypeRequirement(
+    SourceLocation TypenameKWLoc, CXXScopeSpec &SS, SourceLocation NameLoc,
+    const IdentifierInfo *TypeName, TemplateIdAnnotation *TemplateId) {
   assert(((!TypeName && TemplateId) || (TypeName && !TemplateId)) &&
          "Exactly one of TypeName and TemplateId must be specified.");
   TypeSourceInfo *TSI = nullptr;
diff --git a/clang/lib/Sema/SemaExprObjC.cpp b/clang/lib/Sema/SemaExprObjC.cpp
index a8853f634c9cc..3148f0db6e20c 100644
--- a/clang/lib/Sema/SemaExprObjC.cpp
+++ b/clang/lib/Sema/SemaExprObjC.cpp
@@ -663,10 +663,8 @@ ExprResult Sema::BuildObjCBoxedExpr(SourceRange SR, Expr *ValueExpr) {
     }
 
     if (!ValueWithBytesObjCTypeMethod) {
-      IdentifierInfo *II[] = {
-        &Context.Idents.get("valueWithBytes"),
-        &Context.Idents.get("objCType")
-      };
+      const IdentifierInfo *II[] = {&Context.Idents.get("valueWithBytes"),
+                                    &Context.Idents.get("objCType")};
       Selector ValueWithBytesObjCType = Context.Selectors.getSelector(2, II);
 
       // Look for the appropriate method within NSValue.
@@ -2155,13 +2153,12 @@ HandleExprPropertyRefExpr(const ObjCObjectPointerType *OPT,
   return ExprError();
 }
 
-ExprResult Sema::
-ActOnClassPropertyRefExpr(IdentifierInfo &receiverName,
-                          IdentifierInfo &propertyName,
-                          SourceLocation receiverNameLoc,
-                          SourceLocation propertyNameLoc) {
+ExprResult Sema::ActOnClassPropertyRefExpr(const IdentifierInfo &receiverName,
+                                           const IdentifierInfo &propertyName,
+                                           SourceLocation receiverNameLoc,
+                                           SourceLocation propertyNameLoc) {
 
-  IdentifierInfo *receiverNamePtr = &receiverName;
+  const IdentifierInfo *receiverNamePtr = &receiverName;
   ObjCInterfaceDecl *IFace = getObjCInterfaceDecl(receiverNamePtr,
                                                   receiverNameLoc);
 
diff --git a/clang/lib/Sema/SemaObjCProperty.cpp b/clang/lib/Sema/SemaObjCProperty.cpp
index f9e1ad0121e2a..222a65a13dd0b 100644
--- a/clang/lib/Sema/SemaObjCProperty.cpp
+++ b/clang/lib/Sema/SemaObjCProperty.cpp
@@ -419,7 +419,7 @@ Sema::HandlePropertyInClassExtension(Scope *S,
   ObjCCategoryDecl *CDecl = cast<ObjCCategoryDecl>(CurContext);
   // Diagnose if this property is already in continuation class.
   DeclContext *DC = CurContext;
-  IdentifierInfo *PropertyId = FD.D.getIdentifier();
+  const IdentifierInfo *PropertyId = FD.D.getIdentifier();
   ObjCInterfaceDecl *CCPrimary = CDecl->getClassInterface();
 
   // We need to look in the @interface to see if the @property was
@@ -571,7 +571,7 @@ ObjCPropertyDecl *Sema::CreatePropertyDecl(Scope *S,
                                            TypeSourceInfo *TInfo,
                                            tok::ObjCKeywordKind MethodImplKind,
                                            DeclContext *lexicalDC){
-  IdentifierInfo *PropertyId = FD.D.getIdentifier();
+  const IdentifierInfo *PropertyId = FD.D.getIdentifier();
 
   // Property defaults to 'assign' if it is readwrite, unless this is ARC
   // and the type is retainable.
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index c814535ad6bdb..e9efb4721133f 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -7375,7 +7375,7 @@ void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
           llvm::omp::TraitProperty::implementation_extension_allow_templates))
     return;
 
-  IdentifierInfo *BaseII = D.getIdentifier();
+  const IdentifierInfo *BaseII = D.getIdentifier();
   LookupResult Lookup(*this, DeclarationName(BaseII), D.getIdentifierLoc(),
                       LookupOrdinaryName);
   LookupParsedName(Lookup, S, &D.getCXXScopeSpec());
diff --git a/clang/lib/Sema/SemaPseudoObject.cpp b/clang/lib/Sema/SemaPseudoObject.cpp
index 528c261c4a297..82774760b34d4 100644
--- a/clang/lib/Sema/SemaPseudoObject.cpp
+++ b/clang/lib/Sema/SemaPseudoObject.cpp
@@ -613,9 +613,9 @@ bool ObjCPropertyOpBuilder::findGetter() {
       // Must build the getter selector the hard way.
       ObjCMethodDecl *setter = RefExpr->getImplicitPropertySetter();
       assert(setter && "both setter and getter are null - cannot happen");
-      IdentifierInfo *setterName =
-        setter->getSelector().getIdentifierInfoForSlot(0);
-      IdentifierInfo *getterName =
+      const IdentifierInfo *setterName =
+          setter->getSelector().getIdentifierInfoForSlot(0);
+      const IdentifierInfo *getterName =
           &S.Context.Idents.get(setterName->getName().substr(3));
       GetterSelector =
         S.PP.getSelectorTable().getNullarySelector(getterName);
@@ -640,9 +640,9 @@ bool ObjCPropertyOpBuilder::findSetter(bool warn) {
       SetterSelector = setter->getSelector();
       return true;
     } else {
-      IdentifierInfo *getterName =
-        RefExpr->getImplicitPropertyGetter()->getSelector()
-          .getIdentifierInfoForSlot(0);
+      const IdentifierInfo *getterName = RefExpr->getImplicitPropertyGetter()
+                                             ->getSelector()
+                                             .getIdentifierInfoForSlot(0);
       SetterSelector =
         SelectorTable::constructSetterSelector(S.PP.getIdentifierTable(),
                                                S.PP.getSelectorTable(),
@@ -667,7 +667,8 @@ bool ObjCPropertyOpBuilder::findSetter(bool warn) {
         front = isLowercase(front) ? toUppercase(front) : toLowercase(front);
         SmallString<100> PropertyName = thisPropertyName;
         PropertyName[0] = front;
-        IdentifierInfo *AltMember = &S.PP.getIdentifierTable().get(PropertyName);
+        const IdentifierInfo *AltMember =
+            &S.PP.getIdentifierTable().get(PropertyName);
         if (ObjCPropertyDecl *prop1 = IFace->FindPropertyDeclaration(
                 AltMember, prop->getQueryKind()))
           if (prop != prop1 && (prop1->getSetterMethodDecl() == setter)) {
@@ -1126,9 +1127,8 @@ static void CheckKeyForObjCARCConversion(Sema &S, QualType ContainerT,
     return;
   // dictionary subscripting.
   // - (id)objectForKeyedSubscript:(id)key;
-  IdentifierInfo *KeyIdents[] = {
-    &S.Context.Idents.get("objectForKeyedSubscript")
-  };
+  const IdentifierInfo *KeyIdents[] = {
+      &S.Context.Idents.get("objectForKeyedSubscript")};
   Selector GetterSelector = S.Context.Selectors.getSelector(1, KeyIdents);
   ObjCMethodDecl *Getter = S.LookupMethodInObjectType(GetterSelector, ContainerT,
                                                       true /*instance*/);
@@ -1169,16 +1169,14 @@ bool ObjCSubscriptOpBuilder::findAtIndexGetter() {
   if (!arrayRef) {
     // dictionary subscripting.
     // - (id)objectForKeyedSubscript:(id)key;
-    IdentifierInfo *KeyIdents[] = {
-      &S.Context.Idents.get("objectForKeyedSubscript")
-    };
+    const IdentifierInfo *KeyIdents[] = {
+        &S.Context.Idents.get("objectForKeyedSubscript")};
     AtIndexGetterSelector = S.Context.Selectors.getSelector(1, KeyIdents);
   }
   else {
     // - (id)objectAtIndexedSubscript:(size_t)index;
-    IdentifierInfo *KeyIdents[] = {
-      &S.Context.Idents.get("objectAtIndexedSubscript")
-    };
+    const IdentifierInfo *KeyIdents[] = {
+        &S.Context.Idents.get("objectAtIndexedSubscript")};
 
     AtIndexGetterSelector = S.Context.Selectors.getSelector(1, KeyIdents);
   }
@@ -1274,18 +1272,16 @@ bool ObjCSubscriptOpBuilder::findAtIndexSetter() {
   if (!arrayRef) {
     // dictionary subscripting.
     // - (void)setObject:(id)object forKeyedSubscript:(id)key;
-    IdentifierInfo *KeyIdents[] = {
-      &S.Context.Idents.get("setObject"),
-      &S.Context.Idents.get("forKeyedSubscript")
-    };
+    const IdentifierInfo *KeyIdents[] = {
+        &S.Context.Idents.get("setObject"),
+        &S.Context.Idents.get("forKeyedSubscript")};
     AtIndexSetterSelector = S.Context.Selectors.getSelector(2, KeyIdents);
   }
   else {
     // - (void)setObject:(id)object atIndexedSubscript:(NSInteger)index;
-    IdentifierInfo *KeyIdents[] = {
-      &S.Context.Idents.get("setObject"),
-      &S.Context.Idents.get("atIndexedSubscript")
-    };
+    const IdentifierInfo *KeyIdents[] = {
+        &S.Context.Idents.get("setObject"),
+        &S.Context.Idents.get("atIndexedSubscript")};
     AtIndexSetterSelector = S.Context.Selectors.getSelector(2, KeyIdents);
   }
   AtIndexSetter = S.LookupMethodInObjectType(AtIndexSetterSelector, ResultType,
@@ -1474,7 +1470,7 @@ ExprResult MSPropertyOpBuilder::buildGet() {
   }
 
   UnqualifiedId GetterName;
-  IdentifierInfo *II = RefExpr->getPropertyDecl()->getGetterId();
+  const IdentifierInfo *II = RefExpr->getPropertyDecl()->getGetterId();
   GetterName.setIdentifier(II, RefExpr->getMemberLoc());
   CXXScopeSpec SS;
   SS.Adopt(RefExpr->getQualifierLoc());
@@ -1503,7 +1499,7 @@ ExprResult MSPropertyOpBuilder::buildSet(Expr *op, SourceLocation sl,
   }
 
   UnqualifiedId SetterName;
-  IdentifierInfo *II = RefExpr->getPropertyDecl()->getSetterId();
+  const IdentifierInfo *II = RefExpr->getPropertyDecl()->getSetterId();
   SetterName.setIdentifier(II, RefExpr->getMemberLoc());
   CXXScopeSpec SS;
   SS.Adopt(RefExpr->getQualifierLoc());
diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
index e72397adec24f..e53c76e65b03d 100644
--- a/clang/lib/Sema/SemaStmt.cpp
+++ b/clang/lib/Sema/SemaStmt.cpp
@@ -2275,11 +2275,9 @@ Sema::CheckObjCForCollectionOperand(SourceLocation forLoc, Expr *collection) {
     // Otherwise, if we have any useful type information, check that
     // the type declares the appropriate method.
   } else if (iface || !objectType->qual_empty()) {
-    IdentifierInfo *selectorIdents[] = {
-      &Context.Idents.get("countByEnumeratingWithState"),
-      &Context.Idents.get("objects"),
-      &Context.Idents.get("count")
-    };
+    const IdentifierInfo *selectorIdents[] = {
+        &Context.Idents.get("countByEnumeratingWithState"),
+        &Context.Idents.get("objects"), &Context.Idents.get("count")};
     Selector selector = Context.Selectors.getSelector(3, &selectorIdents[0]);
 
     ObjCMethodDecl *method = nullptr;
diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
index 2013799b5eb81..951e5a31cab3b 100644
--- a/clang/lib/Sema/SemaTemplate.cpp
+++ b/clang/lib/Sema/SemaTemplate.cpp
@@ -970,7 +970,7 @@ void Sema::translateTemplateArguments(const ASTTemplateArgsPtr &TemplateArgsIn,
 
 static void maybeDiagnoseTemplateParameterShadow(Sema &SemaRef, Scope *S,
                                                  SourceLocation Loc,
-                                                 IdentifierInfo *Name) {
+                                                 const IdentifierInfo *Name) {
   NamedDecl *PrevDecl = SemaRef.LookupSingleName(
       S, Name, Loc, Sema::LookupOrdinaryName, Sema::ForVisibleRedeclaration);
   if (PrevDecl && PrevDecl->isTemplateParameter())
@@ -1578,7 +1578,7 @@ NamedDecl *Sema::ActOnNonTypeTemplateParameter(Scope *S, Declarator &D,
 
   CheckFunctionOrTemplateParamDeclarator(S, D);
 
-  IdentifierInfo *ParamName = D.getIdentifier();
+  const IdentifierInfo *ParamName = D.getIdentifier();
   bool IsParameterPack = D.hasEllipsis();
   NonTypeTemplateParmDecl *Param = NonTypeTemplateParmDecl::Create(
       Context, Context.getTranslationUnitDecl(), D.getBeginLoc(),
@@ -4702,7 +4702,7 @@ bool Sema::resolveAssumedTemplateNameAsType(Scope *S, TemplateName &Name,
 
 TypeResult Sema::ActOnTemplateIdType(
     Scope *S, CXXScopeSpec &SS, SourceLocation TemplateKWLoc,
-    TemplateTy TemplateD, IdentifierInfo *TemplateII,
+    TemplateTy TemplateD, const IdentifierInfo *TemplateII,
     SourceLocation TemplateIILoc, SourceLocation LAngleLoc,
     ASTTemplateArgsPtr TemplateArgsIn, SourceLocation RAngleLoc,
     bool IsCtorOrDtorName, bool IsClassName,
@@ -9684,10 +9684,9 @@ Decl *Sema::ActOnTemplateDeclarator(Scope *S,
   return NewDecl;
 }
 
-Decl *Sema::ActOnConceptDefinition(Scope *S,
-                              MultiTemplateParamsArg TemplateParameterLists,
-                                   IdentifierInfo *Name, SourceLocation NameLoc,
-                                   Expr *ConstraintExpr) {
+Decl *Sema::ActOnConceptDefinition(
+    Scope *S, MultiTemplateParamsArg TemplateParameterLists,
+    const IdentifierInfo *Name, SourceLocation NameLoc, Expr *ConstraintExpr) {
   DeclContext *DC = CurContext;
 
   if (!DC->getRedeclContext()->isFileContext()) {
@@ -11511,10 +11510,11 @@ DeclResult Sema::ActOnExplicitInstantiation(Scope *S,
   return (Decl*) nullptr;
 }
 
-TypeResult
-Sema::ActOnDependentTag(Scope *S, unsigned TagSpec, TagUseKind TUK,
-                        const CXXScopeSpec &SS, IdentifierInfo *Name,
-                        SourceLocation TagLoc, SourceLocation NameLoc) {
+TypeResult Sema::ActOnDependentTag(Scope *S, unsigned TagSpec, TagUseKind TUK,
+                                   const CXXScopeSpec &SS,
+                                   const IdentifierInfo *Name,
+                                   SourceLocation TagLoc,
+                                   SourceLocation NameLoc) {
   // This has to hold, because SS is expected to be defined.
   assert(Name && "Expected a name in a dependent tag");
 
@@ -11574,14 +11574,10 @@ TypeResult Sema::ActOnTypenameType(Scope *S, SourceLocation TypenameLoc,
 }
 
 TypeResult
-Sema::ActOnTypenameType(Scope *S,
-                        SourceLocation TypenameLoc,
-                        const CXXScopeSpec &SS,
-                        SourceLocation TemplateKWLoc,
-                        TemplateTy TemplateIn,
-                        IdentifierInfo *TemplateII,
-                        SourceLocation TemplateIILoc,
-                        SourceLocation LAngleLoc,
+Sema::ActOnTypenameType(Scope *S, SourceLocation TypenameLoc,
+                        const CXXScopeSpec &SS, SourceLocation TemplateKWLoc,
+                        TemplateTy TemplateIn, const IdentifierInfo *TemplateII,
+                        SourceLocation TemplateIILoc, SourceLocation LAngleLoc,
                         ASTTemplateArgsPtr TemplateArgsIn,
                         SourceLocation RAngleLoc) {
   if (TypenameLoc.isValid() && S && !S->getTemplateParamParent())
@@ -11657,7 +11653,6 @@ Sema::ActOnTypenameType(Scope *S,
   return CreateParsedType(T, TSI);
 }
 
-
 /// Determine whether this failed name lookup should be treated as being
 /// disabled by a usage of std::enable_if.
 static bool isEnableIf(NestedNameSpecifierLoc NNS, const IdentifierInfo &II,
diff --git a/clang/lib/Serialization/ASTCommon.cpp b/clang/lib/Serialization/ASTCommon.cpp
index 6110e287b7fb5..f8d54c0c39890 100644
--- a/clang/lib/Serialization/ASTCommon.cpp
+++ b/clang/lib/Serialization/ASTCommon.cpp
@@ -284,7 +284,7 @@ unsigned serialization::ComputeHash(Selector Sel) {
     ++N;
   unsigned R = 5381;
   for (unsigned I = 0; I != N; ++I)
-    if (IdentifierInfo *II = Sel.getIdentifierInfoForSlot(I))
+    if (const IdentifierInfo *II = Sel.getIdentifierInfoForSlot(I))
       R = llvm::djbHash(II->getName(), R);
   return R;
 }
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 0ca7f6600eee3..4f6987f92fc82 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -916,14 +916,14 @@ ASTSelectorLookupTrait::ReadKey(const unsigned char* d, unsigned) {
   SelectorTable &SelTable = Reader.getContext().Selectors;
   unsigned N =
       endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
-  IdentifierInfo *FirstII = Reader.getLocalIdentifier(
+  const IdentifierInfo *FirstII = Reader.getLocalIdentifier(
       F, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d));
   if (N == 0)
     return SelTable.getNullarySelector(FirstII);
   else if (N == 1)
     return SelTable.getUnarySelector(FirstII);
 
-  SmallVector<IdentifierInfo *, 16> Args;
+  SmallVector<const IdentifierInfo *, 16> Args;
   Args.push_back(FirstII);
   for (unsigned I = 1; I != N; ++I)
     Args.push_back(Reader.getLocalIdentifier(
@@ -987,7 +987,7 @@ ASTIdentifierLookupTraitBase::ReadKey(const unsigned char* d, unsigned n) {
 }
 
 /// Whether the given identifier is "interesting".
-static bool isInterestingIdentifier(ASTReader &Reader, IdentifierInfo &II,
+static bool isInterestingIdentifier(ASTReader &Reader, const IdentifierInfo &II,
                                     bool IsModule) {
   bool IsInteresting =
       II.getNotableIdentifierID() != tok::NotableIdentifierKind::not_notable ||
@@ -2229,7 +2229,7 @@ namespace {
 
 } // namespace
 
-void ASTReader::updateOutOfDateIdentifier(IdentifierInfo &II) {
+void ASTReader::updateOutOfDateIdentifier(const IdentifierInfo &II) {
   // Note that we are loading an identifier.
   Deserializing AnIdentifier(this);
 
@@ -2254,11 +2254,11 @@ void ASTReader::updateOutOfDateIdentifier(IdentifierInfo &II) {
   markIdentifierUpToDate(&II);
 }
 
-void ASTReader::markIdentifierUpToDate(IdentifierInfo *II) {
+void ASTReader::markIdentifierUpToDate(const IdentifierInfo *II) {
   if (!II)
     return;
 
-  II->setOutOfDate(false);
+  const_cast<IdentifierInfo *>(II)->setOutOfDate(false);
 
   // Update the generation for this identifier.
   if (getContext().getLangOpts().Modules)
@@ -10168,7 +10168,7 @@ void ASTReader::FinishedDeserializing() {
 }
 
 void ASTReader::pushExternalDeclIntoScope(NamedDecl *D, DeclarationName Name) {
-  if (IdentifierInfo *II = Name.getAsIdentifierInfo()) {
+  if (const IdentifierInfo *II = Name.getAsIdentifierInfo()) {
     // Remove any fake results before adding any real ones.
     auto It = PendingFakeLookupResults.find(II);
     if (It != PendingFakeLookupResults.end()) {
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index d2afe378bb0c3..ffc53292e3912 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -3629,7 +3629,7 @@ class ASTIdentifierTableTrait {
   }
 
 public:
-  using key_type = IdentifierInfo *;
+  using key_type = const IdentifierInfo *;
   using key_type_ref = key_type;
 
   using data_type = IdentID;
@@ -3661,7 +3661,7 @@ class ASTIdentifierTableTrait {
   }
 
   std::pair<unsigned, unsigned>
-  EmitKeyDataLength(raw_ostream& Out, IdentifierInfo* II, IdentID ID) {
+  EmitKeyDataLength(raw_ostream &Out, const IdentifierInfo *II, IdentID ID) {
     // Record the location of the identifier data. This is used when generating
     // the mapping from persistent IDs to strings.
     Writer.SetIdentifierOffset(II, Out.tell());
@@ -3688,13 +3688,12 @@ class ASTIdentifierTableTrait {
     return emitULEBKeyDataLength(KeyLen, DataLen, Out);
   }
 
-  void EmitKey(raw_ostream& Out, const IdentifierInfo* II,
-               unsigned KeyLen) {
+  void EmitKey(raw_ostream &Out, const IdentifierInfo *II, unsigned KeyLen) {
     Out.write(II->getNameStart(), KeyLen);
   }
 
-  void EmitData(raw_ostream& Out, IdentifierInfo* II,
-                IdentID ID, unsigned) {
+  void EmitData(raw_ostream &Out, const IdentifierInfo *II, IdentID ID,
+                unsigned) {
     using namespace llvm::support;
 
     endian::Writer LE(Out, llvm::endianness::little);
@@ -3776,13 +3775,14 @@ void ASTWriter::WriteIdentifierTable(Preprocessor &PP,
     // for identifiers that appear here for the first time.
     IdentifierOffsets.resize(NextIdentID - FirstIdentID);
     for (auto IdentIDPair : IdentifierIDs) {
-      auto *II = const_cast<IdentifierInfo *>(IdentIDPair.first);
+      const IdentifierInfo *II = IdentIDPair.first;
       IdentID ID = IdentIDPair.second;
       assert(II && "NULL identifier in identifier table");
+
       // Write out identifiers if either the ID is local or the identifier has
       // changed since it was loaded.
-      if (ID >= FirstIdentID || !Chain || !II->isFromAST()
-          || II->hasChangedSinceDeserialization() ||
+      if (ID >= FirstIdentID || !Chain || !II->isFromAST() ||
+          II->hasChangedSinceDeserialization() ||
           (Trait.needDecls() &&
            II->hasFETokenInfoChangedSinceDeserialization()))
         Generator.insert(II, ID, Trait);
diff --git a/clang/lib/StaticAnalyzer/Checkers/CheckObjCDealloc.cpp b/clang/lib/StaticAnalyzer/Checkers/CheckObjCDealloc.cpp
index 978bc0bb082f8..b4390f0b85bbe 100644
--- a/clang/lib/StaticAnalyzer/Checkers/CheckObjCDealloc.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/CheckObjCDealloc.cpp
@@ -768,8 +768,8 @@ void ObjCDeallocChecker::initIdentifierInfoAndSelectors(
   Block_releaseII = &Ctx.Idents.get("_Block_release");
   CIFilterII = &Ctx.Idents.get("CIFilter");
 
-  IdentifierInfo *DeallocII = &Ctx.Idents.get("dealloc");
-  IdentifierInfo *ReleaseII = &Ctx.Idents.get("release");
+  const IdentifierInfo *DeallocII = &Ctx.Idents.get("dealloc");
+  const IdentifierInfo *ReleaseII = &Ctx.Idents.get("release");
   DeallocSel = Ctx.Selectors.getSelector(0, &DeallocII);
   ReleaseSel = Ctx.Selectors.getSelector(0, &ReleaseII);
 }
diff --git a/clang/lib/StaticAnalyzer/Checkers/LocalizationChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/LocalizationChecker.cpp
index 812d787e2e37c..882eb0236a189 100644
--- a/clang/lib/StaticAnalyzer/Checkers/LocalizationChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/LocalizationChecker.cpp
@@ -154,11 +154,11 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UISearchDisplayController, setSearchResultsTitle, 0)
 
   NEW_RECEIVER(UITabBarItem)
-  IdentifierInfo *initWithTitleUITabBarItemTag[] = {
+  const IdentifierInfo *initWithTitleUITabBarItemTag[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("image"),
       &Ctx.Idents.get("tag")};
   ADD_METHOD(UITabBarItem, initWithTitleUITabBarItemTag, 3, 0)
-  IdentifierInfo *initWithTitleUITabBarItemImage[] = {
+  const IdentifierInfo *initWithTitleUITabBarItemImage[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("image"),
       &Ctx.Idents.get("selectedImage")};
   ADD_METHOD(UITabBarItem, initWithTitleUITabBarItemImage, 3, 0)
@@ -171,7 +171,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSStatusItem, setToolTip, 0)
 
   NEW_RECEIVER(UITableViewRowAction)
-  IdentifierInfo *rowActionWithStyleUITableViewRowAction[] = {
+  const IdentifierInfo *rowActionWithStyleUITableViewRowAction[] = {
       &Ctx.Idents.get("rowActionWithStyle"), &Ctx.Idents.get("title"),
       &Ctx.Idents.get("handler")};
   ADD_METHOD(UITableViewRowAction, rowActionWithStyleUITableViewRowAction, 3, 1)
@@ -183,19 +183,19 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   NEW_RECEIVER(NSButton)
   ADD_UNARY_METHOD(NSButton, setTitle, 0)
   ADD_UNARY_METHOD(NSButton, setAlternateTitle, 0)
-  IdentifierInfo *radioButtonWithTitleNSButton[] = {
+  const IdentifierInfo *radioButtonWithTitleNSButton[] = {
       &Ctx.Idents.get("radioButtonWithTitle"), &Ctx.Idents.get("target"),
       &Ctx.Idents.get("action")};
   ADD_METHOD(NSButton, radioButtonWithTitleNSButton, 3, 0)
-  IdentifierInfo *buttonWithTitleNSButtonImage[] = {
+  const IdentifierInfo *buttonWithTitleNSButtonImage[] = {
       &Ctx.Idents.get("buttonWithTitle"), &Ctx.Idents.get("image"),
       &Ctx.Idents.get("target"), &Ctx.Idents.get("action")};
   ADD_METHOD(NSButton, buttonWithTitleNSButtonImage, 4, 0)
-  IdentifierInfo *checkboxWithTitleNSButton[] = {
+  const IdentifierInfo *checkboxWithTitleNSButton[] = {
       &Ctx.Idents.get("checkboxWithTitle"), &Ctx.Idents.get("target"),
       &Ctx.Idents.get("action")};
   ADD_METHOD(NSButton, checkboxWithTitleNSButton, 3, 0)
-  IdentifierInfo *buttonWithTitleNSButtonTarget[] = {
+  const IdentifierInfo *buttonWithTitleNSButtonTarget[] = {
       &Ctx.Idents.get("buttonWithTitle"), &Ctx.Idents.get("target"),
       &Ctx.Idents.get("action")};
   ADD_METHOD(NSButton, buttonWithTitleNSButtonTarget, 3, 0)
@@ -215,8 +215,8 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSTabViewItem, setToolTip, 0)
 
   NEW_RECEIVER(NSBrowser)
-  IdentifierInfo *setTitleNSBrowser[] = {&Ctx.Idents.get("setTitle"),
-                                         &Ctx.Idents.get("ofColumn")};
+  const IdentifierInfo *setTitleNSBrowser[] = {&Ctx.Idents.get("setTitle"),
+                                               &Ctx.Idents.get("ofColumn")};
   ADD_METHOD(NSBrowser, setTitleNSBrowser, 2, 0)
 
   NEW_RECEIVER(UIAccessibilityElement)
@@ -225,14 +225,14 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIAccessibilityElement, setAccessibilityValue, 0)
 
   NEW_RECEIVER(UIAlertAction)
-  IdentifierInfo *actionWithTitleUIAlertAction[] = {
+  const IdentifierInfo *actionWithTitleUIAlertAction[] = {
       &Ctx.Idents.get("actionWithTitle"), &Ctx.Idents.get("style"),
       &Ctx.Idents.get("handler")};
   ADD_METHOD(UIAlertAction, actionWithTitleUIAlertAction, 3, 0)
 
   NEW_RECEIVER(NSPopUpButton)
   ADD_UNARY_METHOD(NSPopUpButton, addItemWithTitle, 0)
-  IdentifierInfo *insertItemWithTitleNSPopUpButton[] = {
+  const IdentifierInfo *insertItemWithTitleNSPopUpButton[] = {
       &Ctx.Idents.get("insertItemWithTitle"), &Ctx.Idents.get("atIndex")};
   ADD_METHOD(NSPopUpButton, insertItemWithTitleNSPopUpButton, 2, 0)
   ADD_UNARY_METHOD(NSPopUpButton, removeItemWithTitle, 0)
@@ -240,7 +240,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSPopUpButton, setTitle, 0)
 
   NEW_RECEIVER(NSTableViewRowAction)
-  IdentifierInfo *rowActionWithStyleNSTableViewRowAction[] = {
+  const IdentifierInfo *rowActionWithStyleNSTableViewRowAction[] = {
       &Ctx.Idents.get("rowActionWithStyle"), &Ctx.Idents.get("title"),
       &Ctx.Idents.get("handler")};
   ADD_METHOD(NSTableViewRowAction, rowActionWithStyleNSTableViewRowAction, 3, 1)
@@ -273,10 +273,10 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSTableColumn, setHeaderToolTip, 0)
 
   NEW_RECEIVER(NSSegmentedControl)
-  IdentifierInfo *setLabelNSSegmentedControl[] = {
+  const IdentifierInfo *setLabelNSSegmentedControl[] = {
       &Ctx.Idents.get("setLabel"), &Ctx.Idents.get("forSegment")};
   ADD_METHOD(NSSegmentedControl, setLabelNSSegmentedControl, 2, 0)
-  IdentifierInfo *setToolTipNSSegmentedControl[] = {
+  const IdentifierInfo *setToolTipNSSegmentedControl[] = {
       &Ctx.Idents.get("setToolTip"), &Ctx.Idents.get("forSegment")};
   ADD_METHOD(NSSegmentedControl, setToolTipNSSegmentedControl, 2, 0)
 
@@ -301,8 +301,8 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSAccessibility, setAccessibilityHelp, 0)
 
   NEW_RECEIVER(NSMatrix)
-  IdentifierInfo *setToolTipNSMatrix[] = {&Ctx.Idents.get("setToolTip"),
-                                          &Ctx.Idents.get("forCell")};
+  const IdentifierInfo *setToolTipNSMatrix[] = {&Ctx.Idents.get("setToolTip"),
+                                                &Ctx.Idents.get("forCell")};
   ADD_METHOD(NSMatrix, setToolTipNSMatrix, 2, 0)
 
   NEW_RECEIVER(NSPrintPanel)
@@ -317,13 +317,13 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSSlider, setTitle, 0)
 
   NEW_RECEIVER(UIMenuItem)
-  IdentifierInfo *initWithTitleUIMenuItem[] = {&Ctx.Idents.get("initWithTitle"),
-                                               &Ctx.Idents.get("action")};
+  const IdentifierInfo *initWithTitleUIMenuItem[] = {
+      &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("action")};
   ADD_METHOD(UIMenuItem, initWithTitleUIMenuItem, 2, 0)
   ADD_UNARY_METHOD(UIMenuItem, setTitle, 0)
 
   NEW_RECEIVER(UIAlertController)
-  IdentifierInfo *alertControllerWithTitleUIAlertController[] = {
+  const IdentifierInfo *alertControllerWithTitleUIAlertController[] = {
       &Ctx.Idents.get("alertControllerWithTitle"), &Ctx.Idents.get("message"),
       &Ctx.Idents.get("preferredStyle")};
   ADD_METHOD(UIAlertController, alertControllerWithTitleUIAlertController, 3, 1)
@@ -331,19 +331,19 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIAlertController, setMessage, 0)
 
   NEW_RECEIVER(UIApplicationShortcutItem)
-  IdentifierInfo *initWithTypeUIApplicationShortcutItemIcon[] = {
+  const IdentifierInfo *initWithTypeUIApplicationShortcutItemIcon[] = {
       &Ctx.Idents.get("initWithType"), &Ctx.Idents.get("localizedTitle"),
       &Ctx.Idents.get("localizedSubtitle"), &Ctx.Idents.get("icon"),
       &Ctx.Idents.get("userInfo")};
   ADD_METHOD(UIApplicationShortcutItem,
              initWithTypeUIApplicationShortcutItemIcon, 5, 1)
-  IdentifierInfo *initWithTypeUIApplicationShortcutItem[] = {
+  const IdentifierInfo *initWithTypeUIApplicationShortcutItem[] = {
       &Ctx.Idents.get("initWithType"), &Ctx.Idents.get("localizedTitle")};
   ADD_METHOD(UIApplicationShortcutItem, initWithTypeUIApplicationShortcutItem,
              2, 1)
 
   NEW_RECEIVER(UIActionSheet)
-  IdentifierInfo *initWithTitleUIActionSheet[] = {
+  const IdentifierInfo *initWithTitleUIActionSheet[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("delegate"),
       &Ctx.Idents.get("cancelButtonTitle"),
       &Ctx.Idents.get("destructiveButtonTitle"),
@@ -353,7 +353,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIActionSheet, setTitle, 0)
 
   NEW_RECEIVER(UIAccessibilityCustomAction)
-  IdentifierInfo *initWithNameUIAccessibilityCustomAction[] = {
+  const IdentifierInfo *initWithNameUIAccessibilityCustomAction[] = {
       &Ctx.Idents.get("initWithName"), &Ctx.Idents.get("target"),
       &Ctx.Idents.get("selector")};
   ADD_METHOD(UIAccessibilityCustomAction,
@@ -382,7 +382,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
 
   NEW_RECEIVER(NSAttributedString)
   ADD_UNARY_METHOD(NSAttributedString, initWithString, 0)
-  IdentifierInfo *initWithStringNSAttributedString[] = {
+  const IdentifierInfo *initWithStringNSAttributedString[] = {
       &Ctx.Idents.get("initWithString"), &Ctx.Idents.get("attributes")};
   ADD_METHOD(NSAttributedString, initWithStringNSAttributedString, 2, 0)
 
@@ -390,7 +390,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSText, setString, 0)
 
   NEW_RECEIVER(UIKeyCommand)
-  IdentifierInfo *keyCommandWithInputUIKeyCommand[] = {
+  const IdentifierInfo *keyCommandWithInputUIKeyCommand[] = {
       &Ctx.Idents.get("keyCommandWithInput"), &Ctx.Idents.get("modifierFlags"),
       &Ctx.Idents.get("action"), &Ctx.Idents.get("discoverabilityTitle")};
   ADD_METHOD(UIKeyCommand, keyCommandWithInputUIKeyCommand, 4, 3)
@@ -400,7 +400,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UILabel, setText, 0)
 
   NEW_RECEIVER(NSAlert)
-  IdentifierInfo *alertWithMessageTextNSAlert[] = {
+  const IdentifierInfo *alertWithMessageTextNSAlert[] = {
       &Ctx.Idents.get("alertWithMessageText"), &Ctx.Idents.get("defaultButton"),
       &Ctx.Idents.get("alternateButton"), &Ctx.Idents.get("otherButton"),
       &Ctx.Idents.get("informativeTextWithFormat")};
@@ -415,13 +415,13 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIMutableApplicationShortcutItem, setLocalizedSubtitle, 0)
 
   NEW_RECEIVER(UIButton)
-  IdentifierInfo *setTitleUIButton[] = {&Ctx.Idents.get("setTitle"),
-                                        &Ctx.Idents.get("forState")};
+  const IdentifierInfo *setTitleUIButton[] = {&Ctx.Idents.get("setTitle"),
+                                              &Ctx.Idents.get("forState")};
   ADD_METHOD(UIButton, setTitleUIButton, 2, 0)
 
   NEW_RECEIVER(NSWindow)
   ADD_UNARY_METHOD(NSWindow, setTitle, 0)
-  IdentifierInfo *minFrameWidthWithTitleNSWindow[] = {
+  const IdentifierInfo *minFrameWidthWithTitleNSWindow[] = {
       &Ctx.Idents.get("minFrameWidthWithTitle"), &Ctx.Idents.get("styleMask")};
   ADD_METHOD(NSWindow, minFrameWidthWithTitleNSWindow, 2, 0)
   ADD_UNARY_METHOD(NSWindow, setMiniwindowTitle, 0)
@@ -430,7 +430,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSPathCell, setPlaceholderString, 0)
 
   NEW_RECEIVER(UIDocumentMenuViewController)
-  IdentifierInfo *addOptionWithTitleUIDocumentMenuViewController[] = {
+  const IdentifierInfo *addOptionWithTitleUIDocumentMenuViewController[] = {
       &Ctx.Idents.get("addOptionWithTitle"), &Ctx.Idents.get("image"),
       &Ctx.Idents.get("order"), &Ctx.Idents.get("handler")};
   ADD_METHOD(UIDocumentMenuViewController,
@@ -442,7 +442,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UINavigationItem, setPrompt, 0)
 
   NEW_RECEIVER(UIAlertView)
-  IdentifierInfo *initWithTitleUIAlertView[] = {
+  const IdentifierInfo *initWithTitleUIAlertView[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("message"),
       &Ctx.Idents.get("delegate"), &Ctx.Idents.get("cancelButtonTitle"),
       &Ctx.Idents.get("otherButtonTitles")};
@@ -474,11 +474,11 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSProgress, setLocalizedAdditionalDescription, 0)
 
   NEW_RECEIVER(NSSegmentedCell)
-  IdentifierInfo *setLabelNSSegmentedCell[] = {&Ctx.Idents.get("setLabel"),
-                                               &Ctx.Idents.get("forSegment")};
+  const IdentifierInfo *setLabelNSSegmentedCell[] = {
+      &Ctx.Idents.get("setLabel"), &Ctx.Idents.get("forSegment")};
   ADD_METHOD(NSSegmentedCell, setLabelNSSegmentedCell, 2, 0)
-  IdentifierInfo *setToolTipNSSegmentedCell[] = {&Ctx.Idents.get("setToolTip"),
-                                                 &Ctx.Idents.get("forSegment")};
+  const IdentifierInfo *setToolTipNSSegmentedCell[] = {
+      &Ctx.Idents.get("setToolTip"), &Ctx.Idents.get("forSegment")};
   ADD_METHOD(NSSegmentedCell, setToolTipNSSegmentedCell, 2, 0)
 
   NEW_RECEIVER(NSUndoManager)
@@ -487,7 +487,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSUndoManager, redoMenuTitleForUndoActionName, 0)
 
   NEW_RECEIVER(NSMenuItem)
-  IdentifierInfo *initWithTitleNSMenuItem[] = {
+  const IdentifierInfo *initWithTitleNSMenuItem[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("action"),
       &Ctx.Idents.get("keyEquivalent")};
   ADD_METHOD(NSMenuItem, initWithTitleNSMenuItem, 3, 0)
@@ -495,11 +495,11 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSMenuItem, setToolTip, 0)
 
   NEW_RECEIVER(NSPopUpButtonCell)
-  IdentifierInfo *initTextCellNSPopUpButtonCell[] = {
+  const IdentifierInfo *initTextCellNSPopUpButtonCell[] = {
       &Ctx.Idents.get("initTextCell"), &Ctx.Idents.get("pullsDown")};
   ADD_METHOD(NSPopUpButtonCell, initTextCellNSPopUpButtonCell, 2, 0)
   ADD_UNARY_METHOD(NSPopUpButtonCell, addItemWithTitle, 0)
-  IdentifierInfo *insertItemWithTitleNSPopUpButtonCell[] = {
+  const IdentifierInfo *insertItemWithTitleNSPopUpButtonCell[] = {
       &Ctx.Idents.get("insertItemWithTitle"), &Ctx.Idents.get("atIndex")};
   ADD_METHOD(NSPopUpButtonCell, insertItemWithTitleNSPopUpButtonCell, 2, 0)
   ADD_UNARY_METHOD(NSPopUpButtonCell, removeItemWithTitle, 0)
@@ -511,11 +511,11 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
 
   NEW_RECEIVER(NSMenu)
   ADD_UNARY_METHOD(NSMenu, initWithTitle, 0)
-  IdentifierInfo *insertItemWithTitleNSMenu[] = {
+  const IdentifierInfo *insertItemWithTitleNSMenu[] = {
       &Ctx.Idents.get("insertItemWithTitle"), &Ctx.Idents.get("action"),
       &Ctx.Idents.get("keyEquivalent"), &Ctx.Idents.get("atIndex")};
   ADD_METHOD(NSMenu, insertItemWithTitleNSMenu, 4, 0)
-  IdentifierInfo *addItemWithTitleNSMenu[] = {
+  const IdentifierInfo *addItemWithTitleNSMenu[] = {
       &Ctx.Idents.get("addItemWithTitle"), &Ctx.Idents.get("action"),
       &Ctx.Idents.get("keyEquivalent")};
   ADD_METHOD(NSMenu, addItemWithTitleNSMenu, 3, 0)
@@ -526,15 +526,15 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
 
   NEW_RECEIVER(NSForm)
   ADD_UNARY_METHOD(NSForm, addEntry, 0)
-  IdentifierInfo *insertEntryNSForm[] = {&Ctx.Idents.get("insertEntry"),
-                                         &Ctx.Idents.get("atIndex")};
+  const IdentifierInfo *insertEntryNSForm[] = {&Ctx.Idents.get("insertEntry"),
+                                               &Ctx.Idents.get("atIndex")};
   ADD_METHOD(NSForm, insertEntryNSForm, 2, 0)
 
   NEW_RECEIVER(NSTextFieldCell)
   ADD_UNARY_METHOD(NSTextFieldCell, setPlaceholderString, 0)
 
   NEW_RECEIVER(NSUserNotificationAction)
-  IdentifierInfo *actionWithIdentifierNSUserNotificationAction[] = {
+  const IdentifierInfo *actionWithIdentifierNSUserNotificationAction[] = {
       &Ctx.Idents.get("actionWithIdentifier"), &Ctx.Idents.get("title")};
   ADD_METHOD(NSUserNotificationAction,
              actionWithIdentifierNSUserNotificationAction, 2, 1)
@@ -544,7 +544,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UITextField, setPlaceholder, 0)
 
   NEW_RECEIVER(UIBarButtonItem)
-  IdentifierInfo *initWithTitleUIBarButtonItem[] = {
+  const IdentifierInfo *initWithTitleUIBarButtonItem[] = {
       &Ctx.Idents.get("initWithTitle"), &Ctx.Idents.get("style"),
       &Ctx.Idents.get("target"), &Ctx.Idents.get("action")};
   ADD_METHOD(UIBarButtonItem, initWithTitleUIBarButtonItem, 4, 0)
@@ -553,16 +553,16 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIViewController, setTitle, 0)
 
   NEW_RECEIVER(UISegmentedControl)
-  IdentifierInfo *insertSegmentWithTitleUISegmentedControl[] = {
+  const IdentifierInfo *insertSegmentWithTitleUISegmentedControl[] = {
       &Ctx.Idents.get("insertSegmentWithTitle"), &Ctx.Idents.get("atIndex"),
       &Ctx.Idents.get("animated")};
   ADD_METHOD(UISegmentedControl, insertSegmentWithTitleUISegmentedControl, 3, 0)
-  IdentifierInfo *setTitleUISegmentedControl[] = {
+  const IdentifierInfo *setTitleUISegmentedControl[] = {
       &Ctx.Idents.get("setTitle"), &Ctx.Idents.get("forSegmentAtIndex")};
   ADD_METHOD(UISegmentedControl, setTitleUISegmentedControl, 2, 0)
 
   NEW_RECEIVER(NSAccessibilityCustomRotorItemResult)
-  IdentifierInfo
+  const IdentifierInfo
       *initWithItemLoadingTokenNSAccessibilityCustomRotorItemResult[] = {
           &Ctx.Idents.get("initWithItemLoadingToken"),
           &Ctx.Idents.get("customLabel")};
@@ -571,7 +571,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSAccessibilityCustomRotorItemResult, setCustomLabel, 0)
 
   NEW_RECEIVER(UIContextualAction)
-  IdentifierInfo *contextualActionWithStyleUIContextualAction[] = {
+  const IdentifierInfo *contextualActionWithStyleUIContextualAction[] = {
       &Ctx.Idents.get("contextualActionWithStyle"), &Ctx.Idents.get("title"),
       &Ctx.Idents.get("handler")};
   ADD_METHOD(UIContextualAction, contextualActionWithStyleUIContextualAction, 3,
@@ -579,7 +579,7 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(UIContextualAction, setTitle, 0)
 
   NEW_RECEIVER(NSAccessibilityCustomRotor)
-  IdentifierInfo *initWithLabelNSAccessibilityCustomRotor[] = {
+  const IdentifierInfo *initWithLabelNSAccessibilityCustomRotor[] = {
       &Ctx.Idents.get("initWithLabel"), &Ctx.Idents.get("itemSearchDelegate")};
   ADD_METHOD(NSAccessibilityCustomRotor,
              initWithLabelNSAccessibilityCustomRotor, 2, 0)
@@ -590,11 +590,11 @@ void NonLocalizedStringChecker::initUIMethods(ASTContext &Ctx) const {
   ADD_UNARY_METHOD(NSWindowTab, setToolTip, 0)
 
   NEW_RECEIVER(NSAccessibilityCustomAction)
-  IdentifierInfo *initWithNameNSAccessibilityCustomAction[] = {
+  const IdentifierInfo *initWithNameNSAccessibilityCustomAction[] = {
       &Ctx.Idents.get("initWithName"), &Ctx.Idents.get("handler")};
   ADD_METHOD(NSAccessibilityCustomAction,
              initWithNameNSAccessibilityCustomAction, 2, 0)
-  IdentifierInfo *initWithNameTargetNSAccessibilityCustomAction[] = {
+  const IdentifierInfo *initWithNameTargetNSAccessibilityCustomAction[] = {
       &Ctx.Idents.get("initWithName"), &Ctx.Idents.get("target"),
       &Ctx.Idents.get("selector")};
   ADD_METHOD(NSAccessibilityCustomAction,
@@ -618,12 +618,12 @@ void NonLocalizedStringChecker::initLocStringsMethods(ASTContext &Ctx) const {
   if (!LSM.empty())
     return;
 
-  IdentifierInfo *LocalizedStringMacro[] = {
+  const IdentifierInfo *LocalizedStringMacro[] = {
       &Ctx.Idents.get("localizedStringForKey"), &Ctx.Idents.get("value"),
       &Ctx.Idents.get("table")};
   LSM_INSERT_SELECTOR("NSBundle", LocalizedStringMacro, 3)
   LSM_INSERT_UNARY("NSDateFormatter", "stringFromDate")
-  IdentifierInfo *LocalizedStringFromDate[] = {
+  const IdentifierInfo *LocalizedStringFromDate[] = {
       &Ctx.Idents.get("localizedStringFromDate"), &Ctx.Idents.get("dateStyle"),
       &Ctx.Idents.get("timeStyle")};
   LSM_INSERT_SELECTOR("NSDateFormatter", LocalizedStringFromDate, 3)
@@ -903,7 +903,7 @@ static inline bool isNSStringType(QualType T, ASTContext &Ctx) {
   if (!Cls)
     return false;
 
-  IdentifierInfo *ClsName = Cls->getIdentifier();
+  const IdentifierInfo *ClsName = Cls->getIdentifier();
 
   // FIXME: Should we walk the chain of classes?
   return ClsName == &Ctx.Idents.get("NSString") ||
diff --git a/clang/lib/StaticAnalyzer/Checkers/NullabilityChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/NullabilityChecker.cpp
index 06f1ad00eaf20..60934e51febe8 100644
--- a/clang/lib/StaticAnalyzer/Checkers/NullabilityChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/NullabilityChecker.cpp
@@ -1082,7 +1082,8 @@ void NullabilityChecker::checkPostObjCMessage(const ObjCMethodCall &M,
       M.getMessageKind() == OCM_PropertyAccess && !C.wasInlined) {
     bool LookupResolved = false;
     if (const MemRegion *ReceiverRegion = getTrackRegion(M.getReceiverSVal())) {
-      if (IdentifierInfo *Ident = M.getSelector().getIdentifierInfoForSlot(0)) {
+      if (const IdentifierInfo *Ident =
+              M.getSelector().getIdentifierInfoForSlot(0)) {
         LookupResolved = true;
         ObjectPropPair Key = std::make_pair(ReceiverRegion, Ident);
         const ConstrainedPropertyVal *PrevPropVal =
diff --git a/clang/lib/StaticAnalyzer/Checkers/ObjCMissingSuperCallChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/ObjCMissingSuperCallChecker.cpp
index 598b368e74d47..03dab4f7ada7a 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ObjCMissingSuperCallChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ObjCMissingSuperCallChecker.cpp
@@ -107,7 +107,7 @@ void ObjCSuperCallChecker::fillSelectors(ASTContext &Ctx,
     assert(Descriptor.ArgumentCount <= 1); // No multi-argument selectors yet.
 
     // Get the selector.
-    IdentifierInfo *II = &Ctx.Idents.get(Descriptor.SelectorName);
+    const IdentifierInfo *II = &Ctx.Idents.get(Descriptor.SelectorName);
 
     Selector Sel = Ctx.Selectors.getSelector(Descriptor.ArgumentCount, &II);
     ClassSelectors.insert(Sel);
diff --git a/clang/lib/StaticAnalyzer/Checkers/ObjCSuperDeallocChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/ObjCSuperDeallocChecker.cpp
index eb40711812e16..a6c4186cb15bb 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ObjCSuperDeallocChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ObjCSuperDeallocChecker.cpp
@@ -26,8 +26,8 @@ namespace {
 class ObjCSuperDeallocChecker
     : public Checker<check::PostObjCMessage, check::PreObjCMessage,
                      check::PreCall, check::Location> {
-  mutable IdentifierInfo *IIdealloc = nullptr;
-  mutable IdentifierInfo *IINSObject = nullptr;
+  mutable const IdentifierInfo *IIdealloc = nullptr;
+  mutable const IdentifierInfo *IINSObject = nullptr;
   mutable Selector SELdealloc;
 
   const BugType DoubleSuperDeallocBugType{
diff --git a/clang/tools/libclang/CIndexCodeCompletion.cpp b/clang/tools/libclang/CIndexCodeCompletion.cpp
index 3c5f390f6d888..850c004680fd9 100644
--- a/clang/tools/libclang/CIndexCodeCompletion.cpp
+++ b/clang/tools/libclang/CIndexCodeCompletion.cpp
@@ -601,15 +601,15 @@ namespace {
       AllocatedResults.Contexts = getContextsForContextKind(contextKind, S);
       
       AllocatedResults.Selector = "";
-      ArrayRef<IdentifierInfo *> SelIdents = Context.getSelIdents();
-      for (ArrayRef<IdentifierInfo *>::iterator I = SelIdents.begin(),
-                                                E = SelIdents.end();
+      ArrayRef<const IdentifierInfo *> SelIdents = Context.getSelIdents();
+      for (ArrayRef<const IdentifierInfo *>::iterator I = SelIdents.begin(),
+                                                      E = SelIdents.end();
            I != E; ++I) {
-        if (IdentifierInfo *selIdent = *I)
+        if (const IdentifierInfo *selIdent = *I)
           AllocatedResults.Selector += selIdent->getName();
         AllocatedResults.Selector += ":";
       }
-      
+
       QualType baseType = Context.getBaseType();
       NamedDecl *D = nullptr;
 
diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ASTUtils.h b/lldb/source/Plugins/ExpressionParser/Clang/ASTUtils.h
index 95e8a600f8382..cefec15a79809 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ASTUtils.h
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ASTUtils.h
@@ -56,7 +56,7 @@ class ExternalASTSourceWrapper : public clang::ExternalSemaSource {
     return m_Source->GetExternalCXXBaseSpecifiers(Offset);
   }
 
-  void updateOutOfDateIdentifier(clang::IdentifierInfo &II) override {
+  void updateOutOfDateIdentifier(const clang::IdentifierInfo &II) override {
     m_Source->updateOutOfDateIdentifier(II);
   }
 
diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangASTSource.cpp b/lldb/source/Plugins/ExpressionParser/Clang/ClangASTSource.cpp
index a95a9e9f01e3f..75493eb10d731 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangASTSource.cpp
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangASTSource.cpp
@@ -713,17 +713,18 @@ bool ClangASTSource::FindObjCMethodDeclsWithOrigin(
   Selector original_selector;
 
   if (decl_name.isObjCZeroArgSelector()) {
-    IdentifierInfo *ident = &original_ctx->Idents.get(decl_name.getAsString());
+    const IdentifierInfo *ident =
+        &original_ctx->Idents.get(decl_name.getAsString());
     original_selector = original_ctx->Selectors.getSelector(0, &ident);
   } else if (decl_name.isObjCOneArgSelector()) {
     const std::string &decl_name_string = decl_name.getAsString();
     std::string decl_name_string_without_colon(decl_name_string.c_str(),
                                                decl_name_string.length() - 1);
-    IdentifierInfo *ident =
+    const IdentifierInfo *ident =
         &original_ctx->Idents.get(decl_name_string_without_colon);
     original_selector = original_ctx->Selectors.getSelector(1, &ident);
   } else {
-    SmallVector<IdentifierInfo *, 4> idents;
+    SmallVector<const IdentifierInfo *, 4> idents;
 
     clang::Selector sel = decl_name.getObjCSelector();
 
diff --git a/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCDeclVendor.cpp b/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCDeclVendor.cpp
index 5d7c5f38d1805..6894cdccaf95a 100644
--- a/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCDeclVendor.cpp
+++ b/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCDeclVendor.cpp
@@ -316,7 +316,7 @@ class ObjCRuntimeMethodType {
     const bool HasRelatedResultType = false;
     const bool for_expression = true;
 
-    std::vector<clang::IdentifierInfo *> selector_components;
+    std::vector<const clang::IdentifierInfo *> selector_components;
 
     const char *name_cursor = name;
     bool is_zero_argument = true;
@@ -335,7 +335,7 @@ class ObjCRuntimeMethodType {
       }
     }
 
-    clang::IdentifierInfo **identifier_infos = selector_components.data();
+    const clang::IdentifierInfo **identifier_infos = selector_components.data();
     if (!identifier_infos) {
       return nullptr;
     }
diff --git a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
index 4a1c8d5765521..ee634d12b3c41 100644
--- a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
+++ b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
@@ -7910,14 +7910,14 @@ bool TypeSystemClang::AddObjCClassProperty(
   if (property_setter_name) {
     std::string property_setter_no_colon(property_setter_name,
                                          strlen(property_setter_name) - 1);
-    clang::IdentifierInfo *setter_ident =
+    const clang::IdentifierInfo *setter_ident =
         &clang_ast.Idents.get(property_setter_no_colon);
     setter_sel = clang_ast.Selectors.getSelector(1, &setter_ident);
   } else if (!(property_attributes & DW_APPLE_PROPERTY_readonly)) {
     std::string setter_sel_string("set");
     setter_sel_string.push_back(::toupper(property_name[0]));
     setter_sel_string.append(&property_name[1]);
-    clang::IdentifierInfo *setter_ident =
+    const clang::IdentifierInfo *setter_ident =
         &clang_ast.Idents.get(setter_sel_string);
     setter_sel = clang_ast.Selectors.getSelector(1, &setter_ident);
   }
@@ -7925,11 +7925,12 @@ bool TypeSystemClang::AddObjCClassProperty(
   property_decl->setPropertyAttributes(ObjCPropertyAttribute::kind_setter);
 
   if (property_getter_name != nullptr) {
-    clang::IdentifierInfo *getter_ident =
+    const clang::IdentifierInfo *getter_ident =
         &clang_ast.Idents.get(property_getter_name);
     getter_sel = clang_ast.Selectors.getSelector(0, &getter_ident);
   } else {
-    clang::IdentifierInfo *getter_ident = &clang_ast.Idents.get(property_name);
+    const clang::IdentifierInfo *getter_ident =
+        &clang_ast.Idents.get(property_name);
     getter_sel = clang_ast.Selectors.getSelector(0, &getter_ident);
   }
   property_decl->setGetterName(getter_sel);
@@ -8091,7 +8092,7 @@ clang::ObjCMethodDecl *TypeSystemClang::AddMethodToObjCObjectType(
     return nullptr;
 
   selector_start++;
-  llvm::SmallVector<clang::IdentifierInfo *, 12> selector_idents;
+  llvm::SmallVector<const clang::IdentifierInfo *, 12> selector_idents;
 
   size_t len = 0;
   const char *start;

From 51f1681424f1a8ccf1e3432d71c341e799597171 Mon Sep 17 00:00:00 2001
From: Owen Pan <owenpiano@gmail.com>
Date: Wed, 10 Apr 2024 19:06:29 -0700
Subject: [PATCH 086/886] [clang-format] Don't merge a short block for
 SBS_Never (#88238)

Also fix unit tests.

Fixes #87484.
---
 clang/lib/Format/FormatToken.h                |  2 +
 clang/lib/Format/UnwrappedLineFormatter.cpp   |  8 +-
 clang/lib/Format/UnwrappedLineParser.cpp      |  7 +-
 clang/unittests/Format/BracesRemoverTest.cpp  |  4 +-
 clang/unittests/Format/FormatTest.cpp         | 84 +++++++++++++++----
 .../Format/FormatTestMacroExpansion.cpp       |  7 +-
 clang/unittests/Format/TokenAnnotatorTest.cpp | 24 ++++++
 7 files changed, 110 insertions(+), 26 deletions(-)

diff --git a/clang/lib/Format/FormatToken.h b/clang/lib/Format/FormatToken.h
index 48b6a9092a8c0..f651e6228c206 100644
--- a/clang/lib/Format/FormatToken.h
+++ b/clang/lib/Format/FormatToken.h
@@ -35,6 +35,8 @@ namespace format {
   TYPE(BinaryOperator)                                                         \
   TYPE(BitFieldColon)                                                          \
   TYPE(BlockComment)                                                           \
+  /* l_brace of a block that is not the body of a (e.g. loop) statement. */    \
+  TYPE(BlockLBrace)                                                            \
   TYPE(BracedListLBrace)                                                       \
   /* The colon at the end of a case label. */                                  \
   TYPE(CaseLabelColon)                                                         \
diff --git a/clang/lib/Format/UnwrappedLineFormatter.cpp b/clang/lib/Format/UnwrappedLineFormatter.cpp
index fb31980ab9f49..4ae54e56331bd 100644
--- a/clang/lib/Format/UnwrappedLineFormatter.cpp
+++ b/clang/lib/Format/UnwrappedLineFormatter.cpp
@@ -796,8 +796,12 @@ class LineJoiner {
       }
     }
 
-    if (const auto *LastNonComment = Line.getLastNonComment();
-        LastNonComment && LastNonComment->is(tok::l_brace)) {
+    if (Line.endsWith(tok::l_brace)) {
+      if (Style.AllowShortBlocksOnASingleLine == FormatStyle::SBS_Never &&
+          Line.First->is(TT_BlockLBrace)) {
+        return 0;
+      }
+
       if (IsSplitBlock && Line.First == Line.Last &&
           I > AnnotatedLines.begin() &&
           (I[-1]->endsWith(tok::kw_else) || IsCtrlStmt(*I[-1]))) {
diff --git a/clang/lib/Format/UnwrappedLineParser.cpp b/clang/lib/Format/UnwrappedLineParser.cpp
index c1f7e2874beb2..603268f771ac5 100644
--- a/clang/lib/Format/UnwrappedLineParser.cpp
+++ b/clang/lib/Format/UnwrappedLineParser.cpp
@@ -395,9 +395,10 @@ bool UnwrappedLineParser::parseLevel(const FormatToken *OpeningBrace,
         ParseDefault();
         continue;
       }
-      if (!InRequiresExpression && FormatTok->isNot(TT_MacroBlockBegin) &&
-          tryToParseBracedList()) {
-        continue;
+      if (!InRequiresExpression && FormatTok->isNot(TT_MacroBlockBegin)) {
+        if (tryToParseBracedList())
+          continue;
+        FormatTok->setFinalizedType(TT_BlockLBrace);
       }
       parseBlock();
       ++StatementCount;
diff --git a/clang/unittests/Format/BracesRemoverTest.cpp b/clang/unittests/Format/BracesRemoverTest.cpp
index 5155eefb9e08c..2e983b887ffcb 100644
--- a/clang/unittests/Format/BracesRemoverTest.cpp
+++ b/clang/unittests/Format/BracesRemoverTest.cpp
@@ -209,7 +209,9 @@ TEST_F(BracesRemoverTest, RemoveBraces) {
   verifyFormat("if (a) {\n"
                "  b;\n"
                "} else {\n"
-               "  { c; }\n"
+               "  {\n"
+               "    c;\n"
+               "  }\n"
                "}",
                Style);
 
diff --git a/clang/unittests/Format/FormatTest.cpp b/clang/unittests/Format/FormatTest.cpp
index f312a9e21158a..4906b3350b5b2 100644
--- a/clang/unittests/Format/FormatTest.cpp
+++ b/clang/unittests/Format/FormatTest.cpp
@@ -52,7 +52,13 @@ TEST_F(FormatTest, FormatsUnwrappedLinesAtFirstFormat) {
 }
 
 TEST_F(FormatTest, FormatsNestedBlockStatements) {
-  verifyFormat("{\n  {\n    {}\n  }\n}", "{{{}}}");
+  verifyFormat("{\n"
+               "  {\n"
+               "    {\n"
+               "    }\n"
+               "  }\n"
+               "}",
+               "{{{}}}");
 }
 
 TEST_F(FormatTest, FormatsNestedCall) {
@@ -5669,7 +5675,10 @@ TEST_F(FormatTest, LayoutCodeInMacroDefinitions) {
                getLLVMStyleWithColumns(14));
 }
 
-TEST_F(FormatTest, LayoutRemainingTokens) { verifyFormat("{}"); }
+TEST_F(FormatTest, LayoutRemainingTokens) {
+  verifyFormat("{\n"
+               "}");
+}
 
 TEST_F(FormatTest, MacroDefinitionInsideStatement) {
   verifyFormat("int x,\n"
@@ -6577,7 +6586,11 @@ TEST_F(FormatTest, FormatAlignInsidePreprocessorElseBlock) {
 }
 
 TEST_F(FormatTest, FormatHashIfNotAtStartOfLine) {
-  verifyFormat("{\n  { a #c; }\n}");
+  verifyFormat("{\n"
+               "  {\n"
+               "    a #c;\n"
+               "  }\n"
+               "}");
 }
 
 TEST_F(FormatTest, FormatUnbalancedStructuralElements) {
@@ -6937,13 +6950,13 @@ TEST_F(FormatTest, FormatNestedBlocksInMacros) {
 }
 
 TEST_F(FormatTest, PutEmptyBlocksIntoOneLine) {
-  verifyFormat("{}");
   verifyFormat("enum E {};");
   verifyFormat("enum E {}");
   FormatStyle Style = getLLVMStyle();
   Style.SpaceInEmptyBlock = true;
   verifyFormat("void f() { }", "void f() {}", Style);
   Style.AllowShortBlocksOnASingleLine = FormatStyle::SBS_Empty;
+  verifyFormat("{ }", Style);
   verifyFormat("while (true) { }", "while (true) {}", Style);
   Style.BreakBeforeBraces = FormatStyle::BS_Custom;
   Style.BraceWrapping.BeforeElse = false;
@@ -11527,10 +11540,18 @@ TEST_F(FormatTest, UnderstandsNewAndDelete) {
                "void new (link p);\n"
                "void delete (link p);");
 
-  verifyFormat("{ p->new(); }\n"
-               "{ p->delete(); }",
-               "{ p->new (); }\n"
-               "{ p->delete (); }");
+  verifyFormat("{\n"
+               "  p->new();\n"
+               "}\n"
+               "{\n"
+               "  p->delete();\n"
+               "}",
+               "{\n"
+               "  p->new ();\n"
+               "}\n"
+               "{\n"
+               "  p->delete ();\n"
+               "}");
 
   FormatStyle AfterPlacementOperator = getLLVMStyle();
   AfterPlacementOperator.SpaceBeforeParens = FormatStyle::SBPO_Custom;
@@ -12352,7 +12373,9 @@ TEST_F(FormatTest, FormatsCasts) {
   // FIXME: single value wrapped with paren will be treated as cast.
   verifyFormat("void f(int i = (kValue)*kMask) {}");
 
-  verifyFormat("{ (void)F; }");
+  verifyFormat("{\n"
+               "  (void)F;\n"
+               "}");
 
   // Don't break after a cast's
   verifyFormat("int aaaaaaaaaaaaaaaaaaaaaaaaaaa =\n"
@@ -13575,7 +13598,8 @@ TEST_F(FormatTest, IncorrectAccessSpecifier) {
   verifyFormat("public\n"
                "B {}");
   verifyFormat("public\n"
-               "{}");
+               "{\n"
+               "}");
   verifyFormat("public\n"
                "B { int x; }");
 }
@@ -13632,10 +13656,31 @@ TEST_F(FormatTest, DoesNotTouchUnwrappedLinesWithErrors) {
 }
 
 TEST_F(FormatTest, IncorrectCodeErrorDetection) {
-  verifyFormat("{\n  {}", "{\n{\n}");
-  verifyFormat("{\n  {}", "{\n  {\n}");
-  verifyFormat("{\n  {}", "{\n  {\n  }");
-  verifyFormat("{\n  {}\n}\n}", "{\n  {\n    }\n  }\n}");
+  verifyFormat("{\n"
+               "  {\n"
+               "  }",
+               "{\n"
+               "{\n"
+               "}");
+  verifyFormat("{\n"
+               "  {\n"
+               "  }",
+               "{\n"
+               "  {\n"
+               "}");
+  verifyFormat("{\n"
+               "  {\n"
+               "  }");
+  verifyFormat("{\n"
+               "  {\n"
+               "  }\n"
+               "}\n"
+               "}",
+               "{\n"
+               "  {\n"
+               "    }\n"
+               "  }\n"
+               "}");
 
   verifyFormat("{\n"
                "  {\n"
@@ -14080,10 +14125,14 @@ TEST_F(FormatTest, FormatsBracedListsInColumnLayout) {
                "}");
   verifyFormat("void foo() {\n"
                "  { // asdf\n"
-               "    { int a; }\n"
+               "    {\n"
+               "      int a;\n"
+               "    }\n"
                "  }\n"
                "  {\n"
-               "    { int b; }\n"
+               "    {\n"
+               "      int b;\n"
+               "    }\n"
                "  }\n"
                "}");
   verifyFormat("namespace n {\n"
@@ -14095,7 +14144,8 @@ TEST_F(FormatTest, FormatsBracedListsInColumnLayout) {
                "      }\n"
                "    }\n"
                "  }\n"
-               "  {}\n"
+               "  {\n"
+               "  }\n"
                "}\n"
                "} // namespace n");
 
diff --git a/clang/unittests/Format/FormatTestMacroExpansion.cpp b/clang/unittests/Format/FormatTestMacroExpansion.cpp
index 85ab6ea3794e8..d391fe3d715c3 100644
--- a/clang/unittests/Format/FormatTestMacroExpansion.cpp
+++ b/clang/unittests/Format/FormatTestMacroExpansion.cpp
@@ -43,9 +43,10 @@ TEST_F(FormatTestMacroExpansion, UnexpandConfiguredMacros) {
                "STMT",
                Style);
   verifyFormat("void f() { ID(a *b); }", Style);
-  verifyFormat(R"(ID(
-    { ID(a *b); });
-)",
+  verifyFormat("ID(\n"
+               "    {\n"
+               "      ID(a *b);\n"
+               "    });",
                Style);
   verifyIncompleteFormat("ID3({, ID(a *b),\n"
                          "    ;\n"
diff --git a/clang/unittests/Format/TokenAnnotatorTest.cpp b/clang/unittests/Format/TokenAnnotatorTest.cpp
index c3153cf6b16f0..da02ced8c7a94 100644
--- a/clang/unittests/Format/TokenAnnotatorTest.cpp
+++ b/clang/unittests/Format/TokenAnnotatorTest.cpp
@@ -2856,6 +2856,30 @@ TEST_F(TokenAnnotatorTest, UnderstandsElaboratedTypeSpecifier) {
   EXPECT_TOKEN(Tokens[7], tok::l_brace, TT_FunctionLBrace);
 }
 
+TEST_F(TokenAnnotatorTest, BlockLBrace) {
+  auto Tokens = annotate("{\n"
+                         "  {\n"
+                         "    foo();\n"
+                         "  }\n"
+                         "}");
+  ASSERT_EQ(Tokens.size(), 9u) << Tokens;
+  EXPECT_TOKEN(Tokens[0], tok::l_brace, TT_BlockLBrace);
+  EXPECT_BRACE_KIND(Tokens[0], BK_Block);
+  EXPECT_TOKEN(Tokens[1], tok::l_brace, TT_BlockLBrace);
+  EXPECT_BRACE_KIND(Tokens[1], BK_Block);
+
+  Tokens = annotate("void bar() {\n"
+                    "  {\n"
+                    "    foo();\n"
+                    "  }\n"
+                    "}");
+  ASSERT_EQ(Tokens.size(), 13u) << Tokens;
+  EXPECT_TOKEN(Tokens[4], tok::l_brace, TT_FunctionLBrace);
+  EXPECT_BRACE_KIND(Tokens[4], BK_Block);
+  EXPECT_TOKEN(Tokens[5], tok::l_brace, TT_BlockLBrace);
+  EXPECT_BRACE_KIND(Tokens[5], BK_Block);
+}
+
 } // namespace
 } // namespace format
 } // namespace clang

From 6b46166ef2612d2a58767447b3db8f0343afb552 Mon Sep 17 00:00:00 2001
From: Jordan Rupprecht <rupprecht@google.com>
Date: Thu, 11 Apr 2024 02:14:07 +0000
Subject: [PATCH 087/886] [llvm][NFC] Suppress `-Wunused-result` call to
 `write`

Commit 87e6f87fe7e343eb656e9b49d30cbb065c086651 adds a call to `::write()`, which may be annotated w/ `warn_unused_result`, leading to `-Wunused-result` failures.
---
 llvm/lib/Support/raw_socket_stream.cpp | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Support/raw_socket_stream.cpp b/llvm/lib/Support/raw_socket_stream.cpp
index 1dcf6352f2cc2..14e2308df4d7e 100644
--- a/llvm/lib/Support/raw_socket_stream.cpp
+++ b/llvm/lib/Support/raw_socket_stream.cpp
@@ -265,7 +265,10 @@ void ListeningSocket::shutdown() {
 
   // Ensure ::poll returns if shutdown is called by a seperate thread
   char Byte = 'A';
-  ::write(PipeFD[1], &Byte, 1);
+  ssize_t written = ::write(PipeFD[1], &Byte, 1);
+
+  // Ignore any write() error
+  (void)written;
 }
 
 ListeningSocket::~ListeningSocket() {

From f4509cf284ced95f31dc7eb63144b4bc47899c43 Mon Sep 17 00:00:00 2001
From: Freddy Ye <freddy.ye@intel.com>
Date: Thu, 11 Apr 2024 10:18:29 +0800
Subject: [PATCH 088/886] [X86][MC] Support enc/dec for SETZUCC and promoted
 SETCC. (#86473)

apx-spec: https://cdrdv2.intel.com/v1/dl/getContent/784266
apx-syntax-recommendation:
https://cdrdv2.intel.com/v1/dl/getContent/817241
---
 .../lib/Target/X86/AsmParser/X86AsmParser.cpp |   1 +
 .../X86/MCTargetDesc/X86MCCodeEmitter.cpp     |   2 +
 llvm/lib/Target/X86/X86InstrAsmAlias.td       |   8 +-
 llvm/lib/Target/X86/X86InstrCMovSetCC.td      |  19 +++
 llvm/lib/Target/X86/X86InstrControl.td        |  24 ++--
 .../MC/Disassembler/X86/apx/evex-format.txt   |  10 ++
 llvm/test/MC/Disassembler/X86/apx/setcc.txt   | 130 ++++++++++++++++++
 llvm/test/MC/Disassembler/X86/apx/setzucc.txt | 130 ++++++++++++++++++
 llvm/test/MC/X86/apx/evex-format-att.s        |  10 ++
 llvm/test/MC/X86/apx/evex-format-intel.s      |  10 ++
 llvm/test/MC/X86/apx/setcc-att.s              | 101 ++++++++++++++
 llvm/test/MC/X86/apx/setcc-intel.s            |  98 +++++++++++++
 llvm/test/MC/X86/apx/setzucc-att.s            | 101 ++++++++++++++
 llvm/test/MC/X86/apx/setzucc-intel.s          |  98 +++++++++++++
 llvm/test/TableGen/x86-fold-tables.inc        |   1 +
 llvm/utils/TableGen/AsmMatcherEmitter.cpp     |  13 +-
 16 files changed, 736 insertions(+), 20 deletions(-)
 create mode 100644 llvm/test/MC/Disassembler/X86/apx/setcc.txt
 create mode 100644 llvm/test/MC/Disassembler/X86/apx/setzucc.txt
 create mode 100644 llvm/test/MC/X86/apx/setcc-att.s
 create mode 100644 llvm/test/MC/X86/apx/setcc-intel.s
 create mode 100644 llvm/test/MC/X86/apx/setzucc-att.s
 create mode 100644 llvm/test/MC/X86/apx/setzucc-intel.s

diff --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index 6401df9f49f03..b05a036fb2f06 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -3287,6 +3287,7 @@ bool X86AsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
 
   // FIXME: Hack to recognize setneb as setne.
   if (PatchedName.starts_with("set") && PatchedName.ends_with("b") &&
+      PatchedName != "setzub" && PatchedName != "setzunb" &&
       PatchedName != "setb" && PatchedName != "setnb")
     PatchedName = PatchedName.substr(0, Name.size()-1);
 
diff --git a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
index 92a14226a0dc0..a5859f98bae02 100644
--- a/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
+++ b/llvm/lib/Target/X86/MCTargetDesc/X86MCCodeEmitter.cpp
@@ -1155,6 +1155,7 @@ X86MCCodeEmitter::emitVEXOpcodePrefix(int MemOperand, const MCInst &MI,
     Prefix.setXX2(MI, MemOperand + X86::AddrIndexReg);
     break;
   }
+  case X86II::MRMXmCC:
   case X86II::MRM0m:
   case X86II::MRM1m:
   case X86II::MRM2m:
@@ -1282,6 +1283,7 @@ X86MCCodeEmitter::emitVEXOpcodePrefix(int MemOperand, const MCInst &MI,
     Prefix.setRR2(MI, CurOp++);
     break;
   }
+  case X86II::MRMXrCC:
   case X86II::MRM0r:
   case X86II::MRM1r:
   case X86II::MRM2r:
diff --git a/llvm/lib/Target/X86/X86InstrAsmAlias.td b/llvm/lib/Target/X86/X86InstrAsmAlias.td
index 6b15213a2e683..d06a0c79b46bb 100644
--- a/llvm/lib/Target/X86/X86InstrAsmAlias.td
+++ b/llvm/lib/Target/X86/X86InstrAsmAlias.td
@@ -760,7 +760,7 @@ def : InstAlias<"xor{q}\t{$imm, %rax|rax, $imm}", (XOR64ri8 RAX, i64i8imm:$imm),
 def : InstAlias<"movq.s\t{$src, $dst|$dst, $src}",
                 (MMX_MOVQ64rr_REV VR64:$dst, VR64:$src), 0>;
 
-//  CMOV SETCC Aliases
+//  CMOV SETCC SETZUCC Aliases
 multiclass CMOV_SETCC_Aliases<string Cond, int CC> {
   def : InstAlias<"cmov"#Cond#"{w}\t{$src, $dst|$dst, $src}",
                   (CMOV16rr GR16:$dst, GR16:$src, CC), 0>;
@@ -787,8 +787,12 @@ let Predicates = [In64BitMode] in {
                   (CMOV64rr_ND GR64:$dst, GR64:$src1, GR64:$src2, CC), 0>;
   def : InstAlias<"cmov"#Cond#"{q}\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                   (CMOV64rm_ND GR64:$dst, GR64:$src1, i64mem:$src2, CC), 0>;
-}
 
+  def : InstAlias<"setzu"#Cond#"\t$dst", (SETZUCCr GR8:$dst, CC), 0>;
+  def : InstAlias<"setzu"#Cond#"\t$dst", (SETZUCCm i8mem:$dst, CC), 0>;
+  def : InstAlias<"set"#Cond#"\t$dst", (SETCCr_EVEX GR8:$dst, CC), 0>;
+  def : InstAlias<"set"#Cond#"\t$dst", (SETCCm_EVEX i8mem:$dst, CC), 0>;
+}
   def : InstAlias<"set"#Cond#"\t$dst", (SETCCr GR8:$dst, CC), 0>;
   def : InstAlias<"set"#Cond#"\t$dst", (SETCCm i8mem:$dst, CC), 0>;
 }
diff --git a/llvm/lib/Target/X86/X86InstrCMovSetCC.td b/llvm/lib/Target/X86/X86InstrCMovSetCC.td
index d41591f68a605..27a0c889a4da3 100644
--- a/llvm/lib/Target/X86/X86InstrCMovSetCC.td
+++ b/llvm/lib/Target/X86/X86InstrCMovSetCC.td
@@ -127,6 +127,25 @@ let Uses = [EFLAGS], isCodeGenOnly = 1, ForceDisassemble = 1 in {
                 TB, Sched<[WriteSETCCStore]>;
 } // Uses = [EFLAGS]
 
+// SetZUCC and promoted SetCC instructions.
+let Uses = [EFLAGS], isCodeGenOnly = 1, ForceDisassemble = 1,
+  hasSideEffects = 0, Predicates = [In64BitMode], Predicates = [HasNDD] in {
+  def SETZUCCr : I<0x40, MRMXrCC, (outs GR8:$dst), (ins ccode:$cond),
+                "setzu${cond}\t$dst", []>,
+                XD, ZU, NoCD8, Sched<[WriteSETCC]>;
+  def SETCCr_EVEX : I<0x40, MRMXrCC, (outs GR8:$dst), (ins ccode:$cond),
+                "set${cond}\t$dst", []>,
+                XD, PL, Sched<[WriteSETCC]>;
+  let mayStore = 1 in {
+    def SETZUCCm : I<0x40, MRMXmCC, (outs), (ins i8mem:$dst, ccode:$cond),
+                  "setzu${cond}\t$dst", []>,
+                  XD, ZU, NoCD8, Sched<[WriteSETCCStore]>;
+    def SETCCm_EVEX : I<0x40, MRMXmCC, (outs), (ins i8mem:$dst, ccode:$cond),
+                  "set${cond}\t$dst", []>,
+                  XD, PL, Sched<[WriteSETCCStore]>;
+  }
+}
+
 // SALC is an undocumented instruction. Information for this instruction can be found
 // here http://www.rcollins.org/secrets/opcodes/SALC.html
 // Set AL if carry. 
diff --git a/llvm/lib/Target/X86/X86InstrControl.td b/llvm/lib/Target/X86/X86InstrControl.td
index 5171c2249dee9..62cc758cc594b 100644
--- a/llvm/lib/Target/X86/X86InstrControl.td
+++ b/llvm/lib/Target/X86/X86InstrControl.td
@@ -167,24 +167,24 @@ let isBranch = 1, isTerminator = 1, isBarrier = 1, isIndirectBranch = 1 in {
   }
 
   let Predicates = [Not64BitMode], AsmVariantName = "att" in {
-    def FARJMP16i  : Iseg16<0xEA, RawFrmImm16, (outs),
-                            (ins i16imm:$off, i16imm:$seg),
-                            "ljmp{w}\t$seg, $off", []>,
-                            OpSize16, Sched<[WriteJump]>;
     def FARJMP32i  : Iseg32<0xEA, RawFrmImm16, (outs),
                             (ins i32imm:$off, i16imm:$seg),
                             "ljmp{l}\t$seg, $off", []>,
                             OpSize32, Sched<[WriteJump]>;
+    def FARJMP16i  : Iseg16<0xEA, RawFrmImm16, (outs),
+                            (ins i16imm:$off, i16imm:$seg),
+                            "ljmp{w}\t$seg, $off", []>,
+                            OpSize16, Sched<[WriteJump]>;
   }
   let mayLoad = 1 in {
     def FARJMP64m  : RI<0xFF, MRM5m, (outs), (ins opaquemem:$dst),
                         "ljmp{q}\t{*}$dst", []>, Sched<[WriteJump]>, Requires<[In64BitMode]>;
 
+    def FARJMP32m  : I<0xFF, MRM5m, (outs), (ins opaquemem:$dst),
+                       "{l}jmp{l}\t{*}$dst", []>, OpSize32, Sched<[WriteJumpLd]>;
     let AsmVariantName = "att" in
     def FARJMP16m  : I<0xFF, MRM5m, (outs), (ins opaquemem:$dst),
                        "ljmp{w}\t{*}$dst", []>, OpSize16, Sched<[WriteJumpLd]>;
-    def FARJMP32m  : I<0xFF, MRM5m, (outs), (ins opaquemem:$dst),
-                       "{l}jmp{l}\t{*}$dst", []>, OpSize32, Sched<[WriteJumpLd]>;
   }
 }
 
@@ -253,21 +253,21 @@ let isCall = 1 in
     }
 
     let Predicates = [Not64BitMode], AsmVariantName = "att" in {
-      def FARCALL16i  : Iseg16<0x9A, RawFrmImm16, (outs),
-                               (ins i16imm:$off, i16imm:$seg),
-                               "lcall{w}\t$seg, $off", []>,
-                               OpSize16, Sched<[WriteJump]>;
       def FARCALL32i  : Iseg32<0x9A, RawFrmImm16, (outs),
                                (ins i32imm:$off, i16imm:$seg),
                                "lcall{l}\t$seg, $off", []>,
                                OpSize32, Sched<[WriteJump]>;
+      def FARCALL16i  : Iseg16<0x9A, RawFrmImm16, (outs),
+                               (ins i16imm:$off, i16imm:$seg),
+                               "lcall{w}\t$seg, $off", []>,
+                               OpSize16, Sched<[WriteJump]>;
     }
 
     let mayLoad = 1 in {
-      def FARCALL16m  : I<0xFF, MRM3m, (outs), (ins opaquemem:$dst),
-                          "lcall{w}\t{*}$dst", []>, OpSize16, Sched<[WriteJumpLd]>;
       def FARCALL32m  : I<0xFF, MRM3m, (outs), (ins opaquemem:$dst),
                           "{l}call{l}\t{*}$dst", []>, OpSize32, Sched<[WriteJumpLd]>;
+      def FARCALL16m  : I<0xFF, MRM3m, (outs), (ins opaquemem:$dst),
+                          "lcall{w}\t{*}$dst", []>, OpSize16, Sched<[WriteJumpLd]>;
     }
   }
 
diff --git a/llvm/test/MC/Disassembler/X86/apx/evex-format.txt b/llvm/test/MC/Disassembler/X86/apx/evex-format.txt
index 1156f5c409922..e9a9f1327a17e 100644
--- a/llvm/test/MC/Disassembler/X86/apx/evex-format.txt
+++ b/llvm/test/MC/Disassembler/X86/apx/evex-format.txt
@@ -215,6 +215,16 @@
 # INTEL: sar	r17, r16, 123
 0x62,0xfc,0xf4,0x10,0xc1,0xf8,0x7b
 
+## MRMXrCC
+# ATT:   setzuo	%r16b
+# INTEL: setzuo	r16b
+0x62,0xfc,0x7f,0x18,0x40,0xc0
+
+## MRMXmCC
+# ATT:   setzuo	(%r16,%r17)
+# INTEL: setzuo	byte ptr [r16 + r17]
+0x62,0xfc,0x7b,0x18,0x40,0x04,0x08
+
 ## NoCD8
 
 # ATT:   {nf}	negq	123(%r16)
diff --git a/llvm/test/MC/Disassembler/X86/apx/setcc.txt b/llvm/test/MC/Disassembler/X86/apx/setcc.txt
new file mode 100644
index 0000000000000..1c00acfb76672
--- /dev/null
+++ b/llvm/test/MC/Disassembler/X86/apx/setcc.txt
@@ -0,0 +1,130 @@
+# RUN: llvm-mc -triple x86_64 -disassemble %s | FileCheck %s --check-prefix=ATT
+# RUN: llvm-mc -triple x86_64 -disassemble -output-asm-variant=1 %s | FileCheck %s --check-prefix=INTEL
+
+# ATT:   {evex}	seto	%al
+# INTEL: {evex}	seto	al
+0x62,0xf4,0x7f,0x08,0x40,0xc0
+
+# ATT:   {evex}	setno	%al
+# INTEL: {evex}	setno	al
+0x62,0xf4,0x7f,0x08,0x41,0xc0
+
+# ATT:   {evex}	setb	%al
+# INTEL: {evex}	setb	al
+0x62,0xf4,0x7f,0x08,0x42,0xc0
+
+# ATT:   {evex}	setae	%al
+# INTEL: {evex}	setae	al
+0x62,0xf4,0x7f,0x08,0x43,0xc0
+
+# ATT:   {evex}	sete	%al
+# INTEL: {evex}	sete	al
+0x62,0xf4,0x7f,0x08,0x44,0xc0
+
+# ATT:   {evex}	setne	%al
+# INTEL: {evex}	setne	al
+0x62,0xf4,0x7f,0x08,0x45,0xc0
+
+# ATT:   {evex}	setbe	%al
+# INTEL: {evex}	setbe	al
+0x62,0xf4,0x7f,0x08,0x46,0xc0
+
+# ATT:   {evex}	seta	%al
+# INTEL: {evex}	seta	al
+0x62,0xf4,0x7f,0x08,0x47,0xc0
+
+# ATT:   {evex}	sets	%al
+# INTEL: {evex}	sets	al
+0x62,0xf4,0x7f,0x08,0x48,0xc0
+
+# ATT:   {evex}	setns	%al
+# INTEL: {evex}	setns	al
+0x62,0xf4,0x7f,0x08,0x49,0xc0
+
+# ATT:   {evex}	setp	%al
+# INTEL: {evex}	setp	al
+0x62,0xf4,0x7f,0x08,0x4a,0xc0
+
+# ATT:   {evex}	setnp	%al
+# INTEL: {evex}	setnp	al
+0x62,0xf4,0x7f,0x08,0x4b,0xc0
+
+# ATT:   {evex}	setl	%al
+# INTEL: {evex}	setl	al
+0x62,0xf4,0x7f,0x08,0x4c,0xc0
+
+# ATT:   {evex}	setge	%al
+# INTEL: {evex}	setge	al
+0x62,0xf4,0x7f,0x08,0x4d,0xc0
+
+# ATT:   {evex}	setle	%al
+# INTEL: {evex}	setle	al
+0x62,0xf4,0x7f,0x08,0x4e,0xc0
+
+# ATT:   {evex}	setg	%al
+# INTEL: {evex}	setg	al
+0x62,0xf4,0x7f,0x08,0x4f,0xc0
+
+# ATT:   {evex}	seto	(%rax)
+# INTEL: {evex}	seto	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x40,0x00
+
+# ATT:   {evex}	setno	(%rax)
+# INTEL: {evex}	setno	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x41,0x00
+
+# ATT:   {evex}	setb	(%rax)
+# INTEL: {evex}	setb	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x42,0x00
+
+# ATT:   {evex}	setae	(%rax)
+# INTEL: {evex}	setae	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x43,0x00
+
+# ATT:   {evex}	sete	(%rax)
+# INTEL: {evex}	sete	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x44,0x00
+
+# ATT:   {evex}	setne	(%rax)
+# INTEL: {evex}	setne	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x45,0x00
+
+# ATT:   {evex}	setbe	(%rax)
+# INTEL: {evex}	setbe	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x46,0x00
+
+# ATT:   {evex}	seta	(%rax)
+# INTEL: {evex}	seta	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x47,0x00
+
+# ATT:   {evex}	sets	(%rax)
+# INTEL: {evex}	sets	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x48,0x00
+
+# ATT:   {evex}	setns	(%rax)
+# INTEL: {evex}	setns	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x49,0x00
+
+# ATT:   {evex}	setp	(%rax)
+# INTEL: {evex}	setp	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4a,0x00
+
+# ATT:   {evex}	setnp	(%rax)
+# INTEL: {evex}	setnp	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4b,0x00
+
+# ATT:   {evex}	setl	(%rax)
+# INTEL: {evex}	setl	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4c,0x00
+
+# ATT:   {evex}	setge	(%rax)
+# INTEL: {evex}	setge	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4d,0x00
+
+# ATT:   {evex}	setle	(%rax)
+# INTEL: {evex}	setle	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4e,0x00
+
+# ATT:   {evex}	setg	(%rax)
+# INTEL: {evex}	setg	byte ptr [rax]
+0x62,0xf4,0x7f,0x08,0x4f,0x00
diff --git a/llvm/test/MC/Disassembler/X86/apx/setzucc.txt b/llvm/test/MC/Disassembler/X86/apx/setzucc.txt
new file mode 100644
index 0000000000000..44aaa4b33cc85
--- /dev/null
+++ b/llvm/test/MC/Disassembler/X86/apx/setzucc.txt
@@ -0,0 +1,130 @@
+# RUN: llvm-mc -triple x86_64 -disassemble %s | FileCheck %s --check-prefix=ATT
+# RUN: llvm-mc -triple x86_64 -disassemble -output-asm-variant=1 %s | FileCheck %s --check-prefix=INTEL
+
+# ATT:   setzuo	%al
+# INTEL: setzuo	al
+0x62,0xf4,0x7f,0x18,0x40,0xc0
+
+# ATT:   setzuno	%al
+# INTEL: setzuno	al
+0x62,0xf4,0x7f,0x18,0x41,0xc0
+
+# ATT:   setzub	%al
+# INTEL: setzub	al
+0x62,0xf4,0x7f,0x18,0x42,0xc0
+
+# ATT:   setzuae	%al
+# INTEL: setzuae	al
+0x62,0xf4,0x7f,0x18,0x43,0xc0
+
+# ATT:   setzue	%al
+# INTEL: setzue	al
+0x62,0xf4,0x7f,0x18,0x44,0xc0
+
+# ATT:   setzune	%al
+# INTEL: setzune	al
+0x62,0xf4,0x7f,0x18,0x45,0xc0
+
+# ATT:   setzube	%al
+# INTEL: setzube	al
+0x62,0xf4,0x7f,0x18,0x46,0xc0
+
+# ATT:   setzua	%al
+# INTEL: setzua	al
+0x62,0xf4,0x7f,0x18,0x47,0xc0
+
+# ATT:   setzus	%al
+# INTEL: setzus	al
+0x62,0xf4,0x7f,0x18,0x48,0xc0
+
+# ATT:   setzuns	%al
+# INTEL: setzuns	al
+0x62,0xf4,0x7f,0x18,0x49,0xc0
+
+# ATT:   setzup	%al
+# INTEL: setzup	al
+0x62,0xf4,0x7f,0x18,0x4a,0xc0
+
+# ATT:   setzunp	%al
+# INTEL: setzunp	al
+0x62,0xf4,0x7f,0x18,0x4b,0xc0
+
+# ATT:   setzul	%al
+# INTEL: setzul	al
+0x62,0xf4,0x7f,0x18,0x4c,0xc0
+
+# ATT:   setzuge	%al
+# INTEL: setzuge	al
+0x62,0xf4,0x7f,0x18,0x4d,0xc0
+
+# ATT:   setzule	%al
+# INTEL: setzule	al
+0x62,0xf4,0x7f,0x18,0x4e,0xc0
+
+# ATT:   setzug	%al
+# INTEL: setzug	al
+0x62,0xf4,0x7f,0x18,0x4f,0xc0
+
+# ATT:   setzuo	(%rax)
+# INTEL: setzuo	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x40,0x00
+
+# ATT:   setzuno	(%rax)
+# INTEL: setzuno	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x41,0x00
+
+# ATT:   setzub	(%rax)
+# INTEL: setzub	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x42,0x00
+
+# ATT:   setzuae	(%rax)
+# INTEL: setzuae	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x43,0x00
+
+# ATT:   setzue	(%rax)
+# INTEL: setzue	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x44,0x00
+
+# ATT:   setzune	(%rax)
+# INTEL: setzune	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x45,0x00
+
+# ATT:   setzube	(%rax)
+# INTEL: setzube	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x46,0x00
+
+# ATT:   setzua	(%rax)
+# INTEL: setzua	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x47,0x00
+
+# ATT:   setzus	(%rax)
+# INTEL: setzus	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x48,0x00
+
+# ATT:   setzuns	(%rax)
+# INTEL: setzuns	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x49,0x00
+
+# ATT:   setzup	(%rax)
+# INTEL: setzup	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4a,0x00
+
+# ATT:   setzunp	(%rax)
+# INTEL: setzunp	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4b,0x00
+
+# ATT:   setzul	(%rax)
+# INTEL: setzul	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4c,0x00
+
+# ATT:   setzuge	(%rax)
+# INTEL: setzuge	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4d,0x00
+
+# ATT:   setzule	(%rax)
+# INTEL: setzule	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4e,0x00
+
+# ATT:   setzug	(%rax)
+# INTEL: setzug	byte ptr [rax]
+0x62,0xf4,0x7f,0x18,0x4f,0x00
diff --git a/llvm/test/MC/X86/apx/evex-format-att.s b/llvm/test/MC/X86/apx/evex-format-att.s
index 36df3f3757dc3..e59039ea2d822 100644
--- a/llvm/test/MC/X86/apx/evex-format-att.s
+++ b/llvm/test/MC/X86/apx/evex-format-att.s
@@ -210,6 +210,16 @@
 # CHECK: encoding: [0x62,0xfc,0xf4,0x10,0xc1,0xf8,0x7b]
          sarq	$123, %r16, %r17
 
+## MRMXrCC
+# CHECK: setzuo	%r16b
+# CHECK: encoding: [0x62,0xfc,0x7f,0x18,0x40,0xc0]
+         setzuo	%r16b
+
+## MRMXmCC
+# CHECK: setzuo	(%r16,%r17)
+# CHECK: encoding: [0x62,0xfc,0x7b,0x18,0x40,0x04,0x08]
+         setzuo	(%r16,%r17)
+
 ## NoCD8
 
 # CHECK: {nf}	negq	123(%r16)
diff --git a/llvm/test/MC/X86/apx/evex-format-intel.s b/llvm/test/MC/X86/apx/evex-format-intel.s
index 2b346e0e85806..42d4c0c0081a7 100644
--- a/llvm/test/MC/X86/apx/evex-format-intel.s
+++ b/llvm/test/MC/X86/apx/evex-format-intel.s
@@ -210,6 +210,16 @@
 # CHECK: encoding: [0x62,0xfc,0xf4,0x10,0xc1,0xf8,0x7b]
          sar	r17, r16, 123
 
+## MRMXrCC
+# CHECK: setzuo	r16b
+# CHECK: encoding: [0x62,0xfc,0x7f,0x18,0x40,0xc0]
+         setzuo r16b
+
+## MRMXmCC
+# CHECK: setzuo byte ptr [r16 + r17]
+# CHECK: encoding: [0x62,0xfc,0x7b,0x18,0x40,0x04,0x08]
+         setzuo byte ptr [r16 + r17]
+
 ## NoCD8
 
 # CHECK: {nf}	neg	qword ptr [r16 + 123]
diff --git a/llvm/test/MC/X86/apx/setcc-att.s b/llvm/test/MC/X86/apx/setcc-att.s
new file mode 100644
index 0000000000000..b5518081a8201
--- /dev/null
+++ b/llvm/test/MC/X86/apx/setcc-att.s
@@ -0,0 +1,101 @@
+# RUN: llvm-mc -triple x86_64 -show-encoding %s | FileCheck %s
+# RUN: not llvm-mc -triple i386 -show-encoding %s 2>&1 | FileCheck %s --check-prefix=ERROR
+
+# ERROR-COUNT-32: error:
+# ERROR-NOT: error:
+# CHECK: {evex}	seto	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x40,0xc0]
+         {evex}	seto	%al
+# CHECK: {evex}	setno	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x41,0xc0]
+         {evex}	setno	%al
+# CHECK: {evex}	setb	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x42,0xc0]
+         {evex}	setb	%al
+# CHECK: {evex}	setae	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x43,0xc0]
+         {evex}	setae	%al
+# CHECK: {evex}	sete	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x44,0xc0]
+         {evex}	sete	%al
+# CHECK: {evex}	setne	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x45,0xc0]
+         {evex}	setne	%al
+# CHECK: {evex}	setbe	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x46,0xc0]
+         {evex}	setbe	%al
+# CHECK: {evex}	seta	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x47,0xc0]
+         {evex}	seta	%al
+# CHECK: {evex}	sets	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x48,0xc0]
+         {evex}	sets	%al
+# CHECK: {evex}	setns	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x49,0xc0]
+         {evex}	setns	%al
+# CHECK: {evex}	setp	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4a,0xc0]
+         {evex}	setp	%al
+# CHECK: {evex}	setnp	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4b,0xc0]
+         {evex}	setnp	%al
+# CHECK: {evex}	setl	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4c,0xc0]
+         {evex}	setl	%al
+# CHECK: {evex}	setge	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4d,0xc0]
+         {evex}	setge	%al
+# CHECK: {evex}	setle	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4e,0xc0]
+         {evex}	setle	%al
+# CHECK: {evex}	setg	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4f,0xc0]
+         {evex}	setg	%al
+# CHECK: {evex}	seto	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x40,0x00]
+         {evex}	seto	(%rax)
+# CHECK: {evex}	setno	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x41,0x00]
+         {evex}	setno	(%rax)
+# CHECK: {evex}	setb	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x42,0x00]
+         {evex}	setb	(%rax)
+# CHECK: {evex}	setae	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x43,0x00]
+         {evex}	setae	(%rax)
+# CHECK: {evex}	sete	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x44,0x00]
+         {evex}	sete	(%rax)
+# CHECK: {evex}	setne	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x45,0x00]
+         {evex}	setne	(%rax)
+# CHECK: {evex}	setbe	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x46,0x00]
+         {evex}	setbe	(%rax)
+# CHECK: {evex}	seta	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x47,0x00]
+         {evex}	seta	(%rax)
+# CHECK: {evex}	sets	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x48,0x00]
+         {evex}	sets	(%rax)
+# CHECK: {evex}	setns	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x49,0x00]
+         {evex}	setns	(%rax)
+# CHECK: {evex}	setp	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4a,0x00]
+         {evex}	setp	(%rax)
+# CHECK: {evex}	setnp	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4b,0x00]
+         {evex}	setnp	(%rax)
+# CHECK: {evex}	setl	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4c,0x00]
+         {evex}	setl	(%rax)
+# CHECK: {evex}	setge	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4d,0x00]
+         {evex}	setge	(%rax)
+# CHECK: {evex}	setle	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4e,0x00]
+         {evex}	setle	(%rax)
+# CHECK: {evex}	setg	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4f,0x00]
+         {evex}	setgb	(%rax)
diff --git a/llvm/test/MC/X86/apx/setcc-intel.s b/llvm/test/MC/X86/apx/setcc-intel.s
new file mode 100644
index 0000000000000..e005c2edb95c4
--- /dev/null
+++ b/llvm/test/MC/X86/apx/setcc-intel.s
@@ -0,0 +1,98 @@
+# RUN: llvm-mc -triple x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 %s | FileCheck %s
+
+# CHECK: {evex}	seto	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x40,0xc0]
+         {evex}	seto	al
+# CHECK: {evex}	setno	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x41,0xc0]
+         {evex}	setno	al
+# CHECK: {evex}	setb	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x42,0xc0]
+         {evex}	setb	al
+# CHECK: {evex}	setae	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x43,0xc0]
+         {evex}	setae	al
+# CHECK: {evex}	sete	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x44,0xc0]
+         {evex}	sete	al
+# CHECK: {evex}	setne	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x45,0xc0]
+         {evex}	setne	al
+# CHECK: {evex}	setbe	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x46,0xc0]
+         {evex}	setbe	al
+# CHECK: {evex}	seta	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x47,0xc0]
+         {evex}	seta	al
+# CHECK: {evex}	sets	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x48,0xc0]
+         {evex}	sets	al
+# CHECK: {evex}	setns	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x49,0xc0]
+         {evex}	setns	al
+# CHECK: {evex}	setp	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4a,0xc0]
+         {evex}	setp	al
+# CHECK: {evex}	setnp	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4b,0xc0]
+         {evex}	setnp	al
+# CHECK: {evex}	setl	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4c,0xc0]
+         {evex}	setl	al
+# CHECK: {evex}	setge	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4d,0xc0]
+         {evex}	setge	al
+# CHECK: {evex}	setle	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4e,0xc0]
+         {evex}	setle	al
+# CHECK: {evex}	setg	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4f,0xc0]
+         {evex}	setg	al
+# CHECK: {evex}	seto	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x40,0x00]
+         {evex}	seto	byte ptr [rax]
+# CHECK: {evex}	setno	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x41,0x00]
+         {evex}	setno	byte ptr [rax]
+# CHECK: {evex}	setb	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x42,0x00]
+         {evex}	setb	byte ptr [rax]
+# CHECK: {evex}	setae	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x43,0x00]
+         {evex}	setae	byte ptr [rax]
+# CHECK: {evex}	sete	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x44,0x00]
+         {evex}	sete	byte ptr [rax]
+# CHECK: {evex}	setne	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x45,0x00]
+         {evex}	setne	byte ptr [rax]
+# CHECK: {evex}	setbe	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x46,0x00]
+         {evex}	setbe	byte ptr [rax]
+# CHECK: {evex}	seta	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x47,0x00]
+         {evex}	seta	byte ptr [rax]
+# CHECK: {evex}	sets	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x48,0x00]
+         {evex}	sets	byte ptr [rax]
+# CHECK: {evex}	setns	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x49,0x00]
+         {evex}	setns	byte ptr [rax]
+# CHECK: {evex}	setp	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4a,0x00]
+         {evex}	setp	byte ptr [rax]
+# CHECK: {evex}	setnp	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4b,0x00]
+         {evex}	setnp	byte ptr [rax]
+# CHECK: {evex}	setl	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4c,0x00]
+         {evex}	setl	byte ptr [rax]
+# CHECK: {evex}	setge	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4d,0x00]
+         {evex}	setge	byte ptr [rax]
+# CHECK: {evex}	setle	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4e,0x00]
+         {evex}	setle	byte ptr [rax]
+# CHECK: {evex}	setg	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x08,0x4f,0x00]
+         {evex}	setg	byte ptr [rax]
diff --git a/llvm/test/MC/X86/apx/setzucc-att.s b/llvm/test/MC/X86/apx/setzucc-att.s
new file mode 100644
index 0000000000000..b4b7a633fa319
--- /dev/null
+++ b/llvm/test/MC/X86/apx/setzucc-att.s
@@ -0,0 +1,101 @@
+# RUN: llvm-mc -triple x86_64 -show-encoding %s | FileCheck %s
+# RUN: not llvm-mc -triple i386 -show-encoding %s 2>&1 | FileCheck %s --check-prefix=ERROR
+
+# ERROR-COUNT-32: error:
+# ERROR-NOT: error:
+# CHECK: setzuo	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x40,0xc0]
+         setzuo	%al
+# CHECK: setzuno	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x41,0xc0]
+         setzuno	%al
+# CHECK: setzub	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x42,0xc0]
+         setzub	%al
+# CHECK: setzuae	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x43,0xc0]
+         setzuae	%al
+# CHECK: setzue	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x44,0xc0]
+         setzue	%al
+# CHECK: setzune	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x45,0xc0]
+         setzune	%al
+# CHECK: setzube	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x46,0xc0]
+         setzube	%al
+# CHECK: setzua	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x47,0xc0]
+         setzua	%al
+# CHECK: setzus	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x48,0xc0]
+         setzus	%al
+# CHECK: setzuns	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x49,0xc0]
+         setzuns	%al
+# CHECK: setzup	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4a,0xc0]
+         setzup	%al
+# CHECK: setzunp	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4b,0xc0]
+         setzunp	%al
+# CHECK: setzul	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4c,0xc0]
+         setzul	%al
+# CHECK: setzuge	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4d,0xc0]
+         setzuge	%al
+# CHECK: setzule	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4e,0xc0]
+         setzule	%al
+# CHECK: setzug	%al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4f,0xc0]
+         setzug	%al
+# CHECK: setzuo	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x40,0x00]
+         setzuo	(%rax)
+# CHECK: setzuno	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x41,0x00]
+         setzuno	(%rax)
+# CHECK: setzub	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x42,0x00]
+         setzub	(%rax)
+# CHECK: setzuae	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x43,0x00]
+         setzuae	(%rax)
+# CHECK: setzue	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x44,0x00]
+         setzue	(%rax)
+# CHECK: setzune	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x45,0x00]
+         setzune	(%rax)
+# CHECK: setzube	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x46,0x00]
+         setzube	(%rax)
+# CHECK: setzua	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x47,0x00]
+         setzua	(%rax)
+# CHECK: setzus	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x48,0x00]
+         setzus	(%rax)
+# CHECK: setzuns	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x49,0x00]
+         setzuns	(%rax)
+# CHECK: setzup	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4a,0x00]
+         setzup	(%rax)
+# CHECK: setzunp	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4b,0x00]
+         setzunp	(%rax)
+# CHECK: setzul	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4c,0x00]
+         setzul	(%rax)
+# CHECK: setzuge	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4d,0x00]
+         setzuge	(%rax)
+# CHECK: setzule	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4e,0x00]
+         setzule	(%rax)
+# CHECK: setzug	(%rax)
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4f,0x00]
+         setzug	(%rax)
diff --git a/llvm/test/MC/X86/apx/setzucc-intel.s b/llvm/test/MC/X86/apx/setzucc-intel.s
new file mode 100644
index 0000000000000..bdefba6ac8d30
--- /dev/null
+++ b/llvm/test/MC/X86/apx/setzucc-intel.s
@@ -0,0 +1,98 @@
+# RUN: llvm-mc -triple x86_64 -show-encoding -x86-asm-syntax=intel -output-asm-variant=1 %s | FileCheck %s
+
+# CHECK: setzuo	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x40,0xc0]
+         setzuo	al
+# CHECK: setzuno	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x41,0xc0]
+         setzuno	al
+# CHECK: setzub	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x42,0xc0]
+         setzub	al
+# CHECK: setzuae	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x43,0xc0]
+         setzuae	al
+# CHECK: setzue	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x44,0xc0]
+         setzue	al
+# CHECK: setzune	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x45,0xc0]
+         setzune	al
+# CHECK: setzube	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x46,0xc0]
+         setzube	al
+# CHECK: setzua	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x47,0xc0]
+         setzua	al
+# CHECK: setzus	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x48,0xc0]
+         setzus	al
+# CHECK: setzuns	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x49,0xc0]
+         setzuns	al
+# CHECK: setzup	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4a,0xc0]
+         setzup	al
+# CHECK: setzunp	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4b,0xc0]
+         setzunp	al
+# CHECK: setzul	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4c,0xc0]
+         setzul	al
+# CHECK: setzuge	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4d,0xc0]
+         setzuge	al
+# CHECK: setzule	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4e,0xc0]
+         setzule	al
+# CHECK: setzug	al
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4f,0xc0]
+         setzug	al
+# CHECK: setzuo	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x40,0x00]
+         setzuo	byte ptr [rax]
+# CHECK: setzuno	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x41,0x00]
+         setzuno	byte ptr [rax]
+# CHECK: setzub	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x42,0x00]
+         setzub	byte ptr [rax]
+# CHECK: setzuae	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x43,0x00]
+         setzuae	byte ptr [rax]
+# CHECK: setzue	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x44,0x00]
+         setzue	byte ptr [rax]
+# CHECK: setzune	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x45,0x00]
+         setzune	byte ptr [rax]
+# CHECK: setzube	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x46,0x00]
+         setzube	byte ptr [rax]
+# CHECK: setzua	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x47,0x00]
+         setzua	byte ptr [rax]
+# CHECK: setzus	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x48,0x00]
+         setzus	byte ptr [rax]
+# CHECK: setzuns	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x49,0x00]
+         setzuns	byte ptr [rax]
+# CHECK: setzup	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4a,0x00]
+         setzup	byte ptr [rax]
+# CHECK: setzunp	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4b,0x00]
+         setzunp	byte ptr [rax]
+# CHECK: setzul	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4c,0x00]
+         setzul	byte ptr [rax]
+# CHECK: setzuge	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4d,0x00]
+         setzuge	byte ptr [rax]
+# CHECK: setzule	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4e,0x00]
+         setzule	byte ptr [rax]
+# CHECK: setzug	byte ptr [rax]
+# CHECK: encoding: [0x62,0xf4,0x7f,0x18,0x4f,0x00]
+         setzug	byte ptr [rax]
diff --git a/llvm/test/TableGen/x86-fold-tables.inc b/llvm/test/TableGen/x86-fold-tables.inc
index 493350d7bd630..6f983d2efabe9 100644
--- a/llvm/test/TableGen/x86-fold-tables.inc
+++ b/llvm/test/TableGen/x86-fold-tables.inc
@@ -468,6 +468,7 @@ static const X86FoldTableEntry Table0[] = {
   {X86::PUSH32r, X86::PUSH32rmm, TB_FOLDED_LOAD},
   {X86::PUSH64r, X86::PUSH64rmm, TB_FOLDED_LOAD},
   {X86::SETCCr, X86::SETCCm, TB_FOLDED_STORE},
+  {X86::SETZUCCr, X86::SETZUCCm, TB_FOLDED_STORE},
   {X86::TAILJMPr, X86::TAILJMPm, TB_FOLDED_LOAD},
   {X86::TAILJMPr64, X86::TAILJMPm64, TB_FOLDED_LOAD},
   {X86::TAILJMPr64_REX, X86::TAILJMPm64_REX, TB_FOLDED_LOAD},
diff --git a/llvm/utils/TableGen/AsmMatcherEmitter.cpp b/llvm/utils/TableGen/AsmMatcherEmitter.cpp
index 8b82ce899a48a..53d49a2900a15 100644
--- a/llvm/utils/TableGen/AsmMatcherEmitter.cpp
+++ b/llvm/utils/TableGen/AsmMatcherEmitter.cpp
@@ -645,12 +645,13 @@ struct MatchableInfo {
     // vex encoding size is smaller. Since X86InstrSSE.td is included ahead
     // of X86InstrAVX512.td, the AVX instruction ID is less than AVX512 ID.
     // We use the ID to sort AVX instruction before AVX512 instruction in
-    // matching table.
-    if (TheDef->isSubClassOf("Instruction") &&
-        TheDef->getValueAsBit("HasPositionOrder") &&
-        RHS.TheDef->isSubClassOf("Instruction") &&
-        RHS.TheDef->getValueAsBit("HasPositionOrder"))
-      return TheDef->getID() < RHS.TheDef->getID();
+    // matching table. As well as InstAlias.
+    if (getResultInst()->TheDef->isSubClassOf("Instruction") &&
+        getResultInst()->TheDef->getValueAsBit("HasPositionOrder") &&
+        RHS.getResultInst()->TheDef->isSubClassOf("Instruction") &&
+        RHS.getResultInst()->TheDef->getValueAsBit("HasPositionOrder"))
+      return getResultInst()->TheDef->getID() <
+             RHS.getResultInst()->TheDef->getID();
 
     // Give matches that require more features higher precedence. This is useful
     // because we cannot define AssemblerPredicates with the negation of

From fd50151180498f0de4fe26ff21d3e3b8accc4de0 Mon Sep 17 00:00:00 2001
From: Jianjian Guan <jacquesguan@me.com>
Date: Thu, 11 Apr 2024 10:23:26 +0800
Subject: [PATCH 089/886] [RISCV] Only support SPLAT_VECTOR for Zvfhmin when
 also enable the scalar extension of half fp (#88275)

---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp |  3 +-
 llvm/test/CodeGen/RISCV/rvv/vsplats-fp.ll   | 35 +++++++++++++++------
 2 files changed, 28 insertions(+), 10 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 944d8b6de895d..1a3ef6feea3e3 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1063,7 +1063,8 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
         setOperationAction({ISD::CONCAT_VECTORS, ISD::INSERT_SUBVECTOR,
                             ISD::EXTRACT_SUBVECTOR, ISD::SCALAR_TO_VECTOR},
                            VT, Custom);
-        setOperationAction(ISD::SPLAT_VECTOR, VT, Custom);
+        if (Subtarget.hasStdExtZfhminOrZhinxmin())
+          setOperationAction(ISD::SPLAT_VECTOR, VT, Custom);
         // load/store
         setOperationAction({ISD::LOAD, ISD::STORE}, VT, Custom);
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/vsplats-fp.ll b/llvm/test/CodeGen/RISCV/rvv/vsplats-fp.ll
index 707ef8a94432d..1aebd32cefca6 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vsplats-fp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vsplats-fp.ll
@@ -1,19 +1,36 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+f,+d,+zfh,+zvfh,+v -target-abi ilp32d -verify-machineinstrs < %s \
-; RUN:   | FileCheck %s --check-prefixes=CHECK,OPTIMIZED
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFH,OPTIMIZED
 ; RUN: llc -mtriple=riscv64 -mattr=+f,+d,+zfh,+zvfh,+v -target-abi lp64d -verify-machineinstrs < %s \
-; RUN:   | FileCheck %s --check-prefixes=CHECK,OPTIMIZED
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFH,OPTIMIZED
 ; RUN: llc -mtriple=riscv32 -mattr=+f,+d,+zfh,+zvfh,+v,+no-optimized-zero-stride-load -target-abi ilp32d -verify-machineinstrs < %s \
-; RUN:   | FileCheck %s --check-prefixes=CHECK,NOT-OPTIMIZED
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFH,NOT-OPTIMIZED
 ; RUN: llc -mtriple=riscv64 -mattr=+f,+d,+zfh,+zvfh,+v,+no-optimized-zero-stride-load -target-abi lp64d -verify-machineinstrs < %s \
-; RUN:   | FileCheck %s --check-prefixes=CHECK,NOT-OPTIMIZED
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFH,NOT-OPTIMIZED
+; RUN: llc -mtriple=riscv32 -mattr=+f,+d,+zfh,+zvfhmin,+v -target-abi ilp32d -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFHMIN,OPTIMIZED
+; RUN: llc -mtriple=riscv64 -mattr=+f,+d,+zfh,+zvfhmin,+v -target-abi lp64d -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFHMIN,OPTIMIZED
+; RUN: llc -mtriple=riscv32 -mattr=+f,+d,+zfh,+zvfhmin,+v,+no-optimized-zero-stride-load -target-abi ilp32d -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFHMIN,NOT-OPTIMIZED
+; RUN: llc -mtriple=riscv64 -mattr=+f,+d,+zfh,+zvfhmin,+v,+no-optimized-zero-stride-load -target-abi lp64d -verify-machineinstrs < %s \
+; RUN:   | FileCheck %s --check-prefixes=CHECK,ZVFHMIN,NOT-OPTIMIZED
 
 define <vscale x 8 x half> @vsplat_nxv8f16(half %f) {
-; CHECK-LABEL: vsplat_nxv8f16:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
-; CHECK-NEXT:    vfmv.v.f v8, fa0
-; CHECK-NEXT:    ret
+; ZVFH-LABEL: vsplat_nxv8f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; ZVFH-NEXT:    vfmv.v.f v8, fa0
+; ZVFH-NEXT:    ret
+;
+; ZVFHMIN-LABEL: vsplat_nxv8f16:
+; ZVFHMIN:       # %bb.0:
+; ZVFHMIN-NEXT:    fcvt.s.h fa5, fa0
+; ZVFHMIN-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
+; ZVFHMIN-NEXT:    vfmv.v.f v12, fa5
+; ZVFHMIN-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVFHMIN-NEXT:    vfncvt.f.f.w v8, v12
+; ZVFHMIN-NEXT:    ret
   %head = insertelement <vscale x 8 x half> poison, half %f, i32 0
   %splat = shufflevector <vscale x 8 x half> %head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
   ret <vscale x 8 x half> %splat

From 999b9e6ddb4324600a46c8f7006acec81fe3af0f Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 10 Apr 2024 19:36:31 -0700
Subject: [PATCH 090/886] [RISCV] Use vector getConstant instead of
 getSplatVector+getConstant. NFC

---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 1a3ef6feea3e3..357432081ddb0 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -4634,9 +4634,7 @@ static SDValue getWideningInterleave(SDValue EvenV, SDValue OddV,
   } else if (Subtarget.hasStdExtZvbb()) {
     // Interleaved = (OddV << VecVT.getScalarSizeInBits()) + EvenV.
     SDValue OffsetVec =
-        DAG.getSplatVector(VecContainerVT, DL,
-                           DAG.getConstant(VecVT.getScalarSizeInBits(), DL,
-                                           Subtarget.getXLenVT()));
+        DAG.getConstant(VecVT.getScalarSizeInBits(), DL, VecContainerVT);
     Interleaved = DAG.getNode(RISCVISD::VWSLL_VL, DL, WideContainerVT, OddV,
                               OffsetVec, Passthru, Mask, VL);
     if (!EvenV.isUndef())

From dda73336ad22bd0b5ecda17040c50fb10fcbe5fb Mon Sep 17 00:00:00 2001
From: Mingming Liu <mingmingl@google.com>
Date: Wed, 10 Apr 2024 19:46:01 -0700
Subject: [PATCH 091/886] [ThinLTO]Record import type in
 GlobalValueSummary::GVFlags (#87597)

The motivating use case is to support import the function declaration
across modules to construct call graph edges for indirect calls [1]
when importing the function definition costs too much compile time
(e.g., the function is too large has no `noinline` attribute).
1. Currently, when the compiled IR module doesn't have a function
definition but its postlink combined summary contains the function
summary or a global alias summary with this function as aliasee, the
function definition will be imported from source module by IRMover. The
implementation is in FunctionImporter::importFunctions [2]
2. In order for FunctionImporter to import a declaration of a function,
both function summary and alias summary need to carry the def / decl
state. Specifically, all existing summary fields doesn't differ across
import modules, but the def / decl state of is decided by
`<ImportModule, Function>`.

This change encodes the def/decl state in `GlobalValueSummary::GVFlags`.

In the subsequent changes
1. The indexing step `computeImportForModule` [3]
will compute the set of definitions and the set of declarations for each
module, and passing on the information to bitcode writer.
2. Bitcode writer will look up the def/decl state and sets the state
when it writes out the flag value. This is demonstrated in
https://github.com/llvm/llvm-project/pull/87600
3. Function importer will read the def/decl state when reading the
combined summary to figure out two sets of global values, and IRMover
will be updated to import the declaration (aka linkGlobalValuePrototype [4])
into the destination module.

- The next change is https://github.com/llvm/llvm-project/pull/87600

[1] mentioned in rfc https://discourse.llvm.org/t/rfc-for-better-call-graph-sort-build-a-more-complete-call-graph-by-adding-more-indirect-call-edges/74029#support-cross-module-function-declaration-import-5
[2] https://github.com/llvm/llvm-project/blob/3b337242ee165554f0017b00671381ec5b1ba855/llvm/lib/Transforms/IPO/FunctionImport.cpp#L1608-L1764
[3] https://github.com/llvm/llvm-project/blob/3b337242ee165554f0017b00671381ec5b1ba855/llvm/lib/Transforms/IPO/FunctionImport.cpp#L856
[4] https://github.com/llvm/llvm-project/blob/3b337242ee165554f0017b00671381ec5b1ba855/llvm/lib/Linker/IRMover.cpp#L605
---
 .../CodeGen/thinlto-distributed-cfi-devirt.ll |  2 +-
 clang/test/CodeGen/thinlto-distributed-cfi.ll |  2 +-
 clang/test/CodeGen/thinlto-funcattr-prop.ll   |  4 +-
 lld/test/ELF/lto/comdat-nodeduplicate.ll      |  8 +-
 llvm/include/llvm/AsmParser/LLParser.h        |  2 +
 llvm/include/llvm/AsmParser/LLToken.h         |  3 +
 llvm/include/llvm/IR/ModuleSummaryIndex.h     | 29 +++++-
 llvm/include/llvm/IR/ModuleSummaryIndexYAML.h | 11 ++-
 llvm/lib/Analysis/ModuleSummaryAnalysis.cpp   | 12 ++-
 llvm/lib/AsmParser/LLLexer.cpp                |  3 +
 llvm/lib/AsmParser/LLParser.cpp               | 33 ++++++-
 llvm/lib/Bitcode/Reader/BitcodeReader.cpp     |  3 +-
 llvm/lib/Bitcode/Writer/BitcodeWriter.cpp     |  2 +
 llvm/lib/IR/AsmWriter.cpp                     | 12 +++
 llvm/lib/IR/ModuleSummaryIndex.cpp            |  4 +
 .../test/Assembler/thinlto-memprof-summary.ll | 20 ++--
 .../thinlto-multiple-summaries-for-guid.ll    |  4 +-
 .../Assembler/thinlto-summary-visibility.ll   |  6 +-
 llvm/test/Assembler/thinlto-summary.ll        | 92 ++++++++++---------
 llvm/test/Assembler/thinlto-vtable-summary.ll |  4 +-
 llvm/test/Bitcode/thinlto-alias.ll            | 10 +-
 .../thinlto-func-summary-vtableref-pgo.ll     |  2 +-
 ...ction-summary-callgraph-profile-summary.ll | 18 ++--
 ...hinlto-function-summary-callgraph-relbf.ll |  2 +-
 .../thinlto-function-summary-refgraph.ll      | 14 +--
 .../thinlto-index-disassembled-by-llvm-dis.ll |  2 +-
 llvm/test/Bitcode/thinlto-type-tests.ll       | 12 +--
 llvm/test/Bitcode/thinlto-type-vcalls.ll      | 24 ++---
 llvm/test/ThinLTO/X86/dot-dumper.ll           | 12 +--
 .../ThinLTO/X86/funcattrs-prop-maythrow.ll    |  8 +-
 .../ThinLTO/X86/funcimport_alwaysinline.ll    |  2 +-
 llvm/test/ThinLTO/X86/load-store-caching.ll   |  4 +-
 .../Transforms/LowerTypeTests/import-unsat.ll |  1 +
 .../Inputs/import-indir.yaml                  |  4 +
 .../WholeProgramDevirt/import-indir.ll        |  9 ++
 35 files changed, 241 insertions(+), 139 deletions(-)

diff --git a/clang/test/CodeGen/thinlto-distributed-cfi-devirt.ll b/clang/test/CodeGen/thinlto-distributed-cfi-devirt.ll
index 2309ed717c2a2..433fd1fe20430 100644
--- a/clang/test/CodeGen/thinlto-distributed-cfi-devirt.ll
+++ b/clang/test/CodeGen/thinlto-distributed-cfi-devirt.ll
@@ -34,7 +34,7 @@
 ; Round trip it through llvm-as
 ; RUN: llvm-dis %t.o.thinlto.bc -o - | llvm-as -o - | llvm-dis -o - | FileCheck %s --check-prefix=CHECK-DIS
 ; CHECK-DIS: ^0 = module: (path: "{{.*}}thinlto-distributed-cfi-devirt.ll.tmp.o", hash: ({{.*}}, {{.*}}, {{.*}}, {{.*}}, {{.*}}))
-; CHECK-DIS: ^1 = gv: (guid: 8346051122425466633, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 18, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTests: (^2), typeCheckedLoadVCalls: (vFuncId: (^2, offset: 8), vFuncId: (^2, offset: 0))))))
+; CHECK-DIS: ^1 = gv: (guid: 8346051122425466633, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 18, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTests: (^2), typeCheckedLoadVCalls: (vFuncId: (^2, offset: 8), vFuncId: (^2, offset: 0))))))
 ; CHECK-DIS: ^2 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi"))))) ; guid = 7004155349499253778
 
 ; RUN: %clang_cc1 -triple x86_64-grtev4-linux-gnu \
diff --git a/clang/test/CodeGen/thinlto-distributed-cfi.ll b/clang/test/CodeGen/thinlto-distributed-cfi.ll
index f5dde2d32a42d..47e56c091a612 100644
--- a/clang/test/CodeGen/thinlto-distributed-cfi.ll
+++ b/clang/test/CodeGen/thinlto-distributed-cfi.ll
@@ -24,7 +24,7 @@
 ; Round trip it through llvm-as
 ; RUN: llvm-dis %t.o.thinlto.bc -o - | llvm-as -o - | llvm-dis -o - | FileCheck %s --check-prefix=CHECK-DIS
 ; CHECK-DIS: ^0 = module: (path: "{{.*}}thinlto-distributed-cfi.ll.tmp.o", hash: ({{.*}}, {{.*}}, {{.*}}, {{.*}}, {{.*}}))
-; CHECK-DIS: ^1 = gv: (guid: 8346051122425466633, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 7, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), typeIdInfo: (typeTests: (^2)))))
+; CHECK-DIS: ^1 = gv: (guid: 8346051122425466633, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 7, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), typeIdInfo: (typeTests: (^2)))))
 ; CHECK-DIS: ^2 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0))) ; guid = 7004155349499253778
 
 ; RUN: %clang_cc1 -triple x86_64-grtev4-linux-gnu \
diff --git a/clang/test/CodeGen/thinlto-funcattr-prop.ll b/clang/test/CodeGen/thinlto-funcattr-prop.ll
index c1274776fe9ce..daaa6e2da8048 100644
--- a/clang/test/CodeGen/thinlto-funcattr-prop.ll
+++ b/clang/test/CodeGen/thinlto-funcattr-prop.ll
@@ -15,9 +15,9 @@
 ; RUN: llvm-dis %t1.o.1.1.promote.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
 ;; Summary for call_extern. Note that llvm-lto2 writes out the index before propagation occurs so call_extern doesn't have its flags updated.
-; CHECK-INDEX: ^2 = gv: (guid: 13959900437860518209, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0), insts: 2, calls: ((callee: ^3)))))
+; CHECK-INDEX: ^2 = gv: (guid: 13959900437860518209, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^3)))))
 ;; Summary for extern
-; CHECK-INDEX: ^3 = gv: (guid: 14959766916849974397, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
+; CHECK-INDEX: ^3 = gv: (guid: 14959766916849974397, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
 
 ;--- a.ll
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
diff --git a/lld/test/ELF/lto/comdat-nodeduplicate.ll b/lld/test/ELF/lto/comdat-nodeduplicate.ll
index eef17d41c5a4d..13d4ab394813b 100644
--- a/lld/test/ELF/lto/comdat-nodeduplicate.ll
+++ b/lld/test/ELF/lto/comdat-nodeduplicate.ll
@@ -56,15 +56,15 @@
 ; IR_AB-DAG: gv: (name: "__profc_foo", {{.*}} guid = [[PROFC:[0-9]+]]
 
 ;; Check extra attributes. b.bc:__profc_foo is prevailing, so it can be internalized.
-; IR_AB-DAG: gv: (guid: [[PROFD]], summaries: (variable: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0),
-; IR_AB-DAG: gv: (guid: [[PROFC]], summaries: (variable: (module: ^0, flags: (linkage: internal, visibility: hidden, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
+; IR_AB-DAG: gv: (guid: [[PROFD]], summaries: (variable: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0),
+; IR_AB-DAG: gv: (guid: [[PROFC]], summaries: (variable: (module: ^0, flags: (linkage: internal, visibility: hidden, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
 
 ; IR_ABC-DAG: gv: (name: "__profd_foo", {{.*}} guid = [[PROFD:[0-9]+]]
 ; IR_ABC-DAG: gv: (name: "__profc_foo", {{.*}} guid = [[PROFC:[0-9]+]]
 
 ;; b.bc:__profc_foo prevails c.bc:__profc_foo, so it is exported and therefore not internalized.
-; IR_ABC-DAG: gv: (guid: [[PROFD]], summaries: (variable: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0),
-; IR_ABC-DAG: gv: (guid: [[PROFC]], summaries: (variable: (module: ^0, flags: (linkage: weak, visibility: hidden, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
+; IR_ABC-DAG: gv: (guid: [[PROFD]], summaries: (variable: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0),
+; IR_ABC-DAG: gv: (guid: [[PROFC]], summaries: (variable: (module: ^0, flags: (linkage: weak, visibility: hidden, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
 
 ;--- a.ll
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
diff --git a/llvm/include/llvm/AsmParser/LLParser.h b/llvm/include/llvm/AsmParser/LLParser.h
index b4e971fea1a13..b2dcdfad0a04b 100644
--- a/llvm/include/llvm/AsmParser/LLParser.h
+++ b/llvm/include/llvm/AsmParser/LLParser.h
@@ -301,6 +301,8 @@ namespace llvm {
                               bool &DSOLocal);
     void parseOptionalDSOLocal(bool &DSOLocal);
     void parseOptionalVisibility(unsigned &Res);
+    bool parseOptionalImportType(lltok::Kind Kind,
+                                 GlobalValueSummary::ImportKind &Res);
     void parseOptionalDLLStorageClass(unsigned &Res);
     bool parseOptionalCallingConv(unsigned &CC);
     bool parseOptionalAlignment(MaybeAlign &Alignment,
diff --git a/llvm/include/llvm/AsmParser/LLToken.h b/llvm/include/llvm/AsmParser/LLToken.h
index 65ccb1b81b3a8..0cbcdcd9ffac7 100644
--- a/llvm/include/llvm/AsmParser/LLToken.h
+++ b/llvm/include/llvm/AsmParser/LLToken.h
@@ -370,6 +370,9 @@ enum Kind {
   kw_live,
   kw_dsoLocal,
   kw_canAutoHide,
+  kw_importType,
+  kw_definition,
+  kw_declaration,
   kw_function,
   kw_insts,
   kw_funcFlags,
diff --git a/llvm/include/llvm/IR/ModuleSummaryIndex.h b/llvm/include/llvm/IR/ModuleSummaryIndex.h
index d5ff15063671d..5d137d4b3553c 100644
--- a/llvm/include/llvm/IR/ModuleSummaryIndex.h
+++ b/llvm/include/llvm/IR/ModuleSummaryIndex.h
@@ -432,6 +432,18 @@ class GlobalValueSummary {
   /// Sububclass discriminator (for dyn_cast<> et al.)
   enum SummaryKind : unsigned { AliasKind, FunctionKind, GlobalVarKind };
 
+  enum ImportKind : unsigned {
+    // The global value definition corresponding to the summary should be
+    // imported from source module
+    Definition = 0,
+
+    // When its definition doesn't exist in the destination module and not
+    // imported (e.g., function is too large to be inlined), the global value
+    // declaration corresponding to the summary should be imported, or the
+    // attributes from summary should be annotated on the function declaration.
+    Declaration = 1,
+  };
+
   /// Group flags (Linkage, NotEligibleToImport, etc.) as a bitfield.
   struct GVFlags {
     /// The linkage type of the associated global value.
@@ -472,14 +484,19 @@ class GlobalValueSummary {
     /// means the symbol was externally visible.
     unsigned CanAutoHide : 1;
 
+    /// This field is written by the ThinLTO indexing step to postlink combined
+    /// summary. The value is interpreted as 'ImportKind' enum defined above.
+    unsigned ImportType : 1;
+
     /// Convenience Constructors
     explicit GVFlags(GlobalValue::LinkageTypes Linkage,
                      GlobalValue::VisibilityTypes Visibility,
                      bool NotEligibleToImport, bool Live, bool IsLocal,
-                     bool CanAutoHide)
+                     bool CanAutoHide, ImportKind ImportType)
         : Linkage(Linkage), Visibility(Visibility),
           NotEligibleToImport(NotEligibleToImport), Live(Live),
-          DSOLocal(IsLocal), CanAutoHide(CanAutoHide) {}
+          DSOLocal(IsLocal), CanAutoHide(CanAutoHide),
+          ImportType(static_cast<unsigned>(ImportType)) {}
   };
 
 private:
@@ -564,6 +581,12 @@ class GlobalValueSummary {
 
   bool canAutoHide() const { return Flags.CanAutoHide; }
 
+  bool shouldImportAsDecl() const {
+    return Flags.ImportType == GlobalValueSummary::ImportKind::Declaration;
+  }
+
+  void setImportKind(ImportKind IK) { Flags.ImportType = IK; }
+
   GlobalValue::VisibilityTypes getVisibility() const {
     return (GlobalValue::VisibilityTypes)Flags.Visibility;
   }
@@ -813,7 +836,7 @@ class FunctionSummary : public GlobalValueSummary {
             GlobalValue::LinkageTypes::AvailableExternallyLinkage,
             GlobalValue::DefaultVisibility,
             /*NotEligibleToImport=*/true, /*Live=*/true, /*IsLocal=*/false,
-            /*CanAutoHide=*/false),
+            /*CanAutoHide=*/false, GlobalValueSummary::ImportKind::Definition),
         /*NumInsts=*/0, FunctionSummary::FFlags{}, /*EntryCount=*/0,
         std::vector<ValueInfo>(), std::move(Edges),
         std::vector<GlobalValue::GUID>(),
diff --git a/llvm/include/llvm/IR/ModuleSummaryIndexYAML.h b/llvm/include/llvm/IR/ModuleSummaryIndexYAML.h
index 33e57e5f2102f..b2747d24c5396 100644
--- a/llvm/include/llvm/IR/ModuleSummaryIndexYAML.h
+++ b/llvm/include/llvm/IR/ModuleSummaryIndexYAML.h
@@ -138,6 +138,7 @@ template <> struct MappingTraits<TypeIdSummary> {
 struct FunctionSummaryYaml {
   unsigned Linkage, Visibility;
   bool NotEligibleToImport, Live, IsLocal, CanAutoHide;
+  unsigned ImportType;
   std::vector<uint64_t> Refs;
   std::vector<uint64_t> TypeTests;
   std::vector<FunctionSummary::VFuncId> TypeTestAssumeVCalls,
@@ -183,6 +184,7 @@ template <> struct MappingTraits<FunctionSummaryYaml> {
     io.mapOptional("Live", summary.Live);
     io.mapOptional("Local", summary.IsLocal);
     io.mapOptional("CanAutoHide", summary.CanAutoHide);
+    io.mapOptional("ImportType", summary.ImportType);
     io.mapOptional("Refs", summary.Refs);
     io.mapOptional("TypeTests", summary.TypeTests);
     io.mapOptional("TypeTestAssumeVCalls", summary.TypeTestAssumeVCalls);
@@ -227,7 +229,8 @@ template <> struct CustomMappingTraits<GlobalValueSummaryMapTy> {
               static_cast<GlobalValue::LinkageTypes>(FSum.Linkage),
               static_cast<GlobalValue::VisibilityTypes>(FSum.Visibility),
               FSum.NotEligibleToImport, FSum.Live, FSum.IsLocal,
-              FSum.CanAutoHide),
+              FSum.CanAutoHide,
+              static_cast<GlobalValueSummary::ImportKind>(FSum.ImportType)),
           /*NumInsts=*/0, FunctionSummary::FFlags{}, /*EntryCount=*/0, Refs,
           ArrayRef<FunctionSummary::EdgeTy>{}, std::move(FSum.TypeTests),
           std::move(FSum.TypeTestAssumeVCalls),
@@ -251,9 +254,9 @@ template <> struct CustomMappingTraits<GlobalValueSummaryMapTy> {
               static_cast<bool>(FSum->flags().NotEligibleToImport),
               static_cast<bool>(FSum->flags().Live),
               static_cast<bool>(FSum->flags().DSOLocal),
-              static_cast<bool>(FSum->flags().CanAutoHide), Refs,
-              FSum->type_tests(), FSum->type_test_assume_vcalls(),
-              FSum->type_checked_load_vcalls(),
+              static_cast<bool>(FSum->flags().CanAutoHide),
+              FSum->flags().ImportType, Refs, FSum->type_tests(),
+              FSum->type_test_assume_vcalls(), FSum->type_checked_load_vcalls(),
               FSum->type_test_assume_const_vcalls(),
               FSum->type_checked_load_const_vcalls()});
           }
diff --git a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
index 3ad0bab827a51..deda1eebb3b57 100644
--- a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
+++ b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
@@ -635,7 +635,8 @@ static void computeFunctionSummary(
                               HasIndirBranchToBlockAddress || HasIFuncCall;
   GlobalValueSummary::GVFlags Flags(
       F.getLinkage(), F.getVisibility(), NotEligibleForImport,
-      /* Live = */ false, F.isDSOLocal(), F.canBeOmittedFromSymbolTable());
+      /* Live = */ false, F.isDSOLocal(), F.canBeOmittedFromSymbolTable(),
+      GlobalValueSummary::ImportKind::Definition);
   FunctionSummary::FFlags FunFlags{
       F.doesNotAccessMemory(), F.onlyReadsMemory() && !F.doesNotAccessMemory(),
       F.hasFnAttribute(Attribute::NoRecurse), F.returnDoesNotAlias(),
@@ -761,7 +762,8 @@ static void computeVariableSummary(ModuleSummaryIndex &Index,
   bool NonRenamableLocal = isNonRenamableLocal(V);
   GlobalValueSummary::GVFlags Flags(
       V.getLinkage(), V.getVisibility(), NonRenamableLocal,
-      /* Live = */ false, V.isDSOLocal(), V.canBeOmittedFromSymbolTable());
+      /* Live = */ false, V.isDSOLocal(), V.canBeOmittedFromSymbolTable(),
+      GlobalValueSummary::Definition);
 
   VTableFuncList VTableFuncs;
   // If splitting is not enabled, then we compute the summary information
@@ -807,7 +809,8 @@ static void computeAliasSummary(ModuleSummaryIndex &Index, const GlobalAlias &A,
   bool NonRenamableLocal = isNonRenamableLocal(A);
   GlobalValueSummary::GVFlags Flags(
       A.getLinkage(), A.getVisibility(), NonRenamableLocal,
-      /* Live = */ false, A.isDSOLocal(), A.canBeOmittedFromSymbolTable());
+      /* Live = */ false, A.isDSOLocal(), A.canBeOmittedFromSymbolTable(),
+      GlobalValueSummary::Definition);
   auto AS = std::make_unique<AliasSummary>(Flags);
   auto AliaseeVI = Index.getValueInfo(Aliasee->getGUID());
   assert(AliaseeVI && "Alias expects aliasee summary to be available");
@@ -887,7 +890,8 @@ ModuleSummaryIndex llvm::buildModuleSummaryIndex(
               GlobalValue::InternalLinkage, GlobalValue::DefaultVisibility,
               /* NotEligibleToImport = */ true,
               /* Live = */ true,
-              /* Local */ GV->isDSOLocal(), GV->canBeOmittedFromSymbolTable());
+              /* Local */ GV->isDSOLocal(), GV->canBeOmittedFromSymbolTable(),
+              GlobalValueSummary::Definition);
           CantBePromoted.insert(GV->getGUID());
           // Create the appropriate summary type.
           if (Function *F = dyn_cast<Function>(GV)) {
diff --git a/llvm/lib/AsmParser/LLLexer.cpp b/llvm/lib/AsmParser/LLLexer.cpp
index 2301a27731eaf..8ded07ffd8bd2 100644
--- a/llvm/lib/AsmParser/LLLexer.cpp
+++ b/llvm/lib/AsmParser/LLLexer.cpp
@@ -737,6 +737,9 @@ lltok::Kind LLLexer::LexIdentifier() {
   KEYWORD(live);
   KEYWORD(dsoLocal);
   KEYWORD(canAutoHide);
+  KEYWORD(importType);
+  KEYWORD(definition);
+  KEYWORD(declaration);
   KEYWORD(function);
   KEYWORD(insts);
   KEYWORD(funcFlags);
diff --git a/llvm/lib/AsmParser/LLParser.cpp b/llvm/lib/AsmParser/LLParser.cpp
index f546e05a5d37d..63104129f8c2d 100644
--- a/llvm/lib/AsmParser/LLParser.cpp
+++ b/llvm/lib/AsmParser/LLParser.cpp
@@ -2083,6 +2083,20 @@ void LLParser::parseOptionalVisibility(unsigned &Res) {
   Lex.Lex();
 }
 
+bool LLParser::parseOptionalImportType(lltok::Kind Kind,
+                                       GlobalValueSummary::ImportKind &Res) {
+  switch (Kind) {
+  default:
+    return tokError("unknown import kind. Expect definition or declaration.");
+  case lltok::kw_definition:
+    Res = GlobalValueSummary::Definition;
+    return false;
+  case lltok::kw_declaration:
+    Res = GlobalValueSummary::Declaration;
+    return false;
+  }
+}
+
 /// parseOptionalDLLStorageClass
 ///   ::= /*empty*/
 ///   ::= 'dllimport'
@@ -9230,7 +9244,8 @@ bool LLParser::parseFunctionSummary(std::string Name, GlobalValue::GUID GUID,
   GlobalValueSummary::GVFlags GVFlags = GlobalValueSummary::GVFlags(
       GlobalValue::ExternalLinkage, GlobalValue::DefaultVisibility,
       /*NotEligibleToImport=*/false,
-      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false);
+      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false,
+      GlobalValueSummary::Definition);
   unsigned InstCount;
   std::vector<FunctionSummary::EdgeTy> Calls;
   FunctionSummary::TypeIdInfo TypeIdInfo;
@@ -9317,7 +9332,8 @@ bool LLParser::parseVariableSummary(std::string Name, GlobalValue::GUID GUID,
   GlobalValueSummary::GVFlags GVFlags = GlobalValueSummary::GVFlags(
       GlobalValue::ExternalLinkage, GlobalValue::DefaultVisibility,
       /*NotEligibleToImport=*/false,
-      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false);
+      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false,
+      GlobalValueSummary::Definition);
   GlobalVarSummary::GVarFlags GVarFlags(/*ReadOnly*/ false,
                                         /* WriteOnly */ false,
                                         /* Constant */ false,
@@ -9375,7 +9391,8 @@ bool LLParser::parseAliasSummary(std::string Name, GlobalValue::GUID GUID,
   GlobalValueSummary::GVFlags GVFlags = GlobalValueSummary::GVFlags(
       GlobalValue::ExternalLinkage, GlobalValue::DefaultVisibility,
       /*NotEligibleToImport=*/false,
-      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false);
+      /*Live=*/false, /*IsLocal=*/false, /*CanAutoHide=*/false,
+      GlobalValueSummary::Definition);
   if (parseToken(lltok::colon, "expected ':' here") ||
       parseToken(lltok::lparen, "expected '(' here") ||
       parseModuleReference(ModulePath) ||
@@ -10161,6 +10178,16 @@ bool LLParser::parseGVFlags(GlobalValueSummary::GVFlags &GVFlags) {
         return true;
       GVFlags.CanAutoHide = Flag;
       break;
+    case lltok::kw_importType:
+      Lex.Lex();
+      if (parseToken(lltok::colon, "expected ':'"))
+        return true;
+      GlobalValueSummary::ImportKind IK;
+      if (parseOptionalImportType(Lex.getKind(), IK))
+        return true;
+      GVFlags.ImportType = static_cast<unsigned>(IK);
+      Lex.Lex();
+      break;
     default:
       return error(Lex.getLoc(), "expected gv flag type");
     }
diff --git a/llvm/lib/Bitcode/Reader/BitcodeReader.cpp b/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
index 92c349525aff5..fe4f0d6dca6c0 100644
--- a/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
+++ b/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
@@ -1141,6 +1141,7 @@ static GlobalValueSummary::GVFlags getDecodedGVSummaryFlags(uint64_t RawFlags,
   // to getDecodedLinkage() will need to be taken into account here as above.
   auto Linkage = GlobalValue::LinkageTypes(RawFlags & 0xF); // 4 bits
   auto Visibility = GlobalValue::VisibilityTypes((RawFlags >> 8) & 3); // 2 bits
+  auto IK = GlobalValueSummary::ImportKind((RawFlags >> 10) & 1);      // 1 bit
   RawFlags = RawFlags >> 4;
   bool NotEligibleToImport = (RawFlags & 0x1) || Version < 3;
   // The Live flag wasn't introduced until version 3. For dead stripping
@@ -1151,7 +1152,7 @@ static GlobalValueSummary::GVFlags getDecodedGVSummaryFlags(uint64_t RawFlags,
   bool AutoHide = (RawFlags & 0x8);
 
   return GlobalValueSummary::GVFlags(Linkage, Visibility, NotEligibleToImport,
-                                     Live, Local, AutoHide);
+                                     Live, Local, AutoHide, IK);
 }
 
 // Decode the flags for GlobalVariable in the summary
diff --git a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
index dd554e422516f..6d01e3b4d8218 100644
--- a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
+++ b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
@@ -1217,6 +1217,8 @@ static uint64_t getEncodedGVSummaryFlags(GlobalValueSummary::GVFlags Flags) {
 
   RawFlags |= (Flags.Visibility << 8); // 2 bits
 
+  RawFlags |= (Flags.ImportType << 10); // 1 bit
+
   return RawFlags;
 }
 
diff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp
index b778a14158ef2..609de920ba7dd 100644
--- a/llvm/lib/IR/AsmWriter.cpp
+++ b/llvm/lib/IR/AsmWriter.cpp
@@ -3306,6 +3306,16 @@ static const char *getVisibilityName(GlobalValue::VisibilityTypes Vis) {
   llvm_unreachable("invalid visibility");
 }
 
+static const char *getImportTypeName(GlobalValueSummary::ImportKind IK) {
+  switch (IK) {
+  case GlobalValueSummary::Definition:
+    return "definition";
+  case GlobalValueSummary::Declaration:
+    return "declaration";
+  }
+  assert(false && "invalid import kind");
+}
+
 void AssemblyWriter::printFunctionSummary(const FunctionSummary *FS) {
   Out << ", insts: " << FS->instCount();
   if (FS->fflags().anyFlagSet())
@@ -3545,6 +3555,8 @@ void AssemblyWriter::printSummary(const GlobalValueSummary &Summary) {
   Out << ", live: " << GVFlags.Live;
   Out << ", dsoLocal: " << GVFlags.DSOLocal;
   Out << ", canAutoHide: " << GVFlags.CanAutoHide;
+  Out << ", importType: "
+      << getImportTypeName(GlobalValueSummary::ImportKind(GVFlags.ImportType));
   Out << ")";
 
   if (Summary.getSummaryKind() == GlobalValueSummary::AliasKind)
diff --git a/llvm/lib/IR/ModuleSummaryIndex.cpp b/llvm/lib/IR/ModuleSummaryIndex.cpp
index 198c730418c72..6713d32fb787c 100644
--- a/llvm/lib/IR/ModuleSummaryIndex.cpp
+++ b/llvm/lib/IR/ModuleSummaryIndex.cpp
@@ -644,6 +644,10 @@ void ModuleSummaryIndex::exportToDot(
         A.addComment("dsoLocal");
       if (Flags.CanAutoHide)
         A.addComment("canAutoHide");
+      if (Flags.ImportType == GlobalValueSummary::ImportKind::Definition)
+        A.addComment("definition");
+      else if (Flags.ImportType == GlobalValueSummary::ImportKind::Declaration)
+        A.addComment("declaration");
       if (GUIDPreservedSymbols.count(SummaryIt.first))
         A.addComment("preserved");
 
diff --git a/llvm/test/Assembler/thinlto-memprof-summary.ll b/llvm/test/Assembler/thinlto-memprof-summary.ll
index b72271bb401fd..69eafc967c2a3 100644
--- a/llvm/test/Assembler/thinlto-memprof-summary.ll
+++ b/llvm/test/Assembler/thinlto-memprof-summary.ll
@@ -5,20 +5,20 @@
 
 ^0 = module: (path: "thinlto-memprof-summary.o", hash: (1369602428, 2747878711, 259090915, 2507395659, 1141468049))
 ;; Function with single alloc, multiple memprof MIBs, no versioning
-^1 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (none), memProf: ((type: notcold, stackIds: (8632435727821051414)), (type: cold, stackIds: (15025054523792398438, 12345678)), (type: hot, stackIds: (987654321))))))))
+^1 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (none), memProf: ((type: notcold, stackIds: (8632435727821051414)), (type: cold, stackIds: (15025054523792398438, 12345678)), (type: hot, stackIds: (987654321))))))))
 ;; Function with callsite stack ids calling above function, no versioning
-^2 = gv: (guid: 25, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^1)), callsites: ((callee: ^1, clones: (0), stackIds: (8632435727821051414)), (callee: ^1, clones: (0), stackIds: (15025054523792398438, 12345678)), (callee: ^1, clones: (0), stackIds: (23456789))))))
+^2 = gv: (guid: 25, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^1)), callsites: ((callee: ^1, clones: (0), stackIds: (8632435727821051414)), (callee: ^1, clones: (0), stackIds: (15025054523792398438, 12345678)), (callee: ^1, clones: (0), stackIds: (23456789))))))
 ;; Function with multiple allocs, multiple memprof MIBs, multiple versions
-^3 = gv: (guid: 26, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (cold, notcold), memProf: ((type: notcold, stackIds: (3456789)), (type: cold, stackIds: (456789)))), (versions: (notcold, cold), memProf: ((type: cold, stackIds: (3456789)), (type: notcold, stackIds: (456789))))))))
+^3 = gv: (guid: 26, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (cold, notcold), memProf: ((type: notcold, stackIds: (3456789)), (type: cold, stackIds: (456789)))), (versions: (notcold, cold), memProf: ((type: cold, stackIds: (3456789)), (type: notcold, stackIds: (456789))))))))
 ;; Function with callsite stack ids calling above function, multiple versions
-^4 = gv: (guid: 27, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^3)), callsites: ((callee: ^3, clones: (0, 1), stackIds: (3456789)), (callee: ^3, clones: (1, 1), stackIds: (456789))))))
+^4 = gv: (guid: 27, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^3)), callsites: ((callee: ^3, clones: (0, 1), stackIds: (3456789)), (callee: ^3, clones: (1, 1), stackIds: (456789))))))
 ;; Function with null callsite stack id (can happen in distributed indexes if callsite not imported)
-^5 = gv: (guid: 28, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), callsites: ((callee: null, clones: (0), stackIds: (8632435727821051414))))))
+^5 = gv: (guid: 28, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), callsites: ((callee: null, clones: (0), stackIds: (8632435727821051414))))))
 
 ; Make sure we get back from llvm-dis what we put in via llvm-as.
 ; CHECK: ^0 = module: (path: "thinlto-memprof-summary.o", hash: (1369602428, 2747878711, 259090915, 2507395659, 1141468049))
-; CHECK: ^1 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (none), memProf: ((type: notcold, stackIds: (8632435727821051414)), (type: cold, stackIds: (15025054523792398438, 12345678)), (type: hot, stackIds: (987654321))))))))
-; CHECK: ^2 = gv: (guid: 25, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^1)), callsites: ((callee: ^1, clones: (0), stackIds: (8632435727821051414)), (callee: ^1, clones: (0), stackIds: (15025054523792398438, 12345678)), (callee: ^1, clones: (0), stackIds: (23456789))))))
-; CHECK: ^3 = gv: (guid: 26, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (cold, notcold), memProf: ((type: notcold, stackIds: (3456789)), (type: cold, stackIds: (456789)))), (versions: (notcold, cold), memProf: ((type: cold, stackIds: (3456789)), (type: notcold, stackIds: (456789))))))))
-; CHECK: ^4 = gv: (guid: 27, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^3)), callsites: ((callee: ^3, clones: (0, 1), stackIds: (3456789)), (callee: ^3, clones: (1, 1), stackIds: (456789))))))
-; CHECK: ^5 = gv: (guid: 28, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), callsites: ((callee: null, clones: (0), stackIds: (8632435727821051414))))))
+; CHECK: ^1 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (none), memProf: ((type: notcold, stackIds: (8632435727821051414)), (type: cold, stackIds: (15025054523792398438, 12345678)), (type: hot, stackIds: (987654321))))))))
+; CHECK: ^2 = gv: (guid: 25, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^1)), callsites: ((callee: ^1, clones: (0), stackIds: (8632435727821051414)), (callee: ^1, clones: (0), stackIds: (15025054523792398438, 12345678)), (callee: ^1, clones: (0), stackIds: (23456789))))))
+; CHECK: ^3 = gv: (guid: 26, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 2, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), allocs: ((versions: (cold, notcold), memProf: ((type: notcold, stackIds: (3456789)), (type: cold, stackIds: (456789)))), (versions: (notcold, cold), memProf: ((type: cold, stackIds: (3456789)), (type: notcold, stackIds: (456789))))))))
+; CHECK: ^4 = gv: (guid: 27, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^3)), callsites: ((callee: ^3, clones: (0, 1), stackIds: (3456789)), (callee: ^3, clones: (1, 1), stackIds: (456789))))))
+; CHECK: ^5 = gv: (guid: 28, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 22, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0), callsites: ((callee: null, clones: (0), stackIds: (8632435727821051414))))))
diff --git a/llvm/test/Assembler/thinlto-multiple-summaries-for-guid.ll b/llvm/test/Assembler/thinlto-multiple-summaries-for-guid.ll
index 4f849fa6e6adc..117280a279d09 100644
--- a/llvm/test/Assembler/thinlto-multiple-summaries-for-guid.ll
+++ b/llvm/test/Assembler/thinlto-multiple-summaries-for-guid.ll
@@ -8,5 +8,5 @@ source_filename = "index.bc"
 ; CHECK: ^0 = module: (path: "[Regular LTO]", hash: (0, 0, 0, 0, 0))
 ^1 = module: (path: "main.bc", hash: (3499594384, 1671013073, 3271036935, 1830411232, 59290952))
 ; CHECK-NEXT: ^1 = module: (path: "main.bc", hash: (3499594384, 1671013073, 3271036935, 1830411232, 59290952))
-^2 = gv: (guid: 13351721993301222997, summaries: (function: (module: ^1, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 1), insts: 1), function: (module: ^1, flags: (linkage: available_externally, notEligibleToImport: 1, live: 1, dsoLocal: 1, canAutoHide: 0), insts: 1)))
-; CHECK-NEXT: ^2 = gv: (guid: 13351721993301222997, summaries: (function: (module: ^1, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 1), insts: 1), function: (module: ^1, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 1, live: 1, dsoLocal: 1, canAutoHide: 0), insts: 1)))
+^2 = gv: (guid: 13351721993301222997, summaries: (function: (module: ^1, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1), function: (module: ^1, flags: (linkage: available_externally, notEligibleToImport: 1, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK-NEXT: ^2 = gv: (guid: 13351721993301222997, summaries: (function: (module: ^1, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1), function: (module: ^1, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 1, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1)))
diff --git a/llvm/test/Assembler/thinlto-summary-visibility.ll b/llvm/test/Assembler/thinlto-summary-visibility.ll
index 77f652a9e97dc..67ddcb961d7ae 100644
--- a/llvm/test/Assembler/thinlto-summary-visibility.ll
+++ b/llvm/test/Assembler/thinlto-summary-visibility.ll
@@ -4,9 +4,9 @@
 ^0 = module: (path: "thinlto-summary-visibility1.o", hash: (1369602428, 2747878711, 259090915, 2507395659, 1141468049))
 ^1 = module: (path: "thinlto-summary-visibility2.o", hash: (2998369023, 4283347029, 1195487472, 2757298015, 1852134156))
 
-; CHECK:      ^2 = gv: (guid: 2, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 10)))
-; CHECK-NEXT: ^3 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: external, visibility: protected, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 10)))
-; CHECK-NEXT: ^4 = gv: (guid: 4, summaries: (function: (module: ^0, flags: (linkage: external, visibility: hidden, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 10)))
+; CHECK:      ^2 = gv: (guid: 2, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 10)))
+; CHECK-NEXT: ^3 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: external, visibility: protected, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 10)))
+; CHECK-NEXT: ^4 = gv: (guid: 4, summaries: (function: (module: ^0, flags: (linkage: external, visibility: hidden, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 10)))
 
 ^2 = gv: (guid: 2, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default), insts: 10)))
 ^3 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: external, visibility: protected), insts: 10)))
diff --git a/llvm/test/Assembler/thinlto-summary.ll b/llvm/test/Assembler/thinlto-summary.ll
index 9eb3c6669780d..05dad2c7acad4 100644
--- a/llvm/test/Assembler/thinlto-summary.ll
+++ b/llvm/test/Assembler/thinlto-summary.ll
@@ -16,13 +16,13 @@
 ^3 = gv: (guid: 2, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 10, calls: ((callee: ^15, relbf: 256, tail: 1)))))
 
 ; Summaries with different linkage types.
-^4 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: internal, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1), insts: 1)))
+^4 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: internal, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, importType: definition), insts: 1)))
 ; Make this one an alias with a forward reference to aliasee.
-^5 = gv: (guid: 4, summaries: (alias: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1), aliasee: ^14)))
+^5 = gv: (guid: 4, summaries: (alias: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1,  importType: definition), aliasee: ^14)))
 ^6 = gv: (guid: 5, summaries: (function: (module: ^0, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 1)))
 ^7 = gv: (guid: 6, summaries: (function: (module: ^0, flags: (linkage: linkonce, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 1)))
 ^8 = gv: (guid: 7, summaries: (function: (module: ^0, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 1)))
-^9 = gv: (guid: 8, summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 1), insts: 1)))
+^9 = gv: (guid: 8, summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 1, importType: definition), insts: 1)))
 ^10 = gv: (guid: 9, summaries: (function: (module: ^0, flags: (linkage: weak, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 1)))
 ^11 = gv: (guid: 10, summaries: (variable: (module: ^0, flags: (linkage: common, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), varFlags: (readonly: 0))))
 ; Test appending globel variable with reference (tests backward reference on
@@ -46,60 +46,64 @@
 ^18 = gv: (guid: 17, summaries: (alias: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1), aliasee: ^14)))
 
 ; Test all types of TypeIdInfo on function summaries.
-^19 = gv: (guid: 18, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 4, typeIdInfo: (typeTests: (^24, ^26)))))
-^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^27, offset: 16))))))
-^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^25, offset: 16))))))
-^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (^27, offset: 16), args: (42)), (vFuncId: (^27, offset: 24)))))))
-^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (^28, offset: 16), args: (42)))))))
+^19 = gv: (guid: 18, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 4, typeIdInfo: (typeTests: (^25, ^27)))))
+^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^28, offset: 16))))))
+^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^26, offset: 16))))))
+^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (^28, offset: 16), args: (42)), (vFuncId: (^28, offset: 24)))))))
+^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (^29, offset: 16), args: (42)))))))
+
+; Function summary with an import type of declaration
+^24 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, importType: declaration), insts: 5)))
 
 ; Test TypeId summaries:
 
-^24 = typeid: (name: "_ZTS1C", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0)))
+^25 = typeid: (name: "_ZTS1C", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0)))
 ; Test TypeId with other optional fields (alignLog2/sizeM1/bitMask/inlineBits)
-^25 = typeid: (name: "_ZTS1B", summary: (typeTestRes: (kind: inline, sizeM1BitWidth: 0, alignLog2: 1, sizeM1: 2, bitMask: 3, inlineBits: 4)))
+^26 = typeid: (name: "_ZTS1B", summary: (typeTestRes: (kind: inline, sizeM1BitWidth: 0, alignLog2: 1, sizeM1: 2, bitMask: 3, inlineBits: 4)))
 ; Test the AllOnes resolution, and all kinds of WholeProgramDevirtResolution
 ; types, including all optional resolution by argument kinds.
-^26 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi")), (offset: 16, wpdRes: (kind: indir, resByArg: (args: (1, 2), byArg: (kind: indir, byte: 2, bit: 3), args: (3), byArg: (kind: uniformRetVal, info: 1), args: (4), byArg: (kind: uniqueRetVal, info: 1), args: (5), byArg: (kind: virtualConstProp)))))))
+^27 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi")), (offset: 16, wpdRes: (kind: indir, resByArg: (args: (1, 2), byArg: (kind: indir, byte: 2, bit: 3), args: (3), byArg: (kind: uniformRetVal, info: 1), args: (4), byArg: (kind: uniqueRetVal, info: 1), args: (5), byArg: (kind: virtualConstProp)))))))
 ; Test the other kinds of type test resoultions
-^27 = typeid: (name: "_ZTS1D", summary: (typeTestRes: (kind: byteArray, sizeM1BitWidth: 0)))
-^28 = typeid: (name: "_ZTS1E", summary: (typeTestRes: (kind: unsat, sizeM1BitWidth: 0)))
-^29 = flags: 8
-^30 = blockcount: 1888
+^28 = typeid: (name: "_ZTS1D", summary: (typeTestRes: (kind: byteArray, sizeM1BitWidth: 0)))
+^29 = typeid: (name: "_ZTS1E", summary: (typeTestRes: (kind: unsat, sizeM1BitWidth: 0)))
+^30 = flags: 8
+^31 = blockcount: 1888
 
 ; Make sure we get back from llvm-dis essentially what we put in via llvm-as.
 ; CHECK: ^0 = module: (path: "thinlto-summary1.o", hash: (1369602428, 2747878711, 259090915, 2507395659, 1141468049))
 ; CHECK: ^1 = module: (path: "thinlto-summary2.o", hash: (2998369023, 4283347029, 1195487472, 2757298015, 1852134156))
-; CHECK: ^2 = gv: (guid: 1, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 10, calls: ((callee: ^15, hotness: hot), (callee: ^17, hotness: cold), (callee: ^16, hotness: none, tail: 1)), refs: (^11, readonly ^13, writeonly ^14))))
+; CHECK: ^2 = gv: (guid: 1, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 10, calls: ((callee: ^15, hotness: hot), (callee: ^17, hotness: cold), (callee: ^16, hotness: none, tail: 1)), refs: (^11, readonly ^13, writeonly ^14))))
 ;; relbf is not emitted since this is a combined summary, and that is only
 ;; emitted for per-module summaries.
-; CHECK: ^3 = gv: (guid: 2, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 10, calls: ((callee: ^15, tail: 1)))))
-; CHECK: ^4 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: internal, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 1)))
-; CHECK: ^5 = gv: (guid: 4, summaries: (alias: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), aliasee: ^14)))
-; CHECK: ^6 = gv: (guid: 5, summaries: (function: (module: ^0, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; CHECK: ^7 = gv: (guid: 6, summaries: (function: (module: ^0, flags: (linkage: linkonce, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; CHECK: ^8 = gv: (guid: 7, summaries: (function: (module: ^0, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; CHECK: ^9 = gv: (guid: 8, summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 1), insts: 1)))
-; CHECK: ^10 = gv: (guid: 9, summaries: (function: (module: ^0, flags: (linkage: weak, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; CHECK: ^11 = gv: (guid: 10, summaries: (variable: (module: ^0, flags: (linkage: common, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
-; CHECK: ^12 = gv: (guid: 11, summaries: (variable: (module: ^0, flags: (linkage: appending, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0), refs: (^4))))
-; CHECK: ^13 = gv: (guid: 12, summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 0, constant: 0))))
-; CHECK: ^14 = gv: (guid: 13, summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
-; CHECK: ^15 = gv: (guid: 14, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 1, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
-; CHECK: ^16 = gv: (guid: 15, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1, funcFlags: (readNone: 1, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 1, noUnwind: 1, mayThrow: 1, hasUnknownCall: 1, mustBeUnreachable: 0))))
-; CHECK: ^17 = gv: (guid: 16, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1, funcFlags: (readNone: 0, readOnly: 1, noRecurse: 0, returnDoesNotAlias: 1, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 1), calls: ((callee: ^15)))))
-; CHECK: ^18 = gv: (guid: 17, summaries: (alias: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), aliasee: ^14)))
-; CHECK: ^19 = gv: (guid: 18, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, typeIdInfo: (typeTests: (^24, ^26)))))
-; CHECK: ^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^27, offset: 16))))))
-; CHECK: ^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^25, offset: 16))))))
-; CHECK: ^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (^27, offset: 16), args: (42)), (vFuncId: (^27, offset: 24)))))))
-; CHECK: ^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (^28, offset: 16), args: (42)))))))
-; CHECK: ^24 = typeid: (name: "_ZTS1C", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0))) ; guid = 1884921850105019584
-; CHECK: ^25 = typeid: (name: "_ZTS1B", summary: (typeTestRes: (kind: inline, sizeM1BitWidth: 0, alignLog2: 1, sizeM1: 2, bitMask: 3, inlineBits: 4))) ; guid = 6203814149063363976
-; CHECK: ^26 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi")), (offset: 16, wpdRes: (kind: indir, resByArg: (args: (1, 2), byArg: (kind: indir, byte: 2, bit: 3), args: (3), byArg: (kind: uniformRetVal, info: 1), args: (4), byArg: (kind: uniqueRetVal, info: 1), args: (5), byArg: (kind: virtualConstProp))))))) ; guid = 7004155349499253778
-; CHECK: ^27 = typeid: (name: "_ZTS1D", summary: (typeTestRes: (kind: byteArray, sizeM1BitWidth: 0))) ; guid = 9614786172484273522
-; CHECK: ^28 = typeid: (name: "_ZTS1E", summary: (typeTestRes: (kind: unsat, sizeM1BitWidth: 0))) ; guid = 17437243864166745132
-; CHECK: ^29 = flags: 8
-; CHECK: ^30 = blockcount: 1888
+; CHECK: ^3 = gv: (guid: 2, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 10, calls: ((callee: ^15, tail: 1)))))
+; CHECK: ^4 = gv: (guid: 3, summaries: (function: (module: ^0, flags: (linkage: internal, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK: ^5 = gv: (guid: 4, summaries: (alias: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), aliasee: ^14)))
+; CHECK: ^6 = gv: (guid: 5, summaries: (function: (module: ^0, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK: ^7 = gv: (guid: 6, summaries: (function: (module: ^0, flags: (linkage: linkonce, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK: ^8 = gv: (guid: 7, summaries: (function: (module: ^0, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK: ^9 = gv: (guid: 8, summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 1, importType: definition), insts: 1)))
+; CHECK: ^10 = gv: (guid: 9, summaries: (function: (module: ^0, flags: (linkage: weak, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; CHECK: ^11 = gv: (guid: 10, summaries: (variable: (module: ^0, flags: (linkage: common, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
+; CHECK: ^12 = gv: (guid: 11, summaries: (variable: (module: ^0, flags: (linkage: appending, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0), refs: (^4))))
+; CHECK: ^13 = gv: (guid: 12, summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 1, writeonly: 0, constant: 0))))
+; CHECK: ^14 = gv: (guid: 13, summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0))))
+; CHECK: ^15 = gv: (guid: 14, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 1, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 1, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
+; CHECK: ^16 = gv: (guid: 15, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1, funcFlags: (readNone: 1, readOnly: 0, noRecurse: 1, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 1, noUnwind: 1, mayThrow: 1, hasUnknownCall: 1, mustBeUnreachable: 0))))
+; CHECK: ^17 = gv: (guid: 16, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1, funcFlags: (readNone: 0, readOnly: 1, noRecurse: 0, returnDoesNotAlias: 1, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 1), calls: ((callee: ^15)))))
+; CHECK: ^18 = gv: (guid: 17, summaries: (alias: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), aliasee: ^14)))
+; CHECK: ^19 = gv: (guid: 18, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, typeIdInfo: (typeTests: (^25, ^27)))))
+; CHECK: ^20 = gv: (guid: 19, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 8, typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (^28, offset: 16))))))
+; CHECK: ^21 = gv: (guid: 20, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (^26, offset: 16))))))
+; CHECK: ^22 = gv: (guid: 21, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 15, typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (^28, offset: 16), args: (42)), (vFuncId: (^28, offset: 24)))))))
+; CHECK: ^23 = gv: (guid: 22, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (^29, offset: 16), args: (42)))))))
+; CHECK: ^24 = gv: (guid: 23, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: declaration), insts: 5)))
+; CHECK: ^25 = typeid: (name: "_ZTS1C", summary: (typeTestRes: (kind: single, sizeM1BitWidth: 0))) ; guid = 1884921850105019584
+; CHECK: ^26 = typeid: (name: "_ZTS1B", summary: (typeTestRes: (kind: inline, sizeM1BitWidth: 0, alignLog2: 1, sizeM1: 2, bitMask: 3, inlineBits: 4))) ; guid = 6203814149063363976
+; CHECK: ^27 = typeid: (name: "_ZTS1A", summary: (typeTestRes: (kind: allOnes, sizeM1BitWidth: 7), wpdResolutions: ((offset: 0, wpdRes: (kind: branchFunnel)), (offset: 8, wpdRes: (kind: singleImpl, singleImplName: "_ZN1A1nEi")), (offset: 16, wpdRes: (kind: indir, resByArg: (args: (1, 2), byArg: (kind: indir, byte: 2, bit: 3), args: (3), byArg: (kind: uniformRetVal, info: 1), args: (4), byArg: (kind: uniqueRetVal, info: 1), args: (5), byArg: (kind: virtualConstProp))))))) ; guid = 7004155349499253778
+; CHECK: ^28 = typeid: (name: "_ZTS1D", summary: (typeTestRes: (kind: byteArray, sizeM1BitWidth: 0))) ; guid = 9614786172484273522
+; CHECK: ^29 = typeid: (name: "_ZTS1E", summary: (typeTestRes: (kind: unsat, sizeM1BitWidth: 0))) ; guid = 17437243864166745132
+; CHECK: ^30 = flags: 8
+; CHECK: ^31 = blockcount: 1888
 
 ; Make sure parsing of a non-summary entry containing a ":" does not fail
 ; after summary parsing, which handles colons differently.
diff --git a/llvm/test/Assembler/thinlto-vtable-summary.ll b/llvm/test/Assembler/thinlto-vtable-summary.ll
index 5a0ff32a83904..80720287f7a09 100644
--- a/llvm/test/Assembler/thinlto-vtable-summary.ll
+++ b/llvm/test/Assembler/thinlto-vtable-summary.ll
@@ -29,9 +29,9 @@ declare i32 @_ZN1C1fEi(ptr, i32)
 
 ^0 = module: (path: "<stdin>", hash: (0, 0, 0, 0, 0))
 ^1 = gv: (name: "_ZN1A1nEi") ; guid = 1621563287929432257
-^2 = gv: (name: "_ZTV1B", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0, vcall_visibility: 0), vTableFuncs: ((virtFunc: ^3, offset: 16), (virtFunc: ^1, offset: 24)), refs: (^3, ^1)))) ; guid = 5283576821522790367
+^2 = gv: (name: "_ZTV1B", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0, vcall_visibility: 0), vTableFuncs: ((virtFunc: ^3, offset: 16), (virtFunc: ^1, offset: 24)), refs: (^3, ^1)))) ; guid = 5283576821522790367
 ^3 = gv: (name: "_ZN1B1fEi") ; guid = 7162046368816414394
-^4 = gv: (name: "_ZTV1C", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 0, writeonly: 0, constant: 0, vcall_visibility: 0), vTableFuncs: ((virtFunc: ^5, offset: 16), (virtFunc: ^1, offset: 24)), refs: (^1, ^5)))) ; guid = 13624023785555846296
+^4 = gv: (name: "_ZTV1C", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 0, writeonly: 0, constant: 0, vcall_visibility: 0), vTableFuncs: ((virtFunc: ^5, offset: 16), (virtFunc: ^1, offset: 24)), refs: (^1, ^5)))) ; guid = 13624023785555846296
 ^5 = gv: (name: "_ZN1C1fEi") ; guid = 14876272565662207556
 ^6 = typeidCompatibleVTable: (name: "_ZTS1A", summary: ((offset: 16, ^2), (offset: 16, ^4))) ; guid = 7004155349499253778
 ^7 = typeidCompatibleVTable: (name: "_ZTS1B", summary: ((offset: 16, ^2))) ; guid = 6203814149063363976
diff --git a/llvm/test/Bitcode/thinlto-alias.ll b/llvm/test/Bitcode/thinlto-alias.ll
index eb794f4e631d1..5dfff0f796198 100644
--- a/llvm/test/Bitcode/thinlto-alias.ll
+++ b/llvm/test/Bitcode/thinlto-alias.ll
@@ -53,11 +53,11 @@ entry:
 declare void @analias(...)
 
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
-; DIS: ^1 = gv: (name: "analias", summaries: (alias: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), aliasee: ^2))) ; guid = 12695095382722328222
-; DIS: ^2 = gv: (name: "aliasee", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1))) ; guid = 17407585008595848568
+; DIS: ^1 = gv: (name: "analias", summaries: (alias: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), aliasee: ^2))) ; guid = 12695095382722328222
+; DIS: ^2 = gv: (name: "aliasee", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1))) ; guid = 17407585008595848568
 
 ; COMBINED-DIS: ^0 = module: (path: "{{.*}}thinlto-alias.ll.tmp.o", hash: (0, 0, 0, 0, 0))
 ; COMBINED-DIS: ^1 = module: (path: "{{.*}}thinlto-alias.ll.tmp2.o", hash: (0, 0, 0, 0, 0))
-; COMBINED-DIS: ^2 = gv: (guid: 12695095382722328222, summaries: (alias: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), aliasee: ^4)))
-; COMBINED-DIS: ^3 = gv: (guid: 15822663052811949562, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, calls: ((callee: ^2)))))
-; COMBINED-DIS: ^4 = gv: (guid: 17407585008595848568, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
+; COMBINED-DIS: ^2 = gv: (guid: 12695095382722328222, summaries: (alias: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), aliasee: ^4)))
+; COMBINED-DIS: ^3 = gv: (guid: 15822663052811949562, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^2)))))
+; COMBINED-DIS: ^4 = gv: (guid: 17407585008595848568, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
diff --git a/llvm/test/Bitcode/thinlto-func-summary-vtableref-pgo.ll b/llvm/test/Bitcode/thinlto-func-summary-vtableref-pgo.ll
index ba3ce9a75ee83..19e228fd5355c 100644
--- a/llvm/test/Bitcode/thinlto-func-summary-vtableref-pgo.ll
+++ b/llvm/test/Bitcode/thinlto-func-summary-vtableref-pgo.ll
@@ -70,5 +70,5 @@ define i32 @_Z4testP4Base(ptr %0) !prof !15 {
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
 ; DIS: ^1 = gv: (guid: 1960855528937986108)
 ; DIS: ^2 = gv: (guid: 5459407273543877811)
-; DIS: ^3 = gv: (name: "_Z4testP4Base", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), calls: ((callee: ^2, hotness: hot)), refs: (readonly ^1)))) ; guid = 15857150948103218965
+; DIS: ^3 = gv: (name: "_Z4testP4Base", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), calls: ((callee: ^2, hotness: hot)), refs: (readonly ^1)))) ; guid = 15857150948103218965
 ; DIS: ^4 = blockcount: 0
diff --git a/llvm/test/Bitcode/thinlto-function-summary-callgraph-profile-summary.ll b/llvm/test/Bitcode/thinlto-function-summary-callgraph-profile-summary.ll
index d7679b6f5af20..563fb18107d35 100644
--- a/llvm/test/Bitcode/thinlto-function-summary-callgraph-profile-summary.ll
+++ b/llvm/test/Bitcode/thinlto-function-summary-callgraph-profile-summary.ll
@@ -150,16 +150,16 @@ declare void @none3() #1
 ; DIS: ^6 = gv: (name: "cold") ; guid = 11668175513417606517
 ; DIS: ^7 = gv: (name: "hot4") ; guid = 13161834114071272798
 ; DIS: ^8 = gv: (name: "none3") ; guid = 16213681105727317812
-; DIS: ^9 = gv: (name: "hot_function", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 16, calls: ((callee: ^5, hotness: hot), (callee: ^6, hotness: cold), (callee: ^4, hotness: hot), (callee: ^7, hotness: cold), (callee: ^10, hotness: none), (callee: ^3, hotness: hot), (callee: ^2, hotness: none), (callee: ^8, hotness: none), (callee: ^1, hotness: critical))))) ; guid = 17381606045411660303
+; DIS: ^9 = gv: (name: "hot_function", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 16, calls: ((callee: ^5, hotness: hot), (callee: ^6, hotness: cold), (callee: ^4, hotness: hot), (callee: ^7, hotness: cold), (callee: ^10, hotness: none), (callee: ^3, hotness: hot), (callee: ^2, hotness: none), (callee: ^8, hotness: none), (callee: ^1, hotness: critical))))) ; guid = 17381606045411660303
 ; DIS: ^10 = gv: (name: "none1") ; guid = 17712061229457633252
 
 ; COMBINED-DIS: ^0 = module: (path: "{{.*}}thinlto-function-summary-callgraph-profile-summary.ll.tmp.o", hash: (0, 0, 0, 0, 0))
 ; COMBINED-DIS: ^1 = module: (path: "{{.*}}thinlto-function-summary-callgraph-profile-summary.ll.tmp2.o", hash: (0, 0, 0, 0, 0))
-; COMBINED-DIS: ^2 = gv: (guid: 3741006263754194003, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^3 = gv: (guid: 5026609803865204483, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^4 = gv: (guid: 8117347573235780485, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^5 = gv: (guid: 9453975128311291976, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^6 = gv: (guid: 11668175513417606517, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^7 = gv: (guid: 16213681105727317812, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; COMBINED-DIS: ^8 = gv: (guid: 17381606045411660303, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 16, calls: ((callee: ^5, hotness: hot), (callee: ^6, hotness: cold), (callee: ^4, hotness: hot), (callee: ^9, hotness: none), (callee: ^3, hotness: hot), (callee: ^2, hotness: none), (callee: ^7, hotness: none)))))
-; COMBINED-DIS: ^9 = gv: (guid: 17712061229457633252, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 1)))
+; COMBINED-DIS: ^2 = gv: (guid: 3741006263754194003, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^3 = gv: (guid: 5026609803865204483, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^4 = gv: (guid: 8117347573235780485, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^5 = gv: (guid: 9453975128311291976, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^6 = gv: (guid: 11668175513417606517, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^7 = gv: (guid: 16213681105727317812, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; COMBINED-DIS: ^8 = gv: (guid: 17381606045411660303, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 16, calls: ((callee: ^5, hotness: hot), (callee: ^6, hotness: cold), (callee: ^4, hotness: hot), (callee: ^9, hotness: none), (callee: ^3, hotness: hot), (callee: ^2, hotness: none), (callee: ^7, hotness: none)))))
+; COMBINED-DIS: ^9 = gv: (guid: 17712061229457633252, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
diff --git a/llvm/test/Bitcode/thinlto-function-summary-callgraph-relbf.ll b/llvm/test/Bitcode/thinlto-function-summary-callgraph-relbf.ll
index ca4f62907cb0a..3c827247a6c9f 100644
--- a/llvm/test/Bitcode/thinlto-function-summary-callgraph-relbf.ll
+++ b/llvm/test/Bitcode/thinlto-function-summary-callgraph-relbf.ll
@@ -39,5 +39,5 @@ declare void @func(...) #1
 
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
 ; DIS: ^1 = gv: (name: "func") ; guid = 7289175272376759421
-; DIS: ^2 = gv: (name: "main", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 3, calls: ((callee: ^1, relbf: 256)), refs: (readonly ^3)))) ; guid = 15822663052811949562
+; DIS: ^2 = gv: (name: "main", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 3, calls: ((callee: ^1, relbf: 256)), refs: (readonly ^3)))) ; guid = 15822663052811949562
 ; DIS: ^3 = gv: (name: "undefinedglob") ; guid = 18036901804029949403
diff --git a/llvm/test/Bitcode/thinlto-function-summary-refgraph.ll b/llvm/test/Bitcode/thinlto-function-summary-refgraph.ll
index fc42b5369644c..c76d70b8c4cc0 100644
--- a/llvm/test/Bitcode/thinlto-function-summary-refgraph.ll
+++ b/llvm/test/Bitcode/thinlto-function-summary-refgraph.ll
@@ -148,18 +148,18 @@ entry:
 ; order, which depends on GUID, and the private function Y GUID will depend
 ; on the path to the test.
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
-; DIS-DAG: = gv: (name: "Z", summaries: (function: (module: ^0, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, calls: ((callee: ^{{.*}}, tail: 1))))) ; guid = 104084381700047393
-; DIS-DAG: = gv: (name: "X", summaries: (function: (module: ^0, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, calls: ((callee: ^{{.*}})), refs: (^{{.*}})))) ; guid = 1881667236089500162
-; DIS-DAG: = gv: (name: "W", summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, calls: ((callee: ^{{.*}}, tail: 1)), refs: (^{{.*}})))) ; guid = 5790125716599269729
+; DIS-DAG: = gv: (name: "Z", summaries: (function: (module: ^0, flags: (linkage: linkonce_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^{{.*}}, tail: 1))))) ; guid = 104084381700047393
+; DIS-DAG: = gv: (name: "X", summaries: (function: (module: ^0, flags: (linkage: available_externally, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^{{.*}})), refs: (^{{.*}})))) ; guid = 1881667236089500162
+; DIS-DAG: = gv: (name: "W", summaries: (function: (module: ^0, flags: (linkage: weak_odr, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^{{.*}}, tail: 1)), refs: (^{{.*}})))) ; guid = 5790125716599269729
 ; DIS-DAG: = gv: (name: "foo") ; guid = 6699318081062747564
 ; DIS-DAG: = gv: (name: "func") ; guid = 7289175272376759421
 ; DIS-DAG: = gv: (name: "func3") ; guid = 11517462787082255043
 ; Check that default value of writeonly attribute is zero for constant variables 
-; DIS-DAG: = gv: (name: "globalvar", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 0, constant: 1)))) ; guid = 12887606300320728018
+; DIS-DAG: = gv: (name: "globalvar", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 1, writeonly: 0, constant: 1)))) ; guid = 12887606300320728018
 ; DIS-DAG: = gv: (name: "func2") ; guid = 14069196320850861797
 ; DIS-DAG: = gv: (name: "llvm.ctpop.i8") ; guid = 15254915475081819833
-; DIS-DAG: = gv: (name: "main", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 9, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}})))) ; guid = 15822663052811949562
-; DIS-DAG: = gv: (name: "bar", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1, constant: 0),  refs: (^{{.*}})))) ; guid = 16434608426314478903
+; DIS-DAG: = gv: (name: "main", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 9, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}})))) ; guid = 15822663052811949562
+; DIS-DAG: = gv: (name: "bar", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), varFlags: (readonly: 1, writeonly: 1, constant: 0),  refs: (^{{.*}})))) ; guid = 16434608426314478903
 ; Don't try to match the exact GUID. Since it is private, the file path
 ; will get hashed, and that will be test dependent.
-; DIS-DAG: = gv: (name: "Y", summaries: (function: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 14, calls: ((callee: ^{{.*}}, tail: 1))))) ; guid =
+; DIS-DAG: = gv: (name: "Y", summaries: (function: (module: ^0, flags: (linkage: private, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 14, calls: ((callee: ^{{.*}}, tail: 1))))) ; guid =
diff --git a/llvm/test/Bitcode/thinlto-index-disassembled-by-llvm-dis.ll b/llvm/test/Bitcode/thinlto-index-disassembled-by-llvm-dis.ll
index 0d6a8e3b4b8d5..3a121c2d5d426 100644
--- a/llvm/test/Bitcode/thinlto-index-disassembled-by-llvm-dis.ll
+++ b/llvm/test/Bitcode/thinlto-index-disassembled-by-llvm-dis.ll
@@ -18,7 +18,7 @@
 ; RUN: llvm-dis --print-thinlto-index-only %t.o -o - | FileCheck %s --check-prefix=DIS
 
 ; DIS: ^0 = module: (path: "{{.*}}thinlto-index-disassembled-by-llvm-dis.ll.tmp
-; DIS: ^1 = gv: (name: "aplusb", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2))) ; guid = 
+; DIS: ^1 = gv: (name: "aplusb", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2))) ; guid =
 
 source_filename = "add.cpp"
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
diff --git a/llvm/test/Bitcode/thinlto-type-tests.ll b/llvm/test/Bitcode/thinlto-type-tests.ll
index 76021a87c52c1..6dc49b849c2fe 100644
--- a/llvm/test/Bitcode/thinlto-type-tests.ll
+++ b/llvm/test/Bitcode/thinlto-type-tests.ll
@@ -37,11 +37,11 @@ declare i1 @llvm.type.test(i8*, metadata) nounwind readnone
 
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
 ; DIS: ^1 = gv: (name: "llvm.type.test") ; guid = 608142985856744218
-; DIS: ^2 = gv: (name: "h", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (16434608426314478903))))) ; guid = 8124147457056772133
-; DIS: ^3 = gv: (name: "g", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, typeIdInfo: (typeTests: (6699318081062747564, 16434608426314478903))))) ; guid = 13146401226427987378
-; DIS: ^4 = gv: (name: "f", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (6699318081062747564))))) ; guid = 14740650423002898831
+; DIS: ^2 = gv: (name: "h", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (16434608426314478903))))) ; guid = 8124147457056772133
+; DIS: ^3 = gv: (name: "g", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, typeIdInfo: (typeTests: (6699318081062747564, 16434608426314478903))))) ; guid = 13146401226427987378
+; DIS: ^4 = gv: (name: "f", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (6699318081062747564))))) ; guid = 14740650423002898831
 
 ; COMBINED-DIS: ^0 = module: (path: "{{.*}}thinlto-type-tests.ll.tmp.o", hash: (0, 0, 0, 0, 0))
-; COMBINED-DIS: ^1 = gv: (guid: 8124147457056772133, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (16434608426314478903)))))
-; COMBINED-DIS: ^2 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, typeIdInfo: (typeTests: (6699318081062747564, 16434608426314478903)))))
-; COMBINED-DIS: ^3 = gv: (guid: 14740650423002898831, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (6699318081062747564)))))
+; COMBINED-DIS: ^1 = gv: (guid: 8124147457056772133, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (16434608426314478903)))))
+; COMBINED-DIS: ^2 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, typeIdInfo: (typeTests: (6699318081062747564, 16434608426314478903)))))
+; COMBINED-DIS: ^3 = gv: (guid: 14740650423002898831, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (6699318081062747564)))))
diff --git a/llvm/test/Bitcode/thinlto-type-vcalls.ll b/llvm/test/Bitcode/thinlto-type-vcalls.ll
index ede5b483bfebf..16c93097101df 100644
--- a/llvm/test/Bitcode/thinlto-type-vcalls.ll
+++ b/llvm/test/Bitcode/thinlto-type-vcalls.ll
@@ -112,19 +112,19 @@ declare {i8*, i1} @llvm.type.checked.load(i8*, i32, metadata)
 
 ; DIS: ^0 = module: (path: "{{.*}}", hash: (0, 0, 0, 0, 0))
 ; DIS: ^1 = gv: (name: "llvm.type.test") ; guid = 608142985856744218
-; DIS: ^2 = gv: (name: "f1", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 8, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16)))))) ; guid = 2072045998141807037
-; DIS: ^3 = gv: (name: "f3", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16)))))) ; guid = 4197650231481825559
+; DIS: ^2 = gv: (name: "f1", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 8, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16)))))) ; guid = 2072045998141807037
+; DIS: ^3 = gv: (name: "f3", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16)))))) ; guid = 4197650231481825559
 ; DIS: ^4 = gv: (name: "llvm.type.checked.load") ; guid = 5568222536364573403
 ; DIS: ^5 = gv: (name: "llvm.assume") ; guid = 6385187066495850096
-; DIS: ^6 = gv: (name: "f2", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 24), vFuncId: (guid: 16434608426314478903, offset: 32)))))) ; guid = 8471399308421654326
-; DIS: ^7 = gv: (name: "f4", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)), (vFuncId: (guid: 6699318081062747564, offset: 24), args: (43))))))) ; guid = 10064745020953272174
-; DIS: ^8 = gv: (name: "f5", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42))))))) ; guid = 11686717102184386164
-; DIS: ^9 = gv: (name: "f6", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (7546896869197086323))))) ; guid = 11834966808443348068
+; DIS: ^6 = gv: (name: "f2", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 24), vFuncId: (guid: 16434608426314478903, offset: 32)))))) ; guid = 8471399308421654326
+; DIS: ^7 = gv: (name: "f4", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)), (vFuncId: (guid: 6699318081062747564, offset: 24), args: (43))))))) ; guid = 10064745020953272174
+; DIS: ^8 = gv: (name: "f5", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42))))))) ; guid = 11686717102184386164
+; DIS: ^9 = gv: (name: "f6", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (7546896869197086323))))) ; guid = 11834966808443348068
 
 ; COMBINED-DIS: ^0 = module: (path: "{{.*}}thinlto-type-vcalls.ll.tmp.o", hash: (0, 0, 0, 0, 0))
-; COMBINED-DIS: ^1 = gv: (guid: 2072045998141807037, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 8, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16))))))
-; COMBINED-DIS: ^2 = gv: (guid: 4197650231481825559, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16))))))
-; COMBINED-DIS: ^3 = gv: (guid: 8471399308421654326, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 24), vFuncId: (guid: 16434608426314478903, offset: 32))))))
-; COMBINED-DIS: ^4 = gv: (guid: 10064745020953272174, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)), (vFuncId: (guid: 6699318081062747564, offset: 24), args: (43)))))))
-; COMBINED-DIS: ^5 = gv: (guid: 11686717102184386164, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)))))))
-; COMBINED-DIS: ^6 = gv: (guid: 11834966808443348068, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 2, typeIdInfo: (typeTests: (7546896869197086323)))))
+; COMBINED-DIS: ^1 = gv: (guid: 2072045998141807037, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 8, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16))))))
+; COMBINED-DIS: ^2 = gv: (guid: 4197650231481825559, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadVCalls: (vFuncId: (guid: 6699318081062747564, offset: 16))))))
+; COMBINED-DIS: ^3 = gv: (guid: 8471399308421654326, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeVCalls: (vFuncId: (guid: 6699318081062747564, offset: 24), vFuncId: (guid: 16434608426314478903, offset: 32))))))
+; COMBINED-DIS: ^4 = gv: (guid: 10064745020953272174, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 15, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeTestAssumeConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)), (vFuncId: (guid: 6699318081062747564, offset: 24), args: (43)))))))
+; COMBINED-DIS: ^5 = gv: (guid: 11686717102184386164, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 0, hasUnknownCall: 1, mustBeUnreachable: 0), typeIdInfo: (typeCheckedLoadConstVCalls: ((vFuncId: (guid: 6699318081062747564, offset: 16), args: (42)))))))
+; COMBINED-DIS: ^6 = gv: (guid: 11834966808443348068, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, typeIdInfo: (typeTests: (7546896869197086323)))))
diff --git a/llvm/test/ThinLTO/X86/dot-dumper.ll b/llvm/test/ThinLTO/X86/dot-dumper.ll
index 53122160b1b25..149039203a68c 100644
--- a/llvm/test/ThinLTO/X86/dot-dumper.ll
+++ b/llvm/test/ThinLTO/X86/dot-dumper.ll
@@ -20,8 +20,8 @@
 ; PERMODULE-NEXT:    color = lightgrey;
 ; PERMODULE-NEXT:    label = "";
 ; PERMODULE-NEXT:    node [style=filled,fillcolor=lightblue];
-; PERMODULE-NEXT:    M0_[[MAIN_ALIAS:[0-9]+]] [style="dotted,filled",shape="box",label="main_alias",fillcolor="red"]; // alias, dead
-; PERMODULE-NEXT:    M0_[[MAIN:[0-9]+]] [shape="record",label="main|extern (inst: 4, ffl: 0000000000)}",fillcolor="red"]; // function, dead
+; PERMODULE-NEXT:    M0_[[MAIN_ALIAS:[0-9]+]] [style="dotted,filled",shape="box",label="main_alias",fillcolor="red"]; // alias, definition, dead
+; PERMODULE-NEXT:    M0_[[MAIN:[0-9]+]] [shape="record",label="main|extern (inst: 4, ffl: 0000000000)}",fillcolor="red"]; // function, definition, dead
 ; PERMODULE-NEXT:    // Edges:
 ; PERMODULE-NEXT:    M0_[[MAIN_ALIAS]] -> M0_[[MAIN]] [style=dotted]; // alias
 ; PERMODULE-NEXT:  }
@@ -39,8 +39,8 @@
 ; COMBINED-NEXT:    color = lightgrey;
 ; COMBINED-NEXT:    label = "dot-dumper{{.*}}1.bc";
 ; COMBINED-NEXT:    node [style=filled,fillcolor=lightblue];
-; COMBINED-NEXT:    M0_[[MAIN_ALIAS:[0-9]+]] [style="dotted,filled",shape="box",label="main_alias",fillcolor="red"]; // alias, dead
-; COMBINED-NEXT:    M0_[[MAIN:[0-9]+]] [shape="record",label="main|extern (inst: 4, ffl: 0000000000)}"]; // function, preserved
+; COMBINED-NEXT:    M0_[[MAIN_ALIAS:[0-9]+]] [style="dotted,filled",shape="box",label="main_alias",fillcolor="red"]; // alias, definition, dead
+; COMBINED-NEXT:    M0_[[MAIN:[0-9]+]] [shape="record",label="main|extern (inst: 4, ffl: 0000000000)}"]; // function, definition, preserved
 ; COMBINED-NEXT:    // Edges:
 ; COMBINED-NEXT:    M0_[[MAIN_ALIAS]] -> M0_[[MAIN]] [style=dotted]; // alias
 ; COMBINED-NEXT:  }
@@ -50,10 +50,10 @@
 ; COMBINED-NEXT:    color = lightgrey;
 ; COMBINED-NEXT:    label = "dot-dumper{{.*}}2.bc";
 ; COMBINED-NEXT:    node [style=filled,fillcolor=lightblue];
-; COMBINED-NEXT:    M1_[[FOO:[0-9]+]] [shape="record",label="foo|extern (inst: 4, ffl: 0000100000)}"]; // function
+; COMBINED-NEXT:    M1_[[FOO:[0-9]+]] [shape="record",label="foo|extern (inst: 4, ffl: 0000100000)}"]; // function, definition
 ; COMBINED-NEXT:    M1_[[A:[0-9]+]] [shape="Mrecord",label="A|extern}"]; // variable, immutable
 ; COMBINED-NEXT:    M1_[[B:[0-9]+]] [shape="Mrecord",label="B|extern}"]; // variable, immutable, constant
-; COMBINED-NEXT:    M1_{{[0-9]+}} [shape="record",label="bar|extern (inst: 1, ffl: 0000000000)}",fillcolor="red"]; // function, dead
+; COMBINED-NEXT:    M1_{{[0-9]+}} [shape="record",label="bar|extern (inst: 1, ffl: 0000000000)}",fillcolor="red"]; // function, definition, dead
 ; COMBINED-NEXT:    // Edges:
 ; COMBINED-NEXT:    M1_[[FOO]] -> M1_[[B]] [style=dashed,color=forestgreen]; // const-ref
 ; COMBINED-NEXT:    M1_[[FOO]] -> M1_[[A]] [style=dashed,color=forestgreen]; // const-ref
diff --git a/llvm/test/ThinLTO/X86/funcattrs-prop-maythrow.ll b/llvm/test/ThinLTO/X86/funcattrs-prop-maythrow.ll
index 6489e952251d0..abfe820075bb9 100644
--- a/llvm/test/ThinLTO/X86/funcattrs-prop-maythrow.ll
+++ b/llvm/test/ThinLTO/X86/funcattrs-prop-maythrow.ll
@@ -48,9 +48,9 @@ define void @caller_nounwind() {
 ; CHECK-DAG: attributes [[ATTR_NOUNWIND]] = { norecurse nounwind }
 ; CHECK-DAG: attributes [[ATTR_MAYTHROW]] = { norecurse }
 
-; SUMMARY-DAG: = gv: (name: "cleanupret", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
-; SUMMARY-DAG: = gv: (name: "resume", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
-; SUMMARY-DAG: = gv: (name: "catchret", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
+; SUMMARY-DAG: = gv: (name: "cleanupret", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
+; SUMMARY-DAG: = gv: (name: "resume", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 4, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
+; SUMMARY-DAG: = gv: (name: "catchret", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 5, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 0, noUnwind: 0, mayThrow: 1, hasUnknownCall: 0, mustBeUnreachable: 0), calls: ((callee: ^{{.*}})), refs: (^{{.*}}))))
 
 ;--- callees.ll
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
@@ -112,4 +112,4 @@ exit:
   ret void
 }
 
-attributes #0 = { nounwind }
\ No newline at end of file
+attributes #0 = { nounwind }
diff --git a/llvm/test/ThinLTO/X86/funcimport_alwaysinline.ll b/llvm/test/ThinLTO/X86/funcimport_alwaysinline.ll
index 90c708fd2d115..67acc2a2892db 100644
--- a/llvm/test/ThinLTO/X86/funcimport_alwaysinline.ll
+++ b/llvm/test/ThinLTO/X86/funcimport_alwaysinline.ll
@@ -23,4 +23,4 @@ entry:
 }
 
 attributes #0 = { alwaysinline nounwind uwtable }
-; CHECK2: ^2 = gv: (guid: {{.*}}, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 1, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
+; CHECK2: ^2 = gv: (guid: {{.*}}, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 1, funcFlags: (readNone: 0, readOnly: 0, noRecurse: 0, returnDoesNotAlias: 0, noInline: 0, alwaysInline: 1, noUnwind: 1, mayThrow: 0, hasUnknownCall: 0, mustBeUnreachable: 0))))
diff --git a/llvm/test/ThinLTO/X86/load-store-caching.ll b/llvm/test/ThinLTO/X86/load-store-caching.ll
index 4fb2d4693042e..b25308bf17616 100644
--- a/llvm/test/ThinLTO/X86/load-store-caching.ll
+++ b/llvm/test/ThinLTO/X86/load-store-caching.ll
@@ -22,5 +22,5 @@ entry:
 }
 
 ; CHECK:       ^0 = module:
-; CHECK-NEXT:  ^1 = gv: (name: "obj", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), varFlags: (readonly: 1, writeonly: 1, constant: 0)))) ; guid =
-; CHECK-NEXT:  ^2 = gv: (name: "foo", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0), insts: 3, refs: (^1)))) ; guid =
+; CHECK-NEXT:  ^1 = gv: (name: "obj", summaries: (variable: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), varFlags: (readonly: 1, writeonly: 1, constant: 0)))) ; guid =
+; CHECK-NEXT:  ^2 = gv: (name: "foo", summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 0, dsoLocal: 1, canAutoHide: 0, importType: definition), insts: 3, refs: (^1)))) ; guid =
diff --git a/llvm/test/Transforms/LowerTypeTests/import-unsat.ll b/llvm/test/Transforms/LowerTypeTests/import-unsat.ll
index 76afe68d21891..f766c2d324163 100644
--- a/llvm/test/Transforms/LowerTypeTests/import-unsat.ll
+++ b/llvm/test/Transforms/LowerTypeTests/import-unsat.ll
@@ -10,6 +10,7 @@
 ; SUMMARY-NEXT:      Live:                true
 ; SUMMARY-NEXT:      Local:               false
 ; SUMMARY-NEXT:      CanAutoHide:         false
+; SUMMARY-NEXT:      ImportType:          0
 ; SUMMARY-NEXT:      TypeTests: [ 123 ]
 ; SUMMARY-NEXT: TypeIdMap:
 ; SUMMARY-NEXT:   typeid1:
diff --git a/llvm/test/Transforms/WholeProgramDevirt/Inputs/import-indir.yaml b/llvm/test/Transforms/WholeProgramDevirt/Inputs/import-indir.yaml
index 30159c5012b08..22533ed636a50 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/Inputs/import-indir.yaml
+++ b/llvm/test/Transforms/WholeProgramDevirt/Inputs/import-indir.yaml
@@ -2,6 +2,7 @@
 GlobalValueMap:
   42:
     - Live: true
+      ImportType: 0
       TypeTestAssumeVCalls:
         - GUID: 123
           Offset: 0
@@ -22,6 +23,9 @@ GlobalValueMap:
             GUID: 456
             Offset: 8
           Args: [24, 12]
+  43:
+    - Live: true
+      ImportType : 1
 TypeIdMap:
   typeid1:
     WPDRes:
diff --git a/llvm/test/Transforms/WholeProgramDevirt/import-indir.ll b/llvm/test/Transforms/WholeProgramDevirt/import-indir.ll
index 1d74a59769787..e4d6f1d52b540 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/import-indir.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/import-indir.ll
@@ -10,6 +10,7 @@
 ; SUMMARY-NEXT:      Live:                true
 ; SUMMARY-NEXT:      Local:               false
 ; SUMMARY-NEXT:      CanAutoHide:         false
+; SUMMARY-NEXT:      ImportType:          0
 ; SUMMARY-NEXT:      TypeTestAssumeVCalls:
 ; SUMMARY-NEXT:        - GUID:            123
 ; SUMMARY-NEXT:          Offset:          0
@@ -30,6 +31,14 @@
 ; SUMMARY-NEXT:            GUID:            456
 ; SUMMARY-NEXT:            Offset:          8
 ; SUMMARY-NEXT:          Args: [ 24, 12 ]
+; SUMMARY-NEXT: 43:
+; SUMMARY-NEXT:   - Linkage:                0
+; SUMMARY-NEXT:     Visibility:             0
+; SUMMARY-NEXT:     NotEligibleToImport:   false
+; SUMMARY-NEXT:     Live:                  true
+; SUMMARY-NEXT:     Local:                 false
+; SUMMARY-NEXT:     CanAutoHide:           false
+; SUMMARY-NEXT:     ImportType:            1
 ; SUMMARY-NEXT: TypeIdMap:
 ; SUMMARY-NEXT:   typeid1:
 ; SUMMARY-NEXT:     TTRes:

From 75edf0c18c777d69df7cfc6462e5233649bd47d4 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Thu, 11 Apr 2024 11:05:55 +0800
Subject: [PATCH 092/886] [NFC] [Serialization] Avoid accessing PendingBodies
 as much as possible

The 'HaveBody' parameter in isConsumerInterestedIn is only used for the function decl if it
doesn't have a body already. It should be relatively less frequent than
the call to isConsumerInterestedIn. So we can delay the computing of
`HaveBdoy` to make it more efficient.
---
 clang/include/clang/Serialization/ASTReader.h |  1 +
 clang/lib/Serialization/ASTReaderDecl.cpp     | 14 ++++++--------
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/clang/include/clang/Serialization/ASTReader.h b/clang/include/clang/Serialization/ASTReader.h
index 5fd55a519c6b0..e8b9f28690d9f 100644
--- a/clang/include/clang/Serialization/ASTReader.h
+++ b/clang/include/clang/Serialization/ASTReader.h
@@ -1492,6 +1492,7 @@ class ASTReader
   getModuleFileLevelDecls(ModuleFile &Mod);
 
 private:
+  bool isConsumerInterestedIn(Decl *D);
   void PassInterestingDeclsToConsumer();
   void PassInterestingDeclToConsumer(Decl *D);
 
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index 78448855fba09..9e49a3780ff41 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -3198,7 +3198,7 @@ inline void ASTReader::LoadedDecl(unsigned Index, Decl *D) {
 /// This routine should return true for anything that might affect
 /// code generation, e.g., inline function definitions, Objective-C
 /// declarations with metadata, etc.
-static bool isConsumerInterestedIn(ASTContext &Ctx, Decl *D, bool HasBody) {
+bool ASTReader::isConsumerInterestedIn(Decl *D) {
   // An ObjCMethodDecl is never considered as "interesting" because its
   // implementation container always is.
 
@@ -3207,7 +3207,7 @@ static bool isConsumerInterestedIn(ASTContext &Ctx, Decl *D, bool HasBody) {
   if (isPartOfPerModuleInitializer(D)) {
     auto *M = D->getImportedOwningModule();
     if (M && M->Kind == Module::ModuleMapModule &&
-        Ctx.DeclMustBeEmitted(D))
+        getContext().DeclMustBeEmitted(D))
       return false;
   }
 
@@ -3222,7 +3222,7 @@ static bool isConsumerInterestedIn(ASTContext &Ctx, Decl *D, bool HasBody) {
            (Var->isThisDeclarationADefinition() == VarDecl::Definition ||
             OMPDeclareTargetDeclAttr::isDeclareTargetDeclaration(Var));
   if (const auto *Func = dyn_cast<FunctionDecl>(D))
-    return Func->doesThisDeclarationHaveABody() || HasBody;
+    return Func->doesThisDeclarationHaveABody() || PendingBodies.count(D);
 
   if (auto *ES = D->getASTContext().getExternalSource())
     if (ES->hasExternalDefinitions(D) == ExternalASTSource::EK_Never)
@@ -4173,7 +4173,7 @@ void ASTReader::PassInterestingDeclsToConsumer() {
   while (!PotentiallyInterestingDecls.empty()) {
     Decl *D = PotentiallyInterestingDecls.front();
     PotentiallyInterestingDecls.pop_front();
-    if (isConsumerInterestedIn(getContext(), D, PendingBodies.count(D)))
+    if (isConsumerInterestedIn(D))
       PassInterestingDeclToConsumer(D);
   }
 }
@@ -4197,8 +4197,7 @@ void ASTReader::loadDeclUpdateRecords(PendingUpdateRecord &Record) {
     // the declaration, then we know it was interesting and we skip the call
     // to isConsumerInterestedIn because it is unsafe to call in the
     // current ASTReader state.
-    bool WasInteresting =
-        Record.JustLoaded || isConsumerInterestedIn(getContext(), D, false);
+    bool WasInteresting = Record.JustLoaded || isConsumerInterestedIn(D);
     for (auto &FileAndOffset : UpdateOffsets) {
       ModuleFile *F = FileAndOffset.first;
       uint64_t Offset = FileAndOffset.second;
@@ -4230,8 +4229,7 @@ void ASTReader::loadDeclUpdateRecords(PendingUpdateRecord &Record) {
 
       // We might have made this declaration interesting. If so, remember that
       // we need to hand it off to the consumer.
-      if (!WasInteresting &&
-          isConsumerInterestedIn(getContext(), D, PendingBodies.count(D))) {
+      if (!WasInteresting && isConsumerInterestedIn(D)) {
         PotentiallyInterestingDecls.push_back(D);
         WasInteresting = true;
       }

From 026165fad70420d85defb5fc9109c138250058ee Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97@outlook.com>
Date: Thu, 11 Apr 2024 11:25:33 +0800
Subject: [PATCH 093/886] [Instrumentation] Support MachineFunction in
 ChangeReporter (#80946)

---
 .../llvm/Passes/StandardInstrumentations.h    | 13 ++-
 llvm/lib/Passes/StandardInstrumentations.cpp  | 82 ++++++++++++++++---
 .../DotCfg/print-changed-dot-cfg.mir          | 24 ++++++
 llvm/test/Other/change-printer.mir            | 21 +++++
 4 files changed, 128 insertions(+), 12 deletions(-)
 create mode 100644 llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
 create mode 100644 llvm/test/Other/change-printer.mir

diff --git a/llvm/include/llvm/Passes/StandardInstrumentations.h b/llvm/include/llvm/Passes/StandardInstrumentations.h
index 8c6a44876d545..b053e6307a653 100644
--- a/llvm/include/llvm/Passes/StandardInstrumentations.h
+++ b/llvm/include/llvm/Passes/StandardInstrumentations.h
@@ -18,6 +18,8 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/StringSet.h"
+#include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/OptBisect.h"
 #include "llvm/IR/PassTimingInfo.h"
@@ -33,6 +35,7 @@ namespace llvm {
 
 class Module;
 class Function;
+class MachineFunction;
 class PassInstrumentationCallbacks;
 
 /// Instrumentation to print IR before/after passes.
@@ -313,6 +316,11 @@ template <typename T> class BlockDataT {
     B.print(SS, nullptr, true, true);
   }
 
+  BlockDataT(const MachineBasicBlock &B) : Label(B.getName().str()), Data(B) {
+    raw_string_ostream SS(Body);
+    B.print(SS);
+  }
+
   bool operator==(const BlockDataT &That) const { return Body == That.Body; }
   bool operator!=(const BlockDataT &That) const { return Body != That.Body; }
 
@@ -364,6 +372,7 @@ template <typename T> class OrderedChangedData {
 class EmptyData {
 public:
   EmptyData(const BasicBlock &) {}
+  EmptyData(const MachineBasicBlock &) {}
 };
 
 // The data saved for comparing functions.
@@ -405,7 +414,8 @@ template <typename T> class IRComparer {
 
 protected:
   // Generate the data for \p F into \p Data.
-  static bool generateFunctionData(IRDataT<T> &Data, const Function &F);
+  template <typename FunctionT>
+  static bool generateFunctionData(IRDataT<T> &Data, const FunctionT &F);
 
   const IRDataT<T> &Before;
   const IRDataT<T> &After;
@@ -475,6 +485,7 @@ class DCData {
 public:
   // Fill the map with the transitions from basic block \p B.
   DCData(const BasicBlock &B);
+  DCData(const MachineBasicBlock &B);
 
   // Return an iterator to the names of the successor blocks.
   StringMap<std::string>::const_iterator begin() const {
diff --git a/llvm/lib/Passes/StandardInstrumentations.cpp b/llvm/lib/Passes/StandardInstrumentations.cpp
index 697988b3fc7c0..c5f0c14885d0c 100644
--- a/llvm/lib/Passes/StandardInstrumentations.cpp
+++ b/llvm/lib/Passes/StandardInstrumentations.cpp
@@ -19,7 +19,9 @@
 #include "llvm/Analysis/CallGraphSCCPass.h"
 #include "llvm/Analysis/LazyCallGraph.h"
 #include "llvm/Analysis/LoopInfo.h"
+#include "llvm/CodeGen/MIRPrinter.h"
 #include "llvm/CodeGen/MachineFunction.h"
+#include "llvm/CodeGen/MachineModuleInfo.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/Module.h"
@@ -180,6 +182,12 @@ const Module *unwrapModule(Any IR, bool Force = false) {
     return F->getParent();
   }
 
+  if (const auto *MF = unwrapIR<MachineFunction>(IR)) {
+    if (!Force && !isFunctionInPrintList(MF->getName()))
+      return nullptr;
+    return MF->getFunction().getParent();
+  }
+
   llvm_unreachable("Unknown IR unit");
 }
 
@@ -215,6 +223,12 @@ void printIR(raw_ostream &OS, const Loop *L) {
   printLoop(const_cast<Loop &>(*L), OS);
 }
 
+void printIR(raw_ostream &OS, const MachineFunction *MF) {
+  if (!isFunctionInPrintList(MF->getName()))
+    return;
+  MF->print(OS);
+}
+
 std::string getIRName(Any IR) {
   if (unwrapIR<Module>(IR))
     return "[module]";
@@ -262,6 +276,9 @@ bool shouldPrintIR(Any IR) {
 
   if (const auto *L = unwrapIR<Loop>(IR))
     return isFunctionInPrintList(L->getHeader()->getParent()->getName());
+
+  if (const auto *MF = unwrapIR<MachineFunction>(IR))
+    return isFunctionInPrintList(MF->getName());
   llvm_unreachable("Unknown wrapped IR type");
 }
 
@@ -275,6 +292,14 @@ void unwrapAndPrint(raw_ostream &OS, Any IR) {
     auto *M = unwrapModule(IR);
     assert(M && "should have unwrapped module");
     printIR(OS, M);
+
+    if (const auto *MF = unwrapIR<MachineFunction>(IR)) {
+      auto &MMI = MF->getMMI();
+      for (const auto &F : *M) {
+        if (auto *MF = MMI.getMachineFunction(F))
+          MF->print(OS);
+      }
+    }
     return;
   }
 
@@ -297,6 +322,11 @@ void unwrapAndPrint(raw_ostream &OS, Any IR) {
     printIR(OS, L);
     return;
   }
+
+  if (const auto *MF = unwrapIR<MachineFunction>(IR)) {
+    printIR(OS, MF);
+    return;
+  }
   llvm_unreachable("Unknown wrapped IR type");
 }
 
@@ -305,7 +335,8 @@ bool isIgnored(StringRef PassID) {
   return isSpecialPass(PassID,
                        {"PassManager", "PassAdaptor", "AnalysisManagerProxy",
                         "DevirtSCCRepeatedPass", "ModuleInlinerWrapperPass",
-                        "VerifierPass", "PrintModulePass"});
+                        "VerifierPass", "PrintModulePass", "PrintMIRPass",
+                        "PrintMIRPreparePass"});
 }
 
 std::string makeHTMLReady(StringRef SR) {
@@ -664,20 +695,38 @@ template <typename T> void IRComparer<T>::analyzeIR(Any IR, IRDataT<T> &Data) {
     return;
   }
 
-  const auto *F = unwrapIR<Function>(IR);
-  if (!F) {
-    const auto *L = unwrapIR<Loop>(IR);
-    assert(L && "Unknown IR unit.");
-    F = L->getHeader()->getParent();
+  if (const auto *F = unwrapIR<Function>(IR)) {
+    generateFunctionData(Data, *F);
+    return;
+  }
+
+  if (const auto *L = unwrapIR<Loop>(IR)) {
+    auto *F = L->getHeader()->getParent();
+    generateFunctionData(Data, *F);
+    return;
   }
-  assert(F && "Unknown IR unit.");
-  generateFunctionData(Data, *F);
+
+  if (const auto *MF = unwrapIR<MachineFunction>(IR)) {
+    generateFunctionData(Data, *MF);
+    return;
+  }
+
+  llvm_unreachable("Unknown IR unit");
+}
+
+static bool shouldGenerateData(const Function &F) {
+  return !F.isDeclaration() && isFunctionInPrintList(F.getName());
+}
+
+static bool shouldGenerateData(const MachineFunction &MF) {
+  return isFunctionInPrintList(MF.getName());
 }
 
 template <typename T>
-bool IRComparer<T>::generateFunctionData(IRDataT<T> &Data, const Function &F) {
-  if (!F.isDeclaration() && isFunctionInPrintList(F.getName())) {
-    FuncDataT<T> FD(F.getEntryBlock().getName().str());
+template <typename FunctionT>
+bool IRComparer<T>::generateFunctionData(IRDataT<T> &Data, const FunctionT &F) {
+  if (shouldGenerateData(F)) {
+    FuncDataT<T> FD(F.front().getName().str());
     int I = 0;
     for (const auto &B : F) {
       std::string BBName = B.getName().str();
@@ -722,6 +771,12 @@ static SmallString<32> getIRFileDisplayName(Any IR) {
     ResultStream << "-loop-";
     stable_hash LoopNameHash = stable_hash_combine_string(L->getName());
     write_hex(ResultStream, LoopNameHash, HexPrintStyle::Lower, MaxHashWidth);
+  } else if (const auto *MF = unwrapIR<MachineFunction>(IR)) {
+    ResultStream << "-machine-function-";
+    stable_hash MachineFunctionNameHash =
+        stable_hash_combine_string(MF->getName());
+    write_hex(ResultStream, MachineFunctionNameHash, HexPrintStyle::Lower,
+              MaxHashWidth);
   } else {
     llvm_unreachable("Unknown wrapped IR type");
   }
@@ -2122,6 +2177,11 @@ DCData::DCData(const BasicBlock &B) {
       addSuccessorLabel(Succ->getName().str(), "");
 }
 
+DCData::DCData(const MachineBasicBlock &B) {
+  for (const MachineBasicBlock *Succ : successors(&B))
+    addSuccessorLabel(Succ->getName().str(), "");
+}
+
 DotCfgChangeReporter::DotCfgChangeReporter(bool Verbose)
     : ChangeReporter<IRDataT<DCData>>(Verbose) {}
 
diff --git a/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir b/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
new file mode 100644
index 0000000000000..340ece93aa02b
--- /dev/null
+++ b/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
@@ -0,0 +1,24 @@
+# REQUIRES: x86-registered-target
+# Simple functionality check.
+# RUN: rm -rf %t && mkdir -p %t
+# RUN: llc -filetype=null -print-changed=dot-cfg -passes=no-op-machine-function -dot-cfg-dir=%t %s
+# RUN: ls %t/*.pdf %t/passes.html | count 3
+
+---
+name:            g
+body:             |
+  bb.0.entry:
+    %0:gr32 = MOV32ri 5
+    $eax = COPY %0
+    RET 0, $eax
+
+...
+---
+name:            f
+body:             |
+  bb.0.entry:
+    %0:gr32 = MOV32ri 7
+    $eax = COPY %0
+    RET 0, $eax
+
+...
diff --git a/llvm/test/Other/change-printer.mir b/llvm/test/Other/change-printer.mir
new file mode 100644
index 0000000000000..5e57da50e625d
--- /dev/null
+++ b/llvm/test/Other/change-printer.mir
@@ -0,0 +1,21 @@
+# REQUIRES: x86-registered-target
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -filetype=null %s \
+# RUN: -p no-op-machine-function -print-changed 2>&1 | FileCheck %s --check-prefix=CHECK-NO-OP
+
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -filetype=null %s \
+# RUN: -p dead-mi-elimination -print-changed 2>&1 | FileCheck %s --check-prefix=CHECK-SIMPLE
+
+---
+name:            test
+body:             |
+  bb.0:
+    %1:gr64 = MOV64ri 0
+    %2:gr64 = MOV64ri 0
+    $eax = COPY %1
+    RET64 implicit $eax
+...
+
+# CHECK-NO-OP: *** IR Dump After NoOpMachineFunctionPass on test omitted because no change ***
+
+# CHECK-SIMPLE: *** IR Dump After DeadMachineInstructionElimPass on test ***
+# CHECK-SIMPLE-NOT: %2:gr64 = MOV64ri 0

From 53003e36e9f4574d06c22611f61f68de32c89c6b Mon Sep 17 00:00:00 2001
From: Sacha Coppey <sacha.coppey@oracle.com>
Date: Thu, 11 Apr 2024 06:19:56 +0200
Subject: [PATCH 094/886] [RISCV] Implement Statepoint and Patchpoint lowering
 to call instructions (#77337)

This patch adds stackmap support for RISC-V with call targets.

Based on patch from https://reviews.llvm.org/D129848.
---
 llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp     |  57 ++++
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   |  11 +
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      |   7 +-
 llvm/test/CodeGen/RISCV/rv64-patchpoint.ll    |  46 ++-
 .../RISCV/rv64-statepoint-call-lowering-x1.ll |  16 ++
 .../RISCV/rv64-statepoint-call-lowering-x2.ll |  23 ++
 .../RISCV/rv64-statepoint-call-lowering.ll    | 262 ++++++++++++++++++
 7 files changed, 419 insertions(+), 3 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x1.ll
 create mode 100644 llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x2.ll
 create mode 100644 llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering.ll

diff --git a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
index 9982a73ee914d..779f179dff619 100644
--- a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
+++ b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
@@ -14,6 +14,7 @@
 #include "MCTargetDesc/RISCVBaseInfo.h"
 #include "MCTargetDesc/RISCVInstPrinter.h"
 #include "MCTargetDesc/RISCVMCExpr.h"
+#include "MCTargetDesc/RISCVMatInt.h"
 #include "MCTargetDesc/RISCVTargetStreamer.h"
 #include "RISCV.h"
 #include "RISCVMachineFunctionInfo.h"
@@ -153,8 +154,35 @@ void RISCVAsmPrinter::LowerPATCHPOINT(MCStreamer &OutStreamer, StackMaps &SM,
 
   PatchPointOpers Opers(&MI);
 
+  const MachineOperand &CalleeMO = Opers.getCallTarget();
   unsigned EncodedBytes = 0;
 
+  if (CalleeMO.isImm()) {
+    uint64_t CallTarget = CalleeMO.getImm();
+    if (CallTarget) {
+      assert((CallTarget & 0xFFFF'FFFF'FFFF) == CallTarget &&
+             "High 16 bits of call target should be zero.");
+      // Materialize the jump address:
+      SmallVector<MCInst, 8> Seq;
+      RISCVMatInt::generateMCInstSeq(CallTarget, *STI, RISCV::X1, Seq);
+      for (MCInst &Inst : Seq) {
+        bool Compressed = EmitToStreamer(OutStreamer, Inst);
+        EncodedBytes += Compressed ? 2 : 4;
+      }
+      bool Compressed = EmitToStreamer(OutStreamer, MCInstBuilder(RISCV::JALR)
+                                                        .addReg(RISCV::X1)
+                                                        .addReg(RISCV::X1)
+                                                        .addImm(0));
+      EncodedBytes += Compressed ? 2 : 4;
+    }
+  } else if (CalleeMO.isGlobal()) {
+    MCOperand CallTargetMCOp;
+    lowerOperand(CalleeMO, CallTargetMCOp);
+    EmitToStreamer(OutStreamer,
+                   MCInstBuilder(RISCV::PseudoCALL).addOperand(CallTargetMCOp));
+    EncodedBytes += 8;
+  }
+
   // Emit padding.
   unsigned NumBytes = Opers.getNumPatchBytes();
   assert(NumBytes >= EncodedBytes &&
@@ -173,6 +201,35 @@ void RISCVAsmPrinter::LowerSTATEPOINT(MCStreamer &OutStreamer, StackMaps &SM,
     assert(PatchBytes % NOPBytes == 0 &&
            "Invalid number of NOP bytes requested!");
     emitNops(PatchBytes / NOPBytes);
+  } else {
+    // Lower call target and choose correct opcode
+    const MachineOperand &CallTarget = SOpers.getCallTarget();
+    MCOperand CallTargetMCOp;
+    switch (CallTarget.getType()) {
+    case MachineOperand::MO_GlobalAddress:
+    case MachineOperand::MO_ExternalSymbol:
+      lowerOperand(CallTarget, CallTargetMCOp);
+      EmitToStreamer(
+          OutStreamer,
+          MCInstBuilder(RISCV::PseudoCALL).addOperand(CallTargetMCOp));
+      break;
+    case MachineOperand::MO_Immediate:
+      CallTargetMCOp = MCOperand::createImm(CallTarget.getImm());
+      EmitToStreamer(OutStreamer, MCInstBuilder(RISCV::JAL)
+                                      .addReg(RISCV::X1)
+                                      .addOperand(CallTargetMCOp));
+      break;
+    case MachineOperand::MO_Register:
+      CallTargetMCOp = MCOperand::createReg(CallTarget.getReg());
+      EmitToStreamer(OutStreamer, MCInstBuilder(RISCV::JALR)
+                                      .addReg(RISCV::X1)
+                                      .addOperand(CallTargetMCOp)
+                                      .addImm(0));
+      break;
+    default:
+      llvm_unreachable("Unsupported operand type in statepoint call target");
+      break;
+    }
   }
 
   auto &Ctx = OutStreamer.getContext();
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 357432081ddb0..3e7bc8c2367de 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -17910,6 +17910,17 @@ RISCVTargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
   case RISCV::PseudoFROUND_D_IN32X:
     return emitFROUND(MI, BB, Subtarget);
   case TargetOpcode::STATEPOINT:
+    // STATEPOINT is a pseudo instruction which has no implicit defs/uses
+    // while jal call instruction (where statepoint will be lowered at the end)
+    // has implicit def. This def is early-clobber as it will be set at
+    // the moment of the call and earlier than any use is read.
+    // Add this implicit dead def here as a workaround.
+    MI.addOperand(*MI.getMF(),
+                  MachineOperand::CreateReg(
+                      RISCV::X1, /*isDef*/ true,
+                      /*isImp*/ true, /*isKill*/ false, /*isDead*/ true,
+                      /*isUndef*/ false, /*isEarlyClobber*/ true));
+    [[fallthrough]];
   case TargetOpcode::STACKMAP:
   case TargetOpcode::PATCHPOINT:
     if (!Subtarget.is64Bit())
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 6b75efe684d91..84d754e3cbcf3 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -1469,9 +1469,12 @@ unsigned RISCVInstrInfo::getInstSizeInBytes(const MachineInstr &MI) const {
   case TargetOpcode::PATCHPOINT:
     // The size of the patchpoint intrinsic is the number of bytes requested
     return PatchPointOpers(&MI).getNumPatchBytes();
-  case TargetOpcode::STATEPOINT:
+  case TargetOpcode::STATEPOINT: {
     // The size of the statepoint intrinsic is the number of bytes requested
-    return StatepointOpers(&MI).getNumPatchBytes();
+    unsigned NumBytes = StatepointOpers(&MI).getNumPatchBytes();
+    // No patch bytes means at most a PseudoCall is emitted
+    return std::max(NumBytes, 8U);
+  }
   default:
     return get(Opcode).getSize();
   }
diff --git a/llvm/test/CodeGen/RISCV/rv64-patchpoint.ll b/llvm/test/CodeGen/RISCV/rv64-patchpoint.ll
index 51c2ae908e842..adf5f9863b79f 100644
--- a/llvm/test/CodeGen/RISCV/rv64-patchpoint.ll
+++ b/llvm/test/CodeGen/RISCV/rv64-patchpoint.ll
@@ -1,12 +1,56 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv64 -debug-entry-values -enable-misched=0 < %s | FileCheck %s
 
+; Trivial patchpoint codegen
+;
+define i64 @trivial_patchpoint_codegen(i64 %p1, i64 %p2, i64 %p3, i64 %p4) {
+; CHECK-LABEL: trivial_patchpoint_codegen:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd s0, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    sd s1, 0(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset s0, -8
+; CHECK-NEXT:    .cfi_offset s1, -16
+; CHECK-NEXT:    mv s0, a0
+; CHECK-NEXT:  .Ltmp0:
+; CHECK-NEXT:    lui ra, 3563
+; CHECK-NEXT:    addiw ra, ra, -577
+; CHECK-NEXT:    slli ra, ra, 12
+; CHECK-NEXT:    addi ra, ra, -259
+; CHECK-NEXT:    slli ra, ra, 12
+; CHECK-NEXT:    addi ra, ra, -1282
+; CHECK-NEXT:    jalr ra
+; CHECK-NEXT:    mv s1, a0
+; CHECK-NEXT:    mv a0, s0
+; CHECK-NEXT:    mv a1, s1
+; CHECK-NEXT:  .Ltmp1:
+; CHECK-NEXT:    lui ra, 3563
+; CHECK-NEXT:    addiw ra, ra, -577
+; CHECK-NEXT:    slli ra, ra, 12
+; CHECK-NEXT:    addi ra, ra, -259
+; CHECK-NEXT:    slli ra, ra, 12
+; CHECK-NEXT:    addi ra, ra, -1281
+; CHECK-NEXT:    jalr ra
+; CHECK-NEXT:    mv a0, s1
+; CHECK-NEXT:    ld s0, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    ld s1, 0(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+entry:
+  %resolveCall2 = inttoptr i64 244837814094590 to i8*
+  %result = tail call i64 (i64, i32, i8*, i32, ...) @llvm.experimental.patchpoint.i64(i64 2, i32 28, i8* %resolveCall2, i32 4, i64 %p1, i64 %p2, i64 %p3, i64 %p4)
+  %resolveCall3 = inttoptr i64 244837814094591 to i8*
+  tail call void (i64, i32, i8*, i32, ...) @llvm.experimental.patchpoint.void(i64 3, i32 28, i8* %resolveCall3, i32 2, i64 %p1, i64 %result)
+  ret i64 %result
+}
+
 ; Test small patchpoints that don't emit calls.
 define void @small_patchpoint_codegen(i64 %p1, i64 %p2, i64 %p3, i64 %p4) {
 ; CHECK-LABEL: small_patchpoint_codegen:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    .cfi_def_cfa_offset 0
-; CHECK-NEXT:  .Ltmp0:
+; CHECK-NEXT:  .Ltmp2:
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    nop
diff --git a/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x1.ll b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x1.ll
new file mode 100644
index 0000000000000..3ba49653cd01e
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x1.ll
@@ -0,0 +1,16 @@
+; RUN: llc -mtriple riscv64 -verify-machineinstrs -stop-after=prologepilog < %s | FileCheck %s
+
+; Check that STATEPOINT instruction has an early clobber implicit def for LR.
+target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "riscv64"
+
+define void @test() "frame-pointer"="all" gc "statepoint-example" {
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(void ()) @return_i1, i32 0, i32 0, i32 0, i32 0) ["gc-live" ()]
+; CHECK: STATEPOINT 0, 0, 0, target-flags(riscv-call) @return_i1, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, 2, 0, csr_ilp32_lp64, implicit-def $x2, implicit-def dead early-clobber $x1
+  ret void
+}
+
+
+declare void @return_i1()
+declare token @llvm.experimental.gc.statepoint.p0(i64, i32, ptr, i32, i32, ...)
diff --git a/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x2.ll b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x2.ll
new file mode 100644
index 0000000000000..9c99f64bcacc0
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering-x2.ll
@@ -0,0 +1,23 @@
+; RUN: llc -mtriple riscv64 -verify-machineinstrs -stop-after=prologepilog < %s | FileCheck %s
+
+; Check that STATEPOINT instruction prefer to use x2 in presense of x8.
+target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "riscv64"
+
+declare void @consume(ptr addrspace(1) %obj)
+
+define i1 @test(ptr addrspace(1) %a) "frame-pointer"="all" gc "statepoint-example" {
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(i1 ()) @return_i1, i32 0, i32 0, i32 0, i32 0) ["gc-live" (ptr addrspace(1) %a)]
+; CHECK: STATEPOINT 0, 0, 0, target-flags(riscv-call) @return_i1, 2, 0, 2, 0, 2, 0, 2, 1, 1, 8, $x8, -32, 2, 0, 2, 1, 0, 0
+  %call1 = call ptr addrspace(1) @llvm.experimental.gc.relocate.p1(token %safepoint_token,  i32 0, i32 0)
+  %call2 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  call void @consume(ptr addrspace(1) %call1)
+  ret i1 %call2
+}
+
+
+declare i1 @return_i1()
+declare token @llvm.experimental.gc.statepoint.p0(i64, i32, ptr, i32, i32, ...)
+declare ptr addrspace(1) @llvm.experimental.gc.relocate.p1(token, i32, i32)
+declare i1 @llvm.experimental.gc.result.i1(token)
diff --git a/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering.ll b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering.ll
new file mode 100644
index 0000000000000..2fa344d4d79a7
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rv64-statepoint-call-lowering.ll
@@ -0,0 +1,262 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs < %s | FileCheck %s
+; A collection of basic functionality tests for statepoint lowering - most
+; interesting cornercases are exercised through the x86 tests.
+
+target datalayout = "e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "riscv64"
+
+%struct = type { i64, i64 }
+
+declare zeroext i1 @return_i1()
+declare zeroext i32 @return_i32()
+declare ptr @return_i32ptr()
+declare float @return_float()
+declare %struct @return_struct()
+declare void @varargf(i32, ...)
+
+define i1 @test_i1_return() gc "statepoint-example" {
+; CHECK-LABEL: test_i1_return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    call return_i1
+; CHECK-NEXT:  .Ltmp0:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+; This is just checking that a i1 gets lowered normally when there's no extra
+; state arguments to the statepoint
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(i1 ()) @return_i1, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  ret i1 %call1
+}
+
+define i32 @test_i32_return() gc "statepoint-example" {
+; CHECK-LABEL: test_i32_return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    call return_i32
+; CHECK-NEXT:  .Ltmp1:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(i32 ()) @return_i32, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call zeroext i32 @llvm.experimental.gc.result.i32(token %safepoint_token)
+  ret i32 %call1
+}
+
+define ptr @test_i32ptr_return() gc "statepoint-example" {
+; CHECK-LABEL: test_i32ptr_return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    call return_i32ptr
+; CHECK-NEXT:  .Ltmp2:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(ptr ()) @return_i32ptr, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call ptr @llvm.experimental.gc.result.p0(token %safepoint_token)
+  ret ptr %call1
+}
+
+define float @test_float_return() gc "statepoint-example" {
+; CHECK-LABEL: test_float_return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    call return_float
+; CHECK-NEXT:  .Ltmp3:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(float ()) @return_float, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call float @llvm.experimental.gc.result.f32(token %safepoint_token)
+  ret float %call1
+}
+
+define %struct @test_struct_return() gc "statepoint-example" {
+; CHECK-LABEL: test_struct_return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    call return_struct
+; CHECK-NEXT:  .Ltmp4:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(%struct ()) @return_struct, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call %struct @llvm.experimental.gc.result.struct(token %safepoint_token)
+  ret %struct %call1
+}
+
+define i1 @test_relocate(ptr addrspace(1) %a) gc "statepoint-example" {
+; CHECK-LABEL: test_relocate:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    sd a0, 0(sp)
+; CHECK-NEXT:    call return_i1
+; CHECK-NEXT:  .Ltmp5:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+; Check that an ununsed relocate has no code-generation impact
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(i1 ()) @return_i1, i32 0, i32 0, i32 0, i32 0) ["gc-live" (ptr addrspace(1) %a)]
+  %call1 = call ptr addrspace(1) @llvm.experimental.gc.relocate.p1(token %safepoint_token,  i32 0, i32 0)
+  %call2 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  ret i1 %call2
+}
+
+define void @test_void_vararg() gc "statepoint-example" {
+; CHECK-LABEL: test_void_vararg:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    li a0, 42
+; CHECK-NEXT:    li a1, 43
+; CHECK-NEXT:    call varargf
+; CHECK-NEXT:  .Ltmp6:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+; Check a statepoint wrapping a *ptr returning vararg function works
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(void (i32, ...)) @varargf, i32 2, i32 0, i32 42, i32 43, i32 0, i32 0)
+  ;; if we try to use the result from a statepoint wrapping a
+  ;; non-void-returning varargf, we will experience a crash.
+  ret void
+}
+
+define i1 @test_i1_return_patchable() gc "statepoint-example" {
+; CHECK-LABEL: test_i1_return_patchable:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    nop
+; CHECK-NEXT:  .Ltmp7:
+; CHECK-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
+; A patchable variant of test_i1_return
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 4, ptr elementtype(i1 ()) null, i32 0, i32 0, i32 0, i32 0)
+  %call1 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  ret i1 %call1
+}
+
+declare void @consume(ptr addrspace(1) %obj)
+
+define i1 @test_cross_bb(ptr addrspace(1) %a, i1 %external_cond) gc "statepoint-example" {
+; CHECK-LABEL: test_cross_bb:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -32
+; CHECK-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    .cfi_offset s0, -16
+; CHECK-NEXT:    andi s0, a1, 1
+; CHECK-NEXT:    sd a0, 8(sp)
+; CHECK-NEXT:    call return_i1
+; CHECK-NEXT:  .Ltmp8:
+; CHECK-NEXT:    beqz s0, .LBB8_2
+; CHECK-NEXT:  # %bb.1: # %left
+; CHECK-NEXT:    ld a1, 8(sp)
+; CHECK-NEXT:    mv s0, a0
+; CHECK-NEXT:    mv a0, a1
+; CHECK-NEXT:    call consume
+; CHECK-NEXT:    mv a0, s0
+; CHECK-NEXT:    j .LBB8_3
+; CHECK-NEXT:  .LBB8_2: # %right
+; CHECK-NEXT:    li a0, 1
+; CHECK-NEXT:  .LBB8_3: # %right
+; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 32
+; CHECK-NEXT:    ret
+entry:
+  %safepoint_token = tail call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(i1 ()) @return_i1, i32 0, i32 0, i32 0, i32 0) ["gc-live" (ptr addrspace(1) %a)]
+  br i1 %external_cond, label %left, label %right
+
+left:
+  %call1 = call ptr addrspace(1) @llvm.experimental.gc.relocate.p1(token %safepoint_token,  i32 0, i32 0)
+  %call2 = call zeroext i1 @llvm.experimental.gc.result.i1(token %safepoint_token)
+  call void @consume(ptr addrspace(1) %call1)
+  ret i1 %call2
+
+right:
+  ret i1 true
+}
+
+%struct2 = type { i64, i64, i64 }
+
+declare void @consume_attributes(i32, ptr nest, i32, ptr byval(%struct2))
+
+define void @test_attributes(ptr byval(%struct2) %s) gc "statepoint-example" {
+; CHECK-LABEL: test_attributes:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -32
+; CHECK-NEXT:    .cfi_def_cfa_offset 32
+; CHECK-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    ld a1, 16(a0)
+; CHECK-NEXT:    sd a1, 16(sp)
+; CHECK-NEXT:    ld a1, 8(a0)
+; CHECK-NEXT:    sd a1, 8(sp)
+; CHECK-NEXT:    ld a0, 0(a0)
+; CHECK-NEXT:    sd a0, 0(sp)
+; CHECK-NEXT:    li a0, 42
+; CHECK-NEXT:    li a1, 17
+; CHECK-NEXT:    mv a2, sp
+; CHECK-NEXT:    li t2, 0
+; CHECK-NEXT:    call consume_attributes
+; CHECK-NEXT:  .Ltmp9:
+; CHECK-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 32
+; CHECK-NEXT:    ret
+entry:
+; Check that arguments with attributes are lowered correctly.
+; We call a function that has a nest argument and a byval argument.
+  %statepoint_token = call token (i64, i32, ptr, i32, i32, ...) @llvm.experimental.gc.statepoint.p0(i64 0, i32 0, ptr elementtype(void (i32, ptr, i32, ptr)) @consume_attributes, i32 4, i32 0, i32 42, ptr nest null, i32 17, ptr byval(%struct2) %s, i32 0, i32 0)
+  ret void
+}
+
+declare token @llvm.experimental.gc.statepoint.p0(i64, i32, ptr, i32, i32, ...)
+declare i1 @llvm.experimental.gc.result.i1(token)
+
+declare i32 @llvm.experimental.gc.result.i32(token)
+
+declare ptr @llvm.experimental.gc.result.p0(token)
+
+declare float @llvm.experimental.gc.result.f32(token)
+
+declare %struct @llvm.experimental.gc.result.struct(token)
+
+
+
+declare ptr addrspace(1) @llvm.experimental.gc.relocate.p1(token, i32, i32)

From b5640369337e98e573c949080ed4a4061ec6ec9a Mon Sep 17 00:00:00 2001
From: Pengcheng Wang <wangpengcheng.pp@bytedance.com>
Date: Thu, 11 Apr 2024 12:20:27 +0800
Subject: [PATCH 095/886] [MachineCombiner][NFC] Split target-dependent
 patterns

We split target-dependent MachineCombiner patterns into their target
folder.

This makes MachineCombiner much more target-independent.

Reviewers:
davemgreen, asavonic, rotateright, RKSimon, lukel97, LuoYuanke, topperc, mshockwave, asi-sc

Reviewed By: topperc, mshockwave

Pull Request: https://github.com/llvm/llvm-project/pull/87991
---
 .../llvm/CodeGen/MachineCombinerPattern.h     | 168 +----
 llvm/include/llvm/CodeGen/TargetInstrInfo.h   |  25 +-
 llvm/lib/CodeGen/MachineCombiner.cpp          |  72 +--
 llvm/lib/CodeGen/TargetInstrInfo.cpp          |  19 +-
 llvm/lib/Target/AArch64/AArch64InstrInfo.cpp  | 572 +++++++++---------
 llvm/lib/Target/AArch64/AArch64InstrInfo.h    | 152 ++++-
 llvm/lib/Target/PowerPC/PPCInstrInfo.cpp      |  77 ++-
 llvm/lib/Target/PowerPC/PPCInstrInfo.h        |  26 +-
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      |  83 +--
 llvm/lib/Target/RISCV/RISCVInstrInfo.h        |  23 +-
 llvm/lib/Target/X86/X86InstrInfo.cpp          |  12 +-
 llvm/lib/Target/X86/X86InstrInfo.h            |  15 +-
 12 files changed, 642 insertions(+), 602 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/MachineCombinerPattern.h b/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
index 41b73eaae0298..3428c4dde5c7f 100644
--- a/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
+++ b/llvm/include/llvm/CodeGen/MachineCombinerPattern.h
@@ -16,8 +16,16 @@
 
 namespace llvm {
 
+/// The combiner's goal may differ based on which pattern it is attempting
+/// to optimize.
+enum class CombinerObjective {
+  MustReduceDepth,            // The data dependency chain must be improved.
+  MustReduceRegisterPressure, // The register pressure must be reduced.
+  Default                     // The critical path must not be lengthened.
+};
+
 /// These are instruction patterns matched by the machine combiner pass.
-enum class MachineCombinerPattern {
+enum MachineCombinerPattern : unsigned {
   // These are commutative variants for reassociating a computation chain. See
   // the comments before getMachineCombinerPatterns() in TargetInstrInfo.cpp.
   REASSOC_AX_BY,
@@ -25,163 +33,7 @@ enum class MachineCombinerPattern {
   REASSOC_XA_BY,
   REASSOC_XA_YB,
 
-  // These are patterns matched by the PowerPC to reassociate FMA chains.
-  REASSOC_XY_AMM_BMM,
-  REASSOC_XMM_AMM_BMM,
-
-  // These are patterns matched by the PowerPC to reassociate FMA and FSUB to
-  // reduce register pressure.
-  REASSOC_XY_BCA,
-  REASSOC_XY_BAC,
-
-  // These are patterns used to reduce the length of dependence chain.
-  SUBADD_OP1,
-  SUBADD_OP2,
-
-  // These are multiply-add patterns matched by the AArch64 machine combiner.
-  MULADDW_OP1,
-  MULADDW_OP2,
-  MULSUBW_OP1,
-  MULSUBW_OP2,
-  MULADDWI_OP1,
-  MULSUBWI_OP1,
-  MULADDX_OP1,
-  MULADDX_OP2,
-  MULSUBX_OP1,
-  MULSUBX_OP2,
-  MULADDXI_OP1,
-  MULSUBXI_OP1,
-  // NEON integers vectors
-  MULADDv8i8_OP1,
-  MULADDv8i8_OP2,
-  MULADDv16i8_OP1,
-  MULADDv16i8_OP2,
-  MULADDv4i16_OP1,
-  MULADDv4i16_OP2,
-  MULADDv8i16_OP1,
-  MULADDv8i16_OP2,
-  MULADDv2i32_OP1,
-  MULADDv2i32_OP2,
-  MULADDv4i32_OP1,
-  MULADDv4i32_OP2,
-
-  MULSUBv8i8_OP1,
-  MULSUBv8i8_OP2,
-  MULSUBv16i8_OP1,
-  MULSUBv16i8_OP2,
-  MULSUBv4i16_OP1,
-  MULSUBv4i16_OP2,
-  MULSUBv8i16_OP1,
-  MULSUBv8i16_OP2,
-  MULSUBv2i32_OP1,
-  MULSUBv2i32_OP2,
-  MULSUBv4i32_OP1,
-  MULSUBv4i32_OP2,
-
-  MULADDv4i16_indexed_OP1,
-  MULADDv4i16_indexed_OP2,
-  MULADDv8i16_indexed_OP1,
-  MULADDv8i16_indexed_OP2,
-  MULADDv2i32_indexed_OP1,
-  MULADDv2i32_indexed_OP2,
-  MULADDv4i32_indexed_OP1,
-  MULADDv4i32_indexed_OP2,
-
-  MULSUBv4i16_indexed_OP1,
-  MULSUBv4i16_indexed_OP2,
-  MULSUBv8i16_indexed_OP1,
-  MULSUBv8i16_indexed_OP2,
-  MULSUBv2i32_indexed_OP1,
-  MULSUBv2i32_indexed_OP2,
-  MULSUBv4i32_indexed_OP1,
-  MULSUBv4i32_indexed_OP2,
-
-  // Floating Point
-  FMULADDH_OP1,
-  FMULADDH_OP2,
-  FMULSUBH_OP1,
-  FMULSUBH_OP2,
-  FMULADDS_OP1,
-  FMULADDS_OP2,
-  FMULSUBS_OP1,
-  FMULSUBS_OP2,
-  FMULADDD_OP1,
-  FMULADDD_OP2,
-  FMULSUBD_OP1,
-  FMULSUBD_OP2,
-  FNMULSUBH_OP1,
-  FNMULSUBS_OP1,
-  FNMULSUBD_OP1,
-  FMLAv1i32_indexed_OP1,
-  FMLAv1i32_indexed_OP2,
-  FMLAv1i64_indexed_OP1,
-  FMLAv1i64_indexed_OP2,
-  FMLAv4f16_OP1,
-  FMLAv4f16_OP2,
-  FMLAv8f16_OP1,
-  FMLAv8f16_OP2,
-  FMLAv2f32_OP2,
-  FMLAv2f32_OP1,
-  FMLAv2f64_OP1,
-  FMLAv2f64_OP2,
-  FMLAv4i16_indexed_OP1,
-  FMLAv4i16_indexed_OP2,
-  FMLAv8i16_indexed_OP1,
-  FMLAv8i16_indexed_OP2,
-  FMLAv2i32_indexed_OP1,
-  FMLAv2i32_indexed_OP2,
-  FMLAv2i64_indexed_OP1,
-  FMLAv2i64_indexed_OP2,
-  FMLAv4f32_OP1,
-  FMLAv4f32_OP2,
-  FMLAv4i32_indexed_OP1,
-  FMLAv4i32_indexed_OP2,
-  FMLSv1i32_indexed_OP2,
-  FMLSv1i64_indexed_OP2,
-  FMLSv4f16_OP1,
-  FMLSv4f16_OP2,
-  FMLSv8f16_OP1,
-  FMLSv8f16_OP2,
-  FMLSv2f32_OP1,
-  FMLSv2f32_OP2,
-  FMLSv2f64_OP1,
-  FMLSv2f64_OP2,
-  FMLSv4i16_indexed_OP1,
-  FMLSv4i16_indexed_OP2,
-  FMLSv8i16_indexed_OP1,
-  FMLSv8i16_indexed_OP2,
-  FMLSv2i32_indexed_OP1,
-  FMLSv2i32_indexed_OP2,
-  FMLSv2i64_indexed_OP1,
-  FMLSv2i64_indexed_OP2,
-  FMLSv4f32_OP1,
-  FMLSv4f32_OP2,
-  FMLSv4i32_indexed_OP1,
-  FMLSv4i32_indexed_OP2,
-
-  FMULv2i32_indexed_OP1,
-  FMULv2i32_indexed_OP2,
-  FMULv2i64_indexed_OP1,
-  FMULv2i64_indexed_OP2,
-  FMULv4i16_indexed_OP1,
-  FMULv4i16_indexed_OP2,
-  FMULv4i32_indexed_OP1,
-  FMULv4i32_indexed_OP2,
-  FMULv8i16_indexed_OP1,
-  FMULv8i16_indexed_OP2,
-
-  // RISCV FMADD, FMSUB, FNMSUB patterns
-  FMADD_AX,
-  FMADD_XA,
-  FMSUB,
-  FNMSUB,
-  SHXADD_ADD_SLLI_OP1,
-  SHXADD_ADD_SLLI_OP2,
-
-  // X86 VNNI
-  DPWSSD,
-
-  FNMADD,
+  TARGET_PATTERN_START
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/CodeGen/TargetInstrInfo.h b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
index 9fd0ebe6956fb..d4a83e3753d98 100644
--- a/llvm/include/llvm/CodeGen/TargetInstrInfo.h
+++ b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
@@ -19,6 +19,7 @@
 #include "llvm/ADT/Uniformity.h"
 #include "llvm/CodeGen/MIRFormatter.h"
 #include "llvm/CodeGen/MachineBasicBlock.h"
+#include "llvm/CodeGen/MachineCombinerPattern.h"
 #include "llvm/CodeGen/MachineCycleAnalysis.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineInstr.h"
@@ -61,7 +62,6 @@ class TargetRegisterClass;
 class TargetRegisterInfo;
 class TargetSchedModel;
 class TargetSubtargetInfo;
-enum class MachineCombinerPattern;
 enum class MachineTraceStrategy;
 
 template <class T> class SmallVectorImpl;
@@ -1191,10 +1191,9 @@ class TargetInstrInfo : public MCInstrInfo {
   /// faster sequence.
   /// \param Root - Instruction that could be combined with one of its operands
   /// \param Patterns - Vector of possible combination patterns
-  virtual bool
-  getMachineCombinerPatterns(MachineInstr &Root,
-                             SmallVectorImpl<MachineCombinerPattern> &Patterns,
-                             bool DoRegPressureReduce) const;
+  virtual bool getMachineCombinerPatterns(MachineInstr &Root,
+                                          SmallVectorImpl<unsigned> &Patterns,
+                                          bool DoRegPressureReduce) const;
 
   /// Return true if target supports reassociation of instructions in machine
   /// combiner pass to reduce register pressure for a given BB.
@@ -1206,13 +1205,17 @@ class TargetInstrInfo : public MCInstrInfo {
 
   /// Fix up the placeholder we may add in genAlternativeCodeSequence().
   virtual void
-  finalizeInsInstrs(MachineInstr &Root, MachineCombinerPattern &P,
+  finalizeInsInstrs(MachineInstr &Root, unsigned &Pattern,
                     SmallVectorImpl<MachineInstr *> &InsInstrs) const {}
 
   /// Return true when a code sequence can improve throughput. It
   /// should be called only for instructions in loops.
   /// \param Pattern - combiner pattern
-  virtual bool isThroughputPattern(MachineCombinerPattern Pattern) const;
+  virtual bool isThroughputPattern(unsigned Pattern) const;
+
+  /// Return the objective of a combiner pattern.
+  /// \param Pattern - combiner pattern
+  virtual CombinerObjective getCombinerObjective(unsigned Pattern) const;
 
   /// Return true if the input \P Inst is part of a chain of dependent ops
   /// that are suitable for reassociation, otherwise return false.
@@ -1256,7 +1259,7 @@ class TargetInstrInfo : public MCInstrInfo {
   /// \param InstIdxForVirtReg - map of virtual register to instruction in
   /// InsInstr that defines it
   virtual void genAlternativeCodeSequence(
-      MachineInstr &Root, MachineCombinerPattern Pattern,
+      MachineInstr &Root, unsigned Pattern,
       SmallVectorImpl<MachineInstr *> &InsInstrs,
       SmallVectorImpl<MachineInstr *> &DelInstrs,
       DenseMap<unsigned, unsigned> &InstIdxForVirtReg) const;
@@ -1270,8 +1273,7 @@ class TargetInstrInfo : public MCInstrInfo {
 
   /// Attempt to reassociate \P Root and \P Prev according to \P Pattern to
   /// reduce critical path length.
-  void reassociateOps(MachineInstr &Root, MachineInstr &Prev,
-                      MachineCombinerPattern Pattern,
+  void reassociateOps(MachineInstr &Root, MachineInstr &Prev, unsigned Pattern,
                       SmallVectorImpl<MachineInstr *> &InsInstrs,
                       SmallVectorImpl<MachineInstr *> &DelInstrs,
                       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const;
@@ -1281,8 +1283,7 @@ class TargetInstrInfo : public MCInstrInfo {
   /// (new root opcode, new prev opcode) that must be used to reassociate \P
   /// Root and \P Prev accoring to \P Pattern.
   std::pair<unsigned, unsigned>
-  getReassociationOpcodes(MachineCombinerPattern Pattern,
-                          const MachineInstr &Root,
+  getReassociationOpcodes(unsigned Pattern, const MachineInstr &Root,
                           const MachineInstr &Prev) const;
 
   /// The limit on resource length extension we accept in MachineCombiner Pass.
diff --git a/llvm/lib/CodeGen/MachineCombiner.cpp b/llvm/lib/CodeGen/MachineCombiner.cpp
index a4c87a7678bd8..ac58162bbfb42 100644
--- a/llvm/lib/CodeGen/MachineCombiner.cpp
+++ b/llvm/lib/CodeGen/MachineCombiner.cpp
@@ -99,17 +99,16 @@ class MachineCombiner : public MachineFunctionPass {
                     const MachineBasicBlock &MBB);
   unsigned getLatency(MachineInstr *Root, MachineInstr *NewRoot,
                       MachineTraceMetrics::Trace BlockTrace);
-  bool
-  improvesCriticalPathLen(MachineBasicBlock *MBB, MachineInstr *Root,
-                          MachineTraceMetrics::Trace BlockTrace,
-                          SmallVectorImpl<MachineInstr *> &InsInstrs,
-                          SmallVectorImpl<MachineInstr *> &DelInstrs,
-                          DenseMap<unsigned, unsigned> &InstrIdxForVirtReg,
-                          MachineCombinerPattern Pattern, bool SlackIsAccurate);
+  bool improvesCriticalPathLen(MachineBasicBlock *MBB, MachineInstr *Root,
+                               MachineTraceMetrics::Trace BlockTrace,
+                               SmallVectorImpl<MachineInstr *> &InsInstrs,
+                               SmallVectorImpl<MachineInstr *> &DelInstrs,
+                               DenseMap<unsigned, unsigned> &InstrIdxForVirtReg,
+                               unsigned Pattern, bool SlackIsAccurate);
   bool reduceRegisterPressure(MachineInstr &Root, MachineBasicBlock *MBB,
                               SmallVectorImpl<MachineInstr *> &InsInstrs,
                               SmallVectorImpl<MachineInstr *> &DelInstrs,
-                              MachineCombinerPattern Pattern);
+                              unsigned Pattern);
   bool preservesResourceLen(MachineBasicBlock *MBB,
                             MachineTraceMetrics::Trace BlockTrace,
                             SmallVectorImpl<MachineInstr *> &InsInstrs,
@@ -123,7 +122,8 @@ class MachineCombiner : public MachineFunctionPass {
                                 MachineTraceMetrics::Trace BlockTrace);
 
   void verifyPatternOrder(MachineBasicBlock *MBB, MachineInstr &Root,
-                          SmallVector<MachineCombinerPattern, 16> &Patterns);
+                          SmallVector<unsigned, 16> &Patterns);
+  CombinerObjective getCombinerObjective(unsigned Pattern);
 };
 }
 
@@ -290,36 +290,17 @@ unsigned MachineCombiner::getLatency(MachineInstr *Root, MachineInstr *NewRoot,
   return NewRootLatency;
 }
 
-/// The combiner's goal may differ based on which pattern it is attempting
-/// to optimize.
-enum class CombinerObjective {
-  MustReduceDepth,            // The data dependency chain must be improved.
-  MustReduceRegisterPressure, // The register pressure must be reduced.
-  Default                     // The critical path must not be lengthened.
-};
-
-static CombinerObjective getCombinerObjective(MachineCombinerPattern P) {
+CombinerObjective MachineCombiner::getCombinerObjective(unsigned Pattern) {
   // TODO: If C++ ever gets a real enum class, make this part of the
   // MachineCombinerPattern class.
-  switch (P) {
+  switch (Pattern) {
   case MachineCombinerPattern::REASSOC_AX_BY:
   case MachineCombinerPattern::REASSOC_AX_YB:
   case MachineCombinerPattern::REASSOC_XA_BY:
   case MachineCombinerPattern::REASSOC_XA_YB:
-  case MachineCombinerPattern::REASSOC_XY_AMM_BMM:
-  case MachineCombinerPattern::REASSOC_XMM_AMM_BMM:
-  case MachineCombinerPattern::SUBADD_OP1:
-  case MachineCombinerPattern::SUBADD_OP2:
-  case MachineCombinerPattern::FMADD_AX:
-  case MachineCombinerPattern::FMADD_XA:
-  case MachineCombinerPattern::FMSUB:
-  case MachineCombinerPattern::FNMSUB:
     return CombinerObjective::MustReduceDepth;
-  case MachineCombinerPattern::REASSOC_XY_BCA:
-  case MachineCombinerPattern::REASSOC_XY_BAC:
-    return CombinerObjective::MustReduceRegisterPressure;
   default:
-    return CombinerObjective::Default;
+    return TII->getCombinerObjective(Pattern);
   }
 }
 
@@ -349,8 +330,7 @@ std::pair<unsigned, unsigned> MachineCombiner::getLatenciesForInstrSequences(
 bool MachineCombiner::reduceRegisterPressure(
     MachineInstr &Root, MachineBasicBlock *MBB,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
-    SmallVectorImpl<MachineInstr *> &DelInstrs,
-    MachineCombinerPattern Pattern) {
+    SmallVectorImpl<MachineInstr *> &DelInstrs, unsigned Pattern) {
   // FIXME: for now, we don't do any check for the register pressure patterns.
   // We treat them as always profitable. But we can do better if we make
   // RegPressureTracker class be aware of TIE attribute. Then we can get an
@@ -368,8 +348,7 @@ bool MachineCombiner::improvesCriticalPathLen(
     MachineTraceMetrics::Trace BlockTrace,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
-    DenseMap<unsigned, unsigned> &InstrIdxForVirtReg,
-    MachineCombinerPattern Pattern,
+    DenseMap<unsigned, unsigned> &InstrIdxForVirtReg, unsigned Pattern,
     bool SlackIsAccurate) {
   // Get depth and latency of NewRoot and Root.
   unsigned NewRootDepth =
@@ -493,13 +472,14 @@ bool MachineCombiner::preservesResourceLen(
 /// \param Pattern is used to call target hook finalizeInsInstrs
 /// \param IncrementalUpdate if true, compute instruction depths incrementally,
 ///                          otherwise invalidate the trace
-static void insertDeleteInstructions(
-    MachineBasicBlock *MBB, MachineInstr &MI,
-    SmallVectorImpl<MachineInstr *> &InsInstrs,
-    SmallVectorImpl<MachineInstr *> &DelInstrs,
-    MachineTraceMetrics::Ensemble *TraceEnsemble,
-    SparseSet<LiveRegUnit> &RegUnits, const TargetInstrInfo *TII,
-    MachineCombinerPattern Pattern, bool IncrementalUpdate) {
+static void
+insertDeleteInstructions(MachineBasicBlock *MBB, MachineInstr &MI,
+                         SmallVectorImpl<MachineInstr *> &InsInstrs,
+                         SmallVectorImpl<MachineInstr *> &DelInstrs,
+                         MachineTraceMetrics::Ensemble *TraceEnsemble,
+                         SparseSet<LiveRegUnit> &RegUnits,
+                         const TargetInstrInfo *TII, unsigned Pattern,
+                         bool IncrementalUpdate) {
   // If we want to fix up some placeholder for some target, do it now.
   // We need this because in genAlternativeCodeSequence, we have not decided the
   // better pattern InsInstrs or DelInstrs, so we don't want generate some
@@ -534,9 +514,9 @@ static void insertDeleteInstructions(
 
 // Check that the difference between original and new latency is decreasing for
 // later patterns. This helps to discover sub-optimal pattern orderings.
-void MachineCombiner::verifyPatternOrder(
-    MachineBasicBlock *MBB, MachineInstr &Root,
-    SmallVector<MachineCombinerPattern, 16> &Patterns) {
+void MachineCombiner::verifyPatternOrder(MachineBasicBlock *MBB,
+                                         MachineInstr &Root,
+                                         SmallVector<unsigned, 16> &Patterns) {
   long PrevLatencyDiff = std::numeric_limits<long>::max();
   (void)PrevLatencyDiff; // Variable is used in assert only.
   for (auto P : Patterns) {
@@ -590,7 +570,7 @@ bool MachineCombiner::combineInstructions(MachineBasicBlock *MBB) {
 
   while (BlockIter != MBB->end()) {
     auto &MI = *BlockIter++;
-    SmallVector<MachineCombinerPattern, 16> Patterns;
+    SmallVector<unsigned, 16> Patterns;
     // The motivating example is:
     //
     //     MUL  Other        MUL_op1 MUL_op2  Other
diff --git a/llvm/lib/CodeGen/TargetInstrInfo.cpp b/llvm/lib/CodeGen/TargetInstrInfo.cpp
index 9fbd516acea8e..7d77e5d1a1ff0 100644
--- a/llvm/lib/CodeGen/TargetInstrInfo.cpp
+++ b/llvm/lib/CodeGen/TargetInstrInfo.cpp
@@ -919,7 +919,7 @@ bool TargetInstrInfo::isReassociationCandidate(const MachineInstr &Inst,
 //    instruction is known to not increase the critical path, then don't match
 //    that pattern.
 bool TargetInstrInfo::getMachineCombinerPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
+    MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
     bool DoRegPressureReduce) const {
   bool Commute;
   if (isReassociationCandidate(Root, Commute)) {
@@ -941,13 +941,17 @@ bool TargetInstrInfo::getMachineCombinerPatterns(
 }
 
 /// Return true when a code sequence can improve loop throughput.
-bool
-TargetInstrInfo::isThroughputPattern(MachineCombinerPattern Pattern) const {
+bool TargetInstrInfo::isThroughputPattern(unsigned Pattern) const {
   return false;
 }
 
+CombinerObjective
+TargetInstrInfo::getCombinerObjective(unsigned Pattern) const {
+  return CombinerObjective::Default;
+}
+
 std::pair<unsigned, unsigned>
-TargetInstrInfo::getReassociationOpcodes(MachineCombinerPattern Pattern,
+TargetInstrInfo::getReassociationOpcodes(unsigned Pattern,
                                          const MachineInstr &Root,
                                          const MachineInstr &Prev) const {
   bool AssocCommutRoot = isAssociativeAndCommutative(Root);
@@ -1036,7 +1040,7 @@ TargetInstrInfo::getReassociationOpcodes(MachineCombinerPattern Pattern,
 // Return a pair of boolean flags showing if the new root and new prev operands
 // must be swapped. See visual example of the rule in
 // TargetInstrInfo::getReassociationOpcodes.
-static std::pair<bool, bool> mustSwapOperands(MachineCombinerPattern Pattern) {
+static std::pair<bool, bool> mustSwapOperands(unsigned Pattern) {
   switch (Pattern) {
   default:
     llvm_unreachable("Unexpected pattern");
@@ -1054,8 +1058,7 @@ static std::pair<bool, bool> mustSwapOperands(MachineCombinerPattern Pattern) {
 /// Attempt the reassociation transformation to reduce critical path length.
 /// See the above comments before getMachineCombinerPatterns().
 void TargetInstrInfo::reassociateOps(
-    MachineInstr &Root, MachineInstr &Prev,
-    MachineCombinerPattern Pattern,
+    MachineInstr &Root, MachineInstr &Prev, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
@@ -1177,7 +1180,7 @@ void TargetInstrInfo::reassociateOps(
 }
 
 void TargetInstrInfo::genAlternativeCodeSequence(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstIdxForVirtReg) const {
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index 9783b33219460..92647cb405252 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -6043,7 +6043,7 @@ bool AArch64InstrInfo::isAssociativeAndCommutative(const MachineInstr &Inst,
 
 /// Find instructions that can be turned into madd.
 static bool getMaddPatterns(MachineInstr &Root,
-                            SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+                            SmallVectorImpl<unsigned> &Patterns) {
   unsigned Opc = Root.getOpcode();
   MachineBasicBlock &MBB = *Root.getParent();
   bool Found = false;
@@ -6064,21 +6064,21 @@ static bool getMaddPatterns(MachineInstr &Root,
   }
 
   auto setFound = [&](int Opcode, int Operand, unsigned ZeroReg,
-                      MachineCombinerPattern Pattern) {
+                      unsigned Pattern) {
     if (canCombineWithMUL(MBB, Root.getOperand(Operand), Opcode, ZeroReg)) {
       Patterns.push_back(Pattern);
       Found = true;
     }
   };
 
-  auto setVFound = [&](int Opcode, int Operand, MachineCombinerPattern Pattern) {
+  auto setVFound = [&](int Opcode, int Operand, unsigned Pattern) {
     if (canCombine(MBB, Root.getOperand(Operand), Opcode)) {
       Patterns.push_back(Pattern);
       Found = true;
     }
   };
 
-  typedef MachineCombinerPattern MCP;
+  typedef AArch64MachineCombinerPattern MCP;
 
   switch (Opc) {
   default:
@@ -6184,7 +6184,7 @@ static bool getMaddPatterns(MachineInstr &Root,
 
 /// Find instructions that can be turned into madd.
 static bool getFMAPatterns(MachineInstr &Root,
-                           SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+                           SmallVectorImpl<unsigned> &Patterns) {
 
   if (!isCombineInstrCandidateFP(Root))
     return false;
@@ -6192,8 +6192,7 @@ static bool getFMAPatterns(MachineInstr &Root,
   MachineBasicBlock &MBB = *Root.getParent();
   bool Found = false;
 
-  auto Match = [&](int Opcode, int Operand,
-                   MachineCombinerPattern Pattern) -> bool {
+  auto Match = [&](int Opcode, int Operand, unsigned Pattern) -> bool {
     if (canCombineWithFMUL(MBB, Root.getOperand(Operand), Opcode)) {
       Patterns.push_back(Pattern);
       return true;
@@ -6201,7 +6200,7 @@ static bool getFMAPatterns(MachineInstr &Root,
     return false;
   };
 
-  typedef MachineCombinerPattern MCP;
+  typedef AArch64MachineCombinerPattern MCP;
 
   switch (Root.getOpcode()) {
   default:
@@ -6327,12 +6326,11 @@ static bool getFMAPatterns(MachineInstr &Root,
 }
 
 static bool getFMULPatterns(MachineInstr &Root,
-                            SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+                            SmallVectorImpl<unsigned> &Patterns) {
   MachineBasicBlock &MBB = *Root.getParent();
   bool Found = false;
 
-  auto Match = [&](unsigned Opcode, int Operand,
-                   MachineCombinerPattern Pattern) -> bool {
+  auto Match = [&](unsigned Opcode, int Operand, unsigned Pattern) -> bool {
     MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
     MachineOperand &MO = Root.getOperand(Operand);
     MachineInstr *MI = nullptr;
@@ -6349,7 +6347,7 @@ static bool getFMULPatterns(MachineInstr &Root,
     return false;
   };
 
-  typedef MachineCombinerPattern MCP;
+  typedef AArch64MachineCombinerPattern MCP;
 
   switch (Root.getOpcode()) {
   default:
@@ -6380,12 +6378,12 @@ static bool getFMULPatterns(MachineInstr &Root,
 }
 
 static bool getFNEGPatterns(MachineInstr &Root,
-                            SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+                            SmallVectorImpl<unsigned> &Patterns) {
   unsigned Opc = Root.getOpcode();
   MachineBasicBlock &MBB = *Root.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
 
-  auto Match = [&](unsigned Opcode, MachineCombinerPattern Pattern) -> bool {
+  auto Match = [&](unsigned Opcode, unsigned Pattern) -> bool {
     MachineOperand &MO = Root.getOperand(1);
     MachineInstr *MI = MRI.getUniqueVRegDef(MO.getReg());
     if (MI != nullptr && (MI->getOpcode() == Opcode) &&
@@ -6404,9 +6402,9 @@ static bool getFNEGPatterns(MachineInstr &Root,
   default:
     break;
   case AArch64::FNEGDr:
-    return Match(AArch64::FMADDDrrr, MachineCombinerPattern::FNMADD);
+    return Match(AArch64::FMADDDrrr, AArch64MachineCombinerPattern::FNMADD);
   case AArch64::FNEGSr:
-    return Match(AArch64::FMADDSrrr, MachineCombinerPattern::FNMADD);
+    return Match(AArch64::FMADDSrrr, AArch64MachineCombinerPattern::FNMADD);
   }
 
   return false;
@@ -6415,116 +6413,115 @@ static bool getFNEGPatterns(MachineInstr &Root,
 /// Return true when a code sequence can improve throughput. It
 /// should be called only for instructions in loops.
 /// \param Pattern - combiner pattern
-bool AArch64InstrInfo::isThroughputPattern(
-    MachineCombinerPattern Pattern) const {
+bool AArch64InstrInfo::isThroughputPattern(unsigned Pattern) const {
   switch (Pattern) {
   default:
     break;
-  case MachineCombinerPattern::FMULADDH_OP1:
-  case MachineCombinerPattern::FMULADDH_OP2:
-  case MachineCombinerPattern::FMULSUBH_OP1:
-  case MachineCombinerPattern::FMULSUBH_OP2:
-  case MachineCombinerPattern::FMULADDS_OP1:
-  case MachineCombinerPattern::FMULADDS_OP2:
-  case MachineCombinerPattern::FMULSUBS_OP1:
-  case MachineCombinerPattern::FMULSUBS_OP2:
-  case MachineCombinerPattern::FMULADDD_OP1:
-  case MachineCombinerPattern::FMULADDD_OP2:
-  case MachineCombinerPattern::FMULSUBD_OP1:
-  case MachineCombinerPattern::FMULSUBD_OP2:
-  case MachineCombinerPattern::FNMULSUBH_OP1:
-  case MachineCombinerPattern::FNMULSUBS_OP1:
-  case MachineCombinerPattern::FNMULSUBD_OP1:
-  case MachineCombinerPattern::FMLAv4i16_indexed_OP1:
-  case MachineCombinerPattern::FMLAv4i16_indexed_OP2:
-  case MachineCombinerPattern::FMLAv8i16_indexed_OP1:
-  case MachineCombinerPattern::FMLAv8i16_indexed_OP2:
-  case MachineCombinerPattern::FMLAv1i32_indexed_OP1:
-  case MachineCombinerPattern::FMLAv1i32_indexed_OP2:
-  case MachineCombinerPattern::FMLAv1i64_indexed_OP1:
-  case MachineCombinerPattern::FMLAv1i64_indexed_OP2:
-  case MachineCombinerPattern::FMLAv4f16_OP2:
-  case MachineCombinerPattern::FMLAv4f16_OP1:
-  case MachineCombinerPattern::FMLAv8f16_OP1:
-  case MachineCombinerPattern::FMLAv8f16_OP2:
-  case MachineCombinerPattern::FMLAv2f32_OP2:
-  case MachineCombinerPattern::FMLAv2f32_OP1:
-  case MachineCombinerPattern::FMLAv2f64_OP1:
-  case MachineCombinerPattern::FMLAv2f64_OP2:
-  case MachineCombinerPattern::FMLAv2i32_indexed_OP1:
-  case MachineCombinerPattern::FMLAv2i32_indexed_OP2:
-  case MachineCombinerPattern::FMLAv2i64_indexed_OP1:
-  case MachineCombinerPattern::FMLAv2i64_indexed_OP2:
-  case MachineCombinerPattern::FMLAv4f32_OP1:
-  case MachineCombinerPattern::FMLAv4f32_OP2:
-  case MachineCombinerPattern::FMLAv4i32_indexed_OP1:
-  case MachineCombinerPattern::FMLAv4i32_indexed_OP2:
-  case MachineCombinerPattern::FMLSv4i16_indexed_OP1:
-  case MachineCombinerPattern::FMLSv4i16_indexed_OP2:
-  case MachineCombinerPattern::FMLSv8i16_indexed_OP1:
-  case MachineCombinerPattern::FMLSv8i16_indexed_OP2:
-  case MachineCombinerPattern::FMLSv1i32_indexed_OP2:
-  case MachineCombinerPattern::FMLSv1i64_indexed_OP2:
-  case MachineCombinerPattern::FMLSv2i32_indexed_OP2:
-  case MachineCombinerPattern::FMLSv2i64_indexed_OP2:
-  case MachineCombinerPattern::FMLSv4f16_OP1:
-  case MachineCombinerPattern::FMLSv4f16_OP2:
-  case MachineCombinerPattern::FMLSv8f16_OP1:
-  case MachineCombinerPattern::FMLSv8f16_OP2:
-  case MachineCombinerPattern::FMLSv2f32_OP2:
-  case MachineCombinerPattern::FMLSv2f64_OP2:
-  case MachineCombinerPattern::FMLSv4i32_indexed_OP2:
-  case MachineCombinerPattern::FMLSv4f32_OP2:
-  case MachineCombinerPattern::FMULv2i32_indexed_OP1:
-  case MachineCombinerPattern::FMULv2i32_indexed_OP2:
-  case MachineCombinerPattern::FMULv2i64_indexed_OP1:
-  case MachineCombinerPattern::FMULv2i64_indexed_OP2:
-  case MachineCombinerPattern::FMULv4i16_indexed_OP1:
-  case MachineCombinerPattern::FMULv4i16_indexed_OP2:
-  case MachineCombinerPattern::FMULv4i32_indexed_OP1:
-  case MachineCombinerPattern::FMULv4i32_indexed_OP2:
-  case MachineCombinerPattern::FMULv8i16_indexed_OP1:
-  case MachineCombinerPattern::FMULv8i16_indexed_OP2:
-  case MachineCombinerPattern::MULADDv8i8_OP1:
-  case MachineCombinerPattern::MULADDv8i8_OP2:
-  case MachineCombinerPattern::MULADDv16i8_OP1:
-  case MachineCombinerPattern::MULADDv16i8_OP2:
-  case MachineCombinerPattern::MULADDv4i16_OP1:
-  case MachineCombinerPattern::MULADDv4i16_OP2:
-  case MachineCombinerPattern::MULADDv8i16_OP1:
-  case MachineCombinerPattern::MULADDv8i16_OP2:
-  case MachineCombinerPattern::MULADDv2i32_OP1:
-  case MachineCombinerPattern::MULADDv2i32_OP2:
-  case MachineCombinerPattern::MULADDv4i32_OP1:
-  case MachineCombinerPattern::MULADDv4i32_OP2:
-  case MachineCombinerPattern::MULSUBv8i8_OP1:
-  case MachineCombinerPattern::MULSUBv8i8_OP2:
-  case MachineCombinerPattern::MULSUBv16i8_OP1:
-  case MachineCombinerPattern::MULSUBv16i8_OP2:
-  case MachineCombinerPattern::MULSUBv4i16_OP1:
-  case MachineCombinerPattern::MULSUBv4i16_OP2:
-  case MachineCombinerPattern::MULSUBv8i16_OP1:
-  case MachineCombinerPattern::MULSUBv8i16_OP2:
-  case MachineCombinerPattern::MULSUBv2i32_OP1:
-  case MachineCombinerPattern::MULSUBv2i32_OP2:
-  case MachineCombinerPattern::MULSUBv4i32_OP1:
-  case MachineCombinerPattern::MULSUBv4i32_OP2:
-  case MachineCombinerPattern::MULADDv4i16_indexed_OP1:
-  case MachineCombinerPattern::MULADDv4i16_indexed_OP2:
-  case MachineCombinerPattern::MULADDv8i16_indexed_OP1:
-  case MachineCombinerPattern::MULADDv8i16_indexed_OP2:
-  case MachineCombinerPattern::MULADDv2i32_indexed_OP1:
-  case MachineCombinerPattern::MULADDv2i32_indexed_OP2:
-  case MachineCombinerPattern::MULADDv4i32_indexed_OP1:
-  case MachineCombinerPattern::MULADDv4i32_indexed_OP2:
-  case MachineCombinerPattern::MULSUBv4i16_indexed_OP1:
-  case MachineCombinerPattern::MULSUBv4i16_indexed_OP2:
-  case MachineCombinerPattern::MULSUBv8i16_indexed_OP1:
-  case MachineCombinerPattern::MULSUBv8i16_indexed_OP2:
-  case MachineCombinerPattern::MULSUBv2i32_indexed_OP1:
-  case MachineCombinerPattern::MULSUBv2i32_indexed_OP2:
-  case MachineCombinerPattern::MULSUBv4i32_indexed_OP1:
-  case MachineCombinerPattern::MULSUBv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMULADDH_OP1:
+  case AArch64MachineCombinerPattern::FMULADDH_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBH_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBH_OP2:
+  case AArch64MachineCombinerPattern::FMULADDS_OP1:
+  case AArch64MachineCombinerPattern::FMULADDS_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBS_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBS_OP2:
+  case AArch64MachineCombinerPattern::FMULADDD_OP1:
+  case AArch64MachineCombinerPattern::FMULADDD_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBD_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBD_OP2:
+  case AArch64MachineCombinerPattern::FNMULSUBH_OP1:
+  case AArch64MachineCombinerPattern::FNMULSUBS_OP1:
+  case AArch64MachineCombinerPattern::FNMULSUBD_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv1i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv1i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv1i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv1i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4f16_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4f16_OP1:
+  case AArch64MachineCombinerPattern::FMLAv8f16_OP1:
+  case AArch64MachineCombinerPattern::FMLAv8f16_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2f32_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2f32_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2f64_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2f64_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4f32_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4f32_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLSv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLSv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv1i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv1i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4f16_OP1:
+  case AArch64MachineCombinerPattern::FMLSv4f16_OP2:
+  case AArch64MachineCombinerPattern::FMLSv8f16_OP1:
+  case AArch64MachineCombinerPattern::FMLSv8f16_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2f32_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2f64_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4f32_OP2:
+  case AArch64MachineCombinerPattern::FMULv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMULv2i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv2i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMULv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMULv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMULv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i8_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i8_OP2:
+  case AArch64MachineCombinerPattern::MULADDv16i8_OP1:
+  case AArch64MachineCombinerPattern::MULADDv16i8_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i16_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i16_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i16_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i16_OP2:
+  case AArch64MachineCombinerPattern::MULADDv2i32_OP1:
+  case AArch64MachineCombinerPattern::MULADDv2i32_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i32_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i32_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i8_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i8_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv16i8_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv16i8_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP2:
     return true;
   } // end switch (Pattern)
   return false;
@@ -6532,8 +6529,7 @@ bool AArch64InstrInfo::isThroughputPattern(
 
 /// Find other MI combine patterns.
 static bool getMiscPatterns(MachineInstr &Root,
-                            SmallVectorImpl<MachineCombinerPattern> &Patterns)
-{
+                            SmallVectorImpl<unsigned> &Patterns) {
   // A - (B + C)  ==>   (A - B) - C  or  (A - C) - B
   unsigned Opc = Root.getOpcode();
   MachineBasicBlock &MBB = *Root.getParent();
@@ -6557,21 +6553,32 @@ static bool getMiscPatterns(MachineInstr &Root,
       canCombine(MBB, Root.getOperand(2), AArch64::ADDSWrr) ||
       canCombine(MBB, Root.getOperand(2), AArch64::ADDXrr) ||
       canCombine(MBB, Root.getOperand(2), AArch64::ADDSXrr)) {
-    Patterns.push_back(MachineCombinerPattern::SUBADD_OP1);
-    Patterns.push_back(MachineCombinerPattern::SUBADD_OP2);
+    Patterns.push_back(AArch64MachineCombinerPattern::SUBADD_OP1);
+    Patterns.push_back(AArch64MachineCombinerPattern::SUBADD_OP2);
     return true;
   }
 
   return false;
 }
 
+CombinerObjective
+AArch64InstrInfo::getCombinerObjective(unsigned Pattern) const {
+  switch (Pattern) {
+  case AArch64MachineCombinerPattern::SUBADD_OP1:
+  case AArch64MachineCombinerPattern::SUBADD_OP2:
+    return CombinerObjective::MustReduceDepth;
+  default:
+    return TargetInstrInfo::getCombinerObjective(Pattern);
+  }
+}
+
 /// Return true when there is potentially a faster code sequence for an
 /// instruction chain ending in \p Root. All potential patterns are listed in
 /// the \p Pattern vector. Pattern should be sorted in priority order since the
 /// pattern evaluator stops checking as soon as it finds a faster sequence.
 
 bool AArch64InstrInfo::getMachineCombinerPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
+    MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
     bool DoRegPressureReduce) const {
   // Integer patterns
   if (getMaddPatterns(Root, Patterns))
@@ -6930,7 +6937,7 @@ genSubAdd2SubSub(MachineFunction &MF, MachineRegisterInfo &MRI,
 /// this function generates the instructions that could replace the
 /// original code sequence
 void AArch64InstrInfo::genAlternativeCodeSequence(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
@@ -6948,25 +6955,25 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     TargetInstrInfo::genAlternativeCodeSequence(Root, Pattern, InsInstrs,
                                                 DelInstrs, InstrIdxForVirtReg);
     return;
-  case MachineCombinerPattern::SUBADD_OP1:
+  case AArch64MachineCombinerPattern::SUBADD_OP1:
     // A - (B + C)
     // ==> (A - B) - C
     genSubAdd2SubSub(MF, MRI, TII, Root, InsInstrs, DelInstrs, 1,
                      InstrIdxForVirtReg);
     break;
-  case MachineCombinerPattern::SUBADD_OP2:
+  case AArch64MachineCombinerPattern::SUBADD_OP2:
     // A - (B + C)
     // ==> (A - C) - B
     genSubAdd2SubSub(MF, MRI, TII, Root, InsInstrs, DelInstrs, 2,
                      InstrIdxForVirtReg);
     break;
-  case MachineCombinerPattern::MULADDW_OP1:
-  case MachineCombinerPattern::MULADDX_OP1:
+  case AArch64MachineCombinerPattern::MULADDW_OP1:
+  case AArch64MachineCombinerPattern::MULADDX_OP1:
     // MUL I=A,B,0
     // ADD R,I,C
     // ==> MADD R,A,B,C
     // --- Create(MADD);
-    if (Pattern == MachineCombinerPattern::MULADDW_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::MULADDW_OP1) {
       Opc = AArch64::MADDWrrr;
       RC = &AArch64::GPR32RegClass;
     } else {
@@ -6975,13 +6982,13 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDW_OP2:
-  case MachineCombinerPattern::MULADDX_OP2:
+  case AArch64MachineCombinerPattern::MULADDW_OP2:
+  case AArch64MachineCombinerPattern::MULADDX_OP2:
     // MUL I=A,B,0
     // ADD R,C,I
     // ==> MADD R,A,B,C
     // --- Create(MADD);
-    if (Pattern == MachineCombinerPattern::MULADDW_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::MULADDW_OP2) {
       Opc = AArch64::MADDWrrr;
       RC = &AArch64::GPR32RegClass;
     } else {
@@ -6990,8 +6997,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDWI_OP1:
-  case MachineCombinerPattern::MULADDXI_OP1: {
+  case AArch64MachineCombinerPattern::MULADDWI_OP1:
+  case AArch64MachineCombinerPattern::MULADDXI_OP1: {
     // MUL I=A,B,0
     // ADD R,I,Imm
     // ==> MOV V, Imm
@@ -6999,7 +7006,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     // --- Create(MADD);
     const TargetRegisterClass *OrrRC;
     unsigned BitSize, OrrOpc, ZeroReg;
-    if (Pattern == MachineCombinerPattern::MULADDWI_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::MULADDWI_OP1) {
       OrrOpc = AArch64::ORRWri;
       OrrRC = &AArch64::GPR32spRegClass;
       BitSize = 32;
@@ -7052,8 +7059,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     MUL = genMaddR(MF, MRI, TII, Root, InsInstrs, 1, Opc, NewVR, RC);
     break;
   }
-  case MachineCombinerPattern::MULSUBW_OP1:
-  case MachineCombinerPattern::MULSUBX_OP1: {
+  case AArch64MachineCombinerPattern::MULSUBW_OP1:
+  case AArch64MachineCombinerPattern::MULSUBX_OP1: {
     // MUL I=A,B,0
     // SUB R,I, C
     // ==> SUB  V, 0, C
@@ -7061,7 +7068,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     // --- Create(MADD);
     const TargetRegisterClass *SubRC;
     unsigned SubOpc, ZeroReg;
-    if (Pattern == MachineCombinerPattern::MULSUBW_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::MULSUBW_OP1) {
       SubOpc = AArch64::SUBWrr;
       SubRC = &AArch64::GPR32spRegClass;
       ZeroReg = AArch64::WZR;
@@ -7085,13 +7092,13 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     MUL = genMaddR(MF, MRI, TII, Root, InsInstrs, 1, Opc, NewVR, RC);
     break;
   }
-  case MachineCombinerPattern::MULSUBW_OP2:
-  case MachineCombinerPattern::MULSUBX_OP2:
+  case AArch64MachineCombinerPattern::MULSUBW_OP2:
+  case AArch64MachineCombinerPattern::MULSUBX_OP2:
     // MUL I=A,B,0
     // SUB R,C,I
     // ==> MSUB R,A,B,C (computes C - A*B)
     // --- Create(MSUB);
-    if (Pattern == MachineCombinerPattern::MULSUBW_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::MULSUBW_OP2) {
       Opc = AArch64::MSUBWrrr;
       RC = &AArch64::GPR32RegClass;
     } else {
@@ -7100,8 +7107,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBWI_OP1:
-  case MachineCombinerPattern::MULSUBXI_OP1: {
+  case AArch64MachineCombinerPattern::MULSUBWI_OP1:
+  case AArch64MachineCombinerPattern::MULSUBXI_OP1: {
     // MUL I=A,B,0
     // SUB R,I, Imm
     // ==> MOV  V, -Imm
@@ -7109,7 +7116,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     // --- Create(MADD);
     const TargetRegisterClass *OrrRC;
     unsigned BitSize, OrrOpc, ZeroReg;
-    if (Pattern == MachineCombinerPattern::MULSUBWI_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::MULSUBWI_OP1) {
       OrrOpc = AArch64::ORRWri;
       OrrRC = &AArch64::GPR32spRegClass;
       BitSize = 32;
@@ -7162,318 +7169,318 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     break;
   }
 
-  case MachineCombinerPattern::MULADDv8i8_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i8_OP1:
     Opc = AArch64::MLAv8i8;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv8i8_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i8_OP2:
     Opc = AArch64::MLAv8i8;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv16i8_OP1:
+  case AArch64MachineCombinerPattern::MULADDv16i8_OP1:
     Opc = AArch64::MLAv16i8;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv16i8_OP2:
+  case AArch64MachineCombinerPattern::MULADDv16i8_OP2:
     Opc = AArch64::MLAv16i8;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i16_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i16_OP1:
     Opc = AArch64::MLAv4i16;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i16_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i16_OP2:
     Opc = AArch64::MLAv4i16;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv8i16_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i16_OP1:
     Opc = AArch64::MLAv8i16;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv8i16_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i16_OP2:
     Opc = AArch64::MLAv8i16;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv2i32_OP1:
+  case AArch64MachineCombinerPattern::MULADDv2i32_OP1:
     Opc = AArch64::MLAv2i32;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv2i32_OP2:
+  case AArch64MachineCombinerPattern::MULADDv2i32_OP2:
     Opc = AArch64::MLAv2i32;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i32_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i32_OP1:
     Opc = AArch64::MLAv4i32;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i32_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i32_OP2:
     Opc = AArch64::MLAv4i32;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
-  case MachineCombinerPattern::MULSUBv8i8_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i8_OP1:
     Opc = AArch64::MLAv8i8;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv8i8,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv8i8_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i8_OP2:
     Opc = AArch64::MLSv8i8;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv16i8_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv16i8_OP1:
     Opc = AArch64::MLAv16i8;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv16i8,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv16i8_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv16i8_OP2:
     Opc = AArch64::MLSv16i8;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i16_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_OP1:
     Opc = AArch64::MLAv4i16;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv4i16,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i16_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_OP2:
     Opc = AArch64::MLSv4i16;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv8i16_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_OP1:
     Opc = AArch64::MLAv8i16;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv8i16,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv8i16_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_OP2:
     Opc = AArch64::MLSv8i16;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv2i32_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_OP1:
     Opc = AArch64::MLAv2i32;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv2i32,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv2i32_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_OP2:
     Opc = AArch64::MLSv2i32;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i32_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_OP1:
     Opc = AArch64::MLAv4i32;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAccNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv4i32,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i32_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_OP2:
     Opc = AArch64::MLSv4i32;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyAcc(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
-  case MachineCombinerPattern::MULADDv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i16_indexed_OP1:
     Opc = AArch64::MLAv4i16_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i16_indexed_OP2:
     Opc = AArch64::MLAv4i16_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv8i16_indexed_OP1:
     Opc = AArch64::MLAv8i16_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv8i16_indexed_OP2:
     Opc = AArch64::MLAv8i16_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv2i32_indexed_OP1:
     Opc = AArch64::MLAv2i32_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv2i32_indexed_OP2:
     Opc = AArch64::MLAv2i32_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULADDv4i32_indexed_OP1:
     Opc = AArch64::MLAv4i32_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::MULADDv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULADDv4i32_indexed_OP2:
     Opc = AArch64::MLAv4i32_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
-  case MachineCombinerPattern::MULSUBv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_indexed_OP1:
     Opc = AArch64::MLAv4i16_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdxNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv4i16,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i16_indexed_OP2:
     Opc = AArch64::MLSv4i16_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_indexed_OP1:
     Opc = AArch64::MLAv8i16_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdxNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv8i16,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv8i16_indexed_OP2:
     Opc = AArch64::MLSv8i16_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_indexed_OP1:
     Opc = AArch64::MLAv2i32_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdxNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv2i32,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv2i32_indexed_OP2:
     Opc = AArch64::MLSv2i32_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP1:
     Opc = AArch64::MLAv4i32_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdxNeg(MF, MRI, TII, Root, InsInstrs,
                                  InstrIdxForVirtReg, 1, Opc, AArch64::NEGv4i32,
                                  RC);
     break;
-  case MachineCombinerPattern::MULSUBv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP2:
     Opc = AArch64::MLSv4i32_indexed;
     RC = &AArch64::FPR128RegClass;
     MUL = genFusedMultiplyIdx(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
   // Floating Point Support
-  case MachineCombinerPattern::FMULADDH_OP1:
+  case AArch64MachineCombinerPattern::FMULADDH_OP1:
     Opc = AArch64::FMADDHrrr;
     RC = &AArch64::FPR16RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULADDS_OP1:
+  case AArch64MachineCombinerPattern::FMULADDS_OP1:
     Opc = AArch64::FMADDSrrr;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULADDD_OP1:
+  case AArch64MachineCombinerPattern::FMULADDD_OP1:
     Opc = AArch64::FMADDDrrr;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
 
-  case MachineCombinerPattern::FMULADDH_OP2:
+  case AArch64MachineCombinerPattern::FMULADDH_OP2:
     Opc = AArch64::FMADDHrrr;
     RC = &AArch64::FPR16RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULADDS_OP2:
+  case AArch64MachineCombinerPattern::FMULADDS_OP2:
     Opc = AArch64::FMADDSrrr;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULADDD_OP2:
+  case AArch64MachineCombinerPattern::FMULADDD_OP2:
     Opc = AArch64::FMADDDrrr;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
-  case MachineCombinerPattern::FMLAv1i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv1i32_indexed_OP1:
     Opc = AArch64::FMLAv1i32_indexed;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv1i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv1i32_indexed_OP2:
     Opc = AArch64::FMLAv1i32_indexed;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLAv1i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv1i64_indexed_OP1:
     Opc = AArch64::FMLAv1i64_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv1i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv1i64_indexed_OP2:
     Opc = AArch64::FMLAv1i64_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLAv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4i16_indexed_OP1:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLAv4i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv4f16_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4f16_OP1:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLAv4f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
-  case MachineCombinerPattern::FMLAv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4i16_indexed_OP2:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLAv4i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv4f16_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4f16_OP2:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLAv4f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
 
-  case MachineCombinerPattern::FMLAv2i32_indexed_OP1:
-  case MachineCombinerPattern::FMLAv2f32_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2f32_OP1:
     RC = &AArch64::FPR64RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv2i32_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP1) {
       Opc = AArch64::FMLAv2i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7483,10 +7490,10 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
                              FMAInstKind::Accumulator);
     }
     break;
-  case MachineCombinerPattern::FMLAv2i32_indexed_OP2:
-  case MachineCombinerPattern::FMLAv2f32_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2f32_OP2:
     RC = &AArch64::FPR64RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv2i32_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv2i32_indexed_OP2) {
       Opc = AArch64::FMLAv2i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7497,35 +7504,35 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMLAv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv8i16_indexed_OP1:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLAv8i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv8f16_OP1:
+  case AArch64MachineCombinerPattern::FMLAv8f16_OP1:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLAv8f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
-  case MachineCombinerPattern::FMLAv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv8i16_indexed_OP2:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLAv8i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
-  case MachineCombinerPattern::FMLAv8f16_OP2:
+  case AArch64MachineCombinerPattern::FMLAv8f16_OP2:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLAv8f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
 
-  case MachineCombinerPattern::FMLAv2i64_indexed_OP1:
-  case MachineCombinerPattern::FMLAv2f64_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv2f64_OP1:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv2i64_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP1) {
       Opc = AArch64::FMLAv2i64_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7535,10 +7542,10 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
                              FMAInstKind::Accumulator);
     }
     break;
-  case MachineCombinerPattern::FMLAv2i64_indexed_OP2:
-  case MachineCombinerPattern::FMLAv2f64_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv2f64_OP2:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv2i64_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv2i64_indexed_OP2) {
       Opc = AArch64::FMLAv2i64_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7549,10 +7556,10 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMLAv4i32_indexed_OP1:
-  case MachineCombinerPattern::FMLAv4f32_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMLAv4f32_OP1:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv4i32_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP1) {
       Opc = AArch64::FMLAv4i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7563,10 +7570,10 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMLAv4i32_indexed_OP2:
-  case MachineCombinerPattern::FMLAv4f32_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLAv4f32_OP2:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLAv4i32_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLAv4i32_indexed_OP2) {
       Opc = AArch64::FMLAv4i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7577,70 +7584,70 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMULSUBH_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBH_OP1:
     Opc = AArch64::FNMSUBHrrr;
     RC = &AArch64::FPR16RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULSUBS_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBS_OP1:
     Opc = AArch64::FNMSUBSrrr;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULSUBD_OP1:
+  case AArch64MachineCombinerPattern::FMULSUBD_OP1:
     Opc = AArch64::FNMSUBDrrr;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
 
-  case MachineCombinerPattern::FNMULSUBH_OP1:
+  case AArch64MachineCombinerPattern::FNMULSUBH_OP1:
     Opc = AArch64::FNMADDHrrr;
     RC = &AArch64::FPR16RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FNMULSUBS_OP1:
+  case AArch64MachineCombinerPattern::FNMULSUBS_OP1:
     Opc = AArch64::FNMADDSrrr;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
-  case MachineCombinerPattern::FNMULSUBD_OP1:
+  case AArch64MachineCombinerPattern::FNMULSUBD_OP1:
     Opc = AArch64::FNMADDDrrr;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC);
     break;
 
-  case MachineCombinerPattern::FMULSUBH_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBH_OP2:
     Opc = AArch64::FMSUBHrrr;
     RC = &AArch64::FPR16RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULSUBS_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBS_OP2:
     Opc = AArch64::FMSUBSrrr;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
-  case MachineCombinerPattern::FMULSUBD_OP2:
+  case AArch64MachineCombinerPattern::FMULSUBD_OP2:
     Opc = AArch64::FMSUBDrrr;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC);
     break;
 
-  case MachineCombinerPattern::FMLSv1i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv1i32_indexed_OP2:
     Opc = AArch64::FMLSv1i32_indexed;
     RC = &AArch64::FPR32RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLSv1i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv1i64_indexed_OP2:
     Opc = AArch64::FMLSv1i64_indexed;
     RC = &AArch64::FPR64RegClass;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLSv4f16_OP1:
-  case MachineCombinerPattern::FMLSv4i16_indexed_OP1: {
+  case AArch64MachineCombinerPattern::FMLSv4f16_OP1:
+  case AArch64MachineCombinerPattern::FMLSv4i16_indexed_OP1: {
     RC = &AArch64::FPR64RegClass;
     Register NewVR = MRI.createVirtualRegister(RC);
     MachineInstrBuilder MIB1 =
@@ -7648,7 +7655,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
             .add(Root.getOperand(2));
     InsInstrs.push_back(MIB1);
     InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
-    if (Pattern == MachineCombinerPattern::FMLSv4f16_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv4f16_OP1) {
       Opc = AArch64::FMLAv4f16;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Accumulator, &NewVR);
@@ -7659,23 +7666,23 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
   }
-  case MachineCombinerPattern::FMLSv4f16_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4f16_OP2:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLSv4f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
-  case MachineCombinerPattern::FMLSv4i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4i16_indexed_OP2:
     RC = &AArch64::FPR64RegClass;
     Opc = AArch64::FMLSv4i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLSv2f32_OP2:
-  case MachineCombinerPattern::FMLSv2i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2f32_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2i32_indexed_OP2:
     RC = &AArch64::FPR64RegClass;
-    if (Pattern == MachineCombinerPattern::FMLSv2i32_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv2i32_indexed_OP2) {
       Opc = AArch64::FMLSv2i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7686,8 +7693,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMLSv8f16_OP1:
-  case MachineCombinerPattern::FMLSv8i16_indexed_OP1: {
+  case AArch64MachineCombinerPattern::FMLSv8f16_OP1:
+  case AArch64MachineCombinerPattern::FMLSv8i16_indexed_OP1: {
     RC = &AArch64::FPR128RegClass;
     Register NewVR = MRI.createVirtualRegister(RC);
     MachineInstrBuilder MIB1 =
@@ -7695,7 +7702,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
             .add(Root.getOperand(2));
     InsInstrs.push_back(MIB1);
     InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
-    if (Pattern == MachineCombinerPattern::FMLSv8f16_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv8f16_OP1) {
       Opc = AArch64::FMLAv8f16;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Accumulator, &NewVR);
@@ -7706,23 +7713,23 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
   }
-  case MachineCombinerPattern::FMLSv8f16_OP2:
+  case AArch64MachineCombinerPattern::FMLSv8f16_OP2:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLSv8f16;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Accumulator);
     break;
-  case MachineCombinerPattern::FMLSv8i16_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv8i16_indexed_OP2:
     RC = &AArch64::FPR128RegClass;
     Opc = AArch64::FMLSv8i16_indexed;
     MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                            FMAInstKind::Indexed);
     break;
 
-  case MachineCombinerPattern::FMLSv2f64_OP2:
-  case MachineCombinerPattern::FMLSv2i64_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2f64_OP2:
+  case AArch64MachineCombinerPattern::FMLSv2i64_indexed_OP2:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLSv2i64_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv2i64_indexed_OP2) {
       Opc = AArch64::FMLSv2i64_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7733,10 +7740,10 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
 
-  case MachineCombinerPattern::FMLSv4f32_OP2:
-  case MachineCombinerPattern::FMLSv4i32_indexed_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4f32_OP2:
+  case AArch64MachineCombinerPattern::FMLSv4i32_indexed_OP2:
     RC = &AArch64::FPR128RegClass;
-    if (Pattern == MachineCombinerPattern::FMLSv4i32_indexed_OP2) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv4i32_indexed_OP2) {
       Opc = AArch64::FMLSv4i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 2, Opc, RC,
                              FMAInstKind::Indexed);
@@ -7746,8 +7753,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
                              FMAInstKind::Accumulator);
     }
     break;
-  case MachineCombinerPattern::FMLSv2f32_OP1:
-  case MachineCombinerPattern::FMLSv2i32_indexed_OP1: {
+  case AArch64MachineCombinerPattern::FMLSv2f32_OP1:
+  case AArch64MachineCombinerPattern::FMLSv2i32_indexed_OP1: {
     RC = &AArch64::FPR64RegClass;
     Register NewVR = MRI.createVirtualRegister(RC);
     MachineInstrBuilder MIB1 =
@@ -7755,7 +7762,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
             .add(Root.getOperand(2));
     InsInstrs.push_back(MIB1);
     InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
-    if (Pattern == MachineCombinerPattern::FMLSv2i32_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv2i32_indexed_OP1) {
       Opc = AArch64::FMLAv2i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed, &NewVR);
@@ -7766,8 +7773,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
   }
-  case MachineCombinerPattern::FMLSv4f32_OP1:
-  case MachineCombinerPattern::FMLSv4i32_indexed_OP1: {
+  case AArch64MachineCombinerPattern::FMLSv4f32_OP1:
+  case AArch64MachineCombinerPattern::FMLSv4i32_indexed_OP1: {
     RC = &AArch64::FPR128RegClass;
     Register NewVR = MRI.createVirtualRegister(RC);
     MachineInstrBuilder MIB1 =
@@ -7775,7 +7782,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
             .add(Root.getOperand(2));
     InsInstrs.push_back(MIB1);
     InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
-    if (Pattern == MachineCombinerPattern::FMLSv4i32_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv4i32_indexed_OP1) {
       Opc = AArch64::FMLAv4i32_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed, &NewVR);
@@ -7786,8 +7793,8 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
   }
-  case MachineCombinerPattern::FMLSv2f64_OP1:
-  case MachineCombinerPattern::FMLSv2i64_indexed_OP1: {
+  case AArch64MachineCombinerPattern::FMLSv2f64_OP1:
+  case AArch64MachineCombinerPattern::FMLSv2i64_indexed_OP1: {
     RC = &AArch64::FPR128RegClass;
     Register NewVR = MRI.createVirtualRegister(RC);
     MachineInstrBuilder MIB1 =
@@ -7795,7 +7802,7 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
             .add(Root.getOperand(2));
     InsInstrs.push_back(MIB1);
     InstrIdxForVirtReg.insert(std::make_pair(NewVR, 0));
-    if (Pattern == MachineCombinerPattern::FMLSv2i64_indexed_OP1) {
+    if (Pattern == AArch64MachineCombinerPattern::FMLSv2i64_indexed_OP1) {
       Opc = AArch64::FMLAv2i64_indexed;
       MUL = genFusedMultiply(MF, MRI, TII, Root, InsInstrs, 1, Opc, RC,
                              FMAInstKind::Indexed, &NewVR);
@@ -7806,47 +7813,52 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     }
     break;
   }
-  case MachineCombinerPattern::FMULv2i32_indexed_OP1:
-  case MachineCombinerPattern::FMULv2i32_indexed_OP2: {
+  case AArch64MachineCombinerPattern::FMULv2i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv2i32_indexed_OP2: {
     unsigned IdxDupOp =
-        (Pattern == MachineCombinerPattern::FMULv2i32_indexed_OP1) ? 1 : 2;
+        (Pattern == AArch64MachineCombinerPattern::FMULv2i32_indexed_OP1) ? 1
+                                                                          : 2;
     genIndexedMultiply(Root, InsInstrs, IdxDupOp, AArch64::FMULv2i32_indexed,
                        &AArch64::FPR128RegClass, MRI);
     break;
   }
-  case MachineCombinerPattern::FMULv2i64_indexed_OP1:
-  case MachineCombinerPattern::FMULv2i64_indexed_OP2: {
+  case AArch64MachineCombinerPattern::FMULv2i64_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv2i64_indexed_OP2: {
     unsigned IdxDupOp =
-        (Pattern == MachineCombinerPattern::FMULv2i64_indexed_OP1) ? 1 : 2;
+        (Pattern == AArch64MachineCombinerPattern::FMULv2i64_indexed_OP1) ? 1
+                                                                          : 2;
     genIndexedMultiply(Root, InsInstrs, IdxDupOp, AArch64::FMULv2i64_indexed,
                        &AArch64::FPR128RegClass, MRI);
     break;
   }
-  case MachineCombinerPattern::FMULv4i16_indexed_OP1:
-  case MachineCombinerPattern::FMULv4i16_indexed_OP2: {
+  case AArch64MachineCombinerPattern::FMULv4i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv4i16_indexed_OP2: {
     unsigned IdxDupOp =
-        (Pattern == MachineCombinerPattern::FMULv4i16_indexed_OP1) ? 1 : 2;
+        (Pattern == AArch64MachineCombinerPattern::FMULv4i16_indexed_OP1) ? 1
+                                                                          : 2;
     genIndexedMultiply(Root, InsInstrs, IdxDupOp, AArch64::FMULv4i16_indexed,
                        &AArch64::FPR128_loRegClass, MRI);
     break;
   }
-  case MachineCombinerPattern::FMULv4i32_indexed_OP1:
-  case MachineCombinerPattern::FMULv4i32_indexed_OP2: {
+  case AArch64MachineCombinerPattern::FMULv4i32_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv4i32_indexed_OP2: {
     unsigned IdxDupOp =
-        (Pattern == MachineCombinerPattern::FMULv4i32_indexed_OP1) ? 1 : 2;
+        (Pattern == AArch64MachineCombinerPattern::FMULv4i32_indexed_OP1) ? 1
+                                                                          : 2;
     genIndexedMultiply(Root, InsInstrs, IdxDupOp, AArch64::FMULv4i32_indexed,
                        &AArch64::FPR128RegClass, MRI);
     break;
   }
-  case MachineCombinerPattern::FMULv8i16_indexed_OP1:
-  case MachineCombinerPattern::FMULv8i16_indexed_OP2: {
+  case AArch64MachineCombinerPattern::FMULv8i16_indexed_OP1:
+  case AArch64MachineCombinerPattern::FMULv8i16_indexed_OP2: {
     unsigned IdxDupOp =
-        (Pattern == MachineCombinerPattern::FMULv8i16_indexed_OP1) ? 1 : 2;
+        (Pattern == AArch64MachineCombinerPattern::FMULv8i16_indexed_OP1) ? 1
+                                                                          : 2;
     genIndexedMultiply(Root, InsInstrs, IdxDupOp, AArch64::FMULv8i16_indexed,
                        &AArch64::FPR128_loRegClass, MRI);
     break;
   }
-  case MachineCombinerPattern::FNMADD: {
+  case AArch64MachineCombinerPattern::FNMADD: {
     MUL = genFNegatedMAD(MF, MRI, TII, Root, InsInstrs);
     break;
   }
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.h b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
index 2f10f80f4bdf7..9a2914891675c 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.h
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
@@ -33,6 +33,146 @@ static const MachineMemOperand::Flags MOStridedAccess =
 
 #define FALKOR_STRIDED_ACCESS_MD "falkor.strided.access"
 
+// AArch64 MachineCombiner patterns
+enum AArch64MachineCombinerPattern : unsigned {
+  // These are patterns used to reduce the length of dependence chain.
+  SUBADD_OP1 = MachineCombinerPattern::TARGET_PATTERN_START,
+  SUBADD_OP2,
+
+  // These are multiply-add patterns matched by the AArch64 machine combiner.
+  MULADDW_OP1,
+  MULADDW_OP2,
+  MULSUBW_OP1,
+  MULSUBW_OP2,
+  MULADDWI_OP1,
+  MULSUBWI_OP1,
+  MULADDX_OP1,
+  MULADDX_OP2,
+  MULSUBX_OP1,
+  MULSUBX_OP2,
+  MULADDXI_OP1,
+  MULSUBXI_OP1,
+  // NEON integers vectors
+  MULADDv8i8_OP1,
+  MULADDv8i8_OP2,
+  MULADDv16i8_OP1,
+  MULADDv16i8_OP2,
+  MULADDv4i16_OP1,
+  MULADDv4i16_OP2,
+  MULADDv8i16_OP1,
+  MULADDv8i16_OP2,
+  MULADDv2i32_OP1,
+  MULADDv2i32_OP2,
+  MULADDv4i32_OP1,
+  MULADDv4i32_OP2,
+
+  MULSUBv8i8_OP1,
+  MULSUBv8i8_OP2,
+  MULSUBv16i8_OP1,
+  MULSUBv16i8_OP2,
+  MULSUBv4i16_OP1,
+  MULSUBv4i16_OP2,
+  MULSUBv8i16_OP1,
+  MULSUBv8i16_OP2,
+  MULSUBv2i32_OP1,
+  MULSUBv2i32_OP2,
+  MULSUBv4i32_OP1,
+  MULSUBv4i32_OP2,
+
+  MULADDv4i16_indexed_OP1,
+  MULADDv4i16_indexed_OP2,
+  MULADDv8i16_indexed_OP1,
+  MULADDv8i16_indexed_OP2,
+  MULADDv2i32_indexed_OP1,
+  MULADDv2i32_indexed_OP2,
+  MULADDv4i32_indexed_OP1,
+  MULADDv4i32_indexed_OP2,
+
+  MULSUBv4i16_indexed_OP1,
+  MULSUBv4i16_indexed_OP2,
+  MULSUBv8i16_indexed_OP1,
+  MULSUBv8i16_indexed_OP2,
+  MULSUBv2i32_indexed_OP1,
+  MULSUBv2i32_indexed_OP2,
+  MULSUBv4i32_indexed_OP1,
+  MULSUBv4i32_indexed_OP2,
+
+  // Floating Point
+  FMULADDH_OP1,
+  FMULADDH_OP2,
+  FMULSUBH_OP1,
+  FMULSUBH_OP2,
+  FMULADDS_OP1,
+  FMULADDS_OP2,
+  FMULSUBS_OP1,
+  FMULSUBS_OP2,
+  FMULADDD_OP1,
+  FMULADDD_OP2,
+  FMULSUBD_OP1,
+  FMULSUBD_OP2,
+  FNMULSUBH_OP1,
+  FNMULSUBS_OP1,
+  FNMULSUBD_OP1,
+  FMLAv1i32_indexed_OP1,
+  FMLAv1i32_indexed_OP2,
+  FMLAv1i64_indexed_OP1,
+  FMLAv1i64_indexed_OP2,
+  FMLAv4f16_OP1,
+  FMLAv4f16_OP2,
+  FMLAv8f16_OP1,
+  FMLAv8f16_OP2,
+  FMLAv2f32_OP2,
+  FMLAv2f32_OP1,
+  FMLAv2f64_OP1,
+  FMLAv2f64_OP2,
+  FMLAv4i16_indexed_OP1,
+  FMLAv4i16_indexed_OP2,
+  FMLAv8i16_indexed_OP1,
+  FMLAv8i16_indexed_OP2,
+  FMLAv2i32_indexed_OP1,
+  FMLAv2i32_indexed_OP2,
+  FMLAv2i64_indexed_OP1,
+  FMLAv2i64_indexed_OP2,
+  FMLAv4f32_OP1,
+  FMLAv4f32_OP2,
+  FMLAv4i32_indexed_OP1,
+  FMLAv4i32_indexed_OP2,
+  FMLSv1i32_indexed_OP2,
+  FMLSv1i64_indexed_OP2,
+  FMLSv4f16_OP1,
+  FMLSv4f16_OP2,
+  FMLSv8f16_OP1,
+  FMLSv8f16_OP2,
+  FMLSv2f32_OP1,
+  FMLSv2f32_OP2,
+  FMLSv2f64_OP1,
+  FMLSv2f64_OP2,
+  FMLSv4i16_indexed_OP1,
+  FMLSv4i16_indexed_OP2,
+  FMLSv8i16_indexed_OP1,
+  FMLSv8i16_indexed_OP2,
+  FMLSv2i32_indexed_OP1,
+  FMLSv2i32_indexed_OP2,
+  FMLSv2i64_indexed_OP1,
+  FMLSv2i64_indexed_OP2,
+  FMLSv4f32_OP1,
+  FMLSv4f32_OP2,
+  FMLSv4i32_indexed_OP1,
+  FMLSv4i32_indexed_OP2,
+
+  FMULv2i32_indexed_OP1,
+  FMULv2i32_indexed_OP2,
+  FMULv2i64_indexed_OP1,
+  FMULv2i64_indexed_OP2,
+  FMULv4i16_indexed_OP1,
+  FMULv4i16_indexed_OP2,
+  FMULv4i32_indexed_OP1,
+  FMULv4i32_indexed_OP2,
+  FMULv8i16_indexed_OP1,
+  FMULv8i16_indexed_OP2,
+
+  FNMADD,
+};
 class AArch64InstrInfo final : public AArch64GenInstrInfo {
   const AArch64RegisterInfo RI;
   const AArch64Subtarget &Subtarget;
@@ -283,17 +423,17 @@ class AArch64InstrInfo final : public AArch64GenInstrInfo {
                             const MachineRegisterInfo *MRI) const override;
   bool optimizeCondBranch(MachineInstr &MI) const override;
 
+  CombinerObjective getCombinerObjective(unsigned Pattern) const override;
   /// Return true when a code sequence can improve throughput. It
   /// should be called only for instructions in loops.
   /// \param Pattern - combiner pattern
-  bool isThroughputPattern(MachineCombinerPattern Pattern) const override;
+  bool isThroughputPattern(unsigned Pattern) const override;
   /// Return true when there is potentially a faster code sequence
   /// for an instruction chain ending in ``Root``. All potential patterns are
   /// listed in the ``Patterns`` array.
-  bool
-  getMachineCombinerPatterns(MachineInstr &Root,
-                             SmallVectorImpl<MachineCombinerPattern> &Patterns,
-                             bool DoRegPressureReduce) const override;
+  bool getMachineCombinerPatterns(MachineInstr &Root,
+                                  SmallVectorImpl<unsigned> &Patterns,
+                                  bool DoRegPressureReduce) const override;
   /// Return true when Inst is associative and commutative so that it can be
   /// reassociated. If Invert is true, then the inverse of Inst operation must
   /// be checked.
@@ -302,7 +442,7 @@ class AArch64InstrInfo final : public AArch64GenInstrInfo {
   /// When getMachineCombinerPatterns() finds patterns, this function generates
   /// the instructions that could replace the original code sequence
   void genAlternativeCodeSequence(
-      MachineInstr &Root, MachineCombinerPattern Pattern,
+      MachineInstr &Root, unsigned Pattern,
       SmallVectorImpl<MachineInstr *> &InsInstrs,
       SmallVectorImpl<MachineInstr *> &DelInstrs,
       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const override;
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
index 5f5eb31a5a85f..93874d65531ae 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -348,9 +348,9 @@ int16_t PPCInstrInfo::getFMAOpIdxInfo(unsigned Opcode) const {
 //  register with D. After the transformation, A and D must be assigned with
 //  same hardware register due to TIE attribute of FMA instructions.
 //
-bool PPCInstrInfo::getFMAPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
-    bool DoRegPressureReduce) const {
+bool PPCInstrInfo::getFMAPatterns(MachineInstr &Root,
+                                  SmallVectorImpl<unsigned> &Patterns,
+                                  bool DoRegPressureReduce) const {
   MachineBasicBlock *MBB = Root.getParent();
   const MachineRegisterInfo *MRI = &MBB->getParent()->getRegInfo();
   const TargetRegisterInfo *TRI = &getRegisterInfo();
@@ -476,7 +476,7 @@ bool PPCInstrInfo::getFMAPatterns(
     if (isLoadFromConstantPool(MULInstrL) && IsUsedOnceR &&
         IsReassociableAddOrSub(*MULInstrR, InfoArrayIdxFSubInst)) {
       LLVM_DEBUG(dbgs() << "add pattern REASSOC_XY_BCA\n");
-      Patterns.push_back(MachineCombinerPattern::REASSOC_XY_BCA);
+      Patterns.push_back(PPCMachineCombinerPattern::REASSOC_XY_BCA);
       return true;
     }
 
@@ -484,7 +484,7 @@ bool PPCInstrInfo::getFMAPatterns(
     if ((isLoadFromConstantPool(MULInstrR) && IsUsedOnceL &&
          IsReassociableAddOrSub(*MULInstrL, InfoArrayIdxFSubInst))) {
       LLVM_DEBUG(dbgs() << "add pattern REASSOC_XY_BAC\n");
-      Patterns.push_back(MachineCombinerPattern::REASSOC_XY_BAC);
+      Patterns.push_back(PPCMachineCombinerPattern::REASSOC_XY_BAC);
       return true;
     }
   }
@@ -511,12 +511,12 @@ bool PPCInstrInfo::getFMAPatterns(
   MachineInstr *Leaf = MRI->getUniqueVRegDef(RegA);
   AddOpIdx = -1;
   if (IsReassociableFMA(*Leaf, AddOpIdx, MulOpIdx, true)) {
-    Patterns.push_back(MachineCombinerPattern::REASSOC_XMM_AMM_BMM);
+    Patterns.push_back(PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM);
     LLVM_DEBUG(dbgs() << "add pattern REASSOC_XMM_AMM_BMM\n");
     return true;
   }
   if (IsReassociableAddOrSub(*Leaf, InfoArrayIdxFAddInst)) {
-    Patterns.push_back(MachineCombinerPattern::REASSOC_XY_AMM_BMM);
+    Patterns.push_back(PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM);
     LLVM_DEBUG(dbgs() << "add pattern REASSOC_XY_AMM_BMM\n");
     return true;
   }
@@ -524,7 +524,7 @@ bool PPCInstrInfo::getFMAPatterns(
 }
 
 void PPCInstrInfo::finalizeInsInstrs(
-    MachineInstr &Root, MachineCombinerPattern &P,
+    MachineInstr &Root, unsigned &Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs) const {
   assert(!InsInstrs.empty() && "Instructions set to be inserted is empty!");
 
@@ -542,12 +542,12 @@ void PPCInstrInfo::finalizeInsInstrs(
   // For now we only need to fix up placeholder for register pressure reduce
   // patterns.
   Register ConstReg = 0;
-  switch (P) {
-  case MachineCombinerPattern::REASSOC_XY_BCA:
+  switch (Pattern) {
+  case PPCMachineCombinerPattern::REASSOC_XY_BCA:
     ConstReg =
         TRI->lookThruCopyLike(Root.getOperand(FirstMulOpIdx).getReg(), MRI);
     break;
-  case MachineCombinerPattern::REASSOC_XY_BAC:
+  case PPCMachineCombinerPattern::REASSOC_XY_BAC:
     ConstReg =
         TRI->lookThruCopyLike(Root.getOperand(FirstMulOpIdx + 1).getReg(), MRI);
     break;
@@ -737,8 +737,21 @@ PPCInstrInfo::getConstantFromConstantPool(MachineInstr *I) const {
   return nullptr;
 }
 
+CombinerObjective PPCInstrInfo::getCombinerObjective(unsigned Pattern) const {
+  switch (Pattern) {
+  case PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM:
+  case PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM:
+    return CombinerObjective::MustReduceDepth;
+  case PPCMachineCombinerPattern::REASSOC_XY_BCA:
+  case PPCMachineCombinerPattern::REASSOC_XY_BAC:
+    return CombinerObjective::MustReduceRegisterPressure;
+  default:
+    return TargetInstrInfo::getCombinerObjective(Pattern);
+  }
+}
+
 bool PPCInstrInfo::getMachineCombinerPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
+    MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
     bool DoRegPressureReduce) const {
   // Using the machine combiner in this way is potentially expensive, so
   // restrict to when aggressive optimizations are desired.
@@ -753,15 +766,15 @@ bool PPCInstrInfo::getMachineCombinerPatterns(
 }
 
 void PPCInstrInfo::genAlternativeCodeSequence(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
   switch (Pattern) {
-  case MachineCombinerPattern::REASSOC_XY_AMM_BMM:
-  case MachineCombinerPattern::REASSOC_XMM_AMM_BMM:
-  case MachineCombinerPattern::REASSOC_XY_BCA:
-  case MachineCombinerPattern::REASSOC_XY_BAC:
+  case PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM:
+  case PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM:
+  case PPCMachineCombinerPattern::REASSOC_XY_BCA:
+  case PPCMachineCombinerPattern::REASSOC_XY_BAC:
     reassociateFMA(Root, Pattern, InsInstrs, DelInstrs, InstrIdxForVirtReg);
     break;
   default:
@@ -773,7 +786,7 @@ void PPCInstrInfo::genAlternativeCodeSequence(
 }
 
 void PPCInstrInfo::reassociateFMA(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
@@ -790,8 +803,8 @@ void PPCInstrInfo::reassociateFMA(
   assert(Idx >= 0 && "Root must be a FMA instruction");
 
   bool IsILPReassociate =
-      (Pattern == MachineCombinerPattern::REASSOC_XY_AMM_BMM) ||
-      (Pattern == MachineCombinerPattern::REASSOC_XMM_AMM_BMM);
+      (Pattern == PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM) ||
+      (Pattern == PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM);
 
   uint16_t AddOpIdx = FMAOpIdxInfo[Idx][InfoArrayIdxAddOpIdx];
   uint16_t FirstMulOpIdx = FMAOpIdxInfo[Idx][InfoArrayIdxMULOpIdx];
@@ -801,18 +814,18 @@ void PPCInstrInfo::reassociateFMA(
   switch (Pattern) {
   default:
     llvm_unreachable("not recognized pattern!");
-  case MachineCombinerPattern::REASSOC_XY_AMM_BMM:
-  case MachineCombinerPattern::REASSOC_XMM_AMM_BMM:
+  case PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM:
+  case PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM:
     Prev = MRI.getUniqueVRegDef(Root.getOperand(AddOpIdx).getReg());
     Leaf = MRI.getUniqueVRegDef(Prev->getOperand(AddOpIdx).getReg());
     break;
-  case MachineCombinerPattern::REASSOC_XY_BAC: {
+  case PPCMachineCombinerPattern::REASSOC_XY_BAC: {
     Register MULReg =
         TRI->lookThruCopyLike(Root.getOperand(FirstMulOpIdx).getReg(), &MRI);
     Leaf = MRI.getVRegDef(MULReg);
     break;
   }
-  case MachineCombinerPattern::REASSOC_XY_BCA: {
+  case PPCMachineCombinerPattern::REASSOC_XY_BCA: {
     Register MULReg = TRI->lookThruCopyLike(
         Root.getOperand(FirstMulOpIdx + 1).getReg(), &MRI);
     Leaf = MRI.getVRegDef(MULReg);
@@ -853,10 +866,10 @@ void PPCInstrInfo::reassociateFMA(
   if (IsILPReassociate)
     GetFMAInstrInfo(*Prev, RegM21, RegM22, RegA21, KillM21, KillM22, KillA21);
 
-  if (Pattern == MachineCombinerPattern::REASSOC_XMM_AMM_BMM) {
+  if (Pattern == PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM) {
     GetFMAInstrInfo(*Leaf, RegM11, RegM12, RegA11, KillM11, KillM12, KillA11);
     GetOperandInfo(Leaf->getOperand(AddOpIdx), RegX, KillX);
-  } else if (Pattern == MachineCombinerPattern::REASSOC_XY_AMM_BMM) {
+  } else if (Pattern == PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM) {
     GetOperandInfo(Leaf->getOperand(1), RegX, KillX);
     GetOperandInfo(Leaf->getOperand(2), RegY, KillY);
   } else {
@@ -881,7 +894,7 @@ void PPCInstrInfo::reassociateFMA(
   }
 
   Register NewVRD = 0;
-  if (Pattern == MachineCombinerPattern::REASSOC_XMM_AMM_BMM) {
+  if (Pattern == PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM) {
     NewVRD = MRI.createVirtualRegister(RC);
     InstrIdxForVirtReg.insert(std::make_pair(NewVRD, 2));
   }
@@ -901,7 +914,7 @@ void PPCInstrInfo::reassociateFMA(
   switch (Pattern) {
   default:
     llvm_unreachable("not recognized pattern!");
-  case MachineCombinerPattern::REASSOC_XY_AMM_BMM: {
+  case PPCMachineCombinerPattern::REASSOC_XY_AMM_BMM: {
     // Create new instructions for insertion.
     MachineInstrBuilder MINewB =
         BuildMI(*MF, Prev->getDebugLoc(), get(FmaOp), NewVRB)
@@ -936,7 +949,7 @@ void PPCInstrInfo::reassociateFMA(
     InsInstrs.push_back(MINewC);
     break;
   }
-  case MachineCombinerPattern::REASSOC_XMM_AMM_BMM: {
+  case PPCMachineCombinerPattern::REASSOC_XMM_AMM_BMM: {
     assert(NewVRD && "new FMA register not created!");
     // Create new instructions for insertion.
     MachineInstrBuilder MINewA =
@@ -980,11 +993,11 @@ void PPCInstrInfo::reassociateFMA(
     InsInstrs.push_back(MINewC);
     break;
   }
-  case MachineCombinerPattern::REASSOC_XY_BAC:
-  case MachineCombinerPattern::REASSOC_XY_BCA: {
+  case PPCMachineCombinerPattern::REASSOC_XY_BAC:
+  case PPCMachineCombinerPattern::REASSOC_XY_BCA: {
     Register VarReg;
     bool KillVarReg = false;
-    if (Pattern == MachineCombinerPattern::REASSOC_XY_BCA) {
+    if (Pattern == PPCMachineCombinerPattern::REASSOC_XY_BCA) {
       VarReg = RegM31;
       KillVarReg = KillM31;
     } else {
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.h b/llvm/lib/Target/PowerPC/PPCInstrInfo.h
index 045932dc0d3ba..1e2687f92c61e 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.h
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.h
@@ -85,6 +85,19 @@ enum SpillOpcodeKey {
   SOK_LastOpcodeSpill // This must be last on the enum.
 };
 
+// PPC MachineCombiner patterns
+enum PPCMachineCombinerPattern : unsigned {
+  // These are patterns matched by the PowerPC to reassociate FMA chains.
+  REASSOC_XY_AMM_BMM = MachineCombinerPattern::TARGET_PATTERN_START,
+  REASSOC_XMM_AMM_BMM,
+
+  // These are patterns matched by the PowerPC to reassociate FMA and FSUB to
+  // reduce register pressure.
+  REASSOC_XY_BCA,
+  REASSOC_XY_BAC,
+
+};
+
 // Define list of load and store spill opcodes.
 #define NoInstr PPC::INSTRUCTION_LIST_END
 #define Pwr8LoadOpcodes                                                        \
@@ -224,7 +237,7 @@ class PPCInstrInfo : public PPCGenInstrInfo {
   ArrayRef<unsigned> getLoadOpcodesForSpillArray() const;
   unsigned getSpillIndex(const TargetRegisterClass *RC) const;
   int16_t getFMAOpIdxInfo(unsigned Opcode) const;
-  void reassociateFMA(MachineInstr &Root, MachineCombinerPattern Pattern,
+  void reassociateFMA(MachineInstr &Root, unsigned Pattern,
                       SmallVectorImpl<MachineInstr *> &InsInstrs,
                       SmallVectorImpl<MachineInstr *> &DelInstrs,
                       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const;
@@ -350,7 +363,7 @@ class PPCInstrInfo : public PPCGenInstrInfo {
   /// When getMachineCombinerPatterns() finds patterns, this function generates
   /// the instructions that could replace the original code sequence
   void genAlternativeCodeSequence(
-      MachineInstr &Root, MachineCombinerPattern Pattern,
+      MachineInstr &Root, unsigned Pattern,
       SmallVectorImpl<MachineInstr *> &InsInstrs,
       SmallVectorImpl<MachineInstr *> &DelInstrs,
       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const override;
@@ -358,15 +371,16 @@ class PPCInstrInfo : public PPCGenInstrInfo {
   /// Return true when there is potentially a faster code sequence for a fma
   /// chain ending in \p Root. All potential patterns are output in the \p
   /// P array.
-  bool getFMAPatterns(MachineInstr &Root,
-                      SmallVectorImpl<MachineCombinerPattern> &P,
+  bool getFMAPatterns(MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
                       bool DoRegPressureReduce) const;
 
+  CombinerObjective getCombinerObjective(unsigned Pattern) const override;
+
   /// Return true when there is potentially a faster code sequence
   /// for an instruction chain ending in <Root>. All potential patterns are
   /// output in the <Pattern> array.
   bool getMachineCombinerPatterns(MachineInstr &Root,
-                                  SmallVectorImpl<MachineCombinerPattern> &P,
+                                  SmallVectorImpl<unsigned> &Patterns,
                                   bool DoRegPressureReduce) const override;
 
   /// On PowerPC, we leverage machine combiner pass to reduce register pressure
@@ -380,7 +394,7 @@ class PPCInstrInfo : public PPCGenInstrInfo {
   /// Fixup the placeholders we put in genAlternativeCodeSequence() for
   /// MachineCombiner.
   void
-  finalizeInsInstrs(MachineInstr &Root, MachineCombinerPattern &P,
+  finalizeInsInstrs(MachineInstr &Root, unsigned &Pattern,
                     SmallVectorImpl<MachineInstr *> &InsInstrs) const override;
 
   bool isAssociativeAndCommutative(const MachineInstr &Inst,
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 84d754e3cbcf3..d78f5bd9dedf3 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -1560,7 +1560,7 @@ MachineTraceStrategy RISCVInstrInfo::getMachineCombinerTraceStrategy() const {
 }
 
 void RISCVInstrInfo::finalizeInsInstrs(
-    MachineInstr &Root, MachineCombinerPattern &P,
+    MachineInstr &Root, unsigned &Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs) const {
   int16_t FrmOpIdx =
       RISCV::getNamedOperandIdx(Root.getOpcode(), RISCV::OpName::frm);
@@ -1748,10 +1748,9 @@ static bool canCombineFPFusedMultiply(const MachineInstr &Root,
   return RISCV::hasEqualFRM(Root, *MI);
 }
 
-static bool
-getFPFusedMultiplyPatterns(MachineInstr &Root,
-                           SmallVectorImpl<MachineCombinerPattern> &Patterns,
-                           bool DoRegPressureReduce) {
+static bool getFPFusedMultiplyPatterns(MachineInstr &Root,
+                                       SmallVectorImpl<unsigned> &Patterns,
+                                       bool DoRegPressureReduce) {
   unsigned Opc = Root.getOpcode();
   bool IsFAdd = isFADD(Opc);
   if (!IsFAdd && !isFSUB(Opc))
@@ -1759,21 +1758,21 @@ getFPFusedMultiplyPatterns(MachineInstr &Root,
   bool Added = false;
   if (canCombineFPFusedMultiply(Root, Root.getOperand(1),
                                 DoRegPressureReduce)) {
-    Patterns.push_back(IsFAdd ? MachineCombinerPattern::FMADD_AX
-                              : MachineCombinerPattern::FMSUB);
+    Patterns.push_back(IsFAdd ? RISCVMachineCombinerPattern::FMADD_AX
+                              : RISCVMachineCombinerPattern::FMSUB);
     Added = true;
   }
   if (canCombineFPFusedMultiply(Root, Root.getOperand(2),
                                 DoRegPressureReduce)) {
-    Patterns.push_back(IsFAdd ? MachineCombinerPattern::FMADD_XA
-                              : MachineCombinerPattern::FNMSUB);
+    Patterns.push_back(IsFAdd ? RISCVMachineCombinerPattern::FMADD_XA
+                              : RISCVMachineCombinerPattern::FNMSUB);
     Added = true;
   }
   return Added;
 }
 
 static bool getFPPatterns(MachineInstr &Root,
-                          SmallVectorImpl<MachineCombinerPattern> &Patterns,
+                          SmallVectorImpl<unsigned> &Patterns,
                           bool DoRegPressureReduce) {
   return getFPFusedMultiplyPatterns(Root, Patterns, DoRegPressureReduce);
 }
@@ -1832,9 +1831,8 @@ static unsigned getSHXADDShiftAmount(unsigned Opc) {
 
 // Look for opportunities to combine (sh3add Z, (add X, (slli Y, 5))) into
 // (sh3add (sh2add Y, Z), X).
-static bool
-getSHXADDPatterns(const MachineInstr &Root,
-                  SmallVectorImpl<MachineCombinerPattern> &Patterns) {
+static bool getSHXADDPatterns(const MachineInstr &Root,
+                              SmallVectorImpl<unsigned> &Patterns) {
   unsigned ShiftAmt = getSHXADDShiftAmount(Root.getOpcode());
   if (!ShiftAmt)
     return false;
@@ -1847,19 +1845,31 @@ getSHXADDPatterns(const MachineInstr &Root,
 
   bool Found = false;
   if (canCombineShiftIntoShXAdd(MBB, AddMI->getOperand(1), ShiftAmt)) {
-    Patterns.push_back(MachineCombinerPattern::SHXADD_ADD_SLLI_OP1);
+    Patterns.push_back(RISCVMachineCombinerPattern::SHXADD_ADD_SLLI_OP1);
     Found = true;
   }
   if (canCombineShiftIntoShXAdd(MBB, AddMI->getOperand(2), ShiftAmt)) {
-    Patterns.push_back(MachineCombinerPattern::SHXADD_ADD_SLLI_OP2);
+    Patterns.push_back(RISCVMachineCombinerPattern::SHXADD_ADD_SLLI_OP2);
     Found = true;
   }
 
   return Found;
 }
 
+CombinerObjective RISCVInstrInfo::getCombinerObjective(unsigned Pattern) const {
+  switch (Pattern) {
+  case RISCVMachineCombinerPattern::FMADD_AX:
+  case RISCVMachineCombinerPattern::FMADD_XA:
+  case RISCVMachineCombinerPattern::FMSUB:
+  case RISCVMachineCombinerPattern::FNMSUB:
+    return CombinerObjective::MustReduceDepth;
+  default:
+    return TargetInstrInfo::getCombinerObjective(Pattern);
+  }
+}
+
 bool RISCVInstrInfo::getMachineCombinerPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
+    MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
     bool DoRegPressureReduce) const {
 
   if (getFPPatterns(Root, Patterns, DoRegPressureReduce))
@@ -1872,8 +1882,7 @@ bool RISCVInstrInfo::getMachineCombinerPatterns(
                                                      DoRegPressureReduce);
 }
 
-static unsigned getFPFusedMultiplyOpcode(unsigned RootOpc,
-                                         MachineCombinerPattern Pattern) {
+static unsigned getFPFusedMultiplyOpcode(unsigned RootOpc, unsigned Pattern) {
   switch (RootOpc) {
   default:
     llvm_unreachable("Unexpected opcode");
@@ -1884,32 +1893,32 @@ static unsigned getFPFusedMultiplyOpcode(unsigned RootOpc,
   case RISCV::FADD_D:
     return RISCV::FMADD_D;
   case RISCV::FSUB_H:
-    return Pattern == MachineCombinerPattern::FMSUB ? RISCV::FMSUB_H
-                                                    : RISCV::FNMSUB_H;
+    return Pattern == RISCVMachineCombinerPattern::FMSUB ? RISCV::FMSUB_H
+                                                         : RISCV::FNMSUB_H;
   case RISCV::FSUB_S:
-    return Pattern == MachineCombinerPattern::FMSUB ? RISCV::FMSUB_S
-                                                    : RISCV::FNMSUB_S;
+    return Pattern == RISCVMachineCombinerPattern::FMSUB ? RISCV::FMSUB_S
+                                                         : RISCV::FNMSUB_S;
   case RISCV::FSUB_D:
-    return Pattern == MachineCombinerPattern::FMSUB ? RISCV::FMSUB_D
-                                                    : RISCV::FNMSUB_D;
+    return Pattern == RISCVMachineCombinerPattern::FMSUB ? RISCV::FMSUB_D
+                                                         : RISCV::FNMSUB_D;
   }
 }
 
-static unsigned getAddendOperandIdx(MachineCombinerPattern Pattern) {
+static unsigned getAddendOperandIdx(unsigned Pattern) {
   switch (Pattern) {
   default:
     llvm_unreachable("Unexpected pattern");
-  case MachineCombinerPattern::FMADD_AX:
-  case MachineCombinerPattern::FMSUB:
+  case RISCVMachineCombinerPattern::FMADD_AX:
+  case RISCVMachineCombinerPattern::FMSUB:
     return 2;
-  case MachineCombinerPattern::FMADD_XA:
-  case MachineCombinerPattern::FNMSUB:
+  case RISCVMachineCombinerPattern::FMADD_XA:
+  case RISCVMachineCombinerPattern::FNMSUB:
     return 1;
   }
 }
 
 static void combineFPFusedMultiply(MachineInstr &Root, MachineInstr &Prev,
-                                   MachineCombinerPattern Pattern,
+                                   unsigned Pattern,
                                    SmallVectorImpl<MachineInstr *> &InsInstrs,
                                    SmallVectorImpl<MachineInstr *> &DelInstrs) {
   MachineFunction *MF = Root.getMF();
@@ -2013,7 +2022,7 @@ genShXAddAddShift(MachineInstr &Root, unsigned AddOpIdx,
 }
 
 void RISCVInstrInfo::genAlternativeCodeSequence(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
@@ -2023,22 +2032,22 @@ void RISCVInstrInfo::genAlternativeCodeSequence(
     TargetInstrInfo::genAlternativeCodeSequence(Root, Pattern, InsInstrs,
                                                 DelInstrs, InstrIdxForVirtReg);
     return;
-  case MachineCombinerPattern::FMADD_AX:
-  case MachineCombinerPattern::FMSUB: {
+  case RISCVMachineCombinerPattern::FMADD_AX:
+  case RISCVMachineCombinerPattern::FMSUB: {
     MachineInstr &Prev = *MRI.getVRegDef(Root.getOperand(1).getReg());
     combineFPFusedMultiply(Root, Prev, Pattern, InsInstrs, DelInstrs);
     return;
   }
-  case MachineCombinerPattern::FMADD_XA:
-  case MachineCombinerPattern::FNMSUB: {
+  case RISCVMachineCombinerPattern::FMADD_XA:
+  case RISCVMachineCombinerPattern::FNMSUB: {
     MachineInstr &Prev = *MRI.getVRegDef(Root.getOperand(2).getReg());
     combineFPFusedMultiply(Root, Prev, Pattern, InsInstrs, DelInstrs);
     return;
   }
-  case MachineCombinerPattern::SHXADD_ADD_SLLI_OP1:
+  case RISCVMachineCombinerPattern::SHXADD_ADD_SLLI_OP1:
     genShXAddAddShift(Root, 1, InsInstrs, DelInstrs, InstrIdxForVirtReg);
     return;
-  case MachineCombinerPattern::SHXADD_ADD_SLLI_OP2:
+  case RISCVMachineCombinerPattern::SHXADD_ADD_SLLI_OP2:
     genShXAddAddShift(Root, 2, InsInstrs, DelInstrs, InstrIdxForVirtReg);
     return;
   }
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.h b/llvm/lib/Target/RISCV/RISCVInstrInfo.h
index 81d9c9db783c0..70fe7da85be0e 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.h
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.h
@@ -49,6 +49,16 @@ unsigned getBrCond(CondCode CC);
 
 } // end of namespace RISCVCC
 
+// RISCV MachineCombiner patterns
+enum RISCVMachineCombinerPattern : unsigned {
+  FMADD_AX = MachineCombinerPattern::TARGET_PATTERN_START,
+  FMADD_XA,
+  FMSUB,
+  FNMSUB,
+  SHXADD_ADD_SLLI_OP1,
+  SHXADD_ADD_SLLI_OP2,
+};
+
 class RISCVInstrInfo : public RISCVGenInstrInfo {
 
 public:
@@ -240,17 +250,18 @@ class RISCVInstrInfo : public RISCVGenInstrInfo {
 
   MachineTraceStrategy getMachineCombinerTraceStrategy() const override;
 
-  bool
-  getMachineCombinerPatterns(MachineInstr &Root,
-                             SmallVectorImpl<MachineCombinerPattern> &Patterns,
-                             bool DoRegPressureReduce) const override;
+  CombinerObjective getCombinerObjective(unsigned Pattern) const override;
+
+  bool getMachineCombinerPatterns(MachineInstr &Root,
+                                  SmallVectorImpl<unsigned> &Patterns,
+                                  bool DoRegPressureReduce) const override;
 
   void
-  finalizeInsInstrs(MachineInstr &Root, MachineCombinerPattern &P,
+  finalizeInsInstrs(MachineInstr &Root, unsigned &Pattern,
                     SmallVectorImpl<MachineInstr *> &InsInstrs) const override;
 
   void genAlternativeCodeSequence(
-      MachineInstr &Root, MachineCombinerPattern Pattern,
+      MachineInstr &Root, unsigned Pattern,
       SmallVectorImpl<MachineInstr *> &InsInstrs,
       SmallVectorImpl<MachineInstr *> &DelInstrs,
       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const override;
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index a5b2e4895eded..510b08f9901a2 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -10578,7 +10578,7 @@ void X86InstrInfo::buildClearRegister(Register Reg, MachineBasicBlock &MBB,
 }
 
 bool X86InstrInfo::getMachineCombinerPatterns(
-    MachineInstr &Root, SmallVectorImpl<MachineCombinerPattern> &Patterns,
+    MachineInstr &Root, SmallVectorImpl<unsigned> &Patterns,
     bool DoRegPressureReduce) const {
   unsigned Opc = Root.getOpcode();
   switch (Opc) {
@@ -10587,7 +10587,7 @@ bool X86InstrInfo::getMachineCombinerPatterns(
   case X86::VPDPWSSDYrr:
   case X86::VPDPWSSDYrm: {
     if (!Subtarget.hasFastDPWSSD()) {
-      Patterns.push_back(MachineCombinerPattern::DPWSSD);
+      Patterns.push_back(X86MachineCombinerPattern::DPWSSD);
       return true;
     }
     break;
@@ -10599,8 +10599,8 @@ bool X86InstrInfo::getMachineCombinerPatterns(
   case X86::VPDPWSSDZr:
   case X86::VPDPWSSDZm: {
    if (Subtarget.hasBWI() && !Subtarget.hasFastDPWSSD()) {
-      Patterns.push_back(MachineCombinerPattern::DPWSSD);
-      return true;
+     Patterns.push_back(X86MachineCombinerPattern::DPWSSD);
+     return true;
     }
     break;
   }
@@ -10700,7 +10700,7 @@ genAlternativeDpCodeSequence(MachineInstr &Root, const TargetInstrInfo &TII,
 }
 
 void X86InstrInfo::genAlternativeCodeSequence(
-    MachineInstr &Root, MachineCombinerPattern Pattern,
+    MachineInstr &Root, unsigned Pattern,
     SmallVectorImpl<MachineInstr *> &InsInstrs,
     SmallVectorImpl<MachineInstr *> &DelInstrs,
     DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const {
@@ -10710,7 +10710,7 @@ void X86InstrInfo::genAlternativeCodeSequence(
     TargetInstrInfo::genAlternativeCodeSequence(Root, Pattern, InsInstrs,
                                                 DelInstrs, InstrIdxForVirtReg);
     return;
-  case MachineCombinerPattern::DPWSSD:
+  case X86MachineCombinerPattern::DPWSSD:
     genAlternativeDpCodeSequence(Root, *this, InsInstrs, DelInstrs,
                                  InstrIdxForVirtReg);
     return;
diff --git a/llvm/lib/Target/X86/X86InstrInfo.h b/llvm/lib/Target/X86/X86InstrInfo.h
index e719be0caf3ee..5407ede69a91c 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.h
+++ b/llvm/lib/Target/X86/X86InstrInfo.h
@@ -26,6 +26,12 @@
 namespace llvm {
 class X86Subtarget;
 
+// X86 MachineCombiner patterns
+enum X86MachineCombinerPattern : unsigned {
+  // X86 VNNI
+  DPWSSD = MachineCombinerPattern::TARGET_PATTERN_START,
+};
+
 namespace X86 {
 
 enum AsmComments {
@@ -607,16 +613,15 @@ class X86InstrInfo final : public X86GenInstrInfo {
   std::optional<DestSourcePair>
   isCopyInstrImpl(const MachineInstr &MI) const override;
 
-  bool
-  getMachineCombinerPatterns(MachineInstr &Root,
-                             SmallVectorImpl<MachineCombinerPattern> &Patterns,
-                             bool DoRegPressureReduce) const override;
+  bool getMachineCombinerPatterns(MachineInstr &Root,
+                                  SmallVectorImpl<unsigned> &Patterns,
+                                  bool DoRegPressureReduce) const override;
 
   /// When getMachineCombinerPatterns() finds potential patterns,
   /// this function generates the instructions that could replace the
   /// original code sequence.
   void genAlternativeCodeSequence(
-      MachineInstr &Root, MachineCombinerPattern Pattern,
+      MachineInstr &Root, unsigned Pattern,
       SmallVectorImpl<MachineInstr *> &InsInstrs,
       SmallVectorImpl<MachineInstr *> &DelInstrs,
       DenseMap<unsigned, unsigned> &InstrIdxForVirtReg) const override;

From efb8cc5ddb03897795dc153a03d0c1548c8ee4a7 Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97@outlook.com>
Date: Thu, 11 Apr 2024 12:27:18 +0800
Subject: [PATCH 096/886] [NewPM] Fix print-changed-dot-cfg failure (#88351)

Fix failure in #80946.
---
 llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir b/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
index 340ece93aa02b..40603630c6134 100644
--- a/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
+++ b/llvm/test/Other/ChangePrinters/DotCfg/print-changed-dot-cfg.mir
@@ -1,7 +1,7 @@
 # REQUIRES: x86-registered-target
 # Simple functionality check.
 # RUN: rm -rf %t && mkdir -p %t
-# RUN: llc -filetype=null -print-changed=dot-cfg -passes=no-op-machine-function -dot-cfg-dir=%t %s
+# RUN: llc -mtriple=x86_64-pc-linux-gnu -filetype=null -print-changed=dot-cfg -passes=no-op-machine-function -dot-cfg-dir=%t %s
 # RUN: ls %t/*.pdf %t/passes.html | count 3
 
 ---

From 3197f9d8b0efc3efdc531421bd11c16305d9b1ff Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Thu, 11 Apr 2024 12:48:52 +0800
Subject: [PATCH 097/886] [InstSimplify] Make sure the simplified value doesn't
 generate poison in threadBinOpOverSelect (#87075)

Alive2: https://alive2.llvm.org/ce/z/y_Jmdn
Fix https://github.com/llvm/llvm-project/issues/87042.
---
 llvm/lib/Analysis/InstructionSimplify.cpp    |  3 +-
 llvm/test/Transforms/InstSimplify/pr87042.ll | 42 ++++++++++++++++++++
 2 files changed, 44 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/Transforms/InstSimplify/pr87042.ll

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 9ff3faff79902..3c943a09a9c23 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -440,7 +440,8 @@ static Value *threadBinOpOverSelect(Instruction::BinaryOps Opcode, Value *LHS,
     // Check that the simplified value has the form "X op Y" where "op" is the
     // same as the original operation.
     Instruction *Simplified = dyn_cast<Instruction>(FV ? FV : TV);
-    if (Simplified && Simplified->getOpcode() == unsigned(Opcode)) {
+    if (Simplified && Simplified->getOpcode() == unsigned(Opcode) &&
+        !Simplified->hasPoisonGeneratingFlags()) {
       // The value that didn't simplify is "UnsimplifiedLHS op UnsimplifiedRHS".
       // We already know that "op" is the same as for the simplified value.  See
       // if the operands match too.  If so, return the simplified value.
diff --git a/llvm/test/Transforms/InstSimplify/pr87042.ll b/llvm/test/Transforms/InstSimplify/pr87042.ll
new file mode 100644
index 0000000000000..800d27c9e6504
--- /dev/null
+++ b/llvm/test/Transforms/InstSimplify/pr87042.ll
@@ -0,0 +1,42 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=instsimplify -S | FileCheck %s
+
+; %or2 cannot be folded into %or1 because %or1 has disjoint.
+; TODO: Can we move the logic into InstCombine and drop the disjoint flag?
+define i64 @test(i1 %cond, i64 %x) {
+; CHECK-LABEL: define i64 @test(
+; CHECK-SAME: i1 [[COND:%.*]], i64 [[X:%.*]]) {
+; CHECK-NEXT:    [[OR1:%.*]] = or disjoint i64 [[X]], 7
+; CHECK-NEXT:    [[SEL1:%.*]] = select i1 [[COND]], i64 [[OR1]], i64 [[X]]
+; CHECK-NEXT:    [[OR2:%.*]] = or i64 [[SEL1]], 7
+; CHECK-NEXT:    ret i64 [[OR2]]
+;
+  %or1 = or disjoint i64 %x, 7
+  %sel1 = select i1 %cond, i64 %or1, i64 %x
+  %or2 = or i64 %sel1, 7
+  ret i64 %or2
+}
+
+define i64 @pr87042(i64 %x) {
+; CHECK-LABEL: define i64 @pr87042(
+; CHECK-SAME: i64 [[X:%.*]]) {
+; CHECK-NEXT:    [[AND1:%.*]] = and i64 [[X]], 65535
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp eq i64 [[AND1]], 0
+; CHECK-NEXT:    [[OR1:%.*]] = or disjoint i64 [[X]], 7
+; CHECK-NEXT:    [[SEL1:%.*]] = select i1 [[CMP1]], i64 [[OR1]], i64 [[X]]
+; CHECK-NEXT:    [[AND2:%.*]] = and i64 [[SEL1]], 16776960
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp eq i64 [[AND2]], 0
+; CHECK-NEXT:    [[OR2:%.*]] = or i64 [[SEL1]], 7
+; CHECK-NEXT:    [[SEL2:%.*]] = select i1 [[CMP2]], i64 [[OR2]], i64 [[SEL1]]
+; CHECK-NEXT:    ret i64 [[SEL2]]
+;
+  %and1 = and i64 %x, 65535
+  %cmp1 = icmp eq i64 %and1, 0
+  %or1 = or disjoint i64 %x, 7
+  %sel1 = select i1 %cmp1, i64 %or1, i64 %x
+  %and2 = and i64 %sel1, 16776960
+  %cmp2 = icmp eq i64 %and2, 0
+  %or2 = or i64 %sel1, 7
+  %sel2 = select i1 %cmp2, i64 %or2, i64 %sel1
+  ret i64 %sel2
+}

From 2bede6873dbe7021b306d3e5bec59d0fba2dd26c Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Wed, 10 Apr 2024 22:03:20 -0700
Subject: [PATCH 098/886] [memprof] Rename RawMemProfReader.{cpp,h} to
 MemProfReader.{cpp,h} (NFC) (#88200)

This patch renames RawMemProfReader.{cpp,h} to MemProfReader.{cpp,h},
respectively.  Also, it re-creates RawMemProfReader.h just to include
MemProfReader.h for compatibility with out-of-tree users.
---
 llvm/include/llvm/ProfileData/MemProfReader.h | 214 ++++++++++++++++++
 .../llvm/ProfileData/RawMemProfReader.h       | 207 +----------------
 llvm/lib/ProfileData/CMakeLists.txt           |   2 +-
 ...RawMemProfReader.cpp => MemProfReader.cpp} |   6 +-
 llvm/tools/llvm-profdata/llvm-profdata.cpp    |   2 +-
 llvm/unittests/ProfileData/MemProfTest.cpp    |   2 +-
 .../secondary/llvm/lib/ProfileData/BUILD.gn   |   2 +-
 7 files changed, 227 insertions(+), 208 deletions(-)
 create mode 100644 llvm/include/llvm/ProfileData/MemProfReader.h
 rename llvm/lib/ProfileData/{RawMemProfReader.cpp => MemProfReader.cpp} (99%)

diff --git a/llvm/include/llvm/ProfileData/MemProfReader.h b/llvm/include/llvm/ProfileData/MemProfReader.h
new file mode 100644
index 0000000000000..89f49a20a6089
--- /dev/null
+++ b/llvm/include/llvm/ProfileData/MemProfReader.h
@@ -0,0 +1,214 @@
+//===- MemProfReader.h - Instrumented memory profiling reader ---*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains support for reading MemProf profiling data.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_PROFILEDATA_MEMPROFREADER_H_
+#define LLVM_PROFILEDATA_MEMPROFREADER_H_
+
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/MapVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
+#include "llvm/DebugInfo/Symbolize/Symbolize.h"
+#include "llvm/IR/GlobalValue.h"
+#include "llvm/Object/Binary.h"
+#include "llvm/Object/ObjectFile.h"
+#include "llvm/ProfileData/InstrProfReader.h"
+#include "llvm/ProfileData/MemProf.h"
+#include "llvm/ProfileData/MemProfData.inc"
+#include "llvm/Support/Error.h"
+#include "llvm/Support/MemoryBuffer.h"
+
+#include <functional>
+
+namespace llvm {
+namespace memprof {
+// A class for memprof profile data populated directly from external
+// sources.
+class MemProfReader {
+public:
+  // The MemProfReader only holds memory profile information.
+  InstrProfKind getProfileKind() const { return InstrProfKind::MemProf; }
+
+  using GuidMemProfRecordPair = std::pair<GlobalValue::GUID, MemProfRecord>;
+  using Iterator = InstrProfIterator<GuidMemProfRecordPair, MemProfReader>;
+  Iterator end() { return Iterator(); }
+  Iterator begin() {
+    Iter = FunctionProfileData.begin();
+    return Iterator(this);
+  }
+
+  // Return a const reference to the internal Id to Frame mappings.
+  const llvm::DenseMap<FrameId, Frame> &getFrameMapping() const {
+    return IdToFrame;
+  }
+
+  // Return a const reference to the internal function profile data.
+  const llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> &
+  getProfileData() const {
+    return FunctionProfileData;
+  }
+
+  virtual Error
+  readNextRecord(GuidMemProfRecordPair &GuidRecord,
+                 std::function<const Frame(const FrameId)> Callback = nullptr) {
+    if (FunctionProfileData.empty())
+      return make_error<InstrProfError>(instrprof_error::empty_raw_profile);
+
+    if (Iter == FunctionProfileData.end())
+      return make_error<InstrProfError>(instrprof_error::eof);
+
+    if (Callback == nullptr)
+      Callback =
+          std::bind(&MemProfReader::idToFrame, this, std::placeholders::_1);
+
+    const IndexedMemProfRecord &IndexedRecord = Iter->second;
+    GuidRecord = {Iter->first, MemProfRecord(IndexedRecord, Callback)};
+    Iter++;
+    return Error::success();
+  }
+
+  // Allow default construction for derived classes which can populate the
+  // contents after construction.
+  MemProfReader() = default;
+  virtual ~MemProfReader() = default;
+
+  // Initialize the MemProfReader with the frame mappings and profile contents.
+  MemProfReader(
+      llvm::DenseMap<FrameId, Frame> FrameIdMap,
+      llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData)
+      : IdToFrame(std::move(FrameIdMap)),
+        FunctionProfileData(std::move(ProfData)) {}
+
+protected:
+  // A helper method to extract the frame from the IdToFrame map.
+  const Frame &idToFrame(const FrameId Id) const {
+    auto It = IdToFrame.find(Id);
+    assert(It != IdToFrame.end() && "Id not found in map.");
+    return It->getSecond();
+  }
+  // A mapping from FrameId (a hash of the contents) to the frame.
+  llvm::DenseMap<FrameId, Frame> IdToFrame;
+  // A mapping from function GUID, hash of the canonical function symbol to the
+  // memprof profile data for that function, i.e allocation and callsite info.
+  llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> FunctionProfileData;
+  // An iterator to the internal function profile data structure.
+  llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord>::iterator Iter;
+};
+
+// Map from id (recorded from sanitizer stack depot) to virtual addresses for
+// each program counter address in the callstack.
+using CallStackMap = llvm::DenseMap<uint64_t, llvm::SmallVector<uint64_t>>;
+
+// Specializes the MemProfReader class to populate the contents from raw binary
+// memprof profiles from instrumentation based profiling.
+class RawMemProfReader final : public MemProfReader {
+public:
+  RawMemProfReader(const RawMemProfReader &) = delete;
+  RawMemProfReader &operator=(const RawMemProfReader &) = delete;
+  virtual ~RawMemProfReader() override = default;
+
+  // Prints the contents of the profile in YAML format.
+  void printYAML(raw_ostream &OS);
+
+  // Return true if the \p DataBuffer starts with magic bytes indicating it is
+  // a raw binary memprof profile.
+  static bool hasFormat(const MemoryBuffer &DataBuffer);
+  // Return true if the file at \p Path starts with magic bytes indicating it is
+  // a raw binary memprof profile.
+  static bool hasFormat(const StringRef Path);
+
+  // Create a RawMemProfReader after sanity checking the contents of the file at
+  // \p Path or the \p Buffer. The binary from which the profile has been
+  // collected is specified via a path in \p ProfiledBinary.
+  static Expected<std::unique_ptr<RawMemProfReader>>
+  create(const Twine &Path, StringRef ProfiledBinary, bool KeepName = false);
+  static Expected<std::unique_ptr<RawMemProfReader>>
+  create(std::unique_ptr<MemoryBuffer> Buffer, StringRef ProfiledBinary,
+         bool KeepName = false);
+
+  // Returns a list of build ids recorded in the segment information.
+  static std::vector<std::string> peekBuildIds(MemoryBuffer *DataBuffer);
+
+  virtual Error
+  readNextRecord(GuidMemProfRecordPair &GuidRecord,
+                 std::function<const Frame(const FrameId)> Callback) override;
+
+  // Constructor for unittests only.
+  RawMemProfReader(std::unique_ptr<llvm::symbolize::SymbolizableModule> Sym,
+                   llvm::SmallVectorImpl<SegmentEntry> &Seg,
+                   llvm::MapVector<uint64_t, MemInfoBlock> &Prof,
+                   CallStackMap &SM, bool KeepName = false)
+      : SegmentInfo(Seg.begin(), Seg.end()), CallstackProfileData(Prof),
+        StackMap(SM), KeepSymbolName(KeepName) {
+    // We don't call initialize here since there is no raw profile to read. The
+    // test should pass in the raw profile as structured data.
+
+    // If there is an error here then the mock symbolizer has not been
+    // initialized properly.
+    if (Error E = symbolizeAndFilterStackFrames(std::move(Sym)))
+      report_fatal_error(std::move(E));
+    if (Error E = mapRawProfileToRecords())
+      report_fatal_error(std::move(E));
+  }
+
+private:
+  RawMemProfReader(object::OwningBinary<object::Binary> &&Bin, bool KeepName)
+      : Binary(std::move(Bin)), KeepSymbolName(KeepName) {}
+  // Initializes the RawMemProfReader with the contents in `DataBuffer`.
+  Error initialize(std::unique_ptr<MemoryBuffer> DataBuffer);
+  // Read and parse the contents of the `DataBuffer` as a binary format profile.
+  Error readRawProfile(std::unique_ptr<MemoryBuffer> DataBuffer);
+  // Initialize the segment mapping information for symbolization.
+  Error setupForSymbolization();
+  // Symbolize and cache all the virtual addresses we encounter in the
+  // callstacks from the raw profile. Also prune callstack frames which we can't
+  // symbolize or those that belong to the runtime. For profile entries where
+  // the entire callstack is pruned, we drop the entry from the profile.
+  Error symbolizeAndFilterStackFrames(
+      std::unique_ptr<llvm::symbolize::SymbolizableModule> Symbolizer);
+  // Construct memprof records for each function and store it in the
+  // `FunctionProfileData` map. A function may have allocation profile data or
+  // callsite data or both.
+  Error mapRawProfileToRecords();
+
+  object::SectionedAddress getModuleOffset(uint64_t VirtualAddress);
+
+  // The profiled binary.
+  object::OwningBinary<object::Binary> Binary;
+  // The preferred load address of the executable segment.
+  uint64_t PreferredTextSegmentAddress = 0;
+  // The base address of the text segment in the process during profiling.
+  uint64_t ProfiledTextSegmentStart = 0;
+  // The limit address of the text segment in the process during profiling.
+  uint64_t ProfiledTextSegmentEnd = 0;
+
+  // The memory mapped segment information for all executable segments in the
+  // profiled binary (filtered from the raw profile using the build id).
+  llvm::SmallVector<SegmentEntry, 2> SegmentInfo;
+
+  // A map from callstack id (same as key in CallStackMap below) to the heap
+  // information recorded for that allocation context.
+  llvm::MapVector<uint64_t, MemInfoBlock> CallstackProfileData;
+  CallStackMap StackMap;
+
+  // Cached symbolization from PC to Frame.
+  llvm::DenseMap<uint64_t, llvm::SmallVector<FrameId>> SymbolizedFrame;
+
+  // Whether to keep the symbol name for each frame after hashing.
+  bool KeepSymbolName = false;
+  // A mapping of the hash to symbol name, only used if KeepSymbolName is true.
+  llvm::DenseMap<uint64_t, std::string> GuidToSymbolName;
+};
+} // namespace memprof
+} // namespace llvm
+
+#endif // LLVM_PROFILEDATA_MEMPROFREADER_H_
diff --git a/llvm/include/llvm/ProfileData/RawMemProfReader.h b/llvm/include/llvm/ProfileData/RawMemProfReader.h
index 6aa5caec65f79..5e06f26fffdc2 100644
--- a/llvm/include/llvm/ProfileData/RawMemProfReader.h
+++ b/llvm/include/llvm/ProfileData/RawMemProfReader.h
@@ -1,6 +1,4 @@
-#ifndef LLVM_PROFILEDATA_RAWMEMPROFREADER_H_
-#define LLVM_PROFILEDATA_RAWMEMPROFREADER_H_
-//===- MemProfReader.h - Instrumented memory profiling reader ---*- C++ -*-===//
+//===- RawMemProfReader.h - Instrumented memory profiling reader *- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -8,207 +6,14 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file contains support for reading MemProf profiling data.
+// This file just includes MemProfReader.h for compatibility with
+// out-of-tree users.
 //
 //===----------------------------------------------------------------------===//
 
-#include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/MapVector.h"
-#include "llvm/ADT/StringRef.h"
-#include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
-#include "llvm/DebugInfo/Symbolize/Symbolize.h"
-#include "llvm/IR/GlobalValue.h"
-#include "llvm/Object/Binary.h"
-#include "llvm/Object/ObjectFile.h"
-#include "llvm/ProfileData/InstrProfReader.h"
-#include "llvm/ProfileData/MemProf.h"
-#include "llvm/ProfileData/MemProfData.inc"
-#include "llvm/Support/Error.h"
-#include "llvm/Support/MemoryBuffer.h"
-
-#include <functional>
-
-namespace llvm {
-namespace memprof {
-// A class for memprof profile data populated directly from external
-// sources.
-// TODO: Rename this file to MemProfReader.h to better reflect the contents.
-class MemProfReader {
-public:
-  // The MemProfReader only holds memory profile information.
-  InstrProfKind getProfileKind() const { return InstrProfKind::MemProf; }
-
-  using GuidMemProfRecordPair = std::pair<GlobalValue::GUID, MemProfRecord>;
-  using Iterator = InstrProfIterator<GuidMemProfRecordPair, MemProfReader>;
-  Iterator end() { return Iterator(); }
-  Iterator begin() {
-    Iter = FunctionProfileData.begin();
-    return Iterator(this);
-  }
-
-  // Return a const reference to the internal Id to Frame mappings.
-  const llvm::DenseMap<FrameId, Frame> &getFrameMapping() const {
-    return IdToFrame;
-  }
-
-  // Return a const reference to the internal function profile data.
-  const llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> &
-  getProfileData() const {
-    return FunctionProfileData;
-  }
-
-  virtual Error
-  readNextRecord(GuidMemProfRecordPair &GuidRecord,
-                 std::function<const Frame(const FrameId)> Callback = nullptr) {
-    if (FunctionProfileData.empty())
-      return make_error<InstrProfError>(instrprof_error::empty_raw_profile);
-
-    if (Iter == FunctionProfileData.end())
-      return make_error<InstrProfError>(instrprof_error::eof);
-
-    if (Callback == nullptr)
-      Callback =
-          std::bind(&MemProfReader::idToFrame, this, std::placeholders::_1);
-
-    const IndexedMemProfRecord &IndexedRecord = Iter->second;
-    GuidRecord = {Iter->first, MemProfRecord(IndexedRecord, Callback)};
-    Iter++;
-    return Error::success();
-  }
-
-  // Allow default construction for derived classes which can populate the
-  // contents after construction.
-  MemProfReader() = default;
-  virtual ~MemProfReader() = default;
-
-  // Initialize the MemProfReader with the frame mappings and profile contents.
-  MemProfReader(
-      llvm::DenseMap<FrameId, Frame> FrameIdMap,
-      llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData)
-      : IdToFrame(std::move(FrameIdMap)),
-        FunctionProfileData(std::move(ProfData)) {}
-
-protected:
-  // A helper method to extract the frame from the IdToFrame map.
-  const Frame &idToFrame(const FrameId Id) const {
-    auto It = IdToFrame.find(Id);
-    assert(It != IdToFrame.end() && "Id not found in map.");
-    return It->getSecond();
-  }
-  // A mapping from FrameId (a hash of the contents) to the frame.
-  llvm::DenseMap<FrameId, Frame> IdToFrame;
-  // A mapping from function GUID, hash of the canonical function symbol to the
-  // memprof profile data for that function, i.e allocation and callsite info.
-  llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> FunctionProfileData;
-  // An iterator to the internal function profile data structure.
-  llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord>::iterator Iter;
-};
-
-// Map from id (recorded from sanitizer stack depot) to virtual addresses for
-// each program counter address in the callstack.
-using CallStackMap = llvm::DenseMap<uint64_t, llvm::SmallVector<uint64_t>>;
-
-// Specializes the MemProfReader class to populate the contents from raw binary
-// memprof profiles from instrumentation based profiling.
-class RawMemProfReader final : public MemProfReader {
-public:
-  RawMemProfReader(const RawMemProfReader &) = delete;
-  RawMemProfReader &operator=(const RawMemProfReader &) = delete;
-  virtual ~RawMemProfReader() override = default;
-
-  // Prints the contents of the profile in YAML format.
-  void printYAML(raw_ostream &OS);
-
-  // Return true if the \p DataBuffer starts with magic bytes indicating it is
-  // a raw binary memprof profile.
-  static bool hasFormat(const MemoryBuffer &DataBuffer);
-  // Return true if the file at \p Path starts with magic bytes indicating it is
-  // a raw binary memprof profile.
-  static bool hasFormat(const StringRef Path);
-
-  // Create a RawMemProfReader after sanity checking the contents of the file at
-  // \p Path or the \p Buffer. The binary from which the profile has been
-  // collected is specified via a path in \p ProfiledBinary.
-  static Expected<std::unique_ptr<RawMemProfReader>>
-  create(const Twine &Path, StringRef ProfiledBinary, bool KeepName = false);
-  static Expected<std::unique_ptr<RawMemProfReader>>
-  create(std::unique_ptr<MemoryBuffer> Buffer, StringRef ProfiledBinary,
-         bool KeepName = false);
-
-  // Returns a list of build ids recorded in the segment information.
-  static std::vector<std::string> peekBuildIds(MemoryBuffer *DataBuffer);
-
-  virtual Error
-  readNextRecord(GuidMemProfRecordPair &GuidRecord,
-                 std::function<const Frame(const FrameId)> Callback) override;
-
-  // Constructor for unittests only.
-  RawMemProfReader(std::unique_ptr<llvm::symbolize::SymbolizableModule> Sym,
-                   llvm::SmallVectorImpl<SegmentEntry> &Seg,
-                   llvm::MapVector<uint64_t, MemInfoBlock> &Prof,
-                   CallStackMap &SM, bool KeepName = false)
-      : SegmentInfo(Seg.begin(), Seg.end()), CallstackProfileData(Prof),
-        StackMap(SM), KeepSymbolName(KeepName) {
-    // We don't call initialize here since there is no raw profile to read. The
-    // test should pass in the raw profile as structured data.
-
-    // If there is an error here then the mock symbolizer has not been
-    // initialized properly.
-    if (Error E = symbolizeAndFilterStackFrames(std::move(Sym)))
-      report_fatal_error(std::move(E));
-    if (Error E = mapRawProfileToRecords())
-      report_fatal_error(std::move(E));
-  }
-
-private:
-  RawMemProfReader(object::OwningBinary<object::Binary> &&Bin, bool KeepName)
-      : Binary(std::move(Bin)), KeepSymbolName(KeepName) {}
-  // Initializes the RawMemProfReader with the contents in `DataBuffer`.
-  Error initialize(std::unique_ptr<MemoryBuffer> DataBuffer);
-  // Read and parse the contents of the `DataBuffer` as a binary format profile.
-  Error readRawProfile(std::unique_ptr<MemoryBuffer> DataBuffer);
-  // Initialize the segment mapping information for symbolization.
-  Error setupForSymbolization();
-  // Symbolize and cache all the virtual addresses we encounter in the
-  // callstacks from the raw profile. Also prune callstack frames which we can't
-  // symbolize or those that belong to the runtime. For profile entries where
-  // the entire callstack is pruned, we drop the entry from the profile.
-  Error symbolizeAndFilterStackFrames(
-      std::unique_ptr<llvm::symbolize::SymbolizableModule> Symbolizer);
-  // Construct memprof records for each function and store it in the
-  // `FunctionProfileData` map. A function may have allocation profile data or
-  // callsite data or both.
-  Error mapRawProfileToRecords();
-
-  object::SectionedAddress getModuleOffset(uint64_t VirtualAddress);
-
-  // The profiled binary.
-  object::OwningBinary<object::Binary> Binary;
-  // The preferred load address of the executable segment.
-  uint64_t PreferredTextSegmentAddress = 0;
-  // The base address of the text segment in the process during profiling.
-  uint64_t ProfiledTextSegmentStart = 0;
-  // The limit address of the text segment in the process during profiling.
-  uint64_t ProfiledTextSegmentEnd = 0;
-
-  // The memory mapped segment information for all executable segments in the
-  // profiled binary (filtered from the raw profile using the build id).
-  llvm::SmallVector<SegmentEntry, 2> SegmentInfo;
-
-  // A map from callstack id (same as key in CallStackMap below) to the heap
-  // information recorded for that allocation context.
-  llvm::MapVector<uint64_t, MemInfoBlock> CallstackProfileData;
-  CallStackMap StackMap;
-
-  // Cached symbolization from PC to Frame.
-  llvm::DenseMap<uint64_t, llvm::SmallVector<FrameId>> SymbolizedFrame;
+#ifndef LLVM_PROFILEDATA_RAWMEMPROFREADER_H_
+#define LLVM_PROFILEDATA_RAWMEMPROFREADER_H_
 
-  // Whether to keep the symbol name for each frame after hashing.
-  bool KeepSymbolName = false;
-  // A mapping of the hash to symbol name, only used if KeepSymbolName is true.
-  llvm::DenseMap<uint64_t, std::string> GuidToSymbolName;
-};
-} // namespace memprof
-} // namespace llvm
+#include "llvm/ProfileData/MemProfReader.h"
 
 #endif // LLVM_PROFILEDATA_RAWMEMPROFREADER_H_
diff --git a/llvm/lib/ProfileData/CMakeLists.txt b/llvm/lib/ProfileData/CMakeLists.txt
index 99617a43fee75..408f9ff01ec87 100644
--- a/llvm/lib/ProfileData/CMakeLists.txt
+++ b/llvm/lib/ProfileData/CMakeLists.txt
@@ -6,8 +6,8 @@ add_llvm_component_library(LLVMProfileData
   InstrProfWriter.cpp
   ItaniumManglingCanonicalizer.cpp
   MemProf.cpp
+  MemProfReader.cpp
   ProfileSummaryBuilder.cpp
-  RawMemProfReader.cpp
   SampleProf.cpp
   SampleProfReader.cpp
   SampleProfWriter.cpp
diff --git a/llvm/lib/ProfileData/RawMemProfReader.cpp b/llvm/lib/ProfileData/MemProfReader.cpp
similarity index 99%
rename from llvm/lib/ProfileData/RawMemProfReader.cpp
rename to llvm/lib/ProfileData/MemProfReader.cpp
index e93fbc72f54eb..4ccec26597c09 100644
--- a/llvm/lib/ProfileData/RawMemProfReader.cpp
+++ b/llvm/lib/ProfileData/MemProfReader.cpp
@@ -32,7 +32,7 @@
 #include "llvm/ProfileData/InstrProf.h"
 #include "llvm/ProfileData/MemProf.h"
 #include "llvm/ProfileData/MemProfData.inc"
-#include "llvm/ProfileData/RawMemProfReader.h"
+#include "llvm/ProfileData/MemProfReader.h"
 #include "llvm/ProfileData/SampleProf.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/Endian.h"
@@ -295,8 +295,8 @@ Error RawMemProfReader::initialize(std::unique_ptr<MemoryBuffer> DataBuffer) {
   // Check whether the profiled binary was built with position independent code
   // (PIC). Perform sanity checks for assumptions we rely on to simplify
   // symbolization.
-  auto* Elf64LEObject = llvm::cast<llvm::object::ELF64LEObjectFile>(ElfObject);
-  const llvm::object::ELF64LEFile& ElfFile = Elf64LEObject->getELFFile();
+  auto *Elf64LEObject = llvm::cast<llvm::object::ELF64LEObjectFile>(ElfObject);
+  const llvm::object::ELF64LEFile &ElfFile = Elf64LEObject->getELFFile();
   auto PHdrsOr = ElfFile.program_headers();
   if (!PHdrsOr)
     return report(
diff --git a/llvm/tools/llvm-profdata/llvm-profdata.cpp b/llvm/tools/llvm-profdata/llvm-profdata.cpp
index 0b78564ccea37..6a70773613b7f 100644
--- a/llvm/tools/llvm-profdata/llvm-profdata.cpp
+++ b/llvm/tools/llvm-profdata/llvm-profdata.cpp
@@ -19,8 +19,8 @@
 #include "llvm/ProfileData/InstrProfReader.h"
 #include "llvm/ProfileData/InstrProfWriter.h"
 #include "llvm/ProfileData/MemProf.h"
+#include "llvm/ProfileData/MemProfReader.h"
 #include "llvm/ProfileData/ProfileCommon.h"
-#include "llvm/ProfileData/RawMemProfReader.h"
 #include "llvm/ProfileData/SampleProfReader.h"
 #include "llvm/ProfileData/SampleProfWriter.h"
 #include "llvm/Support/BalancedPartitioning.h"
diff --git a/llvm/unittests/ProfileData/MemProfTest.cpp b/llvm/unittests/ProfileData/MemProfTest.cpp
index f1aa6f37aa399..9cf307472d656 100644
--- a/llvm/unittests/ProfileData/MemProfTest.cpp
+++ b/llvm/unittests/ProfileData/MemProfTest.cpp
@@ -6,7 +6,7 @@
 #include "llvm/IR/Value.h"
 #include "llvm/Object/ObjectFile.h"
 #include "llvm/ProfileData/MemProfData.inc"
-#include "llvm/ProfileData/RawMemProfReader.h"
+#include "llvm/ProfileData/MemProfReader.h"
 #include "llvm/Support/raw_ostream.h"
 #include "gmock/gmock.h"
 #include "gtest/gtest.h"
diff --git a/llvm/utils/gn/secondary/llvm/lib/ProfileData/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/ProfileData/BUILD.gn
index 51568ac0472c3..9dbfe0f94c1db 100644
--- a/llvm/utils/gn/secondary/llvm/lib/ProfileData/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/ProfileData/BUILD.gn
@@ -16,8 +16,8 @@ static_library("ProfileData") {
     "InstrProfWriter.cpp",
     "ItaniumManglingCanonicalizer.cpp",
     "MemProf.cpp",
+    "MemProfReader.cpp",
     "ProfileSummaryBuilder.cpp",
-    "RawMemProfReader.cpp",
     "SampleProf.cpp",
     "SampleProfReader.cpp",
     "SampleProfWriter.cpp",

From bd32aaa8c9ec2094f605315b3989adc2a567ca98 Mon Sep 17 00:00:00 2001
From: Cyrill Leutwiler <bigcyrill@hotmail.com>
Date: Thu, 11 Apr 2024 07:11:51 +0200
Subject: [PATCH 099/886] [RISCV] Support rv{32, 64}e in the compiler builtins
 (#88252)

Register spills (save/restore) in RISC-V embedded work differently
because there are less registers and different stack alignment.

[GCC equivalent
](https://github.com/gcc-mirror/gcc/blob/master/libgcc/config/riscv/save-restore.S#L298C16-L336)

Follow up from #76777.

---------

Signed-off-by: xermicus <cyrill@parity.io>
---
 compiler-rt/lib/builtins/riscv/restore.S | 42 ++++++++++++++++++++++++
 compiler-rt/lib/builtins/riscv/save.S    | 42 ++++++++++++++++++++++++
 2 files changed, 84 insertions(+)

diff --git a/compiler-rt/lib/builtins/riscv/restore.S b/compiler-rt/lib/builtins/riscv/restore.S
index 73f64a920d669..6f43842c8ca68 100644
--- a/compiler-rt/lib/builtins/riscv/restore.S
+++ b/compiler-rt/lib/builtins/riscv/restore.S
@@ -22,6 +22,8 @@
 
 #if __riscv_xlen == 32
 
+#ifndef __riscv_32e
+
   .globl  __riscv_restore_12
   .type   __riscv_restore_12,@function
 __riscv_restore_12:
@@ -86,8 +88,29 @@ __riscv_restore_0:
   addi    sp, sp, 16
   ret
 
+#else
+
+  .globl  __riscv_restore_2
+  .type   __riscv_restore_2,@function
+  .globl  __riscv_restore_1
+  .type   __riscv_restore_1,@function
+  .globl  __riscv_restore_0
+  .type   __riscv_restore_0,@function
+__riscv_restore_2:
+__riscv_restore_1:
+__riscv_restore_0:
+  lw      s1,  0(sp)
+  lw      s0,  4(sp)
+  lw      ra,  8(sp)
+  addi    sp, sp, 12
+  ret
+
+#endif
+
 #elif __riscv_xlen == 64
 
+#ifndef __riscv_64e
+
   .globl  __riscv_restore_12
   .type   __riscv_restore_12,@function
 __riscv_restore_12:
@@ -161,6 +184,25 @@ __riscv_restore_0:
   addi    sp, sp, 16
   ret
 
+#else
+
+  .globl  __riscv_restore_2
+  .type   __riscv_restore_2,@function
+  .globl  __riscv_restore_1
+  .type   __riscv_restore_1,@function
+  .globl  __riscv_restore_0
+  .type   __riscv_restore_0,@function
+__riscv_restore_2:
+__riscv_restore_1:
+__riscv_restore_0:
+  ld      s1,  0(sp)
+  ld      s0,  8(sp)
+  ld      ra,  16(sp)
+  addi    sp, sp, 24
+  ret
+
+#endif
+
 #else
 # error "xlen must be 32 or 64 for save-restore implementation
 #endif
diff --git a/compiler-rt/lib/builtins/riscv/save.S b/compiler-rt/lib/builtins/riscv/save.S
index 85501aeb4c2e9..3e044179ff7f1 100644
--- a/compiler-rt/lib/builtins/riscv/save.S
+++ b/compiler-rt/lib/builtins/riscv/save.S
@@ -18,6 +18,8 @@
 
 #if __riscv_xlen == 32
 
+#ifndef __riscv_32e
+
   .globl  __riscv_save_12
   .type   __riscv_save_12,@function
 __riscv_save_12:
@@ -92,8 +94,29 @@ __riscv_save_0:
   sw      ra,  12(sp)
   jr      t0
 
+#else
+
+  .globl  __riscv_save_2
+  .type   __riscv_save_2,@function
+  .globl  __riscv_save_1
+  .type   __riscv_save_1,@function
+  .globl  __riscv_save_0
+  .type   __riscv_save_0,@function
+__riscv_save_2:
+__riscv_save_1:
+__riscv_save_0:
+  addi    sp, sp, -12
+  sw      s1,  0(sp)
+  sw      s0,  4(sp)
+  sw      ra,  8(sp)
+  jr      t0
+
+#endif
+
 #elif __riscv_xlen == 64
 
+#ifndef __riscv_64e
+
   .globl  __riscv_save_12
   .type   __riscv_save_12,@function
 __riscv_save_12:
@@ -181,6 +204,25 @@ __riscv_save_0:
   sd     ra, 8(sp)
   jr     t0
 
+#else
+
+  .globl  __riscv_save_2
+  .type   __riscv_save_2,@function
+  .globl  __riscv_save_1
+  .type   __riscv_save_1,@function
+  .globl  __riscv_save_0
+  .type   __riscv_save_0,@function
+__riscv_save_2:
+__riscv_save_1:
+__riscv_save_0:
+  addi   sp, sp, -24
+  sd     s1, 0(sp)
+  sd     s0, 8(sp)
+  sd     ra, 16(sp)
+  jr     t0
+
+#endif
+
 #else
 # error "xlen must be 32 or 64 for save-restore implementation
 #endif

From 5964c944bfe74cee2872cddb66eff22866cdb6ee Mon Sep 17 00:00:00 2001
From: Owen Pan <owenpiano@gmail.com>
Date: Wed, 10 Apr 2024 22:25:37 -0700
Subject: [PATCH 100/886] [clangd] Fix test case due to clang-format bug fix
 (#88352)

See commit 51f1681424f1.
---
 clang-tools-extra/clangd/unittests/HoverTests.cpp | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/clang-tools-extra/clangd/unittests/HoverTests.cpp b/clang-tools-extra/clangd/unittests/HoverTests.cpp
index 35db757b9c15b..5ead74748f550 100644
--- a/clang-tools-extra/clangd/unittests/HoverTests.cpp
+++ b/clang-tools-extra/clangd/unittests/HoverTests.cpp
@@ -1983,10 +1983,14 @@ TEST(Hover, All) {
             HI.Kind = index::SymbolKind::Macro;
             HI.Definition =
                 R"cpp(#define MACRO                                                                  \
-  { return 0; }
+  {                                                                            \
+    return 0;                                                                  \
+  }
 
 // Expands to
-{ return 0; })cpp";
+{
+  return 0;
+})cpp";
           }},
       {
           R"cpp(// Forward class declaration

From 45146082e693415f37413c656e0a0fd13d0e3136 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Wed, 10 Apr 2024 23:11:24 -0700
Subject: [PATCH 101/886] [ELF] relocateNonAlloc & ICF: replace random access
 iterators of relocations with input iterators. NFC

Also replace one `this->file` with a local variable as some function
calls are opaque to the compiler, causing unneeded reload.
---
 lld/ELF/ICF.cpp          | 22 ++++++++++++----------
 lld/ELF/InputSection.cpp | 16 ++++++++--------
 2 files changed, 20 insertions(+), 18 deletions(-)

diff --git a/lld/ELF/ICF.cpp b/lld/ELF/ICF.cpp
index 2551c2e807b73..bfc605c793a92 100644
--- a/lld/ELF/ICF.cpp
+++ b/lld/ELF/ICF.cpp
@@ -239,16 +239,17 @@ bool ICF<ELFT>::constantEq(const InputSection *secA, ArrayRef<RelTy> ra,
                            const InputSection *secB, ArrayRef<RelTy> rb) {
   if (ra.size() != rb.size())
     return false;
-  for (size_t i = 0; i < ra.size(); ++i) {
-    if (ra[i].r_offset != rb[i].r_offset ||
-        ra[i].getType(config->isMips64EL) != rb[i].getType(config->isMips64EL))
+  auto rai = ra.begin(), rae = ra.end(), rbi = rb.begin();
+  for (; rai != rae; ++rai, ++rbi) {
+    if (rai->r_offset != rbi->r_offset ||
+        rai->getType(config->isMips64EL) != rbi->getType(config->isMips64EL))
       return false;
 
-    uint64_t addA = getAddend<ELFT>(ra[i]);
-    uint64_t addB = getAddend<ELFT>(rb[i]);
+    uint64_t addA = getAddend<ELFT>(*rai);
+    uint64_t addB = getAddend<ELFT>(*rbi);
 
-    Symbol &sa = secA->file->getRelocTargetSym(ra[i]);
-    Symbol &sb = secB->file->getRelocTargetSym(rb[i]);
+    Symbol &sa = secA->file->getRelocTargetSym(*rai);
+    Symbol &sb = secB->file->getRelocTargetSym(*rbi);
     if (&sa == &sb) {
       if (addA == addB)
         continue;
@@ -336,10 +337,11 @@ bool ICF<ELFT>::variableEq(const InputSection *secA, ArrayRef<RelTy> ra,
                            const InputSection *secB, ArrayRef<RelTy> rb) {
   assert(ra.size() == rb.size());
 
-  for (size_t i = 0; i < ra.size(); ++i) {
+  auto rai = ra.begin(), rae = ra.end(), rbi = rb.begin();
+  for (; rai != rae; ++rai, ++rbi) {
     // The two sections must be identical.
-    Symbol &sa = secA->file->getRelocTargetSym(ra[i]);
-    Symbol &sb = secB->file->getRelocTargetSym(rb[i]);
+    Symbol &sa = secA->file->getRelocTargetSym(*rai);
+    Symbol &sb = secB->file->getRelocTargetSym(*rbi);
     if (&sa == &sb)
       continue;
 
diff --git a/lld/ELF/InputSection.cpp b/lld/ELF/InputSection.cpp
index c06816bcfd561..c8350652e65a6 100644
--- a/lld/ELF/InputSection.cpp
+++ b/lld/ELF/InputSection.cpp
@@ -918,8 +918,9 @@ void InputSection::relocateNonAlloc(uint8_t *buf, ArrayRef<RelTy> rels) {
       break;
     }
 
-  for (size_t i = 0, relsSize = rels.size(); i != relsSize; ++i) {
-    const RelTy &rel = rels[i];
+  const InputFile *f = this->file;
+  for (auto it = rels.begin(), end = rels.end(); it != end; ++it) {
+    const RelTy &rel = *it;
     const RelType type = rel.getType(config->isMips64EL);
     const uint64_t offset = rel.r_offset;
     uint8_t *bufLoc = buf + offset;
@@ -927,23 +928,22 @@ void InputSection::relocateNonAlloc(uint8_t *buf, ArrayRef<RelTy> rels) {
     if (!RelTy::IsRela)
       addend += target.getImplicitAddend(bufLoc, type);
 
-    Symbol &sym = this->file->getRelocTargetSym(rel);
+    Symbol &sym = f->getRelocTargetSym(rel);
     RelExpr expr = target.getRelExpr(type, sym, bufLoc);
     if (expr == R_NONE)
       continue;
     auto *ds = dyn_cast<Defined>(&sym);
 
     if (emachine == EM_RISCV && type == R_RISCV_SET_ULEB128) {
-      if (++i < relsSize &&
-          rels[i].getType(/*isMips64EL=*/false) == R_RISCV_SUB_ULEB128 &&
-          rels[i].r_offset == offset) {
+      if (++it != end &&
+          it->getType(/*isMips64EL=*/false) == R_RISCV_SUB_ULEB128 &&
+          it->r_offset == offset) {
         uint64_t val;
         if (!ds && tombstone) {
           val = *tombstone;
         } else {
           val = sym.getVA(addend) -
-                (this->file->getRelocTargetSym(rels[i]).getVA(0) +
-                 getAddend<ELFT>(rels[i]));
+                (f->getRelocTargetSym(*it).getVA(0) + getAddend<ELFT>(*it));
         }
         if (overwriteULEB128(bufLoc, val) >= 0x80)
           errorOrWarn(getLocation(offset) + ": ULEB128 value " + Twine(val) +

From 71f1932b842793e5dc7b17051452e8ff2f9219aa Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Thu, 11 Apr 2024 08:20:35 +0200
Subject: [PATCH 102/886] [clang][dataflow] Reland #87320: Propagate locations
 from result objects to initializers. (#88316)

This relands #87320 and additionally removes the now-unused function
`isOriginalRecordConstructor()`, which was causing buildbots to fail.
---
 .../FlowSensitive/DataflowEnvironment.h       |  64 ++-
 .../FlowSensitive/DataflowEnvironment.cpp     | 427 +++++++++++++-----
 clang/lib/Analysis/FlowSensitive/Transfer.cpp | 176 ++++----
 .../TypeErasedDataflowAnalysis.cpp            |  13 +-
 .../FlowSensitive/DataflowEnvironmentTest.cpp |  43 ++
 .../Analysis/FlowSensitive/TransferTest.cpp   | 172 ++++---
 6 files changed, 590 insertions(+), 305 deletions(-)

diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
index 9a65f76cdf56b..706664d7db1c2 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
@@ -30,6 +30,7 @@
 #include "llvm/ADT/MapVector.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/ErrorHandling.h"
+#include <memory>
 #include <type_traits>
 #include <utility>
 
@@ -344,17 +345,6 @@ class Environment {
   /// location of the result object to pass in `this`, even though prvalues are
   /// otherwise not associated with storage locations.
   ///
-  /// FIXME: Currently, this simply returns a stable storage location for `E`,
-  /// but this doesn't do the right thing in scenarios like the following:
-  /// ```
-  /// MyClass c = some_condition()? MyClass(foo) : MyClass(bar);
-  /// ```
-  /// Here, `MyClass(foo)` and `MyClass(bar)` will have two different storage
-  /// locations, when in fact their storage locations should be the same.
-  /// Eventually, we want to propagate storage locations from result objects
-  /// down to the prvalues that initialize them, similar to the way that this is
-  /// done in Clang's CodeGen.
-  ///
   /// Requirements:
   ///  `E` must be a prvalue of record type.
   RecordStorageLocation &
@@ -462,7 +452,13 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values.
-  void initializeFieldsWithValues(RecordStorageLocation &Loc);
+  /// If `Type` is provided, initializes only those fields that are modeled for
+  /// `Type`; this is intended for use in cases where `Loc` is a derived type
+  /// and we only want to initialize the fields of a base type.
+  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type);
+  void initializeFieldsWithValues(RecordStorageLocation &Loc) {
+    initializeFieldsWithValues(Loc, Loc.getType());
+  }
 
   /// Assigns `Val` as the value of `Loc` in the environment.
   void setValue(const StorageLocation &Loc, Value &Val);
@@ -653,6 +649,9 @@ class Environment {
   LLVM_DUMP_METHOD void dump(raw_ostream &OS) const;
 
 private:
+  using PrValueToResultObject =
+      llvm::DenseMap<const Expr *, RecordStorageLocation *>;
+
   // The copy-constructor is for use in fork() only.
   Environment(const Environment &) = default;
 
@@ -682,8 +681,10 @@ class Environment {
   /// Initializes the fields (including synthetic fields) of `Loc` with values,
   /// unless values of the field type are not supported or we hit one of the
   /// limits at which we stop producing values (controlled by `Visited`,
-  /// `Depth`, and `CreatedValuesCount`).
-  void initializeFieldsWithValues(RecordStorageLocation &Loc,
+  /// `Depth`, and `CreatedValuesCount`). If `Type` is different from
+  /// `Loc.getType()`, initializes only those fields that are modeled for
+  /// `Type`.
+  void initializeFieldsWithValues(RecordStorageLocation &Loc, QualType Type,
                                   llvm::DenseSet<QualType> &Visited, int Depth,
                                   int &CreatedValuesCount);
 
@@ -702,22 +703,45 @@ class Environment {
   /// and functions referenced in `FuncDecl`. `FuncDecl` must have a body.
   void initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl);
 
+  static PrValueToResultObject
+  buildResultObjectMap(DataflowAnalysisContext *DACtx,
+                       const FunctionDecl *FuncDecl,
+                       RecordStorageLocation *ThisPointeeLoc,
+                       RecordStorageLocation *LocForRecordReturnVal);
+
   // `DACtx` is not null and not owned by this object.
   DataflowAnalysisContext *DACtx;
 
-  // FIXME: move the fields `CallStack`, `ReturnVal`, `ReturnLoc` and
-  // `ThisPointeeLoc` into a separate call-context object, shared between
-  // environments in the same call.
+  // FIXME: move the fields `CallStack`, `ResultObjectMap`, `ReturnVal`,
+  // `ReturnLoc` and `ThisPointeeLoc` into a separate call-context object,
+  // shared between environments in the same call.
   // https://github.com/llvm/llvm-project/issues/59005
 
   // `DeclContext` of the block being analysed if provided.
   std::vector<const DeclContext *> CallStack;
 
-  // Value returned by the function (if it has non-reference return type).
+  // Maps from prvalues of record type to their result objects. Shared between
+  // all environments for the same function.
+  // FIXME: It's somewhat unsatisfactory that we have to use a `shared_ptr`
+  // here, though the cost is acceptable: The overhead of a `shared_ptr` is
+  // incurred when it is copied, and this happens only relatively rarely (when
+  // we fork the environment). The need for a `shared_ptr` will go away once we
+  // introduce a shared call-context object (see above).
+  std::shared_ptr<PrValueToResultObject> ResultObjectMap;
+
+  // The following three member variables handle various different types of
+  // return values.
+  // - If the return type is not a reference and not a record: Value returned
+  //   by the function.
   Value *ReturnVal = nullptr;
-  // Storage location of the reference returned by the function (if it has
-  // reference return type).
+  // - If the return type is a reference: Storage location of the reference
+  //   returned by the function.
   StorageLocation *ReturnLoc = nullptr;
+  // - If the return type is a record or the function being analyzed is a
+  //   constructor: Storage location into which the return value should be
+  //   constructed.
+  RecordStorageLocation *LocForRecordReturnVal = nullptr;
+
   // The storage location of the `this` pointee. Should only be null if the
   // function being analyzed is only a function and not a method.
   RecordStorageLocation *ThisPointeeLoc = nullptr;
diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index 1bfa7ebcfd50c..bea15ce9bd24d 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -15,6 +15,7 @@
 #include "clang/Analysis/FlowSensitive/DataflowEnvironment.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/DeclCXX.h"
+#include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/AST/Type.h"
 #include "clang/Analysis/FlowSensitive/DataflowLattice.h"
 #include "clang/Analysis/FlowSensitive/Value.h"
@@ -26,6 +27,8 @@
 #include <cassert>
 #include <utility>
 
+#define DEBUG_TYPE "dataflow"
+
 namespace clang {
 namespace dataflow {
 
@@ -354,6 +357,8 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   for (auto *Child : S.children())
     if (Child != nullptr)
       getFieldsGlobalsAndFuncs(*Child, Fields, Vars, Funcs);
+  if (const auto *DefaultArg = dyn_cast<CXXDefaultArgExpr>(&S))
+    getFieldsGlobalsAndFuncs(*DefaultArg->getExpr(), Fields, Vars, Funcs);
   if (const auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(&S))
     getFieldsGlobalsAndFuncs(*DefaultInit->getExpr(), Fields, Vars, Funcs);
 
@@ -386,6 +391,186 @@ getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
   }
 }
 
+namespace {
+
+// Visitor that builds a map from record prvalues to result objects.
+// This traverses the body of the function to be analyzed; for each result
+// object that it encounters, it propagates the storage location of the result
+// object to all record prvalues that can initialize it.
+class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
+public:
+  // `ResultObjectMap` will be filled with a map from record prvalues to result
+  // object. If the function being analyzed returns a record by value,
+  // `LocForRecordReturnVal` is the location to which this record should be
+  // written; otherwise, it is null.
+  explicit ResultObjectVisitor(
+      llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap,
+      RecordStorageLocation *LocForRecordReturnVal,
+      DataflowAnalysisContext &DACtx)
+      : ResultObjectMap(ResultObjectMap),
+        LocForRecordReturnVal(LocForRecordReturnVal), DACtx(DACtx) {}
+
+  bool shouldVisitImplicitCode() { return true; }
+
+  bool shouldVisitLambdaBody() const { return false; }
+
+  // Traverse all member and base initializers of `Ctor`. This function is not
+  // called by `RecursiveASTVisitor`; it should be called manually if we are
+  // analyzing a constructor. `ThisPointeeLoc` is the storage location that
+  // `this` points to.
+  void TraverseConstructorInits(const CXXConstructorDecl *Ctor,
+                                RecordStorageLocation *ThisPointeeLoc) {
+    assert(ThisPointeeLoc != nullptr);
+    for (const CXXCtorInitializer *Init : Ctor->inits()) {
+      Expr *InitExpr = Init->getInit();
+      if (FieldDecl *Field = Init->getMember();
+          Field != nullptr && Field->getType()->isRecordType()) {
+        PropagateResultObject(InitExpr, cast<RecordStorageLocation>(
+                                            ThisPointeeLoc->getChild(*Field)));
+      } else if (Init->getBaseClass()) {
+        PropagateResultObject(InitExpr, ThisPointeeLoc);
+      }
+
+      // Ensure that any result objects within `InitExpr` (e.g. temporaries)
+      // are also propagated to the prvalues that initialize them.
+      TraverseStmt(InitExpr);
+
+      // If this is a `CXXDefaultInitExpr`, also propagate any result objects
+      // within the default expression.
+      if (auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(InitExpr))
+        TraverseStmt(DefaultInit->getExpr());
+    }
+  }
+
+  bool TraverseBindingDecl(BindingDecl *BD) {
+    // `RecursiveASTVisitor` doesn't traverse holding variables for
+    // `BindingDecl`s by itself, so we need to tell it to.
+    if (VarDecl *HoldingVar = BD->getHoldingVar())
+      TraverseDecl(HoldingVar);
+    return RecursiveASTVisitor<ResultObjectVisitor>::TraverseBindingDecl(BD);
+  }
+
+  bool VisitVarDecl(VarDecl *VD) {
+    if (VD->getType()->isRecordType() && VD->hasInit())
+      PropagateResultObject(
+          VD->getInit(),
+          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*VD)));
+    return true;
+  }
+
+  bool VisitMaterializeTemporaryExpr(MaterializeTemporaryExpr *MTE) {
+    if (MTE->getType()->isRecordType())
+      PropagateResultObject(
+          MTE->getSubExpr(),
+          &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*MTE)));
+    return true;
+  }
+
+  bool VisitReturnStmt(ReturnStmt *Return) {
+    Expr *RetValue = Return->getRetValue();
+    if (RetValue != nullptr && RetValue->getType()->isRecordType() &&
+        RetValue->isPRValue())
+      PropagateResultObject(RetValue, LocForRecordReturnVal);
+    return true;
+  }
+
+  bool VisitExpr(Expr *E) {
+    // Clang's AST can have record-type prvalues without a result object -- for
+    // example as full-expressions contained in a compound statement or as
+    // arguments of call expressions. We notice this if we get here and a
+    // storage location has not yet been associated with `E`. In this case,
+    // treat this as if it was a `MaterializeTemporaryExpr`.
+    if (E->isPRValue() && E->getType()->isRecordType() &&
+        !ResultObjectMap.contains(E))
+      PropagateResultObject(
+          E, &cast<RecordStorageLocation>(DACtx.getStableStorageLocation(*E)));
+    return true;
+  }
+
+  // Assigns `Loc` as the result object location of `E`, then propagates the
+  // location to all lower-level prvalues that initialize the same object as
+  // `E` (or one of its base classes or member variables).
+  void PropagateResultObject(Expr *E, RecordStorageLocation *Loc) {
+    if (!E->isPRValue() || !E->getType()->isRecordType()) {
+      assert(false);
+      // Ensure we don't propagate the result object if we hit this in a
+      // release build.
+      return;
+    }
+
+    ResultObjectMap[E] = Loc;
+
+    // The following AST node kinds are "original initializers": They are the
+    // lowest-level AST node that initializes a given object, and nothing
+    // below them can initialize the same object (or part of it).
+    if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
+        isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
+        isa<CXXStdInitializerListExpr>(E)) {
+      return;
+    }
+
+    if (auto *InitList = dyn_cast<InitListExpr>(E)) {
+      if (!InitList->isSemanticForm())
+        return;
+      if (InitList->isTransparent()) {
+        PropagateResultObject(InitList->getInit(0), Loc);
+        return;
+      }
+
+      RecordInitListHelper InitListHelper(InitList);
+
+      for (auto [Base, Init] : InitListHelper.base_inits()) {
+        assert(Base->getType().getCanonicalType() ==
+               Init->getType().getCanonicalType());
+
+        // Storage location for the base class is the same as that of the
+        // derived class because we "flatten" the object hierarchy and put all
+        // fields in `RecordStorageLocation` of the derived class.
+        PropagateResultObject(Init, Loc);
+      }
+
+      for (auto [Field, Init] : InitListHelper.field_inits()) {
+        // Fields of non-record type are handled in
+        // `TransferVisitor::VisitInitListExpr()`.
+        if (!Field->getType()->isRecordType())
+          continue;
+        PropagateResultObject(
+            Init, cast<RecordStorageLocation>(Loc->getChild(*Field)));
+      }
+      return;
+    }
+
+    if (auto *Op = dyn_cast<BinaryOperator>(E); Op && Op->isCommaOp()) {
+      PropagateResultObject(Op->getRHS(), Loc);
+      return;
+    }
+
+    if (auto *Cond = dyn_cast<AbstractConditionalOperator>(E)) {
+      PropagateResultObject(Cond->getTrueExpr(), Loc);
+      PropagateResultObject(Cond->getFalseExpr(), Loc);
+      return;
+    }
+
+    // All other expression nodes that propagate a record prvalue should have
+    // exactly one child.
+    SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
+    LLVM_DEBUG({
+      if (Children.size() != 1)
+        E->dump();
+    });
+    assert(Children.size() == 1);
+    for (Stmt *S : Children)
+      PropagateResultObject(cast<Expr>(S), Loc);
+  }
+
+private:
+  llvm::DenseMap<const Expr *, RecordStorageLocation *> &ResultObjectMap;
+  RecordStorageLocation *LocForRecordReturnVal;
+  DataflowAnalysisContext &DACtx;
+};
+
+} // namespace
+
 Environment::Environment(DataflowAnalysisContext &DACtx)
     : DACtx(&DACtx),
       FlowConditionToken(DACtx.arena().makeFlowConditionToken()) {}
@@ -401,17 +586,23 @@ void Environment::initialize() {
   if (DeclCtx == nullptr)
     return;
 
-  if (const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx)) {
-    assert(FuncDecl->doesThisDeclarationHaveABody());
+  const auto *FuncDecl = dyn_cast<FunctionDecl>(DeclCtx);
+  if (FuncDecl == nullptr)
+    return;
 
-    initFieldsGlobalsAndFuncs(FuncDecl);
+  assert(FuncDecl->doesThisDeclarationHaveABody());
 
-    for (const auto *ParamDecl : FuncDecl->parameters()) {
-      assert(ParamDecl != nullptr);
-      setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
-    }
+  initFieldsGlobalsAndFuncs(FuncDecl);
+
+  for (const auto *ParamDecl : FuncDecl->parameters()) {
+    assert(ParamDecl != nullptr);
+    setStorageLocation(*ParamDecl, createObject(*ParamDecl, nullptr));
   }
 
+  if (FuncDecl->getReturnType()->isRecordType())
+    LocForRecordReturnVal = &cast<RecordStorageLocation>(
+        createStorageLocation(FuncDecl->getReturnType()));
+
   if (const auto *MethodDecl = dyn_cast<CXXMethodDecl>(DeclCtx)) {
     auto *Parent = MethodDecl->getParent();
     assert(Parent != nullptr);
@@ -444,6 +635,12 @@ void Environment::initialize() {
         initializeFieldsWithValues(ThisLoc);
     }
   }
+
+  // We do this below the handling of `CXXMethodDecl` above so that we can
+  // be sure that the storage location for `this` has been set.
+  ResultObjectMap = std::make_shared<PrValueToResultObject>(
+      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
+                           LocForRecordReturnVal));
 }
 
 // FIXME: Add support for resetting globals after function calls to enable
@@ -484,13 +681,18 @@ void Environment::initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl) {
     if (getStorageLocation(*D) != nullptr)
       continue;
 
-    setStorageLocation(*D, createObject(*D));
+    // We don't run transfer functions on the initializers of global variables,
+    // so they won't be associated with a value or storage location. We
+    // therefore intentionally don't pass an initializer to `createObject()`;
+    // in particular, this ensures that `createObject()` will initialize the
+    // fields of record-type variables with values.
+    setStorageLocation(*D, createObject(*D, nullptr));
   }
 
   for (const FunctionDecl *FD : Funcs) {
     if (getStorageLocation(*FD) != nullptr)
       continue;
-    auto &Loc = createStorageLocation(FD->getType());
+    auto &Loc = createStorageLocation(*FD);
     setStorageLocation(*FD, Loc);
   }
 }
@@ -519,6 +721,9 @@ Environment Environment::pushCall(const CallExpr *Call) const {
     }
   }
 
+  if (Call->getType()->isRecordType() && Call->isPRValue())
+    Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
+
   Env.pushCallInternal(Call->getDirectCallee(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
 
@@ -529,6 +734,7 @@ Environment Environment::pushCall(const CXXConstructExpr *Call) const {
   Environment Env(*this);
 
   Env.ThisPointeeLoc = &Env.getResultObjectLocation(*Call);
+  Env.LocForRecordReturnVal = &Env.getResultObjectLocation(*Call);
 
   Env.pushCallInternal(Call->getConstructor(),
                        llvm::ArrayRef(Call->getArgs(), Call->getNumArgs()));
@@ -557,6 +763,10 @@ void Environment::pushCallInternal(const FunctionDecl *FuncDecl,
     const VarDecl *Param = *ParamIt;
     setStorageLocation(*Param, createObject(*Param, Args[ArgIndex]));
   }
+
+  ResultObjectMap = std::make_shared<PrValueToResultObject>(
+      buildResultObjectMap(DACtx, FuncDecl, getThisPointeeStorageLocation(),
+                           LocForRecordReturnVal));
 }
 
 void Environment::popCall(const CallExpr *Call, const Environment &CalleeEnv) {
@@ -600,6 +810,9 @@ bool Environment::equivalentTo(const Environment &Other,
   if (ReturnLoc != Other.ReturnLoc)
     return false;
 
+  if (LocForRecordReturnVal != Other.LocForRecordReturnVal)
+    return false;
+
   if (ThisPointeeLoc != Other.ThisPointeeLoc)
     return false;
 
@@ -623,8 +836,10 @@ LatticeEffect Environment::widen(const Environment &PrevEnv,
   assert(DACtx == PrevEnv.DACtx);
   assert(ReturnVal == PrevEnv.ReturnVal);
   assert(ReturnLoc == PrevEnv.ReturnLoc);
+  assert(LocForRecordReturnVal == PrevEnv.LocForRecordReturnVal);
   assert(ThisPointeeLoc == PrevEnv.ThisPointeeLoc);
   assert(CallStack == PrevEnv.CallStack);
+  assert(ResultObjectMap == PrevEnv.ResultObjectMap);
 
   auto Effect = LatticeEffect::Unchanged;
 
@@ -656,12 +871,16 @@ Environment Environment::join(const Environment &EnvA, const Environment &EnvB,
                               Environment::ValueModel &Model,
                               ExprJoinBehavior ExprBehavior) {
   assert(EnvA.DACtx == EnvB.DACtx);
+  assert(EnvA.LocForRecordReturnVal == EnvB.LocForRecordReturnVal);
   assert(EnvA.ThisPointeeLoc == EnvB.ThisPointeeLoc);
   assert(EnvA.CallStack == EnvB.CallStack);
+  assert(EnvA.ResultObjectMap == EnvB.ResultObjectMap);
 
   Environment JoinedEnv(*EnvA.DACtx);
 
   JoinedEnv.CallStack = EnvA.CallStack;
+  JoinedEnv.ResultObjectMap = EnvA.ResultObjectMap;
+  JoinedEnv.LocForRecordReturnVal = EnvA.LocForRecordReturnVal;
   JoinedEnv.ThisPointeeLoc = EnvA.ThisPointeeLoc;
 
   if (EnvA.ReturnVal == nullptr || EnvB.ReturnVal == nullptr) {
@@ -730,6 +949,12 @@ StorageLocation &Environment::createStorageLocation(const Expr &E) {
 
 void Environment::setStorageLocation(const ValueDecl &D, StorageLocation &Loc) {
   assert(!DeclToLoc.contains(&D));
+  // The only kinds of declarations that may have a "variable" storage location
+  // are declarations of reference type and `BindingDecl`. For all other
+  // declaration, the storage location should be the stable storage location
+  // returned by `createStorageLocation()`.
+  assert(D.getType()->isReferenceType() || isa<BindingDecl>(D) ||
+         &Loc == &createStorageLocation(D));
   DeclToLoc[&D] = &Loc;
 }
 
@@ -764,77 +989,34 @@ StorageLocation *Environment::getStorageLocation(const Expr &E) const {
   return It == ExprToLoc.end() ? nullptr : &*It->second;
 }
 
-// Returns whether a prvalue of record type is the one that originally
-// constructs the object (i.e. it doesn't propagate it from one of its
-// children).
-static bool isOriginalRecordConstructor(const Expr &RecordPRValue) {
-  if (auto *Init = dyn_cast<InitListExpr>(&RecordPRValue))
-    return !Init->isSemanticForm() || !Init->isTransparent();
-  return isa<CXXConstructExpr>(RecordPRValue) || isa<CallExpr>(RecordPRValue) ||
-         isa<LambdaExpr>(RecordPRValue) ||
-         isa<CXXDefaultArgExpr>(RecordPRValue) ||
-         isa<CXXDefaultInitExpr>(RecordPRValue) ||
-         // The framework currently does not propagate the objects created in
-         // the two branches of a `ConditionalOperator` because there is no way
-         // to reconcile their storage locations, which are different. We
-         // therefore claim that the `ConditionalOperator` is the expression
-         // that originally constructs the object.
-         // Ultimately, this will be fixed by propagating locations down from
-         // the result object, rather than up from the original constructor as
-         // we do now (see also the FIXME in the documentation for
-         // `getResultObjectLocation()`).
-         isa<ConditionalOperator>(RecordPRValue);
-}
-
 RecordStorageLocation &
 Environment::getResultObjectLocation(const Expr &RecordPRValue) const {
   assert(RecordPRValue.getType()->isRecordType());
   assert(RecordPRValue.isPRValue());
 
-  // Returns a storage location that we can use if assertions fail.
-  auto FallbackForAssertFailure =
-      [this, &RecordPRValue]() -> RecordStorageLocation & {
+  assert(ResultObjectMap != nullptr);
+  RecordStorageLocation *Loc = ResultObjectMap->lookup(&RecordPRValue);
+  assert(Loc != nullptr);
+  // In release builds, use the "stable" storage location if the map lookup
+  // failed.
+  if (Loc == nullptr)
     return cast<RecordStorageLocation>(
         DACtx->getStableStorageLocation(RecordPRValue));
-  };
-
-  if (isOriginalRecordConstructor(RecordPRValue)) {
-    auto *Val = cast_or_null<RecordValue>(getValue(RecordPRValue));
-    // The builtin transfer function should have created a `RecordValue` for all
-    // original record constructors.
-    assert(Val);
-    if (!Val)
-      return FallbackForAssertFailure();
-    return Val->getLoc();
-  }
-
-  if (auto *Op = dyn_cast<BinaryOperator>(&RecordPRValue);
-      Op && Op->isCommaOp()) {
-    return getResultObjectLocation(*Op->getRHS());
-  }
-
-  // All other expression nodes that propagate a record prvalue should have
-  // exactly one child.
-  llvm::SmallVector<const Stmt *> children(RecordPRValue.child_begin(),
-                                           RecordPRValue.child_end());
-  assert(children.size() == 1);
-  if (children.empty())
-    return FallbackForAssertFailure();
-
-  return getResultObjectLocation(*cast<Expr>(children[0]));
+  return *Loc;
 }
 
 PointerValue &Environment::getOrCreateNullPointerValue(QualType PointeeType) {
   return DACtx->getOrCreateNullPointerValue(PointeeType);
 }
 
-void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc) {
+void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
+                                             QualType Type) {
   llvm::DenseSet<QualType> Visited;
   int CreatedValuesCount = 0;
-  initializeFieldsWithValues(Loc, Visited, 0, CreatedValuesCount);
+  initializeFieldsWithValues(Loc, Type, Visited, 0, CreatedValuesCount);
   if (CreatedValuesCount > MaxCompositeValueSize) {
-    llvm::errs() << "Attempting to initialize a huge value of type: "
-                 << Loc.getType() << '\n';
+    llvm::errs() << "Attempting to initialize a huge value of type: " << Type
+                 << '\n';
   }
 }
 
@@ -848,8 +1030,7 @@ void Environment::setValue(const Expr &E, Value &Val) {
   const Expr &CanonE = ignoreCFGOmittedNodes(E);
 
   if (auto *RecordVal = dyn_cast<RecordValue>(&Val)) {
-    assert(isOriginalRecordConstructor(CanonE) ||
-           &RecordVal->getLoc() == &getResultObjectLocation(CanonE));
+    assert(&RecordVal->getLoc() == &getResultObjectLocation(CanonE));
     (void)RecordVal;
   }
 
@@ -928,7 +1109,8 @@ Value *Environment::createValueUnlessSelfReferential(
   if (Type->isRecordType()) {
     CreatedValuesCount++;
     auto &Loc = cast<RecordStorageLocation>(createStorageLocation(Type));
-    initializeFieldsWithValues(Loc, Visited, Depth, CreatedValuesCount);
+    initializeFieldsWithValues(Loc, Loc.getType(), Visited, Depth,
+                               CreatedValuesCount);
 
     return &refreshRecordValue(Loc, *this);
   }
@@ -960,6 +1142,7 @@ Environment::createLocAndMaybeValue(QualType Ty,
 }
 
 void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
+                                             QualType Type,
                                              llvm::DenseSet<QualType> &Visited,
                                              int Depth,
                                              int &CreatedValuesCount) {
@@ -967,8 +1150,8 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     if (FieldType->isRecordType()) {
       auto &FieldRecordLoc = cast<RecordStorageLocation>(FieldLoc);
       setValue(FieldRecordLoc, create<RecordValue>(FieldRecordLoc));
-      initializeFieldsWithValues(FieldRecordLoc, Visited, Depth + 1,
-                                 CreatedValuesCount);
+      initializeFieldsWithValues(FieldRecordLoc, FieldRecordLoc.getType(),
+                                 Visited, Depth + 1, CreatedValuesCount);
     } else {
       if (!Visited.insert(FieldType.getCanonicalType()).second)
         return;
@@ -979,7 +1162,7 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
     }
   };
 
-  for (const auto &[Field, FieldLoc] : Loc.children()) {
+  for (const FieldDecl *Field : DACtx->getModeledFields(Type)) {
     assert(Field != nullptr);
     QualType FieldType = Field->getType();
 
@@ -988,14 +1171,12 @@ void Environment::initializeFieldsWithValues(RecordStorageLocation &Loc,
                    &createLocAndMaybeValue(FieldType, Visited, Depth + 1,
                                            CreatedValuesCount));
     } else {
+      StorageLocation *FieldLoc = Loc.getChild(*Field);
       assert(FieldLoc != nullptr);
       initField(FieldType, *FieldLoc);
     }
   }
-  for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
-    assert(FieldLoc != nullptr);
-    QualType FieldType = FieldLoc->getType();
-
+  for (const auto &[FieldName, FieldType] : DACtx->getSyntheticFields(Type)) {
     // Synthetic fields cannot have reference type, so we don't need to deal
     // with this case.
     assert(!FieldType->isReferenceType());
@@ -1022,38 +1203,36 @@ StorageLocation &Environment::createObjectInternal(const ValueDecl *D,
     return createObjectInternal(D, Ty.getNonReferenceType(), nullptr);
   }
 
-  Value *Val = nullptr;
-  if (InitExpr) {
-    // In the (few) cases where an expression is intentionally
-    // "uninterpreted", `InitExpr` is not associated with a value.  There are
-    // two ways to handle this situation: propagate the status, so that
-    // uninterpreted initializers result in uninterpreted variables, or
-    // provide a default value. We choose the latter so that later refinements
-    // of the variable can be used for reasoning about the surrounding code.
-    // For this reason, we let this case be handled by the `createValue()`
-    // call below.
-    //
-    // FIXME. If and when we interpret all language cases, change this to
-    // assert that `InitExpr` is interpreted, rather than supplying a
-    // default value (assuming we don't update the environment API to return
-    // references).
-    Val = getValue(*InitExpr);
-
-    if (!Val && isa<ImplicitValueInitExpr>(InitExpr) &&
-        InitExpr->getType()->isPointerType())
-      Val = &getOrCreateNullPointerValue(InitExpr->getType()->getPointeeType());
-  }
-  if (!Val)
-    Val = createValue(Ty);
-
-  if (Ty->isRecordType())
-    return cast<RecordValue>(Val)->getLoc();
-
   StorageLocation &Loc =
       D ? createStorageLocation(*D) : createStorageLocation(Ty);
 
-  if (Val)
-    setValue(Loc, *Val);
+  if (Ty->isRecordType()) {
+    auto &RecordLoc = cast<RecordStorageLocation>(Loc);
+    if (!InitExpr)
+      initializeFieldsWithValues(RecordLoc);
+    refreshRecordValue(RecordLoc, *this);
+  } else {
+    Value *Val = nullptr;
+    if (InitExpr)
+      // In the (few) cases where an expression is intentionally
+      // "uninterpreted", `InitExpr` is not associated with a value.  There are
+      // two ways to handle this situation: propagate the status, so that
+      // uninterpreted initializers result in uninterpreted variables, or
+      // provide a default value. We choose the latter so that later refinements
+      // of the variable can be used for reasoning about the surrounding code.
+      // For this reason, we let this case be handled by the `createValue()`
+      // call below.
+      //
+      // FIXME. If and when we interpret all language cases, change this to
+      // assert that `InitExpr` is interpreted, rather than supplying a
+      // default value (assuming we don't update the environment API to return
+      // references).
+      Val = getValue(*InitExpr);
+    if (!Val)
+      Val = createValue(Ty);
+    if (Val)
+      setValue(Loc, *Val);
+  }
 
   return Loc;
 }
@@ -1072,6 +1251,8 @@ bool Environment::allows(const Formula &F) const {
 
 void Environment::dump(raw_ostream &OS) const {
   llvm::DenseMap<const StorageLocation *, std::string> LocToName;
+  if (LocForRecordReturnVal != nullptr)
+    LocToName[LocForRecordReturnVal] = "(returned record)";
   if (ThisPointeeLoc != nullptr)
     LocToName[ThisPointeeLoc] = "this";
 
@@ -1102,6 +1283,9 @@ void Environment::dump(raw_ostream &OS) const {
       if (auto Iter = LocToName.find(ReturnLoc); Iter != LocToName.end())
         OS << " (" << Iter->second << ")";
       OS << "\n";
+    } else if (Func->getReturnType()->isRecordType() ||
+               isa<CXXConstructorDecl>(Func)) {
+      OS << "LocForRecordReturnVal: " << LocForRecordReturnVal << "\n";
     } else if (!Func->getReturnType()->isVoidType()) {
       if (ReturnVal == nullptr)
         OS << "ReturnVal: nullptr\n";
@@ -1122,6 +1306,22 @@ void Environment::dump() const {
   dump(llvm::dbgs());
 }
 
+Environment::PrValueToResultObject Environment::buildResultObjectMap(
+    DataflowAnalysisContext *DACtx, const FunctionDecl *FuncDecl,
+    RecordStorageLocation *ThisPointeeLoc,
+    RecordStorageLocation *LocForRecordReturnVal) {
+  assert(FuncDecl->doesThisDeclarationHaveABody());
+
+  PrValueToResultObject Map;
+
+  ResultObjectVisitor Visitor(Map, LocForRecordReturnVal, *DACtx);
+  if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(FuncDecl))
+    Visitor.TraverseConstructorInits(Ctor, ThisPointeeLoc);
+  Visitor.TraverseStmt(FuncDecl->getBody());
+
+  return Map;
+}
+
 RecordStorageLocation *getImplicitObjectLocation(const CXXMemberCallExpr &MCE,
                                                  const Environment &Env) {
   Expr *ImplicitObject = MCE.getImplicitObjectArgument();
@@ -1216,24 +1416,11 @@ RecordValue &refreshRecordValue(RecordStorageLocation &Loc, Environment &Env) {
 RecordValue &refreshRecordValue(const Expr &Expr, Environment &Env) {
   assert(Expr.getType()->isRecordType());
 
-  if (Expr.isPRValue()) {
-    if (auto *ExistingVal = Env.get<RecordValue>(Expr)) {
-      auto &NewVal = Env.create<RecordValue>(ExistingVal->getLoc());
-      Env.setValue(Expr, NewVal);
-      Env.setValue(NewVal.getLoc(), NewVal);
-      return NewVal;
-    }
+  if (Expr.isPRValue())
+    refreshRecordValue(Env.getResultObjectLocation(Expr), Env);
 
-    auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
-    Env.setValue(Expr, NewVal);
-    return NewVal;
-  }
-
-  if (auto *Loc = Env.get<RecordStorageLocation>(Expr)) {
-    auto &NewVal = Env.create<RecordValue>(*Loc);
-    Env.setValue(*Loc, NewVal);
-    return NewVal;
-  }
+  if (auto *Loc = Env.get<RecordStorageLocation>(Expr))
+    refreshRecordValue(*Loc, Env);
 
   auto &NewVal = *cast<RecordValue>(Env.createValue(Expr.getType()));
   Env.setStorageLocation(Expr, NewVal.getLoc());
diff --git a/clang/lib/Analysis/FlowSensitive/Transfer.cpp b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
index 0a2e8368d541d..88a9c0eccbebc 100644
--- a/clang/lib/Analysis/FlowSensitive/Transfer.cpp
+++ b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
@@ -460,11 +460,9 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // So make sure we have a value if we didn't propagate one above.
     if (S->isPRValue() && S->getType()->isRecordType()) {
       if (Env.getValue(*S) == nullptr) {
-        Value *Val = Env.createValue(S->getType());
-        // We're guaranteed to always be able to create a value for record
-        // types.
-        assert(Val != nullptr);
-        Env.setValue(*S, *Val);
+        auto &Loc = Env.getResultObjectLocation(*S);
+        Env.initializeFieldsWithValues(Loc);
+        refreshRecordValue(Loc, Env);
       }
     }
   }
@@ -472,6 +470,13 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
   void VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *S) {
     const Expr *InitExpr = S->getExpr();
     assert(InitExpr != nullptr);
+
+    // If this is a prvalue of record type, the handler for `*InitExpr` (if one
+    // exists) will initialize the result object; there is no value to propgate
+    // here.
+    if (S->getType()->isRecordType() && S->isPRValue())
+      return;
+
     propagateValueOrStorageLocation(*InitExpr, *S, Env);
   }
 
@@ -479,6 +484,17 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     const CXXConstructorDecl *ConstructorDecl = S->getConstructor();
     assert(ConstructorDecl != nullptr);
 
+    // `CXXConstructExpr` can have array type if default-initializing an array
+    // of records. We don't handle this specifically beyond potentially inlining
+    // the call.
+    if (!S->getType()->isRecordType()) {
+      transferInlineCall(S, ConstructorDecl);
+      return;
+    }
+
+    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+    Env.setValue(*S, refreshRecordValue(Loc, Env));
+
     if (ConstructorDecl->isCopyOrMoveConstructor()) {
       // It is permissible for a copy/move constructor to have additional
       // parameters as long as they have default arguments defined for them.
@@ -491,24 +507,14 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (ArgLoc == nullptr)
         return;
 
-      if (S->isElidable()) {
-        if (Value *Val = Env.getValue(*ArgLoc))
-          Env.setValue(*S, *Val);
-      } else {
-        auto &Val = *cast<RecordValue>(Env.createValue(S->getType()));
-        Env.setValue(*S, Val);
-        copyRecord(*ArgLoc, Val.getLoc(), Env);
-      }
+      // Even if the copy/move constructor call is elidable, we choose to copy
+      // the record in all cases (which isn't wrong, just potentially not
+      // optimal).
+      copyRecord(*ArgLoc, Loc, Env);
       return;
     }
 
-    // `CXXConstructExpr` can have array type if default-initializing an array
-    // of records, and we currently can't create values for arrays. So check if
-    // we've got a record type.
-    if (S->getType()->isRecordType()) {
-      auto &InitialVal = *cast<RecordValue>(Env.createValue(S->getType()));
-      Env.setValue(*S, InitialVal);
-    }
+    Env.initializeFieldsWithValues(Loc, S->getType());
 
     transferInlineCall(S, ConstructorDecl);
   }
@@ -551,19 +557,15 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       if (S->isGLValue()) {
         Env.setStorageLocation(*S, *LocDst);
       } else if (S->getType()->isRecordType()) {
-        // Make sure that we have a `RecordValue` for this expression so that
-        // `Environment::getResultObjectLocation()` is able to return a location
-        // for it.
-        if (Env.getValue(*S) == nullptr)
-          refreshRecordValue(*S, Env);
+        // Assume that the assignment returns the assigned value.
+        copyRecord(*LocDst, Env.getResultObjectLocation(*S), Env);
       }
 
       return;
     }
 
-    // CXXOperatorCallExpr can be prvalues. Call `VisitCallExpr`() to create
-    // a `RecordValue` for them so that `Environment::getResultObjectLocation()`
-    // can return a value.
+    // `CXXOperatorCallExpr` can be a prvalue. Call `VisitCallExpr`() to
+    // initialize the prvalue's fields with values.
     VisitCallExpr(S);
   }
 
@@ -580,11 +582,6 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     }
   }
 
-  void VisitCXXTemporaryObjectExpr(const CXXTemporaryObjectExpr *S) {
-    if (Value *Val = Env.createValue(S->getType()))
-      Env.setValue(*S, *Val);
-  }
-
   void VisitCallExpr(const CallExpr *S) {
     // Of clang's builtins, only `__builtin_expect` is handled explicitly, since
     // others (like trap, debugtrap, and unreachable) are handled by CFG
@@ -612,13 +609,14 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     } else if (const FunctionDecl *F = S->getDirectCallee()) {
       transferInlineCall(S, F);
 
-      // If this call produces a prvalue of record type, make sure that we have
-      // a `RecordValue` for it. This is required so that
-      // `Environment::getResultObjectLocation()` is able to return a location
-      // for this `CallExpr`.
+      // If this call produces a prvalue of record type, initialize its fields
+      // with values.
       if (S->getType()->isRecordType() && S->isPRValue())
-        if (Env.getValue(*S) == nullptr)
-          refreshRecordValue(*S, Env);
+        if (Env.getValue(*S) == nullptr) {
+          RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+          Env.initializeFieldsWithValues(Loc);
+          Env.setValue(*S, refreshRecordValue(Loc, Env));
+        }
     }
   }
 
@@ -666,8 +664,10 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
     // `getLogicOperatorSubExprValue()`.
     if (S->isGLValue())
       Env.setStorageLocation(*S, Env.createObject(S->getType()));
-    else if (Value *Val = Env.createValue(S->getType()))
-      Env.setValue(*S, *Val);
+    else if (!S->getType()->isRecordType()) {
+      if (Value *Val = Env.createValue(S->getType()))
+        Env.setValue(*S, *Val);
+    }
   }
 
   void VisitInitListExpr(const InitListExpr *S) {
@@ -688,71 +688,51 @@ class TransferVisitor : public ConstStmtVisitor<TransferVisitor> {
       return;
     }
 
-    llvm::DenseMap<const ValueDecl *, StorageLocation *> FieldLocs;
-    RecordInitListHelper InitListHelper(S);
+    RecordStorageLocation &Loc = Env.getResultObjectLocation(*S);
+    Env.setValue(*S, refreshRecordValue(Loc, Env));
 
-    for (auto [Base, Init] : InitListHelper.base_inits()) {
-      assert(Base->getType().getCanonicalType() ==
-             Init->getType().getCanonicalType());
-      auto *BaseVal = Env.get<RecordValue>(*Init);
-      if (!BaseVal)
-        BaseVal = cast<RecordValue>(Env.createValue(Init->getType()));
-      // Take ownership of the fields of the `RecordValue` for the base class
-      // and incorporate them into the "flattened" set of fields for the
-      // derived class.
-      auto Children = BaseVal->getLoc().children();
-      FieldLocs.insert(Children.begin(), Children.end());
-    }
+    // Initialization of base classes and fields of record type happens when we
+    // visit the nested `CXXConstructExpr` or `InitListExpr` for that base class
+    // or field. We therefore only need to deal with fields of non-record type
+    // here.
 
-    for (auto [Field, Init] : InitListHelper.field_inits()) {
-      assert(
-          // The types are same, or
-          Field->getType().getCanonicalType().getUnqualifiedType() ==
-              Init->getType().getCanonicalType().getUnqualifiedType() ||
-          // The field's type is T&, and initializer is T
-          (Field->getType()->isReferenceType() &&
-           Field->getType().getCanonicalType()->getPointeeType() ==
-               Init->getType().getCanonicalType()));
-      auto& Loc = Env.createObject(Field->getType(), Init);
-      FieldLocs.insert({Field, &Loc});
-    }
+    RecordInitListHelper InitListHelper(S);
 
-    // In the case of a union, we don't in general have initializers for all
-    // of the fields. Create storage locations for the remaining fields (but
-    // don't associate them with values).
-    if (Type->isUnionType()) {
-      for (const FieldDecl *Field :
-           Env.getDataflowAnalysisContext().getModeledFields(Type)) {
-        if (auto [it, inserted] = FieldLocs.insert({Field, nullptr}); inserted)
-          it->second = &Env.createStorageLocation(Field->getType());
+    for (auto [Field, Init] : InitListHelper.field_inits()) {
+      if (Field->getType()->isRecordType())
+        continue;
+      if (Field->getType()->isReferenceType()) {
+        assert(Field->getType().getCanonicalType()->getPointeeType() ==
+               Init->getType().getCanonicalType());
+        Loc.setChild(*Field, &Env.createObject(Field->getType(), Init));
+        continue;
       }
+      assert(Field->getType().getCanonicalType().getUnqualifiedType() ==
+             Init->getType().getCanonicalType().getUnqualifiedType());
+      StorageLocation *FieldLoc = Loc.getChild(*Field);
+      // Locations for non-reference fields must always be non-null.
+      assert(FieldLoc != nullptr);
+      Value *Val = Env.getValue(*Init);
+      if (Val == nullptr && isa<ImplicitValueInitExpr>(Init) &&
+          Init->getType()->isPointerType())
+        Val =
+            &Env.getOrCreateNullPointerValue(Init->getType()->getPointeeType());
+      if (Val == nullptr)
+        Val = Env.createValue(Field->getType());
+      if (Val != nullptr)
+        Env.setValue(*FieldLoc, *Val);
     }
 
-    // Check that we satisfy the invariant that a `RecordStorageLoation`
-    // contains exactly the set of modeled fields for that type.
-    // `ModeledFields` includes fields from all the bases, but only the
-    // modeled ones. However, if a class type is initialized with an
-    // `InitListExpr`, all fields in the class, including those from base
-    // classes, are included in the set of modeled fields. The code above
-    // should therefore populate exactly the modeled fields.
-    assert(containsSameFields(
-        Env.getDataflowAnalysisContext().getModeledFields(Type), FieldLocs));
-
-    RecordStorageLocation::SyntheticFieldMap SyntheticFieldLocs;
-    for (const auto &Entry :
-         Env.getDataflowAnalysisContext().getSyntheticFields(Type)) {
-      SyntheticFieldLocs.insert(
-          {Entry.getKey(), &Env.createObject(Entry.getValue())});
+    for (const auto &[FieldName, FieldLoc] : Loc.synthetic_fields()) {
+      QualType FieldType = FieldLoc->getType();
+      if (FieldType->isRecordType()) {
+        Env.initializeFieldsWithValues(*cast<RecordStorageLocation>(FieldLoc));
+      } else {
+        if (Value *Val = Env.createValue(FieldType))
+          Env.setValue(*FieldLoc, *Val);
+      }
     }
 
-    auto &Loc = Env.getDataflowAnalysisContext().createRecordStorageLocation(
-        Type, std::move(FieldLocs), std::move(SyntheticFieldLocs));
-    RecordValue &RecordVal = Env.create<RecordValue>(Loc);
-
-    Env.setValue(Loc, RecordVal);
-
-    Env.setValue(*S, RecordVal);
-
     // FIXME: Implement array initialization.
   }
 
diff --git a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
index 595f70f819ddb..1b73c5d683016 100644
--- a/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
+++ b/clang/lib/Analysis/FlowSensitive/TypeErasedDataflowAnalysis.cpp
@@ -369,17 +369,10 @@ builtinTransferInitializer(const CFGInitializer &Elt,
     ParentLoc->setChild(*Member, InitExprLoc);
   } else if (auto *InitExprVal = Env.getValue(*InitExpr)) {
     assert(MemberLoc != nullptr);
-    if (Member->getType()->isRecordType()) {
-      auto *InitValStruct = cast<RecordValue>(InitExprVal);
-      // FIXME: Rather than performing a copy here, we should really be
-      // initializing the field in place. This would require us to propagate the
-      // storage location of the field to the AST node that creates the
-      // `RecordValue`.
-      copyRecord(InitValStruct->getLoc(),
-                 *cast<RecordStorageLocation>(MemberLoc), Env);
-    } else {
+    // Record-type initializers construct themselves directly into the result
+    // object, so there is no need to handle them here.
+    if (!Member->getType()->isRecordType())
       Env.setValue(*MemberLoc, *InitExprVal);
-    }
   }
 }
 
diff --git a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
index 465a8e21690c4..cc20623f881ff 100644
--- a/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/DataflowEnvironmentTest.cpp
@@ -24,6 +24,7 @@ namespace {
 
 using namespace clang;
 using namespace dataflow;
+using ::clang::dataflow::test::findValueDecl;
 using ::clang::dataflow::test::getFieldValue;
 using ::testing::Contains;
 using ::testing::IsNull;
@@ -199,6 +200,48 @@ TEST_F(EnvironmentTest, JoinRecords) {
   }
 }
 
+TEST_F(EnvironmentTest, DifferentReferenceLocInJoin) {
+  // This tests the case where the storage location for a reference-type
+  // variable is different for two states being joined. We used to believe this
+  // could not happen and therefore had an assertion disallowing this; this test
+  // exists to demonstrate that we can handle this condition without a failing
+  // assertion. See also the discussion here:
+  // https://discourse.llvm.org/t/70086/6
+
+  using namespace ast_matchers;
+
+  std::string Code = R"cc(
+    void f(int &ref) {}
+  )cc";
+
+  auto Unit =
+      tooling::buildASTFromCodeWithArgs(Code, {"-fsyntax-only", "-std=c++11"});
+  auto &Context = Unit->getASTContext();
+
+  ASSERT_EQ(Context.getDiagnostics().getClient()->getNumErrors(), 0U);
+
+  const ValueDecl *Ref = findValueDecl(Context, "ref");
+
+  Environment Env1(DAContext);
+  StorageLocation &Loc1 = Env1.createStorageLocation(Context.IntTy);
+  Env1.setStorageLocation(*Ref, Loc1);
+
+  Environment Env2(DAContext);
+  StorageLocation &Loc2 = Env2.createStorageLocation(Context.IntTy);
+  Env2.setStorageLocation(*Ref, Loc2);
+
+  EXPECT_NE(&Loc1, &Loc2);
+
+  Environment::ValueModel Model;
+  Environment EnvJoined =
+      Environment::join(Env1, Env2, Model, Environment::DiscardExprState);
+
+  // Joining environments with different storage locations for the same
+  // declaration results in the declaration being removed from the joined
+  // environment.
+  EXPECT_EQ(EnvJoined.getStorageLocation(*Ref), nullptr);
+}
+
 TEST_F(EnvironmentTest, InitGlobalVarsFun) {
   using namespace ast_matchers;
 
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index ca055a462a286..00dafb2988c69 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -1582,10 +1582,9 @@ TEST(TransferTest, FieldsDontHaveValuesInConstructorWithBaseClass) {
       [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
          ASTContext &ASTCtx) {
         const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-        // FIXME: The field of the base class should already have been
-        // initialized with a value by the base constructor. This test documents
-        // the current buggy behavior.
-        EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
+        // The field of the base class should already have been initialized with
+        // a value by the base constructor.
+        EXPECT_NE(getFieldValue(Env.getThisPointeeStorageLocation(), "BaseVal",
                                 ASTCtx, Env),
                   nullptr);
         EXPECT_EQ(getFieldValue(Env.getThisPointeeStorageLocation(), "Val",
@@ -2998,8 +2997,12 @@ TEST(TransferTest, ResultObjectLocation) {
 
 TEST(TransferTest, ResultObjectLocationForDefaultArgExpr) {
   std::string Code = R"(
-    struct S {};
-    void funcWithDefaultArg(S s = S());
+    struct Inner {};
+    struct Outer {
+        Inner I = {};
+    };
+
+    void funcWithDefaultArg(Outer O = {});
     void target() {
       funcWithDefaultArg();
       // [[p]]
@@ -3058,13 +3061,7 @@ TEST(TransferTest, ResultObjectLocationForDefaultInitExpr) {
 
         RecordStorageLocation &Loc = Env.getResultObjectLocation(*DefaultInit);
 
-        // FIXME: The result object location for the `CXXDefaultInitExpr` should
-        // be the location of the member variable being initialized, but we
-        // don't do this correctly yet; see also comments in
-        // `builtinTransferInitializer()`.
-        // For the time being, we just document the current erroneous behavior
-        // here (this should be `EXPECT_EQ` when the behavior is fixed).
-        EXPECT_NE(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
+        EXPECT_EQ(&Loc, Env.getThisPointeeStorageLocation()->getChild(*SField));
       });
 }
 
@@ -3101,6 +3098,79 @@ TEST(TransferTest, ResultObjectLocationForCXXOperatorCallExpr) {
       });
 }
 
+TEST(TransferTest, ResultObjectLocationForStdInitializerListExpr) {
+  std::string Code = R"(
+    namespace std {
+    template <typename T>
+    struct initializer_list {};
+    } // namespace std
+
+    void target() {
+      std::initializer_list<int> list = {1};
+      // [[p]]
+    }
+  )";
+
+  using ast_matchers::cxxStdInitializerListExpr;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *StdInitList = selectFirst<CXXStdInitializerListExpr>(
+            "std_init_list",
+            match(cxxStdInitializerListExpr().bind("std_init_list"), ASTCtx));
+        ASSERT_NE(StdInitList, nullptr);
+
+        EXPECT_EQ(&Env.getResultObjectLocation(*StdInitList),
+                  &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "list"));
+      });
+}
+
+TEST(TransferTest, ResultObjectLocationPropagatesThroughConditionalOperator) {
+  std::string Code = R"(
+    struct A {
+      A(int);
+    };
+
+    void target(bool b) {
+      A a = b ? A(0) : A(1);
+      (void)0; // [[p]]
+    }
+  )";
+  using ast_matchers::cxxConstructExpr;
+  using ast_matchers::equals;
+  using ast_matchers::hasArgument;
+  using ast_matchers::integerLiteral;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  using ast_matchers::traverse;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *ConstructExpr0 = selectFirst<CXXConstructExpr>(
+            "construct",
+            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(0))))
+                      .bind("construct"),
+                  ASTCtx));
+        auto *ConstructExpr1 = selectFirst<CXXConstructExpr>(
+            "construct",
+            match(cxxConstructExpr(hasArgument(0, integerLiteral(equals(1))))
+                      .bind("construct"),
+                  ASTCtx));
+
+        auto &ALoc = getLocForDecl<StorageLocation>(ASTCtx, Env, "a");
+        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr0), &ALoc);
+        EXPECT_EQ(&Env.getResultObjectLocation(*ConstructExpr1), &ALoc);
+      });
+}
+
 TEST(TransferTest, StaticCast) {
   std::string Code = R"(
     void target(int Foo) {
@@ -5886,6 +5956,38 @@ TEST(TransferTest, ContextSensitiveReturnRecord) {
       {BuiltinOptions{ContextSensitiveOptions{}}});
 }
 
+TEST(TransferTest, ContextSensitiveReturnSelfReferentialRecord) {
+  std::string Code = R"(
+    struct S {
+      S() { self = this; }
+      S *self;
+    };
+
+    S makeS() {
+      // RVO guarantees that this will be constructed directly into `MyS`.
+      return S();
+    }
+
+    void target() {
+      S MyS = makeS();
+      // [[p]]
+    }
+  )";
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto &MySLoc = getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "MyS");
+
+        auto *SelfVal =
+            cast<PointerValue>(getFieldValue(&MySLoc, "self", ASTCtx, Env));
+        EXPECT_EQ(&SelfVal->getPointeeLoc(), &MySLoc);
+      },
+      {BuiltinOptions{ContextSensitiveOptions{}}});
+}
+
 TEST(TransferTest, ContextSensitiveMethodLiteral) {
   std::string Code = R"(
     class MyClass {
@@ -6830,50 +6932,6 @@ TEST(TransferTest, LambdaCaptureThis) {
       });
 }
 
-TEST(TransferTest, DifferentReferenceLocInJoin) {
-  // This test triggers a case where the storage location for a reference-type
-  // variable is different for two states being joined. We used to believe this
-  // could not happen and therefore had an assertion disallowing this; this test
-  // exists to demonstrate that we can handle this condition without a failing
-  // assertion. See also the discussion here:
-  // https://discourse.llvm.org/t/70086/6
-  std::string Code = R"(
-    namespace std {
-      template <class T> struct initializer_list {
-        const T* begin();
-        const T* end();
-      };
-    }
-
-    void target(char* p, char* end) {
-      while (p != end) {
-        if (*p == ' ') {
-          p++;
-          continue;
-        }
-
-        auto && range = {1, 2};
-        for (auto b = range.begin(), e = range.end(); b != e; ++b) {
-        }
-        (void)0;
-        // [[p]]
-      }
-    }
-  )";
-  runDataflow(
-      Code,
-      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
-         ASTContext &ASTCtx) {
-        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
-
-        // Joining environments with different storage locations for the same
-        // declaration results in the declaration being removed from the joined
-        // environment.
-        const ValueDecl *VD = findValueDecl(ASTCtx, "range");
-        ASSERT_EQ(Env.getStorageLocation(*VD), nullptr);
-      });
-}
-
 // This test verifies correct modeling of a relational dependency that goes
 // through unmodeled functions (the simple `cond()` in this case).
 TEST(TransferTest, ConditionalRelation) {

From 297eca981ea1133388c82ddbfaf9d86391abac65 Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Thu, 11 Apr 2024 08:23:48 +0200
Subject: [PATCH 103/886] [mlir][Interfaces] `ValueBoundsOpInterface`: Add API
 to compare values (#86915)

This commit adds a new public API to `ValueBoundsOpInterface` to compare
values/dims. Supported comparison operators are: LT, LE, EQ, GE, GT.

The new `ValueBoundsOpInterface::compare` API replaces and generalizes
`ValueBoundsOpInterface::areEqual`. Not only does it provide additional
comparison operators, it also works in cases where the difference
between the two values/dims is non-constant. The previous implementation
of `areEqual` used to compute a constant bound of `val1 - val2` (check
if it `== 0` or `!= 0`).

Note: This commit refactors, generalizes and adds a public API for
value/dim comparison. The comparison functionality itself was introduced
in #85895 and is already in use for analyzing `scf.if`.

In the long term, this improvement will allow for a more powerful
analysis of subset ops. A future commit will update
`areOverlappingSlices` to use the new comparison API.
(`areEquivalentSlices` is already using the new API.) This will improve
subset equivalence/disjointness checks with non-constant
offsets/sizes/strides.
---
 .../mlir/Interfaces/ValueBoundsOpInterface.h  |  57 ++++-
 .../SCF/IR/ValueBoundsOpInterfaceImpl.cpp     |  31 +--
 .../lib/Interfaces/ValueBoundsOpInterface.cpp | 237 +++++++++++++-----
 .../value-bounds-op-interface-impl.mlir       |  43 +++-
 .../SCF/value-bounds-op-interface-impl.mlir   |  12 +
 .../value-bounds-op-interface-impl.mlir       |  16 +-
 .../Dialect/Affine/TestReifyValueBounds.cpp   |  79 +++++-
 7 files changed, 361 insertions(+), 114 deletions(-)

diff --git a/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h b/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
index 3543ab52407a3..1d7bc6ea961cc 100644
--- a/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
+++ b/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
@@ -211,7 +211,8 @@ class ValueBoundsConstraintSet
   /// Comparison operator for `ValueBoundsConstraintSet::compare`.
   enum ComparisonOperator { LT, LE, EQ, GT, GE };
 
-  /// Try to prove that, based on the current state of this constraint set
+  /// Populate constraints for lhs/rhs (until the stop condition is met). Then,
+  /// try to prove that, based on the current state of this constraint set
   /// (i.e., without analyzing additional IR or adding new constraints), the
   /// "lhs" value/dim is LE/LT/EQ/GT/GE than the "rhs" value/dim.
   ///
@@ -220,24 +221,37 @@ class ValueBoundsConstraintSet
   /// proven. This could be because the specified relation does in fact not hold
   /// or because there is not enough information in the constraint set. In other
   /// words, if we do not know for sure, this function returns "false".
-  bool compare(Value lhs, std::optional<int64_t> lhsDim, ComparisonOperator cmp,
-               Value rhs, std::optional<int64_t> rhsDim);
+  bool populateAndCompare(OpFoldResult lhs, std::optional<int64_t> lhsDim,
+                          ComparisonOperator cmp, OpFoldResult rhs,
+                          std::optional<int64_t> rhsDim);
+
+  /// Return "true" if "lhs cmp rhs" was proven to hold. Return "false" if the
+  /// specified relation could not be proven. This could be because the
+  /// specified relation does in fact not hold or because there is not enough
+  /// information in the constraint set. In other words, if we do not know for
+  /// sure, this function returns "false".
+  ///
+  /// This function keeps traversing the backward slice of lhs/rhs until could
+  /// prove the relation or until it ran out of IR.
+  static bool compare(OpFoldResult lhs, std::optional<int64_t> lhsDim,
+                      ComparisonOperator cmp, OpFoldResult rhs,
+                      std::optional<int64_t> rhsDim);
+  static bool compare(AffineMap lhs, ValueDimList lhsOperands,
+                      ComparisonOperator cmp, AffineMap rhs,
+                      ValueDimList rhsOperands);
+  static bool compare(AffineMap lhs, ArrayRef<Value> lhsOperands,
+                      ComparisonOperator cmp, AffineMap rhs,
+                      ArrayRef<Value> rhsOperands);
 
   /// Compute whether the given values/dimensions are equal. Return "failure" if
   /// equality could not be determined.
   ///
   /// `dim1`/`dim2` must be `nullopt` if and only if `value1`/`value2` are
   /// index-typed.
-  static FailureOr<bool> areEqual(Value value1, Value value2,
+  static FailureOr<bool> areEqual(OpFoldResult value1, OpFoldResult value2,
                                   std::optional<int64_t> dim1 = std::nullopt,
                                   std::optional<int64_t> dim2 = std::nullopt);
 
-  /// Compute whether the given values/attributes are equal. Return "failure" if
-  /// equality could not be determined.
-  ///
-  /// `ofr1`/`ofr2` must be of index type.
-  static FailureOr<bool> areEqual(OpFoldResult ofr1, OpFoldResult ofr2);
-
   /// Return "true" if the given slices are guaranteed to be overlapping.
   /// Return "false" if the given slices are guaranteed to be non-overlapping.
   /// Return "failure" if unknown.
@@ -294,6 +308,20 @@ class ValueBoundsConstraintSet
 
   ValueBoundsConstraintSet(MLIRContext *ctx, StopConditionFn stopCondition);
 
+  /// Return "true" if, based on the current state of the constraint system,
+  /// "lhs cmp rhs" was proven to hold. Return "false" if the specified relation
+  /// could not be proven. This could be because the specified relation does in
+  /// fact not hold or because there is not enough information in the constraint
+  /// set. In other words, if we do not know for sure, this function returns
+  /// "false".
+  ///
+  /// This function does not analyze any IR and does not populate any additional
+  /// constraints.
+  bool compareValueDims(OpFoldResult lhs, std::optional<int64_t> lhsDim,
+                        ComparisonOperator cmp, OpFoldResult rhs,
+                        std::optional<int64_t> rhsDim);
+  bool comparePos(int64_t lhsPos, ComparisonOperator cmp, int64_t rhsPos);
+
   /// Given an affine map with a single result (and map operands), add a new
   /// column to the constraint set that represents the result of the map.
   /// Traverse additional IR starting from the map operands as needed (as long
@@ -319,6 +347,10 @@ class ValueBoundsConstraintSet
   /// set.
   AffineExpr getPosExpr(int64_t pos);
 
+  /// Return "true" if the given value/dim is mapped (i.e., has a corresponding
+  /// column in the constraint system).
+  bool isMapped(Value value, std::optional<int64_t> dim = std::nullopt) const;
+
   /// Insert a value/dimension into the constraint set. If `isSymbol` is set to
   /// "false", a dimension is added. The value/dimension is added to the
   /// worklist if `addToWorklist` is set.
@@ -338,6 +370,11 @@ class ValueBoundsConstraintSet
   /// dimensions but not for symbols.
   int64_t insert(bool isSymbol = true);
 
+  /// Insert the given affine map and its bound operands as a new column in the
+  /// constraint system. Return the position of the new column. Any operands
+  /// that were not analyzed yet are put on the worklist.
+  int64_t insert(AffineMap map, ValueDimList operands, bool isSymbol = true);
+
   /// Project out the given column in the constraint set.
   void projectOut(int64_t pos);
 
diff --git a/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp b/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
index 72c5aaa230678..087ffc438a830 100644
--- a/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
@@ -58,20 +58,11 @@ struct ForOpInterface
     Value iterArg = forOp.getRegionIterArg(iterArgIdx);
     Value initArg = forOp.getInitArgs()[iterArgIdx];
 
-    // Populate constraints for the yielded value.
-    cstr.populateConstraints(yieldedValue, dim);
-    // Populate constraints for the iter_arg. This is just to ensure that the
-    // iter_arg is mapped in the constraint set, which is a prerequisite for
-    // `compare`. It may lead to a recursive call to this function in case the
-    // iter_arg was not visited when the constraints for the yielded value were
-    // populated, but no additional work is done.
-    cstr.populateConstraints(iterArg, dim);
-
     // An EQ constraint can be added if the yielded value (dimension size)
     // equals the corresponding block argument (dimension size).
-    if (cstr.compare(yieldedValue, dim,
-                     ValueBoundsConstraintSet::ComparisonOperator::EQ, iterArg,
-                     dim)) {
+    if (cstr.populateAndCompare(
+            yieldedValue, dim, ValueBoundsConstraintSet::ComparisonOperator::EQ,
+            iterArg, dim)) {
       if (dim.has_value()) {
         cstr.bound(value)[*dim] == cstr.getExpr(initArg, dim);
       } else {
@@ -113,10 +104,6 @@ struct IfOpInterface
     Value thenValue = ifOp.thenYield().getResults()[resultNum];
     Value elseValue = ifOp.elseYield().getResults()[resultNum];
 
-    // Populate constraints for the yielded value (and all values on the
-    // backward slice, as long as the current stop condition is not satisfied).
-    cstr.populateConstraints(thenValue, dim);
-    cstr.populateConstraints(elseValue, dim);
     auto boundsBuilder = cstr.bound(value);
     if (dim)
       boundsBuilder[*dim];
@@ -125,9 +112,9 @@ struct IfOpInterface
     // If thenValue <= elseValue:
     // * result <= elseValue
     // * result >= thenValue
-    if (cstr.compare(thenValue, dim,
-                     ValueBoundsConstraintSet::ComparisonOperator::LE,
-                     elseValue, dim)) {
+    if (cstr.populateAndCompare(
+            thenValue, dim, ValueBoundsConstraintSet::ComparisonOperator::LE,
+            elseValue, dim)) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(thenValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(elseValue, dim);
@@ -139,9 +126,9 @@ struct IfOpInterface
     // If elseValue <= thenValue:
     // * result <= thenValue
     // * result >= elseValue
-    if (cstr.compare(elseValue, dim,
-                     ValueBoundsConstraintSet::ComparisonOperator::LE,
-                     thenValue, dim)) {
+    if (cstr.populateAndCompare(
+            elseValue, dim, ValueBoundsConstraintSet::ComparisonOperator::LE,
+            thenValue, dim)) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(elseValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(thenValue, dim);
diff --git a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
index 6e3d6dd3c7575..c138056ab41cc 100644
--- a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
+++ b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
@@ -202,6 +202,28 @@ int64_t ValueBoundsConstraintSet::insert(bool isSymbol) {
   return pos;
 }
 
+int64_t ValueBoundsConstraintSet::insert(AffineMap map, ValueDimList operands,
+                                         bool isSymbol) {
+  assert(map.getNumResults() == 1 && "expected affine map with one result");
+  int64_t pos = insert(/*isSymbol=*/false);
+
+  // Add map and operands to the constraint set. Dimensions are converted to
+  // symbols. All operands are added to the worklist (unless they were already
+  // processed).
+  auto mapper = [&](std::pair<Value, std::optional<int64_t>> v) {
+    return getExpr(v.first, v.second);
+  };
+  SmallVector<AffineExpr> dimReplacements = llvm::to_vector(
+      llvm::map_range(ArrayRef(operands).take_front(map.getNumDims()), mapper));
+  SmallVector<AffineExpr> symReplacements = llvm::to_vector(
+      llvm::map_range(ArrayRef(operands).drop_front(map.getNumDims()), mapper));
+  addBound(
+      presburger::BoundType::EQ, pos,
+      map.getResult(0).replaceDimsAndSymbols(dimReplacements, symReplacements));
+
+  return pos;
+}
+
 int64_t ValueBoundsConstraintSet::getPos(Value value,
                                          std::optional<int64_t> dim) const {
 #ifndef NDEBUG
@@ -224,6 +246,13 @@ AffineExpr ValueBoundsConstraintSet::getPosExpr(int64_t pos) {
              : builder.getAffineSymbolExpr(pos - cstr.getNumDimVars());
 }
 
+bool ValueBoundsConstraintSet::isMapped(Value value,
+                                        std::optional<int64_t> dim) const {
+  auto it =
+      valueDimToPosition.find(std::make_pair(value, dim.value_or(kIndexValue)));
+  return it != valueDimToPosition.end();
+}
+
 static Operation *getOwnerOfValue(Value value) {
   if (auto bbArg = dyn_cast<BlockArgument>(value))
     return bbArg.getOwner()->getParentOp();
@@ -560,27 +589,10 @@ void ValueBoundsConstraintSet::populateConstraints(Value value,
 
 int64_t ValueBoundsConstraintSet::populateConstraints(AffineMap map,
                                                       ValueDimList operands) {
-  assert(map.getNumResults() == 1 && "expected affine map with one result");
-  int64_t pos = insert(/*isSymbol=*/false);
-
-  // Add map and operands to the constraint set. Dimensions are converted to
-  // symbols. All operands are added to the worklist (unless they were already
-  // processed).
-  auto mapper = [&](std::pair<Value, std::optional<int64_t>> v) {
-    return getExpr(v.first, v.second);
-  };
-  SmallVector<AffineExpr> dimReplacements = llvm::to_vector(
-      llvm::map_range(ArrayRef(operands).take_front(map.getNumDims()), mapper));
-  SmallVector<AffineExpr> symReplacements = llvm::to_vector(
-      llvm::map_range(ArrayRef(operands).drop_front(map.getNumDims()), mapper));
-  addBound(
-      presburger::BoundType::EQ, pos,
-      map.getResult(0).replaceDimsAndSymbols(dimReplacements, symReplacements));
-
+  int64_t pos = insert(map, operands, /*isSymbol=*/false);
   // Process the backward slice of `operands` (i.e., reverse use-def chain)
   // until `stopCondition` is met.
   processWorklist();
-
   return pos;
 }
 
@@ -600,9 +612,18 @@ ValueBoundsConstraintSet::computeConstantDelta(Value value1, Value value2,
                               {{value1, dim1}, {value2, dim2}});
 }
 
-bool ValueBoundsConstraintSet::compare(Value lhs, std::optional<int64_t> lhsDim,
-                                       ComparisonOperator cmp, Value rhs,
-                                       std::optional<int64_t> rhsDim) {
+bool ValueBoundsConstraintSet::compareValueDims(OpFoldResult lhs,
+                                                std::optional<int64_t> lhsDim,
+                                                ComparisonOperator cmp,
+                                                OpFoldResult rhs,
+                                                std::optional<int64_t> rhsDim) {
+#ifndef NDEBUG
+  if (auto lhsVal = dyn_cast<Value>(lhs))
+    assertValidValueDim(lhsVal, lhsDim);
+  if (auto rhsVal = dyn_cast<Value>(rhs))
+    assertValidValueDim(rhsVal, rhsDim);
+#endif // NDEBUG
+
   // This function returns "true" if "lhs CMP rhs" is proven to hold.
   //
   // Example for ComparisonOperator::LE and index-typed values: We would like to
@@ -621,24 +642,32 @@ bool ValueBoundsConstraintSet::compare(Value lhs, std::optional<int64_t> lhsDim,
 
   // EQ can be expressed as LE and GE.
   if (cmp == EQ)
-    return compare(lhs, lhsDim, ComparisonOperator::LE, rhs, rhsDim) &&
-           compare(lhs, lhsDim, ComparisonOperator::GE, rhs, rhsDim);
+    return compareValueDims(lhs, lhsDim, ComparisonOperator::LE, rhs, rhsDim) &&
+           compareValueDims(lhs, lhsDim, ComparisonOperator::GE, rhs, rhsDim);
 
   // Construct inequality. For the above example: lhs > rhs.
   // `IntegerRelation` inequalities are expressed in the "flattened" form and
   // with ">= 0". I.e., lhs - rhs - 1 >= 0.
-  SmallVector<int64_t> eq(cstr.getNumDimAndSymbolVars() + 1, 0);
+  SmallVector<int64_t> eq(cstr.getNumCols(), 0);
+  auto addToEq = [&](OpFoldResult ofr, std::optional<int64_t> dim,
+                     int64_t factor) {
+    if (auto constVal = ::getConstantIntValue(ofr)) {
+      eq[cstr.getNumCols() - 1] += *constVal * factor;
+    } else {
+      eq[getPos(cast<Value>(ofr), dim)] += factor;
+    }
+  };
   if (cmp == LT || cmp == LE) {
-    ++eq[getPos(lhs, lhsDim)];
-    --eq[getPos(rhs, rhsDim)];
+    addToEq(lhs, lhsDim, 1);
+    addToEq(rhs, rhsDim, -1);
   } else if (cmp == GT || cmp == GE) {
-    --eq[getPos(lhs, lhsDim)];
-    ++eq[getPos(rhs, rhsDim)];
+    addToEq(lhs, lhsDim, -1);
+    addToEq(rhs, rhsDim, 1);
   } else {
     llvm_unreachable("unsupported comparison operator");
   }
   if (cmp == LE || cmp == GE)
-    eq[cstr.getNumDimAndSymbolVars()] -= 1;
+    eq[cstr.getNumCols() - 1] -= 1;
 
   // Add inequality to the constraint set and check if it made the constraint
   // set empty.
@@ -649,40 +678,128 @@ bool ValueBoundsConstraintSet::compare(Value lhs, std::optional<int64_t> lhsDim,
   return isEmpty;
 }
 
+bool ValueBoundsConstraintSet::comparePos(int64_t lhsPos,
+                                          ComparisonOperator cmp,
+                                          int64_t rhsPos) {
+  // This function returns "true" if "lhs CMP rhs" is proven to hold. For
+  // detailed documentation, see `compareValueDims`.
+
+  // EQ can be expressed as LE and GE.
+  if (cmp == EQ)
+    return comparePos(lhsPos, ComparisonOperator::LE, rhsPos) &&
+           comparePos(lhsPos, ComparisonOperator::GE, rhsPos);
+
+  // Construct inequality.
+  SmallVector<int64_t> eq(cstr.getNumCols(), 0);
+  if (cmp == LT || cmp == LE) {
+    ++eq[lhsPos];
+    --eq[rhsPos];
+  } else if (cmp == GT || cmp == GE) {
+    --eq[lhsPos];
+    ++eq[rhsPos];
+  } else {
+    llvm_unreachable("unsupported comparison operator");
+  }
+  if (cmp == LE || cmp == GE)
+    eq[cstr.getNumCols() - 1] -= 1;
+
+  // Add inequality to the constraint set and check if it made the constraint
+  // set empty.
+  int64_t ineqPos = cstr.getNumInequalities();
+  cstr.addInequality(eq);
+  bool isEmpty = cstr.isEmpty();
+  cstr.removeInequality(ineqPos);
+  return isEmpty;
+}
+
+bool ValueBoundsConstraintSet::populateAndCompare(
+    OpFoldResult lhs, std::optional<int64_t> lhsDim, ComparisonOperator cmp,
+    OpFoldResult rhs, std::optional<int64_t> rhsDim) {
+#ifndef NDEBUG
+  if (auto lhsVal = dyn_cast<Value>(lhs))
+    assertValidValueDim(lhsVal, lhsDim);
+  if (auto rhsVal = dyn_cast<Value>(rhs))
+    assertValidValueDim(rhsVal, rhsDim);
+#endif // NDEBUG
+
+  if (auto lhsVal = dyn_cast<Value>(lhs))
+    populateConstraints(lhsVal, lhsDim);
+  if (auto rhsVal = dyn_cast<Value>(rhs))
+    populateConstraints(rhsVal, rhsDim);
+
+  return compareValueDims(lhs, lhsDim, cmp, rhs, rhsDim);
+}
+
+bool ValueBoundsConstraintSet::compare(OpFoldResult lhs,
+                                       std::optional<int64_t> lhsDim,
+                                       ComparisonOperator cmp, OpFoldResult rhs,
+                                       std::optional<int64_t> rhsDim) {
+  auto stopCondition = [&](Value v, std::optional<int64_t> dim,
+                           ValueBoundsConstraintSet &cstr) {
+    // Keep processing as long as lhs/rhs are not mapped.
+    if (auto lhsVal = dyn_cast<Value>(lhs))
+      if (!cstr.isMapped(lhsVal, dim))
+        return false;
+    if (auto rhsVal = dyn_cast<Value>(rhs))
+      if (!cstr.isMapped(rhsVal, dim))
+        return false;
+    // Keep processing as long as the relation cannot be proven.
+    return cstr.compareValueDims(lhs, lhsDim, cmp, rhs, rhsDim);
+  };
+
+  ValueBoundsConstraintSet cstr(lhs.getContext(), stopCondition);
+  return cstr.populateAndCompare(lhs, lhsDim, cmp, rhs, rhsDim);
+}
+
+bool ValueBoundsConstraintSet::compare(AffineMap lhs, ValueDimList lhsOperands,
+                                       ComparisonOperator cmp, AffineMap rhs,
+                                       ValueDimList rhsOperands) {
+  int64_t lhsPos = -1, rhsPos = -1;
+  auto stopCondition = [&](Value v, std::optional<int64_t> dim,
+                           ValueBoundsConstraintSet &cstr) {
+    // Keep processing as long as lhs/rhs were not processed.
+    if (lhsPos >= cstr.positionToValueDim.size() ||
+        rhsPos >= cstr.positionToValueDim.size())
+      return false;
+    // Keep processing as long as the relation cannot be proven.
+    return cstr.comparePos(lhsPos, cmp, rhsPos);
+  };
+  ValueBoundsConstraintSet cstr(lhs.getContext(), stopCondition);
+  lhsPos = cstr.insert(lhs, lhsOperands);
+  rhsPos = cstr.insert(rhs, rhsOperands);
+  cstr.processWorklist();
+  return cstr.comparePos(lhsPos, cmp, rhsPos);
+}
+
+bool ValueBoundsConstraintSet::compare(AffineMap lhs,
+                                       ArrayRef<Value> lhsOperands,
+                                       ComparisonOperator cmp, AffineMap rhs,
+                                       ArrayRef<Value> rhsOperands) {
+  ValueDimList lhsValueDimOperands =
+      llvm::map_to_vector(lhsOperands, [](Value v) {
+        return std::make_pair(v, std::optional<int64_t>());
+      });
+  ValueDimList rhsValueDimOperands =
+      llvm::map_to_vector(rhsOperands, [](Value v) {
+        return std::make_pair(v, std::optional<int64_t>());
+      });
+  return ValueBoundsConstraintSet::compare(lhs, lhsValueDimOperands, cmp, rhs,
+                                           rhsValueDimOperands);
+}
+
 FailureOr<bool>
-ValueBoundsConstraintSet::areEqual(Value value1, Value value2,
+ValueBoundsConstraintSet::areEqual(OpFoldResult value1, OpFoldResult value2,
                                    std::optional<int64_t> dim1,
                                    std::optional<int64_t> dim2) {
-  // Subtract the two values/dimensions from each other. If the result is 0,
-  // both are equal.
-  FailureOr<int64_t> delta = computeConstantDelta(value1, value2, dim1, dim2);
-  if (failed(delta))
-    return failure();
-  return *delta == 0;
-}
-
-FailureOr<bool> ValueBoundsConstraintSet::areEqual(OpFoldResult ofr1,
-                                                   OpFoldResult ofr2) {
-  Builder b(ofr1.getContext());
-  AffineMap map =
-      AffineMap::get(/*dimCount=*/0, /*symbolCount=*/2,
-                     b.getAffineSymbolExpr(0) - b.getAffineSymbolExpr(1));
-  SmallVector<OpFoldResult> ofrOperands;
-  ofrOperands.push_back(ofr1);
-  ofrOperands.push_back(ofr2);
-  SmallVector<Value> valueOperands;
-  AffineMap foldedMap =
-      foldAttributesIntoMap(b, map, ofrOperands, valueOperands);
-  ValueDimList valueDims;
-  for (Value v : valueOperands) {
-    assert(v.getType().isIndex() && "expected index type");
-    valueDims.emplace_back(v, std::nullopt);
-  }
-  FailureOr<int64_t> delta =
-      computeConstantBound(presburger::BoundType::EQ, foldedMap, valueDims);
-  if (failed(delta))
-    return failure();
-  return *delta == 0;
+  if (ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::EQ,
+                                        value2, dim2))
+    return true;
+  if (ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::LT,
+                                        value2, dim2) ||
+      ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::GT,
+                                        value2, dim2))
+    return false;
+  return failure();
 }
 
 FailureOr<bool>
diff --git a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
index 55282e8334abd..10da91870f49d 100644
--- a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
@@ -79,6 +79,17 @@ func.func @composed_affine_apply(%i1 : index) -> (index) {
 }
 
 
+// -----
+
+func.func @are_equal(%i1 : index) {
+  %i2 = affine.apply affine_map<(d0) -> ((d0 floordiv 32) * 16)>(%i1)
+  %i3 = affine.apply affine_map<(d0) -> ((d0 floordiv 32) * 16 + 8)>(%i1)
+  %s = affine.apply affine_map<()[s0, s1] -> (s0 - s1)>()[%i2, %i3]
+  // expected-remark @below{{false}}
+   "test.compare"(%i2, %i3) : (index, index) -> ()
+  return
+}
+
 // -----
 
 // Test for affine::fullyComposeAndCheckIfEqual
@@ -87,6 +98,36 @@ func.func @composed_are_equal(%i1 : index) {
   %i3 = affine.apply affine_map<(d0) -> ((d0 floordiv 32) * 16 + 8)>(%i1)
   %s = affine.apply affine_map<()[s0, s1] -> (s0 - s1)>()[%i2, %i3]
   // expected-remark @below{{different}}
-   "test.are_equal"(%i2, %i3) {compose} : (index, index) -> ()
+   "test.compare"(%i2, %i3) {compose} : (index, index) -> ()
+  return
+}
+
+// -----
+
+func.func @compare_affine_max(%a: index, %b: index) {
+  %0 = affine.max affine_map<()[s0, s1] -> (s0, s1)>()[%a, %b]
+  // expected-remark @below{{true}}
+  "test.compare"(%0, %a) {cmp = "GE"} : (index, index) -> ()
+  // expected-error @below{{unknown}}
+  "test.compare"(%0, %a) {cmp = "GT"} : (index, index) -> ()
+  // expected-remark @below{{false}}
+  "test.compare"(%0, %a) {cmp = "LT"} : (index, index) -> ()
+  // expected-error @below{{unknown}}
+  "test.compare"(%0, %a) {cmp = "LE"} : (index, index) -> ()
+  return
+}
+
+// -----
+
+func.func @compare_affine_min(%a: index, %b: index) {
+  %0 = affine.min affine_map<()[s0, s1] -> (s0, s1)>()[%a, %b]
+  // expected-error @below{{unknown}}
+  "test.compare"(%0, %a) {cmp = "GE"} : (index, index) -> ()
+  // expected-remark @below{{false}}
+  "test.compare"(%0, %a) {cmp = "GT"} : (index, index) -> ()
+  // expected-error @below{{unknown}}
+  "test.compare"(%0, %a) {cmp = "LT"} : (index, index) -> ()
+  // expected-remark @below{{true}}
+  "test.compare"(%0, %a) {cmp = "LE"} : (index, index) -> ()
   return
 }
diff --git a/mlir/test/Dialect/SCF/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/SCF/value-bounds-op-interface-impl.mlir
index 0ea06737886d4..9ab03da1c9a94 100644
--- a/mlir/test/Dialect/SCF/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/SCF/value-bounds-op-interface-impl.mlir
@@ -219,3 +219,15 @@ func.func @scf_if_eq(%a: index, %b: index, %c : i1) {
   "test.some_use"(%reify1) : (index) -> ()
   return
 }
+
+// -----
+
+func.func @compare_scf_for(%a: index, %b: index, %c: index) {
+  scf.for %iv = %a to %b step %c {
+    // expected-remark @below{{true}}
+    "test.compare"(%iv, %a) {cmp = "GE"} : (index, index) -> ()
+    // expected-remark @below{{true}}
+    "test.compare"(%iv, %b) {cmp = "LT"} : (index, index) -> ()
+  }
+  return
+}
diff --git a/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
index 45520da6aeb0b..0c90bcdb42028 100644
--- a/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
@@ -163,8 +163,8 @@ func.func @dynamic_dims_are_equal(%t: tensor<?xf32>) {
   %c0 = arith.constant 0 : index
   %dim0 = tensor.dim %t, %c0 : tensor<?xf32>
   %dim1 = tensor.dim %t, %c0 : tensor<?xf32>
-  // expected-remark @below {{equal}}
-  "test.are_equal"(%dim0, %dim1) : (index, index) -> ()
+  // expected-remark @below {{true}}
+  "test.compare"(%dim0, %dim1) : (index, index) -> ()
   return
 }
 
@@ -175,8 +175,8 @@ func.func @dynamic_dims_are_different(%t: tensor<?xf32>) {
   %c1 = arith.constant 1 : index
   %dim0 = tensor.dim %t, %c0 : tensor<?xf32>
   %val = arith.addi %dim0, %c1 : index
-  // expected-remark @below {{different}}
-  "test.are_equal"(%dim0, %val) : (index, index) -> ()
+  // expected-remark @below {{false}}
+  "test.compare"(%dim0, %val) : (index, index) -> ()
   return
 }
 
@@ -186,8 +186,8 @@ func.func @dynamic_dims_are_maybe_equal_1(%t: tensor<?xf32>) {
   %c0 = arith.constant 0 : index
   %c5 = arith.constant 5 : index
   %dim0 = tensor.dim %t, %c0 : tensor<?xf32>
-  // expected-error @below {{could not determine equality}}
-  "test.are_equal"(%dim0, %c5) : (index, index) -> ()
+  // expected-error @below {{unknown}}
+  "test.compare"(%dim0, %c5) : (index, index) -> ()
   return
 }
 
@@ -198,7 +198,7 @@ func.func @dynamic_dims_are_maybe_equal_2(%t: tensor<?x?xf32>) {
   %c1 = arith.constant 1 : index
   %dim0 = tensor.dim %t, %c0 : tensor<?x?xf32>
   %dim1 = tensor.dim %t, %c1 : tensor<?x?xf32>
-  // expected-error @below {{could not determine equality}}
-  "test.are_equal"(%dim0, %dim1) : (index, index) -> ()
+  // expected-error @below {{unknown}}
+  "test.compare"(%dim0, %dim1) : (index, index) -> ()
   return
 }
diff --git a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
index 4b2b1a06341b7..f38631054fb3c 100644
--- a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
+++ b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
@@ -57,7 +57,7 @@ struct TestReifyValueBounds
 
 } // namespace
 
-FailureOr<BoundType> parseBoundType(const std::string &type) {
+static FailureOr<BoundType> parseBoundType(const std::string &type) {
   if (type == "EQ")
     return BoundType::EQ;
   if (type == "LB")
@@ -67,6 +67,34 @@ FailureOr<BoundType> parseBoundType(const std::string &type) {
   return failure();
 }
 
+static FailureOr<ValueBoundsConstraintSet::ComparisonOperator>
+parseComparisonOperator(const std::string &type) {
+  if (type == "EQ")
+    return ValueBoundsConstraintSet::ComparisonOperator::EQ;
+  if (type == "LT")
+    return ValueBoundsConstraintSet::ComparisonOperator::LT;
+  if (type == "LE")
+    return ValueBoundsConstraintSet::ComparisonOperator::LE;
+  if (type == "GT")
+    return ValueBoundsConstraintSet::ComparisonOperator::GT;
+  if (type == "GE")
+    return ValueBoundsConstraintSet::ComparisonOperator::GE;
+  return failure();
+}
+
+static ValueBoundsConstraintSet::ComparisonOperator
+invertComparisonOperator(ValueBoundsConstraintSet::ComparisonOperator cmp) {
+  if (cmp == ValueBoundsConstraintSet::ComparisonOperator::LT)
+    return ValueBoundsConstraintSet::ComparisonOperator::GE;
+  if (cmp == ValueBoundsConstraintSet::ComparisonOperator::LE)
+    return ValueBoundsConstraintSet::ComparisonOperator::GT;
+  if (cmp == ValueBoundsConstraintSet::ComparisonOperator::GT)
+    return ValueBoundsConstraintSet::ComparisonOperator::LE;
+  if (cmp == ValueBoundsConstraintSet::ComparisonOperator::GE)
+    return ValueBoundsConstraintSet::ComparisonOperator::LT;
+  llvm_unreachable("unsupported comparison operator");
+}
+
 /// Look for "test.reify_bound" ops in the input and replace their results with
 /// the reified values.
 static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
@@ -215,18 +243,34 @@ static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
   return failure(result.wasInterrupted());
 }
 
-/// Look for "test.are_equal" ops and emit errors/remarks.
+/// Look for "test.compare" ops and emit errors/remarks.
 static LogicalResult testEquality(func::FuncOp funcOp) {
   IRRewriter rewriter(funcOp.getContext());
   WalkResult result = funcOp.walk([&](Operation *op) {
-    // Look for test.are_equal ops.
-    if (op->getName().getStringRef() == "test.are_equal") {
+    // Look for test.compare ops.
+    if (op->getName().getStringRef() == "test.compare") {
       if (op->getNumOperands() != 2 || !op->getOperand(0).getType().isIndex() ||
           !op->getOperand(1).getType().isIndex()) {
         op->emitOpError("invalid op");
         return WalkResult::skip();
       }
+
+      // Get comparison operator.
+      std::string cmpStr = "EQ";
+      if (auto cmpAttr = op->getAttrOfType<StringAttr>("cmp"))
+        cmpStr = cmpAttr.str();
+      auto cmpType = parseComparisonOperator(cmpStr);
+      if (failed(cmpType)) {
+        op->emitOpError("invalid comparison operator");
+        return WalkResult::interrupt();
+      }
+
       if (op->hasAttr("compose")) {
+        if (cmpType != ValueBoundsConstraintSet::EQ) {
+          op->emitOpError(
+              "comparison operator must be EQ when 'composed' is specified");
+          return WalkResult::interrupt();
+        }
         FailureOr<int64_t> delta = affine::fullyComposeAndComputeConstantDelta(
             op->getOperand(0), op->getOperand(1));
         if (failed(delta)) {
@@ -236,16 +280,25 @@ static LogicalResult testEquality(func::FuncOp funcOp) {
         } else {
           op->emitRemark("different");
         }
+        return WalkResult::advance();
+      }
+
+      auto compare = [&](ValueBoundsConstraintSet::ComparisonOperator cmp) {
+        return ValueBoundsConstraintSet::compare(
+            /*lhs=*/op->getOperand(0), /*lhsDim=*/std::nullopt, cmp,
+            /*rhs=*/op->getOperand(1), /*rhsDim=*/std::nullopt);
+      };
+      if (compare(*cmpType)) {
+        op->emitRemark("true");
+      } else if (*cmpType != ValueBoundsConstraintSet::EQ &&
+                 compare(invertComparisonOperator(*cmpType))) {
+        op->emitRemark("false");
+      } else if (*cmpType == ValueBoundsConstraintSet::EQ &&
+                 (compare(ValueBoundsConstraintSet::ComparisonOperator::LT) ||
+                  compare(ValueBoundsConstraintSet::ComparisonOperator::GT))) {
+        op->emitRemark("false");
       } else {
-        FailureOr<bool> equal = ValueBoundsConstraintSet::areEqual(
-            op->getOperand(0), op->getOperand(1));
-        if (failed(equal)) {
-          op->emitError("could not determine equality");
-        } else if (*equal) {
-          op->emitRemark("equal");
-        } else {
-          op->emitRemark("different");
-        }
+        op->emitError("unknown");
       }
     }
     return WalkResult::advance();

From 21265f692e4b3b2146b6095cf23122b20e8fa0ed Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Thu, 11 Apr 2024 08:27:12 +0200
Subject: [PATCH 104/886] [mlir][Interfaces] `ValueBoundsOpInterface`: Fix typo
 (#87976)

This was likely a copy-and-paste typo.
---
 mlir/lib/Interfaces/ValueBoundsOpInterface.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
index c138056ab41cc..fa66da4a0def9 100644
--- a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
+++ b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
@@ -205,7 +205,7 @@ int64_t ValueBoundsConstraintSet::insert(bool isSymbol) {
 int64_t ValueBoundsConstraintSet::insert(AffineMap map, ValueDimList operands,
                                          bool isSymbol) {
   assert(map.getNumResults() == 1 && "expected affine map with one result");
-  int64_t pos = insert(/*isSymbol=*/false);
+  int64_t pos = insert(isSymbol);
 
   // Add map and operands to the constraint set. Dimensions are converted to
   // symbols. All operands are added to the worklist (unless they were already

From dc39028906ba4196c3ba544c43ef6b428cf47c51 Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Thu, 11 Apr 2024 14:50:40 +0800
Subject: [PATCH 105/886] [mlir] Fix -Wsign-compare in
 ValueBoundsOpInterface.cpp (NFC)

/llvm-project/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp:762:16:
error: comparison of integers of different signs: 'int64_t' (aka 'long') and 'size_t' (aka 'unsigned long') [-Werror,-Wsign-compare]
        rhsPos >= cstr.positionToValueDim.size())
        ~~~~~~ ^  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/llvm-project/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp:761:16:
error: comparison of integers of different signs: 'int64_t' (aka 'long') and 'size_t' (aka 'unsigned long') [-Werror,-Wsign-compare]
    if (lhsPos >= cstr.positionToValueDim.size() ||
        ~~~~~~ ^  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2 errors generated.
---
 mlir/lib/Interfaces/ValueBoundsOpInterface.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
index fa66da4a0def9..ffa4c0b55cad7 100644
--- a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
+++ b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
@@ -758,8 +758,8 @@ bool ValueBoundsConstraintSet::compare(AffineMap lhs, ValueDimList lhsOperands,
   auto stopCondition = [&](Value v, std::optional<int64_t> dim,
                            ValueBoundsConstraintSet &cstr) {
     // Keep processing as long as lhs/rhs were not processed.
-    if (lhsPos >= cstr.positionToValueDim.size() ||
-        rhsPos >= cstr.positionToValueDim.size())
+    if (size_t(lhsPos) >= cstr.positionToValueDim.size() ||
+        size_t(rhsPos) >= cstr.positionToValueDim.size())
       return false;
     // Keep processing as long as the relation cannot be proven.
     return cstr.comparePos(lhsPos, cmp, rhsPos);

From 3f7f446d3803a699f5964a7429c6e1de0d783452 Mon Sep 17 00:00:00 2001
From: Haohai Wen <haohai.wen@intel.com>
Date: Thu, 11 Apr 2024 15:28:32 +0800
Subject: [PATCH 106/886] [llvm-profgen] Remove temporary perf script files
 (#86668)

The temporary perf script files converted from perf data will occupy
lots
of space for large project. This patch removes them when llvm-profgen
exits normally or receives signals.
---
 llvm/tools/llvm-profgen/PerfReader.cpp | 6 ++++++
 llvm/tools/llvm-profgen/PerfReader.h   | 8 ++++++++
 2 files changed, 14 insertions(+)

diff --git a/llvm/tools/llvm-profgen/PerfReader.cpp b/llvm/tools/llvm-profgen/PerfReader.cpp
index 878147642aa6e..e9442027aed3f 100644
--- a/llvm/tools/llvm-profgen/PerfReader.cpp
+++ b/llvm/tools/llvm-profgen/PerfReader.cpp
@@ -11,6 +11,7 @@
 #include "llvm/DebugInfo/Symbolize/SymbolizableModule.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Process.h"
+#include "llvm/Support/ToolOutputFile.h"
 
 #define DEBUG_TYPE "perf-reader"
 
@@ -375,6 +376,9 @@ PerfScriptReader::convertPerfDataToTrace(ProfiledBinary *Binary,
                                           StringRef(ErrorFile)};    // Stderr
   sys::ExecuteAndWait(PerfPath, ScriptMMapArgs, std::nullopt, Redirects);
 
+  PerfScriptReader::TempFileCleanups.emplace_back(PerfTraceFile);
+  PerfScriptReader::TempFileCleanups.emplace_back(ErrorFile);
+
   // Collect the PIDs
   TraceStream TraceIt(PerfTraceFile);
   std::string PIDs;
@@ -1220,5 +1224,7 @@ void PerfScriptReader::parsePerfTraces() {
     writeUnsymbolizedProfile(OutputFilename);
 }
 
+SmallVector<CleanupInstaller, 2> PerfScriptReader::TempFileCleanups;
+
 } // end namespace sampleprof
 } // end namespace llvm
diff --git a/llvm/tools/llvm-profgen/PerfReader.h b/llvm/tools/llvm-profgen/PerfReader.h
index e9f619350bf97..b821cbe13efae 100644
--- a/llvm/tools/llvm-profgen/PerfReader.h
+++ b/llvm/tools/llvm-profgen/PerfReader.h
@@ -21,6 +21,9 @@ using namespace llvm;
 using namespace sampleprof;
 
 namespace llvm {
+
+class CleanupInstaller;
+
 namespace sampleprof {
 
 // Stream based trace line iterator
@@ -604,6 +607,11 @@ class PerfScriptReader : public PerfReaderBase {
   // Extract perf script type by peaking at the input
   static PerfContent checkPerfScriptType(StringRef FileName);
 
+  // Cleanup installers for temporary files created by perf script command.
+  // Those files will be automatically removed when running destructor or
+  // receiving signals.
+  static SmallVector<CleanupInstaller, 2> TempFileCleanups;
+
 protected:
   // The parsed MMap event
   struct MMapEvent {

From a53674359da8507af539bf879e1b8292e3720eb8 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Thu, 11 Apr 2024 08:45:28 +0100
Subject: [PATCH 107/886] [AArch64] Add ZIP and UZP shuffle costs. (#88150)

This adds some costs for the shuffle instructions that should be lowered
to zip1/zip2/uzp1/uzp2 instructions.
---
 .../Target/AArch64/AArch64ISelLowering.cpp    | 29 --------
 .../Target/AArch64/AArch64PerfectShuffle.h    | 33 ++++++++++
 .../AArch64/AArch64TargetTransformInfo.cpp    | 10 +++
 .../CostModel/AArch64/shuffle-other.ll        | 66 +++++++++----------
 .../CostModel/AArch64/shuffle-store.ll        | 14 ++--
 .../AArch64/vecreduce-shuffle.ll              |  2 +-
 6 files changed, 84 insertions(+), 70 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 744b2cdef504d..80181a77c9d23 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -11851,35 +11851,6 @@ static bool isREVMask(ArrayRef<int> M, EVT VT, unsigned BlockSize) {
   return true;
 }
 
-static bool isZIPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
-  unsigned NumElts = VT.getVectorNumElements();
-  if (NumElts % 2 != 0)
-    return false;
-  WhichResult = (M[0] == 0 ? 0 : 1);
-  unsigned Idx = WhichResult * NumElts / 2;
-  for (unsigned i = 0; i != NumElts; i += 2) {
-    if ((M[i] >= 0 && (unsigned)M[i] != Idx) ||
-        (M[i + 1] >= 0 && (unsigned)M[i + 1] != Idx + NumElts))
-      return false;
-    Idx += 1;
-  }
-
-  return true;
-}
-
-static bool isUZPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
-  unsigned NumElts = VT.getVectorNumElements();
-  WhichResult = (M[0] == 0 ? 0 : 1);
-  for (unsigned i = 0; i != NumElts; ++i) {
-    if (M[i] < 0)
-      continue; // ignore UNDEF indices
-    if ((unsigned)M[i] != 2 * i + WhichResult)
-      return false;
-  }
-
-  return true;
-}
-
 static bool isTRNMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
   unsigned NumElts = VT.getVectorNumElements();
   if (NumElts % 2 != 0)
diff --git a/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h b/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
index 5846fd454b654..7abaead694d11 100644
--- a/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
+++ b/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
@@ -16,6 +16,8 @@
 
 #include "llvm/ADT/ArrayRef.h"
 
+namespace llvm {
+
 // 31 entries have cost 0
 // 756 entries have cost 1
 // 3690 entries have cost 2
@@ -6618,4 +6620,35 @@ static unsigned getPerfectShuffleCost(llvm::ArrayRef<int> M) {
   return (PFEntry >> 30) + 1;
 }
 
+inline bool isZIPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
+  unsigned NumElts = VT.getVectorNumElements();
+  if (NumElts % 2 != 0)
+    return false;
+  WhichResult = (M[0] == 0 ? 0 : 1);
+  unsigned Idx = WhichResult * NumElts / 2;
+  for (unsigned i = 0; i != NumElts; i += 2) {
+    if ((M[i] >= 0 && (unsigned)M[i] != Idx) ||
+        (M[i + 1] >= 0 && (unsigned)M[i + 1] != Idx + NumElts))
+      return false;
+    Idx += 1;
+  }
+
+  return true;
+}
+
+inline bool isUZPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
+  unsigned NumElts = VT.getVectorNumElements();
+  WhichResult = (M[0] == 0 ? 0 : 1);
+  for (unsigned i = 0; i != NumElts; ++i) {
+    if (M[i] < 0)
+      continue; // ignore UNDEF indices
+    if ((unsigned)M[i] != 2 * i + WhichResult)
+      return false;
+  }
+
+  return true;
+}
+
+} // namespace llvm
+
 #endif
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 20150d7386753..bd943de06b4b2 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -3932,6 +3932,16 @@ InstructionCost AArch64TTIImpl::getShuffleCost(
       }))
     return 0;
 
+  // Check for other shuffles that are not SK_ kinds but we have native
+  // instructions for, for example ZIP and UZP.
+  unsigned Unused;
+  if (LT.second.isFixedLengthVector() &&
+      LT.second.getVectorNumElements() == Mask.size() &&
+      (Kind == TTI::SK_PermuteTwoSrc || Kind == TTI::SK_PermuteSingleSrc) &&
+      (isZIPMask(Mask, LT.second, Unused) ||
+       isUZPMask(Mask, LT.second, Unused)))
+    return 1;
+
   if (Kind == TTI::SK_Broadcast || Kind == TTI::SK_Transpose ||
       Kind == TTI::SK_Select || Kind == TTI::SK_PermuteSingleSrc ||
       Kind == TTI::SK_Reverse || Kind == TTI::SK_Splice) {
diff --git a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
index d469ce6305932..6c45ebcb69f40 100644
--- a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
@@ -171,28 +171,28 @@ define void @zip() {
 ; CHECK-LABEL: 'zip'
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 3>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %zipv2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %zip1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %zip2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %zip1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %zip2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %zipv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zipv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zipv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zip1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zip2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %zipv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zipv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zip1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zip2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %zipv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
@@ -276,24 +276,24 @@ define void @zip() {
 
 define void @uzp() {
 ; CHECK-LABEL: 'uzp'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %uzp1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %uzp2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %uzp1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %uzp2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %uzpv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzpv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzpv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzp1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzp2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %uzpv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzpv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzp1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzp2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %uzpv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v4i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v4i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzpv4i32 = shufflevector <4 x i32> undef, <4 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
@@ -360,10 +360,10 @@ define void @uzp() {
 
 define void @multipart() {
 ; CHECK-LABEL: 'multipart'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v16a = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16a = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16b = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %v16c = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %v16d = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16c = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16d = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32a = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32a4 = shufflevector <16 x i32> undef, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32idrev = shufflevector <16 x i32> undef, <16 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 31, i32 30, i32 29, i32 28>
diff --git a/llvm/test/Analysis/CostModel/AArch64/shuffle-store.ll b/llvm/test/Analysis/CostModel/AArch64/shuffle-store.ll
index 12de334574f5c..cd434afddc9a9 100644
--- a/llvm/test/Analysis/CostModel/AArch64/shuffle-store.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/shuffle-store.ll
@@ -5,21 +5,21 @@ target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
 
 define void @vst2(ptr %p) {
 ; CHECK-LABEL: 'vst2'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: store <4 x i8> %v4i8, ptr %p, align 4
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <8 x i8> %v8i8, ptr %p, align 8
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <16 x i8> %v16i8, ptr %p, align 16
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %v32i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: store <32 x i8> %v32i8, ptr %p, align 32
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <4 x i16> %v4i16, ptr %p, align 8
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <8 x i16> %v8i16, ptr %p, align 16
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: store <16 x i16> %v16i16, ptr %p, align 32
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %v32i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: store <32 x i16> %v32i16, ptr %p, align 64
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: store <4 x i32> %v4i32, ptr %p, align 16
diff --git a/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll b/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
index c505cb7b181c2..d69cb75664a8c 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
@@ -416,7 +416,7 @@ define i16 @reduceshuffle_twoin_notlowelt_v16i16(<16 x i16> %a, <16 x i16> %b) {
 
 define i16 @reduceshuffle_twoin_uneven_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; CHECK-LABEL: @reduceshuffle_twoin_uneven_v16i16(
-; CHECK-NEXT:    [[S:%.*]] = shufflevector <16 x i16> [[A:%.*]], <16 x i16> [[B:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22>
+; CHECK-NEXT:    [[S:%.*]] = shufflevector <16 x i16> [[A:%.*]], <16 x i16> [[B:%.*]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 8>
 ; CHECK-NEXT:    [[X:%.*]] = xor <16 x i16> [[S]], <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
 ; CHECK-NEXT:    [[R:%.*]] = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> [[X]])
 ; CHECK-NEXT:    ret i16 [[R]]

From 85bc6de67ef28cd203da0c5abc1485609bea989c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Wed, 10 Apr 2024 11:18:11 +0200
Subject: [PATCH 108/886] Revert "Use setup_host_tool for clang-ast-dump, fixes
 76707"

This reverts commit b4adb42151bbfa80be4cf6d076cbe5edf680693e.

The original commit increased local rebuild times a lot. See
the discussion in
https://github.com/llvm/llvm-project/issues/76707
---
 clang/lib/Tooling/CMakeLists.txt | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/clang/lib/Tooling/CMakeLists.txt b/clang/lib/Tooling/CMakeLists.txt
index 8b4ab0e212964..91e6cbdcbc44f 100644
--- a/clang/lib/Tooling/CMakeLists.txt
+++ b/clang/lib/Tooling/CMakeLists.txt
@@ -53,16 +53,14 @@ else()
     list(APPEND implicitDirs -I ${implicitDir})
   endforeach()
 
-  setup_host_tool(clang-ast-dump CLANG_AST_DUMP clang_ast_dump_exe clang_ast_dump_target)
-
   include(GetClangResourceDir)
   get_clang_resource_dir(resource_dir PREFIX ${LLVM_BINARY_DIR})
   add_custom_command(
       COMMENT Generate ASTNodeAPI.json
       OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/ASTNodeAPI.json
-      DEPENDS ${clang_ast_dump_target} clang-resource-headers
+      DEPENDS clang-ast-dump clang-resource-headers
       COMMAND
-      ${clang_ast_dump_exe}
+      $<TARGET_FILE:clang-ast-dump>
         # Skip this in debug mode because parsing AST.h is too slow
         --skip-processing=${skip_expensive_processing}
         -I ${resource_dir}/include

From d7e0ea205fa111fba46e08f3df2860f76b47acb6 Mon Sep 17 00:00:00 2001
From: Chen Zheng <czhengsz@cn.ibm.com>
Date: Thu, 11 Apr 2024 03:33:04 -0400
Subject: [PATCH 109/886] [PowerPC] add testcase for a xxinsertw bug, NFC

---
 llvm/test/CodeGen/PowerPC/xxinsertw.ll | 36 ++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
 create mode 100644 llvm/test/CodeGen/PowerPC/xxinsertw.ll

diff --git a/llvm/test/CodeGen/PowerPC/xxinsertw.ll b/llvm/test/CodeGen/PowerPC/xxinsertw.ll
new file mode 100644
index 0000000000000..b48eac06a694a
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/xxinsertw.ll
@@ -0,0 +1,36 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mcpu=pwr9 -mtriple=powerpc64-ibm-aix -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
+; RUN:     -stop-after=finalize-isel -verify-machineinstrs < %s | \
+; RUN:   FileCheck %s
+
+define <4 x i1> @foo(i1 %c1, i1 %c2, i1 %c3) {
+  ; CHECK-LABEL: name: foo
+  ; CHECK: bb.0 (%ir-block.0):
+  ; CHECK-NEXT:   liveins: $x3, $x4, $x5
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:g8rc = COPY $x5
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:g8rc = COPY $x4
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:g8rc = COPY $x3
+  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:gprc = COPY [[COPY1]].sub_32
+  ; CHECK-NEXT:   [[MTVSRWZ:%[0-9]+]]:vsfrc = MTVSRWZ killed [[COPY3]]
+  ; CHECK-NEXT:   [[SUBREG_TO_REG:%[0-9]+]]:vrrc = SUBREG_TO_REG 1, killed [[MTVSRWZ]], %subreg.sub_64
+  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:gprc = COPY [[COPY2]].sub_32
+  ; CHECK-NEXT:   [[MTVSRWZ1:%[0-9]+]]:vsfrc = MTVSRWZ killed [[COPY4]]
+  ; CHECK-NEXT:   [[SUBREG_TO_REG1:%[0-9]+]]:vrrc = SUBREG_TO_REG 1, killed [[MTVSRWZ1]], %subreg.sub_64
+  ; CHECK-NEXT:   [[VMRGOW:%[0-9]+]]:vrrc = VMRGOW killed [[SUBREG_TO_REG1]], killed [[SUBREG_TO_REG]]
+  ; CHECK-NEXT:   [[LDtocCPT:%[0-9]+]]:g8rc_and_g8rc_nox0 = LDtocCPT %const.0, $x2 :: (load (s64) from got)
+  ; CHECK-NEXT:   [[LXV:%[0-9]+]]:vsrc = LXV 0, killed [[LDtocCPT]] :: (load (s128) from constant-pool)
+  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:gprc = COPY [[COPY]].sub_32
+  ; CHECK-NEXT:   [[MTVSRWZ2:%[0-9]+]]:vsfrc = MTVSRWZ killed [[COPY5]]
+  ; CHECK-NEXT:   [[SUBREG_TO_REG2:%[0-9]+]]:vsrc = SUBREG_TO_REG 1, killed [[MTVSRWZ2]], %subreg.sub_64
+  ; CHECK-NEXT:   [[XXPERM:%[0-9]+]]:vsrc = XXPERM killed [[VMRGOW]], [[SUBREG_TO_REG2]], killed [[LXV]]
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:vsrc = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[XXINSERTW:%[0-9]+]]:vsrc = XXINSERTW [[DEF]], killed [[XXPERM]], 8
+  ; CHECK-NEXT:   $v2 = COPY [[XXINSERTW]]
+  ; CHECK-NEXT:   BLR8 implicit $lr8, implicit $rm, implicit $v2
+  %1 = insertelement <4 x i1> poison, i1 %c1, i64 0
+  %2 = insertelement <4 x i1> %1, i1 %c2, i64 1
+  %3 = insertelement <4 x i1> %2, i1 %c3, i64 3
+  %4 = shufflevector <4 x i1> %3, <4 x i1> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 3>
+  ret <4 x i1> %4
+}

From 053750c3b42c126eb4620f62cbf4e665803b941d Mon Sep 17 00:00:00 2001
From: Chen Zheng <czhengsz@cn.ibm.com>
Date: Thu, 11 Apr 2024 03:39:07 -0400
Subject: [PATCH 110/886] [PowerPC] Fix the undef register for VECINSERT

If the V2 of the vector_shuffle is undef, the two vector inputs are
expected to be the same when do the VECINSERT transformation. For now
the first operand of VECINSERT is set to undef which is not right.
This patch fixes this bug.
---
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp | 4 +++-
 llvm/test/CodeGen/PowerPC/xxinsertw.ll      | 3 +--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 43e4a34a9b348..52d5b71367059 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -10142,7 +10142,9 @@ SDValue PPCTargetLowering::LowerVECTOR_SHUFFLE(SDValue Op,
   if (Subtarget.hasP9Vector() &&
       PPC::isXXINSERTWMask(SVOp, ShiftElts, InsertAtByte, Swap,
                            isLittleEndian)) {
-    if (Swap)
+    if (V2.isUndef())
+      V2 = V1;
+    else if (Swap)
       std::swap(V1, V2);
     SDValue Conv1 = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1);
     SDValue Conv2 = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V2);
diff --git a/llvm/test/CodeGen/PowerPC/xxinsertw.ll b/llvm/test/CodeGen/PowerPC/xxinsertw.ll
index b48eac06a694a..f944b5a175be4 100644
--- a/llvm/test/CodeGen/PowerPC/xxinsertw.ll
+++ b/llvm/test/CodeGen/PowerPC/xxinsertw.ll
@@ -24,8 +24,7 @@ define <4 x i1> @foo(i1 %c1, i1 %c2, i1 %c3) {
   ; CHECK-NEXT:   [[MTVSRWZ2:%[0-9]+]]:vsfrc = MTVSRWZ killed [[COPY5]]
   ; CHECK-NEXT:   [[SUBREG_TO_REG2:%[0-9]+]]:vsrc = SUBREG_TO_REG 1, killed [[MTVSRWZ2]], %subreg.sub_64
   ; CHECK-NEXT:   [[XXPERM:%[0-9]+]]:vsrc = XXPERM killed [[VMRGOW]], [[SUBREG_TO_REG2]], killed [[LXV]]
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:vsrc = IMPLICIT_DEF
-  ; CHECK-NEXT:   [[XXINSERTW:%[0-9]+]]:vsrc = XXINSERTW [[DEF]], killed [[XXPERM]], 8
+  ; CHECK-NEXT:   [[XXINSERTW:%[0-9]+]]:vsrc = XXINSERTW [[XXPERM]], [[XXPERM]], 8
   ; CHECK-NEXT:   $v2 = COPY [[XXINSERTW]]
   ; CHECK-NEXT:   BLR8 implicit $lr8, implicit $rm, implicit $v2
   %1 = insertelement <4 x i1> poison, i1 %c1, i64 0

From a1cd5e69544ad3e6a865f5e0593ac26195ccb4f7 Mon Sep 17 00:00:00 2001
From: Michael Klemm <michael.klemm@amd.com>
Date: Thu, 11 Apr 2024 10:18:34 +0200
Subject: [PATCH 111/886] [flang] Do not create .f18.mod files for each
 compiled module (#85249)

The default CMake scripts had a copy operation to copy a compiled `.mod`
file to also be available with suffix `.f18.mod`. This seems no longer
needed. Also updated ModFiles.md to point to `-module-suffix`.

---------

Co-authored-by: Kiran Chandramohan <kiranchandramohan@gmail.com>
---
 flang/docs/ModFiles.md         | 4 +++-
 flang/tools/f18/CMakeLists.txt | 7 ++-----
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/flang/docs/ModFiles.md b/flang/docs/ModFiles.md
index e55d72fa3a705..7463454c8563a 100644
--- a/flang/docs/ModFiles.md
+++ b/flang/docs/ModFiles.md
@@ -27,7 +27,9 @@ often use `rm *.mod` to clean up.
 The disadvantage of using the same name as other compilers is that it is not
 clear which compiler created a `.mod` file and files from multiple compilers
 cannot be in the same directory. This could be solved by adding something
-between the module name and extension, e.g. `<modulename>-f18.mod`.
+between the module name and extension, e.g. `<modulename>-f18.mod`.  If this
+is needed, Flang's fc1 accepts the option `-module-suffix` to alter the suffix
+used for the module file.
 
 ## Format
 
diff --git a/flang/tools/f18/CMakeLists.txt b/flang/tools/f18/CMakeLists.txt
index e266055a4bf01..dda3b6887be89 100644
--- a/flang/tools/f18/CMakeLists.txt
+++ b/flang/tools/f18/CMakeLists.txt
@@ -62,11 +62,8 @@ if (NOT CMAKE_CROSSCOMPILING)
         ${FLANG_SOURCE_DIR}/module/${filename}.f90
       DEPENDS flang-new ${FLANG_SOURCE_DIR}/module/${filename}.f90 ${FLANG_SOURCE_DIR}/module/__fortran_builtins.f90 ${depends}
     )
-    add_custom_command(OUTPUT ${base}.f18.mod
-      DEPENDS ${base}.mod
-      COMMAND ${CMAKE_COMMAND} -E copy ${base}.mod ${base}.f18.mod)
-    list(APPEND MODULE_FILES ${base}.mod ${base}.f18.mod)
-    install(FILES ${base}.mod ${base}.f18.mod DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}/flang")
+    list(APPEND MODULE_FILES ${base}.mod)
+    install(FILES ${base}.mod DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}/flang")
   endforeach()
 
   # Special case for omp_lib.mod, because its source comes from openmp/runtime/src/include.

From 9c6e54b154cbbb7da0f45b4ae1e66bcf492151f1 Mon Sep 17 00:00:00 2001
From: Maciej Gabka <maciej.gabka@arm.com>
Date: Thu, 11 Apr 2024 09:32:17 +0100
Subject: [PATCH 112/886] [AArch64] Fix to Neoverse V2 scheduling model
 (#88130)

The size of ROB was incorrecty copied from the Neoverse N2, while it has
actually higher value as descibed in

https://community.arm.com/arm-community-blogs/b/infrastructure-solutions-blog/posts/arm-neoverse-v2-platform-best-in-class-cloud-and-ai-ml-performance
---
 llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td b/llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td
index 4d7f44e7b9b9a..7fed8fed90017 100644
--- a/llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td
+++ b/llvm/lib/Target/AArch64/AArch64SchedNeoverseV2.td
@@ -15,7 +15,7 @@
 
 def NeoverseV2Model : SchedMachineModel {
   let IssueWidth            =  16; // Micro-ops dispatched at a time.
-  let MicroOpBufferSize     = 160; // Entries in micro-op re-order buffer. NOTE: Copied from N2.
+  let MicroOpBufferSize     = 320; // Entries in micro-op re-order buffer.
   let LoadLatency           =   4; // Optimistic load latency.
   let MispredictPenalty     =  10; // Extra cycles for mispredicted branch.  NOTE: Copied from N2.
   let LoopMicroOpBufferSize =  16; // NOTE: Copied from Cortex-A57.

From cd14e7132f18dccd5fc7ed5e60258460bc1352f8 Mon Sep 17 00:00:00 2001
From: Ramkumar Ramachandra <ram.ramachandra@arm.com>
Date: Thu, 11 Apr 2024 09:36:56 +0100
Subject: [PATCH 113/886] lld/AArch64: handle more relocation addends (#87328)

The function getImplicitAddend() is incomplete, as it is possible to
cook up object files with other relocation addends. Although using
llvm-mc or the clang integrated assembler does not produce such object
files, a proprietary assembler known as armasm can:

https://developer.arm.com/documentation/101754/0622/armasm-Legacy-Assembler-Reference

armasm is in a frozen state, but it is still actively used in a lot of
legacy codebases as the directives, macros and operators are very
different from the clang integrated assembler. This makes porting a lot
of legacy code from armasm syntax impractical for a lot of projects.
Some internal testing of projects using open-source clang and lld fell
over at link time when legacy armasm objects were included in the link.

The goal of this patch is to enable people with legacy armasm objects to
be able to use lld as the linker. Sadly armasm uses SHT_REL format
relocations for AArch64 rather than SHT_RELA, which causes lld to reject
the objects. As a frozen project we know the small number of relocations
that the assembler officially supports and won't include (outside the
equivalent of the .reloc directive which I think we can rule out of
scope as that is not commonly used).

The benefit to lld is that it will ease migration from a proprietary to
an open-source toolchain. The drawback is the implementation of a small
number of SHT_REL relocations. Although this patch doesn't aim to
comprehensively cover all possible relocation addends, it does extend
lld to work with the relocation addends that armasm produces, using the
canonical aaelf64 document as a reference:

  https://github.com/ARM-software/abi-aa/blob/main/aaelf64/aaelf64.rst
---
 lld/ELF/Arch/AArch64.cpp                      | 42 +++++++++++++++----
 .../ELF/aarch64-reloc-implicit-addend.test    | 15 ++++++-
 2 files changed, 48 insertions(+), 9 deletions(-)

diff --git a/lld/ELF/Arch/AArch64.cpp b/lld/ELF/Arch/AArch64.cpp
index 017c17c2b03d8..2bf6e2c6c8519 100644
--- a/lld/ELF/Arch/AArch64.cpp
+++ b/lld/ELF/Arch/AArch64.cpp
@@ -69,6 +69,13 @@ struct AArch64Relaxer {
 };
 } // namespace
 
+// Return the bits [Start, End] from Val shifted Start bits.
+// For instance, getBits(0xF0, 4, 8) returns 0xF.
+static uint64_t getBits(uint64_t val, int start, int end) {
+  uint64_t mask = ((uint64_t)1 << (end + 1 - start)) - 1;
+  return (val >> start) & mask;
+}
+
 AArch64::AArch64() {
   copyRel = R_AARCH64_COPY;
   relativeRel = R_AARCH64_RELATIVE;
@@ -219,6 +226,10 @@ int64_t AArch64::getImplicitAddend(const uint8_t *buf, RelType type) const {
   case R_AARCH64_GLOB_DAT:
   case R_AARCH64_JUMP_SLOT:
     return 0;
+  case R_AARCH64_ABS16:
+  case R_AARCH64_PREL16:
+    return SignExtend64<16>(read16(buf));
+  case R_AARCH64_ABS32:
   case R_AARCH64_PREL32:
     return SignExtend64<32>(read32(buf));
   case R_AARCH64_ABS64:
@@ -227,6 +238,30 @@ int64_t AArch64::getImplicitAddend(const uint8_t *buf, RelType type) const {
   case R_AARCH64_IRELATIVE:
   case R_AARCH64_TLS_TPREL64:
     return read64(buf);
+  case R_AARCH64_MOVW_UABS_G0:
+  case R_AARCH64_MOVW_UABS_G0_NC:
+    return getBits(SignExtend64<16>(read16(buf)), 0, 15);
+  case R_AARCH64_MOVW_UABS_G1:
+  case R_AARCH64_MOVW_UABS_G1_NC:
+    return getBits(SignExtend64<32>(read32(buf)), 16, 31);
+  case R_AARCH64_MOVW_UABS_G2:
+  case R_AARCH64_MOVW_UABS_G2_NC:
+    return getBits(read64(buf), 32, 47);
+  case R_AARCH64_MOVW_UABS_G3:
+    return getBits(read64(buf), 48, 63);
+  case R_AARCH64_TSTBR14:
+    return getBits(SignExtend64<32>(read32(buf)), 2, 15);
+  case R_AARCH64_CONDBR19:
+  case R_AARCH64_LD_PREL_LO19:
+    return getBits(SignExtend64<32>(read32(buf)), 2, 20);
+  case R_AARCH64_ADD_ABS_LO12_NC:
+    return getBits(SignExtend64<16>(read16(buf)), 0, 11);
+  case R_AARCH64_ADR_PREL_PG_HI21:
+  case R_AARCH64_ADR_PREL_PG_HI21_NC:
+    return getBits(SignExtend64<32>(read32(buf)), 12, 32);
+  case R_AARCH64_JUMP26:
+  case R_AARCH64_CALL26:
+    return getBits(SignExtend64<32>(read32(buf)), 2, 27);
   default:
     internalLinkerError(getErrorLocation(buf),
                         "cannot read addend for relocation " + toString(type));
@@ -330,13 +365,6 @@ static void write32AArch64Addr(uint8_t *l, uint64_t imm) {
   write32le(l, (read32le(l) & ~mask) | immLo | immHi);
 }
 
-// Return the bits [Start, End] from Val shifted Start bits.
-// For instance, getBits(0xF0, 4, 8) returns 0xF.
-static uint64_t getBits(uint64_t val, int start, int end) {
-  uint64_t mask = ((uint64_t)1 << (end + 1 - start)) - 1;
-  return (val >> start) & mask;
-}
-
 static void or32le(uint8_t *p, int32_t v) { write32le(p, read32le(p) | v); }
 
 // Update the immediate field in a AARCH64 ldr, str, and add instruction.
diff --git a/lld/test/ELF/aarch64-reloc-implicit-addend.test b/lld/test/ELF/aarch64-reloc-implicit-addend.test
index 15f42c4d87b57..804ed97a27371 100644
--- a/lld/test/ELF/aarch64-reloc-implicit-addend.test
+++ b/lld/test/ELF/aarch64-reloc-implicit-addend.test
@@ -1,8 +1,19 @@
 ## Test certain REL relocation types generated by legacy armasm.
 # RUN: yaml2obj %s -o %t.o
-# RUN: not ld.lld %t.o -o /dev/null 2>&1 | FileCheck %s
+# RUN: ld.lld %t.o -o %t
+# RUN: llvm-objdump -s %t | FileCheck %s
 
-# CHECK-COUNT-17: internal linker error: cannot read addend
+# CHECK:      Contents of section .abs:
+# CHECK-NEXT:  [[#%x,]] 29002800 00002700 00000000 0000fcff  ).(...'.........
+# CHECK-NEXT:  [[#%x,]] ffffffff ffff                        ......
+# CHECK-NEXT: Contents of section .uabs:
+# CHECK-NEXT:  [[#%x,]] 40ffffff 40ffffff 20ffffff 20ffffff  @...@... ... ...
+# CHECK-NEXT:  [[#%x,]] 00ffffff 00ffffff                    ........
+# CHECK-NEXT: Contents of section .prel:
+# CHECK-NEXT:  [[#%x,]] 00ffffff fcfeffff f8feffff a0ffffff  ................
+# CHECK-NEXT:  [[#%x,]] 0010009f 0010009f                    ........
+# CHECK-NEXT: Contents of section .branch:
+# CHECK-NEXT:  [[#%x,]] f0ffffff f0ffffff fdffffff fcffff14  ................
 
 ---
 !ELF

From def6174d2a7a5a66b3871d4ce5035a71e513e6ef Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov@fb.com>
Date: Thu, 11 Apr 2024 10:37:57 +0200
Subject: [PATCH 114/886] [BOLT] Emit empty FDE for injected functions

This fixes an issue where `PatchEntries` overwrites function body but
keeps CFI untouched. Existing FDEs thus become invalid. This doesn't
affect unwinding because patched functions are transparent from
EH/unwinding perspective, but it breaks BOLT during disassembling those
functions.

Emit empty FDE for injected functions (emitted to the same address as
.org functions) that take precedence over the original FDE.

This adds eh_frame overhead, but restores the ability to disassemble
.org functions. Note that the overhead is avoided in `-use-old-text`
mode.

Test Plan: updated bolt/test/X86/patch-entries.test

Reviewers: rafaelauler, maksfb, dcci, ayermolo

Reviewed By: maksfb, dcci

Pull Request: https://github.com/llvm/llvm-project/pull/87967
---
 bolt/include/bolt/Core/BinaryFunction.h |  3 ++-
 bolt/test/X86/patch-entries.test        | 23 ++++++++++++++++++++++-
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/bolt/include/bolt/Core/BinaryFunction.h b/bolt/include/bolt/Core/BinaryFunction.h
index bc047fefa3151..26d2d01f86267 100644
--- a/bolt/include/bolt/Core/BinaryFunction.h
+++ b/bolt/include/bolt/Core/BinaryFunction.h
@@ -1402,7 +1402,8 @@ class BinaryFunction {
 
   /// Return true if the function has CFI instructions
   bool hasCFI() const {
-    return !FrameInstructions.empty() || !CIEFrameInstructions.empty();
+    return !FrameInstructions.empty() || !CIEFrameInstructions.empty() ||
+           IsInjected;
   }
 
   /// Return unique number associated with the function.
diff --git a/bolt/test/X86/patch-entries.test b/bolt/test/X86/patch-entries.test
index 54f358f273e79..4a725412dd616 100644
--- a/bolt/test/X86/patch-entries.test
+++ b/bolt/test/X86/patch-entries.test
@@ -7,4 +7,25 @@ REQUIRES: system-linux
 
 RUN: %clang %cflags -no-pie -g %p/Inputs/patch-entries.c -fuse-ld=lld -o %t.exe \
 RUN:   -Wl,-q -I%p/../Inputs
-RUN: llvm-bolt -relocs %t.exe -o %t.out --update-debug-sections --force-patch
+RUN: llvm-bolt -relocs %t.exe -o %t.out --update-debug-sections --force-patch \
+RUN:   --enable-bat
+
+# Check that patched functions can be disassembled (override FDE from the
+# original function)
+# PREAGG: B X:0 #foo.org.0# 1 0
+RUN: link_fdata %s %t.out %t.preagg PREAGG
+RUN: perf2bolt %t.out -p %t.preagg --pa -o %t.yaml --profile-format=yaml \
+RUN:   -print-disasm -print-only=foo.org.0/1 2>&1 | FileCheck %s
+CHECK-NOT: BOLT-WARNING: sizes differ for function foo.org.0/1
+CHECK: Binary Function "foo.org.0/1(*2)" after disassembly {
+
+# Check the expected eh_frame contents
+RUN: llvm-nm --print-size %t.out > %t.foo
+RUN: llvm-objdump %t.out --dwarf=frames >> %t.foo
+RUN: FileCheck %s --input-file %t.foo --check-prefix=CHECK-FOO
+CHECK-FOO: 0000000000[[#%x,FOO:]] [[#%x,OPTSIZE:]] t foo
+CHECK-FOO: 0000000000[[#%x,ORG:]] [[#%x,ORGSIZE:]] t foo.org.0
+# patched FDE comes first
+CHECK-FOO: FDE {{.*}} pc=00[[#%x,ORG]]...00[[#%x,ORG+ORGSIZE]]
+# original FDE comes second
+CHECK-FOO: FDE {{.*}} pc=00[[#%x,ORG]]...00[[#%x,ORG+OPTSIZE]]

From fe3b20d5ab4b47823fb48ad7cfbc47b8224ce826 Mon Sep 17 00:00:00 2001
From: NagyDonat <donat.nagy@ericsson.com>
Date: Thu, 11 Apr 2024 10:44:35 +0200
Subject: [PATCH 115/886] [analyzer] Use CDM::CLibrary instead of
 isGlobalCFunction() (#88267)

This commit updates several checkers to use call descriptions with the
matching mode `CDM::CLibrary` instead of checking
`Call.isGlobalCFunction()` after performing the match. This resolves
several TODOs in various checkers.

Note that both matching with `CDM::CLibrary` and calling
`isGlobalCFunction` leads to `CheckerContext::isCLibraryFunction()`
checks (so this change is close to being NFC), but if it is used via the
matching mode then the checker can automatically recognize the builtin
variants of the matched functions.

I'll also make similar changes in GenericTaintChecker, but that checker
has separate and inconsistent rules for handling the normal and the
builtin variant of several functions (e.g. `memcpy` and
`__builtin_memcpy`), so I'll put those changes into a separate commit.
---
 .../Checkers/BasicObjCFoundationChecks.cpp    |  12 +-
 .../Checkers/PthreadLockChecker.cpp           | 115 +++++++++++-------
 .../Checkers/SimpleStreamChecker.cpp          |  10 +-
 .../StaticAnalyzer/Checkers/StreamChecker.cpp |  69 +++++------
 .../StaticAnalyzer/Checkers/ValistChecker.cpp |  35 +++---
 5 files changed, 130 insertions(+), 111 deletions(-)

diff --git a/clang/lib/StaticAnalyzer/Checkers/BasicObjCFoundationChecks.cpp b/clang/lib/StaticAnalyzer/Checkers/BasicObjCFoundationChecks.cpp
index c72a97cc01e91..80f128b917b20 100644
--- a/clang/lib/StaticAnalyzer/Checkers/BasicObjCFoundationChecks.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/BasicObjCFoundationChecks.cpp
@@ -542,10 +542,10 @@ namespace {
 class CFRetainReleaseChecker : public Checker<check::PreCall> {
   mutable APIMisuse BT{this, "null passed to CF memory management function"};
   const CallDescriptionSet ModelledCalls = {
-      {{"CFRetain"}, 1},
-      {{"CFRelease"}, 1},
-      {{"CFMakeCollectable"}, 1},
-      {{"CFAutorelease"}, 1},
+      {CDM::CLibrary, {"CFRetain"}, 1},
+      {CDM::CLibrary, {"CFRelease"}, 1},
+      {CDM::CLibrary, {"CFMakeCollectable"}, 1},
+      {CDM::CLibrary, {"CFAutorelease"}, 1},
   };
 
 public:
@@ -555,10 +555,6 @@ class CFRetainReleaseChecker : public Checker<check::PreCall> {
 
 void CFRetainReleaseChecker::checkPreCall(const CallEvent &Call,
                                           CheckerContext &C) const {
-  // TODO: Make this check part of CallDescription.
-  if (!Call.isGlobalCFunction())
-    return;
-
   // Check if we called CFRetain/CFRelease/CFMakeCollectable/CFAutorelease.
   if (!ModelledCalls.contains(Call))
     return;
diff --git a/clang/lib/StaticAnalyzer/Checkers/PthreadLockChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/PthreadLockChecker.cpp
index fa8572cf85edf..86530086ff1b2 100644
--- a/clang/lib/StaticAnalyzer/Checkers/PthreadLockChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/PthreadLockChecker.cpp
@@ -87,7 +87,8 @@ class PthreadLockChecker : public Checker<check::PostCall, check::DeadSymbols,
                                               CheckerKind CheckKind) const;
   CallDescriptionMap<FnCheck> PThreadCallbacks = {
       // Init.
-      {{{"pthread_mutex_init"}, 2}, &PthreadLockChecker::InitAnyLock},
+      {{CDM::CLibrary, {"pthread_mutex_init"}, 2},
+       &PthreadLockChecker::InitAnyLock},
       // TODO: pthread_rwlock_init(2 arguments).
       // TODO: lck_mtx_init(3 arguments).
       // TODO: lck_mtx_alloc_init(2 arguments) => returns the mutex.
@@ -95,74 +96,106 @@ class PthreadLockChecker : public Checker<check::PostCall, check::DeadSymbols,
       // TODO: lck_rw_alloc_init(2 arguments) => returns the mutex.
 
       // Acquire.
-      {{{"pthread_mutex_lock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"pthread_rwlock_rdlock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"pthread_rwlock_wrlock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"lck_mtx_lock"}, 1}, &PthreadLockChecker::AcquireXNULock},
-      {{{"lck_rw_lock_exclusive"}, 1}, &PthreadLockChecker::AcquireXNULock},
-      {{{"lck_rw_lock_shared"}, 1}, &PthreadLockChecker::AcquireXNULock},
+      {{CDM::CLibrary, {"pthread_mutex_lock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"pthread_rwlock_rdlock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"pthread_rwlock_wrlock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"lck_mtx_lock"}, 1},
+       &PthreadLockChecker::AcquireXNULock},
+      {{CDM::CLibrary, {"lck_rw_lock_exclusive"}, 1},
+       &PthreadLockChecker::AcquireXNULock},
+      {{CDM::CLibrary, {"lck_rw_lock_shared"}, 1},
+       &PthreadLockChecker::AcquireXNULock},
 
       // Try.
-      {{{"pthread_mutex_trylock"}, 1}, &PthreadLockChecker::TryPthreadLock},
-      {{{"pthread_rwlock_tryrdlock"}, 1}, &PthreadLockChecker::TryPthreadLock},
-      {{{"pthread_rwlock_trywrlock"}, 1}, &PthreadLockChecker::TryPthreadLock},
-      {{{"lck_mtx_try_lock"}, 1}, &PthreadLockChecker::TryXNULock},
-      {{{"lck_rw_try_lock_exclusive"}, 1}, &PthreadLockChecker::TryXNULock},
-      {{{"lck_rw_try_lock_shared"}, 1}, &PthreadLockChecker::TryXNULock},
+      {{CDM::CLibrary, {"pthread_mutex_trylock"}, 1},
+       &PthreadLockChecker::TryPthreadLock},
+      {{CDM::CLibrary, {"pthread_rwlock_tryrdlock"}, 1},
+       &PthreadLockChecker::TryPthreadLock},
+      {{CDM::CLibrary, {"pthread_rwlock_trywrlock"}, 1},
+       &PthreadLockChecker::TryPthreadLock},
+      {{CDM::CLibrary, {"lck_mtx_try_lock"}, 1},
+       &PthreadLockChecker::TryXNULock},
+      {{CDM::CLibrary, {"lck_rw_try_lock_exclusive"}, 1},
+       &PthreadLockChecker::TryXNULock},
+      {{CDM::CLibrary, {"lck_rw_try_lock_shared"}, 1},
+       &PthreadLockChecker::TryXNULock},
 
       // Release.
-      {{{"pthread_mutex_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"pthread_rwlock_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"lck_mtx_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"lck_rw_unlock_exclusive"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"lck_rw_unlock_shared"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"lck_rw_done"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"pthread_mutex_unlock"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"pthread_rwlock_unlock"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"lck_mtx_unlock"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"lck_rw_unlock_exclusive"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"lck_rw_unlock_shared"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"lck_rw_done"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
 
       // Destroy.
-      {{{"pthread_mutex_destroy"}, 1}, &PthreadLockChecker::DestroyPthreadLock},
-      {{{"lck_mtx_destroy"}, 2}, &PthreadLockChecker::DestroyXNULock},
+      {{CDM::CLibrary, {"pthread_mutex_destroy"}, 1},
+       &PthreadLockChecker::DestroyPthreadLock},
+      {{CDM::CLibrary, {"lck_mtx_destroy"}, 2},
+       &PthreadLockChecker::DestroyXNULock},
       // TODO: pthread_rwlock_destroy(1 argument).
       // TODO: lck_rw_destroy(2 arguments).
   };
 
   CallDescriptionMap<FnCheck> FuchsiaCallbacks = {
       // Init.
-      {{{"spin_lock_init"}, 1}, &PthreadLockChecker::InitAnyLock},
+      {{CDM::CLibrary, {"spin_lock_init"}, 1},
+       &PthreadLockChecker::InitAnyLock},
 
       // Acquire.
-      {{{"spin_lock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"spin_lock_save"}, 3}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"sync_mutex_lock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
-      {{{"sync_mutex_lock_with_waiter"}, 1},
+      {{CDM::CLibrary, {"spin_lock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"spin_lock_save"}, 3},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"sync_mutex_lock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"sync_mutex_lock_with_waiter"}, 1},
        &PthreadLockChecker::AcquirePthreadLock},
 
       // Try.
-      {{{"spin_trylock"}, 1}, &PthreadLockChecker::TryFuchsiaLock},
-      {{{"sync_mutex_trylock"}, 1}, &PthreadLockChecker::TryFuchsiaLock},
-      {{{"sync_mutex_timedlock"}, 2}, &PthreadLockChecker::TryFuchsiaLock},
+      {{CDM::CLibrary, {"spin_trylock"}, 1},
+       &PthreadLockChecker::TryFuchsiaLock},
+      {{CDM::CLibrary, {"sync_mutex_trylock"}, 1},
+       &PthreadLockChecker::TryFuchsiaLock},
+      {{CDM::CLibrary, {"sync_mutex_timedlock"}, 2},
+       &PthreadLockChecker::TryFuchsiaLock},
 
       // Release.
-      {{{"spin_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"spin_unlock_restore"}, 3}, &PthreadLockChecker::ReleaseAnyLock},
-      {{{"sync_mutex_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"spin_unlock"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"spin_unlock_restore"}, 3},
+       &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"sync_mutex_unlock"}, 1},
+       &PthreadLockChecker::ReleaseAnyLock},
   };
 
   CallDescriptionMap<FnCheck> C11Callbacks = {
       // Init.
-      {{{"mtx_init"}, 2}, &PthreadLockChecker::InitAnyLock},
+      {{CDM::CLibrary, {"mtx_init"}, 2}, &PthreadLockChecker::InitAnyLock},
 
       // Acquire.
-      {{{"mtx_lock"}, 1}, &PthreadLockChecker::AcquirePthreadLock},
+      {{CDM::CLibrary, {"mtx_lock"}, 1},
+       &PthreadLockChecker::AcquirePthreadLock},
 
       // Try.
-      {{{"mtx_trylock"}, 1}, &PthreadLockChecker::TryC11Lock},
-      {{{"mtx_timedlock"}, 2}, &PthreadLockChecker::TryC11Lock},
+      {{CDM::CLibrary, {"mtx_trylock"}, 1}, &PthreadLockChecker::TryC11Lock},
+      {{CDM::CLibrary, {"mtx_timedlock"}, 2}, &PthreadLockChecker::TryC11Lock},
 
       // Release.
-      {{{"mtx_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
+      {{CDM::CLibrary, {"mtx_unlock"}, 1}, &PthreadLockChecker::ReleaseAnyLock},
 
       // Destroy
-      {{{"mtx_destroy"}, 1}, &PthreadLockChecker::DestroyPthreadLock},
+      {{CDM::CLibrary, {"mtx_destroy"}, 1},
+       &PthreadLockChecker::DestroyPthreadLock},
   };
 
   ProgramStateRef resolvePossiblyDestroyedMutex(ProgramStateRef state,
@@ -258,13 +291,9 @@ REGISTER_MAP_WITH_PROGRAMSTATE(DestroyRetVal, const MemRegion *, SymbolRef)
 
 void PthreadLockChecker::checkPostCall(const CallEvent &Call,
                                        CheckerContext &C) const {
-  // An additional umbrella check that all functions modeled by this checker
-  // are global C functions.
-  // TODO: Maybe make this the default behavior of CallDescription
-  // with exactly one identifier?
   // FIXME: Try to handle cases when the implementation was inlined rather
   // than just giving up.
-  if (!Call.isGlobalCFunction() || C.wasInlined)
+  if (C.wasInlined)
     return;
 
   if (const FnCheck *Callback = PThreadCallbacks.lookup(Call))
diff --git a/clang/lib/StaticAnalyzer/Checkers/SimpleStreamChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/SimpleStreamChecker.cpp
index 50d50562d3e75..5152624d00f46 100644
--- a/clang/lib/StaticAnalyzer/Checkers/SimpleStreamChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/SimpleStreamChecker.cpp
@@ -52,8 +52,8 @@ class SimpleStreamChecker : public Checker<check::PostCall,
                                            check::PreCall,
                                            check::DeadSymbols,
                                            check::PointerEscape> {
-  const CallDescription OpenFn{{"fopen"}, 2};
-  const CallDescription CloseFn{{"fclose"}, 1};
+  const CallDescription OpenFn{CDM::CLibrary, {"fopen"}, 2};
+  const CallDescription CloseFn{CDM::CLibrary, {"fclose"}, 1};
 
   const BugType DoubleCloseBugType{this, "Double fclose",
                                    "Unix Stream API Error"};
@@ -92,9 +92,6 @@ REGISTER_MAP_WITH_PROGRAMSTATE(StreamMap, SymbolRef, StreamState)
 
 void SimpleStreamChecker::checkPostCall(const CallEvent &Call,
                                         CheckerContext &C) const {
-  if (!Call.isGlobalCFunction())
-    return;
-
   if (!OpenFn.matches(Call))
     return;
 
@@ -111,9 +108,6 @@ void SimpleStreamChecker::checkPostCall(const CallEvent &Call,
 
 void SimpleStreamChecker::checkPreCall(const CallEvent &Call,
                                        CheckerContext &C) const {
-  if (!Call.isGlobalCFunction())
-    return;
-
   if (!CloseFn.matches(Call))
     return;
 
diff --git a/clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp
index 31c756ab0c581..bd495cd0f9710 100644
--- a/clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/StreamChecker.cpp
@@ -302,85 +302,88 @@ class StreamChecker : public Checker<check::PreCall, eval::Call,
 
 private:
   CallDescriptionMap<FnDescription> FnDescriptions = {
-      {{{"fopen"}, 2}, {nullptr, &StreamChecker::evalFopen, ArgNone}},
-      {{{"fdopen"}, 2}, {nullptr, &StreamChecker::evalFopen, ArgNone}},
-      {{{"freopen"}, 3},
+      {{CDM::CLibrary, {"fopen"}, 2},
+       {nullptr, &StreamChecker::evalFopen, ArgNone}},
+      {{CDM::CLibrary, {"fdopen"}, 2},
+       {nullptr, &StreamChecker::evalFopen, ArgNone}},
+      {{CDM::CLibrary, {"freopen"}, 3},
        {&StreamChecker::preFreopen, &StreamChecker::evalFreopen, 2}},
-      {{{"tmpfile"}, 0}, {nullptr, &StreamChecker::evalFopen, ArgNone}},
-      {{{"fclose"}, 1},
+      {{CDM::CLibrary, {"tmpfile"}, 0},
+       {nullptr, &StreamChecker::evalFopen, ArgNone}},
+      {{CDM::CLibrary, {"fclose"}, 1},
        {&StreamChecker::preDefault, &StreamChecker::evalFclose, 0}},
-      {{{"fread"}, 4},
+      {{CDM::CLibrary, {"fread"}, 4},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFreadFwrite, _1, _2, _3, _4, true), 3}},
-      {{{"fwrite"}, 4},
+      {{CDM::CLibrary, {"fwrite"}, 4},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFreadFwrite, _1, _2, _3, _4, false), 3}},
-      {{{"fgetc"}, 1},
+      {{CDM::CLibrary, {"fgetc"}, 1},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFgetx, _1, _2, _3, _4, true), 0}},
-      {{{"fgets"}, 3},
+      {{CDM::CLibrary, {"fgets"}, 3},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFgetx, _1, _2, _3, _4, false), 2}},
-      {{{"getc"}, 1},
+      {{CDM::CLibrary, {"getc"}, 1},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFgetx, _1, _2, _3, _4, true), 0}},
-      {{{"fputc"}, 2},
+      {{CDM::CLibrary, {"fputc"}, 2},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFputx, _1, _2, _3, _4, true), 1}},
-      {{{"fputs"}, 2},
+      {{CDM::CLibrary, {"fputs"}, 2},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFputx, _1, _2, _3, _4, false), 1}},
-      {{{"putc"}, 2},
+      {{CDM::CLibrary, {"putc"}, 2},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFputx, _1, _2, _3, _4, true), 1}},
-      {{{"fprintf"}},
+      {{CDM::CLibrary, {"fprintf"}},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFprintf, _1, _2, _3, _4), 0}},
-      {{{"vfprintf"}, 3},
+      {{CDM::CLibrary, {"vfprintf"}, 3},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalFprintf, _1, _2, _3, _4), 0}},
-      {{{"fscanf"}},
+      {{CDM::CLibrary, {"fscanf"}},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFscanf, _1, _2, _3, _4), 0}},
-      {{{"vfscanf"}, 3},
+      {{CDM::CLibrary, {"vfscanf"}, 3},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalFscanf, _1, _2, _3, _4), 0}},
-      {{{"ungetc"}, 2},
+      {{CDM::CLibrary, {"ungetc"}, 2},
        {&StreamChecker::preWrite,
         std::bind(&StreamChecker::evalUngetc, _1, _2, _3, _4), 1}},
-      {{{"getdelim"}, 4},
+      {{CDM::CLibrary, {"getdelim"}, 4},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalGetdelim, _1, _2, _3, _4), 3}},
-      {{{"getline"}, 3},
+      {{CDM::CLibrary, {"getline"}, 3},
        {&StreamChecker::preRead,
         std::bind(&StreamChecker::evalGetdelim, _1, _2, _3, _4), 2}},
-      {{{"fseek"}, 3},
+      {{CDM::CLibrary, {"fseek"}, 3},
        {&StreamChecker::preFseek, &StreamChecker::evalFseek, 0}},
-      {{{"fseeko"}, 3},
+      {{CDM::CLibrary, {"fseeko"}, 3},
        {&StreamChecker::preFseek, &StreamChecker::evalFseek, 0}},
-      {{{"ftell"}, 1},
+      {{CDM::CLibrary, {"ftell"}, 1},
        {&StreamChecker::preWrite, &StreamChecker::evalFtell, 0}},
-      {{{"ftello"}, 1},
+      {{CDM::CLibrary, {"ftello"}, 1},
        {&StreamChecker::preWrite, &StreamChecker::evalFtell, 0}},
-      {{{"fflush"}, 1},
+      {{CDM::CLibrary, {"fflush"}, 1},
        {&StreamChecker::preFflush, &StreamChecker::evalFflush, 0}},
-      {{{"rewind"}, 1},
+      {{CDM::CLibrary, {"rewind"}, 1},
        {&StreamChecker::preDefault, &StreamChecker::evalRewind, 0}},
-      {{{"fgetpos"}, 2},
+      {{CDM::CLibrary, {"fgetpos"}, 2},
        {&StreamChecker::preWrite, &StreamChecker::evalFgetpos, 0}},
-      {{{"fsetpos"}, 2},
+      {{CDM::CLibrary, {"fsetpos"}, 2},
        {&StreamChecker::preDefault, &StreamChecker::evalFsetpos, 0}},
-      {{{"clearerr"}, 1},
+      {{CDM::CLibrary, {"clearerr"}, 1},
        {&StreamChecker::preDefault, &StreamChecker::evalClearerr, 0}},
-      {{{"feof"}, 1},
+      {{CDM::CLibrary, {"feof"}, 1},
        {&StreamChecker::preDefault,
         std::bind(&StreamChecker::evalFeofFerror, _1, _2, _3, _4, ErrorFEof),
         0}},
-      {{{"ferror"}, 1},
+      {{CDM::CLibrary, {"ferror"}, 1},
        {&StreamChecker::preDefault,
         std::bind(&StreamChecker::evalFeofFerror, _1, _2, _3, _4, ErrorFError),
         0}},
-      {{{"fileno"}, 1},
+      {{CDM::CLibrary, {"fileno"}, 1},
        {&StreamChecker::preDefault, &StreamChecker::evalFileno, 0}},
   };
 
@@ -540,8 +543,6 @@ class StreamChecker : public Checker<check::PreCall, eval::Call,
   const FnDescription *lookupFn(const CallEvent &Call) const {
     // Recognize "global C functions" with only integral or pointer arguments
     // (and matching name) as stream functions.
-    if (!Call.isGlobalCFunction())
-      return nullptr;
     for (auto *P : Call.parameters()) {
       QualType T = P->getType();
       if (!T->isIntegralOrEnumerationType() && !T->isPointerType() &&
diff --git a/clang/lib/StaticAnalyzer/Checkers/ValistChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/ValistChecker.cpp
index 2d1b873abf73f..28320f46f237a 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ValistChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ValistChecker.cpp
@@ -100,32 +100,31 @@ class ValistChecker : public Checker<check::PreCall, check::PreStmt<VAArgExpr>,
 };
 
 const SmallVector<ValistChecker::VAListAccepter, 15>
-    ValistChecker::VAListAccepters = {{{{"vfprintf"}, 3}, 2},
-                                      {{{"vfscanf"}, 3}, 2},
-                                      {{{"vprintf"}, 2}, 1},
-                                      {{{"vscanf"}, 2}, 1},
-                                      {{{"vsnprintf"}, 4}, 3},
-                                      {{{"vsprintf"}, 3}, 2},
-                                      {{{"vsscanf"}, 3}, 2},
-                                      {{{"vfwprintf"}, 3}, 2},
-                                      {{{"vfwscanf"}, 3}, 2},
-                                      {{{"vwprintf"}, 2}, 1},
-                                      {{{"vwscanf"}, 2}, 1},
-                                      {{{"vswprintf"}, 4}, 3},
+    ValistChecker::VAListAccepters = {{{CDM::CLibrary, {"vfprintf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vfscanf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vprintf"}, 2}, 1},
+                                      {{CDM::CLibrary, {"vscanf"}, 2}, 1},
+                                      {{CDM::CLibrary, {"vsnprintf"}, 4}, 3},
+                                      {{CDM::CLibrary, {"vsprintf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vsscanf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vfwprintf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vfwscanf"}, 3}, 2},
+                                      {{CDM::CLibrary, {"vwprintf"}, 2}, 1},
+                                      {{CDM::CLibrary, {"vwscanf"}, 2}, 1},
+                                      {{CDM::CLibrary, {"vswprintf"}, 4}, 3},
                                       // vswprintf is the wide version of
                                       // vsnprintf, vsprintf has no wide version
-                                      {{{"vswscanf"}, 3}, 2}};
+                                      {{CDM::CLibrary, {"vswscanf"}, 3}, 2}};
 
-const CallDescription ValistChecker::VaStart({"__builtin_va_start"}, /*Args=*/2,
+const CallDescription ValistChecker::VaStart(CDM::CLibrary,
+                                             {"__builtin_va_start"}, /*Args=*/2,
                                              /*Params=*/1),
-    ValistChecker::VaCopy({"__builtin_va_copy"}, 2),
-    ValistChecker::VaEnd({"__builtin_va_end"}, 1);
+    ValistChecker::VaCopy(CDM::CLibrary, {"__builtin_va_copy"}, 2),
+    ValistChecker::VaEnd(CDM::CLibrary, {"__builtin_va_end"}, 1);
 } // end anonymous namespace
 
 void ValistChecker::checkPreCall(const CallEvent &Call,
                                  CheckerContext &C) const {
-  if (!Call.isGlobalCFunction())
-    return;
   if (VaStart.matches(Call))
     checkVAListStartCall(Call, C, false);
   else if (VaCopy.matches(Call))

From 5c9315f575370393ccc89ef0229743c05f6fe703 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Thu, 11 Apr 2024 10:54:50 +0200
Subject: [PATCH 116/886] Fix complex log1p accuracy with large abs values.
 (#88364)

This ports openxla/xla#10503 by @pearu. In addition to the filecheck
test here, the accuracy was tested with XLA's complex_unary_op_test and
its MLIR emitters.

This is a fixed version of
https://github.com/llvm/llvm-project/pull/88260. The previous version
relied on implementation-specific behavior in the order of evaluation of
maxAbsOfRealPlusOneAndImagMinusOne's operands.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 51 ++++++++++---------
 .../convert-to-standard.mlir                  | 48 ++++++++++-------
 2 files changed, 58 insertions(+), 41 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 9c3c4d96a301e..a6fcf6a758c07 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -570,37 +570,40 @@ struct Log1pOpConversion : public OpConversionPattern<complex::Log1pOp> {
                   ConversionPatternRewriter &rewriter) const override {
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
-    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
     mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
-    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
-    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
+    Value real = b.create<complex::ReOp>(adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(adaptor.getComplex());
 
     Value half = b.create<arith::ConstantOp>(elementType,
                                              b.getFloatAttr(elementType, 0.5));
     Value one = b.create<arith::ConstantOp>(elementType,
                                             b.getFloatAttr(elementType, 1));
-    Value two = b.create<arith::ConstantOp>(elementType,
-                                            b.getFloatAttr(elementType, 2));
-
-    // log1p(a+bi) = .5*log((a+1)^2+b^2) + i*atan2(b, a + 1)
-    // log((a+1)+bi) = .5*log(a*a + 2*a + 1 + b*b) + i*atan2(b, a+1)
-    // log((a+1)+bi) = .5*log1p(a*a + 2*a + b*b) + i*atan2(b, a+1)
-    Value sumSq = b.create<arith::MulFOp>(real, real, fmf.getValue());
-    sumSq = b.create<arith::AddFOp>(
-        sumSq, b.create<arith::MulFOp>(real, two, fmf.getValue()),
-        fmf.getValue());
-    sumSq = b.create<arith::AddFOp>(
-        sumSq, b.create<arith::MulFOp>(imag, imag, fmf.getValue()),
-        fmf.getValue());
-    Value logSumSq =
-        b.create<math::Log1pOp>(elementType, sumSq, fmf.getValue());
-    Value resultReal = b.create<arith::MulFOp>(logSumSq, half, fmf.getValue());
-
-    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf.getValue());
-
-    Value resultImag =
-        b.create<math::Atan2Op>(elementType, imag, realPlusOne, fmf.getValue());
+    Value realPlusOne = b.create<arith::AddFOp>(real, one, fmf);
+    Value absRealPlusOne = b.create<math::AbsFOp>(realPlusOne, fmf);
+    Value absImag = b.create<math::AbsFOp>(imag, fmf);
+
+    Value maxAbs = b.create<arith::MaximumFOp>(absRealPlusOne, absImag, fmf);
+    Value minAbs = b.create<arith::MinimumFOp>(absRealPlusOne, absImag, fmf);
+
+    Value useReal = b.create<arith::CmpFOp>(arith::CmpFPredicate::OGT,
+                                            realPlusOne, absImag, fmf);
+    Value maxMinusOne = b.create<arith::SubFOp>(maxAbs, one, fmf);
+    Value maxAbsOfRealPlusOneAndImagMinusOne =
+        b.create<arith::SelectOp>(useReal, real, maxMinusOne);
+    Value minMaxRatio = b.create<arith::DivFOp>(minAbs, maxAbs, fmf);
+    Value logOfMaxAbsOfRealPlusOneAndImag =
+        b.create<math::Log1pOp>(maxAbsOfRealPlusOneAndImagMinusOne, fmf);
+    Value logOfSqrtPart = b.create<math::Log1pOp>(
+        b.create<arith::MulFOp>(minMaxRatio, minMaxRatio, fmf), fmf);
+    Value r = b.create<arith::AddFOp>(
+        b.create<arith::MulFOp>(half, logOfSqrtPart, fmf),
+        logOfMaxAbsOfRealPlusOneAndImag, fmf);
+    Value resultReal = b.create<arith::SelectOp>(
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, r, r, fmf), minAbs,
+        r);
+    Value resultImag = b.create<math::Atan2Op>(imag, realPlusOne, fmf);
     rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
                                                    resultImag);
     return success();
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index f5d9499eadda4..46dba04a88aa0 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -300,15 +300,22 @@ func.func @complex_log1p(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] : f32
-// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] : f32
-// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] : f32
-// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] : f32
-// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] : f32
-// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] : f32
 // CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] : f32
+// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %[[ONE]] : f32
+// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
+// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] : f32
+// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] : f32
+// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] : f32
+// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] : f32
+// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
@@ -963,15 +970,22 @@ func.func @complex_log1p_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE_HALF:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[TWO:.*]] = arith.constant 2.000000e+00 : f32
-// CHECK: %[[SQ_SUM_0:.*]] = arith.mulf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[TWO_REAL:.*]] = arith.mulf %[[REAL]], %[[TWO]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_SUM_1:.*]] = arith.addf %[[SQ_SUM_0]], %[[TWO_REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_IMAG:.*]] = arith.mulf %[[IMAG]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQ_SUM_2:.*]] = arith.addf %[[SQ_SUM_1]], %[[SQ_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[LOG_SQ_SUM:.*]] = math.log1p %[[SQ_SUM_2]] fastmath<nnan,contract> : f32
-// CHECK: %[[RESULT_REAL:.*]] = arith.mulf %[[LOG_SQ_SUM]], %[[ONE_HALF]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_PLUS_ONE:.*]] = arith.addf %[[REAL]], %[[ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_REAL_PLUS_ONE:.*]] = math.absf %[[REAL_PLUS_ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[CMPF:.*]] = arith.cmpf ogt, %[[REAL_PLUS_ONE]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MAX_MINUS_ONE:.*]] = arith.subf %[[MAX]], %[[ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[SELECT:.*]] = arith.select %[[CMPF]], %0, %[[MAX_MINUS_ONE]] : f32
+// CHECK: %[[MIN_MAX_RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract>  : f32
+// CHECK: %[[LOG_1:.*]] = math.log1p %[[SELECT]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[MIN_MAX_RATIO]], %[[MIN_MAX_RATIO]] fastmath<nnan,contract>  : f32
+// CHECK: %[[LOG_SQ:.*]] = math.log1p %[[RATIO_SQ]] fastmath<nnan,contract>  : f32
+// CHECK: %[[HALF_LOG_SQ:.*]] = arith.mulf %cst, %[[LOG_SQ]] fastmath<nnan,contract>  : f32
+// CHECK: %[[R:.*]] = arith.addf %[[HALF_LOG_SQ]], %[[LOG_1]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ISNAN:.*]] = arith.cmpf uno, %[[R]], %[[R]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.select %[[ISNAN]], %[[MIN]], %[[R]] : f32
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG]], %[[REAL_PLUS_ONE]] fastmath<nnan,contract> : f32
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL]], %[[RESULT_IMAG]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>

From db2fb3d96b217f0d2e139e7816c98d9f95974f25 Mon Sep 17 00:00:00 2001
From: Freddy Ye <freddy.ye@intel.com>
Date: Thu, 11 Apr 2024 16:57:32 +0800
Subject: [PATCH 117/886] [X86] Define __APX_F__ when APX is enabled. (#88343)

Relate gcc patch:
https://gcc.gnu.org/pipermail/gcc-patches/2024-April/648789.html
---
 clang/lib/Basic/Targets/X86.cpp               | 3 +++
 clang/test/Preprocessor/x86_target_features.c | 3 ++-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/clang/lib/Basic/Targets/X86.cpp b/clang/lib/Basic/Targets/X86.cpp
index 1966af17904d6..bf1767c87fe1c 100644
--- a/clang/lib/Basic/Targets/X86.cpp
+++ b/clang/lib/Basic/Targets/X86.cpp
@@ -954,6 +954,9 @@ void X86TargetInfo::getTargetDefines(const LangOptions &Opts,
     Builder.defineMacro("__CCMP__");
   if (HasCF)
     Builder.defineMacro("__CF__");
+  // Condition here is aligned with the feature set of mapxf in Options.td
+  if (HasEGPR && HasPush2Pop2 && HasPPX && HasNDD)
+    Builder.defineMacro("__APX_F__");
 
   // Each case falls through to the previous one here.
   switch (SSELevel) {
diff --git a/clang/test/Preprocessor/x86_target_features.c b/clang/test/Preprocessor/x86_target_features.c
index a1882043910f7..5602c59158fe5 100644
--- a/clang/test/Preprocessor/x86_target_features.c
+++ b/clang/test/Preprocessor/x86_target_features.c
@@ -803,7 +803,8 @@
 // RUN: %clang -target x86_64-unknown-unknown -march=x86-64 -mapx-features=ndd -x c -E -dM -o - %s | FileCheck --check-prefix=NDD %s
 // RUN: %clang -target x86_64-unknown-unknown -march=x86-64 -mapx-features=ccmp -x c -E -dM -o - %s | FileCheck --check-prefix=CCMP %s
 // RUN: %clang -target x86_64-unknown-unknown -march=x86-64 -mapx-features=cf -x c -E -dM -o - %s | FileCheck --check-prefix=CF %s
-// RUN: %clang -target x86_64-unknown-unknown -march=x86-64 -mapxf -x c -E -dM -o - %s | FileCheck --check-prefixes=EGPR,PUSH2POP2,PPX,NDD %s
+// RUN: %clang -target x86_64-unknown-unknown -march=x86-64 -mapxf -x c -E -dM -o - %s | FileCheck --check-prefixes=EGPR,PUSH2POP2,PPX,NDD,APXF %s
+// APXF: #define __APX_F__ 1
 // CCMP: #define __CCMP__ 1
 // CF: #define __CF__ 1
 // EGPR: #define __EGPR__ 1

From 32b95a37083d1fee1a638e292be0aac9a98792fd Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 10:05:53 +0100
Subject: [PATCH 118/886] [VectorCombine][X86] Extend
 shuffle(bitcast(x),bitcast(y)) test coverage

As discussed on #87510 the intention is to fold shuffle(bitcast(x),bitcast(y)) -> bitcast(shuffle(x,y)), but it must not interfere with existing bitcast(shuffle(bitcast(x),bitcast(y))) folds.
---
 .../VectorCombine/X86/shuffle-of-casts.ll     | 67 ++++++++++++++++---
 1 file changed, 56 insertions(+), 11 deletions(-)

diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
index 2031c2d04c601..60c6ff97c58b5 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
@@ -165,6 +165,51 @@ define <8 x double> @interleave_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %
   ret <8 x double> %r
 }
 
+; TODO - bitcasts (same element count)
+
+define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: @concat_bitcast_v4i32_v8f32(
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <4 x float>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <4 x float>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[X0]], <4 x float> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x float> [[R]]
+;
+  %x0 = bitcast <4 x i32> %a0 to <4 x float>
+  %x1 = bitcast <4 x i32> %a1 to <4 x float>
+  %r = shufflevector <4 x float> %x0, <4 x float> %x1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x float> %r
+}
+
+; TODO - bitcasts (lower element count)
+
+define <4 x double> @concat_bitcast_v8i16_v4f64(<8 x i16> %a0, <8 x i16> %a1) {
+; CHECK-LABEL: @concat_bitcast_v8i16_v4f64(
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <8 x i16> [[A0:%.*]] to <2 x double>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <8 x i16> [[A1:%.*]] to <2 x double>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x double> [[X0]], <2 x double> [[X1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    ret <4 x double> [[R]]
+;
+  %x0 = bitcast <8 x i16> %a0 to <2 x double>
+  %x1 = bitcast <8 x i16> %a1 to <2 x double>
+  %r = shufflevector <2 x double> %x0, <2 x double> %x1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  ret <4 x double> %r
+}
+
+; TODO - bitcasts (higher element count)
+
+define <16 x i16> @concat_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: @concat_bitcast_v4i32_v16i16(
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <8 x i16>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <8 x i16>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x i16> [[X0]], <8 x i16> [[X1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    ret <16 x i16> [[R]]
+;
+  %x0 = bitcast <4 x i32> %a0 to <8 x i16>
+  %x1 = bitcast <4 x i32> %a1 to <8 x i16>
+  %r = shufflevector <8 x i16> %x0, <8 x i16> %x1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <16 x i16> %r
+}
+
 ; negative - multiuse
 
 define <8 x i16> @concat_trunc_v4i32_v8i16_multiuse(<4 x i32> %a0, <4 x i32> %a1, ptr %a2) {
@@ -182,19 +227,19 @@ define <8 x i16> @concat_trunc_v4i32_v8i16_multiuse(<4 x i32> %a0, <4 x i32> %a1
   ret <8 x i16> %r
 }
 
-; negative - bitcasts
+; negative - bitcasts (unscalable higher element count)
 
-define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @concat_bitcast_v4i32_v8f32(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <4 x float>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <4 x float>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[X0]], <4 x float> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:    ret <8 x float> [[R]]
+define <16 x i16> @revpair_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: @revpair_bitcast_v4i32_v16i16(
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <8 x i16>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <8 x i16>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x i16> [[X0]], <8 x i16> [[X1]], <16 x i32> <i32 1, i32 0, i32 3, i32 3, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+; CHECK-NEXT:    ret <16 x i16> [[R]]
 ;
-  %x0 = bitcast <4 x i32> %a0 to <4 x float>
-  %x1 = bitcast <4 x i32> %a1 to <4 x float>
-  %r = shufflevector <4 x float> %x0, <4 x float> %x1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-  ret <8 x float> %r
+  %x0 = bitcast <4 x i32> %a0 to <8 x i16>
+  %x1 = bitcast <4 x i32> %a1 to <8 x i16>
+  %r = shufflevector <8 x i16> %x0, <8 x i16> %x1, <16 x i32> <i32 1, i32 0, i32 3, i32 3, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  ret <16 x i16> %r
 }
 
 ; negative - src type mismatch

From 478c42004c2bd4c91a01c47450eca6cdb6b0982d Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Thu, 11 Apr 2024 10:25:56 +0100
Subject: [PATCH 119/886] [VPlan] Update recipe ::clone definitions to use
 cloned tys (NFC).

Update definitions on ::clone in recipe sub-types to use the sub-type as
return type. This avoids typecasts down to the cloned type in some
cases.
---
 llvm/lib/Transforms/Vectorize/VPlan.h | 56 ++++++++++++++-------------
 1 file changed, 30 insertions(+), 26 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 5dc905a3c407b..d86a81d4fb4c7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -888,6 +888,8 @@ class VPSingleDefRecipe : public VPRecipeBase, public VPValue {
     return R && classof(R);
   }
 
+  virtual VPSingleDefRecipe *clone() override = 0;
+
   /// Returns the underlying instruction.
   Instruction *getUnderlyingInstr() {
     return cast<Instruction>(getUnderlyingValue());
@@ -1248,7 +1250,7 @@ class VPInstruction : public VPRecipeWithIRFlags {
 
   VP_CLASSOF_IMPL(VPDef::VPInstructionSC)
 
-  VPRecipeBase *clone() override {
+  VPInstruction *clone() override {
     SmallVector<VPValue *, 2> Operands(operands());
     auto *New = new VPInstruction(Opcode, Operands, getDebugLoc(), Name);
     New->transferFlags(*this);
@@ -1335,7 +1337,7 @@ class VPWidenRecipe : public VPRecipeWithIRFlags {
 
   ~VPWidenRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenRecipe *clone() override {
     auto *R = new VPWidenRecipe(*getUnderlyingInstr(), operands());
     R->transferFlags(*this);
     return R;
@@ -1380,7 +1382,7 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
 
   ~VPWidenCastRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenCastRecipe *clone() override {
     if (auto *UV = getUnderlyingValue())
       return new VPWidenCastRecipe(Opcode, getOperand(0), ResultTy,
                                    *cast<CastInst>(UV));
@@ -1420,7 +1422,7 @@ class VPScalarCastRecipe : public VPSingleDefRecipe {
 
   ~VPScalarCastRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPScalarCastRecipe *clone() override {
     return new VPScalarCastRecipe(Opcode, getOperand(0), ResultTy);
   }
 
@@ -1465,7 +1467,7 @@ class VPWidenCallRecipe : public VPSingleDefRecipe {
 
   ~VPWidenCallRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenCallRecipe *clone() override {
     return new VPWidenCallRecipe(*cast<CallInst>(getUnderlyingInstr()),
                                  operands(), VectorIntrinsicID, getDebugLoc(),
                                  Variant);
@@ -1492,7 +1494,7 @@ struct VPWidenSelectRecipe : public VPSingleDefRecipe {
 
   ~VPWidenSelectRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenSelectRecipe *clone() override {
     return new VPWidenSelectRecipe(*cast<SelectInst>(getUnderlyingInstr()),
                                    operands());
   }
@@ -1540,7 +1542,7 @@ class VPWidenGEPRecipe : public VPRecipeWithIRFlags {
 
   ~VPWidenGEPRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenGEPRecipe *clone() override {
     return new VPWidenGEPRecipe(cast<GetElementPtrInst>(getUnderlyingInstr()),
                                 operands());
   }
@@ -1581,7 +1583,7 @@ class VPVectorPointerRecipe : public VPRecipeWithIRFlags {
     return true;
   }
 
-  VPRecipeBase *clone() override {
+  VPVectorPointerRecipe *clone() override {
     return new VPVectorPointerRecipe(getOperand(0), IndexedTy, IsReverse,
                                      isInBounds(), getDebugLoc());
   }
@@ -1696,7 +1698,7 @@ class VPWidenIntOrFpInductionRecipe : public VPHeaderPHIRecipe {
 
   ~VPWidenIntOrFpInductionRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenIntOrFpInductionRecipe *clone() override {
     return new VPWidenIntOrFpInductionRecipe(IV, getStartValue(),
                                              getStepValue(), IndDesc, Trunc);
   }
@@ -1771,7 +1773,7 @@ class VPWidenPointerInductionRecipe : public VPHeaderPHIRecipe {
 
   ~VPWidenPointerInductionRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenPointerInductionRecipe *clone() override {
     return new VPWidenPointerInductionRecipe(
         cast<PHINode>(getUnderlyingInstr()), getOperand(0), getOperand(1),
         IndDesc, IsScalarAfterVectorization);
@@ -1810,7 +1812,7 @@ class VPWidenPHIRecipe : public VPSingleDefRecipe {
       addOperand(Start);
   }
 
-  VPRecipeBase *clone() override {
+  VPWidenPHIRecipe *clone() override {
     llvm_unreachable("cloning not implemented yet");
   }
 
@@ -1853,7 +1855,7 @@ struct VPFirstOrderRecurrencePHIRecipe : public VPHeaderPHIRecipe {
     return R->getVPDefID() == VPDef::VPFirstOrderRecurrencePHISC;
   }
 
-  VPRecipeBase *clone() override {
+  VPFirstOrderRecurrencePHIRecipe *clone() override {
     return new VPFirstOrderRecurrencePHIRecipe(
         cast<PHINode>(getUnderlyingInstr()), *getOperand(0));
   }
@@ -1893,7 +1895,7 @@ class VPReductionPHIRecipe : public VPHeaderPHIRecipe {
 
   ~VPReductionPHIRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPReductionPHIRecipe *clone() override {
     auto *R =
         new VPReductionPHIRecipe(cast<PHINode>(getUnderlyingInstr()), RdxDesc,
                                  *getOperand(0), IsInLoop, IsOrdered);
@@ -1940,7 +1942,7 @@ class VPBlendRecipe : public VPSingleDefRecipe {
            "Expected an odd number of operands");
   }
 
-  VPRecipeBase *clone() override {
+  VPBlendRecipe *clone() override {
     SmallVector<VPValue *> Ops(operands());
     return new VPBlendRecipe(cast<PHINode>(getUnderlyingValue()), Ops);
   }
@@ -2019,7 +2021,7 @@ class VPInterleaveRecipe : public VPRecipeBase {
   }
   ~VPInterleaveRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPInterleaveRecipe *clone() override {
     return new VPInterleaveRecipe(IG, getAddr(), getStoredValues(), getMask(),
                                   NeedsMaskForGaps);
   }
@@ -2093,7 +2095,7 @@ class VPReductionRecipe : public VPSingleDefRecipe {
 
   ~VPReductionRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPReductionRecipe *clone() override {
     return new VPReductionRecipe(RdxDesc, getUnderlyingInstr(), getChainOp(),
                                  getVecOp(), getCondOp(), IsOrdered);
   }
@@ -2142,7 +2144,7 @@ class VPReplicateRecipe : public VPRecipeWithIRFlags {
 
   ~VPReplicateRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPReplicateRecipe *clone() override {
     auto *Copy =
         new VPReplicateRecipe(getUnderlyingInstr(), operands(), IsUniform,
                               isPredicated() ? getMask() : nullptr);
@@ -2204,7 +2206,7 @@ class VPBranchOnMaskRecipe : public VPRecipeBase {
       addOperand(BlockInMask);
   }
 
-  VPRecipeBase *clone() override {
+  VPBranchOnMaskRecipe *clone() override {
     return new VPBranchOnMaskRecipe(getOperand(0));
   }
 
@@ -2255,7 +2257,7 @@ class VPPredInstPHIRecipe : public VPSingleDefRecipe {
       : VPSingleDefRecipe(VPDef::VPPredInstPHISC, PredV) {}
   ~VPPredInstPHIRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPPredInstPHIRecipe *clone() override {
     return new VPPredInstPHIRecipe(getOperand(0));
   }
 
@@ -2323,7 +2325,7 @@ class VPWidenMemoryInstructionRecipe : public VPRecipeBase {
     setMask(Mask);
   }
 
-  VPRecipeBase *clone() override {
+  VPWidenMemoryInstructionRecipe *clone() override {
     if (isStore())
       return new VPWidenMemoryInstructionRecipe(
           cast<StoreInst>(Ingredient), getAddr(), getStoredValue(), getMask(),
@@ -2399,7 +2401,9 @@ class VPExpandSCEVRecipe : public VPSingleDefRecipe {
 
   ~VPExpandSCEVRecipe() override = default;
 
-  VPRecipeBase *clone() override { return new VPExpandSCEVRecipe(Expr, SE); }
+  VPExpandSCEVRecipe *clone() override {
+    return new VPExpandSCEVRecipe(Expr, SE);
+  }
 
   VP_CLASSOF_IMPL(VPDef::VPExpandSCEVSC)
 
@@ -2426,7 +2430,7 @@ class VPCanonicalIVPHIRecipe : public VPHeaderPHIRecipe {
 
   ~VPCanonicalIVPHIRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPCanonicalIVPHIRecipe *clone() override {
     auto *R = new VPCanonicalIVPHIRecipe(getOperand(0), getDebugLoc());
     R->addOperand(getBackedgeValue());
     return R;
@@ -2484,7 +2488,7 @@ class VPActiveLaneMaskPHIRecipe : public VPHeaderPHIRecipe {
 
   ~VPActiveLaneMaskPHIRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPActiveLaneMaskPHIRecipe *clone() override {
     return new VPActiveLaneMaskPHIRecipe(getOperand(0), getDebugLoc());
   }
 
@@ -2551,7 +2555,7 @@ class VPWidenCanonicalIVRecipe : public VPSingleDefRecipe {
 
   ~VPWidenCanonicalIVRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPWidenCanonicalIVRecipe *clone() override {
     return new VPWidenCanonicalIVRecipe(
         cast<VPCanonicalIVPHIRecipe>(getOperand(0)));
   }
@@ -2602,7 +2606,7 @@ class VPDerivedIVRecipe : public VPSingleDefRecipe {
 
   ~VPDerivedIVRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPDerivedIVRecipe *clone() override {
     return new VPDerivedIVRecipe(Kind, FPBinOp, getStartValue(), getOperand(1),
                                  getStepValue());
   }
@@ -2656,7 +2660,7 @@ class VPScalarIVStepsRecipe : public VPRecipeWithIRFlags {
 
   ~VPScalarIVStepsRecipe() override = default;
 
-  VPRecipeBase *clone() override {
+  VPScalarIVStepsRecipe *clone() override {
     return new VPScalarIVStepsRecipe(
         getOperand(0), getOperand(1), InductionOpcode,
         hasFastMathFlags() ? getFastMathFlags() : FastMathFlags());

From 462e1023838703f1d3e763869afdd72ec5342a33 Mon Sep 17 00:00:00 2001
From: Poseydon42 <vvmposeydon@gmail.com>
Date: Thu, 11 Apr 2024 10:40:52 +0100
Subject: [PATCH 120/886] [InstCombine] Fold (X / C) < X and (X >> C) < X into
 X > 0 (#85555)

Proofs: https://alive2.llvm.org/ce/z/52droC

This resolves #85313.
---
 .../InstCombine/InstCombineCompares.cpp       |  34 +++++
 .../InstCombine/icmp-div-constant.ll          | 141 ++++++++++++++++++
 2 files changed, 175 insertions(+)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 9ff1e3aa5502e..7292bb62702aa 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -7163,6 +7163,40 @@ Instruction *InstCombinerImpl::foldICmpCommutative(ICmpInst::Predicate Pred,
   if (Value *V = foldICmpWithLowBitMaskedVal(Pred, Op0, Op1, Q, *this))
     return replaceInstUsesWith(CxtI, V);
 
+  // Folding (X / Y) pred X => X swap(pred) 0 for constant Y other than 0 or 1
+  {
+    const APInt *Divisor;
+    if (match(Op0, m_UDiv(m_Specific(Op1), m_APInt(Divisor))) &&
+        Divisor->ugt(1)) {
+      return new ICmpInst(ICmpInst::getSwappedPredicate(Pred), Op1,
+                          Constant::getNullValue(Op1->getType()));
+    }
+
+    if (!ICmpInst::isUnsigned(Pred) &&
+        match(Op0, m_SDiv(m_Specific(Op1), m_APInt(Divisor))) &&
+        Divisor->ugt(1)) {
+      return new ICmpInst(ICmpInst::getSwappedPredicate(Pred), Op1,
+                          Constant::getNullValue(Op1->getType()));
+    }
+  }
+
+  // Another case of this fold is (X >> Y) pred X => X swap(pred) 0 if Y != 0
+  {
+    const APInt *Shift;
+    if (match(Op0, m_LShr(m_Specific(Op1), m_APInt(Shift))) &&
+        !Shift->isZero()) {
+      return new ICmpInst(ICmpInst::getSwappedPredicate(Pred), Op1,
+                          Constant::getNullValue(Op1->getType()));
+    }
+
+    if ((Pred == CmpInst::ICMP_SLT || Pred == CmpInst::ICMP_SGE) &&
+        match(Op0, m_AShr(m_Specific(Op1), m_APInt(Shift))) &&
+        !Shift->isZero()) {
+      return new ICmpInst(ICmpInst::getSwappedPredicate(Pred), Op1,
+                          Constant::getNullValue(Op1->getType()));
+    }
+  }
+
   return nullptr;
 }
 
diff --git a/llvm/test/Transforms/InstCombine/icmp-div-constant.ll b/llvm/test/Transforms/InstCombine/icmp-div-constant.ll
index 8dcb96284685f..b047715432d77 100644
--- a/llvm/test/Transforms/InstCombine/icmp-div-constant.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-div-constant.ll
@@ -375,3 +375,144 @@ define i1 @sdiv_eq_smin_use(i32 %x, i32 %y) {
   %r = icmp eq i32 %d, -2147483648
   ret i1 %r
 }
+
+; Fold (X / C) cmp X into X ~cmp 0 (~cmp is the inverse predicate of cmp), for some C != 1
+; Alternative form of this fold is when division is replaced with logic right shift
+
+define i1 @sdiv_x_by_const_cmp_x(i32 %x) {
+; CHECK-LABEL: @sdiv_x_by_const_cmp_x(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %v = sdiv i32 %x, 13
+  %r = icmp eq i32 %v, %x
+  ret i1 %r
+}
+
+define i1 @udiv_x_by_const_cmp_x(i32 %x) {
+; CHECK-LABEL: @udiv_x_by_const_cmp_x(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sgt i32 [[X:%.*]], 0
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %1 = udiv i32 %x, 123
+  %2 = icmp slt i32 %1, %x
+  ret i1 %2
+}
+
+; Same as above but with right shift instead of division (C != 0)
+
+define i1 @lshr_x_by_const_cmp_x(i32 %x) {
+; CHECK-LABEL: @lshr_x_by_const_cmp_x(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[X:%.*]], 0
+; CHECK-NEXT:    ret i1 [[TMP1]]
+;
+  %v = lshr i32 %x, 1
+  %r = icmp eq i32 %v, %x
+  ret i1 %r
+}
+
+define <4 x i1> @lshr_by_const_cmp_sle_value(<4 x i32> %x) {
+; CHECK-LABEL: @lshr_by_const_cmp_sle_value(
+; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[X:%.*]], <i32 -1, i32 -1, i32 -1, i32 -1>
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %v = lshr <4 x i32> %x, <i32 3, i32 3, i32 3, i32 3>
+  %r = icmp sle <4 x i32> %v, %x
+  ret <4 x i1> %r
+}
+
+define i1 @lshr_by_const_cmp_sge_value(i32 %x) {
+; CHECK-LABEL: @lshr_by_const_cmp_sge_value(
+; CHECK-NEXT:    [[R:%.*]] = icmp slt i32 [[X:%.*]], 1
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %v = lshr i32 %x, 3
+  %r = icmp sge i32 %v, %x
+  ret i1 %r
+}
+
+define i1 @ashr_x_by_const_cmp_sge_x(i32 %x) {
+; CHECK-LABEL: @ashr_x_by_const_cmp_sge_x(
+; CHECK-NEXT:    [[R:%.*]] = icmp slt i32 [[X:%.*]], 1
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %v = ashr i32 %x, 5
+  %r = icmp sge i32 %v, %x
+  ret i1 %r
+}
+
+; Negative test - constant is 1
+
+define <2 x i1> @udiv_x_by_const_cmp_eq_value_neg(<2 x i32> %x) {
+; CHECK-LABEL: @udiv_x_by_const_cmp_eq_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = udiv <2 x i32> [[X:%.*]], <i32 1, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i32> [[V]], [[X]]
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %v = udiv <2 x i32> %x, <i32 1, i32 3>
+  %r = icmp eq <2 x i32> %v, %x
+  ret <2 x i1> %r
+}
+
+define <2 x i1> @sdiv_x_by_const_cmp_eq_value_neg(<2 x i32> %x) {
+; CHECK-LABEL: @sdiv_x_by_const_cmp_eq_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = sdiv <2 x i32> [[X:%.*]], <i32 1, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i32> [[V]], [[X]]
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %v = sdiv <2 x i32> %x, <i32 1, i32 3>
+  %r = icmp eq <2 x i32> %v, %x
+  ret <2 x i1> %r
+}
+
+; Negative test - constant is 0
+
+define <2 x i1> @lshr_x_by_const_cmp_slt_value_neg(<2 x i32> %x) {
+; CHECK-LABEL: @lshr_x_by_const_cmp_slt_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = lshr <2 x i32> [[X:%.*]], <i32 0, i32 2>
+; CHECK-NEXT:    [[R:%.*]] = icmp slt <2 x i32> [[V]], [[X]]
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %v = lshr <2 x i32> %x, <i32 0, i32 2>
+  %r = icmp slt <2 x i32> %v, %x
+  ret <2 x i1> %r
+}
+
+; Negative test - unsigned predicate with sdiv
+
+define i1 @sdiv_x_by_const_cmp_ult_value_neg(i32 %x) {
+; CHECK-LABEL: @sdiv_x_by_const_cmp_ult_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = sdiv i32 [[X:%.*]], 3
+; CHECK-NEXT:    [[R:%.*]] = icmp ult i32 [[V]], [[X]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %v = sdiv i32 %x, 3
+  %r = icmp ult i32 %v, %x
+  ret i1 %r
+}
+
+; Negative case - one of the components of a vector is 1
+
+define <4 x i1> @sdiv_x_by_const_cmp_sgt_value_neg(<4 x i32> %x) {
+; CHECK-LABEL: @sdiv_x_by_const_cmp_sgt_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = sdiv <4 x i32> [[X:%.*]], <i32 1, i32 2, i32 3, i32 4>
+; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[V]], [[X]]
+; CHECK-NEXT:    ret <4 x i1> [[R]]
+;
+  %v = sdiv <4 x i32> %x, <i32 1, i32 2, i32 3, i32 4>
+  %r = icmp sgt <4 x i32> %v, %x
+  ret <4 x i1> %r
+}
+
+; Negative case - ashr only allows sge/slt predicates
+
+define i1 @ashr_x_by_const_cmp_sle_value_neg(i32 %x) {
+; CHECK-LABEL: @ashr_x_by_const_cmp_sle_value_neg(
+; CHECK-NEXT:    [[V:%.*]] = ashr i32 [[X:%.*]], 3
+; CHECK-NEXT:    [[R:%.*]] = icmp sle i32 [[V]], [[X]]
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %v = ashr i32 %x, 3
+  %r = icmp sle i32 %v, %x
+  ret i1 %r
+}

From 82ae646eb49cfd762db7db0a74b130970fe45d97 Mon Sep 17 00:00:00 2001
From: Thomas Preud'homme <thomas.preudhomme@arm.com>
Date: Thu, 11 Apr 2024 10:41:07 +0100
Subject: [PATCH 121/886] [llvm-mca] Remove spurious include_directories()
 (#88277)

llvm-mca does not have an include directory so this commit removes the
spurious include_directories directive.
---
 llvm/tools/llvm-mca/CMakeLists.txt | 2 --
 1 file changed, 2 deletions(-)

diff --git a/llvm/tools/llvm-mca/CMakeLists.txt b/llvm/tools/llvm-mca/CMakeLists.txt
index 878a05c51cfb4..4ef8b9afa12a7 100644
--- a/llvm/tools/llvm-mca/CMakeLists.txt
+++ b/llvm/tools/llvm-mca/CMakeLists.txt
@@ -1,5 +1,3 @@
-include_directories(include)
-
 set(LLVM_LINK_COMPONENTS
   AllTargetsAsmParsers
   AllTargetsMCAs          # CustomBehaviour and InstrPostProcess

From e7bc53726459bba3a48b1f529f1fd9472ad9051c Mon Sep 17 00:00:00 2001
From: Andreas Jonson <andjo403@hotmail.com>
Date: Thu, 11 Apr 2024 11:42:59 +0200
Subject: [PATCH 122/886] [IPSCCP] Add range attribute handling (#86747)

Support the new range attribute to infer ConstantRanges in IPSCCP.
---
 .../llvm/Transforms/Utils/SCCPSolver.h        |   4 +
 llvm/lib/Transforms/IPO/SCCP.cpp              |   3 +-
 llvm/lib/Transforms/Utils/SCCPSolver.cpp      |  40 ++++-
 llvm/test/Transforms/SCCP/range-attribute.ll  | 154 ++++++++++++++++++
 4 files changed, 197 insertions(+), 4 deletions(-)
 create mode 100644 llvm/test/Transforms/SCCP/range-attribute.ll

diff --git a/llvm/include/llvm/Transforms/Utils/SCCPSolver.h b/llvm/include/llvm/Transforms/Utils/SCCPSolver.h
index 1a95f80812aab..9f7ccd4a8a32c 100644
--- a/llvm/include/llvm/Transforms/Utils/SCCPSolver.h
+++ b/llvm/include/llvm/Transforms/Utils/SCCPSolver.h
@@ -151,6 +151,10 @@ class SCCPSolver {
   /// works with both scalars and structs.
   void markOverdefined(Value *V);
 
+  /// trackValueOfArgument - Mark the specified argument overdefined unless it
+  /// have range attribute.  This works with both scalars and structs.
+  void trackValueOfArgument(Argument *V);
+
   // isStructLatticeConstant - Return true if all the lattice values
   // corresponding to elements of the structure are constants,
   // false otherwise.
diff --git a/llvm/lib/Transforms/IPO/SCCP.cpp b/llvm/lib/Transforms/IPO/SCCP.cpp
index b1f9b827dcbaf..f8920541e6fd6 100644
--- a/llvm/lib/Transforms/IPO/SCCP.cpp
+++ b/llvm/lib/Transforms/IPO/SCCP.cpp
@@ -144,9 +144,8 @@ static bool runIPSCCP(
     // Assume the function is called.
     Solver.markBlockExecutable(&F.front());
 
-    // Assume nothing about the incoming arguments.
     for (Argument &AI : F.args())
-      Solver.markOverdefined(&AI);
+      Solver.trackValueOfArgument(&AI);
   }
 
   // Determine if we can track any of the module's global variables. If so, add
diff --git a/llvm/lib/Transforms/Utils/SCCPSolver.cpp b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
index 38fc7763c5b20..b82ed25a6d0cf 100644
--- a/llvm/lib/Transforms/Utils/SCCPSolver.cpp
+++ b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
@@ -428,6 +428,13 @@ class SCCPInstVisitor : public InstVisitor<SCCPInstVisitor> {
     return markConstant(ValueState[V], V, C);
   }
 
+  /// markConstantRange - Mark the object as constant range with \p CR. If the
+  /// object is not a constant range with the range \p CR, add it to the
+  /// instruction work list so that the users of the instruction are updated
+  /// later.
+  bool markConstantRange(ValueLatticeElement &IV, Value *V,
+                         const ConstantRange &CR);
+
   // markOverdefined - Make a value be marked as "overdefined". If the
   // value is not already overdefined, add it to the overdefined instruction
   // work list so that the users of the instruction are updated later.
@@ -788,6 +795,17 @@ class SCCPInstVisitor : public InstVisitor<SCCPInstVisitor> {
       markOverdefined(ValueState[V], V);
   }
 
+  void trackValueOfArgument(Argument *A) {
+    if (A->getType()->isIntegerTy()) {
+      if (std::optional<ConstantRange> Range = A->getRange()) {
+        markConstantRange(ValueState[A], A, *Range);
+        return;
+      }
+    }
+    // Assume nothing about the incoming arguments without range.
+    markOverdefined(A);
+  }
+
   bool isStructLatticeConstant(Function *F, StructType *STy);
 
   Constant *getConstant(const ValueLatticeElement &LV, Type *Ty) const;
@@ -873,6 +891,15 @@ bool SCCPInstVisitor::markConstant(ValueLatticeElement &IV, Value *V,
   return true;
 }
 
+bool SCCPInstVisitor::markConstantRange(ValueLatticeElement &IV, Value *V,
+                                        const ConstantRange &CR) {
+  if (!IV.markConstantRange(CR))
+    return false;
+  LLVM_DEBUG(dbgs() << "markConstantRange: " << CR << ": " << *V << '\n');
+  pushToWorkList(IV, V);
+  return true;
+}
+
 bool SCCPInstVisitor::markOverdefined(ValueLatticeElement &IV, Value *V) {
   if (!IV.markOverdefined())
     return false;
@@ -1581,10 +1608,15 @@ void SCCPInstVisitor::visitStoreInst(StoreInst &SI) {
 }
 
 static ValueLatticeElement getValueFromMetadata(const Instruction *I) {
-  if (MDNode *Ranges = I->getMetadata(LLVMContext::MD_range))
-    if (I->getType()->isIntegerTy())
+  if (I->getType()->isIntegerTy()) {
+    if (MDNode *Ranges = I->getMetadata(LLVMContext::MD_range))
       return ValueLatticeElement::getRange(
           getConstantRangeFromMetadata(*Ranges));
+
+    if (const auto *CB = dyn_cast<CallBase>(I))
+      if (std::optional<ConstantRange> Range = CB->getRange())
+        return ValueLatticeElement::getRange(*Range);
+  }
   if (I->hasMetadata(LLVMContext::MD_nonnull))
     return ValueLatticeElement::getNot(
         ConstantPointerNull::get(cast<PointerType>(I->getType())));
@@ -2090,6 +2122,10 @@ const SmallPtrSet<Function *, 16> SCCPSolver::getMRVFunctionsTracked() {
 
 void SCCPSolver::markOverdefined(Value *V) { Visitor->markOverdefined(V); }
 
+void SCCPSolver::trackValueOfArgument(Argument *V) {
+  Visitor->trackValueOfArgument(V);
+}
+
 bool SCCPSolver::isStructLatticeConstant(Function *F, StructType *STy) {
   return Visitor->isStructLatticeConstant(F, STy);
 }
diff --git a/llvm/test/Transforms/SCCP/range-attribute.ll b/llvm/test/Transforms/SCCP/range-attribute.ll
new file mode 100644
index 0000000000000..ae66af91bb8e3
--- /dev/null
+++ b/llvm/test/Transforms/SCCP/range-attribute.ll
@@ -0,0 +1,154 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -passes=ipsccp -S | FileCheck %s
+
+declare void @use(i1)
+declare i32 @get_i32()
+
+define void @range_attribute(i32 range(i32 0, 10) %v) {
+; CHECK-LABEL: @range_attribute(
+; CHECK-NEXT:    call void @use(i1 true)
+; CHECK-NEXT:    [[C2:%.*]] = icmp ult i32 [[V:%.*]], 9
+; CHECK-NEXT:    call void @use(i1 [[C2]])
+; CHECK-NEXT:    call void @use(i1 false)
+; CHECK-NEXT:    [[C4:%.*]] = icmp ugt i32 [[V]], 8
+; CHECK-NEXT:    call void @use(i1 [[C4]])
+; CHECK-NEXT:    ret void
+;
+  %c1 = icmp ult i32 %v, 10
+  call void @use(i1 %c1)
+  %c2 = icmp ult i32 %v, 9
+  call void @use(i1 %c2)
+  %c3 = icmp ugt i32 %v, 9
+  call void @use(i1 %c3)
+  %c4 = icmp ugt i32 %v, 8
+  call void @use(i1 %c4)
+  ret void
+}
+
+define i32 @range_attribute_single(i32 range(i32 0, 1) %v) {
+; CHECK-LABEL: @range_attribute_single(
+; CHECK-NEXT:    ret i32 0
+;
+  ret i32 %v
+}
+
+define void @call_range_attribute() {
+; CHECK-LABEL: @call_range_attribute(
+; CHECK-NEXT:    [[V:%.*]] = call range(i32 0, 10) i32 @get_i32()
+; CHECK-NEXT:    call void @use(i1 true)
+; CHECK-NEXT:    [[C2:%.*]] = icmp ult i32 [[V]], 9
+; CHECK-NEXT:    call void @use(i1 [[C2]])
+; CHECK-NEXT:    call void @use(i1 false)
+; CHECK-NEXT:    [[C4:%.*]] = icmp ugt i32 [[V]], 8
+; CHECK-NEXT:    call void @use(i1 [[C4]])
+; CHECK-NEXT:    ret void
+;
+  %v = call range(i32 0, 10) i32 @get_i32()
+  %c1 = icmp ult i32 %v, 10
+  call void @use(i1 %c1)
+  %c2 = icmp ult i32 %v, 9
+  call void @use(i1 %c2)
+  %c3 = icmp ugt i32 %v, 9
+  call void @use(i1 %c3)
+  %c4 = icmp ugt i32 %v, 8
+  call void @use(i1 %c4)
+  ret void
+}
+
+
+declare range(i32 0, 10) i32 @get_i32_in_range()
+
+define void @call_range_result() {
+; CHECK-LABEL: @call_range_result(
+; CHECK-NEXT:    [[V:%.*]] = call i32 @get_i32_in_range()
+; CHECK-NEXT:    call void @use(i1 true)
+; CHECK-NEXT:    [[C2:%.*]] = icmp ult i32 [[V]], 9
+; CHECK-NEXT:    call void @use(i1 [[C2]])
+; CHECK-NEXT:    call void @use(i1 false)
+; CHECK-NEXT:    [[C4:%.*]] = icmp ugt i32 [[V]], 8
+; CHECK-NEXT:    call void @use(i1 [[C4]])
+; CHECK-NEXT:    ret void
+;
+  %v = call i32 @get_i32_in_range()
+  %c1 = icmp ult i32 %v, 10
+  call void @use(i1 %c1)
+  %c2 = icmp ult i32 %v, 9
+  call void @use(i1 %c2)
+  %c3 = icmp ugt i32 %v, 9
+  call void @use(i1 %c3)
+  %c4 = icmp ugt i32 %v, 8
+  call void @use(i1 %c4)
+  ret void
+}
+
+define internal i1 @ip_cmp_range_attribute(i32 %v) {
+; CHECK-LABEL: @ip_cmp_range_attribute(
+; CHECK-NEXT:    ret i1 undef
+;
+  %c = icmp ult i32 %v, 10
+  ret i1 %c
+}
+
+define i1 @ip_range_attribute(i32 range(i32 0, 10) %v) {
+; CHECK-LABEL: @ip_range_attribute(
+; CHECK-NEXT:    [[C:%.*]] = call i1 @ip_cmp_range_attribute(i32 [[V:%.*]])
+; CHECK-NEXT:    ret i1 true
+;
+  %c = call i1 @ip_cmp_range_attribute(i32 %v)
+  ret i1 %c
+}
+
+define internal i1 @ip_cmp_range_call(i32 %v) {
+; CHECK-LABEL: @ip_cmp_range_call(
+; CHECK-NEXT:    ret i1 undef
+;
+  %c = icmp ult i32 %v, 10
+  ret i1 %c
+}
+
+define i1 @ip_range_call() {
+; CHECK-LABEL: @ip_range_call(
+; CHECK-NEXT:    [[V:%.*]] = call range(i32 0, 10) i32 @get_i32()
+; CHECK-NEXT:    [[C:%.*]] = call i1 @ip_cmp_range_call(i32 [[V]])
+; CHECK-NEXT:    ret i1 true
+;
+  %v = call range(i32 0, 10) i32 @get_i32()
+  %c = call i1 @ip_cmp_range_call(i32 %v)
+  ret i1 %c
+}
+
+define internal i1 @ip_cmp_range_result(i32 %v) {
+; CHECK-LABEL: @ip_cmp_range_result(
+; CHECK-NEXT:    ret i1 undef
+;
+  %c = icmp ult i32 %v, 10
+  ret i1 %c
+}
+
+define i1 @ip_range_result() {
+; CHECK-LABEL: @ip_range_result(
+; CHECK-NEXT:    [[V:%.*]] = call range(i32 0, 10) i32 @get_i32()
+; CHECK-NEXT:    [[C:%.*]] = call i1 @ip_cmp_range_result(i32 [[V]])
+; CHECK-NEXT:    ret i1 true
+;
+  %v = call range(i32 0, 10) i32 @get_i32()
+  %c = call i1 @ip_cmp_range_result(i32 %v)
+  ret i1 %c
+}
+
+define internal i1 @ip_cmp_with_range_attribute(i32 range(i32 0, 10) %v) {
+; CHECK-LABEL: @ip_cmp_with_range_attribute(
+; CHECK-NEXT:    ret i1 undef
+;
+  %c = icmp eq i32 %v, 5
+  ret i1 %c
+}
+
+define i1 @ip_range_attribute_constant() {
+; CHECK-LABEL: @ip_range_attribute_constant(
+; CHECK-NEXT:    [[C:%.*]] = call i1 @ip_cmp_with_range_attribute(i32 5)
+; CHECK-NEXT:    ret i1 true
+;
+  %c = call i1 @ip_cmp_with_range_attribute(i32 5)
+  ret i1 %c
+}

From eef63d3c92766c6f8e78eefb9bb37ae01fbedbfc Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Thu, 11 Apr 2024 10:43:49 +0100
Subject: [PATCH 123/886] [mlir][OpenMP] add missing load for reduction cleanup
 region (#88289)

I missed this before. For by-ref reductions, the private reduction
variable is a pointer to the pointer to the variable. So an extra load
is required to get the right value.

See the "red.private.value.n" loads in the reduction combiner region for
reference.
---
 .../Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp    | 13 +++++++++++--
 .../LLVMIR/openmp-parallel-reduction-cleanup.mlir   |  6 ++++--
 .../LLVMIR/openmp-wsloop-reduction-cleanup.mlir     |  6 ++++--
 3 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index a59677c02fc39..300fc8ba56fc5 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -889,8 +889,17 @@ static LogicalResult inlineReductionCleanup(
 
     // map the argument to the cleanup region
     Block &entry = cleanupRegion.front();
-    moduleTranslation.mapValue(entry.getArgument(0),
-                               privateReductionVariables[i]);
+
+    llvm::Instruction *potentialTerminator =
+        builder.GetInsertBlock()->empty() ? nullptr
+                                          : &builder.GetInsertBlock()->back();
+    if (potentialTerminator && potentialTerminator->isTerminator())
+      builder.SetInsertPoint(potentialTerminator);
+    llvm::Value *reductionVar = builder.CreateLoad(
+        moduleTranslation.convertType(entry.getArgument(0).getType()),
+        privateReductionVariables[i]);
+
+    moduleTranslation.mapValue(entry.getArgument(0), reductionVar);
 
     if (failed(inlineConvertOmpRegions(cleanupRegion, "omp.reduction.cleanup",
                                        builder, moduleTranslation)))
diff --git a/mlir/test/Target/LLVMIR/openmp-parallel-reduction-cleanup.mlir b/mlir/test/Target/LLVMIR/openmp-parallel-reduction-cleanup.mlir
index 9ae4c4ad392b1..b7f71f438e56b 100644
--- a/mlir/test/Target/LLVMIR/openmp-parallel-reduction-cleanup.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-parallel-reduction-cleanup.mlir
@@ -86,8 +86,10 @@
 
 // Cleanup region:
 // CHECK: [[OMP_FINALIZE]]:
-// CHECK: call void @free(ptr %[[PRIV_PTR_I]])
-// CHECK: call void @free(ptr %[[PRIV_PTR_J]])
+// CHECK: %[[PRIV_I:.+]] = load ptr, ptr %[[PRIV_PTR_I]], align 8
+// CHECK: call void @free(ptr %[[PRIV_I]])
+// CHECK: %[[PRIV_J:.+]] = load ptr, ptr %[[PRIV_PTR_J]], align 8
+// CHECK: call void @free(ptr %[[PRIV_J]])
 
 // Reduction function.
 // CHECK: define internal void @[[REDFUNC]]
diff --git a/mlir/test/Target/LLVMIR/openmp-wsloop-reduction-cleanup.mlir b/mlir/test/Target/LLVMIR/openmp-wsloop-reduction-cleanup.mlir
index a1e17afa53e21..3842522934e48 100644
--- a/mlir/test/Target/LLVMIR/openmp-wsloop-reduction-cleanup.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-wsloop-reduction-cleanup.mlir
@@ -63,8 +63,10 @@
 // Weirdly the finalization block is generated before the reduction blocks:
 // CHECK: [[FINALIZE:.+]]:
 // CHECK: call void @__kmpc_barrier
-// CHECK: call void @free(ptr %[[PRIV_PTR_I]])
-// CHECK: call void @free(ptr %[[PRIV_PTR_J]])
+// CHECK: %[[PRIV_I:.+]] = load ptr, ptr %[[PRIV_PTR_I]], align 8
+// CHECK: call void @free(ptr %[[PRIV_I]])
+// CHECK: %[[PRIV_J:.+]] = load ptr, ptr %[[PRIV_PTR_J]], align 8
+// CHECK: call void @free(ptr %[[PRIV_J]])
 // CHECK: ret void
 
 // Non-atomic reduction:

From 6f068b9cf1ac09945c096269f0c6c276d2ec95c4 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Thu, 11 Apr 2024 10:44:09 +0100
Subject: [PATCH 124/886] [flang][OpenMP] Allocate array reduction variables on
 the heap (#87773)

Following up on a review comment:
https://github.com/llvm/llvm-project/pull/84958#discussion_r1527627848

Reductions might be inlined inside of a loop so stack allocations are
not safe.

Normally flang allocates arrays on the stack. Allocatable arrays have a
different type: fir.box<fir.heap<fir.array<...>>> instead of
fir.box<fir.array<...>>. This patch will allocate all arrays on the
heap.

Reductions on allocatable arrays still aren't supported (but I will get
to this soon).
---
 flang/lib/Lower/OpenMP/ReductionProcessor.cpp | 78 +++++++++++++++++--
 .../Lower/OpenMP/parallel-reduction-array.f90 | 20 ++++-
 .../OpenMP/parallel-reduction-array2.f90      | 20 ++++-
 .../test/Lower/OpenMP/parallel-reduction3.f90 | 31 ++++----
 .../wsloop-reduction-array-assumed-shape.f90  | 17 +++-
 .../Lower/OpenMP/wsloop-reduction-array.f90   | 19 ++++-
 .../Lower/OpenMP/wsloop-reduction-array2.f90  | 21 ++++-
 7 files changed, 171 insertions(+), 35 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index 0453c01522779..918edf27baf66 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -20,6 +20,7 @@
 #include "flang/Optimizer/Builder/Todo.h"
 #include "flang/Optimizer/Dialect/FIRType.h"
 #include "flang/Optimizer/HLFIR/HLFIROps.h"
+#include "flang/Optimizer/Support/FatalError.h"
 #include "flang/Parser/tools.h"
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
 #include "llvm/Support/CommandLine.h"
@@ -391,8 +392,60 @@ static void genCombiner(fir::FirOpBuilder &builder, mlir::Location loc,
   TODO(loc, "OpenMP genCombiner for unsupported reduction variable type");
 }
 
+static void
+createReductionCleanupRegion(fir::FirOpBuilder &builder, mlir::Location loc,
+                             mlir::omp::DeclareReductionOp &reductionDecl) {
+  mlir::Type redTy = reductionDecl.getType();
+
+  mlir::Region &cleanupRegion = reductionDecl.getCleanupRegion();
+  assert(cleanupRegion.empty());
+  mlir::Block *block =
+      builder.createBlock(&cleanupRegion, cleanupRegion.end(), {redTy}, {loc});
+  builder.setInsertionPointToEnd(block);
+
+  auto typeError = [loc]() {
+    fir::emitFatalError(loc,
+                        "Attempt to create an omp reduction cleanup region "
+                        "for a type that wasn't allocated",
+                        /*genCrashDiag=*/true);
+  };
+
+  mlir::Type valTy = fir::unwrapRefType(redTy);
+  if (auto boxTy = mlir::dyn_cast_or_null<fir::BaseBoxType>(valTy)) {
+    mlir::Type innerTy = fir::extractSequenceType(boxTy);
+    if (!mlir::isa<fir::SequenceType>(innerTy))
+      typeError();
+
+    mlir::Value arg = block->getArgument(0);
+    arg = builder.loadIfRef(loc, arg);
+    assert(mlir::isa<fir::BaseBoxType>(arg.getType()));
+
+    // Deallocate box
+    // The FIR type system doesn't nesecarrily know that this is a mutable box
+    // if we allocated the thread local array on the heap to avoid looped stack
+    // allocations.
+    mlir::Value addr =
+        hlfir::genVariableRawAddress(loc, builder, hlfir::Entity{arg});
+    mlir::Value isAllocated = builder.genIsNotNullAddr(loc, addr);
+    fir::IfOp ifOp =
+        builder.create<fir::IfOp>(loc, isAllocated, /*withElseRegion=*/false);
+    builder.setInsertionPointToStart(&ifOp.getThenRegion().front());
+
+    mlir::Value cast = builder.createConvert(
+        loc, fir::HeapType::get(fir::dyn_cast_ptrEleTy(addr.getType())), addr);
+    builder.create<fir::FreeMemOp>(loc, cast);
+
+    builder.setInsertionPointAfter(ifOp);
+    builder.create<mlir::omp::YieldOp>(loc);
+    return;
+  }
+
+  typeError();
+}
+
 static mlir::Value
 createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
+                          mlir::omp::DeclareReductionOp &reductionDecl,
                           const ReductionProcessor::ReductionIdentifier redId,
                           mlir::Type type, bool isByRef) {
   mlir::Type ty = fir::unwrapRefType(type);
@@ -419,11 +472,24 @@ createReductionInitRegion(fir::FirOpBuilder &builder, mlir::Location loc,
     // Create the private copy from the initial fir.box:
     hlfir::Entity source = hlfir::Entity{builder.getBlock()->getArgument(0)};
 
-    // TODO: if the whole reduction is nested inside of a loop, this alloca
-    // could lead to a stack overflow (the memory is only freed at the end of
-    // the stack frame). The reduction declare operation needs a deallocation
-    // region to undo the init region.
-    hlfir::Entity temp = createStackTempFromMold(loc, builder, source);
+    // Allocating on the heap in case the whole reduction is nested inside of a
+    // loop
+    // TODO: compare performance here to using allocas - this could be made to
+    // work by inserting stacksave/stackrestore around the reduction in
+    // openmpirbuilder
+    auto [temp, needsDealloc] = createTempFromMold(loc, builder, source);
+    // if needsDealloc isn't statically false, add cleanup region. TODO: always
+    // do this for allocatable boxes because they might have been re-allocated
+    // in the body of the loop/parallel region
+    std::optional<int64_t> cstNeedsDealloc =
+        fir::getIntIfConstant(needsDealloc);
+    assert(cstNeedsDealloc.has_value() &&
+           "createTempFromMold decides this statically");
+    if (cstNeedsDealloc.has_value() && *cstNeedsDealloc != false) {
+      auto insPt = builder.saveInsertionPoint();
+      createReductionCleanupRegion(builder, loc, reductionDecl);
+      builder.restoreInsertionPoint(insPt);
+    }
 
     // Put the temporary inside of a box:
     hlfir::Entity box = hlfir::genVariableBox(loc, builder, temp);
@@ -462,7 +528,7 @@ mlir::omp::DeclareReductionOp ReductionProcessor::createDeclareReduction(
   builder.setInsertionPointToEnd(&decl.getInitializerRegion().back());
 
   mlir::Value init =
-      createReductionInitRegion(builder, loc, redId, type, isByRef);
+      createReductionInitRegion(builder, loc, decl, redId, type, isByRef);
   builder.create<mlir::omp::YieldOp>(loc, init);
 
   builder.createBlock(&decl.getReductionRegion(),
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-array.f90 b/flang/test/Lower/OpenMP/parallel-reduction-array.f90
index 56dcabbb75c3a..26c9d4f085096 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction-array.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction-array.f90
@@ -15,13 +15,15 @@ program reduce
 
 ! CHECK-LABEL:   omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> init {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
-! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.array<3xi32> {bindc_name = ".tmp"}
 ! CHECK:           %[[VAL_2:.*]] = arith.constant 0 : i32
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
 ! CHECK:           %[[VAL_4:.*]] = arith.constant 3 : index
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<3xi32>>, !fir.ref<!fir.array<3xi32>>)
-! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<3xi32>>
+! CHECK:           %[[VAL_1:.*]] = fir.allocmem !fir.array<3xi32> {bindc_name = ".tmp", uniq_name = ""}
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<3xi32>>,
+!fir.shape<1>) -> (!fir.heap<!fir.array<3xi32>>, !fir.heap<!fir.array<3xi32>>)
+! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.heap<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<3xi32>>
 ! CHECK:           hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<3xi32>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
 ! CHECK:           fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
@@ -43,6 +45,18 @@ program reduce
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<3xi32>>) -> !fir.ref<!fir.array<3xi32>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<3xi32>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<3xi32>>) -> !fir.heap<!fir.array<3xi32>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<3xi32>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL:   func.func @_QQmain()
diff --git a/flang/test/Lower/OpenMP/parallel-reduction-array2.f90 b/flang/test/Lower/OpenMP/parallel-reduction-array2.f90
index 94bff410a2f0d..bed04401248be 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction-array2.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction-array2.f90
@@ -15,13 +15,15 @@ program reduce
 
 ! CHECK-LABEL:   omp.declare_reduction @add_reduction_byref_box_3xi32 : !fir.ref<!fir.box<!fir.array<3xi32>>> init {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
-! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.array<3xi32> {bindc_name = ".tmp"}
 ! CHECK:           %[[VAL_2:.*]] = arith.constant 0 : i32
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
 ! CHECK:           %[[VAL_4:.*]] = arith.constant 3 : index
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<3xi32>>, !fir.ref<!fir.array<3xi32>>)
-! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.ref<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<3xi32>>
+! CHECK:           %[[VAL_1:.*]] = fir.allocmem !fir.array<3xi32>
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<3xi32>>,
+!fir.shape<1>) -> (!fir.heap<!fir.array<3xi32>>, !fir.heap<!fir.array<3xi32>>)
+! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.heap<!fir.array<3xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<3xi32>>
 ! CHECK:           hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<3xi32>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<3xi32>>
 ! CHECK:           fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
@@ -43,6 +45,18 @@ program reduce
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<3xi32>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<3xi32>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<3xi32>>) -> !fir.ref<!fir.array<3xi32>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<3xi32>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<3xi32>>) -> !fir.heap<!fir.array<3xi32>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<3xi32>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL:   func.func @_QQmain() attributes {fir.bindc_name = "reduce"} {
diff --git a/flang/test/Lower/OpenMP/parallel-reduction3.f90 b/flang/test/Lower/OpenMP/parallel-reduction3.f90
index b25759713e318..ce6bd17265ddb 100644
--- a/flang/test/Lower/OpenMP/parallel-reduction3.f90
+++ b/flang/test/Lower/OpenMP/parallel-reduction3.f90
@@ -1,15 +1,6 @@
-! NOTE: Assertions have been autogenerated by utils/generate-test-checks.py
-
-! The script is designed to make adding checks to
-! a test case fast, it is *not* designed to be authoritative
-! about what constitutes a good test! The CHECK should be
-! minimized and named to reflect the test intent.
-
 ! RUN: bbc -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s
 ! RUN: %flang_fc1 -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s
 
-
-
 ! CHECK-LABEL:   omp.declare_reduction @add_reduction_byref_box_Uxi32 : !fir.ref<!fir.box<!fir.array<?xi32>>> init {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<?xi32>>>):
 ! CHECK:           %[[VAL_1:.*]] = arith.constant 0 : i32
@@ -17,14 +8,14 @@
 ! CHECK:           %[[VAL_3:.*]] = arith.constant 0 : index
 ! CHECK:           %[[VAL_4:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_3]] : (!fir.box<!fir.array<?xi32>>, index) -> (index, index, index)
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]]#1 : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]] = fir.alloca !fir.array<?xi32>, %[[VAL_4]]#1 {bindc_name = ".tmp"}
-! CHECK:           %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_6]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<?xi32>>, !fir.shape<1>) -> (!fir.box<!fir.array<?xi32>>, !fir.ref<!fir.array<?xi32>>)
+! CHECK:           %[[VAL_6:.*]] = fir.allocmem !fir.array<?xi32>, %[[VAL_4]]#1 {bindc_name = ".tmp", uniq_name = ""}
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_6]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<?xi32>>, !fir.shape<1>) -> (!fir.box<!fir.array<?xi32>>, !fir.heap<!fir.array<?xi32>>)
 ! CHECK:           hlfir.assign %[[VAL_1]] to %[[VAL_7]]#0 : i32, !fir.box<!fir.array<?xi32>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<?xi32>>
 ! CHECK:           fir.store %[[VAL_7]]#0 to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<?xi32>>>
 ! CHECK:           omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<?xi32>>>)
-
-! CHECK-LABEL:   } combiner {
+! CHECK:         } combiner {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<?xi32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.array<?xi32>>>):
 ! CHECK:           %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xi32>>>
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_1]] : !fir.ref<!fir.box<!fir.array<?xi32>>>
@@ -41,6 +32,18 @@
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xi32>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<?xi32>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xi32>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<?xi32>>) -> !fir.ref<!fir.array<?xi32>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<?xi32>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<?xi32>>) -> !fir.heap<!fir.array<?xi32>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<?xi32>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL:   func.func @_QPs(
@@ -122,4 +125,4 @@ subroutine s(x)
     !$omp end parallel do
 
     if (c(1) /= 5050) stop 1
-end subroutine s
\ No newline at end of file
+end subroutine s
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-array-assumed-shape.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-array-assumed-shape.f90
index a1f339faea5cd..8f83a30c9fe78 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-array-assumed-shape.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-array-assumed-shape.f90
@@ -29,8 +29,9 @@ subroutine reduce(r)
 ! CHECK:           %[[VAL_3:.*]] = arith.constant 0 : index
 ! CHECK:           %[[VAL_4:.*]]:3 = fir.box_dims %[[VAL_2]], %[[VAL_3]] : (!fir.box<!fir.array<?xf64>>, index) -> (index, index, index)
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]]#1 : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]] = fir.alloca !fir.array<?xf64>, %[[VAL_4]]#1 {bindc_name = ".tmp"}
-! CHECK:           %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_6]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<?xf64>>, !fir.shape<1>) -> (!fir.box<!fir.array<?xf64>>, !fir.ref<!fir.array<?xf64>>)
+! CHECK:           %[[VAL_6:.*]] = fir.allocmem !fir.array<?xf64>, %[[VAL_4]]#1 {bindc_name = ".tmp", uniq_name = ""}
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_7:.*]]:2 = hlfir.declare %[[VAL_6]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<?xf64>>, !fir.shape<1>) -> (!fir.box<!fir.array<?xf64>>, !fir.heap<!fir.array<?xf64>>)
 ! CHECK:           hlfir.assign %[[VAL_1]] to %[[VAL_7]]#0 : f64, !fir.box<!fir.array<?xf64>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<?xf64>>
 ! CHECK:           fir.store %[[VAL_7]]#0 to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<?xf64>>>
@@ -53,6 +54,18 @@ subroutine reduce(r)
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<f64>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf64>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<?xf64>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf64>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<?xf64>>) -> !fir.ref<!fir.array<?xf64>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<?xf64>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<?xf64>>) -> !fir.heap<!fir.array<?xf64>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<?xf64>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL:   func.func private @_QFPreduce(
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-array.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-array.f90
index a898204c881d9..a08bca9eb283b 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-array.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-array.f90
@@ -16,13 +16,14 @@ program reduce
 
 ! CHECK-LABEL   omp.declare_reduction @add_reduction_byref_box_2xi32 : !fir.ref<!fir.box<!fir.array<2xi32>>> init {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>):
-! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.array<2xi32> {bindc_name = ".tmp"}
 ! CHECK:           %[[VAL_2:.*]] = arith.constant 0 : i32
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
 ! CHECK:           %[[VAL_4:.*]] = arith.constant 2 : index
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<2xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<2xi32>>, !fir.ref<!fir.array<2xi32>>)
-! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.ref<!fir.array<2xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<2xi32>>
+! CHECK:           %[[VAL_1:.*]] = fir.allocmem !fir.array<2xi32> {bindc_name = ".tmp", uniq_name = ""}
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<2xi32>>, !fir.shape<1>) -> (!fir.heap<!fir.array<2xi32>>, !fir.heap<!fir.array<2xi32>>)
+! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.heap<!fir.array<2xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<2xi32>>
 ! CHECK:           hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<2xi32>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<2xi32>>
 ! CHECK:           fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
@@ -45,6 +46,18 @@ program reduce
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<2xi32>>) -> !fir.ref<!fir.array<2xi32>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<2xi32>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<2xi32>>) -> !fir.heap<!fir.array<2xi32>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<2xi32>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL   func.func @_QQmain() attributes {fir.bindc_name = "reduce"} {
diff --git a/flang/test/Lower/OpenMP/wsloop-reduction-array2.f90 b/flang/test/Lower/OpenMP/wsloop-reduction-array2.f90
index f3745c8460915..045208d6f7ffa 100644
--- a/flang/test/Lower/OpenMP/wsloop-reduction-array2.f90
+++ b/flang/test/Lower/OpenMP/wsloop-reduction-array2.f90
@@ -16,19 +16,20 @@ program reduce
 
 ! CHECK-LABEL   omp.declare_reduction @add_reduction_byref_box_2xi32 : !fir.ref<!fir.box<!fir.array<2xi32>>> init {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>):
-! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.array<2xi32> {bindc_name = ".tmp"}
 ! CHECK:           %[[VAL_2:.*]] = arith.constant 0 : i32
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
 ! CHECK:           %[[VAL_4:.*]] = arith.constant 2 : index
 ! CHECK:           %[[VAL_5:.*]] = fir.shape %[[VAL_4]] : (index) -> !fir.shape<1>
-! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.ref<!fir.array<2xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<2xi32>>, !fir.ref<!fir.array<2xi32>>)
-! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.ref<!fir.array<2xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<2xi32>>
+! CHECK:           %[[VAL_1:.*]] = fir.allocmem !fir.array<2xi32> {bindc_name = ".tmp", uniq_name = ""}
+! CHECK:           %[[TRUE:.*]]  = arith.constant true
+! CHECK:           %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_1]](%[[VAL_5]]) {uniq_name = ".tmp"} : (!fir.heap<!fir.array<2xi32>>, !fir.shape<1>) -> (!fir.heap<!fir.array<2xi32>>, !fir.heap<!fir.array<2xi32>>)
+! CHECK:           %[[VAL_7:.*]] = fir.embox %[[VAL_6]]#0(%[[VAL_5]]) : (!fir.heap<!fir.array<2xi32>>, !fir.shape<1>) -> !fir.box<!fir.array<2xi32>>
 ! CHECK:           hlfir.assign %[[VAL_2]] to %[[VAL_7]] : i32, !fir.box<!fir.array<2xi32>>
 ! CHECK:           %[[VAL_8:.*]] = fir.alloca !fir.box<!fir.array<2xi32>>
 ! CHECK:           fir.store %[[VAL_7]] to %[[VAL_8]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
 ! CHECK:           omp.yield(%[[VAL_8]] : !fir.ref<!fir.box<!fir.array<2xi32>>>)
 
-! CHECK-LABEL   } combiner {
+! CHECK:        } combiner {
 ! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>, %[[VAL_1:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>):
 ! CHECK:           %[[VAL_2:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
 ! CHECK:           %[[VAL_3:.*]] = fir.load %[[VAL_1]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
@@ -45,6 +46,18 @@ program reduce
 ! CHECK:             fir.store %[[VAL_13]] to %[[VAL_9]] : !fir.ref<i32>
 ! CHECK:           }
 ! CHECK:           omp.yield(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>)
+! CHECK:         }  cleanup {
+! CHECK:         ^bb0(%[[VAL_0:.*]]: !fir.ref<!fir.box<!fir.array<2xi32>>>):
+! CHECK:           %[[VAL_1:.*]] = fir.load %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<2xi32>>>
+! CHECK:           %[[VAL_2:.*]] = fir.box_addr %[[VAL_1]] : (!fir.box<!fir.array<2xi32>>) -> !fir.ref<!fir.array<2xi32>>
+! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<2xi32>>) -> i64
+! CHECK:           %[[VAL_4:.*]] = arith.constant 0 : i64
+! CHECK:           %[[VAL_5:.*]] = arith.cmpi ne, %[[VAL_3]], %[[VAL_4]] : i64
+! CHECK:           fir.if %[[VAL_5]] {
+! CHECK:             %[[VAL_6:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.array<2xi32>>) -> !fir.heap<!fir.array<2xi32>>
+! CHECK:             fir.freemem %[[VAL_6]] : !fir.heap<!fir.array<2xi32>>
+! CHECK:           }
+! CHECK:           omp.yield
 ! CHECK:         }
 
 ! CHECK-LABEL:   func.func @_QQmain() attributes {fir.bindc_name = "reduce"} {

From b800a9352330a5b3db91d43f2cc6a0ddeda03aa6 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 10:36:02 +0100
Subject: [PATCH 125/886] Fix MSVC "not all control paths return a value"
 warning. NFC.

---
 llvm/lib/IR/AsmWriter.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp
index 609de920ba7dd..941f6a7a7d823 100644
--- a/llvm/lib/IR/AsmWriter.cpp
+++ b/llvm/lib/IR/AsmWriter.cpp
@@ -3313,7 +3313,7 @@ static const char *getImportTypeName(GlobalValueSummary::ImportKind IK) {
   case GlobalValueSummary::Declaration:
     return "declaration";
   }
-  assert(false && "invalid import kind");
+  llvm_unreachable("invalid import kind");
 }
 
 void AssemblyWriter::printFunctionSummary(const FunctionSummary *FS) {

From 759422c6df2dfe42d01bb64b42f43ab57db6e59e Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 10:42:49 +0100
Subject: [PATCH 126/886] [DAG] visitEXTRACT_SUBVECTOR - pull out repeated
 SDLoc. NFC.

---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 20 +++++++++----------
 1 file changed, 9 insertions(+), 11 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 8fe074666a3dc..95c1cde0b9347 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -24433,6 +24433,7 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
   EVT NVT = N->getValueType(0);
   SDValue V = N->getOperand(0);
   uint64_t ExtIdx = N->getConstantOperandVal(1);
+  SDLoc DL(N);
 
   // Extract from UNDEF is UNDEF.
   if (V.isUndef())
@@ -24448,7 +24449,7 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
     if (TLI.isExtractSubvectorCheap(NVT, V.getOperand(0).getValueType(),
                                     V.getConstantOperandVal(1)) &&
         TLI.isOperationLegalOrCustom(ISD::EXTRACT_SUBVECTOR, NVT)) {
-      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, SDLoc(N), NVT, V.getOperand(0),
+      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NVT, V.getOperand(0),
                          V.getOperand(1));
     }
   }
@@ -24457,7 +24458,7 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
   if (V.getOpcode() == ISD::SPLAT_VECTOR)
     if (DAG.isConstantValueOfAnyType(V.getOperand(0)) || V.hasOneUse())
       if (!LegalOperations || TLI.isOperationLegal(ISD::SPLAT_VECTOR, NVT))
-        return DAG.getSplatVector(NVT, SDLoc(N), V.getOperand(0));
+        return DAG.getSplatVector(NVT, DL, V.getOperand(0));
 
   // Try to move vector bitcast after extract_subv by scaling extraction index:
   // extract_subv (bitcast X), Index --> bitcast (extract_subv X, Index')
@@ -24471,10 +24472,9 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
     if ((SrcNumElts % DestNumElts) == 0) {
       unsigned SrcDestRatio = SrcNumElts / DestNumElts;
       ElementCount NewExtEC = NVT.getVectorElementCount() * SrcDestRatio;
-      EVT NewExtVT = EVT::getVectorVT(*DAG.getContext(), SrcVT.getScalarType(),
-                                      NewExtEC);
+      EVT NewExtVT =
+          EVT::getVectorVT(*DAG.getContext(), SrcVT.getScalarType(), NewExtEC);
       if (TLI.isOperationLegalOrCustom(ISD::EXTRACT_SUBVECTOR, NewExtVT)) {
-        SDLoc DL(N);
         SDValue NewIndex = DAG.getVectorIdxConstant(ExtIdx * SrcDestRatio, DL);
         SDValue NewExtract = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NewExtVT,
                                          V.getOperand(0), NewIndex);
@@ -24488,7 +24488,6 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
             NVT.getVectorElementCount().divideCoefficientBy(DestSrcRatio);
         EVT ScalarVT = SrcVT.getScalarType();
         if ((ExtIdx % DestSrcRatio) == 0) {
-          SDLoc DL(N);
           unsigned IndexValScaled = ExtIdx / DestSrcRatio;
           EVT NewExtVT =
               EVT::getVectorVT(*DAG.getContext(), ScalarVT, NewExtEC);
@@ -24536,7 +24535,6 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
     //   v2i8 extract_subvec v8i8 Y, 6
     if (NVT.isFixedLengthVector() && ConcatSrcVT.isFixedLengthVector() &&
         ConcatSrcNumElts % ExtNumElts == 0) {
-      SDLoc DL(N);
       unsigned NewExtIdx = ExtIdx - ConcatOpIdx * ConcatSrcNumElts;
       assert(NewExtIdx + ExtNumElts <= ConcatSrcNumElts &&
              "Trying to extract from >1 concat operand?");
@@ -24575,13 +24573,13 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
         if (NumElems == 1) {
           SDValue Src = V->getOperand(IdxVal);
           if (EltVT != Src.getValueType())
-            Src = DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, Src);
+            Src = DAG.getNode(ISD::TRUNCATE, DL, EltVT, Src);
           return DAG.getBitcast(NVT, Src);
         }
 
         // Extract the pieces from the original build_vector.
-        SDValue BuildVec = DAG.getBuildVector(ExtractVT, SDLoc(N),
-                                              V->ops().slice(IdxVal, NumElems));
+        SDValue BuildVec =
+            DAG.getBuildVector(ExtractVT, DL, V->ops().slice(IdxVal, NumElems));
         return DAG.getBitcast(NVT, BuildVec);
       }
     }
@@ -24608,7 +24606,7 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
       return DAG.getBitcast(NVT, V.getOperand(1));
     }
     return DAG.getNode(
-        ISD::EXTRACT_SUBVECTOR, SDLoc(N), NVT,
+        ISD::EXTRACT_SUBVECTOR, DL, NVT,
         DAG.getBitcast(N->getOperand(0).getValueType(), V.getOperand(0)),
         N->getOperand(1));
   }

From cca30dfb5935e05837e37cced4407a63393c6642 Mon Sep 17 00:00:00 2001
From: nikitalita <nikitalita@protonmail.com>
Date: Thu, 11 Apr 2024 03:08:38 -0700
Subject: [PATCH 127/886] [DebugInfo][ObjectYAML] Remove duplicate "Flags"
 field from LabelSym (#88194)

There was a duplicate flags field mistakenly left in LabelSym. [LabelSym
only has one flags
field](https://github.com/microsoft/microsoft-pdb/blob/805655a28bd8198004be2ac27e6e0290121a5e89/include/cvinfo.h#L3806)
---
 llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp | 1 -
 1 file changed, 1 deletion(-)

diff --git a/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp b/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp
index 64e1a58aa71a8..e1d2700623ec3 100644
--- a/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp
+++ b/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp
@@ -467,7 +467,6 @@ template <> void SymbolRecordImpl<LabelSym>::map(IO &IO) {
   IO.mapOptional("Offset", Symbol.CodeOffset, 0U);
   IO.mapOptional("Segment", Symbol.Segment, uint16_t(0));
   IO.mapRequired("Flags", Symbol.Flags);
-  IO.mapRequired("Flags", Symbol.Flags);
   IO.mapRequired("DisplayName", Symbol.Name);
 }
 

From b60974dc9e5d98054f5a3a0dac7eab70e38bd416 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 11:12:39 +0100
Subject: [PATCH 128/886] [VectorCombine][X86] Extend bitcast(shuffle(x,y))
 test coverage

As discussed on #87510 the intention is only to fold bitcast(shuffle(x,y)) -> shuffle(bitcast(x),bitcast(y)) if we won't actually increase the number of bitcasts (i.e. x or y is already bitcasted from the correct type).
---
 .../Transforms/VectorCombine/X86/shuffle.ll   | 42 +++++++++++++++++++
 1 file changed, 42 insertions(+)

diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
index d1484fd5ab339..5020d37f86f56 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
@@ -121,6 +121,48 @@ define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) {
   ret <16 x i8> %r
 }
 
+; shuffle of 2 operands removes bitcasts
+
+define <4 x i64> @bitcast_shuf_remove_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @bitcast_shuf_remove_bitcasts(
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    ret <4 x i64> [[R]]
+;
+  %bc0 = bitcast <2 x i64> %a0 to <4 x i32>
+  %bc1 = bitcast <2 x i64> %a1 to <4 x i32>
+  %shuf = shufflevector <4 x i32> %bc0, <4 x i32> %bc1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %r = bitcast <8 x i32> %shuf to <4 x i64>
+  ret <4 x i64> %r
+}
+
+; shuffle of 2 operands must reduce bitcasts
+
+define <8 x i32> @bitcast_shuf_one_bitcast(<4 x i32> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @bitcast_shuf_one_bitcast(
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[A1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[TMP1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x i32> [[R]]
+;
+  %bc0 = bitcast <4 x i32> %a0 to <2 x i64>
+  %shuf = shufflevector <2 x i64> %bc0, <2 x i64> %a1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %r = bitcast <4 x i64> %shuf to <8 x i32>
+  ret <8 x i32> %r
+}
+
+; TODO - Negative test - shuffle of 2 operands must not increase bitcasts
+
+define <8 x i32> @bitcast_shuf_too_many_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: @bitcast_shuf_too_many_bitcasts(
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[A0:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A1:%.*]] to <4 x i32>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x i32> [[R]]
+;
+  %shuf = shufflevector <2 x i64> %a0, <2 x i64> %a1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %r = bitcast <4 x i64> %shuf to <8 x i32>
+  ret <8 x i32> %r
+}
+
 define <2 x i64> @PR35454_1(<2 x i64> %v) {
 ; SSE-LABEL: @PR35454_1(
 ; SSE-NEXT:    [[BC:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32>

From a8b461603b3fab3b229ea6552433cb359c30350c Mon Sep 17 00:00:00 2001
From: Adrian Kuegel <akuegel@google.com>
Date: Thu, 11 Apr 2024 10:25:53 +0000
Subject: [PATCH 129/886] [mlir] Apply ClangTidy BugProne fix

forwarding reference passed to std::move(), which may unexpectedly cause
lvalues to be moved; use std::forward() instead.
---
 mlir/lib/Transforms/Utils/DialectConversion.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index 8671c1008902a..f4e34a03d3d09 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -757,7 +757,7 @@ class UnresolvedMaterializationRewrite : public OperationRewrite {
 /// rewrite type and operation among the given rewrites.
 template <typename RewriteTy, typename R>
 static bool hasRewrite(R &&rewrites, Operation *op) {
-  return any_of(std::move(rewrites), [&](auto &rewrite) {
+  return any_of(std::forward(rewrites), [&](auto &rewrite) {
     auto *rewriteTy = dyn_cast<RewriteTy>(rewrite.get());
     return rewriteTy && rewriteTy->getOperation() == op;
   });

From 962534c4b490239269bb2e11d036596826539046 Mon Sep 17 00:00:00 2001
From: Adrian Kuegel <akuegel@google.com>
Date: Thu, 11 Apr 2024 10:28:10 +0000
Subject: [PATCH 130/886] [mlir] Apply ClangTidy BugProne patch

This time for real.
---
 mlir/lib/Transforms/Utils/DialectConversion.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index f4e34a03d3d09..d85938847c776 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -757,7 +757,7 @@ class UnresolvedMaterializationRewrite : public OperationRewrite {
 /// rewrite type and operation among the given rewrites.
 template <typename RewriteTy, typename R>
 static bool hasRewrite(R &&rewrites, Operation *op) {
-  return any_of(std::forward(rewrites), [&](auto &rewrite) {
+  return any_of(std::forward<R>(rewrites), [&](auto &rewrite) {
     auto *rewriteTy = dyn_cast<RewriteTy>(rewrite.get());
     return rewriteTy && rewriteTy->getOperation() == op;
   });

From 364963a0a3935ced1acb2e959ecd08aef39405ef Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Thu, 11 Apr 2024 06:29:51 -0400
Subject: [PATCH 131/886] [BOLT][NFC] Do not assume text section name in more
 places (#88303)

Fixes a couple more places where ".text" is presumed for the main
code section name.
---
 bolt/lib/Rewrite/RewriteInstance.cpp | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index 0c8ee0d417233..1f778d030f7ac 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -556,7 +556,7 @@ Error RewriteInstance::discoverStorage() {
     if (Error E = SectionNameOrErr.takeError())
       return E;
     StringRef SectionName = SectionNameOrErr.get();
-    if (SectionName == ".text") {
+    if (SectionName == BC->getMainCodeSectionName()) {
       BC->OldTextSectionAddress = Section.getAddress();
       BC->OldTextSectionSize = Section.getSize();
 
@@ -1864,7 +1864,8 @@ Error RewriteInstance::readSpecialSections() {
                   "Use -update-debug-sections to keep it.\n";
   }
 
-  HasTextRelocations = (bool)BC->getUniqueSectionByName(".rela.text");
+  HasTextRelocations = (bool)BC->getUniqueSectionByName(
+      ".rela" + std::string(BC->getMainCodeSectionName()));
   HasSymbolTable = (bool)BC->getUniqueSectionByName(".symtab");
   EHFrameSection = BC->getUniqueSectionByName(".eh_frame");
   BuildIDSection = BC->getUniqueSectionByName(".note.gnu.build-id");
@@ -3441,7 +3442,8 @@ void RewriteInstance::emitAndLink() {
   ErrorOr<BinarySection &> TextSection =
       BC->getUniqueSectionByName(BC->getMainCodeSectionName());
   if (BC->HasRelocations && TextSection)
-    BC->renameSection(*TextSection, getOrgSecPrefix() + ".text");
+    BC->renameSection(*TextSection,
+                      getOrgSecPrefix() + BC->getMainCodeSectionName());
 
   //////////////////////////////////////////////////////////////////////////////
   // Assign addresses to new sections.

From e2d482395992d725663543d297f5ab3cc5918fcc Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Thu, 11 Apr 2024 06:35:28 -0400
Subject: [PATCH 132/886] [BOLT][NFC] Make RepRet X86-specific (#88286)

Bolt's RepRet pass is x86-specific, no need to add it for non-x86
targets.
---
 bolt/lib/Rewrite/BinaryPassManager.cpp | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/bolt/lib/Rewrite/BinaryPassManager.cpp b/bolt/lib/Rewrite/BinaryPassManager.cpp
index 6c26bb7957269..be4888ccfa564 100644
--- a/bolt/lib/Rewrite/BinaryPassManager.cpp
+++ b/bolt/lib/Rewrite/BinaryPassManager.cpp
@@ -377,8 +377,9 @@ Error BinaryFunctionPassManager::runAllPasses(BinaryContext &BC) {
 
   Manager.registerPass(std::make_unique<NormalizeCFG>(PrintNormalized));
 
-  Manager.registerPass(std::make_unique<StripRepRet>(NeverPrint),
-                       opts::StripRepRet);
+  if (BC.isX86())
+    Manager.registerPass(std::make_unique<StripRepRet>(NeverPrint),
+                         opts::StripRepRet);
 
   Manager.registerPass(std::make_unique<IdenticalCodeFolding>(PrintICF),
                        opts::ICF);

From a403ad9336a24c459ee79d2cb7675c4b1f32bfd9 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 11:26:41 +0100
Subject: [PATCH 133/886] [VectorCombine] foldBitcastShuffle - limit
 bitcast(shuffle(x,y)) -> shuffle(bitcast(x),bitcast(y))

Only fold bitcast(shuffle(x,y)) -> shuffle(bitcast(x),bitcast(y)) if we won't actually increase the number of bitcasts (i.e. x or y is already bitcasted from the correct type).
---
 llvm/lib/Transforms/Vectorize/VectorCombine.cpp   | 13 ++++++++++++-
 llvm/test/Transforms/VectorCombine/X86/shuffle.ll |  7 +++----
 2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 633b46e2dc8ba..2f9767538e6cb 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -713,6 +713,18 @@ bool VectorCombine::foldBitcastShuffle(Instruction &I) {
   if (SrcTy->getPrimitiveSizeInBits() % DestEltSize != 0)
     return false;
 
+  bool IsUnary = isa<UndefValue>(V1);
+
+  // For binary shuffles, only fold bitcast(shuffle(X,Y))
+  // if it won't increase the number of bitcasts.
+  if (!IsUnary) {
+    auto *BCTy0 = dyn_cast<FixedVectorType>(peekThroughBitcasts(V0)->getType());
+    auto *BCTy1 = dyn_cast<FixedVectorType>(peekThroughBitcasts(V1)->getType());
+    if (!(BCTy0 && BCTy0->getElementType() == DestTy->getElementType()) &&
+        !(BCTy1 && BCTy1->getElementType() == DestTy->getElementType()))
+      return false;
+  }
+
   SmallVector<int, 16> NewMask;
   if (DestEltSize <= SrcEltSize) {
     // The bitcast is from wide to narrow/equal elements. The shuffle mask can
@@ -736,7 +748,6 @@ bool VectorCombine::foldBitcastShuffle(Instruction &I) {
       FixedVectorType::get(DestTy->getScalarType(), NumSrcElts);
   auto *OldShuffleTy =
       FixedVectorType::get(SrcTy->getScalarType(), Mask.size());
-  bool IsUnary = isa<UndefValue>(V1);
   unsigned NumOps = IsUnary ? 1 : 2;
 
   // The new shuffle must not cost more than the old shuffle.
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
index 5020d37f86f56..3d47f373ab77c 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
@@ -149,13 +149,12 @@ define <8 x i32> @bitcast_shuf_one_bitcast(<4 x i32> %a0, <2 x i64> %a1) {
   ret <8 x i32> %r
 }
 
-; TODO - Negative test - shuffle of 2 operands must not increase bitcasts
+; Negative test - shuffle of 2 operands must not increase bitcasts
 
 define <8 x i32> @bitcast_shuf_too_many_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: @bitcast_shuf_too_many_bitcasts(
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[A0:%.*]] to <4 x i32>
-; CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x i64> [[A1:%.*]] to <4 x i32>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP2]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = bitcast <4 x i64> [[SHUF]] to <8 x i32>
 ; CHECK-NEXT:    ret <8 x i32> [[R]]
 ;
   %shuf = shufflevector <2 x i64> %a0, <2 x i64> %a1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>

From 717d3f3974f43d90c1b8829a4077bbc2a2413c83 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 11:42:57 +0100
Subject: [PATCH 134/886] [VectorCombine] foldShuffleOfCastops - add initial
 shuffle(bitcast(x),bitcast(y)) -> bitcast(shuffle(x,y)) support

Just handle cases where the bitcast src/dst element counts are the same (future patches will add shuffle mask scaling)
---
 llvm/lib/Transforms/Vectorize/VectorCombine.cpp            | 7 +++----
 llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll | 7 +++----
 2 files changed, 6 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 2f9767538e6cb..b74fdf27d213a 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -1459,7 +1459,7 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
     return false;
 
   Instruction::CastOps Opcode = C0->getOpcode();
-  if (Opcode == Instruction::BitCast || C0->getSrcTy() != C1->getSrcTy())
+  if (C0->getSrcTy() != C1->getSrcTy())
     return false;
 
   // Handle shuffle(zext_nneg(x), sext(y)) -> sext(shuffle(x,y)) folds.
@@ -1473,10 +1473,9 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
   auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
   auto *CastDstTy = dyn_cast<FixedVectorType>(C0->getDestTy());
   auto *CastSrcTy = dyn_cast<FixedVectorType>(C0->getSrcTy());
-  if (!ShuffleDstTy || !CastDstTy || !CastSrcTy)
+  if (!ShuffleDstTy || !CastDstTy || !CastSrcTy ||
+      CastDstTy->getElementCount() != CastSrcTy->getElementCount())
     return false;
-  assert(CastDstTy->getElementCount() == CastSrcTy->getElementCount() &&
-         "Unexpected src/dst element counts");
 
   auto *NewShuffleDstTy =
       FixedVectorType::get(CastSrcTy->getScalarType(), Mask.size());
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
index 60c6ff97c58b5..97fceacd82758 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
@@ -165,13 +165,12 @@ define <8 x double> @interleave_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %
   ret <8 x double> %r
 }
 
-; TODO - bitcasts (same element count)
+; bitcasts (same element count)
 
 define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: @concat_bitcast_v4i32_v8f32(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <4 x float>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <4 x float>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[X0]], <4 x float> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[R:%.*]] = bitcast <8 x i32> [[TMP1]] to <8 x float>
 ; CHECK-NEXT:    ret <8 x float> [[R]]
 ;
   %x0 = bitcast <4 x i32> %a0 to <4 x float>

From 9d9bb7b1b6e96dc833133dacf1e2c7d9792e640e Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Thu, 11 Apr 2024 12:57:46 +0200
Subject: [PATCH 135/886] Fix complex abs corner cases. (#88373)

The current implementation fails for very small and very large values.
For example, (0, -inf) should return inf, but it returns -inf.

This ports the logic used in XLA. Tested with XLA's
exhaustive_binary_test_f32_f64.
---
 .../ComplexToStandard/ComplexToStandard.cpp   |  54 +--
 .../convert-to-standard.mlir                  | 348 +++++++-----------
 .../ComplexToStandard/full-conversion.mlir    |  34 +-
 3 files changed, 162 insertions(+), 274 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index a6fcf6a758c07..462036e51a1f1 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -26,7 +26,7 @@ namespace mlir {
 using namespace mlir;
 
 namespace {
-// The algorithm is listed in https://dl.acm.org/doi/pdf/10.1145/363717.363780.
+
 struct AbsOpConversion : public OpConversionPattern<complex::AbsOp> {
   using OpConversionPattern<complex::AbsOp>::OpConversionPattern;
 
@@ -35,49 +35,27 @@ struct AbsOpConversion : public OpConversionPattern<complex::AbsOp> {
                   ConversionPatternRewriter &rewriter) const override {
     mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
-    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
 
     Type elementType = op.getType();
-    Value arg = adaptor.getComplex();
-
-    Value zero =
-        b.create<arith::ConstantOp>(elementType, b.getZeroAttr(elementType));
     Value one = b.create<arith::ConstantOp>(elementType,
                                             b.getFloatAttr(elementType, 1.0));
 
-    Value real = b.create<complex::ReOp>(elementType, arg);
-    Value imag = b.create<complex::ImOp>(elementType, arg);
-
-    Value realIsZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero);
-    Value imagIsZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero);
+    Value real = b.create<complex::ReOp>(adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(adaptor.getComplex());
+    Value absReal = b.create<math::AbsFOp>(real, fmf);
+    Value absImag = b.create<math::AbsFOp>(imag, fmf);
 
-    // Real > Imag
-    Value imagDivReal = b.create<arith::DivFOp>(imag, real, fmf.getValue());
-    Value imagSq =
-        b.create<arith::MulFOp>(imagDivReal, imagDivReal, fmf.getValue());
-    Value imagSqPlusOne = b.create<arith::AddFOp>(imagSq, one, fmf.getValue());
-    Value imagSqrt = b.create<math::SqrtOp>(imagSqPlusOne, fmf.getValue());
-    Value realAbs = b.create<math::AbsFOp>(real, fmf.getValue());
-    Value absImag = b.create<arith::MulFOp>(imagSqrt, realAbs, fmf.getValue());
-
-    // Real <= Imag
-    Value realDivImag = b.create<arith::DivFOp>(real, imag, fmf.getValue());
-    Value realSq =
-        b.create<arith::MulFOp>(realDivImag, realDivImag, fmf.getValue());
-    Value realSqPlusOne = b.create<arith::AddFOp>(realSq, one, fmf.getValue());
-    Value realSqrt = b.create<math::SqrtOp>(realSqPlusOne, fmf.getValue());
-    Value imagAbs = b.create<math::AbsFOp>(imag, fmf.getValue());
-    Value absReal = b.create<arith::MulFOp>(realSqrt, imagAbs, fmf.getValue());
-
-    rewriter.replaceOpWithNewOp<arith::SelectOp>(
-        op, realIsZero, imagAbs,
-        b.create<arith::SelectOp>(
-            imagIsZero, realAbs,
-            b.create<arith::SelectOp>(
-                b.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, real, imag),
-                absImag, absReal)));
+    Value max = b.create<arith::MaximumFOp>(absReal, absImag, fmf);
+    Value min = b.create<arith::MinimumFOp>(absReal, absImag, fmf);
+    Value ratio = b.create<arith::DivFOp>(min, max, fmf);
+    Value ratioSq = b.create<arith::MulFOp>(ratio, ratio, fmf);
+    Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
+    Value sqrt = b.create<math::SqrtOp>(ratioSqPlusOne, fmf);
+    Value result = b.create<arith::MulFOp>(max, sqrt, fmf);
+    Value isNaN =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, result, result, fmf);
+    rewriter.replaceOpWithNewOp<arith::SelectOp>(op, isNaN, min, result);
 
     return success();
   }
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 46dba04a88aa0..a1de61d10bb22 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -8,29 +8,21 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   return %abs : f32
 }
 
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[ABS3:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: return %[[ABS3]] : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: return %[[ABS]] : f32
 
 // -----
 
@@ -258,29 +250,21 @@ func.func @complex_log(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg: complex<f32>
   return %log : complex<f32>
 }
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: %[[RESULT_REAL:.*]] = math.log %[[NORM]] : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] : f32
+// CHECK: %[[RESULT:.*]] = arith.mulf %[[MAX]], %[[SQRT]] : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[RESULT]], %[[RESULT]] : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[RESULT]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = math.log %[[ABS]] : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG2]], %[[REAL2]] : f32
@@ -509,30 +493,22 @@ func.func @complex_sign(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[REAL_IS_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
 // CHECK: %[[IMAG_IS_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
 // CHECK: %[[IS_ZERO:.*]] = arith.andi %[[REAL_IS_ZERO]], %[[IMAG_IS_ZERO]] : i1
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL2]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG2]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG2]], %[[REAL2]] : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL2]] : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL2]], %[[IMAG2]] : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG2]] : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL2]], %[[IMAG2]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: %[[REAL_SIGN:.*]] = arith.divf %[[REAL]], %[[NORM]] : f32
-// CHECK: %[[IMAG_SIGN:.*]] = arith.divf %[[IMAG]], %[[NORM]] : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL2]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG2]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[REAL_SIGN:.*]] = arith.divf %[[REAL]], %[[ABS]] : f32
+// CHECK: %[[IMAG_SIGN:.*]] = arith.divf %[[IMAG]], %[[ABS]] : f32
 // CHECK: %[[SIGN:.*]] = complex.create %[[REAL_SIGN]], %[[IMAG_SIGN]] : complex<f32>
 // CHECK: %[[RESULT:.*]] = arith.select %[[IS_ZERO]], %[[ARG]], %[[SIGN]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
@@ -725,29 +701,21 @@ func.func @complex_sqrt(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[VAR0:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[VAR1:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[VAR2:.*]] = math.absf %[[VAR0]] : f32
-// CHECK: %[[CST0:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK: %[[CST1:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[VAR3:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[VAR4:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[VAR5:.*]] = arith.cmpf oeq, %[[VAR3]], %[[CST0]] : f32
-// CHECK: %[[VAR6:.*]] = arith.cmpf oeq, %[[VAR4]], %[[CST0]] : f32
-// CHECK: %[[VAR7:.*]] = arith.divf %[[VAR4]], %[[VAR3]] : f32
-// CHECK: %[[VAR8:.*]] = arith.mulf %[[VAR7]], %[[VAR7]] : f32
-// CHECK: %[[VAR9:.*]] = arith.addf %[[VAR8]], %[[CST1]] : f32
-// CHECK: %[[VAR10:.*]] = math.sqrt %[[VAR9]] : f32
-// CHECK: %[[VAR11:.*]] = math.absf %[[VAR3]] : f32
-// CHECK: %[[VAR12:.*]] = arith.mulf %[[VAR10]], %[[VAR11]] : f32
-// CHECK: %[[VAR13:.*]] = arith.divf %[[VAR3]], %[[VAR4]] : f32
-// CHECK: %[[VAR14:.*]] = arith.mulf %[[VAR13]], %[[VAR13]] : f32
-// CHECK: %[[VAR15:.*]] = arith.addf %[[VAR14]], %[[CST1]] : f32
-// CHECK: %[[VAR16:.*]] = math.sqrt %[[VAR15]] : f32
-// CHECK: %[[VAR17:.*]] = math.absf %[[VAR4]] : f32
-// CHECK: %[[VAR18:.*]] = arith.mulf %[[VAR16]], %[[VAR17]] : f32
-// CHECK: %[[VAR19:.*]] = arith.cmpf ogt, %[[VAR3]], %[[VAR4]] : f32
-// CHECK: %[[VAR20:.*]] = arith.select %[[VAR19]], %[[VAR12]], %[[VAR18]] : f32
-// CHECK: %[[VAR21:.*]] = arith.select %[[VAR6]], %[[VAR11]], %[[VAR20]] : f32
-// CHECK: %[[VAR22:.*]] = arith.select %[[VAR5]], %[[VAR17]], %[[VAR21]] : f32
-// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[VAR22]] : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[ABS]] : f32
 // CHECK: %[[CST2:.*]]  = arith.constant 5.000000e-01 : f32
 // CHECK: %[[VAR24:.*]] = arith.mulf %[[VAR23]], %[[CST2]] : f32
 // CHECK: %[[VAR25:.*]] = math.sqrt %[[VAR24]] : f32
@@ -821,29 +789,21 @@ func.func @complex_abs_with_fmf(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg fastmath<nnan,contract> : complex<f32>
   return %abs : f32
 }
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[ABS3:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: return %[[ABS3]] : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: return %[[ABS]] : f32
 
 // -----
 
@@ -928,29 +888,21 @@ func.func @complex_log_with_fmf(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg fastmath<nnan,contract> : complex<f32>
   return %log : complex<f32>
 }
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG]], %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL]], %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: %[[RESULT_REAL:.*]] = math.log %[[NORM]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = math.log %[[ABS]] fastmath<nnan,contract> : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[RESULT_IMAG:.*]] = math.atan2 %[[IMAG2]], %[[REAL2]] fastmath<nnan,contract> : f32
@@ -1318,29 +1270,21 @@ func.func @complex_atan2_with_fmf(%lhs: complex<f32>,
 // CHECK: %[[VAR187:.*]] = complex.re %[[VAR186]] : complex<f32>
 // CHECK: %[[VAR188:.*]] = complex.im %[[VAR186]] : complex<f32>
 // CHECK: %[[VAR189:.*]] = math.absf %[[VAR187]] fastmath<nnan,contract> : f32
-// CHECK: %[[CST_7:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK: %[[CST_8:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[VAR190:.*]] = complex.re %[[VAR186]] : complex<f32>
-// CHECK: %[[VAR191:.*]] = complex.im %[[VAR186]] : complex<f32>
-// CHECK: %[[VAR192:.*]] = arith.cmpf oeq, %[[VAR190]], %[[CST_7]] : f32
-// CHECK: %[[VAR193:.*]] = arith.cmpf oeq, %[[VAR191]], %[[CST_7]] : f32
-// CHECK: %[[VAR194:.*]] = arith.divf %[[VAR191]], %[[VAR190]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR195:.*]] = arith.mulf %[[VAR194]], %[[VAR194]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR196:.*]] = arith.addf %[[VAR195]], %[[CST_8]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR197:.*]] = math.sqrt %[[VAR196]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR198:.*]] = math.absf %[[VAR190]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR199:.*]] = arith.mulf %[[VAR197]], %[[VAR198]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR200:.*]] = arith.divf %[[VAR190]], %[[VAR191]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR201:.*]] = arith.mulf %[[VAR200]], %[[VAR200]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR202:.*]] = arith.addf %[[VAR201]], %[[CST_8]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR203:.*]] = math.sqrt %[[VAR202]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR204:.*]] = math.absf %[[VAR191]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR205:.*]] = arith.mulf %[[VAR203]], %[[VAR204]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR206:.*]] = arith.cmpf ogt, %[[VAR190]], %[[VAR191]] : f32
-// CHECK: %[[VAR207:.*]] = arith.select %[[VAR206]], %[[VAR199]], %[[VAR205]] : f32
-// CHECK: %[[VAR208:.*]] = arith.select %[[VAR193]], %[[VAR198]], %[[VAR207]] : f32
-// CHECK: %[[VAR209:.*]] = arith.select %[[VAR192]], %[[VAR204]], %[[VAR208]] : f32
-// CHECK: %[[VAR210:.*]] = arith.addf %[[VAR189]], %[[VAR209]] fastmath<nnan,contract> : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[REAL:.*]] = complex.re %[[VAR186]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[VAR186]] : complex<f32>
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract>  : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract>  : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract>  : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract>  : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[VAR210:.*]] = arith.addf %[[VAR189]], %[[ABS]] fastmath<nnan,contract> : f32
 // CHECK: %[[CST_9:.*]] = arith.constant 5.000000e-01 : f32
 // CHECK: %[[VAR211:.*]] = arith.mulf %[[VAR210]], %[[CST_9]] fastmath<nnan,contract> : f32
 // CHECK: %[[VAR212:.*]] = math.sqrt %[[VAR211]] fastmath<nnan,contract> : f32
@@ -1556,29 +1500,21 @@ func.func @complex_atan2_with_fmf(%lhs: complex<f32>,
 // CHECK: %[[VAR413:.*]] = arith.select %[[VAR412]], %[[VAR408]], %[[VAR402]] : f32
 // CHECK: %[[VAR414:.*]] = arith.select %[[VAR412]], %[[VAR409]], %[[VAR403]] : f32
 // CHECK: %[[VAR415:.*]] = complex.create %[[VAR413]], %[[VAR414]] : complex<f32>
-// CHECK: %[[CST_19:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK: %[[CST_20:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[VAR416:.*]] = complex.re %[[VAR415]] : complex<f32>
-// CHECK: %[[VAR417:.*]] = complex.im %[[VAR415]] : complex<f32>
-// CHECK: %[[VAR418:.*]] = arith.cmpf oeq, %[[VAR416]], %[[CST_19]] : f32
-// CHECK: %[[VAR419:.*]] = arith.cmpf oeq, %[[VAR417]], %[[CST_19]] : f32
-// CHECK: %[[VAR420:.*]] = arith.divf %[[VAR417]], %[[VAR416]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR421:.*]] = arith.mulf %[[VAR420]], %[[VAR420]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR422:.*]] = arith.addf %[[VAR421]], %[[CST_20]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR423:.*]] = math.sqrt %[[VAR422]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR424:.*]] = math.absf %[[VAR416]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR425:.*]] = arith.mulf %[[VAR423]], %[[VAR424]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR426:.*]] = arith.divf %[[VAR416]], %[[VAR417]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR427:.*]] = arith.mulf %[[VAR426]], %[[VAR426]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR428:.*]] = arith.addf %[[VAR427]], %[[CST_20]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR429:.*]] = math.sqrt %[[VAR428]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR430:.*]] = math.absf %[[VAR417]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR431:.*]] = arith.mulf %[[VAR429]], %[[VAR430]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR432:.*]] = arith.cmpf ogt, %[[VAR416]], %[[VAR417]] : f32
-// CHECK: %[[VAR433:.*]] = arith.select %[[VAR432]], %[[VAR425]], %[[VAR431]] : f32
-// CHECK: %[[VAR434:.*]] = arith.select %[[VAR419]], %[[VAR424]], %[[VAR433]] : f32
-// CHECK: %[[VAR435:.*]] = arith.select %[[VAR418]], %[[VAR430]], %[[VAR434]] : f32
-// CHECK: %[[VAR436:.*]] = math.log %[[VAR435]] fastmath<nnan,contract> : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[REAL:.*]] = complex.re %[[VAR415]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[VAR415]] : complex<f32>
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[VAR436:.*]] = math.log %[[ABS]] fastmath<nnan,contract> : f32
 // CHECK: %[[VAR437:.*]] = complex.re %[[VAR415]] : complex<f32>
 // CHECK: %[[VAR438:.*]] = complex.im %[[VAR415]] : complex<f32>
 // CHECK: %[[VAR439:.*]] = math.atan2 %[[VAR438]], %[[VAR437]] fastmath<nnan,contract> : f32
@@ -1805,29 +1741,21 @@ func.func @complex_sqrt_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[VAR0:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[VAR1:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[VAR2:.*]] = math.absf %[[VAR0]] fastmath<nnan,contract> : f32
-// CHECK: %[[CST0:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK: %[[CST1:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[VAR3:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[VAR4:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[VAR5:.*]] = arith.cmpf oeq, %[[VAR3]], %[[CST0]] : f32
-// CHECK: %[[VAR6:.*]] = arith.cmpf oeq, %[[VAR4]], %[[CST0]] : f32
-// CHECK: %[[VAR7:.*]] = arith.divf %[[VAR4]], %[[VAR3]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR8:.*]] = arith.mulf %[[VAR7]], %[[VAR7]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR9:.*]] = arith.addf %[[VAR8]], %[[CST1]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR10:.*]] = math.sqrt %[[VAR9]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR11:.*]] = math.absf %[[VAR3]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR12:.*]] = arith.mulf %[[VAR10]], %[[VAR11]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR13:.*]] = arith.divf %[[VAR3]], %[[VAR4]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR14:.*]] = arith.mulf %[[VAR13]], %[[VAR13]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR15:.*]] = arith.addf %[[VAR14]], %[[CST1]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR16:.*]] = math.sqrt %[[VAR15]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR17:.*]] = math.absf %[[VAR4]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR18:.*]] = arith.mulf %[[VAR16]], %[[VAR17]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR19:.*]] = arith.cmpf ogt, %[[VAR3]], %[[VAR4]] : f32
-// CHECK: %[[VAR20:.*]] = arith.select %[[VAR19]], %[[VAR12]], %[[VAR18]] : f32
-// CHECK: %[[VAR21:.*]] = arith.select %[[VAR6]], %[[VAR11]], %[[VAR20]] : f32
-// CHECK: %[[VAR22:.*]] = arith.select %[[VAR5]], %[[VAR17]], %[[VAR21]] : f32
-// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[VAR22]] fastmath<nnan,contract> : f32
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[ABS]] fastmath<nnan,contract> : f32
 // CHECK: %[[CST2:.*]]  = arith.constant 5.000000e-01 : f32
 // CHECK: %[[VAR24:.*]] = arith.mulf %[[VAR23]], %[[CST2]] fastmath<nnan,contract> : f32
 // CHECK: %[[VAR25:.*]] = math.sqrt %[[VAR24]] fastmath<nnan,contract> : f32
@@ -1910,30 +1838,22 @@ func.func @complex_sign_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[REAL_IS_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
 // CHECK: %[[IMAG_IS_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
 // CHECK: %[[IS_ZERO:.*]] = arith.andi %[[REAL_IS_ZERO]], %[[IMAG_IS_ZERO]] : i1
-// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[IS_REAL_ZERO:.*]] = arith.cmpf oeq, %[[REAL2]], %[[ZERO]] : f32
-// CHECK: %[[IS_IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG2]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_DIV_REAL:.*]] = arith.divf %[[IMAG2]], %[[REAL2]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ:.*]] = arith.mulf %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = arith.addf %[[IMAG_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SQRT:.*]] = math.sqrt %[[IMAG_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_ABS:.*]] = math.absf %[[REAL2]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_IMAG:.*]] = arith.mulf %[[IMAG_SQRT]], %[[REAL_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_DIV_IMAG:.*]] = arith.divf %[[REAL2]], %[[IMAG2]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ:.*]] = arith.mulf %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = arith.addf %[[REAL_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_SQRT:.*]] = math.sqrt %[[REAL_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_ABS:.*]] = math.absf %[[IMAG2]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_REAL:.*]] = arith.mulf %[[REAL_SQRT]], %[[IMAG_ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[REAL_GT_IMAG:.*]] = arith.cmpf ogt, %[[REAL2]], %[[IMAG2]] : f32
-// CHECK: %[[ABS1:.*]] = arith.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : f32
-// CHECK: %[[ABS2:.*]] = arith.select %[[IS_IMAG_ZERO]], %[[REAL_ABS]], %[[ABS1]] : f32
-// CHECK: %[[NORM:.*]] = arith.select %[[IS_REAL_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : f32
-// CHECK: %[[REAL_SIGN:.*]] = arith.divf %[[REAL]], %[[NORM]] fastmath<nnan,contract> : f32
-// CHECK: %[[IMAG_SIGN:.*]] = arith.divf %[[IMAG]], %[[NORM]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL2]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG2]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
+// CHECK: %[[REAL_SIGN:.*]] = arith.divf %[[REAL]], %[[ABS]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_SIGN:.*]] = arith.divf %[[IMAG]], %[[ABS]] fastmath<nnan,contract> : f32
 // CHECK: %[[SIGN:.*]] = complex.create %[[REAL_SIGN]], %[[IMAG_SIGN]] : complex<f32>
 // CHECK: %[[RESULT:.*]] = arith.select %[[IS_ZERO]], %[[ARG]], %[[SIGN]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
diff --git a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
index 0f23e20167f49..2649d004a76ac 100644
--- a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
@@ -6,32 +6,22 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg: complex<f32>
   return %abs : f32
 }
-// CHECK: %[[ZERO:.*]] = llvm.mlir.constant(0.000000e+00 : f32) : f32
 // CHECK: %[[ONE:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
 // CHECK: %[[REAL:.*]] = llvm.extractvalue %[[ARG]][0] : ![[C_TY]]
 // CHECK: %[[IMAG:.*]] = llvm.extractvalue %[[ARG]][1] : ![[C_TY]]
-// CHECK: %[[REAL_IS_ZERO:.*]] = llvm.fcmp "oeq" %[[REAL]], %[[ZERO]] : f32
-// CHECK: %[[IMAG_IS_ZERO:.*]] = llvm.fcmp "oeq" %[[IMAG]], %[[ZERO]] : f32
 
-// CHECK: %[[IMAG_DIV_REAL:.*]] = llvm.fdiv %[[IMAG]], %[[REAL]] : f32
-// CHECK: %[[IMAG_SQ:.*]] = llvm.fmul %[[IMAG_DIV_REAL]], %[[IMAG_DIV_REAL]]  : f32
-// CHECK: %[[IMAG_SQ_PLUS_ONE:.*]] = llvm.fadd %[[IMAG_SQ]], %[[ONE]] : f32
-// CHECK: %[[IMAG_SQRT:.*]] = llvm.intr.sqrt(%[[IMAG_SQ_PLUS_ONE]]) : (f32) -> f32
-// CHECK: %[[REAL_ABS:.*]] = llvm.intr.fabs(%[[REAL]]) : (f32) -> f32
-// CHECK: %[[ABS_IMAG:.*]] = llvm.fmul %[[IMAG_SQRT]], %[[REAL_ABS]] : f32
-
-// CHECK: %[[REAL_DIV_IMAG:.*]] = llvm.fdiv %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[REAL_SQ:.*]] = llvm.fmul %[[REAL_DIV_IMAG]], %[[REAL_DIV_IMAG]] : f32
-// CHECK: %[[REAL_SQ_PLUS_ONE:.*]] = llvm.fadd %[[REAL_SQ]], %[[ONE]]  : f32
-// CHECK: %[[REAL_SQRT:.*]] = llvm.intr.sqrt(%[[REAL_SQ_PLUS_ONE]])  : (f32) -> f32
-// CHECK: %[[IMAG_ABS:.*]] = llvm.intr.fabs(%[[IMAG]]) : (f32) -> f32
-// CHECK: %[[ABS_REAL:.*]] = llvm.fmul %[[REAL_SQRT]], %[[IMAG_ABS]]  : f32
-
-// CHECK: %[[REAL_GT_IMAG:.*]] = llvm.fcmp "ogt" %[[REAL]], %[[IMAG]] : f32
-// CHECK: %[[ABS1:.*]] = llvm.select %[[REAL_GT_IMAG]], %[[ABS_IMAG]], %[[ABS_REAL]] : i1, f32
-// CHECK: %[[ABS2:.*]] = llvm.select %[[IMAG_IS_ZERO]], %[[REAL_ABS]], %[[ABS1]] : i1, f32
-// CHECK: %[[NORM:.*]] = llvm.select %[[REAL_IS_ZERO]], %[[IMAG_ABS]], %[[ABS2]] : i1, f32
-// CHECK: llvm.return %[[NORM]] : f32
+// CHECK: %[[ABS_REAL:.*]] = llvm.intr.fabs(%[[REAL]]) : (f32) -> f32
+// CHECK: %[[ABS_IMAG:.*]] = llvm.intr.fabs(%[[IMAG]]) : (f32) -> f32
+// CHECK: %[[MAX:.*]] = llvm.intr.maximum(%[[ABS_REAL]], %[[ABS_IMAG]]) : (f32, f32) -> f32
+// CHECK: %[[MIN:.*]] = llvm.intr.minimum(%[[ABS_REAL]], %[[ABS_IMAG]]) : (f32, f32) -> f32
+// CHECK: %[[RATIO:.*]] = llvm.fdiv %[[MIN]], %[[MAX]] : f32
+// CHECK: %[[RATIO_SQ:.*]] = llvm.fmul %[[RATIO]], %[[RATIO]] : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = llvm.fadd %[[RATIO_SQ]], %[[ONE]] : f32
+// CHECK: %[[SQRT:.*]] = llvm.intr.sqrt(%[[RATIO_SQ_PLUS_ONE]]) : (f32) -> f32
+// CHECK: %[[RESULT:.*]] = llvm.fmul %[[MAX]], %[[SQRT]] : f32
+// CHECK: %[[IS_NAN:.*]] = llvm.fcmp "uno" %[[RESULT]], %11 : f32
+// CHECK: %[[RET:.*]] = llvm.select %[[IS_NAN]], %[[MIN]], %[[RESULT]] : i1, f32
+// CHECK: llvm.return %[[RET]] : f32
 
 // CHECK-LABEL: llvm.func @complex_eq
 // CHECK-SAME: %[[LHS:.*]]: ![[C_TY:.*]], %[[RHS:.*]]: ![[C_TY:.*]])

From b1094776152b68efa05f69b7b833f9cbc0727efc Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Thu, 11 Apr 2024 19:10:53 +0800
Subject: [PATCH 136/886] [InstCombine] Infer nsw/nuw for trunc (#87910)

This patch adds support for inferring trunc's nsw/nuw flags.
---
 clang/test/CodeGen/ms-intrinsics-other.c      |  4 +-
 clang/test/CodeGen/ms-intrinsics.c            |  4 +-
 clang/test/CodeGenOpenCL/builtins-amdgcn.cl   |  2 +-
 clang/test/Headers/__clang_hip_math.hip       |  2 +-
 .../InstCombine/InstCombineCasts.cpp          | 15 +++-
 .../RISCV/riscv-vsetvli-knownbits.ll          |  4 +-
 .../RISCV/riscv-vsetvlimax-knownbits.ll       |  4 +-
 llvm/test/Transforms/InstCombine/add.ll       |  2 +-
 .../Transforms/InstCombine/binop-itofp.ll     | 12 ++--
 .../test/Transforms/InstCombine/bswap-fold.ll |  8 +--
 llvm/test/Transforms/InstCombine/bswap.ll     |  4 +-
 llvm/test/Transforms/InstCombine/cast.ll      | 30 ++++----
 .../Transforms/InstCombine/cmp-intrinsic.ll   |  4 +-
 .../Transforms/InstCombine/compare-signs.ll   |  4 +-
 llvm/test/Transforms/InstCombine/ctpop.ll     |  2 +-
 .../extractelement-inseltpoison.ll            | 10 +--
 .../Transforms/InstCombine/extractelement.ll  | 28 ++++----
 llvm/test/Transforms/InstCombine/ffs-1.ll     |  2 +-
 llvm/test/Transforms/InstCombine/fls.ll       |  2 +-
 .../InstCombine/fold-log2-ceil-idiom.ll       |  2 +-
 .../high-bit-signmask-with-trunc.ll           | 20 +++---
 .../Transforms/InstCombine/icmp-mul-zext.ll   |  2 +-
 .../InstCombine/icmp-of-trunc-ext.ll          | 20 +++---
 .../InstCombine/icmp-topbitssame.ll           |  8 +--
 .../Transforms/InstCombine/insert-trunc.ll    | 14 ++--
 .../Transforms/InstCombine/insertelt-trunc.ll | 34 ++++-----
 .../test/Transforms/InstCombine/known-bits.ll |  2 +-
 .../Transforms/InstCombine/known-non-zero.ll  |  4 +-
 .../InstCombine/known-phi-recurse.ll          | 18 ++---
 .../logical-select-inseltpoison.ll            |  2 +-
 .../Transforms/InstCombine/logical-select.ll  |  2 +-
 .../lshr-trunc-sext-to-ashr-sext.ll           | 12 ++--
 llvm/test/Transforms/InstCombine/lshr.ll      | 22 +++---
 .../merging-multiple-stores-into-successor.ll |  2 +-
 llvm/test/Transforms/InstCombine/narrow.ll    |  2 +-
 .../Transforms/InstCombine/negated-bitmask.ll |  4 +-
 llvm/test/Transforms/InstCombine/pr34349.ll   |  2 +-
 .../InstCombine/reduction-add-sext-zext-i1.ll |  4 +-
 llvm/test/Transforms/InstCombine/sadd_sat.ll  | 38 +++++-----
 .../InstCombine/select-cmp-cttz-ctlz.ll       | 28 ++++----
 .../InstCombine/select-imm-canon.ll           |  6 +-
 llvm/test/Transforms/InstCombine/select.ll    |  2 +-
 .../InstCombine/sext-of-trunc-nsw.ll          |  6 +-
 llvm/test/Transforms/InstCombine/sext.ll      |  2 +-
 llvm/test/Transforms/InstCombine/shift-add.ll |  2 +-
 ...ciation-in-bittest-with-truncation-lshr.ll |  4 +-
 ...ount-reassociation-with-truncation-ashr.ll |  8 +--
 ...ount-reassociation-with-truncation-lshr.ll |  8 +--
 .../Transforms/InstCombine/shift-shift.ll     |  6 +-
 llvm/test/Transforms/InstCombine/shift.ll     |  6 +-
 .../test/Transforms/InstCombine/shl-demand.ll |  2 +-
 ...-test-via-right-shifting-all-other-bits.ll |  4 +-
 .../Transforms/InstCombine/trunc-demand.ll    |  8 +--
 .../InstCombine/trunc-inseltpoison.ll         |  6 +-
 .../InstCombine/trunc-shift-trunc.ll          |  4 +-
 llvm/test/Transforms/InstCombine/trunc.ll     |  6 +-
 .../InstCombine/truncating-saturate.ll        | 20 +++---
 .../Transforms/InstCombine/vector-trunc.ll    |  4 +-
 llvm/test/Transforms/InstCombine/xor-ashr.ll  |  4 +-
 .../zext-ctlz-trunc-to-ctlz-add.ll            |  4 +-
 .../AArch64/deterministic-type-shrinkage.ll   | 16 ++---
 .../LoopVectorize/AArch64/intrinsiccost.ll    |  6 +-
 .../LoopVectorize/X86/intrinsiccost.ll        |  8 +--
 .../Transforms/LoopVectorize/reduction.ll     |  4 +-
 .../PhaseOrdering/AArch64/quant_4x4.ll        | 72 +++++++++----------
 .../Transforms/SLPVectorizer/X86/pr46983.ll   |  2 +-
 66 files changed, 309 insertions(+), 296 deletions(-)

diff --git a/clang/test/CodeGen/ms-intrinsics-other.c b/clang/test/CodeGen/ms-intrinsics-other.c
index 36c40dddcbb4f..0e9dfe34b84cc 100644
--- a/clang/test/CodeGen/ms-intrinsics-other.c
+++ b/clang/test/CodeGen/ms-intrinsics-other.c
@@ -87,7 +87,7 @@ unsigned char test_BitScanForward64(unsigned LONG *Index, unsigned __int64 Mask)
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:
 // CHECK:   [[INDEX:%[0-9]+]] = tail call i64 @llvm.cttz.i64(i64 %Mask, i1 true)
-// CHECK:   [[TRUNC_INDEX:%[0-9]+]] = trunc i64 [[INDEX]] to i32
+// CHECK:   [[TRUNC_INDEX:%[0-9]+]] = trunc nuw nsw i64 [[INDEX]] to i32
 // CHECK:   store i32 [[TRUNC_INDEX]], ptr %Index, align 4
 // CHECK:   br label %[[END_LABEL]]
 
@@ -102,7 +102,7 @@ unsigned char test_BitScanReverse64(unsigned LONG *Index, unsigned __int64 Mask)
 // CHECK:   ret i8 [[RESULT]]
 // CHECK:   [[ISNOTZERO_LABEL]]:
 // CHECK:   [[REVINDEX:%[0-9]+]] = tail call i64 @llvm.ctlz.i64(i64 %Mask, i1 true)
-// CHECK:   [[TRUNC_REVINDEX:%[0-9]+]] = trunc i64 [[REVINDEX]] to i32
+// CHECK:   [[TRUNC_REVINDEX:%[0-9]+]] = trunc nuw nsw i64 [[REVINDEX]] to i32
 // CHECK:   [[INDEX:%[0-9]+]] = xor i32 [[TRUNC_REVINDEX]], 63
 // CHECK:   store i32 [[INDEX]], ptr %Index, align 4
 // CHECK:   br label %[[END_LABEL]]
diff --git a/clang/test/CodeGen/ms-intrinsics.c b/clang/test/CodeGen/ms-intrinsics.c
index 5bb003d1f91fc..6eabd725e2f7c 100644
--- a/clang/test/CodeGen/ms-intrinsics.c
+++ b/clang/test/CodeGen/ms-intrinsics.c
@@ -189,7 +189,7 @@ unsigned char test_BitScanForward64(unsigned long *Index, unsigned __int64 Mask)
 // CHECK-ARM-X64:   ret i8 [[RESULT]]
 // CHECK-ARM-X64:   [[ISNOTZERO_LABEL]]:
 // CHECK-ARM-X64:   [[INDEX:%[0-9]+]] = tail call i64 @llvm.cttz.i64(i64 %Mask, i1 true)
-// CHECK-ARM-X64:   [[TRUNC_INDEX:%[0-9]+]] = trunc i64 [[INDEX]] to i32
+// CHECK-ARM-X64:   [[TRUNC_INDEX:%[0-9]+]] = trunc nuw nsw i64 [[INDEX]] to i32
 // CHECK-ARM-X64:   store i32 [[TRUNC_INDEX]], ptr %Index, align 4
 // CHECK-ARM-X64:   br label %[[END_LABEL]]
 
@@ -204,7 +204,7 @@ unsigned char test_BitScanReverse64(unsigned long *Index, unsigned __int64 Mask)
 // CHECK-ARM-X64:   ret i8 [[RESULT]]
 // CHECK-ARM-X64:   [[ISNOTZERO_LABEL]]:
 // CHECK-ARM-X64:   [[REVINDEX:%[0-9]+]] = tail call i64 @llvm.ctlz.i64(i64 %Mask, i1 true)
-// CHECK-ARM-X64:   [[TRUNC_REVINDEX:%[0-9]+]] = trunc i64 [[REVINDEX]] to i32
+// CHECK-ARM-X64:   [[TRUNC_REVINDEX:%[0-9]+]] = trunc nuw nsw i64 [[REVINDEX]] to i32
 // CHECK-ARM-X64:   [[INDEX:%[0-9]+]] = xor i32 [[TRUNC_REVINDEX]], 63
 // CHECK-ARM-X64:   store i32 [[INDEX]], ptr %Index, align 4
 // CHECK-ARM-X64:   br label %[[END_LABEL]]
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
index 8a4533633706b..bdca97c887867 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
@@ -528,7 +528,7 @@ void test_read_exec_lo(global uint* out) {
 // CHECK-LABEL: @test_read_exec_hi(
 // CHECK: call i64 @llvm.amdgcn.ballot.i64(i1 true)
 // CHECK: lshr i64 [[A:%.*]], 32
-// CHECK: trunc i64 [[B:%.*]] to i32
+// CHECK: trunc nuw i64 [[B:%.*]] to i32
 void test_read_exec_hi(global uint* out) {
   *out = __builtin_amdgcn_read_exec_hi();
 }
diff --git a/clang/test/Headers/__clang_hip_math.hip b/clang/test/Headers/__clang_hip_math.hip
index 37099de74fb8e..2e5f521a5feae 100644
--- a/clang/test/Headers/__clang_hip_math.hip
+++ b/clang/test/Headers/__clang_hip_math.hip
@@ -3703,7 +3703,7 @@ extern "C" __device__ BOOL_TYPE test___signbitf(float x) {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast double [[X:%.*]] to i64
 // CHECK-NEXT:    [[DOTLOBIT:%.*]] = lshr i64 [[TMP0]], 63
-// CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[DOTLOBIT]] to i32
+// CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[DOTLOBIT]] to i32
 // CHECK-NEXT:    ret i32 [[CONV]]
 //
 extern "C" __device__ BOOL_TYPE test___signbit(double x) {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 0652a8ba80b3f..437e9b92c7032 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -897,7 +897,20 @@ Instruction *InstCombinerImpl::visitTrunc(TruncInst &Trunc) {
     }
   }
 
-  return nullptr;
+  bool Changed = false;
+  if (!Trunc.hasNoSignedWrap() &&
+      ComputeMaxSignificantBits(Src, /*Depth=*/0, &Trunc) <= DestWidth) {
+    Trunc.setHasNoSignedWrap(true);
+    Changed = true;
+  }
+  if (!Trunc.hasNoUnsignedWrap() &&
+      MaskedValueIsZero(Src, APInt::getBitsSetFrom(SrcWidth, DestWidth),
+                        /*Depth=*/0, &Trunc)) {
+    Trunc.setHasNoUnsignedWrap(true);
+    Changed = true;
+  }
+
+  return Changed ? &Trunc : nullptr;
 }
 
 Instruction *InstCombinerImpl::transformZExtICmp(ICmpInst *Cmp,
diff --git a/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvli-knownbits.ll b/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvli-knownbits.ll
index 1afae6565fe26..6e0acfd685116 100644
--- a/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvli-knownbits.ll
+++ b/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvli-knownbits.ll
@@ -45,7 +45,7 @@ entry:
 define signext i32 @vsetvl_sext() nounwind #0 {
 ; CHECK-LABEL: @vsetvl_sext(
 ; CHECK-NEXT:    [[A:%.*]] = call i64 @llvm.riscv.vsetvli.i64(i64 1, i64 1, i64 1)
-; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A]] to i32
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw nsw i64 [[A]] to i32
 ; CHECK-NEXT:    ret i32 [[B]]
 ;
   %a = call i64 @llvm.riscv.vsetvli(i64 1, i64 1, i64 1)
@@ -56,7 +56,7 @@ define signext i32 @vsetvl_sext() nounwind #0 {
 define zeroext i32 @vsetvl_zext() nounwind #0 {
 ; CHECK-LABEL: @vsetvl_zext(
 ; CHECK-NEXT:    [[A:%.*]] = call i64 @llvm.riscv.vsetvli.i64(i64 1, i64 1, i64 1)
-; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A]] to i32
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw nsw i64 [[A]] to i32
 ; CHECK-NEXT:    ret i32 [[B]]
 ;
   %a = call i64 @llvm.riscv.vsetvli(i64 1, i64 1, i64 1)
diff --git a/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvlimax-knownbits.ll b/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvlimax-knownbits.ll
index 093ba75e87b5a..811a29c7e5624 100644
--- a/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvlimax-knownbits.ll
+++ b/llvm/test/Transforms/InstCombine/RISCV/riscv-vsetvlimax-knownbits.ll
@@ -45,7 +45,7 @@ entry:
 define signext i32 @vsetvlmax_sext() nounwind #0 {
 ; CHECK-LABEL: @vsetvlmax_sext(
 ; CHECK-NEXT:    [[A:%.*]] = call i64 @llvm.riscv.vsetvlimax.i64(i64 1, i64 1)
-; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A]] to i32
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw nsw i64 [[A]] to i32
 ; CHECK-NEXT:    ret i32 [[B]]
 ;
   %a = call i64 @llvm.riscv.vsetvlimax(i64 1, i64 1)
@@ -56,7 +56,7 @@ define signext i32 @vsetvlmax_sext() nounwind #0 {
 define zeroext i32 @vsetvlmax_zext() nounwind #0 {
 ; CHECK-LABEL: @vsetvlmax_zext(
 ; CHECK-NEXT:    [[A:%.*]] = call i64 @llvm.riscv.vsetvlimax.i64(i64 1, i64 1)
-; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A]] to i32
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw nsw i64 [[A]] to i32
 ; CHECK-NEXT:    ret i32 [[B]]
 ;
   %a = call i64 @llvm.riscv.vsetvlimax(i64 1, i64 1)
diff --git a/llvm/test/Transforms/InstCombine/add.ll b/llvm/test/Transforms/InstCombine/add.ll
index ec3aca26514ca..23eee8547597e 100644
--- a/llvm/test/Transforms/InstCombine/add.ll
+++ b/llvm/test/Transforms/InstCombine/add.ll
@@ -2375,7 +2375,7 @@ define { i64, i64 } @PR57576(i64 noundef %x, i64 noundef %y, i64 noundef %z, i64
 ; CHECK-NEXT:    [[SUB:%.*]] = sub i128 [[XY]], [[ZZ]]
 ; CHECK-NEXT:    [[T:%.*]] = trunc i128 [[SUB]] to i64
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i128 [[SUB]], 64
-; CHECK-NEXT:    [[DOTTR:%.*]] = trunc i128 [[TMP1]] to i64
+; CHECK-NEXT:    [[DOTTR:%.*]] = trunc nuw i128 [[TMP1]] to i64
 ; CHECK-NEXT:    [[DOTNARROW:%.*]] = sub i64 [[DOTTR]], [[W:%.*]]
 ; CHECK-NEXT:    [[R1:%.*]] = insertvalue { i64, i64 } poison, i64 [[T]], 0
 ; CHECK-NEXT:    [[R2:%.*]] = insertvalue { i64, i64 } [[R1]], i64 [[DOTNARROW]], 1
diff --git a/llvm/test/Transforms/InstCombine/binop-itofp.ll b/llvm/test/Transforms/InstCombine/binop-itofp.ll
index cd9ec1e59203f..d72a54e8babc9 100644
--- a/llvm/test/Transforms/InstCombine/binop-itofp.ll
+++ b/llvm/test/Transforms/InstCombine/binop-itofp.ll
@@ -1010,7 +1010,7 @@ define float @test_ui_add_with_signed_constant(i32 %shr.i) {
 define float @missed_nonzero_check_on_constant_for_si_fmul(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @missed_nonzero_check_on_constant_for_si_fmul(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp i16 [[CONV_I]] to float
 ; CHECK-NEXT:    [[MUL3_I_I:%.*]] = call float @llvm.copysign.f32(float 0.000000e+00, float [[CONV1_I]])
 ; CHECK-NEXT:    store i32 [[SEL]], ptr [[G_2345:%.*]], align 4
@@ -1027,7 +1027,7 @@ define float @missed_nonzero_check_on_constant_for_si_fmul(i1 %c, i1 %.b, ptr %g
 define <2 x float> @missed_nonzero_check_on_constant_for_si_fmul_vec(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @missed_nonzero_check_on_constant_for_si_fmul_vec(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
 ; CHECK-NEXT:    [[CONV_I:%.*]] = shufflevector <2 x i16> [[CONV_I_V]], <2 x i16> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp <2 x i16> [[CONV_I]] to <2 x float>
@@ -1048,7 +1048,7 @@ define <2 x float> @missed_nonzero_check_on_constant_for_si_fmul_vec(i1 %c, i1 %
 define float @negzero_check_on_constant_for_si_fmul(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @negzero_check_on_constant_for_si_fmul(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp i16 [[CONV_I]] to float
 ; CHECK-NEXT:    [[TMP1:%.*]] = fneg float [[CONV1_I]]
 ; CHECK-NEXT:    [[MUL3_I_I:%.*]] = call float @llvm.copysign.f32(float 0.000000e+00, float [[TMP1]])
@@ -1066,7 +1066,7 @@ define float @negzero_check_on_constant_for_si_fmul(i1 %c, i1 %.b, ptr %g_2345)
 define <2 x float> @nonzero_check_on_constant_for_si_fmul_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_vec_w_undef(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
 ; CHECK-NEXT:    [[CONV_I:%.*]] = shufflevector <2 x i16> [[CONV_I_V]], <2 x i16> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp <2 x i16> [[CONV_I]] to <2 x float>
@@ -1087,7 +1087,7 @@ define <2 x float> @nonzero_check_on_constant_for_si_fmul_vec_w_undef(i1 %c, i1
 define <2 x float> @nonzero_check_on_constant_for_si_fmul_nz_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_nz_vec_w_undef(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
 ; CHECK-NEXT:    [[CONV_I:%.*]] = shufflevector <2 x i16> [[CONV_I_V]], <2 x i16> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp <2 x i16> [[CONV_I]] to <2 x float>
@@ -1108,7 +1108,7 @@ define <2 x float> @nonzero_check_on_constant_for_si_fmul_nz_vec_w_undef(i1 %c,
 define <2 x float> @nonzero_check_on_constant_for_si_fmul_negz_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
 ; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_negz_vec_w_undef(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
-; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc i32 [[SEL]] to i16
+; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
 ; CHECK-NEXT:    [[CONV_I:%.*]] = shufflevector <2 x i16> [[CONV_I_V]], <2 x i16> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp <2 x i16> [[CONV_I]] to <2 x float>
diff --git a/llvm/test/Transforms/InstCombine/bswap-fold.ll b/llvm/test/Transforms/InstCombine/bswap-fold.ll
index 05933d37057cc..19522168beaf5 100644
--- a/llvm/test/Transforms/InstCombine/bswap-fold.ll
+++ b/llvm/test/Transforms/InstCombine/bswap-fold.ll
@@ -211,7 +211,7 @@ define i64 @variable_shl_not_masked_enough_i64(i64 %x, i64 %n) {
 define i16 @test7(i32 %A) {
 ; CHECK-LABEL: @test7(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[A:%.*]], 16
-; CHECK-NEXT:    [[D:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[D]]
 ;
   %B = tail call i32 @llvm.bswap.i32(i32 %A) nounwind
@@ -223,7 +223,7 @@ define i16 @test7(i32 %A) {
 define <2 x i16> @test7_vector(<2 x i32> %A) {
 ; CHECK-LABEL: @test7_vector(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <2 x i32> [[A:%.*]], <i32 16, i32 16>
-; CHECK-NEXT:    [[D:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i16>
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw <2 x i32> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[D]]
 ;
   %B = tail call <2 x i32> @llvm.bswap.v2i32(<2 x i32> %A) nounwind
@@ -235,7 +235,7 @@ define <2 x i16> @test7_vector(<2 x i32> %A) {
 define i16 @test8(i64 %A) {
 ; CHECK-LABEL: @test8(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[A:%.*]], 48
-; CHECK-NEXT:    [[D:%.*]] = trunc i64 [[TMP1]] to i16
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw i64 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[D]]
 ;
   %B = tail call i64 @llvm.bswap.i64(i64 %A) nounwind
@@ -247,7 +247,7 @@ define i16 @test8(i64 %A) {
 define <2 x i16> @test8_vector(<2 x i64> %A) {
 ; CHECK-LABEL: @test8_vector(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <2 x i64> [[A:%.*]], <i64 48, i64 48>
-; CHECK-NEXT:    [[D:%.*]] = trunc <2 x i64> [[TMP1]] to <2 x i16>
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw <2 x i64> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[D]]
 ;
   %B = tail call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %A) nounwind
diff --git a/llvm/test/Transforms/InstCombine/bswap.ll b/llvm/test/Transforms/InstCombine/bswap.ll
index 21eb170b8c58d..d42583bb5699b 100644
--- a/llvm/test/Transforms/InstCombine/bswap.ll
+++ b/llvm/test/Transforms/InstCombine/bswap.ll
@@ -43,7 +43,7 @@ define i16 @test1_trunc(i32 %i) {
 ; CHECK-NEXT:    [[T3:%.*]] = lshr i32 [[I]], 8
 ; CHECK-NEXT:    [[T4:%.*]] = and i32 [[T3]], 65280
 ; CHECK-NEXT:    [[T5:%.*]] = or disjoint i32 [[T1]], [[T4]]
-; CHECK-NEXT:    [[T13:%.*]] = trunc i32 [[T5]] to i16
+; CHECK-NEXT:    [[T13:%.*]] = trunc nuw i32 [[T5]] to i16
 ; CHECK-NEXT:    ret i16 [[T13]]
 ;
   %t1 = lshr i32 %i, 24
@@ -61,7 +61,7 @@ define i16 @test1_trunc_extra_use(i32 %i) {
 ; CHECK-NEXT:    [[T4:%.*]] = and i32 [[T3]], 65280
 ; CHECK-NEXT:    [[T5:%.*]] = or disjoint i32 [[T1]], [[T4]]
 ; CHECK-NEXT:    call void @extra_use(i32 [[T5]])
-; CHECK-NEXT:    [[T13:%.*]] = trunc i32 [[T5]] to i16
+; CHECK-NEXT:    [[T13:%.*]] = trunc nuw i32 [[T5]] to i16
 ; CHECK-NEXT:    ret i16 [[T13]]
 ;
   %t1 = lshr i32 %i, 24
diff --git a/llvm/test/Transforms/InstCombine/cast.ll b/llvm/test/Transforms/InstCombine/cast.ll
index 97554e9462043..d9c93ba277295 100644
--- a/llvm/test/Transforms/InstCombine/cast.ll
+++ b/llvm/test/Transforms/InstCombine/cast.ll
@@ -1471,7 +1471,7 @@ define i64 @test91(i64 %A) {
 ; ALL-LABEL: @test91(
 ; ALL-NEXT:    [[B:%.*]] = sext i64 [[A:%.*]] to i96
 ; ALL-NEXT:    [[C:%.*]] = lshr i96 [[B]], 48
-; ALL-NEXT:    [[D:%.*]] = trunc i96 [[C]] to i64
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw i96 [[C]] to i64
 ; ALL-NEXT:    ret i64 [[D]]
 ;
   %B = sext i64 %A to i96
@@ -1676,7 +1676,7 @@ define i8 @trunc_lshr_overshift_sext_uses3(i8 %A) {
 define i8 @trunc_lshr_sext_wide_input(i16 %A) {
 ; ALL-LABEL: @trunc_lshr_sext_wide_input(
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr i16 [[A:%.*]], 9
-; ALL-NEXT:    [[D:%.*]] = trunc i16 [[TMP1]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nsw i16 [[TMP1]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i16 %A to i32
@@ -1688,7 +1688,7 @@ define i8 @trunc_lshr_sext_wide_input(i16 %A) {
 define i8 @trunc_lshr_sext_wide_input_exact(i16 %A) {
 ; ALL-LABEL: @trunc_lshr_sext_wide_input_exact(
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr exact i16 [[A:%.*]], 9
-; ALL-NEXT:    [[D:%.*]] = trunc i16 [[TMP1]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nsw i16 [[TMP1]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i16 %A to i32
@@ -1702,7 +1702,7 @@ define <2 x i8> @trunc_lshr_sext_wide_input_uses1(<2 x i16> %A) {
 ; ALL-NEXT:    [[B:%.*]] = sext <2 x i16> [[A:%.*]] to <2 x i32>
 ; ALL-NEXT:    call void @use_v2i32(<2 x i32> [[B]])
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr <2 x i16> [[A]], <i16 9, i16 9>
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i16> [[TMP1]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nsw <2 x i16> [[TMP1]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i16> %A to <2 x i32>
@@ -1747,7 +1747,7 @@ define <2 x i8> @trunc_lshr_sext_wide_input_uses3(<2 x i16> %A) {
 define <2 x i8> @trunc_lshr_overshift_wide_input_sext(<2 x i16> %A) {
 ; ALL-LABEL: @trunc_lshr_overshift_wide_input_sext(
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr <2 x i16> [[A:%.*]], <i16 15, i16 15>
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i16> [[TMP1]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nsw <2 x i16> [[TMP1]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i16> %A to <2 x i32>
@@ -1761,7 +1761,7 @@ define i8 @trunc_lshr_overshift_sext_wide_input_uses1(i16 %A) {
 ; ALL-NEXT:    [[B:%.*]] = sext i16 [[A:%.*]] to i32
 ; ALL-NEXT:    call void @use_i32(i32 [[B]])
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr i16 [[A]], 15
-; ALL-NEXT:    [[D:%.*]] = trunc i16 [[TMP1]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nsw i16 [[TMP1]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i16 %A to i32
@@ -1776,7 +1776,7 @@ define <2 x i8> @trunc_lshr_overshift_sext_wide_input_uses2(<2 x i16> %A) {
 ; ALL-NEXT:    [[TMP1:%.*]] = ashr <2 x i16> [[A:%.*]], <i16 15, i16 15>
 ; ALL-NEXT:    [[C:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i32>
 ; ALL-NEXT:    call void @use_v2i32(<2 x i32> [[C]])
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i16> [[TMP1]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nsw <2 x i16> [[TMP1]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i16> %A to <2 x i32>
@@ -1925,7 +1925,7 @@ define <2 x i8> @trunc_lshr_overshift2_sext(<2 x i8> %A) {
 ; ALL-LABEL: @trunc_lshr_overshift2_sext(
 ; ALL-NEXT:    [[B:%.*]] = sext <2 x i8> [[A:%.*]] to <2 x i32>
 ; ALL-NEXT:    [[C:%.*]] = lshr <2 x i32> [[B]], <i32 25, i32 25>
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i32> [[C]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw <2 x i32> [[C]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i8> %A to <2 x i32>
@@ -1939,7 +1939,7 @@ define i8 @trunc_lshr_overshift2_sext_uses1(i8 %A) {
 ; ALL-NEXT:    [[B:%.*]] = sext i8 [[A:%.*]] to i32
 ; ALL-NEXT:    call void @use_i32(i32 [[B]])
 ; ALL-NEXT:    [[C:%.*]] = lshr i32 [[B]], 25
-; ALL-NEXT:    [[D:%.*]] = trunc i32 [[C]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw i32 [[C]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i8 %A to i32
@@ -1954,7 +1954,7 @@ define <2 x i8> @trunc_lshr_overshift2_sext_uses2(<2 x i8> %A) {
 ; ALL-NEXT:    [[B:%.*]] = sext <2 x i8> [[A:%.*]] to <2 x i32>
 ; ALL-NEXT:    [[C:%.*]] = lshr <2 x i32> [[B]], <i32 25, i32 25>
 ; ALL-NEXT:    call void @use_v2i32(<2 x i32> [[C]])
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i32> [[C]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw <2 x i32> [[C]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i8> %A to <2 x i32>
@@ -1970,7 +1970,7 @@ define i8 @trunc_lshr_overshift2_sext_uses3(i8 %A) {
 ; ALL-NEXT:    call void @use_i32(i32 [[B]])
 ; ALL-NEXT:    [[C:%.*]] = lshr i32 [[B]], 25
 ; ALL-NEXT:    call void @use_i32(i32 [[C]])
-; ALL-NEXT:    [[D:%.*]] = trunc i32 [[C]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw i32 [[C]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i8 %A to i32
@@ -2018,7 +2018,7 @@ define <2 x i8> @trunc_lshr_zext_uniform_undef(<2 x i8> %A) {
 ; ALL-LABEL: @trunc_lshr_zext_uniform_undef(
 ; ALL-NEXT:    [[B:%.*]] = zext <2 x i8> [[A:%.*]] to <2 x i32>
 ; ALL-NEXT:    [[C:%.*]] = lshr <2 x i32> [[B]], <i32 6, i32 undef>
-; ALL-NEXT:    [[D:%.*]] = trunc <2 x i32> [[C]] to <2 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nuw <2 x i32> [[C]] to <2 x i8>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = zext <2 x i8> %A to <2 x i32>
@@ -2042,7 +2042,7 @@ define <3 x i8> @trunc_lshr_zext_nonuniform_undef(<3 x i8> %A) {
 ; ALL-LABEL: @trunc_lshr_zext_nonuniform_undef(
 ; ALL-NEXT:    [[B:%.*]] = zext <3 x i8> [[A:%.*]] to <3 x i32>
 ; ALL-NEXT:    [[C:%.*]] = lshr <3 x i32> [[B]], <i32 6, i32 2, i32 undef>
-; ALL-NEXT:    [[D:%.*]] = trunc <3 x i32> [[C]] to <3 x i8>
+; ALL-NEXT:    [[D:%.*]] = trunc nuw <3 x i32> [[C]] to <3 x i8>
 ; ALL-NEXT:    ret <3 x i8> [[D]]
 ;
   %B = zext <3 x i8> %A to <3 x i32>
@@ -2095,7 +2095,7 @@ define i4 @pr33078_3(i8 %A) {
 ; ALL-LABEL: @pr33078_3(
 ; ALL-NEXT:    [[B:%.*]] = sext i8 [[A:%.*]] to i16
 ; ALL-NEXT:    [[C:%.*]] = lshr i16 [[B]], 12
-; ALL-NEXT:    [[D:%.*]] = trunc i16 [[C]] to i4
+; ALL-NEXT:    [[D:%.*]] = trunc nuw i16 [[C]] to i4
 ; ALL-NEXT:    ret i4 [[D]]
 ;
   %B = sext i8 %A to i16
@@ -2109,7 +2109,7 @@ define i8 @pr33078_4(i3 %x) {
 ; ALL-LABEL: @pr33078_4(
 ; ALL-NEXT:    [[B:%.*]] = sext i3 [[X:%.*]] to i16
 ; ALL-NEXT:    [[C:%.*]] = lshr i16 [[B]], 13
-; ALL-NEXT:    [[D:%.*]] = trunc i16 [[C]] to i8
+; ALL-NEXT:    [[D:%.*]] = trunc nuw nsw i16 [[C]] to i8
 ; ALL-NEXT:    ret i8 [[D]]
 ;
   %B = sext i3 %x to i16
diff --git a/llvm/test/Transforms/InstCombine/cmp-intrinsic.ll b/llvm/test/Transforms/InstCombine/cmp-intrinsic.ll
index 5955650167c21..66cbb2636cbc2 100644
--- a/llvm/test/Transforms/InstCombine/cmp-intrinsic.ll
+++ b/llvm/test/Transforms/InstCombine/cmp-intrinsic.ll
@@ -618,7 +618,7 @@ define i1 @trunc_cttz_false_ult_other_i32_i6(i32 %x) {
 define i1 @trunc_cttz_false_ult_other_i32_i6_extra_use(i32 %x) {
 ; CHECK-LABEL: @trunc_cttz_false_ult_other_i32_i6_extra_use(
 ; CHECK-NEXT:    [[TZ:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 false), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i32 [[TZ]] to i6
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw i32 [[TZ]] to i6
 ; CHECK-NEXT:    call void @use6(i6 [[TRUNC]])
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i6 [[TRUNC]], 7
 ; CHECK-NEXT:    ret i1 [[CMP]]
@@ -720,7 +720,7 @@ define i1 @trunc_ctlz_false_ugt_other_i32_i6(i32 %x) {
 define i1 @trunc_ctlz_false_ugt_other_i32_i6_extra_use(i32 %x) {
 ; CHECK-LABEL: @trunc_ctlz_false_ugt_other_i32_i6_extra_use(
 ; CHECK-NEXT:    [[LZ:%.*]] = tail call i32 @llvm.ctlz.i32(i32 [[X:%.*]], i1 false), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i32 [[LZ]] to i6
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw i32 [[LZ]] to i6
 ; CHECK-NEXT:    call void @use6(i6 [[TRUNC]])
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i6 [[TRUNC]], 4
 ; CHECK-NEXT:    ret i1 [[CMP]]
diff --git a/llvm/test/Transforms/InstCombine/compare-signs.ll b/llvm/test/Transforms/InstCombine/compare-signs.ll
index d7aa710e1ef03..3730d46d5f0f4 100644
--- a/llvm/test/Transforms/InstCombine/compare-signs.ll
+++ b/llvm/test/Transforms/InstCombine/compare-signs.ll
@@ -223,7 +223,7 @@ define <2 x i1> @shift_trunc_signbit_test_vec_uses(<2 x i17> %x, ptr %p1, ptr %p
 ; CHECK-LABEL: @shift_trunc_signbit_test_vec_uses(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr <2 x i17> [[X:%.*]], <i17 4, i17 4>
 ; CHECK-NEXT:    store <2 x i17> [[SH]], ptr [[P1:%.*]], align 8
-; CHECK-NEXT:    [[TR:%.*]] = trunc <2 x i17> [[SH]] to <2 x i13>
+; CHECK-NEXT:    [[TR:%.*]] = trunc nuw <2 x i17> [[SH]] to <2 x i13>
 ; CHECK-NEXT:    store <2 x i13> [[TR]], ptr [[P2:%.*]], align 4
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <2 x i17> [[X]], <i17 -1, i17 -1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
@@ -255,7 +255,7 @@ define i1 @shift_trunc_wrong_shift(i32 %x) {
 define i1 @shift_trunc_wrong_cmp(i32 %x) {
 ; CHECK-LABEL: @shift_trunc_wrong_cmp(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i32 [[X:%.*]], 24
-; CHECK-NEXT:    [[TR:%.*]] = trunc i32 [[SH]] to i8
+; CHECK-NEXT:    [[TR:%.*]] = trunc nuw i32 [[SH]] to i8
 ; CHECK-NEXT:    [[R:%.*]] = icmp slt i8 [[TR]], 1
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/ctpop.ll b/llvm/test/Transforms/InstCombine/ctpop.ll
index dcea5fa87479e..27194724b7d83 100644
--- a/llvm/test/Transforms/InstCombine/ctpop.ll
+++ b/llvm/test/Transforms/InstCombine/ctpop.ll
@@ -397,7 +397,7 @@ define i32 @parity_xor_trunc(i64 %arg, i64 %arg1) {
 ; CHECK-LABEL: @parity_xor_trunc(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor i64 [[ARG1:%.*]], [[ARG:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]]), !range [[RNG5:![0-9]+]]
-; CHECK-NEXT:    [[I4:%.*]] = trunc i64 [[TMP2]] to i32
+; CHECK-NEXT:    [[I4:%.*]] = trunc nuw nsw i64 [[TMP2]] to i32
 ; CHECK-NEXT:    [[I5:%.*]] = and i32 [[I4]], 1
 ; CHECK-NEXT:    ret i32 [[I5]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/extractelement-inseltpoison.ll b/llvm/test/Transforms/InstCombine/extractelement-inseltpoison.ll
index 877aa2e523a31..57e81d2da8989 100644
--- a/llvm/test/Transforms/InstCombine/extractelement-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/extractelement-inseltpoison.ll
@@ -48,7 +48,7 @@ define i32 @bitcasted_inselt_wide_source_zero_elt(i64 %x) {
 ;
 ; BE-LABEL: @bitcasted_inselt_wide_source_zero_elt(
 ; BE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; BE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i32
+; BE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; BE-NEXT:    ret i32 [[R]]
 ;
   %i = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
@@ -64,7 +64,7 @@ define i16 @bitcasted_inselt_wide_source_modulo_elt(i64 %x) {
 ;
 ; BE-LABEL: @bitcasted_inselt_wide_source_modulo_elt(
 ; BE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 48
-; BE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i16
+; BE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i16
 ; BE-NEXT:    ret i16 [[R]]
 ;
   %i = insertelement <2 x i64> poison, i64 %x, i32 1
@@ -76,7 +76,7 @@ define i16 @bitcasted_inselt_wide_source_modulo_elt(i64 %x) {
 define i32 @bitcasted_inselt_wide_source_not_modulo_elt(i64 %x) {
 ; LE-LABEL: @bitcasted_inselt_wide_source_not_modulo_elt(
 ; LE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; LE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i32
+; LE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; LE-NEXT:    ret i32 [[R]]
 ;
 ; BE-LABEL: @bitcasted_inselt_wide_source_not_modulo_elt(
@@ -166,7 +166,7 @@ define i8 @bitcasted_inselt_wide_source_uses(i32 %x) {
 define float @bitcasted_inselt_to_FP(i64 %x) {
 ; LE-LABEL: @bitcasted_inselt_to_FP(
 ; LE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; LE-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; LE-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; LE-NEXT:    [[R:%.*]] = bitcast i32 [[TMP2]] to float
 ; LE-NEXT:    ret float [[R]]
 ;
@@ -218,7 +218,7 @@ define i32 @bitcasted_inselt_from_FP(double %x) {
 ; LE-LABEL: @bitcasted_inselt_from_FP(
 ; LE-NEXT:    [[TMP1:%.*]] = bitcast double [[X:%.*]] to i64
 ; LE-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP1]], 32
-; LE-NEXT:    [[R:%.*]] = trunc i64 [[TMP2]] to i32
+; LE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP2]] to i32
 ; LE-NEXT:    ret i32 [[R]]
 ;
 ; BE-LABEL: @bitcasted_inselt_from_FP(
diff --git a/llvm/test/Transforms/InstCombine/extractelement.ll b/llvm/test/Transforms/InstCombine/extractelement.ll
index bc5dd060a540a..28a4702559c46 100644
--- a/llvm/test/Transforms/InstCombine/extractelement.ll
+++ b/llvm/test/Transforms/InstCombine/extractelement.ll
@@ -50,7 +50,7 @@ define i32 @bitcasted_inselt_wide_source_zero_elt(i64 %x) {
 ;
 ; ANYBE-LABEL: @bitcasted_inselt_wide_source_zero_elt(
 ; ANYBE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; ANYBE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i32
+; ANYBE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; ANYBE-NEXT:    ret i32 [[R]]
 ;
   %i = insertelement <2 x i64> zeroinitializer, i64 %x, i32 0
@@ -66,7 +66,7 @@ define i16 @bitcasted_inselt_wide_source_modulo_elt(i64 %x) {
 ;
 ; ANYBE-LABEL: @bitcasted_inselt_wide_source_modulo_elt(
 ; ANYBE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 48
-; ANYBE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i16
+; ANYBE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i16
 ; ANYBE-NEXT:    ret i16 [[R]]
 ;
   %i = insertelement <2 x i64> undef, i64 %x, i32 1
@@ -78,7 +78,7 @@ define i16 @bitcasted_inselt_wide_source_modulo_elt(i64 %x) {
 define i32 @bitcasted_inselt_wide_source_not_modulo_elt(i64 %x) {
 ; ANYLE-LABEL: @bitcasted_inselt_wide_source_not_modulo_elt(
 ; ANYLE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; ANYLE-NEXT:    [[R:%.*]] = trunc i64 [[TMP1]] to i32
+; ANYLE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; ANYLE-NEXT:    ret i32 [[R]]
 ;
 ; ANYBE-LABEL: @bitcasted_inselt_wide_source_not_modulo_elt(
@@ -168,7 +168,7 @@ define i8 @bitcasted_inselt_wide_source_uses(i32 %x) {
 define float @bitcasted_inselt_to_FP(i64 %x) {
 ; ANYLE-LABEL: @bitcasted_inselt_to_FP(
 ; ANYLE-NEXT:    [[TMP1:%.*]] = lshr i64 [[X:%.*]], 32
-; ANYLE-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; ANYLE-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; ANYLE-NEXT:    [[R:%.*]] = bitcast i32 [[TMP2]] to float
 ; ANYLE-NEXT:    ret float [[R]]
 ;
@@ -220,7 +220,7 @@ define i32 @bitcasted_inselt_from_FP(double %x) {
 ; ANYLE-LABEL: @bitcasted_inselt_from_FP(
 ; ANYLE-NEXT:    [[TMP1:%.*]] = bitcast double [[X:%.*]] to i64
 ; ANYLE-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP1]], 32
-; ANYLE-NEXT:    [[R:%.*]] = trunc i64 [[TMP2]] to i32
+; ANYLE-NEXT:    [[R:%.*]] = trunc nuw i64 [[TMP2]] to i32
 ; ANYLE-NEXT:    ret i32 [[R]]
 ;
 ; ANYBE-LABEL: @bitcasted_inselt_from_FP(
@@ -341,7 +341,7 @@ define i8 @bitcast_scalar_supported_type_index0(i32 %x) {
 ;
 ; ANYBE-LABEL: @bitcast_scalar_supported_type_index0(
 ; ANYBE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 24
-; ANYBE-NEXT:    [[R:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i8
+; ANYBE-NEXT:    [[R:%.*]] = trunc nuw i32 [[EXTELT_OFFSET]] to i8
 ; ANYBE-NEXT:    ret i8 [[R]]
 ;
   %v = bitcast i32 %x to <4 x i8>
@@ -443,7 +443,7 @@ define half @bitcast_fp16vec_index0(i32 %x) {
 ;
 ; ANYBE-LABEL: @bitcast_fp16vec_index0(
 ; ANYBE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
-; ANYBE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; ANYBE-NEXT:    [[TMP1:%.*]] = trunc nuw i32 [[EXTELT_OFFSET]] to i16
 ; ANYBE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
 ; ANYBE-NEXT:    ret half [[R]]
 ;
@@ -455,7 +455,7 @@ define half @bitcast_fp16vec_index0(i32 %x) {
 define half @bitcast_fp16vec_index1(i32 %x) {
 ; ANYLE-LABEL: @bitcast_fp16vec_index1(
 ; ANYLE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
-; ANYLE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; ANYLE-NEXT:    [[TMP1:%.*]] = trunc nuw i32 [[EXTELT_OFFSET]] to i16
 ; ANYLE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to half
 ; ANYLE-NEXT:    ret half [[R]]
 ;
@@ -477,7 +477,7 @@ define bfloat @bitcast_bfp16vec_index0(i32 %x) {
 ;
 ; ANYBE-LABEL: @bitcast_bfp16vec_index0(
 ; ANYBE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
-; ANYBE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; ANYBE-NEXT:    [[TMP1:%.*]] = trunc nuw i32 [[EXTELT_OFFSET]] to i16
 ; ANYBE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to bfloat
 ; ANYBE-NEXT:    ret bfloat [[R]]
 ;
@@ -489,7 +489,7 @@ define bfloat @bitcast_bfp16vec_index0(i32 %x) {
 define bfloat @bitcast_bfp16vec_index1(i32 %x) {
 ; ANYLE-LABEL: @bitcast_bfp16vec_index1(
 ; ANYLE-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i32 [[X:%.*]], 16
-; ANYLE-NEXT:    [[TMP1:%.*]] = trunc i32 [[EXTELT_OFFSET]] to i16
+; ANYLE-NEXT:    [[TMP1:%.*]] = trunc nuw i32 [[EXTELT_OFFSET]] to i16
 ; ANYLE-NEXT:    [[R:%.*]] = bitcast i16 [[TMP1]] to bfloat
 ; ANYLE-NEXT:    ret bfloat [[R]]
 ;
@@ -511,7 +511,7 @@ define float @bitcast_fp32vec_index0(i64 %x) {
 ;
 ; BE64-LABEL: @bitcast_fp32vec_index0(
 ; BE64-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i64 [[X:%.*]], 32
-; BE64-NEXT:    [[TMP1:%.*]] = trunc i64 [[EXTELT_OFFSET]] to i32
+; BE64-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[EXTELT_OFFSET]] to i32
 ; BE64-NEXT:    [[R:%.*]] = bitcast i32 [[TMP1]] to float
 ; BE64-NEXT:    ret float [[R]]
 ;
@@ -528,7 +528,7 @@ define float @bitcast_fp32vec_index0(i64 %x) {
 define float @bitcast_fp32vec_index1(i64 %x) {
 ; LE64-LABEL: @bitcast_fp32vec_index1(
 ; LE64-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i64 [[X:%.*]], 32
-; LE64-NEXT:    [[TMP1:%.*]] = trunc i64 [[EXTELT_OFFSET]] to i32
+; LE64-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[EXTELT_OFFSET]] to i32
 ; LE64-NEXT:    [[R:%.*]] = bitcast i32 [[TMP1]] to float
 ; LE64-NEXT:    ret float [[R]]
 ;
@@ -570,7 +570,7 @@ define double @bitcast_fp64vec_index0(i128 %x) {
 ;
 ; BE128-LABEL: @bitcast_fp64vec_index0(
 ; BE128-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i128 [[X:%.*]], 64
-; BE128-NEXT:    [[TMP1:%.*]] = trunc i128 [[EXTELT_OFFSET]] to i64
+; BE128-NEXT:    [[TMP1:%.*]] = trunc nuw i128 [[EXTELT_OFFSET]] to i64
 ; BE128-NEXT:    [[R:%.*]] = bitcast i64 [[TMP1]] to double
 ; BE128-NEXT:    ret double [[R]]
 ;
@@ -587,7 +587,7 @@ define double @bitcast_fp64vec_index1(i128 %x) {
 ;
 ; LE128-LABEL: @bitcast_fp64vec_index1(
 ; LE128-NEXT:    [[EXTELT_OFFSET:%.*]] = lshr i128 [[X:%.*]], 64
-; LE128-NEXT:    [[TMP1:%.*]] = trunc i128 [[EXTELT_OFFSET]] to i64
+; LE128-NEXT:    [[TMP1:%.*]] = trunc nuw i128 [[EXTELT_OFFSET]] to i64
 ; LE128-NEXT:    [[R:%.*]] = bitcast i64 [[TMP1]] to double
 ; LE128-NEXT:    ret double [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/ffs-1.ll b/llvm/test/Transforms/InstCombine/ffs-1.ll
index a610376da8b59..7cf080765bb1b 100644
--- a/llvm/test/Transforms/InstCombine/ffs-1.ll
+++ b/llvm/test/Transforms/InstCombine/ffs-1.ll
@@ -181,7 +181,7 @@ define i32 @test_simplify15(i64 %x) {
 ;
 ; TARGET-LABEL: @test_simplify15(
 ; TARGET-NEXT:    [[CTTZ:%.*]] = call i64 @llvm.cttz.i64(i64 %x, i1 true), !range !1
-; TARGET-NEXT:    [[TMP1:%.*]] = trunc i64 [[CTTZ]] to i32
+; TARGET-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[CTTZ]] to i32
 ; TARGET-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[TMP1]], 1
 ; TARGET-NEXT:    [[TMP3:%.*]] = icmp eq i64 %x, 0
 ; TARGET-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 0, i32 [[TMP2]]
diff --git a/llvm/test/Transforms/InstCombine/fls.ll b/llvm/test/Transforms/InstCombine/fls.ll
index 8b25a313b6b82..7710093e195a1 100644
--- a/llvm/test/Transforms/InstCombine/fls.ll
+++ b/llvm/test/Transforms/InstCombine/fls.ll
@@ -32,7 +32,7 @@ define i32 @myflsll() {
 define i32 @flsnotconst(i64 %z) {
 ; CHECK-LABEL: @flsnotconst(
 ; CHECK-NEXT:    [[CTLZ:%.*]] = call i64 @llvm.ctlz.i64(i64 [[Z:%.*]], i1 false), !range [[RNG0:![0-9]+]]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[CTLZ]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[CTLZ]] to i32
 ; CHECK-NEXT:    [[GOO:%.*]] = sub nsw i32 64, [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[GOO]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/fold-log2-ceil-idiom.ll b/llvm/test/Transforms/InstCombine/fold-log2-ceil-idiom.ll
index 434d98449f99c..a631aacd97ff9 100644
--- a/llvm/test/Transforms/InstCombine/fold-log2-ceil-idiom.ll
+++ b/llvm/test/Transforms/InstCombine/fold-log2-ceil-idiom.ll
@@ -282,7 +282,7 @@ define i5 @log2_ceil_idiom_trunc_multiuse4(i32 %x) {
 ; CHECK-LABEL: define i5 @log2_ceil_idiom_trunc_multiuse4(
 ; CHECK-SAME: i32 [[X:%.*]]) {
 ; CHECK-NEXT:    [[CTLZ:%.*]] = tail call i32 @llvm.ctlz.i32(i32 [[X]], i1 true), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i32 [[CTLZ]] to i5
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw i32 [[CTLZ]] to i5
 ; CHECK-NEXT:    call void @use5(i5 [[TRUNC]])
 ; CHECK-NEXT:    [[XOR:%.*]] = xor i5 [[TRUNC]], -1
 ; CHECK-NEXT:    [[CTPOP:%.*]] = tail call i32 @llvm.ctpop.i32(i32 [[X]]), !range [[RNG0]]
diff --git a/llvm/test/Transforms/InstCombine/high-bit-signmask-with-trunc.ll b/llvm/test/Transforms/InstCombine/high-bit-signmask-with-trunc.ll
index e87d90909e84a..3ebab115f6543 100644
--- a/llvm/test/Transforms/InstCombine/high-bit-signmask-with-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/high-bit-signmask-with-trunc.ll
@@ -4,7 +4,7 @@
 define i32 @t0(i64 %x) {
 ; CHECK-LABEL: @t0(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = ashr i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    ret i32 [[T1_NEG]]
 ;
   %t0 = lshr i64 %x, 63
@@ -15,7 +15,7 @@ define i32 @t0(i64 %x) {
 define i32 @t1_exact(i64 %x) {
 ; CHECK-LABEL: @t1_exact(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = ashr exact i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    ret i32 [[T1_NEG]]
 ;
   %t0 = lshr exact i64 %x, 63
@@ -26,7 +26,7 @@ define i32 @t1_exact(i64 %x) {
 define i32 @t2(i64 %x) {
 ; CHECK-LABEL: @t2(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = lshr i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nuw nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    ret i32 [[T1_NEG]]
 ;
   %t0 = ashr i64 %x, 63
@@ -37,7 +37,7 @@ define i32 @t2(i64 %x) {
 define i32 @t3_exact(i64 %x) {
 ; CHECK-LABEL: @t3_exact(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = lshr exact i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nuw nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    ret i32 [[T1_NEG]]
 ;
   %t0 = ashr exact i64 %x, 63
@@ -49,7 +49,7 @@ define i32 @t3_exact(i64 %x) {
 define <2 x i32> @t4(<2 x i64> %x) {
 ; CHECK-LABEL: @t4(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = ashr <2 x i64> [[X:%.*]], <i64 63, i64 63>
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc <2 x i64> [[T0_NEG]] to <2 x i32>
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nsw <2 x i64> [[T0_NEG]] to <2 x i32>
 ; CHECK-NEXT:    ret <2 x i32> [[T1_NEG]]
 ;
   %t0 = lshr <2 x i64> %x, <i64 63, i64 63>
@@ -79,7 +79,7 @@ define i32 @t6(i64 %x) {
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = ashr i64 [[X:%.*]], 63
 ; CHECK-NEXT:    [[T0:%.*]] = lshr i64 [[X]], 63
 ; CHECK-NEXT:    call void @use64(i64 [[T0]])
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    ret i32 [[T1_NEG]]
 ;
   %t0 = lshr i64 %x, 63
@@ -92,7 +92,7 @@ define i32 @t6(i64 %x) {
 define i32 @n7(i64 %x) {
 ; CHECK-LABEL: @n7(
 ; CHECK-NEXT:    [[T0:%.*]] = lshr i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1:%.*]] = trunc i64 [[T0]] to i32
+; CHECK-NEXT:    [[T1:%.*]] = trunc nuw nsw i64 [[T0]] to i32
 ; CHECK-NEXT:    call void @use32(i32 [[T1]])
 ; CHECK-NEXT:    [[R:%.*]] = sub nsw i32 0, [[T1]]
 ; CHECK-NEXT:    ret i32 [[R]]
@@ -108,7 +108,7 @@ define i32 @n8(i64 %x) {
 ; CHECK-LABEL: @n8(
 ; CHECK-NEXT:    [[T0:%.*]] = lshr i64 [[X:%.*]], 63
 ; CHECK-NEXT:    call void @use64(i64 [[T0]])
-; CHECK-NEXT:    [[T1:%.*]] = trunc i64 [[T0]] to i32
+; CHECK-NEXT:    [[T1:%.*]] = trunc nuw nsw i64 [[T0]] to i32
 ; CHECK-NEXT:    call void @use32(i32 [[T1]])
 ; CHECK-NEXT:    [[R:%.*]] = sub nsw i32 0, [[T1]]
 ; CHECK-NEXT:    ret i32 [[R]]
@@ -124,7 +124,7 @@ define i32 @n8(i64 %x) {
 define i32 @n9(i64 %x) {
 ; CHECK-LABEL: @n9(
 ; CHECK-NEXT:    [[T0:%.*]] = lshr i64 [[X:%.*]], 62
-; CHECK-NEXT:    [[T1:%.*]] = trunc i64 [[T0]] to i32
+; CHECK-NEXT:    [[T1:%.*]] = trunc nuw nsw i64 [[T0]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = sub nsw i32 0, [[T1]]
 ; CHECK-NEXT:    ret i32 [[R]]
 ;
@@ -137,7 +137,7 @@ define i32 @n9(i64 %x) {
 define i32 @n10(i64 %x) {
 ; CHECK-LABEL: @n10(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = ashr i64 [[X:%.*]], 63
-; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc i64 [[T0_NEG]] to i32
+; CHECK-NEXT:    [[T1_NEG:%.*]] = trunc nsw i64 [[T0_NEG]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = add nsw i32 [[T1_NEG]], 1
 ; CHECK-NEXT:    ret i32 [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/icmp-mul-zext.ll b/llvm/test/Transforms/InstCombine/icmp-mul-zext.ll
index d858c91becb57..aa23a6d27f69b 100644
--- a/llvm/test/Transforms/InstCombine/icmp-mul-zext.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-mul-zext.ll
@@ -60,7 +60,7 @@ define void @PR33765(i8 %beth) {
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[BETH:%.*]] to i32
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nuw nsw i32 [[CONV]], [[CONV]]
 ; CHECK-NEXT:    [[TINKY:%.*]] = load i16, ptr @glob, align 2
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i32 [[MUL]] to i16
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i32 [[MUL]] to i16
 ; CHECK-NEXT:    [[CONV14:%.*]] = and i16 [[TINKY]], [[TMP1]]
 ; CHECK-NEXT:    store i16 [[CONV14]], ptr @glob, align 2
 ; CHECK-NEXT:    ret void
diff --git a/llvm/test/Transforms/InstCombine/icmp-of-trunc-ext.ll b/llvm/test/Transforms/InstCombine/icmp-of-trunc-ext.ll
index 85f67bfa335bb..7f616bbb2a837 100644
--- a/llvm/test/Transforms/InstCombine/icmp-of-trunc-ext.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-of-trunc-ext.ll
@@ -28,8 +28,8 @@ define i1 @icmp_trunc_x_trunc_y_fail_from_illegal1(i256 %x, i256 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i256 [[Y:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[X16:%.*]] = trunc i256 [[X]] to i16
-; CHECK-NEXT:    [[Y16:%.*]] = trunc i256 [[Y]] to i16
+; CHECK-NEXT:    [[X16:%.*]] = trunc nuw i256 [[X]] to i16
+; CHECK-NEXT:    [[Y16:%.*]] = trunc nuw i256 [[Y]] to i16
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq i16 [[X16]], [[Y16]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -49,7 +49,7 @@ define i1 @icmp_trunc_x_trunc_y_illegal_trunc_to_legal_anyways(i123 %x, i32 %y)
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i32 [[Y:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i123 [[X]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i123 [[X]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq i32 [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -89,7 +89,7 @@ define i1 @icmp_trunc_x_trunc_y_3(i64 %x, i32 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i32 [[Y:%.*]], 256
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[X]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[X]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = icmp uge i32 [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -109,7 +109,7 @@ define i1 @icmp_trunc_x_trunc_y_fail_maybe_dirty_upper(i32 %x, i32 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i32 [[Y:%.*]], 65537
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[X16:%.*]] = trunc i32 [[X]] to i16
+; CHECK-NEXT:    [[X16:%.*]] = trunc nuw i32 [[X]] to i16
 ; CHECK-NEXT:    [[Y16:%.*]] = trunc i32 [[Y]] to i16
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne i16 [[X16]], [[Y16]]
 ; CHECK-NEXT:    ret i1 [[R]]
@@ -131,7 +131,7 @@ define i1 @icmp_trunc_x_trunc_y_fail_maybe_dirty_upper_2(i32 %x, i32 %y) {
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
 ; CHECK-NEXT:    [[X16:%.*]] = trunc i32 [[X]] to i16
-; CHECK-NEXT:    [[Y16:%.*]] = trunc i32 [[Y]] to i16
+; CHECK-NEXT:    [[Y16:%.*]] = trunc nuw i32 [[Y]] to i16
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne i16 [[X16]], [[Y16]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -151,7 +151,7 @@ define i1 @icmp_trunc_x_trunc_y_swap0(i33 %x, i32 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i32 [[Y:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i33 [[X]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i33 [[X]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = icmp ule i32 [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -171,7 +171,7 @@ define i1 @icmp_trunc_x_trunc_y_swap1(i33 %x, i32 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i32 [[Y:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i33 [[X]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i33 [[X]] to i32
 ; CHECK-NEXT:    [[R:%.*]] = icmp uge i32 [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -238,7 +238,7 @@ define i1 @icmp_trunc_x_zext_y_3_fail_illegal(i6 %x, i45 %y) {
 ; CHECK-NEXT:    [[Y_LB_ONLY:%.*]] = icmp ult i45 [[Y:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[Y_LB_ONLY]])
 ; CHECK-NEXT:    [[X16:%.*]] = zext i6 [[X:%.*]] to i16
-; CHECK-NEXT:    [[Y16:%.*]] = trunc i45 [[Y]] to i16
+; CHECK-NEXT:    [[Y16:%.*]] = trunc nuw i45 [[Y]] to i16
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne i16 [[Y16]], [[X16]]
 ; CHECK-NEXT:    ret i1 [[R]]
 ;
@@ -254,7 +254,7 @@ define i1 @icmp_trunc_x_zext_y_fail_multiuse(i32 %x, i8 %y) {
 ; CHECK-LABEL: @icmp_trunc_x_zext_y_fail_multiuse(
 ; CHECK-NEXT:    [[X_LB_ONLY:%.*]] = icmp ult i32 [[X:%.*]], 65536
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[X_LB_ONLY]])
-; CHECK-NEXT:    [[X16:%.*]] = trunc i32 [[X]] to i16
+; CHECK-NEXT:    [[X16:%.*]] = trunc nuw i32 [[X]] to i16
 ; CHECK-NEXT:    [[Y16:%.*]] = zext i8 [[Y:%.*]] to i16
 ; CHECK-NEXT:    call void @use(i16 [[Y16]])
 ; CHECK-NEXT:    [[R:%.*]] = icmp ule i16 [[X16]], [[Y16]]
diff --git a/llvm/test/Transforms/InstCombine/icmp-topbitssame.ll b/llvm/test/Transforms/InstCombine/icmp-topbitssame.ll
index 284dc036d11d5..4e11ecbcb8897 100644
--- a/llvm/test/Transforms/InstCombine/icmp-topbitssame.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-topbitssame.ll
@@ -128,7 +128,7 @@ define i1 @wrongimm1(i16 %add) {
 define i1 @wrongimm2(i16 %add) {
 ; CHECK-LABEL: @wrongimm2(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i16 [[ADD:%.*]], 8
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = trunc i16 [[ADD]] to i8
 ; CHECK-NEXT:    [[SHR2_I:%.*]] = ashr i8 [[CONV1_I]], 6
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp eq i8 [[SHR2_I]], [[CONV_I]]
@@ -145,7 +145,7 @@ define i1 @wrongimm2(i16 %add) {
 define i1 @slt(i64 %add) {
 ; CHECK-LABEL: @slt(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i64 [[ADD:%.*]], 32
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i64 [[SH]] to i32
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i64 [[SH]] to i32
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = trunc i64 [[ADD]] to i32
 ; CHECK-NEXT:    [[SHR2_I:%.*]] = ashr i32 [[CONV1_I]], 31
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp slt i32 [[SHR2_I]], [[CONV_I]]
@@ -182,7 +182,7 @@ define i1 @extrause_a(i16 %add) {
 define i1 @extrause_l(i16 %add) {
 ; CHECK-LABEL: @extrause_l(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i16 [[ADD:%.*]], 8
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i16 [[ADD]], 128
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp ult i16 [[TMP1]], 256
 ; CHECK-NEXT:    call void @use(i8 [[CONV_I]])
@@ -200,7 +200,7 @@ define i1 @extrause_l(i16 %add) {
 define i1 @extrause_la(i16 %add) {
 ; CHECK-LABEL: @extrause_la(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i16 [[ADD:%.*]], 8
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = trunc i16 [[ADD]] to i8
 ; CHECK-NEXT:    [[SHR2_I:%.*]] = ashr i8 [[CONV1_I]], 7
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp eq i8 [[SHR2_I]], [[CONV_I]]
diff --git a/llvm/test/Transforms/InstCombine/insert-trunc.ll b/llvm/test/Transforms/InstCombine/insert-trunc.ll
index 3ae128e55b43b..3a160513ccb19 100644
--- a/llvm/test/Transforms/InstCombine/insert-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/insert-trunc.ll
@@ -146,7 +146,7 @@ define <4 x i16> @lshr_same_length_poison_basevec_be(i64 %x) {
 define <4 x i16> @lshr_same_length_poison_basevec_both_endian(i64 %x) {
 ; ALL-LABEL: @lshr_same_length_poison_basevec_both_endian(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <4 x i16> poison, i16 [[T]], i64 0
 ; ALL-NEXT:    ret <4 x i16> [[R]]
 ;
@@ -159,7 +159,7 @@ define <4 x i16> @lshr_same_length_poison_basevec_both_endian(i64 %x) {
 define <4 x i16> @lshr_wrong_index_same_length_poison_basevec(i64 %x) {
 ; ALL-LABEL: @lshr_wrong_index_same_length_poison_basevec(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <4 x i16> poison, i16 [[T]], i64 1
 ; ALL-NEXT:    ret <4 x i16> [[R]]
 ;
@@ -172,7 +172,7 @@ define <4 x i16> @lshr_wrong_index_same_length_poison_basevec(i64 %x) {
 define <8 x i16> @lshr_longer_length_poison_basevec_le(i64 %x) {
 ; ALL-LABEL: @lshr_longer_length_poison_basevec_le(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <8 x i16> poison, i16 [[T]], i64 3
 ; ALL-NEXT:    ret <8 x i16> [[R]]
 ;
@@ -250,7 +250,7 @@ define <4 x i8> @lshr_wrong_index_shorter_length_poison_basevec(i64 %x) {
 define <4 x i8> @lshr_wrong_shift_shorter_length_poison_basevec(i64 %x) {
 ; ALL-LABEL: @lshr_wrong_shift_shorter_length_poison_basevec(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 57
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i8
+; ALL-NEXT:    [[T:%.*]] = trunc nuw nsw i64 [[S]] to i8
 ; ALL-NEXT:    [[R:%.*]] = insertelement <4 x i8> poison, i8 [[T]], i64 0
 ; ALL-NEXT:    ret <4 x i8> [[R]]
 ;
@@ -392,7 +392,7 @@ define <4 x i16> @lshr_same_length_basevec_be(i64 %x, <4 x i16> %v) {
 define <4 x i16> @lshr_same_length_basevec_both_endian(i64 %x, <4 x i16> %v) {
 ; ALL-LABEL: @lshr_same_length_basevec_both_endian(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <4 x i16> [[V:%.*]], i16 [[T]], i64 3
 ; ALL-NEXT:    ret <4 x i16> [[R]]
 ;
@@ -405,7 +405,7 @@ define <4 x i16> @lshr_same_length_basevec_both_endian(i64 %x, <4 x i16> %v) {
 define <4 x i16> @lshr_wrong_index_same_length_basevec(i64 %x, <4 x i16> %v) {
 ; ALL-LABEL: @lshr_wrong_index_same_length_basevec(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <4 x i16> [[V:%.*]], i16 [[T]], i64 1
 ; ALL-NEXT:    ret <4 x i16> [[R]]
 ;
@@ -418,7 +418,7 @@ define <4 x i16> @lshr_wrong_index_same_length_basevec(i64 %x, <4 x i16> %v) {
 define <8 x i16> @lshr_longer_length_basevec_le(i64 %x, <8 x i16> %v) {
 ; ALL-LABEL: @lshr_longer_length_basevec_le(
 ; ALL-NEXT:    [[S:%.*]] = lshr i64 [[X:%.*]], 48
-; ALL-NEXT:    [[T:%.*]] = trunc i64 [[S]] to i16
+; ALL-NEXT:    [[T:%.*]] = trunc nuw i64 [[S]] to i16
 ; ALL-NEXT:    [[R:%.*]] = insertelement <8 x i16> [[V:%.*]], i16 [[T]], i64 3
 ; ALL-NEXT:    ret <8 x i16> [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/insertelt-trunc.ll b/llvm/test/Transforms/InstCombine/insertelt-trunc.ll
index a2721bf13e743..f5f1051ea2014 100644
--- a/llvm/test/Transforms/InstCombine/insertelt-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/insertelt-trunc.ll
@@ -9,7 +9,7 @@ declare void @use_vec(<8 x i16>)
 define <4 x i16> @insert_01_poison_v4i16(i32 %x) {
 ; BE-LABEL: @insert_01_poison_v4i16(
 ; BE-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; BE-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; BE-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; BE-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; BE-NEXT:    [[INS0:%.*]] = insertelement <4 x i16> poison, i16 [[LO16]], i64 0
 ; BE-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[INS0]], i16 [[HI16]], i64 1
@@ -36,7 +36,7 @@ define <8 x i16> @insert_10_poison_v8i16(i32 %x) {
 ;
 ; LE-LABEL: @insert_10_poison_v8i16(
 ; LE-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; LE-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; LE-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; LE-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; LE-NEXT:    [[TMP1:%.*]] = insertelement <8 x i16> poison, i16 [[HI16]], i64 0
 ; LE-NEXT:    [[INS1:%.*]] = insertelement <8 x i16> [[TMP1]], i16 [[LO16]], i64 1
@@ -55,7 +55,7 @@ define <8 x i16> @insert_10_poison_v8i16(i32 %x) {
 define <4 x i32> @insert_12_poison_v4i32(i64 %x) {
 ; ALL-LABEL: @insert_12_poison_v4i32(
 ; ALL-NEXT:    [[HI64:%.*]] = lshr i64 [[X:%.*]], 32
-; ALL-NEXT:    [[HI32:%.*]] = trunc i64 [[HI64]] to i32
+; ALL-NEXT:    [[HI32:%.*]] = trunc nuw i64 [[HI64]] to i32
 ; ALL-NEXT:    [[LO32:%.*]] = trunc i64 [[X]] to i32
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <4 x i32> poison, i32 [[LO32]], i64 1
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i32> [[INS0]], i32 [[HI32]], i64 2
@@ -74,7 +74,7 @@ define <4 x i32> @insert_12_poison_v4i32(i64 %x) {
 define <4 x i16> @insert_21_poison_v4i16(i32 %x) {
 ; ALL-LABEL: @insert_21_poison_v4i16(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[HI16]], i64 1
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[TMP1]], i16 [[LO16]], i64 2
@@ -91,7 +91,7 @@ define <4 x i16> @insert_21_poison_v4i16(i32 %x) {
 define <4 x i32> @insert_23_poison_v4i32(i64 %x) {
 ; BE-LABEL: @insert_23_poison_v4i32(
 ; BE-NEXT:    [[HI64:%.*]] = lshr i64 [[X:%.*]], 32
-; BE-NEXT:    [[HI32:%.*]] = trunc i64 [[HI64]] to i32
+; BE-NEXT:    [[HI32:%.*]] = trunc nuw i64 [[HI64]] to i32
 ; BE-NEXT:    [[LO32:%.*]] = trunc i64 [[X]] to i32
 ; BE-NEXT:    [[INS0:%.*]] = insertelement <4 x i32> poison, i32 [[LO32]], i64 2
 ; BE-NEXT:    [[INS1:%.*]] = insertelement <4 x i32> [[INS0]], i32 [[HI32]], i64 3
@@ -118,7 +118,7 @@ define <4 x i16> @insert_32_poison_v4i16(i32 %x) {
 ;
 ; LE-LABEL: @insert_32_poison_v4i16(
 ; LE-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; LE-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; LE-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; LE-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; LE-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[HI16]], i64 2
 ; LE-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[TMP1]], i16 [[LO16]], i64 3
@@ -140,7 +140,7 @@ define <4 x i16> @insert_32_poison_v4i16(i32 %x) {
 define <2 x i16> @insert_01_v2i16(i32 %x, <2 x i16> %v) {
 ; BE-LABEL: @insert_01_v2i16(
 ; BE-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; BE-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; BE-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; BE-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; BE-NEXT:    [[INS0:%.*]] = insertelement <2 x i16> poison, i16 [[LO16]], i64 0
 ; BE-NEXT:    [[INS1:%.*]] = insertelement <2 x i16> [[INS0]], i16 [[HI16]], i64 1
@@ -163,7 +163,7 @@ define <2 x i16> @insert_01_v2i16(i32 %x, <2 x i16> %v) {
 define <8 x i16> @insert_10_v8i16(i32 %x, <8 x i16> %v) {
 ; ALL-LABEL: @insert_10_v8i16(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[TMP1:%.*]] = insertelement <8 x i16> [[V:%.*]], i16 [[HI16]], i64 0
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <8 x i16> [[TMP1]], i16 [[LO16]], i64 1
@@ -182,7 +182,7 @@ define <8 x i16> @insert_10_v8i16(i32 %x, <8 x i16> %v) {
 define <4 x i32> @insert_12_v4i32(i64 %x, <4 x i32> %v) {
 ; ALL-LABEL: @insert_12_v4i32(
 ; ALL-NEXT:    [[HI64:%.*]] = lshr i64 [[X:%.*]], 32
-; ALL-NEXT:    [[HI32:%.*]] = trunc i64 [[HI64]] to i32
+; ALL-NEXT:    [[HI32:%.*]] = trunc nuw i64 [[HI64]] to i32
 ; ALL-NEXT:    [[LO32:%.*]] = trunc i64 [[X]] to i32
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <4 x i32> [[V:%.*]], i32 [[LO32]], i64 1
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i32> [[INS0]], i32 [[HI32]], i64 2
@@ -201,7 +201,7 @@ define <4 x i32> @insert_12_v4i32(i64 %x, <4 x i32> %v) {
 define <4 x i16> @insert_21_v4i16(i32 %x, <4 x i16> %v) {
 ; ALL-LABEL: @insert_21_v4i16(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> [[V:%.*]], i16 [[HI16]], i64 1
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[TMP1]], i16 [[LO16]], i64 2
@@ -220,7 +220,7 @@ define <4 x i16> @insert_21_v4i16(i32 %x, <4 x i16> %v) {
 define <4 x i32> @insert_23_v4i32(i64 %x, <4 x i32> %v) {
 ; ALL-LABEL: @insert_23_v4i32(
 ; ALL-NEXT:    [[HI64:%.*]] = lshr i64 [[X:%.*]], 32
-; ALL-NEXT:    [[HI32:%.*]] = trunc i64 [[HI64]] to i32
+; ALL-NEXT:    [[HI32:%.*]] = trunc nuw i64 [[HI64]] to i32
 ; ALL-NEXT:    [[LO32:%.*]] = trunc i64 [[X]] to i32
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <4 x i32> [[V:%.*]], i32 [[LO32]], i64 2
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i32> [[INS0]], i32 [[HI32]], i64 3
@@ -239,7 +239,7 @@ define <4 x i32> @insert_23_v4i32(i64 %x, <4 x i32> %v) {
 define <4 x i16> @insert_32_v4i16(i32 %x, <4 x i16> %v) {
 ; ALL-LABEL: @insert_32_v4i16(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> [[V:%.*]], i16 [[HI16]], i64 2
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[TMP1]], i16 [[LO16]], i64 3
@@ -277,7 +277,7 @@ define <4 x i16> @insert_01_v4i16_wrong_shift1(i32 %x) {
 define <4 x i16> @insert_01_v4i16_wrong_op(i32 %x, i32 %y) {
 ; ALL-LABEL: @insert_01_v4i16_wrong_op(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[Y:%.*]] to i16
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <4 x i16> poison, i16 [[LO16]], i64 0
 ; ALL-NEXT:    [[INS1:%.*]] = insertelement <4 x i16> [[INS0]], i16 [[HI16]], i64 1
@@ -296,7 +296,7 @@ define <4 x i16> @insert_01_v4i16_wrong_op(i32 %x, i32 %y) {
 define <8 x i16> @insert_67_v4i16_uses1(i32 %x, <8 x i16> %v) {
 ; ALL-LABEL: @insert_67_v4i16_uses1(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    call void @use(i16 [[HI16]])
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <8 x i16> [[V:%.*]], i16 [[LO16]], i64 6
@@ -318,7 +318,7 @@ define <8 x i16> @insert_67_v4i16_uses1(i32 %x, <8 x i16> %v) {
 define <8 x i16> @insert_76_v4i16_uses2(i32 %x, <8 x i16> %v) {
 ; ALL-LABEL: @insert_76_v4i16_uses2(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    call void @use(i16 [[LO16]])
 ; ALL-NEXT:    [[TMP1:%.*]] = insertelement <8 x i16> [[V:%.*]], i16 [[HI16]], i64 6
@@ -339,7 +339,7 @@ define <8 x i16> @insert_76_v4i16_uses2(i32 %x, <8 x i16> %v) {
 define <8 x i16> @insert_67_v4i16_uses3(i32 %x, <8 x i16> %v) {
 ; ALL-LABEL: @insert_67_v4i16_uses3(
 ; ALL-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; ALL-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; ALL-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; ALL-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; ALL-NEXT:    [[INS0:%.*]] = insertelement <8 x i16> [[V:%.*]], i16 [[LO16]], i64 6
 ; ALL-NEXT:    call void @use_vec(<8 x i16> [[INS0]])
@@ -360,7 +360,7 @@ define <8 x i16> @insert_67_v4i16_uses3(i32 %x, <8 x i16> %v) {
 define <4 x i16> @insert_01_poison_v4i16_high_first(i32 %x) {
 ; BE-LABEL: @insert_01_poison_v4i16_high_first(
 ; BE-NEXT:    [[HI32:%.*]] = lshr i32 [[X:%.*]], 16
-; BE-NEXT:    [[HI16:%.*]] = trunc i32 [[HI32]] to i16
+; BE-NEXT:    [[HI16:%.*]] = trunc nuw i32 [[HI32]] to i16
 ; BE-NEXT:    [[LO16:%.*]] = trunc i32 [[X]] to i16
 ; BE-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[LO16]], i64 0
 ; BE-NEXT:    [[INS0:%.*]] = insertelement <4 x i16> [[TMP1]], i16 [[HI16]], i64 1
diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index e27e4a3eddfbb..d210b19bb7faf 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -455,7 +455,7 @@ define i64 @test_icmp_trunc5(i64 %n) {
 ; CHECK-LABEL: @test_icmp_trunc5(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SHR:%.*]] = ashr i64 [[N:%.*]], 47
-; CHECK-NEXT:    [[CONV1:%.*]] = trunc i64 [[SHR]] to i32
+; CHECK-NEXT:    [[CONV1:%.*]] = trunc nsw i64 [[SHR]] to i32
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[CONV1]], -13
 ; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
diff --git a/llvm/test/Transforms/InstCombine/known-non-zero.ll b/llvm/test/Transforms/InstCombine/known-non-zero.ll
index 7965b47911c41..f1c757cafefb0 100644
--- a/llvm/test/Transforms/InstCombine/known-non-zero.ll
+++ b/llvm/test/Transforms/InstCombine/known-non-zero.ll
@@ -14,7 +14,7 @@ define i32 @test0(i64 %x) {
 ; CHECK-NEXT:    br i1 [[C]], label [[EXIT:%.*]], label [[NON_ZERO:%.*]]
 ; CHECK:       non_zero:
 ; CHECK-NEXT:    [[CTZ:%.*]] = call i64 @llvm.cttz.i64(i64 [[X]], i1 true), !range [[RNG0:![0-9]+]]
-; CHECK-NEXT:    [[CTZ32:%.*]] = trunc i64 [[CTZ]] to i32
+; CHECK-NEXT:    [[CTZ32:%.*]] = trunc nuw nsw i64 [[CTZ]] to i32
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[RES:%.*]] = phi i32 [ [[CTZ32]], [[NON_ZERO]] ], [ 0, [[START:%.*]] ]
@@ -41,7 +41,7 @@ define i32 @test1(i64 %x) {
 ; CHECK-NEXT:    br i1 [[C]], label [[EXIT:%.*]], label [[NON_ZERO:%.*]]
 ; CHECK:       non_zero:
 ; CHECK-NEXT:    [[CTZ:%.*]] = call i64 @llvm.ctlz.i64(i64 [[X]], i1 true), !range [[RNG0]]
-; CHECK-NEXT:    [[CTZ32:%.*]] = trunc i64 [[CTZ]] to i32
+; CHECK-NEXT:    [[CTZ32:%.*]] = trunc nuw nsw i64 [[CTZ]] to i32
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[RES:%.*]] = phi i32 [ [[CTZ32]], [[NON_ZERO]] ], [ 0, [[START:%.*]] ]
diff --git a/llvm/test/Transforms/InstCombine/known-phi-recurse.ll b/llvm/test/Transforms/InstCombine/known-phi-recurse.ll
index 78654155c36cd..d33e08ffaf9b7 100644
--- a/llvm/test/Transforms/InstCombine/known-phi-recurse.ll
+++ b/llvm/test/Transforms/InstCombine/known-phi-recurse.ll
@@ -17,7 +17,7 @@ define i32 @single_entry_phi(i64 %x, i1 %c) {
 ; CHECK-NEXT:    br i1 [[C:%.*]], label [[END:%.*]], label [[BODY]]
 ; CHECK:       end:
 ; CHECK-NEXT:    [[Y:%.*]] = call i64 @llvm.ctpop.i64(i64 [[X:%.*]]), !range [[RNG0:![0-9]+]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[Y]] to i32
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw nsw i64 [[Y]] to i32
 ; CHECK-NEXT:    ret i32 [[TRUNC]]
 ;
 entry:
@@ -37,7 +37,7 @@ define i32 @two_entry_phi_with_constant(i64 %x, i1 %c) {
 ; CHECK-LABEL: @two_entry_phi_with_constant(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[Y:%.*]] = call i64 @llvm.ctpop.i64(i64 [[X:%.*]]), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[Y]] to i32
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw nsw i64 [[Y]] to i32
 ; CHECK-NEXT:    br i1 [[C:%.*]], label [[END:%.*]], label [[BODY:%.*]]
 ; CHECK:       body:
 ; CHECK-NEXT:    br label [[END]]
@@ -62,11 +62,11 @@ define i32 @two_entry_phi_non_constant(i64 %x, i64 %x2, i1 %c) {
 ; CHECK-LABEL: @two_entry_phi_non_constant(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[Y:%.*]] = call i64 @llvm.ctpop.i64(i64 [[X:%.*]]), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[Y]] to i32
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw nsw i64 [[Y]] to i32
 ; CHECK-NEXT:    br i1 [[C:%.*]], label [[END:%.*]], label [[BODY:%.*]]
 ; CHECK:       body:
 ; CHECK-NEXT:    [[Y2:%.*]] = call i64 @llvm.ctpop.i64(i64 [[X2:%.*]]), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC2:%.*]] = trunc i64 [[Y2]] to i32
+; CHECK-NEXT:    [[TRUNC2:%.*]] = trunc nuw nsw i64 [[Y2]] to i32
 ; CHECK-NEXT:    br label [[END]]
 ; CHECK:       end:
 ; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[TRUNC]], [[ENTRY:%.*]] ], [ [[TRUNC2]], [[BODY]] ]
@@ -91,12 +91,12 @@ define i32 @neg_many_branches(i64 %x) {
 ; CHECK-LABEL: @neg_many_branches(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[Y:%.*]] = call i64 @llvm.ctpop.i64(i64 [[X:%.*]]), !range [[RNG0]]
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[Y]] to i32
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw nsw i64 [[Y]] to i32
 ; CHECK-NEXT:    switch i32 [[TRUNC]], label [[END:%.*]] [
-; CHECK-NEXT:    i32 1, label [[ONE:%.*]]
-; CHECK-NEXT:    i32 2, label [[TWO:%.*]]
-; CHECK-NEXT:    i32 3, label [[THREE:%.*]]
-; CHECK-NEXT:    i32 4, label [[FOUR:%.*]]
+; CHECK-NEXT:      i32 1, label [[ONE:%.*]]
+; CHECK-NEXT:      i32 2, label [[TWO:%.*]]
+; CHECK-NEXT:      i32 3, label [[THREE:%.*]]
+; CHECK-NEXT:      i32 4, label [[FOUR:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       one:
 ; CHECK-NEXT:    [[A:%.*]] = add nuw nsw i32 [[TRUNC]], 1
diff --git a/llvm/test/Transforms/InstCombine/logical-select-inseltpoison.ll b/llvm/test/Transforms/InstCombine/logical-select-inseltpoison.ll
index b3d147621b59e..20d60206ebcdf 100644
--- a/llvm/test/Transforms/InstCombine/logical-select-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/logical-select-inseltpoison.ll
@@ -647,7 +647,7 @@ define <4 x i32> @computesignbits_through_shuffles(<4 x float> %x, <4 x float> %
 ; CHECK-NEXT:    [[S3:%.*]] = shufflevector <4 x i32> [[SHUF_OR1]], <4 x i32> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
 ; CHECK-NEXT:    [[S4:%.*]] = shufflevector <4 x i32> [[SHUF_OR1]], <4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
 ; CHECK-NEXT:    [[SHUF_OR2:%.*]] = or <4 x i32> [[S3]], [[S4]]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[SHUF_OR2]] to <4 x i1>
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw <4 x i32> [[SHUF_OR2]] to <4 x i1>
 ; CHECK-NEXT:    [[SEL_V:%.*]] = select <4 x i1> [[TMP1]], <4 x float> [[Z:%.*]], <4 x float> [[X]]
 ; CHECK-NEXT:    [[SEL:%.*]] = bitcast <4 x float> [[SEL_V]] to <4 x i32>
 ; CHECK-NEXT:    ret <4 x i32> [[SEL]]
diff --git a/llvm/test/Transforms/InstCombine/logical-select.ll b/llvm/test/Transforms/InstCombine/logical-select.ll
index c850b87bb2dd4..6e2ed6bf796d0 100644
--- a/llvm/test/Transforms/InstCombine/logical-select.ll
+++ b/llvm/test/Transforms/InstCombine/logical-select.ll
@@ -683,7 +683,7 @@ define <4 x i32> @computesignbits_through_shuffles(<4 x float> %x, <4 x float> %
 ; CHECK-NEXT:    [[S3:%.*]] = shufflevector <4 x i32> [[SHUF_OR1]], <4 x i32> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
 ; CHECK-NEXT:    [[S4:%.*]] = shufflevector <4 x i32> [[SHUF_OR1]], <4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
 ; CHECK-NEXT:    [[SHUF_OR2:%.*]] = or <4 x i32> [[S3]], [[S4]]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[SHUF_OR2]] to <4 x i1>
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw <4 x i32> [[SHUF_OR2]] to <4 x i1>
 ; CHECK-NEXT:    [[SEL_V:%.*]] = select <4 x i1> [[TMP1]], <4 x float> [[Z:%.*]], <4 x float> [[X]]
 ; CHECK-NEXT:    [[SEL:%.*]] = bitcast <4 x float> [[SEL_V]] to <4 x i32>
 ; CHECK-NEXT:    ret <4 x i32> [[SEL]]
diff --git a/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll b/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
index 8d82213d8022f..8e7491ee40370 100644
--- a/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
+++ b/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
@@ -91,7 +91,7 @@ define <2 x i16> @t5_vec_undef(<2 x i8> %x) {
 define i16 @t6_extrause0(i8 %x) {
 ; CHECK-LABEL: @t6_extrause0(
 ; CHECK-NEXT:    [[A:%.*]] = lshr i8 [[X:%.*]], 4
-; CHECK-NEXT:    [[B:%.*]] = trunc i8 [[A]] to i4
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw i8 [[A]] to i4
 ; CHECK-NEXT:    call void @use4(i4 [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext i4 [[B]] to i16
 ; CHECK-NEXT:    ret i16 [[C]]
@@ -157,7 +157,7 @@ define i16 @t10_extrause2(i8 %x) {
 ; CHECK-LABEL: @t10_extrause2(
 ; CHECK-NEXT:    [[A:%.*]] = lshr i8 [[X:%.*]], 4
 ; CHECK-NEXT:    call void @use8(i8 [[A]])
-; CHECK-NEXT:    [[B:%.*]] = trunc i8 [[A]] to i4
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw i8 [[A]] to i4
 ; CHECK-NEXT:    call void @use4(i4 [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext i4 [[B]] to i16
 ; CHECK-NEXT:    ret i16 [[C]]
@@ -189,7 +189,7 @@ define <2 x i16> @t11_extrause2_vec_undef(<2 x i8> %x) {
 define <2 x i10> @wide_source_shifted_signbit(<2 x i32> %x) {
 ; CHECK-LABEL: @wide_source_shifted_signbit(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i32> [[X:%.*]], <i32 24, i32 24>
-; CHECK-NEXT:    [[C:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i10>
+; CHECK-NEXT:    [[C:%.*]] = trunc nsw <2 x i32> [[TMP1]] to <2 x i10>
 ; CHECK-NEXT:    ret <2 x i10> [[C]]
 ;
   %a = lshr <2 x i32> %x, <i32 24, i32 24>
@@ -203,7 +203,7 @@ define i10 @wide_source_shifted_signbit_use1(i32 %x) {
 ; CHECK-NEXT:    [[A:%.*]] = lshr i32 [[X:%.*]], 24
 ; CHECK-NEXT:    call void @use32(i32 [[A]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr i32 [[X]], 24
-; CHECK-NEXT:    [[C:%.*]] = trunc i32 [[TMP1]] to i10
+; CHECK-NEXT:    [[C:%.*]] = trunc nsw i32 [[TMP1]] to i10
 ; CHECK-NEXT:    ret i10 [[C]]
 ;
   %a = lshr i32 %x, 24
@@ -216,7 +216,7 @@ define i10 @wide_source_shifted_signbit_use1(i32 %x) {
 define i10 @wide_source_shifted_signbit_use2(i32 %x) {
 ; CHECK-LABEL: @wide_source_shifted_signbit_use2(
 ; CHECK-NEXT:    [[A:%.*]] = lshr i32 [[X:%.*]], 24
-; CHECK-NEXT:    [[B:%.*]] = trunc i32 [[A]] to i8
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw i32 [[A]] to i8
 ; CHECK-NEXT:    call void @use8(i8 [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext i8 [[B]] to i10
 ; CHECK-NEXT:    ret i10 [[C]]
@@ -256,7 +256,7 @@ define i32 @same_source_shifted_signbit_use1(i32 %x) {
 define i32 @same_source_shifted_signbit_use2(i32 %x) {
 ; CHECK-LABEL: @same_source_shifted_signbit_use2(
 ; CHECK-NEXT:    [[A:%.*]] = lshr i32 [[X:%.*]], 24
-; CHECK-NEXT:    [[B:%.*]] = trunc i32 [[A]] to i8
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw i32 [[A]] to i8
 ; CHECK-NEXT:    call void @use8(i8 [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext i8 [[B]] to i32
 ; CHECK-NEXT:    ret i32 [[C]]
diff --git a/llvm/test/Transforms/InstCombine/lshr.ll b/llvm/test/Transforms/InstCombine/lshr.ll
index 02c2bbc2819b8..7d611ba188d6b 100644
--- a/llvm/test/Transforms/InstCombine/lshr.ll
+++ b/llvm/test/Transforms/InstCombine/lshr.ll
@@ -476,7 +476,7 @@ define i32 @srem2_lshr30(i32 %x) {
 define i12 @trunc_sandwich(i32 %x) {
 ; CHECK-LABEL: @trunc_sandwich(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X:%.*]], 30
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 28
@@ -488,7 +488,7 @@ define i12 @trunc_sandwich(i32 %x) {
 define <2 x i12> @trunc_sandwich_splat_vec(<2 x i32> %x) {
 ; CHECK-LABEL: @trunc_sandwich_splat_vec(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr <2 x i32> [[X:%.*]], <i32 30, i32 30>
-; CHECK-NEXT:    [[R1:%.*]] = trunc <2 x i32> [[SUM_SHIFT]] to <2 x i12>
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw <2 x i32> [[SUM_SHIFT]] to <2 x i12>
 ; CHECK-NEXT:    ret <2 x i12> [[R1]]
 ;
   %sh = lshr <2 x i32> %x, <i32 22, i32 22>
@@ -500,7 +500,7 @@ define <2 x i12> @trunc_sandwich_splat_vec(<2 x i32> %x) {
 define i12 @trunc_sandwich_min_shift1(i32 %x) {
 ; CHECK-LABEL: @trunc_sandwich_min_shift1(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X:%.*]], 21
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 20
@@ -512,7 +512,7 @@ define i12 @trunc_sandwich_min_shift1(i32 %x) {
 define i12 @trunc_sandwich_small_shift1(i32 %x) {
 ; CHECK-LABEL: @trunc_sandwich_small_shift1(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X:%.*]], 20
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    [[R:%.*]] = and i12 [[R1]], 2047
 ; CHECK-NEXT:    ret i12 [[R]]
 ;
@@ -525,7 +525,7 @@ define i12 @trunc_sandwich_small_shift1(i32 %x) {
 define i12 @trunc_sandwich_max_sum_shift(i32 %x) {
 ; CHECK-LABEL: @trunc_sandwich_max_sum_shift(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 20
@@ -537,7 +537,7 @@ define i12 @trunc_sandwich_max_sum_shift(i32 %x) {
 define i12 @trunc_sandwich_max_sum_shift2(i32 %x) {
 ; CHECK-LABEL: @trunc_sandwich_max_sum_shift2(
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 30
@@ -571,7 +571,7 @@ define i12 @trunc_sandwich_use1(i32 %x) {
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i32 [[X:%.*]], 28
 ; CHECK-NEXT:    call void @use(i32 [[SH]])
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X]], 30
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 28
@@ -586,7 +586,7 @@ define <3 x i9> @trunc_sandwich_splat_vec_use1(<3 x i14> %x) {
 ; CHECK-NEXT:    [[SH:%.*]] = lshr <3 x i14> [[X:%.*]], <i14 6, i14 6, i14 6>
 ; CHECK-NEXT:    call void @usevec(<3 x i14> [[SH]])
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr <3 x i14> [[X]], <i14 11, i14 11, i14 11>
-; CHECK-NEXT:    [[R1:%.*]] = trunc <3 x i14> [[SUM_SHIFT]] to <3 x i9>
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw <3 x i14> [[SUM_SHIFT]] to <3 x i9>
 ; CHECK-NEXT:    ret <3 x i9> [[R1]]
 ;
   %sh = lshr <3 x i14> %x, <i14 6, i14 6, i14 6>
@@ -601,7 +601,7 @@ define i12 @trunc_sandwich_min_shift1_use1(i32 %x) {
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i32 [[X:%.*]], 20
 ; CHECK-NEXT:    call void @use(i32 [[SH]])
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X]], 21
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 20
@@ -633,7 +633,7 @@ define i12 @trunc_sandwich_max_sum_shift_use1(i32 %x) {
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i32 [[X:%.*]], 20
 ; CHECK-NEXT:    call void @use(i32 [[SH]])
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X]], 31
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 20
@@ -648,7 +648,7 @@ define i12 @trunc_sandwich_max_sum_shift2_use1(i32 %x) {
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i32 [[X:%.*]], 30
 ; CHECK-NEXT:    call void @use(i32 [[SH]])
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i32 [[X]], 31
-; CHECK-NEXT:    [[R1:%.*]] = trunc i32 [[SUM_SHIFT]] to i12
+; CHECK-NEXT:    [[R1:%.*]] = trunc nuw nsw i32 [[SUM_SHIFT]] to i12
 ; CHECK-NEXT:    ret i12 [[R1]]
 ;
   %sh = lshr i32 %x, 30
diff --git a/llvm/test/Transforms/InstCombine/merging-multiple-stores-into-successor.ll b/llvm/test/Transforms/InstCombine/merging-multiple-stores-into-successor.ll
index 58174f21f767f..866381ff2887f 100644
--- a/llvm/test/Transforms/InstCombine/merging-multiple-stores-into-successor.ll
+++ b/llvm/test/Transforms/InstCombine/merging-multiple-stores-into-successor.ll
@@ -320,7 +320,7 @@ define void @pr46688(i1 %cond, i32 %x, i16 %d, ptr %p1, ptr %p2) {
 ; CHECK-NEXT:    [[THR1_PN:%.*]] = lshr i32 [[THR_PN]], [[X]]
 ; CHECK-NEXT:    [[THR2_PN:%.*]] = lshr i32 [[THR1_PN]], [[X]]
 ; CHECK-NEXT:    [[STOREMERGE:%.*]] = lshr i32 [[THR2_PN]], [[X]]
-; CHECK-NEXT:    [[STOREMERGE1:%.*]] = trunc i32 [[STOREMERGE]] to i16
+; CHECK-NEXT:    [[STOREMERGE1:%.*]] = trunc nuw i32 [[STOREMERGE]] to i16
 ; CHECK-NEXT:    store i16 [[STOREMERGE1]], ptr [[P1:%.*]], align 2
 ; CHECK-NEXT:    store i32 [[STOREMERGE]], ptr [[P2:%.*]], align 4
 ; CHECK-NEXT:    ret void
diff --git a/llvm/test/Transforms/InstCombine/narrow.ll b/llvm/test/Transforms/InstCombine/narrow.ll
index 781974d33bf11..40229f8511f76 100644
--- a/llvm/test/Transforms/InstCombine/narrow.ll
+++ b/llvm/test/Transforms/InstCombine/narrow.ll
@@ -76,7 +76,7 @@ define <2 x i8> @shrink_or_vec(<2 x i16> %a) {
 define i31 @shrink_and(i64 %a) {
 ; CHECK-LABEL: @shrink_and(
 ; CHECK-NEXT:    [[AND:%.*]] = and i64 [[A:%.*]], 42
-; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[AND]] to i31
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw nsw i64 [[AND]] to i31
 ; CHECK-NEXT:    ret i31 [[TRUNC]]
 ;
   %and = and i64 %a, 42
diff --git a/llvm/test/Transforms/InstCombine/negated-bitmask.ll b/llvm/test/Transforms/InstCombine/negated-bitmask.ll
index fe2386bd65c31..9188678186347 100644
--- a/llvm/test/Transforms/InstCombine/negated-bitmask.ll
+++ b/llvm/test/Transforms/InstCombine/negated-bitmask.ll
@@ -70,7 +70,7 @@ define i8 @sub_mask1_trunc_lshr(i64 %a0) {
 ; CHECK-LABEL: @sub_mask1_trunc_lshr(
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[A0:%.*]], 48
 ; CHECK-NEXT:    [[TMP2:%.*]] = ashr i64 [[TMP1]], 63
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP2]] to i8
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nsw i64 [[TMP2]] to i8
 ; CHECK-NEXT:    [[NEG:%.*]] = add nsw i8 [[TMP3]], 10
 ; CHECK-NEXT:    ret i8 [[NEG]]
 ;
@@ -85,7 +85,7 @@ define i32 @sub_sext_mask1_trunc_lshr(i64 %a0) {
 ; CHECK-LABEL: @sub_sext_mask1_trunc_lshr(
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[A0:%.*]], 48
 ; CHECK-NEXT:    [[TMP2:%.*]] = ashr i64 [[TMP1]], 63
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP2]] to i8
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nsw i64 [[TMP2]] to i8
 ; CHECK-NEXT:    [[NARROW:%.*]] = add nsw i8 [[TMP3]], 10
 ; CHECK-NEXT:    [[NEG:%.*]] = zext i8 [[NARROW]] to i32
 ; CHECK-NEXT:    ret i32 [[NEG]]
diff --git a/llvm/test/Transforms/InstCombine/pr34349.ll b/llvm/test/Transforms/InstCombine/pr34349.ll
index ea4afaa245c66..89947650b26b4 100644
--- a/llvm/test/Transforms/InstCombine/pr34349.ll
+++ b/llvm/test/Transforms/InstCombine/pr34349.ll
@@ -7,7 +7,7 @@ define i8 @fast_div_201(i8 %p) {
 ; CHECK-NEXT:    [[V3:%.*]] = zext i8 [[P:%.*]] to i16
 ; CHECK-NEXT:    [[V4:%.*]] = mul nuw nsw i16 [[V3]], 71
 ; CHECK-NEXT:    [[V5:%.*]] = lshr i16 [[V4]], 8
-; CHECK-NEXT:    [[V6:%.*]] = trunc i16 [[V5]] to i8
+; CHECK-NEXT:    [[V6:%.*]] = trunc nuw nsw i16 [[V5]] to i8
 ; CHECK-NEXT:    [[V7:%.*]] = sub i8 [[P]], [[V6]]
 ; CHECK-NEXT:    [[V8:%.*]] = lshr i8 [[V7]], 1
 ; CHECK-NEXT:    [[V13:%.*]] = add nuw i8 [[V8]], [[V6]]
diff --git a/llvm/test/Transforms/InstCombine/reduction-add-sext-zext-i1.ll b/llvm/test/Transforms/InstCombine/reduction-add-sext-zext-i1.ll
index ad55b506a108b..b94be990199bf 100644
--- a/llvm/test/Transforms/InstCombine/reduction-add-sext-zext-i1.ll
+++ b/llvm/test/Transforms/InstCombine/reduction-add-sext-zext-i1.ll
@@ -53,7 +53,7 @@ define i8 @reduce_add_zext_long(<128 x i1> %x) {
 ; CHECK-LABEL: @reduce_add_zext_long(
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <128 x i1> [[X:%.*]] to i128
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i128 @llvm.ctpop.i128(i128 [[TMP1]]), !range [[RNG3:![0-9]+]]
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i128 [[TMP2]] to i8
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw i128 [[TMP2]] to i8
 ; CHECK-NEXT:    [[RES:%.*]] = sub i8 0, [[TMP3]]
 ; CHECK-NEXT:    ret i8 [[RES]]
 ;
@@ -67,7 +67,7 @@ define i8 @reduce_add_zext_long_external_use(<128 x i1> %x) {
 ; CHECK-LABEL: @reduce_add_zext_long_external_use(
 ; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <128 x i1> [[X:%.*]] to i128
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i128 @llvm.ctpop.i128(i128 [[TMP1]]), !range [[RNG3]]
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i128 [[TMP2]] to i8
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw i128 [[TMP2]] to i8
 ; CHECK-NEXT:    [[RES:%.*]] = sub i8 0, [[TMP3]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <128 x i1> [[X]], i64 0
 ; CHECK-NEXT:    [[EXT:%.*]] = sext i1 [[TMP4]] to i8
diff --git a/llvm/test/Transforms/InstCombine/sadd_sat.ll b/llvm/test/Transforms/InstCombine/sadd_sat.ll
index 5ccb6f92b6c72..1cce297122f8a 100644
--- a/llvm/test/Transforms/InstCombine/sadd_sat.ll
+++ b/llvm/test/Transforms/InstCombine/sadd_sat.ll
@@ -79,7 +79,7 @@ define i32 @smul_sat32(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = mul nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
 entry:
@@ -102,7 +102,7 @@ define i32 @smul_sat32_mm(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = mul nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
 entry:
@@ -295,7 +295,7 @@ define signext i4 @sadd_sat4(i4 signext %a, i4 signext %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i32 @llvm.smin.i32(i32 [[ADD]], i32 7)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT10:%.*]] = call i32 @llvm.smax.i32(i32 [[SPEC_STORE_SELECT]], i32 -8)
-; CHECK-NEXT:    [[CONV9:%.*]] = trunc i32 [[SPEC_STORE_SELECT10]] to i4
+; CHECK-NEXT:    [[CONV9:%.*]] = trunc nsw i32 [[SPEC_STORE_SELECT10]] to i4
 ; CHECK-NEXT:    ret i4 [[CONV9]]
 ;
 entry:
@@ -318,7 +318,7 @@ define signext i4 @ssub_sat4(i4 signext %a, i4 signext %b) {
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[CONV]], [[CONV1]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i32 @llvm.smin.i32(i32 [[SUB]], i32 7)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT10:%.*]] = call i32 @llvm.smax.i32(i32 [[SPEC_STORE_SELECT]], i32 -8)
-; CHECK-NEXT:    [[CONV9:%.*]] = trunc i32 [[SPEC_STORE_SELECT10]] to i4
+; CHECK-NEXT:    [[CONV9:%.*]] = trunc nsw i32 [[SPEC_STORE_SELECT10]] to i4
 ; CHECK-NEXT:    ret i4 [[CONV9]]
 ;
 entry:
@@ -465,7 +465,7 @@ define i32 @sadd_sat32_extrause_2(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    call void @use64(i64 [[SPEC_STORE_SELECT]])
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
@@ -490,7 +490,7 @@ define i32 @sadd_sat32_extrause_2_mm(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    call void @use64(i64 [[SPEC_STORE_SELECT]])
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
@@ -513,7 +513,7 @@ define i32 @sadd_sat32_extrause_3(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    call void @use64(i64 [[ADD]])
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
@@ -538,7 +538,7 @@ define i32 @sadd_sat32_extrause_3_mm(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 2147483647)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -2147483648)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    call void @use64(i64 [[ADD]])
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
@@ -561,7 +561,7 @@ define i32 @sadd_sat32_trunc(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smin.i64(i64 [[ADD]], i64 32767)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smax.i64(i64 [[SPEC_STORE_SELECT]], i64 -32768)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
 entry:
@@ -603,7 +603,7 @@ define i8 @sadd_sat8_ext8(i8 %a, i16 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i32 @llvm.smin.i32(i32 [[ADD]], i32 127)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i32 @llvm.smax.i32(i32 [[SPEC_STORE_SELECT]], i32 -128)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i32 [[SPEC_STORE_SELECT8]] to i8
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i32 [[SPEC_STORE_SELECT8]] to i8
 ; CHECK-NEXT:    ret i8 [[CONV7]]
 ;
 entry:
@@ -625,7 +625,7 @@ define i32 @sadd_sat32_zext(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[CONV1:%.*]] = zext i32 [[B:%.*]] to i64
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i64 [[CONV1]], [[CONV]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.umin.i64(i64 [[ADD]], i64 2147483647)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nuw nsw i64 [[SPEC_STORE_SELECT]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
 entry:
@@ -680,7 +680,7 @@ define i32 @ashrA(i64 %a, i32 %b) {
 ; CHECK-LABEL: @ashrA(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i64 [[A:%.*]], 32
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[TMP0]] to i32
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.sadd.sat.i32(i32 [[TMP1]], i32 [[B:%.*]])
 ; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
@@ -698,7 +698,7 @@ define i32 @ashrB(i32 %a, i64 %b) {
 ; CHECK-LABEL: @ashrB(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i64 [[B:%.*]], 32
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[TMP0]] to i32
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.sadd.sat.i32(i32 [[TMP1]], i32 [[A:%.*]])
 ; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
@@ -719,8 +719,8 @@ define i32 @ashrAB(i64 %a, i64 %b) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i64 [[A:%.*]], 32
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[B:%.*]], 32
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw i64 [[TMP0]] to i32
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.sadd.sat.i32(i32 [[TMP2]], i32 [[TMP3]])
 ; CHECK-NEXT:    ret i32 [[TMP4]]
 ;
@@ -744,7 +744,7 @@ define i32 @ashrA31(i64 %a, i32 %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[CONV]], [[CONV1]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call i64 @llvm.smax.i64(i64 [[ADD]], i64 -2147483648)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call i64 @llvm.smin.i64(i64 [[SPEC_STORE_SELECT]], i64 2147483647)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i64 [[SPEC_STORE_SELECT8]] to i32
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i64 [[SPEC_STORE_SELECT8]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV7]]
 ;
 entry:
@@ -763,7 +763,7 @@ define i32 @ashrA33(i64 %a, i32 %b) {
 ; CHECK-LABEL: @ashrA33(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CONV:%.*]] = ashr i64 [[A:%.*]], 33
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[CONV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nsw i64 [[CONV]] to i32
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.sadd.sat.i32(i32 [[TMP0]], i32 [[B:%.*]])
 ; CHECK-NEXT:    ret i32 [[TMP1]]
 ;
@@ -787,7 +787,7 @@ define <2 x i8> @ashrv2i8(<2 x i16> %a, <2 x i8> %b) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add <2 x i16> [[CONV]], [[CONV1]]
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT:%.*]] = call <2 x i16> @llvm.smax.v2i16(<2 x i16> [[ADD]], <2 x i16> <i16 -128, i16 -128>)
 ; CHECK-NEXT:    [[SPEC_STORE_SELECT8:%.*]] = call <2 x i16> @llvm.smin.v2i16(<2 x i16> [[SPEC_STORE_SELECT]], <2 x i16> <i16 127, i16 127>)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc <2 x i16> [[SPEC_STORE_SELECT8]] to <2 x i8>
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw <2 x i16> [[SPEC_STORE_SELECT8]] to <2 x i8>
 ; CHECK-NEXT:    ret <2 x i8> [[CONV7]]
 ;
 entry:
@@ -806,7 +806,7 @@ define <2 x i8> @ashrv2i8_s(<2 x i16> %a, <2 x i8> %b) {
 ; CHECK-LABEL: @ashrv2i8_s(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr <2 x i16> [[A:%.*]], <i16 8, i16 8>
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc <2 x i16> [[TMP0]] to <2 x i8>
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw <2 x i16> [[TMP0]] to <2 x i8>
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> [[TMP1]], <2 x i8> [[B:%.*]])
 ; CHECK-NEXT:    ret <2 x i8> [[TMP2]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll b/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
index 03dd6188ac039..69896f855f5f1 100644
--- a/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
+++ b/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
@@ -220,7 +220,7 @@ define i64 @test6c(i32 %x) {
 define i16 @test1d(i64 %x) {
 ; CHECK-LABEL: @test1d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[CT]] to i16
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[CT]] to i16
 ; CHECK-NEXT:    ret i16 [[CONV]]
 ;
   %ct = tail call i64 @llvm.cttz.i64(i64 %x, i1 true)
@@ -233,7 +233,7 @@ define i16 @test1d(i64 %x) {
 define i32 @test2d(i64 %x) {
 ; CHECK-LABEL: @test2d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %ct = tail call i64 @llvm.cttz.i64(i64 %x, i1 true)
@@ -246,7 +246,7 @@ define i32 @test2d(i64 %x) {
 define i16 @test3d(i32 %x) {
 ; CHECK-LABEL: @test3d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 false), !range [[RNG1]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i32 [[CT]] to i16
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i32 [[CT]] to i16
 ; CHECK-NEXT:    ret i16 [[CAST]]
 ;
   %ct = tail call i32 @llvm.cttz.i32(i32 %x, i1 true)
@@ -259,7 +259,7 @@ define i16 @test3d(i32 %x) {
 define i16 @test4d(i64 %x) {
 ; CHECK-LABEL: @test4d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.ctlz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i16
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i16
 ; CHECK-NEXT:    ret i16 [[CAST]]
 ;
   %ct = tail call i64 @llvm.ctlz.i64(i64 %x, i1 true)
@@ -272,7 +272,7 @@ define i16 @test4d(i64 %x) {
 define i32 @test5d(i64 %x) {
 ; CHECK-LABEL: @test5d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.ctlz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %ct = tail call i64 @llvm.ctlz.i64(i64 %x, i1 true)
@@ -288,7 +288,7 @@ define i32 @not_op_ctlz(i64 %x) {
 ; CHECK-LABEL: @not_op_ctlz(
 ; CHECK-NEXT:    [[N:%.*]] = xor i64 [[X:%.*]], -1
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.ctlz.i64(i64 [[N]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %n = xor i64 %x, -1
@@ -303,7 +303,7 @@ define i32 @not_op_cttz(i64 %x) {
 ; CHECK-LABEL: @not_op_cttz(
 ; CHECK-NEXT:    [[N:%.*]] = xor i64 [[X:%.*]], -1
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[N]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %n = xor i64 %x, -1
@@ -320,7 +320,7 @@ define i32 @not_op_ctlz_wrong_xor_op1(i64 %x) {
 ; CHECK-LABEL: @not_op_ctlz_wrong_xor_op1(
 ; CHECK-NEXT:    [[N:%.*]] = xor i64 [[X:%.*]], -2
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.ctlz.i64(i64 [[N]], i1 true), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp eq i64 [[X]], -1
 ; CHECK-NEXT:    [[R:%.*]] = select i1 [[TOBOOL]], i32 64, i32 [[CAST]]
 ; CHECK-NEXT:    ret i32 [[R]]
@@ -339,7 +339,7 @@ define i32 @not_op_ctlz_wrong_xor_op0(i64 %x, i64 %y) {
 ; CHECK-LABEL: @not_op_ctlz_wrong_xor_op0(
 ; CHECK-NEXT:    [[N:%.*]] = xor i64 [[Y:%.*]], -1
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.ctlz.i64(i64 [[N]], i1 true), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp eq i64 [[X:%.*]], -1
 ; CHECK-NEXT:    [[R:%.*]] = select i1 [[TOBOOL]], i32 64, i32 [[CAST]]
 ; CHECK-NEXT:    ret i32 [[R]]
@@ -358,7 +358,7 @@ define i32 @not_op_cttz_wrong_cmp(i64 %x) {
 ; CHECK-LABEL: @not_op_cttz_wrong_cmp(
 ; CHECK-NEXT:    [[N:%.*]] = xor i64 [[X:%.*]], -1
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[N]], i1 true), !range [[RNG2]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i64 [[CT]] to i32
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i64 [[CT]] to i32
 ; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp eq i64 [[X]], 0
 ; CHECK-NEXT:    [[R:%.*]] = select i1 [[TOBOOL]], i32 64, i32 [[CAST]]
 ; CHECK-NEXT:    ret i32 [[R]]
@@ -374,7 +374,7 @@ define i32 @not_op_cttz_wrong_cmp(i64 %x) {
 define i16 @test6d(i32 %x) {
 ; CHECK-LABEL: @test6d(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.ctlz.i32(i32 [[X:%.*]], i1 false), !range [[RNG1]]
-; CHECK-NEXT:    [[CAST:%.*]] = trunc i32 [[CT]] to i16
+; CHECK-NEXT:    [[CAST:%.*]] = trunc nuw nsw i32 [[CT]] to i16
 ; CHECK-NEXT:    ret i16 [[CAST]]
 ;
   %ct = tail call i32 @llvm.ctlz.i32(i32 %x, i1 true)
@@ -400,7 +400,7 @@ define i64 @select_bug1(i32 %x) {
 define i16 @select_bug2(i32 %x) {
 ; CHECK-LABEL: @select_bug2(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 false), !range [[RNG1]]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i32 [[CT]] to i16
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i32 [[CT]] to i16
 ; CHECK-NEXT:    ret i16 [[CONV]]
 ;
   %ct = tail call i32 @llvm.cttz.i32(i32 %x, i1 false)
@@ -595,7 +595,7 @@ define i64 @test_multiuse_zext_undef(i32 %x, ptr %p) {
 define i16 @test_multiuse_trunc_def(i64 %x, ptr %p) {
 ; CHECK-LABEL: @test_multiuse_trunc_def(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[CT]] to i16
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[CT]] to i16
 ; CHECK-NEXT:    store i16 [[CONV]], ptr [[P:%.*]], align 2
 ; CHECK-NEXT:    ret i16 [[CONV]]
 ;
@@ -610,7 +610,7 @@ define i16 @test_multiuse_trunc_def(i64 %x, ptr %p) {
 define i16 @test_multiuse_trunc_undef(i64 %x, ptr %p) {
 ; CHECK-LABEL: @test_multiuse_trunc_undef(
 ; CHECK-NEXT:    [[CT:%.*]] = tail call i64 @llvm.cttz.i64(i64 [[X:%.*]], i1 false), !range [[RNG2]]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[CT]] to i16
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[CT]] to i16
 ; CHECK-NEXT:    store i16 [[CONV]], ptr [[P:%.*]], align 2
 ; CHECK-NEXT:    ret i16 [[CONV]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/select-imm-canon.ll b/llvm/test/Transforms/InstCombine/select-imm-canon.ll
index cde6329fd1b27..6d57af9d939d9 100644
--- a/llvm/test/Transforms/InstCombine/select-imm-canon.ll
+++ b/llvm/test/Transforms/InstCombine/select-imm-canon.ll
@@ -20,7 +20,7 @@ define i8 @double(i32 %A) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.smax.i32(i32 [[A:%.*]], i32 -128)
 ; CHECK-NEXT:    [[CONV71:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP0]], i32 127)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i32 [[CONV71]] to i8
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i32 [[CONV71]] to i8
 ; CHECK-NEXT:    ret i8 [[CONV7]]
 ;
 entry:
@@ -51,7 +51,7 @@ define i8 @original(i32 %A, i32 %B) {
 ; CHECK-LABEL: @original(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[A:%.*]], i32 -128)
 ; CHECK-NEXT:    [[SPEC_SELECT_I:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP1]], i32 127)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i32 [[SPEC_SELECT_I]] to i8
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i32 [[SPEC_SELECT_I]] to i8
 ; CHECK-NEXT:    ret i8 [[CONV7]]
 ;
   %cmp4.i = icmp slt i32 127, %A
@@ -68,7 +68,7 @@ define i8 @original_logical(i32 %A, i32 %B) {
 ; CHECK-LABEL: @original_logical(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[A:%.*]], i32 -128)
 ; CHECK-NEXT:    [[SPEC_SELECT_I:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP1]], i32 127)
-; CHECK-NEXT:    [[CONV7:%.*]] = trunc i32 [[SPEC_SELECT_I]] to i8
+; CHECK-NEXT:    [[CONV7:%.*]] = trunc nsw i32 [[SPEC_SELECT_I]] to i8
 ; CHECK-NEXT:    ret i8 [[CONV7]]
 ;
   %cmp4.i = icmp slt i32 127, %A
diff --git a/llvm/test/Transforms/InstCombine/select.ll b/llvm/test/Transforms/InstCombine/select.ll
index 05fcf66235295..bd8145ab2a35b 100644
--- a/llvm/test/Transforms/InstCombine/select.ll
+++ b/llvm/test/Transforms/InstCombine/select.ll
@@ -452,7 +452,7 @@ define i64 @test21(i32 %x) {
 define i16 @test22(i32 %x) {
 ; CHECK-LABEL: @test22(
 ; CHECK-NEXT:    [[X_LOBIT:%.*]] = ashr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[RETVAL:%.*]] = trunc i32 [[X_LOBIT]] to i16
+; CHECK-NEXT:    [[RETVAL:%.*]] = trunc nsw i32 [[X_LOBIT]] to i16
 ; CHECK-NEXT:    ret i16 [[RETVAL]]
 ;
   %t = icmp slt i32 %x, 0
diff --git a/llvm/test/Transforms/InstCombine/sext-of-trunc-nsw.ll b/llvm/test/Transforms/InstCombine/sext-of-trunc-nsw.ll
index 5b9334ab93cb4..b992460d0be69 100644
--- a/llvm/test/Transforms/InstCombine/sext-of-trunc-nsw.ll
+++ b/llvm/test/Transforms/InstCombine/sext-of-trunc-nsw.ll
@@ -86,7 +86,7 @@ define i16 @t5_extrause(i8 %x) {
 ; CHECK-LABEL: @t5_extrause(
 ; CHECK-NEXT:    [[A:%.*]] = ashr i8 [[X:%.*]], 5
 ; CHECK-NEXT:    call void @use8(i8 [[A]])
-; CHECK-NEXT:    [[B:%.*]] = trunc i8 [[A]] to i4
+; CHECK-NEXT:    [[B:%.*]] = trunc nsw i8 [[A]] to i4
 ; CHECK-NEXT:    call void @use4(i4 [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext i8 [[A]] to i16
 ; CHECK-NEXT:    ret i16 [[C]]
@@ -134,7 +134,7 @@ define i24 @wide_source_matching_signbits(i32 %x) {
 ; CHECK-LABEL: @wide_source_matching_signbits(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 7
 ; CHECK-NEXT:    [[A:%.*]] = shl nsw i32 -1, [[M]]
-; CHECK-NEXT:    [[C:%.*]] = trunc i32 [[A]] to i24
+; CHECK-NEXT:    [[C:%.*]] = trunc nsw i32 [[A]] to i24
 ; CHECK-NEXT:    ret i24 [[C]]
 ;
   %m = and i32 %x, 7
@@ -194,7 +194,7 @@ define i32 @same_source_matching_signbits_extra_use(i32 %x) {
 ; CHECK-LABEL: @same_source_matching_signbits_extra_use(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 7
 ; CHECK-NEXT:    [[A:%.*]] = shl nsw i32 -1, [[M]]
-; CHECK-NEXT:    [[B:%.*]] = trunc i32 [[A]] to i8
+; CHECK-NEXT:    [[B:%.*]] = trunc nsw i32 [[A]] to i8
 ; CHECK-NEXT:    call void @use8(i8 [[B]])
 ; CHECK-NEXT:    ret i32 [[A]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/sext.ll b/llvm/test/Transforms/InstCombine/sext.ll
index 186745362a448..e3b6058ce7f80 100644
--- a/llvm/test/Transforms/InstCombine/sext.ll
+++ b/llvm/test/Transforms/InstCombine/sext.ll
@@ -385,7 +385,7 @@ define i16 @smear_set_bit_different_dest_type(i32 %x) {
 ; CHECK-LABEL: @smear_set_bit_different_dest_type(
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl i32 [[X:%.*]], 24
 ; CHECK-NEXT:    [[TMP2:%.*]] = ashr i32 [[TMP1]], 31
-; CHECK-NEXT:    [[S:%.*]] = trunc i32 [[TMP2]] to i16
+; CHECK-NEXT:    [[S:%.*]] = trunc nsw i32 [[TMP2]] to i16
 ; CHECK-NEXT:    ret i16 [[S]]
 ;
   %t = trunc i32 %x to i8
diff --git a/llvm/test/Transforms/InstCombine/shift-add.ll b/llvm/test/Transforms/InstCombine/shift-add.ll
index aa3a238e0949c..7f948848844c5 100644
--- a/llvm/test/Transforms/InstCombine/shift-add.ll
+++ b/llvm/test/Transforms/InstCombine/shift-add.ll
@@ -742,7 +742,7 @@ define <3 x i32> @add3_i96(<3 x i32> %0, <3 x i32> %1) {
 ; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <3 x i32> [[TMP1]], i64 2
 ; CHECK-NEXT:    [[TMP14:%.*]] = add i32 [[TMP13]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = lshr i64 [[TMP11]], 32
-; CHECK-NEXT:    [[TMP16:%.*]] = trunc i64 [[TMP15]] to i32
+; CHECK-NEXT:    [[TMP16:%.*]] = trunc nuw nsw i64 [[TMP15]] to i32
 ; CHECK-NEXT:    [[TMP17:%.*]] = add i32 [[TMP14]], [[TMP16]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = insertelement <3 x i32> poison, i32 [[ADD_NARROWED]], i64 0
 ; CHECK-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP11]] to i32
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest-with-truncation-lshr.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest-with-truncation-lshr.ll
index 60a7dce2a8753..a0a3c8edfb4b5 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest-with-truncation-lshr.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest-with-truncation-lshr.ll
@@ -139,7 +139,7 @@ define i1 @n4(i32 %x, i32 %len) {
 ; CHECK-NEXT:    [[T2:%.*]] = add i32 [[LEN]], -16
 ; CHECK-NEXT:    [[T2_WIDE:%.*]] = zext nneg i32 [[T2]] to i64
 ; CHECK-NEXT:    [[T3:%.*]] = lshr i64 262143, [[T2_WIDE]]
-; CHECK-NEXT:    [[T3_TRUNC:%.*]] = trunc i64 [[T3]] to i32
+; CHECK-NEXT:    [[T3_TRUNC:%.*]] = trunc nuw nsw i64 [[T3]] to i32
 ; CHECK-NEXT:    [[T4:%.*]] = and i32 [[T1]], [[T3_TRUNC]]
 ; CHECK-NEXT:    [[T5:%.*]] = icmp ne i32 [[T4]], 0
 ; CHECK-NEXT:    ret i1 [[T5]]
@@ -229,7 +229,7 @@ define <2 x i1> @n8_vec(<2 x i32> %x, <2 x i32> %len) {
 ; CHECK-NEXT:    [[T2:%.*]] = add <2 x i32> [[LEN]], <i32 -16, i32 -16>
 ; CHECK-NEXT:    [[T2_WIDE:%.*]] = zext nneg <2 x i32> [[T2]] to <2 x i64>
 ; CHECK-NEXT:    [[T3:%.*]] = lshr <2 x i64> <i64 131071, i64 262143>, [[T2_WIDE]]
-; CHECK-NEXT:    [[T3_TRUNC:%.*]] = trunc <2 x i64> [[T3]] to <2 x i32>
+; CHECK-NEXT:    [[T3_TRUNC:%.*]] = trunc nuw nsw <2 x i64> [[T3]] to <2 x i32>
 ; CHECK-NEXT:    [[T4:%.*]] = and <2 x i32> [[T1]], [[T3_TRUNC]]
 ; CHECK-NEXT:    [[T5:%.*]] = icmp ne <2 x i32> [[T4]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[T5]]
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
index 6773cbac1d1e8..84dd4c57ebc61 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
@@ -13,7 +13,7 @@
 define i16 @t0(i32 %x, i16 %y) {
 ; CHECK-LABEL: @t0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
@@ -30,7 +30,7 @@ define i16 @t0(i32 %x, i16 %y) {
 define <2 x i16> @t1_vec_splat(<2 x i32> %x, <2 x i16> %y) {
 ; CHECK-LABEL: @t1_vec_splat(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i32> [[X:%.*]], <i32 31, i32 31>
-; CHECK-NEXT:    [[T5:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i16>
+; CHECK-NEXT:    [[T5:%.*]] = trunc nsw <2 x i32> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[T5]]
 ;
   %t0 = sub <2 x i16> <i16 32, i16 32>, %y
@@ -100,7 +100,7 @@ define i16 @t6_extrause0(i32 %x, i16 %y) {
 ; CHECK-NEXT:    [[T3:%.*]] = trunc i32 [[T2]] to i16
 ; CHECK-NEXT:    call void @use16(i16 [[T3]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr i32 [[X]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
@@ -118,7 +118,7 @@ define i16 @t7_extrause1(i32 %x, i16 %y) {
 ; CHECK-NEXT:    [[T4:%.*]] = add i16 [[Y:%.*]], -1
 ; CHECK-NEXT:    call void @use16(i16 [[T4]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
index 63099a8af81f6..214ec88d2e551 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
@@ -13,7 +13,7 @@
 define i16 @t0(i32 %x, i16 %y) {
 ; CHECK-LABEL: @t0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
@@ -30,7 +30,7 @@ define i16 @t0(i32 %x, i16 %y) {
 define <2 x i16> @t1_vec_splat(<2 x i32> %x, <2 x i16> %y) {
 ; CHECK-LABEL: @t1_vec_splat(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <2 x i32> [[X:%.*]], <i32 31, i32 31>
-; CHECK-NEXT:    [[T5:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i16>
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw <2 x i32> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[T5]]
 ;
   %t0 = sub <2 x i16> <i16 32, i16 32>, %y
@@ -100,7 +100,7 @@ define i16 @t6_extrause0(i32 %x, i16 %y) {
 ; CHECK-NEXT:    [[T3:%.*]] = trunc i32 [[T2]] to i16
 ; CHECK-NEXT:    call void @use16(i16 [[T3]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[X]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
@@ -118,7 +118,7 @@ define i16 @t7_extrause1(i32 %x, i16 %y) {
 ; CHECK-NEXT:    [[T4:%.*]] = add i16 [[Y:%.*]], -1
 ; CHECK-NEXT:    call void @use16(i16 [[T4]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[X:%.*]], 31
-; CHECK-NEXT:    [[T5:%.*]] = trunc i32 [[TMP1]] to i16
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw i32 [[TMP1]] to i16
 ; CHECK-NEXT:    ret i16 [[T5]]
 ;
   %t0 = sub i16 32, %y
diff --git a/llvm/test/Transforms/InstCombine/shift-shift.ll b/llvm/test/Transforms/InstCombine/shift-shift.ll
index 8a40863300d45..7c35718601ba7 100644
--- a/llvm/test/Transforms/InstCombine/shift-shift.ll
+++ b/llvm/test/Transforms/InstCombine/shift-shift.ll
@@ -166,7 +166,7 @@ define i8 @shl_trunc_smaller_lshr(i32 %x) {
 define i24 @shl_trunc_bigger_ashr(i32 %x) {
 ; CHECK-LABEL: @shl_trunc_bigger_ashr(
 ; CHECK-NEXT:    [[SH_DIFF:%.*]] = ashr i32 [[X:%.*]], 9
-; CHECK-NEXT:    [[TR_SH_DIFF:%.*]] = trunc i32 [[SH_DIFF]] to i24
+; CHECK-NEXT:    [[TR_SH_DIFF:%.*]] = trunc nsw i32 [[SH_DIFF]] to i24
 ; CHECK-NEXT:    [[LT:%.*]] = and i24 [[TR_SH_DIFF]], -8
 ; CHECK-NEXT:    ret i24 [[LT]]
 ;
@@ -502,7 +502,7 @@ define <2 x i6> @shl_lshr_demand5_undef_left(<2 x i8> %x) {
 ; CHECK-LABEL: @shl_lshr_demand5_undef_left(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <2 x i8> <i8 undef, i8 -108>, [[X:%.*]]
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <2 x i8> [[SHL]], <i8 2, i8 2>
-; CHECK-NEXT:    [[R:%.*]] = trunc <2 x i8> [[LSHR]] to <2 x i6>
+; CHECK-NEXT:    [[R:%.*]] = trunc nuw <2 x i8> [[LSHR]] to <2 x i6>
 ; CHECK-NEXT:    ret <2 x i6> [[R]]
 ;
   %shl = shl <2 x i8> <i8 undef, i8 148>, %x ; 0b1001_0100
@@ -561,7 +561,7 @@ define <2 x i6> @shl_lshr_demand5_nonuniform_vec_both(<2 x i8> %x) {
 ; CHECK-LABEL: @shl_lshr_demand5_nonuniform_vec_both(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <2 x i8> <i8 -104, i8 -108>, [[X:%.*]]
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <2 x i8> [[SHL]], <i8 3, i8 2>
-; CHECK-NEXT:    [[R:%.*]] = trunc <2 x i8> [[LSHR]] to <2 x i6>
+; CHECK-NEXT:    [[R:%.*]] = trunc nuw <2 x i8> [[LSHR]] to <2 x i6>
 ; CHECK-NEXT:    ret <2 x i6> [[R]]
 ;
   %shl = shl <2 x i8> <i8 152, i8 148>, %x ; 0b1001_1000, 0b1001_0100
diff --git a/llvm/test/Transforms/InstCombine/shift.ll b/llvm/test/Transforms/InstCombine/shift.ll
index bef7fc81a7d1f..bb8661919c89f 100644
--- a/llvm/test/Transforms/InstCombine/shift.ll
+++ b/llvm/test/Transforms/InstCombine/shift.ll
@@ -423,7 +423,7 @@ define i32 @test29(i64 %d18) {
 ; CHECK-LABEL: @test29(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr i64 [[D18:%.*]], 63
-; CHECK-NEXT:    [[I101:%.*]] = trunc i64 [[SUM_SHIFT]] to i32
+; CHECK-NEXT:    [[I101:%.*]] = trunc nuw nsw i64 [[SUM_SHIFT]] to i32
 ; CHECK-NEXT:    ret i32 [[I101]]
 ;
 entry:
@@ -437,7 +437,7 @@ define <2 x i32> @test29_uniform(<2 x i64> %d18) {
 ; CHECK-LABEL: @test29_uniform(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SUM_SHIFT:%.*]] = lshr <2 x i64> [[D18:%.*]], <i64 63, i64 63>
-; CHECK-NEXT:    [[I101:%.*]] = trunc <2 x i64> [[SUM_SHIFT]] to <2 x i32>
+; CHECK-NEXT:    [[I101:%.*]] = trunc nuw nsw <2 x i64> [[SUM_SHIFT]] to <2 x i32>
 ; CHECK-NEXT:    ret <2 x i32> [[I101]]
 ;
 entry:
@@ -466,7 +466,7 @@ define <2 x i32> @test29_poison(<2 x i64> %d18) {
 ; CHECK-LABEL: @test29_poison(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[I916:%.*]] = lshr <2 x i64> [[D18:%.*]], <i64 32, i64 poison>
-; CHECK-NEXT:    [[I917:%.*]] = trunc <2 x i64> [[I916]] to <2 x i32>
+; CHECK-NEXT:    [[I917:%.*]] = trunc nuw <2 x i64> [[I916]] to <2 x i32>
 ; CHECK-NEXT:    [[I10:%.*]] = lshr <2 x i32> [[I917]], <i32 31, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[I10]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/shl-demand.ll b/llvm/test/Transforms/InstCombine/shl-demand.ll
index 26175ebbe1535..08e6e74581848 100644
--- a/llvm/test/Transforms/InstCombine/shl-demand.ll
+++ b/llvm/test/Transforms/InstCombine/shl-demand.ll
@@ -222,7 +222,7 @@ define i8 @must_drop_poison(i32 %x, i32 %y)  {
 define i32 @f_t15_t01_t09(i40 %t2) {
 ; CHECK-LABEL: @f_t15_t01_t09(
 ; CHECK-NEXT:    [[SH_DIFF:%.*]] = ashr i40 [[T2:%.*]], 15
-; CHECK-NEXT:    [[TR_SH_DIFF:%.*]] = trunc i40 [[SH_DIFF]] to i32
+; CHECK-NEXT:    [[TR_SH_DIFF:%.*]] = trunc nsw i40 [[SH_DIFF]] to i32
 ; CHECK-NEXT:    [[SHL1:%.*]] = and i32 [[TR_SH_DIFF]], -65536
 ; CHECK-NEXT:    ret i32 [[SHL1]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/sign-bit-test-via-right-shifting-all-other-bits.ll b/llvm/test/Transforms/InstCombine/sign-bit-test-via-right-shifting-all-other-bits.ll
index d3ac6cfa9c601..30fad66bf5218 100644
--- a/llvm/test/Transforms/InstCombine/sign-bit-test-via-right-shifting-all-other-bits.ll
+++ b/llvm/test/Transforms/InstCombine/sign-bit-test-via-right-shifting-all-other-bits.ll
@@ -322,7 +322,7 @@ define i1 @unsigned_sign_bit_extract_with_trunc_extrause(i64 %x) {
 ; CHECK-LABEL: @unsigned_sign_bit_extract_with_trunc_extrause(
 ; CHECK-NEXT:    [[SIGNBIT:%.*]] = lshr i64 [[X:%.*]], 63
 ; CHECK-NEXT:    call void @use64(i64 [[SIGNBIT]])
-; CHECK-NEXT:    [[SIGNBIT_NARROW:%.*]] = trunc i64 [[SIGNBIT]] to i32
+; CHECK-NEXT:    [[SIGNBIT_NARROW:%.*]] = trunc nuw nsw i64 [[SIGNBIT]] to i32
 ; CHECK-NEXT:    call void @use32(i32 [[SIGNBIT_NARROW]])
 ; CHECK-NEXT:    [[ISNEG:%.*]] = icmp slt i64 [[X]], 0
 ; CHECK-NEXT:    ret i1 [[ISNEG]]
@@ -348,7 +348,7 @@ define i1 @signed_sign_bit_extract_trunc_extrause(i64 %x) {
 ; CHECK-LABEL: @signed_sign_bit_extract_trunc_extrause(
 ; CHECK-NEXT:    [[SIGNSMEAR:%.*]] = ashr i64 [[X:%.*]], 63
 ; CHECK-NEXT:    call void @use64(i64 [[SIGNSMEAR]])
-; CHECK-NEXT:    [[SIGNSMEAR_NARROW:%.*]] = trunc i64 [[SIGNSMEAR]] to i32
+; CHECK-NEXT:    [[SIGNSMEAR_NARROW:%.*]] = trunc nsw i64 [[SIGNSMEAR]] to i32
 ; CHECK-NEXT:    call void @use32(i32 [[SIGNSMEAR_NARROW]])
 ; CHECK-NEXT:    [[ISNEG:%.*]] = icmp slt i64 [[X]], 0
 ; CHECK-NEXT:    ret i1 [[ISNEG]]
diff --git a/llvm/test/Transforms/InstCombine/trunc-demand.ll b/llvm/test/Transforms/InstCombine/trunc-demand.ll
index 4f6e79285eaa8..9d7bf589268e2 100644
--- a/llvm/test/Transforms/InstCombine/trunc-demand.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-demand.ll
@@ -36,7 +36,7 @@ define i6 @trunc_lshr_exact_mask(i8 %x) {
 define i6 @trunc_lshr_big_mask(i8 %x) {
 ; CHECK-LABEL: @trunc_lshr_big_mask(
 ; CHECK-NEXT:    [[S:%.*]] = lshr i8 [[X:%.*]], 2
-; CHECK-NEXT:    [[T:%.*]] = trunc i8 [[S]] to i6
+; CHECK-NEXT:    [[T:%.*]] = trunc nuw i8 [[S]] to i6
 ; CHECK-NEXT:    [[R:%.*]] = and i6 [[T]], 31
 ; CHECK-NEXT:    ret i6 [[R]]
 ;
@@ -52,7 +52,7 @@ define i6 @trunc_lshr_use1(i8 %x) {
 ; CHECK-LABEL: @trunc_lshr_use1(
 ; CHECK-NEXT:    [[S:%.*]] = lshr i8 [[X:%.*]], 2
 ; CHECK-NEXT:    call void @use8(i8 [[S]])
-; CHECK-NEXT:    [[T:%.*]] = trunc i8 [[S]] to i6
+; CHECK-NEXT:    [[T:%.*]] = trunc nuw i8 [[S]] to i6
 ; CHECK-NEXT:    [[R:%.*]] = and i6 [[T]], 15
 ; CHECK-NEXT:    ret i6 [[R]]
 ;
@@ -68,7 +68,7 @@ define i6 @trunc_lshr_use1(i8 %x) {
 define i6 @trunc_lshr_use2(i8 %x) {
 ; CHECK-LABEL: @trunc_lshr_use2(
 ; CHECK-NEXT:    [[S:%.*]] = lshr i8 [[X:%.*]], 2
-; CHECK-NEXT:    [[T:%.*]] = trunc i8 [[S]] to i6
+; CHECK-NEXT:    [[T:%.*]] = trunc nuw i8 [[S]] to i6
 ; CHECK-NEXT:    call void @use6(i6 [[T]])
 ; CHECK-NEXT:    [[R:%.*]] = and i6 [[T]], 15
 ; CHECK-NEXT:    ret i6 [[R]]
@@ -157,7 +157,7 @@ define i6 @or_trunc_lshr_more(i8 %x) {
 define i6 @or_trunc_lshr_small_mask(i8 %x) {
 ; CHECK-LABEL: @or_trunc_lshr_small_mask(
 ; CHECK-NEXT:    [[S:%.*]] = lshr i8 [[X:%.*]], 4
-; CHECK-NEXT:    [[T:%.*]] = trunc i8 [[S]] to i6
+; CHECK-NEXT:    [[T:%.*]] = trunc nuw nsw i8 [[S]] to i6
 ; CHECK-NEXT:    [[R:%.*]] = or i6 [[T]], -8
 ; CHECK-NEXT:    ret i6 [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
index 87c90bb91f39e..4c857125365a9 100644
--- a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
@@ -171,7 +171,7 @@ define i32 @test5(i32 %A) {
 define i32 @test6(i64 %A) {
 ; CHECK-LABEL: @test6(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[A:%.*]], 32
-; CHECK-NEXT:    [[D:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    ret i32 [[D]]
 ;
   %B = zext i64 %A to i128
@@ -459,7 +459,7 @@ define <2 x i64> @test12_vec_undef(<2 x i32> %A, <2 x i32> %B) {
 ; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
 ; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
 ; CHECK-NEXT:    [[F:%.*]] = lshr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = trunc nuw nsw <2 x i128> [[F]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
@@ -524,7 +524,7 @@ define <2 x i64> @test13_vec_undef(<2 x i32> %A, <2 x i32> %B) {
 ; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
 ; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
 ; CHECK-NEXT:    [[F:%.*]] = ashr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = trunc nsw <2 x i128> [[F]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = sext <2 x i32> %A to <2 x i128>
diff --git a/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll b/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
index e578b604c9d6a..2c5f428cf98de 100644
--- a/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
@@ -72,7 +72,7 @@ define i8 @trunc_lshr_trunc_outofrange(i64 %a) {
 ; CHECK-LABEL: @trunc_lshr_trunc_outofrange(
 ; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A:%.*]] to i32
 ; CHECK-NEXT:    [[C:%.*]] = lshr i32 [[B]], 25
-; CHECK-NEXT:    [[D:%.*]] = trunc i32 [[C]] to i8
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw nsw i32 [[C]] to i8
 ; CHECK-NEXT:    ret i8 [[D]]
 ;
   %b = trunc i64 %a to i32
@@ -158,7 +158,7 @@ define i8 @trunc_ashr_trunc_outofrange(i64 %a) {
 ; CHECK-LABEL: @trunc_ashr_trunc_outofrange(
 ; CHECK-NEXT:    [[B:%.*]] = trunc i64 [[A:%.*]] to i32
 ; CHECK-NEXT:    [[C:%.*]] = ashr i32 [[B]], 25
-; CHECK-NEXT:    [[D:%.*]] = trunc i32 [[C]] to i8
+; CHECK-NEXT:    [[D:%.*]] = trunc nsw i32 [[C]] to i8
 ; CHECK-NEXT:    ret i8 [[D]]
 ;
   %b = trunc i64 %a to i32
diff --git a/llvm/test/Transforms/InstCombine/trunc.ll b/llvm/test/Transforms/InstCombine/trunc.ll
index 760825d6b1da0..c77d7269f2cf7 100644
--- a/llvm/test/Transforms/InstCombine/trunc.ll
+++ b/llvm/test/Transforms/InstCombine/trunc.ll
@@ -171,7 +171,7 @@ define i32 @test5(i32 %A) {
 define i32 @test6(i64 %A) {
 ; CHECK-LABEL: @test6(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr i64 [[A:%.*]], 32
-; CHECK-NEXT:    [[D:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[D:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    ret i32 [[D]]
 ;
   %B = zext i64 %A to i128
@@ -459,7 +459,7 @@ define <2 x i64> @test12_vec_undef(<2 x i32> %A, <2 x i32> %B) {
 ; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
 ; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
 ; CHECK-NEXT:    [[F:%.*]] = lshr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = trunc nuw nsw <2 x i128> [[F]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
@@ -524,7 +524,7 @@ define <2 x i64> @test13_vec_undef(<2 x i32> %A, <2 x i32> %B) {
 ; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
 ; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
 ; CHECK-NEXT:    [[F:%.*]] = ashr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = trunc nsw <2 x i128> [[F]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = sext <2 x i32> %A to <2 x i128>
diff --git a/llvm/test/Transforms/InstCombine/truncating-saturate.ll b/llvm/test/Transforms/InstCombine/truncating-saturate.ll
index e4df94afd1741..c0111528e2a4d 100644
--- a/llvm/test/Transforms/InstCombine/truncating-saturate.ll
+++ b/llvm/test/Transforms/InstCombine/truncating-saturate.ll
@@ -10,7 +10,7 @@ define i8 @testi16i8(i16 %add) {
 ; CHECK-LABEL: @testi16i8(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i16 @llvm.smax.i16(i16 [[ADD:%.*]], i16 -128)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i16 @llvm.smin.i16(i16 [[TMP1]], i16 127)
-; CHECK-NEXT:    [[COND_I:%.*]] = trunc i16 [[TMP2]] to i8
+; CHECK-NEXT:    [[COND_I:%.*]] = trunc nsw i16 [[TMP2]] to i8
 ; CHECK-NEXT:    ret i8 [[COND_I]]
 ;
   %sh = lshr i16 %add, 8
@@ -29,7 +29,7 @@ define i32 @testi64i32(i64 %add) {
 ; CHECK-LABEL: @testi64i32(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.smax.i64(i64 [[ADD:%.*]], i64 -2147483648)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.smin.i64(i64 [[TMP1]], i64 2147483647)
-; CHECK-NEXT:    [[COND_I:%.*]] = trunc i64 [[TMP2]] to i32
+; CHECK-NEXT:    [[COND_I:%.*]] = trunc nsw i64 [[TMP2]] to i32
 ; CHECK-NEXT:    ret i32 [[COND_I]]
 ;
   %sh = lshr i64 %add, 32
@@ -48,7 +48,7 @@ define i16 @testi32i16i8(i32 %add) {
 ; CHECK-LABEL: @testi32i16i8(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[ADD:%.*]], i32 -128)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.smin.i32(i32 [[TMP1]], i32 127)
-; CHECK-NEXT:    [[R:%.*]] = trunc i32 [[TMP2]] to i16
+; CHECK-NEXT:    [[R:%.*]] = trunc nsw i32 [[TMP2]] to i16
 ; CHECK-NEXT:    ret i16 [[R]]
 ;
   %a = add i32 %add, 128
@@ -64,7 +64,7 @@ define <4 x i16> @testv4i32i16i8(<4 x i32> %add) {
 ; CHECK-LABEL: @testv4i32i16i8(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.smax.v4i32(<4 x i32> [[ADD:%.*]], <4 x i32> <i32 -128, i32 -128, i32 -128, i32 -128>)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <4 x i32> @llvm.smin.v4i32(<4 x i32> [[TMP1]], <4 x i32> <i32 127, i32 127, i32 127, i32 127>)
-; CHECK-NEXT:    [[R:%.*]] = trunc <4 x i32> [[TMP2]] to <4 x i16>
+; CHECK-NEXT:    [[R:%.*]] = trunc nsw <4 x i32> [[TMP2]] to <4 x i16>
 ; CHECK-NEXT:    ret <4 x i16> [[R]]
 ;
   %a = add <4 x i32> %add, <i32 128, i32 128, i32 128, i32 128>
@@ -149,7 +149,7 @@ define <4 x i8> @testv4i16i8(<4 x i16> %add) {
 ; CHECK-LABEL: @testv4i16i8(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i16> @llvm.smax.v4i16(<4 x i16> [[ADD:%.*]], <4 x i16> <i16 -128, i16 -128, i16 -128, i16 -128>)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <4 x i16> @llvm.smin.v4i16(<4 x i16> [[TMP1]], <4 x i16> <i16 127, i16 127, i16 127, i16 127>)
-; CHECK-NEXT:    [[COND_I:%.*]] = trunc <4 x i16> [[TMP2]] to <4 x i8>
+; CHECK-NEXT:    [[COND_I:%.*]] = trunc nsw <4 x i16> [[TMP2]] to <4 x i8>
 ; CHECK-NEXT:    ret <4 x i8> [[COND_I]]
 ;
   %sh = lshr <4 x i16> %add, <i16 8, i16 8, i16 8, i16 8>
@@ -188,7 +188,7 @@ define i8 @testi16i8_revcmp(i16 %add) {
 ; CHECK-LABEL: @testi16i8_revcmp(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i16 @llvm.smax.i16(i16 [[ADD:%.*]], i16 -128)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i16 @llvm.smin.i16(i16 [[TMP1]], i16 127)
-; CHECK-NEXT:    [[COND_I:%.*]] = trunc i16 [[TMP2]] to i8
+; CHECK-NEXT:    [[COND_I:%.*]] = trunc nsw i16 [[TMP2]] to i8
 ; CHECK-NEXT:    ret i8 [[COND_I]]
 ;
   %sh = lshr i16 %add, 8
@@ -207,7 +207,7 @@ define i8 @testi16i8_revselect(i16 %add) {
 ; CHECK-LABEL: @testi16i8_revselect(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i16 @llvm.smax.i16(i16 [[ADD:%.*]], i16 -128)
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i16 @llvm.smin.i16(i16 [[TMP1]], i16 127)
-; CHECK-NEXT:    [[COND_I:%.*]] = trunc i16 [[TMP2]] to i8
+; CHECK-NEXT:    [[COND_I:%.*]] = trunc nsw i16 [[TMP2]] to i8
 ; CHECK-NEXT:    ret i8 [[COND_I]]
 ;
   %sh = lshr i16 %add, 8
@@ -268,7 +268,7 @@ define i16 @differentconsts(i32 %x, i16 %replacement_low, i16 %replacement_high)
 define i8 @badimm1(i16 %add) {
 ; CHECK-LABEL: @badimm1(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i16 [[ADD:%.*]], 9
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw nsw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = trunc i16 [[ADD]] to i8
 ; CHECK-NEXT:    [[SHR2_I:%.*]] = ashr i8 [[CONV1_I]], 7
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp eq i8 [[SHR2_I]], [[CONV_I]]
@@ -292,7 +292,7 @@ define i8 @badimm1(i16 %add) {
 define i8 @badimm2(i16 %add) {
 ; CHECK-LABEL: @badimm2(
 ; CHECK-NEXT:    [[SH:%.*]] = lshr i16 [[ADD:%.*]], 8
-; CHECK-NEXT:    [[CONV_I:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[CONV_I:%.*]] = trunc nuw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = trunc i16 [[ADD]] to i8
 ; CHECK-NEXT:    [[SHR2_I:%.*]] = ashr i8 [[CONV1_I]], 6
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp eq i8 [[SHR2_I]], [[CONV_I]]
@@ -319,7 +319,7 @@ define i8 @badimm3(i16 %add) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i16 [[ADD]], 128
 ; CHECK-NEXT:    [[CMP_NOT_I:%.*]] = icmp ult i16 [[TMP1]], 256
 ; CHECK-NEXT:    [[SHR4_I:%.*]] = ashr i16 [[ADD]], 14
-; CHECK-NEXT:    [[CONV5_I:%.*]] = trunc i16 [[SHR4_I]] to i8
+; CHECK-NEXT:    [[CONV5_I:%.*]] = trunc nsw i16 [[SHR4_I]] to i8
 ; CHECK-NEXT:    [[XOR_I:%.*]] = xor i8 [[CONV5_I]], 127
 ; CHECK-NEXT:    [[COND_I:%.*]] = select i1 [[CMP_NOT_I]], i8 [[CONV1_I]], i8 [[XOR_I]]
 ; CHECK-NEXT:    ret i8 [[COND_I]]
diff --git a/llvm/test/Transforms/InstCombine/vector-trunc.ll b/llvm/test/Transforms/InstCombine/vector-trunc.ll
index eeb5a3fdb7398..bccb12e66eba1 100644
--- a/llvm/test/Transforms/InstCombine/vector-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/vector-trunc.ll
@@ -4,7 +4,7 @@
 define <4 x i16> @trunc_add_nsw(<4 x i32> %0) {
 ; CHECK-LABEL: @trunc_add_nsw(
 ; CHECK-NEXT:    [[TMP2:%.*]] = ashr <4 x i32> [[TMP0:%.*]], <i32 17, i32 17, i32 17, i32 17>
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc <4 x i32> [[TMP2]] to <4 x i16>
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nsw <4 x i32> [[TMP2]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i16> [[TMP3]], <i16 1, i16 1, i16 1, i16 1>
 ; CHECK-NEXT:    ret <4 x i16> [[TMP4]]
 ;
@@ -17,7 +17,7 @@ define <4 x i16> @trunc_add_nsw(<4 x i32> %0) {
 define <4 x i16> @trunc_add_no_nsw(<4 x i32> %0) {
 ; CHECK-LABEL: @trunc_add_no_nsw(
 ; CHECK-NEXT:    [[TMP2:%.*]] = lshr <4 x i32> [[TMP0:%.*]], <i32 16, i32 16, i32 16, i32 16>
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc <4 x i32> [[TMP2]] to <4 x i16>
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw <4 x i32> [[TMP2]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[TMP3]], <i16 1, i16 1, i16 1, i16 1>
 ; CHECK-NEXT:    ret <4 x i16> [[TMP4]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/xor-ashr.ll b/llvm/test/Transforms/InstCombine/xor-ashr.ll
index 32ca8e338c2cc..097e04b3b9cb5 100644
--- a/llvm/test/Transforms/InstCombine/xor-ashr.ll
+++ b/llvm/test/Transforms/InstCombine/xor-ashr.ll
@@ -80,7 +80,7 @@ define <4 x i8> @testv4i16i8_undef(<4 x i16> %add) {
 define i8 @wrongimm(i16 %add) {
 ; CHECK-LABEL: @wrongimm(
 ; CHECK-NEXT:    [[SH:%.*]] = ashr i16 [[ADD:%.*]], 14
-; CHECK-NEXT:    [[T:%.*]] = trunc i16 [[SH]] to i8
+; CHECK-NEXT:    [[T:%.*]] = trunc nsw i16 [[SH]] to i8
 ; CHECK-NEXT:    [[X:%.*]] = xor i8 [[T]], 27
 ; CHECK-NEXT:    ret i8 [[X]]
 ;
@@ -140,7 +140,7 @@ define i16 @extrause_trunc1(i32 %add) {
 define i16 @extrause_trunc2(i32 %add) {
 ; CHECK-LABEL: @extrause_trunc2(
 ; CHECK-NEXT:    [[SH:%.*]] = ashr i32 [[ADD:%.*]], 31
-; CHECK-NEXT:    [[T:%.*]] = trunc i32 [[SH]] to i16
+; CHECK-NEXT:    [[T:%.*]] = trunc nsw i32 [[SH]] to i16
 ; CHECK-NEXT:    call void @use16(i16 [[T]])
 ; CHECK-NEXT:    [[X:%.*]] = xor i16 [[T]], 127
 ; CHECK-NEXT:    ret i16 [[X]]
diff --git a/llvm/test/Transforms/InstCombine/zext-ctlz-trunc-to-ctlz-add.ll b/llvm/test/Transforms/InstCombine/zext-ctlz-trunc-to-ctlz-add.ll
index f082873bf7839..c8eb513a8440b 100644
--- a/llvm/test/Transforms/InstCombine/zext-ctlz-trunc-to-ctlz-add.ll
+++ b/llvm/test/Transforms/InstCombine/zext-ctlz-trunc-to-ctlz-add.ll
@@ -57,7 +57,7 @@ define <2 x i17> @trunc_ctlz_zext_v2i17_v2i32_multiple_uses(<2 x i17> %x) {
 ; CHECK-LABEL: @trunc_ctlz_zext_v2i17_v2i32_multiple_uses(
 ; CHECK-NEXT:    [[Z:%.*]] = zext <2 x i17> [[X:%.*]] to <2 x i32>
 ; CHECK-NEXT:    [[P:%.*]] = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> [[Z]], i1 false), !range [[RNG2:![0-9]+]]
-; CHECK-NEXT:    [[ZZ:%.*]] = trunc <2 x i32> [[P]] to <2 x i17>
+; CHECK-NEXT:    [[ZZ:%.*]] = trunc nuw nsw <2 x i32> [[P]] to <2 x i17>
 ; CHECK-NEXT:    call void @use(<2 x i32> [[P]])
 ; CHECK-NEXT:    ret <2 x i17> [[ZZ]]
 ;
@@ -91,7 +91,7 @@ define i16 @trunc_ctlz_zext_i10_i32(i10 %x) {
 ; CHECK-LABEL: @trunc_ctlz_zext_i10_i32(
 ; CHECK-NEXT:    [[Z:%.*]] = zext i10 [[X:%.*]] to i32
 ; CHECK-NEXT:    [[P:%.*]] = call i32 @llvm.ctlz.i32(i32 [[Z]], i1 false), !range [[RNG3:![0-9]+]]
-; CHECK-NEXT:    [[ZZ:%.*]] = trunc i32 [[P]] to i16
+; CHECK-NEXT:    [[ZZ:%.*]] = trunc nuw nsw i32 [[P]] to i16
 ; CHECK-NEXT:    ret i16 [[ZZ]]
 ;
   %z = zext i10 %x to i32
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll b/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
index 9a7b9f570cf77..ed8d8e15282d5 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
@@ -34,14 +34,14 @@ define void @test_pr25490(i32 %n, ptr noalias nocapture %a, ptr noalias nocaptur
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i16>
 ; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw <16 x i16> [[TMP3]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = lshr <16 x i16> [[TMP5]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
-; CHECK-NEXT:    [[TMP7:%.*]] = trunc <16 x i16> [[TMP6]] to <16 x i8>
+; CHECK-NEXT:    [[TMP7:%.*]] = trunc nuw <16 x i16> [[TMP6]] to <16 x i8>
 ; CHECK-NEXT:    store <16 x i8> [[TMP7]], ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i16>
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw <16 x i16> [[TMP9]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = lshr <16 x i16> [[TMP10]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
-; CHECK-NEXT:    [[TMP12:%.*]] = trunc <16 x i16> [[TMP11]] to <16 x i8>
+; CHECK-NEXT:    [[TMP12:%.*]] = trunc nuw <16 x i16> [[TMP11]] to <16 x i8>
 ; CHECK-NEXT:    store <16 x i8> [[TMP12]], ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -67,14 +67,14 @@ define void @test_pr25490(i32 %n, ptr noalias nocapture %a, ptr noalias nocaptur
 ; CHECK-NEXT:    [[TMP17:%.*]] = zext <8 x i8> [[WIDE_LOAD8]] to <8 x i16>
 ; CHECK-NEXT:    [[TMP18:%.*]] = mul nuw <8 x i16> [[TMP16]], [[TMP17]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = lshr <8 x i16> [[TMP18]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
-; CHECK-NEXT:    [[TMP20:%.*]] = trunc <8 x i16> [[TMP19]] to <8 x i8>
+; CHECK-NEXT:    [[TMP20:%.*]] = trunc nuw <8 x i16> [[TMP19]] to <8 x i8>
 ; CHECK-NEXT:    store <8 x i8> [[TMP20]], ptr [[TMP15]], align 1
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX7]]
 ; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <8 x i8>, ptr [[TMP21]], align 1
 ; CHECK-NEXT:    [[TMP22:%.*]] = zext <8 x i8> [[WIDE_LOAD10]] to <8 x i16>
 ; CHECK-NEXT:    [[TMP23:%.*]] = mul nuw <8 x i16> [[TMP22]], [[TMP17]]
 ; CHECK-NEXT:    [[TMP24:%.*]] = lshr <8 x i16> [[TMP23]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
-; CHECK-NEXT:    [[TMP25:%.*]] = trunc <8 x i16> [[TMP24]] to <8 x i8>
+; CHECK-NEXT:    [[TMP25:%.*]] = trunc nuw <8 x i16> [[TMP24]] to <8 x i8>
 ; CHECK-NEXT:    store <8 x i8> [[TMP25]], ptr [[TMP21]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX7]], 8
 ; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT11]], [[N_VEC5]]
@@ -99,14 +99,14 @@ define void @test_pr25490(i32 %n, ptr noalias nocapture %a, ptr noalias nocaptur
 ; CHECK-NEXT:    [[CONV3:%.*]] = zext i8 [[TMP28]] to i32
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nuw nsw i32 [[CONV3]], [[CONV]]
 ; CHECK-NEXT:    [[SHR_26:%.*]] = lshr i32 [[MUL]], 8
-; CHECK-NEXT:    [[CONV4:%.*]] = trunc i32 [[SHR_26]] to i8
+; CHECK-NEXT:    [[CONV4:%.*]] = trunc nuw i32 [[SHR_26]] to i8
 ; CHECK-NEXT:    store i8 [[CONV4]], ptr [[ARRAYIDX2]], align 1
 ; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    [[TMP29:%.*]] = load i8, ptr [[ARRAYIDX8]], align 1
 ; CHECK-NEXT:    [[CONV9:%.*]] = zext i8 [[TMP29]] to i32
 ; CHECK-NEXT:    [[MUL10:%.*]] = mul nuw nsw i32 [[CONV9]], [[CONV]]
 ; CHECK-NEXT:    [[SHR11_27:%.*]] = lshr i32 [[MUL10]], 8
-; CHECK-NEXT:    [[CONV12:%.*]] = trunc i32 [[SHR11_27]] to i8
+; CHECK-NEXT:    [[CONV12:%.*]] = trunc nuw i32 [[SHR11_27]] to i8
 ; CHECK-NEXT:    store i8 [[CONV12]], ptr [[ARRAYIDX8]], align 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
@@ -172,8 +172,8 @@ define void @test_shrink_zext_in_preheader(ptr noalias %src, ptr noalias %dst, i
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul <16 x i16> [[TMP2]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = lshr <16 x i16> [[TMP5]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
 ; CHECK-NEXT:    [[TMP8:%.*]] = lshr <16 x i16> [[TMP6]], <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
-; CHECK-NEXT:    [[TMP9:%.*]] = trunc <16 x i16> [[TMP7]] to <16 x i8>
-; CHECK-NEXT:    [[TMP10:%.*]] = trunc <16 x i16> [[TMP8]] to <16 x i8>
+; CHECK-NEXT:    [[TMP9:%.*]] = trunc nuw <16 x i16> [[TMP7]] to <16 x i8>
+; CHECK-NEXT:    [[TMP10:%.*]] = trunc nuw <16 x i16> [[TMP8]] to <16 x i8>
 ; CHECK-NEXT:    [[TMP11:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP11]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i64 16
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/intrinsiccost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/intrinsiccost.ll
index b772a3814a64a..0f26092f510ca 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/intrinsiccost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/intrinsiccost.ll
@@ -23,7 +23,7 @@ define void @saddsat(ptr nocapture readonly %pSrc, i16 signext %offset, ptr noca
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 4294967280
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw i64 [[N_VEC]] to i32
 ; CHECK-NEXT:    [[IND_END:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[N_VEC]], 1
 ; CHECK-NEXT:    [[IND_END1:%.*]] = getelementptr i8, ptr [[PSRC:%.*]], i64 [[TMP1]]
@@ -138,7 +138,7 @@ define void @umin(ptr nocapture readonly %pSrc, i8 signext %offset, ptr nocaptur
 ; CHECK:       vec.epilog.iter.check:
 ; CHECK-NEXT:    [[IND_END18:%.*]] = getelementptr i8, ptr [[PDST]], i64 [[N_VEC]]
 ; CHECK-NEXT:    [[IND_END15:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[N_VEC]]
-; CHECK-NEXT:    [[DOTCAST11:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[DOTCAST11:%.*]] = trunc nuw i64 [[N_VEC]] to i32
 ; CHECK-NEXT:    [[IND_END12:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST11]]
 ; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = and i64 [[TMP0]], 24
 ; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp eq i64 [[N_VEC_REMAINING]], 0
@@ -146,7 +146,7 @@ define void @umin(ptr nocapture readonly %pSrc, i8 signext %offset, ptr nocaptur
 ; CHECK:       vec.epilog.ph:
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[N_VEC9:%.*]] = and i64 [[TMP0]], 4294967288
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[N_VEC9]] to i32
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw i64 [[N_VEC9]] to i32
 ; CHECK-NEXT:    [[IND_END10:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST]]
 ; CHECK-NEXT:    [[IND_END14:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[N_VEC9]]
 ; CHECK-NEXT:    [[IND_END17:%.*]] = getelementptr i8, ptr [[PDST]], i64 [[N_VEC9]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/intrinsiccost.ll b/llvm/test/Transforms/LoopVectorize/X86/intrinsiccost.ll
index 23b653bbda380..b5effe73fa73d 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/intrinsiccost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/intrinsiccost.ll
@@ -60,7 +60,7 @@ define void @uaddsat(ptr nocapture readonly %pSrc, i16 signext %offset, ptr noca
 ; CHECK-NEXT:    [[IND_END24:%.*]] = getelementptr i8, ptr [[PDST]], i64 [[TMP14]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = shl nuw nsw i64 [[N_VEC]], 1
 ; CHECK-NEXT:    [[IND_END21:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[TMP15]]
-; CHECK-NEXT:    [[DOTCAST17:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[DOTCAST17:%.*]] = trunc nuw i64 [[N_VEC]] to i32
 ; CHECK-NEXT:    [[IND_END18:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST17]]
 ; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = and i64 [[TMP0]], 56
 ; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp eq i64 [[N_VEC_REMAINING]], 0
@@ -68,7 +68,7 @@ define void @uaddsat(ptr nocapture readonly %pSrc, i16 signext %offset, ptr noca
 ; CHECK:       vec.epilog.ph:
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[N_VEC15:%.*]] = and i64 [[TMP0]], 4294967288
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[N_VEC15]] to i32
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw i64 [[N_VEC15]] to i32
 ; CHECK-NEXT:    [[IND_END16:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = shl nuw nsw i64 [[N_VEC15]], 1
 ; CHECK-NEXT:    [[IND_END20:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[TMP16]]
@@ -183,7 +183,7 @@ define void @fshl(ptr nocapture readonly %pSrc, i8 signext %offset, ptr nocaptur
 ; CHECK:       vec.epilog.iter.check:
 ; CHECK-NEXT:    [[IND_END24:%.*]] = getelementptr i8, ptr [[PDST]], i64 [[N_VEC]]
 ; CHECK-NEXT:    [[IND_END21:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[N_VEC]]
-; CHECK-NEXT:    [[DOTCAST17:%.*]] = trunc i64 [[N_VEC]] to i32
+; CHECK-NEXT:    [[DOTCAST17:%.*]] = trunc nuw i64 [[N_VEC]] to i32
 ; CHECK-NEXT:    [[IND_END18:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST17]]
 ; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = and i64 [[TMP0]], 112
 ; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp eq i64 [[N_VEC_REMAINING]], 0
@@ -191,7 +191,7 @@ define void @fshl(ptr nocapture readonly %pSrc, i8 signext %offset, ptr nocaptur
 ; CHECK:       vec.epilog.ph:
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[N_VEC15:%.*]] = and i64 [[TMP0]], 4294967280
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[N_VEC15]] to i32
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw i64 [[N_VEC15]] to i32
 ; CHECK-NEXT:    [[IND_END16:%.*]] = sub i32 [[BLOCKSIZE]], [[DOTCAST]]
 ; CHECK-NEXT:    [[IND_END20:%.*]] = getelementptr i8, ptr [[PSRC]], i64 [[N_VEC15]]
 ; CHECK-NEXT:    [[IND_END23:%.*]] = getelementptr i8, ptr [[PDST]], i64 [[N_VEC15]]
diff --git a/llvm/test/Transforms/LoopVectorize/reduction.ll b/llvm/test/Transforms/LoopVectorize/reduction.ll
index a47a38510eeeb..b66ce4047ad95 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction.ll
@@ -1205,7 +1205,7 @@ define i64 @reduction_with_phi_with_one_incoming_on_backedge(i16 %n, ptr %A) {
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP1]], 32764
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw nsw i32 [[N_VEC]] to i16
 ; CHECK-NEXT:    [[IND_END:%.*]] = or disjoint i16 [[DOTCAST]], 1
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
@@ -1283,7 +1283,7 @@ define i64 @reduction_with_phi_with_two_incoming_on_backedge(i16 %n, ptr %A) {
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_VEC:%.*]] = and i32 [[TMP1]], 32764
-; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc nuw nsw i32 [[N_VEC]] to i16
 ; CHECK-NEXT:    [[IND_END:%.*]] = or disjoint i16 [[DOTCAST]], 1
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
diff --git a/llvm/test/Transforms/PhaseOrdering/AArch64/quant_4x4.ll b/llvm/test/Transforms/PhaseOrdering/AArch64/quant_4x4.ll
index d18b207e87076..9206893cb2341 100644
--- a/llvm/test/Transforms/PhaseOrdering/AArch64/quant_4x4.ll
+++ b/llvm/test/Transforms/PhaseOrdering/AArch64/quant_4x4.ll
@@ -45,8 +45,8 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[TMP14:%.*]] = mul <8 x i32> [[TMP12]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = lshr <8 x i32> [[TMP13]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
 ; CHECK-NEXT:    [[TMP16:%.*]] = lshr <8 x i32> [[TMP14]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; CHECK-NEXT:    [[TMP17:%.*]] = trunc <8 x i32> [[TMP15]] to <8 x i16>
-; CHECK-NEXT:    [[TMP18:%.*]] = trunc <8 x i32> [[TMP16]] to <8 x i16>
+; CHECK-NEXT:    [[TMP17:%.*]] = trunc nuw <8 x i32> [[TMP15]] to <8 x i16>
+; CHECK-NEXT:    [[TMP18:%.*]] = trunc nuw <8 x i32> [[TMP16]] to <8 x i16>
 ; CHECK-NEXT:    [[TMP19:%.*]] = sub <8 x i16> zeroinitializer, [[TMP17]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = sub <8 x i16> zeroinitializer, [[TMP18]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = add nuw nsw <8 x i32> [[TMP6]], [[TMP1]]
@@ -55,8 +55,8 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[TMP24:%.*]] = mul <8 x i32> [[TMP22]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP25:%.*]] = lshr <8 x i32> [[TMP23]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
 ; CHECK-NEXT:    [[TMP26:%.*]] = lshr <8 x i32> [[TMP24]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
-; CHECK-NEXT:    [[TMP27:%.*]] = trunc <8 x i32> [[TMP25]] to <8 x i16>
-; CHECK-NEXT:    [[TMP28:%.*]] = trunc <8 x i32> [[TMP26]] to <8 x i16>
+; CHECK-NEXT:    [[TMP27:%.*]] = trunc nuw <8 x i32> [[TMP25]] to <8 x i16>
+; CHECK-NEXT:    [[TMP28:%.*]] = trunc nuw <8 x i32> [[TMP26]] to <8 x i16>
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <8 x i1> [[TMP3]], <8 x i16> [[TMP27]], <8 x i16> [[TMP19]]
 ; CHECK-NEXT:    [[PREDPHI34:%.*]] = select <8 x i1> [[TMP4]], <8 x i16> [[TMP28]], <8 x i16> [[TMP20]]
 ; CHECK-NEXT:    store <8 x i16> [[PREDPHI]], ptr [[DCT]], align 2, !alias.scope [[META0]], !noalias [[META3]]
@@ -83,13 +83,13 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV5]], [[CONV]]
 ; CHECK-NEXT:    [[MUL:%.*]] = mul i32 [[ADD]], [[CONV11]]
 ; CHECK-NEXT:    [[SHR:%.*]] = lshr i32 [[MUL]], 16
-; CHECK-NEXT:    [[CONV12:%.*]] = trunc i32 [[SHR]] to i16
+; CHECK-NEXT:    [[CONV12:%.*]] = trunc nuw i32 [[SHR]] to i16
 ; CHECK-NEXT:    br label [[IF_END:%.*]]
 ; CHECK:       if.else:
 ; CHECK-NEXT:    [[ADD21:%.*]] = sub nsw i32 [[CONV5]], [[CONV]]
 ; CHECK-NEXT:    [[MUL25:%.*]] = mul i32 [[ADD21]], [[CONV11]]
 ; CHECK-NEXT:    [[SHR26:%.*]] = lshr i32 [[MUL25]], 16
-; CHECK-NEXT:    [[TMP33:%.*]] = trunc i32 [[SHR26]] to i16
+; CHECK-NEXT:    [[TMP33:%.*]] = trunc nuw i32 [[SHR26]] to i16
 ; CHECK-NEXT:    [[CONV28:%.*]] = sub i16 0, [[TMP33]]
 ; CHECK-NEXT:    br label [[IF_END]]
 ; CHECK:       if.end:
@@ -110,14 +110,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_1:%.*]] = sub nsw i32 [[CONV5_1]], [[CONV_1]]
 ; CHECK-NEXT:    [[MUL25_1:%.*]] = mul i32 [[ADD21_1]], [[CONV11_1]]
 ; CHECK-NEXT:    [[SHR26_1:%.*]] = lshr i32 [[MUL25_1]], 16
-; CHECK-NEXT:    [[TMP37:%.*]] = trunc i32 [[SHR26_1]] to i16
+; CHECK-NEXT:    [[TMP37:%.*]] = trunc nuw i32 [[SHR26_1]] to i16
 ; CHECK-NEXT:    [[CONV28_1:%.*]] = sub i16 0, [[TMP37]]
 ; CHECK-NEXT:    br label [[IF_END_1:%.*]]
 ; CHECK:       if.then.1:
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add nuw nsw i32 [[CONV5_1]], [[CONV_1]]
 ; CHECK-NEXT:    [[MUL_1:%.*]] = mul i32 [[ADD_1]], [[CONV11_1]]
 ; CHECK-NEXT:    [[SHR_1:%.*]] = lshr i32 [[MUL_1]], 16
-; CHECK-NEXT:    [[CONV12_1:%.*]] = trunc i32 [[SHR_1]] to i16
+; CHECK-NEXT:    [[CONV12_1:%.*]] = trunc nuw i32 [[SHR_1]] to i16
 ; CHECK-NEXT:    br label [[IF_END_1]]
 ; CHECK:       if.end.1:
 ; CHECK-NEXT:    [[STOREMERGE_1:%.*]] = phi i16 [ [[CONV28_1]], [[IF_ELSE_1]] ], [ [[CONV12_1]], [[IF_THEN_1]] ]
@@ -138,14 +138,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_2:%.*]] = sub nsw i32 [[CONV5_2]], [[CONV_2]]
 ; CHECK-NEXT:    [[MUL25_2:%.*]] = mul i32 [[ADD21_2]], [[CONV11_2]]
 ; CHECK-NEXT:    [[SHR26_2:%.*]] = lshr i32 [[MUL25_2]], 16
-; CHECK-NEXT:    [[TMP41:%.*]] = trunc i32 [[SHR26_2]] to i16
+; CHECK-NEXT:    [[TMP41:%.*]] = trunc nuw i32 [[SHR26_2]] to i16
 ; CHECK-NEXT:    [[CONV28_2:%.*]] = sub i16 0, [[TMP41]]
 ; CHECK-NEXT:    br label [[IF_END_2:%.*]]
 ; CHECK:       if.then.2:
 ; CHECK-NEXT:    [[ADD_2:%.*]] = add nuw nsw i32 [[CONV5_2]], [[CONV_2]]
 ; CHECK-NEXT:    [[MUL_2:%.*]] = mul i32 [[ADD_2]], [[CONV11_2]]
 ; CHECK-NEXT:    [[SHR_2:%.*]] = lshr i32 [[MUL_2]], 16
-; CHECK-NEXT:    [[CONV12_2:%.*]] = trunc i32 [[SHR_2]] to i16
+; CHECK-NEXT:    [[CONV12_2:%.*]] = trunc nuw i32 [[SHR_2]] to i16
 ; CHECK-NEXT:    br label [[IF_END_2]]
 ; CHECK:       if.end.2:
 ; CHECK-NEXT:    [[STOREMERGE_2:%.*]] = phi i16 [ [[CONV28_2]], [[IF_ELSE_2]] ], [ [[CONV12_2]], [[IF_THEN_2]] ]
@@ -166,14 +166,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_3:%.*]] = sub nsw i32 [[CONV5_3]], [[CONV_3]]
 ; CHECK-NEXT:    [[MUL25_3:%.*]] = mul i32 [[ADD21_3]], [[CONV11_3]]
 ; CHECK-NEXT:    [[SHR26_3:%.*]] = lshr i32 [[MUL25_3]], 16
-; CHECK-NEXT:    [[TMP45:%.*]] = trunc i32 [[SHR26_3]] to i16
+; CHECK-NEXT:    [[TMP45:%.*]] = trunc nuw i32 [[SHR26_3]] to i16
 ; CHECK-NEXT:    [[CONV28_3:%.*]] = sub i16 0, [[TMP45]]
 ; CHECK-NEXT:    br label [[IF_END_3:%.*]]
 ; CHECK:       if.then.3:
 ; CHECK-NEXT:    [[ADD_3:%.*]] = add nuw nsw i32 [[CONV5_3]], [[CONV_3]]
 ; CHECK-NEXT:    [[MUL_3:%.*]] = mul i32 [[ADD_3]], [[CONV11_3]]
 ; CHECK-NEXT:    [[SHR_3:%.*]] = lshr i32 [[MUL_3]], 16
-; CHECK-NEXT:    [[CONV12_3:%.*]] = trunc i32 [[SHR_3]] to i16
+; CHECK-NEXT:    [[CONV12_3:%.*]] = trunc nuw i32 [[SHR_3]] to i16
 ; CHECK-NEXT:    br label [[IF_END_3]]
 ; CHECK:       if.end.3:
 ; CHECK-NEXT:    [[STOREMERGE_3:%.*]] = phi i16 [ [[CONV28_3]], [[IF_ELSE_3]] ], [ [[CONV12_3]], [[IF_THEN_3]] ]
@@ -194,14 +194,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_4:%.*]] = sub nsw i32 [[CONV5_4]], [[CONV_4]]
 ; CHECK-NEXT:    [[MUL25_4:%.*]] = mul i32 [[ADD21_4]], [[CONV11_4]]
 ; CHECK-NEXT:    [[SHR26_4:%.*]] = lshr i32 [[MUL25_4]], 16
-; CHECK-NEXT:    [[TMP49:%.*]] = trunc i32 [[SHR26_4]] to i16
+; CHECK-NEXT:    [[TMP49:%.*]] = trunc nuw i32 [[SHR26_4]] to i16
 ; CHECK-NEXT:    [[CONV28_4:%.*]] = sub i16 0, [[TMP49]]
 ; CHECK-NEXT:    br label [[IF_END_4:%.*]]
 ; CHECK:       if.then.4:
 ; CHECK-NEXT:    [[ADD_4:%.*]] = add nuw nsw i32 [[CONV5_4]], [[CONV_4]]
 ; CHECK-NEXT:    [[MUL_4:%.*]] = mul i32 [[ADD_4]], [[CONV11_4]]
 ; CHECK-NEXT:    [[SHR_4:%.*]] = lshr i32 [[MUL_4]], 16
-; CHECK-NEXT:    [[CONV12_4:%.*]] = trunc i32 [[SHR_4]] to i16
+; CHECK-NEXT:    [[CONV12_4:%.*]] = trunc nuw i32 [[SHR_4]] to i16
 ; CHECK-NEXT:    br label [[IF_END_4]]
 ; CHECK:       if.end.4:
 ; CHECK-NEXT:    [[STOREMERGE_4:%.*]] = phi i16 [ [[CONV28_4]], [[IF_ELSE_4]] ], [ [[CONV12_4]], [[IF_THEN_4]] ]
@@ -222,14 +222,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_5:%.*]] = sub nsw i32 [[CONV5_5]], [[CONV_5]]
 ; CHECK-NEXT:    [[MUL25_5:%.*]] = mul i32 [[ADD21_5]], [[CONV11_5]]
 ; CHECK-NEXT:    [[SHR26_5:%.*]] = lshr i32 [[MUL25_5]], 16
-; CHECK-NEXT:    [[TMP53:%.*]] = trunc i32 [[SHR26_5]] to i16
+; CHECK-NEXT:    [[TMP53:%.*]] = trunc nuw i32 [[SHR26_5]] to i16
 ; CHECK-NEXT:    [[CONV28_5:%.*]] = sub i16 0, [[TMP53]]
 ; CHECK-NEXT:    br label [[IF_END_5:%.*]]
 ; CHECK:       if.then.5:
 ; CHECK-NEXT:    [[ADD_5:%.*]] = add nuw nsw i32 [[CONV5_5]], [[CONV_5]]
 ; CHECK-NEXT:    [[MUL_5:%.*]] = mul i32 [[ADD_5]], [[CONV11_5]]
 ; CHECK-NEXT:    [[SHR_5:%.*]] = lshr i32 [[MUL_5]], 16
-; CHECK-NEXT:    [[CONV12_5:%.*]] = trunc i32 [[SHR_5]] to i16
+; CHECK-NEXT:    [[CONV12_5:%.*]] = trunc nuw i32 [[SHR_5]] to i16
 ; CHECK-NEXT:    br label [[IF_END_5]]
 ; CHECK:       if.end.5:
 ; CHECK-NEXT:    [[STOREMERGE_5:%.*]] = phi i16 [ [[CONV28_5]], [[IF_ELSE_5]] ], [ [[CONV12_5]], [[IF_THEN_5]] ]
@@ -250,14 +250,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_6:%.*]] = sub nsw i32 [[CONV5_6]], [[CONV_6]]
 ; CHECK-NEXT:    [[MUL25_6:%.*]] = mul i32 [[ADD21_6]], [[CONV11_6]]
 ; CHECK-NEXT:    [[SHR26_6:%.*]] = lshr i32 [[MUL25_6]], 16
-; CHECK-NEXT:    [[TMP57:%.*]] = trunc i32 [[SHR26_6]] to i16
+; CHECK-NEXT:    [[TMP57:%.*]] = trunc nuw i32 [[SHR26_6]] to i16
 ; CHECK-NEXT:    [[CONV28_6:%.*]] = sub i16 0, [[TMP57]]
 ; CHECK-NEXT:    br label [[IF_END_6:%.*]]
 ; CHECK:       if.then.6:
 ; CHECK-NEXT:    [[ADD_6:%.*]] = add nuw nsw i32 [[CONV5_6]], [[CONV_6]]
 ; CHECK-NEXT:    [[MUL_6:%.*]] = mul i32 [[ADD_6]], [[CONV11_6]]
 ; CHECK-NEXT:    [[SHR_6:%.*]] = lshr i32 [[MUL_6]], 16
-; CHECK-NEXT:    [[CONV12_6:%.*]] = trunc i32 [[SHR_6]] to i16
+; CHECK-NEXT:    [[CONV12_6:%.*]] = trunc nuw i32 [[SHR_6]] to i16
 ; CHECK-NEXT:    br label [[IF_END_6]]
 ; CHECK:       if.end.6:
 ; CHECK-NEXT:    [[STOREMERGE_6:%.*]] = phi i16 [ [[CONV28_6]], [[IF_ELSE_6]] ], [ [[CONV12_6]], [[IF_THEN_6]] ]
@@ -278,14 +278,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_7:%.*]] = sub nsw i32 [[CONV5_7]], [[CONV_7]]
 ; CHECK-NEXT:    [[MUL25_7:%.*]] = mul i32 [[ADD21_7]], [[CONV11_7]]
 ; CHECK-NEXT:    [[SHR26_7:%.*]] = lshr i32 [[MUL25_7]], 16
-; CHECK-NEXT:    [[TMP61:%.*]] = trunc i32 [[SHR26_7]] to i16
+; CHECK-NEXT:    [[TMP61:%.*]] = trunc nuw i32 [[SHR26_7]] to i16
 ; CHECK-NEXT:    [[CONV28_7:%.*]] = sub i16 0, [[TMP61]]
 ; CHECK-NEXT:    br label [[IF_END_7:%.*]]
 ; CHECK:       if.then.7:
 ; CHECK-NEXT:    [[ADD_7:%.*]] = add nuw nsw i32 [[CONV5_7]], [[CONV_7]]
 ; CHECK-NEXT:    [[MUL_7:%.*]] = mul i32 [[ADD_7]], [[CONV11_7]]
 ; CHECK-NEXT:    [[SHR_7:%.*]] = lshr i32 [[MUL_7]], 16
-; CHECK-NEXT:    [[CONV12_7:%.*]] = trunc i32 [[SHR_7]] to i16
+; CHECK-NEXT:    [[CONV12_7:%.*]] = trunc nuw i32 [[SHR_7]] to i16
 ; CHECK-NEXT:    br label [[IF_END_7]]
 ; CHECK:       if.end.7:
 ; CHECK-NEXT:    [[STOREMERGE_7:%.*]] = phi i16 [ [[CONV28_7]], [[IF_ELSE_7]] ], [ [[CONV12_7]], [[IF_THEN_7]] ]
@@ -306,14 +306,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_8:%.*]] = sub nsw i32 [[CONV5_8]], [[CONV_8]]
 ; CHECK-NEXT:    [[MUL25_8:%.*]] = mul i32 [[ADD21_8]], [[CONV11_8]]
 ; CHECK-NEXT:    [[SHR26_8:%.*]] = lshr i32 [[MUL25_8]], 16
-; CHECK-NEXT:    [[TMP65:%.*]] = trunc i32 [[SHR26_8]] to i16
+; CHECK-NEXT:    [[TMP65:%.*]] = trunc nuw i32 [[SHR26_8]] to i16
 ; CHECK-NEXT:    [[CONV28_8:%.*]] = sub i16 0, [[TMP65]]
 ; CHECK-NEXT:    br label [[IF_END_8:%.*]]
 ; CHECK:       if.then.8:
 ; CHECK-NEXT:    [[ADD_8:%.*]] = add nuw nsw i32 [[CONV5_8]], [[CONV_8]]
 ; CHECK-NEXT:    [[MUL_8:%.*]] = mul i32 [[ADD_8]], [[CONV11_8]]
 ; CHECK-NEXT:    [[SHR_8:%.*]] = lshr i32 [[MUL_8]], 16
-; CHECK-NEXT:    [[CONV12_8:%.*]] = trunc i32 [[SHR_8]] to i16
+; CHECK-NEXT:    [[CONV12_8:%.*]] = trunc nuw i32 [[SHR_8]] to i16
 ; CHECK-NEXT:    br label [[IF_END_8]]
 ; CHECK:       if.end.8:
 ; CHECK-NEXT:    [[STOREMERGE_8:%.*]] = phi i16 [ [[CONV28_8]], [[IF_ELSE_8]] ], [ [[CONV12_8]], [[IF_THEN_8]] ]
@@ -334,14 +334,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_9:%.*]] = sub nsw i32 [[CONV5_9]], [[CONV_9]]
 ; CHECK-NEXT:    [[MUL25_9:%.*]] = mul i32 [[ADD21_9]], [[CONV11_9]]
 ; CHECK-NEXT:    [[SHR26_9:%.*]] = lshr i32 [[MUL25_9]], 16
-; CHECK-NEXT:    [[TMP69:%.*]] = trunc i32 [[SHR26_9]] to i16
+; CHECK-NEXT:    [[TMP69:%.*]] = trunc nuw i32 [[SHR26_9]] to i16
 ; CHECK-NEXT:    [[CONV28_9:%.*]] = sub i16 0, [[TMP69]]
 ; CHECK-NEXT:    br label [[IF_END_9:%.*]]
 ; CHECK:       if.then.9:
 ; CHECK-NEXT:    [[ADD_9:%.*]] = add nuw nsw i32 [[CONV5_9]], [[CONV_9]]
 ; CHECK-NEXT:    [[MUL_9:%.*]] = mul i32 [[ADD_9]], [[CONV11_9]]
 ; CHECK-NEXT:    [[SHR_9:%.*]] = lshr i32 [[MUL_9]], 16
-; CHECK-NEXT:    [[CONV12_9:%.*]] = trunc i32 [[SHR_9]] to i16
+; CHECK-NEXT:    [[CONV12_9:%.*]] = trunc nuw i32 [[SHR_9]] to i16
 ; CHECK-NEXT:    br label [[IF_END_9]]
 ; CHECK:       if.end.9:
 ; CHECK-NEXT:    [[STOREMERGE_9:%.*]] = phi i16 [ [[CONV28_9]], [[IF_ELSE_9]] ], [ [[CONV12_9]], [[IF_THEN_9]] ]
@@ -362,14 +362,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_10:%.*]] = sub nsw i32 [[CONV5_10]], [[CONV_10]]
 ; CHECK-NEXT:    [[MUL25_10:%.*]] = mul i32 [[ADD21_10]], [[CONV11_10]]
 ; CHECK-NEXT:    [[SHR26_10:%.*]] = lshr i32 [[MUL25_10]], 16
-; CHECK-NEXT:    [[TMP73:%.*]] = trunc i32 [[SHR26_10]] to i16
+; CHECK-NEXT:    [[TMP73:%.*]] = trunc nuw i32 [[SHR26_10]] to i16
 ; CHECK-NEXT:    [[CONV28_10:%.*]] = sub i16 0, [[TMP73]]
 ; CHECK-NEXT:    br label [[IF_END_10:%.*]]
 ; CHECK:       if.then.10:
 ; CHECK-NEXT:    [[ADD_10:%.*]] = add nuw nsw i32 [[CONV5_10]], [[CONV_10]]
 ; CHECK-NEXT:    [[MUL_10:%.*]] = mul i32 [[ADD_10]], [[CONV11_10]]
 ; CHECK-NEXT:    [[SHR_10:%.*]] = lshr i32 [[MUL_10]], 16
-; CHECK-NEXT:    [[CONV12_10:%.*]] = trunc i32 [[SHR_10]] to i16
+; CHECK-NEXT:    [[CONV12_10:%.*]] = trunc nuw i32 [[SHR_10]] to i16
 ; CHECK-NEXT:    br label [[IF_END_10]]
 ; CHECK:       if.end.10:
 ; CHECK-NEXT:    [[STOREMERGE_10:%.*]] = phi i16 [ [[CONV28_10]], [[IF_ELSE_10]] ], [ [[CONV12_10]], [[IF_THEN_10]] ]
@@ -390,14 +390,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_11:%.*]] = sub nsw i32 [[CONV5_11]], [[CONV_11]]
 ; CHECK-NEXT:    [[MUL25_11:%.*]] = mul i32 [[ADD21_11]], [[CONV11_11]]
 ; CHECK-NEXT:    [[SHR26_11:%.*]] = lshr i32 [[MUL25_11]], 16
-; CHECK-NEXT:    [[TMP77:%.*]] = trunc i32 [[SHR26_11]] to i16
+; CHECK-NEXT:    [[TMP77:%.*]] = trunc nuw i32 [[SHR26_11]] to i16
 ; CHECK-NEXT:    [[CONV28_11:%.*]] = sub i16 0, [[TMP77]]
 ; CHECK-NEXT:    br label [[IF_END_11:%.*]]
 ; CHECK:       if.then.11:
 ; CHECK-NEXT:    [[ADD_11:%.*]] = add nuw nsw i32 [[CONV5_11]], [[CONV_11]]
 ; CHECK-NEXT:    [[MUL_11:%.*]] = mul i32 [[ADD_11]], [[CONV11_11]]
 ; CHECK-NEXT:    [[SHR_11:%.*]] = lshr i32 [[MUL_11]], 16
-; CHECK-NEXT:    [[CONV12_11:%.*]] = trunc i32 [[SHR_11]] to i16
+; CHECK-NEXT:    [[CONV12_11:%.*]] = trunc nuw i32 [[SHR_11]] to i16
 ; CHECK-NEXT:    br label [[IF_END_11]]
 ; CHECK:       if.end.11:
 ; CHECK-NEXT:    [[STOREMERGE_11:%.*]] = phi i16 [ [[CONV28_11]], [[IF_ELSE_11]] ], [ [[CONV12_11]], [[IF_THEN_11]] ]
@@ -418,14 +418,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_12:%.*]] = sub nsw i32 [[CONV5_12]], [[CONV_12]]
 ; CHECK-NEXT:    [[MUL25_12:%.*]] = mul i32 [[ADD21_12]], [[CONV11_12]]
 ; CHECK-NEXT:    [[SHR26_12:%.*]] = lshr i32 [[MUL25_12]], 16
-; CHECK-NEXT:    [[TMP81:%.*]] = trunc i32 [[SHR26_12]] to i16
+; CHECK-NEXT:    [[TMP81:%.*]] = trunc nuw i32 [[SHR26_12]] to i16
 ; CHECK-NEXT:    [[CONV28_12:%.*]] = sub i16 0, [[TMP81]]
 ; CHECK-NEXT:    br label [[IF_END_12:%.*]]
 ; CHECK:       if.then.12:
 ; CHECK-NEXT:    [[ADD_12:%.*]] = add nuw nsw i32 [[CONV5_12]], [[CONV_12]]
 ; CHECK-NEXT:    [[MUL_12:%.*]] = mul i32 [[ADD_12]], [[CONV11_12]]
 ; CHECK-NEXT:    [[SHR_12:%.*]] = lshr i32 [[MUL_12]], 16
-; CHECK-NEXT:    [[CONV12_12:%.*]] = trunc i32 [[SHR_12]] to i16
+; CHECK-NEXT:    [[CONV12_12:%.*]] = trunc nuw i32 [[SHR_12]] to i16
 ; CHECK-NEXT:    br label [[IF_END_12]]
 ; CHECK:       if.end.12:
 ; CHECK-NEXT:    [[STOREMERGE_12:%.*]] = phi i16 [ [[CONV28_12]], [[IF_ELSE_12]] ], [ [[CONV12_12]], [[IF_THEN_12]] ]
@@ -446,14 +446,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_13:%.*]] = sub nsw i32 [[CONV5_13]], [[CONV_13]]
 ; CHECK-NEXT:    [[MUL25_13:%.*]] = mul i32 [[ADD21_13]], [[CONV11_13]]
 ; CHECK-NEXT:    [[SHR26_13:%.*]] = lshr i32 [[MUL25_13]], 16
-; CHECK-NEXT:    [[TMP85:%.*]] = trunc i32 [[SHR26_13]] to i16
+; CHECK-NEXT:    [[TMP85:%.*]] = trunc nuw i32 [[SHR26_13]] to i16
 ; CHECK-NEXT:    [[CONV28_13:%.*]] = sub i16 0, [[TMP85]]
 ; CHECK-NEXT:    br label [[IF_END_13:%.*]]
 ; CHECK:       if.then.13:
 ; CHECK-NEXT:    [[ADD_13:%.*]] = add nuw nsw i32 [[CONV5_13]], [[CONV_13]]
 ; CHECK-NEXT:    [[MUL_13:%.*]] = mul i32 [[ADD_13]], [[CONV11_13]]
 ; CHECK-NEXT:    [[SHR_13:%.*]] = lshr i32 [[MUL_13]], 16
-; CHECK-NEXT:    [[CONV12_13:%.*]] = trunc i32 [[SHR_13]] to i16
+; CHECK-NEXT:    [[CONV12_13:%.*]] = trunc nuw i32 [[SHR_13]] to i16
 ; CHECK-NEXT:    br label [[IF_END_13]]
 ; CHECK:       if.end.13:
 ; CHECK-NEXT:    [[STOREMERGE_13:%.*]] = phi i16 [ [[CONV28_13]], [[IF_ELSE_13]] ], [ [[CONV12_13]], [[IF_THEN_13]] ]
@@ -474,14 +474,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_14:%.*]] = sub nsw i32 [[CONV5_14]], [[CONV_14]]
 ; CHECK-NEXT:    [[MUL25_14:%.*]] = mul i32 [[ADD21_14]], [[CONV11_14]]
 ; CHECK-NEXT:    [[SHR26_14:%.*]] = lshr i32 [[MUL25_14]], 16
-; CHECK-NEXT:    [[TMP89:%.*]] = trunc i32 [[SHR26_14]] to i16
+; CHECK-NEXT:    [[TMP89:%.*]] = trunc nuw i32 [[SHR26_14]] to i16
 ; CHECK-NEXT:    [[CONV28_14:%.*]] = sub i16 0, [[TMP89]]
 ; CHECK-NEXT:    br label [[IF_END_14:%.*]]
 ; CHECK:       if.then.14:
 ; CHECK-NEXT:    [[ADD_14:%.*]] = add nuw nsw i32 [[CONV5_14]], [[CONV_14]]
 ; CHECK-NEXT:    [[MUL_14:%.*]] = mul i32 [[ADD_14]], [[CONV11_14]]
 ; CHECK-NEXT:    [[SHR_14:%.*]] = lshr i32 [[MUL_14]], 16
-; CHECK-NEXT:    [[CONV12_14:%.*]] = trunc i32 [[SHR_14]] to i16
+; CHECK-NEXT:    [[CONV12_14:%.*]] = trunc nuw i32 [[SHR_14]] to i16
 ; CHECK-NEXT:    br label [[IF_END_14]]
 ; CHECK:       if.end.14:
 ; CHECK-NEXT:    [[STOREMERGE_14:%.*]] = phi i16 [ [[CONV28_14]], [[IF_ELSE_14]] ], [ [[CONV12_14]], [[IF_THEN_14]] ]
@@ -502,14 +502,14 @@ define i32 @quant_4x4(ptr noundef %dct, ptr noundef %mf, ptr noundef %bias) {
 ; CHECK-NEXT:    [[ADD21_15:%.*]] = sub nsw i32 [[CONV5_15]], [[CONV_15]]
 ; CHECK-NEXT:    [[MUL25_15:%.*]] = mul i32 [[ADD21_15]], [[CONV11_15]]
 ; CHECK-NEXT:    [[SHR26_15:%.*]] = lshr i32 [[MUL25_15]], 16
-; CHECK-NEXT:    [[TMP93:%.*]] = trunc i32 [[SHR26_15]] to i16
+; CHECK-NEXT:    [[TMP93:%.*]] = trunc nuw i32 [[SHR26_15]] to i16
 ; CHECK-NEXT:    [[CONV28_15:%.*]] = sub i16 0, [[TMP93]]
 ; CHECK-NEXT:    br label [[IF_END_15]]
 ; CHECK:       if.then.15:
 ; CHECK-NEXT:    [[ADD_15:%.*]] = add nuw nsw i32 [[CONV5_15]], [[CONV_15]]
 ; CHECK-NEXT:    [[MUL_15:%.*]] = mul i32 [[ADD_15]], [[CONV11_15]]
 ; CHECK-NEXT:    [[SHR_15:%.*]] = lshr i32 [[MUL_15]], 16
-; CHECK-NEXT:    [[CONV12_15:%.*]] = trunc i32 [[SHR_15]] to i16
+; CHECK-NEXT:    [[CONV12_15:%.*]] = trunc nuw i32 [[SHR_15]] to i16
 ; CHECK-NEXT:    br label [[IF_END_15]]
 ; CHECK:       if.end.15:
 ; CHECK-NEXT:    [[STOREMERGE_15:%.*]] = phi i16 [ [[CONV28_15]], [[IF_ELSE_15]] ], [ [[CONV12_15]], [[IF_THEN_15]] ]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
index c38f2748a9763..75505f632a43f 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
@@ -55,7 +55,7 @@ define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = mul <4 x i32> [[TMP7]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = lshr <4 x i32> [[TMP8]], <i32 15, i32 15, i32 15, i32 15>
 ; CHECK-NEXT:    [[TMP10:%.*]] = call <4 x i32> @llvm.umin.v4i32(<4 x i32> [[TMP9]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>)
-; CHECK-NEXT:    [[TMP11:%.*]] = trunc <4 x i32> [[TMP10]] to <4 x i8>
+; CHECK-NEXT:    [[TMP11:%.*]] = trunc nuw <4 x i32> [[TMP10]] to <4 x i8>
 ; CHECK-NEXT:    store <4 x i8> [[TMP11]], ptr [[TMP0]], align 1, !tbaa [[TBAA4]]
 ; CHECK-NEXT:    ret void
 ;

From 496de32ee2c34880c7d3396bbd09e45d5d5c8a9e Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97@outlook.com>
Date: Thu, 11 Apr 2024 19:13:06 +0800
Subject: [PATCH 137/886] [NewPM] Remove `MachinePassInfoMixin` (#88243)

Unify the inheritance paths of IR and machine function.
---
 .../llvm/CodeGen/DeadMachineInstructionElim.h |   2 +-
 .../llvm/CodeGen/FreeMachineFunction.h        |   3 +-
 llvm/include/llvm/CodeGen/MIRPrinter.h        |   4 +-
 .../include/llvm/CodeGen/MachinePassManager.h | 108 +++++++-----------
 llvm/include/llvm/Passes/CodeGenPassBuilder.h |   4 +-
 llvm/include/llvm/Passes/PassBuilder.h        |   3 +-
 llvm/lib/Passes/PassBuilder.cpp               |   2 +-
 .../MIR/PassBuilderCallbacksTest.cpp          |   2 +-
 8 files changed, 49 insertions(+), 79 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/DeadMachineInstructionElim.h b/llvm/include/llvm/CodeGen/DeadMachineInstructionElim.h
index b9fe7cfccf9a3..56cfa1e087181 100644
--- a/llvm/include/llvm/CodeGen/DeadMachineInstructionElim.h
+++ b/llvm/include/llvm/CodeGen/DeadMachineInstructionElim.h
@@ -14,7 +14,7 @@
 namespace llvm {
 
 class DeadMachineInstructionElimPass
-    : public MachinePassInfoMixin<DeadMachineInstructionElimPass> {
+    : public PassInfoMixin<DeadMachineInstructionElimPass> {
 public:
   PreservedAnalyses run(MachineFunction &MF,
                         MachineFunctionAnalysisManager &MFAM);
diff --git a/llvm/include/llvm/CodeGen/FreeMachineFunction.h b/llvm/include/llvm/CodeGen/FreeMachineFunction.h
index 77b76c591201a..5f21c6720350b 100644
--- a/llvm/include/llvm/CodeGen/FreeMachineFunction.h
+++ b/llvm/include/llvm/CodeGen/FreeMachineFunction.h
@@ -13,8 +13,7 @@
 
 namespace llvm {
 
-class FreeMachineFunctionPass
-    : public MachinePassInfoMixin<FreeMachineFunctionPass> {
+class FreeMachineFunctionPass : public PassInfoMixin<FreeMachineFunctionPass> {
 public:
   PreservedAnalyses run(MachineFunction &MF,
                         MachineFunctionAnalysisManager &MFAM);
diff --git a/llvm/include/llvm/CodeGen/MIRPrinter.h b/llvm/include/llvm/CodeGen/MIRPrinter.h
index daa0d7e2691f1..d0a11e1c4a2fd 100644
--- a/llvm/include/llvm/CodeGen/MIRPrinter.h
+++ b/llvm/include/llvm/CodeGen/MIRPrinter.h
@@ -24,7 +24,7 @@ class MachineFunction;
 class Module;
 template <typename T> class SmallVectorImpl;
 
-class PrintMIRPreparePass : public MachinePassInfoMixin<PrintMIRPreparePass> {
+class PrintMIRPreparePass : public PassInfoMixin<PrintMIRPreparePass> {
   raw_ostream &OS;
 
 public:
@@ -32,7 +32,7 @@ class PrintMIRPreparePass : public MachinePassInfoMixin<PrintMIRPreparePass> {
   PreservedAnalyses run(Module &M, ModuleAnalysisManager &MFAM);
 };
 
-class PrintMIRPass : public MachinePassInfoMixin<PrintMIRPass> {
+class PrintMIRPass : public PassInfoMixin<PrintMIRPass> {
   raw_ostream &OS;
 
 public:
diff --git a/llvm/include/llvm/CodeGen/MachinePassManager.h b/llvm/include/llvm/CodeGen/MachinePassManager.h
index 8689fd19030f9..4f0b6ba2b1e73 100644
--- a/llvm/include/llvm/CodeGen/MachinePassManager.h
+++ b/llvm/include/llvm/CodeGen/MachinePassManager.h
@@ -36,65 +36,6 @@ class MachineFunction;
 extern template class AnalysisManager<MachineFunction>;
 using MachineFunctionAnalysisManager = AnalysisManager<MachineFunction>;
 
-/// A CRTP mix-in that provides informational APIs needed for machine passes.
-///
-/// This provides some boilerplate for types that are machine passes. It
-/// automatically mixes in \c PassInfoMixin.
-template <typename DerivedT>
-struct MachinePassInfoMixin : public PassInfoMixin<DerivedT> {
-protected:
-  class PropertyChanger {
-    MachineFunction &MF;
-
-    template <typename T>
-    using has_get_required_properties_t =
-        decltype(std::declval<T &>().getRequiredProperties());
-
-    template <typename T>
-    using has_get_set_properties_t =
-        decltype(std::declval<T &>().getSetProperties());
-
-    template <typename T>
-    using has_get_cleared_properties_t =
-        decltype(std::declval<T &>().getClearedProperties());
-
-  public:
-    PropertyChanger(MachineFunction &MF) : MF(MF) {
-#ifndef NDEBUG
-      if constexpr (is_detected<has_get_required_properties_t,
-                                DerivedT>::value) {
-        auto &MFProps = MF.getProperties();
-        auto RequiredProperties = DerivedT::getRequiredProperties();
-        if (!MFProps.verifyRequiredProperties(RequiredProperties)) {
-          errs() << "MachineFunctionProperties required by " << DerivedT::name()
-                 << " pass are not met by function " << MF.getName() << ".\n"
-                 << "Required properties: ";
-          RequiredProperties.print(errs());
-          errs() << "\nCurrent properties: ";
-          MFProps.print(errs());
-          errs() << '\n';
-          report_fatal_error("MachineFunctionProperties check failed");
-        }
-      }
-#endif
-    }
-
-    ~PropertyChanger() {
-      if constexpr (is_detected<has_get_set_properties_t, DerivedT>::value)
-        MF.getProperties().set(DerivedT::getSetProperties());
-      if constexpr (is_detected<has_get_cleared_properties_t, DerivedT>::value)
-        MF.getProperties().reset(DerivedT::getClearedProperties());
-    }
-  };
-
-public:
-  PreservedAnalyses runImpl(MachineFunction &MF,
-                            MachineFunctionAnalysisManager &MFAM) {
-    PropertyChanger PC(MF);
-    return static_cast<DerivedT *>(this)->run(MF, MFAM);
-  }
-};
-
 namespace detail {
 
 template <typename PassT>
@@ -117,8 +58,44 @@ struct MachinePassModel
   MachinePassModel &operator=(const MachinePassModel &) = delete;
   PreservedAnalyses run(MachineFunction &IR,
                         MachineFunctionAnalysisManager &AM) override {
-    return this->Pass.runImpl(IR, AM);
+#ifndef NDEBUG
+    if constexpr (is_detected<has_get_required_properties_t, PassT>::value) {
+      auto &MFProps = IR.getProperties();
+      auto RequiredProperties = PassT::getRequiredProperties();
+      if (!MFProps.verifyRequiredProperties(RequiredProperties)) {
+        errs() << "MachineFunctionProperties required by " << PassT::name()
+               << " pass are not met by function " << IR.getName() << ".\n"
+               << "Required properties: ";
+        RequiredProperties.print(errs());
+        errs() << "\nCurrent properties: ";
+        MFProps.print(errs());
+        errs() << '\n';
+        report_fatal_error("MachineFunctionProperties check failed");
+      }
+    }
+#endif
+
+    auto PA = this->Pass.run(IR, AM);
+
+    if constexpr (is_detected<has_get_set_properties_t, PassT>::value)
+      IR.getProperties().set(PassT::getSetProperties());
+    if constexpr (is_detected<has_get_cleared_properties_t, PassT>::value)
+      IR.getProperties().reset(PassT::getClearedProperties());
+    return PA;
   }
+
+private:
+  template <typename T>
+  using has_get_required_properties_t =
+      decltype(std::declval<T &>().getRequiredProperties());
+
+  template <typename T>
+  using has_get_set_properties_t =
+      decltype(std::declval<T &>().getSetProperties());
+
+  template <typename T>
+  using has_get_cleared_properties_t =
+      decltype(std::declval<T &>().getClearedProperties());
 };
 } // namespace detail
 
@@ -246,20 +223,15 @@ createModuleToMachineFunctionPassAdaptor(MachineFunctionPassT &&Pass) {
 template <>
 template <typename PassT>
 void PassManager<MachineFunction>::addPass(PassT &&Pass) {
-  using PassModelT =
-      detail::PassModel<MachineFunction, PassT, MachineFunctionAnalysisManager>;
   using MachinePassModelT = detail::MachinePassModel<PassT>;
   // Do not use make_unique or emplace_back, they cause too many template
   // instantiations, causing terrible compile times.
-  if constexpr (std::is_base_of_v<MachinePassInfoMixin<PassT>, PassT>) {
-    Passes.push_back(std::unique_ptr<PassConceptT>(
-        new MachinePassModelT(std::forward<PassT>(Pass))));
-  } else if constexpr (std::is_same_v<PassT, PassManager<MachineFunction>>) {
+  if constexpr (std::is_same_v<PassT, PassManager<MachineFunction>>) {
     for (auto &P : Pass.Passes)
       Passes.push_back(std::move(P));
   } else {
-    Passes.push_back(std::unique_ptr<PassConceptT>(
-        new PassModelT(std::forward<PassT>(Pass))));
+    Passes.push_back(std::unique_ptr<MachinePassModelT>(
+        new MachinePassModelT(std::forward<PassT>(Pass))));
   }
 }
 
diff --git a/llvm/include/llvm/Passes/CodeGenPassBuilder.h b/llvm/include/llvm/Passes/CodeGenPassBuilder.h
index ab231dbbb6a68..ed5d8affa7938 100644
--- a/llvm/include/llvm/Passes/CodeGenPassBuilder.h
+++ b/llvm/include/llvm/Passes/CodeGenPassBuilder.h
@@ -87,14 +87,14 @@ namespace llvm {
     }                                                                          \
   };
 #define DUMMY_MACHINE_MODULE_PASS(NAME, PASS_NAME)                             \
-  struct PASS_NAME : public MachinePassInfoMixin<PASS_NAME> {                  \
+  struct PASS_NAME : public PassInfoMixin<PASS_NAME> {                         \
     template <typename... Ts> PASS_NAME(Ts &&...) {}                           \
     PreservedAnalyses run(Module &, ModuleAnalysisManager &) {                 \
       return PreservedAnalyses::all();                                         \
     }                                                                          \
   };
 #define DUMMY_MACHINE_FUNCTION_PASS(NAME, PASS_NAME)                           \
-  struct PASS_NAME : public MachinePassInfoMixin<PASS_NAME> {                  \
+  struct PASS_NAME : public PassInfoMixin<PASS_NAME> {                         \
     template <typename... Ts> PASS_NAME(Ts &&...) {}                           \
     PreservedAnalyses run(MachineFunction &,                                   \
                           MachineFunctionAnalysisManager &) {                  \
diff --git a/llvm/include/llvm/Passes/PassBuilder.h b/llvm/include/llvm/Passes/PassBuilder.h
index d1232124d5d81..c8f643452bb15 100644
--- a/llvm/include/llvm/Passes/PassBuilder.h
+++ b/llvm/include/llvm/Passes/PassBuilder.h
@@ -909,8 +909,7 @@ struct NoOpLoopPass : PassInfoMixin<NoOpLoopPass> {
 };
 
 /// No-op machine function pass which does nothing.
-struct NoOpMachineFunctionPass
-    : public MachinePassInfoMixin<NoOpMachineFunctionPass> {
+struct NoOpMachineFunctionPass : public PassInfoMixin<NoOpMachineFunctionPass> {
   PreservedAnalyses run(MachineFunction &, MachineFunctionAnalysisManager &) {
     return PreservedAnalyses::all();
   }
diff --git a/llvm/lib/Passes/PassBuilder.cpp b/llvm/lib/Passes/PassBuilder.cpp
index 832ee352205a0..8d408ca2363a9 100644
--- a/llvm/lib/Passes/PassBuilder.cpp
+++ b/llvm/lib/Passes/PassBuilder.cpp
@@ -368,7 +368,7 @@ class TriggerVerifierErrorPass
 // A pass requires all MachineFunctionProperties.
 // DO NOT USE THIS EXCEPT FOR TESTING!
 class RequireAllMachineFunctionPropertiesPass
-    : public MachinePassInfoMixin<RequireAllMachineFunctionPropertiesPass> {
+    : public PassInfoMixin<RequireAllMachineFunctionPropertiesPass> {
 public:
   PreservedAnalyses run(MachineFunction &, MachineFunctionAnalysisManager &) {
     return PreservedAnalyses::none();
diff --git a/llvm/unittests/MIR/PassBuilderCallbacksTest.cpp b/llvm/unittests/MIR/PassBuilderCallbacksTest.cpp
index 8e3738dc91920..6fd4e54a929f4 100644
--- a/llvm/unittests/MIR/PassBuilderCallbacksTest.cpp
+++ b/llvm/unittests/MIR/PassBuilderCallbacksTest.cpp
@@ -233,7 +233,7 @@ template <typename DerivedT> class MockAnalysisHandleBase {
 
 template <typename DerivedT> class MockPassHandleBase {
 public:
-  class Pass : public MachinePassInfoMixin<Pass> {
+  class Pass : public PassInfoMixin<Pass> {
     friend MockPassHandleBase;
 
     DerivedT *Handle;

From a6db20f2c39ecb5939890317068d5398c760746d Mon Sep 17 00:00:00 2001
From: Marc Auberer <marc.auberer@chillibits.com>
Date: Thu, 11 Apr 2024 14:15:16 +0200
Subject: [PATCH 138/886] [libcxx] Use generic builtins for popcount, clz and
 ctz (#86563)

Fixes #86556

Use `__builtin_popcountg` instead of `__buildin_popcount{l|ll}`
Use `__builtin_clzg instead` of `__buildin_clz{l|ll}`
Use `__builtin_ctzg instead` of `__builtin_ctz{l|ll}`

The generic variant of the builtins can be used to simplify some logic
with >= Clang 19 or >= GCC 14, where these generic variants are
available. As for backwards compatibility reasons, we can't completely
remove the old logic. Therefore, I left ToDo comments to address this,
as soon as support for pre Clang 19 as well as pre GCC 14 is dropped.

---------

Co-authored-by: Nick Desaulniers <nickdesaulniers@users.noreply.github.com>
---
 libcxx/include/__bit/countl.h   | 11 +++++++++++
 libcxx/include/__bit/countr.h   |  8 +++++++-
 libcxx/include/__bit/popcount.h |  7 +++++++
 3 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/libcxx/include/__bit/countl.h b/libcxx/include/__bit/countl.h
index 396cfc2c3f406..13df8d4e66c40 100644
--- a/libcxx/include/__bit/countl.h
+++ b/libcxx/include/__bit/countl.h
@@ -6,6 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
+// TODO: __builtin_clzg is available since Clang 19 and GCC 14. When support for older versions is dropped, we can
+//  refactor this code to exclusively use __builtin_clzg.
+
 #ifndef _LIBCPP___BIT_COUNTL_H
 #define _LIBCPP___BIT_COUNTL_H
 
@@ -38,6 +41,9 @@ _LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR int __libcpp_cl
 
 #ifndef _LIBCPP_HAS_NO_INT128
 inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR int __libcpp_clz(__uint128_t __x) _NOEXCEPT {
+#  if __has_builtin(__builtin_clzg)
+  return __builtin_clzg(__x);
+#  else
   // The function is written in this form due to C++ constexpr limitations.
   // The algorithm:
   // - Test whether any bit in the high 64-bits is set
@@ -49,12 +55,16 @@ inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR int __libcpp_clz(__uint128_t __x)
   //     zeros in the high 64-bits.
   return ((__x >> 64) == 0) ? (64 + __builtin_clzll(static_cast<unsigned long long>(__x)))
                             : __builtin_clzll(static_cast<unsigned long long>(__x >> 64));
+#  endif
 }
 #endif // _LIBCPP_HAS_NO_INT128
 
 template <class _Tp>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 int __countl_zero(_Tp __t) _NOEXCEPT {
   static_assert(__libcpp_is_unsigned_integer<_Tp>::value, "__countl_zero requires an unsigned integer type");
+#if __has_builtin(__builtin_clzg)
+  return __builtin_clzg(__t, numeric_limits<_Tp>::digits);
+#else  // __has_builtin(__builtin_clzg)
   if (__t == 0)
     return numeric_limits<_Tp>::digits;
 
@@ -79,6 +89,7 @@ _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 int __countl_zero(_Tp __t) _
     }
     return __ret + __iter;
   }
+#endif // __has_builtin(__builtin_clzg)
 }
 
 #if _LIBCPP_STD_VER >= 20
diff --git a/libcxx/include/__bit/countr.h b/libcxx/include/__bit/countr.h
index b6b3ac52ca4e4..724a0bc23801c 100644
--- a/libcxx/include/__bit/countr.h
+++ b/libcxx/include/__bit/countr.h
@@ -6,6 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
+// TODO: __builtin_ctzg is available since Clang 19 and GCC 14. When support for older versions is dropped, we can
+//  refactor this code to exclusively use __builtin_ctzg.
+
 #ifndef _LIBCPP___BIT_COUNTR_H
 #define _LIBCPP___BIT_COUNTR_H
 
@@ -37,9 +40,11 @@ _LIBCPP_NODISCARD inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR int __libcpp_ct
 
 template <class _Tp>
 _LIBCPP_NODISCARD _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 int __countr_zero(_Tp __t) _NOEXCEPT {
+#if __has_builtin(__builtin_ctzg)
+  return __builtin_ctzg(__t, numeric_limits<_Tp>::digits);
+#else  // __has_builtin(__builtin_ctzg)
   if (__t == 0)
     return numeric_limits<_Tp>::digits;
-
   if (sizeof(_Tp) <= sizeof(unsigned int))
     return std::__libcpp_ctz(static_cast<unsigned int>(__t));
   else if (sizeof(_Tp) <= sizeof(unsigned long))
@@ -55,6 +60,7 @@ _LIBCPP_NODISCARD _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 int __coun
     }
     return __ret + std::__libcpp_ctz(static_cast<unsigned long long>(__t));
   }
+#endif // __has_builtin(__builtin_ctzg)
 }
 
 #if _LIBCPP_STD_VER >= 20
diff --git a/libcxx/include/__bit/popcount.h b/libcxx/include/__bit/popcount.h
index b0319cef25189..37b3a3e1f3f2b 100644
--- a/libcxx/include/__bit/popcount.h
+++ b/libcxx/include/__bit/popcount.h
@@ -6,6 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
+// TODO: __builtin_popcountg is available since Clang 19 and GCC 14. When support for older versions is dropped, we can
+//  refactor this code to exclusively use __builtin_popcountg.
+
 #ifndef _LIBCPP___BIT_POPCOUNT_H
 #define _LIBCPP___BIT_POPCOUNT_H
 
@@ -39,6 +42,9 @@ inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR int __libcpp_popcount(unsigned lo
 
 template <__libcpp_unsigned_integer _Tp>
 _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI constexpr int popcount(_Tp __t) noexcept {
+#  if __has_builtin(__builtin_popcountg)
+  return __builtin_popcountg(__t);
+#  else  // __has_builtin(__builtin_popcountg)
   if (sizeof(_Tp) <= sizeof(unsigned int))
     return std::__libcpp_popcount(static_cast<unsigned int>(__t));
   else if (sizeof(_Tp) <= sizeof(unsigned long))
@@ -53,6 +59,7 @@ _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI constexpr int popcount(_Tp __t) noex
     }
     return __ret;
   }
+#  endif // __has_builtin(__builtin_popcountg)
 }
 
 #endif // _LIBCPP_STD_VER >= 20

From 402f15ea92061d94412807887c8115374974967e Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Thu, 11 Apr 2024 14:38:18 +0200
Subject: [PATCH 139/886] [clang][dataflow] Remove deprecated alias
 `ControlFlowContext`. (#88358)

---
 clang/docs/tools/clang-formatted-files.txt    |  1 -
 .../FlowSensitive/ControlFlowContext.h        | 27 -------------------
 2 files changed, 28 deletions(-)
 delete mode 100644 clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h

diff --git a/clang/docs/tools/clang-formatted-files.txt b/clang/docs/tools/clang-formatted-files.txt
index 8fd4fed25a32a..3089438c23d94 100644
--- a/clang/docs/tools/clang-formatted-files.txt
+++ b/clang/docs/tools/clang-formatted-files.txt
@@ -123,7 +123,6 @@ clang/include/clang/Analysis/Analyses/CalledOnceCheck.h
 clang/include/clang/Analysis/Analyses/CFGReachabilityAnalysis.h
 clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
 clang/include/clang/Analysis/FlowSensitive/AdornedCFG.h
-clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h
 clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
 clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h
 clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
diff --git a/clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h b/clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h
deleted file mode 100644
index 3972962d0b2da..0000000000000
--- a/clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h
+++ /dev/null
@@ -1,27 +0,0 @@
-//===-- ControlFlowContext.h ------------------------------------*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-//  This file defines a deprecated alias for AdornedCFG.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_CONTROLFLOWCONTEXT_H
-#define LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_CONTROLFLOWCONTEXT_H
-
-#include "clang/Analysis/FlowSensitive/AdornedCFG.h"
-
-namespace clang {
-namespace dataflow {
-
-// This is a deprecated alias. Use `AdornedCFG` instead.
-using ControlFlowContext = AdornedCFG;
-
-} // namespace dataflow
-} // namespace clang
-
-#endif // LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_CONTROLFLOWCONTEXT_H

From 6fd2fdccf2f28fc155f614eec41f785492aad618 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 14:02:56 +0100
Subject: [PATCH 140/886] [VectorCombine] foldShuffleOfCastops - extend
 shuffle(bitcast(x),bitcast(y)) -> bitcast(shuffle(x,y)) support

Handle shuffle mask scaling handling for cases where the bitcast src/dst element counts are different
---
 clang/test/CodeGen/X86/avx-shuffle-builtins.c |  3 +-
 .../Transforms/Vectorize/VectorCombine.cpp    | 40 ++++++++++++++-----
 .../Transforms/PhaseOrdering/X86/pr67803.ll   |  5 +--
 .../VectorCombine/X86/shuffle-of-casts.ll     | 14 +++----
 .../Transforms/VectorCombine/X86/shuffle.ll   |  5 ++-
 5 files changed, 43 insertions(+), 24 deletions(-)

diff --git a/clang/test/CodeGen/X86/avx-shuffle-builtins.c b/clang/test/CodeGen/X86/avx-shuffle-builtins.c
index 49a56e73230d7..d184d28f3e07a 100644
--- a/clang/test/CodeGen/X86/avx-shuffle-builtins.c
+++ b/clang/test/CodeGen/X86/avx-shuffle-builtins.c
@@ -61,8 +61,7 @@ __m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) {
 
 __m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_permute2f128_si256
-  // X64: shufflevector{{.*}}<i32 0, i32 1, i32 4, i32 5>
-  // X86: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+  // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 4, i32 5>
   return _mm256_permute2f128_si256(a, b, 0x20);
 }
 
diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index b74fdf27d213a..658e8e74fe5b8 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -1448,9 +1448,9 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {
 /// into "castop (shuffle)".
 bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
   Value *V0, *V1;
-  ArrayRef<int> Mask;
+  ArrayRef<int> OldMask;
   if (!match(&I, m_Shuffle(m_OneUse(m_Value(V0)), m_OneUse(m_Value(V1)),
-                           m_Mask(Mask))))
+                           m_Mask(OldMask))))
     return false;
 
   auto *C0 = dyn_cast<CastInst>(V0);
@@ -1473,12 +1473,32 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
   auto *ShuffleDstTy = dyn_cast<FixedVectorType>(I.getType());
   auto *CastDstTy = dyn_cast<FixedVectorType>(C0->getDestTy());
   auto *CastSrcTy = dyn_cast<FixedVectorType>(C0->getSrcTy());
-  if (!ShuffleDstTy || !CastDstTy || !CastSrcTy ||
-      CastDstTy->getElementCount() != CastSrcTy->getElementCount())
+  if (!ShuffleDstTy || !CastDstTy || !CastSrcTy)
     return false;
 
+  unsigned NumSrcElts = CastSrcTy->getNumElements();
+  unsigned NumDstElts = CastDstTy->getNumElements();
+  assert((NumDstElts == NumSrcElts || Opcode == Instruction::BitCast) &&
+         "Only bitcasts expected to alter src/dst element counts");
+
+  SmallVector<int, 16> NewMask;
+  if (NumSrcElts >= NumDstElts) {
+    // The bitcast is from wide to narrow/equal elements. The shuffle mask can
+    // always be expanded to the equivalent form choosing narrower elements.
+    assert(NumSrcElts % NumDstElts == 0 && "Unexpected shuffle mask");
+    unsigned ScaleFactor = NumSrcElts / NumDstElts;
+    narrowShuffleMaskElts(ScaleFactor, OldMask, NewMask);
+  } else {
+    // The bitcast is from narrow elements to wide elements. The shuffle mask
+    // must choose consecutive elements to allow casting first.
+    assert(NumDstElts % NumSrcElts == 0 && "Unexpected shuffle mask");
+    unsigned ScaleFactor = NumDstElts / NumSrcElts;
+    if (!widenShuffleMaskElts(ScaleFactor, OldMask, NewMask))
+      return false;
+  }
+
   auto *NewShuffleDstTy =
-      FixedVectorType::get(CastSrcTy->getScalarType(), Mask.size());
+      FixedVectorType::get(CastSrcTy->getScalarType(), NewMask.size());
 
   // Try to replace a castop with a shuffle if the shuffle is not costly.
   TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;
@@ -1489,11 +1509,11 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
       TTI.getCastInstrCost(C1->getOpcode(), CastDstTy, CastSrcTy,
                            TTI::CastContextHint::None, CostKind);
   OldCost +=
-      TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, CastDstTy, Mask,
-                         CostKind, 0, nullptr, std::nullopt, &I);
+      TTI.getShuffleCost(TargetTransformInfo::SK_PermuteTwoSrc, CastDstTy,
+                         OldMask, CostKind, 0, nullptr, std::nullopt, &I);
 
   InstructionCost NewCost = TTI.getShuffleCost(
-      TargetTransformInfo::SK_PermuteTwoSrc, CastSrcTy, Mask, CostKind);
+      TargetTransformInfo::SK_PermuteTwoSrc, CastSrcTy, NewMask, CostKind);
   NewCost += TTI.getCastInstrCost(Opcode, ShuffleDstTy, NewShuffleDstTy,
                                   TTI::CastContextHint::None, CostKind);
 
@@ -1503,8 +1523,8 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
   if (NewCost > OldCost)
     return false;
 
-  Value *Shuf =
-      Builder.CreateShuffleVector(C0->getOperand(0), C1->getOperand(0), Mask);
+  Value *Shuf = Builder.CreateShuffleVector(C0->getOperand(0),
+                                            C1->getOperand(0), NewMask);
   Value *Cast = Builder.CreateCast(Opcode, Shuf, ShuffleDstTy);
 
   // Intersect flags from the old casts.
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/pr67803.ll b/llvm/test/Transforms/PhaseOrdering/X86/pr67803.ll
index 45e411d733169..36535264bd0f2 100644
--- a/llvm/test/Transforms/PhaseOrdering/X86/pr67803.ll
+++ b/llvm/test/Transforms/PhaseOrdering/X86/pr67803.ll
@@ -17,7 +17,6 @@ define <4 x i64> @PR67803(<4 x i64> %x, <4 x i64> %y, <4 x i64> %a, <4 x i64> %b
 ; CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i32> [[TMP3]] to <32 x i8>
 ; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <32 x i8> [[TMP9]], <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[TMP11:%.*]] = tail call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> [[TMP6]], <16 x i8> [[TMP8]], <16 x i8> [[TMP10]])
-; CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
 ; CHECK-NEXT:    [[TMP13:%.*]] = bitcast <4 x i64> [[A]] to <32 x i8>
 ; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <32 x i8> [[TMP13]], <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; CHECK-NEXT:    [[TMP15:%.*]] = bitcast <4 x i64> [[B]] to <32 x i8>
@@ -25,8 +24,8 @@ define <4 x i64> @PR67803(<4 x i64> %x, <4 x i64> %y, <4 x i64> %a, <4 x i64> %b
 ; CHECK-NEXT:    [[TMP17:%.*]] = bitcast <8 x i32> [[TMP3]] to <32 x i8>
 ; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <32 x i8> [[TMP17]], <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; CHECK-NEXT:    [[TMP19:%.*]] = tail call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> [[TMP14]], <16 x i8> [[TMP16]], <16 x i8> [[TMP18]])
-; CHECK-NEXT:    [[TMP20:%.*]] = bitcast <16 x i8> [[TMP19]] to <2 x i64>
-; CHECK-NEXT:    [[SHUFFLE_I23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> [[TMP20]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <16 x i8> [[TMP11]], <16 x i8> [[TMP19]], <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; CHECK-NEXT:    [[SHUFFLE_I23:%.*]] = bitcast <32 x i8> [[TMP20]] to <4 x i64>
 ; CHECK-NEXT:    ret <4 x i64> [[SHUFFLE_I23]]
 ;
 entry:
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
index 97fceacd82758..4f8cfea146ead 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
@@ -179,13 +179,12 @@ define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
   ret <8 x float> %r
 }
 
-; TODO - bitcasts (lower element count)
+; bitcasts (lower element count)
 
 define <4 x double> @concat_bitcast_v8i16_v4f64(<8 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-LABEL: @concat_bitcast_v8i16_v4f64(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <8 x i16> [[A0:%.*]] to <2 x double>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <8 x i16> [[A1:%.*]] to <2 x double>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x double> [[X0]], <2 x double> [[X1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    [[R:%.*]] = bitcast <16 x i16> [[TMP1]] to <4 x double>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
   %x0 = bitcast <8 x i16> %a0 to <2 x double>
@@ -194,13 +193,12 @@ define <4 x double> @concat_bitcast_v8i16_v4f64(<8 x i16> %a0, <8 x i16> %a1) {
   ret <4 x double> %r
 }
 
-; TODO - bitcasts (higher element count)
+; bitcasts (higher element count)
 
 define <16 x i16> @concat_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: @concat_bitcast_v4i32_v16i16(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <8 x i16>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <8 x i16>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x i16> [[X0]], <8 x i16> [[X1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[R:%.*]] = bitcast <8 x i32> [[TMP1]] to <16 x i16>
 ; CHECK-NEXT:    ret <16 x i16> [[R]]
 ;
   %x0 = bitcast <4 x i32> %a0 to <8 x i16>
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
index 3d47f373ab77c..8337bb37bc549 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
@@ -122,11 +122,14 @@ define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) {
 }
 
 ; shuffle of 2 operands removes bitcasts
+; TODO - can we remove the empty bitcast(bitcast()) ?
 
 define <4 x i64> @bitcast_shuf_remove_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: @bitcast_shuf_remove_bitcasts(
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    ret <4 x i64> [[R]]
+; CHECK-NEXT:    [[SHUF:%.*]] = bitcast <4 x i64> [[R]] to <8 x i32>
+; CHECK-NEXT:    [[R1:%.*]] = bitcast <8 x i32> [[SHUF]] to <4 x i64>
+; CHECK-NEXT:    ret <4 x i64> [[R1]]
 ;
   %bc0 = bitcast <2 x i64> %a0 to <4 x i32>
   %bc1 = bitcast <2 x i64> %a1 to <4 x i32>

From 77dd43570bf7a4bad688de8d8326c34590a0fa94 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Thu, 11 Apr 2024 15:08:54 +0200
Subject: [PATCH 141/886] Fix complex power for large inputs. (#88387)

For example, 1e30^1.2 currently overflows.

Also forward fastmath flags.

This ports XLA's logic and was verified with its test suite. Note that
rsqrt and sqrt are still broken.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 149 +++++++++++-------
 .../convert-to-standard.mlir                  |  21 ++-
 2 files changed, 115 insertions(+), 55 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 462036e51a1f1..9c82e8105f06e 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -989,65 +989,107 @@ struct ConjOpConversion : public OpConversionPattern<complex::ConjOp> {
   }
 };
 
-/// Coverts x^y = (a+bi)^(c+di) to
+/// Converts lhs^y = (a+bi)^(c+di) to
 ///    (a*a+b*b)^(0.5c) * exp(-d*atan2(b,a)) * (cos(q) + i*sin(q)),
 ///    where q = c*atan2(b,a)+0.5d*ln(a*a+b*b)
 static Value powOpConversionImpl(mlir::ImplicitLocOpBuilder &builder,
-                                 ComplexType type, Value a, Value b, Value c,
-                                 Value d) {
+                                 ComplexType type, Value lhs, Value c, Value d,
+                                 arith::FastMathFlags fmf) {
   auto elementType = cast<FloatType>(type.getElementType());
 
-  // Compute (a*a+b*b)^(0.5c).
-  Value aaPbb = builder.create<arith::AddFOp>(
-      builder.create<arith::MulFOp>(a, a), builder.create<arith::MulFOp>(b, b));
-  Value half = builder.create<arith::ConstantOp>(
-      elementType, builder.getFloatAttr(elementType, 0.5));
-  Value halfC = builder.create<arith::MulFOp>(half, c);
-  Value aaPbbTohalfC = builder.create<math::PowFOp>(aaPbb, halfC);
-
-  // Compute exp(-d*atan2(b,a)).
-  Value negD = builder.create<arith::NegFOp>(d);
-  Value argX = builder.create<math::Atan2Op>(b, a);
-  Value negDArgX = builder.create<arith::MulFOp>(negD, argX);
-  Value eToNegDArgX = builder.create<math::ExpOp>(negDArgX);
-
-  // Compute (a*a+b*b)^(0.5c) * exp(-d*atan2(b,a)).
-  Value coeff = builder.create<arith::MulFOp>(aaPbbTohalfC, eToNegDArgX);
-
-  // Compute c*atan2(b,a)+0.5d*ln(a*a+b*b).
-  Value lnAaPbb = builder.create<math::LogOp>(aaPbb);
-  Value halfD = builder.create<arith::MulFOp>(half, d);
-  Value q = builder.create<arith::AddFOp>(
-      builder.create<arith::MulFOp>(c, argX),
-      builder.create<arith::MulFOp>(halfD, lnAaPbb));
-
-  Value cosQ = builder.create<math::CosOp>(q);
-  Value sinQ = builder.create<math::SinOp>(q);
+  Value a = builder.create<complex::ReOp>(lhs);
+  Value b = builder.create<complex::ImOp>(lhs);
+
+  Value abs = builder.create<complex::AbsOp>(lhs, fmf);
+  Value absToC = builder.create<math::PowFOp>(abs, c, fmf);
+
+  Value negD = builder.create<arith::NegFOp>(d, fmf);
+  Value argLhs = builder.create<math::Atan2Op>(b, a, fmf);
+  Value negDArgLhs = builder.create<arith::MulFOp>(negD, argLhs, fmf);
+  Value expNegDArgLhs = builder.create<math::ExpOp>(negDArgLhs, fmf);
+
+  Value coeff = builder.create<arith::MulFOp>(absToC, expNegDArgLhs, fmf);
+  Value lnAbs = builder.create<math::LogOp>(abs, fmf);
+  Value cArgLhs = builder.create<arith::MulFOp>(c, argLhs, fmf);
+  Value dLnAbs = builder.create<arith::MulFOp>(d, lnAbs, fmf);
+  Value q = builder.create<arith::AddFOp>(cArgLhs, dLnAbs, fmf);
+  Value cosQ = builder.create<math::CosOp>(q, fmf);
+  Value sinQ = builder.create<math::SinOp>(q, fmf);
+
+  Value inf = builder.create<arith::ConstantOp>(
+      elementType,
+      builder.getFloatAttr(elementType,
+                           APFloat::getInf(elementType.getFloatSemantics())));
   Value zero = builder.create<arith::ConstantOp>(
-      elementType, builder.getFloatAttr(elementType, 0));
+      elementType, builder.getFloatAttr(elementType, 0.0));
   Value one = builder.create<arith::ConstantOp>(
-      elementType, builder.getFloatAttr(elementType, 1));
-
-  Value xEqZero =
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, aaPbb, zero);
-  Value yGeZero = builder.create<arith::AndIOp>(
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OGE, c, zero),
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, d, zero));
-  Value cEqZero =
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, c, zero);
-  Value complexZero = builder.create<complex::CreateOp>(type, zero, zero);
+      elementType, builder.getFloatAttr(elementType, 1.0));
   Value complexOne = builder.create<complex::CreateOp>(type, one, zero);
-  Value complexOther = builder.create<complex::CreateOp>(
-      type, builder.create<arith::MulFOp>(coeff, cosQ),
-      builder.create<arith::MulFOp>(coeff, sinQ));
+  Value complexZero = builder.create<complex::CreateOp>(type, zero, zero);
+  Value complexInf = builder.create<complex::CreateOp>(type, inf, zero);
 
-  // x^y is 0 if x is 0 and y > 0. 0^0 is defined to be 1.0, see
+  // Case 0:
+  // d^c is 0 if d is 0 and c > 0. 0^0 is defined to be 1.0, see
+  // Branch Cuts for Complex Elementary Functions or Much Ado About
+  // Nothing's Sign Bit, W. Kahan, Section 10.
+  Value absEqZero =
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, abs, zero, fmf);
+  Value dEqZero =
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, d, zero, fmf);
+  Value cEqZero =
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, c, zero, fmf);
+  Value bEqZero =
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, b, zero, fmf);
+
+  Value zeroLeC =
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OLE, zero, c, fmf);
+  Value coeffCosQ = builder.create<arith::MulFOp>(coeff, cosQ, fmf);
+  Value coeffSinQ = builder.create<arith::MulFOp>(coeff, sinQ, fmf);
+  Value complexOneOrZero =
+      builder.create<arith::SelectOp>(cEqZero, complexOne, complexZero);
+  Value coeffCosSin =
+      builder.create<complex::CreateOp>(type, coeffCosQ, coeffSinQ);
+  Value cutoff0 = builder.create<arith::SelectOp>(
+      builder.create<arith::AndIOp>(
+          builder.create<arith::AndIOp>(absEqZero, dEqZero), zeroLeC),
+      complexOneOrZero, coeffCosSin);
+
+  // Case 1:
+  // x^0 is defined to be 1 for any x, see
   // Branch Cuts for Complex Elementary Functions or Much Ado About
   // Nothing's Sign Bit, W. Kahan, Section 10.
-  return builder.create<arith::SelectOp>(
-      builder.create<arith::AndIOp>(xEqZero, yGeZero),
-      builder.create<arith::SelectOp>(cEqZero, complexOne, complexZero),
-      complexOther);
+  Value rhsEqZero = builder.create<arith::AndIOp>(cEqZero, dEqZero);
+  Value cutoff1 =
+      builder.create<arith::SelectOp>(rhsEqZero, complexOne, cutoff0);
+
+  // Case 2:
+  // 1^(c + d*i) = 1 + 0*i
+  Value lhsEqOne = builder.create<arith::AndIOp>(
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, one),
+      bEqZero);
+  Value cutoff2 =
+      builder.create<arith::SelectOp>(lhsEqOne, complexOne, cutoff1);
+
+  // Case 3:
+  // inf^(c + 0*i) = inf + 0*i, c > 0
+  Value lhsEqInf = builder.create<arith::AndIOp>(
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, inf),
+      bEqZero);
+  Value rhsGt0 = builder.create<arith::AndIOp>(
+      dEqZero,
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, c, zero));
+  Value cutoff3 = builder.create<arith::SelectOp>(
+      builder.create<arith::AndIOp>(lhsEqInf, rhsGt0), complexInf, cutoff2);
+
+  // Case 4:
+  // inf^(c + 0*i) = 0 + 0*i, c < 0
+  Value rhsLt0 = builder.create<arith::AndIOp>(
+      dEqZero,
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OLT, c, zero));
+  Value cutoff4 = builder.create<arith::SelectOp>(
+      builder.create<arith::AndIOp>(lhsEqInf, rhsLt0), complexZero, cutoff3);
+
+  return cutoff4;
 }
 
 struct PowOpConversion : public OpConversionPattern<complex::PowOp> {
@@ -1060,12 +1102,11 @@ struct PowOpConversion : public OpConversionPattern<complex::PowOp> {
     auto type = cast<ComplexType>(adaptor.getLhs().getType());
     auto elementType = cast<FloatType>(type.getElementType());
 
-    Value a = builder.create<complex::ReOp>(elementType, adaptor.getLhs());
-    Value b = builder.create<complex::ImOp>(elementType, adaptor.getLhs());
     Value c = builder.create<complex::ReOp>(elementType, adaptor.getRhs());
     Value d = builder.create<complex::ImOp>(elementType, adaptor.getRhs());
 
-    rewriter.replaceOp(op, {powOpConversionImpl(builder, type, a, b, c, d)});
+    rewriter.replaceOp(op, {powOpConversionImpl(builder, type, adaptor.getLhs(),
+                                                c, d, op.getFastmath())});
     return success();
   }
 };
@@ -1080,14 +1121,14 @@ struct RsqrtOpConversion : public OpConversionPattern<complex::RsqrtOp> {
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
 
-    Value a = builder.create<complex::ReOp>(elementType, adaptor.getComplex());
-    Value b = builder.create<complex::ImOp>(elementType, adaptor.getComplex());
     Value c = builder.create<arith::ConstantOp>(
         elementType, builder.getFloatAttr(elementType, -0.5));
     Value d = builder.create<arith::ConstantOp>(
         elementType, builder.getFloatAttr(elementType, 0));
 
-    rewriter.replaceOp(op, {powOpConversionImpl(builder, type, a, b, c, d)});
+    rewriter.replaceOp(op,
+                       {powOpConversionImpl(builder, type, adaptor.getComplex(),
+                                            c, d, op.getFastmath())});
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index a1de61d10bb22..8d2fb09daa87b 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -753,13 +753,32 @@ func.func @complex_conj(%arg: complex<f32>) -> complex<f32> {
 
 // -----
 
-// CHECK-LABEL:   func.func @complex_pow
+// CHECK-LABEL: func.func @complex_pow
+// CHECK-SAME: %[[LHS:.*]]: complex<f32>, %[[RHS:.*]]: complex<f32>
 func.func @complex_pow(%lhs: complex<f32>,
                        %rhs: complex<f32>) -> complex<f32> {
   %pow = complex.pow %lhs, %rhs : complex<f32>
   return %pow : complex<f32>
 }
 
+// CHECK: %[[A:.*]] = complex.re %[[LHS]]
+// CHECK: %[[B:.*]] = complex.im %[[LHS]]
+// CHECK: math.atan2 %[[B]], %[[A]] : f32
+
+// -----
+
+// CHECK-LABEL: func.func @complex_pow_with_fmf
+// CHECK-SAME: %[[LHS:.*]]: complex<f32>, %[[RHS:.*]]: complex<f32>
+func.func @complex_pow_with_fmf(%lhs: complex<f32>,
+                                %rhs: complex<f32>) -> complex<f32> {
+  %pow = complex.pow %lhs, %rhs fastmath<nnan,contract> : complex<f32>
+  return %pow : complex<f32>
+}
+
+// CHECK: %[[A:.*]] = complex.re %[[LHS]]
+// CHECK: %[[B:.*]] = complex.im %[[LHS]]
+// CHECK: math.atan2 %[[B]], %[[A]] fastmath<nnan,contract> : f32
+
 // -----
 
 // CHECK-LABEL:   func.func @complex_rsqrt

From 599adf30afe5802fab80419ec5bb896036a1c8fb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?kadir=20=C3=A7etinkaya?= <kadircet@google.com>
Date: Thu, 11 Apr 2024 15:33:35 +0200
Subject: [PATCH 142/886] [include-cleaner] Dont apply name-match for
 non-owning headers (#82625)

---
 .../include-cleaner/lib/FindHeaders.cpp       |  6 ++++++
 .../unittests/FindHeadersTest.cpp             | 19 +++++++++++++++++++
 2 files changed, 25 insertions(+)

diff --git a/clang-tools-extra/include-cleaner/lib/FindHeaders.cpp b/clang-tools-extra/include-cleaner/lib/FindHeaders.cpp
index fd2de6a17ad4a..7b28d1c252d71 100644
--- a/clang-tools-extra/include-cleaner/lib/FindHeaders.cpp
+++ b/clang-tools-extra/include-cleaner/lib/FindHeaders.cpp
@@ -275,6 +275,12 @@ llvm::SmallVector<Header> headersForSymbol(const Symbol &S,
     // are already ranked in the stdlib mapping.
     if (H.kind() == Header::Standard)
       continue;
+    // Don't apply name match hints to exporting headers. As they usually have
+    // names similar to the original header, e.g. foo_wrapper/foo.h vs
+    // foo/foo.h, but shouldn't be preferred (unless marked as the public
+    // interface).
+    if ((H.Hint & Hints::OriginHeader) == Hints::None)
+      continue;
     if (nameMatch(SymbolName, H))
       H.Hint |= Hints::PreferredHeader;
   }
diff --git a/clang-tools-extra/include-cleaner/unittests/FindHeadersTest.cpp b/clang-tools-extra/include-cleaner/unittests/FindHeadersTest.cpp
index 5a2a41b2d99bd..07302142a13e3 100644
--- a/clang-tools-extra/include-cleaner/unittests/FindHeadersTest.cpp
+++ b/clang-tools-extra/include-cleaner/unittests/FindHeadersTest.cpp
@@ -628,5 +628,24 @@ TEST_F(HeadersForSymbolTest, StandardHeaders) {
                            tooling::stdlib::Header::named("<assert.h>")));
 }
 
+TEST_F(HeadersForSymbolTest, ExporterNoNameMatch) {
+  Inputs.Code = R"cpp(
+    #include "exporter/foo.h"
+    #include "foo_public.h"
+  )cpp";
+  Inputs.ExtraArgs.emplace_back("-I.");
+  // Deliberately named as foo_public to make sure it doesn't get name-match
+  // boost and also gets lexicographically bigger order than "exporter/foo.h".
+  Inputs.ExtraFiles["foo_public.h"] = guard(R"cpp(
+    struct foo {};
+  )cpp");
+  Inputs.ExtraFiles["exporter/foo.h"] = guard(R"cpp(
+    #include "foo_public.h" // IWYU pragma: export
+  )cpp");
+  buildAST();
+  EXPECT_THAT(headersForFoo(), ElementsAre(physicalHeader("foo_public.h"),
+                                           physicalHeader("exporter/foo.h")));
+}
+
 } // namespace
 } // namespace clang::include_cleaner

From 61ea1bc23aa941714be3ec818c922e4ee5a279a3 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 15:23:50 +0100
Subject: [PATCH 143/886] [VectorCombine][X86] Add test coverage for #67803

We are still missing a fold for shuffle(bitcast(sext(x)),bitcast(sext(y))) -> bitcast(sext(shuffle(x,y))) due to foldShuffleOfCastops failing to add new instructions back onto the worklist
---
 .../Transforms/VectorCombine/X86/pr67803.ll   | 33 +++++++++++++++++++
 1 file changed, 33 insertions(+)
 create mode 100644 llvm/test/Transforms/VectorCombine/X86/pr67803.ll

diff --git a/llvm/test/Transforms/VectorCombine/X86/pr67803.ll b/llvm/test/Transforms/VectorCombine/X86/pr67803.ll
new file mode 100644
index 0000000000000..da94bf7f0c907
--- /dev/null
+++ b/llvm/test/Transforms/VectorCombine/X86/pr67803.ll
@@ -0,0 +1,33 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx  | FileCheck %s
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx2 | FileCheck %s
+
+define <4 x i64> @PR67803(<8 x i32> %x, <8 x i32> %y, <8 x float> %a, <8 x float> %b) {
+; CHECK-LABEL: @PR67803(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <8 x i32> [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[CMP_LO:%.*]] = shufflevector <8 x i1> [[CMP]], <8 x i1> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[CMP_HI:%.*]] = shufflevector <8 x i1> [[CMP]], <8 x i1> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[SEXT_LO:%.*]] = sext <4 x i1> [[CMP_LO]] to <4 x i32>
+; CHECK-NEXT:    [[SEXT_HI:%.*]] = sext <4 x i1> [[CMP_HI]] to <4 x i32>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[SEXT_LO]], <4 x i32> [[SEXT_HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[CONCAT:%.*]] = bitcast <8 x i32> [[TMP1]] to <4 x i64>
+; CHECK-NEXT:    [[MASK:%.*]] = bitcast <4 x i64> [[CONCAT]] to <8 x float>
+; CHECK-NEXT:    [[SEL:%.*]] = tail call noundef <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> [[A:%.*]], <8 x float> [[B:%.*]], <8 x float> [[MASK]])
+; CHECK-NEXT:    [[RES:%.*]] = bitcast <8 x float> [[SEL]] to <4 x i64>
+; CHECK-NEXT:    ret <4 x i64> [[RES]]
+;
+entry:
+  %cmp = icmp sgt <8 x i32> %x, %y
+  %cmp.lo = shufflevector <8 x i1> %cmp, <8 x i1> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %cmp.hi = shufflevector <8 x i1> %cmp, <8 x i1> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %sext.lo = sext <4 x i1> %cmp.lo to <4 x i32>
+  %sext.hi = sext <4 x i1> %cmp.hi to <4 x i32>
+  %bitcast.lo = bitcast <4 x i32> %sext.lo to <2 x i64>
+  %bitcast.hi = bitcast <4 x i32> %sext.hi to <2 x i64>
+  %concat = shufflevector <2 x i64> %bitcast.lo, <2 x i64> %bitcast.hi, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %mask = bitcast <4 x i64> %concat to <8 x float>
+  %sel = tail call noundef <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %mask)
+  %res = bitcast <8 x float> %sel to <4 x i64>
+  ret <4 x i64> %res
+}

From ff74236f342c7bc185f56a07bab7bd0cf356c7c6 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 15:47:09 +0100
Subject: [PATCH 144/886] [VectorCombine] foldShuffleOfCastops - ensure we add
 all new instructions onto the worklist

When creating cast(shuffle(x,y)) we were only adding the cast() to the worklist, not the new shuffle, preventing recursive combines.

foldShuffleOfBinops is also failing to do this, but I still need to add test coverage for this.
---
 llvm/lib/Transforms/Vectorize/VectorCombine.cpp   | 3 +++
 llvm/test/Transforms/VectorCombine/X86/pr67803.ll | 5 ++---
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 658e8e74fe5b8..44cba60013afa 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -1440,6 +1440,8 @@ bool VectorCombine::foldShuffleOfBinops(Instruction &I) {
     NewInst->copyIRFlags(B0);
     NewInst->andIRFlags(B1);
   }
+
+  // TODO: Add Shuf0/Shuf1 to WorkList?
   replaceValue(I, *NewBO);
   return true;
 }
@@ -1533,6 +1535,7 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
     NewInst->andIRFlags(C1);
   }
 
+  Worklist.pushValue(Shuf);
   replaceValue(I, *Cast);
   return true;
 }
diff --git a/llvm/test/Transforms/VectorCombine/X86/pr67803.ll b/llvm/test/Transforms/VectorCombine/X86/pr67803.ll
index da94bf7f0c907..69fd6f6a10e2a 100644
--- a/llvm/test/Transforms/VectorCombine/X86/pr67803.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/pr67803.ll
@@ -8,9 +8,8 @@ define <4 x i64> @PR67803(<8 x i32> %x, <8 x i32> %y, <8 x float> %a, <8 x float
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <8 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[CMP_LO:%.*]] = shufflevector <8 x i1> [[CMP]], <8 x i1> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    [[CMP_HI:%.*]] = shufflevector <8 x i1> [[CMP]], <8 x i1> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:    [[SEXT_LO:%.*]] = sext <4 x i1> [[CMP_LO]] to <4 x i32>
-; CHECK-NEXT:    [[SEXT_HI:%.*]] = sext <4 x i1> [[CMP_HI]] to <4 x i32>
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[SEXT_LO]], <4 x i32> [[SEXT_HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i1> [[CMP_LO]], <4 x i1> [[CMP_HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP1:%.*]] = sext <8 x i1> [[TMP0]] to <8 x i32>
 ; CHECK-NEXT:    [[CONCAT:%.*]] = bitcast <8 x i32> [[TMP1]] to <4 x i64>
 ; CHECK-NEXT:    [[MASK:%.*]] = bitcast <4 x i64> [[CONCAT]] to <8 x float>
 ; CHECK-NEXT:    [[SEL:%.*]] = tail call noundef <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> [[A:%.*]], <8 x float> [[B:%.*]], <8 x float> [[MASK]])

From 44718311dee486f1823876e8af9100afcc50041b Mon Sep 17 00:00:00 2001
From: Jakub Kuderski <jakub@nod-labs.com>
Date: Thu, 11 Apr 2024 11:07:17 -0400
Subject: [PATCH 145/886] [mlir][amdgpu] Remove shared memory optimization pass
 (#88225)

This implementation has a number of issues and ultimately does not work
on gfx9.
* It does not reduce bank conflicts with wide memory accesses.
* It does not correctly account for when LDS bank conflicts occur on
amdgpu.
* The implementation is too fragile to be used on real-world code. For
example, the code bails out on any `memref.subview` in the root op, even
when the subview is not a user of any of the `memref.alloc` ops.

I do not see how these can be easily fixed, therefore I think it's
better to delete this code.
---
 .../mlir/Dialect/AMDGPU/CMakeLists.txt        |   1 -
 mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td |  17 --
 .../AMDGPU/TransformOps/AMDGPUTransformOps.h  |  48 ----
 .../AMDGPU/TransformOps/AMDGPUTransformOps.td |  47 ----
 .../AMDGPU/TransformOps/CMakeLists.txt        |   4 -
 .../mlir/Dialect/AMDGPU/Transforms/Passes.h   |   2 +-
 .../mlir/Dialect/AMDGPU/Transforms/Passes.td  |  20 --
 .../Dialect/AMDGPU/Transforms/Transforms.h    |  61 ----
 .../mlir/Dialect/AMDGPU/Transforms/Utils.h    |  24 --
 mlir/include/mlir/InitAllExtensions.h         |   2 -
 mlir/lib/Dialect/AMDGPU/CMakeLists.txt        |   1 -
 mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp  |  15 -
 .../TransformOps/AMDGPUTransformOps.cpp       |  67 -----
 .../AMDGPU/TransformOps/CMakeLists.txt        |  25 --
 .../Dialect/AMDGPU/Transforms/CMakeLists.txt  |   3 -
 .../Transforms/OptimizeSharedMemory.cpp       | 261 ------------------
 mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp  |  39 ---
 .../AMDGPU/optimize_shmem_reads_writes.mlir   |  50 ----
 ...transform_optimize_shmem_reads_writes.mlir |  54 ----
 .../llvm-project-overlay/mlir/BUILD.bazel     |  54 ----
 20 files changed, 1 insertion(+), 794 deletions(-)
 delete mode 100644 mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h
 delete mode 100644 mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td
 delete mode 100644 mlir/include/mlir/Dialect/AMDGPU/TransformOps/CMakeLists.txt
 delete mode 100644 mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h
 delete mode 100644 mlir/include/mlir/Dialect/AMDGPU/Transforms/Utils.h
 delete mode 100644 mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp
 delete mode 100644 mlir/lib/Dialect/AMDGPU/TransformOps/CMakeLists.txt
 delete mode 100644 mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp
 delete mode 100644 mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp
 delete mode 100644 mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir
 delete mode 100644 mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir

diff --git a/mlir/include/mlir/Dialect/AMDGPU/CMakeLists.txt b/mlir/include/mlir/Dialect/AMDGPU/CMakeLists.txt
index 660deb21479d2..9f57627c321fb 100644
--- a/mlir/include/mlir/Dialect/AMDGPU/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/AMDGPU/CMakeLists.txt
@@ -1,3 +1,2 @@
 add_subdirectory(IR)
-add_subdirectory(TransformOps)
 add_subdirectory(Transforms)
diff --git a/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td b/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
index 21942b179a001..3f27e1541cf38 100644
--- a/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
+++ b/mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
@@ -29,23 +29,6 @@ def AMDGPU_Dialect : Dialect {
     "gpu::GPUDialect"
   ];
   let useDefaultAttributePrinterParser = 1;
-
-  let extraClassDeclaration = [{
-    /// Return true if the given MemRefType has an integer address
-    /// space that matches the ROCDL shared memory address space or
-    /// is a gpu::AddressSpaceAttr attribute with value 'workgroup`.
-    static bool hasSharedMemoryAddressSpace(MemRefType type);
-
-    /// Return true if the given Attribute has an integer address
-    /// space that matches the ROCDL shared memory address space or
-    /// is a gpu::AddressSpaceAttr attribute with value 'workgroup`.
-    static bool isSharedMemoryAddressSpace(Attribute type);
-
-    /// Defines the MemRef memory space attribute numeric value that indicates
-    /// a memref is located in shared memory. This should correspond to the
-    /// value used in ROCDL.
-    static constexpr unsigned kSharedMemoryAddressSpace = 3;
-  }];
 }
 
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h b/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h
deleted file mode 100644
index dcf934c71dd1f..0000000000000
--- a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h
+++ /dev/null
@@ -1,48 +0,0 @@
-//===- AMDGPUTransformOps.h - AMDGPU transform ops ---------------*- C++-*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef MLIR_DIALECT_AMDGPU_TRANSFORMOPS_AMDGPUTRANSFORMOPS_H
-#define MLIR_DIALECT_AMDGPU_TRANSFORMOPS_AMDGPUTRANSFORMOPS_H
-
-#include "mlir/Dialect/Func/IR/FuncOps.h"
-#include "mlir/Dialect/Transform/IR/TransformAttrs.h"
-#include "mlir/Dialect/Transform/IR/TransformDialect.h"
-#include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.h"
-#include "mlir/IR/OpImplementation.h"
-#include "mlir/IR/RegionKindInterface.h"
-
-namespace mlir {
-namespace transform {
-class TransformHandleTypeInterface;
-} // namespace transform
-} // namespace mlir
-
-namespace mlir {
-class DialectRegistry;
-
-namespace linalg {
-class LinalgOp;
-} // namespace linalg
-
-namespace scf {
-class ForOp;
-} // namespace scf
-
-namespace amdgpu {
-void registerTransformDialectExtension(DialectRegistry &registry);
-} // namespace amdgpu
-} // namespace mlir
-
-//===----------------------------------------------------------------------===//
-// AMDGPU Transform Operations
-//===----------------------------------------------------------------------===//
-
-#define GET_OP_CLASSES
-#include "mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h.inc"
-
-#endif // MLIR_DIALECT_AMDGPU_TRANSFORMOPS_AMDGPUTRANSFORMOPS_H
diff --git a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td b/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td
deleted file mode 100644
index 8aaa87511a2be..0000000000000
--- a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td
+++ /dev/null
@@ -1,47 +0,0 @@
-//===- AMDGPUTransformOps.td - AMDGPU transform ops --------*- tablegen -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef AMDGPU_TRANSFORM_OPS
-#define AMDGPU_TRANSFORM_OPS
-
-include "mlir/Dialect/Transform/IR/TransformAttrs.td"
-include "mlir/Dialect/Transform/IR/TransformDialect.td"
-include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.td"
-include "mlir/Dialect/Transform/IR/TransformTypes.td"
-
-include "mlir/Interfaces/SideEffectInterfaces.td"
-//===----------------------------------------------------------------------===//
-// ApplyOptimizeSharedMemoryReadsAndWritesOp
-//===----------------------------------------------------------------------===//
-
-def ApplyOptimizeSharedMemoryReadsAndWritesOp :
-  Op<Transform_Dialect, "amdgpu.optimize_shared_memory_reads_and_writes",
-    [DeclareOpInterfaceMethods<MemoryEffectsOpInterface>,
-     TransformOpInterface, TransformEachOpTrait]> {
-  let summary = "Reduce shared memory bank conflicts";
-  let description = [{ This op attempts to optimize GPU Shared memory
-    reads/writes with the goal of avoiding bank conflicts.
-  }];
-
-  let arguments = (ins TransformHandleTypeInterface:$target,
-                    DefaultValuedOptionalAttr<I64Attr, "128">:$sharedMemoryLineSizeBytes,
-                    DefaultValuedOptionalAttr<I64Attr, "128">:$defaultVectorSizeBits);
-  let results = (outs);
-
-  let assemblyFormat = "$target attr-dict `:` functional-type(operands, results)";
-
-  let extraClassDeclaration = [{
-    ::mlir::DiagnosedSilenceableFailure applyToOne(
-        ::mlir::transform::TransformRewriter &rewriter,
-        ::mlir::func::FuncOp funcOp,
-        ::mlir::transform::ApplyToEachResultList &results,
-        ::mlir::transform::TransformState &state);
-  }];
-}
-
-#endif // AMDGPU_TRANSFORM_OPS
diff --git a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/CMakeLists.txt b/mlir/include/mlir/Dialect/AMDGPU/TransformOps/CMakeLists.txt
deleted file mode 100644
index 07bfebc9f96d2..0000000000000
--- a/mlir/include/mlir/Dialect/AMDGPU/TransformOps/CMakeLists.txt
+++ /dev/null
@@ -1,4 +0,0 @@
-set(LLVM_TARGET_DEFINITIONS AMDGPUTransformOps.td)
-mlir_tablegen(AMDGPUTransformOps.h.inc -gen-op-decls)
-mlir_tablegen(AMDGPUTransformOps.cpp.inc -gen-op-defs)
-add_public_tablegen_target(MLIRAMDGPUTransformOpsIncGen) 
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.h b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.h
index ab695756d2a78..8dd5ff1a4b198 100644
--- a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.h
+++ b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.h
@@ -20,7 +20,7 @@ namespace mlir {
 class ConversionTarget;
 namespace amdgpu {
 
-#define GEN_PASS_DECL
+#define GEN_PASS_DECL_AMDGPUEMULATEATOMICSPASS
 #define GEN_PASS_REGISTRATION
 #include "mlir/Dialect/AMDGPU/Transforms/Passes.h.inc"
 
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td
index 67f951fd19d17..e6b27aa842dfc 100644
--- a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td
+++ b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td
@@ -30,24 +30,4 @@ def AmdgpuEmulateAtomicsPass : Pass<"amdgpu-emulate-atomics"> {
                         "Chipset that these operations will run on">];
 }
 
-def OptimizeSharedMemory : Pass<"amdgpu-optimize-shared-memory"> {
-  let summary = "Optimizes accesses to shared memory memrefs in order to reduce bank conflicts.";
-  let description = [{
-    This pass adds a transformation and pass to the AMDGPU dialect that
-    attempts to optimize reads/writes from a memref representing GPU shared
-    memory in order to avoid bank conflicts.
-  }];
-  let dependentDialects = [
-    "memref::MemRefDialect", "vector::VectorDialect"
-  ];
-  let options = [
-    Option<"sharedMemoryLineSizeBytes", "shared-memory-line-size-bytes", "int64_t",
-           /*default=*/"128",
-           "Shared memory line size in bytes">,
-    Option<"defaultVectorSizeBits", "default-vector-size-bits", "int64_t",
-           /*default=*/"128",
-           "Default vector size in bits">,
-  ];
-}
-
 #endif // MLIR_DIALECT_AMDGPU_TRANSFORMS_PASSES_TD_
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h
deleted file mode 100644
index 843cea2c503b9..0000000000000
--- a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h
+++ /dev/null
@@ -1,61 +0,0 @@
-//===- Transforms.h - AMDGPU Dialect transformations -------------*- C++-*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file declares functions that assist transformations for the amdgpu
-// dialect.
-//
-//===----------------------------------------------------------------------===//
-#ifndef MLIR_DIALECT_AMDGPU_TRANSFORMS_TRANSFORMS_H_
-#define MLIR_DIALECT_AMDGPU_TRANSFORMS_TRANSFORMS_H_
-
-#include "mlir/Dialect/Func/IR/FuncOps.h"
-#include "mlir/IR/Operation.h"
-#include "mlir/Support/LogicalResult.h"
-
-namespace mlir {
-class RewriterBase;
-
-namespace amdgpu {
-
-///
-/// Passes
-///
-
-/// Optimizes vectorized accesses to a shared memory buffer specified by
-/// memrefValue. This transformation assumes the following:
-/// 1) All relevant accesses to `memrefValue` are contained with `parentOp`.
-/// 2) The function will fail precondition checks if any subviews are
-/// taken of `memrefValue`. All reads/writes to `memrefValue` should occur
-/// through `memrefValue` directly.
-///
-/// Shared memory bank conflicts occur when multiple threads attempt to read or
-/// write locations assigned to the same shared memory bank. For `2^N` byte
-/// vectorized accesses, we need to be concerned with conflicts among threads
-/// identified as `(tid) -> tid.floordiv(2^{7-N})`. As such, this transformation
-/// changes any indexed memory access (vector.load, memref.load, etc)
-/// such that the final dimension's index value is permuted such that
-/// `newColIndex = oldColIndex % vectorSize +
-/// perm[rowIndex](oldColIndex/vectorSize, rowIndex)` where `rowIndex` is the
-/// index for the second-to last dimension and `perm[rowIndex]` is a permutation
-/// function that depends on the row Index. The permutation function is chosen
-/// to ensure that sequential distributed+vectorized reads/writes down a single
-/// dimension of the memref have minimal conflicts.
-LogicalResult
-optimizeSharedMemoryReadsAndWrites(Operation *parentOp, Value memrefValue,
-                                   int64_t sharedMemoryLineSizeBytes,
-                                   int64_t defaultVectorSizeBits);
-
-std::optional<LogicalResult>
-optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp,
-                                     int64_t sharedMemoryLineSizeBytes,
-                                     int64_t defaultVectorSizeBits);
-
-} // namespace amdgpu
-} // namespace mlir
-
-#endif // MLIR_DIALECT_AMDGPU_TRANSFORMS_TRANSFORMS_H_
diff --git a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Utils.h b/mlir/include/mlir/Dialect/AMDGPU/Transforms/Utils.h
deleted file mode 100644
index 9e5e9589d62f3..0000000000000
--- a/mlir/include/mlir/Dialect/AMDGPU/Transforms/Utils.h
+++ /dev/null
@@ -1,24 +0,0 @@
-//===- Utils.h - Transform utilities -----------------------------*- C++-*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "mlir/IR/Operation.h"
-
-namespace mlir {
-namespace amdgpu {
-
-/// Get and set the indices that the given load/store operation is operating on.
-/// Preconditions:
-/// - The Op must have memory affects.
-/// - Considers memref::LoadOp, vector::LoadOp, and vector::TransferReadOp.
-/// - Considers memref::StoreOp, vector::StoreOp, and vector::TransferWriteOp.
-/// - Excludes subview op.
-std::optional<Operation::operand_range> getIndices(Operation *op);
-void setIndices(Operation *op, ArrayRef<Value> indices);
-
-} // namespace amdgpu
-} // namespace mlir
diff --git a/mlir/include/mlir/InitAllExtensions.h b/mlir/include/mlir/InitAllExtensions.h
index b31fb26f00f8f..7708ca5571de3 100644
--- a/mlir/include/mlir/InitAllExtensions.h
+++ b/mlir/include/mlir/InitAllExtensions.h
@@ -23,7 +23,6 @@
 #include "mlir/Conversion/MemRefToLLVM/MemRefToLLVM.h"
 #include "mlir/Conversion/NVVMToLLVM/NVVMToLLVM.h"
 #include "mlir/Conversion/UBToLLVM/UBToLLVM.h"
-#include "mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h"
 #include "mlir/Dialect/Affine/TransformOps/AffineTransformOps.h"
 #include "mlir/Dialect/Bufferization/TransformOps/BufferizationTransformOps.h"
 #include "mlir/Dialect/Func/Extensions/AllExtensions.h"
@@ -67,7 +66,6 @@ inline void registerAllExtensions(DialectRegistry &registry) {
   ub::registerConvertUBToLLVMInterface(registry);
 
   // Register all transform dialect extensions.
-  amdgpu::registerTransformDialectExtension(registry);
   affine::registerTransformDialectExtension(registry);
   bufferization::registerTransformDialectExtension(registry);
   func::registerTransformDialectExtension(registry);
diff --git a/mlir/lib/Dialect/AMDGPU/CMakeLists.txt b/mlir/lib/Dialect/AMDGPU/CMakeLists.txt
index c47e4c5495c17..31167e6af908b 100644
--- a/mlir/lib/Dialect/AMDGPU/CMakeLists.txt
+++ b/mlir/lib/Dialect/AMDGPU/CMakeLists.txt
@@ -1,4 +1,3 @@
 add_subdirectory(IR)
-add_subdirectory(TransformOps)
 add_subdirectory(Transforms)
 add_subdirectory(Utils)
diff --git a/mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp b/mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp
index 4e72fbf56b80a..2575ad4984814 100644
--- a/mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp
+++ b/mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp
@@ -43,21 +43,6 @@ void AMDGPUDialect::initialize() {
       >();
 }
 
-bool amdgpu::AMDGPUDialect::isSharedMemoryAddressSpace(Attribute memorySpace) {
-  if (!memorySpace)
-    return false;
-  if (auto intAttr = llvm::dyn_cast<IntegerAttr>(memorySpace))
-    return intAttr.getInt() == AMDGPUDialect::kSharedMemoryAddressSpace;
-  if (auto gpuAttr = llvm::dyn_cast<gpu::AddressSpaceAttr>(memorySpace))
-    return gpuAttr.getValue() == gpu::AddressSpace::Workgroup;
-  return false;
-}
-
-bool amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(MemRefType type) {
-  Attribute memorySpace = type.getMemorySpace();
-  return isSharedMemoryAddressSpace(memorySpace);
-}
-
 //===----------------------------------------------------------------------===//
 // 8-bit float ops
 //===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp b/mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp
deleted file mode 100644
index b7e17a9289738..0000000000000
--- a/mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp
+++ /dev/null
@@ -1,67 +0,0 @@
-//===- AMDGPUTransformOps.cpp - Implementation of AMDGPU transform ops-----===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h"
-
-#include "mlir/Dialect/AMDGPU/IR/AMDGPUDialect.h"
-#include "mlir/Dialect/AMDGPU/Transforms/Transforms.h"
-#include "mlir/Dialect/Affine/IR/AffineOps.h"
-#include "mlir/Dialect/Vector/IR/VectorOps.h"
-
-using namespace mlir;
-using namespace mlir::amdgpu;
-using namespace mlir::transform;
-using namespace mlir::func;
-
-#define DEBUG_TYPE "amdgpu-transforms"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define DBGSNL() (llvm::dbgs() << "\n")
-#define LDBG(X) LLVM_DEBUG(DBGS() << (X) << "\n")
-
-DiagnosedSilenceableFailure
-ApplyOptimizeSharedMemoryReadsAndWritesOp::applyToOne(
-    TransformRewriter &rewriter, FuncOp funcOp, ApplyToEachResultList &results,
-    TransformState &state) {
-  optimizeSharedMemoryReadsAndWritesOp(funcOp, getSharedMemoryLineSizeBytes(),
-                                       getDefaultVectorSizeBits());
-  return DiagnosedSilenceableFailure::success();
-}
-
-void ApplyOptimizeSharedMemoryReadsAndWritesOp::getEffects(
-    SmallVectorImpl<MemoryEffects::EffectInstance> &effects) {
-  onlyReadsHandle(getTarget(), effects);
-  modifiesPayload(effects);
-}
-
-//===----------------------------------------------------------------------===//
-// Transform op registration
-//===----------------------------------------------------------------------===//
-
-namespace {
-class AMDGPUTransformDialectExtension
-    : public TransformDialectExtension<AMDGPUTransformDialectExtension> {
-public:
-  AMDGPUTransformDialectExtension() {
-    declareGeneratedDialect<arith::ArithDialect>();
-    declareGeneratedDialect<affine::AffineDialect>();
-    declareGeneratedDialect<amdgpu::AMDGPUDialect>();
-    declareGeneratedDialect<vector::VectorDialect>();
-    registerTransformOps<
-#define GET_OP_LIST
-#include "mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp.inc"
-        >();
-  }
-};
-} // namespace
-
-#define GET_OP_CLASSES
-#include "mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp.inc"
-
-void amdgpu::registerTransformDialectExtension(DialectRegistry &registry) {
-  registry.addExtensions<AMDGPUTransformDialectExtension>();
-}
diff --git a/mlir/lib/Dialect/AMDGPU/TransformOps/CMakeLists.txt b/mlir/lib/Dialect/AMDGPU/TransformOps/CMakeLists.txt
deleted file mode 100644
index c39a3b55eabca..0000000000000
--- a/mlir/lib/Dialect/AMDGPU/TransformOps/CMakeLists.txt
+++ /dev/null
@@ -1,25 +0,0 @@
-add_mlir_dialect_library(MLIRAMDGPUTransformOps
-  AMDGPUTransformOps.cpp
-
-  ADDITIONAL_HEADER_DIRS
-  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/AMDGPU/TransformOps
-
-  DEPENDS
-  MLIRAMDGPUTransformOpsIncGen
-
-  LINK_LIBS PUBLIC
-  MLIRAffineDialect
-  MLIRArithDialect
-  MLIRIR
-  MLIRLinalgDialect
-  MLIRAMDGPUDialect
-  MLIRAMDGPUTransforms
-  MLIRParser
-  MLIRSideEffectInterfaces
-  MLIRSCFDialect
-  MLIRSCFTransforms
-  MLIRTransformDialect
-  MLIRTransformDialectUtils
-  MLIRVectorTransforms
-
-  )
diff --git a/mlir/lib/Dialect/AMDGPU/Transforms/CMakeLists.txt b/mlir/lib/Dialect/AMDGPU/Transforms/CMakeLists.txt
index a955d585b9a1d..0889a21bddc44 100644
--- a/mlir/lib/Dialect/AMDGPU/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/AMDGPU/Transforms/CMakeLists.txt
@@ -1,7 +1,5 @@
 add_mlir_dialect_library(MLIRAMDGPUTransforms
   EmulateAtomics.cpp
-  OptimizeSharedMemory.cpp
-  Utils.cpp
 
   ADDITIONAL_HEADER_DIRS
   {$MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/AMDGPU/Transforms
@@ -19,5 +17,4 @@ add_mlir_dialect_library(MLIRAMDGPUTransforms
   MLIRPass
   MLIRTransforms
   MLIRTransformUtils
-  MLIRVectorDialect
   )
diff --git a/mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp b/mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp
deleted file mode 100644
index 32fab265e03cc..0000000000000
--- a/mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp
+++ /dev/null
@@ -1,261 +0,0 @@
-//===- OptimizeSharedMemory.cpp - MLIR AMDGPU pass implementation ---------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements transforms to optimize accesses to shared memory.
-// It is inspired by
-// https://github.com/llvm/llvm-project/blob/main/mlir/lib/Dialect/NVGPU/Transforms/OptimizeSharedMemory.cpp
-//
-//===----------------------------------------------------------------------===//
-
-#include "mlir/Dialect/AMDGPU/Transforms/Passes.h"
-
-#include "mlir/Dialect/AMDGPU/IR/AMDGPUDialect.h"
-#include "mlir/Dialect/AMDGPU/Transforms/Transforms.h"
-#include "mlir/Dialect/AMDGPU/Transforms/Utils.h"
-#include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Func/IR/FuncOps.h"
-#include "mlir/Dialect/GPU/IR/GPUDialect.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/Vector/IR/VectorOps.h"
-#include "mlir/Interfaces/SideEffectInterfaces.h"
-#include "mlir/Support/LogicalResult.h"
-
-namespace mlir {
-namespace amdgpu {
-#define GEN_PASS_DEF_OPTIMIZESHAREDMEMORY
-#include "mlir/Dialect/AMDGPU/Transforms/Passes.h.inc"
-} // namespace amdgpu
-} // namespace mlir
-
-using namespace mlir;
-using namespace mlir::amdgpu;
-
-/// Uses `srcIndexValue` to permute `tgtIndexValue` via
-/// `result = xor(floordiv(srcIdxVal,permuteEveryN),
-///               floordiv(tgtIdxVal,vectorSize)))
-///            + tgtIdxVal % vectorSize`
-/// This is done using an optimized sequence of `arith` operations.
-static Value permuteVectorOffset(OpBuilder &b, Location loc,
-                                 ArrayRef<Value> indices, MemRefType memrefTy,
-                                 int64_t srcDim, int64_t tgtDim,
-                                 int64_t sharedMemoryLineSizeBytes,
-                                 int64_t defaultVectorSizeBits) {
-  // Adjust the src index to change how often the permutation changes
-  // if necessary.
-  Value src = indices[srcDim];
-
-  // We only want to permute every N iterations of the target dim where N is
-  // ceil(sharedMemoryLineSizeBytes / dimSizeBytes(tgtDim)).
-  const int64_t permuteEveryN = std::max<int64_t>(
-      1, sharedMemoryLineSizeBytes / ((memrefTy.getDimSize(tgtDim) *
-                                       memrefTy.getElementTypeBitWidth()) /
-                                      8));
-
-  // clang-format off
-  // Index bit representation (b0 = least significant bit) for dim(1)
-  // of a `memref<?x?xDT>` is as follows:
-  // N := log2(128/elementSizeBits)
-  // M := log2(dimSize(1))
-  // then
-  // bits[0:N] = sub-vector element offset
-  // bits[N:M] = vector index
-  // clang-format on
-  int64_t n =
-      llvm::Log2_64(defaultVectorSizeBits / memrefTy.getElementTypeBitWidth());
-  int64_t m = llvm::Log2_64(memrefTy.getDimSize(tgtDim));
-
-  // Capture bits[0:(M-N)] of src by first creating a (M-N) mask.
-  int64_t mask = (1LL << (m - n)) - 1;
-  if (permuteEveryN > 1)
-    mask = mask << llvm::Log2_64(permuteEveryN);
-  Value srcBits = b.create<arith::ConstantIndexOp>(loc, mask);
-  srcBits = b.create<arith::AndIOp>(loc, src, srcBits);
-
-  /// Use the src bits to permute the target bits b[N:M] containing the
-  /// vector offset.
-  if (permuteEveryN > 1) {
-    int64_t shlBits = n - llvm::Log2_64(permuteEveryN);
-    if (shlBits > 0) {
-      Value finalShiftVal = b.create<arith::ConstantIndexOp>(loc, shlBits);
-      srcBits = b.createOrFold<arith::ShLIOp>(loc, srcBits, finalShiftVal);
-    } else if (shlBits < 0) {
-      Value finalShiftVal = b.create<arith::ConstantIndexOp>(loc, -1 * shlBits);
-      srcBits = b.createOrFold<arith::ShRUIOp>(loc, srcBits, finalShiftVal);
-    }
-  } else {
-    Value finalShiftVal = b.create<arith::ConstantIndexOp>(loc, n);
-    srcBits = b.createOrFold<arith::ShLIOp>(loc, srcBits, finalShiftVal);
-  }
-
-  Value permutedVectorIdx =
-      b.create<arith::XOrIOp>(loc, indices[tgtDim], srcBits);
-  return permutedVectorIdx;
-}
-
-static void transformIndices(OpBuilder &builder, Location loc,
-                             SmallVector<Value, 4> &indices,
-                             MemRefType memrefTy, int64_t srcDim,
-                             int64_t tgtDim, int64_t sharedMemoryLineSizeBytes,
-                             int64_t defaultVectorSizeBits) {
-  indices[tgtDim] =
-      permuteVectorOffset(builder, loc, indices, memrefTy, srcDim, tgtDim,
-                          sharedMemoryLineSizeBytes, defaultVectorSizeBits);
-}
-
-// Return all operations within `parentOp` that read from or write to
-// `shmMemRef`.
-static LogicalResult
-getShmReadAndWriteOps(Operation *parentOp, Value shmMemRef,
-                      SmallVector<Operation *, 16> &readOps,
-                      SmallVector<Operation *, 16> &writeOps) {
-  parentOp->walk([&](Operation *op) {
-    MemoryEffectOpInterface iface = dyn_cast<MemoryEffectOpInterface>(op);
-    if (!iface)
-      return;
-    std::optional<MemoryEffects::EffectInstance> effect =
-        iface.getEffectOnValue<MemoryEffects::Read>(shmMemRef);
-    if (effect) {
-      readOps.push_back(op);
-      return;
-    }
-    effect = iface.getEffectOnValue<MemoryEffects::Write>(shmMemRef);
-    if (effect)
-      writeOps.push_back(op);
-  });
-
-  // Restrict to a supported set of ops. We also require at least 2D access,
-  // although this could be relaxed.
-  if (llvm::any_of(readOps, [](Operation *op) {
-        return !isa<memref::LoadOp, vector::LoadOp, vector::TransferReadOp>(
-                   op) ||
-               amdgpu::getIndices(op)->size() < 2;
-      }))
-    return failure();
-  if (llvm::any_of(writeOps, [](Operation *op) {
-        return !isa<memref::StoreOp, vector::StoreOp, vector::TransferWriteOp>(
-                   op) ||
-               amdgpu::getIndices(op)->size() < 2;
-      }))
-    return failure();
-
-  return success();
-}
-
-LogicalResult amdgpu::optimizeSharedMemoryReadsAndWrites(
-    Operation *parentOp, Value memrefValue, int64_t sharedMemoryLineSizeBytes,
-    int64_t defaultVectorSizeBits) {
-  auto memRefType = dyn_cast<MemRefType>(memrefValue.getType());
-  if (!memRefType ||
-      !amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(memRefType))
-    return failure();
-
-  // Abort if the given value has any sub-views; we do not do any alias
-  // analysis.
-  bool hasSubView = false;
-  parentOp->walk([&](memref::SubViewOp subView) { hasSubView = true; });
-  if (hasSubView)
-    return failure();
-
-  // Check if this is necessary given the assumption of 128b accesses:
-  // If dim[rank-1] is small enough to fit 8 rows in a 128B line.
-  const int64_t rowSize = memRefType.getDimSize(memRefType.getRank() - 1);
-  const int64_t rowsPerLine =
-      (8 * sharedMemoryLineSizeBytes / memRefType.getElementTypeBitWidth()) /
-      rowSize;
-  const int64_t threadGroupSize =
-      1LL << (7 - llvm::Log2_64(defaultVectorSizeBits / 8));
-  if (rowsPerLine >= threadGroupSize)
-    return failure();
-
-  // Get sets of operations within the function that read/write to shared
-  // memory.
-  SmallVector<Operation *, 16> shmReadOps;
-  SmallVector<Operation *, 16> shmWriteOps;
-  if (failed(getShmReadAndWriteOps(parentOp, memrefValue, shmReadOps,
-                                   shmWriteOps)))
-    return failure();
-
-  if (shmReadOps.empty() || shmWriteOps.empty())
-    return failure();
-
-  OpBuilder builder(parentOp->getContext());
-
-  int64_t tgtDim = memRefType.getRank() - 1;
-  int64_t srcDim = memRefType.getRank() - 2;
-
-  // Transform indices for the ops writing to shared memory.
-  while (!shmWriteOps.empty()) {
-    Operation *shmWriteOp = shmWriteOps.pop_back_val();
-    builder.setInsertionPoint(shmWriteOp);
-
-    auto indices = amdgpu::getIndices(shmWriteOp);
-    SmallVector<Value, 4> transformedIndices(indices->begin(), indices->end());
-    transformIndices(builder, shmWriteOp->getLoc(), transformedIndices,
-                     memRefType, srcDim, tgtDim, sharedMemoryLineSizeBytes,
-                     defaultVectorSizeBits);
-    amdgpu::setIndices(shmWriteOp, transformedIndices);
-  }
-
-  // Transform indices for the ops reading from shared memory.
-  while (!shmReadOps.empty()) {
-    Operation *shmReadOp = shmReadOps.pop_back_val();
-    builder.setInsertionPoint(shmReadOp);
-
-    auto indices = amdgpu::getIndices(shmReadOp);
-    SmallVector<Value, 4> transformedIndices(indices->begin(), indices->end());
-    transformIndices(builder, shmReadOp->getLoc(), transformedIndices,
-                     memRefType, srcDim, tgtDim, sharedMemoryLineSizeBytes,
-                     defaultVectorSizeBits);
-    amdgpu::setIndices(shmReadOp, transformedIndices);
-  }
-
-  return success();
-}
-
-std::optional<LogicalResult>
-amdgpu::optimizeSharedMemoryReadsAndWritesOp(func::FuncOp funcOp,
-                                             int64_t sharedMemoryLineSizeBytes,
-                                             int64_t defaultVectorSizeBits) {
-  SmallVector<memref::AllocOp> shmAllocOps;
-  funcOp.walk([&](memref::AllocOp allocOp) {
-    if (!amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(allocOp.getType()))
-      return;
-    shmAllocOps.push_back(allocOp);
-  });
-  for (auto allocOp : shmAllocOps) {
-    if (failed(amdgpu::optimizeSharedMemoryReadsAndWrites(
-            funcOp, allocOp.getMemref(), sharedMemoryLineSizeBytes,
-            defaultVectorSizeBits)))
-      return failure();
-  }
-  return success();
-}
-
-struct OptimizeSharedMemoryPass
-    : public amdgpu::impl::OptimizeSharedMemoryBase<OptimizeSharedMemoryPass> {
-public:
-  OptimizeSharedMemoryPass() = default;
-  OptimizeSharedMemoryPass(const OptimizeSharedMemoryOptions &options)
-      : OptimizeSharedMemoryBase(options) {}
-  void runOnOperation() override {
-    Operation *op = getOperation();
-    SmallVector<memref::AllocOp> shmAllocOps;
-    op->walk([&](memref::AllocOp allocOp) {
-      if (!amdgpu::AMDGPUDialect::hasSharedMemoryAddressSpace(
-              allocOp.getType()))
-        return;
-      shmAllocOps.push_back(allocOp);
-    });
-    for (auto allocOp : shmAllocOps) {
-      if (failed(optimizeSharedMemoryReadsAndWrites(op, allocOp.getMemref(),
-                                                    sharedMemoryLineSizeBytes,
-                                                    defaultVectorSizeBits)))
-        return;
-    }
-  }
-};
diff --git a/mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp b/mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp
deleted file mode 100644
index 8163eeafdf1f0..0000000000000
--- a/mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp
+++ /dev/null
@@ -1,39 +0,0 @@
-#include "mlir/Dialect/AMDGPU/Transforms/Utils.h"
-
-#include "mlir/Dialect/AMDGPU/IR/AMDGPUDialect.h"
-#include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/Vector/IR/VectorOps.h"
-
-using namespace mlir;
-using namespace mlir::amdgpu;
-
-std::optional<Operation::operand_range> amdgpu::getIndices(Operation *op) {
-  if (auto loadOp = dyn_cast<memref::LoadOp>(op))
-    return loadOp.getIndices();
-  if (auto storeOp = dyn_cast<memref::StoreOp>(op))
-    return storeOp.getIndices();
-  if (auto vectorReadOp = dyn_cast<vector::LoadOp>(op))
-    return vectorReadOp.getIndices();
-  if (auto vectorStoreOp = dyn_cast<vector::StoreOp>(op))
-    return vectorStoreOp.getIndices();
-  if (auto transferReadOp = dyn_cast<vector::TransferReadOp>(op))
-    return transferReadOp.getIndices();
-  if (auto transferWriteOp = dyn_cast<vector::TransferWriteOp>(op))
-    return transferWriteOp.getIndices();
-  return std::nullopt;
-}
-
-void amdgpu::setIndices(Operation *op, ArrayRef<Value> indices) {
-  if (auto loadOp = dyn_cast<memref::LoadOp>(op))
-    return loadOp.getIndicesMutable().assign(indices);
-  if (auto storeOp = dyn_cast<memref::StoreOp>(op))
-    return storeOp.getIndicesMutable().assign(indices);
-  if (auto vectorReadOp = dyn_cast<vector::LoadOp>(op))
-    return vectorReadOp.getIndicesMutable().assign(indices);
-  if (auto vectorStoreOp = dyn_cast<vector::StoreOp>(op))
-    return vectorStoreOp.getIndicesMutable().assign(indices);
-  if (auto transferReadOp = dyn_cast<vector::TransferReadOp>(op))
-    return transferReadOp.getIndicesMutable().assign(indices);
-  if (auto transferWriteOp = dyn_cast<vector::TransferWriteOp>(op))
-    return transferWriteOp.getIndicesMutable().assign(indices);
-}
diff --git a/mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir b/mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir
deleted file mode 100644
index 983eee732e2af..0000000000000
--- a/mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir
+++ /dev/null
@@ -1,50 +0,0 @@
-// RUN: mlir-opt %s --pass-pipeline='builtin.module(func.func(amdgpu-optimize-shared-memory))' | FileCheck %s
-  
-  // CHECK: @optimize_shmem([[arg0:%.+]]: memref<{{.*}}>, [[readRow:%.+]]: index, [[readCol:%.+]]: index, [[writeRow:%.+]]: index, [[writeCol:%.+]]: index, [[fragRow:%.+]]: index, [[fragCol:%.+]]: index, [[fragColPerm:%.+]]: index, [[stRow:%.+]]: index, [[stCol:%.+]]: index)
-  func.func @optimize_shmem(%arg0: memref<4096x4096xf16>,
-                    %readRow: index, %readCol: index,
-                    %writeRow: index, %writeCol: index,
-                    %fragRow: index, %fragCol: index,
-                    %fragColPerm: index,
-                    %stRow: index, %stCol: index) {
-    // CHECK:    %[[cst:.+]] = arith.constant 0.000000e+00 : f16
-    %cst = arith.constant 0.000000e+00 : f16
-
-    // CHECK: [[shmA:%.+]] = memref.alloc
-    // CHECK: [[shmB:%.+]] = memref.alloc
-    %shmA = memref.alloc() {alignment = 64 : i64} : memref<128x32xf16, 3>
-    %shmB = memref.alloc() {alignment = 64 : i64} : memref<256x32xf16, 3>
-
-    %0 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
-    vector.transfer_write %0, %shmB[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<256x32xf16, 3>
-    gpu.barrier
-    gpu.barrier
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]] 
-    %1 = vector.load %shmB[%fragRow, %fragColPerm] : memref<256x32xf16, 3>, vector<8xf16>
-
-    %2 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
-    vector.transfer_write %2, %shmA[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<128x32xf16, 3>
-    gpu.barrier
-    gpu.barrier
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]]
-    %3 = vector.load %shmA[%fragRow, %fragColPerm] : memref<128x32xf16, 3>, vector<8xf16>
-    return
-  }
diff --git a/mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir b/mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir
deleted file mode 100644
index b1bb91ffc2972..0000000000000
--- a/mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir
+++ /dev/null
@@ -1,54 +0,0 @@
-// RUN: mlir-opt %s -transform-interpreter | FileCheck %s
-
-  // CHECK: @optimize_shmem([[arg0:%.+]]: memref<{{.*}}>, [[readRow:%.+]]: index, [[readCol:%.+]]: index, [[writeRow:%.+]]: index, [[writeCol:%.+]]: index, [[fragRow:%.+]]: index, [[fragCol:%.+]]: index, [[fragColPerm:%.+]]: index, [[stRow:%.+]]: index, [[stCol:%.+]]: index)
-  func.func @optimize_shmem(%arg0: memref<4096x4096xf16>,
-                    %readRow: index, %readCol: index,
-                    %writeRow: index, %writeCol: index,
-                    %fragRow: index, %fragCol: index,
-                    %fragColPerm: index,
-                    %stRow: index, %stCol: index) {
-    %cst = arith.constant 0.000000e+00 : f16
-
-    %shmA = memref.alloc() {alignment = 64 : i64} : memref<128x32xf16, 3>
-    %shmB = memref.alloc() {alignment = 64 : i64} : memref<256x32xf16, 3>
-
-    %0 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
-    vector.transfer_write %0, %shmB[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<256x32xf16, 3>
-    gpu.barrier
-    gpu.barrier
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]]
-    %1 = vector.load %shmB[%fragRow, %fragColPerm] : memref<256x32xf16, 3>, vector<8xf16>
-    %2 = vector.transfer_read %arg0[%readRow, %readCol], %cst {in_bounds = [true, true]} : memref<4096x4096xf16>, vector<1x8xf16>
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[stRow:%.+]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[stColPerm:%.+]] = arith.xori [[stCol:%.+]], [[xorBits]]
-    vector.transfer_write %2, %shmA[%writeRow, %writeCol] {in_bounds = [true, true]} : vector<1x8xf16>, memref<128x32xf16, 3>
-    gpu.barrier
-    gpu.barrier
-    // CHECK: [[c6:%.+]] = arith.constant 6 : index
-    // CHECK: [[srcBits:%.+]] = arith.andi [[fragRow]], [[c6]]
-    // CHECK: [[c2:%.+]] = arith.constant 2 : index
-    // CHECK: [[xorBits:%.+]] = arith.shli [[srcBits]], [[c2]]
-    // CHECK: [[fragColPerm:%.+]] = arith.xori [[fragCol:%.+]], [[xorBits]]
-    %3 = vector.load %shmA[%fragRow, %fragColPerm] : memref<128x32xf16, 3>, vector<8xf16>
-    return
-  }
-
-module attributes { transform.with_named_sequence } {
-  transform.named_sequence @__transform_main(%root: !transform.any_op {transform.readonly}) {
-    %0 = transform.structured.match ops{["func.func"]} in %root : (!transform.any_op) -> !transform.any_op
-    transform.amdgpu.optimize_shared_memory_reads_and_writes %0 {sharedMemoryLineSizeBytes = 128, defaultVectorSizeBits = 128}: (!transform.any_op) -> ()
-    transform.yield
-  } // @__transform_main
-} // module
diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 497edcfceffe4..67052fcd39930 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -1552,58 +1552,6 @@ cc_library(
     ],
 )
 
-cc_library(
-    name = "AMDGPUTransformOps",
-    srcs = glob([
-        "lib/Dialect/AMDGPU/TransformOps/*.cpp",
-    ]),
-    hdrs = glob([
-        "include/mlir/Dialect/AMDGPU/TransformOps/*.h",
-    ]),
-    includes = ["include"],
-    deps = [
-        ":AMDGPUDialect",
-        ":AMDGPUTransformOpsIncGen",
-        ":AMDGPUTransforms",
-        ":AffineDialect",
-        ":FuncDialect",
-        ":IR",
-        ":TransformDialect",
-        ":TransformDialectInterfaces",
-        ":VectorDialect",
-    ],
-)
-
-td_library(
-    name = "AMDGPUTransformOpsTdFiles",
-    srcs = glob([
-        "include/mlir/Dialect/AMDGPU/TransformOps/*.td",
-    ]),
-    includes = ["include"],
-    deps = [
-        ":TransformDialectTdFiles",
-    ],
-)
-
-gentbl_cc_library(
-    name = "AMDGPUTransformOpsIncGen",
-    tbl_outs = [
-        (
-            ["-gen-op-decls"],
-            "include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h.inc",
-        ),
-        (
-            ["-gen-op-defs"],
-            "include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp.inc",
-        ),
-    ],
-    tblgen = ":mlir-tblgen",
-    td_file = "include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td",
-    deps = [
-        ":AMDGPUTransformOpsTdFiles",
-    ],
-)
-
 gentbl_cc_library(
     name = "AMDGPUPassIncGen",
     tbl_outs = [
@@ -4787,7 +4735,6 @@ cc_library(
     name = "AllExtensions",
     hdrs = ["include/mlir/InitAllExtensions.h"],
     deps = [
-        ":AMDGPUTransformOps",
         ":AffineTransformOps",
         ":ArithToLLVM",
         ":BufferizationTransformOps",
@@ -9033,7 +8980,6 @@ cc_library(
     deps = [
         ":AMDGPUDialect",
         ":AMDGPUToROCDL",
-        ":AMDGPUTransformOps",
         ":AMDGPUTransforms",
         ":AMXDialect",
         ":AMXTransforms",

From 198ffb85314f7741ed048de67d68ca83bb30e16e Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Thu, 11 Apr 2024 11:23:24 -0400
Subject: [PATCH 146/886] [Clang][Sema] Implement approved resolution for
 CWG2858 (#88042)

The approved resolution for CWG2858 changes
[expr.prim.id.qual] p2 sentence 2 to read:
> A declarative _nested-name-specifier_ shall not have a
_computed-type-specifier_.

This patch implements the approved resolution. Since we don't consider
_nested-name-specifiers_ in friend declarations to be declarative (yet),
it currently isn't possible to write a test that would produce this
diagnostic (`diagnoseQualifiedDeclaration` is never called if the
`DeclContext` can't be computed). Nevertheless, tests were added which
will produce the diagnostic once we start calling
`diagnoseQualifiedDeclaration` for friend declarations.
---
 clang/docs/ReleaseNotes.rst                   |  3 +++
 .../clang/Basic/DiagnosticSemaKinds.td        |  7 +++---
 clang/lib/Sema/SemaDecl.cpp                   | 13 +++++------
 clang/test/CXX/dcl.decl/dcl.meaning/p1-0x.cpp |  4 ++--
 clang/test/CXX/drs/dr28xx.cpp                 | 23 +++++++++++++++++++
 clang/test/Parser/cxx-class.cpp               |  6 ++---
 6 files changed, 40 insertions(+), 16 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index c4a4893aec5cd..93318871fa9f6 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -147,6 +147,9 @@ Resolutions to C++ Defect Reports
   compatibility of two types.
   (`CWG2759: [[no_unique_address] and common initial sequence  <https://cplusplus.github.io/CWG/issues/2759.html>`_).
 
+- Clang now diagnoses declarative nested-name-specifiers with pack-index-specifiers.
+  (`CWG2858: Declarative nested-name-specifiers and pack-index-specifiers <https://cplusplus.github.io/CWG/issues/2858.html>`_).
+
 C Language Changes
 ------------------
 
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 059a8f58da5db..180e913155d67 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -2402,10 +2402,6 @@ def err_selected_explicit_constructor : Error<
 def note_explicit_ctor_deduction_guide_here : Note<
   "explicit %select{constructor|deduction guide}0 declared here">;
 
-// C++11 decltype
-def err_decltype_in_declarator : Error<
-    "'decltype' cannot be used to name a declaration">;
-
 // C++11 auto
 def warn_cxx98_compat_auto_type_specifier : Warning<
   "'auto' type specifier is incompatible with C++98">,
@@ -8313,6 +8309,9 @@ def ext_template_after_declarative_nns : ExtWarn<
 def ext_alias_template_in_declarative_nns : ExtWarn<
   "a declarative nested name specifier cannot name an alias template">,
   InGroup<DiagGroup<"alias-template-in-declaration-name">>;
+def err_computed_type_in_declarative_nns  : Error<
+  "a %select{pack indexing|'decltype'}0 specifier cannot be used in "
+  "a declarative nested name specifier">;
 
 def err_no_typeid_with_fno_rtti : Error<
   "use of typeid requires -frtti">;
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 5a23179dfbbf4..a4699d6ba2c73 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -6335,16 +6335,15 @@ bool Sema::diagnoseQualifiedDeclaration(CXXScopeSpec &SS, DeclContext *DC,
         if (TST->isDependentType() && TST->isTypeAlias())
           Diag(Loc, diag::ext_alias_template_in_declarative_nns)
               << SpecLoc.getLocalSourceRange();
-      } else if (T->isDecltypeType()) {
+      } else if (T->isDecltypeType() || T->getAsAdjusted<PackIndexingType>()) {
         // C++23 [expr.prim.id.qual]p2:
         //   [...] A declarative nested-name-specifier shall not have a
-        //   decltype-specifier.
+        //   computed-type-specifier.
         //
-        // FIXME: This wording appears to be defective as it does not forbid
-        // declarative nested-name-specifiers with pack-index-specifiers.
-        // See https://github.com/cplusplus/CWG/issues/499.
-        Diag(Loc, diag::err_decltype_in_declarator)
-            << SpecLoc.getTypeLoc().getSourceRange();
+        // CWG2858 changed this from 'decltype-specifier' to
+        // 'computed-type-specifier'.
+        Diag(Loc, diag::err_computed_type_in_declarative_nns)
+            << T->isDecltypeType() << SpecLoc.getTypeLoc().getSourceRange();
       }
     }
   } while ((SpecLoc = SpecLoc.getPrefix()));
diff --git a/clang/test/CXX/dcl.decl/dcl.meaning/p1-0x.cpp b/clang/test/CXX/dcl.decl/dcl.meaning/p1-0x.cpp
index fbe9c0895aeae..13be079a40bc3 100644
--- a/clang/test/CXX/dcl.decl/dcl.meaning/p1-0x.cpp
+++ b/clang/test/CXX/dcl.decl/dcl.meaning/p1-0x.cpp
@@ -6,8 +6,8 @@ class foo {
   void func();
 };
 
-int decltype(foo())::i; // expected-error{{'decltype' cannot be used to name a declaration}}
-void decltype(foo())::func() { // expected-error{{'decltype' cannot be used to name a declaration}}
+int decltype(foo())::i; // expected-error{{a 'decltype' specifier cannot be used in a declarative nested name specifier}}
+void decltype(foo())::func() { // expected-error{{a 'decltype' specifier cannot be used in a declarative nested name specifier}}
 }
 
 
diff --git a/clang/test/CXX/drs/dr28xx.cpp b/clang/test/CXX/drs/dr28xx.cpp
index 7f72003d66f1e..9b21d3410a049 100644
--- a/clang/test/CXX/drs/dr28xx.cpp
+++ b/clang/test/CXX/drs/dr28xx.cpp
@@ -58,3 +58,26 @@ void B<int>::g() requires true;
 #endif
 
 } // namespace dr2847
+
+namespace dr2858 { // dr2858: 19
+
+#if __cplusplus > 202302L
+
+template<typename... Ts>
+struct A {
+  // FIXME: The nested-name-specifier in the following friend declarations are declarative,
+  // but we don't treat them as such (yet).
+  friend void Ts...[0]::f();
+  template<typename U>
+  friend void Ts...[0]::g();
+
+  friend struct Ts...[0]::B;
+  // FIXME: The index of the pack-index-specifier is printed as a memory address in the diagnostic.
+  template<typename U>
+  friend struct Ts...[0]::C;
+  // expected-warning-re@-1 {{dependent nested name specifier 'Ts...[{{.*}}]::' for friend template declaration is not supported; ignoring this friend declaration}}
+};
+
+#endif
+
+} // namespace dr2858
diff --git a/clang/test/Parser/cxx-class.cpp b/clang/test/Parser/cxx-class.cpp
index 046d2dd580f02..c90c7e030a8bd 100644
--- a/clang/test/Parser/cxx-class.cpp
+++ b/clang/test/Parser/cxx-class.cpp
@@ -59,14 +59,14 @@ typedef union {
   } y;
 } bug3177;
 
-// check that we don't consume the token after the access specifier 
+// check that we don't consume the token after the access specifier
 // when it's not a colon
 class D {
 public // expected-error{{expected ':'}}
   int i;
 };
 
-// consume the token after the access specifier if it's a semicolon 
+// consume the token after the access specifier if it's a semicolon
 // that was meant to be a colon
 class E {
 public; // expected-error{{expected ':'}}
@@ -281,7 +281,7 @@ struct A {} ::PR41192::a; // ok, no missing ';' here  expected-warning {{extra q
 #if __cplusplus >= 201103L
 struct C;
 struct D { static C c; };
-struct C {} decltype(D())::c; // expected-error {{'decltype' cannot be used to name a declaration}}
+struct C {} decltype(D())::c; // expected-error {{a 'decltype' specifier cannot be used in a declarative nested name specifier}}
 #endif
 }
 

From 298ea9bfd50ca41c77e45065700df06adb6264ae Mon Sep 17 00:00:00 2001
From: Raghu Maddhipatla <7686592+raghavendhra@users.noreply.github.com>
Date: Thu, 11 Apr 2024 10:26:54 -0500
Subject: [PATCH 147/886]  [Flang] [OpenMP] [MLIR] [Lowering] Add lowering
 support for IS_DEVICE_PTR and HAS_DEVICE_ADDR clauses on OMP TARGET
 directive.  (#88206)

Added lowering support for IS_DEVICE_PTR and HAS_DEVICE_ADDR clauses for
OMP TARGET directive and added related tests for these changes.

IS_DEVICE_PTR and HAS_DEVICE_ADDR clauses apply to OMP TARGET directive
OpenMP spec states

The **is_device_ptr** clause indicates that its list items are device
pointers.

The **has_device_addr** clause indicates that its list items already
have device addresses and therefore they may be directly accessed from a
target device.

Whereas USE_DEVICE_PTR and USE_DEVICE_ADDR clauses apply to OMP TARGET
DATA directive and OpenMP spec for them states

Each list item in the **use_device_ptr** clause results in a new list
item that is a device pointer that refers to a device address

Each list item in a **use_device_addr** clause that is present in the
device data environment is treated as if it is implicitly mapped by a
map clause on the construct with a map-type of alloc

Fixed build error caused by Squash merge which needs rebase
---
 flang/lib/Lower/OpenMP/ClauseProcessor.cpp    | 29 +++++++++++++
 flang/lib/Lower/OpenMP/ClauseProcessor.h      | 12 ++++++
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 22 +++++++---
 flang/test/Lower/OpenMP/FIR/target.f90        | 43 ++++++++++++++++++-
 .../Dialect/OpenMP/OpenMPClauseOperands.h     | 16 ++++---
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td | 18 ++++++--
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  |  9 ++--
 mlir/test/Dialect/OpenMP/invalid.mlir         |  2 +-
 mlir/test/Dialect/OpenMP/ops.mlir             |  8 ++--
 9 files changed, 136 insertions(+), 23 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index 0a57a1496289f..fb24c8d1fe3eb 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -751,6 +751,20 @@ bool ClauseProcessor::processDepend(
       });
 }
 
+bool ClauseProcessor::processHasDeviceAddr(
+    llvm::SmallVectorImpl<mlir::Value> &operands,
+    llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
+    llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &isDeviceSymbols)
+    const {
+  return findRepeatableClause<omp::clause::HasDeviceAddr>(
+      [&](const omp::clause::HasDeviceAddr &devAddrClause,
+          const Fortran::parser::CharBlock &) {
+        addUseDeviceClause(converter, devAddrClause.v, operands, isDeviceTypes,
+                           isDeviceLocs, isDeviceSymbols);
+      });
+}
+
 bool ClauseProcessor::processIf(
     omp::clause::If::DirectiveNameModifier directiveName,
     mlir::Value &result) const {
@@ -771,6 +785,20 @@ bool ClauseProcessor::processIf(
   return found;
 }
 
+bool ClauseProcessor::processIsDevicePtr(
+    llvm::SmallVectorImpl<mlir::Value> &operands,
+    llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
+    llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &isDeviceSymbols)
+    const {
+  return findRepeatableClause<omp::clause::IsDevicePtr>(
+      [&](const omp::clause::IsDevicePtr &devPtrClause,
+          const Fortran::parser::CharBlock &) {
+        addUseDeviceClause(converter, devPtrClause.v, operands, isDeviceTypes,
+                           isDeviceLocs, isDeviceSymbols);
+      });
+}
+
 bool ClauseProcessor::processLink(
     llvm::SmallVectorImpl<DeclareTargetCapturePair> &result) const {
   return findRepeatableClause<omp::clause::Link>(
@@ -993,6 +1021,7 @@ bool ClauseProcessor::processUseDevicePtr(
                            useDeviceLocs, useDeviceSymbols);
       });
 }
+
 } // namespace omp
 } // namespace lower
 } // namespace Fortran
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index d31d6a5c20623..df8f4f5310fcb 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -66,6 +66,12 @@ class ClauseProcessor {
   bool processDeviceType(mlir::omp::DeclareTargetDeviceType &result) const;
   bool processFinal(Fortran::lower::StatementContext &stmtCtx,
                     mlir::Value &result) const;
+  bool
+  processHasDeviceAddr(llvm::SmallVectorImpl<mlir::Value> &operands,
+                       llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
+                       llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
+                       llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
+                           &isDeviceSymbols) const;
   bool processHint(mlir::IntegerAttr &result) const;
   bool processMergeable(mlir::UnitAttr &result) const;
   bool processNowait(mlir::UnitAttr &result) const;
@@ -104,6 +110,12 @@ class ClauseProcessor {
   bool processIf(omp::clause::If::DirectiveNameModifier directiveName,
                  mlir::Value &result) const;
   bool
+  processIsDevicePtr(llvm::SmallVectorImpl<mlir::Value> &operands,
+                     llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
+                     llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
+                     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
+                         &isDeviceSymbols) const;
+  bool
   processLink(llvm::SmallVectorImpl<DeclareTargetCapturePair> &result) const;
 
   // This method is used to process a map clause.
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 340921c867246..50ad889052ab0 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -1294,6 +1294,11 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
   llvm::SmallVector<mlir::Type> mapSymTypes;
   llvm::SmallVector<mlir::Location> mapSymLocs;
   llvm::SmallVector<const Fortran::semantics::Symbol *> mapSymbols;
+  llvm::SmallVector<mlir::Value> devicePtrOperands, deviceAddrOperands;
+  llvm::SmallVector<mlir::Type> devicePtrTypes, deviceAddrTypes;
+  llvm::SmallVector<mlir::Location> devicePtrLocs, deviceAddrLocs;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> devicePtrSymbols,
+      deviceAddrSymbols;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
   cp.processIf(llvm::omp::Directive::OMPD_target, ifClauseOperand);
@@ -1303,11 +1308,15 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
   cp.processNowait(nowaitAttr);
   cp.processMap(currentLocation, directive, stmtCtx, mapOperands, &mapSymTypes,
                 &mapSymLocs, &mapSymbols);
+  cp.processIsDevicePtr(devicePtrOperands, devicePtrTypes, devicePtrLocs,
+                        devicePtrSymbols);
+  cp.processHasDeviceAddr(deviceAddrOperands, deviceAddrTypes, deviceAddrLocs,
+                          deviceAddrSymbols);
 
-  cp.processTODO<clause::Private, clause::Firstprivate, clause::IsDevicePtr,
-                 clause::HasDeviceAddr, clause::Reduction, clause::InReduction,
-                 clause::Allocate, clause::UsesAllocators, clause::Defaultmap>(
-      currentLocation, llvm::omp::Directive::OMPD_target);
+  cp.processTODO<clause::Private, clause::Firstprivate, clause::Reduction,
+                 clause::InReduction, clause::Allocate, clause::UsesAllocators,
+                 clause::Defaultmap>(currentLocation,
+                                     llvm::omp::Directive::OMPD_target);
 
   // 5.8.1 Implicit Data-Mapping Attribute Rules
   // The following code follows the implicit data-mapping rules to map all the
@@ -1400,7 +1409,8 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
           ? nullptr
           : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
                                  dependTypeOperands),
-      dependOperands, nowaitAttr, mapOperands);
+      dependOperands, nowaitAttr, devicePtrOperands, deviceAddrOperands,
+      mapOperands);
 
   genBodyOfTargetOp(converter, semaCtx, eval, genNested, targetOp, mapSymTypes,
                     mapSymLocs, mapSymbols, currentLocation);
@@ -2059,6 +2069,8 @@ genOMP(Fortran::lower::AbstractConverter &converter,
         !std::get_if<Fortran::parser::OmpClause::Map>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::UseDevicePtr>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::UseDeviceAddr>(&clause.u) &&
+        !std::get_if<Fortran::parser::OmpClause::IsDevicePtr>(&clause.u) &&
+        !std::get_if<Fortran::parser::OmpClause::HasDeviceAddr>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::ThreadLimit>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::NumTeams>(&clause.u)) {
       TODO(clauseLocation, "OpenMP Block construct clause");
diff --git a/flang/test/Lower/OpenMP/FIR/target.f90 b/flang/test/Lower/OpenMP/FIR/target.f90
index 821196b83c3b9..022327f9c25da 100644
--- a/flang/test/Lower/OpenMP/FIR/target.f90
+++ b/flang/test/Lower/OpenMP/FIR/target.f90
@@ -506,4 +506,45 @@ subroutine omp_target_parallel_do
    !CHECK: omp.terminator
    !CHECK: }
    !$omp end target parallel do
- end subroutine omp_target_parallel_do
+end subroutine omp_target_parallel_do
+
+!===============================================================================
+! Target `is_device_ptr` clause
+!===============================================================================
+
+!CHECK-LABEL: func.func @_QPomp_target_is_device_ptr() {
+subroutine omp_target_is_device_ptr
+   use iso_c_binding, only : c_ptr, c_loc
+   !CHECK: %[[VAL_0:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}> {bindc_name = "a", uniq_name = "_QFomp_target_is_device_ptrEa"}
+   type(c_ptr) :: a
+   !CHECK: %[[VAL_1:.*]] = fir.alloca i32 {bindc_name = "b", fir.target, uniq_name = "_QFomp_target_is_device_ptrEb"}
+   integer, target :: b
+   !CHECK: %[[MAP_0:.*]] = omp.map.info var_ptr(%[[DEV_PTR:.*]] : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) map_clauses(tofrom) capture(ByRef) -> !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>> {name = "a"}
+   !CHECK: %[[MAP_1:.*]] = omp.map.info var_ptr(%[[VAL_0:.*]] : !fir.ref<i32>, i32)   map_clauses(tofrom) capture(ByRef) -> !fir.ref<i32> {name = "b"}
+   !CHECK: %[[MAP_2:.*]] = omp.map.info var_ptr(%[[DEV_PTR:.*]] : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) map_clauses(implicit, exit_release_or_enter_alloc) capture(ByRef) -> !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>> {name = "a"}
+   !CHECK: omp.target is_device_ptr(%[[DEV_PTR:.*]] : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) map_entries(%[[MAP_0:.*]] -> %[[ARG0:.*]], %[[MAP_1:.*]] -> %[[ARG1:.*]], %[[MAP_2:.*]] -> %[[ARG2:.*]] : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.ref<i32>, !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) {
+   !CHECK: ^bb0(%[[ARG0]]: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, %[[ARG1]]: !fir.ref<i32>, %[[ARG2]]: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>):
+   !$omp target map(tofrom: a,b) is_device_ptr(a)
+      !CHECK: {{.*}} = fir.coordinate_of %[[VAL_0:.*]], {{.*}} : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
+      a = c_loc(b)
+   !CHECK: omp.terminator
+   !$omp end target
+   !CHECK: }
+end subroutine omp_target_is_device_ptr
+
+ !===============================================================================
+ ! Target `has_device_addr` clause
+ !===============================================================================
+
+ !CHECK-LABEL: func.func @_QPomp_target_has_device_addr() {
+ subroutine omp_target_has_device_addr
+   !CHECK: %[[VAL_0:.*]] = fir.alloca !fir.box<!fir.ptr<i32>> {bindc_name = "a", uniq_name = "_QFomp_target_has_device_addrEa"}
+   integer, pointer :: a
+   !CHECK: omp.target has_device_addr(%[[VAL_0:.*]] : !fir.ref<!fir.box<!fir.ptr<i32>>>) map_entries({{.*}} -> {{.*}}, {{.*}} -> {{.*}} : !fir.llvm_ptr<!fir.ref<i32>>, !fir.ref<!fir.box<!fir.ptr<i32>>>) {
+   !$omp target has_device_addr(a)
+   !CHECK: {{.*}} = fir.load %[[VAL_0:.*]] : !fir.ref<!fir.box<!fir.ptr<i32>>>
+      a = 10
+   !CHECK: omp.terminator
+   !$omp end target
+   !CHECK: }
+end subroutine omp_target_has_device_addr
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
index 6454076f7593b..4ce7e47da046b 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
@@ -81,6 +81,9 @@ struct GrainsizeClauseOps {
   Value grainsizeVar;
 };
 
+struct HasDeviceAddrOps {
+  llvm::SmallVector<Value> hasDeviceAddrVars;
+};
 struct HintClauseOps {
   IntegerAttr hintAttr;
 };
@@ -94,6 +97,10 @@ struct InReductionClauseOps {
   llvm::SmallVector<Attribute> inReductionDeclSymbols;
 };
 
+struct IsDevicePtrOps {
+  llvm::SmallVector<Value> isDevicePtrVars;
+};
+
 struct LinearClauseOps {
   llvm::SmallVector<Value> linearVars, linearStepVars;
 };
@@ -251,13 +258,12 @@ using SimdLoopClauseOps =
 using SingleClauseOps = detail::Clauses<AllocateClauseOps, CopyprivateClauseOps,
                                         NowaitClauseOps, PrivateClauseOps>;
 
-// TODO `defaultmap`, `has_device_addr`, `is_device_ptr`, `uses_allocators`
-// clauses.
+// TODO `defaultmap`, `uses_allocators` clauses.
 using TargetClauseOps =
     detail::Clauses<AllocateClauseOps, DependClauseOps, DeviceClauseOps,
-                    IfClauseOps, InReductionClauseOps, MapClauseOps,
-                    NowaitClauseOps, PrivateClauseOps, ReductionClauseOps,
-                    ThreadLimitClauseOps>;
+                    HasDeviceAddrOps, IfClauseOps, InReductionClauseOps,
+                    IsDevicePtrOps, MapClauseOps, NowaitClauseOps,
+                    PrivateClauseOps, ReductionClauseOps, ThreadLimitClauseOps>;
 
 using TargetDataClauseOps = detail::Clauses<DeviceClauseOps, IfClauseOps,
                                             MapClauseOps, UseDeviceClauseOps>;
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index a38a82f9cc607..2a8582be38331 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1678,14 +1678,23 @@ def TargetOp : OpenMP_Op<"target", [IsolatedFromAbove, MapClauseOwningOpInterfac
 
     The optional $thread_limit specifies the limit on the number of threads
 
-    The optional $nowait elliminates the implicit barrier so the parent task can make progress
+    The optional $nowait eliminates the implicit barrier so the parent task can make progress
     even if the target task is not yet completed.
 
     The `depends` and `depend_vars` arguments are variadic lists of values
     that specify the dependencies of this particular target task in relation to
     other tasks.
 
-    TODO:  is_device_ptr, defaultmap, in_reduction
+    The optional $is_device_ptr indicates list items are device pointers.
+
+    The optional $has_device_addr indicates that list items already have device
+    addresses, so they may be directly accessed from the target device. This
+    includes array sections.
+
+    The optional $map_operands maps data from the task’s environment to the
+    device environment.
+
+    TODO:  defaultmap, in_reduction
 
   }];
 
@@ -1695,8 +1704,9 @@ def TargetOp : OpenMP_Op<"target", [IsolatedFromAbove, MapClauseOwningOpInterfac
                        OptionalAttr<TaskDependArrayAttr>:$depends,
                        Variadic<OpenMP_PointerLikeType>:$depend_vars,
                        UnitAttr:$nowait,
+                       Variadic<OpenMP_PointerLikeType>:$is_device_ptr,
+                       Variadic<OpenMP_PointerLikeType>:$has_device_addr,
                        Variadic<AnyType>:$map_operands);
-
   let regions = (region AnyRegion:$region);
 
   let builders = [
@@ -1708,6 +1718,8 @@ def TargetOp : OpenMP_Op<"target", [IsolatedFromAbove, MapClauseOwningOpInterfac
     | `device` `(` $device `:` type($device) `)`
     | `thread_limit` `(` $thread_limit `:` type($thread_limit) `)`
     | `nowait` $nowait
+    | `is_device_ptr` `(` $is_device_ptr `:` type($is_device_ptr) `)`
+    | `has_device_addr` `(` $has_device_addr `:` type($has_device_addr) `)`
     | `map_entries` `(` custom<MapEntries>($map_operands, type($map_operands)) `)`
     | `depend` `(` custom<DependVarList>($depend_vars, type($depend_vars), $depends) `)`
     ) $region attr-dict
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 543655338db8c..2d5b2231d2dd5 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1258,10 +1258,11 @@ void TargetOp::build(OpBuilder &builder, OperationState &state,
   // TODO Store clauses in op: allocateVars, allocatorVars, inReductionVars,
   // inReductionDeclSymbols, privateVars, privatizers, reductionVars,
   // reductionByRefAttr, reductionDeclSymbols.
-  TargetOp::build(builder, state, clauses.ifVar, clauses.deviceVar,
-                  clauses.threadLimitVar,
-                  makeArrayAttr(ctx, clauses.dependTypeAttrs),
-                  clauses.dependVars, clauses.nowaitAttr, clauses.mapVars);
+  TargetOp::build(
+      builder, state, clauses.ifVar, clauses.deviceVar, clauses.threadLimitVar,
+      makeArrayAttr(ctx, clauses.dependTypeAttrs), clauses.dependVars,
+      clauses.nowaitAttr, clauses.isDevicePtrVars, clauses.hasDeviceAddrVars,
+      clauses.mapVars);
 }
 
 LogicalResult TargetOp::verify() {
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index 1134db77d5baa..27a440b3f97ca 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -1809,7 +1809,7 @@ func.func @omp_target_depend(%data_var: memref<i32>) {
   // expected-error @below {{op expected as many depend values as depend variables}}
     "omp.target"(%data_var) ({
       "omp.terminator"() : () -> ()
-    }) {depends = [], operandSegmentSizes = array<i32: 0, 0, 0, 1, 0>} : (memref<i32>) -> ()
+    }) {depends = [], operandSegmentSizes = array<i32: 0, 0, 0, 1, 0, 0, 0>} : (memref<i32>) -> ()
    "func.return"() : () -> ()
 }
 
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index e2c255c7a3ccc..ad5b74b84ac70 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -510,22 +510,22 @@ return
 
 
 // CHECK-LABEL: omp_target
-func.func @omp_target(%if_cond : i1, %device : si32,  %num_threads : i32, %map1: memref<?xi32>, %map2: memref<?xi32>) -> () {
+func.func @omp_target(%if_cond : i1, %device : si32,  %num_threads : i32, %device_ptr: memref<i32>, %device_addr: memref<?xi32>, %map1: memref<?xi32>, %map2: memref<?xi32>) -> () {
 
     // Test with optional operands; if_expr, device, thread_limit, private, firstprivate and nowait.
     // CHECK: omp.target if({{.*}}) device({{.*}}) thread_limit({{.*}}) nowait
     "omp.target"(%if_cond, %device, %num_threads) ({
        // CHECK: omp.terminator
        omp.terminator
-    }) {nowait, operandSegmentSizes = array<i32: 1,1,1,0,0>} : ( i1, si32, i32 ) -> ()
+    }) {nowait, operandSegmentSizes = array<i32: 1,1,1,0,0,0,0>} : ( i1, si32, i32 ) -> ()
 
     // Test with optional map clause.
     // CHECK: %[[MAP_A:.*]] = omp.map.info var_ptr(%[[VAL_1:.*]] : memref<?xi32>, tensor<?xi32>)   map_clauses(tofrom) capture(ByRef) -> memref<?xi32> {name = ""}
     // CHECK: %[[MAP_B:.*]] = omp.map.info var_ptr(%[[VAL_2:.*]] : memref<?xi32>, tensor<?xi32>)   map_clauses(exit_release_or_enter_alloc) capture(ByRef) -> memref<?xi32> {name = ""}
-    // CHECK: omp.target map_entries(%[[MAP_A]] -> {{.*}}, %[[MAP_B]] -> {{.*}} : memref<?xi32>, memref<?xi32>) {
+    // CHECK: omp.target is_device_ptr(%[[VAL_4:.*]] : memref<i32>) has_device_addr(%[[VAL_5:.*]] : memref<?xi32>) map_entries(%[[MAP_A]] -> {{.*}}, %[[MAP_B]] -> {{.*}} : memref<?xi32>, memref<?xi32>) {
     %mapv1 = omp.map.info var_ptr(%map1 : memref<?xi32>, tensor<?xi32>)   map_clauses(tofrom) capture(ByRef) -> memref<?xi32> {name = ""}
     %mapv2 = omp.map.info var_ptr(%map2 : memref<?xi32>, tensor<?xi32>)   map_clauses(exit_release_or_enter_alloc) capture(ByRef) -> memref<?xi32> {name = ""}
-    omp.target map_entries(%mapv1 -> %arg0, %mapv2 -> %arg1 : memref<?xi32>, memref<?xi32>) {
+    omp.target map_entries(%mapv1 -> %arg0, %mapv2 -> %arg1 : memref<?xi32>, memref<?xi32>) is_device_ptr(%device_ptr : memref<i32>) has_device_addr(%device_addr : memref<?xi32>) {
     ^bb0(%arg0: memref<?xi32>, %arg1: memref<?xi32>):
       omp.terminator
     }

From ffb5bea2be9f966a39f243a7d8c2f48a1343cb4c Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 11 Apr 2024 16:33:05 +0100
Subject: [PATCH 148/886] [X86] LowerBITREVERSE - support SSE-only GFNI i32/i64
 bitreverse

Support Tremont CPUs which don't have AVX but do have GFNI.

Noticed while trying to workout how to clean up the costmodel for GFNI bitreverse
---
 llvm/lib/Target/X86/X86ISelLowering.cpp    | 17 ++++----
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 46 ++++------------------
 2 files changed, 15 insertions(+), 48 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 52be35aafb0f5..f274da6f6f776 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1286,6 +1286,11 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
       setOperationAction(ISD::CTLZ,             VT, Custom);
     }
 
+    if (Subtarget.hasGFNI()) {
+      setOperationAction(ISD::BITREVERSE,       MVT::i32, Custom);
+      setOperationAction(ISD::BITREVERSE,       MVT::i64, Custom);
+    }
+
     // These might be better off as horizontal vector ops.
     setOperationAction(ISD::ADD,                MVT::i16, Custom);
     setOperationAction(ISD::ADD,                MVT::i32, Custom);
@@ -1496,11 +1501,6 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::TRUNCATE,          MVT::v32i32, Custom);
     setOperationAction(ISD::TRUNCATE,          MVT::v32i64, Custom);
 
-    if (Subtarget.hasGFNI()) {
-      setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
-      setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
-    }
-
     for (auto VT : { MVT::v32i8, MVT::v16i16, MVT::v8i32, MVT::v4i64 }) {
       setOperationAction(ISD::SETCC,           VT, Custom);
       setOperationAction(ISD::CTPOP,           VT, Custom);
@@ -31337,12 +31337,9 @@ static SDValue LowerBITREVERSE(SDValue Op, const X86Subtarget &Subtarget,
   if (VT.is256BitVector() && !Subtarget.hasInt256())
     return splitVectorIntUnary(Op, DAG, DL);
 
-  // Lower i32/i64 to GFNI as vXi8 BITREVERSE + BSWAP
+  // Lower i32/i64 as vXi8 BITREVERSE + BSWAP
   if (!VT.isVector()) {
-
-    assert((VT.getScalarType() == MVT::i32) ||
-           (VT.getScalarType() == MVT::i64));
-
+    assert((VT == MVT::i32 || VT == MVT::i64) && "Only tested for i32/i64");
     MVT VecVT = MVT::getVectorVT(VT, 128 / VT.getSizeInBits());
     SDValue Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VecVT, In);
     Res = DAG.getNode(ISD::BITREVERSE, DL, MVT::v16i8,
diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index 1c5326d35bb00..b22b508db8b28 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -254,24 +254,10 @@ define i32 @test_bitreverse_i32(i32 %a) nounwind {
 ;
 ; GFNISSE-LABEL: test_bitreverse_i32:
 ; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNISSE-NEXT:    bswapl %edi
-; GFNISSE-NEXT:    movl %edi, %eax
-; GFNISSE-NEXT:    andl $252645135, %eax # imm = 0xF0F0F0F
-; GFNISSE-NEXT:    shll $4, %eax
-; GFNISSE-NEXT:    shrl $4, %edi
-; GFNISSE-NEXT:    andl $252645135, %edi # imm = 0xF0F0F0F
-; GFNISSE-NEXT:    orl %eax, %edi
-; GFNISSE-NEXT:    movl %edi, %eax
-; GFNISSE-NEXT:    andl $858993459, %eax # imm = 0x33333333
-; GFNISSE-NEXT:    shrl $2, %edi
-; GFNISSE-NEXT:    andl $858993459, %edi # imm = 0x33333333
-; GFNISSE-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNISSE-NEXT:    movl %eax, %ecx
-; GFNISSE-NEXT:    andl $1431655765, %ecx # imm = 0x55555555
-; GFNISSE-NEXT:    shrl %eax
-; GFNISSE-NEXT:    andl $1431655765, %eax # imm = 0x55555555
-; GFNISSE-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    bswapl %eax
 ; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i32:
@@ -343,26 +329,10 @@ define i64 @test_bitreverse_i64(i64 %a) nounwind {
 ;
 ; GFNISSE-LABEL: test_bitreverse_i64:
 ; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    bswapq %rdi
-; GFNISSE-NEXT:    movq %rdi, %rax
-; GFNISSE-NEXT:    shrq $4, %rax
-; GFNISSE-NEXT:    movabsq $1085102592571150095, %rcx # imm = 0xF0F0F0F0F0F0F0F
-; GFNISSE-NEXT:    andq %rcx, %rax
-; GFNISSE-NEXT:    andq %rcx, %rdi
-; GFNISSE-NEXT:    shlq $4, %rdi
-; GFNISSE-NEXT:    orq %rax, %rdi
-; GFNISSE-NEXT:    movabsq $3689348814741910323, %rax # imm = 0x3333333333333333
-; GFNISSE-NEXT:    movq %rdi, %rcx
-; GFNISSE-NEXT:    andq %rax, %rcx
-; GFNISSE-NEXT:    shrq $2, %rdi
-; GFNISSE-NEXT:    andq %rax, %rdi
-; GFNISSE-NEXT:    leaq (%rdi,%rcx,4), %rax
-; GFNISSE-NEXT:    movabsq $6148914691236517205, %rcx # imm = 0x5555555555555555
-; GFNISSE-NEXT:    movq %rax, %rdx
-; GFNISSE-NEXT:    andq %rcx, %rdx
-; GFNISSE-NEXT:    shrq %rax
-; GFNISSE-NEXT:    andq %rcx, %rax
-; GFNISSE-NEXT:    leaq (%rax,%rdx,2), %rax
+; GFNISSE-NEXT:    movq %rdi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movq %xmm0, %rax
+; GFNISSE-NEXT:    bswapq %rax
 ; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i64:

From 7ab7e7a55f3fce08ccd3cbcae94dabe99dd9e94a Mon Sep 17 00:00:00 2001
From: Xu Zhang <simonzgx@gmail.com>
Date: Thu, 11 Apr 2024 23:49:59 +0800
Subject: [PATCH 149/886] [libc][docs] Generate docs for signal.h & optimized
 is_implemented func (#88028)

Fixes #87835

This patch added the documentation for the POSIX functions according to
[n3096](https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3096.pdf)
Section 7.14, and gives the *docgen.py* script a more elegant *is_implemented*
function.
---
 libc/docs/index.rst           |  1 +
 libc/docs/signal.rst          | 43 +++++++++++++++++++++++++++++++++++
 libc/utils/docgen/docgen.py   | 13 +++++++----
 libc/utils/docgen/signal.json | 29 +++++++++++++++++++++++
 4 files changed, 82 insertions(+), 4 deletions(-)
 create mode 100644 libc/docs/signal.rst
 create mode 100644 libc/utils/docgen/signal.json

diff --git a/libc/docs/index.rst b/libc/docs/index.rst
index 8470c8d9287c2..11d5ae197d718 100644
--- a/libc/docs/index.rst
+++ b/libc/docs/index.rst
@@ -70,6 +70,7 @@ stages there is no ABI stability in any form.
    libc_search
    c23
    ctype
+   signal
 
 .. toctree::
    :hidden:
diff --git a/libc/docs/signal.rst b/libc/docs/signal.rst
new file mode 100644
index 0000000000000..7903bb439cb33
--- /dev/null
+++ b/libc/docs/signal.rst
@@ -0,0 +1,43 @@
+.. include:: check.rst
+
+signal.h Functions
+==================
+
+.. list-table::
+  :widths: auto
+  :align: center
+  :header-rows: 1
+
+  * - Function
+    - Implemented
+    - Standard
+  * - kill
+    - |check|
+    -
+  * - raise
+    - |check|
+    - 7.14.2.1
+  * - sigaction
+    - |check|
+    -
+  * - sigaddset
+    - |check|
+    -
+  * - sigaltstack
+    - |check|
+    -
+  * - sigdelset
+    - |check|
+    -
+  * - sigemptyset
+    - |check|
+    -
+  * - sigfillset
+    - |check|
+    -
+  * - signal
+    - |check|
+    - 7.14.1.1
+  * - sigprocmask
+    - |check|
+    -
diff --git a/libc/utils/docgen/docgen.py b/libc/utils/docgen/docgen.py
index 7411b4506f082..36eb409421b4f 100755
--- a/libc/utils/docgen/docgen.py
+++ b/libc/utils/docgen/docgen.py
@@ -23,12 +23,17 @@ def load_api(hname: str) -> Dict:
 # TODO: we may need to get more sophisticated for less generic implementations.
 # Does libc/src/{hname minus .h suffix}/{fname}.cpp exist?
 def is_implemented(hname: str, fname: str) -> bool:
-    return Path(
+    path = Path(
         Path(__file__).parent.parent.parent,
         "src",
-        hname.rstrip(".h"),
-        fname + ".cpp",
-    ).exists()
+        hname.rstrip(".h")
+    )
+    # Recursively search for the target source file in the subdirectories under
+    # libc/src/{hname}.
+    for _ in path.glob("**/" + fname + ".cpp"):
+        return True
+
+    return False
 
 
 def print_functions(header: str, functions: Dict):
diff --git a/libc/utils/docgen/signal.json b/libc/utils/docgen/signal.json
new file mode 100644
index 0000000000000..976021a803a67
--- /dev/null
+++ b/libc/utils/docgen/signal.json
@@ -0,0 +1,29 @@
+{
+  "macros": [
+    "SIG_DFL",
+    "SIG_ERR",
+    "SIG_IGN",
+    "SIGABRT",
+    "SIGFPE",
+    "SIGILL",
+    "SIGINT",
+    "SIGSEGV",
+    "SIGTERM"
+  ],
+  "functions": {
+    "kill": null,
+    "sigaction": null,
+    "sigaddset": null,
+    "sigaltstack": null,
+    "sigdelset": null,
+    "sigemptyset": null,
+    "sigfillset": null,
+    "sigprocmask": null,
+    "signal": {
+      "defined": "7.14.1.1"
+    },
+    "raise": {
+      "defined": "7.14.2.1"
+    }
+  }
+}

From b63fe0d72e2df3b3c4b9fcb91aea07b2582be195 Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Thu, 11 Apr 2024 17:55:25 +0200
Subject: [PATCH 150/886] [libc++][NFC] Reduce the memory footprint of
 __copy_cv a bit (#87718)

Instead of instantiating `__copy_cv` for every combination of `_From`
and `_To` this only instantiates `__copy_cv` for every `_From` type,
reducing the number of instantiations.
---
 libcxx/include/__type_traits/copy_cv.h | 28 +++++++++++++++-----------
 1 file changed, 16 insertions(+), 12 deletions(-)

diff --git a/libcxx/include/__type_traits/copy_cv.h b/libcxx/include/__type_traits/copy_cv.h
index b1c057ff778b1..d482cb42bffed 100644
--- a/libcxx/include/__type_traits/copy_cv.h
+++ b/libcxx/include/__type_traits/copy_cv.h
@@ -19,28 +19,32 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 
 // Let COPYCV(FROM, TO) be an alias for type TO with the addition of FROM's
 // top-level cv-qualifiers.
-template <class _From, class _To>
+template <class _From>
 struct __copy_cv {
-  using type = _To;
+  template <class _To>
+  using __apply = _To;
 };
 
-template <class _From, class _To>
-struct __copy_cv<const _From, _To> {
-  using type = const _To;
+template <class _From>
+struct __copy_cv<const _From> {
+  template <class _To>
+  using __apply = const _To;
 };
 
-template <class _From, class _To>
-struct __copy_cv<volatile _From, _To> {
-  using type = volatile _To;
+template <class _From>
+struct __copy_cv<volatile _From> {
+  template <class _To>
+  using __apply = volatile _To;
 };
 
-template <class _From, class _To>
-struct __copy_cv<const volatile _From, _To> {
-  using type = const volatile _To;
+template <class _From>
+struct __copy_cv<const volatile _From> {
+  template <class _To>
+  using __apply = const volatile _To;
 };
 
 template <class _From, class _To>
-using __copy_cv_t = typename __copy_cv<_From, _To>::type;
+using __copy_cv_t = typename __copy_cv<_From>::template __apply<_To>;
 
 _LIBCPP_END_NAMESPACE_STD
 

From 72f9881c3ffcf4be6361c3e4312d91c9c8d94a98 Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Thu, 11 Apr 2024 17:09:07 +0100
Subject: [PATCH 151/886] [libclc] Refactor build system to allow in-tree
 builds (#87622)

The previous build system was adding custom "OpenCL" and "LLVM IR"
languages in CMake to build the builtin libraries. This was making it
harder to build in-tree because the tool binaries needed to be present
at configure time.

This commit refactors the build system to use custom commands to build
the bytecode files one by one, and link them all together into the final
bytecode library. It also enables in-tree builds by aliasing the
clang/llvm-link/etc. tool targets to internal targets, which are
imported from the LLVM installation directory when building out of tree.

Diffing (with llvm-diff) all of the final bytecode libraries in an
out-of-tree configuration against those built using the current tip
system shows no changes. Note that there are textual changes to metadata
IDs which confuse regular diff, and that llvm-diff 14 and below may show
false-positives.

This commit also removes a file listed in one of the SOURCEs which
didn't exist and which was preventing the use of
ENABLE_RUNTIME_SUBNORMAL when configuring CMake.
---
 libclc/CMakeLists.txt                         | 277 +++++++++++-------
 libclc/cmake/CMakeCLCCompiler.cmake.in        |   9 -
 libclc/cmake/CMakeCLCInformation.cmake        |  12 -
 libclc/cmake/CMakeDetermineCLCCompiler.cmake  |  18 --
 .../cmake/CMakeDetermineLLAsmCompiler.cmake   |  24 --
 libclc/cmake/CMakeLLAsmCompiler.cmake.in      |  10 -
 libclc/cmake/CMakeLLAsmInformation.cmake      |  12 -
 libclc/cmake/CMakeTestCLCCompiler.cmake       |  56 ----
 libclc/cmake/CMakeTestLLAsmCompiler.cmake     |  56 ----
 libclc/cmake/modules/AddLibclc.cmake          | 156 ++++++++++
 libclc/generic/lib/SOURCES                    |   1 -
 llvm/tools/CMakeLists.txt                     |   3 +
 12 files changed, 322 insertions(+), 312 deletions(-)
 delete mode 100644 libclc/cmake/CMakeCLCCompiler.cmake.in
 delete mode 100644 libclc/cmake/CMakeCLCInformation.cmake
 delete mode 100644 libclc/cmake/CMakeDetermineCLCCompiler.cmake
 delete mode 100644 libclc/cmake/CMakeDetermineLLAsmCompiler.cmake
 delete mode 100644 libclc/cmake/CMakeLLAsmCompiler.cmake.in
 delete mode 100644 libclc/cmake/CMakeLLAsmInformation.cmake
 delete mode 100644 libclc/cmake/CMakeTestCLCCompiler.cmake
 delete mode 100644 libclc/cmake/CMakeTestLLAsmCompiler.cmake
 create mode 100644 libclc/cmake/modules/AddLibclc.cmake

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index c6e3cdf23fe0c..7528228b3b7f9 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -4,6 +4,15 @@ project( libclc VERSION 0.2.0 LANGUAGES CXX C)
 
 set(CMAKE_CXX_STANDARD 17)
 
+# Add path for custom modules
+list( INSERT CMAKE_MODULE_PATH 0 "${PROJECT_SOURCE_DIR}/cmake/modules" )
+
+set( LIBCLC_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR} )
+set( LIBCLC_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR} )
+set( LIBCLC_OBJFILE_DIR ${LIBCLC_BINARY_DIR}/obj.libclc.dir )
+
+include( AddLibclc )
+
 include( GNUInstallDirs )
 set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS
   amdgcn-amdhsa/lib/SOURCES;
@@ -27,31 +36,51 @@ set( LIBCLC_TARGETS_TO_BUILD "all"
 
 option( ENABLE_RUNTIME_SUBNORMAL "Enable runtime linking of subnormal support." OFF )
 
-find_package(LLVM REQUIRED HINTS "${LLVM_CMAKE_DIR}")
-include(AddLLVM)
+if( LIBCLC_STANDALONE_BUILD OR CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR )
+  # Out-of-tree configuration
+  set( LIBCLC_STANDALONE_BUILD TRUE )
 
-message( STATUS "libclc LLVM version: ${LLVM_PACKAGE_VERSION}" )
+  find_package(LLVM REQUIRED HINTS "${LLVM_CMAKE_DIR}")
+  include(AddLLVM)
 
-if( LLVM_PACKAGE_VERSION VERSION_LESS LIBCLC_MIN_LLVM )
-  message( FATAL_ERROR "libclc needs at least LLVM ${LIBCLC_MIN_LLVM}" )
-endif()
+  message( STATUS "libclc LLVM version: ${LLVM_PACKAGE_VERSION}" )
 
-find_program( LLVM_CLANG clang PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
-find_program( LLVM_AS llvm-as PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
-find_program( LLVM_LINK llvm-link PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
-find_program( LLVM_OPT opt PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
-find_program( LLVM_SPIRV llvm-spirv PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
+  if( LLVM_PACKAGE_VERSION VERSION_LESS LIBCLC_MIN_LLVM )
+    message( FATAL_ERROR "libclc needs at least LLVM ${LIBCLC_MIN_LLVM}" )
+  endif()
+
+  # Import required tools as targets
+  foreach( tool clang llvm-as llvm-link opt )
+    find_program( LLVM_TOOL_${tool} ${tool} PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
+    add_executable( libclc::${tool} IMPORTED GLOBAL )
+    set_target_properties( libclc::${tool} PROPERTIES IMPORTED_LOCATION ${LLVM_TOOL_${tool}} )
+  endforeach()
+else()
+  # In-tree configuration
+  set( LIBCLC_STANDALONE_BUILD FALSE )
+
+  set( LLVM_PACKAGE_VERSION ${LLVM_VERSION} )
 
-# Print toolchain
-message( STATUS "libclc toolchain - clang: ${LLVM_CLANG}" )
-message( STATUS "libclc toolchain - llvm-as: ${LLVM_AS}" )
-message( STATUS "libclc toolchain - llvm-link: ${LLVM_LINK}" )
-message( STATUS "libclc toolchain - opt: ${LLVM_OPT}" )
-message( STATUS "libclc toolchain - llvm-spirv: ${LLVM_SPIRV}" )
-if( NOT LLVM_CLANG OR NOT LLVM_OPT OR NOT LLVM_AS OR NOT LLVM_LINK )
+  # Note that we check this later (for both build types) but we can provide a
+  # more useful error message when built in-tree. We assume that LLVM tools are
+  # always available so don't warn here.
+  if( NOT clang IN_LIST LLVM_ENABLE_PROJECTS )
+    message(FATAL_ERROR "Clang is not enabled, but is required to build libclc in-tree")
+  endif()
+
+  foreach( tool clang llvm-as llvm-link opt )
+    add_executable(libclc::${tool} ALIAS ${tool})
+  endforeach()
+endif()
+
+if( NOT TARGET libclc::clang OR NOT TARGET libclc::opt
+    OR NOT TARGET libclc::llvm-as OR NOT TARGET libclc::llvm-link )
   message( FATAL_ERROR "libclc toolchain incomplete!" )
 endif()
 
+# llvm-spirv is an optional dependency, used to build spirv-* targets.
+find_program( LLVM_SPIRV llvm-spirv PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
+
 # List of all targets. Note that some are added dynamically below.
 set( LIBCLC_TARGETS_ALL
   amdgcn--
@@ -90,24 +119,9 @@ if( "spirv-mesa3d-" IN_LIST LIBCLC_TARGETS_TO_BUILD OR "spirv64-mesa3d-" IN_LIST
   endif()
 endif()
 
-set( CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake )
-set( CMAKE_CLC_COMPILER ${LLVM_CLANG} )
-set( CMAKE_CLC_ARCHIVE ${LLVM_LINK} )
-set( CMAKE_LLAsm_PREPROCESSOR ${LLVM_CLANG} )
-set( CMAKE_LLAsm_COMPILER ${LLVM_AS} )
-set( CMAKE_LLAsm_ARCHIVE ${LLVM_LINK} )
-
 # Construct LLVM version define
 set( LLVM_VERSION_DEFINE "-DHAVE_LLVM=0x${LLVM_VERSION_MAJOR}0${LLVM_VERSION_MINOR}" )
 
-
-# LLVM 13 enables standard includes by default
-if( LLVM_PACKAGE_VERSION VERSION_GREATER_EQUAL 13.0.0 )
-  set( CMAKE_LLAsm_FLAGS "${CMAKE_LLAsm_FLAGS} -cl-no-stdinc" )
-  set( CMAKE_CLC_FLAGS "${CMAKE_CLC_FLAGS} -cl-no-stdinc" )
-endif()
-
-enable_language( CLC LLAsm )
 # This needs to be set before any target that needs it
 # We need to use LLVM_INCLUDE_DIRS here, because if we are linking to an
 # llvm build directory, this includes $src/llvm/include which is where all the
@@ -122,7 +136,7 @@ set(LLVM_LINK_COMPONENTS
   IRReader
   Support
 )
-if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
+if( LIBCLC_STANDALONE_BUILD )
   add_llvm_executable( prepare_builtins utils/prepare-builtins.cpp )
 else()
   add_llvm_utility( prepare_builtins utils/prepare-builtins.cpp )
@@ -167,12 +181,14 @@ install( FILES ${CMAKE_CURRENT_BINARY_DIR}/libclc.pc DESTINATION "${CMAKE_INSTAL
 install( DIRECTORY generic/include/clc DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}" )
 
 if( ENABLE_RUNTIME_SUBNORMAL )
-  add_library( subnormal_use_default STATIC
-    generic/lib/subnormal_use_default.ll )
-  add_library( subnormal_disable STATIC
-    generic/lib/subnormal_disable.ll )
-  install( TARGETS subnormal_use_default subnormal_disable ARCHIVE
-    DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
+  foreach( file subnormal_use_default subnormal_disable )
+    link_bc(
+       TARGET ${file}
+       INPUTS ${PROJECT_SOURCE_DIR}/generic/lib/${file}.ll
+    )
+    install( FILES $<TARGET_PROPERTY:${file},TARGET_FILE> ARCHIVE
+      DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
+  endforeach()
 endif()
 
 find_package( Python3 REQUIRED COMPONENTS Interpreter )
@@ -232,19 +248,19 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
 
   # Add the generated convert.cl here to prevent adding the one listed in
   # SOURCES
+  set( objects )   # A "set" of already-added input files
+  set( rel_files ) # Source directory input files, relative to the root dir
+  set( gen_files ) # Generated binary input files, relative to the binary dir
   if( NOT ${ARCH} STREQUAL "spirv" AND NOT ${ARCH} STREQUAL "spirv64" )
     if( NOT ENABLE_RUNTIME_SUBNORMAL AND NOT ${ARCH} STREQUAL "clspv" AND
         NOT ${ARCH} STREQUAL "clspv64" )
-      set( rel_files convert.cl )
-      set( objects convert.cl )
+      list( APPEND gen_files convert.cl )
+      list( APPEND objects convert.cl )
       list( APPEND rel_files generic/lib/subnormal_use_default.ll )
     elseif(${ARCH} STREQUAL "clspv" OR ${ARCH} STREQUAL "clspv64")
-      set( rel_files clspv-convert.cl )
-      set( objects clspv-convert.cl )
+      list( APPEND gen_files clspv-convert.cl )
+      list( APPEND objects clspv-convert.cl )
     endif()
-  else()
-    set( rel_files )
-    set( objects )
   endif()
 
   foreach( l ${source_list} )
@@ -252,46 +268,35 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
     string( REPLACE "\n" ";" file_list ${file_list} )
     get_filename_component( dir ${l} DIRECTORY )
     foreach( f ${file_list} )
-      list( FIND objects ${f} found )
-      if( found EQUAL  -1 )
+      # Only add each file once, so that targets can 'specialize' builtins
+      if( NOT ${f} IN_LIST objects )
         list( APPEND objects ${f} )
         list( APPEND rel_files ${dir}/${f} )
-        # FIXME: This should really go away
-        file( TO_CMAKE_PATH ${PROJECT_SOURCE_DIR}/${dir}/${f} src_loc )
-        get_filename_component( fdir ${src_loc} DIRECTORY )
-
-        set_source_files_properties( ${dir}/${f}
-          PROPERTIES COMPILE_FLAGS "-I ${fdir}" )
       endif()
     endforeach()
   endforeach()
 
   foreach( d ${${t}_devices} )
-    # Some targets don't have a specific GPU to target
-    if( ${d} STREQUAL "none" OR ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
-      set( mcpu )
-      set( arch_suffix "${t}" )
-    else()
-      set( mcpu "-mcpu=${d}" )
-      set( arch_suffix "${d}-${t}" )
+    get_libclc_device_info(
+      TRIPLE ${t}
+      DEVICE ${d}
+      CPU cpu
+      ARCH_SUFFIX arch_suffix
+      CLANG_TRIPLE clang_triple
+    )
+
+    set( mcpu )
+    if( NOT "${cpu}" STREQUAL "" )
+      set( mcpu "-mcpu=${cpu}" )
     endif()
+
     message( STATUS "  device: ${d} ( ${${d}_aliases} )" )
 
-    if ( ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
-      if( ${ARCH} STREQUAL "spirv" )
-        set( t "spir--" )
-      else()
-        set( t "spir64--" )
-      endif()
+    if ( ARCH STREQUAL spirv OR ARCH STREQUAL spirv64 )
       set( build_flags -O0 -finline-hint-functions )
       set( opt_flags )
       set( spvflags --spirv-max-version=1.1 )
-    elseif( ${ARCH} STREQUAL "clspv" )
-      set( t "spir--" )
-      set( build_flags "-Wno-unknown-assumption")
-      set( opt_flags -O3 )
-    elseif( ${ARCH} STREQUAL "clspv64" )
-      set( t "spir64--" )
+    elseif( ARCH STREQUAL clspv OR ARCH STREQUAL clspv64 )
       set( build_flags "-Wno-unknown-assumption")
       set( opt_flags -O3 )
     else()
@@ -299,53 +304,97 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
       set( opt_flags -O3 )
     endif()
 
+    set( LIBCLC_ARCH_OBJFILE_DIR "${LIBCLC_OBJFILE_DIR}/${arch_suffix}" )
+    file( MAKE_DIRECTORY ${LIBCLC_ARCH_OBJFILE_DIR} )
+
+    string( TOUPPER "CLC_${ARCH}" CLC_TARGET_DEFINE )
+
+    list( APPEND build_flags
+      -D__CLC_INTERNAL
+      -D${CLC_TARGET_DEFINE}
+      -I${PROJECT_SOURCE_DIR}/generic/include
+      # FIXME: Fix libclc to not require disabling this noisy warning
+      -Wno-bitwise-conditional-parentheses
+    )
+
+    set( bytecode_files "" )
+    foreach( file IN LISTS gen_files rel_files )
+      # We need to take each file and produce an absolute input file, as well
+      # as a unique architecture-specific output file. We deal with a mix of
+      # different input files, which makes this trickier.
+      if( ${file} IN_LIST gen_files )
+        # Generated files are given just as file names, which we must make
+        # absolute to the binary directory.
+        set( input_file ${CMAKE_CURRENT_BINARY_DIR}/${file} )
+        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${file}.o" )
+      else()
+        # Other files are originally relative to each SOURCE file, which are
+        # then make relative to the libclc root directory. We must normalize
+        # the path (e.g., ironing out any ".."), then make it relative to the
+        # root directory again, and use that relative path component for the
+        # binary path.
+        get_filename_component( abs_path ${file} ABSOLUTE BASE_DIR ${PROJECT_SOURCE_DIR} )
+        file( RELATIVE_PATH root_rel_path ${PROJECT_SOURCE_DIR} ${abs_path} )
+        set( input_file ${PROJECT_SOURCE_DIR}/${file} )
+        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${root_rel_path}.o" )
+      endif()
+
+      get_filename_component( file_dir ${file} DIRECTORY )
+
+      compile_to_bc(
+        TRIPLE ${clang_triple}
+        INPUT ${input_file}
+        OUTPUT ${output_file}
+        EXTRA_OPTS "${mcpu}" -fno-builtin -nostdlib
+                   "${build_flags}" -I${PROJECT_SOURCE_DIR}/${file_dir}
+      )
+      list( APPEND bytecode_files ${output_file} )
+    endforeach()
+
     set( builtins_link_lib_tgt builtins.link.${arch_suffix} )
-    add_library( ${builtins_link_lib_tgt} STATIC ${rel_files} )
-    # Make sure we depend on the pseudo target to prevent
-    # multiple invocations
-    add_dependencies( ${builtins_link_lib_tgt} generate_convert.cl )
-    add_dependencies( ${builtins_link_lib_tgt} clspv-generate_convert.cl )
-    # CMake will turn this include into absolute path
-    target_include_directories( ${builtins_link_lib_tgt} PRIVATE
-      "generic/include" )
-    target_compile_definitions( ${builtins_link_lib_tgt} PRIVATE
-      "__CLC_INTERNAL" )
-    string( TOUPPER "-DCLC_${ARCH}" CLC_TARGET_DEFINE )
-    target_compile_definitions( ${builtins_link_lib_tgt} PRIVATE
-      ${CLC_TARGET_DEFINE} )
-    target_compile_options( ${builtins_link_lib_tgt} PRIVATE  -target
-      ${t} ${mcpu} -fno-builtin -nostdlib ${build_flags} )
-    set_target_properties( ${builtins_link_lib_tgt} PROPERTIES
-      LINKER_LANGUAGE CLC )
-
-    set( obj_suffix ${arch_suffix}.bc )
-    set( builtins_opt_lib_tgt builtins.opt.${obj_suffix} )
-
-    # Add opt target
-    add_custom_command( OUTPUT ${builtins_opt_lib_tgt}
-      COMMAND ${LLVM_OPT} ${opt_flags} -o ${builtins_opt_lib_tgt}
-        $<TARGET_FILE:${builtins_link_lib_tgt}>
-      DEPENDS ${builtins_link_lib_tgt} )
-    add_custom_target( "opt.${obj_suffix}" ALL
-      DEPENDS ${builtins_opt_lib_tgt} )
-
-    if( ${ARCH} STREQUAL "spirv" OR ${ARCH} STREQUAL "spirv64" )
+
+    link_bc(
+      TARGET ${builtins_link_lib_tgt}
+      INPUTS ${bytecode_files}
+    )
+
+    set( builtins_link_lib $<TARGET_PROPERTY:${builtins_link_lib_tgt},TARGET_FILE> )
+
+    if( ARCH STREQUAL spirv OR ARCH STREQUAL spirv64 )
       set( spv_suffix ${arch_suffix}.spv )
-      add_custom_command( OUTPUT "${spv_suffix}"
-        COMMAND ${LLVM_SPIRV} ${spvflags} -o "${spv_suffix}" $<TARGET_FILE:${builtins_link_lib_tgt}>
-        DEPENDS ${builtins_link_lib_tgt} )
+      add_custom_command( OUTPUT ${spv_suffix}
+        COMMAND ${LLVM_SPIRV} ${spvflags} -o ${spv_suffix} ${builtins_link_lib}
+        DEPENDS ${builtins_link_lib_tgt}
+      )
       add_custom_target( "prepare-${spv_suffix}" ALL DEPENDS "${spv_suffix}" )
       install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${spv_suffix}
          DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
     else()
+      set( builtins_opt_lib_tgt builtins.opt.${arch_suffix} )
+
+      # Add opt target
+      add_custom_command( OUTPUT ${builtins_opt_lib_tgt}.bc
+        COMMAND libclc::opt ${opt_flags} -o ${builtins_opt_lib_tgt}.bc
+          ${builtins_link_lib}
+        DEPENDS libclc::opt ${builtins_link_lib_tgt}
+      )
+      add_custom_target( ${builtins_opt_lib_tgt}
+        ALL DEPENDS ${builtins_opt_lib_tgt}.bc
+      )
+      set_target_properties( ${builtins_opt_lib_tgt}
+        PROPERTIES TARGET_FILE ${builtins_opt_lib_tgt}.bc
+      )
+
       # Add prepare target
-      add_custom_command( OUTPUT "${obj_suffix}"
-        COMMAND prepare_builtins -o "${obj_suffix}" ${builtins_opt_lib_tgt}
-        DEPENDS "opt.${obj_suffix}" ${builtins_opt_lib_tgt} prepare_builtins )
-      add_custom_target( "prepare-${obj_suffix}" ALL DEPENDS "${obj_suffix}" )
+      set( obj_suffix ${arch_suffix}.bc )
+      add_custom_command( OUTPUT ${obj_suffix}
+        COMMAND prepare_builtins -o ${obj_suffix}
+          $<TARGET_PROPERTY:${builtins_opt_lib_tgt},TARGET_FILE>
+        DEPENDS ${builtins_opt_lib_tgt} prepare_builtins )
+      add_custom_target( prepare-${obj_suffix} ALL DEPENDS ${obj_suffix} )
 
       # nvptx-- targets don't include workitem builtins
-      if( NOT ${t} MATCHES ".*ptx.*--$" )
+      if( NOT clang_triple MATCHES ".*ptx.*--$" )
         add_test( NAME external-calls-${obj_suffix}
           COMMAND ./check_external_calls.sh ${CMAKE_CURRENT_BINARY_DIR}/${obj_suffix} ${LLVM_TOOLS_BINARY_DIR}
           WORKING_DIRECTORY ${PROJECT_SOURCE_DIR} )
@@ -353,10 +402,10 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
 
       install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${obj_suffix} DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
       foreach( a ${${d}_aliases} )
-        set( alias_suffix "${a}-${t}.bc" )
+        set( alias_suffix "${a}-${clang_triple}.bc" )
         add_custom_target( ${alias_suffix} ALL
           COMMAND ${CMAKE_COMMAND} -E create_symlink ${obj_suffix} ${alias_suffix}
-          DEPENDS "prepare-${obj_suffix}" )
+          DEPENDS prepare-${obj_suffix} )
         install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${alias_suffix} DESTINATION "${CMAKE_INSTALL_DATADIR}/clc" )
       endforeach( a )
     endif()
diff --git a/libclc/cmake/CMakeCLCCompiler.cmake.in b/libclc/cmake/CMakeCLCCompiler.cmake.in
deleted file mode 100644
index 2730b83d9e7d0..0000000000000
--- a/libclc/cmake/CMakeCLCCompiler.cmake.in
+++ /dev/null
@@ -1,9 +0,0 @@
-set(CMAKE_CLC_COMPILER "@CMAKE_CLC_COMPILER@")
-set(CMAKE_CLC_COMPILER_LOADED 1)
-
-set(CMAKE_CLC_SOURCE_FILE_EXTENSIONS cl)
-set(CMAKE_CLC_OUTPUT_EXTENSION .bc)
-set(CMAKE_CLC_OUTPUT_EXTENSION_REPLACE 1)
-set(CMAKE_STATIC_LIBRARY_PREFIX_CLC "")
-set(CMAKE_STATIC_LIBRARY_SUFFIX_CLC ".bc")
-set(CMAKE_CLC_COMPILER_ENV_VAR "CLC_COMPILER")
diff --git a/libclc/cmake/CMakeCLCInformation.cmake b/libclc/cmake/CMakeCLCInformation.cmake
deleted file mode 100644
index 95327e4439722..0000000000000
--- a/libclc/cmake/CMakeCLCInformation.cmake
+++ /dev/null
@@ -1,12 +0,0 @@
-if(NOT CMAKE_CLC_COMPILE_OBJECT)
-  set(CMAKE_CLC_COMPILE_OBJECT
-    "<CMAKE_CLC_COMPILER> <DEFINES> <INCLUDES> <FLAGS> -o <OBJECT> -c <SOURCE> -emit-llvm")
-endif()
-
-if(NOT CMAKE_CLC_CREATE_STATIC_LIBRARY)
-  set(CMAKE_CLC_CREATE_STATIC_LIBRARY
-    "<CMAKE_CLC_ARCHIVE> -o <TARGET> <OBJECTS>")
-endif()
-
-set(CMAKE_INCLUDE_FLAG_CLC "-I")
-set(CMAKE_DEPFILE_FLAGS_CLC "-MD -MT <DEP_TARGET> -MF <DEP_FILE>")
diff --git a/libclc/cmake/CMakeDetermineCLCCompiler.cmake b/libclc/cmake/CMakeDetermineCLCCompiler.cmake
deleted file mode 100644
index 94d85d9e666ad..0000000000000
--- a/libclc/cmake/CMakeDetermineCLCCompiler.cmake
+++ /dev/null
@@ -1,18 +0,0 @@
-include(${CMAKE_ROOT}/Modules/CMakeDetermineCompiler.cmake)
-
-if(NOT CMAKE_CLC_COMPILER)
-  find_program(CMAKE_CLC_COMPILER NAMES clang)
-endif()
-mark_as_advanced(CMAKE_CLC_COMPILER)
-
-if(NOT CMAKE_CLC_ARCHIVE)
-  find_program(CMAKE_CLC_ARCHIVE NAMES llvm-link)
-endif()
-mark_as_advanced(CMAKE_CLC_ARCHIVE)
-
-set(CMAKE_CLC_COMPILER_ENV_VAR "CLC_COMPILER")
-set(CMAKE_CLC_ARCHIVE_ENV_VAR "CLC_LINKER")
-find_file(clc_comp_in CMakeCLCCompiler.cmake.in PATHS ${CMAKE_ROOT}/Modules ${CMAKE_MODULE_PATH})
-# configure all variables set in this file
-configure_file(${clc_comp_in} ${CMAKE_PLATFORM_INFO_DIR}/CMakeCLCCompiler.cmake @ONLY)
-mark_as_advanced(clc_comp_in)
diff --git a/libclc/cmake/CMakeDetermineLLAsmCompiler.cmake b/libclc/cmake/CMakeDetermineLLAsmCompiler.cmake
deleted file mode 100644
index 1c424c79cb453..0000000000000
--- a/libclc/cmake/CMakeDetermineLLAsmCompiler.cmake
+++ /dev/null
@@ -1,24 +0,0 @@
-include(${CMAKE_ROOT}/Modules/CMakeDetermineCompiler.cmake)
-
-if(NOT CMAKE_LLAsm_PREPROCESSOR)
-  find_program(CMAKE_LLAsm_PREPROCESSOR NAMES clang)
-endif()
-mark_as_advanced(CMAKE_LLAsm_PREPROCESSOR)
-
-if(NOT CMAKE_LLAsm_COMPILER)
-  find_program(CMAKE_LLAsm_COMPILER NAMES llvm-as)
-endif()
-mark_as_advanced(CMAKE_LLAsm_ASSEMBLER)
-
-if(NOT CMAKE_LLAsm_ARCHIVE)
-  find_program(CMAKE_LLAsm_ARCHIVE NAMES llvm-link)
-endif()
-mark_as_advanced(CMAKE_LLAsm_ARCHIVE)
-
-set(CMAKE_LLAsm_PREPROCESSOR_ENV_VAR "LL_PREPROCESSOR")
-set(CMAKE_LLAsm_COMPILER_ENV_VAR "LL_ASSEMBLER")
-set(CMAKE_LLAsm_ARCHIVE_ENV_VAR "LL_LINKER")
-find_file(ll_comp_in CMakeLLAsmCompiler.cmake.in PATHS ${CMAKE_ROOT}/Modules ${CMAKE_MODULE_PATH})
-# configure all variables set in this file
-configure_file(${ll_comp_in} ${CMAKE_PLATFORM_INFO_DIR}/CMakeLLAsmCompiler.cmake @ONLY)
-mark_as_advanced(ll_comp_in)
diff --git a/libclc/cmake/CMakeLLAsmCompiler.cmake.in b/libclc/cmake/CMakeLLAsmCompiler.cmake.in
deleted file mode 100644
index 2b00f69234dd2..0000000000000
--- a/libclc/cmake/CMakeLLAsmCompiler.cmake.in
+++ /dev/null
@@ -1,10 +0,0 @@
-set(CMAKE_LLAsm_PREPROCESSOR "@CMAKE_LLAsm_PREPROCESSOR@")
-set(CMAKE_LLAsm_COMPILER "@CMAKE_LLAsm_COMPILER@")
-set(CMAKE_LLAsm_ARCHIVE "@CMAKE_LLAsm_ARCHIVE@")
-set(CMAKE_LLAsm_COMPILER_LOADED 1)
-
-set(CMAKE_LLAsm_SOURCE_FILE_EXTENSIONS ll)
-set(CMAKE_LLAsm_OUTPUT_EXTENSION .bc)
-set(CMAKE_LLAsm_OUTPUT_EXTENSION_REPLACE 1)
-set(CMAKE_STATIC_LIBRARY_PREFIX_LLAsm "")
-set(CMAKE_STATIC_LIBRARY_SUFFIX_LLAsm ".bc")
diff --git a/libclc/cmake/CMakeLLAsmInformation.cmake b/libclc/cmake/CMakeLLAsmInformation.cmake
deleted file mode 100644
index 35ec3081da0f7..0000000000000
--- a/libclc/cmake/CMakeLLAsmInformation.cmake
+++ /dev/null
@@ -1,12 +0,0 @@
-if(NOT CMAKE_LLAsm_COMPILE_OBJECT)
-  set(CMAKE_LLAsm_COMPILE_OBJECT
-    "${CMAKE_LLAsm_PREPROCESSOR} -E -P  <DEFINES> <INCLUDES> <FLAGS> -x cl <SOURCE> -o <OBJECT>.temp"
-    "<CMAKE_LLAsm_COMPILER> -o <OBJECT> <OBJECT>.temp")
-endif()
-
-if(NOT CMAKE_LLAsm_CREATE_STATIC_LIBRARY)
-  set(CMAKE_LLAsm_CREATE_STATIC_LIBRARY
-    "<CMAKE_LLAsm_ARCHIVE> -o <TARGET> <OBJECTS>")
-endif()
-
-set(CMAKE_INCLUDE_FLAG_LLAsm "-I")
diff --git a/libclc/cmake/CMakeTestCLCCompiler.cmake b/libclc/cmake/CMakeTestCLCCompiler.cmake
deleted file mode 100644
index 869fcc3d01ab0..0000000000000
--- a/libclc/cmake/CMakeTestCLCCompiler.cmake
+++ /dev/null
@@ -1,56 +0,0 @@
-if(CMAKE_CLC_COMPILER_FORCED)
-  # The compiler configuration was forced by the user.
-  # Assume the user has configured all compiler information.
-  set(CMAKE_CLC_COMPILER_WORKS TRUE)
-  return()
-endif()
-
-include(CMakeTestCompilerCommon)
-
-# Remove any cached result from an older CMake version.
-# We now store this in CMakeCCompiler.cmake.
-unset(CMAKE_CLC_COMPILER_WORKS CACHE)
-
-# This file is used by EnableLanguage in cmGlobalGenerator to
-# determine that that selected CLC compiler can actually compile
-# and link the most basic of programs. If not, a fatal error
-# is set and cmake stops processing commands and will not generate
-# any makefiles or projects.
-if(NOT CMAKE_CLC_COMPILER_WORKS)
-  PrintTestCompilerStatus("CLC" "")
-  file(WRITE ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/testCLCCompiler.cl
-    "__kernel void test_k(global int * a)\n"
-    "{ *a = 1; }\n")
-  try_compile(CMAKE_CLC_COMPILER_WORKS ${CMAKE_BINARY_DIR}
-    ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/testCLCCompiler.cl
-    # We never generate executable so bypass the link step
-    CMAKE_FLAGS -DCMAKE_CLC_LINK_EXECUTABLE='echo'
-    OUTPUT_VARIABLE __CMAKE_CLC_COMPILER_OUTPUT)
-  # Move result from cache to normal variable.
-  set(CMAKE_CLC_COMPILER_WORKS ${CMAKE_CLC_COMPILER_WORKS})
-  unset(CMAKE_CLC_COMPILER_WORKS CACHE)
-  set(CLC_TEST_WAS_RUN 1)
-endif()
-
-if(NOT CMAKE_CLC_COMPILER_WORKS)
-  PrintTestCompilerStatus("CLC" " -- broken")
-  file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeError.log
-    "Determining if the CLC compiler works failed with "
-    "the following output:\n${__CMAKE_CLC_COMPILER_OUTPUT}\n\n")
-  message(FATAL_ERROR "The CLC compiler \"${CMAKE_CLC_COMPILER}\" "
-    "is not able to compile a simple test program.\nIt fails "
-    "with the following output:\n ${__CMAKE_CLC_COMPILER_OUTPUT}\n\n"
-    "CMake will not be able to correctly generate this project.")
-else()
-  if(CLC_TEST_WAS_RUN)
-    PrintTestCompilerStatus("CLC" " -- works")
-    file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeOutput.log
-      "Determining if the CLC compiler works passed with "
-      "the following output:\n${__CMAKE_CLC_COMPILER_OUTPUT}\n\n")
-  endif()
-
-  include(${CMAKE_PLATFORM_INFO_DIR}/CMakeCLCCompiler.cmake)
-
-endif()
-
-unset(__CMAKE_CLC_COMPILER_OUTPUT)
diff --git a/libclc/cmake/CMakeTestLLAsmCompiler.cmake b/libclc/cmake/CMakeTestLLAsmCompiler.cmake
deleted file mode 100644
index 35948ee07a940..0000000000000
--- a/libclc/cmake/CMakeTestLLAsmCompiler.cmake
+++ /dev/null
@@ -1,56 +0,0 @@
-if(CMAKE_LLAsm_COMPILER_FORCED)
-  # The compiler configuration was forced by the user.
-  # Assume the user has configured all compiler information.
-  set(CMAKE_LLAsm_COMPILER_WORKS TRUE)
-  return()
-endif()
-
-include(CMakeTestCompilerCommon)
-
-# Remove any cached result from an older CMake version.
-# We now store this in CMakeCCompiler.cmake.
-unset(CMAKE_LLAsm_COMPILER_WORKS CACHE)
-
-# This file is used by EnableLanguage in cmGlobalGenerator to
-# determine that that selected llvm assembler can actually compile
-# and link the most basic of programs. If not, a fatal error
-# is set and cmake stops processing commands and will not generate
-# any makefiles or projects.
-if(NOT CMAKE_LLAsm_COMPILER_WORKS)
-  PrintTestCompilerStatus("LLAsm" "")
-  file(WRITE ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/testLLAsmCompiler.ll
-    "define i32 @test() {\n"
-    "ret i32 0 }\n" )
-  try_compile(CMAKE_LLAsm_COMPILER_WORKS ${CMAKE_BINARY_DIR}
-    ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeTmp/testLLAsmCompiler.ll
-    # We never generate executable so bypass the link step
-    CMAKE_FLAGS -DCMAKE_LLAsm_LINK_EXECUTABLE='echo'
-    OUTPUT_VARIABLE __CMAKE_LLAsm_COMPILER_OUTPUT)
-  # Move result from cache to normal variable.
-  set(CMAKE_LLAsm_COMPILER_WORKS ${CMAKE_LLAsm_COMPILER_WORKS})
-  unset(CMAKE_LLAsm_COMPILER_WORKS CACHE)
-  set(LLAsm_TEST_WAS_RUN 1)
-endif()
-
-if(NOT CMAKE_LLAsm_COMPILER_WORKS)
-  PrintTestCompilerStatus("LLAsm" " -- broken")
-  file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeError.log
-    "Determining if the LLAsm compiler works failed with "
-    "the following output:\n${__CMAKE_LLAsm_COMPILER_OUTPUT}\n\n")
-  message(FATAL_ERROR "The LLAsm compiler \"${CMAKE_LLAsm_COMPILER}\" "
-    "is not able to compile a simple test program.\nIt fails "
-    "with the following output:\n ${__CMAKE_LLAsm_COMPILER_OUTPUT}\n\n"
-    "CMake will not be able to correctly generate this project.")
-else()
-  if(LLAsm_TEST_WAS_RUN)
-    PrintTestCompilerStatus("LLAsm" " -- works")
-    file(APPEND ${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/CMakeOutput.log
-      "Determining if the LLAsm compiler works passed with "
-      "the following output:\n${__CMAKE_LLAsm_COMPILER_OUTPUT}\n\n")
-  endif()
-
-  include(${CMAKE_PLATFORM_INFO_DIR}/CMakeLLAsmCompiler.cmake)
-
-endif()
-
-unset(__CMAKE_LLAsm_COMPILER_OUTPUT)
diff --git a/libclc/cmake/modules/AddLibclc.cmake b/libclc/cmake/modules/AddLibclc.cmake
new file mode 100644
index 0000000000000..5e09cde8035c2
--- /dev/null
+++ b/libclc/cmake/modules/AddLibclc.cmake
@@ -0,0 +1,156 @@
+# Compiles an OpenCL C - or assembles an LL file - to bytecode
+#
+# Arguments:
+# * TRIPLE <string>
+#     Target triple for which to compile the bytecode file.
+# * INPUT <string>
+#     File to compile/assemble to bytecode
+# * OUTPUT <string>
+#     Bytecode file to generate
+# * EXTRA_OPTS <string> ...
+#     List of compiler options to use. Note that some are added by default.
+# * DEPENDENCIES <string> ...
+#     List of extra dependencies to inject
+#
+# Depends on the libclc::clang and libclc::llvm-as targets for compiling and
+# assembling, respectively.
+function(compile_to_bc)
+  cmake_parse_arguments(ARG
+    ""
+    "TRIPLE;INPUT;OUTPUT"
+    "EXTRA_OPTS;DEPENDENCIES"
+    ${ARGN}
+  )
+
+  # If this is an LLVM IR file (identified soley by its file suffix),
+  # pre-process it with clang to a temp file, then assemble that to bytecode.
+  set( TMP_SUFFIX )
+  get_filename_component( FILE_EXT ${ARG_INPUT} EXT )
+  if( NOT ${FILE_EXT} STREQUAL ".ll" )
+    # Pass '-c' when not running the preprocessor
+    set( PP_OPTS -c )
+  else()
+    set( PP_OPTS -E;-P )
+    set( TMP_SUFFIX .tmp )
+  endif()
+
+  set( TARGET_ARG )
+  if( ARG_TRIPLE )
+    set( TARGET_ARG "-target" ${ARG_TRIPLE} )
+  endif()
+
+  add_custom_command(
+    OUTPUT ${ARG_OUTPUT}${TMP_SUFFIX}
+    COMMAND libclc::clang
+      ${TARGET_ARG}
+      ${PP_OPTS}
+      ${ARG_EXTRA_OPTS}
+      -MD -MF ${ARG_OUTPUT}.d -MT ${ARG_OUTPUT}${TMP_SUFFIX}
+      # LLVM 13 enables standard includes by default - we don't want
+      # those when pre-processing IR. We disable it unconditionally.
+      $<$<VERSION_GREATER_EQUAL:${LLVM_PACKAGE_VERSION},13.0.0>:-cl-no-stdinc>
+      -emit-llvm
+      -o ${ARG_OUTPUT}${TMP_SUFFIX}
+      -x cl
+      ${ARG_INPUT}
+    DEPENDS
+      libclc::clang
+      ${ARG_INPUT}
+      ${ARG_DEPENDENCIES}
+    DEPFILE ${ARG_OUTPUT}.d
+  )
+
+  if( ${FILE_EXT} STREQUAL ".ll" )
+    add_custom_command(
+      OUTPUT ${ARG_OUTPUT}
+      COMMAND libclc::llvm-as -o ${ARG_OUTPUT} ${ARG_OUTPUT}${TMP_SUFFIX}
+      DEPENDS libclc::llvm-as ${ARG_OUTPUT}${TMP_SUFFIX}
+    )
+  endif()
+endfunction()
+
+# Links together one or more bytecode files
+#
+# Arguments:
+# * TARGET <string>
+#     Custom target to create
+# * INPUT <string> ...
+#     List of bytecode files to link together
+function(link_bc)
+  cmake_parse_arguments(ARG
+    ""
+    "TARGET"
+    "INPUTS"
+    ${ARGN}
+  )
+
+  add_custom_command(
+    OUTPUT ${ARG_TARGET}.bc
+    COMMAND libclc::llvm-link -o ${ARG_TARGET}.bc ${ARG_INPUTS}
+    DEPENDS libclc::llvm-link ${ARG_INPUTS}
+  )
+
+  add_custom_target( ${ARG_TARGET} ALL DEPENDS ${ARG_TARGET}.bc )
+  set_target_properties( ${ARG_TARGET} PROPERTIES TARGET_FILE ${ARG_TARGET}.bc )
+endfunction()
+
+# Decomposes and returns variables based on a libclc triple and architecture
+# combination. Returns data via one or more optional output variables.
+#
+# Arguments:
+# * TRIPLE <string>
+#     libclc target triple to query
+# * DEVICE <string>
+#     libclc device to query
+#
+# Optional Arguments:
+# * CPU <var>
+#     Variable name to be set to the target CPU
+# * ARCH_SUFFIX <var>
+#     Variable name to be set to the triple/architecture suffix
+# * CLANG_TRIPLE <var>
+#     Variable name to be set to the normalized clang triple
+function(get_libclc_device_info)
+  cmake_parse_arguments(ARG
+    ""
+    "TRIPLE;DEVICE;CPU;ARCH_SUFFIX;CLANG_TRIPLE"
+    ""
+    ${ARGN}
+  )
+
+  if( NOT ARG_TRIPLE OR NOT ARG_DEVICE )
+    message( FATAL_ERROR "Must provide both TRIPLE and DEVICE" )
+  endif()
+
+  string( REPLACE "-" ";" TRIPLE  ${ARG_TRIPLE} )
+  list( GET TRIPLE 0 ARCH )
+
+  # Some targets don't have a specific device architecture to target
+  if( ARG_DEVICE STREQUAL none OR ARCH STREQUAL spirv OR ARCH STREQUAL spirv64 )
+    set( cpu )
+    set( arch_suffix "${ARG_TRIPLE}" )
+  else()
+    set( cpu "${ARG_DEVICE}" )
+    set( arch_suffix "${ARG_DEVICE}-${ARG_TRIPLE}" )
+  endif()
+
+  if( ARG_CPU )
+    set( ${ARG_CPU} ${cpu} PARENT_SCOPE )
+  endif()
+
+  if( ARG_ARCH_SUFFIX )
+    set( ${ARG_ARCH_SUFFIX} ${arch_suffix} PARENT_SCOPE )
+  endif()
+
+  # Some libclc targets are not real clang triples: return their canonical
+  # triples.
+  if( ARCH STREQUAL spirv OR ARCH STREQUAL clspv )
+    set( ARG_TRIPLE "spir--" )
+  elseif( ARCH STREQUAL spirv64 OR ARCH STREQUAL clspv64 )
+    set( ARG_TRIPLE "spir64--" )
+  endif()
+
+  if( ARG_CLANG_TRIPLE )
+    set( ${ARG_CLANG_TRIPLE} ${ARG_TRIPLE} PARENT_SCOPE )
+  endif()
+endfunction()
diff --git a/libclc/generic/lib/SOURCES b/libclc/generic/lib/SOURCES
index ee2736b5fbc57..579e909e53d46 100644
--- a/libclc/generic/lib/SOURCES
+++ b/libclc/generic/lib/SOURCES
@@ -48,7 +48,6 @@ cl_khr_int64_extended_atomics/atom_max.cl
 cl_khr_int64_extended_atomics/atom_min.cl
 cl_khr_int64_extended_atomics/atom_or.cl
 cl_khr_int64_extended_atomics/atom_xor.cl
-convert.cl
 common/degrees.cl
 common/mix.cl
 common/radians.cl
diff --git a/llvm/tools/CMakeLists.txt b/llvm/tools/CMakeLists.txt
index cde57367934e4..db66dad5dc0db 100644
--- a/llvm/tools/CMakeLists.txt
+++ b/llvm/tools/CMakeLists.txt
@@ -52,6 +52,9 @@ add_llvm_implicit_projects()
 
 add_llvm_external_project(polly)
 
+# libclc depends on clang
+add_llvm_external_project(libclc)
+
 # Add subprojects specified using LLVM_EXTERNAL_PROJECTS
 foreach(p ${LLVM_EXTERNAL_PROJECTS})
   add_llvm_external_project(${p})

From 0d96422768908a8235f05a5d3b1d43ecc6038004 Mon Sep 17 00:00:00 2001
From: Derek Schuff <dschuff@chromium.org>
Date: Thu, 11 Apr 2024 09:30:48 -0700
Subject: [PATCH 152/886] [Object][Wasm] Move wasm Object tests into their own
 directory (NFC) (#81072)

---
 .../{wasm-bad-data-symbol.yaml => Wasm/bad-data-symbol.yaml}   | 0
 .../bad-metadata-version.yaml}                                 | 0
 llvm/test/Object/Wasm/bad-reloc-type.test                      | 3 +++
 .../{wasm-bad-relocation.yaml => Wasm/bad-relocation.yaml}     | 0
 llvm/test/Object/Wasm/bad-symbol-type.test                     | 3 +++
 .../{wasm-duplicate-name.test => Wasm/duplicate-name.test}     | 0
 .../Object/{wasm-invalid-file.yaml => Wasm/invalid-file.yaml}  | 0
 .../invalid-section-order.test}                                | 2 +-
 .../{wasm-invalid-start.test => Wasm/invalid-start.test}       | 0
 .../linked-namesec-with-linkingsec.yaml}                       | 0
 .../linked-symbol-table.yaml}                                  | 0
 llvm/test/Object/Wasm/missing-version.test                     | 2 ++
 .../{wasm-obj2yaml-tables.test => Wasm/obj2yaml-tables.test}   | 2 +-
 .../relocs-and-producers.yaml}                                 | 0
 llvm/test/Object/Wasm/string-outside-section.test              | 3 +++
 llvm/test/Object/wasm-bad-reloc-type.test                      | 3 ---
 llvm/test/Object/wasm-bad-symbol-type.test                     | 3 ---
 llvm/test/Object/wasm-missing-version.test                     | 2 --
 llvm/test/Object/wasm-string-outside-section.test              | 3 ---
 19 files changed, 13 insertions(+), 13 deletions(-)
 rename llvm/test/Object/{wasm-bad-data-symbol.yaml => Wasm/bad-data-symbol.yaml} (100%)
 rename llvm/test/Object/{wasm-bad-metadata-version.yaml => Wasm/bad-metadata-version.yaml} (100%)
 create mode 100644 llvm/test/Object/Wasm/bad-reloc-type.test
 rename llvm/test/Object/{wasm-bad-relocation.yaml => Wasm/bad-relocation.yaml} (100%)
 create mode 100644 llvm/test/Object/Wasm/bad-symbol-type.test
 rename llvm/test/Object/{wasm-duplicate-name.test => Wasm/duplicate-name.test} (100%)
 rename llvm/test/Object/{wasm-invalid-file.yaml => Wasm/invalid-file.yaml} (100%)
 rename llvm/test/Object/{wasm-invalid-section-order.test => Wasm/invalid-section-order.test} (82%)
 rename llvm/test/Object/{wasm-invalid-start.test => Wasm/invalid-start.test} (100%)
 rename llvm/test/Object/{wasm-linked-namesec-with-linkingsec.yaml => Wasm/linked-namesec-with-linkingsec.yaml} (100%)
 rename llvm/test/Object/{wasm-linked-symbol-table.yaml => Wasm/linked-symbol-table.yaml} (100%)
 create mode 100644 llvm/test/Object/Wasm/missing-version.test
 rename llvm/test/Object/{wasm-obj2yaml-tables.test => Wasm/obj2yaml-tables.test} (98%)
 rename llvm/test/Object/{wasm-relocs-and-producers.yaml => Wasm/relocs-and-producers.yaml} (100%)
 create mode 100644 llvm/test/Object/Wasm/string-outside-section.test
 delete mode 100644 llvm/test/Object/wasm-bad-reloc-type.test
 delete mode 100644 llvm/test/Object/wasm-bad-symbol-type.test
 delete mode 100644 llvm/test/Object/wasm-missing-version.test
 delete mode 100644 llvm/test/Object/wasm-string-outside-section.test

diff --git a/llvm/test/Object/wasm-bad-data-symbol.yaml b/llvm/test/Object/Wasm/bad-data-symbol.yaml
similarity index 100%
rename from llvm/test/Object/wasm-bad-data-symbol.yaml
rename to llvm/test/Object/Wasm/bad-data-symbol.yaml
diff --git a/llvm/test/Object/wasm-bad-metadata-version.yaml b/llvm/test/Object/Wasm/bad-metadata-version.yaml
similarity index 100%
rename from llvm/test/Object/wasm-bad-metadata-version.yaml
rename to llvm/test/Object/Wasm/bad-metadata-version.yaml
diff --git a/llvm/test/Object/Wasm/bad-reloc-type.test b/llvm/test/Object/Wasm/bad-reloc-type.test
new file mode 100644
index 0000000000000..4e210271c0b93
--- /dev/null
+++ b/llvm/test/Object/Wasm/bad-reloc-type.test
@@ -0,0 +1,3 @@
+RUN: not llvm-objdump -s %p/../Inputs/WASM/bad-reloc-type.wasm 2>&1 | FileCheck %s
+
+CHECK: invalid relocation type: 63
diff --git a/llvm/test/Object/wasm-bad-relocation.yaml b/llvm/test/Object/Wasm/bad-relocation.yaml
similarity index 100%
rename from llvm/test/Object/wasm-bad-relocation.yaml
rename to llvm/test/Object/Wasm/bad-relocation.yaml
diff --git a/llvm/test/Object/Wasm/bad-symbol-type.test b/llvm/test/Object/Wasm/bad-symbol-type.test
new file mode 100644
index 0000000000000..5b9770f18bb62
--- /dev/null
+++ b/llvm/test/Object/Wasm/bad-symbol-type.test
@@ -0,0 +1,3 @@
+RUN: not llvm-objdump -s %p/../Inputs/WASM/bad-symbol-type.wasm 2>&1 | FileCheck %s
+
+CHECK: invalid symbol type: 63
diff --git a/llvm/test/Object/wasm-duplicate-name.test b/llvm/test/Object/Wasm/duplicate-name.test
similarity index 100%
rename from llvm/test/Object/wasm-duplicate-name.test
rename to llvm/test/Object/Wasm/duplicate-name.test
diff --git a/llvm/test/Object/wasm-invalid-file.yaml b/llvm/test/Object/Wasm/invalid-file.yaml
similarity index 100%
rename from llvm/test/Object/wasm-invalid-file.yaml
rename to llvm/test/Object/Wasm/invalid-file.yaml
diff --git a/llvm/test/Object/wasm-invalid-section-order.test b/llvm/test/Object/Wasm/invalid-section-order.test
similarity index 82%
rename from llvm/test/Object/wasm-invalid-section-order.test
rename to llvm/test/Object/Wasm/invalid-section-order.test
index 9a67f09150e5e..93cf9eb57059c 100644
--- a/llvm/test/Object/wasm-invalid-section-order.test
+++ b/llvm/test/Object/Wasm/invalid-section-order.test
@@ -1,4 +1,4 @@
-# RUN: not obj2yaml %p/Inputs/WASM/invalid-section-order.wasm 2>&1 | FileCheck %s
+# RUN: not obj2yaml %p/../Inputs/WASM/invalid-section-order.wasm 2>&1 | FileCheck %s
 # CHECK: {{.*}}: out of order section type: 10
 
 # Inputs/WASM/invalid-section-order.wasm is generated from this ll file, by
diff --git a/llvm/test/Object/wasm-invalid-start.test b/llvm/test/Object/Wasm/invalid-start.test
similarity index 100%
rename from llvm/test/Object/wasm-invalid-start.test
rename to llvm/test/Object/Wasm/invalid-start.test
diff --git a/llvm/test/Object/wasm-linked-namesec-with-linkingsec.yaml b/llvm/test/Object/Wasm/linked-namesec-with-linkingsec.yaml
similarity index 100%
rename from llvm/test/Object/wasm-linked-namesec-with-linkingsec.yaml
rename to llvm/test/Object/Wasm/linked-namesec-with-linkingsec.yaml
diff --git a/llvm/test/Object/wasm-linked-symbol-table.yaml b/llvm/test/Object/Wasm/linked-symbol-table.yaml
similarity index 100%
rename from llvm/test/Object/wasm-linked-symbol-table.yaml
rename to llvm/test/Object/Wasm/linked-symbol-table.yaml
diff --git a/llvm/test/Object/Wasm/missing-version.test b/llvm/test/Object/Wasm/missing-version.test
new file mode 100644
index 0000000000000..754658e9d0ddc
--- /dev/null
+++ b/llvm/test/Object/Wasm/missing-version.test
@@ -0,0 +1,2 @@
+# RUN: not llvm-objdump -h %p/../Inputs/WASM/missing-version.wasm 2>&1 | FileCheck %s
+# CHECK: {{.*}}: missing version number
diff --git a/llvm/test/Object/wasm-obj2yaml-tables.test b/llvm/test/Object/Wasm/obj2yaml-tables.test
similarity index 98%
rename from llvm/test/Object/wasm-obj2yaml-tables.test
rename to llvm/test/Object/Wasm/obj2yaml-tables.test
index 870ffd179a1c2..877238437b010 100644
--- a/llvm/test/Object/wasm-obj2yaml-tables.test
+++ b/llvm/test/Object/Wasm/obj2yaml-tables.test
@@ -1,4 +1,4 @@
-RUN: obj2yaml %p/Inputs/WASM/multi-table.wasm | FileCheck %s
+RUN: obj2yaml %p/../Inputs/WASM/multi-table.wasm | FileCheck %s
 
 
 # CHECK:  - Type:            TABLE
diff --git a/llvm/test/Object/wasm-relocs-and-producers.yaml b/llvm/test/Object/Wasm/relocs-and-producers.yaml
similarity index 100%
rename from llvm/test/Object/wasm-relocs-and-producers.yaml
rename to llvm/test/Object/Wasm/relocs-and-producers.yaml
diff --git a/llvm/test/Object/Wasm/string-outside-section.test b/llvm/test/Object/Wasm/string-outside-section.test
new file mode 100644
index 0000000000000..31f4a6080d9bd
--- /dev/null
+++ b/llvm/test/Object/Wasm/string-outside-section.test
@@ -0,0 +1,3 @@
+RUN: not --crash llvm-objdump -s %p/../Inputs/WASM/string-outside-section.wasm 2>&1 | FileCheck %s
+
+CHECK: LLVM ERROR: EOF while reading string
diff --git a/llvm/test/Object/wasm-bad-reloc-type.test b/llvm/test/Object/wasm-bad-reloc-type.test
deleted file mode 100644
index 000acbd55fce5..0000000000000
--- a/llvm/test/Object/wasm-bad-reloc-type.test
+++ /dev/null
@@ -1,3 +0,0 @@
-RUN: not llvm-objdump -s %p/Inputs/WASM/bad-reloc-type.wasm 2>&1 | FileCheck %s
-
-CHECK: invalid relocation type: 63
diff --git a/llvm/test/Object/wasm-bad-symbol-type.test b/llvm/test/Object/wasm-bad-symbol-type.test
deleted file mode 100644
index 4b7c30ad8b738..0000000000000
--- a/llvm/test/Object/wasm-bad-symbol-type.test
+++ /dev/null
@@ -1,3 +0,0 @@
-RUN: not llvm-objdump -s %p/Inputs/WASM/bad-symbol-type.wasm 2>&1 | FileCheck %s
-
-CHECK: invalid symbol type: 63
diff --git a/llvm/test/Object/wasm-missing-version.test b/llvm/test/Object/wasm-missing-version.test
deleted file mode 100644
index beb0b5e9105bf..0000000000000
--- a/llvm/test/Object/wasm-missing-version.test
+++ /dev/null
@@ -1,2 +0,0 @@
-# RUN: not llvm-objdump -h %p/Inputs/WASM/missing-version.wasm 2>&1 | FileCheck %s
-# CHECK: {{.*}}: missing version number
diff --git a/llvm/test/Object/wasm-string-outside-section.test b/llvm/test/Object/wasm-string-outside-section.test
deleted file mode 100644
index 3fa6217bae8e3..0000000000000
--- a/llvm/test/Object/wasm-string-outside-section.test
+++ /dev/null
@@ -1,3 +0,0 @@
-RUN: not --crash llvm-objdump -s %p/Inputs/WASM/string-outside-section.wasm 2>&1 | FileCheck %s
-
-CHECK: LLVM ERROR: EOF while reading string

From 3c4b673af05f53e8a4d1a382b5c86367ea512c9e Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Thu, 11 Apr 2024 12:36:56 -0400
Subject: [PATCH 153/886] [libc++] Fix -Wgnu-include-next in stddef.h (#88214)

As reported in #86843, we must have #pragma GCC system_header before we
use #include_next, otherwise the compiler may not understand that we're
in a system header and may issue a diagnostic for our usage of
---
 libcxx/include/stddef.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/libcxx/include/stddef.h b/libcxx/include/stddef.h
index 470b5408336c6..1583e78e3739b 100644
--- a/libcxx/include/stddef.h
+++ b/libcxx/include/stddef.h
@@ -26,6 +26,10 @@
 
 #include <__config>
 
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
 // Note: This include is outside of header guards because we sometimes get included multiple times
 //       with different defines and the underlying <stddef.h> will know how to deal with that.
 #include_next <stddef.h>
@@ -33,10 +37,6 @@
 #ifndef _LIBCPP_STDDEF_H
 #  define _LIBCPP_STDDEF_H
 
-#  if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
-#    pragma GCC system_header
-#  endif
-
 #  ifdef __cplusplus
 typedef decltype(nullptr) nullptr_t;
 #  endif

From 357f6c7826437f6527db6f99f756a34fb5e0f716 Mon Sep 17 00:00:00 2001
From: abidh <haqadeer@amd.com>
Date: Thu, 11 Apr 2024 17:53:25 +0100
Subject: [PATCH 154/886] [flang] Add design document for debug info
 generation. (#86939)

This document discusses some options where the debug metadata can be
generated. It also goes through various language constructs and explains
how the debug metadata will look like for that construct and how we can
extract that information.

The real point of discussion is how and where to extract the information
about various language features to generate the debug metadata. The
structure of the metadata itself is mostly settled as that is dictated by
the DWARF and structure of LLVM IR metadata. The classic flang
and gfortran generate quite similar DWARF for the various language
constructs.

This document is based on what Kiran posted in
https://reviews.llvm.org/D138534.

---------

Co-authored-by: Tom Eccles <t@freedommail.info>
Co-authored-by: Kiran Chandramohan <kiranchandramohan@gmail.com>
---
 flang/docs/DebugGeneration.md | 442 ++++++++++++++++++++++++++++++++++
 flang/docs/index.md           |   1 +
 2 files changed, 443 insertions(+)
 create mode 100644 flang/docs/DebugGeneration.md

diff --git a/flang/docs/DebugGeneration.md b/flang/docs/DebugGeneration.md
new file mode 100644
index 0000000000000..9409d7e07b104
--- /dev/null
+++ b/flang/docs/DebugGeneration.md
@@ -0,0 +1,442 @@
+# Debug Generation
+
+Application developers spend a significant time debugging the applications that
+they create. Hence it is important that a compiler provide support for a good
+debug experience. DWARF[1] is the standard debugging file format used by
+compilers and debuggers. The LLVM infrastructure supports debug info generation
+using metadata[2]. Support for generating debug metadata is present
+in MLIR by way of MLIR attributes. Flang can leverage these MLIR attributes to
+generate good debug information.
+
+We can break the work for debug generation into two separate tasks:
+1) Line Table generation
+2) Full debug generation
+The support for Fortran Debug in LLVM infrastructure[3] has made great progress
+due to many Fortran frontends adopting LLVM as the backend as well as the
+availability of the Classic Flang compiler.
+
+## Driver Flags
+By default, Flang will not generate any debug or linetable information.
+Debug information will be generated if the following flags are present.
+
+-gline-tables-only, -g1 : Emit debug line number tables only  
+-g : Emit full debug info
+
+## Line Table Generation
+
+There is existing AddDebugFoundationPass which add `FusedLoc` with a
+`SubprogramAttr` on FuncOp. This allows MLIR to generate LLVM IR metadata
+for that function. However, following values are hardcoded at the moment. These
+will instead be passed from the driver.
+
+- Details of the compiler (name and version and git hash).
+- Language Standard. We can set it to Fortran95 for now and periodically
+revise it when full support for later standards is available.
+- Optimisation Level.
+- Type of debug generated (linetable/full debug).
+- Calling Convention: `DW_CC_normal` by default and `DW_CC_program` if it is
+the main program.
+
+`DISubroutineTypeAttr` currently has a fixed type. This will be changed to
+match the signature of the actual function/subroutine.
+
+
+## Full Debug Generation
+
+Full debug info will include metadata to describe functions, variables and
+types. Flang will generate debug metadata in the form of MLIR attributes. These
+attributes will be converted to the format expected by LLVM IR in DebugTranslation[4].
+
+Debug metadata generation can be broken down in 2 steps.
+
+1. MLIR attributes are generated by reading information from AST or FIR. This
+step can happen anytime before or during conversion to LLVM dialect. An example
+of the metadata generated in this step is `DILocalVariableAttr` or
+`DIDerivedTypeAttr`.
+
+2. Changes that can only happen during or after conversion to LLVM dialect. The
+example of this is passing `DIGlobalVariableExpressionAttr` while
+creating `LLVM::GlobalOp`. Another example will be generation of `DbgDeclareOp`
+that is required for local variables. It can only be created after conversion to
+LLVM dialect as it requires LLVM.Ptr type. The changes required for step 2 are
+quite minimal. The bulk of the work happens in step 1.
+
+One design decision that we need to make is to decide where to perform step 1.
+Here are some possible options:
+
+**During conversion to LLVM dialect**
+
+Pros:
+1. Do step 1 and 2 in one place.
+2. No chance of missing any change introduced by an earlier transformation.
+
+Cons:
+1. Passing a lot of information from the driver as discussed in the line table
+section above may muddle interface of FIRToLLVMConversion.
+2. `DeclareOp` is removed before this pass.
+3. Even if `DeclareOp` is retained, creating debug metadata while some ops have
+been converted to LLVMdialect and others are not may cause its own issues. We
+have to walk the ops chain to extract the information which may be problematic
+in this case.
+4. Some source information is lost by this point. Examples include
+information about namelists, source line information about field of derived
+types etc.
+
+**During a pass before conversion to LLVM dialect**
+
+This is similar to what AddDebugFoundationPass is currently doing.
+
+Pros:
+1. One central location dedicated to debug information processing. This can
+result in a cleaner implementation.
+2. Similar to above, less chance of missing any change introduced by an earlier
+transformation.
+
+Cons:
+1. Step 2 still need to happen during conversion to LLVM dialect. But
+changes required for step 2 are quite minimal.
+2. Similar to above, some source information may be lost by this point.
+
+**During Lowering from AST**
+
+Pros
+1. We have better source information.
+
+Cons:
+1. There may be change in the code after lowering which may not be
+reflected in debug information.
+2. Comments on an earlier PR [5] advised against this approach.
+
+## Design
+
+The design below assumes that we are extracting the information from FIR.
+If we generate debug metadata during lowering then the description below
+may need to change. Although the generated metadata remains the same in
+both cases.
+
+The AddDebugFoundationPass will be renamed to AddDebugInfo Pass. The
+information mentioned in the line info section above will be passed to it from
+the driver. This pass will run quite late in the pipeline but before
+`DeclareOp` is removed.
+
+In this pass, we will iterate through the `GlobalOp`, `TypeInfoOp`, `FuncOp`
+and `DeclareOp` to extract the source information and build the MLIR
+attributes. A class will be added to handle conversion of MLIR and FIR types to
+`DITypeAttr`.
+
+Following sections provide details of how various language constructs will be
+handled. In these sections, the LLVM IR metadata and MLIR attributes have been
+used interchangeably. As an example, `DILocalVariableAttr` is an MLIR attribute
+which gets translated to LLVM IR's `DILocalVariable`.
+
+### Variables
+
+#### Local Variables
+  In MLIR, local variables are represented by `DILocalVariableAttr` which
+  stores information like source location and type. They also require a
+  `DbgDeclareOp` which binds `DILocalVariableAttr` with a location.
+
+  In FIR, `DeclareOp` has source information about the variable. The
+  `DeclareOp` will be processed to create `DILocalVariableAttr`. This attr is
+  attached to the memref op of the `DeclareOp` using a `FusedLoc` approach.
+
+  During conversion to LLVM dialect, when an op is encountered that has a
+  `DILocalVariableAttr` in its `FusedLoc`, a `DbgDeclareOp` is created which
+  binds the attr with its location.
+
+  The change in the IR look like as follows:
+
+```
+  original fir
+  %2 = fir.alloca i32  loc(#loc4)
+  %3 = fir.declare %2 {uniq_name = "_QMhelperFchangeEi"}
+
+  Fir with FusedLoc.
+
+  %2 = fir.alloca i32  loc(#loc38)
+  %3 = fir.declare %2 {uniq_name = "_QMhelperFchangeEi"}
+  #di_local_variable5 = #llvm.di_local_variable<name = "i", line = 5, type = #di_basic_type ... >
+  #loc38 = loc(fused<#di_local_variable5>[#loc4])
+
+  After conversion to llvm dialect
+
+  #di_local_variable = #llvm.di_local_variable<name = "i", line = 5, type = #di_basic_type ...>
+  %1 = llvm.alloca %0 x i64
+  llvm.intr.dbg.declare #di_local_variable = %1
+```
+
+#### Function Arguments
+
+Arguments work in similar way, but they present a difficulty that `DeclareOp`'s
+memref points to `BlockArgument`. Unlike the op in local variable case,
+the `BlockArgument` are not handled by the FIRToLLVMLowering. This can easily
+be handled by adding after conversion to LLVM dialect either in FIRToLLVMLowering 
+or in a separate pass.
+
+### Module
+
+In debug metadata, the Fortran module will be represented by `DIModuleAttr`.
+The variables or functions inside module will have scope pointing to the parent module.
+
+```
+module helper
+   real glr
+   ...
+end module helper
+
+!1 = !DICompileUnit(language: DW_LANG_Fortran90 ...)
+!2 = !DIModule(scope: !1, name: "helper" ...)
+!3 = !DIGlobalVariable(scope: !2, name: "glr" ...)
+
+Use of a module results in the following metadata.
+!4 = !DIImportedEntity(tag: DW_TAG_imported_module, entity: !2)
+```
+
+Modules are not first class entities in the FIR. So there is no way to get
+the location where they are declared in source file.
+
+But the information that a variable or function is part of a module
+can be extracted from its mangled name along with name of the module. There is
+a `GlobalOp` generated for each module variable in FIR and there is also a
+`DeclareOp` in each function where the module variable is used.
+
+We will use the `GlobalOp` to generate the `DIModuleAttr` and associated
+`DIGlobalVariableAttr`. A `DeclareOp` for module variable will be used
+to generate `DIImportedEntityAttr`. Care will be taken to avoid generating
+duplicate `DIImportedEntityAttr` entries in same function.
+
+### Derived Types
+
+A derived type will be represented in metadata by `DICompositeType` with a tag of
+`DW_TAG_structure_type`. It will have elements which point to the components.
+
+```
+  type :: t_pair
+    integer :: i
+    real :: x
+  end type
+!1 = !DICompositeType(tag: DW_TAG_structure_type, name: "t_pair", elements: !2 ...)
+!2 = !{!3, !4}
+!3 = !DIDerivedType(tag: DW_TAG_member, scope: !1, name: "i", size: 32, offset: 0, baseType: !5 ...)
+!4 = !DIDerivedType(tag: DW_TAG_member, scope: !1, name: "x", size: 32, offset: 32, baseType: !6 ...)
+!5 = !DIBasicType(tag: DW_TAG_base_type, name: "integer" ...)
+!6 = !DIBasicType(tag: DW_TAG_base_type, name: "real" ...)
+```
+
+In FIR, `RecordType` and `TypeInfoOp` can be used to get information about the
+location of the derived type and the types of its components. We may also use
+`FusedLoc` on `TypeInfoOp` to encode location information for all the components
+of the derived type.
+
+### CommonBlocks
+
+A common block will be represented in metadata by `DICommonBlockAttr` which
+will be used as scope by the variable inside common block. `DIExpression`
+can be used to give the offset of any given variable inside the global storage
+for common block.
+
+```
+integer a, b
+common /test/ a, b
+
+;@test_ = common global [8 x i8] zeroinitializer, !dbg !5, !dbg !6
+!1 = !DISubprogram()
+!2 = !DICommonBlock(scope: !1, name: "test" ...)
+!3 = !DIGlobalVariable(scope: !2, name: "a" ...)
+!4 = !DIExpression()
+!5 = !DIGlobalVariableExpression(var: !3, expr: !4)
+!6 = !DIGlobalVariable(scope: !2, name: "b" ...)
+!7 = !DIExpression(DW_OP_plus_uconst, 4)
+!8 = !DIGlobalVariableExpression(var: !6, expr: !7)
+```
+
+In FIR, a common block results in a `GlobalOp` with common linkage. Every
+function where the common block is used has `DeclareOp` for that variable.
+This `DeclareOp` will point to global storage through
+`CoordinateOp` and `AddrOfOp`. The `CoordinateOp` has the offset of the
+location of this variable in global storage. There is enough information to
+generate the required metadata. Although it requires walking up the chain from
+`DeclaredOp` to locate `CoordinateOp` and `AddrOfOp`.
+
+### Arrays
+
+The type of fixed size array is represented using `DICompositeType`. The
+`DISubrangeAttr` is used to provide bounds in any given dimensions.
+
+```
+integer abc(4,5)
+
+!1 = !DICompositeType(tag: DW_TAG_array_type, baseType: !5, elements: !2 ...)
+!2 = !{ !3, !4 }
+!3 = !DISubrange(lowerBound: 1, upperBound: 4 ...)
+!4 = !DISubrange(lowerBound: 1, upperBound: 5 ...)
+!5 = !DIBasicType(tag: DW_TAG_base_type, name: "integer" ...)
+
+```
+
+#### Adjustable
+
+The debug metadata for the adjustable array looks similar to fixed sized array
+with one change. The bounds are not constant values but point to a
+`DILocalVariableAttr`.
+
+In FIR, the `DeclareOp` points to a `ShapeOp` and we can walk the chain
+to get the value that represents the array bound in any dimension. We will
+create a `DILocalVariableAttr` that will point to that location. This
+variable will be used in the `DISubrangeAttr`. Note that this
+`DILocalVariableAttr` does not correspond to any source variable.
+
+#### Assumed Size
+
+This is treated as raw array. Debug information will not provide any upper bound
+information for the last dimension.
+
+#### Assumed Shape
+The assumed shape array will use the similar representation as fixed size
+array but there will be 2 differences.
+
+1. There will be a `datalocation` field which will be an expression. This will
+enable debugger to get the data pointer from array descriptor.
+
+2. The field in `DISubrangeAttr` for array bounds will be expression which will
+allow the debugger to get the bounds from descriptor.
+
+```
+integer(4), intent(out) :: a(:,:)
+
+!1 = !DICompositeType(tag: DW_TAG_array_type, baseType: !8, elements: !2, dataLocation: !3)
+!2 = !{!5, !7}
+!3 = !DIExpression(DW_OP_push_object_address, DW_OP_deref)
+!4 = !DIExpression(DW_OP_push_object_address, DW_OP_plus_uconst, 32, DW_OP_deref)
+!5 = !DISubrange(lowerBound: !1, upperBound: !4 ...)
+!6 = !DIExpression(DW_OP_push_object_address, DW_OP_plus_uconst, 56, DW_OP_deref)
+!7 = !DISubrange(lowerBound: !1, upperBound: !6, ...)
+!8 = !DIBasicType(tag: DW_TAG_base_type, name: "integer" ...)
+```
+
+In assumed shape case, the rank can be determined from the FIR's `SequenceType`.
+This allows us to generate a `DISubrangeAttr` in each dimension.
+
+#### Assumed Rank
+
+This is currently unsupported in flang. Its representation will be similar to
+array representation for assumed shape array with the following difference.
+
+1. `DICompositeTypeAttr` will have a rank field which will be an expression.
+It will be used to get the rank value from descriptor.
+2. Instead of `DISubrangeType` for each dimension, there will be a single
+`DIGenericSubrange` which will allow debuggers to calculate bounds in any
+dimension.
+
+### Pointers and Allocatables
+The pointer and allocatable will be represented using `DICompositeTypeAttr`. It
+is quirk of DWARF that scalar allocatable or pointer variables will show up in
+the debug info as pointer to scalar while array pointer or allocatable
+variables show up as arrays. The behavior is same in gfortran and classic flang.
+
+```
+  integer, allocatable :: ar(:)
+  integer, pointer :: sc
+
+!1 = !DILocalVariable(name: "sc", type: !2)
+!2 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !3, associated: !9 ...)
+!3 = !DIBasicType(tag: DW_TAG_base_type, name: "integer", ...)
+!4 = !DILocalVariable(name: "ar", type: !5 ...)
+!5 = !DICompositeType(tag: DW_TAG_array_type, baseType: !3, elements: !6, dataLocation: !8, allocated: !9)
+!6 = !{!7}
+!7 = !DISubrange(lowerBound: !10, upperBound: !11 ...)
+!8 = !DIExpression(DW_OP_push_object_address, DW_OP_deref)
+!9 = !DIExpression(DW_OP_push_object_address, DW_OP_deref, DW_OP_lit0, DW_OP_ne)
+!10 = !DIExpression(DW_OP_push_object_address, DW_OP_plus_uconst, 24, DW_OP_deref)
+!11 = !DIExpression(DW_OP_push_object_address, DW_OP_plus_uconst, 32, DW_OP_deref)
+
+```
+
+IN FIR, these variable are represent as <!fir.box<!fir.heap<>> or
+fir.box<!fir.ptr<>>. There is also `allocatable` or `pointer` attribute on
+the `DeclareOp`. This allows us to generate allocated/associated status of
+these variables. The metadata to get the information from the descriptor is
+similar to arrays.
+
+### Strings
+
+The `DIStringTypeAttr` can represent both fixed size and allocatable strings. For
+the allocatable case, the `stringLengthExpression` and `stringLocationExpression`
+are used to provide the length and the location of the string respectively.
+
+```
+  character(len=:), allocatable :: var
+  character(len=20) :: fixed
+
+!1 = !DILocalVariable(name: "var", type: !2)
+!2 = !DIStringType(name: "character(*)", stringLengthExpression: !4, stringLocationExpression: !3 ...)
+!3 = !DIExpression(DW_OP_push_object_address, DW_OP_deref)
+!4 = !DIExpression(DW_OP_push_object_address, DW_OP_plus_uconst, 8)
+
+!5 = !DILocalVariable(name: "fixed", type: !6)
+!6 = !DIStringType(name: "character (20)", size: 160)
+
+```
+
+### Association
+
+They will be treated like normal variables. Although we may require to handle
+the case where the `DeclareOp` of one variable points to the `DeclareOp` of
+another variable (e.g. a => b).
+
+### Namelists
+
+FIR does not seem to have a way to extract information about namelists.
+
+```
+namelist /abc/ x3, y3
+
+(gdb) p abc
+$1 = ( x3 = 100, y3 = 500 )
+(gdb) p x3
+$2 = 100
+(gdb) p y3
+$3 = 500
+```
+
+Even without namelist support, we should be able to see the value of the
+individual variables like `x3` and `y3` in the above example. But we would not
+be able to evaluate the namelist and have the debugger prints the value of all
+the variables in it as shown above for `abc`.
+
+## Missing metadata in MLIR
+
+Some metadata types that are needed for fortran are present in LLVM IR but are
+absent from MLIR. A non comprehensive list is given below.
+
+1. `DICommonBlockAttr`
+2. `DIGenericSubrangeAttr`
+3. `DISubrangeAttr` in MLIR takes IntegerAttr at the moment so only works
+with fixed sizes arrays. It needs to also accept `DIExpressionAttr` or
+`DILocalVariableAttr` to support assumed shape and adjustable arrays.
+4. The `DICompositeTypeAttr` will need to have field for `datalocation`,
+`rank`, `allocated` and `associated`.
+5. `DIStringTypeAttr`
+
+# Testing
+
+- LLVM LIT tests will be added to test:
+  - the driver and ensure that it passes the line table and full debug
+    info generation appropriately.
+  - that the pass works as expected and generates debug info. Test will be
+    with `fir-opt`.
+  - with `flang -fc1` that end-to-end debug info generation works.
+- Manual external tests will be written to ensure that the following works
+  in debug tools
+  - Break at lines.
+  - Break at functions.
+  - print type (ptype) of function names.
+  - print values and types (ptype) of various type of variables
+- Manually run `GDB`'s gdb.fortran testsuite with llvm-flang.
+
+# Resources
+- [1] https://dwarfstd.org/doc/DWARF5.pdf
+- [2] https://llvm.org/docs/LangRef.html#metadata
+- [3] https://archive.fosdem.org/2022/schedule/event/llvm_fortran_debug/
+- [4] https://github.com/llvm/llvm-project/blob/main/mlir/lib/Target/LLVMIR/DebugTranslation.cpp
+- [5] https://github.com/llvm/llvm-project/pull/84202
diff --git a/flang/docs/index.md b/flang/docs/index.md
index 4a0b145df10b0..70478fa0936d0 100644
--- a/flang/docs/index.md
+++ b/flang/docs/index.md
@@ -47,6 +47,7 @@ on how to get in touch with us and to learn more about the current status.
    Character
    ComplexOperations
    ControlFlowGraph
+   DebugGeneration
    Directives
    DoConcurrent
    Extensions

From f135d224a6d078ca3b0722db94e1d772fdbd68ad Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 11 Apr 2024 09:50:41 -0700
Subject: [PATCH 155/886] [SLP][NFC]Add a test with the zext feeding into
 sitofp instructions.

---
 .../X86/sitofp-minbitwidth-node.ll            | 57 +++++++++++++++++++
 1 file changed, 57 insertions(+)
 create mode 100644 llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll

diff --git a/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll b/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll
new file mode 100644
index 0000000000000..00dd9b7e6fb47
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll
@@ -0,0 +1,57 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=slp-vectorizer -mtriple=x86_64 -mcpu=k8 -mattr=+sse4.1 -S < %s | FileCheck %s
+
+define void @foo(ptr %ptr) {
+; CHECK-LABEL: define void @foo(
+; CHECK-SAME: ptr [[PTR:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[GEP0:%.*]] = getelementptr inbounds i8, ptr [[PTR]], i64 328
+; CHECK-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i8, ptr [[PTR]], i64 334
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i16>, ptr [[GEP0]], align 8
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i16> [[TMP1]], <i16 -1, i16 -1>
+; CHECK-NEXT:    [[TMP4:%.*]] = sitofp <2 x i16> [[TMP3]] to <2 x double>
+; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i16>, ptr [[GEP3]], align 2
+; CHECK-NEXT:    [[TMP6:%.*]] = zext <2 x i16> [[TMP5]] to <2 x i32>
+; CHECK-NEXT:    [[TMP2:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i32>
+; CHECK-NEXT:    [[TMP7:%.*]] = sub nsw <2 x i32> [[TMP6]], [[TMP2]]
+; CHECK-NEXT:    [[TMP8:%.*]] = sitofp <2 x i32> [[TMP7]] to <2 x double>
+; CHECK-NEXT:    [[TMP9:%.*]] = fdiv <2 x double> [[TMP4]], [[TMP8]]
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x double> [[TMP9]], i32 0
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <2 x double> [[TMP9]], i32 1
+; CHECK-NEXT:    [[FCMP:%.*]] = fcmp olt double [[TMP11]], [[TMP10]]
+; CHECK-NEXT:    ret void
+;
+  %gep0 = getelementptr inbounds i8, ptr %ptr, i64 328
+  %gep1 = getelementptr inbounds i8, ptr %ptr, i64 330
+
+  %gep3 = getelementptr inbounds i8, ptr %ptr, i64 334
+  %gep4 = getelementptr inbounds i8, ptr %ptr, i64 336
+
+  %ld0 = load i16, ptr %gep0, align 8
+  %ld1 = load i16, ptr %gep1, align 2
+
+  %zext0 = zext i16 %ld0 to i32
+  %zext1 = zext i16 %ld1 to i32
+
+  %xor0 = xor i32 %zext0, 65535
+  %xor1 = xor i32 %zext1, 65535
+
+  %sitofp0 = sitofp i32 %xor0 to double
+  %sitofp1 = sitofp i32 %xor1 to double
+
+  %ld3 = load i16, ptr %gep3, align 2
+  %ld4 = load i16, ptr %gep4, align 8
+
+  %zext3 = zext i16 %ld3 to i32
+  %zext4 = zext i16 %ld4 to i32
+
+  %sub30 = sub nsw i32 %zext3, %zext0
+  %sub41 = sub nsw i32 %zext4, %zext1
+
+  %sitofp30 = sitofp i32 %sub30 to double
+  %sitofp41 = sitofp i32 %sub41 to double
+
+  %fdiv030 = fdiv double %sitofp0, %sitofp30
+  %fdiv141 = fdiv double %sitofp1, %sitofp41
+  %fcmp = fcmp olt double %fdiv141, %fdiv030
+  ret void
+}

From 3749e0d43fb56cf22cc72274c287b7bfdda9821d Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Thu, 11 Apr 2024 09:54:41 -0700
Subject: [PATCH 156/886] [memprof] Use structured binding (NFC) (#88363)

---
 llvm/lib/ProfileData/MemProfReader.cpp     | 20 +++++++++-----------
 llvm/tools/llvm-profdata/llvm-profdata.cpp |  4 ++--
 2 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/ProfileData/MemProfReader.cpp b/llvm/lib/ProfileData/MemProfReader.cpp
index 4ccec26597c09..580867a9083fd 100644
--- a/llvm/lib/ProfileData/MemProfReader.cpp
+++ b/llvm/lib/ProfileData/MemProfReader.cpp
@@ -142,13 +142,13 @@ CallStackMap readStackInfo(const char *Ptr) {
 // any stack ids observed previously map to a different set of program counter
 // addresses.
 bool mergeStackMap(const CallStackMap &From, CallStackMap &To) {
-  for (const auto &IdStack : From) {
-    auto I = To.find(IdStack.first);
+  for (const auto &[Id, Stack] : From) {
+    auto I = To.find(Id);
     if (I == To.end()) {
-      To[IdStack.first] = IdStack.second;
+      To[Id] = Stack;
     } else {
       // Check that the PCs are the same (in order).
-      if (IdStack.second != I->second)
+      if (Stack != I->second)
         return true;
     }
   }
@@ -275,10 +275,10 @@ void RawMemProfReader::printYAML(raw_ostream &OS) {
   }
   // Print out the merged contents of the profiles.
   OS << "  Records:\n";
-  for (const auto &Entry : *this) {
+  for (const auto &[GUID, Record] : *this) {
     OS << "  -\n";
-    OS << "    FunctionGUID: " << Entry.first << "\n";
-    Entry.second.print(OS);
+    OS << "    FunctionGUID: " << GUID << "\n";
+    Record.print(OS);
   }
 }
 
@@ -405,9 +405,7 @@ Error RawMemProfReader::mapRawProfileToRecords() {
 
   // Convert the raw profile callstack data into memprof records. While doing so
   // keep track of related contexts so that we can fill these in later.
-  for (const auto &Entry : CallstackProfileData) {
-    const uint64_t StackId = Entry.first;
-
+  for (const auto &[StackId, MIB] : CallstackProfileData) {
     auto It = StackMap.find(StackId);
     if (It == StackMap.end())
       return make_error<InstrProfError>(
@@ -455,7 +453,7 @@ Error RawMemProfReader::mapRawProfileToRecords() {
       auto Result =
           FunctionProfileData.insert({F.Function, IndexedMemProfRecord()});
       IndexedMemProfRecord &Record = Result.first->second;
-      Record.AllocSites.emplace_back(Callstack, CSId, Entry.second);
+      Record.AllocSites.emplace_back(Callstack, CSId, MIB);
 
       if (!F.IsInlineFrame)
         break;
diff --git a/llvm/tools/llvm-profdata/llvm-profdata.cpp b/llvm/tools/llvm-profdata/llvm-profdata.cpp
index 6a70773613b7f..66a4cea035347 100644
--- a/llvm/tools/llvm-profdata/llvm-profdata.cpp
+++ b/llvm/tools/llvm-profdata/llvm-profdata.cpp
@@ -679,8 +679,8 @@ static void loadInput(const WeightedFile &Input, SymbolRemapper *Remapper,
     }
     const auto &FunctionProfileData = Reader->getProfileData();
     // Add the memprof records into the writer context.
-    for (const auto &I : FunctionProfileData) {
-      WC->Writer.addMemProfRecord(/*Id=*/I.first, /*Record=*/I.second);
+    for (const auto &[GUID, Record] : FunctionProfileData) {
+      WC->Writer.addMemProfRecord(GUID, Record);
     }
     return;
   }

From db9a17a4075d2ba4cf9edfa90018da6c11908e2a Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Thu, 11 Apr 2024 09:56:01 -0700
Subject: [PATCH 157/886] [memprof] Use std::optional (NFC) (#88366)

---
 llvm/lib/ProfileData/InstrProfReader.cpp     | 9 ++++-----
 llvm/unittests/ProfileData/InstrProfTest.cpp | 9 ++++-----
 2 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/llvm/lib/ProfileData/InstrProfReader.cpp b/llvm/lib/ProfileData/InstrProfReader.cpp
index 884334ed070e8..a35366a106a32 100644
--- a/llvm/lib/ProfileData/InstrProfReader.cpp
+++ b/llvm/lib/ProfileData/InstrProfReader.cpp
@@ -33,6 +33,7 @@
 #include <cstdint>
 #include <limits>
 #include <memory>
+#include <optional>
 #include <system_error>
 #include <utility>
 #include <vector>
@@ -1506,13 +1507,11 @@ IndexedInstrProfReader::getMemProfRecord(const uint64_t FuncNameHash) {
 
   // Setup a callback to convert from frame ids to frame using the on-disk
   // FrameData hash table.
-  memprof::FrameId LastUnmappedFrameId = 0;
-  bool HasFrameMappingError = false;
+  std::optional<memprof::FrameId> LastUnmappedFrameId;
   auto IdToFrameCallback = [&](const memprof::FrameId Id) {
     auto FrIter = MemProfFrameTable->find(Id);
     if (FrIter == MemProfFrameTable->end()) {
       LastUnmappedFrameId = Id;
-      HasFrameMappingError = true;
       return memprof::Frame(0, 0, 0, false);
     }
     return *FrIter;
@@ -1521,10 +1520,10 @@ IndexedInstrProfReader::getMemProfRecord(const uint64_t FuncNameHash) {
   memprof::MemProfRecord Record(*Iter, IdToFrameCallback);
 
   // Check that all frame ids were successfully converted to frames.
-  if (HasFrameMappingError) {
+  if (LastUnmappedFrameId) {
     return make_error<InstrProfError>(instrprof_error::hash_mismatch,
                                       "memprof frame not found for frame id " +
-                                          Twine(LastUnmappedFrameId));
+                                          Twine(*LastUnmappedFrameId));
   }
   return Record;
 }
diff --git a/llvm/unittests/ProfileData/InstrProfTest.cpp b/llvm/unittests/ProfileData/InstrProfTest.cpp
index 732f8fd792f8d..82701c47daf71 100644
--- a/llvm/unittests/ProfileData/InstrProfTest.cpp
+++ b/llvm/unittests/ProfileData/InstrProfTest.cpp
@@ -19,6 +19,7 @@
 #include "llvm/Testing/Support/Error.h"
 #include "gtest/gtest.h"
 #include <cstdarg>
+#include <optional>
 
 using namespace llvm;
 using ::testing::EndsWith;
@@ -433,21 +434,19 @@ TEST_F(InstrProfTest, test_memprof) {
   ASSERT_THAT_ERROR(RecordOr.takeError(), Succeeded());
   const memprof::MemProfRecord &Record = RecordOr.get();
 
-  memprof::FrameId LastUnmappedFrameId = 0;
-  bool HasFrameMappingError = false;
+  std::optional<memprof::FrameId> LastUnmappedFrameId;
   auto IdToFrameCallback = [&](const memprof::FrameId Id) {
     auto Iter = IdToFrameMap.find(Id);
     if (Iter == IdToFrameMap.end()) {
       LastUnmappedFrameId = Id;
-      HasFrameMappingError = true;
       return memprof::Frame(0, 0, 0, false);
     }
     return Iter->second;
   };
 
   const memprof::MemProfRecord WantRecord(IndexedMR, IdToFrameCallback);
-  ASSERT_FALSE(HasFrameMappingError)
-      << "could not map frame id: " << LastUnmappedFrameId;
+  ASSERT_FALSE(LastUnmappedFrameId.has_value())
+      << "could not map frame id: " << *LastUnmappedFrameId;
   EXPECT_THAT(WantRecord, EqualsRecord(Record));
 }
 

From 8c3cb6b55b688b767e5d65bcc2891b17322e8d05 Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Thu, 11 Apr 2024 09:56:22 -0700
Subject: [PATCH 158/886] =?UTF-8?q?Reland=20"[lldb][sbdebugger]=20Move=20S?=
 =?UTF-8?q?BDebugger=20Broadcast=20bit=20enum=20into=20ll=E2=80=A6=20(#883?=
 =?UTF-8?q?31)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…db-enumerations.h" (#88324)"

This reverts commit 9f6d08f2566a26144ea1753f80aebb1f2ecfdc63. This broke
the build because of a usage of one of the original SBDebugger broadcast
bits that wasn't updated in the original commit.
---
 lldb/include/lldb/API/SBDebugger.h                        | 7 -------
 lldb/include/lldb/lldb-enumerations.h                     | 8 ++++++++
 .../diagnostic_reporting/TestDiagnosticReporting.py       | 2 +-
 .../progress_reporting/TestProgressReporting.py           | 2 +-
 .../clang_modules/TestClangModuleBuildProgress.py         | 2 +-
 lldb/test/API/macosx/rosetta/TestRosetta.py               | 2 +-
 lldb/tools/lldb-dap/lldb-dap.cpp                          | 4 ++--
 7 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/lldb/include/lldb/API/SBDebugger.h b/lldb/include/lldb/API/SBDebugger.h
index 62b2f91f5076d..cf5409a12a056 100644
--- a/lldb/include/lldb/API/SBDebugger.h
+++ b/lldb/include/lldb/API/SBDebugger.h
@@ -42,13 +42,6 @@ class LLDB_API SBInputReader {
 
 class LLDB_API SBDebugger {
 public:
-  FLAGS_ANONYMOUS_ENUM(){
-      eBroadcastBitProgress = (1 << 0),
-      eBroadcastBitWarning = (1 << 1),
-      eBroadcastBitError = (1 << 2),
-      eBroadcastBitProgressCategory = (1 << 3),
-  };
-
   SBDebugger();
 
   SBDebugger(const lldb::SBDebugger &rhs);
diff --git a/lldb/include/lldb/lldb-enumerations.h b/lldb/include/lldb/lldb-enumerations.h
index 646f7bfda9847..f3b07ea6d2039 100644
--- a/lldb/include/lldb/lldb-enumerations.h
+++ b/lldb/include/lldb/lldb-enumerations.h
@@ -1339,6 +1339,14 @@ enum AddressMaskRange {
   eAddressMaskRangeAll = eAddressMaskRangeAny,
 };
 
+/// Used by the debugger to indicate which events are being broadcasted.
+enum DebuggerBroadcastBit {
+  eBroadcastBitProgress = (1 << 0),
+  eBroadcastBitWarning = (1 << 1),
+  eBroadcastBitError = (1 << 2),
+  eBroadcastBitProgressCategory = (1 << 3),
+};
+
 } // namespace lldb
 
 #endif // LLDB_LLDB_ENUMERATIONS_H
diff --git a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
index 36a3be695628f..6353e3e8cbedb 100644
--- a/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
+++ b/lldb/test/API/functionalities/diagnostic_reporting/TestDiagnosticReporting.py
@@ -15,7 +15,7 @@ def setUp(self):
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
             self.broadcaster,
-            lldb.SBDebugger.eBroadcastBitWarning | lldb.SBDebugger.eBroadcastBitError,
+            lldb.eBroadcastBitWarning | lldb.eBroadcastBitError,
         )
 
     def test_dwarf_symbol_loading_diagnostic_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
index 9af53845ca1b7..98988d7624da3 100644
--- a/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
+++ b/lldb/test/API/functionalities/progress_reporting/TestProgressReporting.py
@@ -13,7 +13,7 @@ def setUp(self):
         TestBase.setUp(self)
         self.broadcaster = self.dbg.GetBroadcaster()
         self.listener = lldbutil.start_listening_from(
-            self.broadcaster, lldb.SBDebugger.eBroadcastBitProgress
+            self.broadcaster, lldb.eBroadcastBitProgress
         )
 
     def test_dwarf_symbol_loading_progress_report(self):
diff --git a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
index 228f676aedf6a..33c7c269c081e 100644
--- a/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
+++ b/lldb/test/API/functionalities/progress_reporting/clang_modules/TestClangModuleBuildProgress.py
@@ -34,7 +34,7 @@ def test_clang_module_build_progress_report(self):
         # other unrelated progress events.
         broadcaster = self.dbg.GetBroadcaster()
         listener = lldbutil.start_listening_from(
-            broadcaster, lldb.SBDebugger.eBroadcastBitProgress
+            broadcaster, lldb.eBroadcastBitProgress
         )
 
         # Trigger module builds.
diff --git a/lldb/test/API/macosx/rosetta/TestRosetta.py b/lldb/test/API/macosx/rosetta/TestRosetta.py
index ce40de475ef16..669db95a1624c 100644
--- a/lldb/test/API/macosx/rosetta/TestRosetta.py
+++ b/lldb/test/API/macosx/rosetta/TestRosetta.py
@@ -49,7 +49,7 @@ def test_rosetta(self):
         if rosetta_debugserver_installed():
             broadcaster = self.dbg.GetBroadcaster()
             listener = lldbutil.start_listening_from(
-                broadcaster, lldb.SBDebugger.eBroadcastBitWarning
+                broadcaster, lldb.eBroadcastBitWarning
             )
 
             target, process, thread, bkpt = lldbutil.run_to_source_breakpoint(
diff --git a/lldb/tools/lldb-dap/lldb-dap.cpp b/lldb/tools/lldb-dap/lldb-dap.cpp
index 55f8c920e6001..25c5ad56e3d6f 100644
--- a/lldb/tools/lldb-dap/lldb-dap.cpp
+++ b/lldb/tools/lldb-dap/lldb-dap.cpp
@@ -420,8 +420,8 @@ void SendStdOutStdErr(lldb::SBProcess &process) {
 
 void ProgressEventThreadFunction() {
   lldb::SBListener listener("lldb-dap.progress.listener");
-  g_dap.debugger.GetBroadcaster().AddListener(
-      listener, lldb::SBDebugger::eBroadcastBitProgress);
+  g_dap.debugger.GetBroadcaster().AddListener(listener,
+                                              lldb::eBroadcastBitProgress);
   g_dap.broadcaster.AddListener(listener, eBroadcastBitStopProgressThread);
   lldb::SBEvent event;
   bool done = false;

From 2ea7ec9737e3ca4e2ce23bf606e79e7066beae0b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Thu, 11 Apr 2024 10:03:29 +0200
Subject: [PATCH 159/886] [clang][Interp][NFC] Expand pointer unittests

Test integral pointers as well.
---
 clang/unittests/AST/Interp/toAPValue.cpp | 56 +++++++++++++++++++-----
 1 file changed, 46 insertions(+), 10 deletions(-)

diff --git a/clang/unittests/AST/Interp/toAPValue.cpp b/clang/unittests/AST/Interp/toAPValue.cpp
index d0dfb40d51495..be7929228d283 100644
--- a/clang/unittests/AST/Interp/toAPValue.cpp
+++ b/clang/unittests/AST/Interp/toAPValue.cpp
@@ -20,7 +20,9 @@ TEST(ToAPValue, Pointers) {
       "  A a[3];\n"
       "};\n"
       "constexpr S d = {{{true, false}, {false, true}, {false, false}}};\n"
-      "constexpr const bool *b = &d.a[1].z;\n";
+      "constexpr const bool *b = &d.a[1].z;\n"
+      "const void *p = (void*)12;\n"
+      "const void *nullp = (void*)0;\n";
 
   auto AST = tooling::buildASTFromCodeWithArgs(
       Code, {"-fexperimental-new-constant-interpreter"});
@@ -41,15 +43,49 @@ TEST(ToAPValue, Pointers) {
     return Prog.getPtrGlobal(*Prog.getGlobal(D));
   };
 
-  const Pointer &GP = getGlobalPtr("b");
-  const Pointer &P = GP.deref<Pointer>();
-  ASSERT_TRUE(P.isLive());
-  APValue A = P.toAPValue();
-  ASSERT_TRUE(A.isLValue());
-  ASSERT_TRUE(A.hasLValuePath());
-  const auto &Path = A.getLValuePath();
-  ASSERT_EQ(Path.size(), 3u);
-  ASSERT_EQ(A.getLValueBase(), getDecl("d"));
+  {
+    const Pointer &GP = getGlobalPtr("b");
+    const Pointer &P = GP.deref<Pointer>();
+    ASSERT_TRUE(P.isLive());
+    APValue A = P.toAPValue();
+    ASSERT_TRUE(A.isLValue());
+    ASSERT_TRUE(A.hasLValuePath());
+    const auto &Path = A.getLValuePath();
+    ASSERT_EQ(Path.size(), 3u);
+    ASSERT_EQ(A.getLValueBase(), getDecl("d"));
+    // FIXME: Also test all path elements.
+  }
+
+  {
+    const ValueDecl *D = getDecl("p");
+    ASSERT_NE(D, nullptr);
+    const Pointer &GP = getGlobalPtr("p");
+    const Pointer &P = GP.deref<Pointer>();
+    ASSERT_TRUE(P.isIntegralPointer());
+    APValue A = P.toAPValue();
+    ASSERT_TRUE(A.isLValue());
+    ASSERT_TRUE(A.getLValueBase().isNull());
+    APSInt I;
+    bool Success = A.toIntegralConstant(I, D->getType(), AST->getASTContext());
+    ASSERT_TRUE(Success);
+    ASSERT_EQ(I, 12);
+  }
+
+  {
+    const ValueDecl *D = getDecl("nullp");
+    ASSERT_NE(D, nullptr);
+    const Pointer &GP = getGlobalPtr("nullp");
+    const Pointer &P = GP.deref<Pointer>();
+    ASSERT_TRUE(P.isIntegralPointer());
+    APValue A = P.toAPValue();
+    ASSERT_TRUE(A.isLValue());
+    ASSERT_TRUE(A.getLValueBase().isNull());
+    ASSERT_TRUE(A.isNullPointer());
+    APSInt I;
+    bool Success = A.toIntegralConstant(I, D->getType(), AST->getASTContext());
+    ASSERT_TRUE(Success);
+    ASSERT_EQ(I, 0);
+  }
 }
 
 TEST(ToAPValue, FunctionPointers) {

From 64c3997939cf2d9b4fd1c24c89724d0b47afcd03 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Thu, 11 Apr 2024 17:53:57 +0200
Subject: [PATCH 160/886] [clang][Interp] Allow initializing static class
 members

We need to handle this when registering global variables.
---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 34 ++++----
 clang/lib/AST/Interp/Interp.cpp          | 99 +++++++++++++++---------
 clang/lib/AST/Interp/Program.cpp         |  2 +-
 clang/test/AST/Interp/cxx23.cpp          | 17 ++--
 clang/test/AST/Interp/records.cpp        | 32 +++++++-
 5 files changed, 118 insertions(+), 66 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 84bacd457c85b..01ec31e4077f7 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -2778,26 +2778,34 @@ bool ByteCodeExprGen<Emitter>::visitVarDecl(const VarDecl *VD) {
   std::optional<PrimType> VarT = classify(VD->getType());
 
   if (Context::shouldBeGloballyIndexed(VD)) {
-    // We've already seen and initialized this global.
-    if (P.getGlobal(VD))
-      return true;
-
-    std::optional<unsigned> GlobalIndex = P.createGlobal(VD, Init);
-
-    if (!GlobalIndex)
-      return false;
-
-    if (Init) {
+    auto initGlobal = [&](unsigned GlobalIndex) -> bool {
+      assert(Init);
       DeclScope<Emitter> LocalScope(this, VD);
 
       if (VarT) {
         if (!this->visit(Init))
           return false;
-        return this->emitInitGlobal(*VarT, *GlobalIndex, VD);
+        return this->emitInitGlobal(*VarT, GlobalIndex, VD);
       }
-      return this->visitGlobalInitializer(Init, *GlobalIndex);
+      return this->visitGlobalInitializer(Init, GlobalIndex);
+    };
+
+    // We've already seen and initialized this global.
+    if (std::optional<unsigned> GlobalIndex = P.getGlobal(VD)) {
+      if (P.getPtrGlobal(*GlobalIndex).isInitialized())
+        return true;
+
+      // The previous attempt at initialization might've been unsuccessful,
+      // so let's try this one.
+      return Init && initGlobal(*GlobalIndex);
     }
-    return true;
+
+    std::optional<unsigned> GlobalIndex = P.createGlobal(VD, Init);
+
+    if (!GlobalIndex)
+      return false;
+
+    return !Init || initGlobal(*GlobalIndex);
   } else {
     VariableScope<Emitter> LocalScope(this);
     if (VarT) {
diff --git a/clang/lib/AST/Interp/Interp.cpp b/clang/lib/AST/Interp/Interp.cpp
index e5e2c932f500b..2607e07432516 100644
--- a/clang/lib/AST/Interp/Interp.cpp
+++ b/clang/lib/AST/Interp/Interp.cpp
@@ -56,22 +56,65 @@ static bool Jf(InterpState &S, CodePtr &PC, int32_t Offset) {
   return true;
 }
 
+static void diagnoseMissingInitializer(InterpState &S, CodePtr OpPC,
+                                       const ValueDecl *VD) {
+  const SourceInfo &E = S.Current->getSource(OpPC);
+  S.FFDiag(E, diag::note_constexpr_var_init_unknown, 1) << VD;
+  S.Note(VD->getLocation(), diag::note_declared_at) << VD->getSourceRange();
+}
+
+static void diagnoseNonConstVariable(InterpState &S, CodePtr OpPC,
+                                     const ValueDecl *VD);
+static bool diagnoseUnknownDecl(InterpState &S, CodePtr OpPC,
+                                const ValueDecl *D) {
+  const SourceInfo &E = S.Current->getSource(OpPC);
+
+  if (isa<ParmVarDecl>(D)) {
+    if (S.getLangOpts().CPlusPlus11) {
+      S.FFDiag(E, diag::note_constexpr_function_param_value_unknown) << D;
+      S.Note(D->getLocation(), diag::note_declared_at) << D->getSourceRange();
+    } else {
+      S.FFDiag(E);
+    }
+  } else if (const auto *VD = dyn_cast<VarDecl>(D)) {
+    if (!VD->getType().isConstQualified()) {
+      diagnoseNonConstVariable(S, OpPC, VD);
+      return false;
+    }
+
+    // const, but no initializer.
+    if (!VD->getAnyInitializer()) {
+      diagnoseMissingInitializer(S, OpPC, VD);
+      return false;
+    }
+  }
+  return false;
+}
+
 static void diagnoseNonConstVariable(InterpState &S, CodePtr OpPC,
                                      const ValueDecl *VD) {
   if (!S.getLangOpts().CPlusPlus)
     return;
 
   const SourceInfo &Loc = S.Current->getSource(OpPC);
+  if (const auto *VarD = dyn_cast<VarDecl>(VD);
+      VarD && VarD->getType().isConstQualified() &&
+      !VarD->getAnyInitializer()) {
+    diagnoseMissingInitializer(S, OpPC, VD);
+    return;
+  }
 
-  if (VD->getType()->isIntegralOrEnumerationType())
+  if (VD->getType()->isIntegralOrEnumerationType()) {
     S.FFDiag(Loc, diag::note_constexpr_ltor_non_const_int, 1) << VD;
-  else
-    S.FFDiag(Loc,
-             S.getLangOpts().CPlusPlus11
-                 ? diag::note_constexpr_ltor_non_constexpr
-                 : diag::note_constexpr_ltor_non_integral,
-             1)
-        << VD << VD->getType();
+    S.Note(VD->getLocation(), diag::note_declared_at);
+    return;
+  }
+
+  S.FFDiag(Loc,
+           S.getLangOpts().CPlusPlus11 ? diag::note_constexpr_ltor_non_constexpr
+                                       : diag::note_constexpr_ltor_non_integral,
+           1)
+      << VD << VD->getType();
   S.Note(VD->getLocation(), diag::note_declared_at);
 }
 
@@ -202,6 +245,9 @@ bool CheckExtern(InterpState &S, CodePtr OpPC, const Pointer &Ptr) {
   if (!Ptr.isExtern())
     return true;
 
+  if (Ptr.isInitialized())
+    return true;
+
   if (!S.checkingPotentialConstantExpression() && S.getLangOpts().CPlusPlus) {
     const auto *VD = Ptr.getDeclDesc()->asValueDecl();
     diagnoseNonConstVariable(S, OpPC, VD);
@@ -369,9 +415,15 @@ bool CheckInitialized(InterpState &S, CodePtr OpPC, const Pointer &Ptr,
   if (const auto *VD = Ptr.getDeclDesc()->asVarDecl();
       VD && VD->hasGlobalStorage()) {
     const SourceInfo &Loc = S.Current->getSource(OpPC);
-    S.FFDiag(Loc, diag::note_constexpr_var_init_non_constant, 1) << VD;
-    S.Note(VD->getLocation(), diag::note_declared_at);
+    if (VD->getAnyInitializer()) {
+      S.FFDiag(Loc, diag::note_constexpr_var_init_non_constant, 1) << VD;
+      S.Note(VD->getLocation(), diag::note_declared_at);
+    } else {
+      diagnoseMissingInitializer(S, OpPC, VD);
+    }
+    return false;
   }
+
   if (!S.checkingPotentialConstantExpression()) {
     S.FFDiag(S.Current->getSource(OpPC), diag::note_constexpr_access_uninit)
         << AK << /*uninitialized=*/true << S.Current->getRange(OpPC);
@@ -598,33 +650,6 @@ bool CheckFloatResult(InterpState &S, CodePtr OpPC, const Floating &Result,
   return true;
 }
 
-static bool diagnoseUnknownDecl(InterpState &S, CodePtr OpPC,
-                                const ValueDecl *D) {
-  const SourceInfo &E = S.Current->getSource(OpPC);
-
-  if (isa<ParmVarDecl>(D)) {
-    if (S.getLangOpts().CPlusPlus11) {
-      S.FFDiag(E, diag::note_constexpr_function_param_value_unknown) << D;
-      S.Note(D->getLocation(), diag::note_declared_at) << D->getSourceRange();
-    } else {
-      S.FFDiag(E);
-    }
-  } else if (const auto *VD = dyn_cast<VarDecl>(D)) {
-    if (!VD->getType().isConstQualified()) {
-      diagnoseNonConstVariable(S, OpPC, VD);
-      return false;
-    }
-
-    // const, but no initializer.
-    if (!VD->getAnyInitializer()) {
-      S.FFDiag(E, diag::note_constexpr_var_init_unknown, 1) << VD;
-      S.Note(VD->getLocation(), diag::note_declared_at) << VD->getSourceRange();
-      return false;
-    }
-  }
-  return false;
-}
-
 /// We aleady know the given DeclRefExpr is invalid for some reason,
 /// now figure out why and print appropriate diagnostics.
 bool CheckDeclRef(InterpState &S, CodePtr OpPC, const DeclRefExpr *DR) {
diff --git a/clang/lib/AST/Interp/Program.cpp b/clang/lib/AST/Interp/Program.cpp
index 82367164743fc..e6f22e79451e9 100644
--- a/clang/lib/AST/Interp/Program.cpp
+++ b/clang/lib/AST/Interp/Program.cpp
@@ -177,7 +177,7 @@ std::optional<unsigned> Program::createGlobal(const ValueDecl *VD,
   bool IsStatic, IsExtern;
   if (const auto *Var = dyn_cast<VarDecl>(VD)) {
     IsStatic = Context::shouldBeGloballyIndexed(VD);
-    IsExtern = !Var->getAnyInitializer();
+    IsExtern = Var->hasExternalStorage();
   } else if (isa<UnnamedGlobalConstantDecl, MSGuidDecl>(VD)) {
     IsStatic = true;
     IsExtern = false;
diff --git a/clang/test/AST/Interp/cxx23.cpp b/clang/test/AST/Interp/cxx23.cpp
index 042e29613aa75..f0325eef6d87c 100644
--- a/clang/test/AST/Interp/cxx23.cpp
+++ b/clang/test/AST/Interp/cxx23.cpp
@@ -5,23 +5,18 @@
 
 /// FIXME: The new interpreter is missing all the 'control flows through...' diagnostics.
 
-constexpr int f(int n) {  // ref20-error {{constexpr function never produces a constant expression}} \
-                          // expected20-error {{constexpr function never produces a constant expression}}
+constexpr int f(int n) {  // ref20-error {{constexpr function never produces a constant expression}}
   static const int m = n; // ref20-note {{control flows through the definition of a static variable}} \
                           // ref20-warning {{is a C++23 extension}} \
-                          // expected20-warning {{is a C++23 extension}} \
-                          // expected20-note {{declared here}} \
+                          // expected20-warning {{is a C++23 extension}}
 
-  return m; // expected20-note {{initializer of 'm' is not a constant expression}}
+  return m;
 }
-constexpr int g(int n) {        // ref20-error {{constexpr function never produces a constant expression}} \
-                                // expected20-error {{constexpr function never produces a constant expression}}
+constexpr int g(int n) {        // ref20-error {{constexpr function never produces a constant expression}}
   thread_local const int m = n; // ref20-note {{control flows through the definition of a thread_local variable}} \
                                 // ref20-warning {{is a C++23 extension}} \
-                                // expected20-warning {{is a C++23 extension}} \
-                                // expected20-note {{declared here}}
-  return m; // expected20-note {{initializer of 'm' is not a constant expression}}
-
+                                // expected20-warning {{is a C++23 extension}}
+  return m;
 }
 
 constexpr int c_thread_local(int n) { // ref20-error {{constexpr function never produces a constant expression}} \
diff --git a/clang/test/AST/Interp/records.cpp b/clang/test/AST/Interp/records.cpp
index 0f76e0cfe9927..f251497ed7018 100644
--- a/clang/test/AST/Interp/records.cpp
+++ b/clang/test/AST/Interp/records.cpp
@@ -1,11 +1,11 @@
-// RUN: %clang_cc1 -fexperimental-new-constant-interpreter -verify=expected,both %s
 // RUN: %clang_cc1 -fexperimental-new-constant-interpreter -std=c++14 -verify=expected,both %s
+// RUN: %clang_cc1 -fexperimental-new-constant-interpreter -std=c++17 -verify=expected,both %s
+// RUN: %clang_cc1 -fexperimental-new-constant-interpreter -std=c++17 -triple i686 -verify=expected,both %s
 // RUN: %clang_cc1 -fexperimental-new-constant-interpreter -std=c++20 -verify=expected,both %s
-// RUN: %clang_cc1 -fexperimental-new-constant-interpreter -triple i686 -verify=expected,both %s
-// RUN: %clang_cc1 -verify=ref,both %s
 // RUN: %clang_cc1 -verify=ref,both -std=c++14 %s
+// RUN: %clang_cc1 -verify=ref,both -std=c++17 %s
+// RUN: %clang_cc1 -verify=ref,both -std=c++17 -triple i686 %s
 // RUN: %clang_cc1 -verify=ref,both -std=c++20 %s
-// RUN: %clang_cc1 -verify=ref,both -triple i686 %s
 
 /// Used to crash.
 struct Empty {};
@@ -1285,3 +1285,27 @@ namespace {
   }
 }
 #endif
+
+namespace pr18633 {
+  struct A1 {
+    static const int sz;
+    static const int sz2;
+  };
+  const int A1::sz2 = 11;
+  template<typename T>
+  void func () {
+    int arr[A1::sz];
+    // both-warning@-1 {{variable length arrays in C++ are a Clang extension}}
+    // both-note@-2 {{initializer of 'sz' is unknown}}
+    // both-note@-9 {{declared here}}
+  }
+  template<typename T>
+  void func2 () {
+    int arr[A1::sz2];
+  }
+  const int A1::sz = 12;
+  void func2() {
+    func<int>();
+    func2<int>();
+  }
+}

From 5122a2c2320c7b14f6585e63b7fc43ac82a550c2 Mon Sep 17 00:00:00 2001
From: Aart Bik <ajcbik@google.com>
Date: Thu, 11 Apr 2024 10:07:24 -0700
Subject: [PATCH 161/886] [mlir][sparse] allow for direct-out passing of sparse
 tensor buffers (#88327)

In order to support various external frameworks (JAX vs PyTorch) we need
a bit more flexibility in [dis]assembling external buffers to and from
sparse tensors in MLIR land. This PR adds a direct-out option that
avoids the rigid pre-allocated for copy-out semantics.

Note that over time, we expect the [dis]assemble operations to converge
into something that supports all sorts of external frameworks. Until
then, this option helps in experimenting with different options.
---
 .../Dialect/SparseTensor/Transforms/Passes.h  |  3 +-
 .../Dialect/SparseTensor/Transforms/Passes.td |  9 ++
 .../Transforms/SparseAssembler.cpp            | 87 ++++++++++++-------
 .../Transforms/SparseTensorConversion.cpp     |  9 +-
 .../Transforms/SparseTensorPasses.cpp         |  3 +-
 .../Dialect/SparseTensor/external_direct.mlir | 52 +++++++++++
 6 files changed, 125 insertions(+), 38 deletions(-)
 create mode 100644 mlir/test/Dialect/SparseTensor/external_direct.mlir

diff --git a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.h b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.h
index 61b07d222d156..d6d038ef65bdf 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.h
+++ b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.h
@@ -60,9 +60,10 @@ enum class SparseEmitStrategy {
 // The SparseAssembler pass.
 //===----------------------------------------------------------------------===//
 
-void populateSparseAssembler(RewritePatternSet &patterns);
+void populateSparseAssembler(RewritePatternSet &patterns, bool directOut);
 
 std::unique_ptr<Pass> createSparseAssembler();
+std::unique_ptr<Pass> createSparseAssembler(bool directOut);
 
 //===----------------------------------------------------------------------===//
 // The SparseReinterpretMap pass.
diff --git a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
index 58e2d6f32386c..4706d5ba2f218 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
@@ -23,12 +23,21 @@ def SparseAssembler : Pass<"sparse-assembler", "ModuleOp"> {
     sparse tensors as numpy arrays from and to Python. Note that eventual
     bufferization decisions (e.g. who [de]allocates the underlying memory)
     should be resolved in agreement with the external runtime.
+
+    By default, the pass uses the [dis]assemble operations to input and output
+    sparse tensors. When the direct-out option is set, however, the output
+    directly returns the MLIR allocated buffers to the external runtime.
   }];
   let constructor = "mlir::createSparseAssembler()";
   let dependentDialects = [
+    "bufferization::BufferizationDialect",
     "sparse_tensor::SparseTensorDialect",
     "tensor::TensorDialect",
   ];
+  let options = [
+    Option<"directOut", "direct-out", "bool",
+      "false", "Directly returns buffers externally">,
+  ];
 }
 
 def SparseReinterpretMap : Pass<"sparse-reinterpret-map", "ModuleOp"> {
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseAssembler.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseAssembler.cpp
index a91d32a23cac9..eafbe95b7aebe 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseAssembler.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseAssembler.cpp
@@ -8,6 +8,7 @@
 
 #include "Utils/CodegenUtils.h"
 
+#include "mlir/Dialect/Bufferization/IR/Bufferization.h"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensor.h"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorStorageLayout.h"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorType.h"
@@ -24,7 +25,7 @@ using namespace sparse_tensor;
 
 // Convert type range to new types range, with sparse tensors externalized.
 static void convTypes(TypeRange types, SmallVectorImpl<Type> &convTypes,
-                      SmallVectorImpl<Type> *extraTypes = nullptr) {
+                      SmallVectorImpl<Type> *extraTypes, bool directOut) {
   for (auto type : types) {
     // All "dense" data passes through unmodified.
     if (!getSparseTensorEncoding(type)) {
@@ -32,31 +33,33 @@ static void convTypes(TypeRange types, SmallVectorImpl<Type> &convTypes,
       continue;
     }
 
-    // Convert the external representation of the position/coordinate array
+    // Convert the external representations of the pos/crd/val arrays.
     const SparseTensorType stt(cast<RankedTensorType>(type));
-    foreachFieldAndTypeInSparseTensor(stt, [&convTypes, extraTypes](
-                                               Type t, FieldIndex,
-                                               SparseTensorFieldKind kind,
-                                               Level, LevelType) {
-      if (kind == SparseTensorFieldKind::CrdMemRef ||
-          kind == SparseTensorFieldKind::PosMemRef ||
-          kind == SparseTensorFieldKind::ValMemRef) {
-        ShapedType st = t.cast<ShapedType>();
-        auto rtp = RankedTensorType::get(st.getShape(), st.getElementType());
-        convTypes.push_back(rtp);
-        if (extraTypes)
-          extraTypes->push_back(rtp);
-      }
-      return true;
-    });
+    foreachFieldAndTypeInSparseTensor(
+        stt, [&convTypes, extraTypes, directOut](Type t, FieldIndex,
+                                                 SparseTensorFieldKind kind,
+                                                 Level, LevelType) {
+          if (kind == SparseTensorFieldKind::PosMemRef ||
+              kind == SparseTensorFieldKind::CrdMemRef ||
+              kind == SparseTensorFieldKind::ValMemRef) {
+            auto rtp = t.cast<ShapedType>();
+            if (!directOut) {
+              rtp = RankedTensorType::get(rtp.getShape(), rtp.getElementType());
+              if (extraTypes)
+                extraTypes->push_back(rtp);
+            }
+            convTypes.push_back(rtp);
+          }
+          return true;
+        });
   }
 }
 
 // Convert input and output values to [dis]assemble ops for sparse tensors.
 static void convVals(OpBuilder &builder, Location loc, TypeRange types,
                      ValueRange fromVals, ValueRange extraVals,
-                     SmallVectorImpl<Value> &toVals, unsigned extra,
-                     bool isIn) {
+                     SmallVectorImpl<Value> &toVals, unsigned extra, bool isIn,
+                     bool directOut) {
   unsigned idx = 0;
   for (auto type : types) {
     // All "dense" data passes through unmodified.
@@ -73,18 +76,29 @@ static void convVals(OpBuilder &builder, Location loc, TypeRange types,
     if (!isIn)
       inputs.push_back(fromVals[idx++]); // The sparse tensor to disassemble
 
-    // Collect the external representations of the pos/crd arrays.
+    // Collect the external representations of the pos/crd/val arrays.
     foreachFieldAndTypeInSparseTensor(stt, [&, isIn](Type t, FieldIndex,
                                                      SparseTensorFieldKind kind,
-                                                     Level, LevelType) {
-      if (kind == SparseTensorFieldKind::CrdMemRef ||
-          kind == SparseTensorFieldKind::PosMemRef ||
+                                                     Level lv, LevelType) {
+      if (kind == SparseTensorFieldKind::PosMemRef ||
+          kind == SparseTensorFieldKind::CrdMemRef ||
           kind == SparseTensorFieldKind::ValMemRef) {
         if (isIn) {
           inputs.push_back(fromVals[idx++]);
+        } else if (directOut) {
+          Value mem;
+          if (kind == SparseTensorFieldKind::PosMemRef)
+            mem = builder.create<sparse_tensor::ToPositionsOp>(loc, inputs[0],
+                                                               lv);
+          else if (kind == SparseTensorFieldKind::CrdMemRef)
+            mem = builder.create<sparse_tensor::ToCoordinatesOp>(loc, inputs[0],
+                                                                 lv);
+          else
+            mem = builder.create<sparse_tensor::ToValuesOp>(loc, inputs[0]);
+          toVals.push_back(mem);
         } else {
-          ShapedType st = t.cast<ShapedType>();
-          auto rtp = RankedTensorType::get(st.getShape(), st.getElementType());
+          ShapedType rtp = t.cast<ShapedType>();
+          rtp = RankedTensorType::get(rtp.getShape(), rtp.getElementType());
           inputs.push_back(extraVals[extra++]);
           retTypes.push_back(rtp);
           cntTypes.push_back(builder.getIndexType());
@@ -97,7 +111,7 @@ static void convVals(OpBuilder &builder, Location loc, TypeRange types,
       // Assemble multiple inputs into a single sparse tensor.
       auto a = builder.create<sparse_tensor::AssembleOp>(loc, rtp, inputs);
       toVals.push_back(a.getResult());
-    } else {
+    } else if (!directOut) {
       // Disassemble a single sparse input into multiple outputs.
       // Note that this includes the counters, which are dropped.
       unsigned len = retTypes.size();
@@ -144,11 +158,14 @@ namespace {
 //   return ..., t1..tn, ...
 // }
 //
-// TODO: refine output sparse tensors to work well with external framework
+// (with a direct-out variant without the disassemble).
 //
 struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
   using OpRewritePattern::OpRewritePattern;
 
+  SparseFuncAssembler(MLIRContext *context, bool dO)
+      : OpRewritePattern(context), directOut(dO) {}
+
   LogicalResult matchAndRewrite(func::FuncOp funcOp,
                                 PatternRewriter &rewriter) const override {
     // Only rewrite public entry methods.
@@ -159,8 +176,8 @@ struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
     SmallVector<Type> inputTypes;
     SmallVector<Type> outputTypes;
     SmallVector<Type> extraTypes;
-    convTypes(funcOp.getArgumentTypes(), inputTypes);
-    convTypes(funcOp.getResultTypes(), outputTypes, &extraTypes);
+    convTypes(funcOp.getArgumentTypes(), inputTypes, nullptr, false);
+    convTypes(funcOp.getResultTypes(), outputTypes, &extraTypes, directOut);
 
     // Only sparse inputs or outputs need a wrapper method.
     if (inputTypes.size() == funcOp.getArgumentTypes().size() &&
@@ -192,7 +209,7 @@ struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
     // Convert inputs.
     SmallVector<Value> inputs;
     convVals(rewriter, loc, funcOp.getArgumentTypes(), body->getArguments(),
-             ValueRange(), inputs, 0, /*isIn=*/true);
+             ValueRange(), inputs, /*extra=*/0, /*isIn=*/true, directOut);
 
     // Call the original, now private method. A subsequent inlining pass can
     // determine whether cloning the method body in place is worthwhile.
@@ -203,7 +220,7 @@ struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
     // Convert outputs and return.
     SmallVector<Value> outputs;
     convVals(rewriter, loc, funcOp.getResultTypes(), call.getResults(),
-             body->getArguments(), outputs, extra, /*isIn=*/false);
+             body->getArguments(), outputs, extra, /*isIn=*/false, directOut);
     rewriter.create<func::ReturnOp>(loc, outputs);
 
     // Finally, migrate a potential c-interface property.
@@ -215,6 +232,9 @@ struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
     }
     return success();
   }
+
+private:
+  const bool directOut;
 };
 
 } // namespace
@@ -223,6 +243,7 @@ struct SparseFuncAssembler : public OpRewritePattern<func::FuncOp> {
 // Public method for populating conversion rules.
 //===----------------------------------------------------------------------===//
 
-void mlir::populateSparseAssembler(RewritePatternSet &patterns) {
-  patterns.add<SparseFuncAssembler>(patterns.getContext());
+void mlir::populateSparseAssembler(RewritePatternSet &patterns,
+                                   bool directOut) {
+  patterns.add<SparseFuncAssembler>(patterns.getContext(), directOut);
 }
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp
index c52fa3751e6b4..f0d162bdb84d9 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorConversion.cpp
@@ -767,6 +767,12 @@ class SparseTensorAssembleConverter : public OpConversionPattern<AssembleOp> {
 };
 
 /// Sparse conversion rule for the sparse_tensor.disassemble operator.
+/// Note that the current implementation simply exposes the buffers to
+/// the external client. This assumes the client only reads the buffers
+/// (usually copying it to the external data structures, such as numpy
+/// arrays). The semantics of the disassemble operation technically
+/// require that the copying is done here already using the out-levels
+/// and out-values clause.
 class SparseTensorDisassembleConverter
     : public OpConversionPattern<DisassembleOp> {
 public:
@@ -774,9 +780,6 @@ class SparseTensorDisassembleConverter
   LogicalResult
   matchAndRewrite(DisassembleOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    // We simply expose the buffers to the external client. This
-    // assumes the client only reads the buffers (usually copying it
-    // to the external data structures, such as numpy arrays).
     Location loc = op->getLoc();
     auto stt = getSparseTensorType(op.getTensor());
     SmallVector<Value> retVal;
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorPasses.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorPasses.cpp
index acea25f023980..b42d58634a36c 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorPasses.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorPasses.cpp
@@ -50,11 +50,12 @@ namespace {
 struct SparseAssembler : public impl::SparseAssemblerBase<SparseAssembler> {
   SparseAssembler() = default;
   SparseAssembler(const SparseAssembler &pass) = default;
+  SparseAssembler(bool dO) { directOut = dO; }
 
   void runOnOperation() override {
     auto *ctx = &getContext();
     RewritePatternSet patterns(ctx);
-    populateSparseAssembler(patterns);
+    populateSparseAssembler(patterns, directOut);
     (void)applyPatternsAndFoldGreedily(getOperation(), std::move(patterns));
   }
 };
diff --git a/mlir/test/Dialect/SparseTensor/external_direct.mlir b/mlir/test/Dialect/SparseTensor/external_direct.mlir
new file mode 100644
index 0000000000000..78c4a295686b3
--- /dev/null
+++ b/mlir/test/Dialect/SparseTensor/external_direct.mlir
@@ -0,0 +1,52 @@
+// RUN: mlir-opt %s --sparse-assembler="direct-out=True" -split-input-file | FileCheck %s
+
+// -----
+
+// CHECK-LABEL: func.func @sparse_in(
+// CHECK-SAME:    %[[B:.*0]]: tensor<?xindex>,
+// CHECK-SAME:    %[[C:.*1]]: tensor<?xindex>,
+// CHECK-SAME:    %[[A:.*]]: tensor<?xf32>) -> tensor<64x64xf32> {
+// CHECK:         %[[I:.*]] = sparse_tensor.assemble (%[[B]], %[[C]]), %[[A]]
+// CHECK:         %[[F:.*]] = call @_internal_sparse_in(%[[I]])
+// CHECK:         return %[[F]] : tensor<64x64xf32>
+// CHECK:       }
+// CHECK:       func.func private @_internal_sparse_in
+#sparse = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : dense, d1 : compressed) }>
+func.func @sparse_in(%arg0: tensor<64x64xf32, #sparse>) -> tensor<64x64xf32> {
+  %0 = sparse_tensor.convert %arg0 : tensor<64x64xf32, #sparse> to tensor<64x64xf32>
+  return %0 : tensor<64x64xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func.func @sparse_out(
+// CHECK-SAME:    %[[X:.*0]]: tensor<64x64xf32>)
+// CHECK:         %[[F:.*]] = call @_internal_sparse_out(%[[X]])
+// CHECK:         %[[P:.*]] = sparse_tensor.positions %[[F]]
+// CHECK:         %[[C:.*]] = sparse_tensor.coordinates %[[F]]
+// CHECK:         %[[V:.*]] = sparse_tensor.values %[[F]]
+// CHECK:         return %[[P]], %[[C]], %[[V]]
+// CHECK:       }
+// CHECK:       func.func private @_internal_sparse_out
+#sparse = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : dense, d1 : compressed) }>
+func.func @sparse_out(%arg0: tensor<64x64xf32>) -> tensor<64x64xf32, #sparse> {
+  %0 = sparse_tensor.convert %arg0 : tensor<64x64xf32> to tensor<64x64xf32, #sparse>
+  return %0 : tensor<64x64xf32, #sparse>
+}
+
+// -----
+
+// CHECK-LABEL: func.func @sparse_out2(
+// CHECK-SAME:    %[[X:.*0]]: tensor<64x64xf32>)
+// CHECK:         %[[F:.*]]:2 = call @_internal_sparse_out2(%[[X]])
+// CHECK:         %[[P:.*]] = sparse_tensor.positions %[[F]]#1
+// CHECK:         %[[C:.*]] = sparse_tensor.coordinates %[[F]]#1
+// CHECK:         %[[V:.*]] = sparse_tensor.values %[[F]]#1
+// CHECK:         return %[[F]]#0, %[[P]], %[[C]], %[[V]]
+// CHECK:       }
+// CHECK:       func.func private @_internal_sparse_out2
+#sparse = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : dense, d1 : compressed) }>
+func.func @sparse_out2(%arg0: tensor<64x64xf32>) -> (tensor<64x64xf32>, tensor<64x64xf32, #sparse>) {
+  %0 = sparse_tensor.convert %arg0 : tensor<64x64xf32> to tensor<64x64xf32, #sparse>
+  return %arg0, %0 : tensor<64x64xf32>, tensor<64x64xf32, #sparse>
+}

From f626a35086d90f25986e3f06e01a54cca91250d8 Mon Sep 17 00:00:00 2001
From: Nick Desaulniers <nickdesaulniers@users.noreply.github.com>
Date: Thu, 11 Apr 2024 10:11:58 -0700
Subject: [PATCH 162/886] [libc] Codify header inclusion policy (#87017)

When supporting "overlay" vs "fullbuild" modes, "what ABI are you
using?" becomes a fundamental question to have concrete answers for.
Overlay mode MUST match the ABI of the system being overlayed onto;
fullbuild more flexible (the only system ABI relevant is the OS kernel).

When implementing llvm-libc we generally prefer the include-what-you use
style of avoiding transitive dependencies (since that makes refactoring
headers more painful, and slows down build times). So what header do you
include for any given type or function declaration? For any given
userspace program, the answer is straightforward. But for llvm-libc
which is trying to support multiple ABIs (at least one per
configuration), the answer is perhaps less clear.

This proposal seeks to add one layer of indirection relative to what's
being done today.

It then converts users of sigset_t and struct epoll_event and the epoll
implemenations over to this convention as an example.
---
 libc/docs/dev/code_style.rst                  | 33 +++++++++++++++++++
 libc/docs/usage_modes.rst                     |  6 +++-
 libc/hdr/CMakeLists.txt                       |  2 ++
 libc/hdr/types/CMakeLists.txt                 | 23 +++++++++++++
 libc/hdr/types/sigset_t.h                     | 21 ++++++++++++
 libc/hdr/types/struct_epoll_event.h           | 21 ++++++++++++
 libc/hdr/types/struct_timespec.h              | 21 ++++++++++++
 libc/src/signal/linux/CMakeLists.txt          | 16 +++++----
 libc/src/signal/linux/raise.cpp               |  5 +--
 libc/src/signal/linux/sigaction.cpp           |  7 ++--
 libc/src/signal/linux/sigaddset.cpp           |  4 +--
 libc/src/signal/linux/sigdelset.cpp           |  4 +--
 libc/src/signal/linux/sigfillset.cpp          |  4 +--
 libc/src/signal/linux/signal_utils.h          |  3 +-
 libc/src/signal/linux/sigprocmask.cpp         |  6 ++--
 libc/src/signal/sigaddset.h                   |  2 +-
 libc/src/signal/sigdelset.h                   |  2 +-
 libc/src/signal/sigemptyset.h                 |  2 +-
 libc/src/signal/sigfillset.h                  |  2 +-
 libc/src/signal/sigprocmask.h                 |  2 +-
 libc/src/sys/epoll/epoll_pwait.h              |  7 ++--
 libc/src/sys/epoll/epoll_pwait2.h             |  9 ++---
 libc/src/sys/epoll/epoll_wait.h               |  5 +--
 libc/src/sys/epoll/linux/CMakeLists.txt       | 14 +++++---
 libc/src/sys/epoll/linux/epoll_pwait.cpp      | 10 ++----
 libc/src/sys/epoll/linux/epoll_pwait2.cpp     | 12 +++----
 libc/src/sys/epoll/linux/epoll_wait.cpp       |  9 ++---
 libc/src/sys/select/linux/select.cpp          |  8 ++---
 .../llvm-project-overlay/libc/BUILD.bazel     | 13 ++++++++
 29 files changed, 200 insertions(+), 73 deletions(-)
 create mode 100644 libc/hdr/types/CMakeLists.txt
 create mode 100644 libc/hdr/types/sigset_t.h
 create mode 100644 libc/hdr/types/struct_epoll_event.h
 create mode 100644 libc/hdr/types/struct_timespec.h

diff --git a/libc/docs/dev/code_style.rst b/libc/docs/dev/code_style.rst
index 22a18b7a4cc1d..ee4e4257c9fa8 100644
--- a/libc/docs/dev/code_style.rst
+++ b/libc/docs/dev/code_style.rst
@@ -186,3 +186,36 @@ We expect contributions to be free of warnings from the `minimum supported
 compiler versions`__ (and newer).
 
 .. __: https://libc.llvm.org/compiler_support.html#minimum-supported-versions
+
+Header Inclusion Policy
+=======================
+
+Because llvm-libc supports
+`Overlay Mode <https://libc.llvm.org/overlay_mode.html>`__ and
+`Fullbuild Mode <https://libc.llvm.org/fullbuild_mode.html>`__ care must be
+taken when ``#include``'ing certain headers.
+
+The ``include/`` directory contains public facing headers that users must
+consume for fullbuild mode. As such, types defined here will have ABI
+implications as these definitions may differ from the underlying system for
+overlay mode and are NEVER appropriate to include in ``libc/src/`` without
+preprocessor guards for ``LLVM_LIBC_FULL_BUILD``.
+
+Consider the case where an implementation in ``libc/src/`` may wish to refer to
+a ``sigset_t``, what header should be included? ``<signal.h>``, ``<spawn.h>``,
+``<sys/select.h>``?
+
+None of the above. Instead, code under ``src/`` should ``#include
+"hdr/types/sigset_t.h"`` which contains preprocessor guards on
+``LLVM_LIBC_FULL_BUILD`` to either include the public type (fullbuild mode) or
+the underlying system header (overlay mode).
+
+Implementations in ``libc/src/`` should NOT be ``#include``'ing using ``<>`` or
+``"include/*``, except for these "proxy" headers that first check for
+``LLVM_LIBC_FULL_BUILD``.
+
+These "proxy" headers are similarly used when referring to preprocessor
+defines. Code under ``libc/src/`` should ``#include`` a proxy header from
+``hdr/``, which contains a guard on ``LLVM_LIBC_FULL_BUILD`` to either include
+our header from ``libc/include/`` (fullbuild) or the corresponding underlying
+system header (overlay).
diff --git a/libc/docs/usage_modes.rst b/libc/docs/usage_modes.rst
index 11c10623b61db..8e5dcca6e0a75 100644
--- a/libc/docs/usage_modes.rst
+++ b/libc/docs/usage_modes.rst
@@ -6,6 +6,10 @@ The libc can used in two different modes:
 
 #. The **overlay** mode: In this mode, the link order semantics are exploited
    to overlay implementations from LLVM's libc over the system libc. See
-   :ref:`overlay_mode` for more information about this mode.
+   :ref:`overlay_mode` for more information about this mode. In this mode, libc
+   uses the ABI of the system it's being overlayed onto. Headers are NOT
+   generated. libllvmlibc.a is the only build artifact.
 #. The **fullbuild** mode: In this mode, LLVM's libc is used as the only libc
    for the binary. See :ref:`fullbuild_mode` for information about this mode.
+   In this mode, libc uses its own ABI. Headers are generated along with a
+   libc.a.
diff --git a/libc/hdr/CMakeLists.txt b/libc/hdr/CMakeLists.txt
index 5a1acd9d17ab4..38ef56e3f04c0 100644
--- a/libc/hdr/CMakeLists.txt
+++ b/libc/hdr/CMakeLists.txt
@@ -40,3 +40,5 @@ add_proxy_header_library(
     libc.include.llvm-libc-macros.fenv_macros
     libc.include.fenv
 )
+
+add_subdirectory(types)
diff --git a/libc/hdr/types/CMakeLists.txt b/libc/hdr/types/CMakeLists.txt
new file mode 100644
index 0000000000000..b685d82fd8cc8
--- /dev/null
+++ b/libc/hdr/types/CMakeLists.txt
@@ -0,0 +1,23 @@
+add_proxy_header_library(
+  sigset_t
+  HDRS
+    sigset_t.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-types.sigset_t
+)
+
+add_proxy_header_library(
+  struct_epoll_event
+  HDRS
+    struct_epoll_event.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-types.struct_epoll_event
+)
+
+add_proxy_header_library(
+  struct_timespec
+  HDRS
+    struct_timespec.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-types.struct_timespec
+)
diff --git a/libc/hdr/types/sigset_t.h b/libc/hdr/types/sigset_t.h
new file mode 100644
index 0000000000000..695ec3029f686
--- /dev/null
+++ b/libc/hdr/types/sigset_t.h
@@ -0,0 +1,21 @@
+//===-- Proxy for sigset_t ------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef LLVM_LIBC_HDR_TYPES_SIGSET_T_H
+#define LLVM_LIBC_HDR_TYPES_SIGSET_T_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-types/sigset_t.h"
+
+#else
+
+#include <signal.h>
+
+#endif // LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_TYPES_SIGSET_T_H
diff --git a/libc/hdr/types/struct_epoll_event.h b/libc/hdr/types/struct_epoll_event.h
new file mode 100644
index 0000000000000..5bb98ce05bb28
--- /dev/null
+++ b/libc/hdr/types/struct_epoll_event.h
@@ -0,0 +1,21 @@
+//===-- Proxy for struct epoll_event --------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef LLVM_LIBC_HDR_TYPES_STRUCT_EPOLL_EVENT_H
+#define LLVM_LIBC_HDR_TYPES_STRUCT_EPOLL_EVENT_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-types/struct_epoll_event.h"
+
+#else
+
+#include <sys/epoll.h>
+
+#endif // LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_TYPES_STRUCT_EPOLL_EVENT_H
diff --git a/libc/hdr/types/struct_timespec.h b/libc/hdr/types/struct_timespec.h
new file mode 100644
index 0000000000000..1f121f3d24d82
--- /dev/null
+++ b/libc/hdr/types/struct_timespec.h
@@ -0,0 +1,21 @@
+//===-- Proxy for struct timespec  ----------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef LLVM_LIBC_HDR_TYPES_STRUCT_TIMESPEC_H
+#define LLVM_LIBC_HDR_TYPES_STRUCT_TIMESPEC_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-types/struct_timespec.h"
+
+#else
+
+#include <time.h>
+
+#endif // LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_TYPES_STRUCT_TIMESPEC_H
diff --git a/libc/src/signal/linux/CMakeLists.txt b/libc/src/signal/linux/CMakeLists.txt
index 77a2453b25a0a..7606b4b21d3dd 100644
--- a/libc/src/signal/linux/CMakeLists.txt
+++ b/libc/src/signal/linux/CMakeLists.txt
@@ -3,6 +3,8 @@ add_header_library(
   HDRS
     signal_utils.h
   DEPENDS
+    libc.hdr.types.sigset_t
+    libc.include.signal
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
 )
@@ -28,7 +30,7 @@ add_entrypoint_object(
     ../raise.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
 )
@@ -57,7 +59,7 @@ add_entrypoint_object(
     ../sigaction.h
   DEPENDS
     .__restore
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
@@ -84,7 +86,7 @@ add_entrypoint_object(
     ../sigprocmask.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
@@ -98,7 +100,7 @@ add_entrypoint_object(
     ../sigemptyset.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.src.errno.errno
 )
 
@@ -110,7 +112,7 @@ add_entrypoint_object(
     ../sigaddset.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.src.errno.errno
 )
 
@@ -133,7 +135,7 @@ add_entrypoint_object(
     ../sigfillset.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.src.errno.errno
 )
 
@@ -145,6 +147,6 @@ add_entrypoint_object(
     ../sigdelset.h
   DEPENDS
     .signal_utils
-    libc.include.signal
+    libc.hdr.types.sigset_t
     libc.src.errno.errno
 )
diff --git a/libc/src/signal/linux/raise.cpp b/libc/src/signal/linux/raise.cpp
index dd6f5eb4b3575..2250df5478444 100644
--- a/libc/src/signal/linux/raise.cpp
+++ b/libc/src/signal/linux/raise.cpp
@@ -7,14 +7,15 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/raise.h"
-#include "src/signal/linux/signal_utils.h"
 
+#include "hdr/types/sigset_t.h"
 #include "src/__support/common.h"
+#include "src/signal/linux/signal_utils.h"
 
 namespace LIBC_NAMESPACE {
 
 LLVM_LIBC_FUNCTION(int, raise, (int sig)) {
-  ::sigset_t sigset;
+  sigset_t sigset;
   block_all_signals(sigset);
   long pid = LIBC_NAMESPACE::syscall_impl<long>(SYS_getpid);
   long tid = LIBC_NAMESPACE::syscall_impl<long>(SYS_gettid);
diff --git a/libc/src/signal/linux/sigaction.cpp b/libc/src/signal/linux/sigaction.cpp
index 7ddc2dc5cbcc7..7b220e5c37f6f 100644
--- a/libc/src/signal/linux/sigaction.cpp
+++ b/libc/src/signal/linux/sigaction.cpp
@@ -7,12 +7,11 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/sigaction.h"
-#include "src/errno/libc_errno.h"
-#include "src/signal/linux/signal_utils.h"
 
+#include "hdr/types/sigset_t.h"
 #include "src/__support/common.h"
-
-#include <signal.h>
+#include "src/errno/libc_errno.h"
+#include "src/signal/linux/signal_utils.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/signal/linux/sigaddset.cpp b/libc/src/signal/linux/sigaddset.cpp
index 536391734e058..8fc5d43180e28 100644
--- a/libc/src/signal/linux/sigaddset.cpp
+++ b/libc/src/signal/linux/sigaddset.cpp
@@ -7,12 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/sigaddset.h"
+
+#include "hdr/types/sigset_t.h"
 #include "src/__support/common.h"
 #include "src/errno/libc_errno.h"
 #include "src/signal/linux/signal_utils.h"
 
-#include <signal.h>
-
 namespace LIBC_NAMESPACE {
 
 LLVM_LIBC_FUNCTION(int, sigaddset, (sigset_t * set, int signum)) {
diff --git a/libc/src/signal/linux/sigdelset.cpp b/libc/src/signal/linux/sigdelset.cpp
index 5cb645e461cf8..997f4574c05d0 100644
--- a/libc/src/signal/linux/sigdelset.cpp
+++ b/libc/src/signal/linux/sigdelset.cpp
@@ -7,12 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/sigdelset.h"
+
+#include "hdr/types/sigset_t.h"
 #include "src/__support/common.h"
 #include "src/errno/libc_errno.h"
 #include "src/signal/linux/signal_utils.h"
 
-#include <signal.h>
-
 namespace LIBC_NAMESPACE {
 
 LLVM_LIBC_FUNCTION(int, sigdelset, (sigset_t * set, int signum)) {
diff --git a/libc/src/signal/linux/sigfillset.cpp b/libc/src/signal/linux/sigfillset.cpp
index e17c85a897ce7..d98bbf7f619cc 100644
--- a/libc/src/signal/linux/sigfillset.cpp
+++ b/libc/src/signal/linux/sigfillset.cpp
@@ -7,12 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/sigfillset.h"
+
+#include "hdr/types/sigset_t.h"
 #include "src/__support/common.h"
 #include "src/errno/libc_errno.h"
 #include "src/signal/linux/signal_utils.h"
 
-#include <signal.h>
-
 namespace LIBC_NAMESPACE {
 
 LLVM_LIBC_FUNCTION(int, sigfillset, (sigset_t * set)) {
diff --git a/libc/src/signal/linux/signal_utils.h b/libc/src/signal/linux/signal_utils.h
index 5e9dd9a5c53ab..3fd0cc0b7b459 100644
--- a/libc/src/signal/linux/signal_utils.h
+++ b/libc/src/signal/linux/signal_utils.h
@@ -9,10 +9,11 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_LINUX_SIGNAL_UTILS_H
 #define LLVM_LIBC_SRC_SIGNAL_LINUX_SIGNAL_UTILS_H
 
+#include "hdr/types/sigset_t.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
 #include "src/__support/common.h"
 
-#include <signal.h>
+#include <signal.h> // sigaction
 #include <stddef.h>
 #include <sys/syscall.h>          // For syscall numbers.
 
diff --git a/libc/src/signal/linux/sigprocmask.cpp b/libc/src/signal/linux/sigprocmask.cpp
index 79a35dd59d75c..0e94efb6400c0 100644
--- a/libc/src/signal/linux/sigprocmask.cpp
+++ b/libc/src/signal/linux/sigprocmask.cpp
@@ -7,13 +7,13 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/signal/sigprocmask.h"
+
+#include "hdr/types/sigset_t.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
+#include "src/__support/common.h"
 #include "src/errno/libc_errno.h"
 #include "src/signal/linux/signal_utils.h"
 
-#include "src/__support/common.h"
-
-#include <signal.h>
 #include <sys/syscall.h> // For syscall numbers.
 
 namespace LIBC_NAMESPACE {
diff --git a/libc/src/signal/sigaddset.h b/libc/src/signal/sigaddset.h
index 626eb20a295c8..c703b46bc6059 100644
--- a/libc/src/signal/sigaddset.h
+++ b/libc/src/signal/sigaddset.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_SIGADDSET_H
 #define LLVM_LIBC_SRC_SIGNAL_SIGADDSET_H
 
-#include <signal.h>
+#include "hdr/types/sigset_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/signal/sigdelset.h b/libc/src/signal/sigdelset.h
index c4fdb9975fa3d..7bdb6e6d18fdd 100644
--- a/libc/src/signal/sigdelset.h
+++ b/libc/src/signal/sigdelset.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_SIGDELSET_H
 #define LLVM_LIBC_SRC_SIGNAL_SIGDELSET_H
 
-#include <signal.h>
+#include "hdr/types/sigset_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/signal/sigemptyset.h b/libc/src/signal/sigemptyset.h
index f3763d1f4f3d4..661fd33b888e0 100644
--- a/libc/src/signal/sigemptyset.h
+++ b/libc/src/signal/sigemptyset.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_SIGEMPTYSET_H
 #define LLVM_LIBC_SRC_SIGNAL_SIGEMPTYSET_H
 
-#include <signal.h>
+#include "hdr/types/sigset_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/signal/sigfillset.h b/libc/src/signal/sigfillset.h
index d8e3168871ea8..2849aacf953b1 100644
--- a/libc/src/signal/sigfillset.h
+++ b/libc/src/signal/sigfillset.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_SIGFILLSET_H
 #define LLVM_LIBC_SRC_SIGNAL_SIGFILLSET_H
 
-#include <signal.h>
+#include "hdr/types/sigset_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/signal/sigprocmask.h b/libc/src/signal/sigprocmask.h
index e0658860579e4..8569578eb68ca 100644
--- a/libc/src/signal/sigprocmask.h
+++ b/libc/src/signal/sigprocmask.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SIGNAL_SIGPROCMASK_H
 #define LLVM_LIBC_SRC_SIGNAL_SIGPROCMASK_H
 
-#include <signal.h>
+#include "hdr/types/sigset_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/epoll_pwait.h b/libc/src/sys/epoll/epoll_pwait.h
index 9dcb55533009f..801aa97610001 100644
--- a/libc/src/sys/epoll/epoll_pwait.h
+++ b/libc/src/sys/epoll/epoll_pwait.h
@@ -9,11 +9,8 @@
 #ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_PWAIT_H
 #define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_PWAIT_H
 
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/sigset_t.h"
-// #include "include/llvm-libc-types/struct_epoll_event.h"
-
-#include <sys/epoll.h>
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_epoll_event.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/epoll_pwait2.h b/libc/src/sys/epoll/epoll_pwait2.h
index 622ede6a0f9f9..7fc528b2fd25d 100644
--- a/libc/src/sys/epoll/epoll_pwait2.h
+++ b/libc/src/sys/epoll/epoll_pwait2.h
@@ -9,12 +9,9 @@
 #ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_PWAIT2_H
 #define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_PWAIT2_H
 
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/sigset_t.h"
-// #include "include/llvm-libc-types/struct_epoll_event.h"
-// #include "include/llvm-libc-types/struct_timespec.h"
-
-#include <sys/epoll.h>
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_epoll_event.h"
+#include "hdr/types/struct_timespec.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/epoll_wait.h b/libc/src/sys/epoll/epoll_wait.h
index d51c9100846ce..b546e91e4c2ee 100644
--- a/libc/src/sys/epoll/epoll_wait.h
+++ b/libc/src/sys/epoll/epoll_wait.h
@@ -9,10 +9,7 @@
 #ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_WAIT_H
 #define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_WAIT_H
 
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/struct_epoll_event.h"
-
-#include <sys/epoll.h>
+#include "hdr/types/struct_epoll_event.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/linux/CMakeLists.txt b/libc/src/sys/epoll/linux/CMakeLists.txt
index a27905d962dc5..586aac7055dc1 100644
--- a/libc/src/sys/epoll/linux/CMakeLists.txt
+++ b/libc/src/sys/epoll/linux/CMakeLists.txt
@@ -5,7 +5,9 @@ add_entrypoint_object(
   HDRS
     ../epoll_wait.h
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.types.sigset_t
+    libc.hdr.types.struct_epoll_event
+    libc.hdr.types.struct_timespec
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
@@ -18,7 +20,9 @@ add_entrypoint_object(
   HDRS
     ../epoll_pwait.h
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.types.sigset_t
+    libc.hdr.types.struct_epoll_event
+    libc.hdr.types.struct_timespec
     libc.include.signal
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
@@ -32,10 +36,12 @@ add_entrypoint_object(
   HDRS
     ../epoll_pwait2.h
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.types.sigset_t
+    libc.hdr.types.struct_epoll_event
+    libc.hdr.types.struct_timespec
     libc.include.signal
-    libc.include.time
     libc.include.sys_syscall
+    libc.include.time
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
 )
diff --git a/libc/src/sys/epoll/linux/epoll_pwait.cpp b/libc/src/sys/epoll/linux/epoll_pwait.cpp
index ee1b4e66e9844..ac012944a9577 100644
--- a/libc/src/sys/epoll/linux/epoll_pwait.cpp
+++ b/libc/src/sys/epoll/linux/epoll_pwait.cpp
@@ -8,17 +8,13 @@
 
 #include "src/sys/epoll/epoll_pwait.h"
 
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_epoll_event.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
 #include "src/__support/common.h"
-
 #include "src/errno/libc_errno.h"
-#include <sys/syscall.h> // For syscall numbers.
 
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/sigset_t.h"
-// #include "include/llvm-libc-types/struct_epoll_event.h"
-
-#include <sys/epoll.h>
+#include <sys/syscall.h> // For syscall numbers.
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/linux/epoll_pwait2.cpp b/libc/src/sys/epoll/linux/epoll_pwait2.cpp
index 671dede2a1058..3c42e38deb22b 100644
--- a/libc/src/sys/epoll/linux/epoll_pwait2.cpp
+++ b/libc/src/sys/epoll/linux/epoll_pwait2.cpp
@@ -8,18 +8,14 @@
 
 #include "src/sys/epoll/epoll_pwait2.h"
 
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_epoll_event.h"
+#include "hdr/types/struct_timespec.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
 #include "src/__support/common.h"
-
 #include "src/errno/libc_errno.h"
-#include <sys/syscall.h> // For syscall numbers.
 
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/sigset_t.h"
-// #include "include/llvm-libc-types/struct_epoll_event.h"
-// #include "include/llvm-libc-types/struct_timespec.h"
-
-#include <sys/epoll.h>
+#include <sys/syscall.h> // For syscall numbers.
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/epoll/linux/epoll_wait.cpp b/libc/src/sys/epoll/linux/epoll_wait.cpp
index 0c43edf764545..18dd6e2b83543 100644
--- a/libc/src/sys/epoll/linux/epoll_wait.cpp
+++ b/libc/src/sys/epoll/linux/epoll_wait.cpp
@@ -8,16 +8,13 @@
 
 #include "src/sys/epoll/epoll_wait.h"
 
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_epoll_event.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
 #include "src/__support/common.h"
 #include "src/errno/libc_errno.h"
-#include <sys/syscall.h> // For syscall numbers.
-
-// TODO: Use this include once the include headers are also using quotes.
-// #include "include/llvm-libc-types/sigset_t.h"
-// #include "include/llvm-libc-types/struct_epoll_event.h"
 
-#include <sys/epoll.h>
+#include <sys/syscall.h> // For syscall numbers.
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/sys/select/linux/select.cpp b/libc/src/sys/select/linux/select.cpp
index 3f387c14ec560..9034b75e5c29e 100644
--- a/libc/src/sys/select/linux/select.cpp
+++ b/libc/src/sys/select/linux/select.cpp
@@ -8,14 +8,14 @@
 
 #include "src/sys/select/select.h"
 
+#include "hdr/types/sigset_t.h"
+#include "hdr/types/struct_timespec.h"
 #include "src/__support/CPP/limits.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
 #include "src/__support/common.h"
-
 #include "src/errno/libc_errno.h"
-#include <signal.h>
-#include <stddef.h> // For size_t
-#include <sys/select.h>
+
+#include <stddef.h>      // For size_t
 #include <sys/syscall.h> // For syscall numbers.
 
 namespace LIBC_NAMESPACE {
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index d38dc3029f74f..c61904a967117 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -3417,6 +3417,15 @@ libc_function(
 
 ############################## sys/epoll targets ###############################
 
+libc_support_library(
+  name = "types_sigset_t",
+  hdrs = ["hdr/types/sigset_t.h"],
+)
+libc_support_library(
+  name = "types_struct_epoll_event",
+  hdrs = ["hdr/types/struct_epoll_event.h"],
+)
+
 libc_function(
     name = "epoll_wait",
     srcs = ["src/sys/epoll/linux/epoll_wait.cpp"],
@@ -3429,6 +3438,8 @@ libc_function(
     deps = [
         ":__support_osutil_syscall",
         ":errno",
+        ":types_sigset_t",
+        ":types_struct_epoll_event",
     ],
 )
 
@@ -3444,6 +3455,8 @@ libc_function(
     deps = [
         ":__support_osutil_syscall",
         ":errno",
+        ":types_sigset_t",
+        ":types_struct_epoll_event",
     ],
 )
 

From 4e6d18f40642c2cc8e124bbe55810b2d9b2ac9c0 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Thu, 11 Apr 2024 13:20:05 -0400
Subject: [PATCH 163/886] [Clang][AST] Track whether template template
 parameters used the 'typename' keyword (#88139)

This patch adds a `Typename` bit-field to `TemplateTemplateParmDecl`
which stores whether the template template parameter was declared with
the `typename` keyword.
---
 clang/docs/ReleaseNotes.rst                   |  4 ++
 clang/include/clang/AST/DeclTemplate.h        | 49 +++++++++++++------
 clang/include/clang/Sema/Sema.h               |  2 +-
 clang/lib/AST/ASTContext.cpp                  |  2 +-
 clang/lib/AST/ASTImporter.cpp                 |  3 +-
 clang/lib/AST/DeclPrinter.cpp                 |  5 +-
 clang/lib/AST/DeclTemplate.cpp                | 18 +++----
 clang/lib/Parse/ParseTemplate.cpp             |  9 ++--
 clang/lib/Sema/SemaTemplate.cpp               | 24 ++++-----
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |  6 ++-
 clang/lib/Serialization/ASTReaderDecl.cpp     |  1 +
 clang/lib/Serialization/ASTWriterDecl.cpp     |  1 +
 clang/unittests/AST/DeclPrinterTest.cpp       |  3 +-
 13 files changed, 75 insertions(+), 52 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 93318871fa9f6..93a380411604b 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -360,6 +360,9 @@ Improvements to Clang's diagnostics
   Added the ``-Wtentative-definition-array`` warning group to cover this.
   Fixes #GH87766
 
+- Clang now uses the correct type-parameter-key (``class`` or ``typename``) when printing
+  template template parameter declarations.
+
 Improvements to Clang's time-trace
 ----------------------------------
 
@@ -535,6 +538,7 @@ Bug Fixes to C++ Support
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
 - Clang now properly preserves ``FoundDecls`` within a ``ConceptReference``. (#GH82628)
+- The presence of the ``typename`` keyword is now stored in ``TemplateTemplateParmDecl``.
 
 Miscellaneous Bug Fixes
 ^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/include/clang/AST/DeclTemplate.h b/clang/include/clang/AST/DeclTemplate.h
index cb598cb81840d..f24e71ff22964 100644
--- a/clang/include/clang/AST/DeclTemplate.h
+++ b/clang/include/clang/AST/DeclTemplate.h
@@ -1581,26 +1581,36 @@ class TemplateTemplateParmDecl final
       DefaultArgStorage<TemplateTemplateParmDecl, TemplateArgumentLoc *>;
   DefArgStorage DefaultArgument;
 
+  /// Whether this template template parameter was declaration with
+  /// the 'typename' keyword.
+  ///
+  /// If false, it was declared with the 'class' keyword.
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned Typename : 1;
+
   /// Whether this parameter is a parameter pack.
-  bool ParameterPack;
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned ParameterPack : 1;
 
   /// Whether this template template parameter is an "expanded"
   /// parameter pack, meaning that it is a pack expansion and we
   /// already know the set of template parameters that expansion expands to.
-  bool ExpandedParameterPack = false;
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned ExpandedParameterPack : 1;
 
   /// The number of parameters in an expanded parameter pack.
   unsigned NumExpandedParams = 0;
 
-  TemplateTemplateParmDecl(DeclContext *DC, SourceLocation L,
-                           unsigned D, unsigned P, bool ParameterPack,
-                           IdentifierInfo *Id, TemplateParameterList *Params)
+  TemplateTemplateParmDecl(DeclContext *DC, SourceLocation L, unsigned D,
+                           unsigned P, bool ParameterPack, IdentifierInfo *Id,
+                           bool Typename, TemplateParameterList *Params)
       : TemplateDecl(TemplateTemplateParm, DC, L, Id, Params),
-        TemplateParmPosition(D, P), ParameterPack(ParameterPack) {}
+        TemplateParmPosition(D, P), Typename(Typename),
+        ParameterPack(ParameterPack), ExpandedParameterPack(false) {}
 
-  TemplateTemplateParmDecl(DeclContext *DC, SourceLocation L,
-                           unsigned D, unsigned P,
-                           IdentifierInfo *Id, TemplateParameterList *Params,
+  TemplateTemplateParmDecl(DeclContext *DC, SourceLocation L, unsigned D,
+                           unsigned P, IdentifierInfo *Id, bool Typename,
+                           TemplateParameterList *Params,
                            ArrayRef<TemplateParameterList *> Expansions);
 
   void anchor() override;
@@ -1613,14 +1623,13 @@ class TemplateTemplateParmDecl final
   static TemplateTemplateParmDecl *Create(const ASTContext &C, DeclContext *DC,
                                           SourceLocation L, unsigned D,
                                           unsigned P, bool ParameterPack,
-                                          IdentifierInfo *Id,
+                                          IdentifierInfo *Id, bool Typename,
                                           TemplateParameterList *Params);
-  static TemplateTemplateParmDecl *Create(const ASTContext &C, DeclContext *DC,
-                                          SourceLocation L, unsigned D,
-                                          unsigned P,
-                                          IdentifierInfo *Id,
-                                          TemplateParameterList *Params,
-                                 ArrayRef<TemplateParameterList *> Expansions);
+  static TemplateTemplateParmDecl *
+  Create(const ASTContext &C, DeclContext *DC, SourceLocation L, unsigned D,
+         unsigned P, IdentifierInfo *Id, bool Typename,
+         TemplateParameterList *Params,
+         ArrayRef<TemplateParameterList *> Expansions);
 
   static TemplateTemplateParmDecl *CreateDeserialized(ASTContext &C,
                                                       unsigned ID);
@@ -1634,6 +1643,14 @@ class TemplateTemplateParmDecl final
   using TemplateParmPosition::setPosition;
   using TemplateParmPosition::getIndex;
 
+  /// Whether this template template parameter was declared with
+  /// the 'typename' keyword.
+  bool wasDeclaredWithTypename() const { return Typename; }
+
+  /// Set whether this template template parameter was declared with
+  /// the 'typename' or 'class' keyword.
+  void setDeclaredWithTypename(bool withTypename) { Typename = withTypename; }
+
   /// Whether this template template parameter is a template
   /// parameter pack.
   ///
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 0ee4f3c8e127f..f2c55b13b6d32 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -9064,7 +9064,7 @@ class Sema final : public SemaBase {
                                            Expr *DefaultArg);
   NamedDecl *ActOnTemplateTemplateParameter(
       Scope *S, SourceLocation TmpLoc, TemplateParameterList *Params,
-      SourceLocation EllipsisLoc, IdentifierInfo *ParamName,
+      bool Typename, SourceLocation EllipsisLoc, IdentifierInfo *ParamName,
       SourceLocation ParamNameLoc, unsigned Depth, unsigned Position,
       SourceLocation EqualLoc, ParsedTemplateArgument DefaultArg);
 
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 2fa6aedca4c6a..6ce233704a588 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -799,7 +799,7 @@ ASTContext::getCanonicalTemplateTemplateParmDecl(
 
   TemplateTemplateParmDecl *CanonTTP = TemplateTemplateParmDecl::Create(
       *this, getTranslationUnitDecl(), SourceLocation(), TTP->getDepth(),
-      TTP->getPosition(), TTP->isParameterPack(), nullptr,
+      TTP->getPosition(), TTP->isParameterPack(), nullptr, /*Typename=*/false,
       TemplateParameterList::Create(*this, SourceLocation(), SourceLocation(),
                                     CanonParams, SourceLocation(),
                                     /*RequiresClause=*/nullptr));
diff --git a/clang/lib/AST/ASTImporter.cpp b/clang/lib/AST/ASTImporter.cpp
index d5ec5ee409156..a5e43fc631667 100644
--- a/clang/lib/AST/ASTImporter.cpp
+++ b/clang/lib/AST/ASTImporter.cpp
@@ -5952,7 +5952,8 @@ ASTNodeImporter::VisitTemplateTemplateParmDecl(TemplateTemplateParmDecl *D) {
           ToD, D, Importer.getToContext(),
           Importer.getToContext().getTranslationUnitDecl(), *LocationOrErr,
           D->getDepth(), D->getPosition(), D->isParameterPack(),
-          (*NameOrErr).getAsIdentifierInfo(), *TemplateParamsOrErr))
+          (*NameOrErr).getAsIdentifierInfo(), D->wasDeclaredWithTypename(),
+          *TemplateParamsOrErr))
     return ToD;
 
   if (D->hasDefaultArgument()) {
diff --git a/clang/lib/AST/DeclPrinter.cpp b/clang/lib/AST/DeclPrinter.cpp
index 6afdb6cfccb14..c66774dd1df15 100644
--- a/clang/lib/AST/DeclPrinter.cpp
+++ b/clang/lib/AST/DeclPrinter.cpp
@@ -1218,7 +1218,10 @@ void DeclPrinter::VisitTemplateDecl(const TemplateDecl *D) {
 
   if (const TemplateTemplateParmDecl *TTP =
         dyn_cast<TemplateTemplateParmDecl>(D)) {
-    Out << "class";
+    if (TTP->wasDeclaredWithTypename())
+      Out << "typename";
+    else
+      Out << "class";
 
     if (TTP->isParameterPack())
       Out << " ...";
diff --git a/clang/lib/AST/DeclTemplate.cpp b/clang/lib/AST/DeclTemplate.cpp
index 571ed81a42e40..5aa2484197372 100644
--- a/clang/lib/AST/DeclTemplate.cpp
+++ b/clang/lib/AST/DeclTemplate.cpp
@@ -805,10 +805,10 @@ void TemplateTemplateParmDecl::anchor() {}
 
 TemplateTemplateParmDecl::TemplateTemplateParmDecl(
     DeclContext *DC, SourceLocation L, unsigned D, unsigned P,
-    IdentifierInfo *Id, TemplateParameterList *Params,
+    IdentifierInfo *Id, bool Typename, TemplateParameterList *Params,
     ArrayRef<TemplateParameterList *> Expansions)
     : TemplateDecl(TemplateTemplateParm, DC, L, Id, Params),
-      TemplateParmPosition(D, P), ParameterPack(true),
+      TemplateParmPosition(D, P), Typename(Typename), ParameterPack(true),
       ExpandedParameterPack(true), NumExpandedParams(Expansions.size()) {
   if (!Expansions.empty())
     std::uninitialized_copy(Expansions.begin(), Expansions.end(),
@@ -819,26 +819,26 @@ TemplateTemplateParmDecl *
 TemplateTemplateParmDecl::Create(const ASTContext &C, DeclContext *DC,
                                  SourceLocation L, unsigned D, unsigned P,
                                  bool ParameterPack, IdentifierInfo *Id,
-                                 TemplateParameterList *Params) {
+                                 bool Typename, TemplateParameterList *Params) {
   return new (C, DC) TemplateTemplateParmDecl(DC, L, D, P, ParameterPack, Id,
-                                              Params);
+                                              Typename, Params);
 }
 
 TemplateTemplateParmDecl *
 TemplateTemplateParmDecl::Create(const ASTContext &C, DeclContext *DC,
                                  SourceLocation L, unsigned D, unsigned P,
-                                 IdentifierInfo *Id,
+                                 IdentifierInfo *Id, bool Typename,
                                  TemplateParameterList *Params,
                                  ArrayRef<TemplateParameterList *> Expansions) {
   return new (C, DC,
               additionalSizeToAlloc<TemplateParameterList *>(Expansions.size()))
-      TemplateTemplateParmDecl(DC, L, D, P, Id, Params, Expansions);
+      TemplateTemplateParmDecl(DC, L, D, P, Id, Typename, Params, Expansions);
 }
 
 TemplateTemplateParmDecl *
 TemplateTemplateParmDecl::CreateDeserialized(ASTContext &C, unsigned ID) {
   return new (C, ID) TemplateTemplateParmDecl(nullptr, SourceLocation(), 0, 0,
-                                              false, nullptr, nullptr);
+                                              false, nullptr, false, nullptr);
 }
 
 TemplateTemplateParmDecl *
@@ -847,7 +847,7 @@ TemplateTemplateParmDecl::CreateDeserialized(ASTContext &C, unsigned ID,
   auto *TTP =
       new (C, ID, additionalSizeToAlloc<TemplateParameterList *>(NumExpansions))
           TemplateTemplateParmDecl(nullptr, SourceLocation(), 0, 0, nullptr,
-                                   nullptr, std::nullopt);
+                                   false, nullptr, std::nullopt);
   TTP->NumExpandedParams = NumExpansions;
   return TTP;
 }
@@ -1469,7 +1469,7 @@ createMakeIntegerSeqParameterList(const ASTContext &C, DeclContext *DC) {
   // template <typename T, ...Ints> class IntSeq
   auto *TemplateTemplateParm = TemplateTemplateParmDecl::Create(
       C, DC, SourceLocation(), /*Depth=*/0, /*Position=*/0,
-      /*ParameterPack=*/false, /*Id=*/nullptr, TPL);
+      /*ParameterPack=*/false, /*Id=*/nullptr, /*Typename=*/false, TPL);
   TemplateTemplateParm->setImplicit(true);
 
   // typename T
diff --git a/clang/lib/Parse/ParseTemplate.cpp b/clang/lib/Parse/ParseTemplate.cpp
index 03257500426e5..b07ce451e878e 100644
--- a/clang/lib/Parse/ParseTemplate.cpp
+++ b/clang/lib/Parse/ParseTemplate.cpp
@@ -805,10 +805,12 @@ NamedDecl *Parser::ParseTemplateTemplateParameter(unsigned Depth,
   // identifier, comma, or greater. Provide a fixit if the identifier, comma,
   // or greater appear immediately or after 'struct'. In the latter case,
   // replace the keyword with 'class'.
+  bool TypenameKeyword = false;
   if (!TryConsumeToken(tok::kw_class)) {
     bool Replace = Tok.isOneOf(tok::kw_typename, tok::kw_struct);
     const Token &Next = Tok.is(tok::kw_struct) ? NextToken() : Tok;
     if (Tok.is(tok::kw_typename)) {
+      TypenameKeyword = true;
       Diag(Tok.getLocation(),
            getLangOpts().CPlusPlus17
                ? diag::warn_cxx14_compat_template_template_param_typename
@@ -878,10 +880,9 @@ NamedDecl *Parser::ParseTemplateTemplateParameter(unsigned Depth,
     }
   }
 
-  return Actions.ActOnTemplateTemplateParameter(getCurScope(), TemplateLoc,
-                                                ParamList, EllipsisLoc,
-                                                ParamName, NameLoc, Depth,
-                                                Position, EqualLoc, DefaultArg);
+  return Actions.ActOnTemplateTemplateParameter(
+      getCurScope(), TemplateLoc, ParamList, TypenameKeyword, EllipsisLoc,
+      ParamName, NameLoc, Depth, Position, EqualLoc, DefaultArg);
 }
 
 /// ParseNonTypeTemplateParameter - Handle the parsing of non-type
diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
index 951e5a31cab3b..e0f5e53dc2481 100644
--- a/clang/lib/Sema/SemaTemplate.cpp
+++ b/clang/lib/Sema/SemaTemplate.cpp
@@ -1630,26 +1630,20 @@ NamedDecl *Sema::ActOnNonTypeTemplateParameter(Scope *S, Declarator &D,
 /// ActOnTemplateTemplateParameter - Called when a C++ template template
 /// parameter (e.g. T in template <template \<typename> class T> class array)
 /// has been parsed. S is the current scope.
-NamedDecl *Sema::ActOnTemplateTemplateParameter(Scope* S,
-                                           SourceLocation TmpLoc,
-                                           TemplateParameterList *Params,
-                                           SourceLocation EllipsisLoc,
-                                           IdentifierInfo *Name,
-                                           SourceLocation NameLoc,
-                                           unsigned Depth,
-                                           unsigned Position,
-                                           SourceLocation EqualLoc,
-                                           ParsedTemplateArgument Default) {
+NamedDecl *Sema::ActOnTemplateTemplateParameter(
+    Scope *S, SourceLocation TmpLoc, TemplateParameterList *Params,
+    bool Typename, SourceLocation EllipsisLoc, IdentifierInfo *Name,
+    SourceLocation NameLoc, unsigned Depth, unsigned Position,
+    SourceLocation EqualLoc, ParsedTemplateArgument Default) {
   assert(S->isTemplateParamScope() &&
          "Template template parameter not in template parameter scope!");
 
   // Construct the parameter object.
   bool IsParameterPack = EllipsisLoc.isValid();
-  TemplateTemplateParmDecl *Param =
-    TemplateTemplateParmDecl::Create(Context, Context.getTranslationUnitDecl(),
-                                     NameLoc.isInvalid()? TmpLoc : NameLoc,
-                                     Depth, Position, IsParameterPack,
-                                     Name, Params);
+  TemplateTemplateParmDecl *Param = TemplateTemplateParmDecl::Create(
+      Context, Context.getTranslationUnitDecl(),
+      NameLoc.isInvalid() ? TmpLoc : NameLoc, Depth, Position, IsParameterPack,
+      Name, Typename, Params);
   Param->setAccess(AS_public);
 
   if (Param->isParameterPack())
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 127a432367b95..2aa696d35d1dd 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -3233,12 +3233,14 @@ TemplateDeclInstantiator::VisitTemplateTemplateParmDecl(
     Param = TemplateTemplateParmDecl::Create(
         SemaRef.Context, Owner, D->getLocation(),
         D->getDepth() - TemplateArgs.getNumSubstitutedLevels(),
-        D->getPosition(), D->getIdentifier(), InstParams, ExpandedParams);
+        D->getPosition(), D->getIdentifier(), D->wasDeclaredWithTypename(),
+        InstParams, ExpandedParams);
   else
     Param = TemplateTemplateParmDecl::Create(
         SemaRef.Context, Owner, D->getLocation(),
         D->getDepth() - TemplateArgs.getNumSubstitutedLevels(),
-        D->getPosition(), D->isParameterPack(), D->getIdentifier(), InstParams);
+        D->getPosition(), D->isParameterPack(), D->getIdentifier(),
+        D->wasDeclaredWithTypename(), InstParams);
   if (D->hasDefaultArgument() && !D->defaultArgumentWasInherited()) {
     NestedNameSpecifierLoc QualifierLoc =
         D->getDefaultArgument().getTemplateQualifierLoc();
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index 9e49a3780ff41..5b7b9f19a106c 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -2715,6 +2715,7 @@ void ASTDeclReader::VisitNonTypeTemplateParmDecl(NonTypeTemplateParmDecl *D) {
 
 void ASTDeclReader::VisitTemplateTemplateParmDecl(TemplateTemplateParmDecl *D) {
   VisitTemplateDecl(D);
+  D->setDeclaredWithTypename(Record.readBool());
   // TemplateParmPosition.
   D->setDepth(Record.readInt());
   D->setPosition(Record.readInt());
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index 86f64bf2a2425..87e773be54fae 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -1921,6 +1921,7 @@ void ASTDeclWriter::VisitTemplateTemplateParmDecl(TemplateTemplateParmDecl *D) {
     Record.push_back(D->getNumExpansionTemplateParameters());
 
   VisitTemplateDecl(D);
+  Record.push_back(D->wasDeclaredWithTypename());
   // TemplateParmPosition.
   Record.push_back(D->getDepth());
   Record.push_back(D->getPosition());
diff --git a/clang/unittests/AST/DeclPrinterTest.cpp b/clang/unittests/AST/DeclPrinterTest.cpp
index f2b027a25621c..c24e442621c92 100644
--- a/clang/unittests/AST/DeclPrinterTest.cpp
+++ b/clang/unittests/AST/DeclPrinterTest.cpp
@@ -1220,8 +1220,7 @@ TEST(DeclPrinter, TestTemplateTemplateParameterWrittenWithTypename) {
   ASSERT_TRUE(PrintedDeclCXX17Matches(
       "template <template <typename> typename Z> void A();",
       functionTemplateDecl(hasName("A")).bind("id"),
-      "template <template <typename> class Z> void A()"));
-  // WRONG: We should use typename if the parameter was written with it.
+      "template <template <typename> typename Z> void A()"));
 }
 
 TEST(DeclPrinter, TestTemplateArgumentList1) {

From 62fa12ad2405cbdf24d5daf489e276c67a38e482 Mon Sep 17 00:00:00 2001
From: Peiming Liu <peiming@google.com>
Date: Thu, 11 Apr 2024 10:26:03 -0700
Subject: [PATCH 164/886] =?UTF-8?q?[mlir][sparse]=20support=20querying=20s?=
 =?UTF-8?q?parse=20buffer=20types=20from=20sparse=20tenso=E2=80=A6=20(#883?=
 =?UTF-8?q?08)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…r encodings.
---
 .../SparseTensor/IR/SparseTensorAttrDefs.td   | 20 ++++++
 .../SparseTensor/IR/SparseTensorType.h        | 12 +---
 .../SparseTensor/IR/SparseTensorDialect.cpp   | 62 ++++++++++++++++---
 3 files changed, 75 insertions(+), 19 deletions(-)

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
index d3be8a3009ba1..4a9b9169ae4b8 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
@@ -401,6 +401,26 @@ def SparseTensorEncodingAttr : SparseTensor_Attr<"SparseTensorEncoding",
     /// the null encoding (since dense-tensors are always all-ordered).
     bool isAllOrdered() const;
 
+    //
+    // Storage type methods.
+    //
+
+    /// Returns the coordinate-overhead MLIR type, defaulting to `IndexType`.
+    Type getCrdElemType() const;
+
+    /// Returns the position-overhead MLIR type, defaulting to `IndexType`.
+    Type getPosElemType() const;
+
+    /// Returns the coordinate-memref MLIR type, an optional tensorDimShape is
+    /// used to refine the leading batch dimensions (if any).
+    MemRefType getCrdMemRefType(
+      std::optional<ArrayRef<int64_t>> tensorDimShape = std::nullopt) const;
+
+    /// Returns the position-memref MLIR type, an optional tensorDimShape is
+    /// used to refine the leading batch dimensions (if any).
+    MemRefType getPosMemRefType(
+      std::optional<ArrayRef<int64_t>> tensorDimShape = std::nullopt) const;
+
     //
     // dimToLvl methods.
     //
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorType.h b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorType.h
index dc770c2e904cb..825d89a408feb 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorType.h
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorType.h
@@ -328,18 +328,10 @@ class SparseTensorType {
   unsigned getPosWidth() const { return enc ? enc.getPosWidth() : 0; }
 
   /// Returns the coordinate-overhead MLIR type, defaulting to `IndexType`.
-  Type getCrdType() const {
-    if (getCrdWidth())
-      return IntegerType::get(getContext(), getCrdWidth());
-    return IndexType::get(getContext());
-  }
+  Type getCrdType() const { return enc.getCrdElemType(); }
 
   /// Returns the position-overhead MLIR type, defaulting to `IndexType`.
-  Type getPosType() const {
-    if (getPosWidth())
-      return IntegerType::get(getContext(), getPosWidth());
-    return IndexType::get(getContext());
-  }
+  Type getPosType() const { return enc.getPosElemType(); }
 
   /// Returns true iff this sparse tensor type has a trailing
   /// COO region starting at the given level. By default, it
diff --git a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
index e4d93c5623b9c..e9058394d33da 100644
--- a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
+++ b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
@@ -61,6 +61,26 @@ static constexpr bool acceptBitWidth(unsigned bitWidth) {
   }
 }
 
+static SmallVector<Size>
+getSparseFieldShape(const SparseTensorEncodingAttr enc,
+                    std::optional<ArrayRef<int64_t>> dimShape) {
+  assert(enc);
+  // With only encoding, we can not determine the static shape for leading
+  // batch levels, we therefore return a dynamic shape memref instead.
+  SmallVector<int64_t> memrefShape(enc.getBatchLvlRank(), ShapedType::kDynamic);
+  if (dimShape.has_value()) {
+    // If the actual tensor shape is provided, we can then refine the leading
+    // batch dimension.
+    SmallVector<int64_t> lvlShape =
+        enc.translateShape(*dimShape, CrdTransDirectionKind::dim2lvl);
+    memrefShape.assign(lvlShape.begin(),
+                       lvlShape.begin() + enc.getBatchLvlRank());
+  }
+  // Another dynamic dimension to store the sparse level.
+  memrefShape.push_back(ShapedType::kDynamic);
+  return memrefShape;
+}
+
 //===----------------------------------------------------------------------===//
 // SparseTensorDialect StorageLayout.
 //===----------------------------------------------------------------------===//
@@ -122,21 +142,17 @@ void sparse_tensor::foreachFieldAndTypeInSparseTensor(
                             LevelType)>
         callback) {
   assert(stt.hasEncoding());
-  // Construct the basic types.
-  const Type crdType = stt.getCrdType();
-  const Type posType = stt.getPosType();
-  const Type eltType = stt.getElementType();
 
-  SmallVector<int64_t> memrefShape = stt.getBatchLvlShape();
-  memrefShape.push_back(ShapedType::kDynamic);
+  SmallVector<int64_t> memrefShape =
+      getSparseFieldShape(stt.getEncoding(), stt.getDimShape());
 
   const Type specType = StorageSpecifierType::get(stt.getEncoding());
   // memref<[batch] x ? x pos>  positions
-  const Type posMemType = MemRefType::get(memrefShape, posType);
+  const Type posMemType = MemRefType::get(memrefShape, stt.getPosType());
   // memref<[batch] x ? x crd>  coordinates
-  const Type crdMemType = MemRefType::get(memrefShape, crdType);
+  const Type crdMemType = MemRefType::get(memrefShape, stt.getCrdType());
   // memref<[batch] x ? x eltType> values
-  const Type valMemType = MemRefType::get(memrefShape, eltType);
+  const Type valMemType = MemRefType::get(memrefShape, stt.getElementType());
 
   StorageLayout(stt).foreachField([specType, posMemType, crdMemType, valMemType,
                                    callback](FieldIndex fieldIdx,
@@ -354,6 +370,34 @@ bool SparseTensorEncodingAttr::isAllOrdered() const {
   return !getImpl() || llvm::all_of(getLvlTypes(), isOrderedLT);
 }
 
+Type SparseTensorEncodingAttr::getCrdElemType() const {
+  if (!getImpl())
+    return nullptr;
+  if (getCrdWidth())
+    return IntegerType::get(getContext(), getCrdWidth());
+  return IndexType::get(getContext());
+}
+
+Type SparseTensorEncodingAttr::getPosElemType() const {
+  if (!getImpl())
+    return nullptr;
+  if (getPosWidth())
+    return IntegerType::get(getContext(), getPosWidth());
+  return IndexType::get(getContext());
+}
+
+MemRefType SparseTensorEncodingAttr::getCrdMemRefType(
+    std::optional<ArrayRef<int64_t>> dimShape) const {
+  SmallVector<Size> shape = getSparseFieldShape(*this, dimShape);
+  return MemRefType::get(shape, getCrdElemType());
+}
+
+MemRefType SparseTensorEncodingAttr::getPosMemRefType(
+    std::optional<ArrayRef<int64_t>> dimShape) const {
+  SmallVector<Size> shape = getSparseFieldShape(*this, dimShape);
+  return MemRefType::get(shape, getPosElemType());
+}
+
 bool SparseTensorEncodingAttr::isIdentity() const {
   return !getImpl() || !getDimToLvl() || getDimToLvl().isIdentity();
 }

From 877ecdf52f40472d31205f107d39097268d3e3e3 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 20 Mar 2024 21:16:59 -0500
Subject: [PATCH 165/886] [ValueTracking] Add tests for computing knownbits
 from `(icmp upred X (and/or X, Y))`; NFC

---
 .../test/Transforms/InstCombine/known-bits.ll | 200 ++++++++++++++++++
 1 file changed, 200 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index d210b19bb7faf..30c18cebb2a3b 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -543,6 +543,206 @@ define i8 @and_eq_bits_must_be_set(i8 %x, i8 %y) {
   ret i8 %r
 }
 
+define i8 @test_icmp_or(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_or(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_OR:%.*]] = or i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[N_OR]], 32
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_or = or i8 %n, %n2
+  %cmp = icmp ult i8 %n_or, 32
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
+define i8 @test_icmp_or2(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_or2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_OR:%.*]] = or i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i8 [[N_OR]], 14
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+; CHECK:       if.else:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+;
+entry:
+  %n_or = or i8 %n, %n2
+  %cmp = icmp uge i8 %n_or, 15
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  ret i8 %other
+if.else:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+}
+
+define i8 @test_icmp_or_fail_bad_range(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_or_fail_bad_range(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_OR:%.*]] = or i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[N_OR]], 33
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_or = or i8 %n, %n2
+  %cmp = icmp ule i8 %n_or, 32
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
+define i8 @test_icmp_or_fail_bad_pred(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_or_fail_bad_pred(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_OR:%.*]] = or i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i8 [[N_OR]], 32
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_or = or i8 %n, %n2
+  %cmp = icmp ugt i8 %n_or, 32
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
+define i8 @test_icmp_and(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_and(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_AND:%.*]] = and i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i8 [[N_AND]], -33
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_and = and i8 %n, %n2
+  %cmp = icmp ugt i8 %n_and, 223
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
+define i8 @test_icmp_and2(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_and2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_AND:%.*]] = and i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[N_AND]], -31
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+; CHECK:       if.else:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+;
+entry:
+  %n_and = and i8 %n, %n2
+  %cmp = icmp ule i8 %n_and, 224
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  ret i8 %other
+if.else:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+}
+
+define i8 @test_icmp_and_fail_bad_range(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_and_fail_bad_range(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_AND:%.*]] = and i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i8 [[N_AND]], -34
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_and = and i8 %n, %n2
+  %cmp = icmp uge i8 %n_and, 223
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
+define i8 @test_icmp_and_fail_bad_pred(i8 %n, i8 %n2, i8 %other) {
+; CHECK-LABEL: @test_icmp_and_fail_bad_pred(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[N_AND:%.*]] = and i8 [[N:%.*]], [[N2:%.*]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i8 [[N_AND]], 31
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+; CHECK:       if.else:
+; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
+;
+entry:
+  %n_and = and i8 %n, %n2
+  %cmp = icmp sge i8 %n_and, 32
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %r = and i8 %n, 32
+  ret i8 %r
+
+if.else:
+  ret i8 %other
+}
+
 define i8 @and_eq_bits_must_be_set2(i8 %x, i8 %y) {
 ; CHECK-LABEL: @and_eq_bits_must_be_set2(
 ; CHECK-NEXT:    [[XY:%.*]] = and i8 [[X:%.*]], [[Y:%.*]]

From b8659600c3af6e5f1a5dd51ac59d6ab2554603d8 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 20 Mar 2024 20:15:47 -0500
Subject: [PATCH 166/886] [ValueTracking] compute knownbits from `(icmp upred X
 (and/or X, Y))`; NFC

`(icmp uge/ugt (and X, Y), C)` implies both `(icmp uge/ugt X, C)` and
`(icmp uge/ugt Y, C)`. We can use this to deduce leading ones in `X`.

`(icmp ule/ult (or X, Y), C)` implies both `(icmp ule/ult X, C)` and
`(icmp ule/ult Y, C)`. We can use this to deduce leading zeros in `X`.

Closes #86059
---
 llvm/lib/Analysis/ValueTracking.cpp           | 48 ++++++++++++++-----
 .../test/Transforms/InstCombine/known-bits.ll | 12 ++---
 2 files changed, 40 insertions(+), 20 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 3a10de72a2756..eb8925c534753 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -701,13 +701,26 @@ static void computeKnownBitsFromCmp(const Value *V, CmpInst::Predicate Pred,
     break;
   }
   default:
-    const APInt *Offset = nullptr;
-    if (match(LHS, m_CombineOr(m_V, m_AddLike(m_V, m_APInt(Offset)))) &&
-        match(RHS, m_APInt(C))) {
-      ConstantRange LHSRange = ConstantRange::makeAllowedICmpRegion(Pred, *C);
-      if (Offset)
-        LHSRange = LHSRange.sub(*Offset);
-      Known = Known.unionWith(LHSRange.toKnownBits());
+    if (match(RHS, m_APInt(C))) {
+      const APInt *Offset = nullptr;
+      if (match(LHS, m_CombineOr(m_V, m_AddLike(m_V, m_APInt(Offset))))) {
+        ConstantRange LHSRange = ConstantRange::makeAllowedICmpRegion(Pred, *C);
+        if (Offset)
+          LHSRange = LHSRange.sub(*Offset);
+        Known = Known.unionWith(LHSRange.toKnownBits());
+      }
+      // X & Y u> C -> X u> C && Y u> C
+      if ((Pred == ICmpInst::ICMP_UGT || Pred == ICmpInst::ICMP_UGE) &&
+          match(LHS, m_c_And(m_V, m_Value()))) {
+        Known.One.setHighBits(
+            (*C + (Pred == ICmpInst::ICMP_UGT)).countLeadingOnes());
+      }
+      // X | Y u< C -> X u< C && Y u< C
+      if ((Pred == ICmpInst::ICMP_ULT || Pred == ICmpInst::ICMP_ULE) &&
+          match(LHS, m_c_Or(m_V, m_Value()))) {
+        Known.Zero.setHighBits(
+            (*C - (Pred == ICmpInst::ICMP_ULT)).countLeadingZeros());
+      }
     }
     break;
   }
@@ -9448,11 +9461,22 @@ void llvm::findValuesAffectedByCondition(
           }
         }
       } else {
-        // Handle (A + C1) u< C2, which is the canonical form of
-        // A > C3 && A < C4.
-        if (match(A, m_AddLike(m_Value(X), m_ConstantInt())) &&
-            match(B, m_ConstantInt()))
-          AddAffected(X);
+        if (match(B, m_ConstantInt())) {
+          // Handle (A + C1) u< C2, which is the canonical form of
+          // A > C3 && A < C4.
+          if (match(A, m_AddLike(m_Value(X), m_ConstantInt())))
+            AddAffected(X);
+
+          Value *Y;
+          // X & Y u> C -> X >u C && Y >u C
+          // X | Y u< C -> X u< C && Y u< C
+          if (ICmpInst::isUnsigned(Pred) &&
+              (match(A, m_And(m_Value(X), m_Value(Y))) ||
+               match(A, m_Or(m_Value(X), m_Value(Y))))) {
+            AddAffected(X);
+            AddAffected(Y);
+          }
+        }
 
         // Handle icmp slt/sgt (bitcast X to int), 0/-1, which is supported
         // by computeKnownFPClass().
diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index 30c18cebb2a3b..fea2f647ee2a8 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -550,8 +550,7 @@ define i8 @test_icmp_or(i8 %n, i8 %n2, i8 %other) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[N_OR]], 32
 ; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ; CHECK:       if.else:
 ; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
 ;
@@ -577,8 +576,7 @@ define i8 @test_icmp_or2(i8 %n, i8 %n2, i8 %other) {
 ; CHECK:       if.then:
 ; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
 ; CHECK:       if.else:
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
 entry:
   %n_or = or i8 %n, %n2
@@ -650,8 +648,7 @@ define i8 @test_icmp_and(i8 %n, i8 %n2, i8 %other) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i8 [[N_AND]], -33
 ; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 32
 ; CHECK:       if.else:
 ; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
 ;
@@ -677,8 +674,7 @@ define i8 @test_icmp_and2(i8 %n, i8 %n2, i8 %other) {
 ; CHECK:       if.then:
 ; CHECK-NEXT:    ret i8 [[OTHER:%.*]]
 ; CHECK:       if.else:
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[N]], 32
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 32
 ;
 entry:
   %n_and = and i8 %n, %n2

From f0ea888e01aabcb131a8931b9e1fe1c5212a6cba Mon Sep 17 00:00:00 2001
From: Christopher Di Bella <cjdb@google.com>
Date: Thu, 11 Apr 2024 10:34:54 -0700
Subject: [PATCH 167/886] [libcxx] applies changes regarding post-commit
 feedback to #75259 (#76534)

Some of the feedback was also relevant to other files, and has been
applied there too.
---
 .../alg.fold/left_folds.pass.cpp              | 21 +++++++++++++------
 .../alg.fold/requirements.compile.pass.cpp    | 17 +++++++++++++--
 .../in_found_result.pass.cpp                  | 12 ++++++++---
 .../algorithms.results/in_fun_result.pass.cpp |  7 ++++++-
 .../in_in_out_result.pass.cpp                 |  5 +++++
 .../algorithms.results/in_in_result.pass.cpp  |  5 +++++
 .../in_out_out_result.pass.cpp                |  5 +++++
 .../algorithms.results/in_out_result.pass.cpp |  5 +++++
 .../in_value_result.pass.cpp                  |  7 ++++++-
 .../min_max_result.pass.cpp                   |  5 +++++
 10 files changed, 76 insertions(+), 13 deletions(-)

diff --git a/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/left_folds.pass.cpp b/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/left_folds.pass.cpp
index b8da26a229fa9..4987ca9cac4ae 100644
--- a/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/left_folds.pass.cpp
+++ b/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/left_folds.pass.cpp
@@ -59,14 +59,14 @@ concept is_in_value_result =
 template <class Result, class T>
 concept is_dangling_with = std::same_as<Result, std::ranges::fold_left_with_iter_result<std::ranges::dangling, T>>;
 
-struct Long {
+struct Integer {
   int value;
 
-  constexpr Long(int const x) : value(x) {}
+  constexpr Integer(int const x) : value(x) {}
 
-  constexpr Long plus(int const x) const { return Long{value + x}; }
+  constexpr Integer plus(int const x) const { return Integer{value + x}; }
 
-  friend constexpr bool operator==(Long const& x, Long const& y) = default;
+  friend constexpr bool operator==(Integer const& x, Integer const& y) = default;
 };
 
 template <std::ranges::input_range R, class T, class F, std::equality_comparable Expected>
@@ -77,6 +77,7 @@ constexpr void check_iterator(R& r, T const& init, F f, Expected const& expected
     assert(result.in == r.end());
     assert(result.value == expected);
   }
+
   {
     auto telemetry                                        = invocable_telemetry();
     auto f2                                               = invocable_with_telemetry(f, telemetry);
@@ -92,6 +93,7 @@ constexpr void check_iterator(R& r, T const& init, F f, Expected const& expected
     std::same_as<Expected> decltype(auto) result = fold_left(r.begin(), r.end(), init, f);
     assert(result == expected);
   }
+
   {
     auto telemetry                               = invocable_telemetry();
     auto f2                                      = invocable_with_telemetry(f, telemetry);
@@ -111,6 +113,7 @@ constexpr void check_lvalue_range(R& r, T const& init, F f, Expected const& expe
     assert(result.in == r.end());
     assert(result.value == expected);
   }
+
   {
     auto telemetry                               = invocable_telemetry();
     auto f2                                      = invocable_with_telemetry(f, telemetry);
@@ -125,6 +128,7 @@ constexpr void check_lvalue_range(R& r, T const& init, F f, Expected const& expe
     std::same_as<Expected> decltype(auto) result = fold_left(r, init, f);
     assert(result == expected);
   }
+
   {
     auto telemetry                               = invocable_telemetry();
     auto f2                                      = invocable_with_telemetry(f, telemetry);
@@ -144,6 +148,7 @@ constexpr void check_rvalue_range(R& r, T const& init, F f, Expected const& expe
     is_dangling_with<Expected> decltype(auto) result = fold_left_with_iter(std::move(r2), init, f);
     assert(result.value == expected);
   }
+
   {
     auto telemetry                                   = invocable_telemetry();
     auto f2                                          = invocable_with_telemetry(f, telemetry);
@@ -160,6 +165,7 @@ constexpr void check_rvalue_range(R& r, T const& init, F f, Expected const& expe
     std::same_as<Expected> decltype(auto) result = fold_left(std::move(r2), init, f);
     assert(result == expected);
   }
+
   {
     auto telemetry                               = invocable_telemetry();
     auto f2                                      = invocable_with_telemetry(f, telemetry);
@@ -186,7 +192,7 @@ constexpr void empty_range_test_case() {
   check(data, -100, std::multiplies(), -100);
 
   check(data | std::views::take_while([](auto) { return false; }), 1.23, std::plus(), 1.23);
-  check(data, Long(52), &Long::plus, Long(52));
+  check(data, Integer(52), &Integer::plus, Integer(52));
 }
 
 constexpr void common_range_test_case() {
@@ -209,7 +215,7 @@ constexpr void common_range_test_case() {
   };
   check(data, 0, fib, fibonacci(data.back()));
 
-  check(data, Long(0), &Long::plus, Long(triangular_sum(data)));
+  check(data, Integer(0), &Integer::plus, Integer(triangular_sum(data)));
 }
 
 constexpr void non_common_range_test_case() {
@@ -269,6 +275,7 @@ void runtime_only_test_case() {
       assert(result.in == data.end());
       assert(result.value == expected);
     }
+
     {
       auto input = std::istringstream(raw_data);
       auto data  = std::views::istream<std::string>(input);
@@ -277,11 +284,13 @@ void runtime_only_test_case() {
       assert(result.in == data.end());
       assert(result.value == expected);
     }
+
     {
       auto input = std::istringstream(raw_data);
       auto data  = std::views::istream<std::string>(input);
       assert(fold_left(data.begin(), data.end(), init, std::plus()) == expected);
     }
+
     {
       auto input = std::istringstream(raw_data);
       auto data  = std::views::istream<std::string>(input);
diff --git a/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/requirements.compile.pass.cpp b/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/requirements.compile.pass.cpp
index cad96d4c10127..261dd6e572e50 100644
--- a/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/requirements.compile.pass.cpp
+++ b/libcxx/test/std/algorithms/alg.nonmodifying/alg.fold/requirements.compile.pass.cpp
@@ -10,8 +10,21 @@
 
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20
 
-// Checks that `std::ranges::fold_left_with_iter`'s requirements reject parameters that don't meet
-// the overloads' constraints.
+// template<input_iterator I, sentinel_for<I> S, class T,
+//          indirectly-binary-left-foldable<T, I> F>
+//   constexpr see below ranges::fold_left_with_iter(I first, S last, T init, F f);
+//
+// template<input_range R, class T, indirectly-binary-left-foldable<T, iterator_t<R>> F>
+//   constexpr see below ranges::fold_left_with_iter(R&& r, T init, F f);
+
+// template<input_iterator I, sentinel_for<I> S, class T,
+//          indirectly-binary-left-foldable<T, I> F>
+//   constexpr see below ranges::fold_left(I first, S last, T init, F f);
+//
+// template<input_range R, class T, indirectly-binary-left-foldable<T, iterator_t<R>> F>
+//   constexpr see below ranges::fold_left(R&& r, T init, F f);
+
+// Checks that the algorithm requirements reject parameters that don't meet the overloads' constraints.
 
 #include <algorithm>
 #include <concepts>
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_found_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_found_result.pass.cpp
index af23a1b051c89..ec104df5feb47 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_found_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_found_result.pass.cpp
@@ -64,6 +64,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_found_result<double> res{10, true};
     assert(res.in == 10);
@@ -72,6 +73,8 @@ constexpr bool test() {
     assert(res2.in.content == 10);
     assert(res2.found == true);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_found_result<MoveOnly> res{MoveOnly{}, false};
     assert(res.in.get() == 1);
@@ -82,10 +85,13 @@ constexpr bool test() {
     assert(res.in.get() == 0);
     assert(!res.found);
   }
-  auto [in, found] = std::ranges::in_found_result<int>{2, false};
-  assert(in == 2);
-  assert(!found);
 
+  // Checks that structured bindings get the correct values.
+  {
+    auto [in, found] = std::ranges::in_found_result<int>{2, false};
+    assert(in == 2);
+    assert(!found);
+  }
   return true;
 }
 
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_fun_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_fun_result.pass.cpp
index f44b5988d849a..4395380ed179a 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_fun_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_fun_result.pass.cpp
@@ -48,7 +48,7 @@ static_assert(std::is_convertible_v<const std::ranges::in_fun_result<int, int>&&
 static_assert(std::is_move_constructible_v<std::ranges::in_fun_result<MoveOnly, int>>);
 static_assert(std::is_move_constructible_v<std::ranges::in_fun_result<int, MoveOnly>>);
 
-// should not copy constructible with move-only type
+// should not be copy constructible with move-only type
 static_assert(!std::is_copy_constructible_v<std::ranges::in_fun_result<MoveOnly, int>>);
 static_assert(!std::is_copy_constructible_v<std::ranges::in_fun_result<int, MoveOnly>>);
 
@@ -64,6 +64,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_fun_result<int, double> res{10, 0.};
     assert(res.in == 10);
@@ -72,6 +73,8 @@ constexpr bool test() {
     assert(res2.in.content == 10);
     assert(res2.fun.content == 0.);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_fun_result<MoveOnly, int> res{MoveOnly{}, 2};
     assert(res.in.get() == 1);
@@ -81,6 +84,8 @@ constexpr bool test() {
     assert(res2.in.get() == 1);
     assert(res2.fun == 2);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [in, fun] = std::ranges::in_fun_result<int, int>{1, 2};
     assert(in == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_in_out_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_in_out_result.pass.cpp
index c9ab2d430b500..0ee09ceb47416 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_in_out_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_in_out_result.pass.cpp
@@ -67,6 +67,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_in_out_result<int, double, float> res{10, 0., 1.f};
     assert(res.in1 == 10);
@@ -77,6 +78,8 @@ constexpr bool test() {
     assert(res2.in2.content == 0.);
     assert(res2.out.content == 1.f);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_in_out_result<MoveOnly, int, int> res1{MoveOnly{}, 0, 0};
     assert(res1.in1.get() == 1);
@@ -84,6 +87,8 @@ constexpr bool test() {
     assert(res1.in1.get() == 0);
     assert(res2.in1.get() == 1);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [in1, in2, out] = std::ranges::in_in_out_result<int, int, int>{1, 2, 3};
     assert(in1 == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_in_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_in_result.pass.cpp
index 64d1106117329..10e6e88327fe1 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_in_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_in_result.pass.cpp
@@ -59,6 +59,7 @@ struct ConstructibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_in_result<int, double> res{10L, 0.};
     assert(res.in1 == 10);
@@ -67,12 +68,16 @@ constexpr bool test() {
     assert(res2.in1.content == 10);
     assert(res2.in2.content == 0.);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_in_result<MoveOnly, int> res{MoveOnly{}, 0};
     assert(res.in1.get() == 1);
     [[maybe_unused]] auto res2 = static_cast<std::ranges::in_in_result<MoveOnly, int>>(std::move(res));
     assert(res.in1.get() == 0);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [in1, in2] = std::ranges::in_in_result<int, int>{1, 2};
     assert(in1 == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_out_out_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_out_out_result.pass.cpp
index 682462bc9a5ba..875080b1900e0 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_out_out_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_out_out_result.pass.cpp
@@ -65,6 +65,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_out_out_result<int, double, float> res{10, 0., 1.f};
     assert(res.in == 10);
@@ -75,6 +76,8 @@ constexpr bool test() {
     assert(res2.out1.content == 0.);
     assert(res2.out2.content == 1.f);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_out_out_result<MoveOnly, int, int> res1{MoveOnly{}, 0, 0};
     assert(res1.in.get() == 1);
@@ -82,6 +85,8 @@ constexpr bool test() {
     assert(res1.in.get() == 0);
     assert(res2.in.get() == 1);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [in, out1, out2] = std::ranges::in_out_out_result<int, int, int>{1, 2, 3};
     assert(in == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_out_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_out_result.pass.cpp
index ca9f4464dee5f..9cfb5f43955cb 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_out_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_out_result.pass.cpp
@@ -59,6 +59,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_out_result<double, int> res{10, 1};
     assert(res.in == 10);
@@ -67,6 +68,8 @@ constexpr bool test() {
     assert(res2.in.content == 10);
     assert(res2.out.content == 1);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_out_result<MoveOnly, int> res{MoveOnly{}, 10};
     assert(res.in.get() == 1);
@@ -77,6 +80,8 @@ constexpr bool test() {
     assert(res2.in.get() == 1);
     assert(res2.out == 10);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [min, max] = std::ranges::in_out_result<int, int>{1, 2};
     assert(min == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/in_value_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/in_value_result.pass.cpp
index f4023b89dce70..b7aba0c88ee95 100644
--- a/libcxx/test/std/algorithms/algorithms.results/in_value_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/in_value_result.pass.cpp
@@ -53,7 +53,7 @@ static_assert(
 static_assert(std::is_move_constructible_v<std::ranges::in_value_result<MoveOnly, int>>);
 static_assert(std::is_move_constructible_v<std::ranges::in_value_result<int, MoveOnly>>);
 
-// should not copy constructible with move-only type
+// should not be copy constructible with move-only type
 static_assert(!std::is_copy_constructible_v<std::ranges::in_value_result<MoveOnly, int>>);
 static_assert(!std::is_copy_constructible_v<std::ranges::in_value_result<int, MoveOnly>>);
 
@@ -71,6 +71,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::in_value_result<int, double> res{10, 0.};
     assert(res.in == 10);
@@ -79,6 +80,8 @@ constexpr bool test() {
     assert(res2.in.content == 10);
     assert(res2.value.content == 0.);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::in_value_result<MoveOnly, int> res{MoveOnly{}, 2};
     assert(res.in.get() == 1);
@@ -88,6 +91,8 @@ constexpr bool test() {
     assert(res2.in.get() == 1);
     assert(res2.value == 2);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [in, value] = std::ranges::in_value_result<int, int>{1, 2};
     assert(in == 1);
diff --git a/libcxx/test/std/algorithms/algorithms.results/min_max_result.pass.cpp b/libcxx/test/std/algorithms/algorithms.results/min_max_result.pass.cpp
index 23c22ec9d3924..bc394926c9548 100644
--- a/libcxx/test/std/algorithms/algorithms.results/min_max_result.pass.cpp
+++ b/libcxx/test/std/algorithms/algorithms.results/min_max_result.pass.cpp
@@ -60,6 +60,7 @@ struct ConvertibleFrom {
 };
 
 constexpr bool test() {
+  // Checks that conversion operations are correct.
   {
     std::ranges::min_max_result<double> res{10, 1};
     assert(res.min == 10);
@@ -68,6 +69,8 @@ constexpr bool test() {
     assert(res2.min.content == 10);
     assert(res2.max.content == 1);
   }
+
+  // Checks that conversions are possible when one of the types is move-only.
   {
     std::ranges::min_max_result<MoveOnly> res{MoveOnly{}, MoveOnly{}};
     assert(res.min.get() == 1);
@@ -78,6 +81,8 @@ constexpr bool test() {
     assert(res2.min.get() == 1);
     assert(res2.max.get() == 1);
   }
+
+  // Checks that structured bindings get the correct values.
   {
     auto [min, max] = std::ranges::min_max_result<int>{1, 2};
     assert(min == 1);

From f6749d8dcb9af5c30affd259f4c4160c54606d91 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 11 Apr 2024 09:59:30 -0700
Subject: [PATCH 168/886] [SLP]Consider unsigned nodes, feeding into sitofp,
 being converted using uitofp.

Need to use uitofp for unsigned nodes, which are part of minbitwidth
analysis, to correctly handle signedness info.
---
 llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp             | 6 ++++++
 .../Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 4ac719af60295..2995cb8435f76 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -9020,6 +9020,9 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
         VecOpcode =
             SrcIt->second.second ? Instruction::SExt : Instruction::ZExt;
       }
+    } else if (VecOpcode == Instruction::SIToFP && SrcIt != MinBWs.end() &&
+               !SrcIt->second.second) {
+      VecOpcode = Instruction::UIToFP;
     }
     auto GetScalarCost = [&](unsigned Idx) -> InstructionCost {
       auto *VI = cast<Instruction>(UniqueValues[Idx]);
@@ -12398,6 +12401,9 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
           VecOpcode =
               SrcIt->second.second ? Instruction::SExt : Instruction::ZExt;
         }
+      } else if (VecOpcode == Instruction::SIToFP && SrcIt != MinBWs.end() &&
+                 !SrcIt->second.second) {
+        VecOpcode = Instruction::UIToFP;
       }
       Value *V = (VecOpcode != ShuffleOrOp && VecOpcode == Instruction::BitCast)
                      ? InVec
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll b/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll
index 00dd9b7e6fb47..a0e17446b6abb 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/sitofp-minbitwidth-node.ll
@@ -8,7 +8,7 @@ define void @foo(ptr %ptr) {
 ; CHECK-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i8, ptr [[PTR]], i64 334
 ; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i16>, ptr [[GEP0]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i16> [[TMP1]], <i16 -1, i16 -1>
-; CHECK-NEXT:    [[TMP4:%.*]] = sitofp <2 x i16> [[TMP3]] to <2 x double>
+; CHECK-NEXT:    [[TMP4:%.*]] = uitofp <2 x i16> [[TMP3]] to <2 x double>
 ; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i16>, ptr [[GEP3]], align 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = zext <2 x i16> [[TMP5]] to <2 x i32>
 ; CHECK-NEXT:    [[TMP2:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i32>

From a417b9b759960cc4d7b49e95ddc2ef129b5abe12 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Thu, 11 Apr 2024 13:46:37 -0400
Subject: [PATCH 169/886] [lldb] Fix call to TemplateTemplateParmDecl::Create
 after #88139 (#88417)

#88139 adds a parameter to `TemplateTemplateParmDecl::Create`, which is
called in LLDB. This adds the missing argument for that parameter.
---
 lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
index ee634d12b3c41..44bd02bd4b367 100644
--- a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
+++ b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
@@ -1682,10 +1682,11 @@ TypeSystemClang::CreateTemplateTemplateParmDecl(const char *template_name) {
   // type that includes a template template argument. Only the name matters for
   // this purpose, so we use dummy values for the other characteristics of the
   // type.
-  return TemplateTemplateParmDecl::Create(
-      ast, decl_ctx, SourceLocation(),
-      /*Depth*/ 0, /*Position*/ 0,
-      /*IsParameterPack*/ false, &identifier_info, template_param_list);
+  return TemplateTemplateParmDecl::Create(ast, decl_ctx, SourceLocation(),
+                                          /*Depth=*/0, /*Position=*/0,
+                                          /*IsParameterPack=*/false,
+                                          &identifier_info, /*Typename=*/false,
+                                          template_param_list);
 }
 
 ClassTemplateSpecializationDecl *

From 03f619da077bca72bb3b0d978a1f50a2fade2750 Mon Sep 17 00:00:00 2001
From: Vincent Lee <thevinster@users.noreply.github.com>
Date: Thu, 11 Apr 2024 10:55:52 -0700
Subject: [PATCH 170/886] [StandardInstrumentations] Support
 -print-after-pass-number option (#87458)

There's already support for `-print-before-pass-number`, so it makes
sense that we also have a `-print-after-pass-number`. This is especially
useful if you want to print the IR after the very last pass without
resorting to `-print-after-all` and combing through stderr or the IR
file directory.
---
 .../llvm/Passes/StandardInstrumentations.h    |  5 +-
 llvm/lib/Passes/StandardInstrumentations.cpp  | 62 +++++++++++++------
 llvm/test/Other/print-at-pass-number.ll       | 14 +++++
 3 files changed, 62 insertions(+), 19 deletions(-)

diff --git a/llvm/include/llvm/Passes/StandardInstrumentations.h b/llvm/include/llvm/Passes/StandardInstrumentations.h
index b053e6307a653..84d1b541171bf 100644
--- a/llvm/include/llvm/Passes/StandardInstrumentations.h
+++ b/llvm/include/llvm/Passes/StandardInstrumentations.h
@@ -67,8 +67,11 @@ class PrintIRInstrumentation {
 
   bool shouldPrintBeforePass(StringRef PassID);
   bool shouldPrintAfterPass(StringRef PassID);
+  bool shouldPrintBeforeCurrentPassNumber();
+  bool shouldPrintAfterCurrentPassNumber();
   bool shouldPrintPassNumbers();
-  bool shouldPrintBeforePassNumber();
+  bool shouldPrintBeforeSomePassNumber();
+  bool shouldPrintAfterSomePassNumber();
 
   void pushPassRunDescriptor(StringRef PassID, Any IR,
                              std::string &DumpIRFilename);
diff --git a/llvm/lib/Passes/StandardInstrumentations.cpp b/llvm/lib/Passes/StandardInstrumentations.cpp
index c5f0c14885d0c..c18b462258623 100644
--- a/llvm/lib/Passes/StandardInstrumentations.cpp
+++ b/llvm/lib/Passes/StandardInstrumentations.cpp
@@ -125,6 +125,11 @@ static cl::opt<unsigned> PrintBeforePassNumber(
     cl::desc("Print IR before the pass with this number as "
              "reported by print-pass-numbers"));
 
+static cl::opt<unsigned>
+    PrintAfterPassNumber("print-after-pass-number", cl::init(0), cl::Hidden,
+                         cl::desc("Print IR after the pass with this number as "
+                                  "reported by print-pass-numbers"));
+
 static cl::opt<std::string> IRDumpDirectory(
     "ir-dump-directory",
     cl::desc("If specified, IR printed using the "
@@ -854,7 +859,9 @@ void PrintIRInstrumentation::printBeforePass(StringRef PassID, Any IR) {
 
   std::string DumpIRFilename;
   if (!IRDumpDirectory.empty() &&
-      (shouldPrintBeforePass(PassID) || shouldPrintAfterPass(PassID)))
+      (shouldPrintBeforePass(PassID) || shouldPrintAfterPass(PassID) ||
+       shouldPrintBeforeCurrentPassNumber() ||
+       shouldPrintAfterCurrentPassNumber()))
     DumpIRFilename = fetchDumpFilename(PassID, IR);
 
   // Saving Module for AfterPassInvalidated operations.
@@ -873,12 +880,15 @@ void PrintIRInstrumentation::printBeforePass(StringRef PassID, Any IR) {
     dbgs() << " Running pass " << CurrentPassNumber << " " << PassID
            << " on " << getIRName(IR) << "\n";
 
-  if (!shouldPrintBeforePass(PassID))
+  if (shouldPrintAfterCurrentPassNumber())
+    pushPassRunDescriptor(PassID, IR, DumpIRFilename);
+
+  if (!shouldPrintBeforePass(PassID) && !shouldPrintBeforeCurrentPassNumber())
     return;
 
   auto WriteIRToStream = [&](raw_ostream &Stream) {
     Stream << "; *** IR Dump Before ";
-    if (shouldPrintBeforePassNumber())
+    if (shouldPrintBeforeSomePassNumber())
       Stream << CurrentPassNumber << "-";
     Stream << PassID << " on " << getIRName(IR) << " ***\n";
     unwrapAndPrint(Stream, IR);
@@ -898,18 +908,21 @@ void PrintIRInstrumentation::printAfterPass(StringRef PassID, Any IR) {
   if (isIgnored(PassID))
     return;
 
-  if (!shouldPrintAfterPass(PassID))
+  if (!shouldPrintAfterPass(PassID) && !shouldPrintAfterCurrentPassNumber())
     return;
 
   auto [M, DumpIRFilename, IRName, StoredPassID] = popPassRunDescriptor(PassID);
   assert(StoredPassID == PassID && "mismatched PassID");
 
-  if (!shouldPrintIR(IR) || !shouldPrintAfterPass(PassID))
+  if (!shouldPrintIR(IR) ||
+      (!shouldPrintAfterPass(PassID) && !shouldPrintAfterCurrentPassNumber()))
     return;
 
   auto WriteIRToStream = [&](raw_ostream &Stream, const StringRef IRName) {
-    Stream << "; *** IR Dump " << StringRef(formatv("After {0}", PassID))
-           << " on " << IRName << " ***\n";
+    Stream << "; *** IR Dump After ";
+    if (shouldPrintAfterSomePassNumber())
+      Stream << CurrentPassNumber << "-";
+    Stream << StringRef(formatv("{0}", PassID)) << " on " << IRName << " ***\n";
     unwrapAndPrint(Stream, IR);
   };
 
@@ -931,14 +944,15 @@ void PrintIRInstrumentation::printAfterPassInvalidated(StringRef PassID) {
   if (isIgnored(PassID))
     return;
 
-  if (!shouldPrintAfterPass(PassID))
+  if (!shouldPrintAfterPass(PassID) && !shouldPrintAfterCurrentPassNumber())
     return;
 
   auto [M, DumpIRFilename, IRName, StoredPassID] = popPassRunDescriptor(PassID);
   assert(StoredPassID == PassID && "mismatched PassID");
   // Additional filtering (e.g. -filter-print-func) can lead to module
   // printing being skipped.
-  if (!M || !shouldPrintAfterPass(PassID))
+  if (!M ||
+      (!shouldPrintAfterPass(PassID) && !shouldPrintAfterCurrentPassNumber()))
     return;
 
   auto WriteIRToStream = [&](raw_ostream &Stream, const Module *M,
@@ -968,10 +982,6 @@ bool PrintIRInstrumentation::shouldPrintBeforePass(StringRef PassID) {
   if (shouldPrintBeforeAll())
     return true;
 
-  if (shouldPrintBeforePassNumber() &&
-      CurrentPassNumber == PrintBeforePassNumber)
-    return true;
-
   StringRef PassName = PIC->getPassNameForClassName(PassID);
   return is_contained(printBeforePasses(), PassName);
 }
@@ -984,26 +994,42 @@ bool PrintIRInstrumentation::shouldPrintAfterPass(StringRef PassID) {
   return is_contained(printAfterPasses(), PassName);
 }
 
+bool PrintIRInstrumentation::shouldPrintBeforeCurrentPassNumber() {
+  return shouldPrintBeforeSomePassNumber() &&
+         (CurrentPassNumber == PrintBeforePassNumber);
+}
+
+bool PrintIRInstrumentation::shouldPrintAfterCurrentPassNumber() {
+  return shouldPrintAfterSomePassNumber() &&
+         (CurrentPassNumber == PrintAfterPassNumber);
+}
+
 bool PrintIRInstrumentation::shouldPrintPassNumbers() {
   return PrintPassNumbers;
 }
 
-bool PrintIRInstrumentation::shouldPrintBeforePassNumber() {
+bool PrintIRInstrumentation::shouldPrintBeforeSomePassNumber() {
   return PrintBeforePassNumber > 0;
 }
 
+bool PrintIRInstrumentation::shouldPrintAfterSomePassNumber() {
+  return PrintAfterPassNumber > 0;
+}
+
 void PrintIRInstrumentation::registerCallbacks(
     PassInstrumentationCallbacks &PIC) {
   this->PIC = &PIC;
 
   // BeforePass callback is not just for printing, it also saves a Module
-  // for later use in AfterPassInvalidated.
-  if (shouldPrintPassNumbers() || shouldPrintBeforePassNumber() ||
-      shouldPrintBeforeSomePass() || shouldPrintAfterSomePass())
+  // for later use in AfterPassInvalidated and keeps tracks of the
+  // CurrentPassNumber.
+  if (shouldPrintPassNumbers() || shouldPrintBeforeSomePassNumber() ||
+      shouldPrintAfterSomePassNumber() || shouldPrintBeforeSomePass() ||
+      shouldPrintAfterSomePass())
     PIC.registerBeforeNonSkippedPassCallback(
         [this](StringRef P, Any IR) { this->printBeforePass(P, IR); });
 
-  if (shouldPrintAfterSomePass()) {
+  if (shouldPrintAfterSomePass() || shouldPrintAfterSomePassNumber()) {
     PIC.registerAfterPassCallback(
         [this](StringRef P, Any IR, const PreservedAnalyses &) {
           this->printAfterPass(P, IR);
diff --git a/llvm/test/Other/print-at-pass-number.ll b/llvm/test/Other/print-at-pass-number.ll
index b9c09a36ca1fa..770b3d703c098 100644
--- a/llvm/test/Other/print-at-pass-number.ll
+++ b/llvm/test/Other/print-at-pass-number.ll
@@ -1,9 +1,13 @@
 ; RUN: opt -passes="loop(indvars,loop-deletion,loop-unroll-full)" -print-pass-numbers -S -o /dev/null %s 2>&1 | FileCheck %s --check-prefix=NUMBER
+; RUN: opt -passes="loop(indvars,loop-deletion,loop-unroll-full)" -print-pass-numbers -filter-print-funcs=baz -S -o /dev/null %s 2>&1 | FileCheck %s --check-prefix=NUMBER-FILTERED
 ; RUN: opt -passes="loop(indvars,loop-deletion,loop-unroll-full)" -print-module-scope -print-before-pass-number=3 -S -o /dev/null %s 2>&1 | FileCheck %s --check-prefix=BEFORE
+; RUN: opt -passes="loop(indvars,loop-deletion,loop-unroll-full)" -print-module-scope -print-after-pass-number=2 -S -o /dev/null %s 2>&1 | FileCheck %s --check-prefix=AFTER
 
 define i32 @bar(i32 %arg) {
 ; BEFORE: *** IR Dump Before 3-IndVarSimplifyPass on bb1 ***
 ; BEFORE: define i32 @bar(i32 %arg) {
+; AFTER:  *** IR Dump After 2-LCSSAPass on bar ***
+; AFTER:  define i32 @bar(i32 %arg) {
 
 bb:
   br label %bb1
@@ -20,8 +24,18 @@ bb4:                                              ; preds = %bb1
   ret i32 %add3
 }
 
+define i32 @baz(i32 %arg) {
+  ret i32 0;
+}
+
 ; NUMBER:  Running pass 1 LoopSimplifyPass on bar
 ; NUMBER-NEXT: Running pass 2 LCSSAPass on bar
 ; NUMBER-NEXT: Running pass 3 IndVarSimplifyPass on bb1
 ; NUMBER-NEXT: Running pass 4 LoopDeletionPass on bb1
+; NUMBER-NEXT: Running pass 5 LoopSimplifyPass on baz
+; NUMBER-NEXT: Running pass 6 LCSSAPass on baz
 ; NUMBER-NOT: Running pass
+
+; NUMBER-FILTERED: Running pass 1 LoopSimplifyPass on baz
+; NUMBER-FILTERED-NEXT: Running pass 2 LCSSAPass on baz
+

From 9029e6ebdfd98a58b5c5646fd534c6c849148cda Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Thu, 11 Apr 2024 17:09:07 +0100
Subject: [PATCH 171/886] [libclc] Refactor build system to allow in-tree
 builds (#87622)

The previous build system was adding custom "OpenCL" and "LLVM IR"
languages in CMake to build the builtin libraries. This was making it
harder to build in-tree because the tool binaries needed to be present
at configure time.

This commit refactors the build system to use custom commands to build
the bytecode files one by one, and link them all together into the final
bytecode library. It also enables in-tree builds by aliasing the
clang/llvm-link/etc. tool targets to internal targets, which are
imported from the LLVM installation directory when building out of tree.

Diffing (with llvm-diff) all of the final bytecode libraries in an
out-of-tree configuration against those built using the current tip
system shows no changes. Note that there are textual changes to metadata
IDs which confuse regular diff, and that llvm-diff 14 and below may show
false-positives.

This commit also removes a file listed in one of the SOURCEs which
didn't exist and which was preventing the use of
ENABLE_RUNTIME_SUBNORMAL when configuring CMake.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 51 +++++++++++++++++--
 .../X86/store-abs-minbitwidth.ll              | 14 +----
 2 files changed, 50 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 2995cb8435f76..55dc0867e2ea9 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -305,7 +305,32 @@ static bool isCommutative(Instruction *I) {
   if (auto *Cmp = dyn_cast<CmpInst>(I))
     return Cmp->isCommutative();
   if (auto *BO = dyn_cast<BinaryOperator>(I))
-    return BO->isCommutative();
+    return BO->isCommutative() ||
+           (BO->getOpcode() == Instruction::Sub &&
+            !BO->hasNUsesOrMore(UsesLimit) &&
+            all_of(
+                BO->uses(),
+                [](const Use &U) {
+                  // Commutative, if icmp eq/ne sub, 0
+                  ICmpInst::Predicate Pred;
+                  if (match(U.getUser(),
+                            m_ICmp(Pred, m_Specific(U.get()), m_Zero())) &&
+                      (Pred == ICmpInst::ICMP_EQ || Pred == ICmpInst::ICMP_NE))
+                    return true;
+                  // Commutative, if abs(sub nsw, true) or abs(sub, false).
+                  ConstantInt *Flag;
+                  return match(U.getUser(),
+                               m_Intrinsic<Intrinsic::abs>(
+                                   m_Specific(U.get()), m_ConstantInt(Flag))) &&
+                         (!cast<Instruction>(U.get())->hasNoSignedWrap() ||
+                          Flag->isOne());
+                })) ||
+           (BO->getOpcode() == Instruction::FSub &&
+            !BO->hasNUsesOrMore(UsesLimit) &&
+            all_of(BO->uses(), [](const Use &U) {
+              return match(U.getUser(),
+                           m_Intrinsic<Intrinsic::fabs>(m_Specific(U.get())));
+            }));
   return I->isCommutative();
 }
 
@@ -6838,7 +6863,7 @@ void BoUpSLP::buildTree_rec(ArrayRef<Value *> VL, unsigned Depth,
 
       // Sort operands of the instructions so that each side is more likely to
       // have the same opcode.
-      if (isa<BinaryOperator>(VL0) && VL0->isCommutative()) {
+      if (isa<BinaryOperator>(VL0) && isCommutative(VL0)) {
         ValueList Left, Right;
         reorderInputsAccordingToOpcode(VL, Left, Right, *this);
         TE->setOperand(0, Left);
@@ -12566,8 +12591,15 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
           static_cast<Instruction::BinaryOps>(E->getOpcode()), LHS,
           RHS);
       propagateIRFlags(V, E->Scalars, VL0, It == MinBWs.end());
-      if (auto *I = dyn_cast<Instruction>(V))
+      if (auto *I = dyn_cast<Instruction>(V)) {
         V = propagateMetadata(I, E->Scalars);
+        // Drop nuw flags for abs(sub(commutative), true).
+        if (!MinBWs.contains(E) && ShuffleOrOp == Instruction::Sub &&
+            any_of(E->Scalars, [](Value *V) {
+              return isCommutative(cast<Instruction>(V));
+            }))
+          I->setHasNoUnsignedWrap(/*b=*/false);
+      }
 
       V = FinalShuffle(V, E, VecTy);
 
@@ -12893,6 +12925,19 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
 
       propagateIRFlags(V0, OpScalars, E->getMainOp(), It == MinBWs.end());
       propagateIRFlags(V1, AltScalars, E->getAltOp(), It == MinBWs.end());
+      auto DropNuwFlag = [&](Value *Vec, unsigned Opcode) {
+        // Drop nuw flags for abs(sub(commutative), true).
+        if (auto *I = dyn_cast<Instruction>(Vec);
+            I && Opcode == Instruction::Sub && !MinBWs.contains(E) &&
+            any_of(E->Scalars, [](Value *V) {
+              auto *I = cast<Instruction>(V);
+              return I->getOpcode() == Instruction::Sub &&
+                     isCommutative(cast<Instruction>(V));
+            }))
+          I->setHasNoUnsignedWrap(/*b=*/false);
+      };
+      DropNuwFlag(V0, E->getOpcode());
+      DropNuwFlag(V1, E->getAltOpcode());
 
       Value *V = Builder.CreateShuffleVector(V0, V1, Mask);
       if (auto *I = dyn_cast<Instruction>(V)) {
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
index df7312e3d2b56..64bfb242752c1 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
@@ -4,19 +4,9 @@
 
 define i32 @test(ptr noalias %in, ptr noalias %inn, ptr %out) {
 ; CHECK-LABEL: @test(
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i8>, ptr [[IN:%.*]], align 1
-; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i8, ptr [[IN]], i64 2
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i8>, ptr [[GEP_2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i8>, ptr [[INN:%.*]], align 1
-; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i8, ptr [[INN]], i64 2
-; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i8>, ptr [[GEP_5]], align 1
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i8> [[TMP2]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i8> [[TMP5]], <4 x i8> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i8>, ptr [[IN:%.*]], align 1
+; CHECK-NEXT:    [[TMP11:%.*]] = load <4 x i8>, ptr [[INN:%.*]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i8> [[TMP7]] to <4 x i16>
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x i8> [[TMP1]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x i8> [[TMP4]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i8> [[TMP9]], <4 x i8> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
 ; CHECK-NEXT:    [[TMP12:%.*]] = sext <4 x i8> [[TMP11]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP13:%.*]] = sub <4 x i16> [[TMP12]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = call <4 x i16> @llvm.abs.v4i16(<4 x i16> [[TMP13]], i1 false)

From d7be9d23012c99b231bd3a4a9dd58ee15f1be5e7 Mon Sep 17 00:00:00 2001
From: Sunil Srivastava <sunil.srivastava@sony.com>
Date: Thu, 11 Apr 2024 11:04:44 -0700
Subject: [PATCH 172/886] PS support for erf/erff/erfl. (#88412)

Co-authored-by: Sunil Srivastava <sunil.srivastava@sony.com>
---
 llvm/lib/Analysis/TargetLibraryInfo.cpp       |  3 ---
 .../tools/llvm-tli-checker/ps4-tli-check.yaml | 22 ++++++++++++++-----
 2 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Analysis/TargetLibraryInfo.cpp b/llvm/lib/Analysis/TargetLibraryInfo.cpp
index 9e17dcaa55925..c62d9daa13ef0 100644
--- a/llvm/lib/Analysis/TargetLibraryInfo.cpp
+++ b/llvm/lib/Analysis/TargetLibraryInfo.cpp
@@ -813,9 +813,6 @@ static void initialize(TargetLibraryInfoImpl &TLI, const Triple &T,
     TLI.setUnavailable(LibFunc_cabs);
     TLI.setUnavailable(LibFunc_cabsf);
     TLI.setUnavailable(LibFunc_cabsl);
-    TLI.setUnavailable(LibFunc_erf);
-    TLI.setUnavailable(LibFunc_erff);
-    TLI.setUnavailable(LibFunc_erfl);
     TLI.setUnavailable(LibFunc_ffs);
     TLI.setUnavailable(LibFunc_flockfile);
     TLI.setUnavailable(LibFunc_fseeko);
diff --git a/llvm/test/tools/llvm-tli-checker/ps4-tli-check.yaml b/llvm/test/tools/llvm-tli-checker/ps4-tli-check.yaml
index 61caf727b0f96..46f6a2d0a554a 100644
--- a/llvm/test/tools/llvm-tli-checker/ps4-tli-check.yaml
+++ b/llvm/test/tools/llvm-tli-checker/ps4-tli-check.yaml
@@ -34,23 +34,23 @@
 #
 # CHECK: << Total TLI yes SDK no:  8
 # CHECK: >> Total TLI no  SDK yes: 0
-# CHECK: == Total TLI yes SDK yes: 235
+# CHECK: == Total TLI yes SDK yes: 238
 #
 # WRONG_DETAIL: << TLI yes SDK no : '_ZdaPv' aka operator delete[](void*)
 # WRONG_DETAIL: >> TLI no  SDK yes: '_ZdaPvj' aka operator delete[](void*, unsigned int)
 # WRONG_DETAIL-COUNT-8: << TLI yes SDK no : {{.*}}__hot_cold_t
 # WRONG_SUMMARY: << Total TLI yes SDK no:  9{{$}}
 # WRONG_SUMMARY: >> Total TLI no  SDK yes: 1{{$}}
-# WRONG_SUMMARY: == Total TLI yes SDK yes: 234
+# WRONG_SUMMARY: == Total TLI yes SDK yes: 237
 #
 ## The -COUNT suffix doesn't care if there are too many matches, so check
 ## the exact count first; the two directives should add up to that.
 ## Yes, this means additions to TLI will fail this test, but the argument
 ## to -COUNT can't be an expression.
-# AVAIL: TLI knows 479 symbols, 243 available
-# AVAIL-COUNT-243: {{^}} available
+# AVAIL: TLI knows 479 symbols, 246 available
+# AVAIL-COUNT-246: {{^}} available
 # AVAIL-NOT:       {{^}} available
-# UNAVAIL-COUNT-236: not available
+# UNAVAIL-COUNT-233: not available
 # UNAVAIL-NOT:       not available
 
 ## This is a large file so it's worth telling lit to stop here.
@@ -347,6 +347,18 @@ DynamicSymbols:
     Type:            STT_FUNC
     Section:         .text
     Binding:         STB_GLOBAL
+  - Name:            erf
+    Type:            STT_FUNC
+    Section:         .text
+    Binding:         STB_GLOBAL
+  - Name:            erff
+    Type:            STT_FUNC
+    Section:         .text
+    Binding:         STB_GLOBAL
+  - Name:            erfl
+    Type:            STT_FUNC
+    Section:         .text
+    Binding:         STB_GLOBAL
   - Name:            exp
     Type:            STT_FUNC
     Section:         .text

From f7cc224044a5121191b7d96825988bd954ce49e2 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 11 Apr 2024 11:08:47 -0700
Subject: [PATCH 173/886] Revert "[libclc] Refactor build system to allow
 in-tree builds (#87622)"

This reverts commit 9029e6ebdfd98a58b5c5646fd534c6c849148cda, which was
committed by mistake with the wrong message and fails  https://lab.llvm.org/buildbot/#/builders/221/builds/21958.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 51 ++-----------------
 .../X86/store-abs-minbitwidth.ll              | 14 ++++-
 2 files changed, 15 insertions(+), 50 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 55dc0867e2ea9..2995cb8435f76 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -305,32 +305,7 @@ static bool isCommutative(Instruction *I) {
   if (auto *Cmp = dyn_cast<CmpInst>(I))
     return Cmp->isCommutative();
   if (auto *BO = dyn_cast<BinaryOperator>(I))
-    return BO->isCommutative() ||
-           (BO->getOpcode() == Instruction::Sub &&
-            !BO->hasNUsesOrMore(UsesLimit) &&
-            all_of(
-                BO->uses(),
-                [](const Use &U) {
-                  // Commutative, if icmp eq/ne sub, 0
-                  ICmpInst::Predicate Pred;
-                  if (match(U.getUser(),
-                            m_ICmp(Pred, m_Specific(U.get()), m_Zero())) &&
-                      (Pred == ICmpInst::ICMP_EQ || Pred == ICmpInst::ICMP_NE))
-                    return true;
-                  // Commutative, if abs(sub nsw, true) or abs(sub, false).
-                  ConstantInt *Flag;
-                  return match(U.getUser(),
-                               m_Intrinsic<Intrinsic::abs>(
-                                   m_Specific(U.get()), m_ConstantInt(Flag))) &&
-                         (!cast<Instruction>(U.get())->hasNoSignedWrap() ||
-                          Flag->isOne());
-                })) ||
-           (BO->getOpcode() == Instruction::FSub &&
-            !BO->hasNUsesOrMore(UsesLimit) &&
-            all_of(BO->uses(), [](const Use &U) {
-              return match(U.getUser(),
-                           m_Intrinsic<Intrinsic::fabs>(m_Specific(U.get())));
-            }));
+    return BO->isCommutative();
   return I->isCommutative();
 }
 
@@ -6863,7 +6838,7 @@ void BoUpSLP::buildTree_rec(ArrayRef<Value *> VL, unsigned Depth,
 
       // Sort operands of the instructions so that each side is more likely to
       // have the same opcode.
-      if (isa<BinaryOperator>(VL0) && isCommutative(VL0)) {
+      if (isa<BinaryOperator>(VL0) && VL0->isCommutative()) {
         ValueList Left, Right;
         reorderInputsAccordingToOpcode(VL, Left, Right, *this);
         TE->setOperand(0, Left);
@@ -12591,15 +12566,8 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
           static_cast<Instruction::BinaryOps>(E->getOpcode()), LHS,
           RHS);
       propagateIRFlags(V, E->Scalars, VL0, It == MinBWs.end());
-      if (auto *I = dyn_cast<Instruction>(V)) {
+      if (auto *I = dyn_cast<Instruction>(V))
         V = propagateMetadata(I, E->Scalars);
-        // Drop nuw flags for abs(sub(commutative), true).
-        if (!MinBWs.contains(E) && ShuffleOrOp == Instruction::Sub &&
-            any_of(E->Scalars, [](Value *V) {
-              return isCommutative(cast<Instruction>(V));
-            }))
-          I->setHasNoUnsignedWrap(/*b=*/false);
-      }
 
       V = FinalShuffle(V, E, VecTy);
 
@@ -12925,19 +12893,6 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
 
       propagateIRFlags(V0, OpScalars, E->getMainOp(), It == MinBWs.end());
       propagateIRFlags(V1, AltScalars, E->getAltOp(), It == MinBWs.end());
-      auto DropNuwFlag = [&](Value *Vec, unsigned Opcode) {
-        // Drop nuw flags for abs(sub(commutative), true).
-        if (auto *I = dyn_cast<Instruction>(Vec);
-            I && Opcode == Instruction::Sub && !MinBWs.contains(E) &&
-            any_of(E->Scalars, [](Value *V) {
-              auto *I = cast<Instruction>(V);
-              return I->getOpcode() == Instruction::Sub &&
-                     isCommutative(cast<Instruction>(V));
-            }))
-          I->setHasNoUnsignedWrap(/*b=*/false);
-      };
-      DropNuwFlag(V0, E->getOpcode());
-      DropNuwFlag(V1, E->getAltOpcode());
 
       Value *V = Builder.CreateShuffleVector(V0, V1, Mask);
       if (auto *I = dyn_cast<Instruction>(V)) {
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
index 64bfb242752c1..df7312e3d2b56 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
@@ -4,9 +4,19 @@
 
 define i32 @test(ptr noalias %in, ptr noalias %inn, ptr %out) {
 ; CHECK-LABEL: @test(
-; CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i8>, ptr [[IN:%.*]], align 1
-; CHECK-NEXT:    [[TMP11:%.*]] = load <4 x i8>, ptr [[INN:%.*]], align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i8>, ptr [[IN:%.*]], align 1
+; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i8, ptr [[IN]], i64 2
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i8>, ptr [[GEP_2]], align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i8>, ptr [[INN:%.*]], align 1
+; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i8, ptr [[INN]], i64 2
+; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i8>, ptr [[GEP_5]], align 1
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i8> [[TMP2]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i8> [[TMP5]], <4 x i8> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
 ; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i8> [[TMP7]] to <4 x i16>
+; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x i8> [[TMP1]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x i8> [[TMP4]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i8> [[TMP9]], <4 x i8> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
 ; CHECK-NEXT:    [[TMP12:%.*]] = sext <4 x i8> [[TMP11]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP13:%.*]] = sub <4 x i16> [[TMP12]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = call <4 x i16> @llvm.abs.v4i16(<4 x i16> [[TMP13]], i1 false)

From 9c4aca2b2ec46895761756e34b191593694e4c47 Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Thu, 11 Apr 2024 11:13:12 -0700
Subject: [PATCH 174/886] [HWASAN] Emit optimization remarks (#88332)

Similar to #88304

SelHWAsan is optimization. We may want to diagnose compiler decisions.
Remarks is the tool for that https://llvm.org/docs/Remarks.html.
---
 .../Instrumentation/HWAddressSanitizer.cpp    | 50 +++++++++++++------
 .../HWAddressSanitizer/pgo-opt-out.ll         | 12 +++--
 2 files changed, 43 insertions(+), 19 deletions(-)

diff --git a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
index 82f60f4feed45..3890aa8ca6ee6 100644
--- a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
@@ -21,6 +21,7 @@
 #include "llvm/Analysis/BlockFrequencyInfo.h"
 #include "llvm/Analysis/DomTreeUpdater.h"
 #include "llvm/Analysis/GlobalsModRef.h"
+#include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/Analysis/PostDominators.h"
 #include "llvm/Analysis/ProfileSummaryInfo.h"
 #include "llvm/Analysis/StackSafetyAnalysis.h"
@@ -1492,23 +1493,42 @@ bool HWAddressSanitizer::instrumentStack(memtag::StackInfo &SInfo,
   return true;
 }
 
+static void emitRemark(const Function &F, OptimizationRemarkEmitter &ORE,
+                       bool Skip) {
+  if (Skip) {
+    ORE.emit([&]() {
+      return OptimizationRemark(DEBUG_TYPE, "Skip", &F)
+             << "Skipped: F=" << ore::NV("Function", &F);
+    });
+  } else {
+    ORE.emit([&]() {
+      return OptimizationRemarkMissed(DEBUG_TYPE, "Sanitize", &F)
+             << "Sanitized: F=" << ore::NV("Function", &F);
+    });
+  }
+}
+
 bool HWAddressSanitizer::selectiveInstrumentationShouldSkip(
     Function &F, FunctionAnalysisManager &FAM) const {
-  if (ClRandomSkipRate.getNumOccurrences()) {
-    std::bernoulli_distribution D(ClRandomSkipRate);
-    return !D(*Rng);
-  }
-  if (!ClHotPercentileCutoff.getNumOccurrences())
-    return false;
-  auto &MAMProxy = FAM.getResult<ModuleAnalysisManagerFunctionProxy>(F);
-  ProfileSummaryInfo *PSI =
-      MAMProxy.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
-  if (!PSI || !PSI->hasProfileSummary()) {
-    ++NumNoProfileSummaryFuncs;
-    return false;
-  }
-  return PSI->isFunctionHotInCallGraphNthPercentile(
-      ClHotPercentileCutoff, &F, FAM.getResult<BlockFrequencyAnalysis>(F));
+  bool Skip = [&]() {
+    if (ClRandomSkipRate.getNumOccurrences()) {
+      std::bernoulli_distribution D(ClRandomSkipRate);
+      return !D(*Rng);
+    }
+    if (!ClHotPercentileCutoff.getNumOccurrences())
+      return false;
+    auto &MAMProxy = FAM.getResult<ModuleAnalysisManagerFunctionProxy>(F);
+    ProfileSummaryInfo *PSI =
+        MAMProxy.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
+    if (!PSI || !PSI->hasProfileSummary()) {
+      ++NumNoProfileSummaryFuncs;
+      return false;
+    }
+    return PSI->isFunctionHotInCallGraphNthPercentile(
+        ClHotPercentileCutoff, &F, FAM.getResult<BlockFrequencyAnalysis>(F));
+  }();
+  emitRemark(F, FAM.getResult<OptimizationRemarkEmitterAnalysis>(F), Skip);
+  return Skip;
 }
 
 void HWAddressSanitizer::sanitizeFunction(Function &F,
diff --git a/llvm/test/Instrumentation/HWAddressSanitizer/pgo-opt-out.ll b/llvm/test/Instrumentation/HWAddressSanitizer/pgo-opt-out.ll
index f2661a02da7a0..d5cb27d9b1ec1 100644
--- a/llvm/test/Instrumentation/HWAddressSanitizer/pgo-opt-out.ll
+++ b/llvm/test/Instrumentation/HWAddressSanitizer/pgo-opt-out.ll
@@ -1,17 +1,21 @@
-; RUN: opt < %s -passes='require<profile-summary>,hwasan' -S -hwasan-percentile-cutoff-hot=700000 | FileCheck %s --check-prefix=HOT70
-; RUN: opt < %s -passes='require<profile-summary>,hwasan' -S -hwasan-percentile-cutoff-hot=990000 | FileCheck %s --check-prefix=HOT99
-; RUN: opt < %s -passes='require<profile-summary>,hwasan' -S -hwasan-random-rate=1.0 | FileCheck %s --check-prefix=ALL
-; RUN: opt < %s -passes='require<profile-summary>,hwasan' -S -hwasan-random-rate=0.0 | FileCheck %s --check-prefix=NONE
+; RUN: opt < %s -passes='require<profile-summary>,hwasan' -pass-remarks=hwasan -pass-remarks-missed=hwasan -S -hwasan-percentile-cutoff-hot=700000 2>&1 | FileCheck %s --check-prefix=HOT70
+; RUN: opt < %s -passes='require<profile-summary>,hwasan' -pass-remarks=hwasan -pass-remarks-missed=hwasan -S -hwasan-percentile-cutoff-hot=990000 2>&1 | FileCheck %s --check-prefix=HOT99
+; RUN: opt < %s -passes='require<profile-summary>,hwasan' -pass-remarks=hwasan -pass-remarks-missed=hwasan -S -hwasan-random-rate=1.0 2>&1 | FileCheck %s --check-prefix=ALL
+; RUN: opt < %s -passes='require<profile-summary>,hwasan' -pass-remarks=hwasan -pass-remarks-missed=hwasan -S -hwasan-random-rate=0.0 2>&1 | FileCheck %s --check-prefix=NONE
 
+; HOT70: remark: <unknown>:0:0: Sanitized: F=sanitized
 ; HOT70: @sanitized
 ; HOT70-NEXT: @__hwasan_tls
 
+; HOT99: remark: <unknown>:0:0: Skipped: F=sanitized
 ; HOT99: @sanitized
 ; HOT99-NEXT: %x = alloca i8, i64 4
 
+; ALL: remark: <unknown>:0:0: Sanitized: F=sanitize
 ; ALL: @sanitized
 ; ALL-NEXT: @__hwasan_tls
 
+; NONE: remark: <unknown>:0:0: Skipped: F=sanitized
 ; NONE: @sanitized
 ; NONE-NEXT: %x = alloca i8, i64 4
 

From b2ea38f9fc2381e7c04e610f6f8ed6786c2da38e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Thu, 11 Apr 2024 18:15:32 +0200
Subject: [PATCH 175/886] [clang][Interp] Fix handling integral function
 pointers

As expected, we need to be a little more careful when the
Function* is created from an integer.
---
 clang/lib/AST/Interp/FunctionPointer.h   | 20 ++++--
 clang/unittests/AST/Interp/toAPValue.cpp | 86 ++++++++++++++++++++----
 2 files changed, 88 insertions(+), 18 deletions(-)

diff --git a/clang/lib/AST/Interp/FunctionPointer.h b/clang/lib/AST/Interp/FunctionPointer.h
index f61f9ded0bf00..840c1101f396b 100644
--- a/clang/lib/AST/Interp/FunctionPointer.h
+++ b/clang/lib/AST/Interp/FunctionPointer.h
@@ -20,14 +20,15 @@ namespace interp {
 class FunctionPointer final {
 private:
   const Function *Func;
+  bool Valid;
 
 public:
-  // FIXME: We might want to track the fact that the Function pointer
-  // has been created from an integer and is most likely garbage anyway.
-  FunctionPointer(uintptr_t IntVal = 0, const Descriptor *Desc = nullptr)
-      : Func(reinterpret_cast<const Function *>(IntVal)) {}
+  FunctionPointer(const Function *Func) : Func(Func), Valid(true) {
+    assert(Func);
+  }
 
-  FunctionPointer(const Function *Func) : Func(Func) { assert(Func); }
+  FunctionPointer(uintptr_t IntVal = 0, const Descriptor *Desc = nullptr)
+      : Func(reinterpret_cast<const Function *>(IntVal)), Valid(false) {}
 
   const Function *getFunction() const { return Func; }
   bool isZero() const { return !Func; }
@@ -37,14 +38,21 @@ class FunctionPointer final {
       return APValue(static_cast<Expr *>(nullptr), CharUnits::Zero(), {},
                      /*OnePastTheEnd=*/false, /*IsNull=*/true);
 
+    if (!Valid)
+      return APValue(static_cast<Expr *>(nullptr),
+                     CharUnits::fromQuantity(getIntegerRepresentation()), {},
+                     /*OnePastTheEnd=*/false, /*IsNull=*/false);
+
     return APValue(Func->getDecl(), CharUnits::Zero(), {},
                    /*OnePastTheEnd=*/false, /*IsNull=*/false);
   }
 
   void print(llvm::raw_ostream &OS) const {
     OS << "FnPtr(";
-    if (Func)
+    if (Func && Valid)
       OS << Func->getName();
+    else if (Func)
+      OS << reinterpret_cast<uintptr_t>(Func);
     else
       OS << "nullptr";
     OS << ")";
diff --git a/clang/unittests/AST/Interp/toAPValue.cpp b/clang/unittests/AST/Interp/toAPValue.cpp
index be7929228d283..e56453aba2c5f 100644
--- a/clang/unittests/AST/Interp/toAPValue.cpp
+++ b/clang/unittests/AST/Interp/toAPValue.cpp
@@ -90,7 +90,8 @@ TEST(ToAPValue, Pointers) {
 
 TEST(ToAPValue, FunctionPointers) {
   constexpr char Code[] = " constexpr bool foo() { return true; }\n"
-                          " constexpr bool (*func)() = foo;\n";
+                          " constexpr bool (*func)() = foo;\n"
+                          " constexpr bool (*nullp)() = nullptr;\n";
 
   auto AST = tooling::buildASTFromCodeWithArgs(
       Code, {"-fexperimental-new-constant-interpreter"});
@@ -112,15 +113,76 @@ TEST(ToAPValue, FunctionPointers) {
     return Prog.getPtrGlobal(*Prog.getGlobal(D));
   };
 
-  const Pointer &GP = getGlobalPtr("func");
-  const FunctionPointer &FP = GP.deref<FunctionPointer>();
-  ASSERT_FALSE(FP.isZero());
-  APValue A = FP.toAPValue();
-  ASSERT_TRUE(A.hasValue());
-  ASSERT_TRUE(A.isLValue());
-  ASSERT_TRUE(A.hasLValuePath());
-  const auto &Path = A.getLValuePath();
-  ASSERT_EQ(Path.size(), 0u);
-  ASSERT_FALSE(A.getLValueBase().isNull());
-  ASSERT_EQ(A.getLValueBase().dyn_cast<const ValueDecl *>(), getDecl("foo"));
+  {
+    const Pointer &GP = getGlobalPtr("func");
+    const FunctionPointer &FP = GP.deref<FunctionPointer>();
+    ASSERT_FALSE(FP.isZero());
+    APValue A = FP.toAPValue();
+    ASSERT_TRUE(A.hasValue());
+    ASSERT_TRUE(A.isLValue());
+    ASSERT_TRUE(A.hasLValuePath());
+    const auto &Path = A.getLValuePath();
+    ASSERT_EQ(Path.size(), 0u);
+    ASSERT_FALSE(A.getLValueBase().isNull());
+    ASSERT_EQ(A.getLValueBase().dyn_cast<const ValueDecl *>(), getDecl("foo"));
+  }
+
+  {
+    const ValueDecl *D = getDecl("nullp");
+    ASSERT_NE(D, nullptr);
+    const Pointer &GP = getGlobalPtr("nullp");
+    const auto &P = GP.deref<FunctionPointer>();
+    APValue A = P.toAPValue();
+    ASSERT_TRUE(A.isLValue());
+    ASSERT_TRUE(A.getLValueBase().isNull());
+    ASSERT_TRUE(A.isNullPointer());
+    APSInt I;
+    bool Success = A.toIntegralConstant(I, D->getType(), AST->getASTContext());
+    ASSERT_TRUE(Success);
+    ASSERT_EQ(I, 0);
+  }
+}
+
+TEST(ToAPValue, FunctionPointersC) {
+  // NB: The declaration of func2 is useless, but it makes us register a global
+  // variable for func.
+  constexpr char Code[] = "const int (* const func)(int *) = (void*)17;\n"
+                          "const int (*func2)(int *) = func;\n";
+  auto AST = tooling::buildASTFromCodeWithArgs(
+      Code, {"-x", "c", "-fexperimental-new-constant-interpreter"});
+
+  auto &Ctx = AST->getASTContext().getInterpContext();
+  Program &Prog = Ctx.getProgram();
+
+  auto getDecl = [&](const char *Name) -> const ValueDecl * {
+    auto Nodes =
+        match(valueDecl(hasName(Name)).bind("var"), AST->getASTContext());
+    assert(Nodes.size() == 1);
+    const auto *D = Nodes[0].getNodeAs<ValueDecl>("var");
+    assert(D);
+    return D;
+  };
+
+  auto getGlobalPtr = [&](const char *Name) -> Pointer {
+    const VarDecl *D = cast<VarDecl>(getDecl(Name));
+    return Prog.getPtrGlobal(*Prog.getGlobal(D));
+  };
+
+  {
+    const ValueDecl *D = getDecl("func");
+    const Pointer &GP = getGlobalPtr("func");
+    ASSERT_TRUE(GP.isLive());
+    const FunctionPointer &FP = GP.deref<FunctionPointer>();
+    ASSERT_FALSE(FP.isZero());
+    APValue A = FP.toAPValue();
+    ASSERT_TRUE(A.hasValue());
+    ASSERT_TRUE(A.isLValue());
+    const auto &Path = A.getLValuePath();
+    ASSERT_EQ(Path.size(), 0u);
+    ASSERT_TRUE(A.getLValueBase().isNull());
+    APSInt I;
+    bool Success = A.toIntegralConstant(I, D->getType(), AST->getASTContext());
+    ASSERT_TRUE(Success);
+    ASSERT_EQ(I, 17);
+  }
 }

From fc52ee336b394d84110184d625cda1d4f84d8098 Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Thu, 11 Apr 2024 20:29:18 +0200
Subject: [PATCH 176/886] [lldb][ClangUserExpression][NFCI] Pass the most
 specific ExecutionContextScope possible into ClangExpressionParser (#87657)

The `ClangExpressionParser` takes an `ExecutionContextScope` which it
uses to query the `Process`/`Target`/`StackFrame` to set various
compiler options in preparation for parsing an expression.

However, `TryParse` constructs the parser with a `Process` or `Target`,
never a `StackFrame`. So when the parser tries to retrieve the current
`StackFrame` from the `exe_scope`, it doesn't succeed. In future patches
we want to query the `StackFrame` from within the
`ClangExpressionParser` constructor.

This patch simplifies `TryParse`, by removing the redundant `exe_scope`
parameter, and instead uses the `exe_ctx` to derive the most fitting
`exe_scope` to pass into `ClangExpressionParser`.

Not entirely sure how to test this. This patch is a prerequisite to get
subsequent patches that set `LangOpts` based on the current `StackFrame`
to work.
---
 .../Clang/ClangUserExpression.cpp             | 35 +++++++------------
 .../Clang/ClangUserExpression.h               |  6 ++--
 2 files changed, 15 insertions(+), 26 deletions(-)

diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.cpp b/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.cpp
index 30bc81c9ed8c1..5776b1e94e072 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.cpp
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.cpp
@@ -553,9 +553,9 @@ bool ClangUserExpression::PrepareForParsing(
 }
 
 bool ClangUserExpression::TryParse(
-    DiagnosticManager &diagnostic_manager, ExecutionContextScope *exe_scope,
-    ExecutionContext &exe_ctx, lldb_private::ExecutionPolicy execution_policy,
-    bool keep_result_in_memory, bool generate_debug_info) {
+    DiagnosticManager &diagnostic_manager, ExecutionContext &exe_ctx,
+    lldb_private::ExecutionPolicy execution_policy, bool keep_result_in_memory,
+    bool generate_debug_info) {
   m_materializer_up = std::make_unique<Materializer>();
 
   ResetDeclMap(exe_ctx, m_result_delegate, keep_result_in_memory);
@@ -574,7 +574,8 @@ bool ClangUserExpression::TryParse(
   }
 
   m_parser = std::make_unique<ClangExpressionParser>(
-      exe_scope, *this, generate_debug_info, m_include_directories, m_filename);
+      exe_ctx.GetBestExecutionContextScope(), *this, generate_debug_info,
+      m_include_directories, m_filename);
 
   unsigned num_errors = m_parser->Parse(diagnostic_manager);
 
@@ -669,15 +670,8 @@ bool ClangUserExpression::Parse(DiagnosticManager &diagnostic_manager,
   // Parse the expression
   //
 
-  Process *process = exe_ctx.GetProcessPtr();
-  ExecutionContextScope *exe_scope = process;
-
-  if (!exe_scope)
-    exe_scope = exe_ctx.GetTargetPtr();
-
-  bool parse_success = TryParse(diagnostic_manager, exe_scope, exe_ctx,
-                                execution_policy, keep_result_in_memory,
-                                generate_debug_info);
+  bool parse_success = TryParse(diagnostic_manager, exe_ctx, execution_policy,
+                                keep_result_in_memory, generate_debug_info);
   // If the expression failed to parse, check if retrying parsing with a loaded
   // C++ module is possible.
   if (!parse_success && shouldRetryWithCppModule(*target, execution_policy)) {
@@ -695,9 +689,8 @@ bool ClangUserExpression::Parse(DiagnosticManager &diagnostic_manager,
       // so recreate those.
       CreateSourceCode(retry_manager, exe_ctx, m_imported_cpp_modules,
                        /*for_completion*/ false);
-      parse_success = TryParse(retry_manager, exe_scope, exe_ctx,
-                               execution_policy, keep_result_in_memory,
-                               generate_debug_info);
+      parse_success = TryParse(retry_manager, exe_ctx, execution_policy,
+                               keep_result_in_memory, generate_debug_info);
       // Return the parse diagnostics if we were successful.
       if (parse_success)
         diagnostic_manager = std::move(retry_manager);
@@ -759,6 +752,7 @@ bool ClangUserExpression::Parse(DiagnosticManager &diagnostic_manager,
     }
   }
 
+  Process *process = exe_ctx.GetProcessPtr();
   if (process && m_jit_start_addr != LLDB_INVALID_ADDRESS)
     m_jit_process_wp = lldb::ProcessWP(process->shared_from_this());
   return true;
@@ -840,13 +834,8 @@ bool ClangUserExpression::Complete(ExecutionContext &exe_ctx,
     DeclMap()->SetLookupsEnabled(true);
   }
 
-  Process *process = exe_ctx.GetProcessPtr();
-  ExecutionContextScope *exe_scope = process;
-
-  if (!exe_scope)
-    exe_scope = exe_ctx.GetTargetPtr();
-
-  ClangExpressionParser parser(exe_scope, *this, false);
+  ClangExpressionParser parser(exe_ctx.GetBestExecutionContextScope(), *this,
+                               false);
 
   // We have to find the source code location where the user text is inside
   // the transformed expression code. When creating the transformed text, we
diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.h b/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.h
index 7a8c095f61183..bc07cbcf9e646 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.h
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangUserExpression.h
@@ -185,9 +185,9 @@ class ClangUserExpression : public LLVMUserExpression {
   /// The parameter have the same meaning as in ClangUserExpression::Parse.
   /// \see ClangUserExpression::Parse
   bool TryParse(DiagnosticManager &diagnostic_manager,
-                ExecutionContextScope *exe_scope, ExecutionContext &exe_ctx,
-                lldb_private::ExecutionPolicy execution_policy, bool keep_result_in_memory,
-                bool generate_debug_info);
+                ExecutionContext &exe_ctx,
+                lldb_private::ExecutionPolicy execution_policy,
+                bool keep_result_in_memory, bool generate_debug_info);
 
   void SetupCppModuleImports(ExecutionContext &exe_ctx);
 

From 38f8fcea3455ef1d77faf9746579f52b4e18dacc Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Thu, 11 Apr 2024 20:30:48 +0200
Subject: [PATCH 177/886] [lldb][ClangExpressionParser] Don't by default enable
 Objecitve-C support when evaluating C++ expressions (#87767)

This patch attempts to decouple C++ expression evaluation from
Objective-C support. We've previously enabled it by default (if a
runtime existed), but that meant we're opting into extra work we only
need to do for Objective-C, which complicates/slows down C++ expression
evaluation. Of course there's a valid use-case for this, which is
calling Objective-C APIs when stopped in C++ frames (which Objective-C++
developers might want to do). In those cases we should really prompt the
user to add the `expr --language objc++` flag. To accomodate a likely
frequent use-case where a user breaks in a system C++ library (without
debug-symbols) but their application is actually an Objective-C app, we
allow Objective-C support in C++ expressions if the current frame
doesn't have debug-info.

This fixes https://github.com/llvm/llvm-project/issues/75443 and allows
us to add more `LangOpts.ObjC` guards around the expression evaluator in
the future (e.g., we could avoid looking into the Objective-C runtime
during C++ expression evaluation, which we currently do
unconditionally).

Depends on https://github.com/llvm/llvm-project/pull/87657
---
 .../Clang/ClangExpressionParser.cpp            |  5 ++++-
 .../diagnostics/TestExprDiagnostics.py         |  2 +-
 .../expression/options/TestExprOptions.py      |  1 -
 .../objc-builtin-types/TestObjCBuiltinTypes.py |  5 +----
 .../Makefile                                   |  4 ++++
 .../TestObjCFromCppFramesWithoutDebugInfo.py   | 18 ++++++++++++++++++
 .../main.cpp                                   |  1 +
 7 files changed, 29 insertions(+), 7 deletions(-)
 create mode 100644 lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/Makefile
 create mode 100644 lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/TestObjCFromCppFramesWithoutDebugInfo.py
 create mode 100644 lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/main.cpp

diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionParser.cpp b/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionParser.cpp
index 822d286cd6c3c..f48bdc730d916 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionParser.cpp
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionParser.cpp
@@ -526,7 +526,10 @@ ClangExpressionParser::ClangExpressionParser(
     [[fallthrough]];
   case lldb::eLanguageTypeC_plus_plus_03:
     lang_opts.CPlusPlus = true;
-    if (process_sp)
+    if (process_sp
+        // We're stopped in a frame without debug-info. The user probably
+        // intends to make global queries (which should include Objective-C).
+        && !(frame_sp && frame_sp->HasDebugInformation()))
       lang_opts.ObjC =
           process_sp->GetLanguageRuntime(lldb::eLanguageTypeObjC) != nullptr;
     break;
diff --git a/lldb/test/API/commands/expression/diagnostics/TestExprDiagnostics.py b/lldb/test/API/commands/expression/diagnostics/TestExprDiagnostics.py
index d64e9897a844f..ddc1c3598480c 100644
--- a/lldb/test/API/commands/expression/diagnostics/TestExprDiagnostics.py
+++ b/lldb/test/API/commands/expression/diagnostics/TestExprDiagnostics.py
@@ -172,7 +172,7 @@ def test_source_locations_from_objc_modules(self):
 
         # Import foundation so that the Obj-C module is loaded (which contains source locations
         # that can be used by LLDB).
-        self.runCmd("expr @import Foundation")
+        self.runCmd("expr --language objective-c++ -- @import Foundation")
         value = frame.EvaluateExpression("NSLog(1);")
         self.assertFalse(value.GetError().Success())
         # LLDB should print the source line that defines NSLog. To not rely on any
diff --git a/lldb/test/API/commands/expression/options/TestExprOptions.py b/lldb/test/API/commands/expression/options/TestExprOptions.py
index 6652c71083a95..01899f3b97cf4 100644
--- a/lldb/test/API/commands/expression/options/TestExprOptions.py
+++ b/lldb/test/API/commands/expression/options/TestExprOptions.py
@@ -59,7 +59,6 @@ def test_expr_options(self):
         self.assertTrue(val.IsValid())
         self.assertFalse(val.GetError().Success())
 
-    @skipIfDarwin
     def test_expr_options_lang(self):
         """These expression language options should work as expected."""
         self.build()
diff --git a/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py b/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
index 1eb7205f1bb46..611c738899905 100644
--- a/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
+++ b/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
@@ -51,7 +51,4 @@ def test_with_python_api(self):
             "expr --language Objective-C++ -- id my_id = 0; my_id",
             patterns=["\(id\) \$.* = nil"],
         )
-        self.expect(
-            "expr --language C++ -- id my_id = 0; my_id",
-            patterns=["\(id\) \$.* = nullptr"],
-        )
+        self.expect("expr --language C++ -- id my_id = 0; my_id", error=True)
diff --git a/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/Makefile b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/Makefile
new file mode 100644
index 0000000000000..7c3c32d6f82df
--- /dev/null
+++ b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/Makefile
@@ -0,0 +1,4 @@
+CXX_SOURCES := main.cpp
+CXXFLAGS_EXTRAS := -g0
+
+include Makefile.rules
diff --git a/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/TestObjCFromCppFramesWithoutDebugInfo.py b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/TestObjCFromCppFramesWithoutDebugInfo.py
new file mode 100644
index 0000000000000..ef8d5540fa4ef
--- /dev/null
+++ b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/TestObjCFromCppFramesWithoutDebugInfo.py
@@ -0,0 +1,18 @@
+"""
+Test that the the expression parser enables ObjC support
+when stopped in a C++ frame without debug-info.
+"""
+
+import lldb
+from lldbsuite.test.decorators import *
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test import lldbutil
+
+
+class TestObjCFromCppFramesWithoutDebugInfo(TestBase):
+    def test(self):
+        self.build()
+        (_, process, _, _) = lldbutil.run_to_name_breakpoint(self, "main")
+
+        self.assertState(process.GetState(), lldb.eStateStopped)
+        self.expect("expr id", error=False)
diff --git a/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/main.cpp b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/main.cpp
new file mode 100644
index 0000000000000..76e8197013aab
--- /dev/null
+++ b/lldb/test/API/lang/objcxx/objc-from-cpp-frames-without-debuginfo/main.cpp
@@ -0,0 +1 @@
+int main() { return 0; }

From b1822ef311b79d409be6f60ce86fccd967d623bc Mon Sep 17 00:00:00 2001
From: XChy <xxs_chy@outlook.com>
Date: Fri, 12 Apr 2024 02:36:32 +0800
Subject: [PATCH 178/886] [SCCP] Refine trunc with nsw/nuw flags (#87926)

Following #85592, add support for nsw/nuw flags of trunc in SCCP.
---
 llvm/lib/Transforms/Utils/SCCPSolver.cpp      |  21 ++
 .../PhaseOrdering/ARM/arm_mult_q15.ll         |   2 +-
 .../test/Transforms/SCCP/conditions-ranges.ll |   6 +-
 llvm/test/Transforms/SCCP/ip-ranges-casts.ll  |   4 +-
 .../Transforms/SCCP/trunc-nuw-nsw-flags.ll    | 179 ++++++++++++++++++
 llvm/test/Transforms/SCCP/widening.ll         |  24 +--
 6 files changed, 218 insertions(+), 18 deletions(-)
 create mode 100644 llvm/test/Transforms/SCCP/trunc-nuw-nsw-flags.ll

diff --git a/llvm/lib/Transforms/Utils/SCCPSolver.cpp b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
index b82ed25a6d0cf..c6029b428ed39 100644
--- a/llvm/lib/Transforms/Utils/SCCPSolver.cpp
+++ b/llvm/lib/Transforms/Utils/SCCPSolver.cpp
@@ -120,6 +120,9 @@ static bool refineInstruction(SCCPSolver &Solver,
   };
 
   if (isa<OverflowingBinaryOperator>(Inst)) {
+    if (Inst.hasNoSignedWrap() && Inst.hasNoUnsignedWrap())
+      return false;
+
     auto RangeA = GetRange(Inst.getOperand(0));
     auto RangeB = GetRange(Inst.getOperand(1));
     if (!Inst.hasNoUnsignedWrap()) {
@@ -146,6 +149,24 @@ static bool refineInstruction(SCCPSolver &Solver,
       Inst.setNonNeg();
       Changed = true;
     }
+  } else if (TruncInst *TI = dyn_cast<TruncInst>(&Inst)) {
+    if (TI->hasNoSignedWrap() && TI->hasNoUnsignedWrap())
+      return false;
+
+    auto Range = GetRange(Inst.getOperand(0));
+    uint64_t DestWidth = TI->getDestTy()->getScalarSizeInBits();
+    if (!TI->hasNoUnsignedWrap()) {
+      if (Range.getActiveBits() <= DestWidth) {
+        TI->setHasNoUnsignedWrap(true);
+        Changed = true;
+      }
+    }
+    if (!TI->hasNoSignedWrap()) {
+      if (Range.getMinSignedBits() <= DestWidth) {
+        TI->setHasNoSignedWrap(true);
+        Changed = true;
+      }
+    }
   }
 
   return Changed;
diff --git a/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll b/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
index 36bcda4c43ca9..c6126727598ef 100644
--- a/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
+++ b/llvm/test/Transforms/PhaseOrdering/ARM/arm_mult_q15.ll
@@ -69,7 +69,7 @@ define void @arm_mult_q15(ptr %pSrcA, ptr %pSrcB, ptr noalias %pDst, i32 %blockS
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV2]], [[CONV]]
 ; CHECK-NEXT:    [[SHR:%.*]] = ashr i32 [[MUL]], 15
 ; CHECK-NEXT:    [[SPEC_SELECT_I:%.*]] = tail call i32 @llvm.smin.i32(i32 [[SHR]], i32 32767)
-; CHECK-NEXT:    [[CONV3:%.*]] = trunc i32 [[SPEC_SELECT_I]] to i16
+; CHECK-NEXT:    [[CONV3:%.*]] = trunc nsw i32 [[SPEC_SELECT_I]] to i16
 ; CHECK-NEXT:    [[INCDEC_PTR4]] = getelementptr inbounds i8, ptr [[PDST_ADDR_04]], i32 2
 ; CHECK-NEXT:    store i16 [[CONV3]], ptr [[PDST_ADDR_04]], align 2
 ; CHECK-NEXT:    [[DEC]] = add i32 [[BLKCNT_06]], -1
diff --git a/llvm/test/Transforms/SCCP/conditions-ranges.ll b/llvm/test/Transforms/SCCP/conditions-ranges.ll
index f17b6c1317263..25719d2bee521 100644
--- a/llvm/test/Transforms/SCCP/conditions-ranges.ll
+++ b/llvm/test/Transforms/SCCP/conditions-ranges.ll
@@ -178,7 +178,7 @@ define i32 @f5(i64 %sz) {
 ; CHECK-NEXT:    br label [[COND_END]]
 ; CHECK:       cond.end:
 ; CHECK-NEXT:    [[COND:%.*]] = phi i64 [ [[DIV]], [[COND_TRUE]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[COND]] to i32
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[COND]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
 ;
 entry:
@@ -759,7 +759,7 @@ define i32 @udiv_1(i64 %sz) {
 ; CHECK-NEXT:    br label [[COND_END]]
 ; CHECK:       cond.end:
 ; CHECK-NEXT:    [[COND:%.*]] = phi i64 [ [[DIV]], [[COND_TRUE]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[COND]] to i32
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[COND]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
 ;
 entry:
@@ -786,7 +786,7 @@ define i32 @udiv_2(i64 %sz) {
 ; CHECK-NEXT:    br label [[COND_END]]
 ; CHECK:       cond.end:
 ; CHECK-NEXT:    [[COND:%.*]] = phi i64 [ 0, [[COND_TRUE]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 [[COND]] to i32
+; CHECK-NEXT:    [[CONV:%.*]] = trunc nuw nsw i64 [[COND]] to i32
 ; CHECK-NEXT:    ret i32 [[CONV]]
 ;
 entry:
diff --git a/llvm/test/Transforms/SCCP/ip-ranges-casts.ll b/llvm/test/Transforms/SCCP/ip-ranges-casts.ll
index b97b3a0c45816..80d90922c2fbd 100644
--- a/llvm/test/Transforms/SCCP/ip-ranges-casts.ll
+++ b/llvm/test/Transforms/SCCP/ip-ranges-casts.ll
@@ -4,7 +4,7 @@
 ; x = [100, 301)
 define internal i1 @f.trunc(i32 %x) {
 ; CHECK-LABEL: @f.trunc(
-; CHECK-NEXT:    [[T_1:%.*]] = trunc i32 [[X:%.*]] to i16
+; CHECK-NEXT:    [[T_1:%.*]] = trunc nuw nsw i32 [[X:%.*]] to i16
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp sgt i16 [[T_1]], 299
 ; CHECK-NEXT:    [[C_4:%.*]] = icmp slt i16 [[T_1]], 101
 ; CHECK-NEXT:    [[RES_1:%.*]] = add nuw nsw i1 false, [[C_2]]
@@ -329,7 +329,7 @@ define i64 @caller.sext_to_zext(i32 %i) {
 ; CHECK-LABEL: @caller.sext_to_zext(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sle i32 [[I:%.*]], 9
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i1 [[CMP]] to i32
-; CHECK-NEXT:    [[T:%.*]] = call i64 @f.sext_to_zext(i32 [[CONV]])
+; CHECK-NEXT:    [[T:%.*]] = call i64 @f.sext_to_zext(i32 [[CONV]]), !range [[RNG0:![0-9]+]]
 ; CHECK-NEXT:    ret i64 [[T]]
 ;
   %cmp = icmp sle i32 %i, 9
diff --git a/llvm/test/Transforms/SCCP/trunc-nuw-nsw-flags.ll b/llvm/test/Transforms/SCCP/trunc-nuw-nsw-flags.ll
new file mode 100644
index 0000000000000..fc3e56011d46c
--- /dev/null
+++ b/llvm/test/Transforms/SCCP/trunc-nuw-nsw-flags.ll
@@ -0,0 +1,179 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=ipsccp -S %s | FileCheck %s
+
+define i16 @range_from_and_nuw(i32 %a) {
+; CHECK-LABEL: define i16 @range_from_and_nuw(
+; CHECK-SAME: i32 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND1:%.*]] = and i32 [[A]], 65535
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nuw i32 [[AND1]] to i16
+; CHECK-NEXT:    ret i16 [[TRUNC1]]
+;
+entry:
+  %and1 = and i32 %a, 65535
+  %trunc1 = trunc i32 %and1 to i16
+  ret i16 %trunc1
+}
+
+define i8 @range_from_or_nsw(i16 %a) {
+; CHECK-LABEL: define i8 @range_from_or_nsw(
+; CHECK-SAME: i16 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND1:%.*]] = or i16 [[A]], -128
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nsw i16 [[AND1]] to i8
+; CHECK-NEXT:    ret i8 [[TRUNC1]]
+;
+entry:
+  %and1 = or i16 %a, 65408
+  %trunc1 = trunc i16 %and1 to i8
+  ret i8 %trunc1
+}
+
+define i16 @range_from_and_nuw_nsw(i32 %a) {
+; CHECK-LABEL: define i16 @range_from_and_nuw_nsw(
+; CHECK-SAME: i32 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND1:%.*]] = and i32 [[A]], 32767
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nuw nsw i32 [[AND1]] to i16
+; CHECK-NEXT:    ret i16 [[TRUNC1]]
+;
+entry:
+  %and1 = and i32 %a, 32767
+  %trunc1 = trunc i32 %and1 to i16
+  ret i16 %trunc1
+}
+
+define <4 x i16> @range_from_and_nuw_vec(<4 x i32> %a) {
+; CHECK-LABEL: define <4 x i16> @range_from_and_nuw_vec(
+; CHECK-SAME: <4 x i32> [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND1:%.*]] = and <4 x i32> [[A]], <i32 65535, i32 65535, i32 65535, i32 65535>
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc <4 x i32> [[AND1]] to <4 x i16>
+; CHECK-NEXT:    ret <4 x i16> [[TRUNC1]]
+;
+entry:
+  %and1 = and <4 x i32> %a, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %trunc1 = trunc <4 x i32> %and1 to <4 x i16>
+  ret <4 x i16> %trunc1
+}
+
+define i4 @range_from_sge_sle(i8 %a) {
+; CHECK-LABEL: define i4 @range_from_sge_sle(
+; CHECK-SAME: i8 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SGT:%.*]] = icmp sge i8 [[A]], 0
+; CHECK-NEXT:    [[SLT:%.*]] = icmp sle i8 [[A]], 15
+; CHECK-NEXT:    [[AND:%.*]] = and i1 [[SGT]], [[SLT]]
+; CHECK-NEXT:    br i1 [[AND]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    [[AND1:%.*]] = and i8 [[A]], 7
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nuw i8 [[A]] to i4
+; CHECK-NEXT:    [[TRUNC2:%.*]] = trunc nuw nsw i8 [[AND1]] to i4
+; CHECK-NEXT:    [[XOR1:%.*]] = xor i4 [[TRUNC1]], [[TRUNC2]]
+; CHECK-NEXT:    ret i4 [[XOR1]]
+; CHECK:       else:
+; CHECK-NEXT:    [[TRUNC3:%.*]] = trunc i8 [[A]] to i4
+; CHECK-NEXT:    ret i4 [[TRUNC3]]
+;
+entry:
+  %sgt = icmp sge i8 %a, 0
+  %slt = icmp sle i8 %a, 15
+  %and = and i1 %sgt, %slt
+  br i1 %and, label %then, label %else
+
+then:
+  %and1 = and i8 %a, 7
+
+  %trunc1 = trunc i8 %a to i4
+  %trunc2 = trunc i8 %and1 to i4
+  %xor1 = xor i4 %trunc1, %trunc2
+  ret i4 %xor1
+
+else:
+  %trunc3 = trunc i8 %a to i4
+  ret i4 %trunc3
+}
+
+define i16 @range_from_sext(i16 %a) {
+; CHECK-LABEL: define i16 @range_from_sext(
+; CHECK-SAME: i16 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SEXT1:%.*]] = sext i16 [[A]] to i32
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nsw i32 [[SEXT1]] to i16
+; CHECK-NEXT:    ret i16 [[TRUNC1]]
+;
+entry:
+  %sext1 = sext i16 %a to i32
+  %trunc1 = trunc i32 %sext1 to i16
+  ret i16 %trunc1
+}
+
+define i16 @range_from_zext(i16 %a) {
+; CHECK-LABEL: define i16 @range_from_zext(
+; CHECK-SAME: i16 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[ZEXT1:%.*]] = zext i16 [[A]] to i32
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nuw i32 [[ZEXT1]] to i16
+; CHECK-NEXT:    ret i16 [[TRUNC1]]
+;
+entry:
+  %zext1 = zext i16 %a to i32
+  %trunc1 = trunc i32 %zext1 to i16
+  ret i16 %trunc1
+}
+
+define i1 @range_from_select_i1_nuw(i1 %c) {
+; CHECK-LABEL: define i1 @range_from_select_i1_nuw(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SELECT1:%.*]] = select i1 [[C]], i16 0, i16 1
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nuw i16 [[SELECT1]] to i1
+; CHECK-NEXT:    ret i1 [[TRUNC1]]
+;
+entry:
+  %select1 = select i1 %c, i16 0, i16 1
+  %trunc1 = trunc i16 %select1 to i1
+  ret i1 %trunc1
+}
+
+define i1 @range_from_select_i1_nsw(i1 %c) {
+; CHECK-LABEL: define i1 @range_from_select_i1_nsw(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SELECT1:%.*]] = select i1 [[C]], i16 0, i16 -1
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc nsw i16 [[SELECT1]] to i1
+; CHECK-NEXT:    ret i1 [[TRUNC1]]
+;
+entry:
+  %select1 = select i1 %c, i16 0, i16 -1
+  %trunc1 = trunc i16 %select1 to i1
+  ret i1 %trunc1
+}
+
+define i1 @range_from_select_i1_fail(i1 %c) {
+; CHECK-LABEL: define i1 @range_from_select_i1_fail(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SELECT1:%.*]] = select i1 [[C]], i16 1, i16 -1
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc i16 [[SELECT1]] to i1
+; CHECK-NEXT:    ret i1 [[TRUNC1]]
+;
+entry:
+  %select1 = select i1 %c, i16 1, i16 -1
+  %trunc1 = trunc i16 %select1 to i1
+  ret i1 %trunc1
+}
+
+define i8 @range_from_trunc_fail(i32 %a) {
+; CHECK-LABEL: define i8 @range_from_trunc_fail(
+; CHECK-SAME: i32 [[A:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TRUNC1:%.*]] = trunc i32 [[A]] to i16
+; CHECK-NEXT:    [[TRUNC2:%.*]] = trunc i16 [[TRUNC1]] to i8
+; CHECK-NEXT:    ret i8 [[TRUNC2]]
+;
+entry:
+  %trunc1 = trunc i32 %a to i16
+  %trunc2 = trunc i16 %trunc1 to i8
+  ret i8 %trunc2
+}
diff --git a/llvm/test/Transforms/SCCP/widening.ll b/llvm/test/Transforms/SCCP/widening.ll
index 2223ca44bccdb..473594eb988da 100644
--- a/llvm/test/Transforms/SCCP/widening.ll
+++ b/llvm/test/Transforms/SCCP/widening.ll
@@ -438,9 +438,9 @@ define void @foo(ptr %arg) {
 ; SCCP-NEXT:    [[TMP:%.*]] = zext i8 undef to i32
 ; SCCP-NEXT:    [[TMP2:%.*]] = load i64, ptr [[ARG:%.*]], align 8
 ; SCCP-NEXT:    switch i32 [[TMP]], label [[BB20:%.*]] [
-; SCCP-NEXT:    i32 1, label [[BB3:%.*]]
-; SCCP-NEXT:    i32 2, label [[BB4:%.*]]
-; SCCP-NEXT:    i32 4, label [[BB19:%.*]]
+; SCCP-NEXT:      i32 1, label [[BB3:%.*]]
+; SCCP-NEXT:      i32 2, label [[BB4:%.*]]
+; SCCP-NEXT:      i32 4, label [[BB19:%.*]]
 ; SCCP-NEXT:    ]
 ; SCCP:       bb3:
 ; SCCP-NEXT:    unreachable
@@ -449,7 +449,7 @@ define void @foo(ptr %arg) {
 ; SCCP-NEXT:    [[TMP6:%.*]] = and i64 [[TMP5]], 3
 ; SCCP-NEXT:    [[TMP7:%.*]] = sub nuw nsw i64 3, [[TMP6]]
 ; SCCP-NEXT:    [[TMP8:%.*]] = shl nuw nsw i64 [[TMP7]], 1
-; SCCP-NEXT:    [[TMP9:%.*]] = trunc i64 [[TMP8]] to i32
+; SCCP-NEXT:    [[TMP9:%.*]] = trunc nuw nsw i64 [[TMP8]] to i32
 ; SCCP-NEXT:    [[TMP10:%.*]] = zext nneg i32 [[TMP9]] to i64
 ; SCCP-NEXT:    br label [[BB11:%.*]]
 ; SCCP:       bb11:
@@ -475,9 +475,9 @@ define void @foo(ptr %arg) {
 ; IPSCCP-NEXT:    [[TMP:%.*]] = zext i8 undef to i32
 ; IPSCCP-NEXT:    [[TMP2:%.*]] = load i64, ptr [[ARG:%.*]], align 8
 ; IPSCCP-NEXT:    switch i32 [[TMP]], label [[BB20:%.*]] [
-; IPSCCP-NEXT:    i32 1, label [[BB3:%.*]]
-; IPSCCP-NEXT:    i32 2, label [[BB4:%.*]]
-; IPSCCP-NEXT:    i32 4, label [[BB19:%.*]]
+; IPSCCP-NEXT:      i32 1, label [[BB3:%.*]]
+; IPSCCP-NEXT:      i32 2, label [[BB4:%.*]]
+; IPSCCP-NEXT:      i32 4, label [[BB19:%.*]]
 ; IPSCCP-NEXT:    ]
 ; IPSCCP:       bb3:
 ; IPSCCP-NEXT:    unreachable
@@ -486,7 +486,7 @@ define void @foo(ptr %arg) {
 ; IPSCCP-NEXT:    [[TMP6:%.*]] = and i64 [[TMP5]], 3
 ; IPSCCP-NEXT:    [[TMP7:%.*]] = sub nuw nsw i64 3, [[TMP6]]
 ; IPSCCP-NEXT:    [[TMP8:%.*]] = shl nuw nsw i64 [[TMP7]], 1
-; IPSCCP-NEXT:    [[TMP9:%.*]] = trunc i64 [[TMP8]] to i32
+; IPSCCP-NEXT:    [[TMP9:%.*]] = trunc nuw nsw i64 [[TMP8]] to i32
 ; IPSCCP-NEXT:    [[TMP10:%.*]] = zext nneg i32 [[TMP9]] to i64
 ; IPSCCP-NEXT:    br label [[BB11:%.*]]
 ; IPSCCP:       bb11:
@@ -615,7 +615,7 @@ define ptr @wobble(ptr %arg, i32 %arg1) align 2 {
 ; SCCP-NEXT:    [[TMP22:%.*]] = icmp eq i32 [[TMP21]], 0
 ; SCCP-NEXT:    br i1 [[TMP22]], label [[BB23:%.*]], label [[BB25:%.*]]
 ; SCCP:       bb23:
-; SCCP-NEXT:    [[TMP24:%.*]] = trunc i32 [[TMP3]] to i16
+; SCCP-NEXT:    [[TMP24:%.*]] = trunc nuw i32 [[TMP3]] to i16
 ; SCCP-NEXT:    store i16 [[TMP24]], ptr [[TMP17]], align 2
 ; SCCP-NEXT:    br label [[BB31]]
 ; SCCP:       bb25:
@@ -639,7 +639,7 @@ define ptr @wobble(ptr %arg, i32 %arg1) align 2 {
 ; SCCP-NEXT:    br i1 [[C_2]], label [[BB39:%.*]], label [[BB58:%.*]]
 ; SCCP:       bb39:
 ; SCCP-NEXT:    [[TMP40:%.*]] = add nsw i32 [[TMP11]], -1
-; SCCP-NEXT:    [[TMP41:%.*]] = trunc i32 [[TMP3]] to i16
+; SCCP-NEXT:    [[TMP41:%.*]] = trunc nuw i32 [[TMP3]] to i16
 ; SCCP-NEXT:    store i16 [[TMP41]], ptr @global.11, align 1
 ; SCCP-NEXT:    [[TMP43:%.*]] = add i32 [[TMP7]], [[TMP40]]
 ; SCCP-NEXT:    [[TMP44:%.*]] = mul i32 [[TMP43]], 4
@@ -697,7 +697,7 @@ define ptr @wobble(ptr %arg, i32 %arg1) align 2 {
 ; IPSCCP-NEXT:    [[TMP22:%.*]] = icmp eq i32 [[TMP21]], 0
 ; IPSCCP-NEXT:    br i1 [[TMP22]], label [[BB23:%.*]], label [[BB25:%.*]]
 ; IPSCCP:       bb23:
-; IPSCCP-NEXT:    [[TMP24:%.*]] = trunc i32 [[TMP3]] to i16
+; IPSCCP-NEXT:    [[TMP24:%.*]] = trunc nuw i32 [[TMP3]] to i16
 ; IPSCCP-NEXT:    store i16 [[TMP24]], ptr [[TMP17]], align 2
 ; IPSCCP-NEXT:    br label [[BB31]]
 ; IPSCCP:       bb25:
@@ -720,7 +720,7 @@ define ptr @wobble(ptr %arg, i32 %arg1) align 2 {
 ; IPSCCP-NEXT:    [[C_2:%.*]] = icmp eq i32 [[TMP11]], 8
 ; IPSCCP-NEXT:    br i1 [[C_2]], label [[BB39:%.*]], label [[BB58:%.*]]
 ; IPSCCP:       bb39:
-; IPSCCP-NEXT:    [[TMP41:%.*]] = trunc i32 [[TMP3]] to i16
+; IPSCCP-NEXT:    [[TMP41:%.*]] = trunc nuw i32 [[TMP3]] to i16
 ; IPSCCP-NEXT:    store i16 [[TMP41]], ptr @global.11, align 1
 ; IPSCCP-NEXT:    [[TMP43:%.*]] = add i32 [[TMP7]], 7
 ; IPSCCP-NEXT:    [[TMP44:%.*]] = mul i32 [[TMP43]], 4

From 5bed6afc21c9ff9dba4aabf5ac74cb12971f634f Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Thu, 11 Apr 2024 14:44:11 -0400
Subject: [PATCH 179/886] [BOLT][NFC] Remove unneeded if (#88322)

No need need to special-case zero. Section 0 will map to section 0.
---
 bolt/lib/Rewrite/RewriteInstance.cpp | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index 1f778d030f7ac..eea66454b289c 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -4419,6 +4419,7 @@ void RewriteInstance::patchELFSectionHeaderTable(ELFObjectFile<ELFT> *File) {
   raw_fd_ostream &OS = Out->os();
   const ELFFile<ELFT> &Obj = File->getELFFile();
 
+  // Mapping from old section indices to new ones
   std::vector<uint32_t> NewSectionIndex;
   std::vector<ELFShdrTy> OutputSections =
       getOutputSections(File, NewSectionIndex);
@@ -4436,10 +4437,8 @@ void RewriteInstance::patchELFSectionHeaderTable(ELFObjectFile<ELFT> *File) {
   // Write all section header entries while patching section references.
   for (ELFShdrTy &Section : OutputSections) {
     Section.sh_link = NewSectionIndex[Section.sh_link];
-    if (Section.sh_type == ELF::SHT_REL || Section.sh_type == ELF::SHT_RELA) {
-      if (Section.sh_info)
-        Section.sh_info = NewSectionIndex[Section.sh_info];
-    }
+    if (Section.sh_type == ELF::SHT_REL || Section.sh_type == ELF::SHT_RELA)
+      Section.sh_info = NewSectionIndex[Section.sh_info];
     OS.write(reinterpret_cast<const char *>(&Section), sizeof(Section));
   }
 

From 6b85fb1ef8586181fbdb02ca11a3bb0e9006aabd Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 11 Apr 2024 14:56:39 -0400
Subject: [PATCH 180/886] [SLP]Consider (f)sub, being operand of
 llvm.(f)abs/icmp eq/ne 0, commutative.

If (f)sub is only operand of llvm.(f)abs or icmp eq/ne 0 (int only), we can consider it as commutative operation, just need to drop wrapping flags for ineteger
operation.

https://alive2.llvm.org/ce/z/GxvxjB for correctness of abs with dropped
flags.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/86196
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 51 +++++++++++++++++--
 .../X86/store-abs-minbitwidth.ll              | 14 +----
 2 files changed, 50 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 2995cb8435f76..55dc0867e2ea9 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -305,7 +305,32 @@ static bool isCommutative(Instruction *I) {
   if (auto *Cmp = dyn_cast<CmpInst>(I))
     return Cmp->isCommutative();
   if (auto *BO = dyn_cast<BinaryOperator>(I))
-    return BO->isCommutative();
+    return BO->isCommutative() ||
+           (BO->getOpcode() == Instruction::Sub &&
+            !BO->hasNUsesOrMore(UsesLimit) &&
+            all_of(
+                BO->uses(),
+                [](const Use &U) {
+                  // Commutative, if icmp eq/ne sub, 0
+                  ICmpInst::Predicate Pred;
+                  if (match(U.getUser(),
+                            m_ICmp(Pred, m_Specific(U.get()), m_Zero())) &&
+                      (Pred == ICmpInst::ICMP_EQ || Pred == ICmpInst::ICMP_NE))
+                    return true;
+                  // Commutative, if abs(sub nsw, true) or abs(sub, false).
+                  ConstantInt *Flag;
+                  return match(U.getUser(),
+                               m_Intrinsic<Intrinsic::abs>(
+                                   m_Specific(U.get()), m_ConstantInt(Flag))) &&
+                         (!cast<Instruction>(U.get())->hasNoSignedWrap() ||
+                          Flag->isOne());
+                })) ||
+           (BO->getOpcode() == Instruction::FSub &&
+            !BO->hasNUsesOrMore(UsesLimit) &&
+            all_of(BO->uses(), [](const Use &U) {
+              return match(U.getUser(),
+                           m_Intrinsic<Intrinsic::fabs>(m_Specific(U.get())));
+            }));
   return I->isCommutative();
 }
 
@@ -6838,7 +6863,7 @@ void BoUpSLP::buildTree_rec(ArrayRef<Value *> VL, unsigned Depth,
 
       // Sort operands of the instructions so that each side is more likely to
       // have the same opcode.
-      if (isa<BinaryOperator>(VL0) && VL0->isCommutative()) {
+      if (isa<BinaryOperator>(VL0) && isCommutative(VL0)) {
         ValueList Left, Right;
         reorderInputsAccordingToOpcode(VL, Left, Right, *this);
         TE->setOperand(0, Left);
@@ -12566,8 +12591,15 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
           static_cast<Instruction::BinaryOps>(E->getOpcode()), LHS,
           RHS);
       propagateIRFlags(V, E->Scalars, VL0, It == MinBWs.end());
-      if (auto *I = dyn_cast<Instruction>(V))
+      if (auto *I = dyn_cast<Instruction>(V)) {
         V = propagateMetadata(I, E->Scalars);
+        // Drop nuw flags for abs(sub(commutative), true).
+        if (!MinBWs.contains(E) && ShuffleOrOp == Instruction::Sub &&
+            any_of(E->Scalars, [](Value *V) {
+              return isCommutative(cast<Instruction>(V));
+            }))
+          I->setHasNoUnsignedWrap(/*b=*/false);
+      }
 
       V = FinalShuffle(V, E, VecTy);
 
@@ -12893,6 +12925,19 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
 
       propagateIRFlags(V0, OpScalars, E->getMainOp(), It == MinBWs.end());
       propagateIRFlags(V1, AltScalars, E->getAltOp(), It == MinBWs.end());
+      auto DropNuwFlag = [&](Value *Vec, unsigned Opcode) {
+        // Drop nuw flags for abs(sub(commutative), true).
+        if (auto *I = dyn_cast<Instruction>(Vec);
+            I && Opcode == Instruction::Sub && !MinBWs.contains(E) &&
+            any_of(E->Scalars, [](Value *V) {
+              auto *I = cast<Instruction>(V);
+              return I->getOpcode() == Instruction::Sub &&
+                     isCommutative(cast<Instruction>(V));
+            }))
+          I->setHasNoUnsignedWrap(/*b=*/false);
+      };
+      DropNuwFlag(V0, E->getOpcode());
+      DropNuwFlag(V1, E->getAltOpcode());
 
       Value *V = Builder.CreateShuffleVector(V0, V1, Mask);
       if (auto *I = dyn_cast<Instruction>(V)) {
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
index df7312e3d2b56..64bfb242752c1 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/store-abs-minbitwidth.ll
@@ -4,19 +4,9 @@
 
 define i32 @test(ptr noalias %in, ptr noalias %inn, ptr %out) {
 ; CHECK-LABEL: @test(
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i8>, ptr [[IN:%.*]], align 1
-; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i8, ptr [[IN]], i64 2
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i8>, ptr [[GEP_2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i8>, ptr [[INN:%.*]], align 1
-; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i8, ptr [[INN]], i64 2
-; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i8>, ptr [[GEP_5]], align 1
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i8> [[TMP2]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i8> [[TMP5]], <4 x i8> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i8>, ptr [[IN:%.*]], align 1
+; CHECK-NEXT:    [[TMP11:%.*]] = load <4 x i8>, ptr [[INN:%.*]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i8> [[TMP7]] to <4 x i16>
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x i8> [[TMP1]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x i8> [[TMP4]], <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i8> [[TMP9]], <4 x i8> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
 ; CHECK-NEXT:    [[TMP12:%.*]] = sext <4 x i8> [[TMP11]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP13:%.*]] = sub <4 x i16> [[TMP12]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = call <4 x i16> @llvm.abs.v4i16(<4 x i16> [[TMP13]], i1 false)

From a5eaec83b3446f217169fdd94e03ec67823af058 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 11 Apr 2024 12:01:38 -0700
Subject: [PATCH 181/886] [SLP]Fix variable redefinition error

---
 llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 55dc0867e2ea9..df891371fdf75 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -12930,9 +12930,9 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
         if (auto *I = dyn_cast<Instruction>(Vec);
             I && Opcode == Instruction::Sub && !MinBWs.contains(E) &&
             any_of(E->Scalars, [](Value *V) {
-              auto *I = cast<Instruction>(V);
-              return I->getOpcode() == Instruction::Sub &&
-                     isCommutative(cast<Instruction>(V));
+              auto *IV = cast<Instruction>(V);
+              return IV->getOpcode() == Instruction::Sub &&
+                     isCommutative(cast<Instruction>(IV));
             }))
           I->setHasNoUnsignedWrap(/*b=*/false);
       };

From 8840992667180d327ff40ee24354ab8b49605aae Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov@fb.com>
Date: Thu, 11 Apr 2024 21:07:36 +0200
Subject: [PATCH 182/886] [BOLT][BAT] Fix handling of split functions

Move BAT parent function lookup outside `getLocationName`, to the
scope where we retrieve `FuncBranchData` linked with the function.

Previously DataAggregator would store branch profile recorded in the
split fragment in `FuncBranchData` associated with the fragment, and
perform name translation in `getLocationName` for symbol name only.
This works for fdata profile which is printed out as-is, but doesn't
work with BAT YAML profile writer which requires a combined profile.

The issue necessitated `fixupBATProfile` which partially addressed the
issue (reassigned inter-fragment calls back into intra-function
branches). However, `fixupBATProfile` fails to address disjoint
profiles (i.e. doesn't merge `FuncBranchData` for fragments back
into parent). This diff eliminates the need for `fixupBATProfile` by
removing the root cause of the issue.

Test Plan: NFC for existing tests

Reviewers: ayermolo, dcci, rafaelauler, maksfb

Reviewed By: maksfb

Pull Request: https://github.com/llvm/llvm-project/pull/87569
---
 bolt/include/bolt/Profile/DataAggregator.h    |  19 ++-
 bolt/lib/Profile/DataAggregator.cpp           | 114 ++++++++----------
 .../X86/yaml-secondary-entry-discriminator.s  |  16 +--
 3 files changed, 66 insertions(+), 83 deletions(-)

diff --git a/bolt/include/bolt/Profile/DataAggregator.h b/bolt/include/bolt/Profile/DataAggregator.h
index 4fbe524b1c385..84f76caae9dbb 100644
--- a/bolt/include/bolt/Profile/DataAggregator.h
+++ b/bolt/include/bolt/Profile/DataAggregator.h
@@ -225,6 +225,10 @@ class DataAggregator : public DataReader {
   /// Aggregation statistics
   uint64_t NumInvalidTraces{0};
   uint64_t NumLongRangeTraces{0};
+  /// Specifies how many samples were recorded in cold areas if we are dealing
+  /// with profiling data collected in a bolted binary. For LBRs, incremented
+  /// for the source of the branch to avoid counting cold activity twice (one
+  /// for source and another for destination).
   uint64_t NumColdSamples{0};
 
   /// Looks into system PATH for Linux Perf and set up the aggregator to use it
@@ -245,14 +249,12 @@ class DataAggregator : public DataReader {
   /// disassembled BinaryFunctions
   BinaryFunction *getBinaryFunctionContainingAddress(uint64_t Address) const;
 
+  /// Perform BAT translation for a given \p Func and return the parent
+  /// BinaryFunction or nullptr.
+  BinaryFunction *getBATParentFunction(const BinaryFunction &Func) const;
+
   /// Retrieve the location name to be used for samples recorded in \p Func.
-  /// If doing BAT translation, link cold parts to the hot part  names (used by
-  /// the original binary).  \p Count specifies how many samples were recorded
-  /// at that location, so we can tally total activity in cold areas if we are
-  /// dealing with profiling data collected in a bolted binary. For LBRs,
-  /// \p Count should only be used for the source of the branch to avoid
-  /// counting cold activity twice (one for source and another for destination).
-  StringRef getLocationName(BinaryFunction &Func, uint64_t Count);
+  StringRef getLocationName(const BinaryFunction &Func) const;
 
   /// Semantic actions - parser hooks to interpret parsed perf samples
   /// Register a sample (non-LBR mode), i.e. a new hit at \p Address
@@ -467,9 +469,6 @@ class DataAggregator : public DataReader {
   std::error_code writeBATYAML(BinaryContext &BC,
                                StringRef OutputFilename) const;
 
-  /// Fixup profile collected on BOLTed binary, namely handle split functions.
-  void fixupBATProfile(BinaryContext &BC);
-
   /// Filter out binaries based on PID
   void filterBinaryMMapInfo();
 
diff --git a/bolt/lib/Profile/DataAggregator.cpp b/bolt/lib/Profile/DataAggregator.cpp
index 71824e2cc0e97..6ebf50dae7d96 100644
--- a/bolt/lib/Profile/DataAggregator.cpp
+++ b/bolt/lib/Profile/DataAggregator.cpp
@@ -604,8 +604,6 @@ Error DataAggregator::readProfile(BinaryContext &BC) {
     // BAT YAML is handled by DataAggregator since normal YAML output requires
     // CFG which is not available in BAT mode.
     if (usesBAT()) {
-      // Postprocess split function profile for BAT
-      fixupBATProfile(BC);
       if (opts::ProfileFormat == opts::ProfileFormatKind::PF_YAML)
         if (std::error_code EC = writeBATYAML(BC, opts::OutputFilename))
           report_error("cannot create output data file", EC);
@@ -664,18 +662,19 @@ DataAggregator::getBinaryFunctionContainingAddress(uint64_t Address) const {
                                                 /*UseMaxSize=*/true);
 }
 
-StringRef DataAggregator::getLocationName(BinaryFunction &Func,
-                                          uint64_t Count) {
+BinaryFunction *
+DataAggregator::getBATParentFunction(const BinaryFunction &Func) const {
+  if (BAT)
+    if (const uint64_t HotAddr = BAT->fetchParentAddress(Func.getAddress()))
+      return getBinaryFunctionContainingAddress(HotAddr);
+  return nullptr;
+}
+
+StringRef DataAggregator::getLocationName(const BinaryFunction &Func) const {
   if (!BAT)
     return Func.getOneName();
 
   const BinaryFunction *OrigFunc = &Func;
-  if (const uint64_t HotAddr = BAT->fetchParentAddress(Func.getAddress())) {
-    NumColdSamples += Count;
-    BinaryFunction *HotFunc = getBinaryFunctionContainingAddress(HotAddr);
-    if (HotFunc)
-      OrigFunc = HotFunc;
-  }
   // If it is a local function, prefer the name containing the file name where
   // the local function was declared
   for (StringRef AlternativeName : OrigFunc->getNames()) {
@@ -690,12 +689,17 @@ StringRef DataAggregator::getLocationName(BinaryFunction &Func,
   return OrigFunc->getOneName();
 }
 
-bool DataAggregator::doSample(BinaryFunction &Func, uint64_t Address,
+bool DataAggregator::doSample(BinaryFunction &OrigFunc, uint64_t Address,
                               uint64_t Count) {
+  BinaryFunction *ParentFunc = getBATParentFunction(OrigFunc);
+  BinaryFunction &Func = ParentFunc ? *ParentFunc : OrigFunc;
+  if (ParentFunc)
+    NumColdSamples += Count;
+
   auto I = NamesToSamples.find(Func.getOneName());
   if (I == NamesToSamples.end()) {
     bool Success;
-    StringRef LocName = getLocationName(Func, Count);
+    StringRef LocName = getLocationName(Func);
     std::tie(I, Success) = NamesToSamples.insert(
         std::make_pair(Func.getOneName(),
                        FuncSampleData(LocName, FuncSampleData::ContainerTy())));
@@ -715,22 +719,12 @@ bool DataAggregator::doIntraBranch(BinaryFunction &Func, uint64_t From,
   FuncBranchData *AggrData = getBranchData(Func);
   if (!AggrData) {
     AggrData = &NamesToBranches[Func.getOneName()];
-    AggrData->Name = getLocationName(Func, Count);
+    AggrData->Name = getLocationName(Func);
     setBranchData(Func, AggrData);
   }
 
-  From -= Func.getAddress();
-  To -= Func.getAddress();
   LLVM_DEBUG(dbgs() << "BOLT-DEBUG: bumpBranchCount: "
                     << formatv("{0} @ {1:x} -> {0} @ {2:x}\n", Func, From, To));
-  if (BAT) {
-    From = BAT->translate(Func.getAddress(), From, /*IsBranchSrc=*/true);
-    To = BAT->translate(Func.getAddress(), To, /*IsBranchSrc=*/false);
-    LLVM_DEBUG(
-        dbgs() << "BOLT-DEBUG: BAT translation on bumpBranchCount: "
-               << formatv("{0} @ {1:x} -> {0} @ {2:x}\n", Func, From, To));
-  }
-
   AggrData->bumpBranchCount(From, To, Count, Mispreds);
   return true;
 }
@@ -744,30 +738,24 @@ bool DataAggregator::doInterBranch(BinaryFunction *FromFunc,
   StringRef SrcFunc;
   StringRef DstFunc;
   if (FromFunc) {
-    SrcFunc = getLocationName(*FromFunc, Count);
+    SrcFunc = getLocationName(*FromFunc);
     FromAggrData = getBranchData(*FromFunc);
     if (!FromAggrData) {
       FromAggrData = &NamesToBranches[FromFunc->getOneName()];
       FromAggrData->Name = SrcFunc;
       setBranchData(*FromFunc, FromAggrData);
     }
-    From -= FromFunc->getAddress();
-    if (BAT)
-      From = BAT->translate(FromFunc->getAddress(), From, /*IsBranchSrc=*/true);
 
     recordExit(*FromFunc, From, Mispreds, Count);
   }
   if (ToFunc) {
-    DstFunc = getLocationName(*ToFunc, 0);
+    DstFunc = getLocationName(*ToFunc);
     ToAggrData = getBranchData(*ToFunc);
     if (!ToAggrData) {
       ToAggrData = &NamesToBranches[ToFunc->getOneName()];
       ToAggrData->Name = DstFunc;
       setBranchData(*ToFunc, ToAggrData);
     }
-    To -= ToFunc->getAddress();
-    if (BAT)
-      To = BAT->translate(ToFunc->getAddress(), To, /*IsBranchSrc=*/false);
 
     recordEntry(*ToFunc, To, Mispreds, Count);
   }
@@ -783,15 +771,32 @@ bool DataAggregator::doInterBranch(BinaryFunction *FromFunc,
 
 bool DataAggregator::doBranch(uint64_t From, uint64_t To, uint64_t Count,
                               uint64_t Mispreds) {
-  BinaryFunction *FromFunc = getBinaryFunctionContainingAddress(From);
-  BinaryFunction *ToFunc = getBinaryFunctionContainingAddress(To);
+  auto handleAddress = [&](uint64_t &Addr, bool IsFrom) -> BinaryFunction * {
+    if (BinaryFunction *Func = getBinaryFunctionContainingAddress(Addr)) {
+      Addr -= Func->getAddress();
+
+      if (BAT)
+        Addr = BAT->translate(Func->getAddress(), Addr, IsFrom);
+
+      if (BinaryFunction *ParentFunc = getBATParentFunction(*Func)) {
+        Func = ParentFunc;
+        if (IsFrom)
+          NumColdSamples += Count;
+      }
+
+      return Func;
+    }
+    return nullptr;
+  };
+
+  BinaryFunction *FromFunc = handleAddress(From, /*IsFrom=*/true);
+  BinaryFunction *ToFunc = handleAddress(To, /*IsFrom=*/false);
   if (!FromFunc && !ToFunc)
     return false;
 
   // Treat recursive control transfers as inter-branches.
-  if (FromFunc == ToFunc && (To != ToFunc->getAddress())) {
-    recordBranch(*FromFunc, From - FromFunc->getAddress(),
-                 To - FromFunc->getAddress(), Count, Mispreds);
+  if (FromFunc == ToFunc && To != 0) {
+    recordBranch(*FromFunc, From, To, Count, Mispreds);
     return doIntraBranch(*FromFunc, From, To, Count, Mispreds);
   }
 
@@ -842,9 +847,14 @@ bool DataAggregator::doTrace(const LBREntry &First, const LBREntry &Second,
                     << FromFunc->getPrintName() << ":"
                     << Twine::utohexstr(First.To) << " to "
                     << Twine::utohexstr(Second.From) << ".\n");
-  for (const std::pair<uint64_t, uint64_t> &Pair : *FTs)
-    doIntraBranch(*FromFunc, Pair.first + FromFunc->getAddress(),
-                  Pair.second + FromFunc->getAddress(), Count, false);
+  BinaryFunction *ParentFunc = getBATParentFunction(*FromFunc);
+  for (auto [From, To] : *FTs) {
+    if (BAT) {
+      From = BAT->translate(FromFunc->getAddress(), From, /*IsBranchSrc=*/true);
+      To = BAT->translate(FromFunc->getAddress(), To, /*IsBranchSrc=*/false);
+    }
+    doIntraBranch(ParentFunc ? *ParentFunc : *FromFunc, From, To, Count, false);
+  }
 
   return true;
 }
@@ -2273,29 +2283,6 @@ DataAggregator::writeAggregatedFile(StringRef OutputFilename) const {
   return std::error_code();
 }
 
-void DataAggregator::fixupBATProfile(BinaryContext &BC) {
-  for (auto &[FuncName, Branches] : NamesToBranches) {
-    BinaryData *BD = BC.getBinaryDataByName(FuncName);
-    assert(BD);
-    uint64_t FuncAddress = BD->getAddress();
-    if (!BAT->isBATFunction(FuncAddress))
-      continue;
-    // Filter out cold fragments
-    if (!BD->getSectionName().equals(BC.getMainCodeSectionName()))
-      continue;
-    // Convert inter-branches between hot and cold fragments into
-    // intra-branches.
-    for (auto &[OffsetFrom, CallToMap] : Branches.InterIndex) {
-      for (auto &[CallToLoc, CallToIdx] : CallToMap) {
-        if (CallToLoc.Name != FuncName)
-          continue;
-        Branches.IntraIndex[OffsetFrom][CallToLoc.Offset] = CallToIdx;
-        Branches.InterIndex[OffsetFrom].erase(CallToLoc);
-      }
-    }
-  }
-}
-
 std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
                                              StringRef OutputFilename) const {
   std::error_code EC;
@@ -2345,9 +2332,6 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
       uint64_t FuncAddress = BD->getAddress();
       if (!BAT->isBATFunction(FuncAddress))
         continue;
-      // Filter out cold fragments
-      if (!BD->getSectionName().equals(BC.getMainCodeSectionName()))
-        continue;
       BinaryFunction *BF = BC.getBinaryFunctionAtAddress(FuncAddress);
       assert(BF);
       YamlBF.Name = FuncName.str();
diff --git a/bolt/test/X86/yaml-secondary-entry-discriminator.s b/bolt/test/X86/yaml-secondary-entry-discriminator.s
index 78e7e55aa98eb..7fd5a716cd68e 100644
--- a/bolt/test/X86/yaml-secondary-entry-discriminator.s
+++ b/bolt/test/X86/yaml-secondary-entry-discriminator.s
@@ -1,5 +1,5 @@
-# This reproduces a bug with BOLT setting incorrect discriminator for
-# secondary entry points in YAML profile.
+## This reproduces a bug with BOLT setting incorrect discriminator for
+## secondary entry points in YAML profile.
 
 # REQUIRES: system-linux
 # RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
@@ -24,7 +24,7 @@
 # CHECK-NEXT:       hash:  {{.*}}
 # CHECK-NEXT:       calls: [ { off: 0x0, fid: 1, disc: 1, cnt: 1, mis: 1 } ]
 
-# Make sure that the profile is attached correctly
+## Make sure that the profile is attached correctly
 # RUN: llvm-bolt %t.exe -o %t.out --data %t.yaml --print-profile \
 # RUN:   --print-only=main | FileCheck %s --check-prefix=CHECK-CFG
 
@@ -33,8 +33,8 @@
 # CHECK-CFG:      callq *%rax # Offset: [[#]] # CallProfile: 1 (1 misses) :
 # CHECK-CFG-NEXT:     { secondary_entry: 1 (1 misses) }
 
-# YAML BAT test of calling BAT secondary entry from non-BAT function
-# Now force-split func and skip main (making it call secondary entries)
+## YAML BAT test of calling BAT secondary entry from non-BAT function
+## Now force-split func and skip main (making it call secondary entries)
 # RUN: llvm-bolt %t.exe -o %t.bat --data %t.fdata --funcs=func \
 # RUN:   --split-functions --split-strategy=all --split-all-cold --enable-bat
 
@@ -45,13 +45,13 @@ func:
   .cfi_startproc
   pushq   %rbp
   movq    %rsp, %rbp
-  # Placeholder code to make splitting profitable
+## Placeholder code to make splitting profitable
 .rept 5
   testq   %rax, %rax
 .endr
 .globl secondary_entry
 secondary_entry:
-  # Placeholder code to make splitting profitable
+## Placeholder code to make splitting profitable
 .rept 5
   testq   %rax, %rax
 .endr
@@ -81,7 +81,7 @@ Lindcall:
   addq    $16, %rsp
   popq    %rbp
   retq
-# For relocations against .text
+## For relocations against .text
   call exit
   .cfi_endproc
   .size	main, .-main

From 3997f0eb8199c89f8224cca1ecdd3d45c1bbe1fc Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov@fb.com>
Date: Thu, 11 Apr 2024 21:15:04 +0200
Subject: [PATCH 183/886] [BOLT] Cover all call sites in writeBATYAML

Call site information setting was conditioned on branch information
presence for a given block. However, it's possible to have sampled
profile lacking one or the other for a given basic block.

Iterate over branch profiles and call profiles independently to cover
all recorded profile data.

Depends on https://github.com/llvm/llvm-project/pull/87569

Test Plan: Updated bolt/test/X86/yaml-secondary-entry-discriminator.s

Reviewers: ayermolo, dcci, maksfb, rafaelauler

Reviewed By: maksfb

Pull Request: https://github.com/llvm/llvm-project/pull/87743
---
 .../bolt/Profile/BoltAddressTranslation.h     |  11 +-
 bolt/lib/Profile/BoltAddressTranslation.cpp   |  20 ----
 bolt/lib/Profile/DataAggregator.cpp           | 113 ++++++++----------
 .../X86/yaml-secondary-entry-discriminator.s  |  58 +++++++++
 4 files changed, 110 insertions(+), 92 deletions(-)

diff --git a/bolt/include/bolt/Profile/BoltAddressTranslation.h b/bolt/include/bolt/Profile/BoltAddressTranslation.h
index 6a0a477f27583..eef05e8a0e681 100644
--- a/bolt/include/bolt/Profile/BoltAddressTranslation.h
+++ b/bolt/include/bolt/Profile/BoltAddressTranslation.h
@@ -119,11 +119,6 @@ class BoltAddressTranslation {
   /// True if a given \p Address is a function with translation table entry.
   bool isBATFunction(uint64_t Address) const { return Maps.count(Address); }
 
-  /// Returns branch offsets grouped by containing basic block in a given
-  /// function.
-  std::unordered_map<uint32_t, std::vector<uint32_t>>
-  getBFBranches(uint64_t FuncOutputAddress) const;
-
   /// For a given \p Symbol in the output binary and known \p InputOffset
   /// return a corresponding pair of parent BinaryFunction and secondary entry
   /// point in it.
@@ -193,7 +188,7 @@ class BoltAddressTranslation {
       EntryTy(unsigned Index, size_t Hash) : Index(Index), Hash(Hash) {}
     };
 
-    std::unordered_map<uint32_t, EntryTy> Map;
+    std::map<uint32_t, EntryTy> Map;
     const EntryTy &getEntry(uint32_t BBInputOffset) const {
       auto It = Map.find(BBInputOffset);
       assert(It != Map.end());
@@ -218,6 +213,10 @@ class BoltAddressTranslation {
     }
 
     size_t getNumBasicBlocks() const { return Map.size(); }
+
+    auto begin() const { return Map.begin(); }
+    auto end() const { return Map.end(); }
+    auto upper_bound(uint32_t Offset) const { return Map.upper_bound(Offset); }
   };
 
   /// Map function output address to its hash and basic blocks hash map.
diff --git a/bolt/lib/Profile/BoltAddressTranslation.cpp b/bolt/lib/Profile/BoltAddressTranslation.cpp
index 59d499f97be72..0141ce189acda 100644
--- a/bolt/lib/Profile/BoltAddressTranslation.cpp
+++ b/bolt/lib/Profile/BoltAddressTranslation.cpp
@@ -582,26 +582,6 @@ void BoltAddressTranslation::saveMetadata(BinaryContext &BC) {
   }
 }
 
-std::unordered_map<uint32_t, std::vector<uint32_t>>
-BoltAddressTranslation::getBFBranches(uint64_t OutputAddress) const {
-  std::unordered_map<uint32_t, std::vector<uint32_t>> Branches;
-  auto FuncIt = Maps.find(OutputAddress);
-  assert(FuncIt != Maps.end());
-  std::vector<uint32_t> InputOffsets;
-  for (const auto &KV : FuncIt->second)
-    InputOffsets.emplace_back(KV.second);
-  // Sort with LSB BRANCHENTRY bit.
-  llvm::sort(InputOffsets);
-  uint32_t BBOffset{0};
-  for (uint32_t InOffset : InputOffsets) {
-    if (InOffset & BRANCHENTRY)
-      Branches[BBOffset].push_back(InOffset >> 1);
-    else
-      BBOffset = InOffset >> 1;
-  }
-  return Branches;
-}
-
 unsigned
 BoltAddressTranslation::getSecondaryEntryPointId(uint64_t Address,
                                                  uint32_t Offset) const {
diff --git a/bolt/lib/Profile/DataAggregator.cpp b/bolt/lib/Profile/DataAggregator.cpp
index 6ebf50dae7d96..0b2a4e86561f3 100644
--- a/bolt/lib/Profile/DataAggregator.cpp
+++ b/bolt/lib/Profile/DataAggregator.cpp
@@ -2341,87 +2341,68 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
       YamlBF.NumBasicBlocks = BAT->getNumBasicBlocks(FuncAddress);
       const BoltAddressTranslation::BBHashMapTy &BlockMap =
           BAT->getBBHashMap(FuncAddress);
+      YamlBF.Blocks.resize(YamlBF.NumBasicBlocks);
 
-      auto addSuccProfile = [&](yaml::bolt::BinaryBasicBlockProfile &YamlBB,
-                                uint64_t SuccOffset, unsigned SuccDataIdx) {
+      for (auto &&[Idx, YamlBB] : llvm::enumerate(YamlBF.Blocks))
+        YamlBB.Index = Idx;
+
+      for (auto BI = BlockMap.begin(), BE = BlockMap.end(); BI != BE; ++BI)
+        YamlBF.Blocks[BI->second.getBBIndex()].Hash = BI->second.getBBHash();
+
+      auto getSuccessorInfo = [&](uint32_t SuccOffset, unsigned SuccDataIdx) {
         const llvm::bolt::BranchInfo &BI = Branches.Data.at(SuccDataIdx);
         yaml::bolt::SuccessorInfo SI;
         SI.Index = BlockMap.getBBIndex(SuccOffset);
         SI.Count = BI.Branches;
         SI.Mispreds = BI.Mispreds;
-        YamlBB.Successors.emplace_back(SI);
+        return SI;
       };
 
-      std::unordered_map<uint32_t, std::vector<uint32_t>> BFBranches =
-          BAT->getBFBranches(FuncAddress);
-
-      auto addCallsProfile = [&](yaml::bolt::BinaryBasicBlockProfile &YamlBB,
-                                 uint64_t Offset) {
-        // Iterate over BRANCHENTRY records in the current block
-        for (uint32_t BranchOffset : BFBranches[Offset]) {
-          if (!Branches.InterIndex.contains(BranchOffset))
-            continue;
-          for (const auto &[CallToLoc, CallToIdx] :
-               Branches.InterIndex.at(BranchOffset)) {
-            const llvm::bolt::BranchInfo &BI = Branches.Data.at(CallToIdx);
-            yaml::bolt::CallSiteInfo YamlCSI;
-            YamlCSI.DestId = 0; // designated for unknown functions
-            YamlCSI.EntryDiscriminator = 0;
-            YamlCSI.Count = BI.Branches;
-            YamlCSI.Mispreds = BI.Mispreds;
-            YamlCSI.Offset = BranchOffset - Offset;
-            BinaryData *CallTargetBD = BC.getBinaryDataByName(CallToLoc.Name);
-            if (!CallTargetBD) {
-              YamlBB.CallSites.emplace_back(YamlCSI);
-              continue;
-            }
-            uint64_t CallTargetAddress = CallTargetBD->getAddress();
-            BinaryFunction *CallTargetBF =
-                BC.getBinaryFunctionAtAddress(CallTargetAddress);
-            if (!CallTargetBF) {
-              YamlBB.CallSites.emplace_back(YamlCSI);
-              continue;
-            }
-            // Calls between hot and cold fragments must be handled in
-            // fixupBATProfile.
-            assert(CallTargetBF != BF && "invalid CallTargetBF");
-            YamlCSI.DestId = CallTargetBF->getFunctionNumber();
-            if (CallToLoc.Offset) {
-              if (BAT->isBATFunction(CallTargetAddress)) {
-                LLVM_DEBUG(dbgs() << "BOLT-DEBUG: Unsupported secondary "
-                                     "entry point in BAT function "
-                                  << CallToLoc.Name << '\n');
-              } else if (const BinaryBasicBlock *CallTargetBB =
-                             CallTargetBF->getBasicBlockAtOffset(
-                                 CallToLoc.Offset)) {
-                // Only record true call information, ignoring returns (normally
-                // won't have a target basic block) and jumps to the landing
-                // pads (not an entry point).
-                if (CallTargetBB->isEntryPoint()) {
-                  YamlCSI.EntryDiscriminator =
-                      CallTargetBF->getEntryIDForSymbol(
-                          CallTargetBB->getLabel());
-                }
-              }
-            }
-            YamlBB.CallSites.emplace_back(YamlCSI);
-          }
-        }
+      auto getCallSiteInfo = [&](Location CallToLoc, unsigned CallToIdx,
+                                 uint32_t Offset) {
+        const llvm::bolt::BranchInfo &BI = Branches.Data.at(CallToIdx);
+        yaml::bolt::CallSiteInfo CSI;
+        CSI.DestId = 0; // designated for unknown functions
+        CSI.EntryDiscriminator = 0;
+        CSI.Count = BI.Branches;
+        CSI.Mispreds = BI.Mispreds;
+        CSI.Offset = Offset;
+        if (BinaryData *BD = BC.getBinaryDataByName(CallToLoc.Name))
+          YAMLProfileWriter::setCSIDestination(BC, CSI, BD->getSymbol(), BAT,
+                                               CallToLoc.Offset);
+        return CSI;
       };
 
       for (const auto &[FromOffset, SuccKV] : Branches.IntraIndex) {
-        yaml::bolt::BinaryBasicBlockProfile YamlBB;
         if (!BlockMap.isInputBlock(FromOffset))
           continue;
-        YamlBB.Index = BlockMap.getBBIndex(FromOffset);
-        YamlBB.Hash = BlockMap.getBBHash(FromOffset);
+        const unsigned Index = BlockMap.getBBIndex(FromOffset);
+        yaml::bolt::BinaryBasicBlockProfile &YamlBB = YamlBF.Blocks[Index];
         for (const auto &[SuccOffset, SuccDataIdx] : SuccKV)
-          addSuccProfile(YamlBB, SuccOffset, SuccDataIdx);
-        addCallsProfile(YamlBB, FromOffset);
-        if (YamlBB.ExecCount || !YamlBB.Successors.empty() ||
-            !YamlBB.CallSites.empty())
-          YamlBF.Blocks.emplace_back(YamlBB);
+          if (BlockMap.isInputBlock(SuccOffset))
+            YamlBB.Successors.emplace_back(
+                getSuccessorInfo(SuccOffset, SuccDataIdx));
+      }
+      for (const auto &[FromOffset, CallTo] : Branches.InterIndex) {
+        auto BlockIt = BlockMap.upper_bound(FromOffset);
+        --BlockIt;
+        const unsigned BlockOffset = BlockIt->first;
+        const unsigned BlockIndex = BlockIt->second.getBBIndex();
+        yaml::bolt::BinaryBasicBlockProfile &YamlBB = YamlBF.Blocks[BlockIndex];
+        const uint32_t Offset = FromOffset - BlockOffset;
+        for (const auto &[CallToLoc, CallToIdx] : CallTo)
+          YamlBB.CallSites.emplace_back(
+              getCallSiteInfo(CallToLoc, CallToIdx, Offset));
+        llvm::sort(YamlBB.CallSites, [](yaml::bolt::CallSiteInfo &A,
+                                        yaml::bolt::CallSiteInfo &B) {
+          return A.Offset < B.Offset;
+        });
       }
+      // Drop blocks without a hash, won't be useful for stale matching.
+      llvm::erase_if(YamlBF.Blocks,
+                     [](const yaml::bolt::BinaryBasicBlockProfile &YamlBB) {
+                       return YamlBB.Hash == (yaml::Hex64)0;
+                     });
       BP.Functions.emplace_back(YamlBF);
     }
   }
diff --git a/bolt/test/X86/yaml-secondary-entry-discriminator.s b/bolt/test/X86/yaml-secondary-entry-discriminator.s
index 7fd5a716cd68e..5d6e291fd7c22 100644
--- a/bolt/test/X86/yaml-secondary-entry-discriminator.s
+++ b/bolt/test/X86/yaml-secondary-entry-discriminator.s
@@ -38,10 +38,62 @@
 # RUN: llvm-bolt %t.exe -o %t.bat --data %t.fdata --funcs=func \
 # RUN:   --split-functions --split-strategy=all --split-all-cold --enable-bat
 
+## Prepare pre-aggregated profile using %t.bat
+# RUN: link_fdata %s %t.bat %t.preagg PREAGG
+## Strip labels used for pre-aggregated profile
+# RUN: llvm-strip -NLcall -NLindcall %t.bat
+
+## Convert pre-aggregated profile using BAT
+# RUN: perf2bolt %t.bat -p %t.preagg --pa -o %t.bat.fdata -w %t.bat.yaml
+
+## Convert BAT fdata into YAML
+# RUN: llvm-bolt %t.exe -data %t.bat.fdata -w %t.bat.fdata-yaml -o /dev/null
+
+## Check fdata YAML - make sure that a direct call has discriminator field
+# RUN: FileCheck %s --input-file %t.bat.fdata-yaml -check-prefix CHECK-BAT-YAML
+
+## Check BAT YAML - make sure that a direct call has discriminator field
+# RUN: FileCheck %s --input-file %t.bat.yaml --check-prefix CHECK-BAT-YAML
+
+## YAML BAT test of calling BAT secondary entry from BAT function
+# RUN: llvm-bolt %t.exe -o %t.bat2 --data %t.fdata --funcs=main,func \
+# RUN:   --split-functions --split-strategy=all --split-all-cold --enable-bat
+
+## Prepare pre-aggregated profile using %t.bat
+# RUN: link_fdata %s %t.bat2 %t.preagg2 PREAGG2
+
+## Strip labels used for pre-aggregated profile
+# RUN: llvm-strip -NLcall -NLindcall %t.bat2
+
+## Convert pre-aggregated profile using BAT
+# RUN: perf2bolt %t.bat2 -p %t.preagg2 --pa -o %t.bat2.fdata -w %t.bat2.yaml
+
+## Convert BAT fdata into YAML
+# RUN: llvm-bolt %t.exe -data %t.bat2.fdata -w %t.bat2.fdata-yaml -o /dev/null
+
+## Check fdata YAML - make sure that a direct call has discriminator field
+# RUN: FileCheck %s --input-file %t.bat2.fdata-yaml -check-prefix CHECK-BAT-YAML
+
+## Check BAT YAML - make sure that a direct call has discriminator field
+# RUN: FileCheck %s --input-file %t.bat2.yaml --check-prefix CHECK-BAT-YAML
+
+# CHECK-BAT-YAML:      - name:    main
+# CHECK-BAT-YAML-NEXT:   fid:     [[#]]
+# CHECK-BAT-YAML-NEXT:   hash:    0xADF270D550151185
+# CHECK-BAT-YAML-NEXT:   exec:    0
+# CHECK-BAT-YAML-NEXT:   nblocks: 4
+# CHECK-BAT-YAML-NEXT:   blocks:
+# CHECK-BAT-YAML:          - bid:   1
+# CHECK-BAT-YAML-NEXT:       insns: [[#]]
+# CHECK-BAT-YAML-NEXT:       hash:  0x36A303CBA4360018
+# CHECK-BAT-YAML-NEXT:       calls: [ { off: 0x0, fid: [[#]], disc: 1, cnt: 1
+
 .globl func
 .type	func, @function
 func:
 # FDATA: 0 [unknown] 0 1 func 0 1 0
+# PREAGG: B X:0 #func# 1 1
+# PREAGG2: B X:0 #func# 1 1
   .cfi_startproc
   pushq   %rbp
   movq    %rsp, %rbp
@@ -71,12 +123,18 @@ main:
   movl    $0, -4(%rbp)
   testq   %rax, %rax
   jne     Lindcall
+.globl Lcall
 Lcall:
   call    secondary_entry
 # FDATA: 1 main #Lcall# 1 secondary_entry 0 1 1
+# PREAGG: B #Lcall# #secondary_entry# 1 1
+# PREAGG2: B #main.cold.0# #func.cold.0# 1 1
+.globl Lindcall
 Lindcall:
   callq   *%rax
 # FDATA: 1 main #Lindcall# 1 secondary_entry 0 1 1
+# PREAGG: B #Lindcall# #secondary_entry# 1 1
+# PREAGG2: B #main.cold.1# #func.cold.0# 1 1
   xorl    %eax, %eax
   addq    $16, %rsp
   popq    %rbp

From 41e8a9b1b2010b8111a0488a03b057cb8e33a199 Mon Sep 17 00:00:00 2001
From: Haojian Wu <hokein.wu@gmail.com>
Date: Thu, 11 Apr 2024 21:15:47 +0200
Subject: [PATCH 184/886] [HLSL] Remove an unnecessary .ll file in
 clang/test/SemaHLSL/. (#87346)

This file seemed to be added unintentionally in
9434c083475e42f47383f3067fe2a155db5c6a30.

From 5d6d8dcd292e0a107b11d378932eee9c2f9ccfc7 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Thu, 11 Apr 2024 13:29:29 -0600
Subject: [PATCH 185/886] [clang][llvm] Remove "implicit-section-name"
 attribute (#87906)

D33412/D33413 introduced this to support a clang pragma to set section
names for a symbol depending on if it would be placed in
bss/data/rodata/text, which may not be known until the backend. However,
for text we know that only functions will go there, so just directly set
the section in clang instead of going through a completely separate
attribute.

Autoupgrade the "implicit-section-name" attribute to directly setting
the section on a Fuction.
---
 clang/lib/CodeGen/CodeGenModule.cpp           |  2 +-
 clang/test/CodeGen/clang-sections-attribute.c |  3 ---
 clang/test/CodeGenCXX/clang-sections.cpp      | 18 ++++++-------
 llvm/lib/CodeGen/TargetInstrInfo.cpp          |  3 +--
 .../CodeGen/TargetLoweringObjectFileImpl.cpp  | 11 +-------
 llvm/lib/IR/AutoUpgrade.cpp                   |  8 ++++++
 llvm/lib/Target/TargetLoweringObjectFile.cpp  |  5 ----
 .../Bitcode/upgrade-implicit-section-name.ll  |  8 ++++++
 .../CodeGen/AArch64/clang-section-macho.ll    | 22 ---------------
 llvm/test/CodeGen/ARM/clang-section.ll        | 24 -----------------
 .../Generic/machine-function-splitter.ll      | 27 +++----------------
 .../basic-block-sections-pragma-sections.ll   |  4 +--
 12 files changed, 31 insertions(+), 104 deletions(-)
 create mode 100644 llvm/test/Bitcode/upgrade-implicit-section-name.ll
 delete mode 100644 llvm/test/CodeGen/AArch64/clang-section-macho.ll

diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index b15031dca4686..73a9cb9d6e042 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -2635,7 +2635,7 @@ void CodeGenModule::setNonAliasAttributes(GlobalDecl GD,
         addUsedGlobal(F);
       if (auto *SA = D->getAttr<PragmaClangTextSectionAttr>())
         if (!D->getAttr<SectionAttr>())
-          F->addFnAttr("implicit-section-name", SA->getName());
+          F->setSection(SA->getName());
 
       llvm::AttrBuilder Attrs(F->getContext());
       if (GetCPUAndFeaturesAttributes(GD, Attrs)) {
diff --git a/clang/test/CodeGen/clang-sections-attribute.c b/clang/test/CodeGen/clang-sections-attribute.c
index 70ed24ed07a28..768bdd4d87649 100644
--- a/clang/test/CodeGen/clang-sections-attribute.c
+++ b/clang/test/CodeGen/clang-sections-attribute.c
@@ -69,8 +69,5 @@ static int int_zvar;
 // CHECK: define internal void @int_fun() #0 section ".int_fun_attr"
 // CHECK: define internal void @int_fun2() #0 section ".int_fun2_attr"
 //
-// Function attributes should not include implicit-section-name.
-// CHECK-NOT: attributes #0 = {{.*}}implicit-section-name
-//
 // No other attribute group should be present in the file.
 // CHECK-NOT: attributes #1
diff --git a/clang/test/CodeGenCXX/clang-sections.cpp b/clang/test/CodeGenCXX/clang-sections.cpp
index a444f2d0cae59..aa159e552b1b3 100644
--- a/clang/test/CodeGenCXX/clang-sections.cpp
+++ b/clang/test/CodeGenCXX/clang-sections.cpp
@@ -81,24 +81,22 @@ int hoo(void) {
 //CHECK: @p ={{.*}} constant i32 7, align 4
 //CHECK: @_ZL5fptrs = internal constant [2 x ptr] [ptr @foo, ptr @goo], align {{4|8}} #3
 
-//CHECK: define{{.*}} i32 @foo() #5 {
-//CHECK: define{{.*}} i32 @goo() #6 {
-//CHECK: declare i32 @zoo(ptr noundef, ptr noundef) #7
-//CHECK: define{{.*}} i32 @hoo() #8 {
+//ELF: define{{.*}} i32 @foo(){{.*}} section "my_text.1" {
+//ELF: define{{.*}} i32 @goo(){{.*}} section "my_text.2" {
+//MACHO: define{{.*}} i32 @foo(){{.*}} section "__TEXT,__mytext1" {
+//MACHO: define{{.*}} i32 @goo(){{.*}} section "__TEXT,__mytext2" {
+
+// ensure zoo/hoo don't have a section
+//CHECK: declare i32 @zoo(ptr noundef, ptr noundef) #6{{$}}
+//CHECK: define{{.*}} i32 @hoo() #5 {
 
 //ELF: attributes #0 = { "bss-section"="my_bss.1" "data-section"="my_data.1" "rodata-section"="my_rodata.1" }
 //ELF: attributes #1 = { "data-section"="my_data.1" "rodata-section"="my_rodata.1" }
 //ELF: attributes #2 = { "bss-section"="my_bss.2" "rodata-section"="my_rodata.1" }
 //ELF: attributes #3 = { "bss-section"="my_bss.2" "data-section"="my_data.2" "relro-section"="my_relro.2" "rodata-section"="my_rodata.2" }
 //ELF: attributes #4 = { "relro-section"="my_relro.2" }
-//ELF: attributes #5 = { {{.*"implicit-section-name"="my_text.1".*}} }
-//ELF: attributes #6 = { {{.*"implicit-section-name"="my_text.2".*}} }
 //MACHO: attributes #0 = { "bss-section"="__BSS,__mybss1" "data-section"="__DATA,__mydata1" "rodata-section"="__RODATA,__myrodata1" }
 //MACHO: attributes #1 = { "data-section"="__DATA,__mydata1" "rodata-section"="__RODATA,__myrodata1" }
 //MACHO: attributes #2 = { "bss-section"="__BSS,__mybss2" "rodata-section"="__RODATA,__myrodata1" }
 //MACHO: attributes #3 = { "bss-section"="__BSS,__mybss2" "data-section"="__DATA,__mydata2" "relro-section"="__RELRO,__myrelro2" "rodata-section"="__RODATA,__myrodata2" }
 //MACHO: attributes #4 = { "relro-section"="__RELRO,__myrelro2" }
-//MACHO: attributes #5 = { {{.*"implicit-section-name"="__TEXT,__mytext1".*}} }
-//MACHO: attributes #6 = { {{.*"implicit-section-name"="__TEXT,__mytext2".*}} }
-//CHECK-NOT: attributes #7 = { {{.*"implicit-section-name".*}} }
-//CHECK-NOT: attributes #8 = { {{.*"implicit-section-name".*}} }
diff --git a/llvm/lib/CodeGen/TargetInstrInfo.cpp b/llvm/lib/CodeGen/TargetInstrInfo.cpp
index 7d77e5d1a1ff0..14b2e4268eb0f 100644
--- a/llvm/lib/CodeGen/TargetInstrInfo.cpp
+++ b/llvm/lib/CodeGen/TargetInstrInfo.cpp
@@ -1473,8 +1473,7 @@ bool TargetInstrInfo::isFunctionSafeToSplit(const MachineFunction &MF) const {
   // since the split part may not be placed in a contiguous region. It may also
   // be more beneficial to augment the linker to ensure contiguous layout of
   // split functions within the same section as specified by the attribute.
-  if (MF.getFunction().hasSection() ||
-      MF.getFunction().hasFnAttribute("implicit-section-name"))
+  if (MF.getFunction().hasSection())
     return false;
 
   // We don't want to proceed further for cold functions
diff --git a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
index 15b59421a0f44..2a77a683a901f 100644
--- a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
@@ -797,10 +797,6 @@ static MCSection *selectExplicitSectionGlobal(
       SectionName = Attrs.getAttribute("data-section").getValueAsString();
     }
   }
-  const Function *F = dyn_cast<Function>(GO);
-  if (F && F->hasFnAttribute("implicit-section-name")) {
-    SectionName = F->getFnAttribute("implicit-section-name").getValueAsString();
-  }
 
   // Infer section flags from the section name if we can.
   Kind = getELFKindForNamedSection(SectionName, Kind);
@@ -934,7 +930,7 @@ MCSection *TargetLoweringObjectFileELF::getUniqueSectionForFunction(
   unsigned Flags = getELFSectionFlags(Kind);
   // If the function's section names is pre-determined via pragma or a
   // section attribute, call selectExplicitSectionGlobal.
-  if (F.hasSection() || F.hasFnAttribute("implicit-section-name"))
+  if (F.hasSection())
     return selectExplicitSectionGlobal(
         &F, Kind, TM, getContext(), getMangler(), NextUniqueID,
         Used.count(&F), /* ForceUnique = */true);
@@ -1298,11 +1294,6 @@ MCSection *TargetLoweringObjectFileMachO::getExplicitSectionGlobal(
     }
   }
 
-  const Function *F = dyn_cast<Function>(GO);
-  if (F && F->hasFnAttribute("implicit-section-name")) {
-    SectionName = F->getFnAttribute("implicit-section-name").getValueAsString();
-  }
-
   // Parse the section specifier and create it if valid.
   StringRef Segment, Section;
   unsigned TAA = 0, StubSize = 0;
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index 0f8c984d5e3c7..2c480fb76ee4d 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -5267,6 +5267,14 @@ void llvm::UpgradeFunctionAttributes(Function &F) {
   F.removeRetAttrs(AttributeFuncs::typeIncompatible(F.getReturnType()));
   for (auto &Arg : F.args())
     Arg.removeAttrs(AttributeFuncs::typeIncompatible(Arg.getType()));
+
+  // Older versions of LLVM treated an "implicit-section-name" attribute
+  // similarly to directly setting the section on a Function.
+  if (Attribute A = F.getFnAttribute("implicit-section-name");
+      A.isValid() && A.isStringAttribute()) {
+    F.setSection(A.getValueAsString());
+    F.removeFnAttr("implicit-section-name");
+  }
 }
 
 static bool isOldLoopArgument(Metadata *MD) {
diff --git a/llvm/lib/Target/TargetLoweringObjectFile.cpp b/llvm/lib/Target/TargetLoweringObjectFile.cpp
index 8db68466490bd..26e9f9eb3f327 100644
--- a/llvm/lib/Target/TargetLoweringObjectFile.cpp
+++ b/llvm/lib/Target/TargetLoweringObjectFile.cpp
@@ -334,11 +334,6 @@ MCSection *TargetLoweringObjectFile::SectionForGlobal(
     }
   }
 
-  if (auto *F = dyn_cast<Function>(GO)) {
-    if (F->hasFnAttribute("implicit-section-name"))
-      return getExplicitSectionGlobal(GO, Kind, TM);
-  }
-
   // Use default section depending on the 'type' of global
   return SelectSectionForGlobal(GO, Kind, TM);
 }
diff --git a/llvm/test/Bitcode/upgrade-implicit-section-name.ll b/llvm/test/Bitcode/upgrade-implicit-section-name.ll
new file mode 100644
index 0000000000000..f22e6a57fa616
--- /dev/null
+++ b/llvm/test/Bitcode/upgrade-implicit-section-name.ll
@@ -0,0 +1,8 @@
+; RUN: llvm-as < %s | llvm-dis - | FileCheck %s
+
+; CHECK: define void @f() section "foo_section"
+; CHECK-NOT: "implicit-section-name"
+
+define void @f() "implicit-section-name"="foo_section" {
+  ret void
+}
diff --git a/llvm/test/CodeGen/AArch64/clang-section-macho.ll b/llvm/test/CodeGen/AArch64/clang-section-macho.ll
deleted file mode 100644
index 406c584af62a9..0000000000000
--- a/llvm/test/CodeGen/AArch64/clang-section-macho.ll
+++ /dev/null
@@ -1,22 +0,0 @@
-;RUN: llc -mtriple=arm64-apple-ios %s -o - | FileCheck %s
-
-define dso_local void @foo() #0 {
-entry:
-  ret void
-}
-
-attributes #0 = { "implicit-section-name"="__TEXT,__mytext" }
-
-; CHECK:      .section	__TEXT,__mytext
-; CHECK-NEXT: .globl	_foo
-
-
-define dso_local void @bar() #1 {
-entry:
-  ret void
-}
-
-attributes #1 = { "implicit-section-name"="__EDATA,zerofill" }
-
-; CHECK:      .section	__EDATA,zerofill
-; CHECK-NEXT: .globl	_bar
diff --git a/llvm/test/CodeGen/ARM/clang-section.ll b/llvm/test/CodeGen/ARM/clang-section.ll
index 4e896a2447dd4..9277d90dba65b 100644
--- a/llvm/test/CodeGen/ARM/clang-section.ll
+++ b/llvm/test/CodeGen/ARM/clang-section.ll
@@ -22,20 +22,6 @@ target triple = "armv7-arm-none-eabi"
 @o = global i32 6, align 4
 @p = constant i32 7, align 4
 
-; Function Attrs: noinline nounwind
-define i32 @foo() #4 {
-entry:
-  %0 = load i32, ptr @b, align 4
-  ret i32 %0
-}
-
-; Function Attrs: noinline
-define i32 @goo() #5 {
-entry:
-  %call = call i32 @zoo(ptr @_ZL1g, ptr @_ZZ3gooE7lstat_h)
-  ret i32 %call
-}
-
 declare i32 @zoo(ptr, ptr) #6
 
 ; Function Attrs: noinline nounwind
@@ -49,8 +35,6 @@ attributes #0 = { "bss-section"="my_bss.1" "data-section"="my_data.1" "rodata-se
 attributes #1 = { "data-section"="my_data.1" "rodata-section"="my_rodata.1" }
 attributes #2 = { "bss-section"="my_bss.2" "rodata-section"="my_rodata.1" }
 attributes #3 = { "bss-section"="my_bss.2" "data-section"="my_data.2" "rodata-section"="my_rodata.2" }
-attributes #4 = { noinline nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="preserve-sign,preserve-sign" "disable-tail-calls"="false" "implicit-section-name"="my_text.1" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a9" "target-features"="+dsp,+fp16,+neon,+vfp3" "unsafe-fp-math"="false" "use-soft-float"="false" }
-attributes #5 = { noinline "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="preserve-sign,preserve-sign" "disable-tail-calls"="false" "implicit-section-name"="my_text.2" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a9" "target-features"="+dsp,+fp16,+neon,+vfp3" "unsafe-fp-math"="false" "use-soft-float"="false" }
 attributes #6 = { "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="preserve-sign,preserve-sign" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a9" "target-features"="+dsp,+fp16,+neon,+vfp3" "unsafe-fp-math"="false" "use-soft-float"="false" }
 attributes #7 = { noinline nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="preserve-sign,preserve-sign" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a9" "target-features"="+dsp,+fp16,+neon,+vfp3" "unsafe-fp-math"="false" "use-soft-float"="false" }
 
@@ -61,14 +45,6 @@ attributes #7 = { noinline nounwind "correctly-rounded-divide-sqrt-fp-math"="fal
 !2 = !{i32 1, !"enumsize_buildattr", i32 2}
 !3 = !{i32 1, !"armlib_unavailable", i32 0}
 
-;CHECK: 	.section	my_text.1,"ax",%progbits
-;CHECK: 	.type	foo,%function
-;CHECK: foo:
-
-;CHECK: 	.section	my_text.2,"ax",%progbits
-;CHECK: 	.type	goo,%function
-;CHECK: goo:
-
 ;CHECK: 	.text
 ;CHECK: 	.type	hoo,%function
 ;CHECK: hoo:
diff --git a/llvm/test/CodeGen/Generic/machine-function-splitter.ll b/llvm/test/CodeGen/Generic/machine-function-splitter.ll
index d7cc3941ee7a3..2097523a61c5f 100644
--- a/llvm/test/CodeGen/Generic/machine-function-splitter.ll
+++ b/llvm/test/CodeGen/Generic/machine-function-splitter.ll
@@ -260,25 +260,6 @@ exit:
   ret i32 %5
 }
 
-define void @foo9(i1 zeroext %0) nounwind #0 !prof !14 {
-;; Check that function with section attribute is not split.
-; MFS-DEFAULTS-LABEL: foo9
-; MFS-DEFAULTS-NOT:   foo9.cold:
-  br i1 %0, label %2, label %4, !prof !17
-
-2:                                                ; preds = %1
-  %3 = call i32 @bar()
-  br label %6
-
-4:                                                ; preds = %1
-  %5 = call i32 @baz()
-  br label %6
-
-6:                                                ; preds = %4, %2
-  %7 = tail call i32 @qux()
-  ret void
-}
-
 define i32 @foo10(i1 zeroext %0) personality ptr @__gxx_personality_v0 !prof !14 {
 ;; Check that nop is inserted just before the EH pad if it's beginning a section.
 ; MFS-DEFAULTS-LABEL:         foo10
@@ -564,10 +545,10 @@ define i32 @foo19(i32 %in) !prof !14 !section_prefix !15 {
 ; MFS-DEFAULTS-LABEL:        foo19
 ; MFS-DEFAULTS:              .section        .text.split.foo19
 ; MFS-DEFAULTS-NEXT:         foo19.cold:
-; MFS-DEFAULTS-X86:            .LJTI18_0
-; MFS-DEFAULTS-AARCH64-NOT:    .LJTI18_0
+; MFS-DEFAULTS-X86:            .LJTI17_0
+; MFS-DEFAULTS-AARCH64-NOT:    .LJTI17_0
 ; MFS-DEFAULTS:              .section        .rodata
-; MFS-DEFAULTS:                .LJTI18_0
+; MFS-DEFAULTS:                .LJTI17_0
   %cmp = icmp sgt i32 %in, 3
   br i1 %cmp, label %hot, label %cold_switch, !prof !17
 
@@ -638,8 +619,6 @@ declare i32 @__gxx_personality_v0(...)
 
 @_ZTIi = external constant ptr
 
-attributes #0 = { "implicit-section-name"="nosplit" }
-
 !llvm.module.flags = !{!0}
 !0 = !{i32 1, !"ProfileSummary", !1}
 !1 = !{!2, !3, !4, !5, !6, !7, !8, !9}
diff --git a/llvm/test/CodeGen/X86/basic-block-sections-pragma-sections.ll b/llvm/test/CodeGen/X86/basic-block-sections-pragma-sections.ll
index 13a4607100a05..da3b4f253147b 100644
--- a/llvm/test/CodeGen/X86/basic-block-sections-pragma-sections.ll
+++ b/llvm/test/CodeGen/X86/basic-block-sections-pragma-sections.ll
@@ -38,7 +38,7 @@
 ;;   return 0;
 ;; }
 
-define dso_local i32 @_Z3fooi(i32 %n) local_unnamed_addr #0 {
+define dso_local i32 @_Z3fooi(i32 %n) local_unnamed_addr section "foo_section" {
 entry:
   %cmp = icmp slt i32 %n, 0
   br i1 %cmp, label %if.then, label %if.end
@@ -52,5 +52,3 @@ if.end:                                           ; preds = %entry
 }
 
 declare dso_local void @exit(i32) local_unnamed_addr
-
-attributes #0 = {"implicit-section-name"="foo_section" }

From 20ed5b1f45871612570d3bd447121ac43e083c6a Mon Sep 17 00:00:00 2001
From: Mingming Liu <mingmingl@google.com>
Date: Thu, 11 Apr 2024 12:47:51 -0700
Subject: [PATCH 186/886] [nfc]Fix test gold/x86/thinlto.ll with GV Summary
 change (#88429)

This gold thinlto test should have been updated in commit dda7333
(https://github.com/llvm/llvm-project/commit/dda73336ad22bd0b5ecda17040c50fb10fcbe5fb).

It's ok to keep {Inputs/emit-llvm.foo.ll, Inputs/emit-llvm.bar.ll} the
current way for `llvm-as` since the field is optional (see example in
https://github.com/llvm/llvm-project/blob/5d6d8dcd292e0a107b11d378932eee9c2f9ccfc7/llvm/test/Assembler/thinlto-summary.ll#L12-L16)
---
 llvm/test/tools/gold/X86/thinlto.ll | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/llvm/test/tools/gold/X86/thinlto.ll b/llvm/test/tools/gold/X86/thinlto.ll
index 1745f0382b817..9d56afe9d58dc 100644
--- a/llvm/test/tools/gold/X86/thinlto.ll
+++ b/llvm/test/tools/gold/X86/thinlto.ll
@@ -123,11 +123,11 @@
 
 ; DIS1: ^0 = module: (path: "{{.*}}thinlto.ll.tmp.o", hash: (0, 0, 0, 0, 0))
 ; DIS1: ^1 = module: (path: "{{.*}}thinlto.ll.tmp2.o", hash: (0, 0, 0, 0, 0))
-; DIS1: ^2 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 1)))
-; DIS1: ^3 = gv: (guid: 14740650423002898831, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 2, calls: ((callee: ^2)))))
+; DIS1: ^2 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^1, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
+; DIS1: ^3 = gv: (guid: 14740650423002898831, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 2, calls: ((callee: ^2)))))
 
 ; DIS2: ^0 = module: (path: "{{.*}}thinlto.ll.tmp2.o", hash: (0, 0, 0, 0, 0))
-; DIS2: ^1 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0), insts: 1)))
+; DIS2: ^1 = gv: (guid: 13146401226427987378, summaries: (function: (module: ^0, flags: (linkage: external, visibility: default, notEligibleToImport: 0, live: 1, dsoLocal: 0, canAutoHide: 0, importType: definition), insts: 1)))
 
 ; COMBINED: <MODULE_STRTAB_BLOCK
 ; COMBINED-NEXT: <ENTRY {{.*}} record string = '{{.*}}/test/tools/gold/X86/Output/thinlto.ll.tmp{{.*}}.o'

From aa80f3ec48419a73aafcc2ff947c6dd1e3734481 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Thu, 11 Apr 2024 16:19:29 -0400
Subject: [PATCH 187/886] Reapply "[Clang][Sema] Fix crash when 'this' is used
 in a dependent class scope function template specialization that instantiates
 to a static member function (#87541)" (#88311)

Reapplies #87541 and addresses the bug which caused expressions naming
overload sets to be incorrectly rebuilt.
---
 clang/docs/ReleaseNotes.rst                   |   2 +
 clang/include/clang/Sema/Sema.h               |  16 ++-
 clang/lib/Sema/SemaExpr.cpp                   |  28 +----
 clang/lib/Sema/SemaExprCXX.cpp                |  61 +++++-----
 clang/lib/Sema/SemaExprMember.cpp             |  63 ++++++++---
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |   8 ++
 clang/lib/Sema/TreeTransform.h                |  65 ++++++-----
 .../SemaTemplate/instantiate-using-decl.cpp   |   2 +-
 ...ms-function-specialization-class-scope.cpp | 106 +++++++++++++++++-
 9 files changed, 252 insertions(+), 99 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 93a380411604b..45a9a79739a4e 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -526,6 +526,8 @@ Bug Fixes to C++ Support
 - Fix an issue caused by not handling invalid cases when substituting into the parameter mapping of a constraint. Fixes (#GH86757).
 - Fixed a bug that prevented member function templates of class templates declared with a deduced return type
   from being explicitly specialized for a given implicit instantiation of the class template.
+- Fixed a crash when ``this`` is used in a dependent class scope function template specialization
+  that instantiates to a static member function.
 
 - Fix crash when inheriting from a cv-qualified type. Fixes:
   (`#35603 <https://github.com/llvm/llvm-project/issues/35603>`_)
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index f2c55b13b6d32..351b8e1f31134 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -5447,7 +5447,8 @@ class Sema final : public SemaBase {
 
   ExprResult BuildDeclarationNameExpr(const CXXScopeSpec &SS, LookupResult &R,
                                       bool NeedsADL,
-                                      bool AcceptInvalidDecl = false);
+                                      bool AcceptInvalidDecl = false,
+                                      bool NeedUnresolved = false);
   ExprResult BuildDeclarationNameExpr(
       const CXXScopeSpec &SS, const DeclarationNameInfo &NameInfo, NamedDecl *D,
       NamedDecl *FoundD = nullptr,
@@ -6590,7 +6591,10 @@ class Sema final : public SemaBase {
                             SourceLocation RParenLoc);
 
   //// ActOnCXXThis -  Parse 'this' pointer.
-  ExprResult ActOnCXXThis(SourceLocation loc);
+  ExprResult ActOnCXXThis(SourceLocation Loc);
+
+  /// Check whether the type of 'this' is valid in the current context.
+  bool CheckCXXThisType(SourceLocation Loc, QualType Type);
 
   /// Build a CXXThisExpr and mark it referenced in the current context.
   Expr *BuildCXXThisExpr(SourceLocation Loc, QualType Type, bool IsImplicit);
@@ -7013,10 +7017,14 @@ class Sema final : public SemaBase {
   ///@{
 
 public:
+  /// Check whether an expression might be an implicit class member access.
+  bool isPotentialImplicitMemberAccess(const CXXScopeSpec &SS, LookupResult &R,
+                                       bool IsAddressOfOperand);
+
   ExprResult BuildPossibleImplicitMemberExpr(
       const CXXScopeSpec &SS, SourceLocation TemplateKWLoc, LookupResult &R,
-      const TemplateArgumentListInfo *TemplateArgs, const Scope *S,
-      UnresolvedLookupExpr *AsULE = nullptr);
+      const TemplateArgumentListInfo *TemplateArgs, const Scope *S);
+
   ExprResult
   BuildImplicitMemberExpr(const CXXScopeSpec &SS, SourceLocation TemplateKWLoc,
                           LookupResult &R,
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 4d4ef9b16381b..846e7d5d3ab92 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -2912,26 +2912,9 @@ Sema::ActOnIdExpression(Scope *S, CXXScopeSpec &SS,
   // to get this right here so that we don't end up making a
   // spuriously dependent expression if we're inside a dependent
   // instance method.
-  if (getLangOpts().CPlusPlus && !R.empty() &&
-      (*R.begin())->isCXXClassMember()) {
-    bool MightBeImplicitMember;
-    if (!IsAddressOfOperand)
-      MightBeImplicitMember = true;
-    else if (!SS.isEmpty())
-      MightBeImplicitMember = false;
-    else if (R.isOverloadedResult())
-      MightBeImplicitMember = false;
-    else if (R.isUnresolvableResult())
-      MightBeImplicitMember = true;
-    else
-      MightBeImplicitMember = isa<FieldDecl>(R.getFoundDecl()) ||
-                              isa<IndirectFieldDecl>(R.getFoundDecl()) ||
-                              isa<MSPropertyDecl>(R.getFoundDecl());
-
-    if (MightBeImplicitMember)
-      return BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc,
-                                             R, TemplateArgs, S);
-  }
+  if (isPotentialImplicitMemberAccess(SS, R, IsAddressOfOperand))
+    return BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs,
+                                           S);
 
   if (TemplateArgs || TemplateKWLoc.isValid()) {
 
@@ -3442,10 +3425,11 @@ static bool ShouldLookupResultBeMultiVersionOverload(const LookupResult &R) {
 
 ExprResult Sema::BuildDeclarationNameExpr(const CXXScopeSpec &SS,
                                           LookupResult &R, bool NeedsADL,
-                                          bool AcceptInvalidDecl) {
+                                          bool AcceptInvalidDecl,
+                                          bool NeedUnresolved) {
   // If this is a single, fully-resolved result and we don't need ADL,
   // just build an ordinary singleton decl ref.
-  if (!NeedsADL && R.isSingleResult() &&
+  if (!NeedUnresolved && !NeedsADL && R.isSingleResult() &&
       !R.getAsSingle<FunctionTemplateDecl>() &&
       !ShouldLookupResultBeMultiVersionOverload(R))
     return BuildDeclarationNameExpr(SS, R.getLookupNameInfo(), R.getFoundDecl(),
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 12f42f66e5e21..763901d4418d2 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -1414,26 +1414,42 @@ bool Sema::CheckCXXThisCapture(SourceLocation Loc, const bool Explicit,
 }
 
 ExprResult Sema::ActOnCXXThis(SourceLocation Loc) {
-  /// C++ 9.3.2: In the body of a non-static member function, the keyword this
-  /// is a non-lvalue expression whose value is the address of the object for
-  /// which the function is called.
+  // C++20 [expr.prim.this]p1:
+  //   The keyword this names a pointer to the object for which an
+  //   implicit object member function is invoked or a non-static
+  //   data member's initializer is evaluated.
   QualType ThisTy = getCurrentThisType();
 
-  if (ThisTy.isNull()) {
-    DeclContext *DC = getFunctionLevelDeclContext();
+  if (CheckCXXThisType(Loc, ThisTy))
+    return ExprError();
 
-    if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
-        Method && Method->isExplicitObjectMemberFunction()) {
-      return Diag(Loc, diag::err_invalid_this_use) << 1;
-    }
+  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
+}
 
-    if (isLambdaCallWithExplicitObjectParameter(CurContext))
-      return Diag(Loc, diag::err_invalid_this_use) << 1;
+bool Sema::CheckCXXThisType(SourceLocation Loc, QualType Type) {
+  if (!Type.isNull())
+    return false;
 
-    return Diag(Loc, diag::err_invalid_this_use) << 0;
+  // C++20 [expr.prim.this]p3:
+  //   If a declaration declares a member function or member function template
+  //   of a class X, the expression this is a prvalue of type
+  //   "pointer to cv-qualifier-seq X" wherever X is the current class between
+  //   the optional cv-qualifier-seq and the end of the function-definition,
+  //   member-declarator, or declarator. It shall not appear within the
+  //   declaration of either a static member function or an explicit object
+  //   member function of the current class (although its type and value
+  //   category are defined within such member functions as they are within
+  //   an implicit object member function).
+  DeclContext *DC = getFunctionLevelDeclContext();
+  if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
+      Method && Method->isExplicitObjectMemberFunction()) {
+    Diag(Loc, diag::err_invalid_this_use) << 1;
+  } else if (isLambdaCallWithExplicitObjectParameter(CurContext)) {
+    Diag(Loc, diag::err_invalid_this_use) << 1;
+  } else {
+    Diag(Loc, diag::err_invalid_this_use) << 0;
   }
-
-  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
+  return true;
 }
 
 Expr *Sema::BuildCXXThisExpr(SourceLocation Loc, QualType Type,
@@ -8626,21 +8642,8 @@ static ExprResult attemptRecovery(Sema &SemaRef,
 
       // Detect and handle the case where the decl might be an implicit
       // member.
-      bool MightBeImplicitMember;
-      if (!Consumer.isAddressOfOperand())
-        MightBeImplicitMember = true;
-      else if (!NewSS.isEmpty())
-        MightBeImplicitMember = false;
-      else if (R.isOverloadedResult())
-        MightBeImplicitMember = false;
-      else if (R.isUnresolvableResult())
-        MightBeImplicitMember = true;
-      else
-        MightBeImplicitMember = isa<FieldDecl>(ND) ||
-                                isa<IndirectFieldDecl>(ND) ||
-                                isa<MSPropertyDecl>(ND);
-
-      if (MightBeImplicitMember)
+      if (SemaRef.isPotentialImplicitMemberAccess(
+              NewSS, R, Consumer.isAddressOfOperand()))
         return SemaRef.BuildPossibleImplicitMemberExpr(
             NewSS, /*TemplateKWLoc*/ SourceLocation(), R,
             /*TemplateArgs*/ nullptr, /*S*/ nullptr);
diff --git a/clang/lib/Sema/SemaExprMember.cpp b/clang/lib/Sema/SemaExprMember.cpp
index 32998ae60eafe..eeac753a34897 100644
--- a/clang/lib/Sema/SemaExprMember.cpp
+++ b/clang/lib/Sema/SemaExprMember.cpp
@@ -61,6 +61,10 @@ enum IMAKind {
   /// The reference is a contextually-permitted abstract member reference.
   IMA_Abstract,
 
+  /// Whether the context is static is dependent on the enclosing template (i.e.
+  /// in a dependent class scope explicit specialization).
+  IMA_Dependent,
+
   /// The reference may be to an unresolved using declaration and the
   /// context is not an instance method.
   IMA_Unresolved_StaticOrExplicitContext,
@@ -91,10 +95,18 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
 
   DeclContext *DC = SemaRef.getFunctionLevelDeclContext();
 
-  bool isStaticOrExplicitContext =
-      SemaRef.CXXThisTypeOverride.isNull() &&
-      (!isa<CXXMethodDecl>(DC) || cast<CXXMethodDecl>(DC)->isStatic() ||
-       cast<CXXMethodDecl>(DC)->isExplicitObjectMemberFunction());
+  bool couldInstantiateToStatic = false;
+  bool isStaticOrExplicitContext = SemaRef.CXXThisTypeOverride.isNull();
+
+  if (auto *MD = dyn_cast<CXXMethodDecl>(DC)) {
+    if (MD->isImplicitObjectMemberFunction()) {
+      isStaticOrExplicitContext = false;
+      // A dependent class scope function template explicit specialization
+      // that is neither declared 'static' nor with an explicit object
+      // parameter could instantiate to a static or non-static member function.
+      couldInstantiateToStatic = MD->getDependentSpecializationInfo();
+    }
+  }
 
   if (R.isUnresolvableResult())
     return isStaticOrExplicitContext ? IMA_Unresolved_StaticOrExplicitContext
@@ -123,6 +135,9 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
   if (Classes.empty())
     return IMA_Static;
 
+  if (couldInstantiateToStatic)
+    return IMA_Dependent;
+
   // C++11 [expr.prim.general]p12:
   //   An id-expression that denotes a non-static data member or non-static
   //   member function of a class can only be used:
@@ -263,32 +278,52 @@ static void diagnoseInstanceReference(Sema &SemaRef,
   }
 }
 
+bool Sema::isPotentialImplicitMemberAccess(const CXXScopeSpec &SS,
+                                           LookupResult &R,
+                                           bool IsAddressOfOperand) {
+  if (!getLangOpts().CPlusPlus)
+    return false;
+  else if (R.empty() || !R.begin()->isCXXClassMember())
+    return false;
+  else if (!IsAddressOfOperand)
+    return true;
+  else if (!SS.isEmpty())
+    return false;
+  else if (R.isOverloadedResult())
+    return false;
+  else if (R.isUnresolvableResult())
+    return true;
+  else
+    return isa<FieldDecl, IndirectFieldDecl, MSPropertyDecl>(R.getFoundDecl());
+}
+
 /// Builds an expression which might be an implicit member expression.
 ExprResult Sema::BuildPossibleImplicitMemberExpr(
     const CXXScopeSpec &SS, SourceLocation TemplateKWLoc, LookupResult &R,
-    const TemplateArgumentListInfo *TemplateArgs, const Scope *S,
-    UnresolvedLookupExpr *AsULE) {
-  switch (ClassifyImplicitMemberAccess(*this, R)) {
+    const TemplateArgumentListInfo *TemplateArgs, const Scope *S) {
+  switch (IMAKind Classification = ClassifyImplicitMemberAccess(*this, R)) {
   case IMA_Instance:
-    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, true, S);
-
   case IMA_Mixed:
   case IMA_Mixed_Unrelated:
   case IMA_Unresolved:
-    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, false,
-                                   S);
-
+    return BuildImplicitMemberExpr(
+        SS, TemplateKWLoc, R, TemplateArgs,
+        /*IsKnownInstance=*/Classification == IMA_Instance, S);
   case IMA_Field_Uneval_Context:
     Diag(R.getNameLoc(), diag::warn_cxx98_compat_non_static_member_use)
       << R.getLookupNameInfo().getName();
     [[fallthrough]];
   case IMA_Static:
   case IMA_Abstract:
+  case IMA_Dependent:
   case IMA_Mixed_StaticOrExplicitContext:
   case IMA_Unresolved_StaticOrExplicitContext:
     if (TemplateArgs || TemplateKWLoc.isValid())
-      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, false, TemplateArgs);
-    return AsULE ? AsULE : BuildDeclarationNameExpr(SS, R, false);
+      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, /*RequiresADL=*/false,
+                                 TemplateArgs);
+    return BuildDeclarationNameExpr(
+        SS, R, /*NeedsADL=*/false, /*AcceptInvalidDecl=*/false,
+        /*NeedUnresolved=*/Classification == IMA_Dependent);
 
   case IMA_Error_StaticOrExplicitContext:
   case IMA_Error_Unrelated:
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 2aa696d35d1dd..707446e132094 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5095,6 +5095,14 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
   EnterExpressionEvaluationContext EvalContext(
       *this, Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 
+  Qualifiers ThisTypeQuals;
+  CXXRecordDecl *ThisContext = nullptr;
+  if (CXXMethodDecl *Method = dyn_cast<CXXMethodDecl>(Function)) {
+    ThisContext = Method->getParent();
+    ThisTypeQuals = Method->getMethodQualifiers();
+  }
+  CXXThisScopeRAII ThisScope(*this, ThisContext, ThisTypeQuals);
+
   // Introduce a new scope where local variable instantiations will be
   // recorded, unless we're actually a member function within a local
   // class, in which case we need to merge our results with the parent
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 33a9356e82f40..284c9173e68ed 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -795,6 +795,9 @@ class TreeTransform {
       ParenExpr *PE, DependentScopeDeclRefExpr *DRE, bool IsAddressOfOperand,
       TypeSourceInfo **RecoveryTSI);
 
+  ExprResult TransformUnresolvedLookupExpr(UnresolvedLookupExpr *E,
+                                           bool IsAddressOfOperand);
+
   StmtResult TransformOMPExecutableDirective(OMPExecutableDirective *S);
 
 // FIXME: We use LLVM_ATTRIBUTE_NOINLINE because inlining causes a ridiculous
@@ -3309,12 +3312,13 @@ class TreeTransform {
 
   /// Build a new C++ "this" expression.
   ///
-  /// By default, builds a new "this" expression without performing any
-  /// semantic analysis. Subclasses may override this routine to provide
-  /// different behavior.
+  /// By default, performs semantic analysis to build a new "this" expression.
+  /// Subclasses may override this routine to provide different behavior.
   ExprResult RebuildCXXThisExpr(SourceLocation ThisLoc,
                                 QualType ThisType,
                                 bool isImplicit) {
+    if (getSema().CheckCXXThisType(ThisLoc, ThisType))
+      return ExprError();
     return getSema().BuildCXXThisExpr(ThisLoc, ThisType, isImplicit);
   }
 
@@ -11351,7 +11355,11 @@ template<typename Derived>
 ExprResult
 TreeTransform<Derived>::TransformAddressOfOperand(Expr *E) {
   if (DependentScopeDeclRefExpr *DRE = dyn_cast<DependentScopeDeclRefExpr>(E))
-    return getDerived().TransformDependentScopeDeclRefExpr(DRE, true, nullptr);
+    return getDerived().TransformDependentScopeDeclRefExpr(
+        DRE, /*IsAddressOfOperand=*/true, nullptr);
+  else if (UnresolvedLookupExpr *ULE = dyn_cast<UnresolvedLookupExpr>(E))
+    return getDerived().TransformUnresolvedLookupExpr(
+        ULE, /*IsAddressOfOperand=*/true);
   else
     return getDerived().TransformExpr(E);
 }
@@ -13057,10 +13065,16 @@ bool TreeTransform<Derived>::TransformOverloadExprDecls(OverloadExpr *Old,
   return false;
 }
 
-template<typename Derived>
+template <typename Derived>
+ExprResult TreeTransform<Derived>::TransformUnresolvedLookupExpr(
+    UnresolvedLookupExpr *Old) {
+  return TransformUnresolvedLookupExpr(Old, /*IsAddressOfOperand=*/false);
+}
+
+template <typename Derived>
 ExprResult
-TreeTransform<Derived>::TransformUnresolvedLookupExpr(
-                                                  UnresolvedLookupExpr *Old) {
+TreeTransform<Derived>::TransformUnresolvedLookupExpr(UnresolvedLookupExpr *Old,
+                                                      bool IsAddressOfOperand) {
   LookupResult R(SemaRef, Old->getName(), Old->getNameLoc(),
                  Sema::LookupOrdinaryName);
 
@@ -13092,26 +13106,8 @@ TreeTransform<Derived>::TransformUnresolvedLookupExpr(
     R.setNamingClass(NamingClass);
   }
 
+  // Rebuild the template arguments, if any.
   SourceLocation TemplateKWLoc = Old->getTemplateKeywordLoc();
-
-  // If we have neither explicit template arguments, nor the template keyword,
-  // it's a normal declaration name or member reference.
-  if (!Old->hasExplicitTemplateArgs() && !TemplateKWLoc.isValid()) {
-    NamedDecl *D = R.getAsSingle<NamedDecl>();
-    // In a C++11 unevaluated context, an UnresolvedLookupExpr might refer to an
-    // instance member. In other contexts, BuildPossibleImplicitMemberExpr will
-    // give a good diagnostic.
-    if (D && D->isCXXInstanceMember()) {
-      return SemaRef.BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc, R,
-                                                     /*TemplateArgs=*/nullptr,
-                                                     /*Scope=*/nullptr);
-    }
-
-    return getDerived().RebuildDeclarationNameExpr(SS, R, Old->requiresADL());
-  }
-
-  // If we have template arguments, rebuild them, then rebuild the
-  // templateid expression.
   TemplateArgumentListInfo TransArgs(Old->getLAngleLoc(), Old->getRAngleLoc());
   if (Old->hasExplicitTemplateArgs() &&
       getDerived().TransformTemplateArguments(Old->getTemplateArgs(),
@@ -13121,6 +13117,23 @@ TreeTransform<Derived>::TransformUnresolvedLookupExpr(
     return ExprError();
   }
 
+  // An UnresolvedLookupExpr can refer to a class member. This occurs e.g. when
+  // a non-static data member is named in an unevaluated operand, or when
+  // a member is named in a dependent class scope function template explicit
+  // specialization that is neither declared static nor with an explicit object
+  // parameter.
+  if (SemaRef.isPotentialImplicitMemberAccess(SS, R, IsAddressOfOperand))
+    return SemaRef.BuildPossibleImplicitMemberExpr(
+        SS, TemplateKWLoc, R,
+        Old->hasExplicitTemplateArgs() ? &TransArgs : nullptr,
+        /*S=*/nullptr);
+
+  // If we have neither explicit template arguments, nor the template keyword,
+  // it's a normal declaration name or member reference.
+  if (!Old->hasExplicitTemplateArgs() && !TemplateKWLoc.isValid())
+    return getDerived().RebuildDeclarationNameExpr(SS, R, Old->requiresADL());
+
+  // If we have template arguments, then rebuild the template-id expression.
   return getDerived().RebuildTemplateIdExpr(SS, TemplateKWLoc, R,
                                             Old->requiresADL(), &TransArgs);
 }
diff --git a/clang/test/SemaTemplate/instantiate-using-decl.cpp b/clang/test/SemaTemplate/instantiate-using-decl.cpp
index 0bbb3ca9c88c8..28d8376438513 100644
--- a/clang/test/SemaTemplate/instantiate-using-decl.cpp
+++ b/clang/test/SemaTemplate/instantiate-using-decl.cpp
@@ -121,7 +121,7 @@ template <typename Scalar> struct Derived : Base<Scalar> {
     (void)&field;
     // expected-error@+1 {{call to non-static member function without an object argument}}
     (void)method;
-    // expected-error@+1 {{call to non-static member function without an object argument}}
+    // expected-error@+1 {{must explicitly qualify name of member function when taking its address}}
     (void)&method;
     // expected-error@+1 {{call to non-static member function without an object argument}}
     method();
diff --git a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
index dcab9bfaeabcb..f7b59f2ac7158 100644
--- a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
+++ b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
@@ -1,7 +1,6 @@
-// RUN: %clang_cc1 -fms-extensions -fsyntax-only -verify %s
-// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -verify %s
+// RUN: %clang_cc1 -fms-extensions -fsyntax-only -Wno-unused-value -verify %s
+// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -Wno-unused-value -verify %s
 
-// expected-no-diagnostics
 class A {
 public:
   template<class U> A(U p) {}
@@ -76,3 +75,104 @@ struct S {
   int f<0>(int);
 };
 }
+
+namespace UsesThis {
+  template<typename T>
+  struct A {
+    int x;
+
+    static inline int y;
+
+    template<typename U = void>
+    static void f();
+
+    template<typename U = void>
+    void g();
+
+    template<typename U>
+    static auto h() -> A*;
+
+    void i();
+
+    static void j();
+
+    template<>
+    void f<int>() {
+      this->x; // expected-error {{invalid use of 'this' outside of a non-static member function}}
+      x; // expected-error {{invalid use of member 'x' in static member function}}
+      A::x; // expected-error {{invalid use of member 'x' in static member function}}
+      +x; // expected-error {{invalid use of member 'x' in static member function}}
+      +A::x; // expected-error {{invalid use of member 'x' in static member function}}
+      &x; // expected-error {{invalid use of member 'x' in static member function}}
+      &A::x;
+      this->y; // expected-error {{invalid use of 'this' outside of a non-static member function}}
+      y;
+      A::y;
+      +y;
+      +A::y;
+      &y;
+      &A::y;
+      f();
+      f<void>();
+      g(); // expected-error {{call to non-static member function without an object argument}}
+      g<void>(); // expected-error {{call to non-static member function without an object argument}}
+      i(); // expected-error {{call to non-static member function without an object argument}}
+      j();
+      &i; // expected-error 2{{must explicitly qualify name of member function when taking its address}}
+      &j;
+      &A::i;
+      &A::j;
+    }
+
+    template<>
+    void g<int>() {
+      this->x;
+      x;
+      A::x;
+      +x;
+      +A::x;
+      &x;
+      &A::x;
+      this->y;
+      y;
+      A::y;
+      +y;
+      +A::y;
+      &y;
+      &A::y;
+      f();
+      f<void>();
+      g();
+      g<void>();
+      i();
+      j();
+      &i; // expected-error 2{{must explicitly qualify name of member function when taking its address}}
+      &j;
+      &A::i;
+      &A::j;
+    }
+
+    template<>
+    auto h<int>() -> decltype(this); // expected-error {{'this' cannot be used in a static member function declaration}}
+  };
+
+  template struct A<int>; // expected-note 3{{in instantiation of}}
+
+  template <typename T>
+  struct Foo {
+    template <typename X>
+    int bar(X x) {
+      return 0;
+    }
+
+    template <>
+    int bar(int x) {
+      return bar(5.0); // ok
+    }
+  };
+
+  void call() {
+    Foo<double> f;
+    f.bar(1);
+  }
+}

From 08e210c6af68005e1bd27488d92de2efba548f58 Mon Sep 17 00:00:00 2001
From: Mingming Liu <mingmingl@google.com>
Date: Thu, 11 Apr 2024 13:28:20 -0700
Subject: [PATCH 188/886] [NFC][IndirectCallProm] Refactor function-based
 conditional devirtualization and indirect call value profile update into one
 helper function (#80762)

* The motivation is to move indirect callee profile update inside the
function-based speculative indirect-call promotion, so that there are
fewer diffs the vtable-based transformation and profile update is
implemented in a follow-up patch.
* The Parent patch is https://github.com/llvm/llvm-project/pull/79381
---
 llvm/include/llvm/ProfileData/InstrProf.h     |  3 +-
 llvm/lib/ProfileData/InstrProf.cpp            |  2 +
 .../Instrumentation/IndirectCallPromotion.cpp | 49 +++++++++++--------
 3 files changed, 32 insertions(+), 22 deletions(-)

diff --git a/llvm/include/llvm/ProfileData/InstrProf.h b/llvm/include/llvm/ProfileData/InstrProf.h
index eb3c10bcba1ca..d5c1ba62911f7 100644
--- a/llvm/include/llvm/ProfileData/InstrProf.h
+++ b/llvm/include/llvm/ProfileData/InstrProf.h
@@ -277,6 +277,8 @@ void annotateValueSite(Module &M, Instruction &Inst,
                        uint32_t MaxMDCount = 3);
 
 /// Same as the above interface but using an ArrayRef, as well as \p Sum.
+/// This function will not annotate !prof metadata on the instruction if the
+/// referenced array is empty.
 void annotateValueSite(Module &M, Instruction &Inst,
                        ArrayRef<InstrProfValueData> VDs, uint64_t Sum,
                        InstrProfValueKind ValueKind, uint32_t MaxMDCount);
@@ -465,7 +467,6 @@ class InstrProfSymtab {
   StringSet<> VTableNames;
   // A map from MD5 keys to function name strings.
   std::vector<std::pair<uint64_t, StringRef>> MD5NameMap;
-
   // A map from MD5 keys to function define. We only populate this map
   // when build the Symtab from a Module.
   std::vector<std::pair<uint64_t, Function *>> MD5FuncMap;
diff --git a/llvm/lib/ProfileData/InstrProf.cpp b/llvm/lib/ProfileData/InstrProf.cpp
index 95f900d0fff1c..a5abf63b010f7 100644
--- a/llvm/lib/ProfileData/InstrProf.cpp
+++ b/llvm/lib/ProfileData/InstrProf.cpp
@@ -1246,6 +1246,8 @@ void annotateValueSite(Module &M, Instruction &Inst,
                        ArrayRef<InstrProfValueData> VDs,
                        uint64_t Sum, InstrProfValueKind ValueKind,
                        uint32_t MaxMDCount) {
+  if (VDs.empty())
+    return;
   LLVMContext &Ctx = M.getContext();
   MDBuilder MDHelper(Ctx);
   SmallVector<Metadata *, 3> Vals;
diff --git a/llvm/lib/Transforms/Instrumentation/IndirectCallPromotion.cpp b/llvm/lib/Transforms/Instrumentation/IndirectCallPromotion.cpp
index 7344fea175171..23a7c6a20aecb 100644
--- a/llvm/lib/Transforms/Instrumentation/IndirectCallPromotion.cpp
+++ b/llvm/lib/Transforms/Instrumentation/IndirectCallPromotion.cpp
@@ -136,11 +136,13 @@ class IndirectCallPromoter {
       const CallBase &CB, const ArrayRef<InstrProfValueData> &ValueDataRef,
       uint64_t TotalCount, uint32_t NumCandidates);
 
-  // Promote a list of targets for one indirect-call callsite. Return
-  // the number of promotions.
-  uint32_t tryToPromote(CallBase &CB,
-                        const std::vector<PromotionCandidate> &Candidates,
-                        uint64_t &TotalCount);
+  // Promote a list of targets for one indirect-call callsite by comparing
+  // indirect callee with functions. Returns true if there are IR
+  // transformations and false otherwise.
+  bool tryToPromoteWithFuncCmp(
+      CallBase &CB, const std::vector<PromotionCandidate> &Candidates,
+      uint64_t TotalCount, ArrayRef<InstrProfValueData> ICallProfDataRef,
+      uint32_t NumCandidates);
 
 public:
   IndirectCallPromoter(Function &Func, InstrProfSymtab *Symtab, bool SamplePGO,
@@ -273,9 +275,10 @@ CallBase &llvm::pgo::promoteIndirectCall(CallBase &CB, Function *DirectCallee,
 }
 
 // Promote indirect-call to conditional direct-call for one callsite.
-uint32_t IndirectCallPromoter::tryToPromote(
+bool IndirectCallPromoter::tryToPromoteWithFuncCmp(
     CallBase &CB, const std::vector<PromotionCandidate> &Candidates,
-    uint64_t &TotalCount) {
+    uint64_t TotalCount, ArrayRef<InstrProfValueData> ICallProfDataRef,
+    uint32_t NumCandidates) {
   uint32_t NumPromoted = 0;
 
   for (const auto &C : Candidates) {
@@ -287,7 +290,22 @@ uint32_t IndirectCallPromoter::tryToPromote(
     NumOfPGOICallPromotion++;
     NumPromoted++;
   }
-  return NumPromoted;
+
+  if (NumPromoted == 0)
+    return false;
+
+  // Adjust the MD.prof metadata. First delete the old one.
+  CB.setMetadata(LLVMContext::MD_prof, nullptr);
+
+  assert(NumPromoted <= ICallProfDataRef.size() &&
+         "Number of promoted functions should not be greater than the number "
+         "of values in profile metadata");
+  // Annotate the remaining value profiles if counter is not zero.
+  if (TotalCount != 0)
+    annotateValueSite(*F.getParent(), CB, ICallProfDataRef.slice(NumPromoted),
+                      TotalCount, IPVK_IndirectCallTarget, NumCandidates);
+
+  return true;
 }
 
 // Traverse all the indirect-call callsite and get the value profile
@@ -305,19 +323,8 @@ bool IndirectCallPromoter::processFunction(ProfileSummaryInfo *PSI) {
       continue;
     auto PromotionCandidates = getPromotionCandidatesForCallSite(
         *CB, ICallProfDataRef, TotalCount, NumCandidates);
-    uint32_t NumPromoted = tryToPromote(*CB, PromotionCandidates, TotalCount);
-    if (NumPromoted == 0)
-      continue;
-
-    Changed = true;
-    // Adjust the MD.prof metadata. First delete the old one.
-    CB->setMetadata(LLVMContext::MD_prof, nullptr);
-    // If all promoted, we don't need the MD.prof metadata.
-    if (TotalCount == 0 || NumPromoted == NumVals)
-      continue;
-    // Otherwise we need update with the un-promoted records back.
-    annotateValueSite(*F.getParent(), *CB, ICallProfDataRef.slice(NumPromoted),
-                      TotalCount, IPVK_IndirectCallTarget, NumCandidates);
+    Changed |= tryToPromoteWithFuncCmp(*CB, PromotionCandidates, TotalCount,
+                                       ICallProfDataRef, NumCandidates);
   }
   return Changed;
 }

From 4df854e1a98ed47c955838d2c94bf2913c50041f Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Thu, 11 Apr 2024 13:38:26 -0700
Subject: [PATCH 189/886] [lldb][lit] Add MallocNanoZone envvar to Darwin ASan
 builds (#88431)

When builds on Darwin are configured with ASan, running tests will emit
a warning about malloc's nano zone (`malloc: nano zone abandoned due to inability to reserve vm space`) that can interfere with the test
output and cause failures. Setting the environment variable
`MallocNanoZone` to 0 will remove this warning and allow the tests to
run as normal.
---
 lldb/test/API/lit.cfg.py   | 1 +
 lldb/test/Shell/lit.cfg.py | 2 ++
 2 files changed, 3 insertions(+)

diff --git a/lldb/test/API/lit.cfg.py b/lldb/test/API/lit.cfg.py
index f9497b632fc50..9ea389c639a01 100644
--- a/lldb/test/API/lit.cfg.py
+++ b/lldb/test/API/lit.cfg.py
@@ -121,6 +121,7 @@ def delete_module_cache(path):
             config.environment["DYLD_INSERT_LIBRARIES"] = find_sanitizer_runtime(
                 "libclang_rt.asan_osx_dynamic.dylib"
             )
+            config.environment["MallocNanoZone"] = "0"
 
     if "Thread" in config.llvm_use_sanitizer:
         config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
diff --git a/lldb/test/Shell/lit.cfg.py b/lldb/test/Shell/lit.cfg.py
index 31afe5151c066..345627835d2bf 100644
--- a/lldb/test/Shell/lit.cfg.py
+++ b/lldb/test/Shell/lit.cfg.py
@@ -52,6 +52,8 @@
 # Enable sanitizer runtime flags.
 config.environment["ASAN_OPTIONS"] = "detect_stack_use_after_return=1"
 config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
+if "Darwin" in config.host_os:
+    config.environment["MallocNanoZone"] = "0"
 
 # Support running the test suite under the lldb-repro wrapper. This makes it
 # possible to capture a test suite run and then rerun all the test from the

From 1f5d130df85c2d0550dc8687ad0fa1d96856c318 Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Thu, 11 Apr 2024 13:46:32 -0700
Subject: [PATCH 190/886] Revert "[lldb][lit] Add MallocNanoZone envvar to
 Darwin ASan builds" (#88436)

Reverts llvm/llvm-project#88431. A config variable for the host OS is
unrecognized on the Linux builds and is causing a build failure.
---
 lldb/test/API/lit.cfg.py   | 1 -
 lldb/test/Shell/lit.cfg.py | 2 --
 2 files changed, 3 deletions(-)

diff --git a/lldb/test/API/lit.cfg.py b/lldb/test/API/lit.cfg.py
index 9ea389c639a01..f9497b632fc50 100644
--- a/lldb/test/API/lit.cfg.py
+++ b/lldb/test/API/lit.cfg.py
@@ -121,7 +121,6 @@ def delete_module_cache(path):
             config.environment["DYLD_INSERT_LIBRARIES"] = find_sanitizer_runtime(
                 "libclang_rt.asan_osx_dynamic.dylib"
             )
-            config.environment["MallocNanoZone"] = "0"
 
     if "Thread" in config.llvm_use_sanitizer:
         config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
diff --git a/lldb/test/Shell/lit.cfg.py b/lldb/test/Shell/lit.cfg.py
index 345627835d2bf..31afe5151c066 100644
--- a/lldb/test/Shell/lit.cfg.py
+++ b/lldb/test/Shell/lit.cfg.py
@@ -52,8 +52,6 @@
 # Enable sanitizer runtime flags.
 config.environment["ASAN_OPTIONS"] = "detect_stack_use_after_return=1"
 config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
-if "Darwin" in config.host_os:
-    config.environment["MallocNanoZone"] = "0"
 
 # Support running the test suite under the lldb-repro wrapper. This makes it
 # possible to capture a test suite run and then rerun all the test from the

From 44de2bb6949f0ca62a2e16506fe3d91be14e6d23 Mon Sep 17 00:00:00 2001
From: Nathan Lanza <nathanlanza@gmail.com>
Date: Thu, 11 Apr 2024 16:56:31 -0400
Subject: [PATCH 191/886] [CIR][cmake] Add support for cmake variable
 CLANG_ENABLE_CIR

Introduce a cmake variable that guards the inclusion of ClangIR into the
build of clang. Guard that we aren't trying to build without MLIR. Add
two subdirectories that, as of now, don't do anything.

Reviewers: bcardosolopes, erichkeane, petrhosek, Ericson2314

Reviewed By: bcardosolopes

Pull Request: https://github.com/llvm/llvm-project/pull/86078
---
 clang/CMakeLists.txt                      | 7 +++++++
 clang/include/clang/CIR/CMakeLists.txt    | 0
 clang/include/clang/CMakeLists.txt        | 3 +++
 clang/include/clang/Config/config.h.cmake | 3 +++
 clang/lib/CIR/CMakeLists.txt              | 0
 clang/lib/CMakeLists.txt                  | 4 ++++
 clang/test/CMakeLists.txt                 | 1 +
 clang/test/lit.site.cfg.py.in             | 1 +
 8 files changed, 19 insertions(+)
 create mode 100644 clang/include/clang/CIR/CMakeLists.txt
 create mode 100644 clang/lib/CIR/CMakeLists.txt

diff --git a/clang/CMakeLists.txt b/clang/CMakeLists.txt
index 284b2af24ddaa..f092766fa19f0 100644
--- a/clang/CMakeLists.txt
+++ b/clang/CMakeLists.txt
@@ -165,6 +165,13 @@ if(CLANG_ENABLE_LIBXML2)
   endif()
 endif()
 
+if(CLANG_ENABLE_CIR)
+  if (NOT "${LLVM_ENABLE_PROJECTS}" MATCHES "MLIR|mlir")
+    message(FATAL_ERROR
+      "Cannot build ClangIR without MLIR in LLVM_ENABLE_PROJECTS")
+  endif()
+endif()
+
 include(CheckIncludeFile)
 check_include_file(sys/resource.h CLANG_HAVE_RLIMITS)
 
diff --git a/clang/include/clang/CIR/CMakeLists.txt b/clang/include/clang/CIR/CMakeLists.txt
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/clang/include/clang/CMakeLists.txt b/clang/include/clang/CMakeLists.txt
index 0dc9ea5ed8ac8..47ac70cd21690 100644
--- a/clang/include/clang/CMakeLists.txt
+++ b/clang/include/clang/CMakeLists.txt
@@ -1,5 +1,8 @@
 add_subdirectory(AST)
 add_subdirectory(Basic)
+if(CLANG_ENABLE_CIR)
+  add_subdirectory(CIR)
+endif()
 add_subdirectory(Driver)
 add_subdirectory(Parse)
 add_subdirectory(Sema)
diff --git a/clang/include/clang/Config/config.h.cmake b/clang/include/clang/Config/config.h.cmake
index 4015ac8040861..27ed69e21562b 100644
--- a/clang/include/clang/Config/config.h.cmake
+++ b/clang/include/clang/Config/config.h.cmake
@@ -83,4 +83,7 @@
 /* Spawn a new process clang.exe for the CC1 tool invocation, when necessary */
 #cmakedefine01 CLANG_SPAWN_CC1
 
+/* Whether CIR is built into Clang */
+#cmakedefine01 CLANG_ENABLE_CIR
+
 #endif
diff --git a/clang/lib/CIR/CMakeLists.txt b/clang/lib/CIR/CMakeLists.txt
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/clang/lib/CMakeLists.txt b/clang/lib/CMakeLists.txt
index 0cac86451f39e..14ba55360fe05 100644
--- a/clang/lib/CMakeLists.txt
+++ b/clang/lib/CMakeLists.txt
@@ -31,3 +31,7 @@ if(CLANG_INCLUDE_TESTS)
 endif()
 add_subdirectory(Interpreter)
 add_subdirectory(Support)
+
+if(CLANG_ENABLE_CIR)
+  add_subdirectory(CIR)
+endif()
diff --git a/clang/test/CMakeLists.txt b/clang/test/CMakeLists.txt
index fcfca354f4a75..df34a5707da33 100644
--- a/clang/test/CMakeLists.txt
+++ b/clang/test/CMakeLists.txt
@@ -9,6 +9,7 @@ llvm_canonicalize_cmake_booleans(
   CLANG_ENABLE_STATIC_ANALYZER
   CLANG_PLUGIN_SUPPORT
   CLANG_SPAWN_CC1
+  CLANG_ENABLE_CIR
   ENABLE_BACKTRACES
   LLVM_ENABLE_ZLIB
   LLVM_ENABLE_ZSTD
diff --git a/clang/test/lit.site.cfg.py.in b/clang/test/lit.site.cfg.py.in
index ef75770a2c3c9..6641811c58839 100644
--- a/clang/test/lit.site.cfg.py.in
+++ b/clang/test/lit.site.cfg.py.in
@@ -27,6 +27,7 @@ config.clang_default_pie_on_linux = @CLANG_DEFAULT_PIE_ON_LINUX@
 config.clang_default_cxx_stdlib = "@CLANG_DEFAULT_CXX_STDLIB@"
 config.clang_staticanalyzer = @CLANG_ENABLE_STATIC_ANALYZER@
 config.clang_staticanalyzer_z3 = @LLVM_WITH_Z3@
+config.clang_enable_cir = @CLANG_ENABLE_CIR@
 config.clang_examples = @CLANG_BUILD_EXAMPLES@
 config.enable_shared = @ENABLE_SHARED@
 config.enable_backtrace = @ENABLE_BACKTRACES@

From a952c123880eb1168f1021b116485e27170d48ca Mon Sep 17 00:00:00 2001
From: Marius Brehler <marius.brehler@iml.fraunhofer.de>
Date: Thu, 11 Apr 2024 22:58:47 +0200
Subject: [PATCH 192/886] [GitHub] Set top level token permission (#87326)

---
 .github/workflows/pr-code-format.yml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/.github/workflows/pr-code-format.yml b/.github/workflows/pr-code-format.yml
index 10b18f245d896..983838858ba43 100644
--- a/.github/workflows/pr-code-format.yml
+++ b/.github/workflows/pr-code-format.yml
@@ -1,4 +1,8 @@
 name: "Check code formatting"
+
+permissions:
+  contents: read
+
 on:
   pull_request:
     branches:

From 84df7a09f8da5809b85fd097015e5ac6cc8a3f88 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Thu, 11 Apr 2024 14:44:55 -0700
Subject: [PATCH 193/886] [clang][modules] Do not resolve `HeaderFileInfo`
 externally in `ASTWriter` (#87848)

Clang uses the `HeaderFileInfo` struct to track bits of information on
header files, which gets used throughout the compiler. We also use this
to compute the set of affecting module maps in `ASTWriter` and in the
end serialize the information into the `HEADER_SEARCH_TABLE` record of a
PCM file, allowing clients to learn about headers from the module. In
doing so, Clang asks for existing `HeaderFileInfo` for all known
`FileEntries`. Note that this asks the loaded PCM files for the
information they have on each header file in question. This seems
unnecessary: we only want to serialize information on header files that
either belong to the current module or that got included textually.
Loaded PCM files can't provide us with any useful information.

For explicit modules with lazy loading (using `-fmodule-map-file=<path>`
with `-fmodule-file=<name>=<path>`) the compiler knows about header
files listed in the module map files on the command-line. This can be a
large number.

Asking for existing `HeaderFileInfo` can trigger deserialization of
`HEADER_SEARCH_TABLE` from loaded PCM files. Keys of the on-disk hash
table consist of the header file size and modification time. However,
with explicit modules Clang zeroes out the modification time. Moreover,
if you import lots of modules, some of their header files end up having
identical sizes. This means lots of hash collisions that can only be
resolved by running the serialized filename through `FileManager` and
comparing equality of the `FileEntry`. This ends up being super
expensive, essentially re-stating lots of the transitively loaded SDK
header files.

This patch cleans up the API for getting `HeaderFileInfo` and makes sure
`ASTWriter` uses the version that doesn't ask loaded PCM files for more
information. This removes the excessive stat traffic coming from
`ASTWriter` hopefully without changing observable behavior.
---
 .../clangd/index/SymbolCollector.cpp          |  2 +-
 clang/include/clang/Lex/HeaderSearch.h        | 24 +++---
 clang/lib/Lex/HeaderSearch.cpp                | 77 ++++++++++---------
 clang/lib/Serialization/ASTWriter.cpp         | 11 +--
 4 files changed, 60 insertions(+), 54 deletions(-)

diff --git a/clang-tools-extra/clangd/index/SymbolCollector.cpp b/clang-tools-extra/clangd/index/SymbolCollector.cpp
index 85b8fc549b016..969f27af6fb1d 100644
--- a/clang-tools-extra/clangd/index/SymbolCollector.cpp
+++ b/clang-tools-extra/clangd/index/SymbolCollector.cpp
@@ -409,7 +409,7 @@ class SymbolCollector::HeaderFileURICache {
     // Framework headers are spelled as <FrameworkName/Foo.h>, not
     // "path/FrameworkName.framework/Headers/Foo.h".
     auto &HS = PP->getHeaderSearchInfo();
-    if (const auto *HFI = HS.getExistingFileInfo(*FE, /*WantExternal*/ false))
+    if (const auto *HFI = HS.getExistingLocalFileInfo(*FE))
       if (!HFI->Framework.empty())
         if (auto Spelling =
                 getFrameworkHeaderIncludeSpelling(*FE, HFI->Framework, HS))
diff --git a/clang/include/clang/Lex/HeaderSearch.h b/clang/include/clang/Lex/HeaderSearch.h
index 855f81f775f8a..c5f90ef4cb368 100644
--- a/clang/include/clang/Lex/HeaderSearch.h
+++ b/clang/include/clang/Lex/HeaderSearch.h
@@ -547,14 +547,15 @@ class HeaderSearch {
   /// Return whether the specified file is a normal header,
   /// a system header, or a C++ friendly system header.
   SrcMgr::CharacteristicKind getFileDirFlavor(FileEntryRef File) {
-    return (SrcMgr::CharacteristicKind)getFileInfo(File).DirInfo;
+    if (const HeaderFileInfo *HFI = getExistingFileInfo(File))
+      return (SrcMgr::CharacteristicKind)HFI->DirInfo;
+    return (SrcMgr::CharacteristicKind)HeaderFileInfo().DirInfo;
   }
 
   /// Mark the specified file as a "once only" file due to
   /// \#pragma once.
   void MarkFileIncludeOnce(FileEntryRef File) {
-    HeaderFileInfo &FI = getFileInfo(File);
-    FI.isPragmaOnce = true;
+    getFileInfo(File).isPragmaOnce = true;
   }
 
   /// Mark the specified file as a system header, e.g. due to
@@ -834,16 +835,17 @@ class HeaderSearch {
 
   unsigned header_file_size() const { return FileInfo.size(); }
 
-  /// Return the HeaderFileInfo structure for the specified FileEntry,
-  /// in preparation for updating it in some way.
+  /// Return the HeaderFileInfo structure for the specified FileEntry, in
+  /// preparation for updating it in some way.
   HeaderFileInfo &getFileInfo(FileEntryRef FE);
 
-  /// Return the HeaderFileInfo structure for the specified FileEntry,
-  /// if it has ever been filled in.
-  /// \param WantExternal Whether the caller wants purely-external header file
-  ///        info (where \p External is true).
-  const HeaderFileInfo *getExistingFileInfo(FileEntryRef FE,
-                                            bool WantExternal = true) const;
+  /// Return the HeaderFileInfo structure for the specified FileEntry, if it has
+  /// ever been filled in (either locally or externally).
+  const HeaderFileInfo *getExistingFileInfo(FileEntryRef FE) const;
+
+  /// Return the headerFileInfo structure for the specified FileEntry, if it has
+  /// ever been filled in locally.
+  const HeaderFileInfo *getExistingLocalFileInfo(FileEntryRef FE) const;
 
   SearchDirIterator search_dir_begin() { return {*this, 0}; }
   SearchDirIterator search_dir_end() { return {*this, SearchDirs.size()}; }
diff --git a/clang/lib/Lex/HeaderSearch.cpp b/clang/lib/Lex/HeaderSearch.cpp
index f0750e5336b6a..0632882b29614 100644
--- a/clang/lib/Lex/HeaderSearch.cpp
+++ b/clang/lib/Lex/HeaderSearch.cpp
@@ -946,9 +946,13 @@ OptionalFileEntryRef HeaderSearch::LookupFile(
       // If we have no includer, that means we're processing a #include
       // from a module build. We should treat this as a system header if we're
       // building a [system] module.
-      bool IncluderIsSystemHeader =
-          Includer ? getFileInfo(*Includer).DirInfo != SrcMgr::C_User :
-          BuildSystemModule;
+      bool IncluderIsSystemHeader = [&]() {
+        if (!Includer)
+          return BuildSystemModule;
+        const HeaderFileInfo *HFI = getExistingFileInfo(*Includer);
+        assert(HFI && "includer without file info");
+        return HFI->DirInfo != SrcMgr::C_User;
+      }();
       if (OptionalFileEntryRef FE = getFileAndSuggestModule(
               TmpDir, IncludeLoc, IncluderAndDir.second, IncluderIsSystemHeader,
               RequestingModule, SuggestedModule)) {
@@ -963,10 +967,11 @@ OptionalFileEntryRef HeaderSearch::LookupFile(
         // Note that we only use one of FromHFI/ToHFI at once, due to potential
         // reallocation of the underlying vector potentially making the first
         // reference binding dangling.
-        HeaderFileInfo &FromHFI = getFileInfo(*Includer);
-        unsigned DirInfo = FromHFI.DirInfo;
-        bool IndexHeaderMapHeader = FromHFI.IndexHeaderMapHeader;
-        StringRef Framework = FromHFI.Framework;
+        const HeaderFileInfo *FromHFI = getExistingFileInfo(*Includer);
+        assert(FromHFI && "includer without file info");
+        unsigned DirInfo = FromHFI->DirInfo;
+        bool IndexHeaderMapHeader = FromHFI->IndexHeaderMapHeader;
+        StringRef Framework = FromHFI->Framework;
 
         HeaderFileInfo &ToHFI = getFileInfo(*FE);
         ToHFI.DirInfo = DirInfo;
@@ -1153,10 +1158,12 @@ OptionalFileEntryRef HeaderSearch::LookupFile(
   // "Foo" is the name of the framework in which the including header was found.
   if (!Includers.empty() && Includers.front().first && !isAngled &&
       !Filename.contains('/')) {
-    HeaderFileInfo &IncludingHFI = getFileInfo(*Includers.front().first);
-    if (IncludingHFI.IndexHeaderMapHeader) {
+    const HeaderFileInfo *IncludingHFI =
+        getExistingFileInfo(*Includers.front().first);
+    assert(IncludingHFI && "includer without file info");
+    if (IncludingHFI->IndexHeaderMapHeader) {
       SmallString<128> ScratchFilename;
-      ScratchFilename += IncludingHFI.Framework;
+      ScratchFilename += IncludingHFI->Framework;
       ScratchFilename += '/';
       ScratchFilename += Filename;
 
@@ -1286,11 +1293,11 @@ OptionalFileEntryRef HeaderSearch::LookupSubframeworkHeader(
   }
 
   // This file is a system header or C++ unfriendly if the old file is.
-  //
-  // Note that the temporary 'DirInfo' is required here, as either call to
-  // getFileInfo could resize the vector and we don't want to rely on order
-  // of evaluation.
-  unsigned DirInfo = getFileInfo(ContextFileEnt).DirInfo;
+  const HeaderFileInfo *ContextHFI = getExistingFileInfo(ContextFileEnt);
+  assert(ContextHFI && "context file without file info");
+  // Note that the temporary 'DirInfo' is required here, as the call to
+  // getFileInfo could resize the vector and might invalidate 'ContextHFI'.
+  unsigned DirInfo = ContextHFI->DirInfo;
   getFileInfo(*File).DirInfo = DirInfo;
 
   FrameworkName.pop_back(); // remove the trailing '/'
@@ -1348,8 +1355,6 @@ static void mergeHeaderFileInfo(HeaderFileInfo &HFI,
     HFI.Framework = OtherHFI.Framework;
 }
 
-/// getFileInfo - Return the HeaderFileInfo structure for the specified
-/// FileEntry.
 HeaderFileInfo &HeaderSearch::getFileInfo(FileEntryRef FE) {
   if (FE.getUID() >= FileInfo.size())
     FileInfo.resize(FE.getUID() + 1);
@@ -1366,27 +1371,20 @@ HeaderFileInfo &HeaderSearch::getFileInfo(FileEntryRef FE) {
   }
 
   HFI->IsValid = true;
-  // We have local information about this header file, so it's no longer
-  // strictly external.
+  // We assume the caller has local information about this header file, so it's
+  // no longer strictly external.
   HFI->External = false;
   return *HFI;
 }
 
-const HeaderFileInfo *
-HeaderSearch::getExistingFileInfo(FileEntryRef FE, bool WantExternal) const {
-  // If we have an external source, ensure we have the latest information.
-  // FIXME: Use a generation count to check whether this is really up to date.
+const HeaderFileInfo *HeaderSearch::getExistingFileInfo(FileEntryRef FE) const {
   HeaderFileInfo *HFI;
   if (ExternalSource) {
-    if (FE.getUID() >= FileInfo.size()) {
-      if (!WantExternal)
-        return nullptr;
+    if (FE.getUID() >= FileInfo.size())
       FileInfo.resize(FE.getUID() + 1);
-    }
 
     HFI = &FileInfo[FE.getUID()];
-    if (!WantExternal && (!HFI->IsValid || HFI->External))
-      return nullptr;
+    // FIXME: Use a generation count to check whether this is really up to date.
     if (!HFI->Resolved) {
       auto ExternalHFI = ExternalSource->GetHeaderFileInfo(FE);
       if (ExternalHFI.IsValid) {
@@ -1395,16 +1393,25 @@ HeaderSearch::getExistingFileInfo(FileEntryRef FE, bool WantExternal) const {
           mergeHeaderFileInfo(*HFI, ExternalHFI);
       }
     }
-  } else if (FE.getUID() >= FileInfo.size()) {
-    return nullptr;
-  } else {
+  } else if (FE.getUID() < FileInfo.size()) {
     HFI = &FileInfo[FE.getUID()];
+  } else {
+    HFI = nullptr;
   }
 
-  if (!HFI->IsValid || (HFI->External && !WantExternal))
-    return nullptr;
+  return (HFI && HFI->IsValid) ? HFI : nullptr;
+}
+
+const HeaderFileInfo *
+HeaderSearch::getExistingLocalFileInfo(FileEntryRef FE) const {
+  HeaderFileInfo *HFI;
+  if (FE.getUID() < FileInfo.size()) {
+    HFI = &FileInfo[FE.getUID()];
+  } else {
+    HFI = nullptr;
+  }
 
-  return HFI;
+  return (HFI && HFI->IsValid && !HFI->External) ? HFI : nullptr;
 }
 
 bool HeaderSearch::isFileMultipleIncludeGuarded(FileEntryRef File) const {
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index ffc53292e3912..330dbd6233502 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -185,8 +185,7 @@ GetAffectingModuleMaps(const Preprocessor &PP, Module *RootModule) {
     if (!File)
       continue;
 
-    const HeaderFileInfo *HFI =
-        HS.getExistingFileInfo(*File, /*WantExternal*/ false);
+    const HeaderFileInfo *HFI = HS.getExistingLocalFileInfo(*File);
     if (!HFI || (HFI->isModuleHeader && !HFI->isCompilingModuleHeader))
       continue;
 
@@ -2052,14 +2051,12 @@ void ASTWriter::WriteHeaderSearch(const HeaderSearch &HS) {
     if (!File)
       continue;
 
-    // Get the file info. This will load info from the external source if
-    // necessary. Skip emitting this file if we have no information on it
-    // as a header file (in which case HFI will be null) or if it hasn't
+    // Get the file info. Skip emitting this file if we have no information on
+    // it as a header file (in which case HFI will be null) or if it hasn't
     // changed since it was loaded. Also skip it if it's for a modular header
     // from a different module; in that case, we rely on the module(s)
     // containing the header to provide this information.
-    const HeaderFileInfo *HFI =
-        HS.getExistingFileInfo(*File, /*WantExternal*/!Chain);
+    const HeaderFileInfo *HFI = HS.getExistingLocalFileInfo(*File);
     if (!HFI || (HFI->isModuleHeader && !HFI->isCompilingModuleHeader))
       continue;
 

From 2837c975f712cc2727dfb939aee8cc9c5b086c1a Mon Sep 17 00:00:00 2001
From: Mike Rice <michael.p.rice@intel.com>
Date: Thu, 11 Apr 2024 15:19:34 -0700
Subject: [PATCH 194/886] Fix typo and formatting in SemaSYCL merge (#13370)

---
 clang/include/clang/Sema/Sema.h     |   8 --
 clang/include/clang/Sema/SemaSYCL.h |   5 +-
 clang/lib/Sema/SemaSYCL.cpp         | 155 +++++++++++++++-------------
 clang/lib/Sema/TreeTransform.h      |   3 +-
 4 files changed, 88 insertions(+), 83 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 68f7fb48c4dc5..e6bedd1509bb4 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -6681,14 +6681,6 @@ class Sema final : public SemaBase {
   void CheckSubscriptAccessOfNoDeref(const ArraySubscriptExpr *E);
   void CheckAddressOfNoDeref(const Expr *E);
 
-public:
-  ExprResult BuildSYCLUniqueStableIdExpr(SourceLocation OpLoc,
-                                         SourceLocation LParen,
-                                         SourceLocation RParen, Expr *E);
-  ExprResult ActOnSYCLUniqueStableIdExpr(SourceLocation OpLoc,
-                                         SourceLocation LParen,
-                                         SourceLocation RParen, Expr *E);
-
   ///@}
 
   //
diff --git a/clang/include/clang/Sema/SemaSYCL.h b/clang/include/clang/Sema/SemaSYCL.h
index 9b6d883996bcb..6a16be6b94818 100644
--- a/clang/include/clang/Sema/SemaSYCL.h
+++ b/clang/include/clang/Sema/SemaSYCL.h
@@ -109,9 +109,7 @@ class SYCLIntegrationHeader {
   /// Signals that emission of __sycl_host_pipe_registration type and
   /// declaration of variable __sycl_host_pipe_registrar of this type in
   /// integration header is required.
-  void addHostPipeRegistration() {
-    NeedToEmitHostPipeRegistration = true;
-  }
+  void addHostPipeRegistration() { NeedToEmitHostPipeRegistration = true; }
 
 private:
   // Kernel actual parameter descriptor.
@@ -217,7 +215,6 @@ class SYCLIntegrationFooter {
 
 class SemaSYCL : public SemaBase {
 private:
-
   // We store SYCL Kernels here and handle separately -- which is a hack.
   // FIXME: It would be best to refactor this.
   llvm::SetVector<Decl *> SyclDeviceDecls;
diff --git a/clang/lib/Sema/SemaSYCL.cpp b/clang/lib/Sema/SemaSYCL.cpp
index 80932bb56e78b..ec137637236d3 100644
--- a/clang/lib/Sema/SemaSYCL.cpp
+++ b/clang/lib/Sema/SemaSYCL.cpp
@@ -226,7 +226,7 @@ ExprResult SemaSYCL::ActOnSYCLBuiltinNumBasesExpr(ParsedType PT) {
 }
 
 ExprResult SemaSYCL::BuildSYCLBuiltinNumBasesExpr(SourceLocation Loc,
-                                              QualType SourceTy) {
+                                                  QualType SourceTy) {
   if (!SourceTy->isDependentType()) {
     if (SemaRef.RequireCompleteType(
             Loc, SourceTy, diag::err_sycl_type_trait_requires_complete_type,
@@ -426,8 +426,7 @@ static void checkSYCLType(SemaSYCL &S, QualType Ty, SourceRange Loc,
 
   // zero length arrays
   if (isZeroSizedArray(S, Ty)) {
-    S.DiagIfDeviceCode(Loc.getBegin(),
-                              diag::err_typecheck_zero_array_size)
+    S.DiagIfDeviceCode(Loc.getBegin(), diag::err_typecheck_zero_array_size)
         << 1;
     Emitting = true;
   }
@@ -617,7 +616,8 @@ class DiagDeviceFunction : public RecursiveASTVisitor<DiagDeviceFunction> {
       // builtin is supported for SYCL
       unsigned BuiltinID = Callee->getBuiltinID();
       if (BuiltinID && !IsSyclMathFunc(BuiltinID)) {
-        StringRef Name = SemaSYCLRef.getASTContext().BuiltinInfo.getName(BuiltinID);
+        StringRef Name =
+            SemaSYCLRef.getASTContext().BuiltinInfo.getName(BuiltinID);
         SemaSYCLRef.Diag(e->getExprLoc(), diag::err_builtin_target_unsupported)
             << Name << "SYCL device";
       }
@@ -799,7 +799,7 @@ class SingleDeviceFunctionTracker {
     if (!CurrentDecl->isDefined() && !CurrentDecl->hasAttr<SYCLKernelAttr>() &&
         !CurrentDecl->hasAttr<SYCLDeviceAttr>())
       Parent.SemaSYCLRef.addFDToReachableFromSyclDevice(CurrentDecl,
-                                                    CallStack.back());
+                                                        CallStack.back());
 
     // If this is a parallel_for_work_item that is declared in the
     // sycl namespace, mark it with the WorkItem scope attribute.
@@ -810,9 +810,8 @@ class SingleDeviceFunctionTracker {
         CurrentDecl->getIdentifier()->getName() == "parallel_for_work_item" &&
         isDeclaredInSYCLNamespace(CurrentDecl) &&
         !CurrentDecl->hasAttr<SYCLScopeAttr>()) {
-      CurrentDecl->addAttr(
-          SYCLScopeAttr::CreateImplicit(Parent.SemaSYCLRef.getASTContext(),
-                                        SYCLScopeAttr::Level::WorkItem));
+      CurrentDecl->addAttr(SYCLScopeAttr::CreateImplicit(
+          Parent.SemaSYCLRef.getASTContext(), SYCLScopeAttr::Level::WorkItem));
     }
 
     // We previously thought we could skip this function if we'd seen it before,
@@ -836,8 +835,8 @@ class SingleDeviceFunctionTracker {
     // Collect attributes for functions that aren't the root kernel.
     if (!CallStack.empty()) {
       bool DirectlyCalled = CallStack.size() == 1;
-      collectSYCLAttributes(Parent.SemaSYCLRef, CurrentDecl, CollectedAttributes,
-                            DirectlyCalled);
+      collectSYCLAttributes(Parent.SemaSYCLRef, CurrentDecl,
+                            CollectedAttributes, DirectlyCalled);
     }
 
     // Calculate the kernel body.  Note the 'isSYCLKernelBodyFunction' only
@@ -1649,7 +1648,7 @@ class SyclKernelFieldChecker : public SyclKernelFieldHandler {
         cast<ClassTemplateSpecializationDecl>(PropListTy->getAsRecordDecl());
     if (AccPropListDecl->getTemplateArgs().size() != 1)
       return SemaSYCLRef.Diag(Loc,
-                          diag::err_sycl_invalid_property_list_param_number)
+                              diag::err_sycl_invalid_property_list_param_number)
              << "accessor_property_list";
 
     const auto TemplArg = AccPropListDecl->getTemplateArgs()[0];
@@ -1679,7 +1678,7 @@ class SyclKernelFieldChecker : public SyclKernelFieldHandler {
         cast<ClassTemplateSpecializationDecl>(PropTy->getAsRecordDecl());
     if (PropDecl->getTemplateArgs().size() != 1)
       return SemaSYCLRef.Diag(Loc,
-                          diag::err_sycl_invalid_property_list_param_number)
+                              diag::err_sycl_invalid_property_list_param_number)
              << "buffer_location";
 
     const auto BufferLoc = PropDecl->getTemplateArgs()[0];
@@ -1709,7 +1708,7 @@ class SyclKernelFieldChecker : public SyclKernelFieldHandler {
          SemaSYCL::isSyclType(Ty, SYCLTypeAttr::annotated_arg)) &&
         (StructFieldDepth > 0 || StructBaseDepth > 0))
       return SemaSYCLRef.Diag(Loc.getBegin(),
-                          diag::err_bad_kernel_param_data_members)
+                              diag::err_bad_kernel_param_data_members)
              << Ty << /*Struct*/ 1;
 
     const RecordDecl *RecD = Ty->getAsRecordDecl();
@@ -1912,8 +1911,8 @@ class SyclKernelDecompMarker : public SyclKernelFieldHandler {
     } else if (PointerStack.pop_back_val()) {
       PointerStack.back() = true;
       if (!RD->hasAttr<SYCLGenerateNewTypeAttr>())
-        RD->addAttr(
-            SYCLGenerateNewTypeAttr::CreateImplicit(SemaSYCLRef.getASTContext()));
+        RD->addAttr(SYCLGenerateNewTypeAttr::CreateImplicit(
+            SemaSYCLRef.getASTContext()));
     }
     return true;
   }
@@ -1942,8 +1941,8 @@ class SyclKernelDecompMarker : public SyclKernelFieldHandler {
     } else if (PointerStack.pop_back_val()) {
       PointerStack.back() = true;
       if (!RD->hasAttr<SYCLGenerateNewTypeAttr>())
-        RD->addAttr(
-            SYCLGenerateNewTypeAttr::CreateImplicit(SemaSYCLRef.getASTContext()));
+        RD->addAttr(SYCLGenerateNewTypeAttr::CreateImplicit(
+            SemaSYCLRef.getASTContext()));
     }
     return true;
   }
@@ -1970,8 +1969,8 @@ class SyclKernelDecompMarker : public SyclKernelFieldHandler {
       PointerStack.pop_back();
     } else if (PointerStack.pop_back_val()) {
       if (!FD->hasAttr<SYCLGenerateNewTypeAttr>())
-        FD->addAttr(
-            SYCLGenerateNewTypeAttr::CreateImplicit(SemaSYCLRef.getASTContext()));
+        FD->addAttr(SYCLGenerateNewTypeAttr::CreateImplicit(
+            SemaSYCLRef.getASTContext()));
       PointerStack.back() = true;
     }
     return true;
@@ -2044,8 +2043,9 @@ class SyclKernelPointerHandler : public SyclKernelFieldHandler {
 
   void createBaseSpecifier(const CXXRecordDecl *Parent, const CXXRecordDecl *RD,
                            const CXXBaseSpecifier &BS) {
-    TypeSourceInfo *TInfo = SemaSYCLRef.getASTContext().getTrivialTypeSourceInfo(
-        QualType(RD->getTypeForDecl(), 0), SourceLocation());
+    TypeSourceInfo *TInfo =
+        SemaSYCLRef.getASTContext().getTrivialTypeSourceInfo(
+            QualType(RD->getTypeForDecl(), 0), SourceLocation());
     CXXBaseSpecifier *ModifiedBase = SemaSYCLRef.SemaRef.CheckBaseSpecifier(
         const_cast<CXXRecordDecl *>(Parent), SourceRange(), BS.isVirtual(),
         BS.getAccessSpecifier(), TInfo, SourceLocation());
@@ -2302,8 +2302,8 @@ class SyclKernelDeclCreator : public SyclKernelFieldHandler {
       const TemplateArgument &AccessModeArg =
           AccessorSpecializationDecl->getTemplateArgs().get(2);
       if (isReadOnlyAccessor(AccessModeArg))
-        Params.back()->addAttr(
-            SYCLAccessorReadonlyAttr::CreateImplicit(SemaSYCLRef.getASTContext()));
+        Params.back()->addAttr(SYCLAccessorReadonlyAttr::CreateImplicit(
+            SemaSYCLRef.getASTContext()));
     }
 
     // Add implicit attribute to parameter decl when it is a read only
@@ -2507,7 +2507,7 @@ class SyclKernelDeclCreator : public SyclKernelFieldHandler {
         SemaSYCLRef.getASTContext(), WrapperClass, SourceLocation(),
         SourceLocation(), /*Id=*/nullptr, FieldTy,
         SemaSYCLRef.getASTContext().getTrivialTypeSourceInfo(FieldTy,
-                                                         SourceLocation()),
+                                                             SourceLocation()),
         /*BW=*/nullptr, /*Mutable=*/false, /*InitStyle=*/ICIS_NoInit);
     Field->setAccess(AS_public);
     WrapperClass->addDecl(Field);
@@ -2674,7 +2674,7 @@ class ESIMDKernelDiagnostics : public SyclKernelFieldHandler {
 
     if (IsESIMD && !isSyclAccessorType(FieldTy))
       return SemaSYCLRef.Diag(KernelLoc,
-                          diag::err_sycl_esimd_not_supported_for_type)
+                              diag::err_sycl_esimd_not_supported_for_type)
              << RecordDecl;
     return true;
   }
@@ -2787,8 +2787,9 @@ class SyclOptReportCreator : public SyclKernelFieldHandler {
     if (KernelArgParent && KernelArgDescription == "decomposed struct/class")
       NameToEmitInDescription = KernelArgParent->getName();
 
-    unsigned KernelArgSize =
-        SemaSYCLRef.getASTContext().getTypeSizeInChars(KernelArgType).getQuantity();
+    unsigned KernelArgSize = SemaSYCLRef.getASTContext()
+                                 .getTypeSizeInChars(KernelArgType)
+                                 .getQuantity();
 
     SemaSYCLRef.getDiagnostics().getSYCLOptReport().AddKernelArgs(
         DC.getKernelDecl(), NameToEmitInDescription,
@@ -2814,8 +2815,9 @@ class SyclOptReportCreator : public SyclKernelFieldHandler {
   void addParam(const CXXBaseSpecifier &, QualType KernelArgType,
                 StringRef KernelArgDescription,
                 bool IsCompilerGeneratedType = false) {
-    unsigned KernelArgSize =
-        SemaSYCLRef.getASTContext().getTypeSizeInChars(KernelArgType).getQuantity();
+    unsigned KernelArgSize = SemaSYCLRef.getASTContext()
+                                 .getTypeSizeInChars(KernelArgType)
+                                 .getQuantity();
     SemaSYCLRef.getDiagnostics().getSYCLOptReport().AddKernelArgs(
         DC.getKernelDecl(), KernelArgType.getAsString(),
         IsCompilerGeneratedType ? "Compiler generated"
@@ -2826,8 +2828,9 @@ class SyclOptReportCreator : public SyclKernelFieldHandler {
 
   // Handles specialization constants.
   void addParam(QualType KernelArgType, std::string KernelArgDescription) {
-    unsigned KernelArgSize =
-        SemaSYCLRef.getASTContext().getTypeSizeInChars(KernelArgType).getQuantity();
+    unsigned KernelArgSize = SemaSYCLRef.getASTContext()
+                                 .getTypeSizeInChars(KernelArgType)
+                                 .getQuantity();
     SemaSYCLRef.getDiagnostics().getSYCLOptReport().AddKernelArgs(
         DC.getKernelDecl(), "", KernelArgType.getAsString(),
         KernelInvocationLoc, KernelArgSize,
@@ -2836,7 +2839,8 @@ class SyclOptReportCreator : public SyclKernelFieldHandler {
 
 public:
   static constexpr const bool VisitInsideSimpleContainers = false;
-  SyclOptReportCreator(SemaSYCL &S, SyclKernelDeclCreator &DC, SourceLocation Loc)
+  SyclOptReportCreator(SemaSYCL &S, SyclKernelDeclCreator &DC,
+                       SourceLocation Loc)
       : SyclKernelFieldHandler(S), DC(DC), KernelInvocationLoc(Loc) {}
 
   bool handleSyclSpecialType(FieldDecl *FD, QualType FieldTy) final {
@@ -3038,8 +3042,8 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
         DeclCreator.getParamVarDeclsForCurrentField()[0];
 
     QualType ParamType = KernelParameter->getOriginalType();
-    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(KernelParameter, ParamType,
-                                                 VK_LValue, KernelCallerSrcLoc);
+    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(
+        KernelParameter, ParamType, VK_LValue, KernelCallerSrcLoc);
     return DRE;
   }
 
@@ -3050,8 +3054,8 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
         DeclCreator.getParamVarDeclsForCurrentField()[0];
 
     QualType ParamType = KernelParameter->getOriginalType();
-    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(KernelParameter, ParamType,
-                                                 VK_LValue, KernelCallerSrcLoc);
+    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(
+        KernelParameter, ParamType, VK_LValue, KernelCallerSrcLoc);
 
     // Struct Type kernel arguments are decomposed. The pointer fields are
     // then wrapped inside a compiler generated struct. Therefore when
@@ -3081,8 +3085,8 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
     ParmVarDecl *KernelParameter =
         DeclCreator.getParamVarDeclsForCurrentField()[0];
     QualType ParamType = KernelParameter->getOriginalType();
-    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(KernelParameter, ParamType,
-                                                 VK_LValue, KernelCallerSrcLoc);
+    Expr *DRE = SemaSYCLRef.SemaRef.BuildDeclRefExpr(
+        KernelParameter, ParamType, VK_LValue, KernelCallerSrcLoc);
 
     // Unwrap the array.
     CXXRecordDecl *WrapperStruct = ParamType->getAsCXXRecordDecl();
@@ -3114,8 +3118,10 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
 
   void addFieldInit(FieldDecl *FD, QualType Ty, MultiExprArg ParamRef,
                     InitializationKind InitKind, InitializedEntity Entity) {
-    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
-    ExprResult Init = InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
+    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind,
+                                   ParamRef);
+    ExprResult Init =
+        InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
 
     InitListExpr *ParentILE = CollectionInitExprs.back();
     ParentILE->updateInit(SemaSYCLRef.getASTContext(), ParentILE->getNumInits(),
@@ -3127,8 +3133,10 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
     InitializedEntity Entity = InitializedEntity::InitializeBase(
         SemaSYCLRef.getASTContext(), &BS, /*IsInheritedVirtualBase*/ false,
         &VarEntity);
-    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind, std::nullopt);
-    ExprResult Init = InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, std::nullopt);
+    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind,
+                                   std::nullopt);
+    ExprResult Init =
+        InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, std::nullopt);
 
     InitListExpr *ParentILE = CollectionInitExprs.back();
     ParentILE->updateInit(SemaSYCLRef.getASTContext(), ParentILE->getNumInits(),
@@ -3141,7 +3149,8 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
         SemaSYCLRef.getASTContext(), &BS, /*IsInheritedVirtualBase*/ false,
         &VarEntity);
     InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind, Args);
-    ExprResult Init = InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, Args);
+    ExprResult Init =
+        InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, Args);
 
     InitListExpr *ParentILE = CollectionInitExprs.back();
     ParentILE->updateInit(SemaSYCLRef.getASTContext(), ParentILE->getNumInits(),
@@ -3157,8 +3166,10 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
         &VarEntity);
 
     Expr *ParamRef = createParamReferenceExpr();
-    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
-    ExprResult Init = InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
+    InitializationSequence InitSeq(SemaSYCLRef.SemaRef, Entity, InitKind,
+                                   ParamRef);
+    ExprResult Init =
+        InitSeq.Perform(SemaSYCLRef.SemaRef, Entity, InitKind, ParamRef);
 
     InitListExpr *ParentILE = CollectionInitExprs.back();
     ParentILE->updateInit(SemaSYCLRef.getASTContext(), ParentILE->getNumInits(),
@@ -3189,7 +3200,8 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
   Expr *createReinterpretCastExpr(Expr *E, QualType To) {
     return CXXReinterpretCastExpr::Create(
         SemaSYCLRef.getASTContext(), To, VK_PRValue, CK_BitCast, E,
-        /*Path=*/nullptr, SemaSYCLRef.getASTContext().getTrivialTypeSourceInfo(To),
+        /*Path=*/nullptr,
+        SemaSYCLRef.getASTContext().getTrivialTypeSourceInfo(To),
         SourceLocation(), SourceLocation(), SourceRange());
   }
 
@@ -3265,13 +3277,13 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
     ResultTy = ResultTy.getNonLValueExprType(SemaSYCLRef.getASTContext());
     llvm::SmallVector<Expr *, 4> ParamStmts;
     const auto *Proto = cast<FunctionProtoType>(Method->getType());
-    SemaSYCLRef.SemaRef.GatherArgumentsForCall(KernelCallerSrcLoc, Method, Proto, 0,
-                                           ParamDREs, ParamStmts);
+    SemaSYCLRef.SemaRef.GatherArgumentsForCall(KernelCallerSrcLoc, Method,
+                                               Proto, 0, ParamDREs, ParamStmts);
     // [kernel_obj or wrapper object].accessor.__init(_ValueType*,
     // range<int>, range<int>, id<int>)
     AddTo.push_back(CXXMemberCallExpr::Create(
-        SemaSYCLRef.getASTContext(), MethodME, ParamStmts, ResultTy, VK, KernelCallerSrcLoc,
-        FPOptionsOverride()));
+        SemaSYCLRef.getASTContext(), MethodME, ParamStmts, ResultTy, VK,
+        KernelCallerSrcLoc, FPOptionsOverride()));
   }
 
   // Creates an empty InitListExpr of the correct number of child-inits
@@ -3291,8 +3303,9 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
   }
 
   InitListExpr *createInitListExpr(QualType InitTy, uint64_t NumChildInits) {
-    InitListExpr *ILE = new (SemaSYCLRef.getASTContext()) InitListExpr(
-        SemaSYCLRef.getASTContext(), KernelCallerSrcLoc, {}, KernelCallerSrcLoc);
+    InitListExpr *ILE = new (SemaSYCLRef.getASTContext())
+        InitListExpr(SemaSYCLRef.getASTContext(), KernelCallerSrcLoc, {},
+                     KernelCallerSrcLoc);
     ILE->reserveInits(SemaSYCLRef.getASTContext(), NumChildInits);
     ILE->setType(InitTy);
 
@@ -3358,10 +3371,10 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
 
   // Generate __init call for kernel handler argument
   void handleSpecialType(QualType KernelHandlerTy) {
-    DeclRefExpr *KernelHandlerCloneRef =
-        DeclRefExpr::Create(SemaSYCLRef.getASTContext(), NestedNameSpecifierLoc(),
-                            KernelCallerSrcLoc, KernelHandlerClone, false,
-                            DeclarationNameInfo(), KernelHandlerTy, VK_LValue);
+    DeclRefExpr *KernelHandlerCloneRef = DeclRefExpr::Create(
+        SemaSYCLRef.getASTContext(), NestedNameSpecifierLoc(),
+        KernelCallerSrcLoc, KernelHandlerClone, false, DeclarationNameInfo(),
+        KernelHandlerTy, VK_LValue);
     const auto *RecordDecl =
         KernelHandlerClone->getType()->getAsCXXRecordDecl();
     MemberExprBases.push_back(KernelHandlerCloneRef);
@@ -3563,9 +3576,9 @@ class SyclKernelBodyCreator : public SyclKernelFieldHandler {
                              DeclCreator.getKernelDecl(), KernelHandlerArg);
 
     // Add declaration statement to openCL kernel body
-    Stmt *DS =
-        new (SemaSYCLRef.getASTContext()) DeclStmt(DeclGroupRef(KernelHandlerClone),
-                                       KernelCallerSrcLoc, KernelCallerSrcLoc);
+    Stmt *DS = new (SemaSYCLRef.getASTContext())
+        DeclStmt(DeclGroupRef(KernelHandlerClone), KernelCallerSrcLoc,
+                 KernelCallerSrcLoc);
     BodyStmts.push_back(DS);
 
     // Generate
@@ -3708,7 +3721,8 @@ class SyclKernelIntHeaderCreator : public SyclKernelFieldHandler {
 
 public:
   static constexpr const bool VisitInsideSimpleContainers = false;
-  SyclKernelIntHeaderCreator(bool IsESIMD, SemaSYCL &S, SYCLIntegrationHeader &H,
+  SyclKernelIntHeaderCreator(bool IsESIMD, SemaSYCL &S,
+                             SYCLIntegrationHeader &H,
                              const CXXRecordDecl *KernelObj, QualType NameType,
                              FunctionDecl *KernelFunc)
       : SyclKernelFieldHandler(S), Header(H) {
@@ -3851,7 +3865,8 @@ class SyclKernelIntHeaderCreator : public SyclKernelFieldHandler {
   }
 
   bool nextElement(QualType ET, uint64_t Index) final {
-    int64_t Size = SemaSYCLRef.getASTContext().getTypeSizeInChars(ET).getQuantity();
+    int64_t Size =
+        SemaSYCLRef.getASTContext().getTypeSizeInChars(ET).getQuantity();
     CurOffset = ArrayBaseOffsets.back() + Size * Index;
     return true;
   }
@@ -4064,7 +4079,7 @@ class SYCLKernelNameTypeVisitor
 };
 
 void SemaSYCL::CheckSYCLKernelCall(FunctionDecl *KernelFunc,
-                               ArrayRef<const Expr *> Args) {
+                                   ArrayRef<const Expr *> Args) {
   QualType KernelNameType =
       calculateKernelNameType(getASTContext(), KernelFunc);
   SYCLKernelNameTypeVisitor KernelNameTypeVisitor(
@@ -4074,7 +4089,7 @@ void SemaSYCL::CheckSYCLKernelCall(FunctionDecl *KernelFunc,
 
   // FIXME: In place until the library works around its 'host' invocation
   // issues.
-  if (SemaRef.LangOpts.SYCLIsDevice)
+  if (!SemaRef.LangOpts.SYCLIsDevice)
     return;
 
   const CXXRecordDecl *KernelObj =
@@ -4227,7 +4242,7 @@ void SemaSYCL::SetSYCLKernelNames() {
 //
 //
 void SemaSYCL::ConstructOpenCLKernel(FunctionDecl *KernelCallerFunc,
-                                 MangleContext &MC) {
+                                     MangleContext &MC) {
   // The first argument to the KernelCallerFunc is the lambda object.
   const CXXRecordDecl *KernelObj =
       GetSYCLKernelObjectType(KernelCallerFunc)->getAsCXXRecordDecl();
@@ -4274,7 +4289,8 @@ void SemaSYCL::ConstructOpenCLKernel(FunctionDecl *KernelCallerFunc,
                                     CallOperator);
   SyclKernelIntHeaderCreator int_header(
       IsSIMDKernel, *this, getSyclIntegrationHeader(), KernelObj,
-      calculateKernelNameType(getASTContext(), KernelCallerFunc), KernelCallerFunc);
+      calculateKernelNameType(getASTContext(), KernelCallerFunc),
+      KernelCallerFunc);
 
   SyclKernelIntFooterCreator int_footer(*this, getSyclIntegrationFooter());
   SyclOptReportCreator opt_report(*this, kernel_decl, KernelObj->getLocation());
@@ -4605,7 +4621,8 @@ SemaSYCL::DiagIfDeviceCode(SourceLocation Loc, unsigned DiagID,
       return SemaDiagnosticBuilder::K_ImmediateWithCallStack;
     if (!FD)
       return SemaDiagnosticBuilder::K_Nop;
-    if (SemaRef.getEmissionStatus(FD) == Sema::FunctionEmissionStatus::Emitted) {
+    if (SemaRef.getEmissionStatus(FD) ==
+        Sema::FunctionEmissionStatus::Emitted) {
       // Skip the diagnostic if we know it won't be emitted.
       if ((SemaRef.getEmissionReason(FD) & Reason) ==
           Sema::DeviceDiagnosticReason::None)
@@ -4711,9 +4728,9 @@ void SemaSYCL::deepTypeCheckForDevice(SourceLocation UsedAt,
 }
 
 void SemaSYCL::finalizeSYCLDelayedAnalysis(const FunctionDecl *Caller,
-                                       const FunctionDecl *Callee,
-                                       SourceLocation Loc,
-                                       DeviceDiagnosticReason Reason) {
+                                           const FunctionDecl *Callee,
+                                           SourceLocation Loc,
+                                           DeviceDiagnosticReason Reason) {
   Callee = Callee->getMostRecentDecl();
 
   // If the reason for the emission of this diagnostic is not SYCL-specific,
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index b976e5d41b95e..81d2a3b1f7f5e 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -2642,8 +2642,7 @@ class TreeTransform {
   ExprResult RebuildSYCLUniqueStableIdExpr(SourceLocation OpLoc,
                                            SourceLocation LParen,
                                            SourceLocation RParen, Expr *E) {
-    return getSema().SYCL().BuildUniqueStableIdExpr(OpLoc, LParen, RParen,
-                                                        E);
+    return getSema().SYCL().BuildUniqueStableIdExpr(OpLoc, LParen, RParen, E);
   }
 
   /// Build a new predefined expression.

From 95fbd8d19dff10b5a734e3db1b29cba2da7a983f Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Thu, 11 Apr 2024 15:22:35 -0700
Subject: [PATCH 195/886] [clangd] Don't ignore external HFI in
 `SymbolCollector` (#88446)

The `FileEntry` corresponds to a `FileID` containing the
`SourceLocation` of a `NamedDecl` which (I think) might've been
deserialized from a PCM file. Considering external `HeaderFileInfo` here
is most likely the right thing to do here in order to get the correct
spelling in case the current compiler instance has not register this
file as a header yet.
---
 clang-tools-extra/clangd/index/SymbolCollector.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang-tools-extra/clangd/index/SymbolCollector.cpp b/clang-tools-extra/clangd/index/SymbolCollector.cpp
index 969f27af6fb1d..5c4e2150cf312 100644
--- a/clang-tools-extra/clangd/index/SymbolCollector.cpp
+++ b/clang-tools-extra/clangd/index/SymbolCollector.cpp
@@ -409,7 +409,7 @@ class SymbolCollector::HeaderFileURICache {
     // Framework headers are spelled as <FrameworkName/Foo.h>, not
     // "path/FrameworkName.framework/Headers/Foo.h".
     auto &HS = PP->getHeaderSearchInfo();
-    if (const auto *HFI = HS.getExistingLocalFileInfo(*FE))
+    if (const auto *HFI = HS.getExistingFileInfo(*FE))
       if (!HFI->Framework.empty())
         if (auto Spelling =
                 getFrameworkHeaderIncludeSpelling(*FE, HFI->Framework, HS))

From 9a36077e4db30c7da603620762036d4a430e4e62 Mon Sep 17 00:00:00 2001
From: jimingham <jingham@apple.com>
Date: Thu, 11 Apr 2024 15:22:58 -0700
Subject: [PATCH 196/886] Fix error in unrecognized register name handling for
 "SBFrame.register" (#88047)

The code returned lldb.SBValue() when you passed in an unrecognized
register name. But referring to "lldb" is apparently not legal within
the module.

I changed this to just return SBValue(), but then this construct:

(lldb) script
>>> for reg_set in lldb.target.process.thread[0].frames[0].register
...    print(reg)

Runs forever printing "No Value". The __getitem__(key) gets called with
a monotonically increasing by 1 series of integers. I don't know why
Python decided the class we defined should have a generator that returns
positive integers in order, but we can add a more useful one here by
returning an iterator over the flattened list of registers.

Note, the not very aptly named "SBFrame.registers" is an iterator over
register sets, not registers, so the two are not redundant.
---
 lldb/bindings/interface/SBFrameExtensions.i  | 12 +++++++++++-
 lldb/test/API/python_api/frame/TestFrames.py | 12 ++++++++++++
 2 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/lldb/bindings/interface/SBFrameExtensions.i b/lldb/bindings/interface/SBFrameExtensions.i
index 43b22ed7a6b32..e0472280666ab 100644
--- a/lldb/bindings/interface/SBFrameExtensions.i
+++ b/lldb/bindings/interface/SBFrameExtensions.i
@@ -44,6 +44,16 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                 def __init__(self, regs):
                     self.regs = regs
 
+                def __iter__(self):
+                    return self.get_registers()
+
+                def get_registers(self):
+                    for i in range(0,len(self.regs)):
+                        rs = self.regs[i]
+                        for j in range (0,rs.num_children):
+                            reg = rs.GetChildAtIndex(j)
+                            yield reg
+                          
                 def __getitem__(self, key):
                     if type(key) is str:
                         for i in range(0,len(self.regs)):
@@ -52,7 +62,7 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                                 reg = rs.GetChildAtIndex(j)
                                 if reg.name == key: return reg
                     else:
-                        return lldb.SBValue()
+                        return SBValue()
 
             return registers_access(self.registers)
 
diff --git a/lldb/test/API/python_api/frame/TestFrames.py b/lldb/test/API/python_api/frame/TestFrames.py
index a82b129bc8099..dfa96d51830ba 100644
--- a/lldb/test/API/python_api/frame/TestFrames.py
+++ b/lldb/test/API/python_api/frame/TestFrames.py
@@ -73,7 +73,19 @@ def test_get_arg_vals_for_call_stack(self):
                 gpr_reg_set = lldbutil.get_GPRs(frame)
                 pc_value = gpr_reg_set.GetChildMemberWithName("pc")
                 self.assertTrue(pc_value, "We should have a valid PC.")
+                # Make sure we can also get this from the "register" property:
+                iterator_pc_value = 0
+                found_pc = False
+                for reg in frame.register:
+                    if reg.name == "pc":
+                        found_pc = True
+                        iterator_pc_value = int(reg.GetValue(), 0)
+                        break
+
                 pc_value_int = int(pc_value.GetValue(), 0)
+                self.assertTrue(found_pc, "Found the PC value in the register list")
+                self.assertEqual(iterator_pc_value, pc_value_int, "The methods of finding pc match")
+    
                 # Make sure on arm targets we dont mismatch PC value on the basis of thumb bit.
                 # Frame PC will not have thumb bit set in case of a thumb
                 # instruction as PC.

From 4036a6946e5420bb77a93037c83732be600b9b0b Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <1802579+farzonl@users.noreply.github.com>
Date: Thu, 11 Apr 2024 15:26:25 -0700
Subject: [PATCH 197/886] [HLSL] move rcp to cgbuiltins (#88401)

Removing the intrinsic because there is no opCodes for rcp in DXIL or
SPIR-V.
Moving means we don't have to re-implement this feature for each
backend.

fixes #87784

Co-authored-by: Farzon Lotfi <farzon@farzon.com>
---
 clang/lib/CodeGen/CGBuiltin.cpp               |  13 +-
 clang/test/CodeGenHLSL/builtins/rcp.hlsl      | 127 ++++++++++++------
 .../Target/DirectX/DXILIntrinsicExpansion.cpp |  22 ---
 llvm/test/CodeGen/DirectX/rcp.ll              |  52 -------
 .../test/CodeGen/SPIRV/hlsl-intrinsics/rcp.ll | 126 +++++++++++++++++
 5 files changed, 224 insertions(+), 116 deletions(-)
 delete mode 100644 llvm/test/CodeGen/DirectX/rcp.ll
 create mode 100644 llvm/test/CodeGen/SPIRV/hlsl-intrinsics/rcp.ll

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index c052367d28782..c7b219dcfcec5 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18303,9 +18303,16 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
     Value *Op0 = EmitScalarExpr(E->getArg(0));
     if (!E->getArg(0)->getType()->hasFloatingRepresentation())
       llvm_unreachable("rcp operand must have a float representation");
-    return Builder.CreateIntrinsic(
-        /*ReturnType=*/Op0->getType(), Intrinsic::dx_rcp,
-        ArrayRef<Value *>{Op0}, nullptr, "dx.rcp");
+    llvm::Type *Ty = Op0->getType();
+    llvm::Type *EltTy = Ty->getScalarType();
+    Constant *One =
+        Ty->isVectorTy()
+            ? ConstantVector::getSplat(
+                  ElementCount::getFixed(
+                      dyn_cast<FixedVectorType>(Ty)->getNumElements()),
+                  ConstantFP::get(EltTy, 1.0))
+            : ConstantFP::get(EltTy, 1.0);
+    return Builder.CreateFDiv(One, Op0, "hlsl.rcp");
   }
   case Builtin::BI__builtin_hlsl_elementwise_rsqrt: {
     Value *Op0 = EmitScalarExpr(E->getArg(0));
diff --git a/clang/test/CodeGenHLSL/builtins/rcp.hlsl b/clang/test/CodeGenHLSL/builtins/rcp.hlsl
index 9b8406e1f0b60..eb89bcc4c7c01 100644
--- a/clang/test/CodeGenHLSL/builtins/rcp.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/rcp.hlsl
@@ -1,53 +1,102 @@
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -fnative-half-type \
 // RUN:   -emit-llvm -disable-llvm-passes -o - | FileCheck %s \ 
-// RUN:   --check-prefixes=CHECK,NATIVE_HALF
+// RUN:   --check-prefixes=CHECK,DXIL_CHECK,DXIL_NATIVE_HALF,NATIVE_HALF
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,NO_HALF
-
-// NATIVE_HALF: define noundef half @
-// NATIVE_HALF: %dx.rcp = call half @llvm.dx.rcp.f16(
-// NATIVE_HALF: ret half %dx.rcp
-// NO_HALF: define noundef float @"?test_rcp_half@@YA$halff@$halff@@Z"(
-// NO_HALF: %dx.rcp = call float @llvm.dx.rcp.f32(
-// NO_HALF: ret float %dx.rcp
+// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,DXIL_CHECK,NO_HALF,DXIL_NO_HALF
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-unknown-vulkan-compute %s -fnative-half-type \
+// RUN:   -emit-llvm -disable-llvm-passes -o - | FileCheck %s \ 
+// RUN:   --check-prefixes=CHECK,NATIVE_HALF,SPIR_NATIVE_HALF,SPIR_CHECK
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-unknown-vulkan-compute %s -emit-llvm -disable-llvm-passes \
+// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,NO_HALF,SPIR_NO_HALF,SPIR_CHECK
+
+// DXIL_NATIVE_HALF: define noundef half @
+// SPIR_NATIVE_HALF: define spir_func noundef half @
+// NATIVE_HALF: %hlsl.rcp = fdiv half 0xH3C00, %{{.*}} 
+// NATIVE_HALF: ret half %hlsl.rcp
+// DXIL_NO_HALF: define noundef float @
+// SPIR_NO_HALF: define spir_func noundef float @
+// NO_HALF: %hlsl.rcp = fdiv float 1.000000e+00, %{{.*}}
+// NO_HALF: ret float %hlsl.rcp
 half test_rcp_half(half p0) { return rcp(p0); }
-// NATIVE_HALF: define noundef <2 x half> @
-// NATIVE_HALF: %dx.rcp = call <2 x half> @llvm.dx.rcp.v2f16
-// NATIVE_HALF: ret <2 x half> %dx.rcp
-// NO_HALF: define noundef <2 x float> @
-// NO_HALF: %dx.rcp = call <2 x float> @llvm.dx.rcp.v2f32(
-// NO_HALF: ret <2 x float> %dx.rcp
+
+// DXIL_NATIVE_HALF: define noundef <2 x half> @
+// SPIR_NATIVE_HALF: define spir_func noundef <2 x half> @
+// NATIVE_HALF: %hlsl.rcp = fdiv <2 x half> <half  0xH3C00, half  0xH3C00>, %{{.*}} 
+// NATIVE_HALF: ret <2 x half> %hlsl.rcp
+// DXIL_NO_HALF: define noundef <2 x float> @
+// SPIR_NO_HALF: define spir_func noundef <2 x float> @
+// NO_HALF: %hlsl.rcp = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// NO_HALF: ret <2 x float> %hlsl.rcp
 half2 test_rcp_half2(half2 p0) { return rcp(p0); }
-// NATIVE_HALF: define noundef <3 x half> @
-// NATIVE_HALF: %dx.rcp = call <3 x half> @llvm.dx.rcp.v3f16
-// NATIVE_HALF: ret <3 x half> %dx.rcp
-// NO_HALF: define noundef <3 x float> @
-// NO_HALF: %dx.rcp = call <3 x float> @llvm.dx.rcp.v3f32(
-// NO_HALF: ret <3 x float> %dx.rcp
+
+// DXIL_NATIVE_HALF: define noundef <3 x half> @
+// SPIR_NATIVE_HALF: define spir_func noundef <3 x half> @
+// NATIVE_HALF: %hlsl.rcp = fdiv <3 x half> <half  0xH3C00, half  0xH3C00, half  0xH3C00>, %{{.*}} 
+// NATIVE_HALF: ret <3 x half> %hlsl.rcp
+// DXIL_NO_HALF: define noundef <3 x float> @
+// SPIR_NO_HALF: define spir_func noundef <3 x float> @
+// NO_HALF: %hlsl.rcp = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// NO_HALF: ret <3 x float> %hlsl.rcp
 half3 test_rcp_half3(half3 p0) { return rcp(p0); }
-// NATIVE_HALF: define noundef <4 x half> @
-// NATIVE_HALF: %dx.rcp = call <4 x half> @llvm.dx.rcp.v4f16
-// NATIVE_HALF: ret <4 x half> %dx.rcp
-// NO_HALF: define noundef <4 x float> @
-// NO_HALF: %dx.rcp = call <4 x float> @llvm.dx.rcp.v4f32(
-// NO_HALF: ret <4 x float> %dx.rcp
+
+// DXIL_NATIVE_HALF: define noundef <4 x half> @
+// SPIR_NATIVE_HALF: define spir_func noundef <4 x half> @
+// NATIVE_HALF: %hlsl.rcp = fdiv <4 x half> <half  0xH3C00, half  0xH3C00, half  0xH3C00, half  0xH3C00>, %{{.*}} 
+// NATIVE_HALF: ret <4 x half> %hlsl.rcp
+// DXIL_NO_HALF: define noundef <4 x float> @
+// SPIR_NO_HALF: define spir_func noundef <4 x float> @
+// NO_HALF: %hlsl.rcp = fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// NO_HALF: ret <4 x float> %hlsl.rcp
 half4 test_rcp_half4(half4 p0) { return rcp(p0); }
 
-// CHECK: define noundef float @
-// CHECK: %dx.rcp = call float @llvm.dx.rcp.f32(
-// CHECK: ret float %dx.rcp
+// DXIL_CHECK: define noundef float @
+// SPIR_CHECK: define spir_func noundef float @
+// CHECK: %hlsl.rcp = fdiv float 1.000000e+00, %{{.*}}
+// CHECK: ret float %hlsl.rcp
 float test_rcp_float(float p0) { return rcp(p0); }
-// CHECK: define noundef <2 x float> @
-// CHECK: %dx.rcp = call <2 x float> @llvm.dx.rcp.v2f32
-// CHECK: ret <2 x float> %dx.rcp
+
+// DXIL_CHECK: define noundef <2 x float> @
+// SPIR_CHECK: define spir_func noundef <2 x float> @
+// CHECK: %hlsl.rcp = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// CHECK: ret <2 x float> %hlsl.rcp
 float2 test_rcp_float2(float2 p0) { return rcp(p0); }
-// CHECK: define noundef <3 x float> @
-// CHECK: %dx.rcp = call <3 x float> @llvm.dx.rcp.v3f32
-// CHECK: ret <3 x float> %dx.rcp
+
+// DXIL_CHECK: define noundef <3 x float> @
+// SPIR_CHECK: define spir_func noundef <3 x float> @
+// CHECK: %hlsl.rcp = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// CHECK: ret <3 x float> %hlsl.rcp
 float3 test_rcp_float3(float3 p0) { return rcp(p0); }
-// CHECK: define noundef <4 x float> @
-// CHECK: %dx.rcp = call <4 x float> @llvm.dx.rcp.v4f32
-// CHECK: ret <4 x float> %dx.rcp
+
+// DXIL_CHECK: define noundef <4 x float> @
+// SPIR_CHECK: define spir_func noundef <4 x float> @
+// CHECK: %hlsl.rcp = fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %{{.*}}
+// CHECK: ret <4 x float> %hlsl.rcp
 float4 test_rcp_float4(float4 p0) { return rcp(p0); }
+
+// DXIL_CHECK: define noundef double @
+// SPIR_CHECK: define spir_func noundef double @
+// CHECK: %hlsl.rcp = fdiv double 1.000000e+00, %{{.*}} 
+// CHECK: ret double %hlsl.rcp
+double test_rcp_double(double p0) { return rcp(p0); }
+
+// DXIL_CHECK: define noundef <2 x double> @
+// SPIR_CHECK: define spir_func noundef <2 x double> @
+// CHECK: %hlsl.rcp = fdiv <2 x double> <double 1.000000e+00, double 1.000000e+00>, %{{.*}}
+// CHECK: ret <2 x double> %hlsl.rcp
+double2 test_rcp_double2(double2 p0) { return rcp(p0); }
+
+// DXIL_CHECK: define noundef <3 x double> @
+// SPIR_CHECK: define spir_func noundef <3 x double> @
+// CHECK: %hlsl.rcp = fdiv <3 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %{{.*}}
+// CHECK: ret <3 x double> %hlsl.rcp
+double3 test_rcp_double3(double3 p0) { return rcp(p0); }
+
+// DXIL_CHECK: define noundef <4 x double> @
+// SPIR_CHECK: define spir_func noundef <4 x double> @
+// CHECK: %hlsl.rcp = fdiv <4 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %{{.*}}
+// CHECK: ret <4 x double> %hlsl.rcp
+double4 test_rcp_double4(double4 p0) { return rcp(p0); }
diff --git a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
index 3e2d10f5ee7a2..4d99bc0069005 100644
--- a/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
+++ b/llvm/lib/Target/DirectX/DXILIntrinsicExpansion.cpp
@@ -42,7 +42,6 @@ static bool isIntrinsicExpansion(Function &F) {
   case Intrinsic::dx_clamp:
   case Intrinsic::dx_uclamp:
   case Intrinsic::dx_lerp:
-  case Intrinsic::dx_rcp:
   case Intrinsic::dx_sdot:
   case Intrinsic::dx_udot:
     return true;
@@ -218,25 +217,6 @@ static bool expandPowIntrinsic(CallInst *Orig) {
   return true;
 }
 
-static bool expandRcpIntrinsic(CallInst *Orig) {
-  Value *X = Orig->getOperand(0);
-  IRBuilder<> Builder(Orig->getParent());
-  Builder.SetInsertPoint(Orig);
-  Type *Ty = X->getType();
-  Type *EltTy = Ty->getScalarType();
-  Constant *One =
-      Ty->isVectorTy()
-          ? ConstantVector::getSplat(
-                ElementCount::getFixed(
-                    dyn_cast<FixedVectorType>(Ty)->getNumElements()),
-                ConstantFP::get(EltTy, 1.0))
-          : ConstantFP::get(EltTy, 1.0);
-  auto *Result = Builder.CreateFDiv(One, X, "dx.rcp");
-  Orig->replaceAllUsesWith(Result);
-  Orig->eraseFromParent();
-  return true;
-}
-
 static Intrinsic::ID getMaxForClamp(Type *ElemTy,
                                     Intrinsic::ID ClampIntrinsic) {
   if (ClampIntrinsic == Intrinsic::dx_uclamp)
@@ -300,8 +280,6 @@ static bool expandIntrinsic(Function &F, CallInst *Orig) {
     return expandClampIntrinsic(Orig, F.getIntrinsicID());
   case Intrinsic::dx_lerp:
     return expandLerpIntrinsic(Orig);
-  case Intrinsic::dx_rcp:
-    return expandRcpIntrinsic(Orig);
   case Intrinsic::dx_sdot:
   case Intrinsic::dx_udot:
     return expandIntegerDot(Orig, F.getIntrinsicID());
diff --git a/llvm/test/CodeGen/DirectX/rcp.ll b/llvm/test/CodeGen/DirectX/rcp.ll
deleted file mode 100644
index 65abe832db53f..0000000000000
--- a/llvm/test/CodeGen/DirectX/rcp.ll
+++ /dev/null
@@ -1,52 +0,0 @@
-; RUN: opt -S -dxil-op-lower < %s | FileCheck %s
-
-; Make sure dxil operation function calls for rcp are generated for float, double, and half.
-
-; CHECK-LABEL: rcp_float4
-; CHECK: fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %{{.*}}
-define noundef <4 x float> @rcp_float4(<4 x float> noundef %p0) {
-entry:
-  %p0.addr = alloca <4 x float>, align 16
-  store <4 x float> %p0, ptr %p0.addr, align 16
-  %0 = load <4 x float>, ptr %p0.addr, align 16
-  %dx.rcp = call <4 x float> @llvm.dx.rcp.v4f32(<4 x float> %0)
-  ret <4 x float> %dx.rcp
-}
-
-; CHECK-LABEL: rcp_double4
-; CHECK: fdiv <4 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %{{.*}}
-define noundef <4 x double> @rcp_double4(<4 x double> noundef %p0) {
-entry:
-  %p0.addr = alloca <4 x double>, align 16
-  store <4 x double> %p0, ptr %p0.addr, align 16
-  %0 = load <4 x double>, ptr %p0.addr, align 16
-  %dx.rcp = call <4 x double> @llvm.dx.rcp.v4f64(<4 x double> %0)
-  ret <4 x double> %dx.rcp
-}
-
-; CHECK-LABEL: rcp_half4
-; CHECK: fdiv <4 x half> <half  0xH3C00, half  0xH3C00, half  0xH3C00, half  0xH3C00>, %{{.*}} 
-define noundef <4 x half> @rcp_half4(<4 x half> noundef %p0) {
-entry:
-  %p0.addr = alloca <4 x half>, align 16
-  store <4 x half> %p0, ptr %p0.addr, align 16
-  %0 = load <4 x half>, ptr %p0.addr, align 16
-  %dx.rcp = call <4 x half> @llvm.dx.rcp.v4f16(<4 x half> %0)
-  ret <4 x half> %dx.rcp
-}
-
-; CHECK-LABEL: rcp_half
-; CHECK: fdiv half 0xH3C00, %{{.*}} 
-define noundef half @rcp_half(half noundef %p0) {
-entry:
-  %p0.addr = alloca half, align 2
-  store half %p0, ptr %p0.addr, align 2
-  %0 = load half, ptr %p0.addr, align 2
-  %dx.rcp = call half @llvm.dx.rcp.f16(half %0)
-  ret half %dx.rcp
-}
-
-declare half @llvm.dx.rcp.f16(half)
-declare <4 x half> @llvm.dx.rcp.v4f16(<4 x half>)
-declare <4 x float> @llvm.dx.rcp.v4f32(<4 x float>)
-declare <4 x double> @llvm.dx.rcp.v4f64(<4 x double>)
diff --git a/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/rcp.ll b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/rcp.ll
new file mode 100644
index 0000000000000..95962c0fdc969
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/rcp.ll
@@ -0,0 +1,126 @@
+ ; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s
+
+; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
+; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
+; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
+; CHECK-DAG: %[[#vec2_float_16:]] = OpTypeVector %[[#float_16]] 2
+; CHECK-DAG: %[[#vec2_float_32:]] = OpTypeVector %[[#float_32]] 2
+; CHECK-DAG: %[[#vec2_float_64:]] = OpTypeVector %[[#float_64]] 2
+; CHECK-DAG: %[[#vec3_float_16:]] = OpTypeVector %[[#float_16]] 3
+; CHECK-DAG: %[[#vec3_float_32:]] = OpTypeVector %[[#float_32]] 3
+; CHECK-DAG: %[[#vec3_float_64:]] = OpTypeVector %[[#float_64]] 3
+; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
+; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
+; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
+; CHECK-DAG: %[[#const_f64_1:]] = OpConstant %[[#float_64]] 1
+; CHECK-DAG: %[[#const_f32_1:]] = OpConstant %[[#float_32:]] 1
+; CHECK-DAG: %[[#const_f16_1:]] = OpConstant %[[#float_16:]] 1
+
+; CHECK-DAG: %[[#vec2_const_ones_f16:]] = OpConstantComposite %[[#vec2_float_16:]] %[[#const_f16_1:]] %[[#const_f16_1:]]
+; CHECK-DAG: %[[#vec3_const_ones_f16:]] = OpConstantComposite %[[#vec3_float_16:]] %[[#const_f16_1:]] %[[#const_f16_1:]] %[[#const_f16_1:]]
+; CHECK-DAG: %[[#vec4_const_ones_f16:]] = OpConstantComposite %[[#vec4_float_16:]] %[[#const_f16_1:]] %[[#const_f16_1:]] %[[#const_f16_1:]] %[[#const_f16_1:]]
+
+; CHECK-DAG: %[[#vec2_const_ones_f32:]] = OpConstantComposite %[[#vec2_float_32:]] %[[#const_f32_1:]] %[[#const_f32_1:]]
+; CHECK-DAG: %[[#vec3_const_ones_f32:]] = OpConstantComposite %[[#vec3_float_32:]] %[[#const_f32_1:]] %[[#const_f32_1:]] %[[#const_f32_1:]]
+; CHECK-DAG: %[[#vec4_const_ones_f32:]] = OpConstantComposite %[[#vec4_float_32:]] %[[#const_f32_1:]] %[[#const_f32_1:]] %[[#const_f32_1:]] %[[#const_f32_1:]]
+
+; CHECK-DAG: %[[#vec2_const_ones_f64:]] = OpConstantComposite %[[#vec2_float_64:]] %[[#const_f64_1:]] %[[#const_f64_1:]]
+; CHECK-DAG: %[[#vec3_const_ones_f64:]] = OpConstantComposite %[[#vec3_float_64:]] %[[#const_f64_1:]] %[[#const_f64_1:]] %[[#const_f64_1:]]
+; CHECK-DAG: %[[#vec4_const_ones_f64:]] = OpConstantComposite %[[#vec4_float_64:]] %[[#const_f64_1:]] %[[#const_f64_1:]] %[[#const_f64_1:]] %[[#const_f64_1:]]
+
+
+define spir_func noundef half @test_rcp_half(half noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_16:]]
+  ; CHECK: OpFDiv %[[#float_16:]] %[[#const_f16_1:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv half 0xH3C00, %p0
+  ret half %hlsl.rcp
+}
+
+define spir_func noundef <2 x half> @test_rcp_half2(<2 x half> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_16:]]
+  ; CHECK: OpFDiv %[[#vec2_float_16:]] %[[#vec2_const_ones_f16:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <2 x half> <half 0xH3C00, half 0xH3C00>, %p0
+  ret <2 x half> %hlsl.rcp
+}
+
+define spir_func noundef <3 x half> @test_rcp_half3(<3 x half> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_16:]]
+  ; CHECK: OpFDiv %[[#vec3_float_16:]] %[[#vec3_const_ones_f16:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <3 x half> <half 0xH3C00, half 0xH3C00, half 0xH3C00>, %p0
+  ret <3 x half> %hlsl.rcp
+}
+
+define spir_func noundef <4 x half> @test_rcp_half4(<4 x half> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_16:]]
+  ; CHECK: OpFDiv %[[#vec4_float_16:]] %[[#vec4_const_ones_f16:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <4 x half> <half 0xH3C00, half 0xH3C00, half 0xH3C00, half 0xH3C00>, %p0
+  ret <4 x half> %hlsl.rcp
+}
+
+define spir_func noundef float @test_rcp_float(float noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_32:]]
+  ; CHECK: OpFDiv %[[#float_32:]] %[[#const_f32_1:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv float 1.000000e+00, %p0
+  ret float %hlsl.rcp
+}
+
+define spir_func noundef <2 x float> @test_rcp_float2(<2 x float> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_32:]]
+  ; CHECK: OpFDiv %[[#vec2_float_32:]] %[[#vec2_const_ones_f32:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, %p0
+  ret <2 x float> %hlsl.rcp
+}
+
+define spir_func noundef <3 x float> @test_rcp_float3(<3 x float> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_32:]]
+  ; CHECK: OpFDiv %[[#vec3_float_32:]] %[[#vec3_const_ones_f32:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <3 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %p0
+  ret <3 x float> %hlsl.rcp
+}
+
+define spir_func noundef <4 x float> @test_rcp_float4(<4 x float> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_32:]]
+  ; CHECK: OpFDiv %[[#vec4_float_32:]] %[[#vec4_const_ones_f32:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %p0
+  ret <4 x float> %hlsl.rcp
+}
+
+define spir_func noundef double @test_rcp_double(double noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#float_64:]]
+  ; CHECK: OpFDiv %[[#float_64:]] %[[#const_f64_1:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv double 1.000000e+00, %p0
+  ret double %hlsl.rcp
+}
+
+define spir_func noundef <2 x double> @test_rcp_double2(<2 x double> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec2_float_64:]]
+  ; CHECK: OpFDiv %[[#vec2_float_64:]] %[[#vec2_const_ones_f64:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <2 x double> <double 1.000000e+00, double 1.000000e+00>, %p0
+  ret <2 x double> %hlsl.rcp
+}
+
+define spir_func noundef <3 x double> @test_rcp_double3(<3 x double> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec3_float_64:]]
+  ; CHECK: OpFDiv %[[#vec3_float_64:]] %[[#vec3_const_ones_f64:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <3 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %p0
+  ret <3 x double> %hlsl.rcp
+}
+
+define spir_func noundef <4 x double> @test_rcp_double4(<4 x double> noundef %p0) #0 {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#vec4_float_64:]]
+  ; CHECK: OpFDiv %[[#vec4_float_64:]] %[[#vec4_const_ones_f64:]] %[[#arg0:]]
+  %hlsl.rcp = fdiv <4 x double> <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %p0
+  ret <4 x double> %hlsl.rcp
+}

From 00a4f091bad199053b33af983a5abbf2c5abe949 Mon Sep 17 00:00:00 2001
From: Cyndy Ishida <cyndy_ishida@apple.com>
Date: Thu, 11 Apr 2024 15:51:27 -0700
Subject: [PATCH 198/886] [clang][docs] Modernize attribute docs for darwin
 specifics (#88448)

* Generally recommend target triples. But replace `m*version-min` with
`mtargetos`.
* Also include test coverage for -mtargetos=visionos
---
 clang/include/clang/Basic/AttrDocs.td | 37 ++++++++++++++++++---------
 clang/test/Driver/mtargetos-darwin.c  |  2 ++
 2 files changed, 27 insertions(+), 12 deletions(-)

diff --git a/clang/include/clang/Basic/AttrDocs.td b/clang/include/clang/Basic/AttrDocs.td
index 0ca4ea377fc36..8687c4f57d3f8 100644
--- a/clang/include/clang/Basic/AttrDocs.td
+++ b/clang/include/clang/Basic/AttrDocs.td
@@ -1604,27 +1604,40 @@ specifies availability for the current target platform, the availability
 attributes are ignored. Supported platforms are:
 
 ``ios``
-  Apple's iOS operating system. The minimum deployment target is specified by
-  the ``-mios-version-min=*version*`` or ``-miphoneos-version-min=*version*``
-  command-line arguments.
+  Apple's iOS operating system. The minimum deployment target is specified 
+  as part of the ``-target *arch*-apple-ios*version*`` command line argument. 
+  Alternatively, it can be specified by the ``-mtargetos=ios*version*`` 
+  command-line argument.
 
 ``macos``
-  Apple's macOS operating system. The minimum deployment target is
-  specified by the ``-mmacosx-version-min=*version*`` command-line argument.
-  ``macosx`` is supported for backward-compatibility reasons, but it is
-  deprecated.
+  Apple's macOS operating system. The minimum deployment target is specified 
+  as part of the ``-target *arch*-apple-macos*version*`` command line argument. 
+  Alternatively, it can be specified by the ``-mtargetos=macos*version*`` 
+  command-line argument. ``macosx`` is supported for 
+  backward-compatibility reasons, but it is deprecated.
 
 ``tvos``
-  Apple's tvOS operating system. The minimum deployment target is specified by
-  the ``-mtvos-version-min=*version*`` command-line argument.
+  Apple's tvOS operating system. The minimum deployment target is specified 
+  as part of the ``-target *arch*-apple-tvos*version*`` command line argument. 
+  Alternatively, it can be specified by the ``-mtargetos=tvos*version*`` 
+  command-line argument.
 
 ``watchos``
-  Apple's watchOS operating system. The minimum deployment target is specified by
-  the ``-mwatchos-version-min=*version*`` command-line argument.
+  Apple's watchOS operating system. The minimum deployment target is specified
+  as part of the ``-target *arch*-apple-watchos*version*`` command line argument. 
+  Alternatively, it can be specified by the ``-mtargetos=watchos*version*`` 
+  command-line argument.
+
+``visionos``
+  Apple's visionOS operating system. The minimum deployment target is specified
+  as part of the ``-target *arch*-apple-visionos*version*`` command line argument. 
+  Alternatively, it can be specified by the ``-mtargetos=visionos*version*`` 
+  command-line argument.
 
 ``driverkit``
   Apple's DriverKit userspace kernel extensions. The minimum deployment target
-  is specified as part of the triple.
+  is specified as part of the ``-target *arch*-apple-driverkit*version*`` 
+  command line argument.
 
 A declaration can typically be used even when deploying back to a platform
 version prior to when the declaration was introduced. When this happens, the
diff --git a/clang/test/Driver/mtargetos-darwin.c b/clang/test/Driver/mtargetos-darwin.c
index e706be37a371f..7e86ab15279b9 100644
--- a/clang/test/Driver/mtargetos-darwin.c
+++ b/clang/test/Driver/mtargetos-darwin.c
@@ -4,6 +4,7 @@
 // RUN: %clang -mtargetos=ios14-macabi -arch arm64 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=MACCATALYST %s
 // RUN: %clang -mtargetos=tvos14 -arch arm64 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=TVOS %s
 // RUN: %clang -mtargetos=watchos7.1 -arch arm64 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=WATCHOS %s
+// RUN: %clang -mtargetos=visionos1 -arch arm64 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=VISIONOS %s
 
 // RUN: not %clang -target arm64-apple-ios14 -mtargetos=ios14 -arch arm64 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=NOMIX1 %s
 // RUN: not %clang -mtargetos=ios14 -arch arm64 -miphoneos-version-min=14 -c %s -o %t.o -### 2>&1 | FileCheck --check-prefix=NOMIX2 %s
@@ -19,6 +20,7 @@
 // MACCATALYST: "-cc1" "-triple" "arm64-apple-ios14.0.0-macabi"
 // TVOS: "-cc1" "-triple" "arm64-apple-tvos14.0.0"
 // WATCHOS: "-cc1" "-triple" "arm64-apple-watchos7.1.0"
+// VISIONOS: "-cc1" "-triple" "arm64-apple-xros1.0.0"
 
 // NOMIX1: error: cannot specify '-mtargetos=ios14' along with '-target arm64-apple-ios14'
 // NOMIX2: error: cannot specify '-miphoneos-version-min=14' along with '-mtargetos=ios14'

From 43d0891d3bb1fc40ff5dcea91c28d1582978caff Mon Sep 17 00:00:00 2001
From: Maksim Panchenko <maks@fb.com>
Date: Thu, 11 Apr 2024 16:11:00 -0700
Subject: [PATCH 199/886] [BOLT] Fix handling of trailing entries in jump
 tables (#88444)

If a jump table has entries at the end that are a result of
__builtin_unreachable() targets, BOLT can confuse them with function
pointers. In such case, we should exclude these targets from the table
as we risk incorrectly updating the function pointers. It is safe to
exclude them as branching on such targets is considered an undefined
behavior.
---
 bolt/lib/Core/BinaryContext.cpp      |  26 ++++-
 bolt/test/runtime/X86/jt-confusion.s | 164 +++++++++++++++++++++++++++
 2 files changed, 185 insertions(+), 5 deletions(-)
 create mode 100644 bolt/test/runtime/X86/jt-confusion.s

diff --git a/bolt/lib/Core/BinaryContext.cpp b/bolt/lib/Core/BinaryContext.cpp
index 47eae964e816c..7c2d8c52287be 100644
--- a/bolt/lib/Core/BinaryContext.cpp
+++ b/bolt/lib/Core/BinaryContext.cpp
@@ -555,6 +555,9 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
                                      const uint64_t NextJTAddress,
                                      JumpTable::AddressesType *EntriesAsAddress,
                                      bool *HasEntryInFragment) const {
+  // Target address of __builtin_unreachable.
+  const uint64_t UnreachableAddress = BF.getAddress() + BF.getSize();
+
   // Is one of the targets __builtin_unreachable?
   bool HasUnreachable = false;
 
@@ -564,9 +567,15 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
   // Number of targets other than __builtin_unreachable.
   uint64_t NumRealEntries = 0;
 
-  auto addEntryAddress = [&](uint64_t EntryAddress) {
-    if (EntriesAsAddress)
-      EntriesAsAddress->emplace_back(EntryAddress);
+  // Size of the jump table without trailing __builtin_unreachable entries.
+  size_t TrimmedSize = 0;
+
+  auto addEntryAddress = [&](uint64_t EntryAddress, bool Unreachable = false) {
+    if (!EntriesAsAddress)
+      return;
+    EntriesAsAddress->emplace_back(EntryAddress);
+    if (!Unreachable)
+      TrimmedSize = EntriesAsAddress->size();
   };
 
   ErrorOr<const BinarySection &> Section = getSectionForAddress(Address);
@@ -618,8 +627,8 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
             : *getPointerAtAddress(EntryAddress);
 
     // __builtin_unreachable() case.
-    if (Value == BF.getAddress() + BF.getSize()) {
-      addEntryAddress(Value);
+    if (Value == UnreachableAddress) {
+      addEntryAddress(Value, /*Unreachable*/ true);
       HasUnreachable = true;
       LLVM_DEBUG(dbgs() << formatv("OK: {0:x} __builtin_unreachable\n", Value));
       continue;
@@ -673,6 +682,13 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
     addEntryAddress(Value);
   }
 
+  // Trim direct/normal jump table to exclude trailing unreachable entries that
+  // can collide with a function address.
+  if (Type == JumpTable::JTT_NORMAL && EntriesAsAddress &&
+      TrimmedSize != EntriesAsAddress->size() &&
+      getBinaryFunctionAtAddress(UnreachableAddress))
+    EntriesAsAddress->resize(TrimmedSize);
+
   // It's a jump table if the number of real entries is more than 1, or there's
   // one real entry and one or more special targets. If there are only multiple
   // special targets, then it's not a jump table.
diff --git a/bolt/test/runtime/X86/jt-confusion.s b/bolt/test/runtime/X86/jt-confusion.s
new file mode 100644
index 0000000000000..f15c83b35b6a4
--- /dev/null
+++ b/bolt/test/runtime/X86/jt-confusion.s
@@ -0,0 +1,164 @@
+# REQUIRES: system-linux
+
+# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
+# RUN: llvm-strip --strip-unneeded %t.o
+# RUN: %clang %cflags -no-pie -nostartfiles -nostdlib -lc %t.o -o %t.exe -Wl,-q
+
+# RUN: llvm-bolt %t.exe -o %t.exe.bolt --relocs=1 --lite=0
+
+# RUN: %t.exe.bolt
+
+## Check that BOLT's jump table detection diffrentiates between
+## __builtin_unreachable() targets and function pointers.
+
+## The test case was built from the following two source files and
+## modiffied for standalone build. main became _start, etc.
+## $ $(CC) a.c -O1 -S -o a.s
+## $ $(CC) b.c -O0 -S -o b.s
+
+## a.c:
+
+## typedef int (*fptr)(int);
+## void check_fptr(fptr, int);
+##
+## int foo(int a) {
+##   check_fptr(foo, 0);
+##   switch (a) {
+##   default:
+##     __builtin_unreachable();
+##   case 0:
+##     return 3;
+##   case 1:
+##     return 5;
+##   case 2:
+##     return 7;
+##   case 3:
+##     return 11;
+##   case 4:
+##     return 13;
+##   case 5:
+##     return 17;
+##   }
+##   return 0;
+## }
+##
+## int main(int argc) {
+##   check_fptr(main, 1);
+##   return foo(argc);
+## }
+##
+## const fptr funcs[2] = {foo, main};
+
+## b.c.:
+
+## typedef int (*fptr)(int);
+## extern const fptr funcs[2];
+##
+## #define assert(C) { if (!(C)) (*(unsigned long long *)0) = 0; }
+## void check_fptr(fptr f, int i) {
+##   assert(f == funcs[i]);
+## }
+
+
+	.text
+	.globl	foo
+	.type	foo, @function
+foo:
+.LFB0:
+	.cfi_startproc
+	pushq	%rbx
+	.cfi_def_cfa_offset 16
+	.cfi_offset 3, -16
+	movl	%edi, %ebx
+	movl	$0, %esi
+	movl	$foo, %edi
+	call	check_fptr
+	movl	%ebx, %ebx
+	jmp	*.L4(,%rbx,8)
+.L8:
+	movl	$5, %eax
+	jmp	.L1
+.L7:
+	movl	$7, %eax
+	jmp	.L1
+.L6:
+	movl	$11, %eax
+	jmp	.L1
+.L5:
+	movl	$13, %eax
+	jmp	.L1
+.L3:
+	movl	$17, %eax
+	jmp	.L1
+.L10:
+	movl	$3, %eax
+.L1:
+	popq	%rbx
+	.cfi_def_cfa_offset 8
+	ret
+	.cfi_endproc
+.LFE0:
+	.size	foo, .-foo
+	.globl	_start
+	.type	_start, @function
+_start:
+.LFB1:
+	.cfi_startproc
+	pushq	%rbx
+	.cfi_def_cfa_offset 16
+	.cfi_offset 3, -16
+	movl	%edi, %ebx
+	movl	$1, %esi
+	movl	$_start, %edi
+	call	check_fptr
+	movl	$1, %edi
+	call	foo
+	popq	%rbx
+	.cfi_def_cfa_offset 8
+  callq exit@PLT
+	.cfi_endproc
+.LFE1:
+	.size	_start, .-_start
+	.globl	check_fptr
+	.type	check_fptr, @function
+check_fptr:
+.LFB2:
+	.cfi_startproc
+	pushq	%rbp
+	.cfi_def_cfa_offset 16
+	.cfi_offset 6, -16
+	movq	%rsp, %rbp
+	.cfi_def_cfa_register 6
+	movq	%rdi, -8(%rbp)
+	movl	%esi, -12(%rbp)
+	movl	-12(%rbp), %eax
+	cltq
+	movq	funcs(,%rax,8), %rax
+	cmpq	%rax, -8(%rbp)
+	je	.L33
+	movl	$0, %eax
+	movq	$0, (%rax)
+.L33:
+	nop
+	popq	%rbp
+	.cfi_def_cfa 7, 8
+	ret
+	.cfi_endproc
+
+	.section	.rodata
+	.align 8
+	.align 4
+.L4:
+	.quad	.L10
+	.quad	.L8
+	.quad	.L7
+	.quad	.L6
+	.quad	.L5
+	.quad	.L3
+
+	.globl	funcs
+	.type	funcs, @object
+	.size	funcs, 16
+funcs:
+	.quad	foo
+	.quad	_start

From 2f55056c89dd1ccb1aa69a9aed68048a59413d4d Mon Sep 17 00:00:00 2001
From: Mark Rowe <mrowe@bdash.net.nz>
Date: Thu, 11 Apr 2024 16:13:54 -0700
Subject: [PATCH 200/886] [compiler-rt] Don't explicitly ad-hoc code sign
 dylibs if using Apple's new linker (#88323)

Apple's new linker reports itself as ld rather than ld64 and does not
match the version detection regex.

Invert the logic to look only for older versions of ld64. This ensures
the runtime dylibs are left with a linker-generated code signature that
tools such as `strip` will preserve.

Co-authored-by: Mark Rowe <markrowe@chromium.org>
---
 compiler-rt/cmake/Modules/AddCompilerRT.cmake | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/compiler-rt/cmake/Modules/AddCompilerRT.cmake b/compiler-rt/cmake/Modules/AddCompilerRT.cmake
index e0400a8ea9522..dc36f16c661c2 100644
--- a/compiler-rt/cmake/Modules/AddCompilerRT.cmake
+++ b/compiler-rt/cmake/Modules/AddCompilerRT.cmake
@@ -405,11 +405,14 @@ function(add_compiler_rt_runtime name type)
         if (HAD_ERROR)
           message(FATAL_ERROR "${CMAKE_LINKER} failed with status ${HAD_ERROR}")
         endif()
-        set(NEED_EXPLICIT_ADHOC_CODESIGN 1)
+        set(NEED_EXPLICIT_ADHOC_CODESIGN 0)
+        # Apple introduced a new linker by default in Xcode 15. This linker reports itself as ld
+        # rather than ld64 and does not match this version regex. That's ok since it never needs
+        # the explicit ad-hoc code signature.
         if ("${LD_V_OUTPUT}" MATCHES ".*ld64-([0-9.]+).*")
           string(REGEX REPLACE ".*ld64-([0-9.]+).*" "\\1" HOST_LINK_VERSION ${LD_V_OUTPUT})
-          if (HOST_LINK_VERSION VERSION_GREATER_EQUAL 609)
-            set(NEED_EXPLICIT_ADHOC_CODESIGN 0)
+          if (HOST_LINK_VERSION VERSION_LESS 609)
+            set(NEED_EXPLICIT_ADHOC_CODESIGN 1)
           endif()
         endif()
         if (NEED_EXPLICIT_ADHOC_CODESIGN)

From 5fb821560afd0543e4875c3c81d58f68f73eb03b Mon Sep 17 00:00:00 2001
From: Michael Jones <michaelrj@google.com>
Date: Thu, 11 Apr 2024 16:26:49 -0700
Subject: [PATCH 201/886] [libc] add remaining epoll functions, pipe (#84587)

The epoll_wait functions need the rest of the epoll functions (create,
ctl) to be available to actually test them, as well as pipe to create a
usable file descriptor. This patch adds epoll_create, epoll_create1,
epoll_ctl, and pipe. These have tests, and the tests for epoll_wait,
epoll_pwait, and epoll_pwait2 (currently disabled) are updated to use
these newly available functions.
---
 libc/config/linux/x86_64/entrypoints.txt      |   4 +
 libc/hdr/CMakeLists.txt                       |  18 +++
 libc/hdr/signal_macros.h                      |  22 ++++
 libc/hdr/sys_epoll_macros.h                   |  22 ++++
 libc/include/CMakeLists.txt                   |   1 +
 libc/include/llvm-libc-macros/CMakeLists.txt  |   6 +
 .../llvm-libc-macros/linux/CMakeLists.txt     |   6 +
 .../llvm-libc-macros/linux/sys-epoll-macros.h |  40 +++++++
 .../llvm-libc-macros/sys-epoll-macros.h       |  16 +++
 libc/include/llvm-libc-types/sigset_t.h       |   4 +-
 .../llvm-libc-types/struct_epoll_event.h      |   6 +-
 libc/include/sys/epoll.h.def                  |   2 +
 libc/spec/posix.td                            |   5 +
 libc/src/sys/epoll/CMakeLists.txt             |  21 ++++
 libc/src/sys/epoll/epoll_create.h             |  18 +++
 libc/src/sys/epoll/epoll_create1.h            |  18 +++
 libc/src/sys/epoll/epoll_ctl.h                |  21 ++++
 libc/src/sys/epoll/epoll_pwait.h              |   4 +-
 libc/src/sys/epoll/linux/CMakeLists.txt       |  45 +++++++-
 libc/src/sys/epoll/linux/epoll_create.cpp     |  38 +++++++
 libc/src/sys/epoll/linux/epoll_create1.cpp    |  31 ++++++
 libc/src/sys/epoll/linux/epoll_ctl.cpp        |  34 ++++++
 libc/src/sys/epoll/linux/epoll_pwait.cpp      |   5 +-
 libc/src/sys/epoll/linux/epoll_pwait2.cpp     |   5 +-
 libc/src/sys/epoll/linux/epoll_wait.cpp       |   5 +-
 libc/src/unistd/CMakeLists.txt                |   7 ++
 libc/src/unistd/linux/CMakeLists.txt          |  13 +++
 libc/src/unistd/linux/pipe.cpp                |  33 ++++++
 libc/src/unistd/pipe.h                        |  18 +++
 libc/test/src/sys/epoll/linux/CMakeLists.txt  |  68 +++++++++++-
 .../sys/epoll/linux/epoll_create1_test.cpp    |  31 ++++++
 .../src/sys/epoll/linux/epoll_create_test.cpp |  26 +++++
 .../src/sys/epoll/linux/epoll_ctl_test.cpp    |  47 ++++++++
 .../src/sys/epoll/linux/epoll_pwait2_test.cpp |  47 +++++++-
 .../src/sys/epoll/linux/epoll_pwait_test.cpp  |  42 ++++++-
 .../src/sys/epoll/linux/epoll_wait_test.cpp   |  37 ++++++-
 libc/test/src/unistd/CMakeLists.txt           |  15 +++
 libc/test/src/unistd/pipe_test.cpp            |  26 +++++
 .../llvm-project-overlay/libc/BUILD.bazel     | 103 ++++++++++++++++--
 .../libc/test/src/sys/epoll/BUILD.bazel       |  61 +++++++++++
 40 files changed, 933 insertions(+), 38 deletions(-)
 create mode 100644 libc/hdr/signal_macros.h
 create mode 100644 libc/hdr/sys_epoll_macros.h
 create mode 100644 libc/include/llvm-libc-macros/linux/sys-epoll-macros.h
 create mode 100644 libc/include/llvm-libc-macros/sys-epoll-macros.h
 create mode 100644 libc/src/sys/epoll/epoll_create.h
 create mode 100644 libc/src/sys/epoll/epoll_create1.h
 create mode 100644 libc/src/sys/epoll/epoll_ctl.h
 create mode 100644 libc/src/sys/epoll/linux/epoll_create.cpp
 create mode 100644 libc/src/sys/epoll/linux/epoll_create1.cpp
 create mode 100644 libc/src/sys/epoll/linux/epoll_ctl.cpp
 create mode 100644 libc/src/unistd/linux/pipe.cpp
 create mode 100644 libc/src/unistd/pipe.h
 create mode 100644 libc/test/src/sys/epoll/linux/epoll_create1_test.cpp
 create mode 100644 libc/test/src/sys/epoll/linux/epoll_create_test.cpp
 create mode 100644 libc/test/src/sys/epoll/linux/epoll_ctl_test.cpp
 create mode 100644 libc/test/src/unistd/pipe_test.cpp

diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index 6bb53cb76220f..8fdd4575e27e2 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -215,6 +215,9 @@ set(TARGET_LIBC_ENTRYPOINTS
     libc.src.stdio.fileno
 
     # sys/epoll.h entrypoints
+    libc.src.sys.epoll.epoll_create
+    libc.src.sys.epoll.epoll_create1
+    libc.src.sys.epoll.epoll_ctl
     libc.src.sys.epoll.epoll_wait
     libc.src.sys.epoll.epoll_pwait
     # TODO: Need to check if pwait2 is available before providing.
@@ -308,6 +311,7 @@ set(TARGET_LIBC_ENTRYPOINTS
     libc.src.unistd.link
     libc.src.unistd.linkat
     libc.src.unistd.lseek
+    libc.src.unistd.pipe
     libc.src.unistd.pread
     libc.src.unistd.pwrite
     libc.src.unistd.read
diff --git a/libc/hdr/CMakeLists.txt b/libc/hdr/CMakeLists.txt
index 38ef56e3f04c0..fb7c342f92b78 100644
--- a/libc/hdr/CMakeLists.txt
+++ b/libc/hdr/CMakeLists.txt
@@ -41,4 +41,22 @@ add_proxy_header_library(
     libc.include.fenv
 )
 
+add_proxy_header_library(
+  signal_macros
+  HDRS
+    signal_macros.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-macros.signal_macros
+    libc.include.signal
+)
+
+add_proxy_header_library(
+  sys_epoll_macros
+  HDRS
+    sys_epoll_macros.h
+  FULL_BUILD_DEPENDS
+    libc.include.sys_epoll
+    libc.include.llvm-libc-macros.sys_epoll_macros
+)
+
 add_subdirectory(types)
diff --git a/libc/hdr/signal_macros.h b/libc/hdr/signal_macros.h
new file mode 100644
index 0000000000000..867d17a4ca8c5
--- /dev/null
+++ b/libc/hdr/signal_macros.h
@@ -0,0 +1,22 @@
+//===-- Definition of macros from signal.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_HDR_SIGNAL_MACROS_H
+#define LLVM_LIBC_HDR_SIGNAL_MACROS_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-macros/signal-macros.h"
+
+#else // Overlay mode
+
+#include <signal.h>
+
+#endif // LLVM_LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_SIGNAL_MACROS_H
diff --git a/libc/hdr/sys_epoll_macros.h b/libc/hdr/sys_epoll_macros.h
new file mode 100644
index 0000000000000..db047f1616343
--- /dev/null
+++ b/libc/hdr/sys_epoll_macros.h
@@ -0,0 +1,22 @@
+//===-- Definition of macros from sys/epoll.h -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_HDR_SYS_EPOLL_MACROS_H
+#define LLVM_LIBC_HDR_SYS_EPOLL_MACROS_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-macros/sys-epoll-macros.h"
+
+#else // Overlay mode
+
+#include <sys/epoll.h>
+
+#endif // LLVM_LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_SYS/EPOLL_MACROS_H
diff --git a/libc/include/CMakeLists.txt b/libc/include/CMakeLists.txt
index 02c7dc8fbc0b3..b85366c8deafe 100644
--- a/libc/include/CMakeLists.txt
+++ b/libc/include/CMakeLists.txt
@@ -379,6 +379,7 @@ add_gen_header(
     .llvm-libc-types.struct_epoll_event
     .llvm-libc-types.struct_epoll_data
     .llvm-libc-types.sigset_t
+    .llvm-libc-macros.sys_epoll_macros
 )
 
 add_gen_header(
diff --git a/libc/include/llvm-libc-macros/CMakeLists.txt b/libc/include/llvm-libc-macros/CMakeLists.txt
index 635ccadfb49e7..382cb8ee417e1 100644
--- a/libc/include/llvm-libc-macros/CMakeLists.txt
+++ b/libc/include/llvm-libc-macros/CMakeLists.txt
@@ -143,6 +143,12 @@ add_macro_header(
     sys-auxv-macros.h
 )
 
+add_macro_header(
+  sys_epoll_macros
+  HDR
+    sys-epoll-macros.h
+)
+
 add_macro_header(
   sys_ioctl_macros
   HDR
diff --git a/libc/include/llvm-libc-macros/linux/CMakeLists.txt b/libc/include/llvm-libc-macros/linux/CMakeLists.txt
index d934c3a0fca6f..4ee429d1db166 100644
--- a/libc/include/llvm-libc-macros/linux/CMakeLists.txt
+++ b/libc/include/llvm-libc-macros/linux/CMakeLists.txt
@@ -10,6 +10,12 @@ add_header(
     sched-macros.h
 )
 
+add_header(
+  sys_epoll_macros
+  HDR
+    sys-epoll-macros.h
+)
+
 add_header(
   sys_ioctl_macros
   HDR
diff --git a/libc/include/llvm-libc-macros/linux/sys-epoll-macros.h b/libc/include/llvm-libc-macros/linux/sys-epoll-macros.h
new file mode 100644
index 0000000000000..f73d690908fd0
--- /dev/null
+++ b/libc/include/llvm-libc-macros/linux/sys-epoll-macros.h
@@ -0,0 +1,40 @@
+//===-- Macros defined in sys/epoll.h header file -------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_EPOLL_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_EPOLL_MACROS_H
+
+#include "fcntl-macros.h"
+
+// These are also defined in <linux/eventpoll.h> but that also contains a
+// different definition of the epoll_event struct that is different from the
+// userspace version.
+
+#define EPOLL_CLOEXEC O_CLOEXEC
+
+#define EPOLL_CTL_ADD 1
+#define EPOLL_CTL_DEL 2
+#define EPOLL_CTL_MOD 3
+
+#define EPOLLIN 0x1
+#define EPOLLPRI 0x2
+#define EPOLLOUT 0x4
+#define EPOLLERR 0x8
+#define EPOLLHUP 0x10
+#define EPOLLRDNORM 0x40
+#define EPOLLRDBAND 0x80
+#define EPOLLWRNORM 0x100
+#define EPOLLWRBAND 0x200
+#define EPOLLMSG 0x400
+#define EPOLLRDHUP 0x2000
+#define EPOLLEXCLUSIVE 0x10000000
+#define EPOLLWAKEUP 0x20000000
+#define EPOLLONESHOT 0x40000000
+#define EPOLLET 0x80000000
+
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_EPOLL_MACROS_H
diff --git a/libc/include/llvm-libc-macros/sys-epoll-macros.h b/libc/include/llvm-libc-macros/sys-epoll-macros.h
new file mode 100644
index 0000000000000..8212df252b8e7
--- /dev/null
+++ b/libc/include/llvm-libc-macros/sys-epoll-macros.h
@@ -0,0 +1,16 @@
+//===-- Macros defined in sys/epoll.h header file -------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_MACROS_SYS_EPOLL_MACROS_H
+#define LLVM_LIBC_MACROS_SYS_EPOLL_MACROS_H
+
+#ifdef __linux__
+#include "linux/sys-epoll-macros.h"
+#endif
+
+#endif // LLVM_LIBC_MACROS_SYS_EPOLL_MACROS_H
diff --git a/libc/include/llvm-libc-types/sigset_t.h b/libc/include/llvm-libc-types/sigset_t.h
index 311a92b823ff5..5c3e4e6f0431b 100644
--- a/libc/include/llvm-libc-types/sigset_t.h
+++ b/libc/include/llvm-libc-types/sigset_t.h
@@ -13,8 +13,8 @@
 
 // This definition can be adjusted/specialized for different targets and
 // platforms as necessary. This definition works for Linux on most targets.
-typedef struct {
+struct sigset_t {
   unsigned long __signals[__NSIGSET_WORDS];
-} sigset_t;
+};
 
 #endif // LLVM_LIBC_TYPES_SIGSET_T_H
diff --git a/libc/include/llvm-libc-types/struct_epoll_event.h b/libc/include/llvm-libc-types/struct_epoll_event.h
index 66cf86c1e2a00..b7685df20bd7e 100644
--- a/libc/include/llvm-libc-types/struct_epoll_event.h
+++ b/libc/include/llvm-libc-types/struct_epoll_event.h
@@ -11,7 +11,11 @@
 
 #include "llvm-libc-types/struct_epoll_data.h"
 
-typedef struct epoll_event {
+typedef struct
+#ifdef __x86_64__
+    [[gnu::packed]] // Necessary for compatibility.
+#endif
+    epoll_event {
   __UINT32_TYPE__ events;
   epoll_data_t data;
 } epoll_event;
diff --git a/libc/include/sys/epoll.h.def b/libc/include/sys/epoll.h.def
index 85f7d9ad6091f..98d96bf429251 100644
--- a/libc/include/sys/epoll.h.def
+++ b/libc/include/sys/epoll.h.def
@@ -11,6 +11,8 @@
 
 #include "__llvm-libc-common.h"
 
+#include "llvm-libc-macros/sys-epoll-macros.h"
+
 %%public_api()
 
 #endif // LLVM_LIBC_SYS_EPOLL_H
diff --git a/libc/spec/posix.td b/libc/spec/posix.td
index 45f7ecfe84e98..7095a3964ee3f 100644
--- a/libc/spec/posix.td
+++ b/libc/spec/posix.td
@@ -572,6 +572,11 @@ def POSIX : StandardSpec<"POSIX"> {
           RetValSpec<OffTType>,
           [ArgSpec<IntType>, ArgSpec<OffTType>, ArgSpec<IntType>]
         >,
+        FunctionSpec<
+          "pipe",
+          RetValSpec<IntType>,
+          [ArgSpec<IntPtr>] //TODO: make this int[2]
+        >,
         FunctionSpec<
           "pread",
           RetValSpec<SSizeTType>,
diff --git a/libc/src/sys/epoll/CMakeLists.txt b/libc/src/sys/epoll/CMakeLists.txt
index d4991a238e2a7..e785f535bf0cf 100644
--- a/libc/src/sys/epoll/CMakeLists.txt
+++ b/libc/src/sys/epoll/CMakeLists.txt
@@ -2,6 +2,27 @@ if(EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/${LIBC_TARGET_OS})
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/${LIBC_TARGET_OS})
 endif()
 
+add_entrypoint_object(
+  epoll_create
+  ALIAS
+  DEPENDS
+    .${LIBC_TARGET_OS}.epoll_create
+)
+
+add_entrypoint_object(
+  epoll_create1
+  ALIAS
+  DEPENDS
+    .${LIBC_TARGET_OS}.epoll_create1
+)
+
+add_entrypoint_object(
+  epoll_ctl
+  ALIAS
+  DEPENDS
+    .${LIBC_TARGET_OS}.epoll_ctl
+)
+
 add_entrypoint_object(
   epoll_wait
   ALIAS
diff --git a/libc/src/sys/epoll/epoll_create.h b/libc/src/sys/epoll/epoll_create.h
new file mode 100644
index 0000000000000..a1eeabd567e4c
--- /dev/null
+++ b/libc/src/sys/epoll/epoll_create.h
@@ -0,0 +1,18 @@
+//===-- Implementation header for epoll_create function ---------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE_H
+#define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE_H
+
+namespace LIBC_NAMESPACE {
+
+int epoll_create(int size);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE_H
diff --git a/libc/src/sys/epoll/epoll_create1.h b/libc/src/sys/epoll/epoll_create1.h
new file mode 100644
index 0000000000000..70f446b592660
--- /dev/null
+++ b/libc/src/sys/epoll/epoll_create1.h
@@ -0,0 +1,18 @@
+//===-- Implementation header for epoll_create1 function --------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE1_H
+#define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE1_H
+
+namespace LIBC_NAMESPACE {
+
+int epoll_create1(int flags);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CREATE1_H
diff --git a/libc/src/sys/epoll/epoll_ctl.h b/libc/src/sys/epoll/epoll_ctl.h
new file mode 100644
index 0000000000000..42402949d06b3
--- /dev/null
+++ b/libc/src/sys/epoll/epoll_ctl.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for epoll_ctl function ------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CTL_H
+#define LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CTL_H
+
+#include "hdr/types/struct_epoll_event.h"
+
+namespace LIBC_NAMESPACE {
+
+// TODO: event should be nullable
+int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_SYS_EPOLL_EPOLL_CTL_H
diff --git a/libc/src/sys/epoll/epoll_pwait.h b/libc/src/sys/epoll/epoll_pwait.h
index 801aa97610001..bcae72b3e2c26 100644
--- a/libc/src/sys/epoll/epoll_pwait.h
+++ b/libc/src/sys/epoll/epoll_pwait.h
@@ -15,8 +15,8 @@
 namespace LIBC_NAMESPACE {
 
 // TODO: sigmask should be nullable
-int epoll_pwait(int epfd, epoll_event *events, int maxevents, int timeout,
-                const sigset_t *sigmask);
+int epoll_pwait(int epfd, struct epoll_event *events, int maxevents,
+                int timeout, const sigset_t *sigmask);
 
 } // namespace LIBC_NAMESPACE
 
diff --git a/libc/src/sys/epoll/linux/CMakeLists.txt b/libc/src/sys/epoll/linux/CMakeLists.txt
index 586aac7055dc1..4e661b262b85b 100644
--- a/libc/src/sys/epoll/linux/CMakeLists.txt
+++ b/libc/src/sys/epoll/linux/CMakeLists.txt
@@ -1,3 +1,41 @@
+add_entrypoint_object(
+  epoll_create
+  SRCS
+    epoll_create.cpp
+  HDRS
+    ../epoll_create.h
+  DEPENDS
+    libc.include.sys_syscall
+    libc.src.__support.OSUtil.osutil
+    libc.src.errno.errno
+)
+
+add_entrypoint_object(
+  epoll_create1
+  SRCS
+    epoll_create1.cpp
+  HDRS
+    ../epoll_create1.h
+  DEPENDS
+    libc.include.sys_syscall
+    libc.src.__support.OSUtil.osutil
+    libc.src.errno.errno
+)
+
+add_entrypoint_object(
+  epoll_ctl
+  SRCS
+    epoll_ctl.cpp
+  HDRS
+    ../epoll_ctl.h
+  DEPENDS
+    libc.hdr.types.struct_epoll_event
+    libc.hdr.sys_epoll_macros
+    libc.include.sys_syscall
+    libc.src.__support.OSUtil.osutil
+    libc.src.errno.errno
+)
+
 add_entrypoint_object(
   epoll_wait
   SRCS
@@ -23,7 +61,8 @@ add_entrypoint_object(
     libc.hdr.types.sigset_t
     libc.hdr.types.struct_epoll_event
     libc.hdr.types.struct_timespec
-    libc.include.signal
+    libc.hdr.sys_epoll_macros
+    libc.hdr.signal_macros
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
@@ -39,9 +78,9 @@ add_entrypoint_object(
     libc.hdr.types.sigset_t
     libc.hdr.types.struct_epoll_event
     libc.hdr.types.struct_timespec
-    libc.include.signal
+    libc.hdr.sys_epoll_macros
+    libc.hdr.signal_macros
     libc.include.sys_syscall
-    libc.include.time
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
 )
diff --git a/libc/src/sys/epoll/linux/epoll_create.cpp b/libc/src/sys/epoll/linux/epoll_create.cpp
new file mode 100644
index 0000000000000..d4995c476074e
--- /dev/null
+++ b/libc/src/sys/epoll/linux/epoll_create.cpp
@@ -0,0 +1,38 @@
+//===---------- Linux implementation of the epoll_create function ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/sys/epoll/epoll_create.h"
+
+#include "src/__support/OSUtil/syscall.h" // For internal syscall function.
+#include "src/__support/common.h"
+#include "src/errno/libc_errno.h"
+#include <sys/syscall.h> // For syscall numbers.
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, epoll_create, ([[maybe_unused]] int size)) {
+#ifdef SYS_epoll_create
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(SYS_epoll_create, size);
+#elif defined(SYS_epoll_create1)
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(SYS_epoll_create1, 0);
+#else
+#error                                                                         \
+    "epoll_create and epoll_create1 are unavailable. Unable to build epoll_create."
+#endif
+
+  // A negative return value indicates an error with the magnitude of the
+  // value being the error code.
+  if (ret < 0) {
+    libc_errno = -ret;
+    return -1;
+  }
+
+  return ret;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/sys/epoll/linux/epoll_create1.cpp b/libc/src/sys/epoll/linux/epoll_create1.cpp
new file mode 100644
index 0000000000000..30f1a9910274d
--- /dev/null
+++ b/libc/src/sys/epoll/linux/epoll_create1.cpp
@@ -0,0 +1,31 @@
+//===---------- Linux implementation of the epoll_create1 function --------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/sys/epoll/epoll_create1.h"
+
+#include "src/__support/OSUtil/syscall.h" // For internal syscall function.
+#include "src/__support/common.h"
+#include "src/errno/libc_errno.h"
+#include <sys/syscall.h> // For syscall numbers.
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, epoll_create1, (int flags)) {
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(SYS_epoll_create1, flags);
+
+  // A negative return value indicates an error with the magnitude of the
+  // value being the error code.
+  if (ret < 0) {
+    libc_errno = -ret;
+    return -1;
+  }
+
+  return ret;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/sys/epoll/linux/epoll_ctl.cpp b/libc/src/sys/epoll/linux/epoll_ctl.cpp
new file mode 100644
index 0000000000000..c111c5e4e0ee0
--- /dev/null
+++ b/libc/src/sys/epoll/linux/epoll_ctl.cpp
@@ -0,0 +1,34 @@
+//===---------- Linux implementation of the epoll_ctl function ----------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/sys/epoll/epoll_ctl.h"
+
+#include "hdr/types/struct_epoll_event.h"
+#include "src/__support/OSUtil/syscall.h" // For internal syscall function.
+#include "src/__support/common.h"
+#include "src/errno/libc_errno.h"
+#include <sys/syscall.h> // For syscall numbers.
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, epoll_ctl,
+                   (int epfd, int op, int fd, epoll_event *event)) {
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(SYS_epoll_ctl, epfd, op, fd,
+                                              reinterpret_cast<long>(event));
+
+  // A negative return value indicates an error with the magnitude of the
+  // value being the error code.
+  if (ret < 0) {
+    libc_errno = -ret;
+    return -1;
+  }
+
+  return ret;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/sys/epoll/linux/epoll_pwait.cpp b/libc/src/sys/epoll/linux/epoll_pwait.cpp
index ac012944a9577..8f498d18d547d 100644
--- a/libc/src/sys/epoll/linux/epoll_pwait.cpp
+++ b/libc/src/sys/epoll/linux/epoll_pwait.cpp
@@ -8,6 +8,7 @@
 
 #include "src/sys/epoll/epoll_pwait.h"
 
+#include "hdr/signal_macros.h" // for NSIG
 #include "hdr/types/sigset_t.h"
 #include "hdr/types/struct_epoll_event.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
@@ -23,7 +24,7 @@ LLVM_LIBC_FUNCTION(int, epoll_pwait,
                     int timeout, const sigset_t *sigmask)) {
   int ret = LIBC_NAMESPACE::syscall_impl<int>(
       SYS_epoll_pwait, epfd, reinterpret_cast<long>(events), maxevents, timeout,
-      reinterpret_cast<long>(sigmask), sizeof(sigset_t));
+      reinterpret_cast<long>(sigmask), NSIG / 8);
 
   // A negative return value indicates an error with the magnitude of the
   // value being the error code.
@@ -32,7 +33,7 @@ LLVM_LIBC_FUNCTION(int, epoll_pwait,
     return -1;
   }
 
-  return 0;
+  return ret;
 }
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/src/sys/epoll/linux/epoll_pwait2.cpp b/libc/src/sys/epoll/linux/epoll_pwait2.cpp
index 3c42e38deb22b..bd33cb6325cea 100644
--- a/libc/src/sys/epoll/linux/epoll_pwait2.cpp
+++ b/libc/src/sys/epoll/linux/epoll_pwait2.cpp
@@ -8,6 +8,7 @@
 
 #include "src/sys/epoll/epoll_pwait2.h"
 
+#include "hdr/signal_macros.h" // for NSIG
 #include "hdr/types/sigset_t.h"
 #include "hdr/types/struct_epoll_event.h"
 #include "hdr/types/struct_timespec.h"
@@ -25,7 +26,7 @@ LLVM_LIBC_FUNCTION(int, epoll_pwait2,
   int ret = LIBC_NAMESPACE::syscall_impl<int>(
       SYS_epoll_pwait2, epfd, reinterpret_cast<long>(events), maxevents,
       reinterpret_cast<long>(timeout), reinterpret_cast<long>(sigmask),
-      sizeof(sigset_t));
+      NSIG / 8);
 
   // A negative return value indicates an error with the magnitude of the
   // value being the error code.
@@ -34,7 +35,7 @@ LLVM_LIBC_FUNCTION(int, epoll_pwait2,
     return -1;
   }
 
-  return 0;
+  return ret;
 }
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/src/sys/epoll/linux/epoll_wait.cpp b/libc/src/sys/epoll/linux/epoll_wait.cpp
index 18dd6e2b83543..95238d872d52f 100644
--- a/libc/src/sys/epoll/linux/epoll_wait.cpp
+++ b/libc/src/sys/epoll/linux/epoll_wait.cpp
@@ -8,6 +8,7 @@
 
 #include "src/sys/epoll/epoll_wait.h"
 
+#include "hdr/signal_macros.h" // for NSIG
 #include "hdr/types/sigset_t.h"
 #include "hdr/types/struct_epoll_event.h"
 #include "src/__support/OSUtil/syscall.h" // For internal syscall function.
@@ -27,7 +28,7 @@ LLVM_LIBC_FUNCTION(int, epoll_wait,
 #elif defined(SYS_epoll_pwait)
   int ret = LIBC_NAMESPACE::syscall_impl<int>(
       SYS_epoll_pwait, epfd, reinterpret_cast<long>(events), maxevents, timeout,
-      reinterpret_cast<long>(nullptr), sizeof(sigset_t));
+      reinterpret_cast<long>(nullptr), NSIG / 8);
 #else
 #error "epoll_wait and epoll_pwait are unavailable. Unable to build epoll_wait."
 #endif
@@ -38,7 +39,7 @@ LLVM_LIBC_FUNCTION(int, epoll_wait,
     return -1;
   }
 
-  return 0;
+  return ret;
 }
 
 } // namespace LIBC_NAMESPACE
diff --git a/libc/src/unistd/CMakeLists.txt b/libc/src/unistd/CMakeLists.txt
index 58b44f5307ca7..77db76518350c 100644
--- a/libc/src/unistd/CMakeLists.txt
+++ b/libc/src/unistd/CMakeLists.txt
@@ -160,6 +160,13 @@ add_entrypoint_object(
     .${LIBC_TARGET_OS}.lseek
 )
 
+add_entrypoint_object(
+  pipe
+  ALIAS
+  DEPENDS
+    .${LIBC_TARGET_OS}.pipe
+)
+
 add_entrypoint_object(
   pread
   ALIAS
diff --git a/libc/src/unistd/linux/CMakeLists.txt b/libc/src/unistd/linux/CMakeLists.txt
index df85d44e9e9ed..7d831f9c29c74 100644
--- a/libc/src/unistd/linux/CMakeLists.txt
+++ b/libc/src/unistd/linux/CMakeLists.txt
@@ -273,6 +273,19 @@ add_entrypoint_object(
     libc.src.errno.errno
 )
 
+add_entrypoint_object(
+  pipe
+  SRCS
+    pipe.cpp
+  HDRS
+    ../pipe.h
+  DEPENDS
+    libc.include.unistd
+    libc.include.sys_syscall
+    libc.src.__support.OSUtil.osutil
+    libc.src.errno.errno
+)
+
 add_entrypoint_object(
   pread
   SRCS
diff --git a/libc/src/unistd/linux/pipe.cpp b/libc/src/unistd/linux/pipe.cpp
new file mode 100644
index 0000000000000..b4e8b9b7d9c85
--- /dev/null
+++ b/libc/src/unistd/linux/pipe.cpp
@@ -0,0 +1,33 @@
+//===-- Linux implementation of pipe --------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/unistd/pipe.h"
+
+#include "src/__support/OSUtil/syscall.h" // For internal syscall function.
+#include "src/__support/common.h"
+#include "src/errno/libc_errno.h"
+#include <sys/syscall.h> // For syscall numbers.
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pipe, (int pipefd[2])) {
+#ifdef SYS_pipe
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(SYS_pipe,
+                                              reinterpret_cast<long>(pipefd));
+#elif defined(SYS_pipe2)
+  int ret = LIBC_NAMESPACE::syscall_impl<int>(
+      SYS_pipe2, reinterpret_cast<long>(pipefd), 0);
+#endif
+  if (ret < 0) {
+    libc_errno = -ret;
+    return -1;
+  }
+  return ret;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/unistd/pipe.h b/libc/src/unistd/pipe.h
new file mode 100644
index 0000000000000..0e20bb49c0584
--- /dev/null
+++ b/libc/src/unistd/pipe.h
@@ -0,0 +1,18 @@
+//===-- Implementation header for pipe --------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_UNISTD_PIPE_H
+#define LLVM_LIBC_SRC_UNISTD_PIPE_H
+
+namespace LIBC_NAMESPACE {
+
+int pipe(int pipefd[2]);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_UNISTD_PIPE_H
diff --git a/libc/test/src/sys/epoll/linux/CMakeLists.txt b/libc/test/src/sys/epoll/linux/CMakeLists.txt
index e346af2328db2..8f4b698ef1411 100644
--- a/libc/test/src/sys/epoll/linux/CMakeLists.txt
+++ b/libc/test/src/sys/epoll/linux/CMakeLists.txt
@@ -1,4 +1,50 @@
 add_custom_target(libc_sys_epoll_unittests)
+
+add_libc_unittest(
+  epoll_create_test
+  SUITE
+    libc_sys_epoll_unittests
+  SRCS
+    epoll_create_test.cpp
+  DEPENDS
+    libc.include.sys_epoll
+    libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create
+    libc.src.unistd.close
+    libc.test.UnitTest.ErrnoSetterMatcher
+)
+
+add_libc_unittest(
+  epoll_create1_test
+  SUITE
+    libc_sys_epoll_unittests
+  SRCS
+    epoll_create1_test.cpp
+  DEPENDS
+    libc.hdr.sys_epoll_macros
+    libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create1
+    libc.src.unistd.close
+    libc.test.UnitTest.ErrnoSetterMatcher
+)
+
+add_libc_unittest(
+  epoll_ctl_test
+  SUITE
+    libc_sys_epoll_unittests
+  SRCS
+    epoll_ctl_test.cpp
+  DEPENDS
+    libc.hdr.sys_epoll_macros
+    libc.hdr.types.struct_epoll_event
+    libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create1
+    libc.src.sys.epoll.epoll_ctl
+    libc.src.unistd.pipe
+    libc.src.unistd.close
+    libc.test.UnitTest.ErrnoSetterMatcher
+)
+
 add_libc_unittest(
   epoll_wait_test
   SUITE
@@ -6,9 +52,14 @@ add_libc_unittest(
   SRCS
     epoll_wait_test.cpp
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.sys_epoll_macros
+    libc.hdr.types.struct_epoll_event
     libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create1
+    libc.src.sys.epoll.epoll_ctl
     libc.src.sys.epoll.epoll_wait
+    libc.src.unistd.pipe
+    libc.src.unistd.close
     libc.test.UnitTest.ErrnoSetterMatcher
 )
 
@@ -19,9 +70,14 @@ add_libc_unittest(
   SRCS
     epoll_pwait_test.cpp
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.sys_epoll_macros
+    libc.hdr.types.struct_epoll_event
     libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create1
+    libc.src.sys.epoll.epoll_ctl
     libc.src.sys.epoll.epoll_pwait
+    libc.src.unistd.pipe
+    libc.src.unistd.close
     libc.test.UnitTest.ErrnoSetterMatcher
 )
 
@@ -32,8 +88,14 @@ add_libc_unittest(
   SRCS
     epoll_pwait2_test.cpp
   DEPENDS
-    libc.include.sys_epoll
+    libc.hdr.sys_epoll_macros
+    libc.hdr.types.struct_epoll_event
+    libc.hdr.types.struct_timespec
     libc.src.errno.errno
+    libc.src.sys.epoll.epoll_create1
+    libc.src.sys.epoll.epoll_ctl
     libc.src.sys.epoll.epoll_pwait2
+    libc.src.unistd.pipe
+    libc.src.unistd.close
     libc.test.UnitTest.ErrnoSetterMatcher
 )
diff --git a/libc/test/src/sys/epoll/linux/epoll_create1_test.cpp b/libc/test/src/sys/epoll/linux/epoll_create1_test.cpp
new file mode 100644
index 0000000000000..4059afe16b807
--- /dev/null
+++ b/libc/test/src/sys/epoll/linux/epoll_create1_test.cpp
@@ -0,0 +1,31 @@
+//===-- Unittests for epoll_create1 ---------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#include "hdr/sys_epoll_macros.h"
+#include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create1.h"
+#include "src/unistd/close.h"
+#include "test/UnitTest/ErrnoSetterMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
+
+TEST(LlvmLibcEpollCreate1Test, Basic) {
+  int fd = LIBC_NAMESPACE::epoll_create1(0);
+  ASSERT_GT(fd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(fd), Succeeds());
+}
+
+TEST(LlvmLibcEpollCreate1Test, CloseOnExecute) {
+  int fd = LIBC_NAMESPACE::epoll_create1(EPOLL_CLOEXEC);
+  ASSERT_GT(fd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(fd), Succeeds());
+}
diff --git a/libc/test/src/sys/epoll/linux/epoll_create_test.cpp b/libc/test/src/sys/epoll/linux/epoll_create_test.cpp
new file mode 100644
index 0000000000000..fdcdcf8eb4271
--- /dev/null
+++ b/libc/test/src/sys/epoll/linux/epoll_create_test.cpp
@@ -0,0 +1,26 @@
+//===-- Unittests for epoll_create ----------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create.h"
+#include "src/unistd/close.h"
+#include "test/UnitTest/ErrnoSetterMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
+
+TEST(LlvmLibcEpollCreateTest, Basic) {
+  int fd = LIBC_NAMESPACE::epoll_create(1);
+  ASSERT_GT(fd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(fd), Succeeds());
+}
+
+TEST(LlvmLibcEpollCreateTest, Fails) {
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_create(0), Fails(EINVAL));
+}
diff --git a/libc/test/src/sys/epoll/linux/epoll_ctl_test.cpp b/libc/test/src/sys/epoll/linux/epoll_ctl_test.cpp
new file mode 100644
index 0000000000000..fa2d358c57966
--- /dev/null
+++ b/libc/test/src/sys/epoll/linux/epoll_ctl_test.cpp
@@ -0,0 +1,47 @@
+//===-- Unittests for epoll_ctl -------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "hdr/sys_epoll_macros.h"
+#include "hdr/types/struct_epoll_event.h"
+#include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create1.h"
+#include "src/sys/epoll/epoll_ctl.h"
+#include "src/unistd/close.h"
+#include "src/unistd/pipe.h"
+#include "test/UnitTest/ErrnoSetterMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
+
+TEST(LlvmLibcEpollCtlTest, Basic) {
+  int epfd = LIBC_NAMESPACE::epoll_create1(0);
+  ASSERT_GT(epfd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  int pipefd[2];
+
+  ASSERT_THAT(LIBC_NAMESPACE::pipe(pipefd), Succeeds());
+
+  epoll_event event;
+  event.events = EPOLLOUT;
+  event.data.fd = pipefd[0];
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_ADD, pipefd[0], &event),
+              Succeeds());
+
+  // adding the same file fail.
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_ADD, pipefd[0], &event),
+              Fails(EEXIST));
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_DEL, pipefd[0], &event),
+              Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[0]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[1]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(epfd), Succeeds());
+}
diff --git a/libc/test/src/sys/epoll/linux/epoll_pwait2_test.cpp b/libc/test/src/sys/epoll/linux/epoll_pwait2_test.cpp
index 83fe12bb8eb65..2f4c9854be1a0 100644
--- a/libc/test/src/sys/epoll/linux/epoll_pwait2_test.cpp
+++ b/libc/test/src/sys/epoll/linux/epoll_pwait2_test.cpp
@@ -5,16 +5,53 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
+#include "hdr/sys_epoll_macros.h"
+#include "hdr/types/struct_epoll_event.h"
+#include "hdr/types/struct_timespec.h"
 #include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create1.h"
+#include "src/sys/epoll/epoll_ctl.h"
 #include "src/sys/epoll/epoll_pwait2.h"
+#include "src/unistd/close.h"
+#include "src/unistd/pipe.h"
 #include "test/UnitTest/ErrnoSetterMatcher.h"
 #include "test/UnitTest/Test.h"
 
 using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
 
-TEST(LlvmLibcEpollWaitTest, Basic) {
-  EXPECT_THAT(LIBC_NAMESPACE::epoll_pwait2(-1, nullptr, 0, nullptr, nullptr),
-              returns(EQ(-1ul)).with_errno(EQ(EINVAL)));
-}
+TEST(LlvmLibcEpollPwaitTest, Basic) {
+  int epfd = LIBC_NAMESPACE::epoll_create1(0);
+  ASSERT_GT(epfd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  int pipefd[2];
+
+  ASSERT_THAT(LIBC_NAMESPACE::pipe(pipefd), Succeeds());
+
+  epoll_event event;
+  event.events = EPOLLOUT;
+  event.data.fd = pipefd[0];
+
+  timespec time_spec;
+  time_spec.tv_sec = 0;
+  time_spec.tv_nsec = 0;
 
-// TODO: Complete these tests when epoll_create is implemented.
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_ADD, pipefd[0], &event),
+              Succeeds());
+
+  // Timeout of 0 causes immediate return. We just need to check that the
+  // interface works, we're not testing the kernel behavior here.
+  ASSERT_THAT(
+      LIBC_NAMESPACE::epoll_pwait2(epfd, &event, 1, &time_spec, nullptr),
+      Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_pwait2(-1, &event, 1, &time_spec, nullptr),
+              Fails(EBADF));
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_DEL, pipefd[0], &event),
+              Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[0]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[1]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(epfd), Succeeds());
+}
diff --git a/libc/test/src/sys/epoll/linux/epoll_pwait_test.cpp b/libc/test/src/sys/epoll/linux/epoll_pwait_test.cpp
index 217facbfcebb9..8e14aea8b9d57 100644
--- a/libc/test/src/sys/epoll/linux/epoll_pwait_test.cpp
+++ b/libc/test/src/sys/epoll/linux/epoll_pwait_test.cpp
@@ -5,16 +5,48 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
+#include "hdr/sys_epoll_macros.h"
+#include "hdr/types/struct_epoll_event.h"
 #include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create1.h"
+#include "src/sys/epoll/epoll_ctl.h"
 #include "src/sys/epoll/epoll_pwait.h"
+#include "src/unistd/close.h"
+#include "src/unistd/pipe.h"
+
 #include "test/UnitTest/ErrnoSetterMatcher.h"
 #include "test/UnitTest/Test.h"
 
 using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
 
-TEST(LlvmLibcEpollWaitTest, Basic) {
-  EXPECT_THAT(LIBC_NAMESPACE::epoll_pwait(-1, nullptr, 0, 0, nullptr),
-              returns(EQ(-1ul)).with_errno(EQ(EINVAL)));
-}
+TEST(LlvmLibcEpollPwaitTest, Basic) {
+  int epfd = LIBC_NAMESPACE::epoll_create1(0);
+  ASSERT_GT(epfd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  int pipefd[2];
+
+  ASSERT_THAT(LIBC_NAMESPACE::pipe(pipefd), Succeeds());
+
+  epoll_event event;
+  event.events = EPOLLOUT;
+  event.data.fd = pipefd[0];
 
-// TODO: Complete these tests when epoll_create is implemented.
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_ADD, pipefd[0], &event),
+              Succeeds());
+
+  // Timeout of 0 causes immediate return. We just need to check that the
+  // interface works, we're not testing the kernel behavior here.
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_pwait(epfd, &event, 1, 0, nullptr),
+              Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_pwait(-1, &event, 1, 0, nullptr),
+              Fails(EBADF));
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_DEL, pipefd[0], &event),
+              Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[0]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[1]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(epfd), Succeeds());
+}
diff --git a/libc/test/src/sys/epoll/linux/epoll_wait_test.cpp b/libc/test/src/sys/epoll/linux/epoll_wait_test.cpp
index 57fef3f690b62..f9e855a891b05 100644
--- a/libc/test/src/sys/epoll/linux/epoll_wait_test.cpp
+++ b/libc/test/src/sys/epoll/linux/epoll_wait_test.cpp
@@ -5,16 +5,45 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
+#include "hdr/sys_epoll_macros.h"
+#include "hdr/types/struct_epoll_event.h"
 #include "src/errno/libc_errno.h"
+#include "src/sys/epoll/epoll_create1.h"
+#include "src/sys/epoll/epoll_ctl.h"
 #include "src/sys/epoll/epoll_wait.h"
+#include "src/unistd/close.h"
+#include "src/unistd/pipe.h"
 #include "test/UnitTest/ErrnoSetterMatcher.h"
 #include "test/UnitTest/Test.h"
 
 using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
 
 TEST(LlvmLibcEpollWaitTest, Basic) {
-  EXPECT_THAT(LIBC_NAMESPACE::epoll_wait(-1, nullptr, 0, 0),
-              returns(EQ(-1ul)).with_errno(EQ(EINVAL)));
-}
+  int epfd = LIBC_NAMESPACE::epoll_create1(0);
+  ASSERT_GT(epfd, 0);
+  ASSERT_ERRNO_SUCCESS();
+
+  int pipefd[2];
+
+  ASSERT_THAT(LIBC_NAMESPACE::pipe(pipefd), Succeeds());
+
+  epoll_event event;
+  event.events = EPOLLOUT;
+  event.data.fd = pipefd[0];
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_ADD, pipefd[0], &event),
+              Succeeds());
 
-// TODO: Complete these tests when epoll_create is implemented.
+  // Timeout of 0 causes immediate return. We just need to check that the
+  // interface works, we're not testing the kernel behavior here.
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_wait(epfd, &event, 1, 0), Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_wait(-1, &event, 1, 0), Fails(EBADF));
+
+  ASSERT_THAT(LIBC_NAMESPACE::epoll_ctl(epfd, EPOLL_CTL_DEL, pipefd[0], &event),
+              Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[0]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[1]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(epfd), Succeeds());
+}
diff --git a/libc/test/src/unistd/CMakeLists.txt b/libc/test/src/unistd/CMakeLists.txt
index 9de5081f0094e..de3e8d9ccbb62 100644
--- a/libc/test/src/unistd/CMakeLists.txt
+++ b/libc/test/src/unistd/CMakeLists.txt
@@ -212,6 +212,21 @@ add_libc_unittest(
     libc.test.UnitTest.ErrnoSetterMatcher
 )
 
+add_libc_unittest(
+  pipe_test
+  SUITE
+    libc_unistd_unittests
+  SRCS
+    pipe_test.cpp
+  DEPENDS
+    libc.include.errno
+    libc.include.unistd
+    libc.src.errno.errno
+    libc.src.unistd.close
+    libc.src.unistd.pipe
+    libc.test.UnitTest.ErrnoSetterMatcher
+)
+
 add_libc_unittest(
   rmdir_test
   SUITE
diff --git a/libc/test/src/unistd/pipe_test.cpp b/libc/test/src/unistd/pipe_test.cpp
new file mode 100644
index 0000000000000..9c633dea1dba5
--- /dev/null
+++ b/libc/test/src/unistd/pipe_test.cpp
@@ -0,0 +1,26 @@
+//===-- Unittests for pipe ------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#include "src/unistd/close.h"
+#include "src/unistd/pipe.h"
+
+#include "test/UnitTest/ErrnoSetterMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using namespace LIBC_NAMESPACE::testing::ErrnoSetterMatcher;
+
+TEST(LlvmLibcPipeTest, SmokeTest) {
+
+  int pipefd[2];
+
+  ASSERT_THAT(LIBC_NAMESPACE::pipe(pipefd), Succeeds());
+
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[0]), Succeeds());
+  ASSERT_THAT(LIBC_NAMESPACE::close(pipefd[1]), Succeeds());
+}
+
+// TODO: Functionality tests
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index c61904a967117..9270a9d75b01c 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -106,7 +106,7 @@ libc_support_library(
     hdrs = ["include/llvm-libc-macros/linux/fcntl-macros.h"],
 )
 
-############################ Proxy Header Files ################################
+########################### Macro Proxy Header Files ###########################
 
 libc_support_library(
     name = "hdr_math_macros",
@@ -118,6 +118,33 @@ libc_support_library(
     hdrs = ["hdr/fenv_macros.h"],
 )
 
+libc_support_library(
+    name = "hdr_signal_macros",
+    hdrs = ["hdr/signal_macros.h"],
+)
+
+libc_support_library(
+    name = "hdr_sys_epoll_macros",
+    hdrs = ["hdr/sys_epoll_macros.h"],
+)
+
+############################ Type Proxy Header Files ###########################
+
+libc_support_library(
+    name = "types_sigset_t",
+    hdrs = ["hdr/types/sigset_t.h"],
+)
+
+libc_support_library(
+    name = "types_struct_epoll_event",
+    hdrs = ["hdr/types/struct_epoll_event.h"],
+)
+
+libc_support_library(
+    name = "types_struct_timespec",
+    hdrs = ["hdr/types/struct_timespec.h"],
+)
+
 ############################### Support libraries ##############################
 
 libc_support_library(
@@ -2941,6 +2968,17 @@ libc_function(
     ],
 )
 
+libc_function(
+    name = "pipe",
+    srcs = ["src/unistd/linux/pipe.cpp"],
+    hdrs = ["src/unistd/pipe.h"],
+    deps = [
+        ":__support_common",
+        ":__support_osutil_syscall",
+        ":errno",
+    ],
+)
+
 libc_function(
     name = "lseek",
     srcs = ["src/unistd/linux/lseek.cpp"],
@@ -3417,13 +3455,51 @@ libc_function(
 
 ############################## sys/epoll targets ###############################
 
-libc_support_library(
-  name = "types_sigset_t",
-  hdrs = ["hdr/types/sigset_t.h"],
+libc_function(
+    name = "epoll_create",
+    srcs = ["src/sys/epoll/linux/epoll_create.cpp"],
+    hdrs = ["src/sys/epoll/epoll_create.h"],
+    target_compatible_with = select({
+        "@platforms//os:linux": [],
+        "//conditions:default": ["@platforms//:incompatible"],
+    }),
+    weak = True,
+    deps = [
+        ":__support_osutil_syscall",
+        ":errno",
+    ],
 )
-libc_support_library(
-  name = "types_struct_epoll_event",
-  hdrs = ["hdr/types/struct_epoll_event.h"],
+
+libc_function(
+    name = "epoll_create1",
+    srcs = ["src/sys/epoll/linux/epoll_create1.cpp"],
+    hdrs = ["src/sys/epoll/epoll_create1.h"],
+    target_compatible_with = select({
+        "@platforms//os:linux": [],
+        "//conditions:default": ["@platforms//:incompatible"],
+    }),
+    weak = True,
+    deps = [
+        ":__support_osutil_syscall",
+        ":errno",
+    ],
+)
+
+libc_function(
+    name = "epoll_ctl",
+    srcs = ["src/sys/epoll/linux/epoll_ctl.cpp"],
+    hdrs = ["src/sys/epoll/epoll_ctl.h"],
+    target_compatible_with = select({
+        "@platforms//os:linux": [],
+        "//conditions:default": ["@platforms//:incompatible"],
+    }),
+    weak = True,
+    deps = [
+        ":__support_osutil_syscall",
+        ":errno",
+        ":hdr_sys_epoll_macros",
+        ":types_struct_epoll_event",
+    ],
 )
 
 libc_function(
@@ -3438,6 +3514,8 @@ libc_function(
     deps = [
         ":__support_osutil_syscall",
         ":errno",
+        ":hdr_signal_macros",
+        ":hdr_sys_epoll_macros",
         ":types_sigset_t",
         ":types_struct_epoll_event",
     ],
@@ -3455,6 +3533,8 @@ libc_function(
     deps = [
         ":__support_osutil_syscall",
         ":errno",
+        ":hdr_signal_macros",
+        ":hdr_sys_epoll_macros",
         ":types_sigset_t",
         ":types_struct_epoll_event",
     ],
@@ -3466,9 +3546,18 @@ libc_function(
 #     name = "epoll_pwait2",
 #     srcs = ["src/sys/epoll/linux/epoll_pwait2.cpp"],
 #     hdrs = ["src/sys/epoll/epoll_pwait2.h"],
+#     target_compatible_with = select({
+#         "@platforms//os:linux": [],
+#         "//conditions:default": ["@platforms//:incompatible"],
+#     }),
 #     weak = True,
 #     deps = [
 #         ":__support_osutil_syscall",
 #         ":errno",
+#         ":hdr_signal_macros",
+#         ":hdr_sys_epoll_macros",
+#         ":types_sigset_t",
+#         ":types_struct_epoll_event",
+#         ":types_struct_timespec",
 #     ],
 # )
diff --git a/utils/bazel/llvm-project-overlay/libc/test/src/sys/epoll/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/src/sys/epoll/BUILD.bazel
index e6b63235757e5..b090bde35b88d 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/src/sys/epoll/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/src/sys/epoll/BUILD.bazel
@@ -10,11 +10,55 @@ package(default_visibility = ["//visibility:public"])
 
 licenses(["notice"])
 
+libc_test(
+    name = "epoll_create_test",
+    srcs = ["linux/epoll_create_test.cpp"],
+    libc_function_deps = [
+        "//libc:epoll_create",
+        "//libc:close",
+    ],
+)
+
+libc_test(
+    name = "epoll_create1_test",
+    srcs = ["linux/epoll_create1_test.cpp"],
+    libc_function_deps = [
+        "//libc:epoll_create1",
+        "//libc:close",
+    ],
+    deps = [
+        "//libc:hdr_sys_epoll_macros",
+    ],
+)
+
+libc_test(
+    name = "epoll_ctl_test",
+    srcs = ["linux/epoll_ctl_test.cpp"],
+    libc_function_deps = [
+        "//libc:epoll_create1",
+        "//libc:epoll_ctl",
+        "//libc:pipe",
+        "//libc:close",
+    ],
+    deps = [
+        "//libc:hdr_sys_epoll_macros",
+        "//libc:types_struct_epoll_event",
+    ],
+)
+
 libc_test(
     name = "epoll_wait_test",
     srcs = ["linux/epoll_wait_test.cpp"],
     libc_function_deps = [
         "//libc:epoll_wait",
+        "//libc:epoll_create1",
+        "//libc:epoll_ctl",
+        "//libc:pipe",
+        "//libc:close",
+    ],
+    deps = [
+        "//libc:hdr_sys_epoll_macros",
+        "//libc:types_struct_epoll_event",
     ],
 )
 
@@ -23,6 +67,14 @@ libc_test(
     srcs = ["linux/epoll_pwait_test.cpp"],
     libc_function_deps = [
         "//libc:epoll_pwait",
+        "//libc:epoll_create1",
+        "//libc:epoll_ctl",
+        "//libc:pipe",
+        "//libc:close",
+    ],
+    deps = [
+        "//libc:hdr_sys_epoll_macros",
+        "//libc:types_struct_epoll_event",
     ],
 )
 
@@ -33,5 +85,14 @@ libc_test(
 #     srcs = ["linux/epoll_pwait2_test.cpp"],
 #     libc_function_deps = [
 #         "//libc:epoll_pwait2",
+#         "//libc:epoll_create1",
+#         "//libc:epoll_ctl",
+#         "//libc:pipe",
+#         "//libc:close",
+#     ],
+#     deps = [
+#         "//libc:hdr_sys_epoll_macros",
+#         "//libc:types_struct_epoll_event",
+#         "//libc:types_struct_timespec",
 #     ],
 # )

From 7aa371687ace40b85f04e21956e03f1e93052b56 Mon Sep 17 00:00:00 2001
From: Nathan Ridge <zeratul976@hotmail.com>
Date: Thu, 11 Apr 2024 19:34:53 -0400
Subject: [PATCH 202/886] [clangd] Avoid using CompletionItemKind.Text for
 macro completions from the index (#88236)

This was fixed in https://github.com/clangd/clangd/issues/1484 for Sema
completions but the fix did not apply to index completions.

Fixes https://github.com/clangd/clangd/issues/2002
---
 clang-tools-extra/clangd/CodeComplete.cpp        | 16 +++++++++++++---
 .../clangd/unittests/CodeCompleteTests.cpp       |  7 +++++--
 clang-tools-extra/clangd/unittests/TestIndex.cpp |  7 ++++++-
 clang-tools-extra/clangd/unittests/TestIndex.h   |  4 +++-
 4 files changed, 27 insertions(+), 7 deletions(-)

diff --git a/clang-tools-extra/clangd/CodeComplete.cpp b/clang-tools-extra/clangd/CodeComplete.cpp
index 9e321dce4c504..89eee392837af 100644
--- a/clang-tools-extra/clangd/CodeComplete.cpp
+++ b/clang-tools-extra/clangd/CodeComplete.cpp
@@ -89,7 +89,11 @@ const CodeCompleteOptions::CodeCompletionRankingModel
 
 namespace {
 
-CompletionItemKind toCompletionItemKind(index::SymbolKind Kind) {
+// Note: changes to this function should also be reflected in the
+// CodeCompletionResult overload where appropriate.
+CompletionItemKind
+toCompletionItemKind(index::SymbolKind Kind,
+                     const llvm::StringRef *Signature = nullptr) {
   using SK = index::SymbolKind;
   switch (Kind) {
   case SK::Unknown:
@@ -99,7 +103,10 @@ CompletionItemKind toCompletionItemKind(index::SymbolKind Kind) {
   case SK::NamespaceAlias:
     return CompletionItemKind::Module;
   case SK::Macro:
-    return CompletionItemKind::Text;
+    // Use macro signature (if provided) to tell apart function-like and
+    // object-like macros.
+    return Signature && Signature->contains('(') ? CompletionItemKind::Function
+                                                 : CompletionItemKind::Constant;
   case SK::Enum:
     return CompletionItemKind::Enum;
   case SK::Struct:
@@ -150,6 +157,8 @@ CompletionItemKind toCompletionItemKind(index::SymbolKind Kind) {
   llvm_unreachable("Unhandled clang::index::SymbolKind.");
 }
 
+// Note: changes to this function should also be reflected in the
+// index::SymbolKind overload where appropriate.
 CompletionItemKind toCompletionItemKind(const CodeCompletionResult &Res,
                                         CodeCompletionContext::Kind CtxKind) {
   if (Res.Declaration)
@@ -379,7 +388,8 @@ struct CodeCompletionBuilder {
       if (Completion.Scope.empty())
         Completion.Scope = std::string(C.IndexResult->Scope);
       if (Completion.Kind == CompletionItemKind::Missing)
-        Completion.Kind = toCompletionItemKind(C.IndexResult->SymInfo.Kind);
+        Completion.Kind = toCompletionItemKind(C.IndexResult->SymInfo.Kind,
+                                               &C.IndexResult->Signature);
       if (Completion.Name.empty())
         Completion.Name = std::string(C.IndexResult->Name);
       if (Completion.FilterText.empty())
diff --git a/clang-tools-extra/clangd/unittests/CodeCompleteTests.cpp b/clang-tools-extra/clangd/unittests/CodeCompleteTests.cpp
index 49337bddf98d5..8fbac73cb653b 100644
--- a/clang-tools-extra/clangd/unittests/CodeCompleteTests.cpp
+++ b/clang-tools-extra/clangd/unittests/CodeCompleteTests.cpp
@@ -671,7 +671,8 @@ TEST(CompletionTest, Kinds) {
           #define MACRO 10
           int X = ^
       )cpp",
-      {func("indexFunction"), var("indexVariable"), cls("indexClass")});
+      {func("indexFunction"), var("indexVariable"), cls("indexClass"),
+       macro("indexObjMacro"), macro("indexFuncMacro", "(x, y)")});
   EXPECT_THAT(Results.Completions,
               AllOf(has("function", CompletionItemKind::Function),
                     has("variable", CompletionItemKind::Variable),
@@ -680,7 +681,9 @@ TEST(CompletionTest, Kinds) {
                     has("MACRO", CompletionItemKind::Constant),
                     has("indexFunction", CompletionItemKind::Function),
                     has("indexVariable", CompletionItemKind::Variable),
-                    has("indexClass", CompletionItemKind::Class)));
+                    has("indexClass", CompletionItemKind::Class),
+                    has("indexObjMacro", CompletionItemKind::Constant),
+                    has("indexFuncMacro", CompletionItemKind::Function)));
 
   Results = completions("nam^");
   EXPECT_THAT(Results.Completions,
diff --git a/clang-tools-extra/clangd/unittests/TestIndex.cpp b/clang-tools-extra/clangd/unittests/TestIndex.cpp
index 278336bdde2ee..b13a5d32d1752 100644
--- a/clang-tools-extra/clangd/unittests/TestIndex.cpp
+++ b/clang-tools-extra/clangd/unittests/TestIndex.cpp
@@ -38,7 +38,7 @@ static std::string replace(llvm::StringRef Haystack, llvm::StringRef Needle,
 // Helpers to produce fake index symbols for memIndex() or completions().
 // USRFormat is a regex replacement string for the unqualified part of the USR.
 Symbol sym(llvm::StringRef QName, index::SymbolKind Kind,
-           llvm::StringRef USRFormat) {
+           llvm::StringRef USRFormat, llvm::StringRef Signature) {
   Symbol Sym;
   std::string USR = "c:"; // We synthesize a few simple cases of USRs by hand!
   size_t Pos = QName.rfind("::");
@@ -55,6 +55,7 @@ Symbol sym(llvm::StringRef QName, index::SymbolKind Kind,
   Sym.SymInfo.Kind = Kind;
   Sym.Flags |= Symbol::IndexedForCodeCompletion;
   Sym.Origin = SymbolOrigin::Static;
+  Sym.Signature = Signature;
   return Sym;
 }
 
@@ -86,6 +87,10 @@ Symbol conceptSym(llvm::StringRef Name) {
   return sym(Name, index::SymbolKind::Concept, "@CT@\\0");
 }
 
+Symbol macro(llvm::StringRef Name, llvm::StringRef ArgList) {
+  return sym(Name, index::SymbolKind::Macro, "@macro@\\0", ArgList);
+}
+
 Symbol objcSym(llvm::StringRef Name, index::SymbolKind Kind,
                llvm::StringRef USRPrefix) {
   Symbol Sym;
diff --git a/clang-tools-extra/clangd/unittests/TestIndex.h b/clang-tools-extra/clangd/unittests/TestIndex.h
index 9280b0b12a67f..0699b29392d72 100644
--- a/clang-tools-extra/clangd/unittests/TestIndex.h
+++ b/clang-tools-extra/clangd/unittests/TestIndex.h
@@ -20,7 +20,7 @@ Symbol symbol(llvm::StringRef QName);
 // Helpers to produce fake index symbols with proper SymbolID.
 // USRFormat is a regex replacement string for the unqualified part of the USR.
 Symbol sym(llvm::StringRef QName, index::SymbolKind Kind,
-           llvm::StringRef USRFormat);
+           llvm::StringRef USRFormat, llvm::StringRef Signature = {});
 // Creats a function symbol assuming no function arg.
 Symbol func(llvm::StringRef Name);
 // Creates a class symbol.
@@ -35,6 +35,8 @@ Symbol var(llvm::StringRef Name);
 Symbol ns(llvm::StringRef Name);
 // Create a C++20 concept symbol.
 Symbol conceptSym(llvm::StringRef Name);
+// Create a macro symbol.
+Symbol macro(llvm::StringRef Name, llvm::StringRef ArgList = {});
 
 // Create an Objective-C symbol.
 Symbol objcSym(llvm::StringRef Name, index::SymbolKind Kind,

From c42a262560a659d1cfb9ec46712ad66420314f2d Mon Sep 17 00:00:00 2001
From: Alexandre Eichenberger <alexe@us.ibm.com>
Date: Thu, 11 Apr 2024 20:04:26 -0400
Subject: [PATCH 203/886] [MLIR] Bug Fix: affine.prefetch replaceAffineOp
 invoked during canonicalization (#88346)

Signed-off-by: Alexandre Eichenberger <alexe@us.ibm.com>
---
 mlir/lib/Dialect/Affine/IR/AffineOps.cpp   |  5 ++---
 mlir/test/Dialect/Affine/canonicalize.mlir | 14 ++++++++++++++
 2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/mlir/lib/Dialect/Affine/IR/AffineOps.cpp b/mlir/lib/Dialect/Affine/IR/AffineOps.cpp
index c591e5056480c..c9c0a7b4cc686 100644
--- a/mlir/lib/Dialect/Affine/IR/AffineOps.cpp
+++ b/mlir/lib/Dialect/Affine/IR/AffineOps.cpp
@@ -1487,9 +1487,8 @@ void SimplifyAffineOp<AffinePrefetchOp>::replaceAffineOp(
     PatternRewriter &rewriter, AffinePrefetchOp prefetch, AffineMap map,
     ArrayRef<Value> mapOperands) const {
   rewriter.replaceOpWithNewOp<AffinePrefetchOp>(
-      prefetch, prefetch.getMemref(), map, mapOperands,
-      prefetch.getLocalityHint(), prefetch.getIsWrite(),
-      prefetch.getIsDataCache());
+      prefetch, prefetch.getMemref(), map, mapOperands, prefetch.getIsWrite(),
+      prefetch.getLocalityHint(), prefetch.getIsDataCache());
 }
 template <>
 void SimplifyAffineOp<AffineStoreOp>::replaceAffineOp(
diff --git a/mlir/test/Dialect/Affine/canonicalize.mlir b/mlir/test/Dialect/Affine/canonicalize.mlir
index 7c0930eedc856..730808dbbdf18 100644
--- a/mlir/test/Dialect/Affine/canonicalize.mlir
+++ b/mlir/test/Dialect/Affine/canonicalize.mlir
@@ -1452,3 +1452,17 @@ func.func @mod_of_mod(%lb: index, %ub: index, %step: index) -> (index, index) {
   %1 = affine.apply affine_map<()[s0, s1, s2] -> ((s0 - ((s0 - s2) mod s1) - s2) mod s1)> ()[%ub, %step, %lb]
   return %0, %1 : index, index
 }
+
+// -----
+
+// CHECK-LABEL:  func.func @prefetch_canonicalize
+// CHECK-SAME:   ([[PARAM_0_:%.+]]: memref<512xf32>) {
+func.func @prefetch_canonicalize(%arg0: memref<512xf32>) -> () {
+  // CHECK: affine.for [[I_0_:%.+]] = 0 to 8 {
+  affine.for %arg3 = 0 to 8  {
+    %1 = affine.apply affine_map<()[s0] -> (s0 * 64)>()[%arg3]
+    // CHECK: affine.prefetch [[PARAM_0_]][symbol([[I_0_]]) * 64], read, locality<3>, data : memref<512xf32>
+    affine.prefetch %arg0[%1], read, locality<3>, data : memref<512xf32>
+  }
+  return
+}

From 56954a53e58282d7584e31ec14a2b1052cd861e8 Mon Sep 17 00:00:00 2001
From: pawelszczerbuk <153013546+pawelszczerbuk@users.noreply.github.com>
Date: Thu, 11 Apr 2024 17:09:44 -0700
Subject: [PATCH 204/886] [MLIR][LoopPipelining] Improve schedule verifier, so
 it checks also operands of nested operations (#88450)

`verifySchedule` was not looking at the operands of nested operations,
which caused incorrect schedule to be allowed in some cases, potentially
leading to crash during expansion.
There is also minor fix in `cloneAndUpdateOperands` in the pipeline
expander that prevents double visit of the cloned op. This one has no
functional impact, so no test for it.
---
 .../Dialect/SCF/Transforms/LoopPipelining.cpp | 18 +++++++++++----
 mlir/test/Dialect/SCF/loop-pipelining.mlir    | 23 +++++++++++++++++++
 2 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp b/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
index 9eda1a4597ba4..82ec95d31f525 100644
--- a/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
+++ b/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
@@ -203,6 +203,17 @@ bool LoopPipelinerInternal::initializeLoopInfo(
   return true;
 }
 
+/// Find operands of all the nested operations within `op`.
+static SetVector<Value> getNestedOperands(Operation *op) {
+  SetVector<Value> operands;
+  op->walk([&](Operation *nestedOp) {
+    for (Value operand : nestedOp->getOperands()) {
+      operands.insert(operand);
+    }
+  });
+  return operands;
+}
+
 /// Compute unrolled cycles of each op (consumer) and verify that each op is
 /// scheduled after its operands (producers) while adjusting for the distance
 /// between producer and consumer.
@@ -219,7 +230,7 @@ bool LoopPipelinerInternal::verifySchedule() {
   }
   for (Operation *consumer : opOrder) {
     int64_t consumerCycle = unrolledCyles[consumer];
-    for (Value operand : consumer->getOperands()) {
+    for (Value operand : getNestedOperands(consumer)) {
       auto [producer, distance] = getDefiningOpAndDistance(operand);
       if (!producer)
         continue;
@@ -245,9 +256,8 @@ static Operation *
 cloneAndUpdateOperands(RewriterBase &rewriter, Operation *op,
                        function_ref<void(OpOperand *newOperand)> callback) {
   Operation *clone = rewriter.clone(*op);
-  for (OpOperand &operand : clone->getOpOperands())
-    callback(&operand);
-  clone->walk([&](Operation *nested) {
+  clone->walk<WalkOrder::PreOrder>([&](Operation *nested) {
+    // 'clone' itself will be visited first.
     for (OpOperand &operand : nested->getOpOperands()) {
       Operation *def = operand.get().getDefiningOp();
       if ((def && !clone->isAncestor(def)) || isa<BlockArgument>(operand.get()))
diff --git a/mlir/test/Dialect/SCF/loop-pipelining.mlir b/mlir/test/Dialect/SCF/loop-pipelining.mlir
index 8d6f454d18751..46e7feca4329e 100644
--- a/mlir/test/Dialect/SCF/loop-pipelining.mlir
+++ b/mlir/test/Dialect/SCF/loop-pipelining.mlir
@@ -846,3 +846,26 @@ func.func @invalid_schedule2(%A: memref<?xf32>, %result: memref<?xf32>) {
   }  { __test_pipelining_loop__ }
   return
 }
+
+// -----
+
+func.func @invalid_schedule3(%A: memref<?xf32>, %result: memref<?xf32>, %ext: index) {
+  %c0 = arith.constant 0 : index
+  %c1 = arith.constant 1 : index
+  %c4 = arith.constant 4 : index
+  %r = scf.for %i0 = %c0 to %c4 step %c1 iter_args(%idx = %c0) -> (index) {
+    %cnd = arith.cmpi slt, %ext, %c4 { __test_pipelining_stage__ = 0, __test_pipelining_op_order__ = 0 } : index
+    // expected-error@+1 {{operation scheduled before its operands}}
+    %idx1 = scf.if %cnd -> (index) {
+      %idxinc = arith.addi %idx, %c1 : index
+      scf.yield %idxinc : index
+    } else {
+      scf.yield %idx : index
+    } { __test_pipelining_stage__ = 0, __test_pipelining_op_order__ = 1 }
+    %A_elem = memref.load %A[%idx1] { __test_pipelining_stage__ = 0, __test_pipelining_op_order__ = 2 } : memref<?xf32>
+    %idx2 = arith.addi %idx1, %c1 { __test_pipelining_stage__ = 1, __test_pipelining_op_order__ = 3 } : index
+    memref.store %A_elem, %result[%idx1] { __test_pipelining_stage__ = 2, __test_pipelining_op_order__ = 4 } : memref<?xf32>
+    scf.yield %idx2 : index
+  }  { __test_pipelining_loop__ }
+  return
+}

From bf1d7b8df287d69ee265b91be40dec37267b2d5c Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Thu, 11 Apr 2024 16:55:22 -0700
Subject: [PATCH 205/886] [RISCV] Give HWASAN_CHECK_MEMACCESS_SHORTGRANULES
 pseudo the same code size as PseudoCall.

This is converted to a PseudoCall in RISCVAsmPrinter.cpp so it should
have the same size of 8 bytes.
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.td b/llvm/lib/Target/RISCV/RISCVInstrInfo.td
index 2cfad7f7c0611..f9dadc6c0d489 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.td
@@ -1861,7 +1861,7 @@ def : Pat<(trap), (UNIMP)>;
 def : Pat<(debugtrap), (EBREAK)>;
 
 let Predicates = [IsRV64], Uses = [X5],
-    Defs = [X1, X6, X7, X28, X29, X30, X31] in
+    Defs = [X1, X6, X7, X28, X29, X30, X31], Size = 8 in
 def HWASAN_CHECK_MEMACCESS_SHORTGRANULES
   : Pseudo<(outs), (ins GPRJALR:$ptr, i32imm:$accessinfo),
            [(int_hwasan_check_memaccess_shortgranules (i64 X5), GPRJALR:$ptr,

From 09d51a841dcfbc41c3d7f3274b109b5f9fb09bb0 Mon Sep 17 00:00:00 2001
From: "Felix (Ting Wang)" <Ting.Wang.SH@ibm.com>
Date: Fri, 12 Apr 2024 08:18:01 +0800
Subject: [PATCH 206/886] [PowerPC][AIX] Enable aix-small-local-dynamic-tls
 target attribute (#86641)

Following the aix-small-local-exec-tls target attribute, this patch adds
a target attribute for an AIX-specific option in llc that informs the
compiler that it can use a faster access sequence for the local-dynamic
TLS model (formally named aix-small-local-dynamic-tls) when TLS
variables are less than ~32KB in size.

The patch either produces an addi/la with a displacement off of module
handle (return value from .__tls_get_mod) when the address is
calculated, or it produces an addi/la followed by a load/store when the
address is calculated and used for further accesses.

---------

Co-authored-by: Amy Kwan <amy.kwan1@ibm.com>
---
 .../MCTargetDesc/PPCXCOFFObjectWriter.cpp     |    4 +
 llvm/lib/Target/PowerPC/PPC.td                |    9 +
 llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp     |   20 +-
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp   |   32 +-
 llvm/lib/Target/PowerPC/PPCMCInstLower.cpp    |   13 +-
 llvm/lib/Target/PowerPC/PPCSubtarget.cpp      |   26 +-
 ...aix-small-local-dynamic-tls-largeaccess.ll |  182 +-
 .../aix-small-local-dynamic-tls-types.ll      | 1601 ++++++++++++++---
 ...ix-small-local-exec-tls-opt-IRattribute.ll |   25 +-
 .../check-aix-small-local-exec-tls-opt.ll     |   14 +-
 10 files changed, 1530 insertions(+), 396 deletions(-)

diff --git a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCXCOFFObjectWriter.cpp b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCXCOFFObjectWriter.cpp
index f4998e9b9dcba..714ce64a39391 100644
--- a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCXCOFFObjectWriter.cpp
+++ b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCXCOFFObjectWriter.cpp
@@ -71,6 +71,8 @@ std::pair<uint8_t, uint8_t> PPCXCOFFObjectWriter::getRelocTypeAndSignSize(
       return {XCOFF::RelocationType::R_TOCL, SignAndSizeForHalf16};
     case MCSymbolRefExpr::VK_PPC_AIX_TLSLE:
       return {XCOFF::RelocationType::R_TLS_LE, SignAndSizeForHalf16};
+    case MCSymbolRefExpr::VK_PPC_AIX_TLSLD:
+      return {XCOFF::RelocationType::R_TLS_LD, SignAndSizeForHalf16};
     }
   } break;
   case PPC::fixup_ppc_half16ds:
@@ -86,6 +88,8 @@ std::pair<uint8_t, uint8_t> PPCXCOFFObjectWriter::getRelocTypeAndSignSize(
       return {XCOFF::RelocationType::R_TOCL, 15};
     case MCSymbolRefExpr::VK_PPC_AIX_TLSLE:
       return {XCOFF::RelocationType::R_TLS_LE, 15};
+    case MCSymbolRefExpr::VK_PPC_AIX_TLSLD:
+      return {XCOFF::RelocationType::R_TLS_LD, 15};
     }
   } break;
   case PPC::fixup_ppc_br24:
diff --git a/llvm/lib/Target/PowerPC/PPC.td b/llvm/lib/Target/PowerPC/PPC.td
index 535616d33a803..b962ed28d7200 100644
--- a/llvm/lib/Target/PowerPC/PPC.td
+++ b/llvm/lib/Target/PowerPC/PPC.td
@@ -329,6 +329,15 @@ def FeatureAIXLocalExecTLS :
                    "Produce a TOC-free local-exec TLS sequence for this function "
                    "for 64-bit AIX">;
 
+// Specifies that local-dynamic TLS accesses in any function with this target
+// attribute should use the optimized sequence (where the offset is an immediate
+// off the module-handle for which the linker might add fix-up code for if the
+// immediate is too large).
+def FeatureAIXLocalDynamicTLS :
+  SubtargetFeature<"aix-small-local-dynamic-tls", "HasAIXSmallLocalDynamicTLS",
+                   "true", "Produce a faster local-dynamic TLS sequence for this "
+                   "function for 64-bit AIX">;
+
 def FeaturePredictableSelectIsExpensive :
   SubtargetFeature<"predictable-select-expensive",
                    "PredictableSelectIsExpensive",
diff --git a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
index 16942c6893a16..1c57b92057fff 100644
--- a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
+++ b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
@@ -803,7 +803,8 @@ void PPCAsmPrinter::emitInstruction(const MachineInstr *MI) {
   MCInst TmpInst;
   const bool IsPPC64 = Subtarget->isPPC64();
   const bool IsAIX = Subtarget->isAIXABI();
-  const bool HasAIXSmallLocalExecTLS = Subtarget->hasAIXSmallLocalExecTLS();
+  const bool HasAIXSmallLocalTLS = Subtarget->hasAIXSmallLocalExecTLS() ||
+                                   Subtarget->hasAIXSmallLocalDynamicTLS();
   const Module *M = MF->getFunction().getParent();
   PICLevel::Level PL = M->getPICLevel();
 
@@ -1612,11 +1613,11 @@ void PPCAsmPrinter::emitInstruction(const MachineInstr *MI) {
   case PPC::LFD:
   case PPC::STFD:
   case PPC::ADDI8: {
-    // A faster non-TOC-based local-exec sequence is represented by `addi`
-    // or a load/store instruction (that directly loads or stores off of the
-    // thread pointer) with an immediate operand having the MO_TPREL_FLAG.
+    // A faster non-TOC-based local-[exec|dynamic] sequence is represented by
+    // `addi` or a load/store instruction (that directly loads or stores off of
+    // the thread pointer) with an immediate operand having the MO_TPREL_FLAG.
     // Such instructions do not otherwise arise.
-    if (!HasAIXSmallLocalExecTLS)
+    if (!HasAIXSmallLocalTLS)
       break;
     bool IsMIADDI8 = MI->getOpcode() == PPC::ADDI8;
     unsigned OpNum = IsMIADDI8 ? 2 : 1;
@@ -1624,7 +1625,7 @@ void PPCAsmPrinter::emitInstruction(const MachineInstr *MI) {
     unsigned Flag = MO.getTargetFlags();
     if (Flag == PPCII::MO_TPREL_FLAG ||
         Flag == PPCII::MO_GOT_TPREL_PCREL_FLAG ||
-        Flag == PPCII::MO_TPREL_PCREL_FLAG) {
+        Flag == PPCII::MO_TPREL_PCREL_FLAG || Flag == PPCII::MO_TLSLD_FLAG) {
       LowerPPCMachineInstrToMCInst(MI, TmpInst, *this);
 
       const MCExpr *Expr = getAdjustedLocalExecExpr(MO, MO.getOffset());
@@ -1672,7 +1673,12 @@ const MCExpr *PPCAsmPrinter::getAdjustedLocalExecExpr(const MachineOperand &MO,
 
   assert(MO.isGlobal() && "Only expecting a global MachineOperand here!");
   const GlobalValue *GValue = MO.getGlobal();
-  assert(TM.getTLSModel(GValue) == TLSModel::LocalExec &&
+  // TODO: Handle the aix-small-local-dynamic-tls non-zero offset case.
+  TLSModel::Model Model = TM.getTLSModel(GValue);
+  if (Model == TLSModel::LocalDynamic) {
+    return nullptr;
+  }
+  assert(Model == TLSModel::LocalExec &&
          "Only local-exec accesses are handled!");
 
   bool IsGlobalADeclaration = GValue->isDeclarationForLinker();
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 52d5b71367059..d27932f2915fb 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -153,10 +153,10 @@ static SDValue widenVec(SelectionDAG &DAG, SDValue Vec, const SDLoc &dl);
 
 static const char AIXSSPCanaryWordName[] = "__ssp_canary_word";
 
-// A faster local-exec TLS access sequence (enabled with the
-// -maix-small-local-exec-tls option) can be produced for TLS variables;
-// consistent with the IBM XL compiler, we apply a max size of slightly under
-// 32KB.
+// A faster local-[exec|dynamic] TLS access sequence (enabled with the
+// -maix-small-local-[exec|dynamic]-tls option) can be produced for TLS
+// variables; consistent with the IBM XL compiler, we apply a max size of
+// slightly under 32KB.
 constexpr uint64_t AIXSmallTlsPolicySizeLimit = 32751;
 
 // FIXME: Remove this once the bug has been fixed!
@@ -3434,6 +3434,14 @@ SDValue PPCTargetLowering::LowerGlobalTLSAddressAIX(SDValue Op,
   }
 
   if (Model == TLSModel::LocalDynamic) {
+    bool HasAIXSmallLocalDynamicTLS = Subtarget.hasAIXSmallLocalDynamicTLS();
+
+    // We do not implement the 32-bit version of the faster access sequence
+    // for local-dynamic that is controlled by -maix-small-local-dynamic-tls.
+    if (!Is64Bit && HasAIXSmallLocalDynamicTLS)
+      report_fatal_error("The small-local-dynamic TLS access sequence is "
+                         "currently only supported on AIX (64-bit mode).");
+
     // For local-dynamic on AIX, we need to generate one TOC entry for each
     // variable offset, and a single module-handle TOC entry for the entire
     // file.
@@ -3454,6 +3462,22 @@ SDValue PPCTargetLowering::LowerGlobalTLSAddressAIX(SDValue Op,
     SDValue ModuleHandle =
         DAG.getNode(PPCISD::TLSLD_AIX, dl, PtrVT, ModuleHandleTOC);
 
+    // With the -maix-small-local-dynamic-tls option, produce a faster access
+    // sequence for local-dynamic TLS variables where the offset from the
+    // module-handle is encoded as an immediate operand.
+    //
+    // We only utilize the faster local-dynamic access sequence when the TLS
+    // variable has a size within the policy limit. We treat types that are
+    // not sized or are empty as being over the policy size limit.
+    if (HasAIXSmallLocalDynamicTLS) {
+      Type *GVType = GV->getValueType();
+      if (GVType->isSized() && !GVType->isEmptyTy() &&
+          GV->getParent()->getDataLayout().getTypeAllocSize(GVType) <=
+              AIXSmallTlsPolicySizeLimit)
+        return DAG.getNode(PPCISD::Lo, dl, PtrVT, VariableOffsetTGA,
+                           ModuleHandle);
+    }
+
     return DAG.getNode(ISD::ADD, dl, PtrVT, ModuleHandle, VariableOffset);
   }
 
diff --git a/llvm/lib/Target/PowerPC/PPCMCInstLower.cpp b/llvm/lib/Target/PowerPC/PPCMCInstLower.cpp
index 9a3ca5a782936..c05bb37e58bf6 100644
--- a/llvm/lib/Target/PowerPC/PPCMCInstLower.cpp
+++ b/llvm/lib/Target/PowerPC/PPCMCInstLower.cpp
@@ -96,15 +96,18 @@ static MCOperand GetSymbolRef(const MachineOperand &MO, const MCSymbol *Symbol,
     RefKind = MCSymbolRefExpr::VK_PPC_GOT_TLSLD_PCREL;
   else if (MO.getTargetFlags() == PPCII::MO_GOT_TPREL_PCREL_FLAG)
     RefKind = MCSymbolRefExpr::VK_PPC_GOT_TPREL_PCREL;
-  else if (MO.getTargetFlags() == PPCII::MO_TPREL_FLAG) {
+  else if (MO.getTargetFlags() == PPCII::MO_TPREL_FLAG ||
+           MO.getTargetFlags() == PPCII::MO_TLSLD_FLAG) {
     assert(MO.isGlobal() && "Only expecting a global MachineOperand here!");
     TLSModel::Model Model = TM.getTLSModel(MO.getGlobal());
-    // For the local-exec TLS model, we may generate the offset from the TLS
-    // base as an immediate operand (instead of using a TOC entry).
-    // Set the relocation type in case the result is used for purposes other
-    // than a TOC reference. In TOC reference cases, this result is discarded.
+    // For the local-[exec|dynamic] TLS model, we may generate the offset from
+    // the TLS base as an immediate operand (instead of using a TOC entry). Set
+    // the relocation type in case the result is used for purposes other than a
+    // TOC reference. In TOC reference cases, this result is discarded.
     if (Model == TLSModel::LocalExec)
       RefKind = MCSymbolRefExpr::VK_PPC_AIX_TLSLE;
+    else if (Model == TLSModel::LocalDynamic)
+      RefKind = MCSymbolRefExpr::VK_PPC_AIX_TLSLD;
   }
 
   const MachineInstr *MI = MO.getParent();
diff --git a/llvm/lib/Target/PowerPC/PPCSubtarget.cpp b/llvm/lib/Target/PowerPC/PPCSubtarget.cpp
index 653d9bda99192..d1722555f1fcb 100644
--- a/llvm/lib/Target/PowerPC/PPCSubtarget.cpp
+++ b/llvm/lib/Target/PowerPC/PPCSubtarget.cpp
@@ -124,22 +124,22 @@ void PPCSubtarget::initSubtargetFeatures(StringRef CPU, StringRef TuneCPU,
   // Determine endianness.
   IsLittleEndian = TM.isLittleEndian();
 
-  if (HasAIXSmallLocalExecTLS) {
+  if (HasAIXSmallLocalExecTLS || HasAIXSmallLocalDynamicTLS) {
     if (!TargetTriple.isOSAIX() || !IsPPC64)
-      report_fatal_error(
-          "The aix-small-local-exec-tls attribute is only supported on AIX in "
-          "64-bit mode.\n",
-          false);
-    // The aix-small-local-exec-tls attribute should only be used with
+      report_fatal_error("The aix-small-local-[exec|dynamic]-tls attribute is "
+                         "only supported on AIX in "
+                         "64-bit mode.\n",
+                         false);
+    // The aix-small-local-[exec|dynamic]-tls attribute should only be used with
     // -data-sections, as having data sections turned off with this option
-    // is not ideal for performance. Moreover, the small-local-exec-tls region
-    // is a limited resource, and should not be used for variables that may
-    // be replaced.
+    // is not ideal for performance. Moreover, the
+    // small-local-[exec|dynamic]-tls region is a limited resource, and should
+    // not be used for variables that may be replaced.
     if (!TM.getDataSections())
-      report_fatal_error(
-          "The aix-small-local-exec-tls attribute can only be specified with "
-          "-data-sections.\n",
-          false);
+      report_fatal_error("The aix-small-local-[exec|dynamic]-tls attribute can "
+                         "only be specified with "
+                         "-data-sections.\n",
+                         false);
   }
 }
 
diff --git a/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-largeaccess.ll b/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-largeaccess.ll
index eb16bae67150e..7db1048c258cd 100644
--- a/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-largeaccess.ll
+++ b/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-largeaccess.ll
@@ -1,9 +1,9 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
 ; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
-; RUN:      -mtriple powerpc64-ibm-aix-xcoff < %s \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff < %s \
 ; RUN:      | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-SMALLCM64
 ; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
-; RUN:      -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
 ; RUN:      < %s | FileCheck %s \
 ; RUN:      --check-prefix=SMALL-LOCAL-DYNAMIC-LARGECM64
 
@@ -39,27 +39,23 @@ define signext i32 @test1() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r6, L..C1(r2) # target-flags(ppc-tlsld) @ElementIntTLS2
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r7, L..C2(r2) # target-flags(ppc-tlsld) @ElementIntTLS3
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r8, L..C3(r2) # target-flags(ppc-tlsld) @ElementIntTLS4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r9, L..C4(r2) # target-flags(ppc-tlsld) @ElementIntTLS5
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r6, 4
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r5, L..C5(r2) # target-flags(ppc-tlsld) @ElementIntTLSv1
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 1
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r6, r3, r6
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r7, r3, r7
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r8, r3, r8
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r9, r3, r9
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stwux r4, r3, r5
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 24(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 2
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 320(r6)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 3
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 324(r7)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 88
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 328(r8)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 332(r9)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r5, 1
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r4, ElementIntTLSv1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r5, ElementIntTLSv1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r5, ElementIntTLS2[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r6, 24(r4)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 2
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 320(r5)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r4, ElementIntTLS3[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r5, 3
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r5, 324(r4)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r4, ElementIntTLS4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementIntTLS5[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r6, 328(r4)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 88
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 332(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 102
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
@@ -71,34 +67,25 @@ define signext i32 @test1() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C1@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C2@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r6, 4
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r8, L..C3@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r9, L..C4@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r7, L..C2@l(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r8, L..C3@l(r8)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r9, L..C4@l(r9)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r5, L..C1@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C5@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 1
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r6, L..C5@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r7, r3, r7
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r8, r3, r8
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r9, r3, r9
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r6, r3, r6
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stwux r4, r3, r5
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 4
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r4, 24(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 2
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 320(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 3
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 324(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 88
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r4, 328(r8)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 332(r9)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r5, 1
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r4, ElementIntTLSv1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r5, ElementIntTLSv1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r5, ElementIntTLS2[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r6, 24(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 2
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r4, 320(r5)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r4, ElementIntTLS3[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r5, 3
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r5, 324(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r4, ElementIntTLS4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementIntTLS5[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r6, 328(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 88
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r4, 332(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 102
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
@@ -144,31 +131,26 @@ define i64 @test2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C6(r2) # target-flags(ppc-tlsld) @ElementLongTLS6
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mr r6, r3
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 212
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r4, r6, r4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r3, 424(r4)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C7(r2) # target-flags(ppc-tlsld) @ElementLongTLS2
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementLongTLS6[UL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 212
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 424(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementLongTLS2[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 203
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 1200(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C8(r2) # target-flags(ppc-tlsgdm) @MyTLSGDVar
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsgd) @MyTLSGDVar
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C1(r2) # target-flags(ppc-tlsgdm) @MyTLSGDVar
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C2(r2) # target-flags(ppc-tlsgd) @MyTLSGDVar
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_addr[PR]
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 44
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 440(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C10(r2) # target-flags(ppc-tlsld) @ElementLongTLS3
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementLongTLS3[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 6
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 2000(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @ElementLongTLS4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 100
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r4, r6, r4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r3, 6800(r4)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C12(r2) # target-flags(ppc-tlsld) @ElementLongTLS5
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementLongTLS4[TL]@ld(r6)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 100
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 6800(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ElementLongTLS5[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 882
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, 8400(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 1191
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
@@ -182,40 +164,30 @@ define i64 @test2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C6@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 212
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mr r6, r3
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C6@l(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r6, r3
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementLongTLS6[UL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 424(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C7@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementLongTLS2[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 203
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C7@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 1200(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C8@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r4, L..C9@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C8@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C9@l(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C1@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r4, L..C2@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C2@l(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_addr[PR]
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 44
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 440(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C10@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementLongTLS3[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 6
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C10@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 2000(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C11@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementLongTLS4[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 100
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C11@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 6800(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C12@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ElementLongTLS5[TL]@ld(r6)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 882
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C12@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, 8400(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 1191
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
@@ -257,27 +229,23 @@ define signext i32 @test3() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r8, L..C3(r2) # target-flags(ppc-tlsld) @ElementIntTLSv2
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r6, L..C1(r2) # target-flags(ppc-tlsld) @ElementIntTLS2
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r7, L..C2(r2) # target-flags(ppc-tlsld) @ElementIntTLS3
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r8, L..C3(r2) # target-flags(ppc-tlsld) @ElementIntTLS4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r9, L..C4(r2) # target-flags(ppc-tlsld) @ElementIntTLS5
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r5, L..C13(r2) # target-flags(ppc-tlsld) @ElementIntTLSv2
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 1
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r6, r3, r6
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r7, r3, r7
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r8, r3, r8
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r9, r3, r9
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stwux r4, r3, r5
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r5, ElementIntTLS2[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r6, ElementIntTLS3[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r7, ElementIntTLS4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r9, ElementIntTLS5[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stwux r4, r3, r8
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r4, 4
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 24(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 2
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 320(r6)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 320(r5)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 3
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 324(r7)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 324(r6)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 88
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 328(r8)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r4, 328(r7)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stw r3, 332(r9)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    li r3, 102
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
@@ -290,29 +258,21 @@ define signext i32 @test3() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C13@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C3@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C2@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r8, L..C3@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r9, L..C4@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r7, L..C2@l(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r8, L..C3@l(r8)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r9, L..C4@l(r9)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r6, L..C3@l(r6)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r5, L..C13@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C5@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 1
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r6, L..C5@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r7, r3, r7
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r8, r3, r8
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r9, r3, r9
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r6, r3, r6
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stwux r4, r3, r5
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r5, ElementIntTLS2[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r7, ElementIntTLS3[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r8, ElementIntTLS4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r9, ElementIntTLS5[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stwux r4, r3, r6
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r4, 4
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r4, 24(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 2
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 320(r6)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 320(r5)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stw r3, 324(r7)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    li r3, 88
diff --git a/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-types.ll b/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-types.ll
index d996d86a23d86..161a58a90296a 100644
--- a/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-types.ll
+++ b/llvm/test/CodeGen/PowerPC/aix-small-local-dynamic-tls-types.ll
@@ -1,18 +1,30 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
 ; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
-; RUN:      -mtriple powerpc64-ibm-aix-xcoff < %s \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff < %s \
 ; RUN:      | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-SMALLCM64
 ; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
+; RUN:      < %s | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-LARGECM64
+; RUN: llc  -O0 -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff < %s \
+; RUN:      | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-SMALLCM64-O0
+; RUN: llc  -O0 -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
+; RUN:      -mattr=+aix-small-local-dynamic-tls -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
+; RUN:      < %s | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-LARGECM64-O0
+; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
+; RUN:      -mtriple powerpc64-ibm-aix-xcoff < %s \
+; RUN:      | FileCheck %s --check-prefix=DEFAULT-LOCAL-DYNAMIC-SMALLCM64
+; RUN: llc  -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
 ; RUN:      -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
 ; RUN:      < %s | FileCheck %s \
-; RUN:      --check-prefix=SMALL-LOCAL-DYNAMIC-LARGECM64
+; RUN:      --check-prefix=DEFAULT-LOCAL-DYNAMIC-LARGECM64
 ; RUN: llc  -O0 -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
 ; RUN:      -mtriple powerpc64-ibm-aix-xcoff < %s \
-; RUN:      | FileCheck %s --check-prefix=SMALL-LOCAL-DYNAMIC-SMALLCM64-O0
+; RUN:      | FileCheck %s --check-prefix=DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0
 ; RUN: llc  -O0 -verify-machineinstrs -mcpu=pwr7 -ppc-asm-full-reg-names \
 ; RUN:      -mtriple powerpc64-ibm-aix-xcoff --code-model=large \
 ; RUN:      < %s | FileCheck %s \
-; RUN:      --check-prefix=SMALL-LOCAL-DYNAMIC-LARGECM64-O0
+; RUN:      --check-prefix=DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0
 
 declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull) #1
 @tlv_int_init = local_unnamed_addr global i32 87, align 4
@@ -27,7 +39,8 @@ declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull) #1
 @internal_tlv_double = internal thread_local(localdynamic) global double 1.000000e+00, align 8
 
 %struct.anon = type { i32 }
-@ThreadLocalStruct = thread_local(localdynamic) global %struct.anon zeroinitializer, align 1
+@ThreadLocalStructAlign1 = thread_local(localdynamic) global %struct.anon zeroinitializer, align 1
+@ThreadLocalStructAlign4 = thread_local(localdynamic) global %struct.anon zeroinitializer, align 4
 @a = thread_local(localdynamic) global [87 x i32] zeroinitializer, align 4
 
 define nonnull ptr @AddrTest1() local_unnamed_addr {
@@ -38,8 +51,7 @@ define nonnull ptr @AddrTest1() local_unnamed_addr {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, a[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r3, 12
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
@@ -51,12 +63,10 @@ define nonnull ptr @AddrTest1() local_unnamed_addr {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C1@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C1@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, a[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r3, 12
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
@@ -70,8 +80,7 @@ define nonnull ptr @AddrTest1() local_unnamed_addr {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, a[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r3, 12
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -81,27 +90,212 @@ define nonnull ptr @AddrTest1() local_unnamed_addr {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: AddrTest1:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C0@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, a[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r3, 12
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: AddrTest1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r3, 12
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: AddrTest1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C1@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r3, 12
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: AddrTest1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r3, 12
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: AddrTest1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C0@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r3, 12
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @a)
   %arrayidx = getelementptr inbounds [87 x i32], ptr %tlv_addr, i64 0, i64 3
   ret ptr %arrayidx
 }
 
+; Demonstrates when offset is zero, addi for the offset is optimized out.
+define nonnull ptr @AddrTest2() local_unnamed_addr {
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-LABEL: AddrTest2:
+; SMALL-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, a[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-LARGECM64-LABEL: AddrTest2:
+; SMALL-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, a[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: AddrTest2:
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, a[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: AddrTest2:
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, a[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: AddrTest2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: AddrTest2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C1@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: AddrTest2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C1(r2) # target-flags(ppc-tlsld) @a
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: AddrTest2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C0@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+entry:
+  %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @a)
+  %arrayidx = getelementptr inbounds [87 x i32], ptr %tlv_addr, i64 0, i64 0
+  ret ptr %arrayidx
+}
+
 define signext i32 @testUnaligned() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-LABEL: testUnaligned:
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
@@ -110,8 +304,8 @@ define signext i32 @testUnaligned() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C2(r2) # target-flags(ppc-tlsld) @ThreadLocalStruct
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwax r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    la r3, ThreadLocalStructAlign1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwa r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -122,12 +316,11 @@ define signext i32 @testUnaligned() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C2@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C2@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwax r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    la r3, ThreadLocalStructAlign1[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwa r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -140,8 +333,7 @@ define signext i32 @testUnaligned() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C2(r2) # target-flags(ppc-tlsld) @ThreadLocalStruct
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, ThreadLocalStructAlign1[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -151,26 +343,213 @@ define signext i32 @testUnaligned() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testUnaligned:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C2@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C2@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, ThreadLocalStructAlign1[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testUnaligned:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C2(r2) # target-flags(ppc-tlsld) @ThreadLocalStructAlign1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testUnaligned:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C2@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C2@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testUnaligned:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C2(r2) # target-flags(ppc-tlsld) @ThreadLocalStructAlign1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testUnaligned:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C2@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C2@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
-  %tlv_addr = call align 1 ptr @llvm.threadlocal.address.p0(ptr align 1 @ThreadLocalStruct)
-  %x = getelementptr inbounds %struct.anon, ptr %tlv_addr, i32 0, i32 0
-  %value = load i32, ptr %x, align 1
-  ret i32 %value
+  %tlv_addr1 = call align 1 ptr @llvm.threadlocal.address.p0(ptr align 1 @ThreadLocalStructAlign1)
+  %x = getelementptr inbounds %struct.anon, ptr %tlv_addr1, i32 0, i32 0
+  %value1 = load i32, ptr %x, align 1
+  ret i32 %value1
+}
+
+define signext i32 @testAligned() {
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-LABEL: testAligned:
+; SMALL-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwa r3, ThreadLocalStructAlign4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-LARGECM64-LABEL: testAligned:
+; SMALL-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwa r3, ThreadLocalStructAlign4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testAligned:
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, ThreadLocalStructAlign4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testAligned:
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, ThreadLocalStructAlign4[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testAligned:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C3(r2) # target-flags(ppc-tlsld) @ThreadLocalStructAlign4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testAligned:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C3@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C3@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testAligned:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C3(r2) # target-flags(ppc-tlsld) @ThreadLocalStructAlign4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testAligned:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C3@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C3@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+entry:
+  %tlv_addr1 = call align 1 ptr @llvm.threadlocal.address.p0(ptr align 4 @ThreadLocalStructAlign4)
+  %x = getelementptr inbounds %struct.anon, ptr %tlv_addr1, i32 0, i32 0
+  %value1 = load i32, ptr %x, align 1
+  ret i32 %value1
 }
 
 define void @testChar(i8 noundef signext %x) {
@@ -182,8 +561,7 @@ define void @testChar(i8 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C3(r2) # target-flags(ppc-tlsld) @tlv_char
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stbx r6, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stb r6, tlv_char[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -195,12 +573,10 @@ define void @testChar(i8 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mr r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C3@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C3@l(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stbx r6, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stb r6, tlv_char[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -217,8 +593,7 @@ define void @testChar(i8 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mr r4, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r5, L..C3(r2) # target-flags(ppc-tlsld) @tlv_char
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r4, r5
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r4, tlv_char[TL]@ld(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stb r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -232,21 +607,91 @@ define void @testChar(i8 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C3@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, 48(r1) # 8-byte Folded Reload
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mr r4, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, L..C3@l(r5)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r4, r5
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r4, tlv_char[TL]@ld(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stb r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testChar:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mr r6, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C4(r2) # target-flags(ppc-tlsld) @tlv_char
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stbx r6, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testChar:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mr r6, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C4@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C4@l(r7)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stbx r6, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testChar:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mr r4, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r5, L..C4(r2) # target-flags(ppc-tlsld) @tlv_char
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r4, r5
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stb r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testChar:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C4@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, 48(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mr r4, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, L..C4@l(r5)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r4, r5
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stb r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 1 ptr @llvm.threadlocal.address.p0(ptr align 1 @tlv_char)
   store i8 %x, ptr %tlv_addr, align 1
@@ -262,8 +707,7 @@ define void @testShort(i16 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C4(r2) # target-flags(ppc-tlsld) @tlv_short
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    sthx r6, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    sth r6, tlv_short[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -275,12 +719,10 @@ define void @testShort(i16 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mr r6, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C4@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C4@l(r7)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    sthx r6, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    sth r6, tlv_short[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -297,8 +739,7 @@ define void @testShort(i16 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mr r4, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r5, L..C4(r2) # target-flags(ppc-tlsld) @tlv_short
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r4, r5
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r4, tlv_short[TL]@ld(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    sth r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -312,21 +753,91 @@ define void @testShort(i16 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C4@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, 48(r1) # 8-byte Folded Reload
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mr r4, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, L..C4@l(r5)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r4, r5
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r4, tlv_short[TL]@ld(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    sth r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testShort:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mr r6, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C5(r2) # target-flags(ppc-tlsld) @tlv_short
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    sthx r6, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testShort:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mr r6, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r7, L..C5@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C5@l(r7)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    sthx r6, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testShort:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mr r4, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r5, L..C5(r2) # target-flags(ppc-tlsld) @tlv_short
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r4, r5
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    sth r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testShort:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stw r3, 60(r1) # 4-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C5@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, 48(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mr r4, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 60(r1) # 4-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r5, L..C5@l(r5)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r4, r5
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    sth r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 2 ptr @llvm.threadlocal.address.p0(ptr align 2 @tlv_short)
   store i16 %x, ptr %tlv_addr, align 2
@@ -341,8 +852,7 @@ define signext i32 @testInt1() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C5(r2) # target-flags(ppc-tlsld) @tlv_int
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwax r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwa r3, tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -353,12 +863,10 @@ define signext i32 @testInt1() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C5@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C5@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwax r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwa r3, tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -371,8 +879,7 @@ define signext i32 @testInt1() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C5(r2) # target-flags(ppc-tlsld) @tlv_int
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -382,21 +889,81 @@ define signext i32 @testInt1() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testInt1:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C5@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C5@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testInt1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C6(r2) # target-flags(ppc-tlsld) @tlv_int
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testInt1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C6@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C6@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwax r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testInt1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C6(r2) # target-flags(ppc-tlsld) @tlv_int
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testInt1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C6@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C6@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @tlv_int)
   %value = load i32, ptr %tlv_addr, align 4
@@ -411,9 +978,8 @@ define signext i32 @testInt2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C6(r2) # target-flags(ppc-tlsld) @internal_tlv_int
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwzx r3, r3, r4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C7(r2) # @tlv_int_init
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C1(r2) # @tlv_int_init
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwz r3, internal_tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwz r4, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r4, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    extsw r3, r3
@@ -427,14 +993,12 @@ define signext i32 @testInt2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C6@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C6@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwzx r3, r3, r4
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r4, L..C7@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C7@l(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r4, L..C1@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C1@l(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwz r3, internal_tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwz r4, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r4, r3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    extsw r3, r3
@@ -450,10 +1014,9 @@ define signext i32 @testInt2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C6(r2) # target-flags(ppc-tlsld) @internal_tlv_int
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, internal_tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r4, 0(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C7(r2) # @tlv_int_init
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C1(r2) # @tlv_int_init
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    extsw r3, r3
@@ -465,26 +1028,104 @@ define signext i32 @testInt2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testInt2:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C6@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C6@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, internal_tlv_int[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r4, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C7@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C7@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    extsw r3, r3
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testInt2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C7(r2) # target-flags(ppc-tlsld) @internal_tlv_int
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwzx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C8(r2) # @tlv_int_init
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwz r4, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    add r3, r4, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    extsw r3, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testInt2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C7@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C7@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwzx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r4, L..C8@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C8@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwz r4, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    add r3, r4, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    extsw r3, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testInt2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C7(r2) # target-flags(ppc-tlsld) @internal_tlv_int
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r4, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C8(r2) # @tlv_int_init
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    extsw r3, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testInt2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C7@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C7@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r4, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C8@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C8@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    extsw r3, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @internal_tlv_int)
   %tlv_val = load i32, ptr %tlv_addr, align 4
@@ -501,8 +1142,7 @@ define signext i64 @testLong1() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C8(r2) # target-flags(ppc-tlsld) @tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -513,12 +1153,10 @@ define signext i64 @testLong1() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C8@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C8@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -531,8 +1169,7 @@ define signext i64 @testLong1() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C8(r2) # target-flags(ppc-tlsld) @tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -542,21 +1179,81 @@ define signext i64 @testLong1() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong1:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C8@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C8@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testLong1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testLong1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C9@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C9@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testLong1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C9@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C9@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @tlv_long)
   %value = load i64, ptr %tlv_addr, align 4
@@ -571,10 +1268,9 @@ define void @testLong2(i64 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @internal_tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r5, r3, r4
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r5, r5, 9
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdx r5, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, internal_tlv_long[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r4, r4, 9
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r4, internal_tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -585,14 +1281,12 @@ define void @testLong2(i64 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C9@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C9@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r5, r3, r4
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r5, r5, 9
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdx r5, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, internal_tlv_long[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r4, r4, 9
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r4, internal_tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -605,8 +1299,7 @@ define void @testLong2(i64 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @internal_tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r4, internal_tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r3, 9
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r3, 0(r4)
@@ -618,23 +1311,91 @@ define void @testLong2(i64 noundef signext %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong2:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C9@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C9@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r4, internal_tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r3, 9
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 0(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testLong2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @internal_tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r5, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r5, r5, 9
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdx r5, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testLong2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C10@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C10@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r5, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r5, r5, 9
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdx r5, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testLong2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @internal_tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r4, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r3, 9
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C10@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C10@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r4, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r3, 9
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @internal_tlv_long)
   %value = load i64, ptr %tlv_addr, align 8
@@ -651,8 +1412,7 @@ define i32 @testLong3() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C8(r2) # target-flags(ppc-tlsld) @tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -663,12 +1423,10 @@ define i32 @testLong3() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C8@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C8@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -681,8 +1439,7 @@ define i32 @testLong3() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C8(r2) # target-flags(ppc-tlsld) @tlv_long
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    clrldi r3, r3, 32
@@ -694,23 +1451,87 @@ define i32 @testLong3() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong3:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C8@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C8@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, tlv_long[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    clrldi r3, r3, 32
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testLong3:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ldx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testLong3:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C9@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C9@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ldx r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testLong3:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C9(r2) # target-flags(ppc-tlsld) @tlv_long
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testLong3:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C9@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C9@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $r3 killed $r3 killed $x3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @tlv_long)
   %value = load i64, ptr %tlv_addr, align 8
@@ -723,18 +1544,17 @@ define void @testFloat1(float noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    vspltisw v2, 1
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    vspltisw v3, 8
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xvcvsxwdp vs0, vs34
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @tlv_float
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfsx f1, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfs f1, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    fadds f0, f1, f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xvcvsxwdp vs1, vs35
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    fadds f0, f0, f1
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfsx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfs f0, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -744,20 +1564,18 @@ define void @testFloat1(float noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    vspltisw v2, 1
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C10@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    vspltisw v3, 8
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    xvcvsxwdp vs0, vs34
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C10@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfsx f1, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfs f1, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    fadds f0, f1, f0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    xvcvsxwdp vs1, vs35
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    fadds f0, f0, f1
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfsx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfs f0, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -770,13 +1588,12 @@ define void @testFloat1(float noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @tlv_float
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f0, 0(r3)
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C11(r2) # %const.1
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C2(r2) # %const.1
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f1, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fadds f0, f0, f1
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C12(r2) # %const.0
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C3(r2) # %const.0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f1, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fadds f0, f0, f1
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfs f0, 0(r3)
@@ -788,30 +1605,120 @@ define void @testFloat1(float noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testFloat1:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C10@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C10@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f0, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C11@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C11@l(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C2@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C2@l(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f1, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fadds f0, f0, f1
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C12@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C12@l(r4)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C3@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C3@l(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f1, 0(r4)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fadds f0, f0, f1
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfs f0, 0(r3)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testFloat1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    vspltisw v2, 1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    vspltisw v3, 8
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xvcvsxwdp vs0, vs34
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @tlv_float
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfsx f1, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    fadds f0, f1, f0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xvcvsxwdp vs1, vs35
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfsx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testFloat1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    vspltisw v2, 1
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C11@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    vspltisw v3, 8
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    xvcvsxwdp vs0, vs34
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C11@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfsx f1, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    fadds f0, f1, f0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    xvcvsxwdp vs1, vs35
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfsx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testFloat1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @tlv_float
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C12(r2) # %const.1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f1, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C13(r2) # %const.0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f1, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testFloat1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C11@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C11@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C12@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C12@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f1, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r4, L..C13@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C13@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f1, 0(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fadds f0, f0, f1
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @tlv_float)
   %value = load float, ptr %tlv_addr, align 4
@@ -829,8 +1736,7 @@ define i32 @testFloat2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @tlv_float
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfsx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfs f0, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r1, 60
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xscvdpsxws f0, f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfiwx f0, 0, r3
@@ -845,12 +1751,10 @@ define i32 @testFloat2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C10@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C10@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfsx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfs f0, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r1, 60
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    xscvdpsxws f0, f0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfiwx f0, 0, r3
@@ -867,8 +1771,7 @@ define i32 @testFloat2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C10(r2) # target-flags(ppc-tlsld) @tlv_float
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f0, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fctiwz f0, f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfd f0, 56(r1)
@@ -884,14 +1787,10 @@ define i32 @testFloat2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C10@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 48(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C10@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, tlv_float[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f0, 0(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fctiwz f0, f0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfd f0, 56(r1)
@@ -901,6 +1800,86 @@ define i32 @testFloat2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testFloat2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @tlv_float
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfsx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r1, 60
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwz r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testFloat2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C11@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C11@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfsx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r1, 60
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwz r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testFloat2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @tlv_float
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    fctiwz f0, f0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfd f0, 56(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwa r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testFloat2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C11@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 48(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 48(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C11@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfs f0, 0(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    fctiwz f0, f0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfd f0, 56(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwa r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @tlv_float)
   %value = load float, ptr %tlv_addr, align 4
@@ -916,8 +1895,7 @@ define void @testDouble1(double noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @internal_tlv_double
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfdx f1, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfd f1, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
@@ -928,12 +1906,10 @@ define void @testDouble1(double noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C11@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C11@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfdx f1, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfd f1, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
@@ -946,8 +1922,7 @@ define void @testDouble1(double noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C13(r2) # target-flags(ppc-tlsld) @internal_tlv_double
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stxsdx f1, 0, r3
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
@@ -957,21 +1932,81 @@ define void @testDouble1(double noundef %x) {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testDouble1:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C13@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -48(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C13@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stxsdx f1, 0, r3
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 48
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testDouble1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C12(r2) # target-flags(ppc-tlsld) @internal_tlv_double
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfdx f1, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testDouble1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C12@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C12@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfdx f1, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testDouble1:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -48(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C14(r2) # target-flags(ppc-tlsld) @internal_tlv_double
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stxsdx f1, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 48
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testDouble1:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C14@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C14@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    add r3, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stxsdx f1, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @internal_tlv_double)
   store double %x, ptr %tlv_addr, align 8
@@ -986,8 +2021,7 @@ define i32 @testDouble2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C11(r2) # target-flags(ppc-tlsld) @internal_tlv_double
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfdx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfd f0, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r1, 60
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xscvdpsxws f0, f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfiwx f0, 0, r3
@@ -1002,12 +2036,10 @@ define i32 @testDouble2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -64(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C11@u(r2)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C11@l(r6)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfdx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfd f0, internal_tlv_double[TL]@ld(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r1, 60
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    xscvdpsxws f0, f0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfiwx f0, 0, r3
@@ -1024,8 +2056,9 @@ define i32 @testDouble2() {
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C13(r2) # target-flags(ppc-tlsld) @internal_tlv_double
-; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfdx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    la r3, internal_tlv_double[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfd f0, 0(r3)
+; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    # kill: def $f1 killed $f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    xscvdpsxws f0, f0
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r1, 52
 ; SMALL-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfiwx f0, 0, r3
@@ -1039,25 +2072,103 @@ define i32 @testDouble2() {
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testDouble2:
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -80(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 96(r1)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C13@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -64(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 80(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C0@u(r2)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C0@l(r3)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C13@l(r4)
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfdx f0, r3, r4
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    la r3, internal_tlv_double[TL]@ld(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfd f0, 0(r3)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    # kill: def $f1 killed $f0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    xscvdpsxws f0, f0
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r1, 68
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r1, 52
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfiwx f0, 0, r3
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 68(r1)
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 52(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    clrldi r3, r3, 32
-; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 80
+; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 64
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
 ; SMALL-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-LABEL: testDouble2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r4, L..C12(r2) # target-flags(ppc-tlsld) @internal_tlv_double
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lfdx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r3, r1, 60
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    lwz r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-LABEL: testDouble2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r3, L..C0@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addis r6, L..C12@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r3, L..C0@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r4, L..C12@l(r6)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lfdx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r3, r1, 60
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    lwz r3, 60(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-LABEL: testDouble2:
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stdu r1, -64(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    std r0, 80(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r3, L..C0(r2) # target-flags(ppc-tlsldm) @"_$TLSML"
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r4, L..C14(r2) # target-flags(ppc-tlsld) @internal_tlv_double
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lfdx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r3, r1, 52
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    lwz r3, 52(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    addi r1, r1, 64
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-SMALLCM64-O0-NEXT:    blr
+;
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-LABEL: testDouble2:
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0:       # %bb.0: # %entry
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mflr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stdu r1, -80(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r0, 96(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C14@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    std r3, 56(r1) # 8-byte Folded Spill
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addis r3, L..C1@u(r2)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r3, L..C1@l(r3)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    bla .__tls_get_mod[PR]
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, 56(r1) # 8-byte Folded Reload
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r4, L..C14@l(r4)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lfdx f0, r3, r4
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    xscvdpsxws f0, f0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r3, r1, 68
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    stfiwx f0, 0, r3
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    lwz r3, 68(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    clrldi r3, r3, 32
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    addi r1, r1, 80
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    ld r0, 16(r1)
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    mtlr r0
+; DEFAULT-LOCAL-DYNAMIC-LARGECM64-O0-NEXT:    blr
 entry:
   %tlv_addr = tail call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @internal_tlv_double)
   %value = load double, ptr %tlv_addr, align 8
diff --git a/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt-IRattribute.ll b/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt-IRattribute.ll
index dc78ae8436df6..fc49e9416803b 100644
--- a/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt-IRattribute.ll
+++ b/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt-IRattribute.ll
@@ -8,23 +8,32 @@
 ; RUN:   -data-sections=false < %s 2>&1 | \
 ; RUN: FileCheck %s --check-prefix=CHECK-UNSUPPORTED-NO-DATASEC
 
-define dso_local signext i32 @testWithIRAttr() #0 {
+define dso_local signext i32 @testWithIRAttrLocalExec() #0 {
 entry:
   ret i32 0
 }
-; Check that the aix-small-local-exec-tls attribute is not supported on Linux and AIX (32-bit).
-; CHECK-NOT-SUPPORTED: The aix-small-local-exec-tls attribute is only supported on AIX in 64-bit mode.
 
-; Check that the aix-small-local-exec-tls attribute is only supported when
+define dso_local signext i32 @testWithIRAttrLocalDynamic() #1 {
+entry:
+  ret i32 0
+}
+
+; Check that the aix-small-local-[exec|dynamic]-tls attribute is not supported on Linux and AIX (32-bit).
+; CHECK-NOT-SUPPORTED: The aix-small-local-[exec|dynamic]-tls attribute is only supported on AIX in 64-bit mode.
+
+; Check that the aix-small-local-[exec|dynamic]-tls attribute is only supported when
 ; data sections are enabled.
-; CHECK-UNSUPPORTED-NO-DATASEC: The aix-small-local-exec-tls attribute can only be specified with -data-sections.
+; CHECK-UNSUPPORTED-NO-DATASEC: The aix-small-local-[exec|dynamic]-tls attribute can only be specified with -data-sections.
 
 ; Make sure that the test was actually compiled successfully after using the
-; aix-small-local-exec-tls attribute.
-; CHECK-LABEL: testWithIRAttr:
+; aix-small-local-[exec|dynamic]-tls attribute.
+; CHECK-LABEL: testWithIRAttrLocalExec:
+; CHECK:        li r3, 0
+; CHECK-NEXT:   blr
+; CHECK-LABEL: testWithIRAttrLocalDynamic:
 ; CHECK:        li r3, 0
 ; CHECK-NEXT:   blr
-
 
 attributes #0 = { "target-features"="+aix-small-local-exec-tls" }
+attributes #1 = { "target-features"="+aix-small-local-dynamic-tls" }
 
diff --git a/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt.ll b/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt.ll
index 5104b7c6c8398..25bab5b5efcd2 100644
--- a/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt.ll
+++ b/llvm/test/CodeGen/PowerPC/check-aix-small-local-exec-tls-opt.ll
@@ -6,17 +6,25 @@
 ; RUN: not llc -mtriple powerpc64le-unknown-linux-gnu -mattr=+aix-small-local-exec-tls \
 ; RUN:   -ppc-asm-full-reg-names < %s 2>&1 | \
 ; RUN:   FileCheck %s --check-prefix=CHECK-NOT-SUPPORTED
+; RUN: llc -mtriple powerpc64-ibm-aix-xcoff -mattr=+aix-small-local-dynamic-tls \
+; RUN:   -ppc-asm-full-reg-names < %s | FileCheck %s
+; RUN: not llc -mtriple powerpc-ibm-aix-xcoff -mattr=+aix-small-local-dynamic-tls \
+; RUN:   -ppc-asm-full-reg-names < %s 2>&1 | \
+; RUN:   FileCheck %s --check-prefix=CHECK-NOT-SUPPORTED
+; RUN: not llc -mtriple powerpc64le-unknown-linux-gnu -mattr=+aix-small-local-dynamic-tls \
+; RUN:   -ppc-asm-full-reg-names < %s 2>&1 | \
+; RUN:   FileCheck %s --check-prefix=CHECK-NOT-SUPPORTED
 
 define dso_local signext i32 @testNoIRAttr() {
 entry:
   ret i32 0
 }
 
-; Check that the aix-small-local-exec-tls attribute is not supported on Linux and AIX (32-bit).
-; CHECK-NOT-SUPPORTED: The aix-small-local-exec-tls attribute is only supported on AIX in 64-bit mode.
+; Check that the aix-small-local-[exec|dynamic]-tls attribute is not supported on Linux and AIX (32-bit).
+; CHECK-NOT-SUPPORTED: The aix-small-local-[exec|dynamic]-tls attribute is only supported on AIX in 64-bit mode.
 
 ; Make sure that the test was actually compiled successfully after using the
-; aix-small-local-exec-tls attribute.
+; aix-small-local-[exec|dynamic]-tls attribute.
 ; CHECK-LABEL: testNoIRAttr:
 ; CHECK:        li r3, 0
 ; CHECK-NEXT:   blr

From 86df55efa1242e308535679cf05c2793ff38f3d9 Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang@intel.com>
Date: Fri, 12 Apr 2024 09:21:17 +0800
Subject: [PATCH 207/886] [X86][BF16] Promote vector ADD/SUB/MUL/DIV to f32
 (#87858)

---
 llvm/lib/Target/X86/X86ISelLowering.cpp |  18 +-
 llvm/test/CodeGen/X86/bfloat.ll         | 286 ++----------------------
 2 files changed, 31 insertions(+), 273 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index f274da6f6f776..b7cb4b7dafeb6 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1978,10 +1978,8 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::STRICT_FP_ROUND, MVT::v16f16, Custom);
     setOperationAction(ISD::FP_EXTEND, MVT::v16f32, Custom);
     setOperationAction(ISD::STRICT_FP_EXTEND, MVT::v16f32, Custom);
-    for (unsigned Opc : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV}) {
-      setOperationPromotedToType(Opc, MVT::v16f16, MVT::v16f32);
+    for (unsigned Opc : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV})
       setOperationPromotedToType(Opc, MVT::v32f16, MVT::v32f32);
-    }
 
     for (auto VT : { MVT::v16i32, MVT::v8i64, MVT::v16f32, MVT::v8f64 }) {
       setOperationAction(ISD::MLOAD,               VT, Legal);
@@ -2296,15 +2294,15 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::bf16, Custom);
     for (auto VT : {MVT::v8bf16, MVT::v16bf16}) {
       setF16Action(VT, Expand);
-      setOperationAction(ISD::FADD, VT, Expand);
-      setOperationAction(ISD::FSUB, VT, Expand);
-      setOperationAction(ISD::FMUL, VT, Expand);
-      setOperationAction(ISD::FDIV, VT, Expand);
       setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
       setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
       setOperationAction(ISD::INSERT_SUBVECTOR, VT, Legal);
       setOperationAction(ISD::CONCAT_VECTORS, VT, Custom);
     }
+    for (unsigned Opc : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV}) {
+      setOperationPromotedToType(Opc, MVT::v8bf16, MVT::v8f32);
+      setOperationPromotedToType(Opc, MVT::v16bf16, MVT::v16f32);
+    }
     setOperationAction(ISD::FP_ROUND, MVT::v8bf16, Custom);
     addLegalFPImmediate(APFloat::getZero(APFloat::BFloat()));
   }
@@ -2312,10 +2310,8 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
   if (!Subtarget.useSoftFloat() && Subtarget.hasBF16()) {
     addRegisterClass(MVT::v32bf16, &X86::VR512RegClass);
     setF16Action(MVT::v32bf16, Expand);
-    setOperationAction(ISD::FADD, MVT::v32bf16, Expand);
-    setOperationAction(ISD::FSUB, MVT::v32bf16, Expand);
-    setOperationAction(ISD::FMUL, MVT::v32bf16, Expand);
-    setOperationAction(ISD::FDIV, MVT::v32bf16, Expand);
+    for (unsigned Opc : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV})
+      setOperationPromotedToType(Opc, MVT::v32bf16, MVT::v32f32);
     setOperationAction(ISD::BUILD_VECTOR, MVT::v32bf16, Custom);
     setOperationAction(ISD::FP_ROUND, MVT::v16bf16, Custom);
     setOperationAction(ISD::VECTOR_SHUFFLE, MVT::v32bf16, Custom);
diff --git a/llvm/test/CodeGen/X86/bfloat.ll b/llvm/test/CodeGen/X86/bfloat.ll
index 6e3020d628edf..8a2109a1c78df 100644
--- a/llvm/test/CodeGen/X86/bfloat.ll
+++ b/llvm/test/CodeGen/X86/bfloat.ll
@@ -525,101 +525,13 @@ define bfloat @fold_ext_trunc2(bfloat %a) nounwind {
 define <8 x bfloat> @addv(<8 x bfloat> %a, <8 x bfloat> %b) nounwind {
 ; X86-LABEL: addv:
 ; X86:       # %bb.0:
-; X86-NEXT:    pushl %ebp
-; X86-NEXT:    pushl %ebx
-; X86-NEXT:    pushl %edi
-; X86-NEXT:    pushl %esi
-; X86-NEXT:    vmovw %xmm1, %eax
-; X86-NEXT:    shll $16, %eax
-; X86-NEXT:    vmovd %eax, %xmm2
-; X86-NEXT:    vmovw %xmm0, %eax
-; X86-NEXT:    shll $16, %eax
-; X86-NEXT:    vmovd %eax, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %ecx
-; X86-NEXT:    vpextrw $1, %xmm1, %eax
-; X86-NEXT:    shll $16, %eax
-; X86-NEXT:    vmovd %eax, %xmm2
-; X86-NEXT:    vpextrw $1, %xmm0, %eax
-; X86-NEXT:    shll $16, %eax
-; X86-NEXT:    vmovd %eax, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %eax
-; X86-NEXT:    vpextrw $2, %xmm1, %edx
-; X86-NEXT:    shll $16, %edx
-; X86-NEXT:    vmovd %edx, %xmm2
-; X86-NEXT:    vpextrw $2, %xmm0, %edx
-; X86-NEXT:    shll $16, %edx
-; X86-NEXT:    vmovd %edx, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %edx
-; X86-NEXT:    vpextrw $3, %xmm1, %esi
-; X86-NEXT:    shll $16, %esi
-; X86-NEXT:    vmovd %esi, %xmm2
-; X86-NEXT:    vpextrw $3, %xmm0, %esi
-; X86-NEXT:    shll $16, %esi
-; X86-NEXT:    vmovd %esi, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %esi
-; X86-NEXT:    vpextrw $4, %xmm1, %edi
-; X86-NEXT:    shll $16, %edi
-; X86-NEXT:    vmovd %edi, %xmm2
-; X86-NEXT:    vpextrw $4, %xmm0, %edi
-; X86-NEXT:    shll $16, %edi
-; X86-NEXT:    vmovd %edi, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %ebx
-; X86-NEXT:    vpextrw $5, %xmm1, %edi
-; X86-NEXT:    shll $16, %edi
-; X86-NEXT:    vmovd %edi, %xmm2
-; X86-NEXT:    vpextrw $5, %xmm0, %edi
-; X86-NEXT:    shll $16, %edi
-; X86-NEXT:    vmovd %edi, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; X86-NEXT:    vmovw %xmm2, %edi
-; X86-NEXT:    vpextrw $6, %xmm1, %ebp
-; X86-NEXT:    shll $16, %ebp
-; X86-NEXT:    vmovd %ebp, %xmm2
-; X86-NEXT:    vpextrw $6, %xmm0, %ebp
-; X86-NEXT:    shll $16, %ebp
-; X86-NEXT:    vmovd %ebp, %xmm3
-; X86-NEXT:    vaddss %xmm2, %xmm3, %xmm3
-; X86-NEXT:    vmovw %ecx, %xmm2
-; X86-NEXT:    vcvtneps2bf16 %xmm3, %xmm3
-; X86-NEXT:    vmovw %xmm3, %ecx
-; X86-NEXT:    vmovw %ebx, %xmm3
-; X86-NEXT:    vpextrw $7, %xmm1, %ebx
-; X86-NEXT:    shll $16, %ebx
-; X86-NEXT:    vmovd %ebx, %xmm1
-; X86-NEXT:    vpextrw $7, %xmm0, %ebx
-; X86-NEXT:    shll $16, %ebx
-; X86-NEXT:    vmovd %ebx, %xmm0
-; X86-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; X86-NEXT:    vmovw %ecx, %xmm1
-; X86-NEXT:    vcvtneps2bf16 %xmm0, %xmm0
-; X86-NEXT:    vmovw %xmm0, %ecx
-; X86-NEXT:    vmovw %ecx, %xmm0
-; X86-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; X86-NEXT:    vmovw %edi, %xmm1
-; X86-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; X86-NEXT:    vmovw %edx, %xmm3
-; X86-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; X86-NEXT:    vmovw %esi, %xmm1
-; X86-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; X86-NEXT:    vmovw %eax, %xmm3
-; X86-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; X86-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; X86-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; X86-NEXT:    popl %esi
-; X86-NEXT:    popl %edi
-; X86-NEXT:    popl %ebx
-; X86-NEXT:    popl %ebp
+; X86-NEXT:    vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; X86-NEXT:    vpslld $16, %ymm1, %ymm1
+; X86-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; X86-NEXT:    vpslld $16, %ymm0, %ymm0
+; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; X86-NEXT:    vcvtneps2bf16 %ymm0, %xmm0
+; X86-NEXT:    vzeroupper
 ; X86-NEXT:    retl
 ;
 ; SSE2-LABEL: addv:
@@ -756,176 +668,26 @@ define <8 x bfloat> @addv(<8 x bfloat> %a, <8 x bfloat> %b) nounwind {
 ; SSE2-NEXT:    popq %rbp
 ; SSE2-NEXT:    retq
 ;
-; FP16-LABEL: addv:
-; FP16:       # %bb.0:
-; FP16-NEXT:    vmovw %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm2
-; FP16-NEXT:    vmovw %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm3
-; FP16-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; FP16-NEXT:    vcvtneps2bf16 %xmm2, %xmm2
-; FP16-NEXT:    vmovw %xmm2, %eax
-; FP16-NEXT:    vmovw %eax, %xmm2
-; FP16-NEXT:    vpextrw $1, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm3
-; FP16-NEXT:    vpextrw $1, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm4
-; FP16-NEXT:    vaddss %xmm3, %xmm4, %xmm3
-; FP16-NEXT:    vcvtneps2bf16 %xmm3, %xmm3
-; FP16-NEXT:    vmovw %xmm3, %eax
-; FP16-NEXT:    vmovw %eax, %xmm3
-; FP16-NEXT:    vpextrw $2, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm4
-; FP16-NEXT:    vpextrw $2, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm5
-; FP16-NEXT:    vaddss %xmm4, %xmm5, %xmm4
-; FP16-NEXT:    vcvtneps2bf16 %xmm4, %xmm4
-; FP16-NEXT:    vmovw %xmm4, %eax
-; FP16-NEXT:    vmovw %eax, %xmm4
-; FP16-NEXT:    vpextrw $3, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm5
-; FP16-NEXT:    vpextrw $3, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm6
-; FP16-NEXT:    vaddss %xmm5, %xmm6, %xmm5
-; FP16-NEXT:    vcvtneps2bf16 %xmm5, %xmm5
-; FP16-NEXT:    vmovw %xmm5, %eax
-; FP16-NEXT:    vmovw %eax, %xmm5
-; FP16-NEXT:    vpextrw $4, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm6
-; FP16-NEXT:    vpextrw $4, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm7
-; FP16-NEXT:    vaddss %xmm6, %xmm7, %xmm6
-; FP16-NEXT:    vcvtneps2bf16 %xmm6, %xmm6
-; FP16-NEXT:    vmovw %xmm6, %eax
-; FP16-NEXT:    vmovw %eax, %xmm6
-; FP16-NEXT:    vpextrw $5, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm7
-; FP16-NEXT:    vpextrw $5, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm8
-; FP16-NEXT:    vaddss %xmm7, %xmm8, %xmm7
-; FP16-NEXT:    vcvtneps2bf16 %xmm7, %xmm7
-; FP16-NEXT:    vmovw %xmm7, %eax
-; FP16-NEXT:    vmovw %eax, %xmm7
-; FP16-NEXT:    vpextrw $6, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm8
-; FP16-NEXT:    vpextrw $6, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm9
-; FP16-NEXT:    vaddss %xmm8, %xmm9, %xmm8
-; FP16-NEXT:    vcvtneps2bf16 %xmm8, %xmm8
-; FP16-NEXT:    vmovw %xmm8, %eax
-; FP16-NEXT:    vmovw %eax, %xmm8
-; FP16-NEXT:    vpextrw $7, %xmm1, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm1
-; FP16-NEXT:    vpextrw $7, %xmm0, %eax
-; FP16-NEXT:    shll $16, %eax
-; FP16-NEXT:    vmovd %eax, %xmm0
-; FP16-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; FP16-NEXT:    vcvtneps2bf16 %xmm0, %xmm0
-; FP16-NEXT:    vmovw %xmm0, %eax
-; FP16-NEXT:    vmovw %eax, %xmm0
-; FP16-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; FP16-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
-; FP16-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; FP16-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; FP16-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; FP16-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; FP16-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; FP16-NEXT:    retq
+; F16-LABEL: addv:
+; F16:       # %bb.0:
+; F16-NEXT:    vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; F16-NEXT:    vpslld $16, %ymm1, %ymm1
+; F16-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; F16-NEXT:    vpslld $16, %ymm0, %ymm0
+; F16-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; F16-NEXT:    vcvtneps2bf16 %ymm0, %xmm0
+; F16-NEXT:    vzeroupper
+; F16-NEXT:    retq
 ;
 ; AVXNC-LABEL: addv:
 ; AVXNC:       # %bb.0:
-; AVXNC-NEXT:    vpextrw $7, %xmm1, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm2
-; AVXNC-NEXT:    vpextrw $7, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %eax
-; AVXNC-NEXT:    vpextrw $6, %xmm1, %ecx
-; AVXNC-NEXT:    shll $16, %ecx
-; AVXNC-NEXT:    vmovd %ecx, %xmm2
-; AVXNC-NEXT:    vpextrw $6, %xmm0, %ecx
-; AVXNC-NEXT:    shll $16, %ecx
-; AVXNC-NEXT:    vmovd %ecx, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %ecx
-; AVXNC-NEXT:    vpextrw $5, %xmm1, %edx
-; AVXNC-NEXT:    shll $16, %edx
-; AVXNC-NEXT:    vmovd %edx, %xmm2
-; AVXNC-NEXT:    vpextrw $5, %xmm0, %edx
-; AVXNC-NEXT:    shll $16, %edx
-; AVXNC-NEXT:    vmovd %edx, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %edx
-; AVXNC-NEXT:    vpextrw $4, %xmm1, %esi
-; AVXNC-NEXT:    shll $16, %esi
-; AVXNC-NEXT:    vmovd %esi, %xmm2
-; AVXNC-NEXT:    vpextrw $4, %xmm0, %esi
-; AVXNC-NEXT:    shll $16, %esi
-; AVXNC-NEXT:    vmovd %esi, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %esi
-; AVXNC-NEXT:    vpextrw $3, %xmm1, %edi
-; AVXNC-NEXT:    shll $16, %edi
-; AVXNC-NEXT:    vmovd %edi, %xmm2
-; AVXNC-NEXT:    vpextrw $3, %xmm0, %edi
-; AVXNC-NEXT:    shll $16, %edi
-; AVXNC-NEXT:    vmovd %edi, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %edi
-; AVXNC-NEXT:    vpextrw $2, %xmm1, %r8d
-; AVXNC-NEXT:    shll $16, %r8d
-; AVXNC-NEXT:    vmovd %r8d, %xmm2
-; AVXNC-NEXT:    vpextrw $2, %xmm0, %r8d
-; AVXNC-NEXT:    shll $16, %r8d
-; AVXNC-NEXT:    vmovd %r8d, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm2, %r8d
-; AVXNC-NEXT:    vpextrw $1, %xmm1, %r9d
-; AVXNC-NEXT:    shll $16, %r9d
-; AVXNC-NEXT:    vmovd %r9d, %xmm2
-; AVXNC-NEXT:    vpextrw $1, %xmm0, %r9d
-; AVXNC-NEXT:    shll $16, %r9d
-; AVXNC-NEXT:    vmovd %r9d, %xmm3
-; AVXNC-NEXT:    vaddss %xmm2, %xmm3, %xmm2
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm1, %r9d
-; AVXNC-NEXT:    shll $16, %r9d
-; AVXNC-NEXT:    vmovd %r9d, %xmm1
-; AVXNC-NEXT:    vmovd %xmm0, %r9d
-; AVXNC-NEXT:    shll $16, %r9d
-; AVXNC-NEXT:    vmovd %r9d, %xmm0
-; AVXNC-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; AVXNC-NEXT:    {vex} vcvtneps2bf16 %xmm0, %xmm0
-; AVXNC-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; AVXNC-NEXT:    vpinsrw $2, %r8d, %xmm0, %xmm0
-; AVXNC-NEXT:    vpinsrw $3, %edi, %xmm0, %xmm0
-; AVXNC-NEXT:    vpinsrw $4, %esi, %xmm0, %xmm0
-; AVXNC-NEXT:    vpinsrw $5, %edx, %xmm0, %xmm0
-; AVXNC-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0
-; AVXNC-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; AVXNC-NEXT:    vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVXNC-NEXT:    vpslld $16, %ymm1, %ymm1
+; AVXNC-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVXNC-NEXT:    vpslld $16, %ymm0, %ymm0
+; AVXNC-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; AVXNC-NEXT:    {vex} vcvtneps2bf16 %ymm0, %xmm0
+; AVXNC-NEXT:    vzeroupper
 ; AVXNC-NEXT:    retq
   %add = fadd <8 x bfloat> %a, %b
   ret <8 x bfloat> %add

From 27ce513788e15c95f8c623566fbfca9304bb1844 Mon Sep 17 00:00:00 2001
From: Corbin Robeck <corbin.robeck@amd.com>
Date: Thu, 11 Apr 2024 22:05:01 -0400
Subject: [PATCH 208/886] [AMDGPU] Add Clang builtins for amdgcn s_ttrace
 intrinsics (#88076)

---
 clang/include/clang/Basic/BuiltinsAMDGPU.def  |  2 ++
 .../CodeGenOpenCL/builtins-amdgcn-gfx12.cl    | 22 +++++++++++++++++++
 llvm/include/llvm/IR/IntrinsicsAMDGPU.td      |  3 +++
 3 files changed, 27 insertions(+)

diff --git a/clang/include/clang/Basic/BuiltinsAMDGPU.def b/clang/include/clang/Basic/BuiltinsAMDGPU.def
index c660582cc98e6..3e21a2fe2ac6b 100644
--- a/clang/include/clang/Basic/BuiltinsAMDGPU.def
+++ b/clang/include/clang/Basic/BuiltinsAMDGPU.def
@@ -61,6 +61,7 @@ BUILTIN(__builtin_amdgcn_s_waitcnt, "vIi", "n")
 BUILTIN(__builtin_amdgcn_s_sendmsg, "vIiUi", "n")
 BUILTIN(__builtin_amdgcn_s_sendmsghalt, "vIiUi", "n")
 BUILTIN(__builtin_amdgcn_s_barrier, "v", "n")
+BUILTIN(__builtin_amdgcn_s_ttracedata, "vi", "n")
 BUILTIN(__builtin_amdgcn_wave_barrier, "v", "n")
 BUILTIN(__builtin_amdgcn_sched_barrier, "vIi", "n")
 BUILTIN(__builtin_amdgcn_sched_group_barrier, "vIiIiIi", "n")
@@ -267,6 +268,7 @@ TARGET_BUILTIN(__builtin_amdgcn_dot4_f32_bf8_bf8, "fUiUif", "nc", "dot11-insts")
 TARGET_BUILTIN(__builtin_amdgcn_permlane16, "UiUiUiUiUiIbIb", "nc", "gfx10-insts")
 TARGET_BUILTIN(__builtin_amdgcn_permlanex16, "UiUiUiUiUiIbIb", "nc", "gfx10-insts")
 TARGET_BUILTIN(__builtin_amdgcn_mov_dpp8, "UiUiIUi", "nc", "gfx10-insts")
+TARGET_BUILTIN(__builtin_amdgcn_s_ttracedata_imm, "vIs", "n", "gfx10-insts")
 
 //===----------------------------------------------------------------------===//
 // Raytracing builtins.
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx12.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx12.cl
index ebd367bba0cdc..26c0ee4830623 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx12.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx12.cl
@@ -235,3 +235,25 @@ unsigned test_s_get_barrier_state(int a)
   unsigned State = __builtin_amdgcn_s_get_barrier_state(a);
   return State;
 }
+
+// CHECK-LABEL: @test_s_ttracedata(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    call void @llvm.amdgcn.s.ttracedata(i32 1)
+// CHECK-NEXT:    ret void
+//
+void test_s_ttracedata()
+{
+  __builtin_amdgcn_s_ttracedata(1);
+}
+
+// CHECK-LABEL: @test_s_ttracedata_imm(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    call void @llvm.amdgcn.s.ttracedata.imm(i16 1)
+// CHECK-NEXT:    ret void
+//
+void test_s_ttracedata_imm()
+{
+  __builtin_amdgcn_s_ttracedata_imm(1);
+}
+
+
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 6bbc13f1de86e..ee9a5d7a34398 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -1887,9 +1887,12 @@ def int_amdgcn_s_setprio :
                                 IntrHasSideEffects]>;
 
 def int_amdgcn_s_ttracedata :
+  ClangBuiltin<"__builtin_amdgcn_s_ttracedata">,
   DefaultAttrsIntrinsic<[], [llvm_i32_ty],
                         [IntrNoMem, IntrHasSideEffects]>;
+
 def int_amdgcn_s_ttracedata_imm :
+  ClangBuiltin<"__builtin_amdgcn_s_ttracedata_imm">,
   DefaultAttrsIntrinsic<[], [llvm_i16_ty],
                         [IntrNoMem, IntrHasSideEffects, ImmArg<ArgIndex<0>>]>;
 

From 6b6f272f355a2cbc753e9057dd7aefb1a74a0409 Mon Sep 17 00:00:00 2001
From: Feng Zou <feng.zou@intel.com>
Date: Fri, 12 Apr 2024 10:13:56 +0800
Subject: [PATCH 209/886] [DAGCombiner] Require same type of splat & element
 for build_vector (#88284)

Only allow to change build_vector to concat_vector when the splat type
and vector element type is same. It's to fix assertion of failing to
bitcast types of different sizes.
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |  9 ++++---
 llvm/test/CodeGen/X86/buildvec-bitcast.ll     | 25 +++++++++++++++++++
 2 files changed, 31 insertions(+), 3 deletions(-)
 create mode 100644 llvm/test/CodeGen/X86/buildvec-bitcast.ll

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 95c1cde0b9347..24f50b87c4cf2 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -23428,7 +23428,10 @@ SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
   // If X is a build_vector itself, the concat can become a larger build_vector.
   // TODO: Maybe this is useful for non-splat too?
   if (!LegalOperations) {
-    if (SDValue Splat = cast<BuildVectorSDNode>(N)->getSplatValue()) {
+    SDValue Splat = cast<BuildVectorSDNode>(N)->getSplatValue();
+    // Only change build_vector to a concat_vector if the splat value type is
+    // same as the vector element type.
+    if (Splat && Splat.getValueType() == VT.getVectorElementType()) {
       Splat = peekThroughBitcasts(Splat);
       EVT SrcVT = Splat.getValueType();
       if (SrcVT.isVector()) {
@@ -23437,8 +23440,8 @@ SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
                                      SrcVT.getVectorElementType(), NumElts);
         if (!LegalTypes || TLI.isTypeLegal(NewVT)) {
           SmallVector<SDValue, 8> Ops(N->getNumOperands(), Splat);
-          SDValue Concat = DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(N),
-                                       NewVT, Ops);
+          SDValue Concat =
+              DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(N), NewVT, Ops);
           return DAG.getBitcast(VT, Concat);
         }
       }
diff --git a/llvm/test/CodeGen/X86/buildvec-bitcast.ll b/llvm/test/CodeGen/X86/buildvec-bitcast.ll
new file mode 100644
index 0000000000000..980fd2dee0915
--- /dev/null
+++ b/llvm/test/CodeGen/X86/buildvec-bitcast.ll
@@ -0,0 +1,25 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=x86_64 -mattr=avx512bw | FileCheck %s
+
+; Verify that the DAGCombiner doesn't change build_vector to concat_vectors if
+; the vector element type is different than splat type. The example here:
+;   v8i1 = build_vector (i8 (bitcast (v8i1 X))), ..., (i8 (bitcast (v8i1 X))))
+
+define <8 x i1> @foo(<8 x i1> %mask.i1) {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+entry:
+  %0 = and <8 x i1> %mask.i1, <i1 true, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false>
+  %1 = bitcast <8 x i1> %0 to i8
+  %2 = icmp ne i8 %1, 0
+  %insert54 = insertelement <8 x i1> zeroinitializer, i1 %2, i64 0
+  %splat55 = shufflevector <8 x i1> %insert54, <8 x i1> zeroinitializer, <8 x i32> zeroinitializer
+  %3 = and <8 x i1> %0, %splat55
+  br label %end
+
+end:                           ; preds = %entry
+  %4 = select <8 x i1> %3, <8 x i1> zeroinitializer, <8 x i1> zeroinitializer
+  ret <8 x i1> %4
+}

From 3a2817749eb8a2210320aa551fa20a5f9c8c32e3 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Thu, 11 Apr 2024 19:47:58 -0700
Subject: [PATCH 210/886] [llvm-exegesis] Let ArrayRef infer the array size
 (NFC) (#88241)

---
 llvm/tools/llvm-exegesis/lib/X86/Target.cpp | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/llvm/tools/llvm-exegesis/lib/X86/Target.cpp b/llvm/tools/llvm-exegesis/lib/X86/Target.cpp
index 31bec2545926b..dbb2bf409316e 100644
--- a/llvm/tools/llvm-exegesis/lib/X86/Target.cpp
+++ b/llvm/tools/llvm-exegesis/lib/X86/Target.cpp
@@ -769,11 +769,9 @@ class ExegesisX86Target : public ExegesisTarget {
 
   ArrayRef<unsigned> getUnavailableRegisters() const override {
     if (DisableUpperSSERegisters)
-      return ArrayRef(kUnavailableRegistersSSE,
-                      sizeof(kUnavailableRegistersSSE) /
-                          sizeof(kUnavailableRegistersSSE[0]));
+      return ArrayRef(kUnavailableRegistersSSE);
 
-    return ArrayRef(kUnavailableRegisters, std::size(kUnavailableRegisters));
+    return ArrayRef(kUnavailableRegisters);
   }
 
   bool allowAsBackToBack(const Instruction &Instr) const override {

From 334e07f18e6f2fbc84579f7cac3cdd29d0d7cce0 Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen@sifive.com>
Date: Fri, 12 Apr 2024 10:59:06 +0800
Subject: [PATCH 211/886] [RISCV] Disallow target attribute use in
 multiversioning (#85899)

For RISC-V target only `target_clones` and `target_version` can enable
function multiversion(FMV).

This patch make target attribute trigger redefinition instead of emit
FMV.

Here is spec
https://github.com/riscv-non-isa/riscv-c-api-doc/blob/master/riscv-c-api.md#__attribute__targetattr-string
---
 clang/lib/Sema/SemaDecl.cpp         | 8 +++++++-
 clang/test/Sema/attr-target-riscv.c | 6 ++++++
 2 files changed, 13 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/Sema/attr-target-riscv.c

diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index a4699d6ba2c73..691e170cd0514 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -11911,8 +11911,14 @@ static bool CheckMultiVersionFunction(Sema &S, FunctionDecl *NewFD,
     return false;
   }
 
+  const llvm::Triple &T = S.getASTContext().getTargetInfo().getTriple();
+
   // Target attribute on AArch64 is not used for multiversioning
-  if (NewTA && S.getASTContext().getTargetInfo().getTriple().isAArch64())
+  if (NewTA && T.isAArch64())
+    return false;
+
+  // Target attribute on RISCV is not used for multiversioning
+  if (NewTA && T.isRISCV())
     return false;
 
   if (!OldDecl || !OldDecl->getAsFunction() ||
diff --git a/clang/test/Sema/attr-target-riscv.c b/clang/test/Sema/attr-target-riscv.c
new file mode 100644
index 0000000000000..ed4e2915d6c6e
--- /dev/null
+++ b/clang/test/Sema/attr-target-riscv.c
@@ -0,0 +1,6 @@
+// RUN: %clang_cc1 -triple riscv64-linux-gnu -target-feature +i -fsyntax-only -verify -std=c2x %s
+
+//expected-note@+1 {{previous definition is here}}
+int __attribute__((target("arch=rv64g"))) foo(void) { return 0; }
+//expected-error@+1 {{redefinition of 'foo'}}
+int __attribute__((target("arch=rv64gc"))) foo(void) { return 0; }

From c24efffabbf96c7a138439bb8e219850c4d78887 Mon Sep 17 00:00:00 2001
From: Cyndy Ishida <cyndy_ishida@apple.com>
Date: Thu, 11 Apr 2024 20:01:55 -0700
Subject: [PATCH 212/886] [InstallAPI] Handle zippered frameworks (#88205)

A zippered framework is a single framework that can be loaded in both
macOS and macatalyst processes. Broadly to InstallAPI, it means the same
interface can represent two separate platforms.

A dylib's symbol table does not distinguish between macOS/macCatalyst.
  `InstallAPI` provides the ability for the tbd file to distinct
symbols between them.
The verifier handles this special logic by tracking all unavailable and
obsoleted APIs in this context and checking against those when
determining dylib symbols with no matching declaration.

* If there exists an available decl for either platform, do not warn.
* If there is no available decl, emit a diagnostic and print the source
location for both decls.
---
 .../clang/Basic/DiagnosticInstallAPIKinds.td  |   4 +
 .../include/clang/InstallAPI/DylibVerifier.h  |  32 +-
 clang/lib/InstallAPI/DylibVerifier.cpp        |  98 ++-
 .../Inputs/MacOSX13.0.sdk/SDKSettings.json    |  19 +
 .../test/InstallAPI/diagnostics-zippered.test | 765 ++++++++++++++++++
 .../InstallAPI/driver-invalid-options.test    |   7 +
 clang/tools/clang-installapi/Options.cpp      |  52 +-
 clang/tools/clang-installapi/Options.h        |   3 +
 8 files changed, 964 insertions(+), 16 deletions(-)
 create mode 100644 clang/test/InstallAPI/Inputs/MacOSX13.0.sdk/SDKSettings.json
 create mode 100644 clang/test/InstallAPI/diagnostics-zippered.test

diff --git a/clang/include/clang/Basic/DiagnosticInstallAPIKinds.td b/clang/include/clang/Basic/DiagnosticInstallAPIKinds.td
index 0a477da7186b0..396bff0146a37 100644
--- a/clang/include/clang/Basic/DiagnosticInstallAPIKinds.td
+++ b/clang/include/clang/Basic/DiagnosticInstallAPIKinds.td
@@ -20,6 +20,10 @@ def warn_no_such_excluded_header_file : Warning<"no such excluded %select{public
 def warn_glob_did_not_match: Warning<"glob '%0' did not match any header file">, InGroup<InstallAPIViolation>;
 def err_no_such_umbrella_header_file : Error<"%select{public|private|project}1 umbrella header file not found in input: '%0'">;
 def err_cannot_find_reexport : Error<"cannot find re-exported %select{framework|library}0: '%1'">;
+def err_no_matching_target : Error<"no matching target found for target variant '%0'">;
+def err_unsupported_vendor : Error<"vendor '%0' is not supported: '%1'">;
+def err_unsupported_environment : Error<"environment '%0' is not supported: '%1'">;
+def err_unsupported_os : Error<"os '%0' is not supported: '%1'">;
 } // end of command line category.
 
 let CategoryName = "Verification" in {
diff --git a/clang/include/clang/InstallAPI/DylibVerifier.h b/clang/include/clang/InstallAPI/DylibVerifier.h
index a3df25f10de4b..31de212fc423a 100644
--- a/clang/include/clang/InstallAPI/DylibVerifier.h
+++ b/clang/include/clang/InstallAPI/DylibVerifier.h
@@ -28,6 +28,16 @@ enum class VerificationMode {
 using LibAttrs = llvm::StringMap<ArchitectureSet>;
 using ReexportedInterfaces = llvm::SmallVector<llvm::MachO::InterfaceFile, 8>;
 
+// Pointers to information about a zippered declaration used for
+// querying and reporting violations against different
+// declarations that all map to the same symbol.
+struct ZipperedDeclSource {
+  const FrontendAttrs *FA;
+  clang::SourceManager *SrcMgr;
+  Target T;
+};
+using ZipperedDeclSources = std::vector<ZipperedDeclSource>;
+
 /// Service responsible to tracking state of verification across the
 /// lifetime of InstallAPI.
 /// As declarations are collected during AST traversal, they are
@@ -68,10 +78,10 @@ class DylibVerifier : llvm::MachO::RecordVisitor {
   DylibVerifier() = default;
 
   DylibVerifier(llvm::MachO::Records &&Dylib, ReexportedInterfaces &&Reexports,
-                DiagnosticsEngine *Diag, VerificationMode Mode, bool Demangle,
-                StringRef DSYMPath)
+                DiagnosticsEngine *Diag, VerificationMode Mode, bool Zippered,
+                bool Demangle, StringRef DSYMPath)
       : Dylib(std::move(Dylib)), Reexports(std::move(Reexports)), Mode(Mode),
-        Demangle(Demangle), DSYMPath(DSYMPath),
+        Zippered(Zippered), Demangle(Demangle), DSYMPath(DSYMPath),
         Exports(std::make_unique<SymbolSet>()), Ctx(VerifierContext{Diag}) {}
 
   Result verify(GlobalRecord *R, const FrontendAttrs *FA);
@@ -118,6 +128,15 @@ class DylibVerifier : llvm::MachO::RecordVisitor {
   /// symbols should be omitted from the text-api file.
   bool shouldIgnoreReexport(const Record *R, SymbolContext &SymCtx) const;
 
+  // Ignore and omit unavailable symbols in zippered libraries.
+  bool shouldIgnoreZipperedAvailability(const Record *R, SymbolContext &SymCtx);
+
+  // Check if an internal declaration in zippered library has an
+  // external declaration for a different platform. This results
+  // in the symbol being in a "seperate" platform slice.
+  bool shouldIgnoreInternalZipperedSymbol(const Record *R,
+                                          const SymbolContext &SymCtx) const;
+
   /// Compare the visibility declarations to the linkage of symbol found in
   /// dylib.
   Result compareVisibility(const Record *R, SymbolContext &SymCtx,
@@ -173,6 +192,9 @@ class DylibVerifier : llvm::MachO::RecordVisitor {
   // Controls what class of violations to report.
   VerificationMode Mode = VerificationMode::Invalid;
 
+  // Library is zippered.
+  bool Zippered = false;
+
   // Attempt to demangle when reporting violations.
   bool Demangle = false;
 
@@ -182,6 +204,10 @@ class DylibVerifier : llvm::MachO::RecordVisitor {
   // Valid symbols in final text file.
   std::unique_ptr<SymbolSet> Exports = std::make_unique<SymbolSet>();
 
+  // Unavailable or obsoleted declarations for a zippered library.
+  // These are cross referenced against symbols in the dylib.
+  llvm::StringMap<ZipperedDeclSources> DeferredZipperedSymbols;
+
   // Track current state of verification while traversing AST.
   VerifierContext Ctx;
 
diff --git a/clang/lib/InstallAPI/DylibVerifier.cpp b/clang/lib/InstallAPI/DylibVerifier.cpp
index 4fa2d4e9292c7..84d9b5892e88d 100644
--- a/clang/lib/InstallAPI/DylibVerifier.cpp
+++ b/clang/lib/InstallAPI/DylibVerifier.cpp
@@ -176,7 +176,13 @@ void DylibVerifier::addSymbol(const Record *R, SymbolContext &SymCtx,
 
 bool DylibVerifier::shouldIgnoreObsolete(const Record *R, SymbolContext &SymCtx,
                                          const Record *DR) {
-  return SymCtx.FA->Avail.isObsoleted();
+  if (!SymCtx.FA->Avail.isObsoleted())
+    return false;
+
+  if (Zippered)
+    DeferredZipperedSymbols[SymCtx.SymbolName].emplace_back(ZipperedDeclSource{
+        SymCtx.FA, &Ctx.Diag->getSourceManager(), Ctx.Target});
+  return true;
 }
 
 bool DylibVerifier::shouldIgnoreReexport(const Record *R,
@@ -195,6 +201,28 @@ bool DylibVerifier::shouldIgnoreReexport(const Record *R,
   return false;
 }
 
+bool DylibVerifier::shouldIgnoreInternalZipperedSymbol(
+    const Record *R, const SymbolContext &SymCtx) const {
+  if (!Zippered)
+    return false;
+
+  return Exports->findSymbol(SymCtx.Kind, SymCtx.SymbolName,
+                             SymCtx.ObjCIFKind) != nullptr;
+}
+
+bool DylibVerifier::shouldIgnoreZipperedAvailability(const Record *R,
+                                                     SymbolContext &SymCtx) {
+  if (!(Zippered && SymCtx.FA->Avail.isUnavailable()))
+    return false;
+
+  // Collect source location incase there is an exported symbol to diagnose
+  // during `verifyRemainingSymbols`.
+  DeferredZipperedSymbols[SymCtx.SymbolName].emplace_back(
+      ZipperedDeclSource{SymCtx.FA, SourceManagers.back().get(), Ctx.Target});
+
+  return true;
+}
+
 bool DylibVerifier::compareObjCInterfaceSymbols(const Record *R,
                                                 SymbolContext &SymCtx,
                                                 const ObjCInterfaceRecord *DR) {
@@ -294,6 +322,9 @@ DylibVerifier::Result DylibVerifier::compareVisibility(const Record *R,
     if (shouldIgnorePrivateExternAttr(SymCtx.FA->D))
       return Result::Ignore;
 
+    if (shouldIgnoreInternalZipperedSymbol(R, SymCtx))
+      return Result::Ignore;
+
     unsigned ID;
     Result Outcome;
     if (Mode == VerificationMode::ErrorsAndWarnings) {
@@ -321,6 +352,9 @@ DylibVerifier::Result DylibVerifier::compareAvailability(const Record *R,
   if (!SymCtx.FA->Avail.isUnavailable())
     return Result::Valid;
 
+  if (shouldIgnoreZipperedAvailability(R, SymCtx))
+    return Result::Ignore;
+
   const bool IsDeclAvailable = SymCtx.FA->Avail.isUnavailable();
 
   switch (Mode) {
@@ -588,13 +622,58 @@ void DylibVerifier::visitSymbolInDylib(const Record &R, SymbolContext &SymCtx) {
     }
   }
 
+  const bool IsLinkerSymbol = SymbolName.starts_with("$ld$");
+
+  if (R.isVerified()) {
+    // Check for unavailable symbols.
+    // This should only occur in the zippered case where we ignored
+    // availability until all headers have been parsed.
+    auto It = DeferredZipperedSymbols.find(SymCtx.SymbolName);
+    if (It == DeferredZipperedSymbols.end()) {
+      updateState(Result::Valid);
+      return;
+    }
+
+    ZipperedDeclSources Locs;
+    for (const ZipperedDeclSource &ZSource : It->second) {
+      if (ZSource.FA->Avail.isObsoleted()) {
+        updateState(Result::Ignore);
+        return;
+      }
+      if (ZSource.T.Arch != Ctx.Target.Arch)
+        continue;
+      Locs.emplace_back(ZSource);
+    }
+    assert(Locs.size() == 2 && "Expected two decls for zippered symbol");
+
+    // Print violating declarations per platform.
+    for (const ZipperedDeclSource &ZSource : Locs) {
+      unsigned DiagID = 0;
+      if (Mode == VerificationMode::Pedantic || IsLinkerSymbol) {
+        updateState(Result::Invalid);
+        DiagID = diag::err_header_availability_mismatch;
+      } else if (Mode == VerificationMode::ErrorsAndWarnings) {
+        updateState(Result::Ignore);
+        DiagID = diag::warn_header_availability_mismatch;
+      } else {
+        updateState(Result::Ignore);
+        return;
+      }
+      // Bypass emitDiag banner and print the target everytime.
+      Ctx.Diag->setSourceManager(ZSource.SrcMgr);
+      Ctx.Diag->Report(diag::warn_target) << getTargetTripleName(ZSource.T);
+      Ctx.Diag->Report(ZSource.FA->Loc, DiagID)
+          << getAnnotatedName(&R, SymCtx) << ZSource.FA->Avail.isUnavailable()
+          << ZSource.FA->Avail.isUnavailable();
+    }
+    return;
+  }
+
   if (shouldIgnoreCpp(SymbolName, R.isWeakDefined())) {
     updateState(Result::Valid);
     return;
   }
 
-  const bool IsLinkerSymbol = SymbolName.starts_with("$ld$");
-
   // All checks at this point classify as some kind of violation.
   // The different verification modes dictate whether they are reported to the
   // user.
@@ -647,8 +726,6 @@ void DylibVerifier::visitSymbolInDylib(const Record &R, SymbolContext &SymCtx) {
 }
 
 void DylibVerifier::visitGlobal(const GlobalRecord &R) {
-  if (R.isVerified())
-    return;
   SymbolContext SymCtx;
   SimpleSymbol Sym = parseSymbol(R.getName());
   SymCtx.SymbolName = Sym.Name;
@@ -658,8 +735,6 @@ void DylibVerifier::visitGlobal(const GlobalRecord &R) {
 
 void DylibVerifier::visitObjCIVar(const ObjCIVarRecord &R,
                                   const StringRef Super) {
-  if (R.isVerified())
-    return;
   SymbolContext SymCtx;
   SymCtx.SymbolName = ObjCIVarRecord::createScopedName(Super, R.getName());
   SymCtx.Kind = EncodeKind::ObjectiveCInstanceVariable;
@@ -679,8 +754,6 @@ void DylibVerifier::accumulateSrcLocForDylibSymbols() {
 }
 
 void DylibVerifier::visitObjCInterface(const ObjCInterfaceRecord &R) {
-  if (R.isVerified())
-    return;
   SymbolContext SymCtx;
   SymCtx.SymbolName = R.getName();
   SymCtx.ObjCIFKind = assignObjCIFSymbolKind(&R);
@@ -713,9 +786,12 @@ DylibVerifier::Result DylibVerifier::verifyRemainingSymbols() {
 
   DWARFContext DWARFInfo;
   DWARFCtx = &DWARFInfo;
-  Ctx.DiscoveredFirstError = false;
-  Ctx.PrintArch = true;
+  Ctx.Target = Target(Architecture::AK_unknown, PlatformType::PLATFORM_UNKNOWN);
   for (std::shared_ptr<RecordsSlice> Slice : Dylib) {
+    if (Ctx.Target.Arch == Slice->getTarget().Arch)
+      continue;
+    Ctx.DiscoveredFirstError = false;
+    Ctx.PrintArch = true;
     Ctx.Target = Slice->getTarget();
     Ctx.DylibSlice = Slice.get();
     Slice->visit(*this);
diff --git a/clang/test/InstallAPI/Inputs/MacOSX13.0.sdk/SDKSettings.json b/clang/test/InstallAPI/Inputs/MacOSX13.0.sdk/SDKSettings.json
new file mode 100644
index 0000000000000..258d8288fc6b4
--- /dev/null
+++ b/clang/test/InstallAPI/Inputs/MacOSX13.0.sdk/SDKSettings.json
@@ -0,0 +1,19 @@
+{
+  "DefaultVariant": "macos", "DisplayName": "macOS 13",
+  "Version": "13.0",
+  "MaximumDeploymentTarget": "13.0.99",
+  "PropertyConditionFallbackNames": [], "VersionMap": {
+    "iOSMac_macOS": {
+      "16.1": "13.0",
+      "15.0": "12.0",
+      "13.1": "10.15",
+      "14.0": "11.0"
+    },
+    "macOS_iOSMac": {
+      "13.0": "16.1",
+      "12.0": "15.0",
+      "11.0": "14.0",
+      "10.15": "13.1"
+    }
+  }
+}
diff --git a/clang/test/InstallAPI/diagnostics-zippered.test b/clang/test/InstallAPI/diagnostics-zippered.test
new file mode 100644
index 0000000000000..59fd0bb673a3d
--- /dev/null
+++ b/clang/test/InstallAPI/diagnostics-zippered.test
@@ -0,0 +1,765 @@
+; RUN: rm -rf %t
+; RUN: split-file %s %t
+; RUN: cp -r %S/Inputs/Foundation/Foundation.framework %t/System/Library/Frameworks/
+; RUN: sed -e "s|DSTROOT|%/t|g" %t/inputs.json.in > %t/inputs.json
+; RUN: yaml2obj %t/Mismatch.yaml -o %t/System/Library/Frameworks/Mismatch.framework/Mismatch
+
+; RUN: not clang-installapi --target=arm64-apple-macos13 \
+; RUN: --target=x86_64-apple-macos13 -darwin-target-variant x86_64-apple-ios16.0-macabi \
+; RUN: -darwin-target-variant arm64-apple-ios16.0-macabi \
+; RUN: -F %t/System/Library/Frameworks -isysroot %S/Inputs/MacOSX13.0.sdk \
+; RUN: -install_name /System/Library/Frameworks/Mismatch.framework/Versions/A/Mismatch \
+; RUN: -current_version 1 -compatibility_version 1 %t/inputs.json -o %t/Mismatch.tbd \
+; RUN: --verify-against=%t/System/Library/Frameworks/Mismatch.framework/Mismatch \
+; RUN: --verify-mode=Pedantic 2>&1 | FileCheck %s 
+
+; CHECK:      warning: violations found for x86_64-apple-macos13
+; CHECK:      error: declaration has external linkage, but dynamic library doesn't have symbol 'foo_arch_x86'
+; CHECK-NEXT: extern int foo_arch_x86();
+
+; CHECK:      error: symbol exported in dynamic library, but marked hidden in declaration 'baz'
+; CHECK-NEXT: int baz() __attribute__((visibility("hidden")));
+
+; CHECK-NOT:  inlinedFunc
+
+; CHECK:      error: declaration has external linkage, but symbol has internal linkage in dynamic library 'bar'
+; CHECK-NEXT: extern int bar();
+
+; CHECK:      error: declaration has external linkage, but dynamic library doesn't have symbol 'NSFoo'
+; CHECK-NEXT: @interface NSFoo
+
+; CHECK:      error: declaration has external linkage, but dynamic library doesn't have symbol 'NSFoo.nsint'
+; CHECK-NEXT: int nsint;
+
+; CHECK:      error: declaration has external linkage, but dynamic library doesn't have symbol 'NSFooExcept'
+; CHECK-NEXT: @interface NSFooExcept
+ 
+; CHECK:      warning: violations found for arm64-apple-macos13
+; CHECK:      error: declaration has external linkage, but dynamic library doesn't have symbol 'foo_arch_arm'
+; CHECK-NEXT: extern int foo_arch_arm();
+ 
+; CHECK:      warning: violations found for x86_64 
+; CHECK-NEXT: error: no declaration found for exported symbol '_fooBar' in dynamic library
+ 
+; CHECK:      warning: violations found for arm64 
+; CHECK-NEXT: error: no declaration found for exported symbol '_fooBar' in dynamic library
+; CHECK:      warning: violations found for arm64-apple-macos13
+; CHECK-NEXT: error: declaration 'unavailableSymbol' is marked unavailable, but symbol is exported in dynamic library
+; CHECK-NEXT: extern int unavailableSymbol
+; CHECK:      warning: violations found for arm64-apple-ios16.0-macabi
+; CHECK-NEXT: error: declaration 'unavailableSymbol' is marked unavailable, but symbol is exported in dynamic library
+; CHECK-NEXT: extern int unavailableSymbol
+
+
+;--- System/Library/Frameworks/Mismatch.framework/Headers/Mismatch.h
+#if defined(__x86_64__)
+extern int foo_arch_x86();
+#else
+extern int foo_arch_arm();
+#endif
+
+int foo();
+int baz() __attribute__((visibility("hidden")));
+
+__attribute__((visibility("hidden"))) inline int inlinedFunc() { return 1; }
+
+;--- System/Library/Frameworks/Mismatch.framework/PrivateHeaders/Mismatch_Private.h
+#include <Foundation/Foundation.h>
+extern int bar();
+
+@interface NSFoo : NSObject {
+  int nsint;
+}
+@end
+
+__attribute__((objc_exception))
+@interface NSFooExcept : NSObject
+@end
+
+extern int unavailableSymbol __attribute__((
+    availability(macosx, unavailable), availability(macCatalyst, unavailable)));
+
+
+// Generated from source:
+// *** 
+//     int inlinedFunc() { return 1; }
+//     int foo() { return 1; }
+//     __attribute__((visibility("hidden"))) int bar() { return 1; }
+//     int baz() { return 1; }
+//     int fooBar = 1;
+//     int unavailableSymbol = 1;
+// *** 
+// xcrun -sdk macosx clang -target x86_64-apple-macos13 --target-variant=x86_64-apple-ios16-macabi -dynamiclib
+// xcrun -sdk macosx clang -target arm64-apple-macos13 --target-variant=arm64-apple-ios16-macabi -dynamiclib
+// then lipo'ed together. 
+
+;--- Mismatch.yaml
+--- !fat-mach-o
+FatHeader:
+  magic:           0xCAFEBABE
+  nfat_arch:       2
+FatArchs:
+  - cputype:         0x1000007
+    cpusubtype:      0x3
+    offset:          0x1000
+    size:            20768
+    align:           12
+  - cputype:         0x100000C
+    cpusubtype:      0x0
+    offset:          0x8000
+    size:            49984
+    align:           14
+Slices:
+  - !mach-o
+    FileHeader:
+      magic:           0xFEEDFACF
+      cputype:         0x1000007
+      cpusubtype:      0x3
+      filetype:        0x6
+      ncmds:           15
+      sizeofcmds:      1584
+      flags:           0x100085
+      reserved:        0x0
+    LoadCommands:
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         312
+        segname:         __TEXT
+        vmaddr:          0
+        vmsize:          12288
+        fileoff:         0
+        filesize:        12288
+        maxprot:         5
+        initprot:        5
+        nsects:          3
+        flags:           0
+        Sections:
+          - sectname:        __text
+            segname:         __TEXT
+            addr:            0x1090
+            size:            59
+            offset:          0x1090
+            align:           4
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x80000400
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         554889E5B8010000005DC30F1F440000554889E5B8010000005DC30F1F440000554889E5B8010000005DC30F1F440000554889E5B8010000005DC3
+          - sectname:        __unwind_info
+            segname:         __TEXT
+            addr:            0x10CC
+            size:            4152
+            offset:          0x10CC
+            align:           2
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         010000001C000000010000002000000000000000200000000200000000000001901000003800000038000000CB1000000000000038000000030000000C0001001000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
+          - sectname:        __eh_frame
+            segname:         __TEXT
+            addr:            0x2108
+            size:            24
+            offset:          0x2108
+            align:           3
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x6000000B
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         1400000000000000017A520001781001100C070890010000
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         __DATA_CONST
+        vmaddr:          12288
+        vmsize:          4096
+        fileoff:         12288
+        filesize:        4096
+        maxprot:         3
+        initprot:        3
+        nsects:          1
+        flags:           16
+        Sections:
+          - sectname:        __objc_imageinfo
+            segname:         __DATA_CONST
+            addr:            0x3000
+            size:            8
+            offset:          0x3000
+            align:           0
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         '0000000040000000'
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         __DATA
+        vmaddr:          16384
+        vmsize:          4096
+        fileoff:         16384
+        filesize:        4096
+        maxprot:         3
+        initprot:        3
+        nsects:          1
+        flags:           0
+        Sections:
+          - sectname:        __data
+            segname:         __DATA
+            addr:            0x4000
+            size:            8
+            offset:          0x4000
+            align:           2
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         '0100000001000000'
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         72
+        segname:         __LINKEDIT
+        vmaddr:          20480
+        vmsize:          288
+        fileoff:         20480
+        filesize:        288
+        maxprot:         1
+        initprot:        1
+        nsects:          0
+        flags:           0
+      - cmd:             LC_DYLD_INFO_ONLY
+        cmdsize:         48
+        rebase_off:      0
+        rebase_size:     0
+        bind_off:        0
+        bind_size:       0
+        weak_bind_off:   0
+        weak_bind_size:  0
+        lazy_bind_off:   0
+        lazy_bind_size:  0
+        export_off:      20480
+        export_size:     88
+      - cmd:             LC_SYMTAB
+        cmdsize:         24
+        symoff:          20576
+        nsyms:           7
+        stroff:          20688
+        strsize:         80
+      - cmd:             LC_DYSYMTAB
+        cmdsize:         80
+        ilocalsym:       0
+        nlocalsym:       1
+        iextdefsym:      1
+        nextdefsym:      5
+        iundefsym:       6
+        nundefsym:       1
+        tocoff:          0
+        ntoc:            0
+        modtaboff:       0
+        nmodtab:         0
+        extrefsymoff:    0
+        nextrefsyms:     0
+        indirectsymoff:  0
+        nindirectsyms:   0
+        extreloff:       0
+        nextrel:         0
+        locreloff:       0
+        nlocrel:         0
+      - cmd:             LC_ID_DYLIB
+        cmdsize:         96
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 65536
+          compatibility_version: 65536
+        Content:         '/System/Library/Frameworks/Mismatch.framework/Versions/A/Mismatch'
+        ZeroPadBytes:    7
+      - cmd:             LC_UUID
+        cmdsize:         24
+        uuid:            4C4C4421-5555-3144-A16C-E4AA63D66775
+      - cmd:             LC_BUILD_VERSION
+        cmdsize:         32
+        platform:        1
+        minos:           851968
+        sdk:             983040
+        ntools:          1
+        Tools:
+          - tool:            4
+            version:         1245184
+      - cmd:             LC_BUILD_VERSION
+        cmdsize:         32
+        platform:        6
+        minos:           1048576
+        sdk:             1048576
+        ntools:          1
+        Tools:
+          - tool:            4
+            version:         1245184
+      - cmd:             LC_LOAD_DYLIB
+        cmdsize:         56
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 14942208
+          compatibility_version: 65536
+        Content:         '/usr/lib/libobjc.A.dylib'
+        ZeroPadBytes:    8
+      - cmd:             LC_LOAD_DYLIB
+        cmdsize:         56
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 88539136
+          compatibility_version: 65536
+        Content:         '/usr/lib/libSystem.B.dylib'
+        ZeroPadBytes:    6
+      - cmd:             LC_FUNCTION_STARTS
+        cmdsize:         16
+        dataoff:         20568
+        datasize:        8
+      - cmd:             LC_DATA_IN_CODE
+        cmdsize:         16
+        dataoff:         20576
+        datasize:        0
+    LinkEditData:
+      ExportTrie:
+        TerminalSize:    0
+        NodeOffset:      0
+        Name:            ''
+        Flags:           0x0
+        Address:         0x0
+        Other:           0x0
+        ImportName:      ''
+        Children:
+          - TerminalSize:    0
+            NodeOffset:      5
+            Name:            _
+            Flags:           0x0
+            Address:         0x0
+            Other:           0x0
+            ImportName:      ''
+            Children:
+              - TerminalSize:    3
+                NodeOffset:      49
+                Name:            inlinedFunc
+                Flags:           0x0
+                Address:         0x1090
+                Other:           0x0
+                ImportName:      ''
+              - TerminalSize:    3
+                NodeOffset:      54
+                Name:            foo
+                Flags:           0x0
+                Address:         0x10A0
+                Other:           0x0
+                ImportName:      ''
+                Children:
+                  - TerminalSize:    4
+                    NodeOffset:      64
+                    Name:            Bar
+                    Flags:           0x0
+                    Address:         0x4000
+                    Other:           0x0
+                    ImportName:      ''
+              - TerminalSize:    4
+                NodeOffset:      70
+                Name:            unavailableSymbol
+                Flags:           0x0
+                Address:         0x4004
+                Other:           0x0
+                ImportName:      ''
+              - TerminalSize:    3
+                NodeOffset:      76
+                Name:            baz
+                Flags:           0x0
+                Address:         0x10C0
+                Other:           0x0
+                ImportName:      ''
+      NameList:
+        - n_strx:          20
+          n_type:          0x1E
+          n_sect:          1
+          n_desc:          0
+          n_value:         4272
+        - n_strx:          2
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4240
+        - n_strx:          15
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4256
+        - n_strx:          25
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4288
+        - n_strx:          30
+          n_type:          0xF
+          n_sect:          5
+          n_desc:          0
+          n_value:         16384
+        - n_strx:          38
+          n_type:          0xF
+          n_sect:          5
+          n_desc:          0
+          n_value:         16388
+        - n_strx:          57
+          n_type:          0x1
+          n_sect:          0
+          n_desc:          512
+          n_value:         0
+      StringTable:
+        - ' '
+        - _inlinedFunc
+        - _foo
+        - _bar
+        - _baz
+        - _fooBar
+        - _unavailableSymbol
+        - dyld_stub_binder
+        - ''
+        - ''
+        - ''
+        - ''
+        - ''
+        - ''
+      FunctionStarts:  [ 0x1090, 0x10A0, 0x10B0, 0x10C0 ]
+  - !mach-o
+    FileHeader:
+      magic:           0xFEEDFACF
+      cputype:         0x100000C
+      cpusubtype:      0x0
+      filetype:        0x6
+      ncmds:           16
+      sizeofcmds:      1104
+      flags:           0x100085
+      reserved:        0x0
+    LoadCommands:
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         232
+        segname:         __TEXT
+        vmaddr:          0
+        vmsize:          16384
+        fileoff:         0
+        filesize:        16384
+        maxprot:         5
+        initprot:        5
+        nsects:          2
+        flags:           0
+        Sections:
+          - sectname:        __text
+            segname:         __TEXT
+            addr:            0x1050
+            size:            32
+            offset:          0x1050
+            align:           2
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x80000400
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         20008052C0035FD620008052C0035FD620008052C0035FD620008052C0035FD6
+          - sectname:        __unwind_info
+            segname:         __TEXT
+            addr:            0x1070
+            size:            4152
+            offset:          0x1070
+            align:           2
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         010000001C000000010000002000000000000000200000000200000000000002501000003800000038000000701000000000000038000000030000000C0001001000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         __DATA_CONST
+        vmaddr:          16384
+        vmsize:          16384
+        fileoff:         16384
+        filesize:        16384
+        maxprot:         3
+        initprot:        3
+        nsects:          1
+        flags:           16
+        Sections:
+          - sectname:        __objc_imageinfo
+            segname:         __DATA_CONST
+            addr:            0x4000
+            size:            8
+            offset:          0x4000
+            align:           0
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         '0000000040000000'
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         __DATA
+        vmaddr:          32768
+        vmsize:          16384
+        fileoff:         32768
+        filesize:        16384
+        maxprot:         3
+        initprot:        3
+        nsects:          1
+        flags:           0
+        Sections:
+          - sectname:        __data
+            segname:         __DATA
+            addr:            0x8000
+            size:            8
+            offset:          0x8000
+            align:           2
+            reloff:          0x0
+            nreloc:          0
+            flags:           0x0
+            reserved1:       0x0
+            reserved2:       0x0
+            reserved3:       0x0
+            content:         '0100000001000000'
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         72
+        segname:         __LINKEDIT
+        vmaddr:          49152
+        vmsize:          832
+        fileoff:         49152
+        filesize:        832
+        maxprot:         1
+        initprot:        1
+        nsects:          0
+        flags:           0
+      - cmd:             LC_DYLD_INFO_ONLY
+        cmdsize:         48
+        rebase_off:      0
+        rebase_size:     0
+        bind_off:        0
+        bind_size:       0
+        weak_bind_off:   0
+        weak_bind_size:  0
+        lazy_bind_off:   0
+        lazy_bind_size:  0
+        export_off:      49152
+        export_size:     88
+      - cmd:             LC_SYMTAB
+        cmdsize:         24
+        symoff:          49248
+        nsyms:           7
+        stroff:          49360
+        strsize:         80
+      - cmd:             LC_DYSYMTAB
+        cmdsize:         80
+        ilocalsym:       0
+        nlocalsym:       1
+        iextdefsym:      1
+        nextdefsym:      5
+        iundefsym:       6
+        nundefsym:       1
+        tocoff:          0
+        ntoc:            0
+        modtaboff:       0
+        nmodtab:         0
+        extrefsymoff:    0
+        nextrefsyms:     0
+        indirectsymoff:  0
+        nindirectsyms:   0
+        extreloff:       0
+        nextrel:         0
+        locreloff:       0
+        nlocrel:         0
+      - cmd:             LC_ID_DYLIB
+        cmdsize:         96
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 65536
+          compatibility_version: 65536
+        Content:         '/System/Library/Frameworks/Mismatch.framework/Versions/A/Mismatch'
+        ZeroPadBytes:    7
+      - cmd:             LC_UUID
+        cmdsize:         24
+        uuid:            4C4C4428-5555-3144-A12B-7130A3C946D2
+      - cmd:             LC_BUILD_VERSION
+        cmdsize:         32
+        platform:        1
+        minos:           851968
+        sdk:             983040
+        ntools:          1
+        Tools:
+          - tool:            4
+            version:         1245184
+      - cmd:             LC_BUILD_VERSION
+        cmdsize:         32
+        platform:        6
+        minos:           1048576
+        sdk:             1048576
+        ntools:          1
+        Tools:
+          - tool:            4
+            version:         1245184
+      - cmd:             LC_LOAD_DYLIB
+        cmdsize:         56
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 14942208
+          compatibility_version: 65536
+        Content:         '/usr/lib/libobjc.A.dylib'
+        ZeroPadBytes:    8
+      - cmd:             LC_LOAD_DYLIB
+        cmdsize:         56
+        dylib:
+          name:            24
+          timestamp:       0
+          current_version: 88539136
+          compatibility_version: 65536
+        Content:         '/usr/lib/libSystem.B.dylib'
+        ZeroPadBytes:    6
+      - cmd:             LC_FUNCTION_STARTS
+        cmdsize:         16
+        dataoff:         49240
+        datasize:        8
+      - cmd:             LC_DATA_IN_CODE
+        cmdsize:         16
+        dataoff:         49248
+        datasize:        0
+      - cmd:             LC_CODE_SIGNATURE
+        cmdsize:         16
+        dataoff:         49440
+        datasize:        544
+    LinkEditData:
+      ExportTrie:
+        TerminalSize:    0
+        NodeOffset:      0
+        Name:            ''
+        Flags:           0x0
+        Address:         0x0
+        Other:           0x0
+        ImportName:      ''
+        Children:
+          - TerminalSize:    0
+            NodeOffset:      5
+            Name:            _
+            Flags:           0x0
+            Address:         0x0
+            Other:           0x0
+            ImportName:      ''
+            Children:
+              - TerminalSize:    3
+                NodeOffset:      49
+                Name:            inlinedFunc
+                Flags:           0x0
+                Address:         0x1050
+                Other:           0x0
+                ImportName:      ''
+              - TerminalSize:    3
+                NodeOffset:      54
+                Name:            foo
+                Flags:           0x0
+                Address:         0x1058
+                Other:           0x0
+                ImportName:      ''
+                Children:
+                  - TerminalSize:    4
+                    NodeOffset:      64
+                    Name:            Bar
+                    Flags:           0x0
+                    Address:         0x8000
+                    Other:           0x0
+                    ImportName:      ''
+              - TerminalSize:    4
+                NodeOffset:      70
+                Name:            unavailableSymbol
+                Flags:           0x0
+                Address:         0x8004
+                Other:           0x0
+                ImportName:      ''
+              - TerminalSize:    3
+                NodeOffset:      76
+                Name:            baz
+                Flags:           0x0
+                Address:         0x1068
+                Other:           0x0
+                ImportName:      ''
+      NameList:
+        - n_strx:          20
+          n_type:          0x1E
+          n_sect:          1
+          n_desc:          0
+          n_value:         4192
+        - n_strx:          2
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4176
+        - n_strx:          15
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4184
+        - n_strx:          25
+          n_type:          0xF
+          n_sect:          1
+          n_desc:          0
+          n_value:         4200
+        - n_strx:          30
+          n_type:          0xF
+          n_sect:          4
+          n_desc:          0
+          n_value:         32768
+        - n_strx:          38
+          n_type:          0xF
+          n_sect:          4
+          n_desc:          0
+          n_value:         32772
+        - n_strx:          57
+          n_type:          0x1
+          n_sect:          0
+          n_desc:          512
+          n_value:         0
+      StringTable:
+        - ' '
+        - _inlinedFunc
+        - _foo
+        - _bar
+        - _baz
+        - _fooBar
+        - _unavailableSymbol
+        - dyld_stub_binder
+        - ''
+        - ''
+        - ''
+        - ''
+        - ''
+        - ''
+      FunctionStarts:  [ 0x1050, 0x1058, 0x1060, 0x1068 ]
+...
+
+;--- inputs.json.in
+{
+  "headers": [ 
+  {
+    "path" : "DSTROOT/System/Library/Frameworks/Mismatch.framework/PrivateHeaders/Mismatch_Private.h",
+    "type" : "private"
+  }, 
+  {
+    "path" : "DSTROOT/System/Library/Frameworks/Mismatch.framework/Headers/Mismatch.h",
+    "type" : "public"
+  }
+  ],
+  "version": "3"
+}
diff --git a/clang/test/InstallAPI/driver-invalid-options.test b/clang/test/InstallAPI/driver-invalid-options.test
index 2b2c551fca202..8efefd1b8c594 100644
--- a/clang/test/InstallAPI/driver-invalid-options.test
+++ b/clang/test/InstallAPI/driver-invalid-options.test
@@ -19,3 +19,10 @@
 // RUN: -isysroot /no/such/path -o tmp.tbd 2> %t 
 // RUN: FileCheck --check-prefix INVALID_ISYSROOT -input-file %t %s 
 // INVALID_ISYSROOT: error: no such sysroot directory: {{.*}}no/such/path'
+
+/// Check invalid target-variant.
+// RUN: not clang-installapi -target x86_64-apple-macosx12 \
+// RUN: -darwin-target-variant arm64-apple-ios15-macabi \
+// RUN: -install_name tmp.dylib -o tmp.tbd 2> %t 
+// RUN: FileCheck --check-prefix INVALID_TARGET_VARIANT -input-file %t %s
+// INVALID_TARGET_VARIANT: error: no matching target found for target variant 'arm64-apple-ios15-macabi'
diff --git a/clang/tools/clang-installapi/Options.cpp b/clang/tools/clang-installapi/Options.cpp
index d1f9118128009..120ff3da89986 100644
--- a/clang/tools/clang-installapi/Options.cpp
+++ b/clang/tools/clang-installapi/Options.cpp
@@ -104,7 +104,7 @@ bool Options::processDriverOptions(InputArgList &Args) {
   auto *ArgArch = Args.getLastArgNoClaim(drv::OPT_arch);
   auto *ArgTarget = Args.getLastArgNoClaim(drv::OPT_target);
   auto *ArgTargetVariant =
-      Args.getLastArgNoClaim(drv::OPT_darwin_target_variant_triple);
+      Args.getLastArgNoClaim(drv::OPT_darwin_target_variant);
   if (ArgArch && (ArgTarget || ArgTargetVariant)) {
     Diags->Report(clang::diag::err_drv_argument_not_allowed_with)
         << ArgArch->getAsString(Args)
@@ -135,6 +135,53 @@ bool Options::processDriverOptions(InputArgList &Args) {
     }
   }
 
+  // Capture target variants.
+  DriverOpts.Zippered = ArgTargetVariant != nullptr;
+  for (Arg *A : Args.filtered(drv::OPT_darwin_target_variant)) {
+    A->claim();
+    Triple Variant(A->getValue());
+    if (Variant.getVendor() != Triple::Apple) {
+      Diags->Report(diag::err_unsupported_vendor)
+          << Variant.getVendorName() << A->getAsString(Args);
+      return false;
+    }
+
+    switch (Variant.getOS()) {
+    default:
+      Diags->Report(diag::err_unsupported_os)
+          << Variant.getOSName() << A->getAsString(Args);
+      return false;
+    case Triple::MacOSX:
+    case Triple::IOS:
+      break;
+    }
+
+    switch (Variant.getEnvironment()) {
+    default:
+      Diags->Report(diag::err_unsupported_environment)
+          << Variant.getEnvironmentName() << A->getAsString(Args);
+      return false;
+    case Triple::UnknownEnvironment:
+    case Triple::MacABI:
+      break;
+    }
+
+    Target TAPIVariant(Variant);
+    // See if there is a matching --target option for this --target-variant
+    // option.
+    auto It = find_if(DriverOpts.Targets, [&](const auto &T) {
+      return (T.first.Arch == TAPIVariant.Arch) &&
+             (T.first.Platform != PlatformType::PLATFORM_UNKNOWN);
+    });
+
+    if (It == DriverOpts.Targets.end()) {
+      Diags->Report(diag::err_no_matching_target) << Variant.str();
+      return false;
+    }
+
+    DriverOpts.Targets[TAPIVariant] = Variant;
+  }
+
   DriverOpts.Verbose = Args.hasArgNoClaim(drv::OPT_v);
 
   return true;
@@ -803,7 +850,8 @@ InstallAPIContext Options::createContext() {
 
   Ctx.Verifier = std::make_unique<DylibVerifier>(
       std::move(*Slices), std::move(ReexportedIFs), Diags,
-      DriverOpts.VerifyMode, DriverOpts.Demangle, DriverOpts.DSYMPath);
+      DriverOpts.VerifyMode, DriverOpts.Zippered, DriverOpts.Demangle,
+      DriverOpts.DSYMPath);
   return Ctx;
 }
 
diff --git a/clang/tools/clang-installapi/Options.h b/clang/tools/clang-installapi/Options.h
index 6da1469de2c89..984366c94e91c 100644
--- a/clang/tools/clang-installapi/Options.h
+++ b/clang/tools/clang-installapi/Options.h
@@ -75,6 +75,9 @@ struct DriverOptions {
   /// \brief Verification mode for comparing symbols.
   VerificationMode VerifyMode = VerificationMode::Pedantic;
 
+  /// \brief Whether the library is zippered.
+  bool Zippered = false;
+
   /// \brief Print demangled symbols when reporting errors.
   bool Demangle = false;
 

From 9f14914753599f3879e4c273191959e2f1b3632c Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Thu, 11 Apr 2024 20:15:43 -0700
Subject: [PATCH 213/886] Revert "Fix error in unrecognized register name
 handling for "SBFrame.register"" (#88468)

Reverts llvm/llvm-project#88047. TestFrames.py is failing on x86_64
GreenDragon:
https://green.lab.llvm.org/job/llvm.org/view/LLDB/job/lldb-cmake/983/
---
 lldb/bindings/interface/SBFrameExtensions.i  | 12 +-----------
 lldb/test/API/python_api/frame/TestFrames.py | 12 ------------
 2 files changed, 1 insertion(+), 23 deletions(-)

diff --git a/lldb/bindings/interface/SBFrameExtensions.i b/lldb/bindings/interface/SBFrameExtensions.i
index e0472280666ab..43b22ed7a6b32 100644
--- a/lldb/bindings/interface/SBFrameExtensions.i
+++ b/lldb/bindings/interface/SBFrameExtensions.i
@@ -44,16 +44,6 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                 def __init__(self, regs):
                     self.regs = regs
 
-                def __iter__(self):
-                    return self.get_registers()
-
-                def get_registers(self):
-                    for i in range(0,len(self.regs)):
-                        rs = self.regs[i]
-                        for j in range (0,rs.num_children):
-                            reg = rs.GetChildAtIndex(j)
-                            yield reg
-                          
                 def __getitem__(self, key):
                     if type(key) is str:
                         for i in range(0,len(self.regs)):
@@ -62,7 +52,7 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                                 reg = rs.GetChildAtIndex(j)
                                 if reg.name == key: return reg
                     else:
-                        return SBValue()
+                        return lldb.SBValue()
 
             return registers_access(self.registers)
 
diff --git a/lldb/test/API/python_api/frame/TestFrames.py b/lldb/test/API/python_api/frame/TestFrames.py
index dfa96d51830ba..a82b129bc8099 100644
--- a/lldb/test/API/python_api/frame/TestFrames.py
+++ b/lldb/test/API/python_api/frame/TestFrames.py
@@ -73,19 +73,7 @@ def test_get_arg_vals_for_call_stack(self):
                 gpr_reg_set = lldbutil.get_GPRs(frame)
                 pc_value = gpr_reg_set.GetChildMemberWithName("pc")
                 self.assertTrue(pc_value, "We should have a valid PC.")
-                # Make sure we can also get this from the "register" property:
-                iterator_pc_value = 0
-                found_pc = False
-                for reg in frame.register:
-                    if reg.name == "pc":
-                        found_pc = True
-                        iterator_pc_value = int(reg.GetValue(), 0)
-                        break
-
                 pc_value_int = int(pc_value.GetValue(), 0)
-                self.assertTrue(found_pc, "Found the PC value in the register list")
-                self.assertEqual(iterator_pc_value, pc_value_int, "The methods of finding pc match")
-    
                 # Make sure on arm targets we dont mismatch PC value on the basis of thumb bit.
                 # Frame PC will not have thumb bit set in case of a thumb
                 # instruction as PC.

From c1a44568911b75c98675a3c19a55b4741d861437 Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Thu, 11 Apr 2024 20:21:16 -0700
Subject: [PATCH 214/886] =?UTF-8?q?Reland=20"[lldb][lit]=20Add=20MallocNan?=
 =?UTF-8?q?oZone=20envvar=20to=20Darwin=20ASan=20builds"=20=E2=80=A6=20(#8?=
 =?UTF-8?q?8442)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…(#88436)"

This reverts commit 1f5d130df85c2d0550dc8687ad0fa1d96856c318. The
original commit checks that the host system is "Darwin" before setting
the `MallocNanoZone` envvar, but on the Shell lit config this attribute
does not exist at the point where it is being checked which leads to a
build failure.

This commit checks the host OS correctly.
---
 lldb/test/API/lit.cfg.py   | 1 +
 lldb/test/Shell/lit.cfg.py | 2 ++
 2 files changed, 3 insertions(+)

diff --git a/lldb/test/API/lit.cfg.py b/lldb/test/API/lit.cfg.py
index f9497b632fc50..9ea389c639a01 100644
--- a/lldb/test/API/lit.cfg.py
+++ b/lldb/test/API/lit.cfg.py
@@ -121,6 +121,7 @@ def delete_module_cache(path):
             config.environment["DYLD_INSERT_LIBRARIES"] = find_sanitizer_runtime(
                 "libclang_rt.asan_osx_dynamic.dylib"
             )
+            config.environment["MallocNanoZone"] = "0"
 
     if "Thread" in config.llvm_use_sanitizer:
         config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
diff --git a/lldb/test/Shell/lit.cfg.py b/lldb/test/Shell/lit.cfg.py
index 31afe5151c066..290569576ac80 100644
--- a/lldb/test/Shell/lit.cfg.py
+++ b/lldb/test/Shell/lit.cfg.py
@@ -52,6 +52,8 @@
 # Enable sanitizer runtime flags.
 config.environment["ASAN_OPTIONS"] = "detect_stack_use_after_return=1"
 config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
+if platform.system() == "Darwin":
+    config.environment["MallocNanoZone"] = "0"
 
 # Support running the test suite under the lldb-repro wrapper. This makes it
 # possible to capture a test suite run and then rerun all the test from the

From b45c9c313c5107b1942cd325e8ab3b4235948a08 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Fri, 12 Apr 2024 07:44:58 +0400
Subject: [PATCH 215/886] [clang][NFC] Move more functions to `SemaHLSL`
 (#88354)

A follow-up to #87912. I'm moving more HLSL-related functions from
`Sema` to `SemaHLSL`. I'm also dropping `HLSL` from their names in the
process.
---
 clang/include/clang/Sema/Sema.h     |  15 ---
 clang/include/clang/Sema/SemaHLSL.h |  27 +++-
 clang/lib/Parse/ParseHLSL.cpp       |  10 +-
 clang/lib/Sema/SemaDecl.cpp         | 130 +------------------
 clang/lib/Sema/SemaDeclAttr.cpp     |  54 +-------
 clang/lib/Sema/SemaHLSL.cpp         | 186 +++++++++++++++++++++++++++-
 6 files changed, 218 insertions(+), 204 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 351b8e1f31134..c4a603cc5a4a7 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -2940,13 +2940,6 @@ class Sema final : public SemaBase {
                                       QualType NewT, QualType OldT);
   void CheckMain(FunctionDecl *FD, const DeclSpec &D);
   void CheckMSVCRTEntryPoint(FunctionDecl *FD);
-  void ActOnHLSLTopLevelFunction(FunctionDecl *FD);
-  void CheckHLSLEntryPoint(FunctionDecl *FD);
-  void CheckHLSLSemanticAnnotation(FunctionDecl *EntryPoint, const Decl *Param,
-                                   const HLSLAnnotationAttr *AnnotationAttr);
-  void DiagnoseHLSLAttrStageMismatch(
-      const Attr *A, HLSLShaderAttr::ShaderType Stage,
-      std::initializer_list<HLSLShaderAttr::ShaderType> AllowedStages);
   Attr *getImplicitCodeSegOrSectionAttrForFunction(const FunctionDecl *FD,
                                                    bool IsDefinition);
   void CheckFunctionOrTemplateParamDeclarator(Scope *S, Declarator &D);
@@ -3708,14 +3701,6 @@ class Sema final : public SemaBase {
                           StringRef UuidAsWritten, MSGuidDecl *GuidDecl);
 
   BTFDeclTagAttr *mergeBTFDeclTagAttr(Decl *D, const BTFDeclTagAttr &AL);
-  HLSLNumThreadsAttr *mergeHLSLNumThreadsAttr(Decl *D,
-                                              const AttributeCommonInfo &AL,
-                                              int X, int Y, int Z);
-  HLSLShaderAttr *mergeHLSLShaderAttr(Decl *D, const AttributeCommonInfo &AL,
-                                      HLSLShaderAttr::ShaderType ShaderType);
-  HLSLParamModifierAttr *
-  mergeHLSLParamModifierAttr(Decl *D, const AttributeCommonInfo &AL,
-                             HLSLParamModifierAttr::Spelling Spelling);
 
   WebAssemblyImportNameAttr *
   mergeImportNameAttr(Decl *D, const WebAssemblyImportNameAttr &AL);
diff --git a/clang/include/clang/Sema/SemaHLSL.h b/clang/include/clang/Sema/SemaHLSL.h
index acc675963c23a..34acaf19517f2 100644
--- a/clang/include/clang/Sema/SemaHLSL.h
+++ b/clang/include/clang/Sema/SemaHLSL.h
@@ -13,12 +13,16 @@
 #ifndef LLVM_CLANG_SEMA_SEMAHLSL_H
 #define LLVM_CLANG_SEMA_SEMAHLSL_H
 
+#include "clang/AST/Attr.h"
+#include "clang/AST/Decl.h"
 #include "clang/AST/DeclBase.h"
 #include "clang/AST/Expr.h"
+#include "clang/Basic/AttributeCommonInfo.h"
 #include "clang/Basic/IdentifierTable.h"
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/SemaBase.h"
+#include <initializer_list>
 
 namespace clang {
 
@@ -26,10 +30,25 @@ class SemaHLSL : public SemaBase {
 public:
   SemaHLSL(Sema &S);
 
-  Decl *ActOnStartHLSLBuffer(Scope *BufferScope, bool CBuffer,
-                             SourceLocation KwLoc, IdentifierInfo *Ident,
-                             SourceLocation IdentLoc, SourceLocation LBrace);
-  void ActOnFinishHLSLBuffer(Decl *Dcl, SourceLocation RBrace);
+  Decl *ActOnStartBuffer(Scope *BufferScope, bool CBuffer, SourceLocation KwLoc,
+                         IdentifierInfo *Ident, SourceLocation IdentLoc,
+                         SourceLocation LBrace);
+  void ActOnFinishBuffer(Decl *Dcl, SourceLocation RBrace);
+  HLSLNumThreadsAttr *mergeNumThreadsAttr(Decl *D,
+                                          const AttributeCommonInfo &AL, int X,
+                                          int Y, int Z);
+  HLSLShaderAttr *mergeShaderAttr(Decl *D, const AttributeCommonInfo &AL,
+                                  HLSLShaderAttr::ShaderType ShaderType);
+  HLSLParamModifierAttr *
+  mergeParamModifierAttr(Decl *D, const AttributeCommonInfo &AL,
+                         HLSLParamModifierAttr::Spelling Spelling);
+  void ActOnTopLevelFunction(FunctionDecl *FD);
+  void CheckEntryPoint(FunctionDecl *FD);
+  void CheckSemanticAnnotation(FunctionDecl *EntryPoint, const Decl *Param,
+                               const HLSLAnnotationAttr *AnnotationAttr);
+  void DiagnoseAttrStageMismatch(
+      const Attr *A, HLSLShaderAttr::ShaderType Stage,
+      std::initializer_list<HLSLShaderAttr::ShaderType> AllowedStages);
 };
 
 } // namespace clang
diff --git a/clang/lib/Parse/ParseHLSL.cpp b/clang/lib/Parse/ParseHLSL.cpp
index 5afc958600fa5..d97985d42369a 100644
--- a/clang/lib/Parse/ParseHLSL.cpp
+++ b/clang/lib/Parse/ParseHLSL.cpp
@@ -72,9 +72,9 @@ Decl *Parser::ParseHLSLBuffer(SourceLocation &DeclEnd) {
     return nullptr;
   }
 
-  Decl *D = Actions.HLSL().ActOnStartHLSLBuffer(
-      getCurScope(), IsCBuffer, BufferLoc, Identifier, IdentifierLoc,
-      T.getOpenLocation());
+  Decl *D = Actions.HLSL().ActOnStartBuffer(getCurScope(), IsCBuffer, BufferLoc,
+                                            Identifier, IdentifierLoc,
+                                            T.getOpenLocation());
 
   while (Tok.isNot(tok::r_brace) && Tok.isNot(tok::eof)) {
     // FIXME: support attribute on constants inside cbuffer/tbuffer.
@@ -88,7 +88,7 @@ Decl *Parser::ParseHLSLBuffer(SourceLocation &DeclEnd) {
       T.skipToEnd();
       DeclEnd = T.getCloseLocation();
       BufferScope.Exit();
-      Actions.HLSL().ActOnFinishHLSLBuffer(D, DeclEnd);
+      Actions.HLSL().ActOnFinishBuffer(D, DeclEnd);
       return nullptr;
     }
   }
@@ -96,7 +96,7 @@ Decl *Parser::ParseHLSLBuffer(SourceLocation &DeclEnd) {
   T.consumeClose();
   DeclEnd = T.getCloseLocation();
   BufferScope.Exit();
-  Actions.HLSL().ActOnFinishHLSLBuffer(D, DeclEnd);
+  Actions.HLSL().ActOnFinishBuffer(D, DeclEnd);
 
   Actions.ProcessDeclAttributeList(Actions.CurScope, D, Attrs);
   return D;
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 691e170cd0514..afd9e407946a0 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -45,6 +45,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/SmallString.h"
@@ -2972,10 +2973,10 @@ static bool mergeDeclAttribute(Sema &S, NamedDecl *D,
   else if (const auto *BTFA = dyn_cast<BTFDeclTagAttr>(Attr))
     NewAttr = S.mergeBTFDeclTagAttr(D, *BTFA);
   else if (const auto *NT = dyn_cast<HLSLNumThreadsAttr>(Attr))
-    NewAttr =
-        S.mergeHLSLNumThreadsAttr(D, *NT, NT->getX(), NT->getY(), NT->getZ());
+    NewAttr = S.HLSL().mergeNumThreadsAttr(D, *NT, NT->getX(), NT->getY(),
+                                           NT->getZ());
   else if (const auto *SA = dyn_cast<HLSLShaderAttr>(Attr))
-    NewAttr = S.mergeHLSLShaderAttr(D, *SA, SA->getType());
+    NewAttr = S.HLSL().mergeShaderAttr(D, *SA, SA->getType());
   else if (isa<SuppressAttr>(Attr))
     // Do nothing. Each redeclaration should be suppressed separately.
     NewAttr = nullptr;
@@ -10808,10 +10809,10 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
   if (getLangOpts().HLSL && D.isFunctionDefinition()) {
     // Any top level function could potentially be specified as an entry.
     if (!NewFD->isInvalidDecl() && S->getDepth() == 0 && Name.isIdentifier())
-      ActOnHLSLTopLevelFunction(NewFD);
+      HLSL().ActOnTopLevelFunction(NewFD);
 
     if (NewFD->hasAttr<HLSLShaderAttr>())
-      CheckHLSLEntryPoint(NewFD);
+      HLSL().CheckEntryPoint(NewFD);
   }
 
   // If this is the first declaration of a library builtin function, add
@@ -12665,125 +12666,6 @@ void Sema::CheckMSVCRTEntryPoint(FunctionDecl *FD) {
   }
 }
 
-void Sema::ActOnHLSLTopLevelFunction(FunctionDecl *FD) {
-  auto &TargetInfo = getASTContext().getTargetInfo();
-
-  if (FD->getName() != TargetInfo.getTargetOpts().HLSLEntry)
-    return;
-
-  StringRef Env = TargetInfo.getTriple().getEnvironmentName();
-  HLSLShaderAttr::ShaderType ShaderType;
-  if (HLSLShaderAttr::ConvertStrToShaderType(Env, ShaderType)) {
-    if (const auto *Shader = FD->getAttr<HLSLShaderAttr>()) {
-      // The entry point is already annotated - check that it matches the
-      // triple.
-      if (Shader->getType() != ShaderType) {
-        Diag(Shader->getLocation(), diag::err_hlsl_entry_shader_attr_mismatch)
-            << Shader;
-        FD->setInvalidDecl();
-      }
-    } else {
-      // Implicitly add the shader attribute if the entry function isn't
-      // explicitly annotated.
-      FD->addAttr(HLSLShaderAttr::CreateImplicit(Context, ShaderType,
-                                                 FD->getBeginLoc()));
-    }
-  } else {
-    switch (TargetInfo.getTriple().getEnvironment()) {
-    case llvm::Triple::UnknownEnvironment:
-    case llvm::Triple::Library:
-      break;
-    default:
-      llvm_unreachable("Unhandled environment in triple");
-    }
-  }
-}
-
-void Sema::CheckHLSLEntryPoint(FunctionDecl *FD) {
-  const auto *ShaderAttr = FD->getAttr<HLSLShaderAttr>();
-  assert(ShaderAttr && "Entry point has no shader attribute");
-  HLSLShaderAttr::ShaderType ST = ShaderAttr->getType();
-
-  switch (ST) {
-  case HLSLShaderAttr::Pixel:
-  case HLSLShaderAttr::Vertex:
-  case HLSLShaderAttr::Geometry:
-  case HLSLShaderAttr::Hull:
-  case HLSLShaderAttr::Domain:
-  case HLSLShaderAttr::RayGeneration:
-  case HLSLShaderAttr::Intersection:
-  case HLSLShaderAttr::AnyHit:
-  case HLSLShaderAttr::ClosestHit:
-  case HLSLShaderAttr::Miss:
-  case HLSLShaderAttr::Callable:
-    if (const auto *NT = FD->getAttr<HLSLNumThreadsAttr>()) {
-      DiagnoseHLSLAttrStageMismatch(NT, ST,
-                                    {HLSLShaderAttr::Compute,
-                                     HLSLShaderAttr::Amplification,
-                                     HLSLShaderAttr::Mesh});
-      FD->setInvalidDecl();
-    }
-    break;
-
-  case HLSLShaderAttr::Compute:
-  case HLSLShaderAttr::Amplification:
-  case HLSLShaderAttr::Mesh:
-    if (!FD->hasAttr<HLSLNumThreadsAttr>()) {
-      Diag(FD->getLocation(), diag::err_hlsl_missing_numthreads)
-          << HLSLShaderAttr::ConvertShaderTypeToStr(ST);
-      FD->setInvalidDecl();
-    }
-    break;
-  }
-
-  for (ParmVarDecl *Param : FD->parameters()) {
-    if (const auto *AnnotationAttr = Param->getAttr<HLSLAnnotationAttr>()) {
-      CheckHLSLSemanticAnnotation(FD, Param, AnnotationAttr);
-    } else {
-      // FIXME: Handle struct parameters where annotations are on struct fields.
-      // See: https://github.com/llvm/llvm-project/issues/57875
-      Diag(FD->getLocation(), diag::err_hlsl_missing_semantic_annotation);
-      Diag(Param->getLocation(), diag::note_previous_decl) << Param;
-      FD->setInvalidDecl();
-    }
-  }
-  // FIXME: Verify return type semantic annotation.
-}
-
-void Sema::CheckHLSLSemanticAnnotation(
-    FunctionDecl *EntryPoint, const Decl *Param,
-    const HLSLAnnotationAttr *AnnotationAttr) {
-  auto *ShaderAttr = EntryPoint->getAttr<HLSLShaderAttr>();
-  assert(ShaderAttr && "Entry point has no shader attribute");
-  HLSLShaderAttr::ShaderType ST = ShaderAttr->getType();
-
-  switch (AnnotationAttr->getKind()) {
-  case attr::HLSLSV_DispatchThreadID:
-  case attr::HLSLSV_GroupIndex:
-    if (ST == HLSLShaderAttr::Compute)
-      return;
-    DiagnoseHLSLAttrStageMismatch(AnnotationAttr, ST,
-                                  {HLSLShaderAttr::Compute});
-    break;
-  default:
-    llvm_unreachable("Unknown HLSLAnnotationAttr");
-  }
-}
-
-void Sema::DiagnoseHLSLAttrStageMismatch(
-    const Attr *A, HLSLShaderAttr::ShaderType Stage,
-    std::initializer_list<HLSLShaderAttr::ShaderType> AllowedStages) {
-  SmallVector<StringRef, 8> StageStrings;
-  llvm::transform(AllowedStages, std::back_inserter(StageStrings),
-                  [](HLSLShaderAttr::ShaderType ST) {
-                    return StringRef(
-                        HLSLShaderAttr::ConvertShaderTypeToStr(ST));
-                  });
-  Diag(A->getLoc(), diag::err_hlsl_attr_unsupported_in_stage)
-      << A << HLSLShaderAttr::ConvertShaderTypeToStr(Stage)
-      << (AllowedStages.size() != 1) << join(StageStrings, ", ");
-}
-
 bool Sema::CheckForConstantInitializer(Expr *Init, QualType DclT) {
   // FIXME: Need strict checking.  In C89, we need to check for
   // any assignment, increment, decrement, function-calls, or
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index 8bce04640e748..b91064e28e415 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -39,6 +39,7 @@
 #include "clang/Sema/ParsedAttr.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/StringExtras.h"
@@ -7238,24 +7239,11 @@ static void handleHLSLNumThreadsAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     return;
   }
 
-  HLSLNumThreadsAttr *NewAttr = S.mergeHLSLNumThreadsAttr(D, AL, X, Y, Z);
+  HLSLNumThreadsAttr *NewAttr = S.HLSL().mergeNumThreadsAttr(D, AL, X, Y, Z);
   if (NewAttr)
     D->addAttr(NewAttr);
 }
 
-HLSLNumThreadsAttr *Sema::mergeHLSLNumThreadsAttr(Decl *D,
-                                                  const AttributeCommonInfo &AL,
-                                                  int X, int Y, int Z) {
-  if (HLSLNumThreadsAttr *NT = D->getAttr<HLSLNumThreadsAttr>()) {
-    if (NT->getX() != X || NT->getY() != Y || NT->getZ() != Z) {
-      Diag(NT->getLocation(), diag::err_hlsl_attribute_param_mismatch) << AL;
-      Diag(AL.getLoc(), diag::note_conflicting_attribute);
-    }
-    return nullptr;
-  }
-  return ::new (Context) HLSLNumThreadsAttr(Context, AL, X, Y, Z);
-}
-
 static bool isLegalTypeForHLSLSV_DispatchThreadID(QualType T) {
   if (!T->hasUnsignedIntegerRepresentation())
     return false;
@@ -7299,24 +7287,11 @@ static void handleHLSLShaderAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
 
   // FIXME: check function match the shader stage.
 
-  HLSLShaderAttr *NewAttr = S.mergeHLSLShaderAttr(D, AL, ShaderType);
+  HLSLShaderAttr *NewAttr = S.HLSL().mergeShaderAttr(D, AL, ShaderType);
   if (NewAttr)
     D->addAttr(NewAttr);
 }
 
-HLSLShaderAttr *
-Sema::mergeHLSLShaderAttr(Decl *D, const AttributeCommonInfo &AL,
-                          HLSLShaderAttr::ShaderType ShaderType) {
-  if (HLSLShaderAttr *NT = D->getAttr<HLSLShaderAttr>()) {
-    if (NT->getType() != ShaderType) {
-      Diag(NT->getLocation(), diag::err_hlsl_attribute_param_mismatch) << AL;
-      Diag(AL.getLoc(), diag::note_conflicting_attribute);
-    }
-    return nullptr;
-  }
-  return HLSLShaderAttr::Create(Context, ShaderType, AL);
-}
-
 static void handleHLSLResourceBindingAttr(Sema &S, Decl *D,
                                           const ParsedAttr &AL) {
   StringRef Space = "space0";
@@ -7391,34 +7366,13 @@ static void handleHLSLResourceBindingAttr(Sema &S, Decl *D,
 
 static void handleHLSLParamModifierAttr(Sema &S, Decl *D,
                                         const ParsedAttr &AL) {
-  HLSLParamModifierAttr *NewAttr = S.mergeHLSLParamModifierAttr(
+  HLSLParamModifierAttr *NewAttr = S.HLSL().mergeParamModifierAttr(
       D, AL,
       static_cast<HLSLParamModifierAttr::Spelling>(AL.getSemanticSpelling()));
   if (NewAttr)
     D->addAttr(NewAttr);
 }
 
-HLSLParamModifierAttr *
-Sema::mergeHLSLParamModifierAttr(Decl *D, const AttributeCommonInfo &AL,
-                                 HLSLParamModifierAttr::Spelling Spelling) {
-  // We can only merge an `in` attribute with an `out` attribute. All other
-  // combinations of duplicated attributes are ill-formed.
-  if (HLSLParamModifierAttr *PA = D->getAttr<HLSLParamModifierAttr>()) {
-    if ((PA->isIn() && Spelling == HLSLParamModifierAttr::Keyword_out) ||
-        (PA->isOut() && Spelling == HLSLParamModifierAttr::Keyword_in)) {
-      D->dropAttr<HLSLParamModifierAttr>();
-      SourceRange AdjustedRange = {PA->getLocation(), AL.getRange().getEnd()};
-      return HLSLParamModifierAttr::Create(
-          Context, /*MergedSpelling=*/true, AdjustedRange,
-          HLSLParamModifierAttr::Keyword_inout);
-    }
-    Diag(AL.getLoc(), diag::err_hlsl_duplicate_parameter_modifier) << AL;
-    Diag(PA->getLocation(), diag::note_conflicting_attribute);
-    return nullptr;
-  }
-  return HLSLParamModifierAttr::Create(Context, AL);
-}
-
 static void handleMSInheritanceAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   if (!S.LangOpts.CPlusPlus) {
     S.Diag(AL.getLoc(), diag::err_attribute_not_supported_in_lang)
diff --git a/clang/lib/Sema/SemaHLSL.cpp b/clang/lib/Sema/SemaHLSL.cpp
index 681849d6e6c8a..bb9e37f18d370 100644
--- a/clang/lib/Sema/SemaHLSL.cpp
+++ b/clang/lib/Sema/SemaHLSL.cpp
@@ -9,17 +9,25 @@
 //===----------------------------------------------------------------------===//
 
 #include "clang/Sema/SemaHLSL.h"
+#include "clang/Basic/DiagnosticSema.h"
+#include "clang/Basic/LLVM.h"
+#include "clang/Basic/TargetInfo.h"
 #include "clang/Sema/Sema.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/TargetParser/Triple.h"
+#include <iterator>
 
 using namespace clang;
 
 SemaHLSL::SemaHLSL(Sema &S) : SemaBase(S) {}
 
-Decl *SemaHLSL::ActOnStartHLSLBuffer(Scope *BufferScope, bool CBuffer,
-                                     SourceLocation KwLoc,
-                                     IdentifierInfo *Ident,
-                                     SourceLocation IdentLoc,
-                                     SourceLocation LBrace) {
+Decl *SemaHLSL::ActOnStartBuffer(Scope *BufferScope, bool CBuffer,
+                                 SourceLocation KwLoc, IdentifierInfo *Ident,
+                                 SourceLocation IdentLoc,
+                                 SourceLocation LBrace) {
   // For anonymous namespace, take the location of the left brace.
   DeclContext *LexicalParent = SemaRef.getCurLexicalContext();
   HLSLBufferDecl *Result = HLSLBufferDecl::Create(
@@ -31,8 +39,174 @@ Decl *SemaHLSL::ActOnStartHLSLBuffer(Scope *BufferScope, bool CBuffer,
   return Result;
 }
 
-void SemaHLSL::ActOnFinishHLSLBuffer(Decl *Dcl, SourceLocation RBrace) {
+void SemaHLSL::ActOnFinishBuffer(Decl *Dcl, SourceLocation RBrace) {
   auto *BufDecl = cast<HLSLBufferDecl>(Dcl);
   BufDecl->setRBraceLoc(RBrace);
   SemaRef.PopDeclContext();
 }
+
+HLSLNumThreadsAttr *SemaHLSL::mergeNumThreadsAttr(Decl *D,
+                                                  const AttributeCommonInfo &AL,
+                                                  int X, int Y, int Z) {
+  if (HLSLNumThreadsAttr *NT = D->getAttr<HLSLNumThreadsAttr>()) {
+    if (NT->getX() != X || NT->getY() != Y || NT->getZ() != Z) {
+      Diag(NT->getLocation(), diag::err_hlsl_attribute_param_mismatch) << AL;
+      Diag(AL.getLoc(), diag::note_conflicting_attribute);
+    }
+    return nullptr;
+  }
+  return ::new (getASTContext())
+      HLSLNumThreadsAttr(getASTContext(), AL, X, Y, Z);
+}
+
+HLSLShaderAttr *
+SemaHLSL::mergeShaderAttr(Decl *D, const AttributeCommonInfo &AL,
+                          HLSLShaderAttr::ShaderType ShaderType) {
+  if (HLSLShaderAttr *NT = D->getAttr<HLSLShaderAttr>()) {
+    if (NT->getType() != ShaderType) {
+      Diag(NT->getLocation(), diag::err_hlsl_attribute_param_mismatch) << AL;
+      Diag(AL.getLoc(), diag::note_conflicting_attribute);
+    }
+    return nullptr;
+  }
+  return HLSLShaderAttr::Create(getASTContext(), ShaderType, AL);
+}
+
+HLSLParamModifierAttr *
+SemaHLSL::mergeParamModifierAttr(Decl *D, const AttributeCommonInfo &AL,
+                                 HLSLParamModifierAttr::Spelling Spelling) {
+  // We can only merge an `in` attribute with an `out` attribute. All other
+  // combinations of duplicated attributes are ill-formed.
+  if (HLSLParamModifierAttr *PA = D->getAttr<HLSLParamModifierAttr>()) {
+    if ((PA->isIn() && Spelling == HLSLParamModifierAttr::Keyword_out) ||
+        (PA->isOut() && Spelling == HLSLParamModifierAttr::Keyword_in)) {
+      D->dropAttr<HLSLParamModifierAttr>();
+      SourceRange AdjustedRange = {PA->getLocation(), AL.getRange().getEnd()};
+      return HLSLParamModifierAttr::Create(
+          getASTContext(), /*MergedSpelling=*/true, AdjustedRange,
+          HLSLParamModifierAttr::Keyword_inout);
+    }
+    Diag(AL.getLoc(), diag::err_hlsl_duplicate_parameter_modifier) << AL;
+    Diag(PA->getLocation(), diag::note_conflicting_attribute);
+    return nullptr;
+  }
+  return HLSLParamModifierAttr::Create(getASTContext(), AL);
+}
+
+void SemaHLSL::ActOnTopLevelFunction(FunctionDecl *FD) {
+  auto &TargetInfo = getASTContext().getTargetInfo();
+
+  if (FD->getName() != TargetInfo.getTargetOpts().HLSLEntry)
+    return;
+
+  StringRef Env = TargetInfo.getTriple().getEnvironmentName();
+  HLSLShaderAttr::ShaderType ShaderType;
+  if (HLSLShaderAttr::ConvertStrToShaderType(Env, ShaderType)) {
+    if (const auto *Shader = FD->getAttr<HLSLShaderAttr>()) {
+      // The entry point is already annotated - check that it matches the
+      // triple.
+      if (Shader->getType() != ShaderType) {
+        Diag(Shader->getLocation(), diag::err_hlsl_entry_shader_attr_mismatch)
+            << Shader;
+        FD->setInvalidDecl();
+      }
+    } else {
+      // Implicitly add the shader attribute if the entry function isn't
+      // explicitly annotated.
+      FD->addAttr(HLSLShaderAttr::CreateImplicit(getASTContext(), ShaderType,
+                                                 FD->getBeginLoc()));
+    }
+  } else {
+    switch (TargetInfo.getTriple().getEnvironment()) {
+    case llvm::Triple::UnknownEnvironment:
+    case llvm::Triple::Library:
+      break;
+    default:
+      llvm_unreachable("Unhandled environment in triple");
+    }
+  }
+}
+
+void SemaHLSL::CheckEntryPoint(FunctionDecl *FD) {
+  const auto *ShaderAttr = FD->getAttr<HLSLShaderAttr>();
+  assert(ShaderAttr && "Entry point has no shader attribute");
+  HLSLShaderAttr::ShaderType ST = ShaderAttr->getType();
+
+  switch (ST) {
+  case HLSLShaderAttr::Pixel:
+  case HLSLShaderAttr::Vertex:
+  case HLSLShaderAttr::Geometry:
+  case HLSLShaderAttr::Hull:
+  case HLSLShaderAttr::Domain:
+  case HLSLShaderAttr::RayGeneration:
+  case HLSLShaderAttr::Intersection:
+  case HLSLShaderAttr::AnyHit:
+  case HLSLShaderAttr::ClosestHit:
+  case HLSLShaderAttr::Miss:
+  case HLSLShaderAttr::Callable:
+    if (const auto *NT = FD->getAttr<HLSLNumThreadsAttr>()) {
+      DiagnoseAttrStageMismatch(NT, ST,
+                                {HLSLShaderAttr::Compute,
+                                 HLSLShaderAttr::Amplification,
+                                 HLSLShaderAttr::Mesh});
+      FD->setInvalidDecl();
+    }
+    break;
+
+  case HLSLShaderAttr::Compute:
+  case HLSLShaderAttr::Amplification:
+  case HLSLShaderAttr::Mesh:
+    if (!FD->hasAttr<HLSLNumThreadsAttr>()) {
+      Diag(FD->getLocation(), diag::err_hlsl_missing_numthreads)
+          << HLSLShaderAttr::ConvertShaderTypeToStr(ST);
+      FD->setInvalidDecl();
+    }
+    break;
+  }
+
+  for (ParmVarDecl *Param : FD->parameters()) {
+    if (const auto *AnnotationAttr = Param->getAttr<HLSLAnnotationAttr>()) {
+      CheckSemanticAnnotation(FD, Param, AnnotationAttr);
+    } else {
+      // FIXME: Handle struct parameters where annotations are on struct fields.
+      // See: https://github.com/llvm/llvm-project/issues/57875
+      Diag(FD->getLocation(), diag::err_hlsl_missing_semantic_annotation);
+      Diag(Param->getLocation(), diag::note_previous_decl) << Param;
+      FD->setInvalidDecl();
+    }
+  }
+  // FIXME: Verify return type semantic annotation.
+}
+
+void SemaHLSL::CheckSemanticAnnotation(
+    FunctionDecl *EntryPoint, const Decl *Param,
+    const HLSLAnnotationAttr *AnnotationAttr) {
+  auto *ShaderAttr = EntryPoint->getAttr<HLSLShaderAttr>();
+  assert(ShaderAttr && "Entry point has no shader attribute");
+  HLSLShaderAttr::ShaderType ST = ShaderAttr->getType();
+
+  switch (AnnotationAttr->getKind()) {
+  case attr::HLSLSV_DispatchThreadID:
+  case attr::HLSLSV_GroupIndex:
+    if (ST == HLSLShaderAttr::Compute)
+      return;
+    DiagnoseAttrStageMismatch(AnnotationAttr, ST, {HLSLShaderAttr::Compute});
+    break;
+  default:
+    llvm_unreachable("Unknown HLSLAnnotationAttr");
+  }
+}
+
+void SemaHLSL::DiagnoseAttrStageMismatch(
+    const Attr *A, HLSLShaderAttr::ShaderType Stage,
+    std::initializer_list<HLSLShaderAttr::ShaderType> AllowedStages) {
+  SmallVector<StringRef, 8> StageStrings;
+  llvm::transform(AllowedStages, std::back_inserter(StageStrings),
+                  [](HLSLShaderAttr::ShaderType ST) {
+                    return StringRef(
+                        HLSLShaderAttr::ConvertShaderTypeToStr(ST));
+                  });
+  Diag(A->getLoc(), diag::err_hlsl_attr_unsupported_in_stage)
+      << A << HLSLShaderAttr::ConvertShaderTypeToStr(Stage)
+      << (AllowedStages.size() != 1) << join(StageStrings, ", ");
+}

From 505a9ae81d620b92284f7b2cbe874c936cf19583 Mon Sep 17 00:00:00 2001
From: Sirraide <aeternalmail@gmail.com>
Date: Fri, 12 Apr 2024 06:49:53 +0200
Subject: [PATCH 216/886] [Clang] Look through type sugar when accessing
 FunctionProtoType (#88428)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This fixes a bug introduced by #84473: if a lambda’s type is type sugar
(e.g. an `AttributedType`), we need to use `getAs()` instead of `cast()`
to retrieve the `FunctionProtoType`.
---
 clang/lib/Sema/SemaExpr.cpp                |  2 +-
 clang/lib/Sema/SemaExprCXX.cpp             |  2 +-
 clang/test/SemaCXX/cxx2b-deducing-this.cpp | 12 ++++++++++++
 clang/test/SemaCXX/lambda-expressions.cpp  |  9 +++++++++
 4 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 846e7d5d3ab92..ec5ca2b9352ed 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -20704,7 +20704,7 @@ static void FixDependencyOfIdExpressionsInLambdaWithDependentObjectParameter(
       if (MD->getType().isNull())
         continue;
 
-      const auto *Ty = cast<FunctionProtoType>(MD->getType());
+      const auto *Ty = MD->getType()->getAs<FunctionProtoType>();
       if (!Ty || !MD->isExplicitObjectMemberFunction() ||
           !Ty->getParamType(0)->isDependentType())
         continue;
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 763901d4418d2..ac39cb99fc934 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -1488,7 +1488,7 @@ void Sema::MarkThisReferenced(CXXThisExpr *This) {
         if (MD->getType().isNull())
           return false;
 
-        const auto *Ty = cast<FunctionProtoType>(MD->getType());
+        const auto *Ty = MD->getType()->getAs<FunctionProtoType>();
         return Ty && MD->isExplicitObjectMemberFunction() &&
                Ty->getParamType(0)->isDependentType();
       }
diff --git a/clang/test/SemaCXX/cxx2b-deducing-this.cpp b/clang/test/SemaCXX/cxx2b-deducing-this.cpp
index c14c971afd235..5f29a955e053c 100644
--- a/clang/test/SemaCXX/cxx2b-deducing-this.cpp
+++ b/clang/test/SemaCXX/cxx2b-deducing-this.cpp
@@ -312,6 +312,18 @@ void TestMutationInLambda() {
       l1();
       l2();
     }
+
+    // Check that we don't crash if the lambda has type sugar.
+    const auto l15 = [=](this auto&&) [[clang::annotate_type("foo")]] [[clang::annotate_type("bar")]] {
+        return x;
+    };
+
+    const auto l16 = [=]() [[clang::annotate_type("foo")]] [[clang::annotate_type("bar")]] {
+        return x;
+    };
+
+    l15();
+    l16();
 }
 
 struct Over_Call_Func_Example {
diff --git a/clang/test/SemaCXX/lambda-expressions.cpp b/clang/test/SemaCXX/lambda-expressions.cpp
index 151d74f21d64d..6be338064452e 100644
--- a/clang/test/SemaCXX/lambda-expressions.cpp
+++ b/clang/test/SemaCXX/lambda-expressions.cpp
@@ -762,3 +762,12 @@ template auto t::operator()<int>(int a) const; // expected-note {{in instantiati
 
 }
 #endif
+
+namespace GH84473_bug {
+void f1() {
+  int b;
+  (void) [=] [[gnu::regcall]] () { // expected-warning {{an attribute specifier sequence in this position is a C++23 extension}}
+    (void) b;
+  };
+}
+}

From f21ead06750b670cf8ce72d6666e3550b04056a2 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Fri, 12 Apr 2024 12:51:58 +0800
Subject: [PATCH 217/886] [C++20] [Modules] [Reduced BMI] Remove unreachable
 decls GMF in redued BMI (#88359)

Following of https://github.com/llvm/llvm-project/pull/76930

This follows the idea of "only writes what we writes", which I think is
the most natural and efficient way to implement this optimization.

We start writing the BMI from the first declaration in module purview
instead of the global module fragment, so that everything in the GMF
untouched won't be written in the BMI naturally.

The exception is, as I said in
https://github.com/llvm/llvm-project/pull/76930, when we write a
declaration we need to write its decl context, and when we write the
decl context, we need to write everything from it. So when we see
`std::vector`, we basically need to write everything under namespace
std. This violates our intention. To fix this, this patch delays the
writing of namespace in the GMF.

From my local measurement, the size of the BMI decrease to 90M from 112M
for a local modules build. I think this is significant.

This feature will be covered under the experimental reduced BMI so that
it won't affect any existing users. So I'd like to land this when the CI
gets green.

Documents will be added seperately.
---
 .../include/clang/Serialization/ASTBitCodes.h |   4 +
 clang/include/clang/Serialization/ASTReader.h |  14 ++
 clang/include/clang/Serialization/ASTWriter.h |  26 +++-
 clang/lib/Serialization/ASTReader.cpp         |  23 +++
 clang/lib/Serialization/ASTReaderDecl.cpp     |   9 ++
 clang/lib/Serialization/ASTWriter.cpp         | 139 +++++++++++++++---
 clang/lib/Serialization/ASTWriterDecl.cpp     |  27 +++-
 .../module.glob.frag/cxx20-10-4-ex2.cppm      |  58 ++++++++
 .../inconsistent-deduction-guide-linkage.cppm |   8 +
 clang/test/Modules/named-modules-adl-2.cppm   |   7 +-
 clang/test/Modules/named-modules-adl.cppm     |   7 +-
 .../test/Modules/placement-new-reachable.cpp  |   9 +-
 clang/test/Modules/polluted-operator.cppm     |  15 +-
 clang/test/Modules/pr62589.cppm               |   4 +
 clang/test/Modules/preferred_name.cppm        |   2 +
 .../redundant-template-default-arg3.cpp       |  10 ++
 .../template-function-specialization.cpp      |   8 +-
 17 files changed, 337 insertions(+), 33 deletions(-)
 create mode 100644 clang/test/CXX/module/module.glob.frag/cxx20-10-4-ex2.cppm

diff --git a/clang/include/clang/Serialization/ASTBitCodes.h b/clang/include/clang/Serialization/ASTBitCodes.h
index f762116fea956..500098dd3dab1 100644
--- a/clang/include/clang/Serialization/ASTBitCodes.h
+++ b/clang/include/clang/Serialization/ASTBitCodes.h
@@ -698,6 +698,10 @@ enum ASTRecordTypes {
   /// Record code for an unterminated \#pragma clang assume_nonnull begin
   /// recorded in a preamble.
   PP_ASSUME_NONNULL_LOC = 67,
+
+  /// Record code for lexical and visible block for delayed namespace in
+  /// reduced BMI.
+  DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD = 68,
 };
 
 /// Record types used within a source manager block.
diff --git a/clang/include/clang/Serialization/ASTReader.h b/clang/include/clang/Serialization/ASTReader.h
index e8b9f28690d9f..6656c1c58dec9 100644
--- a/clang/include/clang/Serialization/ASTReader.h
+++ b/clang/include/clang/Serialization/ASTReader.h
@@ -517,6 +517,20 @@ class ASTReader
   /// in the chain.
   DeclUpdateOffsetsMap DeclUpdateOffsets;
 
+  using DelayedNamespaceOffsetMapTy = llvm::DenseMap<
+      serialization::DeclID,
+      std::pair</*LexicalOffset*/ uint64_t, /*VisibleOffset*/ uint64_t>>;
+
+  /// Mapping from global declaration IDs to the lexical and visible block
+  /// offset for delayed namespace in reduced BMI.
+  ///
+  /// We can't use the existing DeclUpdate mechanism since the DeclUpdate
+  /// may only be applied in an outer most read. However, we need to know
+  /// whether or not a DeclContext has external storage during the recursive
+  /// reading. So we need to apply the offset immediately after we read the
+  /// namespace as if it is not delayed.
+  DelayedNamespaceOffsetMapTy DelayedNamespaceOffsetMap;
+
   struct PendingUpdateRecord {
     Decl *D;
     serialization::GlobalDeclID ID;
diff --git a/clang/include/clang/Serialization/ASTWriter.h b/clang/include/clang/Serialization/ASTWriter.h
index 214eb3601148b..443f770310470 100644
--- a/clang/include/clang/Serialization/ASTWriter.h
+++ b/clang/include/clang/Serialization/ASTWriter.h
@@ -201,6 +201,16 @@ class ASTWriter : public ASTDeserializationListener,
   /// The declarations and types to emit.
   std::queue<DeclOrType> DeclTypesToEmit;
 
+  /// The delayed namespace to emit. Only meaningful for reduced BMI.
+  ///
+  /// In reduced BMI, we want to elide the unreachable declarations in
+  /// the global module fragment. However, in ASTWriterDecl, when we see
+  /// a namespace, all the declarations in the namespace would be emitted.
+  /// So the optimization become meaningless. To solve the issue, we
+  /// delay recording all the declarations until we emit all the declarations.
+  /// Then we can safely record the reached declarations only.
+  llvm::SmallVector<NamespaceDecl *, 16> DelayedNamespace;
+
   /// The first ID number we can use for our own declarations.
   serialization::DeclID FirstDeclID = serialization::NUM_PREDEF_DECL_IDS;
 
@@ -529,7 +539,8 @@ class ASTWriter : public ASTDeserializationListener,
   void WriteType(QualType T);
 
   bool isLookupResultExternal(StoredDeclsList &Result, DeclContext *DC);
-  bool isLookupResultEntirelyExternal(StoredDeclsList &Result, DeclContext *DC);
+  bool isLookupResultEntirelyExternalOrUnreachable(StoredDeclsList &Result,
+                                                   DeclContext *DC);
 
   void GenerateNameLookupTable(const DeclContext *DC,
                                llvm::SmallVectorImpl<char> &LookupTable);
@@ -704,6 +715,15 @@ class ASTWriter : public ASTDeserializationListener,
   /// declaration.
   serialization::DeclID getDeclID(const Decl *D);
 
+  /// Whether or not the declaration got emitted. If not, it wouldn't be
+  /// emitted.
+  ///
+  /// This may only be called after we've done the job to write the
+  /// declarations (marked by DoneWritingDeclsAndTypes).
+  ///
+  /// A declaration may only be omitted in reduced BMI.
+  bool wasDeclEmitted(const Decl *D) const;
+
   unsigned getAnonymousDeclarationNumber(const NamedDecl *D);
 
   /// Add a string to the given record.
@@ -798,6 +818,10 @@ class ASTWriter : public ASTDeserializationListener,
     return WritingModule && WritingModule->isNamedModule();
   }
 
+  bool isGeneratingReducedBMI() const { return GeneratingReducedBMI; }
+
+  bool getDoneWritingDeclsAndTypes() const { return DoneWritingDeclsAndTypes; }
+
 private:
   // ASTDeserializationListener implementation
   void ReaderInitialized(ASTReader *Reader) override;
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 4f6987f92fc82..ce96ce2bdbcce 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -3795,6 +3795,29 @@ llvm::Error ASTReader::ReadASTBlock(ModuleFile &F,
       }
       break;
 
+    case DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD: {
+      if (Record.size() % 3 != 0)
+        return llvm::createStringError(
+            std::errc::illegal_byte_sequence,
+            "invalid DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD block in AST "
+            "file");
+      for (unsigned I = 0, N = Record.size(); I != N; I += 3) {
+        GlobalDeclID ID = getGlobalDeclID(F, Record[I]);
+
+        uint64_t BaseOffset = F.DeclsBlockStartOffset;
+        assert(BaseOffset && "Invalid DeclsBlockStartOffset for module file!");
+        uint64_t LexicalOffset = Record[I + 1] ? BaseOffset + Record[I + 1] : 0;
+        uint64_t VisibleOffset = Record[I + 2] ? BaseOffset + Record[I + 2] : 0;
+
+        DelayedNamespaceOffsetMap[ID] = {LexicalOffset, VisibleOffset};
+
+        assert(!GetExistingDecl(ID) &&
+               "We shouldn't load the namespace in the front of delayed "
+               "namespace lexical and visible block");
+      }
+      break;
+    }
+
     case OBJC_CATEGORIES_MAP:
       if (F.LocalNumObjCCategoriesInMap != 0)
         return llvm::createStringError(
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index 5b7b9f19a106c..ff72fc30c718f 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -4125,6 +4125,15 @@ Decl *ASTReader::ReadDeclRecord(DeclID ID) {
   // offsets for its tables of lexical and visible declarations.
   if (auto *DC = dyn_cast<DeclContext>(D)) {
     std::pair<uint64_t, uint64_t> Offsets = Reader.VisitDeclContext(DC);
+
+    // Get the lexical and visible block for the delayed namespace.
+    // It is sufficient to judge if ID is in DelayedNamespaceOffsetMap.
+    // But it may be more efficient to filter the other cases.
+    if (!Offsets.first && !Offsets.second && isa<NamespaceDecl>(D))
+      if (auto Iter = DelayedNamespaceOffsetMap.find(ID);
+          Iter != DelayedNamespaceOffsetMap.end())
+        Offsets = Iter->second;
+
     if (Offsets.first &&
         ReadLexicalDeclContextStorage(*Loc.F, DeclsCursor, Offsets.first, DC))
       return nullptr;
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 330dbd6233502..d9ba10ab60878 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -869,6 +869,7 @@ void ASTWriter::WriteBlockInfoBlock() {
   RECORD(WEAK_UNDECLARED_IDENTIFIERS);
   RECORD(PENDING_IMPLICIT_INSTANTIATIONS);
   RECORD(UPDATE_VISIBLE);
+  RECORD(DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD);
   RECORD(DECL_UPDATE_OFFSETS);
   RECORD(DECL_UPDATES);
   RECORD(CUDA_SPECIAL_DECL_REFS);
@@ -3026,10 +3027,12 @@ void ASTWriter::WriteSubmodules(Module *WritingModule) {
       Stream.EmitRecordWithBlob(ConfigMacroAbbrev, Record, CM);
     }
 
-    // Emit the initializers, if any.
+    // Emit the reachable initializers.
+    // The initializer may only be unreachable in reduced BMI.
     RecordData Inits;
     for (Decl *D : Context->getModuleInitializers(Mod))
-      Inits.push_back(GetDeclRef(D));
+      if (wasDeclEmitted(D))
+        Inits.push_back(GetDeclRef(D));
     if (!Inits.empty())
       Stream.EmitRecord(SUBMODULE_INITIALIZERS, Inits);
 
@@ -3208,6 +3211,9 @@ uint64_t ASTWriter::WriteDeclContextLexicalBlock(ASTContext &Context,
   uint64_t Offset = Stream.GetCurrentBitNo();
   SmallVector<uint32_t, 128> KindDeclPairs;
   for (const auto *D : DC->decls()) {
+    if (DoneWritingDeclsAndTypes && !wasDeclEmitted(D))
+      continue;
+
     KindDeclPairs.push_back(D->getKind());
     KindDeclPairs.push_back(GetDeclRef(D));
   }
@@ -3862,8 +3868,14 @@ class ASTDeclContextNameLookupTrait {
   data_type getData(const Coll &Decls) {
     unsigned Start = DeclIDs.size();
     for (NamedDecl *D : Decls) {
-      DeclIDs.push_back(
-          Writer.GetDeclRef(getDeclForLocalLookup(Writer.getLangOpts(), D)));
+      NamedDecl *DeclForLocalLookup =
+          getDeclForLocalLookup(Writer.getLangOpts(), D);
+
+      if (Writer.getDoneWritingDeclsAndTypes() &&
+          !Writer.wasDeclEmitted(DeclForLocalLookup))
+        continue;
+
+      DeclIDs.push_back(Writer.GetDeclRef(DeclForLocalLookup));
     }
     return std::make_pair(Start, DeclIDs.size());
   }
@@ -3972,11 +3984,20 @@ bool ASTWriter::isLookupResultExternal(StoredDeclsList &Result,
          DC->hasNeedToReconcileExternalVisibleStorage();
 }
 
-bool ASTWriter::isLookupResultEntirelyExternal(StoredDeclsList &Result,
-                                               DeclContext *DC) {
-  for (auto *D : Result.getLookupResult())
-    if (!getDeclForLocalLookup(getLangOpts(), D)->isFromASTFile())
-      return false;
+bool ASTWriter::isLookupResultEntirelyExternalOrUnreachable(
+    StoredDeclsList &Result, DeclContext *DC) {
+  for (auto *D : Result.getLookupResult()) {
+    auto *LocalD = getDeclForLocalLookup(getLangOpts(), D);
+    if (LocalD->isFromASTFile())
+      continue;
+
+    // We can only be sure whether the local declaration is reachable
+    // after we done writing the declarations and types.
+    if (DoneWritingDeclsAndTypes && !wasDeclEmitted(LocalD))
+      continue;
+
+    return false;
+  }
 
   return true;
 }
@@ -4014,8 +4035,17 @@ ASTWriter::GenerateNameLookupTable(const DeclContext *ConstDC,
     // don't need to write an entry for the name at all. If we can't
     // write out a lookup set without performing more deserialization,
     // just skip this entry.
-    if (isLookupResultExternal(Result, DC) &&
-        isLookupResultEntirelyExternal(Result, DC))
+    //
+    // Also in reduced BMI, we'd like to avoid writing unreachable
+    // declarations in GMF, so we need to avoid writing declarations
+    // that entirely external or unreachable.
+    //
+    // FIMXE: It looks sufficient to test
+    // isLookupResultEntirelyExternalOrUnreachable here. But due to bug we have
+    // to test isLookupResultExternal here. See
+    // https://github.com/llvm/llvm-project/issues/61065 for details.
+    if ((GeneratingReducedBMI || isLookupResultExternal(Result, DC)) &&
+        isLookupResultEntirelyExternalOrUnreachable(Result, DC))
       continue;
 
     // We also skip empty results. If any of the results could be external and
@@ -4206,9 +4236,15 @@ uint64_t ASTWriter::WriteDeclContextVisibleBlock(ASTContext &Context,
         continue;
       }
 
-      for (NamedDecl *ND : Result)
-        if (!ND->isFromASTFile())
-          GetDeclRef(ND);
+      for (NamedDecl *ND : Result) {
+        if (ND->isFromASTFile())
+          continue;
+
+        if (DoneWritingDeclsAndTypes && !wasDeclEmitted(ND))
+          continue;
+
+        GetDeclRef(ND);
+      }
     }
 
     return 0;
@@ -4976,9 +5012,18 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
   const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
 
   // Force all top level declarations to be emitted.
-  for (const auto *D : TU->noload_decls())
-    if (!D->isFromASTFile())
-      GetDeclRef(D);
+  //
+  // We start emitting top level declarations from the module purview to
+  // implement the eliding unreachable declaration feature.
+  for (const auto *D : TU->noload_decls()) {
+    if (D->isFromASTFile())
+      continue;
+
+    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
+      continue;
+
+    GetDeclRef(D);
+  }
 
   // If the translation unit has an anonymous namespace, and we don't already
   // have an update block for it, write it as an update block.
@@ -5288,24 +5333,59 @@ void ASTWriter::WriteDeclAndTypes(ASTContext &Context) {
         WriteDecl(Context, DOT.getDecl());
     }
   } while (!DeclUpdates.empty());
-  Stream.ExitBlock();
 
   DoneWritingDeclsAndTypes = true;
 
+  // DelayedNamespace is only meaningful in reduced BMI.
+  // See the comments of DelayedNamespace for details.
+  assert(DelayedNamespace.empty() || GeneratingReducedBMI);
+  RecordData DelayedNamespaceRecord;
+  for (NamespaceDecl *NS : DelayedNamespace) {
+    uint64_t LexicalOffset = WriteDeclContextLexicalBlock(Context, NS);
+    uint64_t VisibleOffset = WriteDeclContextVisibleBlock(Context, NS);
+
+    // Write the offset relative to current block.
+    if (LexicalOffset)
+      LexicalOffset -= DeclTypesBlockStartOffset;
+
+    if (VisibleOffset)
+      VisibleOffset -= DeclTypesBlockStartOffset;
+
+    DelayedNamespaceRecord.push_back(getDeclID(NS));
+    DelayedNamespaceRecord.push_back(LexicalOffset);
+    DelayedNamespaceRecord.push_back(VisibleOffset);
+  }
+
+  // The process of writing lexical and visible block for delayed namespace
+  // shouldn't introduce any new decls, types or update to emit.
+  assert(DeclTypesToEmit.empty());
+  assert(DeclUpdates.empty());
+
+  Stream.ExitBlock();
+
   // These things can only be done once we've written out decls and types.
   WriteTypeDeclOffsets();
   if (!DeclUpdatesOffsetsRecord.empty())
     Stream.EmitRecord(DECL_UPDATE_OFFSETS, DeclUpdatesOffsetsRecord);
 
+  if (!DelayedNamespaceRecord.empty())
+    Stream.EmitRecord(DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD,
+                      DelayedNamespaceRecord);
+
   const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
   // Create a lexical update block containing all of the declarations in the
   // translation unit that do not come from other AST files.
   SmallVector<uint32_t, 128> NewGlobalKindDeclPairs;
   for (const auto *D : TU->noload_decls()) {
-    if (!D->isFromASTFile()) {
-      NewGlobalKindDeclPairs.push_back(D->getKind());
-      NewGlobalKindDeclPairs.push_back(GetDeclRef(D));
-    }
+    if (D->isFromASTFile())
+      continue;
+
+    // In reduced BMI, skip unreached declarations.
+    if (!wasDeclEmitted(D))
+      continue;
+
+    NewGlobalKindDeclPairs.push_back(D->getKind());
+    NewGlobalKindDeclPairs.push_back(GetDeclRef(D));
   }
 
   auto Abv = std::make_shared<llvm::BitCodeAbbrev>();
@@ -5814,6 +5894,21 @@ DeclID ASTWriter::getDeclID(const Decl *D) {
   return DeclIDs[D];
 }
 
+bool ASTWriter::wasDeclEmitted(const Decl *D) const {
+  assert(D);
+
+  assert(DoneWritingDeclsAndTypes &&
+         "wasDeclEmitted should only be called after writing declarations");
+
+  if (D->isFromASTFile())
+    return true;
+
+  bool Emitted = DeclIDs.contains(D);
+  assert((Emitted || GeneratingReducedBMI) &&
+         "The declaration can only be omitted in reduced BMI.");
+  return Emitted;
+}
+
 void ASTWriter::associateDeclWithFile(const Decl *D, DeclID ID) {
   assert(ID);
   assert(D);
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index 87e773be54fae..4b417909751fb 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -1719,6 +1719,15 @@ void ASTDeclWriter::VisitClassTemplateDecl(ClassTemplateDecl *D) {
 
   if (D->isFirstDecl())
     AddTemplateSpecializations(D);
+
+  // Force emitting the corresponding deduction guide in reduced BMI mode.
+  // Otherwise, the deduction guide may be optimized out incorrectly.
+  if (Writer.isGeneratingReducedBMI()) {
+    auto Name = Context.DeclarationNames.getCXXDeductionGuideName(D);
+    for (auto *DG : D->getDeclContext()->noload_lookup(Name))
+      Writer.GetDeclRef(DG);
+  }
+
   Code = serialization::DECL_CLASS_TEMPLATE;
 }
 
@@ -1963,8 +1972,22 @@ void ASTDeclWriter::VisitDeclContext(DeclContext *DC) {
                 "You need to update the serializer after you change the "
                 "DeclContextBits");
 
-  Record.AddOffset(Writer.WriteDeclContextLexicalBlock(Context, DC));
-  Record.AddOffset(Writer.WriteDeclContextVisibleBlock(Context, DC));
+  uint64_t LexicalOffset = 0;
+  uint64_t VisibleOffset = 0;
+
+  if (Writer.isGeneratingReducedBMI() && isa<NamespaceDecl>(DC) &&
+      cast<NamespaceDecl>(DC)->isFromExplicitGlobalModule()) {
+    // In reduced BMI, delay writing lexical and visible block for namespace
+    // in the global module fragment. See the comments of DelayedNamespace for
+    // details.
+    Writer.DelayedNamespace.push_back(cast<NamespaceDecl>(DC));
+  } else {
+    LexicalOffset = Writer.WriteDeclContextLexicalBlock(Context, DC);
+    VisibleOffset = Writer.WriteDeclContextVisibleBlock(Context, DC);
+  }
+
+  Record.AddOffset(LexicalOffset);
+  Record.AddOffset(VisibleOffset);
 }
 
 const Decl *ASTWriter::getFirstLocalDecl(const Decl *D) {
diff --git a/clang/test/CXX/module/module.glob.frag/cxx20-10-4-ex2.cppm b/clang/test/CXX/module/module.glob.frag/cxx20-10-4-ex2.cppm
new file mode 100644
index 0000000000000..edb208cdf4de4
--- /dev/null
+++ b/clang/test/CXX/module/module.glob.frag/cxx20-10-4-ex2.cppm
@@ -0,0 +1,58 @@
+// RUN: rm -rf %t
+// RUN: mkdir %t
+// RUN: split-file %s %t
+
+// RUN: %clang_cc1 -std=c++20 %t/std-10-4-ex2-interface.cppm -emit-reduced-module-interface \
+// RUN:     -o %t/M.pcm -Wno-unused-value
+// RUN: %clang_cc1 -std=c++20 %t/std-10-4-ex2-implementation.cpp -fmodule-file=M=%t/M.pcm \
+// RUN:     -fsyntax-only -verify -Wno-unused-value
+
+//--- std-10-4-ex2.h
+
+namespace N {
+struct X {};
+int d();
+int e();
+inline int f(X, int = d()) { return e(); }
+int g(X);
+int h(X);
+} // namespace N
+
+//--- std-10-4-ex2-interface.cppm
+module;
+#include "std-10-4-ex2.h"
+export module M;
+
+template <typename T> int use_f() {
+  N::X x;           // N::X, N, and  ::  are decl-reachable from use_f
+  return f(x, 123); // N::f is decl-reachable from use_f,
+                    // N::e is indirectly decl-reachable from use_f
+                    //   because it is decl-reachable from N::f, and
+                    // N::d is decl-reachable from use_f
+                    //   because it is decl-reachable from N::f
+                    //   even though it is not used in this call
+}
+
+template <typename T> int use_g() {
+  N::X x;             // N::X, N, and :: are decl-reachable from use_g
+  return g((T(), x)); // N::g is not decl-reachable from use_g
+}
+
+template <typename T> int use_h() {
+  N::X x;             // N::X, N, and :: are decl-reachable from use_h
+  return h((T(), x)); // N::h is not decl-reachable from use_h, but
+                      // N::h is decl-reachable from use_h<int>
+}
+
+int k = use_h<int>();
+// use_h<int> is decl-reachable from k, so
+// N::h is decl-reachable from k
+
+//--- std-10-4-ex2-implementation.cpp
+module M;
+
+int a = use_f<int>();
+int b = use_g<int>();
+// expected-error@std-10-4-ex2-interface.cppm:17 {{use of undeclared identifier 'g'}}
+// expected-note@-2 {{in instantiation of function template specialization 'use_g<int>' requested here}}
+int c = use_h<int>();
diff --git a/clang/test/Modules/inconsistent-deduction-guide-linkage.cppm b/clang/test/Modules/inconsistent-deduction-guide-linkage.cppm
index 3991e47ce2151..eaf033d1d9283 100644
--- a/clang/test/Modules/inconsistent-deduction-guide-linkage.cppm
+++ b/clang/test/Modules/inconsistent-deduction-guide-linkage.cppm
@@ -26,6 +26,10 @@ module;
 #include "C.h"
 export module B;
 
+namespace foo {
+    export using foo::bar;
+}
+
 //--- C.h
 namespace foo {
   template<class T, class U> struct bar { // expected-error {{declaration of 'bar' in module baz:A follows declaration in the global module}} // expected-note {{previous declaration is here}}
@@ -40,6 +44,10 @@ module;
 #include "E.h"
 export module D;
 
+namespace foo {
+    export using foo::bar;
+}
+
 //--- D-part.cppm
 export module D:part;
 import D;
diff --git a/clang/test/Modules/named-modules-adl-2.cppm b/clang/test/Modules/named-modules-adl-2.cppm
index a14b9a68d74e4..88f58d85bc7c7 100644
--- a/clang/test/Modules/named-modules-adl-2.cppm
+++ b/clang/test/Modules/named-modules-adl-2.cppm
@@ -7,7 +7,7 @@
 // RUN: %clang_cc1 -std=c++20 %t/c.cppm -fmodule-file=a=%t/a.pcm -fmodule-file=b=%t/b.pcm -fsyntax-only -verify
 
 // RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-reduced-module-interface -o %t/a.pcm
-// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -emit-reduced-module-interface -o %t/b.pcm
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -emit-reduced-module-interface -o %t/b.pcm -DREDUCED
 // RUN: %clang_cc1 -std=c++20 %t/c.cppm -fmodule-file=a=%t/a.pcm -fmodule-file=b=%t/b.pcm -fsyntax-only -verify
 
 //--- a.cppm
@@ -35,6 +35,11 @@ void b() {
 	a(s());
 }
 
+#ifdef REDUCED
+// Mention it to avoid the compiler optimizing it out.
+using ::operator+;
+#endif
+
 //--- c.cppm
 // expected-no-diagnostics
 export module c;
diff --git a/clang/test/Modules/named-modules-adl.cppm b/clang/test/Modules/named-modules-adl.cppm
index ef250023f91e7..079d816200b2c 100644
--- a/clang/test/Modules/named-modules-adl.cppm
+++ b/clang/test/Modules/named-modules-adl.cppm
@@ -5,7 +5,7 @@
 // RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-module-interface -o %t/a.pcm
 // RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
 
-// RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-reduced-module-interface -o %t/a.pcm
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-reduced-module-interface -o %t/a.pcm -DREDUCED
 // RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
 
 //--- a.h
@@ -28,6 +28,11 @@ void a(T x) {
 	n::s() + x;
 }
 
+#ifdef REDUCED
+// Use it to make sure it is not optimized out in reduced BMI.
+using n::operator+;
+#endif
+
 //--- b.cppm
 // expected-no-diagnostics
 export module b;
diff --git a/clang/test/Modules/placement-new-reachable.cpp b/clang/test/Modules/placement-new-reachable.cpp
index 6b495a60306bc..2440294704742 100644
--- a/clang/test/Modules/placement-new-reachable.cpp
+++ b/clang/test/Modules/placement-new-reachable.cpp
@@ -31,12 +31,17 @@ export struct B {
     void *ptr;
 };
 
+// The use of operator new in the current module unit is only in the non-inline
+// function definitions. So it may be optimized out.
+using ::operator new;
+
 //--- Use.cpp
 // expected-no-diagnostics
 import A;
 void bar(int *);
 void foo(void *ptr) {
-    A<int>(nullptr); // Good. It should be OK to construct A.
-    void *p = ::operator new(sizeof(int), ptr); // Bad. The function shouldn't be visible here.
+    A<int> a(nullptr); // Good. It should be OK to construct A.
+    B b(nullptr);
+    void *p = ::operator new(sizeof(int), ptr); // Bad. The placement allocation in module A is not visible.
     void *q = new (ptr) int(43); // Good. We don't call the placement allocation function directly.
 }
diff --git a/clang/test/Modules/polluted-operator.cppm b/clang/test/Modules/polluted-operator.cppm
index 2179fa098064a..45cc5e37d6a64 100644
--- a/clang/test/Modules/polluted-operator.cppm
+++ b/clang/test/Modules/polluted-operator.cppm
@@ -12,7 +12,8 @@
 // RUN:   -emit-module-interface -DSKIP_ODR_CHECK_IN_GMF -o %t/b.pcm -verify
 
 // RUN: %clang_cc1 -std=c++20 -emit-reduced-module-interface %t/a.cppm -o %t/a.pcm
-// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fprebuilt-module-path=%t -emit-reduced-module-interface -o %t/b.pcm -verify
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fprebuilt-module-path=%t -emit-reduced-module-interface \
+// RUN:     -o %t/b.pcm -verify -DREDUCED
 
 //--- foo.h
 
@@ -53,18 +54,26 @@ module;
 #include "foo.h"
 #include "bar.h"
 export module a;
+export namespace std {
+  using std::variant;
+  using std::_Traits;
+  using std::operator&&;
+}
 
 //--- b.cppm
 module;
 #include "bar.h"
 export module b;
 import a;
+export namespace std {
+  using std::variant;
+  using std::_Traits;
+  using std::operator&&;
+}
 
 #ifdef SKIP_ODR_CHECK_IN_GMF
 // expected-no-diagnostics
 #else
 // expected-error@* {{has different definitions in different modules; first difference is defined here found data member '_S_copy_ctor' with an initializer}}
 // expected-note@* {{but in 'a.<global>' found data member '_S_copy_ctor' with a different initializer}}
-// expected-error@* {{from module 'a.<global>' is not present in definition of 'variant<_Types...>' provided earlier}}
-// expected-note@* {{declaration of 'swap' does not match}}
 #endif
diff --git a/clang/test/Modules/pr62589.cppm b/clang/test/Modules/pr62589.cppm
index c5aec3ed81846..54f2ecef22e18 100644
--- a/clang/test/Modules/pr62589.cppm
+++ b/clang/test/Modules/pr62589.cppm
@@ -73,6 +73,10 @@ export module a;
 export using ::a;
 export using ::a_view;
 
+// We need to mention the 'operator==' explicitly to make sure it won't be
+// discarded.
+export using ::operator==;
+
 //--- b.cpp
 // expected-no-diagnostics
 import a;
diff --git a/clang/test/Modules/preferred_name.cppm b/clang/test/Modules/preferred_name.cppm
index 2f17058678455..806781a81c5ca 100644
--- a/clang/test/Modules/preferred_name.cppm
+++ b/clang/test/Modules/preferred_name.cppm
@@ -42,6 +42,7 @@ inline foo_templ<char> bar()
 module;
 #include "foo.h"
 export module A;
+export using ::foo_templ;
 
 //--- Use.cppm
 // expected-no-diagnostics
@@ -49,6 +50,7 @@ module;
 #include "foo.h"
 export module Use;
 import A;
+export using ::foo_templ;
 
 //--- Use1.cpp
 import A;         // expected-warning@foo.h:8 {{attribute declaration must precede definition}}
diff --git a/clang/test/Modules/redundant-template-default-arg3.cpp b/clang/test/Modules/redundant-template-default-arg3.cpp
index e4464c40e9768..da6fe01a5357b 100644
--- a/clang/test/Modules/redundant-template-default-arg3.cpp
+++ b/clang/test/Modules/redundant-template-default-arg3.cpp
@@ -91,6 +91,16 @@ int v9;
 module;
 #include "foo.h"
 export module foo;
+export using ::v;
+export using ::v2;
+export using ::my_array;
+export using ::v3;
+export using ::v4;
+export using ::v5;
+export using ::v6;
+export using ::v7;
+export using ::v8;
+export using ::v9;
 
 //--- use.cpp
 import foo;
diff --git a/clang/test/Modules/template-function-specialization.cpp b/clang/test/Modules/template-function-specialization.cpp
index 1b6bf2de6ba1d..d5d7d7e812398 100644
--- a/clang/test/Modules/template-function-specialization.cpp
+++ b/clang/test/Modules/template-function-specialization.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++20 -fprebuilt-module-path=%t %t/Use.cpp -verify -fsyntax-only
 
 // RUN: %clang_cc1 -std=c++20 -emit-reduced-module-interface %t/foo.cppm -o %t/foo.pcm
-// RUN: %clang_cc1 -std=c++20 -fprebuilt-module-path=%t %t/Use.cpp -verify -fsyntax-only
+// RUN: %clang_cc1 -std=c++20 -fprebuilt-module-path=%t %t/Use.cpp -verify -fsyntax-only -DREDUCED
 
 //--- foo.cppm
 module;
@@ -51,10 +51,16 @@ import foo;
 void use() {
   foo<short>();
   foo<int>();
+#ifdef REDUCED
+  // In reduced BMI, the foo2 template function is optimized out.
+  foo2<short>(); // expected-error {{use of undeclared identifier 'foo2'}}
+  foo2<int>();   // expected-error {{use of undeclared identifier 'foo2'}}
+#else
   foo2<short>(); // expected-error {{missing '#include'; 'foo2' must be declared before it is used}}
                  // expected-note@* {{declaration here is not visible}}
   foo2<int>();   // expected-error {{missing '#include'; 'foo2' must be declared before it is used}}
                  // expected-note@* {{declaration here is not visible}}
+#endif
   foo3<short>();
   foo3<int>();
 

From 506ff547924ab81123865964b7d998a5a378330a Mon Sep 17 00:00:00 2001
From: NimishMishra <42909663+NimishMishra@users.noreply.github.com>
Date: Thu, 11 Apr 2024 22:12:14 -0700
Subject: [PATCH 218/886] [flang][Semantics] Fix updating flags of
 threadprivate symbols in presence of default clause (#78283)

Current semantic checks of default clause incorrectly update symbol
flags related to threadprivate symbols. This patch adds an additional
check to skip such updation should a symbol be already declared
threadprivate.

Fixes https://github.com/llvm/llvm-project/issues/78282
---
 flang/lib/Semantics/resolve-directives.cpp |  5 ++++
 flang/test/Lower/OpenMP/default-clause.f90 | 29 ++++++++++++++++++++++
 2 files changed, 34 insertions(+)

diff --git a/flang/lib/Semantics/resolve-directives.cpp b/flang/lib/Semantics/resolve-directives.cpp
index 27af192f606be..f4ac7f198d854 100644
--- a/flang/lib/Semantics/resolve-directives.cpp
+++ b/flang/lib/Semantics/resolve-directives.cpp
@@ -2021,6 +2021,11 @@ void OmpAttributeVisitor::Post(const parser::Name &name) {
         }
       }
     }
+
+    if (Symbol * found{currScope().FindSymbol(name.source)}) {
+      if (found->test(semantics::Symbol::Flag::OmpThreadprivate))
+        return;
+    }
     std::vector<Symbol *> defaultDSASymbols;
     for (int dirDepth{0}; dirDepth < (int)dirContext_.size(); ++dirDepth) {
       DirContext &dirContext = dirContext_[dirDepth];
diff --git a/flang/test/Lower/OpenMP/default-clause.f90 b/flang/test/Lower/OpenMP/default-clause.f90
index f86b51aef4e02..9a47e561338f1 100644
--- a/flang/test/Lower/OpenMP/default-clause.f90
+++ b/flang/test/Lower/OpenMP/default-clause.f90
@@ -383,3 +383,32 @@ subroutine skipped_default_clause_checks()
           iii=it(11)
        !$omp end parallel
 end subroutine
+
+!CHECK: func.func @_QPthreadprivate_with_default() {
+!CHECK: %[[VAR_I:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFthreadprivate_with_defaultEi"}
+!CHECK: %[[VAR_I_DECLARE:.*]] = hlfir.declare %[[VAR_I]] {uniq_name = "_QFthreadprivate_with_defaultEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+!CHECK: %[[BLK_ADDR:.*]] = fir.address_of(@blk_) : !fir.ref<!fir.array<4xi8>>
+!CHECK: %[[BLK_THREADPRIVATE_OUTER:.*]] = omp.threadprivate %[[BLK_ADDR]] : !fir.ref<!fir.array<4xi8>> -> !fir.ref<!fir.array<4xi8>>
+!CHECK: %[[CONVERT:.*]] = fir.convert %[[BLK_THREADPRIVATE_OUTER]] : (!fir.ref<!fir.array<4xi8>>) -> !fir.ref<!fir.array<?xi8>>
+!CHECK: %[[VAR_C:.*]] = arith.constant 0 : index
+!CHECK: %[[BLK_REF:.*]] = fir.coordinate_of %[[CONVERT]], %[[VAR_C]] : (!fir.ref<!fir.array<?xi8>>, index) -> !fir.ref<i8>
+!CHECK: %[[CONVERT:.*]] = fir.convert %[[BLK_REF]] : (!fir.ref<i8>) -> !fir.ref<i32>
+!CHECK: %[[VAR_X_DECLARE:.*]] = hlfir.declare %[[CONVERT]] {uniq_name = "_QFthreadprivate_with_defaultEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+!CHECK: omp.parallel {
+!CHECK:   %[[BLK_THREADPRIVATE_INNER:.*]] = omp.threadprivate %[[BLK_ADDR]] : !fir.ref<!fir.array<4xi8>> -> !fir.ref<!fir.array<4xi8>>
+!CHECK:   %[[CONVERT_INNER:.*]] = fir.convert %[[BLK_THREADPRIVATE_INNER]] : (!fir.ref<!fir.array<4xi8>>) -> !fir.ref<!fir.array<?xi8>>
+!CHECK:   %[[VAR_C_INNER:.*]] = arith.constant 0 : index
+!CHECK:   %[[BLK_REF_INNER:.*]] = fir.coordinate_of %[[CONVERT_INNER]], %[[VAR_C_INNER]] : (!fir.ref<!fir.array<?xi8>>, index) -> !fir.ref<i8>
+!CHECK:   %[[CONVERT_INNER:.*]] = fir.convert %[[BLK_REF_INNER]] : (!fir.ref<i8>) -> !fir.ref<i32>
+!CHECK:   %[[VAR_X_DECLARE_INNER:.*]] = hlfir.declare %[[CONVERT_INNER]] {uniq_name = "_QFthreadprivate_with_defaultEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+subroutine threadprivate_with_default
+        integer :: x
+        common /blk/ x 
+        !$omp threadprivate (/blk/)
+
+        !$omp parallel do default(private)
+           do i = 1, 4
+              x = 4
+           end do
+        !$omp end parallel do
+end subroutine

From b502dc596ba3ed1e857e163e21ed6426125760ef Mon Sep 17 00:00:00 2001
From: Sirraide <aeternalmail@gmail.com>
Date: Fri, 12 Apr 2024 07:12:57 +0200
Subject: [PATCH 219/886] [Clang] Fix test broken due to unsupported calling
 convention (#88472)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

#88428 ended up breaking CI because it included a test that uses the
`regcall` calling convention, which isn’t supported on all targets; I’ve
moved it into a separate file that sets the triple.
---
 clang/test/SemaCXX/gh84473.cpp            | 11 +++++++++++
 clang/test/SemaCXX/lambda-expressions.cpp |  9 ---------
 2 files changed, 11 insertions(+), 9 deletions(-)
 create mode 100644 clang/test/SemaCXX/gh84473.cpp

diff --git a/clang/test/SemaCXX/gh84473.cpp b/clang/test/SemaCXX/gh84473.cpp
new file mode 100644
index 0000000000000..72f5e9c45b0c4
--- /dev/null
+++ b/clang/test/SemaCXX/gh84473.cpp
@@ -0,0 +1,11 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -std=c++23 -fsyntax-only -verify %s
+// expected-no-diagnostics
+
+namespace GH84473_bug {
+void f1() {
+  int b;
+  (void) [=] [[gnu::regcall]] () {
+    (void) b;
+  };
+}
+}
diff --git a/clang/test/SemaCXX/lambda-expressions.cpp b/clang/test/SemaCXX/lambda-expressions.cpp
index 6be338064452e..151d74f21d64d 100644
--- a/clang/test/SemaCXX/lambda-expressions.cpp
+++ b/clang/test/SemaCXX/lambda-expressions.cpp
@@ -762,12 +762,3 @@ template auto t::operator()<int>(int a) const; // expected-note {{in instantiati
 
 }
 #endif
-
-namespace GH84473_bug {
-void f1() {
-  int b;
-  (void) [=] [[gnu::regcall]] () { // expected-warning {{an attribute specifier sequence in this position is a C++23 extension}}
-    (void) b;
-  };
-}
-}

From 469758c944dfb492102ded9e50eac5e478f85f6a Mon Sep 17 00:00:00 2001
From: Dominik Adamski <dominik.adamski@amd.com>
Date: Fri, 12 Apr 2024 09:11:26 +0200
Subject: [PATCH 220/886] [Flang] Add options frtlib-add-rpath and resource-dir
 (#88280)

frtlib-add-rpath option is used by clang and should also be visible in
flang. It is already handled by the toolchains used by both clang and
flang.

Reported issue: https://github.com/llvm/llvm-project/issues/82553
---
 clang/include/clang/Driver/Options.td         |  6 ++--
 clang/lib/Driver/ToolChains/Flang.cpp         |  4 +++
 .../Driver/arch-specific-libdir-rpath.f95     | 32 +++++++++++++++++++
 flang/test/Driver/driver-help-hidden.f90      |  3 ++
 flang/test/Driver/driver-help.f90             |  2 ++
 5 files changed, 45 insertions(+), 2 deletions(-)
 create mode 100644 flang/test/Driver/arch-specific-libdir-rpath.f95

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 7ac36222644aa..400f72f2250a2 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -5464,21 +5464,23 @@ def rdynamic : Flag<["-"], "rdynamic">, Group<Link_Group>,
   Visibility<[ClangOption, FlangOption]>;
 def resource_dir : Separate<["-"], "resource-dir">,
   Flags<[NoXarchOption, HelpHidden]>,
-  Visibility<[ClangOption, CC1Option, CLOption, DXCOption]>,
+  Visibility<[ClangOption, CC1Option, CLOption, DXCOption, FlangOption, FC1Option]>,
   HelpText<"The directory which holds the compiler resource files">,
   MarshallingInfoString<HeaderSearchOpts<"ResourceDir">>;
 def resource_dir_EQ : Joined<["-"], "resource-dir=">, Flags<[NoXarchOption]>,
-  Visibility<[ClangOption, CLOption, DXCOption]>,
+  Visibility<[ClangOption, CLOption, DXCOption, FlangOption]>,
   Alias<resource_dir>;
 def rpath : Separate<["-"], "rpath">, Flags<[LinkerInput]>, Group<Link_Group>,
   Visibility<[ClangOption, CLOption, DXCOption, FlangOption]>;
 def rtlib_EQ : Joined<["-", "--"], "rtlib=">, Visibility<[ClangOption, CLOption]>,
   HelpText<"Compiler runtime library to use">;
 def frtlib_add_rpath: Flag<["-"], "frtlib-add-rpath">, Flags<[NoArgumentUnused]>,
+  Visibility<[ClangOption, FlangOption]>,
   HelpText<"Add -rpath with architecture-specific resource directory to the linker flags. "
   "When --hip-link is specified, also add -rpath with HIP runtime library directory to the linker flags">;
 def fno_rtlib_add_rpath: Flag<["-"], "fno-rtlib-add-rpath">,
   Flags<[NoArgumentUnused]>,
+  Visibility<[ClangOption, FlangOption]>,
   HelpText<"Do not add -rpath with architecture-specific resource directory to the linker flags. "
   "When --hip-link is specified, do not add -rpath with HIP runtime library directory to the linker flags">;
 def offload_add_rpath: Flag<["--"], "offload-add-rpath">,
diff --git a/clang/lib/Driver/ToolChains/Flang.cpp b/clang/lib/Driver/ToolChains/Flang.cpp
index b00068c8098b9..b46bac24503ce 100644
--- a/clang/lib/Driver/ToolChains/Flang.cpp
+++ b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -786,6 +786,10 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
     }
   }
 
+  // Pass the path to compiler resource files.
+  CmdArgs.push_back("-resource-dir");
+  CmdArgs.push_back(D.ResourceDir.c_str());
+
   // Offloading related options
   addOffloadOptions(C, Inputs, JA, Args, CmdArgs);
 
diff --git a/flang/test/Driver/arch-specific-libdir-rpath.f95 b/flang/test/Driver/arch-specific-libdir-rpath.f95
new file mode 100644
index 0000000000000..cc09938f7d1e2
--- /dev/null
+++ b/flang/test/Driver/arch-specific-libdir-rpath.f95
@@ -0,0 +1,32 @@
+! REQUIRES: x86-registered-target
+! Test that the driver adds an arch-specific subdirectory in
+! {RESOURCE_DIR}/lib/linux to the linker search path and to '-rpath'
+!
+! Test the default behavior when neither -frtlib-add-rpath nor
+! -fno-rtlib-add-rpath is specified, which is to skip -rpath
+! RUN: %flang %s -### --target=x86_64-linux \
+! RUN:     -resource-dir=%S/../../../clang/test/Driver/Inputs/resource_dir_with_arch_subdir 2>&1 \
+! RUN:   | FileCheck --check-prefixes=RESDIR,LIBPATH-X86_64,NO-RPATH-X86_64 %s
+!
+! Test that -rpath is not added under -fno-rtlib-add-rpath
+! RUN: %flang %s -### --target=x86_64-linux \
+! RUN:     -resource-dir=%S/../../../clang/test/Driver/Inputs/resource_dir_with_arch_subdir \
+! RUN:     -fno-rtlib-add-rpath 2>&1 \
+! RUN:   | FileCheck --check-prefixes=RESDIR,LIBPATH-X86_64,NO-RPATH-X86_64 %s
+!
+! Test that -rpath is added
+!
+! Add LIBPATH, RPATH for OpenMP
+!
+! RUN: %flang %s -### --target=x86_64-linux -fopenmp \
+! RUN:     -resource-dir=%S/../../../clang/test/Driver/Inputs/resource_dir_with_arch_subdir \
+! RUN:     -frtlib-add-rpath 2>&1 \
+! RUN:   | FileCheck --check-prefixes=RESDIR,LIBPATH-X86_64,RPATH-X86_64 %s
+!
+!
+! RESDIR: "-resource-dir" "[[RESDIR:[^"]*]]"
+!
+! LIBPATH-X86_64: -L[[RESDIR]]{{(/|\\\\)lib(/|\\\\)linux(/|\\\\)x86_64}}
+! RPATH-X86_64:   "-rpath" "[[RESDIR]]{{(/|\\\\)lib(/|\\\\)linux(/|\\\\)x86_64}}"
+!
+! NO-RPATH-X86_64-NOT:   "-rpath" "[[RESDIR]]{{(/|\\\\)lib(/|\\\\)linux(/|\\\\)x86_64}}"
diff --git a/flang/test/Driver/driver-help-hidden.f90 b/flang/test/Driver/driver-help-hidden.f90
index 35b188ad3a9ec..eb00949c2a951 100644
--- a/flang/test/Driver/driver-help-hidden.f90
+++ b/flang/test/Driver/driver-help-hidden.f90
@@ -68,6 +68,7 @@
 ! CHECK-NEXT: -fno-lto                Disable LTO mode (default)
 ! CHECK-NEXT: -fno-ppc-native-vector-element-order
 ! CHECK-NEXT:                         Specifies PowerPC non-native vector element order
+! CHECK-NEXT: -fno-rtlib-add-rpath Do not add -rpath with architecture-specific resource directory to the linker flags. When --hip-link is specified, do not add -rpath with HIP runtime library directory to the linker flags
 ! CHECK-NEXT: -fno-signed-zeros       Allow optimizations that ignore the sign of floating point zeros
 ! CHECK-NEXT: -fno-stack-arrays       Allocate array temporaries on the heap (default)
 ! CHECK-NEXT: -fno-version-loops-for-stride
@@ -93,6 +94,7 @@
 ! CHECK-NEXT:                         Specifies PowerPC native vector element order (default)
 ! CHECK-NEXT: -freciprocal-math       Allow division operations to be reassociated
 ! CHECK-NEXT: -fropi                  Generate read-only position independent code (ARM only)
+! CHECK-NEXT: -frtlib-add-rpath Add -rpath with architecture-specific resource directory to the linker flags. When --hip-link is specified, also add -rpath with HIP runtime library directory to the linker flags
 ! CHECK-NEXT: -frwpi                  Generate read-write position independent code (ARM only)
 ! CHECK-NEXT: -fsave-optimization-record=<format>
 ! CHECK-NEXT:                         Generate an optimization record file in a specific format
@@ -145,6 +147,7 @@
 ! CHECK-NEXT: -print-target-triple    Print the normalized target triple
 ! CHECK-NEXT: -pthread                Support POSIX threads in generated code
 ! CHECK-NEXT: -P                      Disable linemarker output in -E mode
+! CHECK-NEXT: -resource-dir <value>   The directory which holds the compiler resource files
 ! CHECK-NEXT: -Rpass-analysis=<value> Report transformation analysis from optimization passes whose name matches the given POSIX regular expression
 ! CHECK-NEXT: -Rpass-missed=<value>   Report missed transformations by optimization passes whose name matches the given POSIX regular expression
 ! CHECK-NEXT: -Rpass=<value>          Report transformations performed by optimization passes whose name matches the given POSIX regular expression
diff --git a/flang/test/Driver/driver-help.f90 b/flang/test/Driver/driver-help.f90
index d4dab55f40e8b..05f50ea348c22 100644
--- a/flang/test/Driver/driver-help.f90
+++ b/flang/test/Driver/driver-help.f90
@@ -60,6 +60,7 @@
 ! HELP-NEXT: -fno-lto                Disable LTO mode (default)
 ! HELP-NEXT: -fno-ppc-native-vector-element-order
 ! HELP-NEXT:                         Specifies PowerPC non-native vector element order
+! HELP-NEXT:  -fno-rtlib-add-rpath Do not add -rpath with architecture-specific resource directory to the linker flags. When --hip-link is specified, do not add -rpath with HIP runtime library directory to the linker flags
 ! HELP-NEXT: -fno-signed-zeros       Allow optimizations that ignore the sign of floating point zeros
 ! HELP-NEXT: -fno-stack-arrays       Allocate array temporaries on the heap (default)
 ! HELP-NEXT: -fno-version-loops-for-stride
@@ -81,6 +82,7 @@
 ! HELP-NEXT:                         Specifies PowerPC native vector element order (default)
 ! HELP-NEXT: -freciprocal-math       Allow division operations to be reassociated
 ! HELP-NEXT: -fropi                  Generate read-only position independent code (ARM only)
+! HELP-NEXT: -frtlib-add-rpath Add -rpath with architecture-specific resource directory to the linker flags. When --hip-link is specified, also add -rpath with HIP runtime library directory to the linker flags
 ! HELP-NEXT: -frwpi                  Generate read-write position independent code (ARM only)
 ! HELP-NEXT: -fsave-optimization-record=<format>
 ! HELP-NEXT:                         Generate an optimization record file in a specific format

From 1b8830c56abd01c9ab70bbbb71a00c2506cf2116 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Fri, 12 Apr 2024 08:52:18 +0200
Subject: [PATCH 221/886] [clang][Interp] Fix comparing to integral function
 pointers

We need to account for the fact that the Function pointer is not
accessible here. Add FunctionPointer::isWeak() for that.
---
 clang/lib/AST/Interp/FunctionPointer.h |  6 ++++++
 clang/lib/AST/Interp/Interp.h          |  2 +-
 clang/test/AST/Interp/c.c              | 10 ++++++++--
 3 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/clang/lib/AST/Interp/FunctionPointer.h b/clang/lib/AST/Interp/FunctionPointer.h
index 840c1101f396b..c2ea295b82bdf 100644
--- a/clang/lib/AST/Interp/FunctionPointer.h
+++ b/clang/lib/AST/Interp/FunctionPointer.h
@@ -32,6 +32,12 @@ class FunctionPointer final {
 
   const Function *getFunction() const { return Func; }
   bool isZero() const { return !Func; }
+  bool isWeak() const {
+    if (!Func || !Valid)
+      return false;
+
+    return Func->getDecl()->isWeak();
+  }
 
   APValue toAPValue() const {
     if (!Func)
diff --git a/clang/lib/AST/Interp/Interp.h b/clang/lib/AST/Interp/Interp.h
index c7012aa4ec680..4182254357eb9 100644
--- a/clang/lib/AST/Interp/Interp.h
+++ b/clang/lib/AST/Interp/Interp.h
@@ -758,7 +758,7 @@ inline bool CmpHelperEQ<FunctionPointer>(InterpState &S, CodePtr OpPC,
 
   // We cannot compare against weak declarations at compile time.
   for (const auto &FP : {LHS, RHS}) {
-    if (!FP.isZero() && FP.getFunction()->getDecl()->isWeak()) {
+    if (FP.isWeak()) {
       const SourceInfo &Loc = S.Current->getSource(OpPC);
       S.FFDiag(Loc, diag::note_constexpr_pointer_weak_comparison)
           << FP.toDiagnosticString(S.getCtx());
diff --git a/clang/test/AST/Interp/c.c b/clang/test/AST/Interp/c.c
index cdecd3e83a997..e0b18120fd211 100644
--- a/clang/test/AST/Interp/c.c
+++ b/clang/test/AST/Interp/c.c
@@ -1,7 +1,7 @@
 // RUN: %clang_cc1 -triple x86_64-linux -fexperimental-new-constant-interpreter -verify=expected,all -std=c11 -Wcast-qual %s
-// RUN: %clang_cc1 -triple x86_64-linux -fexperimental-new-constant-interpreter -pedantic -verify=pedantic-expected,all -std=c11 -Wcast-qual %s
+// RUN: %clang_cc1 -triple x86_64-linux -fexperimental-new-constant-interpreter -pedantic -verify=pedantic,pedantic-expected,all -std=c11 -Wcast-qual %s
 // RUN: %clang_cc1 -triple x86_64-linux -verify=ref,all -std=c11 -Wcast-qual %s
-// RUN: %clang_cc1 -triple x86_64-linux -pedantic -verify=pedantic-ref,all -std=c11 -Wcast-qual %s
+// RUN: %clang_cc1 -triple x86_64-linux -pedantic -verify=pedantic,pedantic-ref,all -std=c11 -Wcast-qual %s
 
 typedef __INTPTR_TYPE__ intptr_t;
 typedef __PTRDIFF_TYPE__ ptrdiff_t;
@@ -227,3 +227,9 @@ int castViaInt[*(int*)(unsigned long)"test"]; // ref-error {{variable length arr
                                               // pedantic-ref-error {{variable length array}} \
                                               // expected-error {{variable length array}} \
                                               // pedantic-expected-error {{variable length array}}
+
+const void (*const funcp)(void) = (void*)123; // pedantic-warning {{converts between void pointer and function pointer}}
+_Static_assert(funcp == (void*)0, ""); // all-error {{failed due to requirement 'funcp == (void *)0'}} \
+                                       // pedantic-warning {{expression is not an integer constant expression}}
+_Static_assert(funcp == (void*)123, ""); // pedantic-warning {{equality comparison between function pointer and void pointer}} \
+                                         // pedantic-warning {{expression is not an integer constant expression}}

From ff9bc3a0fa4e3790c437e62450903bf399df37f9 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Fri, 12 Apr 2024 09:42:04 +0200
Subject: [PATCH 222/886] Fix overflows in complex sqrt lowering. (#88480)

This ports XLA's complex sqrt lowering. The accuracy was tested with its
exhaustive_unary_test_complex test.

Note: rsqrt is still broken.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 166 ++++++----
 .../convert-to-standard.mlir                  | 296 +++++++++++-------
 .../ComplexToStandard/full-conversion.mlir    |   2 +-
 3 files changed, 280 insertions(+), 184 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 9c82e8105f06e..0664b053fc9e6 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -27,35 +27,52 @@ using namespace mlir;
 
 namespace {
 
+// Returns the absolute value or its square root.
+Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
+                 ImplicitLocOpBuilder &b, bool returnSqrt = false) {
+  Value one = b.create<arith::ConstantOp>(real.getType(),
+                                          b.getFloatAttr(real.getType(), 1.0));
+
+  Value absReal = b.create<math::AbsFOp>(real, fmf);
+  Value absImag = b.create<math::AbsFOp>(imag, fmf);
+
+  Value max = b.create<arith::MaximumFOp>(absReal, absImag, fmf);
+  Value min = b.create<arith::MinimumFOp>(absReal, absImag, fmf);
+  Value ratio = b.create<arith::DivFOp>(min, max, fmf);
+  Value ratioSq = b.create<arith::MulFOp>(ratio, ratio, fmf);
+  Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
+  Value result;
+
+  if (returnSqrt) {
+    Value quarter = b.create<arith::ConstantOp>(
+        real.getType(), b.getFloatAttr(real.getType(), 0.25));
+    // sqrt(sqrt(a*b)) would avoid the pow, but will overflow more easily.
+    Value sqrt = b.create<math::SqrtOp>(max, fmf);
+    Value p025 = b.create<math::PowFOp>(ratioSqPlusOne, quarter, fmf);
+    result = b.create<arith::MulFOp>(sqrt, p025, fmf);
+  } else {
+    Value sqrt = b.create<math::SqrtOp>(ratioSqPlusOne, fmf);
+    result = b.create<arith::MulFOp>(max, sqrt, fmf);
+  }
+
+  Value isNaN =
+      b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, result, result, fmf);
+  return b.create<arith::SelectOp>(isNaN, min, result);
+}
+
 struct AbsOpConversion : public OpConversionPattern<complex::AbsOp> {
   using OpConversionPattern<complex::AbsOp>::OpConversionPattern;
 
   LogicalResult
   matchAndRewrite(complex::AbsOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
+    ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
     arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
 
-    Type elementType = op.getType();
-    Value one = b.create<arith::ConstantOp>(elementType,
-                                            b.getFloatAttr(elementType, 1.0));
-
     Value real = b.create<complex::ReOp>(adaptor.getComplex());
     Value imag = b.create<complex::ImOp>(adaptor.getComplex());
-    Value absReal = b.create<math::AbsFOp>(real, fmf);
-    Value absImag = b.create<math::AbsFOp>(imag, fmf);
-
-    Value max = b.create<arith::MaximumFOp>(absReal, absImag, fmf);
-    Value min = b.create<arith::MinimumFOp>(absReal, absImag, fmf);
-    Value ratio = b.create<arith::DivFOp>(min, max, fmf);
-    Value ratioSq = b.create<arith::MulFOp>(ratio, ratio, fmf);
-    Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
-    Value sqrt = b.create<math::SqrtOp>(ratioSqPlusOne, fmf);
-    Value result = b.create<arith::MulFOp>(max, sqrt, fmf);
-    Value isNaN =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, result, result, fmf);
-    rewriter.replaceOpWithNewOp<arith::SelectOp>(op, isNaN, min, result);
+    rewriter.replaceOp(op, computeAbs(real, imag, fmf, b));
 
     return success();
   }
@@ -829,60 +846,71 @@ struct SqrtOpConversion : public OpConversionPattern<complex::SqrtOp> {
   LogicalResult
   matchAndRewrite(complex::SqrtOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
+    ImplicitLocOpBuilder b(op.getLoc(), rewriter);
 
     auto type = cast<ComplexType>(op.getType());
-    Type elementType = type.getElementType();
-    Value arg = adaptor.getComplex();
-    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
-
-    Value zero =
-        b.create<arith::ConstantOp>(elementType, b.getZeroAttr(elementType));
-
-    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
-    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
-
-    Value absLhs = b.create<math::AbsFOp>(real, fmf);
-    Value absArg = b.create<complex::AbsOp>(elementType, arg, fmf);
-    Value addAbs = b.create<arith::AddFOp>(absLhs, absArg, fmf);
+    auto elementType = type.getElementType().cast<FloatType>();
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
 
+    auto cst = [&](APFloat v) {
+      return b.create<arith::ConstantOp>(elementType,
+                                         b.getFloatAttr(elementType, v));
+    };
+    const auto &floatSemantics = elementType.getFloatSemantics();
+    Value zero = cst(APFloat::getZero(floatSemantics));
     Value half = b.create<arith::ConstantOp>(elementType,
                                              b.getFloatAttr(elementType, 0.5));
-    Value halfAddAbs = b.create<arith::MulFOp>(addAbs, half, fmf);
-    Value sqrtAddAbs = b.create<math::SqrtOp>(halfAddAbs, fmf);
-
-    Value realIsNegative =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OLT, real, zero);
-    Value imagIsNegative =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OLT, imag, zero);
-
-    Value resultReal = sqrtAddAbs;
-
-    Value imagDivTwoResultReal = b.create<arith::DivFOp>(
-        imag, b.create<arith::AddFOp>(resultReal, resultReal, fmf), fmf);
-
-    Value negativeResultReal = b.create<arith::NegFOp>(resultReal);
 
+    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
+    Value absSqrt = computeAbs(real, imag, fmf, b, /*returnSqrt=*/true);
+    Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
+    Value sqrtArg = b.create<arith::MulFOp>(argArg, half, fmf);
+    Value cos = b.create<math::CosOp>(sqrtArg, fmf);
+    Value sin = b.create<math::SinOp>(sqrtArg, fmf);
+    // sin(atan2(0, inf)) = 0, sqrt(abs(inf)) = inf, but we can't multiply
+    // 0 * inf.
+    Value sinIsZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, sin, zero, fmf);
+
+    Value resultReal = b.create<arith::MulFOp>(absSqrt, cos, fmf);
     Value resultImag = b.create<arith::SelectOp>(
-        realIsNegative,
-        b.create<arith::SelectOp>(imagIsNegative, negativeResultReal,
-                                  resultReal),
-        imagDivTwoResultReal);
-
-    resultReal = b.create<arith::SelectOp>(
-        realIsNegative,
-        b.create<arith::DivFOp>(
-            imag, b.create<arith::AddFOp>(resultImag, resultImag, fmf), fmf),
-        resultReal);
-
-    Value realIsZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero);
-    Value imagIsZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero);
-    Value argIsZero = b.create<arith::AndIOp>(realIsZero, imagIsZero);
-
-    resultReal = b.create<arith::SelectOp>(argIsZero, zero, resultReal);
-    resultImag = b.create<arith::SelectOp>(argIsZero, zero, resultImag);
+        sinIsZero, zero, b.create<arith::MulFOp>(absSqrt, sin, fmf));
+    if (!arith::bitEnumContainsAll(fmf, arith::FastMathFlags::nnan |
+                                            arith::FastMathFlags::ninf)) {
+      Value inf = cst(APFloat::getInf(floatSemantics));
+      Value negInf = cst(APFloat::getInf(floatSemantics, true));
+      Value nan = cst(APFloat::getNaN(floatSemantics));
+      Value absImag = b.create<math::AbsFOp>(elementType, imag, fmf);
+
+      Value absImagIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absImag, inf, fmf);
+      Value absImagIsNotInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::ONE, absImag, inf, fmf);
+      Value realIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, inf, fmf);
+      Value realIsNegInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, negInf, fmf);
+
+      resultReal = b.create<arith::SelectOp>(
+          b.create<arith::AndIOp>(realIsNegInf, absImagIsNotInf), zero,
+          resultReal);
+      resultReal = b.create<arith::SelectOp>(
+          b.create<arith::OrIOp>(absImagIsInf, realIsInf), inf, resultReal);
+
+      Value imagSignInf = b.create<math::CopySignOp>(inf, imag, fmf);
+      resultImag = b.create<arith::SelectOp>(
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, absSqrt, absSqrt),
+          nan, resultImag);
+      resultImag = b.create<arith::SelectOp>(
+          b.create<arith::OrIOp>(absImagIsInf, realIsNegInf), imagSignInf,
+          resultImag);
+    }
+
+    Value resultIsZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absSqrt, zero, fmf);
+    resultReal = b.create<arith::SelectOp>(resultIsZero, zero, resultReal);
+    resultImag = b.create<arith::SelectOp>(resultIsZero, zero, resultImag);
 
     rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
                                                    resultImag);
@@ -1065,7 +1093,7 @@ static Value powOpConversionImpl(mlir::ImplicitLocOpBuilder &builder,
   // Case 2:
   // 1^(c + d*i) = 1 + 0*i
   Value lhsEqOne = builder.create<arith::AndIOp>(
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, one),
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, one, fmf),
       bEqZero);
   Value cutoff2 =
       builder.create<arith::SelectOp>(lhsEqOne, complexOne, cutoff1);
@@ -1073,11 +1101,11 @@ static Value powOpConversionImpl(mlir::ImplicitLocOpBuilder &builder,
   // Case 3:
   // inf^(c + 0*i) = inf + 0*i, c > 0
   Value lhsEqInf = builder.create<arith::AndIOp>(
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, inf),
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, a, inf, fmf),
       bEqZero);
   Value rhsGt0 = builder.create<arith::AndIOp>(
       dEqZero,
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, c, zero));
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OGT, c, zero, fmf));
   Value cutoff3 = builder.create<arith::SelectOp>(
       builder.create<arith::AndIOp>(lhsEqInf, rhsGt0), complexInf, cutoff2);
 
@@ -1085,7 +1113,7 @@ static Value powOpConversionImpl(mlir::ImplicitLocOpBuilder &builder,
   // inf^(c + 0*i) = 0 + 0*i, c < 0
   Value rhsLt0 = builder.create<arith::AndIOp>(
       dEqZero,
-      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OLT, c, zero));
+      builder.create<arith::CmpFOp>(arith::CmpFPredicate::OLT, c, zero, fmf));
   Value cutoff4 = builder.create<arith::SelectOp>(
       builder.create<arith::AndIOp>(lhsEqInf, rhsLt0), complexZero, cutoff3);
 
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 8d2fb09daa87b..b22c1acacaea1 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -8,9 +8,9 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   return %abs : f32
 }
 
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
@@ -250,9 +250,9 @@ func.func @complex_log(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg: complex<f32>
   return %log : complex<f32>
 }
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
@@ -493,9 +493,9 @@ func.func @complex_sign(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[REAL_IS_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
 // CHECK: %[[IMAG_IS_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
 // CHECK: %[[IS_ZERO:.*]] = arith.andi %[[REAL_IS_ZERO]], %[[IMAG_IS_ZERO]] : i1
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL2]] : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG2]] : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
@@ -697,45 +697,95 @@ func.func @complex_sqrt(%arg: complex<f32>) -> complex<f32> {
   return %sqrt : complex<f32>
 }
 
-// CHECK: %[[CST:.*]]  = arith.constant 0.000000e+00 : f32
-// CHECK: %[[VAR0:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[VAR1:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[VAR2:.*]] = math.absf %[[VAR0]] : f32
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[HALF:.*]] = arith.constant 5.000000e-01 : f32
+// CHECK: %[[RE:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IM:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
-// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] : f32
-// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
-// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] : f32
+// CHECK: %[[ABSRE:.*]] = math.absf %[[RE]] : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABSRE]], %[[ABSIM]] : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABSRE]], %[[ABSIM]] : f32
 // CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] : f32
 // CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] : f32
 // CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] : f32
-// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] : f32
-// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] : f32
-// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] : f32
-// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
-// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[ABS]] : f32
-// CHECK: %[[CST2:.*]]  = arith.constant 5.000000e-01 : f32
-// CHECK: %[[VAR24:.*]] = arith.mulf %[[VAR23]], %[[CST2]] : f32
-// CHECK: %[[VAR25:.*]] = math.sqrt %[[VAR24]] : f32
-// CHECK: %[[VAR26:.*]] = arith.cmpf olt, %[[VAR0]], %cst : f32
-// CHECK: %[[VAR27:.*]] = arith.cmpf olt, %[[VAR1]], %cst : f32
-// CHECK: %[[VAR28:.*]] = arith.addf %[[VAR25]], %[[VAR25]] : f32
-// CHECK: %[[VAR29:.*]] = arith.divf %[[VAR1]], %[[VAR28]] : f32
-// CHECK: %[[VAR30:.*]] = arith.negf %[[VAR25]] : f32
-// CHECK: %[[VAR31:.*]] = arith.select %[[VAR27]], %[[VAR30]], %[[VAR25]] : f32
-// CHECK: %[[VAR32:.*]] = arith.select %[[VAR26]], %[[VAR31]], %[[VAR29]] : f32
-// CHECK: %[[VAR33:.*]] = arith.addf %[[VAR32]], %[[VAR32]] : f32
-// CHECK: %[[VAR34:.*]] = arith.divf %[[VAR1]], %[[VAR33]] : f32
-// CHECK: %[[VAR35:.*]] = arith.select %[[VAR26]], %[[VAR34]], %[[VAR25]] : f32
-// CHECK: %[[VAR36:.*]] = arith.cmpf oeq, %[[VAR0]], %cst : f32
-// CHECK: %[[VAR37:.*]] = arith.cmpf oeq, %[[VAR1]], %cst : f32
-// CHECK: %[[VAR38:.*]] = arith.andi %[[VAR36]], %[[VAR37]] : i1
-// CHECK: %[[VAR39:.*]] = arith.select %[[VAR38]], %cst, %[[VAR35]] : f32
-// CHECK: %[[VAR40:.*]] = arith.select %[[VAR38]], %cst, %[[VAR32]] : f32
-// CHECK: %[[VAR41:.*]] = complex.create %[[VAR39]], %[[VAR40]] : complex<f32>
-// CHECK: return %[[VAR41]] : complex<f32>
+// CHECK: %[[QUARTER:.*]] = arith.constant 2.500000e-01 : f32
+// CHECK: %[[SQRT_MAX:.*]] = math.sqrt %[[MAX]] : f32
+// CHECK: %[[POW:.*]] = math.powf %[[RATIO_SQ_PLUS_ONE]], %[[QUARTER]] : f32
+// CHECK: %[[SQRT_ABS_OR_NAN:.*]] = arith.mulf %[[SQRT_MAX]], %[[POW]] : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS_OR_NAN]], %[[SQRT_ABS_OR_NAN]] : f32
+// CHECK: %[[SQRT_ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[SQRT_ABS_OR_NAN]] : f32
+// CHECK: %[[ARGARG:.*]] = math.atan2 %[[IM]], %[[RE]] : f32
+// CHECK: %[[SQRTARG:.*]] = arith.mulf %[[ARGARG]], %[[HALF]] : f32
+// CHECK: %[[COS:.*]] = math.cos %[[SQRTARG]] : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[SQRTARG]] : f32
+// CHECK: %[[SIN_ZERO:.*]] = arith.cmpf oeq, %[[SIN]], %[[ZERO]] : f32
+// CHECK: %[[RESULT_RE:.*]] = arith.mulf %[[SQRT_ABS]], %[[COS]] : f32
+// CHECK: %[[RESULT_IM:.*]] = arith.mulf %[[SQRT_ABS]], %[[SIN]] : f32
+// CHECK: %[[RESULT_IM2:.*]] = arith.select %[[SIN_ZERO]], %[[ZERO]], %[[RESULT_IM]] : f32
+// CHECK: %[[INF:.*]] = arith.constant 0x7F800000 : f32
+// CHECK: %[[NINF:.*]] = arith.constant 0xFF800000 : f32
+// CHECK: %[[NAN:.*]] = arith.constant 0x7FC00000 : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] : f32
+// CHECK: %[[ABSIMINF:.*]] = arith.cmpf oeq, %[[ABSIM]], %[[INF]] : f32
+// CHECK: %[[ABSIMNOTINF:.*]] = arith.cmpf one, %[[ABSIM]], %[[INF]] : f32
+// CHECK: %[[REINF:.*]] = arith.cmpf oeq, %[[RE]], %[[INF]] : f32
+// CHECK: %[[RENINF:.*]] = arith.cmpf oeq, %[[RE]], %[[NINF]] : f32
+// CHECK: %[[RESULT_RE_ZERO:.*]] = arith.andi %[[RENINF]], %[[ABSIMNOTINF]] : i1
+// CHECK: %[[RESULT_RE2:.*]] = arith.select %[[RESULT_RE_ZERO]], %[[ZERO]], %[[RESULT_RE]] : f32
+// CHECK: %[[RESUL_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[REINF]] : i1
+// CHECK: %[[RESULT_RE3:.*]] = arith.select %[[RESUL_IM_INF]], %[[INF]], %[[RESULT_RE2]] : f32
+// CHECK: %[[INF_IM_SIGN:.*]] = math.copysign %[[INF]], %[[IM]] : f32
+// CHECK: %[[RESULT_IM_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS]], %[[SQRT_ABS]] : f32
+// CHECK: %[[RESULT_IM3:.*]] = arith.select %[[RESULT_IM_NAN]], %[[NAN]], %[[RESULT_IM2]] : f32
+// CHECK: %[[RESULT_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[RENINF]] : i1
+// CHECK: %[[RESULT_IM4:.*]] = arith.select %[[RESULT_IM_INF]], %[[INF_IM_SIGN]], %[[RESULT_IM3]] : f32
+// CHECK: %[[RESULT_ZERO:.*]] = arith.cmpf oeq, %[[SQRT_ABS]], %[[ZERO]] : f32
+// CHECK: %[[RESULT_RE4:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_RE3]] : f32
+// CHECK: %[[RESULT_IM5:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_IM4]] : f32
+// CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_RE4]], %[[RESULT_IM5]] : complex<f32>
+// CHECK: return %[[RESULT]] : complex<f32>
+
+// -----
+
+// CHECK-LABEL: func @complex_sqrt_nnan_ninf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_sqrt_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
+  %sqrt = complex.sqrt %arg fastmath<nnan,ninf> : complex<f32>
+  return %sqrt : complex<f32>
+}
+
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[HALF:.*]] = arith.constant 5.000000e-01 : f32
+// CHECK: %[[RE:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IM:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[ABSRE:.*]] = math.absf %[[RE]] fastmath<nnan,ninf> : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] fastmath<nnan,ninf> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,ninf> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,ninf> : f32
+// CHECK: %[[QUARTER:.*]] = arith.constant 2.500000e-01 : f32
+// CHECK: %[[SQRT_MAX:.*]] = math.sqrt %[[MAX]] fastmath<nnan,ninf> : f32
+// CHECK: %[[POW:.*]] = math.powf %[[RATIO_SQ_PLUS_ONE]], %[[QUARTER]] fastmath<nnan,ninf> : f32
+// CHECK: %[[SQRT_ABS_OR_NAN:.*]] = arith.mulf %[[SQRT_MAX]], %[[POW]] fastmath<nnan,ninf> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS_OR_NAN]], %[[SQRT_ABS_OR_NAN]] fastmath<nnan,ninf> : f32
+// CHECK: %[[SQRT_ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[SQRT_ABS_OR_NAN]] : f32
+// CHECK: %[[ARGARG:.*]] = math.atan2 %[[IM]], %[[RE]] fastmath<nnan,ninf> : f32
+// CHECK: %[[SQRTARG:.*]] = arith.mulf %[[ARGARG]], %[[HALF]] fastmath<nnan,ninf> : f32
+// CHECK: %[[COS:.*]] = math.cos %[[SQRTARG]] fastmath<nnan,ninf> : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[SQRTARG]] fastmath<nnan,ninf> : f32
+// CHECK: %[[SIN_ZERO:.*]] = arith.cmpf oeq, %[[SIN]], %[[ZERO]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RESULT_RE:.*]] = arith.mulf %[[SQRT_ABS]], %[[COS]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RESULT_IM:.*]] = arith.mulf %[[SQRT_ABS]], %[[SIN]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RESULT_IM2:.*]] = arith.select %[[SIN_ZERO]], %[[ZERO]], %[[RESULT_IM]] : f32
+// CHECK: %[[RESULT_ZERO:.*]] = arith.cmpf oeq, %[[SQRT_ABS]], %[[ZERO]] fastmath<nnan,ninf> : f32
+// CHECK: %[[RESULT_RE2:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_RE]] : f32
+// CHECK: %[[RESULT_IM3:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_IM2]] : f32
+// CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_RE2]], %[[RESULT_IM3]] : complex<f32>
+// CHECK: return %[[RESULT]] : complex<f32>
 
 // -----
 
@@ -808,9 +858,9 @@ func.func @complex_abs_with_fmf(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg fastmath<nnan,contract> : complex<f32>
   return %abs : f32
 }
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
@@ -907,9 +957,9 @@ func.func @complex_log_with_fmf(%arg: complex<f32>) -> complex<f32> {
   %log = complex.log %arg fastmath<nnan,contract> : complex<f32>
   return %log : complex<f32>
 }
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
@@ -1285,44 +1335,53 @@ func.func @complex_atan2_with_fmf(%lhs: complex<f32>,
 // CHECK: %[[VAR184:.*]] = complex.im %[[VAR179]] : complex<f32>
 // CHECK: %[[VAR185:.*]] = arith.addf %[[VAR183]], %[[VAR184]] fastmath<nnan,contract> : f32
 // CHECK: %[[VAR186:.*]] = complex.create %[[VAR182]], %[[VAR185]] : complex<f32>
-// CHECK: %[[CST_6:.*]] = arith.constant 0.000000e+00 : f32
-// CHECK: %[[VAR187:.*]] = complex.re %[[VAR186]] : complex<f32>
-// CHECK: %[[VAR188:.*]] = complex.im %[[VAR186]] : complex<f32>
-// CHECK: %[[VAR189:.*]] = math.absf %[[VAR187]] fastmath<nnan,contract> : f32
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[HALF:.*]] = arith.constant 5.000000e-01 : f32
+// CHECK: %[[RE:.*]] = complex.re %[[VAR186]] : complex<f32>
+// CHECK: %[[IM:.*]] = complex.im %[[VAR186]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[REAL:.*]] = complex.re %[[VAR186]] : complex<f32>
-// CHECK: %[[IMAG:.*]] = complex.im %[[VAR186]] : complex<f32>
-// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract>  : f32
-// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract>  : f32
-// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract>  : f32
-// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract>  : f32
-// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract>  : f32
-// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract>  : f32
-// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
-// CHECK: %[[VAR210:.*]] = arith.addf %[[VAR189]], %[[ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[CST_9:.*]] = arith.constant 5.000000e-01 : f32
-// CHECK: %[[VAR211:.*]] = arith.mulf %[[VAR210]], %[[CST_9]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR212:.*]] = math.sqrt %[[VAR211]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR213:.*]] = arith.cmpf olt, %[[VAR187]], %[[CST_6]] : f32
-// CHECK: %[[VAR214:.*]] = arith.cmpf olt, %[[VAR188]], %[[CST_6]] : f32
-// CHECK: %[[VAR215:.*]] = arith.addf %[[VAR212]], %[[VAR212]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR216:.*]] = arith.divf %[[VAR188]], %[[VAR215]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR217:.*]] = arith.negf %[[VAR212]] : f32
-// CHECK: %[[VAR218:.*]] = arith.select %[[VAR214]], %[[VAR217]], %[[VAR212]] : f32
-// CHECK: %[[VAR219:.*]] = arith.select %[[VAR213]], %[[VAR218]], %[[VAR216]] : f32
-// CHECK: %[[VAR220:.*]] = arith.addf %[[VAR219]], %[[VAR219]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR221:.*]] = arith.divf %[[VAR188]], %[[VAR220]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR222:.*]] = arith.select %[[VAR213]], %[[VAR221]], %[[VAR212]] : f32
-// CHECK: %[[VAR223:.*]] = arith.cmpf oeq, %[[VAR187]], %[[CST_6]] : f32
-// CHECK: %[[VAR224:.*]] = arith.cmpf oeq, %[[VAR188]], %[[CST_6]] : f32
-// CHECK: %[[VAR225:.*]] = arith.andi %[[VAR223]], %[[VAR224]] : i1
-// CHECK: %[[VAR226:.*]] = arith.select %[[VAR225]], %[[CST_6]], %[[VAR222]] : f32
-// CHECK: %[[VAR227:.*]] = arith.select %[[VAR225]], %[[CST_6]], %[[VAR219]] : f32
-// CHECK: %[[VAR228:.*]] = complex.create %[[VAR226]], %[[VAR227]] : complex<f32>
+// CHECK: %[[ABSRE:.*]] = math.absf %[[RE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
+// CHECK: %[[QUARTER:.*]] = arith.constant 2.500000e-01 : f32
+// CHECK: %[[SQRT_MAX:.*]] = math.sqrt %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[POW:.*]] = math.powf %[[RATIO_SQ_PLUS_ONE]], %[[QUARTER]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT_ABS_OR_NAN:.*]] = arith.mulf %[[SQRT_MAX]], %[[POW]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS_OR_NAN]], %[[SQRT_ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT_ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[SQRT_ABS_OR_NAN]] : f32
+// CHECK: %[[ARGARG:.*]] = math.atan2 %[[IM]], %[[RE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRTARG:.*]] = arith.mulf %[[ARGARG]], %[[HALF]] fastmath<nnan,contract> : f32
+// CHECK: %[[COS:.*]] = math.cos %[[SQRTARG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[SQRTARG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN_ZERO:.*]] = arith.cmpf oeq, %[[SIN]], %[[ZERO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE:.*]] = arith.mulf %[[SQRT_ABS]], %[[COS]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM:.*]] = arith.mulf %[[SQRT_ABS]], %[[SIN]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM2:.*]] = arith.select %[[SIN_ZERO]], %[[ZERO]], %[[RESULT_IM]] : f32
+// CHECK: %[[INF:.*]] = arith.constant 0x7F800000 : f32
+// CHECK: %[[NINF:.*]] = arith.constant 0xFF800000 : f32
+// CHECK: %[[NAN:.*]] = arith.constant 0x7FC00000 : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIMINF:.*]] = arith.cmpf oeq, %[[ABSIM]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIMNOTINF:.*]] = arith.cmpf one, %[[ABSIM]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[REINF:.*]] = arith.cmpf oeq, %[[RE]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[RENINF:.*]] = arith.cmpf oeq, %[[RE]], %[[NINF]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE_ZERO:.*]] = arith.andi %[[RENINF]], %[[ABSIMNOTINF]] : i1
+// CHECK: %[[RESULT_RE2:.*]] = arith.select %[[RESULT_RE_ZERO]], %[[ZERO]], %[[RESULT_RE]] : f32
+// CHECK: %[[RESUL_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[REINF]] : i1
+// CHECK: %[[RESULT_RE3:.*]] = arith.select %[[RESUL_IM_INF]], %[[INF]], %[[RESULT_RE2]] : f32
+// CHECK: %[[INF_IM_SIGN:.*]] = math.copysign %[[INF]], %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS]], %[[SQRT_ABS]] : f32
+// CHECK: %[[RESULT_IM3:.*]] = arith.select %[[RESULT_IM_NAN]], %[[NAN]], %[[RESULT_IM2]] : f32
+// CHECK: %[[RESULT_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[RENINF]] : i1
+// CHECK: %[[RESULT_IM4:.*]] = arith.select %[[RESULT_IM_INF]], %[[INF_IM_SIGN]], %[[RESULT_IM3]] : f32
+// CHECK: %[[RESULT_ZERO:.*]] = arith.cmpf oeq, %[[SQRT_ABS]], %[[ZERO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE4:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_RE3]] : f32
+// CHECK: %[[RESULT_IM5:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_IM4]] : f32
+// CHECK: %[[VAR228:.*]] = complex.create %[[RESULT_RE4]], %[[RESULT_IM5]] : complex<f32>
 // CHECK: %[[CST_10:.*]] = arith.constant 0.000000e+00 : f32
 // CHECK: %[[CST_11:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[VAR229:.*]] = complex.create %[[CST_10]], %[[CST_11]] : complex<f32>
@@ -1519,9 +1578,9 @@ func.func @complex_atan2_with_fmf(%lhs: complex<f32>,
 // CHECK: %[[VAR413:.*]] = arith.select %[[VAR412]], %[[VAR408]], %[[VAR402]] : f32
 // CHECK: %[[VAR414:.*]] = arith.select %[[VAR412]], %[[VAR409]], %[[VAR403]] : f32
 // CHECK: %[[VAR415:.*]] = complex.create %[[VAR413]], %[[VAR414]] : complex<f32>
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL:.*]] = complex.re %[[VAR415]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[VAR415]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
@@ -1756,45 +1815,54 @@ func.func @complex_sqrt_with_fmf(%arg: complex<f32>) -> complex<f32> {
   return %sqrt : complex<f32>
 }
 
-// CHECK: %[[CST:.*]]  = arith.constant 0.000000e+00 : f32
-// CHECK: %[[VAR0:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[VAR1:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[VAR2:.*]] = math.absf %[[VAR0]] fastmath<nnan,contract> : f32
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[HALF:.*]] = arith.constant 5.000000e-01 : f32
+// CHECK: %[[RE:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IM:.*]] = complex.im %[[ARG]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
-// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSRE:.*]] = math.absf %[[RE]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[MAX:.*]] = arith.maximumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,contract> : f32
+// CHECK: %[[MIN:.*]] = arith.minimumf %[[ABSRE]], %[[ABSIM]] fastmath<nnan,contract> : f32
 // CHECK: %[[RATIO:.*]] = arith.divf %[[MIN]], %[[MAX]] fastmath<nnan,contract> : f32
 // CHECK: %[[RATIO_SQ:.*]] = arith.mulf %[[RATIO]], %[[RATIO]] fastmath<nnan,contract> : f32
 // CHECK: %[[RATIO_SQ_PLUS_ONE:.*]] = arith.addf %[[RATIO_SQ]], %[[ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[SQRT:.*]] = math.sqrt %[[RATIO_SQ_PLUS_ONE]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS_OR_NAN:.*]] = arith.mulf %[[MAX]], %[[SQRT]] fastmath<nnan,contract> : f32
-// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[ABS_OR_NAN]], %[[ABS_OR_NAN]] fastmath<nnan,contract> : f32
-// CHECK: %[[ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[ABS_OR_NAN]] : f32
-// CHECK: %[[VAR23:.*]] = arith.addf %[[VAR2]], %[[ABS]] fastmath<nnan,contract> : f32
-// CHECK: %[[CST2:.*]]  = arith.constant 5.000000e-01 : f32
-// CHECK: %[[VAR24:.*]] = arith.mulf %[[VAR23]], %[[CST2]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR25:.*]] = math.sqrt %[[VAR24]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR26:.*]] = arith.cmpf olt, %[[VAR0]], %cst : f32
-// CHECK: %[[VAR27:.*]] = arith.cmpf olt, %[[VAR1]], %cst : f32
-// CHECK: %[[VAR28:.*]] = arith.addf %[[VAR25]], %[[VAR25]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR29:.*]] = arith.divf %[[VAR1]], %[[VAR28]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR30:.*]] = arith.negf %[[VAR25]] : f32
-// CHECK: %[[VAR31:.*]] = arith.select %[[VAR27]], %[[VAR30]], %[[VAR25]] : f32
-// CHECK: %[[VAR32:.*]] = arith.select %[[VAR26]], %[[VAR31]], %[[VAR29]] : f32
-// CHECK: %[[VAR33:.*]] = arith.addf %[[VAR32]], %[[VAR32]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR34:.*]] = arith.divf %[[VAR1]], %[[VAR33]] fastmath<nnan,contract> : f32
-// CHECK: %[[VAR35:.*]] = arith.select %[[VAR26]], %[[VAR34]], %[[VAR25]] : f32
-// CHECK: %[[VAR36:.*]] = arith.cmpf oeq, %[[VAR0]], %cst : f32
-// CHECK: %[[VAR37:.*]] = arith.cmpf oeq, %[[VAR1]], %cst : f32
-// CHECK: %[[VAR38:.*]] = arith.andi %[[VAR36]], %[[VAR37]] : i1
-// CHECK: %[[VAR39:.*]] = arith.select %[[VAR38]], %cst, %[[VAR35]] : f32
-// CHECK: %[[VAR40:.*]] = arith.select %[[VAR38]], %cst, %[[VAR32]] : f32
-// CHECK: %[[VAR41:.*]] = complex.create %[[VAR39]], %[[VAR40]] : complex<f32>
-// CHECK: return %[[VAR41]] : complex<f32>
+// CHECK: %[[QUARTER:.*]] = arith.constant 2.500000e-01 : f32
+// CHECK: %[[SQRT_MAX:.*]] = math.sqrt %[[MAX]] fastmath<nnan,contract> : f32
+// CHECK: %[[POW:.*]] = math.powf %[[RATIO_SQ_PLUS_ONE]], %[[QUARTER]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT_ABS_OR_NAN:.*]] = arith.mulf %[[SQRT_MAX]], %[[POW]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS_OR_NAN]], %[[SQRT_ABS_OR_NAN]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRT_ABS:.*]] = arith.select %[[IS_NAN]], %[[MIN]], %[[SQRT_ABS_OR_NAN]] : f32
+// CHECK: %[[ARGARG:.*]] = math.atan2 %[[IM]], %[[RE]] fastmath<nnan,contract> : f32
+// CHECK: %[[SQRTARG:.*]] = arith.mulf %[[ARGARG]], %[[HALF]] fastmath<nnan,contract> : f32
+// CHECK: %[[COS:.*]] = math.cos %[[SQRTARG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[SQRTARG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN_ZERO:.*]] = arith.cmpf oeq, %[[SIN]], %[[ZERO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE:.*]] = arith.mulf %[[SQRT_ABS]], %[[COS]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM:.*]] = arith.mulf %[[SQRT_ABS]], %[[SIN]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM2:.*]] = arith.select %[[SIN_ZERO]], %[[ZERO]], %[[RESULT_IM]] : f32
+// CHECK: %[[INF:.*]] = arith.constant 0x7F800000 : f32
+// CHECK: %[[NINF:.*]] = arith.constant 0xFF800000 : f32
+// CHECK: %[[NAN:.*]] = arith.constant 0x7FC00000 : f32
+// CHECK: %[[ABSIM:.*]] = math.absf %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIMINF:.*]] = arith.cmpf oeq, %[[ABSIM]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABSIMNOTINF:.*]] = arith.cmpf one, %[[ABSIM]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[REINF:.*]] = arith.cmpf oeq, %[[RE]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[RENINF:.*]] = arith.cmpf oeq, %[[RE]], %[[NINF]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE_ZERO:.*]] = arith.andi %[[RENINF]], %[[ABSIMNOTINF]] : i1
+// CHECK: %[[RESULT_RE2:.*]] = arith.select %[[RESULT_RE_ZERO]], %[[ZERO]], %[[RESULT_RE]] : f32
+// CHECK: %[[RESUL_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[REINF]] : i1
+// CHECK: %[[RESULT_RE3:.*]] = arith.select %[[RESUL_IM_INF]], %[[INF]], %[[RESULT_RE2]] : f32
+// CHECK: %[[INF_IM_SIGN:.*]] = math.copysign %[[INF]], %[[IM]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_IM_NAN:.*]] = arith.cmpf uno, %[[SQRT_ABS]], %[[SQRT_ABS]] : f32
+// CHECK: %[[RESULT_IM3:.*]] = arith.select %[[RESULT_IM_NAN]], %[[NAN]], %[[RESULT_IM2]] : f32
+// CHECK: %[[RESULT_IM_INF:.*]] = arith.ori %[[ABSIMINF]], %[[RENINF]] : i1
+// CHECK: %[[RESULT_IM4:.*]] = arith.select %[[RESULT_IM_INF]], %[[INF_IM_SIGN]], %[[RESULT_IM3]] : f32
+// CHECK: %[[RESULT_ZERO:.*]] = arith.cmpf oeq, %[[SQRT_ABS]], %[[ZERO]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_RE4:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_RE3]] : f32
+// CHECK: %[[RESULT_IM5:.*]] = arith.select %[[RESULT_ZERO]], %[[ZERO]], %[[RESULT_IM4]] : f32
+// CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_RE4]], %[[RESULT_IM5]] : complex<f32>
+// CHECK: return %[[RESULT]] : complex<f32>
 
 // -----
 
@@ -1857,9 +1925,9 @@ func.func @complex_sign_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[REAL_IS_ZERO:.*]] = arith.cmpf oeq, %[[REAL]], %[[ZERO]] : f32
 // CHECK: %[[IMAG_IS_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
 // CHECK: %[[IS_ZERO:.*]] = arith.andi %[[REAL_IS_ZERO]], %[[IMAG_IS_ZERO]] : i1
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[REAL2:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG2:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL2]] fastmath<nnan,contract> : f32
 // CHECK: %[[ABS_IMAG:.*]] = math.absf %[[IMAG2]] fastmath<nnan,contract> : f32
 // CHECK: %[[MAX:.*]] = arith.maximumf %[[ABS_REAL]], %[[ABS_IMAG]] fastmath<nnan,contract> : f32
diff --git a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
index 2649d004a76ac..110a78631fb95 100644
--- a/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/full-conversion.mlir
@@ -6,10 +6,10 @@ func.func @complex_abs(%arg: complex<f32>) -> f32 {
   %abs = complex.abs %arg: complex<f32>
   return %abs : f32
 }
-// CHECK: %[[ONE:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
 // CHECK: %[[REAL:.*]] = llvm.extractvalue %[[ARG]][0] : ![[C_TY]]
 // CHECK: %[[IMAG:.*]] = llvm.extractvalue %[[ARG]][1] : ![[C_TY]]
 
+// CHECK: %[[ONE:.*]] = llvm.mlir.constant(1.000000e+00 : f32) : f32
 // CHECK: %[[ABS_REAL:.*]] = llvm.intr.fabs(%[[REAL]]) : (f32) -> f32
 // CHECK: %[[ABS_IMAG:.*]] = llvm.intr.fabs(%[[IMAG]]) : (f32) -> f32
 // CHECK: %[[MAX:.*]] = llvm.intr.maximum(%[[ABS_REAL]], %[[ABS_IMAG]]) : (f32, f32) -> f32

From 1e7763557bbace002b6bb811167942ebebf50f99 Mon Sep 17 00:00:00 2001
From: wanglei <wanglei@loongson.cn>
Date: Fri, 12 Apr 2024 16:15:02 +0800
Subject: [PATCH 223/886] [LoongArch] Add support for getNumberOfRegisters()
 (#88372)

The `TTI` hooks are used during vectorization for calculating register
pressure. The default implementation defined wrong value for register
number (all register class are 8 registers).

This patch also defines LoongArch's own register classes.
---
 .../LoongArchTargetTransformInfo.cpp          | 41 +++++++++++++++++
 .../LoongArch/LoongArchTargetTransformInfo.h  |  4 ++
 .../LoopVectorize/LoongArch/reg-usage.ll      | 44 +++++++++++++++++++
 3 files changed, 89 insertions(+)
 create mode 100644 llvm/test/Transforms/LoopVectorize/LoongArch/reg-usage.ll

diff --git a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
index d47dded9ea6ec..add1c60d89d21 100644
--- a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.cpp
@@ -40,4 +40,45 @@ TypeSize LoongArchTTIImpl::getRegisterBitWidth(
   llvm_unreachable("Unsupported register kind");
 }
 
+unsigned LoongArchTTIImpl::getNumberOfRegisters(unsigned ClassID) const {
+  switch (ClassID) {
+  case LoongArchRegisterClass::GPRRC:
+    // 30 = 32 GPRs - r0 (zero register) - r21 (non-allocatable)
+    return 30;
+  case LoongArchRegisterClass::FPRRC:
+    return ST->hasBasicF() ? 32 : 0;
+  case LoongArchRegisterClass::VRRC:
+    return ST->hasExtLSX() ? 32 : 0;
+  }
+  llvm_unreachable("unknown register class");
+}
+
+unsigned LoongArchTTIImpl::getRegisterClassForType(bool Vector,
+                                                   Type *Ty) const {
+  if (Vector)
+    return LoongArchRegisterClass::VRRC;
+  if (!Ty)
+    return LoongArchRegisterClass::GPRRC;
+
+  Type *ScalarTy = Ty->getScalarType();
+  if ((ScalarTy->isFloatTy() && ST->hasBasicF()) ||
+      (ScalarTy->isDoubleTy() && ST->hasBasicD())) {
+    return LoongArchRegisterClass::FPRRC;
+  }
+
+  return LoongArchRegisterClass::GPRRC;
+}
+
+const char *LoongArchTTIImpl::getRegisterClassName(unsigned ClassID) const {
+  switch (ClassID) {
+  case LoongArchRegisterClass::GPRRC:
+    return "LoongArch::GPRRC";
+  case LoongArchRegisterClass::FPRRC:
+    return "LoongArch::FPRRC";
+  case LoongArchRegisterClass::VRRC:
+    return "LoongArch::VRRC";
+  }
+  llvm_unreachable("unknown register class");
+}
+
 // TODO: Implement more hooks to provide TTI machinery for LoongArch.
diff --git a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
index d296c9ed576fb..34c18163bbdb6 100644
--- a/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
+++ b/llvm/lib/Target/LoongArch/LoongArchTargetTransformInfo.h
@@ -28,6 +28,7 @@ class LoongArchTTIImpl : public BasicTTIImplBase<LoongArchTTIImpl> {
   typedef TargetTransformInfo TTI;
   friend BaseT;
 
+  enum LoongArchRegisterClass { GPRRC, FPRRC, VRRC };
   const LoongArchSubtarget *ST;
   const LoongArchTargetLowering *TLI;
 
@@ -40,6 +41,9 @@ class LoongArchTTIImpl : public BasicTTIImplBase<LoongArchTTIImpl> {
         TLI(ST->getTargetLowering()) {}
 
   TypeSize getRegisterBitWidth(TargetTransformInfo::RegisterKind K) const;
+  unsigned getNumberOfRegisters(unsigned ClassID) const;
+  unsigned getRegisterClassForType(bool Vector, Type *Ty = nullptr) const;
+  const char *getRegisterClassName(unsigned ClassID) const;
 
   // TODO: Implement more hooks to provide TTI machinery for LoongArch.
 };
diff --git a/llvm/test/Transforms/LoopVectorize/LoongArch/reg-usage.ll b/llvm/test/Transforms/LoopVectorize/LoongArch/reg-usage.ll
new file mode 100644
index 0000000000000..f45a2f0f5b7e8
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/LoongArch/reg-usage.ll
@@ -0,0 +1,44 @@
+; REQUIRES: asserts
+; RUN: opt --passes=loop-vectorize --mtriple loongarch64-linux-gnu \
+; RUN:   --mattr=+lsx -debug-only=loop-vectorize --force-vector-width=1 \
+; RUN:   -S < %s 2>&1 | FileCheck %s --check-prefix=CHECK-SCALAR
+; RUN: opt --passes=loop-vectorize --mtriple loongarch64-linux-gnu \
+; RUN:   --mattr=+lsx -debug-only=loop-vectorize --force-vector-width=4 \
+; RUN:   -S < %s 2>&1 | FileCheck %s --check-prefix=CHECK-VECTOR
+
+define void @bar(ptr %A, i32 signext %n) {
+; CHECK-LABEL: bar
+; CHECK-SCALAR:      LV(REG): Found max usage: 2 item
+; CHECK-SCALAR-NEXT: LV(REG): RegisterClass: LoongArch::GPRRC, 2 registers
+; CHECK-SCALAR-NEXT: LV(REG): RegisterClass: LoongArch::FPRRC, 1 registers
+; CHECK-SCALAR-NEXT: LV(REG): Found invariant usage: 1 item
+; CHECK-SCALAR-NEXT: LV(REG): RegisterClass: LoongArch::GPRRC, 1 registers
+; CHECK-SCALAR-NEXT: LV: The target has 30 registers of LoongArch::GPRRC register class
+; CHECK-SCALAR-NEXT: LV: The target has 32 registers of LoongArch::FPRRC register class
+; CHECK-VECTOR:      LV(REG): Found max usage: 1 item
+; CHECK-VECTOR-NEXT: LV(REG): RegisterClass: LoongArch::VRRC, 3 registers
+; CHECK-VECTOR-NEXT: LV(REG): Found invariant usage: 1 item
+; CHECK-VECTOR-NEXT: LV(REG): RegisterClass: LoongArch::GPRRC, 1 registers
+; CHECK-VECTOR-NEXT: LV: The target has 32 registers of LoongArch::VRRC register class
+
+entry:
+  %cmp4 = icmp sgt i32 %n, 0
+  br i1 %cmp4, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:                               ; preds = %entry
+  %wide.trip.count = zext nneg i32 %n to i64
+  br label %for.body
+
+for.cond.cleanup:                                 ; preds = %for.body, %entry
+  ret void
+
+for.body:                                         ; preds = %for.body.preheader, %for.body
+  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
+  %0 = trunc i64 %indvars.iv to i32
+  %conv = sitofp i32 %0 to float
+  %arrayidx = getelementptr inbounds float, ptr %A, i64 %indvars.iv
+  store float %conv, ptr %arrayidx, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}

From a5ed14bc8e122fa5ac0aa81f8d8390931bd6b4e4 Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang@intel.com>
Date: Fri, 12 Apr 2024 16:16:56 +0800
Subject: [PATCH 224/886] [X86][BF16] Fix a typo, NFCI

VR128X is not enabled under AVXNECONVERT.
---
 llvm/lib/Target/X86/X86InstrSSE.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 2b391b60f2c9b..5d799fc00df92 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -8314,7 +8314,7 @@ let Predicates = [HasAVXNECONVERT] in {
        f256mem>, T8;
   defm VCVTNEPS2BF16 : VCVTNEPS2BF16_BASE, VEX, T8, XS, ExplicitVEXPrefix;
 
-  def : Pat<(v8bf16 (X86cvtneps2bf16 (v4f32 VR128X:$src))),
+  def : Pat<(v8bf16 (X86cvtneps2bf16 (v4f32 VR128:$src))),
             (VCVTNEPS2BF16rr VR128:$src)>;
   def : Pat<(v8bf16 (X86cvtneps2bf16 (loadv4f32 addr:$src))),
             (VCVTNEPS2BF16rm addr:$src)>;

From ad4e1aba3fd12f81de71ce3985ae66ff80773d90 Mon Sep 17 00:00:00 2001
From: jeanPerier <jperier@nvidia.com>
Date: Fri, 12 Apr 2024 10:29:01 +0200
Subject: [PATCH 225/886] [flang] Pass VALUE CHARACTER arg by register in
 BIND(C) calls (#87774)

Fortran mandates "CHARACTER(1), VALUE" be passed as a C "char" in calls
to BIND(C) procedures (F'2023 18.3.7 (4)). Lowering passed them by
memory instead. Update call interface lowering code to pass them by
register. Fix related test and update it to use HLFIR.
---
 flang/lib/Lower/CallInterface.cpp   |  24 ++++---
 flang/lib/Lower/ConvertCall.cpp     |  16 +++--
 flang/lib/Lower/ConvertVariable.cpp |   9 ++-
 flang/test/Lower/call-by-value.f90  | 106 +++++++++++++++++-----------
 4 files changed, 101 insertions(+), 54 deletions(-)

diff --git a/flang/lib/Lower/CallInterface.cpp b/flang/lib/Lower/CallInterface.cpp
index 29cdb3cff589b..05a0c10c70974 100644
--- a/flang/lib/Lower/CallInterface.cpp
+++ b/flang/lib/Lower/CallInterface.cpp
@@ -1136,14 +1136,22 @@ class Fortran::lower::CallInterfaceImpl {
       addPassedArg(PassEntityBy::Box, entity, characteristics);
     } else if (dynamicType.category() ==
                Fortran::common::TypeCategory::Character) {
-      // Pass as fir.box_char
-      mlir::Type boxCharTy =
-          fir::BoxCharType::get(&mlirContext, dynamicType.kind());
-      addFirOperand(boxCharTy, nextPassedArgPosition(), Property::BoxChar,
-                    attrs);
-      addPassedArg(isValueAttr ? PassEntityBy::CharBoxValueAttribute
-                               : PassEntityBy::BoxChar,
-                   entity, characteristics);
+      if (isValueAttr && isBindC) {
+        // Pass as fir.char<1>
+        mlir::Type charTy =
+            fir::CharacterType::getSingleton(&mlirContext, dynamicType.kind());
+        addFirOperand(charTy, nextPassedArgPosition(), Property::Value, attrs);
+        addPassedArg(PassEntityBy::Value, entity, characteristics);
+      } else {
+        // Pass as fir.box_char
+        mlir::Type boxCharTy =
+            fir::BoxCharType::get(&mlirContext, dynamicType.kind());
+        addFirOperand(boxCharTy, nextPassedArgPosition(), Property::BoxChar,
+                      attrs);
+        addPassedArg(isValueAttr ? PassEntityBy::CharBoxValueAttribute
+                                 : PassEntityBy::BoxChar,
+                     entity, characteristics);
+      }
     } else {
       // Pass as fir.ref unless it's by VALUE and BIND(C). Also pass-by-value
       // for numerical/logical scalar without OPTIONAL so that the behavior is
diff --git a/flang/lib/Lower/ConvertCall.cpp b/flang/lib/Lower/ConvertCall.cpp
index 315a3f6736aa1..c6f7d3410ad5c 100644
--- a/flang/lib/Lower/ConvertCall.cpp
+++ b/flang/lib/Lower/ConvertCall.cpp
@@ -1494,11 +1494,19 @@ genUserCall(Fortran::lower::PreparedActualArguments &loweredActuals,
           value =
               hlfir::Entity{genRecordCPtrValueArg(builder, loc, value, eleTy)};
         }
-      } else if (fir::isa_derived(value.getFortranElementType())) {
-        // BIND(C), VALUE derived type. The derived type value must really
+      } else if (fir::isa_derived(value.getFortranElementType()) ||
+                 value.isCharacter()) {
+        // BIND(C), VALUE derived type or character. The value must really
         // be loaded here.
-        auto [derived, cleanup] = hlfir::convertToValue(loc, builder, value);
-        mlir::Value loadedValue = fir::getBase(derived);
+        auto [exv, cleanup] = hlfir::convertToValue(loc, builder, value);
+        mlir::Value loadedValue = fir::getBase(exv);
+        // Character actual arguments may have unknown length or a length longer
+        // than one. Cast the memory ref to the dummy type so that the load is
+        // valid and only loads what is needed.
+        if (mlir::Type baseTy = fir::dyn_cast_ptrEleTy(loadedValue.getType()))
+          if (fir::isa_char(baseTy))
+            loadedValue = builder.createConvert(
+                loc, fir::ReferenceType::get(argTy), loadedValue);
         if (fir::isa_ref_type(loadedValue.getType()))
           loadedValue = builder.create<fir::LoadOp>(loc, loadedValue);
         caller.placeInput(arg, loadedValue);
diff --git a/flang/lib/Lower/ConvertVariable.cpp b/flang/lib/Lower/ConvertVariable.cpp
index 01473f7f6cd72..2d2d9eba905bd 100644
--- a/flang/lib/Lower/ConvertVariable.cpp
+++ b/flang/lib/Lower/ConvertVariable.cpp
@@ -2100,10 +2100,15 @@ void Fortran::lower::mapSymbolAttributes(
   if (ba.isChar()) {
     if (arg) {
       assert(!preAlloc && "dummy cannot be pre-allocated");
-      if (arg.getType().isa<fir::BoxCharType>())
+      if (mlir::isa<fir::BoxCharType>(arg.getType())) {
         std::tie(addr, len) = charHelp.createUnboxChar(arg);
-      else if (!addr)
+      } else if (mlir::isa<fir::CharacterType>(arg.getType())) {
+        // fir.char<1> passed by value (BIND(C) with VALUE attribute).
+        addr = builder.create<fir::AllocaOp>(loc, arg.getType());
+        builder.create<fir::StoreOp>(loc, arg, addr);
+      } else if (!addr) {
         addr = arg;
+      }
       // Ensure proper type is given to array/scalar that was transmitted as a
       // fir.boxchar arg or is a statement function actual argument with
       // a different length than the dummy.
diff --git a/flang/test/Lower/call-by-value.f90 b/flang/test/Lower/call-by-value.f90
index e489ea432305f..d4694f4575d08 100644
--- a/flang/test/Lower/call-by-value.f90
+++ b/flang/test/Lower/call-by-value.f90
@@ -1,9 +1,10 @@
 ! Test for PassBy::Value
-! RUN: bbc -emit-fir -hlfir=false %s -o - | FileCheck %s
+! RUN: bbc -emit-fir %s -o - | FileCheck %s
 
 !CHECK-LABEL: func @_QQmain()
-!CHECK: %[[LOGICAL:.*]] = fir.alloca !fir.logical<1>
 !CHECK: %false = arith.constant false
+!CHECK: %[[LOGICAL_ALLOC:.*]] = fir.alloca !fir.logical<1>
+!CHECK: %[[LOGICAL:.*]] = fir.declare %[[LOGICAL_ALLOC]]
 !CHECK: %[[VALUE:.*]] = fir.convert %false : (i1) -> !fir.logical<1>
 !CHECK: fir.store %[[VALUE]] to %[[LOGICAL]]
 !CHECK: %[[LOAD:.*]] = fir.load %[[LOGICAL]]
@@ -23,39 +24,42 @@ end subroutine omp_set_nested
   call omp_set_nested(do_nested)
 end program call_by_value
 
-! CHECK-LABEL: func.func @test_integer_value(
-! CHECK-SAME:                                %[[VAL_0:.*]]: i32 {fir.bindc_name = "x"}) attributes {fir.bindc_name = "test_integer_value"} {
-! CHECK:         %[[VAL_1:.*]] = fir.alloca i32
-! CHECK:         fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<i32>
-! CHECK:         fir.call @_QPinternal_call(%[[VAL_1]]) {{.*}}: (!fir.ref<i32>) -> ()
-! CHECK:         return
-! CHECK:       }
+! CHECK-LABEL:   func.func @test_integer_value(
+! CHECK-SAME:                                  %[[VAL_0:.*]]: i32
+! CHECK:           %[[VAL_1:.*]] = fir.alloca i32
+! CHECK:           fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<i32>
+! CHECK:           %[[VAL_2:.*]] = fir.declare %[[VAL_1]]
+! CHECK:           fir.call @_QPinternal_call(%[[VAL_2]]) {{.*}}: (!fir.ref<i32>) -> ()
+! CHECK:           return
+! CHECK:         }
 
 subroutine test_integer_value(x) bind(c)
   integer, value :: x
   call internal_call(x)
 end
+! CHECK-LABEL:   func.func @test_real_value(
+! CHECK-SAME:                               %[[VAL_0:.*]]: f32
+! CHECK:           %[[VAL_1:.*]] = fir.alloca f32
+! CHECK:           fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<f32>
+! CHECK:           %[[VAL_2:.*]] = fir.declare %[[VAL_1]]
+! CHECK:           fir.call @_QPinternal_call2(%[[VAL_2]]) {{.*}}: (!fir.ref<f32>) -> ()
+! CHECK:           return
+! CHECK:         }
 
-! CHECK-LABEL: func.func @test_real_value(
-! CHECK-SAME:                             %[[VAL_0:.*]]: f32 {fir.bindc_name = "x"}) attributes {fir.bindc_name = "test_real_value"} {
-! CHECK:         %[[VAL_1:.*]] = fir.alloca f32
-! CHECK:         fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<f32>
-! CHECK:         fir.call @_QPinternal_call2(%[[VAL_1]]) {{.*}}: (!fir.ref<f32>) -> ()
-! CHECK:         return
-! CHECK:       }
 
 subroutine test_real_value(x) bind(c)
   real, value :: x
   call internal_call2(x)
 end
+! CHECK-LABEL:   func.func @test_complex_value(
+! CHECK-SAME:                                  %[[VAL_0:.*]]: !fir.complex<4>
+! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.complex<4>
+! CHECK:           fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<!fir.complex<4>>
+! CHECK:           %[[VAL_2:.*]] = fir.declare %[[VAL_1]]
+! CHECK:           fir.call @_QPinternal_call3(%[[VAL_2]]) {{.*}}: (!fir.ref<!fir.complex<4>>) -> ()
+! CHECK:           return
+! CHECK:         }
 
-! CHECK-LABEL: func.func @test_complex_value(
-! CHECK-SAME:                                %[[VAL_0:.*]]: !fir.complex<4> {fir.bindc_name = "x"}) attributes {fir.bindc_name = "test_complex_value"} {
-! CHECK:         %[[VAL_1:.*]] = fir.alloca !fir.complex<4>
-! CHECK:         fir.store %[[VAL_0]] to %[[VAL_1]] : !fir.ref<!fir.complex<4>>
-! CHECK:         fir.call @_QPinternal_call3(%[[VAL_1]]) {{.*}}: (!fir.ref<!fir.complex<4>>) -> ()
-! CHECK:         return
-! CHECK:       }
 
 subroutine test_complex_value(x) bind(c)
   complex, value :: x
@@ -63,12 +67,13 @@ subroutine test_complex_value(x) bind(c)
 end
 
 ! CHECK-LABEL:   func.func @test_char_value(
-! CHECK-SAME:                               %[[VAL_0:.*]]: !fir.boxchar<1> {fir.bindc_name = "x"}) attributes {fir.bindc_name = "test_char_value"} {
-! CHECK:           %[[VAL_1:.*]]:2 = fir.unboxchar %[[VAL_0]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
-! CHECK:           %[[VAL_3:.*]] = fir.convert %[[VAL_1]]#0 : (!fir.ref<!fir.char<1,?>>) -> !fir.ref<!fir.char<1>>
-! CHECK:           %[[VAL_2:.*]] = arith.constant 1 : index
-! CHECK:           %[[VAL_5:.*]] = fir.emboxchar %[[VAL_3]], %[[VAL_2]] : (!fir.ref<!fir.char<1>>, index) -> !fir.boxchar<1>
-! CHECK:           fir.call @_QPinternal_call4(%[[VAL_5]]) fastmath<contract> : (!fir.boxchar<1>) -> ()
+! CHECK-SAME:                               %[[VAL_0:.*]]: !fir.char<1>
+! CHECK:           %[[VAL_1:.*]] = arith.constant 1 : index
+! CHECK:           %[[VAL_2:.*]] = fir.alloca !fir.char<1>
+! CHECK:           fir.store %[[VAL_0]] to %[[VAL_2]] : !fir.ref<!fir.char<1>>
+! CHECK:           %[[VAL_3:.*]] = fir.declare %[[VAL_2]] typeparams %[[VAL_1]]
+! CHECK:           %[[VAL_4:.*]] = fir.emboxchar %[[VAL_3]], %[[VAL_1]] : (!fir.ref<!fir.char<1>>, index) -> !fir.boxchar<1>
+! CHECK:           fir.call @_QPinternal_call4(%[[VAL_4]]) {{.*}}: (!fir.boxchar<1>) -> ()
 ! CHECK:           return
 ! CHECK:         }
 
@@ -77,19 +82,40 @@ subroutine test_char_value(x) bind(c)
   call internal_call4(x)
 end
 
-! CHECK-LABEL: func.func @_QPtest_cptr_value(
-! CHECK-SAME:                                %[[VAL_0:.*]]: !fir.ref<i64> {fir.bindc_name = "x"}) {
-! CHECK:         %[[VAL_1:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
-! CHECK:         %[[VAL_2:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
-! CHECK:         %[[VAL_3:.*]] = fir.coordinate_of %[[VAL_1]], %[[VAL_2]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
-! CHECK:         %[[VAL_4:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<i64>) -> i64
-! CHECK:         fir.store %[[VAL_4]] to %[[VAL_3]] : !fir.ref<i64>
-! CHECK:         fir.call @_QPinternal_call5(%[[VAL_1]]) {{.*}}: (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) -> ()
-! CHECK:         return
-! CHECK:       }
+! CHECK-LABEL:   func.func @_QPtest_call_char_value(
+! CHECK-SAME:                                       %[[VAL_0:.*]]: !fir.boxchar<1>
+! CHECK:           %[[VAL_1:.*]]:2 = fir.unboxchar %[[VAL_0]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
+! CHECK:           %[[VAL_2:.*]] = fir.declare %[[VAL_1]]#0 typeparams %[[VAL_1]]#1
+! CHECK:           %[[VAL_3:.*]] = fir.emboxchar %[[VAL_2]], %[[VAL_1]]#1 : (!fir.ref<!fir.char<1,?>>, index) -> !fir.boxchar<1>
+! CHECK:           %[[VAL_4:.*]] = fir.convert %[[VAL_2]] : (!fir.ref<!fir.char<1,?>>) -> !fir.ref<!fir.char<1>>
+! CHECK:           %[[VAL_5:.*]] = fir.load %[[VAL_4]] : !fir.ref<!fir.char<1>>
+! CHECK:           fir.call @test_char_value(%[[VAL_5]]) {{.*}}: (!fir.char<1>) -> ()
+! CHECK:           return
+! CHECK:         }
+subroutine test_call_char_value(x)
+  character(*) :: x
+  interface
+    subroutine test_char_value(x) bind(c)
+      character(1), value :: x
+    end
+  end interface
+  call test_char_value(x)
+end subroutine
+
+! CHECK-LABEL:   func.func @_QPtest_cptr_value(
+! CHECK-SAME:                                  %[[VAL_0:.*]]: !fir.ref<i64>
+! CHECK:           %[[VAL_1:.*]] = fir.alloca !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
+! CHECK:           %[[VAL_2:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
+! CHECK:           %[[VAL_3:.*]] = fir.coordinate_of %[[VAL_1]], %[[VAL_2]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
+! CHECK:           %[[VAL_4:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<i64>) -> i64
+! CHECK:           fir.store %[[VAL_4]] to %[[VAL_3]] : !fir.ref<i64>
+! CHECK:           %[[VAL_5:.*]] = fir.declare %[[VAL_1]]
+! CHECK:           fir.call @_QPinternal_call5(%[[VAL_5]]) fastmath<contract> : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) -> ()
+! CHECK:           return
+! CHECK:         }
 
 subroutine test_cptr_value(x)
-  use iso_c_binding
+  use iso_c_binding, only: c_ptr
   type(c_ptr), value :: x
   call internal_call5(x)
 end

From d488b2225d484027eb2c21a06d27decf008e878b Mon Sep 17 00:00:00 2001
From: Beal Wang <colorfishes@outlook.com>
Date: Fri, 12 Apr 2024 16:37:50 +0800
Subject: [PATCH 226/886] [mlir][ods] Do not print default-valued properties
 when the value is equal to the default (#87970)

This diff causes the `tblgen`-erated printProperties() function to skip
printing a `DefaultValuedAttr` property when the value is equal to the
default.

Co-authored-by: Biao Wang <biaow@nvidia.com>
---
 .../include/mlir/Dialect/XeGPU/IR/XeGPUOps.td |  3 +-
 mlir/include/mlir/IR/OpDefinition.h           | 31 ++++++++++++-------
 mlir/lib/IR/Operation.cpp                     | 30 ++++++++++++++----
 mlir/test/IR/properties.mlir                  |  5 +++
 mlir/test/lib/Dialect/Test/TestOps.td         | 17 +++++++---
 mlir/tools/mlir-tblgen/OpFormatGen.cpp        | 23 +++++++++++++-
 6 files changed, 85 insertions(+), 24 deletions(-)

diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
index 93c56ad05b432..b8ebd1a40c607 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
@@ -27,7 +27,8 @@ class XeGPU_Op<string mnemonic, list<Trait> traits = []>:
 
   code extraBaseClassDeclaration = [{
     void printProperties(::mlir::MLIRContext *ctx,
-            ::mlir::OpAsmPrinter &p, const Properties &prop) {
+            ::mlir::OpAsmPrinter &p, const Properties &prop,
+            ::mlir::ArrayRef<::llvm::StringRef> elidedProps) {
       Attribute propAttr = getPropertiesAsAttr(ctx, prop);
       if (propAttr)
         p << "<" << propAttr << ">";
diff --git a/mlir/include/mlir/IR/OpDefinition.h b/mlir/include/mlir/IR/OpDefinition.h
index 2d1dee2303e8f..f4355c8ce26ac 100644
--- a/mlir/include/mlir/IR/OpDefinition.h
+++ b/mlir/include/mlir/IR/OpDefinition.h
@@ -226,8 +226,10 @@ class OpState {
   static ParseResult genericParseProperties(OpAsmParser &parser,
                                             Attribute &result);
 
-  /// Print the properties as a Attribute.
-  static void genericPrintProperties(OpAsmPrinter &p, Attribute properties);
+  /// Print the properties as a Attribute with names not included within
+  /// 'elidedProps'
+  static void genericPrintProperties(OpAsmPrinter &p, Attribute properties,
+                                     ArrayRef<StringRef> elidedProps = {});
 
   /// Print an operation name, eliding the dialect prefix if necessary.
   static void printOpName(Operation *op, OpAsmPrinter &p,
@@ -1805,10 +1807,13 @@ class Op : public OpState, public Traits<ConcreteType>... {
   template <typename T>
   using detect_has_print = llvm::is_detected<has_print, T>;
 
-  /// Trait to check if printProperties(OpAsmPrinter, T) exist
+  /// Trait to check if printProperties(OpAsmPrinter, T, ArrayRef<StringRef>)
+  /// exist
   template <typename T, typename... Args>
-  using has_print_properties = decltype(printProperties(
-      std::declval<OpAsmPrinter &>(), std::declval<T>()));
+  using has_print_properties =
+      decltype(printProperties(std::declval<OpAsmPrinter &>(),
+                               std::declval<T>(),
+                               std::declval<ArrayRef<StringRef>>()));
   template <typename T>
   using detect_has_print_properties =
       llvm::is_detected<has_print_properties, T>;
@@ -1974,16 +1979,18 @@ class Op : public OpState, public Traits<ConcreteType>... {
   static void populateDefaultProperties(OperationName opName,
                                         InferredProperties<T> &properties) {}
 
-  /// Print the operation properties. Unless overridden, this method will try to
-  /// dispatch to a `printProperties` free-function if it exists, and otherwise
-  /// by converting the properties to an Attribute.
+  /// Print the operation properties with names not included within
+  /// 'elidedProps'. Unless overridden, this method will try to dispatch to a
+  /// `printProperties` free-function if it exists, and otherwise by converting
+  /// the properties to an Attribute.
   template <typename T>
   static void printProperties(MLIRContext *ctx, OpAsmPrinter &p,
-                              const T &properties) {
+                              const T &properties,
+                              ArrayRef<StringRef> elidedProps = {}) {
     if constexpr (detect_has_print_properties<T>::value)
-      return printProperties(p, properties);
-    genericPrintProperties(p,
-                           ConcreteType::getPropertiesAsAttr(ctx, properties));
+      return printProperties(p, properties, elidedProps);
+    genericPrintProperties(
+        p, ConcreteType::getPropertiesAsAttr(ctx, properties), elidedProps);
   }
 
   /// Parser the properties. Unless overridden, this method will print by
diff --git a/mlir/lib/IR/Operation.cpp b/mlir/lib/IR/Operation.cpp
index ca5ff9f72e3e2..db903d540761b 100644
--- a/mlir/lib/IR/Operation.cpp
+++ b/mlir/lib/IR/Operation.cpp
@@ -790,15 +790,33 @@ void OpState::printOpName(Operation *op, OpAsmPrinter &p,
 /// Parse properties as a Attribute.
 ParseResult OpState::genericParseProperties(OpAsmParser &parser,
                                             Attribute &result) {
-  if (parser.parseLess() || parser.parseAttribute(result) ||
-      parser.parseGreater())
-    return failure();
+  if (succeeded(parser.parseOptionalLess())) { // The less is optional.
+    if (parser.parseAttribute(result) || parser.parseGreater())
+      return failure();
+  }
   return success();
 }
 
-/// Print the properties as a Attribute.
-void OpState::genericPrintProperties(OpAsmPrinter &p, Attribute properties) {
-  p << "<" << properties << ">";
+/// Print the properties as a Attribute with names not included within
+/// 'elidedProps'
+void OpState::genericPrintProperties(OpAsmPrinter &p, Attribute properties,
+                                     ArrayRef<StringRef> elidedProps) {
+  auto dictAttr = dyn_cast_or_null<::mlir::DictionaryAttr>(properties);
+  if (dictAttr && !elidedProps.empty()) {
+    ArrayRef<NamedAttribute> attrs = dictAttr.getValue();
+    llvm::SmallDenseSet<StringRef> elidedAttrsSet(elidedProps.begin(),
+                                                  elidedProps.end());
+    bool atLeastOneAttr = llvm::any_of(attrs, [&](NamedAttribute attr) {
+      return !elidedAttrsSet.contains(attr.getName().strref());
+    });
+    if (atLeastOneAttr) {
+      p << "<";
+      p.printOptionalAttrDict(dictAttr.getValue(), elidedProps);
+      p << ">";
+    }
+  } else {
+    p << "<" << properties << ">";
+  }
 }
 
 /// Emit an error about fatal conditions with this operation, reporting up to
diff --git a/mlir/test/IR/properties.mlir b/mlir/test/IR/properties.mlir
index 3c4bd57859ef9..1d22cb1940f24 100644
--- a/mlir/test/IR/properties.mlir
+++ b/mlir/test/IR/properties.mlir
@@ -33,3 +33,8 @@ test.using_property_in_custom [1, 4, 20]
 // GENERIC-SAME: second = 4
 // GENERIC-SAME: }>
 test.using_property_ref_in_custom 1 + 4 = 5
+
+// CHECK:   test.with_default_valued_properties {{$}}
+// GENERIC: "test.with_default_valued_properties"()
+// GENERIC-SAME:  <{a = 0 : i32}>
+test.with_default_valued_properties <{a = 0 : i32}>
diff --git a/mlir/test/lib/Dialect/Test/TestOps.td b/mlir/test/lib/Dialect/Test/TestOps.td
index e6c3601d08dad..edca05fde5a52 100644
--- a/mlir/test/lib/Dialect/Test/TestOps.td
+++ b/mlir/test/lib/Dialect/Test/TestOps.td
@@ -2909,7 +2909,8 @@ def TestOpWithNiceProperties : TEST_Op<"with_nice_properties"> {
   );
   let extraClassDeclaration = [{
     void printProperties(::mlir::MLIRContext *ctx, ::mlir::OpAsmPrinter &p,
-                         const Properties &prop);
+                         const Properties &prop,
+                         ::mlir::ArrayRef<::llvm::StringRef> elidedProps);
     static ::mlir::ParseResult  parseProperties(::mlir::OpAsmParser &parser,
                                      ::mlir::OperationState &result);
     static ::mlir::LogicalResult readFromMlirBytecode(
@@ -2938,7 +2939,8 @@ def TestOpWithNiceProperties : TEST_Op<"with_nice_properties"> {
       writer.writeVarInt(prop.value);
     }
     void TestOpWithNiceProperties::printProperties(::mlir::MLIRContext *ctx,
-            ::mlir::OpAsmPrinter &p, const Properties &prop) {
+            ::mlir::OpAsmPrinter &p, const Properties &prop,
+            ::mlir::ArrayRef<::llvm::StringRef> elidedProps) {
       customPrintProperties(p, prop.prop);
     }
     ::mlir::ParseResult TestOpWithNiceProperties::parseProperties(
@@ -2971,7 +2973,8 @@ def TestOpWithVersionedProperties : TEST_Op<"with_versioned_properties"> {
   );
   let extraClassDeclaration = [{
     void printProperties(::mlir::MLIRContext *ctx, ::mlir::OpAsmPrinter &p,
-                         const Properties &prop);
+                         const Properties &prop,
+                         ::mlir::ArrayRef<::llvm::StringRef> elidedProps);
     static ::mlir::ParseResult  parseProperties(::mlir::OpAsmParser &parser,
                                      ::mlir::OperationState &result);
     static ::mlir::LogicalResult readFromMlirBytecode(
@@ -2983,7 +2986,8 @@ def TestOpWithVersionedProperties : TEST_Op<"with_versioned_properties"> {
   }];
   let extraClassDefinition = [{
     void TestOpWithVersionedProperties::printProperties(::mlir::MLIRContext *ctx,
-            ::mlir::OpAsmPrinter &p, const Properties &prop) {
+            ::mlir::OpAsmPrinter &p, const Properties &prop,
+            ::mlir::ArrayRef<::llvm::StringRef> elidedProps) {
       customPrintProperties(p, prop.prop);
     }
     ::mlir::ParseResult TestOpWithVersionedProperties::parseProperties(
@@ -2997,6 +3001,11 @@ def TestOpWithVersionedProperties : TEST_Op<"with_versioned_properties"> {
   }];
 }
 
+def TestOpWithDefaultValuedProperties : TEST_Op<"with_default_valued_properties"> {
+  let assemblyFormat = "prop-dict attr-dict";
+  let arguments = (ins DefaultValuedAttr<I32Attr, "0">:$a);
+}
+
 //===----------------------------------------------------------------------===//
 // Test Dataflow
 //===----------------------------------------------------------------------===//
diff --git a/mlir/tools/mlir-tblgen/OpFormatGen.cpp b/mlir/tools/mlir-tblgen/OpFormatGen.cpp
index c8e0476d45b9a..5963b5e689da7 100644
--- a/mlir/tools/mlir-tblgen/OpFormatGen.cpp
+++ b/mlir/tools/mlir-tblgen/OpFormatGen.cpp
@@ -1775,9 +1775,30 @@ const char *enumAttrBeginPrinterCode = R"(
 /// Generate the printer for the 'prop-dict' directive.
 static void genPropDictPrinter(OperationFormat &fmt, Operator &op,
                                MethodBody &body) {
+  body << "  ::llvm::SmallVector<::llvm::StringRef, 2> elidedProps;\n";
+  // Add code to check attributes for equality with the default value
+  // for attributes with the elidePrintingDefaultValue bit set.
+  for (const NamedAttribute &namedAttr : op.getAttributes()) {
+    const Attribute &attr = namedAttr.attr;
+    if (!attr.isDerivedAttr() && attr.hasDefaultValue()) {
+      const StringRef &name = namedAttr.name;
+      FmtContext fctx;
+      fctx.withBuilder("odsBuilder");
+      std::string defaultValue = std::string(
+          tgfmt(attr.getConstBuilderTemplate(), &fctx, attr.getDefaultValue()));
+      body << "  {\n";
+      body << "     ::mlir::Builder odsBuilder(getContext());\n";
+      body << "     ::mlir::Attribute attr = " << op.getGetterName(name)
+           << "Attr();\n";
+      body << "     if(attr && (attr == " << defaultValue << "))\n";
+      body << "       elidedProps.push_back(\"" << name << "\");\n";
+      body << "  }\n";
+    }
+  }
+
   body << "  _odsPrinter << \" \";\n"
        << "  printProperties(this->getContext(), _odsPrinter, "
-          "getProperties());\n";
+          "getProperties(), elidedProps);\n";
 }
 
 /// Generate the printer for the 'attr-dict' directive.

From d019b9aaaaa60c858ee82fdbaf5de16e048e9db2 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Fri, 12 Apr 2024 12:20:22 +0300
Subject: [PATCH 227/886] [clang][NFC] Refactor `CXXSpecialMember`

In preparation for `SemaCUDA`, which requires this enum to be forward-declarable.
---
 clang/include/clang/Sema/Sema.h               |  64 ++-
 clang/lib/Sema/SemaAccess.cpp                 |  14 +-
 clang/lib/Sema/SemaCUDA.cpp                   |   2 +-
 clang/lib/Sema/SemaDecl.cpp                   |  51 +-
 clang/lib/Sema/SemaDeclCXX.cpp                | 522 ++++++++++--------
 clang/lib/Sema/SemaInit.cpp                   |   6 +-
 clang/lib/Sema/SemaLookup.cpp                 |  78 +--
 clang/lib/Sema/SemaOverload.cpp               |  16 +-
 clang/lib/Sema/SemaTemplateInstantiate.cpp    |   7 +-
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |   8 +-
 clang/lib/Sema/SemaType.cpp                   |   3 +-
 11 files changed, 424 insertions(+), 347 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index c4a603cc5a4a7..90c9166282703 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -60,6 +60,7 @@
 #include "clang/Sema/TypoCorrection.h"
 #include "clang/Sema/Weak.h"
 #include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/SmallBitVector.h"
 #include "llvm/ADT/SmallPtrSet.h"
@@ -423,6 +424,17 @@ enum class TemplateDeductionResult {
   AlreadyDiagnosed
 };
 
+/// Kinds of C++ special members.
+enum class CXXSpecialMemberKind {
+  DefaultConstructor,
+  CopyConstructor,
+  MoveConstructor,
+  CopyAssignment,
+  MoveAssignment,
+  Destructor,
+  Invalid
+};
+
 /// Sema - This implements semantic analysis and AST building for C.
 /// \nosubgrouping
 class Sema final : public SemaBase {
@@ -4077,22 +4089,11 @@ class Sema final : public SemaBase {
       SourceRange SpecificationRange, ArrayRef<ParsedType> DynamicExceptions,
       ArrayRef<SourceRange> DynamicExceptionRanges, Expr *NoexceptExpr);
 
-  /// Kinds of C++ special members.
-  enum CXXSpecialMember {
-    CXXDefaultConstructor,
-    CXXCopyConstructor,
-    CXXMoveConstructor,
-    CXXCopyAssignment,
-    CXXMoveAssignment,
-    CXXDestructor,
-    CXXInvalid
-  };
-
   class InheritedConstructorInfo;
 
   /// Determine if a special member function should have a deleted
   /// definition when it is defaulted.
-  bool ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
+  bool ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMemberKind CSM,
                                  InheritedConstructorInfo *ICI = nullptr,
                                  bool Diagnose = false);
 
@@ -4458,7 +4459,7 @@ class Sema final : public SemaBase {
   void CheckExplicitlyDefaultedFunction(Scope *S, FunctionDecl *MD);
 
   bool CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
-                                             CXXSpecialMember CSM,
+                                             CXXSpecialMemberKind CSM,
                                              SourceLocation DefaultLoc);
   void CheckDelayedMemberExceptionSpecs();
 
@@ -4618,7 +4619,7 @@ class Sema final : public SemaBase {
   void CheckCXXDefaultArguments(FunctionDecl *FD);
   void CheckExtraCXXDefaultArguments(Declarator &D);
 
-  CXXSpecialMember getSpecialMember(const CXXMethodDecl *MD) {
+  CXXSpecialMemberKind getSpecialMember(const CXXMethodDecl *MD) {
     return getDefaultedFunctionKind(MD).asSpecialMember();
   }
 
@@ -4645,7 +4646,8 @@ class Sema final : public SemaBase {
                                    AccessSpecifier AS,
                                    const ParsedAttr &MSPropertyAttr);
 
-  void DiagnoseNontrivial(const CXXRecordDecl *Record, CXXSpecialMember CSM);
+  void DiagnoseNontrivial(const CXXRecordDecl *Record,
+                          CXXSpecialMemberKind CSM);
 
   enum TrivialABIHandling {
     /// The triviality of a method unaffected by "trivial_abi".
@@ -4655,26 +4657,31 @@ class Sema final : public SemaBase {
     TAH_ConsiderTrivialABI
   };
 
-  bool SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
+  bool SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMemberKind CSM,
                               TrivialABIHandling TAH = TAH_IgnoreTrivialABI,
                               bool Diagnose = false);
 
   /// For a defaulted function, the kind of defaulted function that it is.
   class DefaultedFunctionKind {
+    LLVM_PREFERRED_TYPE(CXXSpecialMemberKind)
     unsigned SpecialMember : 8;
     unsigned Comparison : 8;
 
   public:
     DefaultedFunctionKind()
-        : SpecialMember(CXXInvalid),
+        : SpecialMember(llvm::to_underlying(CXXSpecialMemberKind::Invalid)),
           Comparison(llvm::to_underlying(DefaultedComparisonKind::None)) {}
-    DefaultedFunctionKind(CXXSpecialMember CSM)
-        : SpecialMember(CSM),
+    DefaultedFunctionKind(CXXSpecialMemberKind CSM)
+        : SpecialMember(llvm::to_underlying(CSM)),
           Comparison(llvm::to_underlying(DefaultedComparisonKind::None)) {}
     DefaultedFunctionKind(DefaultedComparisonKind Comp)
-        : SpecialMember(CXXInvalid), Comparison(llvm::to_underlying(Comp)) {}
+        : SpecialMember(llvm::to_underlying(CXXSpecialMemberKind::Invalid)),
+          Comparison(llvm::to_underlying(Comp)) {}
 
-    bool isSpecialMember() const { return SpecialMember != CXXInvalid; }
+    bool isSpecialMember() const {
+      return static_cast<CXXSpecialMemberKind>(SpecialMember) !=
+             CXXSpecialMemberKind::Invalid;
+    }
     bool isComparison() const {
       return static_cast<DefaultedComparisonKind>(Comparison) !=
              DefaultedComparisonKind::None;
@@ -4684,8 +4691,8 @@ class Sema final : public SemaBase {
       return isSpecialMember() || isComparison();
     }
 
-    CXXSpecialMember asSpecialMember() const {
-      return static_cast<CXXSpecialMember>(SpecialMember);
+    CXXSpecialMemberKind asSpecialMember() const {
+      return static_cast<CXXSpecialMemberKind>(SpecialMember);
     }
     DefaultedComparisonKind asComparison() const {
       return static_cast<DefaultedComparisonKind>(Comparison);
@@ -4693,7 +4700,8 @@ class Sema final : public SemaBase {
 
     /// Get the index of this function kind for use in diagnostics.
     unsigned getDiagnosticIndex() const {
-      static_assert(CXXInvalid > CXXDestructor,
+      static_assert(llvm::to_underlying(CXXSpecialMemberKind::Invalid) >
+                        llvm::to_underlying(CXXSpecialMemberKind::Destructor),
                     "invalid should have highest index");
       static_assert((unsigned)DefaultedComparisonKind::None == 0,
                     "none should be equal to zero");
@@ -4799,7 +4807,7 @@ class Sema final : public SemaBase {
   /// definition in this translation unit.
   llvm::MapVector<NamedDecl *, SourceLocation> UndefinedButUsed;
 
-  typedef llvm::PointerIntPair<CXXRecordDecl *, 3, CXXSpecialMember>
+  typedef llvm::PointerIntPair<CXXRecordDecl *, 3, CXXSpecialMemberKind>
       SpecialMemberDecl;
 
   /// The C++ special members which we are currently in the process of
@@ -7493,7 +7501,7 @@ class Sema final : public SemaBase {
   };
 
   SpecialMemberOverloadResult
-  LookupSpecialMember(CXXRecordDecl *D, CXXSpecialMember SM, bool ConstArg,
+  LookupSpecialMember(CXXRecordDecl *D, CXXSpecialMemberKind SM, bool ConstArg,
                       bool VolatileArg, bool RValueThis, bool ConstThis,
                       bool VolatileThis);
 
@@ -10014,7 +10022,7 @@ class Sema final : public SemaBase {
       unsigned NumCallArgs;
 
       /// The special member being declared or defined.
-      CXXSpecialMember SpecialMember;
+      CXXSpecialMemberKind SpecialMember;
     };
 
     ArrayRef<TemplateArgument> template_arguments() const {
@@ -13109,7 +13117,7 @@ class Sema final : public SemaBase {
   /// The result of this call is implicit CUDA target attribute(s) attached to
   /// the member declaration.
   bool inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
-                                               CXXSpecialMember CSM,
+                                               CXXSpecialMemberKind CSM,
                                                CXXMethodDecl *MemberDecl,
                                                bool ConstRHS, bool Diagnose);
 
diff --git a/clang/lib/Sema/SemaAccess.cpp b/clang/lib/Sema/SemaAccess.cpp
index 4af3c0f30a8e8..6a707eeb66d01 100644
--- a/clang/lib/Sema/SemaAccess.cpp
+++ b/clang/lib/Sema/SemaAccess.cpp
@@ -10,8 +10,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "clang/Basic/Specifiers.h"
-#include "clang/Sema/SemaInternal.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/CXXInheritance.h"
 #include "clang/AST/DeclCXX.h"
@@ -19,9 +17,12 @@
 #include "clang/AST/DeclObjC.h"
 #include "clang/AST/DependentDiagnostic.h"
 #include "clang/AST/ExprCXX.h"
+#include "clang/Basic/Specifiers.h"
 #include "clang/Sema/DelayedDiagnostic.h"
 #include "clang/Sema/Initialization.h"
 #include "clang/Sema/Lookup.h"
+#include "clang/Sema/SemaInternal.h"
+#include "llvm/ADT/STLForwardCompat.h"
 
 using namespace clang;
 using namespace sema;
@@ -1658,21 +1659,24 @@ Sema::AccessResult Sema::CheckConstructorAccess(SourceLocation UseLoc,
   case InitializedEntity::EK_Base:
     PD = PDiag(diag::err_access_base_ctor);
     PD << Entity.isInheritedVirtualBase()
-       << Entity.getBaseSpecifier()->getType() << getSpecialMember(Constructor);
+       << Entity.getBaseSpecifier()->getType()
+       << llvm::to_underlying(getSpecialMember(Constructor));
     break;
 
   case InitializedEntity::EK_Member:
   case InitializedEntity::EK_ParenAggInitMember: {
     const FieldDecl *Field = cast<FieldDecl>(Entity.getDecl());
     PD = PDiag(diag::err_access_field_ctor);
-    PD << Field->getType() << getSpecialMember(Constructor);
+    PD << Field->getType()
+       << llvm::to_underlying(getSpecialMember(Constructor));
     break;
   }
 
   case InitializedEntity::EK_LambdaCapture: {
     StringRef VarName = Entity.getCapturedVarName();
     PD = PDiag(diag::err_access_lambda_capture);
-    PD << VarName << Entity.getType() << getSpecialMember(Constructor);
+    PD << VarName << Entity.getType()
+       << llvm::to_underlying(getSpecialMember(Constructor));
     break;
   }
 
diff --git a/clang/lib/Sema/SemaCUDA.cpp b/clang/lib/Sema/SemaCUDA.cpp
index 4d4f4b6a2d4d9..1596222e3d1da 100644
--- a/clang/lib/Sema/SemaCUDA.cpp
+++ b/clang/lib/Sema/SemaCUDA.cpp
@@ -358,7 +358,7 @@ resolveCalleeCUDATargetConflict(Sema::CUDAFunctionTarget Target1,
 }
 
 bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
-                                                   CXXSpecialMember CSM,
+                                                   CXXSpecialMemberKind CSM,
                                                    CXXMethodDecl *MemberDecl,
                                                    bool ConstRHS,
                                                    bool Diagnose) {
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index afd9e407946a0..b448bacd49f32 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -48,6 +48,7 @@
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/TargetParser/Triple.h"
@@ -4049,13 +4050,13 @@ bool Sema::MergeFunctionDecl(FunctionDecl *New, NamedDecl *&OldD, Scope *S,
         } else {
           Diag(NewMethod->getLocation(),
                diag::err_definition_of_implicitly_declared_member)
-            << New << getSpecialMember(OldMethod);
+              << New << llvm::to_underlying(getSpecialMember(OldMethod));
           return true;
         }
       } else if (OldMethod->getFirstDecl()->isExplicitlyDefaulted() && !isFriend) {
         Diag(NewMethod->getLocation(),
              diag::err_definition_of_explicitly_defaulted_member)
-          << getSpecialMember(OldMethod);
+            << llvm::to_underlying(getSpecialMember(OldMethod));
         return true;
       }
     }
@@ -18847,22 +18848,22 @@ bool Sema::CheckNontrivialField(FieldDecl *FD) {
       // because otherwise we'll never get complaints about
       // copy constructors.
 
-      CXXSpecialMember member = CXXInvalid;
+      CXXSpecialMemberKind member = CXXSpecialMemberKind::Invalid;
       // We're required to check for any non-trivial constructors. Since the
       // implicit default constructor is suppressed if there are any
       // user-declared constructors, we just need to check that there is a
       // trivial default constructor and a trivial copy constructor. (We don't
       // worry about move constructors here, since this is a C++98 check.)
       if (RDecl->hasNonTrivialCopyConstructor())
-        member = CXXCopyConstructor;
+        member = CXXSpecialMemberKind::CopyConstructor;
       else if (!RDecl->hasTrivialDefaultConstructor())
-        member = CXXDefaultConstructor;
+        member = CXXSpecialMemberKind::DefaultConstructor;
       else if (RDecl->hasNonTrivialCopyAssignment())
-        member = CXXCopyAssignment;
+        member = CXXSpecialMemberKind::CopyAssignment;
       else if (RDecl->hasNonTrivialDestructor())
-        member = CXXDestructor;
+        member = CXXSpecialMemberKind::Destructor;
 
-      if (member != CXXInvalid) {
+      if (member != CXXSpecialMemberKind::Invalid) {
         if (!getLangOpts().CPlusPlus11 &&
             getLangOpts().ObjCAutoRefCount && RDecl->hasObjectMember()) {
           // Objective-C++ ARC: it is an error to have a non-trivial field of
@@ -18879,10 +18880,13 @@ bool Sema::CheckNontrivialField(FieldDecl *FD) {
           }
         }
 
-        Diag(FD->getLocation(), getLangOpts().CPlusPlus11 ?
-               diag::warn_cxx98_compat_nontrivial_union_or_anon_struct_member :
-               diag::err_illegal_union_or_anon_struct_member)
-          << FD->getParent()->isUnion() << FD->getDeclName() << member;
+        Diag(
+            FD->getLocation(),
+            getLangOpts().CPlusPlus11
+                ? diag::warn_cxx98_compat_nontrivial_union_or_anon_struct_member
+                : diag::err_illegal_union_or_anon_struct_member)
+            << FD->getParent()->isUnion() << FD->getDeclName()
+            << llvm::to_underlying(member);
         DiagnoseNontrivial(RDecl, member);
         return !getLangOpts().CPlusPlus11;
       }
@@ -19132,10 +19136,10 @@ static void ComputeSelectedDestructor(Sema &S, CXXRecordDecl *Record) {
 static bool AreSpecialMemberFunctionsSameKind(ASTContext &Context,
                                               CXXMethodDecl *M1,
                                               CXXMethodDecl *M2,
-                                              Sema::CXXSpecialMember CSM) {
+                                              CXXSpecialMemberKind CSM) {
   // We don't want to compare templates to non-templates: See
   // https://github.com/llvm/llvm-project/issues/59206
-  if (CSM == Sema::CXXDefaultConstructor)
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor)
     return bool(M1->getDescribedFunctionTemplate()) ==
            bool(M2->getDescribedFunctionTemplate());
   // FIXME: better resolve CWG
@@ -19158,7 +19162,7 @@ static bool AreSpecialMemberFunctionsSameKind(ASTContext &Context,
 ///   [CWG2595], if any, are satisfied is more constrained.
 static void SetEligibleMethods(Sema &S, CXXRecordDecl *Record,
                                ArrayRef<CXXMethodDecl *> Methods,
-                               Sema::CXXSpecialMember CSM) {
+                               CXXSpecialMemberKind CSM) {
   SmallVector<bool, 4> SatisfactionStatus;
 
   for (CXXMethodDecl *Method : Methods) {
@@ -19216,7 +19220,8 @@ static void SetEligibleMethods(Sema &S, CXXRecordDecl *Record,
     // DR1734 and DR1496.
     if (!AnotherMethodIsMoreConstrained) {
       Method->setIneligibleOrNotSelected(false);
-      Record->addedEligibleSpecialMemberFunction(Method, 1 << CSM);
+      Record->addedEligibleSpecialMemberFunction(Method,
+                                                 1 << llvm::to_underlying(CSM));
     }
   }
 }
@@ -19255,13 +19260,15 @@ static void ComputeSpecialMemberFunctionsEligiblity(Sema &S,
   }
 
   SetEligibleMethods(S, Record, DefaultConstructors,
-                     Sema::CXXDefaultConstructor);
-  SetEligibleMethods(S, Record, CopyConstructors, Sema::CXXCopyConstructor);
-  SetEligibleMethods(S, Record, MoveConstructors, Sema::CXXMoveConstructor);
+                     CXXSpecialMemberKind::DefaultConstructor);
+  SetEligibleMethods(S, Record, CopyConstructors,
+                     CXXSpecialMemberKind::CopyConstructor);
+  SetEligibleMethods(S, Record, MoveConstructors,
+                     CXXSpecialMemberKind::MoveConstructor);
   SetEligibleMethods(S, Record, CopyAssignmentOperators,
-                     Sema::CXXCopyAssignment);
+                     CXXSpecialMemberKind::CopyAssignment);
   SetEligibleMethods(S, Record, MoveAssignmentOperators,
-                     Sema::CXXMoveAssignment);
+                     CXXSpecialMemberKind::MoveAssignment);
 }
 
 void Sema::ActOnFields(Scope *S, SourceLocation RecLoc, Decl *EnclosingDecl,
@@ -19630,7 +19637,7 @@ void Sema::ActOnFields(Scope *S, SourceLocation RecLoc, Decl *EnclosingDecl,
     if (CXXRecord) {
       auto *Dtor = CXXRecord->getDestructor();
       if (Dtor && Dtor->isImplicit() &&
-          ShouldDeleteSpecialMember(Dtor, CXXDestructor)) {
+          ShouldDeleteSpecialMember(Dtor, CXXSpecialMemberKind::Destructor)) {
         CXXRecord->setImplicitDestructorIsDeleted();
         SetDeclDeleted(Dtor, CXXRecord->getLocation());
       }
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 858951580ea45..51c14443d2d8f 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -46,6 +46,7 @@
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/ScopeExit.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringExtras.h"
@@ -657,13 +658,13 @@ bool Sema::MergeCXXFunctionDecl(FunctionDecl *New, FunctionDecl *Old,
   // is ill-formed. This can only happen for constructors.
   if (isa<CXXConstructorDecl>(New) &&
       New->getMinRequiredArguments() < Old->getMinRequiredArguments()) {
-    CXXSpecialMember NewSM = getSpecialMember(cast<CXXMethodDecl>(New)),
-                     OldSM = getSpecialMember(cast<CXXMethodDecl>(Old));
+    CXXSpecialMemberKind NewSM = getSpecialMember(cast<CXXMethodDecl>(New)),
+                         OldSM = getSpecialMember(cast<CXXMethodDecl>(Old));
     if (NewSM != OldSM) {
       ParmVarDecl *NewParam = New->getParamDecl(New->getMinRequiredArguments());
       assert(NewParam->hasDefaultArg());
       Diag(NewParam->getLocation(), diag::err_default_arg_makes_ctor_special)
-        << NewParam->getDefaultArgRange() << NewSM;
+          << NewParam->getDefaultArgRange() << llvm::to_underlying(NewSM);
       Diag(Old->getLocation(), diag::note_previous_declaration);
     }
   }
@@ -6779,7 +6780,7 @@ void Sema::propagateDLLAttrToBaseClassTemplate(
 ///
 /// If the function is both a default constructor and a copy / move constructor
 /// (due to having a default argument for the first parameter), this picks
-/// CXXDefaultConstructor.
+/// CXXSpecialMemberKind::DefaultConstructor.
 ///
 /// FIXME: Check that case is properly handled by all callers.
 Sema::DefaultedFunctionKind
@@ -6787,23 +6788,23 @@ Sema::getDefaultedFunctionKind(const FunctionDecl *FD) {
   if (auto *MD = dyn_cast<CXXMethodDecl>(FD)) {
     if (const CXXConstructorDecl *Ctor = dyn_cast<CXXConstructorDecl>(FD)) {
       if (Ctor->isDefaultConstructor())
-        return Sema::CXXDefaultConstructor;
+        return CXXSpecialMemberKind::DefaultConstructor;
 
       if (Ctor->isCopyConstructor())
-        return Sema::CXXCopyConstructor;
+        return CXXSpecialMemberKind::CopyConstructor;
 
       if (Ctor->isMoveConstructor())
-        return Sema::CXXMoveConstructor;
+        return CXXSpecialMemberKind::MoveConstructor;
     }
 
     if (MD->isCopyAssignmentOperator())
-      return Sema::CXXCopyAssignment;
+      return CXXSpecialMemberKind::CopyAssignment;
 
     if (MD->isMoveAssignmentOperator())
-      return Sema::CXXMoveAssignment;
+      return CXXSpecialMemberKind::MoveAssignment;
 
     if (isa<CXXDestructorDecl>(FD))
-      return Sema::CXXDestructor;
+      return CXXSpecialMemberKind::Destructor;
   }
 
   switch (FD->getDeclName().getCXXOverloadedOperator()) {
@@ -6843,26 +6844,26 @@ static void DefineDefaultedFunction(Sema &S, FunctionDecl *FD,
     return S.DefineDefaultedComparison(DefaultLoc, FD, DFK.asComparison());
 
   switch (DFK.asSpecialMember()) {
-  case Sema::CXXDefaultConstructor:
+  case CXXSpecialMemberKind::DefaultConstructor:
     S.DefineImplicitDefaultConstructor(DefaultLoc,
                                        cast<CXXConstructorDecl>(FD));
     break;
-  case Sema::CXXCopyConstructor:
+  case CXXSpecialMemberKind::CopyConstructor:
     S.DefineImplicitCopyConstructor(DefaultLoc, cast<CXXConstructorDecl>(FD));
     break;
-  case Sema::CXXCopyAssignment:
+  case CXXSpecialMemberKind::CopyAssignment:
     S.DefineImplicitCopyAssignment(DefaultLoc, cast<CXXMethodDecl>(FD));
     break;
-  case Sema::CXXDestructor:
+  case CXXSpecialMemberKind::Destructor:
     S.DefineImplicitDestructor(DefaultLoc, cast<CXXDestructorDecl>(FD));
     break;
-  case Sema::CXXMoveConstructor:
+  case CXXSpecialMemberKind::MoveConstructor:
     S.DefineImplicitMoveConstructor(DefaultLoc, cast<CXXConstructorDecl>(FD));
     break;
-  case Sema::CXXMoveAssignment:
+  case CXXSpecialMemberKind::MoveAssignment:
     S.DefineImplicitMoveAssignment(DefaultLoc, cast<CXXMethodDecl>(FD));
     break;
-  case Sema::CXXInvalid:
+  case CXXSpecialMemberKind::Invalid:
     llvm_unreachable("Invalid special member.");
   }
 }
@@ -7182,9 +7183,9 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
 
     // For an explicitly defaulted or deleted special member, we defer
     // determining triviality until the class is complete. That time is now!
-    CXXSpecialMember CSM = getSpecialMember(M);
+    CXXSpecialMemberKind CSM = getSpecialMember(M);
     if (!M->isImplicit() && !M->isUserProvided()) {
-      if (CSM != CXXInvalid) {
+      if (CSM != CXXSpecialMemberKind::Invalid) {
         M->setTrivial(SpecialMemberIsTrivial(M, CSM));
         // Inform the class that we've finished declaring this member.
         Record->finishedDefaultedOrDeletedMember(M);
@@ -7197,8 +7198,10 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
 
     // Set triviality for the purpose of calls if this is a user-provided
     // copy/move constructor or destructor.
-    if ((CSM == CXXCopyConstructor || CSM == CXXMoveConstructor ||
-         CSM == CXXDestructor) && M->isUserProvided()) {
+    if ((CSM == CXXSpecialMemberKind::CopyConstructor ||
+         CSM == CXXSpecialMemberKind::MoveConstructor ||
+         CSM == CXXSpecialMemberKind::Destructor) &&
+        M->isUserProvided()) {
       M->setTrivialForCall(HasTrivialABI);
       Record->setTrivialForCallFlags(M);
     }
@@ -7207,8 +7210,9 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
         M->hasAttr<DLLExportAttr>()) {
       if (getLangOpts().isCompatibleWithMSVC(LangOptions::MSVC2015) &&
           M->isTrivial() &&
-          (CSM == CXXDefaultConstructor || CSM == CXXCopyConstructor ||
-           CSM == CXXDestructor))
+          (CSM == CXXSpecialMemberKind::DefaultConstructor ||
+           CSM == CXXSpecialMemberKind::CopyConstructor ||
+           CSM == CXXSpecialMemberKind::Destructor))
         M->dropAttr<DLLExportAttr>();
 
       if (M->hasAttr<DLLExportAttr>()) {
@@ -7220,8 +7224,8 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
     // Define defaulted constexpr virtual functions that override a base class
     // function right away.
     // FIXME: We can defer doing this until the vtable is marked as used.
-    if (CSM != CXXInvalid && !M->isDeleted() && M->isDefaulted() &&
-        M->isConstexpr() && M->size_overridden_methods())
+    if (CSM != CXXSpecialMemberKind::Invalid && !M->isDeleted() &&
+        M->isDefaulted() && M->isConstexpr() && M->size_overridden_methods())
       DefineDefaultedFunction(*this, M, M->getLocation());
 
     if (!Incomplete)
@@ -7343,15 +7347,18 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
 /// \param ConstRHS True if this is a copy operation with a const object
 ///        on its RHS, that is, if the argument to the outer special member
 ///        function is 'const' and this is not a field marked 'mutable'.
-static Sema::SpecialMemberOverloadResult lookupCallFromSpecialMember(
-    Sema &S, CXXRecordDecl *Class, Sema::CXXSpecialMember CSM,
-    unsigned FieldQuals, bool ConstRHS) {
+static Sema::SpecialMemberOverloadResult
+lookupCallFromSpecialMember(Sema &S, CXXRecordDecl *Class,
+                            CXXSpecialMemberKind CSM, unsigned FieldQuals,
+                            bool ConstRHS) {
   unsigned LHSQuals = 0;
-  if (CSM == Sema::CXXCopyAssignment || CSM == Sema::CXXMoveAssignment)
+  if (CSM == CXXSpecialMemberKind::CopyAssignment ||
+      CSM == CXXSpecialMemberKind::MoveAssignment)
     LHSQuals = FieldQuals;
 
   unsigned RHSQuals = FieldQuals;
-  if (CSM == Sema::CXXDefaultConstructor || CSM == Sema::CXXDestructor)
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor ||
+      CSM == CXXSpecialMemberKind::Destructor)
     RHSQuals = 0;
   else if (ConstRHS)
     RHSQuals |= Qualifiers::Const;
@@ -7447,12 +7454,10 @@ class Sema::InheritedConstructorInfo {
 
 /// Is the special member function which would be selected to perform the
 /// specified operation on the specified class type a constexpr constructor?
-static bool
-specialMemberIsConstexpr(Sema &S, CXXRecordDecl *ClassDecl,
-                         Sema::CXXSpecialMember CSM, unsigned Quals,
-                         bool ConstRHS,
-                         CXXConstructorDecl *InheritedCtor = nullptr,
-                         Sema::InheritedConstructorInfo *Inherited = nullptr) {
+static bool specialMemberIsConstexpr(
+    Sema &S, CXXRecordDecl *ClassDecl, CXXSpecialMemberKind CSM, unsigned Quals,
+    bool ConstRHS, CXXConstructorDecl *InheritedCtor = nullptr,
+    Sema::InheritedConstructorInfo *Inherited = nullptr) {
   // Suppress duplicate constraint checking here, in case a constraint check
   // caused us to decide to do this.  Any truely recursive checks will get
   // caught during these checks anyway.
@@ -7461,16 +7466,16 @@ specialMemberIsConstexpr(Sema &S, CXXRecordDecl *ClassDecl,
   // If we're inheriting a constructor, see if we need to call it for this base
   // class.
   if (InheritedCtor) {
-    assert(CSM == Sema::CXXDefaultConstructor);
+    assert(CSM == CXXSpecialMemberKind::DefaultConstructor);
     auto BaseCtor =
         Inherited->findConstructorForBase(ClassDecl, InheritedCtor).first;
     if (BaseCtor)
       return BaseCtor->isConstexpr();
   }
 
-  if (CSM == Sema::CXXDefaultConstructor)
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor)
     return ClassDecl->hasConstexprDefaultConstructor();
-  if (CSM == Sema::CXXDestructor)
+  if (CSM == CXXSpecialMemberKind::Destructor)
     return ClassDecl->hasConstexprDestructor();
 
   Sema::SpecialMemberOverloadResult SMOR =
@@ -7485,8 +7490,8 @@ specialMemberIsConstexpr(Sema &S, CXXRecordDecl *ClassDecl,
 /// Determine whether the specified special member function would be constexpr
 /// if it were implicitly defined.
 static bool defaultedSpecialMemberIsConstexpr(
-    Sema &S, CXXRecordDecl *ClassDecl, Sema::CXXSpecialMember CSM,
-    bool ConstArg, CXXConstructorDecl *InheritedCtor = nullptr,
+    Sema &S, CXXRecordDecl *ClassDecl, CXXSpecialMemberKind CSM, bool ConstArg,
+    CXXConstructorDecl *InheritedCtor = nullptr,
     Sema::InheritedConstructorInfo *Inherited = nullptr) {
   if (!S.getLangOpts().CPlusPlus11)
     return false;
@@ -7495,7 +7500,7 @@ static bool defaultedSpecialMemberIsConstexpr(
   // In the definition of a constexpr constructor [...]
   bool Ctor = true;
   switch (CSM) {
-  case Sema::CXXDefaultConstructor:
+  case CXXSpecialMemberKind::DefaultConstructor:
     if (Inherited)
       break;
     // Since default constructor lookup is essentially trivial (and cannot
@@ -7506,23 +7511,23 @@ static bool defaultedSpecialMemberIsConstexpr(
     // constructor is constexpr to determine whether the type is a literal type.
     return ClassDecl->defaultedDefaultConstructorIsConstexpr();
 
-  case Sema::CXXCopyConstructor:
-  case Sema::CXXMoveConstructor:
+  case CXXSpecialMemberKind::CopyConstructor:
+  case CXXSpecialMemberKind::MoveConstructor:
     // For copy or move constructors, we need to perform overload resolution.
     break;
 
-  case Sema::CXXCopyAssignment:
-  case Sema::CXXMoveAssignment:
+  case CXXSpecialMemberKind::CopyAssignment:
+  case CXXSpecialMemberKind::MoveAssignment:
     if (!S.getLangOpts().CPlusPlus14)
       return false;
     // In C++1y, we need to perform overload resolution.
     Ctor = false;
     break;
 
-  case Sema::CXXDestructor:
+  case CXXSpecialMemberKind::Destructor:
     return ClassDecl->defaultedDestructorIsConstexpr();
 
-  case Sema::CXXInvalid:
+  case CXXSpecialMemberKind::Invalid:
     return false;
   }
 
@@ -7534,7 +7539,7 @@ static bool defaultedSpecialMemberIsConstexpr(
   // will be initialized (if the constructor isn't deleted), we just don't know
   // which one.
   if (Ctor && ClassDecl->isUnion())
-    return CSM == Sema::CXXDefaultConstructor
+    return CSM == CXXSpecialMemberKind::DefaultConstructor
                ? ClassDecl->hasInClassInitializer() ||
                      !ClassDecl->hasVariantMembers()
                : true;
@@ -7575,7 +7580,8 @@ static bool defaultedSpecialMemberIsConstexpr(
     for (const auto *F : ClassDecl->fields()) {
       if (F->isInvalidDecl())
         continue;
-      if (CSM == Sema::CXXDefaultConstructor && F->hasInClassInitializer())
+      if (CSM == CXXSpecialMemberKind::DefaultConstructor &&
+          F->hasInClassInitializer())
         continue;
       QualType BaseType = S.Context.getBaseElementType(F->getType());
       if (const RecordType *RecordTy = BaseType->getAs<RecordType>()) {
@@ -7584,7 +7590,7 @@ static bool defaultedSpecialMemberIsConstexpr(
                                       BaseType.getCVRQualifiers(),
                                       ConstArg && !F->isMutable()))
           return false;
-      } else if (CSM == Sema::CXXDefaultConstructor) {
+      } else if (CSM == CXXSpecialMemberKind::DefaultConstructor) {
         return false;
       }
     }
@@ -7615,9 +7621,10 @@ struct ComputingExceptionSpec {
 }
 
 static Sema::ImplicitExceptionSpecification
-ComputeDefaultedSpecialMemberExceptionSpec(
-    Sema &S, SourceLocation Loc, CXXMethodDecl *MD, Sema::CXXSpecialMember CSM,
-    Sema::InheritedConstructorInfo *ICI);
+ComputeDefaultedSpecialMemberExceptionSpec(Sema &S, SourceLocation Loc,
+                                           CXXMethodDecl *MD,
+                                           CXXSpecialMemberKind CSM,
+                                           Sema::InheritedConstructorInfo *ICI);
 
 static Sema::ImplicitExceptionSpecification
 ComputeDefaultedComparisonExceptionSpec(Sema &S, SourceLocation Loc,
@@ -7641,7 +7648,7 @@ computeImplicitExceptionSpec(Sema &S, SourceLocation Loc, FunctionDecl *FD) {
   Sema::InheritedConstructorInfo ICI(
       S, Loc, CD->getInheritedConstructor().getShadowDecl());
   return ComputeDefaultedSpecialMemberExceptionSpec(
-      S, Loc, CD, Sema::CXXDefaultConstructor, &ICI);
+      S, Loc, CD, CXXSpecialMemberKind::DefaultConstructor, &ICI);
 }
 
 static FunctionProtoType::ExtProtoInfo getImplicitMethodEPI(Sema &S,
@@ -7693,11 +7700,11 @@ void Sema::CheckExplicitlyDefaultedFunction(Scope *S, FunctionDecl *FD) {
 }
 
 bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
-                                                 CXXSpecialMember CSM,
+                                                 CXXSpecialMemberKind CSM,
                                                  SourceLocation DefaultLoc) {
   CXXRecordDecl *RD = MD->getParent();
 
-  assert(MD->isExplicitlyDefaulted() && CSM != CXXInvalid &&
+  assert(MD->isExplicitlyDefaulted() && CSM != CXXSpecialMemberKind::Invalid &&
          "not an explicitly-defaulted special member");
 
   // Defer all checking for special members of a dependent type.
@@ -7723,21 +7730,22 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
   bool DeleteOnTypeMismatch = getLangOpts().CPlusPlus20 && First;
   bool ShouldDeleteForTypeMismatch = false;
   unsigned ExpectedParams = 1;
-  if (CSM == CXXDefaultConstructor || CSM == CXXDestructor)
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor ||
+      CSM == CXXSpecialMemberKind::Destructor)
     ExpectedParams = 0;
   if (MD->getNumExplicitParams() != ExpectedParams) {
     // This checks for default arguments: a copy or move constructor with a
     // default argument is classified as a default constructor, and assignment
     // operations and destructors can't have default arguments.
     Diag(MD->getLocation(), diag::err_defaulted_special_member_params)
-      << CSM << MD->getSourceRange();
+        << llvm::to_underlying(CSM) << MD->getSourceRange();
     HadError = true;
   } else if (MD->isVariadic()) {
     if (DeleteOnTypeMismatch)
       ShouldDeleteForTypeMismatch = true;
     else {
       Diag(MD->getLocation(), diag::err_defaulted_special_member_variadic)
-        << CSM << MD->getSourceRange();
+          << llvm::to_underlying(CSM) << MD->getSourceRange();
       HadError = true;
     }
   }
@@ -7745,13 +7753,14 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
   const FunctionProtoType *Type = MD->getType()->castAs<FunctionProtoType>();
 
   bool CanHaveConstParam = false;
-  if (CSM == CXXCopyConstructor)
+  if (CSM == CXXSpecialMemberKind::CopyConstructor)
     CanHaveConstParam = RD->implicitCopyConstructorHasConstParam();
-  else if (CSM == CXXCopyAssignment)
+  else if (CSM == CXXSpecialMemberKind::CopyAssignment)
     CanHaveConstParam = RD->implicitCopyAssignmentHasConstParam();
 
   QualType ReturnType = Context.VoidTy;
-  if (CSM == CXXCopyAssignment || CSM == CXXMoveAssignment) {
+  if (CSM == CXXSpecialMemberKind::CopyAssignment ||
+      CSM == CXXSpecialMemberKind::MoveAssignment) {
     // Check for return type matching.
     ReturnType = Type->getReturnType();
     QualType ThisType = MD->getFunctionObjectParameterType();
@@ -7765,7 +7774,8 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
 
     if (!Context.hasSameType(ReturnType, ExpectedReturnType)) {
       Diag(MD->getLocation(), diag::err_defaulted_special_member_return_type)
-        << (CSM == CXXMoveAssignment) << ExpectedReturnType;
+          << (CSM == CXXSpecialMemberKind::MoveAssignment)
+          << ExpectedReturnType;
       HadError = true;
     }
 
@@ -7775,7 +7785,8 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
         ShouldDeleteForTypeMismatch = true;
       else {
         Diag(MD->getLocation(), diag::err_defaulted_special_member_quals)
-          << (CSM == CXXMoveAssignment) << getLangOpts().CPlusPlus14;
+            << (CSM == CXXSpecialMemberKind::MoveAssignment)
+            << getLangOpts().CPlusPlus14;
         HadError = true;
       }
     }
@@ -7793,7 +7804,8 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
       else {
         Diag(MD->getLocation(),
              diag::err_defaulted_special_member_explicit_object_mismatch)
-            << (CSM == CXXMoveAssignment) << RD << MD->getSourceRange();
+            << (CSM == CXXSpecialMemberKind::MoveAssignment) << RD
+            << MD->getSourceRange();
         HadError = true;
       }
     }
@@ -7815,7 +7827,8 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
         ShouldDeleteForTypeMismatch = true;
       else {
         Diag(MD->getLocation(),
-             diag::err_defaulted_special_member_volatile_param) << CSM;
+             diag::err_defaulted_special_member_volatile_param)
+            << llvm::to_underlying(CSM);
         HadError = true;
       }
     }
@@ -7823,23 +7836,25 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
     if (HasConstParam && !CanHaveConstParam) {
       if (DeleteOnTypeMismatch)
         ShouldDeleteForTypeMismatch = true;
-      else if (CSM == CXXCopyConstructor || CSM == CXXCopyAssignment) {
+      else if (CSM == CXXSpecialMemberKind::CopyConstructor ||
+               CSM == CXXSpecialMemberKind::CopyAssignment) {
         Diag(MD->getLocation(),
              diag::err_defaulted_special_member_copy_const_param)
-          << (CSM == CXXCopyAssignment);
+            << (CSM == CXXSpecialMemberKind::CopyAssignment);
         // FIXME: Explain why this special member can't be const.
         HadError = true;
       } else {
         Diag(MD->getLocation(),
              diag::err_defaulted_special_member_move_const_param)
-          << (CSM == CXXMoveAssignment);
+            << (CSM == CXXSpecialMemberKind::MoveAssignment);
         HadError = true;
       }
     }
   } else if (ExpectedParams) {
     // A copy assignment operator can take its argument by value, but a
     // defaulted one cannot.
-    assert(CSM == CXXCopyAssignment && "unexpected non-ref argument");
+    assert(CSM == CXXSpecialMemberKind::CopyAssignment &&
+           "unexpected non-ref argument");
     Diag(MD->getLocation(), diag::err_defaulted_copy_assign_not_ref);
     HadError = true;
   }
@@ -7874,12 +7889,12 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
         if (!MD->isConsteval() && RD->getNumVBases()) {
           Diag(MD->getBeginLoc(),
                diag::err_incorrect_defaulted_constexpr_with_vb)
-              << CSM;
+              << llvm::to_underlying(CSM);
           for (const auto &I : RD->vbases())
             Diag(I.getBeginLoc(), diag::note_constexpr_virtual_base_here);
         } else {
           Diag(MD->getBeginLoc(), diag::err_incorrect_defaulted_constexpr)
-              << CSM << MD->isConsteval();
+              << llvm::to_underlying(CSM) << MD->isConsteval();
         }
         HadError = true;
         // FIXME: Explain why the special member can't be constexpr.
@@ -7912,9 +7927,11 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
     if (First) {
       SetDeclDeleted(MD, MD->getLocation());
       if (!inTemplateInstantiation() && !HadError) {
-        Diag(MD->getLocation(), diag::warn_defaulted_method_deleted) << CSM;
+        Diag(MD->getLocation(), diag::warn_defaulted_method_deleted)
+            << llvm::to_underlying(CSM);
         if (ShouldDeleteForTypeMismatch) {
-          Diag(MD->getLocation(), diag::note_deleted_type_mismatch) << CSM;
+          Diag(MD->getLocation(), diag::note_deleted_type_mismatch)
+              << llvm::to_underlying(CSM);
         } else if (ShouldDeleteSpecialMember(MD, CSM, nullptr,
                                              /*Diagnose*/ true) &&
                    DefaultLoc.isValid()) {
@@ -7924,13 +7941,15 @@ bool Sema::CheckExplicitlyDefaultedSpecialMember(CXXMethodDecl *MD,
       }
       if (ShouldDeleteForTypeMismatch && !HadError) {
         Diag(MD->getLocation(),
-             diag::warn_cxx17_compat_defaulted_method_type_mismatch) << CSM;
+             diag::warn_cxx17_compat_defaulted_method_type_mismatch)
+            << llvm::to_underlying(CSM);
       }
     } else {
       // C++11 [dcl.fct.def.default]p4:
       //   [For a] user-provided explicitly-defaulted function [...] if such a
       //   function is implicitly defined as deleted, the program is ill-formed.
-      Diag(MD->getLocation(), diag::err_out_of_line_default_deletes) << CSM;
+      Diag(MD->getLocation(), diag::err_out_of_line_default_deletes)
+          << llvm::to_underlying(CSM);
       assert(!ShouldDeleteForTypeMismatch && "deleted non-first decl");
       ShouldDeleteSpecialMember(MD, CSM, nullptr, /*Diagnose*/true);
       HadError = true;
@@ -9271,28 +9290,28 @@ template<typename Derived>
 struct SpecialMemberVisitor {
   Sema &S;
   CXXMethodDecl *MD;
-  Sema::CXXSpecialMember CSM;
+  CXXSpecialMemberKind CSM;
   Sema::InheritedConstructorInfo *ICI;
 
   // Properties of the special member, computed for convenience.
   bool IsConstructor = false, IsAssignment = false, ConstArg = false;
 
-  SpecialMemberVisitor(Sema &S, CXXMethodDecl *MD, Sema::CXXSpecialMember CSM,
+  SpecialMemberVisitor(Sema &S, CXXMethodDecl *MD, CXXSpecialMemberKind CSM,
                        Sema::InheritedConstructorInfo *ICI)
       : S(S), MD(MD), CSM(CSM), ICI(ICI) {
     switch (CSM) {
-    case Sema::CXXDefaultConstructor:
-    case Sema::CXXCopyConstructor:
-    case Sema::CXXMoveConstructor:
+    case CXXSpecialMemberKind::DefaultConstructor:
+    case CXXSpecialMemberKind::CopyConstructor:
+    case CXXSpecialMemberKind::MoveConstructor:
       IsConstructor = true;
       break;
-    case Sema::CXXCopyAssignment:
-    case Sema::CXXMoveAssignment:
+    case CXXSpecialMemberKind::CopyAssignment:
+    case CXXSpecialMemberKind::MoveAssignment:
       IsAssignment = true;
       break;
-    case Sema::CXXDestructor:
+    case CXXSpecialMemberKind::Destructor:
       break;
-    case Sema::CXXInvalid:
+    case CXXSpecialMemberKind::Invalid:
       llvm_unreachable("invalid special member kind");
     }
 
@@ -9307,7 +9326,8 @@ struct SpecialMemberVisitor {
 
   /// Is this a "move" special member?
   bool isMove() const {
-    return CSM == Sema::CXXMoveConstructor || CSM == Sema::CXXMoveAssignment;
+    return CSM == CXXSpecialMemberKind::MoveConstructor ||
+           CSM == CXXSpecialMemberKind::MoveAssignment;
   }
 
   /// Look up the corresponding special member in the given class.
@@ -9322,7 +9342,7 @@ struct SpecialMemberVisitor {
   Sema::SpecialMemberOverloadResult lookupInheritedCtor(CXXRecordDecl *Class) {
     if (!ICI)
       return {};
-    assert(CSM == Sema::CXXDefaultConstructor);
+    assert(CSM == CXXSpecialMemberKind::DefaultConstructor);
     auto *BaseCtor =
       cast<CXXConstructorDecl>(MD)->getInheritedConstructor().getConstructor();
     if (auto *MD = ICI->findConstructorForBase(Class, BaseCtor).first)
@@ -9392,15 +9412,15 @@ struct SpecialMemberDeletionInfo
   bool AllFieldsAreConst;
 
   SpecialMemberDeletionInfo(Sema &S, CXXMethodDecl *MD,
-                            Sema::CXXSpecialMember CSM,
+                            CXXSpecialMemberKind CSM,
                             Sema::InheritedConstructorInfo *ICI, bool Diagnose)
       : SpecialMemberVisitor(S, MD, CSM, ICI), Diagnose(Diagnose),
         Loc(MD->getLocation()), AllFieldsAreConst(true) {}
 
   bool inUnion() const { return MD->getParent()->isUnion(); }
 
-  Sema::CXXSpecialMember getEffectiveCSM() {
-    return ICI ? Sema::CXXInvalid : CSM;
+  CXXSpecialMemberKind getEffectiveCSM() {
+    return ICI ? CXXSpecialMemberKind::Invalid : CSM;
   }
 
   bool shouldDeleteForVariantObjCPtrMember(FieldDecl *FD, QualType FieldType);
@@ -9466,7 +9486,7 @@ bool SpecialMemberDeletionInfo::shouldDeleteForSubobjectCall(
     // must be accessible and non-deleted, but need not be trivial. Such a
     // destructor is never actually called, but is semantically checked as
     // if it were.
-    if (CSM == Sema::CXXDefaultConstructor) {
+    if (CSM == CXXSpecialMemberKind::DefaultConstructor) {
       // [class.default.ctor]p2:
       //   A defaulted default constructor for class X is defined as deleted if
       //   - X is a union that has a variant member with a non-trivial default
@@ -9487,15 +9507,16 @@ bool SpecialMemberDeletionInfo::shouldDeleteForSubobjectCall(
     if (Field) {
       S.Diag(Field->getLocation(),
              diag::note_deleted_special_member_class_subobject)
-        << getEffectiveCSM() << MD->getParent() << /*IsField*/true
-        << Field << DiagKind << IsDtorCallInCtor << /*IsObjCPtr*/false;
+          << llvm::to_underlying(getEffectiveCSM()) << MD->getParent()
+          << /*IsField*/ true << Field << DiagKind << IsDtorCallInCtor
+          << /*IsObjCPtr*/ false;
     } else {
       CXXBaseSpecifier *Base = Subobj.get<CXXBaseSpecifier*>();
       S.Diag(Base->getBeginLoc(),
              diag::note_deleted_special_member_class_subobject)
-          << getEffectiveCSM() << MD->getParent() << /*IsField*/ false
-          << Base->getType() << DiagKind << IsDtorCallInCtor
-          << /*IsObjCPtr*/false;
+          << llvm::to_underlying(getEffectiveCSM()) << MD->getParent()
+          << /*IsField*/ false << Base->getType() << DiagKind
+          << IsDtorCallInCtor << /*IsObjCPtr*/ false;
     }
 
     if (DiagKind == 1)
@@ -9527,8 +9548,8 @@ bool SpecialMemberDeletionInfo::shouldDeleteForClassSubobject(
   // C++11 [class.dtor]p5:
   // -- any direct or virtual base class [...] has a type with a destructor
   //    that is deleted or inaccessible
-  if (!(CSM == Sema::CXXDefaultConstructor &&
-        Field && Field->hasInClassInitializer()) &&
+  if (!(CSM == CXXSpecialMemberKind::DefaultConstructor && Field &&
+        Field->hasInClassInitializer()) &&
       shouldDeleteForSubobjectCall(Subobj, lookupIn(Class, Quals, IsMutable),
                                    false))
     return true;
@@ -9538,8 +9559,8 @@ bool SpecialMemberDeletionInfo::shouldDeleteForClassSubobject(
   //    type with a destructor that is deleted or inaccessible
   if (IsConstructor) {
     Sema::SpecialMemberOverloadResult SMOR =
-        S.LookupSpecialMember(Class, Sema::CXXDestructor,
-                              false, false, false, false, false);
+        S.LookupSpecialMember(Class, CXXSpecialMemberKind::Destructor, false,
+                              false, false, false, false);
     if (shouldDeleteForSubobjectCall(Subobj, SMOR, true))
       return true;
   }
@@ -9557,15 +9578,16 @@ bool SpecialMemberDeletionInfo::shouldDeleteForVariantObjCPtrMember(
 
   // Don't make the defaulted default constructor defined as deleted if the
   // member has an in-class initializer.
-  if (CSM == Sema::CXXDefaultConstructor && FD->hasInClassInitializer())
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor &&
+      FD->hasInClassInitializer())
     return false;
 
   if (Diagnose) {
     auto *ParentClass = cast<CXXRecordDecl>(FD->getParent());
-    S.Diag(FD->getLocation(),
-           diag::note_deleted_special_member_class_subobject)
-        << getEffectiveCSM() << ParentClass << /*IsField*/true
-        << FD << 4 << /*IsDtorCallInCtor*/false << /*IsObjCPtr*/true;
+    S.Diag(FD->getLocation(), diag::note_deleted_special_member_class_subobject)
+        << llvm::to_underlying(getEffectiveCSM()) << ParentClass
+        << /*IsField*/ true << FD << 4 << /*IsDtorCallInCtor*/ false
+        << /*IsObjCPtr*/ true;
   }
 
   return true;
@@ -9590,9 +9612,9 @@ bool SpecialMemberDeletionInfo::shouldDeleteForBase(CXXBaseSpecifier *Base) {
     if (BaseCtor->isDeleted() && Diagnose) {
       S.Diag(Base->getBeginLoc(),
              diag::note_deleted_special_member_class_subobject)
-          << getEffectiveCSM() << MD->getParent() << /*IsField*/ false
-          << Base->getType() << /*Deleted*/ 1 << /*IsDtorCallInCtor*/ false
-          << /*IsObjCPtr*/false;
+          << llvm::to_underlying(getEffectiveCSM()) << MD->getParent()
+          << /*IsField*/ false << Base->getType() << /*Deleted*/ 1
+          << /*IsDtorCallInCtor*/ false << /*IsObjCPtr*/ false;
       S.NoteDeletedFunction(BaseCtor);
     }
     return BaseCtor->isDeleted();
@@ -9609,7 +9631,7 @@ bool SpecialMemberDeletionInfo::shouldDeleteForField(FieldDecl *FD) {
   if (inUnion() && shouldDeleteForVariantObjCPtrMember(FD, FieldType))
     return true;
 
-  if (CSM == Sema::CXXDefaultConstructor) {
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor) {
     // For a default constructor, all references must be initialized in-class
     // and, if a union, it must have a non-const member.
     if (FieldType->isReferenceType() && !FD->hasInClassInitializer()) {
@@ -9632,7 +9654,7 @@ bool SpecialMemberDeletionInfo::shouldDeleteForField(FieldDecl *FD) {
 
     if (inUnion() && !FieldType.isConstQualified())
       AllFieldsAreConst = false;
-  } else if (CSM == Sema::CXXCopyConstructor) {
+  } else if (CSM == CXXSpecialMemberKind::CopyConstructor) {
     // For a copy constructor, data members must not be of rvalue reference
     // type.
     if (FieldType->isRValueReferenceType()) {
@@ -9683,8 +9705,8 @@ bool SpecialMemberDeletionInfo::shouldDeleteForField(FieldDecl *FD) {
       }
 
       // At least one member in each anonymous union must be non-const
-      if (CSM == Sema::CXXDefaultConstructor && AllVariantFieldsAreConst &&
-          !FieldRecord->field_empty()) {
+      if (CSM == CXXSpecialMemberKind::DefaultConstructor &&
+          AllVariantFieldsAreConst && !FieldRecord->field_empty()) {
         if (Diagnose)
           S.Diag(FieldRecord->getLocation(),
                  diag::note_deleted_default_ctor_all_const)
@@ -9712,7 +9734,8 @@ bool SpecialMemberDeletionInfo::shouldDeleteForField(FieldDecl *FD) {
 bool SpecialMemberDeletionInfo::shouldDeleteForAllConstMembers() {
   // This is a silly definition, because it gives an empty union a deleted
   // default constructor. Don't do that.
-  if (CSM == Sema::CXXDefaultConstructor && inUnion() && AllFieldsAreConst) {
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor && inUnion() &&
+      AllFieldsAreConst) {
     bool AnyFields = false;
     for (auto *F : MD->getParent()->fields())
       if ((AnyFields = !F->isUnnamedBitfield()))
@@ -9731,7 +9754,8 @@ bool SpecialMemberDeletionInfo::shouldDeleteForAllConstMembers() {
 /// Determine whether a defaulted special member function should be defined as
 /// deleted, as specified in C++11 [class.ctor]p5, C++11 [class.copy]p11,
 /// C++11 [class.copy]p23, and C++11 [class.dtor]p5.
-bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
+bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD,
+                                     CXXSpecialMemberKind CSM,
                                      InheritedConstructorInfo *ICI,
                                      bool Diagnose) {
   if (MD->isInvalidDecl())
@@ -9748,7 +9772,8 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
   //   assignment operator.
   // C++2a adds back these operators if the lambda has no lambda-capture.
   if (RD->isLambda() && !RD->lambdaIsDefaultConstructibleAndAssignable() &&
-      (CSM == CXXDefaultConstructor || CSM == CXXCopyAssignment)) {
+      (CSM == CXXSpecialMemberKind::DefaultConstructor ||
+       CSM == CXXSpecialMemberKind::CopyAssignment)) {
     if (Diagnose)
       Diag(RD->getLocation(), diag::note_lambda_decl);
     return true;
@@ -9757,16 +9782,16 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
   // For an anonymous struct or union, the copy and assignment special members
   // will never be used, so skip the check. For an anonymous union declared at
   // namespace scope, the constructor and destructor are used.
-  if (CSM != CXXDefaultConstructor && CSM != CXXDestructor &&
-      RD->isAnonymousStructOrUnion())
+  if (CSM != CXXSpecialMemberKind::DefaultConstructor &&
+      CSM != CXXSpecialMemberKind::Destructor && RD->isAnonymousStructOrUnion())
     return false;
 
   // C++11 [class.copy]p7, p18:
   //   If the class definition declares a move constructor or move assignment
   //   operator, an implicitly declared copy constructor or copy assignment
   //   operator is defined as deleted.
-  if (MD->isImplicit() &&
-      (CSM == CXXCopyConstructor || CSM == CXXCopyAssignment)) {
+  if (MD->isImplicit() && (CSM == CXXSpecialMemberKind::CopyConstructor ||
+                           CSM == CXXSpecialMemberKind::CopyAssignment)) {
     CXXMethodDecl *UserDeclaredMove = nullptr;
 
     // In Microsoft mode up to MSVC 2013, a user-declared move only causes the
@@ -9777,7 +9802,8 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
         !getLangOpts().isCompatibleWithMSVC(LangOptions::MSVC2015);
 
     if (RD->hasUserDeclaredMoveConstructor() &&
-        (!DeletesOnlyMatchingCopy || CSM == CXXCopyConstructor)) {
+        (!DeletesOnlyMatchingCopy ||
+         CSM == CXXSpecialMemberKind::CopyConstructor)) {
       if (!Diagnose) return true;
 
       // Find any user-declared move constructor.
@@ -9789,7 +9815,8 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
       }
       assert(UserDeclaredMove);
     } else if (RD->hasUserDeclaredMoveAssignment() &&
-               (!DeletesOnlyMatchingCopy || CSM == CXXCopyAssignment)) {
+               (!DeletesOnlyMatchingCopy ||
+                CSM == CXXSpecialMemberKind::CopyAssignment)) {
       if (!Diagnose) return true;
 
       // Find any user-declared move assignment operator.
@@ -9805,8 +9832,8 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
     if (UserDeclaredMove) {
       Diag(UserDeclaredMove->getLocation(),
            diag::note_deleted_copy_user_declared_move)
-        << (CSM == CXXCopyAssignment) << RD
-        << UserDeclaredMove->isMoveAssignmentOperator();
+          << (CSM == CXXSpecialMemberKind::CopyAssignment) << RD
+          << UserDeclaredMove->isMoveAssignmentOperator();
       return true;
     }
   }
@@ -9817,7 +9844,7 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD, CXXSpecialMember CSM,
   // C++11 [class.dtor]p5:
   // -- for a virtual destructor, lookup of the non-array deallocation function
   //    results in an ambiguity or in a function that is deleted or inaccessible
-  if (CSM == CXXDestructor && MD->isVirtual()) {
+  if (CSM == CXXSpecialMemberKind::Destructor && MD->isVirtual()) {
     FunctionDecl *OperatorDelete = nullptr;
     DeclarationName Name =
       Context.DeclarationNames.getCXXOperatorName(OO_Delete);
@@ -9891,7 +9918,7 @@ void Sema::DiagnoseDeletedDefaultedFunction(FunctionDecl *FD) {
 /// If \p ForCall is true, look at CXXRecord::HasTrivialSpecialMembersForCall to
 /// determine whether the special member is trivial.
 static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
-                                     Sema::CXXSpecialMember CSM, unsigned Quals,
+                                     CXXSpecialMemberKind CSM, unsigned Quals,
                                      bool ConstRHS,
                                      Sema::TrivialABIHandling TAH,
                                      CXXMethodDecl **Selected) {
@@ -9899,10 +9926,10 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
     *Selected = nullptr;
 
   switch (CSM) {
-  case Sema::CXXInvalid:
+  case CXXSpecialMemberKind::Invalid:
     llvm_unreachable("not a special member");
 
-  case Sema::CXXDefaultConstructor:
+  case CXXSpecialMemberKind::DefaultConstructor:
     // C++11 [class.ctor]p5:
     //   A default constructor is trivial if:
     //    - all the [direct subobjects] have trivial default constructors
@@ -9931,7 +9958,7 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
 
     return false;
 
-  case Sema::CXXDestructor:
+  case CXXSpecialMemberKind::Destructor:
     // C++11 [class.dtor]p5:
     //   A destructor is trivial if:
     //    - all the direct [subobjects] have trivial destructors
@@ -9948,7 +9975,7 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
 
     return false;
 
-  case Sema::CXXCopyConstructor:
+  case CXXSpecialMemberKind::CopyConstructor:
     // C++11 [class.copy]p12:
     //   A copy constructor is trivial if:
     //    - the constructor selected to copy each direct [subobject] is trivial
@@ -9969,7 +9996,7 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
     //   struct B { mutable A a; };
     goto NeedOverloadResolution;
 
-  case Sema::CXXCopyAssignment:
+  case CXXSpecialMemberKind::CopyAssignment:
     // C++11 [class.copy]p25:
     //   A copy assignment operator is trivial if:
     //    - the assignment operator selected to copy each direct [subobject] is
@@ -9984,8 +10011,8 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
     // treat that as a language defect.
     goto NeedOverloadResolution;
 
-  case Sema::CXXMoveConstructor:
-  case Sema::CXXMoveAssignment:
+  case CXXSpecialMemberKind::MoveConstructor:
+  case CXXSpecialMemberKind::MoveAssignment:
   NeedOverloadResolution:
     Sema::SpecialMemberOverloadResult SMOR =
         lookupCallFromSpecialMember(S, RD, CSM, Quals, ConstRHS);
@@ -10009,7 +10036,8 @@ static bool findTrivialSpecialMember(Sema &S, CXXRecordDecl *RD,
       *Selected = SMOR.getMethod();
 
     if (TAH == Sema::TAH_ConsiderTrivialABI &&
-        (CSM == Sema::CXXCopyConstructor || CSM == Sema::CXXMoveConstructor))
+        (CSM == CXXSpecialMemberKind::CopyConstructor ||
+         CSM == CXXSpecialMemberKind::MoveConstructor))
       return SMOR.getMethod()->isTrivialForCall();
     return SMOR.getMethod()->isTrivial();
   }
@@ -10047,9 +10075,10 @@ enum TrivialSubobjectKind {
 /// Check whether the special member selected for a given type would be trivial.
 static bool checkTrivialSubobjectCall(Sema &S, SourceLocation SubobjLoc,
                                       QualType SubType, bool ConstRHS,
-                                      Sema::CXXSpecialMember CSM,
+                                      CXXSpecialMemberKind CSM,
                                       TrivialSubobjectKind Kind,
-                                      Sema::TrivialABIHandling TAH, bool Diagnose) {
+                                      Sema::TrivialABIHandling TAH,
+                                      bool Diagnose) {
   CXXRecordDecl *SubRD = SubType->getAsCXXRecordDecl();
   if (!SubRD)
     return true;
@@ -10063,27 +10092,28 @@ static bool checkTrivialSubobjectCall(Sema &S, SourceLocation SubobjLoc,
     if (ConstRHS)
       SubType.addConst();
 
-    if (!Selected && CSM == Sema::CXXDefaultConstructor) {
+    if (!Selected && CSM == CXXSpecialMemberKind::DefaultConstructor) {
       S.Diag(SubobjLoc, diag::note_nontrivial_no_def_ctor)
         << Kind << SubType.getUnqualifiedType();
       if (CXXConstructorDecl *CD = findUserDeclaredCtor(SubRD))
         S.Diag(CD->getLocation(), diag::note_user_declared_ctor);
     } else if (!Selected)
       S.Diag(SubobjLoc, diag::note_nontrivial_no_copy)
-        << Kind << SubType.getUnqualifiedType() << CSM << SubType;
+          << Kind << SubType.getUnqualifiedType() << llvm::to_underlying(CSM)
+          << SubType;
     else if (Selected->isUserProvided()) {
       if (Kind == TSK_CompleteObject)
         S.Diag(Selected->getLocation(), diag::note_nontrivial_user_provided)
-          << Kind << SubType.getUnqualifiedType() << CSM;
+            << Kind << SubType.getUnqualifiedType() << llvm::to_underlying(CSM);
       else {
         S.Diag(SubobjLoc, diag::note_nontrivial_user_provided)
-          << Kind << SubType.getUnqualifiedType() << CSM;
+            << Kind << SubType.getUnqualifiedType() << llvm::to_underlying(CSM);
         S.Diag(Selected->getLocation(), diag::note_declared_at);
       }
     } else {
       if (Kind != TSK_CompleteObject)
         S.Diag(SubobjLoc, diag::note_nontrivial_subobject)
-          << Kind << SubType.getUnqualifiedType() << CSM;
+            << Kind << SubType.getUnqualifiedType() << llvm::to_underlying(CSM);
 
       // Explain why the defaulted or deleted special member isn't trivial.
       S.SpecialMemberIsTrivial(Selected, CSM, Sema::TAH_IgnoreTrivialABI,
@@ -10097,8 +10127,7 @@ static bool checkTrivialSubobjectCall(Sema &S, SourceLocation SubobjLoc,
 /// Check whether the members of a class type allow a special member to be
 /// trivial.
 static bool checkTrivialClassMembers(Sema &S, CXXRecordDecl *RD,
-                                     Sema::CXXSpecialMember CSM,
-                                     bool ConstArg,
+                                     CXXSpecialMemberKind CSM, bool ConstArg,
                                      Sema::TrivialABIHandling TAH,
                                      bool Diagnose) {
   for (const auto *FI : RD->fields()) {
@@ -10119,7 +10148,8 @@ static bool checkTrivialClassMembers(Sema &S, CXXRecordDecl *RD,
     //   A default constructor is trivial if [...]
     //    -- no non-static data member of its class has a
     //       brace-or-equal-initializer
-    if (CSM == Sema::CXXDefaultConstructor && FI->hasInClassInitializer()) {
+    if (CSM == CXXSpecialMemberKind::DefaultConstructor &&
+        FI->hasInClassInitializer()) {
       if (Diagnose)
         S.Diag(FI->getLocation(), diag::note_nontrivial_default_member_init)
             << FI;
@@ -10148,10 +10178,12 @@ static bool checkTrivialClassMembers(Sema &S, CXXRecordDecl *RD,
 
 /// Diagnose why the specified class does not have a trivial special member of
 /// the given kind.
-void Sema::DiagnoseNontrivial(const CXXRecordDecl *RD, CXXSpecialMember CSM) {
+void Sema::DiagnoseNontrivial(const CXXRecordDecl *RD,
+                              CXXSpecialMemberKind CSM) {
   QualType Ty = Context.getRecordType(RD);
 
-  bool ConstArg = (CSM == CXXCopyConstructor || CSM == CXXCopyAssignment);
+  bool ConstArg = (CSM == CXXSpecialMemberKind::CopyConstructor ||
+                   CSM == CXXSpecialMemberKind::CopyAssignment);
   checkTrivialSubobjectCall(*this, RD->getLocation(), Ty, ConstArg, CSM,
                             TSK_CompleteObject, TAH_IgnoreTrivialABI,
                             /*Diagnose*/true);
@@ -10160,9 +10192,10 @@ void Sema::DiagnoseNontrivial(const CXXRecordDecl *RD, CXXSpecialMember CSM) {
 /// Determine whether a defaulted or deleted special member function is trivial,
 /// as specified in C++11 [class.ctor]p5, C++11 [class.copy]p12,
 /// C++11 [class.copy]p25, and C++11 [class.dtor]p5.
-bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
+bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMemberKind CSM,
                                   TrivialABIHandling TAH, bool Diagnose) {
-  assert(!MD->isUserProvided() && CSM != CXXInvalid && "not special enough");
+  assert(!MD->isUserProvided() && CSM != CXXSpecialMemberKind::Invalid &&
+         "not special enough");
 
   CXXRecordDecl *RD = MD->getParent();
 
@@ -10172,13 +10205,13 @@ bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
   //   A [special member] is trivial if [...] its parameter-type-list is
   //   equivalent to the parameter-type-list of an implicit declaration [...]
   switch (CSM) {
-  case CXXDefaultConstructor:
-  case CXXDestructor:
+  case CXXSpecialMemberKind::DefaultConstructor:
+  case CXXSpecialMemberKind::Destructor:
     // Trivial default constructors and destructors cannot have parameters.
     break;
 
-  case CXXCopyConstructor:
-  case CXXCopyAssignment: {
+  case CXXSpecialMemberKind::CopyConstructor:
+  case CXXSpecialMemberKind::CopyAssignment: {
     const ParmVarDecl *Param0 = MD->getNonObjectParameter(0);
     const ReferenceType *RT = Param0->getType()->getAs<ReferenceType>();
 
@@ -10207,8 +10240,8 @@ bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
     break;
   }
 
-  case CXXMoveConstructor:
-  case CXXMoveAssignment: {
+  case CXXSpecialMemberKind::MoveConstructor:
+  case CXXSpecialMemberKind::MoveAssignment: {
     // Trivial move operations always have non-cv-qualified parameters.
     const ParmVarDecl *Param0 = MD->getNonObjectParameter(0);
     const RValueReferenceType *RT =
@@ -10223,7 +10256,7 @@ bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
     break;
   }
 
-  case CXXInvalid:
+  case CXXSpecialMemberKind::Invalid:
     llvm_unreachable("not a special member");
   }
 
@@ -10272,7 +10305,7 @@ bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
   // C++11 [class.dtor]p5:
   //   A destructor is trivial if [...]
   //    -- the destructor is not virtual
-  if (CSM == CXXDestructor && MD->isVirtual()) {
+  if (CSM == CXXSpecialMemberKind::Destructor && MD->isVirtual()) {
     if (Diagnose)
       Diag(MD->getLocation(), diag::note_nontrivial_virtual_dtor) << RD;
     return false;
@@ -10281,7 +10314,8 @@ bool Sema::SpecialMemberIsTrivial(CXXMethodDecl *MD, CXXSpecialMember CSM,
   // C++11 [class.ctor]p5, C++11 [class.copy]p12, C++11 [class.copy]p25:
   //   A [special member] for class X is trivial if [...]
   //    -- class X has no virtual functions and no virtual base classes
-  if (CSM != CXXDestructor && MD->getParent()->isDynamicClass()) {
+  if (CSM != CXXSpecialMemberKind::Destructor &&
+      MD->getParent()->isDynamicClass()) {
     if (!Diagnose)
       return false;
 
@@ -13760,7 +13794,7 @@ struct SpecialMemberExceptionSpecInfo
   Sema::ImplicitExceptionSpecification ExceptSpec;
 
   SpecialMemberExceptionSpecInfo(Sema &S, CXXMethodDecl *MD,
-                                 Sema::CXXSpecialMember CSM,
+                                 CXXSpecialMemberKind CSM,
                                  Sema::InheritedConstructorInfo *ICI,
                                  SourceLocation Loc)
       : SpecialMemberVisitor(S, MD, CSM, ICI), Loc(Loc), ExceptSpec(S) {}
@@ -13793,7 +13827,8 @@ bool SpecialMemberExceptionSpecInfo::visitBase(CXXBaseSpecifier *Base) {
 }
 
 bool SpecialMemberExceptionSpecInfo::visitField(FieldDecl *FD) {
-  if (CSM == Sema::CXXDefaultConstructor && FD->hasInClassInitializer()) {
+  if (CSM == CXXSpecialMemberKind::DefaultConstructor &&
+      FD->hasInClassInitializer()) {
     Expr *E = FD->getInClassInitializer();
     if (!E)
       // FIXME: It's a little wasteful to build and throw away a
@@ -13852,7 +13887,7 @@ ExplicitSpecifier Sema::ActOnExplicitBoolSpecifier(Expr *ExplicitExpr) {
 
 static Sema::ImplicitExceptionSpecification
 ComputeDefaultedSpecialMemberExceptionSpec(
-    Sema &S, SourceLocation Loc, CXXMethodDecl *MD, Sema::CXXSpecialMember CSM,
+    Sema &S, SourceLocation Loc, CXXMethodDecl *MD, CXXSpecialMemberKind CSM,
     Sema::InheritedConstructorInfo *ICI) {
   ComputingExceptionSpec CES(S, MD, Loc);
 
@@ -13902,7 +13937,7 @@ struct DeclaringSpecialMember {
   Sema::ContextRAII SavedContext;
   bool WasAlreadyBeingDeclared;
 
-  DeclaringSpecialMember(Sema &S, CXXRecordDecl *RD, Sema::CXXSpecialMember CSM)
+  DeclaringSpecialMember(Sema &S, CXXRecordDecl *RD, CXXSpecialMemberKind CSM)
       : S(S), D(RD, CSM), SavedContext(S, RD) {
     WasAlreadyBeingDeclared = !S.SpecialMembersBeingDeclared.insert(D).second;
     if (WasAlreadyBeingDeclared)
@@ -13992,13 +14027,13 @@ CXXConstructorDecl *Sema::DeclareImplicitDefaultConstructor(
   assert(ClassDecl->needsImplicitDefaultConstructor() &&
          "Should not build implicit default constructor!");
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXDefaultConstructor);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::DefaultConstructor);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXDefaultConstructor,
-                                                     false);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::DefaultConstructor, false);
 
   // Create the actual constructor declaration.
   CanQualType ClassType
@@ -14020,10 +14055,10 @@ CXXConstructorDecl *Sema::DeclareImplicitDefaultConstructor(
   setupImplicitSpecialMemberType(DefaultCon, Context.VoidTy, std::nullopt);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXDefaultConstructor,
-                                            DefaultCon,
-                                            /* ConstRHS */ false,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::DefaultConstructor, DefaultCon,
+        /* ConstRHS */ false,
+        /* Diagnose */ false);
 
   // We don't need to use SpecialMemberIsTrivial here; triviality for default
   // constructors is easy to compute.
@@ -14035,7 +14070,8 @@ CXXConstructorDecl *Sema::DeclareImplicitDefaultConstructor(
   Scope *S = getScopeForContext(ClassDecl);
   CheckImplicitSpecialMemberDeclaration(S, DefaultCon);
 
-  if (ShouldDeleteSpecialMember(DefaultCon, CXXDefaultConstructor))
+  if (ShouldDeleteSpecialMember(DefaultCon,
+                                CXXSpecialMemberKind::DefaultConstructor))
     SetDeclDeleted(DefaultCon, ClassLoc);
 
   if (S)
@@ -14127,10 +14163,10 @@ Sema::findInheritingConstructor(SourceLocation Loc,
   // from which it was inherited.
   InheritedConstructorInfo ICI(*this, Loc, Shadow);
 
-  bool Constexpr =
-      BaseCtor->isConstexpr() &&
-      defaultedSpecialMemberIsConstexpr(*this, Derived, CXXDefaultConstructor,
-                                        false, BaseCtor, &ICI);
+  bool Constexpr = BaseCtor->isConstexpr() &&
+                   defaultedSpecialMemberIsConstexpr(
+                       *this, Derived, CXXSpecialMemberKind::DefaultConstructor,
+                       false, BaseCtor, &ICI);
 
   CXXConstructorDecl *DerivedCtor = CXXConstructorDecl::Create(
       Context, Derived, UsingLoc, NameInfo, TInfo->getType(), TInfo,
@@ -14174,7 +14210,8 @@ Sema::findInheritingConstructor(SourceLocation Loc,
   DerivedCtor->setParams(ParamDecls);
   Derived->addDecl(DerivedCtor);
 
-  if (ShouldDeleteSpecialMember(DerivedCtor, CXXDefaultConstructor, &ICI))
+  if (ShouldDeleteSpecialMember(DerivedCtor,
+                                CXXSpecialMemberKind::DefaultConstructor, &ICI))
     SetDeclDeleted(DerivedCtor, UsingLoc);
 
   return DerivedCtor;
@@ -14183,8 +14220,9 @@ Sema::findInheritingConstructor(SourceLocation Loc,
 void Sema::NoteDeletedInheritingConstructor(CXXConstructorDecl *Ctor) {
   InheritedConstructorInfo ICI(*this, Ctor->getLocation(),
                                Ctor->getInheritedConstructor().getShadowDecl());
-  ShouldDeleteSpecialMember(Ctor, CXXDefaultConstructor, &ICI,
-                            /*Diagnose*/true);
+  ShouldDeleteSpecialMember(Ctor, CXXSpecialMemberKind::DefaultConstructor,
+                            &ICI,
+                            /*Diagnose*/ true);
 }
 
 void Sema::DefineInheritingConstructor(SourceLocation CurrentLocation,
@@ -14275,13 +14313,13 @@ CXXDestructorDecl *Sema::DeclareImplicitDestructor(CXXRecordDecl *ClassDecl) {
   //   inline public member of its class.
   assert(ClassDecl->needsImplicitDestructor());
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXDestructor);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::Destructor);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXDestructor,
-                                                     false);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::Destructor, false);
 
   // Create the actual destructor declaration.
   CanQualType ClassType
@@ -14303,10 +14341,10 @@ CXXDestructorDecl *Sema::DeclareImplicitDestructor(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(Destructor, Context.VoidTy, std::nullopt);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXDestructor,
-                                            Destructor,
-                                            /* ConstRHS */ false,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::Destructor, Destructor,
+        /* ConstRHS */ false,
+        /* Diagnose */ false);
 
   // We don't need to use SpecialMemberIsTrivial here; triviality for
   // destructors is easy to compute.
@@ -14324,7 +14362,7 @@ CXXDestructorDecl *Sema::DeclareImplicitDestructor(CXXRecordDecl *ClassDecl) {
   // the definition of the class, because its validity depends on the alignment
   // of the class. We'll check this from ActOnFields once the class is complete.
   if (ClassDecl->isCompleteDefinition() &&
-      ShouldDeleteSpecialMember(Destructor, CXXDestructor))
+      ShouldDeleteSpecialMember(Destructor, CXXSpecialMemberKind::Destructor))
     SetDeclDeleted(Destructor, ClassLoc);
 
   // Introduce this destructor into its scope.
@@ -14905,7 +14943,8 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
   // operators taking an object instead of a reference are allowed.
   assert(ClassDecl->needsImplicitCopyAssignment());
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXCopyAssignment);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::CopyAssignment);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
@@ -14922,9 +14961,8 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
 
   ArgType = Context.getLValueReferenceType(ArgType);
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXCopyAssignment,
-                                                     Const);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::CopyAssignment, Const);
 
   //   An implicitly-declared copy assignment operator is an inline public
   //   member of its class.
@@ -14945,10 +14983,10 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(CopyAssignment, RetType, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXCopyAssignment,
-                                            CopyAssignment,
-                                            /* ConstRHS */ Const,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::CopyAssignment, CopyAssignment,
+        /* ConstRHS */ Const,
+        /* Diagnose */ false);
 
   // Add the parameter to the operator.
   ParmVarDecl *FromParam = ParmVarDecl::Create(Context, CopyAssignment,
@@ -14959,9 +14997,10 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
   CopyAssignment->setParams(FromParam);
 
   CopyAssignment->setTrivial(
-    ClassDecl->needsOverloadResolutionForCopyAssignment()
-      ? SpecialMemberIsTrivial(CopyAssignment, CXXCopyAssignment)
-      : ClassDecl->hasTrivialCopyAssignment());
+      ClassDecl->needsOverloadResolutionForCopyAssignment()
+          ? SpecialMemberIsTrivial(CopyAssignment,
+                                   CXXSpecialMemberKind::CopyAssignment)
+          : ClassDecl->hasTrivialCopyAssignment());
 
   // Note that we have added this copy-assignment operator.
   ++getASTContext().NumImplicitCopyAssignmentOperatorsDeclared;
@@ -14969,7 +15008,8 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
   Scope *S = getScopeForContext(ClassDecl);
   CheckImplicitSpecialMemberDeclaration(S, CopyAssignment);
 
-  if (ShouldDeleteSpecialMember(CopyAssignment, CXXCopyAssignment)) {
+  if (ShouldDeleteSpecialMember(CopyAssignment,
+                                CXXSpecialMemberKind::CopyAssignment)) {
     ClassDecl->setImplicitCopyAssignmentIsDeleted();
     SetDeclDeleted(CopyAssignment, ClassLoc);
   }
@@ -15256,7 +15296,8 @@ void Sema::DefineImplicitCopyAssignment(SourceLocation CurrentLocation,
 CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   assert(ClassDecl->needsImplicitMoveAssignment());
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXMoveAssignment);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::MoveAssignment);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
@@ -15272,9 +15313,8 @@ CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   QualType RetType = Context.getLValueReferenceType(ArgType);
   ArgType = Context.getRValueReferenceType(ArgType);
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXMoveAssignment,
-                                                     false);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::MoveAssignment, false);
 
   //   An implicitly-declared move assignment operator is an inline public
   //   member of its class.
@@ -15295,10 +15335,10 @@ CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(MoveAssignment, RetType, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXMoveAssignment,
-                                            MoveAssignment,
-                                            /* ConstRHS */ false,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::MoveAssignment, MoveAssignment,
+        /* ConstRHS */ false,
+        /* Diagnose */ false);
 
   // Add the parameter to the operator.
   ParmVarDecl *FromParam = ParmVarDecl::Create(Context, MoveAssignment,
@@ -15309,9 +15349,10 @@ CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   MoveAssignment->setParams(FromParam);
 
   MoveAssignment->setTrivial(
-    ClassDecl->needsOverloadResolutionForMoveAssignment()
-      ? SpecialMemberIsTrivial(MoveAssignment, CXXMoveAssignment)
-      : ClassDecl->hasTrivialMoveAssignment());
+      ClassDecl->needsOverloadResolutionForMoveAssignment()
+          ? SpecialMemberIsTrivial(MoveAssignment,
+                                   CXXSpecialMemberKind::MoveAssignment)
+          : ClassDecl->hasTrivialMoveAssignment());
 
   // Note that we have added this copy-assignment operator.
   ++getASTContext().NumImplicitMoveAssignmentOperatorsDeclared;
@@ -15319,7 +15360,8 @@ CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   Scope *S = getScopeForContext(ClassDecl);
   CheckImplicitSpecialMemberDeclaration(S, MoveAssignment);
 
-  if (ShouldDeleteSpecialMember(MoveAssignment, CXXMoveAssignment)) {
+  if (ShouldDeleteSpecialMember(MoveAssignment,
+                                CXXSpecialMemberKind::MoveAssignment)) {
     ClassDecl->setImplicitMoveAssignmentIsDeleted();
     SetDeclDeleted(MoveAssignment, ClassLoc);
   }
@@ -15368,10 +15410,10 @@ static void checkMoveAssignmentForRepeatedMove(Sema &S, CXXRecordDecl *Class,
       // If we're not actually going to call a move assignment for this base,
       // or the selected move assignment is trivial, skip it.
       Sema::SpecialMemberOverloadResult SMOR =
-        S.LookupSpecialMember(Base, Sema::CXXMoveAssignment,
-                              /*ConstArg*/false, /*VolatileArg*/false,
-                              /*RValueThis*/true, /*ConstThis*/false,
-                              /*VolatileThis*/false);
+          S.LookupSpecialMember(Base, CXXSpecialMemberKind::MoveAssignment,
+                                /*ConstArg*/ false, /*VolatileArg*/ false,
+                                /*RValueThis*/ true, /*ConstThis*/ false,
+                                /*VolatileThis*/ false);
       if (!SMOR.getMethod() || SMOR.getMethod()->isTrivial() ||
           !SMOR.getMethod()->isMoveAssignmentOperator())
         continue;
@@ -15648,7 +15690,8 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
   //   constructor, one is declared implicitly.
   assert(ClassDecl->needsImplicitCopyConstructor());
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXCopyConstructor);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::CopyConstructor);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
@@ -15666,9 +15709,8 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
 
   ArgType = Context.getLValueReferenceType(ArgType);
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXCopyConstructor,
-                                                     Const);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::CopyConstructor, Const);
 
   DeclarationName Name
     = Context.DeclarationNames.getCXXConstructorName(
@@ -15691,10 +15733,10 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
   setupImplicitSpecialMemberType(CopyConstructor, Context.VoidTy, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXCopyConstructor,
-                                            CopyConstructor,
-                                            /* ConstRHS */ Const,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::CopyConstructor, CopyConstructor,
+        /* ConstRHS */ Const,
+        /* Diagnose */ false);
 
   // During template instantiation of special member functions we need a
   // reliable TypeSourceInfo for the parameter types in order to allow functions
@@ -15712,14 +15754,16 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
 
   CopyConstructor->setTrivial(
       ClassDecl->needsOverloadResolutionForCopyConstructor()
-          ? SpecialMemberIsTrivial(CopyConstructor, CXXCopyConstructor)
+          ? SpecialMemberIsTrivial(CopyConstructor,
+                                   CXXSpecialMemberKind::CopyConstructor)
           : ClassDecl->hasTrivialCopyConstructor());
 
   CopyConstructor->setTrivialForCall(
       ClassDecl->hasAttr<TrivialABIAttr>() ||
       (ClassDecl->needsOverloadResolutionForCopyConstructor()
-           ? SpecialMemberIsTrivial(CopyConstructor, CXXCopyConstructor,
-             TAH_ConsiderTrivialABI)
+           ? SpecialMemberIsTrivial(CopyConstructor,
+                                    CXXSpecialMemberKind::CopyConstructor,
+                                    TAH_ConsiderTrivialABI)
            : ClassDecl->hasTrivialCopyConstructorForCall()));
 
   // Note that we have declared this constructor.
@@ -15728,7 +15772,8 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
   Scope *S = getScopeForContext(ClassDecl);
   CheckImplicitSpecialMemberDeclaration(S, CopyConstructor);
 
-  if (ShouldDeleteSpecialMember(CopyConstructor, CXXCopyConstructor)) {
+  if (ShouldDeleteSpecialMember(CopyConstructor,
+                                CXXSpecialMemberKind::CopyConstructor)) {
     ClassDecl->setImplicitCopyConstructorIsDeleted();
     SetDeclDeleted(CopyConstructor, ClassLoc);
   }
@@ -15793,7 +15838,8 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
                                                     CXXRecordDecl *ClassDecl) {
   assert(ClassDecl->needsImplicitMoveConstructor());
 
-  DeclaringSpecialMember DSM(*this, ClassDecl, CXXMoveConstructor);
+  DeclaringSpecialMember DSM(*this, ClassDecl,
+                             CXXSpecialMemberKind::MoveConstructor);
   if (DSM.isAlreadyBeingDeclared())
     return nullptr;
 
@@ -15807,9 +15853,8 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
     ArgType = Context.getAddrSpaceQualType(ClassType, AS);
   ArgType = Context.getRValueReferenceType(ArgType);
 
-  bool Constexpr = defaultedSpecialMemberIsConstexpr(*this, ClassDecl,
-                                                     CXXMoveConstructor,
-                                                     false);
+  bool Constexpr = defaultedSpecialMemberIsConstexpr(
+      *this, ClassDecl, CXXSpecialMemberKind::MoveConstructor, false);
 
   DeclarationName Name
     = Context.DeclarationNames.getCXXConstructorName(
@@ -15833,10 +15878,10 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
   setupImplicitSpecialMemberType(MoveConstructor, Context.VoidTy, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(ClassDecl, CXXMoveConstructor,
-                                            MoveConstructor,
-                                            /* ConstRHS */ false,
-                                            /* Diagnose */ false);
+    inferCUDATargetForImplicitSpecialMember(
+        ClassDecl, CXXSpecialMemberKind::MoveConstructor, MoveConstructor,
+        /* ConstRHS */ false,
+        /* Diagnose */ false);
 
   // Add the parameter to the constructor.
   ParmVarDecl *FromParam = ParmVarDecl::Create(Context, MoveConstructor,
@@ -15848,13 +15893,15 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
 
   MoveConstructor->setTrivial(
       ClassDecl->needsOverloadResolutionForMoveConstructor()
-          ? SpecialMemberIsTrivial(MoveConstructor, CXXMoveConstructor)
+          ? SpecialMemberIsTrivial(MoveConstructor,
+                                   CXXSpecialMemberKind::MoveConstructor)
           : ClassDecl->hasTrivialMoveConstructor());
 
   MoveConstructor->setTrivialForCall(
       ClassDecl->hasAttr<TrivialABIAttr>() ||
       (ClassDecl->needsOverloadResolutionForMoveConstructor()
-           ? SpecialMemberIsTrivial(MoveConstructor, CXXMoveConstructor,
+           ? SpecialMemberIsTrivial(MoveConstructor,
+                                    CXXSpecialMemberKind::MoveConstructor,
                                     TAH_ConsiderTrivialABI)
            : ClassDecl->hasTrivialMoveConstructorForCall()));
 
@@ -15864,7 +15911,8 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
   Scope *S = getScopeForContext(ClassDecl);
   CheckImplicitSpecialMemberDeclaration(S, MoveConstructor);
 
-  if (ShouldDeleteSpecialMember(MoveConstructor, CXXMoveConstructor)) {
+  if (ShouldDeleteSpecialMember(MoveConstructor,
+                                CXXSpecialMemberKind::MoveConstructor)) {
     ClassDecl->setImplicitMoveConstructorIsDeleted();
     SetDeclDeleted(MoveConstructor, ClassLoc);
   }
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index a75e9925a4314..6b8ce0f633a3e 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -31,6 +31,7 @@
 #include "llvm/ADT/APInt.h"
 #include "llvm/ADT/FoldingSet.h"
 #include "llvm/ADT/PointerIntPair.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
@@ -10023,8 +10024,9 @@ bool InitializationSequence::Diagnose(Sema &S,
         // implicit.
         if (S.isImplicitlyDeleted(Best->Function))
           S.Diag(Kind.getLocation(), diag::err_ovl_deleted_special_init)
-            << S.getSpecialMember(cast<CXXMethodDecl>(Best->Function))
-            << DestType << ArgsRange;
+              << llvm::to_underlying(
+                     S.getSpecialMember(cast<CXXMethodDecl>(Best->Function)))
+              << DestType << ArgsRange;
         else
           S.Diag(Kind.getLocation(), diag::err_ovl_deleted_init)
               << DestType << ArgsRange;
diff --git a/clang/lib/Sema/SemaLookup.cpp b/clang/lib/Sema/SemaLookup.cpp
index 38237ee578079..d65f52b8efe81 100644
--- a/clang/lib/Sema/SemaLookup.cpp
+++ b/clang/lib/Sema/SemaLookup.cpp
@@ -37,6 +37,7 @@
 #include "clang/Sema/TemplateDeduction.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/TinyPtrVector.h"
 #include "llvm/ADT/edit_distance.h"
@@ -3341,21 +3342,20 @@ void Sema::LookupOverloadedOperatorName(OverloadedOperatorKind Op, Scope *S,
   Functions.append(Operators.begin(), Operators.end());
 }
 
-Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
-                                                           CXXSpecialMember SM,
-                                                           bool ConstArg,
-                                                           bool VolatileArg,
-                                                           bool RValueThis,
-                                                           bool ConstThis,
-                                                           bool VolatileThis) {
+Sema::SpecialMemberOverloadResult
+Sema::LookupSpecialMember(CXXRecordDecl *RD, CXXSpecialMemberKind SM,
+                          bool ConstArg, bool VolatileArg, bool RValueThis,
+                          bool ConstThis, bool VolatileThis) {
   assert(CanDeclareSpecialMemberFunction(RD) &&
          "doing special member lookup into record that isn't fully complete");
   RD = RD->getDefinition();
   if (RValueThis || ConstThis || VolatileThis)
-    assert((SM == CXXCopyAssignment || SM == CXXMoveAssignment) &&
+    assert((SM == CXXSpecialMemberKind::CopyAssignment ||
+            SM == CXXSpecialMemberKind::MoveAssignment) &&
            "constructors and destructors always have unqualified lvalue this");
   if (ConstArg || VolatileArg)
-    assert((SM != CXXDefaultConstructor && SM != CXXDestructor) &&
+    assert((SM != CXXSpecialMemberKind::DefaultConstructor &&
+            SM != CXXSpecialMemberKind::Destructor) &&
            "parameter-less special members can't have qualified arguments");
 
   // FIXME: Get the caller to pass in a location for the lookup.
@@ -3363,7 +3363,7 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
 
   llvm::FoldingSetNodeID ID;
   ID.AddPointer(RD);
-  ID.AddInteger(SM);
+  ID.AddInteger(llvm::to_underlying(SM));
   ID.AddInteger(ConstArg);
   ID.AddInteger(VolatileArg);
   ID.AddInteger(RValueThis);
@@ -3382,7 +3382,7 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
   Result = new (Result) SpecialMemberOverloadResultEntry(ID);
   SpecialMemberCache.InsertNode(Result, InsertPoint);
 
-  if (SM == CXXDestructor) {
+  if (SM == CXXSpecialMemberKind::Destructor) {
     if (RD->needsImplicitDestructor()) {
       runWithSufficientStackSpace(RD->getLocation(), [&] {
         DeclareImplicitDestructor(RD);
@@ -3406,7 +3406,7 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
   QualType ArgType = CanTy;
   ExprValueKind VK = VK_LValue;
 
-  if (SM == CXXDefaultConstructor) {
+  if (SM == CXXSpecialMemberKind::DefaultConstructor) {
     Name = Context.DeclarationNames.getCXXConstructorName(CanTy);
     NumArgs = 0;
     if (RD->needsImplicitDefaultConstructor()) {
@@ -3415,7 +3415,8 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
       });
     }
   } else {
-    if (SM == CXXCopyConstructor || SM == CXXMoveConstructor) {
+    if (SM == CXXSpecialMemberKind::CopyConstructor ||
+        SM == CXXSpecialMemberKind::MoveConstructor) {
       Name = Context.DeclarationNames.getCXXConstructorName(CanTy);
       if (RD->needsImplicitCopyConstructor()) {
         runWithSufficientStackSpace(RD->getLocation(), [&] {
@@ -3453,7 +3454,8 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
     // Possibly an XValue is actually correct in the case of move, but
     // there is no semantic difference for class types in this restricted
     // case.
-    if (SM == CXXCopyConstructor || SM == CXXCopyAssignment)
+    if (SM == CXXSpecialMemberKind::CopyConstructor ||
+        SM == CXXSpecialMemberKind::CopyAssignment)
       VK = VK_LValue;
     else
       VK = VK_PRValue;
@@ -3461,7 +3463,7 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
 
   OpaqueValueExpr FakeArg(LookupLoc, ArgType, VK);
 
-  if (SM != CXXDefaultConstructor) {
+  if (SM != CXXSpecialMemberKind::DefaultConstructor) {
     NumArgs = 1;
     Arg = &FakeArg;
   }
@@ -3487,7 +3489,7 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
     // type, rather than because there's some other declared constructor.
     // Every class has a copy/move constructor, copy/move assignment, and
     // destructor.
-    assert(SM == CXXDefaultConstructor &&
+    assert(SM == CXXSpecialMemberKind::DefaultConstructor &&
            "lookup for a constructor or assignment operator was empty");
     Result->setMethod(nullptr);
     Result->setKind(SpecialMemberOverloadResult::NoMemberOrDeleted);
@@ -3505,7 +3507,8 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
     DeclAccessPair Cand = DeclAccessPair::make(CandDecl, AS_public);
     auto CtorInfo = getConstructorInfo(Cand);
     if (CXXMethodDecl *M = dyn_cast<CXXMethodDecl>(Cand->getUnderlyingDecl())) {
-      if (SM == CXXCopyAssignment || SM == CXXMoveAssignment)
+      if (SM == CXXSpecialMemberKind::CopyAssignment ||
+          SM == CXXSpecialMemberKind::MoveAssignment)
         AddMethodCandidate(M, Cand, RD, ThisTy, Classification,
                            llvm::ArrayRef(&Arg, NumArgs), OCS, true);
       else if (CtorInfo)
@@ -3517,7 +3520,8 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
                              /*SuppressUserConversions*/ true);
     } else if (FunctionTemplateDecl *Tmpl =
                  dyn_cast<FunctionTemplateDecl>(Cand->getUnderlyingDecl())) {
-      if (SM == CXXCopyAssignment || SM == CXXMoveAssignment)
+      if (SM == CXXSpecialMemberKind::CopyAssignment ||
+          SM == CXXSpecialMemberKind::MoveAssignment)
         AddMethodTemplateCandidate(Tmpl, Cand, RD, nullptr, ThisTy,
                                    Classification,
                                    llvm::ArrayRef(&Arg, NumArgs), OCS, true);
@@ -3563,8 +3567,8 @@ Sema::SpecialMemberOverloadResult Sema::LookupSpecialMember(CXXRecordDecl *RD,
 /// Look up the default constructor for the given class.
 CXXConstructorDecl *Sema::LookupDefaultConstructor(CXXRecordDecl *Class) {
   SpecialMemberOverloadResult Result =
-    LookupSpecialMember(Class, CXXDefaultConstructor, false, false, false,
-                        false, false);
+      LookupSpecialMember(Class, CXXSpecialMemberKind::DefaultConstructor,
+                          false, false, false, false, false);
 
   return cast_or_null<CXXConstructorDecl>(Result.getMethod());
 }
@@ -3574,9 +3578,9 @@ CXXConstructorDecl *Sema::LookupCopyingConstructor(CXXRecordDecl *Class,
                                                    unsigned Quals) {
   assert(!(Quals & ~(Qualifiers::Const | Qualifiers::Volatile)) &&
          "non-const, non-volatile qualifiers for copy ctor arg");
-  SpecialMemberOverloadResult Result =
-    LookupSpecialMember(Class, CXXCopyConstructor, Quals & Qualifiers::Const,
-                        Quals & Qualifiers::Volatile, false, false, false);
+  SpecialMemberOverloadResult Result = LookupSpecialMember(
+      Class, CXXSpecialMemberKind::CopyConstructor, Quals & Qualifiers::Const,
+      Quals & Qualifiers::Volatile, false, false, false);
 
   return cast_or_null<CXXConstructorDecl>(Result.getMethod());
 }
@@ -3584,9 +3588,9 @@ CXXConstructorDecl *Sema::LookupCopyingConstructor(CXXRecordDecl *Class,
 /// Look up the moving constructor for the given class.
 CXXConstructorDecl *Sema::LookupMovingConstructor(CXXRecordDecl *Class,
                                                   unsigned Quals) {
-  SpecialMemberOverloadResult Result =
-    LookupSpecialMember(Class, CXXMoveConstructor, Quals & Qualifiers::Const,
-                        Quals & Qualifiers::Volatile, false, false, false);
+  SpecialMemberOverloadResult Result = LookupSpecialMember(
+      Class, CXXSpecialMemberKind::MoveConstructor, Quals & Qualifiers::Const,
+      Quals & Qualifiers::Volatile, false, false, false);
 
   return cast_or_null<CXXConstructorDecl>(Result.getMethod());
 }
@@ -3618,11 +3622,10 @@ CXXMethodDecl *Sema::LookupCopyingAssignment(CXXRecordDecl *Class,
          "non-const, non-volatile qualifiers for copy assignment arg");
   assert(!(ThisQuals & ~(Qualifiers::Const | Qualifiers::Volatile)) &&
          "non-const, non-volatile qualifiers for copy assignment this");
-  SpecialMemberOverloadResult Result =
-    LookupSpecialMember(Class, CXXCopyAssignment, Quals & Qualifiers::Const,
-                        Quals & Qualifiers::Volatile, RValueThis,
-                        ThisQuals & Qualifiers::Const,
-                        ThisQuals & Qualifiers::Volatile);
+  SpecialMemberOverloadResult Result = LookupSpecialMember(
+      Class, CXXSpecialMemberKind::CopyAssignment, Quals & Qualifiers::Const,
+      Quals & Qualifiers::Volatile, RValueThis, ThisQuals & Qualifiers::Const,
+      ThisQuals & Qualifiers::Volatile);
 
   return Result.getMethod();
 }
@@ -3634,11 +3637,10 @@ CXXMethodDecl *Sema::LookupMovingAssignment(CXXRecordDecl *Class,
                                             unsigned ThisQuals) {
   assert(!(ThisQuals & ~(Qualifiers::Const | Qualifiers::Volatile)) &&
          "non-const, non-volatile qualifiers for copy assignment this");
-  SpecialMemberOverloadResult Result =
-    LookupSpecialMember(Class, CXXMoveAssignment, Quals & Qualifiers::Const,
-                        Quals & Qualifiers::Volatile, RValueThis,
-                        ThisQuals & Qualifiers::Const,
-                        ThisQuals & Qualifiers::Volatile);
+  SpecialMemberOverloadResult Result = LookupSpecialMember(
+      Class, CXXSpecialMemberKind::MoveAssignment, Quals & Qualifiers::Const,
+      Quals & Qualifiers::Volatile, RValueThis, ThisQuals & Qualifiers::Const,
+      ThisQuals & Qualifiers::Volatile);
 
   return Result.getMethod();
 }
@@ -3651,8 +3653,8 @@ CXXMethodDecl *Sema::LookupMovingAssignment(CXXRecordDecl *Class,
 /// \returns The destructor for this class.
 CXXDestructorDecl *Sema::LookupDestructor(CXXRecordDecl *Class) {
   return cast_or_null<CXXDestructorDecl>(
-      LookupSpecialMember(Class, CXXDestructor, false, false, false, false,
-                          false)
+      LookupSpecialMember(Class, CXXSpecialMemberKind::Destructor, false, false,
+                          false, false, false)
           .getMethod());
 }
 
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 3808af37ff54a..1f674dd4bb0fb 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -36,6 +36,7 @@
 #include "clang/Sema/TemplateDeduction.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
@@ -11955,25 +11956,25 @@ static void DiagnoseBadTarget(Sema &S, OverloadCandidate *Cand) {
   CXXMethodDecl *Meth = dyn_cast<CXXMethodDecl>(Callee);
   if (Meth != nullptr && Meth->isImplicit()) {
     CXXRecordDecl *ParentClass = Meth->getParent();
-    Sema::CXXSpecialMember CSM;
+    CXXSpecialMemberKind CSM;
 
     switch (FnKindPair.first) {
     default:
       return;
     case oc_implicit_default_constructor:
-      CSM = Sema::CXXDefaultConstructor;
+      CSM = CXXSpecialMemberKind::DefaultConstructor;
       break;
     case oc_implicit_copy_constructor:
-      CSM = Sema::CXXCopyConstructor;
+      CSM = CXXSpecialMemberKind::CopyConstructor;
       break;
     case oc_implicit_move_constructor:
-      CSM = Sema::CXXMoveConstructor;
+      CSM = CXXSpecialMemberKind::MoveConstructor;
       break;
     case oc_implicit_copy_assignment:
-      CSM = Sema::CXXCopyAssignment;
+      CSM = CXXSpecialMemberKind::CopyAssignment;
       break;
     case oc_implicit_move_assignment:
-      CSM = Sema::CXXMoveAssignment;
+      CSM = CXXSpecialMemberKind::MoveAssignment;
       break;
     };
 
@@ -15064,7 +15065,8 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
         DefaultedFunctionKind DFK = getDefaultedFunctionKind(DeletedFD);
         if (DFK.isSpecialMember()) {
           Diag(OpLoc, diag::err_ovl_deleted_special_oper)
-            << Args[0]->getType() << DFK.asSpecialMember();
+              << Args[0]->getType()
+              << llvm::to_underlying(DFK.asSpecialMember());
         } else {
           assert(DFK.isComparison());
           Diag(OpLoc, diag::err_ovl_deleted_comparison)
diff --git a/clang/lib/Sema/SemaTemplateInstantiate.cpp b/clang/lib/Sema/SemaTemplateInstantiate.cpp
index a265fd1c46a63..7cd428de0bb32 100644
--- a/clang/lib/Sema/SemaTemplateInstantiate.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiate.cpp
@@ -36,6 +36,7 @@
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateDeduction.h"
 #include "clang/Sema/TemplateInstCallback.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/TimeProfiler.h"
@@ -1142,7 +1143,8 @@ void Sema::PrintInstantiationStack() {
     case CodeSynthesisContext::DeclaringSpecialMember:
       Diags.Report(Active->PointOfInstantiation,
                    diag::note_in_declaration_of_implicit_special_member)
-        << cast<CXXRecordDecl>(Active->Entity) << Active->SpecialMember;
+          << cast<CXXRecordDecl>(Active->Entity)
+          << llvm::to_underlying(Active->SpecialMember);
       break;
 
     case CodeSynthesisContext::DeclaringImplicitEqualityComparison:
@@ -1160,7 +1162,8 @@ void Sema::PrintInstantiationStack() {
         auto *MD = cast<CXXMethodDecl>(FD);
         Diags.Report(Active->PointOfInstantiation,
                      diag::note_member_synthesized_at)
-            << MD->isExplicitlyDefaulted() << DFK.asSpecialMember()
+            << MD->isExplicitlyDefaulted()
+            << llvm::to_underlying(DFK.asSpecialMember())
             << Context.getTagDeclType(MD->getParent());
       } else if (DFK.isComparison()) {
         QualType RecordType = FD->getParamDecl(0)
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 707446e132094..7f510d34d671e 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5123,10 +5123,10 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
     assert(PatternDecl->isDefaulted() &&
            "Special member needs to be defaulted");
     auto PatternSM = getDefaultedFunctionKind(PatternDecl).asSpecialMember();
-    if (!(PatternSM == Sema::CXXCopyConstructor ||
-          PatternSM == Sema::CXXCopyAssignment ||
-          PatternSM == Sema::CXXMoveConstructor ||
-          PatternSM == Sema::CXXMoveAssignment))
+    if (!(PatternSM == CXXSpecialMemberKind::CopyConstructor ||
+          PatternSM == CXXSpecialMemberKind::CopyAssignment ||
+          PatternSM == CXXSpecialMemberKind::MoveConstructor ||
+          PatternSM == CXXSpecialMemberKind::MoveAssignment))
       return;
 
     auto *NewRec = dyn_cast<CXXRecordDecl>(Function->getDeclContext());
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index 8762744396f4d..b8a1518fbe00a 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -9738,7 +9738,8 @@ bool Sema::RequireLiteralType(SourceLocation Loc, QualType T,
                                     : diag::note_non_literal_nontrivial_dtor)
           << RD;
       if (!Dtor->isUserProvided())
-        SpecialMemberIsTrivial(Dtor, CXXDestructor, TAH_IgnoreTrivialABI,
+        SpecialMemberIsTrivial(Dtor, CXXSpecialMemberKind::Destructor,
+                               TAH_IgnoreTrivialABI,
                                /*Diagnose*/ true);
     }
   }

From 9250aedb5cabf0e44c694f0c8e68ae03e937be89 Mon Sep 17 00:00:00 2001
From: Andreas Jonson <andjo403@hotmail.com>
Date: Fri, 12 Apr 2024 11:21:48 +0200
Subject: [PATCH 228/886] [SCEV] Add range attribute handling (#88449)

---
 llvm/lib/Analysis/ScalarEvolution.cpp         |  9 ++-
 .../ScalarEvolution/range-attribute.ll        | 74 +++++++++++++++++++
 2 files changed, 82 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/Analysis/ScalarEvolution/range-attribute.ll

diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 9fcce797f5597..95440dda3b4c0 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -6382,9 +6382,16 @@ uint32_t ScalarEvolution::getMinTrailingZeros(const SCEV *S) {
 
 /// Helper method to assign a range to V from metadata present in the IR.
 static std::optional<ConstantRange> GetRangeFromMetadata(Value *V) {
-  if (Instruction *I = dyn_cast<Instruction>(V))
+  if (Instruction *I = dyn_cast<Instruction>(V)) {
     if (MDNode *MD = I->getMetadata(LLVMContext::MD_range))
       return getConstantRangeFromMetadata(*MD);
+    if (const auto *CB = dyn_cast<CallBase>(V))
+      if (std::optional<ConstantRange> Range = CB->getRange())
+        return Range;
+  }
+  if (auto *A = dyn_cast<Argument>(V))
+    if (std::optional<ConstantRange> Range = A->getRange())
+      return Range;
 
   return std::nullopt;
 }
diff --git a/llvm/test/Analysis/ScalarEvolution/range-attribute.ll b/llvm/test/Analysis/ScalarEvolution/range-attribute.ll
new file mode 100644
index 0000000000000..ea22695e58d1d
--- /dev/null
+++ b/llvm/test/Analysis/ScalarEvolution/range-attribute.ll
@@ -0,0 +1,74 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -disable-output "-passes=print<scalar-evolution>" -scalar-evolution-classify-expressions=0 < %s 2>&1 | FileCheck %s
+
+define i32 @slt_trip_count_with_range_attr(i32 range(i32 1, 100) %limit) {
+;
+; CHECK-LABEL: 'slt_trip_count_with_range_attr'
+; CHECK-NEXT:  Determining loop execution counts for: @slt_trip_count_with_range_attr
+; CHECK-NEXT:  Loop %loop: backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is i32 98
+; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: Trip multiple is 1
+;
+ entry:
+  br label %loop
+
+ loop:
+  %index = phi i32 [ 0, %entry ], [ %index.inc, %loop ]
+  %index.inc = add i32 %index, 1
+  %continue = icmp slt i32 %index.inc, %limit
+  br i1 %continue, label %loop, label %loop.exit
+
+ loop.exit:
+  ret i32 0
+}
+
+declare i32 @get_i32()
+
+define i32 @slt_trip_count_with_range_call() {
+;
+; CHECK-LABEL: 'slt_trip_count_with_range_call'
+; CHECK-NEXT:  Determining loop execution counts for: @slt_trip_count_with_range_call
+; CHECK-NEXT:  Loop %loop: backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is i32 98
+; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: Trip multiple is 1
+;
+ entry:
+  %limit = call range(i32 1, 100) i32 @get_i32()
+  br label %loop
+
+ loop:
+  %index = phi i32 [ 0, %entry ], [ %index.inc, %loop ]
+  %index.inc = add i32 %index, 1
+  %continue = icmp slt i32 %index.inc, %limit
+  br i1 %continue, label %loop, label %loop.exit
+
+ loop.exit:
+  ret i32 0
+}
+
+declare range(i32 1, 100) i32 @get_i32_in_range()
+
+define i32 @slt_trip_count_with_range_result() {
+;
+; CHECK-LABEL: 'slt_trip_count_with_range_result'
+; CHECK-NEXT:  Determining loop execution counts for: @slt_trip_count_with_range_result
+; CHECK-NEXT:  Loop %loop: backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is i32 98
+; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is (-1 + %limit)<nsw>
+; CHECK-NEXT:  Loop %loop: Trip multiple is 1
+;
+ entry:
+  %limit = call i32 @get_i32_in_range()
+  br label %loop
+
+ loop:
+  %index = phi i32 [ 0, %entry ], [ %index.inc, %loop ]
+  %index.inc = add i32 %index, 1
+  %continue = icmp slt i32 %index.inc, %limit
+  br i1 %continue, label %loop, label %loop.exit
+
+ loop.exit:
+  ret i32 0
+}

From 5d6b00929bd026d1448a5dd3a8341ea99b3befe1 Mon Sep 17 00:00:00 2001
From: AtariDreams <gfunni234@gmail.com>
Date: Fri, 12 Apr 2024 05:24:03 -0400
Subject: [PATCH 229/886] [NFC] Replace m_Sub(m_Zero(), X) with m_Neg(X)
 (#88461)

---
 llvm/lib/Analysis/InstructionSimplify.cpp                | 2 +-
 llvm/lib/Analysis/ValueTracking.cpp                      | 4 ++--
 llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp           | 2 +-
 llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp | 7 +++----
 4 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 3c943a09a9c23..1aaa4d658803b 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -5052,7 +5052,7 @@ static Value *simplifyGEPInst(Type *SrcTy, Value *Ptr,
 
       // gep (gep V, C), (sub 0, V) -> C
       if (match(Indices.back(),
-                m_Sub(m_Zero(), m_PtrToInt(m_Specific(StrippedBasePtr)))) &&
+                m_Neg(m_PtrToInt(m_Specific(StrippedBasePtr)))) &&
           !BasePtrOffset.isZero()) {
         auto *CI = ConstantInt::get(GEPTy->getContext(), BasePtrOffset);
         return ConstantExpr::getIntToPtr(CI, GEPTy);
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index eb8925c534753..ca257cc6aefdf 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -8016,12 +8016,12 @@ bool llvm::isKnownNegation(const Value *X, const Value *Y, bool NeedNSW) {
 
   // X = sub (0, Y) || X = sub nsw (0, Y)
   if ((!NeedNSW && match(X, m_Sub(m_ZeroInt(), m_Specific(Y)))) ||
-      (NeedNSW && match(X, m_NSWSub(m_ZeroInt(), m_Specific(Y)))))
+      (NeedNSW && match(X, m_NSWNeg(m_Specific(Y)))))
     return true;
 
   // Y = sub (0, X) || Y = sub nsw (0, X)
   if ((!NeedNSW && match(Y, m_Sub(m_ZeroInt(), m_Specific(X)))) ||
-      (NeedNSW && match(Y, m_NSWSub(m_ZeroInt(), m_Specific(X)))))
+      (NeedNSW && match(Y, m_NSWNeg(m_Specific(X)))))
     return true;
 
   // X = sub (A, B), Y = sub (B, A) || X = sub nsw (A, B), Y = sub nsw (B, A)
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 1dd0fa49a460f..e002cf1d1de38 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -4308,7 +4308,7 @@ bool SelectionDAG::isKnownToBeAPowerOfTwo(SDValue Val, unsigned Depth) const {
   //    x & -x -> non-zero pow2
   // so if we find the pattern return whether we know `x` is non-zero.
   SDValue X;
-  if (sd_match(Val, m_And(m_Value(X), m_Sub(m_Zero(), m_Deferred(X)))))
+  if (sd_match(Val, m_And(m_Value(X), m_Neg(m_Deferred(X)))))
     return isKnownNeverZero(X, Depth);
 
   if (Val.getOpcode() == ISD::ZERO_EXTEND)
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 8c698e52b5a0e..4dc1319f1c437 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -1572,8 +1572,7 @@ Instruction *InstCombinerImpl::visitSDiv(BinaryOperator &I) {
     // -X / C --> X / -C (if the negation doesn't overflow).
     // TODO: This could be enhanced to handle arbitrary vector constants by
     //       checking if all elements are not the min-signed-val.
-    if (!Op1C->isMinSignedValue() &&
-        match(Op0, m_NSWSub(m_Zero(), m_Value(X)))) {
+    if (!Op1C->isMinSignedValue() && match(Op0, m_NSWNeg(m_Value(X)))) {
       Constant *NegC = ConstantInt::get(Ty, -(*Op1C));
       Instruction *BO = BinaryOperator::CreateSDiv(X, NegC);
       BO->setIsExact(I.isExact());
@@ -1583,7 +1582,7 @@ Instruction *InstCombinerImpl::visitSDiv(BinaryOperator &I) {
 
   // -X / Y --> -(X / Y)
   Value *Y;
-  if (match(&I, m_SDiv(m_OneUse(m_NSWSub(m_Zero(), m_Value(X))), m_Value(Y))))
+  if (match(&I, m_SDiv(m_OneUse(m_NSWNeg(m_Value(X))), m_Value(Y))))
     return BinaryOperator::CreateNSWNeg(
         Builder.CreateSDiv(X, Y, I.getName(), I.isExact()));
 
@@ -2173,7 +2172,7 @@ Instruction *InstCombinerImpl::visitSRem(BinaryOperator &I) {
 
   // -X srem Y --> -(X srem Y)
   Value *X, *Y;
-  if (match(&I, m_SRem(m_OneUse(m_NSWSub(m_Zero(), m_Value(X))), m_Value(Y))))
+  if (match(&I, m_SRem(m_OneUse(m_NSWNeg(m_Value(X))), m_Value(Y))))
     return BinaryOperator::CreateNSWNeg(Builder.CreateSRem(X, Y));
 
   // If the sign bits of both operands are zero (i.e. we can prove they are

From 0e6367236413e50599b49aff15b086088f152769 Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Fri, 12 Apr 2024 10:35:48 +0100
Subject: [PATCH 230/886] [MLIR][OpenMP] Add omp.loop_nest operation (#87083)

This patch introduces an operation intended to hold loop information
associated to the `omp.distribute`, `omp.simdloop`, `omp.taskloop` and
`omp.wsloop` operations. This is a stopgap solution to unblock work on
transitioning these operations to becoming wrappers, as discussed in
[this
RFC](https://discourse.llvm.org/t/rfc-representing-combined-composite-constructs-in-the-openmp-dialect/76986).

Long-term, this operation will likely be replaced by
`omp.canonical_loop`, which is being designed to address missing support
for loop transformations, etc.
---
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td | 66 +++++++++++++++-
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  | 71 +++++++++++++++++
 mlir/test/Dialect/OpenMP/invalid.mlir         | 37 +++++++++
 mlir/test/Dialect/OpenMP/ops.mlir             | 79 +++++++++++++++++++
 4 files changed, 252 insertions(+), 1 deletion(-)

diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 2a8582be38331..a2eb39a8e6077 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -524,6 +524,70 @@ def SingleOp : OpenMP_Op<"single", [AttrSizedOperandSegments]> {
   let hasVerifier = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// Loop Nest
+//===----------------------------------------------------------------------===//
+
+def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
+                        AllTypesMatch<["lowerBound", "upperBound", "step"]>,
+                        ParentOneOf<["DistributeOp", "SimdLoopOp", "TaskloopOp",
+                                     "WsloopOp"]>,
+                        RecursiveMemoryEffects]> {
+  let summary = "rectangular loop nest";
+  let description = [{
+    This operation represents a collapsed rectangular loop nest. For each
+    rectangular loop of the nest represented by an instance of this operation,
+    lower and upper bounds, as well as a step variable, must be defined.
+
+    The lower and upper bounds specify a half-open range: the range includes the
+    lower bound but does not include the upper bound. If the `inclusive`
+    attribute is specified then the upper bound is also included.
+
+    The body region can contain any number of blocks. The region is terminated
+    by an `omp.yield` instruction without operands. The induction variables,
+    represented as entry block arguments to the loop nest operation's single
+    region, match the types of the `lowerBound`, `upperBound` and `step`
+    arguments.
+
+    ```mlir
+    omp.loop_nest (%i1, %i2) : i32 = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
+      %a = load %arrA[%i1, %i2] : memref<?x?xf32>
+      %b = load %arrB[%i1, %i2] : memref<?x?xf32>
+      %sum = arith.addf %a, %b : f32
+      store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
+      omp.yield
+    }
+    ```
+
+    This is a temporary simplified definition of a loop based on existing OpenMP
+    loop operations intended to serve as a stopgap solution until the long-term
+    representation of canonical loops is defined. Specifically, this operation
+    is intended to serve as a unique source for loop information during the
+    transition to making `omp.distribute`, `omp.simdloop`, `omp.taskloop` and
+    `omp.wsloop` wrapper operations. It is not intended to help with the
+    addition of support for loop transformations, non-rectangular loops and
+    non-perfectly nested loops.
+  }];
+
+  let arguments = (ins Variadic<IntLikeType>:$lowerBound,
+                       Variadic<IntLikeType>:$upperBound,
+                       Variadic<IntLikeType>:$step,
+                       UnitAttr:$inclusive);
+
+  let regions = (region AnyRegion:$region);
+
+  let extraClassDeclaration = [{
+    /// Returns the number of loops in the loop nest.
+    unsigned getNumLoops() { return getLowerBound().size(); }
+
+    /// Returns the induction variables of the loop nest.
+    ArrayRef<BlockArgument> getIVs() { return getRegion().getArguments(); }
+  }];
+
+  let hasCustomAssemblyFormat = 1;
+  let hasVerifier = 1;
+}
+
 //===----------------------------------------------------------------------===//
 // 2.9.2 Workshare Loop Construct
 //===----------------------------------------------------------------------===//
@@ -743,7 +807,7 @@ def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
 
 def YieldOp : OpenMP_Op<"yield",
     [Pure, ReturnLike, Terminator,
-     ParentOneOf<["WsloopOp", "DeclareReductionOp",
+     ParentOneOf<["LoopNestOp", "WsloopOp", "DeclareReductionOp",
      "AtomicUpdateOp", "SimdLoopOp", "PrivateClauseOp"]>]> {
   let summary = "loop yield and termination operation";
   let description = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 2d5b2231d2dd5..4ab78bf331f80 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1862,6 +1862,77 @@ LogicalResult TaskloopOp::verify() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// LoopNestOp
+//===----------------------------------------------------------------------===//
+
+ParseResult LoopNestOp::parse(OpAsmParser &parser, OperationState &result) {
+  // Parse an opening `(` followed by induction variables followed by `)`
+  SmallVector<OpAsmParser::Argument> ivs;
+  SmallVector<OpAsmParser::UnresolvedOperand> lbs, ubs;
+  Type loopVarType;
+  if (parser.parseArgumentList(ivs, OpAsmParser::Delimiter::Paren) ||
+      parser.parseColonType(loopVarType) ||
+      // Parse loop bounds.
+      parser.parseEqual() ||
+      parser.parseOperandList(lbs, ivs.size(), OpAsmParser::Delimiter::Paren) ||
+      parser.parseKeyword("to") ||
+      parser.parseOperandList(ubs, ivs.size(), OpAsmParser::Delimiter::Paren))
+    return failure();
+
+  for (auto &iv : ivs)
+    iv.type = loopVarType;
+
+  // Parse "inclusive" flag.
+  if (succeeded(parser.parseOptionalKeyword("inclusive")))
+    result.addAttribute("inclusive",
+                        UnitAttr::get(parser.getBuilder().getContext()));
+
+  // Parse step values.
+  SmallVector<OpAsmParser::UnresolvedOperand> steps;
+  if (parser.parseKeyword("step") ||
+      parser.parseOperandList(steps, ivs.size(), OpAsmParser::Delimiter::Paren))
+    return failure();
+
+  // Parse the body.
+  Region *region = result.addRegion();
+  if (parser.parseRegion(*region, ivs))
+    return failure();
+
+  // Resolve operands.
+  if (parser.resolveOperands(lbs, loopVarType, result.operands) ||
+      parser.resolveOperands(ubs, loopVarType, result.operands) ||
+      parser.resolveOperands(steps, loopVarType, result.operands))
+    return failure();
+
+  // Parse the optional attribute list.
+  return parser.parseOptionalAttrDict(result.attributes);
+}
+
+void LoopNestOp::print(OpAsmPrinter &p) {
+  Region &region = getRegion();
+  auto args = region.getArguments();
+  p << " (" << args << ") : " << args[0].getType() << " = (" << getLowerBound()
+    << ") to (" << getUpperBound() << ") ";
+  if (getInclusive())
+    p << "inclusive ";
+  p << "step (" << getStep() << ") ";
+  p.printRegion(region, /*printEntryBlockArgs=*/false);
+}
+
+LogicalResult LoopNestOp::verify() {
+  if (getLowerBound().size() != getIVs().size())
+    return emitOpError() << "number of range arguments and IVs do not match";
+
+  for (auto [lb, iv] : llvm::zip_equal(getLowerBound(), getIVs())) {
+    if (lb.getType() != iv.getType())
+      return emitOpError()
+             << "range argument type does not match corresponding IV type";
+  }
+
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // WsloopOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index 27a440b3f97ca..3802fbde534d6 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -87,6 +87,43 @@ func.func @proc_bind_once() {
 
 // -----
 
+func.func @invalid_parent(%lb : index, %ub : index, %step : index) {
+  // expected-error@+1 {{op expects parent op to be one of 'omp.distribute, omp.simdloop, omp.taskloop, omp.wsloop'}}
+  omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+    omp.yield
+  }
+}
+
+// -----
+
+func.func @type_mismatch(%lb : index, %ub : index, %step : index) {
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // expected-error@+1 {{range argument type does not match corresponding IV type}}
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb0(%iv2: i32):
+      omp.yield
+    }) : (index, index, index) -> ()
+    omp.yield
+  }
+}
+
+// -----
+
+func.func @iv_number_mismatch(%lb : index, %ub : index, %step : index) {
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // expected-error@+1 {{number of range arguments and IVs do not match}}
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb0(%iv1 : index, %iv2 : index):
+      omp.yield
+    }) : (index, index, index) -> ()
+    omp.yield
+  }
+}
+
+// -----
+
 func.func @inclusive_not_a_clause(%lb : index, %ub : index, %step : index) {
   // expected-error @below {{expected 'for'}}
   omp.wsloop nowait inclusive
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index ad5b74b84ac70..851d44ad984ee 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -133,6 +133,85 @@ func.func @omp_parallel_pretty(%data_var : memref<i32>, %if_cond : i1, %num_thre
   return
 }
 
+// CHECK-LABEL: omp_loop_nest
+func.func @omp_loop_nest(%lb : index, %ub : index, %step : index) -> () {
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) step (%{{.*}})
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb0(%iv2: index):
+      omp.yield
+    }) : (index, index, index) -> ()
+    omp.yield
+  }
+
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) inclusive step (%{{.*}})
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb0(%iv2: index):
+      omp.yield
+    }) {inclusive} : (index, index, index) -> ()
+    omp.yield
+  }
+
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}, %{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}})
+    "omp.loop_nest" (%lb, %lb, %ub, %ub, %step, %step) ({
+    ^bb0(%iv2: index, %iv3: index):
+      omp.yield
+    }) : (index, index, index, index, index, index) -> ()
+    omp.yield
+  }
+
+  return
+}
+
+// CHECK-LABEL: omp_loop_nest_pretty
+func.func @omp_loop_nest_pretty(%lb : index, %ub : index, %step : index) -> () {
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) step (%{{.*}})
+    omp.loop_nest (%iv2) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+    omp.yield
+  }
+
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) inclusive step (%{{.*}})
+    omp.loop_nest (%iv2) : index = (%lb) to (%ub) inclusive step (%step) {
+      omp.yield
+    }
+    omp.yield
+  }
+
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}})
+    omp.loop_nest (%iv2, %iv3) : index = (%lb, %lb) to (%ub, %ub) step (%step, %step) {
+      omp.yield
+    }
+    omp.yield
+  }
+
+  return
+}
+
 // CHECK-LABEL: omp_wsloop
 func.func @omp_wsloop(%lb : index, %ub : index, %step : index, %data_var : memref<i32>, %linear_var : i32, %chunk_var : i32) -> () {
 

From f5b2d24b59f168eaeed08b5f45bf0c8dfcc1c292 Mon Sep 17 00:00:00 2001
From: Benjamin Kramer <benny.kra@googlemail.com>
Date: Fri, 12 Apr 2024 11:47:53 +0200
Subject: [PATCH 231/886] [NVPTX] Move roundeven on bf16 into the sm_90 has it
 but sm_80 doesn't bucket

---
 llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp  |  1 +
 llvm/test/CodeGen/NVPTX/bf16-instructions.ll | 12 ++++++++++++
 2 files changed, 13 insertions(+)

diff --git a/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp b/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
index e1cdbd6a4b4dc..44b61a937d64e 100644
--- a/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXISelLowering.cpp
@@ -447,6 +447,7 @@ NVPTXTargetLowering::NVPTXTargetLowering(const NVPTXTargetMachine &TM,
       case ISD::FFLOOR:
       case ISD::FNEARBYINT:
       case ISD::FRINT:
+      case ISD::FROUNDEVEN:
       case ISD::FTRUNC:
         IsOpSupported = STI.getSmVersion() >= 90 && STI.getPTXVersion() >= 78;
         break;
diff --git a/llvm/test/CodeGen/NVPTX/bf16-instructions.ll b/llvm/test/CodeGen/NVPTX/bf16-instructions.ll
index 884860712632d..4cbab8bf24166 100644
--- a/llvm/test/CodeGen/NVPTX/bf16-instructions.ll
+++ b/llvm/test/CodeGen/NVPTX/bf16-instructions.ll
@@ -330,3 +330,15 @@ define bfloat @test_uitofp_i64(i64 %a) {
   %r = uitofp i64 %a to bfloat
   ret bfloat %r
 }
+
+; CHECK-LABEL: test_roundeven(
+; CHECK:      ld.param.b16      [[A:%rs[0-9]+]], [test_roundeven_param_0];
+; SM80:       cvt.rni.f32.f32   [[F:%f[0-9]+]]
+; SM80:       cvt.rn.bf16.f32   [[R:%rs[0-9]+]], [[F]];
+; SM90:       cvt.rni.bf16.bf16 [[R:%rs[0-9]+]], [[A]];
+; CHECK:      st.param.b16      [func_retval0+0], [[R]];
+; CHECK:      ret;
+define bfloat @test_roundeven(bfloat %a) {
+  %r = call bfloat @llvm.roundeven.bf16(bfloat %a)
+  ret bfloat %r
+}

From b24af43fdfa1b1242b7cb77540462212227c57c4 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Fri, 12 Apr 2024 10:57:01 +0100
Subject: [PATCH 232/886] [AArch64] Improve scheduling latency into Bundles
 (#86310)

By default the scheduling info of instructions into a BUNDLE are given a
latency of 0 as they operate on the implicit register of the bundle.
This modifies that for AArch64 so that the latency is adjusted to use
the latency from the instruction in the bundle instead. This essentially
assumes that the bundled instructions are executed in a single cycle,
which for AArch64 is probably OK considering they are mostly used for
MOVPFX bundles, where this can help create slightly better scheduling
especially for in-order cores.
---
 .../llvm/CodeGen/TargetSubtargetInfo.h        |   4 +-
 llvm/lib/CodeGen/MachinePipeliner.cpp         |   3 +-
 llvm/lib/CodeGen/ScheduleDAGInstrs.cpp        |   8 +-
 .../SelectionDAG/ScheduleDAGSDNodes.cpp       |   2 +-
 llvm/lib/Target/AArch64/AArch64Subtarget.cpp  |  39 ++++++
 llvm/lib/Target/AArch64/AArch64Subtarget.h    |   3 +
 llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp    |   5 +-
 llvm/lib/Target/AMDGPU/GCNSubtarget.h         |   3 +-
 llvm/lib/Target/Hexagon/HexagonSubtarget.cpp  |   6 +-
 llvm/lib/Target/Hexagon/HexagonSubtarget.h    |   3 +-
 ...x-deinterleaving-add-mull-scalable-fast.ll |   6 +-
 .../CodeGen/AArch64/llvm-ir-to-intrinsic.ll   |   2 +-
 llvm/test/CodeGen/AArch64/misched-bundle.mir  |  36 +++---
 .../AArch64/sve-fixed-length-build-vector.ll  |   2 +-
 .../AArch64/sve-fixed-length-fp-arith.ll      |  24 ++--
 .../AArch64/sve-fixed-length-fp-fma.ll        |   6 +-
 .../AArch64/sve-fixed-length-fp-minmax.ll     |  24 ++--
 .../AArch64/sve-fixed-length-int-arith.ll     |   8 +-
 .../AArch64/sve-fixed-length-int-minmax.ll    |  32 ++---
 .../AArch64/sve-fixed-length-int-mulh.ll      |  16 +--
 .../AArch64/sve-fixed-length-int-rem.ll       |  96 +++++++-------
 .../AArch64/sve-fixed-length-int-shifts.ll    |  24 ++--
 llvm/test/CodeGen/AArch64/sve-fpext-load.ll   |   2 +-
 llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll   |  70 +++++-----
 llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll   | 122 +++++++++---------
 .../AArch64/sve-gather-scatter-addr-opts.ll   |  10 +-
 .../AArch64/sve-masked-gather-legalize.ll     |   4 +-
 llvm/test/CodeGen/AArch64/sve-smulo-sdnode.ll |  24 ++--
 llvm/test/CodeGen/AArch64/sve-split-fcvt.ll   |  34 ++---
 ...aming-mode-fixed-length-fp-extend-trunc.ll |   6 +-
 ...sve-streaming-mode-fixed-length-int-div.ll |   8 +-
 ...sve-streaming-mode-fixed-length-int-rem.ll |  32 ++---
 ...e-streaming-mode-fixed-length-int-to-fp.ll |  20 +--
 llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll |  16 +--
 .../test/CodeGen/AArch64/sve-vecreduce-dot.ll |   2 +-
 llvm/test/CodeGen/AArch64/sve2-xar.ll         |   2 +-
 36 files changed, 377 insertions(+), 327 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/TargetSubtargetInfo.h b/llvm/include/llvm/CodeGen/TargetSubtargetInfo.h
index 7f8ed5c501989..5023e29ce145a 100644
--- a/llvm/include/llvm/CodeGen/TargetSubtargetInfo.h
+++ b/llvm/include/llvm/CodeGen/TargetSubtargetInfo.h
@@ -235,7 +235,9 @@ class TargetSubtargetInfo : public MCSubtargetInfo {
   // and UseOpIdx are the indices of the operands in Def and Use, respectively.
   // Otherwise, either may be -1.
   virtual void adjustSchedDependency(SUnit *Def, int DefOpIdx, SUnit *Use,
-                                     int UseOpIdx, SDep &Dep) const {}
+                                     int UseOpIdx, SDep &Dep,
+                                     const TargetSchedModel *SchedModel) const {
+  }
 
   // For use with PostRAScheduling: get the anti-dependence breaking that should
   // be performed before post-RA scheduling.
diff --git a/llvm/lib/CodeGen/MachinePipeliner.cpp b/llvm/lib/CodeGen/MachinePipeliner.cpp
index 9305040678324..4f7d9d070cee6 100644
--- a/llvm/lib/CodeGen/MachinePipeliner.cpp
+++ b/llvm/lib/CodeGen/MachinePipeliner.cpp
@@ -919,7 +919,8 @@ void SwingSchedulerDAG::updatePhiDependences() {
           if (!MI->isPHI()) {
             SDep Dep(SU, SDep::Data, Reg);
             Dep.setLatency(0);
-            ST.adjustSchedDependency(SU, 0, &I, MO.getOperandNo(), Dep);
+            ST.adjustSchedDependency(SU, 0, &I, MO.getOperandNo(), Dep,
+                                     &SchedModel);
             I.addPred(Dep);
           } else {
             HasPhiUse = Reg;
diff --git a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
index 51ede7992af53..c848ce4af86cc 100644
--- a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -282,7 +282,7 @@ void ScheduleDAGInstrs::addPhysRegDataDeps(SUnit *SU, unsigned OperIdx) {
       } else {
         Dep.setLatency(0);
       }
-      ST.adjustSchedDependency(SU, OperIdx, UseSU, UseOpIdx, Dep);
+      ST.adjustSchedDependency(SU, OperIdx, UseSU, UseOpIdx, Dep, &SchedModel);
       UseSU->addPred(Dep);
     }
   }
@@ -323,7 +323,8 @@ void ScheduleDAGInstrs::addPhysRegDeps(SUnit *SU, unsigned OperIdx) {
           Dep.setLatency(
               SchedModel.computeOutputLatency(MI, OperIdx, DefInstr));
         }
-        ST.adjustSchedDependency(SU, OperIdx, DefSU, I->OpIdx, Dep);
+        ST.adjustSchedDependency(SU, OperIdx, DefSU, I->OpIdx, Dep,
+                                 &SchedModel);
         DefSU->addPred(Dep);
       }
     }
@@ -453,7 +454,8 @@ void ScheduleDAGInstrs::addVRegDefDeps(SUnit *SU, unsigned OperIdx) {
         SDep Dep(SU, SDep::Data, Reg);
         Dep.setLatency(SchedModel.computeOperandLatency(MI, OperIdx, Use,
                                                         I->OperandIndex));
-        ST.adjustSchedDependency(SU, OperIdx, UseSU, I->OperandIndex, Dep);
+        ST.adjustSchedDependency(SU, OperIdx, UseSU, I->OperandIndex, Dep,
+                                 &SchedModel);
         UseSU->addPred(Dep);
       }
 
diff --git a/llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp b/llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
index 862fce6959f1a..379456e3a9b6c 100644
--- a/llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
@@ -512,7 +512,7 @@ void ScheduleDAGSDNodes::AddSchedEdges() {
         Dep.setLatency(OpLatency);
         if (!isChain && !UnitLatencies) {
           computeOperandLatency(OpN, N, i, Dep);
-          ST.adjustSchedDependency(OpSU, DefIdx, &SU, i, Dep);
+          ST.adjustSchedDependency(OpSU, DefIdx, &SU, i, Dep, nullptr);
         }
 
         if (!SU.addPred(Dep) && !Dep.isCtrl() && OpSU->NumRegDefsLeft > 1) {
diff --git a/llvm/lib/Target/AArch64/AArch64Subtarget.cpp b/llvm/lib/Target/AArch64/AArch64Subtarget.cpp
index bb268b2ba926c..9eb1ecbb03158 100644
--- a/llvm/lib/Target/AArch64/AArch64Subtarget.cpp
+++ b/llvm/lib/Target/AArch64/AArch64Subtarget.cpp
@@ -472,6 +472,45 @@ void AArch64Subtarget::overrideSchedPolicy(MachineSchedPolicy &Policy,
   Policy.DisableLatencyHeuristic = DisableLatencySchedHeuristic;
 }
 
+void AArch64Subtarget::adjustSchedDependency(
+    SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx, SDep &Dep,
+    const TargetSchedModel *SchedModel) const {
+  if (!SchedModel || Dep.getKind() != SDep::Kind::Data || !Dep.getReg() ||
+      !Def->isInstr() || !Use->isInstr() ||
+      (Def->getInstr()->getOpcode() != TargetOpcode::BUNDLE &&
+       Use->getInstr()->getOpcode() != TargetOpcode::BUNDLE))
+    return;
+
+  // If the Def is a BUNDLE, find the last instruction in the bundle that defs
+  // the register.
+  const MachineInstr *DefMI = Def->getInstr();
+  if (DefMI->getOpcode() == TargetOpcode::BUNDLE) {
+    Register Reg = DefMI->getOperand(DefOpIdx).getReg();
+    for (const auto &Op : const_mi_bundle_ops(*DefMI)) {
+      if (Op.isReg() && Op.isDef() && Op.getReg() == Reg) {
+        DefMI = Op.getParent();
+        DefOpIdx = Op.getOperandNo();
+      }
+    }
+  }
+
+  // If the Use is a BUNDLE, find the first instruction that uses the Reg.
+  const MachineInstr *UseMI = Use->getInstr();
+  if (UseMI->getOpcode() == TargetOpcode::BUNDLE) {
+    Register Reg = UseMI->getOperand(UseOpIdx).getReg();
+    for (const auto &Op : const_mi_bundle_ops(*UseMI)) {
+      if (Op.isReg() && Op.isUse() && Op.getReg() == Reg) {
+        UseMI = Op.getParent();
+        UseOpIdx = Op.getOperandNo();
+        break;
+      }
+    }
+  }
+
+  Dep.setLatency(
+      SchedModel->computeOperandLatency(DefMI, DefOpIdx, UseMI, UseOpIdx));
+}
+
 bool AArch64Subtarget::enableEarlyIfConversion() const {
   return EnableEarlyIfConvert;
 }
diff --git a/llvm/lib/Target/AArch64/AArch64Subtarget.h b/llvm/lib/Target/AArch64/AArch64Subtarget.h
index 95bef7a76bcab..c5ebd4c6cc615 100644
--- a/llvm/lib/Target/AArch64/AArch64Subtarget.h
+++ b/llvm/lib/Target/AArch64/AArch64Subtarget.h
@@ -354,6 +354,9 @@ class AArch64Subtarget final : public AArch64GenSubtargetInfo {
 
   void overrideSchedPolicy(MachineSchedPolicy &Policy,
                            unsigned NumRegionInstrs) const override;
+  void adjustSchedDependency(SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx,
+                             SDep &Dep,
+                             const TargetSchedModel *SchedModel) const override;
 
   bool enableEarlyIfConversion() const override;
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
index 8f0eae362ecae..36e453f04426d 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -860,8 +860,9 @@ unsigned GCNSubtarget::getMaxNumVGPRs(const MachineFunction &MF) const {
   return getBaseMaxNumVGPRs(F, MFI.getWavesPerEU());
 }
 
-void GCNSubtarget::adjustSchedDependency(SUnit *Def, int DefOpIdx, SUnit *Use,
-                                         int UseOpIdx, SDep &Dep) const {
+void GCNSubtarget::adjustSchedDependency(
+    SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx, SDep &Dep,
+    const TargetSchedModel *SchedModel) const {
   if (Dep.getKind() != SDep::Kind::Data || !Dep.getReg() ||
       !Def->isInstr() || !Use->isInstr())
     return;
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index 04ff53a6647bd..8a4a46ce50d1d 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -1500,7 +1500,8 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   }
 
   void adjustSchedDependency(SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx,
-                             SDep &Dep) const override;
+                             SDep &Dep,
+                             const TargetSchedModel *SchedModel) const override;
 
   // \returns true if it's beneficial on this subtarget for the scheduler to
   // cluster stores as well as loads.
diff --git a/llvm/lib/Target/Hexagon/HexagonSubtarget.cpp b/llvm/lib/Target/Hexagon/HexagonSubtarget.cpp
index 2d320e6b0cad7..da8ab5c4b21bb 100644
--- a/llvm/lib/Target/Hexagon/HexagonSubtarget.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonSubtarget.cpp
@@ -437,9 +437,9 @@ bool HexagonSubtarget::useAA() const {
 
 /// Perform target specific adjustments to the latency of a schedule
 /// dependency.
-void HexagonSubtarget::adjustSchedDependency(SUnit *Src, int SrcOpIdx,
-                                             SUnit *Dst, int DstOpIdx,
-                                             SDep &Dep) const {
+void HexagonSubtarget::adjustSchedDependency(
+    SUnit *Src, int SrcOpIdx, SUnit *Dst, int DstOpIdx, SDep &Dep,
+    const TargetSchedModel *SchedModel) const {
   if (!Src->isInstr() || !Dst->isInstr())
     return;
 
diff --git a/llvm/lib/Target/Hexagon/HexagonSubtarget.h b/llvm/lib/Target/Hexagon/HexagonSubtarget.h
index e56007ee21e75..da3f8f56ccc4a 100644
--- a/llvm/lib/Target/Hexagon/HexagonSubtarget.h
+++ b/llvm/lib/Target/Hexagon/HexagonSubtarget.h
@@ -308,7 +308,8 @@ class HexagonSubtarget : public HexagonGenSubtargetInfo {
   /// Perform target specific adjustments to the latency of a schedule
   /// dependency.
   void adjustSchedDependency(SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx,
-                             SDep &Dep) const override;
+                             SDep &Dep,
+                             const TargetSchedModel *SchedModel) const override;
 
   unsigned getVectorLength() const {
     assert(useHVXOps());
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
index aac43f7b960fe..1751f99b3e92c 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
@@ -190,8 +190,8 @@ define <vscale x 4 x double> @mul_add_rot_mull(<vscale x 4 x double> %a, <vscale
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    uzp1 z24.d, z2.d, z3.d
 ; CHECK-NEXT:    uzp2 z25.d, z0.d, z1.d
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uzp2 z2.d, z2.d, z3.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uzp1 z0.d, z0.d, z1.d
 ; CHECK-NEXT:    uzp1 z26.d, z6.d, z7.d
 ; CHECK-NEXT:    fmul z1.d, z24.d, z25.d
@@ -199,12 +199,12 @@ define <vscale x 4 x double> @mul_add_rot_mull(<vscale x 4 x double> %a, <vscale
 ; CHECK-NEXT:    uzp2 z25.d, z4.d, z5.d
 ; CHECK-NEXT:    uzp1 z4.d, z4.d, z5.d
 ; CHECK-NEXT:    uzp2 z5.d, z6.d, z7.d
-; CHECK-NEXT:    fmla z1.d, p0/m, z2.d, z0.d
 ; CHECK-NEXT:    fmla z3.d, p0/m, z26.d, z25.d
+; CHECK-NEXT:    fmla z1.d, p0/m, z2.d, z0.d
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    fmla z2.d, p0/m, z5.d, z4.d
-; CHECK-NEXT:    fnmls z2.d, p0/m, z24.d, z0.d
 ; CHECK-NEXT:    fmla z1.d, p0/m, z26.d, z4.d
+; CHECK-NEXT:    fnmls z2.d, p0/m, z24.d, z0.d
 ; CHECK-NEXT:    fmls z1.d, p0/m, z5.d, z25.d
 ; CHECK-NEXT:    zip1 z0.d, z2.d, z1.d
 ; CHECK-NEXT:    zip2 z1.d, z2.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll b/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
index f8cba4ce4b63b..ab15bf564ec42 100644
--- a/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
+++ b/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
@@ -1145,8 +1145,8 @@ define <vscale x 4 x i64> @fshl_rot_illegal_i64(<vscale x 4 x i64> %a, <vscale x
 ; CHECK-NEXT:    subr z3.d, z3.d, #0 // =0x0
 ; CHECK-NEXT:    and z4.d, z4.d, #0x3f
 ; CHECK-NEXT:    and z2.d, z2.d, #0x3f
-; CHECK-NEXT:    and z3.d, z3.d, #0x3f
 ; CHECK-NEXT:    and z5.d, z5.d, #0x3f
+; CHECK-NEXT:    and z3.d, z3.d, #0x3f
 ; CHECK-NEXT:    lslr z4.d, p0/m, z4.d, z0.d
 ; CHECK-NEXT:    lsr z0.d, p0/m, z0.d, z2.d
 ; CHECK-NEXT:    movprfx z2, z1
diff --git a/llvm/test/CodeGen/AArch64/misched-bundle.mir b/llvm/test/CodeGen/AArch64/misched-bundle.mir
index a947c04a4229e..9adcd2904a250 100644
--- a/llvm/test/CodeGen/AArch64/misched-bundle.mir
+++ b/llvm/test/CodeGen/AArch64/misched-bundle.mir
@@ -46,9 +46,9 @@
 # CHECK-NEXT:   # rdefs left       : 0
 # CHECK-NEXT:   Latency            : 3
 # CHECK-NEXT:   Depth              : 0
-# CHECK-NEXT:   Height             : 0
+# CHECK-NEXT:   Height             : 7
 # CHECK-NEXT:   Successors:
-# CHECK-NEXT:     SU(7): Data Latency=0 Reg=$z3
+# CHECK-NEXT:     SU(7): Data Latency=3 Reg=$z3
 # CHECK-NEXT:     SU(9): Ord  Latency=0 Memory
 # CHECK-NEXT:     SU(8): Ord  Latency=0 Memory
 # CHECK-NEXT:   Single Issue       : false;
@@ -58,9 +58,9 @@
 # CHECK-NEXT:   # rdefs left       : 0
 # CHECK-NEXT:   Latency            : 3
 # CHECK-NEXT:   Depth              : 0
-# CHECK-NEXT:   Height             : 0
+# CHECK-NEXT:   Height             : 7
 # CHECK-NEXT:   Successors:
-# CHECK-NEXT:     SU(7): Data Latency=0 Reg=$z4
+# CHECK-NEXT:     SU(7): Data Latency=3 Reg=$z4
 # CHECK-NEXT:     SU(9): Ord  Latency=0 Memory
 # CHECK-NEXT:     SU(8): Ord  Latency=0 Memory
 # CHECK-NEXT:   Single Issue       : false;
@@ -70,9 +70,9 @@
 # CHECK-NEXT:   # rdefs left       : 0
 # CHECK-NEXT:   Latency            : 3
 # CHECK-NEXT:   Depth              : 0
-# CHECK-NEXT:   Height             : 0
+# CHECK-NEXT:   Height             : 7
 # CHECK-NEXT:   Successors:
-# CHECK-NEXT:     SU(7): Data Latency=0 Reg=$z5
+# CHECK-NEXT:     SU(7): Data Latency=3 Reg=$z5
 # CHECK-NEXT:     SU(9): Ord  Latency=0 Memory
 # CHECK-NEXT:     SU(8): Ord  Latency=0 Memory
 # CHECK-NEXT:   Single Issue       : false;
@@ -98,15 +98,15 @@
 # CHECK-NEXT:   # rdefs left       : 0
 # CHECK-NEXT:   Latency            : 1
 # CHECK-NEXT:   Depth              : 3
-# CHECK-NEXT:   Height             : 0
+# CHECK-NEXT:   Height             : 4
 # CHECK-NEXT:   Predecessors:
 # CHECK-NEXT:     SU(6): Anti Latency=0
-# CHECK-NEXT:     SU(5): Data Latency=0 Reg=$z5
-# CHECK-NEXT:     SU(4): Data Latency=0 Reg=$z4
-# CHECK-NEXT:     SU(3): Data Latency=0 Reg=$z3
+# CHECK-NEXT:     SU(5): Data Latency=3 Reg=$z5
+# CHECK-NEXT:     SU(4): Data Latency=3 Reg=$z4
+# CHECK-NEXT:     SU(3): Data Latency=3 Reg=$z3
 # CHECK-NEXT:     SU(1): Out  Latency=1
 # CHECK-NEXT:   Successors:
-# CHECK-NEXT:     SU(9): Data Latency=0 Reg=$z1
+# CHECK-NEXT:     SU(9): Data Latency=4 Reg=$z1
 # CHECK-NEXT:   Single Issue       : false;
 # CHECK-NEXT: SU(8):   ST1H killed renamable $z0, renamable $p0, renamable $x0, renamable $x10 :: (store unknown-size, align 1)
 # CHECK-NEXT:   # preds left       : 7
@@ -135,7 +135,7 @@
 # CHECK-NEXT:   Height             : 0
 # CHECK-NEXT:   Predecessors:
 # CHECK-NEXT:     SU(8): Ord  Latency=0 Memory
-# CHECK-NEXT:     SU(7): Data Latency=0 Reg=$z1
+# CHECK-NEXT:     SU(7): Data Latency=4 Reg=$z1
 # CHECK-NEXT:     SU(5): Ord  Latency=0 Memory
 # CHECK-NEXT:     SU(4): Ord  Latency=0 Memory
 # CHECK-NEXT:     SU(3): Ord  Latency=0 Memory
@@ -159,24 +159,24 @@ body:             |
   bb.0.entry:
     liveins: $p0, $x0, $x1, $x2, $x10, $x11, $x12, $x13
 
+
     ; CHECK-LABEL: name: test
     ; CHECK: liveins: $p0, $x0, $x1, $x2, $x10, $x11, $x12, $x13
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: renamable $z0 = LD1H renamable $p0, renamable $x1, renamable $x10 :: (load unknown-size, align 1)
     ; CHECK-NEXT: renamable $z1 = LD1H renamable $p0, renamable $x2, renamable $x10 :: (load unknown-size, align 1)
     ; CHECK-NEXT: renamable $z2 = LD1H renamable $p0, renamable $x0, renamable $x10 :: (load unknown-size, align 1)
-    ; CHECK-NEXT: $z0 = FMAD_ZPmZZ_H renamable $p0, killed $z0, renamable $z1, killed renamable $z2
     ; CHECK-NEXT: renamable $z3 = LD1H renamable $p0, renamable $x11, renamable $x10 :: (load unknown-size, align 1)
     ; CHECK-NEXT: renamable $z4 = LD1H renamable $p0, renamable $x12, renamable $x10 :: (load unknown-size, align 1)
     ; CHECK-NEXT: renamable $z5 = LD1H renamable $p0, renamable $x13, renamable $x10 :: (load unknown-size, align 1)
-    ; CHECK-NEXT: ST1H killed renamable $z0, renamable $p0, renamable $x0, renamable $x10 :: (store unknown-size, align 1)
-    ; CHECK-NEXT: BUNDLE implicit-def $z1, implicit-def $q1, implicit-def $d1, implicit-def $s1, implicit-def $h1, implicit-def $b1, implicit $z5, implicit $p0, implicit $z4, implicit $z3 {
+    ; CHECK-NEXT: $z0 = FMAD_ZPmZZ_H renamable $p0, killed $z0, killed renamable $z1, killed renamable $z2
+    ; CHECK-NEXT: BUNDLE implicit-def $z1, implicit-def $q1, implicit-def $d1, implicit-def $s1, implicit-def $h1, implicit-def $b1, implicit $z5, implicit $p0, implicit killed $z4, implicit killed $z3 {
     ; CHECK-NEXT:   $z1 = MOVPRFX_ZZ $z5
-    ; CHECK-NEXT:   $z1 = FMLA_ZPmZZ_H renamable $p0, internal $z1, renamable $z4, renamable $z3
+    ; CHECK-NEXT:   $z1 = FMLA_ZPmZZ_H renamable $p0, internal killed $z1, killed renamable $z4, killed renamable $z3
     ; CHECK-NEXT: }
-    ; CHECK-NEXT: ST1H renamable $z1, renamable $p0, renamable $x13, renamable $x10 :: (store unknown-size, align 1)
+    ; CHECK-NEXT: ST1H killed renamable $z0, renamable $p0, renamable $x0, renamable $x10 :: (store unknown-size, align 1)
+    ; CHECK-NEXT: ST1H killed renamable $z1, renamable $p0, renamable $x13, renamable $x10 :: (store unknown-size, align 1)
     ; CHECK-NEXT: RET_ReallyLR
-
     renamable $z0 = LD1H renamable $p0, renamable $x1, renamable $x10 :: (load unknown-size)
     renamable $z1 = LD1H renamable $p0, renamable $x2, renamable $x10 :: (load unknown-size)
     renamable $z2 = LD1H renamable $p0, renamable $x0, renamable $x10 :: (load unknown-size)
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
index 0e3307d161729..ad482118ec0bb 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
@@ -43,7 +43,7 @@ define void @build_vector_minus2_dec32_v4i64(ptr %a) #0 {
 ; VBITS_GE_256-LABEL: build_vector_minus2_dec32_v4i64:
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
-; VBITS_GE_256-NEXT:    mov x8, #-32
+; VBITS_GE_256-NEXT:    mov x8, #-32 // =0xffffffffffffffe0
 ; VBITS_GE_256-NEXT:    index z0.d, #-2, x8
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-arith.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-arith.ll
index 64c4eea0ee5f3..6fbae7edfec0a 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-arith.ll
@@ -57,8 +57,8 @@ define void @fadd_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fadd z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fadd_v32f16:
@@ -156,8 +156,8 @@ define void @fadd_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fadd z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fadd_v16f32:
@@ -255,8 +255,8 @@ define void @fadd_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fadd z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fadd z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fadd_v8f64:
@@ -662,8 +662,8 @@ define void @fma_v32f16(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.h, p0/m, z1.h, z2.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.h, p0/m, z3.h, z4.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v32f16:
@@ -773,8 +773,8 @@ define void @fma_v16f32(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.s, p0/m, z1.s, z2.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.s, p0/m, z3.s, z4.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v16f32:
@@ -883,8 +883,8 @@ define void @fma_v8f64(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.d, p0/m, z1.d, z2.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.d, p0/m, z3.d, z4.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v8f64:
@@ -992,8 +992,8 @@ define void @fmul_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fmul z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fmul z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmul_v32f16:
@@ -1091,8 +1091,8 @@ define void @fmul_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fmul z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fmul z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmul_v16f32:
@@ -1190,8 +1190,8 @@ define void @fmul_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fmul z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fmul z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmul_v8f64:
@@ -1829,8 +1829,8 @@ define void @fsub_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fsub z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fsub z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fsub_v32f16:
@@ -1928,8 +1928,8 @@ define void @fsub_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fsub z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fsub z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fsub_v16f32:
@@ -2027,8 +2027,8 @@ define void @fsub_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fsub z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    fsub z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fsub_v8f64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-fma.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-fma.ll
index d3713b989301b..e1ec5ee5f6137 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-fma.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-fma.ll
@@ -66,8 +66,8 @@ define void @fma_v32f16(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.h, p0/m, z1.h, z2.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.h, p0/m, z3.h, z4.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v32f16:
@@ -183,8 +183,8 @@ define void @fma_v16f32(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.s, p0/m, z1.s, z2.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.s, p0/m, z3.s, z4.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v16f32:
@@ -299,8 +299,8 @@ define void @fma_v8f64(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    fmad z0.d, p0/m, z1.d, z2.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z5
 ; VBITS_GE_256-NEXT:    fmla z1.d, p0/m, z3.d, z4.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fma_v8f64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-minmax.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-minmax.ll
index d20f083066a83..de60deeafaf32 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-minmax.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-minmax.ll
@@ -57,8 +57,8 @@ define void @fmaxnm_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmaxnm z0.h, p0/m, z0.h, z1.h
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmaxnm z1.h, p0/m, z1.h, z3.h
-; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmaxnm_v32f16:
@@ -156,8 +156,8 @@ define void @fmaxnm_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmaxnm z0.s, p0/m, z0.s, z1.s
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmaxnm z1.s, p0/m, z1.s, z3.s
-; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmaxnm_v16f32:
@@ -255,8 +255,8 @@ define void @fmaxnm_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmaxnm z0.d, p0/m, z0.d, z1.d
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmaxnm z1.d, p0/m, z1.d, z3.d
-; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmaxnm_v8f64:
@@ -358,8 +358,8 @@ define void @fminnm_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fminnm z0.h, p0/m, z0.h, z1.h
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fminnm z1.h, p0/m, z1.h, z3.h
-; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fminnm_v32f16:
@@ -457,8 +457,8 @@ define void @fminnm_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fminnm z0.s, p0/m, z0.s, z1.s
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fminnm z1.s, p0/m, z1.s, z3.s
-; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fminnm_v16f32:
@@ -556,8 +556,8 @@ define void @fminnm_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fminnm z0.d, p0/m, z0.d, z1.d
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fminnm z1.d, p0/m, z1.d, z3.d
-; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fminnm_v8f64:
@@ -659,8 +659,8 @@ define void @fmax_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmax z0.h, p0/m, z0.h, z1.h
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmax z1.h, p0/m, z1.h, z3.h
-; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmax_v32f16:
@@ -758,8 +758,8 @@ define void @fmax_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmax z0.s, p0/m, z0.s, z1.s
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmax z1.s, p0/m, z1.s, z3.s
-; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmax_v16f32:
@@ -857,8 +857,8 @@ define void @fmax_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmax z0.d, p0/m, z0.d, z1.d
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmax z1.d, p0/m, z1.d, z3.d
-; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmax_v8f64:
@@ -960,8 +960,8 @@ define void @fmin_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmin z0.h, p0/m, z0.h, z1.h
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmin z1.h, p0/m, z1.h, z3.h
-; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_EQ_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmin_v32f16:
@@ -1059,8 +1059,8 @@ define void @fmin_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmin z0.s, p0/m, z0.s, z1.s
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmin z1.s, p0/m, z1.s, z3.s
-; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_EQ_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmin_v16f32:
@@ -1158,8 +1158,8 @@ define void @fmin_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_EQ_256-NEXT:    fmin z0.d, p0/m, z0.d, z1.d
 ; VBITS_EQ_256-NEXT:    movprfx z1, z2
 ; VBITS_EQ_256-NEXT:    fmin z1.d, p0/m, z1.d, z3.d
-; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_EQ_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_EQ_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fmin_v8f64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
index 8f2aeade48a0f..b45a62f4e7581 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
@@ -458,8 +458,8 @@ define void @mul_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mul z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    mul z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: mul_v64i8:
@@ -557,8 +557,8 @@ define void @mul_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    mul z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: mul_v32i16:
@@ -656,8 +656,8 @@ define void @mul_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    mul z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: mul_v16i32:
@@ -761,8 +761,8 @@ define void @mul_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    mul z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: mul_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-minmax.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-minmax.ll
index 296e4818b1cd7..4926684ddc2de 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-minmax.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-minmax.ll
@@ -57,8 +57,8 @@ define void @smax_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smax z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smax z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smax_v64i8:
@@ -156,8 +156,8 @@ define void @smax_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smax z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smax z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smax_v32i16:
@@ -255,8 +255,8 @@ define void @smax_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smax z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smax_v16i32:
@@ -362,8 +362,8 @@ define void @smax_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smax z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smax z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smax_v8i64:
@@ -465,8 +465,8 @@ define void @smin_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smin z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smin z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smin_v64i8:
@@ -564,8 +564,8 @@ define void @smin_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smin z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smin z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smin_v32i16:
@@ -663,8 +663,8 @@ define void @smin_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smin z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smin_v16i32:
@@ -770,8 +770,8 @@ define void @smin_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smin z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smin z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smin_v8i64:
@@ -873,8 +873,8 @@ define void @umax_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umax z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umax z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umax_v64i8:
@@ -972,8 +972,8 @@ define void @umax_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umax z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umax_v32i16:
@@ -1071,8 +1071,8 @@ define void @umax_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umax z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umax_v16i32:
@@ -1178,8 +1178,8 @@ define void @umax_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umax z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umax_v8i64:
@@ -1281,8 +1281,8 @@ define void @umin_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umin z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umin z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umin_v64i8:
@@ -1380,8 +1380,8 @@ define void @umin_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umin z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umin_v32i16:
@@ -1479,8 +1479,8 @@ define void @umin_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umin z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umin_v16i32:
@@ -1586,8 +1586,8 @@ define void @umin_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umin z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umin z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umin_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-mulh.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-mulh.ll
index 995da2be0cc17..331aab17a4cea 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-mulh.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-mulh.ll
@@ -82,8 +82,8 @@ define void @smulh_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smulh z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smulh z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smulh_v64i8:
@@ -215,8 +215,8 @@ define void @smulh_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smulh z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smulh z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smulh_v32i16:
@@ -346,8 +346,8 @@ define void @smulh_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smulh z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smulh z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smulh_v16i32:
@@ -479,8 +479,8 @@ define void @smulh_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    smulh z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    smulh z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: smulh_v8i64:
@@ -615,8 +615,8 @@ define void @umulh_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umulh z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umulh z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umulh_v64i8:
@@ -749,8 +749,8 @@ define void @umulh_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umulh z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umulh z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umulh_v32i16:
@@ -882,8 +882,8 @@ define void @umulh_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umulh z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umulh z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umulh_v16i32:
@@ -1013,8 +1013,8 @@ define void @umulh_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    umulh z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    umulh z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: umulh_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
index 45960617f1d93..38444d83c1d7e 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
@@ -533,9 +533,9 @@ define void @srem_v128i16(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <2 x i32> @srem_v2i32(<2 x i32> %op1, <2 x i32> %op2) vscale_range(1,0) #0 {
 ; CHECK-LABEL: srem_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
-; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls v0.2s, v2.2s, v1.2s
@@ -549,9 +549,9 @@ define <2 x i32> @srem_v2i32(<2 x i32> %op1, <2 x i32> %op2) vscale_range(1,0) #
 define <4 x i32> @srem_v4i32(<4 x i32> %op1, <4 x i32> %op2) vscale_range(1,0) #0 {
 ; CHECK-LABEL: srem_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
-; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls v0.4s, v2.4s, v1.4s
@@ -582,25 +582,25 @@ define void @srem_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @srem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: srem_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
+; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
-; VBITS_GE_128-NEXT:    movprfx z4, z1
-; VBITS_GE_128-NEXT:    sdiv z4.s, p0/m, z4.s, z0.s
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    sdiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
+; VBITS_GE_128-NEXT:    movprfx z4, z1
+; VBITS_GE_128-NEXT:    sdiv z4.s, p0/m, z4.s, z0.s
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    sdiv z7.s, p0/m, z7.s, z6.s
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    sdiv z18.s, p0/m, z18.s, z17.s
-; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
-; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    sdiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
 ; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
-; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
+; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
+; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
 ; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: srem_v16i32:
@@ -609,17 +609,17 @@ define void @srem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x1, x8, lsl #2]
-; VBITS_GE_256-NEXT:    movprfx z2, z0
-; VBITS_GE_256-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    ld1w { z3.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1w { z4.s }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    movprfx z2, z0
+; VBITS_GE_256-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z5, z3
 ; VBITS_GE_256-NEXT:    sdiv z5.s, p0/m, z5.s, z4.s
 ; VBITS_GE_256-NEXT:    mls z0.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z3
 ; VBITS_GE_256-NEXT:    mls z1.s, p0/m, z5.s, z4.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: srem_v16i32:
@@ -681,8 +681,8 @@ define <1 x i64> @srem_v1i64(<1 x i64> %op1, <1 x i64> %op2) vscale_range(1,0) #
 ; CHECK-LABEL: srem_v1i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl1
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -698,8 +698,8 @@ define <2 x i64> @srem_v2i64(<2 x i64> %op1, <2 x i64> %op2) vscale_range(1,0) #
 ; CHECK-LABEL: srem_v2i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl2
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -733,16 +733,16 @@ define void @srem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
-; VBITS_GE_128-NEXT:    movprfx z4, z1
-; VBITS_GE_128-NEXT:    sdiv z4.d, p0/m, z4.d, z0.d
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    sdiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
+; VBITS_GE_128-NEXT:    movprfx z4, z1
+; VBITS_GE_128-NEXT:    sdiv z4.d, p0/m, z4.d, z0.d
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    sdiv z7.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    sdiv z18.d, p0/m, z18.d, z17.d
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    sdiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
 ; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    msb z0.d, p0/m, z4.d, z1.d
@@ -758,17 +758,17 @@ define void @srem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    movprfx z2, z0
-; VBITS_GE_256-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    movprfx z2, z0
+; VBITS_GE_256-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z5, z3
 ; VBITS_GE_256-NEXT:    sdiv z5.d, p0/m, z5.d, z4.d
 ; VBITS_GE_256-NEXT:    mls z0.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z3
 ; VBITS_GE_256-NEXT:    mls z1.d, p0/m, z5.d, z4.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: srem_v8i64:
@@ -1351,9 +1351,9 @@ define void @urem_v128i16(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <2 x i32> @urem_v2i32(<2 x i32> %op1, <2 x i32> %op2) vscale_range(1,0) #0 {
 ; CHECK-LABEL: urem_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
-; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls v0.2s, v2.2s, v1.2s
@@ -1367,9 +1367,9 @@ define <2 x i32> @urem_v2i32(<2 x i32> %op1, <2 x i32> %op2) vscale_range(1,0) #
 define <4 x i32> @urem_v4i32(<4 x i32> %op1, <4 x i32> %op2) vscale_range(1,0) #0 {
 ; CHECK-LABEL: urem_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
-; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls v0.4s, v2.4s, v1.4s
@@ -1400,25 +1400,25 @@ define void @urem_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @urem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: urem_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
+; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
-; VBITS_GE_128-NEXT:    movprfx z4, z1
-; VBITS_GE_128-NEXT:    udiv z4.s, p0/m, z4.s, z0.s
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    udiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
+; VBITS_GE_128-NEXT:    movprfx z4, z1
+; VBITS_GE_128-NEXT:    udiv z4.s, p0/m, z4.s, z0.s
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    udiv z7.s, p0/m, z7.s, z6.s
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    udiv z18.s, p0/m, z18.s, z17.s
-; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
-; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    udiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
 ; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
-; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
+; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
+; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
 ; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: urem_v16i32:
@@ -1427,17 +1427,17 @@ define void @urem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x1, x8, lsl #2]
-; VBITS_GE_256-NEXT:    movprfx z2, z0
-; VBITS_GE_256-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    ld1w { z3.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1w { z4.s }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    movprfx z2, z0
+; VBITS_GE_256-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z5, z3
 ; VBITS_GE_256-NEXT:    udiv z5.s, p0/m, z5.s, z4.s
 ; VBITS_GE_256-NEXT:    mls z0.s, p0/m, z2.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z3
 ; VBITS_GE_256-NEXT:    mls z1.s, p0/m, z5.s, z4.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: urem_v16i32:
@@ -1499,8 +1499,8 @@ define <1 x i64> @urem_v1i64(<1 x i64> %op1, <1 x i64> %op2) vscale_range(1,0) #
 ; CHECK-LABEL: urem_v1i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl1
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -1516,8 +1516,8 @@ define <2 x i64> @urem_v2i64(<2 x i64> %op1, <2 x i64> %op2) vscale_range(1,0) #
 ; CHECK-LABEL: urem_v2i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl2
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -1551,16 +1551,16 @@ define void @urem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
-; VBITS_GE_128-NEXT:    movprfx z4, z1
-; VBITS_GE_128-NEXT:    udiv z4.d, p0/m, z4.d, z0.d
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    udiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
+; VBITS_GE_128-NEXT:    movprfx z4, z1
+; VBITS_GE_128-NEXT:    udiv z4.d, p0/m, z4.d, z0.d
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    udiv z7.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    udiv z18.d, p0/m, z18.d, z17.d
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    udiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
 ; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    msb z0.d, p0/m, z4.d, z1.d
@@ -1576,17 +1576,17 @@ define void @urem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    movprfx z2, z0
-; VBITS_GE_256-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    movprfx z2, z0
+; VBITS_GE_256-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z5, z3
 ; VBITS_GE_256-NEXT:    udiv z5.d, p0/m, z5.d, z4.d
 ; VBITS_GE_256-NEXT:    mls z0.d, p0/m, z2.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z3
 ; VBITS_GE_256-NEXT:    mls z1.d, p0/m, z5.d, z4.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: urem_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-shifts.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-shifts.ll
index 7184cc9705973..0fa8c8f50e29c 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-shifts.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-shifts.ll
@@ -59,8 +59,8 @@ define void @ashr_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    asr z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    asr z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ashr_v64i8:
@@ -160,8 +160,8 @@ define void @ashr_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    asr z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    asr z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ashr_v32i16:
@@ -261,8 +261,8 @@ define void @ashr_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    asr z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    asr z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ashr_v16i32:
@@ -362,8 +362,8 @@ define void @ashr_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    asr z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    asr z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ashr_v8i64:
@@ -467,8 +467,8 @@ define void @lshr_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsr z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsr z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: lshr_v64i8:
@@ -568,8 +568,8 @@ define void @lshr_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsr z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsr z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: lshr_v32i16:
@@ -669,8 +669,8 @@ define void @lshr_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsr z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsr z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: lshr_v16i32:
@@ -770,8 +770,8 @@ define void @lshr_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsr z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsr z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: lshr_v8i64:
@@ -873,8 +873,8 @@ define void @shl_v64i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsl z0.b, p0/m, z0.b, z1.b
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsl z1.b, p0/m, z1.b, z3.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: shl_v64i8:
@@ -972,8 +972,8 @@ define void @shl_v32i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsl z0.h, p0/m, z0.h, z1.h
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsl z1.h, p0/m, z1.h, z3.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: shl_v32i16:
@@ -1071,8 +1071,8 @@ define void @shl_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsl z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsl z1.s, p0/m, z1.s, z3.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: shl_v16i32:
@@ -1170,8 +1170,8 @@ define void @shl_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    lsl z0.d, p0/m, z0.d, z1.d
 ; VBITS_GE_256-NEXT:    movprfx z1, z2
 ; VBITS_GE_256-NEXT:    lsl z1.d, p0/m, z1.d, z3.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: shl_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fpext-load.ll b/llvm/test/CodeGen/AArch64/sve-fpext-load.ll
index 78decf5ed1e3c..aec2150124c3f 100644
--- a/llvm/test/CodeGen/AArch64/sve-fpext-load.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fpext-load.ll
@@ -46,13 +46,13 @@ define <vscale x 8 x double> @ext8_f16_f64(ptr %ptr, i64 %index) {
 ; CHECK-NEXT:    uunpkhi z1.d, z1.s
 ; CHECK-NEXT:    uunpklo z3.d, z0.s
 ; CHECK-NEXT:    uunpkhi z4.d, z0.s
+; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z2.h
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    fcvt z2.d, p0/m, z3.h
 ; CHECK-NEXT:    movprfx z3, z4
 ; CHECK-NEXT:    fcvt z3.d, p0/m, z4.h
-; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    ret
   %load = load <vscale x 8 x half>, ptr %ptr, align 4
   %load.ext = fpext <vscale x 8 x half> %load to <vscale x 8 x double>
diff --git a/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll b/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
index 43113df07c079..813f1601e809e 100644
--- a/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
@@ -74,12 +74,12 @@ define <vscale x 8 x i32> @test_signed_v8f32_v8i32(<vscale x 8 x float> %f) {
 ; CHECK-NEXT:    mov w8, #1325400063 // =0x4effffff
 ; CHECK-NEXT:    mov z6.s, #0x7fffffff
 ; CHECK-NEXT:    mov z4.s, w8
-; CHECK-NEXT:    movprfx z5, z1
-; CHECK-NEXT:    fcvtzs z5.s, p0/m, z1.s
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzs z2.s, p0/m, z0.s
+; CHECK-NEXT:    movprfx z5, z1
+; CHECK-NEXT:    fcvtzs z5.s, p0/m, z1.s
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z0.s, z4.s
 ; CHECK-NEXT:    fcmgt p4.s, p0/z, z1.s, z4.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -134,17 +134,17 @@ define <vscale x 8 x i16> @test_signed_v8f32_v8i16(<vscale x 8 x float> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #-956301312 // =0xc7000000
-; CHECK-NEXT:    movprfx z4, z0
-; CHECK-NEXT:    fcvtzs z4.s, p0/m, z0.s
+; CHECK-NEXT:    mov z5.s, #32767 // =0x7fff
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    mov w8, #65024 // =0xfe00
-; CHECK-NEXT:    mov z5.s, #32767 // =0x7fff
 ; CHECK-NEXT:    movk w8, #18175, lsl #16
 ; CHECK-NEXT:    mov z3.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.s, p0/m, z1.s
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzs z4.s, p0/m, z0.s
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z3.s
 ; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z3.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -206,13 +206,13 @@ define <vscale x 4 x i64> @test_signed_v4f32_v4i64(<vscale x 4 x float> %f) {
 ; CHECK-NEXT:    mov w8, #1593835519 // =0x5effffff
 ; CHECK-NEXT:    mov z3.d, #0x8000000000000000
 ; CHECK-NEXT:    mov z4.s, w8
-; CHECK-NEXT:    movprfx z5, z0
-; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.s
 ; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.s
+; CHECK-NEXT:    movprfx z5, z0
+; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.s
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z4.s
 ; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z4.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -283,13 +283,13 @@ define <vscale x 4 x i32> @test_signed_v4f64_v4i32(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    mov x8, #281474972516352 // =0xffffffc00000
 ; CHECK-NEXT:    mov z6.d, #0x7fffffff
 ; CHECK-NEXT:    movk x8, #16863, lsl #48
-; CHECK-NEXT:    movprfx z5, z0
-; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.d
 ; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z5, z0
+; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.d
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -324,24 +324,24 @@ define <vscale x 8 x i32> @test_signed_v8f64_v8i32(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4476578029606273024 // =0xc1e0000000000000
-; CHECK-NEXT:    movprfx z6, z1
-; CHECK-NEXT:    fcvtzs z6.d, p0/m, z1.d
+; CHECK-NEXT:    mov z26.d, #0x7fffffff
 ; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    mov x8, #281474972516352 // =0xffffffc00000
-; CHECK-NEXT:    movprfx z7, z0
-; CHECK-NEXT:    fcvtzs z7.d, p0/m, z0.d
 ; CHECK-NEXT:    movk x8, #16863, lsl #48
-; CHECK-NEXT:    mov z26.d, #0x7fffffff
-; CHECK-NEXT:    movprfx z24, z3
-; CHECK-NEXT:    fcvtzs z24.d, p0/m, z3.d
 ; CHECK-NEXT:    mov z5.d, x8
-; CHECK-NEXT:    movprfx z25, z2
-; CHECK-NEXT:    fcvtzs z25.d, p0/m, z2.d
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, z4.d
 ; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, z4.d
 ; CHECK-NEXT:    mov z4.d, #0xffffffff80000000
+; CHECK-NEXT:    movprfx z6, z1
+; CHECK-NEXT:    fcvtzs z6.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z7, z0
+; CHECK-NEXT:    fcvtzs z7.d, p0/m, z0.d
+; CHECK-NEXT:    movprfx z24, z3
+; CHECK-NEXT:    fcvtzs z24.d, p0/m, z3.d
+; CHECK-NEXT:    movprfx z25, z2
+; CHECK-NEXT:    fcvtzs z25.d, p0/m, z2.d
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z1.d, z5.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z0.d, z5.d
 ; CHECK-NEXT:    fcmgt p7.d, p0/z, z3.d, z5.d
@@ -389,17 +389,17 @@ define <vscale x 4 x i16> @test_signed_v4f64_v4i16(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4548635623644200960 // =0xc0e0000000000000
-; CHECK-NEXT:    movprfx z4, z0
-; CHECK-NEXT:    fcvtzs z4.d, p0/m, z0.d
+; CHECK-NEXT:    mov z5.d, #32767 // =0x7fff
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    mov x8, #281200098803712 // =0xffc000000000
-; CHECK-NEXT:    mov z5.d, #32767 // =0x7fff
 ; CHECK-NEXT:    movk x8, #16607, lsl #48
 ; CHECK-NEXT:    mov z3.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z0.d
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z3.d
 ; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z3.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -434,23 +434,23 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4548635623644200960 // =0xc0e0000000000000
+; CHECK-NEXT:    mov z25.d, #32767 // =0x7fff
+; CHECK-NEXT:    mov z4.d, x8
+; CHECK-NEXT:    mov x8, #281200098803712 // =0xffc000000000
+; CHECK-NEXT:    movk x8, #16607, lsl #48
+; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, z4.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, z4.d
+; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, z4.d
+; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    movprfx z5, z3
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z3.d
 ; CHECK-NEXT:    mov z4.d, x8
-; CHECK-NEXT:    mov x8, #281200098803712 // =0xffc000000000
 ; CHECK-NEXT:    movprfx z6, z2
 ; CHECK-NEXT:    fcvtzs z6.d, p0/m, z2.d
-; CHECK-NEXT:    movk x8, #16607, lsl #48
 ; CHECK-NEXT:    movprfx z7, z1
 ; CHECK-NEXT:    fcvtzs z7.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z24, z0
 ; CHECK-NEXT:    fcvtzs z24.d, p0/m, z0.d
-; CHECK-NEXT:    mov z25.d, #32767 // =0x7fff
-; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, z4.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, z4.d
-; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, z4.d
-; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, z4.d
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z3.d, z4.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z2.d, z4.d
@@ -528,12 +528,12 @@ define <vscale x 4 x i64> @test_signed_v4f64_v4i64(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    mov x8, #4890909195324358655 // =0x43dfffffffffffff
 ; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    mov z4.d, x8
-; CHECK-NEXT:    movprfx z5, z1
-; CHECK-NEXT:    fcvtzs z5.d, p0/m, z1.d
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, z2.d
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzs z2.d, p0/m, z0.d
+; CHECK-NEXT:    movprfx z5, z1
+; CHECK-NEXT:    fcvtzs z5.d, p0/m, z1.d
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    fcmgt p4.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -629,13 +629,13 @@ define <vscale x 8 x i32> @test_signed_v8f16_v8i32(<vscale x 8 x half> %f) {
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
 ; CHECK-NEXT:    mov z3.s, #0x80000000
 ; CHECK-NEXT:    mov z4.h, w8
-; CHECK-NEXT:    movprfx z5, z0
-; CHECK-NEXT:    fcvtzs z5.s, p0/m, z0.h
 ; CHECK-NEXT:    mov z6.s, #0x7fffffff
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z0.h, z2.h
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.s, p0/m, z1.h
+; CHECK-NEXT:    movprfx z5, z0
+; CHECK-NEXT:    fcvtzs z5.s, p0/m, z0.h
 ; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z4.h
 ; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z4.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -742,13 +742,13 @@ define <vscale x 4 x i64> @test_signed_v4f16_v4i64(<vscale x 4 x half> %f) {
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
 ; CHECK-NEXT:    mov z3.d, #0x8000000000000000
 ; CHECK-NEXT:    mov z4.h, w8
-; CHECK-NEXT:    movprfx z5, z0
-; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.h
 ; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z0.h, z2.h
 ; CHECK-NEXT:    movprfx z2, z1
 ; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.h
+; CHECK-NEXT:    movprfx z5, z0
+; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.h
 ; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z4.h
 ; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z4.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
diff --git a/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll b/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
index c9e06fd9f3414..c56c0b37888dc 100644
--- a/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
@@ -16,10 +16,10 @@ define <vscale x 2 x i32> @test_signed_v2f32_v2i32(<vscale x 2 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1333788671 // =0x4f7fffff
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    movprfx z2, z0
+; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.d, #0xffffffff
@@ -35,10 +35,10 @@ define <vscale x 4 x i32> @test_signed_v4f32_v4i32(<vscale x 4 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #1333788671 // =0x4f7fffff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.s, p0/m, z0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.s, p0/m, z0.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    mov z1.s, p1/m, #0 // =0x0
@@ -54,13 +54,13 @@ define <vscale x 8 x i32> @test_signed_v8f32_v8i32(<vscale x 8 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #1333788671 // =0x4f7fffff
+; CHECK-NEXT:    mov z4.s, w8
+; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fcmge p2.s, p0/z, z1.s, #0.0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.s
-; CHECK-NEXT:    mov z4.s, w8
 ; CHECK-NEXT:    movprfx z3, z1
 ; CHECK-NEXT:    fcvtzu z3.s, p0/m, z1.s
-; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
-; CHECK-NEXT:    fcmge p2.s, p0/z, z1.s, #0.0
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z0.s, z4.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -81,11 +81,11 @@ define <vscale x 4 x i16> @test_signed_v4f32_v4i16(<vscale x 4 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #65280 // =0xff00
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.s
 ; CHECK-NEXT:    movk w8, #18303, lsl #16
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    movprfx z2, z0
+; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.s, #65535 // =0xffff
@@ -101,14 +101,14 @@ define <vscale x 8 x i16> @test_signed_v8f32_v8i16(<vscale x 8 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #65280 // =0xff00
-; CHECK-NEXT:    movprfx z3, z1
-; CHECK-NEXT:    fcvtzu z3.s, p0/m, z1.s
 ; CHECK-NEXT:    movk w8, #18303, lsl #16
-; CHECK-NEXT:    movprfx z4, z0
-; CHECK-NEXT:    fcvtzu z4.s, p0/m, z0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, #0.0
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    movprfx z3, z1
+; CHECK-NEXT:    fcvtzu z3.s, p0/m, z1.s
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzu z4.s, p0/m, z0.s
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -129,10 +129,10 @@ define <vscale x 2 x i64> @test_signed_v2f32_v2i64(<vscale x 2 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1602224127 // =0x5f7fffff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    mov z1.d, p1/m, #0 // =0x0
@@ -150,13 +150,13 @@ define <vscale x 4 x i64> @test_signed_v4f32_v4i64(<vscale x 4 x float> %f) {
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-NEXT:    mov w8, #1602224127 // =0x5f7fffff
+; CHECK-NEXT:    mov z4.s, w8
+; CHECK-NEXT:    fcmge p1.s, p0/z, z2.s, #0.0
+; CHECK-NEXT:    fcmge p2.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z2.s
 ; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fcvtzu z1.d, p0/m, z3.s
-; CHECK-NEXT:    mov z4.s, w8
-; CHECK-NEXT:    fcmge p1.s, p0/z, z2.s, #0.0
-; CHECK-NEXT:    fcmge p2.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    fcmgt p3.s, p0/z, z2.s, z4.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -185,11 +185,11 @@ define <vscale x 2 x i32> @test_signed_v2f64_v2i32(<vscale x 2 x double> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.d
 ; CHECK-NEXT:    movk x8, #16879, lsl #48
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    movprfx z2, z0
+; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    mov z0.d, #0xffffffff
@@ -205,14 +205,14 @@ define <vscale x 4 x i32> @test_signed_v4f64_v4i32(<vscale x 4 x double> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
-; CHECK-NEXT:    movprfx z3, z1
-; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
 ; CHECK-NEXT:    movk x8, #16879, lsl #48
-; CHECK-NEXT:    movprfx z4, z0
-; CHECK-NEXT:    fcvtzu z4.d, p0/m, z0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    movprfx z3, z1
+; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzu z4.d, p0/m, z0.d
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -240,20 +240,20 @@ define <vscale x 8 x i32> @test_signed_v8f64_v8i32(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
+; CHECK-NEXT:    movk x8, #16879, lsl #48
+; CHECK-NEXT:    mov z4.d, x8
+; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
+; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, #0.0
+; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, #0.0
 ; CHECK-NEXT:    movprfx z5, z1
 ; CHECK-NEXT:    fcvtzu z5.d, p0/m, z1.d
-; CHECK-NEXT:    movk x8, #16879, lsl #48
 ; CHECK-NEXT:    movprfx z6, z0
 ; CHECK-NEXT:    fcvtzu z6.d, p0/m, z0.d
 ; CHECK-NEXT:    movprfx z7, z3
 ; CHECK-NEXT:    fcvtzu z7.d, p0/m, z3.d
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z24, z2
 ; CHECK-NEXT:    fcvtzu z24.d, p0/m, z2.d
-; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
-; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
-; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, #0.0
-; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, #0.0
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    mov z0.d, #0xffffffff
@@ -288,14 +288,14 @@ define <vscale x 4 x i16> @test_signed_v4f64_v4i16(<vscale x 4 x double> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281337537757184 // =0xffe000000000
-; CHECK-NEXT:    movprfx z3, z1
-; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
 ; CHECK-NEXT:    movk x8, #16623, lsl #48
-; CHECK-NEXT:    movprfx z4, z0
-; CHECK-NEXT:    fcvtzu z4.d, p0/m, z0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    movprfx z3, z1
+; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzu z4.d, p0/m, z0.d
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -323,20 +323,20 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281337537757184 // =0xffe000000000
+; CHECK-NEXT:    movk x8, #16623, lsl #48
+; CHECK-NEXT:    mov z4.d, x8
+; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, #0.0
+; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, #0.0
+; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, #0.0
+; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, #0.0
 ; CHECK-NEXT:    movprfx z5, z3
 ; CHECK-NEXT:    fcvtzu z5.d, p0/m, z3.d
-; CHECK-NEXT:    movk x8, #16623, lsl #48
 ; CHECK-NEXT:    movprfx z6, z2
 ; CHECK-NEXT:    fcvtzu z6.d, p0/m, z2.d
 ; CHECK-NEXT:    movprfx z7, z1
 ; CHECK-NEXT:    fcvtzu z7.d, p0/m, z1.d
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z24, z0
 ; CHECK-NEXT:    fcvtzu z24.d, p0/m, z0.d
-; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, #0.0
-; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, #0.0
-; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, #0.0
-; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, #0.0
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z3.d, z4.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z2.d, z4.d
 ; CHECK-NEXT:    mov z2.d, #65535 // =0xffff
@@ -372,10 +372,10 @@ define <vscale x 2 x i64> @test_signed_v2f64_v2i64(<vscale x 2 x double> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #4895412794951729151 // =0x43efffffffffffff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.d, p0/z, z0.d, z2.d
 ; CHECK-NEXT:    mov z1.d, p1/m, #0 // =0x0
@@ -391,13 +391,13 @@ define <vscale x 4 x i64> @test_signed_v4f64_v4i64(<vscale x 4 x double> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #4895412794951729151 // =0x43efffffffffffff
+; CHECK-NEXT:    mov z4.d, x8
+; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    fcmge p2.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.d
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z3, z1
 ; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
-; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, #0.0
-; CHECK-NEXT:    fcmge p2.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmgt p3.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -429,10 +429,10 @@ define <vscale x 2 x i32> @test_signed_v2f16_v2i32(<vscale x 2 x half> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    movprfx z2, z0
+; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    mov z0.d, #0xffffffff
@@ -448,10 +448,10 @@ define <vscale x 4 x i32> @test_signed_v4f16_v4i32(<vscale x 4 x half> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.s, p0/m, z0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.s, p0/m, z0.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z0.h, z2.h
 ; CHECK-NEXT:    mov z1.s, p1/m, #0 // =0x0
@@ -469,13 +469,13 @@ define <vscale x 8 x i32> @test_signed_v8f16_v8i32(<vscale x 8 x half> %f) {
 ; CHECK-NEXT:    uunpklo z2.s, z0.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
+; CHECK-NEXT:    mov z4.h, w8
+; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
+; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z2.h
 ; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fcvtzu z1.s, p0/m, z3.h
-; CHECK-NEXT:    mov z4.h, w8
-; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
-; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fcmgt p3.h, p0/z, z2.h, z4.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
@@ -494,10 +494,10 @@ define <vscale x 4 x i16> @test_signed_v4f16_v4i16(<vscale x 4 x half> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    movprfx z2, z0
+; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    mov z0.s, #65535 // =0xffff
@@ -513,10 +513,10 @@ define <vscale x 8 x i16> @test_signed_v8f16_v8i16(<vscale x 8 x half> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.h, p0/m, z0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.h, p0/m, z0.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z0.h, z2.h
 ; CHECK-NEXT:    mov z1.h, p1/m, #0 // =0x0
@@ -532,10 +532,10 @@ define <vscale x 2 x i64> @test_signed_v2f16_v2i64(<vscale x 2 x half> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    movprfx z1, z0
-; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    movprfx z1, z0
+; CHECK-NEXT:    fcvtzu z1.d, p0/m, z0.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z0.h, z2.h
 ; CHECK-NEXT:    mov z1.d, p1/m, #0 // =0x0
@@ -553,13 +553,13 @@ define <vscale x 4 x i64> @test_signed_v4f16_v4i64(<vscale x 4 x half> %f) {
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
+; CHECK-NEXT:    mov z4.h, w8
+; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
+; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z2.h
 ; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fcvtzu z1.d, p0/m, z3.h
-; CHECK-NEXT:    mov z4.h, w8
-; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
-; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fcmgt p3.h, p0/z, z2.h, z4.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
diff --git a/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll b/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
index 1a6e4814cee6d..5c4c9463528b8 100644
--- a/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
+++ b/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
@@ -71,18 +71,18 @@ define <vscale x 4 x i8> @gather_i8_index_offset_8(ptr %base, i64 %offset, <vsca
 define void @scatter_f16_index_offset_var(ptr %base, i64 %offset, i64 %scale, <vscale x 4 x i1> %pg, <vscale x 4 x half> %data) #0 {
 ; CHECK-LABEL: scatter_f16_index_offset_var:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x1
-; CHECK-NEXT:    movprfx z4, z2
-; CHECK-NEXT:    mla z4.d, p1/m, z1.d, z2.d
 ; CHECK-NEXT:    punpklo p2.h, p0.b
 ; CHECK-NEXT:    uunpklo z3.d, z0.s
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    movprfx z4, z2
+; CHECK-NEXT:    mla z4.d, p1/m, z1.d, z2.d
 ; CHECK-NEXT:    incd z1.d
-; CHECK-NEXT:    st1h { z3.d }, p2, [x0, z4.d, lsl #1]
 ; CHECK-NEXT:    mad z1.d, p1/m, z2.d, z2.d
+; CHECK-NEXT:    st1h { z3.d }, p2, [x0, z4.d, lsl #1]
 ; CHECK-NEXT:    st1h { z0.d }, p0, [x0, z1.d, lsl #1]
 ; CHECK-NEXT:    ret
   %t0 = insertelement <vscale x 4 x i64> undef, i64 %offset, i32 0
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll b/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
index 3ecfa7e6c0b18..e40d65efb158b 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
@@ -145,10 +145,10 @@ define <vscale x 4 x double> @masked_gather_nxv4f64(ptr %base, <vscale x 4 x i16
 ; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    movprfx z1, z0
 ; CHECK-NEXT:    sxth z1.s, p1/m, z0.s
-; CHECK-NEXT:    sunpklo z0.d, z1.s
 ; CHECK-NEXT:    punpklo p1.h, p0.b
-; CHECK-NEXT:    sunpkhi z1.d, z1.s
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    sunpklo z0.d, z1.s
+; CHECK-NEXT:    sunpkhi z1.d, z1.s
 ; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0, z0.d, lsl #3]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0, z1.d, lsl #3]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-smulo-sdnode.ll b/llvm/test/CodeGen/AArch64/sve-smulo-sdnode.ll
index 871c74a1de14c..a2429d4975b7a 100644
--- a/llvm/test/CodeGen/AArch64/sve-smulo-sdnode.ll
+++ b/llvm/test/CodeGen/AArch64/sve-smulo-sdnode.ll
@@ -129,9 +129,9 @@ define <vscale x 64 x i8> @smulo_nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i
 ; CHECK-NEXT:    mul z27.b, p0/m, z27.b, z5.b
 ; CHECK-NEXT:    smulh z3.b, p0/m, z3.b, z7.b
 ; CHECK-NEXT:    smulh z0.b, p0/m, z0.b, z4.b
-; CHECK-NEXT:    asr z4.b, z25.b, #7
 ; CHECK-NEXT:    smulh z2.b, p0/m, z2.b, z6.b
 ; CHECK-NEXT:    smulh z1.b, p0/m, z1.b, z5.b
+; CHECK-NEXT:    asr z4.b, z25.b, #7
 ; CHECK-NEXT:    asr z5.b, z24.b, #7
 ; CHECK-NEXT:    asr z6.b, z26.b, #7
 ; CHECK-NEXT:    asr z7.b, z27.b, #7
@@ -140,13 +140,13 @@ define <vscale x 64 x i8> @smulo_nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i
 ; CHECK-NEXT:    cmpne p3.b, p0/z, z2.b, z6.b
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z1.b, z7.b
 ; CHECK-NEXT:    mov z25.b, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z26.b, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z24.b, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z27.b, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z26.b, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z27.d
-; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 64 x i8>, <vscale x 64 x i1> } @llvm.smul.with.overflow.nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i8> %y)
   %b = extractvalue { <vscale x 64 x i8>, <vscale x 64 x i1> } %a, 0
@@ -262,9 +262,9 @@ define <vscale x 32 x i16> @smulo_nxv32i16(<vscale x 32 x i16> %x, <vscale x 32
 ; CHECK-NEXT:    mul z27.h, p0/m, z27.h, z5.h
 ; CHECK-NEXT:    smulh z3.h, p0/m, z3.h, z7.h
 ; CHECK-NEXT:    smulh z0.h, p0/m, z0.h, z4.h
-; CHECK-NEXT:    asr z4.h, z25.h, #15
 ; CHECK-NEXT:    smulh z2.h, p0/m, z2.h, z6.h
 ; CHECK-NEXT:    smulh z1.h, p0/m, z1.h, z5.h
+; CHECK-NEXT:    asr z4.h, z25.h, #15
 ; CHECK-NEXT:    asr z5.h, z24.h, #15
 ; CHECK-NEXT:    asr z6.h, z26.h, #15
 ; CHECK-NEXT:    asr z7.h, z27.h, #15
@@ -273,13 +273,13 @@ define <vscale x 32 x i16> @smulo_nxv32i16(<vscale x 32 x i16> %x, <vscale x 32
 ; CHECK-NEXT:    cmpne p3.h, p0/z, z2.h, z6.h
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z1.h, z7.h
 ; CHECK-NEXT:    mov z25.h, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z26.h, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z24.h, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z27.h, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z26.h, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z27.d
-; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 32 x i16>, <vscale x 32 x i1> } @llvm.smul.with.overflow.nxv32i16(<vscale x 32 x i16> %x, <vscale x 32 x i16> %y)
   %b = extractvalue { <vscale x 32 x i16>, <vscale x 32 x i1> } %a, 0
@@ -374,9 +374,9 @@ define <vscale x 16 x i32> @smulo_nxv16i32(<vscale x 16 x i32> %x, <vscale x 16
 ; CHECK-NEXT:    mul z27.s, p0/m, z27.s, z5.s
 ; CHECK-NEXT:    smulh z3.s, p0/m, z3.s, z7.s
 ; CHECK-NEXT:    smulh z0.s, p0/m, z0.s, z4.s
-; CHECK-NEXT:    asr z4.s, z25.s, #31
 ; CHECK-NEXT:    smulh z2.s, p0/m, z2.s, z6.s
 ; CHECK-NEXT:    smulh z1.s, p0/m, z1.s, z5.s
+; CHECK-NEXT:    asr z4.s, z25.s, #31
 ; CHECK-NEXT:    asr z5.s, z24.s, #31
 ; CHECK-NEXT:    asr z6.s, z26.s, #31
 ; CHECK-NEXT:    asr z7.s, z27.s, #31
@@ -385,13 +385,13 @@ define <vscale x 16 x i32> @smulo_nxv16i32(<vscale x 16 x i32> %x, <vscale x 16
 ; CHECK-NEXT:    cmpne p3.s, p0/z, z2.s, z6.s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, z7.s
 ; CHECK-NEXT:    mov z25.s, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z26.s, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z24.s, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z27.s, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z26.s, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z27.d
-; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 16 x i32>, <vscale x 16 x i1> } @llvm.smul.with.overflow.nxv16i32(<vscale x 16 x i32> %x, <vscale x 16 x i32> %y)
   %b = extractvalue { <vscale x 16 x i32>, <vscale x 16 x i1> } %a, 0
@@ -465,9 +465,9 @@ define <vscale x 8 x i64> @smulo_nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i6
 ; CHECK-NEXT:    mul z27.d, p0/m, z27.d, z5.d
 ; CHECK-NEXT:    smulh z3.d, p0/m, z3.d, z7.d
 ; CHECK-NEXT:    smulh z0.d, p0/m, z0.d, z4.d
-; CHECK-NEXT:    asr z4.d, z25.d, #63
 ; CHECK-NEXT:    smulh z2.d, p0/m, z2.d, z6.d
 ; CHECK-NEXT:    smulh z1.d, p0/m, z1.d, z5.d
+; CHECK-NEXT:    asr z4.d, z25.d, #63
 ; CHECK-NEXT:    asr z5.d, z24.d, #63
 ; CHECK-NEXT:    asr z6.d, z26.d, #63
 ; CHECK-NEXT:    asr z7.d, z27.d, #63
@@ -476,13 +476,13 @@ define <vscale x 8 x i64> @smulo_nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i6
 ; CHECK-NEXT:    cmpne p3.d, p0/z, z2.d, z6.d
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, z7.d
 ; CHECK-NEXT:    mov z25.d, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z26.d, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z24.d, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z27.d, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z26.d, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z27.d
-; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 8 x i64>, <vscale x 8 x i1> } @llvm.smul.with.overflow.nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i64> %y)
   %b = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i1> } %a, 0
diff --git a/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll b/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
index 44796606e7a1a..3997409172d03 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
@@ -6,8 +6,8 @@
 define <vscale x 8 x float> @fcvts_nxv8f16(<vscale x 8 x half> %a) {
 ; CHECK-LABEL: fcvts_nxv8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.s, z0.h
 ; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    uunpklo z1.s, z0.h
 ; CHECK-NEXT:    uunpkhi z2.s, z0.h
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.s, p0/m, z1.h
@@ -21,8 +21,8 @@ define <vscale x 8 x float> @fcvts_nxv8f16(<vscale x 8 x half> %a) {
 define <vscale x 4 x double> @fcvtd_nxv4f16(<vscale x 4 x half> %a) {
 ; CHECK-LABEL: fcvtd_nxv4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z1.h
@@ -43,13 +43,13 @@ define <vscale x 8 x double> @fcvtd_nxv8f16(<vscale x 8 x half> %a) {
 ; CHECK-NEXT:    uunpkhi z1.d, z1.s
 ; CHECK-NEXT:    uunpklo z3.d, z0.s
 ; CHECK-NEXT:    uunpkhi z4.d, z0.s
+; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z2.h
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    fcvt z2.d, p0/m, z3.h
 ; CHECK-NEXT:    movprfx z3, z4
 ; CHECK-NEXT:    fcvt z3.d, p0/m, z4.h
-; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    ret
   %res = fpext <vscale x 8 x half> %a to <vscale x 8 x double>
   ret <vscale x 8 x double> %res
@@ -58,8 +58,8 @@ define <vscale x 8 x double> @fcvtd_nxv8f16(<vscale x 8 x half> %a) {
 define <vscale x 4 x double> @fcvtd_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtd_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z1.s
@@ -73,17 +73,17 @@ define <vscale x 4 x double> @fcvtd_nxv4f32(<vscale x 4 x float> %a) {
 define <vscale x 8 x double> @fcvtd_nxv8f32(<vscale x 8 x float> %a) {
 ; CHECK-LABEL: fcvtd_nxv8f32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z4.d, z1.s
+; CHECK-NEXT:    uunpkhi z5.d, z1.s
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z2.s
-; CHECK-NEXT:    movprfx z2, z4
-; CHECK-NEXT:    fcvt z2.d, p0/m, z4.s
-; CHECK-NEXT:    uunpkhi z5.d, z1.s
 ; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fcvt z1.d, p0/m, z3.s
+; CHECK-NEXT:    movprfx z2, z4
+; CHECK-NEXT:    fcvt z2.d, p0/m, z4.s
 ; CHECK-NEXT:    movprfx z3, z5
 ; CHECK-NEXT:    fcvt z3.d, p0/m, z5.s
 ; CHECK-NEXT:    ret
@@ -195,8 +195,8 @@ define <vscale x 8 x i16> @fcvtzs_h_nxv8f64(<vscale x 8 x double> %a) {
 define <vscale x 4 x i64> @fcvtzs_d_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtzs_d_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z1.s
@@ -210,17 +210,17 @@ define <vscale x 4 x i64> @fcvtzs_d_nxv4f32(<vscale x 4 x float> %a) {
 define <vscale x 16 x i32> @fcvtzs_s_nxv16f16(<vscale x 16 x half> %a) {
 ; CHECK-LABEL: fcvtzs_s_nxv16f16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z2.s, z0.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z4.s, z1.h
+; CHECK-NEXT:    uunpkhi z5.s, z1.h
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z2.h
-; CHECK-NEXT:    movprfx z2, z4
-; CHECK-NEXT:    fcvtzs z2.s, p0/m, z4.h
-; CHECK-NEXT:    uunpkhi z5.s, z1.h
 ; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fcvtzs z1.s, p0/m, z3.h
+; CHECK-NEXT:    movprfx z2, z4
+; CHECK-NEXT:    fcvtzs z2.s, p0/m, z4.h
 ; CHECK-NEXT:    movprfx z3, z5
 ; CHECK-NEXT:    fcvtzs z3.s, p0/m, z5.h
 ; CHECK-NEXT:    ret
@@ -247,8 +247,8 @@ define <vscale x 4 x i32> @fcvtzu_s_nxv4f64(<vscale x 4 x double> %a) {
 define <vscale x 4 x i64> @fcvtzu_d_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtzu_d_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z1.s
@@ -301,13 +301,13 @@ define <vscale x 16 x float> @scvtf_s_nxv16i8(<vscale x 16 x i8> %a) {
 ; CHECK-NEXT:    sunpkhi z1.s, z1.h
 ; CHECK-NEXT:    sunpklo z3.s, z0.h
 ; CHECK-NEXT:    sunpkhi z4.s, z0.h
+; CHECK-NEXT:    scvtf z1.s, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z2.s
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    scvtf z2.s, p0/m, z3.s
 ; CHECK-NEXT:    movprfx z3, z4
 ; CHECK-NEXT:    scvtf z3.s, p0/m, z4.s
-; CHECK-NEXT:    scvtf z1.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 16 x i8> %a to <vscale x 16 x float>
   ret <vscale x 16 x float> %res
@@ -316,8 +316,8 @@ define <vscale x 16 x float> @scvtf_s_nxv16i8(<vscale x 16 x i8> %a) {
 define <vscale x 4 x double> @scvtf_d_nxv4i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: scvtf_d_nxv4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    sunpklo z1.d, z0.s
 ; CHECK-NEXT:    sunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z1.d
@@ -378,8 +378,8 @@ define <vscale x 8 x half> @ucvtf_h_nxv8i64(<vscale x 8 x i64> %a) {
 define <vscale x 4 x double> @ucvtf_d_nxv4i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: ucvtf_d_nxv4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
index 37cc79011a986..28e02da53af43 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
@@ -239,10 +239,10 @@ define void @fcvt_v16f16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov x8, #4 // =0x4
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z0.h
 ; CHECK-NEXT:    ld1h { z4.d }, p0/z, [x0, x8, lsl #1]
-; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    mov x8, #6 // =0x6
-; CHECK-NEXT:    fcvt z2.d, p0/m, z2.h
+; CHECK-NEXT:    fcvt z1.d, p0/m, z1.h
 ; CHECK-NEXT:    ld1h { z5.d }, p0/z, [x0, x8, lsl #1]
+; CHECK-NEXT:    fcvt z2.d, p0/m, z2.h
 ; CHECK-NEXT:    mov x8, #2 // =0x2
 ; CHECK-NEXT:    fcvt z3.d, p0/m, z3.h
 ; CHECK-NEXT:    ld1h { z7.d }, p0/z, [x0, x8, lsl #1]
@@ -255,8 +255,8 @@ define void @fcvt_v16f16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    stp q2, q3, [x1, #64]
 ; CHECK-NEXT:    movprfx z2, z7
 ; CHECK-NEXT:    fcvt z2.d, p0/m, z7.h
-; CHECK-NEXT:    stp q1, q2, [x1]
 ; CHECK-NEXT:    stp q4, q0, [x1, #32]
+; CHECK-NEXT:    stp q1, q2, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x half>, ptr %a
   %res = fpext <16 x half> %op1 to <16 x double>
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
index 5e3ce0ddebe2e..fcf4f21c6ea84 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
@@ -750,14 +750,14 @@ define void @udiv_constantsplat_v8i32(ptr %a)  {
 ; SVE-NEXT:    mov z0.s, w8
 ; SVE-NEXT:    movprfx z3, z1
 ; SVE-NEXT:    umulh z3.s, p0/m, z3.s, z0.s
-; SVE-NEXT:    sub z1.s, z1.s, z3.s
 ; SVE-NEXT:    umulh z0.s, p0/m, z0.s, z2.s
-; SVE-NEXT:    lsr z1.s, z1.s, #1
+; SVE-NEXT:    sub z1.s, z1.s, z3.s
 ; SVE-NEXT:    sub z2.s, z2.s, z0.s
-; SVE-NEXT:    add z1.s, z1.s, z3.s
+; SVE-NEXT:    lsr z1.s, z1.s, #1
 ; SVE-NEXT:    lsr z2.s, z2.s, #1
-; SVE-NEXT:    lsr z1.s, z1.s, #6
+; SVE-NEXT:    add z1.s, z1.s, z3.s
 ; SVE-NEXT:    add z0.s, z2.s, z0.s
+; SVE-NEXT:    lsr z1.s, z1.s, #6
 ; SVE-NEXT:    lsr z0.s, z0.s, #6
 ; SVE-NEXT:    stp q1, q0, [x0]
 ; SVE-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
index eb95a410209b4..4a1209b942f4a 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
@@ -139,7 +139,6 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    sdiv z5.s, p0/m, z5.s, z4.s
 ; CHECK-NEXT:    ldr q3, [x0]
 ; CHECK-NEXT:    ldr q4, [x1]
-; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    mov z18.d, z3.d
 ; CHECK-NEXT:    mov z17.d, z4.d
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
@@ -154,6 +153,7 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    sdivr z7.s, p0/m, z7.s, z16.s
 ; CHECK-NEXT:    sunpklo z18.s, z18.h
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    sunpklo z17.s, z17.h
 ; CHECK-NEXT:    sdivr z19.s, p0/m, z19.s, z20.s
 ; CHECK-NEXT:    sunpklo z20.h, z3.b
@@ -172,21 +172,21 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    sdivr z18.s, p0/m, z18.s, z20.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
+; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z5.h
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
-; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
+; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
-; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
 ; CHECK-NEXT:    splice z19.h, p0, z19.h, z18.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
 ; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z6.b, p0, z6.b, z2.b
 ; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
+; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    mls z2.b, p1/m, z7.b, z4.b
-; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i8>, ptr %a
@@ -293,8 +293,8 @@ define <2 x i32> @srem_v2i32(<2 x i32> %op1, <2 x i32> %op2) {
 ; CHECK-LABEL: srem_v2i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl2
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls z0.s, p0/m, z2.s, z1.s
@@ -308,8 +308,8 @@ define <4 x i32> @srem_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 ; CHECK-LABEL: srem_v4i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls z0.s, p0/m, z2.s, z1.s
@@ -345,8 +345,8 @@ define <1 x i64> @srem_v1i64(<1 x i64> %op1, <1 x i64> %op2) {
 ; CHECK-LABEL: srem_v1i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl1
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -360,8 +360,8 @@ define <2 x i64> @srem_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 ; CHECK-LABEL: srem_v2i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl2
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    sdiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -528,7 +528,6 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    udiv z5.s, p0/m, z5.s, z4.s
 ; CHECK-NEXT:    ldr q3, [x0]
 ; CHECK-NEXT:    ldr q4, [x1]
-; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    mov z18.d, z3.d
 ; CHECK-NEXT:    mov z17.d, z4.d
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
@@ -543,6 +542,7 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    udivr z7.s, p0/m, z7.s, z16.s
 ; CHECK-NEXT:    uunpklo z18.s, z18.h
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    uunpklo z17.s, z17.h
 ; CHECK-NEXT:    udivr z19.s, p0/m, z19.s, z20.s
 ; CHECK-NEXT:    uunpklo z20.h, z3.b
@@ -561,21 +561,21 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    udivr z18.s, p0/m, z18.s, z20.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
+; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z5.h
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
-; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
+; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
-; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
 ; CHECK-NEXT:    splice z19.h, p0, z19.h, z18.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
 ; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z6.b, p0, z6.b, z2.b
 ; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
+; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
 ; CHECK-NEXT:    movprfx z2, z3
 ; CHECK-NEXT:    mls z2.b, p1/m, z7.b, z4.b
-; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i8>, ptr %a
@@ -682,8 +682,8 @@ define <2 x i32> @urem_v2i32(<2 x i32> %op1, <2 x i32> %op2) {
 ; CHECK-LABEL: urem_v2i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl2
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls z0.s, p0/m, z2.s, z1.s
@@ -697,8 +697,8 @@ define <4 x i32> @urem_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 ; CHECK-LABEL: urem_v4i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.s, p0/m, z2.s, z1.s
 ; CHECK-NEXT:    mls z0.s, p0/m, z2.s, z1.s
@@ -734,8 +734,8 @@ define <1 x i64> @urem_v1i64(<1 x i64> %op1, <1 x i64> %op2) {
 ; CHECK-LABEL: urem_v1i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl1
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
@@ -749,8 +749,8 @@ define <2 x i64> @urem_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 ; CHECK-LABEL: urem_v2i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl2
-; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    udiv z2.d, p0/m, z2.d, z1.d
 ; CHECK-NEXT:    mls z0.d, p0/m, z2.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
index c110e89326cc0..5c5cf68135bf8 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
@@ -221,21 +221,21 @@ define void @ucvtf_v16i16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z1.d, z1.s
 ; CHECK-NEXT:    ucvtf z2.d, p0/m, z2.d
-; CHECK-NEXT:    ucvtf z3.d, p0/m, z3.d
 ; CHECK-NEXT:    uunpklo z7.d, z7.s
+; CHECK-NEXT:    ucvtf z3.d, p0/m, z3.d
 ; CHECK-NEXT:    uunpklo z5.d, z5.s
+; CHECK-NEXT:    uunpklo z6.d, z6.s
 ; CHECK-NEXT:    ucvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
-; CHECK-NEXT:    uunpklo z6.d, z6.s
 ; CHECK-NEXT:    ucvtf z1.d, p0/m, z1.d
 ; CHECK-NEXT:    ucvtf z5.d, p0/m, z5.d
 ; CHECK-NEXT:    stp q2, q4, [x1, #64]
 ; CHECK-NEXT:    movprfx z2, z6
 ; CHECK-NEXT:    ucvtf z2.d, p0/m, z6.d
-; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q0, q5, [x1, #96]
 ; CHECK-NEXT:    movprfx z0, z7
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z7.d
+; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q3, q0, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
@@ -481,10 +481,10 @@ define <8 x half> @ucvtf_v8i64_v8f16(ptr %a) {
 ; CHECK-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    splice z1.s, p0, z1.s, z0.s
-; CHECK-NEXT:    movprfx z0, z1
-; CHECK-NEXT:    fcvt z0.h, p1/m, z1.s
 ; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    movprfx z0, z1
+; CHECK-NEXT:    fcvt z0.h, p1/m, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
 ; CHECK-NEXT:    fcvt z1.h, p1/m, z2.s
 ; CHECK-NEXT:    uzp1 z2.h, z0.h, z0.h
@@ -794,21 +794,21 @@ define void @scvtf_v16i16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
 ; CHECK-NEXT:    scvtf z2.d, p0/m, z2.d
-; CHECK-NEXT:    scvtf z3.d, p0/m, z3.d
 ; CHECK-NEXT:    sunpklo z7.d, z7.s
+; CHECK-NEXT:    scvtf z3.d, p0/m, z3.d
 ; CHECK-NEXT:    sunpklo z5.d, z5.s
+; CHECK-NEXT:    sunpklo z6.d, z6.s
 ; CHECK-NEXT:    scvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
-; CHECK-NEXT:    sunpklo z6.d, z6.s
 ; CHECK-NEXT:    scvtf z1.d, p0/m, z1.d
 ; CHECK-NEXT:    scvtf z5.d, p0/m, z5.d
 ; CHECK-NEXT:    stp q2, q4, [x1, #64]
 ; CHECK-NEXT:    movprfx z2, z6
 ; CHECK-NEXT:    scvtf z2.d, p0/m, z6.d
-; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q0, q5, [x1, #96]
 ; CHECK-NEXT:    movprfx z0, z7
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z7.d
+; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q3, q0, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
@@ -982,16 +982,16 @@ define void @scvtf_v16i32_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
 ; CHECK-NEXT:    ext z6.b, z6.b, z4.b, #8
-; CHECK-NEXT:    sunpklo z4.d, z4.s
 ; CHECK-NEXT:    ext z7.b, z7.b, z5.b, #8
+; CHECK-NEXT:    sunpklo z4.d, z4.s
 ; CHECK-NEXT:    sunpklo z5.d, z5.s
 ; CHECK-NEXT:    sunpklo z2.d, z2.s
 ; CHECK-NEXT:    sunpklo z3.d, z3.s
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    sunpklo z6.d, z6.s
+; CHECK-NEXT:    sunpklo z7.d, z7.s
 ; CHECK-NEXT:    scvtf z1.d, p0/m, z1.d
 ; CHECK-NEXT:    scvtf z4.d, p0/m, z4.d
-; CHECK-NEXT:    sunpklo z7.d, z7.s
 ; CHECK-NEXT:    scvtf z2.d, p0/m, z2.d
 ; CHECK-NEXT:    scvtf z3.d, p0/m, z3.d
 ; CHECK-NEXT:    stp q1, q3, [x1, #64]
diff --git a/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll b/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
index 45d796f68c424..36d64725742e5 100644
--- a/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
+++ b/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
@@ -120,16 +120,16 @@ define <vscale x 64 x i8> @umulo_nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i
 ; CHECK-NEXT:    umulh z27.b, p0/m, z27.b, z5.b
 ; CHECK-NEXT:    mul z3.b, p0/m, z3.b, z7.b
 ; CHECK-NEXT:    mul z0.b, p0/m, z0.b, z4.b
-; CHECK-NEXT:    cmpne p1.b, p0/z, z25.b, #0
 ; CHECK-NEXT:    mul z2.b, p0/m, z2.b, z6.b
 ; CHECK-NEXT:    mul z1.b, p0/m, z1.b, z5.b
+; CHECK-NEXT:    cmpne p1.b, p0/z, z25.b, #0
 ; CHECK-NEXT:    cmpne p2.b, p0/z, z24.b, #0
 ; CHECK-NEXT:    cmpne p3.b, p0/z, z26.b, #0
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z27.b, #0
 ; CHECK-NEXT:    mov z0.b, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z2.b, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z3.b, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.b, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z2.b, p3/m, #0 // =0x0
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 64 x i8>, <vscale x 64 x i1> } @llvm.umul.with.overflow.nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i8> %y)
   %b = extractvalue { <vscale x 64 x i8>, <vscale x 64 x i1> } %a, 0
@@ -237,16 +237,16 @@ define <vscale x 32 x i16> @umulo_nxv32i16(<vscale x 32 x i16> %x, <vscale x 32
 ; CHECK-NEXT:    umulh z27.h, p0/m, z27.h, z5.h
 ; CHECK-NEXT:    mul z3.h, p0/m, z3.h, z7.h
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z4.h
-; CHECK-NEXT:    cmpne p1.h, p0/z, z25.h, #0
 ; CHECK-NEXT:    mul z2.h, p0/m, z2.h, z6.h
 ; CHECK-NEXT:    mul z1.h, p0/m, z1.h, z5.h
+; CHECK-NEXT:    cmpne p1.h, p0/z, z25.h, #0
 ; CHECK-NEXT:    cmpne p2.h, p0/z, z24.h, #0
 ; CHECK-NEXT:    cmpne p3.h, p0/z, z26.h, #0
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z27.h, #0
 ; CHECK-NEXT:    mov z0.h, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z2.h, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z3.h, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.h, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z2.h, p3/m, #0 // =0x0
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 32 x i16>, <vscale x 32 x i1> } @llvm.umul.with.overflow.nxv32i16(<vscale x 32 x i16> %x, <vscale x 32 x i16> %y)
   %b = extractvalue { <vscale x 32 x i16>, <vscale x 32 x i1> } %a, 0
@@ -334,16 +334,16 @@ define <vscale x 16 x i32> @umulo_nxv16i32(<vscale x 16 x i32> %x, <vscale x 16
 ; CHECK-NEXT:    umulh z27.s, p0/m, z27.s, z5.s
 ; CHECK-NEXT:    mul z3.s, p0/m, z3.s, z7.s
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z4.s
-; CHECK-NEXT:    cmpne p1.s, p0/z, z25.s, #0
 ; CHECK-NEXT:    mul z2.s, p0/m, z2.s, z6.s
 ; CHECK-NEXT:    mul z1.s, p0/m, z1.s, z5.s
+; CHECK-NEXT:    cmpne p1.s, p0/z, z25.s, #0
 ; CHECK-NEXT:    cmpne p2.s, p0/z, z24.s, #0
 ; CHECK-NEXT:    cmpne p3.s, p0/z, z26.s, #0
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z27.s, #0
 ; CHECK-NEXT:    mov z0.s, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z2.s, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z3.s, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.s, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z2.s, p3/m, #0 // =0x0
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 16 x i32>, <vscale x 16 x i1> } @llvm.umul.with.overflow.nxv16i32(<vscale x 16 x i32> %x, <vscale x 16 x i32> %y)
   %b = extractvalue { <vscale x 16 x i32>, <vscale x 16 x i1> } %a, 0
@@ -411,16 +411,16 @@ define <vscale x 8 x i64> @umulo_nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i6
 ; CHECK-NEXT:    umulh z27.d, p0/m, z27.d, z5.d
 ; CHECK-NEXT:    mul z3.d, p0/m, z3.d, z7.d
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z4.d
-; CHECK-NEXT:    cmpne p1.d, p0/z, z25.d, #0
 ; CHECK-NEXT:    mul z2.d, p0/m, z2.d, z6.d
 ; CHECK-NEXT:    mul z1.d, p0/m, z1.d, z5.d
+; CHECK-NEXT:    cmpne p1.d, p0/z, z25.d, #0
 ; CHECK-NEXT:    cmpne p2.d, p0/z, z24.d, #0
 ; CHECK-NEXT:    cmpne p3.d, p0/z, z26.d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z27.d, #0
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
+; CHECK-NEXT:    mov z2.d, p3/m, #0 // =0x0
 ; CHECK-NEXT:    mov z3.d, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z2.d, p3/m, #0 // =0x0
 ; CHECK-NEXT:    ret
   %a = call { <vscale x 8 x i64>, <vscale x 8 x i1> } @llvm.umul.with.overflow.nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i64> %y)
   %b = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i1> } %a, 0
diff --git a/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll b/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
index 4d0d2a991d0e5..194d1071301d4 100644
--- a/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
+++ b/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
@@ -31,8 +31,8 @@ define i32 @test(<vscale x 32 x i8> %bin.rdx, <vscale x 32 x i8> %bin.rdx2)  {
 ; CHECK-NEXT:    sunpkhi z24.s, z3.h
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z2.s
 ; CHECK-NEXT:    sunpkhi z2.s, z7.h
-; CHECK-NEXT:    sunpklo z3.s, z3.h
 ; CHECK-NEXT:    sunpklo z7.s, z7.h
+; CHECK-NEXT:    sunpklo z3.s, z3.h
 ; CHECK-NEXT:    mla z0.s, p0/m, z25.s, z24.s
 ; CHECK-NEXT:    mad z2.s, p0/m, z6.s, z4.s
 ; CHECK-NEXT:    mad z1.s, p0/m, z3.s, z26.s
diff --git a/llvm/test/CodeGen/AArch64/sve2-xar.ll b/llvm/test/CodeGen/AArch64/sve2-xar.ll
index 62680522bab93..e297ade6b9ae1 100644
--- a/llvm/test/CodeGen/AArch64/sve2-xar.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-xar.ll
@@ -155,8 +155,8 @@ define <vscale x 2 x i64> @xar_nxv2i64_l_neg1(<vscale x 2 x i64> %x, <vscale x 2
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    subr z2.d, z2.d, #0 // =0x0
 ; CHECK-NEXT:    eor z0.d, z0.d, z1.d
-; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    and z3.d, z3.d, #0x3f
+; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    movprfx z1, z0
 ; CHECK-NEXT:    lsl z1.d, p0/m, z1.d, z3.d
 ; CHECK-NEXT:    lsr z0.d, p0/m, z0.d, z2.d

From 5fc8a190b37906cba3f175440abd79dabd8acc73 Mon Sep 17 00:00:00 2001
From: wanglei <wanglei@loongson.cn>
Date: Fri, 12 Apr 2024 16:49:52 +0800
Subject: [PATCH 233/886] [LoongArch] Pre commit test for #88371. NFC

---
 .../CodeGenPrepare/LoongArch/lit.local.cfg    |  2 ++
 .../CodeGenPrepare/LoongArch/splitgep.ll      | 34 +++++++++++++++++++
 2 files changed, 36 insertions(+)
 create mode 100644 llvm/test/Transforms/CodeGenPrepare/LoongArch/lit.local.cfg
 create mode 100644 llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll

diff --git a/llvm/test/Transforms/CodeGenPrepare/LoongArch/lit.local.cfg b/llvm/test/Transforms/CodeGenPrepare/LoongArch/lit.local.cfg
new file mode 100644
index 0000000000000..cc24278acbb41
--- /dev/null
+++ b/llvm/test/Transforms/CodeGenPrepare/LoongArch/lit.local.cfg
@@ -0,0 +1,2 @@
+if not "LoongArch" in config.root.targets:
+    config.unsupported = True
diff --git a/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll b/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll
new file mode 100644
index 0000000000000..ca2a16a0ca7ec
--- /dev/null
+++ b/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll
@@ -0,0 +1,34 @@
+; RUN: opt --mtriple=loongarch64 -S --passes='require<profile-summary>,function(codegenprepare)' %s | FileCheck %s
+
+; Check that we have deterministic output
+define void @test(ptr %sp, ptr %t, i32 %n) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    %s = load ptr, ptr %sp, align 8
+; CHECK-NEXT:    br label %while_cond
+
+entry:
+  %s = load ptr, ptr %sp
+  br label %while_cond
+
+while_cond:
+  %phi = phi i32 [ 0, %entry ], [ %i, %while_body ]
+  %gep0 = getelementptr [65536 x i32], ptr %s, i64 0, i64 20000
+  %gep1 = getelementptr [65536 x i32], ptr %s, i64 0, i64 20001
+  %gep2 = getelementptr [65536 x i32], ptr %t, i64 0, i64 20000
+  %gep3 = getelementptr [65536 x i32], ptr %t, i64 0, i64 20001
+  %cmp = icmp slt i32 %phi, %n
+  br i1 %cmp, label %while_body, label %while_end
+
+while_body:
+  %i = add i32 %phi, 1
+  store i32 %i, ptr %gep0
+  store i32 %phi, ptr %gep1
+  store i32 %i, ptr %gep2
+  store i32 %phi, ptr %gep3
+  br label %while_cond
+
+while_end:
+  ret void
+}
+

From c39df496d73621238ba87de235d9a61e2b2203b9 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Fri, 12 Apr 2024 13:05:49 +0300
Subject: [PATCH 234/886] [clang][NFC] Refactor `CUDAFunctionTarget`

Refactor `CUDAFunctionTarget` into a scoped enum at namespace scope, so that it can be forward declared. This is done in preparation for `SemaCUDA`.
---
 clang/include/clang/Sema/Sema.h |  29 ++++---
 clang/lib/Sema/SemaCUDA.cpp     | 146 ++++++++++++++++++--------------
 clang/lib/Sema/SemaDecl.cpp     |   8 +-
 clang/lib/Sema/SemaDeclAttr.cpp |  15 ++--
 clang/lib/Sema/SemaExpr.cpp     |  16 ++--
 clang/lib/Sema/SemaExprCXX.cpp  |   3 +-
 clang/lib/Sema/SemaOverload.cpp |  14 +--
 clang/lib/Sema/SemaStmt.cpp     |   3 +-
 clang/lib/Sema/SemaType.cpp     |  32 +++----
 9 files changed, 145 insertions(+), 121 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 90c9166282703..00888b7f7a738 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -435,6 +435,14 @@ enum class CXXSpecialMemberKind {
   Invalid
 };
 
+enum class CUDAFunctionTarget {
+  Device,
+  Global,
+  Host,
+  HostDevice,
+  InvalidTarget
+};
+
 /// Sema - This implements semantic analysis and AST building for C.
 /// \nosubgrouping
 class Sema final : public SemaBase {
@@ -3663,20 +3671,12 @@ class Sema final : public SemaBase {
   InternalLinkageAttr *mergeInternalLinkageAttr(Decl *D,
                                                 const InternalLinkageAttr &AL);
 
-  enum CUDAFunctionTarget {
-    CFT_Device,
-    CFT_Global,
-    CFT_Host,
-    CFT_HostDevice,
-    CFT_InvalidTarget
-  };
-
   /// Check validaty of calling convention attribute \p attr. If \p FD
   /// is not null pointer, use \p FD to determine the CUDA/HIP host/device
   /// target. Otherwise, it is specified by \p CFT.
-  bool CheckCallingConvAttr(const ParsedAttr &attr, CallingConv &CC,
-                            const FunctionDecl *FD = nullptr,
-                            CUDAFunctionTarget CFT = CFT_InvalidTarget);
+  bool CheckCallingConvAttr(
+      const ParsedAttr &attr, CallingConv &CC, const FunctionDecl *FD = nullptr,
+      CUDAFunctionTarget CFT = CUDAFunctionTarget::InvalidTarget);
 
   void AddParameterABIAttr(Decl *D, const AttributeCommonInfo &CI,
                            ParameterABI ABI);
@@ -12967,7 +12967,8 @@ class Sema final : public SemaBase {
   /// Example usage:
   ///
   ///  // Variable-length arrays are not allowed in CUDA device code.
-  ///  if (CUDADiagIfDeviceCode(Loc, diag::err_cuda_vla) << CurrentCUDATarget())
+  ///  if (CUDADiagIfDeviceCode(Loc, diag::err_cuda_vla)
+  ///     << llvm::to_underlying(CurrentCUDATarget()))
   ///    return ExprError();
   ///  // Otherwise, continue parsing as normal.
   SemaDiagnosticBuilder CUDADiagIfDeviceCode(SourceLocation Loc,
@@ -12983,7 +12984,7 @@ class Sema final : public SemaBase {
   /// function.
   ///
   /// Use this rather than examining the function's attributes yourself -- you
-  /// will get it wrong.  Returns CFT_Host if D is null.
+  /// will get it wrong.  Returns CUDAFunctionTarget::Host if D is null.
   CUDAFunctionTarget IdentifyCUDATarget(const FunctionDecl *D,
                                         bool IgnoreImplicitHDAttr = false);
   CUDAFunctionTarget IdentifyCUDATarget(const ParsedAttributesView &Attrs);
@@ -13008,7 +13009,7 @@ class Sema final : public SemaBase {
   /// Define the current global CUDA host/device context where a function may be
   /// called. Only used when a function is called outside of any functions.
   struct CUDATargetContext {
-    CUDAFunctionTarget Target = CFT_HostDevice;
+    CUDAFunctionTarget Target = CUDAFunctionTarget::HostDevice;
     CUDATargetContextKind Kind = CTCK_Unknown;
     Decl *D = nullptr;
   } CurCUDATargetCtx;
diff --git a/clang/lib/Sema/SemaCUDA.cpp b/clang/lib/Sema/SemaCUDA.cpp
index 1596222e3d1da..9d6d709e262ad 100644
--- a/clang/lib/Sema/SemaCUDA.cpp
+++ b/clang/lib/Sema/SemaCUDA.cpp
@@ -22,6 +22,7 @@
 #include "clang/Sema/SemaDiagnostic.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallVector.h"
 #include <optional>
 using namespace clang;
@@ -64,8 +65,7 @@ ExprResult Sema::ActOnCUDAExecConfigExpr(Scope *S, SourceLocation LLLLoc,
                        /*IsExecConfig=*/true);
 }
 
-Sema::CUDAFunctionTarget
-Sema::IdentifyCUDATarget(const ParsedAttributesView &Attrs) {
+CUDAFunctionTarget Sema::IdentifyCUDATarget(const ParsedAttributesView &Attrs) {
   bool HasHostAttr = false;
   bool HasDeviceAttr = false;
   bool HasGlobalAttr = false;
@@ -90,18 +90,18 @@ Sema::IdentifyCUDATarget(const ParsedAttributesView &Attrs) {
   }
 
   if (HasInvalidTargetAttr)
-    return CFT_InvalidTarget;
+    return CUDAFunctionTarget::InvalidTarget;
 
   if (HasGlobalAttr)
-    return CFT_Global;
+    return CUDAFunctionTarget::Global;
 
   if (HasHostAttr && HasDeviceAttr)
-    return CFT_HostDevice;
+    return CUDAFunctionTarget::HostDevice;
 
   if (HasDeviceAttr)
-    return CFT_Device;
+    return CUDAFunctionTarget::Device;
 
-  return CFT_Host;
+  return CUDAFunctionTarget::Host;
 }
 
 template <typename A>
@@ -120,43 +120,43 @@ Sema::CUDATargetContextRAII::CUDATargetContextRAII(Sema &S_,
   assert(K == CTCK_InitGlobalVar);
   auto *VD = dyn_cast_or_null<VarDecl>(D);
   if (VD && VD->hasGlobalStorage() && !VD->isStaticLocal()) {
-    auto Target = CFT_Host;
+    auto Target = CUDAFunctionTarget::Host;
     if ((hasAttr<CUDADeviceAttr>(VD, /*IgnoreImplicit=*/true) &&
          !hasAttr<CUDAHostAttr>(VD, /*IgnoreImplicit=*/true)) ||
         hasAttr<CUDASharedAttr>(VD, /*IgnoreImplicit=*/true) ||
         hasAttr<CUDAConstantAttr>(VD, /*IgnoreImplicit=*/true))
-      Target = CFT_Device;
+      Target = CUDAFunctionTarget::Device;
     S.CurCUDATargetCtx = {Target, K, VD};
   }
 }
 
 /// IdentifyCUDATarget - Determine the CUDA compilation target for this function
-Sema::CUDAFunctionTarget Sema::IdentifyCUDATarget(const FunctionDecl *D,
-                                                  bool IgnoreImplicitHDAttr) {
+CUDAFunctionTarget Sema::IdentifyCUDATarget(const FunctionDecl *D,
+                                            bool IgnoreImplicitHDAttr) {
   // Code that lives outside a function gets the target from CurCUDATargetCtx.
   if (D == nullptr)
     return CurCUDATargetCtx.Target;
 
   if (D->hasAttr<CUDAInvalidTargetAttr>())
-    return CFT_InvalidTarget;
+    return CUDAFunctionTarget::InvalidTarget;
 
   if (D->hasAttr<CUDAGlobalAttr>())
-    return CFT_Global;
+    return CUDAFunctionTarget::Global;
 
   if (hasAttr<CUDADeviceAttr>(D, IgnoreImplicitHDAttr)) {
     if (hasAttr<CUDAHostAttr>(D, IgnoreImplicitHDAttr))
-      return CFT_HostDevice;
-    return CFT_Device;
+      return CUDAFunctionTarget::HostDevice;
+    return CUDAFunctionTarget::Device;
   } else if (hasAttr<CUDAHostAttr>(D, IgnoreImplicitHDAttr)) {
-    return CFT_Host;
+    return CUDAFunctionTarget::Host;
   } else if ((D->isImplicit() || !D->isUserProvided()) &&
              !IgnoreImplicitHDAttr) {
     // Some implicit declarations (like intrinsic functions) are not marked.
     // Set the most lenient target on them for maximal flexibility.
-    return CFT_HostDevice;
+    return CUDAFunctionTarget::HostDevice;
   }
 
-  return CFT_Host;
+  return CUDAFunctionTarget::Host;
 }
 
 /// IdentifyTarget - Determine the CUDA compilation target for this variable.
@@ -181,10 +181,10 @@ Sema::CUDAVariableTarget Sema::IdentifyCUDATarget(const VarDecl *Var) {
   //  - on device side in device or global functions
   if (auto *FD = dyn_cast<FunctionDecl>(Var->getDeclContext())) {
     switch (IdentifyCUDATarget(FD)) {
-    case CFT_HostDevice:
+    case CUDAFunctionTarget::HostDevice:
       return CVT_Both;
-    case CFT_Device:
-    case CFT_Global:
+    case CUDAFunctionTarget::Device:
+    case CUDAFunctionTarget::Global:
       return CVT_Device;
     default:
       return CVT_Host;
@@ -230,7 +230,7 @@ Sema::IdentifyCUDAPreference(const FunctionDecl *Caller,
   // trivial ctor/dtor without device attr to be used. Non-trivial ctor/dtor
   // will be diagnosed by checkAllowedCUDAInitializer.
   if (Caller == nullptr && CurCUDATargetCtx.Kind == CTCK_InitGlobalVar &&
-      CurCUDATargetCtx.Target == CFT_Device &&
+      CurCUDATargetCtx.Target == CUDAFunctionTarget::Device &&
       (isa<CXXConstructorDecl>(Callee) || isa<CXXDestructorDecl>(Callee)))
     return CFP_HostDevice;
 
@@ -239,40 +239,47 @@ Sema::IdentifyCUDAPreference(const FunctionDecl *Caller,
 
   // If one of the targets is invalid, the check always fails, no matter what
   // the other target is.
-  if (CallerTarget == CFT_InvalidTarget || CalleeTarget == CFT_InvalidTarget)
+  if (CallerTarget == CUDAFunctionTarget::InvalidTarget ||
+      CalleeTarget == CUDAFunctionTarget::InvalidTarget)
     return CFP_Never;
 
   // (a) Can't call global from some contexts until we support CUDA's
   // dynamic parallelism.
-  if (CalleeTarget == CFT_Global &&
-      (CallerTarget == CFT_Global || CallerTarget == CFT_Device))
+  if (CalleeTarget == CUDAFunctionTarget::Global &&
+      (CallerTarget == CUDAFunctionTarget::Global ||
+       CallerTarget == CUDAFunctionTarget::Device))
     return CFP_Never;
 
   // (b) Calling HostDevice is OK for everyone.
-  if (CalleeTarget == CFT_HostDevice)
+  if (CalleeTarget == CUDAFunctionTarget::HostDevice)
     return CFP_HostDevice;
 
   // (c) Best case scenarios
   if (CalleeTarget == CallerTarget ||
-      (CallerTarget == CFT_Host && CalleeTarget == CFT_Global) ||
-      (CallerTarget == CFT_Global && CalleeTarget == CFT_Device))
+      (CallerTarget == CUDAFunctionTarget::Host &&
+       CalleeTarget == CUDAFunctionTarget::Global) ||
+      (CallerTarget == CUDAFunctionTarget::Global &&
+       CalleeTarget == CUDAFunctionTarget::Device))
     return CFP_Native;
 
   // HipStdPar mode is special, in that assessing whether a device side call to
   // a host target is deferred to a subsequent pass, and cannot unambiguously be
   // adjudicated in the AST, hence we optimistically allow them to pass here.
   if (getLangOpts().HIPStdPar &&
-      (CallerTarget == CFT_Global || CallerTarget == CFT_Device ||
-       CallerTarget == CFT_HostDevice) &&
-      CalleeTarget == CFT_Host)
+      (CallerTarget == CUDAFunctionTarget::Global ||
+       CallerTarget == CUDAFunctionTarget::Device ||
+       CallerTarget == CUDAFunctionTarget::HostDevice) &&
+      CalleeTarget == CUDAFunctionTarget::Host)
     return CFP_HostDevice;
 
   // (d) HostDevice behavior depends on compilation mode.
-  if (CallerTarget == CFT_HostDevice) {
+  if (CallerTarget == CUDAFunctionTarget::HostDevice) {
     // It's OK to call a compilation-mode matching function from an HD one.
-    if ((getLangOpts().CUDAIsDevice && CalleeTarget == CFT_Device) ||
+    if ((getLangOpts().CUDAIsDevice &&
+         CalleeTarget == CUDAFunctionTarget::Device) ||
         (!getLangOpts().CUDAIsDevice &&
-         (CalleeTarget == CFT_Host || CalleeTarget == CFT_Global)))
+         (CalleeTarget == CUDAFunctionTarget::Host ||
+          CalleeTarget == CUDAFunctionTarget::Global)))
       return CFP_SameSide;
 
     // Calls from HD to non-mode-matching functions (i.e., to host functions
@@ -283,9 +290,12 @@ Sema::IdentifyCUDAPreference(const FunctionDecl *Caller,
   }
 
   // (e) Calling across device/host boundary is not something you should do.
-  if ((CallerTarget == CFT_Host && CalleeTarget == CFT_Device) ||
-      (CallerTarget == CFT_Device && CalleeTarget == CFT_Host) ||
-      (CallerTarget == CFT_Global && CalleeTarget == CFT_Host))
+  if ((CallerTarget == CUDAFunctionTarget::Host &&
+       CalleeTarget == CUDAFunctionTarget::Device) ||
+      (CallerTarget == CUDAFunctionTarget::Device &&
+       CalleeTarget == CUDAFunctionTarget::Host) ||
+      (CallerTarget == CUDAFunctionTarget::Global &&
+       CalleeTarget == CUDAFunctionTarget::Host))
     return CFP_Never;
 
   llvm_unreachable("All cases should've been handled by now.");
@@ -337,16 +347,16 @@ void Sema::EraseUnwantedCUDAMatches(
 /// \param ResolvedTarget with a target that resolves for both calls.
 /// \return true if there's a conflict, false otherwise.
 static bool
-resolveCalleeCUDATargetConflict(Sema::CUDAFunctionTarget Target1,
-                                Sema::CUDAFunctionTarget Target2,
-                                Sema::CUDAFunctionTarget *ResolvedTarget) {
+resolveCalleeCUDATargetConflict(CUDAFunctionTarget Target1,
+                                CUDAFunctionTarget Target2,
+                                CUDAFunctionTarget *ResolvedTarget) {
   // Only free functions and static member functions may be global.
-  assert(Target1 != Sema::CFT_Global);
-  assert(Target2 != Sema::CFT_Global);
+  assert(Target1 != CUDAFunctionTarget::Global);
+  assert(Target2 != CUDAFunctionTarget::Global);
 
-  if (Target1 == Sema::CFT_HostDevice) {
+  if (Target1 == CUDAFunctionTarget::HostDevice) {
     *ResolvedTarget = Target2;
-  } else if (Target2 == Sema::CFT_HostDevice) {
+  } else if (Target2 == CUDAFunctionTarget::HostDevice) {
     *ResolvedTarget = Target1;
   } else if (Target1 != Target2) {
     return true;
@@ -422,7 +432,8 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
         if (Diagnose) {
           Diag(ClassDecl->getLocation(),
                diag::note_implicit_member_target_infer_collision)
-              << (unsigned)CSM << *InferredTarget << BaseMethodTarget;
+              << (unsigned)CSM << llvm::to_underlying(*InferredTarget)
+              << llvm::to_underlying(BaseMethodTarget);
         }
         MemberDecl->addAttr(CUDAInvalidTargetAttr::CreateImplicit(Context));
         return true;
@@ -465,7 +476,8 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
         if (Diagnose) {
           Diag(ClassDecl->getLocation(),
                diag::note_implicit_member_target_infer_collision)
-              << (unsigned)CSM << *InferredTarget << FieldMethodTarget;
+              << (unsigned)CSM << llvm::to_underlying(*InferredTarget)
+              << llvm::to_underlying(FieldMethodTarget);
         }
         MemberDecl->addAttr(CUDAInvalidTargetAttr::CreateImplicit(Context));
         return true;
@@ -478,9 +490,9 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
   // it's the least restrictive option that can be invoked from any target.
   bool NeedsH = true, NeedsD = true;
   if (InferredTarget) {
-    if (*InferredTarget == CFT_Device)
+    if (*InferredTarget == CUDAFunctionTarget::Device)
       NeedsH = false;
-    else if (*InferredTarget == CFT_Host)
+    else if (*InferredTarget == CUDAFunctionTarget::Host)
       NeedsD = false;
   }
 
@@ -677,9 +689,10 @@ void Sema::checkAllowedCUDAInitializer(VarDecl *VD) {
     }
     if (InitFn) {
       CUDAFunctionTarget InitFnTarget = IdentifyCUDATarget(InitFn);
-      if (InitFnTarget != CFT_Host && InitFnTarget != CFT_HostDevice) {
+      if (InitFnTarget != CUDAFunctionTarget::Host &&
+          InitFnTarget != CUDAFunctionTarget::HostDevice) {
         Diag(VD->getLocation(), diag::err_ref_bad_target_global_initializer)
-            << InitFnTarget << InitFn;
+            << llvm::to_underlying(InitFnTarget) << InitFn;
         Diag(InitFn->getLocation(), diag::note_previous_decl) << InitFn;
         VD->setInvalidDecl();
       }
@@ -699,8 +712,9 @@ void Sema::CUDARecordImplicitHostDeviceFuncUsedByDevice(
   CUDAFunctionTarget CallerTarget = IdentifyCUDATarget(Caller);
 
   // Record whether an implicit host device function is used on device side.
-  if (CallerTarget != CFT_Device && CallerTarget != CFT_Global &&
-      (CallerTarget != CFT_HostDevice ||
+  if (CallerTarget != CUDAFunctionTarget::Device &&
+      CallerTarget != CUDAFunctionTarget::Global &&
+      (CallerTarget != CUDAFunctionTarget::HostDevice ||
        (isCUDAImplicitHostDeviceFunction(Caller) &&
         !getASTContext().CUDAImplicitHostDeviceFunUsedByDevice.count(Caller))))
     return;
@@ -806,10 +820,10 @@ Sema::SemaDiagnosticBuilder Sema::CUDADiagIfDeviceCode(SourceLocation Loc,
     if (!CurFunContext)
       return SemaDiagnosticBuilder::K_Nop;
     switch (CurrentCUDATarget()) {
-    case CFT_Global:
-    case CFT_Device:
+    case CUDAFunctionTarget::Global:
+    case CUDAFunctionTarget::Device:
       return SemaDiagnosticBuilder::K_Immediate;
-    case CFT_HostDevice:
+    case CUDAFunctionTarget::HostDevice:
       // An HD function counts as host code if we're compiling for host, and
       // device code if we're compiling for device.  Defer any errors in device
       // mode until the function is known-emitted.
@@ -836,9 +850,9 @@ Sema::SemaDiagnosticBuilder Sema::CUDADiagIfHostCode(SourceLocation Loc,
     if (!CurFunContext)
       return SemaDiagnosticBuilder::K_Nop;
     switch (CurrentCUDATarget()) {
-    case CFT_Host:
+    case CUDAFunctionTarget::Host:
       return SemaDiagnosticBuilder::K_Immediate;
-    case CFT_HostDevice:
+    case CUDAFunctionTarget::HostDevice:
       // An HD function counts as host code if we're compiling for host, and
       // device code if we're compiling for device.  Defer any errors in device
       // mode until the function is known-emitted.
@@ -911,8 +925,8 @@ bool Sema::CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee) {
     return true;
 
   SemaDiagnosticBuilder(DiagKind, Loc, diag::err_ref_bad_target, Caller, *this)
-      << IdentifyCUDATarget(Callee) << /*function*/ 0 << Callee
-      << IdentifyCUDATarget(Caller);
+      << llvm::to_underlying(IdentifyCUDATarget(Callee)) << /*function*/ 0
+      << Callee << llvm::to_underlying(IdentifyCUDATarget(Caller));
   if (!Callee->getBuiltinID())
     SemaDiagnosticBuilder(DiagKind, Callee->getLocation(),
                           diag::note_previous_decl, Caller, *this)
@@ -995,19 +1009,21 @@ void Sema::checkCUDATargetOverload(FunctionDecl *NewFD,
     // HD/global functions "exist" in some sense on both the host and device, so
     // should have the same implementation on both sides.
     if (NewTarget != OldTarget &&
-        ((NewTarget == CFT_HostDevice &&
+        ((NewTarget == CUDAFunctionTarget::HostDevice &&
           !(LangOpts.OffloadImplicitHostDeviceTemplates &&
             isCUDAImplicitHostDeviceFunction(NewFD) &&
-            OldTarget == CFT_Device)) ||
-         (OldTarget == CFT_HostDevice &&
+            OldTarget == CUDAFunctionTarget::Device)) ||
+         (OldTarget == CUDAFunctionTarget::HostDevice &&
           !(LangOpts.OffloadImplicitHostDeviceTemplates &&
             isCUDAImplicitHostDeviceFunction(OldFD) &&
-            NewTarget == CFT_Device)) ||
-         (NewTarget == CFT_Global) || (OldTarget == CFT_Global)) &&
+            NewTarget == CUDAFunctionTarget::Device)) ||
+         (NewTarget == CUDAFunctionTarget::Global) ||
+         (OldTarget == CUDAFunctionTarget::Global)) &&
         !IsOverload(NewFD, OldFD, /* UseMemberUsingDeclRules = */ false,
                     /* ConsiderCudaAttrs = */ false)) {
       Diag(NewFD->getLocation(), diag::err_cuda_ovl_target)
-          << NewTarget << NewFD->getDeclName() << OldTarget << OldFD;
+          << llvm::to_underlying(NewTarget) << NewFD->getDeclName()
+          << llvm::to_underlying(OldTarget) << OldFD;
       Diag(OldFD->getLocation(), diag::note_previous_declaration);
       NewFD->setInvalidDecl();
       break;
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index b448bacd49f32..720e56692359b 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -20666,11 +20666,11 @@ Sema::FunctionEmissionStatus Sema::getEmissionStatus(const FunctionDecl *FD,
     // when compiling for host, device and global functions are never emitted.
     // (Technically, we do emit a host-side stub for global functions, but this
     // doesn't count for our purposes here.)
-    Sema::CUDAFunctionTarget T = IdentifyCUDATarget(FD);
-    if (LangOpts.CUDAIsDevice && T == Sema::CFT_Host)
+    CUDAFunctionTarget T = IdentifyCUDATarget(FD);
+    if (LangOpts.CUDAIsDevice && T == CUDAFunctionTarget::Host)
       return FunctionEmissionStatus::CUDADiscarded;
     if (!LangOpts.CUDAIsDevice &&
-        (T == Sema::CFT_Device || T == Sema::CFT_Global))
+        (T == CUDAFunctionTarget::Device || T == CUDAFunctionTarget::Global))
       return FunctionEmissionStatus::CUDADiscarded;
 
     if (IsEmittedForExternalSymbol())
@@ -20691,5 +20691,5 @@ bool Sema::shouldIgnoreInHostDeviceCheck(FunctionDecl *Callee) {
   // for host, only HD functions actually called from the host get marked as
   // known-emitted.
   return LangOpts.CUDA && !LangOpts.CUDAIsDevice &&
-         IdentifyCUDATarget(Callee) == CFT_Global;
+         IdentifyCUDATarget(Callee) == CUDAFunctionTarget::Global;
 }
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index b91064e28e415..56c9d90c9b52b 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -42,6 +42,7 @@
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/IR/Assumptions.h"
 #include "llvm/MC/MCSectionMachO.h"
@@ -5099,7 +5100,7 @@ static void handleSharedAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   }
   if (S.getLangOpts().CUDA && VD->hasLocalStorage() &&
       S.CUDADiagIfHostCode(AL.getLoc(), diag::err_cuda_host_shared)
-          << S.CurrentCUDATarget())
+          << llvm::to_underlying(S.CurrentCUDATarget()))
     return;
   D->addAttr(::new (S.Context) CUDASharedAttr(S.Context, AL));
 }
@@ -5493,22 +5494,22 @@ bool Sema::CheckCallingConvAttr(const ParsedAttr &Attrs, CallingConv &CC,
   // on their host/device attributes.
   if (LangOpts.CUDA) {
     auto *Aux = Context.getAuxTargetInfo();
-    assert(FD || CFT != CFT_InvalidTarget);
+    assert(FD || CFT != CUDAFunctionTarget::InvalidTarget);
     auto CudaTarget = FD ? IdentifyCUDATarget(FD) : CFT;
     bool CheckHost = false, CheckDevice = false;
     switch (CudaTarget) {
-    case CFT_HostDevice:
+    case CUDAFunctionTarget::HostDevice:
       CheckHost = true;
       CheckDevice = true;
       break;
-    case CFT_Host:
+    case CUDAFunctionTarget::Host:
       CheckHost = true;
       break;
-    case CFT_Device:
-    case CFT_Global:
+    case CUDAFunctionTarget::Device:
+    case CUDAFunctionTarget::Global:
       CheckDevice = true;
       break;
-    case CFT_InvalidTarget:
+    case CUDAFunctionTarget::InvalidTarget:
       llvm_unreachable("unexpected cuda target");
     }
     auto *HostTI = LangOpts.CUDAIsDevice ? Aux : &TI;
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index ec5ca2b9352ed..18284e0c3e987 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -53,6 +53,7 @@
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/ConvertUTF.h"
@@ -17307,7 +17308,8 @@ ExprResult Sema::BuildVAArgExpr(SourceLocation BuiltinLoc,
   if (getLangOpts().CUDA && getLangOpts().CUDAIsDevice) {
     if (const FunctionDecl *F = dyn_cast<FunctionDecl>(CurContext)) {
       CUDAFunctionTarget T = IdentifyCUDATarget(F);
-      if (T == CFT_Global || T == CFT_Device || T == CFT_HostDevice)
+      if (T == CUDAFunctionTarget::Global || T == CUDAFunctionTarget::Device ||
+          T == CUDAFunctionTarget::HostDevice)
         return ExprError(Diag(E->getBeginLoc(), diag::err_va_arg_in_device));
     }
   }
@@ -19182,14 +19184,16 @@ MarkVarDeclODRUsed(ValueDecl *V, SourceLocation Loc, Sema &SemaRef,
     auto VarTarget = SemaRef.IdentifyCUDATarget(Var);
     auto UserTarget = SemaRef.IdentifyCUDATarget(FD);
     if (VarTarget == Sema::CVT_Host &&
-        (UserTarget == Sema::CFT_Device || UserTarget == Sema::CFT_HostDevice ||
-         UserTarget == Sema::CFT_Global)) {
+        (UserTarget == CUDAFunctionTarget::Device ||
+         UserTarget == CUDAFunctionTarget::HostDevice ||
+         UserTarget == CUDAFunctionTarget::Global)) {
       // Diagnose ODR-use of host global variables in device functions.
       // Reference of device global variables in host functions is allowed
       // through shadow variables therefore it is not diagnosed.
       if (SemaRef.LangOpts.CUDAIsDevice && !SemaRef.LangOpts.HIPStdPar) {
         SemaRef.targetDiag(Loc, diag::err_ref_bad_target)
-            << /*host*/ 2 << /*variable*/ 1 << Var << UserTarget;
+            << /*host*/ 2 << /*variable*/ 1 << Var
+            << llvm::to_underlying(UserTarget);
         SemaRef.targetDiag(Var->getLocation(),
                            Var->getType().isConstQualified()
                                ? diag::note_cuda_const_var_unpromoted
@@ -19197,8 +19201,8 @@ MarkVarDeclODRUsed(ValueDecl *V, SourceLocation Loc, Sema &SemaRef,
       }
     } else if (VarTarget == Sema::CVT_Device &&
                !Var->hasAttr<CUDASharedAttr>() &&
-               (UserTarget == Sema::CFT_Host ||
-                UserTarget == Sema::CFT_HostDevice)) {
+               (UserTarget == CUDAFunctionTarget::Host ||
+                UserTarget == CUDAFunctionTarget::HostDevice)) {
       // Record a CUDA/HIP device side variable if it is ODR-used
       // by host code. This is done conservatively, when the variable is
       // referenced in any of the following contexts:
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index ac39cb99fc934..ce9d5c26e2185 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -44,6 +44,7 @@
 #include "clang/Sema/TemplateDeduction.h"
 #include "llvm/ADT/APInt.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/TypeSize.h"
@@ -884,7 +885,7 @@ ExprResult Sema::BuildCXXThrow(SourceLocation OpLoc, Expr *Ex,
   // Exceptions aren't allowed in CUDA device code.
   if (getLangOpts().CUDA)
     CUDADiagIfDeviceCode(OpLoc, diag::err_cuda_device_exceptions)
-        << "throw" << CurrentCUDATarget();
+        << "throw" << llvm::to_underlying(CurrentCUDATarget());
 
   if (getCurScope() && getCurScope()->isOpenMPSimdDirectiveScope())
     Diag(OpLoc, diag::err_omp_simd_region_cannot_use_stmt) << "throw";
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 1f674dd4bb0fb..e1155dc2d5d28 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -1549,10 +1549,10 @@ static bool IsOverloadOrOverrideImpl(Sema &SemaRef, FunctionDecl *New,
     // Don't allow overloading of destructors.  (In theory we could, but it
     // would be a giant change to clang.)
     if (!isa<CXXDestructorDecl>(New)) {
-      Sema::CUDAFunctionTarget NewTarget = SemaRef.IdentifyCUDATarget(New),
-                               OldTarget = SemaRef.IdentifyCUDATarget(Old);
-      if (NewTarget != Sema::CFT_InvalidTarget) {
-        assert((OldTarget != Sema::CFT_InvalidTarget) &&
+      CUDAFunctionTarget NewTarget = SemaRef.IdentifyCUDATarget(New),
+                         OldTarget = SemaRef.IdentifyCUDATarget(Old);
+      if (NewTarget != CUDAFunctionTarget::InvalidTarget) {
+        assert((OldTarget != CUDAFunctionTarget::InvalidTarget) &&
                "Unexpected invalid target.");
 
         // Allow overloading of functions with same signature and different CUDA
@@ -11938,8 +11938,8 @@ static void DiagnoseBadTarget(Sema &S, OverloadCandidate *Cand) {
   FunctionDecl *Caller = S.getCurFunctionDecl(/*AllowLambda=*/true);
   FunctionDecl *Callee = Cand->Function;
 
-  Sema::CUDAFunctionTarget CallerTarget = S.IdentifyCUDATarget(Caller),
-                           CalleeTarget = S.IdentifyCUDATarget(Callee);
+  CUDAFunctionTarget CallerTarget = S.IdentifyCUDATarget(Caller),
+                     CalleeTarget = S.IdentifyCUDATarget(Callee);
 
   std::string FnDesc;
   std::pair<OverloadCandidateKind, OverloadCandidateSelect> FnKindPair =
@@ -11949,7 +11949,7 @@ static void DiagnoseBadTarget(Sema &S, OverloadCandidate *Cand) {
   S.Diag(Callee->getLocation(), diag::note_ovl_candidate_bad_target)
       << (unsigned)FnKindPair.first << (unsigned)ocs_non_template
       << FnDesc /* Ignored */
-      << CalleeTarget << CallerTarget;
+      << llvm::to_underlying(CalleeTarget) << llvm::to_underlying(CallerTarget);
 
   // This could be an implicit constructor for which we could not infer the
   // target due to a collsion. Diagnose that case.
diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
index e53c76e65b03d..1c2f6120f6218 100644
--- a/clang/lib/Sema/SemaStmt.cpp
+++ b/clang/lib/Sema/SemaStmt.cpp
@@ -37,6 +37,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
@@ -4574,7 +4575,7 @@ StmtResult Sema::ActOnCXXTryBlock(SourceLocation TryLoc, Stmt *TryBlock,
   // Exceptions aren't allowed in CUDA device code.
   if (getLangOpts().CUDA)
     CUDADiagIfDeviceCode(TryLoc, diag::err_cuda_device_exceptions)
-        << "try" << CurrentCUDATarget();
+        << "try" << llvm::to_underlying(CurrentCUDATarget());
 
   if (getCurScope() && getCurScope()->isOpenMPSimdDirectiveScope())
     Diag(TryLoc, diag::err_omp_simd_region_cannot_use_stmt) << "try";
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index b8a1518fbe00a..b3f6078952f6e 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -37,6 +37,7 @@
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateInstCallback.h"
 #include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringExtras.h"
@@ -376,11 +377,10 @@ enum TypeAttrLocation {
 static void
 processTypeAttrs(TypeProcessingState &state, QualType &type,
                  TypeAttrLocation TAL, const ParsedAttributesView &attrs,
-                 Sema::CUDAFunctionTarget CFT = Sema::CFT_HostDevice);
+                 CUDAFunctionTarget CFT = CUDAFunctionTarget::HostDevice);
 
 static bool handleFunctionTypeAttr(TypeProcessingState &state, ParsedAttr &attr,
-                                   QualType &type,
-                                   Sema::CUDAFunctionTarget CFT);
+                                   QualType &type, CUDAFunctionTarget CFT);
 
 static bool handleMSPointerTypeQualifierAttr(TypeProcessingState &state,
                                              ParsedAttr &attr, QualType &type);
@@ -627,7 +627,7 @@ static void distributeFunctionTypeAttr(TypeProcessingState &state,
 static bool distributeFunctionTypeAttrToInnermost(
     TypeProcessingState &state, ParsedAttr &attr,
     ParsedAttributesView &attrList, QualType &declSpecType,
-    Sema::CUDAFunctionTarget CFT) {
+    CUDAFunctionTarget CFT) {
   Declarator &declarator = state.getDeclarator();
 
   // Put it on the innermost function chunk, if there is one.
@@ -644,10 +644,10 @@ static bool distributeFunctionTypeAttrToInnermost(
 
 /// A function type attribute was written in the decl spec.  Try to
 /// apply it somewhere.
-static void
-distributeFunctionTypeAttrFromDeclSpec(TypeProcessingState &state,
-                                       ParsedAttr &attr, QualType &declSpecType,
-                                       Sema::CUDAFunctionTarget CFT) {
+static void distributeFunctionTypeAttrFromDeclSpec(TypeProcessingState &state,
+                                                   ParsedAttr &attr,
+                                                   QualType &declSpecType,
+                                                   CUDAFunctionTarget CFT) {
   state.saveDeclSpecAttrs();
 
   // Try to distribute to the innermost.
@@ -664,9 +664,10 @@ distributeFunctionTypeAttrFromDeclSpec(TypeProcessingState &state,
 /// Try to apply it somewhere.
 /// `Attrs` is the attribute list containing the declaration (either of the
 /// declarator or the declaration).
-static void distributeFunctionTypeAttrFromDeclarator(
-    TypeProcessingState &state, ParsedAttr &attr, QualType &declSpecType,
-    Sema::CUDAFunctionTarget CFT) {
+static void distributeFunctionTypeAttrFromDeclarator(TypeProcessingState &state,
+                                                     ParsedAttr &attr,
+                                                     QualType &declSpecType,
+                                                     CUDAFunctionTarget CFT) {
   Declarator &declarator = state.getDeclarator();
 
   // Try to distribute to the innermost.
@@ -694,7 +695,7 @@ static void distributeFunctionTypeAttrFromDeclarator(
 /// declarator or the declaration).
 static void distributeTypeAttrsFromDeclarator(TypeProcessingState &state,
                                               QualType &declSpecType,
-                                              Sema::CUDAFunctionTarget CFT) {
+                                              CUDAFunctionTarget CFT) {
   // The called functions in this loop actually remove things from the current
   // list, so iterating over the existing list isn't possible.  Instead, make a
   // non-owning copy and iterate over that.
@@ -2734,7 +2735,7 @@ QualType Sema::BuildArrayType(QualType T, ArraySizeModifier ASM,
       bool IsCUDADevice = (getLangOpts().CUDA && getLangOpts().CUDAIsDevice);
       targetDiag(Loc,
                  IsCUDADevice ? diag::err_cuda_vla : diag::err_vla_unsupported)
-          << (IsCUDADevice ? CurrentCUDATarget() : 0);
+          << (IsCUDADevice ? llvm::to_underlying(CurrentCUDATarget()) : 0);
     } else if (sema::FunctionScopeInfo *FSI = getCurFunction()) {
       // VLAs are supported on this target, but we may need to do delayed
       // checking that the VLA is not being used within a coroutine.
@@ -8028,8 +8029,7 @@ static bool handleArmStateAttribute(Sema &S,
 /// Process an individual function attribute.  Returns true to
 /// indicate that the attribute was handled, false if it wasn't.
 static bool handleFunctionTypeAttr(TypeProcessingState &state, ParsedAttr &attr,
-                                   QualType &type,
-                                   Sema::CUDAFunctionTarget CFT) {
+                                   QualType &type, CUDAFunctionTarget CFT) {
   Sema &S = state.getSema();
 
   FunctionTypeUnwrapper unwrapped(S, type);
@@ -8863,7 +8863,7 @@ static void HandleHLSLParamModifierAttr(QualType &CurType,
 static void processTypeAttrs(TypeProcessingState &state, QualType &type,
                              TypeAttrLocation TAL,
                              const ParsedAttributesView &attrs,
-                             Sema::CUDAFunctionTarget CFT) {
+                             CUDAFunctionTarget CFT) {
 
   state.setParsedNoDeref(false);
   if (attrs.empty())

From 3bb147e05365680d0a954d823a7c014f357448e7 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 12 Apr 2024 10:54:35 +0100
Subject: [PATCH 235/886] [X86] combine-sse41-intrinsics.ll - add AVX2 test
 coverage

---
 .../CodeGen/X86/combine-sse41-intrinsics.ll   | 52 +++++++++++++------
 1 file changed, 36 insertions(+), 16 deletions(-)

diff --git a/llvm/test/CodeGen/X86/combine-sse41-intrinsics.ll b/llvm/test/CodeGen/X86/combine-sse41-intrinsics.ll
index cbb5bd09c2399..6034a24099e4d 100644
--- a/llvm/test/CodeGen/X86/combine-sse41-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/combine-sse41-intrinsics.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx    | FileCheck %s --check-prefixes=CHECK,AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx    | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2   | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
 
 
 define <2 x double> @test_x86_sse41_blend_pd(<2 x double> %a0, <2 x double> %a1) {
@@ -146,12 +147,18 @@ define <16 x i8> @demandedelts_pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8>
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: demandedelts_pblendvb:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    retq
+; AVX1-LABEL: demandedelts_pblendvb:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: demandedelts_pblendvb:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastb %xmm0, %xmm0
+; AVX2-NEXT:    retq
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <16 x i32> zeroinitializer
   %2 = shufflevector <16 x i8> %a1, <16 x i8> undef, <16 x i32> zeroinitializer
   %3 = shufflevector <16 x i8> %a2, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -193,15 +200,28 @@ define <4 x float> @demandedbits_uitofp_blendvps(<4 x float> %a0, <4 x float> %a
 ; SSE-NEXT:    movaps %xmm3, %xmm0
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: demandedbits_uitofp_blendvps:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],mem[1],xmm2[2],mem[3],xmm2[4],mem[5],xmm2[6],mem[7]
-; AVX-NEXT:    vpsrld $16, %xmm2, %xmm2
-; AVX-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],mem[1],xmm2[2],mem[3],xmm2[4],mem[5],xmm2[6],mem[7]
-; AVX-NEXT:    vsubps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
-; AVX-NEXT:    vaddps %xmm2, %xmm3, %xmm2
-; AVX-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    retq
+; AVX1-LABEL: demandedbits_uitofp_blendvps:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],mem[1],xmm2[2],mem[3],xmm2[4],mem[5],xmm2[6],mem[7]
+; AVX1-NEXT:    vpsrld $16, %xmm2, %xmm2
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],mem[1],xmm2[2],mem[3],xmm2[4],mem[5],xmm2[6],mem[7]
+; AVX1-NEXT:    vsubps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT:    vaddps %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: demandedbits_uitofp_blendvps:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [1258291200,1258291200,1258291200,1258291200]
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],xmm3[1],xmm2[2],xmm3[3],xmm2[4],xmm3[5],xmm2[6],xmm3[7]
+; AVX2-NEXT:    vpsrld $16, %xmm2, %xmm2
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm4 = [1392508928,1392508928,1392508928,1392508928]
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2],xmm4[3],xmm2[4],xmm4[5],xmm2[6],xmm4[7]
+; AVX2-NEXT:    vbroadcastss {{.*#+}} xmm4 = [5.49764202E+11,5.49764202E+11,5.49764202E+11,5.49764202E+11]
+; AVX2-NEXT:    vsubps %xmm4, %xmm2, %xmm2
+; AVX2-NEXT:    vaddps %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
   %cvt = uitofp <4 x i32> %a2 to <4 x float>
   %sel = tail call noundef <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %cvt)
   ret <4 x float> %sel

From 8c0f52e9d5a99bf96bb64ac23b5893482c292527 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 12 Apr 2024 11:23:51 +0100
Subject: [PATCH 236/886] [DAG] Fold
 extract_subvector(insert_subvector(x,y,c1),c2) --> extract_subvector(y,c2-c1)
 (#87925)

If the extract_subvector is cheap, attempt to extract directly from an inserted subvector
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |   17 +
 .../any_extend_vector_inreg_of_broadcast.ll   |   42 +-
 llvm/test/CodeGen/X86/dpbusd.ll               |    2 +-
 llvm/test/CodeGen/X86/dpbusd_i4.ll            |    2 +-
 .../vector-interleaved-load-i16-stride-3.ll   | 1300 ++++++++---------
 .../vector-interleaved-store-i8-stride-7.ll   |  602 ++++----
 .../zero_extend_vector_inreg_of_broadcast.ll  |   28 +-
 7 files changed, 987 insertions(+), 1006 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 24f50b87c4cf2..d9c6c28d5dac9 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -24463,6 +24463,23 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
       if (!LegalOperations || TLI.isOperationLegal(ISD::SPLAT_VECTOR, NVT))
         return DAG.getSplatVector(NVT, DL, V.getOperand(0));
 
+  // extract_subvector(insert_subvector(x,y,c1),c2)
+  //  --> extract_subvector(y,c2-c1)
+  // iff we're just extracting from the inserted subvector.
+  if (V.getOpcode() == ISD::INSERT_SUBVECTOR) {
+    SDValue InsSub = V.getOperand(1);
+    EVT InsSubVT = InsSub.getValueType();
+    unsigned NumInsElts = InsSubVT.getVectorMinNumElements();
+    unsigned InsIdx = V.getConstantOperandVal(2);
+    unsigned NumSubElts = NVT.getVectorMinNumElements();
+    if (InsIdx <= ExtIdx && (ExtIdx + NumSubElts) <= (InsIdx + NumInsElts) &&
+        TLI.isExtractSubvectorCheap(NVT, InsSubVT, ExtIdx - InsIdx)) {
+      SDLoc DL(N);
+      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NVT, InsSub,
+                         DAG.getVectorIdxConstant(ExtIdx - InsIdx, DL));
+    }
+  }
+
   // Try to move vector bitcast after extract_subv by scaling extraction index:
   // extract_subv (bitcast X), Index --> bitcast (extract_subv X, Index')
   if (V.getOpcode() == ISD::BITCAST &&
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
index 4242d8483e723..39c7ce1413d1b 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -3507,13 +3507,12 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm2
-; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512F-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3523,13 +3522,12 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512DQ-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3768,10 +3766,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -3784,10 +3782,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4147,9 +4145,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %xmm0
@@ -4161,9 +4159,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/dpbusd.ll b/llvm/test/CodeGen/X86/dpbusd.ll
index fbea08eb1e550..04d7a9691b645 100644
--- a/llvm/test/CodeGen/X86/dpbusd.ll
+++ b/llvm/test/CodeGen/X86/dpbusd.ll
@@ -26,7 +26,7 @@ define i32 @no_dpbusd(ptr%a, ptr%b, i32 %c, i32 %n) {
 ; AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
 ; AVX512-NEXT:    vpmaddwd %ymm0, %ymm1, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/dpbusd_i4.ll b/llvm/test/CodeGen/X86/dpbusd_i4.ll
index 906fead7f8db5..a212f99680ef4 100644
--- a/llvm/test/CodeGen/X86/dpbusd_i4.ll
+++ b/llvm/test/CodeGen/X86/dpbusd_i4.ll
@@ -86,7 +86,7 @@ define i32 @mul_sext_i4i4(<16 x i4> %a, <16 x i4> %b, i32 %c) {
 ; CHECK-NEXT:    vpsraw $12, %ymm0, %ymm0
 ; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; CHECK-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 1436922f9dd11..6d5fc9ed0ab5b 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -1828,22 +1828,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-LABEL: load_i16_stride3_vf32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1857,14 +1857,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1885,21 +1885,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1908,22 +1906,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512-FCP:       # %bb.0:
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1937,14 +1935,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1965,21 +1963,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1988,22 +1984,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2017,14 +2013,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2045,21 +2041,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -2068,22 +2062,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ-FCP:       # %bb.0:
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2097,14 +2091,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2125,21 +2119,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -3500,688 +3492,668 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-LABEL: load_i16_stride3_vf64:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512DQ-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512DQ-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512DQ-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512DQ-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512DQ-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512DQ-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 8b6ba51506ab7..8091afbbfd70c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -1246,29 +1246,28 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
-; AVX512BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
 ; AVX512BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
-; AVX512BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
-; AVX512BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
+; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
+; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
+; AVX512BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX512BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
 ; AVX512BW-FCP-NEXT:    vzeroupper
 ; AVX512BW-FCP-NEXT:    retq
 ;
@@ -1326,29 +1325,28 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512DQ-BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512DQ-BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
 ; AVX512DQ-BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
-; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
+; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
 ; AVX512DQ-BW-FCP-NEXT:    vzeroupper
 ; AVX512DQ-BW-FCP-NEXT:    retq
   %in.vec0 = load <8 x i8>, ptr %in.vecptr0, align 64
@@ -2053,77 +2051,76 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-NEXT:    vmovdqa (%rdi), %xmm4
-; AVX512-NEXT:    vmovdqa (%rsi), %xmm5
-; AVX512-NEXT:    vmovdqa (%rdx), %xmm6
-; AVX512-NEXT:    vmovdqa (%rcx), %xmm7
-; AVX512-NEXT:    vmovdqa (%r8), %xmm0
-; AVX512-NEXT:    vmovdqa (%r10), %xmm1
-; AVX512-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
-; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
-; AVX512-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
-; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
-; AVX512-NEXT:    vpor %xmm6, %xmm4, %xmm4
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm5
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
-; AVX512-NEXT:    vpor %xmm6, %xmm5, %xmm5
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
-; AVX512-NEXT:    vpor %ymm4, %ymm6, %ymm4
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
+; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
+; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
+; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
+; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
+; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
+; AVX512-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
+; AVX512-NEXT:    vporq %zmm10, %zmm11, %zmm10
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
+; AVX512-NEXT:    vpandn %ymm12, %ymm13, %ymm12
+; AVX512-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
+; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
+; AVX512-NEXT:    vporq %zmm12, %zmm11, %zmm11
+; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
+; AVX512-NEXT:    vpor %ymm8, %ymm9, %ymm8
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
 ; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
-; AVX512-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
-; AVX512-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512-NEXT:    vporq %zmm3, %zmm2, %zmm2
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512-NEXT:    vpandn %ymm3, %ymm6, %ymm3
-; AVX512-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
-; AVX512-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
-; AVX512-NEXT:    vporq %zmm1, %zmm0, %zmm0
-; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
-; AVX512-NEXT:    vmovdqa %xmm5, 96(%rax)
-; AVX512-NEXT:    vmovdqa64 %zmm0, (%rax)
-; AVX512-NEXT:    vmovdqa %ymm4, 64(%rax)
+; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
+; AVX512-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512-NEXT:    vmovdqa64 %zmm11, (%rax)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
@@ -2131,70 +2128,69 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm2
-; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm3
-; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm4
-; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm5
-; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm1
-; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm0
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
-; AVX512-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
-; AVX512-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
-; AVX512-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
-; AVX512-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
-; AVX512-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
-; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
-; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
-; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
-; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
-; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
-; AVX512-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
-; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
-; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
-; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
-; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
-; AVX512-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
-; AVX512-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
+; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512-FCP-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
+; AVX512-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
+; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
+; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
+; AVX512-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
+; AVX512-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
+; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
+; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
+; AVX512-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
+; AVX512-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
+; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
+; AVX512-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
@@ -2202,77 +2198,76 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm4
-; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm5
-; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm6
-; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm7
-; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm0
-; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm1
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
-; AVX512DQ-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
-; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
-; AVX512DQ-NEXT:    vpor %xmm6, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vextracti128 $1, %ymm0, %xmm5
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
-; AVX512DQ-NEXT:    vpor %xmm6, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
-; AVX512DQ-NEXT:    vpor %ymm4, %ymm6, %ymm4
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512DQ-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512DQ-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vporq %zmm10, %zmm11, %zmm10
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
+; AVX512DQ-NEXT:    vpandn %ymm12, %ymm13, %ymm12
+; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
+; AVX512DQ-NEXT:    vporq %zmm12, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
+; AVX512DQ-NEXT:    vpor %ymm8, %ymm9, %ymm8
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
-; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512DQ-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512DQ-NEXT:    vporq %zmm3, %zmm2, %zmm2
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512DQ-NEXT:    vpandn %ymm3, %ymm6, %ymm3
-; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vporq %zmm1, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
-; AVX512DQ-NEXT:    vmovdqa %xmm5, 96(%rax)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rax)
-; AVX512DQ-NEXT:    vmovdqa %ymm4, 64(%rax)
+; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512DQ-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512DQ-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
+; AVX512DQ-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512DQ-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm11, (%rax)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
@@ -2280,70 +2275,69 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm3
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm5
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm1
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm0
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
-; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
-; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
-; AVX512DQ-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
-; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
-; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
-; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
-; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
-; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
-; AVX512DQ-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
-; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
+; AVX512DQ-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
+; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512DQ-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
+; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
+; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
+; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
+; AVX512DQ-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
+; AVX512DQ-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
+; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512DQ-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512DQ-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index 11f422d671541..99e8cdb179c8d 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -4026,10 +4026,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F-FAST:       # %bb.0:
-; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512F-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4062,10 +4062,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ-FAST:       # %bb.0:
-; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4541,9 +4541,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4559,9 +4559,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0

From 78eac466095c205988ef1e2380033d042a169a3b Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Fri, 12 Apr 2024 12:42:41 +0100
Subject: [PATCH 237/886] [Flang][OpenMP][Lower] Use clause operand structures
 (#86802)

This patch updates Flang lowering to use the new set of OpenMP clause
operand structures and their groupings into directive-specific sets of
clause operands.

It simplifies the passing of information from the clause processor and
the creation of operations.

The `DataSharingProcessor` is slightly modified to not hold delayed
privatization state. Instead, optional arguments are added to
`processStep1` which are only passed when delayed privatization is used.
This enables using the clause operand structure for `private` and
removes the need for the ad-hoc `DelayedPrivatizationInfo` structure.

The processing of the `schedule` clause is updated to process the
`chunk` modifier rather than requiring two separate calls to the
`ClauseProcessor`.

Lowering of a block-associated `ordered` construct is updated to emit a
TODO error if the `simd` clause is specified, since it is not currently
supported by the `ClauseProcessor` or later compilation stages.

Removed processing of `schedule` from `omp.simdloop`, as it doesn't
apply to `simd` constructs.
---
 flang/lib/Lower/OpenMP/ClauseProcessor.cpp    | 273 +++++----
 flang/lib/Lower/OpenMP/ClauseProcessor.h      | 109 ++--
 .../lib/Lower/OpenMP/DataSharingProcessor.cpp |  38 +-
 flang/lib/Lower/OpenMP/DataSharingProcessor.h |  45 +-
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 536 +++++++-----------
 .../Dialect/OpenMP/OpenMPClauseOperands.h     |  10 +-
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |   4 +
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  |   6 +
 8 files changed, 459 insertions(+), 562 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index fb24c8d1fe3eb..ae0d8bd37228d 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -162,14 +162,13 @@ getIfClauseOperand(Fortran::lower::AbstractConverter &converter,
                                     ifVal);
 }
 
-static void
-addUseDeviceClause(Fortran::lower::AbstractConverter &converter,
-                   const omp::ObjectList &objects,
-                   llvm::SmallVectorImpl<mlir::Value> &operands,
-                   llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
-                   llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
-                   llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-                       &useDeviceSymbols) {
+static void addUseDeviceClause(
+    Fortran::lower::AbstractConverter &converter,
+    const omp::ObjectList &objects,
+    llvm::SmallVectorImpl<mlir::Value> &operands,
+    llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
+    llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSyms) {
   genObjectList(objects, converter, operands);
   for (mlir::Value &operand : operands) {
     checkMapType(operand.getLoc(), operand.getType());
@@ -177,25 +176,24 @@ addUseDeviceClause(Fortran::lower::AbstractConverter &converter,
     useDeviceLocs.push_back(operand.getLoc());
   }
   for (const omp::Object &object : objects)
-    useDeviceSymbols.push_back(object.id());
+    useDeviceSyms.push_back(object.id());
 }
 
 static void convertLoopBounds(Fortran::lower::AbstractConverter &converter,
                               mlir::Location loc,
-                              llvm::SmallVectorImpl<mlir::Value> &lowerBound,
-                              llvm::SmallVectorImpl<mlir::Value> &upperBound,
-                              llvm::SmallVectorImpl<mlir::Value> &step,
+                              mlir::omp::CollapseClauseOps &result,
                               std::size_t loopVarTypeSize) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   // The types of lower bound, upper bound, and step are converted into the
   // type of the loop variable if necessary.
   mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize);
-  for (unsigned it = 0; it < (unsigned)lowerBound.size(); it++) {
-    lowerBound[it] =
-        firOpBuilder.createConvert(loc, loopVarType, lowerBound[it]);
-    upperBound[it] =
-        firOpBuilder.createConvert(loc, loopVarType, upperBound[it]);
-    step[it] = firOpBuilder.createConvert(loc, loopVarType, step[it]);
+  for (unsigned it = 0; it < (unsigned)result.loopLBVar.size(); it++) {
+    result.loopLBVar[it] =
+        firOpBuilder.createConvert(loc, loopVarType, result.loopLBVar[it]);
+    result.loopUBVar[it] =
+        firOpBuilder.createConvert(loc, loopVarType, result.loopUBVar[it]);
+    result.loopStepVar[it] =
+        firOpBuilder.createConvert(loc, loopVarType, result.loopStepVar[it]);
   }
 }
 
@@ -205,9 +203,7 @@ static void convertLoopBounds(Fortran::lower::AbstractConverter &converter,
 
 bool ClauseProcessor::processCollapse(
     mlir::Location currentLocation, Fortran::lower::pft::Evaluation &eval,
-    llvm::SmallVectorImpl<mlir::Value> &lowerBound,
-    llvm::SmallVectorImpl<mlir::Value> &upperBound,
-    llvm::SmallVectorImpl<mlir::Value> &step,
+    mlir::omp::CollapseClauseOps &result,
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv) const {
   bool found = false;
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
@@ -238,15 +234,15 @@ bool ClauseProcessor::processCollapse(
         std::get_if<Fortran::parser::LoopControl::Bounds>(&loopControl->u);
     assert(bounds && "Expected bounds for worksharing do loop");
     Fortran::lower::StatementContext stmtCtx;
-    lowerBound.push_back(fir::getBase(converter.genExprValue(
+    result.loopLBVar.push_back(fir::getBase(converter.genExprValue(
         *Fortran::semantics::GetExpr(bounds->lower), stmtCtx)));
-    upperBound.push_back(fir::getBase(converter.genExprValue(
+    result.loopUBVar.push_back(fir::getBase(converter.genExprValue(
         *Fortran::semantics::GetExpr(bounds->upper), stmtCtx)));
     if (bounds->step) {
-      step.push_back(fir::getBase(converter.genExprValue(
+      result.loopStepVar.push_back(fir::getBase(converter.genExprValue(
           *Fortran::semantics::GetExpr(bounds->step), stmtCtx)));
     } else { // If `step` is not present, assume it as `1`.
-      step.push_back(firOpBuilder.createIntegerConstant(
+      result.loopStepVar.push_back(firOpBuilder.createIntegerConstant(
           currentLocation, firOpBuilder.getIntegerType(32), 1));
     }
     iv.push_back(bounds->name.thing.symbol);
@@ -257,8 +253,7 @@ bool ClauseProcessor::processCollapse(
         &*std::next(doConstructEval->getNestedEvaluations().begin());
   } while (collapseValue > 0);
 
-  convertLoopBounds(converter, currentLocation, lowerBound, upperBound, step,
-                    loopVarTypeSize);
+  convertLoopBounds(converter, currentLocation, result, loopVarTypeSize);
 
   return found;
 }
@@ -286,7 +281,7 @@ bool ClauseProcessor::processDefault() const {
 }
 
 bool ClauseProcessor::processDevice(Fortran::lower::StatementContext &stmtCtx,
-                                    mlir::Value &result) const {
+                                    mlir::omp::DeviceClauseOps &result) const {
   const Fortran::parser::CharBlock *source = nullptr;
   if (auto *clause = findUniqueClause<omp::clause::Device>(&source)) {
     mlir::Location clauseLocation = converter.genLocation(*source);
@@ -298,25 +293,26 @@ bool ClauseProcessor::processDevice(Fortran::lower::StatementContext &stmtCtx,
       }
     }
     const auto &deviceExpr = std::get<omp::SomeExpr>(clause->t);
-    result = fir::getBase(converter.genExprValue(deviceExpr, stmtCtx));
+    result.deviceVar =
+        fir::getBase(converter.genExprValue(deviceExpr, stmtCtx));
     return true;
   }
   return false;
 }
 
 bool ClauseProcessor::processDeviceType(
-    mlir::omp::DeclareTargetDeviceType &result) const {
+    mlir::omp::DeviceTypeClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::DeviceType>()) {
     // Case: declare target ... device_type(any | host | nohost)
     switch (clause->v) {
     case omp::clause::DeviceType::DeviceTypeDescription::Nohost:
-      result = mlir::omp::DeclareTargetDeviceType::nohost;
+      result.deviceType = mlir::omp::DeclareTargetDeviceType::nohost;
       break;
     case omp::clause::DeviceType::DeviceTypeDescription::Host:
-      result = mlir::omp::DeclareTargetDeviceType::host;
+      result.deviceType = mlir::omp::DeclareTargetDeviceType::host;
       break;
     case omp::clause::DeviceType::DeviceTypeDescription::Any:
-      result = mlir::omp::DeclareTargetDeviceType::any;
+      result.deviceType = mlir::omp::DeclareTargetDeviceType::any;
       break;
     }
     return true;
@@ -325,7 +321,7 @@ bool ClauseProcessor::processDeviceType(
 }
 
 bool ClauseProcessor::processFinal(Fortran::lower::StatementContext &stmtCtx,
-                                   mlir::Value &result) const {
+                                   mlir::omp::FinalClauseOps &result) const {
   const Fortran::parser::CharBlock *source = nullptr;
   if (auto *clause = findUniqueClause<omp::clause::Final>(&source)) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
@@ -333,100 +329,108 @@ bool ClauseProcessor::processFinal(Fortran::lower::StatementContext &stmtCtx,
 
     mlir::Value finalVal =
         fir::getBase(converter.genExprValue(clause->v, stmtCtx));
-    result = firOpBuilder.createConvert(clauseLocation,
-                                        firOpBuilder.getI1Type(), finalVal);
+    result.finalVar = firOpBuilder.createConvert(
+        clauseLocation, firOpBuilder.getI1Type(), finalVal);
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processHint(mlir::IntegerAttr &result) const {
+bool ClauseProcessor::processHint(mlir::omp::HintClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Hint>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
     int64_t hintValue = *Fortran::evaluate::ToInt64(clause->v);
-    result = firOpBuilder.getI64IntegerAttr(hintValue);
+    result.hintAttr = firOpBuilder.getI64IntegerAttr(hintValue);
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processMergeable(mlir::UnitAttr &result) const {
-  return markClauseOccurrence<omp::clause::Mergeable>(result);
+bool ClauseProcessor::processMergeable(
+    mlir::omp::MergeableClauseOps &result) const {
+  return markClauseOccurrence<omp::clause::Mergeable>(result.mergeableAttr);
 }
 
-bool ClauseProcessor::processNowait(mlir::UnitAttr &result) const {
-  return markClauseOccurrence<omp::clause::Nowait>(result);
+bool ClauseProcessor::processNowait(mlir::omp::NowaitClauseOps &result) const {
+  return markClauseOccurrence<omp::clause::Nowait>(result.nowaitAttr);
 }
 
-bool ClauseProcessor::processNumTeams(Fortran::lower::StatementContext &stmtCtx,
-                                      mlir::Value &result) const {
+bool ClauseProcessor::processNumTeams(
+    Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::NumTeamsClauseOps &result) const {
   // TODO Get lower and upper bounds for num_teams when parser is updated to
   // accept both.
   if (auto *clause = findUniqueClause<omp::clause::NumTeams>()) {
     // auto lowerBound = std::get<std::optional<ExprTy>>(clause->t);
     auto &upperBound = std::get<ExprTy>(clause->t);
-    result = fir::getBase(converter.genExprValue(upperBound, stmtCtx));
+    result.numTeamsUpperVar =
+        fir::getBase(converter.genExprValue(upperBound, stmtCtx));
     return true;
   }
   return false;
 }
 
 bool ClauseProcessor::processNumThreads(
-    Fortran::lower::StatementContext &stmtCtx, mlir::Value &result) const {
+    Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::NumThreadsClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::NumThreads>()) {
     // OMPIRBuilder expects `NUM_THREADS` clause as a `Value`.
-    result = fir::getBase(converter.genExprValue(clause->v, stmtCtx));
+    result.numThreadsVar =
+        fir::getBase(converter.genExprValue(clause->v, stmtCtx));
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processOrdered(mlir::IntegerAttr &result) const {
+bool ClauseProcessor::processOrdered(
+    mlir::omp::OrderedClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Ordered>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
     int64_t orderedClauseValue = 0l;
     if (clause->v.has_value())
       orderedClauseValue = *Fortran::evaluate::ToInt64(*clause->v);
-    result = firOpBuilder.getI64IntegerAttr(orderedClauseValue);
+    result.orderedAttr = firOpBuilder.getI64IntegerAttr(orderedClauseValue);
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processPriority(Fortran::lower::StatementContext &stmtCtx,
-                                      mlir::Value &result) const {
+bool ClauseProcessor::processPriority(
+    Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::PriorityClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Priority>()) {
-    result = fir::getBase(converter.genExprValue(clause->v, stmtCtx));
+    result.priorityVar =
+        fir::getBase(converter.genExprValue(clause->v, stmtCtx));
     return true;
   }
   return false;
 }
 
 bool ClauseProcessor::processProcBind(
-    mlir::omp::ClauseProcBindKindAttr &result) const {
+    mlir::omp::ProcBindClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::ProcBind>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-    result = genProcBindKindAttr(firOpBuilder, *clause);
+    result.procBindKindAttr = genProcBindKindAttr(firOpBuilder, *clause);
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processSafelen(mlir::IntegerAttr &result) const {
+bool ClauseProcessor::processSafelen(
+    mlir::omp::SafelenClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Safelen>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
     const std::optional<std::int64_t> safelenVal =
         Fortran::evaluate::ToInt64(clause->v);
-    result = firOpBuilder.getI64IntegerAttr(*safelenVal);
+    result.safelenAttr = firOpBuilder.getI64IntegerAttr(*safelenVal);
     return true;
   }
   return false;
 }
 
 bool ClauseProcessor::processSchedule(
-    mlir::omp::ClauseScheduleKindAttr &valAttr,
-    mlir::omp::ScheduleModifierAttr &modifierAttr,
-    mlir::UnitAttr &simdModifierAttr) const {
+    Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::ScheduleClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Schedule>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
     mlir::MLIRContext *context = firOpBuilder.getContext();
@@ -451,53 +455,51 @@ bool ClauseProcessor::processSchedule(
       break;
     }
 
-    mlir::omp::ScheduleModifier scheduleModifier = getScheduleModifier(*clause);
+    result.scheduleValAttr =
+        mlir::omp::ClauseScheduleKindAttr::get(context, scheduleKind);
 
+    mlir::omp::ScheduleModifier scheduleModifier = getScheduleModifier(*clause);
     if (scheduleModifier != mlir::omp::ScheduleModifier::none)
-      modifierAttr =
+      result.scheduleModAttr =
           mlir::omp::ScheduleModifierAttr::get(context, scheduleModifier);
 
     if (getSimdModifier(*clause) != mlir::omp::ScheduleModifier::none)
-      simdModifierAttr = firOpBuilder.getUnitAttr();
+      result.scheduleSimdAttr = firOpBuilder.getUnitAttr();
 
-    valAttr = mlir::omp::ClauseScheduleKindAttr::get(context, scheduleKind);
-    return true;
-  }
-  return false;
-}
-
-bool ClauseProcessor::processScheduleChunk(
-    Fortran::lower::StatementContext &stmtCtx, mlir::Value &result) const {
-  if (auto *clause = findUniqueClause<omp::clause::Schedule>()) {
     if (const auto &chunkExpr = std::get<omp::MaybeExpr>(clause->t))
-      result = fir::getBase(converter.genExprValue(*chunkExpr, stmtCtx));
+      result.scheduleChunkVar =
+          fir::getBase(converter.genExprValue(*chunkExpr, stmtCtx));
+
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processSimdlen(mlir::IntegerAttr &result) const {
+bool ClauseProcessor::processSimdlen(
+    mlir::omp::SimdlenClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::Simdlen>()) {
     fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
     const std::optional<std::int64_t> simdlenVal =
         Fortran::evaluate::ToInt64(clause->v);
-    result = firOpBuilder.getI64IntegerAttr(*simdlenVal);
+    result.simdlenAttr = firOpBuilder.getI64IntegerAttr(*simdlenVal);
     return true;
   }
   return false;
 }
 
 bool ClauseProcessor::processThreadLimit(
-    Fortran::lower::StatementContext &stmtCtx, mlir::Value &result) const {
+    Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::ThreadLimitClauseOps &result) const {
   if (auto *clause = findUniqueClause<omp::clause::ThreadLimit>()) {
-    result = fir::getBase(converter.genExprValue(clause->v, stmtCtx));
+    result.threadLimitVar =
+        fir::getBase(converter.genExprValue(clause->v, stmtCtx));
     return true;
   }
   return false;
 }
 
-bool ClauseProcessor::processUntied(mlir::UnitAttr &result) const {
-  return markClauseOccurrence<omp::clause::Untied>(result);
+bool ClauseProcessor::processUntied(mlir::omp::UntiedClauseOps &result) const {
+  return markClauseOccurrence<omp::clause::Untied>(result.untiedAttr);
 }
 
 //===----------------------------------------------------------------------===//
@@ -505,13 +507,12 @@ bool ClauseProcessor::processUntied(mlir::UnitAttr &result) const {
 //===----------------------------------------------------------------------===//
 
 bool ClauseProcessor::processAllocate(
-    llvm::SmallVectorImpl<mlir::Value> &allocatorOperands,
-    llvm::SmallVectorImpl<mlir::Value> &allocateOperands) const {
+    mlir::omp::AllocateClauseOps &result) const {
   return findRepeatableClause<omp::clause::Allocate>(
       [&](const omp::clause::Allocate &clause,
           const Fortran::parser::CharBlock &) {
-        genAllocateClause(converter, clause, allocatorOperands,
-                          allocateOperands);
+        genAllocateClause(converter, clause, result.allocatorVars,
+                          result.allocateVars);
       });
 }
 
@@ -660,10 +661,9 @@ createCopyFunc(mlir::Location loc, Fortran::lower::AbstractConverter &converter,
   return funcOp;
 }
 
-bool ClauseProcessor::processCopyPrivate(
+bool ClauseProcessor::processCopyprivate(
     mlir::Location currentLocation,
-    llvm::SmallVectorImpl<mlir::Value> &copyPrivateVars,
-    llvm::SmallVectorImpl<mlir::Attribute> &copyPrivateFuncs) const {
+    mlir::omp::CopyprivateClauseOps &result) const {
   auto addCopyPrivateVar = [&](Fortran::semantics::Symbol *sym) {
     mlir::Value symVal = converter.getSymbolAddress(*sym);
     auto declOp = symVal.getDefiningOp<hlfir::DeclareOp>();
@@ -690,10 +690,10 @@ bool ClauseProcessor::processCopyPrivate(
       cpVar = alloca;
     }
 
-    copyPrivateVars.push_back(cpVar);
+    result.copyprivateVars.push_back(cpVar);
     mlir::func::FuncOp funcOp =
         createCopyFunc(currentLocation, converter, cpVar.getType(), attrs);
-    copyPrivateFuncs.push_back(mlir::SymbolRefAttr::get(funcOp));
+    result.copyprivateFuncs.push_back(mlir::SymbolRefAttr::get(funcOp));
   };
 
   bool hasCopyPrivate = findRepeatableClause<clause::Copyprivate>(
@@ -714,9 +714,7 @@ bool ClauseProcessor::processCopyPrivate(
   return hasCopyPrivate;
 }
 
-bool ClauseProcessor::processDepend(
-    llvm::SmallVectorImpl<mlir::Attribute> &dependTypeOperands,
-    llvm::SmallVectorImpl<mlir::Value> &dependOperands) const {
+bool ClauseProcessor::processDepend(mlir::omp::DependClauseOps &result) const {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
 
   return findRepeatableClause<omp::clause::Depend>(
@@ -731,7 +729,7 @@ bool ClauseProcessor::processDepend(
 
         mlir::omp::ClauseTaskDependAttr dependTypeOperand =
             genDependKindAttr(firOpBuilder, kind);
-        dependTypeOperands.append(objects.size(), dependTypeOperand);
+        result.dependTypeAttrs.append(objects.size(), dependTypeOperand);
 
         for (const omp::Object &object : objects) {
           assert(object.ref() && "Expecting designator");
@@ -746,13 +744,13 @@ bool ClauseProcessor::processDepend(
 
           Fortran::semantics::Symbol *sym = object.id();
           const mlir::Value variable = converter.getSymbolAddress(*sym);
-          dependOperands.push_back(variable);
+          result.dependVars.push_back(variable);
         }
       });
 }
 
 bool ClauseProcessor::processHasDeviceAddr(
-    llvm::SmallVectorImpl<mlir::Value> &operands,
+    mlir::omp::HasDeviceAddrClauseOps &result,
     llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
     llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &isDeviceSymbols)
@@ -760,14 +758,14 @@ bool ClauseProcessor::processHasDeviceAddr(
   return findRepeatableClause<omp::clause::HasDeviceAddr>(
       [&](const omp::clause::HasDeviceAddr &devAddrClause,
           const Fortran::parser::CharBlock &) {
-        addUseDeviceClause(converter, devAddrClause.v, operands, isDeviceTypes,
-                           isDeviceLocs, isDeviceSymbols);
+        addUseDeviceClause(converter, devAddrClause.v, result.hasDeviceAddrVars,
+                           isDeviceTypes, isDeviceLocs, isDeviceSymbols);
       });
 }
 
 bool ClauseProcessor::processIf(
     omp::clause::If::DirectiveNameModifier directiveName,
-    mlir::Value &result) const {
+    mlir::omp::IfClauseOps &result) const {
   bool found = false;
   findRepeatableClause<omp::clause::If>(
       [&](const omp::clause::If &clause,
@@ -778,7 +776,7 @@ bool ClauseProcessor::processIf(
         // Assume that, at most, a single 'if' clause will be applicable to the
         // given directive.
         if (operand) {
-          result = operand;
+          result.ifVar = operand;
           found = true;
         }
       });
@@ -786,7 +784,7 @@ bool ClauseProcessor::processIf(
 }
 
 bool ClauseProcessor::processIsDevicePtr(
-    llvm::SmallVectorImpl<mlir::Value> &operands,
+    mlir::omp::IsDevicePtrClauseOps &result,
     llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
     llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &isDeviceSymbols)
@@ -794,8 +792,8 @@ bool ClauseProcessor::processIsDevicePtr(
   return findRepeatableClause<omp::clause::IsDevicePtr>(
       [&](const omp::clause::IsDevicePtr &devPtrClause,
           const Fortran::parser::CharBlock &) {
-        addUseDeviceClause(converter, devPtrClause.v, operands, isDeviceTypes,
-                           isDeviceLocs, isDeviceSymbols);
+        addUseDeviceClause(converter, devPtrClause.v, result.isDevicePtrVars,
+                           isDeviceTypes, isDeviceLocs, isDeviceSymbols);
       });
 }
 
@@ -835,12 +833,10 @@ createMapInfoOp(fir::FirOpBuilder &builder, mlir::Location loc,
 
 bool ClauseProcessor::processMap(
     mlir::Location currentLocation, const llvm::omp::Directive &directive,
-    Fortran::lower::StatementContext &stmtCtx,
-    llvm::SmallVectorImpl<mlir::Value> &mapOperands,
-    llvm::SmallVectorImpl<mlir::Type> *mapSymTypes,
+    Fortran::lower::StatementContext &stmtCtx, mlir::omp::MapClauseOps &result,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *mapSyms,
     llvm::SmallVectorImpl<mlir::Location> *mapSymLocs,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *mapSymbols)
-    const {
+    llvm::SmallVectorImpl<mlir::Type> *mapSymTypes) const {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   return findRepeatableClause<omp::clause::Map>(
       [&](const omp::clause::Map &clause,
@@ -915,25 +911,23 @@ bool ClauseProcessor::processMap(
                   mapTypeBits),
               mlir::omp::VariableCaptureKind::ByRef, symAddr.getType());
 
-          mapOperands.push_back(mapOp);
-          if (mapSymTypes)
-            mapSymTypes->push_back(symAddr.getType());
+          result.mapVars.push_back(mapOp);
+
+          if (mapSyms)
+            mapSyms->push_back(object.id());
           if (mapSymLocs)
             mapSymLocs->push_back(symAddr.getLoc());
-
-          if (mapSymbols)
-            mapSymbols->push_back(object.id());
+          if (mapSymTypes)
+            mapSymTypes->push_back(symAddr.getType());
         }
       });
 }
 
 bool ClauseProcessor::processReduction(
-    mlir::Location currentLocation,
-    llvm::SmallVectorImpl<mlir::Value> &outReductionVars,
-    llvm::SmallVectorImpl<mlir::Type> &outReductionTypes,
-    llvm::SmallVectorImpl<mlir::Attribute> &outReductionDeclSymbols,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-        *outReductionSymbols) const {
+    mlir::Location currentLocation, mlir::omp::ReductionClauseOps &result,
+    llvm::SmallVectorImpl<mlir::Type> *outReductionTypes,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *outReductionSyms)
+    const {
   return findRepeatableClause<omp::clause::Reduction>(
       [&](const omp::clause::Reduction &clause,
           const Fortran::parser::CharBlock &) {
@@ -943,30 +937,31 @@ bool ClauseProcessor::processReduction(
         // whether to do the reduction byref.
         llvm::SmallVector<mlir::Value> reductionVars;
         llvm::SmallVector<mlir::Attribute> reductionDeclSymbols;
-        llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSymbols;
+        llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
         ReductionProcessor rp;
         rp.addDeclareReduction(currentLocation, converter, clause,
                                reductionVars, reductionDeclSymbols,
-                               outReductionSymbols ? &reductionSymbols
-                                                   : nullptr);
+                               outReductionSyms ? &reductionSyms : nullptr);
 
         // Copy local lists into the output.
-        llvm::copy(reductionVars, std::back_inserter(outReductionVars));
+        llvm::copy(reductionVars, std::back_inserter(result.reductionVars));
         llvm::copy(reductionDeclSymbols,
-                   std::back_inserter(outReductionDeclSymbols));
-        if (outReductionSymbols)
-          llvm::copy(reductionSymbols,
-                     std::back_inserter(*outReductionSymbols));
-
-        outReductionTypes.reserve(outReductionTypes.size() +
-                                  reductionVars.size());
-        llvm::transform(reductionVars, std::back_inserter(outReductionTypes),
-                        [](mlir::Value v) { return v.getType(); });
+                   std::back_inserter(result.reductionDeclSymbols));
+
+        if (outReductionTypes) {
+          outReductionTypes->reserve(outReductionTypes->size() +
+                                     reductionVars.size());
+          llvm::transform(reductionVars, std::back_inserter(*outReductionTypes),
+                          [](mlir::Value v) { return v.getType(); });
+        }
+
+        if (outReductionSyms)
+          llvm::copy(reductionSyms, std::back_inserter(*outReductionSyms));
       });
 }
 
 bool ClauseProcessor::processSectionsReduction(
-    mlir::Location currentLocation) const {
+    mlir::Location currentLocation, mlir::omp::ReductionClauseOps &) const {
   return findRepeatableClause<omp::clause::Reduction>(
       [&](const omp::clause::Reduction &, const Fortran::parser::CharBlock &) {
         TODO(currentLocation, "OMPC_Reduction");
@@ -995,30 +990,30 @@ bool ClauseProcessor::processEnter(
 }
 
 bool ClauseProcessor::processUseDeviceAddr(
-    llvm::SmallVectorImpl<mlir::Value> &operands,
+    mlir::omp::UseDeviceClauseOps &result,
     llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
     llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSymbols)
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSyms)
     const {
   return findRepeatableClause<omp::clause::UseDeviceAddr>(
       [&](const omp::clause::UseDeviceAddr &clause,
           const Fortran::parser::CharBlock &) {
-        addUseDeviceClause(converter, clause.v, operands, useDeviceTypes,
-                           useDeviceLocs, useDeviceSymbols);
+        addUseDeviceClause(converter, clause.v, result.useDeviceAddrVars,
+                           useDeviceTypes, useDeviceLocs, useDeviceSyms);
       });
 }
 
 bool ClauseProcessor::processUseDevicePtr(
-    llvm::SmallVectorImpl<mlir::Value> &operands,
+    mlir::omp::UseDeviceClauseOps &result,
     llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
     llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSymbols)
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSyms)
     const {
   return findRepeatableClause<omp::clause::UseDevicePtr>(
       [&](const omp::clause::UseDevicePtr &clause,
           const Fortran::parser::CharBlock &) {
-        addUseDeviceClause(converter, clause.v, operands, useDeviceTypes,
-                           useDeviceLocs, useDeviceSymbols);
+        addUseDeviceClause(converter, clause.v, result.useDevicePtrVars,
+                           useDeviceTypes, useDeviceLocs, useDeviceSyms);
       });
 }
 
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index df8f4f5310fcb..aa2c14b61e756 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -37,7 +37,7 @@ namespace omp {
 /// corresponding clause if it is present in the clause list. Otherwise, they
 /// will return `false` to signal that the clause was not found.
 ///
-/// The intended use is of this class is to move clause processing outside of
+/// The intended use of this class is to move clause processing outside of
 /// construct processing, since the same clauses can appear attached to
 /// different constructs and constructs can be combined, so that code
 /// duplication is minimized.
@@ -56,61 +56,51 @@ class ClauseProcessor {
   // 'Unique' clauses: They can appear at most once in the clause list.
   bool processCollapse(
       mlir::Location currentLocation, Fortran::lower::pft::Evaluation &eval,
-      llvm::SmallVectorImpl<mlir::Value> &lowerBound,
-      llvm::SmallVectorImpl<mlir::Value> &upperBound,
-      llvm::SmallVectorImpl<mlir::Value> &step,
+      mlir::omp::CollapseClauseOps &result,
       llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv) const;
   bool processDefault() const;
   bool processDevice(Fortran::lower::StatementContext &stmtCtx,
-                     mlir::Value &result) const;
-  bool processDeviceType(mlir::omp::DeclareTargetDeviceType &result) const;
+                     mlir::omp::DeviceClauseOps &result) const;
+  bool processDeviceType(mlir::omp::DeviceTypeClauseOps &result) const;
   bool processFinal(Fortran::lower::StatementContext &stmtCtx,
-                    mlir::Value &result) const;
+                    mlir::omp::FinalClauseOps &result) const;
   bool
-  processHasDeviceAddr(llvm::SmallVectorImpl<mlir::Value> &operands,
+  processHasDeviceAddr(mlir::omp::HasDeviceAddrClauseOps &result,
                        llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
                        llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
                        llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
                            &isDeviceSymbols) const;
-  bool processHint(mlir::IntegerAttr &result) const;
-  bool processMergeable(mlir::UnitAttr &result) const;
-  bool processNowait(mlir::UnitAttr &result) const;
+  bool processHint(mlir::omp::HintClauseOps &result) const;
+  bool processMergeable(mlir::omp::MergeableClauseOps &result) const;
+  bool processNowait(mlir::omp::NowaitClauseOps &result) const;
   bool processNumTeams(Fortran::lower::StatementContext &stmtCtx,
-                       mlir::Value &result) const;
+                       mlir::omp::NumTeamsClauseOps &result) const;
   bool processNumThreads(Fortran::lower::StatementContext &stmtCtx,
-                         mlir::Value &result) const;
-  bool processOrdered(mlir::IntegerAttr &result) const;
+                         mlir::omp::NumThreadsClauseOps &result) const;
+  bool processOrdered(mlir::omp::OrderedClauseOps &result) const;
   bool processPriority(Fortran::lower::StatementContext &stmtCtx,
-                       mlir::Value &result) const;
-  bool processProcBind(mlir::omp::ClauseProcBindKindAttr &result) const;
-  bool processSafelen(mlir::IntegerAttr &result) const;
-  bool processSchedule(mlir::omp::ClauseScheduleKindAttr &valAttr,
-                       mlir::omp::ScheduleModifierAttr &modifierAttr,
-                       mlir::UnitAttr &simdModifierAttr) const;
-  bool processScheduleChunk(Fortran::lower::StatementContext &stmtCtx,
-                            mlir::Value &result) const;
-  bool processSimdlen(mlir::IntegerAttr &result) const;
+                       mlir::omp::PriorityClauseOps &result) const;
+  bool processProcBind(mlir::omp::ProcBindClauseOps &result) const;
+  bool processSafelen(mlir::omp::SafelenClauseOps &result) const;
+  bool processSchedule(Fortran::lower::StatementContext &stmtCtx,
+                       mlir::omp::ScheduleClauseOps &result) const;
+  bool processSimdlen(mlir::omp::SimdlenClauseOps &result) const;
   bool processThreadLimit(Fortran::lower::StatementContext &stmtCtx,
-                          mlir::Value &result) const;
-  bool processUntied(mlir::UnitAttr &result) const;
+                          mlir::omp::ThreadLimitClauseOps &result) const;
+  bool processUntied(mlir::omp::UntiedClauseOps &result) const;
 
   // 'Repeatable' clauses: They can appear multiple times in the clause list.
-  bool
-  processAllocate(llvm::SmallVectorImpl<mlir::Value> &allocatorOperands,
-                  llvm::SmallVectorImpl<mlir::Value> &allocateOperands) const;
+  bool processAllocate(mlir::omp::AllocateClauseOps &result) const;
   bool processCopyin() const;
-  bool processCopyPrivate(
-      mlir::Location currentLocation,
-      llvm::SmallVectorImpl<mlir::Value> &copyPrivateVars,
-      llvm::SmallVectorImpl<mlir::Attribute> &copyPrivateFuncs) const;
-  bool processDepend(llvm::SmallVectorImpl<mlir::Attribute> &dependTypeOperands,
-                     llvm::SmallVectorImpl<mlir::Value> &dependOperands) const;
+  bool processCopyprivate(mlir::Location currentLocation,
+                          mlir::omp::CopyprivateClauseOps &result) const;
+  bool processDepend(mlir::omp::DependClauseOps &result) const;
   bool
   processEnter(llvm::SmallVectorImpl<DeclareTargetCapturePair> &result) const;
   bool processIf(omp::clause::If::DirectiveNameModifier directiveName,
-                 mlir::Value &result) const;
+                 mlir::omp::IfClauseOps &result) const;
   bool
-  processIsDevicePtr(llvm::SmallVectorImpl<mlir::Value> &operands,
+  processIsDevicePtr(mlir::omp::IsDevicePtrClauseOps &result,
                      llvm::SmallVectorImpl<mlir::Type> &isDeviceTypes,
                      llvm::SmallVectorImpl<mlir::Location> &isDeviceLocs,
                      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
@@ -119,43 +109,42 @@ class ClauseProcessor {
   processLink(llvm::SmallVectorImpl<DeclareTargetCapturePair> &result) const;
 
   // This method is used to process a map clause.
-  // The optional parameters - mapSymTypes, mapSymLocs & mapSymbols are used to
+  // The optional parameters - mapSymTypes, mapSymLocs & mapSyms are used to
   // store the original type, location and Fortran symbol for the map operands.
   // They may be used later on to create the block_arguments for some of the
   // target directives that require it.
-  bool processMap(mlir::Location currentLocation,
-                  const llvm::omp::Directive &directive,
-                  Fortran::lower::StatementContext &stmtCtx,
-                  llvm::SmallVectorImpl<mlir::Value> &mapOperands,
-                  llvm::SmallVectorImpl<mlir::Type> *mapSymTypes = nullptr,
-                  llvm::SmallVectorImpl<mlir::Location> *mapSymLocs = nullptr,
-                  llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-                      *mapSymbols = nullptr) const;
-  bool
-  processReduction(mlir::Location currentLocation,
-                   llvm::SmallVectorImpl<mlir::Value> &reductionVars,
-                   llvm::SmallVectorImpl<mlir::Type> &reductionTypes,
-                   llvm::SmallVectorImpl<mlir::Attribute> &reductionDeclSymbols,
-                   llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-                       *reductionSymbols = nullptr) const;
-  bool processSectionsReduction(mlir::Location currentLocation) const;
+  bool processMap(
+      mlir::Location currentLocation, const llvm::omp::Directive &directive,
+      Fortran::lower::StatementContext &stmtCtx,
+      mlir::omp::MapClauseOps &result,
+      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *mapSyms =
+          nullptr,
+      llvm::SmallVectorImpl<mlir::Location> *mapSymLocs = nullptr,
+      llvm::SmallVectorImpl<mlir::Type> *mapSymTypes = nullptr) const;
+  bool processReduction(
+      mlir::Location currentLocation, mlir::omp::ReductionClauseOps &result,
+      llvm::SmallVectorImpl<mlir::Type> *reductionTypes = nullptr,
+      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *reductionSyms =
+          nullptr) const;
+  bool processSectionsReduction(mlir::Location currentLocation,
+                                mlir::omp::ReductionClauseOps &result) const;
   bool processTo(llvm::SmallVectorImpl<DeclareTargetCapturePair> &result) const;
   bool
-  processUseDeviceAddr(llvm::SmallVectorImpl<mlir::Value> &operands,
+  processUseDeviceAddr(mlir::omp::UseDeviceClauseOps &result,
                        llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
                        llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
                        llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-                           &useDeviceSymbols) const;
+                           &useDeviceSyms) const;
   bool
-  processUseDevicePtr(llvm::SmallVectorImpl<mlir::Value> &operands,
+  processUseDevicePtr(mlir::omp::UseDeviceClauseOps &result,
                       llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
                       llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
                       llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-                          &useDeviceSymbols) const;
+                          &useDeviceSyms) const;
 
   template <typename T>
   bool processMotionClauses(Fortran::lower::StatementContext &stmtCtx,
-                            llvm::SmallVectorImpl<mlir::Value> &mapOperands);
+                            mlir::omp::MapClauseOps &result);
 
   // Call this method for these clauses that should be supported but are not
   // implemented yet. It triggers a compilation error if any of the given
@@ -197,7 +186,7 @@ class ClauseProcessor {
 template <typename T>
 bool ClauseProcessor::processMotionClauses(
     Fortran::lower::StatementContext &stmtCtx,
-    llvm::SmallVectorImpl<mlir::Value> &mapOperands) {
+    mlir::omp::MapClauseOps &result) {
   return findRepeatableClause<T>(
       [&](const T &clause, const Fortran::parser::CharBlock &source) {
         mlir::Location clauseLocation = converter.genLocation(source);
@@ -239,7 +228,7 @@ bool ClauseProcessor::processMotionClauses(
                   mapTypeBits),
               mlir::omp::VariableCaptureKind::ByRef, symAddr.getType());
 
-          mapOperands.push_back(mapOp);
+          result.mapVars.push_back(mapOp);
         }
       });
 }
diff --git a/flang/lib/Lower/OpenMP/DataSharingProcessor.cpp b/flang/lib/Lower/OpenMP/DataSharingProcessor.cpp
index e114ab9f4548a..5a42e6a6aa417 100644
--- a/flang/lib/Lower/OpenMP/DataSharingProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/DataSharingProcessor.cpp
@@ -23,11 +23,13 @@ namespace Fortran {
 namespace lower {
 namespace omp {
 
-void DataSharingProcessor::processStep1() {
+void DataSharingProcessor::processStep1(
+    mlir::omp::PrivateClauseOps *clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms) {
   collectSymbolsForPrivatization();
   collectDefaultSymbols();
-  privatize();
-  defaultPrivatize();
+  privatize(clauseOps, privateSyms);
+  defaultPrivatize(clauseOps, privateSyms);
   insertBarrier();
 }
 
@@ -299,14 +301,16 @@ void DataSharingProcessor::collectDefaultSymbols() {
   }
 }
 
-void DataSharingProcessor::privatize() {
+void DataSharingProcessor::privatize(
+    mlir::omp::PrivateClauseOps *clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms) {
   for (const Fortran::semantics::Symbol *sym : privatizedSymbols) {
     if (const auto *commonDet =
             sym->detailsIf<Fortran::semantics::CommonBlockDetails>()) {
       for (const auto &mem : commonDet->objects())
-        doPrivatize(&*mem);
+        doPrivatize(&*mem, clauseOps, privateSyms);
     } else
-      doPrivatize(sym);
+      doPrivatize(sym, clauseOps, privateSyms);
   }
 }
 
@@ -323,7 +327,9 @@ void DataSharingProcessor::copyLastPrivatize(mlir::Operation *op) {
     }
 }
 
-void DataSharingProcessor::defaultPrivatize() {
+void DataSharingProcessor::defaultPrivatize(
+    mlir::omp::PrivateClauseOps *clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms) {
   for (const Fortran::semantics::Symbol *sym : defaultSymbols) {
     if (!Fortran::semantics::IsProcedure(*sym) &&
         !sym->GetUltimate().has<Fortran::semantics::DerivedTypeDetails>() &&
@@ -331,11 +337,14 @@ void DataSharingProcessor::defaultPrivatize() {
         !symbolsInNestedRegions.contains(sym) &&
         !symbolsInParentRegions.contains(sym) &&
         !privatizedSymbols.contains(sym))
-      doPrivatize(sym);
+      doPrivatize(sym, clauseOps, privateSyms);
   }
 }
 
-void DataSharingProcessor::doPrivatize(const Fortran::semantics::Symbol *sym) {
+void DataSharingProcessor::doPrivatize(
+    const Fortran::semantics::Symbol *sym,
+    mlir::omp::PrivateClauseOps *clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms) {
   if (!useDelayedPrivatization) {
     cloneSymbol(sym);
     copyFirstPrivateSymbol(sym);
@@ -419,10 +428,13 @@ void DataSharingProcessor::doPrivatize(const Fortran::semantics::Symbol *sym) {
     return result;
   }();
 
-  delayedPrivatizationInfo.privatizers.push_back(
-      mlir::SymbolRefAttr::get(privatizerOp));
-  delayedPrivatizationInfo.originalAddresses.push_back(hsb.getAddr());
-  delayedPrivatizationInfo.symbols.push_back(sym);
+  if (clauseOps) {
+    clauseOps->privatizers.push_back(mlir::SymbolRefAttr::get(privatizerOp));
+    clauseOps->privateVars.push_back(hsb.getAddr());
+  }
+
+  if (privateSyms)
+    privateSyms->push_back(sym);
 }
 
 } // namespace omp
diff --git a/flang/lib/Lower/OpenMP/DataSharingProcessor.h b/flang/lib/Lower/OpenMP/DataSharingProcessor.h
index 1cbc825fd5e11..c11ee299c5d08 100644
--- a/flang/lib/Lower/OpenMP/DataSharingProcessor.h
+++ b/flang/lib/Lower/OpenMP/DataSharingProcessor.h
@@ -19,28 +19,17 @@
 #include "flang/Parser/parse-tree.h"
 #include "flang/Semantics/symbol.h"
 
+namespace mlir {
+namespace omp {
+struct PrivateClauseOps;
+} // namespace omp
+} // namespace mlir
+
 namespace Fortran {
 namespace lower {
 namespace omp {
 
 class DataSharingProcessor {
-public:
-  /// Collects all the information needed for delayed privatization. This can be
-  /// used by ops with data-sharing clauses to properly generate their regions
-  /// (e.g. add region arguments) and map the original SSA values to their
-  /// corresponding OMP region operands.
-  struct DelayedPrivatizationInfo {
-    // The list of symbols referring to delayed privatizer ops (i.e.
-    // `omp.private` ops).
-    llvm::SmallVector<mlir::SymbolRefAttr> privatizers;
-    // SSA values that correspond to "original" values being privatized.
-    // "Original" here means the SSA value outside the OpenMP region from which
-    // a clone is created inside the region.
-    llvm::SmallVector<mlir::Value> originalAddresses;
-    // Fortran symbols corresponding to the above SSA values.
-    llvm::SmallVector<const Fortran::semantics::Symbol *> symbols;
-  };
-
 private:
   bool hasLastPrivateOp;
   mlir::OpBuilder::InsertPoint lastPrivIP;
@@ -57,7 +46,6 @@ class DataSharingProcessor {
   Fortran::lower::pft::Evaluation &eval;
   bool useDelayedPrivatization;
   Fortran::lower::SymMap *symTable;
-  DelayedPrivatizationInfo delayedPrivatizationInfo;
 
   bool needBarrier();
   void collectSymbols(Fortran::semantics::Symbol::Flag flag);
@@ -67,9 +55,16 @@ class DataSharingProcessor {
   void collectSymbolsForPrivatization();
   void insertBarrier();
   void collectDefaultSymbols();
-  void privatize();
-  void defaultPrivatize();
-  void doPrivatize(const Fortran::semantics::Symbol *sym);
+  void privatize(
+      mlir::omp::PrivateClauseOps *clauseOps,
+      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms);
+  void defaultPrivatize(
+      mlir::omp::PrivateClauseOps *clauseOps,
+      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms);
+  void doPrivatize(
+      const Fortran::semantics::Symbol *sym,
+      mlir::omp::PrivateClauseOps *clauseOps,
+      llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *privateSyms);
   void copyLastPrivatize(mlir::Operation *op);
   void insertLastPrivateCompare(mlir::Operation *op);
   void cloneSymbol(const Fortran::semantics::Symbol *sym);
@@ -103,17 +98,15 @@ class DataSharingProcessor {
   // Step2 performs the copying for lastprivates and requires knowledge of the
   // MLIR operation to insert the last private update. Step2 adds
   // dealocation code as well.
-  void processStep1();
+  void processStep1(mlir::omp::PrivateClauseOps *clauseOps = nullptr,
+                    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
+                        *privateSyms = nullptr);
   void processStep2(mlir::Operation *op, bool isLoop);
 
   void setLoopIV(mlir::Value iv) {
     assert(!loopIV && "Loop iteration variable already set");
     loopIV = iv;
   }
-
-  const DelayedPrivatizationInfo &getDelayedPrivatizationInfo() const {
-    return delayedPrivatizationInfo;
-  }
 };
 
 } // namespace omp
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 50ad889052ab0..3dcfe0fd775dc 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -730,19 +730,25 @@ genMasterOp(Fortran::lower::AbstractConverter &converter,
             mlir::Location currentLocation) {
   return genOpWithBody<mlir::omp::MasterOp>(
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested),
-      /*resultTypes=*/mlir::TypeRange());
+          .setGenNested(genNested));
 }
 
 static mlir::omp::OrderedRegionOp
 genOrderedRegionOp(Fortran::lower::AbstractConverter &converter,
                    Fortran::semantics::SemanticsContext &semaCtx,
                    Fortran::lower::pft::Evaluation &eval, bool genNested,
-                   mlir::Location currentLocation) {
+                   mlir::Location currentLocation,
+                   const Fortran::parser::OmpClauseList &clauseList) {
+  mlir::omp::OrderedRegionClauseOps clauseOps;
+
+  ClauseProcessor cp(converter, semaCtx, clauseList);
+  cp.processTODO<clause::Simd>(currentLocation,
+                               llvm::omp::Directive::OMPD_ordered);
+
   return genOpWithBody<mlir::omp::OrderedRegionOp>(
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(genNested),
-      /*simd=*/false);
+      clauseOps);
 }
 
 static mlir::omp::ParallelOp
@@ -753,77 +759,62 @@ genParallelOp(Fortran::lower::AbstractConverter &converter,
               mlir::Location currentLocation,
               const Fortran::parser::OmpClauseList &clauseList,
               bool outerCombined = false) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value ifClauseOperand, numThreadsClauseOperand;
-  mlir::omp::ClauseProcBindKindAttr procBindKindAttr;
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands,
-      reductionVars;
+  mlir::omp::ParallelClauseOps clauseOps;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> privateSyms;
   llvm::SmallVector<mlir::Type> reductionTypes;
-  llvm::SmallVector<mlir::Attribute> reductionDeclSymbols;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSymbols;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_parallel, ifClauseOperand);
-  cp.processNumThreads(stmtCtx, numThreadsClauseOperand);
-  cp.processProcBind(procBindKindAttr);
+  cp.processIf(llvm::omp::Directive::OMPD_parallel, clauseOps);
+  cp.processNumThreads(stmtCtx, clauseOps);
+  cp.processProcBind(clauseOps);
   cp.processDefault();
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processAllocate(clauseOps);
+
   if (!outerCombined)
-    cp.processReduction(currentLocation, reductionVars, reductionTypes,
-                        reductionDeclSymbols, &reductionSymbols);
+    cp.processReduction(currentLocation, clauseOps, &reductionTypes,
+                        &reductionSyms);
+
+  if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
+    clauseOps.reductionByRefAttr = firOpBuilder.getUnitAttr();
 
   auto reductionCallback = [&](mlir::Operation *op) {
-    llvm::SmallVector<mlir::Location> locs(reductionVars.size(),
+    llvm::SmallVector<mlir::Location> locs(clauseOps.reductionVars.size(),
                                            currentLocation);
-    auto *block = converter.getFirOpBuilder().createBlock(&op->getRegion(0), {},
-                                                          reductionTypes, locs);
+    auto *block =
+        firOpBuilder.createBlock(&op->getRegion(0), {}, reductionTypes, locs);
     for (auto [arg, prv] :
-         llvm::zip_equal(reductionSymbols, block->getArguments())) {
+         llvm::zip_equal(reductionSyms, block->getArguments())) {
       converter.bindSymbol(*arg, prv);
     }
-    return reductionSymbols;
+    return reductionSyms;
   };
 
-  mlir::UnitAttr byrefAttr;
-  if (ReductionProcessor::doReductionByRef(reductionVars))
-    byrefAttr = converter.getFirOpBuilder().getUnitAttr();
-
   OpWithBodyGenInfo genInfo =
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(genNested)
           .setOuterCombined(outerCombined)
           .setClauses(&clauseList)
-          .setReductions(&reductionSymbols, &reductionTypes)
+          .setReductions(&reductionSyms, &reductionTypes)
           .setGenRegionEntryCb(reductionCallback);
 
-  if (!enableDelayedPrivatization) {
-    return genOpWithBody<mlir::omp::ParallelOp>(
-        genInfo,
-        /*resultTypes=*/mlir::TypeRange(), ifClauseOperand,
-        numThreadsClauseOperand, allocateOperands, allocatorOperands,
-        reductionVars,
-        reductionDeclSymbols.empty()
-            ? nullptr
-            : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                   reductionDeclSymbols),
-        procBindKindAttr, /*private_vars=*/llvm::SmallVector<mlir::Value>{},
-        /*privatizers=*/nullptr, byrefAttr);
-  }
+  if (!enableDelayedPrivatization)
+    return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
 
   bool privatize = !outerCombined;
   DataSharingProcessor dsp(converter, semaCtx, clauseList, eval,
                            /*useDelayedPrivatization=*/true, &symTable);
 
   if (privatize)
-    dsp.processStep1();
-
-  const auto &delayedPrivatizationInfo = dsp.getDelayedPrivatizationInfo();
+    dsp.processStep1(&clauseOps, &privateSyms);
 
   auto genRegionEntryCB = [&](mlir::Operation *op) {
     auto parallelOp = llvm::cast<mlir::omp::ParallelOp>(op);
 
-    llvm::SmallVector<mlir::Location> reductionLocs(reductionVars.size(),
-                                                    currentLocation);
+    llvm::SmallVector<mlir::Location> reductionLocs(
+        clauseOps.reductionVars.size(), currentLocation);
 
     mlir::OperandRange privateVars = parallelOp.getPrivateVars();
     mlir::Region &region = parallelOp.getRegion();
@@ -838,12 +829,12 @@ genParallelOp(Fortran::lower::AbstractConverter &converter,
     llvm::transform(privateVars, std::back_inserter(privateVarLocs),
                     [](mlir::Value v) { return v.getLoc(); });
 
-    converter.getFirOpBuilder().createBlock(&region, /*insertPt=*/{},
-                                            privateVarTypes, privateVarLocs);
+    firOpBuilder.createBlock(&region, /*insertPt=*/{}, privateVarTypes,
+                             privateVarLocs);
 
     llvm::SmallVector<const Fortran::semantics::Symbol *> allSymbols =
-        reductionSymbols;
-    allSymbols.append(delayedPrivatizationInfo.symbols);
+        reductionSyms;
+    allSymbols.append(privateSyms);
     for (auto [arg, prv] : llvm::zip_equal(allSymbols, region.getArguments())) {
       converter.bindSymbol(*arg, prv);
     }
@@ -853,26 +844,7 @@ genParallelOp(Fortran::lower::AbstractConverter &converter,
 
   // TODO Merge with the reduction CB.
   genInfo.setGenRegionEntryCb(genRegionEntryCB).setDataSharingProcessor(&dsp);
-
-  llvm::SmallVector<mlir::Attribute> privatizers(
-      delayedPrivatizationInfo.privatizers.begin(),
-      delayedPrivatizationInfo.privatizers.end());
-
-  return genOpWithBody<mlir::omp::ParallelOp>(
-      genInfo,
-      /*resultTypes=*/mlir::TypeRange(), ifClauseOperand,
-      numThreadsClauseOperand, allocateOperands, allocatorOperands,
-      reductionVars,
-      reductionDeclSymbols.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 reductionDeclSymbols),
-      procBindKindAttr, delayedPrivatizationInfo.originalAddresses,
-      delayedPrivatizationInfo.privatizers.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 privatizers),
-      byrefAttr);
+  return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
 }
 
 static mlir::omp::SectionOp
@@ -896,28 +868,21 @@ genSingleOp(Fortran::lower::AbstractConverter &converter,
             mlir::Location currentLocation,
             const Fortran::parser::OmpClauseList &beginClauseList,
             const Fortran::parser::OmpClauseList &endClauseList) {
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands;
-  llvm::SmallVector<mlir::Value> copyPrivateVars;
-  llvm::SmallVector<mlir::Attribute> copyPrivateFuncs;
-  mlir::UnitAttr nowaitAttr;
+  mlir::omp::SingleClauseOps clauseOps;
 
   ClauseProcessor cp(converter, semaCtx, beginClauseList);
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processAllocate(clauseOps);
+  // TODO Support delayed privatization.
 
   ClauseProcessor ecp(converter, semaCtx, endClauseList);
-  ecp.processNowait(nowaitAttr);
-  ecp.processCopyPrivate(currentLocation, copyPrivateVars, copyPrivateFuncs);
+  ecp.processNowait(clauseOps);
+  ecp.processCopyprivate(currentLocation, clauseOps);
 
   return genOpWithBody<mlir::omp::SingleOp>(
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(genNested)
           .setClauses(&beginClauseList),
-      allocateOperands, allocatorOperands, copyPrivateVars,
-      copyPrivateFuncs.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 copyPrivateFuncs),
-      nowaitAttr);
+      clauseOps);
 }
 
 static mlir::omp::TaskOp
@@ -927,21 +892,19 @@ genTaskOp(Fortran::lower::AbstractConverter &converter,
           mlir::Location currentLocation,
           const Fortran::parser::OmpClauseList &clauseList) {
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value ifClauseOperand, finalClauseOperand, priorityClauseOperand;
-  mlir::UnitAttr untiedAttr, mergeableAttr;
-  llvm::SmallVector<mlir::Attribute> dependTypeOperands;
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands,
-      dependOperands;
+  mlir::omp::TaskClauseOps clauseOps;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_task, ifClauseOperand);
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processIf(llvm::omp::Directive::OMPD_task, clauseOps);
+  cp.processAllocate(clauseOps);
   cp.processDefault();
-  cp.processFinal(stmtCtx, finalClauseOperand);
-  cp.processUntied(untiedAttr);
-  cp.processMergeable(mergeableAttr);
-  cp.processPriority(stmtCtx, priorityClauseOperand);
-  cp.processDepend(dependTypeOperands, dependOperands);
+  cp.processFinal(stmtCtx, clauseOps);
+  cp.processUntied(clauseOps);
+  cp.processMergeable(clauseOps);
+  cp.processPriority(stmtCtx, clauseOps);
+  cp.processDepend(clauseOps);
+  // TODO Support delayed privatization.
+
   cp.processTODO<clause::InReduction, clause::Detach, clause::Affinity>(
       currentLocation, llvm::omp::Directive::OMPD_task);
 
@@ -949,14 +912,7 @@ genTaskOp(Fortran::lower::AbstractConverter &converter,
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(genNested)
           .setClauses(&clauseList),
-      ifClauseOperand, finalClauseOperand, untiedAttr, mergeableAttr,
-      /*in_reduction_vars=*/mlir::ValueRange(),
-      /*in_reductions=*/nullptr, priorityClauseOperand,
-      dependTypeOperands.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 dependTypeOperands),
-      dependOperands, allocateOperands, allocatorOperands);
+      clauseOps);
 }
 
 static mlir::omp::TaskgroupOp
@@ -965,17 +921,18 @@ genTaskgroupOp(Fortran::lower::AbstractConverter &converter,
                Fortran::lower::pft::Evaluation &eval, bool genNested,
                mlir::Location currentLocation,
                const Fortran::parser::OmpClauseList &clauseList) {
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands;
+  mlir::omp::TaskgroupClauseOps clauseOps;
+
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processAllocate(clauseOps);
   cp.processTODO<clause::TaskReduction>(currentLocation,
                                         llvm::omp::Directive::OMPD_taskgroup);
+
   return genOpWithBody<mlir::omp::TaskgroupOp>(
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(genNested)
           .setClauses(&clauseList),
-      /*task_reduction_vars=*/mlir::ValueRange(),
-      /*task_reductions=*/nullptr, allocateOperands, allocatorOperands);
+      clauseOps);
 }
 
 // This helper function implements the functionality of "promoting"
@@ -996,8 +953,7 @@ genTaskgroupOp(Fortran::lower::AbstractConverter &converter,
 //  clause. Support for such list items in a use_device_ptr clause
 //  is deprecated."
 static void promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
-    llvm::SmallVectorImpl<mlir::Value> &devicePtrOperands,
-    llvm::SmallVectorImpl<mlir::Value> &deviceAddrOperands,
+    mlir::omp::UseDeviceClauseOps &clauseOps,
     llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
     llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
@@ -1010,9 +966,10 @@ static void promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
 
   // Iterate over our use_device_ptr list and shift all non-cptr arguments into
   // use_device_addr.
-  for (auto *it = devicePtrOperands.begin(); it != devicePtrOperands.end();) {
+  for (auto *it = clauseOps.useDevicePtrVars.begin();
+       it != clauseOps.useDevicePtrVars.end();) {
     if (!fir::isa_builtin_cptr_type(fir::unwrapRefType(it->getType()))) {
-      deviceAddrOperands.push_back(*it);
+      clauseOps.useDeviceAddrVars.push_back(*it);
       // We have to shuffle the symbols around as well, to maintain
       // the correct Input -> BlockArg for use_device_ptr/use_device_addr.
       // NOTE: However, as map's do not seem to be included currently
@@ -1020,11 +977,11 @@ static void promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
       // future alterations. I believe the reason they are not currently
       // is that the BlockArg assign/lowering needs to be extended
       // to a greater set of types.
-      auto idx = std::distance(devicePtrOperands.begin(), it);
+      auto idx = std::distance(clauseOps.useDevicePtrVars.begin(), it);
       moveElementToBack(idx, useDeviceTypes);
       moveElementToBack(idx, useDeviceLocs);
       moveElementToBack(idx, useDeviceSymbols);
-      it = devicePtrOperands.erase(it);
+      it = clauseOps.useDevicePtrVars.erase(it);
       continue;
     }
     ++it;
@@ -1038,20 +995,19 @@ genTargetDataOp(Fortran::lower::AbstractConverter &converter,
                 mlir::Location currentLocation,
                 const Fortran::parser::OmpClauseList &clauseList) {
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value ifClauseOperand, deviceOperand;
-  llvm::SmallVector<mlir::Value> mapOperands, devicePtrOperands,
-      deviceAddrOperands;
+  mlir::omp::TargetDataClauseOps clauseOps;
   llvm::SmallVector<mlir::Type> useDeviceTypes;
   llvm::SmallVector<mlir::Location> useDeviceLocs;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> useDeviceSymbols;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> useDeviceSyms;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_target_data, ifClauseOperand);
-  cp.processDevice(stmtCtx, deviceOperand);
-  cp.processUseDevicePtr(devicePtrOperands, useDeviceTypes, useDeviceLocs,
-                         useDeviceSymbols);
-  cp.processUseDeviceAddr(deviceAddrOperands, useDeviceTypes, useDeviceLocs,
-                          useDeviceSymbols);
+  cp.processIf(llvm::omp::Directive::OMPD_target_data, clauseOps);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processUseDevicePtr(clauseOps, useDeviceTypes, useDeviceLocs,
+                         useDeviceSyms);
+  cp.processUseDeviceAddr(clauseOps, useDeviceTypes, useDeviceLocs,
+                          useDeviceSyms);
+
   // This function implements the deprecated functionality of use_device_ptr
   // that allows users to provide non-CPTR arguments to it with the caveat
   // that the compiler will treat them as use_device_addr. A lot of legacy
@@ -1063,17 +1019,16 @@ genTargetDataOp(Fortran::lower::AbstractConverter &converter,
   // ordering.
   // TODO: Perhaps create a user provideable compiler option that will
   // re-introduce a hard-error rather than a warning in these cases.
-  promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
-      devicePtrOperands, deviceAddrOperands, useDeviceTypes, useDeviceLocs,
-      useDeviceSymbols);
+  promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(clauseOps, useDeviceTypes,
+                                                useDeviceLocs, useDeviceSyms);
   cp.processMap(currentLocation, llvm::omp::Directive::OMPD_target_data,
-                stmtCtx, mapOperands);
+                stmtCtx, clauseOps);
 
   auto dataOp = converter.getFirOpBuilder().create<mlir::omp::TargetDataOp>(
-      currentLocation, ifClauseOperand, deviceOperand, devicePtrOperands,
-      deviceAddrOperands, mapOperands);
+      currentLocation, clauseOps);
+
   genBodyOfTargetDataOp(converter, semaCtx, eval, genNested, dataOp,
-                        useDeviceTypes, useDeviceLocs, useDeviceSymbols,
+                        useDeviceTypes, useDeviceLocs, useDeviceSyms,
                         currentLocation);
   return dataOp;
 }
@@ -1086,10 +1041,7 @@ static OpTy genTargetEnterExitDataUpdateOp(
     const Fortran::parser::OmpClauseList &clauseList) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value ifClauseOperand, deviceOperand;
-  mlir::UnitAttr nowaitAttr;
-  llvm::SmallVector<mlir::Value> mapOperands, dependOperands;
-  llvm::SmallVector<mlir::Attribute> dependTypeOperands;
+  mlir::omp::TargetEnterExitUpdateDataClauseOps clauseOps;
 
   // GCC 9.3.0 emits a (probably) bogus warning about an unused variable.
   [[maybe_unused]] llvm::omp::Directive directive;
@@ -1104,25 +1056,19 @@ static OpTy genTargetEnterExitDataUpdateOp(
   }
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(directive, ifClauseOperand);
-  cp.processDevice(stmtCtx, deviceOperand);
-  cp.processDepend(dependTypeOperands, dependOperands);
-  cp.processNowait(nowaitAttr);
+  cp.processIf(directive, clauseOps);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processDepend(clauseOps);
+  cp.processNowait(clauseOps);
 
   if constexpr (std::is_same_v<OpTy, mlir::omp::TargetUpdateOp>) {
-    cp.processMotionClauses<clause::To>(stmtCtx, mapOperands);
-    cp.processMotionClauses<clause::From>(stmtCtx, mapOperands);
+    cp.processMotionClauses<clause::To>(stmtCtx, clauseOps);
+    cp.processMotionClauses<clause::From>(stmtCtx, clauseOps);
   } else {
-    cp.processMap(currentLocation, directive, stmtCtx, mapOperands);
+    cp.processMap(currentLocation, directive, stmtCtx, clauseOps);
   }
 
-  return firOpBuilder.create<OpTy>(
-      currentLocation, ifClauseOperand, deviceOperand,
-      dependTypeOperands.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 dependTypeOperands),
-      dependOperands, nowaitAttr, mapOperands);
+  return firOpBuilder.create<OpTy>(currentLocation, clauseOps);
 }
 
 // This functions creates a block for the body of the targetOp's region. It adds
@@ -1132,9 +1078,9 @@ genBodyOfTargetOp(Fortran::lower::AbstractConverter &converter,
                   Fortran::semantics::SemanticsContext &semaCtx,
                   Fortran::lower::pft::Evaluation &eval, bool genNested,
                   mlir::omp::TargetOp &targetOp,
-                  llvm::ArrayRef<mlir::Type> mapSymTypes,
+                  llvm::ArrayRef<const Fortran::semantics::Symbol *> mapSyms,
                   llvm::ArrayRef<mlir::Location> mapSymLocs,
-                  llvm::ArrayRef<const Fortran::semantics::Symbol *> mapSymbols,
+                  llvm::ArrayRef<mlir::Type> mapSymTypes,
                   const mlir::Location &currentLocation) {
   assert(mapSymTypes.size() == mapSymLocs.size());
 
@@ -1163,7 +1109,7 @@ genBodyOfTargetOp(Fortran::lower::AbstractConverter &converter,
   };
 
   // Bind the symbols to their corresponding block arguments.
-  for (auto [argIndex, argSymbol] : llvm::enumerate(mapSymbols)) {
+  for (auto [argIndex, argSymbol] : llvm::enumerate(mapSyms)) {
     const mlir::BlockArgument &arg = region.getArgument(argIndex);
     // Avoid capture of a reference to a structured binding.
     const Fortran::semantics::Symbol *sym = argSymbol;
@@ -1287,31 +1233,25 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
             const Fortran::parser::OmpClauseList &clauseList,
             llvm::omp::Directive directive, bool outerCombined = false) {
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value ifClauseOperand, deviceOperand, threadLimitOperand;
-  mlir::UnitAttr nowaitAttr;
-  llvm::SmallVector<mlir::Attribute> dependTypeOperands;
-  llvm::SmallVector<mlir::Value> mapOperands, dependOperands;
-  llvm::SmallVector<mlir::Type> mapSymTypes;
-  llvm::SmallVector<mlir::Location> mapSymLocs;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> mapSymbols;
-  llvm::SmallVector<mlir::Value> devicePtrOperands, deviceAddrOperands;
-  llvm::SmallVector<mlir::Type> devicePtrTypes, deviceAddrTypes;
-  llvm::SmallVector<mlir::Location> devicePtrLocs, deviceAddrLocs;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> devicePtrSymbols,
-      deviceAddrSymbols;
+  mlir::omp::TargetClauseOps clauseOps;
+  llvm::SmallVector<mlir::Type> mapTypes, devicePtrTypes, deviceAddrTypes;
+  llvm::SmallVector<mlir::Location> mapLocs, devicePtrLocs, deviceAddrLocs;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> mapSyms, devicePtrSyms,
+      deviceAddrSyms;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_target, ifClauseOperand);
-  cp.processDevice(stmtCtx, deviceOperand);
-  cp.processThreadLimit(stmtCtx, threadLimitOperand);
-  cp.processDepend(dependTypeOperands, dependOperands);
-  cp.processNowait(nowaitAttr);
-  cp.processMap(currentLocation, directive, stmtCtx, mapOperands, &mapSymTypes,
-                &mapSymLocs, &mapSymbols);
-  cp.processIsDevicePtr(devicePtrOperands, devicePtrTypes, devicePtrLocs,
-                        devicePtrSymbols);
-  cp.processHasDeviceAddr(deviceAddrOperands, deviceAddrTypes, deviceAddrLocs,
-                          deviceAddrSymbols);
+  cp.processIf(llvm::omp::Directive::OMPD_target, clauseOps);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processThreadLimit(stmtCtx, clauseOps);
+  cp.processDepend(clauseOps);
+  cp.processNowait(clauseOps);
+  cp.processMap(currentLocation, directive, stmtCtx, clauseOps, &mapSyms,
+                &mapLocs, &mapTypes);
+  cp.processIsDevicePtr(clauseOps, devicePtrTypes, devicePtrLocs,
+                        devicePtrSyms);
+  cp.processHasDeviceAddr(clauseOps, deviceAddrTypes, deviceAddrLocs,
+                          deviceAddrSyms);
+  // TODO Support delayed privatization.
 
   cp.processTODO<clause::Private, clause::Firstprivate, clause::Reduction,
                  clause::InReduction, clause::Allocate, clause::UsesAllocators,
@@ -1323,7 +1263,7 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
   // symbols used inside the region that have not been explicitly mapped using
   // the map clause.
   auto captureImplicitMap = [&](const Fortran::semantics::Symbol &sym) {
-    if (llvm::find(mapSymbols, &sym) == mapSymbols.end()) {
+    if (llvm::find(mapSyms, &sym) == mapSyms.end()) {
       mlir::Value baseOp = converter.getSymbolAddress(sym);
       if (!baseOp)
         if (const auto *details = sym.template detailsIf<
@@ -1394,26 +1334,20 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
                 mapFlag),
             captureKind, baseOp.getType());
 
-        mapOperands.push_back(mapOp);
-        mapSymTypes.push_back(baseOp.getType());
-        mapSymLocs.push_back(baseOp.getLoc());
-        mapSymbols.push_back(&sym);
+        clauseOps.mapVars.push_back(mapOp);
+        mapSyms.push_back(&sym);
+        mapLocs.push_back(baseOp.getLoc());
+        mapTypes.push_back(baseOp.getType());
       }
     }
   };
   Fortran::lower::pft::visitAllSymbols(eval, captureImplicitMap);
 
   auto targetOp = converter.getFirOpBuilder().create<mlir::omp::TargetOp>(
-      currentLocation, ifClauseOperand, deviceOperand, threadLimitOperand,
-      dependTypeOperands.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 dependTypeOperands),
-      dependOperands, nowaitAttr, devicePtrOperands, deviceAddrOperands,
-      mapOperands);
+      currentLocation, clauseOps);
 
-  genBodyOfTargetOp(converter, semaCtx, eval, genNested, targetOp, mapSymTypes,
-                    mapSymLocs, mapSymbols, currentLocation);
+  genBodyOfTargetOp(converter, semaCtx, eval, genNested, targetOp, mapSyms,
+                    mapLocs, mapTypes, currentLocation);
 
   return targetOp;
 }
@@ -1426,17 +1360,16 @@ genTeamsOp(Fortran::lower::AbstractConverter &converter,
            const Fortran::parser::OmpClauseList &clauseList,
            bool outerCombined = false) {
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value numTeamsClauseOperand, ifClauseOperand, threadLimitClauseOperand;
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands,
-      reductionVars;
-  llvm::SmallVector<mlir::Attribute> reductionDeclSymbols;
+  mlir::omp::TeamsClauseOps clauseOps;
 
   ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_teams, ifClauseOperand);
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processIf(llvm::omp::Directive::OMPD_teams, clauseOps);
+  cp.processAllocate(clauseOps);
   cp.processDefault();
-  cp.processNumTeams(stmtCtx, numTeamsClauseOperand);
-  cp.processThreadLimit(stmtCtx, threadLimitClauseOperand);
+  cp.processNumTeams(stmtCtx, clauseOps);
+  cp.processThreadLimit(stmtCtx, clauseOps);
+  // TODO Support delayed privatization.
+
   cp.processTODO<clause::Reduction>(currentLocation,
                                     llvm::omp::Directive::OMPD_teams);
 
@@ -1445,30 +1378,20 @@ genTeamsOp(Fortran::lower::AbstractConverter &converter,
           .setGenNested(genNested)
           .setOuterCombined(outerCombined)
           .setClauses(&clauseList),
-      /*num_teams_lower=*/nullptr, numTeamsClauseOperand, ifClauseOperand,
-      threadLimitClauseOperand, allocateOperands, allocatorOperands,
-      reductionVars,
-      reductionDeclSymbols.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(converter.getFirOpBuilder().getContext(),
-                                 reductionDeclSymbols));
+      clauseOps);
 }
 
 /// Extract the list of function and variable symbols affected by the given
 /// 'declare target' directive and return the intended device type for them.
-static mlir::omp::DeclareTargetDeviceType getDeclareTargetInfo(
+static void getDeclareTargetInfo(
     Fortran::lower::AbstractConverter &converter,
     Fortran::semantics::SemanticsContext &semaCtx,
     Fortran::lower::pft::Evaluation &eval,
     const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
+    mlir::omp::DeclareTargetClauseOps &clauseOps,
     llvm::SmallVectorImpl<DeclareTargetCapturePair> &symbolAndClause) {
-
-  // The default capture type
-  mlir::omp::DeclareTargetDeviceType deviceType =
-      mlir::omp::DeclareTargetDeviceType::any;
   const auto &spec = std::get<Fortran::parser::OmpDeclareTargetSpecifier>(
       declareTargetConstruct.t);
-
   if (const auto *objectList{
           Fortran::parser::Unwrap<Fortran::parser::OmpObjectList>(spec.u)}) {
     ObjectList objects{makeObjects(*objectList, semaCtx)};
@@ -1489,12 +1412,10 @@ static mlir::omp::DeclareTargetDeviceType getDeclareTargetInfo(
     cp.processTo(symbolAndClause);
     cp.processEnter(symbolAndClause);
     cp.processLink(symbolAndClause);
-    cp.processDeviceType(deviceType);
+    cp.processDeviceType(clauseOps);
     cp.processTODO<clause::Indirect>(converter.getCurrentLocation(),
                                      llvm::omp::Directive::OMPD_declare_target);
   }
-
-  return deviceType;
 }
 
 static void collectDeferredDeclareTargets(
@@ -1504,9 +1425,10 @@ static void collectDeferredDeclareTargets(
     const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
     llvm::SmallVectorImpl<Fortran::lower::OMPDeferredDeclareTargetInfo>
         &deferredDeclareTarget) {
+  mlir::omp::DeclareTargetClauseOps clauseOps;
   llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
-  mlir::omp::DeclareTargetDeviceType devType = getDeclareTargetInfo(
-      converter, semaCtx, eval, declareTargetConstruct, symbolAndClause);
+  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
+                       clauseOps, symbolAndClause);
   // Return the device type only if at least one of the targets for the
   // directive is a function or subroutine
   mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
@@ -1516,8 +1438,9 @@ static void collectDeferredDeclareTargets(
         std::get<const Fortran::semantics::Symbol &>(symClause)));
 
     if (!op) {
-      deferredDeclareTarget.push_back(
-          {std::get<0>(symClause), devType, std::get<1>(symClause)});
+      deferredDeclareTarget.push_back({std::get<0>(symClause),
+                                       clauseOps.deviceType,
+                                       std::get<1>(symClause)});
     }
   }
 }
@@ -1529,9 +1452,10 @@ getDeclareTargetFunctionDevice(
     Fortran::lower::pft::Evaluation &eval,
     const Fortran::parser::OpenMPDeclareTargetConstruct
         &declareTargetConstruct) {
+  mlir::omp::DeclareTargetClauseOps clauseOps;
   llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
-  mlir::omp::DeclareTargetDeviceType deviceType = getDeclareTargetInfo(
-      converter, semaCtx, eval, declareTargetConstruct, symbolAndClause);
+  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
+                       clauseOps, symbolAndClause);
 
   // Return the device type only if at least one of the targets for the
   // directive is a function or subroutine
@@ -1541,7 +1465,7 @@ getDeclareTargetFunctionDevice(
         std::get<const Fortran::semantics::Symbol &>(symClause)));
 
     if (mlir::isa_and_nonnull<mlir::func::FuncOp>(op))
-      return deviceType;
+      return clauseOps.deviceType;
   }
 
   return std::nullopt;
@@ -1571,12 +1495,14 @@ genOmpSimpleStandalone(Fortran::lower::AbstractConverter &converter,
   case llvm::omp::Directive::OMPD_barrier:
     firOpBuilder.create<mlir::omp::BarrierOp>(currentLocation);
     break;
-  case llvm::omp::Directive::OMPD_taskwait:
-    ClauseProcessor(converter, semaCtx, opClauseList)
-        .processTODO<clause::Depend, clause::Nowait>(
-            currentLocation, llvm::omp::Directive::OMPD_taskwait);
-    firOpBuilder.create<mlir::omp::TaskwaitOp>(currentLocation);
+  case llvm::omp::Directive::OMPD_taskwait: {
+    mlir::omp::TaskwaitClauseOps clauseOps;
+    ClauseProcessor cp(converter, semaCtx, opClauseList);
+    cp.processTODO<clause::Depend, clause::Nowait>(
+        currentLocation, llvm::omp::Directive::OMPD_taskwait);
+    firOpBuilder.create<mlir::omp::TaskwaitOp>(currentLocation, clauseOps);
     break;
+  }
   case llvm::omp::Directive::OMPD_taskyield:
     firOpBuilder.create<mlir::omp::TaskyieldOp>(currentLocation);
     break;
@@ -1711,32 +1637,21 @@ createSimdLoop(Fortran::lower::AbstractConverter &converter,
   dsp.processStep1();
 
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value scheduleChunkClauseOperand, ifClauseOperand;
-  llvm::SmallVector<mlir::Value> lowerBound, upperBound, step, reductionVars;
-  llvm::SmallVector<mlir::Value> alignedVars, nontemporalVars;
+  mlir::omp::SimdLoopClauseOps clauseOps;
   llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
-  llvm::SmallVector<mlir::Type> reductionTypes;
-  llvm::SmallVector<mlir::Attribute> reductionDeclSymbols;
-  mlir::omp::ClauseOrderKindAttr orderClauseOperand;
-  mlir::IntegerAttr simdlenClauseOperand, safelenClauseOperand;
 
   ClauseProcessor cp(converter, semaCtx, loopOpClauseList);
-  cp.processCollapse(loc, eval, lowerBound, upperBound, step, iv);
-  cp.processScheduleChunk(stmtCtx, scheduleChunkClauseOperand);
-  cp.processReduction(loc, reductionVars, reductionTypes, reductionDeclSymbols);
-  cp.processIf(llvm::omp::Directive::OMPD_simd, ifClauseOperand);
-  cp.processSimdlen(simdlenClauseOperand);
-  cp.processSafelen(safelenClauseOperand);
+  cp.processCollapse(loc, eval, clauseOps, iv);
+  cp.processReduction(loc, clauseOps);
+  cp.processIf(llvm::omp::Directive::OMPD_simd, clauseOps);
+  cp.processSimdlen(clauseOps);
+  cp.processSafelen(clauseOps);
+  clauseOps.loopInclusiveAttr = firOpBuilder.getUnitAttr();
+  // TODO Support delayed privatization.
+
   cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
                  clause::Nontemporal, clause::Order>(loc, ompDirective);
 
-  mlir::TypeRange resultType;
-  auto simdLoopOp = firOpBuilder.create<mlir::omp::SimdLoopOp>(
-      loc, resultType, lowerBound, upperBound, step, alignedVars,
-      /*alignment_values=*/nullptr, ifClauseOperand, nontemporalVars,
-      orderClauseOperand, simdlenClauseOperand, safelenClauseOperand,
-      /*inclusive=*/firOpBuilder.getUnitAttr());
-
   auto *nestedEval = getCollapsedLoopEval(
       eval, Fortran::lower::getCollapseValue(loopOpClauseList));
 
@@ -1744,11 +1659,12 @@ createSimdLoop(Fortran::lower::AbstractConverter &converter,
     return genLoopVars(op, converter, loc, iv);
   };
 
-  createBodyOfOp<mlir::omp::SimdLoopOp>(
-      simdLoopOp, OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
-                      .setClauses(&loopOpClauseList)
-                      .setDataSharingProcessor(&dsp)
-                      .setGenRegionEntryCb(ivCallback));
+  genOpWithBody<mlir::omp::SimdLoopOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
+          .setClauses(&loopOpClauseList)
+          .setDataSharingProcessor(&dsp)
+          .setGenRegionEntryCb(ivCallback),
+      clauseOps);
 }
 
 static void createWsloop(Fortran::lower::AbstractConverter &converter,
@@ -1763,77 +1679,50 @@ static void createWsloop(Fortran::lower::AbstractConverter &converter,
   dsp.processStep1();
 
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value scheduleChunkClauseOperand;
-  llvm::SmallVector<mlir::Value> lowerBound, upperBound, step, reductionVars;
-  llvm::SmallVector<mlir::Value> linearVars, linearStepVars;
+  mlir::omp::WsloopClauseOps clauseOps;
   llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
   llvm::SmallVector<mlir::Type> reductionTypes;
-  llvm::SmallVector<mlir::Attribute> reductionDeclSymbols;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSymbols;
-  mlir::omp::ClauseOrderKindAttr orderClauseOperand;
-  mlir::omp::ClauseScheduleKindAttr scheduleValClauseOperand;
-  mlir::UnitAttr nowaitClauseOperand, byrefOperand, scheduleSimdClauseOperand;
-  mlir::IntegerAttr orderedClauseOperand;
-  mlir::omp::ScheduleModifierAttr scheduleModClauseOperand;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
 
   ClauseProcessor cp(converter, semaCtx, beginClauseList);
-  cp.processCollapse(loc, eval, lowerBound, upperBound, step, iv);
-  cp.processScheduleChunk(stmtCtx, scheduleChunkClauseOperand);
-  cp.processReduction(loc, reductionVars, reductionTypes, reductionDeclSymbols,
-                      &reductionSymbols);
-  cp.processTODO<clause::Linear, clause::Order>(loc, ompDirective);
-
-  if (ReductionProcessor::doReductionByRef(reductionVars))
-    byrefOperand = firOpBuilder.getUnitAttr();
-
-  auto wsLoopOp = firOpBuilder.create<mlir::omp::WsloopOp>(
-      loc, lowerBound, upperBound, step, linearVars, linearStepVars,
-      reductionVars,
-      reductionDeclSymbols.empty()
-          ? nullptr
-          : mlir::ArrayAttr::get(firOpBuilder.getContext(),
-                                 reductionDeclSymbols),
-      scheduleValClauseOperand, scheduleChunkClauseOperand,
-      /*schedule_modifiers=*/nullptr,
-      /*simd_modifier=*/nullptr, nowaitClauseOperand, byrefOperand,
-      orderedClauseOperand, orderClauseOperand,
-      /*inclusive=*/firOpBuilder.getUnitAttr());
-
-  // Handle attribute based clauses.
-  if (cp.processOrdered(orderedClauseOperand))
-    wsLoopOp.setOrderedValAttr(orderedClauseOperand);
-
-  if (cp.processSchedule(scheduleValClauseOperand, scheduleModClauseOperand,
-                         scheduleSimdClauseOperand)) {
-    wsLoopOp.setScheduleValAttr(scheduleValClauseOperand);
-    wsLoopOp.setScheduleModifierAttr(scheduleModClauseOperand);
-    wsLoopOp.setSimdModifierAttr(scheduleSimdClauseOperand);
-  }
+  cp.processCollapse(loc, eval, clauseOps, iv);
+  cp.processSchedule(stmtCtx, clauseOps);
+  cp.processReduction(loc, clauseOps, &reductionTypes, &reductionSyms);
+  cp.processOrdered(clauseOps);
+  clauseOps.loopInclusiveAttr = firOpBuilder.getUnitAttr();
+  // TODO Support delayed privatization.
+
+  if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
+    clauseOps.reductionByRefAttr = firOpBuilder.getUnitAttr();
+
+  cp.processTODO<clause::Allocate, clause::Linear, clause::Order>(loc,
+                                                                  ompDirective);
+
   // In FORTRAN `nowait` clause occur at the end of `omp do` directive.
   // i.e
   // !$omp do
   // <...>
   // !$omp end do nowait
   if (endClauseList) {
-    if (ClauseProcessor(converter, semaCtx, *endClauseList)
-            .processNowait(nowaitClauseOperand))
-      wsLoopOp.setNowaitAttr(nowaitClauseOperand);
+    ClauseProcessor ecp(converter, semaCtx, *endClauseList);
+    ecp.processNowait(clauseOps);
   }
 
   auto *nestedEval = getCollapsedLoopEval(
       eval, Fortran::lower::getCollapseValue(beginClauseList));
 
   auto ivCallback = [&](mlir::Operation *op) {
-    return genLoopAndReductionVars(op, converter, loc, iv, reductionSymbols,
+    return genLoopAndReductionVars(op, converter, loc, iv, reductionSyms,
                                    reductionTypes);
   };
 
-  createBodyOfOp<mlir::omp::WsloopOp>(
-      wsLoopOp, OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
-                    .setClauses(&beginClauseList)
-                    .setDataSharingProcessor(&dsp)
-                    .setReductions(&reductionSymbols, &reductionTypes)
-                    .setGenRegionEntryCb(ivCallback));
+  genOpWithBody<mlir::omp::WsloopOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
+          .setClauses(&beginClauseList)
+          .setDataSharingProcessor(&dsp)
+          .setReductions(&reductionSyms, &reductionTypes)
+          .setGenRegionEntryCb(ivCallback),
+      clauseOps);
 }
 
 static void createSimdWsloop(
@@ -1921,10 +1810,11 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
                    Fortran::lower::pft::Evaluation &eval,
                    const Fortran::parser::OpenMPDeclareTargetConstruct
                        &declareTargetConstruct) {
+  mlir::omp::DeclareTargetClauseOps clauseOps;
   llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
   mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
-  mlir::omp::DeclareTargetDeviceType deviceType = getDeclareTargetInfo(
-      converter, semaCtx, eval, declareTargetConstruct, symbolAndClause);
+  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
+                       clauseOps, symbolAndClause);
 
   for (const DeclareTargetCapturePair &symClause : symbolAndClause) {
     mlir::Operation *op = mod.lookupSymbol(converter.mangleName(
@@ -1938,7 +1828,8 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
 
     markDeclareTarget(
         op, converter,
-        std::get<mlir::omp::DeclareTargetCaptureClause>(symClause), deviceType);
+        std::get<mlir::omp::DeclareTargetCaptureClause>(symClause),
+        clauseOps.deviceType);
   }
 }
 
@@ -2072,7 +1963,8 @@ genOMP(Fortran::lower::AbstractConverter &converter,
         !std::get_if<Fortran::parser::OmpClause::IsDevicePtr>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::HasDeviceAddr>(&clause.u) &&
         !std::get_if<Fortran::parser::OmpClause::ThreadLimit>(&clause.u) &&
-        !std::get_if<Fortran::parser::OmpClause::NumTeams>(&clause.u)) {
+        !std::get_if<Fortran::parser::OmpClause::NumTeams>(&clause.u) &&
+        !std::get_if<Fortran::parser::OmpClause::Simd>(&clause.u)) {
       TODO(clauseLocation, "OpenMP Block construct clause");
     }
   }
@@ -2092,7 +1984,7 @@ genOMP(Fortran::lower::AbstractConverter &converter,
     break;
   case llvm::omp::Directive::OMPD_ordered:
     genOrderedRegionOp(converter, semaCtx, eval, /*genNested=*/true,
-                       currentLocation);
+                       currentLocation, beginClauseList);
     break;
   case llvm::omp::Directive::OMPD_parallel:
     genParallelOp(converter, symTable, semaCtx, eval, /*genNested=*/true,
@@ -2183,7 +2075,6 @@ genOMP(Fortran::lower::AbstractConverter &converter,
        const Fortran::parser::OpenMPCriticalConstruct &criticalConstruct) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   mlir::Location currentLocation = converter.getCurrentLocation();
-  mlir::IntegerAttr hintClauseOp;
   std::string name;
   const Fortran::parser::OmpCriticalDirective &cd =
       std::get<Fortran::parser::OmpCriticalDirective>(criticalConstruct.t);
@@ -2192,21 +2083,28 @@ genOMP(Fortran::lower::AbstractConverter &converter,
         std::get<std::optional<Fortran::parser::Name>>(cd.t).value().ToString();
   }
 
-  const auto &clauseList = std::get<Fortran::parser::OmpClauseList>(cd.t);
-  ClauseProcessor(converter, semaCtx, clauseList).processHint(hintClauseOp);
-
   mlir::omp::CriticalOp criticalOp = [&]() {
     if (name.empty()) {
       return firOpBuilder.create<mlir::omp::CriticalOp>(
           currentLocation, mlir::FlatSymbolRefAttr());
     }
+
     mlir::ModuleOp module = firOpBuilder.getModule();
     mlir::OpBuilder modBuilder(module.getBodyRegion());
     auto global = module.lookupSymbol<mlir::omp::CriticalDeclareOp>(name);
-    if (!global)
-      global = modBuilder.create<mlir::omp::CriticalDeclareOp>(
-          currentLocation,
-          mlir::StringAttr::get(firOpBuilder.getContext(), name), hintClauseOp);
+    if (!global) {
+      mlir::omp::CriticalClauseOps clauseOps;
+      const auto &clauseList = std::get<Fortran::parser::OmpClauseList>(cd.t);
+
+      ClauseProcessor cp(converter, semaCtx, clauseList);
+      cp.processHint(clauseOps);
+      clauseOps.nameAttr =
+          mlir::StringAttr::get(firOpBuilder.getContext(), name);
+
+      global = modBuilder.create<mlir::omp::CriticalDeclareOp>(currentLocation,
+                                                               clauseOps);
+    }
+
     return firOpBuilder.create<mlir::omp::CriticalOp>(
         currentLocation, mlir::FlatSymbolRefAttr::get(firOpBuilder.getContext(),
                                                       global.getSymName()));
@@ -2323,8 +2221,7 @@ genOMP(Fortran::lower::AbstractConverter &converter,
        Fortran::lower::pft::Evaluation &eval,
        const Fortran::parser::OpenMPSectionsConstruct &sectionsConstruct) {
   mlir::Location currentLocation = converter.getCurrentLocation();
-  llvm::SmallVector<mlir::Value> allocateOperands, allocatorOperands;
-  mlir::UnitAttr nowaitClauseOperand;
+  mlir::omp::SectionsClauseOps clauseOps;
   const auto &beginSectionsDirective =
       std::get<Fortran::parser::OmpBeginSectionsDirective>(sectionsConstruct.t);
   const auto &sectionsClauseList =
@@ -2333,8 +2230,9 @@ genOMP(Fortran::lower::AbstractConverter &converter,
   // Process clauses before optional omp.parallel, so that new variables are
   // allocated outside of the parallel region
   ClauseProcessor cp(converter, semaCtx, sectionsClauseList);
-  cp.processSectionsReduction(currentLocation);
-  cp.processAllocate(allocatorOperands, allocateOperands);
+  cp.processSectionsReduction(currentLocation, clauseOps);
+  cp.processAllocate(clauseOps);
+  // TODO Support delayed privatization.
 
   llvm::omp::Directive dir =
       std::get<Fortran::parser::OmpSectionsDirective>(beginSectionsDirective.t)
@@ -2351,16 +2249,14 @@ genOMP(Fortran::lower::AbstractConverter &converter,
     const auto &endSectionsClauseList =
         std::get<Fortran::parser::OmpClauseList>(endSectionsDirective.t);
     ClauseProcessor(converter, semaCtx, endSectionsClauseList)
-        .processNowait(nowaitClauseOperand);
+        .processNowait(clauseOps);
   }
 
   // SECTIONS construct
   genOpWithBody<mlir::omp::SectionsOp>(
       OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
           .setGenNested(false),
-      /*reduction_vars=*/mlir::ValueRange(),
-      /*reductions=*/nullptr, allocateOperands, allocatorOperands,
-      nowaitClauseOperand);
+      clauseOps);
 
   const auto &sectionBlocks =
       std::get<Fortran::parser::OmpSectionBlocks>(sectionsConstruct.t);
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
index 4ce7e47da046b..304a9740d91ed 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
@@ -81,7 +81,7 @@ struct GrainsizeClauseOps {
   Value grainsizeVar;
 };
 
-struct HasDeviceAddrOps {
+struct HasDeviceAddrClauseOps {
   llvm::SmallVector<Value> hasDeviceAddrVars;
 };
 struct HintClauseOps {
@@ -97,7 +97,7 @@ struct InReductionClauseOps {
   llvm::SmallVector<Attribute> inReductionDeclSymbols;
 };
 
-struct IsDevicePtrOps {
+struct IsDevicePtrClauseOps {
   llvm::SmallVector<Value> isDevicePtrVars;
 };
 
@@ -234,6 +234,8 @@ using DistributeClauseOps =
     detail::Clauses<AllocateClauseOps, DistScheduleClauseOps, OrderClauseOps,
                     PrivateClauseOps>;
 
+using LoopNestClauseOps = detail::Clauses<CollapseClauseOps, LoopRelatedOps>;
+
 // TODO `filter` clause.
 using MaskedClauseOps = detail::Clauses<>;
 
@@ -261,8 +263,8 @@ using SingleClauseOps = detail::Clauses<AllocateClauseOps, CopyprivateClauseOps,
 // TODO `defaultmap`, `uses_allocators` clauses.
 using TargetClauseOps =
     detail::Clauses<AllocateClauseOps, DependClauseOps, DeviceClauseOps,
-                    HasDeviceAddrOps, IfClauseOps, InReductionClauseOps,
-                    IsDevicePtrOps, MapClauseOps, NowaitClauseOps,
+                    HasDeviceAddrClauseOps, IfClauseOps, InReductionClauseOps,
+                    IsDevicePtrClauseOps, MapClauseOps, NowaitClauseOps,
                     PrivateClauseOps, ReductionClauseOps, ThreadLimitClauseOps>;
 
 using TargetDataClauseOps = detail::Clauses<DeviceClauseOps, IfClauseOps,
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index a2eb39a8e6077..596b2f5e4444f 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -574,6 +574,10 @@ def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
                        Variadic<IntLikeType>:$step,
                        UnitAttr:$inclusive);
 
+  let builders = [
+    OpBuilder<(ins CArg<"const LoopNestClauseOps &">:$clauses)>
+  ];
+
   let regions = (region AnyRegion:$region);
 
   let extraClassDeclaration = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 4ab78bf331f80..35fb174046a3a 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1920,6 +1920,12 @@ void LoopNestOp::print(OpAsmPrinter &p) {
   p.printRegion(region, /*printEntryBlockArgs=*/false);
 }
 
+void LoopNestOp::build(OpBuilder &builder, OperationState &state,
+                       const LoopNestClauseOps &clauses) {
+  LoopNestOp::build(builder, state, clauses.loopLBVar, clauses.loopUBVar,
+                    clauses.loopStepVar, clauses.loopInclusiveAttr);
+}
+
 LogicalResult LoopNestOp::verify() {
   if (getLowerBound().size() != getIVs().size())
     return emitOpError() << "number of range arguments and IVs do not match";

From 33779b861d748ed18b8d4f9cbad3a84deaa95e9f Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Fri, 12 Apr 2024 07:41:43 -0400
Subject: [PATCH 238/886] Reword a diagnostic; NFC

This diagnostic used to talk about complex integer types but is issued
for use of the increment or decrement operators on any complex type,
not just integral ones.

The diagnostic also had zero test coverage, so new coverage is added
along with the rewording.
---
 .../clang/Basic/DiagnosticSemaKinds.td        |  4 ++--
 clang/lib/Sema/SemaExpr.cpp                   |  4 ++--
 clang/test/Sema/complex-inc-dec.c             | 24 +++++++++++++++++++
 3 files changed, 28 insertions(+), 4 deletions(-)
 create mode 100644 clang/test/Sema/complex-inc-dec.c

diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 180e913155d67..774d2b53a3825 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -7588,8 +7588,8 @@ def ext_gnu_ptr_func_arith : Extension<
   InGroup<GNUPointerArith>;
 def err_readonly_message_assignment : Error<
   "assigning to 'readonly' return result of an Objective-C message not allowed">;
-def ext_integer_increment_complex : Extension<
-  "ISO C does not support '++'/'--' on complex integer type %0">;
+def ext_increment_complex : Extension<
+  "'%select{--|++}0' on an object of complex type is a Clang extension">;
 def ext_integer_complement_complex : Extension<
   "ISO C does not support '~' for complex conjugation of %0">;
 def err_nosetter_property_assignment : Error<
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 18284e0c3e987..b294d2bd9f53f 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -14859,8 +14859,8 @@ static QualType CheckIncrementDecrementOperand(Sema &S, Expr *Op,
       return QualType();
   } else if (ResType->isAnyComplexType()) {
     // C99 does not support ++/-- on complex types, we allow as an extension.
-    S.Diag(OpLoc, diag::ext_integer_increment_complex)
-      << ResType << Op->getSourceRange();
+    S.Diag(OpLoc, diag::ext_increment_complex)
+      << IsInc << Op->getSourceRange();
   } else if (ResType->isPlaceholderType()) {
     ExprResult PR = S.CheckPlaceholderExpr(Op);
     if (PR.isInvalid()) return QualType();
diff --git a/clang/test/Sema/complex-inc-dec.c b/clang/test/Sema/complex-inc-dec.c
new file mode 100644
index 0000000000000..ebc23c36f40c8
--- /dev/null
+++ b/clang/test/Sema/complex-inc-dec.c
@@ -0,0 +1,24 @@
+// RUN: %clang_cc1 -verify -pedantic -std=c99 %s
+
+void func(void) {
+  _Complex float cf;
+  _Complex double cd;
+  _Complex long double cld;
+
+  ++cf;  // expected-warning {{'++' on an object of complex type is a Clang extension}}
+  ++cd;  // expected-warning {{'++' on an object of complex type is a Clang extension}}
+  ++cld; // expected-warning {{'++' on an object of complex type is a Clang extension}}
+
+  --cf;  // expected-warning {{'--' on an object of complex type is a Clang extension}}
+  --cd;  // expected-warning {{'--' on an object of complex type is a Clang extension}}
+  --cld; // expected-warning {{'--' on an object of complex type is a Clang extension}}
+
+  cf++;  // expected-warning {{'++' on an object of complex type is a Clang extension}}
+  cd++;  // expected-warning {{'++' on an object of complex type is a Clang extension}}
+  cld++; // expected-warning {{'++' on an object of complex type is a Clang extension}}
+
+  cf--;  // expected-warning {{'--' on an object of complex type is a Clang extension}}
+  cd--;  // expected-warning {{'--' on an object of complex type is a Clang extension}}
+  cld--; // expected-warning {{'--' on an object of complex type is a Clang extension}}
+}
+

From 5ebeaf211551f9410e17b1134126daadd8b163ac Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Fri, 12 Apr 2024 16:02:57 +0400
Subject: [PATCH 239/886] [clang] Add tests for some CWG 5xx issues (#87909)

This patch covers
[CWG393](https://cplusplus.github.io/CWG/issues/393.html) "Pointer to
array of unknown bound in template argument list in parameter",
[CWG528](https://cplusplus.github.io/CWG/issues/528.html) "Why are
incomplete class types not allowed with `typeid`?",
[CWG550](https://cplusplus.github.io/CWG/issues/550.html) "Pointer to
array of unknown bound in parameter declarations",
[CWG553](https://cplusplus.github.io/CWG/issues/553.html) "Problems with
friend allocation and deallocation functions",
[CWG555](https://cplusplus.github.io/CWG/issues/555.html)
"Pseudo-destructor name lookup",
[CWG560](https://cplusplus.github.io/CWG/issues/560.html) "Use of the
`typename` keyword in return types".

CWG393 is on this list, because CWG550 is marked as a duplicate of
CWG393.

Test for CWG553 has been already written, but it was missing a status
comment. As a drive-by fix, I'm adding missing status comments to
CWG1584 and CWG1903 as well.

CWG555 used CWG466 test, and also a variation of that test to test
references. CWG466 is now testing non-reference non-pointer case.

CWG560 showcases again that converting warnings to errors in DR tests
via `-pedantic-errors` doesn't make things more clear. By default that
test is accepted with an extension warning since we implemented
[P0634R3](https://wg21.link/p0634r3) "Down with `typename`!" in Clang
16.
---
 clang/test/CXX/drs/dr15xx.cpp |  2 +-
 clang/test/CXX/drs/dr19xx.cpp |  2 +-
 clang/test/CXX/drs/dr3xx.cpp  | 19 ++++++++
 clang/test/CXX/drs/dr4xx.cpp  | 60 +++++++++++------------
 clang/test/CXX/drs/dr5xx.cpp  | 89 +++++++++++++++++++++++++++++++++++
 clang/www/cxx_dr_status.html  | 16 +++----
 6 files changed, 149 insertions(+), 39 deletions(-)

diff --git a/clang/test/CXX/drs/dr15xx.cpp b/clang/test/CXX/drs/dr15xx.cpp
index 195c0fa610d57..6e3ad41c748fb 100644
--- a/clang/test/CXX/drs/dr15xx.cpp
+++ b/clang/test/CXX/drs/dr15xx.cpp
@@ -555,7 +555,7 @@ auto DR1579_lambda_invalid = []() -> GenericMoveOnly<char> {
 #endif
 } // end namespace dr1579
 
-namespace dr1584 {
+namespace dr1584 { // dr1584: 7 drafting 2015-05
 #if __cplusplus >= 201103L
   // Deducing function types from cv-qualified types
   template<typename T> void f(const T *); // #dr1584-f
diff --git a/clang/test/CXX/drs/dr19xx.cpp b/clang/test/CXX/drs/dr19xx.cpp
index 716b1476831ed..f8c1581f08540 100644
--- a/clang/test/CXX/drs/dr19xx.cpp
+++ b/clang/test/CXX/drs/dr19xx.cpp
@@ -34,7 +34,7 @@ namespace dr1902 { // dr1902: 3.7
 #endif
 }
 
-namespace dr1903 {
+namespace dr1903 { // dr1903: 2.7
   namespace A {
     struct a {};
     int a;
diff --git a/clang/test/CXX/drs/dr3xx.cpp b/clang/test/CXX/drs/dr3xx.cpp
index 483ebf7a08aad..d5cfebccf8134 100644
--- a/clang/test/CXX/drs/dr3xx.cpp
+++ b/clang/test/CXX/drs/dr3xx.cpp
@@ -1568,6 +1568,25 @@ namespace dr391 { // dr391: 2.8 c++11
 }
 
 // dr392 is in dr392.cpp
+
+namespace dr393 { // dr393: 2.7
+template <typename T>
+struct S {};
+
+void f1(S<int (*)[]>);
+void f2(S<int (&)[]>);
+void g(int(*S<int>::*)[]);
+
+template<typename T>
+void sp_assert_convertible( T* ) {}
+
+template<typename T, typename U>
+void h() {
+  T (*p) [] = (U(*)[])0;
+  sp_assert_convertible<T[]>( (U(*)[])0 );
+}
+} // namespace dr393
+
 // dr394: na
 
 namespace dr395 { // dr395: 3.0
diff --git a/clang/test/CXX/drs/dr4xx.cpp b/clang/test/CXX/drs/dr4xx.cpp
index 34dd638c1d9b0..9b7a9777202f2 100644
--- a/clang/test/CXX/drs/dr4xx.cpp
+++ b/clang/test/CXX/drs/dr4xx.cpp
@@ -2,8 +2,9 @@
 // RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++11 %s -verify=expected,cxx98-14,cxx98-17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++14 %s -verify=expected,cxx98-14,cxx98-17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++17 %s -verify=expected,since-cxx17,cxx98-17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
-// RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++20 %s -verify=expected,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
-// RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++23 %s -verify=expected,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
+// RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++20 %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
+// RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++23 %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
+// RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++2c %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 
 // FIXME: __SIZE_TYPE__ expands to 'long long' on some targets.
 __extension__ typedef __SIZE_TYPE__ size_t;
@@ -949,33 +950,34 @@ namespace dr460 { // dr460: yes
 // dr465: na
 
 namespace dr466 { // dr466: 2.8
-  typedef int I;
-  typedef const int CI;
-  typedef volatile int VI;
-  void f(int *a, CI *b, VI *c) {
-    a->~I();
-    a->~CI();
-    a->~VI();
-    a->I::~I();
-    a->CI::~CI();
-    a->VI::~VI();
-
-    a->CI::~VI(); // allowed by changes to [expr.id.prim.qual]/2 introduced in P1131R2
-
-    b->~I();
-    b->~CI();
-    b->~VI();
-    b->I::~I();
-    b->CI::~CI();
-    b->VI::~VI();
-
-    c->~I();
-    c->~CI();
-    c->~VI();
-    c->I::~I();
-    c->CI::~CI();
-    c->VI::~VI();
-  }
+typedef int I;
+typedef const int CI;
+typedef volatile int VI;
+void g(int a, CI b, VI c) {
+// since-cxx20-warning@-1 {{volatile-qualified parameter type 'VI' (aka 'volatile int') is deprecated}}
+  a.~I();
+  a.~CI();
+  a.~VI();
+  a.I::~I();
+  a.CI::~CI();
+  a.VI::~VI();
+
+  a.CI::~VI(); // allowed by changes to [expr.id.prim.qual]/2 introduced in P1131R2
+
+  b.~I();
+  b.~CI();
+  b.~VI();
+  b.I::~I();
+  b.CI::~CI();
+  b.VI::~VI();
+
+  c.~I();
+  c.~CI();
+  c.~VI();
+  c.I::~I();
+  c.CI::~CI();
+  c.VI::~VI();
+}
 }
 
 namespace dr467 { // dr467: yes
diff --git a/clang/test/CXX/drs/dr5xx.cpp b/clang/test/CXX/drs/dr5xx.cpp
index 0ea306a041167..d473f4827455c 100644
--- a/clang/test/CXX/drs/dr5xx.cpp
+++ b/clang/test/CXX/drs/dr5xx.cpp
@@ -18,6 +18,10 @@ namespace std {
 void *operator new(size_t, std::align_val_t); // #dr5xx-global-operator-new-aligned
 #endif
 
+namespace std {
+  struct type_info;
+}
+
 namespace dr500 { // dr500: dup 372
   class D;
   class A {
@@ -265,6 +269,18 @@ namespace dr527 { // dr527: na
   int ax = a.x, bx = b.x, cx = c.x, dx = d.x, ex = E::e->x, fx = F::f->x;
 }
 
+namespace dr528 { // dr528: 2.7
+
+struct S; // #dr528-S
+
+void f() {
+  typeid(S);
+  // expected-error@-1 {{'typeid' of incomplete type 'S'}}
+  //   expected-note@#dr528-S {{forward declaration of 'dr528::S'}}
+}
+
+} // namespace dr528
+
 namespace dr530 { // dr530: yes
   template<int*> struct S { enum { N = 1 }; };
   template<void(*)()> struct T { enum { N = 1 }; };
@@ -618,6 +634,8 @@ namespace dr548 { // dr548: dup 482
   template void dr548::f<int>();
 }
 
+// dr550: dup 393
+
 namespace dr551 { // dr551: yes c++11
   // FIXME: This obviously should apply in C++98 mode too.
   template<typename T> void f() {}
@@ -641,6 +659,7 @@ namespace dr552 { // dr552: yes
   X<Y, 0> x;
 }
 
+// dr553: 2.7
 struct dr553_class {
   friend void *operator new(size_t, dr553_class);
 };
@@ -661,6 +680,62 @@ namespace dr553 {
 }
 
 // dr554: na
+
+namespace dr555 { // dr555: 2.8
+typedef int I;
+typedef const int CI;
+typedef volatile int VI;
+void f(int *a, CI *b, VI *c) {
+  a->~I();
+  a->~CI();
+  a->~VI();
+  a->I::~I();
+  a->CI::~CI();
+  a->VI::~VI();
+
+  a->CI::~VI(); // allowed by changes to [expr.id.prim.qual]/2 introduced in P1131R2
+
+  b->~I();
+  b->~CI();
+  b->~VI();
+  b->I::~I();
+  b->CI::~CI();
+  b->VI::~VI();
+
+  c->~I();
+  c->~CI();
+  c->~VI();
+  c->I::~I();
+  c->CI::~CI();
+  c->VI::~VI();
+}
+
+void g(int &a, CI &b, VI &c) {
+  a.~I();
+  a.~CI();
+  a.~VI();
+  a.I::~I();
+  a.CI::~CI();
+  a.VI::~VI();
+
+  a.CI::~VI(); // allowed by changes to [expr.id.prim.qual]/2 introduced in P1131R2
+
+  b.~I();
+  b.~CI();
+  b.~VI();
+  b.I::~I();
+  b.CI::~CI();
+  b.VI::~VI();
+
+  c.~I();
+  c.~CI();
+  c.~VI();
+  c.I::~I();
+  c.CI::~CI();
+  c.VI::~VI();
+}
+} // namespace dr555
+
 // dr556: na
 
 namespace dr557 { // dr557: 3.1
@@ -689,6 +764,20 @@ namespace dr558 { // dr558: 2.9
 
 template<typename> struct dr559 { typedef int T; dr559::T u; }; // dr559: yes
 
+namespace dr560 { // dr560: 16
+
+template <class T>
+struct Outer {
+  struct Inner {
+    Inner* self();
+  };
+};
+template <class T>
+Outer<T>::Inner* Outer<T>::Inner::self() { return this; }
+// cxx98-17-error@-1 {{missing 'typename' prior to dependent type name Outer<T>::Inner; implicit 'typename' is a C++20 extension}}
+
+} // namespace dr560
+
 namespace dr561 { // dr561: yes
   template<typename T> void f(int);
   template<typename T> void g(T t) {
diff --git a/clang/www/cxx_dr_status.html b/clang/www/cxx_dr_status.html
index 6daa5ad2d1316..83b71e7c122d1 100755
--- a/clang/www/cxx_dr_status.html
+++ b/clang/www/cxx_dr_status.html
@@ -2398,7 +2398,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/393.html">393</a></td>
     <td>CD4</td>
     <td>Pointer to array of unknown bound in template argument list in parameter</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 2.7</td>
   </tr>
   <tr id="394">
     <td><a href="https://cplusplus.github.io/CWG/issues/394.html">394</a></td>
@@ -3208,7 +3208,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/528.html">528</a></td>
     <td>NAD</td>
     <td>Why are incomplete class types not allowed with <TT>typeid</TT>?</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 2.7</td>
   </tr>
   <tr class="open" id="529">
     <td><a href="https://cplusplus.github.io/CWG/issues/529.html">529</a></td>
@@ -3342,7 +3342,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/550.html">550</a></td>
     <td>dup</td>
     <td>Pointer to array of unknown bound in parameter declarations</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Duplicate of <a href="#393">393</a></td>
   </tr>
   <tr id="551">
     <td><a href="https://cplusplus.github.io/CWG/issues/551.html">551</a></td>
@@ -3360,7 +3360,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/553.html">553</a></td>
     <td>NAD</td>
     <td>Problems with friend allocation and deallocation functions</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 2.7</td>
   </tr>
   <tr id="554">
     <td><a href="https://cplusplus.github.io/CWG/issues/554.html">554</a></td>
@@ -3372,7 +3372,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/555.html">555</a></td>
     <td>CD5</td>
     <td>Pseudo-destructor name lookup</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 2.8</td>
   </tr>
   <tr id="556">
     <td><a href="https://cplusplus.github.io/CWG/issues/556.html">556</a></td>
@@ -3402,7 +3402,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/560.html">560</a></td>
     <td>NAD</td>
     <td>Use of the <TT>typename</TT> keyword in return types</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 16</td>
   </tr>
   <tr id="561">
     <td><a href="https://cplusplus.github.io/CWG/issues/561.html">561</a></td>
@@ -9312,7 +9312,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/1584.html">1584</a></td>
     <td>drafting</td>
     <td>Deducing function types from cv-qualified types</td>
-    <td align="center">Not resolved</td>
+    <td title="Clang 7 implements 2015-05 resolution" align="center">Not Resolved*</td>
   </tr>
   <tr id="1585">
     <td><a href="https://cplusplus.github.io/CWG/issues/1585.html">1585</a></td>
@@ -11226,7 +11226,7 @@ <h2 id="cxxdr">C++ defect report implementation status</h2>
     <td><a href="https://cplusplus.github.io/CWG/issues/1903.html">1903</a></td>
     <td>CD4</td>
     <td>What declarations are introduced by a non-member <I>using-declaration</I>?</td>
-    <td class="unknown" align="center">Unknown</td>
+    <td class="full" align="center">Clang 2.7</td>
   </tr>
   <tr id="1904">
     <td><a href="https://cplusplus.github.io/CWG/issues/1904.html">1904</a></td>

From ab037c4ff3452a680efb758ccacb6ee210ce333b Mon Sep 17 00:00:00 2001
From: SahilPatidar <patidarsahil2001@gmail.com>
Date: Fri, 12 Apr 2024 17:34:54 +0530
Subject: [PATCH 240/886] [DAG] computeKnownBits - add ISD::ABDU/ISD::ABDS
 handling #84905 (#88253)

Resolve #84905
---
 .../lib/CodeGen/SelectionDAG/SelectionDAG.cpp |  12 ++
 llvm/test/CodeGen/AArch64/abd-combine.ll      |  12 +-
 .../CodeGen/AArch64/arm64-neon-aba-abd.ll     | 157 ++++++++++++++++++
 3 files changed, 174 insertions(+), 7 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index e002cf1d1de38..ddb12a101e02c 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -3397,6 +3397,18 @@ KnownBits SelectionDAG::computeKnownBits(SDValue Op, const APInt &DemandedElts,
     Known = KnownBits::mulhs(Known, Known2);
     break;
   }
+  case ISD::ABDU: {
+    Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
+    Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
+    Known = KnownBits::abdu(Known, Known2);
+    break;
+  }
+  case ISD::ABDS: {
+    Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
+    Known2 = computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
+    Known = KnownBits::abds(Known, Known2);
+    break;
+  }
   case ISD::UMUL_LOHI: {
     assert((Op.getResNo() == 0 || Op.getResNo() == 1) && "Unknown result");
     Known = computeKnownBits(Op.getOperand(1), DemandedElts, Depth + 1);
diff --git a/llvm/test/CodeGen/AArch64/abd-combine.ll b/llvm/test/CodeGen/AArch64/abd-combine.ll
index cc29a59288c7d..e48680f4be98b 100644
--- a/llvm/test/CodeGen/AArch64/abd-combine.ll
+++ b/llvm/test/CodeGen/AArch64/abd-combine.ll
@@ -530,13 +530,11 @@ define <8 x i16> @abds_i_reassoc(<8 x i16> %src1) {
 define <1 x i64> @recursive() {
 ; CHECK-LABEL: recursive:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    movi v0.2d, #0xffffffffffffffff
-; CHECK-NEXT:    movi v1.8b, #1
-; CHECK-NEXT:    uabd v2.8b, v1.8b, v0.8b
-; CHECK-NEXT:    uabdl v0.8h, v1.8b, v0.8b
-; CHECK-NEXT:    dup v1.8b, v2.b[0]
-; CHECK-NEXT:    saddlp v0.1d, v0.2s
-; CHECK-NEXT:    orr v0.8b, v1.8b, v0.8b
+; CHECK-NEXT:    movi v0.8b, #254
+; CHECK-NEXT:    ushll v1.8h, v0.8b, #0
+; CHECK-NEXT:    dup v0.8b, v0.b[0]
+; CHECK-NEXT:    saddlp v1.1d, v1.2s
+; CHECK-NEXT:    orr v0.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    ret
   %1 = tail call <8 x i8> @llvm.aarch64.neon.umax.v8i8(<8 x i8> zeroinitializer, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>)
   %2 = tail call <8 x i8> @llvm.aarch64.neon.uabd.v8i8(<8 x i8> %1, <8 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>)
diff --git a/llvm/test/CodeGen/AArch64/arm64-neon-aba-abd.ll b/llvm/test/CodeGen/AArch64/arm64-neon-aba-abd.ll
index a13eac9b0a5e6..cc9732b83e10c 100644
--- a/llvm/test/CodeGen/AArch64/arm64-neon-aba-abd.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-neon-aba-abd.ll
@@ -291,3 +291,160 @@ define <2 x double> @test_fabd_v2f64(<2 x double> %lhs, <2 x double> %rhs) {
   %abd = call <2 x double> @llvm.aarch64.neon.fabd.v2f64(<2 x double> %lhs, <2 x double> %rhs)
   ret <2 x double> %abd
 }
+
+define <8 x i16> @test_uabd_knownbits_vec8i16(<8 x i16> %lhs, <8 x i16> %rhs) {
+; CHECK-LABEL: test_uabd_knownbits_vec8i16:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v2.8h, #15
+; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT:    uabd v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    rev64 v0.8h, v0.8h
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    ret
+  %and1 = and <8 x i16> %lhs, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %and2 = and <8 x i16> %rhs, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %uabd = call <8 x i16> @llvm.aarch64.neon.uabd.v8i16(<8 x i16> %and1, <8 x i16> %and2)
+  %suff = shufflevector <8 x i16> %uabd, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %res = and <8 x i16> %suff, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  ret <8 x i16> %res
+}
+
+define <4 x i32> @knownbits_uabd_mask_and_shuffle_lshr(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_uabd_mask_and_shuffle_lshr:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    ushr v0.4s, v0.4s, #17
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %2 = and <4 x i32> %a1, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %3 = call <4 x i32> @llvm.aarch64.neon.uabd.v4i32(<4 x i32> %1, <4 x i32> %2)
+  %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %5 = lshr <4 x i32> %4, <i32 17, i32 17, i32 17, i32 17>
+  ret <4 x i32> %5
+}
+
+define <4 x i32> @knownbits_mask_and_shuffle_lshr(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_mask_and_shuffle_lshr:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 32767, i32 32767, i32 32767, i32 32767>
+  %2 = and <4 x i32> %a1, <i32 32767, i32 32767, i32 32767, i32 32767>
+  %3 = call <4 x i32> @llvm.aarch64.neon.uabd.v4i32(<4 x i32> %1, <4 x i32> %2)
+  %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %5 = lshr <4 x i32> %4, <i32 17, i32 17, i32 17, i32 17>
+  ret <4 x i32> %5
+}
+
+define <4 x i32> @test_sabd_knownbits_vec4i32(<4 x i32> %lhs, <4 x i32> %rhs) {
+; CHECK-LABEL: test_sabd_knownbits_vec4i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    adrp x8, .LCPI31_0
+; CHECK-NEXT:    adrp x9, .LCPI31_1
+; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI31_0]
+; CHECK-NEXT:    ldr q3, [x9, :lo12:.LCPI31_1]
+; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v3.16b
+; CHECK-NEXT:    sabd v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    movi v1.2d, #0x0000ff000000ff
+; CHECK-NEXT:    mov v0.s[1], v0.s[0]
+; CHECK-NEXT:    trn2 v0.4s, v0.4s, v0.4s
+; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
+; CHECK-NEXT:    ret
+  %and1 = and <4 x i32> %lhs, <i32 255, i32 -1, i32 -1, i32 255>
+  %and2 = and <4 x i32> %rhs, <i32 255, i32 255, i32 -1, i32 -1>
+  %abd = call <4 x i32> @llvm.aarch64.neon.sabd.v4i32(<4 x i32> %and1, <4 x i32> %and2)
+  %s = shufflevector <4 x i32> %abd, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 3, i32 3>
+  %4 = and <4 x i32> %s, <i32 255, i32 255, i32 255, i32 255>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @knownbits_sabd_and_mask(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_sabd_and_mask:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    adrp x8, .LCPI32_0
+; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI32_0]
+; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT:    sabd v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    zip2 v0.4s, v0.4s, v0.4s
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 -1, i32 -1, i32 255, i32 4085>
+  %2 = and <4 x i32> %a1, <i32 -1, i32 -1, i32 255, i32 4085>
+  %3 = call <4 x i32> @llvm.aarch64.neon.sabd.v4i32(<4 x i32> %1, <4 x i32> %2)
+  %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @knownbits_sabd_and_or_mask(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_sabd_and_or_mask:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 -1, i32 -1, i32 255, i32 4085>
+  %2 = or <4 x i32> %1, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %3 = and <4 x i32> %a1, <i32 -1, i32 -1, i32 255, i32 4085>
+  %4 = or <4 x i32> %3, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %5 = call <4 x i32> @llvm.aarch64.neon.uabd.v4i32(<4 x i32> %2, <4 x i32> %4)
+  %6 = shufflevector <4 x i32> %5, <4 x i32> undef, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
+  ret <4 x i32> %6
+}
+
+define <4 x i32> @knownbits_sabd_and_xor_mask(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_sabd_and_xor_mask:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    adrp x8, .LCPI34_0
+; CHECK-NEXT:    movi v3.2d, #0x00ffff0000ffff
+; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI34_0]
+; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT:    eor v0.16b, v0.16b, v3.16b
+; CHECK-NEXT:    eor v1.16b, v1.16b, v3.16b
+; CHECK-NEXT:    sabd v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    zip2 v0.4s, v0.4s, v0.4s
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 -1, i32 -1, i32 255, i32 4085>
+  %2 = xor <4 x i32> %1, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %3 = and <4 x i32> %a1, <i32 -1, i32 -1, i32 255, i32 4085>
+  %4 = xor <4 x i32> %3, <i32 65535, i32 65535, i32 65535, i32 65535>
+  %5 = call <4 x i32> @llvm.aarch64.neon.sabd.v4i32(<4 x i32> %2, <4 x i32> %4)
+  %6 = shufflevector <4 x i32> %5, <4 x i32> undef, <4 x i32> <i32 2, i32 2, i32 3, i32 3>
+  ret <4 x i32> %6
+}
+
+define <4 x i32> @knownbits_sabd_and_shl_mask(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_sabd_and_shl_mask:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 -65536, i32 -7, i32 -7, i32 -65536>
+  %2 = shl <4 x i32> %1, <i32 17, i32 17, i32 17, i32 17>
+  %3 = and <4 x i32> %a1, <i32 -65536, i32 -7, i32 -7, i32 -65536>
+  %4 = shl <4 x i32> %3, <i32 17, i32 17, i32 17, i32 17>
+  %5 = call <4 x i32> @llvm.aarch64.neon.sabd.v4i32(<4 x i32> %2, <4 x i32> %4)
+  %6 = shufflevector <4 x i32> %5, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 3, i32 3>
+  ret <4 x i32> %6
+}
+
+define <4 x i32> @knownbits_sabd_and_mul_mask(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: knownbits_sabd_and_mul_mask:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    adrp x8, .LCPI36_0
+; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI36_0]
+; CHECK-NEXT:    and v3.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v2.16b, v1.16b, v2.16b
+; CHECK-NEXT:    mul v0.4s, v0.4s, v3.4s
+; CHECK-NEXT:    mul v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    sabd v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    mov v0.s[1], v0.s[0]
+; CHECK-NEXT:    trn2 v0.4s, v0.4s, v0.4s
+; CHECK-NEXT:    ret
+  %1 = and <4 x i32> %a0, <i32 -65536, i32 -7, i32 -7, i32 -65536>
+  %2 = mul <4 x i32> %a0, %1
+  %3 = and <4 x i32> %a1, <i32 -65536, i32 -7, i32 -7, i32 -65536>
+  %4 = mul <4 x i32> %a1, %3
+  %5 = call <4 x i32> @llvm.aarch64.neon.sabd.v4i32(<4 x i32> %2, <4 x i32> %4)
+  %6 = shufflevector <4 x i32> %5, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 3, i32 3>
+  ret <4 x i32> %6
+}

From 6ec467297db272d5ae8f76e695296bd4945dedc8 Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Fri, 12 Apr 2024 08:34:43 -0400
Subject: [PATCH 241/886] [BOLT][NFC] Adjust misleading comment & formatting
 (#88409)

This originally dealt with tbss, but now handles any bss-like section.
So the comment is inaccurate. Also, the `{}` on the messaging seem
unnecessary.
---
 bolt/lib/Rewrite/RewriteInstance.cpp | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index eea66454b289c..cbbe7d00d35dc 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -4300,7 +4300,7 @@ RewriteInstance::getOutputSections(ELFObjectFile<ELFT> *File,
   for (auto &SectionKV : OutputSections) {
     ELFShdrTy &Section = SectionKV.second;
 
-    // Ignore TLS sections as they don't take any space in the file.
+    // Ignore NOBITS sections as they don't take any space in the file.
     if (Section.sh_type == ELF::SHT_NOBITS)
       continue;
 
@@ -4308,10 +4308,9 @@ RewriteInstance::getOutputSections(ELFObjectFile<ELFT> *File,
     // placed in different loadable segments.
     if (PrevSection &&
         PrevSection->sh_offset + PrevSection->sh_size > Section.sh_offset) {
-      if (opts::Verbosity > 1) {
+      if (opts::Verbosity > 1)
         BC->outs() << "BOLT-INFO: adjusting size for section "
                    << PrevBinSec->getOutputName() << '\n';
-      }
       PrevSection->sh_size = Section.sh_offset - PrevSection->sh_offset;
     }
 

From f220d26eb1ab5b588215acf538ca82bb3c8798cc Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Fri, 12 Apr 2024 13:43:48 +0100
Subject: [PATCH 242/886] [flang][runtime] Add ACCESS library procedure
 (#88395)

This is a GNU extension:
https://gcc.gnu.org/onlinedocs/gfortran/ACCESS.html

Used in SALMON:
https://salmon-tddft.jp/download.html

Unfortunately the intrinsic takes a file path to operate on so there
isn't an easy way to make the test robust. The unit test expects to be
able to create, set read write and execute permissions, and delete files
called
  `std::filesystem::temp_directory_path() / <test_name>.<pid>`

The test will fail if a file already exists with that name.

I have not implemented the intrinsic on Windows because this is wrapping
a POSIX system call and Windows doesn't support all of the permission
bits tested by the intrinsic. I don't have a Windows machine easily
available to check if Gfortran implements this intrinsic on Windows.
---
 flang/docs/Intrinsics.md                 |   8 +
 flang/include/flang/Runtime/extensions.h |   7 +
 flang/runtime/extensions.cpp             |  73 +++++
 flang/unittests/Runtime/AccessTest.cpp   | 390 +++++++++++++++++++++++
 flang/unittests/Runtime/CMakeLists.txt   |   1 +
 5 files changed, 479 insertions(+)
 create mode 100644 flang/unittests/Runtime/AccessTest.cpp

diff --git a/flang/docs/Intrinsics.md b/flang/docs/Intrinsics.md
index ccb93e104dab6..848619cb65d90 100644
--- a/flang/docs/Intrinsics.md
+++ b/flang/docs/Intrinsics.md
@@ -657,6 +657,14 @@ CALL CO_REDUCE
 CALL CO_SUM
 ```
 
+### Inquiry Functions
+ACCESS (GNU extension) is not supported on Windows. Otherwise:
+```
+CHARACTER(LEN=*) :: path = 'path/to/file'
+IF (ACCESS(path, 'rwx')) &
+  ...
+```
+
 ## Non-standard intrinsics
 ### PGI
 ```
diff --git a/flang/include/flang/Runtime/extensions.h b/flang/include/flang/Runtime/extensions.h
index 7d0952206fc19..fef651f3b2eed 100644
--- a/flang/include/flang/Runtime/extensions.h
+++ b/flang/include/flang/Runtime/extensions.h
@@ -44,5 +44,12 @@ std::int64_t RTNAME(Signal)(std::int64_t number, void (*handler)(int));
 // GNU extension subroutine SLEEP(SECONDS)
 void RTNAME(Sleep)(std::int64_t seconds);
 
+// GNU extension function ACCESS(NAME, MODE)
+// TODO: not supported on Windows
+#ifndef _WIN32
+std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
+    std::int64_t nameLength, const char *mode, std::int64_t modeLength);
+#endif
+
 } // extern "C"
 #endif // FORTRAN_RUNTIME_EXTENSIONS_H_
diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 3ac98000335d7..12498b502ae1c 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -17,6 +17,7 @@
 #include "flang/Runtime/entry-names.h"
 #include "flang/Runtime/io-api.h"
 #include <chrono>
+#include <cstring>
 #include <ctime>
 #include <signal.h>
 #include <thread>
@@ -138,5 +139,77 @@ void RTNAME(Sleep)(std::int64_t seconds) {
   std::this_thread::sleep_for(std::chrono::seconds(seconds));
 }
 
+// TODO: not supported on Windows
+#ifndef _WIN32
+std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
+    std::int64_t nameLength, const char *mode, std::int64_t modeLength) {
+  std::int64_t ret{-1};
+  if (nameLength <= 0 || modeLength <= 0 || !name || !mode) {
+    return ret;
+  }
+
+  // ensure name is null terminated
+  char *newName{nullptr};
+  if (name[nameLength - 1] != '\0') {
+    newName = static_cast<char *>(std::malloc(nameLength + 1));
+    std::memcpy(newName, name, nameLength);
+    newName[nameLength] = '\0';
+    name = newName;
+  }
+
+  // calculate mode
+  bool read{false};
+  bool write{false};
+  bool execute{false};
+  bool exists{false};
+  int imode{0};
+
+  for (std::int64_t i = 0; i < modeLength; ++i) {
+    switch (mode[i]) {
+    case 'r':
+      read = true;
+      break;
+    case 'w':
+      write = true;
+      break;
+    case 'x':
+      execute = true;
+      break;
+    case ' ':
+      exists = true;
+      break;
+    default:
+      // invalid mode
+      goto cleanup;
+    }
+  }
+  if (!read && !write && !execute && !exists) {
+    // invalid mode
+    goto cleanup;
+  }
+
+  if (!read && !write && !execute) {
+    imode = F_OK;
+  } else {
+    if (read) {
+      imode |= R_OK;
+    }
+    if (write) {
+      imode |= W_OK;
+    }
+    if (execute) {
+      imode |= X_OK;
+    }
+  }
+  ret = access(name, imode);
+
+cleanup:
+  if (newName) {
+    free(newName);
+  }
+  return ret;
+}
+#endif
+
 } // namespace Fortran::runtime
 } // extern "C"
diff --git a/flang/unittests/Runtime/AccessTest.cpp b/flang/unittests/Runtime/AccessTest.cpp
new file mode 100644
index 0000000000000..5c8a634192d92
--- /dev/null
+++ b/flang/unittests/Runtime/AccessTest.cpp
@@ -0,0 +1,390 @@
+//===-- flang/unittests/Runtime/AccessTest.cpp ----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// TODO: ACCESS is not yet implemented on Windows
+#ifndef _WIN32
+
+#include "CrashHandlerFixture.h"
+#include "gtest/gtest.h"
+#include "flang/Runtime/extensions.h"
+
+#include <fcntl.h>
+#include <filesystem>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+namespace {
+
+struct AccessTests : public CrashHandlerFixture {};
+
+struct AccessType {
+  bool read{false};
+  bool write{false};
+  bool execute{false};
+  bool exists{false};
+};
+
+} // namespace
+
+static std::string addPIDSuffix(const char *name) {
+  std::stringstream ss;
+  ss << name;
+  ss << '.';
+
+  ss << getpid();
+
+  return ss.str();
+}
+
+static std::filesystem::path createTemporaryFile(
+    const char *name, const AccessType &accessType) {
+  std::filesystem::path path{
+      std::filesystem::temp_directory_path() / addPIDSuffix(name)};
+
+  // O_CREAT | O_EXCL enforces that this file is newly created by this call.
+  // This feels risky. If we don't have permission to create files in the
+  // temporary directory or if the files already exist, the test will fail.
+  // But we can't use std::tmpfile() because we need a path to the file and
+  // to control the filesystem permissions
+  mode_t mode{0};
+  if (accessType.read) {
+    mode |= S_IRUSR;
+  }
+  if (accessType.write) {
+    mode |= S_IWUSR;
+  }
+  if (accessType.execute) {
+    mode |= S_IXUSR;
+  }
+
+  int file = open(path.c_str(), O_CREAT | O_EXCL, mode);
+  if (file == -1) {
+    return {};
+  }
+
+  close(file);
+
+  return path;
+}
+
+static std::int64_t callAccess(
+    const std::filesystem::path &path, const AccessType &accessType) {
+  const char *cpath{path.c_str()};
+  std::int64_t pathlen = std::strlen(cpath);
+
+  std::string mode;
+  if (accessType.read) {
+    mode += 'r';
+  }
+  if (accessType.write) {
+    mode += 'w';
+  }
+  if (accessType.execute) {
+    mode += 'x';
+  }
+  if (accessType.exists) {
+    mode += ' ';
+  }
+
+  const char *cmode = mode.c_str();
+  std::int64_t modelen = std::strlen(cmode);
+
+  return FORTRAN_PROCEDURE_NAME(access)(cpath, pathlen, cmode, modelen);
+}
+
+TEST(AccessTests, TestExists) {
+  AccessType accessType;
+  accessType.exists = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotExists) {
+  std::filesystem::path nonExistant{addPIDSuffix(__func__)};
+  ASSERT_FALSE(std::filesystem::exists(nonExistant));
+
+  AccessType accessType;
+  accessType.exists = true;
+  std::int64_t res = callAccess(nonExistant, accessType);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestRead) {
+  AccessType accessType;
+  accessType.read = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotRead) {
+  AccessType accessType;
+  accessType.read = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestWrite) {
+  AccessType accessType;
+  accessType.write = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotWrite) {
+  AccessType accessType;
+  accessType.write = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestReadWrite) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite0) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite1) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite2) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestExecute) {
+  AccessType accessType;
+  accessType.execute = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotExecute) {
+  AccessType accessType;
+  accessType.execute = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestRWX) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX0) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = false;
+  accessType.execute = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX1) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+  accessType.execute = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX2) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = false;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX3) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+  accessType.execute = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX4) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = true;
+  accessType.execute = true;
+
+  std::filesystem::path path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  std::filesystem::remove(path);
+
+  ASSERT_NE(res, 0);
+}
+
+#endif // !_WIN32
diff --git a/flang/unittests/Runtime/CMakeLists.txt b/flang/unittests/Runtime/CMakeLists.txt
index 23f02aa751246..f7caacad3a598 100644
--- a/flang/unittests/Runtime/CMakeLists.txt
+++ b/flang/unittests/Runtime/CMakeLists.txt
@@ -1,4 +1,5 @@
 add_flang_unittest(FlangRuntimeTests
+  AccessTest.cpp
   Allocatable.cpp
   ArrayConstructor.cpp
   BufferTest.cpp

From 7e7468c9ed4e29fe8327525923fd8ab0169b1c07 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Fri, 12 Apr 2024 13:50:39 +0100
Subject: [PATCH 243/886] Revert "[flang][runtime] Add ACCESS library
 procedure" (#88507)

Reverts llvm/llvm-project#88395

This broke the powerpc buildbot. That build doesn't support using
`std::filesystem` in flang unit tests.
---
 flang/docs/Intrinsics.md                 |   8 -
 flang/include/flang/Runtime/extensions.h |   7 -
 flang/runtime/extensions.cpp             |  73 -----
 flang/unittests/Runtime/AccessTest.cpp   | 390 -----------------------
 flang/unittests/Runtime/CMakeLists.txt   |   1 -
 5 files changed, 479 deletions(-)
 delete mode 100644 flang/unittests/Runtime/AccessTest.cpp

diff --git a/flang/docs/Intrinsics.md b/flang/docs/Intrinsics.md
index 848619cb65d90..ccb93e104dab6 100644
--- a/flang/docs/Intrinsics.md
+++ b/flang/docs/Intrinsics.md
@@ -657,14 +657,6 @@ CALL CO_REDUCE
 CALL CO_SUM
 ```
 
-### Inquiry Functions
-ACCESS (GNU extension) is not supported on Windows. Otherwise:
-```
-CHARACTER(LEN=*) :: path = 'path/to/file'
-IF (ACCESS(path, 'rwx')) &
-  ...
-```
-
 ## Non-standard intrinsics
 ### PGI
 ```
diff --git a/flang/include/flang/Runtime/extensions.h b/flang/include/flang/Runtime/extensions.h
index fef651f3b2eed..7d0952206fc19 100644
--- a/flang/include/flang/Runtime/extensions.h
+++ b/flang/include/flang/Runtime/extensions.h
@@ -44,12 +44,5 @@ std::int64_t RTNAME(Signal)(std::int64_t number, void (*handler)(int));
 // GNU extension subroutine SLEEP(SECONDS)
 void RTNAME(Sleep)(std::int64_t seconds);
 
-// GNU extension function ACCESS(NAME, MODE)
-// TODO: not supported on Windows
-#ifndef _WIN32
-std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
-    std::int64_t nameLength, const char *mode, std::int64_t modeLength);
-#endif
-
 } // extern "C"
 #endif // FORTRAN_RUNTIME_EXTENSIONS_H_
diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 12498b502ae1c..3ac98000335d7 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -17,7 +17,6 @@
 #include "flang/Runtime/entry-names.h"
 #include "flang/Runtime/io-api.h"
 #include <chrono>
-#include <cstring>
 #include <ctime>
 #include <signal.h>
 #include <thread>
@@ -139,77 +138,5 @@ void RTNAME(Sleep)(std::int64_t seconds) {
   std::this_thread::sleep_for(std::chrono::seconds(seconds));
 }
 
-// TODO: not supported on Windows
-#ifndef _WIN32
-std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
-    std::int64_t nameLength, const char *mode, std::int64_t modeLength) {
-  std::int64_t ret{-1};
-  if (nameLength <= 0 || modeLength <= 0 || !name || !mode) {
-    return ret;
-  }
-
-  // ensure name is null terminated
-  char *newName{nullptr};
-  if (name[nameLength - 1] != '\0') {
-    newName = static_cast<char *>(std::malloc(nameLength + 1));
-    std::memcpy(newName, name, nameLength);
-    newName[nameLength] = '\0';
-    name = newName;
-  }
-
-  // calculate mode
-  bool read{false};
-  bool write{false};
-  bool execute{false};
-  bool exists{false};
-  int imode{0};
-
-  for (std::int64_t i = 0; i < modeLength; ++i) {
-    switch (mode[i]) {
-    case 'r':
-      read = true;
-      break;
-    case 'w':
-      write = true;
-      break;
-    case 'x':
-      execute = true;
-      break;
-    case ' ':
-      exists = true;
-      break;
-    default:
-      // invalid mode
-      goto cleanup;
-    }
-  }
-  if (!read && !write && !execute && !exists) {
-    // invalid mode
-    goto cleanup;
-  }
-
-  if (!read && !write && !execute) {
-    imode = F_OK;
-  } else {
-    if (read) {
-      imode |= R_OK;
-    }
-    if (write) {
-      imode |= W_OK;
-    }
-    if (execute) {
-      imode |= X_OK;
-    }
-  }
-  ret = access(name, imode);
-
-cleanup:
-  if (newName) {
-    free(newName);
-  }
-  return ret;
-}
-#endif
-
 } // namespace Fortran::runtime
 } // extern "C"
diff --git a/flang/unittests/Runtime/AccessTest.cpp b/flang/unittests/Runtime/AccessTest.cpp
deleted file mode 100644
index 5c8a634192d92..0000000000000
--- a/flang/unittests/Runtime/AccessTest.cpp
+++ /dev/null
@@ -1,390 +0,0 @@
-//===-- flang/unittests/Runtime/AccessTest.cpp ----------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// TODO: ACCESS is not yet implemented on Windows
-#ifndef _WIN32
-
-#include "CrashHandlerFixture.h"
-#include "gtest/gtest.h"
-#include "flang/Runtime/extensions.h"
-
-#include <fcntl.h>
-#include <filesystem>
-#include <sys/stat.h>
-#include <sys/types.h>
-#include <unistd.h>
-
-namespace {
-
-struct AccessTests : public CrashHandlerFixture {};
-
-struct AccessType {
-  bool read{false};
-  bool write{false};
-  bool execute{false};
-  bool exists{false};
-};
-
-} // namespace
-
-static std::string addPIDSuffix(const char *name) {
-  std::stringstream ss;
-  ss << name;
-  ss << '.';
-
-  ss << getpid();
-
-  return ss.str();
-}
-
-static std::filesystem::path createTemporaryFile(
-    const char *name, const AccessType &accessType) {
-  std::filesystem::path path{
-      std::filesystem::temp_directory_path() / addPIDSuffix(name)};
-
-  // O_CREAT | O_EXCL enforces that this file is newly created by this call.
-  // This feels risky. If we don't have permission to create files in the
-  // temporary directory or if the files already exist, the test will fail.
-  // But we can't use std::tmpfile() because we need a path to the file and
-  // to control the filesystem permissions
-  mode_t mode{0};
-  if (accessType.read) {
-    mode |= S_IRUSR;
-  }
-  if (accessType.write) {
-    mode |= S_IWUSR;
-  }
-  if (accessType.execute) {
-    mode |= S_IXUSR;
-  }
-
-  int file = open(path.c_str(), O_CREAT | O_EXCL, mode);
-  if (file == -1) {
-    return {};
-  }
-
-  close(file);
-
-  return path;
-}
-
-static std::int64_t callAccess(
-    const std::filesystem::path &path, const AccessType &accessType) {
-  const char *cpath{path.c_str()};
-  std::int64_t pathlen = std::strlen(cpath);
-
-  std::string mode;
-  if (accessType.read) {
-    mode += 'r';
-  }
-  if (accessType.write) {
-    mode += 'w';
-  }
-  if (accessType.execute) {
-    mode += 'x';
-  }
-  if (accessType.exists) {
-    mode += ' ';
-  }
-
-  const char *cmode = mode.c_str();
-  std::int64_t modelen = std::strlen(cmode);
-
-  return FORTRAN_PROCEDURE_NAME(access)(cpath, pathlen, cmode, modelen);
-}
-
-TEST(AccessTests, TestExists) {
-  AccessType accessType;
-  accessType.exists = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotExists) {
-  std::filesystem::path nonExistant{addPIDSuffix(__func__)};
-  ASSERT_FALSE(std::filesystem::exists(nonExistant));
-
-  AccessType accessType;
-  accessType.exists = true;
-  std::int64_t res = callAccess(nonExistant, accessType);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestRead) {
-  AccessType accessType;
-  accessType.read = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotRead) {
-  AccessType accessType;
-  accessType.read = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestWrite) {
-  AccessType accessType;
-  accessType.write = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotWrite) {
-  AccessType accessType;
-  accessType.write = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.write = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestReadWrite) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotReadWrite0) {
-  AccessType accessType;
-  accessType.read = false;
-  accessType.write = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotReadWrite1) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotReadWrite2) {
-  AccessType accessType;
-  accessType.read = false;
-  accessType.write = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestExecute) {
-  AccessType accessType;
-  accessType.execute = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotExecute) {
-  AccessType accessType;
-  accessType.execute = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestRWX) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_EQ(res, 0);
-}
-
-TEST(AccessTests, TestNotRWX0) {
-  AccessType accessType;
-  accessType.read = false;
-  accessType.write = false;
-  accessType.execute = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotRWX1) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = false;
-  accessType.execute = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotRWX2) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = false;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotRWX3) {
-  AccessType accessType;
-  accessType.read = true;
-  accessType.write = false;
-  accessType.execute = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-TEST(AccessTests, TestNotRWX4) {
-  AccessType accessType;
-  accessType.read = false;
-  accessType.write = true;
-  accessType.execute = true;
-
-  std::filesystem::path path = createTemporaryFile(__func__, accessType);
-  ASSERT_FALSE(path.empty());
-
-  accessType.read = true;
-  accessType.write = true;
-  accessType.execute = true;
-  std::int64_t res = callAccess(path, accessType);
-
-  std::filesystem::remove(path);
-
-  ASSERT_NE(res, 0);
-}
-
-#endif // !_WIN32
diff --git a/flang/unittests/Runtime/CMakeLists.txt b/flang/unittests/Runtime/CMakeLists.txt
index f7caacad3a598..23f02aa751246 100644
--- a/flang/unittests/Runtime/CMakeLists.txt
+++ b/flang/unittests/Runtime/CMakeLists.txt
@@ -1,5 +1,4 @@
 add_flang_unittest(FlangRuntimeTests
-  AccessTest.cpp
   Allocatable.cpp
   ArrayConstructor.cpp
   BufferTest.cpp

From ea3d0db130b9a6c4678c11dd8bc48e5630624b62 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 12 Apr 2024 13:51:41 +0100
Subject: [PATCH 244/886] [VectorCombine] foldShuffleOfCastops - ensure we can
 scale shuffle masks between bitcasted vector types

Don't just assert that the src/dst vector element counts are multiples of one another - in general IR this can actually happen.

Reported by @mikaelholmen
---
 llvm/lib/Transforms/Vectorize/VectorCombine.cpp   |  6 ++++++
 .../VectorCombine/X86/shuffle-of-casts.ll         | 15 +++++++++++++++
 2 files changed, 21 insertions(+)

diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 44cba60013afa..653e0ff7a3ece 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -1483,6 +1483,12 @@ bool VectorCombine::foldShuffleOfCastops(Instruction &I) {
   assert((NumDstElts == NumSrcElts || Opcode == Instruction::BitCast) &&
          "Only bitcasts expected to alter src/dst element counts");
 
+  // Check for bitcasting of unscalable vector types.
+  // e.g. <32 x i40> -> <40 x i32>
+  if (NumDstElts != NumSrcElts && (NumSrcElts % NumDstElts) != 0 &&
+      (NumDstElts % NumSrcElts) != 0)
+    return false;
+
   SmallVector<int, 16> NewMask;
   if (NumSrcElts >= NumDstElts) {
     // The bitcast is from wide to narrow/equal elements. The shuffle mask can
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
index 4f8cfea146ead..7d9f7e390b9c0 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
@@ -239,6 +239,21 @@ define <16 x i16> @revpair_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
   ret <16 x i16> %r
 }
 
+; negative - bitcasts (unscalable element counts)
+
+define <4 x i32> @shuffle_bitcast_v32i40_v4i32(<32 x i40> %a0, <32 x i40> %a1) {
+; CHECK-LABEL: @shuffle_bitcast_v32i40_v4i32(
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <32 x i40> [[A0:%.*]] to <40 x i32>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <32 x i40> [[A1:%.*]] to <40 x i32>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <40 x i32> [[X0]], <40 x i32> [[X1]], <4 x i32> <i32 0, i32 42, i32 poison, i32 poison>
+; CHECK-NEXT:    ret <4 x i32> [[R]]
+;
+  %x0 = bitcast <32 x i40> %a0 to <40 x i32>
+  %x1 = bitcast <32 x i40> %a1 to <40 x i32>
+  %r = shufflevector <40 x i32> %x0, <40 x i32> %x1, <4 x i32> <i32 0, i32 42, i32 poison, i32 poison>
+  ret <4 x i32> %r
+}
+
 ; negative - src type mismatch
 
 define <8 x i32> @concat_sext_v4i8_v4i16_v8i32(<4 x i8> %a0, <4 x i16> %a1) {

From 3fa830804e1b90d38d0a62084b5b65e70c83c345 Mon Sep 17 00:00:00 2001
From: Brandon Wu <brandon.wu@sifive.com>
Date: Fri, 12 Apr 2024 21:11:45 +0800
Subject: [PATCH 245/886] Revert "[RISCV] RISCV vector calling convention (2/2)
 (#79096)" (#88511)

This reverts commit 29e8bfc13c6078ed07e6474e8c9634c42aa2f6f4.
This patch didn't handle vector return type correctly.
---
 .../Target/RISCV/GISel/RISCVCallLowering.cpp  |  55 +++---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 180 ++++--------------
 llvm/lib/Target/RISCV/RISCVISelLowering.h     |  56 +-----
 llvm/test/CodeGen/RISCV/rvv/calling-conv.ll   |  87 ---------
 .../RISCV/rvv/vector-deinterleave-load.ll     |   6 +-
 .../CodeGen/RISCV/rvv/vector-deinterleave.ll  |  19 +-
 6 files changed, 83 insertions(+), 320 deletions(-)

diff --git a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
index 8af4bc658409d..45e19cdea300b 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
@@ -34,15 +34,14 @@ struct RISCVOutgoingValueAssigner : public CallLowering::OutgoingValueAssigner {
   // Whether this is assigning args for a return.
   bool IsRet;
 
-  RVVArgDispatcher &RVVDispatcher;
+  // true if assignArg has been called for a mask argument, false otherwise.
+  bool AssignedFirstMaskArg = false;
 
 public:
   RISCVOutgoingValueAssigner(
-      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet,
-      RVVArgDispatcher &RVVDispatcher)
+      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet)
       : CallLowering::OutgoingValueAssigner(nullptr),
-        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet),
-        RVVDispatcher(RVVDispatcher) {}
+        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet) {}
 
   bool assignArg(unsigned ValNo, EVT OrigVT, MVT ValVT, MVT LocVT,
                  CCValAssign::LocInfo LocInfo,
@@ -52,9 +51,16 @@ struct RISCVOutgoingValueAssigner : public CallLowering::OutgoingValueAssigner {
     const DataLayout &DL = MF.getDataLayout();
     const RISCVSubtarget &Subtarget = MF.getSubtarget<RISCVSubtarget>();
 
+    std::optional<unsigned> FirstMaskArgument;
+    if (Subtarget.hasVInstructions() && !AssignedFirstMaskArg &&
+        ValVT.isVector() && ValVT.getVectorElementType() == MVT::i1) {
+      FirstMaskArgument = ValNo;
+      AssignedFirstMaskArg = true;
+    }
+
     if (RISCVAssignFn(DL, Subtarget.getTargetABI(), ValNo, ValVT, LocVT,
                       LocInfo, Flags, State, Info.IsFixed, IsRet, Info.Ty,
-                      *Subtarget.getTargetLowering(), RVVDispatcher))
+                      *Subtarget.getTargetLowering(), FirstMaskArgument))
       return true;
 
     StackSize = State.getStackSize();
@@ -175,15 +181,14 @@ struct RISCVIncomingValueAssigner : public CallLowering::IncomingValueAssigner {
   // Whether this is assigning args from a return.
   bool IsRet;
 
-  RVVArgDispatcher &RVVDispatcher;
+  // true if assignArg has been called for a mask argument, false otherwise.
+  bool AssignedFirstMaskArg = false;
 
 public:
   RISCVIncomingValueAssigner(
-      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet,
-      RVVArgDispatcher &RVVDispatcher)
+      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet)
       : CallLowering::IncomingValueAssigner(nullptr),
-        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet),
-        RVVDispatcher(RVVDispatcher) {}
+        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet) {}
 
   bool assignArg(unsigned ValNo, EVT OrigVT, MVT ValVT, MVT LocVT,
                  CCValAssign::LocInfo LocInfo,
@@ -196,9 +201,16 @@ struct RISCVIncomingValueAssigner : public CallLowering::IncomingValueAssigner {
     if (LocVT.isScalableVector())
       MF.getInfo<RISCVMachineFunctionInfo>()->setIsVectorCall();
 
+    std::optional<unsigned> FirstMaskArgument;
+    if (Subtarget.hasVInstructions() && !AssignedFirstMaskArg &&
+        ValVT.isVector() && ValVT.getVectorElementType() == MVT::i1) {
+      FirstMaskArgument = ValNo;
+      AssignedFirstMaskArg = true;
+    }
+
     if (RISCVAssignFn(DL, Subtarget.getTargetABI(), ValNo, ValVT, LocVT,
                       LocInfo, Flags, State, /*IsFixed=*/true, IsRet, Info.Ty,
-                      *Subtarget.getTargetLowering(), RVVDispatcher))
+                      *Subtarget.getTargetLowering(), FirstMaskArgument))
       return true;
 
     StackSize = State.getStackSize();
@@ -408,11 +420,9 @@ bool RISCVCallLowering::lowerReturnVal(MachineIRBuilder &MIRBuilder,
   SmallVector<ArgInfo, 4> SplitRetInfos;
   splitToValueTypes(OrigRetInfo, SplitRetInfos, DL, CC);
 
-  RVVArgDispatcher Dispatcher{&MF, getTLI<RISCVTargetLowering>(),
-                              F.getReturnType()};
   RISCVOutgoingValueAssigner Assigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/true, Dispatcher);
+      /*IsRet=*/true);
   RISCVOutgoingValueHandler Handler(MIRBuilder, MF.getRegInfo(), Ret);
   return determineAndHandleAssignments(Handler, Assigner, SplitRetInfos,
                                        MIRBuilder, CC, F.isVarArg());
@@ -521,7 +531,6 @@ bool RISCVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
   CallingConv::ID CC = F.getCallingConv();
 
   SmallVector<ArgInfo, 32> SplitArgInfos;
-  SmallVector<Type *, 4> TypeList;
   unsigned Index = 0;
   for (auto &Arg : F.args()) {
     // Construct the ArgInfo object from destination register and argument type.
@@ -533,15 +542,12 @@ bool RISCVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
     // correspondingly and appended to SplitArgInfos.
     splitToValueTypes(AInfo, SplitArgInfos, DL, CC);
 
-    TypeList.push_back(Arg.getType());
-
     ++Index;
   }
 
-  RVVArgDispatcher Dispatcher{&MF, getTLI<RISCVTargetLowering>(), TypeList};
   RISCVIncomingValueAssigner Assigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/false, Dispatcher);
+      /*IsRet=*/false);
   RISCVFormalArgHandler Handler(MIRBuilder, MF.getRegInfo());
 
   SmallVector<CCValAssign, 16> ArgLocs;
@@ -579,13 +585,11 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
 
   SmallVector<ArgInfo, 32> SplitArgInfos;
   SmallVector<ISD::OutputArg, 8> Outs;
-  SmallVector<Type *, 4> TypeList;
   for (auto &AInfo : Info.OrigArgs) {
     // Handle any required unmerging of split value types from a given VReg into
     // physical registers. ArgInfo objects are constructed correspondingly and
     // appended to SplitArgInfos.
     splitToValueTypes(AInfo, SplitArgInfos, DL, CC);
-    TypeList.push_back(AInfo.Ty);
   }
 
   // TODO: Support tail calls.
@@ -603,10 +607,9 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
   Call.addRegMask(TRI->getCallPreservedMask(MF, Info.CallConv));
 
-  RVVArgDispatcher ArgDispatcher{&MF, getTLI<RISCVTargetLowering>(), TypeList};
   RISCVOutgoingValueAssigner ArgAssigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/false, ArgDispatcher);
+      /*IsRet=*/false);
   RISCVOutgoingValueHandler ArgHandler(MIRBuilder, MF.getRegInfo(), Call);
   if (!determineAndHandleAssignments(ArgHandler, ArgAssigner, SplitArgInfos,
                                      MIRBuilder, CC, Info.IsVarArg))
@@ -634,11 +637,9 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   SmallVector<ArgInfo, 4> SplitRetInfos;
   splitToValueTypes(Info.OrigRet, SplitRetInfos, DL, CC);
 
-  RVVArgDispatcher RetDispatcher{&MF, getTLI<RISCVTargetLowering>(),
-                                 F.getReturnType()};
   RISCVIncomingValueAssigner RetAssigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/true, RetDispatcher);
+      /*IsRet=*/true);
   RISCVCallReturnHandler RetHandler(MIRBuilder, MF.getRegInfo(), Call);
   if (!determineAndHandleAssignments(RetHandler, RetAssigner, SplitRetInfos,
                                      MIRBuilder, CC, Info.IsVarArg))
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 3e7bc8c2367de..5a572002091ff 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -22,7 +22,6 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/MemoryLocation.h"
 #include "llvm/Analysis/VectorUtils.h"
-#include "llvm/CodeGen/Analysis.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
@@ -18079,12 +18078,33 @@ static bool CC_RISCVAssign2XLen(unsigned XLen, CCState &State, CCValAssign VA1,
   return false;
 }
 
+static unsigned allocateRVVReg(MVT ValVT, unsigned ValNo,
+                               std::optional<unsigned> FirstMaskArgument,
+                               CCState &State, const RISCVTargetLowering &TLI) {
+  const TargetRegisterClass *RC = TLI.getRegClassFor(ValVT);
+  if (RC == &RISCV::VRRegClass) {
+    // Assign the first mask argument to V0.
+    // This is an interim calling convention and it may be changed in the
+    // future.
+    if (FirstMaskArgument && ValNo == *FirstMaskArgument)
+      return State.AllocateReg(RISCV::V0);
+    return State.AllocateReg(ArgVRs);
+  }
+  if (RC == &RISCV::VRM2RegClass)
+    return State.AllocateReg(ArgVRM2s);
+  if (RC == &RISCV::VRM4RegClass)
+    return State.AllocateReg(ArgVRM4s);
+  if (RC == &RISCV::VRM8RegClass)
+    return State.AllocateReg(ArgVRM8s);
+  llvm_unreachable("Unhandled register class for ValueType");
+}
+
 // Implements the RISC-V calling convention. Returns true upon failure.
 bool RISCV::CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
                      MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
                      ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
                      bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-                     RVVArgDispatcher &RVVDispatcher) {
+                     std::optional<unsigned> FirstMaskArgument) {
   unsigned XLen = DL.getLargestLegalIntTypeSizeInBits();
   assert(XLen == 32 || XLen == 64);
   MVT XLenVT = XLen == 32 ? MVT::i32 : MVT::i64;
@@ -18253,7 +18273,7 @@ bool RISCV::CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
   else if (ValVT == MVT::f64 && !UseGPRForF64)
     Reg = State.AllocateReg(ArgFPR64s);
   else if (ValVT.isVector()) {
-    Reg = RVVDispatcher.getNextPhysReg();
+    Reg = allocateRVVReg(ValVT, ValNo, FirstMaskArgument, State, TLI);
     if (!Reg) {
       // For return values, the vector must be passed fully via registers or
       // via the stack.
@@ -18339,13 +18359,9 @@ void RISCVTargetLowering::analyzeInputArgs(
   unsigned NumArgs = Ins.size();
   FunctionType *FType = MF.getFunction().getFunctionType();
 
-  SmallVector<Type *, 4> TypeList;
-  if (IsRet)
-    TypeList.push_back(MF.getFunction().getReturnType());
-  else
-    for (const Argument &Arg : MF.getFunction().args())
-      TypeList.push_back(Arg.getType());
-  RVVArgDispatcher Dispatcher{&MF, this, TypeList};
+  std::optional<unsigned> FirstMaskArgument;
+  if (Subtarget.hasVInstructions())
+    FirstMaskArgument = preAssignMask(Ins);
 
   for (unsigned i = 0; i != NumArgs; ++i) {
     MVT ArgVT = Ins[i].VT;
@@ -18360,7 +18376,7 @@ void RISCVTargetLowering::analyzeInputArgs(
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (Fn(MF.getDataLayout(), ABI, i, ArgVT, ArgVT, CCValAssign::Full,
            ArgFlags, CCInfo, /*IsFixed=*/true, IsRet, ArgTy, *this,
-           Dispatcher)) {
+           FirstMaskArgument)) {
       LLVM_DEBUG(dbgs() << "InputArg #" << i << " has unhandled type "
                         << ArgVT << '\n');
       llvm_unreachable(nullptr);
@@ -18374,13 +18390,9 @@ void RISCVTargetLowering::analyzeOutputArgs(
     CallLoweringInfo *CLI, RISCVCCAssignFn Fn) const {
   unsigned NumArgs = Outs.size();
 
-  SmallVector<Type *, 4> TypeList;
-  if (IsRet)
-    TypeList.push_back(MF.getFunction().getReturnType());
-  else if (CLI)
-    for (const TargetLowering::ArgListEntry &Arg : CLI->getArgs())
-      TypeList.push_back(Arg.Ty);
-  RVVArgDispatcher Dispatcher{&MF, this, TypeList};
+  std::optional<unsigned> FirstMaskArgument;
+  if (Subtarget.hasVInstructions())
+    FirstMaskArgument = preAssignMask(Outs);
 
   for (unsigned i = 0; i != NumArgs; i++) {
     MVT ArgVT = Outs[i].VT;
@@ -18390,7 +18402,7 @@ void RISCVTargetLowering::analyzeOutputArgs(
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (Fn(MF.getDataLayout(), ABI, i, ArgVT, ArgVT, CCValAssign::Full,
            ArgFlags, CCInfo, Outs[i].IsFixed, IsRet, OrigTy, *this,
-           Dispatcher)) {
+           FirstMaskArgument)) {
       LLVM_DEBUG(dbgs() << "OutputArg #" << i << " has unhandled type "
                         << ArgVT << "\n");
       llvm_unreachable(nullptr);
@@ -18571,7 +18583,7 @@ bool RISCV::CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI,
                             ISD::ArgFlagsTy ArgFlags, CCState &State,
                             bool IsFixed, bool IsRet, Type *OrigTy,
                             const RISCVTargetLowering &TLI,
-                            RVVArgDispatcher &RVVDispatcher) {
+                            std::optional<unsigned> FirstMaskArgument) {
   if (LocVT == MVT::i32 || LocVT == MVT::i64) {
     if (unsigned Reg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
       State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
@@ -18649,14 +18661,13 @@ bool RISCV::CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI,
   }
 
   if (LocVT.isVector()) {
-    MCPhysReg AllocatedVReg = RVVDispatcher.getNextPhysReg();
-    if (AllocatedVReg) {
+    if (unsigned Reg =
+            allocateRVVReg(ValVT, ValNo, FirstMaskArgument, State, TLI)) {
       // Fixed-length vectors are located in the corresponding scalable-vector
       // container types.
       if (ValVT.isFixedLengthVector())
         LocVT = TLI.getContainerForFixedLengthVector(LocVT);
-      State.addLoc(
-          CCValAssign::getReg(ValNo, ValVT, AllocatedVReg, LocVT, LocInfo));
+      State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
     } else {
       // Try and pass the address via a "fast" GPR.
       if (unsigned GPRReg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
@@ -19284,15 +19295,17 @@ bool RISCVTargetLowering::CanLowerReturn(
   SmallVector<CCValAssign, 16> RVLocs;
   CCState CCInfo(CallConv, IsVarArg, MF, RVLocs, Context);
 
-  RVVArgDispatcher Dispatcher{&MF, this, MF.getFunction().getReturnType()};
+  std::optional<unsigned> FirstMaskArgument;
+  if (Subtarget.hasVInstructions())
+    FirstMaskArgument = preAssignMask(Outs);
 
   for (unsigned i = 0, e = Outs.size(); i != e; ++i) {
     MVT VT = Outs[i].VT;
     ISD::ArgFlagsTy ArgFlags = Outs[i].Flags;
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (RISCV::CC_RISCV(MF.getDataLayout(), ABI, i, VT, VT, CCValAssign::Full,
-                        ArgFlags, CCInfo, /*IsFixed=*/true, /*IsRet=*/true,
-                        nullptr, *this, Dispatcher))
+                 ArgFlags, CCInfo, /*IsFixed=*/true, /*IsRet=*/true, nullptr,
+                 *this, FirstMaskArgument))
       return false;
   }
   return true;
@@ -21089,119 +21102,6 @@ unsigned RISCVTargetLowering::getMinimumJumpTableEntries() const {
   return Subtarget.getMinimumJumpTableEntries();
 }
 
-void RVVArgDispatcher::constructArgInfos(ArrayRef<Type *> TypeList) {
-  const DataLayout &DL = MF->getDataLayout();
-  const Function &F = MF->getFunction();
-  LLVMContext &Context = F.getContext();
-
-  bool FirstVMaskAssigned = false;
-  for (Type *Ty : TypeList) {
-    StructType *STy = dyn_cast<StructType>(Ty);
-    if (STy && STy->containsHomogeneousScalableVectorTypes()) {
-      Type *ElemTy = STy->getTypeAtIndex(0U);
-      EVT VT = TLI->getValueType(DL, ElemTy);
-      MVT RegisterVT =
-          TLI->getRegisterTypeForCallingConv(Context, F.getCallingConv(), VT);
-
-      RVVArgInfos.push_back({STy->getNumElements(), RegisterVT, false});
-    } else {
-      SmallVector<EVT, 4> ValueVTs;
-      ComputeValueVTs(*TLI, DL, Ty, ValueVTs);
-
-      for (unsigned Value = 0, NumValues = ValueVTs.size(); Value != NumValues;
-           ++Value) {
-        EVT VT = ValueVTs[Value];
-        MVT RegisterVT =
-            TLI->getRegisterTypeForCallingConv(Context, F.getCallingConv(), VT);
-        unsigned NumRegs =
-            TLI->getNumRegistersForCallingConv(Context, F.getCallingConv(), VT);
-
-        // Skip non-RVV register type
-        if (!RegisterVT.isVector())
-          continue;
-
-        if (RegisterVT.isFixedLengthVector())
-          RegisterVT = TLI->getContainerForFixedLengthVector(RegisterVT);
-
-        if (!FirstVMaskAssigned &&
-            RegisterVT.getVectorElementType() == MVT::i1) {
-          RVVArgInfos.push_back({1, RegisterVT, true});
-          FirstVMaskAssigned = true;
-          --NumRegs;
-        }
-
-        RVVArgInfos.insert(RVVArgInfos.end(), NumRegs, {1, RegisterVT, false});
-      }
-    }
-  }
-}
-
-void RVVArgDispatcher::allocatePhysReg(unsigned NF, unsigned LMul,
-                                       unsigned StartReg) {
-  assert((StartReg % LMul) == 0 &&
-         "Start register number should be multiple of lmul");
-  const MCPhysReg *VRArrays;
-  switch (LMul) {
-  default:
-    report_fatal_error("Invalid lmul");
-  case 1:
-    VRArrays = ArgVRs;
-    break;
-  case 2:
-    VRArrays = ArgVRM2s;
-    break;
-  case 4:
-    VRArrays = ArgVRM4s;
-    break;
-  case 8:
-    VRArrays = ArgVRM8s;
-    break;
-  }
-
-  for (unsigned i = 0; i < NF; ++i)
-    if (StartReg)
-      AllocatedPhysRegs.push_back(VRArrays[(StartReg - 8) / LMul + i]);
-    else
-      AllocatedPhysRegs.push_back(MCPhysReg());
-}
-
-/// This function determines if each RVV argument is passed by register, if the
-/// argument can be assigned to a VR, then give it a specific register.
-/// Otherwise, assign the argument to 0 which is a invalid MCPhysReg.
-void RVVArgDispatcher::compute() {
-  uint32_t AssignedMap = 0;
-  auto allocate = [&](const RVVArgInfo &ArgInfo) {
-    // Allocate first vector mask argument to V0.
-    if (ArgInfo.FirstVMask) {
-      AllocatedPhysRegs.push_back(RISCV::V0);
-      return;
-    }
-
-    unsigned RegsNeeded = divideCeil(
-        ArgInfo.VT.getSizeInBits().getKnownMinValue(), RISCV::RVVBitsPerBlock);
-    unsigned TotalRegsNeeded = ArgInfo.NF * RegsNeeded;
-    for (unsigned StartReg = 0; StartReg + TotalRegsNeeded <= NumArgVRs;
-         StartReg += RegsNeeded) {
-      uint32_t Map = ((1 << TotalRegsNeeded) - 1) << StartReg;
-      if ((AssignedMap & Map) == 0) {
-        allocatePhysReg(ArgInfo.NF, RegsNeeded, StartReg + 8);
-        AssignedMap |= Map;
-        return;
-      }
-    }
-
-    allocatePhysReg(ArgInfo.NF, RegsNeeded, 0);
-  };
-
-  for (unsigned i = 0; i < RVVArgInfos.size(); ++i)
-    allocate(RVVArgInfos[i]);
-}
-
-MCPhysReg RVVArgDispatcher::getNextPhysReg() {
-  assert(CurIdx < AllocatedPhysRegs.size() && "Index out of range");
-  return AllocatedPhysRegs[CurIdx++];
-}
-
 namespace llvm::RISCVVIntrinsicsTable {
 
 #define GET_RISCVVIntrinsicsTable_IMPL
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.h b/llvm/lib/Target/RISCV/RISCVISelLowering.h
index c28552354bf42..ace5b3fd2b95b 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.h
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.h
@@ -24,7 +24,6 @@ namespace llvm {
 class InstructionCost;
 class RISCVSubtarget;
 struct RISCVRegisterInfo;
-class RVVArgDispatcher;
 
 namespace RISCVISD {
 // clang-format off
@@ -876,7 +875,7 @@ class RISCVTargetLowering : public TargetLowering {
                                ISD::ArgFlagsTy ArgFlags, CCState &State,
                                bool IsFixed, bool IsRet, Type *OrigTy,
                                const RISCVTargetLowering &TLI,
-                               RVVArgDispatcher &RVVDispatcher);
+                               std::optional<unsigned> FirstMaskArgument);
 
 private:
   void analyzeInputArgs(MachineFunction &MF, CCState &CCInfo,
@@ -1016,68 +1015,19 @@ class RISCVTargetLowering : public TargetLowering {
   unsigned getMinimumJumpTableEntries() const override;
 };
 
-/// As per the spec, the rules for passing vector arguments are as follows:
-///
-/// 1. For the first vector mask argument, use v0 to pass it.
-/// 2. For vector data arguments or rest vector mask arguments, starting from
-/// the v8 register, if a vector register group between v8-v23 that has not been
-/// allocated can be found and the first register number is a multiple of LMUL,
-/// then allocate this vector register group to the argument and mark these
-/// registers as allocated. Otherwise, pass it by reference and are replaced in
-/// the argument list with the address.
-/// 3. For tuple vector data arguments, starting from the v8 register, if
-/// NFIELDS consecutive vector register groups between v8-v23 that have not been
-/// allocated can be found and the first register number is a multiple of LMUL,
-/// then allocate these vector register groups to the argument and mark these
-/// registers as allocated. Otherwise, pass it by reference and are replaced in
-/// the argument list with the address.
-class RVVArgDispatcher {
-public:
-  static constexpr unsigned NumArgVRs = 16;
-
-  struct RVVArgInfo {
-    unsigned NF;
-    MVT VT;
-    bool FirstVMask = false;
-  };
-
-  RVVArgDispatcher(const MachineFunction *MF, const RISCVTargetLowering *TLI,
-                   ArrayRef<Type *> TypeList)
-      : MF(MF), TLI(TLI) {
-    constructArgInfos(TypeList);
-    compute();
-  }
-
-  MCPhysReg getNextPhysReg();
-
-private:
-  SmallVector<RVVArgInfo, 4> RVVArgInfos;
-  SmallVector<MCPhysReg, 4> AllocatedPhysRegs;
-
-  const MachineFunction *MF = nullptr;
-  const RISCVTargetLowering *TLI = nullptr;
-
-  unsigned CurIdx = 0;
-
-  void constructArgInfos(ArrayRef<Type *> TypeList);
-  void compute();
-  void allocatePhysReg(unsigned NF = 1, unsigned LMul = 1,
-                       unsigned StartReg = 0);
-};
-
 namespace RISCV {
 
 bool CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
               MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
               ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
               bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-              RVVArgDispatcher &RVVDispatcher);
+              std::optional<unsigned> FirstMaskArgument);
 
 bool CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
                      MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
                      ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
                      bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-                     RVVArgDispatcher &RVVDispatcher);
+                     std::optional<unsigned> FirstMaskArgument);
 
 bool CC_RISCV_GHC(unsigned ValNo, MVT ValVT, MVT LocVT,
                   CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags,
diff --git a/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll b/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
index 90edb994ce822..78e8700a9feff 100644
--- a/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
@@ -162,90 +162,3 @@ define void @caller_tuple_argument({<vscale x 4 x i32>, <vscale x 4 x i32>} %x)
 }
 
 declare void @callee_tuple_argument({<vscale x 4 x i32>, <vscale x 4 x i32>})
-
-; %0 -> v8
-; %1 -> v9
-define <vscale x 1 x i64> @case1(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1) {
-; CHECK-LABEL: case1:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v8, v9
-; CHECK-NEXT:    ret
-  %a = add <vscale x 1 x i64> %0, %1
-  ret <vscale x 1 x i64> %a
-}
-
-; %0 -> v8
-; %1 -> v10-v11
-; %2 -> v9
-define <vscale x 1 x i64> @case2_1(<vscale x 1 x i64> %0, <vscale x 2 x i64> %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case2_1:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v8, v9
-; CHECK-NEXT:    ret
-  %a = add <vscale x 1 x i64> %0, %2
-  ret <vscale x 1 x i64> %a
-}
-define <vscale x 2 x i64> @case2_2(<vscale x 1 x i64> %0, <vscale x 2 x i64> %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case2_2:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v10, v10
-; CHECK-NEXT:    ret
-  %a = add <vscale x 2 x i64> %1, %1
-  ret <vscale x 2 x i64> %a
-}
-
-; %0 -> v8
-; %1 -> {v10-v11, v12-v13}
-; %2 -> v9
-define <vscale x 1 x i64> @case3_1(<vscale x 1 x i64> %0, {<vscale x 2 x i64>, <vscale x 2 x i64>} %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case3_1:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v8, v9
-; CHECK-NEXT:    ret
-  %add = add <vscale x 1 x i64> %0, %2
-  ret <vscale x 1 x i64> %add
-}
-define <vscale x 2 x i64> @case3_2(<vscale x 1 x i64> %0, {<vscale x 2 x i64>, <vscale x 2 x i64>} %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case3_2:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v10, v12
-; CHECK-NEXT:    ret
-  %a = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } %1, 0
-  %b = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } %1, 1
-  %add = add <vscale x 2 x i64> %a, %b
-  ret <vscale x 2 x i64> %add
-}
-
-; %0 -> v8
-; %1 -> {by-ref, by-ref}
-; %2 -> v9
-define <vscale x 8 x i64> @case4_1(<vscale x 1 x i64> %0, {<vscale x 8 x i64>, <vscale x 8 x i64>} %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case4_1:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    csrr a1, vlenb
-; CHECK-NEXT:    slli a1, a1, 3
-; CHECK-NEXT:    add a1, a0, a1
-; CHECK-NEXT:    vl8re64.v v8, (a1)
-; CHECK-NEXT:    vl8re64.v v16, (a0)
-; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v16, v8
-; CHECK-NEXT:    ret
-  %a = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i64> } %1, 0
-  %b = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i64> } %1, 1
-  %add = add <vscale x 8 x i64> %a, %b
-  ret <vscale x 8 x i64> %add
-}
-define <vscale x 1 x i64> @case4_2(<vscale x 1 x i64> %0, {<vscale x 8 x i64>, <vscale x 8 x i64>} %1, <vscale x 1 x i64> %2) {
-; CHECK-LABEL: case4_2:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
-; CHECK-NEXT:    vadd.vv v8, v8, v9
-; CHECK-NEXT:    ret
-  %add = add <vscale x 1 x i64> %0, %2
-  ret <vscale x 1 x i64> %add
-}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
index 6a712080fda74..a320aecc6fce4 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
@@ -18,10 +18,10 @@ define {<vscale x 16 x i1>, <vscale x 16 x i1>} @vector_deinterleave_load_nxv16i
 ; CHECK-NEXT:    vmerge.vim v14, v10, 1, v0
 ; CHECK-NEXT:    vmv1r.v v0, v8
 ; CHECK-NEXT:    vmerge.vim v12, v10, 1, v0
-; CHECK-NEXT:    vnsrl.wi v10, v12, 0
-; CHECK-NEXT:    vmsne.vi v8, v10, 0
+; CHECK-NEXT:    vnsrl.wi v8, v12, 0
+; CHECK-NEXT:    vmsne.vi v0, v8, 0
 ; CHECK-NEXT:    vnsrl.wi v10, v12, 8
-; CHECK-NEXT:    vmsne.vi v9, v10, 0
+; CHECK-NEXT:    vmsne.vi v8, v10, 0
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 32 x i1>, ptr %p
   %retval = call {<vscale x 16 x i1>, <vscale x 16 x i1>} @llvm.experimental.vector.deinterleave2.nxv32i1(<vscale x 32 x i1> %vec)
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
index d98597fabcd95..ef4baf34d23f0 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
@@ -8,18 +8,18 @@ define {<vscale x 16 x i1>, <vscale x 16 x i1>} @vector_deinterleave_nxv16i1_nxv
 ; CHECK-LABEL: vector_deinterleave_nxv16i1_nxv32i1:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
-; CHECK-NEXT:    vmv.v.i v8, 0
-; CHECK-NEXT:    vmerge.vim v12, v8, 1, v0
+; CHECK-NEXT:    vmv.v.i v10, 0
+; CHECK-NEXT:    vmerge.vim v8, v10, 1, v0
 ; CHECK-NEXT:    csrr a0, vlenb
 ; CHECK-NEXT:    srli a0, a0, 2
 ; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, ma
 ; CHECK-NEXT:    vslidedown.vx v0, v0, a0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
-; CHECK-NEXT:    vmerge.vim v14, v8, 1, v0
-; CHECK-NEXT:    vnsrl.wi v10, v12, 0
-; CHECK-NEXT:    vmsne.vi v8, v10, 0
-; CHECK-NEXT:    vnsrl.wi v10, v12, 8
-; CHECK-NEXT:    vmsne.vi v9, v10, 0
+; CHECK-NEXT:    vmerge.vim v10, v10, 1, v0
+; CHECK-NEXT:    vnsrl.wi v12, v8, 0
+; CHECK-NEXT:    vmsne.vi v0, v12, 0
+; CHECK-NEXT:    vnsrl.wi v12, v8, 8
+; CHECK-NEXT:    vmsne.vi v8, v12, 0
 ; CHECK-NEXT:    ret
 %retval = call {<vscale x 16 x i1>, <vscale x 16 x i1>} @llvm.experimental.vector.deinterleave2.nxv32i1(<vscale x 32 x i1> %vec)
 ret {<vscale x 16 x i1>, <vscale x 16 x i1>} %retval
@@ -102,13 +102,12 @@ define {<vscale x 64 x i1>, <vscale x 64 x i1>} @vector_deinterleave_nxv64i1_nxv
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, ma
 ; CHECK-NEXT:    vnsrl.wi v28, v8, 0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, ma
-; CHECK-NEXT:    vmsne.vi v7, v24, 0
+; CHECK-NEXT:    vmsne.vi v0, v24, 0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, ma
 ; CHECK-NEXT:    vnsrl.wi v24, v16, 8
 ; CHECK-NEXT:    vnsrl.wi v28, v8, 8
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, ma
-; CHECK-NEXT:    vmsne.vi v9, v24, 0
-; CHECK-NEXT:    vmv1r.v v8, v7
+; CHECK-NEXT:    vmsne.vi v8, v24, 0
 ; CHECK-NEXT:    ret
 %retval = call {<vscale x 64 x i1>, <vscale x 64 x i1>} @llvm.experimental.vector.deinterleave2.nxv128i1(<vscale x 128 x i1> %vec)
 ret {<vscale x 64 x i1>, <vscale x 64 x i1>} %retval

From b4df0da9e8421c8026bd09980b8a6acd7a6ce8c9 Mon Sep 17 00:00:00 2001
From: Shilei Tian <i@tianshilei.me>
Date: Fri, 12 Apr 2024 09:34:46 -0400
Subject: [PATCH 246/886] [AMDGPU] Fix a potential wrong return value
 indicating whether a pass modifies a function (#88197)

When the alloca is too big for vectorization, the function could have
already
been modified in previous iteration of the `for` loop.
---
 llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp    |  2 +-
 llvm/test/CodeGen/AMDGPU/half-alloca-promotion.ll | 11 +++++++++++
 2 files changed, 12 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/half-alloca-promotion.ll

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
index 6f3cdf54dceec..c0846b123d187 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
@@ -336,7 +336,7 @@ bool AMDGPUPromoteAllocaImpl::run(Function &F, bool PromoteToLDS) {
     if (AllocaCost > VectorizationBudget) {
       LLVM_DEBUG(dbgs() << "  Alloca too big for vectorization: " << *AI
                         << "\n");
-      return false;
+      return Changed;
     }
 
     if (tryPromoteAllocaToVector(*AI)) {
diff --git a/llvm/test/CodeGen/AMDGPU/half-alloca-promotion.ll b/llvm/test/CodeGen/AMDGPU/half-alloca-promotion.ll
new file mode 100644
index 0000000000000..cfec49f3652fb
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/half-alloca-promotion.ll
@@ -0,0 +1,11 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes="amdgpu-promote-alloca-to-vector" -o - %s
+; We don't really need to check anything here because with expensive check, this
+; test case crashes. The correctness of the pass is beyond the scope.
+
+define fastcc void @foo() {
+entry:
+  %det = alloca [4 x i32], align 16, addrspace(5)
+  %trkltPosTmpYZ = alloca [2 x float], align 4, addrspace(5)
+  %trkltCovTmp = alloca [3 x float], align 4, addrspace(5)
+  ret void
+}

From 2a5ba4fb895931b7dc86f4304b0e64153a1175d4 Mon Sep 17 00:00:00 2001
From: Xiaoyang Liu <siujoeng.lau@gmail.com>
Date: Fri, 12 Apr 2024 10:04:21 -0400
Subject: [PATCH 247/886] [libc++] LWG3643: Missing `constexpr` in
 `std::counted_iterator` (#87901)

This pull request implements LWG3643: Missing constexpr in
std::counted_iterator. Specifically, one overload of
std::counted_operator::operator++ was not marked as constexpr,
despite being eligible for it after the introduction of try-block
support in constexpr functions in C++20.
---
 libcxx/docs/Status/Cxx23Issues.csv            |  2 +-
 libcxx/include/__iterator/counted_iterator.h  |  2 +-
 .../counted.iterator/increment.pass.cpp       | 45 +++++++++----------
 3 files changed, 24 insertions(+), 25 deletions(-)

diff --git a/libcxx/docs/Status/Cxx23Issues.csv b/libcxx/docs/Status/Cxx23Issues.csv
index a212d56685c00..fa82e67ac7db7 100644
--- a/libcxx/docs/Status/Cxx23Issues.csv
+++ b/libcxx/docs/Status/Cxx23Issues.csv
@@ -150,7 +150,7 @@
 "`3619 <https://wg21.link/LWG3619>`__","Specification of ``vformat_to`` contains ill-formed ``formatted_size`` calls","February 2022","|Nothing to do|","","|format|"
 "`3621 <https://wg21.link/LWG3621>`__","Remove feature-test macro ``__cpp_lib_monadic_optional`` ","February 2022","|Complete|","15.0"
 "`3632 <https://wg21.link/LWG3632>`__","``unique_ptr`` ""Mandates: This constructor is not selected by class template argument deduction""","February 2022","|Nothing to do|",""
-"`3643 <https://wg21.link/LWG3643>`__","Missing ``constexpr`` in ``std::counted_iterator`` ","February 2022","","","|ranges|"
+"`3643 <https://wg21.link/LWG3643>`__","Missing ``constexpr`` in ``std::counted_iterator`` ","February 2022","|Complete|","19.0","|ranges|"
 "`3648 <https://wg21.link/LWG3648>`__","``format`` should not print ``bool`` with ``'c'`` ","February 2022","|Complete|","15.0","|format|"
 "`3649 <https://wg21.link/LWG3649>`__","[fund.ts.v2] Reinstate and bump ``__cpp_lib_experimental_memory_resource`` feature test macro","February 2022","",""
 "`3650 <https://wg21.link/LWG3650>`__","Are ``std::basic_string`` 's ``iterator`` and ``const_iterator`` constexpr iterators?","February 2022","|Nothing to do|",""
diff --git a/libcxx/include/__iterator/counted_iterator.h b/libcxx/include/__iterator/counted_iterator.h
index 008c52fa87ce0..ea2832e3b978d 100644
--- a/libcxx/include/__iterator/counted_iterator.h
+++ b/libcxx/include/__iterator/counted_iterator.h
@@ -129,7 +129,7 @@ class counted_iterator
     return *this;
   }
 
-  _LIBCPP_HIDE_FROM_ABI decltype(auto) operator++(int) {
+  _LIBCPP_HIDE_FROM_ABI constexpr decltype(auto) operator++(int) {
     _LIBCPP_ASSERT_UNCATEGORIZED(__count_ > 0, "Iterator already at or past end.");
     --__count_;
 #  ifndef _LIBCPP_HAS_NO_EXCEPTIONS
diff --git a/libcxx/test/std/iterators/predef.iterators/counted.iterator/increment.pass.cpp b/libcxx/test/std/iterators/predef.iterators/counted.iterator/increment.pass.cpp
index 66d90a4187448..2fea2561ebf7f 100644
--- a/libcxx/test/std/iterators/predef.iterators/counted.iterator/increment.pass.cpp
+++ b/libcxx/test/std/iterators/predef.iterators/counted.iterator/increment.pass.cpp
@@ -9,7 +9,7 @@
 // UNSUPPORTED: c++03, c++11, c++14, c++17
 
 // constexpr counted_iterator& operator++();
-// decltype(auto) operator++(int);
+// constexpr decltype(auto) operator++(int);
 // constexpr counted_iterator operator++(int)
 //   requires forward_iterator<I>;
 
@@ -62,6 +62,26 @@ concept PlusEnabled = requires(Iter& iter) {
 constexpr bool test() {
   int buffer[8] = {1, 2, 3, 4, 5, 6, 7, 8};
 
+  {
+    using Counted = std::counted_iterator<InputOrOutputArchetype>;
+    std::counted_iterator iter(InputOrOutputArchetype{buffer}, 8);
+
+    iter++;
+    assert((++iter).base().ptr == buffer + 2);
+
+    ASSERT_SAME_TYPE(decltype(iter++), void);
+    ASSERT_SAME_TYPE(decltype(++iter), Counted&);
+  }
+  {
+    using Counted = std::counted_iterator<cpp20_input_iterator<int*>>;
+    std::counted_iterator iter(cpp20_input_iterator<int*>{buffer}, 8);
+
+    iter++;
+    assert(++iter == Counted(cpp20_input_iterator<int*>{buffer + 2}, 6));
+
+    ASSERT_SAME_TYPE(decltype(iter++), void);
+    ASSERT_SAME_TYPE(decltype(++iter), Counted&);
+  }
   {
     using Counted = std::counted_iterator<forward_iterator<int*>>;
     std::counted_iterator iter(forward_iterator<int*>{buffer}, 8);
@@ -95,29 +115,8 @@ int main(int, char**) {
   test();
   static_assert(test());
 
-  int buffer[8] = {1, 2, 3, 4, 5, 6, 7, 8};
-
-  {
-    using Counted = std::counted_iterator<InputOrOutputArchetype>;
-    std::counted_iterator iter(InputOrOutputArchetype{buffer}, 8);
-
-    iter++;
-    assert((++iter).base().ptr == buffer + 2);
-
-    ASSERT_SAME_TYPE(decltype(iter++), void);
-    ASSERT_SAME_TYPE(decltype(++iter), Counted&);
-  }
-  {
-    using Counted = std::counted_iterator<cpp20_input_iterator<int*>>;
-    std::counted_iterator iter(cpp20_input_iterator<int*>{buffer}, 8);
-
-    iter++;
-    assert(++iter == Counted(cpp20_input_iterator<int*>{buffer + 2}, 6));
-
-    ASSERT_SAME_TYPE(decltype(iter++), void);
-    ASSERT_SAME_TYPE(decltype(++iter), Counted&);
-  }
 #ifndef TEST_HAS_NO_EXCEPTIONS
+  int buffer[8] = {1, 2, 3, 4, 5, 6, 7, 8};
   {
     using Counted = std::counted_iterator<ThrowsOnInc<int*>>;
     std::counted_iterator iter(ThrowsOnInc<int*>{buffer}, 8);

From 5fa58e28f34f4b5b41dd2869212ac917fb42b2e3 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 06:55:47 -0700
Subject: [PATCH 248/886] [RISCV] Split PseudoVFADD, PseudoVFSUB, and
 PseudoVFRSUB by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 26 ++++++++++---------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |  6 ++---
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  6 ++---
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    | 16 +++++++++---
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 16 +++++++++---
 llvm/lib/Target/RISCV/RISCVScheduleV.td       | 16 ++++++------
 6 files changed, 52 insertions(+), 34 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index cf9a31c23a06e..ebe8a29fdfa85 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2917,16 +2917,17 @@ multiclass VPseudoVMAX_VV_VF {
 
 multiclass VPseudoVALU_VV_VF_RM {
   foreach m = MxListF in {
-    defm "" : VPseudoBinaryFV_VV_RM<m>,
-              SchedBinary<"WriteVFALUV", "ReadVFALUV", "ReadVFALUV", m.MX,
-                          forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm "" : VPseudoBinaryFV_VV_RM<m, "", sew=e>,
+                SchedBinary<"WriteVFALUV", "ReadVFALUV", "ReadVFALUV", m.MX, e,
+                            forceMergeOpRead=true>;
   }
 
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoBinaryV_VF_RM<m, f>,
+      defm "" : VPseudoBinaryV_VF_RM<m, f, "", sew=f.SEW>,
                 SchedBinary<"WriteVFALUF", "ReadVFALUV", "ReadVFALUF", m.MX,
-                            forceMergeOpRead=true>;
+                            f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -2934,9 +2935,9 @@ multiclass VPseudoVALU_VV_VF_RM {
 multiclass VPseudoVALU_VF_RM {
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoBinaryV_VF_RM<m, f>,
+      defm "" : VPseudoBinaryV_VF_RM<m, f, "", sew=f.SEW>,
                 SchedBinary<"WriteVFALUF", "ReadVFALUV", "ReadVFALUF", m.MX,
-                            forceMergeOpRead=true>;
+                            f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -7061,11 +7062,12 @@ defm : VPatBinaryV_WV_WX_WI_RM<"int_riscv_vnclip", "PseudoVNCLIP",
 //===----------------------------------------------------------------------===//
 // 13.2. Vector Single-Width Floating-Point Add/Subtract Instructions
 //===----------------------------------------------------------------------===//
-defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfadd", "PseudoVFADD",
-                            AllFloatVectors>;
-defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfsub", "PseudoVFSUB",
-                            AllFloatVectors>;
-defm : VPatBinaryV_VX_RM<"int_riscv_vfrsub", "PseudoVFRSUB", AllFloatVectors>;
+defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfadd", "PseudoVFADD", AllFloatVectors,
+                            isSEWAware = 1>;
+defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfsub", "PseudoVFSUB", AllFloatVectors,
+                            isSEWAware = 1>;
+defm : VPatBinaryV_VX_RM<"int_riscv_vfrsub", "PseudoVFRSUB", AllFloatVectors,
+                         isSEWAware = 1>;
 
 //===----------------------------------------------------------------------===//
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index da761ae856706..0d5cd93984af4 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -1204,9 +1204,9 @@ foreach mti = AllMasks in {
 // 13. Vector Floating-Point Instructions
 
 // 13.2. Vector Single-Width Floating-Point Add/Subtract Instructions
-defm : VPatBinaryFPSDNode_VV_VF_RM<any_fadd, "PseudoVFADD">;
-defm : VPatBinaryFPSDNode_VV_VF_RM<any_fsub, "PseudoVFSUB">;
-defm : VPatBinaryFPSDNode_R_VF_RM<any_fsub, "PseudoVFRSUB">;
+defm : VPatBinaryFPSDNode_VV_VF_RM<any_fadd, "PseudoVFADD", isSEWAware=1>;
+defm : VPatBinaryFPSDNode_VV_VF_RM<any_fsub, "PseudoVFSUB", isSEWAware=1>;
+defm : VPatBinaryFPSDNode_R_VF_RM<any_fsub, "PseudoVFRSUB", isSEWAware=1>;
 
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
 defm : VPatWidenBinaryFPSDNode_VV_VF_WV_WF_RM<fadd, "PseudoVFWADD">;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index b721dcd989885..befe8d07e5750 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -2425,9 +2425,9 @@ foreach vtiToWti = AllWidenableIntVectors in
 // 13. Vector Floating-Point Instructions
 
 // 13.2. Vector Single-Width Floating-Point Add/Subtract Instructions
-defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fadd_vl, "PseudoVFADD">;
-defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fsub_vl, "PseudoVFSUB">;
-defm : VPatBinaryFPVL_R_VF_RM<any_riscv_fsub_vl, "PseudoVFRSUB">;
+defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fadd_vl, "PseudoVFADD", isSEWAware=1>;
+defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fsub_vl, "PseudoVFSUB", isSEWAware=1>;
+defm : VPatBinaryFPVL_R_VF_RM<any_riscv_fsub_vl, "PseudoVFRSUB", isSEWAware=1>;
 
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
 defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwadd_vl, riscv_vfwadd_w_vl, "PseudoVFWADD">;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 54a13889ee698..f14766c984391 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -732,12 +732,20 @@ foreach mx = SchedMxListW in {
 }
 
 // 13. Vector Floating-Point Instructions
+foreach mx = SchedMxListF in {
+  foreach sew = SchedSEWSet<mx, isF=1>.val in {
+    defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
+    defvar IsWorstCase = SiFive7IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
+    let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxList in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFALUV",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFALUF",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulV",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulF",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddV",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
@@ -1137,8 +1145,8 @@ defm "" : LMULReadAdvanceW<"ReadVNClipV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 
 // 14. Vector Floating-Point Instructions
-defm "" : LMULReadAdvance<"ReadVFALUV", 0>;
-defm "" : LMULReadAdvance<"ReadVFALUF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 54016959d348e..4d81bafdf88b8 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -486,12 +486,20 @@ foreach mx = SchedMxList in {
 }
 
 // 13. Vector Floating-Point Instructions
+foreach mx = SchedMxListF in {
+  foreach sew = SchedSEWSet<mx, isF=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
+    let Latency = 6, ReleaseAtCycles = [LMulLat] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFALUV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFALUF",    [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulF",    [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddV", [SiFiveP600VectorArith], mx, IsWorstCase>;
@@ -925,8 +933,8 @@ defm "" : LMULReadAdvanceW<"ReadVNClipV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 
 // 14. Vector Floating-Point Instructions
-defm "" : LMULReadAdvance<"ReadVFALUV", 0>;
-defm "" : LMULReadAdvance<"ReadVFALUF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 379622d4ca834..9b9c72e0767af 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -397,8 +397,8 @@ defm "" : LMULSchedWritesW<"WriteVNClipI">;
 
 // 13. Vector Floating-Point Instructions
 // 13.2. Vector Single-Width Floating-Point Add/Subtract Instructions
-defm "" : LMULSchedWrites<"WriteVFALUV">;
-defm "" : LMULSchedWrites<"WriteVFALUF">;
+defm "" : LMULSEWSchedWritesF<"WriteVFALUV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFALUF">;
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
 defm "" : LMULSchedWritesFW<"WriteVFWALUV">;
 defm "" : LMULSchedWritesFW<"WriteVFWALUF">;
@@ -622,8 +622,8 @@ defm "" : LMULSchedReadsW<"ReadVNClipX">;
 
 // 13. Vector Floating-Point Instructions
 // 13.2. Vector Single-Width Floating-Point Add/Subtract Instructions
-defm "" : LMULSchedReads<"ReadVFALUV">;
-defm "" : LMULSchedReads<"ReadVFALUF">;
+defm "" : LMULSEWSchedReadsF<"ReadVFALUV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFALUF">;
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
 defm "" : LMULSchedReadsFW<"ReadVFWALUV">;
 defm "" : LMULSchedReadsFW<"ReadVFWALUF">;
@@ -868,8 +868,8 @@ defm "" : LMULWriteResW<"WriteVNClipX", []>;
 defm "" : LMULWriteResW<"WriteVNClipI", []>;
 
 // 13. Vector Floating-Point Instructions
-defm "" : LMULWriteRes<"WriteVFALUV", []>;
-defm "" : LMULWriteRes<"WriteVFALUF", []>;
+defm "" : LMULSEWWriteResF<"WriteVFALUV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFALUF", []>;
 defm "" : LMULWriteResFW<"WriteVFWALUV", []>;
 defm "" : LMULWriteResFW<"WriteVFWALUF", []>;
 defm "" : LMULWriteRes<"WriteVFMulV", []>;
@@ -1024,8 +1024,8 @@ defm "" : LMULReadAdvanceW<"ReadVNClipV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 
 // 13. Vector Floating-Point Instructions
-defm "" : LMULReadAdvance<"ReadVFALUV", 0>;
-defm "" : LMULReadAdvance<"ReadVFALUF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;

From aece68269c8cb28e16dc5c4ec9aabecb807a0438 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 08:38:32 -0700
Subject: [PATCH 249/886] [RISCV] Split PseudoVFWADD, PseudoVFWSUB, and
 PseudoVFWMUL by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      |  46 ++++++-
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 121 ++++++++++--------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |  12 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  43 ++++---
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |  24 ++--
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td |  24 ++--
 llvm/lib/Target/RISCV/RISCVScheduleV.td       |  44 ++++---
 7 files changed, 203 insertions(+), 111 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index d78f5bd9dedf3..3b76e7158ffd8 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -3038,14 +3038,13 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
 #define CASE_WIDEOP_OPCODE_LMULS(OP)                                           \
   CASE_WIDEOP_OPCODE_COMMON(OP, MF8):                                          \
   case CASE_WIDEOP_OPCODE_LMULS_MF4(OP)
-// clang-format on
 
 #define CASE_WIDEOP_CHANGE_OPCODE_COMMON(OP, LMUL)                             \
   case RISCV::PseudoV##OP##_##LMUL##_TIED:                                     \
     NewOpc = RISCV::PseudoV##OP##_##LMUL;                                      \
     break;
 
-#define CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4(OP)                                 \
+#define CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4(OP)                                \
   CASE_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF4)                                    \
   CASE_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF2)                                    \
   CASE_WIDEOP_CHANGE_OPCODE_COMMON(OP, M1)                                     \
@@ -3056,6 +3055,41 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
   CASE_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF8)                                    \
   CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4(OP)
 
+// FP Widening Ops may by SEW aware. Create SEW aware cases for these cases.
+#define CASE_FP_WIDEOP_OPCODE_COMMON(OP, LMUL, SEW)                            \
+  RISCV::PseudoV##OP##_##LMUL##_##SEW##_TIED
+
+#define CASE_FP_WIDEOP_OPCODE_LMULS_MF4(OP)                                    \
+  CASE_FP_WIDEOP_OPCODE_COMMON(OP, MF4, E16):                                  \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, MF2, E16):                             \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, MF2, E32):                             \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M1, E16):                              \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M1, E32):                              \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M2, E16):                              \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M2, E32):                              \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M4, E16):                              \
+  case CASE_FP_WIDEOP_OPCODE_COMMON(OP, M4, E32)                               \
+
+#define CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, LMUL, SEW)                     \
+  case RISCV::PseudoV##OP##_##LMUL##_##SEW##_TIED:                             \
+    NewOpc = RISCV::PseudoV##OP##_##LMUL##_##SEW;                              \
+    break;
+
+#define CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS_MF4(OP)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF4, E16)                            \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF2, E16)                            \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, MF2, E32)                            \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M1, E16)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M1, E32)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M2, E16)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M2, E32)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M4, E16)                             \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON(OP, M4, E32)                             \
+
+#define CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS(OP)                                 \
+  CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS_MF4(OP)
+// clang-format on
+
 MachineInstr *RISCVInstrInfo::convertToThreeAddress(MachineInstr &MI,
                                                     LiveVariables *LV,
                                                     LiveIntervals *LIS) const {
@@ -3063,8 +3097,8 @@ MachineInstr *RISCVInstrInfo::convertToThreeAddress(MachineInstr &MI,
   switch (MI.getOpcode()) {
   default:
     return nullptr;
-  case CASE_WIDEOP_OPCODE_LMULS_MF4(FWADD_WV):
-  case CASE_WIDEOP_OPCODE_LMULS_MF4(FWSUB_WV): {
+  case CASE_FP_WIDEOP_OPCODE_LMULS_MF4(FWADD_WV):
+  case CASE_FP_WIDEOP_OPCODE_LMULS_MF4(FWSUB_WV): {
     assert(RISCVII::hasVecPolicyOp(MI.getDesc().TSFlags) &&
            MI.getNumExplicitOperands() == 7 &&
            "Expect 7 explicit operands rd, rs2, rs1, rm, vl, sew, policy");
@@ -3077,8 +3111,8 @@ MachineInstr *RISCVInstrInfo::convertToThreeAddress(MachineInstr &MI,
     switch (MI.getOpcode()) {
     default:
       llvm_unreachable("Unexpected opcode");
-    CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4(FWADD_WV)
-    CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4(FWSUB_WV)
+    CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS_MF4(FWADD_WV)
+    CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS_MF4(FWSUB_WV)
     }
     // clang-format on
 
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index ebe8a29fdfa85..1097eb4fb6431 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2213,11 +2213,13 @@ multiclass VPseudoTiedBinaryRoundingMode<VReg RetClass,
                                          DAGOperand Op2Class,
                                          LMULInfo MInfo,
                                          string Constraint = "",
+                                         int sew = 0,
                                          int TargetConstraintType = 1> {
+    defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
     let VLMul = MInfo.value in {
-    def "_" # MInfo.MX # "_TIED":
+    def suffix # "_TIED":
       VPseudoTiedBinaryNoMaskRoundingMode<RetClass, Op2Class, Constraint, TargetConstraintType>;
-    def "_" # MInfo.MX # "_MASK_TIED" :
+    def suffix # "_MASK_TIED" :
       VPseudoTiedBinaryMaskRoundingMode<RetClass, Op2Class, Constraint, TargetConstraintType>,
       RISCVMaskedPseudo<MaskIdx=2>;
   }
@@ -2334,9 +2336,9 @@ multiclass VPseudoBinaryW_VV<LMULInfo m> {
                            "@earlyclobber $rd", TargetConstraintType=3>;
 }
 
-multiclass VPseudoBinaryW_VV_RM<LMULInfo m> {
+multiclass VPseudoBinaryW_VV_RM<LMULInfo m, int sew = 0> {
   defm _VV : VPseudoBinaryRoundingMode<m.wvrclass, m.vrclass, m.vrclass, m,
-                                      "@earlyclobber $rd",  UsesVXRM=0,
+                                      "@earlyclobber $rd", sew, UsesVXRM=0,
                                       TargetConstraintType=3>;
 }
 
@@ -2356,10 +2358,10 @@ multiclass VPseudoBinaryW_VF<LMULInfo m, FPR_Info f> {
                                    "@earlyclobber $rd">;
 }
 
-multiclass VPseudoBinaryW_VF_RM<LMULInfo m, FPR_Info f> {
+multiclass VPseudoBinaryW_VF_RM<LMULInfo m, FPR_Info f, int sew = 0> {
   defm "_V" # f.FX : VPseudoBinaryRoundingMode<m.wvrclass, m.vrclass,
                                                f.fprclass, m,
-                                               "@earlyclobber $rd",
+                                               "@earlyclobber $rd", sew,
                                                UsesVXRM=0,
                                                TargetConstraintType=3>;
 }
@@ -2371,11 +2373,13 @@ multiclass VPseudoBinaryW_WV<LMULInfo m> {
                                "@earlyclobber $rd", TargetConstraintType=3>;
 }
 
-multiclass VPseudoBinaryW_WV_RM<LMULInfo m> {
+multiclass VPseudoBinaryW_WV_RM<LMULInfo m, int sew = 0> {
   defm _WV : VPseudoBinaryRoundingMode<m.wvrclass, m.wvrclass, m.vrclass, m,
-                                       "@earlyclobber $rd", UsesVXRM=0, TargetConstraintType=3>;
+                                       "@earlyclobber $rd", sew, UsesVXRM = 0,
+                                       TargetConstraintType = 3>;
   defm _WV : VPseudoTiedBinaryRoundingMode<m.wvrclass, m.vrclass, m,
-                                           "@earlyclobber $rd", TargetConstraintType=3>;
+                                           "@earlyclobber $rd", sew,
+                                           TargetConstraintType = 3>;
 }
 
 multiclass VPseudoBinaryW_WX<LMULInfo m> {
@@ -2387,11 +2391,11 @@ multiclass VPseudoBinaryW_WF<LMULInfo m, FPR_Info f, int TargetConstraintType =
                                    f.fprclass, m, /*Constraint*/ "", TargetConstraintType=TargetConstraintType>;
 }
 
-multiclass VPseudoBinaryW_WF_RM<LMULInfo m, FPR_Info f> {
+multiclass VPseudoBinaryW_WF_RM<LMULInfo m, FPR_Info f, int sew = 0> {
   defm "_W" # f.FX : VPseudoBinaryRoundingMode<m.wvrclass, m.wvrclass,
                                                f.fprclass, m,
                                                Constraint="",
-                                               sew=0,
+                                               sew=sew,
                                                UsesVXRM=0,
                                                TargetConstraintType=3>;
 }
@@ -2979,16 +2983,17 @@ multiclass VPseudoVWMUL_VV_VX {
 
 multiclass VPseudoVWMUL_VV_VF_RM {
   foreach m = MxListFW in {
-    defm "" : VPseudoBinaryW_VV_RM<m>,
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+    defm "" : VPseudoBinaryW_VV_RM<m, sew=e>,
               SchedBinary<"WriteVFWMulV", "ReadVFWMulV", "ReadVFWMulV", m.MX,
-                          forceMergeOpRead=true>;
+                          e, forceMergeOpRead=true>;
   }
 
   foreach f = FPListW in {
     foreach m = f.MxListFW in {
-      defm "" : VPseudoBinaryW_VF_RM<m, f>,
+      defm "" : VPseudoBinaryW_VF_RM<m, f, sew=f.SEW>,
                 SchedBinary<"WriteVFWMulF", "ReadVFWMulV", "ReadVFWMulF", m.MX,
-                          forceMergeOpRead=true>;
+                          f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -3007,31 +3012,33 @@ multiclass VPseudoVWALU_WV_WX {
 
 multiclass VPseudoVFWALU_VV_VF_RM {
   foreach m = MxListFW in {
-    defm "" : VPseudoBinaryW_VV_RM<m>,
-              SchedBinary<"WriteVFWALUV", "ReadVFWALUV", "ReadVFWALUV", m.MX,
-                          forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm "" : VPseudoBinaryW_VV_RM<m, sew=e>,
+                SchedBinary<"WriteVFWALUV", "ReadVFWALUV", "ReadVFWALUV", m.MX,
+                            e, forceMergeOpRead=true>;
   }
 
   foreach f = FPListW in {
     foreach m = f.MxListFW in {
-      defm "" : VPseudoBinaryW_VF_RM<m, f>,
+      defm "" : VPseudoBinaryW_VF_RM<m, f, sew=f.SEW>,
                 SchedBinary<"WriteVFWALUF", "ReadVFWALUV", "ReadVFWALUF", m.MX,
-                          forceMergeOpRead=true>;
+                          f.SEW, forceMergeOpRead=true>;
     }
   }
 }
 
 multiclass VPseudoVFWALU_WV_WF_RM {
   foreach m = MxListFW in {
-    defm "" : VPseudoBinaryW_WV_RM<m>,
-              SchedBinary<"WriteVFWALUV", "ReadVFWALUV", "ReadVFWALUV", m.MX,
-                          forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm "" : VPseudoBinaryW_WV_RM<m, sew=e>,
+                SchedBinary<"WriteVFWALUV", "ReadVFWALUV", "ReadVFWALUV", m.MX,
+                            e, forceMergeOpRead=true>;
   }
   foreach f = FPListW in {
     foreach m = f.MxListFW in {
-      defm "" : VPseudoBinaryW_WF_RM<m, f>,
+      defm "" : VPseudoBinaryW_WF_RM<m, f, sew=f.SEW>,
                 SchedBinary<"WriteVFWALUF", "ReadVFWALUV", "ReadVFWALUF", m.MX,
-                          forceMergeOpRead=true>;
+                            f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -5028,13 +5035,16 @@ multiclass VPatBinaryW_VV<string intrinsic, string instruction,
 }
 
 multiclass VPatBinaryW_VV_RM<string intrinsic, string instruction,
-                             list<VTypeInfoToWide> vtilist> {
+                             list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
+    defvar name = !if(isSEWAware,
+                      instruction # "_VV_" # Vti.LMul.MX # "_E" # Vti.SEW,
+                      instruction # "_VV_" # Vti.LMul.MX);
     let Predicates = !listconcat(GetVTypePredicates<Vti>.Predicates,
                                  GetVTypePredicates<Wti>.Predicates) in
-    defm : VPatBinaryRoundingMode<intrinsic, instruction # "_VV_" # Vti.LMul.MX,
+    defm : VPatBinaryRoundingMode<intrinsic, name,
                                   Wti.Vector, Vti.Vector, Vti.Vector, Vti.Mask,
                                   Vti.Log2SEW, Wti.RegClass,
                                   Vti.RegClass, Vti.RegClass>;
@@ -5057,14 +5067,17 @@ multiclass VPatBinaryW_VX<string intrinsic, string instruction,
 }
 
 multiclass VPatBinaryW_VX_RM<string intrinsic, string instruction,
-                          list<VTypeInfoToWide> vtilist> {
+                          list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
     defvar kind = "V"#Vti.ScalarSuffix;
+    defvar name = !if(isSEWAware,
+                      instruction#"_"#kind#"_"#Vti.LMul.MX # "_E" # Vti.SEW,
+                      instruction#"_"#kind#"_"#Vti.LMul.MX);
     let Predicates = !listconcat(GetVTypePredicates<Vti>.Predicates,
                                  GetVTypePredicates<Wti>.Predicates) in
-    defm : VPatBinaryRoundingMode<intrinsic, instruction#"_"#kind#"_"#Vti.LMul.MX,
+    defm : VPatBinaryRoundingMode<intrinsic, name,
                                   Wti.Vector, Vti.Vector, Vti.Scalar, Vti.Mask,
                                   Vti.Log2SEW, Wti.RegClass,
                                   Vti.RegClass, Vti.ScalarRegClass>;
@@ -5101,27 +5114,30 @@ multiclass VPatBinaryW_WV<string intrinsic, string instruction,
 }
 
 multiclass VPatBinaryW_WV_RM<string intrinsic, string instruction,
-                             list<VTypeInfoToWide> vtilist> {
+                             list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
+    defvar name = !if(isSEWAware,
+                      instruction # "_WV_" # Vti.LMul.MX # "_E" # Vti.SEW,
+                      instruction # "_WV_" # Vti.LMul.MX);
     let Predicates = !listconcat(GetVTypePredicates<Vti>.Predicates,
                                  GetVTypePredicates<Wti>.Predicates) in {
-      def : VPatTiedBinaryNoMaskRoundingMode<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
+      def : VPatTiedBinaryNoMaskRoundingMode<intrinsic, name,
                                              Wti.Vector, Vti.Vector,
                                              Vti.Log2SEW, Wti.RegClass, Vti.RegClass>;
-      def : VPatBinaryNoMaskTURoundingMode<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
+      def : VPatBinaryNoMaskTURoundingMode<intrinsic, name,
                                            Wti.Vector, Wti.Vector, Vti.Vector, Vti.Log2SEW,
                                            Wti.RegClass, Wti.RegClass, Vti.RegClass>;
       let AddedComplexity = 1 in {
-      def : VPatTiedBinaryNoMaskTURoundingMode<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
+      def : VPatTiedBinaryNoMaskTURoundingMode<intrinsic, name,
                                                Wti.Vector, Vti.Vector,
                                                Vti.Log2SEW, Wti.RegClass, Vti.RegClass>;
-      def : VPatTiedBinaryMaskRoundingMode<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
+      def : VPatTiedBinaryMaskRoundingMode<intrinsic, name,
                                            Wti.Vector, Vti.Vector, Vti.Mask,
                                            Vti.Log2SEW, Wti.RegClass, Vti.RegClass>;
       }
-      def : VPatBinaryMaskTARoundingMode<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
+      def : VPatBinaryMaskTARoundingMode<intrinsic, name,
                                          Wti.Vector, Wti.Vector, Vti.Vector, Vti.Mask,
                                          Vti.Log2SEW, Wti.RegClass,
                                          Wti.RegClass, Vti.RegClass>;
@@ -5145,14 +5161,17 @@ multiclass VPatBinaryW_WX<string intrinsic, string instruction,
 }
 
 multiclass VPatBinaryW_WX_RM<string intrinsic, string instruction,
-                             list<VTypeInfoToWide> vtilist> {
+                             list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
     defvar kind = "W"#Vti.ScalarSuffix;
+    defvar name = !if(isSEWAware,
+                      instruction#"_"#kind#"_"#Vti.LMul.MX#"_E"#Vti.SEW,
+                      instruction#"_"#kind#"_"#Vti.LMul.MX);
     let Predicates = !listconcat(GetVTypePredicates<Vti>.Predicates,
                                  GetVTypePredicates<Wti>.Predicates) in
-    defm : VPatBinaryRoundingMode<intrinsic, instruction#"_"#kind#"_"#Vti.LMul.MX,
+    defm : VPatBinaryRoundingMode<intrinsic, name,
                                   Wti.Vector, Wti.Vector, Vti.Scalar, Vti.Mask,
                                   Vti.Log2SEW, Wti.RegClass,
                                   Wti.RegClass, Vti.ScalarRegClass>;
@@ -5417,20 +5436,22 @@ multiclass VPatBinaryW_VV_VX<string intrinsic, string instruction,
     : VPatBinaryW_VV<intrinsic, instruction, vtilist>,
       VPatBinaryW_VX<intrinsic, instruction, vtilist>;
 
-multiclass VPatBinaryW_VV_VX_RM<string intrinsic, string instruction,
-                                list<VTypeInfoToWide> vtilist>
-    : VPatBinaryW_VV_RM<intrinsic, instruction, vtilist>,
-      VPatBinaryW_VX_RM<intrinsic, instruction, vtilist>;
+multiclass
+    VPatBinaryW_VV_VX_RM<string intrinsic, string instruction,
+                         list<VTypeInfoToWide> vtilist, bit isSEWAware = 0>
+    : VPatBinaryW_VV_RM<intrinsic, instruction, vtilist, isSEWAware>,
+      VPatBinaryW_VX_RM<intrinsic, instruction, vtilist, isSEWAware>;
 
 multiclass VPatBinaryW_WV_WX<string intrinsic, string instruction,
                              list<VTypeInfoToWide> vtilist>
     : VPatBinaryW_WV<intrinsic, instruction, vtilist>,
       VPatBinaryW_WX<intrinsic, instruction, vtilist>;
 
-multiclass VPatBinaryW_WV_WX_RM<string intrinsic, string instruction,
-                                list<VTypeInfoToWide> vtilist>
-    : VPatBinaryW_WV_RM<intrinsic, instruction, vtilist>,
-      VPatBinaryW_WX_RM<intrinsic, instruction, vtilist>;
+multiclass
+    VPatBinaryW_WV_WX_RM<string intrinsic, string instruction,
+                         list<VTypeInfoToWide> vtilist, bit isSEWAware = 0>
+    : VPatBinaryW_WV_RM<intrinsic, instruction, vtilist, isSEWAware>,
+      VPatBinaryW_WX_RM<intrinsic, instruction, vtilist, isSEWAware>;
 
 multiclass VPatBinaryV_WV_WX_WI<string intrinsic, string instruction,
                                 list<VTypeInfoToWide> vtilist>
@@ -7073,13 +7094,13 @@ defm : VPatBinaryV_VX_RM<"int_riscv_vfrsub", "PseudoVFRSUB", AllFloatVectors,
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
 //===----------------------------------------------------------------------===//
 defm : VPatBinaryW_VV_VX_RM<"int_riscv_vfwadd", "PseudoVFWADD",
-                            AllWidenableFloatVectors>;
+                            AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatBinaryW_VV_VX_RM<"int_riscv_vfwsub", "PseudoVFWSUB",
-                            AllWidenableFloatVectors>;
+                            AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatBinaryW_WV_WX_RM<"int_riscv_vfwadd_w", "PseudoVFWADD",
-                            AllWidenableFloatVectors>;
+                            AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatBinaryW_WV_WX_RM<"int_riscv_vfwsub_w", "PseudoVFWSUB",
-                            AllWidenableFloatVectors>;
+                            AllWidenableFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
@@ -7095,7 +7116,7 @@ defm : VPatBinaryV_VX_RM<"int_riscv_vfrdiv", "PseudoVFRDIV",
 // 13.5. Vector Widening Floating-Point Multiply
 //===----------------------------------------------------------------------===//
 defm : VPatBinaryW_VV_VX_RM<"int_riscv_vfwmul", "PseudoVFWMUL",
-                            AllWidenableFloatVectors>;
+                            AllWidenableFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 0d5cd93984af4..37acd9667a546 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -621,7 +621,7 @@ multiclass VPatWidenBinaryFPSDNode_VV_VF_RM<SDNode op, string instruction_name>
                     (wti.Vector (riscv_fpextend_vl_oneuse
                                      (vti.Vector vti.RegClass:$rs1),
                                      (vti.Mask true_mask), (XLenVT srcvalue)))),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX#"_E"#vti.SEW)
                   (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2,
                   vti.RegClass:$rs1,
                    // Value to indicate no rounding mode change in
@@ -634,7 +634,7 @@ multiclass VPatWidenBinaryFPSDNode_VV_VF_RM<SDNode op, string instruction_name>
                     (wti.Vector (riscv_fpextend_vl_oneuse
                                      (vti.Vector (SplatFPOp (vti.Scalar vti.ScalarRegClass:$rs1))),
                                      (vti.Mask true_mask), (XLenVT srcvalue)))),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2,
                    vti.ScalarRegClass:$rs1,
                    // Value to indicate no rounding mode change in
@@ -645,7 +645,7 @@ multiclass VPatWidenBinaryFPSDNode_VV_VF_RM<SDNode op, string instruction_name>
                                      (vti.Vector vti.RegClass:$rs2),
                                      (vti.Mask true_mask), (XLenVT srcvalue))),
                     (wti.Vector (SplatFPOp (fpext_oneuse (vti.Scalar vti.ScalarRegClass:$rs1))))),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (wti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs2,
                    vti.ScalarRegClass:$rs1,
                    // Value to indicate no rounding mode change in
@@ -666,7 +666,7 @@ multiclass VPatWidenBinaryFPSDNode_WV_WF_RM<SDNode op, string instruction_name>
                     (wti.Vector (riscv_fpextend_vl_oneuse
                                      (vti.Vector vti.RegClass:$rs1),
                                      (vti.Mask true_mask), (XLenVT srcvalue)))),
-                (!cast<Instruction>(instruction_name#"_WV_"#vti.LMul.MX#"_TIED")
+                (!cast<Instruction>(instruction_name#"_WV_"#vti.LMul.MX#"_E"#vti.SEW#"_TIED")
                    wti.RegClass:$rs2, vti.RegClass:$rs1,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -677,7 +677,7 @@ multiclass VPatWidenBinaryFPSDNode_WV_WF_RM<SDNode op, string instruction_name>
                     (wti.Vector (riscv_fpextend_vl_oneuse
                                      (vti.Vector (SplatFPOp vti.ScalarRegClass:$rs1)),
                                      (vti.Mask true_mask), (XLenVT srcvalue)))),
-                (!cast<Instruction>(instruction_name#"_W"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_W"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (wti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs2,
                    vti.ScalarRegClass:$rs1,
                    // Value to indicate no rounding mode change in
@@ -686,7 +686,7 @@ multiclass VPatWidenBinaryFPSDNode_WV_WF_RM<SDNode op, string instruction_name>
                    vti.AVL, vti.Log2SEW, TA_MA)>;
       def : Pat<(op (wti.Vector wti.RegClass:$rs2),
                     (wti.Vector (SplatFPOp (fpext_oneuse (vti.Scalar vti.ScalarRegClass:$rs1))))),
-                (!cast<Instruction>(instruction_name#"_W"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_W"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (wti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs2,
                    vti.ScalarRegClass:$rs1,
                    // Value to indicate no rounding mode change in
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index befe8d07e5750..d2f575459e320 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -725,23 +725,27 @@ multiclass VPatTiedBinaryNoMaskVL_V_RM<SDNode vop,
                                        string suffix,
                                        ValueType result_type,
                                        ValueType op2_type,
-                                       int sew,
+                                       int log2sew,
                                        LMULInfo vlmul,
                                        VReg result_reg_class,
-                                       VReg op2_reg_class> {
+                                       VReg op2_reg_class,
+                                       bit isSEWAware = 0> {
+  defvar name = !if(isSEWAware,
+                    instruction_name#"_"#suffix#"_"#vlmul.MX#"_E"#!shl(1, log2sew)#"_TIED",
+                    instruction_name#"_"#suffix#"_"#vlmul.MX#"_TIED");
   def : Pat<(result_type (vop
                          (result_type result_reg_class:$rs1),
                          (op2_type op2_reg_class:$rs2),
                          srcvalue,
                          true_mask,
                          VLOpFrag)),
-        (!cast<Instruction>(instruction_name#"_"#suffix#"_"# vlmul.MX#"_TIED")
+        (!cast<Instruction>(name)
                      result_reg_class:$rs1,
                      op2_reg_class:$rs2,
                      // Value to indicate no rounding mode change in
                      // RISCVInsertReadWriteCSR
                      FRM_DYN,
-                     GPR:$vl, sew, TAIL_AGNOSTIC)>;
+                     GPR:$vl, log2sew, TAIL_AGNOSTIC)>;
   // Tail undisturbed
   def : Pat<(riscv_vmerge_vl true_mask,
              (result_type (vop
@@ -751,13 +755,13 @@ multiclass VPatTiedBinaryNoMaskVL_V_RM<SDNode vop,
                            true_mask,
                            VLOpFrag)),
              result_reg_class:$rs1, result_reg_class:$rs1, VLOpFrag),
-            (!cast<Instruction>(instruction_name#"_"#suffix#"_"# vlmul.MX#"_TIED")
+            (!cast<Instruction>(name)
                      result_reg_class:$rs1,
                      op2_reg_class:$rs2,
                      // Value to indicate no rounding mode change in
                      // RISCVInsertReadWriteCSR
                      FRM_DYN,
-                     GPR:$vl, sew, TU_MU)>;
+                     GPR:$vl, log2sew, TU_MU)>;
 }
 
 class VPatBinaryVL_XI<SDPatternOperator vop,
@@ -1591,7 +1595,8 @@ multiclass VPatBinaryFPWVL_VV_VF<SDNode vop, string instruction_name> {
   }
 }
 
-multiclass VPatBinaryFPWVL_VV_VF_RM<SDNode vop, string instruction_name> {
+multiclass VPatBinaryFPWVL_VV_VF_RM<SDNode vop, string instruction_name,
+                                    bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar vti = fvtiToFWti.Vti;
     defvar wti = fvtiToFWti.Wti;
@@ -1600,11 +1605,11 @@ multiclass VPatBinaryFPWVL_VV_VF_RM<SDNode vop, string instruction_name> {
       def : VPatBinaryVL_V_RM<vop, instruction_name, "VV",
                                        wti.Vector, vti.Vector, vti.Vector, vti.Mask,
                                        vti.Log2SEW, vti.LMul, wti.RegClass, vti.RegClass,
-                                       vti.RegClass>;
+                                       vti.RegClass, isSEWAware>;
       def : VPatBinaryVL_VF_RM<vop, instruction_name#"_V"#vti.ScalarSuffix,
                                         wti.Vector, vti.Vector, vti.Vector, vti.Mask,
                                         vti.Log2SEW, vti.LMul, wti.RegClass, vti.RegClass,
-                                        vti.ScalarRegClass>;
+                                        vti.ScalarRegClass, isSEWAware>;
     }
   }
 }
@@ -1631,8 +1636,9 @@ multiclass VPatBinaryFPWVL_VV_VF_WV_WF<SDNode vop, SDNode vop_w, string instruct
   }
 }
 
-multiclass VPatBinaryFPWVL_VV_VF_WV_WF_RM<SDNode vop, SDNode vop_w, string instruction_name>
-    : VPatBinaryFPWVL_VV_VF_RM<vop, instruction_name> {
+multiclass VPatBinaryFPWVL_VV_VF_WV_WF_RM<
+    SDNode vop, SDNode vop_w, string instruction_name, bit isSEWAware = 0>
+    : VPatBinaryFPWVL_VV_VF_RM<vop, instruction_name, isSEWAware> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar vti = fvtiToFWti.Vti;
     defvar wti = fvtiToFWti.Wti;
@@ -1640,15 +1646,16 @@ multiclass VPatBinaryFPWVL_VV_VF_WV_WF_RM<SDNode vop, SDNode vop_w, string instr
                                  GetVTypePredicates<wti>.Predicates) in {
       defm : VPatTiedBinaryNoMaskVL_V_RM<vop_w, instruction_name, "WV",
                                          wti.Vector, vti.Vector, vti.Log2SEW,
-                                         vti.LMul, wti.RegClass, vti.RegClass>;
+                                         vti.LMul, wti.RegClass, vti.RegClass,
+                                         isSEWAware>;
       def : VPatBinaryVL_V_RM<vop_w, instruction_name, "WV",
                                        wti.Vector, wti.Vector, vti.Vector, vti.Mask,
                                        vti.Log2SEW, vti.LMul, wti.RegClass, wti.RegClass,
-                                       vti.RegClass>;
+                                       vti.RegClass, isSEWAware>;
       def : VPatBinaryVL_VF_RM<vop_w, instruction_name#"_W"#vti.ScalarSuffix,
                                         wti.Vector, wti.Vector, vti.Vector, vti.Mask,
                                         vti.Log2SEW, vti.LMul, wti.RegClass, wti.RegClass,
-                                        vti.ScalarRegClass>;
+                                        vti.ScalarRegClass, isSEWAware>;
     }
   }
 }
@@ -2430,8 +2437,10 @@ defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fsub_vl, "PseudoVFSUB", isSEWAware=1>;
 defm : VPatBinaryFPVL_R_VF_RM<any_riscv_fsub_vl, "PseudoVFRSUB", isSEWAware=1>;
 
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
-defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwadd_vl, riscv_vfwadd_w_vl, "PseudoVFWADD">;
-defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwsub_vl, riscv_vfwsub_w_vl, "PseudoVFWSUB">;
+defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwadd_vl, riscv_vfwadd_w_vl,
+                                      "PseudoVFWADD", isSEWAware=1>;
+defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwsub_vl, riscv_vfwsub_w_vl,
+                                      "PseudoVFWSUB", isSEWAware=1>;
 
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fmul_vl, "PseudoVFMUL">;
@@ -2439,7 +2448,7 @@ defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fdiv_vl, "PseudoVFDIV", isSEWAware=1>;
 defm : VPatBinaryFPVL_R_VF_RM<any_riscv_fdiv_vl, "PseudoVFRDIV", isSEWAware=1>;
 
 // 13.5. Vector Widening Floating-Point Multiply Instructions
-defm : VPatBinaryFPWVL_VV_VF_RM<riscv_vfwmul_vl, "PseudoVFWMUL">;
+defm : VPatBinaryFPWVL_VV_VF_RM<riscv_vfwmul_vl, "PseudoVFWMUL", isSEWAware=1>;
 
 // 13.6 Vector Single-Width Floating-Point Fused Multiply-Add Instructions.
 defm : VPatFPMulAddVL_VV_VF_RM<any_riscv_vfmadd_vl,  "PseudoVFMADD">;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index f14766c984391..2b6fc5e59f803 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -790,18 +790,26 @@ foreach mx = SchedMxListW in {
     defm "" : LMULWriteResMX<"WriteVFWCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
 }
+foreach mx = SchedMxListFW in {
+  foreach sew = SchedSEWSet<mx, isF=1, isWidening=1>.val in {
+    defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
+    defvar IsWorstCase = SiFive7IsWorstCaseMXSEW<mx, sew, SchedMxListFW, isF=1>.c;
+    let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxListFW in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxListFW>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFWALUV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWMulAddV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWMulAddF",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulF",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWALUF",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
 }
 // Narrowing
@@ -1147,14 +1155,14 @@ defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 // 14. Vector Floating-Point Instructions
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 4d81bafdf88b8..00e92cde85227 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -541,16 +541,24 @@ foreach mx = SchedMxListFW in {
     defm "" : LMULWriteResMX<"WriteVFWCvtFToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
   }
 }
+foreach mx = SchedMxListFW in {
+  foreach sew = SchedSEWSet<mx, isF=1, isWidening=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListFW, isF=1>.c;
+    let Latency = 6, ReleaseAtCycles = [LMulLat] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxListFW in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxListFW>.c;
   let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFWALUV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWMulAddV", [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWMulAddF", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulF",    [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWALUF",    [SiFiveP600VectorArith], mx, IsWorstCase>;
   }
 }
 // Narrowing
@@ -935,14 +943,14 @@ defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 // 14. Vector Floating-Point Instructions
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 9b9c72e0767af..8b5bd7a0c28a6 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -253,6 +253,18 @@ multiclass LMULReadAdvanceFW<string name, int val, list<SchedWrite> writes = []>
   : LMULReadAdvanceImpl<name, val, writes>;
 class LMULSchedWriteListFW<list<string> names> : LMULSchedWriteListImpl<names, SchedMxListFW>;
 
+multiclass LMULSEWSchedWritesFW<string name>
+    : LMULSEWSchedWritesImpl<name, SchedMxListFW, isF = 1, isWidening = 1>;
+multiclass LMULSEWSchedReadsFW<string name>
+    : LMULSEWSchedReadsImpl<name, SchedMxListFW, isF = 1, isWidening = 1>;
+multiclass LMULSEWWriteResFW<string name, list<ProcResourceKind> resources>
+    : LMULSEWWriteResImpl<name, resources, SchedMxListFW, isF = 1,
+                          isWidening = 1>;
+multiclass
+    LMULSEWReadAdvanceFW<string name, int val, list<SchedWrite> writes = []>
+    : LMULSEWReadAdvanceImpl<name, val, writes, SchedMxListFW, isF = 1,
+                             isWidening = 1>;
+
 // 3.6 Vector Byte Length vlenb
 def WriteRdVLENB      : SchedWrite;
 
@@ -400,16 +412,16 @@ defm "" : LMULSchedWritesW<"WriteVNClipI">;
 defm "" : LMULSEWSchedWritesF<"WriteVFALUV">;
 defm "" : LMULSEWSchedWritesF<"WriteVFALUF">;
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
-defm "" : LMULSchedWritesFW<"WriteVFWALUV">;
-defm "" : LMULSchedWritesFW<"WriteVFWALUF">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWALUV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWALUF">;
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 defm "" : LMULSchedWrites<"WriteVFMulV">;
 defm "" : LMULSchedWrites<"WriteVFMulF">;
 defm "" : LMULSEWSchedWritesF<"WriteVFDivV">;
 defm "" : LMULSEWSchedWritesF<"WriteVFDivF">;
 // 13.5. Vector Widening Floating-Point Multiply
-defm "" : LMULSchedWritesFW<"WriteVFWMulV">;
-defm "" : LMULSchedWritesFW<"WriteVFWMulF">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWMulV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWMulF">;
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
 defm "" : LMULSchedWrites<"WriteVFMulAddV">;
 defm "" : LMULSchedWrites<"WriteVFMulAddF">;
@@ -625,16 +637,16 @@ defm "" : LMULSchedReadsW<"ReadVNClipX">;
 defm "" : LMULSEWSchedReadsF<"ReadVFALUV">;
 defm "" : LMULSEWSchedReadsF<"ReadVFALUF">;
 // 13.3. Vector Widening Floating-Point Add/Subtract Instructions
-defm "" : LMULSchedReadsFW<"ReadVFWALUV">;
-defm "" : LMULSchedReadsFW<"ReadVFWALUF">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWALUV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWALUF">;
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 defm "" : LMULSchedReads<"ReadVFMulV">;
 defm "" : LMULSchedReads<"ReadVFMulF">;
 defm "" : LMULSEWSchedReadsF<"ReadVFDivV">;
 defm "" : LMULSEWSchedReadsF<"ReadVFDivF">;
 // 13.5. Vector Widening Floating-Point Multiply
-defm "" : LMULSchedReadsFW<"ReadVFWMulV">;
-defm "" : LMULSchedReadsFW<"ReadVFWMulF">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWMulV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWMulF">;
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
 defm "" : LMULSchedReads<"ReadVFMulAddV">;
 defm "" : LMULSchedReads<"ReadVFMulAddF">;
@@ -870,14 +882,14 @@ defm "" : LMULWriteResW<"WriteVNClipI", []>;
 // 13. Vector Floating-Point Instructions
 defm "" : LMULSEWWriteResF<"WriteVFALUV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFALUF", []>;
-defm "" : LMULWriteResFW<"WriteVFWALUV", []>;
-defm "" : LMULWriteResFW<"WriteVFWALUF", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWALUV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWALUF", []>;
 defm "" : LMULWriteRes<"WriteVFMulV", []>;
 defm "" : LMULWriteRes<"WriteVFMulF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFDivV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFDivF", []>;
-defm "" : LMULWriteResFW<"WriteVFWMulV", []>;
-defm "" : LMULWriteResFW<"WriteVFWMulF", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWMulV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWMulF", []>;
 defm "" : LMULWriteRes<"WriteVFMulAddV", []>;
 defm "" : LMULWriteRes<"WriteVFMulAddF", []>;
 defm "" : LMULWriteResFW<"WriteVFWMulAddV", []>;
@@ -1026,14 +1038,14 @@ defm "" : LMULReadAdvanceW<"ReadVNClipX", 0>;
 // 13. Vector Floating-Point Instructions
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWALUF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;

From d309d7e4fe51a0ff0ce6430a8af585e2af2772a3 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 08:57:15 -0700
Subject: [PATCH 250/886] [RISCV] Split PseudoVFMUL by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td  | 13 +++++++------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td    |  2 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td    |  2 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td       |  8 ++++----
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td    |  9 +++++----
 llvm/lib/Target/RISCV/RISCVScheduleV.td          | 16 ++++++++--------
 6 files changed, 26 insertions(+), 24 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index 1097eb4fb6431..c00c650d2d368 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2832,16 +2832,17 @@ multiclass VPseudoVDIV_VV_VX {
 
 multiclass VPseudoVFMUL_VV_VF_RM {
   foreach m = MxListF in {
-    defm "" : VPseudoBinaryFV_VV_RM<m>,
-              SchedBinary<"WriteVFMulV", "ReadVFMulV", "ReadVFMulV", m.MX,
-                          forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm "" : VPseudoBinaryFV_VV_RM<m, "", sew=e>,
+                SchedBinary<"WriteVFMulV", "ReadVFMulV", "ReadVFMulV", m.MX, e,
+                            forceMergeOpRead=true>;
   }
 
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoBinaryV_VF_RM<m, f>,
+      defm "" : VPseudoBinaryV_VF_RM<m, f, "", sew=f.SEW>,
                 SchedBinary<"WriteVFMulF", "ReadVFMulV", "ReadVFMulF", m.MX,
-                            forceMergeOpRead=true>;
+                            f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -7106,7 +7107,7 @@ defm : VPatBinaryW_WV_WX_RM<"int_riscv_vfwsub_w", "PseudoVFWSUB",
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 //===----------------------------------------------------------------------===//
 defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfmul", "PseudoVFMUL",
-                            AllFloatVectors>;
+                            AllFloatVectors, isSEWAware=1>;
 defm : VPatBinaryV_VV_VX_RM<"int_riscv_vfdiv", "PseudoVFDIV",
                             AllFloatVectors, isSEWAware=1>;
 defm : VPatBinaryV_VX_RM<"int_riscv_vfrdiv", "PseudoVFRDIV",
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 37acd9667a546..26847e1bdd829 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -1213,7 +1213,7 @@ defm : VPatWidenBinaryFPSDNode_VV_VF_WV_WF_RM<fadd, "PseudoVFWADD">;
 defm : VPatWidenBinaryFPSDNode_VV_VF_WV_WF_RM<fsub, "PseudoVFWSUB">;
 
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
-defm : VPatBinaryFPSDNode_VV_VF_RM<any_fmul, "PseudoVFMUL">;
+defm : VPatBinaryFPSDNode_VV_VF_RM<any_fmul, "PseudoVFMUL", isSEWAware=1>;
 defm : VPatBinaryFPSDNode_VV_VF_RM<any_fdiv, "PseudoVFDIV", isSEWAware=1>;
 defm : VPatBinaryFPSDNode_R_VF_RM<any_fdiv, "PseudoVFRDIV", isSEWAware=1>;
 
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index d2f575459e320..3225260dc6f77 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -2443,7 +2443,7 @@ defm : VPatBinaryFPWVL_VV_VF_WV_WF_RM<riscv_vfwsub_vl, riscv_vfwsub_w_vl,
                                       "PseudoVFWSUB", isSEWAware=1>;
 
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
-defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fmul_vl, "PseudoVFMUL">;
+defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fmul_vl, "PseudoVFMUL", isSEWAware=1>;
 defm : VPatBinaryFPVL_VV_VF_RM<any_riscv_fdiv_vl, "PseudoVFDIV", isSEWAware=1>;
 defm : VPatBinaryFPVL_R_VF_RM<any_riscv_fdiv_vl, "PseudoVFRDIV", isSEWAware=1>;
 
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 2b6fc5e59f803..af65b706b0e20 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -739,6 +739,8 @@ foreach mx = SchedMxListF in {
     let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -746,8 +748,6 @@ foreach mx = SchedMxList in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFMulV",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMulF",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddV",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddF",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFRecpV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
@@ -1157,8 +1157,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 00e92cde85227..5819ce5aa3153 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -493,6 +493,9 @@ foreach mx = SchedMxListF in {
     let Latency = 6, ReleaseAtCycles = [LMulLat] in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUV",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+
     }
   }
 }
@@ -500,8 +503,6 @@ foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFMulV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMulF",    [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddV", [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMulAddF", [SiFiveP600VectorArith], mx, IsWorstCase>;
   }
@@ -945,8 +946,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 8b5bd7a0c28a6..9cb3cd1c3804d 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -415,8 +415,8 @@ defm "" : LMULSEWSchedWritesF<"WriteVFALUF">;
 defm "" : LMULSEWSchedWritesFW<"WriteVFWALUV">;
 defm "" : LMULSEWSchedWritesFW<"WriteVFWALUF">;
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
-defm "" : LMULSchedWrites<"WriteVFMulV">;
-defm "" : LMULSchedWrites<"WriteVFMulF">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMulV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMulF">;
 defm "" : LMULSEWSchedWritesF<"WriteVFDivV">;
 defm "" : LMULSEWSchedWritesF<"WriteVFDivF">;
 // 13.5. Vector Widening Floating-Point Multiply
@@ -640,8 +640,8 @@ defm "" : LMULSEWSchedReadsF<"ReadVFALUF">;
 defm "" : LMULSEWSchedReadsFW<"ReadVFWALUV">;
 defm "" : LMULSEWSchedReadsFW<"ReadVFWALUF">;
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
-defm "" : LMULSchedReads<"ReadVFMulV">;
-defm "" : LMULSchedReads<"ReadVFMulF">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMulV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMulF">;
 defm "" : LMULSEWSchedReadsF<"ReadVFDivV">;
 defm "" : LMULSEWSchedReadsF<"ReadVFDivF">;
 // 13.5. Vector Widening Floating-Point Multiply
@@ -884,8 +884,8 @@ defm "" : LMULSEWWriteResF<"WriteVFALUV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFALUF", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWALUV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWALUF", []>;
-defm "" : LMULWriteRes<"WriteVFMulV", []>;
-defm "" : LMULWriteRes<"WriteVFMulF", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMulV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMulF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFDivV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFDivF", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulV", []>;
@@ -1040,8 +1040,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFALUV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFALUF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWALUF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;

From c6b7944be4dfbb1fb35301c670812726845acaa7 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 10:52:54 -0700
Subject: [PATCH 251/886] [RISCV] Split single width floating point fused
 multiple-add pseudo instructions by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      | 254 +++++++++++-------
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    |  93 ++++---
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |   2 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |   4 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |   8 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td |  19 +-
 llvm/lib/Target/RISCV/RISCVScheduleV.td       |  16 +-
 7 files changed, 246 insertions(+), 150 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 3b76e7158ffd8..0eed604e251b8 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -2712,31 +2712,54 @@ std::string RISCVInstrInfo::createMIROperandComment(
 }
 
 // clang-format off
-#define CASE_VFMA_OPCODE_COMMON(OP, TYPE, LMUL)                                \
+#define CASE_VMA_OPCODE_COMMON(OP, TYPE, LMUL)                                 \
   RISCV::PseudoV##OP##_##TYPE##_##LMUL
 
-#define CASE_VFMA_OPCODE_LMULS_M1(OP, TYPE)                                    \
-  CASE_VFMA_OPCODE_COMMON(OP, TYPE, M1):                                       \
-  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M2):                                  \
-  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M4):                                  \
-  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M8)
+#define CASE_VMA_OPCODE_LMULS_M1(OP, TYPE)                                     \
+  CASE_VMA_OPCODE_COMMON(OP, TYPE, M1):                                        \
+  case CASE_VMA_OPCODE_COMMON(OP, TYPE, M2):                                   \
+  case CASE_VMA_OPCODE_COMMON(OP, TYPE, M4):                                   \
+  case CASE_VMA_OPCODE_COMMON(OP, TYPE, M8)
 
-#define CASE_VFMA_OPCODE_LMULS_MF2(OP, TYPE)                                   \
-  CASE_VFMA_OPCODE_COMMON(OP, TYPE, MF2):                                      \
-  case CASE_VFMA_OPCODE_LMULS_M1(OP, TYPE)
+#define CASE_VMA_OPCODE_LMULS_MF2(OP, TYPE)                                    \
+  CASE_VMA_OPCODE_COMMON(OP, TYPE, MF2):                                       \
+  case CASE_VMA_OPCODE_LMULS_M1(OP, TYPE)
 
-#define CASE_VFMA_OPCODE_LMULS_MF4(OP, TYPE)                                   \
-  CASE_VFMA_OPCODE_COMMON(OP, TYPE, MF4):                                      \
-  case CASE_VFMA_OPCODE_LMULS_MF2(OP, TYPE)
+#define CASE_VMA_OPCODE_LMULS_MF4(OP, TYPE)                                    \
+  CASE_VMA_OPCODE_COMMON(OP, TYPE, MF4):                                       \
+  case CASE_VMA_OPCODE_LMULS_MF2(OP, TYPE)
 
-#define CASE_VFMA_OPCODE_LMULS(OP, TYPE)                                       \
-  CASE_VFMA_OPCODE_COMMON(OP, TYPE, MF8):                                      \
-  case CASE_VFMA_OPCODE_LMULS_MF4(OP, TYPE)
+#define CASE_VMA_OPCODE_LMULS(OP, TYPE)                                        \
+  CASE_VMA_OPCODE_COMMON(OP, TYPE, MF8):                                       \
+  case CASE_VMA_OPCODE_LMULS_MF4(OP, TYPE)
+
+// VFMA instructions are SEW specific.
+#define CASE_VFMA_OPCODE_COMMON(OP, TYPE, LMUL, SEW)                           \
+  RISCV::PseudoV##OP##_##TYPE##_##LMUL##_##SEW
+
+#define CASE_VFMA_OPCODE_LMULS_M1(OP, TYPE, SEW)                               \
+  CASE_VFMA_OPCODE_COMMON(OP, TYPE, M1, SEW):                                  \
+  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M2, SEW):                             \
+  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M4, SEW):                             \
+  case CASE_VFMA_OPCODE_COMMON(OP, TYPE, M8, SEW)
+
+#define CASE_VFMA_OPCODE_LMULS_MF2(OP, TYPE, SEW)                              \
+  CASE_VFMA_OPCODE_COMMON(OP, TYPE, MF2, SEW):                                 \
+  case CASE_VFMA_OPCODE_LMULS_M1(OP, TYPE, SEW)
+
+#define CASE_VFMA_OPCODE_LMULS_MF4(OP, TYPE, SEW)                              \
+  CASE_VFMA_OPCODE_COMMON(OP, TYPE, MF4, SEW):                                 \
+  case CASE_VFMA_OPCODE_LMULS_MF2(OP, TYPE, SEW)
+
+#define CASE_VFMA_OPCODE_VV(OP)                                                \
+  CASE_VFMA_OPCODE_LMULS_MF4(OP, VV, E16):                                     \
+  case CASE_VFMA_OPCODE_LMULS_MF2(OP, VV, E32):                                \
+  case CASE_VFMA_OPCODE_LMULS_M1(OP, VV, E64)
 
 #define CASE_VFMA_SPLATS(OP)                                                   \
-  CASE_VFMA_OPCODE_LMULS_MF4(OP, VFPR16):                                      \
-  case CASE_VFMA_OPCODE_LMULS_MF2(OP, VFPR32):                                 \
-  case CASE_VFMA_OPCODE_LMULS_M1(OP, VFPR64)
+  CASE_VFMA_OPCODE_LMULS_MF4(OP, VFPR16, E16):                                 \
+  case CASE_VFMA_OPCODE_LMULS_MF2(OP, VFPR32, E32):                            \
+  case CASE_VFMA_OPCODE_LMULS_M1(OP, VFPR64, E64)
 // clang-format on
 
 bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
@@ -2776,16 +2799,16 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
   case CASE_VFMA_SPLATS(FNMSUB):
   case CASE_VFMA_SPLATS(FNMACC):
   case CASE_VFMA_SPLATS(FNMSAC):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMSAC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMSAC, VV):
-  case CASE_VFMA_OPCODE_LMULS(MADD, VX):
-  case CASE_VFMA_OPCODE_LMULS(NMSUB, VX):
-  case CASE_VFMA_OPCODE_LMULS(MACC, VX):
-  case CASE_VFMA_OPCODE_LMULS(NMSAC, VX):
-  case CASE_VFMA_OPCODE_LMULS(MACC, VV):
-  case CASE_VFMA_OPCODE_LMULS(NMSAC, VV): {
+  case CASE_VFMA_OPCODE_VV(FMACC):
+  case CASE_VFMA_OPCODE_VV(FMSAC):
+  case CASE_VFMA_OPCODE_VV(FNMACC):
+  case CASE_VFMA_OPCODE_VV(FNMSAC):
+  case CASE_VMA_OPCODE_LMULS(MADD, VX):
+  case CASE_VMA_OPCODE_LMULS(NMSUB, VX):
+  case CASE_VMA_OPCODE_LMULS(MACC, VX):
+  case CASE_VMA_OPCODE_LMULS(NMSAC, VX):
+  case CASE_VMA_OPCODE_LMULS(MACC, VV):
+  case CASE_VMA_OPCODE_LMULS(NMSAC, VV): {
     // If the tail policy is undisturbed we can't commute.
     assert(RISCVII::hasVecPolicyOp(MI.getDesc().TSFlags));
     if ((MI.getOperand(MI.getNumExplicitOperands() - 1).getImm() & 1) == 0)
@@ -2800,12 +2823,12 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
       return false;
     return true;
   }
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMSUB, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMSUB, VV):
-  case CASE_VFMA_OPCODE_LMULS(MADD, VV):
-  case CASE_VFMA_OPCODE_LMULS(NMSUB, VV): {
+  case CASE_VFMA_OPCODE_VV(FMADD):
+  case CASE_VFMA_OPCODE_VV(FMSUB):
+  case CASE_VFMA_OPCODE_VV(FNMADD):
+  case CASE_VFMA_OPCODE_VV(FNMSUB):
+  case CASE_VMA_OPCODE_LMULS(MADD, VV):
+  case CASE_VMA_OPCODE_LMULS(NMSUB, VV): {
     // If the tail policy is undisturbed we can't commute.
     assert(RISCVII::hasVecPolicyOp(MI.getDesc().TSFlags));
     if ((MI.getOperand(MI.getNumExplicitOperands() - 1).getImm() & 1) == 0)
@@ -2874,33 +2897,68 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
   return TargetInstrInfo::findCommutedOpIndices(MI, SrcOpIdx1, SrcOpIdx2);
 }
 
-#define CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, LMUL)               \
+// clang-format off
+#define CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, LMUL)                \
   case RISCV::PseudoV##OLDOP##_##TYPE##_##LMUL:                                \
     Opc = RISCV::PseudoV##NEWOP##_##TYPE##_##LMUL;                             \
     break;
 
-#define CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE)                   \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M1)                       \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M2)                       \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M4)                       \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M8)
+#define CASE_VMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE)                    \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M1)                        \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M2)                        \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M4)                        \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M8)
+
+#define CASE_VMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE)                   \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF2)                       \
+  CASE_VMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE)
+
+#define CASE_VMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE)                   \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF4)                       \
+  CASE_VMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE)
+
+#define CASE_VMA_CHANGE_OPCODE_LMULS(OLDOP, NEWOP, TYPE)                       \
+  CASE_VMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF8)                       \
+  CASE_VMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE)
+
+#define CASE_VMA_CHANGE_OPCODE_SPLATS(OLDOP, NEWOP)                            \
+  CASE_VMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, VFPR16)                       \
+  CASE_VMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VFPR32)                       \
+  CASE_VMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, VFPR64)
+
+// VFMA depends on SEW.
+#define CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, LMUL, SEW)          \
+  case RISCV::PseudoV##OLDOP##_##TYPE##_##LMUL##_##SEW:                        \
+    Opc = RISCV::PseudoV##NEWOP##_##TYPE##_##LMUL##_##SEW;                     \
+    break;
+
+#define CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE, SEW)              \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M1, SEW)                  \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M2, SEW)                  \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M4, SEW)                  \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, M8, SEW)
+
+#define CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE, SEW)             \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF2, SEW)                 \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE, SEW)
 
-#define CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE)                  \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF2)                      \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE)
+#define CASE_VFMA_CHANGE_OPCODE_VV(OLDOP, NEWOP)                               \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, VV, MF4, E16)                   \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E16)                     \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E32)
 
-#define CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE)                  \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF4)                      \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE)
+#define CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE, SEW)             \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF4, SEW)                 \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, TYPE, SEW)
 
-#define CASE_VFMA_CHANGE_OPCODE_LMULS(OLDOP, NEWOP, TYPE)                      \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF8)                      \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE)
+#define CASE_VFMA_CHANGE_OPCODE_LMULS(OLDOP, NEWOP, TYPE, SEW)                 \
+  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF8, SEW)                 \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE, SEW)
 
 #define CASE_VFMA_CHANGE_OPCODE_SPLATS(OLDOP, NEWOP)                           \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, VFPR16)                      \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VFPR32)                      \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, VFPR64)
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, VFPR16, E16)                 \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VFPR32, E32)                 \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, VFPR64, E64)
 
 MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
                                                      bool NewMI,
@@ -2939,16 +2997,16 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
   case CASE_VFMA_SPLATS(FNMADD):
   case CASE_VFMA_SPLATS(FNMSAC):
   case CASE_VFMA_SPLATS(FNMSUB):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMSAC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMSAC, VV):
-  case CASE_VFMA_OPCODE_LMULS(MADD, VX):
-  case CASE_VFMA_OPCODE_LMULS(NMSUB, VX):
-  case CASE_VFMA_OPCODE_LMULS(MACC, VX):
-  case CASE_VFMA_OPCODE_LMULS(NMSAC, VX):
-  case CASE_VFMA_OPCODE_LMULS(MACC, VV):
-  case CASE_VFMA_OPCODE_LMULS(NMSAC, VV): {
+  case CASE_VFMA_OPCODE_VV(FMACC):
+  case CASE_VFMA_OPCODE_VV(FMSAC):
+  case CASE_VFMA_OPCODE_VV(FNMACC):
+  case CASE_VFMA_OPCODE_VV(FNMSAC):
+  case CASE_VMA_OPCODE_LMULS(MADD, VX):
+  case CASE_VMA_OPCODE_LMULS(NMSUB, VX):
+  case CASE_VMA_OPCODE_LMULS(MACC, VX):
+  case CASE_VMA_OPCODE_LMULS(NMSAC, VX):
+  case CASE_VMA_OPCODE_LMULS(MACC, VV):
+  case CASE_VMA_OPCODE_LMULS(NMSAC, VV): {
     // It only make sense to toggle these between clobbering the
     // addend/subtrahend/minuend one of the multiplicands.
     assert((OpIdx1 == 1 || OpIdx2 == 1) && "Unexpected opcode index");
@@ -2965,16 +3023,16 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
       CASE_VFMA_CHANGE_OPCODE_SPLATS(FNMADD, FNMACC)
       CASE_VFMA_CHANGE_OPCODE_SPLATS(FNMSAC, FNMSUB)
       CASE_VFMA_CHANGE_OPCODE_SPLATS(FNMSUB, FNMSAC)
-      CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FMACC, FMADD, VV)
-      CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FMSAC, FMSUB, VV)
-      CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FNMACC, FNMADD, VV)
-      CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FNMSAC, FNMSUB, VV)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(MACC, MADD, VX)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(MADD, MACC, VX)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VX)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VX)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(MACC, MADD, VV)
-      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VV)
+      CASE_VFMA_CHANGE_OPCODE_VV(FMACC, FMADD)
+      CASE_VFMA_CHANGE_OPCODE_VV(FMSAC, FMSUB)
+      CASE_VFMA_CHANGE_OPCODE_VV(FNMACC, FNMADD)
+      CASE_VFMA_CHANGE_OPCODE_VV(FNMSAC, FNMSUB)
+      CASE_VMA_CHANGE_OPCODE_LMULS(MACC, MADD, VX)
+      CASE_VMA_CHANGE_OPCODE_LMULS(MADD, MACC, VX)
+      CASE_VMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VX)
+      CASE_VMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VX)
+      CASE_VMA_CHANGE_OPCODE_LMULS(MACC, MADD, VV)
+      CASE_VMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VV)
     }
 
     auto &WorkingMI = cloneIfNew(MI);
@@ -2982,12 +3040,12 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
     return TargetInstrInfo::commuteInstructionImpl(WorkingMI, /*NewMI=*/false,
                                                    OpIdx1, OpIdx2);
   }
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FMSUB, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS_MF4(FNMSUB, VV):
-  case CASE_VFMA_OPCODE_LMULS(MADD, VV):
-  case CASE_VFMA_OPCODE_LMULS(NMSUB, VV): {
+  case CASE_VFMA_OPCODE_VV(FMADD):
+  case CASE_VFMA_OPCODE_VV(FMSUB):
+  case CASE_VFMA_OPCODE_VV(FNMADD):
+  case CASE_VFMA_OPCODE_VV(FNMSUB):
+  case CASE_VMA_OPCODE_LMULS(MADD, VV):
+  case CASE_VMA_OPCODE_LMULS(NMSUB, VV): {
     assert((OpIdx1 == 1 || OpIdx2 == 1) && "Unexpected opcode index");
     // If one of the operands, is the addend we need to change opcode.
     // Otherwise we're just swapping 2 of the multiplicands.
@@ -2996,12 +3054,12 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
       switch (MI.getOpcode()) {
         default:
           llvm_unreachable("Unexpected opcode");
-        CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FMADD, FMACC, VV)
-        CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FMSUB, FMSAC, VV)
-        CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FNMADD, FNMACC, VV)
-        CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(FNMSUB, FNMSAC, VV)
-        CASE_VFMA_CHANGE_OPCODE_LMULS(MADD, MACC, VV)
-        CASE_VFMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VV)
+        CASE_VFMA_CHANGE_OPCODE_VV(FMADD, FMACC)
+        CASE_VFMA_CHANGE_OPCODE_VV(FMSUB, FMSAC)
+        CASE_VFMA_CHANGE_OPCODE_VV(FNMADD, FNMACC)
+        CASE_VFMA_CHANGE_OPCODE_VV(FNMSUB, FNMSAC)
+        CASE_VMA_CHANGE_OPCODE_LMULS(MADD, MACC, VV)
+        CASE_VMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VV)
       }
 
       auto &WorkingMI = cloneIfNew(MI);
@@ -3017,12 +3075,17 @@ MachineInstr *RISCVInstrInfo::commuteInstructionImpl(MachineInstr &MI,
   return TargetInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
 }
 
-#undef CASE_VFMA_CHANGE_OPCODE_SPLATS
-#undef CASE_VFMA_CHANGE_OPCODE_LMULS
-#undef CASE_VFMA_CHANGE_OPCODE_COMMON
-#undef CASE_VFMA_SPLATS
-#undef CASE_VFMA_OPCODE_LMULS
+#undef CASE_VMA_OPCODE_COMMON
+#undef CASE_VMA_OPCODE_LMULS_M1
+#undef CASE_VMA_OPCODE_LMULS_MF2
+#undef CASE_VMA_OPCODE_LMULS_MF4
+#undef CASE_VMA_OPCODE_LMULS
 #undef CASE_VFMA_OPCODE_COMMON
+#undef CASE_VFMA_OPCODE_LMULS_M1
+#undef CASE_VFMA_OPCODE_LMULS_MF2
+#undef CASE_VFMA_OPCODE_LMULS_MF4
+#undef CASE_VFMA_OPCODE_VV
+#undef CASE_VFMA_SPLATS
 
 // clang-format off
 #define CASE_WIDEOP_OPCODE_COMMON(OP, LMUL)                                    \
@@ -3190,10 +3253,17 @@ MachineInstr *RISCVInstrInfo::convertToThreeAddress(MachineInstr &MI,
   return MIB;
 }
 
-#undef CASE_WIDEOP_CHANGE_OPCODE_LMULS
-#undef CASE_WIDEOP_CHANGE_OPCODE_COMMON
-#undef CASE_WIDEOP_OPCODE_LMULS
 #undef CASE_WIDEOP_OPCODE_COMMON
+#undef CASE_WIDEOP_OPCODE_LMULS_MF4
+#undef CASE_WIDEOP_OPCODE_LMULS
+#undef CASE_WIDEOP_CHANGE_OPCODE_COMMON
+#undef CASE_WIDEOP_CHANGE_OPCODE_LMULS_MF4
+#undef CASE_WIDEOP_CHANGE_OPCODE_LMULS
+#undef CASE_FP_WIDEOP_OPCODE_COMMON
+#undef CASE_FP_WIDEOP_OPCODE_LMULS_MF4
+#undef CASE_FP_WIDEOP_CHANGE_OPCODE_COMMON
+#undef CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS_MF4
+#undef CASE_FP_WIDEOP_CHANGE_OPCODE_LMULS
 
 void RISCVInstrInfo::mulImm(MachineFunction &MF, MachineBasicBlock &MBB,
                             MachineBasicBlock::iterator II, const DebugLoc &DL,
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index c00c650d2d368..f1159ee633eaa 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -3234,15 +3234,17 @@ multiclass VPseudoTernaryWithPolicyRoundingMode<VReg RetClass,
                                                 DAGOperand Op2Class,
                                                 LMULInfo MInfo,
                                                 string Constraint = "",
+                                                int sew = 0,
                                                 bit Commutable = 0,
                                                 int TargetConstraintType = 1> {
   let VLMul = MInfo.value in {
+    defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
     let isCommutable = Commutable in
-    def "_" # MInfo.MX :
+    def suffix :
         VPseudoTernaryNoMaskWithPolicyRoundingMode<RetClass, Op1Class,
                                                    Op2Class, Constraint,
                                                    TargetConstraintType>;
-    def "_" # MInfo.MX # "_MASK" :
+    def suffix # "_MASK" :
         VPseudoBinaryMaskPolicyRoundingMode<RetClass, Op1Class,
                                             Op2Class, Constraint,
                                             UsesVXRM_=0,
@@ -3256,9 +3258,9 @@ multiclass VPseudoTernaryV_VV_AAXA<LMULInfo m, string Constraint = ""> {
                                       Constraint, Commutable=1>;
 }
 
-multiclass VPseudoTernaryV_VV_AAXA_RM<LMULInfo m, string Constraint = ""> {
+multiclass VPseudoTernaryV_VV_AAXA_RM<LMULInfo m, string Constraint = "", int sew = 0> {
   defm _VV : VPseudoTernaryWithPolicyRoundingMode<m.vrclass, m.vrclass, m.vrclass, m,
-                                                  Constraint, Commutable=1>;
+                                                  Constraint, sew, Commutable=1>;
 }
 
 multiclass VPseudoTernaryV_VX_AAXA<LMULInfo m, string Constraint = ""> {
@@ -3272,10 +3274,11 @@ multiclass VPseudoTernaryV_VF_AAXA<LMULInfo m, FPR_Info f, string Constraint = "
                                               Commutable=1>;
 }
 
-multiclass VPseudoTernaryV_VF_AAXA_RM<LMULInfo m, FPR_Info f, string Constraint = ""> {
+multiclass VPseudoTernaryV_VF_AAXA_RM<LMULInfo m, FPR_Info f,
+                                      string Constraint = "", int sew = 0> {
   defm "_V" # f.FX : VPseudoTernaryWithPolicyRoundingMode<m.vrclass, f.fprclass,
                                                           m.vrclass, m, Constraint,
-                                                          Commutable=1>;
+                                                          sew, Commutable=1>;
 }
 
 multiclass VPseudoTernaryW_VV<LMULInfo m> {
@@ -3345,16 +3348,17 @@ multiclass VPseudoVMAC_VV_VX_AAXA<string Constraint = ""> {
 
 multiclass VPseudoVMAC_VV_VF_AAXA_RM<string Constraint = ""> {
   foreach m = MxListF in {
-    defm "" : VPseudoTernaryV_VV_AAXA_RM<m, Constraint>,
-              SchedTernary<"WriteVFMulAddV", "ReadVFMulAddV", "ReadVFMulAddV",
-                           "ReadVFMulAddV", m.MX>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm "" : VPseudoTernaryV_VV_AAXA_RM<m, Constraint, sew=e>,
+                SchedTernary<"WriteVFMulAddV", "ReadVFMulAddV", "ReadVFMulAddV",
+                             "ReadVFMulAddV", m.MX, e>;
   }
 
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoTernaryV_VF_AAXA_RM<m, f, Constraint>,
+      defm "" : VPseudoTernaryV_VF_AAXA_RM<m, f, Constraint, sew=f.SEW>,
                 SchedTernary<"WriteVFMulAddF", "ReadVFMulAddV", "ReadVFMulAddF",
-                             "ReadVFMulAddV", m.MX>;
+                             "ReadVFMulAddV", m.MX, f.SEW>;
     }
   }
 }
@@ -4461,23 +4465,26 @@ class VPatTernaryNoMaskWithPolicyRoundingMode<string intrinsic,
                                   ValueType result_type,
                                   ValueType op1_type,
                                   ValueType op2_type,
-                                  int sew,
+                                  int log2sew,
                                   LMULInfo vlmul,
                                   VReg result_reg_class,
                                   RegisterClass op1_reg_class,
-                                  DAGOperand op2_kind> :
+                                  DAGOperand op2_kind,
+                                  bit isSEWAware = 0> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic)
                     (result_type result_reg_class:$rs3),
                     (op1_type op1_reg_class:$rs1),
                     (op2_type op2_kind:$rs2),
                     (XLenVT timm:$round),
                     VLOpFrag, (XLenVT timm:$policy))),
-                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX)
+                   (!cast<Instruction>(!if(isSEWAware,
+                          inst#"_"#kind#"_"#vlmul.MX#"_E"#!shl(1, log2sew),
+                          inst#"_"#kind#"_"#vlmul.MX))
                     result_reg_class:$rs3,
                     (op1_type op1_reg_class:$rs1),
                     op2_kind:$rs2,
                     (XLenVT timm:$round),
-                    GPR:$vl, sew, (XLenVT timm:$policy))>;
+                    GPR:$vl, log2sew, (XLenVT timm:$policy))>;
 
 class VPatTernaryMask<string intrinsic,
                       string inst,
@@ -4536,11 +4543,12 @@ class VPatTernaryMaskPolicyRoundingMode<string intrinsic,
                                         ValueType op1_type,
                                         ValueType op2_type,
                                         ValueType mask_type,
-                                        int sew,
+                                        int log2sew,
                                         LMULInfo vlmul,
                                         VReg result_reg_class,
                                         RegisterClass op1_reg_class,
-                                        DAGOperand op2_kind> :
+                                        DAGOperand op2_kind,
+                                        bit isSEWAware = 0> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic#"_mask")
                     (result_type result_reg_class:$rs3),
                     (op1_type op1_reg_class:$rs1),
@@ -4548,13 +4556,15 @@ class VPatTernaryMaskPolicyRoundingMode<string intrinsic,
                     (mask_type V0),
                     (XLenVT timm:$round),
                     VLOpFrag, (XLenVT timm:$policy))),
-                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX # "_MASK")
+                   (!cast<Instruction>(!if(isSEWAware,
+                          inst#"_"#kind#"_"#vlmul.MX#"_E"#!shl(1, log2sew) # "_MASK",
+                          inst#"_"#kind#"_"#vlmul.MX # "_MASK"))
                     result_reg_class:$rs3,
                     (op1_type op1_reg_class:$rs1),
                     op2_kind:$rs2,
                     (mask_type V0),
                     (XLenVT timm:$round),
-                    GPR:$vl, sew, (XLenVT timm:$policy))>;
+                    GPR:$vl, log2sew, (XLenVT timm:$policy))>;
 
 class VPatTernaryMaskTA<string intrinsic,
                         string inst,
@@ -5564,15 +5574,16 @@ multiclass VPatTernaryWithPolicyRoundingMode<string intrinsic,
                                              LMULInfo vlmul,
                                              VReg result_reg_class,
                                              RegisterClass op1_reg_class,
-                                             DAGOperand op2_kind> {
+                                             DAGOperand op2_kind,
+                                             bit isSEWAware = 0> {
   def : VPatTernaryNoMaskWithPolicyRoundingMode<intrinsic, inst, kind, result_type,
                                                 op1_type, op2_type, sew, vlmul,
                                                 result_reg_class, op1_reg_class,
-                                                op2_kind>;
+                                                op2_kind, isSEWAware>;
   def : VPatTernaryMaskPolicyRoundingMode<intrinsic, inst, kind, result_type, op1_type,
                                                 op2_type, mask_type, sew, vlmul,
                                                 result_reg_class, op1_reg_class,
-                                                op2_kind>;
+                                                op2_kind, isSEWAware>;
 }
 
 multiclass VPatTernaryTA<string intrinsic,
@@ -5626,13 +5637,13 @@ multiclass VPatTernaryV_VV_AAXA<string intrinsic, string instruction,
 }
 
 multiclass VPatTernaryV_VV_AAXA_RM<string intrinsic, string instruction,
-                                list<VTypeInfo> vtilist> {
+                                list<VTypeInfo> vtilist, bit isSEWAware = 0> {
   foreach vti = vtilist in
     let Predicates = GetVTypePredicates<vti>.Predicates in
     defm : VPatTernaryWithPolicyRoundingMode<intrinsic, instruction, "VV",
                                              vti.Vector, vti.Vector, vti.Vector, vti.Mask,
                                              vti.Log2SEW, vti.LMul, vti.RegClass,
-                                             vti.RegClass, vti.RegClass>;
+                                             vti.RegClass, vti.RegClass, isSEWAware>;
 }
 
 multiclass VPatTernaryV_VX<string intrinsic, string instruction,
@@ -5657,14 +5668,14 @@ multiclass VPatTernaryV_VX_AAXA<string intrinsic, string instruction,
 }
 
 multiclass VPatTernaryV_VX_AAXA_RM<string intrinsic, string instruction,
-                           list<VTypeInfo> vtilist> {
+                           list<VTypeInfo> vtilist, bit isSEWAware = 0> {
   foreach vti = vtilist in
     let Predicates = GetVTypePredicates<vti>.Predicates in
     defm : VPatTernaryWithPolicyRoundingMode<intrinsic, instruction,
                                              "V"#vti.ScalarSuffix,
                                              vti.Vector, vti.Scalar, vti.Vector, vti.Mask,
                                              vti.Log2SEW, vti.LMul, vti.RegClass,
-                                             vti.ScalarRegClass, vti.RegClass>;
+                                             vti.ScalarRegClass, vti.RegClass, isSEWAware>;
 }
 
 multiclass VPatTernaryV_VI<string intrinsic, string instruction,
@@ -5742,9 +5753,9 @@ multiclass VPatTernaryV_VV_VX_AAXA<string intrinsic, string instruction,
       VPatTernaryV_VX_AAXA<intrinsic, instruction, vtilist>;
 
 multiclass VPatTernaryV_VV_VX_AAXA_RM<string intrinsic, string instruction,
-                              list<VTypeInfo> vtilist>
-    : VPatTernaryV_VV_AAXA_RM<intrinsic, instruction, vtilist>,
-      VPatTernaryV_VX_AAXA_RM<intrinsic, instruction, vtilist>;
+                              list<VTypeInfo> vtilist, bit isSEWAware = 0>
+    : VPatTernaryV_VV_AAXA_RM<intrinsic, instruction, vtilist, isSEWAware>,
+      VPatTernaryV_VX_AAXA_RM<intrinsic, instruction, vtilist, isSEWAware>;
 
 multiclass VPatTernaryV_VX_VI<string intrinsic, string instruction,
                               list<VTypeInfo> vtilist, Operand Imm_type = simm5>
@@ -7122,14 +7133,22 @@ defm : VPatBinaryW_VV_VX_RM<"int_riscv_vfwmul", "PseudoVFWMUL",
 //===----------------------------------------------------------------------===//
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmacc", "PseudoVFMACC", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmacc", "PseudoVFNMACC", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmsac", "PseudoVFMSAC", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmsac", "PseudoVFNMSAC", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmadd", "PseudoVFMADD", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmadd", "PseudoVFNMADD", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmsub", "PseudoVFMSUB", AllFloatVectors>;
-defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmsub", "PseudoVFNMSUB", AllFloatVectors>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmacc", "PseudoVFMACC",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmacc", "PseudoVFNMACC",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmsac", "PseudoVFMSAC",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmsac", "PseudoVFNMSAC",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmadd", "PseudoVFMADD",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmadd", "PseudoVFNMADD",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfmsub", "PseudoVFMSUB",
+                                  AllFloatVectors, isSEWAware=1>;
+defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmsub", "PseudoVFNMSUB",
+                                  AllFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 26847e1bdd829..d9dc373a628d6 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -1224,7 +1224,7 @@ defm : VPatWidenBinaryFPSDNode_VV_VF_RM<fmul, "PseudoVFWMUL">;
 foreach fvti = AllFloatVectors in {
   // NOTE: We choose VFMADD because it has the most commuting freedom. So it
   // works best with how TwoAddressInstructionPass tries commuting.
-  defvar suffix = fvti.LMul.MX;
+  defvar suffix = fvti.LMul.MX # "_E" # fvti.SEW;
   let Predicates = GetVTypePredicates<fvti>.Predicates in {
     def : Pat<(fvti.Vector (any_fma fvti.RegClass:$rs1, fvti.RegClass:$rd,
                                     fvti.RegClass:$rs2)),
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index 3225260dc6f77..a6f24438daa60 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1850,7 +1850,7 @@ multiclass VPatFPMulAddVL_VV_VF<SDPatternOperator vop, string instruction_name>
 
 multiclass VPatFPMulAddVL_VV_VF_RM<SDPatternOperator vop, string instruction_name> {
   foreach vti = AllFloatVectors in {
-  defvar suffix = vti.LMul.MX;
+  defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
   let Predicates = GetVTypePredicates<vti>.Predicates in {
     def : Pat<(vti.Vector (vop vti.RegClass:$rs1, vti.RegClass:$rd,
                                vti.RegClass:$rs2, (vti.Mask V0),
@@ -1916,7 +1916,7 @@ multiclass VPatFPMulAccVL_VV_VF<PatFrag vop, string instruction_name> {
 
 multiclass VPatFPMulAccVL_VV_VF_RM<PatFrag vop, string instruction_name> {
   foreach vti = AllFloatVectors in {
-  defvar suffix = vti.LMul.MX;
+  defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
   let Predicates = GetVTypePredicates<vti>.Predicates in {
     def : Pat<(riscv_vmerge_vl (vti.Mask V0),
                            (vti.Vector (vop vti.RegClass:$rs1, vti.RegClass:$rs2,
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index af65b706b0e20..16bfb509e0856 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -741,6 +741,8 @@ foreach mx = SchedMxListF in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -748,8 +750,6 @@ foreach mx = SchedMxList in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFMulAddV",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMulAddF",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFRecpV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtIToFV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtFToIV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
@@ -1163,8 +1163,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 5819ce5aa3153..0896f3358d230 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -499,13 +499,20 @@ foreach mx = SchedMxListF in {
     }
   }
 }
+
+foreach mx = SchedMxListF in {
+  foreach sew = SchedSEWSet<mx, isF=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
+    let Latency = 6, ReleaseAtCycles = [LMulLat] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
-  let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFMulAddV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMulAddF", [SiFiveP600VectorArith], mx, IsWorstCase>;
-  }
   let Latency = 3, ReleaseAtCycles = [LMulLat] in {
     defm "" : LMULWriteResMX<"WriteVFCvtIToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtFToIV", [SiFiveP600VectorArith], mx, IsWorstCase>;
@@ -952,8 +959,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 9cb3cd1c3804d..40090d1db8874 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -423,8 +423,8 @@ defm "" : LMULSEWSchedWritesF<"WriteVFDivF">;
 defm "" : LMULSEWSchedWritesFW<"WriteVFWMulV">;
 defm "" : LMULSEWSchedWritesFW<"WriteVFWMulF">;
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
-defm "" : LMULSchedWrites<"WriteVFMulAddV">;
-defm "" : LMULSchedWrites<"WriteVFMulAddF">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMulAddV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMulAddF">;
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
 defm "" : LMULSchedWritesFW<"WriteVFWMulAddV">;
 defm "" : LMULSchedWritesFW<"WriteVFWMulAddF">;
@@ -648,8 +648,8 @@ defm "" : LMULSEWSchedReadsF<"ReadVFDivF">;
 defm "" : LMULSEWSchedReadsFW<"ReadVFWMulV">;
 defm "" : LMULSEWSchedReadsFW<"ReadVFWMulF">;
 // 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
-defm "" : LMULSchedReads<"ReadVFMulAddV">;
-defm "" : LMULSchedReads<"ReadVFMulAddF">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMulAddV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMulAddF">;
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
 defm "" : LMULSchedReadsFW<"ReadVFWMulAddV">;
 defm "" : LMULSchedReadsFW<"ReadVFWMulAddF">;
@@ -890,8 +890,8 @@ defm "" : LMULSEWWriteResF<"WriteVFDivV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFDivF", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulF", []>;
-defm "" : LMULWriteRes<"WriteVFMulAddV", []>;
-defm "" : LMULWriteRes<"WriteVFMulAddF", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMulAddV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMulAddF", []>;
 defm "" : LMULWriteResFW<"WriteVFWMulAddV", []>;
 defm "" : LMULWriteResFW<"WriteVFWMulAddF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFSqrtV", []>;
@@ -1046,8 +1046,8 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFDivV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFDivF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;

From 43248ffea7b8de3a33b11768e8c21d2434252528 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 11:58:29 -0700
Subject: [PATCH 252/886] [RISCV] Split widening floating point fused
 multiple-add pseudo instructions by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      |  4 +-
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 68 ++++++++++---------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td | 24 ++++---
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  5 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |  8 +--
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 14 ++--
 llvm/lib/Target/RISCV/RISCVScheduleV.td       | 16 ++---
 .../CodeGen/RISCV/rvv/fixed-vectors-fmf.ll    |  4 +-
 .../RISCV/rvv/pass-fast-math-flags-sdnode.ll  |  4 +-
 9 files changed, 74 insertions(+), 73 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 0eed604e251b8..004a228ccc131 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -2943,9 +2943,9 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
   CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE, SEW)
 
 #define CASE_VFMA_CHANGE_OPCODE_VV(OLDOP, NEWOP)                               \
-  CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, VV, MF4, E16)                   \
   CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E16)                     \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E32)
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E32)                     \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, VV, E64)
 
 #define CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, TYPE, SEW)             \
   CASE_VFMA_CHANGE_OPCODE_COMMON(OLDOP, NEWOP, TYPE, MF4, SEW)                 \
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index f1159ee633eaa..77b40999fc6a4 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -3287,10 +3287,10 @@ multiclass VPseudoTernaryW_VV<LMULInfo m> {
                                       constraint, /*Commutable*/ 0, TargetConstraintType=3>;
 }
 
-multiclass VPseudoTernaryW_VV_RM<LMULInfo m> {
+multiclass VPseudoTernaryW_VV_RM<LMULInfo m, int sew = 0> {
   defvar constraint = "@earlyclobber $rd";
   defm _VV : VPseudoTernaryWithPolicyRoundingMode<m.wvrclass, m.vrclass, m.vrclass, m,
-                                                  constraint, /* Commutable */ 0,
+                                                  constraint, sew, /* Commutable */ 0,
                                                   TargetConstraintType=3>;
 }
 
@@ -3306,11 +3306,11 @@ multiclass VPseudoTernaryW_VF<LMULInfo m, FPR_Info f, int TargetConstraintType =
                                               m.vrclass, m, constraint, /*Commutable*/ 0, TargetConstraintType>;
 }
 
-multiclass VPseudoTernaryW_VF_RM<LMULInfo m, FPR_Info f> {
+multiclass VPseudoTernaryW_VF_RM<LMULInfo m, FPR_Info f, int sew = 0> {
   defvar constraint = "@earlyclobber $rd";
   defm "_V" # f.FX : VPseudoTernaryWithPolicyRoundingMode<m.wvrclass, f.fprclass,
                                                           m.vrclass, m, constraint,
-                                                          /* Commutable */ 0,
+                                                          sew, /* Commutable */ 0,
                                                           TargetConstraintType=3>;
 }
 
@@ -3396,16 +3396,17 @@ multiclass VPseudoVWMAC_VX {
 
 multiclass VPseudoVWMAC_VV_VF_RM {
   foreach m = MxListFW in {
-    defm "" : VPseudoTernaryW_VV_RM<m>,
-              SchedTernary<"WriteVFWMulAddV", "ReadVFWMulAddV",
-                           "ReadVFWMulAddV", "ReadVFWMulAddV", m.MX>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm "" : VPseudoTernaryW_VV_RM<m, sew=e>,
+                SchedTernary<"WriteVFWMulAddV", "ReadVFWMulAddV",
+                             "ReadVFWMulAddV", "ReadVFWMulAddV", m.MX, e>;
   }
 
   foreach f = FPListW in {
     foreach m = f.MxListFW in {
-      defm "" : VPseudoTernaryW_VF_RM<m, f>,
+      defm "" : VPseudoTernaryW_VF_RM<m, f, sew=f.SEW>,
                 SchedTernary<"WriteVFWMulAddF", "ReadVFWMulAddV",
-                             "ReadVFWMulAddF", "ReadVFWMulAddV", m.MX>;
+                             "ReadVFWMulAddF", "ReadVFWMulAddV", m.MX, f.SEW>;
     }
   }
 }
@@ -3413,17 +3414,18 @@ multiclass VPseudoVWMAC_VV_VF_RM {
 multiclass VPseudoVWMAC_VV_VF_BF_RM {
   foreach m = MxListFW in {
     defvar mx = m.MX;
-    defm "" : VPseudoTernaryW_VV_RM<m>,
-              SchedTernary<"WriteVFWMulAddV", "ReadVFWMulAddV",
-                           "ReadVFWMulAddV", "ReadVFWMulAddV", mx>;
+    foreach e = SchedSEWSet<mx, isF=1, isWidening=1>.val in
+      defm "" : VPseudoTernaryW_VV_RM<m, sew=e>,
+                SchedTernary<"WriteVFWMulAddV", "ReadVFWMulAddV",
+                             "ReadVFWMulAddV", "ReadVFWMulAddV", mx, e>;
   }
 
   foreach f = BFPListW in {
     foreach m = f.MxListFW in {
       defvar mx = m.MX;
-      defm "" : VPseudoTernaryW_VF_RM<m, f>,
+      defm "" : VPseudoTernaryW_VF_RM<m, f, sew=f.SEW>,
                 SchedTernary<"WriteVFWMulAddF", "ReadVFWMulAddV",
-                             "ReadVFWMulAddF", "ReadVFWMulAddV", mx>;
+                             "ReadVFWMulAddF", "ReadVFWMulAddV", mx, f.SEW>;
     }
   }
 }
@@ -5703,7 +5705,7 @@ multiclass VPatTernaryW_VV<string intrinsic, string instruction,
 }
 
 multiclass VPatTernaryW_VV_RM<string intrinsic, string instruction,
-                           list<VTypeInfoToWide> vtilist> {
+                           list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach vtiToWti = vtilist in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
@@ -5712,7 +5714,8 @@ multiclass VPatTernaryW_VV_RM<string intrinsic, string instruction,
     defm : VPatTernaryWithPolicyRoundingMode<intrinsic, instruction, "VV",
                                              wti.Vector, vti.Vector, vti.Vector,
                                              vti.Mask, vti.Log2SEW, vti.LMul,
-                                             wti.RegClass, vti.RegClass, vti.RegClass>;
+                                             wti.RegClass, vti.RegClass,
+                                             vti.RegClass, isSEWAware>;
   }
 }
 
@@ -5731,19 +5734,18 @@ multiclass VPatTernaryW_VX<string intrinsic, string instruction,
   }
 }
 
-multiclass VPatTernaryW_VX_RM<string intrinsic, string instruction,
-                           list<VTypeInfoToWide> vtilist> {
+multiclass
+    VPatTernaryW_VX_RM<string intrinsic, string instruction,
+                       list<VTypeInfoToWide> vtilist, bit isSEWAware = 0> {
   foreach vtiToWti = vtilist in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
-                                 GetVTypePredicates<wti>.Predicates) in
-    defm : VPatTernaryWithPolicyRoundingMode<intrinsic, instruction,
-                                             "V"#vti.ScalarSuffix,
-                                             wti.Vector, vti.Scalar, vti.Vector,
-                                             vti.Mask, vti.Log2SEW, vti.LMul,
-                                             wti.RegClass, vti.ScalarRegClass,
-                                             vti.RegClass>;
+                                 GetVTypePredicates<wti>.Predicates) in defm
+        : VPatTernaryWithPolicyRoundingMode<
+              intrinsic, instruction, "V" #vti.ScalarSuffix, wti.Vector,
+              vti.Scalar, vti.Vector, vti.Mask, vti.Log2SEW, vti.LMul,
+              wti.RegClass, vti.ScalarRegClass, vti.RegClass, isSEWAware>;
   }
 }
 
@@ -5775,9 +5777,9 @@ multiclass VPatTernaryW_VV_VX<string intrinsic, string instruction,
       VPatTernaryW_VX<intrinsic, instruction, vtilist>;
 
 multiclass VPatTernaryW_VV_VX_RM<string intrinsic, string instruction,
-                              list<VTypeInfoToWide> vtilist>
-    : VPatTernaryW_VV_RM<intrinsic, instruction, vtilist>,
-      VPatTernaryW_VX_RM<intrinsic, instruction, vtilist>;
+                              list<VTypeInfoToWide> vtilist, bit isSEWAware = 1>
+    : VPatTernaryW_VV_RM<intrinsic, instruction, vtilist, isSEWAware>,
+      VPatTernaryW_VX_RM<intrinsic, instruction, vtilist, isSEWAware>;
 
 multiclass VPatBinaryM_VV_VX<string intrinsic, string instruction,
                              list<VTypeInfo> vtilist>
@@ -7154,16 +7156,16 @@ defm : VPatTernaryV_VV_VX_AAXA_RM<"int_riscv_vfnmsub", "PseudoVFNMSUB",
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
 defm : VPatTernaryW_VV_VX_RM<"int_riscv_vfwmacc", "PseudoVFWMACC",
-                             AllWidenableFloatVectors>;
+                             AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatTernaryW_VV_VX_RM<"int_riscv_vfwnmacc", "PseudoVFWNMACC",
-                             AllWidenableFloatVectors>;
+                             AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatTernaryW_VV_VX_RM<"int_riscv_vfwmsac", "PseudoVFWMSAC",
-                             AllWidenableFloatVectors>;
+                             AllWidenableFloatVectors, isSEWAware=1>;
 defm : VPatTernaryW_VV_VX_RM<"int_riscv_vfwnmsac", "PseudoVFWNMSAC",
-                             AllWidenableFloatVectors>;
+                             AllWidenableFloatVectors, isSEWAware=1>;
 let Predicates = [HasStdExtZvfbfwma] in
 defm : VPatTernaryW_VV_VX_RM<"int_riscv_vfwmaccbf16", "PseudoVFWMACCBF16",
-                              AllWidenableBFloatToFloatVectors>;
+                              AllWidenableBFloatToFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.8. Vector Floating-Point Square-Root Instruction
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index d9dc373a628d6..7c77449b4f6e1 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -706,6 +706,7 @@ multiclass VPatWidenFPMulAccSDNode_VV_VF_RM<string instruction_name> {
   foreach vtiToWti = AllWidenableFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
+    defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
                                  GetVTypePredicates<wti>.Predicates) in {
       def : Pat<(fma (wti.Vector (riscv_fpextend_vl_oneuse
@@ -715,7 +716,7 @@ multiclass VPatWidenFPMulAccSDNode_VV_VF_RM<string instruction_name> {
                                       (vti.Vector vti.RegClass:$rs2),
                                       (vti.Mask true_mask), (XLenVT srcvalue))),
                      (wti.Vector wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_VV_"#suffix)
                    wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -727,7 +728,7 @@ multiclass VPatWidenFPMulAccSDNode_VV_VF_RM<string instruction_name> {
                                       (vti.Vector vti.RegClass:$rs2),
                                       (vti.Mask true_mask), (XLenVT srcvalue))),
                      (wti.Vector wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -741,6 +742,7 @@ multiclass VPatWidenFPNegMulAccSDNode_VV_VF_RM<string instruction_name> {
   foreach vtiToWti = AllWidenableFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
+    defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
                                  GetVTypePredicates<wti>.Predicates) in {
       def : Pat<(fma (fneg (wti.Vector (riscv_fpextend_vl_oneuse
@@ -749,7 +751,7 @@ multiclass VPatWidenFPNegMulAccSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      (fneg wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_VV_"#suffix)
                    wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -760,7 +762,7 @@ multiclass VPatWidenFPNegMulAccSDNode_VV_VF_RM<string instruction_name> {
                                             (vti.Vector vti.RegClass:$rs2),
                                             (vti.Mask true_mask), (XLenVT srcvalue)))),
                      (fneg wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -770,7 +772,7 @@ multiclass VPatWidenFPNegMulAccSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      (fneg wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -784,6 +786,7 @@ multiclass VPatWidenFPMulSacSDNode_VV_VF_RM<string instruction_name> {
   foreach vtiToWti = AllWidenableFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
+    defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
                                  GetVTypePredicates<wti>.Predicates) in {
       def : Pat<(fma (wti.Vector (riscv_fpextend_vl_oneuse
@@ -792,7 +795,7 @@ multiclass VPatWidenFPMulSacSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      (fneg wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_VV_"#suffix)
                    wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -802,7 +805,7 @@ multiclass VPatWidenFPMulSacSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      (fneg wti.RegClass:$rd)),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -816,6 +819,7 @@ multiclass VPatWidenFPNegMulSacSDNode_VV_VF_RM<string instruction_name> {
   foreach vtiToWti = AllWidenableFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
+    defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
                                  GetVTypePredicates<wti>.Predicates) in {
       def : Pat<(fma (fneg (wti.Vector (riscv_fpextend_vl_oneuse
@@ -824,7 +828,7 @@ multiclass VPatWidenFPNegMulSacSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      wti.RegClass:$rd),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_VV_"#suffix)
                    wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -835,7 +839,7 @@ multiclass VPatWidenFPNegMulSacSDNode_VV_VF_RM<string instruction_name> {
                                             (vti.Vector vti.RegClass:$rs2),
                                             (vti.Mask true_mask), (XLenVT srcvalue)))),
                      wti.RegClass:$rd),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
@@ -845,7 +849,7 @@ multiclass VPatWidenFPNegMulSacSDNode_VV_VF_RM<string instruction_name> {
                      (riscv_fpextend_vl_oneuse (vti.Vector vti.RegClass:$rs2),
                                                (vti.Mask true_mask), (XLenVT srcvalue)),
                      wti.RegClass:$rd),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix)
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    // Value to indicate no rounding mode change in
                    // RISCVInsertReadWriteCSR
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index a6f24438daa60..322c055306e86 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1994,13 +1994,14 @@ multiclass VPatWidenFPMulAccVL_VV_VF_RM<SDNode vop, string instruction_name> {
   foreach vtiToWti = AllWidenableFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
+    defvar suffix = vti.LMul.MX # "_E" # vti.SEW;
     let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
                                  GetVTypePredicates<wti>.Predicates) in {
       def : Pat<(vop (vti.Vector vti.RegClass:$rs1),
                      (vti.Vector vti.RegClass:$rs2),
                      (wti.Vector wti.RegClass:$rd), (vti.Mask V0),
                      VLOpFrag),
-                (!cast<Instruction>(instruction_name#"_VV_"#vti.LMul.MX #"_MASK")
+                (!cast<Instruction>(instruction_name#"_VV_"#suffix#"_MASK")
                    wti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
                    (vti.Mask V0),
                    // Value to indicate no rounding mode change in
@@ -2011,7 +2012,7 @@ multiclass VPatWidenFPMulAccVL_VV_VF_RM<SDNode vop, string instruction_name> {
                      (vti.Vector vti.RegClass:$rs2),
                      (wti.Vector wti.RegClass:$rd), (vti.Mask V0),
                      VLOpFrag),
-                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#vti.LMul.MX #"_MASK")
+                (!cast<Instruction>(instruction_name#"_V"#vti.ScalarSuffix#"_"#suffix#"_MASK")
                    wti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                    (vti.Mask V0),
                    // Value to indicate no rounding mode change in
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 16bfb509e0856..1dd4afa4ee48b 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -799,6 +799,8 @@ foreach mx = SchedMxListFW in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -806,10 +808,8 @@ foreach mx = SchedMxListFW in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxListFW>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFWMulAddV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFWCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulAddF",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
 }
 // Narrowing
@@ -1165,8 +1165,8 @@ defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 0896f3358d230..f63dcf20e683d 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -558,17 +558,11 @@ foreach mx = SchedMxListFW in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWALUF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
     }
   }
 }
-foreach mx = SchedMxListFW in {
-  defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
-  defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxListFW>.c;
-  let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFWMulAddV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWMulAddF", [SiFiveP600VectorArith], mx, IsWorstCase>;
-  }
-}
 // Narrowing
 foreach mx = SchedMxListW in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
@@ -961,8 +955,8 @@ defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 40090d1db8874..b82e86e9fcd58 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -426,8 +426,8 @@ defm "" : LMULSEWSchedWritesFW<"WriteVFWMulF">;
 defm "" : LMULSEWSchedWritesF<"WriteVFMulAddV">;
 defm "" : LMULSEWSchedWritesF<"WriteVFMulAddF">;
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
-defm "" : LMULSchedWritesFW<"WriteVFWMulAddV">;
-defm "" : LMULSchedWritesFW<"WriteVFWMulAddF">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWMulAddV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWMulAddF">;
 // 13.8. Vector Floating-Point Square-Root Instruction
 defm "" : LMULSEWSchedWritesF<"WriteVFSqrtV">;
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
@@ -651,8 +651,8 @@ defm "" : LMULSEWSchedReadsFW<"ReadVFWMulF">;
 defm "" : LMULSEWSchedReadsF<"ReadVFMulAddV">;
 defm "" : LMULSEWSchedReadsF<"ReadVFMulAddF">;
 // 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
-defm "" : LMULSchedReadsFW<"ReadVFWMulAddV">;
-defm "" : LMULSchedReadsFW<"ReadVFWMulAddF">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWMulAddV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWMulAddF">;
 // 13.8. Vector Floating-Point Square-Root Instruction
 defm "" : LMULSEWSchedReadsF<"ReadVFSqrtV">;
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
@@ -892,8 +892,8 @@ defm "" : LMULSEWWriteResFW<"WriteVFWMulV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFMulAddV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFMulAddF", []>;
-defm "" : LMULWriteResFW<"WriteVFWMulAddV", []>;
-defm "" : LMULWriteResFW<"WriteVFWMulAddF", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWMulAddV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWMulAddF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFSqrtV", []>;
 defm "" : LMULWriteRes<"WriteVFRecpV", []>;
 defm "" : LMULWriteRes<"WriteVFMinMaxV", []>;
@@ -1048,8 +1048,8 @@ defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWMulAddF", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fmf.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fmf.ll
index 1c1440c3e6d0e..a4851e9838fbf 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fmf.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fmf.ll
@@ -9,8 +9,8 @@ define <2 x double> @foo(<2 x double> %x, <2 x double> %y) {
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:vr = COPY $v9
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:vr = COPY $v8
-  ; CHECK-NEXT:   [[PseudoVFADD_VV_M1_:%[0-9]+]]:vr = nnan ninf nsz arcp contract afn reassoc nofpexcept PseudoVFADD_VV_M1 $noreg, [[COPY1]], [[COPY]], 7, 2, 6 /* e64 */, 1 /* ta, mu */, implicit $frm
-  ; CHECK-NEXT:   $v8 = COPY [[PseudoVFADD_VV_M1_]]
+  ; CHECK-NEXT:   [[PseudoVFADD_VV_M1_E64_:%[0-9]+]]:vr = nnan ninf nsz arcp contract afn reassoc nofpexcept PseudoVFADD_VV_M1_E64 $noreg, [[COPY1]], [[COPY]], 7, 2, 6 /* e64 */, 1 /* ta, mu */, implicit $frm
+  ; CHECK-NEXT:   $v8 = COPY [[PseudoVFADD_VV_M1_E64_]]
   ; CHECK-NEXT:   PseudoRET implicit $v8
   %1 = fadd fast <2 x double> %x, %y
   ret <2 x double> %1
diff --git a/llvm/test/CodeGen/RISCV/rvv/pass-fast-math-flags-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/pass-fast-math-flags-sdnode.ll
index 1ad8e2d66392a..8457f3d2c149c 100644
--- a/llvm/test/CodeGen/RISCV/rvv/pass-fast-math-flags-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/pass-fast-math-flags-sdnode.ll
@@ -15,8 +15,8 @@ define <vscale x 1 x double> @foo(<vscale x 1 x double> %x, <vscale x 1 x double
   ; CHECK-NEXT:   [[SLLI:%[0-9]+]]:gpr = SLLI [[COPY]], 32
   ; CHECK-NEXT:   [[SRLI:%[0-9]+]]:gprnox0 = SRLI killed [[SLLI]], 32
   ; CHECK-NEXT:   $v0 = COPY [[COPY1]]
-  ; CHECK-NEXT:   [[PseudoVFMUL_VV_M1_MASK:%[0-9]+]]:vrnov0 = nnan ninf nsz arcp contract afn reassoc nofpexcept PseudoVFMUL_VV_M1_MASK $noreg, [[COPY3]], [[COPY2]], $v0, 7, killed [[SRLI]], 6 /* e64 */, 1 /* ta, mu */, implicit $frm
-  ; CHECK-NEXT:   $v8 = COPY [[PseudoVFMUL_VV_M1_MASK]]
+  ; CHECK-NEXT:   [[PseudoVFMUL_VV_M1_E64_MASK:%[0-9]+]]:vrnov0 = nnan ninf nsz arcp contract afn reassoc nofpexcept PseudoVFMUL_VV_M1_E64_MASK $noreg, [[COPY3]], [[COPY2]], $v0, 7, killed [[SRLI]], 6 /* e64 */, 1 /* ta, mu */, implicit $frm
+  ; CHECK-NEXT:   $v8 = COPY [[PseudoVFMUL_VV_M1_E64_MASK]]
   ; CHECK-NEXT:   PseudoRET implicit $v8
   %1 = call fast <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> %x, <vscale x 1 x double> %y, <vscale x 1 x i1> %m, i32 %vl)
   ret <vscale x 1 x double> %1

From b8e1ff322104e5445371328a4fbc9c1d2bf82f01 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Thu, 4 Apr 2024 12:16:39 -0700
Subject: [PATCH 253/886] [RISCV] Split PseudoVFRSQRT7 and PseudoVFREC7 by SEW

Co-authored-by: Wang Pengcheng <wangpengcheng.pp@bytedance.com>
---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 62 +++++++++++--------
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |  4 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 11 +++-
 llvm/lib/Target/RISCV/RISCVScheduleV.td       |  8 +--
 4 files changed, 52 insertions(+), 33 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index 77b40999fc6a4..a5f56a1c209c6 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2577,14 +2577,16 @@ multiclass VPseudoVSQR_V_RM {
 multiclass VPseudoVRCP_V {
   foreach m = MxListF in {
     defvar mx = m.MX;
-    let VLMul = m.value in {
-      def "_V_" # mx
-          : VPseudoUnaryNoMask<m.vrclass, m.vrclass>,
-            SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, forceMergeOpRead=true>;
-      def "_V_" # mx # "_MASK"
-          : VPseudoUnaryMask<m.vrclass, m.vrclass>,
-            RISCVMaskedPseudo<MaskIdx = 2>,
-            SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<mx, isF=1>.val in {
+      let VLMul = m.value in {
+        def "_V_" # mx # "_E" # e
+            : VPseudoUnaryNoMask<m.vrclass, m.vrclass>,
+              SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, e, forceMergeOpRead=true>;
+        def "_V_" # mx # "_E" # e # "_MASK"
+            : VPseudoUnaryMask<m.vrclass, m.vrclass>,
+              RISCVMaskedPseudo<MaskIdx = 2>,
+              SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, e, forceMergeOpRead=true>;
+      }
     }
   }
 }
@@ -2592,14 +2594,16 @@ multiclass VPseudoVRCP_V {
 multiclass VPseudoVRCP_V_RM {
   foreach m = MxListF in {
     defvar mx = m.MX;
-    let VLMul = m.value in {
-      def "_V_" # mx
-          : VPseudoUnaryNoMaskRoundingMode<m.vrclass, m.vrclass>,
-            SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, forceMergeOpRead=true>;
-      def "_V_" # mx # "_MASK"
-          : VPseudoUnaryMaskRoundingMode<m.vrclass, m.vrclass>,
-            RISCVMaskedPseudo<MaskIdx = 2>,
-            SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<mx, isF=1>.val in {
+      let VLMul = m.value in {
+        def "_V_" # mx # "_E" # e
+            : VPseudoUnaryNoMaskRoundingMode<m.vrclass, m.vrclass>,
+              SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, e, forceMergeOpRead=true>;
+        def "_V_" # mx # "_E" # e # "_MASK"
+            : VPseudoUnaryMaskRoundingMode<m.vrclass, m.vrclass>,
+              RISCVMaskedPseudo<MaskIdx = 2>,
+              SchedUnary<"WriteVFRecpV", "ReadVFRecpV", mx, e, forceMergeOpRead=true>;
+      }
     }
   }
 }
@@ -3939,12 +3943,16 @@ class VPatUnaryNoMask<string intrinsic_name,
                       int log2sew,
                       LMULInfo vlmul,
                       VReg result_reg_class,
-                      VReg op2_reg_class> :
+                      VReg op2_reg_class,
+                      bit isSEWAware = 0> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic_name)
                    (result_type result_reg_class:$merge),
                    (op2_type op2_reg_class:$rs2),
                    VLOpFrag)),
-                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX)
+                   (!cast<Instruction>(
+                     !if(isSEWAware,
+                         inst#"_"#kind#"_"#vlmul.MX#"_E"#!shl(1, log2sew),
+                         inst#"_"#kind#"_"#vlmul.MX))
                    (result_type result_reg_class:$merge),
                    (op2_type op2_reg_class:$rs2),
                    GPR:$vl, log2sew, TU_MU)>;
@@ -3983,13 +3991,17 @@ class VPatUnaryMask<string intrinsic_name,
                     int log2sew,
                     LMULInfo vlmul,
                     VReg result_reg_class,
-                    VReg op2_reg_class> :
+                    VReg op2_reg_class,
+                    bit isSEWAware = 0> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic_name#"_mask")
                    (result_type result_reg_class:$merge),
                    (op2_type op2_reg_class:$rs2),
                    (mask_type V0),
                    VLOpFrag, (XLenVT timm:$policy))),
-                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX#"_MASK")
+                   (!cast<Instruction>(
+                      !if(isSEWAware,
+                          inst#"_"#kind#"_"#vlmul.MX#"_E"#!shl(1, log2sew)#"_MASK",
+                          inst#"_"#kind#"_"#vlmul.MX#"_MASK"))
                    (result_type result_reg_class:$merge),
                    (op2_type op2_reg_class:$rs2),
                    (mask_type V0), GPR:$vl, log2sew, (XLenVT timm:$policy))>;
@@ -4681,15 +4693,15 @@ multiclass VPatUnaryV_VF<string intrinsic, string instruction, string suffix,
 }
 
 multiclass VPatUnaryV_V<string intrinsic, string instruction,
-                        list<VTypeInfo> vtilist> {
+                        list<VTypeInfo> vtilist, bit isSEWAware = 0> {
   foreach vti = vtilist in {
     let Predicates = GetVTypePredicates<vti>.Predicates in {
       def : VPatUnaryNoMask<intrinsic, instruction, "V",
                             vti.Vector, vti.Vector, vti.Log2SEW,
-                            vti.LMul, vti.RegClass, vti.RegClass>;
+                            vti.LMul, vti.RegClass, vti.RegClass, isSEWAware>;
       def : VPatUnaryMask<intrinsic, instruction, "V",
                           vti.Vector, vti.Vector, vti.Mask, vti.Log2SEW,
-                          vti.LMul, vti.RegClass, vti.RegClass>;
+                          vti.LMul, vti.RegClass, vti.RegClass, isSEWAware>;
     }
   }
 }
@@ -7175,12 +7187,12 @@ defm : VPatUnaryV_V_RM<"int_riscv_vfsqrt", "PseudoVFSQRT", AllFloatVectors, isSE
 //===----------------------------------------------------------------------===//
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
 //===----------------------------------------------------------------------===//
-defm : VPatUnaryV_V<"int_riscv_vfrsqrt7", "PseudoVFRSQRT7", AllFloatVectors>;
+defm : VPatUnaryV_V<"int_riscv_vfrsqrt7", "PseudoVFRSQRT7", AllFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.10. Vector Floating-Point Reciprocal Estimate Instruction
 //===----------------------------------------------------------------------===//
-defm : VPatUnaryV_V_RM<"int_riscv_vfrec7", "PseudoVFREC7", AllFloatVectors>;
+defm : VPatUnaryV_V_RM<"int_riscv_vfrec7", "PseudoVFREC7", AllFloatVectors, isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.11. Vector Floating-Point Min/Max Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 1dd4afa4ee48b..2a13cb4beac14 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -743,6 +743,7 @@ foreach mx = SchedMxListF in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF",  [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV",   [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -750,7 +751,6 @@ foreach mx = SchedMxList in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFRecpV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtIToFV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtFToIV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
@@ -1168,7 +1168,7 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
-defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxF", 0>;
 defm "" : LMULReadAdvance<"ReadVFSgnjV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index f63dcf20e683d..80090a0027149 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -510,6 +510,14 @@ foreach mx = SchedMxListF in {
     }
   }
 }
+foreach mx = SchedMxListF in {
+  foreach sew = SchedSEWSet<mx, isF=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
+    let Latency = 2, ReleaseAtCycles = [LMulLat] in
+    defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+  }
+}
 foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
@@ -520,7 +528,6 @@ foreach mx = SchedMxList in {
   let Latency = 2, ReleaseAtCycles = [LMulLat] in {
     defm "" : LMULWriteResMX<"WriteVFCmpV",  [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCmpF",  [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFRecpV", [SiFiveP600VectorArith], mx, IsWorstCase>;
   }
   let Latency = 1, ReleaseAtCycles = [LMulLat] in {
     defm "" : LMULWriteResMX<"WriteVFSgnjV",   [SiFiveP600VectorArith], mx, IsWorstCase>;
@@ -958,7 +965,7 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
-defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
+defm "" : LMULSEWReadAdvance<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index b82e86e9fcd58..6070482ea1757 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -432,7 +432,7 @@ defm "" : LMULSEWSchedWritesFW<"WriteVFWMulAddF">;
 defm "" : LMULSEWSchedWritesF<"WriteVFSqrtV">;
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
 // 13.10. Vector Floating-Point Reciprocal Estimate Instruction
-defm "" : LMULSchedWrites<"WriteVFRecpV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFRecpV">;
 // 13.11. Vector Floating-Point MIN/MAX Instructions
 defm "" : LMULSchedWrites<"WriteVFMinMaxV">;
 defm "" : LMULSchedWrites<"WriteVFMinMaxF">;
@@ -657,7 +657,7 @@ defm "" : LMULSEWSchedReadsFW<"ReadVFWMulAddF">;
 defm "" : LMULSEWSchedReadsF<"ReadVFSqrtV">;
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
 // 13.10. Vector Floating-Point Reciprocal Estimate Instruction
-defm "" : LMULSchedReads<"ReadVFRecpV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFRecpV">;
 // 13.11. Vector Floating-Point MIN/MAX Instructions
 defm "" : LMULSchedReads<"ReadVFMinMaxV">;
 defm "" : LMULSchedReads<"ReadVFMinMaxF">;
@@ -895,7 +895,7 @@ defm "" : LMULSEWWriteResF<"WriteVFMulAddF", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulAddV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulAddF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFSqrtV", []>;
-defm "" : LMULWriteRes<"WriteVFRecpV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFRecpV", []>;
 defm "" : LMULWriteRes<"WriteVFMinMaxV", []>;
 defm "" : LMULWriteRes<"WriteVFMinMaxF", []>;
 defm "" : LMULWriteRes<"WriteVFSgnjV", []>;
@@ -1051,7 +1051,7 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
-defm "" : LMULReadAdvance<"ReadVFRecpV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMinMaxF", 0>;
 defm "" : LMULReadAdvance<"ReadVFSgnjV", 0>;

From f4e3226a6c2bd89247192e4aae395a8deb3868b1 Mon Sep 17 00:00:00 2001
From: Xiaoyang Liu <siujoeng.lau@gmail.com>
Date: Fri, 12 Apr 2024 10:09:05 -0400
Subject: [PATCH 254/886] [libc++][ranges] LWG3736: move_iterator missing
 disable_sized_sentinel_for specialization (#85611)

This pull request implements LWG3736: move_iterator missing
disable_sized_sentinel_for specialization.
---
 libcxx/docs/Status/Cxx23Issues.csv            |  2 +-
 libcxx/include/__iterator/move_iterator.h     |  6 ++++
 libcxx/include/iterator                       |  5 +++
 .../sized_sentinel.compile.pass.cpp           | 32 +++++++++++++++++++
 4 files changed, 44 insertions(+), 1 deletion(-)
 create mode 100644 libcxx/test/std/iterators/predef.iterators/move.iterators/sized_sentinel.compile.pass.cpp

diff --git a/libcxx/docs/Status/Cxx23Issues.csv b/libcxx/docs/Status/Cxx23Issues.csv
index fa82e67ac7db7..04f6e23375acd 100644
--- a/libcxx/docs/Status/Cxx23Issues.csv
+++ b/libcxx/docs/Status/Cxx23Issues.csv
@@ -201,7 +201,7 @@
 "`3677 <https://wg21.link/LWG3677>`__","Is a cv-qualified ``pair`` specially handled in uses-allocator construction?", "November 2022","|Complete|","18.0",""
 "`3717 <https://wg21.link/LWG3717>`__","``common_view::end`` should improve ``random_access_range`` case", "November 2022","","","|ranges|"
 "`3732 <https://wg21.link/LWG3732>`__","``prepend_range`` and ``append_range`` can't be amortized constant time", "November 2022","|Nothing to do|","","|ranges|"
-"`3736 <https://wg21.link/LWG3736>`__","``move_iterator`` missing ``disable_sized_sentinel_for`` specialization", "November 2022","","","|ranges|"
+"`3736 <https://wg21.link/LWG3736>`__","``move_iterator`` missing ``disable_sized_sentinel_for`` specialization", "November 2022","|Complete|","19.0","|ranges|"
 "`3737 <https://wg21.link/LWG3737>`__","``take_view::sentinel`` should provide ``operator-``", "November 2022","","","|ranges|"
 "`3738 <https://wg21.link/LWG3738>`__","Missing preconditions for ``take_view`` constructor", "November 2022","|Complete|","16.0","|ranges|"
 "`3743 <https://wg21.link/LWG3743>`__","``ranges::to``'s reserve may be ill-formed", "November 2022","","","|ranges|"
diff --git a/libcxx/include/__iterator/move_iterator.h b/libcxx/include/__iterator/move_iterator.h
index eefd5b3748454..6382c947060bf 100644
--- a/libcxx/include/__iterator/move_iterator.h
+++ b/libcxx/include/__iterator/move_iterator.h
@@ -330,6 +330,12 @@ operator+(typename move_iterator<_Iter>::difference_type __n, const move_iterato
 }
 #endif // _LIBCPP_STD_VER >= 20
 
+#if _LIBCPP_STD_VER >= 20
+template <class _Iter1, class _Iter2>
+  requires(!sized_sentinel_for<_Iter1, _Iter2>)
+inline constexpr bool disable_sized_sentinel_for<move_iterator<_Iter1>, move_iterator<_Iter2>> = true;
+#endif // _LIBCPP_STD_VER >= 20
+
 template <class _Iter>
 inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX17 move_iterator<_Iter> make_move_iterator(_Iter __i) {
   return move_iterator<_Iter>(std::move(__i));
diff --git a/libcxx/include/iterator b/libcxx/include/iterator
index 5779bf828711b..1b9e7eaf0c1e8 100644
--- a/libcxx/include/iterator
+++ b/libcxx/include/iterator
@@ -475,6 +475,11 @@ constexpr move_iterator<Iterator> operator+(   // constexpr in C++17
 template <class Iterator>   // constexpr in C++17
 constexpr  move_iterator<Iterator> make_move_iterator(const Iterator& i);
 
+template<class Iterator1, class Iterator2>
+    requires (!sized_sentinel_for<Iterator1, Iterator2>)
+  inline constexpr bool disable_sized_sentinel_for<move_iterator<Iterator1>,    // since C++20
+                                                    move_iterator<Iterator2>> = true;
+
 template<semiregular S>
 class move_sentinel {
 public:
diff --git a/libcxx/test/std/iterators/predef.iterators/move.iterators/sized_sentinel.compile.pass.cpp b/libcxx/test/std/iterators/predef.iterators/move.iterators/sized_sentinel.compile.pass.cpp
new file mode 100644
index 0000000000000..cb49086dd6802
--- /dev/null
+++ b/libcxx/test/std/iterators/predef.iterators/move.iterators/sized_sentinel.compile.pass.cpp
@@ -0,0 +1,32 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+#include <iterator>
+
+#include "test_iterators.h"
+
+using sized_it = random_access_iterator<int*>;
+static_assert(std::sized_sentinel_for<sized_it, sized_it>);
+static_assert(std::sized_sentinel_for<std::move_iterator<sized_it>, std::move_iterator<sized_it>>);
+
+struct unsized_it {
+  using value_type      = int;
+  using difference_type = std::ptrdiff_t;
+
+  value_type& operator*() const;
+  bool operator==(const unsized_it&) const;
+  difference_type operator-(const unsized_it&) const { return 0; }
+};
+
+template <>
+inline constexpr bool std::disable_sized_sentinel_for<unsized_it, unsized_it> = true;
+
+static_assert(!std::sized_sentinel_for<unsized_it, unsized_it>);
+static_assert(!std::sized_sentinel_for<std::move_iterator<unsized_it>, std::move_iterator<unsized_it>>);

From 9d8ccbc1be78fb4ba0ad1c6ead2dc04b24c66668 Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Fri, 12 Apr 2024 16:13:54 +0200
Subject: [PATCH 255/886] [lldb][test] Add tests for evaluating local variables
 whose name clashes with Objective-C types (#87807)

Depends on https://github.com/llvm/llvm-project/pull/87767
---
 .../objc-builtin-types/TestObjCBuiltinTypes.py      | 13 +++++++++++++
 .../API/lang/objcxx/objc-builtin-types/main.cpp     |  8 +++++++-
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py b/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
index 611c738899905..3cdca31b8969b 100644
--- a/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
+++ b/lldb/test/API/lang/objcxx/objc-builtin-types/TestObjCBuiltinTypes.py
@@ -13,6 +13,7 @@ def setUp(self):
         # Find the line numbers to break inside main().
         self.main_source = "main.cpp"
         self.break_line = line_number(self.main_source, "// Set breakpoint here.")
+        self.bar_break_line = line_number(self.main_source, "return id + Class")
 
     @add_test_categories(["pyapi"])
     def test_with_python_api(self):
@@ -26,6 +27,11 @@ def test_with_python_api(self):
         bpt = target.BreakpointCreateByLocation(self.main_source, self.break_line)
         self.assertTrue(bpt, VALID_BREAKPOINT)
 
+        bar_bpt = target.BreakpointCreateByLocation(
+            self.main_source, self.bar_break_line
+        )
+        self.assertTrue(bar_bpt, VALID_BREAKPOINT)
+
         # Now launch the process, and do not stop at entry point.
         process = target.LaunchSimple(None, None, self.get_process_working_directory())
 
@@ -52,3 +58,10 @@ def test_with_python_api(self):
             patterns=["\(id\) \$.* = nil"],
         )
         self.expect("expr --language C++ -- id my_id = 0; my_id", error=True)
+
+        lldbutil.continue_to_breakpoint(process, bar_bpt)
+
+        self.expect_expr("id", result_value="12", result_type="int")
+        self.expect_expr("Class", result_value="15", result_type="int")
+        self.expect("expr --language Objective-C++ -- id", error=True)
+        self.expect("expr --language Objective-C++ -- Class", error=True)
diff --git a/lldb/test/API/lang/objcxx/objc-builtin-types/main.cpp b/lldb/test/API/lang/objcxx/objc-builtin-types/main.cpp
index 6dd8cbc6e9fef..5b35ec0f0b8c9 100644
--- a/lldb/test/API/lang/objcxx/objc-builtin-types/main.cpp
+++ b/lldb/test/API/lang/objcxx/objc-builtin-types/main.cpp
@@ -2,8 +2,14 @@ namespace ns {
   typedef int id;
 };
 
+int bar() {
+  int id = 12;
+  int Class = 15;
+  return id + Class;
+}
+
 int main()
 {
   ns::id foo = 0;
-  return foo; // Set breakpoint here.
+  return foo + bar(); // Set breakpoint here.
 }

From bcf047a4ed9afc49f851156698a6d277ab2c00ea Mon Sep 17 00:00:00 2001
From: Bjorn Pettersson <bjorn.a.pettersson@ericsson.com>
Date: Wed, 3 Apr 2024 19:00:17 +0200
Subject: [PATCH 256/886] [ARM][PowerPC] Add regression tests for narrowing
 load in TargetLowering::SimplifySetCC

These test cases show some miscomplies for big-endian when dealing
with non byte-sized loads. One part of the problem is that LLVM IR
isn't really telling where the padding goes for non byte-sized
loads/stores. So currently TargetLowering::SimplifySetCC can't assume
anything about it. But the implementation also do not consider that
the TypeStoreSize could be larger than the TypeSize, resulting in
the offset calculation being wrong for big-endian.

Pre-commit for https://github.com/llvm/llvm-project/pull/87646
---
 .../CodeGen/ARM/simplifysetcc_narrow_load.ll  | 665 ++++++++++++++++++
 .../PowerPC/simplifysetcc_narrow_load.ll      | 390 ++++++++++
 2 files changed, 1055 insertions(+)
 create mode 100644 llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
 create mode 100644 llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll

diff --git a/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll b/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
new file mode 100644
index 0000000000000..2cba4b46f9a8a
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
@@ -0,0 +1,665 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -O1 -mtriple arm -o - %s | FileCheck --check-prefix CHECK-LE %s
+; RUN: llc -O1 -mtriple armv7 -o - %s | FileCheck --check-prefix CHECK-V7-LE %s
+; RUN: llc -O1 -mtriple armeb -o - %s | FileCheck --check-prefix CHECK-BE %s
+; RUN: llc -O1 -mtriple armv7eb -o - %s | FileCheck --check-prefix CHECK-V7-BE %s
+
+; A collection of regression tests to verify the load-narrowing part of
+; TargetLowering::SimplifySetCC (and/or other similar rewrites such as
+; combining AND+LOAD into ZEXTLOAD).
+;
+; Using both arm and armv7 to show that alignment restrictions are
+; considered for the narrowed load (armv7 is a bit more relaxed when it
+; comes to unaligned memory accesses).
+
+;--------------------------------------------------------------------------
+; Test non byte-sized types.
+;
+; As long as LLVM IR isn't defining where the padding goes we can't really
+; optimize these (without adding a target lowering hook that can inform
+; ISel about which bits are padding).
+; --------------------------------------------------------------------------
+
+define i1 @test_129_15_0(ptr %y) {
+; CHECK-LE-LABEL: test_129_15_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0]
+; CHECK-LE-NEXT:    mov r1, #255
+; CHECK-LE-NEXT:    orr r1, r1, #32512
+; CHECK-LE-NEXT:    ands r0, r0, r1
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_129_15_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-LE-NEXT:    bfc r0, #15, #17
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_129_15_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #14]
+; CHECK-BE-NEXT:    mov r1, #255
+; CHECK-BE-NEXT:    orr r1, r1, #32512
+; CHECK-BE-NEXT:    ands r0, r0, r1
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_129_15_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #14]
+; CHECK-V7-BE-NEXT:    bfc r0, #15, #17
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i129, ptr %y
+  %b = and i129 %a, u0x7fff
+  %cmp = icmp ne i129 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_126_20_4(ptr %y) {
+; CHECK-LE-LABEL: test_126_20_4:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldr r0, [r0]
+; CHECK-LE-NEXT:    mvn r1, #15
+; CHECK-LE-NEXT:    sub r1, r1, #-16777216
+; CHECK-LE-NEXT:    ands r0, r0, r1
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_126_20_4:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldr r0, [r0]
+; CHECK-V7-LE-NEXT:    movw r1, #65520
+; CHECK-V7-LE-NEXT:    movt r1, #255
+; CHECK-V7-LE-NEXT:    ands r0, r0, r1
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_126_20_4:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldr r0, [r0, #12]
+; CHECK-BE-NEXT:    mvn r1, #15
+; CHECK-BE-NEXT:    sub r1, r1, #-16777216
+; CHECK-BE-NEXT:    ands r0, r0, r1
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_126_20_4:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldr r0, [r0, #12]
+; CHECK-V7-BE-NEXT:    movw r1, #65520
+; CHECK-V7-BE-NEXT:    movt r1, #255
+; CHECK-V7-BE-NEXT:    ands r0, r0, r1
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i126, ptr %y
+  %b = and i126 %a, u0xfffff0
+  %cmp = icmp ne i126 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_8_0(ptr %y) {
+; CHECK-LE-LABEL: test_33_8_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_33_8_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_33_8_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_33_8_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0xff
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_32(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_32:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0, #4]
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_33_1_32:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0, #4]
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_33_1_32:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldr r0, [r0]
+; CHECK-BE-NEXT:    lsr r0, r0, #24
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_33_1_32:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldr r0, [r0]
+; CHECK-V7-BE-NEXT:    lsr r0, r0, #24
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x100000000
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_31(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_31:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-LE-NEXT:    lsr r0, r0, #7
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_33_1_31:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-V7-LE-NEXT:    lsr r0, r0, #7
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_33_1_31:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    lsr r0, r0, #7
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_33_1_31:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-BE-NEXT:    lsr r0, r0, #7
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x80000000
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_0(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0]
+; CHECK-LE-NEXT:    and r0, r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_33_1_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-LE-NEXT:    and r0, r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_33_1_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-BE-NEXT:    and r0, r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_33_1_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    and r0, r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x1
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+;--------------------------------------------------------------------------
+; Test byte-sized types.
+;--------------------------------------------------------------------------
+
+
+define i1 @test_128_20_4(ptr %y) {
+; CHECK-LE-LABEL: test_128_20_4:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldr r0, [r0]
+; CHECK-LE-NEXT:    mvn r1, #15
+; CHECK-LE-NEXT:    sub r1, r1, #-16777216
+; CHECK-LE-NEXT:    ands r0, r0, r1
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_128_20_4:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldr r0, [r0]
+; CHECK-V7-LE-NEXT:    movw r1, #65520
+; CHECK-V7-LE-NEXT:    movt r1, #255
+; CHECK-V7-LE-NEXT:    ands r0, r0, r1
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_128_20_4:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldr r0, [r0, #12]
+; CHECK-BE-NEXT:    mvn r1, #15
+; CHECK-BE-NEXT:    sub r1, r1, #-16777216
+; CHECK-BE-NEXT:    ands r0, r0, r1
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_128_20_4:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldr r0, [r0, #12]
+; CHECK-V7-BE-NEXT:    movw r1, #65520
+; CHECK-V7-BE-NEXT:    movt r1, #255
+; CHECK-V7-BE-NEXT:    ands r0, r0, r1
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i128, ptr %y
+  %b = and i128 %a, u0xfffff0
+  %cmp = icmp ne i128 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_48_16_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_48_16_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_48_16_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_8(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_8:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0, #1]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_48_16_8:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0, #1]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_48_16_8:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #3]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_48_16_8:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff00
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_16(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_16:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0, #2]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_48_16_16:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0, #2]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_48_16_16:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #2]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_48_16_16:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #2]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff0000
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_32(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_32:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_48_16_32:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_48_16_32:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_48_16_32:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff00000000
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_17_0(ptr %y) {
+; CHECK-LE-LABEL: test_48_17_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldr r0, [r0]
+; CHECK-LE-NEXT:    ldr r1, .LCPI11_0
+; CHECK-LE-NEXT:    ands r0, r0, r1
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+; CHECK-LE-NEXT:    .p2align 2
+; CHECK-LE-NEXT:  @ %bb.1:
+; CHECK-LE-NEXT:  .LCPI11_0:
+; CHECK-LE-NEXT:    .long 131071 @ 0x1ffff
+;
+; CHECK-V7-LE-LABEL: test_48_17_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldr r0, [r0]
+; CHECK-V7-LE-NEXT:    bfc r0, #17, #15
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_48_17_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldr r1, [r0]
+; CHECK-BE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-BE-NEXT:    orr r0, r0, r1, lsl #16
+; CHECK-BE-NEXT:    ldr r1, .LCPI11_0
+; CHECK-BE-NEXT:    ands r0, r0, r1
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+; CHECK-BE-NEXT:    .p2align 2
+; CHECK-BE-NEXT:  @ %bb.1:
+; CHECK-BE-NEXT:  .LCPI11_0:
+; CHECK-BE-NEXT:    .long 131071 @ 0x1ffff
+;
+; CHECK-V7-BE-LABEL: test_48_17_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldr r1, [r0]
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #4]
+; CHECK-V7-BE-NEXT:    orr r0, r0, r1, lsl #16
+; CHECK-V7-BE-NEXT:    bfc r0, #17, #15
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0x1ffff
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_40_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_40_16_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_40_16_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_40_16_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #3]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_40_16_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i40, ptr %y
+  %b = and i40 %a, u0xffff
+  %cmp = icmp ne i40 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_40_1_32(ptr %y) {
+; CHECK-LE-LABEL: test_40_1_32:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0, #4]
+; CHECK-LE-NEXT:    and r0, r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_40_1_32:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0, #4]
+; CHECK-V7-LE-NEXT:    and r0, r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_40_1_32:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldr r0, [r0]
+; CHECK-BE-NEXT:    mov r1, #1
+; CHECK-BE-NEXT:    and r0, r1, r0, lsr #24
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_40_1_32:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldr r0, [r0]
+; CHECK-V7-BE-NEXT:    ubfx r0, r0, #24, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i40, ptr %y
+  %b = and i40 %a, u0x100000000
+  %cmp = icmp ne i40 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_24_16_0:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrh r0, [r0]
+; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_24_16_0:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-LE-NEXT:    cmp r0, #0
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_24_16_0:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0, #1]
+; CHECK-BE-NEXT:    cmp r0, #0
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_24_16_0:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #1]
+; CHECK-V7-BE-NEXT:    cmp r0, #0
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xffff
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_8(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_8:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0, #1]
+; CHECK-LE-NEXT:    lsls r0, r0, #8
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_24_8_8:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0, #1]
+; CHECK-V7-LE-NEXT:    lsls r0, r0, #8
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_24_8_8:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrb r0, [r0, #1]
+; CHECK-BE-NEXT:    lsls r0, r0, #8
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_24_8_8:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #1]
+; CHECK-V7-BE-NEXT:    lsls r0, r0, #8
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff00
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_12(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_12:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r1, [r0, #2]
+; CHECK-LE-NEXT:    ldrh r0, [r0]
+; CHECK-LE-NEXT:    orr r0, r0, r1, lsl #16
+; CHECK-LE-NEXT:    ands r0, r0, #1044480
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_24_8_12:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r1, [r0, #2]
+; CHECK-V7-LE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-LE-NEXT:    orr r0, r0, r1, lsl #16
+; CHECK-V7-LE-NEXT:    ands r0, r0, #1044480
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_24_8_12:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrh r0, [r0]
+; CHECK-BE-NEXT:    mov r1, #1044480
+; CHECK-BE-NEXT:    ands r0, r1, r0, lsl #8
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_24_8_12:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0]
+; CHECK-V7-BE-NEXT:    mov r1, #1044480
+; CHECK-V7-BE-NEXT:    ands r0, r1, r0, lsl #8
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff000
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_16(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_16:
+; CHECK-LE:       @ %bb.0:
+; CHECK-LE-NEXT:    ldrb r0, [r0, #2]
+; CHECK-LE-NEXT:    lsls r0, r0, #16
+; CHECK-LE-NEXT:    movne r0, #1
+; CHECK-LE-NEXT:    mov pc, lr
+;
+; CHECK-V7-LE-LABEL: test_24_8_16:
+; CHECK-V7-LE:       @ %bb.0:
+; CHECK-V7-LE-NEXT:    ldrb r0, [r0, #2]
+; CHECK-V7-LE-NEXT:    lsls r0, r0, #16
+; CHECK-V7-LE-NEXT:    movwne r0, #1
+; CHECK-V7-LE-NEXT:    bx lr
+;
+; CHECK-BE-LABEL: test_24_8_16:
+; CHECK-BE:       @ %bb.0:
+; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    lsls r0, r0, #16
+; CHECK-BE-NEXT:    movne r0, #1
+; CHECK-BE-NEXT:    mov pc, lr
+;
+; CHECK-V7-BE-LABEL: test_24_8_16:
+; CHECK-V7-BE:       @ %bb.0:
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-BE-NEXT:    lsls r0, r0, #16
+; CHECK-V7-BE-NEXT:    movwne r0, #1
+; CHECK-V7-BE-NEXT:    bx lr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff0000
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
diff --git a/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll b/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll
new file mode 100644
index 0000000000000..1a03423fe6aec
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll
@@ -0,0 +1,390 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -O1 -mtriple ppc32le -o - %s | FileCheck --check-prefix CHECK-LE %s
+; RUN: llc -O1 -mtriple ppc32 -o - %s | FileCheck --check-prefix CHECK-BE %s
+
+; A collection of regression tests to verify the load-narrowing part of
+; TargetLowering::SimplifySetCC (and/or other similar rewrites such as
+; combining AND+LOAD into ZEXTLOAD).
+
+
+;--------------------------------------------------------------------------
+; Test non byte-sized types.
+;
+; As long as LLVM IR isn't defining where the padding goes we can't really
+; optimize these (without adding a target lowering hook that can inform
+; ISel about which bits are padding).
+; --------------------------------------------------------------------------
+
+define i1 @test_129_15_0(ptr %y) {
+; CHECK-LE-LABEL: test_129_15_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 0(3)
+; CHECK-LE-NEXT:    clrlwi 3, 3, 17
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_129_15_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 14(3)
+; CHECK-BE-NEXT:    clrlwi 3, 3, 17
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i129, ptr %y
+  %b = and i129 %a, u0x7fff
+  %cmp = icmp ne i129 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_126_20_4(ptr %y) {
+; CHECK-LE-LABEL: test_126_20_4:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lwz 3, 0(3)
+; CHECK-LE-NEXT:    rlwinm 3, 3, 0, 8, 27
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_126_20_4:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lwz 3, 12(3)
+; CHECK-BE-NEXT:    rlwinm 3, 3, 0, 8, 27
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i126, ptr %y
+  %b = and i126 %a, u0xfffff0
+  %cmp = icmp ne i126 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_8_0(ptr %y) {
+; CHECK-LE-LABEL: test_33_8_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 0(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_33_8_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lbz 3, 3(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0xff
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_32(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_32:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 4(3)
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_33_1_32:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lwz 3, 0(3)
+; CHECK-BE-NEXT:    srwi 3, 3, 24
+; CHECK-BE-NEXT:    blr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x100000000
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_31(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_31:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 3(3)
+; CHECK-LE-NEXT:    srwi 3, 3, 7
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_33_1_31:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lbz 3, 0(3)
+; CHECK-BE-NEXT:    srwi 3, 3, 7
+; CHECK-BE-NEXT:    blr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x80000000
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_33_1_0(ptr %y) {
+; CHECK-LE-LABEL: test_33_1_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 0(3)
+; CHECK-LE-NEXT:    clrlwi 3, 3, 31
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_33_1_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lbz 3, 3(3)
+; CHECK-BE-NEXT:    clrlwi 3, 3, 31
+; CHECK-BE-NEXT:    blr
+  %a = load i33, ptr %y
+  %b = and i33 %a, u0x1
+  %cmp = icmp ne i33 %b, 0
+  ret i1 %cmp
+}
+
+;--------------------------------------------------------------------------
+; Test byte-sized types.
+;--------------------------------------------------------------------------
+
+
+define i1 @test_128_20_4(ptr %y) {
+; CHECK-LE-LABEL: test_128_20_4:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lwz 3, 0(3)
+; CHECK-LE-NEXT:    rlwinm 3, 3, 0, 8, 27
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_128_20_4:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lwz 3, 12(3)
+; CHECK-BE-NEXT:    rlwinm 3, 3, 0, 8, 27
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i128, ptr %y
+  %b = and i128 %a, u0xfffff0
+  %cmp = icmp ne i128 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 0(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_48_16_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 4(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_8(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_8:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 1(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_48_16_8:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 3(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff00
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_16(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_16:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 2(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_48_16_16:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 2(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff0000
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_16_32(ptr %y) {
+; CHECK-LE-LABEL: test_48_16_32:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 4(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_48_16_32:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 0(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0xffff00000000
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_48_17_0(ptr %y) {
+; CHECK-LE-LABEL: test_48_17_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lwz 3, 0(3)
+; CHECK-LE-NEXT:    clrlwi 3, 3, 15
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_48_17_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 4, 4(3)
+; CHECK-BE-NEXT:    lwz 3, 0(3)
+; CHECK-BE-NEXT:    clrlwi 4, 4, 16
+; CHECK-BE-NEXT:    rlwimi 4, 3, 16, 15, 15
+; CHECK-BE-NEXT:    addic 3, 4, -1
+; CHECK-BE-NEXT:    subfe 3, 3, 4
+; CHECK-BE-NEXT:    blr
+  %a = load i48, ptr %y
+  %b = and i48 %a, u0x1ffff
+  %cmp = icmp ne i48 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_40_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_40_16_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 0(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_40_16_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 3(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i40, ptr %y
+  %b = and i40 %a, u0xffff
+  %cmp = icmp ne i40 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_40_1_32(ptr %y) {
+; CHECK-LE-LABEL: test_40_1_32:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 4(3)
+; CHECK-LE-NEXT:    clrlwi 3, 3, 31
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_40_1_32:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lwz 3, 0(3)
+; CHECK-BE-NEXT:    rlwinm 3, 3, 8, 31, 31
+; CHECK-BE-NEXT:    blr
+  %a = load i40, ptr %y
+  %b = and i40 %a, u0x100000000
+  %cmp = icmp ne i40 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_16_0(ptr %y) {
+; CHECK-LE-LABEL: test_24_16_0:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 3, 0(3)
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_24_16_0:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 1(3)
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xffff
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_8(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_8:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 1(3)
+; CHECK-LE-NEXT:    slwi 3, 3, 8
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_24_8_8:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lbz 3, 1(3)
+; CHECK-BE-NEXT:    slwi 3, 3, 8
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff00
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_12(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_12:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lhz 4, 0(3)
+; CHECK-LE-NEXT:    lbz 3, 2(3)
+; CHECK-LE-NEXT:    rlwinm 4, 4, 0, 16, 19
+; CHECK-LE-NEXT:    rlwimi 4, 3, 16, 12, 15
+; CHECK-LE-NEXT:    addic 3, 4, -1
+; CHECK-LE-NEXT:    subfe 3, 3, 4
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_24_8_12:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lhz 3, 0(3)
+; CHECK-BE-NEXT:    rlwinm 3, 3, 8, 12, 19
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff000
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}
+
+define i1 @test_24_8_16(ptr %y) {
+; CHECK-LE-LABEL: test_24_8_16:
+; CHECK-LE:       # %bb.0:
+; CHECK-LE-NEXT:    lbz 3, 2(3)
+; CHECK-LE-NEXT:    slwi 3, 3, 16
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
+; CHECK-LE-NEXT:    blr
+;
+; CHECK-BE-LABEL: test_24_8_16:
+; CHECK-BE:       # %bb.0:
+; CHECK-BE-NEXT:    lbz 3, 0(3)
+; CHECK-BE-NEXT:    slwi 3, 3, 16
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
+; CHECK-BE-NEXT:    blr
+  %a = load i24, ptr %y
+  %b = and i24 %a, u0xff0000
+  %cmp = icmp ne i24 %b, 0
+  ret i1 %cmp
+}

From 33e6b488beda80dddb68016a132ee1f0a3c04aa1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bj=C3=B6rn=20Pettersson?= <bjorn.a.pettersson@ericsson.com>
Date: Fri, 12 Apr 2024 16:18:12 +0200
Subject: [PATCH 257/886] [SelectionDAG] Fix and improve
 TargetLowering::SimplifySetCC (#87646)

The load narrowing part of TargetLowering::SimplifySetCC is updated
according to this:

1) The offset calculation (for big endian) did not work properly for
   non byte-sized types. This is basically solved by an early exit
   if the memory type isn't byte-sized. But the code is also corrected
   to use the store size when calculating the offset.
2) To still allow some optimizations for non-byte-sized types the
   TargetLowering::isPaddedAtMostSignificantBitsWhenStored hook is
   added. By default it assumes that scalar integer types are padded
   starting at the most significant bits, if the type needs padding
   when being stored to memory.
3) Allow optimizing when isPaddedAtMostSignificantBitsWhenStored is
   true, as that hook makes it possible for TargetLowering to know
   how the non byte-sized value is aligned in memory.
4) Update the algorithm to always search for a narrowed load with
   a power-of-2 byte-sized type. In the past the algorithm started
   with the the width of the original load, and then divided it by
   two for each iteration. But for a type such as i48 that would
   just end up trying to narrow the load into a i24 or i12 load,
   and then we would fail sooner or later due to not finding a
   newVT that fulfilled newVT.isRound().
   With this new approach we can narrow the i48 load into either
   an i8, i16 or i32 load. By checking if such a load is allowed
(e.g. alignment wise) for any "multiple of 8 offset", then we can find
   more opportunities for the optimization to trigger. So even for a
   byte-sized type such as i32 we may now end up narrowing the load
   into loading the 16 bits starting at offset 8 (if that is allowed
   by the target). The old algorithm did not even consider that case.
5) Also start using getObjectPtrOffset instead of getMemBasePlusOffset
   when creating the new ptr. This way we get "nsw" on the add.
---
 llvm/include/llvm/CodeGen/TargetLowering.h    |  7 ++
 .../CodeGen/SelectionDAG/TargetLowering.cpp   | 68 +++++++++++--------
 .../CodeGen/ARM/simplifysetcc_narrow_load.ll  | 33 +++++----
 .../PowerPC/simplifysetcc_narrow_load.ll      | 38 ++++-------
 4 files changed, 79 insertions(+), 67 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index a4dc097446186..980d75ad91a0d 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -1804,6 +1804,13 @@ class TargetLoweringBase {
   /// where the sext is redundant, and use x directly.
   virtual bool shouldRemoveRedundantExtend(SDValue Op) const { return true; }
 
+  /// Indicates if any padding is guaranteed to go at the most significant bits
+  /// when storing the type to memory and the type size isn't equal to the store
+  /// size.
+  bool isPaddedAtMostSignificantBitsWhenStored(EVT VT) const {
+    return VT.isScalarInteger() && !VT.isByteSized();
+  }
+
   /// When splitting a value of the specified type into parts, does the Lo
   /// or Hi part come first?  This usually follows the endianness, except
   /// for ppcf128, where the Hi part always comes first.
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 409d66adfd67d..4939976b8e287 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -4621,7 +4621,10 @@ SDValue TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
       LoadSDNode *Lod = cast<LoadSDNode>(N0.getOperand(0));
       APInt bestMask;
       unsigned bestWidth = 0, bestOffset = 0;
-      if (Lod->isSimple() && Lod->isUnindexed()) {
+      if (Lod->isSimple() && Lod->isUnindexed() &&
+          (Lod->getMemoryVT().isByteSized() ||
+           isPaddedAtMostSignificantBitsWhenStored(Lod->getMemoryVT()))) {
+        unsigned memWidth = Lod->getMemoryVT().getStoreSizeInBits();
         unsigned origWidth = N0.getValueSizeInBits();
         unsigned maskWidth = origWidth;
         // We can narrow (e.g.) 16-bit extending loads on 32-bit target to
@@ -4629,40 +4632,51 @@ SDValue TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
         if (Lod->getExtensionType() != ISD::NON_EXTLOAD)
           origWidth = Lod->getMemoryVT().getSizeInBits();
         const APInt &Mask = N0.getConstantOperandAPInt(1);
-        for (unsigned width = origWidth / 2; width>=8; width /= 2) {
+        // Only consider power-of-2 widths (and at least one byte) as candiates
+        // for the narrowed load.
+        for (unsigned width = 8; width < origWidth; width *= 2) {
+          EVT newVT = EVT::getIntegerVT(*DAG.getContext(), width);
+          if (!shouldReduceLoadWidth(Lod, ISD::NON_EXTLOAD, newVT))
+            continue;
           APInt newMask = APInt::getLowBitsSet(maskWidth, width);
-          for (unsigned offset=0; offset<origWidth/width; offset++) {
+          // Avoid accessing any padding here for now (we could use memWidth
+          // instead of origWidth here otherwise).
+          unsigned maxOffset = origWidth - width;
+          for (unsigned offset = 0; offset <= maxOffset; offset += 8) {
             if (Mask.isSubsetOf(newMask)) {
-              if (Layout.isLittleEndian())
-                bestOffset = (uint64_t)offset * (width/8);
-              else
-                bestOffset = (origWidth/width - offset - 1) * (width/8);
-              bestMask = Mask.lshr(offset * (width/8) * 8);
-              bestWidth = width;
-              break;
+              unsigned ptrOffset =
+                  Layout.isLittleEndian() ? offset : memWidth - width - offset;
+              unsigned IsFast = 0;
+              Align NewAlign = commonAlignment(Lod->getAlign(), ptrOffset / 8);
+              if (allowsMemoryAccess(
+                      *DAG.getContext(), Layout, newVT, Lod->getAddressSpace(),
+                      NewAlign, Lod->getMemOperand()->getFlags(), &IsFast) &&
+                  IsFast) {
+                bestOffset = ptrOffset / 8;
+                bestMask = Mask.lshr(offset);
+                bestWidth = width;
+                break;
+              }
             }
-            newMask <<= width;
+            newMask <<= 8;
           }
+          if (bestWidth)
+            break;
         }
       }
       if (bestWidth) {
         EVT newVT = EVT::getIntegerVT(*DAG.getContext(), bestWidth);
-        if (newVT.isRound() &&
-            shouldReduceLoadWidth(Lod, ISD::NON_EXTLOAD, newVT)) {
-          SDValue Ptr = Lod->getBasePtr();
-          if (bestOffset != 0)
-            Ptr = DAG.getMemBasePlusOffset(Ptr, TypeSize::getFixed(bestOffset),
-                                           dl);
-          SDValue NewLoad =
-              DAG.getLoad(newVT, dl, Lod->getChain(), Ptr,
-                          Lod->getPointerInfo().getWithOffset(bestOffset),
-                          Lod->getOriginalAlign());
-          return DAG.getSetCC(dl, VT,
-                              DAG.getNode(ISD::AND, dl, newVT, NewLoad,
-                                      DAG.getConstant(bestMask.trunc(bestWidth),
-                                                      dl, newVT)),
-                              DAG.getConstant(0LL, dl, newVT), Cond);
-        }
+        SDValue Ptr = Lod->getBasePtr();
+        if (bestOffset != 0)
+          Ptr = DAG.getObjectPtrOffset(dl, Ptr, TypeSize::getFixed(bestOffset));
+        SDValue NewLoad =
+            DAG.getLoad(newVT, dl, Lod->getChain(), Ptr,
+                        Lod->getPointerInfo().getWithOffset(bestOffset),
+                        Lod->getOriginalAlign());
+        SDValue And =
+            DAG.getNode(ISD::AND, dl, newVT, NewLoad,
+                        DAG.getConstant(bestMask.trunc(bestWidth), dl, newVT));
+        return DAG.getSetCC(dl, VT, And, DAG.getConstant(0LL, dl, newVT), Cond);
       }
     }
 
diff --git a/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll b/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
index 2cba4b46f9a8a..838da59f9e412 100644
--- a/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
+++ b/llvm/test/CodeGen/ARM/simplifysetcc_narrow_load.ll
@@ -40,7 +40,9 @@ define i1 @test_129_15_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_129_15_0:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    ldrh r0, [r0, #14]
+; CHECK-BE-NEXT:    ldr r1, [r0, #12]
+; CHECK-BE-NEXT:    ldrb r0, [r0, #16]
+; CHECK-BE-NEXT:    orr r0, r0, r1, lsl #8
 ; CHECK-BE-NEXT:    mov r1, #255
 ; CHECK-BE-NEXT:    orr r1, r1, #32512
 ; CHECK-BE-NEXT:    ands r0, r0, r1
@@ -49,7 +51,7 @@ define i1 @test_129_15_0(ptr %y) {
 ;
 ; CHECK-V7-BE-LABEL: test_129_15_0:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #14]
+; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #15]
 ; CHECK-V7-BE-NEXT:    bfc r0, #15, #17
 ; CHECK-V7-BE-NEXT:    cmp r0, #0
 ; CHECK-V7-BE-NEXT:    movwne r0, #1
@@ -119,14 +121,14 @@ define i1 @test_33_8_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_8_0:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-BE-NEXT:    ldrb r0, [r0, #4]
 ; CHECK-BE-NEXT:    cmp r0, #0
 ; CHECK-BE-NEXT:    movne r0, #1
 ; CHECK-BE-NEXT:    mov pc, lr
 ;
 ; CHECK-V7-BE-LABEL: test_33_8_0:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #4]
 ; CHECK-V7-BE-NEXT:    cmp r0, #0
 ; CHECK-V7-BE-NEXT:    movwne r0, #1
 ; CHECK-V7-BE-NEXT:    bx lr
@@ -179,13 +181,13 @@ define i1 @test_33_1_31(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_1_31:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    ldrb r0, [r0, #1]
 ; CHECK-BE-NEXT:    lsr r0, r0, #7
 ; CHECK-BE-NEXT:    mov pc, lr
 ;
 ; CHECK-V7-BE-LABEL: test_33_1_31:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #1]
 ; CHECK-V7-BE-NEXT:    lsr r0, r0, #7
 ; CHECK-V7-BE-NEXT:    bx lr
   %a = load i33, ptr %y
@@ -209,13 +211,13 @@ define i1 @test_33_1_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_1_0:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-BE-NEXT:    ldrb r0, [r0, #4]
 ; CHECK-BE-NEXT:    and r0, r0, #1
 ; CHECK-BE-NEXT:    mov pc, lr
 ;
 ; CHECK-V7-BE-LABEL: test_33_1_0:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #3]
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0, #4]
 ; CHECK-V7-BE-NEXT:    and r0, r0, #1
 ; CHECK-V7-BE-NEXT:    bx lr
   %a = load i33, ptr %y
@@ -309,7 +311,7 @@ define i1 @test_48_16_8(ptr %y) {
 ; CHECK-LE-LABEL: test_48_16_8:
 ; CHECK-LE:       @ %bb.0:
 ; CHECK-LE-NEXT:    ldrh r0, [r0, #1]
-; CHECK-LE-NEXT:    cmp r0, #0
+; CHECK-LE-NEXT:    lsls r0, r0, #8
 ; CHECK-LE-NEXT:    movne r0, #1
 ; CHECK-LE-NEXT:    mov pc, lr
 ;
@@ -444,9 +446,7 @@ define i1 @test_48_17_0(ptr %y) {
 ;
 ; CHECK-V7-BE-LABEL: test_48_17_0:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldr r1, [r0]
-; CHECK-V7-BE-NEXT:    ldrh r0, [r0, #4]
-; CHECK-V7-BE-NEXT:    orr r0, r0, r1, lsl #16
+; CHECK-V7-BE-NEXT:    ldr r0, [r0, #2]
 ; CHECK-V7-BE-NEXT:    bfc r0, #17, #15
 ; CHECK-V7-BE-NEXT:    cmp r0, #0
 ; CHECK-V7-BE-NEXT:    movwne r0, #1
@@ -506,15 +506,14 @@ define i1 @test_40_1_32(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_40_1_32:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    ldr r0, [r0]
-; CHECK-BE-NEXT:    mov r1, #1
-; CHECK-BE-NEXT:    and r0, r1, r0, lsr #24
+; CHECK-BE-NEXT:    ldrb r0, [r0]
+; CHECK-BE-NEXT:    and r0, r0, #1
 ; CHECK-BE-NEXT:    mov pc, lr
 ;
 ; CHECK-V7-BE-LABEL: test_40_1_32:
 ; CHECK-V7-BE:       @ %bb.0:
-; CHECK-V7-BE-NEXT:    ldr r0, [r0]
-; CHECK-V7-BE-NEXT:    ubfx r0, r0, #24, #1
+; CHECK-V7-BE-NEXT:    ldrb r0, [r0]
+; CHECK-V7-BE-NEXT:    and r0, r0, #1
 ; CHECK-V7-BE-NEXT:    bx lr
   %a = load i40, ptr %y
   %b = and i40 %a, u0x100000000
diff --git a/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll b/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll
index 1a03423fe6aec..49b8e2bc2f7b4 100644
--- a/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll
+++ b/llvm/test/CodeGen/PowerPC/simplifysetcc_narrow_load.ll
@@ -26,7 +26,7 @@ define i1 @test_129_15_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_129_15_0:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lhz 3, 14(3)
+; CHECK-BE-NEXT:    lhz 3, 15(3)
 ; CHECK-BE-NEXT:    clrlwi 3, 3, 17
 ; CHECK-BE-NEXT:    addic 4, 3, -1
 ; CHECK-BE-NEXT:    subfe 3, 4, 3
@@ -69,7 +69,7 @@ define i1 @test_33_8_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_8_0:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lbz 3, 3(3)
+; CHECK-BE-NEXT:    lbz 3, 4(3)
 ; CHECK-BE-NEXT:    addic 4, 3, -1
 ; CHECK-BE-NEXT:    subfe 3, 4, 3
 ; CHECK-BE-NEXT:    blr
@@ -105,7 +105,7 @@ define i1 @test_33_1_31(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_1_31:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lbz 3, 0(3)
+; CHECK-BE-NEXT:    lbz 3, 1(3)
 ; CHECK-BE-NEXT:    srwi 3, 3, 7
 ; CHECK-BE-NEXT:    blr
   %a = load i33, ptr %y
@@ -123,7 +123,7 @@ define i1 @test_33_1_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_33_1_0:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lbz 3, 3(3)
+; CHECK-BE-NEXT:    lbz 3, 4(3)
 ; CHECK-BE-NEXT:    clrlwi 3, 3, 31
 ; CHECK-BE-NEXT:    blr
   %a = load i33, ptr %y
@@ -250,12 +250,10 @@ define i1 @test_48_17_0(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_48_17_0:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lhz 4, 4(3)
-; CHECK-BE-NEXT:    lwz 3, 0(3)
-; CHECK-BE-NEXT:    clrlwi 4, 4, 16
-; CHECK-BE-NEXT:    rlwimi 4, 3, 16, 15, 15
-; CHECK-BE-NEXT:    addic 3, 4, -1
-; CHECK-BE-NEXT:    subfe 3, 3, 4
+; CHECK-BE-NEXT:    lwz 3, 2(3)
+; CHECK-BE-NEXT:    clrlwi 3, 3, 15
+; CHECK-BE-NEXT:    addic 4, 3, -1
+; CHECK-BE-NEXT:    subfe 3, 4, 3
 ; CHECK-BE-NEXT:    blr
   %a = load i48, ptr %y
   %b = and i48 %a, u0x1ffff
@@ -292,8 +290,8 @@ define i1 @test_40_1_32(ptr %y) {
 ;
 ; CHECK-BE-LABEL: test_40_1_32:
 ; CHECK-BE:       # %bb.0:
-; CHECK-BE-NEXT:    lwz 3, 0(3)
-; CHECK-BE-NEXT:    rlwinm 3, 3, 8, 31, 31
+; CHECK-BE-NEXT:    lbz 3, 0(3)
+; CHECK-BE-NEXT:    clrlwi 3, 3, 31
 ; CHECK-BE-NEXT:    blr
   %a = load i40, ptr %y
   %b = and i40 %a, u0x100000000
@@ -325,7 +323,6 @@ define i1 @test_24_8_8(ptr %y) {
 ; CHECK-LE-LABEL: test_24_8_8:
 ; CHECK-LE:       # %bb.0:
 ; CHECK-LE-NEXT:    lbz 3, 1(3)
-; CHECK-LE-NEXT:    slwi 3, 3, 8
 ; CHECK-LE-NEXT:    addic 4, 3, -1
 ; CHECK-LE-NEXT:    subfe 3, 4, 3
 ; CHECK-LE-NEXT:    blr
@@ -333,7 +330,6 @@ define i1 @test_24_8_8(ptr %y) {
 ; CHECK-BE-LABEL: test_24_8_8:
 ; CHECK-BE:       # %bb.0:
 ; CHECK-BE-NEXT:    lbz 3, 1(3)
-; CHECK-BE-NEXT:    slwi 3, 3, 8
 ; CHECK-BE-NEXT:    addic 4, 3, -1
 ; CHECK-BE-NEXT:    subfe 3, 4, 3
 ; CHECK-BE-NEXT:    blr
@@ -346,18 +342,16 @@ define i1 @test_24_8_8(ptr %y) {
 define i1 @test_24_8_12(ptr %y) {
 ; CHECK-LE-LABEL: test_24_8_12:
 ; CHECK-LE:       # %bb.0:
-; CHECK-LE-NEXT:    lhz 4, 0(3)
-; CHECK-LE-NEXT:    lbz 3, 2(3)
-; CHECK-LE-NEXT:    rlwinm 4, 4, 0, 16, 19
-; CHECK-LE-NEXT:    rlwimi 4, 3, 16, 12, 15
-; CHECK-LE-NEXT:    addic 3, 4, -1
-; CHECK-LE-NEXT:    subfe 3, 3, 4
+; CHECK-LE-NEXT:    lhz 3, 1(3)
+; CHECK-LE-NEXT:    rlwinm 3, 3, 0, 20, 27
+; CHECK-LE-NEXT:    addic 4, 3, -1
+; CHECK-LE-NEXT:    subfe 3, 4, 3
 ; CHECK-LE-NEXT:    blr
 ;
 ; CHECK-BE-LABEL: test_24_8_12:
 ; CHECK-BE:       # %bb.0:
 ; CHECK-BE-NEXT:    lhz 3, 0(3)
-; CHECK-BE-NEXT:    rlwinm 3, 3, 8, 12, 19
+; CHECK-BE-NEXT:    rlwinm 3, 3, 0, 20, 27
 ; CHECK-BE-NEXT:    addic 4, 3, -1
 ; CHECK-BE-NEXT:    subfe 3, 4, 3
 ; CHECK-BE-NEXT:    blr
@@ -371,7 +365,6 @@ define i1 @test_24_8_16(ptr %y) {
 ; CHECK-LE-LABEL: test_24_8_16:
 ; CHECK-LE:       # %bb.0:
 ; CHECK-LE-NEXT:    lbz 3, 2(3)
-; CHECK-LE-NEXT:    slwi 3, 3, 16
 ; CHECK-LE-NEXT:    addic 4, 3, -1
 ; CHECK-LE-NEXT:    subfe 3, 4, 3
 ; CHECK-LE-NEXT:    blr
@@ -379,7 +372,6 @@ define i1 @test_24_8_16(ptr %y) {
 ; CHECK-BE-LABEL: test_24_8_16:
 ; CHECK-BE:       # %bb.0:
 ; CHECK-BE-NEXT:    lbz 3, 0(3)
-; CHECK-BE-NEXT:    slwi 3, 3, 16
 ; CHECK-BE-NEXT:    addic 4, 3, -1
 ; CHECK-BE-NEXT:    subfe 3, 4, 3
 ; CHECK-BE-NEXT:    blr

From 2c2377d3a9305b86ab110a4f8390b2d16bff9510 Mon Sep 17 00:00:00 2001
From: Adam Fowler <adamfowler71@gmail.com>
Date: Fri, 12 Apr 2024 15:39:22 +0100
Subject: [PATCH 258/886] Add lldb-dap to Swift distributions (#88482)

This includes the lldb-dap executable in the MacOS and Linux
distributions of Swift. Currently there is a commit in the Apple repo to
do this for just MacOS https://github.com/apple/llvm-project/pull/8176.
This PR extends this to both Linux and MacOS and brings the change
upstream.

@JDevlieghere @adrian-prantl
---
 lldb/cmake/caches/Apple-lldb-Linux.cmake | 1 +
 lldb/cmake/caches/Apple-lldb-macOS.cmake | 1 +
 2 files changed, 2 insertions(+)

diff --git a/lldb/cmake/caches/Apple-lldb-Linux.cmake b/lldb/cmake/caches/Apple-lldb-Linux.cmake
index bfa660d8654b7..b936929afc42f 100644
--- a/lldb/cmake/caches/Apple-lldb-Linux.cmake
+++ b/lldb/cmake/caches/Apple-lldb-Linux.cmake
@@ -5,6 +5,7 @@ set(LLVM_DISTRIBUTION_COMPONENTS
   lldb
   liblldb
   lldb-argdumper
+  lldb-dap
   lldb-server
   lldb-python-scripts
   CACHE STRING "")
diff --git a/lldb/cmake/caches/Apple-lldb-macOS.cmake b/lldb/cmake/caches/Apple-lldb-macOS.cmake
index 2aef41157bab1..5155024ac3a02 100644
--- a/lldb/cmake/caches/Apple-lldb-macOS.cmake
+++ b/lldb/cmake/caches/Apple-lldb-macOS.cmake
@@ -22,6 +22,7 @@ set(LLVM_DISTRIBUTION_COMPONENTS
   lldb
   liblldb
   lldb-argdumper
+  lldb-dap
   darwin-debug
   debugserver
   CACHE STRING "")

From 986d0dba85f3239671c5151d9817b73de08664a0 Mon Sep 17 00:00:00 2001
From: yronglin <yronglin777@gmail.com>
Date: Fri, 12 Apr 2024 22:39:40 +0800
Subject: [PATCH 259/886] [Clang] Fix AST dump for {CXXDefaultArgExpr,
 CXXDefaultInitExpr} (#88269)

This PR fix a AST dump issue since
https://github.com/llvm/llvm-project/pull/80001

When Clang dumps `CXXDefaultArgExpr`/`CXXDefaultInitExpr`, there has no
recursively dump the complete `CXXDefaultArgExpr`/`CXXDefaultInitExpr`.

Since this PR, Clang will recursively dump a
`CXXDefaultArgExpr`/`CXXDefaultInitExpr` node, even if the node has no
rewritten init.

*Consider*:
```
struct A {
  int arr[1];
};

struct B {
  const A &a = A{{0}};
};

void test() {
  B b{};
}
```

*Before*:
```
`-FunctionDecl <line:9:1, line:11:1> line:9:6 test 'void ()'
  `-CompoundStmt <col:13, line:11:1>
    `-DeclStmt <line:10:3, col:8>
      `-VarDecl <col:3, col:7> col:5 b 'B' listinit
        `-InitListExpr <col:6, col:7> 'B'
          `-CXXDefaultInitExpr <col:7> 'const A' lvalue has rewritten init
            `-ExprWithCleanups <line:6:16, col:21> 'const A' lvalue
```

*After*:
```
`-FunctionDecl 0x15a9455a8 <line:9:1, line:11:1> line:9:6 test 'void ()'
  `-CompoundStmt 0x15a945850 <col:13, line:11:1>
    `-DeclStmt 0x15a945838 <line:10:3, col:8>
      `-VarDecl 0x15a945708 <col:3, col:7> col:5 b 'B' listinit
        `-InitListExpr 0x15a9457b0 <col:6, col:7> 'B'
          `-CXXDefaultInitExpr 0x15a9457f8 <col:7> 'const A' lvalue has rewritten init
            `-ExprWithCleanups 0x15a945568 <line:6:16, col:21> 'const A' lvalue
              `-MaterializeTemporaryExpr 0x15a945500 <col:16, col:21> 'const A' lvalue extended by Field 0x15a945160 'a' 'const A &'
                `-ImplicitCastExpr 0x15a9454e8 <col:16, col:21> 'const A' <NoOp>
                  `-CXXFunctionalCastExpr 0x15a9454c0 <col:16, col:21> 'A' functional cast to A <NoOp>
                    `-InitListExpr 0x15a9452c0 <col:17, col:21> 'A'
                      `-InitListExpr 0x15a945308 <col:18, col:20> 'int[1]'
                        `-IntegerLiteral 0x15a945210 <col:19> 'int' 0
```

---------

Signed-off-by: yronglin <yronglin777@gmail.com>
---
 clang/include/clang/AST/ASTNodeTraverser.h    |    8 +
 clang/include/clang/AST/JSONNodeDumper.h      |    2 +
 clang/lib/AST/JSONNodeDumper.cpp              |    8 +
 clang/lib/AST/TextNodeDumper.cpp              |   14 +-
 clang/test/AST/ast-dump-default-arg-json.cpp  | 2091 +++++++++++++++++
 clang/test/AST/ast-dump-default-init-json.cpp |  929 ++++++++
 clang/test/AST/ast-dump-default-init.cpp      |   21 +
 .../test/AST/ast-dump-for-range-lifetime.cpp  |   21 +
 clang/unittests/AST/ASTTraverserTest.cpp      |    8 +
 9 files changed, 3090 insertions(+), 12 deletions(-)
 create mode 100644 clang/test/AST/ast-dump-default-arg-json.cpp
 create mode 100644 clang/test/AST/ast-dump-default-init-json.cpp
 create mode 100644 clang/test/AST/ast-dump-default-init.cpp

diff --git a/clang/include/clang/AST/ASTNodeTraverser.h b/clang/include/clang/AST/ASTNodeTraverser.h
index 94e7dd817809d..a6e3b05b00596 100644
--- a/clang/include/clang/AST/ASTNodeTraverser.h
+++ b/clang/include/clang/AST/ASTNodeTraverser.h
@@ -932,6 +932,14 @@ class ASTNodeTraverser
       Visit(TArg);
   }
 
+  void VisitCXXDefaultArgExpr(const CXXDefaultArgExpr *Node) {
+    Visit(Node->getExpr());
+  }
+
+  void VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *Node) {
+    Visit(Node->getExpr());
+  }
+
   // Implements Visit methods for Attrs.
 #include "clang/AST/AttrNodeTraverse.inc"
 };
diff --git a/clang/include/clang/AST/JSONNodeDumper.h b/clang/include/clang/AST/JSONNodeDumper.h
index 7a60f362650ca..55bd583e304e8 100644
--- a/clang/include/clang/AST/JSONNodeDumper.h
+++ b/clang/include/clang/AST/JSONNodeDumper.h
@@ -310,6 +310,8 @@ class JSONNodeDumper
   void VisitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *MTE);
   void VisitCXXDependentScopeMemberExpr(const CXXDependentScopeMemberExpr *ME);
   void VisitRequiresExpr(const RequiresExpr *RE);
+  void VisitCXXDefaultArgExpr(const CXXDefaultArgExpr *Node);
+  void VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *Node);
 
   void VisitObjCEncodeExpr(const ObjCEncodeExpr *OEE);
   void VisitObjCMessageExpr(const ObjCMessageExpr *OME);
diff --git a/clang/lib/AST/JSONNodeDumper.cpp b/clang/lib/AST/JSONNodeDumper.cpp
index fb3494393f755..78115200f4f0d 100644
--- a/clang/lib/AST/JSONNodeDumper.cpp
+++ b/clang/lib/AST/JSONNodeDumper.cpp
@@ -1576,6 +1576,14 @@ void JSONNodeDumper::VisitMaterializeTemporaryExpr(
   attributeOnlyIfTrue("boundToLValueRef", MTE->isBoundToLvalueReference());
 }
 
+void JSONNodeDumper::VisitCXXDefaultArgExpr(const CXXDefaultArgExpr *Node) {
+  attributeOnlyIfTrue("hasRewrittenInit", Node->hasRewrittenInit());
+}
+
+void JSONNodeDumper::VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *Node) {
+  attributeOnlyIfTrue("hasRewrittenInit", Node->hasRewrittenInit());
+}
+
 void JSONNodeDumper::VisitCXXDependentScopeMemberExpr(
     const CXXDependentScopeMemberExpr *DSME) {
   JOS.attribute("isArrow", DSME->isArrow());
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 085a7f51ce99a..f4b2df7730ae4 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -1450,23 +1450,13 @@ void TextNodeDumper::VisitExpressionTraitExpr(const ExpressionTraitExpr *Node) {
 }
 
 void TextNodeDumper::VisitCXXDefaultArgExpr(const CXXDefaultArgExpr *Node) {
-  if (Node->hasRewrittenInit()) {
+  if (Node->hasRewrittenInit())
     OS << " has rewritten init";
-    AddChild([=] {
-      ColorScope Color(OS, ShowColors, StmtColor);
-      Visit(Node->getExpr());
-    });
-  }
 }
 
 void TextNodeDumper::VisitCXXDefaultInitExpr(const CXXDefaultInitExpr *Node) {
-  if (Node->hasRewrittenInit()) {
+  if (Node->hasRewrittenInit())
     OS << " has rewritten init";
-    AddChild([=] {
-      ColorScope Color(OS, ShowColors, StmtColor);
-      Visit(Node->getExpr());
-    });
-  }
 }
 
 void TextNodeDumper::VisitMaterializeTemporaryExpr(
diff --git a/clang/test/AST/ast-dump-default-arg-json.cpp b/clang/test/AST/ast-dump-default-arg-json.cpp
new file mode 100644
index 0000000000000..b6a138934caf9
--- /dev/null
+++ b/clang/test/AST/ast-dump-default-arg-json.cpp
@@ -0,0 +1,2091 @@
+// RUN: %clang_cc1 -std=c++23 -triple x86_64-linux-gnu -fsyntax-only -ast-dump=json %s | FileCheck %s
+
+struct S {
+  int arr[1];
+  const int *begin() const { return arr; }
+  const int *end() const { return &arr[1]; }
+  S() {}
+  ~S() {}
+};
+
+S func(const int &, const S &s = S());
+
+void test() {
+  for (auto v : func(1)) {}
+}
+
+// NOTE: CHECK lines have been autogenerated by gen_ast_dump_json_test.py
+
+
+// CHECK-NOT: {{^}}Dumping
+// CHECK:  "kind": "TranslationUnitDecl",
+// CHECK-NEXT:  "loc": {},
+// CHECK-NEXT:  "range": {
+// CHECK-NEXT:   "begin": {},
+// CHECK-NEXT:   "end": {}
+// CHECK-NEXT:  },
+// CHECK-NEXT:  "inner": [
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__int128_t",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__int128"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "BuiltinType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__int128"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__uint128_t",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "unsigned __int128"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "BuiltinType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "unsigned __int128"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__NSConstantString",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__NSConstantString_tag"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "RecordType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__NSConstantString_tag"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "decl": {
+// CHECK-NEXT:       "id": "0x{{.*}}",
+// CHECK-NEXT:       "kind": "CXXRecordDecl",
+// CHECK-NEXT:       "name": "__NSConstantString_tag"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__builtin_ms_va_list",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "char *"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "PointerType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "char *"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "BuiltinType",
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "char"
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__builtin_va_list",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__va_list_tag[1]"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "ConstantArrayType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__va_list_tag[1]"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "size": 1,
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "RecordType",
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "__va_list_tag"
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "decl": {
+// CHECK-NEXT:         "id": "0x{{.*}}",
+// CHECK-NEXT:         "kind": "CXXRecordDecl",
+// CHECK-NEXT:         "name": "__va_list_tag"
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "CXXRecordDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 110,
+// CHECK-NEXT:     "file": "{{.*}}",
+// CHECK-NEXT:     "line": 3,
+// CHECK-NEXT:     "col": 8,
+// CHECK-NEXT:     "tokLen": 1
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 103,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 6
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 235,
+// CHECK-NEXT:      "line": 9,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isReferenced": true,
+// CHECK-NEXT:    "name": "S",
+// CHECK-NEXT:    "tagUsed": "struct",
+// CHECK-NEXT:    "completeDefinition": true,
+// CHECK-NEXT:    "definitionData": {
+// CHECK-NEXT:     "canConstDefaultInit": true,
+// CHECK-NEXT:     "copyAssign": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "copyCtor": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "defaultCtor": {
+// CHECK-NEXT:      "defaultedIsConstexpr": true,
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "nonTrivial": true,
+// CHECK-NEXT:      "userProvided": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "dtor": {
+// CHECK-NEXT:      "nonTrivial": true,
+// CHECK-NEXT:      "userDeclared": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "hasUserDeclaredConstructor": true,
+// CHECK-NEXT:     "isStandardLayout": true,
+// CHECK-NEXT:     "moveAssign": {},
+// CHECK-NEXT:     "moveCtor": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXRecordDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 110,
+// CHECK-NEXT:       "line": 3,
+// CHECK-NEXT:       "col": 8,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 103,
+// CHECK-NEXT:        "col": 1,
+// CHECK-NEXT:        "tokLen": 6
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 110,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isImplicit": true,
+// CHECK-NEXT:      "isReferenced": true,
+// CHECK-NEXT:      "name": "S",
+// CHECK-NEXT:      "tagUsed": "struct"
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "FieldDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 120,
+// CHECK-NEXT:       "line": 4,
+// CHECK-NEXT:       "col": 7,
+// CHECK-NEXT:       "tokLen": 3
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 116,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 3
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 125,
+// CHECK-NEXT:        "col": 12,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isReferenced": true,
+// CHECK-NEXT:      "name": "arr",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "int[1]"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXMethodDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 141,
+// CHECK-NEXT:       "line": 5,
+// CHECK-NEXT:       "col": 14,
+// CHECK-NEXT:       "tokLen": 5
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 130,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 5
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 169,
+// CHECK-NEXT:        "col": 42,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isUsed": true,
+// CHECK-NEXT:      "name": "begin",
+// CHECK-NEXT:      "mangledName": "_ZNK1S5beginEv",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "const int *() const"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "CompoundStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 155,
+// CHECK-NEXT:          "col": 28,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 169,
+// CHECK-NEXT:          "col": 42,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "ReturnStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 157,
+// CHECK-NEXT:            "col": 30,
+// CHECK-NEXT:            "tokLen": 6
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 164,
+// CHECK-NEXT:            "col": 37,
+// CHECK-NEXT:            "tokLen": 3
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "ImplicitCastExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 164,
+// CHECK-NEXT:              "col": 37,
+// CHECK-NEXT:              "tokLen": 3
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 164,
+// CHECK-NEXT:              "col": 37,
+// CHECK-NEXT:              "tokLen": 3
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "castKind": "ArrayToPointerDecay",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "MemberExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 164,
+// CHECK-NEXT:                "col": 37,
+// CHECK-NEXT:                "tokLen": 3
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 164,
+// CHECK-NEXT:                "col": 37,
+// CHECK-NEXT:                "tokLen": 3
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int[1]"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "lvalue",
+// CHECK-NEXT:              "name": "arr",
+// CHECK-NEXT:              "isArrow": true,
+// CHECK-NEXT:              "referencedMemberDecl": "0x{{.*}}",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "CXXThisExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 164,
+// CHECK-NEXT:                  "col": 37,
+// CHECK-NEXT:                  "tokLen": 3
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 164,
+// CHECK-NEXT:                  "col": 37,
+// CHECK-NEXT:                  "tokLen": 3
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "const S *"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "implicit": true
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXMethodDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 184,
+// CHECK-NEXT:       "line": 6,
+// CHECK-NEXT:       "col": 14,
+// CHECK-NEXT:       "tokLen": 3
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 173,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 5
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 214,
+// CHECK-NEXT:        "col": 44,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isUsed": true,
+// CHECK-NEXT:      "name": "end",
+// CHECK-NEXT:      "mangledName": "_ZNK1S3endEv",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "const int *() const"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "CompoundStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 196,
+// CHECK-NEXT:          "col": 26,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 214,
+// CHECK-NEXT:          "col": 44,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "ReturnStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 198,
+// CHECK-NEXT:            "col": 28,
+// CHECK-NEXT:            "tokLen": 6
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 211,
+// CHECK-NEXT:            "col": 41,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "UnaryOperator",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 205,
+// CHECK-NEXT:              "col": 35,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 211,
+// CHECK-NEXT:              "col": 41,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "isPostfix": false,
+// CHECK-NEXT:            "opcode": "&",
+// CHECK-NEXT:            "canOverflow": false,
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "ArraySubscriptExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 206,
+// CHECK-NEXT:                "col": 36,
+// CHECK-NEXT:                "tokLen": 3
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 211,
+// CHECK-NEXT:                "col": 41,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "lvalue",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 206,
+// CHECK-NEXT:                  "col": 36,
+// CHECK-NEXT:                  "tokLen": 3
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 206,
+// CHECK-NEXT:                  "col": 36,
+// CHECK-NEXT:                  "tokLen": 3
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "const int *"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "castKind": "ArrayToPointerDecay",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "MemberExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 206,
+// CHECK-NEXT:                    "col": 36,
+// CHECK-NEXT:                    "tokLen": 3
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 206,
+// CHECK-NEXT:                    "col": 36,
+// CHECK-NEXT:                    "tokLen": 3
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "const int[1]"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "lvalue",
+// CHECK-NEXT:                  "name": "arr",
+// CHECK-NEXT:                  "isArrow": true,
+// CHECK-NEXT:                  "referencedMemberDecl": "0x{{.*}}",
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "CXXThisExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 206,
+// CHECK-NEXT:                      "col": 36,
+// CHECK-NEXT:                      "tokLen": 3
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 206,
+// CHECK-NEXT:                      "col": 36,
+// CHECK-NEXT:                      "tokLen": 3
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "const S *"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "prvalue",
+// CHECK-NEXT:                    "implicit": true
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               },
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "IntegerLiteral",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 210,
+// CHECK-NEXT:                  "col": 40,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 210,
+// CHECK-NEXT:                  "col": 40,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "int"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "value": "1"
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXConstructorDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 218,
+// CHECK-NEXT:       "line": 7,
+// CHECK-NEXT:       "col": 3,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 218,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 223,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isUsed": true,
+// CHECK-NEXT:      "name": "S",
+// CHECK-NEXT:      "mangledName": "_ZN1SC1Ev",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "void ()"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "CompoundStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 222,
+// CHECK-NEXT:          "col": 7,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 223,
+// CHECK-NEXT:          "col": 8,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXDestructorDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 227,
+// CHECK-NEXT:       "line": 8,
+// CHECK-NEXT:       "col": 3,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 227,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 233,
+// CHECK-NEXT:        "col": 9,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isUsed": true,
+// CHECK-NEXT:      "name": "~S",
+// CHECK-NEXT:      "mangledName": "_ZN1SD1Ev",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "void () noexcept"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "CompoundStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 232,
+// CHECK-NEXT:          "col": 8,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 233,
+// CHECK-NEXT:          "col": 9,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXConstructorDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 110,
+// CHECK-NEXT:       "line": 3,
+// CHECK-NEXT:       "col": 8,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 110,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 110,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isImplicit": true,
+// CHECK-NEXT:      "name": "S",
+// CHECK-NEXT:      "mangledName": "_ZN1SC1ERKS_",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "void (const S &)"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inline": true,
+// CHECK-NEXT:      "constexpr": true,
+// CHECK-NEXT:      "explicitlyDefaulted": "default",
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "ParmVarDecl",
+// CHECK-NEXT:        "loc": {
+// CHECK-NEXT:         "offset": 110,
+// CHECK-NEXT:         "col": 8,
+// CHECK-NEXT:         "tokLen": 1
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 110,
+// CHECK-NEXT:          "col": 8,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 110,
+// CHECK-NEXT:          "col": 8,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "const S &"
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "FunctionDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 241,
+// CHECK-NEXT:     "line": 11,
+// CHECK-NEXT:     "col": 3,
+// CHECK-NEXT:     "tokLen": 4
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 239,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 275,
+// CHECK-NEXT:      "col": 37,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isUsed": true,
+// CHECK-NEXT:    "name": "func",
+// CHECK-NEXT:    "mangledName": "_Z4funcRKiRK1S",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "S (const int &, const S &)"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "ParmVarDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 257,
+// CHECK-NEXT:       "col": 19,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 246,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 5
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 256,
+// CHECK-NEXT:        "col": 18,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "const int &"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "ParmVarDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 268,
+// CHECK-NEXT:       "col": 30,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 259,
+// CHECK-NEXT:        "col": 21,
+// CHECK-NEXT:        "tokLen": 5
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 274,
+// CHECK-NEXT:        "col": 36,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "name": "s",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "const S &"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "init": "c",
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "ExprWithCleanups",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 272,
+// CHECK-NEXT:          "col": 34,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 274,
+// CHECK-NEXT:          "col": 36,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "const S"
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "valueCategory": "lvalue",
+// CHECK-NEXT:        "cleanupsHaveSideEffects": true,
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 272,
+// CHECK-NEXT:            "col": 34,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 274,
+// CHECK-NEXT:            "col": 36,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "type": {
+// CHECK-NEXT:           "qualType": "const S"
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "valueCategory": "lvalue",
+// CHECK-NEXT:          "storageDuration": "full expression",
+// CHECK-NEXT:          "boundToLValueRef": true,
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "ImplicitCastExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 272,
+// CHECK-NEXT:              "col": 34,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 274,
+// CHECK-NEXT:              "col": 36,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const S"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "castKind": "NoOp",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "CXXBindTemporaryExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 272,
+// CHECK-NEXT:                "col": 34,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 274,
+// CHECK-NEXT:                "col": 36,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "S"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "prvalue",
+// CHECK-NEXT:              "temp": "0x{{.*}}",
+// CHECK-NEXT:              "dtor": {
+// CHECK-NEXT:               "id": "0x{{.*}}",
+// CHECK-NEXT:               "kind": "CXXDestructorDecl",
+// CHECK-NEXT:               "name": "~S",
+// CHECK-NEXT:               "type": {
+// CHECK-NEXT:                "qualType": "void () noexcept"
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "CXXTemporaryObjectExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 272,
+// CHECK-NEXT:                  "col": 34,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 274,
+// CHECK-NEXT:                  "col": 36,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "S"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "ctorType": {
+// CHECK-NEXT:                 "qualType": "void ()"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "hadMultipleCandidates": true,
+// CHECK-NEXT:                "constructionKind": "complete"
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "FunctionDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 284,
+// CHECK-NEXT:     "line": 13,
+// CHECK-NEXT:     "col": 6,
+// CHECK-NEXT:     "tokLen": 4
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 279,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 4
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 321,
+// CHECK-NEXT:      "line": 15,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "name": "test",
+// CHECK-NEXT:    "mangledName": "_Z4testv",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "void ()"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CompoundStmt",
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 291,
+// CHECK-NEXT:        "line": 13,
+// CHECK-NEXT:        "col": 13,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 321,
+// CHECK-NEXT:        "line": 15,
+// CHECK-NEXT:        "col": 1,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "CXXForRangeStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 295,
+// CHECK-NEXT:          "line": 14,
+// CHECK-NEXT:          "col": 3,
+// CHECK-NEXT:          "tokLen": 3
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 319,
+// CHECK-NEXT:          "col": 27,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {},
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "DeclStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 309,
+// CHECK-NEXT:            "col": 17,
+// CHECK-NEXT:            "tokLen": 4
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 309,
+// CHECK-NEXT:            "col": 17,
+// CHECK-NEXT:            "tokLen": 4
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "VarDecl",
+// CHECK-NEXT:            "loc": {
+// CHECK-NEXT:             "offset": 309,
+// CHECK-NEXT:             "col": 17,
+// CHECK-NEXT:             "tokLen": 4
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 309,
+// CHECK-NEXT:              "col": 17,
+// CHECK-NEXT:              "tokLen": 4
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 315,
+// CHECK-NEXT:              "col": 23,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "isImplicit": true,
+// CHECK-NEXT:            "isUsed": true,
+// CHECK-NEXT:            "name": "__range1",
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "S &&"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "init": "c",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "ExprWithCleanups",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 309,
+// CHECK-NEXT:                "col": 17,
+// CHECK-NEXT:                "tokLen": 4
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 315,
+// CHECK-NEXT:                "col": 23,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "S"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "xvalue",
+// CHECK-NEXT:              "cleanupsHaveSideEffects": true,
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 309,
+// CHECK-NEXT:                  "col": 17,
+// CHECK-NEXT:                  "tokLen": 4
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 315,
+// CHECK-NEXT:                  "col": 23,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "S"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "xvalue",
+// CHECK-NEXT:                "extendingDecl": {
+// CHECK-NEXT:                 "id": "0x{{.*}}",
+// CHECK-NEXT:                 "kind": "VarDecl",
+// CHECK-NEXT:                 "name": "__range1",
+// CHECK-NEXT:                 "type": {
+// CHECK-NEXT:                  "qualType": "S &&"
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "storageDuration": "automatic",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "CXXBindTemporaryExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 309,
+// CHECK-NEXT:                    "col": 17,
+// CHECK-NEXT:                    "tokLen": 4
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 315,
+// CHECK-NEXT:                    "col": 23,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "S"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "prvalue",
+// CHECK-NEXT:                  "temp": "0x{{.*}}",
+// CHECK-NEXT:                  "dtor": {
+// CHECK-NEXT:                   "id": "0x{{.*}}",
+// CHECK-NEXT:                   "kind": "CXXDestructorDecl",
+// CHECK-NEXT:                   "name": "~S",
+// CHECK-NEXT:                   "type": {
+// CHECK-NEXT:                    "qualType": "void () noexcept"
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "CallExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 309,
+// CHECK-NEXT:                      "col": 17,
+// CHECK-NEXT:                      "tokLen": 4
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 315,
+// CHECK-NEXT:                      "col": 23,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "S"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "prvalue",
+// CHECK-NEXT:                    "inner": [
+// CHECK-NEXT:                     {
+// CHECK-NEXT:                      "id": "0x{{.*}}",
+// CHECK-NEXT:                      "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                      "range": {
+// CHECK-NEXT:                       "begin": {
+// CHECK-NEXT:                        "offset": 309,
+// CHECK-NEXT:                        "col": 17,
+// CHECK-NEXT:                        "tokLen": 4
+// CHECK-NEXT:                       },
+// CHECK-NEXT:                       "end": {
+// CHECK-NEXT:                        "offset": 309,
+// CHECK-NEXT:                        "col": 17,
+// CHECK-NEXT:                        "tokLen": 4
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "type": {
+// CHECK-NEXT:                       "qualType": "S (*)(const int &, const S &)"
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "valueCategory": "prvalue",
+// CHECK-NEXT:                      "castKind": "FunctionToPointerDecay",
+// CHECK-NEXT:                      "inner": [
+// CHECK-NEXT:                       {
+// CHECK-NEXT:                        "id": "0x{{.*}}",
+// CHECK-NEXT:                        "kind": "DeclRefExpr",
+// CHECK-NEXT:                        "range": {
+// CHECK-NEXT:                         "begin": {
+// CHECK-NEXT:                          "offset": 309,
+// CHECK-NEXT:                          "col": 17,
+// CHECK-NEXT:                          "tokLen": 4
+// CHECK-NEXT:                         },
+// CHECK-NEXT:                         "end": {
+// CHECK-NEXT:                          "offset": 309,
+// CHECK-NEXT:                          "col": 17,
+// CHECK-NEXT:                          "tokLen": 4
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "type": {
+// CHECK-NEXT:                         "qualType": "S (const int &, const S &)"
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "valueCategory": "lvalue",
+// CHECK-NEXT:                        "referencedDecl": {
+// CHECK-NEXT:                         "id": "0x{{.*}}",
+// CHECK-NEXT:                         "kind": "FunctionDecl",
+// CHECK-NEXT:                         "name": "func",
+// CHECK-NEXT:                         "type": {
+// CHECK-NEXT:                          "qualType": "S (const int &, const S &)"
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        }
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      ]
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     {
+// CHECK-NEXT:                      "id": "0x{{.*}}",
+// CHECK-NEXT:                      "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:                      "range": {
+// CHECK-NEXT:                       "begin": {
+// CHECK-NEXT:                        "offset": 314,
+// CHECK-NEXT:                        "col": 22,
+// CHECK-NEXT:                        "tokLen": 1
+// CHECK-NEXT:                       },
+// CHECK-NEXT:                       "end": {
+// CHECK-NEXT:                        "offset": 314,
+// CHECK-NEXT:                        "col": 22,
+// CHECK-NEXT:                        "tokLen": 1
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "type": {
+// CHECK-NEXT:                       "qualType": "const int"
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "valueCategory": "lvalue",
+// CHECK-NEXT:                      "extendingDecl": {
+// CHECK-NEXT:                       "id": "0x{{.*}}",
+// CHECK-NEXT:                       "kind": "VarDecl",
+// CHECK-NEXT:                       "name": "__range1",
+// CHECK-NEXT:                       "type": {
+// CHECK-NEXT:                        "qualType": "S &&"
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "storageDuration": "automatic",
+// CHECK-NEXT:                      "boundToLValueRef": true,
+// CHECK-NEXT:                      "inner": [
+// CHECK-NEXT:                       {
+// CHECK-NEXT:                        "id": "0x{{.*}}",
+// CHECK-NEXT:                        "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                        "range": {
+// CHECK-NEXT:                         "begin": {
+// CHECK-NEXT:                          "offset": 314,
+// CHECK-NEXT:                          "col": 22,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         },
+// CHECK-NEXT:                         "end": {
+// CHECK-NEXT:                          "offset": 314,
+// CHECK-NEXT:                          "col": 22,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "type": {
+// CHECK-NEXT:                         "qualType": "const int"
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "valueCategory": "prvalue",
+// CHECK-NEXT:                        "castKind": "NoOp",
+// CHECK-NEXT:                        "inner": [
+// CHECK-NEXT:                         {
+// CHECK-NEXT:                          "id": "0x{{.*}}",
+// CHECK-NEXT:                          "kind": "IntegerLiteral",
+// CHECK-NEXT:                          "range": {
+// CHECK-NEXT:                           "begin": {
+// CHECK-NEXT:                            "offset": 314,
+// CHECK-NEXT:                            "col": 22,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           },
+// CHECK-NEXT:                           "end": {
+// CHECK-NEXT:                            "offset": 314,
+// CHECK-NEXT:                            "col": 22,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           }
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "type": {
+// CHECK-NEXT:                           "qualType": "int"
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "valueCategory": "prvalue",
+// CHECK-NEXT:                          "value": "1"
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        ]
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      ]
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     {
+// CHECK-NEXT:                      "id": "0x{{.*}}",
+// CHECK-NEXT:                      "kind": "CXXDefaultArgExpr",
+// CHECK-NEXT:                      "range": {
+// CHECK-NEXT:                       "begin": {},
+// CHECK-NEXT:                       "end": {}
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "type": {
+// CHECK-NEXT:                       "qualType": "const S"
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "valueCategory": "lvalue",
+// CHECK-NEXT:                      "hasRewrittenInit": true,
+// CHECK-NEXT:                      "inner": [
+// CHECK-NEXT:                       {
+// CHECK-NEXT:                        "id": "0x{{.*}}",
+// CHECK-NEXT:                        "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:                        "range": {
+// CHECK-NEXT:                         "begin": {
+// CHECK-NEXT:                          "offset": 272,
+// CHECK-NEXT:                          "line": 11,
+// CHECK-NEXT:                          "col": 34,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         },
+// CHECK-NEXT:                         "end": {
+// CHECK-NEXT:                          "offset": 274,
+// CHECK-NEXT:                          "col": 36,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "type": {
+// CHECK-NEXT:                         "qualType": "const S"
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "valueCategory": "lvalue",
+// CHECK-NEXT:                        "extendingDecl": {
+// CHECK-NEXT:                         "id": "0x{{.*}}",
+// CHECK-NEXT:                         "kind": "VarDecl",
+// CHECK-NEXT:                         "name": "__range1",
+// CHECK-NEXT:                         "type": {
+// CHECK-NEXT:                          "qualType": "S &&"
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "storageDuration": "automatic",
+// CHECK-NEXT:                        "boundToLValueRef": true,
+// CHECK-NEXT:                        "inner": [
+// CHECK-NEXT:                         {
+// CHECK-NEXT:                          "id": "0x{{.*}}",
+// CHECK-NEXT:                          "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                          "range": {
+// CHECK-NEXT:                           "begin": {
+// CHECK-NEXT:                            "offset": 272,
+// CHECK-NEXT:                            "col": 34,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           },
+// CHECK-NEXT:                           "end": {
+// CHECK-NEXT:                            "offset": 274,
+// CHECK-NEXT:                            "col": 36,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           }
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "type": {
+// CHECK-NEXT:                           "qualType": "const S"
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "valueCategory": "prvalue",
+// CHECK-NEXT:                          "castKind": "NoOp",
+// CHECK-NEXT:                          "inner": [
+// CHECK-NEXT:                           {
+// CHECK-NEXT:                            "id": "0x{{.*}}",
+// CHECK-NEXT:                            "kind": "CXXBindTemporaryExpr",
+// CHECK-NEXT:                            "range": {
+// CHECK-NEXT:                             "begin": {
+// CHECK-NEXT:                              "offset": 272,
+// CHECK-NEXT:                              "col": 34,
+// CHECK-NEXT:                              "tokLen": 1
+// CHECK-NEXT:                             },
+// CHECK-NEXT:                             "end": {
+// CHECK-NEXT:                              "offset": 274,
+// CHECK-NEXT:                              "col": 36,
+// CHECK-NEXT:                              "tokLen": 1
+// CHECK-NEXT:                             }
+// CHECK-NEXT:                            },
+// CHECK-NEXT:                            "type": {
+// CHECK-NEXT:                             "qualType": "S"
+// CHECK-NEXT:                            },
+// CHECK-NEXT:                            "valueCategory": "prvalue",
+// CHECK-NEXT:                            "temp": "0x{{.*}}",
+// CHECK-NEXT:                            "dtor": {
+// CHECK-NEXT:                             "id": "0x{{.*}}",
+// CHECK-NEXT:                             "kind": "CXXDestructorDecl",
+// CHECK-NEXT:                             "name": "~S",
+// CHECK-NEXT:                             "type": {
+// CHECK-NEXT:                              "qualType": "void () noexcept"
+// CHECK-NEXT:                             }
+// CHECK-NEXT:                            },
+// CHECK-NEXT:                            "inner": [
+// CHECK-NEXT:                             {
+// CHECK-NEXT:                              "id": "0x{{.*}}",
+// CHECK-NEXT:                              "kind": "CXXTemporaryObjectExpr",
+// CHECK-NEXT:                              "range": {
+// CHECK-NEXT:                               "begin": {
+// CHECK-NEXT:                                "offset": 272,
+// CHECK-NEXT:                                "col": 34,
+// CHECK-NEXT:                                "tokLen": 1
+// CHECK-NEXT:                               },
+// CHECK-NEXT:                               "end": {
+// CHECK-NEXT:                                "offset": 274,
+// CHECK-NEXT:                                "col": 36,
+// CHECK-NEXT:                                "tokLen": 1
+// CHECK-NEXT:                               }
+// CHECK-NEXT:                              },
+// CHECK-NEXT:                              "type": {
+// CHECK-NEXT:                               "qualType": "S"
+// CHECK-NEXT:                              },
+// CHECK-NEXT:                              "valueCategory": "prvalue",
+// CHECK-NEXT:                              "ctorType": {
+// CHECK-NEXT:                               "qualType": "void ()"
+// CHECK-NEXT:                              },
+// CHECK-NEXT:                              "hadMultipleCandidates": true,
+// CHECK-NEXT:                              "constructionKind": "complete"
+// CHECK-NEXT:                             }
+// CHECK-NEXT:                            ]
+// CHECK-NEXT:                           }
+// CHECK-NEXT:                          ]
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        ]
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      ]
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    ]
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "DeclStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "line": 14,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "VarDecl",
+// CHECK-NEXT:            "loc": {
+// CHECK-NEXT:             "offset": 307,
+// CHECK-NEXT:             "col": 15,
+// CHECK-NEXT:             "tokLen": 1
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "isImplicit": true,
+// CHECK-NEXT:            "isUsed": true,
+// CHECK-NEXT:            "name": "__begin1",
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "init": "c",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "CXXMemberCallExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int *"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "prvalue",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "MemberExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "<bound member function type>"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "name": "begin",
+// CHECK-NEXT:                "isArrow": false,
+// CHECK-NEXT:                "referencedMemberDecl": "0x{{.*}}",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "const S"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "lvalue",
+// CHECK-NEXT:                  "castKind": "NoOp",
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "DeclRefExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "S"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "lvalue",
+// CHECK-NEXT:                    "referencedDecl": {
+// CHECK-NEXT:                     "id": "0x{{.*}}",
+// CHECK-NEXT:                     "kind": "VarDecl",
+// CHECK-NEXT:                     "name": "__range1",
+// CHECK-NEXT:                     "type": {
+// CHECK-NEXT:                      "qualType": "S &&"
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    }
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "DeclStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "VarDecl",
+// CHECK-NEXT:            "loc": {
+// CHECK-NEXT:             "offset": 307,
+// CHECK-NEXT:             "col": 15,
+// CHECK-NEXT:             "tokLen": 1
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "isImplicit": true,
+// CHECK-NEXT:            "isUsed": true,
+// CHECK-NEXT:            "name": "__end1",
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "init": "c",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "CXXMemberCallExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int *"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "prvalue",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "MemberExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "<bound member function type>"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "name": "end",
+// CHECK-NEXT:                "isArrow": false,
+// CHECK-NEXT:                "referencedMemberDecl": "0x{{.*}}",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "const S"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "lvalue",
+// CHECK-NEXT:                  "castKind": "NoOp",
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "DeclRefExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "S"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "lvalue",
+// CHECK-NEXT:                    "referencedDecl": {
+// CHECK-NEXT:                     "id": "0x{{.*}}",
+// CHECK-NEXT:                     "kind": "VarDecl",
+// CHECK-NEXT:                     "name": "__range1",
+// CHECK-NEXT:                     "type": {
+// CHECK-NEXT:                      "qualType": "S &&"
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    }
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "BinaryOperator",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "type": {
+// CHECK-NEXT:           "qualType": "bool"
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "valueCategory": "prvalue",
+// CHECK-NEXT:          "opcode": "!=",
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "ImplicitCastExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "castKind": "LValueToRValue",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "DeclRefExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int *"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "lvalue",
+// CHECK-NEXT:              "referencedDecl": {
+// CHECK-NEXT:               "id": "0x{{.*}}",
+// CHECK-NEXT:               "kind": "VarDecl",
+// CHECK-NEXT:               "name": "__begin1",
+// CHECK-NEXT:               "type": {
+// CHECK-NEXT:                "qualType": "const int *"
+// CHECK-NEXT:               }
+// CHECK-NEXT:              }
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           },
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "ImplicitCastExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "castKind": "LValueToRValue",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "DeclRefExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const int *"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "lvalue",
+// CHECK-NEXT:              "referencedDecl": {
+// CHECK-NEXT:               "id": "0x{{.*}}",
+// CHECK-NEXT:               "kind": "VarDecl",
+// CHECK-NEXT:               "name": "__end1",
+// CHECK-NEXT:               "type": {
+// CHECK-NEXT:                "qualType": "const int *"
+// CHECK-NEXT:               }
+// CHECK-NEXT:              }
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "UnaryOperator",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 307,
+// CHECK-NEXT:            "col": 15,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "type": {
+// CHECK-NEXT:           "qualType": "const int *"
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "valueCategory": "lvalue",
+// CHECK-NEXT:          "isPostfix": false,
+// CHECK-NEXT:          "opcode": "++",
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "DeclRefExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const int *"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "lvalue",
+// CHECK-NEXT:            "referencedDecl": {
+// CHECK-NEXT:             "id": "0x{{.*}}",
+// CHECK-NEXT:             "kind": "VarDecl",
+// CHECK-NEXT:             "name": "__begin1",
+// CHECK-NEXT:             "type": {
+// CHECK-NEXT:              "qualType": "const int *"
+// CHECK-NEXT:             }
+// CHECK-NEXT:            }
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "DeclStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 300,
+// CHECK-NEXT:            "col": 8,
+// CHECK-NEXT:            "tokLen": 4
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 316,
+// CHECK-NEXT:            "col": 24,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "VarDecl",
+// CHECK-NEXT:            "loc": {
+// CHECK-NEXT:             "offset": 305,
+// CHECK-NEXT:             "col": 13,
+// CHECK-NEXT:             "tokLen": 1
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 300,
+// CHECK-NEXT:              "col": 8,
+// CHECK-NEXT:              "tokLen": 4
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 307,
+// CHECK-NEXT:              "col": 15,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "name": "v",
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "int"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "init": "c",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "ImplicitCastExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 307,
+// CHECK-NEXT:                "col": 15,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "int"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "prvalue",
+// CHECK-NEXT:              "castKind": "LValueToRValue",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "UnaryOperator",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 307,
+// CHECK-NEXT:                  "col": 15,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "const int"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "lvalue",
+// CHECK-NEXT:                "isPostfix": false,
+// CHECK-NEXT:                "opcode": "*",
+// CHECK-NEXT:                "canOverflow": false,
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 307,
+// CHECK-NEXT:                    "col": 15,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "const int *"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "prvalue",
+// CHECK-NEXT:                  "castKind": "LValueToRValue",
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "DeclRefExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 307,
+// CHECK-NEXT:                      "col": 15,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "const int *"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "lvalue",
+// CHECK-NEXT:                    "referencedDecl": {
+// CHECK-NEXT:                     "id": "0x{{.*}}",
+// CHECK-NEXT:                     "kind": "VarDecl",
+// CHECK-NEXT:                     "name": "__begin1",
+// CHECK-NEXT:                     "type": {
+// CHECK-NEXT:                      "qualType": "const int *"
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    }
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         },
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "CompoundStmt",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 318,
+// CHECK-NEXT:            "col": 26,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 319,
+// CHECK-NEXT:            "col": 27,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          }
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   }
+// CHECK-NEXT:  ]
+// CHECK-NEXT: }
diff --git a/clang/test/AST/ast-dump-default-init-json.cpp b/clang/test/AST/ast-dump-default-init-json.cpp
new file mode 100644
index 0000000000000..1058b4e3ea4d9
--- /dev/null
+++ b/clang/test/AST/ast-dump-default-init-json.cpp
@@ -0,0 +1,929 @@
+// RUN: %clang_cc1 -triple x86_64-linux-gnu -fsyntax-only -ast-dump=json %s | FileCheck %s
+
+struct A {
+  int arr[1];
+};
+
+struct B {
+  const A &a = A{{0}};
+};
+
+void test() {
+  B b{};
+}
+
+// NOTE: CHECK lines have been autogenerated by gen_ast_dump_json_test.py
+
+
+// CHECK-NOT: {{^}}Dumping
+// CHECK:  "kind": "TranslationUnitDecl",
+// CHECK-NEXT:  "loc": {},
+// CHECK-NEXT:  "range": {
+// CHECK-NEXT:   "begin": {},
+// CHECK-NEXT:   "end": {}
+// CHECK-NEXT:  },
+// CHECK-NEXT:  "inner": [
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__int128_t",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__int128"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "BuiltinType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__int128"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__uint128_t",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "unsigned __int128"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "BuiltinType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "unsigned __int128"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__NSConstantString",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__NSConstantString_tag"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "RecordType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__NSConstantString_tag"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "decl": {
+// CHECK-NEXT:       "id": "0x{{.*}}",
+// CHECK-NEXT:       "kind": "CXXRecordDecl",
+// CHECK-NEXT:       "name": "__NSConstantString_tag"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__builtin_ms_va_list",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "char *"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "PointerType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "char *"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "BuiltinType",
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "char"
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "TypedefDecl",
+// CHECK-NEXT:    "loc": {},
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {},
+// CHECK-NEXT:     "end": {}
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isImplicit": true,
+// CHECK-NEXT:    "name": "__builtin_va_list",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "__va_list_tag[1]"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "ConstantArrayType",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "__va_list_tag[1]"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "size": 1,
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "RecordType",
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "__va_list_tag"
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "decl": {
+// CHECK-NEXT:         "id": "0x{{.*}}",
+// CHECK-NEXT:         "kind": "CXXRecordDecl",
+// CHECK-NEXT:         "name": "__va_list_tag"
+// CHECK-NEXT:        }
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "CXXRecordDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 99,
+// CHECK-NEXT:     "file": "{{.*}}",
+// CHECK-NEXT:     "line": 3,
+// CHECK-NEXT:     "col": 8,
+// CHECK-NEXT:     "tokLen": 1
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 92,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 6
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 117,
+// CHECK-NEXT:      "line": 5,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isReferenced": true,
+// CHECK-NEXT:    "name": "A",
+// CHECK-NEXT:    "tagUsed": "struct",
+// CHECK-NEXT:    "completeDefinition": true,
+// CHECK-NEXT:    "definitionData": {
+// CHECK-NEXT:     "canPassInRegisters": true,
+// CHECK-NEXT:     "copyAssign": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "copyCtor": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "defaultCtor": {
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "dtor": {
+// CHECK-NEXT:      "irrelevant": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "isAggregate": true,
+// CHECK-NEXT:     "isLiteral": true,
+// CHECK-NEXT:     "isPOD": true,
+// CHECK-NEXT:     "isStandardLayout": true,
+// CHECK-NEXT:     "isTrivial": true,
+// CHECK-NEXT:     "isTriviallyCopyable": true,
+// CHECK-NEXT:     "moveAssign": {
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "moveCtor": {
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXRecordDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 99,
+// CHECK-NEXT:       "line": 3,
+// CHECK-NEXT:       "col": 8,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 92,
+// CHECK-NEXT:        "col": 1,
+// CHECK-NEXT:        "tokLen": 6
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 99,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isImplicit": true,
+// CHECK-NEXT:      "name": "A",
+// CHECK-NEXT:      "tagUsed": "struct"
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "FieldDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 109,
+// CHECK-NEXT:       "line": 4,
+// CHECK-NEXT:       "col": 7,
+// CHECK-NEXT:       "tokLen": 3
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 105,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 3
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 114,
+// CHECK-NEXT:        "col": 12,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "name": "arr",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "int[1]"
+// CHECK-NEXT:      }
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXDestructorDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 99,
+// CHECK-NEXT:       "line": 3,
+// CHECK-NEXT:       "col": 8,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 99,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 99,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isImplicit": true,
+// CHECK-NEXT:      "isReferenced": true,
+// CHECK-NEXT:      "name": "~A",
+// CHECK-NEXT:      "mangledName": "_ZN1AD1Ev",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "void () noexcept"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inline": true,
+// CHECK-NEXT:      "explicitlyDefaulted": "default"
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "CXXRecordDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 128,
+// CHECK-NEXT:     "line": 7,
+// CHECK-NEXT:     "col": 8,
+// CHECK-NEXT:     "tokLen": 1
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 121,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 6
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 155,
+// CHECK-NEXT:      "line": 9,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "isReferenced": true,
+// CHECK-NEXT:    "name": "B",
+// CHECK-NEXT:    "tagUsed": "struct",
+// CHECK-NEXT:    "completeDefinition": true,
+// CHECK-NEXT:    "definitionData": {
+// CHECK-NEXT:     "canConstDefaultInit": true,
+// CHECK-NEXT:     "canPassInRegisters": true,
+// CHECK-NEXT:     "copyAssign": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "copyCtor": {
+// CHECK-NEXT:      "hasConstParam": true,
+// CHECK-NEXT:      "implicitHasConstParam": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "defaultCtor": {
+// CHECK-NEXT:      "defaultedIsConstexpr": true,
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "isConstexpr": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "nonTrivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "dtor": {
+// CHECK-NEXT:      "irrelevant": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "hasConstexprNonCopyMoveConstructor": true,
+// CHECK-NEXT:     "isAggregate": true,
+// CHECK-NEXT:     "isLiteral": true,
+// CHECK-NEXT:     "isTriviallyCopyable": true,
+// CHECK-NEXT:     "moveAssign": {
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "moveCtor": {
+// CHECK-NEXT:      "exists": true,
+// CHECK-NEXT:      "needsImplicit": true,
+// CHECK-NEXT:      "simple": true,
+// CHECK-NEXT:      "trivial": true
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CXXRecordDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 128,
+// CHECK-NEXT:       "line": 7,
+// CHECK-NEXT:       "col": 8,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 121,
+// CHECK-NEXT:        "col": 1,
+// CHECK-NEXT:        "tokLen": 6
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 128,
+// CHECK-NEXT:        "col": 8,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "isImplicit": true,
+// CHECK-NEXT:      "name": "B",
+// CHECK-NEXT:      "tagUsed": "struct"
+// CHECK-NEXT:     },
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "FieldDecl",
+// CHECK-NEXT:      "loc": {
+// CHECK-NEXT:       "offset": 143,
+// CHECK-NEXT:       "line": 8,
+// CHECK-NEXT:       "col": 12,
+// CHECK-NEXT:       "tokLen": 1
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 134,
+// CHECK-NEXT:        "col": 3,
+// CHECK-NEXT:        "tokLen": 5
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 152,
+// CHECK-NEXT:        "col": 21,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "name": "a",
+// CHECK-NEXT:      "type": {
+// CHECK-NEXT:       "qualType": "const A &"
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "hasInClassInitializer": true,
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "ExprWithCleanups",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 147,
+// CHECK-NEXT:          "col": 16,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 152,
+// CHECK-NEXT:          "col": 21,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "type": {
+// CHECK-NEXT:         "qualType": "const A"
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "valueCategory": "lvalue",
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 147,
+// CHECK-NEXT:            "col": 16,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 152,
+// CHECK-NEXT:            "col": 21,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "type": {
+// CHECK-NEXT:           "qualType": "const A"
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "valueCategory": "lvalue",
+// CHECK-NEXT:          "extendingDecl": {
+// CHECK-NEXT:           "id": "0x{{.*}}",
+// CHECK-NEXT:           "kind": "FieldDecl",
+// CHECK-NEXT:           "name": "a",
+// CHECK-NEXT:           "type": {
+// CHECK-NEXT:            "qualType": "const A &"
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "storageDuration": "automatic",
+// CHECK-NEXT:          "boundToLValueRef": true,
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "ImplicitCastExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 147,
+// CHECK-NEXT:              "col": 16,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 152,
+// CHECK-NEXT:              "col": 21,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "const A"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "castKind": "NoOp",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "CXXFunctionalCastExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 147,
+// CHECK-NEXT:                "col": 16,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 152,
+// CHECK-NEXT:                "col": 21,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "A"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "prvalue",
+// CHECK-NEXT:              "castKind": "NoOp",
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "InitListExpr",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 148,
+// CHECK-NEXT:                  "col": 17,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 152,
+// CHECK-NEXT:                  "col": 21,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "A"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "prvalue",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "InitListExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 149,
+// CHECK-NEXT:                    "col": 18,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 151,
+// CHECK-NEXT:                    "col": 20,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "int[1]"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "prvalue",
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "IntegerLiteral",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 150,
+// CHECK-NEXT:                      "col": 19,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 150,
+// CHECK-NEXT:                      "col": 19,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "int"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "prvalue",
+// CHECK-NEXT:                    "value": "0"
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   },
+// CHECK-NEXT:   {
+// CHECK-NEXT:    "id": "0x{{.*}}",
+// CHECK-NEXT:    "kind": "FunctionDecl",
+// CHECK-NEXT:    "loc": {
+// CHECK-NEXT:     "offset": 164,
+// CHECK-NEXT:     "line": 11,
+// CHECK-NEXT:     "col": 6,
+// CHECK-NEXT:     "tokLen": 4
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "range": {
+// CHECK-NEXT:     "begin": {
+// CHECK-NEXT:      "offset": 159,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 4
+// CHECK-NEXT:     },
+// CHECK-NEXT:     "end": {
+// CHECK-NEXT:      "offset": 182,
+// CHECK-NEXT:      "line": 13,
+// CHECK-NEXT:      "col": 1,
+// CHECK-NEXT:      "tokLen": 1
+// CHECK-NEXT:     }
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "name": "test",
+// CHECK-NEXT:    "mangledName": "_Z4testv",
+// CHECK-NEXT:    "type": {
+// CHECK-NEXT:     "qualType": "void ()"
+// CHECK-NEXT:    },
+// CHECK-NEXT:    "inner": [
+// CHECK-NEXT:     {
+// CHECK-NEXT:      "id": "0x{{.*}}",
+// CHECK-NEXT:      "kind": "CompoundStmt",
+// CHECK-NEXT:      "range": {
+// CHECK-NEXT:       "begin": {
+// CHECK-NEXT:        "offset": 171,
+// CHECK-NEXT:        "line": 11,
+// CHECK-NEXT:        "col": 13,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       },
+// CHECK-NEXT:       "end": {
+// CHECK-NEXT:        "offset": 182,
+// CHECK-NEXT:        "line": 13,
+// CHECK-NEXT:        "col": 1,
+// CHECK-NEXT:        "tokLen": 1
+// CHECK-NEXT:       }
+// CHECK-NEXT:      },
+// CHECK-NEXT:      "inner": [
+// CHECK-NEXT:       {
+// CHECK-NEXT:        "id": "0x{{.*}}",
+// CHECK-NEXT:        "kind": "DeclStmt",
+// CHECK-NEXT:        "range": {
+// CHECK-NEXT:         "begin": {
+// CHECK-NEXT:          "offset": 175,
+// CHECK-NEXT:          "line": 12,
+// CHECK-NEXT:          "col": 3,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         },
+// CHECK-NEXT:         "end": {
+// CHECK-NEXT:          "offset": 180,
+// CHECK-NEXT:          "col": 8,
+// CHECK-NEXT:          "tokLen": 1
+// CHECK-NEXT:         }
+// CHECK-NEXT:        },
+// CHECK-NEXT:        "inner": [
+// CHECK-NEXT:         {
+// CHECK-NEXT:          "id": "0x{{.*}}",
+// CHECK-NEXT:          "kind": "VarDecl",
+// CHECK-NEXT:          "loc": {
+// CHECK-NEXT:           "offset": 177,
+// CHECK-NEXT:           "col": 5,
+// CHECK-NEXT:           "tokLen": 1
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "range": {
+// CHECK-NEXT:           "begin": {
+// CHECK-NEXT:            "offset": 175,
+// CHECK-NEXT:            "col": 3,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           },
+// CHECK-NEXT:           "end": {
+// CHECK-NEXT:            "offset": 179,
+// CHECK-NEXT:            "col": 7,
+// CHECK-NEXT:            "tokLen": 1
+// CHECK-NEXT:           }
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "name": "b",
+// CHECK-NEXT:          "type": {
+// CHECK-NEXT:           "qualType": "B"
+// CHECK-NEXT:          },
+// CHECK-NEXT:          "init": "list",
+// CHECK-NEXT:          "inner": [
+// CHECK-NEXT:           {
+// CHECK-NEXT:            "id": "0x{{.*}}",
+// CHECK-NEXT:            "kind": "InitListExpr",
+// CHECK-NEXT:            "range": {
+// CHECK-NEXT:             "begin": {
+// CHECK-NEXT:              "offset": 178,
+// CHECK-NEXT:              "col": 6,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             },
+// CHECK-NEXT:             "end": {
+// CHECK-NEXT:              "offset": 179,
+// CHECK-NEXT:              "col": 7,
+// CHECK-NEXT:              "tokLen": 1
+// CHECK-NEXT:             }
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "type": {
+// CHECK-NEXT:             "qualType": "B"
+// CHECK-NEXT:            },
+// CHECK-NEXT:            "valueCategory": "prvalue",
+// CHECK-NEXT:            "inner": [
+// CHECK-NEXT:             {
+// CHECK-NEXT:              "id": "0x{{.*}}",
+// CHECK-NEXT:              "kind": "CXXDefaultInitExpr",
+// CHECK-NEXT:              "range": {
+// CHECK-NEXT:               "begin": {
+// CHECK-NEXT:                "offset": 179,
+// CHECK-NEXT:                "col": 7,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               },
+// CHECK-NEXT:               "end": {
+// CHECK-NEXT:                "offset": 179,
+// CHECK-NEXT:                "col": 7,
+// CHECK-NEXT:                "tokLen": 1
+// CHECK-NEXT:               }
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "type": {
+// CHECK-NEXT:               "qualType": "const A"
+// CHECK-NEXT:              },
+// CHECK-NEXT:              "valueCategory": "lvalue",
+// CHECK-NEXT:              "hasRewrittenInit": true,
+// CHECK-NEXT:              "inner": [
+// CHECK-NEXT:               {
+// CHECK-NEXT:                "id": "0x{{.*}}",
+// CHECK-NEXT:                "kind": "ExprWithCleanups",
+// CHECK-NEXT:                "range": {
+// CHECK-NEXT:                 "begin": {
+// CHECK-NEXT:                  "offset": 147,
+// CHECK-NEXT:                  "line": 8,
+// CHECK-NEXT:                  "col": 16,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 },
+// CHECK-NEXT:                 "end": {
+// CHECK-NEXT:                  "offset": 152,
+// CHECK-NEXT:                  "col": 21,
+// CHECK-NEXT:                  "tokLen": 1
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "type": {
+// CHECK-NEXT:                 "qualType": "const A"
+// CHECK-NEXT:                },
+// CHECK-NEXT:                "valueCategory": "lvalue",
+// CHECK-NEXT:                "inner": [
+// CHECK-NEXT:                 {
+// CHECK-NEXT:                  "id": "0x{{.*}}",
+// CHECK-NEXT:                  "kind": "MaterializeTemporaryExpr",
+// CHECK-NEXT:                  "range": {
+// CHECK-NEXT:                   "begin": {
+// CHECK-NEXT:                    "offset": 147,
+// CHECK-NEXT:                    "col": 16,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   },
+// CHECK-NEXT:                   "end": {
+// CHECK-NEXT:                    "offset": 152,
+// CHECK-NEXT:                    "col": 21,
+// CHECK-NEXT:                    "tokLen": 1
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "type": {
+// CHECK-NEXT:                   "qualType": "const A"
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "valueCategory": "lvalue",
+// CHECK-NEXT:                  "extendingDecl": {
+// CHECK-NEXT:                   "id": "0x{{.*}}",
+// CHECK-NEXT:                   "kind": "FieldDecl",
+// CHECK-NEXT:                   "name": "a",
+// CHECK-NEXT:                   "type": {
+// CHECK-NEXT:                    "qualType": "const A &"
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  },
+// CHECK-NEXT:                  "storageDuration": "automatic",
+// CHECK-NEXT:                  "boundToLValueRef": true,
+// CHECK-NEXT:                  "inner": [
+// CHECK-NEXT:                   {
+// CHECK-NEXT:                    "id": "0x{{.*}}",
+// CHECK-NEXT:                    "kind": "ImplicitCastExpr",
+// CHECK-NEXT:                    "range": {
+// CHECK-NEXT:                     "begin": {
+// CHECK-NEXT:                      "offset": 147,
+// CHECK-NEXT:                      "col": 16,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     },
+// CHECK-NEXT:                     "end": {
+// CHECK-NEXT:                      "offset": 152,
+// CHECK-NEXT:                      "col": 21,
+// CHECK-NEXT:                      "tokLen": 1
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "type": {
+// CHECK-NEXT:                     "qualType": "const A"
+// CHECK-NEXT:                    },
+// CHECK-NEXT:                    "valueCategory": "prvalue",
+// CHECK-NEXT:                    "castKind": "NoOp",
+// CHECK-NEXT:                    "inner": [
+// CHECK-NEXT:                     {
+// CHECK-NEXT:                      "id": "0x{{.*}}",
+// CHECK-NEXT:                      "kind": "CXXFunctionalCastExpr",
+// CHECK-NEXT:                      "range": {
+// CHECK-NEXT:                       "begin": {
+// CHECK-NEXT:                        "offset": 147,
+// CHECK-NEXT:                        "col": 16,
+// CHECK-NEXT:                        "tokLen": 1
+// CHECK-NEXT:                       },
+// CHECK-NEXT:                       "end": {
+// CHECK-NEXT:                        "offset": 152,
+// CHECK-NEXT:                        "col": 21,
+// CHECK-NEXT:                        "tokLen": 1
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "type": {
+// CHECK-NEXT:                       "qualType": "A"
+// CHECK-NEXT:                      },
+// CHECK-NEXT:                      "valueCategory": "prvalue",
+// CHECK-NEXT:                      "castKind": "NoOp",
+// CHECK-NEXT:                      "inner": [
+// CHECK-NEXT:                       {
+// CHECK-NEXT:                        "id": "0x{{.*}}",
+// CHECK-NEXT:                        "kind": "InitListExpr",
+// CHECK-NEXT:                        "range": {
+// CHECK-NEXT:                         "begin": {
+// CHECK-NEXT:                          "offset": 148,
+// CHECK-NEXT:                          "col": 17,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         },
+// CHECK-NEXT:                         "end": {
+// CHECK-NEXT:                          "offset": 152,
+// CHECK-NEXT:                          "col": 21,
+// CHECK-NEXT:                          "tokLen": 1
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "type": {
+// CHECK-NEXT:                         "qualType": "A"
+// CHECK-NEXT:                        },
+// CHECK-NEXT:                        "valueCategory": "prvalue",
+// CHECK-NEXT:                        "inner": [
+// CHECK-NEXT:                         {
+// CHECK-NEXT:                          "id": "0x{{.*}}",
+// CHECK-NEXT:                          "kind": "InitListExpr",
+// CHECK-NEXT:                          "range": {
+// CHECK-NEXT:                           "begin": {
+// CHECK-NEXT:                            "offset": 149,
+// CHECK-NEXT:                            "col": 18,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           },
+// CHECK-NEXT:                           "end": {
+// CHECK-NEXT:                            "offset": 151,
+// CHECK-NEXT:                            "col": 20,
+// CHECK-NEXT:                            "tokLen": 1
+// CHECK-NEXT:                           }
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "type": {
+// CHECK-NEXT:                           "qualType": "int[1]"
+// CHECK-NEXT:                          },
+// CHECK-NEXT:                          "valueCategory": "prvalue",
+// CHECK-NEXT:                          "inner": [
+// CHECK-NEXT:                           {
+// CHECK-NEXT:                            "id": "0x{{.*}}",
+// CHECK-NEXT:                            "kind": "IntegerLiteral",
+// CHECK-NEXT:                            "range": {
+// CHECK-NEXT:                             "begin": {
+// CHECK-NEXT:                              "offset": 150,
+// CHECK-NEXT:                              "col": 19,
+// CHECK-NEXT:                              "tokLen": 1
+// CHECK-NEXT:                             },
+// CHECK-NEXT:                             "end": {
+// CHECK-NEXT:                              "offset": 150,
+// CHECK-NEXT:                              "col": 19,
+// CHECK-NEXT:                              "tokLen": 1
+// CHECK-NEXT:                             }
+// CHECK-NEXT:                            },
+// CHECK-NEXT:                            "type": {
+// CHECK-NEXT:                             "qualType": "int"
+// CHECK-NEXT:                            },
+// CHECK-NEXT:                            "valueCategory": "prvalue",
+// CHECK-NEXT:                            "value": "0"
+// CHECK-NEXT:                           }
+// CHECK-NEXT:                          ]
+// CHECK-NEXT:                         }
+// CHECK-NEXT:                        ]
+// CHECK-NEXT:                       }
+// CHECK-NEXT:                      ]
+// CHECK-NEXT:                     }
+// CHECK-NEXT:                    ]
+// CHECK-NEXT:                   }
+// CHECK-NEXT:                  ]
+// CHECK-NEXT:                 }
+// CHECK-NEXT:                ]
+// CHECK-NEXT:               }
+// CHECK-NEXT:              ]
+// CHECK-NEXT:             }
+// CHECK-NEXT:            ]
+// CHECK-NEXT:           }
+// CHECK-NEXT:          ]
+// CHECK-NEXT:         }
+// CHECK-NEXT:        ]
+// CHECK-NEXT:       }
+// CHECK-NEXT:      ]
+// CHECK-NEXT:     }
+// CHECK-NEXT:    ]
+// CHECK-NEXT:   }
+// CHECK-NEXT:  ]
+// CHECK-NEXT: }
diff --git a/clang/test/AST/ast-dump-default-init.cpp b/clang/test/AST/ast-dump-default-init.cpp
new file mode 100644
index 0000000000000..9fe945ee6e932
--- /dev/null
+++ b/clang/test/AST/ast-dump-default-init.cpp
@@ -0,0 +1,21 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -ast-dump %s | FileCheck %s
+
+struct A {
+  int arr[1];
+};
+
+struct B {
+  const A &a = A{{0}};
+};
+
+void test() {
+  B b{};
+}
+// CHECK: -CXXDefaultInitExpr 0x{{[^ ]*}} <{{.*}}> 'const A' lvalue has rewritten init
+// CHECK-NEXT:  `-ExprWithCleanups 0x{{[^ ]*}} <{{.*}}> 'const A' lvalue
+// CHECK-NEXT:    `-MaterializeTemporaryExpr 0x{{[^ ]*}} <{{.*}}> 'const A' lvalue extended by Field 0x{{[^ ]*}} 'a' 'const A &'
+// CHECK-NEXT:      `-ImplicitCastExpr 0x{{[^ ]*}} <{{.*}}> 'const A' <NoOp>
+// CHECK-NEXT:        `-CXXFunctionalCastExpr 0x{{[^ ]*}} <{{.*}}> 'A' functional cast to A <NoOp>
+// CHECK-NEXT:          `-InitListExpr 0x{{[^ ]*}} <{{.*}}> 'A'
+// CHECK-NEXT:            `-InitListExpr 0x{{[^ ]*}} <{{.*}}> 'int[1]'
+// CHECK-NEXT:              `-IntegerLiteral 0x{{[^ ]*}} <{{.*}}> 'int' 0
diff --git a/clang/test/AST/ast-dump-for-range-lifetime.cpp b/clang/test/AST/ast-dump-for-range-lifetime.cpp
index 88b838268be2e..0e92b6990ed50 100644
--- a/clang/test/AST/ast-dump-for-range-lifetime.cpp
+++ b/clang/test/AST/ast-dump-for-range-lifetime.cpp
@@ -132,6 +132,9 @@ void test4() {
   // CHECK-NEXT:  |       | `-DeclRefExpr {{.*}} 'int (&(const A &))[3]' lvalue Function {{.*}} 'default_arg_fn' 'int (&(const A &))[3]'
   // CHECK-NEXT:  |       `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const A':'const P2718R0::A' lvalue has rewritten init
   // CHECK-NEXT:  |         `-MaterializeTemporaryExpr {{.*}} 'const A':'const P2718R0::A' lvalue extended by Var {{.*}} '__range1' 'int (&)[3]'
+  // CHECK-NEXT:  |           `-ImplicitCastExpr {{.*}} 'const A':'const P2718R0::A' <NoOp>
+  // CHECK-NEXT:  |             `-CXXBindTemporaryExpr {{.*}} 'A':'P2718R0::A' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |               `-CXXTemporaryObjectExpr {{.*}} 'A':'P2718R0::A' 'void ()'
   for (auto e : default_arg_fn()) 
     bar(e);
 }
@@ -179,10 +182,19 @@ void test5() {
   // CHECK-NEXT:  |               |       |       |     `-CXXTemporaryObjectExpr {{.*}} 'A':'P2718R0::A' 'void ()'
   // CHECK-NEXT:  |               |       |       `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |               |       |         `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'int (&)[3]'
+  // CHECK-NEXT:  |               |       |           `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |               |       |             `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |               |       |               `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   // CHECK-NEXT:  |               |       `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |               |         `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'int (&)[3]'
+  // CHECK-NEXT:  |               |           `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |               |             `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |               |               `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   // CHECK-NEXT:  |               `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |                 `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'int (&)[3]'
+  // CHECK-NEXT:  |                   `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |                     `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |                       `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   for (auto e : default_arg_fn(foo(foo(foo(A())))))
     bar(e);
 }
@@ -219,10 +231,19 @@ void test6() {
   // CHECK-NEXT:  |           |       |       |     `-CXXTemporaryObjectExpr {{.*}} 'C':'P2718R0::C' 'void ()'
   // CHECK-NEXT:  |           |       |       `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |           |       |         `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'C &&'
+  // CHECK-NEXT:  |           |       |           `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |           |       |             `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |           |       |               `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   // CHECK-NEXT:  |           |       `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |           |         `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'C &&'
+  // CHECK-NEXT:  |           |           `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |           |             `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |           |               `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   // CHECK-NEXT:  |           `-CXXDefaultArgExpr {{.*}} <<invalid sloc>> 'const DefaultA':'const P2718R0::DefaultA' lvalue has rewritten init
   // CHECK-NEXT:  |             `-MaterializeTemporaryExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' lvalue extended by Var {{.*}} '__range1' 'C &&'
+  // CHECK-NEXT:  |               `-ImplicitCastExpr {{.*}} 'const DefaultA':'const P2718R0::DefaultA' <NoOp>
+  // CHECK-NEXT:  |                 `-CXXBindTemporaryExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' (CXXTemporary {{.*}})
+  // CHECK-NEXT:  |                   `-CXXTemporaryObjectExpr {{.*}} 'DefaultA':'P2718R0::DefaultA' 'void ()'
   for (auto e : C(0, C(0, C(0, C()))))
     bar(e);
 }
diff --git a/clang/unittests/AST/ASTTraverserTest.cpp b/clang/unittests/AST/ASTTraverserTest.cpp
index 362f37655a0ae..8b6e3e90c0ea6 100644
--- a/clang/unittests/AST/ASTTraverserTest.cpp
+++ b/clang/unittests/AST/ASTTraverserTest.cpp
@@ -368,6 +368,8 @@ FunctionDecl 'stringConstruct'
   |             |-ImplicitCastExpr
   |             | `-StringLiteral
   |             `-CXXDefaultArgExpr
+  |               `-UnaryOperator
+  |                 `-IntegerLiteral
   `-ExprWithCleanups
     `-CXXOperatorCallExpr
       |-ImplicitCastExpr
@@ -378,6 +380,8 @@ FunctionDecl 'stringConstruct'
           |-ImplicitCastExpr
           | `-StringLiteral
           `-CXXDefaultArgExpr
+            `-UnaryOperator
+              `-IntegerLiteral
 )cpp");
 
     EXPECT_EQ(dumpASTString(TK_IgnoreUnlessSpelledInSource,
@@ -415,6 +419,8 @@ FunctionDecl 'overloadCall'
   |             |-ImplicitCastExpr
   |             | `-StringLiteral
   |             `-CXXDefaultArgExpr
+  |               `-UnaryOperator
+  |                 `-IntegerLiteral
   `-CXXMemberCallExpr
     `-MemberExpr
       `-ParenExpr
@@ -1219,6 +1225,7 @@ CXXRecordDecl 'Record'
 | | `-IntegerLiteral
 | |-CXXCtorInitializer 'm_i2'
 | | `-CXXDefaultInitExpr
+| |   `-IntegerLiteral
 | |-CXXCtorInitializer 'm_s'
 | | `-CXXConstructExpr
 | `-CompoundStmt
@@ -1485,6 +1492,7 @@ CallExpr
 | `-DeclRefExpr 'hasDefaultArg'
 |-IntegerLiteral
 `-CXXDefaultArgExpr
+  `-IntegerLiteral
 )cpp");
     EXPECT_EQ(dumpASTString(TK_IgnoreUnlessSpelledInSource,
                             BN[0].getNodeAs<CallExpr>("funcCall")),

From a27ab3f54377a0addeef1ec9ba7c826fd885a304 Mon Sep 17 00:00:00 2001
From: Dominik Adamski <dominik.adamski@amd.com>
Date: Fri, 12 Apr 2024 16:41:34 +0200
Subject: [PATCH 260/886] [Flang][AMDGPU] Add rocm-path flag (#88190)

ROCm installation path is used for finding and automatically linking
required bitcode libraries for OpenMP AMDGPU offload.

Reported issue: https://github.com/llvm/llvm-project/issues/82553
---
 clang/include/clang/Driver/Options.td                       | 3 ++-
 .../Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/asanrtl.bc | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/hip.bc     | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ockl.bc    | 0
 .../bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_off.bc   | 0
 .../bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_on.bc    | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_off.bc      | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_on.bc       | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_off.bc  | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_on.bc   | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1010.bc | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1011.bc | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1012.bc | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_803.bc  | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_900.bc  | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_908.bc  | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_off.bc  | 0
 .../rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_on.bc   | 0
 .../amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_off.bc   | 0
 .../amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_on.bc    | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ocml.bc    | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/opencl.bc  | 0
 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/asanrtl.bc     | 0
 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/hip.bc         | 0
 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ockl.bc        | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_abi_version_400.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_abi_version_500.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_abi_version_600.bc      | 0
 .../rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_off.bc  | 0
 .../rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_on.bc   | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_off.bc   | 0
 .../Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_on.bc    | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_finite_only_off.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_finite_only_on.bc       | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1010.bc     | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1011.bc     | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1012.bc     | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_803.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_900.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_isa_version_908.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_off.bc      | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_on.bc       | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_off.bc  | 0
 .../Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_on.bc   | 0
 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ocml.bc        | 0
 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/opencl.bc      | 0
 flang/test/Driver/Inputs/rocm/bin/.hipVersion               | 6 ++++++
 flang/test/Driver/Inputs/rocm/include/hip/hip_runtime.h     | 0
 flang/test/Driver/Inputs/rocm/share/hip/hipVersion          | 6 ++++++
 flang/test/Driver/driver-help-hidden.f90                    | 1 +
 flang/test/Driver/driver-help.f90                           | 1 +
 flang/test/Driver/omp-driver-offload.f90                    | 6 ++++++
 52 files changed, 22 insertions(+), 1 deletion(-)
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/asanrtl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/hip.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ockl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1010.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1011.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1012.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_803.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_900.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_908.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ocml.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/opencl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/asanrtl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/hip.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ockl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_400.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_500.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_600.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1010.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1011.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1012.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_803.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_900.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_908.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_off.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_on.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ocml.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/amdgcn/bitcode/opencl.bc
 create mode 100644 flang/test/Driver/Inputs/rocm/bin/.hipVersion
 create mode 100644 flang/test/Driver/Inputs/rocm/include/hip/hip_runtime.h
 create mode 100644 flang/test/Driver/Inputs/rocm/share/hip/hipVersion

diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 400f72f2250a2..9a0b5d3304ca6 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -1341,7 +1341,8 @@ def hip_link : Flag<["--"], "hip-link">, Group<opencl_Group>,
   HelpText<"Link clang-offload-bundler bundles for HIP">;
 def no_hip_rt: Flag<["-"], "no-hip-rt">, Group<hip_Group>,
   HelpText<"Do not link against HIP runtime libraries">;
-def rocm_path_EQ : Joined<["--"], "rocm-path=">, Group<hip_Group>,
+def rocm_path_EQ : Joined<["--"], "rocm-path=">,
+  Visibility<[FlangOption]>, Group<hip_Group>,
   HelpText<"ROCm installation path, used for finding and automatically linking required bitcode libraries.">;
 def hip_path_EQ : Joined<["--"], "hip-path=">, Group<hip_Group>,
   HelpText<"HIP runtime installation path, used for finding HIP version and adding HIP include path.">;
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/asanrtl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/asanrtl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/hip.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/hip.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ockl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ockl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_correctly_rounded_sqrt_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_daz_opt_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_finite_only_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1010.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1010.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1011.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1011.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1012.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_1012.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_803.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_803.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_900.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_900.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_908.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_isa_version_908.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_unsafe_math_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/oclc_wavefrontsize64_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ocml.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/ocml.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/opencl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode-no-abi-ver/opencl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/asanrtl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/asanrtl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/hip.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/hip.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ockl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ockl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_400.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_400.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_500.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_500.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_600.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_abi_version_600.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_correctly_rounded_sqrt_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_daz_opt_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_finite_only_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1010.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1010.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1011.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1011.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1012.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_1012.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_803.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_803.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_900.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_900.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_908.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_isa_version_908.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_unsafe_math_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_off.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_off.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_on.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/oclc_wavefrontsize64_on.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ocml.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/ocml.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/opencl.bc b/flang/test/Driver/Inputs/rocm/amdgcn/bitcode/opencl.bc
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/bin/.hipVersion b/flang/test/Driver/Inputs/rocm/bin/.hipVersion
new file mode 100644
index 0000000000000..677293c09139f
--- /dev/null
+++ b/flang/test/Driver/Inputs/rocm/bin/.hipVersion
@@ -0,0 +1,6 @@
+# Auto-generated by cmake
+# NOTE: The trailing whitespace is added on purpose to verify that these
+# whitespaces are trimmed before paring.
+HIP_VERSION_MAJOR=3 
+HIP_VERSION_MINOR=6
+HIP_VERSION_PATCH=20214-a2917cd
diff --git a/flang/test/Driver/Inputs/rocm/include/hip/hip_runtime.h b/flang/test/Driver/Inputs/rocm/include/hip/hip_runtime.h
new file mode 100644
index 0000000000000..e69de29bb2d1d
diff --git a/flang/test/Driver/Inputs/rocm/share/hip/hipVersion b/flang/test/Driver/Inputs/rocm/share/hip/hipVersion
new file mode 100644
index 0000000000000..677293c09139f
--- /dev/null
+++ b/flang/test/Driver/Inputs/rocm/share/hip/hipVersion
@@ -0,0 +1,6 @@
+# Auto-generated by cmake
+# NOTE: The trailing whitespace is added on purpose to verify that these
+# whitespaces are trimmed before paring.
+HIP_VERSION_MAJOR=3 
+HIP_VERSION_MINOR=6
+HIP_VERSION_PATCH=20214-a2917cd
diff --git a/flang/test/Driver/driver-help-hidden.f90 b/flang/test/Driver/driver-help-hidden.f90
index eb00949c2a951..de2fe3048f993 100644
--- a/flang/test/Driver/driver-help-hidden.f90
+++ b/flang/test/Driver/driver-help-hidden.f90
@@ -148,6 +148,7 @@
 ! CHECK-NEXT: -pthread                Support POSIX threads in generated code
 ! CHECK-NEXT: -P                      Disable linemarker output in -E mode
 ! CHECK-NEXT: -resource-dir <value>   The directory which holds the compiler resource files
+! CHECK-NEXT: --rocm-path=<value> ROCm installation path, used for finding and automatically linking required bitcode libraries.
 ! CHECK-NEXT: -Rpass-analysis=<value> Report transformation analysis from optimization passes whose name matches the given POSIX regular expression
 ! CHECK-NEXT: -Rpass-missed=<value>   Report missed transformations by optimization passes whose name matches the given POSIX regular expression
 ! CHECK-NEXT: -Rpass=<value>          Report transformations performed by optimization passes whose name matches the given POSIX regular expression
diff --git a/flang/test/Driver/driver-help.f90 b/flang/test/Driver/driver-help.f90
index 05f50ea348c22..b258eb59c1862 100644
--- a/flang/test/Driver/driver-help.f90
+++ b/flang/test/Driver/driver-help.f90
@@ -135,6 +135,7 @@
 ! HELP-NEXT: -print-target-triple    Print the normalized target triple
 ! HELP-NEXT: -pthread                Support POSIX threads in generated code
 ! HELP-NEXT: -P                      Disable linemarker output in -E mode
+! HELP-NEXT:  --rocm-path=<value> ROCm installation path, used for finding and automatically linking required bitcode libraries.
 ! HELP-NEXT: -Rpass-analysis=<value> Report transformation analysis from optimization passes whose name matches the given POSIX regular expression
 ! HELP-NEXT: -Rpass-missed=<value>   Report missed transformations by optimization passes whose name matches the given POSIX regular expression
 ! HELP-NEXT: -Rpass=<value>          Report transformations performed by optimization passes whose name matches the given POSIX regular expression
diff --git a/flang/test/Driver/omp-driver-offload.f90 b/flang/test/Driver/omp-driver-offload.f90
index 7e9a73627cd75..8f48ca75114ce 100644
--- a/flang/test/Driver/omp-driver-offload.f90
+++ b/flang/test/Driver/omp-driver-offload.f90
@@ -201,3 +201,9 @@
 ! RUN:      -nogpulibc %s 2>&1 \
 ! RUN:   | FileCheck --check-prefix=NO-LIBC-GPU-AMDGPU %s
 ! NO-LIBC-GPU-AMDGPU-NOT: "-lcgpu-amdgpu"
+
+! RUN:   %flang -### -v --target=x86_64-unknown-linux-gnu -fopenmp  \
+! RUN:      --offload-arch=gfx900 \
+! RUN:      --rocm-path=%S/Inputs/rocm %s 2>&1 \
+! RUN:   | FileCheck --check-prefix=ROCM-PATH %s
+! ROCM-PATH: Found HIP installation: {{.*Inputs.*rocm}}, version 3.6.20214-a2917cd

From 5752e3196bc52fdac70e3650abc703570ff6209b Mon Sep 17 00:00:00 2001
From: Vincent Belliard <81770341+v-bulle@users.noreply.github.com>
Date: Fri, 12 Apr 2024 07:48:41 -0700
Subject: [PATCH 261/886] [lldb] fix dead lock in TypeCategoryMap.cpp (#87540)

FormatManager::GetCategoryForLanguage and
FormatManager::GetCategory(can_create = true) can be called concurrently
and they both take the TypeCategory::m_map_mutex and the
FormatManager::m_language_categories_mutex but in reverse order.

On one thread, GetCategoryForLanguage takes m_language_categories_mutex
and then ends calling TypeCategoryMap::Get which takes m_map_mutex

On another thread GetCategory calls TypeCategoryMap::Add which takes
m_map_mutex and then calls FormatManager::Changed() which takes
m_language_categories_mutex

If both threads are running concurrently, we have a dead lock.

The patch releases the m_map_mutex before calling Changed which avoids
the dead lock.

---------

Co-authored-by: Vincent Belliard <v-bulle@github.com>
---
 .../source/DataFormatters/TypeCategoryMap.cpp | 40 ++++++++++++++-----
 1 file changed, 29 insertions(+), 11 deletions(-)

diff --git a/lldb/source/DataFormatters/TypeCategoryMap.cpp b/lldb/source/DataFormatters/TypeCategoryMap.cpp
index fd76bab95826a..ce2cf369b5be5 100644
--- a/lldb/source/DataFormatters/TypeCategoryMap.cpp
+++ b/lldb/source/DataFormatters/TypeCategoryMap.cpp
@@ -25,19 +25,31 @@ TypeCategoryMap::TypeCategoryMap(IFormatChangeListener *lst)
 }
 
 void TypeCategoryMap::Add(KeyType name, const TypeCategoryImplSP &entry) {
-  std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
-  m_map[name] = entry;
+  {
+    std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
+    m_map[name] = entry;
+  }
+  // Release the mutex to avoid a potential deadlock between
+  // TypeCategoryMap::m_map_mutex and
+  // FormatManager::m_language_categories_mutex which can be acquired in
+  // reverse order when calling FormatManager::Changed.
   if (listener)
     listener->Changed();
 }
 
 bool TypeCategoryMap::Delete(KeyType name) {
-  std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
-  MapIterator iter = m_map.find(name);
-  if (iter == m_map.end())
-    return false;
-  m_map.erase(name);
-  Disable(name);
+  {
+    std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
+    MapIterator iter = m_map.find(name);
+    if (iter == m_map.end())
+      return false;
+    m_map.erase(name);
+    Disable(name);
+  }
+  // Release the mutex to avoid a potential deadlock between
+  // TypeCategoryMap::m_map_mutex and
+  // FormatManager::m_language_categories_mutex which can be acquired in
+  // reverse order when calling FormatManager::Changed.
   if (listener)
     listener->Changed();
   return true;
@@ -123,9 +135,15 @@ void TypeCategoryMap::DisableAllCategories() {
 }
 
 void TypeCategoryMap::Clear() {
-  std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
-  m_map.clear();
-  m_active_categories.clear();
+  {
+    std::lock_guard<std::recursive_mutex> guard(m_map_mutex);
+    m_map.clear();
+    m_active_categories.clear();
+  }
+  // Release the mutex to avoid a potential deadlock between
+  // TypeCategoryMap::m_map_mutex and
+  // FormatManager::m_language_categories_mutex which can be acquired in
+  // reverse order when calling FormatManager::Changed.
   if (listener)
     listener->Changed();
 }

From 4dd20b0728223ff20f4817edc2f89901385cb990 Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Fri, 12 Apr 2024 11:35:23 -0400
Subject: [PATCH 262/886] [BOLT][NFC] Refactor relocation loop (#88424)

Use the std `if () continue;` idiom before falling into the
processing.
---
 bolt/lib/Rewrite/RewriteInstance.cpp | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index cbbe7d00d35dc..f22bede002da5 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -2306,9 +2306,13 @@ void RewriteInstance::processRelocations() {
     return;
 
   for (const SectionRef &Section : InputFile->sections()) {
-    if (cantFail(Section.getRelocatedSection()) != InputFile->section_end() &&
-        !BinarySection(*BC, Section).isAllocatable())
-      readRelocations(Section);
+    section_iterator SecIter = cantFail(Section.getRelocatedSection());
+    if (SecIter == InputFile->section_end())
+      continue;
+    if (BinarySection(*BC, Section).isAllocatable())
+      continue;
+
+    readRelocations(Section);
   }
 
   if (NumFailedRelocations)

From 040efafa9fef101924d3cf67db20f6429ce1c871 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Fri, 12 Apr 2024 08:45:03 -0700
Subject: [PATCH 263/886] [RISCV] Support uimm32 immediates in
 RISCVInstrInfo::movImm for RV32. (#88464)

This allows us to support larger stack offsets for FrameLowering.

Fixes #88365.
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp | 11 ++++++++--
 llvm/test/CodeGen/RISCV/pr88365.ll       | 26 ++++++++++++++++++++++++
 2 files changed, 35 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/pr88365.ll

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 004a228ccc131..508f607fab20f 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -767,8 +767,15 @@ void RISCVInstrInfo::movImm(MachineBasicBlock &MBB,
                             bool DstIsDead) const {
   Register SrcReg = RISCV::X0;
 
-  if (!STI.is64Bit() && !isInt<32>(Val))
-    report_fatal_error("Should only materialize 32-bit constants for RV32");
+  // For RV32, allow a sign or unsigned 32 bit value.
+  if (!STI.is64Bit() && !isInt<32>(Val)) {
+    // If have a uimm32 it will still fit in a register so we can allow it.
+    if (!isUInt<32>(Val))
+      report_fatal_error("Should only materialize 32-bit constants for RV32");
+
+    // Sign extend for generateInstSeq.
+    Val = SignExtend64<32>(Val);
+  }
 
   RISCVMatInt::InstSeq Seq = RISCVMatInt::generateInstSeq(Val, STI);
   assert(!Seq.empty());
diff --git a/llvm/test/CodeGen/RISCV/pr88365.ll b/llvm/test/CodeGen/RISCV/pr88365.ll
new file mode 100644
index 0000000000000..73010fdf40447
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/pr88365.ll
@@ -0,0 +1,26 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=riscv32 | FileCheck %s
+
+define void @foo() {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi sp, sp, -2032
+; CHECK-NEXT:    .cfi_def_cfa_offset 2032
+; CHECK-NEXT:    sw ra, 2028(sp) # 4-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -4
+; CHECK-NEXT:    li a0, -2048
+; CHECK-NEXT:    sub sp, sp, a0
+; CHECK-NEXT:    .cfi_def_cfa_offset -16
+; CHECK-NEXT:    addi a0, sp, 4
+; CHECK-NEXT:    call use
+; CHECK-NEXT:    li a0, -2048
+; CHECK-NEXT:    add sp, sp, a0
+; CHECK-NEXT:    lw ra, 2028(sp) # 4-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 2032
+; CHECK-NEXT:    ret
+  %1 = alloca [1073741818 x i32], align 4
+  call void @use(ptr %1)
+  ret void
+}
+
+declare void @use(ptr)

From 72dfee114bd38dc4b424d392bc14cd6b7dfb79e5 Mon Sep 17 00:00:00 2001
From: Francis Visoiu Mistrih <890283+francisvm@users.noreply.github.com>
Date: Fri, 12 Apr 2024 08:45:47 -0700
Subject: [PATCH 264/886] [RISCV] Remove mayLoad = 1 from store-conditional
 (#88470)

lr.[wd]: mayLoad = 1, mayStore = 0
sc.[wd]: mayLoad = 0, mayStore = 1

all other AMOs: mayLoad = 1, mayStore = 1
---
 llvm/lib/Target/RISCV/RISCVInstrInfoA.td | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoA.td b/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
index 36842ceb49bfb..814e0ddf111e6 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
@@ -30,6 +30,19 @@ multiclass LR_r_aq_rl<bits<3> funct3, string opcodestr> {
   def _AQ_RL : LR_r<1, 1, funct3, opcodestr # ".aqrl">;
 }
 
+let hasSideEffects = 0, mayLoad = 0, mayStore = 1 in
+class SC_r<bit aq, bit rl, bits<3> funct3, string opcodestr>
+    : RVInstRAtomic<0b00011, aq, rl, funct3, OPC_AMO,
+                    (outs GPR:$rd), (ins GPRMemZeroOffset:$rs1, GPR:$rs2),
+                    opcodestr, "$rd, $rs2, $rs1">;
+
+multiclass SC_r_aq_rl<bits<3> funct3, string opcodestr> {
+  def ""     : SC_r<0, 0, funct3, opcodestr>;
+  def _AQ    : SC_r<1, 0, funct3, opcodestr # ".aq">;
+  def _RL    : SC_r<0, 1, funct3, opcodestr # ".rl">;
+  def _AQ_RL : SC_r<1, 1, funct3, opcodestr # ".aqrl">;
+}
+
 let hasSideEffects = 0, mayLoad = 1, mayStore = 1 in
 class AMO_rr<bits<5> funct5, bit aq, bit rl, bits<3> funct3, string opcodestr>
     : RVInstRAtomic<funct5, aq, rl, funct3, OPC_AMO,
@@ -49,7 +62,7 @@ multiclass AMO_rr_aq_rl<bits<5> funct5, bits<3> funct3, string opcodestr> {
 
 let Predicates = [HasStdExtAOrZalrsc], IsSignExtendingOpW = 1 in {
 defm LR_W       : LR_r_aq_rl<0b010, "lr.w">, Sched<[WriteAtomicLDW, ReadAtomicLDW]>;
-defm SC_W       : AMO_rr_aq_rl<0b00011, 0b010, "sc.w">,
+defm SC_W       : SC_r_aq_rl<0b010, "sc.w">,
                   Sched<[WriteAtomicSTW, ReadAtomicSTW, ReadAtomicSTW]>;
 } // Predicates = [HasStdExtAOrZalrsc], IsSignExtendingOpW = 1
 
@@ -76,7 +89,7 @@ defm AMOMAXU_W  : AMO_rr_aq_rl<0b11100, 0b010, "amomaxu.w">,
 
 let Predicates = [HasStdExtAOrZalrsc, IsRV64] in {
 defm LR_D       : LR_r_aq_rl<0b011, "lr.d">, Sched<[WriteAtomicLDD, ReadAtomicLDD]>;
-defm SC_D       : AMO_rr_aq_rl<0b00011, 0b011, "sc.d">,
+defm SC_D       : SC_r_aq_rl<0b011, "sc.d">,
                   Sched<[WriteAtomicSTD, ReadAtomicSTD, ReadAtomicSTD]>;
 } // Predicates = [HasStdExtAOrZalrsc, IsRV64]
 

From e0a628715a8464e220c8ba9e9aaaf2561139198a Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Fri, 12 Apr 2024 23:47:20 +0800
Subject: [PATCH 265/886] [ValueTracking] Convert `isKnownNonZero` to use
 SimplifyQuery (#85863)

This patch converts `isKnownNonZero` to use SimplifyQuery. Then we can
use the context information from `DomCondCache`.

Fixes https://github.com/llvm/llvm-project/issues/85823.
Alive2: https://alive2.llvm.org/ce/z/QUvHVj
---
 clang/lib/CodeGen/CGCall.cpp                  |   3 +-
 llvm/include/llvm/Analysis/ValueTracking.h    |   6 +-
 llvm/lib/Analysis/BasicAliasAnalysis.cpp      |   3 +-
 llvm/lib/Analysis/InstructionSimplify.cpp     |  29 ++--
 llvm/lib/Analysis/LazyValueInfo.cpp           |   5 +-
 llvm/lib/Analysis/Loads.cpp                   |   6 +-
 llvm/lib/Analysis/ScalarEvolution.cpp         |   2 +-
 llvm/lib/Analysis/ValueTracking.cpp           |  17 +--
 llvm/lib/CodeGen/CodeGenPrepare.cpp           |   2 +-
 .../Transforms/IPO/AttributorAttributes.cpp   |   5 +-
 llvm/lib/Transforms/IPO/FunctionAttrs.cpp     |   2 +-
 .../InstCombine/InstCombineAddSub.cpp         |   2 +-
 .../InstCombine/InstCombineAndOrXor.cpp       |   8 +-
 .../InstCombine/InstCombineCalls.cpp          |  13 +-
 .../InstCombine/InstCombineCompares.cpp       |  21 ++-
 .../Transforms/InstCombine/InstCombinePHI.cpp |   3 +-
 .../InstCombine/InstructionCombining.cpp      |   2 +-
 .../Instrumentation/MemorySanitizer.cpp       |   4 +-
 .../Utils/PromoteMemoryToRegister.cpp         |   2 +-
 .../lib/Transforms/Utils/SimplifyLibCalls.cpp |  20 +--
 .../Transforms/Vectorize/VectorCombine.cpp    |   3 +-
 llvm/test/Transforms/InstCombine/icmp-dom.ll  | 131 ++++++++++++++++++
 llvm/unittests/Analysis/ValueTrackingTest.cpp |  15 +-
 23 files changed, 213 insertions(+), 91 deletions(-)

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 3f5463a9a70e9..7a0bc6fa77b88 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -4124,7 +4124,8 @@ static bool isProvablyNull(llvm::Value *addr) {
 }
 
 static bool isProvablyNonNull(Address Addr, CodeGenFunction &CGF) {
-  return llvm::isKnownNonZero(Addr.getBasePointer(), CGF.CGM.getDataLayout());
+  return llvm::isKnownNonZero(Addr.getBasePointer(), /*Depth=*/0,
+                              CGF.CGM.getDataLayout());
 }
 
 /// Emit the actual writing-back of a writeback.
diff --git a/llvm/include/llvm/Analysis/ValueTracking.h b/llvm/include/llvm/Analysis/ValueTracking.h
index 3970efba18cc8..9db0894162afc 100644
--- a/llvm/include/llvm/Analysis/ValueTracking.h
+++ b/llvm/include/llvm/Analysis/ValueTracking.h
@@ -124,11 +124,7 @@ bool isOnlyUsedInZeroEqualityComparison(const Instruction *CxtI);
 /// specified, perform context-sensitive analysis and return true if the
 /// pointer couldn't possibly be null at the specified instruction.
 /// Supports values with integer or pointer type and vectors of integers.
-bool isKnownNonZero(const Value *V, const DataLayout &DL, unsigned Depth = 0,
-                    AssumptionCache *AC = nullptr,
-                    const Instruction *CxtI = nullptr,
-                    const DominatorTree *DT = nullptr,
-                    bool UseInstrInfo = true);
+bool isKnownNonZero(const Value *V, unsigned Depth, const SimplifyQuery &Q);
 
 /// Return true if the two given values are negation.
 /// Currently can recoginze Value pair:
diff --git a/llvm/lib/Analysis/BasicAliasAnalysis.cpp b/llvm/lib/Analysis/BasicAliasAnalysis.cpp
index 8dfc4b2a1cb1f..b082dfe8fbd21 100644
--- a/llvm/lib/Analysis/BasicAliasAnalysis.cpp
+++ b/llvm/lib/Analysis/BasicAliasAnalysis.cpp
@@ -1283,7 +1283,8 @@ AliasResult BasicAAResult::aliasGEP(
     // VarIndex = Scale*V.
     const VariableGEPIndex &Var = DecompGEP1.VarIndices[0];
     if (Var.Val.TruncBits == 0 &&
-        isKnownNonZero(Var.Val.V, DL, 0, &AC, Var.CxtI, DT)) {
+        isKnownNonZero(Var.Val.V, /*Depth=*/0,
+                       SimplifyQuery(DL, DT, &AC, Var.CxtI))) {
       // Check if abs(V*Scale) >= abs(Scale) holds in the presence of
       // potentially wrapping math.
       auto MultiplyByScaleNoWrap = [](const VariableGEPIndex &Var) {
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 1aaa4d658803b..e83657df20f2a 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -1586,12 +1586,10 @@ static Value *simplifyUnsignedRangeCheck(ICmpInst *ZeroICmp,
     if (match(UnsignedICmp,
               m_c_ICmp(UnsignedPred, m_Specific(Y), m_Specific(A)))) {
       if (UnsignedPred == ICmpInst::ICMP_UGE && IsAnd &&
-          EqPred == ICmpInst::ICMP_NE &&
-          isKnownNonZero(B, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+          EqPred == ICmpInst::ICMP_NE && isKnownNonZero(B, /*Depth=*/0, Q))
         return UnsignedICmp;
       if (UnsignedPred == ICmpInst::ICMP_ULT && !IsAnd &&
-          EqPred == ICmpInst::ICMP_EQ &&
-          isKnownNonZero(B, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+          EqPred == ICmpInst::ICMP_EQ && isKnownNonZero(B, /*Depth=*/0, Q))
         return UnsignedICmp;
     }
   }
@@ -1609,13 +1607,13 @@ static Value *simplifyUnsignedRangeCheck(ICmpInst *ZeroICmp,
   // X > Y && Y == 0  -->  Y == 0  iff X != 0
   // X > Y || Y == 0  -->  X > Y   iff X != 0
   if (UnsignedPred == ICmpInst::ICMP_UGT && EqPred == ICmpInst::ICMP_EQ &&
-      isKnownNonZero(X, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+      isKnownNonZero(X, /*Depth=*/0, Q))
     return IsAnd ? ZeroICmp : UnsignedICmp;
 
   // X <= Y && Y != 0  -->  X <= Y  iff X != 0
   // X <= Y || Y != 0  -->  Y != 0  iff X != 0
   if (UnsignedPred == ICmpInst::ICMP_ULE && EqPred == ICmpInst::ICMP_NE &&
-      isKnownNonZero(X, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+      isKnownNonZero(X, /*Depth=*/0, Q))
     return IsAnd ? UnsignedICmp : ZeroICmp;
 
   // The transforms below here are expected to be handled more generally with
@@ -2821,11 +2819,10 @@ static Constant *computePointerICmp(CmpInst::Predicate Pred, Value *LHS,
     // the other operand can not be based on the alloc - if it were, then
     // the cmp itself would be a capture.
     Value *MI = nullptr;
-    if (isAllocLikeFn(LHS, TLI) &&
-        llvm::isKnownNonZero(RHS, DL, 0, nullptr, CxtI, DT))
+    if (isAllocLikeFn(LHS, TLI) && llvm::isKnownNonZero(RHS, /*Depth=*/0, Q))
       MI = LHS;
     else if (isAllocLikeFn(RHS, TLI) &&
-             llvm::isKnownNonZero(LHS, DL, 0, nullptr, CxtI, DT))
+             llvm::isKnownNonZero(LHS, /*Depth=*/0, Q))
       MI = RHS;
     if (MI) {
       // FIXME: This is incorrect, see PR54002. While we can assume that the
@@ -2981,12 +2978,12 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     return getTrue(ITy);
   case ICmpInst::ICMP_EQ:
   case ICmpInst::ICMP_ULE:
-    if (isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT, Q.IIQ.UseInstrInfo))
+    if (isKnownNonZero(LHS, /*Depth=*/0, Q))
       return getFalse(ITy);
     break;
   case ICmpInst::ICMP_NE:
   case ICmpInst::ICMP_UGT:
-    if (isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT, Q.IIQ.UseInstrInfo))
+    if (isKnownNonZero(LHS, /*Depth=*/0, Q))
       return getTrue(ITy);
     break;
   case ICmpInst::ICMP_SLT: {
@@ -3001,8 +2998,7 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     KnownBits LHSKnown = computeKnownBits(LHS, /* Depth */ 0, Q);
     if (LHSKnown.isNegative())
       return getTrue(ITy);
-    if (LHSKnown.isNonNegative() &&
-        isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
+    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, /*Depth=*/0, Q))
       return getFalse(ITy);
     break;
   }
@@ -3018,8 +3014,7 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     KnownBits LHSKnown = computeKnownBits(LHS, /* Depth */ 0, Q);
     if (LHSKnown.isNegative())
       return getFalse(ITy);
-    if (LHSKnown.isNonNegative() &&
-        isKnownNonZero(LHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT))
+    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, /*Depth=*/0, Q))
       return getTrue(ITy);
     break;
   }
@@ -3172,7 +3167,7 @@ static Value *simplifyICmpWithBinOpOnLHS(CmpInst::Predicate Pred,
   const APInt *C;
   if ((match(LBO, m_LShr(m_Specific(RHS), m_APInt(C))) && *C != 0) ||
       (match(LBO, m_UDiv(m_Specific(RHS), m_APInt(C))) && *C != 1)) {
-    if (isKnownNonZero(RHS, Q.DL, 0, Q.AC, Q.CxtI, Q.DT)) {
+    if (isKnownNonZero(RHS, /*Depth=*/0, Q)) {
       switch (Pred) {
       default:
         break;
@@ -3405,7 +3400,7 @@ static Value *simplifyICmpWithBinOp(CmpInst::Predicate Pred, Value *LHS,
       bool NUW = Q.IIQ.hasNoUnsignedWrap(LBO) && Q.IIQ.hasNoUnsignedWrap(RBO);
       bool NSW = Q.IIQ.hasNoSignedWrap(LBO) && Q.IIQ.hasNoSignedWrap(RBO);
       if (!NUW || (ICmpInst::isSigned(Pred) && !NSW) ||
-          !isKnownNonZero(LBO->getOperand(0), Q.DL))
+          !isKnownNonZero(LBO->getOperand(0), /*Depth=*/0, Q))
         break;
       if (Value *V = simplifyICmpInst(Pred, LBO->getOperand(1),
                                       RBO->getOperand(1), Q, MaxRecurse - 1))
diff --git a/llvm/lib/Analysis/LazyValueInfo.cpp b/llvm/lib/Analysis/LazyValueInfo.cpp
index 6cded828c25f4..3223b0564e6c9 100644
--- a/llvm/lib/Analysis/LazyValueInfo.cpp
+++ b/llvm/lib/Analysis/LazyValueInfo.cpp
@@ -645,7 +645,7 @@ LazyValueInfoImpl::solveBlockValueImpl(Value *Val, BasicBlock *BB) {
   // instruction is placed, even if it could legally be hoisted much higher.
   // That is unfortunate.
   PointerType *PT = dyn_cast<PointerType>(BBI->getType());
-  if (PT && isKnownNonZero(BBI, DL))
+  if (PT && isKnownNonZero(BBI, /*Depth=*/0, DL))
     return ValueLatticeElement::getNot(ConstantPointerNull::get(PT));
 
   if (BBI->getType()->isIntegerTy()) {
@@ -1863,7 +1863,8 @@ LazyValueInfo::getPredicateAt(unsigned Pred, Value *V, Constant *C,
   Module *M = CxtI->getModule();
   const DataLayout &DL = M->getDataLayout();
   if (V->getType()->isPointerTy() && C->isNullValue() &&
-      isKnownNonZero(V->stripPointerCastsSameRepresentation(), DL)) {
+      isKnownNonZero(V->stripPointerCastsSameRepresentation(), /*Depth=*/0,
+                     DL)) {
     if (Pred == ICmpInst::ICMP_EQ)
       return LazyValueInfo::False;
     else if (Pred == ICmpInst::ICMP_NE)
diff --git a/llvm/lib/Analysis/Loads.cpp b/llvm/lib/Analysis/Loads.cpp
index 5916d2ab48ece..b5403408cf2ab 100644
--- a/llvm/lib/Analysis/Loads.cpp
+++ b/llvm/lib/Analysis/Loads.cpp
@@ -99,7 +99,8 @@ static bool isDereferenceableAndAlignedPointer(
                                                           CheckForFreed));
   if (KnownDerefBytes.getBoolValue() && KnownDerefBytes.uge(Size) &&
       !CheckForFreed)
-    if (!CheckForNonNull || isKnownNonZero(V, DL, 0, AC, CtxI, DT)) {
+    if (!CheckForNonNull ||
+        isKnownNonZero(V, /*Depth=*/0, SimplifyQuery(DL, DT, AC, CtxI))) {
       // As we recursed through GEPs to get here, we've incrementally checked
       // that each step advanced by a multiple of the alignment. If our base is
       // properly aligned, then the original offset accessed must also be.
@@ -133,7 +134,8 @@ static bool isDereferenceableAndAlignedPointer(
     if (getObjectSize(V, ObjSize, DL, TLI, Opts)) {
       APInt KnownDerefBytes(Size.getBitWidth(), ObjSize);
       if (KnownDerefBytes.getBoolValue() && KnownDerefBytes.uge(Size) &&
-          isKnownNonZero(V, DL, 0, AC, CtxI, DT) && !V->canBeFreed()) {
+          isKnownNonZero(V, /*Depth=*/0, SimplifyQuery(DL, DT, AC, CtxI)) &&
+          !V->canBeFreed()) {
         // As we recursed through GEPs to get here, we've incrementally
         // checked that each step advanced by a multiple of the alignment. If
         // our base is properly aligned, then the original offset accessed
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 95440dda3b4c0..1c98b0295e525 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -6900,7 +6900,7 @@ const ConstantRange &ScalarEvolution::getRangeRef(
         uint64_t Rem = MaxVal.urem(Align);
         MaxVal -= APInt(BitWidth, Rem);
         APInt MinVal = APInt::getZero(BitWidth);
-        if (llvm::isKnownNonZero(V, DL))
+        if (llvm::isKnownNonZero(V, /*Depth=*/0, DL))
           MinVal = Align;
         ConservativeResult = ConservativeResult.intersectWith(
             ConstantRange::getNonEmpty(MinVal, MaxVal + 1), RangeType);
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index ca257cc6aefdf..b83e7e6769c20 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -274,16 +274,6 @@ bool llvm::isKnownToBeAPowerOfTwo(const Value *V, const DataLayout &DL,
 static bool isKnownNonZero(const Value *V, const APInt &DemandedElts,
                            unsigned Depth, const SimplifyQuery &Q);
 
-static bool isKnownNonZero(const Value *V, unsigned Depth,
-                           const SimplifyQuery &Q);
-
-bool llvm::isKnownNonZero(const Value *V, const DataLayout &DL, unsigned Depth,
-                          AssumptionCache *AC, const Instruction *CxtI,
-                          const DominatorTree *DT, bool UseInstrInfo) {
-  return ::isKnownNonZero(
-      V, Depth, SimplifyQuery(DL, DT, AC, safeCxtI(V, CxtI), UseInstrInfo));
-}
-
 bool llvm::isKnownNonNegative(const Value *V, const SimplifyQuery &SQ,
                               unsigned Depth) {
   return computeKnownBits(V, Depth, SQ).isNonNegative();
@@ -298,7 +288,7 @@ bool llvm::isKnownPositive(const Value *V, const SimplifyQuery &SQ,
   // this updated.
   KnownBits Known = computeKnownBits(V, Depth, SQ);
   return Known.isNonNegative() &&
-         (Known.isNonZero() || ::isKnownNonZero(V, Depth, SQ));
+         (Known.isNonZero() || isKnownNonZero(V, Depth, SQ));
 }
 
 bool llvm::isKnownNegative(const Value *V, const SimplifyQuery &SQ,
@@ -3093,11 +3083,12 @@ bool isKnownNonZero(const Value *V, const APInt &DemandedElts, unsigned Depth,
   return false;
 }
 
-bool isKnownNonZero(const Value *V, unsigned Depth, const SimplifyQuery &Q) {
+bool llvm::isKnownNonZero(const Value *V, unsigned Depth,
+                          const SimplifyQuery &Q) {
   auto *FVTy = dyn_cast<FixedVectorType>(V->getType());
   APInt DemandedElts =
       FVTy ? APInt::getAllOnes(FVTy->getNumElements()) : APInt(1, 1);
-  return isKnownNonZero(V, DemandedElts, Depth, Q);
+  return ::isKnownNonZero(V, DemandedElts, Depth, Q);
 }
 
 /// If the pair of operators are the same invertible function, return the
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index e657872c38284..22dbb3198a9f1 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -2314,7 +2314,7 @@ static bool despeculateCountZeros(IntrinsicInst *CountZeros,
 
   // Bail if the value is never zero.
   Use &Op = CountZeros->getOperandUse(0);
-  if (isKnownNonZero(Op, *DL))
+  if (isKnownNonZero(Op, /*Depth=*/0, *DL))
     return false;
 
   // The intrinsic will be sunk behind a compare against zero and branch.
diff --git a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
index ff680e998e71d..f27d8d64a1040 100644
--- a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
+++ b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
@@ -2452,8 +2452,9 @@ bool AANonNull::isImpliedByIR(Attributor &A, const IRPosition &IRP,
   }
 
   if (llvm::any_of(Worklist, [&](AA::ValueAndContext VAC) {
-        return !isKnownNonZero(VAC.getValue(), A.getDataLayout(), 0, AC,
-                               VAC.getCtxI(), DT);
+        return !isKnownNonZero(
+            VAC.getValue(), /*Depth=*/0,
+            SimplifyQuery(A.getDataLayout(), DT, AC, VAC.getCtxI()));
       }))
     return false;
 
diff --git a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
index 7ebf265e17ba1..14612b251d1a4 100644
--- a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
+++ b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
@@ -1175,7 +1175,7 @@ static bool isReturnNonNull(Function *F, const SCCNodeSet &SCCNodes,
     Value *RetVal = FlowsToReturn[i];
 
     // If this value is locally known to be non-null, we're good
-    if (isKnownNonZero(RetVal, DL))
+    if (isKnownNonZero(RetVal, /*Depth=*/0, DL))
       continue;
 
     // Otherwise, we need to look upwards since we can't make any local
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
index fa1e2280991fd..07c50d866544b 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -988,7 +988,7 @@ Instruction *InstCombinerImpl::foldAddWithConstant(BinaryOperator &Add) {
   if (C->isOne()) {
     if (match(Op0, m_ZExt(m_Add(m_Value(X), m_AllOnes())))) {
       const SimplifyQuery Q = SQ.getWithInstruction(&Add);
-      if (llvm::isKnownNonZero(X, DL, 0, Q.AC, Q.CxtI, Q.DT))
+      if (llvm::isKnownNonZero(X, /*Depth=*/0, Q))
         return new ZExtInst(X, Ty);
     }
   }
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index c0cf1a7db7260..2c0c4ee46e809 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -1031,10 +1031,6 @@ static Value *foldUnsignedUnderflowCheck(ICmpInst *ZeroICmp,
       !ICmpInst::isEquality(EqPred))
     return nullptr;
 
-  auto IsKnownNonZero = [&](Value *V) {
-    return isKnownNonZero(V, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT);
-  };
-
   ICmpInst::Predicate UnsignedPred;
 
   Value *A, *B;
@@ -1043,9 +1039,9 @@ static Value *foldUnsignedUnderflowCheck(ICmpInst *ZeroICmp,
       match(ZeroCmpOp, m_c_Add(m_Specific(A), m_Value(B))) &&
       (ZeroICmp->hasOneUse() || UnsignedICmp->hasOneUse())) {
     auto GetKnownNonZeroAndOther = [&](Value *&NonZero, Value *&Other) {
-      if (!IsKnownNonZero(NonZero))
+      if (!isKnownNonZero(NonZero, /*Depth=*/0, Q))
         std::swap(NonZero, Other);
-      return IsKnownNonZero(NonZero);
+      return isKnownNonZero(NonZero, /*Depth=*/0, Q);
     };
 
     // Given  ZeroCmpOp = (A + B)
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index b1017f4d6bc17..b35ab8751a211 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -601,8 +601,8 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) {
   // then change the 'ZeroIsPoison' parameter to 'true'
   // because we know the zero behavior can't affect the result.
   if (!Known.One.isZero() ||
-      isKnownNonZero(Op0, IC.getDataLayout(), 0, &IC.getAssumptionCache(), &II,
-                     &IC.getDominatorTree())) {
+      isKnownNonZero(Op0, /*Depth=*/0,
+                     IC.getSimplifyQuery().getWithInstruction(&II))) {
     if (!match(II.getArgOperand(1), m_One()))
       return IC.replaceOperand(II, 1, IC.Builder.getTrue());
   }
@@ -2061,7 +2061,8 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
     // See if we can deduce non-null.
     if (!CI.hasRetAttr(Attribute::NonNull) &&
         (Known.isNonZero() ||
-         isKnownNonZero(II, DL, /*Depth*/ 0, &AC, II, &DT))) {
+         isKnownNonZero(II, /*Depth=*/0,
+                        getSimplifyQuery().getWithInstruction(II)))) {
       CI.addRetAttr(Attribute::NonNull);
       Changed = true;
     }
@@ -3648,7 +3649,8 @@ Instruction *InstCombinerImpl::visitCallBase(CallBase &Call) {
   for (Value *V : Call.args()) {
     if (V->getType()->isPointerTy() &&
         !Call.paramHasAttr(ArgNo, Attribute::NonNull) &&
-        isKnownNonZero(V, DL, 0, &AC, &Call, &DT))
+        isKnownNonZero(V, /*Depth=*/0,
+                       getSimplifyQuery().getWithInstruction(&Call)))
       ArgNos.push_back(ArgNo);
     ArgNo++;
   }
@@ -3828,7 +3830,8 @@ Instruction *InstCombinerImpl::visitCallBase(CallBase &Call) {
 
         // isKnownNonNull -> nonnull attribute
         if (!GCR.hasRetAttr(Attribute::NonNull) &&
-            isKnownNonZero(DerivedPtr, DL, 0, &AC, &Call, &DT)) {
+            isKnownNonZero(DerivedPtr, /*Depth=*/0,
+                           getSimplifyQuery().getWithInstruction(&Call))) {
           GCR.addRetAttr(Attribute::NonNull);
           // We discovered new fact, re-check users.
           Worklist.pushUsersToWorkList(GCR);
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 7292bb62702aa..90550cdbdf891 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -1273,12 +1273,12 @@ Instruction *InstCombinerImpl::foldICmpWithZero(ICmpInst &Cmp) {
 
       // if X non-zero and NoOverflow(X * Y)
       //    (icmp eq/ne Y)
-      if (!XKnown.One.isZero() || isKnownNonZero(X, DL, 0, Q.AC, Q.CxtI, Q.DT))
+      if (!XKnown.One.isZero() || isKnownNonZero(X, /*Depth=*/0, Q))
         return new ICmpInst(Pred, Y, Cmp.getOperand(1));
 
       // if Y non-zero and NoOverflow(X * Y)
       //    (icmp eq/ne X)
-      if (!YKnown.One.isZero() || isKnownNonZero(Y, DL, 0, Q.AC, Q.CxtI, Q.DT))
+      if (!YKnown.One.isZero() || isKnownNonZero(Y, /*Depth=*/0, Q))
         return new ICmpInst(Pred, X, Cmp.getOperand(1));
     }
     // Note, we are skipping cases:
@@ -3087,7 +3087,7 @@ Instruction *InstCombinerImpl::foldICmpAddConstant(ICmpInst &Cmp,
   // (X + -1) <u C --> X <=u C (if X is never null)
   if (Pred == CmpInst::ICMP_ULT && C2->isAllOnes()) {
     const SimplifyQuery Q = SQ.getWithInstruction(&Cmp);
-    if (llvm::isKnownNonZero(X, DL, 0, Q.AC, Q.CxtI, Q.DT))
+    if (llvm::isKnownNonZero(X, /*Depth=*/0, Q))
       return new ICmpInst(ICmpInst::ICMP_ULE, X, ConstantInt::get(Ty, C));
   }
 
@@ -4275,8 +4275,7 @@ static Value *foldICmpWithLowBitMaskedVal(ICmpInst::Predicate Pred, Value *Op0,
 
       // Look for: x & ~Mask pred ~Mask
       if (isMaskOrZero(X, /*Not=*/true, Q)) {
-        return !ICmpInst::isSigned(Pred) ||
-               isKnownNonZero(X, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT);
+        return !ICmpInst::isSigned(Pred) || isKnownNonZero(X, /*Depth=*/0, Q);
       }
       return false;
     }
@@ -4780,8 +4779,7 @@ static Instruction *foldICmpXorXX(ICmpInst &I, const SimplifyQuery &Q,
   // icmp (X ^ Y_NonZero) s>= X --> icmp (X ^ Y_NonZero) s> X
   // icmp (X ^ Y_NonZero) s<= X --> icmp (X ^ Y_NonZero) s< X
   CmpInst::Predicate PredOut = CmpInst::getStrictPredicate(Pred);
-  if (PredOut != Pred &&
-      isKnownNonZero(A, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+  if (PredOut != Pred && isKnownNonZero(A, /*Depth=*/0, Q))
     return new ICmpInst(PredOut, Op0, Op1);
 
   return nullptr;
@@ -5064,11 +5062,11 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
     return new ICmpInst(Pred, C, D);
   // (A - B) u>=/u< A --> B u>/u<= A  iff B != 0
   if (A == Op1 && (Pred == ICmpInst::ICMP_UGE || Pred == ICmpInst::ICMP_ULT) &&
-      isKnownNonZero(B, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+      isKnownNonZero(B, /*Depth=*/0, Q))
     return new ICmpInst(CmpInst::getFlippedStrictnessPredicate(Pred), B, A);
   // C u<=/u> (C - D) --> C u</u>= D  iff B != 0
   if (C == Op0 && (Pred == ICmpInst::ICMP_ULE || Pred == ICmpInst::ICMP_UGT) &&
-      isKnownNonZero(D, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT))
+      isKnownNonZero(D, /*Depth=*/0, Q))
     return new ICmpInst(CmpInst::getFlippedStrictnessPredicate(Pred), C, D);
 
   // icmp (A-B), (C-B) -> icmp A, C for equalities or if there is no overflow.
@@ -5110,14 +5108,13 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
         //    X * Z eq/ne Y * Z -> X eq/ne Y
         if (ZKnown.countMaxTrailingZeros() == 0)
           return new ICmpInst(Pred, X, Y);
-        NonZero = !ZKnown.One.isZero() ||
-                  isKnownNonZero(Z, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT);
+        NonZero = !ZKnown.One.isZero() || isKnownNonZero(Z, /*Depth=*/0, Q);
         // if Z != 0 and nsw(X * Z) and nsw(Y * Z)
         //    X * Z eq/ne Y * Z -> X eq/ne Y
         if (NonZero && BO0 && BO1 && Op0HasNSW && Op1HasNSW)
           return new ICmpInst(Pred, X, Y);
       } else
-        NonZero = isKnownNonZero(Z, Q.DL, /*Depth=*/0, Q.AC, Q.CxtI, Q.DT);
+        NonZero = isKnownNonZero(Z, /*Depth=*/0, Q);
 
       // If Z != 0 and nuw(X * Z) and nuw(Y * Z)
       //    X * Z u{lt/le/gt/ge}/eq/ne Y * Z -> X u{lt/le/gt/ge}/eq/ne Y
diff --git a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
index 46bca4b722a03..9838e2aa9f3a2 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
@@ -1537,7 +1537,8 @@ Instruction *InstCombinerImpl::visitPHINode(PHINode &PN) {
       for (unsigned I = 0, E = PN.getNumIncomingValues(); I != E; ++I) {
         Instruction *CtxI = PN.getIncomingBlock(I)->getTerminator();
         Value *VA = PN.getIncomingValue(I);
-        if (isKnownNonZero(VA, DL, 0, &AC, CtxI, &DT)) {
+        if (isKnownNonZero(VA, 0,
+                           getSimplifyQuery().getWithInstruction(CtxI))) {
           if (!NonZeroConst)
             NonZeroConst = getAnyNonZeroConstInt(PN);
           if (NonZeroConst != VA) {
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 7c40fb4fc8608..4cfafa7ac8006 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1431,7 +1431,7 @@ Instruction *InstCombinerImpl::foldFBinOpOfIntCastsFromSign(
     if (OpsKnown[OpNo].hasKnownBits() &&
         OpsKnown[OpNo].getKnownBits(SQ).isNonZero())
       return true;
-    return isKnownNonZero(IntOps[OpNo], SQ.DL);
+    return isKnownNonZero(IntOps[OpNo], /*Depth=*/0, SQ);
   };
 
   auto IsNonNeg = [&](unsigned OpNo) -> bool {
diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index ee3531bbd68df..a72b0ee9a08e0 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -1281,7 +1281,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
         // ignored.
         return;
       }
-      if (llvm::isKnownNonZero(ConvertedShadow, DL)) {
+      if (llvm::isKnownNonZero(ConvertedShadow, /*Depth=*/0, DL)) {
         // Copy origin as the value is definitely uninitialized.
         paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
                     OriginAlignment);
@@ -1427,7 +1427,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
           // Skip, value is initialized or const shadow is ignored.
           continue;
         }
-        if (llvm::isKnownNonZero(ConvertedShadow, DL)) {
+        if (llvm::isKnownNonZero(ConvertedShadow, /*Depth=*/0, DL)) {
           // Report as the value is definitely uninitialized.
           insertWarningFn(IRB, ShadowData.Origin);
           if (!MS.Recover)
diff --git a/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp b/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
index adcf161b313b2..f376b5f7d68d4 100644
--- a/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
+++ b/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
@@ -459,7 +459,7 @@ static void convertMetadataToAssumes(LoadInst *LI, Value *Val,
   // we can only do this if the value is known non-poison.
   if (AC && LI->getMetadata(LLVMContext::MD_nonnull) &&
       LI->getMetadata(LLVMContext::MD_noundef) &&
-      !isKnownNonZero(Val, DL, 0, AC, LI, DT))
+      !isKnownNonZero(Val, /*Depth=*/0, SimplifyQuery(DL, DT, AC, LI)))
     addAssumeNonNull(AC, LI);
 }
 
diff --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
index 2e68a9c01898c..7e9e91606fe22 100644
--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -305,7 +305,7 @@ static void annotateNonNullAndDereferenceable(CallInst *CI, ArrayRef<unsigned> A
   if (ConstantInt *LenC = dyn_cast<ConstantInt>(Size)) {
     annotateNonNullNoUndefBasedOnAccess(CI, ArgNos);
     annotateDereferenceableBytes(CI, ArgNos, LenC->getZExtValue());
-  } else if (isKnownNonZero(Size, DL)) {
+  } else if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
     annotateNonNullNoUndefBasedOnAccess(CI, ArgNos);
     const APInt *X, *Y;
     uint64_t DerefMin = 1;
@@ -394,7 +394,7 @@ Value *LibCallSimplifier::optimizeStrNCat(CallInst *CI, IRBuilderBase &B) {
   Value *Size = CI->getArgOperand(2);
   uint64_t Len;
   annotateNonNullNoUndefBasedOnAccess(CI, 0);
-  if (isKnownNonZero(Size, DL))
+  if (isKnownNonZero(Size, /*Depth=*/0, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 1);
 
   // We don't do anything if length is not constant.
@@ -613,7 +613,7 @@ Value *LibCallSimplifier::optimizeStrNCmp(CallInst *CI, IRBuilderBase &B) {
   if (Str1P == Str2P) // strncmp(x,x,n)  -> 0
     return ConstantInt::get(CI->getType(), 0);
 
-  if (isKnownNonZero(Size, DL))
+  if (isKnownNonZero(Size, /*Depth=*/0, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, {0, 1});
   // Get the length argument if it is constant.
   uint64_t Length;
@@ -749,7 +749,7 @@ Value *LibCallSimplifier::optimizeStpCpy(CallInst *CI, IRBuilderBase &B) {
 
 Value *LibCallSimplifier::optimizeStrLCpy(CallInst *CI, IRBuilderBase &B) {
   Value *Size = CI->getArgOperand(2);
-  if (isKnownNonZero(Size, DL))
+  if (isKnownNonZero(Size, /*Depth=*/0, DL))
     // Like snprintf, the function stores into the destination only when
     // the size argument is nonzero.
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
@@ -833,7 +833,7 @@ Value *LibCallSimplifier::optimizeStringNCpy(CallInst *CI, bool RetEnd,
   Value *Src = CI->getArgOperand(1);
   Value *Size = CI->getArgOperand(2);
 
-  if (isKnownNonZero(Size, DL)) {
+  if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
     // Both st{p,r}ncpy(D, S, N) access the source and destination arrays
     // only when N is nonzero.
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
@@ -926,7 +926,7 @@ Value *LibCallSimplifier::optimizeStringLength(CallInst *CI, IRBuilderBase &B,
   Type *CharTy = B.getIntNTy(CharSize);
 
   if (isOnlyUsedInZeroEqualityComparison(CI) &&
-      (!Bound || isKnownNonZero(Bound, DL))) {
+      (!Bound || isKnownNonZero(Bound, /*Depth=*/0, DL))) {
     // Fold strlen:
     //   strlen(x) != 0 --> *x != 0
     //   strlen(x) == 0 --> *x == 0
@@ -1047,7 +1047,7 @@ Value *LibCallSimplifier::optimizeStrNLen(CallInst *CI, IRBuilderBase &B) {
   if (Value *V = optimizeStringLength(CI, B, 8, Bound))
     return V;
 
-  if (isKnownNonZero(Bound, DL))
+  if (isKnownNonZero(Bound, /*Depth=*/0, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
   return nullptr;
 }
@@ -1291,7 +1291,7 @@ Value *LibCallSimplifier::optimizeMemChr(CallInst *CI, IRBuilderBase &B) {
   Value *SrcStr = CI->getArgOperand(0);
   Value *Size = CI->getArgOperand(2);
 
-  if (isKnownNonZero(Size, DL)) {
+  if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
     if (isOnlyUsedInEqualityComparison(CI, SrcStr))
       return memChrToCharCompare(CI, Size, B, DL);
@@ -2976,7 +2976,7 @@ Value *LibCallSimplifier::optimizeStrToInt(CallInst *CI, IRBuilderBase &B,
     // It would be readonly too, except that it still may write to errno.
     CI->addParamAttr(0, Attribute::NoCapture);
     EndPtr = nullptr;
-  } else if (!isKnownNonZero(EndPtr, DL))
+  } else if (!isKnownNonZero(EndPtr, /*Depth=*/0, DL))
     return nullptr;
 
   StringRef Str;
@@ -3402,7 +3402,7 @@ Value *LibCallSimplifier::optimizeSnPrintF(CallInst *CI, IRBuilderBase &B) {
     return V;
   }
 
-  if (isKnownNonZero(CI->getOperand(1), DL))
+  if (isKnownNonZero(CI->getOperand(1), /*Depth=*/0, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
   return nullptr;
 }
diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 653e0ff7a3ece..e0e2f50c89ada 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -885,7 +885,8 @@ bool VectorCombine::scalarizeVPIntrinsic(Instruction &I) {
   else
     SafeToSpeculate = isSafeToSpeculativelyExecuteWithOpcode(
         *FunctionalOpcode, &VPI, nullptr, &AC, &DT);
-  if (!SafeToSpeculate && !isKnownNonZero(EVL, *DL, 0, &AC, &VPI, &DT))
+  if (!SafeToSpeculate &&
+      !isKnownNonZero(EVL, /*Depth=*/0, SimplifyQuery(*DL, &DT, &AC, &VPI)))
     return false;
 
   Value *ScalarVal =
diff --git a/llvm/test/Transforms/InstCombine/icmp-dom.ll b/llvm/test/Transforms/InstCombine/icmp-dom.ll
index f4b9022d14349..83cedd5ea9cb4 100644
--- a/llvm/test/Transforms/InstCombine/icmp-dom.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-dom.ll
@@ -403,3 +403,134 @@ truelabel:
 falselabel:
   ret i8 0
 }
+
+define i1 @and_mask1_eq(i32 %conv) {
+; CHECK-LABEL: @and_mask1_eq(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[CONV:%.*]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[AND]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    ret i1 false
+; CHECK:       else:
+; CHECK-NEXT:    ret i1 false
+;
+entry:
+  %and = and i32 %conv, 1
+  %cmp = icmp eq i32 %and, 0
+  br i1 %cmp, label %then, label %else
+
+then:
+  ret i1 0
+
+else:
+  %and1 = and i32 %conv, 3
+  %cmp1 = icmp eq i32 %and1, 0
+  ret i1 %cmp1
+}
+
+define i1 @and_mask1_ne(i32 %conv) {
+; CHECK-LABEL: @and_mask1_ne(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[CONV:%.*]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[AND]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    ret i1 false
+; CHECK:       else:
+; CHECK-NEXT:    ret i1 true
+;
+entry:
+  %and = and i32 %conv, 1
+  %cmp = icmp eq i32 %and, 0
+  br i1 %cmp, label %then, label %else
+
+then:
+  ret i1 0
+
+else:
+  %and1 = and i32 %conv, 3
+  %cmp1 = icmp ne i32 %and1, 0
+  ret i1 %cmp1
+}
+
+define i1 @and_mask2(i32 %conv) {
+; CHECK-LABEL: @and_mask2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[CONV:%.*]], 4
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[AND]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    ret i1 false
+; CHECK:       else:
+; CHECK-NEXT:    [[AND1:%.*]] = and i32 [[CONV]], 3
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp eq i32 [[AND1]], 0
+; CHECK-NEXT:    ret i1 [[CMP1]]
+;
+entry:
+  %and = and i32 %conv, 4
+  %cmp = icmp eq i32 %and, 0
+  br i1 %cmp, label %then, label %else
+
+then:
+  ret i1 0
+
+else:
+  %and1 = and i32 %conv, 3
+  %cmp1 = icmp eq i32 %and1, 0
+  ret i1 %cmp1
+}
+
+; TODO: %cmp1 can be folded into false.
+
+define i1 @and_mask3(i32 %conv) {
+; CHECK-LABEL: @and_mask3(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[CONV:%.*]], 3
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[AND]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    ret i1 false
+; CHECK:       else:
+; CHECK-NEXT:    [[AND1:%.*]] = and i32 [[CONV]], 7
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp eq i32 [[AND1]], 0
+; CHECK-NEXT:    ret i1 [[CMP1]]
+;
+entry:
+  %and = and i32 %conv, 3
+  %cmp = icmp eq i32 %and, 0
+  br i1 %cmp, label %then, label %else
+
+then:
+  ret i1 0
+
+else:
+  %and1 = and i32 %conv, 7
+  %cmp1 = icmp eq i32 %and1, 0
+  ret i1 %cmp1
+}
+
+define i1 @and_mask4(i32 %conv) {
+; CHECK-LABEL: @and_mask4(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[CONV:%.*]], 4
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[AND]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[THEN:%.*]], label [[ELSE:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    ret i1 false
+; CHECK:       else:
+; CHECK-NEXT:    ret i1 false
+;
+entry:
+  %and = and i32 %conv, 4
+  %cmp = icmp eq i32 %and, 0
+  br i1 %cmp, label %then, label %else
+
+then:
+  ret i1 0
+
+else:
+  %and1 = and i32 %conv, 7
+  %cmp1 = icmp eq i32 %and1, 0
+  ret i1 %cmp1
+}
diff --git a/llvm/unittests/Analysis/ValueTrackingTest.cpp b/llvm/unittests/Analysis/ValueTrackingTest.cpp
index b4d2270d70703..6b1044876d366 100644
--- a/llvm/unittests/Analysis/ValueTrackingTest.cpp
+++ b/llvm/unittests/Analysis/ValueTrackingTest.cpp
@@ -2055,7 +2055,8 @@ TEST_F(ValueTrackingTest, isNonZeroRecurrence) {
   )");
   const DataLayout &DL = M->getDataLayout();
   AssumptionCache AC(*F);
-  EXPECT_TRUE(isKnownNonZero(A, DL, 0, &AC, CxtI));
+  EXPECT_TRUE(isKnownNonZero(A, /*Depth=*/0,
+                             SimplifyQuery(DL, /*DT=*/nullptr, &AC, CxtI)));
 }
 
 TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond) {
@@ -2078,8 +2079,10 @@ TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond) {
   AssumptionCache AC(*F);
   DominatorTree DT(*F);
   const DataLayout &DL = M->getDataLayout();
-  EXPECT_EQ(isKnownNonZero(A, DL, 0, &AC, CxtI, &DT), true);
-  EXPECT_EQ(isKnownNonZero(A, DL, 0, &AC, CxtI2, &DT), false);
+  const SimplifyQuery SQ(DL, &DT, &AC);
+  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI)), true);
+  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI2)),
+            false);
 }
 
 TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond2) {
@@ -2102,8 +2105,10 @@ TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond2) {
   AssumptionCache AC(*F);
   DominatorTree DT(*F);
   const DataLayout &DL = M->getDataLayout();
-  EXPECT_EQ(isKnownNonZero(A, DL, 0, &AC, CxtI, &DT), true);
-  EXPECT_EQ(isKnownNonZero(A, DL, 0, &AC, CxtI2, &DT), false);
+  const SimplifyQuery SQ(DL, &DT, &AC);
+  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI)), true);
+  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI2)),
+            false);
 }
 
 TEST_F(ValueTrackingTest, IsImpliedConditionAnd) {

From 6f1e23b47d428d792866993ed26f4173d479d43d Mon Sep 17 00:00:00 2001
From: Kunwar Grover <groverkss@gmail.com>
Date: Fri, 12 Apr 2024 17:56:46 +0200
Subject: [PATCH 266/886] [MLIR][Bufferization] Choose default memory space in
 tensor copy insertion (#88500)

Tensor copy insertion currently uses memory_space = 0 when creating a
tensor copy using alloc_tensor. This memory space should instead be the
default memory space provided in bufferization options.
---
 .../Dialect/Bufferization/IR/BufferizableOpInterface.cpp   | 7 ++++---
 .../Bufferization/Transforms/tensor-copy-insertion.mlir    | 6 +++---
 mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir           | 2 +-
 3 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp b/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
index 55c9299c58eff..c2b2b99fc0083 100644
--- a/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
+++ b/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
@@ -193,10 +193,11 @@ FailureOr<Value> bufferization::allocateTensorForShapedValue(
   FailureOr<BaseMemRefType> copyBufferType = getBufferType(tensor, options);
   if (failed(copyBufferType))
     return failure();
-  Attribute memorySpace = copyBufferType->getMemorySpace();
+  std::optional<Attribute> memorySpace = copyBufferType->getMemorySpace();
   if (!memorySpace)
-    memorySpace = b.getI64IntegerAttr(0);
-  allocTensorOp.setMemorySpaceAttr(memorySpace);
+    memorySpace = options.defaultMemorySpaceFn(tensorType);
+  if (memorySpace.has_value())
+    allocTensorOp.setMemorySpaceAttr(memorySpace.value());
   return allocTensorOp.getResult();
 }
 
diff --git a/mlir/test/Dialect/Bufferization/Transforms/tensor-copy-insertion.mlir b/mlir/test/Dialect/Bufferization/Transforms/tensor-copy-insertion.mlir
index 72cf08df5978c..55dcca193cbe4 100644
--- a/mlir/test/Dialect/Bufferization/Transforms/tensor-copy-insertion.mlir
+++ b/mlir/test/Dialect/Bufferization/Transforms/tensor-copy-insertion.mlir
@@ -34,7 +34,7 @@ func.func @do_not_copy_undefined_tensor(%f: f32, %idx: index)
 {
   // The second alloc_tensor should not have a copy operand.
   // CHECK: bufferization.alloc_tensor() : tensor<5xf32>
-  // CHECK: bufferization.alloc_tensor() {memory_space = 0 : i64} : tensor<5xf32>
+  // CHECK: bufferization.alloc_tensor() : tensor<5xf32>
   %0 = bufferization.alloc_tensor() : tensor<5xf32>
   %1 = tensor.insert %f into %0[%idx] : tensor<5xf32>
   return %0, %1 : tensor<5xf32>, tensor<5xf32>
@@ -46,7 +46,7 @@ func.func @do_not_copy_undefined_tensor(%f: f32, %idx: index)
 func.func @do_not_copy_when_overwritten(%t: tensor<5xf32>, %f: f32)
   -> (tensor<5xf32>, tensor<5xf32>)
 {
-  // CHECK: %[[alloc:.*]] = bufferization.alloc_tensor() {memory_space = 0 : i64} : tensor<5xf32>
+  // CHECK: %[[alloc:.*]] = bufferization.alloc_tensor() : tensor<5xf32>
   // CHECK: linalg.generic {{.*}} outs(%[[alloc]] : tensor<5xf32>)
   %r = linalg.generic {
     indexing_maps = [affine_map<(d0) -> (d0)>],
@@ -65,7 +65,7 @@ func.func @do_not_copy_when_result_not_read(%t: tensor<5xf32>, %f: f32)
   -> (tensor<3xf32>)
 {
   %0 = tensor.extract_slice %t[0][3][1] : tensor<5xf32> to tensor<3xf32>
-  // CHECK: %[[alloc:.*]] = bufferization.alloc_tensor() {memory_space = 0 : i64} : tensor<3xf32>
+  // CHECK: %[[alloc:.*]] = bufferization.alloc_tensor() : tensor<3xf32>
   // CHECK: linalg.generic {{.*}} outs(%[[alloc]] : tensor<3xf32>)
   %r = linalg.generic {
     indexing_maps = [affine_map<(d0) -> (d0)>],
diff --git a/mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir b/mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir
index 6fc2db481c6fd..fcd69bea426d6 100644
--- a/mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir
+++ b/mlir/test/Dialect/SparseTensor/sparse_sddmm.mlir
@@ -63,7 +63,7 @@ func.func @fold_yield_direct_zero() -> tensor<32xf64> {
 // CHECK-DAG:       %[[VAL_5:.*]] = arith.constant 1 : index
 // CHECK-DAG:       %[[VAL_6:.*]] = arith.constant dense<0.000000e+00> : tensor<8x8xf64>
 // CHECK-DAG:       %[[VAL_7:.*]] = bufferization.alloc_tensor() copy(%[[VAL_6]]) : tensor<8x8xf64>
-// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.alloc_tensor() copy(%[[VAL_6]]) {memory_space = 0 : i64} : tensor<8x8xf64>
+// CHECK-DAG:       %[[VAL_8:.*]] = bufferization.alloc_tensor() copy(%[[VAL_6]]) : tensor<8x8xf64>
 // CHECK-DAG:       %[[VAL_9:.*]] = bufferization.to_memref %[[VAL_1]] : memref<8x8xf64>
 // CHECK-DAG:       %[[VAL_10:.*]] = bufferization.to_memref %[[VAL_2]] : memref<8x8xf64>
 // CHECK-DAG:       %[[VAL_11:.*]] = sparse_tensor.positions %[[VAL_0]] {level = 0 : index} : tensor<8x8xf64, #sparse{{[0-9]*}}> to memref<?xindex>

From b614e5b0340f783ad355899248c52cb22a04b014 Mon Sep 17 00:00:00 2001
From: Dominik Steenken <dost@de.ibm.com>
Date: Fri, 12 Apr 2024 18:03:57 +0200
Subject: [PATCH 267/886] [SystemZ] Add missing (dis-)assembly tests. (#88498)

### Assembly Tests
This adds assembly tests for
```
lcdfr, lndfr, lpdfr
```

### Disassembly Tests
This also adds a number of disassembly tests for the following
menmonics:
```
cgit, cgrt, cit, clfit, clgit, clgrt, clgt, clrt, clt, crt, lcdfr, lndfr, lpdfr
```
For `cutfu` / `cu12`, it also adds a check for bad register choice.
---
 llvm/test/MC/Disassembler/SystemZ/insns.txt   | 96 +++++++++++++++++++
 .../MC/Disassembler/SystemZ/invalid-regs.txt  |  7 ++
 llvm/test/MC/SystemZ/insn-good.s              | 30 ++++++
 3 files changed, 133 insertions(+)

diff --git a/llvm/test/MC/Disassembler/SystemZ/insns.txt b/llvm/test/MC/Disassembler/SystemZ/insns.txt
index f2f942ed5d36e..23714dfc3a8e6 100644
--- a/llvm/test/MC/Disassembler/SystemZ/insns.txt
+++ b/llvm/test/MC/Disassembler/SystemZ/insns.txt
@@ -2956,6 +2956,12 @@
 # CHECK: cgib %r0, 0, 15, 0
 0xec 0x0f 0x00 0x00 0x00 0xfc
 
+# CHECK: cgit %r0, 0, 0
+0xec 0x00 0x00 0x00 0x00 0x70
+
+# CHECK: cgit %r0, 0, 15
+0xec 0x00 0x00 0x00 0xf0 0x70
+
 # CHECK: cgith %r0, 0
 0xec 0x00 0x00 0x00 0x20 0x70
 
@@ -3052,6 +3058,12 @@
 # CHECK: cgrb %r0, %r0, 15, 0
 0xec 0x00 0x00 0x00 0xf0 0xe4
 
+# CHECK: cgrt %r0, %r1, 0
+0xb9 0x60 0x00 0x01
+
+# CHECK: cgrt %r0, %r1, 15
+0xb9 0x60 0xf0 0x01
+
 # CHECK: cgrth %r0, %r1
 0xb9 0x60 0x20 0x01
 
@@ -3430,6 +3442,12 @@
 # CHECK: cih %r15, 0
 0xcc 0xfd 0x00 0x00 0x00 0x00
 
+# CHECK: cit %r0, 0, 0
+0xec 0x00 0x00 0x00 0x00 0x72
+
+# CHECK: cit %r0, 0, 15
+0xec 0x00 0x00 0x00 0xf0 0x72
+
 # CHECK: cith %r0, 0
 0xec 0x00 0x00 0x00 0x20 0x72
 
@@ -3862,6 +3880,12 @@
 # CHECK: clfi %r15, 0
 0xc2 0xff 0x00 0x00 0x00 0x00
 
+# CHECK: clfit %r0, 0, 0
+0xec 0x00 0x00 0x00 0x00 0x73
+
+# CHECK: clfit %r0, 0, 15
+0xec 0x00 0x00 0x00 0xf0 0x73
+
 # CHECK: clfith %r0, 0
 0xec 0x00 0x00 0x00 0x20 0x73
 
@@ -3910,6 +3934,12 @@
 # CHECK: clg %r15, 0
 0xe3 0xf0 0x00 0x00 0x00 0x21
 
+# CHECK: clgit %r0, 0, 0
+0xec 0x00 0x00 0x00 0x00 0x71
+
+# CHECK: clgit %r0, 0, 15
+0xec 0x00 0x00 0x00 0xf0 0x71
+
 # CHECK: clgith %r0, 0
 0xec 0x00 0x00 0x00 0x20 0x71
 
@@ -4447,6 +4477,12 @@
 # CHECK: clrb %r0, %r0, 15, 0
 0xec 0x00 0x00 0x00 0xf0 0xf7
 
+# CHECK: clgrt %r0, %r1, 0
+0xb9 0x61 0x00 0x01
+
+# CHECK: clgrt %r0, %r1, 15
+0xb9 0x61 0xf0 0x01
+
 # CHECK: clgrth %r0, %r1
 0xb9 0x61 0x20 0x01
 
@@ -4465,6 +4501,12 @@
 # CHECK: clgrtle %r0, %r1
 0xb9 0x61 0xc0 0x01
 
+# CHECK: clrt %r0, %r1, 0
+0xb9 0x73 0x00 0x01
+
+# CHECK: clrt %r0, %r1, 15
+0xb9 0x73 0xf0 0x01
+
 # CHECK: clrth %r0, %r1
 0xb9 0x73 0x20 0x01
 
@@ -4483,6 +4525,12 @@
 # CHECK: clrtle %r0, %r1
 0xb9 0x73 0xc0 0x01
 
+# CHECK: clt %r0, 0, -524288
+0xeb 0x00 0x00 0x00 0x80 0x23
+
+# CHECK: clt %r0, 15, 32768
+0xeb 0x0f 0x00 0x00 0x08 0x23
+
 # CHECK: clth %r0, -524288
 0xeb 0x02 0x00 0x00 0x80 0x23
 
@@ -4525,6 +4573,12 @@
 # CHECK: cltle %r1, 0
 0xeb 0x1c 0x00 0x00 0x00 0x23
 
+# CHECK: clgt %r0, 0, -524288
+0xeb 0x00 0x00 0x00 0x80 0x2b
+
+# CHECK: clgt %r0, 15, 32768
+0xeb 0x0f 0x00 0x00 0x08 0x2b
+
 # CHECK: clgth %r0, -524288
 0xeb 0x02 0x00 0x00 0x80 0x2b
 
@@ -4792,6 +4846,12 @@
 # CHECK: crdte %r4, %r5, %r6, 7
 0xb9 0x8f 0x57 0x46
 
+# CHECK: crt %r0, %r1, 0
+0xb9 0x72 0x00 0x01
+
+# CHECK: crt %r0, %r1, 15
+0xb9 0x72 0xf0 0x01
+
 # CHECK: crth %r0, %r1
 0xb9 0x72 0x20 0x01
 
@@ -7972,6 +8032,18 @@
 # CHECK: lcdbr %f15, %f9
 0xb3 0x13 0x00 0xf9
 
+# CHECK: lcdfr %f0, %f9
+0xb3 0x73 0x00 0x09
+
+# CHECK: lcdfr %f0, %f15
+0xb3 0x73 0x00 0x0f
+
+# CHECK: lcdfr %f15, %f0
+0xb3 0x73 0x00 0xf0
+
+# CHECK: lcdfr %f15, %f9
+0xb3 0x73 0x00 0xf9
+
 # CHECK: lcdr %f0, %f9
 0x23 0x09
 
@@ -9570,6 +9642,18 @@
 # CHECK: lndbr %f15, %f9
 0xb3 0x11 0x00 0xf9
 
+# CHECK: lndfr %f0, %f9
+0xb3 0x71 0x00 0x09
+
+# CHECK: lndfr %f0, %f15
+0xb3 0x71 0x00 0x0f
+
+# CHECK: lndfr %f15, %f0
+0xb3 0x71 0x00 0xf0
+
+# CHECK: lndfr %f15, %f9
+0xb3 0x71 0x00 0xf9
+
 # CHECK: lndr %f0, %f9
 0x21 0x09
 
@@ -9909,6 +9993,18 @@
 # CHECK: lpdbr %f15, %f9
 0xb3 0x10 0x00 0xf9
 
+# CHECK: lpdfr %f0, %f9
+0xb3 0x70 0x00 0x09
+
+# CHECK: lpdfr %f0, %f15
+0xb3 0x70 0x00 0x0f
+
+# CHECK: lpdfr %f15, %f0
+0xb3 0x70 0x00 0xf0
+
+# CHECK: lpdfr %f15, %f9
+0xb3 0x70 0x00 0xf9
+
 # CHECK: lpdg %r0, 0, 0
 0xc8 0x05 0x00 0x00 0x00 0x00
 
diff --git a/llvm/test/MC/Disassembler/SystemZ/invalid-regs.txt b/llvm/test/MC/Disassembler/SystemZ/invalid-regs.txt
index 12440677ba34a..0ce960d5ddcc3 100644
--- a/llvm/test/MC/Disassembler/SystemZ/invalid-regs.txt
+++ b/llvm/test/MC/Disassembler/SystemZ/invalid-regs.txt
@@ -20,3 +20,10 @@
 # CHECK-NEXT: 0xb9 0x97 0x00 0x18
 # CHECK-NEXT: ^
 0xb9 0x97 0x00 0x18
+
+# This would be "cutfu %r1, %r2", but %r1 is invalid.
+#
+# CHECK-NEXT: warning: invalid instruction encoding
+# CHECK-NEXT: 0xb2 0xa7 0x00 0x12
+# CHECK-NEXT: ^
+0xb2 0xa7 0x00 0x12
diff --git a/llvm/test/MC/SystemZ/insn-good.s b/llvm/test/MC/SystemZ/insn-good.s
index 91120f7f97aa9..f5dd672f9dd8c 100644
--- a/llvm/test/MC/SystemZ/insn-good.s
+++ b/llvm/test/MC/SystemZ/insn-good.s
@@ -9038,6 +9038,36 @@
 	lcdbr	%f15,%f0
 	lcdbr	%f15,%f9
 
+#CHECK: lcdfr	%f0, %f9                # encoding: [0xb3,0x73,0x00,0x09]
+#CHECK: lcdfr	%f0, %f15               # encoding: [0xb3,0x73,0x00,0x0f]
+#CHECK: lcdfr	%f15, %f0               # encoding: [0xb3,0x73,0x00,0xf0]
+#CHECK: lcdfr	%f15, %f9               # encoding: [0xb3,0x73,0x00,0xf9]
+
+	lcdfr	%f0,%f9
+	lcdfr	%f0,%f15
+	lcdfr	%f15,%f0
+	lcdfr	%f15,%f9
+
+#CHECK: lndfr	%f0, %f9                # encoding: [0xb3,0x71,0x00,0x09]
+#CHECK: lndfr	%f0, %f15               # encoding: [0xb3,0x71,0x00,0x0f]
+#CHECK: lndfr	%f15, %f0               # encoding: [0xb3,0x71,0x00,0xf0]
+#CHECK: lndfr	%f15, %f9               # encoding: [0xb3,0x71,0x00,0xf9]
+
+	lndfr	%f0,%f9
+	lndfr	%f0,%f15
+	lndfr	%f15,%f0
+	lndfr	%f15,%f9
+
+#CHECK: lpdfr	%f0, %f9                # encoding: [0xb3,0x70,0x00,0x09]
+#CHECK: lpdfr	%f0, %f15               # encoding: [0xb3,0x70,0x00,0x0f]
+#CHECK: lpdfr	%f15, %f0               # encoding: [0xb3,0x70,0x00,0xf0]
+#CHECK: lpdfr	%f15, %f9               # encoding: [0xb3,0x70,0x00,0xf9]
+
+	lpdfr	%f0,%f9
+	lpdfr	%f0,%f15
+	lpdfr	%f15,%f0
+	lpdfr	%f15,%f9
+
 #CHECK: lcdr	%f0, %f9                # encoding: [0x23,0x09]
 #CHECK: lcdr	%f0, %f15               # encoding: [0x23,0x0f]
 #CHECK: lcdr	%f15, %f0               # encoding: [0x23,0xf0]

From b794dc23255505dd7735f995b8ff1192305a072e Mon Sep 17 00:00:00 2001
From: Dominik Steenken <dost@de.ibm.com>
Date: Fri, 12 Apr 2024 18:05:30 +0200
Subject: [PATCH 268/886] [SystemZ] Add custom handling of legal vectors with
 reduce-add. (#88495)

This commit skips the expansion of the `vector.reduce.add` intrinsic on
vector-enabled SystemZ targets in order to introduce custom handling of
`vector.reduce.add` for legal vector types using the VSUM instructions.
This is limited to full vectors with scalar types up to `i32` due to
performance concerns.

It also adds testing for the generation of such custom handling, and
adapts the related cost computation, as well as the testing for that.

The expected result is a performance boost in certain benchmarks that
make heavy use of `vector.reduce.add` with other benchmarks remaining
constant.

For instance, the assembly for `vector.reduce.add<4 x i32>` changes from
```hlasm
        vmrlg   %v0, %v24, %v24
        vaf     %v0, %v24, %v0
        vrepf   %v1, %v0, 1
        vaf     %v0, %v0, %v1
        vlgvf   %r2, %v0, 0
```
to
```hlasm
        vgbm    %v0, 0
        vsumqf  %v0, %v24, %v0
        vlgvf   %r2, %v0, 3
```
---
 .../Target/SystemZ/SystemZISelLowering.cpp    |  43 +++
 llvm/lib/Target/SystemZ/SystemZISelLowering.h |   1 +
 .../SystemZ/SystemZTargetTransformInfo.cpp    |  31 +-
 .../SystemZ/SystemZTargetTransformInfo.h      |   2 +
 .../Analysis/CostModel/SystemZ/reduce-add.ll  |  24 +-
 .../test/CodeGen/SystemZ/vec-reduce-add-01.ll | 289 ++++++++++++++++++
 6 files changed, 373 insertions(+), 17 deletions(-)
 create mode 100644 llvm/test/CodeGen/SystemZ/vec-reduce-add-01.ll

diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
index efffd669d268b..3b3057f8fe964 100644
--- a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
@@ -16,6 +16,7 @@
 #include "SystemZMachineFunctionInfo.h"
 #include "SystemZTargetMachine.h"
 #include "llvm/CodeGen/CallingConvLower.h"
+#include "llvm/CodeGen/ISDOpcodes.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
@@ -23,6 +24,7 @@
 #include "llvm/IR/Intrinsics.h"
 #include "llvm/IR/IntrinsicsS390.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/KnownBits.h"
 #include <cctype>
 #include <optional>
@@ -451,6 +453,10 @@ SystemZTargetLowering::SystemZTargetLowering(const TargetMachine &TM,
       setOperationAction(ISD::SRL, VT, Custom);
       setOperationAction(ISD::ROTL, VT, Custom);
 
+      // Add ISD::VECREDUCE_ADD as custom in order to implement
+      // it with VZERO+VSUM
+      setOperationAction(ISD::VECREDUCE_ADD, VT, Custom);
+
       // Map SETCCs onto one of VCE, VCH or VCHL, swapping the operands
       // and inverting the result as necessary.
       setOperationAction(ISD::SETCC, VT, Custom);
@@ -6167,6 +6173,8 @@ SDValue SystemZTargetLowering::LowerOperation(SDValue Op,
     return lowerOR(Op, DAG);
   case ISD::CTPOP:
     return lowerCTPOP(Op, DAG);
+  case ISD::VECREDUCE_ADD:
+    return lowerVECREDUCE_ADD(Op, DAG);
   case ISD::ATOMIC_FENCE:
     return lowerATOMIC_FENCE(Op, DAG);
   case ISD::ATOMIC_SWAP:
@@ -9600,3 +9608,38 @@ SDValue SystemZTargetLowering::lowerGET_ROUNDING(SDValue Op,
 
   return DAG.getMergeValues({RetVal, Chain}, dl);
 }
+
+SDValue SystemZTargetLowering::lowerVECREDUCE_ADD(SDValue Op,
+                                                  SelectionDAG &DAG) const {
+  EVT VT = Op.getValueType();
+  Op = Op.getOperand(0);
+  EVT OpVT = Op.getValueType();
+
+  assert(OpVT.isVector() && "Operand type for VECREDUCE_ADD is not a vector.");
+
+  SDLoc DL(Op);
+
+  // load a 0 vector for the third operand of VSUM.
+  SDValue Zero = DAG.getSplatBuildVector(OpVT, DL, DAG.getConstant(0, DL, VT));
+
+  // execute VSUM.
+  switch (OpVT.getScalarSizeInBits()) {
+  case 8:
+  case 16:
+    Op = DAG.getNode(SystemZISD::VSUM, DL, MVT::v4i32, Op, Zero);
+    LLVM_FALLTHROUGH;
+  case 32:
+  case 64:
+    Op = DAG.getNode(SystemZISD::VSUM, DL, MVT::i128, Op,
+                     DAG.getBitcast(Op.getValueType(), Zero));
+    break;
+  case 128:
+    break; // VSUM over v1i128 should not happen and would be a noop
+  default:
+    llvm_unreachable("Unexpected scalar size.");
+  }
+  // Cast to original vector type, retrieve last element.
+  return DAG.getNode(
+      ISD::EXTRACT_VECTOR_ELT, DL, VT, DAG.getBitcast(OpVT, Op),
+      DAG.getConstant(OpVT.getVectorNumElements() - 1, DL, MVT::i32));
+}
diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.h b/llvm/lib/Target/SystemZ/SystemZISelLowering.h
index 7140287a886cc..2290a7d62e89f 100644
--- a/llvm/lib/Target/SystemZ/SystemZISelLowering.h
+++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.h
@@ -696,6 +696,7 @@ class SystemZTargetLowering : public TargetLowering {
   SDValue lowerBITCAST(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerOR(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerCTPOP(SDValue Op, SelectionDAG &DAG) const;
+  SDValue lowerVECREDUCE_ADD(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerATOMIC_FENCE(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerATOMIC_LDST_I128(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerATOMIC_LOAD_OP(SDValue Op, SelectionDAG &DAG,
diff --git a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
index 17e534f405c08..4c9e78c05dbca 100644
--- a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp
@@ -19,6 +19,7 @@
 #include "llvm/CodeGen/CostTable.h"
 #include "llvm/CodeGen/TargetLowering.h"
 #include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Intrinsics.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/MathExtras.h"
 
@@ -1293,18 +1294,14 @@ getVectorIntrinsicInstrCost(Intrinsic::ID ID, Type *RetTy,
   if (ID == Intrinsic::vector_reduce_add) {
     // Retrieve number and size of elements for the vector op.
     auto *VTy = cast<FixedVectorType>(ParamTys.front());
-    unsigned NumElements = VTy->getNumElements();
     unsigned ScalarSize = VTy->getScalarSizeInBits();
     // For scalar sizes >128 bits, we fall back to the generic cost estimate.
     if (ScalarSize > SystemZ::VectorBits)
       return -1;
-    // A single vector register can hold this many elements.
-    unsigned MaxElemsPerVector = SystemZ::VectorBits / ScalarSize;
     // This many vector regs are needed to represent the input elements (V).
     unsigned VectorRegsNeeded = getNumVectorRegs(VTy);
     // This many instructions are needed for the final sum of vector elems (S).
-    unsigned LastVectorHandling =
-        2 * Log2_32_Ceil(std::min(NumElements, MaxElemsPerVector));
+    unsigned LastVectorHandling = (ScalarSize < 32) ? 3 : 2;
     // We use vector adds to create a sum vector, which takes
     // V/2 + V/4 + ... = V - 1 operations.
     // Then, we need S operations to sum up the elements of that sum vector,
@@ -1324,3 +1321,27 @@ SystemZTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     return Cost;
   return BaseT::getIntrinsicInstrCost(ICA, CostKind);
 }
+
+bool SystemZTTIImpl::shouldExpandReduction(const IntrinsicInst *II) const {
+  // Always expand on Subtargets without vector instructions
+  if (!ST->hasVector())
+    return true;
+
+  // Always expand for operands that do not fill one vector reg
+  auto *Type = cast<FixedVectorType>(II->getOperand(0)->getType());
+  unsigned NumElts = Type->getNumElements();
+  unsigned ScalarSize = Type->getScalarSizeInBits();
+  unsigned MaxElts = SystemZ::VectorBits / ScalarSize;
+  if (NumElts < MaxElts)
+    return true;
+
+  // Otherwise
+  switch (II->getIntrinsicID()) {
+  // Do not expand vector.reduce.add
+  case Intrinsic::vector_reduce_add:
+    // Except for i64, since the performance benefit is dubious there
+    return ScalarSize >= 64;
+  default:
+    return true;
+  }
+}
diff --git a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
index 1d824d353d8fb..696d887c1d5db 100644
--- a/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
+++ b/llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
@@ -127,6 +127,8 @@ class SystemZTTIImpl : public BasicTTIImplBase<SystemZTTIImpl> {
 
   InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
                                         TTI::TargetCostKind CostKind);
+
+  bool shouldExpandReduction(const IntrinsicInst *II) const;
   /// @}
 };
 
diff --git a/llvm/test/Analysis/CostModel/SystemZ/reduce-add.ll b/llvm/test/Analysis/CostModel/SystemZ/reduce-add.ll
index 061e5ece44a4e..90b5b746c914a 100644
--- a/llvm/test/Analysis/CostModel/SystemZ/reduce-add.ll
+++ b/llvm/test/Analysis/CostModel/SystemZ/reduce-add.ll
@@ -7,19 +7,19 @@ define void @reduce(ptr %src, ptr %dst) {
 ; CHECK:  Cost Model: Found an estimated cost of 5 for instruction: %R8_64 = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %V8_64)
 ; CHECK:  Cost Model: Found an estimated cost of 9 for instruction: %R16_64 = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %V16_64)
 ; CHECK:  Cost Model: Found an estimated cost of 2 for instruction: %R2_32 = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %V2_32)
-; CHECK:  Cost Model: Found an estimated cost of 4 for instruction: %R4_32 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %V4_32)
-; CHECK:  Cost Model: Found an estimated cost of 5 for instruction: %R8_32 = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %V8_32)
-; CHECK:  Cost Model: Found an estimated cost of 7 for instruction: %R16_32 = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %V16_32)
-; CHECK:  Cost Model: Found an estimated cost of 2 for instruction: %R2_16 = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> %V2_16)
-; CHECK:  Cost Model: Found an estimated cost of 4 for instruction: %R4_16 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %V4_16)
-; CHECK:  Cost Model: Found an estimated cost of 6 for instruction: %R8_16 = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %V8_16)
-; CHECK:  Cost Model: Found an estimated cost of 7 for instruction: %R16_16 = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %V16_16)
-; CHECK:  Cost Model: Found an estimated cost of 2 for instruction: %R2_8 = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %V2_8)
-; CHECK:  Cost Model: Found an estimated cost of 4 for instruction: %R4_8 = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %V4_8)
-; CHECK:  Cost Model: Found an estimated cost of 6 for instruction: %R8_8 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %V8_8)
-; CHECK:  Cost Model: Found an estimated cost of 8 for instruction: %R16_8 = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %V16_8)
+; CHECK:  Cost Model: Found an estimated cost of 2 for instruction: %R4_32 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %V4_32)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R8_32 = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %V8_32)
+; CHECK:  Cost Model: Found an estimated cost of 5 for instruction: %R16_32 = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %V16_32)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R2_16 = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> %V2_16)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R4_16 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %V4_16)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R8_16 = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %V8_16)
+; CHECK:  Cost Model: Found an estimated cost of 4 for instruction: %R16_16 = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %V16_16)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R2_8 = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %V2_8)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R4_8 = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %V4_8)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R8_8 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %V8_8)
+; CHECK:  Cost Model: Found an estimated cost of 3 for instruction: %R16_8 = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %V16_8)
 ;
-; CHECK:  Cost Model: Found an estimated cost of 15 for instruction: %R128_8 = call i8 @llvm.vector.reduce.add.v128i8(<128 x i8> %V128_8)
+; CHECK:  Cost Model: Found an estimated cost of 10 for instruction: %R128_8 = call i8 @llvm.vector.reduce.add.v128i8(<128 x i8> %V128_8)
 ; CHECK:  Cost Model: Found an estimated cost of 20 for instruction: %R4_256 = call i256 @llvm.vector.reduce.add.v4i256(<4 x i256> %V4_256)
 
   ; REDUCEADD64
diff --git a/llvm/test/CodeGen/SystemZ/vec-reduce-add-01.ll b/llvm/test/CodeGen/SystemZ/vec-reduce-add-01.ll
new file mode 100644
index 0000000000000..56b151d7f9412
--- /dev/null
+++ b/llvm/test/CodeGen/SystemZ/vec-reduce-add-01.ll
@@ -0,0 +1,289 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; Test vector add reduction instrinsic
+;
+; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z16 | FileCheck %s
+
+; 1 vector length
+declare i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %a)
+declare i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %a)
+declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a)
+declare i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a)
+declare i128 @llvm.vector.reduce.add.v1i128(<1 x i128> %a)
+; 2 vector lengths
+declare i8 @llvm.vector.reduce.add.v32i8(<32 x i8> %a)
+declare i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %a)
+declare i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %a)
+declare i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %a)
+declare i128 @llvm.vector.reduce.add.v2i128(<2 x i128> %a)
+; ; TODO
+; ; 4 vector lengths
+declare i8 @llvm.vector.reduce.add.v64i8(<64 x i8> %a)
+declare i16 @llvm.vector.reduce.add.v32i16(<32 x i16> %a)
+declare i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %a)
+declare i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %a)
+declare i128 @llvm.vector.reduce.add.v4i128(<4 x i128> %a)
+; ; Subvector lengths
+declare i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a)
+declare i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a)
+declare i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %a)
+declare i64 @llvm.vector.reduce.add.v1i64(<1 x i64> %a)
+
+define i8 @f1_1(<16 x i8> %a) {
+; CHECK-LABEL: f1_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vgbm %v0, 0
+; CHECK-NEXT:    vsumb %v1, %v24, %v0
+; CHECK-NEXT:    vsumqf %v0, %v1, %v0
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %a)
+  ret i8 %redadd
+}
+
+define i16 @f1_2(<8 x i16> %a) {
+; CHECK-LABEL: f1_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vgbm %v0, 0
+; CHECK-NEXT:    vsumh %v1, %v24, %v0
+; CHECK-NEXT:    vsumqf %v0, %v1, %v0
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %a)
+  ret i16 %redadd
+}
+
+define i32 @f1_3(<4 x i32> %a) {
+; CHECK-LABEL: f1_3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vgbm %v0, 0
+; CHECK-NEXT:    vsumqf %v0, %v24, %v0
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a)
+  ret i32 %redadd
+}
+
+define i64 @f1_4(<2 x i64> %a) {
+; CHECK-LABEL: f1_4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vrepg %v0, %v24, 1
+; CHECK-NEXT:    vag %v0, %v24, %v0
+; CHECK-NEXT:    vlgvg %r2, %v0, 0
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a)
+  ret i64 %redadd
+}
+
+define i128 @f1_5(<1 x i128> %a) {
+; CHECK-LABEL: f1_5:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vst %v24, 0(%r2), 3
+; CHECK-NEXT:    br %r14
+  %redadd = call i128 @llvm.vector.reduce.add.v1i128(<1 x i128> %a)
+  ret i128 %redadd
+}
+
+define i8 @f2_1(<32 x i8> %a) {
+; CHECK-LABEL: f2_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vab %v0, %v24, %v26
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumb %v0, %v0, %v1
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i8 @llvm.vector.reduce.add.v32i8(<32 x i8> %a)
+  ret i8 %redadd
+}
+
+define i16 @f2_2(<16 x i16> %a) {
+; CHECK-LABEL: f2_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vah %v0, %v24, %v26
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumh %v0, %v0, %v1
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %a)
+  ret i16 %redadd
+}
+
+define i32 @f2_3(<8 x i32> %a) {
+; CHECK-LABEL: f2_3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vaf %v0, %v24, %v26
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %a)
+  ret i32 %redadd
+}
+
+define i64 @f2_4(<4 x i64> %a) {
+; CHECK-LABEL: f2_4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vag %v0, %v24, %v26
+; CHECK-NEXT:    vrepg %v1, %v0, 1
+; CHECK-NEXT:    vag %v0, %v0, %v1
+; CHECK-NEXT:    vlgvg %r2, %v0, 0
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %a)
+  ret i64 %redadd
+}
+
+define i128 @f2_5(<2 x i128> %a) {
+; CHECK-LABEL: f2_5:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl %v0, 16(%r3), 3
+; CHECK-NEXT:    vl %v1, 0(%r3), 3
+; CHECK-NEXT:    vaq %v0, %v1, %v0
+; CHECK-NEXT:    vst %v0, 0(%r2), 3
+; CHECK-NEXT:    br %r14
+  %redadd = call i128 @llvm.vector.reduce.add.v2i128(<2 x i128> %a)
+  ret i128 %redadd
+}
+
+define i8 @f3_1(<64 x i8> %a) {
+; CHECK-LABEL: f3_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vab %v0, %v26, %v30
+; CHECK-NEXT:    vab %v1, %v24, %v28
+; CHECK-NEXT:    vab %v0, %v1, %v0
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumb %v0, %v0, %v1
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i8 @llvm.vector.reduce.add.v64i8(<64 x i8> %a)
+  ret i8 %redadd
+}
+
+define i16 @f3_2(<32 x i16> %a) {
+; CHECK-LABEL: f3_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vah %v0, %v26, %v30
+; CHECK-NEXT:    vah %v1, %v24, %v28
+; CHECK-NEXT:    vah %v0, %v1, %v0
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumh %v0, %v0, %v1
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i16 @llvm.vector.reduce.add.v32i16(<32 x i16> %a)
+  ret i16 %redadd
+}
+
+define i32 @f3_3(<16 x i32> %a) {
+; CHECK-LABEL: f3_3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vaf %v0, %v26, %v30
+; CHECK-NEXT:    vaf %v1, %v24, %v28
+; CHECK-NEXT:    vaf %v0, %v1, %v0
+; CHECK-NEXT:    vgbm %v1, 0
+; CHECK-NEXT:    vsumqf %v0, %v0, %v1
+; CHECK-NEXT:    vlgvf %r2, %v0, 3
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %a)
+  ret i32 %redadd
+}
+
+define i64 @f3_4(<8 x i64> %a) {
+; CHECK-LABEL: f3_4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vag %v0, %v26, %v30
+; CHECK-NEXT:    vag %v1, %v24, %v28
+; CHECK-NEXT:    vag %v0, %v1, %v0
+; CHECK-NEXT:    vrepg %v1, %v0, 1
+; CHECK-NEXT:    vag %v0, %v0, %v1
+; CHECK-NEXT:    vlgvg %r2, %v0, 0
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %a)
+  ret i64 %redadd
+}
+
+define i128 @f3_5(<4 x i128> %a) {
+; CHECK-LABEL: f3_5:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl %v0, 32(%r3), 3
+; CHECK-NEXT:    vl %v1, 0(%r3), 3
+; CHECK-NEXT:    vl %v2, 48(%r3), 3
+; CHECK-NEXT:    vl %v3, 16(%r3), 3
+; CHECK-NEXT:    vaq %v2, %v3, %v2
+; CHECK-NEXT:    vaq %v0, %v1, %v0
+; CHECK-NEXT:    vaq %v0, %v0, %v2
+; CHECK-NEXT:    vst %v0, 0(%r2), 3
+; CHECK-NEXT:    br %r14
+  %redadd = call i128 @llvm.vector.reduce.add.v4i128(<4 x i128> %a)
+  ret i128 %redadd
+}
+
+
+define i8 @f4_1(<8 x i8> %a) {
+; CHECK-LABEL: f4_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpkg %v0, %v24, %v24
+; CHECK-NEXT:    vab %v0, %v24, %v0
+; CHECK-NEXT:    vpkf %v1, %v0, %v0
+; CHECK-NEXT:    vab %v0, %v0, %v1
+; CHECK-NEXT:    vrepb %v1, %v0, 1
+; CHECK-NEXT:    vab %v0, %v0, %v1
+; CHECK-NEXT:    vlgvb %r2, %v0, 0
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a)
+  ret i8 %redadd
+}
+
+define i16 @f4_2(<4 x i16> %a) {
+; CHECK-LABEL: f4_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpkg %v0, %v24, %v24
+; CHECK-NEXT:    vah %v0, %v24, %v0
+; CHECK-NEXT:    vreph %v1, %v0, 1
+; CHECK-NEXT:    vah %v0, %v0, %v1
+; CHECK-NEXT:    vlgvh %r2, %v0, 0
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+  %redadd = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a)
+  ret i16 %redadd
+}
+
+define i32 @f4_3(<2 x i32> %a) {
+; CHECK-LABEL: f4_3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vrepf %v0, %v24, 1
+; CHECK-NEXT:    vaf %v0, %v24, %v0
+; CHECK-NEXT:    vlgvf %r2, %v0, 0
+; CHECK-NEXT:    # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %a)
+  ret i32 %redadd
+}
+
+define i64 @f4_4(<1 x i64> %a) {
+; CHECK-LABEL: f4_4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vlgvg %r2, %v24, 0
+; CHECK-NEXT:    br %r14
+
+  %redadd = call i64 @llvm.vector.reduce.add.v1i64(<1 x i64> %a)
+  ret i64 %redadd
+}

From 8d468c132eed7ffe34d601b224220efd51655eb3 Mon Sep 17 00:00:00 2001
From: Evan Wilde <etceterawilde@gmail.com>
Date: Fri, 12 Apr 2024 09:05:59 -0700
Subject: [PATCH 269/886] NFC: Make clang resource headers an interface library
 (#88317)

Making the clang resource headers into an interface library instead of a
custom target means that we can attach the header search paths to the
library. Targets that "link" against this library will automatically
have the appropriate paths added to their header search paths to find
them. Then downstream projects that embed a copy of clang can query the
generated `ClangTargets.cmake` file for the header location instead of
attempting to compute them.
---
 clang/lib/Headers/CMakeLists.txt | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/clang/lib/Headers/CMakeLists.txt b/clang/lib/Headers/CMakeLists.txt
index 97104ccd8db59..e6ae4e19e81db 100644
--- a/clang/lib/Headers/CMakeLists.txt
+++ b/clang/lib/Headers/CMakeLists.txt
@@ -437,14 +437,14 @@ foreach( f ${generated_files} )
 endforeach( f )
 
 function(add_header_target target_name file_list)
-  add_custom_target(${target_name} DEPENDS ${file_list})
+  add_library(${target_name} INTERFACE ${file_list})
   set_target_properties(${target_name} PROPERTIES
     FOLDER "Misc"
     RUNTIME_OUTPUT_DIRECTORY "${output_dir}")
 endfunction()
 
 # The catch-all clang-resource-headers target
-add_custom_target("clang-resource-headers" ALL DEPENDS ${out_files})
+add_library(clang-resource-headers INTERFACE ${out_files})
 set_target_properties("clang-resource-headers" PROPERTIES
   FOLDER "Misc"
   RUNTIME_OUTPUT_DIRECTORY "${output_dir}")
@@ -501,6 +501,10 @@ add_header_target("windows-resource-headers" ${windows_only_files})
 add_header_target("utility-resource-headers" ${utility_files})
 
 get_clang_resource_dir(header_install_dir SUBDIR include)
+target_include_directories(clang-resource-headers INTERFACE
+  $<BUILD_INTERFACE:${output_dir}>
+  $<INSTALL_INTERFACE:${header_install_dir}>)
+set_property(GLOBAL APPEND PROPERTY CLANG_EXPORTS clang-resource-headers)
 
 #############################################################
 # Install rules for the catch-all clang-resource-headers target

From 05d8b5e62d7a1cf2f94582346d6fd2d39667a26b Mon Sep 17 00:00:00 2001
From: JOSTAR <52376093+shenjunjiekoda@users.noreply.github.com>
Date: Sat, 13 Apr 2024 00:15:33 +0800
Subject: [PATCH 270/886] [analyzer] Support `PointerType` in
 `getCXXRecordDecl` for `ContainerModeling` (#87787)

---
 .../lib/StaticAnalyzer/Checkers/ContainerModeling.cpp  |  4 ++++
 clang/test/Analysis/invalidated-iterator.cpp           | 10 +++++++++-
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
index 65a2ec4076fdf..009c0d3fb9368 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
@@ -770,6 +770,10 @@ const CXXRecordDecl *getCXXRecordDecl(ProgramStateRef State,
     Type = RefT->getPointeeType();
   }
 
+  if (const auto *PtrT = Type->getAs<PointerType>()) {
+    Type = PtrT->getPointeeType();
+  }
+
   return Type->getUnqualifiedDesugaredType()->getAsCXXRecordDecl();
 }
 
diff --git a/clang/test/Analysis/invalidated-iterator.cpp b/clang/test/Analysis/invalidated-iterator.cpp
index 778a8e01d9938..c940dbf7276d3 100644
--- a/clang/test/Analysis/invalidated-iterator.cpp
+++ b/clang/test/Analysis/invalidated-iterator.cpp
@@ -130,6 +130,14 @@ struct cont_with_ptr_iterator {
   T* erase(T*);
 };
 
+void invalidated_access_via_end_iterator_after_push_back() {
+  cont_with_ptr_iterator<int> C;
+  C.push_back(1);
+  auto i = C.end();
+  C.push_back(2);
+  auto j = i[-1]; // expected-warning{{Invalidated iterator accessed}}
+}
+
 void invalidated_dereference_end_ptr_iterator(cont_with_ptr_iterator<int> &C) {
   auto i = C.begin();
   C.erase(i);
@@ -196,4 +204,4 @@ void invalidated_subscript_end_ptr_iterator(cont_with_ptr_iterator<int> &C) {
   auto i = C.begin();
   C.erase(i);
   (void) i[1]; // expected-warning{{Invalidated iterator accessed}}
-}
+}
\ No newline at end of file

From 37575f5262d0e37a1e0b10c59fb8f59182cfe951 Mon Sep 17 00:00:00 2001
From: Jordan Rupprecht <rupprecht@google.com>
Date: Fri, 12 Apr 2024 16:27:53 +0000
Subject: [PATCH 271/886] [NFC][ValueTracking] Fix Wunused-variable

For e0a628715a8464e220c8ba9e9aaaf2561139198a
---
 llvm/lib/Analysis/InstructionSimplify.cpp | 2 --
 1 file changed, 2 deletions(-)

diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index e83657df20f2a..4e6e666922671 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -2713,8 +2713,6 @@ static Constant *computePointerICmp(CmpInst::Predicate Pred, Value *LHS,
   assert(LHS->getType() == RHS->getType() && "Must have same types");
   const DataLayout &DL = Q.DL;
   const TargetLibraryInfo *TLI = Q.TLI;
-  const DominatorTree *DT = Q.DT;
-  const Instruction *CxtI = Q.CxtI;
 
   // We can only fold certain predicates on pointer comparisons.
   switch (Pred) {

From 54a6798e0a3630e705ed32dbbd63414a16331085 Mon Sep 17 00:00:00 2001
From: Edwin Vane <revane@google.com>
Date: Fri, 12 Apr 2024 12:49:25 -0400
Subject: [PATCH 272/886] [clang-tidy] Simplify RenamerClangTidyCheck API
 (#88268)

Some functions allow a null SourceManager, no SourceManager, or a
SourceManager in an inconsistent argument position. Since SourceManager
is generally not null and it doesn't make sense to apply renaming
without one, these inconsistencies are now gone.
---
 .../utils/RenamerClangTidyCheck.cpp           | 46 +++++++++++--------
 .../clang-tidy/utils/RenamerClangTidyCheck.h  | 11 +++--
 2 files changed, 32 insertions(+), 25 deletions(-)

diff --git a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
index ad8048e2a92b7..962a243ce94d4 100644
--- a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
+++ b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.cpp
@@ -169,14 +169,14 @@ class RenamerClangTidyCheckPPCallbacks : public PPCallbacks {
       return;
     if (SM.isWrittenInCommandLineFile(MacroNameTok.getLocation()))
       return;
-    Check->checkMacro(SM, MacroNameTok, Info);
+    Check->checkMacro(MacroNameTok, Info, SM);
   }
 
   /// MacroExpands calls expandMacro for macros in the main file
   void MacroExpands(const Token &MacroNameTok, const MacroDefinition &MD,
                     SourceRange /*Range*/,
                     const MacroArgs * /*Args*/) override {
-    Check->expandMacro(MacroNameTok, MD.getMacroInfo());
+    Check->expandMacro(MacroNameTok, MD.getMacroInfo(), SM);
   }
 
 private:
@@ -187,7 +187,7 @@ class RenamerClangTidyCheckPPCallbacks : public PPCallbacks {
 class RenamerClangTidyVisitor
     : public RecursiveASTVisitor<RenamerClangTidyVisitor> {
 public:
-  RenamerClangTidyVisitor(RenamerClangTidyCheck *Check, const SourceManager *SM,
+  RenamerClangTidyVisitor(RenamerClangTidyCheck *Check, const SourceManager &SM,
                           bool AggressiveDependentMemberLookup)
       : Check(Check), SM(SM),
         AggressiveDependentMemberLookup(AggressiveDependentMemberLookup) {}
@@ -258,7 +258,7 @@ class RenamerClangTidyVisitor
     // Fix overridden methods
     if (const auto *Method = dyn_cast<CXXMethodDecl>(Decl)) {
       if (const CXXMethodDecl *Overridden = getOverrideMethod(Method)) {
-        Check->addUsage(Overridden, Method->getLocation());
+        Check->addUsage(Overridden, Method->getLocation(), SM);
         return true; // Don't try to add the actual decl as a Failure.
       }
     }
@@ -268,7 +268,7 @@ class RenamerClangTidyVisitor
     if (isa<ClassTemplateSpecializationDecl>(Decl))
       return true;
 
-    Check->checkNamedDecl(Decl, *SM);
+    Check->checkNamedDecl(Decl, SM);
     return true;
   }
 
@@ -385,7 +385,7 @@ class RenamerClangTidyVisitor
 
 private:
   RenamerClangTidyCheck *Check;
-  const SourceManager *SM;
+  const SourceManager &SM;
   const bool AggressiveDependentMemberLookup;
 };
 
@@ -415,7 +415,7 @@ void RenamerClangTidyCheck::registerPPCallbacks(
 
 void RenamerClangTidyCheck::addUsage(
     const RenamerClangTidyCheck::NamingCheckId &Decl, SourceRange Range,
-    const SourceManager *SourceMgr) {
+    const SourceManager &SourceMgr) {
   // Do nothing if the provided range is invalid.
   if (Range.isInvalid())
     return;
@@ -425,8 +425,7 @@ void RenamerClangTidyCheck::addUsage(
   // spelling location to different source locations, and we only want to fix
   // the token once, before it is expanded by the macro.
   SourceLocation FixLocation = Range.getBegin();
-  if (SourceMgr)
-    FixLocation = SourceMgr->getSpellingLoc(FixLocation);
+  FixLocation = SourceMgr.getSpellingLoc(FixLocation);
   if (FixLocation.isInvalid())
     return;
 
@@ -440,15 +439,15 @@ void RenamerClangTidyCheck::addUsage(
   if (!Failure.shouldFix())
     return;
 
-  if (SourceMgr && SourceMgr->isWrittenInScratchSpace(FixLocation))
+  if (SourceMgr.isWrittenInScratchSpace(FixLocation))
     Failure.FixStatus = RenamerClangTidyCheck::ShouldFixStatus::InsideMacro;
 
-  if (!utils::rangeCanBeFixed(Range, SourceMgr))
+  if (!utils::rangeCanBeFixed(Range, &SourceMgr))
     Failure.FixStatus = RenamerClangTidyCheck::ShouldFixStatus::InsideMacro;
 }
 
 void RenamerClangTidyCheck::addUsage(const NamedDecl *Decl, SourceRange Range,
-                                     const SourceManager *SourceMgr) {
+                                     const SourceManager &SourceMgr) {
   // Don't keep track for non-identifier names.
   auto *II = Decl->getIdentifier();
   if (!II)
@@ -489,18 +488,24 @@ void RenamerClangTidyCheck::checkNamedDecl(const NamedDecl *Decl,
   }
 
   Failure.Info = std::move(Info);
-  addUsage(Decl, Range, &SourceMgr);
+  addUsage(Decl, Range, SourceMgr);
 }
 
 void RenamerClangTidyCheck::check(const MatchFinder::MatchResult &Result) {
-  RenamerClangTidyVisitor Visitor(this, Result.SourceManager,
+  if (!Result.SourceManager) {
+    // In principle SourceManager is not null but going only by the definition
+    // of MatchResult it must be handled. Cannot rename anything without a
+    // SourceManager.
+    return;
+  }
+  RenamerClangTidyVisitor Visitor(this, *Result.SourceManager,
                                   AggressiveDependentMemberLookup);
   Visitor.TraverseAST(*Result.Context);
 }
 
-void RenamerClangTidyCheck::checkMacro(const SourceManager &SourceMgr,
-                                       const Token &MacroNameTok,
-                                       const MacroInfo *MI) {
+void RenamerClangTidyCheck::checkMacro(const Token &MacroNameTok,
+                                       const MacroInfo *MI,
+                                       const SourceManager &SourceMgr) {
   std::optional<FailureInfo> MaybeFailure =
       getMacroFailureInfo(MacroNameTok, SourceMgr);
   if (!MaybeFailure)
@@ -515,11 +520,12 @@ void RenamerClangTidyCheck::checkMacro(const SourceManager &SourceMgr,
     Failure.FixStatus = ShouldFixStatus::FixInvalidIdentifier;
 
   Failure.Info = std::move(Info);
-  addUsage(ID, Range);
+  addUsage(ID, Range, SourceMgr);
 }
 
 void RenamerClangTidyCheck::expandMacro(const Token &MacroNameTok,
-                                        const MacroInfo *MI) {
+                                        const MacroInfo *MI,
+                                        const SourceManager &SourceMgr) {
   StringRef Name = MacroNameTok.getIdentifierInfo()->getName();
   NamingCheckId ID(MI->getDefinitionLoc(), Name);
 
@@ -528,7 +534,7 @@ void RenamerClangTidyCheck::expandMacro(const Token &MacroNameTok,
     return;
 
   SourceRange Range(MacroNameTok.getLocation(), MacroNameTok.getEndLoc());
-  addUsage(ID, Range);
+  addUsage(ID, Range, SourceMgr);
 }
 
 static std::string
diff --git a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.h b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.h
index 38228fb59bf62..be5b6f0c7f767 100644
--- a/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.h
+++ b/clang-tools-extra/clang-tidy/utils/RenamerClangTidyCheck.h
@@ -108,18 +108,19 @@ class RenamerClangTidyCheck : public ClangTidyCheck {
       llvm::DenseMap<NamingCheckId, NamingCheckFailure>;
 
   /// Check Macros for style violations.
-  void checkMacro(const SourceManager &SourceMgr, const Token &MacroNameTok,
-                  const MacroInfo *MI);
+  void checkMacro(const Token &MacroNameTok, const MacroInfo *MI,
+                  const SourceManager &SourceMgr);
 
   /// Add a usage of a macro if it already has a violation.
-  void expandMacro(const Token &MacroNameTok, const MacroInfo *MI);
+  void expandMacro(const Token &MacroNameTok, const MacroInfo *MI,
+                   const SourceManager &SourceMgr);
 
   void addUsage(const RenamerClangTidyCheck::NamingCheckId &Decl,
-                SourceRange Range, const SourceManager *SourceMgr = nullptr);
+                SourceRange Range, const SourceManager &SourceMgr);
 
   /// Convenience method when the usage to be added is a NamedDecl.
   void addUsage(const NamedDecl *Decl, SourceRange Range,
-                const SourceManager *SourceMgr = nullptr);
+                const SourceManager &SourceMgr);
 
   void checkNamedDecl(const NamedDecl *Decl, const SourceManager &SourceMgr);
 

From 3652b2a877c3691b524a19bc38b338f1ba1dde37 Mon Sep 17 00:00:00 2001
From: Mike Rice <michael.p.rice@intel.com>
Date: Fri, 12 Apr 2024 10:01:01 -0700
Subject: [PATCH 273/886] [clang][CodeGen][OpenMP] Fix casting of atomic update
 of ptr types (#88215)

In 4d5e834c5b7f0ccccd90a6d543e182df602f6bc8, casts were removed for
pointers but one case was missed. Add missing check.
---
 clang/lib/CodeGen/CGAtomic.cpp              |  6 +++++-
 clang/test/OpenMP/atomic_update_codegen.cpp | 11 +++++++++++
 2 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index d35ce0409d723..07452b18a85ea 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -1806,7 +1806,11 @@ void AtomicInfo::EmitAtomicUpdateOp(
                                              /*NumReservedValues=*/2);
   PHI->addIncoming(OldVal, CurBB);
   Address NewAtomicAddr = CreateTempAlloca();
-  Address NewAtomicIntAddr = castToAtomicIntPointer(NewAtomicAddr);
+  Address NewAtomicIntAddr =
+      shouldCastToInt(NewAtomicAddr.getElementType(), /*CmpXchg=*/true)
+          ? castToAtomicIntPointer(NewAtomicAddr)
+          : NewAtomicAddr;
+
   if ((LVal.isBitField() && BFI.Size != ValueSizeInBits) ||
       requiresMemSetZero(getAtomicAddress().getElementType())) {
     CGF.Builder.CreateStore(PHI, NewAtomicIntAddr);
diff --git a/clang/test/OpenMP/atomic_update_codegen.cpp b/clang/test/OpenMP/atomic_update_codegen.cpp
index ce0765118922a..fe745590a9f91 100644
--- a/clang/test/OpenMP/atomic_update_codegen.cpp
+++ b/clang/test/OpenMP/atomic_update_codegen.cpp
@@ -27,6 +27,7 @@ long double ldv, ldx;
 _Complex int civ, cix;
 _Complex float cfv, cfx;
 _Complex double cdv, cdx;
+char *cpx;
 
 typedef int int4 __attribute__((__vector_size__(16)));
 int4 int4x;
@@ -851,6 +852,16 @@ int main(void) {
 // CHECK: call{{.*}} @__kmpc_flush(
 #pragma omp atomic seq_cst
   rix = dv / rix;
+
+// CHECK: [[LD_CPX:%.+]] = load atomic ptr, ptr @cpx monotonic
+// CHECK: br label %[[CONT:.+]]
+// CHECK: [[CONT]]
+// CHECK: [[PHI:%.+]] = phi ptr
+// CHECK: [[RES:%.+]] = cmpxchg ptr @cpx,
+// CHECK: br i1 %{{.+}}, label %[[EXIT:.+]], label %[[CONT]]
+  #pragma omp atomic update
+  cpx += 1;
+
   return 0;
 }
 

From d61edecbfd099143e0e2617505fec921524938f8 Mon Sep 17 00:00:00 2001
From: Owen Pan <owenpiano@gmail.com>
Date: Fri, 12 Apr 2024 10:12:24 -0700
Subject: [PATCH 274/886] [clang-format] Fix a regression in
 ContinuationIndenter (#88414)

Commit d06b92391513 caused a regression that breaks after a block
comment adjacent to a function paramter that follows.

Fixes #86573.
---
 clang/lib/Format/ContinuationIndenter.cpp     | 8 +++++++-
 clang/unittests/Format/FormatTestComments.cpp | 4 ++++
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/clang/lib/Format/ContinuationIndenter.cpp b/clang/lib/Format/ContinuationIndenter.cpp
index 700bce35c8683..ad0e2c3c620c3 100644
--- a/clang/lib/Format/ContinuationIndenter.cpp
+++ b/clang/lib/Format/ContinuationIndenter.cpp
@@ -684,7 +684,13 @@ void ContinuationIndenter::addTokenOnCurrentLine(LineState &State, bool DryRun,
         // arguments to function calls. We do this by ensuring that either all
         // arguments (including any lambdas) go on the same line as the function
         // call, or we break before the first argument.
-        auto PrevNonComment = Current.getPreviousNonComment();
+        const auto *Prev = Current.Previous;
+        if (!Prev)
+          return false;
+        // For example, `/*Newline=*/false`.
+        if (Prev->is(TT_BlockComment) && Current.SpacesRequiredBefore == 0)
+          return false;
+        const auto *PrevNonComment = Current.getPreviousNonComment();
         if (!PrevNonComment || PrevNonComment->isNot(tok::l_paren))
           return false;
         if (Current.isOneOf(tok::comment, tok::l_paren, TT_LambdaLSquare))
diff --git a/clang/unittests/Format/FormatTestComments.cpp b/clang/unittests/Format/FormatTestComments.cpp
index d705cf34d8af0..d2baace6a7d80 100644
--- a/clang/unittests/Format/FormatTestComments.cpp
+++ b/clang/unittests/Format/FormatTestComments.cpp
@@ -376,6 +376,10 @@ TEST_F(FormatTestComments, RemovesTrailingWhitespaceOfComments) {
 TEST_F(FormatTestComments, UnderstandsBlockComments) {
   verifyFormat("f(/*noSpaceAfterParameterNamingComment=*/true);");
   verifyFormat("void f() { g(/*aaa=*/x, /*bbb=*/!y, /*c=*/::c); }");
+  verifyFormat("fooooooooooooooooooooooooooooo(\n"
+               "    /*qq_=*/move(q), [this, b](bar<void(uint32_t)> b) {},\n"
+               "    c);",
+               getLLVMStyleWithColumns(60));
   EXPECT_EQ("f(aaaaaaaaaaaaaaaaaaaaaaaaa, /* Trailing comment for aa... */\n"
             "  bbbbbbbbbbbbbbbbbbbbbbbbb);",
             format("f(aaaaaaaaaaaaaaaaaaaaaaaaa ,   \\\n"

From dcd097c475163b9bd41ff009a9157e86c6f2f171 Mon Sep 17 00:00:00 2001
From: Stanislav Mekhanoshin <rampitec@users.noreply.github.com>
Date: Fri, 12 Apr 2024 10:15:59 -0700
Subject: [PATCH 275/886] Add IIT_V6 to support 6-element vectors in
 intrinsics. (#88196)

Needed for the future patch.
---
 llvm/include/llvm/IR/Intrinsics.td | 1 +
 llvm/lib/IR/Function.cpp           | 4 ++++
 2 files changed, 5 insertions(+)

diff --git a/llvm/include/llvm/IR/Intrinsics.td b/llvm/include/llvm/IR/Intrinsics.td
index f0723a633f0fc..bdd8465883fcf 100644
--- a/llvm/include/llvm/IR/Intrinsics.td
+++ b/llvm/include/llvm/IR/Intrinsics.td
@@ -319,6 +319,7 @@ def IIT_FUNCREF : IIT_VT<funcref, 55>;
 def IIT_I2 : IIT_Int<2, 57>;
 def IIT_I4 : IIT_Int<4, 58>;
 def IIT_AARCH64_SVCOUNT : IIT_VT<aarch64svcount, 59>;
+def IIT_V6 : IIT_Vec<6, 60>;
 }
 
 defvar IIT_all_FixedTypes = !filter(iit, IIT_all,
diff --git a/llvm/lib/IR/Function.cpp b/llvm/lib/IR/Function.cpp
index b5fda9bb3d129..96953ac49c19b 100644
--- a/llvm/lib/IR/Function.cpp
+++ b/llvm/lib/IR/Function.cpp
@@ -1172,6 +1172,10 @@ static void DecodeIITType(unsigned &NextElt, ArrayRef<unsigned char> Infos,
     OutputTable.push_back(IITDescriptor::getVector(4, IsScalableVector));
     DecodeIITType(NextElt, Infos, Info, OutputTable);
     return;
+  case IIT_V6:
+    OutputTable.push_back(IITDescriptor::getVector(6, IsScalableVector));
+    DecodeIITType(NextElt, Infos, Info, OutputTable);
+    return;
   case IIT_V8:
     OutputTable.push_back(IITDescriptor::getVector(8, IsScalableVector));
     DecodeIITType(NextElt, Infos, Info, OutputTable);

From 6a85cf8fc0437d4885fc829948befe32c1e5a21d Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Fri, 12 Apr 2024 10:22:21 -0700
Subject: [PATCH 276/886] [SelectionDAG] Verify SPLAT_VECTOR nodes when they
 are created. (#88305)

This applies the same rules we have for the scalar operands of a
BUILD_VECTOR where the scalar type must match the element type or for
integer vectors we allow the scalar type to be larger than the element
type. Hexagon uses i32 for an FP zero vector so we allow that as an
exception.
---
 llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index ddb12a101e02c..412e1de9fc41c 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -6032,6 +6032,17 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
     if (N1.getValueType().getScalarType() == MVT::i1)
       return getNode(ISD::VECREDUCE_AND, DL, VT, N1);
     break;
+  case ISD::SPLAT_VECTOR:
+    assert(VT.isVector() && "Wrong return type!");
+    // FIXME: Hexagon uses i32 scalar for a floating point zero vector so allow
+    // that for now.
+    assert((VT.getVectorElementType() == N1.getValueType() ||
+            (VT.isFloatingPoint() && N1.getValueType() == MVT::i32) ||
+            (VT.getVectorElementType().isInteger() &&
+             N1.getValueType().isInteger() &&
+             VT.getVectorElementType().bitsLE(N1.getValueType()))) &&
+           "Wrong operand type!");
+    break;
   }
 
   SDNode *N;

From 7cafe04e0d74c1e9f7e3871a0bcdf4ccb1c89f0c Mon Sep 17 00:00:00 2001
From: Xiaoyang Liu <siujoeng.lau@gmail.com>
Date: Fri, 12 Apr 2024 13:25:22 -0400
Subject: [PATCH 277/886] [libc++] P3029R1: Better `mdspan`'s CTAD (#87873)

## Abstract

This pull request implements [P3029R1](https://wg21.link/P3029R1). The
paper discusses the current behavior of `mdspan`'s most common
pointer-indices CTAD, where the `Extents` template parameter is deduced
as `dextents` (dynamic extents), even when passing compile-time constant
values. The author believes this behavior is suboptimal, as it doesn't
take advantage of the compile-time information. The proposed change
suggests deducing static extents if `integral_constant`-like constants
are passed, resulting in more intuitive syntax and less error-prone
code.

## Reference

- [P3029R1](https://wg21.link/P3029R1)
- [Draft C++ Standard: [span.syn]](https://eel.is/c++draft/span.syn)
- [Draft C++ Standard: [mdspan.syn]](https://eel.is/c++draft/mdspan.syn)
---
 libcxx/docs/ReleaseNotes/19.rst               |  1 +
 libcxx/docs/Status/Cxx2cPapers.csv            |  2 +-
 libcxx/include/__mdspan/mdspan.h              |  7 +++
 libcxx/include/mdspan                         |  6 ++-
 libcxx/include/span                           | 44 ++++++++++++++++++-
 .../views/mdspan/mdspan/deduction.pass.cpp    | 18 +++++++-
 .../views.span/span.cons/deduct.pass.cpp      | 11 +++++
 7 files changed, 84 insertions(+), 5 deletions(-)

diff --git a/libcxx/docs/ReleaseNotes/19.rst b/libcxx/docs/ReleaseNotes/19.rst
index 81c05b9112bd2..7bc0148c9ff0a 100644
--- a/libcxx/docs/ReleaseNotes/19.rst
+++ b/libcxx/docs/ReleaseNotes/19.rst
@@ -47,6 +47,7 @@ Implemented Papers
 - P3142R0 - Printing Blank Lines with ``println`` (as DR against C++23)
 - P2302R4 - ``std::ranges::contains``
 - P1659R3 - ``std::ranges::starts_with`` and ``std::ranges::ends_with``
+- P3029R1 - Better ``mdspan``'s CTAD
 
 Improvements and New Features
 -----------------------------
diff --git a/libcxx/docs/Status/Cxx2cPapers.csv b/libcxx/docs/Status/Cxx2cPapers.csv
index fa11da62bc080..8516070c72b8f 100644
--- a/libcxx/docs/Status/Cxx2cPapers.csv
+++ b/libcxx/docs/Status/Cxx2cPapers.csv
@@ -61,5 +61,5 @@
 "`P1068R11 <https://wg21.link/P1068R11>`__","LWG","Vector API for random number generation","Tokyo March 2024","","",""
 "`P2944R3 <https://wg21.link/P2944R3>`__","LWG","Comparisons for ``reference_wrapper``","Tokyo March 2024","","",""
 "`P2642R6 <https://wg21.link/P2642R6>`__","LWG","Padded ``mdspan`` layouts","Tokyo March 2024","","",""
-"`P3029R1 <https://wg21.link/P3029R1>`__","LWG","Better ``mdspan``'s CTAD","Tokyo March 2024","","",""
+"`P3029R1 <https://wg21.link/P3029R1>`__","LWG","Better ``mdspan``'s CTAD","Tokyo March 2024","|Complete|","19.0",""
 "","","","","","",""
diff --git a/libcxx/include/__mdspan/mdspan.h b/libcxx/include/__mdspan/mdspan.h
index d9a0108d6d350..09665af316e71 100644
--- a/libcxx/include/__mdspan/mdspan.h
+++ b/libcxx/include/__mdspan/mdspan.h
@@ -266,10 +266,17 @@ class mdspan {
   friend class mdspan;
 };
 
+#  if _LIBCPP_STD_VER >= 26
+template <class _ElementType, class... _OtherIndexTypes>
+  requires((is_convertible_v<_OtherIndexTypes, size_t> && ...) && (sizeof...(_OtherIndexTypes) > 0))
+explicit mdspan(_ElementType*,
+                _OtherIndexTypes...) -> mdspan<_ElementType, extents<size_t, __maybe_static_ext<_OtherIndexTypes>...>>;
+#  else
 template <class _ElementType, class... _OtherIndexTypes>
   requires((is_convertible_v<_OtherIndexTypes, size_t> && ...) && (sizeof...(_OtherIndexTypes) > 0))
 explicit mdspan(_ElementType*, _OtherIndexTypes...)
     -> mdspan<_ElementType, dextents<size_t, sizeof...(_OtherIndexTypes)>>;
+#  endif
 
 template <class _Pointer>
   requires(is_pointer_v<remove_reference_t<_Pointer>>)
diff --git a/libcxx/include/mdspan b/libcxx/include/mdspan
index c13d9eef001ac..8d443f4acd1dd 100644
--- a/libcxx/include/mdspan
+++ b/libcxx/include/mdspan
@@ -370,7 +370,11 @@ namespace std {
   template<class ElementType, class... Integrals>
     requires((is_convertible_v<Integrals, size_t> && ...) && sizeof...(Integrals) > 0)
     explicit mdspan(ElementType*, Integrals...)
-      -> mdspan<ElementType, dextents<size_t, sizeof...(Integrals)>>;
+      -> mdspan<ElementType, dextents<size_t, sizeof...(Integrals)>>;            // until C++26
+  template<class ElementType, class... Integrals>
+    requires((is_convertible_v<Integrals, size_t> && ...) && sizeof...(Integrals) > 0)
+    explicit mdspan(ElementType*, Integrals...)
+      -> mdspan<ElementType, extents<size_t, maybe-static-ext<Integrals>...>>;  // since C++26
 
   template<class ElementType, class OtherIndexType, size_t N>
     mdspan(ElementType*, span<OtherIndexType, N>)
diff --git a/libcxx/include/span b/libcxx/include/span
index c0fe25ddb4beb..0307edcb55c30 100644
--- a/libcxx/include/span
+++ b/libcxx/include/span
@@ -18,6 +18,20 @@ namespace std {
 // constants
 inline constexpr size_t dynamic_extent = numeric_limits<size_t>::max();
 
+template<class T>
+  concept integral-constant-like =                           // exposition only, since C++26
+    is_integral_v<decltype(T::value)> &&
+    !is_same_v<bool, remove_const_t<decltype(T::value)>> &&
+    convertible_to<T, decltype(T::value)> &&
+    equality_comparable_with<T, decltype(T::value)> &&
+    bool_constant<T() == T::value>::value &&
+    bool_constant<static_cast<decltype(T::value)>(T()) == T::value>::value;
+
+template<class T>
+  constexpr size_t maybe-static-ext = dynamic_extent;        // exposition only, since C++26
+template<integral-constant-like T>
+  constexpr size_t maybe-static-ext<T> = {T::value};
+
 // [views.span], class template span
 template <class ElementType, size_t Extent = dynamic_extent>
     class span;
@@ -110,7 +124,9 @@ private:
 };
 
 template<class It, class EndOrSize>
-    span(It, EndOrSize) -> span<remove_reference_t<iter_reference_t<_It>>>;
+    span(It, EndOrSize) -> span<remove_reference_t<iter_reference_t<_It>>>;                             // until C++26
+template<class It, class EndOrSize>
+    span(It, EndOrSize) -> span<remove_reference_t<iter_reference_t<It>>, maybe-static-ext<EndOrSize>>; // since C++26
 
 template<class T, size_t N>
     span(T (&)[N]) -> span<T, N>;
@@ -129,6 +145,8 @@ template<class R>
 */
 
 #include <__assert>
+#include <__concepts/convertible_to.h>
+#include <__concepts/equality_comparable.h>
 #include <__config>
 #include <__fwd/array.h>
 #include <__fwd/span.h>
@@ -143,15 +161,19 @@ template<class R>
 #include <__ranges/enable_borrowed_range.h>
 #include <__ranges/enable_view.h>
 #include <__ranges/size.h>
+#include <__type_traits/integral_constant.h>
 #include <__type_traits/is_array.h>
 #include <__type_traits/is_const.h>
 #include <__type_traits/is_convertible.h>
+#include <__type_traits/is_integral.h>
+#include <__type_traits/is_same.h>
+#include <__type_traits/remove_const.h>
 #include <__type_traits/remove_cv.h>
 #include <__type_traits/remove_cvref.h>
 #include <__type_traits/remove_reference.h>
 #include <__type_traits/type_identity.h>
 #include <__utility/forward.h>
-#include <cstddef>      // for byte
+#include <cstddef> // for byte
 #include <initializer_list>
 #include <stdexcept>
 #include <version>
@@ -563,8 +585,26 @@ _LIBCPP_HIDE_FROM_ABI auto as_writable_bytes(span<_Tp, _Extent> __s) noexcept {
   return __s.__as_writable_bytes();
 }
 
+#  if _LIBCPP_STD_VER >= 26
+template <class _Tp>
+concept __integral_constant_like =
+    is_integral_v<decltype(_Tp::value)> && !is_same_v<bool, remove_const_t<decltype(_Tp::value)>> &&
+    convertible_to<_Tp, decltype(_Tp::value)> && equality_comparable_with<_Tp, decltype(_Tp::value)> &&
+    bool_constant<_Tp() == _Tp::value>::value &&
+    bool_constant<static_cast<decltype(_Tp::value)>(_Tp()) == _Tp::value>::value;
+
+template <class _Tp>
+inline constexpr size_t __maybe_static_ext = dynamic_extent;
+
+template <__integral_constant_like _Tp>
+inline constexpr size_t __maybe_static_ext<_Tp> = {_Tp::value};
+
+template <contiguous_iterator _It, class _EndOrSize>
+span(_It, _EndOrSize) -> span<remove_reference_t<iter_reference_t<_It>>, __maybe_static_ext<_EndOrSize>>;
+#  else
 template <contiguous_iterator _It, class _EndOrSize>
 span(_It, _EndOrSize) -> span<remove_reference_t<iter_reference_t<_It>>>;
+#  endif
 
 template <class _Tp, size_t _Sz>
 span(_Tp (&)[_Sz]) -> span<_Tp, _Sz>;
diff --git a/libcxx/test/std/containers/views/mdspan/mdspan/deduction.pass.cpp b/libcxx/test/std/containers/views/mdspan/mdspan/deduction.pass.cpp
index 557829dcbad9b..876a3e84d6957 100644
--- a/libcxx/test/std/containers/views/mdspan/mdspan/deduction.pass.cpp
+++ b/libcxx/test/std/containers/views/mdspan/mdspan/deduction.pass.cpp
@@ -22,7 +22,11 @@
 //  template<class ElementType, class... Integrals>
 //    requires((is_convertible_v<Integrals, size_t> && ...) && sizeof...(Integrals) > 0)
 //    explicit mdspan(ElementType*, Integrals...)
-//      -> mdspan<ElementType, dextents<size_t, sizeof...(Integrals)>>;
+//      -> mdspan<ElementType, dextents<size_t, sizeof...(Integrals)>>;            // until C++26
+//  template<class ElementType, class... Integrals>
+//    requires((is_convertible_v<Integrals, size_t> && ...) && sizeof...(Integrals) > 0)
+//    explicit mdspan(ElementType*, Integrals...)
+//      -> mdspan<ElementType, extents<size_t, maybe-static-ext<Integrals>...>>;  // since C++26
 //
 //  template<class ElementType, class OtherIndexType, size_t N>
 //    mdspan(ElementType*, span<OtherIndexType, N>)
@@ -102,6 +106,18 @@ constexpr bool test_no_layout_deduction_guides(const H& handle, const A&) {
   // deduction from pointer and integral like
   ASSERT_SAME_TYPE(decltype(std::mdspan(handle, 5, SizeTIntType(6))), std::mdspan<T, std::dextents<size_t, 2>>);
 
+#if _LIBCPP_STD_VER >= 26
+  // P3029R1: deduction from `integral_constant`
+  ASSERT_SAME_TYPE(
+      decltype(std::mdspan(handle, std::integral_constant<size_t, 5>{})), std::mdspan<T, std::extents<size_t, 5>>);
+  ASSERT_SAME_TYPE(decltype(std::mdspan(handle, std::integral_constant<size_t, 5>{}, std::dynamic_extent)),
+                   std::mdspan<T, std::extents<size_t, 5, std::dynamic_extent>>);
+  ASSERT_SAME_TYPE(
+      decltype(std::mdspan(
+          handle, std::integral_constant<size_t, 5>{}, std::dynamic_extent, std::integral_constant<size_t, 7>{})),
+      std::mdspan<T, std::extents<size_t, 5, std::dynamic_extent, 7>>);
+#endif
+
   std::array<char, 3> exts;
   // deduction from pointer and array
   ASSERT_SAME_TYPE(decltype(std::mdspan(handle, exts)), std::mdspan<T, std::dextents<size_t, 3>>);
diff --git a/libcxx/test/std/containers/views/views.span/span.cons/deduct.pass.cpp b/libcxx/test/std/containers/views/views.span/span.cons/deduct.pass.cpp
index 398862dccdb6a..92bb7c9caea83 100644
--- a/libcxx/test/std/containers/views/views.span/span.cons/deduct.pass.cpp
+++ b/libcxx/test/std/containers/views/views.span/span.cons/deduct.pass.cpp
@@ -31,6 +31,7 @@
 #include <iterator>
 #include <memory>
 #include <string>
+#include <type_traits>
 
 #include "test_macros.h"
 
@@ -48,6 +49,16 @@ void test_iterator_sentinel() {
   assert(s.size() == std::size(arr));
   assert(s.data() == std::data(arr));
   }
+
+#if _LIBCPP_STD_VER >= 26
+  // P3029R1: deduction from `integral_constant`
+  {
+    std::span s{std::begin(arr), std::integral_constant<size_t, 3>{}};
+    ASSERT_SAME_TYPE(decltype(s), std::span<int, 3>);
+    assert(s.size() == std::size(arr));
+    assert(s.data() == std::data(arr));
+  }
+#endif
 }
 
 void test_c_array() {

From fe59cb256289d9f2f6ebe75571389f3cc90bbdec Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 09:43:10 -0700
Subject: [PATCH 278/886] [clang][deps] NFCI: Extract `FilenameForLookup` into
 new function

This is an NFC change split from https://github.com/llvm/llvm-project/pull/68645.
---
 .../DependencyScanningFilesystem.h            |  4 ++
 .../DependencyScanningFilesystem.cpp          | 41 ++++++++++++-------
 2 files changed, 30 insertions(+), 15 deletions(-)

diff --git a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
index 9a522a3e2fe25..4cd0f958fcff8 100644
--- a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
+++ b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
@@ -402,6 +402,10 @@ class DependencyScanningWorkerFilesystem
   llvm::ErrorOr<std::string> WorkingDirForCacheLookup;
 
   void updateWorkingDirForCacheLookup();
+
+  llvm::ErrorOr<StringRef>
+  tryGetFilenameForLookup(StringRef OriginalFilename,
+                          llvm::SmallVectorImpl<char> &PathBuf) const;
 };
 
 } // end namespace dependencies
diff --git a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
index 9b7812a1adb9e..c66780d50fa18 100644
--- a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
+++ b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
@@ -233,24 +233,15 @@ DependencyScanningWorkerFilesystem::computeAndStoreResult(
 llvm::ErrorOr<EntryRef>
 DependencyScanningWorkerFilesystem::getOrCreateFileSystemEntry(
     StringRef OriginalFilename) {
-  StringRef FilenameForLookup;
   SmallString<256> PathBuf;
-  if (llvm::sys::path::is_absolute_gnu(OriginalFilename)) {
-    FilenameForLookup = OriginalFilename;
-  } else if (!WorkingDirForCacheLookup) {
-    return WorkingDirForCacheLookup.getError();
-  } else {
-    StringRef RelFilename = OriginalFilename;
-    RelFilename.consume_front("./");
-    PathBuf = *WorkingDirForCacheLookup;
-    llvm::sys::path::append(PathBuf, RelFilename);
-    FilenameForLookup = PathBuf.str();
-  }
-  assert(llvm::sys::path::is_absolute_gnu(FilenameForLookup));
+  auto FilenameForLookup = tryGetFilenameForLookup(OriginalFilename, PathBuf);
+  if (!FilenameForLookup)
+    return FilenameForLookup.getError();
+
   if (const auto *Entry =
-          findEntryByFilenameWithWriteThrough(FilenameForLookup))
+          findEntryByFilenameWithWriteThrough(*FilenameForLookup))
     return EntryRef(OriginalFilename, *Entry).unwrapError();
-  auto MaybeEntry = computeAndStoreResult(OriginalFilename, FilenameForLookup);
+  auto MaybeEntry = computeAndStoreResult(OriginalFilename, *FilenameForLookup);
   if (!MaybeEntry)
     return MaybeEntry.getError();
   return EntryRef(OriginalFilename, *MaybeEntry).unwrapError();
@@ -351,4 +342,24 @@ void DependencyScanningWorkerFilesystem::updateWorkingDirForCacheLookup() {
          llvm::sys::path::is_absolute_gnu(*WorkingDirForCacheLookup));
 }
 
+llvm::ErrorOr<StringRef>
+DependencyScanningWorkerFilesystem::tryGetFilenameForLookup(
+    StringRef OriginalFilename, llvm::SmallVectorImpl<char> &PathBuf) const {
+  StringRef FilenameForLookup;
+  if (llvm::sys::path::is_absolute_gnu(OriginalFilename)) {
+    FilenameForLookup = OriginalFilename;
+  } else if (!WorkingDirForCacheLookup) {
+    return WorkingDirForCacheLookup.getError();
+  } else {
+    StringRef RelFilename = OriginalFilename;
+    RelFilename.consume_front("./");
+    PathBuf.assign(WorkingDirForCacheLookup->begin(),
+                   WorkingDirForCacheLookup->end());
+    llvm::sys::path::append(PathBuf, RelFilename);
+    FilenameForLookup = StringRef{PathBuf.begin(), PathBuf.size()};
+  }
+  assert(llvm::sys::path::is_absolute_gnu(FilenameForLookup));
+  return FilenameForLookup;
+}
+
 const char DependencyScanningWorkerFilesystem::ID = 0;

From edd7fed9da48c0e708cce9bd4d305ae43d8bd77c Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 10:09:04 -0700
Subject: [PATCH 279/886] [llvm][vfs] NFCI: Remove `const` from
 `VFS::getRealPath()`

This is an NFC change split from https://github.com/llvm/llvm-project/pull/68645.
---
 llvm/include/llvm/Support/VirtualFileSystem.h | 10 ++++-----
 llvm/lib/Support/FileCollector.cpp            |  2 +-
 llvm/lib/Support/VirtualFileSystem.cpp        | 21 ++++++++-----------
 .../Support/VirtualFileSystemTest.cpp         |  2 +-
 4 files changed, 16 insertions(+), 19 deletions(-)

diff --git a/llvm/include/llvm/Support/VirtualFileSystem.h b/llvm/include/llvm/Support/VirtualFileSystem.h
index 770ca8764426a..4b1ca0c3d262b 100644
--- a/llvm/include/llvm/Support/VirtualFileSystem.h
+++ b/llvm/include/llvm/Support/VirtualFileSystem.h
@@ -298,7 +298,7 @@ class FileSystem : public llvm::ThreadSafeRefCountedBase<FileSystem>,
   /// symlinks. For real file system, this uses `llvm::sys::fs::real_path`.
   /// This returns errc::operation_not_permitted if not implemented by subclass.
   virtual std::error_code getRealPath(const Twine &Path,
-                                      SmallVectorImpl<char> &Output) const;
+                                      SmallVectorImpl<char> &Output);
 
   /// Check whether a file exists. Provided for convenience.
   bool exists(const Twine &Path);
@@ -393,7 +393,7 @@ class OverlayFileSystem : public RTTIExtends<OverlayFileSystem, FileSystem> {
   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
   std::error_code isLocal(const Twine &Path, bool &Result) override;
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override;
+                              SmallVectorImpl<char> &Output) override;
 
   using iterator = FileSystemList::reverse_iterator;
   using const_iterator = FileSystemList::const_reverse_iterator;
@@ -453,7 +453,7 @@ class ProxyFileSystem : public RTTIExtends<ProxyFileSystem, FileSystem> {
     return FS->setCurrentWorkingDirectory(Path);
   }
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override {
+                              SmallVectorImpl<char> &Output) override {
     return FS->getRealPath(Path, Output);
   }
   std::error_code isLocal(const Twine &Path, bool &Result) override {
@@ -622,7 +622,7 @@ class InMemoryFileSystem : public RTTIExtends<InMemoryFileSystem, FileSystem> {
   /// This doesn't resolve symlinks as they are not supported in in-memory file
   /// system.
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override;
+                              SmallVectorImpl<char> &Output) override;
   std::error_code isLocal(const Twine &Path, bool &Result) override;
   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
 
@@ -1047,7 +1047,7 @@ class RedirectingFileSystem
   ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
 
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override;
+                              SmallVectorImpl<char> &Output) override;
 
   llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override;
 
diff --git a/llvm/lib/Support/FileCollector.cpp b/llvm/lib/Support/FileCollector.cpp
index 92bcdf00b8a82..be0b06b0aff80 100644
--- a/llvm/lib/Support/FileCollector.cpp
+++ b/llvm/lib/Support/FileCollector.cpp
@@ -281,7 +281,7 @@ class FileCollectorFileSystem : public vfs::FileSystem {
   }
 
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override {
+                              SmallVectorImpl<char> &Output) override {
     auto EC = FS->getRealPath(Path, Output);
     if (!EC) {
       Collector->addFile(Path);
diff --git a/llvm/lib/Support/VirtualFileSystem.cpp b/llvm/lib/Support/VirtualFileSystem.cpp
index 057f8eae0552c..32b480028e71b 100644
--- a/llvm/lib/Support/VirtualFileSystem.cpp
+++ b/llvm/lib/Support/VirtualFileSystem.cpp
@@ -138,7 +138,7 @@ std::error_code FileSystem::makeAbsolute(SmallVectorImpl<char> &Path) const {
 }
 
 std::error_code FileSystem::getRealPath(const Twine &Path,
-                                        SmallVectorImpl<char> &Output) const {
+                                        SmallVectorImpl<char> &Output) {
   return errc::operation_not_permitted;
 }
 
@@ -275,7 +275,7 @@ class RealFileSystem : public FileSystem {
   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
   std::error_code isLocal(const Twine &Path, bool &Result) override;
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override;
+                              SmallVectorImpl<char> &Output) override;
 
 protected:
   void printImpl(raw_ostream &OS, PrintType Type,
@@ -357,9 +357,8 @@ std::error_code RealFileSystem::isLocal(const Twine &Path, bool &Result) {
   return llvm::sys::fs::is_local(adjustPath(Path, Storage), Result);
 }
 
-std::error_code
-RealFileSystem::getRealPath(const Twine &Path,
-                            SmallVectorImpl<char> &Output) const {
+std::error_code RealFileSystem::getRealPath(const Twine &Path,
+                                            SmallVectorImpl<char> &Output) {
   SmallString<256> Storage;
   return llvm::sys::fs::real_path(adjustPath(Path, Storage), Output);
 }
@@ -471,9 +470,8 @@ std::error_code OverlayFileSystem::isLocal(const Twine &Path, bool &Result) {
   return errc::no_such_file_or_directory;
 }
 
-std::error_code
-OverlayFileSystem::getRealPath(const Twine &Path,
-                               SmallVectorImpl<char> &Output) const {
+std::error_code OverlayFileSystem::getRealPath(const Twine &Path,
+                                               SmallVectorImpl<char> &Output) {
   for (const auto &FS : FSList)
     if (FS->exists(Path))
       return FS->getRealPath(Path, Output);
@@ -1157,9 +1155,8 @@ std::error_code InMemoryFileSystem::setCurrentWorkingDirectory(const Twine &P) {
   return {};
 }
 
-std::error_code
-InMemoryFileSystem::getRealPath(const Twine &Path,
-                                SmallVectorImpl<char> &Output) const {
+std::error_code InMemoryFileSystem::getRealPath(const Twine &Path,
+                                                SmallVectorImpl<char> &Output) {
   auto CWD = getCurrentWorkingDirectory();
   if (!CWD || CWD->empty())
     return errc::operation_not_permitted;
@@ -2535,7 +2532,7 @@ RedirectingFileSystem::openFileForRead(const Twine &OriginalPath) {
 
 std::error_code
 RedirectingFileSystem::getRealPath(const Twine &OriginalPath,
-                                   SmallVectorImpl<char> &Output) const {
+                                   SmallVectorImpl<char> &Output) {
   SmallString<256> Path;
   OriginalPath.toVector(Path);
 
diff --git a/llvm/unittests/Support/VirtualFileSystemTest.cpp b/llvm/unittests/Support/VirtualFileSystemTest.cpp
index 695b09343257f..49a2e19e4f74c 100644
--- a/llvm/unittests/Support/VirtualFileSystemTest.cpp
+++ b/llvm/unittests/Support/VirtualFileSystemTest.cpp
@@ -82,7 +82,7 @@ class DummyFileSystem : public vfs::FileSystem {
   }
   // Map any symlink to "/symlink".
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override {
+                              SmallVectorImpl<char> &Output) override {
     auto I = findEntry(Path);
     if (I == FilesAndDirs.end())
       return make_error_code(llvm::errc::no_such_file_or_directory);

From c11976f525f3b9b9dc6080f0b23d2ef1ec5fe8fd Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 10:26:58 -0700
Subject: [PATCH 280/886] [clang][deps] Create separate unittest directory

This is an NFC change split from https://github.com/llvm/llvm-project/pull/68645.
---
 clang/unittests/Tooling/CMakeLists.txt                          | 2 +-
 .../Tooling/{ => DependencyScanning}/DependencyScannerTest.cpp  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
 rename clang/unittests/Tooling/{ => DependencyScanning}/DependencyScannerTest.cpp (99%)

diff --git a/clang/unittests/Tooling/CMakeLists.txt b/clang/unittests/Tooling/CMakeLists.txt
index 5a10a6b285390..2ff493ef5fc32 100644
--- a/clang/unittests/Tooling/CMakeLists.txt
+++ b/clang/unittests/Tooling/CMakeLists.txt
@@ -13,7 +13,6 @@ add_clang_unittest(ToolingTests
   CastExprTest.cpp
   CommentHandlerTest.cpp
   CompilationDatabaseTest.cpp
-  DependencyScannerTest.cpp
   DiagnosticsYamlTest.cpp
   ExecutionTest.cpp
   FixItTest.cpp
@@ -24,6 +23,7 @@ add_clang_unittest(ToolingTests
   LookupTest.cpp
   QualTypeNamesTest.cpp
   RangeSelectorTest.cpp
+  DependencyScanning/DependencyScannerTest.cpp
   RecursiveASTVisitorTests/Attr.cpp
   RecursiveASTVisitorTests/BitfieldInitializer.cpp
   RecursiveASTVisitorTests/CallbacksLeaf.cpp
diff --git a/clang/unittests/Tooling/DependencyScannerTest.cpp b/clang/unittests/Tooling/DependencyScanning/DependencyScannerTest.cpp
similarity index 99%
rename from clang/unittests/Tooling/DependencyScannerTest.cpp
rename to clang/unittests/Tooling/DependencyScanning/DependencyScannerTest.cpp
index 8735fcad20046..ec0e143be4a20 100644
--- a/clang/unittests/Tooling/DependencyScannerTest.cpp
+++ b/clang/unittests/Tooling/DependencyScanning/DependencyScannerTest.cpp
@@ -1,4 +1,4 @@
-//===- unittest/Tooling/ToolingTest.cpp - Tooling unit tests --------------===//
+//===- DependencyScannerTest.cpp ------------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.

From a11a4324bb27c01e7a005e1a7f49fb8284098e8c Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 10:34:42 -0700
Subject: [PATCH 281/886] [clang][deps] Cache `VFS::getRealPath()` (#68645)

This PR starts caching calls to
`DependencyScanningWorkerFilesystem::getRealPath()` that we use whenever
we canonicalize module map path. In the case of the real VFS, this
functions performs an expensive syscall that we'd like to do as rarely
as possible.

This PR keeps the real path out of `CachedFileSystemEntry`, since that's
**immutable**; populating the real path on creation of this data
structure (every stat/open) would be expensive.
---
 .../DependencyScanningFilesystem.h            |  71 +++++++-
 .../DependencyScanningFilesystem.cpp          | 109 +++++++++++--
 clang/unittests/Tooling/CMakeLists.txt        |   1 +
 .../DependencyScanningFilesystemTest.cpp      | 152 ++++++++++++++++++
 4 files changed, 314 insertions(+), 19 deletions(-)
 create mode 100644 clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp

diff --git a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
index 4cd0f958fcff8..8b6f149c7cb26 100644
--- a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
+++ b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
@@ -142,6 +142,8 @@ class CachedFileSystemEntry {
   CachedFileContents *Contents;
 };
 
+using CachedRealPath = llvm::ErrorOr<std::string>;
+
 /// This class is a shared cache, that caches the 'stat' and 'open' calls to the
 /// underlying real file system, and the scanned preprocessor directives of
 /// files.
@@ -154,9 +156,11 @@ class DependencyScanningFilesystemSharedCache {
     /// The mutex that needs to be locked before mutation of any member.
     mutable std::mutex CacheLock;
 
-    /// Map from filenames to cached entries.
-    llvm::StringMap<const CachedFileSystemEntry *, llvm::BumpPtrAllocator>
-        EntriesByFilename;
+    /// Map from filenames to cached entries and real paths.
+    llvm::StringMap<
+        std::pair<const CachedFileSystemEntry *, const CachedRealPath *>,
+        llvm::BumpPtrAllocator>
+        CacheByFilename;
 
     /// Map from unique IDs to cached entries.
     llvm::DenseMap<llvm::sys::fs::UniqueID, const CachedFileSystemEntry *>
@@ -168,6 +172,9 @@ class DependencyScanningFilesystemSharedCache {
     /// The backing storage for cached contents.
     llvm::SpecificBumpPtrAllocator<CachedFileContents> ContentsStorage;
 
+    /// The backing storage for cached real paths.
+    llvm::SpecificBumpPtrAllocator<CachedRealPath> RealPathStorage;
+
     /// Returns entry associated with the filename or nullptr if none is found.
     const CachedFileSystemEntry *findEntryByFilename(StringRef Filename) const;
 
@@ -194,6 +201,17 @@ class DependencyScanningFilesystemSharedCache {
     const CachedFileSystemEntry &
     getOrInsertEntryForFilename(StringRef Filename,
                                 const CachedFileSystemEntry &Entry);
+
+    /// Returns the real path associated with the filename or nullptr if none is
+    /// found.
+    const CachedRealPath *findRealPathByFilename(StringRef Filename) const;
+
+    /// Returns the real path associated with the filename if there is some.
+    /// Otherwise, constructs new one with the given one, associates it with the
+    /// filename and returns the result.
+    const CachedRealPath &
+    getOrEmplaceRealPathForFilename(StringRef Filename,
+                                    llvm::ErrorOr<StringRef> RealPath);
   };
 
   DependencyScanningFilesystemSharedCache();
@@ -210,14 +228,17 @@ class DependencyScanningFilesystemSharedCache {
 /// This class is a local cache, that caches the 'stat' and 'open' calls to the
 /// underlying real file system.
 class DependencyScanningFilesystemLocalCache {
-  llvm::StringMap<const CachedFileSystemEntry *, llvm::BumpPtrAllocator> Cache;
+  llvm::StringMap<
+      std::pair<const CachedFileSystemEntry *, const CachedRealPath *>,
+      llvm::BumpPtrAllocator>
+      Cache;
 
 public:
   /// Returns entry associated with the filename or nullptr if none is found.
   const CachedFileSystemEntry *findEntryByFilename(StringRef Filename) const {
     assert(llvm::sys::path::is_absolute_gnu(Filename));
     auto It = Cache.find(Filename);
-    return It == Cache.end() ? nullptr : It->getValue();
+    return It == Cache.end() ? nullptr : It->getValue().first;
   }
 
   /// Associates the given entry with the filename and returns the given entry
@@ -226,9 +247,40 @@ class DependencyScanningFilesystemLocalCache {
   insertEntryForFilename(StringRef Filename,
                          const CachedFileSystemEntry &Entry) {
     assert(llvm::sys::path::is_absolute_gnu(Filename));
-    const auto *InsertedEntry = Cache.insert({Filename, &Entry}).first->second;
-    assert(InsertedEntry == &Entry && "entry already present");
-    return *InsertedEntry;
+    auto [It, Inserted] = Cache.insert({Filename, {&Entry, nullptr}});
+    auto &[CachedEntry, CachedRealPath] = It->getValue();
+    if (!Inserted) {
+      // The file is already present in the local cache. If we got here, it only
+      // contains the real path. Let's make sure the entry is populated too.
+      assert((!CachedEntry && CachedRealPath) && "entry already present");
+      CachedEntry = &Entry;
+    }
+    return *CachedEntry;
+  }
+
+  /// Returns real path associated with the filename or nullptr if none is
+  /// found.
+  const CachedRealPath *findRealPathByFilename(StringRef Filename) const {
+    assert(llvm::sys::path::is_absolute_gnu(Filename));
+    auto It = Cache.find(Filename);
+    return It == Cache.end() ? nullptr : It->getValue().second;
+  }
+
+  /// Associates the given real path with the filename and returns the given
+  /// entry pointer (for convenience).
+  const CachedRealPath &
+  insertRealPathForFilename(StringRef Filename,
+                            const CachedRealPath &RealPath) {
+    assert(llvm::sys::path::is_absolute_gnu(Filename));
+    auto [It, Inserted] = Cache.insert({Filename, {nullptr, &RealPath}});
+    auto &[CachedEntry, CachedRealPath] = It->getValue();
+    if (!Inserted) {
+      // The file is already present in the local cache. If we got here, it only
+      // contains the entry. Let's make sure the real path is populated too.
+      assert((!CachedRealPath && CachedEntry) && "real path already present");
+      CachedRealPath = &RealPath;
+    }
+    return *CachedRealPath;
   }
 };
 
@@ -296,6 +348,9 @@ class DependencyScanningWorkerFilesystem
   llvm::ErrorOr<std::unique_ptr<llvm::vfs::File>>
   openFileForRead(const Twine &Path) override;
 
+  std::error_code getRealPath(const Twine &Path,
+                              SmallVectorImpl<char> &Output) override;
+
   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
 
   /// Returns entry for the given filename.
diff --git a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
index c66780d50fa18..84185c931b552 100644
--- a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
+++ b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
@@ -113,8 +113,8 @@ DependencyScanningFilesystemSharedCache::CacheShard::findEntryByFilename(
     StringRef Filename) const {
   assert(llvm::sys::path::is_absolute_gnu(Filename));
   std::lock_guard<std::mutex> LockGuard(CacheLock);
-  auto It = EntriesByFilename.find(Filename);
-  return It == EntriesByFilename.end() ? nullptr : It->getValue();
+  auto It = CacheByFilename.find(Filename);
+  return It == CacheByFilename.end() ? nullptr : It->getValue().first;
 }
 
 const CachedFileSystemEntry *
@@ -130,11 +130,16 @@ DependencyScanningFilesystemSharedCache::CacheShard::
     getOrEmplaceEntryForFilename(StringRef Filename,
                                  llvm::ErrorOr<llvm::vfs::Status> Stat) {
   std::lock_guard<std::mutex> LockGuard(CacheLock);
-  auto Insertion = EntriesByFilename.insert({Filename, nullptr});
-  if (Insertion.second)
-    Insertion.first->second =
+  auto [It, Inserted] = CacheByFilename.insert({Filename, {nullptr, nullptr}});
+  auto &[CachedEntry, CachedRealPath] = It->getValue();
+  if (!CachedEntry) {
+    // The entry is not present in the shared cache. Either the cache doesn't
+    // know about the file at all, or it only knows about its real path.
+    assert((Inserted || CachedRealPath) && "existing file with empty pair");
+    CachedEntry =
         new (EntryStorage.Allocate()) CachedFileSystemEntry(std::move(Stat));
-  return *Insertion.first->second;
+  }
+  return *CachedEntry;
 }
 
 const CachedFileSystemEntry &
@@ -142,16 +147,17 @@ DependencyScanningFilesystemSharedCache::CacheShard::getOrEmplaceEntryForUID(
     llvm::sys::fs::UniqueID UID, llvm::vfs::Status Stat,
     std::unique_ptr<llvm::MemoryBuffer> Contents) {
   std::lock_guard<std::mutex> LockGuard(CacheLock);
-  auto Insertion = EntriesByUID.insert({UID, nullptr});
-  if (Insertion.second) {
+  auto [It, Inserted] = EntriesByUID.insert({UID, nullptr});
+  auto &CachedEntry = It->getSecond();
+  if (Inserted) {
     CachedFileContents *StoredContents = nullptr;
     if (Contents)
       StoredContents = new (ContentsStorage.Allocate())
           CachedFileContents(std::move(Contents));
-    Insertion.first->second = new (EntryStorage.Allocate())
+    CachedEntry = new (EntryStorage.Allocate())
         CachedFileSystemEntry(std::move(Stat), StoredContents);
   }
-  return *Insertion.first->second;
+  return *CachedEntry;
 }
 
 const CachedFileSystemEntry &
@@ -159,7 +165,40 @@ DependencyScanningFilesystemSharedCache::CacheShard::
     getOrInsertEntryForFilename(StringRef Filename,
                                 const CachedFileSystemEntry &Entry) {
   std::lock_guard<std::mutex> LockGuard(CacheLock);
-  return *EntriesByFilename.insert({Filename, &Entry}).first->getValue();
+  auto [It, Inserted] = CacheByFilename.insert({Filename, {&Entry, nullptr}});
+  auto &[CachedEntry, CachedRealPath] = It->getValue();
+  if (!Inserted || !CachedEntry)
+    CachedEntry = &Entry;
+  return *CachedEntry;
+}
+
+const CachedRealPath *
+DependencyScanningFilesystemSharedCache::CacheShard::findRealPathByFilename(
+    StringRef Filename) const {
+  assert(llvm::sys::path::is_absolute_gnu(Filename));
+  std::lock_guard<std::mutex> LockGuard(CacheLock);
+  auto It = CacheByFilename.find(Filename);
+  return It == CacheByFilename.end() ? nullptr : It->getValue().second;
+}
+
+const CachedRealPath &DependencyScanningFilesystemSharedCache::CacheShard::
+    getOrEmplaceRealPathForFilename(StringRef Filename,
+                                    llvm::ErrorOr<llvm::StringRef> RealPath) {
+  std::lock_guard<std::mutex> LockGuard(CacheLock);
+
+  const CachedRealPath *&StoredRealPath = CacheByFilename[Filename].second;
+  if (!StoredRealPath) {
+    auto OwnedRealPath = [&]() -> CachedRealPath {
+      if (!RealPath)
+        return RealPath.getError();
+      return RealPath->str();
+    }();
+
+    StoredRealPath = new (RealPathStorage.Allocate())
+        CachedRealPath(std::move(OwnedRealPath));
+  }
+
+  return *StoredRealPath;
 }
 
 static bool shouldCacheStatFailures(StringRef Filename) {
@@ -321,6 +360,54 @@ DependencyScanningWorkerFilesystem::openFileForRead(const Twine &Path) {
   return DepScanFile::create(Result.get());
 }
 
+std::error_code
+DependencyScanningWorkerFilesystem::getRealPath(const Twine &Path,
+                                                SmallVectorImpl<char> &Output) {
+  SmallString<256> OwnedFilename;
+  StringRef OriginalFilename = Path.toStringRef(OwnedFilename);
+
+  SmallString<256> PathBuf;
+  auto FilenameForLookup = tryGetFilenameForLookup(OriginalFilename, PathBuf);
+  if (!FilenameForLookup)
+    return FilenameForLookup.getError();
+
+  auto HandleCachedRealPath =
+      [&Output](const CachedRealPath &RealPath) -> std::error_code {
+    if (!RealPath)
+      return RealPath.getError();
+    Output.assign(RealPath->begin(), RealPath->end());
+    return {};
+  };
+
+  // If we already have the result in local cache, no work required.
+  if (const auto *RealPath =
+          LocalCache.findRealPathByFilename(*FilenameForLookup))
+    return HandleCachedRealPath(*RealPath);
+
+  // If we have the result in the shared cache, cache it locally.
+  auto &Shard = SharedCache.getShardForFilename(*FilenameForLookup);
+  if (const auto *ShardRealPath =
+          Shard.findRealPathByFilename(*FilenameForLookup)) {
+    const auto &RealPath = LocalCache.insertRealPathForFilename(
+        *FilenameForLookup, *ShardRealPath);
+    return HandleCachedRealPath(RealPath);
+  }
+
+  // If we don't know the real path, compute it...
+  std::error_code EC = getUnderlyingFS().getRealPath(OriginalFilename, Output);
+  llvm::ErrorOr<llvm::StringRef> ComputedRealPath = EC;
+  if (!EC)
+    ComputedRealPath = StringRef{Output.data(), Output.size()};
+
+  // ...and try to write it into the shared cache. In case some other thread won
+  // this race and already wrote its own result there, just adopt it. Write
+  // whatever is in the shared cache into the local one.
+  const auto &RealPath = Shard.getOrEmplaceRealPathForFilename(
+      *FilenameForLookup, ComputedRealPath);
+  return HandleCachedRealPath(
+      LocalCache.insertRealPathForFilename(*FilenameForLookup, RealPath));
+}
+
 std::error_code DependencyScanningWorkerFilesystem::setCurrentWorkingDirectory(
     const Twine &Path) {
   std::error_code EC = ProxyFileSystem::setCurrentWorkingDirectory(Path);
diff --git a/clang/unittests/Tooling/CMakeLists.txt b/clang/unittests/Tooling/CMakeLists.txt
index 2ff493ef5fc32..0eb612f8d9498 100644
--- a/clang/unittests/Tooling/CMakeLists.txt
+++ b/clang/unittests/Tooling/CMakeLists.txt
@@ -24,6 +24,7 @@ add_clang_unittest(ToolingTests
   QualTypeNamesTest.cpp
   RangeSelectorTest.cpp
   DependencyScanning/DependencyScannerTest.cpp
+  DependencyScanning/DependencyScanningFilesystemTest.cpp
   RecursiveASTVisitorTests/Attr.cpp
   RecursiveASTVisitorTests/BitfieldInitializer.cpp
   RecursiveASTVisitorTests/CallbacksLeaf.cpp
diff --git a/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
new file mode 100644
index 0000000000000..50cad72d223e3
--- /dev/null
+++ b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
@@ -0,0 +1,152 @@
+//===- DependencyScanningFilesystemTest.cpp -------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/Support/VirtualFileSystem.h"
+#include "gtest/gtest.h"
+
+using namespace clang::tooling::dependencies;
+
+namespace {
+struct InstrumentingFilesystem
+    : llvm::RTTIExtends<InstrumentingFilesystem, llvm::vfs::ProxyFileSystem> {
+  unsigned NumStatusCalls = 0;
+  unsigned NumGetRealPathCalls = 0;
+
+  using llvm::RTTIExtends<InstrumentingFilesystem,
+                          llvm::vfs::ProxyFileSystem>::RTTIExtends;
+
+  llvm::ErrorOr<llvm::vfs::Status> status(const llvm::Twine &Path) override {
+    ++NumStatusCalls;
+    return ProxyFileSystem::status(Path);
+  }
+
+  std::error_code getRealPath(const llvm::Twine &Path,
+                              llvm::SmallVectorImpl<char> &Output) override {
+    ++NumGetRealPathCalls;
+    return ProxyFileSystem::getRealPath(Path, Output);
+  }
+};
+} // namespace
+
+TEST(DependencyScanningWorkerFilesystem, CacheStatusFailures) {
+  auto InMemoryFS = llvm::makeIntrusiveRefCnt<llvm::vfs::InMemoryFileSystem>();
+
+  auto InstrumentingFS =
+      llvm::makeIntrusiveRefCnt<InstrumentingFilesystem>(InMemoryFS);
+
+  DependencyScanningFilesystemSharedCache SharedCache;
+  DependencyScanningWorkerFilesystem DepFS(SharedCache, InstrumentingFS);
+  DependencyScanningWorkerFilesystem DepFS2(SharedCache, InstrumentingFS);
+
+  DepFS.status("/foo.c");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 1u);
+
+  DepFS.status("/foo.c");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 1u); // Cached, no increase.
+
+  DepFS.status("/bar.c");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 2u);
+
+  DepFS2.status("/foo.c");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 2u); // Shared cache.
+}
+
+TEST(DependencyScanningFilesystem, CacheGetRealPath) {
+  auto InMemoryFS = llvm::makeIntrusiveRefCnt<llvm::vfs::InMemoryFileSystem>();
+  InMemoryFS->setCurrentWorkingDirectory("/");
+  InMemoryFS->addFile("/foo", 0, llvm::MemoryBuffer::getMemBuffer(""));
+  InMemoryFS->addFile("/bar", 0, llvm::MemoryBuffer::getMemBuffer(""));
+
+  auto InstrumentingFS =
+      llvm::makeIntrusiveRefCnt<InstrumentingFilesystem>(InMemoryFS);
+
+  DependencyScanningFilesystemSharedCache SharedCache;
+  DependencyScanningWorkerFilesystem DepFS(SharedCache, InstrumentingFS);
+  DependencyScanningWorkerFilesystem DepFS2(SharedCache, InstrumentingFS);
+
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/foo", Result);
+    EXPECT_EQ(InstrumentingFS->NumGetRealPathCalls, 1u);
+  }
+
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/foo", Result);
+    EXPECT_EQ(InstrumentingFS->NumGetRealPathCalls, 1u); // Cached, no increase.
+  }
+
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/bar", Result);
+    EXPECT_EQ(InstrumentingFS->NumGetRealPathCalls, 2u);
+  }
+
+  {
+    llvm::SmallString<128> Result;
+    DepFS2.getRealPath("/foo", Result);
+    EXPECT_EQ(InstrumentingFS->NumGetRealPathCalls, 2u); // Shared cache.
+  }
+}
+
+TEST(DependencyScanningFilesystem, RealPathAndStatusInvariants) {
+  auto InMemoryFS = llvm::makeIntrusiveRefCnt<llvm::vfs::InMemoryFileSystem>();
+  InMemoryFS->setCurrentWorkingDirectory("/");
+  InMemoryFS->addFile("/foo.c", 0, llvm::MemoryBuffer::getMemBuffer(""));
+  InMemoryFS->addFile("/bar.c", 0, llvm::MemoryBuffer::getMemBuffer(""));
+
+  auto InstrumentingFS =
+      llvm::makeIntrusiveRefCnt<InstrumentingFilesystem>(InMemoryFS);
+
+  DependencyScanningFilesystemSharedCache SharedCache;
+  DependencyScanningWorkerFilesystem DepFS(SharedCache, InstrumentingFS);
+
+  // Success.
+  {
+    DepFS.status("/foo.c");
+
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/foo.c", Result);
+  }
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/bar.c", Result);
+
+    DepFS.status("/bar.c");
+  }
+
+  // Failure.
+  {
+    DepFS.status("/foo.m");
+
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/foo.m", Result);
+  }
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/bar.m", Result);
+
+    DepFS.status("/bar.m");
+  }
+
+  // Failure without caching.
+  {
+    DepFS.status("/foo");
+
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/foo", Result);
+  }
+  {
+    llvm::SmallString<128> Result;
+    DepFS.getRealPath("/bar", Result);
+
+    DepFS.status("/bar");
+  }
+}

From 2620ccc3cdebcaa53b96b5f4d79ffe3951534e07 Mon Sep 17 00:00:00 2001
From: Kalesh Singh <kaleshsingh96@gmail.com>
Date: Fri, 12 Apr 2024 10:43:59 -0700
Subject: [PATCH 282/886] ANDROID: x86_64: Set default max-page-size to 16kB
 (#87413)

Android now supports both 4kB and 16kB page sizes. The vast majority of
android apps are developed on x86_64 machines. In order to provide
emulators that support larger page sizes, Android emulates the page-size
in x86_64 to support testing apps for large page size support.

For this reason, update Android x86_64 ELFs default max-page-size to
16384 to support both 4kB and 16kB page-size devices.

Increase max-page-size raises concerns of increased disk space and extra
VMA slab memory.

In Android, RO partitions use sparse images, so that the holes on ELFs
don't allocate blocks on disk; and PackageManager ensures to punch holes
in ELF-paddings on the /data partition when apps are installed.

Extra VMA slab memory is addressed by the bionic loader, which extends
segment VMAs to cover the gaps between consecutive segment mappings, to
avoid the extra VMAs needed for the gap PROT_NONE mappings (---p). This
optimization is done in the crt_pad_segment note [1] is present in the
ELF.

[1]
https://cs.android.com/android/platform/superproject/main/+/189e480390ef13199d59e1fb54078e8b78ea6f79:bionic/libc/arch-common/bionic/crt_pad_segment.S

---------

Signed-off-by: Kalesh Singh <kaleshsingh@google.com>
Co-authored-by: Kalesh Singh <kaleshsingh@google.com>
---
 clang/lib/Driver/ToolChains/Linux.cpp |  3 ++-
 clang/test/Driver/android-link.cpp    | 38 +++++++++++++++++----------
 2 files changed, 26 insertions(+), 15 deletions(-)

diff --git a/clang/lib/Driver/ToolChains/Linux.cpp b/clang/lib/Driver/ToolChains/Linux.cpp
index 6c2f23e57bce0..fb65881061eff 100644
--- a/clang/lib/Driver/ToolChains/Linux.cpp
+++ b/clang/lib/Driver/ToolChains/Linux.cpp
@@ -244,8 +244,9 @@ Linux::Linux(const Driver &D, const llvm::Triple &Triple, const ArgList &Args)
       // Android ARM uses max-page-size=4096 to reduce VMA usage.
       ExtraOpts.push_back("-z");
       ExtraOpts.push_back("max-page-size=4096");
-    } else if (Triple.isAArch64()) {
+    } else if (Triple.isAArch64() || Triple.getArch() == llvm::Triple::x86_64) {
       // Android AArch64 uses max-page-size=16384 to support 4k/16k page sizes.
+      // Android emulates a 16k page size for app testing on x86_64 machines.
       ExtraOpts.push_back("-z");
       ExtraOpts.push_back("max-page-size=16384");
     }
diff --git a/clang/test/Driver/android-link.cpp b/clang/test/Driver/android-link.cpp
index f9bdd00507d7b..ab7dae5405587 100644
--- a/clang/test/Driver/android-link.cpp
+++ b/clang/test/Driver/android-link.cpp
@@ -1,26 +1,36 @@
-// Check that we add relevant linker flags for Android ARM/AArch64.
+// Check that we add relevant linker flags for Android ARM/AArch64/i386/x86_64.
 
-// RUN: %clang -### -target arm-linux-androideabi %s 2>&1 | \
-// RUN:   FileCheck --check-prefix=MAX-PAGE-SIZE %s
+// RUN: %clang -### --target=arm-linux-androideabi %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=MAX-PAGE-SIZE-4KB %s
 
-// RUN: %clang -target aarch64-none-linux-android \
+// RUN: %clang --target=aarch64-none-linux-android \
 // RUN:   -### -v %s 2> %t
 // RUN: FileCheck -check-prefix=GENERIC-ARM < %t %s
-//
-// RUN: %clang -target aarch64-none-linux-android \
+
+// RUN: %clang --target=aarch64-none-linux-android \
 // RUN:   -mcpu=cortex-a53 -### -v %s 2> %t
 // RUN: FileCheck -check-prefix=CORTEX-A53 < %t %s
-//
-// RUN: %clang -target aarch64-none-linux-android \
+
+// RUN: %clang --target=aarch64-none-linux-android \
 // RUN:   -mcpu=cortex-a57 -### -v %s 2> %t
 // RUN: FileCheck -check-prefix=CORTEX-A57 < %t %s
-//
-// RUN: %clang -target aarch64-none-linux-android \
+
+// RUN: %clang --target=aarch64-none-linux-android \
 // RUN:   -### -v %s 2> %t
-// RUN: FileCheck -check-prefix=MAX-PAGE-SIZE-AARCH64 < %t %s
-//
+// RUN: FileCheck -check-prefix=MAX-PAGE-SIZE-16KB < %t %s
+
+// RUN: %clang -### --target=i386-none-linux-android %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=NO-MAX-PAGE-SIZE-16KB %s
+
+// RUN: %clang -### --target=x86_64-none-linux-gnu %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=NO-MAX-PAGE-SIZE-16KB %s
+
+// RUN: %clang -### --target=x86_64-none-linux-android %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=MAX-PAGE-SIZE-16KB %s
+
 // GENERIC-ARM: --fix-cortex-a53-843419
 // CORTEX-A53: --fix-cortex-a53-843419
 // CORTEX-A57-NOT: --fix-cortex-a53-843419
-// MAX-PAGE-SIZE: "-z" "max-page-size=4096"
-// MAX-PAGE-SIZE-AARCH64: "-z" "max-page-size=16384"
+// MAX-PAGE-SIZE-4KB: "-z" "max-page-size=4096"
+// MAX-PAGE-SIZE-16KB: "-z" "max-page-size=16384"
+// NO-MAX-PAGE-SIZE-16KB-NOT: "-z" "max-page-size=16384"

From 446d38c65f72fd5d242a64182b05683577f683d3 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 10:48:54 -0700
Subject: [PATCH 283/886] [lldb] Fix test build failure

Caused by commit edd7fed9da48c0e708cce9bd4d305ae43d8bd77c
---
 lldb/unittests/Host/FileSystemTest.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lldb/unittests/Host/FileSystemTest.cpp b/lldb/unittests/Host/FileSystemTest.cpp
index 4ed9beff4d303..3b5ee7c8bc223 100644
--- a/lldb/unittests/Host/FileSystemTest.cpp
+++ b/lldb/unittests/Host/FileSystemTest.cpp
@@ -74,7 +74,7 @@ class DummyFileSystem : public vfs::FileSystem {
   }
   // Map any symlink to "/symlink".
   std::error_code getRealPath(const Twine &Path,
-                              SmallVectorImpl<char> &Output) const override {
+                              SmallVectorImpl<char> &Output) override {
     auto I = FilesAndDirs.find(Path.str());
     if (I == FilesAndDirs.end())
       return make_error_code(llvm::errc::no_such_file_or_directory);

From 9022d9c102a8c47616fbeb6a45b015db6bd2c17a Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Fri, 12 Apr 2024 10:49:39 -0700
Subject: [PATCH 284/886] [xray] Apply default visibility to
 __xray_CustomEvent/__xray_TypedEvent

`__xray_customevent` and `__xray_typedevent` are built-in functions in
Clang. With -fxray-instrument, they are lowered to `__xray_CustomEvent`
(with 2 arguments) or `__xray_TypedEvent` (with 3 arguments).

xray patching is supported for shared objects, but they may contain
`__xray_customevent` and `__xray_typedevent` references that need to be
satisfied by default visibility definitions exported by the executable.

lld since df54f627fad789ccb11c72a9fddf116decbeba0e, like GNU ld, catches
the scenario at link time.
---
 compiler-rt/lib/xray/xray_trampoline_AArch64.S | 4 ++--
 compiler-rt/lib/xray/xray_trampoline_x86_64.S  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/compiler-rt/lib/xray/xray_trampoline_AArch64.S b/compiler-rt/lib/xray/xray_trampoline_AArch64.S
index 6f10dda3602b5..536a79e0d150e 100644
--- a/compiler-rt/lib/xray/xray_trampoline_AArch64.S
+++ b/compiler-rt/lib/xray/xray_trampoline_AArch64.S
@@ -138,8 +138,9 @@ ASM_SYMBOL(__xray_ArgLoggerEntry):
   ret
 ASM_SIZE(__xray_ArgLoggerEntry)
 
+// __xray_*Event have default visibility so that they can be referenced by user
+// DSOs that do not link against the runtime.
 .global ASM_SYMBOL(__xray_CustomEvent)
-ASM_HIDDEN(__xray_CustomEvent)
 ASM_TYPE_FUNCTION(__xray_CustomEvent)
 ASM_SYMBOL(__xray_CustomEvent):
   SAVE_REGISTERS
@@ -153,7 +154,6 @@ ASM_SYMBOL(__xray_CustomEvent):
 ASM_SIZE(__xray_CustomEvent)
 
 .global ASM_SYMBOL(__xray_TypedEvent)
-ASM_HIDDEN(__xray_TypedEvent)
 ASM_TYPE_FUNCTION(__xray_TypedEvent)
 ASM_SYMBOL(__xray_TypedEvent):
   SAVE_REGISTERS
diff --git a/compiler-rt/lib/xray/xray_trampoline_x86_64.S b/compiler-rt/lib/xray/xray_trampoline_x86_64.S
index 0f00bcc41508d..ff3ac91071a60 100644
--- a/compiler-rt/lib/xray/xray_trampoline_x86_64.S
+++ b/compiler-rt/lib/xray/xray_trampoline_x86_64.S
@@ -257,8 +257,9 @@ LOCAL_LABEL(arg1entryFail):
 
 //===----------------------------------------------------------------------===//
 
+// __xray_*Event have default visibility so that they can be referenced by user
+// DSOs that do not link against the runtime.
 	.global ASM_SYMBOL(__xray_CustomEvent)
-	ASM_HIDDEN(__xray_CustomEvent)
 	.align 16, 0x90
 	ASM_TYPE_FUNCTION(__xray_CustomEvent)
 # LLVM-MCA-BEGIN __xray_CustomEvent
@@ -284,7 +285,6 @@ LOCAL_LABEL(customEventCleanup):
 //===----------------------------------------------------------------------===//
 
 	.global ASM_SYMBOL(__xray_TypedEvent)
-	ASM_HIDDEN(__xray_TypedEvent)
 	.align 16, 0x90
 	ASM_TYPE_FUNCTION(__xray_TypedEvent)
 # LLVM-MCA-BEGIN __xray_TypedEvent

From 4b0beb4f5ec42aea58461df7994e2fa40f335bb6 Mon Sep 17 00:00:00 2001
From: jimingham <jingham@apple.com>
Date: Fri, 12 Apr 2024 10:51:49 -0700
Subject: [PATCH 285/886] =?UTF-8?q?Reapply=20"Fix=20error=20in=20unrecogni?=
 =?UTF-8?q?zed=20register=20name=20handling=20for=20"SBFram=E2=80=A6e.regi?=
 =?UTF-8?q?ster""=20(#88468)"=20(#88535)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The only change is a fix for the "register" iterator test to not rely on
particular register names.

I mistook where the artificial "pc" register is generated. It isn't
added to the register list or the register sets (except on arm where
that's the name of the actual register), so I can't use it in this test.
I instead just assert that the "register" generator produces the same
list as flattening the register sets from "registers".

This reverts commit 9f14914753599f3879e4c273191959e2f1b3632c.
---
 lldb/bindings/interface/SBFrameExtensions.i  | 12 +++++++++++-
 lldb/test/API/python_api/frame/TestFrames.py | 16 ++++++++++++++--
 2 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/lldb/bindings/interface/SBFrameExtensions.i b/lldb/bindings/interface/SBFrameExtensions.i
index 43b22ed7a6b32..e0472280666ab 100644
--- a/lldb/bindings/interface/SBFrameExtensions.i
+++ b/lldb/bindings/interface/SBFrameExtensions.i
@@ -44,6 +44,16 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                 def __init__(self, regs):
                     self.regs = regs
 
+                def __iter__(self):
+                    return self.get_registers()
+
+                def get_registers(self):
+                    for i in range(0,len(self.regs)):
+                        rs = self.regs[i]
+                        for j in range (0,rs.num_children):
+                            reg = rs.GetChildAtIndex(j)
+                            yield reg
+                          
                 def __getitem__(self, key):
                     if type(key) is str:
                         for i in range(0,len(self.regs)):
@@ -52,7 +62,7 @@ STRING_EXTENSION_OUTSIDE(SBFrame)
                                 reg = rs.GetChildAtIndex(j)
                                 if reg.name == key: return reg
                     else:
-                        return lldb.SBValue()
+                        return SBValue()
 
             return registers_access(self.registers)
 
diff --git a/lldb/test/API/python_api/frame/TestFrames.py b/lldb/test/API/python_api/frame/TestFrames.py
index a82b129bc8099..0ec12206024ab 100644
--- a/lldb/test/API/python_api/frame/TestFrames.py
+++ b/lldb/test/API/python_api/frame/TestFrames.py
@@ -73,10 +73,12 @@ def test_get_arg_vals_for_call_stack(self):
                 gpr_reg_set = lldbutil.get_GPRs(frame)
                 pc_value = gpr_reg_set.GetChildMemberWithName("pc")
                 self.assertTrue(pc_value, "We should have a valid PC.")
-                pc_value_int = int(pc_value.GetValue(), 0)
+                               
+    
                 # Make sure on arm targets we dont mismatch PC value on the basis of thumb bit.
                 # Frame PC will not have thumb bit set in case of a thumb
                 # instruction as PC.
+                pc_value_int = int(pc_value.GetValue(), 0)
                 if self.getArchitecture() in ["arm", "armv7", "armv7k"]:
                     pc_value_int &= ~1
                 self.assertEqual(
@@ -91,7 +93,17 @@ def test_get_arg_vals_for_call_stack(self):
                     frame.GetSP(),
                     "SP gotten as a value should equal frame's GetSP",
                 )
-
+                # Test that the "register" property's flat list matches the list from
+                # the "registers" property that returns register sets:
+                register_regs = set()
+                flattened_regs = set()
+                for reg_set in frame.registers:
+                    for reg in reg_set:
+                        flattened_regs.add(reg.name)
+                for reg in frame.register:
+                    register_regs.add(reg.name)
+                self.assertEqual(register_regs, flattened_regs, "register matches registers")
+                        
             print("---", file=session)
             process.Continue()
 

From 83dc41992dc602070f5429e2717352f60aad931c Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Fri, 12 Apr 2024 11:14:20 -0700
Subject: [PATCH 286/886] [memprof] Clean up writer traits (NFC) (#88549)

RecordWriter does not live past the end of writeMemProfRecords, so it
can be safely on stack.

The constructor of FrameWriter does not take any parameter, so we can
let OnDiskChainedHashTableGenerator::Emit (with a single parameter)
default-construct an instance of the writer trait inside Emit.
---
 llvm/lib/ProfileData/InstrProfWriter.cpp | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/llvm/lib/ProfileData/InstrProfWriter.cpp b/llvm/lib/ProfileData/InstrProfWriter.cpp
index 7c56cde3e6ced..ede042dd6b9c3 100644
--- a/llvm/lib/ProfileData/InstrProfWriter.cpp
+++ b/llvm/lib/ProfileData/InstrProfWriter.cpp
@@ -428,14 +428,13 @@ static uint64_t writeMemProfRecords(
     llvm::MapVector<GlobalValue::GUID, memprof::IndexedMemProfRecord>
         &MemProfRecordData,
     memprof::MemProfSchema *Schema) {
-  auto RecordWriter =
-      std::make_unique<memprof::RecordWriterTrait>(memprof::Version1);
-  RecordWriter->Schema = Schema;
+  memprof::RecordWriterTrait RecordWriter(memprof::Version1);
+  RecordWriter.Schema = Schema;
   OnDiskChainedHashTableGenerator<memprof::RecordWriterTrait>
       RecordTableGenerator;
   for (auto &[GUID, Record] : MemProfRecordData) {
     // Insert the key (func hash) and value (memprof record).
-    RecordTableGenerator.insert(GUID, Record, *RecordWriter.get());
+    RecordTableGenerator.insert(GUID, Record, RecordWriter);
   }
   // Release the memory of this MapVector as it is no longer needed.
   MemProfRecordData.clear();
@@ -443,14 +442,13 @@ static uint64_t writeMemProfRecords(
   // The call to Emit invokes RecordWriterTrait::EmitData which destructs
   // the memprof record copies owned by the RecordTableGenerator. This works
   // because the RecordTableGenerator is not used after this point.
-  return RecordTableGenerator.Emit(OS.OS, *RecordWriter);
+  return RecordTableGenerator.Emit(OS.OS, RecordWriter);
 }
 
 // Serialize MemProfFrameData.  Return FrameTableOffset.
 static uint64_t writeMemProfFrames(
     ProfOStream &OS,
     llvm::MapVector<memprof::FrameId, memprof::Frame> &MemProfFrameData) {
-  auto FrameWriter = std::make_unique<memprof::FrameWriterTrait>();
   OnDiskChainedHashTableGenerator<memprof::FrameWriterTrait>
       FrameTableGenerator;
   for (auto &[FrameId, Frame] : MemProfFrameData) {
@@ -460,7 +458,7 @@ static uint64_t writeMemProfFrames(
   // Release the memory of this MapVector as it is no longer needed.
   MemProfFrameData.clear();
 
-  return FrameTableGenerator.Emit(OS.OS, *FrameWriter);
+  return FrameTableGenerator.Emit(OS.OS);
 }
 
 static Error writeMemProfV0(

From 0318ce8552896df7b180ee5481463eada65c755e Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Fri, 12 Apr 2024 14:15:22 -0400
Subject: [PATCH 287/886] [C99] Claim conformance for _Complex support (#88161)

There's so much overlap between the cited papers so this condenses the
status page into a single entry rather than trying to test conformance
against multiple papers doing conflicting things.
---
 clang/test/C/C99/n809.c   | 122 ++++++++++++++++++++++++++++++++++++++
 clang/test/C/C99/n809_2.c |  64 ++++++++++++++++++++
 clang/test/C/C99/n809_3.c |  12 ++++
 clang/www/c_status.html   |  37 +++++-------
 4 files changed, 213 insertions(+), 22 deletions(-)
 create mode 100644 clang/test/C/C99/n809.c
 create mode 100644 clang/test/C/C99/n809_2.c
 create mode 100644 clang/test/C/C99/n809_3.c

diff --git a/clang/test/C/C99/n809.c b/clang/test/C/C99/n809.c
new file mode 100644
index 0000000000000..7297443a777cc
--- /dev/null
+++ b/clang/test/C/C99/n809.c
@@ -0,0 +1,122 @@
+// RUN: %clang_cc1 -verify -std=c99 %s
+
+/* WG14 N620, N638, N657, N694, N809: Partial
+ * Complex and imaginary support in <complex.h>
+ *
+ * NB: Clang supports _Complex but not _Imaginary. In C99, _Complex support is
+ * required outside of freestanding, but _Imaginary support is fully optional.
+ * In C11, both are made fully optional.
+ *
+ * NB: _Complex support requires an underlying support library such as
+ * compiler-rt to provide functions like __divsc3. Compiler-rt is not supported
+ * on Windows.
+ *
+ * Because the functionality is so intertwined between the various papers,
+ * we're testing all of the functionality in one file.
+ */
+
+// Demonstrate that we support spelling complex floating-point objects.
+float _Complex f1;
+_Complex float f2;
+
+double _Complex d1;
+_Complex double d2;
+
+long double _Complex ld1;
+_Complex long double ld2;
+
+// Show that we don't support spelling imaginary types.
+float _Imaginary fi1; // expected-error {{imaginary types are not supported}}
+_Imaginary float fi2; // expected-error {{imaginary types are not supported}}
+
+double _Imaginary di1; // expected-error {{imaginary types are not supported}}
+_Imaginary double di2; // expected-error {{imaginary types are not supported}}
+
+long double _Imaginary ldi1; // expected-error {{imaginary types are not supported}}
+_Imaginary long double ldi2; // expected-error {{imaginary types are not supported}}
+
+// Each complex type has the same representation and alignment as an array
+// containing two elements of the corresponding real type. Note, it is not
+// mandatory that the alignment of a structure containing an array of two
+// elements has the same alignment as an array of two elements outside of a
+// structure, but this is a property Clang supports.
+_Static_assert(sizeof(float _Complex) == sizeof(struct { float mem[2]; }), "");
+_Static_assert(_Alignof(float _Complex) == _Alignof(struct { float mem[2]; }), "");
+
+_Static_assert(sizeof(double _Complex) == sizeof(struct { double mem[2]; }), "");
+_Static_assert(_Alignof(double _Complex) == _Alignof(struct { double mem[2]; }), "");
+
+_Static_assert(sizeof(long double _Complex) == sizeof(struct { long double mem[2]; }), "");
+_Static_assert(_Alignof(long double _Complex) == _Alignof(struct { long double mem[2]; }), "");
+
+// The first element corresponds to the real part and the second element
+// corresponds to the imaginary part.
+_Static_assert(__real((float _Complex){ 1.0f, 2.0f }) == 1.0f, "");
+_Static_assert(__imag((float _Complex){ 1.0f, 2.0f }) == 2.0f, "");
+
+_Static_assert(__real((double _Complex){ 1.0, 2.0 }) == 1.0, "");
+_Static_assert(__imag((double _Complex){ 1.0, 2.0 }) == 2.0, "");
+
+_Static_assert(__real((long double _Complex){ 1.0L, 2.0L }) == 1.0L, "");
+_Static_assert(__imag((long double _Complex){ 1.0L, 2.0L }) == 2.0L, "");
+
+// When a real value is converted to a complex value, the real part follows the
+// usual conversion rules and the imaginary part should be zero.
+_Static_assert(__real((float _Complex)1.0f) == 1.0f, "");
+_Static_assert(__imag((float _Complex)1.0f) == 0.0f, "");
+
+_Static_assert(__real((double _Complex)1.0f) == 1.0, "");
+_Static_assert(__imag((double _Complex)1.0f) == 0.0, "");
+
+_Static_assert(__real((long double _Complex)1.0f) == 1.0L, "");
+_Static_assert(__imag((long double _Complex)1.0f) == 0.0L, "");
+
+// When a complex value is converted to a real value, the real part follows the
+// usual conversion rules and the imaginary part is discarded.
+_Static_assert((float)(float _Complex){ 1.0f, 2.0f } == 1.0f, "");
+_Static_assert((double)(float _Complex){ 1.0f, 2.0f } == 1.0, "");
+_Static_assert((long double)(float _Complex){ 1.0f, 2.0f } == 1.0L, "");
+
+// Complex values are only equal if both the real and imaginary parts are equal.
+_Static_assert((float _Complex){ 1.0f, 2.0f } == (float _Complex){ 1.0f, 2.0f }, "");
+_Static_assert((double _Complex){ 1.0, 2.0 } == (double _Complex){ 1.0, 2.0 }, "");
+_Static_assert((long double _Complex){ 1.0L, 2.0L } == (long double _Complex){ 1.0L, 2.0L }, "");
+
+_Static_assert((float _Complex){ 1.0f, 2.0f } != (float _Complex){ 2.0f, 0.0f }, "");
+_Static_assert((double _Complex){ 1.0, 2.0 } != (double _Complex){ 2.0, 0.0 }, "");
+_Static_assert((long double _Complex){ 1.0L, 2.0L } != (long double _Complex){ 2.0L, 0.0L }, "");
+
+// You cannot use relational operator on complex values.
+int i1 = (float _Complex){ 1.0f, 2.0f } < 10;        // expected-error {{invalid operands to binary expression}}
+int i2 = (double _Complex){ 1.0f, 2.0f } > 10;       // expected-error {{invalid operands to binary expression}}
+int i3 = (long double _Complex){ 1.0f, 2.0f } <= 10; // expected-error {{invalid operands to binary expression}}
+int i4 = (float _Complex){ 1.0f, 2.0f } >= 10;       // expected-error {{invalid operands to binary expression}}
+
+// As a type specifier, _Complex cannot appear alone; however, we support it as
+// an extension by assuming _Complex double.
+_Complex c = 1.0f; // expected-warning {{plain '_Complex' requires a type specifier; assuming '_Complex double'}}
+// Because we don't support imaginary types, we don't extend the extension to
+// that type specifier.
+// FIXME: the warning diagnostic here is incorrect and should not be emitted.
+_Imaginary i = 1.0f; // expected-warning {{plain '_Complex' requires a type specifier; assuming '_Complex double'}} \
+                        expected-error {{imaginary types are not supported}}
+
+void func(void) {
+#pragma clang diagnostic push
+#pragma clang diagnostic warning "-Wpedantic"
+  // Increment and decrement operators have a constraint that their operand be
+  // a real type; Clang supports this as an extension on complex types as well.
+  _Complex float cf = 0.0f;
+
+  cf++; // expected-warning {{'++' on an object of complex type is a Clang extension}}
+  ++cf; // expected-warning {{'++' on an object of complex type is a Clang extension}}
+
+  cf--; // expected-warning {{'--' on an object of complex type is a Clang extension}}
+  --cf; // expected-warning {{'--' on an object of complex type is a Clang extension}}
+
+  // However, unary + and - are fine, as is += 1.
+  (void)-cf;
+  (void)+cf;
+  cf += 1;
+#pragma clang diagnostic pop
+}
diff --git a/clang/test/C/C99/n809_2.c b/clang/test/C/C99/n809_2.c
new file mode 100644
index 0000000000000..3bf163126521a
--- /dev/null
+++ b/clang/test/C/C99/n809_2.c
@@ -0,0 +1,64 @@
+// RUN: %clang_cc1 -ast-dump -std=c99 %s | FileCheck %s
+
+void variadic(int i, ...);
+
+void func(void) {
+  // CHECK: FunctionDecl {{.*}} func 'void (void)'
+
+  // Show that we correctly convert between two complex domains.
+  _Complex float cf = 1.0f;
+  _Complex double cd;
+
+  cd = cf;
+  // CHECK: BinaryOperator {{.*}} '_Complex double' '='
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cd'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex double' <FloatingComplexCast>
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cf'
+
+  cf = cd;
+  // CHECK: BinaryOperator {{.*}} '_Complex float' '='
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cf'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex float' <FloatingComplexCast>
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex double' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cd'
+
+  // Show that we correctly convert to the common type of a complex and real.
+  // This should convert the _Complex float to a _Complex double ("without
+  // change of domain" c.f. C99 6.3.1.8p1).
+  (void)(cf + 1.0);
+  // CHECK: BinaryOperator {{.*}} '_Complex double' '+'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex double' <FloatingComplexCast>
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cf'
+  // CHECK-NEXT: FloatingLiteral {{.*}} 'double' 1.0
+
+  // This should convert the float constant to double, then produce a
+  // _Complex double.
+  (void)(cd + 1.0f);
+  // CHECK: BinaryOperator {{.*}} '_Complex double' '+'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex double' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cd'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} 'double' <FloatingCast>
+  // CHECK-NEXT: FloatingLiteral {{.*}} 'float' 1.0
+
+  // This should convert the int constant to float, then produce a
+  // _Complex float.
+  (void)(cf + 1);
+  // CHECK: BinaryOperator {{.*}} '_Complex float' '+'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cf'
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} 'float' <IntegralToFloating>
+  // CHECK-NEXT: IntegerLiteral {{.*}} 'int' 1
+
+  // Show that we do not promote a _Complex float to _Complex double as part of
+  // the default argument promotions when passing to a variadic function.
+  variadic(1, cf);
+  // CHECK: CallExpr
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} <FunctionToPointerDecay>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'variadic'
+  // CHECK-NEXT: IntegerLiteral {{.*}} 'int' 1
+  // CHECK-NEXT: ImplicitCastExpr {{.*}} '_Complex float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr {{.*}} 'cf'
+}
+
diff --git a/clang/test/C/C99/n809_3.c b/clang/test/C/C99/n809_3.c
new file mode 100644
index 0000000000000..f1283f5fe1632
--- /dev/null
+++ b/clang/test/C/C99/n809_3.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -emit-llvm -std=c99 %s -o - | FileCheck %s
+
+// Demonstrate that statics are properly zero initialized.
+static _Complex float f_global;
+void func(void) {
+  static _Complex double d_local;
+  d_local = f_global;
+}
+
+// CHECK-DAG: @func.d_local = internal global { double, double } zeroinitializer
+// CHECK-DAG: @f_global = internal global { float, float } zeroinitializer
+
diff --git a/clang/www/c_status.html b/clang/www/c_status.html
index 411f55447be77..9893170ae8473 100644
--- a/clang/www/c_status.html
+++ b/clang/www/c_status.html
@@ -132,29 +132,22 @@ <h2 id="c99">C99 implementation status</h2>
       <td>Unknown</td>
       <td class="full" align="center">Yes</td>
     </tr>
-    <tr id="complex">
-      <td rowspan="6">complex and imaginary support in &lt;complex.h&gt;</td>
+    <tr>
+      <td>complex and imaginary support in &lt;complex.h&gt;</td>
+      <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n693.ps">N693</a></td>
+      <td class="partial" align="center">
+        <details><summary>Partial</summary>
+          Clang supports <code>_Complex</code> type specifiers but does not
+          support <code>_Imaginary</code> type specifiers. Support for
+          <code>_Imaginary</code> is optional in C99 and Clang does not claim
+          conformance to Annex G.<br />
+          <br />
+          <code>_Complex</code> support requires an underlying support library
+          such as compiler-rt to provide functions like <code>__divsc3</code>,
+          but compiler-rt is not supported on Windows.
+        </details>
+      </td>
     </tr>
-      <tr>
-        <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n620.ps">N620</a></td>
-        <td class="unknown" align="center">Unknown</td>
-      </tr>
-      <tr>
-        <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n638.ps">N638</a></td>
-        <td class="unknown" align="center">Unknown</td>
-      </tr>
-      <tr>
-        <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n657.ps">N657</a></td>
-        <td class="unknown" align="center">Unknown</td>
-      </tr>
-      <tr>
-        <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n694.ps">N694</a></td>
-        <td class="unknown" align="center">Unknown</td>
-      </tr>
-      <tr>
-        <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n809.ps">N809</a></td>
-        <td class="unknown" align="center">Unknown</td>
-      </tr>
     <tr>
       <td>type-generic math macros in &lt;tgmath.h&gt;</td>
       <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n693.ps">N693</a></td>

From b074f25329501487e312b59e463a2d5f743090f8 Mon Sep 17 00:00:00 2001
From: elizabethandrews <elizabeth.andrews@intel.com>
Date: Fri, 12 Apr 2024 14:41:49 -0400
Subject: [PATCH 288/886] [NFC][Clang] Fix static analyzer concern (#88179)

Fix static analyzer concerns about dereferencing
null values.
---
 clang/lib/AST/Interp/InterpState.h | 6 +++++-
 clang/lib/Sema/SemaAPINotes.cpp    | 2 ++
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/clang/lib/AST/Interp/InterpState.h b/clang/lib/AST/Interp/InterpState.h
index 8f84bf6ed2eaf..c17cfad11b1e2 100644
--- a/clang/lib/AST/Interp/InterpState.h
+++ b/clang/lib/AST/Interp/InterpState.h
@@ -89,7 +89,11 @@ class InterpState final : public State, public SourceMapper {
 
   /// Delegates source mapping to the mapper.
   SourceInfo getSource(const Function *F, CodePtr PC) const override {
-    return M ? M->getSource(F, PC) : F->getSource(PC);
+    if (M)
+      return M->getSource(F, PC);
+
+    assert(F && "Function cannot be null");
+    return F->getSource(PC);
   }
 
   Context &getContext() const { return Ctx; }
diff --git a/clang/lib/Sema/SemaAPINotes.cpp b/clang/lib/Sema/SemaAPINotes.cpp
index a3128306c664f..4c445f28bba8c 100644
--- a/clang/lib/Sema/SemaAPINotes.cpp
+++ b/clang/lib/Sema/SemaAPINotes.cpp
@@ -463,6 +463,8 @@ static void ProcessAPINotes(Sema &S, FunctionOrMethod AnyFunc,
     D = MD;
   }
 
+  assert((FD || MD) && "Expecting Function or ObjCMethod");
+
   // Nullability of return type.
   if (Info.NullabilityAudited)
     applyNullability(S, D, Info.getReturnTypeInfo(), Metadata);

From 8eba9c398a5284660081a6a49ba2b23ef271d778 Mon Sep 17 00:00:00 2001
From: "Demchuk, Tennyson" <tennyson.demchuk@intel.com>
Date: Fri, 12 Apr 2024 12:07:08 -0700
Subject: [PATCH 289/886] Fix build failure.

---
 libclc/CMakeLists.txt                |   2 -
 libclc/cmake/modules/AddLibclc.cmake | 159 ---------------------------
 libclc/generic/lib/SOURCES           |   3 -
 3 files changed, 164 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index eb293ad5e0bd5..bdd5866a7ef07 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -64,12 +64,10 @@ set( LIBCLC_TEST_TARGETS_ALL
 set( LIBCLC_ROOT_DIR ${CMAKE_CURRENT_SOURCE_DIR} )
 set( LIBCLC_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR} )
 
-# $&$ maybe delete
 # mesa3d environment is only available since LLVM 4.0
 if( LLVM_PACKAGE_VERSION VERSION_GREATER_EQUAL 4.0.0 )
   list( APPEND LIBCLC_TARGETS_ALL amdgcn-mesa-mesa3d )
 endif()
-# $&$ end delete
 
 set( LIBCLC_TARGETS_TO_BUILD "all"
     CACHE STRING "Semicolon-separated list of targets to build, or 'all'." )
diff --git a/libclc/cmake/modules/AddLibclc.cmake b/libclc/cmake/modules/AddLibclc.cmake
index ac6a0c240393d..c458efb528b8f 100644
--- a/libclc/cmake/modules/AddLibclc.cmake
+++ b/libclc/cmake/modules/AddLibclc.cmake
@@ -1,4 +1,3 @@
-<<<<<<< HEAD
 function(add_libclc_alias alias target)
   cmake_parse_arguments(ARG "" "" PARENT_TARGET "" ${ARGN})
 
@@ -248,161 +247,3 @@ function(libclc_configure_lib_source OUT_LIST)
   set( ${OUT_LIST} ${rel_files} PARENT_SCOPE )
 
 endfunction(libclc_configure_lib_source OUT_LIST)
-=======
-# Compiles an OpenCL C - or assembles an LL file - to bytecode
-#
-# Arguments:
-# * TRIPLE <string>
-#     Target triple for which to compile the bytecode file.
-# * INPUT <string>
-#     File to compile/assemble to bytecode
-# * OUTPUT <string>
-#     Bytecode file to generate
-# * EXTRA_OPTS <string> ...
-#     List of compiler options to use. Note that some are added by default.
-# * DEPENDENCIES <string> ...
-#     List of extra dependencies to inject
-#
-# Depends on the libclc::clang and libclc::llvm-as targets for compiling and
-# assembling, respectively.
-function(compile_to_bc)
-  cmake_parse_arguments(ARG
-    ""
-    "TRIPLE;INPUT;OUTPUT"
-    "EXTRA_OPTS;DEPENDENCIES"
-    ${ARGN}
-  )
-
-  # If this is an LLVM IR file (identified soley by its file suffix),
-  # pre-process it with clang to a temp file, then assemble that to bytecode.
-  set( TMP_SUFFIX )
-  get_filename_component( FILE_EXT ${ARG_INPUT} EXT )
-  if( NOT ${FILE_EXT} STREQUAL ".ll" )
-    # Pass '-c' when not running the preprocessor
-    set( PP_OPTS -c )
-  else()
-    set( PP_OPTS -E;-P )
-    set( TMP_SUFFIX .tmp )
-  endif()
-
-  set( TARGET_ARG )
-  if( ARG_TRIPLE )
-    set( TARGET_ARG "-target" ${ARG_TRIPLE} )
-  endif()
-
-  add_custom_command(
-    OUTPUT ${ARG_OUTPUT}${TMP_SUFFIX}
-    COMMAND libclc::clang
-      ${TARGET_ARG}
-      ${PP_OPTS}
-      ${ARG_EXTRA_OPTS}
-      -MD -MF ${ARG_OUTPUT}.d -MT ${ARG_OUTPUT}${TMP_SUFFIX}
-      # LLVM 13 enables standard includes by default - we don't want
-      # those when pre-processing IR. We disable it unconditionally.
-      $<$<VERSION_GREATER_EQUAL:${LLVM_PACKAGE_VERSION},13.0.0>:-cl-no-stdinc>
-      -emit-llvm
-      -o ${ARG_OUTPUT}${TMP_SUFFIX}
-      -x cl
-      ${ARG_INPUT}
-    DEPENDS
-      libclc::clang
-      ${ARG_INPUT}
-      ${ARG_DEPENDENCIES}
-    DEPFILE ${ARG_OUTPUT}.d
-  )
-
-  if( ${FILE_EXT} STREQUAL ".ll" )
-    add_custom_command(
-      OUTPUT ${ARG_OUTPUT}
-      COMMAND libclc::llvm-as -o ${ARG_OUTPUT} ${ARG_OUTPUT}${TMP_SUFFIX}
-      DEPENDS libclc::llvm-as ${ARG_OUTPUT}${TMP_SUFFIX}
-    )
-  endif()
-endfunction()
-
-# Links together one or more bytecode files
-#
-# Arguments:
-# * TARGET <string>
-#     Custom target to create
-# * INPUT <string> ...
-#     List of bytecode files to link together
-function(link_bc)
-  cmake_parse_arguments(ARG
-    ""
-    "TARGET"
-    "INPUTS"
-    ${ARGN}
-  )
-
-  add_custom_command(
-    OUTPUT ${ARG_TARGET}.bc
-    COMMAND libclc::llvm-link -o ${ARG_TARGET}.bc ${ARG_INPUTS}
-    DEPENDS libclc::llvm-link ${ARG_INPUTS}
-  )
-
-  add_custom_target( ${ARG_TARGET} ALL DEPENDS ${ARG_TARGET}.bc )
-  set_target_properties( ${ARG_TARGET} PROPERTIES TARGET_FILE ${ARG_TARGET}.bc )
-endfunction()
-
-# Decomposes and returns variables based on a libclc triple and architecture
-# combination. Returns data via one or more optional output variables.
-#
-# Arguments:
-# * TRIPLE <string>
-#     libclc target triple to query
-# * DEVICE <string>
-#     libclc device to query
-#
-# Optional Arguments:
-# * CPU <var>
-#     Variable name to be set to the target CPU
-# * ARCH_SUFFIX <var>
-#     Variable name to be set to the triple/architecture suffix
-# * CLANG_TRIPLE <var>
-#     Variable name to be set to the normalized clang triple
-function(get_libclc_device_info)
-  cmake_parse_arguments(ARG
-    ""
-    "TRIPLE;DEVICE;CPU;ARCH_SUFFIX;CLANG_TRIPLE"
-    ""
-    ${ARGN}
-  )
-
-  if( NOT ARG_TRIPLE OR NOT ARG_DEVICE )
-    message( FATAL_ERROR "Must provide both TRIPLE and DEVICE" )
-  endif()
-
-  string( REPLACE "-" ";" TRIPLE  ${ARG_TRIPLE} )
-  list( GET TRIPLE 0 ARCH )
-
-  # Some targets don't have a specific device architecture to target
-  if( ARG_DEVICE STREQUAL none OR ARCH STREQUAL spirv OR ARCH STREQUAL spirv64 )
-    set( cpu )
-    set( arch_suffix "${ARG_TRIPLE}" )
-  else()
-    set( cpu "${ARG_DEVICE}" )
-    set( arch_suffix "${ARG_DEVICE}-${ARG_TRIPLE}" )
-  endif()
-
-  if( ARG_CPU )
-    set( ${ARG_CPU} ${cpu} PARENT_SCOPE )
-  endif()
-
-  if( ARG_ARCH_SUFFIX )
-    set( ${ARG_ARCH_SUFFIX} ${arch_suffix} PARENT_SCOPE )
-  endif()
-
-  # Some libclc targets are not real clang triples: return their canonical
-  # triples.
-  if( ARCH STREQUAL spirv OR ARCH STREQUAL clspv )
-    set( ARG_TRIPLE "spir--" )
-  elseif( ARCH STREQUAL spirv64 OR ARCH STREQUAL clspv64 )
-    set( ARG_TRIPLE "spir64--" )
-  endif()
-
-  if( ARG_CLANG_TRIPLE )
-    set( ${ARG_CLANG_TRIPLE} ${ARG_TRIPLE} PARENT_SCOPE )
-  endif()
-endfunction()
->>>>>>> 72f9881c3ffcf4be6361c3e4312d91c9c8d94a98
diff --git a/libclc/generic/lib/SOURCES b/libclc/generic/lib/SOURCES
index 3c88b71b010ab..0a7558459be24 100644
--- a/libclc/generic/lib/SOURCES
+++ b/libclc/generic/lib/SOURCES
@@ -46,10 +46,7 @@ cl_khr_int64_extended_atomics/atom_max.cl
 cl_khr_int64_extended_atomics/atom_min.cl
 cl_khr_int64_extended_atomics/atom_or.cl
 cl_khr_int64_extended_atomics/atom_xor.cl
-<<<<<<< HEAD
 convert-clc.cl
-=======
->>>>>>> 72f9881c3ffcf4be6361c3e4312d91c9c8d94a98
 common/degrees.cl
 common/mix.cl
 common/radians.cl

From c777c011a709dffd4fa5e79cad7947b7c3405d02 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Thorsten=20Sch=C3=BCtt?= <schuett@gmail.com>
Date: Fri, 12 Apr 2024 21:10:19 +0200
Subject: [PATCH 290/886] [GlobalIsel] Import vscale (#88240)

https://github.com/llvm/llvm-project/pull/84542
---
 llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp  |  4 ++
 .../AArch64/GlobalISel/irtranslator-vscale.ll | 52 +++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-vscale.ll

diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index 312e564f5d802..9b4575f7f34d4 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -2550,6 +2550,10 @@ bool IRTranslator::translateKnownIntrinsic(const CallInst &CI, Intrinsic::ID ID,
     MIRBuilder.buildInstr(TargetOpcode::G_RESET_FPMODE, {}, {});
     return true;
   }
+  case Intrinsic::vscale: {
+    MIRBuilder.buildVScale(getOrCreateVReg(CI), 1);
+    return true;
+  }
   case Intrinsic::prefetch: {
     Value *Addr = CI.getOperand(0);
     unsigned RW = cast<ConstantInt>(CI.getOperand(1))->getZExtValue();
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-vscale.ll b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-vscale.ll
new file mode 100644
index 0000000000000..d06e2c64c2457
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/irtranslator-vscale.ll
@@ -0,0 +1,52 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+; RUN: llc -O0 -mtriple=aarch64-linux-gnu -global-isel -stop-after=irtranslator %s -o - | FileCheck %s
+
+define i64 @call_vscale_i64() {
+  ; CHECK-LABEL: name: call_vscale_i64
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[VSCALE:%[0-9]+]]:_(s64) = G_VSCALE i64 1
+  ; CHECK-NEXT:   $x0 = COPY [[VSCALE]](s64)
+  ; CHECK-NEXT:   RET_ReallyLR implicit $x0
+entry:
+  %vscale = call i64 @llvm.vscale.64()
+  ret i64 %vscale
+}
+
+define i64 @call_vscale_i32() {
+  ; CHECK-LABEL: name: call_vscale_i32
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[VSCALE:%[0-9]+]]:_(s32) = G_VSCALE i32 1
+  ; CHECK-NEXT:   [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[VSCALE]](s32)
+  ; CHECK-NEXT:   $x0 = COPY [[ZEXT]](s64)
+  ; CHECK-NEXT:   RET_ReallyLR implicit $x0
+entry:
+  %vscale = call i32 @llvm.vscale.32()
+  %zext = zext i32 %vscale to i64
+  ret i64 %zext
+}
+
+define i64 @call_vscale_i16() {
+  ; CHECK-LABEL: name: call_vscale_i16
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[VSCALE:%[0-9]+]]:_(s16) = G_VSCALE i16 1
+  ; CHECK-NEXT:   [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[VSCALE]](s16)
+  ; CHECK-NEXT:   $x0 = COPY [[ZEXT]](s64)
+  ; CHECK-NEXT:   RET_ReallyLR implicit $x0
+entry:
+  %vscale = call i16 @llvm.vscale.16()
+  %zext = zext i16 %vscale to i64
+  ret i64 %zext
+}
+
+define i64 @call_vscale_i8() {
+  ; CHECK-LABEL: name: call_vscale_i8
+  ; CHECK: bb.1.entry:
+  ; CHECK-NEXT:   [[VSCALE:%[0-9]+]]:_(s8) = G_VSCALE i8 1
+  ; CHECK-NEXT:   [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[VSCALE]](s8)
+  ; CHECK-NEXT:   $x0 = COPY [[ZEXT]](s64)
+  ; CHECK-NEXT:   RET_ReallyLR implicit $x0
+entry:
+  %vscale = call i8 @llvm.vscale.8()
+  %zext = zext i8 %vscale to i64
+  ret i64 %zext
+}

From a9111d4a26514e169ca57c11e6533ca7c5408832 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Fri, 12 Apr 2024 12:33:19 -0700
Subject: [PATCH 291/886] [clang] Remove unused variable in test

---
 .../DependencyScanning/DependencyScanningFilesystemTest.cpp  | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
index 50cad72d223e3..697b7d70ff035 100644
--- a/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
+++ b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
@@ -102,11 +102,8 @@ TEST(DependencyScanningFilesystem, RealPathAndStatusInvariants) {
   InMemoryFS->addFile("/foo.c", 0, llvm::MemoryBuffer::getMemBuffer(""));
   InMemoryFS->addFile("/bar.c", 0, llvm::MemoryBuffer::getMemBuffer(""));
 
-  auto InstrumentingFS =
-      llvm::makeIntrusiveRefCnt<InstrumentingFilesystem>(InMemoryFS);
-
   DependencyScanningFilesystemSharedCache SharedCache;
-  DependencyScanningWorkerFilesystem DepFS(SharedCache, InstrumentingFS);
+  DependencyScanningWorkerFilesystem DepFS(SharedCache, InMemoryFS);
 
   // Success.
   {

From 6dbd4bb3b4c00d62bf1552e54e4370c6816ea060 Mon Sep 17 00:00:00 2001
From: Jing Wang <2019426+jingw@users.noreply.github.com>
Date: Fri, 12 Apr 2024 12:41:06 -0700
Subject: [PATCH 292/886] Add txtpb to the list of supported TextProto
 extensions (#88355)

According to
https://protobuf.dev/reference/protobuf/textformat-spec/#text-format-files,
txtpb is the canonical extension

Co-authored-by: Jing Wang <99jingw@gmail.com>
---
 clang/docs/ClangFormat.rst               | 2 +-
 clang/lib/Format/Format.cpp              | 6 +++++-
 clang/test/Format/lit.local.cfg          | 1 +
 clang/tools/clang-format/ClangFormat.cpp | 2 +-
 4 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/clang/docs/ClangFormat.rst b/clang/docs/ClangFormat.rst
index 80dc38a075c8f..dbd9c91ae508e 100644
--- a/clang/docs/ClangFormat.rst
+++ b/clang/docs/ClangFormat.rst
@@ -54,7 +54,7 @@ to format C/C++/Java/JavaScript/JSON/Objective-C/Protobuf/C# code.
                                        Objective-C: .m .mm
                                        Proto: .proto .protodevel
                                        TableGen: .td
-                                       TextProto: .textpb .pb.txt .textproto .asciipb
+                                       TextProto: .txtpb .textpb .pb.txt .textproto .asciipb
                                        Verilog: .sv .svh .v .vh
     --cursor=<uint>                - The position of the cursor when invoking
                                      clang-format from an editor integration
diff --git a/clang/lib/Format/Format.cpp b/clang/lib/Format/Format.cpp
index 89e6c19b0af45..ccb2c9190e2ef 100644
--- a/clang/lib/Format/Format.cpp
+++ b/clang/lib/Format/Format.cpp
@@ -3891,7 +3891,11 @@ static FormatStyle::LanguageKind getLanguageByFileName(StringRef FileName) {
       FileName.ends_with_insensitive(".protodevel")) {
     return FormatStyle::LK_Proto;
   }
-  if (FileName.ends_with_insensitive(".textpb") ||
+  // txtpb is the canonical extension, and textproto is the legacy canonical
+  // extension
+  // https://protobuf.dev/reference/protobuf/textformat-spec/#text-format-files
+  if (FileName.ends_with_insensitive(".txtpb") ||
+      FileName.ends_with_insensitive(".textpb") ||
       FileName.ends_with_insensitive(".pb.txt") ||
       FileName.ends_with_insensitive(".textproto") ||
       FileName.ends_with_insensitive(".asciipb")) {
diff --git a/clang/test/Format/lit.local.cfg b/clang/test/Format/lit.local.cfg
index 3851397a3f7cc..3eb0f54ceaa6f 100644
--- a/clang/test/Format/lit.local.cfg
+++ b/clang/test/Format/lit.local.cfg
@@ -12,6 +12,7 @@ config.suffixes = [
     ".proto",
     ".protodevel",
     ".pb.txt",
+    ".txtpb",
     ".textproto",
     ".textpb",
     ".asciipb",
diff --git a/clang/tools/clang-format/ClangFormat.cpp b/clang/tools/clang-format/ClangFormat.cpp
index ed401135ad843..feb733fe3c9e0 100644
--- a/clang/tools/clang-format/ClangFormat.cpp
+++ b/clang/tools/clang-format/ClangFormat.cpp
@@ -92,7 +92,7 @@ static cl::opt<std::string> AssumeFileName(
              "  Objective-C: .m .mm\n"
              "  Proto: .proto .protodevel\n"
              "  TableGen: .td\n"
-             "  TextProto: .textpb .pb.txt .textproto .asciipb\n"
+             "  TextProto: .txtpb .textpb .pb.txt .textproto .asciipb\n"
              "  Verilog: .sv .svh .v .vh"),
     cl::init("<stdin>"), cl::cat(ClangFormatCategory));
 

From 4078afc6d23e25df6baedad61b224ef86a94d42f Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek <Krzysztof.Parzyszek@amd.com>
Date: Fri, 12 Apr 2024 14:52:24 -0500
Subject: [PATCH 293/886] [LLVM][OpenMP] Add "nowait" clause as valid for
 "workshare" (#88426)

Add the "nowait" clause to the list of allowed clauses for the "workshare"
directive. This will make it consistent with other directives (which are
shared between C/C++ and Fortran).

The parser will still reject "nowait" on "!$omp workshare", so this has no
effect on accepting/rejecting Fortran source code.
---
 flang/lib/Semantics/check-omp-structure.h         | 1 +
 flang/test/Semantics/OpenMP/clause-validity01.f90 | 3 +++
 llvm/include/llvm/Frontend/OpenMP/OMP.td          | 3 +++
 3 files changed, 7 insertions(+)

diff --git a/flang/lib/Semantics/check-omp-structure.h b/flang/lib/Semantics/check-omp-structure.h
index 33243d926cf16..8287653458e1c 100644
--- a/flang/lib/Semantics/check-omp-structure.h
+++ b/flang/lib/Semantics/check-omp-structure.h
@@ -42,6 +42,7 @@ static const OmpDirectiveSet noWaitClauseNotAllowedSet{
     Directive::OMPD_do_simd,
     Directive::OMPD_sections,
     Directive::OMPD_single,
+    Directive::OMPD_workshare,
 };
 } // namespace omp
 } // namespace llvm
diff --git a/flang/test/Semantics/OpenMP/clause-validity01.f90 b/flang/test/Semantics/OpenMP/clause-validity01.f90
index 977a529864615..74f154bb0ad67 100644
--- a/flang/test/Semantics/OpenMP/clause-validity01.f90
+++ b/flang/test/Semantics/OpenMP/clause-validity01.f90
@@ -342,6 +342,9 @@
   a = 1.0
   !ERROR: COPYPRIVATE clause is not allowed on the END WORKSHARE directive
   !$omp end workshare nowait copyprivate(a)
+  !$omp workshare nowait
+  !ERROR: NOWAIT clause is not allowed on the WORKSHARE directive, use it on OMP END WORKSHARE directive
+  !$omp end workshare
   !$omp end parallel
 
 ! 2.8.1 simd-clause -> safelen-clause |
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index d9a931438b429..e91169e8da1aa 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -2115,6 +2115,9 @@ def OMP_scope : Directive<"scope"> {
   let association = AS_Block;
 }
 def OMP_Workshare : Directive<"workshare"> {
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_NoWait>
+  ];
   let association = AS_Block;
 }
 def OMP_ParallelWorkshare : Directive<"parallel workshare"> {

From 41ff91e143a72dcd5755c9ad2160cf3c333b50c6 Mon Sep 17 00:00:00 2001
From: OverMighty <its.overmighty@gmail.com>
Date: Fri, 12 Apr 2024 21:01:36 +0100
Subject: [PATCH 294/886] Reapply "[libc][math][c23] Add remaining linux/*
 entrypoints for {,u}fromfp{,x}* (#86692)" (#88567)

This reverts commit 8a071678a9091d536eae29912ca7be6238105956.

The test failure on 32-bit Arm should have been fixed by #86892.

cc @nickdesaulniers @lntue
---
 libc/config/linux/aarch64/entrypoints.txt | 16 ++++++++++++++++
 libc/config/linux/arm/entrypoints.txt     | 12 ++++++++++++
 libc/config/linux/riscv/entrypoints.txt   | 16 ++++++++++++++++
 3 files changed, 44 insertions(+)

diff --git a/libc/config/linux/aarch64/entrypoints.txt b/libc/config/linux/aarch64/entrypoints.txt
index f5f5c437685a2..2952baacdd67f 100644
--- a/libc/config/linux/aarch64/entrypoints.txt
+++ b/libc/config/linux/aarch64/entrypoints.txt
@@ -398,6 +398,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.frexp
     libc.src.math.frexpf
     libc.src.math.frexpl
+    libc.src.math.fromfp
+    libc.src.math.fromfpf
+    libc.src.math.fromfpl
+    libc.src.math.fromfpx
+    libc.src.math.fromfpxf
+    libc.src.math.fromfpxl
     libc.src.math.hypot
     libc.src.math.hypotf
     libc.src.math.ilogb
@@ -480,6 +486,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.trunc
     libc.src.math.truncf
     libc.src.math.truncl
+    libc.src.math.ufromfp
+    libc.src.math.ufromfpf
+    libc.src.math.ufromfpl
+    libc.src.math.ufromfpx
+    libc.src.math.ufromfpxf
+    libc.src.math.ufromfpxl
 )
 
 if(LIBC_TYPES_HAS_FLOAT128)
@@ -502,6 +514,8 @@ if(LIBC_TYPES_HAS_FLOAT128)
     libc.src.math.fminimum_mag_numf128
     libc.src.math.fmodf128
     libc.src.math.frexpf128
+    libc.src.math.fromfpf128
+    libc.src.math.fromfpxf128
     libc.src.math.ilogbf128
     libc.src.math.ldexpf128
     libc.src.math.llogbf128
@@ -519,6 +533,8 @@ if(LIBC_TYPES_HAS_FLOAT128)
     libc.src.math.roundf128
     libc.src.math.sqrtf128
     libc.src.math.truncf128
+    libc.src.math.ufromfpf128
+    libc.src.math.ufromfpxf128
   )
 endif()
 
diff --git a/libc/config/linux/arm/entrypoints.txt b/libc/config/linux/arm/entrypoints.txt
index fca50735d320b..35fd588a9a6c4 100644
--- a/libc/config/linux/arm/entrypoints.txt
+++ b/libc/config/linux/arm/entrypoints.txt
@@ -265,6 +265,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.frexp
     libc.src.math.frexpf
     libc.src.math.frexpl
+    libc.src.math.fromfp
+    libc.src.math.fromfpf
+    libc.src.math.fromfpl
+    libc.src.math.fromfpx
+    libc.src.math.fromfpxf
+    libc.src.math.fromfpxl
     libc.src.math.hypot
     libc.src.math.hypotf
     libc.src.math.ilogb
@@ -347,6 +353,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.trunc
     libc.src.math.truncf
     libc.src.math.truncl
+    libc.src.math.ufromfp
+    libc.src.math.ufromfpf
+    libc.src.math.ufromfpl
+    libc.src.math.ufromfpx
+    libc.src.math.ufromfpxf
+    libc.src.math.ufromfpxl
 )
 
 set(TARGET_LLVMLIBC_ENTRYPOINTS
diff --git a/libc/config/linux/riscv/entrypoints.txt b/libc/config/linux/riscv/entrypoints.txt
index 71289789158f4..47c03a61c45a9 100644
--- a/libc/config/linux/riscv/entrypoints.txt
+++ b/libc/config/linux/riscv/entrypoints.txt
@@ -406,6 +406,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.frexp
     libc.src.math.frexpf
     libc.src.math.frexpl
+    libc.src.math.fromfp
+    libc.src.math.fromfpf
+    libc.src.math.fromfpl
+    libc.src.math.fromfpx
+    libc.src.math.fromfpxf
+    libc.src.math.fromfpxl
     libc.src.math.hypot
     libc.src.math.hypotf
     libc.src.math.ilogb
@@ -488,6 +494,12 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.math.trunc
     libc.src.math.truncf
     libc.src.math.truncl
+    libc.src.math.ufromfp
+    libc.src.math.ufromfpf
+    libc.src.math.ufromfpl
+    libc.src.math.ufromfpx
+    libc.src.math.ufromfpxf
+    libc.src.math.ufromfpxl
 )
 
 if(LIBC_TYPES_HAS_FLOAT128)
@@ -510,6 +522,8 @@ if(LIBC_TYPES_HAS_FLOAT128)
     libc.src.math.fminimum_mag_numf128
     libc.src.math.fmodf128
     libc.src.math.frexpf128
+    libc.src.math.fromfpf128
+    libc.src.math.fromfpxf128
     libc.src.math.ilogbf128
     libc.src.math.ldexpf128
     libc.src.math.llogbf128
@@ -527,6 +541,8 @@ if(LIBC_TYPES_HAS_FLOAT128)
     libc.src.math.roundf128
     libc.src.math.sqrtf128
     libc.src.math.truncf128
+    libc.src.math.ufromfpf128
+    libc.src.math.ufromfpxf128
   )
 endif()
 

From 5300a6731e98fbcf7bca68374e934de737166698 Mon Sep 17 00:00:00 2001
From: Jonathan Peyton <jonathan.l.peyton@intel.com>
Date: Fri, 12 Apr 2024 15:13:59 -0500
Subject: [PATCH 295/886] [OpenMP] Fix re-locking hang found in issue 86684
 (#88539)

This was initially reported here (including stacktraces):
https://stackoverflow.com/questions/78183545/does-compiling-imagick-with-openmp-enabled-in-freebsd-13-2-cause-sched-yield

If `__kmp_register_library_startup()` detects that another instance of
the library is present, `__kmp_is_address_mapped()` is eventually
called. which uses `kmpc_alloc()` to allocate memory. This function
calls `__kmp_entry_thread()` to access the thread-local memory pool,
which is a bad idea during initialization. This macro internally calls
`__kmp_get_global_thread_id_reg()` which sets the bootstrap lock at the
beginning (before calling `__kmp_register_library_startup()`).

The fix is to use `KMP_INTERNAL_MALLOC()`/`KMP_INTERNAL_FREE()` instead
of `kmpc_malloc()`/`kmpc_free()`. `KMP_INTERNAL_MALLOC` and
`KMP_INTERNAL_FREE` do not use any bootstrap locks. They just translate
to `malloc()`/`free()` and are meant to be used during library
initialization before other library-specific allocators have been
initialized.

Fixes: #86684
---
 openmp/runtime/src/z_Linux_util.cpp | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/openmp/runtime/src/z_Linux_util.cpp b/openmp/runtime/src/z_Linux_util.cpp
index be47be0dcec2a..11ce083f4801d 100644
--- a/openmp/runtime/src/z_Linux_util.cpp
+++ b/openmp/runtime/src/z_Linux_util.cpp
@@ -2141,10 +2141,10 @@ int __kmp_is_address_mapped(void *addr) {
   // We pass from number of vm entry's semantic
   // to size of whole entry map list.
   lstsz = lstsz * 4 / 3;
-  buf = reinterpret_cast<char *>(kmpc_malloc(lstsz));
+  buf = reinterpret_cast<char *>(KMP_INTERNAL_MALLOC(lstsz));
   rc = sysctl(mib, 4, buf, &lstsz, NULL, 0);
   if (rc < 0) {
-    kmpc_free(buf);
+    KMP_INTERNAL_FREE(buf);
     return 0;
   }
 
@@ -2168,7 +2168,7 @@ int __kmp_is_address_mapped(void *addr) {
     }
     lw += cursz;
   }
-  kmpc_free(buf);
+  KMP_INTERNAL_FREE(buf);
 #elif KMP_OS_DRAGONFLY
   char err[_POSIX2_LINE_MAX];
   kinfo_proc *proc;
@@ -2234,12 +2234,12 @@ int __kmp_is_address_mapped(void *addr) {
     return 0;
   }
 
-  buf = kmpc_malloc(sz);
+  buf = KMP_INTERNAL_MALLOC(sz);
 
   while (sz > 0 && (rd = pread(file, buf, sz, 0)) == sz) {
     void *newbuf;
     sz <<= 1;
-    newbuf = kmpc_realloc(buf, sz);
+    newbuf = KMP_INTERNAL_REALLOC(buf, sz);
     buf = newbuf;
   }
 
@@ -2255,7 +2255,7 @@ int __kmp_is_address_mapped(void *addr) {
     }
   }
 
-  kmpc_free(map);
+  KMP_INTERNAL_FREE(map);
   close(file);
   KMP_INTERNAL_FREE(name);
 #elif KMP_OS_DARWIN

From dfafe3822b1e59f938c256c25a92a7b555f78469 Mon Sep 17 00:00:00 2001
From: Alexander Richardson <alexrichardson@google.com>
Date: Fri, 12 Apr 2024 13:20:30 -0700
Subject: [PATCH 296/886] Reland "[compiler-rt] Allow running tests without
 installing first"

Currently, the testsuite uses the default runtimes path to find the
runtimes libraries which may or may not match the just-built runtimes.
This change uses the `-resource-dir` flag for clang whenever
`COMPILER_RT_TEST_STANDALONE_BUILD_LIBS` is set to ensure that we are
actually testing the currently built libraries rather than the ones
bundled with `${COMPILER_RT_TEST_COMPILER}`.

The existing logic works fine when clang and compiler-rt share the same
build directory ``-DLLVM_ENABLE_PROJECTS=clang;compiler-rt`, but when
building compiler-rt separately we need to tell the compiler used for
the tests where it can find the just-built libraries.

This reduces the fixes check-all failures to one in my configuration:
```
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo -G Ninja
-DCMAKE_C_COMPILER=$HOME/output/upstream-llvm/bin/clang
-DCMAKE_CXX_COMPILER=$HOME/output/upstream-llvm/bin/clang++
-DCOMPILER_RT_INCLUDE_TESTS=ON
-DLLVM_EXTERNAL_LIT=$HOME/build/upstream-llvm-project-build/bin/llvm-lit
-DLLVM_CMAKE_DIR=$HOME/output/upstream-llvm
-DCOMPILER_RT_DEBUG=OFF
-S $HOME/src/upstream-llvm-project/compiler-rt
-B $HOME/src/upstream-llvm-project/compiler-rt/cmake-build-all-sanitizers
```

This relands the previous PR with fixes for Windows.
Depends on https://github.com/llvm/llvm-project/pull/88074 to be merged
first for GCC buildbots.

Pull Request: https://github.com/llvm/llvm-project/pull/88075
---
 compiler-rt/CMakeLists.txt                    | 23 +++++
 compiler-rt/cmake/Modules/AddCompilerRT.cmake |  1 +
 compiler-rt/test/CMakeLists.txt               |  8 --
 compiler-rt/test/fuzzer/lit.cfg.py            |  1 +
 compiler-rt/test/lit.common.cfg.py            | 89 ++++++++++++++-----
 compiler-rt/test/lit.common.configured.in     |  1 +
 compiler-rt/test/safestack/lit.cfg.py         | 10 ++-
 7 files changed, 101 insertions(+), 32 deletions(-)

diff --git a/compiler-rt/CMakeLists.txt b/compiler-rt/CMakeLists.txt
index f4e92f14db857..8649507ce1c79 100644
--- a/compiler-rt/CMakeLists.txt
+++ b/compiler-rt/CMakeLists.txt
@@ -584,6 +584,29 @@ string(APPEND COMPILER_RT_TEST_COMPILER_CFLAGS " ${stdlib_flag}")
 string(REPLACE " " ";" COMPILER_RT_UNITTEST_CFLAGS "${COMPILER_RT_TEST_COMPILER_CFLAGS}")
 set(COMPILER_RT_UNITTEST_LINK_FLAGS ${COMPILER_RT_UNITTEST_CFLAGS})
 
+option(COMPILER_RT_TEST_STANDALONE_BUILD_LIBS
+  "When set to ON and testing in a standalone build, test the runtime \
+  libraries built by this standalone build rather than the runtime libraries \
+  shipped with the compiler (used for testing). When set to OFF and testing \
+  in a standalone build, test the runtime libraries shipped with the compiler \
+  (used for testing). This option has no effect if the compiler and this \
+  build are configured to use the same runtime library path."
+  ON)
+if (COMPILER_RT_TEST_STANDALONE_BUILD_LIBS)
+  # Ensure that the unit tests can find the sanitizer headers prior to installation.
+  list(APPEND COMPILER_RT_UNITTEST_CFLAGS "-I${CMAKE_CURRENT_LIST_DIR}/include")
+  # Ensure that unit tests link against the just-built runtime libraries instead
+  # of the ones bundled with the compiler by overriding the resource directory.
+  #
+  if ("${COMPILER_RT_TEST_COMPILER_ID}" MATCHES "Clang")
+    list(APPEND COMPILER_RT_UNITTEST_LINK_FLAGS "-resource-dir=${COMPILER_RT_OUTPUT_DIR}")
+  endif()
+  get_compiler_rt_output_dir(${COMPILER_RT_DEFAULT_TARGET_ARCH} rtlib_dir)
+  if (NOT WIN32)
+    list(APPEND COMPILER_RT_UNITTEST_LINK_FLAGS "-Wl,-rpath,${rtlib_dir}")
+  endif()
+endif()
+
 if(COMPILER_RT_USE_LLVM_UNWINDER)
   # We're linking directly against the libunwind that we're building so don't
   # try to link in the toolchain's default libunwind which may be missing.
diff --git a/compiler-rt/cmake/Modules/AddCompilerRT.cmake b/compiler-rt/cmake/Modules/AddCompilerRT.cmake
index dc36f16c661c2..6e0d9dbff65a9 100644
--- a/compiler-rt/cmake/Modules/AddCompilerRT.cmake
+++ b/compiler-rt/cmake/Modules/AddCompilerRT.cmake
@@ -771,6 +771,7 @@ function(configure_compiler_rt_lit_site_cfg input output)
   get_compiler_rt_output_dir(${COMPILER_RT_DEFAULT_TARGET_ARCH} output_dir)
 
   string(REPLACE ${CMAKE_CFG_INTDIR} ${LLVM_BUILD_MODE} COMPILER_RT_RESOLVED_TEST_COMPILER ${COMPILER_RT_TEST_COMPILER})
+  string(REPLACE ${CMAKE_CFG_INTDIR} ${LLVM_BUILD_MODE} COMPILER_RT_RESOLVED_OUTPUT_DIR ${COMPILER_RT_OUTPUT_DIR})
   string(REPLACE ${CMAKE_CFG_INTDIR} ${LLVM_BUILD_MODE} COMPILER_RT_RESOLVED_LIBRARY_OUTPUT_DIR ${output_dir})
 
   configure_lit_site_cfg(${input} ${output})
diff --git a/compiler-rt/test/CMakeLists.txt b/compiler-rt/test/CMakeLists.txt
index c186be1e44fd9..edc007aaf477a 100644
--- a/compiler-rt/test/CMakeLists.txt
+++ b/compiler-rt/test/CMakeLists.txt
@@ -1,14 +1,6 @@
 # Needed for lit support in standalone builds.
 include(AddLLVM)
 
-option(COMPILER_RT_TEST_STANDALONE_BUILD_LIBS
-  "When set to ON and testing in a standalone build, test the runtime \
-  libraries built by this standalone build rather than the runtime libraries \
-  shipped with the compiler (used for testing). When set to OFF and testing \
-  in a standalone build, test the runtime libraries shipped with the compiler \
-  (used for testing). This option has no effect if the compiler and this \
-  build are configured to use the same runtime library path."
-  ON)
 pythonize_bool(COMPILER_RT_TEST_STANDALONE_BUILD_LIBS)
 
 pythonize_bool(LLVM_ENABLE_EXPENSIVE_CHECKS)
diff --git a/compiler-rt/test/fuzzer/lit.cfg.py b/compiler-rt/test/fuzzer/lit.cfg.py
index 9084254b3b15c..0c0550ac729d6 100644
--- a/compiler-rt/test/fuzzer/lit.cfg.py
+++ b/compiler-rt/test/fuzzer/lit.cfg.py
@@ -106,6 +106,7 @@ def generate_compiler_cmd(is_cpp=True, fuzzer_enabled=True, msan_enabled=False):
     return " ".join(
         [
             compiler_cmd,
+            config.target_cflags,
             std_cmd,
             "-O2 -gline-tables-only",
             sanitizers_cmd,
diff --git a/compiler-rt/test/lit.common.cfg.py b/compiler-rt/test/lit.common.cfg.py
index 0ac20a9831d95..28f126a11b16b 100644
--- a/compiler-rt/test/lit.common.cfg.py
+++ b/compiler-rt/test/lit.common.cfg.py
@@ -14,6 +14,27 @@
 import lit.util
 
 
+def get_path_from_clang(args, allow_failure):
+    clang_cmd = [
+        config.clang.strip(),
+        f"--target={config.target_triple}",
+        *args,
+    ]
+    path = None
+    try:
+        result = subprocess.run(
+            clang_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True
+        )
+        path = result.stdout.decode().strip()
+    except subprocess.CalledProcessError as e:
+        msg = f"Failed to run {clang_cmd}\nrc:{e.returncode}\nstdout:{e.stdout}\ne.stderr{e.stderr}"
+        if allow_failure:
+            lit_config.warning(msg)
+        else:
+            lit_config.fatal(msg)
+    return path, clang_cmd
+
+
 def find_compiler_libdir():
     """
     Returns the path to library resource directory used
@@ -26,26 +47,6 @@ def find_compiler_libdir():
         # TODO: Support other compilers.
         return None
 
-    def get_path_from_clang(args, allow_failure):
-        clang_cmd = [
-            config.clang.strip(),
-            f"--target={config.target_triple}",
-        ]
-        clang_cmd.extend(args)
-        path = None
-        try:
-            result = subprocess.run(
-                clang_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True
-            )
-            path = result.stdout.decode().strip()
-        except subprocess.CalledProcessError as e:
-            msg = f"Failed to run {clang_cmd}\nrc:{e.returncode}\nstdout:{e.stdout}\ne.stderr{e.stderr}"
-            if allow_failure:
-                lit_config.warning(msg)
-            else:
-                lit_config.fatal(msg)
-        return path, clang_cmd
-
     # Try using `-print-runtime-dir`. This is only supported by very new versions of Clang.
     # so allow failure here.
     runtime_dir, clang_cmd = get_path_from_clang(
@@ -130,6 +131,7 @@ def push_dynamic_library_lookup_path(config, new_path):
     config.available_features.add("shell")
 
 target_is_msvc = bool(re.match(r".*-windows-msvc$", config.target_triple))
+target_is_windows = bool(re.match(r".*-windows.*$", config.target_triple))
 
 compiler_id = getattr(config, "compiler_id", None)
 if compiler_id == "Clang":
@@ -168,10 +170,53 @@ def push_dynamic_library_lookup_path(config, new_path):
             r"/i386(?=-[^/]+$)", "/x86_64", config.compiler_rt_libdir
         )
 
+
+# Check if the test compiler resource dir matches the local build directory
+# (which happens with -DLLVM_ENABLE_PROJECTS=clang;compiler-rt) or if we are
+# using an installed clang to test compiler-rt standalone. In the latter case
+# we may need to override the resource dir to match the path of the just-built
+# compiler-rt libraries.
+test_cc_resource_dir, _ = get_path_from_clang(
+    shlex.split(config.target_cflags) + ["-print-resource-dir"], allow_failure=True
+)
+# Normalize the path for comparison
+if test_cc_resource_dir is not None:
+    test_cc_resource_dir = os.path.realpath(test_cc_resource_dir)
+if lit_config.debug:
+    lit_config.note(f"Resource dir for {config.clang} is {test_cc_resource_dir}")
+local_build_resource_dir = os.path.realpath(config.compiler_rt_output_dir)
+if test_cc_resource_dir != local_build_resource_dir and config.test_standalone_build_libs:
+    if config.compiler_id == "Clang":
+        if lit_config.debug:
+            lit_config.note(
+                f"Overriding test compiler resource dir to use "
+                f'libraries in "{config.compiler_rt_libdir}"'
+            )
+        # Ensure that we use the just-built static libraries when linking by
+        # overriding the Clang resource directory. Additionally, we want to use
+        # the builtin headers shipped with clang (e.g. stdint.h), so we
+        # explicitly add this as an include path (since the headers are not
+        # going to be in the current compiler-rt build directory).
+        # We also tell the linker to add an RPATH entry for the local library
+        # directory so that the just-built shared libraries are used.
+        config.target_cflags += f" -nobuiltininc"
+        config.target_cflags += f" -I{config.compiler_rt_src_root}/include"
+        config.target_cflags += f" -idirafter {test_cc_resource_dir}/include"
+        config.target_cflags += f" -resource-dir={config.compiler_rt_output_dir}"
+        if not target_is_windows:
+            # Avoid passing -rpath on Windows where it is not supported.
+            config.target_cflags += f" -Wl,-rpath,{config.compiler_rt_libdir}"
+    else:
+        lit_config.warning(
+            f"Cannot override compiler-rt library directory with non-Clang "
+            f"compiler: {config.compiler_id}"
+        )
+
+
 # Ask the compiler for the path to libraries it is going to use. If this
 # doesn't match config.compiler_rt_libdir then it means we might be testing the
 # compiler's own runtime libraries rather than the ones we just built.
-# Warn about about this and handle appropriately.
+# Warn about this and handle appropriately.
 compiler_libdir = find_compiler_libdir()
 if compiler_libdir:
     compiler_rt_libdir_real = os.path.realpath(config.compiler_rt_libdir)
@@ -182,7 +227,7 @@ def push_dynamic_library_lookup_path(config, new_path):
             f'compiler-rt libdir:  "{compiler_rt_libdir_real}"'
         )
         if config.test_standalone_build_libs:
-            # Use just built runtime libraries, i.e. the the libraries this built just built.
+            # Use just built runtime libraries, i.e. the libraries this build just built.
             if not config.test_suite_supports_overriding_runtime_lib_path:
                 # Test suite doesn't support this configuration.
                 # TODO(dliew): This should be an error but it seems several bots are
diff --git a/compiler-rt/test/lit.common.configured.in b/compiler-rt/test/lit.common.configured.in
index fff5dc6cc7509..8889b816b149f 100644
--- a/compiler-rt/test/lit.common.configured.in
+++ b/compiler-rt/test/lit.common.configured.in
@@ -27,6 +27,7 @@ set_default("compiler_id", "@COMPILER_RT_TEST_COMPILER_ID@")
 set_default("python_executable", "@Python3_EXECUTABLE@")
 set_default("compiler_rt_debug", @COMPILER_RT_DEBUG_PYBOOL@)
 set_default("compiler_rt_intercept_libdispatch", @COMPILER_RT_INTERCEPT_LIBDISPATCH_PYBOOL@)
+set_default("compiler_rt_output_dir", "@COMPILER_RT_RESOLVED_OUTPUT_DIR@")
 set_default("compiler_rt_libdir", "@COMPILER_RT_RESOLVED_LIBRARY_OUTPUT_DIR@")
 set_default("emulator", "@COMPILER_RT_EMULATOR@")
 set_default("asan_shadow_scale", "@COMPILER_RT_ASAN_SHADOW_SCALE@")
diff --git a/compiler-rt/test/safestack/lit.cfg.py b/compiler-rt/test/safestack/lit.cfg.py
index adf27a0d7e5ea..aadb8bf0d5c77 100644
--- a/compiler-rt/test/safestack/lit.cfg.py
+++ b/compiler-rt/test/safestack/lit.cfg.py
@@ -13,10 +13,16 @@
 
 # Add clang substitutions.
 config.substitutions.append(
-    ("%clang_nosafestack ", config.clang + " -O0 -fno-sanitize=safe-stack ")
+    (
+        "%clang_nosafestack ",
+        config.clang + config.target_cflags + " -O0 -fno-sanitize=safe-stack ",
+    )
 )
 config.substitutions.append(
-    ("%clang_safestack ", config.clang + " -O0 -fsanitize=safe-stack ")
+    (
+        "%clang_safestack ",
+        config.clang + config.target_cflags + " -O0 -fsanitize=safe-stack ",
+    )
 )
 
 if config.lto_supported:

From 588987405a979f724ab0aa7ad788cc5e87af3649 Mon Sep 17 00:00:00 2001
From: Artem Chikin <achikin@apple.com>
Date: Fri, 12 Apr 2024 13:34:47 -0700
Subject: [PATCH 297/886] [llvm][vfs] Make vfs::FileSystem::exists() virtual
 NFC (#88575)

Allow a `vfs::FileSystem` to provide a more efficient implementation of
`exists()` if they are able to. The existing `FileSystem`
implementations continue to default to using `status()` except that
overlay, proxy, and redirecting filesystems are taught to forward calls
to `exists()` correctly to their wrapped/external filesystem.

Co-authored-by: Ben Langmuir <blangmuir@apple.com>
---
 llvm/include/llvm/Support/VirtualFileSystem.h |   8 +-
 llvm/lib/Support/VirtualFileSystem.cpp        |  56 ++++++
 .../Support/VirtualFileSystemTest.cpp         | 165 ++++++++++++++++++
 3 files changed, 227 insertions(+), 2 deletions(-)

diff --git a/llvm/include/llvm/Support/VirtualFileSystem.h b/llvm/include/llvm/Support/VirtualFileSystem.h
index 4b1ca0c3d262b..0113d6b7da25d 100644
--- a/llvm/include/llvm/Support/VirtualFileSystem.h
+++ b/llvm/include/llvm/Support/VirtualFileSystem.h
@@ -300,8 +300,9 @@ class FileSystem : public llvm::ThreadSafeRefCountedBase<FileSystem>,
   virtual std::error_code getRealPath(const Twine &Path,
                                       SmallVectorImpl<char> &Output);
 
-  /// Check whether a file exists. Provided for convenience.
-  bool exists(const Twine &Path);
+  /// Check whether \p Path exists. By default this uses \c status(), but
+  /// filesystems may provide a more efficient implementation if available.
+  virtual bool exists(const Twine &Path);
 
   /// Is the file mounted on a local filesystem?
   virtual std::error_code isLocal(const Twine &Path, bool &Result);
@@ -386,6 +387,7 @@ class OverlayFileSystem : public RTTIExtends<OverlayFileSystem, FileSystem> {
   void pushOverlay(IntrusiveRefCntPtr<FileSystem> FS);
 
   llvm::ErrorOr<Status> status(const Twine &Path) override;
+  bool exists(const Twine &Path) override;
   llvm::ErrorOr<std::unique_ptr<File>>
   openFileForRead(const Twine &Path) override;
   directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
@@ -439,6 +441,7 @@ class ProxyFileSystem : public RTTIExtends<ProxyFileSystem, FileSystem> {
   llvm::ErrorOr<Status> status(const Twine &Path) override {
     return FS->status(Path);
   }
+  bool exists(const Twine &Path) override { return FS->exists(Path); }
   llvm::ErrorOr<std::unique_ptr<File>>
   openFileForRead(const Twine &Path) override {
     return FS->openFileForRead(Path);
@@ -1044,6 +1047,7 @@ class RedirectingFileSystem
          bool UseExternalNames, FileSystem &ExternalFS);
 
   ErrorOr<Status> status(const Twine &Path) override;
+  bool exists(const Twine &Path) override;
   ErrorOr<std::unique_ptr<File>> openFileForRead(const Twine &Path) override;
 
   std::error_code getRealPath(const Twine &Path,
diff --git a/llvm/lib/Support/VirtualFileSystem.cpp b/llvm/lib/Support/VirtualFileSystem.cpp
index 32b480028e71b..921af30bfcde9 100644
--- a/llvm/lib/Support/VirtualFileSystem.cpp
+++ b/llvm/lib/Support/VirtualFileSystem.cpp
@@ -438,6 +438,15 @@ ErrorOr<Status> OverlayFileSystem::status(const Twine &Path) {
   return make_error_code(llvm::errc::no_such_file_or_directory);
 }
 
+bool OverlayFileSystem::exists(const Twine &Path) {
+  // FIXME: handle symlinks that cross file systems
+  for (iterator I = overlays_begin(), E = overlays_end(); I != E; ++I) {
+    if ((*I)->exists(Path))
+      return true;
+  }
+  return false;
+}
+
 ErrorOr<std::unique_ptr<File>>
 OverlayFileSystem::openFileForRead(const llvm::Twine &Path) {
   // FIXME: handle symlinks that cross file systems
@@ -2428,6 +2437,53 @@ ErrorOr<Status> RedirectingFileSystem::status(const Twine &OriginalPath) {
   return S;
 }
 
+bool RedirectingFileSystem::exists(const Twine &OriginalPath) {
+  SmallString<256> Path;
+  OriginalPath.toVector(Path);
+
+  if (makeAbsolute(Path))
+    return false;
+
+  if (Redirection == RedirectKind::Fallback) {
+    // Attempt to find the original file first, only falling back to the
+    // mapped file if that fails.
+    if (ExternalFS->exists(Path))
+      return true;
+  }
+
+  ErrorOr<RedirectingFileSystem::LookupResult> Result = lookupPath(Path);
+  if (!Result) {
+    // Was not able to map file, fallthrough to using the original path if
+    // that was the specified redirection type.
+    if (Redirection == RedirectKind::Fallthrough &&
+        isFileNotFound(Result.getError()))
+      return ExternalFS->exists(Path);
+    return false;
+  }
+
+  std::optional<StringRef> ExtRedirect = Result->getExternalRedirect();
+  if (!ExtRedirect) {
+    assert(isa<RedirectingFileSystem::DirectoryEntry>(Result->E));
+    return true;
+  }
+
+  SmallString<256> RemappedPath((*ExtRedirect).str());
+  if (makeAbsolute(RemappedPath))
+    return false;
+
+  if (ExternalFS->exists(RemappedPath))
+    return true;
+
+  if (Redirection == RedirectKind::Fallthrough) {
+    // Mapped the file but it wasn't found in the underlying filesystem,
+    // fallthrough to using the original path if that was the specified
+    // redirection type.
+    return ExternalFS->exists(Path);
+  }
+
+  return false;
+}
+
 namespace {
 
 /// Provide a file wrapper with an overriden status.
diff --git a/llvm/unittests/Support/VirtualFileSystemTest.cpp b/llvm/unittests/Support/VirtualFileSystemTest.cpp
index 49a2e19e4f74c..e9b4ac3d92e1d 100644
--- a/llvm/unittests/Support/VirtualFileSystemTest.cpp
+++ b/llvm/unittests/Support/VirtualFileSystemTest.cpp
@@ -201,6 +201,21 @@ class ErrorDummyFileSystem : public DummyFileSystem {
   }
 };
 
+/// A version of \c DummyFileSystem that aborts on \c status() to test that
+/// \c exists() is being used.
+class NoStatusDummyFileSystem : public DummyFileSystem {
+public:
+  ErrorOr<vfs::Status> status(const Twine &Path) override {
+    llvm::report_fatal_error(
+        "unexpected call to NoStatusDummyFileSystem::status");
+  }
+
+  bool exists(const Twine &Path) override {
+    auto Status = DummyFileSystem::status(Path);
+    return Status && Status->exists();
+  }
+};
+
 /// Replace back-slashes by front-slashes.
 std::string getPosixPath(const Twine &S) {
   SmallString<128> Result;
@@ -968,6 +983,30 @@ TEST(OverlayFileSystemTest, PrintOutput) {
             Output);
 }
 
+TEST(OverlayFileSystemTest, Exists) {
+  IntrusiveRefCntPtr<DummyFileSystem> Lower(new NoStatusDummyFileSystem());
+  IntrusiveRefCntPtr<DummyFileSystem> Upper(new NoStatusDummyFileSystem());
+  IntrusiveRefCntPtr<vfs::OverlayFileSystem> O(
+      new vfs::OverlayFileSystem(Lower));
+  O->pushOverlay(Upper);
+
+  Lower->addDirectory("/both");
+  Upper->addDirectory("/both");
+  Lower->addRegularFile("/both/lower_file");
+  Upper->addRegularFile("/both/upper_file");
+  Lower->addDirectory("/lower");
+  Upper->addDirectory("/upper");
+
+  EXPECT_TRUE(O->exists("/both"));
+  EXPECT_TRUE(O->exists("/both"));
+  EXPECT_TRUE(O->exists("/both/lower_file"));
+  EXPECT_TRUE(O->exists("/both/upper_file"));
+  EXPECT_TRUE(O->exists("/lower"));
+  EXPECT_TRUE(O->exists("/upper"));
+  EXPECT_FALSE(O->exists("/both/nope"));
+  EXPECT_FALSE(O->exists("/nope"));
+}
+
 TEST(ProxyFileSystemTest, Basic) {
   IntrusiveRefCntPtr<vfs::InMemoryFileSystem> Base(
       new vfs::InMemoryFileSystem());
@@ -3364,6 +3403,11 @@ TEST(RedirectingFileSystemTest, ExternalPaths) {
       SeenPaths.push_back(Dir.str());
       return ProxyFileSystem::dir_begin(Dir, EC);
     }
+
+    bool exists(const Twine &Path) override {
+      SeenPaths.push_back(Path.str());
+      return ProxyFileSystem::exists(Path);
+    }
   };
 
   std::error_code EC;
@@ -3395,3 +3439,124 @@ TEST(RedirectingFileSystemTest, ExternalPaths) {
 
   EXPECT_EQ(CheckFS->SeenPaths, Expected);
 }
+
+TEST(RedirectingFileSystemTest, Exists) {
+  IntrusiveRefCntPtr<DummyFileSystem> Dummy(new NoStatusDummyFileSystem());
+  auto YAML =
+    MemoryBuffer::getMemBuffer("{\n"
+                               "  'version': 0,\n"
+                               "  'roots': [\n"
+                               "    {\n"
+                               "      'type': 'directory-remap',\n"
+                               "      'name': '/dremap',\n"
+                               "      'external-contents': '/a',\n"
+                               "    },"
+                               "    {\n"
+                               "      'type': 'directory-remap',\n"
+                               "      'name': '/dmissing',\n"
+                               "      'external-contents': '/dmissing',\n"
+                               "    },"
+                               "    {\n"
+                               "      'type': 'directory',\n"
+                               "      'name': '/both',\n"
+                               "      'contents': [\n"
+                               "        {\n"
+                               "          'type': 'file',\n"
+                               "          'name': 'vfile',\n"
+                               "          'external-contents': '/c'\n"
+                               "        }\n"
+                               "      ]\n"
+                               "    },\n"
+                               "    {\n"
+                               "      'type': 'directory',\n"
+                               "      'name': '/vdir',\n"
+                               "      'contents': ["
+                               "        {\n"
+                               "          'type': 'directory-remap',\n"
+                               "          'name': 'dremap',\n"
+                               "          'external-contents': '/b'\n"
+                               "        },\n"
+                               "        {\n"
+                               "          'type': 'file',\n"
+                               "          'name': 'missing',\n"
+                               "          'external-contents': '/missing'\n"
+                               "        },\n"
+                               "        {\n"
+                               "          'type': 'file',\n"
+                               "          'name': 'vfile',\n"
+                               "          'external-contents': '/c'\n"
+                               "        }]\n"
+                               "    }]\n"
+                               "}");
+
+  Dummy->addDirectory("/a");
+  Dummy->addRegularFile("/a/foo");
+  Dummy->addDirectory("/b");
+  Dummy->addRegularFile("/c");
+  Dummy->addRegularFile("/both/foo");
+
+  auto Redirecting = vfs::RedirectingFileSystem::create(
+							std::move(YAML), nullptr, "", nullptr, Dummy);
+
+  EXPECT_TRUE(Redirecting->exists("/dremap"));
+  EXPECT_FALSE(Redirecting->exists("/dmissing"));
+  EXPECT_FALSE(Redirecting->exists("/unknown"));
+  EXPECT_TRUE(Redirecting->exists("/both"));
+  EXPECT_TRUE(Redirecting->exists("/both/foo"));
+  EXPECT_TRUE(Redirecting->exists("/both/vfile"));
+  EXPECT_TRUE(Redirecting->exists("/vdir"));
+  EXPECT_TRUE(Redirecting->exists("/vdir/dremap"));
+  EXPECT_FALSE(Redirecting->exists("/vdir/missing"));
+  EXPECT_TRUE(Redirecting->exists("/vdir/vfile"));
+  EXPECT_FALSE(Redirecting->exists("/vdir/unknown"));
+}
+
+TEST(RedirectingFileSystemTest, ExistsFallback) {
+  IntrusiveRefCntPtr<DummyFileSystem> Dummy(new NoStatusDummyFileSystem());
+  auto YAML =
+    MemoryBuffer::getMemBuffer("{\n"
+                               "  'version': 0,\n"
+                               "  'redirecting-with': 'fallback',\n"
+                               "  'roots': [\n"
+                               "    {\n"
+                               "      'type': 'file',\n"
+                               "      'name': '/fallback',\n"
+                               "      'external-contents': '/missing',\n"
+                               "    },"
+                               "  ]\n"
+                               "}");
+
+  Dummy->addRegularFile("/fallback");
+
+  auto Redirecting = vfs::RedirectingFileSystem::create(
+							std::move(YAML), nullptr, "", nullptr, Dummy);
+
+  EXPECT_TRUE(Redirecting->exists("/fallback"));
+  EXPECT_FALSE(Redirecting->exists("/missing"));
+}
+
+TEST(RedirectingFileSystemTest, ExistsRedirectOnly) {
+  IntrusiveRefCntPtr<DummyFileSystem> Dummy(new NoStatusDummyFileSystem());
+  auto YAML =
+    MemoryBuffer::getMemBuffer("{\n"
+                               "  'version': 0,\n"
+                               "  'redirecting-with': 'redirect-only',\n"
+                               "  'roots': [\n"
+                               "    {\n"
+                               "      'type': 'file',\n"
+                               "      'name': '/vfile',\n"
+                               "      'external-contents': '/a',\n"
+                               "    },"
+                               "  ]\n"
+                               "}");
+
+  Dummy->addRegularFile("/a");
+  Dummy->addRegularFile("/b");
+
+  auto Redirecting = vfs::RedirectingFileSystem::create(
+							std::move(YAML), nullptr, "", nullptr, Dummy);
+
+  EXPECT_FALSE(Redirecting->exists("/a"));
+  EXPECT_FALSE(Redirecting->exists("/b"));
+  EXPECT_TRUE(Redirecting->exists("/vfile"));
+}

From 62624a4bfed4ea468dcb677b799aee3518a24bf7 Mon Sep 17 00:00:00 2001
From: "Harrison,Hao" <57025411+TSWorld1314@users.noreply.github.com>
Date: Sat, 13 Apr 2024 04:43:58 +0800
Subject: [PATCH 298/886] [Support] Fix the issue where the character being
 saved in Unicode causes a warning to be treated as an error in Visual Studio
 2022. (#88513)

Fix the issue where the character being saved in Unicode causes a
warning to be treated as an error in Visual Studio 2022.

![image](https://github.com/llvm/llvm-project/assets/57025411/07353525-6520-4b74-b4f5-5b3f5afc47e1)
---
 llvm/lib/Support/Parallel.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Support/Parallel.cpp b/llvm/lib/Support/Parallel.cpp
index 9b14b05b52116..af35947192c0d 100644
--- a/llvm/lib/Support/Parallel.cpp
+++ b/llvm/lib/Support/Parallel.cpp
@@ -57,7 +57,7 @@ class ThreadPoolExecutor : public Executor {
     Threads.resize(1);
     std::lock_guard<std::mutex> Lock(Mutex);
     // Use operator[] before creating the thread to avoid data race in .size()
-    // in “safe libc++” mode.
+    // in 'safe libc++' mode.
     auto &Thread0 = Threads[0];
     Thread0 = std::thread([this, S] {
       for (unsigned I = 1; I < ThreadCount; ++I) {

From 8a4b7de91dc334c828674aa2cad927c6ffb9cf37 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Fri, 12 Apr 2024 12:45:47 -0700
Subject: [PATCH 299/886] [SLP][NFC]Make TTIRef capture argument instead of
 outer declaration.

---
 llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index df891371fdf75..7b5b0efca8d3f 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -4895,14 +4895,10 @@ void BoUpSLP::reorderTopToBottom() {
   // Maps a TreeEntry to the reorder indices of external users.
   DenseMap<const TreeEntry *, SmallVector<OrdersType, 1>>
       ExternalUserReorderMap;
-  // FIXME: Workaround for syntax error reported by MSVC buildbots.
-  TargetTransformInfo &TTIRef = *TTI;
   // Find all reorderable nodes with the given VF.
   // Currently the are vectorized stores,loads,extracts + some gathering of
   // extracts.
-  for_each(VectorizableTree, [this, &TTIRef, &VFToOrderedEntries,
-                              &GathersToOrders, &ExternalUserReorderMap,
-                              &AltShufflesToOrders, &PhisToOrders](
+  for_each(VectorizableTree, [&, &TTIRef = *TTI](
                                  const std::unique_ptr<TreeEntry> &TE) {
     // Look for external users that will probably be vectorized.
     SmallVector<OrdersType, 1> ExternalUserReorderIndices =
@@ -9288,11 +9284,9 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
       (void)E;
       return TTI->getInstructionCost(VI, CostKind);
     };
-    // FIXME: Workaround for syntax error reported by MSVC buildbots.
-    TargetTransformInfo &TTIRef = *TTI;
     // Need to clear CommonCost since the final shuffle cost is included into
     // vector cost.
-    auto GetVectorCost = [&](InstructionCost) {
+    auto GetVectorCost = [&, &TTIRef = *TTI](InstructionCost) {
       // VecCost is equal to sum of the cost of creating 2 vectors
       // and the cost of creating shuffle.
       InstructionCost VecCost = 0;

From 0f82469314f34a086669dfcd190a9f89260fbee5 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Fri, 12 Apr 2024 22:07:29 +0100
Subject: [PATCH 300/886] [Passes] Run SimpleLoopUnswitch after introducing
 invariant branches. (#81271)

IndVars may be able to replace a loop dependent condition with a loop
invariant one, but loop-unswitch runs before IndVars, so the invariant
check remains in the loop.

For an example, consider a read-only loop with a bounds check:
https://godbolt.org/z/8cdj4qhbG

This patch uses a approach similar to the way extra cleanup passes are
run on demand after vectorization (added in acea6e9cfa4c4a0e8678c7).

It introduces a new ShouldRunExtraSimpleLoopUnswitch analysis marker,
which IndVars can use to indicate that extra unswitching is beneficial.

ExtraSimpleLoopUnswitchPassManager uses this analysis to determine
whether to run its passes on a loop.

Compile-time impact (geomean) ranges from +0.0% to 0.02%
https://llvm-compile-time-tracker.com/compare.php?from=138c0beb109ffe47f75a0fe8c4dc2cdabe8a6532&to=19e6e99eeb280d426907ea73a21b139ba7225627&stat=instructions%3Au

Compile-time impact (geomean) of unconditionally running
SimpleLoopUnswitch ranges from +0.05% - +0.16%

https://llvm-compile-time-tracker.com/compare.php?from=138c0beb109ffe47f75a0fe8c4dc2cdabe8a6532&to=2930dfd5accdce2e6f8d5146ae4d626add2065a2&stat=instructions:u

Unconditionally running SimpleLoopUnswitch seems to indicate that there
are multiple other scenarios where we fail to run unswitching when
opportunities remain.


Fixes https://github.com/llvm/llvm-project/issues/85551.

PR: https://github.com/llvm/llvm-project/pull/81271
---
 .../Transforms/Scalar/SimpleLoopUnswitch.h    |  35 ++++++
 .../llvm/Transforms/Utils/SimplifyIndVar.h    |  17 ++-
 llvm/lib/Passes/PassBuilderPipelines.cpp      |   7 ++
 llvm/lib/Passes/PassRegistry.def              |   3 +
 llvm/lib/Transforms/Scalar/IndVarSimplify.cpp |  17 ++-
 .../Transforms/Scalar/SimpleLoopUnswitch.cpp  |   1 +
 llvm/lib/Transforms/Utils/SimplifyIndVar.cpp  |  20 +++-
 .../AArch64/hoist-runtime-checks.ll           | 106 ++++++++++++++----
 8 files changed, 169 insertions(+), 37 deletions(-)

diff --git a/llvm/include/llvm/Transforms/Scalar/SimpleLoopUnswitch.h b/llvm/include/llvm/Transforms/Scalar/SimpleLoopUnswitch.h
index b97ee23fc0e65..a40b0a2a9ab72 100644
--- a/llvm/include/llvm/Transforms/Scalar/SimpleLoopUnswitch.h
+++ b/llvm/include/llvm/Transforms/Scalar/SimpleLoopUnswitch.h
@@ -12,6 +12,7 @@
 #include "llvm/ADT/STLFunctionalExtras.h"
 #include "llvm/Analysis/LoopAnalysisManager.h"
 #include "llvm/IR/PassManager.h"
+#include "llvm/Transforms/Scalar/LoopPassManager.h"
 
 namespace llvm {
 
@@ -20,6 +21,40 @@ class Loop;
 class StringRef;
 class raw_ostream;
 
+struct ShouldRunExtraSimpleLoopUnswitch
+    : public AnalysisInfoMixin<ShouldRunExtraSimpleLoopUnswitch> {
+  static AnalysisKey Key;
+  struct Result {
+    bool invalidate(Loop &L, const PreservedAnalyses &PA,
+                    LoopAnalysisManager::Invalidator &) {
+      // Check whether the analysis has been explicitly invalidated. Otherwise,
+      // it remains preserved.
+      auto PAC = PA.getChecker<ShouldRunExtraSimpleLoopUnswitch>();
+      return !PAC.preservedWhenStateless();
+    }
+  };
+
+  Result run(Loop &L, LoopAnalysisManager &AM,
+             LoopStandardAnalysisResults &AR) {
+    return Result();
+  }
+
+  static bool isRequired() { return true; }
+};
+
+struct ExtraSimpleLoopUnswitchPassManager : public LoopPassManager {
+  PreservedAnalyses run(Loop &L, LoopAnalysisManager &AM,
+                        LoopStandardAnalysisResults &AR, LPMUpdater &U) {
+    auto PA = PreservedAnalyses::all();
+    if (AM.getCachedResult<ShouldRunExtraSimpleLoopUnswitch>(L))
+      PA.intersect(LoopPassManager::run(L, AM, AR, U));
+    PA.abandon<ShouldRunExtraSimpleLoopUnswitch>();
+    return PA;
+  }
+
+  static bool isRequired() { return true; }
+};
+
 /// This pass transforms loops that contain branches or switches on loop-
 /// invariant conditions to have multiple loops. For example, it turns the left
 /// into the right code:
diff --git a/llvm/include/llvm/Transforms/Utils/SimplifyIndVar.h b/llvm/include/llvm/Transforms/Utils/SimplifyIndVar.h
index ff60811b61685..bd1718dd8cad7 100644
--- a/llvm/include/llvm/Transforms/Utils/SimplifyIndVar.h
+++ b/llvm/include/llvm/Transforms/Utils/SimplifyIndVar.h
@@ -15,6 +15,8 @@
 #ifndef LLVM_TRANSFORMS_UTILS_SIMPLIFYINDVAR_H
 #define LLVM_TRANSFORMS_UTILS_SIMPLIFYINDVAR_H
 
+#include <utility>
+
 namespace llvm {
 
 class Type;
@@ -46,11 +48,16 @@ class IVVisitor {
 };
 
 /// simplifyUsersOfIV - Simplify instructions that use this induction variable
-/// by using ScalarEvolution to analyze the IV's recurrence.
-bool simplifyUsersOfIV(PHINode *CurrIV, ScalarEvolution *SE, DominatorTree *DT,
-                       LoopInfo *LI, const TargetTransformInfo *TTI,
-                       SmallVectorImpl<WeakTrackingVH> &Dead,
-                       SCEVExpander &Rewriter, IVVisitor *V = nullptr);
+/// by using ScalarEvolution to analyze the IV's recurrence. Returns a pair
+/// where the first entry indicates that the function makes changes and the
+/// second entry indicates that it introduced new opportunities for loop
+/// unswitching.
+std::pair<bool, bool> simplifyUsersOfIV(PHINode *CurrIV, ScalarEvolution *SE,
+                                        DominatorTree *DT, LoopInfo *LI,
+                                        const TargetTransformInfo *TTI,
+                                        SmallVectorImpl<WeakTrackingVH> &Dead,
+                                        SCEVExpander &Rewriter,
+                                        IVVisitor *V = nullptr);
 
 /// SimplifyLoopIVs - Simplify users of induction variables within this
 /// loop. This does not actually change or add IVs.
diff --git a/llvm/lib/Passes/PassBuilderPipelines.cpp b/llvm/lib/Passes/PassBuilderPipelines.cpp
index cb892e30c4a0b..3bb2ce0ae3460 100644
--- a/llvm/lib/Passes/PassBuilderPipelines.cpp
+++ b/llvm/lib/Passes/PassBuilderPipelines.cpp
@@ -651,6 +651,13 @@ PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
   LPM2.addPass(LoopIdiomRecognizePass());
   LPM2.addPass(IndVarSimplifyPass());
 
+  {
+    ExtraSimpleLoopUnswitchPassManager ExtraPasses;
+    ExtraPasses.addPass(SimpleLoopUnswitchPass(/* NonTrivial */ Level ==
+                                               OptimizationLevel::O3));
+    LPM2.addPass(std::move(ExtraPasses));
+  }
+
   invokeLateLoopOptimizationsEPCallbacks(LPM2, Level);
 
   LPM2.addPass(LoopDeletionPass());
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index d15f58d7adfae..2fbc7f7d88ba3 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -587,6 +587,9 @@ LOOP_ANALYSIS("ddg", DDGAnalysis())
 LOOP_ANALYSIS("iv-users", IVUsersAnalysis())
 LOOP_ANALYSIS("no-op-loop", NoOpLoopAnalysis())
 LOOP_ANALYSIS("pass-instrumentation", PassInstrumentationAnalysis(PIC))
+LOOP_ANALYSIS("should-run-extra-simple-loop-unswitch",
+                  ShouldRunExtraSimpleLoopUnswitch())
+
 #undef LOOP_ANALYSIS
 
 #ifndef LOOP_PASS
diff --git a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
index 38104afa7f783..ba392e187b8be 100644
--- a/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
+++ b/llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
@@ -70,6 +70,7 @@
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/Transforms/Scalar/SimpleLoopUnswitch.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
 #include "llvm/Transforms/Utils/LoopUtils.h"
@@ -137,6 +138,8 @@ class IndVarSimplify {
   SmallVector<WeakTrackingVH, 16> DeadInsts;
   bool WidenIndVars;
 
+  bool RunUnswitching = false;
+
   bool handleFloatingPointIV(Loop *L, PHINode *PH);
   bool rewriteNonIntegerIVs(Loop *L);
 
@@ -170,6 +173,8 @@ class IndVarSimplify {
   }
 
   bool run(Loop *L);
+
+  bool runUnswitching() const { return RunUnswitching; }
 };
 
 } // end anonymous namespace
@@ -615,9 +620,11 @@ bool IndVarSimplify::simplifyAndExtend(Loop *L,
       // Information about sign/zero extensions of CurrIV.
       IndVarSimplifyVisitor Visitor(CurrIV, SE, TTI, DT);
 
-      Changed |= simplifyUsersOfIV(CurrIV, SE, DT, LI, TTI, DeadInsts, Rewriter,
-                                   &Visitor);
+      const auto &[C, U] = simplifyUsersOfIV(CurrIV, SE, DT, LI, TTI, DeadInsts,
+                                             Rewriter, &Visitor);
 
+      Changed |= C;
+      RunUnswitching |= U;
       if (Visitor.WI.WidestNativeType) {
         WideIVs.push_back(Visitor.WI);
       }
@@ -1874,6 +1881,7 @@ bool IndVarSimplify::predicateLoopExits(Loop *L, SCEVExpander &Rewriter) {
     if (OldCond->use_empty())
       DeadInsts.emplace_back(OldCond);
     Changed = true;
+    RunUnswitching = true;
   }
 
   return Changed;
@@ -2059,6 +2067,11 @@ PreservedAnalyses IndVarSimplifyPass::run(Loop &L, LoopAnalysisManager &AM,
 
   auto PA = getLoopPassPreservedAnalyses();
   PA.preserveSet<CFGAnalyses>();
+  if (IVS.runUnswitching()) {
+    AM.getResult<ShouldRunExtraSimpleLoopUnswitch>(L, AR);
+    PA.preserve<ShouldRunExtraSimpleLoopUnswitch>();
+  }
+
   if (AR.MSSA)
     PA.preserve<MemorySSAAnalysis>();
   return PA;
diff --git a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
index 64b850a6a5c2f..d763b1ee0aa1b 100644
--- a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
+++ b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
@@ -133,6 +133,7 @@ static cl::opt<unsigned> InjectInvariantConditionHotnesThreshold(
                          "not-taken 1/<this option> times or less."),
     cl::init(16));
 
+AnalysisKey ShouldRunExtraSimpleLoopUnswitch::Key;
 namespace {
 struct CompareDesc {
   BranchInst *Term;
diff --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
index f6d72265305d7..440fe0790d795 100644
--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
@@ -60,6 +60,7 @@ namespace {
     SmallVectorImpl<WeakTrackingVH> &DeadInsts;
 
     bool Changed = false;
+    bool RunUnswitching = false;
 
   public:
     SimplifyIndvar(Loop *Loop, ScalarEvolution *SE, DominatorTree *DT,
@@ -72,6 +73,7 @@ namespace {
     }
 
     bool hasChanged() const { return Changed; }
+    bool runUnswitching() const { return RunUnswitching; }
 
     /// Iteratively perform simplification on a worklist of users of the
     /// specified induction variable. This is the top-level driver that applies
@@ -233,6 +235,7 @@ bool SimplifyIndvar::makeIVComparisonInvariant(ICmpInst *ICmp,
   ICmp->setPredicate(InvariantPredicate);
   ICmp->setOperand(0, NewLHS);
   ICmp->setOperand(1, NewRHS);
+  RunUnswitching = true;
   return true;
 }
 
@@ -993,14 +996,18 @@ void IVVisitor::anchor() { }
 
 /// Simplify instructions that use this induction variable
 /// by using ScalarEvolution to analyze the IV's recurrence.
-bool simplifyUsersOfIV(PHINode *CurrIV, ScalarEvolution *SE, DominatorTree *DT,
-                       LoopInfo *LI, const TargetTransformInfo *TTI,
-                       SmallVectorImpl<WeakTrackingVH> &Dead,
-                       SCEVExpander &Rewriter, IVVisitor *V) {
+///  Returns a pair where the first entry indicates that the function makes
+///  changes and the second entry indicates that it introduced new opportunities
+///  for loop unswitching.
+std::pair<bool, bool> simplifyUsersOfIV(PHINode *CurrIV, ScalarEvolution *SE,
+                                        DominatorTree *DT, LoopInfo *LI,
+                                        const TargetTransformInfo *TTI,
+                                        SmallVectorImpl<WeakTrackingVH> &Dead,
+                                        SCEVExpander &Rewriter, IVVisitor *V) {
   SimplifyIndvar SIV(LI->getLoopFor(CurrIV->getParent()), SE, DT, LI, TTI,
                      Rewriter, Dead);
   SIV.simplifyUsers(CurrIV, V);
-  return SIV.hasChanged();
+  return {SIV.hasChanged(), SIV.runUnswitching()};
 }
 
 /// Simplify users of induction variables within this
@@ -1014,8 +1021,9 @@ bool simplifyLoopIVs(Loop *L, ScalarEvolution *SE, DominatorTree *DT,
 #endif
   bool Changed = false;
   for (BasicBlock::iterator I = L->getHeader()->begin(); isa<PHINode>(I); ++I) {
-    Changed |=
+    const auto &[C, _] =
         simplifyUsersOfIV(cast<PHINode>(I), SE, DT, LI, TTI, Dead, Rewriter);
+    Changed |= C;
   }
   return Changed;
 }
diff --git a/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll b/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
index a140e17a0dd15..55dd28b70170b 100644
--- a/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
+++ b/llvm/test/Transforms/PhaseOrdering/AArch64/hoist-runtime-checks.ll
@@ -14,24 +14,50 @@ define i32 @read_only_loop_with_runtime_check(ptr noundef %array, i32 noundef %c
 ; CHECK-NEXT:    [[TMP0:%.*]] = zext i32 [[N]] to i64
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
 ; CHECK-NEXT:    [[DOTNOT_NOT:%.*]] = icmp ult i32 [[TMP1]], [[COUNT]]
+; CHECK-NEXT:    br i1 [[DOTNOT_NOT]], label [[FOR_BODY_PREHEADER10:%.*]], label [[IF_THEN:%.*]]
+; CHECK:       for.body.preheader10:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 8
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY_PREHEADER13:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 4294967288
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI11:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[ARRAY]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i64 16
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP4]] = add <4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP5]] = add <4 x i32> [[WIDE_LOAD12]], [[VEC_PHI11]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP5]], [[TMP4]]
+; CHECK-NEXT:    [[TMP7:%.*]] = tail call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY_PREHEADER13]]
+; CHECK:       for.body.preheader13:
+; CHECK-NEXT:    [[INDVARS_IV_PH:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER10]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[SUM_07_PH:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER10]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
-; CHECK-NEXT:    [[SUM_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[IF_END:%.*]] ]
+; CHECK-NEXT:    [[SUM_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[SUM_0_LCSSA]]
 ; CHECK:       for.body:
-; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[IF_END]] ]
-; CHECK-NEXT:    [[SUM_07:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[ADD]], [[IF_END]] ]
-; CHECK-NEXT:    br i1 [[DOTNOT_NOT]], label [[IF_END]], label [[IF_THEN:%.*]]
-; CHECK:       if.then:
-; CHECK-NEXT:    tail call void @llvm.trap()
-; CHECK-NEXT:    unreachable
-; CHECK:       if.end:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[INDVARS_IV_PH]], [[FOR_BODY_PREHEADER13]] ]
+; CHECK-NEXT:    [[SUM_07:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[SUM_07_PH]], [[FOR_BODY_PREHEADER13]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[ARRAY]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[ADD]] = add nsw i32 [[TMP2]], [[SUM_07]]
+; CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[ADD]] = add nsw i32 [[TMP8]], [[SUM_07]]
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[TMP0]]
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       if.then:
+; CHECK-NEXT:    tail call void @llvm.trap()
+; CHECK-NEXT:    unreachable
 ;
 entry:
   %array.addr = alloca ptr, align 8
@@ -103,24 +129,50 @@ define dso_local noundef i32 @sum_prefix_with_sum(ptr %s.coerce0, i64 %s.coerce1
 ; CHECK:       for.body.preheader:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
 ; CHECK-NEXT:    [[DOTNOT_NOT:%.*]] = icmp ult i64 [[TMP0]], [[S_COERCE1]]
+; CHECK-NEXT:    br i1 [[DOTNOT_NOT]], label [[ENTRY:%.*]], label [[COND_FALSE_I:%.*]], !prof [[PROF4:![0-9]+]]
+; CHECK:       for.body.preheader8:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 8
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY_PREHEADER11:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -8
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI9:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[S_COERCE0]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i64 16
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP3]] = add <4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP4]] = add <4 x i32> [[WIDE_LOAD10]], [[VEC_PHI9]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[SPAN_CHECKED_ACCESS_EXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP4]], [[TMP3]]
+; CHECK-NEXT:    [[ADD:%.*]] = tail call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY_PREHEADER11]]
+; CHECK:       for.body.preheader11:
+; CHECK-NEXT:    [[I_07_PH:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[N_VEC]], [[SPAN_CHECKED_ACCESS_EXIT]] ]
+; CHECK-NEXT:    [[RET_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[ADD]], [[SPAN_CHECKED_ACCESS_EXIT]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
 ; CHECK:       for.cond.cleanup:
-; CHECK-NEXT:    [[RET_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[SPAN_CHECKED_ACCESS_EXIT:%.*]] ]
-; CHECK-NEXT:    ret i32 [[RET_0_LCSSA]]
+; CHECK-NEXT:    [[RET_0_LCSSA1:%.*]] = phi i32 [ 0, [[ENTRY1:%.*]] ], [ [[ADD]], [[SPAN_CHECKED_ACCESS_EXIT]] ], [ [[ADD1:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    ret i32 [[RET_0_LCSSA1]]
 ; CHECK:       for.body:
-; CHECK-NEXT:    [[I_07:%.*]] = phi i64 [ [[INC:%.*]], [[SPAN_CHECKED_ACCESS_EXIT]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-NEXT:    [[RET_06:%.*]] = phi i32 [ [[ADD]], [[SPAN_CHECKED_ACCESS_EXIT]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[DOTNOT_NOT]], label [[SPAN_CHECKED_ACCESS_EXIT]], label [[COND_FALSE_I:%.*]], !prof [[PROF0:![0-9]+]]
-; CHECK:       cond.false.i:
-; CHECK-NEXT:    tail call void @llvm.trap()
-; CHECK-NEXT:    unreachable
-; CHECK:       span_checked_access.exit:
+; CHECK-NEXT:    [[I_07:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_BODY1]] ], [ [[I_07_PH]], [[FOR_BODY_PREHEADER11]] ]
+; CHECK-NEXT:    [[RET_06:%.*]] = phi i32 [ [[ADD1]], [[FOR_BODY1]] ], [ [[RET_0_LCSSA]], [[FOR_BODY_PREHEADER11]] ]
 ; CHECK-NEXT:    [[ARRAYIDX_I:%.*]] = getelementptr inbounds i32, ptr [[S_COERCE0]], i64 [[I_07]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX_I]], align 4
-; CHECK-NEXT:    [[ADD]] = add nsw i32 [[TMP7]], [[RET_06]]
+; CHECK-NEXT:    [[ADD1]] = add nsw i32 [[TMP7]], [[RET_06]]
 ; CHECK-NEXT:    [[INC]] = add nuw i64 [[I_07]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK:       cond.false.i:
+; CHECK-NEXT:    tail call void @llvm.trap()
+; CHECK-NEXT:    unreachable
 ;
 entry:
   %s = alloca %"class.std::__1::span", align 8
@@ -176,7 +228,7 @@ define hidden noundef nonnull align 4 dereferenceable(4) ptr @span_checked_acces
 ; CHECK-NEXT:    [[__SIZE__I:%.*]] = getelementptr inbounds i8, ptr [[THIS]], i64 8
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[__SIZE__I]], align 8
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP0]], [[__IDX]]
-; CHECK-NEXT:    br i1 [[CMP]], label [[COND_END:%.*]], label [[COND_FALSE:%.*]], !prof [[PROF0]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[COND_END:%.*]], label [[COND_FALSE:%.*]], !prof [[PROF4]]
 ; CHECK:       cond.false:
 ; CHECK-NEXT:    tail call void @llvm.trap()
 ; CHECK-NEXT:    unreachable
@@ -237,5 +289,11 @@ declare void @llvm.trap()
 
 declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture)
 ;.
-; CHECK: [[PROF0]] = !{!"branch_weights", i32 2000, i32 1}
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; CHECK: [[PROF4]] = !{!"branch_weights", i32 2000, i32 1}
+; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]], [[META2]]}
+; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META2]], [[META1]]}
 ;.

From c6cd4608c8c4a5356ae41ef05e67df3dc5d9500b Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze@braunis.de>
Date: Fri, 12 Apr 2024 14:10:07 -0700
Subject: [PATCH 301/886] IRVerifier: Allow GlobalValue as
 llvm.threadlocal.address operand (#88321)

Loosen `llvm.threadlocal.address` verifier checks to allow any
`GlobalValue` with `isThreadLocal()` set to true.
---
 llvm/docs/LangRef.rst                  |  3 ++-
 llvm/lib/IR/Verifier.cpp               |  8 +++---
 llvm/test/Verifier/threadlocal-pass.ll | 13 ++++++++++
 llvm/test/Verifier/threadlocal.ll      | 34 ++++++++++++++++++++++++++
 4 files changed, 53 insertions(+), 5 deletions(-)
 create mode 100644 llvm/test/Verifier/threadlocal-pass.ll
 create mode 100644 llvm/test/Verifier/threadlocal.ll

diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index f6ada292b93b1..9592929d79feb 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -28130,7 +28130,8 @@ Syntax:
 Arguments:
 """"""""""
 
-The first argument is a thread local :ref:`global variable <globalvars>`.
+The `llvm.threadlocal.address` intrinsic requires a global value argument (a
+:ref:`global variable <globalvars>` or alias) that is thread local.
 
 Semantics:
 """"""""""
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 4092f0cb12ffe..25cb99f0f9bd8 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -6226,10 +6226,10 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
   }
   case Intrinsic::threadlocal_address: {
     const Value &Arg0 = *Call.getArgOperand(0);
-    Check(isa<GlobalVariable>(Arg0),
-          "llvm.threadlocal.address first argument must be a GlobalVariable");
-    Check(cast<GlobalVariable>(Arg0).isThreadLocal(),
-          "llvm.threadlocal.address operand isThreadLocal() must no be false");
+    Check(isa<GlobalValue>(Arg0),
+          "llvm.threadlocal.address first argument must be a GlobalValue");
+    Check(cast<GlobalValue>(Arg0).isThreadLocal(),
+          "llvm.threadlocal.address operand isThreadLocal() must be true");
     break;
   }
   };
diff --git a/llvm/test/Verifier/threadlocal-pass.ll b/llvm/test/Verifier/threadlocal-pass.ll
new file mode 100644
index 0000000000000..0cee4af824397
--- /dev/null
+++ b/llvm/test/Verifier/threadlocal-pass.ll
@@ -0,0 +1,13 @@
+; RUN: opt -passes=verify -S < %s | FileCheck %s
+
+@var = thread_local global i32 0
+@alias = thread_local alias i32, ptr @var
+
+; CHECK-LABEL: @should_pass
+define void @should_pass() {
+  %p0 = call ptr @llvm.threadlocal.address(ptr @var)
+  store i32 42, ptr %p0, align 4
+  %p1 = call ptr @llvm.threadlocal.address(ptr @alias)
+  store i32 13, ptr %p1, align 4
+  ret void
+}
diff --git a/llvm/test/Verifier/threadlocal.ll b/llvm/test/Verifier/threadlocal.ll
new file mode 100644
index 0000000000000..8af23c361cc4b
--- /dev/null
+++ b/llvm/test/Verifier/threadlocal.ll
@@ -0,0 +1,34 @@
+; RUN: not llvm-as %s -o /dev/null 2>&1 | FileCheck %s
+
+@var = global i32 0
+@tlsvar = thread_local addrspace(1) global i32 0
+
+define void @fail0(ptr %arg) {
+; CHECK: llvm.threadlocal.address first argument must be a GlobalValue
+  %p0 = call ptr @llvm.threadlocal.address(ptr %arg)
+  store i32 42, ptr %p0, align 4
+  ret void
+}
+
+define void @fail1() {
+; CHECK: llvm.threadlocal.address first argument must be a GlobalValue
+  %p0 = call ptr @llvm.threadlocal.address.p0(ptr addrspacecast (ptr addrspace(1) @tlsvar to ptr addrspace(0)))
+  store i32 42, ptr %p0, align 4
+  ret void
+}
+
+
+
+define void @fail2() {
+; CHECK: llvm.threadlocal.address operand isThreadLocal() must be true
+  %p0 = call ptr @llvm.threadlocal.address(ptr @var)
+  store i32 42, ptr %p0, align 4
+  ret void
+}
+
+define void @fail3() {
+; CHECK: llvm.threadlocal.address operand isThreadLocal() must be true
+  %p0 = call ptr @llvm.threadlocal.address(ptr @fail2)
+  store i32 42, ptr %p0, align 4
+  ret void
+}

From daa88364df15683bfa0e58d62cc96d33cd49d34b Mon Sep 17 00:00:00 2001
From: Erich Keane <ekeane@nvidia.com>
Date: Fri, 12 Apr 2024 14:13:31 -0700
Subject: [PATCH 302/886] [OpenACC] Implement 'if' clause for Compute
 Constructs (#88411)

Like with the 'default' clause, this is being applied to only Compute
Constructs for now. The 'if' clause takes a condition expression which
is used as a runtime value.

This is not a particularly complex semantic implementation, as there
isn't much to this clause, other than its interactions with 'self',
  which will be managed in the patch to implement that.
---
 clang/include/clang/AST/ASTNodeTraverser.h    |   3 +-
 clang/include/clang/AST/OpenACCClause.h       |  81 +++++++++++-
 clang/include/clang/Basic/OpenACCClauses.def  |  21 +++
 clang/include/clang/Parse/Parser.h            |   5 +
 clang/include/clang/Sema/SemaOpenACC.h        |  28 +++-
 clang/lib/AST/OpenACCClause.cpp               |  39 ++++++
 clang/lib/AST/StmtProfile.cpp                 |  19 ++-
 clang/lib/AST/TextNodeDumper.cpp              |   5 +
 clang/lib/Parse/ParseOpenACC.cpp              |  32 +++--
 clang/lib/Sema/SemaOpenACC.cpp                |  73 +++++++++--
 clang/lib/Sema/TreeTransform.h                |  14 ++
 clang/lib/Serialization/ASTReader.cpp         |   7 +-
 clang/lib/Serialization/ASTWriter.cpp         |   7 +-
 clang/test/ParserOpenACC/parse-clauses.c      |   2 -
 .../compute-construct-clause-ast.cpp          | 120 +++++++++++++++++-
 .../SemaOpenACC/compute-construct-if-clause.c |  62 +++++++++
 .../compute-construct-if-clause.cpp           |  33 +++++
 clang/tools/libclang/CIndex.cpp               |  43 +++++++
 18 files changed, 557 insertions(+), 37 deletions(-)
 create mode 100644 clang/include/clang/Basic/OpenACCClauses.def
 create mode 100644 clang/test/SemaOpenACC/compute-construct-if-clause.c
 create mode 100644 clang/test/SemaOpenACC/compute-construct-if-clause.cpp

diff --git a/clang/include/clang/AST/ASTNodeTraverser.h b/clang/include/clang/AST/ASTNodeTraverser.h
index a6e3b05b00596..f5c47d8a7c211 100644
--- a/clang/include/clang/AST/ASTNodeTraverser.h
+++ b/clang/include/clang/AST/ASTNodeTraverser.h
@@ -243,7 +243,8 @@ class ASTNodeTraverser
   void Visit(const OpenACCClause *C) {
     getNodeDelegate().AddChild([=] {
       getNodeDelegate().Visit(C);
-      // TODO OpenACC: Switch on clauses that have children, and add them.
+      for (const auto *S : C->children())
+        Visit(S);
     });
   }
 
diff --git a/clang/include/clang/AST/OpenACCClause.h b/clang/include/clang/AST/OpenACCClause.h
index 27e4e1a12c983..712ad46c0f238 100644
--- a/clang/include/clang/AST/OpenACCClause.h
+++ b/clang/include/clang/AST/OpenACCClause.h
@@ -14,6 +14,7 @@
 #ifndef LLVM_CLANG_AST_OPENACCCLAUSE_H
 #define LLVM_CLANG_AST_OPENACCCLAUSE_H
 #include "clang/AST/ASTContext.h"
+#include "clang/AST/StmtIterator.h"
 #include "clang/Basic/OpenACCKinds.h"
 
 namespace clang {
@@ -34,6 +35,17 @@ class OpenACCClause {
 
   static bool classof(const OpenACCClause *) { return true; }
 
+  using child_iterator = StmtIterator;
+  using const_child_iterator = ConstStmtIterator;
+  using child_range = llvm::iterator_range<child_iterator>;
+  using const_child_range = llvm::iterator_range<const_child_iterator>;
+
+  child_range children();
+  const_child_range children() const {
+    auto Children = const_cast<OpenACCClause *>(this)->children();
+    return const_child_range(Children.begin(), Children.end());
+  }
+
   virtual ~OpenACCClause() = default;
 };
 
@@ -49,6 +61,13 @@ class OpenACCClauseWithParams : public OpenACCClause {
 
 public:
   SourceLocation getLParenLoc() const { return LParenLoc; }
+
+  child_range children() {
+    return child_range(child_iterator(), child_iterator());
+  }
+  const_child_range children() const {
+    return const_child_range(const_child_iterator(), const_child_iterator());
+  }
 };
 
 /// A 'default' clause, has the optional 'none' or 'present' argument.
@@ -81,6 +100,51 @@ class OpenACCDefaultClause : public OpenACCClauseWithParams {
                                       SourceLocation EndLoc);
 };
 
+/// Represents one of the handful of classes that has an optional/required
+/// 'condition' expression as an argument.
+class OpenACCClauseWithCondition : public OpenACCClauseWithParams {
+  Expr *ConditionExpr = nullptr;
+
+protected:
+  OpenACCClauseWithCondition(OpenACCClauseKind K, SourceLocation BeginLoc,
+                             SourceLocation LParenLoc, Expr *ConditionExpr,
+                             SourceLocation EndLoc)
+      : OpenACCClauseWithParams(K, BeginLoc, LParenLoc, EndLoc),
+        ConditionExpr(ConditionExpr) {}
+
+public:
+  bool hasConditionExpr() const { return ConditionExpr; }
+  const Expr *getConditionExpr() const { return ConditionExpr; }
+  Expr *getConditionExpr() { return ConditionExpr; }
+
+  child_range children() {
+    if (ConditionExpr)
+      return child_range(reinterpret_cast<Stmt **>(&ConditionExpr),
+                         reinterpret_cast<Stmt **>(&ConditionExpr + 1));
+    return child_range(child_iterator(), child_iterator());
+  }
+
+  const_child_range children() const {
+    if (ConditionExpr)
+      return const_child_range(
+          reinterpret_cast<Stmt *const *>(&ConditionExpr),
+          reinterpret_cast<Stmt *const *>(&ConditionExpr + 1));
+    return const_child_range(const_child_iterator(), const_child_iterator());
+  }
+};
+
+/// An 'if' clause, which has a required condition expression.
+class OpenACCIfClause : public OpenACCClauseWithCondition {
+protected:
+  OpenACCIfClause(SourceLocation BeginLoc, SourceLocation LParenLoc,
+                  Expr *ConditionExpr, SourceLocation EndLoc);
+
+public:
+  static OpenACCIfClause *Create(const ASTContext &C, SourceLocation BeginLoc,
+                                 SourceLocation LParenLoc, Expr *ConditionExpr,
+                                 SourceLocation EndLoc);
+};
+
 template <class Impl> class OpenACCClauseVisitor {
   Impl &getDerived() { return static_cast<Impl &>(*this); }
 
@@ -98,6 +162,9 @@ template <class Impl> class OpenACCClauseVisitor {
     case OpenACCClauseKind::Default:
       VisitOpenACCDefaultClause(*cast<OpenACCDefaultClause>(C));
       return;
+    case OpenACCClauseKind::If:
+      VisitOpenACCIfClause(*cast<OpenACCIfClause>(C));
+      return;
     case OpenACCClauseKind::Finalize:
     case OpenACCClauseKind::IfPresent:
     case OpenACCClauseKind::Seq:
@@ -106,7 +173,6 @@ template <class Impl> class OpenACCClauseVisitor {
     case OpenACCClauseKind::Worker:
     case OpenACCClauseKind::Vector:
     case OpenACCClauseKind::NoHost:
-    case OpenACCClauseKind::If:
     case OpenACCClauseKind::Self:
     case OpenACCClauseKind::Copy:
     case OpenACCClauseKind::UseDevice:
@@ -145,9 +211,13 @@ template <class Impl> class OpenACCClauseVisitor {
     llvm_unreachable("Invalid Clause kind");
   }
 
-  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause) {
-    return getDerived().VisitOpenACCDefaultClause(Clause);
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
+      const OpenACC##CLAUSE_NAME##Clause &Clause) {                            \
+    return getDerived().VisitOpenACC##CLAUSE_NAME##Clause(Clause);             \
   }
+
+#include "clang/Basic/OpenACCClauses.def"
 };
 
 class OpenACCClausePrinter final
@@ -165,7 +235,10 @@ class OpenACCClausePrinter final
   }
   OpenACCClausePrinter(raw_ostream &OS) : OS(OS) {}
 
-  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause);
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
+      const OpenACC##CLAUSE_NAME##Clause &Clause);
+#include "clang/Basic/OpenACCClauses.def"
 };
 
 } // namespace clang
diff --git a/clang/include/clang/Basic/OpenACCClauses.def b/clang/include/clang/Basic/OpenACCClauses.def
new file mode 100644
index 0000000000000..7fd2720e02ce2
--- /dev/null
+++ b/clang/include/clang/Basic/OpenACCClauses.def
@@ -0,0 +1,21 @@
+//===-- OpenACCClauses.def - List of implemented OpenACC Clauses -- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines a list of currently implemented OpenACC Clauses (and
+// eventually, the entire list) in a way that makes generating 'visitor' and
+// other lists easier.
+//
+// The primary macro is a single-argument version taking the name of the Clause
+// as used in Clang source (so `Default` instead of `default`).
+//
+// VISIT_CLAUSE(CLAUSE_NAME)
+
+VISIT_CLAUSE(Default)
+VISIT_CLAUSE(If)
+
+#undef VISIT_CLAUSE
diff --git a/clang/include/clang/Parse/Parser.h b/clang/include/clang/Parse/Parser.h
index 3a055c10ffb38..db90d38c7dacf 100644
--- a/clang/include/clang/Parse/Parser.h
+++ b/clang/include/clang/Parse/Parser.h
@@ -3611,6 +3611,9 @@ class Parser : public CodeCompletionHandler {
   OpenACCClauseParseResult OpenACCCannotContinue();
   OpenACCClauseParseResult OpenACCSuccess(OpenACCClause *Clause);
 
+  using OpenACCConditionExprParseResult =
+      std::pair<ExprResult, OpenACCParseCanContinue>;
+
   /// Parses the OpenACC directive (the entire pragma) including the clause
   /// list, but does not produce the main AST node.
   OpenACCDirectiveParseInfo ParseOpenACCDirective();
@@ -3657,6 +3660,8 @@ class Parser : public CodeCompletionHandler {
   bool ParseOpenACCGangArgList();
   /// Parses a 'gang-arg', used for the 'gang' clause.
   bool ParseOpenACCGangArg();
+  /// Parses a 'condition' expr, ensuring it results in a
+  ExprResult ParseOpenACCConditionExpr();
 
 private:
   //===--------------------------------------------------------------------===//
diff --git a/clang/include/clang/Sema/SemaOpenACC.h b/clang/include/clang/Sema/SemaOpenACC.h
index 27aaee164a288..c1fe0f5b9c0f6 100644
--- a/clang/include/clang/Sema/SemaOpenACC.h
+++ b/clang/include/clang/Sema/SemaOpenACC.h
@@ -40,7 +40,11 @@ class SemaOpenACC : public SemaBase {
       OpenACCDefaultClauseKind DefaultClauseKind;
     };
 
-    std::variant<DefaultDetails> Details;
+    struct ConditionDetails {
+      Expr *ConditionExpr;
+    };
+
+    std::variant<DefaultDetails, ConditionDetails> Details;
 
   public:
     OpenACCParsedClause(OpenACCDirectiveKind DirKind,
@@ -63,6 +67,16 @@ class SemaOpenACC : public SemaBase {
       return std::get<DefaultDetails>(Details).DefaultClauseKind;
     }
 
+    const Expr *getConditionExpr() const {
+      return const_cast<OpenACCParsedClause *>(this)->getConditionExpr();
+    }
+
+    Expr *getConditionExpr() {
+      assert(ClauseKind == OpenACCClauseKind::If &&
+             "Parsed clause kind does not have a condition expr");
+      return std::get<ConditionDetails>(Details).ConditionExpr;
+    }
+
     void setLParenLoc(SourceLocation EndLoc) { LParenLoc = EndLoc; }
     void setEndLoc(SourceLocation EndLoc) { ClauseRange.setEnd(EndLoc); }
 
@@ -71,6 +85,18 @@ class SemaOpenACC : public SemaBase {
              "Parsed clause is not a default clause");
       Details = DefaultDetails{DefKind};
     }
+
+    void setConditionDetails(Expr *ConditionExpr) {
+      assert(ClauseKind == OpenACCClauseKind::If &&
+             "Parsed clause kind does not have a condition expr");
+      // In C++ we can count on this being a 'bool', but in C this gets left as
+      // some sort of scalar that codegen will have to take care of converting.
+      assert((!ConditionExpr || ConditionExpr->isInstantiationDependent() ||
+              ConditionExpr->getType()->isScalarType()) &&
+             "Condition expression type not scalar/dependent");
+
+      Details = ConditionDetails{ConditionExpr};
+    }
   };
 
   SemaOpenACC(Sema &S);
diff --git a/clang/lib/AST/OpenACCClause.cpp b/clang/lib/AST/OpenACCClause.cpp
index c83128b60e3ac..2bc8d14fa129b 100644
--- a/clang/lib/AST/OpenACCClause.cpp
+++ b/clang/lib/AST/OpenACCClause.cpp
@@ -13,6 +13,7 @@
 
 #include "clang/AST/OpenACCClause.h"
 #include "clang/AST/ASTContext.h"
+#include "clang/AST/Expr.h"
 
 using namespace clang;
 
@@ -27,6 +28,40 @@ OpenACCDefaultClause *OpenACCDefaultClause::Create(const ASTContext &C,
   return new (Mem) OpenACCDefaultClause(K, BeginLoc, LParenLoc, EndLoc);
 }
 
+OpenACCIfClause *OpenACCIfClause::Create(const ASTContext &C,
+                                         SourceLocation BeginLoc,
+                                         SourceLocation LParenLoc,
+                                         Expr *ConditionExpr,
+                                         SourceLocation EndLoc) {
+  void *Mem = C.Allocate(sizeof(OpenACCIfClause), alignof(OpenACCIfClause));
+  return new (Mem) OpenACCIfClause(BeginLoc, LParenLoc, ConditionExpr, EndLoc);
+}
+
+OpenACCIfClause::OpenACCIfClause(SourceLocation BeginLoc,
+                                 SourceLocation LParenLoc, Expr *ConditionExpr,
+                                 SourceLocation EndLoc)
+    : OpenACCClauseWithCondition(OpenACCClauseKind::If, BeginLoc, LParenLoc,
+                                 ConditionExpr, EndLoc) {
+  assert(ConditionExpr && "if clause requires condition expr");
+  assert((ConditionExpr->isInstantiationDependent() ||
+          ConditionExpr->getType()->isScalarType()) &&
+         "Condition expression type not scalar/dependent");
+}
+
+OpenACCClause::child_range OpenACCClause::children() {
+  switch (getClauseKind()) {
+  default:
+    assert(false && "Clause children function not implemented");
+    break;
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  case OpenACCClauseKind::CLAUSE_NAME:                                         \
+    return cast<OpenACC##CLAUSE_NAME##Clause>(this)->children();
+
+#include "clang/Basic/OpenACCClauses.def"
+  }
+  return child_range(child_iterator(), child_iterator());
+}
+
 //===----------------------------------------------------------------------===//
 //  OpenACC clauses printing methods
 //===----------------------------------------------------------------------===//
@@ -34,3 +69,7 @@ void OpenACCClausePrinter::VisitOpenACCDefaultClause(
     const OpenACCDefaultClause &C) {
   OS << "default(" << C.getDefaultClauseKind() << ")";
 }
+
+void OpenACCClausePrinter::VisitOpenACCIfClause(const OpenACCIfClause &C) {
+  OS << "if(" << C.getConditionExpr() << ")";
+}
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 01e1d1cc8289b..44b635bfb8433 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2445,9 +2445,10 @@ void StmtProfiler::VisitTemplateArgument(const TemplateArgument &Arg) {
 namespace {
 class OpenACCClauseProfiler
     : public OpenACCClauseVisitor<OpenACCClauseProfiler> {
+  StmtProfiler &Profiler;
 
 public:
-  OpenACCClauseProfiler() = default;
+  OpenACCClauseProfiler(StmtProfiler &P) : Profiler(P) {}
 
   void VisitOpenACCClauseList(ArrayRef<const OpenACCClause *> Clauses) {
     for (const OpenACCClause *Clause : Clauses) {
@@ -2456,12 +2457,24 @@ class OpenACCClauseProfiler
       Visit(Clause);
     }
   }
-  void VisitOpenACCDefaultClause(const OpenACCDefaultClause &Clause);
+
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
+      const OpenACC##CLAUSE_NAME##Clause &Clause);
+
+#include "clang/Basic/OpenACCClauses.def"
 };
 
 /// Nothing to do here, there are no sub-statements.
 void OpenACCClauseProfiler::VisitOpenACCDefaultClause(
     const OpenACCDefaultClause &Clause) {}
+
+void OpenACCClauseProfiler::VisitOpenACCIfClause(
+    const OpenACCIfClause &Clause) {
+  assert(Clause.hasConditionExpr() &&
+         "if clause requires a valid condition expr");
+  Profiler.VisitStmt(Clause.getConditionExpr());
+}
 } // namespace
 
 void StmtProfiler::VisitOpenACCComputeConstruct(
@@ -2469,7 +2482,7 @@ void StmtProfiler::VisitOpenACCComputeConstruct(
   // VisitStmt handles children, so the AssociatedStmt is handled.
   VisitStmt(S);
 
-  OpenACCClauseProfiler P;
+  OpenACCClauseProfiler P{*this};
   P.VisitOpenACCClauseList(S->clauses());
 }
 
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index f4b2df7730ae4..0973e63183bd7 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -397,6 +397,11 @@ void TextNodeDumper::Visit(const OpenACCClause *C) {
     case OpenACCClauseKind::Default:
       OS << '(' << cast<OpenACCDefaultClause>(C)->getDefaultClauseKind() << ')';
       break;
+    case OpenACCClauseKind::If:
+      // The condition expression will be printed as a part of the 'children',
+      // but print 'clause' here so it is clear what is happening from the dump.
+      OS << " clause";
+      break;
     default:
       // Nothing to do here.
       break;
diff --git a/clang/lib/Parse/ParseOpenACC.cpp b/clang/lib/Parse/ParseOpenACC.cpp
index b487a1968d1ec..91f2b8afcf0c2 100644
--- a/clang/lib/Parse/ParseOpenACC.cpp
+++ b/clang/lib/Parse/ParseOpenACC.cpp
@@ -535,14 +535,6 @@ bool ClauseHasRequiredParens(OpenACCDirectiveKind DirKind,
   return getClauseParensKind(DirKind, Kind) == ClauseParensKind::Required;
 }
 
-ExprResult ParseOpenACCConditionalExpr(Parser &P) {
-  // FIXME: It isn't clear if the spec saying 'condition' means the same as
-  // it does in an if/while/etc (See ParseCXXCondition), however as it was
-  // written with Fortran/C in mind, we're going to assume it just means an
-  // 'expression evaluating to boolean'.
-  return P.getActions().CorrectDelayedTyposInExpr(P.ParseExpression());
-}
-
 // Skip until we see the end of pragma token, but don't consume it. This is us
 // just giving up on the rest of the pragma so we can continue executing. We
 // have to do this because 'SkipUntil' considers paren balancing, which isn't
@@ -595,6 +587,23 @@ Parser::OpenACCClauseParseResult Parser::OpenACCSuccess(OpenACCClause *Clause) {
   return {Clause, OpenACCParseCanContinue::Can};
 }
 
+ExprResult Parser::ParseOpenACCConditionExpr() {
+  // FIXME: It isn't clear if the spec saying 'condition' means the same as
+  // it does in an if/while/etc (See ParseCXXCondition), however as it was
+  // written with Fortran/C in mind, we're going to assume it just means an
+  // 'expression evaluating to boolean'.
+  ExprResult ER = getActions().CorrectDelayedTyposInExpr(ParseExpression());
+
+  if (!ER.isUsable())
+    return ER;
+
+  Sema::ConditionResult R =
+      getActions().ActOnCondition(getCurScope(), ER.get()->getExprLoc(),
+                                  ER.get(), Sema::ConditionKind::Boolean);
+
+  return R.isInvalid() ? ExprError() : R.get().second;
+}
+
 // OpenACC 3.3, section 1.7:
 // To simplify the specification and convey appropriate constraint information,
 // a pqr-list is a comma-separated list of pdr items. The one exception is a
@@ -842,12 +851,15 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
       break;
     }
     case OpenACCClauseKind::If: {
-      ExprResult CondExpr = ParseOpenACCConditionalExpr(*this);
+      ExprResult CondExpr = ParseOpenACCConditionExpr();
+      ParsedClause.setConditionDetails(CondExpr.isUsable() ? CondExpr.get()
+                                                           : nullptr);
 
       if (CondExpr.isInvalid()) {
         Parens.skipToEnd();
         return OpenACCCanContinue();
       }
+
       break;
     }
     case OpenACCClauseKind::CopyIn:
@@ -964,7 +976,7 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
       switch (ClauseKind) {
       case OpenACCClauseKind::Self: {
         assert(DirKind != OpenACCDirectiveKind::Update);
-        ExprResult CondExpr = ParseOpenACCConditionalExpr(*this);
+        ExprResult CondExpr = ParseOpenACCConditionExpr();
 
         if (CondExpr.isInvalid()) {
           Parens.skipToEnd();
diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
index a6f4453e525d0..1249136c87650 100644
--- a/clang/lib/Sema/SemaOpenACC.cpp
+++ b/clang/lib/Sema/SemaOpenACC.cpp
@@ -55,12 +55,49 @@ bool doesClauseApplyToDirective(OpenACCDirectiveKind DirectiveKind,
     default:
       return false;
     }
+  case OpenACCClauseKind::If:
+    switch (DirectiveKind) {
+    case OpenACCDirectiveKind::Parallel:
+    case OpenACCDirectiveKind::Serial:
+    case OpenACCDirectiveKind::Kernels:
+    case OpenACCDirectiveKind::Data:
+    case OpenACCDirectiveKind::EnterData:
+    case OpenACCDirectiveKind::ExitData:
+    case OpenACCDirectiveKind::HostData:
+    case OpenACCDirectiveKind::Init:
+    case OpenACCDirectiveKind::Shutdown:
+    case OpenACCDirectiveKind::Set:
+    case OpenACCDirectiveKind::Update:
+    case OpenACCDirectiveKind::Wait:
+    case OpenACCDirectiveKind::ParallelLoop:
+    case OpenACCDirectiveKind::SerialLoop:
+    case OpenACCDirectiveKind::KernelsLoop:
+      return true;
+    default:
+      return false;
+    }
   default:
     // Do nothing so we can go to the 'unimplemented' diagnostic instead.
     return true;
   }
   llvm_unreachable("Invalid clause kind");
 }
+
+bool checkAlreadyHasClauseOfKind(
+    SemaOpenACC &S, ArrayRef<const OpenACCClause *> ExistingClauses,
+    SemaOpenACC::OpenACCParsedClause &Clause) {
+  const auto *Itr = llvm::find_if(ExistingClauses, [&](const OpenACCClause *C) {
+    return C->getClauseKind() == Clause.getClauseKind();
+  });
+  if (Itr != ExistingClauses.end()) {
+    S.Diag(Clause.getBeginLoc(), diag::err_acc_duplicate_clause_disallowed)
+        << Clause.getDirectiveKind() << Clause.getClauseKind();
+    S.Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+    return true;
+  }
+  return false;
+}
+
 } // namespace
 
 SemaOpenACC::SemaOpenACC(Sema &S) : SemaBase(S) {}
@@ -97,22 +134,38 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     // At most one 'default' clause may appear, and it must have a value of
     // either 'none' or 'present'.
     // Second half of the sentence is diagnosed during parsing.
-    auto Itr = llvm::find_if(ExistingClauses, [](const OpenACCClause *C) {
-      return C->getClauseKind() == OpenACCClauseKind::Default;
-    });
-
-    if (Itr != ExistingClauses.end()) {
-      Diag(Clause.getBeginLoc(),
-                   diag::err_acc_duplicate_clause_disallowed)
-          << Clause.getDirectiveKind() << Clause.getClauseKind();
-      Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+    if (checkAlreadyHasClauseOfKind(*this, ExistingClauses, Clause))
       return nullptr;
-    }
 
     return OpenACCDefaultClause::Create(
         getASTContext(), Clause.getDefaultClauseKind(), Clause.getBeginLoc(),
         Clause.getLParenLoc(), Clause.getEndLoc());
   }
+
+  case OpenACCClauseKind::If: {
+    // Restrictions only properly implemented on 'compute' constructs, and
+    // 'compute' constructs are the only construct that can do anything with
+    // this yet, so skip/treat as unimplemented in this case.
+    if (Clause.getDirectiveKind() != OpenACCDirectiveKind::Parallel &&
+        Clause.getDirectiveKind() != OpenACCDirectiveKind::Serial &&
+        Clause.getDirectiveKind() != OpenACCDirectiveKind::Kernels)
+      break;
+
+    // There is no prose in the standard that says duplicates aren't allowed,
+    // but this diagnostic is present in other compilers, as well as makes
+    // sense.
+    if (checkAlreadyHasClauseOfKind(*this, ExistingClauses, Clause))
+      return nullptr;
+
+    // The parser has ensured that we have a proper condition expr, so there
+    // isn't really much to do here.
+
+    // TODO OpenACC: When we implement 'self', this clauses causes us to
+    // 'ignore' the self clause, so we should implement a warning here.
+    return OpenACCIfClause::Create(
+        getASTContext(), Clause.getBeginLoc(), Clause.getLParenLoc(),
+        Clause.getConditionExpr(), Clause.getEndLoc());
+  }
   default:
     break;
   }
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 284c9173e68ed..ee6bd228b459e 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -11103,6 +11103,20 @@ OpenACCClause *TreeTransform<Derived>::TransformOpenACCClause(
     ParsedClause.setDefaultDetails(
         cast<OpenACCDefaultClause>(OldClause)->getDefaultClauseKind());
     break;
+  case OpenACCClauseKind::If: {
+    Expr *Cond = const_cast<Expr *>(
+        cast<OpenACCIfClause>(OldClause)->getConditionExpr());
+    assert(Cond && "If constructed with invalid Condition");
+    Sema::ConditionResult Res =
+        TransformCondition(Cond->getExprLoc(), /*Var=*/nullptr, Cond,
+                           Sema::ConditionKind::Boolean);
+
+    if (Res.isInvalid() || !Res.get().second)
+      return nullptr;
+
+    ParsedClause.setConditionDetails(Res.get().second);
+    break;
+  }
   default:
     assert(false && "Unhandled OpenACC clause in TreeTransform");
     return nullptr;
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index ce96ce2bdbcce..a47c7a393d632 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -11787,6 +11787,12 @@ OpenACCClause *ASTRecordReader::readOpenACCClause() {
     return OpenACCDefaultClause::Create(getContext(), DCK, BeginLoc, LParenLoc,
                                         EndLoc);
   }
+  case OpenACCClauseKind::If: {
+    SourceLocation LParenLoc = readSourceLocation();
+    Expr *CondExpr = readSubExpr();
+    return OpenACCIfClause::Create(getContext(), BeginLoc, LParenLoc, CondExpr,
+                                   EndLoc);
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
@@ -11795,7 +11801,6 @@ OpenACCClause *ASTRecordReader::readOpenACCClause() {
   case OpenACCClauseKind::Worker:
   case OpenACCClauseKind::Vector:
   case OpenACCClauseKind::NoHost:
-  case OpenACCClauseKind::If:
   case OpenACCClauseKind::Self:
   case OpenACCClauseKind::Copy:
   case OpenACCClauseKind::UseDevice:
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index d9ba10ab60878..d45695784354a 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -7517,6 +7517,12 @@ void ASTRecordWriter::writeOpenACCClause(const OpenACCClause *C) {
     writeEnum(DC->getDefaultClauseKind());
     return;
   }
+  case OpenACCClauseKind::If: {
+    const auto *IC = cast<OpenACCIfClause>(C);
+    writeSourceLocation(IC->getLParenLoc());
+    AddStmt(const_cast<Expr *>(IC->getConditionExpr()));
+    return;
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
@@ -7525,7 +7531,6 @@ void ASTRecordWriter::writeOpenACCClause(const OpenACCClause *C) {
   case OpenACCClauseKind::Worker:
   case OpenACCClauseKind::Vector:
   case OpenACCClauseKind::NoHost:
-  case OpenACCClauseKind::If:
   case OpenACCClauseKind::Self:
   case OpenACCClauseKind::Copy:
   case OpenACCClauseKind::UseDevice:
diff --git a/clang/test/ParserOpenACC/parse-clauses.c b/clang/test/ParserOpenACC/parse-clauses.c
index b363a0cb1362b..2369df58308a7 100644
--- a/clang/test/ParserOpenACC/parse-clauses.c
+++ b/clang/test/ParserOpenACC/parse-clauses.c
@@ -283,11 +283,9 @@ void IfClause() {
 
   int i, j;
 
-  // expected-warning@+1{{OpenACC clause 'if' not yet implemented, clause ignored}}
 #pragma acc serial if(i > j)
   for(;;){}
 
-  // expected-warning@+2{{OpenACC clause 'if' not yet implemented, clause ignored}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial if(1+5>3), seq
   for(;;){}
diff --git a/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
index bd80103445028..018f0b68c7810 100644
--- a/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
+++ b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
@@ -6,8 +6,10 @@
 
 #ifndef PCH_HELPER
 #define PCH_HELPER
-void NormalFunc() {
+void NormalFunc(int i, float f) {
   // CHECK: FunctionDecl{{.*}}NormalFunc
+  // CHECK-NEXT: ParmVarDecl
+  // CHECK-NEXT: ParmVarDecl
   // CHECK-NEXT: CompoundStmt
 #pragma acc parallel default(none)
   while(true);
@@ -24,6 +26,20 @@ void NormalFunc() {
   // CHECK-NEXT: WhileStmt
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
+
+#pragma acc kernels if( i < f)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: BinaryOperator{{.*}} 'bool' '<'
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'float' <IntegralToFloating>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'int' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'int' lvalue ParmVar{{.*}} 'i' 'int'
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'float' lvalue ParmVar{{.*}} 'f' 'float'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
 }
 
 template<typename T>
@@ -51,24 +67,120 @@ void TemplFunc() {
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
 
+#pragma acc parallel if(T::SomeFloat < typename T::IntTy{})
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: BinaryOperator{{.*}} '<dependent type>' '<'
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: CXXUnresolvedConstructExpr{{.*}} 'typename T::IntTy' 'typename T::IntTy'
+  // CHECK-NEXT: InitListExpr{{.*}} 'void'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc serial if(typename T::IntTy{})
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: CXXUnresolvedConstructExpr{{.*}} 'typename T::IntTy' 'typename T::IntTy'
+  // CHECK-NEXT: InitListExpr{{.*}} 'void'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc kernels if(T::SomeFloat)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc parallel if(T::BC)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
   // Match the instantiation:
   // CHECK: FunctionDecl{{.*}}TemplFunc{{.*}}implicit_instantiation
-  // CHECK-NEXT: TemplateArgument type 'int'
-  // CHECK-NEXT: BuiltinType
+  // CHECK-NEXT: TemplateArgument type 'InstTy'
+  // CHECK-NEXT: RecordType{{.*}} 'InstTy'
+  // CHECK-NEXT: CXXRecord{{.*}} 'InstTy'
   // CHECK-NEXT: CompoundStmt
+
   // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
   // CHECK-NEXT: default(none)
   // CHECK-NEXT: WhileStmt
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
+
   // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
   // CHECK-NEXT: default(present)
   // CHECK-NEXT: WhileStmt
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: BinaryOperator{{.*}} 'bool' '<'
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'float' <IntegralToFloating>
+  // CHECK-NEXT: CXXFunctionalCastExpr{{.*}}'typename InstTy::IntTy':'int' functional cast to typename struct InstTy::IntTy <NoOp>
+  // CHECK-NEXT: InitListExpr {{.*}}'typename InstTy::IntTy':'int'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <IntegralToBoolean>
+  // CHECK-NEXT: CXXFunctionalCastExpr{{.*}}'typename InstTy::IntTy':'int' functional cast to typename struct InstTy::IntTy <NoOp>
+  // CHECK-NEXT: InitListExpr {{.*}}'typename InstTy::IntTy':'int'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}} 'bool' <UserDefinedConversion>
+  // CHECK-NEXT: CXXMemberCallExpr{{.*}} 'bool'
+  // CHECK-NEXT: MemberExpr{{.*}} .operator bool
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const BoolConversion' lvalue Var{{.*}} 'BC' 'const BoolConversion'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
 }
 
+struct BoolConversion{ operator bool() const;};
+struct InstTy {
+  using IntTy = int;
+  static constexpr float SomeFloat = 5.0;
+  static constexpr BoolConversion BC;
+};
+
 void Instantiate() {
-  TemplFunc<int>();
+  TemplFunc<InstTy>();
 }
 #endif
diff --git a/clang/test/SemaOpenACC/compute-construct-if-clause.c b/clang/test/SemaOpenACC/compute-construct-if-clause.c
new file mode 100644
index 0000000000000..767b8414b3a68
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-if-clause.c
@@ -0,0 +1,62 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+void BoolExpr(int *I, float *F) {
+
+  typedef struct {} SomeStruct;
+  int Array[5];
+
+  struct C{};
+  // expected-error@+1{{expected expression}}
+#pragma acc parallel if (struct C f())
+  while(0);
+
+  // expected-error@+1{{unexpected type name 'SomeStruct': expected expression}}
+#pragma acc serial if (SomeStruct)
+  while(0);
+
+  // expected-error@+1{{unexpected type name 'SomeStruct': expected expression}}
+#pragma acc serial if (SomeStruct())
+  while(0);
+
+  SomeStruct S;
+  // expected-error@+1{{statement requires expression of scalar type ('SomeStruct' invalid)}}
+#pragma acc serial if (S)
+  while(0);
+
+  // expected-warning@+1{{address of array 'Array' will always evaluate to 'true'}}
+#pragma acc kernels if (Array)
+  while(0);
+
+  // expected-warning@+4{{incompatible pointer types assigning to 'int *' from 'float *'}}
+  // expected-warning@+3{{using the result of an assignment as a condition without parentheses}}
+  // expected-note@+2{{place parentheses around the assignment to silence this warning}}
+  // expected-note@+1{{use '==' to turn this assignment into an equality comparison}}
+#pragma acc kernels if (I = F)
+  while(0);
+
+#pragma acc parallel if (I)
+  while(0);
+
+#pragma acc serial if (F)
+  while(0);
+
+#pragma acc kernels if (*I < *F)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'data' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'if' not yet implemented}}
+#pragma acc data if (*I < *F)
+  while(0);
+  // expected-warning@+2{{OpenACC construct 'parallel loop' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'if' not yet implemented}}
+#pragma acc parallel loop if (*I < *F)
+  while(0);
+  // expected-warning@+2{{OpenACC construct 'serial loop' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'if' not yet implemented}}
+#pragma acc serial loop if (*I < *F)
+  while(0);
+  // expected-warning@+2{{OpenACC construct 'kernels loop' not yet implemented}}
+  // expected-warning@+1{{OpenACC clause 'if' not yet implemented}}
+#pragma acc kernels loop if (*I < *F)
+  while(0);
+}
diff --git a/clang/test/SemaOpenACC/compute-construct-if-clause.cpp b/clang/test/SemaOpenACC/compute-construct-if-clause.cpp
new file mode 100644
index 0000000000000..2a9bb8a3d5c4e
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-if-clause.cpp
@@ -0,0 +1,33 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+struct NoBoolConversion{};
+struct BoolConversion{
+  operator bool();
+};
+
+template <typename T, typename U>
+void BoolExpr() {
+
+  // expected-error@+1{{value of type 'NoBoolConversion' is not contextually convertible to 'bool'}}
+#pragma acc parallel if (NoBoolConversion{})
+  while(0);
+
+  // expected-error@+2{{no member named 'NotValid' in 'NoBoolConversion'}}
+  // expected-note@#INST{{in instantiation of function template specialization}}
+#pragma acc parallel if (T::NotValid)
+  while(0);
+
+#pragma acc parallel if (BoolConversion{})
+  while(0);
+
+  // expected-error@+1{{value of type 'NoBoolConversion' is not contextually convertible to 'bool'}}
+#pragma acc parallel if (T{})
+  while(0);
+
+#pragma acc parallel if (U{})
+  while(0);
+}
+
+void Instantiate() {
+  BoolExpr<NoBoolConversion, BoolConversion>(); // #INST
+}
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index 00ddc4f24500f..936e8872513ae 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -28,6 +28,7 @@
 #include "clang/AST/Expr.h"
 #include "clang/AST/ExprCXX.h"
 #include "clang/AST/Mangle.h"
+#include "clang/AST/OpenACCClause.h"
 #include "clang/AST/OpenMPClause.h"
 #include "clang/AST/OperationKinds.h"
 #include "clang/AST/StmtVisitor.h"
@@ -2115,6 +2116,7 @@ class MemberRefVisit : public VisitorJob {
 };
 class EnqueueVisitor : public ConstStmtVisitor<EnqueueVisitor, void>,
                        public ConstAttrVisitor<EnqueueVisitor, void> {
+  friend class OpenACCClauseEnqueue;
   friend class OMPClauseEnqueue;
   VisitorWorkList &WL;
   CXCursor Parent;
@@ -2171,6 +2173,7 @@ class EnqueueVisitor : public ConstStmtVisitor<EnqueueVisitor, void>,
   void VisitConceptSpecializationExpr(const ConceptSpecializationExpr *E);
   void VisitRequiresExpr(const RequiresExpr *E);
   void VisitCXXParenListInitExpr(const CXXParenListInitExpr *E);
+  void VisitOpenACCComputeConstruct(const OpenACCComputeConstruct *D);
   void VisitOMPExecutableDirective(const OMPExecutableDirective *D);
   void VisitOMPLoopBasedDirective(const OMPLoopBasedDirective *D);
   void VisitOMPLoopDirective(const OMPLoopDirective *D);
@@ -2269,6 +2272,7 @@ class EnqueueVisitor : public ConstStmtVisitor<EnqueueVisitor, void>,
   void AddDecl(const Decl *D, bool isFirst = true);
   void AddTypeLoc(TypeSourceInfo *TI);
   void EnqueueChildren(const Stmt *S);
+  void EnqueueChildren(const OpenACCClause *S);
   void EnqueueChildren(const OMPClause *S);
   void EnqueueChildren(const AnnotateAttr *A);
 };
@@ -2771,6 +2775,38 @@ void EnqueueVisitor::EnqueueChildren(const OMPClause *S) {
   std::reverse(I, E);
 }
 
+namespace {
+class OpenACCClauseEnqueue : public OpenACCClauseVisitor<OpenACCClauseEnqueue> {
+  EnqueueVisitor &Visitor;
+
+public:
+  OpenACCClauseEnqueue(EnqueueVisitor &V) : Visitor(V) {}
+
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  void VisitOpenACC##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &C);
+#include "clang/Basic/OpenACCClauses.def"
+};
+
+void OpenACCClauseEnqueue::VisitOpenACCDefaultClause(
+    const OpenACCDefaultClause &C) {}
+void OpenACCClauseEnqueue::VisitOpenACCIfClause(const OpenACCIfClause &C) {
+  Visitor.AddStmt(C.getConditionExpr());
+}
+} // namespace
+
+void EnqueueVisitor::EnqueueChildren(const OpenACCClause *C) {
+  unsigned size = WL.size();
+  OpenACCClauseEnqueue Visitor(*this);
+  Visitor.Visit(C);
+
+  if (size == WL.size())
+    return;
+  // Now reverse the entries we just added.  This will match the DFS
+  // ordering performed by the worklist.
+  VisitorWorkList::iterator I = WL.begin() + size, E = WL.end();
+  std::reverse(I, E);
+}
+
 void EnqueueVisitor::EnqueueChildren(const AnnotateAttr *A) {
   unsigned size = WL.size();
   for (const Expr *Arg : A->args()) {
@@ -3386,6 +3422,13 @@ void EnqueueVisitor::VisitOMPTargetTeamsDistributeSimdDirective(
   VisitOMPLoopDirective(D);
 }
 
+void EnqueueVisitor::VisitOpenACCComputeConstruct(
+    const OpenACCComputeConstruct *C) {
+  EnqueueChildren(C);
+  for (auto *Clause : C->clauses())
+    EnqueueChildren(Clause);
+}
+
 void EnqueueVisitor::VisitAnnotateAttr(const AnnotateAttr *A) {
   EnqueueChildren(A);
 }

From 7b6b023121fe7d612dc571de0ff3dcaaf8477765 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Fri, 12 Apr 2024 14:28:05 -0700
Subject: [PATCH 303/886] [flang][cuda] Fix crash in semantic (#88577)

Fix for #88451

Do not perform semantic check about data transfer on assignment
statement in device context.
---
 flang/lib/Semantics/check-cuda.cpp |  8 +++++++-
 flang/test/Semantics/cuf11.cuf     | 12 ++++++++++++
 2 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/flang/lib/Semantics/check-cuda.cpp b/flang/lib/Semantics/check-cuda.cpp
index e0a796972441b..2cb15437a235a 100644
--- a/flang/lib/Semantics/check-cuda.cpp
+++ b/flang/lib/Semantics/check-cuda.cpp
@@ -480,10 +480,16 @@ void CUDAChecker::Enter(const parser::CUFKernelDoConstruct &x) {
 }
 
 void CUDAChecker::Enter(const parser::AssignmentStmt &x) {
+  auto lhsLoc{std::get<parser::Variable>(x.t).GetSource()};
+  const auto &scope{context_.FindScope(lhsLoc)};
+  const Scope &progUnit{GetProgramUnitContaining(scope)};
+  if (IsCUDADeviceContext(&progUnit)) {
+    return; // Data transfer with assignment is only perform on host.
+  }
+
   const evaluate::Assignment *assign{semantics::GetAssignment(x)};
   int nbLhs{evaluate::GetNbOfCUDASymbols(assign->lhs)};
   int nbRhs{evaluate::GetNbOfCUDASymbols(assign->rhs)};
-  auto lhsLoc{std::get<parser::Variable>(x.t).GetSource()};
 
   // device to host transfer with more than one device object on the rhs is not
   // legal.
diff --git a/flang/test/Semantics/cuf11.cuf b/flang/test/Semantics/cuf11.cuf
index 96108e2b24556..b915c7246ed94 100644
--- a/flang/test/Semantics/cuf11.cuf
+++ b/flang/test/Semantics/cuf11.cuf
@@ -1,5 +1,17 @@
 ! RUN: %python %S/test_errors.py %s %flang_fc1
 
+module mod1
+contains
+  attributes(global) subroutine sub1(adev)
+    real :: adev(10)
+    integer :: tid
+    tid = threadIdx%x
+! Use to crash the compiler. Make sure we have the proper semantic error.
+!ERROR: Actual argument for 'i=' has bad type 'REAL(4)'
+    adev(tid + 1) = scale(real(tid), 2.0) 
+  end subroutine sub1
+end module
+
 subroutine sub1()
   real, device :: adev(10), bdev(10)
   real :: ahost(10)

From 900be9013fdc3bab9fce906f8a71e59ecd8873b4 Mon Sep 17 00:00:00 2001
From: Victor Toni <ViToni@users.noreply.github.com>
Date: Fri, 12 Apr 2024 23:33:35 +0200
Subject: [PATCH 304/886] Fix typos (#88565)

---
 clang/lib/StaticAnalyzer/Core/ExprEngineCXX.cpp         | 2 +-
 libc/src/__support/str_to_float.h                       | 8 ++++----
 llvm/lib/Analysis/LoopCacheAnalysis.cpp                 | 2 +-
 llvm/lib/Support/APInt.cpp                              | 2 +-
 llvm/lib/Transforms/Scalar/LoopInterchange.cpp          | 2 +-
 llvm/test/Transforms/JumpThreading/free_instructions.ll | 2 +-
 llvm/utils/lint/common_lint.py                          | 2 +-
 7 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/clang/lib/StaticAnalyzer/Core/ExprEngineCXX.cpp b/clang/lib/StaticAnalyzer/Core/ExprEngineCXX.cpp
index 504fd7f05e0f9..c50db1e0e2f86 100644
--- a/clang/lib/StaticAnalyzer/Core/ExprEngineCXX.cpp
+++ b/clang/lib/StaticAnalyzer/Core/ExprEngineCXX.cpp
@@ -229,7 +229,7 @@ SVal ExprEngine::computeObjectUnderConstruction(
         // We are on the top frame of the analysis. We do not know where is the
         // object returned to. Conjure a symbolic region for the return value.
         // TODO: We probably need a new MemRegion kind to represent the storage
-        // of that SymbolicRegion, so that we cound produce a fancy symbol
+        // of that SymbolicRegion, so that we could produce a fancy symbol
         // instead of an anonymous conjured symbol.
         // TODO: Do we need to track the region to avoid having it dead
         // too early? It does die too early, at least in C++17, but because
diff --git a/libc/src/__support/str_to_float.h b/libc/src/__support/str_to_float.h
index cd0c07629f876..fa466ca8492fc 100644
--- a/libc/src/__support/str_to_float.h
+++ b/libc/src/__support/str_to_float.h
@@ -695,10 +695,10 @@ LIBC_INLINE FloatConvertReturn<T> decimal_exp_to_float(
     // If the mantissa is truncated, then the result may be off by the LSB, so
     // check if rounding the mantissa up changes the result. If not, then it's
     // safe, else use the fallback.
-    auto secound_output = eisel_lemire<T>({mantissa + 1, exp10}, round);
-    if (secound_output.has_value()) {
-      if (opt_output->mantissa == secound_output->mantissa &&
-          opt_output->exponent == secound_output->exponent) {
+    auto second_output = eisel_lemire<T>({mantissa + 1, exp10}, round);
+    if (second_output.has_value()) {
+      if (opt_output->mantissa == second_output->mantissa &&
+          opt_output->exponent == second_output->exponent) {
         return {opt_output.value(), 0};
       }
     }
diff --git a/llvm/lib/Analysis/LoopCacheAnalysis.cpp b/llvm/lib/Analysis/LoopCacheAnalysis.cpp
index c3a56639b5c8f..284d8d16d264e 100644
--- a/llvm/lib/Analysis/LoopCacheAnalysis.cpp
+++ b/llvm/lib/Analysis/LoopCacheAnalysis.cpp
@@ -315,7 +315,7 @@ CacheCostTy IndexedReference::computeRefCost(const Loop &L,
     RefCost = TripCount;
 
     int Index = getSubscriptIndex(L);
-    assert(Index >= 0 && "Cound not locate a valid Index");
+    assert(Index >= 0 && "Could not locate a valid Index");
 
     for (unsigned I = Index + 1; I < getNumSubscripts() - 1; ++I) {
       const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(getSubscript(I));
diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index 18feca4c05533..24e136fcb9c7f 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -2663,7 +2663,7 @@ int APInt::tcDivide(WordType *lhs, const WordType *rhs,
   return false;
 }
 
-/// Shift a bignum left Cound bits in-place. Shifted in bits are zero. There are
+/// Shift a bignum left Count bits in-place. Shifted in bits are zero. There are
 /// no restrictions on Count.
 void APInt::tcShiftLeft(WordType *Dst, unsigned Words, unsigned Count) {
   // Don't bother performing a no-op shift.
diff --git a/llvm/lib/Transforms/Scalar/LoopInterchange.cpp b/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
index 277f530ee25fc..400973fd9fc91 100644
--- a/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
@@ -976,7 +976,7 @@ bool LoopInterchangeLegality::canInterchangeLoops(unsigned InnerLoopId,
       }
 
   if (!findInductions(InnerLoop, InnerLoopInductions)) {
-    LLVM_DEBUG(dbgs() << "Cound not find inner loop induction variables.\n");
+    LLVM_DEBUG(dbgs() << "Could not find inner loop induction variables.\n");
     return false;
   }
 
diff --git a/llvm/test/Transforms/JumpThreading/free_instructions.ll b/llvm/test/Transforms/JumpThreading/free_instructions.ll
index 3485bf8c98da8..48bedc72c2d31 100644
--- a/llvm/test/Transforms/JumpThreading/free_instructions.ll
+++ b/llvm/test/Transforms/JumpThreading/free_instructions.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -passes=jump-threading -jump-threading-threshold=3 < %s | FileCheck %s
 
-; TODO: In this test case, only the stores in %join should cound towards
+; TODO: In this test case, only the stores in %join should count towards
 ; the jump threading threshold, as everything else are free instructions.
 define i32 @free_instructions(i1 %c, ptr %p) {
 ; CHECK-LABEL: @free_instructions(
diff --git a/llvm/utils/lint/common_lint.py b/llvm/utils/lint/common_lint.py
index 1bf1695659d88..07b8fc5899e55 100644
--- a/llvm/utils/lint/common_lint.py
+++ b/llvm/utils/lint/common_lint.py
@@ -97,7 +97,7 @@ def RunLintOverAllFiles(linter, filenames):
     for filename in filenames:
         file = open(filename, "r")
         if not file:
-            print("Cound not open %s" % filename)
+            print("Could not open %s" % filename)
             continue
         lines = file.readlines()
         lint.extend(linter.RunOnFile(filename, lines))

From 2996cb7970fb536d7db1088d571dc8d2f9ad35fa Mon Sep 17 00:00:00 2001
From: Michael Flanders <flanders.michaelk@gmail.com>
Date: Fri, 12 Apr 2024 14:35:10 -0700
Subject: [PATCH 305/886] [libc][docs] use docgen to generate stdbit rst docs
 (#88540)

Closes #88066.

Compared to before, the function names in the stdbit table are sorted by
function name, not order-of-appearance in the standard. Since macros
aren't printed by docgen.py and are still a TODO in the code, they are
also not printed in the new stdbit.h docs.

Adds some checks to docgen.py for conditions that tripped me up.

Add code to docgen.py to add the include of the `|check|` rewriter,
since all other generated files need it.
---
 libc/docs/stdbit.rst          | 350 +++++++++++++++++++++-------------
 libc/utils/docgen/docgen.py   |   8 +
 libc/utils/docgen/stdbit.json | 234 +++++++++++++++++++++++
 3 files changed, 461 insertions(+), 131 deletions(-)
 create mode 100644 libc/utils/docgen/stdbit.json

diff --git a/libc/docs/stdbit.rst b/libc/docs/stdbit.rst
index d42f793824629..0a12b2b6d7b51 100644
--- a/libc/docs/stdbit.rst
+++ b/libc/docs/stdbit.rst
@@ -1,135 +1,223 @@
-=========================
-Bitwise Utility Functions
-=========================
-
 .. include:: check.rst
 
----------------
-Source Location
----------------
-
--   The main source for bitwise utility functions is located at:
-    ``libc/src/stdbit``.
-
--   The source for internal helpers used to implement these is located at:
-    ``libc/src/__support/CPP/bit.h``.
-
--   The tests are located at:
-    ``libc/test/src/stdbit/``, ``libc/test/include/stdbit_test.cpp``, and
-    ``src/__support/CPP/bit_test.cpp``.
-
----------------------
-Implementation Status
----------------------
-
-Functions
-=========
-
-..
-  Do not order these, they are as they appear in the standard.
-
-============================ =========
-Function Name                Available
-============================ =========
-stdc_leading_zeros_uc        |check|
-stdc_leading_zeros_us        |check|
-stdc_leading_zeros_ui        |check|
-stdc_leading_zeros_ul        |check|
-stdc_leading_zeros_ull       |check|
-stdc_leading_ones_uc         |check|
-stdc_leading_ones_us         |check|
-stdc_leading_ones_ui         |check|
-stdc_leading_ones_ul         |check|
-stdc_leading_ones_ull        |check|
-stdc_trailing_zeros_uc       |check|
-stdc_trailing_zeros_us       |check|
-stdc_trailing_zeros_ui       |check|
-stdc_trailing_zeros_ul       |check|
-stdc_trailing_zeros_ull      |check|
-stdc_trailing_ones_uc        |check|
-stdc_trailing_ones_us        |check|
-stdc_trailing_ones_ui        |check|
-stdc_trailing_ones_ul        |check|
-stdc_trailing_ones_ull       |check|
-stdc_first_leading_zero_uc   |check|
-stdc_first_leading_zero_us   |check|
-stdc_first_leading_zero_ui   |check|
-stdc_first_leading_zero_ul   |check|
-stdc_first_leading_zero_ull  |check|
-stdc_first_leading_one_uc    |check|
-stdc_first_leading_one_us    |check|
-stdc_first_leading_one_ui    |check|
-stdc_first_leading_one_ul    |check|
-stdc_first_leading_one_ull   |check|
-stdc_first_trailing_zero_uc  |check|
-stdc_first_trailing_zero_us  |check|
-stdc_first_trailing_zero_ui  |check|
-stdc_first_trailing_zero_ul  |check|
-stdc_first_trailing_zero_ull |check|
-stdc_first_trailing_one_uc   |check|
-stdc_first_trailing_one_us   |check|
-stdc_first_trailing_one_ui   |check|
-stdc_first_trailing_one_ul   |check|
-stdc_first_trailing_one_ull  |check|
-stdc_count_zeros_uc          |check|
-stdc_count_zeros_us          |check|
-stdc_count_zeros_ui          |check|
-stdc_count_zeros_ul          |check|
-stdc_count_zeros_ull         |check|
-stdc_count_ones_uc           |check|
-stdc_count_ones_us           |check|
-stdc_count_ones_ui           |check|
-stdc_count_ones_ul           |check|
-stdc_count_ones_ull          |check|
-stdc_has_single_bit_uc       |check|
-stdc_has_single_bit_us       |check|
-stdc_has_single_bit_ui       |check|
-stdc_has_single_bit_ul       |check|
-stdc_has_single_bit_ull      |check|
-stdc_bit_width_uc            |check|
-stdc_bit_width_us            |check|
-stdc_bit_width_ui            |check|
-stdc_bit_width_ul            |check|
-stdc_bit_width_ull           |check|
-stdc_bit_floor_uc            |check|
-stdc_bit_floor_us            |check|
-stdc_bit_floor_ui            |check|
-stdc_bit_floor_ul            |check|
-stdc_bit_floor_ull           |check|
-stdc_bit_ceil_uc             |check|
-stdc_bit_ceil_us             |check|
-stdc_bit_ceil_ui             |check|
-stdc_bit_ceil_ul             |check|
-stdc_bit_ceil_ull            |check|
-============================ =========
-
-
-Macros
-======
+stdbit.h Functions
+==================
 
-=========================  =========
-Macro Name                 Available
-=========================  =========
-__STDC_VERSION_STDBIT_H__  |check|
-__STDC_ENDIAN_LITTLE__     |check|
-__STDC_ENDIAN_BIG__        |check|
-__STDC_ENDIAN_NATIVE__     |check|
-stdc_leading_zeros         |check|
-stdc_leading_ones          |check|
-stdc_trailing_zeros        |check|
-stdc_trailing_ones         |check|
-stdc_first_leading_zero    |check|
-stdc_first_leading_one     |check|
-stdc_first_trailing_zero   |check|
-stdc_first_trailing_one    |check|
-stdc_count_zeros           |check|
-stdc_count_ones            |check|
-stdc_has_single_bit        |check|
-stdc_bit_width             |check|
-stdc_bit_floor             |check|
-stdc_bit_ceil              |check|
-=========================  =========
+.. list-table::
+  :widths: auto
+  :align: center
+  :header-rows: 1
 
-Standards
-=========
-stdbit.h was specified as part of C23 in section 7.18 "Bit and byte utilities."
+  * - Function
+    - Implemented
+    - Standard
+  * - stdc_bit_ceil_uc
+    - |check|
+    - 7.18.16
+  * - stdc_bit_ceil_ui
+    - |check|
+    - 7.18.16
+  * - stdc_bit_ceil_ul
+    - |check|
+    - 7.18.16
+  * - stdc_bit_ceil_ull
+    - |check|
+    - 7.18.16
+  * - stdc_bit_ceil_us
+    - |check|
+    - 7.18.16
+  * - stdc_bit_floor_uc
+    - |check|
+    - 7.18.15
+  * - stdc_bit_floor_ui
+    - |check|
+    - 7.18.15
+  * - stdc_bit_floor_ul
+    - |check|
+    - 7.18.15
+  * - stdc_bit_floor_ull
+    - |check|
+    - 7.18.15
+  * - stdc_bit_floor_us
+    - |check|
+    - 7.18.15
+  * - stdc_bit_width_uc
+    - |check|
+    - 7.18.14
+  * - stdc_bit_width_ui
+    - |check|
+    - 7.18.14
+  * - stdc_bit_width_ul
+    - |check|
+    - 7.18.14
+  * - stdc_bit_width_ull
+    - |check|
+    - 7.18.14
+  * - stdc_bit_width_us
+    - |check|
+    - 7.18.14
+  * - stdc_count_ones_uc
+    - |check|
+    - 7.18.12
+  * - stdc_count_ones_ui
+    - |check|
+    - 7.18.12
+  * - stdc_count_ones_ul
+    - |check|
+    - 7.18.12
+  * - stdc_count_ones_ull
+    - |check|
+    - 7.18.12
+  * - stdc_count_ones_us
+    - |check|
+    - 7.18.12
+  * - stdc_count_zeros_uc
+    - |check|
+    - 7.18.11
+  * - stdc_count_zeros_ui
+    - |check|
+    - 7.18.11
+  * - stdc_count_zeros_ul
+    - |check|
+    - 7.18.11
+  * - stdc_count_zeros_ull
+    - |check|
+    - 7.18.11
+  * - stdc_count_zeros_us
+    - |check|
+    - 7.18.11
+  * - stdc_first_leading_one_uc
+    - |check|
+    - 7.18.8
+  * - stdc_first_leading_one_ui
+    - |check|
+    - 7.18.8
+  * - stdc_first_leading_one_ul
+    - |check|
+    - 7.18.8
+  * - stdc_first_leading_one_ull
+    - |check|
+    - 7.18.8
+  * - stdc_first_leading_one_us
+    - |check|
+    - 7.18.8
+  * - stdc_first_leading_zero_uc
+    - |check|
+    - 7.18.7
+  * - stdc_first_leading_zero_ui
+    - |check|
+    - 7.18.7
+  * - stdc_first_leading_zero_ul
+    - |check|
+    - 7.18.7
+  * - stdc_first_leading_zero_ull
+    - |check|
+    - 7.18.7
+  * - stdc_first_leading_zero_us
+    - |check|
+    - 7.18.7
+  * - stdc_first_trailing_one_uc
+    - |check|
+    - 7.18.10
+  * - stdc_first_trailing_one_ui
+    - |check|
+    - 7.18.10
+  * - stdc_first_trailing_one_ul
+    - |check|
+    - 7.18.10
+  * - stdc_first_trailing_one_ull
+    - |check|
+    - 7.18.10
+  * - stdc_first_trailing_one_us
+    - |check|
+    - 7.18.10
+  * - stdc_first_trailing_zero_uc
+    - |check|
+    - 7.18.9
+  * - stdc_first_trailing_zero_ui
+    - |check|
+    - 7.18.9
+  * - stdc_first_trailing_zero_ul
+    - |check|
+    - 7.18.9
+  * - stdc_first_trailing_zero_ull
+    - |check|
+    - 7.18.9
+  * - stdc_first_trailing_zero_us
+    - |check|
+    - 7.18.9
+  * - stdc_has_single_bit_uc
+    - |check|
+    - 7.18.13
+  * - stdc_has_single_bit_ui
+    - |check|
+    - 7.18.13
+  * - stdc_has_single_bit_ul
+    - |check|
+    - 7.18.13
+  * - stdc_has_single_bit_ull
+    - |check|
+    - 7.18.13
+  * - stdc_has_single_bit_us
+    - |check|
+    - 7.18.13
+  * - stdc_leading_ones_uc
+    - |check|
+    - 7.18.4
+  * - stdc_leading_ones_ui
+    - |check|
+    - 7.18.4
+  * - stdc_leading_ones_ul
+    - |check|
+    - 7.18.4
+  * - stdc_leading_ones_ull
+    - |check|
+    - 7.18.4
+  * - stdc_leading_ones_us
+    - |check|
+    - 7.18.4
+  * - stdc_leading_zeros_uc
+    - |check|
+    - 7.18.3
+  * - stdc_leading_zeros_ui
+    - |check|
+    - 7.18.3
+  * - stdc_leading_zeros_ul
+    - |check|
+    - 7.18.3
+  * - stdc_leading_zeros_ull
+    - |check|
+    - 7.18.3
+  * - stdc_leading_zeros_us
+    - |check|
+    - 7.18.3
+  * - stdc_trailing_ones_uc
+    - |check|
+    - 7.18.6
+  * - stdc_trailing_ones_ui
+    - |check|
+    - 7.18.6
+  * - stdc_trailing_ones_ul
+    - |check|
+    - 7.18.6
+  * - stdc_trailing_ones_ull
+    - |check|
+    - 7.18.6
+  * - stdc_trailing_ones_us
+    - |check|
+    - 7.18.6
+  * - stdc_trailing_zeros_uc
+    - |check|
+    - 7.18.5
+  * - stdc_trailing_zeros_ui
+    - |check|
+    - 7.18.5
+  * - stdc_trailing_zeros_ul
+    - |check|
+    - 7.18.5
+  * - stdc_trailing_zeros_ull
+    - |check|
+    - 7.18.5
+  * - stdc_trailing_zeros_us
+    - |check|
+    - 7.18.5
diff --git a/libc/utils/docgen/docgen.py b/libc/utils/docgen/docgen.py
index 36eb409421b4f..23d45305fe513 100755
--- a/libc/utils/docgen/docgen.py
+++ b/libc/utils/docgen/docgen.py
@@ -28,6 +28,13 @@ def is_implemented(hname: str, fname: str) -> bool:
         "src",
         hname.rstrip(".h")
     )
+
+    if not path.exists():
+        raise FileNotFoundError(f"implementation dir does not exist: {path}")
+
+    if not path.is_dir():
+        raise NotADirectoryError(f"implementation dir is not a dir: {path}")
+
     # Recursively search for the target source file in the subdirectories under
     # libc/src/{hname}.
     for _ in path.glob("**/" + fname + ".cpp"):
@@ -53,6 +60,7 @@ def print_functions(header: str, functions: Dict):
 
 
 def print_header(header: str, api: Dict):
+    print(".. include:: check.rst\n")
     fns = f"{header} Functions"
     print(fns)
     print("=" * (len(fns)))
diff --git a/libc/utils/docgen/stdbit.json b/libc/utils/docgen/stdbit.json
new file mode 100644
index 0000000000000..9dda0cb0f5383
--- /dev/null
+++ b/libc/utils/docgen/stdbit.json
@@ -0,0 +1,234 @@
+{
+  "macros": [
+    "__STDC_VERSION_STDBIT_H__",
+    "__STDC_ENDIAN_LITTLE__",
+    "__STDC_ENDIAN_BIG__",
+    "__STDC_ENDIAN_NATIVE__",
+    "stdc_leading_zeros",
+    "stdc_leading_ones",
+    "stdc_trailing_zeros",
+    "stdc_trailing_ones",
+    "stdc_first_leading_zero",
+    "stdc_first_leading_one",
+    "stdc_first_trailing_zero",
+    "stdc_first_trailing_one",
+    "stdc_count_zeros",
+    "stdc_count_ones",
+    "stdc_has_single_bit",
+    "stdc_bit_width",
+    "stdc_bit_floor",
+    "stdc_bit_ceil"
+  ],
+  "functions": {
+    "stdc_leading_zeros_uc": {
+      "defined": "7.18.3"
+    },
+    "stdc_leading_zeros_us": {
+      "defined": "7.18.3"
+    },
+    "stdc_leading_zeros_ui": {
+      "defined": "7.18.3"
+    },
+    "stdc_leading_zeros_ul": {
+      "defined": "7.18.3"
+    },
+    "stdc_leading_zeros_ull": {
+      "defined": "7.18.3"
+    },
+    "stdc_leading_ones_uc": {
+      "defined": "7.18.4"
+    },
+    "stdc_leading_ones_us": {
+      "defined": "7.18.4"
+    },
+    "stdc_leading_ones_ui": {
+      "defined": "7.18.4"
+    },
+    "stdc_leading_ones_ul": {
+      "defined": "7.18.4"
+    },
+    "stdc_leading_ones_ull": {
+      "defined": "7.18.4"
+    },
+    "stdc_trailing_zeros_uc": {
+      "defined": "7.18.5"
+    },
+    "stdc_trailing_zeros_us": {
+      "defined": "7.18.5"
+    },
+    "stdc_trailing_zeros_ui": {
+      "defined": "7.18.5"
+    },
+    "stdc_trailing_zeros_ul": {
+      "defined": "7.18.5"
+    },
+    "stdc_trailing_zeros_ull": {
+      "defined": "7.18.5"
+    },
+    "stdc_trailing_ones_uc": {
+      "defined": "7.18.6"
+    },
+    "stdc_trailing_ones_us": {
+      "defined": "7.18.6"
+    },
+    "stdc_trailing_ones_ui": {
+      "defined": "7.18.6"
+    },
+    "stdc_trailing_ones_ul": {
+      "defined": "7.18.6"
+    },
+    "stdc_trailing_ones_ull": {
+      "defined": "7.18.6"
+    },
+    "stdc_first_leading_zero_uc": {
+      "defined": "7.18.7"
+    },
+    "stdc_first_leading_zero_us": {
+      "defined": "7.18.7"
+    },
+    "stdc_first_leading_zero_ui": {
+      "defined": "7.18.7"
+    },
+    "stdc_first_leading_zero_ul": {
+      "defined": "7.18.7"
+    },
+    "stdc_first_leading_zero_ull": {
+      "defined": "7.18.7"
+    },
+    "stdc_first_leading_one_uc": {
+      "defined": "7.18.8"
+    },
+    "stdc_first_leading_one_us": {
+      "defined": "7.18.8"
+    },
+    "stdc_first_leading_one_ui": {
+      "defined": "7.18.8"
+    },
+    "stdc_first_leading_one_ul": {
+      "defined": "7.18.8"
+    },
+    "stdc_first_leading_one_ull": {
+      "defined": "7.18.8"
+    },
+    "stdc_first_trailing_zero_uc": {
+      "defined": "7.18.9"
+    },
+    "stdc_first_trailing_zero_us": {
+      "defined": "7.18.9"
+    },
+    "stdc_first_trailing_zero_ui": {
+      "defined": "7.18.9"
+    },
+    "stdc_first_trailing_zero_ul": {
+      "defined": "7.18.9"
+    },
+    "stdc_first_trailing_zero_ull": {
+      "defined": "7.18.9"
+    },
+    "stdc_first_trailing_one_uc": {
+      "defined": "7.18.10"
+    },
+    "stdc_first_trailing_one_us": {
+      "defined": "7.18.10"
+    },
+    "stdc_first_trailing_one_ui": {
+      "defined": "7.18.10"
+    },
+    "stdc_first_trailing_one_ul": {
+      "defined": "7.18.10"
+    },
+    "stdc_first_trailing_one_ull": {
+      "defined": "7.18.10"
+    },
+    "stdc_count_zeros_uc": {
+      "defined": "7.18.11"
+    },
+    "stdc_count_zeros_us": {
+      "defined": "7.18.11"
+    },
+    "stdc_count_zeros_ui": {
+      "defined": "7.18.11"
+    },
+    "stdc_count_zeros_ul": {
+      "defined": "7.18.11"
+    },
+    "stdc_count_zeros_ull": {
+      "defined": "7.18.11"
+    },
+    "stdc_count_ones_uc": {
+      "defined": "7.18.12"
+    },
+    "stdc_count_ones_us": {
+      "defined": "7.18.12"
+    },
+    "stdc_count_ones_ui": {
+      "defined": "7.18.12"
+    },
+    "stdc_count_ones_ul": {
+      "defined": "7.18.12"
+    },
+    "stdc_count_ones_ull": {
+      "defined": "7.18.12"
+    },
+    "stdc_has_single_bit_uc": {
+      "defined": "7.18.13"
+    },
+    "stdc_has_single_bit_us": {
+      "defined": "7.18.13"
+    },
+    "stdc_has_single_bit_ui": {
+      "defined": "7.18.13"
+    },
+    "stdc_has_single_bit_ul": {
+      "defined": "7.18.13"
+    },
+    "stdc_has_single_bit_ull": {
+      "defined": "7.18.13"
+    },
+    "stdc_bit_width_uc": {
+      "defined": "7.18.14"
+    },
+    "stdc_bit_width_us": {
+      "defined": "7.18.14"
+    },
+    "stdc_bit_width_ui": {
+      "defined": "7.18.14"
+    },
+    "stdc_bit_width_ul": {
+      "defined": "7.18.14"
+    },
+    "stdc_bit_width_ull": {
+      "defined": "7.18.14"
+    },
+    "stdc_bit_floor_uc": {
+      "defined": "7.18.15"
+    },
+    "stdc_bit_floor_us": {
+      "defined": "7.18.15"
+    },
+    "stdc_bit_floor_ui": {
+      "defined": "7.18.15"
+    },
+    "stdc_bit_floor_ul": {
+      "defined": "7.18.15"
+    },
+    "stdc_bit_floor_ull": {
+      "defined": "7.18.15"
+    },
+    "stdc_bit_ceil_uc": {
+      "defined": "7.18.16"
+    },
+    "stdc_bit_ceil_us": {
+      "defined": "7.18.16"
+    },
+    "stdc_bit_ceil_ui": {
+      "defined": "7.18.16"
+    },
+    "stdc_bit_ceil_ul": {
+      "defined": "7.18.16"
+    },
+    "stdc_bit_ceil_ull": {
+      "defined": "7.18.16"
+    }
+  }
+}

From eaa3947106e50d6d837e6681fb55cf92c2bb0f9a Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Fri, 12 Apr 2024 14:32:37 -0700
Subject: [PATCH 306/886] [NFC] Remove unused type alias in OpenACC

Leftover from a previous commit, this ends up not being used, so remove
it.
---
 clang/include/clang/Parse/Parser.h | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/clang/include/clang/Parse/Parser.h b/clang/include/clang/Parse/Parser.h
index db90d38c7dacf..c719218731c35 100644
--- a/clang/include/clang/Parse/Parser.h
+++ b/clang/include/clang/Parse/Parser.h
@@ -3611,9 +3611,6 @@ class Parser : public CodeCompletionHandler {
   OpenACCClauseParseResult OpenACCCannotContinue();
   OpenACCClauseParseResult OpenACCSuccess(OpenACCClause *Clause);
 
-  using OpenACCConditionExprParseResult =
-      std::pair<ExprResult, OpenACCParseCanContinue>;
-
   /// Parses the OpenACC directive (the entire pragma) including the clause
   /// list, but does not produce the main AST node.
   OpenACCDirectiveParseInfo ParseOpenACCDirective();

From a472e6473e6921f22427d41c3d29e9a26eb4ab32 Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Fri, 12 Apr 2024 14:41:20 -0700
Subject: [PATCH 307/886] [NFC][OpenACC] Rename visitor functions to omit
 redundant OpenACC

While working on a followup patch, it became clear that this extra bit
of 'OpenACC' before each clause name was redundant with the visitors, so
remove it to make this a little less verbose.
---
 clang/include/clang/AST/OpenACCClause.h | 11 +++++------
 clang/lib/AST/OpenACCClause.cpp         |  4 ++--
 clang/lib/AST/StmtProfile.cpp           |  6 +++---
 clang/tools/libclang/CIndex.cpp         |  5 ++---
 4 files changed, 12 insertions(+), 14 deletions(-)

diff --git a/clang/include/clang/AST/OpenACCClause.h b/clang/include/clang/AST/OpenACCClause.h
index 712ad46c0f238..401b8e904a1b7 100644
--- a/clang/include/clang/AST/OpenACCClause.h
+++ b/clang/include/clang/AST/OpenACCClause.h
@@ -160,10 +160,10 @@ template <class Impl> class OpenACCClauseVisitor {
 
     switch (C->getClauseKind()) {
     case OpenACCClauseKind::Default:
-      VisitOpenACCDefaultClause(*cast<OpenACCDefaultClause>(C));
+      VisitDefaultClause(*cast<OpenACCDefaultClause>(C));
       return;
     case OpenACCClauseKind::If:
-      VisitOpenACCIfClause(*cast<OpenACCIfClause>(C));
+      VisitIfClause(*cast<OpenACCIfClause>(C));
       return;
     case OpenACCClauseKind::Finalize:
     case OpenACCClauseKind::IfPresent:
@@ -212,9 +212,9 @@ template <class Impl> class OpenACCClauseVisitor {
   }
 
 #define VISIT_CLAUSE(CLAUSE_NAME)                                              \
-  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
+  void Visit##CLAUSE_NAME##Clause(                                             \
       const OpenACC##CLAUSE_NAME##Clause &Clause) {                            \
-    return getDerived().VisitOpenACC##CLAUSE_NAME##Clause(Clause);             \
+    return getDerived().Visit##CLAUSE_NAME##Clause(Clause);                    \
   }
 
 #include "clang/Basic/OpenACCClauses.def"
@@ -236,8 +236,7 @@ class OpenACCClausePrinter final
   OpenACCClausePrinter(raw_ostream &OS) : OS(OS) {}
 
 #define VISIT_CLAUSE(CLAUSE_NAME)                                              \
-  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
-      const OpenACC##CLAUSE_NAME##Clause &Clause);
+  void Visit##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &Clause);
 #include "clang/Basic/OpenACCClauses.def"
 };
 
diff --git a/clang/lib/AST/OpenACCClause.cpp b/clang/lib/AST/OpenACCClause.cpp
index 2bc8d14fa129b..446e470032047 100644
--- a/clang/lib/AST/OpenACCClause.cpp
+++ b/clang/lib/AST/OpenACCClause.cpp
@@ -65,11 +65,11 @@ OpenACCClause::child_range OpenACCClause::children() {
 //===----------------------------------------------------------------------===//
 //  OpenACC clauses printing methods
 //===----------------------------------------------------------------------===//
-void OpenACCClausePrinter::VisitOpenACCDefaultClause(
+void OpenACCClausePrinter::VisitDefaultClause(
     const OpenACCDefaultClause &C) {
   OS << "default(" << C.getDefaultClauseKind() << ")";
 }
 
-void OpenACCClausePrinter::VisitOpenACCIfClause(const OpenACCIfClause &C) {
+void OpenACCClausePrinter::VisitIfClause(const OpenACCIfClause &C) {
   OS << "if(" << C.getConditionExpr() << ")";
 }
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 44b635bfb8433..68ba7475f4ca4 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2459,17 +2459,17 @@ class OpenACCClauseProfiler
   }
 
 #define VISIT_CLAUSE(CLAUSE_NAME)                                              \
-  void VisitOpenACC##CLAUSE_NAME##Clause(                                      \
+  void Visit##CLAUSE_NAME##Clause(                                      \
       const OpenACC##CLAUSE_NAME##Clause &Clause);
 
 #include "clang/Basic/OpenACCClauses.def"
 };
 
 /// Nothing to do here, there are no sub-statements.
-void OpenACCClauseProfiler::VisitOpenACCDefaultClause(
+void OpenACCClauseProfiler::VisitDefaultClause(
     const OpenACCDefaultClause &Clause) {}
 
-void OpenACCClauseProfiler::VisitOpenACCIfClause(
+void OpenACCClauseProfiler::VisitIfClause(
     const OpenACCIfClause &Clause) {
   assert(Clause.hasConditionExpr() &&
          "if clause requires a valid condition expr");
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index 936e8872513ae..b999750c0f5e3 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -2787,9 +2787,8 @@ class OpenACCClauseEnqueue : public OpenACCClauseVisitor<OpenACCClauseEnqueue> {
 #include "clang/Basic/OpenACCClauses.def"
 };
 
-void OpenACCClauseEnqueue::VisitOpenACCDefaultClause(
-    const OpenACCDefaultClause &C) {}
-void OpenACCClauseEnqueue::VisitOpenACCIfClause(const OpenACCIfClause &C) {
+void OpenACCClauseEnqueue::VisitDefaultClause(const OpenACCDefaultClause &C) {}
+void OpenACCClauseEnqueue::VisitIfClause(const OpenACCIfClause &C) {
   Visitor.AddStmt(C.getConditionExpr());
 }
 } // namespace

From db8e182952573fa9198424a6ed1b1bb1676eae62 Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Fri, 12 Apr 2024 14:44:20 -0700
Subject: [PATCH 308/886] [NFC] Clang-format previous OpenACC Patch

---
 clang/lib/AST/OpenACCClause.cpp | 3 +--
 clang/lib/AST/StmtProfile.cpp   | 6 ++----
 2 files changed, 3 insertions(+), 6 deletions(-)

diff --git a/clang/lib/AST/OpenACCClause.cpp b/clang/lib/AST/OpenACCClause.cpp
index 446e470032047..dcb512cb51417 100644
--- a/clang/lib/AST/OpenACCClause.cpp
+++ b/clang/lib/AST/OpenACCClause.cpp
@@ -65,8 +65,7 @@ OpenACCClause::child_range OpenACCClause::children() {
 //===----------------------------------------------------------------------===//
 //  OpenACC clauses printing methods
 //===----------------------------------------------------------------------===//
-void OpenACCClausePrinter::VisitDefaultClause(
-    const OpenACCDefaultClause &C) {
+void OpenACCClausePrinter::VisitDefaultClause(const OpenACCDefaultClause &C) {
   OS << "default(" << C.getDefaultClauseKind() << ")";
 }
 
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 68ba7475f4ca4..d2aac1e640380 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2459,8 +2459,7 @@ class OpenACCClauseProfiler
   }
 
 #define VISIT_CLAUSE(CLAUSE_NAME)                                              \
-  void Visit##CLAUSE_NAME##Clause(                                      \
-      const OpenACC##CLAUSE_NAME##Clause &Clause);
+  void Visit##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &Clause);
 
 #include "clang/Basic/OpenACCClauses.def"
 };
@@ -2469,8 +2468,7 @@ class OpenACCClauseProfiler
 void OpenACCClauseProfiler::VisitDefaultClause(
     const OpenACCDefaultClause &Clause) {}
 
-void OpenACCClauseProfiler::VisitIfClause(
-    const OpenACCIfClause &Clause) {
+void OpenACCClauseProfiler::VisitIfClause(const OpenACCIfClause &Clause) {
   assert(Clause.hasConditionExpr() &&
          "if clause requires a valid condition expr");
   Profiler.VisitStmt(Clause.getConditionExpr());

From 779ba60417b467a6d2d25101b11711c009694315 Mon Sep 17 00:00:00 2001
From: Artem Chikin <achikin@apple.com>
Date: Fri, 12 Apr 2024 14:46:56 -0700
Subject: [PATCH 309/886] [clang][deps] Overload `Filesystem::exists` in
 `DependencyScanningFilesystem` to have it use cached `status` (#88152)

As-is, calls to `exists()` fallback on the implementation in
`ProxyFileSystem::exists` which explicitly calls out to the underlying
`FS`, which for the `DependencyScanningFilesystem` (overlay) is the real
underlying filesystem.

Instead, directly overloading `exists` allows us to have it rely on the
cached `status` behavior used elsewhere by the
`DependencyScanningFilesystem`.
---
 .../DependencyScanningFilesystem.h            |  4 +++
 .../DependencyScanningFilesystem.cpp          | 11 ++++++++
 .../DependencyScanningFilesystemTest.cpp      | 27 +++++++++++++++++++
 3 files changed, 42 insertions(+)

diff --git a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
index 8b6f149c7cb26..f7b4510d7f7be 100644
--- a/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
+++ b/clang/include/clang/Tooling/DependencyScanning/DependencyScanningFilesystem.h
@@ -365,6 +365,10 @@ class DependencyScanningWorkerFilesystem
   /// false if not (i.e. this entry is not a file or its scan fails).
   bool ensureDirectiveTokensArePopulated(EntryRef Entry);
 
+  /// Check whether \p Path exists. By default checks cached result of \c
+  /// status(), and falls back on FS if unable to do so.
+  bool exists(const Twine &Path) override;
+
 private:
   /// For a filename that's not yet associated with any entry in the caches,
   /// uses the underlying filesystem to either look up the entry based in the
diff --git a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
index 84185c931b552..0cab17a342440 100644
--- a/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
+++ b/clang/lib/Tooling/DependencyScanning/DependencyScanningFilesystem.cpp
@@ -300,6 +300,17 @@ DependencyScanningWorkerFilesystem::status(const Twine &Path) {
   return Result->getStatus();
 }
 
+bool DependencyScanningWorkerFilesystem::exists(const Twine &Path) {
+  // While some VFS overlay filesystems may implement more-efficient
+  // mechanisms for `exists` queries, `DependencyScanningWorkerFilesystem`
+  // typically wraps `RealFileSystem` which does not specialize `exists`,
+  // so it is not likely to benefit from such optimizations. Instead,
+  // it is more-valuable to have this query go through the
+  // cached-`status` code-path of the `DependencyScanningWorkerFilesystem`.
+  llvm::ErrorOr<llvm::vfs::Status> Status = status(Path);
+  return Status && Status->exists();
+}
+
 namespace {
 
 /// The VFS that is used by clang consumes the \c CachedFileSystemEntry using
diff --git a/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
index 697b7d70ff035..87bb67cfd9327 100644
--- a/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
+++ b/clang/unittests/Tooling/DependencyScanning/DependencyScanningFilesystemTest.cpp
@@ -18,6 +18,7 @@ struct InstrumentingFilesystem
     : llvm::RTTIExtends<InstrumentingFilesystem, llvm::vfs::ProxyFileSystem> {
   unsigned NumStatusCalls = 0;
   unsigned NumGetRealPathCalls = 0;
+  unsigned NumExistsCalls = 0;
 
   using llvm::RTTIExtends<InstrumentingFilesystem,
                           llvm::vfs::ProxyFileSystem>::RTTIExtends;
@@ -32,6 +33,11 @@ struct InstrumentingFilesystem
     ++NumGetRealPathCalls;
     return ProxyFileSystem::getRealPath(Path, Output);
   }
+
+  bool exists(const llvm::Twine &Path) override {
+    ++NumExistsCalls;
+    return ProxyFileSystem::exists(Path);
+  }
 };
 } // namespace
 
@@ -147,3 +153,24 @@ TEST(DependencyScanningFilesystem, RealPathAndStatusInvariants) {
     DepFS.status("/bar");
   }
 }
+
+TEST(DependencyScanningFilesystem, CacheStatOnExists) {
+  auto InMemoryFS = llvm::makeIntrusiveRefCnt<llvm::vfs::InMemoryFileSystem>();
+  auto InstrumentingFS =
+      llvm::makeIntrusiveRefCnt<InstrumentingFilesystem>(InMemoryFS);
+  InMemoryFS->setCurrentWorkingDirectory("/");
+  InMemoryFS->addFile("/foo", 0, llvm::MemoryBuffer::getMemBuffer(""));
+  InMemoryFS->addFile("/bar", 0, llvm::MemoryBuffer::getMemBuffer(""));
+  DependencyScanningFilesystemSharedCache SharedCache;
+  DependencyScanningWorkerFilesystem DepFS(SharedCache, InstrumentingFS);
+
+  DepFS.status("/foo");
+  DepFS.status("/foo");
+  DepFS.status("/bar");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 2u);
+
+  DepFS.exists("/foo");
+  DepFS.exists("/bar");
+  EXPECT_EQ(InstrumentingFS->NumStatusCalls, 2u);
+  EXPECT_EQ(InstrumentingFS->NumExistsCalls, 0u);
+}

From 03c393db1821ef4c362c4c228492c6fb03f8876f Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Fri, 12 Apr 2024 14:48:39 -0700
Subject: [PATCH 310/886] Fix build failure from a472e6473e69

Looks like I forgot to do build CIndex.cpp when validating myself!
---
 clang/tools/libclang/CIndex.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index b999750c0f5e3..f304786ff9dff 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -2783,7 +2783,7 @@ class OpenACCClauseEnqueue : public OpenACCClauseVisitor<OpenACCClauseEnqueue> {
   OpenACCClauseEnqueue(EnqueueVisitor &V) : Visitor(V) {}
 
 #define VISIT_CLAUSE(CLAUSE_NAME)                                              \
-  void VisitOpenACC##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &C);
+  void Visit##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &C);
 #include "clang/Basic/OpenACCClauses.def"
 };
 

From 0d2bb7f017f13ceae793fab7d83d3e67e8d8d8f8 Mon Sep 17 00:00:00 2001
From: Tom Stellard <tstellar@redhat.com>
Date: Fri, 12 Apr 2024 15:10:04 -0700
Subject: [PATCH 311/886] [CMake][Release] Disable PGO (#88465)

Running the test-release.sh script with PGO enabled causes build errors
like:

ld.lld: error: Function Import: link error: linking module flags
'ProfileSummary': IDs have conflicting values

I believe this a build system bug due to the PGO profile data being
generated unconditionally. If you run `ninja check-all` and then `ninja
install` like we do in test-release.sh, then the profile data is
regenerated during `ninja install` and some of the clang tools which are
not test dependencies get build during the ninja install step with
different profile data. When these tools link against the LLVM
libraries, like libSupport, we end up with these errors.
---
 clang/cmake/caches/Release.cmake | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/cmake/caches/Release.cmake b/clang/cmake/caches/Release.cmake
index 1ca9138b98073..bd1f688d61a7e 100644
--- a/clang/cmake/caches/Release.cmake
+++ b/clang/cmake/caches/Release.cmake
@@ -4,7 +4,7 @@
 
 # General Options
 set(LLVM_RELEASE_ENABLE_LTO THIN CACHE STRING "")
-set(LLVM_RELEASE_ENABLE_PGO ON CACHE BOOL "")
+set(LLVM_RELEASE_ENABLE_PGO OFF CACHE BOOL "")
 
 set(CMAKE_BUILD_TYPE RELEASE CACHE STRING "")
 

From 9bd10853e505b61f5fe2e3b3592c76787a06aa31 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Sat, 13 Apr 2024 00:24:12 +0200
Subject: [PATCH 312/886] AMDGPU: Undo atomicrmw add/sub/xor 0 -> atomicrmw or
 canonicalization (#87533)

InstCombine transforms add of 0 to or of 0. For system atomics, this is
problematic because while PCIe supports add, it does not support the
other operations. Undo this for system scope atomics.
---
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     |  27 +-
 .../AMDGPU/global_atomics_i32_system.ll       | 723 ++++++++++--------
 ...expand-atomicrmw-integer-ops-0-to-add-0.ll | 130 ++++
 3 files changed, 572 insertions(+), 308 deletions(-)
 create mode 100644 llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomicrmw-integer-ops-0-to-add-0.ll

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 14948ef9ea8d1..6fde73119f815 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -16027,6 +16027,19 @@ SITargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *RMW) const {
       SSID == RMW->getContext().getOrInsertSyncScopeID("one-as");
 
   switch (RMW->getOperation()) {
+  case AtomicRMWInst::Sub:
+  case AtomicRMWInst::Or:
+  case AtomicRMWInst::Xor: {
+    // Atomic sub/or/xor do not work over PCI express, but atomic add
+    // does. InstCombine transforms these with 0 to or, so undo that.
+    if (HasSystemScope && AMDGPU::isFlatGlobalAddrSpace(AS)) {
+      if (Constant *ConstVal = dyn_cast<Constant>(RMW->getValOperand());
+          ConstVal && ConstVal->isNullValue())
+        return AtomicExpansionKind::Expand;
+    }
+
+    break;
+  }
   case AtomicRMWInst::FAdd: {
     Type *Ty = RMW->getType();
 
@@ -16312,14 +16325,24 @@ bool SITargetLowering::checkForPhysRegDependency(
 }
 
 void SITargetLowering::emitExpandAtomicRMW(AtomicRMWInst *AI) const {
+  AtomicRMWInst::BinOp Op = AI->getOperation();
+
+  if (Op == AtomicRMWInst::Sub || Op == AtomicRMWInst::Or ||
+      Op == AtomicRMWInst::Xor) {
+    // atomicrmw or %ptr, 0 -> atomicrmw add %ptr, 0
+    assert(cast<Constant>(AI->getValOperand())->isNullValue() &&
+           "this cannot be replaced with add");
+    AI->setOperation(AtomicRMWInst::Add);
+    return;
+  }
+
   assert(Subtarget->hasAtomicFaddInsts() &&
          "target should have atomic fadd instructions");
   assert(AI->getType()->isFloatTy() &&
          AI->getPointerAddressSpace() == AMDGPUAS::FLAT_ADDRESS &&
          "generic atomicrmw expansion only supports FP32 operand in flat "
          "address space");
-  assert(AI->getOperation() == AtomicRMWInst::FAdd &&
-         "only fadd is supported for now");
+  assert(Op == AtomicRMWInst::FAdd && "only fadd is supported for now");
 
   // Given: atomicrmw fadd ptr %addr, float %val ordering
   //
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
index 99d02ffaa523d..4598d23e088b5 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
@@ -1403,6 +1403,43 @@ define amdgpu_gfx i32 @global_atomic_sub_i32_ret_offset_scalar(ptr addrspace(1)
   ret i32 %result
 }
 
+define i32 @global_atomic_sub_0_i32_ret(ptr addrspace(1) %ptr) {
+; SI-LABEL: global_atomic_sub_0_i32_ret:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, 0
+; SI-NEXT:    v_mov_b32_e32 v2, 0
+; SI-NEXT:    s_mov_b32 s4, s6
+; SI-NEXT:    s_mov_b32 s5, s6
+; SI-NEXT:    buffer_atomic_add v2, v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_wbinvl1
+; SI-NEXT:    v_mov_b32_e32 v0, v2
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    s_setpc_b64 s[30:31]
+;
+; VI-LABEL: global_atomic_sub_0_i32_ret:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, 0
+; VI-NEXT:    flat_atomic_add v0, v[0:1], v2 glc
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    buffer_wbinvl1_vol
+; VI-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-LABEL: global_atomic_sub_0_i32_ret:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    global_atomic_add v0, v[0:1], v2, off glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_wbinvl1_vol
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+  %result = atomicrmw sub ptr addrspace(1) %ptr, i32 0 seq_cst
+  ret i32 %result
+}
+
 ; ---------------------------------------------------------------------
 ; atomicrmw and
 ; ---------------------------------------------------------------------
@@ -1767,7 +1804,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB40_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB41_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1782,7 +1819,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB40_1
+; SI-NEXT:    s_cbranch_execnz .LBB41_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -1793,7 +1830,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB40_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB41_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1805,7 +1842,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB40_1
+; VI-NEXT:    s_cbranch_execnz .LBB41_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -1815,7 +1852,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB40_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB41_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1827,7 +1864,7 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB40_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB41_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -1845,7 +1882,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB41_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB42_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1860,7 +1897,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB41_1
+; SI-NEXT:    s_cbranch_execnz .LBB42_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -1873,7 +1910,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB41_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB42_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1885,7 +1922,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB41_1
+; VI-NEXT:    s_cbranch_execnz .LBB42_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -1895,7 +1932,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB41_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB42_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, v4, v2
@@ -1907,7 +1944,7 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB41_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB42_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -1926,7 +1963,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB42_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB43_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -1941,7 +1978,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB42_1
+; SI-NEXT:    s_cbranch_execnz .LBB43_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -1953,7 +1990,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB42_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB43_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
@@ -1965,7 +2002,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB42_1
+; VI-NEXT:    s_cbranch_execnz .LBB43_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, v3
@@ -1976,7 +2013,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB42_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB43_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -1988,7 +2025,7 @@ define i32 @global_atomic_nand_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB42_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB43_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -2007,7 +2044,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB43_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB44_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -2022,7 +2059,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB43_1
+; SI-NEXT:    s_cbranch_execnz .LBB44_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -2036,7 +2073,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v0, v[3:4]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB43_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB44_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, v0
@@ -2048,7 +2085,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB43_1
+; VI-NEXT:    s_cbranch_execnz .LBB44_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -2058,7 +2095,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB43_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB44_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -2070,7 +2107,7 @@ define i32 @global_atomic_nand_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB43_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB44_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -2095,7 +2132,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB44_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB45_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_and_b32_e32 v0, s34, v1
@@ -2110,7 +2147,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB44_1
+; SI-NEXT:    s_cbranch_execnz .LBB45_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -2128,7 +2165,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB44_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB45_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_and_b32_e32 v2, s6, v3
@@ -2140,7 +2177,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB44_1
+; VI-NEXT:    s_cbranch_execnz .LBB45_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -2151,7 +2188,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB44_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB45_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v0, s6, v1
@@ -2163,7 +2200,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB44_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB45_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -2186,7 +2223,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB45_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB46_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_and_b32_e32 v0, s34, v1
@@ -2201,7 +2238,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB45_1
+; SI-NEXT:    s_cbranch_execnz .LBB46_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -2221,7 +2258,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    v_mov_b32_e32 v1, s35
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB45_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB46_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_and_b32_e32 v2, s6, v3
@@ -2233,7 +2270,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB45_1
+; VI-NEXT:    s_cbranch_execnz .LBB46_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -2244,7 +2281,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB45_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB46_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v0, s6, v1
@@ -2256,7 +2293,7 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB45_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB46_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -2280,7 +2317,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB46_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB47_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -2295,7 +2332,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB46_1
+; SI-NEXT:    s_cbranch_execnz .LBB47_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -2315,7 +2352,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB46_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB47_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -2327,7 +2364,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB46_1
+; VI-NEXT:    s_cbranch_execnz .LBB47_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -2338,7 +2375,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB46_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB47_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -2350,7 +2387,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB46_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB47_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -2373,7 +2410,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB47_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB48_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -2388,7 +2425,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB47_1
+; SI-NEXT:    s_cbranch_execnz .LBB48_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -2408,7 +2445,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v2, s35
 ; VI-NEXT:    flat_load_dword v0, v[1:2]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB47_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB48_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -2420,7 +2457,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB47_1
+; VI-NEXT:    s_cbranch_execnz .LBB48_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -2431,7 +2468,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB47_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB48_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -2443,7 +2480,7 @@ define amdgpu_gfx i32 @global_atomic_nand_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB47_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB48_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -2802,6 +2839,43 @@ define amdgpu_gfx i32 @global_atomic_or_i32_ret_offset_scalar(ptr addrspace(1) i
   ret i32 %result
 }
 
+define i32 @global_atomic_or_0_i32_ret(ptr addrspace(1) %ptr) {
+; SI-LABEL: global_atomic_or_0_i32_ret:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, 0
+; SI-NEXT:    v_mov_b32_e32 v2, 0
+; SI-NEXT:    s_mov_b32 s4, s6
+; SI-NEXT:    s_mov_b32 s5, s6
+; SI-NEXT:    buffer_atomic_add v2, v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_wbinvl1
+; SI-NEXT:    v_mov_b32_e32 v0, v2
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    s_setpc_b64 s[30:31]
+;
+; VI-LABEL: global_atomic_or_0_i32_ret:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, 0
+; VI-NEXT:    flat_atomic_add v0, v[0:1], v2 glc
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    buffer_wbinvl1_vol
+; VI-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-LABEL: global_atomic_or_0_i32_ret:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    global_atomic_add v0, v[0:1], v2, off glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_wbinvl1_vol
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+  %result = atomicrmw or ptr addrspace(1) %ptr, i32 0 seq_cst
+  ret i32 %result
+}
+
 ; ---------------------------------------------------------------------
 ; atomicrmw xor
 ; ---------------------------------------------------------------------
@@ -3152,6 +3226,43 @@ define amdgpu_gfx i32 @global_atomic_xor_i32_ret_offset_scalar(ptr addrspace(1)
   ret i32 %result
 }
 
+define i32 @global_atomic_xor_0_i32_ret(ptr addrspace(1) %ptr) {
+; SI-LABEL: global_atomic_xor_0_i32_ret:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s6, 0
+; SI-NEXT:    v_mov_b32_e32 v2, 0
+; SI-NEXT:    s_mov_b32 s4, s6
+; SI-NEXT:    s_mov_b32 s5, s6
+; SI-NEXT:    buffer_atomic_add v2, v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    buffer_wbinvl1
+; SI-NEXT:    v_mov_b32_e32 v0, v2
+; SI-NEXT:    s_waitcnt expcnt(0)
+; SI-NEXT:    s_setpc_b64 s[30:31]
+;
+; VI-LABEL: global_atomic_xor_0_i32_ret:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, 0
+; VI-NEXT:    flat_atomic_add v0, v[0:1], v2 glc
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    buffer_wbinvl1_vol
+; VI-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX9-LABEL: global_atomic_xor_0_i32_ret:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    global_atomic_add v0, v[0:1], v2, off glc
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    buffer_wbinvl1_vol
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+  %result = atomicrmw xor ptr addrspace(1) %ptr, i32 0 seq_cst
+  ret i32 %result
+}
+
 ; ---------------------------------------------------------------------
 ; atomicrmw max
 ; ---------------------------------------------------------------------
@@ -3166,7 +3277,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB64_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB67_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3180,7 +3291,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB64_1
+; SI-NEXT:    s_cbranch_execnz .LBB67_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -3191,7 +3302,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB64_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB67_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3202,7 +3313,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB64_1
+; VI-NEXT:    s_cbranch_execnz .LBB67_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3212,7 +3323,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB64_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB67_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3223,7 +3334,7 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB64_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB67_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3241,7 +3352,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB65_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB68_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3255,7 +3366,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB65_1
+; SI-NEXT:    s_cbranch_execnz .LBB68_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -3268,7 +3379,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB65_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB68_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3279,7 +3390,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB65_1
+; VI-NEXT:    s_cbranch_execnz .LBB68_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3289,7 +3400,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB65_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB68_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_i32_e32 v3, v4, v2
@@ -3300,7 +3411,7 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB65_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB68_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3319,7 +3430,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB66_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB69_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -3333,7 +3444,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB66_1
+; SI-NEXT:    s_cbranch_execnz .LBB69_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -3345,7 +3456,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB66_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB69_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
@@ -3356,7 +3467,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB66_1
+; VI-NEXT:    s_cbranch_execnz .LBB69_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, v3
@@ -3367,7 +3478,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB66_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB69_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -3378,7 +3489,7 @@ define i32 @global_atomic_max_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB66_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB69_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -3397,7 +3508,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB67_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB70_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -3411,7 +3522,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB67_1
+; SI-NEXT:    s_cbranch_execnz .LBB70_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -3425,7 +3536,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v0, v[3:4]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB67_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB70_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, v0
@@ -3436,7 +3547,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB67_1
+; VI-NEXT:    s_cbranch_execnz .LBB70_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3446,7 +3557,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB67_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB70_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -3457,7 +3568,7 @@ define i32 @global_atomic_max_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB67_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB70_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -3482,7 +3593,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB68_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB71_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_i32_e32 v0, s34, v1
@@ -3496,7 +3607,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB68_1
+; SI-NEXT:    s_cbranch_execnz .LBB71_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -3514,7 +3625,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB68_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB71_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_i32_e32 v2, s6, v3
@@ -3525,7 +3636,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB68_1
+; VI-NEXT:    s_cbranch_execnz .LBB71_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3536,7 +3647,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB68_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB71_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_i32_e32 v0, s6, v1
@@ -3547,7 +3658,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB68_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB71_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3570,7 +3681,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB69_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB72_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_i32_e32 v0, s34, v1
@@ -3584,7 +3695,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB69_1
+; SI-NEXT:    s_cbranch_execnz .LBB72_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -3604,7 +3715,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; VI-NEXT:    v_mov_b32_e32 v1, s35
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB69_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB72_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_i32_e32 v2, s6, v3
@@ -3615,7 +3726,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB69_1
+; VI-NEXT:    s_cbranch_execnz .LBB72_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3626,7 +3737,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB69_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB72_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_i32_e32 v0, s6, v1
@@ -3637,7 +3748,7 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB69_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB72_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3661,7 +3772,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB70_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB73_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -3675,7 +3786,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB70_1
+; SI-NEXT:    s_cbranch_execnz .LBB73_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -3695,7 +3806,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB70_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB73_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -3706,7 +3817,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB70_1
+; VI-NEXT:    s_cbranch_execnz .LBB73_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3717,7 +3828,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB70_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB73_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -3728,7 +3839,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_scalar(ptr addrspace(1) inreg %
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB70_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB73_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3751,7 +3862,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB71_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB74_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -3765,7 +3876,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB71_1
+; SI-NEXT:    s_cbranch_execnz .LBB74_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -3785,7 +3896,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v2, s35
 ; VI-NEXT:    flat_load_dword v0, v[1:2]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB71_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB74_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -3796,7 +3907,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB71_1
+; VI-NEXT:    s_cbranch_execnz .LBB74_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -3807,7 +3918,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB71_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB74_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -3818,7 +3929,7 @@ define amdgpu_gfx i32 @global_atomic_max_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB71_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB74_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -3843,7 +3954,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB72_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB75_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_i32_e32 v0, s2, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -3856,7 +3967,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB72_1
+; SI-NEXT:    s_cbranch_execnz .LBB75_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_endpgm
 ;
@@ -3877,7 +3988,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
-; VI-NEXT:  .LBB72_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB75_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_max_i32_e32 v2, s2, v3
 ; VI-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
@@ -3887,7 +3998,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB72_1
+; VI-NEXT:    s_cbranch_execnz .LBB75_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_endpgm
 ;
@@ -3905,7 +4016,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s3
-; GFX9-NEXT:  .LBB72_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB75_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_max_i32_e32 v0, s2, v1
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
@@ -3915,7 +4026,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB72_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB75_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
@@ -3942,7 +4053,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB73_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB76_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_i32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -3955,7 +4066,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB73_1
+; SI-NEXT:    s_cbranch_execnz .LBB76_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -3983,7 +4094,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
-; VI-NEXT:  .LBB73_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB76_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_max_i32_e32 v2, s4, v3
@@ -3993,7 +4104,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB73_1
+; VI-NEXT:    s_cbranch_execnz .LBB76_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -4016,7 +4127,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB73_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB76_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_max_i32_e32 v2, s2, v3
@@ -4026,7 +4137,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB73_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB76_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
@@ -4056,7 +4167,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB74_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB77_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_i32_e32 v0, s2, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -4069,7 +4180,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB74_1
+; SI-NEXT:    s_cbranch_execnz .LBB77_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_endpgm
 ;
@@ -4088,7 +4199,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
-; VI-NEXT:  .LBB74_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB77_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_max_i32_e32 v2, s2, v3
 ; VI-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
@@ -4098,7 +4209,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB74_1
+; VI-NEXT:    s_cbranch_execnz .LBB77_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_endpgm
 ;
@@ -4116,7 +4227,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s3
-; GFX9-NEXT:  .LBB74_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB77_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_max_i32_e32 v0, s2, v1
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
@@ -4126,7 +4237,7 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB74_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB77_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
@@ -4152,7 +4263,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB75_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB78_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_i32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -4165,7 +4276,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB75_1
+; SI-NEXT:    s_cbranch_execnz .LBB78_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -4191,7 +4302,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB75_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB78_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_max_i32_e32 v2, s4, v3
@@ -4201,7 +4312,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB75_1
+; VI-NEXT:    s_cbranch_execnz .LBB78_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -4224,7 +4335,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB75_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB78_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_max_i32_e32 v2, s2, v3
@@ -4234,7 +4345,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB75_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB78_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
@@ -4261,7 +4372,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB76_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB79_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4275,7 +4386,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB76_1
+; SI-NEXT:    s_cbranch_execnz .LBB79_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -4286,7 +4397,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB76_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB79_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4297,7 +4408,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB76_1
+; VI-NEXT:    s_cbranch_execnz .LBB79_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4307,7 +4418,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB76_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB79_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4318,7 +4429,7 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB76_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB79_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4336,7 +4447,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB77_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB80_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4350,7 +4461,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB77_1
+; SI-NEXT:    s_cbranch_execnz .LBB80_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -4363,7 +4474,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB77_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB80_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4374,7 +4485,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB77_1
+; VI-NEXT:    s_cbranch_execnz .LBB80_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4384,7 +4495,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB77_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB80_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_u32_e32 v3, v4, v2
@@ -4395,7 +4506,7 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB77_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB80_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4414,7 +4525,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB78_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB81_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -4428,7 +4539,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB78_1
+; SI-NEXT:    s_cbranch_execnz .LBB81_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -4440,7 +4551,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB78_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB81_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
@@ -4451,7 +4562,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB78_1
+; VI-NEXT:    s_cbranch_execnz .LBB81_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, v3
@@ -4462,7 +4573,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB78_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB81_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -4473,7 +4584,7 @@ define i32 @global_atomic_umax_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB78_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB81_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -4492,7 +4603,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB79_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB82_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -4506,7 +4617,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB79_1
+; SI-NEXT:    s_cbranch_execnz .LBB82_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -4520,7 +4631,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v0, v[3:4]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB79_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB82_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, v0
@@ -4531,7 +4642,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB79_1
+; VI-NEXT:    s_cbranch_execnz .LBB82_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4541,7 +4652,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB79_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB82_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -4552,7 +4663,7 @@ define i32 @global_atomic_umax_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB79_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB82_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -4577,7 +4688,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB80_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB83_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_u32_e32 v0, s34, v1
@@ -4591,7 +4702,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB80_1
+; SI-NEXT:    s_cbranch_execnz .LBB83_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -4609,7 +4720,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB80_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB83_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_u32_e32 v2, s6, v3
@@ -4620,7 +4731,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB80_1
+; VI-NEXT:    s_cbranch_execnz .LBB83_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4631,7 +4742,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB80_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB83_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_u32_e32 v0, s6, v1
@@ -4642,7 +4753,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB80_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB83_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4665,7 +4776,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB81_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB84_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_max_u32_e32 v0, s34, v1
@@ -4679,7 +4790,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB81_1
+; SI-NEXT:    s_cbranch_execnz .LBB84_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -4699,7 +4810,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    v_mov_b32_e32 v1, s35
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB81_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB84_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_max_u32_e32 v2, s6, v3
@@ -4710,7 +4821,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB81_1
+; VI-NEXT:    s_cbranch_execnz .LBB84_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4721,7 +4832,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB81_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB84_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_max_u32_e32 v0, s6, v1
@@ -4732,7 +4843,7 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB81_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB84_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4756,7 +4867,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB82_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB85_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -4770,7 +4881,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB82_1
+; SI-NEXT:    s_cbranch_execnz .LBB85_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -4790,7 +4901,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB82_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB85_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -4801,7 +4912,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB82_1
+; VI-NEXT:    s_cbranch_execnz .LBB85_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4812,7 +4923,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB82_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB85_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -4823,7 +4934,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB82_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB85_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4846,7 +4957,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB83_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB86_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -4860,7 +4971,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB83_1
+; SI-NEXT:    s_cbranch_execnz .LBB86_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -4880,7 +4991,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v2, s35
 ; VI-NEXT:    flat_load_dword v0, v[1:2]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB83_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB86_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -4891,7 +5002,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB83_1
+; VI-NEXT:    s_cbranch_execnz .LBB86_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -4902,7 +5013,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB83_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB86_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -4913,7 +5024,7 @@ define amdgpu_gfx i32 @global_atomic_umax_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB83_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB86_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -4938,7 +5049,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB84_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB87_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_u32_e32 v0, s2, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -4951,7 +5062,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB84_1
+; SI-NEXT:    s_cbranch_execnz .LBB87_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_endpgm
 ;
@@ -4972,7 +5083,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
-; VI-NEXT:  .LBB84_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB87_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_max_u32_e32 v2, s2, v3
 ; VI-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
@@ -4982,7 +5093,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB84_1
+; VI-NEXT:    s_cbranch_execnz .LBB87_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_endpgm
 ;
@@ -5000,7 +5111,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s3
-; GFX9-NEXT:  .LBB84_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB87_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_max_u32_e32 v0, s2, v1
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
@@ -5010,7 +5121,7 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB84_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB87_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
@@ -5037,7 +5148,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB85_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB88_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_u32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -5050,7 +5161,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB85_1
+; SI-NEXT:    s_cbranch_execnz .LBB88_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -5078,7 +5189,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
-; VI-NEXT:  .LBB85_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB88_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_max_u32_e32 v2, s4, v3
@@ -5088,7 +5199,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB85_1
+; VI-NEXT:    s_cbranch_execnz .LBB88_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -5111,7 +5222,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB85_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB88_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_max_u32_e32 v2, s2, v3
@@ -5121,7 +5232,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB85_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB88_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
@@ -5152,7 +5263,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB86_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB89_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_max_u32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -5165,7 +5276,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB86_1
+; SI-NEXT:    s_cbranch_execnz .LBB89_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -5191,7 +5302,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB86_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB89_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_max_u32_e32 v2, s4, v3
@@ -5201,7 +5312,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB86_1
+; VI-NEXT:    s_cbranch_execnz .LBB89_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -5224,7 +5335,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB86_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB89_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_max_u32_e32 v2, s2, v3
@@ -5234,7 +5345,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB86_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB89_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
@@ -5261,7 +5372,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB87_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB90_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5275,7 +5386,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB87_1
+; SI-NEXT:    s_cbranch_execnz .LBB90_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -5286,7 +5397,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB87_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB90_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5297,7 +5408,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB87_1
+; VI-NEXT:    s_cbranch_execnz .LBB90_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5307,7 +5418,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB87_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB90_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5318,7 +5429,7 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB87_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB90_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5336,7 +5447,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB88_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB91_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5350,7 +5461,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB88_1
+; SI-NEXT:    s_cbranch_execnz .LBB91_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -5363,7 +5474,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB88_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB91_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5374,7 +5485,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB88_1
+; VI-NEXT:    s_cbranch_execnz .LBB91_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5384,7 +5495,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB88_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB91_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_u32_e32 v3, v4, v2
@@ -5395,7 +5506,7 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB88_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB91_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5414,7 +5525,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB89_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB92_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -5428,7 +5539,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB89_1
+; SI-NEXT:    s_cbranch_execnz .LBB92_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -5440,7 +5551,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB89_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB92_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
@@ -5451,7 +5562,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB89_1
+; VI-NEXT:    s_cbranch_execnz .LBB92_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, v3
@@ -5462,7 +5573,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB89_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB92_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -5473,7 +5584,7 @@ define i32 @global_atomic_umin_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB89_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB92_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -5492,7 +5603,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB90_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB93_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -5506,7 +5617,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB90_1
+; SI-NEXT:    s_cbranch_execnz .LBB93_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -5520,7 +5631,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v0, v[3:4]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB90_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB93_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, v0
@@ -5531,7 +5642,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB90_1
+; VI-NEXT:    s_cbranch_execnz .LBB93_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5541,7 +5652,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB90_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB93_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -5552,7 +5663,7 @@ define i32 @global_atomic_umin_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB90_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB93_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -5577,7 +5688,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB91_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB94_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_u32_e32 v0, s34, v1
@@ -5591,7 +5702,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB91_1
+; SI-NEXT:    s_cbranch_execnz .LBB94_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -5609,7 +5720,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB91_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB94_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_u32_e32 v2, s6, v3
@@ -5620,7 +5731,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB91_1
+; VI-NEXT:    s_cbranch_execnz .LBB94_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5631,7 +5742,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB91_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB94_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_u32_e32 v0, s6, v1
@@ -5642,7 +5753,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB91_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB94_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5665,7 +5776,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB92_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB95_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_u32_e32 v0, s34, v1
@@ -5679,7 +5790,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB92_1
+; SI-NEXT:    s_cbranch_execnz .LBB95_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -5699,7 +5810,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    v_mov_b32_e32 v1, s35
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB92_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB95_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_u32_e32 v2, s6, v3
@@ -5710,7 +5821,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB92_1
+; VI-NEXT:    s_cbranch_execnz .LBB95_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5721,7 +5832,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB92_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB95_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_u32_e32 v0, s6, v1
@@ -5732,7 +5843,7 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB92_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB95_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5756,7 +5867,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB93_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB96_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -5770,7 +5881,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB93_1
+; SI-NEXT:    s_cbranch_execnz .LBB96_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -5790,7 +5901,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB93_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB96_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -5801,7 +5912,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB93_1
+; VI-NEXT:    s_cbranch_execnz .LBB96_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5812,7 +5923,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB93_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB96_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -5823,7 +5934,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_scalar(ptr addrspace(1) inreg
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB93_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB96_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5846,7 +5957,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB94_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB97_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -5860,7 +5971,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB94_1
+; SI-NEXT:    s_cbranch_execnz .LBB97_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -5880,7 +5991,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v2, s35
 ; VI-NEXT:    flat_load_dword v0, v[1:2]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB94_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB97_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -5891,7 +6002,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB94_1
+; VI-NEXT:    s_cbranch_execnz .LBB97_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5902,7 +6013,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB94_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB97_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -5913,7 +6024,7 @@ define amdgpu_gfx i32 @global_atomic_umin_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB94_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB97_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -5936,7 +6047,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB95_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB98_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -5950,7 +6061,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB95_1
+; SI-NEXT:    s_cbranch_execnz .LBB98_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -5961,7 +6072,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB95_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB98_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -5972,7 +6083,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB95_1
+; VI-NEXT:    s_cbranch_execnz .LBB98_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -5982,7 +6093,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB95_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB98_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -5993,7 +6104,7 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB95_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB98_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6011,7 +6122,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB96_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB99_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -6025,7 +6136,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v4, v5
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB96_1
+; SI-NEXT:    s_cbranch_execnz .LBB99_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -6038,7 +6149,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v4, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB96_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB99_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -6049,7 +6160,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB96_1
+; VI-NEXT:    s_cbranch_execnz .LBB99_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6059,7 +6170,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v4, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB96_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB99_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_i32_e32 v3, v4, v2
@@ -6070,7 +6181,7 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB96_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB99_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6089,7 +6200,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB97_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB100_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -6103,7 +6214,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB97_1
+; SI-NEXT:    s_cbranch_execnz .LBB100_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -6115,7 +6226,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB97_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB100_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v3
@@ -6126,7 +6237,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB97_1
+; VI-NEXT:    s_cbranch_execnz .LBB100_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    v_mov_b32_e32 v0, v3
@@ -6137,7 +6248,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB97_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB100_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -6148,7 +6259,7 @@ define i32 @global_atomic_min_i32_ret(ptr addrspace(1) %ptr, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB97_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB100_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -6167,7 +6278,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_mov_b32 s5, s6
 ; SI-NEXT:    buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
 ; SI-NEXT:    s_mov_b64 s[8:9], 0
-; SI-NEXT:  .LBB98_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB101_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v5, v3
@@ -6181,7 +6292,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v5
 ; SI-NEXT:    s_or_b64 s[8:9], vcc, s[8:9]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[8:9]
-; SI-NEXT:    s_cbranch_execnz .LBB98_1
+; SI-NEXT:    s_cbranch_execnz .LBB101_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; SI-NEXT:    v_mov_b32_e32 v0, v3
@@ -6195,7 +6306,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v1, vcc
 ; VI-NEXT:    flat_load_dword v0, v[3:4]
 ; VI-NEXT:    s_mov_b64 s[4:5], 0
-; VI-NEXT:  .LBB98_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB101_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v1, v0
@@ -6206,7 +6317,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
 ; VI-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; VI-NEXT:    s_cbranch_execnz .LBB98_1
+; VI-NEXT:    s_cbranch_execnz .LBB101_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6216,7 +6327,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off offset:16
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
-; GFX9-NEXT:  .LBB98_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB101_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v4, v3
@@ -6227,7 +6338,7 @@ define i32 @global_atomic_min_i32_ret_offset(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB98_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB101_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v3
@@ -6252,7 +6363,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB99_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB102_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_i32_e32 v0, s34, v1
@@ -6266,7 +6377,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB99_1
+; SI-NEXT:    s_cbranch_execnz .LBB102_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -6284,7 +6395,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB99_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB102_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_i32_e32 v2, s6, v3
@@ -6295,7 +6406,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB99_1
+; VI-NEXT:    s_cbranch_execnz .LBB102_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6306,7 +6417,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB99_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB102_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_i32_e32 v0, s6, v1
@@ -6317,7 +6428,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB99_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB102_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6340,7 +6451,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB100_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB103_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_min_i32_e32 v0, s34, v1
@@ -6354,7 +6465,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB100_1
+; SI-NEXT:    s_cbranch_execnz .LBB103_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v4, 1
@@ -6374,7 +6485,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; VI-NEXT:    v_mov_b32_e32 v1, s35
 ; VI-NEXT:    flat_load_dword v3, v[0:1]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB100_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB103_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_min_i32_e32 v2, s6, v3
@@ -6385,7 +6496,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB100_1
+; VI-NEXT:    s_cbranch_execnz .LBB103_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6396,7 +6507,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    global_load_dword v1, v2, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB100_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB103_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_min_i32_e32 v0, s6, v1
@@ -6407,7 +6518,7 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB100_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB103_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6431,7 +6542,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB101_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB104_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -6445,7 +6556,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB101_1
+; SI-NEXT:    s_cbranch_execnz .LBB104_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -6465,7 +6576,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; VI-NEXT:    v_mov_b32_e32 v1, s4
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB101_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB104_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -6476,7 +6587,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB101_1
+; VI-NEXT:    s_cbranch_execnz .LBB104_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6487,7 +6598,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5]
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB101_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB104_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -6498,7 +6609,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_scalar(ptr addrspace(1) inreg %
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB101_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB104_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6521,7 +6632,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    s_mov_b32 s6, -1
 ; SI-NEXT:    buffer_load_dword v0, off, s[4:7], 0 offset:16
 ; SI-NEXT:    s_mov_b64 s[36:37], 0
-; SI-NEXT:  .LBB102_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB105_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v2, v0
@@ -6535,7 +6646,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; SI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v2
 ; SI-NEXT:    s_or_b64 s[36:37], vcc, s[36:37]
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[36:37]
-; SI-NEXT:    s_cbranch_execnz .LBB102_1
+; SI-NEXT:    s_cbranch_execnz .LBB105_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[36:37]
 ; SI-NEXT:    v_readlane_b32 s7, v3, 1
@@ -6555,7 +6666,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_mov_b32_e32 v2, s35
 ; VI-NEXT:    flat_load_dword v0, v[1:2]
 ; VI-NEXT:    s_mov_b64 s[34:35], 0
-; VI-NEXT:  .LBB102_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB105_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v4, v0
@@ -6566,7 +6677,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
 ; VI-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; VI-NEXT:    s_cbranch_execnz .LBB102_1
+; VI-NEXT:    s_cbranch_execnz .LBB105_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; VI-NEXT:    s_setpc_b64 s[30:31]
@@ -6577,7 +6688,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    global_load_dword v0, v1, s[4:5] offset:16
 ; GFX9-NEXT:    s_mov_b64 s[34:35], 0
-; GFX9-NEXT:  .LBB102_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB105_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
@@ -6588,7 +6699,7 @@ define amdgpu_gfx i32 @global_atomic_min_i32_ret_offset_scalar(ptr addrspace(1)
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[34:35], vcc, s[34:35]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[34:35]
-; GFX9-NEXT:    s_cbranch_execnz .LBB102_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB105_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[34:35]
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
@@ -6613,7 +6724,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB103_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB106_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_min_i32_e32 v0, s2, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -6626,7 +6737,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB103_1
+; SI-NEXT:    s_cbranch_execnz .LBB106_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_endpgm
 ;
@@ -6647,7 +6758,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
-; VI-NEXT:  .LBB103_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB106_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_min_i32_e32 v2, s2, v3
 ; VI-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
@@ -6657,7 +6768,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB103_1
+; VI-NEXT:    s_cbranch_execnz .LBB106_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_endpgm
 ;
@@ -6675,7 +6786,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s3
-; GFX9-NEXT:  .LBB103_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB106_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_min_i32_e32 v0, s2, v1
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 glc
@@ -6685,7 +6796,7 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB103_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB106_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
@@ -6712,7 +6823,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB104_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB107_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_min_i32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -6725,7 +6836,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB104_1
+; SI-NEXT:    s_cbranch_execnz .LBB107_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -6753,7 +6864,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
-; VI-NEXT:  .LBB104_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB107_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_min_i32_e32 v2, s4, v3
@@ -6763,7 +6874,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB104_1
+; VI-NEXT:    s_cbranch_execnz .LBB107_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -6786,7 +6897,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB104_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB107_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_min_i32_e32 v2, s2, v3
@@ -6796,7 +6907,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB104_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB107_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
@@ -6822,7 +6933,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB105_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB108_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_min_i32_e32 v0, s2, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -6835,7 +6946,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB105_1
+; SI-NEXT:    s_cbranch_execnz .LBB108_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_endpgm
 ;
@@ -6850,7 +6961,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s3
-; VI-NEXT:  .LBB105_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB108_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_min_i32_e32 v2, s2, v3
 ; VI-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] glc
@@ -6860,7 +6971,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB105_1
+; VI-NEXT:    s_cbranch_execnz .LBB108_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_endpgm
 ;
@@ -6874,7 +6985,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_load_dword s5, s[2:3], 0x0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s5
-; GFX9-NEXT:  .LBB105_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB108_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_min_i32_e32 v0, s4, v1
 ; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[2:3] glc
@@ -6884,7 +6995,7 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
 ; GFX9-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, v0
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; GFX9-NEXT:    s_cbranch_execnz .LBB105_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB108_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_endpgm
 entry:
@@ -6909,7 +7020,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    v_mov_b32_e32 v1, s6
 ; SI-NEXT:    s_mov_b32 s6, -1
-; SI-NEXT:  .LBB106_1: ; %atomicrmw.start
+; SI-NEXT:  .LBB109_1: ; %atomicrmw.start
 ; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; SI-NEXT:    v_min_i32_e32 v0, s8, v1
 ; SI-NEXT:    s_waitcnt expcnt(0)
@@ -6922,7 +7033,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; SI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; SI-NEXT:    v_mov_b32_e32 v1, v2
 ; SI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; SI-NEXT:    s_cbranch_execnz .LBB106_1
+; SI-NEXT:    s_cbranch_execnz .LBB109_1
 ; SI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; SI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; SI-NEXT:    s_mov_b32 s7, 0xf000
@@ -6948,7 +7059,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_mov_b32_e32 v1, s7
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v2, s5
-; VI-NEXT:  .LBB106_1: ; %atomicrmw.start
+; VI-NEXT:  .LBB109_1: ; %atomicrmw.start
 ; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; VI-NEXT:    v_mov_b32_e32 v3, v2
 ; VI-NEXT:    v_min_i32_e32 v2, s4, v3
@@ -6958,7 +7069,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; VI-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v3
 ; VI-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
 ; VI-NEXT:    s_andn2_b64 exec, exec, s[0:1]
-; VI-NEXT:    s_cbranch_execnz .LBB106_1
+; VI-NEXT:    s_cbranch_execnz .LBB109_1
 ; VI-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; VI-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; VI-NEXT:    v_mov_b32_e32 v0, s2
@@ -6981,7 +7092,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, s3
-; GFX9-NEXT:  .LBB106_1: ; %atomicrmw.start
+; GFX9-NEXT:  .LBB109_1: ; %atomicrmw.start
 ; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v0
 ; GFX9-NEXT:    v_min_i32_e32 v2, s2, v3
@@ -6991,7 +7102,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v3
 ; GFX9-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
 ; GFX9-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX9-NEXT:    s_cbranch_execnz .LBB106_1
+; GFX9-NEXT:    s_cbranch_execnz .LBB109_1
 ; GFX9-NEXT:  ; %bb.2: ; %atomicrmw.end
 ; GFX9-NEXT:    s_or_b64 exec, exec, s[4:5]
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
diff --git a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomicrmw-integer-ops-0-to-add-0.ll b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomicrmw-integer-ops-0-to-add-0.ll
new file mode 100644
index 0000000000000..e32c218e9fe56
--- /dev/null
+++ b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomicrmw-integer-ops-0-to-add-0.ll
@@ -0,0 +1,130 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=atomic-expand %s | FileCheck %s
+
+; Test that system scoped atomicrmw or 0 is transformed to add 0.
+
+; Transform to add
+define i32 @test_atomicrmw_or_0_global_system(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(1) [[PTR]], i32 0 seq_cst, align 4, !foo.md [[META0:![0-9]+]]
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(1) %ptr, i32 0 seq_cst, !foo.md !0
+  ret i32 %res
+}
+
+; Transform to add
+define i32 @test_atomicrmw_or_0_global_one_as(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_global_one_as(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(1) [[PTR]], i32 0 syncscope("one-as") seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(1) %ptr, i32 0 syncscope("one-as") seq_cst
+  ret i32 %res
+}
+
+; Transform to add
+define i32 @test_atomicrmw_or_0_flat_system(ptr %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_flat_system(
+; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr [[PTR]], i32 0 seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr %ptr, i32 0 seq_cst
+  ret i32 %res
+}
+
+; Transform to add
+define i32 @test_atomicrmw_or_0_as999_system(ptr addrspace(999) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_as999_system(
+; CHECK-SAME: ptr addrspace(999) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(999) [[PTR]], i32 0 seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(999) %ptr, i32 0 seq_cst
+  ret i32 %res
+}
+
+; Leave as-is, only system scope should be changed.
+define i32 @test_atomicrmw_or_0_global_agent(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_global_agent(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw or ptr addrspace(1) [[PTR]], i32 0 syncscope("agent") seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(1) %ptr, i32 0 syncscope("agent") seq_cst
+  ret i32 %res
+}
+
+; Leave as-is, LDS atomics aren't relevant.
+define i32 @test_atomicrmw_or_0_local(ptr addrspace(3) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_0_local(
+; CHECK-SAME: ptr addrspace(3) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw or ptr addrspace(3) [[PTR]], i32 0 seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(3) %ptr, i32 0 seq_cst
+  ret i32 %res
+}
+
+; Leave non-0 values alone.
+define i32 @test_atomicrmw_or_1_global_system(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_1_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw or ptr addrspace(1) [[PTR]], i32 1 seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(1) %ptr, i32 1 seq_cst
+  ret i32 %res
+}
+
+define i32 @test_atomicrmw_or_var_global_system(ptr addrspace(1) %ptr, i32 %val) {
+; CHECK-LABEL: define i32 @test_atomicrmw_or_var_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]], i32 [[VAL:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw or ptr addrspace(1) [[PTR]], i32 [[VAL]] seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw or ptr addrspace(1) %ptr, i32 %val seq_cst
+  ret i32 %res
+}
+
+; Leave as-is
+define i32 @test_atomicrmw_add_0_global_system(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_add_0_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(1) [[PTR]], i32 0 seq_cst, align 4
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw add ptr addrspace(1) %ptr, i32 0 seq_cst
+  ret i32 %res
+}
+
+; Transform to add
+define i32 @test_atomicrmw_sub_0_global_system(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_sub_0_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(1) [[PTR]], i32 0 seq_cst, align 4, !foo.md [[META0]]
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw sub ptr addrspace(1) %ptr, i32 0 seq_cst, !foo.md !0
+  ret i32 %res
+}
+
+; Transform to add
+define i32 @test_atomicrmw_xor_0_global_system(ptr addrspace(1) %ptr) {
+; CHECK-LABEL: define i32 @test_atomicrmw_xor_0_global_system(
+; CHECK-SAME: ptr addrspace(1) [[PTR:%.*]]) {
+; CHECK-NEXT:    [[RES:%.*]] = atomicrmw add ptr addrspace(1) [[PTR]], i32 0 seq_cst, align 4, !foo.md [[META0]]
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %res = atomicrmw xor ptr addrspace(1) %ptr, i32 0 seq_cst, !foo.md !0
+  ret i32 %res
+}
+
+
+!0 = !{}
+;.
+; CHECK: [[META0]] = !{}
+;.

From b9bed1f1106e3116e8bea38806c511b57b956929 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Fri, 12 Apr 2024 16:44:46 -0700
Subject: [PATCH 313/886] [test] Improve tests for alias/ifunc attributes

---
 clang/test/Sema/alias-unused-win.cpp | 31 +++++++++++++++++
 clang/test/Sema/alias-unused.c       |  7 ----
 clang/test/Sema/alias-unused.cpp     | 50 ++++++++++++++++++++++++++++
 3 files changed, 81 insertions(+), 7 deletions(-)
 create mode 100644 clang/test/Sema/alias-unused-win.cpp
 delete mode 100644 clang/test/Sema/alias-unused.c
 create mode 100644 clang/test/Sema/alias-unused.cpp

diff --git a/clang/test/Sema/alias-unused-win.cpp b/clang/test/Sema/alias-unused-win.cpp
new file mode 100644
index 0000000000000..47c96d4117517
--- /dev/null
+++ b/clang/test/Sema/alias-unused-win.cpp
@@ -0,0 +1,31 @@
+// RUN: %clang_cc1 -triple %ms_abi_triple -Wunused -x c -verify %s
+// RUN: %clang_cc1 -triple %ms_abi_triple -Wunused -verify=expected,cxx %s
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+static int f(void) { return 42; } // cxx-warning{{unused function 'f'}}
+int g(void) __attribute__((alias("f")));
+
+static int foo [] = { 42, 0xDEAD }; // cxx-warning{{variable 'foo' is not needed and will not be emitted}}
+extern typeof(foo) bar __attribute__((unused, alias("foo")));
+
+static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
+static float __attribute__((overloadable)) f0(float x) { return x; } // expected-warning{{unused function 'f0'}}
+int g0(void) __attribute__((alias("?f0@@YAHH@Z")));
+
+#ifdef __cplusplus
+/// https://github.com/llvm/llvm-project/issues/88593
+/// We report a warning in C++ mode because the internal linkage `resolver` gets
+/// mangled as it does not have a language linkage. GCC does not mangle
+/// `resolver` or report a warning.
+static int f1() { return 42; } // cxx-warning{{unused function 'f1'}}
+int g1(void) __attribute__((alias("?f1@@YAHXZ")));
+}
+
+namespace ns {
+static int f3(int) { return 42; } // cxx-warning{{unused function 'f3'}}
+static int f3() { return 42; } // cxx-warning{{unused function 'f3'}}
+int g3() __attribute__((alias("?f3@ns@@YAHXZ")));
+}
+#endif
diff --git a/clang/test/Sema/alias-unused.c b/clang/test/Sema/alias-unused.c
deleted file mode 100644
index de9fc8cc73766..0000000000000
--- a/clang/test/Sema/alias-unused.c
+++ /dev/null
@@ -1,7 +0,0 @@
-// RUN: %clang_cc1 -triple x86_64-linux-gnu -Wunneeded-internal-declaration -x c -verify %s
-// expected-no-diagnostics
-static int f(void) { return 42; }
-int g(void) __attribute__((alias("f")));
-
-static int foo [] = { 42, 0xDEAD };
-extern typeof(foo) bar __attribute__((unused, alias("foo")));
diff --git a/clang/test/Sema/alias-unused.cpp b/clang/test/Sema/alias-unused.cpp
new file mode 100644
index 0000000000000..509d3e26225d7
--- /dev/null
+++ b/clang/test/Sema/alias-unused.cpp
@@ -0,0 +1,50 @@
+// RUN: %clang_cc1 -triple %itanium_abi_triple -Wunused -x c -verify %s
+// RUN: %clang_cc1 -triple %itanium_abi_triple -Wunused -verify=expected,cxx %s
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+static int f(void) { return 42; }
+int g(void) __attribute__((alias("f")));
+
+static int foo [] = { 42, 0xDEAD }; // cxx-warning{{variable 'foo' is not needed and will not be emitted}}
+extern typeof(foo) bar __attribute__((unused, alias("foo")));
+
+/// https://github.com/llvm/llvm-project/issues/88593
+/// We report a warning in C++ mode because the internal linkage `resolver` gets
+/// mangled as it does not have a language linkage. GCC does not mangle
+/// `resolver` or report a warning.
+static int (*resolver(void))(void) { return f; } // expected-warning{{unused function 'resolver'}}
+int ifunc(void) __attribute__((ifunc("resolver")));
+
+static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
+static float __attribute__((overloadable)) f0(float x) { return x; } // expected-warning{{unused function 'f0'}}
+int g0(void) __attribute__((alias("_ZL2f0i")));
+
+#ifdef __cplusplus
+static int f1() { return 42; } // expected-warning{{unused function 'f1'}}
+int g1(void) __attribute__((alias("_ZL2f1v")));
+}
+
+static int f2(int) { return 42; } // expected-warning{{unused function 'f2'}}
+static int f2() { return 42; } // expected-warning{{unused function 'f2'}}
+int g2() __attribute__((alias("_ZL2f2v")));
+
+static int (*resolver1())() { return f; } // expected-warning{{unused function 'resolver1'}}
+static int (*resolver1(int))() { return f; } // expected-warning{{unused function 'resolver1'}}
+int ifunc1() __attribute__((ifunc("_ZL9resolver1i")));
+
+/// TODO: We should report "unused function" for f3(int).
+namespace ns {
+static int f3(int) { return 42; } // cxx-warning{{unused function 'f3'}}
+static int f3() { return 42; } // cxx-warning{{unused function 'f3'}}
+int g3() __attribute__((alias("_ZN2nsL2f3Ev")));
+}
+
+template <class T>
+static void *f4(T) { return nullptr; }
+static void *f4() { return nullptr; } // cxx-warning{{unused function 'f4'}}
+extern void g4_int() __attribute__((ifunc("_ZL2f4IiEPvT_")));
+extern void g4_char() __attribute__((ifunc("_ZL2f4IcEPcT_"))); // rejected by CodeGen
+void *use4 = f4(0);
+#endif

From 9e95c4947d31670ddd3abad1051eda06729c96b3 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Sat, 13 Apr 2024 09:01:17 +0900
Subject: [PATCH 314/886] [DSE] Fix non-determinism due to address reuse
 (#84943)

The malloc->calloc fold creates a new MemoryAccess, which may end of at
the same address as a previously deleted access inside SkipStores.

To the most part, this is not a problem, because SkipStores is normally
only used together with MemDefs. Neither the old malloc access nor the
new calloc access will be part of MemDefs, so there is no problem here.

However, SkipStores is also used in one more place: In the main DSE
loop, ToCheck entries are checked against it. Fix this by not using
SkipStores here, and instead using a separate set to track deletions
inside this loop. This way it is not affected by the calloc optimization
that happens outside it.

This is all pretty ugly, but I haven't found another good way to fix it.
Suggestions welcome.

No test case as I don't have a reliable DSE-only test-case for this.

Fixes https://github.com/llvm/llvm-project/issues/84458.
---
 .../Scalar/DeadStoreElimination.cpp           | 22 ++++++++++++++-----
 1 file changed, 17 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
index bfc8bd5970bf2..ed4212d29cef7 100644
--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -1699,7 +1699,9 @@ struct DSEState {
 
   /// Delete dead memory defs and recursively add their operands to ToRemove if
   /// they became dead.
-  void deleteDeadInstruction(Instruction *SI) {
+  void
+  deleteDeadInstruction(Instruction *SI,
+                        SmallPtrSetImpl<MemoryAccess *> *Deleted = nullptr) {
     MemorySSAUpdater Updater(&MSSA);
     SmallVector<Instruction *, 32> NowDeadInsts;
     NowDeadInsts.push_back(SI);
@@ -1720,6 +1722,8 @@ struct DSEState {
         if (IsMemDef) {
           auto *MD = cast<MemoryDef>(MA);
           SkipStores.insert(MD);
+          if (Deleted)
+            Deleted->insert(MD);
           if (auto *SI = dyn_cast<StoreInst>(MD->getMemoryInst())) {
             if (SI->getValueOperand()->getType()->isPointerTy()) {
               const Value *UO = getUnderlyingObject(SI->getValueOperand());
@@ -2168,6 +2172,11 @@ static bool eliminateDeadStores(Function &F, AliasAnalysis &AA, MemorySSA &MSSA,
     unsigned PartialLimit = MemorySSAPartialStoreLimit;
     // Worklist of MemoryAccesses that may be killed by KillingDef.
     SmallSetVector<MemoryAccess *, 8> ToCheck;
+    // Track MemoryAccesses that have been deleted in the loop below, so we can
+    // skip them. Don't use SkipStores for this, which may contain reused
+    // MemoryAccess addresses.
+    SmallPtrSet<MemoryAccess *, 8> Deleted;
+    [[maybe_unused]] unsigned OrigNumSkipStores = State.SkipStores.size();
     ToCheck.insert(KillingDef->getDefiningAccess());
 
     bool Shortend = false;
@@ -2175,7 +2184,7 @@ static bool eliminateDeadStores(Function &F, AliasAnalysis &AA, MemorySSA &MSSA,
     // Check if MemoryAccesses in the worklist are killed by KillingDef.
     for (unsigned I = 0; I < ToCheck.size(); I++) {
       MemoryAccess *Current = ToCheck[I];
-      if (State.SkipStores.count(Current))
+      if (Deleted.contains(Current))
         continue;
 
       std::optional<MemoryAccess *> MaybeDeadAccess = State.getDomMemoryDef(
@@ -2222,7 +2231,7 @@ static bool eliminateDeadStores(Function &F, AliasAnalysis &AA, MemorySSA &MSSA,
           continue;
         LLVM_DEBUG(dbgs() << "DSE: Remove Dead Store:\n  DEAD: " << *DeadI
                           << "\n  KILLER: " << *KillingI << '\n');
-        State.deleteDeadInstruction(DeadI);
+        State.deleteDeadInstruction(DeadI, &Deleted);
         ++NumFastStores;
         MadeChange = true;
       } else {
@@ -2259,7 +2268,7 @@ static bool eliminateDeadStores(Function &F, AliasAnalysis &AA, MemorySSA &MSSA,
               Shortend = true;
               // Remove killing store and remove any outstanding overlap
               // intervals for the updated store.
-              State.deleteDeadInstruction(KillingSI);
+              State.deleteDeadInstruction(KillingSI, &Deleted);
               auto I = State.IOLs.find(DeadSI->getParent());
               if (I != State.IOLs.end())
                 I->second.erase(DeadSI);
@@ -2271,13 +2280,16 @@ static bool eliminateDeadStores(Function &F, AliasAnalysis &AA, MemorySSA &MSSA,
         if (OR == OW_Complete) {
           LLVM_DEBUG(dbgs() << "DSE: Remove Dead Store:\n  DEAD: " << *DeadI
                             << "\n  KILLER: " << *KillingI << '\n');
-          State.deleteDeadInstruction(DeadI);
+          State.deleteDeadInstruction(DeadI, &Deleted);
           ++NumFastStores;
           MadeChange = true;
         }
       }
     }
 
+    assert(State.SkipStores.size() - OrigNumSkipStores == Deleted.size() &&
+           "SkipStores and Deleted out of sync?");
+
     // Check if the store is a no-op.
     if (!Shortend && State.storeIsNoop(KillingDef, KillingUndObj)) {
       LLVM_DEBUG(dbgs() << "DSE: Remove No-Op Store:\n  DEAD: " << *KillingI

From f172bfcdc229ed6f3e8a0298dab3713fb763ff28 Mon Sep 17 00:00:00 2001
From: Volodymyr Sapsai <vsapsai@apple.com>
Date: Fri, 12 Apr 2024 18:27:40 -0700
Subject: [PATCH 315/886] [unused-includes][ADT] Don't include DJB.h as it's
 not used. NFC. (#88211)

StringMap uses xxHash instead of DJB.
---
 llvm/include/llvm/ADT/StringMap.h | 1 -
 1 file changed, 1 deletion(-)

diff --git a/llvm/include/llvm/ADT/StringMap.h b/llvm/include/llvm/ADT/StringMap.h
index 32e611467ac58..453d91349e358 100644
--- a/llvm/include/llvm/ADT/StringMap.h
+++ b/llvm/include/llvm/ADT/StringMap.h
@@ -17,7 +17,6 @@
 #include "llvm/ADT/StringMapEntry.h"
 #include "llvm/ADT/iterator.h"
 #include "llvm/Support/AllocatorBase.h"
-#include "llvm/Support/DJB.h"
 #include "llvm/Support/PointerLikeTypeTraits.h"
 #include <initializer_list>
 #include <iterator>

From f22c30063ddcb7f8e8c4fb0496f515be7f9408b8 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Fri, 12 Apr 2024 20:30:15 -0700
Subject: [PATCH 316/886] [test] alias-unused.cpp: set triple to Linux

Otherwise the test would fail on Darwin and other platforms that use
Itanium ABI but do not support alias/ifunc.
---
 clang/test/Sema/alias-unused.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/test/Sema/alias-unused.cpp b/clang/test/Sema/alias-unused.cpp
index 509d3e26225d7..dc8e46f072d74 100644
--- a/clang/test/Sema/alias-unused.cpp
+++ b/clang/test/Sema/alias-unused.cpp
@@ -1,5 +1,5 @@
-// RUN: %clang_cc1 -triple %itanium_abi_triple -Wunused -x c -verify %s
-// RUN: %clang_cc1 -triple %itanium_abi_triple -Wunused -verify=expected,cxx %s
+// RUN: %clang_cc1 -triple x86_64-linux -Wunused -x c -verify %s
+// RUN: %clang_cc1 -triple x86_64-linux -Wunused -verify=expected,cxx %s
 
 #ifdef __cplusplus
 extern "C" {

From aef28100294d04c2253a1cb8bbf552fa296dea4e Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Fri, 12 Apr 2024 20:53:44 -0700
Subject: [PATCH 317/886] [ELF] Relocations: Avoid MIPS check for ELFCLASS64
 with if constexpr. NFC

After inlining, `scanSection` is significantly longer (more than 100+
instructions on x86-64 built with Clang) when `i` does not always
increment by one (MIPS).
---
 lld/ELF/Relocations.cpp | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/lld/ELF/Relocations.cpp b/lld/ELF/Relocations.cpp
index 55274344f8811..04db413a6609f 100644
--- a/lld/ELF/Relocations.cpp
+++ b/lld/ELF/Relocations.cpp
@@ -1419,11 +1419,16 @@ template <class ELFT, class RelTy> void RelocationScanner::scanOne(RelTy *&i) {
   uint32_t symIndex = rel.getSymbol(config->isMips64EL);
   Symbol &sym = sec->getFile<ELFT>()->getSymbol(symIndex);
   RelType type;
-  if (config->mipsN32Abi) {
-    type = getMipsN32RelType(i);
-  } else {
+  if constexpr (ELFT::Is64Bits) {
     type = rel.getType(config->isMips64EL);
     ++i;
+  } else {
+    if (config->mipsN32Abi) {
+      type = getMipsN32RelType(i);
+    } else {
+      type = rel.getType(config->isMips64EL);
+      ++i;
+    }
   }
   // Get an offset in an output section this relocation is applied to.
   uint64_t offset = getter.get(rel.r_offset);

From 94b3c192d3468752e6bc2f20957d433bccdb8ff3 Mon Sep 17 00:00:00 2001
From: Jim Lin <jim@andestech.com>
Date: Fri, 12 Apr 2024 23:39:30 -0500
Subject: [PATCH 318/886] [M68k][clang] Enable frame pointer optimization by
 default (#87264)

Enable frame pointer optimization by default to match it with gcc.

Fixes: https://github.com/llvm/llvm-project/issues/75013
---
 clang/lib/Driver/ToolChains/CommonArgs.cpp | 1 +
 clang/test/Driver/frame-pointer-elim.c     | 6 ++++++
 2 files changed, 7 insertions(+)

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.cpp b/clang/lib/Driver/ToolChains/CommonArgs.cpp
index 62a53b85ce098..f10aa4dfaa9dd 100644
--- a/clang/lib/Driver/ToolChains/CommonArgs.cpp
+++ b/clang/lib/Driver/ToolChains/CommonArgs.cpp
@@ -114,6 +114,7 @@ static bool useFramePointerForTargetByDefault(const llvm::opt::ArgList &Args,
   case llvm::Triple::csky:
   case llvm::Triple::loongarch32:
   case llvm::Triple::loongarch64:
+  case llvm::Triple::m68k:
     return !clang::driver::tools::areOptimizationsEnabled(Args);
   default:
     break;
diff --git a/clang/test/Driver/frame-pointer-elim.c b/clang/test/Driver/frame-pointer-elim.c
index 847c5b7b0b01f..e1b0a468ab825 100644
--- a/clang/test/Driver/frame-pointer-elim.c
+++ b/clang/test/Driver/frame-pointer-elim.c
@@ -125,6 +125,12 @@
 // RUN: %clang -### -target sparc64 -S -O1 %s 2>&1 | \
 // RUN:   FileCheck --check-prefix=KEEP-NONE %s
 
+// M68k targets omit the frame pointer when optimizations are enabled.
+// RUN: %clang -### -target m68k -S %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=KEEP-ALL %s
+// RUN: %clang -### -target m68k -S -O1 %s 2>&1 | \
+// RUN:   FileCheck --check-prefix=KEEP-NONE %s
+
 // For AAarch32 (A32, T32) linux targets, default omit frame pointer when
 // optimizations are enabled.
 // RUN: %clang -### -target arm-linux-gnueabihf- -marm -S %s 2>&1 | \

From 0a6f6df5b0c3d0f2a42f013bf5cafb9b5020dcac Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 08:54:25 +0400
Subject: [PATCH 319/886] [clang] Introduce `SemaCUDA` (#88559)

This patch moves CUDA-related `Sema` function into new `SemaCUDA` class,
following the recent example of SYCL, OpenACC, and HLSL. This is a part
of the effort to split Sema. Additional context can be found in
https://github.com/llvm/llvm-project/pull/82217,
https://github.com/llvm/llvm-project/pull/84184,
https://github.com/llvm/llvm-project/pull/87634.
---
 clang/include/clang/Basic/Cuda.h              |   8 +
 clang/include/clang/Sema/Sema.h               | 304 +-----------------
 clang/include/clang/Sema/SemaBase.h           |   2 +-
 clang/include/clang/Sema/SemaCUDA.h           | 304 ++++++++++++++++++
 clang/include/clang/Serialization/ASTReader.h |   2 +-
 clang/lib/Parse/ParseDecl.cpp                 |   4 +-
 clang/lib/Parse/ParseExpr.cpp                 |   7 +-
 clang/lib/Parse/ParsePragma.cpp               |   5 +-
 clang/lib/Sema/Sema.cpp                       |  18 +-
 clang/lib/Sema/SemaBase.cpp                   |   5 +-
 clang/lib/Sema/SemaCUDA.cpp                   | 277 ++++++++--------
 clang/lib/Sema/SemaDecl.cpp                   |  19 +-
 clang/lib/Sema/SemaDeclAttr.cpp               |  12 +-
 clang/lib/Sema/SemaDeclCXX.cpp                |  21 +-
 clang/lib/Sema/SemaExpr.cpp                   |  17 +-
 clang/lib/Sema/SemaExprCXX.cpp                |  27 +-
 clang/lib/Sema/SemaLambda.cpp                 |   9 +-
 clang/lib/Sema/SemaOverload.cpp               |  60 ++--
 clang/lib/Sema/SemaStmt.cpp                   |   5 +-
 clang/lib/Sema/SemaTemplate.cpp               |  15 +-
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |   3 +-
 clang/lib/Sema/SemaType.cpp                   |   9 +-
 clang/lib/Serialization/ASTReader.cpp         |   5 +-
 clang/lib/Serialization/ASTWriter.cpp         |   5 +-
 24 files changed, 604 insertions(+), 539 deletions(-)
 create mode 100644 clang/include/clang/Sema/SemaCUDA.h

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index 3e77a74c7c009..acc6bb6581d85 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -126,6 +126,14 @@ enum class CudaArch {
   HIPDefault = CudaArch::GFX906,
 };
 
+enum class CUDAFunctionTarget {
+  Device,
+  Global,
+  Host,
+  HostDevice,
+  InvalidTarget
+};
+
 static inline bool IsNVIDIAGpuArch(CudaArch A) {
   return A >= CudaArch::SM_20 && A < CudaArch::GFX600;
 }
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 00888b7f7a738..6b9789334811e 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -38,6 +38,7 @@
 #include "clang/AST/TypeOrdering.h"
 #include "clang/Basic/BitmaskEnum.h"
 #include "clang/Basic/Builtins.h"
+#include "clang/Basic/Cuda.h"
 #include "clang/Basic/DarwinSDKInfo.h"
 #include "clang/Basic/ExpressionTraits.h"
 #include "clang/Basic/Module.h"
@@ -183,6 +184,7 @@ class Preprocessor;
 class PseudoDestructorTypeStorage;
 class PseudoObjectExpr;
 class QualType;
+class SemaCUDA;
 class SemaHLSL;
 class SemaOpenACC;
 class SemaSYCL;
@@ -435,14 +437,6 @@ enum class CXXSpecialMemberKind {
   Invalid
 };
 
-enum class CUDAFunctionTarget {
-  Device,
-  Global,
-  Host,
-  HostDevice,
-  InvalidTarget
-};
-
 /// Sema - This implements semantic analysis and AST building for C.
 /// \nosubgrouping
 class Sema final : public SemaBase {
@@ -486,8 +480,7 @@ class Sema final : public SemaBase {
   // 35. Code Completion (SemaCodeComplete.cpp)
   // 36. FixIt Helpers (SemaFixItUtils.cpp)
   // 37. Name Lookup for RISC-V Vector Intrinsic (SemaRISCVVectorLookup.cpp)
-  // 38. CUDA (SemaCUDA.cpp)
-  // 39. OpenMP Directives and Clauses (SemaOpenMP.cpp)
+  // 38. OpenMP Directives and Clauses (SemaOpenMP.cpp)
 
   /// \name Semantic Analysis
   /// Implementations are in Sema.cpp
@@ -981,9 +974,19 @@ class Sema final : public SemaBase {
     return DelayedDiagnostics.push(pool);
   }
 
+  /// Diagnostics that are emitted only if we discover that the given function
+  /// must be codegen'ed.  Because handling these correctly adds overhead to
+  /// compilation, this is currently only enabled for CUDA compilations.
+  SemaDiagnosticBuilder::DeferredDiagnosticsType DeviceDeferredDiags;
+
   /// CurContext - This is the current declaration context of parsing.
   DeclContext *CurContext;
 
+  SemaCUDA &CUDA() {
+    assert(CUDAPtr);
+    return *CUDAPtr;
+  }
+
   SemaHLSL &HLSL() {
     assert(HLSLPtr);
     return *HLSLPtr;
@@ -1029,6 +1032,7 @@ class Sema final : public SemaBase {
 
   mutable IdentifierInfo *Ident_super;
 
+  std::unique_ptr<SemaCUDA> CUDAPtr;
   std::unique_ptr<SemaHLSL> HLSLPtr;
   std::unique_ptr<SemaOpenACC> OpenACCPtr;
   std::unique_ptr<SemaSYCL> SYCLPtr;
@@ -12908,258 +12912,6 @@ class Sema final : public SemaBase {
   //
   //
 
-  /// \name CUDA
-  /// Implementations are in SemaCUDA.cpp
-  ///@{
-
-public:
-  /// Increments our count of the number of times we've seen a pragma forcing
-  /// functions to be __host__ __device__.  So long as this count is greater
-  /// than zero, all functions encountered will be __host__ __device__.
-  void PushForceCUDAHostDevice();
-
-  /// Decrements our count of the number of times we've seen a pragma forcing
-  /// functions to be __host__ __device__.  Returns false if the count is 0
-  /// before incrementing, so you can emit an error.
-  bool PopForceCUDAHostDevice();
-
-  ExprResult ActOnCUDAExecConfigExpr(Scope *S, SourceLocation LLLLoc,
-                                     MultiExprArg ExecConfig,
-                                     SourceLocation GGGLoc);
-
-  /// Diagnostics that are emitted only if we discover that the given function
-  /// must be codegen'ed.  Because handling these correctly adds overhead to
-  /// compilation, this is currently only enabled for CUDA compilations.
-  SemaDiagnosticBuilder::DeferredDiagnosticsType DeviceDeferredDiags;
-
-  /// A pair of a canonical FunctionDecl and a SourceLocation.  When used as the
-  /// key in a hashtable, both the FD and location are hashed.
-  struct FunctionDeclAndLoc {
-    CanonicalDeclPtr<const FunctionDecl> FD;
-    SourceLocation Loc;
-  };
-
-  /// FunctionDecls and SourceLocations for which CheckCUDACall has emitted a
-  /// (maybe deferred) "bad call" diagnostic.  We use this to avoid emitting the
-  /// same deferred diag twice.
-  llvm::DenseSet<FunctionDeclAndLoc> LocsWithCUDACallDiags;
-
-  /// An inverse call graph, mapping known-emitted functions to one of their
-  /// known-emitted callers (plus the location of the call).
-  ///
-  /// Functions that we can tell a priori must be emitted aren't added to this
-  /// map.
-  llvm::DenseMap</* Callee = */ CanonicalDeclPtr<const FunctionDecl>,
-                 /* Caller = */ FunctionDeclAndLoc>
-      DeviceKnownEmittedFns;
-
-  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
-  /// context is "used as device code".
-  ///
-  /// - If CurContext is a __host__ function, does not emit any diagnostics
-  ///   unless \p EmitOnBothSides is true.
-  /// - If CurContext is a __device__ or __global__ function, emits the
-  ///   diagnostics immediately.
-  /// - If CurContext is a __host__ __device__ function and we are compiling for
-  ///   the device, creates a diagnostic which is emitted if and when we realize
-  ///   that the function will be codegen'ed.
-  ///
-  /// Example usage:
-  ///
-  ///  // Variable-length arrays are not allowed in CUDA device code.
-  ///  if (CUDADiagIfDeviceCode(Loc, diag::err_cuda_vla)
-  ///     << llvm::to_underlying(CurrentCUDATarget()))
-  ///    return ExprError();
-  ///  // Otherwise, continue parsing as normal.
-  SemaDiagnosticBuilder CUDADiagIfDeviceCode(SourceLocation Loc,
-                                             unsigned DiagID);
-
-  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
-  /// context is "used as host code".
-  ///
-  /// Same as CUDADiagIfDeviceCode, with "host" and "device" switched.
-  SemaDiagnosticBuilder CUDADiagIfHostCode(SourceLocation Loc, unsigned DiagID);
-
-  /// Determines whether the given function is a CUDA device/host/kernel/etc.
-  /// function.
-  ///
-  /// Use this rather than examining the function's attributes yourself -- you
-  /// will get it wrong.  Returns CUDAFunctionTarget::Host if D is null.
-  CUDAFunctionTarget IdentifyCUDATarget(const FunctionDecl *D,
-                                        bool IgnoreImplicitHDAttr = false);
-  CUDAFunctionTarget IdentifyCUDATarget(const ParsedAttributesView &Attrs);
-
-  enum CUDAVariableTarget {
-    CVT_Device,  /// Emitted on device side with a shadow variable on host side
-    CVT_Host,    /// Emitted on host side only
-    CVT_Both,    /// Emitted on both sides with different addresses
-    CVT_Unified, /// Emitted as a unified address, e.g. managed variables
-  };
-  /// Determines whether the given variable is emitted on host or device side.
-  CUDAVariableTarget IdentifyCUDATarget(const VarDecl *D);
-
-  /// Defines kinds of CUDA global host/device context where a function may be
-  /// called.
-  enum CUDATargetContextKind {
-    CTCK_Unknown,       /// Unknown context
-    CTCK_InitGlobalVar, /// Function called during global variable
-                        /// initialization
-  };
-
-  /// Define the current global CUDA host/device context where a function may be
-  /// called. Only used when a function is called outside of any functions.
-  struct CUDATargetContext {
-    CUDAFunctionTarget Target = CUDAFunctionTarget::HostDevice;
-    CUDATargetContextKind Kind = CTCK_Unknown;
-    Decl *D = nullptr;
-  } CurCUDATargetCtx;
-
-  struct CUDATargetContextRAII {
-    Sema &S;
-    CUDATargetContext SavedCtx;
-    CUDATargetContextRAII(Sema &S_, CUDATargetContextKind K, Decl *D);
-    ~CUDATargetContextRAII() { S.CurCUDATargetCtx = SavedCtx; }
-  };
-
-  /// Gets the CUDA target for the current context.
-  CUDAFunctionTarget CurrentCUDATarget() {
-    return IdentifyCUDATarget(dyn_cast<FunctionDecl>(CurContext));
-  }
-
-  static bool isCUDAImplicitHostDeviceFunction(const FunctionDecl *D);
-
-  // CUDA function call preference. Must be ordered numerically from
-  // worst to best.
-  enum CUDAFunctionPreference {
-    CFP_Never,      // Invalid caller/callee combination.
-    CFP_WrongSide,  // Calls from host-device to host or device
-                    // function that do not match current compilation
-                    // mode.
-    CFP_HostDevice, // Any calls to host/device functions.
-    CFP_SameSide,   // Calls from host-device to host or device
-                    // function matching current compilation mode.
-    CFP_Native,     // host-to-host or device-to-device calls.
-  };
-
-  /// Identifies relative preference of a given Caller/Callee
-  /// combination, based on their host/device attributes.
-  /// \param Caller function which needs address of \p Callee.
-  ///               nullptr in case of global context.
-  /// \param Callee target function
-  ///
-  /// \returns preference value for particular Caller/Callee combination.
-  CUDAFunctionPreference IdentifyCUDAPreference(const FunctionDecl *Caller,
-                                                const FunctionDecl *Callee);
-
-  /// Determines whether Caller may invoke Callee, based on their CUDA
-  /// host/device attributes.  Returns false if the call is not allowed.
-  ///
-  /// Note: Will return true for CFP_WrongSide calls.  These may appear in
-  /// semantically correct CUDA programs, but only if they're never codegen'ed.
-  bool IsAllowedCUDACall(const FunctionDecl *Caller,
-                         const FunctionDecl *Callee) {
-    return IdentifyCUDAPreference(Caller, Callee) != CFP_Never;
-  }
-
-  /// May add implicit CUDAHostAttr and CUDADeviceAttr attributes to FD,
-  /// depending on FD and the current compilation settings.
-  void maybeAddCUDAHostDeviceAttrs(FunctionDecl *FD,
-                                   const LookupResult &Previous);
-
-  /// May add implicit CUDAConstantAttr attribute to VD, depending on VD
-  /// and current compilation settings.
-  void MaybeAddCUDAConstantAttr(VarDecl *VD);
-
-  /// Check whether we're allowed to call Callee from the current context.
-  ///
-  /// - If the call is never allowed in a semantically-correct program
-  ///   (CFP_Never), emits an error and returns false.
-  ///
-  /// - If the call is allowed in semantically-correct programs, but only if
-  ///   it's never codegen'ed (CFP_WrongSide), creates a deferred diagnostic to
-  ///   be emitted if and when the caller is codegen'ed, and returns true.
-  ///
-  ///   Will only create deferred diagnostics for a given SourceLocation once,
-  ///   so you can safely call this multiple times without generating duplicate
-  ///   deferred errors.
-  ///
-  /// - Otherwise, returns true without emitting any diagnostics.
-  bool CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee);
-
-  void CUDACheckLambdaCapture(CXXMethodDecl *D, const sema::Capture &Capture);
-
-  /// Set __device__ or __host__ __device__ attributes on the given lambda
-  /// operator() method.
-  ///
-  /// CUDA lambdas by default is host device function unless it has explicit
-  /// host or device attribute.
-  void CUDASetLambdaAttrs(CXXMethodDecl *Method);
-
-  /// Record \p FD if it is a CUDA/HIP implicit host device function used on
-  /// device side in device compilation.
-  void CUDARecordImplicitHostDeviceFuncUsedByDevice(const FunctionDecl *FD);
-
-  /// Finds a function in \p Matches with highest calling priority
-  /// from \p Caller context and erases all functions with lower
-  /// calling priority.
-  void EraseUnwantedCUDAMatches(
-      const FunctionDecl *Caller,
-      SmallVectorImpl<std::pair<DeclAccessPair, FunctionDecl *>> &Matches);
-
-  /// Given a implicit special member, infer its CUDA target from the
-  /// calls it needs to make to underlying base/field special members.
-  /// \param ClassDecl the class for which the member is being created.
-  /// \param CSM the kind of special member.
-  /// \param MemberDecl the special member itself.
-  /// \param ConstRHS true if this is a copy operation with a const object on
-  ///        its RHS.
-  /// \param Diagnose true if this call should emit diagnostics.
-  /// \return true if there was an error inferring.
-  /// The result of this call is implicit CUDA target attribute(s) attached to
-  /// the member declaration.
-  bool inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
-                                               CXXSpecialMemberKind CSM,
-                                               CXXMethodDecl *MemberDecl,
-                                               bool ConstRHS, bool Diagnose);
-
-  /// \return true if \p CD can be considered empty according to CUDA
-  /// (E.2.3.1 in CUDA 7.5 Programming guide).
-  bool isEmptyCudaConstructor(SourceLocation Loc, CXXConstructorDecl *CD);
-  bool isEmptyCudaDestructor(SourceLocation Loc, CXXDestructorDecl *CD);
-
-  // \brief Checks that initializers of \p Var satisfy CUDA restrictions. In
-  // case of error emits appropriate diagnostic and invalidates \p Var.
-  //
-  // \details CUDA allows only empty constructors as initializers for global
-  // variables (see E.2.3.1, CUDA 7.5). The same restriction also applies to all
-  // __shared__ variables whether they are local or not (they all are implicitly
-  // static in CUDA). One exception is that CUDA allows constant initializers
-  // for __constant__ and __device__ variables.
-  void checkAllowedCUDAInitializer(VarDecl *VD);
-
-  /// Check whether NewFD is a valid overload for CUDA. Emits
-  /// diagnostics and invalidates NewFD if not.
-  void checkCUDATargetOverload(FunctionDecl *NewFD,
-                               const LookupResult &Previous);
-  /// Copies target attributes from the template TD to the function FD.
-  void inheritCUDATargetAttrs(FunctionDecl *FD, const FunctionTemplateDecl &TD);
-
-  /// Returns the name of the launch configuration function.  This is the name
-  /// of the function that will be called to configure kernel call, with the
-  /// parameters specified via <<<>>>.
-  std::string getCudaConfigureFuncName() const;
-
-private:
-  unsigned ForceCUDAHostDeviceDepth = 0;
-
-  ///@}
-
-  //
-  //
-  // -------------------------------------------------------------------------
-  //
-  //
-
   /// \name OpenMP Directives and Clauses
   /// Implementations are in SemaOpenMP.cpp
   ///@{
@@ -14546,32 +14298,4 @@ std::unique_ptr<sema::RISCVIntrinsicManager>
 CreateRISCVIntrinsicManager(Sema &S);
 } // end namespace clang
 
-namespace llvm {
-// Hash a FunctionDeclAndLoc by looking at both its FunctionDecl and its
-// SourceLocation.
-template <> struct DenseMapInfo<clang::Sema::FunctionDeclAndLoc> {
-  using FunctionDeclAndLoc = clang::Sema::FunctionDeclAndLoc;
-  using FDBaseInfo =
-      DenseMapInfo<clang::CanonicalDeclPtr<const clang::FunctionDecl>>;
-
-  static FunctionDeclAndLoc getEmptyKey() {
-    return {FDBaseInfo::getEmptyKey(), clang::SourceLocation()};
-  }
-
-  static FunctionDeclAndLoc getTombstoneKey() {
-    return {FDBaseInfo::getTombstoneKey(), clang::SourceLocation()};
-  }
-
-  static unsigned getHashValue(const FunctionDeclAndLoc &FDL) {
-    return hash_combine(FDBaseInfo::getHashValue(FDL.FD),
-                        FDL.Loc.getHashValue());
-  }
-
-  static bool isEqual(const FunctionDeclAndLoc &LHS,
-                      const FunctionDeclAndLoc &RHS) {
-    return LHS.FD == RHS.FD && LHS.Loc == RHS.Loc;
-  }
-};
-} // namespace llvm
-
 #endif
diff --git a/clang/include/clang/Sema/SemaBase.h b/clang/include/clang/Sema/SemaBase.h
index ff718022fca03..3220f71dd797e 100644
--- a/clang/include/clang/Sema/SemaBase.h
+++ b/clang/include/clang/Sema/SemaBase.h
@@ -146,7 +146,7 @@ class SemaBase {
     ///   if (SemaDiagnosticBuilder(...) << foo << bar)
     ///     return ExprError();
     ///
-    /// But see CUDADiagIfDeviceCode() and CUDADiagIfHostCode() -- you probably
+    /// But see DiagIfDeviceCode() and DiagIfHostCode() -- you probably
     /// want to use these instead of creating a SemaDiagnosticBuilder yourself.
     operator bool() const { return isImmediate(); }
 
diff --git a/clang/include/clang/Sema/SemaCUDA.h b/clang/include/clang/Sema/SemaCUDA.h
new file mode 100644
index 0000000000000..63dc3f4da240b
--- /dev/null
+++ b/clang/include/clang/Sema/SemaCUDA.h
@@ -0,0 +1,304 @@
+//===----- SemaCUDA.h ----- Semantic Analysis for CUDA constructs ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file declares semantic analysis for CUDA constructs.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_SEMACUDA_H
+#define LLVM_CLANG_SEMA_SEMACUDA_H
+
+#include "clang/AST/Decl.h"
+#include "clang/AST/DeclCXX.h"
+#include "clang/AST/Redeclarable.h"
+#include "clang/Basic/Cuda.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Sema/Lookup.h"
+#include "clang/Sema/Ownership.h"
+#include "clang/Sema/ParsedAttr.h"
+#include "clang/Sema/Scope.h"
+#include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaBase.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallVector.h"
+#include <string>
+
+namespace clang {
+
+enum class CUDAFunctionTarget;
+
+class SemaCUDA : public SemaBase {
+public:
+  SemaCUDA(Sema &S);
+
+  /// Increments our count of the number of times we've seen a pragma forcing
+  /// functions to be __host__ __device__.  So long as this count is greater
+  /// than zero, all functions encountered will be __host__ __device__.
+  void PushForceHostDevice();
+
+  /// Decrements our count of the number of times we've seen a pragma forcing
+  /// functions to be __host__ __device__.  Returns false if the count is 0
+  /// before incrementing, so you can emit an error.
+  bool PopForceHostDevice();
+
+  ExprResult ActOnExecConfigExpr(Scope *S, SourceLocation LLLLoc,
+                                 MultiExprArg ExecConfig,
+                                 SourceLocation GGGLoc);
+
+  /// A pair of a canonical FunctionDecl and a SourceLocation.  When used as the
+  /// key in a hashtable, both the FD and location are hashed.
+  struct FunctionDeclAndLoc {
+    CanonicalDeclPtr<const FunctionDecl> FD;
+    SourceLocation Loc;
+  };
+
+  /// FunctionDecls and SourceLocations for which CheckCall has emitted a
+  /// (maybe deferred) "bad call" diagnostic.  We use this to avoid emitting the
+  /// same deferred diag twice.
+  llvm::DenseSet<FunctionDeclAndLoc> LocsWithCUDACallDiags;
+
+  /// An inverse call graph, mapping known-emitted functions to one of their
+  /// known-emitted callers (plus the location of the call).
+  ///
+  /// Functions that we can tell a priori must be emitted aren't added to this
+  /// map.
+  llvm::DenseMap</* Callee = */ CanonicalDeclPtr<const FunctionDecl>,
+                 /* Caller = */ FunctionDeclAndLoc>
+      DeviceKnownEmittedFns;
+
+  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
+  /// context is "used as device code".
+  ///
+  /// - If CurContext is a __host__ function, does not emit any diagnostics
+  ///   unless \p EmitOnBothSides is true.
+  /// - If CurContext is a __device__ or __global__ function, emits the
+  ///   diagnostics immediately.
+  /// - If CurContext is a __host__ __device__ function and we are compiling for
+  ///   the device, creates a diagnostic which is emitted if and when we realize
+  ///   that the function will be codegen'ed.
+  ///
+  /// Example usage:
+  ///
+  ///  // Variable-length arrays are not allowed in CUDA device code.
+  ///  if (DiagIfDeviceCode(Loc, diag::err_cuda_vla) << CurrentTarget())
+  ///    return ExprError();
+  ///  // Otherwise, continue parsing as normal.
+  SemaDiagnosticBuilder DiagIfDeviceCode(SourceLocation Loc, unsigned DiagID);
+
+  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
+  /// context is "used as host code".
+  ///
+  /// Same as DiagIfDeviceCode, with "host" and "device" switched.
+  SemaDiagnosticBuilder DiagIfHostCode(SourceLocation Loc, unsigned DiagID);
+
+  /// Determines whether the given function is a CUDA device/host/kernel/etc.
+  /// function.
+  ///
+  /// Use this rather than examining the function's attributes yourself -- you
+  /// will get it wrong.  Returns CUDAFunctionTarget::Host if D is null.
+  CUDAFunctionTarget IdentifyTarget(const FunctionDecl *D,
+                                    bool IgnoreImplicitHDAttr = false);
+  CUDAFunctionTarget IdentifyTarget(const ParsedAttributesView &Attrs);
+
+  enum CUDAVariableTarget {
+    CVT_Device,  /// Emitted on device side with a shadow variable on host side
+    CVT_Host,    /// Emitted on host side only
+    CVT_Both,    /// Emitted on both sides with different addresses
+    CVT_Unified, /// Emitted as a unified address, e.g. managed variables
+  };
+  /// Determines whether the given variable is emitted on host or device side.
+  CUDAVariableTarget IdentifyTarget(const VarDecl *D);
+
+  /// Defines kinds of CUDA global host/device context where a function may be
+  /// called.
+  enum CUDATargetContextKind {
+    CTCK_Unknown,       /// Unknown context
+    CTCK_InitGlobalVar, /// Function called during global variable
+                        /// initialization
+  };
+
+  /// Define the current global CUDA host/device context where a function may be
+  /// called. Only used when a function is called outside of any functions.
+  struct CUDATargetContext {
+    CUDAFunctionTarget Target = CUDAFunctionTarget::HostDevice;
+    CUDATargetContextKind Kind = CTCK_Unknown;
+    Decl *D = nullptr;
+  } CurCUDATargetCtx;
+
+  struct CUDATargetContextRAII {
+    SemaCUDA &S;
+    SemaCUDA::CUDATargetContext SavedCtx;
+    CUDATargetContextRAII(SemaCUDA &S_, SemaCUDA::CUDATargetContextKind K,
+                          Decl *D);
+    ~CUDATargetContextRAII() { S.CurCUDATargetCtx = SavedCtx; }
+  };
+
+  /// Gets the CUDA target for the current context.
+  CUDAFunctionTarget CurrentTarget() {
+    return IdentifyTarget(dyn_cast<FunctionDecl>(SemaRef.CurContext));
+  }
+
+  static bool isImplicitHostDeviceFunction(const FunctionDecl *D);
+
+  // CUDA function call preference. Must be ordered numerically from
+  // worst to best.
+  enum CUDAFunctionPreference {
+    CFP_Never,      // Invalid caller/callee combination.
+    CFP_WrongSide,  // Calls from host-device to host or device
+                    // function that do not match current compilation
+                    // mode.
+    CFP_HostDevice, // Any calls to host/device functions.
+    CFP_SameSide,   // Calls from host-device to host or device
+                    // function matching current compilation mode.
+    CFP_Native,     // host-to-host or device-to-device calls.
+  };
+
+  /// Identifies relative preference of a given Caller/Callee
+  /// combination, based on their host/device attributes.
+  /// \param Caller function which needs address of \p Callee.
+  ///               nullptr in case of global context.
+  /// \param Callee target function
+  ///
+  /// \returns preference value for particular Caller/Callee combination.
+  CUDAFunctionPreference IdentifyPreference(const FunctionDecl *Caller,
+                                            const FunctionDecl *Callee);
+
+  /// Determines whether Caller may invoke Callee, based on their CUDA
+  /// host/device attributes.  Returns false if the call is not allowed.
+  ///
+  /// Note: Will return true for CFP_WrongSide calls.  These may appear in
+  /// semantically correct CUDA programs, but only if they're never codegen'ed.
+  bool IsAllowedCall(const FunctionDecl *Caller, const FunctionDecl *Callee) {
+    return IdentifyPreference(Caller, Callee) != CFP_Never;
+  }
+
+  /// May add implicit CUDAHostAttr and CUDADeviceAttr attributes to FD,
+  /// depending on FD and the current compilation settings.
+  void maybeAddHostDeviceAttrs(FunctionDecl *FD, const LookupResult &Previous);
+
+  /// May add implicit CUDAConstantAttr attribute to VD, depending on VD
+  /// and current compilation settings.
+  void MaybeAddConstantAttr(VarDecl *VD);
+
+  /// Check whether we're allowed to call Callee from the current context.
+  ///
+  /// - If the call is never allowed in a semantically-correct program
+  ///   (CFP_Never), emits an error and returns false.
+  ///
+  /// - If the call is allowed in semantically-correct programs, but only if
+  ///   it's never codegen'ed (CFP_WrongSide), creates a deferred diagnostic to
+  ///   be emitted if and when the caller is codegen'ed, and returns true.
+  ///
+  ///   Will only create deferred diagnostics for a given SourceLocation once,
+  ///   so you can safely call this multiple times without generating duplicate
+  ///   deferred errors.
+  ///
+  /// - Otherwise, returns true without emitting any diagnostics.
+  bool CheckCall(SourceLocation Loc, FunctionDecl *Callee);
+
+  void CheckLambdaCapture(CXXMethodDecl *D, const sema::Capture &Capture);
+
+  /// Set __device__ or __host__ __device__ attributes on the given lambda
+  /// operator() method.
+  ///
+  /// CUDA lambdas by default is host device function unless it has explicit
+  /// host or device attribute.
+  void SetLambdaAttrs(CXXMethodDecl *Method);
+
+  /// Record \p FD if it is a CUDA/HIP implicit host device function used on
+  /// device side in device compilation.
+  void RecordImplicitHostDeviceFuncUsedByDevice(const FunctionDecl *FD);
+
+  /// Finds a function in \p Matches with highest calling priority
+  /// from \p Caller context and erases all functions with lower
+  /// calling priority.
+  void EraseUnwantedMatches(
+      const FunctionDecl *Caller,
+      llvm::SmallVectorImpl<std::pair<DeclAccessPair, FunctionDecl *>>
+          &Matches);
+
+  /// Given a implicit special member, infer its CUDA target from the
+  /// calls it needs to make to underlying base/field special members.
+  /// \param ClassDecl the class for which the member is being created.
+  /// \param CSM the kind of special member.
+  /// \param MemberDecl the special member itself.
+  /// \param ConstRHS true if this is a copy operation with a const object on
+  ///        its RHS.
+  /// \param Diagnose true if this call should emit diagnostics.
+  /// \return true if there was an error inferring.
+  /// The result of this call is implicit CUDA target attribute(s) attached to
+  /// the member declaration.
+  bool inferTargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
+                                           CXXSpecialMemberKind CSM,
+                                           CXXMethodDecl *MemberDecl,
+                                           bool ConstRHS, bool Diagnose);
+
+  /// \return true if \p CD can be considered empty according to CUDA
+  /// (E.2.3.1 in CUDA 7.5 Programming guide).
+  bool isEmptyConstructor(SourceLocation Loc, CXXConstructorDecl *CD);
+  bool isEmptyDestructor(SourceLocation Loc, CXXDestructorDecl *CD);
+
+  // \brief Checks that initializers of \p Var satisfy CUDA restrictions. In
+  // case of error emits appropriate diagnostic and invalidates \p Var.
+  //
+  // \details CUDA allows only empty constructors as initializers for global
+  // variables (see E.2.3.1, CUDA 7.5). The same restriction also applies to all
+  // __shared__ variables whether they are local or not (they all are implicitly
+  // static in CUDA). One exception is that CUDA allows constant initializers
+  // for __constant__ and __device__ variables.
+  void checkAllowedInitializer(VarDecl *VD);
+
+  /// Check whether NewFD is a valid overload for CUDA. Emits
+  /// diagnostics and invalidates NewFD if not.
+  void checkTargetOverload(FunctionDecl *NewFD, const LookupResult &Previous);
+  /// Copies target attributes from the template TD to the function FD.
+  void inheritTargetAttrs(FunctionDecl *FD, const FunctionTemplateDecl &TD);
+
+  /// Returns the name of the launch configuration function.  This is the name
+  /// of the function that will be called to configure kernel call, with the
+  /// parameters specified via <<<>>>.
+  std::string getConfigureFuncName() const;
+
+private:
+  unsigned ForceHostDeviceDepth = 0;
+
+  friend class ASTReader;
+  friend class ASTWriter;
+};
+
+} // namespace clang
+
+namespace llvm {
+// Hash a FunctionDeclAndLoc by looking at both its FunctionDecl and its
+// SourceLocation.
+template <> struct DenseMapInfo<clang::SemaCUDA::FunctionDeclAndLoc> {
+  using FunctionDeclAndLoc = clang::SemaCUDA::FunctionDeclAndLoc;
+  using FDBaseInfo =
+      DenseMapInfo<clang::CanonicalDeclPtr<const clang::FunctionDecl>>;
+
+  static FunctionDeclAndLoc getEmptyKey() {
+    return {FDBaseInfo::getEmptyKey(), clang::SourceLocation()};
+  }
+
+  static FunctionDeclAndLoc getTombstoneKey() {
+    return {FDBaseInfo::getTombstoneKey(), clang::SourceLocation()};
+  }
+
+  static unsigned getHashValue(const FunctionDeclAndLoc &FDL) {
+    return hash_combine(FDBaseInfo::getHashValue(FDL.FD),
+                        FDL.Loc.getHashValue());
+  }
+
+  static bool isEqual(const FunctionDeclAndLoc &LHS,
+                      const FunctionDeclAndLoc &RHS) {
+    return LHS.FD == RHS.FD && LHS.Loc == RHS.Loc;
+  }
+};
+} // namespace llvm
+
+#endif // LLVM_CLANG_SEMA_SEMACUDA_H
diff --git a/clang/include/clang/Serialization/ASTReader.h b/clang/include/clang/Serialization/ASTReader.h
index 6656c1c58dec9..e3fde887f99cb 100644
--- a/clang/include/clang/Serialization/ASTReader.h
+++ b/clang/include/clang/Serialization/ASTReader.h
@@ -873,7 +873,7 @@ class ASTReader
 
   /// Our current depth in #pragma cuda force_host_device begin/end
   /// macros.
-  unsigned ForceCUDAHostDeviceDepth = 0;
+  unsigned ForceHostDeviceDepth = 0;
 
   /// The IDs of the declarations Sema stores directly.
   ///
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index 583232f2d610d..951e2210031a1 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -26,6 +26,7 @@
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaDiagnostic.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallString.h"
@@ -2664,7 +2665,8 @@ Decl *Parser::ParseDeclarationAfterDeclaratorAndAttributes(
     }
   }
 
-  Sema::CUDATargetContextRAII X(Actions, Sema::CTCK_InitGlobalVar, ThisDecl);
+  SemaCUDA::CUDATargetContextRAII X(Actions.CUDA(),
+                                    SemaCUDA::CTCK_InitGlobalVar, ThisDecl);
   switch (TheInitKind) {
   // Parse declarator '=' initializer.
   case InitKind::Equal: {
diff --git a/clang/lib/Parse/ParseExpr.cpp b/clang/lib/Parse/ParseExpr.cpp
index d08e675604d19..473ec9afd6018 100644
--- a/clang/lib/Parse/ParseExpr.cpp
+++ b/clang/lib/Parse/ParseExpr.cpp
@@ -30,6 +30,7 @@
 #include "clang/Sema/EnterExpressionEvaluationContext.h"
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaSYCL.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "llvm/ADT/SmallVector.h"
@@ -2129,10 +2130,8 @@ Parser::ParsePostfixExpressionSuffix(ExprResult LHS) {
         }
 
         if (!LHS.isInvalid()) {
-          ExprResult ECResult = Actions.ActOnCUDAExecConfigExpr(getCurScope(),
-                                    OpenLoc,
-                                    ExecConfigExprs,
-                                    CloseLoc);
+          ExprResult ECResult = Actions.CUDA().ActOnExecConfigExpr(
+              getCurScope(), OpenLoc, ExecConfigExprs, CloseLoc);
           if (ECResult.isInvalid())
             LHS = ExprError();
           else
diff --git a/clang/lib/Parse/ParsePragma.cpp b/clang/lib/Parse/ParsePragma.cpp
index 0f692e2146a49..3979f75b6020d 100644
--- a/clang/lib/Parse/ParsePragma.cpp
+++ b/clang/lib/Parse/ParsePragma.cpp
@@ -21,6 +21,7 @@
 #include "clang/Parse/RAIIObjectsForParser.h"
 #include "clang/Sema/EnterExpressionEvaluationContext.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/StringSwitch.h"
 #include <optional>
@@ -3900,8 +3901,8 @@ void PragmaForceCUDAHostDeviceHandler::HandlePragma(
   }
 
   if (Info->isStr("begin"))
-    Actions.PushForceCUDAHostDevice();
-  else if (!Actions.PopForceCUDAHostDevice())
+    Actions.CUDA().PushForceHostDevice();
+  else if (!Actions.CUDA().PopForceHostDevice())
     PP.Diag(FirstTok.getLocation(),
             diag::err_pragma_cannot_end_force_cuda_host_device);
 
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index a2ea66f339c8e..8de202f4f7a0c 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -41,6 +41,7 @@
 #include "clang/Sema/RISCVIntrinsicManager.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaConsumer.h"
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
@@ -199,6 +200,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       LateTemplateParser(nullptr), LateTemplateParserCleanup(nullptr),
       OpaqueParser(nullptr), CurContext(nullptr), ExternalSource(nullptr),
       CurScope(nullptr), Ident_super(nullptr),
+      CUDAPtr(std::make_unique<SemaCUDA>(*this)),
       HLSLPtr(std::make_unique<SemaHLSL>(*this)),
       OpenACCPtr(std::make_unique<SemaOpenACC>(*this)),
       SYCLPtr(std::make_unique<SemaSYCL>(*this)),
@@ -1635,15 +1637,15 @@ bool Sema::hasUncompilableErrorOccurred() const {
 // Print notes showing how we can reach FD starting from an a priori
 // known-callable function.
 static void emitCallStackNotes(Sema &S, const FunctionDecl *FD) {
-  auto FnIt = S.DeviceKnownEmittedFns.find(FD);
-  while (FnIt != S.DeviceKnownEmittedFns.end()) {
+  auto FnIt = S.CUDA().DeviceKnownEmittedFns.find(FD);
+  while (FnIt != S.CUDA().DeviceKnownEmittedFns.end()) {
     // Respect error limit.
     if (S.Diags.hasFatalErrorOccurred())
       return;
     DiagnosticBuilder Builder(
         S.Diags.Report(FnIt->second.Loc, diag::note_called_by));
     Builder << FnIt->second.FD;
-    FnIt = S.DeviceKnownEmittedFns.find(FnIt->second.FD);
+    FnIt = S.CUDA().DeviceKnownEmittedFns.find(FnIt->second.FD);
   }
 }
 
@@ -1747,7 +1749,7 @@ class DeferredDiagnosticsEmitter
         (ShouldEmitRootNode || InOMPDeviceContext))
       S.finalizeOpenMPDelayedAnalysis(Caller, FD, Loc);
     if (Caller)
-      S.DeviceKnownEmittedFns[FD] = {Caller, Loc};
+      S.CUDA().DeviceKnownEmittedFns[FD] = {Caller, Loc};
     // Always emit deferred diagnostics for the direct users. This does not
     // lead to explosion of diagnostics since each user is visited at most
     // twice.
@@ -1836,8 +1838,8 @@ void Sema::emitDeferredDiags() {
 // which other not-known-emitted functions.
 //
 // When we see something which is illegal if the current function is emitted
-// (usually by way of CUDADiagIfDeviceCode, CUDADiagIfHostCode, or
-// CheckCUDACall), we first check if the current function is known-emitted.  If
+// (usually by way of DiagIfDeviceCode, DiagIfHostCode, or
+// CheckCall), we first check if the current function is known-emitted.  If
 // so, we immediately output the diagnostic.
 //
 // Otherwise, we "defer" the diagnostic.  It sits in Sema::DeviceDeferredDiags
@@ -1900,8 +1902,8 @@ Sema::targetDiag(SourceLocation Loc, unsigned DiagID, const FunctionDecl *FD) {
                ? diagIfOpenMPDeviceCode(Loc, DiagID, FD)
                : diagIfOpenMPHostCode(Loc, DiagID, FD);
   if (getLangOpts().CUDA)
-    return getLangOpts().CUDAIsDevice ? CUDADiagIfDeviceCode(Loc, DiagID)
-                                      : CUDADiagIfHostCode(Loc, DiagID);
+    return getLangOpts().CUDAIsDevice ? CUDA().DiagIfDeviceCode(Loc, DiagID)
+                                      : CUDA().DiagIfHostCode(Loc, DiagID);
 
   if (getLangOpts().SYCLIsDevice)
     return SYCL().DiagIfDeviceCode(Loc, DiagID);
diff --git a/clang/lib/Sema/SemaBase.cpp b/clang/lib/Sema/SemaBase.cpp
index 95c0cfbe283b0..0442fb2929e3c 100644
--- a/clang/lib/Sema/SemaBase.cpp
+++ b/clang/lib/Sema/SemaBase.cpp
@@ -1,5 +1,6 @@
 #include "clang/Sema/SemaBase.h"
 #include "clang/Sema/Sema.h"
+#include "clang/Sema/SemaCUDA.h"
 
 namespace clang {
 
@@ -70,8 +71,8 @@ Sema::SemaDiagnosticBuilder SemaBase::Diag(SourceLocation Loc, unsigned DiagID,
   }
 
   SemaDiagnosticBuilder DB = getLangOpts().CUDAIsDevice
-                                 ? SemaRef.CUDADiagIfDeviceCode(Loc, DiagID)
-                                 : SemaRef.CUDADiagIfHostCode(Loc, DiagID);
+                                 ? SemaRef.CUDA().DiagIfDeviceCode(Loc, DiagID)
+                                 : SemaRef.CUDA().DiagIfHostCode(Loc, DiagID);
   SetIsLastErrorImmediate(DB.isImmediate());
   return DB;
 }
diff --git a/clang/lib/Sema/SemaCUDA.cpp b/clang/lib/Sema/SemaCUDA.cpp
index 9d6d709e262ad..80ea43dc5316e 100644
--- a/clang/lib/Sema/SemaCUDA.cpp
+++ b/clang/lib/Sema/SemaCUDA.cpp
@@ -10,6 +10,7 @@
 ///
 //===----------------------------------------------------------------------===//
 
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/ExprCXX.h"
@@ -27,6 +28,8 @@
 #include <optional>
 using namespace clang;
 
+SemaCUDA::SemaCUDA(Sema &S) : SemaBase(S) {}
+
 template <typename AttrT> static bool hasExplicitAttr(const VarDecl *D) {
   if (!D)
     return false;
@@ -35,37 +38,37 @@ template <typename AttrT> static bool hasExplicitAttr(const VarDecl *D) {
   return false;
 }
 
-void Sema::PushForceCUDAHostDevice() {
+void SemaCUDA::PushForceHostDevice() {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
-  ForceCUDAHostDeviceDepth++;
+  ForceHostDeviceDepth++;
 }
 
-bool Sema::PopForceCUDAHostDevice() {
+bool SemaCUDA::PopForceHostDevice() {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
-  if (ForceCUDAHostDeviceDepth == 0)
+  if (ForceHostDeviceDepth == 0)
     return false;
-  ForceCUDAHostDeviceDepth--;
+  ForceHostDeviceDepth--;
   return true;
 }
 
-ExprResult Sema::ActOnCUDAExecConfigExpr(Scope *S, SourceLocation LLLLoc,
+ExprResult SemaCUDA::ActOnExecConfigExpr(Scope *S, SourceLocation LLLLoc,
                                          MultiExprArg ExecConfig,
                                          SourceLocation GGGLoc) {
-  FunctionDecl *ConfigDecl = Context.getcudaConfigureCallDecl();
+  FunctionDecl *ConfigDecl = getASTContext().getcudaConfigureCallDecl();
   if (!ConfigDecl)
     return ExprError(Diag(LLLLoc, diag::err_undeclared_var_use)
-                     << getCudaConfigureFuncName());
+                     << getConfigureFuncName());
   QualType ConfigQTy = ConfigDecl->getType();
 
-  DeclRefExpr *ConfigDR = new (Context)
-      DeclRefExpr(Context, ConfigDecl, false, ConfigQTy, VK_LValue, LLLLoc);
-  MarkFunctionReferenced(LLLLoc, ConfigDecl);
+  DeclRefExpr *ConfigDR = new (getASTContext()) DeclRefExpr(
+      getASTContext(), ConfigDecl, false, ConfigQTy, VK_LValue, LLLLoc);
+  SemaRef.MarkFunctionReferenced(LLLLoc, ConfigDecl);
 
-  return BuildCallExpr(S, ConfigDR, LLLLoc, ExecConfig, GGGLoc, nullptr,
-                       /*IsExecConfig=*/true);
+  return SemaRef.BuildCallExpr(S, ConfigDR, LLLLoc, ExecConfig, GGGLoc, nullptr,
+                               /*IsExecConfig=*/true);
 }
 
-CUDAFunctionTarget Sema::IdentifyCUDATarget(const ParsedAttributesView &Attrs) {
+CUDAFunctionTarget SemaCUDA::IdentifyTarget(const ParsedAttributesView &Attrs) {
   bool HasHostAttr = false;
   bool HasDeviceAttr = false;
   bool HasGlobalAttr = false;
@@ -112,12 +115,11 @@ static bool hasAttr(const Decl *D, bool IgnoreImplicitAttr) {
          });
 }
 
-Sema::CUDATargetContextRAII::CUDATargetContextRAII(Sema &S_,
-                                                   CUDATargetContextKind K,
-                                                   Decl *D)
+SemaCUDA::CUDATargetContextRAII::CUDATargetContextRAII(
+    SemaCUDA &S_, SemaCUDA::CUDATargetContextKind K, Decl *D)
     : S(S_) {
   SavedCtx = S.CurCUDATargetCtx;
-  assert(K == CTCK_InitGlobalVar);
+  assert(K == SemaCUDA::CTCK_InitGlobalVar);
   auto *VD = dyn_cast_or_null<VarDecl>(D);
   if (VD && VD->hasGlobalStorage() && !VD->isStaticLocal()) {
     auto Target = CUDAFunctionTarget::Host;
@@ -130,8 +132,8 @@ Sema::CUDATargetContextRAII::CUDATargetContextRAII(Sema &S_,
   }
 }
 
-/// IdentifyCUDATarget - Determine the CUDA compilation target for this function
-CUDAFunctionTarget Sema::IdentifyCUDATarget(const FunctionDecl *D,
+/// IdentifyTarget - Determine the CUDA compilation target for this function
+CUDAFunctionTarget SemaCUDA::IdentifyTarget(const FunctionDecl *D,
                                             bool IgnoreImplicitHDAttr) {
   // Code that lives outside a function gets the target from CurCUDATargetCtx.
   if (D == nullptr)
@@ -160,7 +162,7 @@ CUDAFunctionTarget Sema::IdentifyCUDATarget(const FunctionDecl *D,
 }
 
 /// IdentifyTarget - Determine the CUDA compilation target for this variable.
-Sema::CUDAVariableTarget Sema::IdentifyCUDATarget(const VarDecl *Var) {
+SemaCUDA::CUDAVariableTarget SemaCUDA::IdentifyTarget(const VarDecl *Var) {
   if (Var->hasAttr<HIPManagedAttr>())
     return CVT_Unified;
   // Only constexpr and const variabless with implicit constant attribute
@@ -180,7 +182,7 @@ Sema::CUDAVariableTarget Sema::IdentifyCUDATarget(const VarDecl *Var) {
   //  - on both sides in host device functions
   //  - on device side in device or global functions
   if (auto *FD = dyn_cast<FunctionDecl>(Var->getDeclContext())) {
-    switch (IdentifyCUDATarget(FD)) {
+    switch (IdentifyTarget(FD)) {
     case CUDAFunctionTarget::HostDevice:
       return CVT_Both;
     case CUDAFunctionTarget::Device:
@@ -221,21 +223,21 @@ Sema::CUDAVariableTarget Sema::IdentifyCUDATarget(const VarDecl *Var) {
 // | hd | h  | SS  | WS  | (d) |
 // | hd | hd | HD  | HD  | (b) |
 
-Sema::CUDAFunctionPreference
-Sema::IdentifyCUDAPreference(const FunctionDecl *Caller,
+SemaCUDA::CUDAFunctionPreference
+SemaCUDA::IdentifyPreference(const FunctionDecl *Caller,
                              const FunctionDecl *Callee) {
   assert(Callee && "Callee must be valid.");
 
   // Treat ctor/dtor as host device function in device var initializer to allow
   // trivial ctor/dtor without device attr to be used. Non-trivial ctor/dtor
-  // will be diagnosed by checkAllowedCUDAInitializer.
+  // will be diagnosed by checkAllowedInitializer.
   if (Caller == nullptr && CurCUDATargetCtx.Kind == CTCK_InitGlobalVar &&
       CurCUDATargetCtx.Target == CUDAFunctionTarget::Device &&
       (isa<CXXConstructorDecl>(Callee) || isa<CXXDestructorDecl>(Callee)))
     return CFP_HostDevice;
 
-  CUDAFunctionTarget CallerTarget = IdentifyCUDATarget(Caller);
-  CUDAFunctionTarget CalleeTarget = IdentifyCUDATarget(Callee);
+  CUDAFunctionTarget CallerTarget = IdentifyTarget(Caller);
+  CUDAFunctionTarget CalleeTarget = IdentifyTarget(Callee);
 
   // If one of the targets is invalid, the check always fails, no matter what
   // the other target is.
@@ -309,13 +311,13 @@ template <typename AttrT> static bool hasImplicitAttr(const FunctionDecl *D) {
   return D->isImplicit();
 }
 
-bool Sema::isCUDAImplicitHostDeviceFunction(const FunctionDecl *D) {
+bool SemaCUDA::isImplicitHostDeviceFunction(const FunctionDecl *D) {
   bool IsImplicitDevAttr = hasImplicitAttr<CUDADeviceAttr>(D);
   bool IsImplicitHostAttr = hasImplicitAttr<CUDAHostAttr>(D);
   return IsImplicitDevAttr && IsImplicitHostAttr;
 }
 
-void Sema::EraseUnwantedCUDAMatches(
+void SemaCUDA::EraseUnwantedMatches(
     const FunctionDecl *Caller,
     SmallVectorImpl<std::pair<DeclAccessPair, FunctionDecl *>> &Matches) {
   if (Matches.size() <= 1)
@@ -325,7 +327,7 @@ void Sema::EraseUnwantedCUDAMatches(
 
   // Gets the CUDA function preference for a call from Caller to Match.
   auto GetCFP = [&](const Pair &Match) {
-    return IdentifyCUDAPreference(Caller, Match.second);
+    return IdentifyPreference(Caller, Match.second);
   };
 
   // Find the best call preference among the functions in Matches.
@@ -367,7 +369,7 @@ resolveCalleeCUDATargetConflict(CUDAFunctionTarget Target1,
   return false;
 }
 
-bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
+bool SemaCUDA::inferTargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
                                                    CXXSpecialMemberKind CSM,
                                                    CXXMethodDecl *MemberDecl,
                                                    bool ConstRHS,
@@ -388,7 +390,7 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
 
   // We're going to invoke special member lookup; mark that these special
   // members are called from this one, and not from its caller.
-  ContextRAII MethodContext(*this, MemberDecl);
+  Sema::ContextRAII MethodContext(SemaRef, MemberDecl);
 
   // Look for special members in base classes that should be invoked from here.
   // Infer the target of this member base on the ones it should call.
@@ -412,17 +414,17 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
 
     CXXRecordDecl *BaseClassDecl = cast<CXXRecordDecl>(BaseType->getDecl());
     Sema::SpecialMemberOverloadResult SMOR =
-        LookupSpecialMember(BaseClassDecl, CSM,
-                            /* ConstArg */ ConstRHS,
-                            /* VolatileArg */ false,
-                            /* RValueThis */ false,
-                            /* ConstThis */ false,
-                            /* VolatileThis */ false);
+        SemaRef.LookupSpecialMember(BaseClassDecl, CSM,
+                                    /* ConstArg */ ConstRHS,
+                                    /* VolatileArg */ false,
+                                    /* RValueThis */ false,
+                                    /* ConstThis */ false,
+                                    /* VolatileThis */ false);
 
     if (!SMOR.getMethod())
       continue;
 
-    CUDAFunctionTarget BaseMethodTarget = IdentifyCUDATarget(SMOR.getMethod());
+    CUDAFunctionTarget BaseMethodTarget = IdentifyTarget(SMOR.getMethod());
     if (!InferredTarget) {
       InferredTarget = BaseMethodTarget;
     } else {
@@ -435,7 +437,8 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
               << (unsigned)CSM << llvm::to_underlying(*InferredTarget)
               << llvm::to_underlying(BaseMethodTarget);
         }
-        MemberDecl->addAttr(CUDAInvalidTargetAttr::CreateImplicit(Context));
+        MemberDecl->addAttr(
+            CUDAInvalidTargetAttr::CreateImplicit(getASTContext()));
         return true;
       }
     }
@@ -448,25 +451,24 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
     }
 
     const RecordType *FieldType =
-        Context.getBaseElementType(F->getType())->getAs<RecordType>();
+        getASTContext().getBaseElementType(F->getType())->getAs<RecordType>();
     if (!FieldType) {
       continue;
     }
 
     CXXRecordDecl *FieldRecDecl = cast<CXXRecordDecl>(FieldType->getDecl());
     Sema::SpecialMemberOverloadResult SMOR =
-        LookupSpecialMember(FieldRecDecl, CSM,
-                            /* ConstArg */ ConstRHS && !F->isMutable(),
-                            /* VolatileArg */ false,
-                            /* RValueThis */ false,
-                            /* ConstThis */ false,
-                            /* VolatileThis */ false);
+        SemaRef.LookupSpecialMember(FieldRecDecl, CSM,
+                                    /* ConstArg */ ConstRHS && !F->isMutable(),
+                                    /* VolatileArg */ false,
+                                    /* RValueThis */ false,
+                                    /* ConstThis */ false,
+                                    /* VolatileThis */ false);
 
     if (!SMOR.getMethod())
       continue;
 
-    CUDAFunctionTarget FieldMethodTarget =
-        IdentifyCUDATarget(SMOR.getMethod());
+    CUDAFunctionTarget FieldMethodTarget = IdentifyTarget(SMOR.getMethod());
     if (!InferredTarget) {
       InferredTarget = FieldMethodTarget;
     } else {
@@ -479,7 +481,8 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
               << (unsigned)CSM << llvm::to_underlying(*InferredTarget)
               << llvm::to_underlying(FieldMethodTarget);
         }
-        MemberDecl->addAttr(CUDAInvalidTargetAttr::CreateImplicit(Context));
+        MemberDecl->addAttr(
+            CUDAInvalidTargetAttr::CreateImplicit(getASTContext()));
         return true;
       }
     }
@@ -499,16 +502,16 @@ bool Sema::inferCUDATargetForImplicitSpecialMember(CXXRecordDecl *ClassDecl,
   // We either setting attributes first time, or the inferred ones must match
   // previously set ones.
   if (NeedsD && !HasD)
-    MemberDecl->addAttr(CUDADeviceAttr::CreateImplicit(Context));
+    MemberDecl->addAttr(CUDADeviceAttr::CreateImplicit(getASTContext()));
   if (NeedsH && !HasH)
-    MemberDecl->addAttr(CUDAHostAttr::CreateImplicit(Context));
+    MemberDecl->addAttr(CUDAHostAttr::CreateImplicit(getASTContext()));
 
   return false;
 }
 
-bool Sema::isEmptyCudaConstructor(SourceLocation Loc, CXXConstructorDecl *CD) {
+bool SemaCUDA::isEmptyConstructor(SourceLocation Loc, CXXConstructorDecl *CD) {
   if (!CD->isDefined() && CD->isTemplateInstantiation())
-    InstantiateFunctionDefinition(Loc, CD->getFirstDecl());
+    SemaRef.InstantiateFunctionDefinition(Loc, CD->getFirstDecl());
 
   // (E.2.3.1, CUDA 7.5) A constructor for a class type is considered
   // empty at a point in the translation unit, if it is either a
@@ -536,7 +539,7 @@ bool Sema::isEmptyCudaConstructor(SourceLocation Loc, CXXConstructorDecl *CD) {
   if (!llvm::all_of(CD->inits(), [&](const CXXCtorInitializer *CI) {
         if (const CXXConstructExpr *CE =
                 dyn_cast<CXXConstructExpr>(CI->getInit()))
-          return isEmptyCudaConstructor(Loc, CE->getConstructor());
+          return isEmptyConstructor(Loc, CE->getConstructor());
         return false;
       }))
     return false;
@@ -544,13 +547,13 @@ bool Sema::isEmptyCudaConstructor(SourceLocation Loc, CXXConstructorDecl *CD) {
   return true;
 }
 
-bool Sema::isEmptyCudaDestructor(SourceLocation Loc, CXXDestructorDecl *DD) {
+bool SemaCUDA::isEmptyDestructor(SourceLocation Loc, CXXDestructorDecl *DD) {
   // No destructor -> no problem.
   if (!DD)
     return true;
 
   if (!DD->isDefined() && DD->isTemplateInstantiation())
-    InstantiateFunctionDefinition(Loc, DD->getFirstDecl());
+    SemaRef.InstantiateFunctionDefinition(Loc, DD->getFirstDecl());
 
   // (E.2.3.1, CUDA 7.5) A destructor for a class type is considered
   // empty at a point in the translation unit, if it is either a
@@ -579,7 +582,7 @@ bool Sema::isEmptyCudaDestructor(SourceLocation Loc, CXXDestructorDecl *DD) {
   // destructors for all base classes...
   if (!llvm::all_of(ClassDecl->bases(), [&](const CXXBaseSpecifier &BS) {
         if (CXXRecordDecl *RD = BS.getType()->getAsCXXRecordDecl())
-          return isEmptyCudaDestructor(Loc, RD->getDestructor());
+          return isEmptyDestructor(Loc, RD->getDestructor());
         return true;
       }))
     return false;
@@ -589,7 +592,7 @@ bool Sema::isEmptyCudaDestructor(SourceLocation Loc, CXXDestructorDecl *DD) {
         if (CXXRecordDecl *RD = Field->getType()
                                     ->getBaseElementTypeUnsafe()
                                     ->getAsCXXRecordDecl())
-          return isEmptyCudaDestructor(Loc, RD->getDestructor());
+          return isEmptyDestructor(Loc, RD->getDestructor());
         return true;
       }))
     return false;
@@ -620,7 +623,7 @@ bool IsDependentVar(VarDecl *VD) {
 // __shared__ variables whether they are local or not (they all are implicitly
 // static in CUDA). One exception is that CUDA allows constant initializers
 // for __constant__ and __device__ variables.
-bool HasAllowedCUDADeviceStaticInitializer(Sema &S, VarDecl *VD,
+bool HasAllowedCUDADeviceStaticInitializer(SemaCUDA &S, VarDecl *VD,
                                            CUDAInitializerCheckKind CheckKind) {
   assert(!VD->isInvalidDecl() && VD->hasGlobalStorage());
   assert(!IsDependentVar(VD) && "do not check dependent var");
@@ -629,30 +632,30 @@ bool HasAllowedCUDADeviceStaticInitializer(Sema &S, VarDecl *VD,
     if (!Init)
       return true;
     if (const auto *CE = dyn_cast<CXXConstructExpr>(Init)) {
-      return S.isEmptyCudaConstructor(VD->getLocation(), CE->getConstructor());
+      return S.isEmptyConstructor(VD->getLocation(), CE->getConstructor());
     }
     return false;
   };
   auto IsConstantInit = [&](const Expr *Init) {
     assert(Init);
-    ASTContext::CUDAConstantEvalContextRAII EvalCtx(S.Context,
+    ASTContext::CUDAConstantEvalContextRAII EvalCtx(S.getASTContext(),
                                                     /*NoWronSidedVars=*/true);
-    return Init->isConstantInitializer(S.Context,
+    return Init->isConstantInitializer(S.getASTContext(),
                                        VD->getType()->isReferenceType());
   };
   auto HasEmptyDtor = [&](VarDecl *VD) {
     if (const auto *RD = VD->getType()->getAsCXXRecordDecl())
-      return S.isEmptyCudaDestructor(VD->getLocation(), RD->getDestructor());
+      return S.isEmptyDestructor(VD->getLocation(), RD->getDestructor());
     return true;
   };
   if (CheckKind == CICK_Shared)
     return IsEmptyInit(Init) && HasEmptyDtor(VD);
-  return S.LangOpts.GPUAllowDeviceInit ||
+  return S.getLangOpts().GPUAllowDeviceInit ||
          ((IsEmptyInit(Init) || IsConstantInit(Init)) && HasEmptyDtor(VD));
 }
 } // namespace
 
-void Sema::checkAllowedCUDAInitializer(VarDecl *VD) {
+void SemaCUDA::checkAllowedInitializer(VarDecl *VD) {
   // Return early if VD is inside a non-instantiated template function since
   // the implicit constructor is not defined yet.
   if (const FunctionDecl *FD =
@@ -688,7 +691,7 @@ void Sema::checkAllowedCUDAInitializer(VarDecl *VD) {
       InitFn = CE->getDirectCallee();
     }
     if (InitFn) {
-      CUDAFunctionTarget InitFnTarget = IdentifyCUDATarget(InitFn);
+      CUDAFunctionTarget InitFnTarget = IdentifyTarget(InitFn);
       if (InitFnTarget != CUDAFunctionTarget::Host &&
           InitFnTarget != CUDAFunctionTarget::HostDevice) {
         Diag(VD->getLocation(), diag::err_ref_bad_target_global_initializer)
@@ -700,22 +703,22 @@ void Sema::checkAllowedCUDAInitializer(VarDecl *VD) {
   }
 }
 
-void Sema::CUDARecordImplicitHostDeviceFuncUsedByDevice(
+void SemaCUDA::RecordImplicitHostDeviceFuncUsedByDevice(
     const FunctionDecl *Callee) {
-  FunctionDecl *Caller = getCurFunctionDecl(/*AllowLambda=*/true);
+  FunctionDecl *Caller = SemaRef.getCurFunctionDecl(/*AllowLambda=*/true);
   if (!Caller)
     return;
 
-  if (!isCUDAImplicitHostDeviceFunction(Callee))
+  if (!isImplicitHostDeviceFunction(Callee))
     return;
 
-  CUDAFunctionTarget CallerTarget = IdentifyCUDATarget(Caller);
+  CUDAFunctionTarget CallerTarget = IdentifyTarget(Caller);
 
   // Record whether an implicit host device function is used on device side.
   if (CallerTarget != CUDAFunctionTarget::Device &&
       CallerTarget != CUDAFunctionTarget::Global &&
       (CallerTarget != CUDAFunctionTarget::HostDevice ||
-       (isCUDAImplicitHostDeviceFunction(Caller) &&
+       (isImplicitHostDeviceFunction(Caller) &&
         !getASTContext().CUDAImplicitHostDeviceFunUsedByDevice.count(Caller))))
     return;
 
@@ -731,18 +734,18 @@ void Sema::CUDARecordImplicitHostDeviceFuncUsedByDevice(
 //    system header, in which case we leave the constexpr function unattributed.
 //
 // In addition, all function decls are treated as __host__ __device__ when
-// ForceCUDAHostDeviceDepth > 0 (corresponding to code within a
+// ForceHostDeviceDepth > 0 (corresponding to code within a
 //   #pragma clang force_cuda_host_device_begin/end
 // pair).
-void Sema::maybeAddCUDAHostDeviceAttrs(FunctionDecl *NewD,
+void SemaCUDA::maybeAddHostDeviceAttrs(FunctionDecl *NewD,
                                        const LookupResult &Previous) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
 
-  if (ForceCUDAHostDeviceDepth > 0) {
+  if (ForceHostDeviceDepth > 0) {
     if (!NewD->hasAttr<CUDAHostAttr>())
-      NewD->addAttr(CUDAHostAttr::CreateImplicit(Context));
+      NewD->addAttr(CUDAHostAttr::CreateImplicit(getASTContext()));
     if (!NewD->hasAttr<CUDADeviceAttr>())
-      NewD->addAttr(CUDADeviceAttr::CreateImplicit(Context));
+      NewD->addAttr(CUDADeviceAttr::CreateImplicit(getASTContext()));
     return;
   }
 
@@ -753,8 +756,8 @@ void Sema::maybeAddCUDAHostDeviceAttrs(FunctionDecl *NewD,
       !NewD->hasAttr<CUDAGlobalAttr>() &&
       (NewD->getDescribedFunctionTemplate() ||
        NewD->isFunctionTemplateSpecialization())) {
-    NewD->addAttr(CUDAHostAttr::CreateImplicit(Context));
-    NewD->addAttr(CUDADeviceAttr::CreateImplicit(Context));
+    NewD->addAttr(CUDAHostAttr::CreateImplicit(getASTContext()));
+    NewD->addAttr(CUDADeviceAttr::CreateImplicit(getASTContext()));
     return;
   }
 
@@ -771,8 +774,9 @@ void Sema::maybeAddCUDAHostDeviceAttrs(FunctionDecl *NewD,
     FunctionDecl *OldD = D->getAsFunction();
     return OldD && OldD->hasAttr<CUDADeviceAttr>() &&
            !OldD->hasAttr<CUDAHostAttr>() &&
-           !IsOverload(NewD, OldD, /* UseMemberUsingDeclRules = */ false,
-                       /* ConsiderCudaAttrs = */ false);
+           !SemaRef.IsOverload(NewD, OldD,
+                               /* UseMemberUsingDeclRules = */ false,
+                               /* ConsiderCudaAttrs = */ false);
   };
   auto It = llvm::find_if(Previous, IsMatchingDeviceFn);
   if (It != Previous.end()) {
@@ -781,7 +785,7 @@ void Sema::maybeAddCUDAHostDeviceAttrs(FunctionDecl *NewD,
     // in a system header, in which case we simply return without making NewD
     // host+device.
     NamedDecl *Match = *It;
-    if (!getSourceManager().isInSystemHeader(Match->getLocation())) {
+    if (!SemaRef.getSourceManager().isInSystemHeader(Match->getLocation())) {
       Diag(NewD->getLocation(),
            diag::err_cuda_unattributed_constexpr_cannot_overload_device)
           << NewD;
@@ -791,14 +795,14 @@ void Sema::maybeAddCUDAHostDeviceAttrs(FunctionDecl *NewD,
     return;
   }
 
-  NewD->addAttr(CUDAHostAttr::CreateImplicit(Context));
-  NewD->addAttr(CUDADeviceAttr::CreateImplicit(Context));
+  NewD->addAttr(CUDAHostAttr::CreateImplicit(getASTContext()));
+  NewD->addAttr(CUDADeviceAttr::CreateImplicit(getASTContext()));
 }
 
 // TODO: `__constant__` memory may be a limited resource for certain targets.
 // A safeguard may be needed at the end of compilation pipeline if
 // `__constant__` memory usage goes beyond limit.
-void Sema::MaybeAddCUDAConstantAttr(VarDecl *VD) {
+void SemaCUDA::MaybeAddConstantAttr(VarDecl *VD) {
   // Do not promote dependent variables since the cotr/dtor/initializer are
   // not determined. Do it after instantiation.
   if (getLangOpts().CUDAIsDevice && !VD->hasAttr<CUDAConstantAttr>() &&
@@ -812,14 +816,15 @@ void Sema::MaybeAddCUDAConstantAttr(VarDecl *VD) {
   }
 }
 
-Sema::SemaDiagnosticBuilder Sema::CUDADiagIfDeviceCode(SourceLocation Loc,
-                                                       unsigned DiagID) {
+SemaBase::SemaDiagnosticBuilder SemaCUDA::DiagIfDeviceCode(SourceLocation Loc,
+                                                           unsigned DiagID) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
-  FunctionDecl *CurFunContext = getCurFunctionDecl(/*AllowLambda=*/true);
+  FunctionDecl *CurFunContext =
+      SemaRef.getCurFunctionDecl(/*AllowLambda=*/true);
   SemaDiagnosticBuilder::Kind DiagKind = [&] {
     if (!CurFunContext)
       return SemaDiagnosticBuilder::K_Nop;
-    switch (CurrentCUDATarget()) {
+    switch (CurrentTarget()) {
     case CUDAFunctionTarget::Global:
     case CUDAFunctionTarget::Device:
       return SemaDiagnosticBuilder::K_Immediate;
@@ -829,27 +834,29 @@ Sema::SemaDiagnosticBuilder Sema::CUDADiagIfDeviceCode(SourceLocation Loc,
       // mode until the function is known-emitted.
       if (!getLangOpts().CUDAIsDevice)
         return SemaDiagnosticBuilder::K_Nop;
-      if (IsLastErrorImmediate && Diags.getDiagnosticIDs()->isBuiltinNote(DiagID))
+      if (SemaRef.IsLastErrorImmediate &&
+          getDiagnostics().getDiagnosticIDs()->isBuiltinNote(DiagID))
         return SemaDiagnosticBuilder::K_Immediate;
-      return (getEmissionStatus(CurFunContext) ==
-              FunctionEmissionStatus::Emitted)
+      return (SemaRef.getEmissionStatus(CurFunContext) ==
+              Sema::FunctionEmissionStatus::Emitted)
                  ? SemaDiagnosticBuilder::K_ImmediateWithCallStack
                  : SemaDiagnosticBuilder::K_Deferred;
     default:
       return SemaDiagnosticBuilder::K_Nop;
     }
   }();
-  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, CurFunContext, *this);
+  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, CurFunContext, SemaRef);
 }
 
-Sema::SemaDiagnosticBuilder Sema::CUDADiagIfHostCode(SourceLocation Loc,
+Sema::SemaDiagnosticBuilder SemaCUDA::DiagIfHostCode(SourceLocation Loc,
                                                      unsigned DiagID) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
-  FunctionDecl *CurFunContext = getCurFunctionDecl(/*AllowLambda=*/true);
+  FunctionDecl *CurFunContext =
+      SemaRef.getCurFunctionDecl(/*AllowLambda=*/true);
   SemaDiagnosticBuilder::Kind DiagKind = [&] {
     if (!CurFunContext)
       return SemaDiagnosticBuilder::K_Nop;
-    switch (CurrentCUDATarget()) {
+    switch (CurrentTarget()) {
     case CUDAFunctionTarget::Host:
       return SemaDiagnosticBuilder::K_Immediate;
     case CUDAFunctionTarget::HostDevice:
@@ -858,40 +865,41 @@ Sema::SemaDiagnosticBuilder Sema::CUDADiagIfHostCode(SourceLocation Loc,
       // mode until the function is known-emitted.
       if (getLangOpts().CUDAIsDevice)
         return SemaDiagnosticBuilder::K_Nop;
-      if (IsLastErrorImmediate && Diags.getDiagnosticIDs()->isBuiltinNote(DiagID))
+      if (SemaRef.IsLastErrorImmediate &&
+          getDiagnostics().getDiagnosticIDs()->isBuiltinNote(DiagID))
         return SemaDiagnosticBuilder::K_Immediate;
-      return (getEmissionStatus(CurFunContext) ==
-              FunctionEmissionStatus::Emitted)
+      return (SemaRef.getEmissionStatus(CurFunContext) ==
+              Sema::FunctionEmissionStatus::Emitted)
                  ? SemaDiagnosticBuilder::K_ImmediateWithCallStack
                  : SemaDiagnosticBuilder::K_Deferred;
     default:
       return SemaDiagnosticBuilder::K_Nop;
     }
   }();
-  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, CurFunContext, *this);
+  return SemaDiagnosticBuilder(DiagKind, Loc, DiagID, CurFunContext, SemaRef);
 }
 
-bool Sema::CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee) {
+bool SemaCUDA::CheckCall(SourceLocation Loc, FunctionDecl *Callee) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
   assert(Callee && "Callee may not be null.");
 
-  const auto &ExprEvalCtx = currentEvaluationContext();
+  const auto &ExprEvalCtx = SemaRef.currentEvaluationContext();
   if (ExprEvalCtx.isUnevaluated() || ExprEvalCtx.isConstantEvaluated())
     return true;
 
   // FIXME: Is bailing out early correct here?  Should we instead assume that
   // the caller is a global initializer?
-  FunctionDecl *Caller = getCurFunctionDecl(/*AllowLambda=*/true);
+  FunctionDecl *Caller = SemaRef.getCurFunctionDecl(/*AllowLambda=*/true);
   if (!Caller)
     return true;
 
   // If the caller is known-emitted, mark the callee as known-emitted.
   // Otherwise, mark the call in our call graph so we can traverse it later.
-  bool CallerKnownEmitted =
-      getEmissionStatus(Caller) == FunctionEmissionStatus::Emitted;
+  bool CallerKnownEmitted = SemaRef.getEmissionStatus(Caller) ==
+                            Sema::FunctionEmissionStatus::Emitted;
   SemaDiagnosticBuilder::Kind DiagKind = [this, Caller, Callee,
                                           CallerKnownEmitted] {
-    switch (IdentifyCUDAPreference(Caller, Callee)) {
+    switch (IdentifyPreference(Caller, Callee)) {
     case CFP_Never:
     case CFP_WrongSide:
       assert(Caller && "Never/wrongSide calls require a non-null caller");
@@ -908,7 +916,7 @@ bool Sema::CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee) {
 
   if (DiagKind == SemaDiagnosticBuilder::K_Nop) {
     // For -fgpu-rdc, keep track of external kernels used by host functions.
-    if (LangOpts.CUDAIsDevice && LangOpts.GPURelocatableDeviceCode &&
+    if (getLangOpts().CUDAIsDevice && getLangOpts().GPURelocatableDeviceCode &&
         Callee->hasAttr<CUDAGlobalAttr>() && !Callee->isDefined() &&
         (!Caller || (!Caller->getDescribedFunctionTemplate() &&
                      getASTContext().GetGVALinkageForFunction(Caller) ==
@@ -924,12 +932,13 @@ bool Sema::CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee) {
   if (!LocsWithCUDACallDiags.insert({Caller, Loc}).second)
     return true;
 
-  SemaDiagnosticBuilder(DiagKind, Loc, diag::err_ref_bad_target, Caller, *this)
-      << llvm::to_underlying(IdentifyCUDATarget(Callee)) << /*function*/ 0
-      << Callee << llvm::to_underlying(IdentifyCUDATarget(Caller));
+  SemaDiagnosticBuilder(DiagKind, Loc, diag::err_ref_bad_target, Caller,
+                        SemaRef)
+      << llvm::to_underlying(IdentifyTarget(Callee)) << /*function*/ 0 << Callee
+      << llvm::to_underlying(IdentifyTarget(Caller));
   if (!Callee->getBuiltinID())
     SemaDiagnosticBuilder(DiagKind, Callee->getLocation(),
-                          diag::note_previous_decl, Caller, *this)
+                          diag::note_previous_decl, Caller, SemaRef)
         << Callee;
   return DiagKind != SemaDiagnosticBuilder::K_Immediate &&
          DiagKind != SemaDiagnosticBuilder::K_ImmediateWithCallStack;
@@ -940,7 +949,7 @@ bool Sema::CheckCUDACall(SourceLocation Loc, FunctionDecl *Callee) {
 // defined and uses the capture by reference when the lambda is called. When
 // the capture and use happen on different sides, the capture is invalid and
 // should be diagnosed.
-void Sema::CUDACheckLambdaCapture(CXXMethodDecl *Callee,
+void SemaCUDA::CheckLambdaCapture(CXXMethodDecl *Callee,
                                   const sema::Capture &Capture) {
   // In host compilation we only need to check lambda functions emitted on host
   // side. In such lambda functions, a reference capture is invalid only
@@ -950,12 +959,12 @@ void Sema::CUDACheckLambdaCapture(CXXMethodDecl *Callee,
   // kernel cannot pass a lambda back to a host function since we cannot
   // define a kernel argument type which can hold the lambda before the lambda
   // itself is defined.
-  if (!LangOpts.CUDAIsDevice)
+  if (!getLangOpts().CUDAIsDevice)
     return;
 
   // File-scope lambda can only do init captures for global variables, which
   // results in passing by value for these global variables.
-  FunctionDecl *Caller = getCurFunctionDecl(/*AllowLambda=*/true);
+  FunctionDecl *Caller = SemaRef.getCurFunctionDecl(/*AllowLambda=*/true);
   if (!Caller)
     return;
 
@@ -972,7 +981,7 @@ void Sema::CUDACheckLambdaCapture(CXXMethodDecl *Callee,
   auto DiagKind = SemaDiagnosticBuilder::K_Deferred;
   if (Capture.isVariableCapture() && !getLangOpts().HIPStdPar) {
     SemaDiagnosticBuilder(DiagKind, Capture.getLocation(),
-                          diag::err_capture_bad_target, Callee, *this)
+                          diag::err_capture_bad_target, Callee, SemaRef)
         << Capture.getVariable();
   } else if (Capture.isThisCapture()) {
     // Capture of this pointer is allowed since this pointer may be pointing to
@@ -981,28 +990,28 @@ void Sema::CUDACheckLambdaCapture(CXXMethodDecl *Callee,
     // accessible on device side.
     SemaDiagnosticBuilder(DiagKind, Capture.getLocation(),
                           diag::warn_maybe_capture_bad_target_this_ptr, Callee,
-                          *this);
+                          SemaRef);
   }
 }
 
-void Sema::CUDASetLambdaAttrs(CXXMethodDecl *Method) {
+void SemaCUDA::SetLambdaAttrs(CXXMethodDecl *Method) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
   if (Method->hasAttr<CUDAHostAttr>() || Method->hasAttr<CUDADeviceAttr>())
     return;
-  Method->addAttr(CUDADeviceAttr::CreateImplicit(Context));
-  Method->addAttr(CUDAHostAttr::CreateImplicit(Context));
+  Method->addAttr(CUDADeviceAttr::CreateImplicit(getASTContext()));
+  Method->addAttr(CUDAHostAttr::CreateImplicit(getASTContext()));
 }
 
-void Sema::checkCUDATargetOverload(FunctionDecl *NewFD,
+void SemaCUDA::checkTargetOverload(FunctionDecl *NewFD,
                                    const LookupResult &Previous) {
   assert(getLangOpts().CUDA && "Should only be called during CUDA compilation");
-  CUDAFunctionTarget NewTarget = IdentifyCUDATarget(NewFD);
+  CUDAFunctionTarget NewTarget = IdentifyTarget(NewFD);
   for (NamedDecl *OldND : Previous) {
     FunctionDecl *OldFD = OldND->getAsFunction();
     if (!OldFD)
       continue;
 
-    CUDAFunctionTarget OldTarget = IdentifyCUDATarget(OldFD);
+    CUDAFunctionTarget OldTarget = IdentifyTarget(OldFD);
     // Don't allow HD and global functions to overload other functions with the
     // same signature.  We allow overloading based on CUDA attributes so that
     // functions can have different implementations on the host and device, but
@@ -1010,17 +1019,17 @@ void Sema::checkCUDATargetOverload(FunctionDecl *NewFD,
     // should have the same implementation on both sides.
     if (NewTarget != OldTarget &&
         ((NewTarget == CUDAFunctionTarget::HostDevice &&
-          !(LangOpts.OffloadImplicitHostDeviceTemplates &&
-            isCUDAImplicitHostDeviceFunction(NewFD) &&
+          !(getLangOpts().OffloadImplicitHostDeviceTemplates &&
+            isImplicitHostDeviceFunction(NewFD) &&
             OldTarget == CUDAFunctionTarget::Device)) ||
          (OldTarget == CUDAFunctionTarget::HostDevice &&
-          !(LangOpts.OffloadImplicitHostDeviceTemplates &&
-            isCUDAImplicitHostDeviceFunction(OldFD) &&
+          !(getLangOpts().OffloadImplicitHostDeviceTemplates &&
+            isImplicitHostDeviceFunction(OldFD) &&
             NewTarget == CUDAFunctionTarget::Device)) ||
          (NewTarget == CUDAFunctionTarget::Global) ||
          (OldTarget == CUDAFunctionTarget::Global)) &&
-        !IsOverload(NewFD, OldFD, /* UseMemberUsingDeclRules = */ false,
-                    /* ConsiderCudaAttrs = */ false)) {
+        !SemaRef.IsOverload(NewFD, OldFD, /* UseMemberUsingDeclRules = */ false,
+                            /* ConsiderCudaAttrs = */ false)) {
       Diag(NewFD->getLocation(), diag::err_cuda_ovl_target)
           << llvm::to_underlying(NewTarget) << NewFD->getDeclName()
           << llvm::to_underlying(OldTarget) << OldFD;
@@ -1041,21 +1050,21 @@ static void copyAttrIfPresent(Sema &S, FunctionDecl *FD,
   }
 }
 
-void Sema::inheritCUDATargetAttrs(FunctionDecl *FD,
+void SemaCUDA::inheritTargetAttrs(FunctionDecl *FD,
                                   const FunctionTemplateDecl &TD) {
   const FunctionDecl &TemplateFD = *TD.getTemplatedDecl();
-  copyAttrIfPresent<CUDAGlobalAttr>(*this, FD, TemplateFD);
-  copyAttrIfPresent<CUDAHostAttr>(*this, FD, TemplateFD);
-  copyAttrIfPresent<CUDADeviceAttr>(*this, FD, TemplateFD);
+  copyAttrIfPresent<CUDAGlobalAttr>(SemaRef, FD, TemplateFD);
+  copyAttrIfPresent<CUDAHostAttr>(SemaRef, FD, TemplateFD);
+  copyAttrIfPresent<CUDADeviceAttr>(SemaRef, FD, TemplateFD);
 }
 
-std::string Sema::getCudaConfigureFuncName() const {
+std::string SemaCUDA::getConfigureFuncName() const {
   if (getLangOpts().HIP)
     return getLangOpts().HIPUseNewLaunchAPI ? "__hipPushCallConfiguration"
                                             : "hipConfigureCall";
 
   // New CUDA kernel launch sequence.
-  if (CudaFeatureEnabled(Context.getTargetInfo().getSDKVersion(),
+  if (CudaFeatureEnabled(getASTContext().getTargetInfo().getSDKVersion(),
                          CudaFeature::CUDA_USES_NEW_LAUNCH))
     return "__cudaPushCallConfiguration";
 
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 720e56692359b..17032d1370521 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -45,6 +45,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
@@ -10595,12 +10596,12 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
 
     // We do not add HD attributes to specializations here because
     // they may have different constexpr-ness compared to their
-    // templates and, after maybeAddCUDAHostDeviceAttrs() is applied,
+    // templates and, after maybeAddHostDeviceAttrs() is applied,
     // may end up with different effective targets. Instead, a
     // specialization inherits its target attributes from its template
     // in the CheckFunctionTemplateSpecialization() call below.
     if (getLangOpts().CUDA && !isFunctionTemplateSpecialization)
-      maybeAddCUDAHostDeviceAttrs(NewFD, Previous);
+      CUDA().maybeAddHostDeviceAttrs(NewFD, Previous);
 
     // Handle explict specializations of function templates
     // and friend function declarations with an explicit
@@ -10898,12 +10899,12 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
 
   if (getLangOpts().CUDA) {
     IdentifierInfo *II = NewFD->getIdentifier();
-    if (II && II->isStr(getCudaConfigureFuncName()) &&
+    if (II && II->isStr(CUDA().getConfigureFuncName()) &&
         !NewFD->isInvalidDecl() &&
         NewFD->getDeclContext()->getRedeclContext()->isTranslationUnit()) {
       if (!R->castAs<FunctionType>()->getReturnType()->isScalarType())
         Diag(NewFD->getLocation(), diag::err_config_scalar_return)
-            << getCudaConfigureFuncName();
+            << CUDA().getConfigureFuncName();
       Context.setcudaConfigureCallDecl(NewFD);
     }
 
@@ -12398,7 +12399,7 @@ bool Sema::CheckFunctionDeclaration(Scope *S, FunctionDecl *NewFD,
     }
 
     if (!Redeclaration && LangOpts.CUDA)
-      checkCUDATargetOverload(NewFD, Previous);
+      CUDA().checkTargetOverload(NewFD, Previous);
   }
 
   // Check if the function definition uses any AArch64 SME features without
@@ -14415,7 +14416,7 @@ StmtResult Sema::ActOnCXXForRangeIdentifier(Scope *S, SourceLocation IdentLoc,
 void Sema::CheckCompleteVariableDeclaration(VarDecl *var) {
   if (var->isInvalidDecl()) return;
 
-  MaybeAddCUDAConstantAttr(var);
+  CUDA().MaybeAddConstantAttr(var);
 
   if (getLangOpts().OpenCL) {
     // OpenCL v2.0 s6.12.5 - Every block variable declaration must have an
@@ -14829,7 +14830,7 @@ void Sema::FinalizeDeclaration(Decl *ThisDecl) {
   // variables whether they are local or not. CUDA also allows
   // constant initializers for __constant__ and __device__ variables.
   if (getLangOpts().CUDA)
-    checkAllowedCUDAInitializer(VD);
+    CUDA().checkAllowedInitializer(VD);
 
   // Grab the dllimport or dllexport attribute off of the VarDecl.
   const InheritableAttr *DLLAttr = getDLLAttr(VD);
@@ -20666,7 +20667,7 @@ Sema::FunctionEmissionStatus Sema::getEmissionStatus(const FunctionDecl *FD,
     // when compiling for host, device and global functions are never emitted.
     // (Technically, we do emit a host-side stub for global functions, but this
     // doesn't count for our purposes here.)
-    CUDAFunctionTarget T = IdentifyCUDATarget(FD);
+    CUDAFunctionTarget T = CUDA().IdentifyTarget(FD);
     if (LangOpts.CUDAIsDevice && T == CUDAFunctionTarget::Host)
       return FunctionEmissionStatus::CUDADiscarded;
     if (!LangOpts.CUDAIsDevice &&
@@ -20691,5 +20692,5 @@ bool Sema::shouldIgnoreInHostDeviceCheck(FunctionDecl *Callee) {
   // for host, only HD functions actually called from the host get marked as
   // known-emitted.
   return LangOpts.CUDA && !LangOpts.CUDAIsDevice &&
-         IdentifyCUDATarget(Callee) == CUDAFunctionTarget::Global;
+         CUDA().IdentifyTarget(Callee) == CUDAFunctionTarget::Global;
 }
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index 56c9d90c9b52b..b7b1fbc625a15 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -39,6 +39,7 @@
 #include "clang/Sema/ParsedAttr.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "llvm/ADT/STLExtras.h"
@@ -5099,8 +5100,8 @@ static void handleSharedAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     return;
   }
   if (S.getLangOpts().CUDA && VD->hasLocalStorage() &&
-      S.CUDADiagIfHostCode(AL.getLoc(), diag::err_cuda_host_shared)
-          << llvm::to_underlying(S.CurrentCUDATarget()))
+      S.CUDA().DiagIfHostCode(AL.getLoc(), diag::err_cuda_host_shared)
+          << llvm::to_underlying(S.CUDA().CurrentTarget()))
     return;
   D->addAttr(::new (S.Context) CUDASharedAttr(S.Context, AL));
 }
@@ -5189,8 +5190,9 @@ static void handleCallConvAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   // Diagnostic is emitted elsewhere: here we store the (valid) AL
   // in the Decl node for syntactic reasoning, e.g., pretty-printing.
   CallingConv CC;
-  if (S.CheckCallingConvAttr(AL, CC, /*FD*/ nullptr,
-                             S.IdentifyCUDATarget(dyn_cast<FunctionDecl>(D))))
+  if (S.CheckCallingConvAttr(
+          AL, CC, /*FD*/ nullptr,
+          S.CUDA().IdentifyTarget(dyn_cast<FunctionDecl>(D))))
     return;
 
   if (!isa<ObjCMethodDecl>(D)) {
@@ -5495,7 +5497,7 @@ bool Sema::CheckCallingConvAttr(const ParsedAttr &Attrs, CallingConv &CC,
   if (LangOpts.CUDA) {
     auto *Aux = Context.getAuxTargetInfo();
     assert(FD || CFT != CUDAFunctionTarget::InvalidTarget);
-    auto CudaTarget = FD ? IdentifyCUDATarget(FD) : CFT;
+    auto CudaTarget = FD ? CUDA().IdentifyTarget(FD) : CFT;
     bool CheckHost = false, CheckDevice = false;
     switch (CudaTarget) {
     case CUDAFunctionTarget::HostDevice:
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 51c14443d2d8f..1fe10375222c5 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -42,6 +42,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/ArrayRef.h"
@@ -9876,15 +9877,15 @@ bool Sema::ShouldDeleteSpecialMember(CXXMethodDecl *MD,
     // failed.
     // For inherited constructors (non-null ICI), CSM may be passed so that MD
     // is treated as certain special member, which may not reflect what special
-    // member MD really is. However inferCUDATargetForImplicitSpecialMember
+    // member MD really is. However inferTargetForImplicitSpecialMember
     // expects CSM to match MD, therefore recalculate CSM.
     assert(ICI || CSM == getSpecialMember(MD));
     auto RealCSM = CSM;
     if (ICI)
       RealCSM = getSpecialMember(MD);
 
-    return inferCUDATargetForImplicitSpecialMember(RD, RealCSM, MD,
-                                                   SMI.ConstArg, Diagnose);
+    return CUDA().inferTargetForImplicitSpecialMember(RD, RealCSM, MD,
+                                                      SMI.ConstArg, Diagnose);
   }
 
   return false;
@@ -14055,7 +14056,7 @@ CXXConstructorDecl *Sema::DeclareImplicitDefaultConstructor(
   setupImplicitSpecialMemberType(DefaultCon, Context.VoidTy, std::nullopt);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::DefaultConstructor, DefaultCon,
         /* ConstRHS */ false,
         /* Diagnose */ false);
@@ -14341,7 +14342,7 @@ CXXDestructorDecl *Sema::DeclareImplicitDestructor(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(Destructor, Context.VoidTy, std::nullopt);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::Destructor, Destructor,
         /* ConstRHS */ false,
         /* Diagnose */ false);
@@ -14983,7 +14984,7 @@ CXXMethodDecl *Sema::DeclareImplicitCopyAssignment(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(CopyAssignment, RetType, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::CopyAssignment, CopyAssignment,
         /* ConstRHS */ Const,
         /* Diagnose */ false);
@@ -15335,7 +15336,7 @@ CXXMethodDecl *Sema::DeclareImplicitMoveAssignment(CXXRecordDecl *ClassDecl) {
   setupImplicitSpecialMemberType(MoveAssignment, RetType, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::MoveAssignment, MoveAssignment,
         /* ConstRHS */ false,
         /* Diagnose */ false);
@@ -15733,7 +15734,7 @@ CXXConstructorDecl *Sema::DeclareImplicitCopyConstructor(
   setupImplicitSpecialMemberType(CopyConstructor, Context.VoidTy, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::CopyConstructor, CopyConstructor,
         /* ConstRHS */ Const,
         /* Diagnose */ false);
@@ -15878,7 +15879,7 @@ CXXConstructorDecl *Sema::DeclareImplicitMoveConstructor(
   setupImplicitSpecialMemberType(MoveConstructor, Context.VoidTy, ArgType);
 
   if (getLangOpts().CUDA)
-    inferCUDATargetForImplicitSpecialMember(
+    CUDA().inferTargetForImplicitSpecialMember(
         ClassDecl, CXXSpecialMemberKind::MoveConstructor, MoveConstructor,
         /* ConstRHS */ false,
         /* Diagnose */ false);
@@ -16184,7 +16185,7 @@ ExprResult Sema::BuildCXXConstructExpr(
              DeclInitType->getBaseElementTypeUnsafe()->getAsCXXRecordDecl()) &&
          "given constructor for wrong type");
   MarkFunctionReferenced(ConstructLoc, Constructor);
-  if (getLangOpts().CUDA && !CheckCUDACall(ConstructLoc, Constructor))
+  if (getLangOpts().CUDA && !CUDA().CheckCall(ConstructLoc, Constructor))
     return ExprError();
 
   return CheckForImmediateInvocation(
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index b294d2bd9f53f..823bf36d88bc9 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -49,6 +49,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaFixItUtils.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
@@ -308,7 +309,7 @@ bool Sema::DiagnoseUseOfDecl(NamedDecl *D, ArrayRef<SourceLocation> Locs,
         DeduceReturnType(FD, Loc))
       return true;
 
-    if (getLangOpts().CUDA && !CheckCUDACall(Loc, FD))
+    if (getLangOpts().CUDA && !CUDA().CheckCall(Loc, FD))
       return true;
 
   }
@@ -17307,7 +17308,7 @@ ExprResult Sema::BuildVAArgExpr(SourceLocation BuiltinLoc,
   // CUDA device code does not support varargs.
   if (getLangOpts().CUDA && getLangOpts().CUDAIsDevice) {
     if (const FunctionDecl *F = dyn_cast<FunctionDecl>(CurContext)) {
-      CUDAFunctionTarget T = IdentifyCUDATarget(F);
+      CUDAFunctionTarget T = CUDA().IdentifyTarget(F);
       if (T == CUDAFunctionTarget::Global || T == CUDAFunctionTarget::Device ||
           T == CUDAFunctionTarget::HostDevice)
         return ExprError(Diag(E->getBeginLoc(), diag::err_va_arg_in_device));
@@ -18961,7 +18962,7 @@ void Sema::MarkFunctionReferenced(SourceLocation Loc, FunctionDecl *Func,
     checkSpecializationReachability(Loc, Func);
 
   if (getLangOpts().CUDA)
-    CheckCUDACall(Loc, Func);
+    CUDA().CheckCall(Loc, Func);
 
   // If we need a definition, try to create one.
   if (NeedDefinition && !Func->getBody()) {
@@ -19108,7 +19109,7 @@ void Sema::MarkFunctionReferenced(SourceLocation Loc, FunctionDecl *Func,
   // side. Therefore keep trying until it is recorded.
   if (LangOpts.OffloadImplicitHostDeviceTemplates && LangOpts.CUDAIsDevice &&
       !getASTContext().CUDAImplicitHostDeviceFunUsedByDevice.count(Func))
-    CUDARecordImplicitHostDeviceFuncUsedByDevice(Func);
+    CUDA().RecordImplicitHostDeviceFuncUsedByDevice(Func);
 
   // If this is the first "real" use, act on that.
   if (OdrUse == OdrUseContext::Used && !Func->isUsed(/*CheckUsedAttr=*/false)) {
@@ -19181,9 +19182,9 @@ MarkVarDeclODRUsed(ValueDecl *V, SourceLocation Loc, Sema &SemaRef,
 
   if (SemaRef.LangOpts.CUDA && Var->hasGlobalStorage()) {
     auto *FD = dyn_cast_or_null<FunctionDecl>(SemaRef.CurContext);
-    auto VarTarget = SemaRef.IdentifyCUDATarget(Var);
-    auto UserTarget = SemaRef.IdentifyCUDATarget(FD);
-    if (VarTarget == Sema::CVT_Host &&
+    auto VarTarget = SemaRef.CUDA().IdentifyTarget(Var);
+    auto UserTarget = SemaRef.CUDA().IdentifyTarget(FD);
+    if (VarTarget == SemaCUDA::CVT_Host &&
         (UserTarget == CUDAFunctionTarget::Device ||
          UserTarget == CUDAFunctionTarget::HostDevice ||
          UserTarget == CUDAFunctionTarget::Global)) {
@@ -19199,7 +19200,7 @@ MarkVarDeclODRUsed(ValueDecl *V, SourceLocation Loc, Sema &SemaRef,
                                ? diag::note_cuda_const_var_unpromoted
                                : diag::note_cuda_host_var);
       }
-    } else if (VarTarget == Sema::CVT_Device &&
+    } else if (VarTarget == SemaCUDA::CVT_Device &&
                !Var->hasAttr<CUDASharedAttr>() &&
                (UserTarget == CUDAFunctionTarget::Host ||
                 UserTarget == CUDAFunctionTarget::HostDevice)) {
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index ce9d5c26e2185..8911257a6f614 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -38,6 +38,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/SemaLambda.h"
 #include "clang/Sema/Template.h"
@@ -884,8 +885,8 @@ ExprResult Sema::BuildCXXThrow(SourceLocation OpLoc, Expr *Ex,
 
   // Exceptions aren't allowed in CUDA device code.
   if (getLangOpts().CUDA)
-    CUDADiagIfDeviceCode(OpLoc, diag::err_cuda_device_exceptions)
-        << "throw" << llvm::to_underlying(CurrentCUDATarget());
+    CUDA().DiagIfDeviceCode(OpLoc, diag::err_cuda_device_exceptions)
+        << "throw" << llvm::to_underlying(CUDA().CurrentTarget());
 
   if (getCurScope() && getCurScope()->isOpenMPSimdDirectiveScope())
     Diag(OpLoc, diag::err_omp_simd_region_cannot_use_stmt) << "throw";
@@ -1708,17 +1709,17 @@ bool Sema::isUsualDeallocationFunction(const CXXMethodDecl *Method) {
   // [CUDA] Ignore this function, if we can't call it.
   const FunctionDecl *Caller = getCurFunctionDecl(/*AllowLambda=*/true);
   if (getLangOpts().CUDA) {
-    auto CallPreference = IdentifyCUDAPreference(Caller, Method);
+    auto CallPreference = CUDA().IdentifyPreference(Caller, Method);
     // If it's not callable at all, it's not the right function.
-    if (CallPreference < CFP_WrongSide)
+    if (CallPreference < SemaCUDA::CFP_WrongSide)
       return false;
-    if (CallPreference == CFP_WrongSide) {
+    if (CallPreference == SemaCUDA::CFP_WrongSide) {
       // Maybe. We have to check if there are better alternatives.
       DeclContext::lookup_result R =
           Method->getDeclContext()->lookup(Method->getDeclName());
       for (const auto *D : R) {
         if (const auto *FD = dyn_cast<FunctionDecl>(D)) {
-          if (IdentifyCUDAPreference(Caller, FD) > CFP_WrongSide)
+          if (CUDA().IdentifyPreference(Caller, FD) > SemaCUDA::CFP_WrongSide)
             return false;
         }
       }
@@ -1737,7 +1738,7 @@ bool Sema::isUsualDeallocationFunction(const CXXMethodDecl *Method) {
   return llvm::none_of(PreventedBy, [&](const FunctionDecl *FD) {
     assert(FD->getNumParams() == 1 &&
            "Only single-operand functions should be in PreventedBy");
-    return IdentifyCUDAPreference(Caller, FD) >= CFP_HostDevice;
+    return CUDA().IdentifyPreference(Caller, FD) >= SemaCUDA::CFP_HostDevice;
   });
 }
 
@@ -1774,7 +1775,7 @@ namespace {
     UsualDeallocFnInfo(Sema &S, DeclAccessPair Found)
         : Found(Found), FD(dyn_cast<FunctionDecl>(Found->getUnderlyingDecl())),
           Destroying(false), HasSizeT(false), HasAlignValT(false),
-          CUDAPref(Sema::CFP_Native) {
+          CUDAPref(SemaCUDA::CFP_Native) {
       // A function template declaration is never a usual deallocation function.
       if (!FD)
         return;
@@ -1800,7 +1801,7 @@ namespace {
 
       // In CUDA, determine how much we'd like / dislike to call this.
       if (S.getLangOpts().CUDA)
-        CUDAPref = S.IdentifyCUDAPreference(
+        CUDAPref = S.CUDA().IdentifyPreference(
             S.getCurFunctionDecl(/*AllowLambda=*/true), FD);
     }
 
@@ -1831,7 +1832,7 @@ namespace {
     DeclAccessPair Found;
     FunctionDecl *FD;
     bool Destroying, HasSizeT, HasAlignValT;
-    Sema::CUDAFunctionPreference CUDAPref;
+    SemaCUDA::CUDAFunctionPreference CUDAPref;
   };
 }
 
@@ -1855,7 +1856,7 @@ static UsualDeallocFnInfo resolveDeallocationOverload(
   for (auto I = R.begin(), E = R.end(); I != E; ++I) {
     UsualDeallocFnInfo Info(S, I.getPair());
     if (!Info || !isNonPlacementDeallocationFunction(S, Info.FD) ||
-        Info.CUDAPref == Sema::CFP_Never)
+        Info.CUDAPref == SemaCUDA::CFP_Never)
       continue;
 
     if (!Best) {
@@ -2956,8 +2957,8 @@ bool Sema::FindAllocationFunctions(SourceLocation StartLoc, SourceRange Range,
     }
 
     if (getLangOpts().CUDA)
-      EraseUnwantedCUDAMatches(getCurFunctionDecl(/*AllowLambda=*/true),
-                               Matches);
+      CUDA().EraseUnwantedMatches(getCurFunctionDecl(/*AllowLambda=*/true),
+                                  Matches);
   } else {
     // C++1y [expr.new]p22:
     //   For a non-placement allocation function, the normal deallocation
diff --git a/clang/lib/Sema/SemaLambda.cpp b/clang/lib/Sema/SemaLambda.cpp
index 5b95bae567b72..35a51c6c2328d 100644
--- a/clang/lib/Sema/SemaLambda.cpp
+++ b/clang/lib/Sema/SemaLambda.cpp
@@ -9,17 +9,18 @@
 //  This file implements semantic analysis for C++ lambda expressions.
 //
 //===----------------------------------------------------------------------===//
-#include "clang/Sema/DeclSpec.h"
+#include "clang/Sema/SemaLambda.h"
 #include "TypeLocBuilder.h"
 #include "clang/AST/ASTLambda.h"
 #include "clang/AST/ExprCXX.h"
 #include "clang/Basic/TargetInfo.h"
+#include "clang/Sema/DeclSpec.h"
 #include "clang/Sema/Initialization.h"
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
-#include "clang/Sema/SemaLambda.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/STLExtras.h"
 #include <optional>
@@ -1393,7 +1394,7 @@ void Sema::ActOnStartOfLambdaDefinition(LambdaIntroducer &Intro,
 
   // CUDA lambdas get implicit host and device attributes.
   if (getLangOpts().CUDA)
-    CUDASetLambdaAttrs(Method);
+    CUDA().SetLambdaAttrs(Method);
 
   // OpenMP lambdas might get assumumption attributes.
   if (LangOpts.OpenMP)
@@ -2136,7 +2137,7 @@ ExprResult Sema::BuildLambdaExpr(SourceLocation StartLoc, SourceLocation EndLoc,
       CaptureInits.push_back(Init.get());
 
       if (LangOpts.CUDA)
-        CUDACheckLambdaCapture(CallOperator, From);
+        CUDA().CheckLambdaCapture(CallOperator, From);
     }
 
     Class->setCaptures(Context, Captures);
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index e1155dc2d5d28..397e7681828f3 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -31,6 +31,7 @@
 #include "clang/Sema/Initialization.h"
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/Overload.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateDeduction.h"
@@ -1549,8 +1550,8 @@ static bool IsOverloadOrOverrideImpl(Sema &SemaRef, FunctionDecl *New,
     // Don't allow overloading of destructors.  (In theory we could, but it
     // would be a giant change to clang.)
     if (!isa<CXXDestructorDecl>(New)) {
-      CUDAFunctionTarget NewTarget = SemaRef.IdentifyCUDATarget(New),
-                         OldTarget = SemaRef.IdentifyCUDATarget(Old);
+      CUDAFunctionTarget NewTarget = SemaRef.CUDA().IdentifyTarget(New),
+                         OldTarget = SemaRef.CUDA().IdentifyTarget(Old);
       if (NewTarget != CUDAFunctionTarget::InvalidTarget) {
         assert((OldTarget != CUDAFunctionTarget::InvalidTarget) &&
                "Unexpected invalid target.");
@@ -7100,7 +7101,7 @@ void Sema::AddOverloadCandidate(
     // inferred for the member automatically, based on the bases and fields of
     // the class.
     if (!(Caller && Caller->isImplicit()) &&
-        !IsAllowedCUDACall(Caller, Function)) {
+        !CUDA().IsAllowedCall(Caller, Function)) {
       Candidate.Viable = false;
       Candidate.FailureKind = ovl_fail_bad_target;
       return;
@@ -7618,7 +7619,8 @@ Sema::AddMethodCandidate(CXXMethodDecl *Method, DeclAccessPair FoundDecl,
 
   // (CUDA B.1): Check for invalid calls between targets.
   if (getLangOpts().CUDA)
-    if (!IsAllowedCUDACall(getCurFunctionDecl(/*AllowLambda=*/true), Method)) {
+    if (!CUDA().IsAllowedCall(getCurFunctionDecl(/*AllowLambda=*/true),
+                              Method)) {
       Candidate.Viable = false;
       Candidate.FailureKind = ovl_fail_bad_target;
       return;
@@ -10440,7 +10442,7 @@ bool clang::isBetterOverloadCandidate(
   // If other rules cannot determine which is better, CUDA preference will be
   // used again to determine which is better.
   //
-  // TODO: Currently IdentifyCUDAPreference does not return correct values
+  // TODO: Currently IdentifyPreference does not return correct values
   // for functions called in global variable initializers due to missing
   // correct context about device/host. Therefore we can only enforce this
   // rule when there is a caller. We should enforce this rule for functions
@@ -10452,14 +10454,14 @@ bool clang::isBetterOverloadCandidate(
   if (S.getLangOpts().CUDA && Cand1.Function && Cand2.Function &&
       S.getLangOpts().GPUExcludeWrongSideOverloads) {
     if (FunctionDecl *Caller = S.getCurFunctionDecl(/*AllowLambda=*/true)) {
-      bool IsCallerImplicitHD = Sema::isCUDAImplicitHostDeviceFunction(Caller);
+      bool IsCallerImplicitHD = SemaCUDA::isImplicitHostDeviceFunction(Caller);
       bool IsCand1ImplicitHD =
-          Sema::isCUDAImplicitHostDeviceFunction(Cand1.Function);
+          SemaCUDA::isImplicitHostDeviceFunction(Cand1.Function);
       bool IsCand2ImplicitHD =
-          Sema::isCUDAImplicitHostDeviceFunction(Cand2.Function);
-      auto P1 = S.IdentifyCUDAPreference(Caller, Cand1.Function);
-      auto P2 = S.IdentifyCUDAPreference(Caller, Cand2.Function);
-      assert(P1 != Sema::CFP_Never && P2 != Sema::CFP_Never);
+          SemaCUDA::isImplicitHostDeviceFunction(Cand2.Function);
+      auto P1 = S.CUDA().IdentifyPreference(Caller, Cand1.Function);
+      auto P2 = S.CUDA().IdentifyPreference(Caller, Cand2.Function);
+      assert(P1 != SemaCUDA::CFP_Never && P2 != SemaCUDA::CFP_Never);
       // The implicit HD function may be a function in a system header which
       // is forced by pragma. In device compilation, if we prefer HD candidates
       // over wrong-sided candidates, overloading resolution may change, which
@@ -10473,8 +10475,8 @@ bool clang::isBetterOverloadCandidate(
       auto EmitThreshold =
           (S.getLangOpts().CUDAIsDevice && IsCallerImplicitHD &&
            (IsCand1ImplicitHD || IsCand2ImplicitHD))
-              ? Sema::CFP_Never
-              : Sema::CFP_WrongSide;
+              ? SemaCUDA::CFP_Never
+              : SemaCUDA::CFP_WrongSide;
       auto Cand1Emittable = P1 > EmitThreshold;
       auto Cand2Emittable = P2 > EmitThreshold;
       if (Cand1Emittable && !Cand2Emittable)
@@ -10758,8 +10760,8 @@ bool clang::isBetterOverloadCandidate(
   // to determine which is better.
   if (S.getLangOpts().CUDA && Cand1.Function && Cand2.Function) {
     FunctionDecl *Caller = S.getCurFunctionDecl(/*AllowLambda=*/true);
-    return S.IdentifyCUDAPreference(Caller, Cand1.Function) >
-           S.IdentifyCUDAPreference(Caller, Cand2.Function);
+    return S.CUDA().IdentifyPreference(Caller, Cand1.Function) >
+           S.CUDA().IdentifyPreference(Caller, Cand2.Function);
   }
 
   // General member function overloading is handled above, so this only handles
@@ -10891,15 +10893,15 @@ OverloadCandidateSet::BestViableFunction(Sema &S, SourceLocation Loc,
         llvm::any_of(Candidates, [&](OverloadCandidate *Cand) {
           // Check viable function only.
           return Cand->Viable && Cand->Function &&
-                 S.IdentifyCUDAPreference(Caller, Cand->Function) ==
-                     Sema::CFP_SameSide;
+                 S.CUDA().IdentifyPreference(Caller, Cand->Function) ==
+                     SemaCUDA::CFP_SameSide;
         });
     if (ContainsSameSideCandidate) {
       auto IsWrongSideCandidate = [&](OverloadCandidate *Cand) {
         // Check viable function only to avoid unnecessary data copying/moving.
         return Cand->Viable && Cand->Function &&
-               S.IdentifyCUDAPreference(Caller, Cand->Function) ==
-                   Sema::CFP_WrongSide;
+               S.CUDA().IdentifyPreference(Caller, Cand->Function) ==
+                   SemaCUDA::CFP_WrongSide;
       };
       llvm::erase_if(Candidates, IsWrongSideCandidate);
     }
@@ -11938,8 +11940,8 @@ static void DiagnoseBadTarget(Sema &S, OverloadCandidate *Cand) {
   FunctionDecl *Caller = S.getCurFunctionDecl(/*AllowLambda=*/true);
   FunctionDecl *Callee = Cand->Function;
 
-  CUDAFunctionTarget CallerTarget = S.IdentifyCUDATarget(Caller),
-                     CalleeTarget = S.IdentifyCUDATarget(Callee);
+  CUDAFunctionTarget CallerTarget = S.CUDA().IdentifyTarget(Caller),
+                     CalleeTarget = S.CUDA().IdentifyTarget(Callee);
 
   std::string FnDesc;
   std::pair<OverloadCandidateKind, OverloadCandidateSelect> FnKindPair =
@@ -11986,9 +11988,9 @@ static void DiagnoseBadTarget(Sema &S, OverloadCandidate *Cand) {
       }
     }
 
-    S.inferCUDATargetForImplicitSpecialMember(ParentClass, CSM, Meth,
-                                              /* ConstRHS */ ConstRHS,
-                                              /* Diagnose */ true);
+    S.CUDA().inferTargetForImplicitSpecialMember(ParentClass, CSM, Meth,
+                                                 /* ConstRHS */ ConstRHS,
+                                                 /* Diagnose */ true);
   }
 }
 
@@ -13060,7 +13062,7 @@ class AddressOfFunctionResolver {
       if (S.getLangOpts().CUDA) {
         FunctionDecl *Caller = S.getCurFunctionDecl(/*AllowLambda=*/true);
         if (!(Caller && Caller->isImplicit()) &&
-            !S.IsAllowedCUDACall(Caller, FunDecl))
+            !S.CUDA().IsAllowedCall(Caller, FunDecl))
           return false;
       }
       if (FunDecl->isMultiVersion()) {
@@ -13180,8 +13182,8 @@ class AddressOfFunctionResolver {
   }
 
   void EliminateSuboptimalCudaMatches() {
-    S.EraseUnwantedCUDAMatches(S.getCurFunctionDecl(/*AllowLambda=*/true),
-                               Matches);
+    S.CUDA().EraseUnwantedMatches(S.getCurFunctionDecl(/*AllowLambda=*/true),
+                                  Matches);
   }
 
 public:
@@ -13335,8 +13337,8 @@ Sema::resolveAddressOfSingleOverloadCandidate(Expr *E, DeclAccessPair &Pair) {
   // Return positive for better, negative for worse, 0 for equal preference.
   auto CheckCUDAPreference = [&](FunctionDecl *FD1, FunctionDecl *FD2) {
     FunctionDecl *Caller = getCurFunctionDecl(/*AllowLambda=*/true);
-    return static_cast<int>(IdentifyCUDAPreference(Caller, FD1)) -
-           static_cast<int>(IdentifyCUDAPreference(Caller, FD2));
+    return static_cast<int>(CUDA().IdentifyPreference(Caller, FD1)) -
+           static_cast<int>(CUDA().IdentifyPreference(Caller, FD2));
   };
 
   auto CheckMoreConstrained = [&](FunctionDecl *FD1,
diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
index 1c2f6120f6218..d28c24cfdfd33 100644
--- a/clang/lib/Sema/SemaStmt.cpp
+++ b/clang/lib/Sema/SemaStmt.cpp
@@ -33,6 +33,7 @@
 #include "clang/Sema/Ownership.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
@@ -4574,8 +4575,8 @@ StmtResult Sema::ActOnCXXTryBlock(SourceLocation TryLoc, Stmt *TryBlock,
 
   // Exceptions aren't allowed in CUDA device code.
   if (getLangOpts().CUDA)
-    CUDADiagIfDeviceCode(TryLoc, diag::err_cuda_device_exceptions)
-        << "try" << llvm::to_underlying(CurrentCUDATarget());
+    CUDA().DiagIfDeviceCode(TryLoc, diag::err_cuda_device_exceptions)
+        << "try" << llvm::to_underlying(CUDA().CurrentTarget());
 
   if (getCurScope() && getCurScope()->isOpenMPSimdDirectiveScope())
     Diag(TryLoc, diag::err_omp_simd_region_cannot_use_stmt) << "try";
diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
index e0f5e53dc2481..95171359f0ab1 100644
--- a/clang/lib/Sema/SemaTemplate.cpp
+++ b/clang/lib/Sema/SemaTemplate.cpp
@@ -33,6 +33,7 @@
 #include "clang/Sema/Overload.h"
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateDeduction.h"
@@ -10155,9 +10156,9 @@ bool Sema::CheckFunctionTemplateSpecialization(
       // take target attributes into account, we reject candidates
       // here that have a different target.
       if (LangOpts.CUDA &&
-          IdentifyCUDATarget(Specialization,
-                             /* IgnoreImplicitHDAttr = */ true) !=
-              IdentifyCUDATarget(FD, /* IgnoreImplicitHDAttr = */ true)) {
+          CUDA().IdentifyTarget(Specialization,
+                                /* IgnoreImplicitHDAttr = */ true) !=
+              CUDA().IdentifyTarget(FD, /* IgnoreImplicitHDAttr = */ true)) {
         FailedCandidates.addCandidate().set(
             I.getPair(), FunTmpl->getTemplatedDecl(),
             MakeDeductionFailureInfo(
@@ -10328,7 +10329,7 @@ bool Sema::CheckFunctionTemplateSpecialization(
   // virtue e.g. of being constexpr, and it passes these implicit
   // attributes on to its specializations.)
   if (LangOpts.CUDA)
-    inheritCUDATargetAttrs(FD, *Specialization->getPrimaryTemplate());
+    CUDA().inheritTargetAttrs(FD, *Specialization->getPrimaryTemplate());
 
   // The "previous declaration" for this function template specialization is
   // the prior function template specialization.
@@ -11364,9 +11365,9 @@ DeclResult Sema::ActOnExplicitInstantiation(Scope *S,
     // target attributes into account, we reject candidates here that
     // have a different target.
     if (LangOpts.CUDA &&
-        IdentifyCUDATarget(Specialization,
-                           /* IgnoreImplicitHDAttr = */ true) !=
-            IdentifyCUDATarget(D.getDeclSpec().getAttributes())) {
+        CUDA().IdentifyTarget(Specialization,
+                              /* IgnoreImplicitHDAttr = */ true) !=
+            CUDA().IdentifyTarget(D.getDeclSpec().getAttributes())) {
       FailedCandidates.addCandidate().set(
           P.getPair(), FunTmpl->getTemplatedDecl(),
           MakeDeductionFailureInfo(
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 7f510d34d671e..c0469a47ab8b6 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -26,6 +26,7 @@
 #include "clang/Sema/Initialization.h"
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateInstCallback.h"
@@ -5537,7 +5538,7 @@ void Sema::InstantiateVariableInitializer(
   }
 
   if (getLangOpts().CUDA)
-    checkAllowedCUDAInitializer(Var);
+    CUDA().checkAllowedInitializer(Var);
 }
 
 /// Instantiate the definition of the given variable from its
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index b3f6078952f6e..404c4e8e31b55 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -33,6 +33,7 @@
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateInstCallback.h"
@@ -2735,7 +2736,7 @@ QualType Sema::BuildArrayType(QualType T, ArraySizeModifier ASM,
       bool IsCUDADevice = (getLangOpts().CUDA && getLangOpts().CUDAIsDevice);
       targetDiag(Loc,
                  IsCUDADevice ? diag::err_cuda_vla : diag::err_vla_unsupported)
-          << (IsCUDADevice ? llvm::to_underlying(CurrentCUDATarget()) : 0);
+          << (IsCUDADevice ? llvm::to_underlying(CUDA().CurrentTarget()) : 0);
     } else if (sema::FunctionScopeInfo *FSI = getCurFunction()) {
       // VLAs are supported on this target, but we may need to do delayed
       // checking that the VLA is not being used within a coroutine.
@@ -3618,7 +3619,7 @@ static QualType GetDeclSpecTypeForDeclarator(TypeProcessingState &state,
   // D.getDeclarationAttributes()) because those are always C++11 attributes,
   // and those don't get distributed.
   distributeTypeAttrsFromDeclarator(
-      state, T, SemaRef.IdentifyCUDATarget(D.getAttributes()));
+      state, T, SemaRef.CUDA().IdentifyTarget(D.getAttributes()));
 
   // Find the deduced type in this type. Look in the trailing return type if we
   // have one, otherwise in the DeclSpec type.
@@ -4139,7 +4140,7 @@ static CallingConv getCCForDeclaratorChunk(
       // handleFunctionTypeAttr.
       CallingConv CC;
       if (!S.CheckCallingConvAttr(AL, CC, /*FunctionDecl=*/nullptr,
-                                  S.IdentifyCUDATarget(D.getAttributes())) &&
+                                  S.CUDA().IdentifyTarget(D.getAttributes())) &&
           (!FTI.isVariadic || supportsVariadicCall(CC))) {
         return CC;
       }
@@ -5825,7 +5826,7 @@ static TypeSourceInfo *GetFullTypeForDeclarator(TypeProcessingState &state,
 
     // See if there are any attributes on this declarator chunk.
     processTypeAttrs(state, T, TAL_DeclChunk, DeclType.getAttrs(),
-                     S.IdentifyCUDATarget(D.getAttributes()));
+                     S.CUDA().IdentifyTarget(D.getAttributes()));
 
     if (DeclType.Kind != DeclaratorChunk::Paren) {
       if (ExpectNoDerefChunk && !IsNoDerefableChunk(DeclType))
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index a47c7a393d632..8c4b460970ad2 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -78,6 +78,7 @@
 #include "clang/Sema/ObjCMethodList.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/Sema.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/Weak.h"
 #include "clang/Serialization/ASTBitCodes.h"
 #include "clang/Serialization/ASTDeserializationListener.h"
@@ -3995,7 +3996,7 @@ llvm::Error ASTReader::ReadASTBlock(ModuleFile &F,
       if (Record.size() != 1)
         return llvm::createStringError(std::errc::illegal_byte_sequence,
                                        "invalid cuda pragma options record");
-      ForceCUDAHostDeviceDepth = Record[0];
+      ForceHostDeviceDepth = Record[0];
       break;
 
     case ALIGN_PACK_PRAGMA_OPTIONS: {
@@ -8274,7 +8275,7 @@ void ASTReader::UpdateSema() {
             PragmaMSPointersToMembersState,
         PointersToMembersPragmaLocation);
   }
-  SemaObj->ForceCUDAHostDeviceDepth = ForceCUDAHostDeviceDepth;
+  SemaObj->CUDA().ForceHostDeviceDepth = ForceHostDeviceDepth;
 
   if (PragmaAlignPackCurrentValue) {
     // The bottom of the stack might have a default value. It must be adjusted
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index d45695784354a..85b7fd5535a1b 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -65,6 +65,7 @@
 #include "clang/Sema/IdentifierResolver.h"
 #include "clang/Sema/ObjCMethodList.h"
 #include "clang/Sema/Sema.h"
+#include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/Weak.h"
 #include "clang/Serialization/ASTBitCodes.h"
 #include "clang/Serialization/ASTReader.h"
@@ -4335,8 +4336,8 @@ void ASTWriter::WriteOpenCLExtensions(Sema &SemaRef) {
   Stream.EmitRecord(OPENCL_EXTENSIONS, Record);
 }
 void ASTWriter::WriteCUDAPragmas(Sema &SemaRef) {
-  if (SemaRef.ForceCUDAHostDeviceDepth > 0) {
-    RecordData::value_type Record[] = {SemaRef.ForceCUDAHostDeviceDepth};
+  if (SemaRef.CUDA().ForceHostDeviceDepth > 0) {
+    RecordData::value_type Record[] = {SemaRef.CUDA().ForceHostDeviceDepth};
     Stream.EmitRecord(CUDA_PRAGMA_FORCE_HOST_DEVICE_DEPTH, Record);
   }
 }

From 6dcb60481323c871556a5a14ed887a5867ece114 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 08:55:09 +0400
Subject: [PATCH 320/886] [clang] Implement
 `__is_pointer_interconvertible_base_of()` (#88473)

This patch implements intrinsic that supports
`std::is_pointer_interconvertible_base_of` type trait from
[P0466R5](https://wg21.link/p0466r5) "Layout-compatibility and
Pointer-interconvertibility Traits".

Normative wording:
> Comment: If `Base` and Derived are non-union class types and are not
(possibly _cv_-qualified) versions of the same type, `Derived` is a
complete type.
> Condition: `Derived` is unambiguously derived from `Base` without
regard to _cv_-qualifiers, and each object of type `Derived` is
pointer-interconvertible (6.7.2 [basic.compound]) with its `Base`
subobject, or `Base` and `Derived` are not unions and name the same
class type without regard to _cv_-qualifiers.

The paper also express the following intent:
> Note that `is_pointer_interconvertible_base_of_v<T,T>` is always true
under this wording, even though `T` is not derived from itself.

I find the treatment of unions in the wording contradictory to this
intent, and I'm not able to find anything relevant in minutes or on the
reflector. That said, this patch implements what the wording says, since
it's very explicit about unions.
---
 clang/docs/ReleaseNotes.rst              |  3 +-
 clang/include/clang/Basic/TokenKinds.def |  1 +
 clang/include/clang/Sema/Sema.h          |  2 +
 clang/lib/Sema/SemaChecking.cpp          | 21 ++++++
 clang/lib/Sema/SemaExprCXX.cpp           | 16 +++++
 clang/test/SemaCXX/type-traits.cpp       | 91 +++++++++++++++++++++++-
 6 files changed, 132 insertions(+), 2 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 45a9a79739a4e..399d5ba796d15 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -98,7 +98,8 @@ C++20 Feature Support
   behavior can use the flag '-Xclang -fno-skip-odr-check-in-gmf'.
   (#GH79240).
 
-- Implemented the `__is_layout_compatible` intrinsic to support
+- Implemented the `__is_layout_compatible` and `__is_pointer_interconvertible_base_of`
+  intrinsics to support
   `P0466R5: Layout-compatibility and Pointer-interconvertibility Traits <https://wg21.link/P0466R5>`_.
 
 - Clang now implements [module.import]p7 fully. Clang now will import module
diff --git a/clang/include/clang/Basic/TokenKinds.def b/clang/include/clang/Basic/TokenKinds.def
index 800af0e6d0448..a27fbed358a60 100644
--- a/clang/include/clang/Basic/TokenKinds.def
+++ b/clang/include/clang/Basic/TokenKinds.def
@@ -521,6 +521,7 @@ TYPE_TRAIT_1(__is_union, IsUnion, KEYCXX)
 TYPE_TRAIT_1(__has_unique_object_representations,
              HasUniqueObjectRepresentations, KEYCXX)
 TYPE_TRAIT_2(__is_layout_compatible, IsLayoutCompatible, KEYCXX)
+TYPE_TRAIT_2(__is_pointer_interconvertible_base_of, IsPointerInterconvertibleBaseOf, KEYCXX)
 
 #define TRANSFORM_TYPE_TRAIT_DEF(_, Trait) KEYWORD(__##Trait, KEYCXX)
 #include "clang/Basic/TransformTypeTraits.def"
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 6b9789334811e..3b252f94dcbee 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -1999,6 +1999,8 @@ class Sema final : public SemaBase {
   };
 
   bool IsLayoutCompatible(QualType T1, QualType T2) const;
+  bool IsPointerInterconvertibleBaseOf(const TypeSourceInfo *Base,
+                                       const TypeSourceInfo *Derived);
 
   bool CheckFunctionCall(FunctionDecl *FDecl, CallExpr *TheCall,
                          const FunctionProtoType *Proto);
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index abfd9a3031577..8e21811b67d90 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -19710,6 +19710,27 @@ bool Sema::IsLayoutCompatible(QualType T1, QualType T2) const {
   return isLayoutCompatible(getASTContext(), T1, T2);
 }
 
+//===-------------- Pointer interconvertibility ----------------------------//
+
+bool Sema::IsPointerInterconvertibleBaseOf(const TypeSourceInfo *Base,
+                                           const TypeSourceInfo *Derived) {
+  QualType BaseT = Base->getType()->getCanonicalTypeUnqualified();
+  QualType DerivedT = Derived->getType()->getCanonicalTypeUnqualified();
+
+  if (BaseT->isStructureOrClassType() && DerivedT->isStructureOrClassType() &&
+      getASTContext().hasSameType(BaseT, DerivedT))
+    return true;
+
+  if (!IsDerivedFrom(Derived->getTypeLoc().getBeginLoc(), DerivedT, BaseT))
+    return false;
+
+  // Per [basic.compound]/4.3, containing object has to be standard-layout.
+  if (DerivedT->getAsCXXRecordDecl()->isStandardLayout())
+    return true;
+
+  return false;
+}
+
 //===--- CHECK: pointer_with_type_tag attribute: datatypes should match ----//
 
 /// Given a type tag expression find the type tag itself.
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 8911257a6f614..1cfd3e56a583e 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -6100,6 +6100,22 @@ static bool EvaluateBinaryTypeTrait(Sema &Self, TypeTrait BTT, const TypeSourceI
       Self.Diag(Rhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
           << 1 << tok::kw___is_layout_compatible;
     return Self.IsLayoutCompatible(LhsT, RhsT);
+  }
+  case BTT_IsPointerInterconvertibleBaseOf: {
+    if (LhsT->isStructureOrClassType() && RhsT->isStructureOrClassType() &&
+        !Self.getASTContext().hasSameUnqualifiedType(LhsT, RhsT)) {
+      Self.RequireCompleteType(Rhs->getTypeLoc().getBeginLoc(), RhsT,
+                               diag::err_incomplete_type);
+    }
+
+    if (LhsT->isVariableArrayType())
+      Self.Diag(Lhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
+          << 1 << tok::kw___is_pointer_interconvertible_base_of;
+    if (RhsT->isVariableArrayType())
+      Self.Diag(Rhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
+          << 1 << tok::kw___is_pointer_interconvertible_base_of;
+
+    return Self.IsPointerInterconvertibleBaseOf(Lhs, Rhs);
   }
     default: llvm_unreachable("not a BTT");
   }
diff --git a/clang/test/SemaCXX/type-traits.cpp b/clang/test/SemaCXX/type-traits.cpp
index 421d3007d27ff..d43701c3d976e 100644
--- a/clang/test/SemaCXX/type-traits.cpp
+++ b/clang/test/SemaCXX/type-traits.cpp
@@ -1740,7 +1740,7 @@ void is_layout_compatible(int n)
   static_assert(!__is_layout_compatible(void, int));
   static_assert(__is_layout_compatible(void, const void));
   static_assert(__is_layout_compatible(void, volatile void));
-  static_assert(__is_layout_compatible(const int, volatile int));
+  static_assert(__is_layout_compatible(const void, volatile void));
   static_assert(__is_layout_compatible(int, int));
   static_assert(__is_layout_compatible(int, const int));
   static_assert(__is_layout_compatible(int, volatile int));
@@ -1839,6 +1839,95 @@ void is_layout_compatible(int n)
   static_assert(!__is_layout_compatible(EnumClassForward, int));
 }
 
+namespace IPIBO {
+struct Base {};
+struct Base2 {};
+struct Base3 : Base {};
+struct Base3Virtual : virtual Base {};
+struct Derived : Base {};
+struct DerivedIndirect : Base3 {};
+struct DerivedMultiple : Base, Base2 {};
+struct DerivedAmbiguous : Base, Base3 {};
+/* expected-warning@-1 {{direct base 'Base' is inaccessible due to ambiguity:
+    struct IPIBO::DerivedAmbiguous -> Base
+    struct IPIBO::DerivedAmbiguous -> Base3 -> Base}} */
+struct DerivedPrivate : private Base {};
+struct DerivedVirtual : virtual Base {};
+
+union Union {};
+union UnionIncomplete;
+struct StructIncomplete; // #StructIncomplete
+
+void is_pointer_interconvertible_base_of(int n)
+{
+  static_assert(__is_pointer_interconvertible_base_of(Base, Derived));
+  static_assert(!__is_pointer_interconvertible_base_of(Base2, Derived));
+  static_assert(__is_pointer_interconvertible_base_of(Base, DerivedIndirect));
+  static_assert(__is_pointer_interconvertible_base_of(Base, DerivedMultiple));
+  static_assert(!__is_pointer_interconvertible_base_of(Base3, DerivedMultiple));
+  static_assert(!__is_pointer_interconvertible_base_of(Base, DerivedAmbiguous));
+  static_assert(__is_pointer_interconvertible_base_of(Base, DerivedPrivate));
+  static_assert(!__is_pointer_interconvertible_base_of(Base, DerivedVirtual));
+  static_assert(!__is_pointer_interconvertible_base_of(Union, Union));
+  static_assert(!__is_pointer_interconvertible_base_of(UnionIncomplete, UnionIncomplete));
+  static_assert(__is_pointer_interconvertible_base_of(StructIncomplete, StructIncomplete));
+  static_assert(__is_pointer_interconvertible_base_of(StructIncomplete, const StructIncomplete));
+  static_assert(__is_pointer_interconvertible_base_of(StructIncomplete, volatile StructIncomplete));
+  static_assert(__is_pointer_interconvertible_base_of(const StructIncomplete, volatile StructIncomplete));
+  static_assert(!__is_pointer_interconvertible_base_of(StructIncomplete, Derived));
+  static_assert(!__is_pointer_interconvertible_base_of(Base, StructIncomplete));
+  // expected-error@-1 {{incomplete type 'StructIncomplete' where a complete type is required}}
+  //   expected-note@#StructIncomplete {{forward declaration of 'IPIBO::StructIncomplete'}}
+  static_assert(!__is_pointer_interconvertible_base_of(CStruct2, CppStructNonStandardByBase2));
+  static_assert(!__is_pointer_interconvertible_base_of(void, void));
+  static_assert(!__is_pointer_interconvertible_base_of(void, int));
+  static_assert(!__is_pointer_interconvertible_base_of(void, const void));
+  static_assert(!__is_pointer_interconvertible_base_of(void, volatile void));
+  static_assert(!__is_pointer_interconvertible_base_of(const void, volatile void));
+  static_assert(!__is_pointer_interconvertible_base_of(int, int));
+  static_assert(!__is_pointer_interconvertible_base_of(int, const int));
+  static_assert(!__is_pointer_interconvertible_base_of(int, volatile int));
+  static_assert(!__is_pointer_interconvertible_base_of(const int, volatile int));
+  static_assert(!__is_pointer_interconvertible_base_of(int *, int * __restrict));
+  static_assert(!__is_pointer_interconvertible_base_of(int, _Atomic int));
+  static_assert(!__is_pointer_interconvertible_base_of(_Atomic(int), _Atomic int));
+  static_assert(!__is_pointer_interconvertible_base_of(int, unsigned int));
+  static_assert(!__is_pointer_interconvertible_base_of(char, unsigned char));
+  static_assert(!__is_pointer_interconvertible_base_of(char, signed char));
+  static_assert(!__is_pointer_interconvertible_base_of(unsigned char, signed char));
+  using function_type = void();
+  using function_type2 = void(char);
+  static_assert(!__is_pointer_interconvertible_base_of(const function_type, const function_type));
+  // expected-warning@-1 {{'const' qualifier on function type 'function_type' (aka 'void ()') has no effect}}
+  // expected-warning@-2 {{'const' qualifier on function type 'function_type' (aka 'void ()') has no effect}}
+  static_assert(!__is_pointer_interconvertible_base_of(function_type, const function_type));
+  // expected-warning@-1 {{'const' qualifier on function type 'function_type' (aka 'void ()') has no effect}}
+  static_assert(!__is_pointer_interconvertible_base_of(const function_type, const function_type2));
+  // expected-warning@-1 {{'const' qualifier on function type 'function_type' (aka 'void ()') has no effect}}
+  // expected-warning@-2 {{'const' qualifier on function type 'function_type2' (aka 'void (char)') has no effect}}
+  static_assert(!__is_pointer_interconvertible_base_of(int CStruct2::*, int CStruct2::*));
+  static_assert(!__is_pointer_interconvertible_base_of(int CStruct2::*, char CStruct2::*));
+  static_assert(!__is_pointer_interconvertible_base_of(void(CStruct2::*)(int), void(CStruct2::*)(int)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(CStruct2::*)(int), void(CStruct2::*)(char)));
+  static_assert(!__is_pointer_interconvertible_base_of(int[], int[]));
+  static_assert(!__is_pointer_interconvertible_base_of(int[], double[]));
+  static_assert(!__is_pointer_interconvertible_base_of(int[2], int[2]));
+  static_assert(!__is_pointer_interconvertible_base_of(int[n], int[2]));
+  // expected-error@-1 {{variable length arrays are not supported in '__is_pointer_interconvertible_base_of'}}
+  static_assert(!__is_pointer_interconvertible_base_of(int[n], int[n]));
+  // expected-error@-1 {{variable length arrays are not supported in '__is_pointer_interconvertible_base_of'}}
+  // expected-error@-2 {{variable length arrays are not supported in '__is_pointer_interconvertible_base_of'}}
+  static_assert(!__is_pointer_interconvertible_base_of(int&, int&));
+  static_assert(!__is_pointer_interconvertible_base_of(int&, char&));
+  static_assert(!__is_pointer_interconvertible_base_of(void(int), void(int)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(int), void(char)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(&)(int), void(&)(int)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(&)(int), void(&)(char)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(*)(int), void(*)(int)));
+  static_assert(!__is_pointer_interconvertible_base_of(void(*)(int), void(*)(char)));
+}
+}
+
 void is_signed()
 {
   //static_assert(__is_signed(char));

From 79dca25f4a0fefd47e9e37d9ce47d84dc0b3bedb Mon Sep 17 00:00:00 2001
From: Cyndy Ishida <cyndy_ishida@apple.com>
Date: Fri, 12 Apr 2024 22:05:12 -0700
Subject: [PATCH 321/886] [InstallAPI] Replace std::string -> StringRef

---
 clang/tools/clang-installapi/Options.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/tools/clang-installapi/Options.cpp b/clang/tools/clang-installapi/Options.cpp
index 120ff3da89986..3dc61476ce09d 100644
--- a/clang/tools/clang-installapi/Options.cpp
+++ b/clang/tools/clang-installapi/Options.cpp
@@ -372,7 +372,7 @@ bool Options::addFilePaths(InputArgList &Args, PathSeq &Headers,
       }
       // Sort headers to ensure deterministic behavior.
       sort(*InputHeadersOrErr);
-      for (std::string &H : *InputHeadersOrErr)
+      for (StringRef H : *InputHeadersOrErr)
         Headers.emplace_back(std::move(H));
     } else
       Headers.emplace_back(Path);
@@ -690,7 +690,7 @@ InstallAPIContext Options::createContext() {
     FrameworkName = getFrameworkNameFromInstallName(LinkerOpts.InstallName);
 
   // Process inputs.
-  for (const std::string &ListPath : DriverOpts.FileLists) {
+  for (const StringRef ListPath : DriverOpts.FileLists) {
     auto Buffer = FM->getBufferForFile(ListPath);
     if (auto Err = Buffer.getError()) {
       Diags->Report(diag::err_cannot_open_file) << ListPath << Err.message();

From ed128c7df9b4e60bfd814dc9fd22de1dde4a2c1c Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 08:28:02 +0300
Subject: [PATCH 322/886] [clang][NFC] Replace `dr` with `cwg` in DR test suite

This patch is breaking `dr` to `cwg` equivalence in our terminology, making room for tests for LWG issues that concern compiler intrinsics.
---
 clang/test/CXX/drs/{dr118.cpp => cwg118.cpp}  |   2 +-
 clang/test/CXX/drs/{dr124.cpp => cwg124.cpp}  |  18 +-
 clang/test/CXX/drs/{dr158.cpp => cwg158.cpp}  |   2 +-
 .../test/CXX/drs/{dr1748.cpp => cwg1748.cpp}  |   2 +-
 .../test/CXX/drs/{dr177x.cpp => cwg177x.cpp}  |   8 +-
 .../test/CXX/drs/{dr1807.cpp => cwg1807.cpp}  |  10 +-
 clang/test/CXX/drs/{dr185.cpp => cwg185.cpp}  |  12 +-
 clang/test/CXX/drs/{dr193.cpp => cwg193.cpp}  |  18 +-
 clang/test/CXX/drs/{dr199.cpp => cwg199.cpp}  |  12 +-
 clang/test/CXX/drs/{dr201.cpp => cwg201.cpp}  |  16 +-
 clang/test/CXX/drs/{dr210.cpp => cwg210.cpp}  |   8 +-
 .../test/CXX/drs/{dr2335.cpp => cwg2335.cpp}  |  26 +-
 .../test/CXX/drs/{dr2390.cpp => cwg2390.cpp}  |   2 +-
 .../test/CXX/drs/{dr2504.cpp => cwg2504.cpp}  |   8 +-
 clang/test/CXX/drs/{dr292.cpp => cwg292.cpp}  |   8 +-
 clang/test/CXX/drs/{dr392.cpp => cwg392.cpp}  |  10 +-
 clang/test/CXX/drs/{dr412.cpp => cwg412.cpp}  |   2 +-
 clang/test/CXX/drs/{dr438.cpp => cwg438.cpp}  |   6 +-
 clang/test/CXX/drs/{dr439.cpp => cwg439.cpp}  |   6 +-
 clang/test/CXX/drs/{dr441.cpp => cwg441.cpp}  |  18 +-
 clang/test/CXX/drs/{dr462.cpp => cwg462.cpp}  |  10 +-
 clang/test/CXX/drs/{dr492.cpp => cwg492.cpp}  |   6 +-
 clang/test/CXX/drs/{dr519.cpp => cwg519.cpp}  |   6 +-
 clang/test/CXX/drs/{dr571.cpp => cwg571.cpp}  |   6 +-
 clang/test/CXX/drs/{dr605.cpp => cwg605.cpp}  |   6 +-
 clang/test/CXX/drs/{dr650.cpp => cwg650.cpp}  |  12 +-
 clang/test/CXX/drs/{dr653.cpp => cwg653.cpp}  |   6 +-
 clang/test/CXX/drs/{dr658.cpp => cwg658.cpp}  |   6 +-
 clang/test/CXX/drs/{dr661.cpp => cwg661.cpp}  |   8 +-
 clang/test/CXX/drs/{dr672.cpp => cwg672.cpp}  |   8 +-
 clang/test/CXX/drs/{dr593.cpp => cwgr593.cpp} |   6 +-
 clang/test/CXX/drs/dr0xx.cpp                  | 842 +++++++++---------
 clang/test/CXX/drs/dr10xx.cpp                 |  36 +-
 clang/test/CXX/drs/dr11xx.cpp                 |  24 +-
 clang/test/CXX/drs/dr12xx.cpp                 |  34 +-
 clang/test/CXX/drs/dr13xx.cpp                 | 234 ++---
 clang/test/CXX/drs/dr14xx.cpp                 | 238 ++---
 clang/test/CXX/drs/dr15xx.cpp                 | 370 ++++----
 clang/test/CXX/drs/dr16xx.cpp                 | 290 +++---
 clang/test/CXX/drs/dr17xx.cpp                 |  64 +-
 clang/test/CXX/drs/dr18xx.cpp                 | 100 +--
 clang/test/CXX/drs/dr19xx.cpp                 |  68 +-
 clang/test/CXX/drs/dr1xx.cpp                  | 614 ++++++-------
 clang/test/CXX/drs/dr20xx.cpp                 | 152 ++--
 clang/test/CXX/drs/dr21xx.cpp                 | 104 +--
 clang/test/CXX/drs/dr22xx.cpp                 |  46 +-
 clang/test/CXX/drs/dr23xx.cpp                 |  80 +-
 clang/test/CXX/drs/dr24xx.cpp                 |   8 +-
 clang/test/CXX/drs/dr25xx.cpp                 |  66 +-
 clang/test/CXX/drs/dr26xx.cpp                 |  76 +-
 clang/test/CXX/drs/dr27xx.cpp                 |  18 +-
 clang/test/CXX/drs/dr28xx.cpp                 |   8 +-
 clang/test/CXX/drs/dr2xx.cpp                  | 510 +++++------
 clang/test/CXX/drs/dr3xx.cpp                  | 774 ++++++++--------
 clang/test/CXX/drs/dr4xx.cpp                  | 558 ++++++------
 clang/test/CXX/drs/dr5xx.cpp                  | 444 ++++-----
 clang/test/CXX/drs/dr6xx.cpp                  | 446 +++++-----
 clang/test/CXX/drs/dr7xx.cpp                  | 120 +--
 clang/test/CXX/drs/dr8xx.cpp                  |  12 +-
 clang/test/CXX/drs/dr9xx.cpp                  |  62 +-
 clang/www/make_cxx_dr_status                  |   6 +-
 61 files changed, 3339 insertions(+), 3339 deletions(-)
 rename clang/test/CXX/drs/{dr118.cpp => cwg118.cpp} (98%)
 rename clang/test/CXX/drs/{dr124.cpp => cwg124.cpp} (83%)
 rename clang/test/CXX/drs/{dr158.cpp => cwg158.cpp} (98%)
 rename clang/test/CXX/drs/{dr1748.cpp => cwg1748.cpp} (98%)
 rename clang/test/CXX/drs/{dr177x.cpp => cwg177x.cpp} (95%)
 rename clang/test/CXX/drs/{dr1807.cpp => cwg1807.cpp} (83%)
 rename clang/test/CXX/drs/{dr185.cpp => cwg185.cpp} (85%)
 rename clang/test/CXX/drs/{dr193.cpp => cwg193.cpp} (79%)
 rename clang/test/CXX/drs/{dr199.cpp => cwg199.cpp} (85%)
 rename clang/test/CXX/drs/{dr201.cpp => cwg201.cpp} (82%)
 rename clang/test/CXX/drs/{dr210.cpp => cwg210.cpp} (91%)
 rename clang/test/CXX/drs/{dr2335.cpp => cwg2335.cpp} (56%)
 rename clang/test/CXX/drs/{dr2390.cpp => cwg2390.cpp} (98%)
 rename clang/test/CXX/drs/{dr2504.cpp => cwg2504.cpp} (90%)
 rename clang/test/CXX/drs/{dr292.cpp => cwg292.cpp} (91%)
 rename clang/test/CXX/drs/{dr392.cpp => cwg392.cpp} (88%)
 rename clang/test/CXX/drs/{dr412.cpp => cwg412.cpp} (99%)
 rename clang/test/CXX/drs/{dr438.cpp => cwg438.cpp} (94%)
 rename clang/test/CXX/drs/{dr439.cpp => cwg439.cpp} (94%)
 rename clang/test/CXX/drs/{dr441.cpp => cwg441.cpp} (76%)
 rename clang/test/CXX/drs/{dr462.cpp => cwg462.cpp} (87%)
 rename clang/test/CXX/drs/{dr492.cpp => cwg492.cpp} (94%)
 rename clang/test/CXX/drs/{dr519.cpp => cwg519.cpp} (95%)
 rename clang/test/CXX/drs/{dr571.cpp => cwg571.cpp} (91%)
 rename clang/test/CXX/drs/{dr605.cpp => cwg605.cpp} (91%)
 rename clang/test/CXX/drs/{dr650.cpp => cwg650.cpp} (86%)
 rename clang/test/CXX/drs/{dr653.cpp => cwg653.cpp} (92%)
 rename clang/test/CXX/drs/{dr658.cpp => cwg658.cpp} (92%)
 rename clang/test/CXX/drs/{dr661.cpp => cwg661.cpp} (91%)
 rename clang/test/CXX/drs/{dr672.cpp => cwg672.cpp} (91%)
 rename clang/test/CXX/drs/{dr593.cpp => cwgr593.cpp} (93%)

diff --git a/clang/test/CXX/drs/dr118.cpp b/clang/test/CXX/drs/cwg118.cpp
similarity index 98%
rename from clang/test/CXX/drs/dr118.cpp
rename to clang/test/CXX/drs/cwg118.cpp
index 58aa3912c8010..04e19ce050788 100644
--- a/clang/test/CXX/drs/dr118.cpp
+++ b/clang/test/CXX/drs/cwg118.cpp
@@ -3,7 +3,7 @@
 // RUN: %clang_cc1 -triple x86_64-linux -std=c++14 %s -pedantic-errors -emit-llvm -o - | FileCheck %s --implicit-check-not " call "
 // RUN: %clang_cc1 -triple x86_64-linux -std=c++1z %s -pedantic-errors -emit-llvm -o - | FileCheck %s --implicit-check-not " call "
 
-// dr118: yes
+// cwg118: yes
 
 struct S {
   virtual void f();
diff --git a/clang/test/CXX/drs/dr124.cpp b/clang/test/CXX/drs/cwg124.cpp
similarity index 83%
rename from clang/test/CXX/drs/dr124.cpp
rename to clang/test/CXX/drs/cwg124.cpp
index c07beb11709c7..fef3c6085c375 100644
--- a/clang/test/CXX/drs/dr124.cpp
+++ b/clang/test/CXX/drs/cwg124.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr124 { // dr124: 2.7
+namespace cwg124 { // cwg124: 2.7
 
 extern void full_expr_fence() NOTHROW;
 
@@ -32,20 +32,20 @@ void f() {
   full_expr_fence();
 }
 
-// CHECK-LABEL: define {{.*}} void @dr124::f()()
-// CHECK:         call void @dr124::full_expr_fence()
+// CHECK-LABEL: define {{.*}} void @cwg124::f()()
+// CHECK:         call void @cwg124::full_expr_fence()
 // CHECK:         br label %arrayctor.loop
 // CHECK-LABEL: arrayctor.loop:
-// CHECK:         call void @dr124::A::A()
-// CHECK:         call void @dr124::B::B(dr124::A)
-// CHECK:         call void @dr124::A::~A()
+// CHECK:         call void @cwg124::A::A()
+// CHECK:         call void @cwg124::B::B(cwg124::A)
+// CHECK:         call void @cwg124::A::~A()
 // CHECK:         br {{.*}}, label %arrayctor.cont, label %arrayctor.loop
 // CHECK-LABEL: arrayctor.cont:
-// CHECK:         call void @dr124::full_expr_fence()
+// CHECK:         call void @cwg124::full_expr_fence()
 // CHECK:         br label %arraydestroy.body
 // CHECK-LABEL: arraydestroy.body:
-// CHECK:         call void @dr124::B::~B()
+// CHECK:         call void @cwg124::B::~B()
 // CHECK-LABEL: }
 
 
-} // namespace dr124
+} // namespace cwg124
diff --git a/clang/test/CXX/drs/dr158.cpp b/clang/test/CXX/drs/cwg158.cpp
similarity index 98%
rename from clang/test/CXX/drs/dr158.cpp
rename to clang/test/CXX/drs/cwg158.cpp
index a0a8bd05baee3..9301c790297e9 100644
--- a/clang/test/CXX/drs/dr158.cpp
+++ b/clang/test/CXX/drs/cwg158.cpp
@@ -3,7 +3,7 @@
 // RUN: %clang_cc1 -triple x86_64-linux -std=c++14 %s -O3 -disable-llvm-passes -pedantic-errors -emit-llvm -o - | FileCheck %s
 // RUN: %clang_cc1 -triple x86_64-linux -std=c++1z %s -O3 -disable-llvm-passes -pedantic-errors -emit-llvm -o - | FileCheck %s
 
-// dr158: yes
+// cwg158: yes
 
 // CHECK-LABEL: define {{.*}} @_Z1f
 const int *f(const int * const *p, int **q) {
diff --git a/clang/test/CXX/drs/dr1748.cpp b/clang/test/CXX/drs/cwg1748.cpp
similarity index 98%
rename from clang/test/CXX/drs/dr1748.cpp
rename to clang/test/CXX/drs/cwg1748.cpp
index 7e04f402d2667..f216963d69f2a 100644
--- a/clang/test/CXX/drs/dr1748.cpp
+++ b/clang/test/CXX/drs/cwg1748.cpp
@@ -3,7 +3,7 @@
 // RUN: %clang_cc1 -std=c++14 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | FileCheck %s
 // RUN: %clang_cc1 -std=c++1z %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | FileCheck %s
 
-// dr1748: 3.7
+// cwg1748: 3.7
 
 // FIXME: __SIZE_TYPE__ expands to 'long long' on some targets.
 __extension__ typedef __SIZE_TYPE__ size_t;
diff --git a/clang/test/CXX/drs/dr177x.cpp b/clang/test/CXX/drs/cwg177x.cpp
similarity index 95%
rename from clang/test/CXX/drs/dr177x.cpp
rename to clang/test/CXX/drs/cwg177x.cpp
index 7b96ff0996c72..cc62bdac4cf06 100644
--- a/clang/test/CXX/drs/dr177x.cpp
+++ b/clang/test/CXX/drs/cwg177x.cpp
@@ -4,9 +4,9 @@
 // RUN: %clang_cc1 -std=c++1z %s -fexceptions -fcxx-exceptions -pedantic-errors -ast-dump | FileCheck %s --check-prefixes=CHECK,CXX11,CXX14
 // RUN: %clang_cc1 -std=c++1z %s -fexceptions -fcxx-exceptions -pedantic-errors -triple i386-windows-pc -ast-dump | FileCheck %s --check-prefixes=CHECK,CXX11,CXX14
 
-namespace dr1772 { // dr1772: 14
+namespace cwg1772 { // cwg1772: 14
   // __func__ in a lambda should name operator(), not the containing function.
-  // CHECK: NamespaceDecl{{.+}}dr1772
+  // CHECK: NamespaceDecl{{.+}}cwg1772
 #if __cplusplus >= 201103L
   auto x = []() { __func__; };
   // CXX11: LambdaExpr
@@ -30,10 +30,10 @@ namespace dr1772 { // dr1772: 14
 #endif // __cplusplus >= 201103L
 }
 
-namespace dr1779 { // dr1779: 14
+namespace cwg1779 { // cwg1779: 14
   // __func__ in a function template, member function template, or generic
   //  lambda should have a dependent type.
-  // CHECK: NamespaceDecl{{.+}}dr1779
+  // CHECK: NamespaceDecl{{.+}}cwg1779
 
   template<typename T>
   void FuncTemplate() {
diff --git a/clang/test/CXX/drs/dr1807.cpp b/clang/test/CXX/drs/cwg1807.cpp
similarity index 83%
rename from clang/test/CXX/drs/dr1807.cpp
rename to clang/test/CXX/drs/cwg1807.cpp
index 81e1e8ca640e3..59edacc49658c 100644
--- a/clang/test/CXX/drs/dr1807.cpp
+++ b/clang/test/CXX/drs/cwg1807.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK,SINCE-CXX11
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK,SINCE-CXX11
 
-namespace dr1807 { // dr1807: 3.0
+namespace cwg1807 { // cwg1807: 3.0
 struct S {
   S() {}
   ~S() {}
@@ -17,12 +17,12 @@ void f() {
 }
 }
 
-// CHECK-LABEL:      define dso_local void @dr1807::f()
-// CHECK:              invoke void @dr1807::S::S(){{.+}}
+// CHECK-LABEL:      define dso_local void @cwg1807::f()
+// CHECK:              invoke void @cwg1807::S::S(){{.+}}
 // CHECK-NEXT:         {{.+}} unwind label %lpad
 // CHECK-LABEL:      lpad:
 // CHECK:              br {{.+}}, label {{.+}}, label %arraydestroy.body
 // CHECK-LABEL:      arraydestroy.body:         
 // CHECK:              [[ARRAYDESTROY_ELEMENT:%.*]] = getelementptr {{.+}}, i64 -1
-// CXX98-NEXT:         invoke void @dr1807::S::~S()({{.*}}[[ARRAYDESTROY_ELEMENT]])
-// SINCE-CXX11-NEXT:   call void @dr1807::S::~S()({{.*}}[[ARRAYDESTROY_ELEMENT]])
+// CXX98-NEXT:         invoke void @cwg1807::S::~S()({{.*}}[[ARRAYDESTROY_ELEMENT]])
+// SINCE-CXX11-NEXT:   call void @cwg1807::S::~S()({{.*}}[[ARRAYDESTROY_ELEMENT]])
diff --git a/clang/test/CXX/drs/dr185.cpp b/clang/test/CXX/drs/cwg185.cpp
similarity index 85%
rename from clang/test/CXX/drs/dr185.cpp
rename to clang/test/CXX/drs/cwg185.cpp
index aff00f1a8764a..8ab5bc5d28f82 100644
--- a/clang/test/CXX/drs/dr185.cpp
+++ b/clang/test/CXX/drs/cwg185.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr185 { // dr185: 2.7
+namespace cwg185 { // cwg185: 2.7
 struct A {
   mutable int value;
   explicit A(int i) : value(i) {}
@@ -20,11 +20,11 @@ int foo() {
   return n.value;
 }
 
-// CHECK-LABEL: define {{.*}} i32 @dr185::foo()
-// CHECK:         call void @dr185::A::A(int)(ptr {{[^,]*}} %ref.tmp, {{.*}})
+// CHECK-LABEL: define {{.*}} i32 @cwg185::foo()
+// CHECK:         call void @cwg185::A::A(int)(ptr {{[^,]*}} %ref.tmp, {{.*}})
 // CHECK:         store ptr %ref.tmp, ptr %t
 // CHECK-NOT:     %t =
-// CHECK:         [[DR185_T:%.+]] = load ptr, ptr %t
-// CHECK:         call void @llvm.memcpy.p0.p0.i64(ptr {{[^,]*}} %n, ptr {{[^,]*}} [[DR185_T]], {{.*}})
+// CHECK:         [[CWG185_T:%.+]] = load ptr, ptr %t
+// CHECK:         call void @llvm.memcpy.p0.p0.i64(ptr {{[^,]*}} %n, ptr {{[^,]*}} [[CWG185_T]], {{.*}})
 // CHECK-LABEL: }
-} // namespace dr185
+} // namespace cwg185
diff --git a/clang/test/CXX/drs/dr193.cpp b/clang/test/CXX/drs/cwg193.cpp
similarity index 79%
rename from clang/test/CXX/drs/dr193.cpp
rename to clang/test/CXX/drs/cwg193.cpp
index c010dad50e403..0a6ac38941d20 100644
--- a/clang/test/CXX/drs/dr193.cpp
+++ b/clang/test/CXX/drs/cwg193.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr193 { // dr193: 2.7
+namespace cwg193 { // cwg193: 2.7
 struct A {
   ~A() NOTHROW {}
 };
@@ -35,12 +35,12 @@ void foo() {
 }
 
 // skipping over D1 (complete object destructor)
-// CHECK-LABEL: define {{.*}} void @dr193::D::~D(){{.*}}
-// CHECK-LABEL: define {{.*}} void @dr193::D::~D(){{.*}}
-// CHECK-NOT:     call void @dr193::A::~A()
-// CHECK-NOT:     call void @dr193::B::~B()
-// CHECK:         call void @dr193::C::~C()
-// CHECK:         call void @dr193::B::~B()
-// CHECK:         call void @dr193::A::~A()
+// CHECK-LABEL: define {{.*}} void @cwg193::D::~D(){{.*}}
+// CHECK-LABEL: define {{.*}} void @cwg193::D::~D(){{.*}}
+// CHECK-NOT:     call void @cwg193::A::~A()
+// CHECK-NOT:     call void @cwg193::B::~B()
+// CHECK:         call void @cwg193::C::~C()
+// CHECK:         call void @cwg193::B::~B()
+// CHECK:         call void @cwg193::A::~A()
 // CHECK-LABEL: }
-} // namespace dr193
+} // namespace cwg193
diff --git a/clang/test/CXX/drs/dr199.cpp b/clang/test/CXX/drs/cwg199.cpp
similarity index 85%
rename from clang/test/CXX/drs/dr199.cpp
rename to clang/test/CXX/drs/cwg199.cpp
index 7517d79680c6f..5d2e5110786f1 100644
--- a/clang/test/CXX/drs/dr199.cpp
+++ b/clang/test/CXX/drs/cwg199.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr199 { // dr199: 2.8
+namespace cwg199 { // cwg199: 2.8
 struct A {
   ~A() NOTHROW {}
 };
@@ -25,9 +25,9 @@ void foo() {
   A(), B();
 }
 
-// CHECK-LABEL: define {{.*}} void @dr199::foo()
-// CHECK-NOT:     call void @dr199::A::~A()
-// CHECK:         call void @dr199::B::~B()
-// CHECK:         call void @dr199::A::~A()
+// CHECK-LABEL: define {{.*}} void @cwg199::foo()
+// CHECK-NOT:     call void @cwg199::A::~A()
+// CHECK:         call void @cwg199::B::~B()
+// CHECK:         call void @cwg199::A::~A()
 // CHECK-LABEL: }
-} // namespace dr199
+} // namespace cwg199
diff --git a/clang/test/CXX/drs/dr201.cpp b/clang/test/CXX/drs/cwg201.cpp
similarity index 82%
rename from clang/test/CXX/drs/dr201.cpp
rename to clang/test/CXX/drs/cwg201.cpp
index 7e864981e13be..b6cf92a1fc748 100644
--- a/clang/test/CXX/drs/dr201.cpp
+++ b/clang/test/CXX/drs/cwg201.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr201 { // dr201: 2.8
+namespace cwg201 { // cwg201: 2.8
 
 extern void full_expr_fence() NOTHROW;
 
@@ -31,12 +31,12 @@ void foo() {
   full_expr_fence();
 }
 
-// CHECK-LABEL: define {{.*}} void @dr201::foo()
-// CHECK:         call void @dr201::full_expr_fence()
-// CHECK:         call void @dr201::B::B(dr201::A)
-// CHECK:         call void @dr201::A::~A()
-// CHECK:         call void @dr201::full_expr_fence()
-// CHECK:         call void @dr201::B::~B()
+// CHECK-LABEL: define {{.*}} void @cwg201::foo()
+// CHECK:         call void @cwg201::full_expr_fence()
+// CHECK:         call void @cwg201::B::B(cwg201::A)
+// CHECK:         call void @cwg201::A::~A()
+// CHECK:         call void @cwg201::full_expr_fence()
+// CHECK:         call void @cwg201::B::~B()
 // CHECK-LABEL: }
 
-} // namespace dr201
+} // namespace cwg201
diff --git a/clang/test/CXX/drs/dr210.cpp b/clang/test/CXX/drs/cwg210.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr210.cpp
rename to clang/test/CXX/drs/cwg210.cpp
index 156ee81093b43..2c3cf61a6a5b1 100644
--- a/clang/test/CXX/drs/dr210.cpp
+++ b/clang/test/CXX/drs/cwg210.cpp
@@ -13,7 +13,7 @@
 #pragma clang diagnostic pop
 #endif
 
-namespace dr210 { // dr210: 2.7
+namespace cwg210 { // cwg210: 2.7
 struct B {
   long i;
   B();
@@ -33,9 +33,9 @@ void toss(const B* b) {
   throw *b;
 }
 
-// CHECK-LABEL: define {{.*}} void @dr210::toss(dr210::B const*)
+// CHECK-LABEL: define {{.*}} void @cwg210::toss(cwg210::B const*)
 // CHECK:         %[[EXCEPTION:.*]] = call ptr @__cxa_allocate_exception(i64 16)
-// CHECK:         call void @__cxa_throw(ptr %[[EXCEPTION]], ptr @typeinfo for dr210::B, ptr @dr210::B::~B())
+// CHECK:         call void @__cxa_throw(ptr %[[EXCEPTION]], ptr @typeinfo for cwg210::B, ptr @cwg210::B::~B())
 // CHECK-LABEL: }
 
-} // namespace dr210
+} // namespace cwg210
diff --git a/clang/test/CXX/drs/dr2335.cpp b/clang/test/CXX/drs/cwg2335.cpp
similarity index 56%
rename from clang/test/CXX/drs/dr2335.cpp
rename to clang/test/CXX/drs/cwg2335.cpp
index 33eebb2c4a5c5..8b00a9d2d98a5 100644
--- a/clang/test/CXX/drs/dr2335.cpp
+++ b/clang/test/CXX/drs/cwg2335.cpp
@@ -10,7 +10,7 @@
 // expected-no-diagnostics
 #endif
 
-namespace dr2335 { // dr2335: no drafting 2018-06
+namespace cwg2335 { // cwg2335: no drafting 2018-06
 // FIXME: current consensus is that the examples are well-formed.
 #if __cplusplus >= 201402L
 namespace ex1 {
@@ -25,24 +25,24 @@ namespace ex2 {
 template <int> struct X {};
 template <class T> struct partition_indices {
   static auto compute_right() { return X<I>(); }
-  // since-cxx14-error@-1 {{no member 'I' in 'dr2335::ex2::partition_indices<int>'; it has not yet been instantiated}}
-  //   since-cxx14-note@#dr2335-ex2-right {{in instantiation of member function 'dr2335::ex2::partition_indices<int>::compute_right' requested here}}
-  //   since-cxx14-note@#dr2335-ex2-inst {{in instantiation of template class 'dr2335::ex2::partition_indices<int>' requested here}}
-  //   since-cxx14-note@#dr2335-ex2-I {{not-yet-instantiated member is declared here}}
-  static constexpr auto right = compute_right; // #dr2335-ex2-right
-  static constexpr int I = sizeof(T); // #dr2335-ex2-I
+  // since-cxx14-error@-1 {{no member 'I' in 'cwg2335::ex2::partition_indices<int>'; it has not yet been instantiated}}
+  //   since-cxx14-note@#cwg2335-ex2-right {{in instantiation of member function 'cwg2335::ex2::partition_indices<int>::compute_right' requested here}}
+  //   since-cxx14-note@#cwg2335-ex2-inst {{in instantiation of template class 'cwg2335::ex2::partition_indices<int>' requested here}}
+  //   since-cxx14-note@#cwg2335-ex2-I {{not-yet-instantiated member is declared here}}
+  static constexpr auto right = compute_right; // #cwg2335-ex2-right
+  static constexpr int I = sizeof(T); // #cwg2335-ex2-I
 };
-template struct partition_indices<int>; // #dr2335-ex2-inst
+template struct partition_indices<int>; // #cwg2335-ex2-inst
 } // namespace ex2
 
 namespace ex3 {
 struct partition_indices {
-  static auto compute_right() {} // #dr2335-compute_right
-  static constexpr auto right = compute_right; // #dr2335-ex3-right
+  static auto compute_right() {} // #cwg2335-compute_right
+  static constexpr auto right = compute_right; // #cwg2335-ex3-right
   // since-cxx14-error@-1 {{function 'compute_right' with deduced return type cannot be used before it is defined}}
-  //   since-cxx14-note@#dr2335-compute_right {{'compute_right' declared here}}
-  // since-cxx14-error@#dr2335-ex3-right {{declaration of variable 'right' with deduced type 'const auto' requires an initializer}}
+  //   since-cxx14-note@#cwg2335-compute_right {{'compute_right' declared here}}
+  // since-cxx14-error@#cwg2335-ex3-right {{declaration of variable 'right' with deduced type 'const auto' requires an initializer}}
 };
 } // namespace ex3
 #endif
-} // namespace dr2335
+} // namespace cwg2335
diff --git a/clang/test/CXX/drs/dr2390.cpp b/clang/test/CXX/drs/cwg2390.cpp
similarity index 98%
rename from clang/test/CXX/drs/dr2390.cpp
rename to clang/test/CXX/drs/cwg2390.cpp
index 3931365b568ce..41bbd0d1c5499 100644
--- a/clang/test/CXX/drs/dr2390.cpp
+++ b/clang/test/CXX/drs/cwg2390.cpp
@@ -1,6 +1,6 @@
 // RUN: %clang_cc1 -E -P %s -o - | FileCheck %s
 
-// dr2390: 14
+// cwg2390: 14
 
 namespace PR48462 {
 // Test that macro expansion of the builtin argument works.
diff --git a/clang/test/CXX/drs/dr2504.cpp b/clang/test/CXX/drs/cwg2504.cpp
similarity index 90%
rename from clang/test/CXX/drs/dr2504.cpp
rename to clang/test/CXX/drs/cwg2504.cpp
index 686ea73cd6a0e..fa775df327cbe 100644
--- a/clang/test/CXX/drs/dr2504.cpp
+++ b/clang/test/CXX/drs/cwg2504.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK,SINCE-CXX11
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK,SINCE-CXX11
 
-namespace dr2504 { // dr2504: no
+namespace cwg2504 { // cwg2504: no
 #if __cplusplus >= 201103L
 struct V { V() = default; V(int); };
 struct Q { Q(); };
@@ -27,11 +27,11 @@ void foo() { C c; } // bar is not invoked, because the V subobject is not initia
 //        we are not supposed to unconditionally call `bar()` and call a constructor
 //        inherited from `V`.
 
-// SINCE-CXX11-LABEL: define linkonce_odr void @dr2504::B::B()
+// SINCE-CXX11-LABEL: define linkonce_odr void @cwg2504::B::B()
 // SINCE-CXX11-NOT:     br
-// SINCE-CXX11:         call noundef i32 @dr2504::bar()
+// SINCE-CXX11:         call noundef i32 @cwg2504::bar()
 // SINCE-CXX11-NOT:     br
-// SINCE-CXX11:         call void @dr2504::A::A(int)
+// SINCE-CXX11:         call void @cwg2504::A::A(int)
 // SINCE-CXX11-LABEL: }
 
 // CHECK: {{.*}}
diff --git a/clang/test/CXX/drs/dr292.cpp b/clang/test/CXX/drs/cwg292.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr292.cpp
rename to clang/test/CXX/drs/cwg292.cpp
index 19caeef291fa7..b05d3b92d6275 100644
--- a/clang/test/CXX/drs/dr292.cpp
+++ b/clang/test/CXX/drs/cwg292.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -disable-llvm-passes -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -disable-llvm-passes -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr292 { // dr292: 2.9
+namespace cwg292 { // cwg292: 2.9
 
 extern int g();
 
@@ -18,13 +18,13 @@ void f() {
   new A(g());
 }
 
-// CHECK-LABEL: define {{.*}} void @dr292::f()()
+// CHECK-LABEL: define {{.*}} void @cwg292::f()()
 // CHECK:         %[[CALL:.+]] = call {{.*}} @operator new(unsigned long)({{.*}})
-// CHECK:         invoke {{.*}} i32 @dr292::g()()
+// CHECK:         invoke {{.*}} i32 @cwg292::g()()
 // CHECK-NEXT:           to {{.*}} unwind label %lpad
 // CHECK-LABEL: lpad:
 // CHECK:         call void @operator delete(void*)(ptr {{.*}} %[[CALL]])
 // CHECK-LABEL: eh.resume:
 // CHECK-LABEL: }
 
-} // namespace dr292
+} // namespace cwg292
diff --git a/clang/test/CXX/drs/dr392.cpp b/clang/test/CXX/drs/cwg392.cpp
similarity index 88%
rename from clang/test/CXX/drs/dr392.cpp
rename to clang/test/CXX/drs/cwg392.cpp
index 26e6259f71961..e118dd7bdb5ca 100644
--- a/clang/test/CXX/drs/dr392.cpp
+++ b/clang/test/CXX/drs/cwg392.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr392 { // dr392: 2.8
+namespace cwg392 { // cwg392: 2.8
 
 struct A {
   operator bool() NOTHROW;
@@ -32,9 +32,9 @@ void f()
   if (C().get()) {}
 }
 
-} // namespace dr392
+} // namespace cwg392
 
-// CHECK-LABEL: define {{.*}} void @dr392::f()()
-// CHECK:         call {{.*}} i1 @dr392::A::operator bool()
-// CHECK:         call void @dr392::C::~C()
+// CHECK-LABEL: define {{.*}} void @cwg392::f()()
+// CHECK:         call {{.*}} i1 @cwg392::A::operator bool()
+// CHECK:         call void @cwg392::C::~C()
 // CHECK-LABEL: }
diff --git a/clang/test/CXX/drs/dr412.cpp b/clang/test/CXX/drs/cwg412.cpp
similarity index 99%
rename from clang/test/CXX/drs/dr412.cpp
rename to clang/test/CXX/drs/cwg412.cpp
index 8ea29135d1df8..7e75bececac87 100644
--- a/clang/test/CXX/drs/dr412.cpp
+++ b/clang/test/CXX/drs/cwg412.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -verify -fexceptions -fcxx-exceptions -pedantic-errors -DNOEXCEPT=noexcept -DBAD_ALLOC=
 // RUN: %clang_cc1 -std=c++2c %s -verify -fexceptions -fcxx-exceptions -pedantic-errors -DNOEXCEPT=noexcept -DBAD_ALLOC=
 
-// dr412: 3.4
+// cwg412: 3.4
 // lwg404: yes
 // lwg2340: yes
 
diff --git a/clang/test/CXX/drs/dr438.cpp b/clang/test/CXX/drs/cwg438.cpp
similarity index 94%
rename from clang/test/CXX/drs/dr438.cpp
rename to clang/test/CXX/drs/cwg438.cpp
index a6ed39b88c242..5f2fb7c70d879 100644
--- a/clang/test/CXX/drs/dr438.cpp
+++ b/clang/test/CXX/drs/cwg438.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr438 { // dr438: 2.7
+namespace cwg438 { // cwg438: 2.7
 
 void f() {
   long A[2];
@@ -14,9 +14,9 @@ void f() {
   A[A[0]] = 1;
 }
 
-} // namespace dr438
+} // namespace cwg438
 
-// CHECK-LABEL: define {{.*}} void @dr438::f()()
+// CHECK-LABEL: define {{.*}} void @cwg438::f()()
 // CHECK:         [[A:%.+]] = alloca [2 x i64]
 // CHECK:         {{.+}} = getelementptr inbounds [2 x i64], ptr [[A]], i64 0, i64 0
 // CHECK:         [[ARRAYIDX1:%.+]] = getelementptr inbounds [2 x i64], ptr [[A]], i64 0, i64 0
diff --git a/clang/test/CXX/drs/dr439.cpp b/clang/test/CXX/drs/cwg439.cpp
similarity index 94%
rename from clang/test/CXX/drs/dr439.cpp
rename to clang/test/CXX/drs/cwg439.cpp
index 46960af93bb9a..e409b803797fa 100644
--- a/clang/test/CXX/drs/dr439.cpp
+++ b/clang/test/CXX/drs/cwg439.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr439 { // dr439: 2.7
+namespace cwg439 { // cwg439: 2.7
 
 void f() {
   int* p1 = new int;
@@ -14,12 +14,12 @@ void f() {
   bool b = p1 == p2; // b will have the value true.
 }
 
-} // namespace dr439
+} // namespace cwg439
 
 // We're checking that p2 was copied from p1, and then was carried over
 // to the comparison without change.
 
-// CHECK-LABEL: define {{.*}} void @dr439::f()()
+// CHECK-LABEL: define {{.*}} void @cwg439::f()()
 // CHECK:         [[P1:%.+]] = alloca ptr, align 8
 // CHECK-NEXT:    [[P2:%.+]] = alloca ptr, align 8
 // CHECK:         [[TEMP0:%.+]] = load ptr, ptr [[P1]]
diff --git a/clang/test/CXX/drs/dr441.cpp b/clang/test/CXX/drs/cwg441.cpp
similarity index 76%
rename from clang/test/CXX/drs/dr441.cpp
rename to clang/test/CXX/drs/cwg441.cpp
index 6504bba689d22..5f566f2301936 100644
--- a/clang/test/CXX/drs/dr441.cpp
+++ b/clang/test/CXX/drs/cwg441.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr441 { // dr441: 2.7
+namespace cwg441 { // cwg441: 2.7
 
 struct A {
   A() {}
@@ -17,21 +17,21 @@ int i;
 int& ir = i;
 int* ip = &i;
 
-} // namespace dr441
+} // namespace cwg441
 
-// CHECK-DAG:   @dr441::dynamic_init = global %"struct.dr441::A" zeroinitializer
-// CHECK-DAG:   @dr441::i = global i32 0
-// CHECK-DAG:   @dr441::ir = constant ptr @dr441::i
-// CHECK-DAG:   @dr441::ip = global ptr @dr441::i
-// CHECK-DAG:   @llvm.global_ctors = appending global [{{.+}}] [{ {{.+}} } { {{.+}}, ptr @_GLOBAL__sub_I_dr441.cpp, {{.+}} }]
+// CHECK-DAG:   @cwg441::dynamic_init = global %"struct.cwg441::A" zeroinitializer
+// CHECK-DAG:   @cwg441::i = global i32 0
+// CHECK-DAG:   @cwg441::ir = constant ptr @cwg441::i
+// CHECK-DAG:   @cwg441::ip = global ptr @cwg441::i
+// CHECK-DAG:   @llvm.global_ctors = appending global [{{.+}}] [{ {{.+}} } { {{.+}}, ptr @_GLOBAL__sub_I_cwg441.cpp, {{.+}} }]
 
 // CHECK-LABEL: define {{.*}} void @__cxx_global_var_init()
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    call void @dr441::A::A()({{.*}} @dr441::dynamic_init)
+// CHECK-NEXT:    call void @cwg441::A::A()({{.*}} @cwg441::dynamic_init)
 // CHECK-NEXT:    ret void
 // CHECK-NEXT:  }
 
-// CHECK-LABEL: define {{.*}} void @_GLOBAL__sub_I_dr441.cpp()
+// CHECK-LABEL: define {{.*}} void @_GLOBAL__sub_I_cwg441.cpp()
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    call void @__cxx_global_var_init()
 // CHECK-NEXT:    ret void
diff --git a/clang/test/CXX/drs/dr462.cpp b/clang/test/CXX/drs/cwg462.cpp
similarity index 87%
rename from clang/test/CXX/drs/dr462.cpp
rename to clang/test/CXX/drs/cwg462.cpp
index 2b268778ea10d..bdbcacd733bbb 100644
--- a/clang/test/CXX/drs/dr462.cpp
+++ b/clang/test/CXX/drs/cwg462.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr462 { // dr462: 2.7
+namespace cwg462 { // cwg462: 2.7
 
 struct A {
   ~A() NOTHROW {}
@@ -25,9 +25,9 @@ void f() {
   full_expr_fence();
 }
 
-} // namespace dr462
+} // namespace cwg462
 
-// CHECK-LABEL: define {{.*}} void @dr462::f()()
-// CHECK:         call void @dr462::full_expr_fence()()
-// CHECK:         call void @dr462::A::~A()
+// CHECK-LABEL: define {{.*}} void @cwg462::f()()
+// CHECK:         call void @cwg462::full_expr_fence()()
+// CHECK:         call void @cwg462::A::~A()
 // CHECK-LABEL: }
diff --git a/clang/test/CXX/drs/dr492.cpp b/clang/test/CXX/drs/cwg492.cpp
similarity index 94%
rename from clang/test/CXX/drs/dr492.cpp
rename to clang/test/CXX/drs/cwg492.cpp
index f53f1cb541240..7fc46b04d72bd 100644
--- a/clang/test/CXX/drs/dr492.cpp
+++ b/clang/test/CXX/drs/cwg492.cpp
@@ -18,7 +18,7 @@ struct type_info {
 }; 
 }
 
-namespace dr492 { // dr492: 2.7
+namespace cwg492 { // cwg492: 2.7
 
 void f() {
   typeid(int).name();
@@ -27,9 +27,9 @@ void f() {
   typeid(const volatile int).name();
 }
 
-} // namespace dr492
+} // namespace cwg492
 
-// CHECK-LABEL: define {{.*}} void @dr492::f()()
+// CHECK-LABEL: define {{.*}} void @cwg492::f()()
 // CHECK:         {{.*}} = call {{.*}} @std::type_info::name() const({{.*}} @typeinfo for int)
 // CHECK-NEXT:    {{.*}} = call {{.*}} @std::type_info::name() const({{.*}} @typeinfo for int)
 // CHECK-NEXT:    {{.*}} = call {{.*}} @std::type_info::name() const({{.*}} @typeinfo for int)
diff --git a/clang/test/CXX/drs/dr519.cpp b/clang/test/CXX/drs/cwg519.cpp
similarity index 95%
rename from clang/test/CXX/drs/dr519.cpp
rename to clang/test/CXX/drs/cwg519.cpp
index 67c01d95ef7c6..ce8a1cc95f600 100644
--- a/clang/test/CXX/drs/dr519.cpp
+++ b/clang/test/CXX/drs/cwg519.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr519 { // dr519: 2.7
+namespace cwg519 { // cwg519: 2.7
 void f() {
   int *a = 0;
   void *v = a;
@@ -16,12 +16,12 @@ void f() {
   int *b = static_cast<int*>(w);
   bool c2 = b == static_cast<int *>(0);
 }
-} // namespace dr519
+} // namespace cwg519
 
 // We're checking that `null`s that were initially stored in `a` and `w`
 // are simply copied over all the way to respective comparisons with `null`.
 
-// CHECK-LABEL: define {{.*}} void @dr519::f()()
+// CHECK-LABEL: define {{.*}} void @cwg519::f()()
 // CHECK:         store ptr null, ptr [[A:%.+]],
 // CHECK-NEXT:    [[TEMP_A:%.+]] = load ptr, ptr [[A]] 
 // CHECK-NEXT:    store ptr [[TEMP_A]], ptr [[V:%.+]],
diff --git a/clang/test/CXX/drs/dr571.cpp b/clang/test/CXX/drs/cwg571.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr571.cpp
rename to clang/test/CXX/drs/cwg571.cpp
index 19a85b7ddc350..9f0f455fb7276 100644
--- a/clang/test/CXX/drs/dr571.cpp
+++ b/clang/test/CXX/drs/cwg571.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr571 { // dr571: 2.7
+namespace cwg571 { // cwg571: 2.7
   typedef int &ir;
   int n;
   const ir r = n;
@@ -16,5 +16,5 @@ namespace dr571 { // dr571: 2.7
 
 // Entities have external linkage by default.
 
-// CHECK: @dr571::r = constant ptr @dr571::n
-// CHECK: @dr571::r2 = constant ptr @dr571::n
+// CHECK: @cwg571::r = constant ptr @cwg571::n
+// CHECK: @cwg571::r2 = constant ptr @cwg571::n
diff --git a/clang/test/CXX/drs/dr605.cpp b/clang/test/CXX/drs/cwg605.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr605.cpp
rename to clang/test/CXX/drs/cwg605.cpp
index 6c212d8dabc06..2fd9e8155bf77 100644
--- a/clang/test/CXX/drs/dr605.cpp
+++ b/clang/test/CXX/drs/cwg605.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr605 { // dr605: 2.7
+namespace cwg605 { // cwg605: 2.7
 
 template <class T>
 static T f(T t) {}
@@ -18,6 +18,6 @@ void g(int a) {
   f(a);
 }
 
-} // namespace dr605
+} // namespace cwg605
 
-// CHECK: define internal {{.*}} i32 @int dr605::f<int>(int)
+// CHECK: define internal {{.*}} i32 @int cwg605::f<int>(int)
diff --git a/clang/test/CXX/drs/dr650.cpp b/clang/test/CXX/drs/cwg650.cpp
similarity index 86%
rename from clang/test/CXX/drs/dr650.cpp
rename to clang/test/CXX/drs/cwg650.cpp
index 715b4fdf04a7f..dcb844095b059 100644
--- a/clang/test/CXX/drs/dr650.cpp
+++ b/clang/test/CXX/drs/cwg650.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr650 { // dr650: 2.8
+namespace cwg650 { // cwg650: 2.8
 
 struct Q {
   ~Q() NOTHROW;
@@ -31,10 +31,10 @@ const S& f() {
   return (R(), S());
 }
 
-} // namespace dr650
+} // namespace cwg650
 
-// CHECK-LABEL: define {{.*}} @dr650::f()()
-// CHECK:         call void @dr650::S::~S()
-// CHECK:         call void @dr650::R::~R()
-// CHECK:         call void @dr650::Q::~Q()
+// CHECK-LABEL: define {{.*}} @cwg650::f()()
+// CHECK:         call void @cwg650::S::~S()
+// CHECK:         call void @cwg650::R::~R()
+// CHECK:         call void @cwg650::Q::~Q()
 // CHECK-LABEL: }
diff --git a/clang/test/CXX/drs/dr653.cpp b/clang/test/CXX/drs/cwg653.cpp
similarity index 92%
rename from clang/test/CXX/drs/dr653.cpp
rename to clang/test/CXX/drs/cwg653.cpp
index fd1f0153bfb74..3aeb394347ea6 100644
--- a/clang/test/CXX/drs/dr653.cpp
+++ b/clang/test/CXX/drs/cwg653.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr653 { // dr653: 2.7
+namespace cwg653 { // cwg653: 2.7
 
 union U {
   int a;
@@ -18,8 +18,8 @@ void f(U u) {
   v = u;
 }
 
-} // namespace dr653
+} // namespace cwg653
 
-// CHECK-LABEL: define {{.*}} void @dr653::f(dr653::U)
+// CHECK-LABEL: define {{.*}} void @cwg653::f(cwg653::U)
 // CHECK:         call void @llvm.memcpy.p0.p0.i64(ptr {{.*}} %v, ptr {{.*}} %u, {{.*}})
 // CHECK-LABEL: }
diff --git a/clang/test/CXX/drs/dr658.cpp b/clang/test/CXX/drs/cwg658.cpp
similarity index 92%
rename from clang/test/CXX/drs/dr658.cpp
rename to clang/test/CXX/drs/cwg658.cpp
index 51034c2af3bf3..2f7f1ad7deda5 100644
--- a/clang/test/CXX/drs/dr658.cpp
+++ b/clang/test/CXX/drs/cwg658.cpp
@@ -6,17 +6,17 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr658 { // dr658: 2.7
+namespace cwg658 { // cwg658: 2.7
 
 void f(int* p1) {
   char* p2 = reinterpret_cast<char*>(p1);
 }
 
-} // namespace dr658
+} // namespace cwg658
 
 // We're checking that p1 is stored into p2 without changes.
 
-// CHECK-LABEL: define {{.*}} void @dr658::f(int*)(ptr noundef %p1)
+// CHECK-LABEL: define {{.*}} void @cwg658::f(int*)(ptr noundef %p1)
 // CHECK:         [[P1_ADDR:%.+]] = alloca ptr, align 8
 // CHECK-NEXT:    [[P2:%.+]] = alloca ptr, align 8
 // CHECK:         store ptr %p1, ptr [[P1_ADDR]]
diff --git a/clang/test/CXX/drs/dr661.cpp b/clang/test/CXX/drs/cwg661.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr661.cpp
rename to clang/test/CXX/drs/cwg661.cpp
index 4e97bb7088476..5572131710708 100644
--- a/clang/test/CXX/drs/dr661.cpp
+++ b/clang/test/CXX/drs/cwg661.cpp
@@ -6,9 +6,9 @@
 // RUN: %clang_cc1 -std=c++23 %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 // RUN: %clang_cc1 -std=c++2c %s -triple x86_64-linux-gnu -emit-llvm -o - -fexceptions -fcxx-exceptions -pedantic-errors | llvm-cxxfilt -n | FileCheck %s --check-prefixes CHECK
 
-namespace dr661 {
+namespace cwg661 {
 
-void f(int a, int b) { // dr661: 2.7
+void f(int a, int b) { // cwg661: 2.7
   a == b;
   a != b;
   a < b;
@@ -17,9 +17,9 @@ void f(int a, int b) { // dr661: 2.7
   a >= b;
 }
 
-} // namespace dr661
+} // namespace cwg661
 
-// CHECK-LABEL: define {{.*}} void @dr661::f(int, int)
+// CHECK-LABEL: define {{.*}} void @cwg661::f(int, int)
 // CHECK:         icmp eq
 // CHECK:         icmp ne
 // CHECK:         icmp slt
diff --git a/clang/test/CXX/drs/dr672.cpp b/clang/test/CXX/drs/cwg672.cpp
similarity index 91%
rename from clang/test/CXX/drs/dr672.cpp
rename to clang/test/CXX/drs/cwg672.cpp
index d5f0530ecbc9d..07a64eaaf8269 100644
--- a/clang/test/CXX/drs/dr672.cpp
+++ b/clang/test/CXX/drs/cwg672.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr672 { // dr672: 2.7
+namespace cwg672 { // cwg672: 2.7
 
 struct A {
   A() NOTHROW;
@@ -22,11 +22,11 @@ void f() {
   A *a = new A;
 }
 
-} // namespace dr672
+} // namespace cwg672
 
-// CHECK-LABEL: define {{.*}} void @dr672::f()()
+// CHECK-LABEL: define {{.*}} void @cwg672::f()()
 // CHECK:         [[A:%.+]] = alloca ptr
 // CHECK:         [[CALL:%.+]] = call {{.*}} ptr @operator new(unsigned long)
-// CHECK:         call void @dr672::A::A()
+// CHECK:         call void @cwg672::A::A()
 // CHECK:         store ptr [[CALL]], ptr [[A]]
 // CHECK-LABEL: }
diff --git a/clang/test/CXX/drs/dr593.cpp b/clang/test/CXX/drs/cwgr593.cpp
similarity index 93%
rename from clang/test/CXX/drs/dr593.cpp
rename to clang/test/CXX/drs/cwgr593.cpp
index 4998af966ebb9..d747f4e4a1617 100644
--- a/clang/test/CXX/drs/dr593.cpp
+++ b/clang/test/CXX/drs/cwgr593.cpp
@@ -12,7 +12,7 @@
 #define NOTHROW noexcept(true)
 #endif
 
-namespace dr593 { // dr593: 2.8
+namespace cwg593 { // cwg593: 2.8
 
 void f();
 void fence() NOTHROW;
@@ -29,7 +29,7 @@ void g() {
   A();
 }
 
-} // namespace dr593
+} // namespace cwg593
 
-// CHECK:      call void @dr593::fence()()
+// CHECK:      call void @cwg593::fence()()
 // CHECK-NEXT: invoke void @__cxa_rethrow()
diff --git a/clang/test/CXX/drs/dr0xx.cpp b/clang/test/CXX/drs/dr0xx.cpp
index 5959f0a0c8dd6..a304862885c64 100644
--- a/clang/test/CXX/drs/dr0xx.cpp
+++ b/clang/test/CXX/drs/dr0xx.cpp
@@ -5,44 +5,44 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors -triple %itanium_abi_triple
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors -triple %itanium_abi_triple
 
-namespace dr1 { // dr1: no
-  namespace X { extern "C" void dr1_f(int a = 1); }
-  namespace Y { extern "C" void dr1_f(int a = 1); }
-  using X::dr1_f; using Y::dr1_f;
+namespace cwg1 { // cwg1: no
+  namespace X { extern "C" void cwg1_f(int a = 1); }
+  namespace Y { extern "C" void cwg1_f(int a = 1); }
+  using X::cwg1_f; using Y::cwg1_f;
   void g() {
-    dr1_f(0);
+    cwg1_f(0);
     // FIXME: This should be rejected, due to the ambiguous default argument.
-    dr1_f();
+    cwg1_f();
   }
   namespace X {
-    using Y::dr1_f;
+    using Y::cwg1_f;
     void h() {
-      dr1_f(0);
+      cwg1_f(0);
       // FIXME: This should be rejected, due to the ambiguous default argument.
-      dr1_f();
+      cwg1_f();
     }
   }
 
   namespace X {
     void z(int);
   }
-  void X::z(int = 1) {} // #dr1-z
+  void X::z(int = 1) {} // #cwg1-z
   namespace X {
     void z(int = 1);
     // expected-error@-1 {{redefinition of default argument}}
-    //   expected-note@#dr1-z {{previous definition is here}}
+    //   expected-note@#cwg1-z {{previous definition is here}}
   }
 
   void i(int = 1);
   void j() {
     void i(int = 1);
-    using dr1::i;
+    using cwg1::i;
     i(0);
     // FIXME: This should be rejected, due to the ambiguous default argument.
     i();
   }
   void k() {
-    using dr1::i;
+    using cwg1::i;
     void i(int = 1);
     i(0);
     // FIXME: This should be rejected, due to the ambiguous default argument.
@@ -50,27 +50,27 @@ namespace dr1 { // dr1: no
   }
 }
 
-namespace dr3 { // dr3: yes
+namespace cwg3 { // cwg3: yes
   template<typename T> struct A {};
-  template<typename T> void f(T) { A<T> a; } // #dr3-f-T
+  template<typename T> void f(T) { A<T> a; } // #cwg3-f-T
   template void f(int);
   template<> struct A<int> {};
-  // expected-error@-1 {{explicit specialization of 'dr3::A<int>' after instantiation}}
-  //   expected-note@#dr3-f-T {{implicit instantiation first required here}}
+  // expected-error@-1 {{explicit specialization of 'cwg3::A<int>' after instantiation}}
+  //   expected-note@#cwg3-f-T {{implicit instantiation first required here}}
 }
 
-namespace dr4 { // dr4: 2.8
+namespace cwg4 { // cwg4: 2.8
   extern "C" {
-    static void dr4_f(int) {}
-    static void dr4_f(float) {}
-    void dr4_g(int) {} // #dr4-g-int
-    void dr4_g(float) {}
-    // expected-error@-1 {{conflicting types for 'dr4_g'}}
-    //   expected-note@#dr4-g-int {{previous definition is here}}
+    static void cwg4_f(int) {}
+    static void cwg4_f(float) {}
+    void cwg4_g(int) {} // #cwg4-g-int
+    void cwg4_g(float) {}
+    // expected-error@-1 {{conflicting types for 'cwg4_g'}}
+    //   expected-note@#cwg4-g-int {{previous definition is here}}
   }
 }
 
-namespace dr5 { // dr5: 3.1
+namespace cwg5 { // cwg5: 3.1
   struct A {} a;
   struct B {
     B(const A&);
@@ -84,24 +84,24 @@ namespace dr5 { // dr5: 3.1
   const C c = e;
 }
 
-namespace dr7 { // dr7: 3.4
+namespace cwg7 { // cwg7: 3.4
   class A { public: ~A(); };
-  class B : virtual private A {}; // #dr7-B
-  class C : public B {} c; // #dr7-C
-  // expected-error@#dr7-C {{inherited virtual base class 'A' has private destructor}}
-  //   expected-note@#dr7-C {{in implicit default constructor for 'dr7::C' first required here}}
-  //   expected-note@#dr7-B {{declared private here}}
-  // expected-error@#dr7-C {{inherited virtual base class 'A' has private destructor}}
-  //   expected-note@#dr7-C {{in implicit destructor for 'dr7::C' first required here}}
-  //   expected-note@#dr7-B {{declared private here}}
+  class B : virtual private A {}; // #cwg7-B
+  class C : public B {} c; // #cwg7-C
+  // expected-error@#cwg7-C {{inherited virtual base class 'A' has private destructor}}
+  //   expected-note@#cwg7-C {{in implicit default constructor for 'cwg7::C' first required here}}
+  //   expected-note@#cwg7-B {{declared private here}}
+  // expected-error@#cwg7-C {{inherited virtual base class 'A' has private destructor}}
+  //   expected-note@#cwg7-C {{in implicit destructor for 'cwg7::C' first required here}}
+  //   expected-note@#cwg7-B {{declared private here}}
   class VeryDerivedC : public B, virtual public A {} vdc;
 
-  class X { ~X(); }; // #dr7-X
+  class X { ~X(); }; // #cwg7-X
   class Y : X { ~Y() {} };
   // expected-error@-1 {{base class 'X' has private destructor}}
-  //   expected-note@#dr7-X {{implicitly declared private here}}
+  //   expected-note@#cwg7-X {{implicitly declared private here}}
 
-  namespace PR16370 { // This regressed the first time DR7 was fixed.
+  namespace PR16370 { // This regressed the first time CWG7 was fixed.
     struct S1 { virtual ~S1(); };
     struct S2 : S1 {};
     struct S3 : S2 {};
@@ -114,7 +114,7 @@ namespace dr7 { // dr7: 3.4
   }
 }
 
-namespace dr8 { // dr8: dup 45
+namespace cwg8 { // cwg8: dup 45
   class A {
     struct U;
     static const int k = 5;
@@ -126,23 +126,23 @@ namespace dr8 { // dr8: dup 45
   A::T<A::U, A::k, &A::f> *A::g() { return 0; }
 }
 
-namespace dr9 { // dr9: 2.8
+namespace cwg9 { // cwg9: 2.8
   struct B {
   protected:
-    int m; // #dr9-m
+    int m; // #cwg9-m
     friend int R1();
   };
-  struct N : protected B { // #dr9-N
+  struct N : protected B { // #cwg9-N
     friend int R2();
   } n;
   int R1() { return n.m; }
-  // expected-error@-1 {{'m' is a protected member of 'dr9::B'}}
-  //   expected-note@#dr9-N {{constrained by protected inheritance here}}
-  //   expected-note@#dr9-m {{member is declared here}}
+  // expected-error@-1 {{'m' is a protected member of 'cwg9::B'}}
+  //   expected-note@#cwg9-N {{constrained by protected inheritance here}}
+  //   expected-note@#cwg9-m {{member is declared here}}
   int R2() { return n.m; }
 }
 
-namespace dr10 { // dr10: dup 45
+namespace cwg10 { // cwg10: dup 45
   class A {
     struct B {
       A::B *p;
@@ -150,7 +150,7 @@ namespace dr10 { // dr10: dup 45
   };
 }
 
-namespace dr11 { // dr11: yes
+namespace cwg11 { // cwg11: yes
   template<typename T> struct A : T {
     using typename T::U;
     U u;
@@ -164,19 +164,19 @@ namespace dr11 { // dr11: yes
   A<X> ax;
 }
 
-namespace dr12 { // dr12: sup 239
+namespace cwg12 { // cwg12: sup 239
   enum E { e };
   E &f(E, E = e);
   void g() {
     int &f(int, E = e);
-    // Under DR12, these call two different functions.
-    // Under DR239, they call the same function.
+    // Under CWG12, these call two different functions.
+    // Under CWG239, they call the same function.
     int &b = f(e);
     int &c = f(1);
   }
 }
 
-namespace dr13 { // dr13: no
+namespace cwg13 { // cwg13: no
   extern "C" void f(int);
   void g(char);
 
@@ -191,60 +191,60 @@ namespace dr13 { // dr13: no
   int a4 = h(g);
 }
 
-namespace dr14 { // dr14: 3.4
-  namespace X { extern "C" int dr14_f(); }
-  namespace Y { extern "C" int dr14_f(); }
+namespace cwg14 { // cwg14: 3.4
+  namespace X { extern "C" int cwg14_f(); }
+  namespace Y { extern "C" int cwg14_f(); }
   using namespace X;
   using namespace Y;
-  int k = dr14_f();
+  int k = cwg14_f();
 
   class C {
     int k;
-    friend int Y::dr14_f();
+    friend int Y::cwg14_f();
   } c;
   namespace Z {
-    extern "C" int dr14_f() { return c.k; }
+    extern "C" int cwg14_f() { return c.k; }
   }
 
-  namespace X { typedef int T; typedef int U; } // #dr14-X-U
-  namespace Y { typedef int T; typedef long U; } // #dr14-Y-U
+  namespace X { typedef int T; typedef int U; } // #cwg14-X-U
+  namespace Y { typedef int T; typedef long U; } // #cwg14-Y-U
   T t; // ok, same type both times
   U u;
   // expected-error@-1 {{reference to 'U' is ambiguous}}
-  //   expected-note@#dr14-X-U {{candidate found by name lookup is 'dr14::X::U'}}
-  //   expected-note@#dr14-Y-U {{candidate found by name lookup is 'dr14::Y::U'}}
+  //   expected-note@#cwg14-X-U {{candidate found by name lookup is 'cwg14::X::U'}}
+  //   expected-note@#cwg14-Y-U {{candidate found by name lookup is 'cwg14::Y::U'}}
 }
 
-namespace dr15 { // dr15: yes
-  template<typename T> void f(int); // #dr15-f-decl-first
+namespace cwg15 { // cwg15: yes
+  template<typename T> void f(int); // #cwg15-f-decl-first
   template<typename T> void f(int = 0);
   // expected-error@-1 {{default arguments cannot be added to a function template that has already been declared}}
-  //   expected-note@#dr15-f-decl-first {{previous template declaration is here}}
+  //   expected-note@#cwg15-f-decl-first {{previous template declaration is here}}
 }
 
-namespace dr16 { // dr16: 2.8
-  class A { // #dr16-A
-    void f(); // #dr16-A-f-decl
+namespace cwg16 { // cwg16: 2.8
+  class A { // #cwg16-A
+    void f(); // #cwg16-A-f-decl
     friend class C;
   };
-  class B : A {}; // #dr16-B
+  class B : A {}; // #cwg16-B
   class C : B {
     void g() {
       f();
-      // expected-error@-1 {{'f' is a private member of 'dr16::A'}}
-      //   expected-note@#dr16-B {{constrained by implicitly private inheritance here}}
-      //   expected-note@#dr16-A-f-decl {{member is declared here}}
-      A::f(); // #dr16-A-f-call
-      // expected-error@#dr16-A-f-call {{'A' is a private member of 'dr16::A'}}
-      //   expected-note@#dr16-B {{constrained by implicitly private inheritance here}}
-      //   expected-note@#dr16-A {{member is declared here}}
-      // expected-error@#dr16-A-f-call {{cannot cast 'dr16::C' to its private base class 'dr16::A'}}
-      //   expected-note@#dr16-B {{implicitly declared private here}}
+      // expected-error@-1 {{'f' is a private member of 'cwg16::A'}}
+      //   expected-note@#cwg16-B {{constrained by implicitly private inheritance here}}
+      //   expected-note@#cwg16-A-f-decl {{member is declared here}}
+      A::f(); // #cwg16-A-f-call
+      // expected-error@#cwg16-A-f-call {{'A' is a private member of 'cwg16::A'}}
+      //   expected-note@#cwg16-B {{constrained by implicitly private inheritance here}}
+      //   expected-note@#cwg16-A {{member is declared here}}
+      // expected-error@#cwg16-A-f-call {{cannot cast 'cwg16::C' to its private base class 'cwg16::A'}}
+      //   expected-note@#cwg16-B {{implicitly declared private here}}
     }
   };
 }
 
-namespace dr17 { // dr17: yes
+namespace cwg17 { // cwg17: yes
   class A {
     int n;
     int f();
@@ -257,38 +257,38 @@ namespace dr17 { // dr17: yes
   };
 }
 
-// dr18: sup 577
+// cwg18: sup 577
 
-namespace dr19 { // dr19: 3.1
+namespace cwg19 { // cwg19: 3.1
   struct A {
-    int n; // #dr19-n
+    int n; // #cwg19-n
   };
-  struct B : protected A { // #dr19-B
+  struct B : protected A { // #cwg19-B
   };
   struct C : B {} c;
   struct D : B {
     int get1() { return c.n; }
-    // expected-error@-1 {{'n' is a protected member of 'dr19::A'}}
-    //   expected-note@#dr19-B {{constrained by protected inheritance here}}
-    //   expected-note@#dr19-n {{member is declared here}}
+    // expected-error@-1 {{'n' is a protected member of 'cwg19::A'}}
+    //   expected-note@#cwg19-B {{constrained by protected inheritance here}}
+    //   expected-note@#cwg19-n {{member is declared here}}
     int get2() { return ((A&)c).n; } // ok, A is an accessible base of B from here
   };
 }
 
-namespace dr20 { // dr20: 2.8
+namespace cwg20 { // cwg20: 2.8
   class X {
   public:
     X();
   private:
-    X(const X&); // #dr20-X-ctor
+    X(const X&); // #cwg20-X-ctor
   };
   X &f();
   X x = f();
-  // expected-error@-1 {{calling a private constructor of class 'dr20::X'}}
-  //   expected-note@#dr20-X-ctor {{declared private here}}
+  // expected-error@-1 {{calling a private constructor of class 'cwg20::X'}}
+  //   expected-note@#cwg20-X-ctor {{declared private here}}
 }
 
-namespace dr21 { // dr21: 3.4
+namespace cwg21 { // cwg21: 3.4
   template<typename T> struct A;
   struct X {
     template<typename T = int> friend struct A;
@@ -298,25 +298,25 @@ namespace dr21 { // dr21: 3.4
   };
 }
 
-namespace dr22 { // dr22: sup 481
-  template<typename dr22_T = dr22_T> struct X;
-  // expected-error@-1 {{unknown type name 'dr22_T'}}
+namespace cwg22 { // cwg22: sup 481
+  template<typename cwg22_T = cwg22_T> struct X;
+  // expected-error@-1 {{unknown type name 'cwg22_T'}}
   typedef int T;
   template<typename T = T> struct Y;
 }
 
-namespace dr23 { // dr23: yes
-  template<typename T> void f(T, T); // #dr23-f-T-T
-  template<typename T> void f(T, int); // #dr23-f-T-int
+namespace cwg23 { // cwg23: yes
+  template<typename T> void f(T, T); // #cwg23-f-T-T
+  template<typename T> void f(T, int); // #cwg23-f-T-int
   void g() { f(0, 0); }
   // expected-error@-1 {{call to 'f' is ambiguous}}
-  //   expected-note@#dr23-f-T-T {{candidate function [with T = int]}}
-  //   expected-note@#dr23-f-T-int {{candidate function [with T = int]}}
+  //   expected-note@#cwg23-f-T-T {{candidate function [with T = int]}}
+  //   expected-note@#cwg23-f-T-int {{candidate function [with T = int]}}
 }
 
-// dr24: na
+// cwg24: na
 
-namespace dr25 { // dr25: yes
+namespace cwg25 { // cwg25: yes
   struct A {
     void f() throw(int);
     // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
@@ -351,7 +351,7 @@ namespace dr25 { // dr25: yes
   }
 }
 
-namespace dr26 { // dr26: yes
+namespace cwg26 { // cwg26: yes
   struct A { A(A, const A & = A()); };
   // expected-error@-1 {{copy constructor must pass its first argument by reference}}
   struct B {
@@ -371,77 +371,77 @@ namespace dr26 { // dr26: yes
   };
 }
 
-namespace dr27 { // dr27: yes
+namespace cwg27 { // cwg27: yes
   enum E { e } n;
   E &m = true ? n : n;
 }
 
-// dr28: na lib
+// cwg28: na lib
 
-namespace dr29 { // dr29: 3.4
-  void dr29_f0(); // #dr29-f0
-  void g0() { void dr29_f0(); }
-  extern "C++" void g0_cxx() { void dr29_f0(); }
-  extern "C" void g0_c() { void dr29_f0(); }
-  // expected-error@-1 {{declaration of 'dr29_f0' has a different language linkage}}
-  //   expected-note@#dr29-f0 {{previous declaration is here}}
+namespace cwg29 { // cwg29: 3.4
+  void cwg29_f0(); // #cwg29-f0
+  void g0() { void cwg29_f0(); }
+  extern "C++" void g0_cxx() { void cwg29_f0(); }
+  extern "C" void g0_c() { void cwg29_f0(); }
+  // expected-error@-1 {{declaration of 'cwg29_f0' has a different language linkage}}
+  //   expected-note@#cwg29-f0 {{previous declaration is here}}
 
-  extern "C" void dr29_f1(); // #dr29-f1
-  void g1() { void dr29_f1(); }
-  extern "C" void g1_c() { void dr29_f1(); }
-  extern "C++" void g1_cxx() { void dr29_f1(); }
-  // expected-error@-1 {{declaration of 'dr29_f1' has a different language linkage}}
-  //   expected-note@#dr29-f1 {{previous declaration is here}}
+  extern "C" void cwg29_f1(); // #cwg29-f1
+  void g1() { void cwg29_f1(); }
+  extern "C" void g1_c() { void cwg29_f1(); }
+  extern "C++" void g1_cxx() { void cwg29_f1(); }
+  // expected-error@-1 {{declaration of 'cwg29_f1' has a different language linkage}}
+  //   expected-note@#cwg29-f1 {{previous declaration is here}}
 
-  void g2() { void dr29_f2(); } // #dr29-f2
-  extern "C" void dr29_f2();
-  // expected-error@-1 {{declaration of 'dr29_f2' has a different language linkage}}
-  //   expected-note@#dr29-f2 {{previous declaration is here}}
+  void g2() { void cwg29_f2(); } // #cwg29-f2
+  extern "C" void cwg29_f2();
+  // expected-error@-1 {{declaration of 'cwg29_f2' has a different language linkage}}
+  //   expected-note@#cwg29-f2 {{previous declaration is here}}
 
-  extern "C" void g3() { void dr29_f3(); } // #dr29-f3
-  extern "C++" void dr29_f3();
-  // expected-error@-1 {{declaration of 'dr29_f3' has a different language linkage}}
-  //   expected-note@#dr29-f3 {{previous declaration is here}}
+  extern "C" void g3() { void cwg29_f3(); } // #cwg29-f3
+  extern "C++" void cwg29_f3();
+  // expected-error@-1 {{declaration of 'cwg29_f3' has a different language linkage}}
+  //   expected-note@#cwg29-f3 {{previous declaration is here}}
 
-  extern "C++" void g4() { void dr29_f4(); } // #dr29-f4
-  extern "C" void dr29_f4();
-  // expected-error@-1 {{declaration of 'dr29_f4' has a different language linkage}}
-  //   expected-note@#dr29-f4 {{previous declaration is here}}
+  extern "C++" void g4() { void cwg29_f4(); } // #cwg29-f4
+  extern "C" void cwg29_f4();
+  // expected-error@-1 {{declaration of 'cwg29_f4' has a different language linkage}}
+  //   expected-note@#cwg29-f4 {{previous declaration is here}}
 
   extern "C" void g5();
-  extern "C++" void dr29_f5();
+  extern "C++" void cwg29_f5();
   void g5() {
-    void dr29_f5(); // ok, g5 is extern "C" but we're not inside the linkage-specification here.
+    void cwg29_f5(); // ok, g5 is extern "C" but we're not inside the linkage-specification here.
   }
 
   extern "C++" void g6();
-  extern "C" void dr29_f6();
+  extern "C" void cwg29_f6();
   void g6() {
-    void dr29_f6(); // ok, g6 is extern "C" but we're not inside the linkage-specification here.
+    void cwg29_f6(); // ok, g6 is extern "C" but we're not inside the linkage-specification here.
   }
 
   extern "C" void g7();
-  extern "C++" void dr29_f7(); // #dr29-f7
+  extern "C++" void cwg29_f7(); // #cwg29-f7
   extern "C" void g7() {
-    void dr29_f7();
-    // expected-error@-1 {{declaration of 'dr29_f7' has a different language linkage}}
-    //   expected-note@#dr29-f7 {{previous declaration is here}}
+    void cwg29_f7();
+    // expected-error@-1 {{declaration of 'cwg29_f7' has a different language linkage}}
+    //   expected-note@#cwg29-f7 {{previous declaration is here}}
   }
 
   extern "C++" void g8();
-  extern "C" void dr29_f8(); // #dr29-f8
+  extern "C" void cwg29_f8(); // #cwg29-f8
   extern "C++" void g8() {
-    void dr29_f8();
-    // expected-error@-1 {{declaration of 'dr29_f8' has a different language linkage}}
-    //   expected-note@#dr29-f8 {{previous declaration is here}}
+    void cwg29_f8();
+    // expected-error@-1 {{declaration of 'cwg29_f8' has a different language linkage}}
+    //   expected-note@#cwg29-f8 {{previous declaration is here}}
   }
 }
 
-namespace dr30 { // dr30: sup 468 c++11
+namespace cwg30 { // cwg30: sup 468 c++11
   struct A {
     template<int> static int f();
   } a, *p = &a;
-  // FIXME: It's not clear whether DR468 applies to C++98 too.
+  // FIXME: It's not clear whether CWG468 applies to C++98 too.
   int x = A::template f<0>();
   // cxx98-error@-1 {{'template' keyword outside of a template}}
   int y = a.template f<0>();
@@ -450,29 +450,29 @@ namespace dr30 { // dr30: sup 468 c++11
   // cxx98-error@-1 {{'template' keyword outside of a template}}
 }
 
-namespace dr31 { // dr31: 2.8
+namespace cwg31 { // cwg31: 2.8
   class X {
   private:
-    void operator delete(void*); // #dr31-delete
+    void operator delete(void*); // #cwg31-delete
   };
   // We would call X::operator delete if X() threw (even though it can't,
   // and even though we allocated the X using ::operator delete).
   X *p = new X;
-  // expected-error@-1 {{'operator delete' is a private member of 'dr31::X'}}
-  //   expected-note@#dr31-delete {{declared private here}}
+  // expected-error@-1 {{'operator delete' is a private member of 'cwg31::X'}}
+  //   expected-note@#cwg31-delete {{declared private here}}
 }
 
-// dr32: na
+// cwg32: na
 
-namespace dr33 { // dr33: 9
-  namespace X { struct S; void f(void (*)(S)); } // #dr33-f-S
-  namespace Y { struct T; void f(void (*)(T)); } // #dr33-f-T
+namespace cwg33 { // cwg33: 9
+  namespace X { struct S; void f(void (*)(S)); } // #cwg33-f-S
+  namespace Y { struct T; void f(void (*)(T)); } // #cwg33-f-T
   void g(X::S);
   template<typename Z> Z g(Y::T);
   void h() { f(&g); }
   // expected-error@-1 {{call to 'f' is ambiguous}}
-  //   expected-note@#dr33-f-S {{candidate function}}
-  //   expected-note@#dr33-f-T {{candidate function}}
+  //   expected-note@#cwg33-f-S {{candidate function}}
+  //   expected-note@#cwg33-f-T {{candidate function}}
 
   template<typename T> void t(X::S);
   template<typename T, typename U = void> void u(X::S);
@@ -507,10 +507,10 @@ namespace dr33 { // dr33: 9
   }
 }
 
-// dr34: na
-// dr35: dup 178
+// cwg34: na
+// cwg35: dup 178
 
-namespace dr36 { // dr36: 2.8
+namespace cwg36 { // cwg36: 2.8
 namespace example1 {
   namespace A {
     int i;
@@ -540,25 +540,25 @@ namespace example2 {
 
   struct D : virtual B, virtual C
   {
-    using B::i; // #dr36-ex2-B-i-first
+    using B::i; // #cwg36-ex2-B-i-first
     using B::i;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex2-B-i-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex2-B-i-first {{previous using declaration}}
 
-    using C::i; // #dr36-ex2-C-i-first
+    using C::i; // #cwg36-ex2-C-i-first
     using C::i;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex2-C-i-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex2-C-i-first {{previous using declaration}}
 
-    using B::j; // #dr36-ex2-B-j-first
+    using B::j; // #cwg36-ex2-B-j-first
     using B::j;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex2-B-j-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex2-B-j-first {{previous using declaration}}
 
-    using C::j; // #dr36-ex2-C-j-first
+    using C::j; // #cwg36-ex2-C-j-first
     using C::j;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex2-C-j-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex2-C-j-first {{previous using declaration}}
   };
 }
 
@@ -578,25 +578,25 @@ namespace example3 {
   template<typename T>
   struct D : virtual B<T>, virtual C<T>
   {
-    using B<T>::i; // #dr36-ex3-B-i-first
+    using B<T>::i; // #cwg36-ex3-B-i-first
     using B<T>::i;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex3-B-i-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex3-B-i-first {{previous using declaration}}
 
-    using C<T>::i; // #dr36-ex3-C-i-first
+    using C<T>::i; // #cwg36-ex3-C-i-first
     using C<T>::i;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex3-C-i-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex3-C-i-first {{previous using declaration}}
 
-    using B<T>::j; // #dr36-ex3-B-j-first
+    using B<T>::j; // #cwg36-ex3-B-j-first
     using B<T>::j;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex3-B-j-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex3-B-j-first {{previous using declaration}}
 
-    using C<T>::j; // #dr36-ex3-C-j-first
+    using C<T>::j; // #cwg36-ex3-C-j-first
     using C<T>::j;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-ex3-C-j-first {{previous using declaration}}
+    //   expected-note@#cwg36-ex3-C-j-first {{previous using declaration}}
   };
 }
 namespace example4 {
@@ -607,23 +607,23 @@ namespace example4 {
 
   template<typename T>
   struct G : E<T> {
-    using E<T>::k; // #dr36-E-k-first
+    using E<T>::k; // #cwg36-E-k-first
     using E<T>::k;
     // expected-error@-1 {{redeclaration of using declaration}}
-    //   expected-note@#dr36-E-k-first {{previous using declaration}}
+    //   expected-note@#cwg36-E-k-first {{previous using declaration}}
   };
 }
 }
 
-// dr37: sup 475
+// cwg37: sup 475
 
-namespace dr38 { // dr38: yes
+namespace cwg38 { // cwg38: yes
   template<typename T> struct X {};
   template<typename T> X<T> operator+(X<T> a, X<T> b) { return a; }
   template X<int> operator+<int>(X<int>, X<int>);
 }
 
-namespace dr39 { // dr39: no
+namespace cwg39 { // cwg39: no
   namespace example1 {
     struct A { int &f(int); };
     struct B : A {
@@ -635,16 +635,16 @@ namespace dr39 { // dr39: no
 
   namespace example2 {
     struct A {
-      int &x(int); // #dr39-A-x-decl
-      static int &y(int); // #dr39-A-y-decl
+      int &x(int); // #cwg39-A-x-decl
+      static int &y(int); // #cwg39-A-y-decl
     };
     struct V {
       int &z(int);
     };
     struct B : A, virtual V {
-      using A::x; // #dr39-using-A-x
+      using A::x; // #cwg39-using-A-x
       float &x(float);
-      using A::y; // #dr39-using-A-y
+      using A::y; // #cwg39-using-A-y
       static float &y(float);
       using V::z;
       float &z(float);
@@ -652,18 +652,18 @@ namespace dr39 { // dr39: no
     struct C : A, B, virtual V {} c;
     /* expected-warning@-1
     {{direct base 'A' is inaccessible due to ambiguity:
-    struct dr39::example2::C -> A
-    struct dr39::example2::C -> B -> A}} */
+    struct cwg39::example2::C -> A
+    struct cwg39::example2::C -> B -> A}} */
     int &x = c.x(0);
     // expected-error@-1 {{member 'x' found in multiple base classes of different types}}
-    //   expected-note@#dr39-A-x-decl {{member found by ambiguous name lookup}}
-    //   expected-note@#dr39-using-A-x {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg39-A-x-decl {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg39-using-A-x {{member found by ambiguous name lookup}}
 
     // FIXME: This is valid, because we find the same static data member either way.
     int &y = c.y(0);
     // expected-error@-1 {{member 'y' found in multiple base classes of different types}}
-    //   expected-note@#dr39-A-y-decl {{member found by ambiguous name lookup}}
-    //   expected-note@#dr39-using-A-y {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg39-A-y-decl {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg39-using-A-y {{member found by ambiguous name lookup}}
     int &z = c.z(0);
   }
 
@@ -676,63 +676,63 @@ namespace dr39 { // dr39: no
   }
 
   namespace example4 {
-    struct A { int n; }; // #dr39-ex4-A-n
+    struct A { int n; }; // #cwg39-ex4-A-n
     struct B : A {};
     struct C : A {};
     struct D : B, C { int f() { return n; } };
     /* expected-error@-1
     {{non-static member 'n' found in multiple base-class subobjects of type 'A':
-    struct dr39::example4::D -> B -> A
-    struct dr39::example4::D -> C -> A}} */
-    //   expected-note@#dr39-ex4-A-n {{member found by ambiguous name lookup}}
+    struct cwg39::example4::D -> B -> A
+    struct cwg39::example4::D -> C -> A}} */
+    //   expected-note@#cwg39-ex4-A-n {{member found by ambiguous name lookup}}
   }
 
   namespace PR5916 {
     // FIXME: This is valid.
-    struct A { int n; }; // #dr39-A-n
+    struct A { int n; }; // #cwg39-A-n
     struct B : A {};
     struct C : A {};
     struct D : B, C {};
-    int k = sizeof(D::n); // #dr39-sizeof
-    /* expected-error@#dr39-sizeof
+    int k = sizeof(D::n); // #cwg39-sizeof
+    /* expected-error@#cwg39-sizeof
     {{non-static member 'n' found in multiple base-class subobjects of type 'A':
-    struct dr39::PR5916::D -> B -> A
-    struct dr39::PR5916::D -> C -> A}} */
-    //   expected-note@#dr39-A-n {{member found by ambiguous name lookup}}
+    struct cwg39::PR5916::D -> B -> A
+    struct cwg39::PR5916::D -> C -> A}} */
+    //   expected-note@#cwg39-A-n {{member found by ambiguous name lookup}}
 
-    // expected-error@#dr39-sizeof {{unknown type name}}
+    // expected-error@#cwg39-sizeof {{unknown type name}}
 #if __cplusplus >= 201103L
     decltype(D::n) n;
     /* expected-error@-1
     {{non-static member 'n' found in multiple base-class subobjects of type 'A':
-    struct dr39::PR5916::D -> B -> A
-    struct dr39::PR5916::D -> C -> A}} */
-    //   expected-note@#dr39-A-n {{member found by ambiguous name lookup}}
+    struct cwg39::PR5916::D -> B -> A
+    struct cwg39::PR5916::D -> C -> A}} */
+    //   expected-note@#cwg39-A-n {{member found by ambiguous name lookup}}
 #endif
   }
 }
 
-// dr40: na
+// cwg40: na
 
-namespace dr41 { // dr41: yes
+namespace cwg41 { // cwg41: yes
   struct S f(S);
 }
 
-namespace dr42 { // dr42: yes
+namespace cwg42 { // cwg42: yes
   struct A { static const int k = 0; };
   struct B : A { static const int k = A::k; };
 }
 
-// dr43: na
+// cwg43: na
 
-namespace dr44 { // dr44: sup 727
+namespace cwg44 { // cwg44: sup 727
   struct A {
     template<int> void f();
     template<> void f<0>();
   };
 }
 
-namespace dr45 { // dr45: yes
+namespace cwg45 { // cwg45: yes
   class A {
     class B {};
     class C : B {};
@@ -740,27 +740,27 @@ namespace dr45 { // dr45: yes
   };
 }
 
-namespace dr46 { // dr46: yes
+namespace cwg46 { // cwg46: yes
   template<typename> struct A { template<typename> struct B {}; };
   template template struct A<int>::B<int>;
   // expected-error@-1 {{expected unqualified-id}}
 }
 
-namespace dr47 { // dr47: sup 329
+namespace cwg47 { // cwg47: sup 329
   template<typename T> struct A {
-    friend void f() { T t; } // #dr47-f
+    friend void f() { T t; } // #cwg47-f
     // expected-error@-1 {{redefinition of 'f'}}
-    //   expected-note@#dr47-b {{in instantiation of template class 'dr47::A<float>' requested here}}
-    //   expected-note@#dr47-f {{previous definition is here}}
+    //   expected-note@#cwg47-b {{in instantiation of template class 'cwg47::A<float>' requested here}}
+    //   expected-note@#cwg47-f {{previous definition is here}}
   };
   A<int> a;
-  A<float> b; // #dr47-b
+  A<float> b; // #cwg47-b
 
   void f();
   void g() { f(); }
 }
 
-namespace dr48 { // dr48: yes
+namespace cwg48 { // cwg48: yes
   namespace {
     struct S {
       static const int m = 0;
@@ -776,45 +776,45 @@ namespace dr48 { // dr48: yes
   const int &c = S::o;
 }
 
-namespace dr49 { // dr49: 2.8
-  template<int*> struct A {}; // #dr49-A
+namespace cwg49 { // cwg49: 2.8
+  template<int*> struct A {}; // #cwg49-A
   int k;
 #if __has_feature(cxx_constexpr)
   constexpr
 #endif
-  int *const p = &k; // #dr49-p
+  int *const p = &k; // #cwg49-p
   A<&k> a;
-  A<p> b; // #dr49-b
-  // cxx98-error@#dr49-b {{non-type template argument referring to object 'p' with internal linkage is a C++11 extension}}
-  //   cxx98-note@#dr49-p {{non-type template argument refers to object here}}
-  // cxx98-14-error@#dr49-b {{non-type template argument for template parameter of pointer type 'int *' must have its address taken}}
-  //   cxx98-14-note@#dr49-A {{template parameter is declared here}}
-  int *q = &k; // #dr49-q
-  A<q> c; // #dr49-c
-  // cxx98-error@#dr49-c {{non-type template argument for template parameter of pointer type 'int *' must have its address taken}}
-  //   cxx98-note@#dr49-A {{template parameter is declared here}}
-  // cxx11-14-error@#dr49-c {{non-type template argument of type 'int *' is not a constant expression}}
-  //   cxx11-14-note@#dr49-c {{read of non-constexpr variable 'q' is not allowed in a constant expression}}
-  //   cxx11-14-note@#dr49-q {{declared here}}
-  //   cxx11-14-note@#dr49-A {{template parameter is declared here}}
-  // since-cxx17-error@#dr49-c {{non-type template argument is not a constant expression}}
-  //   since-cxx17-note@#dr49-c {{read of non-constexpr variable 'q' is not allowed in a constant expression}}
-  //   since-cxx17-note@#dr49-q {{declared here}}
-}
-
-namespace dr50 { // dr50: yes
-  struct X; // #dr50-X
+  A<p> b; // #cwg49-b
+  // cxx98-error@#cwg49-b {{non-type template argument referring to object 'p' with internal linkage is a C++11 extension}}
+  //   cxx98-note@#cwg49-p {{non-type template argument refers to object here}}
+  // cxx98-14-error@#cwg49-b {{non-type template argument for template parameter of pointer type 'int *' must have its address taken}}
+  //   cxx98-14-note@#cwg49-A {{template parameter is declared here}}
+  int *q = &k; // #cwg49-q
+  A<q> c; // #cwg49-c
+  // cxx98-error@#cwg49-c {{non-type template argument for template parameter of pointer type 'int *' must have its address taken}}
+  //   cxx98-note@#cwg49-A {{template parameter is declared here}}
+  // cxx11-14-error@#cwg49-c {{non-type template argument of type 'int *' is not a constant expression}}
+  //   cxx11-14-note@#cwg49-c {{read of non-constexpr variable 'q' is not allowed in a constant expression}}
+  //   cxx11-14-note@#cwg49-q {{declared here}}
+  //   cxx11-14-note@#cwg49-A {{template parameter is declared here}}
+  // since-cxx17-error@#cwg49-c {{non-type template argument is not a constant expression}}
+  //   since-cxx17-note@#cwg49-c {{read of non-constexpr variable 'q' is not allowed in a constant expression}}
+  //   since-cxx17-note@#cwg49-q {{declared here}}
+}
+
+namespace cwg50 { // cwg50: yes
+  struct X; // #cwg50-X
   extern X *p;
   X *q = (X*)p;
   X *r = static_cast<X*>(p);
   X *s = const_cast<X*>(p);
   X *t = reinterpret_cast<X*>(p);
   X *u = dynamic_cast<X*>(p);
-  // expected-error@-1 {{'dr50::X' is an incomplete type}}
-  //   expected-note@#dr50-X {{forward declaration of 'dr50::X'}}
+  // expected-error@-1 {{'cwg50::X' is an incomplete type}}
+  //   expected-note@#cwg50-X {{forward declaration of 'cwg50::X'}}
 }
 
-namespace dr51 { // dr51: 2.8
+namespace cwg51 { // cwg51: 2.8
   struct A {};
   struct B : A {};
   struct S {
@@ -824,57 +824,57 @@ namespace dr51 { // dr51: 2.8
   A &a = s;
 }
 
-namespace dr52 { // dr52: 2.8
-  struct A { int n; }; // #dr52-A
-  struct B : private A {} b; // #dr52-B
-  int k = b.A::n; // #dr52-k
+namespace cwg52 { // cwg52: 2.8
+  struct A { int n; }; // #cwg52-A
+  struct B : private A {} b; // #cwg52-B
+  int k = b.A::n; // #cwg52-k
   // FIXME: This first diagnostic is very strangely worded, and seems to be bogus.
-  // expected-error@#dr52-k {{'A' is a private member of 'dr52::A'}}
-  //   expected-note@#dr52-B {{constrained by private inheritance here}}
-  //   expected-note@#dr52-A {{member is declared here}}
-  // expected-error@#dr52-k {{cannot cast 'struct B' to its private base class 'dr52::A'}}
-  //   expected-note@#dr52-B {{declared private here}}
+  // expected-error@#cwg52-k {{'A' is a private member of 'cwg52::A'}}
+  //   expected-note@#cwg52-B {{constrained by private inheritance here}}
+  //   expected-note@#cwg52-A {{member is declared here}}
+  // expected-error@#cwg52-k {{cannot cast 'struct B' to its private base class 'cwg52::A'}}
+  //   expected-note@#cwg52-B {{declared private here}}
 }
 
-namespace dr53 { // dr53: yes
+namespace cwg53 { // cwg53: yes
   int n = 0;
   enum E { e } x = static_cast<E>(n);
 }
 
-namespace dr54 { // dr54: 2.8
+namespace cwg54 { // cwg54: 2.8
   struct A { int a; } a;
   struct V { int v; } v;
-  struct B : private A, virtual V { int b; } b; // #dr54-B
+  struct B : private A, virtual V { int b; } b; // #cwg54-B
 
   A &sab = static_cast<A&>(b);
   // expected-error@-1 {{cannot cast 'struct B' to its private base class 'A'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  //   expected-note@#cwg54-B {{declared private here}}
   A *spab = static_cast<A*>(&b);
   // expected-error@-1 {{cannot cast 'struct B' to its private base class 'A'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  //   expected-note@#cwg54-B {{declared private here}}
   int A::*smab = static_cast<int A::*>(&B::b);
-  // expected-error@-1 {{cannot cast 'dr54::B' to its private base class 'dr54::A'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  // expected-error@-1 {{cannot cast 'cwg54::B' to its private base class 'cwg54::A'}}
+  //   expected-note@#cwg54-B {{declared private here}}
   B &sba = static_cast<B&>(a);
-  // expected-error@-1 {{cannot cast private base class 'dr54::A' to 'dr54::B'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  // expected-error@-1 {{cannot cast private base class 'cwg54::A' to 'cwg54::B'}}
+  //   expected-note@#cwg54-B {{declared private here}}
   B *spba = static_cast<B*>(&a);
-  // expected-error@-1 {{cannot cast private base class 'dr54::A' to 'dr54::B'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  // expected-error@-1 {{cannot cast private base class 'cwg54::A' to 'cwg54::B'}}
+  //   expected-note@#cwg54-B {{declared private here}}
   int B::*smba = static_cast<int B::*>(&A::a);
-  // expected-error@-1 {{cannot cast private base class 'dr54::A' to 'dr54::B'}}
-  //   expected-note@#dr54-B {{declared private here}}
+  // expected-error@-1 {{cannot cast private base class 'cwg54::A' to 'cwg54::B'}}
+  //   expected-note@#cwg54-B {{declared private here}}
 
   V &svb = static_cast<V&>(b);
   V *spvb = static_cast<V*>(&b);
   int V::*smvb = static_cast<int V::*>(&B::b);
-  // expected-error@-1 {{conversion from pointer to member of class 'dr54::B' to pointer to member of class 'dr54::V' via virtual base 'dr54::V' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg54::B' to pointer to member of class 'cwg54::V' via virtual base 'cwg54::V' is not allowed}}
   B &sbv = static_cast<B&>(v);
-  // expected-error@-1 {{cannot cast 'struct V' to 'B &' via virtual base 'dr54::V'}}
+  // expected-error@-1 {{cannot cast 'struct V' to 'B &' via virtual base 'cwg54::V'}}
   B *spbv = static_cast<B*>(&v);
-  // expected-error@-1 {{cannot cast 'dr54::V *' to 'B *' via virtual base 'dr54::V'}}
+  // expected-error@-1 {{cannot cast 'cwg54::V *' to 'B *' via virtual base 'cwg54::V'}}
   int B::*smbv = static_cast<int B::*>(&V::v);
-  // expected-error@-1 {{conversion from pointer to member of class 'dr54::V' to pointer to member of class 'dr54::B' via virtual base 'dr54::V' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg54::V' to pointer to member of class 'cwg54::B' via virtual base 'cwg54::V' is not allowed}}
 
   A &cab = (A&)(b);
   A *cpab = (A*)(&b);
@@ -886,37 +886,37 @@ namespace dr54 { // dr54: 2.8
   V &cvb = (V&)(b);
   V *cpvb = (V*)(&b);
   int V::*cmvb = (int V::*)(&B::b);
-  // expected-error@-1 {{conversion from pointer to member of class 'dr54::B' to pointer to member of class 'dr54::V' via virtual base 'dr54::V' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg54::B' to pointer to member of class 'cwg54::V' via virtual base 'cwg54::V' is not allowed}}
   B &cbv = (B&)(v);
-  // expected-error@-1 {{cannot cast 'struct V' to 'B &' via virtual base 'dr54::V'}}
+  // expected-error@-1 {{cannot cast 'struct V' to 'B &' via virtual base 'cwg54::V'}}
   B *cpbv = (B*)(&v);
-  // expected-error@-1 {{cannot cast 'dr54::V *' to 'B *' via virtual base 'dr54::V'}}
+  // expected-error@-1 {{cannot cast 'cwg54::V *' to 'B *' via virtual base 'cwg54::V'}}
   int B::*cmbv = (int B::*)(&V::v);
-  // expected-error@-1 {{conversion from pointer to member of class 'dr54::V' to pointer to member of class 'dr54::B' via virtual base 'dr54::V' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg54::V' to pointer to member of class 'cwg54::B' via virtual base 'cwg54::V' is not allowed}}
 }
 
-namespace dr55 { // dr55: yes
+namespace cwg55 { // cwg55: yes
   enum E { e = 5 };
   int test[(e + 1 == 6) ? 1 : -1];
 }
 
-namespace dr56 { // dr56: yes
+namespace cwg56 { // cwg56: yes
   struct A {
-    typedef int T; // #dr56-typedef-int-T-first
+    typedef int T; // #cwg56-typedef-int-T-first
     typedef int T;
     // expected-error@-1 {{redefinition of 'T'}}
-    //   expected-note@#dr56-typedef-int-T-first {{previous definition is here}}
+    //   expected-note@#cwg56-typedef-int-T-first {{previous definition is here}}
   };
   struct B {
     struct X;
-    typedef X X; // #dr56-typedef-X-X-first
+    typedef X X; // #cwg56-typedef-X-X-first
     typedef X X;
     // expected-error@-1 {{redefinition of 'X'}}
-    //   expected-note@#dr56-typedef-X-X-first {{previous definition is here}}
+    //   expected-note@#cwg56-typedef-X-X-first {{previous definition is here}}
   };
 }
 
-namespace dr58 { // dr58: 3.1
+namespace cwg58 { // cwg58: 3.1
   // FIXME: Ideally, we should have a CodeGen test for this.
 #if __cplusplus >= 201103L
   enum E1 { E1_0 = 0, E1_1 = 1 };
@@ -927,54 +927,54 @@ namespace dr58 { // dr58: 3.1
 #endif
 }
 
-namespace dr59 { // dr59: yes
+namespace cwg59 { // cwg59: yes
 #pragma clang diagnostic push
 #pragma clang diagnostic ignored "-Wdeprecated-volatile"
   template<typename T> struct convert_to { operator T() const; };
-  struct A {}; // #dr59-A
-  struct B : A {}; // #dr59-B
+  struct A {}; // #cwg59-A
+  struct B : A {}; // #cwg59-B
 
   A a1 = convert_to<A>();
   A a2 = convert_to<A&>();
   A a3 = convert_to<const A>();
   A a4 = convert_to<const volatile A>();
-  // cxx98-14-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::A'}}
-  //   cxx98-14-note@#dr59-A {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile dr59::A') would lose volatile qualifier}}
-  //   cxx11-14-note@#dr59-A {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile dr59::A') would lose const and volatile qualifiers}}
-  //   cxx98-14-note@#dr59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // cxx98-14-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::A'}}
+  //   cxx98-14-note@#cwg59-A {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile cwg59::A') would lose volatile qualifier}}
+  //   cxx11-14-note@#cwg59-A {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile cwg59::A') would lose const and volatile qualifiers}}
+  //   cxx98-14-note@#cwg59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
   A a5 = convert_to<const volatile A&>();
-  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::A'}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile dr59::A') would lose volatile qualifier}}
-  //   since-cxx11-note@#dr59-A {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile dr59::A') would lose const and volatile qualifiers}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::A'}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile cwg59::A') would lose volatile qualifier}}
+  //   since-cxx11-note@#cwg59-A {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile cwg59::A') would lose const and volatile qualifiers}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
 
   B b1 = convert_to<B>();
   B b2 = convert_to<B&>();
   B b3 = convert_to<const B>();
   B b4 = convert_to<const volatile B>();
-  // cxx98-14-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::B'}}
-  //   cxx98-14-note@#dr59-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile dr59::B') would lose volatile qualifier}}
-  //   cxx11-14-note@#dr59-B {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile dr59::B') would lose const and volatile qualifiers}}
-  //   cxx98-14-note@#dr59-B {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // cxx98-14-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::B'}}
+  //   cxx98-14-note@#cwg59-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile cwg59::B') would lose volatile qualifier}}
+  //   cxx11-14-note@#cwg59-B {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile cwg59::B') would lose const and volatile qualifiers}}
+  //   cxx98-14-note@#cwg59-B {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
   B b5 = convert_to<const volatile B&>();
-  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::B'}}
-  //   expected-note@#dr59-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile dr59::B') would lose volatile qualifier}}
-  //   since-cxx11-note@#dr59-B {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile dr59::B') would lose const and volatile qualifiers}}
-  //   expected-note@#dr59-B {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::B'}}
+  //   expected-note@#cwg59-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const volatile cwg59::B') would lose volatile qualifier}}
+  //   since-cxx11-note@#cwg59-B {{candidate constructor (the implicit move constructor) not viable: 1st argument ('const volatile cwg59::B') would lose const and volatile qualifiers}}
+  //   expected-note@#cwg59-B {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
 
   A c1 = convert_to<B>();
   A c2 = convert_to<B&>();
   A c3 = convert_to<const B>();
   A c4 = convert_to<const volatile B>();
-  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::B'}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'const volatile dr59::B' to 'const A &' for 1st argument}}
-  //   since-cxx11-note@#dr59-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'const volatile dr59::B' to 'A &&' for 1st argument}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::B'}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'const volatile cwg59::B' to 'const A &' for 1st argument}}
+  //   since-cxx11-note@#cwg59-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'const volatile cwg59::B' to 'A &&' for 1st argument}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
   A c5 = convert_to<const volatile B&>();
-  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile dr59::B'}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'const volatile dr59::B' to 'const A &' for 1st argument}}
-  //   since-cxx11-note@#dr59-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'const volatile dr59::B' to 'A &&' for 1st argument}}
-  //   expected-note@#dr59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+  // expected-error@-1 {{no viable constructor copying variable of type 'const volatile cwg59::B'}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'const volatile cwg59::B' to 'const A &' for 1st argument}}
+  //   since-cxx11-note@#cwg59-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'const volatile cwg59::B' to 'A &&' for 1st argument}}
+  //   expected-note@#cwg59-A {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
 
   int n1 = convert_to<int>();
   int n2 = convert_to<int&>();
@@ -984,14 +984,14 @@ namespace dr59 { // dr59: yes
 #pragma clang diagnostic pop
 }
 
-namespace dr60 { // dr60: yes
+namespace cwg60 { // cwg60: yes
   void f(int &);
   int &f(...);
   const int k = 0;
   int &n = f(k);
 }
 
-namespace dr61 { // dr61: 3.4
+namespace cwg61 { // cwg61: 3.4
   struct X {
     static void f();
   } x;
@@ -1008,7 +1008,7 @@ namespace dr61 { // dr61: 3.4
   // expected-error@-1 {{cannot create a non-constant pointer to member function}}
 }
 
-namespace dr62 { // dr62: 2.9
+namespace cwg62 { // cwg62: 2.9
   struct A {
     struct { int n; } b;
   };
@@ -1019,7 +1019,7 @@ namespace dr62 { // dr62: 2.9
   X<A> x1;
   A a = get<A>();
 
-  typedef struct { } *NoNameForLinkagePtr; // #dr62-unnamed
+  typedef struct { } *NoNameForLinkagePtr; // #cwg62-unnamed
   NoNameForLinkagePtr noNameForLinkagePtr;
 
   struct Danger {
@@ -1028,19 +1028,19 @@ namespace dr62 { // dr62: 2.9
 
   X<NoNameForLinkagePtr> x2;
   // cxx98-error@-1 {{template argument uses unnamed type}}
-  //   cxx98-note@#dr62-unnamed {{unnamed type used in template argument was declared here}}
+  //   cxx98-note@#cwg62-unnamed {{unnamed type used in template argument was declared here}}
   X<const NoNameForLinkagePtr> x3;
   // cxx98-error@-1 {{template argument uses unnamed type}}
-  //   cxx98-note@#dr62-unnamed {{unnamed type used in template argument was declared here}}
+  //   cxx98-note@#cwg62-unnamed {{unnamed type used in template argument was declared here}}
   NoNameForLinkagePtr p1 = get<NoNameForLinkagePtr>();
   // cxx98-error@-1 {{template argument uses unnamed type}}
-  //   cxx98-note@#dr62-unnamed {{unnamed type used in template argument was declared here}}
+  //   cxx98-note@#cwg62-unnamed {{unnamed type used in template argument was declared here}}
   NoNameForLinkagePtr p2 = get<const NoNameForLinkagePtr>();
   // cxx98-error@-1 {{template argument uses unnamed type}}
-  //   cxx98-note@#dr62-unnamed {{unnamed type used in template argument was declared here}}
+  //   cxx98-note@#cwg62-unnamed {{unnamed type used in template argument was declared here}}
   int n1 = take(noNameForLinkagePtr);
   // cxx98-error@-1 {{template argument uses unnamed type}}
-  //   cxx98-note@#dr62-unnamed {{unnamed type used in template argument was declared here}}
+  //   cxx98-note@#cwg62-unnamed {{unnamed type used in template argument was declared here}}
 
   X<Danger> x4;
 
@@ -1061,13 +1061,13 @@ namespace dr62 { // dr62: 2.9
   }
 }
 
-namespace dr63 { // dr63: yes
+namespace cwg63 { // cwg63: yes
   template<typename T> struct S { typename T::error e; };
   extern S<int> *p;
   void *q = p;
 }
 
-namespace dr64 { // dr64: yes
+namespace cwg64 { // cwg64: yes
   template<class T> void f(T);
   template<class T> void f(T*);
   template<> void f(int*);
@@ -1075,11 +1075,11 @@ namespace dr64 { // dr64: yes
   template<> void f(int);
 }
 
-// dr65: na
+// cwg65: na
 
-namespace dr66 { // dr66: no
+namespace cwg66 { // cwg66: no
   namespace X {
-    int f(int n); // #dr66-f-first
+    int f(int n); // #cwg66-f-first
   }
   using X::f;
   namespace X {
@@ -1089,36 +1089,36 @@ namespace dr66 { // dr66: no
   // FIXME: The first two calls here should be accepted.
   int a = f();
   // expected-error@-1 {{no matching function for call to 'f'}}
-  //   expected-note@#dr66-f-first {{candidate function not viable: requires single argument 'n', but no arguments were provided}}
+  //   expected-note@#cwg66-f-first {{candidate function not viable: requires single argument 'n', but no arguments were provided}}
   int b = f(1);
   int c = f(1, 2);
   // expected-error@-1 {{no matching function for call to 'f'}}
-  //   expected-note@#dr66-f-first {{candidate function not viable: requires single argument 'n', but 2 arguments were provided}}
+  //   expected-note@#cwg66-f-first {{candidate function not viable: requires single argument 'n', but 2 arguments were provided}}
 }
 
-// dr67: na
+// cwg67: na
 
-namespace dr68 { // dr68: 2.8
+namespace cwg68 { // cwg68: 2.8
   template<typename T> struct X {};
-  struct ::dr68::X<int> x1;
-  struct ::dr68::template X<int> x2;
+  struct ::cwg68::X<int> x1;
+  struct ::cwg68::template X<int> x2;
   // cxx98-error@-1 {{'template' keyword outside of a template}}
   struct Y {
     friend struct X<int>;
-    friend struct ::dr68::X<char>;
-    friend struct ::dr68::template X<double>;
+    friend struct ::cwg68::X<char>;
+    friend struct ::cwg68::template X<double>;
     // cxx98-error@-1 {{'template' keyword outside of a template}}
   };
   template<typename>
   struct Z {
-    friend struct ::dr68::template X<double>;
-    friend typename ::dr68::X<double>;
-    // cxx98-error@-1 {{unelaborated friend declaration is a C++11 extension; specify 'struct' to befriend 'typename ::dr68::X<double>'}}
+    friend struct ::cwg68::template X<double>;
+    friend typename ::cwg68::X<double>;
+    // cxx98-error@-1 {{unelaborated friend declaration is a C++11 extension; specify 'struct' to befriend 'typename ::cwg68::X<double>'}}
   };
 }
 
-namespace dr69 { // dr69: 9
-  template<typename T> static void f() {} // #dr69-f
+namespace cwg69 { // cwg69: 9
+  template<typename T> static void f() {} // #cwg69-f
   // FIXME: Should we warn here?
   inline void g() { f<int>(); }
   extern template void f<char>();
@@ -1127,21 +1127,21 @@ namespace dr69 { // dr69: 9
   template<void(*)()> struct Q {};
   Q<&f<int> > q;
   // cxx98-error@-1 {{non-type template argument referring to function 'f<int>' with internal linkage is a C++11 extension}}
-  //   cxx98-note@#dr69-f {{non-type template argument refers to function here}}
+  //   cxx98-note@#cwg69-f {{non-type template argument refers to function here}}
 }
 
-namespace dr70 { // dr70: yes
+namespace cwg70 { // cwg70: yes
   template<int> struct A {};
   template<int I, int J> int f(int (&)[I + J], A<I>, A<J>);
   int arr[7];
   int k = f(arr, A<3>(), A<4>());
 }
 
-// dr71: na
-// dr72: dup 69
+// cwg71: na
+// cwg72: dup 69
 
 #if __cplusplus >= 201103L
-namespace dr73 { // dr73: sup 1652
+namespace cwg73 { // cwg73: sup 1652
   int a, b;
   static_assert(&a + 1 != &b, "");
   // expected-error@-1 {{static assertion expression is not an integral constant expression}}
@@ -1149,43 +1149,43 @@ namespace dr73 { // dr73: sup 1652
 }
 #endif
 
-namespace dr74 { // dr74: yes
+namespace cwg74 { // cwg74: yes
   enum E { k = 5 };
   int (*p)[k] = new int[k][k];
 }
 
-namespace dr75 { // dr75: yes
+namespace cwg75 { // cwg75: yes
   struct S {
     static int n = 0;
     // expected-error@-1 {{non-const static data member must be initialized out of line}}
   };
 }
 
-namespace dr76 { // dr76: yes
+namespace cwg76 { // cwg76: yes
   const volatile int n = 1;
-  int arr[n]; // #dr76-vla
-  // expected-error@#dr76-vla {{variable length arrays in C++ are a Clang extension}}
-  //   expected-note@#dr76-vla {{read of volatile-qualified type 'const volatile int' is not allowed in a constant expression}}
-  // expected-error@#dr76-vla {{variable length array declaration not allowed at file scope}}
+  int arr[n]; // #cwg76-vla
+  // expected-error@#cwg76-vla {{variable length arrays in C++ are a Clang extension}}
+  //   expected-note@#cwg76-vla {{read of volatile-qualified type 'const volatile int' is not allowed in a constant expression}}
+  // expected-error@#cwg76-vla {{variable length array declaration not allowed at file scope}}
 }
 
-namespace dr77 { // dr77: yes
+namespace cwg77 { // cwg77: yes
   struct A {
     struct B {};
     friend struct B;
   };
 }
 
-namespace dr78 { // dr78: sup ????
-  // Under DR78, this is valid, because 'k' has static storage duration, so is
+namespace cwg78 { // cwg78: sup ????
+  // Under CWG78, this is valid, because 'k' has static storage duration, so is
   // zero-initialized.
   const int k;
   // expected-error@-1 {{default initialization of an object of const type 'const int'}}
 }
 
-// dr79: na
+// cwg79: na
 
-namespace dr80 { // dr80: 2.9
+namespace cwg80 { // cwg80: 2.9
   struct A {
     int A;
   };
@@ -1205,22 +1205,22 @@ namespace dr80 { // dr80: 2.9
   };
 }
 
-// dr81: na
-// dr82: dup 48
+// cwg81: na
+// cwg82: dup 48
 
-namespace dr83 { // dr83: yes
+namespace cwg83 { // cwg83: yes
   int &f(const char*);
   char &f(char *);
   int &k = f("foo");
 }
 
-namespace dr84 { // dr84: yes
+namespace cwg84 { // cwg84: yes
   struct B;
   struct A { operator B() const; };
   struct C {};
   struct B {
-    B(B&); // #dr84-copy-ctor
-    B(C); // #dr84-ctor-from-C
+    B(B&); // #cwg84-copy-ctor
+    B(C); // #cwg84-ctor-from-C
     operator C() const;
   };
   A a;
@@ -1228,78 +1228,78 @@ namespace dr84 { // dr84: yes
   // here. In C++17, we initialize the B object directly using 'A::operator B()'.
   B b = a;
   // cxx98-14-error@-1 {{no viable constructor copying variable of type 'B'}}
-  //   cxx98-14-note@#dr84-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
-  //   cxx98-14-note@#dr84-ctor-from-C {{candidate constructor not viable: no known conversion from 'B' to 'C' for 1st argument}}
+  //   cxx98-14-note@#cwg84-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
+  //   cxx98-14-note@#cwg84-ctor-from-C {{candidate constructor not viable: no known conversion from 'B' to 'C' for 1st argument}}
 }
 
-namespace dr85 { // dr85: 3.4
+namespace cwg85 { // cwg85: 3.4
   struct A {
     struct B;
-    struct B {}; // #dr85-B-def
+    struct B {}; // #cwg85-B-def
     struct B;
     // expected-error@-1 {{class member cannot be redeclared}}
-    //   expected-note@#dr85-B-def {{previous declaration is here}}
+    //   expected-note@#cwg85-B-def {{previous declaration is here}}
 
     union U;
-    union U {}; // #dr85-U-def
+    union U {}; // #cwg85-U-def
     union U;
     // expected-error@-1 {{class member cannot be redeclared}}
-    //   expected-note@#dr85-U-def {{previous declaration is here}}
+    //   expected-note@#cwg85-U-def {{previous declaration is here}}
 
 #if __cplusplus >= 201103L
     enum E1 : int;
-    enum E1 : int { e1 }; // #dr85-E1-def
+    enum E1 : int { e1 }; // #cwg85-E1-def
     enum E1 : int;
     // expected-error@-1 {{class member cannot be redeclared}}
-    //   expected-note@#dr85-E1-def {{previous declaration is here}}
+    //   expected-note@#cwg85-E1-def {{previous declaration is here}}
 
     enum class E2;
-    enum class E2 { e2 }; // #dr85-E2-def
+    enum class E2 { e2 }; // #cwg85-E2-def
     enum class E2;
     // expected-error@-1 {{class member cannot be redeclared}}
-    //   expected-note@#dr85-E2-def {{previous declaration is here}}
+    //   expected-note@#cwg85-E2-def {{previous declaration is here}}
 #endif
   };
 
   template <typename T>
   struct C {
-    struct B {}; // #dr85-C-B-def
+    struct B {}; // #cwg85-C-B-def
     struct B;
     // expected-error@-1 {{class member cannot be redeclared}}
-    //   expected-note@#dr85-C-B-def {{previous declaration is here}}
+    //   expected-note@#cwg85-C-B-def {{previous declaration is here}}
   };
 }
 
-// dr86: dup 446
+// cwg86: dup 446
 
-namespace dr87 { // dr87: no
-  // FIXME: Superseded by dr1975
+namespace cwg87 { // cwg87: no
+  // FIXME: Superseded by cwg1975
   template<typename T> struct X {};
   // FIXME: This is invalid.
   X<void() throw()> x;
-  // This is valid under dr87 but not under dr1975.
+  // This is valid under cwg87 but not under cwg1975.
   X<void(void() throw())> y;
 }
 
-namespace dr88 { // dr88: 2.8
+namespace cwg88 { // cwg88: 2.8
   template<typename T> struct S {
-    static const int a = 1; // #dr88-a
+    static const int a = 1; // #cwg88-a
     static const int b;
   };
   template<> const int S<int>::a = 4;
   // expected-error@-1 {{static data member 'a' already has an initializer}}
-  //   expected-note@#dr88-a {{previous initialization is here}}
+  //   expected-note@#cwg88-a {{previous initialization is here}}
   template<> const int S<int>::b = 4;
 }
 
-// dr89: na
+// cwg89: na
 
-namespace dr90 { // dr90: yes
+namespace cwg90 { // cwg90: yes
   struct A {
-    template<typename T> friend void dr90_f(T);
+    template<typename T> friend void cwg90_f(T);
   };
   struct B : A {
-    template<typename T> friend void dr90_g(T);
+    template<typename T> friend void cwg90_g(T);
     struct C {};
     union D {};
   };
@@ -1307,41 +1307,41 @@ namespace dr90 { // dr90: yes
   struct F : B::C {};
 
   void test() {
-    dr90_f(A());
-    dr90_f(B());
-    dr90_f(B::C());
-    // expected-error@-1 {{use of undeclared identifier 'dr90_f'}}
-    dr90_f(B::D());
-    // expected-error@-1 {{use of undeclared identifier 'dr90_f'}}
-    dr90_f(E());
-    dr90_f(F());
-    // expected-error@-1 {{use of undeclared identifier 'dr90_f'}}
-
-    dr90_g(A());
-    // expected-error@-1 {{use of undeclared identifier 'dr90_g'}}
-    dr90_g(B());
-    dr90_g(B::C());
-    dr90_g(B::D());
-    dr90_g(E());
-    dr90_g(F());
-    // expected-error@-1 {{use of undeclared identifier 'dr90_g'}}
+    cwg90_f(A());
+    cwg90_f(B());
+    cwg90_f(B::C());
+    // expected-error@-1 {{use of undeclared identifier 'cwg90_f'}}
+    cwg90_f(B::D());
+    // expected-error@-1 {{use of undeclared identifier 'cwg90_f'}}
+    cwg90_f(E());
+    cwg90_f(F());
+    // expected-error@-1 {{use of undeclared identifier 'cwg90_f'}}
+
+    cwg90_g(A());
+    // expected-error@-1 {{use of undeclared identifier 'cwg90_g'}}
+    cwg90_g(B());
+    cwg90_g(B::C());
+    cwg90_g(B::D());
+    cwg90_g(E());
+    cwg90_g(F());
+    // expected-error@-1 {{use of undeclared identifier 'cwg90_g'}}
   }
 }
 
-namespace dr91 { // dr91: yes
+namespace cwg91 { // cwg91: yes
   union U { friend int f(U); };
   int k = f(U());
 }
 
-namespace dr92 { // dr92: 4 c++17
+namespace cwg92 { // cwg92: 4 c++17
   void f() throw(int, float);
   // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
   //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
-  void (*p)() throw(int) = &f; // #dr92-p
-  // since-cxx17-error@#dr92-p {{ISO C++17 does not allow dynamic exception specifications}}
-  //   since-cxx17-note@#dr92-p {{use 'noexcept(false)' instead}}
-  // cxx98-14-error@#dr92-p {{target exception specification is not superset of source}}
-  // since-cxx17-warning@#dr92-p {{target exception specification is not superset of source}}
+  void (*p)() throw(int) = &f; // #cwg92-p
+  // since-cxx17-error@#cwg92-p {{ISO C++17 does not allow dynamic exception specifications}}
+  //   since-cxx17-note@#cwg92-p {{use 'noexcept(false)' instead}}
+  // cxx98-14-error@#cwg92-p {{target exception specification is not superset of source}}
+  // since-cxx17-warning@#cwg92-p {{target exception specification is not superset of source}}
   void (*q)() throw(int);
   // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
   //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
@@ -1349,17 +1349,17 @@ namespace dr92 { // dr92: 4 c++17
   // cxx98-14-error@-1 {{exception specifications are not allowed beyond a single level of indirection}}
   // since-cxx17-error@-2 {{cannot initialize a variable of type 'void (**)() throw()' with an rvalue of type 'void (**)() throw(int)'}}
 
-  void g(void() throw()); // #dr92-g
+  void g(void() throw()); // #cwg92-g
   // cxx98-14-warning@-1 {{mangled name of 'g' will change in C++17 due to non-throwing exception specification in function signature}}
   void h() throw() {
     g(f);
     // cxx98-14-error@-1 {{target exception specification is not superset of source}}
     // since-cxx17-error@-2 {{no matching function for call to 'g'}}
-    //   since-cxx17-note@#dr92-g {{candidate function not viable: no known conversion from 'void () throw(int, float)' to 'void (*)() throw()' for 1st argument}}
+    //   since-cxx17-note@#cwg92-g {{candidate function not viable: no known conversion from 'void () throw(int, float)' to 'void (*)() throw()' for 1st argument}}
     g(q);
     // cxx98-14-error@-1 {{target exception specification is not superset of source}}
     // since-cxx17-error@-2 {{no matching function for call to 'g'}}
-    //   since-cxx17-note@#dr92-g {{candidate function not viable: no known conversion from 'void (*)() throw(int)' to 'void (*)() throw()' for 1st argument}}
+    //   since-cxx17-note@#cwg92-g {{candidate function not viable: no known conversion from 'void (*)() throw(int)' to 'void (*)() throw()' for 1st argument}}
   }
 
   // Prior to C++17, this is OK because the exception specification is not
@@ -1376,31 +1376,31 @@ namespace dr92 { // dr92: 4 c++17
   Y<&h> yp; // ok
 }
 
-// dr93: na
+// cwg93: na
 
-namespace dr94 { // dr94: yes
+namespace cwg94 { // cwg94: yes
   struct A { static const int n = 5; };
   int arr[A::n];
 }
 
-namespace dr95 { // dr95: 3.3
+namespace cwg95 { // cwg95: 3.3
   struct A;
   struct B;
   namespace N {
     class C {
       friend struct A;
       friend struct B;
-      static void f(); // #dr95-C-f
+      static void f(); // #cwg95-C-f
     };
-    struct A *p; // dr95::A, not dr95::N::A.
+    struct A *p; // cwg95::A, not cwg95::N::A.
   }
   A *q = N::p; // ok, same type
   struct B { void f() { N::C::f(); } };
-  // expected-error@-1 {{'f' is a private member of 'dr95::N::C'}}
-  //   expected-note@#dr95-C-f {{implicitly declared private here}}
+  // expected-error@-1 {{'f' is a private member of 'cwg95::N::C'}}
+  //   expected-note@#cwg95-C-f {{implicitly declared private here}}
 }
 
-namespace dr96 { // dr96: no
+namespace cwg96 { // cwg96: no
   struct A {
     void f(int);
     template<typename T> int f(T);
@@ -1420,42 +1420,42 @@ namespace dr96 { // dr96: no
   }
 }
 
-namespace dr97 { // dr97: yes
+namespace cwg97 { // cwg97: yes
   struct A {
     static const int a = false;
     static const int b = !a;
   };
 }
 
-namespace dr98 { // dr98: yes
+namespace cwg98 { // cwg98: yes
   void test(int n) {
     switch (n) {
-      try { // #dr98-try
+      try { // #cwg98-try
         case 0:
         // expected-error@-1 {{cannot jump from switch statement to this case label}}
-        //   expected-note@#dr98-try {{jump bypasses initialization of try block}}
+        //   expected-note@#cwg98-try {{jump bypasses initialization of try block}}
         x:
           throw n;
-      } catch (...) { // #dr98-catch
+      } catch (...) { // #cwg98-catch
         case 1:
         // expected-error@-1 {{cannot jump from switch statement to this case label}}
-        //   expected-note@#dr98-catch {{jump bypasses initialization of catch block}}
+        //   expected-note@#cwg98-catch {{jump bypasses initialization of catch block}}
         y:
           throw n;
       }
       case 2:
         goto x;
         // expected-error@-1 {{cannot jump from this goto statement to its label}}
-        //   expected-note@#dr98-try {{jump bypasses initialization of try block}}
+        //   expected-note@#cwg98-try {{jump bypasses initialization of try block}}
       case 3:
         goto y;
         // expected-error@-1 {{cannot jump from this goto statement to its label}}
-        //   expected-note@#dr98-catch {{jump bypasses initialization of catch block}}
+        //   expected-note@#cwg98-catch {{jump bypasses initialization of catch block}}
     }
   }
 }
 
-namespace dr99 { // dr99: sup 214
+namespace cwg99 { // cwg99: sup 214
   template<typename T> void f(T&);
   template<typename T> int &f(const T&);
   const int n = 0;
diff --git a/clang/test/CXX/drs/dr10xx.cpp b/clang/test/CXX/drs/dr10xx.cpp
index 77c59078414c6..58d552942c77c 100644
--- a/clang/test/CXX/drs/dr10xx.cpp
+++ b/clang/test/CXX/drs/dr10xx.cpp
@@ -14,26 +14,26 @@ namespace std {
   };
 }
 
-namespace dr1004 { // dr1004: 5
+namespace cwg1004 { // cwg1004: 5
   template<typename> struct A {};
   template<typename> struct B1 {};
   template<template<typename> class> struct B2 {};
-  template<typename X> void f(); // #dr1004-f-1
-  template<template<typename> class X> void f(); // #dr1004-f-2
-  template<template<typename> class X> void g(); // #dr1004-g-1
-  template<typename X> void g(); // #dr1004-g-2
+  template<typename X> void f(); // #cwg1004-f-1
+  template<template<typename> class X> void f(); // #cwg1004-f-2
+  template<template<typename> class X> void g(); // #cwg1004-g-1
+  template<typename X> void g(); // #cwg1004-g-2
   struct C : A<int> {
     B1<A> b1a;
     B2<A> b2a;
     void h() {
       f<A>();
       // expected-error@-1 {{call to 'f' is ambiguous}}
-      //   expected-note@#dr1004-f-1 {{candidate function [with X = dr1004::A<int>]}}
-      //   expected-note@#dr1004-f-2 {{candidate function [with X = dr1004::A]}}
+      //   expected-note@#cwg1004-f-1 {{candidate function [with X = cwg1004::A<int>]}}
+      //   expected-note@#cwg1004-f-2 {{candidate function [with X = cwg1004::A]}}
       g<A>();
       // expected-error@-1 {{call to 'g' is ambiguous}}
-      //   expected-note@#dr1004-g-1 {{candidate function [with X = dr1004::A]}}
-      //   expected-note@#dr1004-g-2 {{candidate function [with X = dr1004::A<int>]}}
+      //   expected-note@#cwg1004-g-1 {{candidate function [with X = cwg1004::A]}}
+      //   expected-note@#cwg1004-g-2 {{candidate function [with X = cwg1004::A<int>]}}
     }
   };
 
@@ -41,17 +41,17 @@ namespace dr1004 { // dr1004: 5
   // name lookup of "T::template A" names the constructor.
   template<class T, template<class> class U = T::template A> struct Third { };
   // expected-error@-1 {{is a constructor name}}
-  //   expected-note@#dr1004-t {{in instantiation of default argument}}
-  Third<A<int> > t; // #dr1004-t
+  //   expected-note@#cwg1004-t {{in instantiation of default argument}}
+  Third<A<int> > t; // #cwg1004-t
 }
 
-namespace dr1042 { // dr1042: 3.5
+namespace cwg1042 { // cwg1042: 3.5
 #if __cplusplus >= 201402L
   // C++14 added an attribute that we can test the semantics of.
-  using foo [[deprecated]] = int; // #dr1042-using
+  using foo [[deprecated]] = int; // #cwg1042-using
   foo f = 12;
   // since-cxx14-warning@-1 {{'foo' is deprecated}}
-  //   since-cxx14-note@#dr1042-using {{'foo' has been explicitly marked deprecated here}}
+  //   since-cxx14-note@#cwg1042-using {{'foo' has been explicitly marked deprecated here}}
 #elif __cplusplus >= 201103L
   // C++11 did not have any attributes that could be applied to an alias
   // declaration, so the best we can test is that we accept an empty attribute
@@ -60,7 +60,7 @@ namespace dr1042 { // dr1042: 3.5
 #endif
 }
 
-namespace dr1048 { // dr1048: 3.6
+namespace cwg1048 { // cwg1048: 3.6
   struct A {};
   const A f();
   A g();
@@ -78,20 +78,20 @@ namespace dr1048 { // dr1048: 3.6
 #endif
 }
 
-namespace dr1054 { // dr1054: no
+namespace cwg1054 { // cwg1054: no
   // FIXME: Test is incomplete.
   struct A {} volatile a;
   void f() {
     // FIXME: This is wrong: an lvalue-to-rvalue conversion is applied here,
     // which copy-initializes a temporary from 'a'. Therefore this is
     // ill-formed because A does not have a volatile copy constructor.
-    // (We might want to track this aspect under dr1383 instead?)
+    // (We might want to track this aspect under cwg1383 instead?)
     a;
     // expected-warning@-1 {{expression result unused; assign into a variable to force a volatile load}}
   }
 }
 
-namespace dr1070 { // dr1070: 3.5
+namespace cwg1070 { // cwg1070: 3.5
 #if __cplusplus >= 201103L
   struct A {
     A(std::initializer_list<int>);
diff --git a/clang/test/CXX/drs/dr11xx.cpp b/clang/test/CXX/drs/dr11xx.cpp
index a71a105c7eb20..46a0e526be390 100644
--- a/clang/test/CXX/drs/dr11xx.cpp
+++ b/clang/test/CXX/drs/dr11xx.cpp
@@ -4,27 +4,27 @@
 // RUN: %clang_cc1 -std=c++17 %s -verify=expected -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++2a %s -verify=expected -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr1111 { // dr1111: 3.2
+namespace cwg1111 { // cwg1111: 3.2
 namespace example1 {
-template <typename> struct set; // #dr1111-struct-set
+template <typename> struct set; // #cwg1111-struct-set
 
 struct X {
-  template <typename T> void set(const T &value); // #dr1111-func-set
+  template <typename T> void set(const T &value); // #cwg1111-func-set
 };
 void foo() {
   X x;
   // FIXME: should we backport C++11 behavior?
   x.set<double>(3.2);
   // cxx98-error@-1 {{lookup of 'set' in member access expression is ambiguous; using member of 'X'}}
-  //   cxx98-note@#dr1111-func-set {{lookup in the object type 'X' refers here}}
-  //   cxx98-note@#dr1111-struct-set {{lookup from the current scope refers here}}
+  //   cxx98-note@#cwg1111-func-set {{lookup in the object type 'X' refers here}}
+  //   cxx98-note@#cwg1111-struct-set {{lookup from the current scope refers here}}
 }
 
 struct Y {};
 void bar() {
   Y y;
   y.set<double>(3.2);
-  // expected-error@-1 {{no member named 'set' in 'dr1111::example1::Y'}}
+  // expected-error@-1 {{no member named 'set' in 'cwg1111::example1::Y'}}
 }
 } // namespace example1
 
@@ -42,14 +42,14 @@ void baz() {
   a.operator A();
 }
 } // namespace example2
-} // namespace dr1111
+} // namespace cwg1111
 
-namespace dr1113 { // dr1113: partial
+namespace cwg1113 { // cwg1113: partial
   namespace named {
-    extern int a; // #dr1113-a
+    extern int a; // #cwg1113-a
     static int a;
     // expected-error@-1 {{static declaration of 'a' follows non-static}}
-    //   expected-note@#dr1113-a {{previous declaration is here}}
+    //   expected-note@#cwg1113-a {{previous declaration is here}}
   }
   namespace {
     extern int a;
@@ -57,7 +57,7 @@ namespace dr1113 { // dr1113: partial
     int b = a;
   }
 
-  // FIXME: Per DR1113 and DR4, this is ill-formed due to ambiguity: the second
+  // FIXME: Per CWG1113 and CWG4, this is ill-formed due to ambiguity: the second
   // 'f' has internal linkage, and so does not have C language linkage, so is
   // not a redeclaration of the first 'f'.
   //
@@ -71,4 +71,4 @@ namespace dr1113 { // dr1113: partial
   void g() { f(); }
 }
 
-// dr1150: na
+// cwg1150: na
diff --git a/clang/test/CXX/drs/dr12xx.cpp b/clang/test/CXX/drs/dr12xx.cpp
index da5dd02a00677..cdfbc6d672658 100644
--- a/clang/test/CXX/drs/dr12xx.cpp
+++ b/clang/test/CXX/drs/dr12xx.cpp
@@ -5,9 +5,9 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,since-cxx17,since-cxx14,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx17,since-cxx14,since-cxx11,since-cxx23 -fexceptions -fcxx-exceptions -pedantic-errors
 
-// dr1200: na
+// cwg1200: na
 
-namespace dr1213 { // dr1213: 7
+namespace cwg1213 { // cwg1213: 7
 #if __cplusplus >= 201103L
   using T = int[3];
   int &&r = T{}[1];
@@ -32,7 +32,7 @@ namespace dr1213 { // dr1213: 7
 }
 
 #if __cplusplus >= 201103L
-namespace dr1223 { // dr1223: 17 drafting 2023-05-12
+namespace cwg1223 { // cwg1223: 17 drafting 2023-05-12
 struct M;
 template <typename T>
 struct V;
@@ -80,31 +80,31 @@ void g() {
     A b(auto ()->C);
     static_assert(sizeof(B ()->C[1] == sizeof(int)), "");
     sizeof(auto () -> C[1]);
-    // since-cxx11-error@-1 {{function cannot return array type 'C[1]' (aka 'dr1223::BB[1]')}}
+    // since-cxx11-error@-1 {{function cannot return array type 'C[1]' (aka 'cwg1223::BB[1]')}}
 }
 
 }
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr1227 { // dr1227: 3.0
+namespace cwg1227 { // cwg1227: 3.0
 template <class T> struct A { using X = typename T::X; };
 // since-cxx11-error@-1 {{type 'int' cannot be used prior to '::' because it has no members}}
-//   since-cxx11-note@#dr1227-g {{in instantiation of template class 'dr1227::A<int>' requested here}}
-//   since-cxx11-note@#dr1227-g-int {{while substituting explicitly-specified template arguments into function template 'g'}}
+//   since-cxx11-note@#cwg1227-g {{in instantiation of template class 'cwg1227::A<int>' requested here}}
+//   since-cxx11-note@#cwg1227-g-int {{while substituting explicitly-specified template arguments into function template 'g'}}
 template <class T> typename T::X f(typename A<T>::X);
 template <class T> void f(...) { }
-template <class T> auto g(typename A<T>::X) -> typename T::X; // #dr1227-g
+template <class T> auto g(typename A<T>::X) -> typename T::X; // #cwg1227-g
 template <class T> void g(...) { }
 
 void h() {
   f<int>(0); // OK, substituting return type causes deduction to fail
-  g<int>(0); // #dr1227-g-int
+  g<int>(0); // #cwg1227-g-int
 }
 }
 #endif
 
-namespace dr1250 { // dr1250: 3.9
+namespace cwg1250 { // cwg1250: 3.9
 struct Incomplete;
 
 struct Base {
@@ -116,7 +116,7 @@ struct Derived : Base {
 };
 }
 
-namespace dr1265 { // dr1265: 5
+namespace cwg1265 { // cwg1265: 5
 #if __cplusplus >= 201103L
   auto a = 0, b() -> int;
   // since-cxx11-error@-1 {{declaration with trailing return type must be the only declaration in its group}}
@@ -136,9 +136,9 @@ namespace dr1265 { // dr1265: 5
 #endif
 }
 
-// dr1291: na
+// cwg1291: na
 
-namespace dr1295 { // dr1295: 4
+namespace cwg1295 { // cwg1295: 4
   struct X {
     unsigned bitfield : 4;
   };
@@ -150,11 +150,11 @@ namespace dr1295 { // dr1295: 4
   unsigned const &r2 = static_cast<unsigned &&>(x.bitfield);
   // cxx98-error@-1 {{rvalue references are a C++11 extension}}
 
-  template<unsigned &r> struct Y {}; // #dr1295-Y
-  Y<x.bitfield> y; // #dr1295-y
+  template<unsigned &r> struct Y {}; // #cwg1295-Y
+  Y<x.bitfield> y; // #cwg1295-y
   // cxx98-14-error@-1 {{non-type template argument does not refer to any declaration}}
-  //   cxx98-14-note@#dr1295-Y {{template parameter is declared here}}
-  // since-cxx17-error@#dr1295-y {{reference cannot bind to bit-field in converted constant expression}}
+  //   cxx98-14-note@#cwg1295-Y {{template parameter is declared here}}
+  // since-cxx17-error@#cwg1295-y {{reference cannot bind to bit-field in converted constant expression}}
 
 #if __cplusplus >= 201103L
   const unsigned other = 0;
diff --git a/clang/test/CXX/drs/dr13xx.cpp b/clang/test/CXX/drs/dr13xx.cpp
index d8e3b5d87bd14..dad82c4e2829f 100644
--- a/clang/test/CXX/drs/dr13xx.cpp
+++ b/clang/test/CXX/drs/dr13xx.cpp
@@ -17,18 +17,18 @@ namespace std {
 }
 
 #if __cplusplus >= 201103L
-namespace dr1305 { // dr1305: 3.0
-struct Incomplete; // #dr1305-Incomplete
+namespace cwg1305 { // cwg1305: 3.0
+struct Incomplete; // #cwg1305-Incomplete
 struct Complete {};
 
 int incomplete = alignof(Incomplete(&)[]);
 // since-cxx11-error@-1 {{invalid application of 'alignof' to an incomplete type 'Incomplete'}}
-//   since-cxx11-note@#dr1305-Incomplete {{forward declaration of 'dr1305::Incomplete'}}
+//   since-cxx11-note@#cwg1305-Incomplete {{forward declaration of 'cwg1305::Incomplete'}}
 int complete = alignof(Complete(&)[]);
 }
 #endif
 
-namespace dr1307 { // dr1307: 14
+namespace cwg1307 { // cwg1307: 14
 #if __cplusplus >= 201103L
 void f(int const (&)[2]);
 void f(int const (&)[3]);
@@ -38,11 +38,11 @@ void caller() {
   f({1, 2, 3});
 }
 #endif // __cplusplus >= 201103L
-} // namespace dr1307
+} // namespace cwg1307
 
-// dr1308: sup 1330
+// cwg1308: sup 1330
 
-namespace dr1310 { // dr1310: 5
+namespace cwg1310 { // cwg1310: 5
   struct S {} * sp = new S::S;
   // expected-error@-1 {{qualified reference to 'S' is a constructor name rather than a type in this context}}
   void f() {
@@ -126,8 +126,8 @@ namespace dr1310 { // dr1310: 5
   void wt_test() {
     typename W::W w2a;
     // expected-error@-1 {{ISO C++ specifies that qualified reference to 'W' is a constructor name rather than a type in this context, despite preceding 'typename' keyword}}
-    //   cxx98-note@#dr1310-W-int {{in instantiation of function template specialization 'dr1310::wt_test<dr1310::W<int> >' requested here}}
-    //   since-cxx11-note@#dr1310-W-int {{in instantiation of function template specialization 'dr1310::wt_test<dr1310::W<int>>' requested here}}
+    //   cxx98-note@#cwg1310-W-int {{in instantiation of function template specialization 'cwg1310::wt_test<cwg1310::W<int> >' requested here}}
+    //   since-cxx11-note@#cwg1310-W-int {{in instantiation of function template specialization 'cwg1310::wt_test<cwg1310::W<int>>' requested here}}
     typename W::template W<int> w4;
     // expected-error@-1 {{ISO C++ specifies that qualified reference to 'W' is a constructor name rather than a template name in this context, despite preceding 'template' keyword}}
     TTy<typename W::W> tt2;
@@ -148,11 +148,11 @@ namespace dr1310 { // dr1310: 5
     (void)w.template W<int>::W::n;
     (void)w.template W<int>::template W<int>::n;
   }
-  template void wt_test<W<int> >(); // #dr1310-W-int
+  template void wt_test<W<int> >(); // #cwg1310-W-int
   template void wt_test_good<W<int> >();
 }
 
-namespace dr1315 { // dr1315: partial
+namespace cwg1315 { // cwg1315: partial
   template <int I, int J> struct A {};
   template <int I> struct A<I + 5, I * 2> {};
   // expected-error@-1 {{class template partial specialization contains a template parameter that cannot be deduced; this partial specialization will never be used}}
@@ -160,7 +160,7 @@ namespace dr1315 { // dr1315: partial
   template <int I> struct A<I, I> {};
 
   template <int I, int J, int K> struct B;
-  template <int I, int K> struct B<I, I * 2, K> {}; // #dr1315-B-1
+  template <int I, int K> struct B<I, I * 2, K> {}; // #cwg1315-B-1
   B<1, 2, 3> b1;
 
   // Multiple declarations with the same dependent expression are equivalent
@@ -169,13 +169,13 @@ namespace dr1315 { // dr1315: partial
   B<1, 2, 2>::type b2;
 
   // Multiple declarations with differing dependent expressions are unordered.
-  template <int I, int K> struct B<I, I + 1, K> {}; // #dr1315-B-2
+  template <int I, int K> struct B<I, I + 1, K> {}; // #cwg1315-B-2
   B<1, 2, 4> b3;
   // expected-error@-1 {{ambiguous partial specializations of 'B<1, 2, 4>'}}
-  //   expected-note@#dr1315-B-1 {{partial specialization matches [with I = 1, K = 4]}}
-  //   expected-note@#dr1315-B-2 {{partial specialization matches [with I = 1, K = 4]}}
+  //   expected-note@#cwg1315-B-1 {{partial specialization matches [with I = 1, K = 4]}}
+  //   expected-note@#cwg1315-B-2 {{partial specialization matches [with I = 1, K = 4]}}
 
-  // FIXME: Under dr1315, this is perhaps valid, but that is not clear: this
+  // FIXME: Under cwg1315, this is perhaps valid, but that is not clear: this
   // fails the "more specialized than the primary template" test because the
   // dependent type of T::value is not the same as 'int'.
   // A core issue will be opened to decide what is supposed to happen here.
@@ -184,7 +184,7 @@ namespace dr1315 { // dr1315: partial
   // expected-error@-1 {{type of specialized non-type template argument depends on a template parameter of the partial specialization}}
 }
 
-namespace dr1330 { // dr1330: 4 c++11
+namespace cwg1330 { // cwg1330: 4 c++11
   // exception-specifications are parsed in a context where the class is complete.
   struct A {
     void f() throw(T) {}
@@ -204,7 +204,7 @@ namespace dr1330 { // dr1330: 4 c++11
   //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
   void (A::*af2)() throw() = &A::f;
   // cxx98-14-error@-1 {{target exception specification is not superset of source}}
-  // since-cxx17-error@-2 {{cannot initialize a variable of type 'void (dr1330::A::*)() throw()' with an rvalue of type 'void (dr1330::A::*)() throw(T)': different exception specifications}}
+  // since-cxx17-error@-2 {{cannot initialize a variable of type 'void (cwg1330::A::*)() throw()' with an rvalue of type 'void (cwg1330::A::*)() throw(T)': different exception specifications}}
 
 #if __cplusplus >= 201103L
   static_assert(noexcept(A().g()), "");
@@ -252,7 +252,7 @@ namespace dr1330 { // dr1330: 4 c++11
   void (B<P>::*bpf3)() = &B<P>::f;
   void (B<P>::*bpf4)() throw() = &B<P>::f;
   // cxx98-14-error@-1 {{target exception specification is not superset of source}}
-  // since-cxx17-error@-2 {{cannot initialize a variable of type 'void (B<P>::*)() throw()' with an rvalue of type 'void (dr1330::B<dr1330::P>::*)() throw(T, typename P::type)': different exception specifications}}
+  // since-cxx17-error@-2 {{cannot initialize a variable of type 'void (B<P>::*)() throw()' with an rvalue of type 'void (cwg1330::B<cwg1330::P>::*)() throw(T, typename P::type)': different exception specifications}}
 
 #if __cplusplus >= 201103L
   static_assert(noexcept(B<P>().g()), "");
@@ -260,73 +260,73 @@ namespace dr1330 { // dr1330: 4 c++11
   static_assert(!noexcept(B<Q>().g()), "");
 #endif
 
-  template<typename T> int f() throw(typename T::error) { return 0; } // #dr1330-f
-  // expected-error@#dr1330-f {{type 'int' cannot be used prior to '::' because it has no members}}
-  //   cxx98-note@#dr1330-f-int {{in instantiation of function template specialization 'dr1330::f<int>' requested here}}
-  //   since-cxx11-note@#dr1330-f-int {{in instantiation of exception specification for 'f<int>' requested here}}
-  // cxx98-14-error@#dr1330-f {{type 'short' cannot be used prior to '::' because it has no members}}
-  //   cxx98-14-note@#dr1330-f-short {{in instantiation of function template specialization 'dr1330::f<short>' requested here}}
-  //   cxx11-14-note@#dr1330-f {{in instantiation of exception specification for 'f<short>' requested here}}
-  // since-cxx11-error@#dr1330-f {{type 'char' cannot be used prior to '::' because it has no members}}
-  //   since-cxx11-note@#dr1330-f-char {{in instantiation of exception specification for 'f<char>' requested here}}
-  // since-cxx11-error@#dr1330-f {{type 'float' cannot be used prior to '::' because it has no members}}
-  //   since-cxx11-note@#dr1330-f-float {{in instantiation of exception specification for 'f<float>' requested here}}
-  // since-cxx17-error@#dr1330-f {{ISO C++17 does not allow dynamic exception specifications}}
-  //   since-cxx17-note@#dr1330-f {{use 'noexcept(false)' instead}}
+  template<typename T> int f() throw(typename T::error) { return 0; } // #cwg1330-f
+  // expected-error@#cwg1330-f {{type 'int' cannot be used prior to '::' because it has no members}}
+  //   cxx98-note@#cwg1330-f-int {{in instantiation of function template specialization 'cwg1330::f<int>' requested here}}
+  //   since-cxx11-note@#cwg1330-f-int {{in instantiation of exception specification for 'f<int>' requested here}}
+  // cxx98-14-error@#cwg1330-f {{type 'short' cannot be used prior to '::' because it has no members}}
+  //   cxx98-14-note@#cwg1330-f-short {{in instantiation of function template specialization 'cwg1330::f<short>' requested here}}
+  //   cxx11-14-note@#cwg1330-f {{in instantiation of exception specification for 'f<short>' requested here}}
+  // since-cxx11-error@#cwg1330-f {{type 'char' cannot be used prior to '::' because it has no members}}
+  //   since-cxx11-note@#cwg1330-f-char {{in instantiation of exception specification for 'f<char>' requested here}}
+  // since-cxx11-error@#cwg1330-f {{type 'float' cannot be used prior to '::' because it has no members}}
+  //   since-cxx11-note@#cwg1330-f-float {{in instantiation of exception specification for 'f<float>' requested here}}
+  // since-cxx17-error@#cwg1330-f {{ISO C++17 does not allow dynamic exception specifications}}
+  //   since-cxx17-note@#cwg1330-f {{use 'noexcept(false)' instead}}
 
   // An exception-specification is needed even if the function is only used in
   // an unevaluated operand.
-  int f1 = sizeof(f<int>()); // #dr1330-f-int
+  int f1 = sizeof(f<int>()); // #cwg1330-f-int
 #if __cplusplus >= 201103L
-  decltype(f<char>()) f2; // #dr1330-f-char
-  bool f3 = noexcept(f<float>()); /// #dr1330-f-float
+  decltype(f<char>()) f2; // #cwg1330-f-char
+  bool f3 = noexcept(f<float>()); /// #cwg1330-f-float
 #endif
   // In C++17 onwards, substituting explicit template arguments into the
   // function type substitutes into the exception specification (because it's
   // part of the type). In earlier languages, we don't notice there's a problem
   // until we've already started to instantiate.
-  template int f<short>(); // #dr1330-f-short
+  template int f<short>(); // #cwg1330-f-short
   // since-cxx17-error@-1 {{explicit instantiation of 'f' does not refer to a function template, variable template, member function, member class, or static data member}}
-  //   since-cxx17-note@#dr1330-f {{candidate template ignored: substitution failure [with T = short]: type 'short' cannot be used prior to '::' because it has no members}}
+  //   since-cxx17-note@#cwg1330-f {{candidate template ignored: substitution failure [with T = short]: type 'short' cannot be used prior to '::' because it has no members}}
 
   template<typename T> struct C {
-    C() throw(typename T::type); // #dr1330-C
+    C() throw(typename T::type); // #cwg1330-C
     // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
     //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
-    // cxx98-error@#dr1330-C {{type 'void' cannot be used prior to '::' because it has no members}}
-    //   cxx98-note@#dr1330-C-void {{in instantiation of template class 'dr1330::C<void>' requested here}}
-    // expected-error@#dr1330-C {{type 'int' cannot be used prior to '::' because it has no members}}
-    //   cxx98-note@#dr1330-C-int {{in instantiation of template class 'dr1330::C<int>' requested here}}
-    //   since-cxx11-note@#dr1330-C-int {{in instantiation of exception specification for 'C' requested here}}
-    //   since-cxx11-note@#dr1330-e {{in evaluation of exception specification for 'dr1330::E::E' needed here}}
+    // cxx98-error@#cwg1330-C {{type 'void' cannot be used prior to '::' because it has no members}}
+    //   cxx98-note@#cwg1330-C-void {{in instantiation of template class 'cwg1330::C<void>' requested here}}
+    // expected-error@#cwg1330-C {{type 'int' cannot be used prior to '::' because it has no members}}
+    //   cxx98-note@#cwg1330-C-int {{in instantiation of template class 'cwg1330::C<int>' requested here}}
+    //   since-cxx11-note@#cwg1330-C-int {{in instantiation of exception specification for 'C' requested here}}
+    //   since-cxx11-note@#cwg1330-e {{in evaluation of exception specification for 'cwg1330::E::E' needed here}}
   };
-  struct D : C<void> {}; // #dr1330-C-void
+  struct D : C<void> {}; // #cwg1330-C-void
   void f(D &d) { d = d; } // ok
 
-  struct E : C<int> {}; // #dr1330-C-int
-  E e; // #dr1330-e
+  struct E : C<int> {}; // #cwg1330-C-int
+  E e; // #cwg1330-e
 }
 
-// dr1334: sup 1719
+// cwg1334: sup 1719
 
-namespace dr1341 { // dr1341: sup P0683R1
+namespace cwg1341 { // cwg1341: sup P0683R1
 #if __cplusplus >= 202002L
 int a;
-const int b = 0; // #dr1341-b
+const int b = 0; // #cwg1341-b
 struct S {
   int x1 : 8 = 42;
   int x2 : 8 { 42 };
   int y1 : true ? 8 : a = 42;
   int y2 : true ? 8 : b = 42;
   // since-cxx20-error@-1 {{cannot assign to variable 'b' with const-qualified type 'const int'}}
-  //   since-cxx20-note@#dr1341-b {{variable 'b' declared const here}}
+  //   since-cxx20-note@#cwg1341-b {{variable 'b' declared const here}}
   int y3 : (true ? 8 : b) = 42;
   int z : 1 || new int { 0 };
 };
 #endif
 }
 
-namespace dr1346 { // dr1346: 3.5
+namespace cwg1346 { // cwg1346: 3.5
   auto a(1);
   // cxx98-error@-1 {{'auto' type specifier is a C++11 extension}}
   auto b(1, 2);
@@ -345,9 +345,9 @@ namespace dr1346 { // dr1346: 3.5
     auto x(ts...);
     // cxx98-error@-1 {{'auto' type specifier is a C++11 extension}}
     // expected-error@-2 {{initializer for variable 'x' with type 'auto' is empty}}
-    //   expected-note@#dr1346-f {{in instantiation of function template specialization 'dr1346::f<>' requested here}}
+    //   expected-note@#cwg1346-f {{in instantiation of function template specialization 'cwg1346::f<>' requested here}}
   }
-  template void f(); // #dr1346-f
+  template void f(); // #cwg1346-f
 
 #if __cplusplus >= 201103L
   void init_capture() {
@@ -369,7 +369,7 @@ namespace dr1346 { // dr1346: 3.5
 #endif
 }
 
-namespace dr1347 { // dr1347: 3.1
+namespace cwg1347 { // cwg1347: 3.1
   auto x = 5, *y = &x;
   // cxx98-error@-1 {{'auto' type specifier is a C++11 extension}}
   auto z = y, *q = y;
@@ -383,7 +383,7 @@ namespace dr1347 { // dr1347: 3.1
 #endif
 }
 
-namespace dr1350 { // dr1350: 3.5
+namespace cwg1350 { // cwg1350: 3.5
 #if __cplusplus >= 201103L
 struct NoexceptCtor {
   NoexceptCtor(int) noexcept {}
@@ -452,12 +452,12 @@ struct D4 : NoexceptCtor, ThrowingDefaultArgTemplate {
 
 static_assert(!__is_nothrow_constructible(D4, int), "");
 #endif
-} // namespace dr1350
+} // namespace cwg1350
 
-namespace dr1358 { // dr1358: 3.1
+namespace cwg1358 { // cwg1358: 3.1
 #if __cplusplus >= 201103L
   struct Lit { constexpr operator int() const { return 0; } };
-  struct NonLit { NonLit(); operator int(); }; // #dr1358-NonLit
+  struct NonLit { NonLit(); operator int(); }; // #cwg1358-NonLit
   struct NonConstexprConv { constexpr operator int() const; };
   struct Virt { virtual int f(int) const; };
 
@@ -486,83 +486,83 @@ namespace dr1358 { // dr1358: 3.1
     int member;
     constexpr B(NonLit u) : member(u) {}
     // cxx11-20-error@-1 {{constexpr constructor's 1st parameter type 'NonLit' is not a literal type}}
-    //   cxx11-20-note@#dr1358-NonLit {{'NonLit' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
+    //   cxx11-20-note@#cwg1358-NonLit {{'NonLit' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
     constexpr NonLit f(NonLit u) const { return NonLit(); }
     // cxx11-20-error@-1 {{constexpr function's return type 'NonLit' is not a literal type}}
-    //   cxx11-20-note@#dr1358-NonLit {{'NonLit' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
+    //   cxx11-20-note@#cwg1358-NonLit {{'NonLit' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
   };
 #endif
 }
 
-namespace dr1359 { // dr1359: 3.5
+namespace cwg1359 { // cwg1359: 3.5
 #if __cplusplus >= 201103L
   union A { constexpr A() = default; };
-  union B { constexpr B() = default; int a; }; // #dr1359-B
+  union B { constexpr B() = default; int a; }; // #cwg1359-B
   // cxx11-17-error@-1 {{defaulted definition of default constructor cannot be marked constexpr before C++23}}
-  union C { constexpr C() = default; int a, b; }; // #dr1359-C
+  union C { constexpr C() = default; int a, b; }; // #cwg1359-C
   // cxx11-17-error@-1 {{defaulted definition of default constructor cannot be marked constexpr}} 
   struct X { constexpr X() = default; union {}; };
   // since-cxx11-error@-1 {{declaration does not declare anything}}
-  struct Y { constexpr Y() = default; union { int a; }; }; // #dr1359-Y
+  struct Y { constexpr Y() = default; union { int a; }; }; // #cwg1359-Y
   // cxx11-17-error@-1 {{defaulted definition of default constructor cannot be marked constexpr}}
 
   constexpr A a = A();
   constexpr B b = B();
   // cxx11-17-error@-1 {{no matching constructor for initialization of 'B'}}
-  //   cxx11-17-note@#dr1359-B {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-  //   cxx11-17-note@#dr1359-B {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-B {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-B {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
   constexpr C c = C();
   // cxx11-17-error@-1 {{no matching constructor for initialization of 'C'}}
-  //   cxx11-17-note@#dr1359-C {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-  //   cxx11-17-note@#dr1359-C {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-C {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-C {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
   constexpr X x = X();
   constexpr Y y = Y();
   // cxx11-17-error@-1 {{no matching constructor for initialization of 'Y'}}
-  //   cxx11-17-note@#dr1359-Y {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-  //   cxx11-17-note@#dr1359-Y {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-Y {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1359-Y {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
 #endif
 }
 
-namespace dr1388 { // dr1388: 4
-  template<typename A, typename ...T> void f(T..., A); // #dr1388-f
+namespace cwg1388 { // cwg1388: 4
+  template<typename A, typename ...T> void f(T..., A); // #cwg1388-f
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
-  template<typename ...T> void g(T..., int); // #dr1388-g
+  template<typename ...T> void g(T..., int); // #cwg1388-g
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
-  template<typename ...T, typename A> void h(T..., A); // #dr1388-h
+  template<typename ...T, typename A> void h(T..., A); // #cwg1388-h
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
 
   void test_f() {
     f(0); // ok, trailing parameter pack deduced to empty
     f(0, 0);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr1388-f {{candidate function [with A = int, T = <>] not viable: requires 1 argument, but 2 were provided}}
+    //   expected-note@#cwg1388-f {{candidate function [with A = int, T = <>] not viable: requires 1 argument, but 2 were provided}}
     f<int>(0);
     f<int>(0, 0);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr1388-f {{candidate function [with A = int, T = <>] not viable: requires 1 argument, but 2 were provided}}
+    //   expected-note@#cwg1388-f {{candidate function [with A = int, T = <>] not viable: requires 1 argument, but 2 were provided}}
     f<int, int>(0, 0);
     f<int, int, int>(0, 0);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr1388-f {{candidate function [with A = int, T = <int, int>] not viable: requires 3 arguments, but 2 were provided}}
+    //   expected-note@#cwg1388-f {{candidate function [with A = int, T = <int, int>] not viable: requires 3 arguments, but 2 were provided}}
 
     g(0);
     g(0, 0);
     // expected-error@-1 {{no matching function for call to 'g'}}
-    //   expected-note@#dr1388-g {{candidate function [with T = <>] not viable: requires 1 argument, but 2 were provided}}
+    //   expected-note@#cwg1388-g {{candidate function [with T = <>] not viable: requires 1 argument, but 2 were provided}}
     g<>(0);
     g<int>(0);
     // expected-error@-1 {{no matching function for call to 'g'}}
-    //   expected-note@#dr1388-g {{candidate function [with T = <int>] not viable: requires 2 arguments, but 1 was provided}}
+    //   expected-note@#cwg1388-g {{candidate function [with T = <int>] not viable: requires 2 arguments, but 1 was provided}}
     g<int>(0, 0);
 
     h(0);
     h(0, 0);
     // expected-error@-1 {{no matching function for call to 'h'}}
-    //   expected-note@#dr1388-h {{candidate function [with T = <>, A = int] not viable: requires 1 argument, but 2 were provided}}
+    //   expected-note@#cwg1388-h {{candidate function [with T = <>, A = int] not viable: requires 1 argument, but 2 were provided}}
     h<int>(0, 0);
     h<int, int>(0, 0);
     // expected-error@-1 {{no matching function for call to 'h'}}
-    //   expected-note@#dr1388-h {{candidate template ignored: couldn't infer template argument 'A'}}
+    //   expected-note@#cwg1388-h {{candidate template ignored: couldn't infer template argument 'A'}}
   }
 
   // A non-trailing parameter pack is still a non-deduced context, even though
@@ -570,40 +570,40 @@ namespace dr1388 { // dr1388: 4
   template<typename T, typename U> struct pair {};
   template<typename ...T> struct tuple { typedef char type; }; //
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
-  template<typename ...T, typename ...U> void f_pair_1(pair<T, U>..., int); // #dr1388-f-1
+  template<typename ...T, typename ...U> void f_pair_1(pair<T, U>..., int); // #cwg1388-f-1
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
   // cxx98-error@-2 {{variadic templates are a C++11 extension}}
   template<typename ...T, typename U> void f_pair_2(pair<T, char>..., U);
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
-  template<typename ...T, typename ...U> void f_pair_3(pair<T, U>..., tuple<U...>); // #dr1388-f-3
+  template<typename ...T, typename ...U> void f_pair_3(pair<T, U>..., tuple<U...>); // #cwg1388-f-3
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
   // cxx98-error@-2 {{variadic templates are a C++11 extension}}
-  template<typename ...T> void f_pair_4(pair<T, char>..., T...); // #dr1388-f-4
+  template<typename ...T> void f_pair_4(pair<T, char>..., T...); // #cwg1388-f-4
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
   void g(pair<int, char> a, pair<long, char> b, tuple<char, char> c) {
     f_pair_1<int, long>(a, b, 0);
     // expected-error@-1 {{no matching function for call to 'f_pair_1'}}
-    //   expected-note@#dr1388-f-1 {{candidate template ignored: substitution failure [with T = <int, long>]: deduced incomplete pack <(no value), (no value)> for template parameter 'U'}}
+    //   expected-note@#cwg1388-f-1 {{candidate template ignored: substitution failure [with T = <int, long>]: deduced incomplete pack <(no value), (no value)> for template parameter 'U'}}
     f_pair_2<int, long>(a, b, 0);
     f_pair_3<int, long>(a, b, c);
     f_pair_3<int, long>(a, b, tuple<char>());
     // expected-error@-1 {{no matching function for call to 'f_pair_3'}}
-    //   expected-note@#dr1388-f-3 {{candidate template ignored: deduced packs of different lengths for parameter 'U' (<(no value), (no value)> vs. <char>)}}
+    //   expected-note@#cwg1388-f-3 {{candidate template ignored: deduced packs of different lengths for parameter 'U' (<(no value), (no value)> vs. <char>)}}
     f_pair_4<int, long>(a, b, 0, 0L);
     f_pair_4<int, long>(a, b, 0, 0L, "foo");
     // expected-error@-1 {{no matching function for call to 'f_pair_4'}}
-    //   expected-note@#dr1388-f-4 {{candidate template ignored: deduced packs of different lengths for parameter 'T' (<int, long> vs. <int, long, const char *>)}}
+    //   expected-note@#cwg1388-f-4 {{candidate template ignored: deduced packs of different lengths for parameter 'T' (<int, long> vs. <int, long, const char *>)}}
   }
 }
 
-namespace dr1391 { // dr1391: partial
+namespace cwg1391 { // cwg1391: partial
   struct A {}; struct B : A {};
-  template<typename T> struct C { C(int); typename T::error error; }; // #dr1391-C
-  // expected-error@#dr1391-C {{type 'int' cannot be used prior to '::' because it has no members}}
-  //   expected-note@#dr1391-b {{in instantiation of template class 'dr1391::C<int>' requested here}}
-  //   expected-note@#dr1391-b {{while substituting deduced template arguments into function template 'b' [with T = int]}}
-  // expected-error@#dr1391-C {{type 'double' cannot be used prior to '::' because it has no members}}
-  //   expected-note@#dr1391-c {{in instantiation of template class 'dr1391::C<double>' requested here}}
+  template<typename T> struct C { C(int); typename T::error error; }; // #cwg1391-C
+  // expected-error@#cwg1391-C {{type 'int' cannot be used prior to '::' because it has no members}}
+  //   expected-note@#cwg1391-b {{in instantiation of template class 'cwg1391::C<int>' requested here}}
+  //   expected-note@#cwg1391-b {{while substituting deduced template arguments into function template 'b' [with T = int]}}
+  // expected-error@#cwg1391-C {{type 'double' cannot be used prior to '::' because it has no members}}
+  //   expected-note@#cwg1391-c {{in instantiation of template class 'cwg1391::C<double>' requested here}}
   template<typename T> struct D {};
 
   // No deduction is performed for parameters with no deducible template-parameters, therefore types do not need to match.
@@ -644,42 +644,42 @@ namespace dr1391 { // dr1391: partial
   void test_b() {
     b(0, 0); // ok, deduction fails prior to forming a conversion sequence and instantiating C<int>
     // FIXME: The "while substituting" note should point at the overload candidate.
-    b<int>(0, 0); // #dr1391-b
+    b<int>(0, 0); // #cwg1391-b
   }
 
   template<typename T> struct Id { typedef T type; };
   template<typename T> void c(T, typename Id<C<T> >::type);
   void test_c() {
     // Implicit conversion sequences for dependent types are checked later.
-    c(0.0, 0); // #dr1391-c
+    c(0.0, 0); // #cwg1391-c
   }
 
   namespace partial_ordering {
     // FIXME: Second template should be considered more specialized because non-dependent parameter is ignored.
-    template<typename T> int a(T, short) = delete; // #dr1391-a-short
+    template<typename T> int a(T, short) = delete; // #cwg1391-a-short
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
-    template<typename T> int a(T*, char); // #dr1391-a-char
+    template<typename T> int a(T*, char); // #cwg1391-a-char
     int test_a = a((int*)0, 0);
     // expected-error@-1 {{call to 'a' is ambiguous}} FIXME
-    //   expected-note@#dr1391-a-short {{candidate function [with T = int *] has been explicitly deleted}}
-    //   expected-note@#dr1391-a-char {{candidate function [with T = int]}}
+    //   expected-note@#cwg1391-a-short {{candidate function [with T = int *] has been explicitly deleted}}
+    //   expected-note@#cwg1391-a-char {{candidate function [with T = int]}}
 
     // FIXME: Second template should be considered more specialized:
     // deducing #1 from #2 ignores the second P/A pair, so deduction succeeds,
     // deducing #2 from #1 fails to deduce T, so deduction fails.
-    template<typename T> int b(T, int) = delete; // #dr1391-b-int
+    template<typename T> int b(T, int) = delete; // #cwg1391-b-int
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
-    template<typename T, typename U> int b(T*, U); // #dr1391-b-U
+    template<typename T, typename U> int b(T*, U); // #cwg1391-b-U
     int test_b = b((int*)0, 0);
     // expected-error@-1 {{call to 'b' is ambiguous}} FIXME
-    //   expected-note@#dr1391-b-int {{candidate function [with T = int *] has been explicitly deleted}}
-    //   expected-note@#dr1391-b-U {{candidate function [with T = int, U = int]}}
+    //   expected-note@#cwg1391-b-int {{candidate function [with T = int *] has been explicitly deleted}}
+    //   expected-note@#cwg1391-b-U {{candidate function [with T = int, U = int]}}
 
     // Unintended consequences: because partial ordering does not consider
     // explicit template arguments, and deduction from a non-dependent type
     // vacuously succeeds, a non-dependent template is less specialized than
     // anything else!
-    // According to DR1391, this is ambiguous!
+    // According to CWG1391, this is ambiguous!
     template<typename T> int c(int);
     template<typename T> int c(T);
     int test_c1 = c(0); // ok
@@ -687,46 +687,46 @@ namespace dr1391 { // dr1391: partial
   }
 }
 
-namespace dr1394 { // dr1394: 15
+namespace cwg1394 { // cwg1394: 15
 #if __cplusplus >= 201103L
 struct Incomplete;
 Incomplete f(Incomplete) = delete; // well-formed
 #endif
 }
 
-namespace dr1395 { // dr1395: 16
+namespace cwg1395 { // cwg1395: 16
 #if __cplusplus >= 201103L
   template <typename T, typename... U> void f(T, U...);
   template <typename T> void f(T);
   void h(int i) {
-    // This is made ambiguous by dr692, but made valid again by dr1395.
+    // This is made ambiguous by cwg692, but made valid again by cwg1395.
     f(&i);
   }
 #endif
 }
 
-namespace dr1397 { // dr1397: 3.2
+namespace cwg1397 { // cwg1397: 3.2
 #if __cplusplus >= 201103L
 struct A {
 // cxx11-error@-1 {{default member initializer for 'p' needed within definition of enclosing class 'A' outside of member functions}}
-//   cxx11-note@#dr1397-p {{in evaluation of exception specification for 'dr1397::A::A' needed here}}
-//   cxx11-note@#dr1397-p {{default member initializer declared here}}
-  void *p = A{}; // #dr1397-p
+//   cxx11-note@#cwg1397-p {{in evaluation of exception specification for 'cwg1397::A::A' needed here}}
+//   cxx11-note@#cwg1397-p {{default member initializer declared here}}
+  void *p = A{}; // #cwg1397-p
   // since-cxx14-error@-1 {{default member initializer for 'p' needed within definition of enclosing class 'A' outside of member functions}}
   //   since-cxx14-note@-2 {{default member initializer declared here}}
   operator void*() const { return nullptr; }
 };
 #endif
-} // namespace dr1397
+} // namespace cwg1397
 
-namespace dr1399 { // dr1399: dup 1388
-  template<typename ...T> void f(T..., int, T...) {} // #dr1399-f 
+namespace cwg1399 { // cwg1399: dup 1388
+  template<typename ...T> void f(T..., int, T...) {} // #cwg1399-f 
   // cxx98-error@-1 {{variadic templates are a C++11 extension}}
   void g() {
     f(0);
     f<int>(0, 0, 0);
     f(0, 0, 0);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr1399-f {{candidate template ignored: deduced packs of different lengths for parameter 'T' (<> vs. <int, int>)}}
+    //   expected-note@#cwg1399-f {{candidate template ignored: deduced packs of different lengths for parameter 'T' (<> vs. <int, int>)}}
   }
 }
diff --git a/clang/test/CXX/drs/dr14xx.cpp b/clang/test/CXX/drs/dr14xx.cpp
index ed6dda731fd51..9ff9a68dc13c3 100644
--- a/clang/test/CXX/drs/dr14xx.cpp
+++ b/clang/test/CXX/drs/dr14xx.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11,since-cxx14,since-cxx20 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++2c %s -verify=expected,since-cxx11,since-cxx14,since-cxx20 -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr1413 { // dr1413: 12
+namespace cwg1413 { // cwg1413: 12
   template<int> struct Check {
     typedef int type;
   };
@@ -21,28 +21,28 @@ namespace dr1413 { // dr1413: 12
       // expected-error@-1 {{use of undeclared identifier 'var1'}}
 
       // ok, variable declaration
-      Check<true ? 0 : a>::type *var2; // #dr1413-var2
+      Check<true ? 0 : a>::type *var2; // #cwg1413-var2
       Check<true ? 0 : b>::type *var3;
       // expected-error@-1 {{use of undeclared identifier 'var3'}}
-      //   expected-note@#dr1413-var2 {{'var2' declared here}}
+      //   expected-note@#cwg1413-var2 {{'var2' declared here}}
       Check<true ? 0 : ((void)c, 0)>::type *var4;
       // expected-error@-1 {{use of undeclared identifier 'var4'}}
-      //   expected-note@#dr1413-var2 {{'var2' declared here}}
+      //   expected-note@#cwg1413-var2 {{'var2' declared here}}
 
       // value-dependent because of the implied type-dependent 'this->', not because of 'd'
       Check<true ? 0 : (d(), 0)>::type *var5;
       // expected-error@-1 {{use of undeclared identifier 'var5'}}
-      //   expected-note@#dr1413-var2 {{'var2' declared here}}
+      //   expected-note@#cwg1413-var2 {{'var2' declared here}}
 
       // value-dependent because of the value-dependent '&' operator, not because of 'A::d'
       Check<true ? 0 : (&A::d(), 0)>::type *var5;
       // expected-error@-1 {{use of undeclared identifier 'var5'}}
-      //   expected-note@#dr1413-var2 {{'var2' declared here}}
+      //   expected-note@#cwg1413-var2 {{'var2' declared here}}
     }
   };
 }
 
-namespace dr1423 { // dr1423: 11
+namespace cwg1423 { // cwg1423: 11
 #if __cplusplus >= 201103L
   bool b1 = nullptr;
   // since-cxx11-error@-1 {{cannot initialize a variable of type 'bool' with an rvalue of type 'std::nullptr_t'}}
@@ -55,9 +55,9 @@ namespace dr1423 { // dr1423: 11
 #endif
 }
 
-// dr1425: na abi
+// cwg1425: na abi
 
-namespace dr1432 { // dr1432: 16
+namespace cwg1432 { // cwg1432: 16
 #if __cplusplus >= 201103L
   template<typename T> T declval();
 
@@ -78,7 +78,7 @@ namespace dr1432 { // dr1432: 16
 #endif
 }
 
-namespace dr1443 { // dr1443: yes
+namespace cwg1443 { // cwg1443: yes
 struct A {
   int i;
   A() { void foo(int=i); }
@@ -86,7 +86,7 @@ struct A {
 };
 }
 
-namespace dr1460 { // dr1460: 3.5
+namespace cwg1460 { // cwg1460: 3.5
 #if __cplusplus >= 201103L
   namespace DRExample {
     union A {
@@ -121,23 +121,23 @@ namespace dr1460 { // dr1460: 3.5
   }
 
   union A {};
-  union B { int n; }; // #dr1460-B
+  union B { int n; }; // #cwg1460-B
   union C { int n = 0; };
   struct D { union {}; };
   // expected-error@-1 {{declaration does not declare anything}}
-  struct E { union { int n; }; }; // #dr1460-E
+  struct E { union { int n; }; }; // #cwg1460-E
   struct F { union { int n = 0; }; };
 
   struct X {
     friend constexpr A::A() noexcept;
     friend constexpr B::B() noexcept;
     // cxx11-17-error@-1 {{constexpr declaration of 'B' follows non-constexpr declaration}}
-    //   cxx11-17-note@#dr1460-B {{previous declaration is here}}
+    //   cxx11-17-note@#cwg1460-B {{previous declaration is here}}
     friend constexpr C::C() noexcept;
     friend constexpr D::D() noexcept;
     friend constexpr E::E() noexcept;
     // cxx11-17-error@-1 {{constexpr declaration of 'E' follows non-constexpr declaration}}
-    //   cxx11-17-note@#dr1460-E {{previous declaration is here}}
+    //   cxx11-17-note@#cwg1460-E {{previous declaration is here}}
     friend constexpr F::F() noexcept;
   };
 
@@ -167,63 +167,63 @@ namespace dr1460 { // dr1460: 3.5
       union {
         int n = 0;
       };
-      union { // #dr1460-H-union
+      union { // #cwg1460-H-union
         int m;
       };
       constexpr H() {}
       // cxx11-17-error@-1 {{constexpr constructor that does not initialize all members is a C++20 extension}}
-      //   cxx11-17-note@#dr1460-H-union {{member not initialized by constructor}}
+      //   cxx11-17-note@#cwg1460-H-union {{member not initialized by constructor}}
       constexpr H(bool) : m(1) {}
       constexpr H(char) : n(1) {}
       // cxx11-17-error@-1 {{constexpr constructor that does not initialize all members is a C++20 extension}}
-      //   cxx11-17-note@#dr1460-H-union {{member not initialized by constructor}}
+      //   cxx11-17-note@#cwg1460-H-union {{member not initialized by constructor}}
       constexpr H(double) : m(1), n(1) {}
     };
   }
 
 #if __cplusplus >= 201403L
   template<typename T> constexpr bool check() {
-    T t; // #dr1460-t
+    T t; // #cwg1460-t
     return true;
   }
   static_assert(check<A>(), "");
-  static_assert(check<B>(), ""); // #dr1460-check-B
+  static_assert(check<B>(), ""); // #cwg1460-check-B
   // cxx14-17-error@-1 {{static assertion expression is not an integral constant expression}}
-  //   cxx14-17-note@#dr1460-t {{non-constexpr constructor 'B' cannot be used in a constant expression}}
-  //   cxx14-17-note@#dr1460-check-B {{in call to 'check<dr1460::B>()'}}
-  //   cxx14-17-note@#dr1460-B {{declared here}}
+  //   cxx14-17-note@#cwg1460-t {{non-constexpr constructor 'B' cannot be used in a constant expression}}
+  //   cxx14-17-note@#cwg1460-check-B {{in call to 'check<cwg1460::B>()'}}
+  //   cxx14-17-note@#cwg1460-B {{declared here}}
   static_assert(check<C>(), "");
   static_assert(check<D>(), "");
-  static_assert(check<E>(), ""); // #dr1460-check-E
+  static_assert(check<E>(), ""); // #cwg1460-check-E
   // cxx14-17-error@-1 {{static assertion expression is not an integral constant expression}}
-  //   cxx14-17-note@#dr1460-t {{non-constexpr constructor 'E' cannot be used in a constant expression}}
-  //   cxx14-17-note@#dr1460-check-E {{in call to 'check<dr1460::E>()'}}
-  //   cxx14-17-note@#dr1460-E {{declared here}}
+  //   cxx14-17-note@#cwg1460-t {{non-constexpr constructor 'E' cannot be used in a constant expression}}
+  //   cxx14-17-note@#cwg1460-check-E {{in call to 'check<cwg1460::E>()'}}
+  //   cxx14-17-note@#cwg1460-E {{declared here}}
   static_assert(check<F>(), "");
 #endif
 
   union G {
-    int a = 0; // #dr1460-G-a
+    int a = 0; // #cwg1460-G-a
     int b = 0;
     // expected-error@-1 {{initializing multiple members of union}}
-    //   expected-note@#dr1460-G-a {{previous initialization is here}}
+    //   expected-note@#cwg1460-G-a {{previous initialization is here}}
   };
   union H {
     union {
-      int a = 0; // #dr1460-H-a
+      int a = 0; // #cwg1460-H-a
     };
     union {
       int b = 0;
       // expected-error@-1 {{initializing multiple members of union}}
-      //   expected-note@#dr1460-H-a {{previous initialization is here}}
+      //   expected-note@#cwg1460-H-a {{previous initialization is here}}
     };
   };
   struct I {
     union {
-      int a = 0; // #dr1460-I-a
+      int a = 0; // #cwg1460-I-a
       int b = 0;
       // expected-error@-1 {{initializing multiple members of union}}
-      //   expected-note@#dr1460-I-a {{previous initialization is here}}
+      //   expected-note@#cwg1460-I-a {{previous initialization is here}}
     };
   };
   struct J {
@@ -374,9 +374,9 @@ namespace std {
 } // std
 #endif
 
-namespace dr1467 {  // dr1467: 3.7 c++11
+namespace cwg1467 {  // cwg1467: 3.7 c++11
 #if __cplusplus >= 201103L
-  // Note that the change to [over.best.ics] was partially undone by DR2076;
+  // Note that the change to [over.best.ics] was partially undone by CWG2076;
   // the resulting rule is tested with the tests for that change.
 
   // List-initialization of aggregate from same-type object
@@ -441,12 +441,12 @@ namespace dr1467 {  // dr1467: 3.7 c++11
     X x;
     X x2{x};
 
-    void f1(int); // #dr1467-f1
-    void f1(std::initializer_list<long>) = delete; // #dr1467-f1-deleted
+    void f1(int); // #cwg1467-f1
+    void f1(std::initializer_list<long>) = delete; // #cwg1467-f1-deleted
     void g1() { f1({42}); }
     // since-cxx11-error@-1 {{call to deleted function 'f1'}}
-    //   since-cxx11-note@#dr1467-f1 {{candidate function}}
-    //   since-cxx11-note@#dr1467-f1-deleted {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f1 {{candidate function}}
+    //   since-cxx11-note@#cwg1467-f1-deleted {{candidate function has been explicitly deleted}}
 
     template <class T, class U>
     struct Pair {
@@ -456,12 +456,12 @@ namespace dr1467 {  // dr1467: 3.7 c++11
       String(const char *);
     };
 
-    void f2(Pair<const char *, const char *>); // #dr1467-f2
-    void f2(std::initializer_list<String>) = delete; // #dr1467-f2-deleted
+    void f2(Pair<const char *, const char *>); // #cwg1467-f2
+    void f2(std::initializer_list<String>) = delete; // #cwg1467-f2-deleted
     void g2() { f2({"foo", "bar"}); }
     // since-cxx11-error@-1 {{call to deleted function 'f2'}}
-    //   since-cxx11-note@#dr1467-f2 {{candidate function}}
-    //   since-cxx11-note@#dr1467-f2-deleted {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f2 {{candidate function}}
+    //   since-cxx11-note@#cwg1467-f2-deleted {{candidate function has been explicitly deleted}}
   } // dr_example
 
   namespace nonaggregate {
@@ -522,88 +522,88 @@ namespace dr1467 {  // dr1467: 3.7 c++11
   // When the array size is 4 the call will attempt to bind an lvalue to an
   // rvalue and fail. Therefore #2 will be called. (rsmith will bring this
   // issue to CWG)
-  void f(const char(&&)[4]);              // #dr1467-f-char-4
-  void f(const char(&&)[5]) = delete;     // #dr1467-f-char-5
-  void f(const wchar_t(&&)[4]);           // #dr1467-f-wchar-4
-  void f(const wchar_t(&&)[5]) = delete;  // #dr1467-f-wchar-5
+  void f(const char(&&)[4]);              // #cwg1467-f-char-4
+  void f(const char(&&)[5]) = delete;     // #cwg1467-f-char-5
+  void f(const wchar_t(&&)[4]);           // #cwg1467-f-wchar-4
+  void f(const wchar_t(&&)[5]) = delete;  // #cwg1467-f-wchar-5
 #if __cplusplus >= 202002L
-  void f2(const char8_t(&&)[4]);          // #dr1467-f2-char8-4
-  void f2(const char8_t(&&)[5]) = delete; // #dr1467-f2-char8-5
+  void f2(const char8_t(&&)[4]);          // #cwg1467-f2-char8-4
+  void f2(const char8_t(&&)[5]) = delete; // #cwg1467-f2-char8-5
 #endif
-  void f(const char16_t(&&)[4]);          // #dr1467-f-char16-4
-  void f(const char16_t(&&)[5]) = delete; // #dr1467-f-char16-5
-  void f(const char32_t(&&)[4]);          // #dr1467-f-char32-4
-  void f(const char32_t(&&)[5]) = delete; // #dr1467-f-char32-5
+  void f(const char16_t(&&)[4]);          // #cwg1467-f-char16-4
+  void f(const char16_t(&&)[5]) = delete; // #cwg1467-f-char16-5
+  void f(const char32_t(&&)[4]);          // #cwg1467-f-char32-4
+  void f(const char32_t(&&)[5]) = delete; // #cwg1467-f-char32-5
   void g() {
     f({"abc"});
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1467-f-char-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx11-note@#dr1467-f-char-4 {{candidate function not viable: expects an rvalue for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f-char-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
     f({((("abc")))});
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1467-f-char-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx11-note@#dr1467-f-char-4 {{candidate function not viable: expects an rvalue for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f-char-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char[4]' to 'const char32_t' for 1st argument}}
     f({L"abc"});
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1467-f-wchar-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx11-note@#dr1467-f-char-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-4 {{candidate function not viable: expects an rvalue for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char32_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f-char-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-4 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-5 {{candidate function not viable: no known conversion from 'const wchar_t[4]' to 'const char32_t' for 1st argument}}
 #if __cplusplus >= 202002L
     f2({u8"abc"});
     // since-cxx20-error@-1 {{call to deleted function 'f2'}}
-    //   since-cxx20-note@#dr1467-f2-char8-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx20-note@#dr1467-f2-char8-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx20-note@#cwg1467-f2-char8-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx20-note@#cwg1467-f2-char8-4 {{candidate function not viable: expects an rvalue for 1st argument}}
 #endif
     f({uR"(abc)"});
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1467-f-char16-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx11-note@#dr1467-f-char-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-4 {{candidate function not viable: expects an rvalue for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char32_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f-char-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-4 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char32_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-5 {{candidate function not viable: no known conversion from 'const char16_t[4]' to 'const char32_t' for 1st argument}}
     f({(UR"(abc)")});
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1467-f-char32-5 {{candidate function has been explicitly deleted}}
-    //   since-cxx11-note@#dr1467-f-char-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const wchar_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char16_t' for 1st argument}}
-    //   since-cxx11-note@#dr1467-f-char32-4 {{candidate function not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-5 {{candidate function has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1467-f-char-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-wchar-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const wchar_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-4 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char16-5 {{candidate function not viable: no known conversion from 'const char32_t[4]' to 'const char16_t' for 1st argument}}
+    //   since-cxx11-note@#cwg1467-f-char32-4 {{candidate function not viable: expects an rvalue for 1st argument}}
   }
   } // namespace StringLiterals
 #endif
-} // dr1467
+} // cwg1467
 
-namespace dr1479 { // dr1479: 3.1
+namespace cwg1479 { // cwg1479: 3.1
 #if __cplusplus >= 201103L
   int operator"" _a(const char*, std::size_t = 0);
   // since-cxx11-error@-1 {{literal operator cannot have a default argument}}
 #endif
 }
 
-namespace dr1482 { // dr1482: 3.0
+namespace cwg1482 { // cwg1482: 3.0
                    // NB: sup 2516, test reused there
 #if __cplusplus >= 201103L
 template <typename T> struct S {
@@ -612,11 +612,11 @@ template <typename T> struct S {
 enum E2 : S<E2>::I { e };
 // since-cxx11-error@-1 {{use of undeclared identifier 'E2'}}
 #endif
-} // namespace dr1482
+} // namespace cwg1482
 
-namespace dr1487 { // dr1487: 3.3
+namespace cwg1487 { // cwg1487: 3.3
 #if __cplusplus >= 201103L
-struct A { // #dr1482-A
+struct A { // #cwg1482-A
   struct B {
     using A::A;
     // since-cxx11-error@-1 {{using declaration refers into 'A::', which is not a base class of 'B'}}
@@ -624,7 +624,7 @@ struct A { // #dr1482-A
 
   struct C : A {
   // since-cxx11-error@-1 {{base class has incomplete type}}
-  //   since-cxx11-note@#dr1482-A {{definition of 'dr1487::A' is not complete until the closing '}'}}
+  //   since-cxx11-note@#cwg1482-A {{definition of 'cwg1487::A' is not complete until the closing '}'}}
     using A::A;
     // since-cxx11-error@-1 {{using declaration refers into 'A::', which is not a base class of 'C'}}
   };
@@ -636,9 +636,9 @@ struct D : A {
   using A::A;
 };
 #endif
-} // namespace dr1487
+} // namespace cwg1487
 
-namespace dr1490 {  // dr1490: 3.7 c++11
+namespace cwg1490 {  // cwg1490: 3.7 c++11
 #if __cplusplus >= 201103L
   // List-initialization from a string literal
 
@@ -646,51 +646,51 @@ namespace dr1490 {  // dr1490: 3.7 c++11
   std::initializer_list<char>{"abc"};
   // since-cxx11-error@-1 {{expected unqualified-id}}}
 #endif
-} // dr1490
+} // cwg1490
 
-namespace dr1495 { // dr1495: 4
+namespace cwg1495 { // cwg1495: 4
 #if __cplusplus >= 201103L
   // Deduction succeeds in both directions.
-  template<typename T, typename U> struct A {}; // #dr1495-A
+  template<typename T, typename U> struct A {}; // #cwg1495-A
   template<typename T, typename U> struct A<U, T> {};
   // since-cxx11-error@-1 {{class template partial specialization is not more specialized than the primary template}}
-  //   since-cxx11-note@#dr1495-A {{template is declared here}}
+  //   since-cxx11-note@#cwg1495-A {{template is declared here}}
 
   // Primary template is more specialized.
-  template<typename, typename...> struct B {}; // #dr1495-B
+  template<typename, typename...> struct B {}; // #cwg1495-B
   template<typename ...Ts> struct B<Ts...> {};
   // since-cxx11-error@-1 {{class template partial specialization is not more specialized than the primary template}}
-  //   since-cxx11-note@#dr1495-B {{template is declared here}}
+  //   since-cxx11-note@#cwg1495-B {{template is declared here}}
 
   // Deduction fails in both directions.
-  template<int, typename, typename ...> struct C {}; // #dr1495-C
+  template<int, typename, typename ...> struct C {}; // #cwg1495-C
   template<typename ...Ts> struct C<0, Ts...> {};
   // since-cxx11-error@-1 {{class template partial specialization is not more specialized than the primary template}}
-  //   since-cxx11-note@#dr1495-C {{template is declared here}}
+  //   since-cxx11-note@#cwg1495-C {{template is declared here}}
 
 #if __cplusplus >= 201402L
   // Deduction succeeds in both directions.
-  template<typename T, typename U> int a; // #dr1495-a
+  template<typename T, typename U> int a; // #cwg1495-a
   template<typename T, typename U> int a<U, T>;
   // since-cxx14-error@-1 {{variable template partial specialization is not more specialized than the primary template}}
-  //   since-cxx14-note@#dr1495-a {{template is declared here}}
+  //   since-cxx14-note@#cwg1495-a {{template is declared here}}
 
   // Primary template is more specialized.
-  template<typename, typename...> int b; // #dr1495-b
+  template<typename, typename...> int b; // #cwg1495-b
   template<typename ...Ts> int b<Ts...>;
   // since-cxx14-error@-1 {{variable template partial specialization is not more specialized than the primary template}}
-  //   since-cxx14-note@#dr1495-b {{template is declared here}}
+  //   since-cxx14-note@#cwg1495-b {{template is declared here}}
 
   // Deduction fails in both directions.
-  template<int, typename, typename ...> int c; // #dr1495-c
+  template<int, typename, typename ...> int c; // #cwg1495-c
   template<typename ...Ts> int c<0, Ts...>;
   // since-cxx14-error@-1 {{variable template partial specialization is not more specialized than the primary template}}
-  //   since-cxx14-note@#dr1495-c {{template is declared here}}
+  //   since-cxx14-note@#cwg1495-c {{template is declared here}}
 #endif
 #endif
 }
 
-namespace dr1496 { // dr1496: no
+namespace cwg1496 { // cwg1496: no
 #if __cplusplus >= 201103L
 struct A {
     A() = delete;
diff --git a/clang/test/CXX/drs/dr15xx.cpp b/clang/test/CXX/drs/dr15xx.cpp
index 6e3ad41c748fb..21a392a5141e3 100644
--- a/clang/test/CXX/drs/dr15xx.cpp
+++ b/clang/test/CXX/drs/dr15xx.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-unknown %s -verify=expected,since-cxx23,since-cxx20,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++2c -triple x86_64-unknown-unknown %s -verify=expected,since-cxx23,since-cxx20,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr1512 { // dr1512: 4
+namespace cwg1512 { // cwg1512: 4
   void f(char *p) {
     if (p > 0) {}
     // expected-error@-1 {{ordered comparison between pointer and zero ('char *' and 'int')}}
@@ -33,11 +33,11 @@ namespace dr1512 { // dr1512: 4
   template<typename A, typename B, typename C> void composite_pointer_type_is_ord() {
     composite_pointer_type_is_base<A, B, C>();
 
-    typedef __typeof(val<A>() < val<B>()) cmp; // #dr1512-lt 
-    // since-cxx17-warning@#dr1512-lt {{ordered comparison of function pointers ('int (*)() noexcept' and 'int (*)()')}}
-    //   since-cxx17-note@#dr1512-noexcept-1st {{in instantiation of function template specialization 'dr1512::composite_pointer_type_is_ord<int (*)() noexcept, int (*)(), int (*)()>' requested here}}
-    // since-cxx17-warning@#dr1512-lt {{ordered comparison of function pointers ('int (*)()' and 'int (*)() noexcept')}}
-    //   since-cxx17-note@#dr1512-noexcept-2nd {{in instantiation of function template specialization 'dr1512::composite_pointer_type_is_ord<int (*)(), int (*)() noexcept, int (*)()>' requested here}}
+    typedef __typeof(val<A>() < val<B>()) cmp; // #cwg1512-lt 
+    // since-cxx17-warning@#cwg1512-lt {{ordered comparison of function pointers ('int (*)() noexcept' and 'int (*)()')}}
+    //   since-cxx17-note@#cwg1512-noexcept-1st {{in instantiation of function template specialization 'cwg1512::composite_pointer_type_is_ord<int (*)() noexcept, int (*)(), int (*)()>' requested here}}
+    // since-cxx17-warning@#cwg1512-lt {{ordered comparison of function pointers ('int (*)()' and 'int (*)() noexcept')}}
+    //   since-cxx17-note@#cwg1512-noexcept-2nd {{in instantiation of function template specialization 'cwg1512::composite_pointer_type_is_ord<int (*)(), int (*)() noexcept, int (*)()>' requested here}}
     typedef __typeof(val<A>() <= val<B>()) cmp;
     // since-cxx17-warning@-1 {{ordered comparison of function pointers ('int (*)() noexcept' and 'int (*)()')}}
     // since-cxx17-warning@-2 {{ordered comparison of function pointers ('int (*)()' and 'int (*)() noexcept')}}
@@ -95,8 +95,8 @@ namespace dr1512 { // dr1512: 4
     // since-cxx20-warning@-1 {{volatile-qualified return type 'volatile int' is deprecated}}
 
 #if __cplusplus >= 201703L
-    composite_pointer_type_is_ord<int (*)() noexcept, int (*)(), int (*)()>(); // #dr1512-noexcept-1st
-    composite_pointer_type_is_ord<int (*)(), int (*)() noexcept, int (*)()>(); // #dr1512-noexcept-2nd
+    composite_pointer_type_is_ord<int (*)() noexcept, int (*)(), int (*)()>(); // #cwg1512-noexcept-1st
+    composite_pointer_type_is_ord<int (*)(), int (*)() noexcept, int (*)()>(); // #cwg1512-noexcept-2nd
     composite_pointer_type_is_unord<int (A::*)() noexcept, int (A::*)(), int (A::*)()>();
     composite_pointer_type_is_unord<int (A::*)(), int (A::*)() noexcept, int (A::*)()>();
     // FIXME: This looks like a standard defect; these should probably all have type 'int (B::*)()'.
@@ -129,7 +129,7 @@ namespace dr1512 { // dr1512: 4
   }
 
 #if __cplusplus >= 201103L
-  template<typename T> struct Wrap { operator T(); }; // #dr1512-Wrap
+  template<typename T> struct Wrap { operator T(); }; // #cwg1512-Wrap
   void test_overload() {
     using nullptr_t = decltype(nullptr);
     void(Wrap<nullptr_t>() == Wrap<nullptr_t>());
@@ -143,75 +143,75 @@ namespace dr1512 { // dr1512: 4
     void(Wrap<nullptr_t>() >= Wrap<nullptr_t>());
     // since-cxx11-error@-1 {{invalid operands to binary expression ('Wrap<nullptr_t>' (aka 'Wrap<std::nullptr_t>') and 'Wrap<nullptr_t>' (aka 'Wrap<std::nullptr_t>'))}}
 
-    // Under dr1213, this is ill-formed: we select the builtin operator<(int*, int*)
+    // Under cwg1213, this is ill-formed: we select the builtin operator<(int*, int*)
     // but then only convert as far as 'nullptr_t', which we then can't convert to 'int*'.
     void(Wrap<nullptr_t>() == Wrap<int*>());
     void(Wrap<nullptr_t>() != Wrap<int*>());
     void(Wrap<nullptr_t>() < Wrap<int*>());
     // since-cxx11-error@-1 {{invalid operands to binary expression ('Wrap<nullptr_t>' (aka 'Wrap<std::nullptr_t>') and 'Wrap<int *>')}}
-    //   since-cxx11-note@#dr1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
-    //   since-cxx11-note@#dr1512-Wrap {{second operand was implicitly converted to type 'int *'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{second operand was implicitly converted to type 'int *'}}
     void(Wrap<nullptr_t>() > Wrap<int*>());
     // since-cxx11-error@-1 {{invalid operands}}
-    //   since-cxx11-note@#dr1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
-    //   since-cxx11-note@#dr1512-Wrap{{second operand was implicitly converted to type 'int *'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
+    //   since-cxx11-note@#cwg1512-Wrap{{second operand was implicitly converted to type 'int *'}}
     void(Wrap<nullptr_t>() <= Wrap<int*>());
     // since-cxx11-error@-1 {{invalid operands}}
-    //   since-cxx11-note@#dr1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
-    //   since-cxx11-note@#dr1512-Wrap {{second operand was implicitly converted to type 'int *'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{second operand was implicitly converted to type 'int *'}}
     void(Wrap<nullptr_t>() >= Wrap<int*>());
     // since-cxx11-error@-1 {{invalid operands}}
-    //   since-cxx11-note@#dr1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
-    //   since-cxx11-note@#dr1512-Wrap {{second operand was implicitly converted to type 'int *'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{first operand was implicitly converted to type 'std::nullptr_t'}}
+    //   since-cxx11-note@#cwg1512-Wrap {{second operand was implicitly converted to type 'int *'}}
   }
 #endif
 }
 
-namespace dr1514 { // dr1514: 11
+namespace cwg1514 { // cwg1514: 11
 #if __cplusplus >= 201103L
   struct S {
-    enum E : int {}; // #dr1514-E
+    enum E : int {}; // #cwg1514-E
     enum E : int {};
     // since-cxx11-error@-1 {{redefinition of 'E'}}
-    //   since-cxx11-note@#dr1514-E {{previous definition is here}}
+    //   since-cxx11-note@#cwg1514-E {{previous definition is here}}
   };
   S::E se; // OK, complete type, not zero-width bitfield.
 
-  // The behavior in other contexts is superseded by DR1966.
+  // The behavior in other contexts is superseded by CWG1966.
 #endif
 }
 
-namespace dr1518 { // dr1518: 4
+namespace cwg1518 { // cwg1518: 4
 #if __cplusplus >= 201103L
-struct Z0 { // #dr1518-Z0
-  explicit Z0() = default; // #dr1518-Z0-ctor
+struct Z0 { // #cwg1518-Z0
+  explicit Z0() = default; // #cwg1518-Z0-ctor
 };
-struct Z { // #dr1518-Z
-  explicit Z(); // #dr1518-Z-ctor
-  explicit Z(int); // #dr1518-Z-int
-  explicit Z(int, int); // #dr1518-Z-int-int
+struct Z { // #cwg1518-Z
+  explicit Z(); // #cwg1518-Z-ctor
+  explicit Z(int); // #cwg1518-Z-int
+  explicit Z(int, int); // #cwg1518-Z-int-int
 };
-template <class T> int Eat(T); // #dr1518-Eat
+template <class T> int Eat(T); // #cwg1518-Eat
 Z0 a;
 Z0 b{};
 Z0 c = {};
 // since-cxx11-error@-1 {{chosen constructor is explicit in copy-initialization}}
-//   since-cxx11-note@#dr1518-Z0-ctor {{explicit constructor declared here}}
+//   since-cxx11-note@#cwg1518-Z0-ctor {{explicit constructor declared here}}
 int i = Eat<Z0>({});
 // since-cxx11-error@-1 {{no matching function for call to 'Eat'}}
-//   since-cxx11-note@#dr1518-Eat {{candidate function template not viable: cannot convert initializer list argument to 'Z0'}}
+//   since-cxx11-note@#cwg1518-Eat {{candidate function template not viable: cannot convert initializer list argument to 'Z0'}}
 
 Z c2 = {};
 // since-cxx11-error@-1 {{chosen constructor is explicit in copy-initialization}}
-//   since-cxx11-note@#dr1518-Z-ctor {{explicit constructor declared here}}
+//   since-cxx11-note@#cwg1518-Z-ctor {{explicit constructor declared here}}
 int i2 = Eat<Z>({});
 // since-cxx11-error@-1 {{no matching function for call to 'Eat'}}
-//   since-cxx11-note@#dr1518-Eat {{candidate function template not viable: cannot convert initializer list argument to 'Z'}}
+//   since-cxx11-note@#cwg1518-Eat {{candidate function template not viable: cannot convert initializer list argument to 'Z'}}
 Z a1 = 1;
 // since-cxx11-error@-1 {{no viable conversion from 'int' to 'Z'}}
-//   since-cxx11-note@#dr1518-Z {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int' to 'const Z &' for 1st argument}}
-//   since-cxx11-note@#dr1518-Z {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int' to 'Z &&' for 1st argument}}
-//   since-cxx11-note@#dr1518-Z-int {{explicit constructor is not a candidate}}
+//   since-cxx11-note@#cwg1518-Z {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int' to 'const Z &' for 1st argument}}
+//   since-cxx11-note@#cwg1518-Z {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int' to 'Z &&' for 1st argument}}
+//   since-cxx11-note@#cwg1518-Z-int {{explicit constructor is not a candidate}}
 Z a3 = Z(1);
 Z a2(1);
 Z *p = new Z(1);
@@ -219,129 +219,129 @@ Z a4 = (Z)1;
 Z a5 = static_cast<Z>(1);
 Z a6 = {4, 3};
 // since-cxx11-error@-1 {{chosen constructor is explicit in copy-initialization}}
-//   since-cxx11-note@#dr1518-Z-int-int {{explicit constructor declared here}}
+//   since-cxx11-note@#cwg1518-Z-int-int {{explicit constructor declared here}}
 
-struct UserProvidedBaseCtor { // #dr1518-U
+struct UserProvidedBaseCtor { // #cwg1518-U
   UserProvidedBaseCtor() {}
 };
-struct DoesntInheritCtor : UserProvidedBaseCtor { // #dr1518-D-U
+struct DoesntInheritCtor : UserProvidedBaseCtor { // #cwg1518-D-U
   int x;
 };
 DoesntInheritCtor I{{}, 42};
 // cxx11-14-error@-1 {{no matching constructor for initialization of 'DoesntInheritCtor'}}
-//   cxx11-14-note@#dr1518-D-U {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
-//   cxx11-14-note@#dr1518-D-U {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
-//   cxx11-14-note@#dr1518-D-U {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 2 were provided}}
+//   cxx11-14-note@#cwg1518-D-U {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
+//   cxx11-14-note@#cwg1518-D-U {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
+//   cxx11-14-note@#cwg1518-D-U {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 2 were provided}}
 
-struct BaseCtor { BaseCtor() = default; }; // #dr1518-BC
-struct InheritsCtor : BaseCtor { // #dr1518-I
-  using BaseCtor::BaseCtor;      // #dr1518-I-using
+struct BaseCtor { BaseCtor() = default; }; // #cwg1518-BC
+struct InheritsCtor : BaseCtor { // #cwg1518-I
+  using BaseCtor::BaseCtor;      // #cwg1518-I-using
   int x;
 };
 InheritsCtor II = {{}, 42};
 // since-cxx11-error@-1 {{no matching constructor for initialization of 'InheritsCtor'}}
-//   since-cxx11-note@#dr1518-BC {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
-//   since-cxx11-note@#dr1518-I-using {{constructor from base class 'BaseCtor' inherited here}}
-//   since-cxx11-note@#dr1518-BC {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
-//   since-cxx11-note@#dr1518-I-using {{constructor from base class 'BaseCtor' inherited here}}
-//   since-cxx11-note@#dr1518-I {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
-//   since-cxx11-note@#dr1518-I {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
-//   since-cxx11-note@#dr1518-I {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 2 were provided}}
+//   since-cxx11-note@#cwg1518-BC {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
+//   since-cxx11-note@#cwg1518-I-using {{constructor from base class 'BaseCtor' inherited here}}
+//   since-cxx11-note@#cwg1518-BC {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
+//   since-cxx11-note@#cwg1518-I-using {{constructor from base class 'BaseCtor' inherited here}}
+//   since-cxx11-note@#cwg1518-I {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
+//   since-cxx11-note@#cwg1518-I {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
+//   since-cxx11-note@#cwg1518-I {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 2 were provided}}
 
 namespace std_example {
   struct A {
-    explicit A() = default; // #dr1518-A
+    explicit A() = default; // #cwg1518-A
   };
 
   struct B : A {
-    explicit B() = default; // #dr1518-B
+    explicit B() = default; // #cwg1518-B
   };
 
   struct C {
-    explicit C(); // #dr1518-C
+    explicit C(); // #cwg1518-C
   };
 
   struct D : A {
     C c;
-    explicit D() = default; // #dr1518-D
+    explicit D() = default; // #cwg1518-D
   };
 
   template <typename T> void f() {
     T t; // ok
     T u{}; // ok
-    T v = {}; // #dr1518-v
-    // since-cxx11-error@#dr1518-v {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-f-A {{in instantiation of function template specialization 'dr1518::std_example::f<dr1518::std_example::A>' requested here}}
-    //   since-cxx11-note@#dr1518-A {{explicit constructor declared here}}
-    // since-cxx11-error@#dr1518-v {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-f-B {{in instantiation of function template specialization 'dr1518::std_example::f<dr1518::std_example::B>' requested here}}
-    //   since-cxx11-note@#dr1518-B {{explicit constructor declared here}}
-    // since-cxx11-error@#dr1518-v {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-f-C {{in instantiation of function template specialization 'dr1518::std_example::f<dr1518::std_example::C>' requested here}}
-    //   since-cxx11-note@#dr1518-C {{explicit constructor declared here}}
-    // since-cxx11-error@#dr1518-v {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-f-D {{in instantiation of function template specialization 'dr1518::std_example::f<dr1518::std_example::D>' requested here}}
-    //   since-cxx11-note@#dr1518-D {{explicit constructor declared here}}
+    T v = {}; // #cwg1518-v
+    // since-cxx11-error@#cwg1518-v {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-f-A {{in instantiation of function template specialization 'cwg1518::std_example::f<cwg1518::std_example::A>' requested here}}
+    //   since-cxx11-note@#cwg1518-A {{explicit constructor declared here}}
+    // since-cxx11-error@#cwg1518-v {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-f-B {{in instantiation of function template specialization 'cwg1518::std_example::f<cwg1518::std_example::B>' requested here}}
+    //   since-cxx11-note@#cwg1518-B {{explicit constructor declared here}}
+    // since-cxx11-error@#cwg1518-v {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-f-C {{in instantiation of function template specialization 'cwg1518::std_example::f<cwg1518::std_example::C>' requested here}}
+    //   since-cxx11-note@#cwg1518-C {{explicit constructor declared here}}
+    // since-cxx11-error@#cwg1518-v {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-f-D {{in instantiation of function template specialization 'cwg1518::std_example::f<cwg1518::std_example::D>' requested here}}
+    //   since-cxx11-note@#cwg1518-D {{explicit constructor declared here}}
   }
   template <typename T> void g() {
-    void x(T t); // #dr1518-x
-    x({}); // #dr1518-x-call
-    // since-cxx11-error@#dr1518-x-call {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-g-A {{in instantiation of function template specialization 'dr1518::std_example::g<dr1518::std_example::A>' requested here}}
-    //   since-cxx11-note@#dr1518-A {{explicit constructor declared here}}
-    //   since-cxx11-note@#dr1518-x {{passing argument to parameter 't' here}}
-    // since-cxx11-error@#dr1518-x-call {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-g-B {{in instantiation of function template specialization 'dr1518::std_example::g<dr1518::std_example::B>' requested here}}
-    //   since-cxx11-note@#dr1518-B {{explicit constructor declared here}}
-    //   since-cxx11-note@#dr1518-x {{passing argument to parameter 't' here}}
-    // since-cxx11-error@#dr1518-x-call {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-g-C {{in instantiation of function template specialization 'dr1518::std_example::g<dr1518::std_example::C>' requested here}}
-    //   since-cxx11-note@#dr1518-C {{explicit constructor declared here}}
-    //   since-cxx11-note@#dr1518-x {{passing argument to parameter 't' here}}
-    // since-cxx11-error@#dr1518-x-call {{chosen constructor is explicit in copy-initialization}}
-    //   since-cxx11-note@#dr1518-g-D {{in instantiation of function template specialization 'dr1518::std_example::g<dr1518::std_example::D>' requested here}}
-    //   since-cxx11-note@#dr1518-D {{explicit constructor declared here}}
-    //   since-cxx11-note@#dr1518-x {{passing argument to parameter 't' here}}
+    void x(T t); // #cwg1518-x
+    x({}); // #cwg1518-x-call
+    // since-cxx11-error@#cwg1518-x-call {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-g-A {{in instantiation of function template specialization 'cwg1518::std_example::g<cwg1518::std_example::A>' requested here}}
+    //   since-cxx11-note@#cwg1518-A {{explicit constructor declared here}}
+    //   since-cxx11-note@#cwg1518-x {{passing argument to parameter 't' here}}
+    // since-cxx11-error@#cwg1518-x-call {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-g-B {{in instantiation of function template specialization 'cwg1518::std_example::g<cwg1518::std_example::B>' requested here}}
+    //   since-cxx11-note@#cwg1518-B {{explicit constructor declared here}}
+    //   since-cxx11-note@#cwg1518-x {{passing argument to parameter 't' here}}
+    // since-cxx11-error@#cwg1518-x-call {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-g-C {{in instantiation of function template specialization 'cwg1518::std_example::g<cwg1518::std_example::C>' requested here}}
+    //   since-cxx11-note@#cwg1518-C {{explicit constructor declared here}}
+    //   since-cxx11-note@#cwg1518-x {{passing argument to parameter 't' here}}
+    // since-cxx11-error@#cwg1518-x-call {{chosen constructor is explicit in copy-initialization}}
+    //   since-cxx11-note@#cwg1518-g-D {{in instantiation of function template specialization 'cwg1518::std_example::g<cwg1518::std_example::D>' requested here}}
+    //   since-cxx11-note@#cwg1518-D {{explicit constructor declared here}}
+    //   since-cxx11-note@#cwg1518-x {{passing argument to parameter 't' here}}
   }
 
   void test() {
-    f<A>(); // #dr1518-f-A
-    f<B>(); // #dr1518-f-B
-    f<C>(); // #dr1518-f-C
-    f<D>(); // #dr1518-f-D
-    g<A>(); // #dr1518-g-A
-    g<B>(); // #dr1518-g-B
-    g<C>(); // #dr1518-g-C
-    g<D>(); // #dr1518-g-D
+    f<A>(); // #cwg1518-f-A
+    f<B>(); // #cwg1518-f-B
+    f<C>(); // #cwg1518-f-C
+    f<D>(); // #cwg1518-f-D
+    g<A>(); // #cwg1518-g-A
+    g<B>(); // #cwg1518-g-B
+    g<C>(); // #cwg1518-g-C
+    g<D>(); // #cwg1518-g-D
   }
 }
 #endif // __cplusplus >= 201103L
 }
 
-namespace dr1550 { // dr1550: 3.4
+namespace cwg1550 { // cwg1550: 3.4
   int f(bool b, int n) {
     return (b ? (throw 0) : n) + (b ? n : (throw 0));
   }
 }
 
-namespace dr1558 { // dr1558: 12
+namespace cwg1558 { // cwg1558: 12
 #if __cplusplus >= 201103L
   template<class T, class...> using first_of = T;
-  template<class T> first_of<void, typename T::type> f(int); // #dr1558-f 
-  template<class T> void f(...) = delete; // #dr1558-f-deleted
+  template<class T> first_of<void, typename T::type> f(int); // #cwg1558-f 
+  template<class T> void f(...) = delete; // #cwg1558-f-deleted
 
   struct X { typedef void type; };
   void test() {
     f<X>(0);
     f<int>(0);
     // since-cxx11-error@-1 {{call to deleted function 'f'}}
-    //   since-cxx11-note@#dr1558-f-deleted {{candidate function [with T = int] has been explicitly deleted}}
-    //   since-cxx11-note@#dr1558-f {{candidate template ignored: substitution failure [with T = int]: type 'int' cannot be used prior to '::' because it has no members}}
+    //   since-cxx11-note@#cwg1558-f-deleted {{candidate function [with T = int] has been explicitly deleted}}
+    //   since-cxx11-note@#cwg1558-f {{candidate template ignored: substitution failure [with T = int]: type 'int' cannot be used prior to '::' because it has no members}}
   }
 #endif
 }
 
-namespace dr1560 { // dr1560: 3.5
+namespace cwg1560 { // cwg1560: 3.5
   void f(bool b, int n) {
     (b ? throw 0 : n) = (b ? n : throw 0) = 0;
   }
@@ -350,7 +350,7 @@ namespace dr1560 { // dr1560: 3.5
   const X &x = true ? get() : throw 0;
 }
 
-namespace dr1563 { // dr1563: yes
+namespace cwg1563 { // cwg1563: yes
 #if __cplusplus >= 201103L
   double bar(double) { return 0.0; }
   float bar(float) { return 0.0f; }
@@ -360,7 +360,7 @@ namespace dr1563 { // dr1563: yes
 #endif
 }
 
-namespace dr1567 { // dr1567: 3.3
+namespace cwg1567 { // cwg1567: 3.3
 #if __cplusplus >= 201103L
 struct B;
 struct A {
@@ -368,12 +368,12 @@ struct A {
   A(const B&) = delete;
   A(A&&);
   A(B&&) = delete;
-  A(int); // #dr1567-A-int
+  A(int); // #cwg1567-A-int
 };
 
-struct B: A { // #dr1567-B
-  using A::A; // #dr1567-using-A
-  B(double); // #dr1567-B-double
+struct B: A { // #cwg1567-B
+  using A::A; // #cwg1567-using-A
+  B(double); // #cwg1567-B-double
 };
 
 A a{0};
@@ -384,22 +384,22 @@ B b3{B{1.0}};
 // Good, copy/move ctors are not inherited
 B b4{a};
 // since-cxx11-error@-1 {{no matching constructor for initialization of 'B'}}
-//   since-cxx11-note@#dr1567-A-int {{candidate inherited constructor not viable: no known conversion from 'A' to 'int' for 1st argument}}
-//   since-cxx11-note@#dr1567-using-A {{constructor from base class 'A' inherited here}}
-//   since-cxx11-note@#dr1567-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'A' to 'const B' for 1st argument}}
-//   since-cxx11-note@#dr1567-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'A' to 'B' for 1st argument}}
-//   since-cxx11-note@#dr1567-B-double {{candidate constructor not viable: no known conversion from 'A' to 'double' for 1st argument}}
+//   since-cxx11-note@#cwg1567-A-int {{candidate inherited constructor not viable: no known conversion from 'A' to 'int' for 1st argument}}
+//   since-cxx11-note@#cwg1567-using-A {{constructor from base class 'A' inherited here}}
+//   since-cxx11-note@#cwg1567-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'A' to 'const B' for 1st argument}}
+//   since-cxx11-note@#cwg1567-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'A' to 'B' for 1st argument}}
+//   since-cxx11-note@#cwg1567-B-double {{candidate constructor not viable: no known conversion from 'A' to 'double' for 1st argument}}
 B b5{A{0}};
 // since-cxx11-error@-1 {{no matching constructor for initialization of 'B'}}
-//   since-cxx11-note@#dr1567-A-int {{candidate inherited constructor not viable: no known conversion from 'A' to 'int' for 1st argument}}
-//   since-cxx11-note@#dr1567-using-A {{constructor from base class 'A' inherited here}}
-//   since-cxx11-note@#dr1567-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'A' to 'const B' for 1st argument}}
-//   since-cxx11-note@#dr1567-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'A' to 'B' for 1st argument}}
-//   since-cxx11-note@#dr1567-B-double {{candidate constructor not viable: no known conversion from 'A' to 'double' for 1st argument}}
+//   since-cxx11-note@#cwg1567-A-int {{candidate inherited constructor not viable: no known conversion from 'A' to 'int' for 1st argument}}
+//   since-cxx11-note@#cwg1567-using-A {{constructor from base class 'A' inherited here}}
+//   since-cxx11-note@#cwg1567-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'A' to 'const B' for 1st argument}}
+//   since-cxx11-note@#cwg1567-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'A' to 'B' for 1st argument}}
+//   since-cxx11-note@#cwg1567-B-double {{candidate constructor not viable: no known conversion from 'A' to 'double' for 1st argument}}
 #endif
 }
 
-namespace dr1573 { // dr1573: 3.9
+namespace cwg1573 { // cwg1573: 3.9
 #if __cplusplus >= 201103L
   // ellipsis is inherited (p0136r1 supersedes this part).
   struct A { A(); A(int, char, ...); };
@@ -407,38 +407,38 @@ namespace dr1573 { // dr1573: 3.9
   B b(1, 'x', 4.0, "hello"); // ok
 
   // inherited constructor is effectively constexpr if the user-written constructor would be
-  struct C { C(); constexpr C(int) {} }; // #dr1573-C
+  struct C { C(); constexpr C(int) {} }; // #cwg1573-C
   struct D : C { using C::C; };
   constexpr D d = D(0); // ok
-  struct E : C { using C::C; A a; }; // #dr1573-E
+  struct E : C { using C::C; A a; }; // #cwg1573-E
   constexpr E e = E(0);
   // since-cxx11-error@-1 {{constexpr variable cannot have non-literal type 'const E'}}
-  //   since-cxx11-note@#dr1573-E {{'E' is not literal because it has data member 'a' of non-literal type 'A'}}
+  //   since-cxx11-note@#cwg1573-E {{'E' is not literal because it has data member 'a' of non-literal type 'A'}}
 
   // FIXME: This diagnostic is pretty bad; we should explain that the problem
   // is that F::c would be initialized by a non-constexpr constructor.
-  struct F : C { using C::C; C c; }; // #dr1573-F
+  struct F : C { using C::C; C c; }; // #cwg1573-F
   constexpr F f = F(0);
   // since-cxx11-error@-1 {{constexpr variable 'f' must be initialized by a constant expression}}
   //   cxx11-20-note@-2 {{constructor inherited from base class 'C' cannot be used in a constant expression; derived class cannot be implicitly initialized}}
   //   since-cxx23-note@-3 {{in implicit initialization for inherited constructor of 'F'}}
-  //   since-cxx23-note@#dr1573-F {{non-constexpr constructor 'C' cannot be used in a constant expression}}
-  //   cxx11-20-note@#dr1573-F {{declared here}}
-  //   since-cxx23-note@#dr1573-C {{declared here}}
+  //   since-cxx23-note@#cwg1573-F {{non-constexpr constructor 'C' cannot be used in a constant expression}}
+  //   cxx11-20-note@#cwg1573-F {{declared here}}
+  //   since-cxx23-note@#cwg1573-C {{declared here}}
 
   // inherited constructor is effectively deleted if the user-written constructor would be
   struct G { G(int); };
-  struct H : G { using G::G; G g; }; // #dr1573-H
+  struct H : G { using G::G; G g; }; // #cwg1573-H
   H h(0);
   // since-cxx11-error@-1 {{constructor inherited by 'H' from base class 'G' is implicitly deleted}}
-  //   since-cxx11-note@#dr1573-H {{constructor inherited by 'H' is implicitly deleted because field 'g' has no default constructor}}
+  //   since-cxx11-note@#cwg1573-H {{constructor inherited by 'H' is implicitly deleted because field 'g' has no default constructor}}
 
   // deleted definition of constructor is inherited
-  struct I { I(int) = delete; }; // #dr1573-I
+  struct I { I(int) = delete; }; // #cwg1573-I
   struct J : I { using I::I; };
   J j(0);
   // since-cxx11-error@-1 {{call to deleted constructor of 'J'}}
-  //   since-cxx11-note@#dr1573-I {{'I' has been explicitly marked deleted here}}
+  //   since-cxx11-note@#cwg1573-I {{'I' has been explicitly marked deleted here}}
 #endif
 }
 
@@ -483,18 +483,18 @@ namespace std {
 } // std
 #endif
 
-namespace dr1579 { // dr1579: 3.9
+namespace cwg1579 { // cwg1579: 3.9
 #if __cplusplus >= 201103L
 template<class T>
 struct GenericMoveOnly {
   GenericMoveOnly();
-  template<class U> GenericMoveOnly(const GenericMoveOnly<U> &) = delete; // #dr1579-deleted-U
-  GenericMoveOnly(const int &) = delete; // #dr1579-deleted-int
+  template<class U> GenericMoveOnly(const GenericMoveOnly<U> &) = delete; // #cwg1579-deleted-U
+  GenericMoveOnly(const int &) = delete; // #cwg1579-deleted-int
   template<class U> GenericMoveOnly(GenericMoveOnly<U> &&);
   GenericMoveOnly(int &&);
 };
 
-GenericMoveOnly<float> DR1579_Eligible(GenericMoveOnly<char> CharMO) {
+GenericMoveOnly<float> CWG1579_Eligible(GenericMoveOnly<char> CharMO) {
   int i;
   GenericMoveOnly<char> GMO;
 
@@ -510,7 +510,7 @@ GenericMoveOnly<float> DR1579_Eligible(GenericMoveOnly<char> CharMO) {
 
 GenericMoveOnly<char> GlobalMO;
 
-GenericMoveOnly<float> DR1579_Ineligible(int &AnInt,
+GenericMoveOnly<float> CWG1579_Ineligible(int &AnInt,
                                           GenericMoveOnly<char> &CharMO) {
   static GenericMoveOnly<char> StaticMove;
   extern GenericMoveOnly<char> ExternMove;
@@ -518,63 +518,63 @@ GenericMoveOnly<float> DR1579_Ineligible(int &AnInt,
   if (0)
     return AnInt;
     // since-cxx11-error@-1 {{conversion function from 'int' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-int {{'GenericMoveOnly' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-int {{'GenericMoveOnly' has been explicitly marked deleted here}}
   else if (0)
     return GlobalMO;
     // since-cxx11-error@-1 {{conversion function from 'GenericMoveOnly<char>' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
   else if (0)
     return StaticMove;
     // since-cxx11-error@-1 {{conversion function from 'GenericMoveOnly<char>' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
   else if (0)
     return ExternMove;
     // since-cxx11-error@-1 {{conversion function from 'GenericMoveOnly<char>' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
   else if (0)
     return AnInt;
     // since-cxx11-error@-1 {{conversion function from 'int' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-int {{'GenericMoveOnly' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-int {{'GenericMoveOnly' has been explicitly marked deleted here}}
   else
     return CharMO;
     // since-cxx11-error@-1 {{conversion function from 'GenericMoveOnly<char>' to 'GenericMoveOnly<float>' invokes a deleted function}}
-    //   since-cxx11-note@#dr1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
+    //   since-cxx11-note@#cwg1579-deleted-U {{'GenericMoveOnly<char>' has been explicitly marked deleted here}}
 }
 
-auto DR1579_lambda_valid = [](GenericMoveOnly<float> mo) ->
+auto CWG1579_lambda_valid = [](GenericMoveOnly<float> mo) ->
   GenericMoveOnly<char> {
   return mo;
 };
 
-auto DR1579_lambda_invalid = []() -> GenericMoveOnly<char> {
+auto CWG1579_lambda_invalid = []() -> GenericMoveOnly<char> {
   static GenericMoveOnly<float> mo;
   return mo;
   // since-cxx11-error@-1 {{conversion function from 'GenericMoveOnly<float>' to 'GenericMoveOnly<char>' invokes a deleted function}}
-  //   since-cxx11-note@#dr1579-deleted-U {{'GenericMoveOnly<float>' has been explicitly marked deleted here}}
+  //   since-cxx11-note@#cwg1579-deleted-U {{'GenericMoveOnly<float>' has been explicitly marked deleted here}}
 };
 #endif
-} // end namespace dr1579
+} // end namespace cwg1579
 
-namespace dr1584 { // dr1584: 7 drafting 2015-05
+namespace cwg1584 { // cwg1584: 7 drafting 2015-05
 #if __cplusplus >= 201103L
   // Deducing function types from cv-qualified types
-  template<typename T> void f(const T *); // #dr1584-f
+  template<typename T> void f(const T *); // #cwg1584-f
   template<typename T> void g(T *, const T * = 0);
   template<typename T> void h(T *) { T::error; }
   // since-cxx11-error@-1 {{type 'void ()' cannot be used prior to '::' because it has no members}}
-  //   since-cxx11-note@#dr1584-h {{in instantiation of function template specialization 'dr1584::h<void ()>' requested here}}
+  //   since-cxx11-note@#cwg1584-h {{in instantiation of function template specialization 'cwg1584::h<void ()>' requested here}}
   template<typename T> void h(const T *);
   void i() {
     f(&i);
     // since-cxx11-error@-1 {{no matching function for call to 'f'}}
-    //   since-cxx11-note@#dr1584-f {{candidate template ignored: could not match 'const T *' against 'void (*)()'}}
+    //   since-cxx11-note@#cwg1584-f {{candidate template ignored: could not match 'const T *' against 'void (*)()'}}
     g(&i);
-    h(&i); // #dr1584-h
+    h(&i); // #cwg1584-h
   }
 #endif
 }
 
-namespace dr1589 {   // dr1589: 3.7 c++11
+namespace cwg1589 {   // cwg1589: 3.7 c++11
 #if __cplusplus >= 201103L
   // Ambiguous ranking of list-initialization sequences
 
@@ -595,33 +595,33 @@ namespace dr1589 {   // dr1589: 3.7 c++11
 
   namespace with_error {
     void f0(long);
-    void f0(std::initializer_list<int>);          // #dr1589-f0-ilist
-    void f0(std::initializer_list<int>, int = 0); // #dr1589-f0-ilist-int
+    void f0(std::initializer_list<int>);          // #cwg1589-f0-ilist
+    void f0(std::initializer_list<int>, int = 0); // #cwg1589-f0-ilist-int
     void g0() { f0({1L}); }
     // since-cxx11-error@-1 {{call to 'f0' is ambiguous}}
-    //   since-cxx11-note@#dr1589-f0-ilist {{candidate function}}
-    //   since-cxx11-note@#dr1589-f0-ilist-int {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f0-ilist {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f0-ilist-int {{candidate function}}
 
     void f1(int);
-    void f1(std::initializer_list<long>);          // #dr1589-f1-ilist
-    void f1(std::initializer_list<long>, int = 0); // #dr1589-f1-ilist-long
+    void f1(std::initializer_list<long>);          // #cwg1589-f1-ilist
+    void f1(std::initializer_list<long>, int = 0); // #cwg1589-f1-ilist-long
     void g1() { f1({42}); }
     // since-cxx11-error@-1 {{call to 'f1' is ambiguous}}
-    //   since-cxx11-note@#dr1589-f1-ilist {{candidate function}}
-    //   since-cxx11-note@#dr1589-f1-ilist-long {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f1-ilist {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f1-ilist-long {{candidate function}}
 
     void f2(std::pair<const char*, const char*>);
-    void f2(std::initializer_list<std::string>);          // #dr1589-f2-ilist
-    void f2(std::initializer_list<std::string>, int = 0); // #dr1589-f2-ilist-int
+    void f2(std::initializer_list<std::string>);          // #cwg1589-f2-ilist
+    void f2(std::initializer_list<std::string>, int = 0); // #cwg1589-f2-ilist-int
     void g2() { f2({"foo","bar"}); }
     // since-cxx11-error@-1 {{call to 'f2' is ambiguous}}
-    //   since-cxx11-note@#dr1589-f2-ilist {{candidate function}}
-    //   since-cxx11-note@#dr1589-f2-ilist-int {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f2-ilist {{candidate function}}
+    //   since-cxx11-note@#cwg1589-f2-ilist-int {{candidate function}}
   }
 #endif
-} // dr1589
+} // cwg1589
 
-namespace dr1591 {  //dr1591. Deducing array bound and element type from initializer list
+namespace cwg1591 {  //cwg1591. Deducing array bound and element type from initializer list
 #if __cplusplus >= 201103L
   template<class T, int N> int h(T const(&)[N]);
   int X = h({1,2,3});              // T deduced to int, N deduced to 3
@@ -630,10 +630,10 @@ namespace dr1591 {  //dr1591. Deducing array bound and element type from initial
   int Y = j({42});                 // T deduced to int, array bound not considered
 
   struct Aggr { int i; int j; };
-  template<int N> int k(Aggr const(&)[N]); // #dr1591-k
+  template<int N> int k(Aggr const(&)[N]); // #cwg1591-k
   int Y0 = k({1,2,3});
   // since-cxx11-error@-1 {{no matching function for call to 'k'}}
-  //   since-cxx11-note@#dr1591-k {{candidate function [with N = 3] not viable: no known conversion from 'int' to 'const Aggr' for 1st argument}}
+  //   since-cxx11-note@#cwg1591-k {{candidate function [with N = 3] not viable: no known conversion from 'int' to 'const Aggr' for 1st argument}}
   int Z = k({{1},{2},{3}});        // OK, N deduced to 3
 
   template<int M, int N> int m(int const(&)[M][N]);
@@ -644,31 +644,31 @@ namespace dr1591 {  //dr1591. Deducing array bound and element type from initial
   
   
   namespace check_multi_dim_arrays {
-    template<class T, int N, int M, int O> int ***f(const T (&a)[N][M][O]); // #dr1591-f-3
-    template<class T, int N, int M> int **f(const T (&a)[N][M]); // #dr1591-f-2
+    template<class T, int N, int M, int O> int ***f(const T (&a)[N][M][O]); // #cwg1591-f-3
+    template<class T, int N, int M> int **f(const T (&a)[N][M]); // #cwg1591-f-2
    
-   template<class T, int N> int *f(const T (&a)[N]); // #dr1591-f-1
+   template<class T, int N> int *f(const T (&a)[N]); // #cwg1591-f-1
     int ***p3 = f({  {  {1,2}, {3, 4}  }, {  {5,6}, {7, 8}  }, {  {9,10}, {11, 12}  } });
     int ***p33 = f({  {  {1,2}, {3, 4}  }, {  {5,6}, {7, 8}  }, {  {9,10}, {11, 12, 13}  } });
     // since-cxx11-error@-1 {{no matching function for call to 'f'}}
-    //   since-cxx11-note@#dr1591-f-2 {{candidate template ignored: couldn't infer template argument 'T'}}
-    //   since-cxx11-note@#dr1591-f-1 {{candidate template ignored: couldn't infer template argument 'T'}}
-    //   since-cxx11-note@#dr1591-f-3 {{candidate template ignored: deduced conflicting values for parameter 'O' (2 vs. 3)}}
+    //   since-cxx11-note@#cwg1591-f-2 {{candidate template ignored: couldn't infer template argument 'T'}}
+    //   since-cxx11-note@#cwg1591-f-1 {{candidate template ignored: couldn't infer template argument 'T'}}
+    //   since-cxx11-note@#cwg1591-f-3 {{candidate template ignored: deduced conflicting values for parameter 'O' (2 vs. 3)}}
     int **p2 = f({  {1,2,3}, {3, 4, 5}  });
     int **p22 = f({  {1,2}, {3, 4}  });
     int *p1 = f({1, 2, 3});
   }
   namespace check_multi_dim_arrays_rref {
-    template<class T, int N, int M, int O> int ***g(T (&&a)[N][M][O]); // #dr1591-g-3
-    template<class T, int N, int M> int **g(T (&&a)[N][M]); // #dr1591-g-2
+    template<class T, int N, int M, int O> int ***g(T (&&a)[N][M][O]); // #cwg1591-g-3
+    template<class T, int N, int M> int **g(T (&&a)[N][M]); // #cwg1591-g-2
    
-    template<class T, int N> int *g(T (&&a)[N]); // #dr1591-g-1
+    template<class T, int N> int *g(T (&&a)[N]); // #cwg1591-g-1
     int ***p3 = g({  {  {1,2}, {3, 4}  }, {  {5,6}, {7, 8}  }, {  {9,10}, {11, 12}  } });
     int ***p33 = g({  {  {1,2}, {3, 4}  }, {  {5,6}, {7, 8}  }, {  {9,10}, {11, 12, 13}  } });
     // since-cxx11-error@-1 {{no matching function for call to 'g'}}
-    //   since-cxx11-note@#dr1591-g-2 {{candidate template ignored: couldn't infer template argument 'T'}}
-    //   since-cxx11-note@#dr1591-g-1 {{candidate template ignored: couldn't infer template argument 'T'}}
-    //   since-cxx11-note@#dr1591-g-3 {{candidate template ignored: deduced conflicting values for parameter 'O' (2 vs. 3)}}
+    //   since-cxx11-note@#cwg1591-g-2 {{candidate template ignored: couldn't infer template argument 'T'}}
+    //   since-cxx11-note@#cwg1591-g-1 {{candidate template ignored: couldn't infer template argument 'T'}}
+    //   since-cxx11-note@#cwg1591-g-3 {{candidate template ignored: deduced conflicting values for parameter 'O' (2 vs. 3)}}
     int **p2 = g({  {1,2,3}, {3, 4, 5}  });
     int **p22 = g({  {1,2}, {3, 4}  });
     int *p1 = g({1, 2, 3});
@@ -684,8 +684,8 @@ namespace dr1591 {  //dr1591. Deducing array bound and element type from initial
     
     template<class T, int N> int *i(T (&&)[N]);  // #1
     template<class T> char *i(std::initializer_list<T> &&);  // #2
-    template<class T, int N, int M> int **i(T (&&)[N][M]); // #3 #dr1591-i-2
-    template<class T, int N> char **i(std::initializer_list<T> (&&)[N]); // #4 #dr1591-i-1
+    template<class T, int N, int M> int **i(T (&&)[N][M]); // #3 #cwg1591-i-2
+    template<class T, int N> char **i(std::initializer_list<T> (&&)[N]); // #4 #cwg1591-i-1
 
     template<class T> short *i(T (&&)[2]);  // #5
 
@@ -697,11 +697,11 @@ namespace dr1591 {  //dr1591. Deducing array bound and element type from initial
 
     void *pv1 = i({ {1, 2, 3}, {4, 5, 6} }); // ambiguous btw 3 & 4
     // since-cxx11-error@-1 {{call to 'i' is ambiguous}} 
-    //   since-cxx11-note@#dr1591-i-2 {{candidate function [with T = int, N = 2, M = 3]}}
-    //   since-cxx11-note@#dr1591-i-1 {{candidate function [with T = int, N = 2]}}
+    //   since-cxx11-note@#cwg1591-i-2 {{candidate function [with T = int, N = 2, M = 3]}}
+    //   since-cxx11-note@#cwg1591-i-1 {{candidate function [with T = int, N = 2]}}
     char **pcc = i({ {1}, {2, 3} }); // OK #4
 
     short *ps = i(Arr<int>{1, 2});  // OK #5
   }
 #endif
-} // dr1591
+} // cwg1591
diff --git a/clang/test/CXX/drs/dr16xx.cpp b/clang/test/CXX/drs/dr16xx.cpp
index f4d6c04fb8e07..6d7bb7619f8b8 100644
--- a/clang/test/CXX/drs/dr16xx.cpp
+++ b/clang/test/CXX/drs/dr16xx.cpp
@@ -25,7 +25,7 @@ namespace std {
 } // std
 #endif
 
-namespace dr1601 { // dr1601: 10
+namespace cwg1601 { // cwg1601: 10
 enum E : char { e };
 // cxx98-error@-1 {{enumeration types with a fixed underlying type are a C++11 extension}}
 void f(char);
@@ -33,9 +33,9 @@ void f(int);
 void g() {
   f(e);
 }
-} // namespace dr1601
+} // namespace cwg1601
 
-namespace dr1606 { // dr1606: 3.1
+namespace cwg1606 { // cwg1606: 3.1
 #if __cplusplus >= 201103L
   std::size_t test() {
     int i = 1;
@@ -44,16 +44,16 @@ namespace dr1606 { // dr1606: 3.1
     return sizeof(f);
   }
 #endif
-} // namespace dr1606
+} // namespace cwg1606
 
-namespace dr1611 { // dr1611: dup 1658
+namespace cwg1611 { // cwg1611: dup 1658
   struct A { A(int); };
   struct B : virtual A { virtual void f() = 0; };
   struct C : B { C() : A(0) {} void f(); };
   C c;
 }
 
-namespace dr1631 {  // dr1631: 3.7
+namespace cwg1631 {  // cwg1631: 3.7
 #if __cplusplus >= 201103L
   // Incorrect overload resolution for single-element initializer-list
 
@@ -70,24 +70,24 @@ namespace dr1631 {  // dr1631: 3.7
 
   namespace with_error {
     void f(B, int);           // TODO: expected- note {{candidate function}}
-    void f(int, A);           // #dr1631-f
-    void f(int, A, int = 0);  // #dr1631-f-int
+    void f(int, A);           // #cwg1631-f
+    void f(int, A, int = 0);  // #cwg1631-f-int
 
     void test() {
       f({0}, {{1}});
       // since-cxx11-error@-1 {{call to 'f' is ambiguous}}
-      //   since-cxx11-note@#dr1631-f {{candidate function}}
-      //   since-cxx11-note@#dr1631-f-int {{candidate function}}
+      //   since-cxx11-note@#cwg1631-f {{candidate function}}
+      //   since-cxx11-note@#cwg1631-f-int {{candidate function}}
     }
   }
 #endif
 }
 
-namespace dr1638 { // dr1638: 3.1
+namespace cwg1638 { // cwg1638: 3.1
 #if __cplusplus >= 201103L
   template<typename T> struct A {
-    enum class E; // #dr1638-E
-    enum class F : T; // #dr1638-F
+    enum class E; // #cwg1638-E
+    enum class F : T; // #cwg1638-F
   };
 
   template<> enum class A<int>::E;
@@ -100,13 +100,13 @@ namespace dr1638 { // dr1638: 3.1
 
   template<> enum class A<short>::F;
   // since-cxx11-error@-1 {{enumeration redeclared with different underlying type 'int' (was 'short')}}
-  //   since-cxx11-note@#dr1638-F {{previous declaration is here}}
+  //   since-cxx11-note@#cwg1638-F {{previous declaration is here}}
   template<> enum class A<char>::E : char;
   // since-cxx11-error@-1 {{enumeration redeclared with different underlying type 'char' (was 'int')}}
-  //   since-cxx11-note@#dr1638-E {{previous declaration is here}}
+  //   since-cxx11-note@#cwg1638-E {{previous declaration is here}}
   template<> enum class A<char>::F : int;
   // since-cxx11-error@-1 {{enumeration redeclared with different underlying type 'int' (was 'char')}}
-  //   since-cxx11-note@#dr1638-F {{previous declaration is here}}
+  //   since-cxx11-note@#cwg1638-F {{previous declaration is here}}
 
   enum class A<unsigned>::E;
   // since-cxx11-error@-1 {{template specialization requires 'template<>'}}
@@ -124,34 +124,34 @@ namespace dr1638 { // dr1638: 3.1
 #endif
 }
 
-namespace dr1645 { // dr1645: 3.9
+namespace cwg1645 { // cwg1645: 3.9
 #if __cplusplus >= 201103L
   struct A {
-    constexpr A(int, float = 0); // #dr1645-int-float
-    explicit A(int, int = 0); // #dr1645-int-int
-    A(int, int, int = 0) = delete; // #dr1645-int-int-int
+    constexpr A(int, float = 0); // #cwg1645-int-float
+    explicit A(int, int = 0); // #cwg1645-int-int
+    A(int, int, int = 0) = delete; // #cwg1645-int-int-int
   };
 
   struct B : A {
-    using A::A; // #dr1645-using
+    using A::A; // #cwg1645-using
   };
 
   constexpr B a(0);
   // since-cxx11-error@-1 {{call to constructor of 'const B' is ambiguous}}
-  //   since-cxx11-note@#dr1645-int-float {{candidate inherited constructor}}
-  //   since-cxx11-note@#dr1645-using {{constructor from base class 'A' inherited here}}
-  //   since-cxx11-note@#dr1645-int-int {{candidate inherited constructor}}
-  //   since-cxx11-note@#dr1645-using {{constructor from base class 'A' inherited here}}
+  //   since-cxx11-note@#cwg1645-int-float {{candidate inherited constructor}}
+  //   since-cxx11-note@#cwg1645-using {{constructor from base class 'A' inherited here}}
+  //   since-cxx11-note@#cwg1645-int-int {{candidate inherited constructor}}
+  //   since-cxx11-note@#cwg1645-using {{constructor from base class 'A' inherited here}}
   constexpr B b(0, 0);
   // since-cxx11-error@-1 {{call to constructor of 'const B' is ambiguous}}
-  //   since-cxx11-note@#dr1645-int-int {{candidate inherited constructor}}
-  //   since-cxx11-note@#dr1645-using {{constructor from base class 'A' inherited here}}
-  //   since-cxx11-note@#dr1645-int-int-int {{candidate inherited constructor has been explicitly deleted}}
-  //   since-cxx11-note@#dr1645-using {{constructor from base class 'A' inherited here}}
+  //   since-cxx11-note@#cwg1645-int-int {{candidate inherited constructor}}
+  //   since-cxx11-note@#cwg1645-using {{constructor from base class 'A' inherited here}}
+  //   since-cxx11-note@#cwg1645-int-int-int {{candidate inherited constructor has been explicitly deleted}}
+  //   since-cxx11-note@#cwg1645-using {{constructor from base class 'A' inherited here}}
 #endif
 }
 
-namespace dr1652 { // dr1652: 3.6
+namespace cwg1652 { // cwg1652: 3.6
   int a, b;
   int arr[&a + 1 == &b ? 1 : 2];
   // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
@@ -159,7 +159,7 @@ namespace dr1652 { // dr1652: 3.6
   // expected-error@-3 {{variable length array declaration not allowed at file scope}}
 }
 
-namespace dr1653 { // dr1653: 4 c++17
+namespace cwg1653 { // cwg1653: 4 c++17
   void f(bool b) {
     ++b;
     // cxx98-14-warning@-1 {{incrementing expression of type bool is deprecated and incompatible with C++17}}
@@ -176,10 +176,10 @@ namespace dr1653 { // dr1653: 4 c++17
   }
 }
 
-namespace dr1658 { // dr1658: 5
+namespace cwg1658 { // cwg1658: 5
   namespace DefCtor {
-    class A { A(); }; // #dr1658-A1
-    class B { ~B(); }; // #dr1658-B1
+    class A { A(); }; // #cwg1658-A1
+    class B { ~B(); }; // #cwg1658-B1
 
     // The stars align! An abstract class does not construct its virtual bases.
     struct C : virtual A { C(); virtual void foo() = 0; };
@@ -190,76 +190,76 @@ namespace dr1658 { // dr1658: 5
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
 
     // In all other cases, we are not so lucky.
-    struct E : A { E(); virtual void foo() = 0; }; // #dr1658-E1
-    E::E() = default; // #dr1658-E1-ctor
+    struct E : A { E(); virtual void foo() = 0; }; // #cwg1658-E1
+    E::E() = default; // #cwg1658-E1-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-2 {{base class 'A' has private default constructor}}
-    //   cxx98-note@-3 {{in defaulted default constructor for 'dr1658::DefCtor::E' first required here}}
-    //   cxx98-note@#dr1658-A1 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-E1-ctor {{defaulting this default constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-E1 {{default constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible default constructor}}
-    struct F : virtual A { F(); }; // #dr1658-F1
-    F::F() = default; // #dr1658-F1-ctor
+    //   cxx98-note@-3 {{in defaulted default constructor for 'cwg1658::DefCtor::E' first required here}}
+    //   cxx98-note@#cwg1658-A1 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-E1-ctor {{defaulting this default constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-E1 {{default constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible default constructor}}
+    struct F : virtual A { F(); }; // #cwg1658-F1
+    F::F() = default; // #cwg1658-F1-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-2 {{inherited virtual base class 'A' has private default constructor}}
-    //   cxx98-note@-3 {{in defaulted default constructor for 'dr1658::DefCtor::F' first required here}}
-    //   cxx98-note@#dr1658-A1 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-F1-ctor {{defaulting this default constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-F1 {{default constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible default constructor}}
+    //   cxx98-note@-3 {{in defaulted default constructor for 'cwg1658::DefCtor::F' first required here}}
+    //   cxx98-note@#cwg1658-A1 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-F1-ctor {{defaulting this default constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-F1 {{default constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible default constructor}}
 
-    struct G : B { G(); virtual void foo() = 0; }; // #dr1658-G1
-    G::G() = default; // #dr1658-G1-ctor
+    struct G : B { G(); virtual void foo() = 0; }; // #cwg1658-G1
+    G::G() = default; // #cwg1658-G1-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
-    // cxx98-error@#dr1658-G1 {{base class 'B' has private destructor}}
-    //   cxx98-note@#dr1658-G1-ctor {{in defaulted default constructor for 'dr1658::DefCtor::G' first required here}}
-    //   cxx98-note@#dr1658-B1 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-G1-ctor {{defaulting this default constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-G1 {{default constructor of 'G' is implicitly deleted because base class 'B' has an inaccessible destructor}}
-    struct H : virtual B { H(); }; // #dr1658-H1
-    H::H() = default; // #dr1658-H1-ctor
+    // cxx98-error@#cwg1658-G1 {{base class 'B' has private destructor}}
+    //   cxx98-note@#cwg1658-G1-ctor {{in defaulted default constructor for 'cwg1658::DefCtor::G' first required here}}
+    //   cxx98-note@#cwg1658-B1 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-G1-ctor {{defaulting this default constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-G1 {{default constructor of 'G' is implicitly deleted because base class 'B' has an inaccessible destructor}}
+    struct H : virtual B { H(); }; // #cwg1658-H1
+    H::H() = default; // #cwg1658-H1-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
-    // cxx98-error@#dr1658-H1 {{base class 'B' has private destructor}}
-    //   cxx98-note@#dr1658-H1-ctor {{in defaulted default constructor for 'dr1658::DefCtor::H' first required here}}
-    //   cxx98-note@#dr1658-B1 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-H1-ctor {{defaulting this default constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-H1 {{default constructor of 'H' is implicitly deleted because base class 'B' has an inaccessible destructor}}
+    // cxx98-error@#cwg1658-H1 {{base class 'B' has private destructor}}
+    //   cxx98-note@#cwg1658-H1-ctor {{in defaulted default constructor for 'cwg1658::DefCtor::H' first required here}}
+    //   cxx98-note@#cwg1658-B1 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-H1-ctor {{defaulting this default constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-H1 {{default constructor of 'H' is implicitly deleted because base class 'B' has an inaccessible destructor}}
   }
 
   namespace Dtor {
-    class B { ~B(); }; // #dr1658-B2
+    class B { ~B(); }; // #cwg1658-B2
 
     struct D : virtual B { ~D(); virtual void foo() = 0; };
     D::~D() = default; // ok, not deleted
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
 
-    struct G : B { ~G(); virtual void foo() = 0; }; // #dr1658-G2
-    G::~G() = default; // #dr1658-G2-dtor
+    struct G : B { ~G(); virtual void foo() = 0; }; // #cwg1658-G2
+    G::~G() = default; // #cwg1658-G2-dtor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
-    // cxx98-error@#dr1658-G2 {{base class 'B' has private destructor}}
-    //   cxx98-note@#dr1658-G2-dtor {{in defaulted destructor for 'dr1658::Dtor::G' first required here}}
-    //   cxx98-note@#dr1658-B2 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-G2-dtor {{defaulting this destructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-G2 {{destructor of 'G' is implicitly deleted because base class 'B' has an inaccessible destructor}}
-    struct H : virtual B { ~H(); }; // #dr1658-H2
-    H::~H() = default; // #dr1658-H2-dtor
+    // cxx98-error@#cwg1658-G2 {{base class 'B' has private destructor}}
+    //   cxx98-note@#cwg1658-G2-dtor {{in defaulted destructor for 'cwg1658::Dtor::G' first required here}}
+    //   cxx98-note@#cwg1658-B2 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-G2-dtor {{defaulting this destructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-G2 {{destructor of 'G' is implicitly deleted because base class 'B' has an inaccessible destructor}}
+    struct H : virtual B { ~H(); }; // #cwg1658-H2
+    H::~H() = default; // #cwg1658-H2-dtor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
-    // cxx98-error@#dr1658-H2 {{base class 'B' has private destructor}}
-    //   cxx98-note@#dr1658-H2-dtor {{in defaulted destructor for 'dr1658::Dtor::H' first required here}}
-    //   cxx98-note@#dr1658-B2 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-H2-dtor {{defaulting this destructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-H2 {{destructor of 'H' is implicitly deleted because base class 'B' has an inaccessible destructor}}
+    // cxx98-error@#cwg1658-H2 {{base class 'B' has private destructor}}
+    //   cxx98-note@#cwg1658-H2-dtor {{in defaulted destructor for 'cwg1658::Dtor::H' first required here}}
+    //   cxx98-note@#cwg1658-B2 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-H2-dtor {{defaulting this destructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-H2 {{destructor of 'H' is implicitly deleted because base class 'B' has an inaccessible destructor}}
   }
 
   namespace MemInit {
-    struct A { A(int); }; // #dr1658-A3
+    struct A { A(int); }; // #cwg1658-A3
     struct B : virtual A {
       B() {}
       virtual void f() = 0;
     };
     struct C : virtual A {
       C() {}
-      // expected-error@-1 {{constructor for 'dr1658::MemInit::C' must explicitly initialize the base class 'A' which does not have a default constructor}}
-      //   expected-note@#dr1658-A3 {{'dr1658::MemInit::A' declared here}}
+      // expected-error@-1 {{constructor for 'cwg1658::MemInit::C' must explicitly initialize the base class 'A' which does not have a default constructor}}
+      //   expected-note@#cwg1658-A3 {{'cwg1658::MemInit::A' declared here}}
     };
   }
 
@@ -277,7 +277,7 @@ namespace dr1658 { // dr1658: 5
   }
 
   namespace CopyCtor {
-    class A { A(const A&); A(A&&); }; // #dr1658-A5
+    class A { A(const A&); A(A&&); }; // #cwg1658-A5
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
 
     struct C : virtual A { C(const C&); C(C&&); virtual void foo() = 0; };
@@ -288,46 +288,46 @@ namespace dr1658 { // dr1658: 5
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
     // cxx98-error@-2 {{defaulted function definitions are a C++11 extension}}
 
-    struct E : A { E(const E&); E(E&&); virtual void foo() = 0; }; // #dr1658-E5
+    struct E : A { E(const E&); E(E&&); virtual void foo() = 0; }; // #cwg1658-E5
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
-    E::E(const E&) = default; // #dr1658-E5-copy-ctor
+    E::E(const E&) = default; // #cwg1658-E5-copy-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-2 {{base class 'A' has private copy constructor}}
-    //   cxx98-note@-3 {{in defaulted copy constructor for 'dr1658::CopyCtor::E' first required here}}
-    //   cxx98-note@#dr1658-A5 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-E5-copy-ctor {{defaulting this copy constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-E5 {{copy constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible copy constructor}}
-    E::E(E&&) = default; // #dr1658-E5-move-ctor
+    //   cxx98-note@-3 {{in defaulted copy constructor for 'cwg1658::CopyCtor::E' first required here}}
+    //   cxx98-note@#cwg1658-A5 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-E5-copy-ctor {{defaulting this copy constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-E5 {{copy constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible copy constructor}}
+    E::E(E&&) = default; // #cwg1658-E5-move-ctor
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
     // cxx98-error@-2 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-3 {{base class 'A' has private move constructor}}
-    //   cxx98-note@-4 {{in defaulted move constructor for 'dr1658::CopyCtor::E' first required here}}
-    //   cxx98-note@#dr1658-A5 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-E5-move-ctor {{defaulting this move constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-E5 {{move constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible move constructor}}
-    struct F : virtual A { F(const F&); F(F&&); }; // #dr1658-F5
+    //   cxx98-note@-4 {{in defaulted move constructor for 'cwg1658::CopyCtor::E' first required here}}
+    //   cxx98-note@#cwg1658-A5 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-E5-move-ctor {{defaulting this move constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-E5 {{move constructor of 'E' is implicitly deleted because base class 'A' has an inaccessible move constructor}}
+    struct F : virtual A { F(const F&); F(F&&); }; // #cwg1658-F5
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
-    F::F(const F&) = default; // #dr1658-F5-copy-ctor
+    F::F(const F&) = default; // #cwg1658-F5-copy-ctor
     // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-2 {{inherited virtual base class 'A' has private copy constructor}}
-    //   cxx98-note@-3 {{in defaulted copy constructor for 'dr1658::CopyCtor::F' first required here}}
-    //   cxx98-note@#dr1658-A5 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-F5-copy-ctor {{defaulting this copy constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-F5 {{copy constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible copy constructor}}
-    F::F(F&&) = default; // #dr1658-F5-move-ctor
+    //   cxx98-note@-3 {{in defaulted copy constructor for 'cwg1658::CopyCtor::F' first required here}}
+    //   cxx98-note@#cwg1658-A5 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-F5-copy-ctor {{defaulting this copy constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-F5 {{copy constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible copy constructor}}
+    F::F(F&&) = default; // #cwg1658-F5-move-ctor
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
     // cxx98-error@-2 {{defaulted function definitions are a C++11 extension}}
     // cxx98-error@-3 {{inherited virtual base class 'A' has private move constructor}}
-    //   cxx98-note@-4 {{in defaulted move constructor for 'dr1658::CopyCtor::F' first required here}}
-    //   cxx98-note@#dr1658-A5 {{implicitly declared private here}}
-    // since-cxx11-error@#dr1658-F5-move-ctor {{defaulting this move constructor would delete it after its first declaration}}
-    //   since-cxx11-note@#dr1658-F5 {{move constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible move constructor}}
+    //   cxx98-note@-4 {{in defaulted move constructor for 'cwg1658::CopyCtor::F' first required here}}
+    //   cxx98-note@#cwg1658-A5 {{implicitly declared private here}}
+    // since-cxx11-error@#cwg1658-F5-move-ctor {{defaulting this move constructor would delete it after its first declaration}}
+    //   since-cxx11-note@#cwg1658-F5 {{move constructor of 'F' is implicitly deleted because base class 'A' has an inaccessible move constructor}}
   }
 
-  // assignment case is superseded by dr2180
+  // assignment case is superseded by cwg2180
 }
 
-namespace dr1672 { // dr1672: 7
+namespace cwg1672 { // cwg1672: 7
   struct Empty {};
   struct A : Empty {};
   struct B { Empty e; };
@@ -352,9 +352,9 @@ namespace dr1672 { // dr1672: 7
   static_assert(!__is_standard_layout(Y<X>), "");
 }
 
-namespace dr1684 { // dr1684: 3.6
+namespace cwg1684 { // cwg1684: 3.6
 #if __cplusplus >= 201103L
-  struct NonLiteral { // #dr1684-struct
+  struct NonLiteral { // #cwg1684-struct
     NonLiteral();
     constexpr int f() { return 0; }
     // cxx11-warning@-1 {{'constexpr' non-static member function will not be implicitly 'const' in C++14; add 'const' to avoid a change in behavior}}
@@ -362,34 +362,34 @@ namespace dr1684 { // dr1684: 3.6
   constexpr int f(NonLiteral &) { return 0; }
   constexpr int f(NonLiteral) { return 0; }
   // cxx11-20-error@-1 {{constexpr function's 1st parameter type 'NonLiteral' is not a literal type}}
-  //   cxx11-20-note@#dr1684-struct {{'NonLiteral' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
+  //   cxx11-20-note@#cwg1684-struct {{'NonLiteral' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
 #endif
 }
 
-namespace dr1687 { // dr1687: 7
+namespace cwg1687 { // cwg1687: 7
   template<typename T> struct To {
-    operator T(); // #dr1687-op-T
+    operator T(); // #cwg1687-op-T
   };
 
   int *a = To<int*>() + 100.0;
   // expected-error@-1 {{invalid operands to binary expression ('To<int *>' and 'double')}}
-  //   expected-note@#dr1687-op-T {{first operand was implicitly converted to type 'int *'}}
-  //   since-cxx20-note@#dr1687-op-T {{second operand was implicitly converted to type 'dr1687::E2'}}
+  //   expected-note@#cwg1687-op-T {{first operand was implicitly converted to type 'int *'}}
+  //   since-cxx20-note@#cwg1687-op-T {{second operand was implicitly converted to type 'cwg1687::E2'}}
   int *b = To<int*>() + To<double>();
   // expected-error@-1 {{invalid operands to binary expression ('To<int *>' and 'To<double>')}}
-  //   expected-note@#dr1687-op-T {{first operand was implicitly converted to type 'int *'}}
-  //   expected-note@#dr1687-op-T {{second operand was implicitly converted to type 'double'}}
+  //   expected-note@#cwg1687-op-T {{first operand was implicitly converted to type 'int *'}}
+  //   expected-note@#cwg1687-op-T {{second operand was implicitly converted to type 'double'}}
 
 #if __cplusplus >= 202002L
   enum E1 {};
   enum E2 {};
   auto c = To<E1>() <=> To<E2>();
   // since-cxx20-error@-1 {{invalid operands to binary expression ('To<E1>' and 'To<E2>')}}
-  //   since-cxx20-note@#dr1687-op-T {{operand was implicitly converted to type 'dr1687::E}}
+  //   since-cxx20-note@#cwg1687-op-T {{operand was implicitly converted to type 'cwg1687::E}}
 #endif
 }
 
-namespace dr1690 { // dr1690: 9
+namespace cwg1690 { // cwg1690: 9
   // See also the various tests in "CXX/basic/basic.lookup/basic.lookup.argdep".
 #if __cplusplus >= 201103L
   namespace N {
@@ -404,7 +404,7 @@ namespace dr1690 { // dr1690: 9
 #endif
 }
 
-namespace dr1691 { // dr1691: 9
+namespace cwg1691 { // cwg1691: 9
 #if __cplusplus >= 201103L
   namespace N {
     namespace M {
@@ -412,19 +412,19 @@ namespace dr1691 { // dr1691: 9
       void f(E);
     }
     enum M::E : int {};
-    void g(M::E); // #dr1691-g
+    void g(M::E); // #cwg1691-g
   }
   void test() {
     N::M::E e;
     f(e); // ok
     g(e);
     // since-cxx11-error@-1 {{use of undeclared identifier 'g'; did you mean 'N::g'?}}
-    //   since-cxx11-note@#dr1691-g {{'N::g' declared here}}
+    //   since-cxx11-note@#cwg1691-g {{'N::g' declared here}}
   }
 #endif
 }
 
-namespace dr1692 { // dr1692: 9
+namespace cwg1692 { // cwg1692: 9
   namespace N {
     struct A {
       struct B {
@@ -439,7 +439,7 @@ namespace dr1692 { // dr1692: 9
   }
 }
 
-namespace dr1696 { // dr1696: 7
+namespace cwg1696 { // cwg1696: 7
   namespace std_examples {
 #if __cplusplus >= 201402L
     extern struct A a;
@@ -456,66 +456,66 @@ namespace dr1696 { // dr1696: 7
   struct A { A(); ~A(); };
 #if __cplusplus >= 201103L
   struct B {
-    A &&a; // #dr1696-a
+    A &&a; // #cwg1696-a
     B() : a{} {}
     // since-cxx11-error@-1 {{reference member 'a' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-    //   since-cxx11-note@#dr1696-a {{reference member declared here}}
+    //   since-cxx11-note@#cwg1696-a {{reference member declared here}}
   } b;
 #endif
 
   struct C {
     C();
-    const A &a; // #dr1696-C-a
+    const A &a; // #cwg1696-C-a
   };
   C::C() : a(A()) {}
   // expected-error@-1 {{reference member 'a' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   expected-note@#dr1696-C-a {{reference member declared here}}
+  //   expected-note@#cwg1696-C-a {{reference member declared here}}
 
 #if __cplusplus >= 201103L
-  // This is OK in C++14 onwards, per DR1815, though we don't support that yet:
+  // This is OK in C++14 onwards, per CWG1815, though we don't support that yet:
   //   D1 d1 = {};
   // is equivalent to
   //   D1 d1 = {A()};
   // ... which lifetime-extends the A temporary.
   struct D1 {
   // cxx11-error@-1 {{reference member 'a' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   cxx11-note@#dr1696-d1 {{in implicit default constructor for 'dr1696::D1' first required here}}
-  //   cxx11-note@#dr1696-D1-a {{initializing field 'a' with default member initializer}}
-    const A &a = A(); // #dr1696-D1-a
+  //   cxx11-note@#cwg1696-d1 {{in implicit default constructor for 'cwg1696::D1' first required here}}
+  //   cxx11-note@#cwg1696-D1-a {{initializing field 'a' with default member initializer}}
+    const A &a = A(); // #cwg1696-D1-a
   };
-  D1 d1 = {}; // #dr1696-d1
+  D1 d1 = {}; // #cwg1696-d1
   // since-cxx14-warning@-1 {{lifetime extension of temporary created by aggregate initialization using a default member initializer is not yet supported; lifetime of temporary will end at the end of the full-expression}}
-  //   since-cxx14-note@#dr1696-D1-a {{initializing field 'a' with default member initializer}}
+  //   since-cxx14-note@#cwg1696-D1-a {{initializing field 'a' with default member initializer}}
 
   struct D2 {
-    const A &a = A(); // #dr1696-D2-a
+    const A &a = A(); // #cwg1696-D2-a
     D2() {}
     // since-cxx11-error@-1 {{reference member 'a' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-    //   since-cxx11-note@#dr1696-D2-a {{initializing field 'a' with default member initializer}}
+    //   since-cxx11-note@#cwg1696-D2-a {{initializing field 'a' with default member initializer}}
   };
 
   struct D3 {
   // since-cxx11-error@-1 {{reference member 'a' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   since-cxx11-note@#dr1696-d3 {{in implicit default constructor for 'dr1696::D3' first required here}}
-  //   since-cxx11-note@#dr1696-D3-a {{initializing field 'a' with default member initializer}}
-    const A &a = A(); // #dr1696-D3-a
+  //   since-cxx11-note@#cwg1696-d3 {{in implicit default constructor for 'cwg1696::D3' first required here}}
+  //   since-cxx11-note@#cwg1696-D3-a {{initializing field 'a' with default member initializer}}
+    const A &a = A(); // #cwg1696-D3-a
   };
-  D3 d3; // #dr1696-d3
+  D3 d3; // #cwg1696-d3
 
   struct haslist1 {
-    std::initializer_list<int> il; // #dr1696-il-1
+    std::initializer_list<int> il; // #cwg1696-il-1
     haslist1(int i) : il{i, 2, 3} {}
     // since-cxx11-error@-1 {{backing array for 'std::initializer_list' member 'il' is a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-    //   since-cxx11-note@#dr1696-il-1 {{'std::initializer_list' member declared here}}
+    //   since-cxx11-note@#cwg1696-il-1 {{'std::initializer_list' member declared here}}
   };
 
   struct haslist2 {
-    std::initializer_list<int> il; // #dr1696-il-2
+    std::initializer_list<int> il; // #cwg1696-il-2
     haslist2();
   };
   haslist2::haslist2() : il{1, 2} {}
   // since-cxx11-error@-1 {{backing array for 'std::initializer_list' member 'il' is a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   since-cxx11-note@#dr1696-il-2 {{'std::initializer_list' member declared here}}
+  //   since-cxx11-note@#cwg1696-il-2 {{'std::initializer_list' member declared here}}
 
   struct haslist3 {
     std::initializer_list<int> il = {1, 2, 3};
@@ -523,17 +523,17 @@ namespace dr1696 { // dr1696: 7
 
   struct haslist4 {
   // since-cxx11-error@-1 {{backing array for 'std::initializer_list' member 'il' is a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   since-cxx11-note@#dr1696-hl4 {{in implicit default constructor for 'dr1696::haslist4' first required here}}
-  //   since-cxx11-note@#dr1696-il-4 {{initializing field 'il' with default member initializer}}
-    std::initializer_list<int> il = {1, 2, 3}; // #dr1696-il-4
+  //   since-cxx11-note@#cwg1696-hl4 {{in implicit default constructor for 'cwg1696::haslist4' first required here}}
+  //   since-cxx11-note@#cwg1696-il-4 {{initializing field 'il' with default member initializer}}
+    std::initializer_list<int> il = {1, 2, 3}; // #cwg1696-il-4
   };
-  haslist4 hl4; // #dr1696-hl4
+  haslist4 hl4; // #cwg1696-hl4
 
   struct haslist5 {
-    std::initializer_list<int> il = {1, 2, 3}; // #dr1696-il-5
+    std::initializer_list<int> il = {1, 2, 3}; // #cwg1696-il-5
     haslist5() {}
     // since-cxx11-error@-1 {{backing array for 'std::initializer_list' member 'il' is a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-    //   since-cxx11-note@#dr1696-il-5 {{nitializing field 'il' with default member initializer}}
+    //   since-cxx11-note@#cwg1696-il-5 {{nitializing field 'il' with default member initializer}}
   };
 #endif
 }
diff --git a/clang/test/CXX/drs/dr17xx.cpp b/clang/test/CXX/drs/dr17xx.cpp
index d3cb5e58f06b3..fb53a56923b10 100644
--- a/clang/test/CXX/drs/dr17xx.cpp
+++ b/clang/test/CXX/drs/dr17xx.cpp
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++2c %s -verify=expected,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr1710 { // dr1710: no
+namespace cwg1710 { // cwg1710: no
 // FIXME: all of the following is well-formed
 template <typename T> struct D1 : T::template B<int>::template C<int> {};
 template <typename T> struct D2 : T::B<int>::template C<int> {};
@@ -16,9 +16,9 @@ template <typename T> struct D3 : T::template B<int>::C<int> {};
 template <typename T> struct D4 : T::B<int>::C<int> {};
 // expected-error@-1 {{use 'template' keyword to treat 'B' as a dependent template name}}
 // expected-error@-2 {{use 'template' keyword to treat 'C' as a dependent template name}}
-} // namespace dr1710
+} // namespace cwg1710
 
-namespace dr1715 { // dr1715: 3.9
+namespace cwg1715 { // cwg1715: 3.9
 #if __cplusplus >= 201103L
   struct B {
     template<class T> B(T, typename T::Q);
@@ -32,21 +32,21 @@ namespace dr1715 { // dr1715: 3.9
   struct D : B {
     using B::B;
   };
-  struct E : B { // #dr1715-E
-    template<class T> E(T t, typename T::Q q) : B(t, q) {} // #dr1715-E-ctor
+  struct E : B { // #cwg1715-E
+    template<class T> E(T t, typename T::Q q) : B(t, q) {} // #cwg1715-E-ctor
   };
 
   B b(S(), 1);
   D d(S(), 2);
   E e(S(), 3);
   // since-cxx11-error@-1 {{no matching constructor for initialization of 'E'}}
-  //   since-cxx11-note@#dr1715-E-ctor {{candidate template ignored: substitution failure [with T = S]: 'Q' is a private member of 'dr1715::S'}}
-  //   since-cxx11-note@#dr1715-E {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
-  //   since-cxx11-note@#dr1715-E {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
+  //   since-cxx11-note@#cwg1715-E-ctor {{candidate template ignored: substitution failure [with T = S]: 'Q' is a private member of 'cwg1715::S'}}
+  //   since-cxx11-note@#cwg1715-E {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 2 were provided}}
+  //   since-cxx11-note@#cwg1715-E {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 2 were provided}}
 #endif
 }
 
-namespace dr1719 { // dr1719: 19
+namespace cwg1719 { // cwg1719: 19
 #if __cplusplus >= 201103L
 struct CStruct {
   int one;
@@ -72,9 +72,9 @@ static_assert(__is_layout_compatible(const int, volatile int), "");
 static_assert(__is_layout_compatible(CStruct, CStructWithQualifiers), "");
 static_assert(__is_layout_compatible(int[], const volatile int[]), "");
 #endif
-} // namespace dr1719
+} // namespace cwg1719
 
-namespace dr1722 { // dr1722: 9
+namespace cwg1722 { // cwg1722: 9
 #if __cplusplus >= 201103L
 void f() {
   const auto lambda = [](int x) { return x + 1; };
@@ -84,9 +84,9 @@ void f() {
       "Lambda-to-function-pointer conversion is expected to be noexcept");
 }
 #endif
-} // namespace dr1722
+} // namespace cwg1722
 
-namespace dr1734 { // dr1734: no
+namespace cwg1734 { // cwg1734: no
 #if __cplusplus >= 201103L
 struct A {
     A(const A&) = delete;
@@ -98,7 +98,7 @@ static_assert(__is_trivially_copyable(A), "");
 #endif
 }
 
-namespace dr1736 { // dr1736: 3.9
+namespace cwg1736 { // cwg1736: 3.9
 #if __cplusplus >= 201103L
 struct S {
   template <class T> S(T t) {
@@ -107,17 +107,17 @@ struct S {
     };
     typename T::type value;
     // since-cxx11-error@-1 {{type 'int' cannot be used prior to '::' because it has no members}}
-    //   since-cxx11-note@#dr1736-l {{in instantiation of function template specialization 'dr1736::S::S<int>' requested here}}
-    //   since-cxx11-note@#dr1736-s {{in instantiation of function template specialization 'dr1736::S::S<dr1736::Q>' requested here}}
-    L l(value); // #dr1736-l
+    //   since-cxx11-note@#cwg1736-l {{in instantiation of function template specialization 'cwg1736::S::S<int>' requested here}}
+    //   since-cxx11-note@#cwg1736-s {{in instantiation of function template specialization 'cwg1736::S::S<cwg1736::Q>' requested here}}
+    L l(value); // #cwg1736-l
   }
 };
 struct Q { typedef int type; } q;
-S s(q); // #dr1736-s
+S s(q); // #cwg1736-s
 #endif
 }
 
-namespace dr1738 { // dr1738: sup P0136R1
+namespace cwg1738 { // cwg1738: sup P0136R1
 #if __cplusplus >= 201103L
 struct A {
   template <typename T>
@@ -134,9 +134,9 @@ template B::B(int, double);
 #endif
 }
 
-// dr1748 is in dr1748.cpp
+// cwg1748 is in cwg1748.cpp
 
-namespace dr1753 { // dr1753: 11
+namespace cwg1753 { // cwg1753: 11
   typedef int T;
   struct A { typedef int T; };
   namespace B { typedef int T; }
@@ -145,9 +145,9 @@ namespace dr1753 { // dr1753: 11
     n.~T();
     n.T::~T();
 
-    n.dr1753::~T();
-    // expected-error@-1 {{'dr1753' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T' (aka 'int')}}
-    n.dr1753::T::~T();
+    n.cwg1753::~T();
+    // expected-error@-1 {{'cwg1753' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T' (aka 'int')}}
+    n.cwg1753::T::~T();
 
     n.A::~T();
     // expected-error@-1 {{the type of object expression ('T' (aka 'int')) does not match the type being destroyed ('A') in pseudo-destructor expression}}
@@ -167,7 +167,7 @@ namespace dr1753 { // dr1753: 11
   }
 }
 
-namespace dr1756 { // dr1756: 3.7
+namespace cwg1756 { // cwg1756: 3.7
 #if __cplusplus >= 201103L
   // Direct-list-initialization of a non-class object
   
@@ -178,7 +178,7 @@ namespace dr1756 { // dr1756: 3.7
 #endif
 }
 
-namespace dr1758 { // dr1758: 3.7
+namespace cwg1758 { // cwg1758: 3.7
 #if __cplusplus >= 201103L
   // Explicit conversion in copy/move list initialization
 
@@ -197,7 +197,7 @@ namespace dr1758 { // dr1758: 3.7
 #endif
 }
 
-namespace dr1762 { // dr1762: 14
+namespace cwg1762 { // cwg1762: 14
 #if __cplusplus >= 201103L
   float operator ""_E(const char *);
   float operator ""E(const char *);
@@ -206,9 +206,9 @@ namespace dr1762 { // dr1762: 14
 #endif
 }
 
-// dr1772 is in dr177x.cpp
+// cwg1772 is in cwg177x.cpp
 
-namespace dr1778 { // dr1778: 9
+namespace cwg1778 { // cwg1778: 9
   // Superseded by P1286R2.
 #if __cplusplus >= 201103L
   struct A { A() noexcept(true) = default; };
@@ -223,9 +223,9 @@ namespace dr1778 { // dr1778: 9
 #endif
 }
 
-// dr1779 is in dr177x.cpp
+// cwg1779 is in cwg177x.cpp
 
-namespace dr1794 { // dr1794: yes
+namespace cwg1794 { // cwg1794: yes
                    // NB: dup 1710
 #if __cplusplus >= 201103L
 template <template <typename> class Template> struct Internal {
@@ -238,4 +238,4 @@ using Instantiate = Template<Arg>;
 template <template <typename> class Template, typename Argument>
 using Bind = Instantiate<Internal<Template>::template Bind, Argument>;
 #endif
-} // namespace dr1794
+} // namespace cwg1794
diff --git a/clang/test/CXX/drs/dr18xx.cpp b/clang/test/CXX/drs/dr18xx.cpp
index e78730e8992cf..3a2248a1af551 100644
--- a/clang/test/CXX/drs/dr18xx.cpp
+++ b/clang/test/CXX/drs/dr18xx.cpp
@@ -11,24 +11,24 @@
 // cxx98-error@-1 {{variadic macros are a C99 feature}}
 #endif
 
-namespace dr1800 { // dr1800: 2.9
+namespace cwg1800 { // cwg1800: 2.9
 struct A { union { int n; }; };
 static_assert(__is_same(__decltype(&A::n), int A::*), "");
-} // namespace dr1800
+} // namespace cwg1800
 
-namespace dr1801 { // dr1801: 2.8
+namespace cwg1801 { // cwg1801: 2.8
 static union {
   int i;
 };
 
-template <int &> struct S {}; // #dr1801-S
-S<i> V; // #dr1801-S-i
+template <int &> struct S {}; // #cwg1801-S
+S<i> V; // #cwg1801-S-i
 // cxx98-14-error@-1 {{non-type template argument does not refer to any declaration}}
-//   cxx98-14-note@#dr1801-S {{template parameter is declared here}}
-// cxx17-error@#dr1801-S-i {{non-type template argument refers to subobject '.i'}}
+//   cxx98-14-note@#cwg1801-S {{template parameter is declared here}}
+// cxx17-error@#cwg1801-S-i {{non-type template argument refers to subobject '.i'}}
 }
 
-namespace dr1802 { // dr1802: 3.1
+namespace cwg1802 { // cwg1802: 3.1
 #if __cplusplus >= 201103L
 // Using a Wikipedia example of surrogate pair:
 // https://en.wikipedia.org/wiki/UTF-16#Examples
@@ -37,9 +37,9 @@ static_assert(a[0] == 0xD801, "");
 static_assert(a[1] == 0xDC37, "");
 static_assert(a[2] == 0x0, "");
 #endif
-} // namespace dr1802
+} // namespace cwg1802
 
-namespace dr1803 { // dr1803: 2.9
+namespace cwg1803 { // cwg1803: 2.9
 #if __cplusplus >= 201103L
 struct A {
   enum E : int;
@@ -50,9 +50,9 @@ struct A {
   enum struct ES {};
 };
 #endif
-} // namespace dr1803
+} // namespace cwg1803
 
-namespace dr1804 { // dr1804: 2.7
+namespace cwg1804 { // cwg1804: 2.7
 template <typename, typename>
 struct A {
   void f1();
@@ -162,11 +162,11 @@ void A<double, U>::C<V>::f4() {
   D d;
   d.i = 0;
 }
-} // namespace dr1804
+} // namespace cwg1804
 
-// dr1807 is in dr1807.cpp
+// cwg1807 is in cwg1807.cpp
 
-namespace dr1812 { // dr1812: no
+namespace cwg1812 { // cwg1812: no
                    // NB: dup 1710
 #if __cplusplus >= 201103L
 template <typename T> struct A {
@@ -174,9 +174,9 @@ template <typename T> struct A {
   // since-cxx11-error@-1 {{use 'template' keyword to treat 'C' as a dependent template name}}
 };
 #endif
-} // namespace dr1812
+} // namespace cwg1812
 
-namespace dr1813 { // dr1813: 7
+namespace cwg1813 { // cwg1813: 7
   struct B { int i; };
   struct C : B {};
   struct D : C {};
@@ -198,7 +198,7 @@ namespace dr1813 { // dr1813: 7
   static_assert(!__is_standard_layout(U), "");
 }
 
-namespace dr1814 { // dr1814: yes
+namespace cwg1814 { // cwg1814: yes
 #if __cplusplus >= 201103L
   void test() {
     auto lam = [](int x = 42) { return x; };
@@ -206,23 +206,23 @@ namespace dr1814 { // dr1814: yes
 #endif
 }
 
-namespace dr1815 { // dr1815: no
+namespace cwg1815 { // cwg1815: no
 #if __cplusplus >= 201402L
   // FIXME: needs codegen test
-  struct A { int &&r = 0; }; // #dr1815-A 
+  struct A { int &&r = 0; }; // #cwg1815-A 
   A a = {};
   // since-cxx14-warning@-1 {{lifetime extension of temporary created by aggregate initialization using a default member initializer is not yet supported; lifetime of temporary will end at the end of the full-expression}} FIXME
-  //   since-cxx14-note@#dr1815-A {{initializing field 'r' with default member initializer}}
+  //   since-cxx14-note@#cwg1815-A {{initializing field 'r' with default member initializer}}
 
-  struct B { int &&r = 0; }; // #dr1815-B
+  struct B { int &&r = 0; }; // #cwg1815-B
   // since-cxx14-error@-1 {{reference member 'r' binds to a temporary object whose lifetime would be shorter than the lifetime of the constructed object}}
-  //   since-cxx14-note@#dr1815-B {{initializing field 'r' with default member initializer}}
-  //   since-cxx14-note@#dr1815-b {{in implicit default constructor for 'dr1815::B' first required here}}
-  B b; // #dr1815-b
+  //   since-cxx14-note@#cwg1815-B {{initializing field 'r' with default member initializer}}
+  //   since-cxx14-note@#cwg1815-b {{in implicit default constructor for 'cwg1815::B' first required here}}
+  B b; // #cwg1815-b
 #endif
 }
 
-namespace dr1821 { // dr1821: 2.9
+namespace cwg1821 { // cwg1821: 2.9
 struct A {
   template <typename> struct B {
     void f();
@@ -236,9 +236,9 @@ struct A {
   void C::f() {}
   // expected-error@-1 {{non-friend class member 'f' cannot have a qualified name}}
 };
-} // namespace dr1821
+} // namespace cwg1821
 
-namespace dr1822 { // dr1822: yes
+namespace cwg1822 { // cwg1822: yes
 #if __cplusplus >= 201103L
   double a;
   auto x = [] (int a) {
@@ -247,7 +247,7 @@ namespace dr1822 { // dr1822: yes
 #endif
 }
 
-namespace dr1837 { // dr1837: 3.3
+namespace cwg1837 { // cwg1837: 3.3
 #if __cplusplus >= 201103L
   template <typename T>
   struct Fish { static const bool value = true; };
@@ -291,7 +291,7 @@ namespace dr1837 { // dr1837: 3.3
 #endif
 }
 
-namespace dr1872 { // dr1872: 9
+namespace cwg1872 { // cwg1872: 9
 #if __cplusplus >= 201103L
   template<typename T> struct A : T {
     constexpr int f() const { return 0; }
@@ -320,7 +320,7 @@ namespace dr1872 { // dr1872: 9
 #endif
 }
 
-namespace dr1878 { // dr1878: 18
+namespace cwg1878 { // cwg1878: 18
 #if __cplusplus >= 201402L
 #if __cplusplus >= 202002L
 template <typename T>
@@ -355,7 +355,7 @@ struct S {
 #endif
 }
 
-namespace dr1881 { // dr1881: 7
+namespace cwg1881 { // cwg1881: 7
   struct A { int a : 4; };
   struct B : A { int b : 3; };
   static_assert(__is_standard_layout(A), "");
@@ -367,17 +367,17 @@ namespace dr1881 { // dr1881: 7
   static_assert(!__is_standard_layout(D), "");
 }
 
-namespace dr1890 { // dr1890: no drafting 2018-06-04
+namespace cwg1890 { // cwg1890: no drafting 2018-06-04
 // FIXME: current consensus for CWG2335 is that the examples are well-formed.
 namespace ex1 {
 #if __cplusplus >= 201402L
 struct A {
   struct B {
-    auto foo() { return 0; } // #dr1890-foo
+    auto foo() { return 0; } // #cwg1890-foo
   };
   decltype(B().foo()) x;
   // since-cxx14-error@-1 {{function 'foo' with deduced return type cannot be used before it is defined}}
-  //   since-cxx14-note@#dr1890-foo {{'foo' declared here}}
+  //   since-cxx14-note@#cwg1890-foo {{'foo' declared here}}
 };
 #endif
 } // namespace ex1
@@ -389,17 +389,17 @@ struct Bar {
     int a = 0;
   };
   static_assert(__is_constructible(Baz), "");
-  // since-cxx11-error@-1 {{static assertion failed due to requirement '__is_constructible(dr1890::ex2::Bar::Baz)'}}
+  // since-cxx11-error@-1 {{static assertion failed due to requirement '__is_constructible(cwg1890::ex2::Bar::Baz)'}}
 };
 #endif
 } // namespace ex2
-} // namespace dr1890
+} // namespace cwg1890
 
-void dr1891() { // dr1891: 4
+void cwg1891() { // cwg1891: 4
 #if __cplusplus >= 201103L
   int n;
-  auto a = []{}; // #dr1891-a
-  auto b = [=]{ return n; }; // #dr1891-b
+  auto a = []{}; // #cwg1891-a
+  auto b = [=]{ return n; }; // #cwg1891-b
   typedef decltype(a) A;
   typedef decltype(b) B;
 
@@ -410,31 +410,31 @@ void dr1891() { // dr1891: 4
   // C++20 allows default construction for non-capturing lambdas (P0624R2).
   A x;
   // cxx11-17-error@-1 {{no matching constructor for initialization of 'A' (aka '(lambda at}}
-  //   cxx11-17-note@#dr1891-a {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-  //   cxx11-17-note@#dr1891-a {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1891-a {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   cxx11-17-note@#cwg1891-a {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
   B y;
   // since-cxx11-error@-1 {{no matching constructor for initialization of 'B' (aka '(lambda at}}
-  //   since-cxx11-note@#dr1891-b {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-  //   since-cxx11-note@#dr1891-b {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   since-cxx11-note@#cwg1891-b {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+  //   since-cxx11-note@#cwg1891-b {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
 
   // C++20 allows assignment for non-capturing lambdas (P0624R2).
   a = a;
   // cxx11-17-error-re@-1 {{{{object of type '\(lambda at .+\)' cannot be assigned because its copy assignment operator is implicitly deleted}}}}
-  //   cxx11-17-note@#dr1891-a {{lambda expression begins here}}
+  //   cxx11-17-note@#cwg1891-a {{lambda expression begins here}}
   a = static_cast<A&&>(a);
   // cxx11-17-error-re@-1 {{{{object of type '\(lambda at .+\)' cannot be assigned because its copy assignment operator is implicitly deleted}}}}
-  //   cxx11-17-note@#dr1891-a {{lambda expression begins here}}
+  //   cxx11-17-note@#cwg1891-a {{lambda expression begins here}}
   b = b;
   // since-cxx11-error-re@-1 {{{{object of type '\(lambda at .+\)' cannot be assigned because its copy assignment operator is implicitly deleted}}}}
-  //   since-cxx11-note@#dr1891-b {{lambda expression begins here}}
+  //   since-cxx11-note@#cwg1891-b {{lambda expression begins here}}
   b = static_cast<B&&>(b);
   // since-cxx11-error-re@-1 {{{{object of type '\(lambda at .+\)' cannot be assigned because its copy assignment operator is implicitly deleted}}}}
-  //   since-cxx11-note@#dr1891-b {{lambda expression begins here}}
+  //   since-cxx11-note@#cwg1891-b {{lambda expression begins here}}
 #endif
 }
 
-namespace dr1894 { // dr1894: 3.8
-                   // NB: reusing part of dr407 test
+namespace cwg1894 { // cwg1894: 3.8
+                   // NB: reusing part of cwg407 test
 namespace A {
   struct S {};
 }
diff --git a/clang/test/CXX/drs/dr19xx.cpp b/clang/test/CXX/drs/dr19xx.cpp
index f8c1581f08540..3e4f82813f51b 100644
--- a/clang/test/CXX/drs/dr19xx.cpp
+++ b/clang/test/CXX/drs/dr19xx.cpp
@@ -8,13 +8,13 @@
 
 namespace std { struct type_info; }
 
-namespace dr1902 { // dr1902: 3.7
+namespace cwg1902 { // cwg1902: 3.7
   struct A {};
   struct B {
-    B(A); // #dr1902-B-A
-    B() = delete; // #dr1902-B-ctor
+    B(A); // #cwg1902-B-A
+    B() = delete; // #cwg1902-B-ctor
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
-    B(const B&) = delete; // #dr1902-B-copy-ctor
+    B(const B&) = delete; // #cwg1902-B-copy-ctor
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
     operator A();
   };
@@ -22,19 +22,19 @@ namespace dr1902 { // dr1902: 3.7
   extern B b1;
   B b2(b1);
   // expected-error@-1 {{call to deleted constructor of 'B'}}
-  //   expected-note@#dr1902-B-copy-ctor {{'B' has been explicitly marked deleted here}}
+  //   expected-note@#cwg1902-B-copy-ctor {{'B' has been explicitly marked deleted here}}
 
 #if __cplusplus >= 201103L
   // This is ambiguous, even though calling the B(const B&) constructor would
   // both directly and indirectly call a deleted function.
   B b({});
   // since-cxx11-error@-1 {{call to constructor of 'B' is ambiguous}}
-  //   since-cxx11-note@#dr1902-B-A {{candidate constructor}}
-  //   since-cxx11-note@#dr1902-B-copy-ctor {{candidate constructor has been explicitly deleted}}
+  //   since-cxx11-note@#cwg1902-B-A {{candidate constructor}}
+  //   since-cxx11-note@#cwg1902-B-copy-ctor {{candidate constructor has been explicitly deleted}}
 #endif
 }
 
-namespace dr1903 { // dr1903: 2.7
+namespace cwg1903 { // cwg1903: 2.7
   namespace A {
     struct a {};
     int a;
@@ -59,7 +59,7 @@ namespace dr1903 { // dr1903: 2.7
   }
 }
 
-namespace dr1909 { // dr1909: 3.7
+namespace cwg1909 { // cwg1909: 3.7
   struct A {
     template<typename T> struct A {};
     // expected-error@-1 {{member 'A' has the same name as its class}}
@@ -80,7 +80,7 @@ namespace dr1909 { // dr1909: 3.7
   };
 }
 
-namespace dr1940 { // dr1940: 3.5
+namespace cwg1940 { // cwg1940: 3.5
 #if __cplusplus >= 201103L
 static union {
   static_assert(true, "");  // ok
@@ -91,7 +91,7 @@ static union {
 #endif
 }
 
-namespace dr1941 { // dr1941: 3.9
+namespace cwg1941 { // cwg1941: 3.9
 #if __cplusplus >= 201402L
 template<typename X>
 struct base {
@@ -121,7 +121,7 @@ derived d2(42, 9);
 #endif
 }
 
-namespace dr1947 { // dr1947: 3.5
+namespace cwg1947 { // cwg1947: 3.5
 #if __cplusplus >= 201402L
 unsigned o = 0'01;  // ok
 unsigned b = 0b'01;
@@ -132,26 +132,26 @@ unsigned x = 0x'01;
 }
 
 #if __cplusplus >= 201103L
-// dr1948: 3.5
+// cwg1948: 3.5
 // FIXME: This diagnostic could be improved.
 void *operator new(__SIZE_TYPE__) noexcept { return nullptr; }
 // since-cxx11-error@-1 {{exception specification in declaration does not match previous declaration}}
 #endif
 
-namespace dr1959 { // dr1959: 3.9
+namespace cwg1959 { // cwg1959: 3.9
 #if __cplusplus >= 201103L
   struct b;
   struct c;
   struct a {
     a() = default;
-    a(const a &) = delete; // #dr1959-copy-ctor
+    a(const a &) = delete; // #cwg1959-copy-ctor
     a(const b &) = delete; // not inherited
-    a(c &&) = delete; // #dr1959-move-ctor
-    template<typename T> a(T) = delete; // #dr1959-temp-ctor
+    a(c &&) = delete; // #cwg1959-move-ctor
+    template<typename T> a(T) = delete; // #cwg1959-temp-ctor
   };
 
-  struct b : a { // #dr1959-b
-    using a::a; // #dr1959-using-a
+  struct b : a { // #cwg1959-b
+    using a::a; // #cwg1959-using-a
   };
 
   a x;
@@ -160,15 +160,15 @@ namespace dr1959 { // dr1959: 3.9
   // where the base class is reference-related to the argument type.
   b y = x;
   // since-cxx11-error@-1 {{no viable conversion from 'a' to 'b'}}
-  //   since-cxx11-note@#dr1959-move-ctor {{candidate inherited constructor not viable: no known conversion from 'a' to 'c &&' for 1st argument}}
-  //   since-cxx11-note@#dr1959-using-a {{constructor from base class 'a' inherited here}}
-  //   since-cxx11-note@#dr1959-b {{candidate constructor (the implicit copy constructor) not viable: cannot bind base class object of type 'a' to derived class reference 'const b &' for 1st argument}}
-  //   since-cxx11-note@#dr1959-temp-ctor {{candidate template ignored: instantiation would take its own class type by value}}
-  //   since-cxx11-note@#dr1959-using-a {{constructor from base class 'a' inherited here}}
+  //   since-cxx11-note@#cwg1959-move-ctor {{candidate inherited constructor not viable: no known conversion from 'a' to 'c &&' for 1st argument}}
+  //   since-cxx11-note@#cwg1959-using-a {{constructor from base class 'a' inherited here}}
+  //   since-cxx11-note@#cwg1959-b {{candidate constructor (the implicit copy constructor) not viable: cannot bind base class object of type 'a' to derived class reference 'const b &' for 1st argument}}
+  //   since-cxx11-note@#cwg1959-temp-ctor {{candidate template ignored: instantiation would take its own class type by value}}
+  //   since-cxx11-note@#cwg1959-using-a {{constructor from base class 'a' inherited here}}
   b z = z;
   // since-cxx11-error@-1 {{call to implicitly-deleted copy constructor of 'b'}}
-  //   since-cxx11-note@#dr1959-b {{copy constructor of 'b' is implicitly deleted because base class 'a' has a deleted copy constructor}}
-  //   since-cxx11-note@#dr1959-copy-ctor {{'a' has been explicitly marked deleted here}}
+  //   since-cxx11-note@#cwg1959-b {{copy constructor of 'b' is implicitly deleted because base class 'a' has a deleted copy constructor}}
+  //   since-cxx11-note@#cwg1959-copy-ctor {{'a' has been explicitly marked deleted here}}
 
   struct c : a {
     using a::a;
@@ -181,7 +181,7 @@ namespace dr1959 { // dr1959: 3.9
 #endif
 }
 
-namespace dr1960 { // dr1960: no
+namespace cwg1960 { // cwg1960: no
 struct A {
 void f() {}
 protected:
@@ -202,7 +202,7 @@ using A::g;
 };
 }
 
-namespace dr1966 { // dr1966: 11
+namespace cwg1966 { // cwg1966: 11
 #if __cplusplus >= 201103L
   struct A {
     enum E : int {1};
@@ -215,12 +215,12 @@ namespace dr1966 { // dr1966: 11
   auto *p3 = true ? new enum G : int {};
   // since-cxx11-error@-1 {{ISO C++ forbids forward references to 'enum' types}}
   // since-cxx11-error@-2 {{allocation of incomplete type 'enum G'}}
-  //   since-cxx11-note@-3 {{forward declaration of 'dr1966::G'}}
+  //   since-cxx11-note@-3 {{forward declaration of 'cwg1966::G'}}
   auto h() -> enum E : int {};
   // since-cxx11-error@-1 {{non-defining declaration of enumeration with a fixed underlying type is only permitted as a standalone declaration}}
 
   enum X : enum Y : int {} {};
-  // since-cxx11-error@-1 {{'dr1966::Y' cannot be defined in a type specifier}}
+  // since-cxx11-error@-1 {{'cwg1966::Y' cannot be defined in a type specifier}}
   struct Q {
     // FIXME: can we emit something nicer than that?
     enum X : enum Y : int {} {};
@@ -231,9 +231,9 @@ namespace dr1966 { // dr1966: 11
 #endif
 }
 
-namespace dr1968 { // dr1968: no
+namespace cwg1968 { // cwg1968: no
 #if __cplusplus >= 201103L
-  // FIXME: According to DR1968, both of these should be considered
+  // FIXME: According to CWG1968, both of these should be considered
   // non-constant.
   static_assert(&typeid(int) == &typeid(int), "");
 
@@ -242,7 +242,7 @@ namespace dr1968 { // dr1968: no
 #endif
 }
 
-namespace dr1991 { // dr1991: 3.9
+namespace cwg1991 { // cwg1991: 3.9
 #if __cplusplus >= 201103L
   struct A {
     A(int, int) = delete;
@@ -260,4 +260,4 @@ namespace dr1991 { // dr1991: 3.9
 #endif
 }
 
-// dr1994: dup 529
+// cwg1994: dup 529
diff --git a/clang/test/CXX/drs/dr1xx.cpp b/clang/test/CXX/drs/dr1xx.cpp
index d55033cef1b64..5b497dda047d6 100644
--- a/clang/test/CXX/drs/dr1xx.cpp
+++ b/clang/test/CXX/drs/dr1xx.cpp
@@ -5,61 +5,61 @@
 // RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-unknown %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-unknown %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr100 { // dr100: yes
-  template<const char (*)[4]> struct A {}; // #dr100-A
-  template<const char (&)[4]> struct B {}; // #dr100-B
-  template<const char *> struct C {}; // #dr100-C
-  template<const char &> struct D {}; // #dr100-D
-  A<&"foo"> a; // #dr100-a
-  // cxx98-14-error@#dr100-a {{non-type template argument does not refer to any declaration}}
-  //   cxx98-14-note@#dr100-A {{template parameter is declared here}}
-  // since-cxx17-error@#dr100-a {{pointer to string literal is not allowed in a template argument}}
-  B<"bar"> b; // #dr100-b
-  // cxx98-14-error@#dr100-b {{non-type template argument does not refer to any declaration}}
-  //   cxx98-14-note@#dr100-B {{template parameter is declared here}}
-  // since-cxx17-error@#dr100-b {{reference to string literal is not allowed in a template argument}}
-  C<"baz"> c; // #dr100-c
-  // cxx98-14-error@#dr100-c {{non-type template argument does not refer to any declaration}}
-  //   cxx98-14-note@#dr100-C {{template parameter is declared here}}
-  // since-cxx17-error@#dr100-c {{pointer to subobject of string literal is not allowed in a template argument}}
-  D<*"quux"> d; // #dr100-d
-  // cxx98-14-error@#dr100-d {{non-type template argument does not refer to any declaration}}
-  //   cxx98-14-note@#dr100-D {{template parameter is declared here}}
-  // since-cxx17-error@#dr100-d {{reference to subobject of string literal is not allowed in a template argument}}
-}
-
-namespace dr101 { // dr101: 3.5
-  extern "C" void dr101_f();
+namespace cwg100 { // cwg100: yes
+  template<const char (*)[4]> struct A {}; // #cwg100-A
+  template<const char (&)[4]> struct B {}; // #cwg100-B
+  template<const char *> struct C {}; // #cwg100-C
+  template<const char &> struct D {}; // #cwg100-D
+  A<&"foo"> a; // #cwg100-a
+  // cxx98-14-error@#cwg100-a {{non-type template argument does not refer to any declaration}}
+  //   cxx98-14-note@#cwg100-A {{template parameter is declared here}}
+  // since-cxx17-error@#cwg100-a {{pointer to string literal is not allowed in a template argument}}
+  B<"bar"> b; // #cwg100-b
+  // cxx98-14-error@#cwg100-b {{non-type template argument does not refer to any declaration}}
+  //   cxx98-14-note@#cwg100-B {{template parameter is declared here}}
+  // since-cxx17-error@#cwg100-b {{reference to string literal is not allowed in a template argument}}
+  C<"baz"> c; // #cwg100-c
+  // cxx98-14-error@#cwg100-c {{non-type template argument does not refer to any declaration}}
+  //   cxx98-14-note@#cwg100-C {{template parameter is declared here}}
+  // since-cxx17-error@#cwg100-c {{pointer to subobject of string literal is not allowed in a template argument}}
+  D<*"quux"> d; // #cwg100-d
+  // cxx98-14-error@#cwg100-d {{non-type template argument does not refer to any declaration}}
+  //   cxx98-14-note@#cwg100-D {{template parameter is declared here}}
+  // since-cxx17-error@#cwg100-d {{reference to subobject of string literal is not allowed in a template argument}}
+}
+
+namespace cwg101 { // cwg101: 3.5
+  extern "C" void cwg101_f();
   typedef unsigned size_t;
   namespace X {
-    extern "C" void dr101_f();
+    extern "C" void cwg101_f();
     typedef unsigned size_t;
   }
-  using X::dr101_f;
+  using X::cwg101_f;
   using X::size_t;
-  extern "C" void dr101_f();
+  extern "C" void cwg101_f();
   typedef unsigned size_t;
 }
 
-namespace dr102 { // dr102: yes
+namespace cwg102 { // cwg102: yes
   namespace A {
     template<typename T> T f(T a, T b) { return a + b; }
     // expected-error@-1 {{call to function 'operator+' that is neither visible in the template definition nor found by argument-dependent lookup}}
-    //   expected-note@#dr102-instantiation {{in instantiation of function template specialization 'dr102::A::f<dr102::B::S>' requested here}}
-    //   expected-note@#dr102-operator-plus {{'operator+' should be declared prior to the call site or in namespace 'dr102::B'}}
+    //   expected-note@#cwg102-instantiation {{in instantiation of function template specialization 'cwg102::A::f<cwg102::B::S>' requested here}}
+    //   expected-note@#cwg102-operator-plus {{'operator+' should be declared prior to the call site or in namespace 'cwg102::B'}}
   }
   namespace B {
     struct S {};
   }
-  B::S operator+(B::S, B::S); // #dr102-operator-plus
-  template B::S A::f(B::S, B::S); // #dr102-instantiation
+  B::S operator+(B::S, B::S); // #cwg102-operator-plus
+  template B::S A::f(B::S, B::S); // #cwg102-instantiation
 }
 
-// dr103: na
-// dr104: na lib
-// dr105: na
+// cwg103: na
+// cwg104: na lib
+// cwg105: na
 
-namespace dr106 { // dr106: sup 540
+namespace cwg106 { // cwg106: sup 540
   typedef int &r1;
   typedef r1 &r1;
   typedef const r1 r1;
@@ -75,12 +75,12 @@ namespace dr106 { // dr106: sup 540
   // expected-warning@-1 {{'const' qualifier on reference type 'r2' (aka 'const int &') has no effect}}
 }
 
-namespace dr107 { // dr107: yes
+namespace cwg107 { // cwg107: yes
   struct S {};
   extern "C" S operator+(S, S) { return S(); }
 }
 
-namespace dr108 { // dr108: 2.9
+namespace cwg108 { // cwg108: 2.9
   template<typename T> struct A {
     struct B { typedef int X; };
     B::X x;
@@ -91,7 +91,7 @@ namespace dr108 { // dr108: 2.9
   template<> struct A<int>::B { int X; };
 }
 
-namespace dr109 { // dr109: yes
+namespace cwg109 { // cwg109: yes
   struct A { template<typename T> void f(T); };
   template<typename T> struct B : T {
     using T::template f;
@@ -108,37 +108,37 @@ namespace dr109 { // dr109: yes
   };
 }
 
-namespace dr111 { // dr111: dup 535
+namespace cwg111 { // cwg111: dup 535
   struct A { A(); A(volatile A&, int = 0); A(A&, const char * = "foo"); };
-  struct B : A { B(); }; // #dr111-B
+  struct B : A { B(); }; // #cwg111-B
   const B b1;
   B b2(b1);
   // expected-error@-1 {{no matching constructor for initialization of 'B'}}
-  //   expected-note@#dr111-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const B') would lose const qualifier}}
-  //   expected-note@#dr111-B {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
+  //   expected-note@#cwg111-B {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('const B') would lose const qualifier}}
+  //   expected-note@#cwg111-B {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
 }
 
-namespace dr112 { // dr112: yes
+namespace cwg112 { // cwg112: yes
   struct T { int n; };
   typedef T Arr[1];
 
-  const T a1[1] = {}; // #dr112-a1
+  const T a1[1] = {}; // #cwg112-a1
   volatile T a2[1] = {};
-  const Arr a3 = {}; // #dr112-a3
+  const Arr a3 = {}; // #cwg112-a3
   volatile Arr a4 = {};
   template<const volatile T*> struct X {};
   // FIXME: Test this somehow in C++11 and on.
   X<a1> x1;
   // cxx98-error@-1 {{non-type template argument referring to object 'a1' with internal linkage is a C++11 extension}}
-  //   cxx98-note@#dr112-a1 {{non-type template argument refers to object here}}
+  //   cxx98-note@#cwg112-a1 {{non-type template argument refers to object here}}
   X<a2> x2;
   X<a3> x3;
   // cxx98-error@-1 {{non-type template argument referring to object 'a3' with internal linkage is a C++11 extension}}
-  //   cxx98-note@#dr112-a3 {{non-type template argument refers to object here}}
+  //   cxx98-note@#cwg112-a3 {{non-type template argument refers to object here}}
   X<a4> x4;
 }
 
-namespace dr113 { // dr113: yes
+namespace cwg113 { // cwg113: yes
   extern void (*p)();
   void f() {
     no_such_function();
@@ -149,50 +149,50 @@ namespace dr113 { // dr113: yes
   void (*p)() = &g;
 }
 
-namespace dr114 { // dr114: yes
+namespace cwg114 { // cwg114: yes
   struct A {
-    virtual void f(int) = 0; // #dr114-A-f
+    virtual void f(int) = 0; // #cwg114-A-f
   };
   struct B : A {
     template<typename T> void f(T);
     void g() { f(0); }
   } b;
   // expected-error@-1 {{variable type 'struct B' is an abstract class}}
-  //   expected-note@#dr114-A-f {{unimplemented pure virtual method 'f' in 'B'}}
+  //   expected-note@#cwg114-A-f {{unimplemented pure virtual method 'f' in 'B'}}
 }
 
-namespace dr115 { // dr115: 3.0
-  template<typename T> int f(T); // #dr115-f
-  template<typename T> int g(T); // #dr115-g
-  template<typename T> int g(T, int); // #dr115-g-int
+namespace cwg115 { // cwg115: 3.0
+  template<typename T> int f(T); // #cwg115-f
+  template<typename T> int g(T); // #cwg115-g
+  template<typename T> int g(T, int); // #cwg115-g-int
 
   int k1 = f(&f);
   // expected-error@-1 {{no matching function for call to 'f'}}
-  //   expected-note@#dr115-f {{candidate template ignored: couldn't infer template argument 'T'}}
+  //   expected-note@#cwg115-f {{candidate template ignored: couldn't infer template argument 'T'}}
   int k2 = f(&f<int>);
   int k3 = f(&g<int>);
   // expected-error@-1 {{no matching function for call to 'f'}}
-  //   expected-note@#dr115-f {{candidate template ignored: couldn't infer template argument 'T'}}
+  //   expected-note@#cwg115-f {{candidate template ignored: couldn't infer template argument 'T'}}
 
   void h() {
     (void)&f;
     // expected-error@-1 {{address of overloaded function 'f' cannot be cast to type 'void'}}
-    //   expected-note@#dr115-f {{candidate function template}}
+    //   expected-note@#cwg115-f {{candidate function template}}
     (void)&f<int>;
     (void)&g<int>;
     // expected-error@-1 {{address of overloaded function 'g' cannot be cast to type 'void'}}
-    //   expected-note@#dr115-g-int {{candidate function template}}
-    //   expected-note@#dr115-g {{candidate function template}}
+    //   expected-note@#cwg115-g-int {{candidate function template}}
+    //   expected-note@#cwg115-g {{candidate function template}}
 
     &f;
     // expected-error@-1 {{reference to overloaded function could not be resolved; did you mean to call it?}}
-    //   expected-note@#dr115-f {{possible target for call}}
+    //   expected-note@#cwg115-f {{possible target for call}}
     &f<int>;
     // expected-warning@-1 {{expression result unused}}
     &g<int>;
     // expected-error@-1 {{reference to overloaded function could not be resolved; did you mean to call it?}}
-    //   expected-note@#dr115-g-int {{possible target for call}}
-    //   expected-note@#dr115-g {{possible target for call}}
+    //   expected-note@#cwg115-g-int {{possible target for call}}
+    //   expected-note@#cwg115-g {{possible target for call}}
   }
 
   struct S {
@@ -251,43 +251,43 @@ namespace dr115 { // dr115: 3.0
 
 #if __cplusplus >= 201103L
   // Special case kicks in only if a template argument list is specified.
-  template<typename T=int> void with_default(); // #dr115-with-default
+  template<typename T=int> void with_default(); // #cwg115-with-default
   int k10 = f(&with_default);
   // expected-error@-1 {{no matching function for call to 'f'}}
-  //   expected-note@#dr115-f {{candidate template ignored: couldn't infer template argument 'T'}}
+  //   expected-note@#cwg115-f {{candidate template ignored: couldn't infer template argument 'T'}}
   int k11 = f(&with_default<>);
   void k() {
     (void)&with_default;
     // expected-error@-1 {{address of overloaded function 'with_default' cannot be cast to type 'void'}}
-    //   expected-note@#dr115-with-default {{candidate function template}}
+    //   expected-note@#cwg115-with-default {{candidate function template}}
     (void)&with_default<>;
     &with_default;
     // expected-error@-1 {{reference to overloaded function could not be resolved; did you mean to call it?}}
-    //   expected-note@#dr115-with-default {{possible target for call}}
+    //   expected-note@#cwg115-with-default {{possible target for call}}
     &with_default<>;
     // expected-warning@-1 {{expression result unused}}
   }
 #endif
 }
 
-namespace dr116 { // dr116: yes
+namespace cwg116 { // cwg116: yes
   template<int> struct A {};
-  template<int N> void f(A<N>) {} // #dr116-f-N
+  template<int N> void f(A<N>) {} // #cwg116-f-N
   template<int M> void f(A<M>) {}
   // expected-error@-1 {{redefinition of 'f'}}
-  //   expected-note@#dr116-f-N {{previous definition is here}}
-  template<typename T> void f(A<sizeof(T)>) {} // #dr116-f-T
+  //   expected-note@#cwg116-f-N {{previous definition is here}}
+  template<typename T> void f(A<sizeof(T)>) {} // #cwg116-f-T
   template<typename U> void f(A<sizeof(U)>) {}
   // expected-error@-1 {{redefinition of 'f'}}
-  //   expected-note@#dr116-f-T {{previous definition is here}}
+  //   expected-note@#cwg116-f-T {{previous definition is here}}
 }
 
-// dr117: na
-// dr118 is in dr118.cpp
-// dr119: na
-// dr120: na
+// cwg117: na
+// cwg118 is in cwg118.cpp
+// cwg119: na
+// cwg120: na
 
-namespace dr121 { // dr121: yes
+namespace cwg121 { // cwg121: yes
   struct X {
     template<typename T> struct Y {};
   };
@@ -300,32 +300,32 @@ namespace dr121 { // dr121: yes
   Z<X> z;
 }
 
-namespace dr122 { // dr122: yes
+namespace cwg122 { // cwg122: yes
   template<typename T> void f();
   void g() { f<int>(); }
 }
 
-// dr123: na
-// dr124 is in dr124.cpp
+// cwg123: na
+// cwg124 is in cwg124.cpp
 
-// dr125: yes
-struct dr125_A { struct dr125_B {}; }; // #dr125_B
-dr125_A::dr125_B dr125_C();
-namespace dr125_B { dr125_A dr125_C(); }
-namespace dr125 {
+// cwg125: yes
+struct cwg125_A { struct cwg125_B {}; }; // #cwg125_B
+cwg125_A::cwg125_B cwg125_C();
+namespace cwg125_B { cwg125_A cwg125_C(); }
+namespace cwg125 {
   struct X {
-    friend dr125_A::dr125_B (::dr125_C)(); // ok
-    friend dr125_A (::dr125_B::dr125_C)(); // ok
-    friend dr125_A::dr125_B::dr125_C(); // #dr125_C
-    // expected-error@#dr125_C {{missing return type for function 'dr125_C'; did you mean the constructor name 'dr125_B'?}}
-    // cxx98-error@#dr125_C {{'dr125_B' is missing exception specification 'throw()'}}
-    //   cxx98-note@#dr125_B {{previous declaration is here}}
-    // since-cxx11-error@#dr125_C {{'dr125_B' is missing exception specification 'noexcept'}}
-    //   since-cxx11-note@#dr125_B {{previous declaration is here}}
+    friend cwg125_A::cwg125_B (::cwg125_C)(); // ok
+    friend cwg125_A (::cwg125_B::cwg125_C)(); // ok
+    friend cwg125_A::cwg125_B::cwg125_C(); // #cwg125_C
+    // expected-error@#cwg125_C {{missing return type for function 'cwg125_C'; did you mean the constructor name 'cwg125_B'?}}
+    // cxx98-error@#cwg125_C {{'cwg125_B' is missing exception specification 'throw()'}}
+    //   cxx98-note@#cwg125_B {{previous declaration is here}}
+    // since-cxx11-error@#cwg125_C {{'cwg125_B' is missing exception specification 'noexcept'}}
+    //   since-cxx11-note@#cwg125_B {{previous declaration is here}}
   };
 }
 
-namespace dr126 { // dr126: partial
+namespace cwg126 { // cwg126: partial
   // FIXME: We do not yet generate correct code for this change:
   // eg:
   //   catch (void*&) should catch void* but not int*
@@ -350,23 +350,23 @@ namespace dr126 { // dr126: partial
   struct A {
     virtual void cp() throw(C*);
     virtual void dp() throw(C*);
-    virtual void ep() throw(C*); // #dr126-ep
-    virtual void fp() throw(C*); // #dr126-fp
+    virtual void ep() throw(C*); // #cwg126-ep
+    virtual void fp() throw(C*); // #cwg126-fp
     virtual void gp() throw(C*);
-    virtual void hp() throw(C*); // #dr126-hp
+    virtual void hp() throw(C*); // #cwg126-hp
 
     virtual void cr() throw(C&);
     virtual void dr() throw(C&);
-    virtual void er() throw(C&); // #dr126-er
-    virtual void fr() throw(C&); // #dr126-fr
+    virtual void er() throw(C&); // #cwg126-er
+    virtual void fr() throw(C&); // #cwg126-fr
     virtual void gr() throw(C&);
-    virtual void hr() throw(C&); // #dr126-hr
+    virtual void hr() throw(C&); // #cwg126-hr
 
     virtual void pv() throw(void*);
 
     virtual void np() throw(C*);
     virtual void npm() throw(int C::*);
-    virtual void nr() throw(C*&); // #dr126-nr
+    virtual void nr() throw(C*&); // #cwg126-nr
     virtual void ncr() throw(C*const&);
 
     virtual void ref1() throw(C *const&);
@@ -374,36 +374,36 @@ namespace dr126 { // dr126: partial
 
     virtual void v() throw(int);
     virtual void w() throw(const int);
-    virtual void x() throw(int*); // #dr126-x
+    virtual void x() throw(int*); // #cwg126-x
     virtual void y() throw(const int*);
-    virtual void z() throw(int); // #dr126-z
+    virtual void z() throw(int); // #cwg126-z
   };
   struct B : A {
     virtual void cp() throw(C*);
     virtual void dp() throw(D*);
     virtual void ep() throw(E*);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-ep {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-ep {{overridden virtual function is here}}
     virtual void fp() throw(F*);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-fp {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-fp {{overridden virtual function is here}}
     virtual void gp() throw(G*);
     virtual void hp() throw(H*);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-hp {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-hp {{overridden virtual function is here}}
 
     virtual void cr() throw(C&);
     virtual void dr() throw(D&);
     virtual void er() throw(E&);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-er {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-er {{overridden virtual function is here}}
     virtual void fr() throw(F&);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-fr {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-fr {{overridden virtual function is here}}
     virtual void gr() throw(G&);
     virtual void hr() throw(H&);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-hr {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-hr {{overridden virtual function is here}}
 
     virtual void pv() throw(C*);
 
@@ -413,7 +413,7 @@ namespace dr126 { // dr126: partial
     virtual void npm() throw(nullptr_t&);
     virtual void nr() throw(nullptr_t);
     // cxx11-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx11-14-note@#dr126-nr {{overridden virtual function is here}}
+    //   cxx11-14-note@#cwg126-nr {{overridden virtual function is here}}
     virtual void ncr() throw(nullptr_t);
 #endif // __cplusplus >= 201103L
 
@@ -424,11 +424,11 @@ namespace dr126 { // dr126: partial
     virtual void w() throw(int);
     virtual void x() throw(const int*);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-x {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-x {{overridden virtual function is here}}
     virtual void y() throw(int*); // ok
     virtual void z() throw(long);
     // cxx98-14-error@-1 {{exception specification of overriding function is more lax than base version}}
-    //   cxx98-14-note@#dr126-z {{overridden virtual function is here}}
+    //   cxx98-14-note@#cwg126-z {{overridden virtual function is here}}
   };
 #endif // __cplusplus <= 201402L
   void f() throw(int);
@@ -436,37 +436,37 @@ namespace dr126 { // dr126: partial
   //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
 }
 
-namespace dr127 { // dr127: 2.9
+namespace cwg127 { // cwg127: 2.9
   __extension__ typedef __decltype(sizeof(0)) size_t;
   template<typename T> struct A {
     A() { throw 0; }
     void *operator new(size_t, const char * = 0);
-    void operator delete(void *, const char *) { T::error; } // #dr127-delete-const-char
-    // expected-error@#dr127-delete-const-char {{type 'void' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr127-p {{in instantiation of member function 'dr127::A<void>::operator delete' requested here}}
-    // expected-error@#dr127-delete-const-char {{type 'int' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr127-q {{in instantiation of member function 'dr127::A<int>::operator delete' requested here}}
+    void operator delete(void *, const char *) { T::error; } // #cwg127-delete-const-char
+    // expected-error@#cwg127-delete-const-char {{type 'void' cannot be used prior to '::' because it has no members}}
+    //   expected-note@#cwg127-p {{in instantiation of member function 'cwg127::A<void>::operator delete' requested here}}
+    // expected-error@#cwg127-delete-const-char {{type 'int' cannot be used prior to '::' because it has no members}}
+    //   expected-note@#cwg127-q {{in instantiation of member function 'cwg127::A<int>::operator delete' requested here}}
     void operator delete(void *) { T::error; }
   };
-  A<void> *p = new A<void>; // #dr127-p
-  A<int> *q = new ("") A<int>; // #dr127-q
+  A<void> *p = new A<void>; // #cwg127-p
+  A<int> *q = new ("") A<int>; // #cwg127-q
 }
 
-namespace dr128 { // dr128: yes
+namespace cwg128 { // cwg128: yes
   enum E1 { e1 } x = e1;
   enum E2 { e2 } y = static_cast<E2>(x), z = static_cast<E2>(e1);
 }
 
-// dr129: dup 616
-// dr130: na
+// cwg129: dup 616
+// cwg130: na
 
-namespace dr131 { // dr131: sup P1949
+namespace cwg131 { // cwg131: sup P1949
   const char *a_with_\u0e8c = "\u0e8c";
   const char *b_with_\u0e8d = "\u0e8d";
   const char *c_with_\u0e8e = "\u0e8e";
 }
 
-namespace dr132 { // dr132: no
+namespace cwg132 { // cwg132: no
   void f() {
     extern struct {} x; // ok
     extern struct S {} y; // FIXME: This is invalid.
@@ -474,10 +474,10 @@ namespace dr132 { // dr132: no
   static enum { E } e;
 }
 
-// dr133: dup 87
-// dr134: na
+// cwg133: dup 87
+// cwg134: na
 
-namespace dr135 { // dr135: yes
+namespace cwg135 { // cwg135: yes
   struct A {
     A f(A a) { return a; }
     friend A g(A a) { return a; }
@@ -485,55 +485,55 @@ namespace dr135 { // dr135: yes
   };
 }
 
-namespace dr136 { // dr136: 3.4
-  void f(int, int, int = 0); // #dr136-f
-  void g(int, int, int); // #dr136-g
+namespace cwg136 { // cwg136: 3.4
+  void f(int, int, int = 0); // #cwg136-f
+  void g(int, int, int); // #cwg136-g
   struct A {
     friend void f(int, int = 0, int);
     // expected-error@-1 {{friend declaration specifying a default argument must be the only declaration}}
-    //   expected-note@#dr136-f {{previous declaration is here}}
+    //   expected-note@#cwg136-f {{previous declaration is here}}
     friend void g(int, int, int = 0);
     // expected-error@-1 {{friend declaration specifying a default argument must be the only declaration}}
-    //   expected-note@#dr136-g {{previous declaration is here}}
+    //   expected-note@#cwg136-g {{previous declaration is here}}
     friend void h(int, int, int = 0);
     // expected-error@-1 {{friend declaration specifying a default argument must be a definition}}
-    friend void i(int, int, int = 0) {} // #dr136-A-i
+    friend void i(int, int, int = 0) {} // #cwg136-A-i
     friend void j(int, int, int = 0) {}
     operator int();
   };
   void i(int, int, int);
   // expected-error@-1 {{friend declaration specifying a default argument must be the only declaration}}
-  //   expected-note@#dr136-A-i {{previous declaration is here}}
+  //   expected-note@#cwg136-A-i {{previous declaration is here}}
   void q() {
     j(A(), A()); // ok, has default argument
   }
-  extern "C" void k(int, int, int, int); // #dr136-k 
+  extern "C" void k(int, int, int, int); // #cwg136-k 
   namespace NSA {
   struct A {
-    friend void dr136::k(int, int, int, int = 0);
+    friend void cwg136::k(int, int, int, int = 0);
     // expected-error@-1 {{friend declaration specifying a default argument must be the only declaration}}
-    //   expected-note@#dr136-k {{previous declaration is here}}
+    //   expected-note@#cwg136-k {{previous declaration is here}}
   };
   }
   namespace NSB {
   struct A {
-    friend void dr136::k(int, int, int = 0, int); // #dr136-friend-k
-    // expected-error@#dr136-friend-k {{friend declaration specifying a default argument must be the only declaration}}
-    //   expected-note@#dr136-k {{previous declaration is here}}
-    // expected-error@#dr136-friend-k {{missing default argument on parameter}}
+    friend void cwg136::k(int, int, int = 0, int); // #cwg136-friend-k
+    // expected-error@#cwg136-friend-k {{friend declaration specifying a default argument must be the only declaration}}
+    //   expected-note@#cwg136-k {{previous declaration is here}}
+    // expected-error@#cwg136-friend-k {{missing default argument on parameter}}
   };
   }
   struct B {
-    void f(int); // #dr136-B-f
+    void f(int); // #cwg136-B-f
   };
   struct C {
     friend void B::f(int = 0);
     // expected-error@-1 {{friend declaration specifying a default argument must be the only declaration}}
-    //   expected-note@#dr136-B-f {{previous declaration is here}}
+    //   expected-note@#cwg136-B-f {{previous declaration is here}}
   };
 }
 
-namespace dr137 { // dr137: yes
+namespace cwg137 { // cwg137: yes
   extern void *p;
   extern const void *cp;
   extern volatile void *vp;
@@ -557,13 +557,13 @@ namespace dr137 { // dr137: yes
   const volatile int *cvqcv = static_cast<const volatile int*>(cvp);
 }
 
-namespace dr139 { // dr139: yes
+namespace cwg139 { // cwg139: yes
   namespace example1 {
-    typedef int f; // #dr139-typedef-f
+    typedef int f; // #cwg139-typedef-f
     struct A {
       friend void f(A &);
       // expected-error@-1 {{redefinition of 'f' as different kind of symbol}}
-      //   expected-note@#dr139-typedef-f {{previous definition is here}}
+      //   expected-note@#cwg139-typedef-f {{previous definition is here}}
     };
   }
 
@@ -579,34 +579,34 @@ namespace dr139 { // dr139: yes
   }
 }
 
-namespace dr140 { // dr140: yes
-  void f(int *const) {} // #dr140-f-first
+namespace cwg140 { // cwg140: yes
+  void f(int *const) {} // #cwg140-f-first
   void f(int[3]) {}
   // expected-error@-1 {{redefinition of 'f'}}
-  //   expected-note@#dr140-f-first {{previous definition is here}}
+  //   expected-note@#cwg140-f-first {{previous definition is here}}
   void g(const int);
   void g(int n) { n = 2; }
 }
 
-namespace dr141 { // dr141: 3.1
+namespace cwg141 { // cwg141: 3.1
   template<typename T> void f();
-  template<typename T> struct S { int n; }; // #dr141-S
+  template<typename T> struct S { int n; }; // #cwg141-S
   struct A : S<int> {
     template<typename T> void f();
-    template<typename T> struct S {}; // #dr141-A-S
+    template<typename T> struct S {}; // #cwg141-A-S
   } a;
   struct B : S<int> {} b;
   void g() {
     a.f<int>();
-    (void)a.S<int>::n; // #dr141-a
-    // cxx98-error@#dr141-a {{lookup of 'S' in member access expression is ambiguous; using member of 'struct A'}}
-    //   cxx98-note@#dr141-A-S {{lookup in the object type 'struct A' refers here}}
-    //   cxx98-note@#dr141-S {{lookup from the current scope refers here}}
-    // expected-error@#dr141-a {{no member named 'n' in 'dr141::A::S<int>'; did you mean '::dr141::S<int>::n'?}}
-    //   expected-note@#dr141-S {{'::dr141::S<int>::n' declared here}}
+    (void)a.S<int>::n; // #cwg141-a
+    // cxx98-error@#cwg141-a {{lookup of 'S' in member access expression is ambiguous; using member of 'struct A'}}
+    //   cxx98-note@#cwg141-A-S {{lookup in the object type 'struct A' refers here}}
+    //   cxx98-note@#cwg141-S {{lookup from the current scope refers here}}
+    // expected-error@#cwg141-a {{no member named 'n' in 'cwg141::A::S<int>'; did you mean '::cwg141::S<int>::n'?}}
+    //   expected-note@#cwg141-S {{'::cwg141::S<int>::n' declared here}}
     // FIXME: we issue a useful diagnostic first, then some bogus ones.
     b.f<int>();
-    // expected-error@-1 {{no member named 'f' in 'dr141::B'}}
+    // expected-error@-1 {{no member named 'f' in 'cwg141::B'}}
     // expected-error@-2 +{{}}
     (void)b.S<int>::n;
   }
@@ -630,60 +630,60 @@ namespace dr141 { // dr141: 3.1
   void i() { C<X>().i(); } // ok!!
 }
 
-namespace dr142 { // dr142: 2.8
-  class B { // #dr142-B
+namespace cwg142 { // cwg142: 2.8
+  class B { // #cwg142-B
   public:
-    int mi; // #dr142-B-mi
-    static int si; // #dr142-B-si
+    int mi; // #cwg142-B-mi
+    static int si; // #cwg142-B-si
   };
-  class D : private B { // #dr142-D
+  class D : private B { // #cwg142-D
   };
   class DD : public D {
     void f();
   };
   void DD::f() {
     mi = 3;
-    // expected-error@-1 {{'mi' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B-mi {{member is declared here}}
+    // expected-error@-1 {{'mi' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B-mi {{member is declared here}}
     si = 3;
-    // expected-error@-1 {{'si' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B-si {{member is declared here}}
+    // expected-error@-1 {{'si' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B-si {{member is declared here}}
     B b_old;
-    // expected-error@-1 {{'B' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B {{member is declared here}}
-    dr142::B b;
+    // expected-error@-1 {{'B' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B {{member is declared here}}
+    cwg142::B b;
     b.mi = 3;
     b.si = 3;
     B::si = 3;
-    // expected-error@-1 {{'B' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B {{member is declared here}}
-    dr142::B::si = 3;
-    B *bp1_old = this; // #dr142-bp1_old
-    // expected-error@#dr142-bp1_old {{'B' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B {{member is declared here}}
-    // expected-error@#dr142-bp1_old {{cannot cast 'dr142::DD' to its private base class 'B'}}
-    //   expected-note@#dr142-D {{declared private here}}
-    dr142::B *bp1 = this;
-    // expected-error@-1 {{cannot cast 'dr142::DD' to its private base class 'dr142::B'}}
-    //   expected-note@#dr142-D {{declared private here}}
-    B *bp2_old = (B*)this; // #dr142-bp2_old
-    // expected-error@#dr142-bp2_old {{'B' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B {{member is declared here}}
-    // expected-error@#dr142-bp2_old {{'B' is a private member of 'dr142::B'}}
-    //   expected-note@#dr142-D {{constrained by private inheritance here}}
-    //   expected-note@#dr142-B {{member is declared here}}
-    dr142::B *bp2 = (dr142::B*)this;
+    // expected-error@-1 {{'B' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B {{member is declared here}}
+    cwg142::B::si = 3;
+    B *bp1_old = this; // #cwg142-bp1_old
+    // expected-error@#cwg142-bp1_old {{'B' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B {{member is declared here}}
+    // expected-error@#cwg142-bp1_old {{cannot cast 'cwg142::DD' to its private base class 'B'}}
+    //   expected-note@#cwg142-D {{declared private here}}
+    cwg142::B *bp1 = this;
+    // expected-error@-1 {{cannot cast 'cwg142::DD' to its private base class 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{declared private here}}
+    B *bp2_old = (B*)this; // #cwg142-bp2_old
+    // expected-error@#cwg142-bp2_old {{'B' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B {{member is declared here}}
+    // expected-error@#cwg142-bp2_old {{'B' is a private member of 'cwg142::B'}}
+    //   expected-note@#cwg142-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg142-B {{member is declared here}}
+    cwg142::B *bp2 = (cwg142::B*)this;
     bp2->mi = 3;
   }
 }
 
-namespace dr143 { // dr143: yes
+namespace cwg143 { // cwg143: yes
   namespace A { struct X; }
   namespace B { void f(A::X); }
   namespace A {
@@ -695,7 +695,7 @@ namespace dr143 { // dr143: yes
   }
 }
 
-namespace dr145 { // dr145: yes
+namespace cwg145 { // cwg145: yes
   void f(bool b) {
     ++b;
     // cxx98-14-warning@-1 {{incrementing expression of type bool is deprecated and incompatible with C++17}}
@@ -706,7 +706,7 @@ namespace dr145 { // dr145: yes
   }
 }
 
-namespace dr147 { // dr147: yes
+namespace cwg147 { // cwg147: yes
   namespace example1 {
     template<typename> struct A {
       template<typename T> A(T);
@@ -734,15 +734,15 @@ namespace dr147 { // dr147: yes
   }
 }
 
-namespace dr148 { // dr148: yes
+namespace cwg148 { // cwg148: yes
   struct A { int A::*p; };
   int check1[__is_pod(int(A::*)) ? 1 : -1];
   int check2[__is_pod(A) ? 1 : -1];
 }
 
-// dr149: na
+// cwg149: na
 
-namespace dr151 { // dr151: 3.1
+namespace cwg151 { // cwg151: 3.1
   struct X {};
   typedef int X::*p;
 #if __cplusplus < 201103L
@@ -754,28 +754,28 @@ namespace dr151 { // dr151: 3.1
 #undef fold
 }
 
-namespace dr152 { // dr152: yes
+namespace cwg152 { // cwg152: yes
   struct A {
-    A(); // #dr152-A-ctor
-    explicit A(const A&); // #dr152-A-explicit-ctor
+    A(); // #cwg152-A-ctor
+    explicit A(const A&); // #cwg152-A-explicit-ctor
   };
   A a1 = A();
   // cxx98-14-error@-1 {{no matching constructor for initialization of 'A'}}
-  //   cxx98-14-note@#dr152-A-explicit-ctor {{explicit constructor is not a candidate}}
-  //   cxx98-14-note@#dr152-A-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
+  //   cxx98-14-note@#cwg152-A-explicit-ctor {{explicit constructor is not a candidate}}
+  //   cxx98-14-note@#cwg152-A-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
   A a2((A()));
 
   A &f();
   A a3 = f();
   // expected-error@-1 {{no matching constructor for initialization of 'A'}}
-  //   expected-note@#dr152-A-explicit-ctor {{explicit constructor is not a candidate}}
-  //   expected-note@#dr152-A-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
+  //   expected-note@#cwg152-A-explicit-ctor {{explicit constructor is not a candidate}}
+  //   expected-note@#cwg152-A-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
   A a4(f());
 }
 
-// dr153: na
+// cwg153: na
 
-namespace dr154 { // dr154: yes
+namespace cwg154 { // cwg154: yes
   union { int a; };
   // expected-error@-1 {{nonymous unions at namespace or global scope must be declared 'static'}}
   namespace {
@@ -784,29 +784,29 @@ namespace dr154 { // dr154: yes
   static union { int c; };
 }
 
-namespace dr155 { // dr155: dup 632
+namespace cwg155 { // cwg155: dup 632
   struct S { int n; } s = { { 1 } };
   // expected-warning@-1 {{braces around scalar initializer}}
 }
 
-// dr158 is in dr158.cpp
+// cwg158 is in cwg158.cpp
 
-namespace dr159 { // dr159: 3.5
+namespace cwg159 { // cwg159: 3.5
   namespace X { void f(); }
   void f();
-  void dr159::f() {}
+  void cwg159::f() {}
   // expected-warning@-1 {{extra qualification on member 'f'}}
-  void dr159::X::f() {}
+  void cwg159::X::f() {}
 }
 
-// dr160: na
+// cwg160: na
 
-namespace dr161 { // dr161: 3.1
+namespace cwg161 { // cwg161: 3.1
   class A {
   protected:
-    struct B { int n; } b; // #dr161-B
+    struct B { int n; } b; // #cwg161-B
     static B bs;
-    void f(); // #dr161-f
+    void f(); // #cwg161-f
     static void sf();
   };
   struct C : A {};
@@ -816,18 +816,18 @@ namespace dr161 { // dr161: 3.1
       B b1;
       C::B b2; // ok, accessible as a member of A
       (void)&C::b;
-      // expected-error@-1 {{'b' is a protected member of 'dr161::A'}}
-      //   expected-note@#dr161-B {{declared protected here}}
+      // expected-error@-1 {{'b' is a protected member of 'cwg161::A'}}
+      //   expected-note@#cwg161-B {{declared protected here}}
       (void)&C::bs;
       (void)c.b;
-      // expected-error@-1 {{'b' is a protected member of 'dr161::A'}}
-      //   expected-note@#dr161-B {{declared protected here}}
+      // expected-error@-1 {{'b' is a protected member of 'cwg161::A'}}
+      //   expected-note@#cwg161-B {{declared protected here}}
       (void)c.bs;
       f();
       sf();
       c.f();
       // expected-error@-1 {{protected}}
-      //   expected-note@#dr161-f {{declared protected here}}
+      //   expected-note@#cwg161-f {{declared protected here}}
       c.sf();
       A::f();
       D::f();
@@ -838,7 +838,7 @@ namespace dr161 { // dr161: 3.1
   };
 }
 
-namespace dr162 { // dr162: no
+namespace cwg162 { // cwg162: no
   struct A {
     char &f(char);
     static int &f(int);
@@ -857,9 +857,9 @@ namespace dr162 { // dr162: no
   // expected-error@-1 {{reference to overloaded function could not be resolved; did you mean to call it?}}
 }
 
-// dr163: na
+// cwg163: na
 
-namespace dr164 { // dr164: yes
+namespace cwg164 { // cwg164: yes
   void f(int);
   template <class T> int g(T t) { return f(t); }
 
@@ -869,35 +869,35 @@ namespace dr164 { // dr164: yes
   int k = g(e);
 }
 
-namespace dr165 { // dr165: no
+namespace cwg165 { // cwg165: no
   namespace N {
     struct A { friend struct B; };
     void f() { void g(); }
   }
-  // FIXME: dr1477 says this is ok, dr165 says it's ill-formed
+  // FIXME: cwg1477 says this is ok, cwg165 says it's ill-formed
   struct N::B {};
-  // FIXME: dr165 says this is ill-formed, but the argument in dr1477 says it's ok
+  // FIXME: cwg165 says this is ill-formed, but the argument in cwg1477 says it's ok
   void N::g() {}
 }
 
-namespace dr166 { // dr166: 2.9
+namespace cwg166 { // cwg166: 2.9
   namespace A { class X; }
 
   template<typename T> int f(T t) { return t.n; }
   int g(A::X);
   template<typename T> int h(T t) { return t.n; }
-  // expected-error@-1 {{'n' is a private member of 'dr166::A::X'}}
-  //   expected-note@#dr166-h-instantiation {{in instantiation of function template specialization 'dr166::h<dr166::A::X>' requested here}}
-  //   expected-note@#dr166-X-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg166::A::X'}}
+  //   expected-note@#cwg166-h-instantiation {{in instantiation of function template specialization 'cwg166::h<cwg166::A::X>' requested here}}
+  //   expected-note@#cwg166-X-n {{implicitly declared private here}}
   int i(A::X);
 
   namespace A {
     class X {
       friend int f<X>(X);
-      friend int dr166::g(X);
+      friend int cwg166::g(X);
       friend int h(X);
       friend int i(X);
-      int n; // #dr166-X-n
+      int n; // #cwg166-X-n
     };
 
     int h(X x) { return x.n; }
@@ -906,15 +906,15 @@ namespace dr166 { // dr166: 2.9
 
   template int f(A::X);
   int g(A::X x) { return x.n; }
-  template int h(A::X); // #dr166-h-instantiation
+  template int h(A::X); // #cwg166-h-instantiation
   int i(A::X x) { return x.n; }
-  // expected-error@-1 {{'n' is a private member of 'dr166::A::X'}}
-  //   expected-note@#dr166-X-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg166::A::X'}}
+  //   expected-note@#cwg166-X-n {{implicitly declared private here}}
 }
 
-// dr167: sup 1012
+// cwg167: sup 1012
 
-namespace dr168 { // dr168: no
+namespace cwg168 { // cwg168: no
   extern "C" typedef int (*p)();
   extern "C++" typedef int (*q)();
   struct S {
@@ -924,7 +924,7 @@ namespace dr168 { // dr168: no
   q b = &S::f;
 }
 
-namespace dr169 { // dr169: yes
+namespace cwg169 { // cwg169: yes
   template<typename> struct A { int n; };
   struct B {
     template<typename> struct C;
@@ -943,18 +943,18 @@ namespace dr169 { // dr169: yes
   };
 }
 
-namespace { // dr171: 3.4
-  int dr171a;
+namespace { // cwg171: 3.4
+  int cwg171a;
 }
-int dr171b; // #dr171b-int
-namespace dr171 {
-  extern "C" void dr171a();
-  extern "C" void dr171b();
-  // expected-error@-1 {{declaration of 'dr171b' with C language linkage conflicts with declaration in global scope}}
-  //   expected-note@#dr171b-int {{declared in global scope here}}
+int cwg171b; // #cwg171b-int
+namespace cwg171 {
+  extern "C" void cwg171a();
+  extern "C" void cwg171b();
+  // expected-error@-1 {{declaration of 'cwg171b' with C language linkage conflicts with declaration in global scope}}
+  //   expected-note@#cwg171b-int {{declared in global scope here}}
 }
 
-namespace dr172 { // dr172: yes
+namespace cwg172 { // cwg172: yes
   enum { zero };
   int check1[-1 < zero ? 1 : -1];
 
@@ -988,38 +988,38 @@ namespace dr172 { // dr172: yes
   int check8b[-f > 0 ? 1 : -1];
 }
 
-namespace dr173 { // dr173: yes
+namespace cwg173 { // cwg173: yes
   int check[('0' + 1 == '1' && '0' + 2 == '2' && '0' + 3 == '3' &&
              '0' + 4 == '4' && '0' + 5 == '5' && '0' + 6 == '6' &&
              '0' + 7 == '7' && '0' + 8 == '8' && '0' + 9 == '9') ? 1 : -1];
 }
 
-// dr174: sup 1012
+// cwg174: sup 1012
 
-namespace dr175 { // dr175: 2.8
-  struct A {}; // #dr175-A
-  struct B : private A {}; // #dr175-B
+namespace cwg175 { // cwg175: 2.8
+  struct A {}; // #cwg175-A
+  struct B : private A {}; // #cwg175-B
   struct C : B {
     A a;
-    // expected-error@-1 {{'A' is a private member of 'dr175::A'}}
-    //   expected-note@#dr175-B {{constrained by private inheritance here}}
-    //   expected-note@#dr175-A {{member is declared here}}
-    dr175::A b;
+    // expected-error@-1 {{'A' is a private member of 'cwg175::A'}}
+    //   expected-note@#cwg175-B {{constrained by private inheritance here}}
+    //   expected-note@#cwg175-A {{member is declared here}}
+    cwg175::A b;
   };
 }
 
-namespace dr176 { // dr176: 3.1
+namespace cwg176 { // cwg176: 3.1
   template<typename T> class Y;
   template<> class Y<int> {
     void f() {
-      typedef Y A; // #dr176-A-first
+      typedef Y A; // #cwg176-A-first
       typedef Y<char> A;
       // expected-error@-1 {{typedef redefinition with different types ('Y<char>' vs 'Y<int>')}}
-      //   expected-note@#dr176-A-first {{previous definition is here}}
+      //   expected-note@#cwg176-A-first {{previous definition is here}}
     }
   };
 
-  template<typename T> struct Base {}; // #dr176-Base
+  template<typename T> struct Base {}; // #cwg176-Base
   template<typename T> struct Derived : public Base<T> {
     void f() {
       typedef typename Derived::template Base<T> A;
@@ -1031,45 +1031,45 @@ namespace dr176 { // dr176: 3.1
   template<typename T> struct Derived2 : Base<int>, Base<char> {
     typename Derived2::Base b;
     // expected-error@-1 {{member 'Base' found in multiple base classes of different types}}
-    //   expected-note@#dr176-Base {{member type 'dr176::Base<int>' found by ambiguous name lookup}}
-    //   expected-note@#dr176-Base {{member type 'dr176::Base<char>' found by ambiguous name lookup}}
+    //   expected-note@#cwg176-Base {{member type 'cwg176::Base<int>' found by ambiguous name lookup}}
+    //   expected-note@#cwg176-Base {{member type 'cwg176::Base<char>' found by ambiguous name lookup}}
     typename Derived2::Base<double> d;
   };
 
-  template<typename T> class X { // #dr176-X
+  template<typename T> class X { // #cwg176-X
     X *p1;
     X<T> *p2;
     X<int> *p3;
-    dr176::X *p4; // #dr176-p4
-    // cxx98-14-error@#dr176-p4 {{use of class template 'dr176::X' requires template arguments}}
-    //  cxx98-14-note@#dr176-X {{template is declared here}}
-    // since-cxx17-error@#dr176-p4 {{use of class template 'X' requires template arguments; argument deduction not allowed in non-static class member}}
-    //  since-cxx17-note@#dr176-X {{template is declared here}}
+    cwg176::X *p4; // #cwg176-p4
+    // cxx98-14-error@#cwg176-p4 {{use of class template 'cwg176::X' requires template arguments}}
+    //  cxx98-14-note@#cwg176-X {{template is declared here}}
+    // since-cxx17-error@#cwg176-p4 {{use of class template 'X' requires template arguments; argument deduction not allowed in non-static class member}}
+    //  since-cxx17-note@#cwg176-X {{template is declared here}}
   };
 }
 
-namespace dr177 { // dr177: yes
+namespace cwg177 { // cwg177: yes
   struct B {};
   struct A {
-    A(A &); // #dr177-A-copy-ctor
-    A(const B &); // #dr177-A-ctor-from-B
+    A(A &); // #cwg177-A-copy-ctor
+    A(const B &); // #cwg177-A-ctor-from-B
   };
   B b;
   A a = b;
   // cxx98-14-error@-1 {{no viable constructor copying variable of type 'A'}}
-  //   cxx98-14-note@#dr177-A-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
-  //   cxx98-14-note@#dr177-A-ctor-from-B {{candidate constructor not viable: no known conversion from 'A' to 'const B &' for 1st argument}}
+  //   cxx98-14-note@#cwg177-A-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
+  //   cxx98-14-note@#cwg177-A-ctor-from-B {{candidate constructor not viable: no known conversion from 'A' to 'const B &' for 1st argument}}
 
-  struct C { C(C&); }; // #dr177-C-copy-ctor
+  struct C { C(C&); }; // #cwg177-C-copy-ctor
   struct D : C {};
   struct E { operator D(); };
   E e;
   C c = e;
   // expected-error@-1 {{no viable constructor copying variable of type 'D'}}
-  //   expected-note@#dr177-C-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
+  //   expected-note@#cwg177-C-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
 }
 
-namespace dr178 { // dr178: yes
+namespace cwg178 { // cwg178: yes
   int check[int() == 0 ? 1 : -1];
 #if __cplusplus >= 201103L
   static_assert(int{} == 0, "");
@@ -1082,13 +1082,13 @@ namespace dr178 { // dr178: yes
 #endif
 }
 
-namespace dr179 { // dr179: yes
+namespace cwg179 { // cwg179: yes
   void f();
   int n = &f - &f;
   // expected-error@-1 {{arithmetic on pointers to the function type 'void ()'}}
 }
 
-namespace dr180 { // dr180: 2.8
+namespace cwg180 { // cwg180: 2.8
   template<typename T> struct X : T, T::some_base {
     X() : T::some_type_that_might_be_T(), T::some_base() {}
     friend class T::some_class;
@@ -1098,7 +1098,7 @@ namespace dr180 { // dr180: 2.8
   };
 }
 
-namespace dr181 { // dr181: yes
+namespace cwg181 { // cwg181: yes
   namespace X {
     template <template X<class T> > struct A { };
     // expected-error@-1 +{{}}
@@ -1112,7 +1112,7 @@ namespace dr181 { // dr181: yes
   }
 }
 
-namespace dr182 { // dr182: 14
+namespace cwg182 { // cwg182: 14
   template <class T> struct C {
     void f();
     void g();
@@ -1135,7 +1135,7 @@ namespace dr182 { // dr182: 14
   }
 }
 
-namespace dr183 { // dr183: sup 382
+namespace cwg183 { // cwg183: sup 382
   template<typename T> struct A {};
   template<typename T> struct B {
     typedef int X;
@@ -1146,7 +1146,7 @@ namespace dr183 { // dr183: sup 382
   };
 }
 
-namespace dr184 { // dr184: yes
+namespace cwg184 { // cwg184: yes
   template<typename T = float> struct B {};
 
   template<template<typename TT = float> class T> struct A {
@@ -1154,10 +1154,10 @@ namespace dr184 { // dr184: yes
     void g();
   };
 
-  template<template<typename TT> class T> void A<T>::f() { // #dr184-T
+  template<template<typename TT> class T> void A<T>::f() { // #cwg184-T
     T<> t;
     // expected-error@-1 {{too few template arguments for template template parameter 'T'}}
-    //   expected-note@#dr184-T {{template is declared here}}
+    //   expected-note@#cwg184-T {{template is declared here}}
   }
 
   template<template<typename TT = char> class T> void A<T>::g() {
@@ -1169,25 +1169,25 @@ namespace dr184 { // dr184: yes
   void h() { A<B>().g(); }
 }
 
-// dr185 is in dr185.cpp
+// cwg185 is in cwg185.cpp
 
-namespace dr187 { // dr187: sup 481
+namespace cwg187 { // cwg187: sup 481
   const int Z = 1;
   template<int X = Z, int Z = X> struct A;
   typedef A<> T;
   typedef A<1, 1> T;
 }
 
-namespace dr188 { // dr188: yes
+namespace cwg188 { // cwg188: yes
   char c[10];
   int check[sizeof(0, c) == 10 ? 1 : -1];
 }
 
-// dr190 FIXME: add codegen test for tbaa
+// cwg190 FIXME: add codegen test for tbaa
 //              or implement C++20 std::is_layout_compatible and test it this way
 
-int dr191_j;
-namespace dr191 { // dr191: yes
+int cwg191_j;
+namespace cwg191 { // cwg191: yes
   namespace example1 {
     struct outer {
       static int i;
@@ -1208,7 +1208,7 @@ namespace dr191 { // dr191: yes
       void f() {
         struct local2 {
           void g() {
-            dr191_j = 5;
+            cwg191_j = 5;
           }
         };
       }
@@ -1216,9 +1216,9 @@ namespace dr191 { // dr191: yes
   }
 }
 
-// dr193 is in dr193.cpp
+// cwg193 is in cwg193.cpp
 
-namespace dr194 { // dr194: yes
+namespace cwg194 { // cwg194: yes
   struct A {
     A();
     void A();
@@ -1234,7 +1234,7 @@ namespace dr194 { // dr194: yes
   };
 }
 
-namespace dr195 { // dr195: yes
+namespace cwg195 { // cwg195: yes
   void f();
   int *p = (int*)&f;
   // cxx98-error@-1 {{cast between pointer-to-function and pointer-to-object is an extension}}
@@ -1242,14 +1242,14 @@ namespace dr195 { // dr195: yes
   // cxx98-error@-1 {{cast between pointer-to-function and pointer-to-object is an extension}}
 }
 
-namespace dr197 { // dr197: yes
+namespace cwg197 { // cwg197: yes
   char &f(char);
 
   template <class T> void g(T t) {
     char &a = f(1);
     char &b = f(T(1));
     // expected-error@-1 {{non-const lvalue reference to type 'char' cannot bind to a value of unrelated type 'int'}}
-    //   expected-note@#dr197-g-e-call {{in instantiation of function template specialization 'dr197::g<dr197::E>' requested here}}
+    //   expected-note@#cwg197-g-e-call {{in instantiation of function template specialization 'cwg197::g<cwg197::E>' requested here}}
     char &c = f(t);
     // expected-error@-1 {{non-const lvalue reference to type 'char' cannot bind to a value of unrelated type 'int'}}
   }
@@ -1262,11 +1262,11 @@ namespace dr197 { // dr197: yes
   void h() {
     g('a');
     g(2);
-    g(e); // #dr197-g-e-call
+    g(e); // #cwg197-g-e-call
   }
 }
 
-namespace dr198 { // dr198: yes
+namespace cwg198 { // cwg198: yes
   struct A {
     int n;
     struct B {
@@ -1291,4 +1291,4 @@ namespace dr198 { // dr198: yes
   };
 }
 
-// dr199 is in dr199.cpp
+// cwg199 is in cwg199.cpp
diff --git a/clang/test/CXX/drs/dr20xx.cpp b/clang/test/CXX/drs/dr20xx.cpp
index 96f2572cfb524..291a77e0cc71d 100644
--- a/clang/test/CXX/drs/dr20xx.cpp
+++ b/clang/test/CXX/drs/dr20xx.cpp
@@ -11,7 +11,7 @@
 // cxx98-error@-1 {{variadic macros are a C99 feature}}
 #endif
 
-namespace dr2007 { // dr2007: 3.4
+namespace cwg2007 { // cwg2007: 3.4
 template<typename T> struct A { typename T::error e; };
 template<typename T> struct B { };
 B<A<void> > b1;
@@ -20,24 +20,24 @@ int a = b2[0];
 // cxx98-error@-1 {{type 'B<A<void> >' does not provide a subscript operator}}
 // since-cxx11-error@-2 {{type 'B<A<void>>' does not provide a subscript operator}}
 int b = __builtin_addressof(b2)->foo;
-// cxx98-error@-1 {{no member named 'foo' in 'dr2007::B<dr2007::A<void> >'}}
-// since-cxx11-error@-2 {{no member named 'foo' in 'dr2007::B<dr2007::A<void>>'}}
+// cxx98-error@-1 {{no member named 'foo' in 'cwg2007::B<cwg2007::A<void> >'}}
+// since-cxx11-error@-2 {{no member named 'foo' in 'cwg2007::B<cwg2007::A<void>>'}}
 }
 
-// dr2009: na
+// cwg2009: na
 
-namespace dr2026 { // dr2026: 11
+namespace cwg2026 { // cwg2026: 11
   template<int> struct X {};
 
-  const int a = a + 1; // #dr2026-a
+  const int a = a + 1; // #cwg2026-a
   // expected-warning@-1 {{variable 'a' is uninitialized when used within its own initialization}}
-  X<a> xa; // #dr2026-xa
+  X<a> xa; // #cwg2026-xa
   // cxx98-error@-1 {{non-type template argument of type 'int' is not an integral constant expression}}
   //   cxx98-note@-2 {{initializer of 'a' is not a constant expression}}
-  //   cxx98-note@#dr2026-a {{declared here}}
-  // since-cxx11-error@#dr2026-xa {{non-type template argument is not a constant expression}}
-  //   since-cxx11-note@#dr2026-xa {{initializer of 'a' is not a constant expression}}
-  //   since-cxx11-note@#dr2026-a {{declared here}}
+  //   cxx98-note@#cwg2026-a {{declared here}}
+  // since-cxx11-error@#cwg2026-xa {{non-type template argument is not a constant expression}}
+  //   since-cxx11-note@#cwg2026-xa {{initializer of 'a' is not a constant expression}}
+  //   since-cxx11-note@#cwg2026-a {{declared here}}
 
 #if __cplusplus >= 201103L
   constexpr int b = b;
@@ -59,15 +59,15 @@ namespace dr2026 { // dr2026: 11
 #endif
 
   void f() {
-    static const int e = e + 1; // #dr2026-e
+    static const int e = e + 1; // #cwg2026-e
     // expected-warning@-1 {{static variable 'e' is suspiciously used within its own initialization}}
-    X<e> xe; // #dr2026-xe
+    X<e> xe; // #cwg2026-xe
     // cxx98-error@-1 {{non-type template argument of type 'int' is not an integral constant expression}}
     //   cxx98-note@-2 {{initializer of 'e' is not a constant expression}}
-    //   cxx98-note@#dr2026-e {{declared here}}
-    // since-cxx11-error@#dr2026-xe {{non-type template argument is not a constant expression}}
-    //   since-cxx11-note@#dr2026-xe {{initializer of 'e' is not a constant expression}}
-    //   since-cxx11-note@#dr2026-e {{declared here}}
+    //   cxx98-note@#cwg2026-e {{declared here}}
+    // since-cxx11-error@#cwg2026-xe {{non-type template argument is not a constant expression}}
+    //   since-cxx11-note@#cwg2026-xe {{initializer of 'e' is not a constant expression}}
+    //   since-cxx11-note@#cwg2026-e {{declared here}}
 
 #if __cplusplus >= 201103L
     static constexpr int f = f;
@@ -90,7 +90,7 @@ namespace dr2026 { // dr2026: 11
   }
 }
 
-namespace dr2049 { // dr2049: 18 drafting P2308R1
+namespace cwg2049 { // cwg2049: 18 drafting P2308R1
 #if __cplusplus >= 202302L
 template <int* x = {}> struct X {};
 X<> a;
@@ -99,7 +99,7 @@ static_assert(__is_same(decltype(a), decltype(b)));
 #endif
 }
 
-namespace dr2061 { // dr2061: yes
+namespace cwg2061 { // cwg2061: yes
 #if __cplusplus >= 201103L
   namespace A {
     inline namespace b {
@@ -127,7 +127,7 @@ namespace dr2061 { // dr2061: yes
 #endif // C++11
 }
 
-namespace dr2076 { // dr2076: 13
+namespace cwg2076 { // cwg2076: 13
 #if __cplusplus >= 201103L
   namespace std_example {
     struct A { A(int); };
@@ -149,8 +149,8 @@ namespace dr2076 { // dr2076: 13
     operator string_view() const;
   };
 
-  void foo(const string &); // #dr2076-foo 
-  void bar(string_view); // #dr2076-bar
+  void foo(const string &); // #cwg2076-foo 
+  void bar(string_view); // #cwg2076-bar
 
   void func(const string &arg) {
     // An argument in one set of braces is subject to user-defined conversions;
@@ -161,27 +161,27 @@ namespace dr2076 { // dr2076: 13
     foo({{arg}});
     foo({{{arg}}});
     // since-cxx11-error@-1 {{no matching function}}
-    //   since-cxx11-note@#dr2076-foo  {{cannot convert initializer list}}
+    //   since-cxx11-note@#cwg2076-foo  {{cannot convert initializer list}}
     bar(arg);
     bar({arg});
     bar({{arg}});
     // since-cxx11-error@-1 {{no matching function}}
-    //   since-cxx11-note@#dr2076-bar {{cannot convert initializer list}}
+    //   since-cxx11-note@#cwg2076-bar {{cannot convert initializer list}}
     bar({{{arg}}});
     // since-cxx11-error@-1 {{no matching function}}
-    //   since-cxx11-note@#dr2076-bar {{cannot convert initializer list}}
+    //   since-cxx11-note@#cwg2076-bar {{cannot convert initializer list}}
   }
 #endif
 }
 
-namespace dr2082 { // dr2082: 11
+namespace cwg2082 { // cwg2082: 11
   void test1(int x, int = sizeof(x)); // ok
 #if __cplusplus >= 201103L
   void test2(int x, int = decltype(x){}); // ok
 #endif
 }
 
-namespace dr2083 { // dr2083: partial
+namespace cwg2083 { // cwg2083: partial
 #if __cplusplus >= 201103L
   void non_const_mem_ptr() {
     struct A {
@@ -204,23 +204,23 @@ namespace dr2083 { // dr2083: partial
   const int a = 1;
   int b;
   // Note, references only get special odr-use / constant initializxer
-  // treatment in C++11 onwards. We continue to apply that even after DR2083.
+  // treatment in C++11 onwards. We continue to apply that even after CWG2083.
   void ref_to_non_const() {
     int c;
-    const int &ra = a; // #dr2083-ra
-    int &rb = b; // #dr2083-rb
-    int &rc = c; // #dr2083-rc
+    const int &ra = a; // #cwg2083-ra
+    int &rb = b; // #cwg2083-rb
+    int &rc = c; // #cwg2083-rc
     struct A {
       int f() {
         int a = ra;
-        // cxx98-error@-1 {{reference to local variable 'ra' declared in enclosing function 'dr2083::ref_to_non_const'}}
-        //   cxx98-note@#dr2083-ra {{'ra' declared here}}
+        // cxx98-error@-1 {{reference to local variable 'ra' declared in enclosing function 'cwg2083::ref_to_non_const'}}
+        //   cxx98-note@#cwg2083-ra {{'ra' declared here}}
         int b = rb;
-        // cxx98-error@-1 {{reference to local variable 'rb' declared in enclosing function 'dr2083::ref_to_non_const'}}
-        //   cxx98-note@#dr2083-rb {{'rb' declared here}}
+        // cxx98-error@-1 {{reference to local variable 'rb' declared in enclosing function 'cwg2083::ref_to_non_const'}}
+        //   cxx98-note@#cwg2083-rb {{'rb' declared here}}
         int c = rc;
-        // expected-error@-1 {{reference to local variable 'rc' declared in enclosing function 'dr2083::ref_to_non_const'}}
-        //   expected-note@#dr2083-rc {{'rc' declared here}}
+        // expected-error@-1 {{reference to local variable 'rc' declared in enclosing function 'cwg2083::ref_to_non_const'}}
+        //   expected-note@#cwg2083-rc {{'rc' declared here}}
         return a + b + c;
       }
     };
@@ -244,9 +244,9 @@ namespace dr2083 { // dr2083: partial
     constexpr NoMut1 nm1 = {1, 2};
     constexpr NoMut2 nm2 = {1, 2};
     constexpr NoMut3 nm3 = {1, 2};
-    constexpr Mut1 m1 = {1, 2}; // #dr2083-m1
-    constexpr Mut2 m2 = {1, 2}; // #dr2083-m2
-    constexpr Mut3 m3 = {1, 2}; // #dr2083-m3
+    constexpr Mut1 m1 = {1, 2}; // #cwg2083-m1
+    constexpr Mut2 m2 = {1, 2}; // #cwg2083-m2
+    constexpr Mut3 m3 = {1, 2}; // #cwg2083-m3
     struct A {
       void f() {
         static_assert(nm1.a == 1, "");
@@ -254,14 +254,14 @@ namespace dr2083 { // dr2083: partial
         static_assert(nm3.a == 1, "");
         // Can't even access a non-mutable member of a variable containing mutable fields.
         static_assert(m1.a == 1, "");
-        // since-cxx11-error@-1 {{reference to local variable 'm1' declared in enclosing function 'dr2083::mutable_subobjects'}}
-        //   since-cxx11-note@#dr2083-m1 {{'m1' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'm1' declared in enclosing function 'cwg2083::mutable_subobjects'}}
+        //   since-cxx11-note@#cwg2083-m1 {{'m1' declared here}}
         static_assert(m2.m.a == 1, "");
-        // since-cxx11-error@-1 {{reference to local variable 'm2' declared in enclosing function 'dr2083::mutable_subobjects'}}
-        //   since-cxx11-note@#dr2083-m2 {{'m2' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'm2' declared in enclosing function 'cwg2083::mutable_subobjects'}}
+        //   since-cxx11-note@#cwg2083-m2 {{'m2' declared here}}
         static_assert(m3.a == 1, "");
-        // since-cxx11-error@-1 {{reference to local variable 'm3' declared in enclosing function 'dr2083::mutable_subobjects'}}
-        //   since-cxx11-note@#dr2083-m3 {{'m3' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'm3' declared in enclosing function 'cwg2083::mutable_subobjects'}}
+        //   since-cxx11-note@#cwg2083-m3 {{'m3' declared here}}
       }
     };
   }
@@ -274,7 +274,7 @@ namespace dr2083 { // dr2083: partial
 #if __cplusplus >= 201103L
     constexpr
 #endif
-      A a = {}; // #dr2083-a
+      A a = {}; // #cwg2083-a
     struct B {
       void f() {
         ellipsis(n);
@@ -282,8 +282,8 @@ namespace dr2083 { // dr2083: partial
         // conversion, it calls a constructor and binds 'a' to a reference, so
         // it results in an odr-use.
         ellipsis(a);
-        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr2083::ellipsis'}}
-        //   expected-note@#dr2083-a {{'a' declared here}}
+        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg2083::ellipsis'}}
+        //   expected-note@#cwg2083-a {{'a' declared here}}
       }
     };
   }
@@ -291,7 +291,7 @@ namespace dr2083 { // dr2083: partial
 #if __cplusplus >= 201103L
   void volatile_lval() {
     struct A { int n; };
-    constexpr A a = {0}; // #dr2083-a2
+    constexpr A a = {0}; // #cwg2083-a2
     struct B {
       void f() {
         // An lvalue-to-rvalue conversion of a volatile lvalue always results
@@ -300,8 +300,8 @@ namespace dr2083 { // dr2083: partial
         int x = a.*p;
         volatile int A::*q = p;
         int y = a.*q;
-        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr2083::volatile_lval'}}
-        //   since-cxx11-note@#dr2083-a2 {{'a' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg2083::volatile_lval'}}
+        //   since-cxx11-note@#cwg2083-a2 {{'a' declared here}}
       }
     };
   }
@@ -309,45 +309,45 @@ namespace dr2083 { // dr2083: partial
 
   void discarded_lval() {
     struct A { int x; mutable int y; volatile int z; };
-    A a; // #dr2083-a-3
-    int &r = a.x; // #dr2083-r
+    A a; // #cwg2083-a-3
+    int &r = a.x; // #cwg2083-r
     struct B {
       void f() {
         // FIXME: We emit more errors than we should be. They are explictly marked below.
         a.x;
         // expected-warning@-1 {{expression result unused}}
-        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
+        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
         a.*&A::x;
         // expected-warning@-1 {{expression result unused}}
-        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
-        true ? a.x : a.y; // #dr2083-ternary
+        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
+        true ? a.x : a.y; // #cwg2083-ternary
         // expected-warning@-1 {{expression result unused}}
-        // expected-error@#dr2083-ternary {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
-        // expected-error@#dr2083-ternary {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
+        // expected-error@#cwg2083-ternary {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
+        // expected-error@#cwg2083-ternary {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
         (void)a.x;
-        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
+        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
         a.x, discarded_lval();
         // expected-warning@-1 {{left operand of comma operator has no effect}}
-        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}} FIXME
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
+        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}} FIXME
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
 
         // 'volatile' qualifier triggers an lvalue-to-rvalue conversion.
         a.z;
         // cxx98-warning@-1 {{expression result unused; assign into a variable to force a volatile load}}
-        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'dr2083::discarded_lval'}}
-        //   expected-note@#dr2083-a-3 {{'a' declared here}}
+        // expected-error@-2 {{reference to local variable 'a' declared in enclosing function 'cwg2083::discarded_lval'}}
+        //   expected-note@#cwg2083-a-3 {{'a' declared here}}
 
         // References always get "loaded" to determine what they reference,
         // even if the result is discarded.
         r;
         // expected-warning@-1 {{expression result unused}}
-        // expected-error@-2 {{reference to local variable 'r' declared in enclosing function 'dr2083::discarded_lval'}}
-        //   expected-note@#dr2083-r {{'r' declared here}}
+        // expected-error@-2 {{reference to local variable 'r' declared in enclosing function 'cwg2083::discarded_lval'}}
+        //   expected-note@#cwg2083-r {{'r' declared here}}
       }
     };
   }
@@ -355,11 +355,11 @@ namespace dr2083 { // dr2083: partial
   namespace dr_example_1 {
     extern int globx;
     int main() {
-      const int &x = globx; // #dr2083-x
+      const int &x = globx; // #cwg2083-x
       struct A {
         const int *foo() { return &x; }
-        // cxx98-error@-1 {{reference to local variable 'x' declared in enclosing function 'dr2083::dr_example_1::main'}}
-        //   cxx98-note@#dr2083-x {{'x' declared here}}
+        // cxx98-error@-1 {{reference to local variable 'x' declared in enclosing function 'cwg2083::dr_example_1::main'}}
+        //   cxx98-note@#cwg2083-x {{'x' declared here}}
       } a;
       return *a.foo();
     }
@@ -400,7 +400,7 @@ namespace dr2083 { // dr2083: partial
 #endif
 }
 
-namespace dr2094 { // dr2094: 5
+namespace cwg2094 { // cwg2094: 5
   struct A { int n; };
   struct B { volatile int n; };
   static_assert(__is_trivially_copyable(volatile int), "");
@@ -419,4 +419,4 @@ namespace dr2094 { // dr2094: 5
   static_assert(__is_trivially_assignable(B, const B&), "");
 }
 
-// dr2096: dup 2598
+// cwg2096: dup 2598
diff --git a/clang/test/CXX/drs/dr21xx.cpp b/clang/test/CXX/drs/dr21xx.cpp
index a7e50df3f374b..4fab10c279aa4 100644
--- a/clang/test/CXX/drs/dr21xx.cpp
+++ b/clang/test/CXX/drs/dr21xx.cpp
@@ -11,7 +11,7 @@
 // cxx98-error@-1 {{variadic macros are a C99 feature}}
 #endif
 
-namespace dr2100 { // dr2100: 12
+namespace cwg2100 { // cwg2100: 12
   template<const int *P, bool = true> struct X {};
   template<typename T> struct A {
     static const int n = 1;
@@ -19,10 +19,10 @@ namespace dr2100 { // dr2100: 12
       return X<&n>::n; // ok, value-dependent
     }
     int g() {
-      static const int n = 2; // #dr2100-n
+      static const int n = 2; // #cwg2100-n
       return X<&n>::n; // ok, value-dependent
       // cxx98-14-error@-1 {{non-type template argument refers to object 'n' that does not have linkage}}
-      //   cxx98-14-note@#dr2100-n {{non-type template argument refers to object here}}
+      //   cxx98-14-note@#cwg2100-n {{non-type template argument refers to object here}}
     }
   };
   template<const int *P> struct X<P> {
@@ -41,12 +41,12 @@ namespace dr2100 { // dr2100: 12
     static const int n = 1;
     int f() {
       return Y<n>::declared_later;
-      // expected-error@-1 {{no member named 'declared_later' in 'dr2100::Y<1>'}}
+      // expected-error@-1 {{no member named 'declared_later' in 'cwg2100::Y<1>'}}
     }
     int g() {
       static const int n = 2;
       return Y<n>::declared_later;
-      // expected-error@-1 {{no member named 'declared_later' in 'dr2100::Y<2>'}}
+      // expected-error@-1 {{no member named 'declared_later' in 'cwg2100::Y<2>'}}
     }
   };
   template<int N> struct Y<N> {
@@ -54,22 +54,22 @@ namespace dr2100 { // dr2100: 12
   };
 }
 
-namespace dr2103 { // dr2103: yes
+namespace cwg2103 { // cwg2103: yes
   void f() {
     int a;
-    int &r = a; // #dr2103-r
+    int &r = a; // #cwg2103-r
     struct Inner {
       void f() {
         int &s = r;
-        // expected-error@-1 {{reference to local variable 'r' declared in enclosing function 'dr2103::f'}}
-        //   expected-note@#dr2103-r {{'r' declared here}}
+        // expected-error@-1 {{reference to local variable 'r' declared in enclosing function 'cwg2103::f'}}
+        //   expected-note@#cwg2103-r {{'r' declared here}}
         (void)s;
       }
     };
   }
 }
 
-namespace dr2120 { // dr2120: 7
+namespace cwg2120 { // cwg2120: 7
   struct A {};
   struct B : A {};
   struct C { A a; };
@@ -80,7 +80,7 @@ namespace dr2120 { // dr2120: 7
   static_assert(!__is_standard_layout(E), "");
 }
 
-namespace dr2126 { // dr2126: 12
+namespace cwg2126 { // cwg2126: 12
 #if __cplusplus >= 201103L
   struct A { int n; };
 
@@ -88,16 +88,16 @@ namespace dr2126 { // dr2126: 12
   A &b = (A &)(const A &)A{1};   // const temporary
   A &&c = (A &&)(const A &)A{1}; // const temporary
 
-  A &&d = {1};                   // non-const temporary #dr21260-d
-  const A &e = (A &)(A &&) A{1}; // non-const temporary #dr21260-e
-  A &&f = (A &&)(A &&) A{1};     // non-const temporary #dr21260-f
+  A &&d = {1};                   // non-const temporary #cwg21260-d
+  const A &e = (A &)(A &&) A{1}; // non-const temporary #cwg21260-e
+  A &&f = (A &&)(A &&) A{1};     // non-const temporary #cwg21260-f
 
   constexpr const A &g = {1};    // const temporary
-  constexpr A &&h = {1};         // non-const temporary #dr21260-h
+  constexpr A &&h = {1};         // non-const temporary #cwg21260-h
 
   struct B { const A &a; };
-  B i = {{1}};           // extending decl not usable in constant expr #dr21260-i
-  const B j = {{1}};     // extending decl not usable in constant expr #dr21260-j
+  B i = {{1}};           // extending decl not usable in constant expr #cwg21260-i
+  const B j = {{1}};     // extending decl not usable in constant expr #cwg21260-j
   constexpr B k = {{1}}; // extending decl usable in constant expr
 
   static_assert(a.n == 1, "");
@@ -106,33 +106,33 @@ namespace dr2126 { // dr2126: 12
   static_assert(d.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of temporary is not allowed in a constant expression outside the expression that created the temporary}}
-  //   since-cxx11-note@#dr21260-d {{temporary created here}}
+  //   since-cxx11-note@#cwg21260-d {{temporary created here}}
   static_assert(e.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of temporary is not allowed in a constant expression outside the expression that created the temporary}}
-  //   since-cxx11-note@#dr21260-e {{temporary created here}}
+  //   since-cxx11-note@#cwg21260-e {{temporary created here}}
   static_assert(f.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of temporary is not allowed in a constant expression outside the expression that created the temporary}}
-  //   since-cxx11-note@#dr21260-f {{temporary created here}}
+  //   since-cxx11-note@#cwg21260-f {{temporary created here}}
   static_assert(g.n == 1, "");
   static_assert(h.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of temporary is not allowed in a constant expression outside the expression that created the temporary}}
-  //   since-cxx11-note@#dr21260-h {{temporary created here}}
+  //   since-cxx11-note@#cwg21260-h {{temporary created here}}
   static_assert(i.a.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of non-constexpr variable 'i' is not allowed in a constant expression}}
-  //   since-cxx11-note@#dr21260-i {{declared here}}
+  //   since-cxx11-note@#cwg21260-i {{declared here}}
   static_assert(j.a.n == 1, "");
   // since-cxx11-error@-1 {{static assertion expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{read of temporary is not allowed in a constant expression outside the expression that created the temporary}}
-  //   since-cxx11-note@#dr21260-j {{temporary created here}}
+  //   since-cxx11-note@#cwg21260-j {{temporary created here}}
   static_assert(k.a.n == 1, "");
 #endif
 }
 
-namespace dr2140 { // dr2140: 9
+namespace cwg2140 { // cwg2140: 9
 #if __cplusplus >= 201103L
   union U { int a; decltype(nullptr) b; };
   constexpr int *test(U u) {
@@ -142,7 +142,7 @@ namespace dr2140 { // dr2140: 9
 #endif
 }
 
-namespace dr2141 { // dr2141: 17
+namespace cwg2141 { // cwg2141: 17
 struct A{};
 
 template <typename T>
@@ -175,34 +175,34 @@ void foo() {
 }
 }
 
-namespace dr2157 { // dr2157: 11
+namespace cwg2157 { // cwg2157: 11
 #if __cplusplus >= 201103L
   enum E : int;
   struct X {
-    enum dr2157::E : int();
+    enum cwg2157::E : int();
     // since-cxx11-error@-1 {{ISO C++ only allows ':' in member enumeration declaration to introduce a fixed underlying type, not an anonymous bit-field}}
   };
 #endif
 }
 
-// dr2165: na
+// cwg2165: na
 
-namespace dr2170 { // dr2170: 9
+namespace cwg2170 { // cwg2170: 9
 #if __cplusplus >= 201103L
   void f() {
-    constexpr int arr[3] = {1, 2, 3}; // #dr2170-arr
+    constexpr int arr[3] = {1, 2, 3}; // #cwg2170-arr
     struct S {
       int get(int n) { return arr[n]; }
       const int &get_ref(int n) { return arr[n]; }
       // since-cxx11-warning@-1 {{reference to stack memory associated with local variable 'arr' returned}} FIXME
-      // since-cxx11-error@-2 {{reference to local variable 'arr' declared in enclosing function 'dr2170::f'}}
-      //   since-cxx11-note@#dr2170-arr {{'arr' declared here}}
+      // since-cxx11-error@-2 {{reference to local variable 'arr' declared in enclosing function 'cwg2170::f'}}
+      //   since-cxx11-note@#cwg2170-arr {{'arr' declared here}}
     };
   }
 #endif
 }
 
-namespace dr2171 { // dr2171: 15
+namespace cwg2171 { // cwg2171: 15
 #if __cplusplus >= 201103L
 
 struct NonConstCopy {
@@ -227,40 +227,40 @@ static_assert(!__is_trivially_assignable(NonConstCopy &&, NonConstCopy), "");
 static_assert(!__is_trivially_assignable(NonConstCopy &&, NonConstCopy &&), "");
 
 #endif
-} // namespace dr2171
+} // namespace cwg2171
 
-namespace dr2180 { // dr2180: yes
+namespace cwg2180 { // cwg2180: yes
   class A {
-    A &operator=(const A &); // #dr2180-A-copy
-    A &operator=(A &&); // #dr2180-A-move
+    A &operator=(const A &); // #cwg2180-A-copy
+    A &operator=(A &&); // #cwg2180-A-move
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
   };
 
-  struct B : virtual A { // #dr2180-B
+  struct B : virtual A { // #cwg2180-B
     B &operator=(const B &);
     B &operator=(B &&);
     // cxx98-error@-1 {{rvalue references are a C++11 extension}}
     virtual void foo() = 0;
   };
-  B &B::operator=(const B&) = default; // #dr2180-B-copy
+  B &B::operator=(const B&) = default; // #cwg2180-B-copy
   // cxx98-error@-1 {{defaulted function definitions are a C++11 extension}}
-  // cxx98-error@-2 {{'operator=' is a private member of 'dr2180::A'}}
-  //   cxx98-note@-3 {{in defaulted copy assignment operator for 'dr2180::B' first required here}}
-  //   cxx98-note@#dr2180-A-copy {{implicitly declared private here}}
-  // since-cxx11-error@#dr2180-B-copy {{defaulting this copy assignment operator would delete it after its first declaration}}
-  //   since-cxx11-note@#dr2180-B {{copy assignment operator of 'B' is implicitly deleted because base class 'A' has an inaccessible copy assignment operator}}
-  B &B::operator=(B&&) = default; // #dr2180-B-move
+  // cxx98-error@-2 {{'operator=' is a private member of 'cwg2180::A'}}
+  //   cxx98-note@-3 {{in defaulted copy assignment operator for 'cwg2180::B' first required here}}
+  //   cxx98-note@#cwg2180-A-copy {{implicitly declared private here}}
+  // since-cxx11-error@#cwg2180-B-copy {{defaulting this copy assignment operator would delete it after its first declaration}}
+  //   since-cxx11-note@#cwg2180-B {{copy assignment operator of 'B' is implicitly deleted because base class 'A' has an inaccessible copy assignment operator}}
+  B &B::operator=(B&&) = default; // #cwg2180-B-move
   // cxx98-error@-1 {{rvalue references are a C++11 extension}}
   // cxx98-error@-2 {{defaulted function definitions are a C++11 extension}}
-  // cxx98-error@-3 {{'operator=' is a private member of 'dr2180::A'}}
-  //   cxx98-note@-4 {{in defaulted move assignment operator for 'dr2180::B' first required here}}
-  //   cxx98-note@#dr2180-A-move {{implicitly declared private here}}
-  // since-cxx11-error@#dr2180-B-move {{defaulting this move assignment operator would delete it after its first declaration}}
-  //   since-cxx11-note@#dr2180-B {{move assignment operator of 'B' is implicitly deleted because base class 'A' has an inaccessible move assignment operator}}
+  // cxx98-error@-3 {{'operator=' is a private member of 'cwg2180::A'}}
+  //   cxx98-note@-4 {{in defaulted move assignment operator for 'cwg2180::B' first required here}}
+  //   cxx98-note@#cwg2180-A-move {{implicitly declared private here}}
+  // since-cxx11-error@#cwg2180-B-move {{defaulting this move assignment operator would delete it after its first declaration}}
+  //   since-cxx11-note@#cwg2180-B {{move assignment operator of 'B' is implicitly deleted because base class 'A' has an inaccessible move assignment operator}}
 }
 
-namespace dr2199 { // dr2199: 3.8
-                   // NB: reusing part of dr407 test
+namespace cwg2199 { // cwg2199: 3.8
+                   // NB: reusing part of cwg407 test
 namespace A {
   struct S {};
 }
diff --git a/clang/test/CXX/drs/dr22xx.cpp b/clang/test/CXX/drs/dr22xx.cpp
index 3a13cb0471a75..797c3ed8546ef 100644
--- a/clang/test/CXX/drs/dr22xx.cpp
+++ b/clang/test/CXX/drs/dr22xx.cpp
@@ -8,7 +8,7 @@
 
 
 #if __cplusplus >= 201103L
-namespace dr2211 { // dr2211: 8
+namespace cwg2211 { // cwg2211: 8
 void f() {
   int a;
   auto f = [a](int a) { (void)a; };
@@ -19,15 +19,15 @@ void f() {
 }
 #endif
 
-namespace dr2213 { // dr2213: yes
+namespace cwg2213 { // cwg2213: yes
 template <typename T, typename U>
 struct A;
 
 template <typename U>
 struct A<int, U>;
-} // namespace dr2213
+} // namespace cwg2213
 
-namespace dr2229 { // dr2229: 7
+namespace cwg2229 { // cwg2229: 7
 struct AnonBitfieldQualifiers {
   const unsigned : 1;
   // expected-error@-1 {{anonymous bit-field cannot have qualifiers}}
@@ -43,7 +43,7 @@ struct AnonBitfieldQualifiers {
 };
 }
 
-namespace dr2233 { // dr2233: 11
+namespace cwg2233 { // cwg2233: 11
 #if __cplusplus >= 201103L
 template <typename... T>
 void f(int i = 0, T... args) {}
@@ -107,7 +107,7 @@ namespace MultilevelSpecialization {
   };
   template<> template<int a, int b>
     void B<int, int>::f(int i, int (&arr1)[a], int (&arr2)[b]) {}
-    // since-cxx11-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'dr2233::MultilevelSpecialization::B<int, int>'}}
+    // since-cxx11-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'cwg2233::MultilevelSpecialization::B<int, int>'}}
   template<> template<>
     void B<int, int>::f<1, 1>(int i, int (&arr1a)[1], int (&arr2a)[1]) {}
 }
@@ -130,12 +130,12 @@ namespace CheckAfterMerging2 {
   void h() { f<int>(); }
 }
 #endif
-} // namespace dr2233
+} // namespace cwg2233
 
-namespace dr2267 { // dr2267: no
+namespace cwg2267 { // cwg2267: no
 #if __cplusplus >= 201103L
 struct A {} a;
-struct B { explicit B(const A&); }; // #dr2267-struct-B
+struct B { explicit B(const A&); }; // #cwg2267-struct-B
 
 struct D { D(); };
 struct C { explicit operator D(); } c;
@@ -144,9 +144,9 @@ B b1(a);
 const B &b2{a}; // FIXME ill-formed
 const B &b3(a);
 // since-cxx11-error@-1 {{no viable conversion from 'struct A' to 'const B'}}
-//   since-cxx11-note@#dr2267-struct-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'struct A' to 'const B &' for 1st argument}}
-//   since-cxx11-note@#dr2267-struct-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'struct A' to 'B &&' for 1st argument}}
-//   since-cxx11-note@#dr2267-struct-B {{explicit constructor is not a candidate}}
+//   since-cxx11-note@#cwg2267-struct-B {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'struct A' to 'const B &' for 1st argument}}
+//   since-cxx11-note@#cwg2267-struct-B {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'struct A' to 'B &&' for 1st argument}}
+//   since-cxx11-note@#cwg2267-struct-B {{explicit constructor is not a candidate}}
 
 D d1(c);
 const D &d2{c}; // FIXME ill-formed
@@ -154,34 +154,34 @@ const D &d3(c); // FIXME ill-formed
 #endif
 }
 
-namespace dr2273 { // dr2273: 3.3
+namespace cwg2273 { // cwg2273: 3.3
 #if __cplusplus >= 201103L
 struct A {
-  A(int = 0) = delete; // #dr2273-A
+  A(int = 0) = delete; // #cwg2273-A
 };
 
-struct B : A { // #dr2273-B
+struct B : A { // #cwg2273-B
   using A::A;
 };
 
 B b;
 // since-cxx11-error@-1 {{call to implicitly-deleted default constructor of 'B'}}
-//   since-cxx11-note@#dr2273-B {{default constructor of 'B' is implicitly deleted because base class 'A' has a deleted default constructor}}
-//   since-cxx11-note@#dr2273-A {{'A' has been explicitly marked deleted here}}
+//   since-cxx11-note@#cwg2273-B {{default constructor of 'B' is implicitly deleted because base class 'A' has a deleted default constructor}}
+//   since-cxx11-note@#cwg2273-A {{'A' has been explicitly marked deleted here}}
 #endif
 }
 
-namespace dr2277 { // dr2277: partial
+namespace cwg2277 { // cwg2277: partial
 #if __cplusplus >= 201103L
 struct A {
   A(int, int = 0);
-  void f(int, int = 0); // #dr2277-A-f
+  void f(int, int = 0); // #cwg2277-A-f
 };
 struct B : A {
   B(int);
   using A::A;
 
-  void f(int); // #dr2277-B-f
+  void f(int); // #cwg2277-B-f
   using A::f;
 };
 
@@ -189,13 +189,13 @@ void g() {
   B b{0};
   b.f(0); // FIXME: this is well-formed for the same reason as initialization of 'b' above
   // since-cxx11-error@-1 {{call to member function 'f' is ambiguous}}
-  //   since-cxx11-note@#dr2277-A-f {{candidate function}}
-  //   since-cxx11-note@#dr2277-B-f {{candidate function}}
+  //   since-cxx11-note@#cwg2277-A-f {{candidate function}}
+  //   since-cxx11-note@#cwg2277-B-f {{candidate function}}
 }
 #endif
 }
 
-namespace dr2292 { // dr2292: 9
+namespace cwg2292 { // cwg2292: 9
 #if __cplusplus >= 201103L
   template<typename T> using id = T;
   void test(int *p) {
diff --git a/clang/test/CXX/drs/dr23xx.cpp b/clang/test/CXX/drs/dr23xx.cpp
index 812d95a39b638..db5b7c3cd3c9a 100644
--- a/clang/test/CXX/drs/dr23xx.cpp
+++ b/clang/test/CXX/drs/dr23xx.cpp
@@ -7,7 +7,7 @@
 // RUN: %clang_cc1 -std=c++2c %s -verify=expected,since-cxx11,since-cxx14,since-cxx17,since-cxx20 -fexceptions -fcxx-exceptions -pedantic-errors 2>&1 | FileCheck %s
 
 #if __cplusplus >= 201103L
-namespace dr2303 { // dr2303: 12
+namespace cwg2303 { // cwg2303: 12
 template <typename... T>
 struct A;
 template <>
@@ -17,12 +17,12 @@ struct A<T, Ts...> : A<Ts...> {};
 struct B : A<int, int> {};
 struct C : A<int, int>, A<int> {};
 /* since-cxx11-warning@-1 {{direct base 'A<int>' is inaccessible due to ambiguity:
-    struct dr2303::C -> A<int, int> -> A<int>
-    struct dr2303::C -> A<int>}} */
+    struct cwg2303::C -> A<int, int> -> A<int>
+    struct cwg2303::C -> A<int>}} */
 struct D : A<int>, A<int, int> {};
 /* since-cxx11-warning@-1 {{direct base 'A<int>' is inaccessible due to ambiguity:
-    struct dr2303::D -> A<int>
-    struct dr2303::D -> A<int, int> -> A<int>}} */
+    struct cwg2303::D -> A<int>
+    struct cwg2303::D -> A<int, int> -> A<int>}} */
 struct E : A<int, int> {};
 struct F : B, E {};
 
@@ -41,17 +41,17 @@ void g() {
   f(D{});
   f(F{});
   /* since-cxx11-error@-1 {{ambiguous conversion from derived class 'const F' to base class 'const A<int, int>':
-    struct dr2303::F -> B -> A<int, int>
-    struct dr2303::F -> E -> A<int, int>}} */
+    struct cwg2303::F -> B -> A<int, int>
+    struct cwg2303::F -> E -> A<int, int>}} */
 }
-} // namespace dr2303
+} // namespace cwg2303
 #endif
 
-// dr2331: na
-// dr2335 is in dr2335.cxx
+// cwg2331: na
+// cwg2335 is in cwg2335.cxx
 
 #if __cplusplus >= 201103L
-namespace dr2338 { // dr2338: 12
+namespace cwg2338 { // cwg2338: 12
 namespace B {
 enum E : bool { Zero, One };
 static_assert((int)(E)2 == 1, "");
@@ -60,17 +60,17 @@ namespace D {
 enum class E : bool { Zero, One };
 static_assert((int)(E)2 == 1, "");
 } // namespace D
-} // namespace dr2338
+} // namespace cwg2338
 #endif
 
-namespace dr2346 { // dr2346: 11
+namespace cwg2346 { // cwg2346: 11
   void test() {
     const int i2 = 0;
     extern void h2b(int x = i2 + 0); // ok, not odr-use
   }
 }
 
-namespace dr2352 { // dr2352: 10
+namespace cwg2352 { // cwg2352: 10
   int **p;
   const int *const *const &f1() { return p; }
   int *const *const &f2() { return p; }
@@ -106,7 +106,7 @@ namespace dr2352 { // dr2352: 10
 #endif
 }
 
-namespace dr2353 { // dr2353: 9
+namespace cwg2353 { // cwg2353: 9
   struct X {
     static const int n = 0;
   };
@@ -138,16 +138,16 @@ namespace dr2353 { // dr2353: 9
 #pragma clang __debug dump not_use_2
 }
 
-namespace dr2354 { // dr2354: 15
+namespace cwg2354 { // cwg2354: 15
 #if __cplusplus >= 201103L
 enum alignas(64) A {};
 // since-cxx11-error@-1 {{'alignas' attribute cannot be applied to an enumeration}}
 enum struct alignas(64) B {};
 // since-cxx11-error@-1 {{'alignas' attribute cannot be applied to an enumeration}}
 #endif
-} // namespace dr2354
+} // namespace cwg2354
 
-namespace dr2356 { // dr2356: 4
+namespace cwg2356 { // cwg2356: 4
 #if __cplusplus >= 201103L
 struct A {
   A();
@@ -159,11 +159,11 @@ struct B : A {
   B(const B &);                   // #3
   B(B &&) = default;              // #4, implicitly deleted
   // since-cxx11-warning@-1 {{explicitly defaulted move constructor is implicitly deleted}}
-  //   since-cxx11-note@#dr2356-X {{move constructor of 'B' is implicitly deleted because field 'x' has a deleted move constructor}}
-  //   since-cxx11-note@#dr2356-X {{'X' has been explicitly marked deleted here}}
+  //   since-cxx11-note@#cwg2356-X {{move constructor of 'B' is implicitly deleted because field 'x' has a deleted move constructor}}
+  //   since-cxx11-note@#cwg2356-X {{'X' has been explicitly marked deleted here}}
   //   since-cxx11-note@-4 {{replace 'default' with 'delete'}}
 
-  struct X { X(X &&) = delete; } x; // #dr2356-X
+  struct X { X(X &&) = delete; } x; // #cwg2356-X
 };
 extern B b1;
 B b2 = static_cast<B&&>(b1);      // calls #3: #1, #2, and #4 are not viable
@@ -173,7 +173,7 @@ B b3 = C();                       // calls #3
 }
 
 #if __cplusplus >= 201402L
-namespace dr2358 { // dr2358: 16
+namespace cwg2358 { // cwg2358: 16
   void f2() {
     int i = 1;
     void g1(int = [xxx=1] { return xxx; }());  // OK
@@ -185,7 +185,7 @@ namespace dr2358 { // dr2358: 16
 
 // CWG2363 was closed as NAD, but its resolution does affirm that
 // a friend declaration cannot have an opaque-enumm-specifier.
-namespace dr2363 { // dr2363: yes
+namespace cwg2363 { // cwg2363: yes
 #if __cplusplus >= 201103L
 enum class E0;
 enum E1 : int;
@@ -215,9 +215,9 @@ struct A {
   // expected-note@-3 {{remove 'enum class' to befriend an enum}}
 };
 #endif
-} // namespace dr2363
+} // namespace cwg2363
 
-namespace dr2370 { // dr2370: no
+namespace cwg2370 { // cwg2370: no
 namespace N {
 typedef int type;
 void g(type);
@@ -229,33 +229,33 @@ class C {
   // friend void N::g(type);
   friend void N::h(N_type);
 };
-} // namespace dr2370
+} // namespace cwg2370
 
 #if __cplusplus >= 201702L
 // Otherwise, if the qualified-id std::tuple_size<E> names a complete class
 // type **with a member value**, the expression std::tuple_size<E>::value shall
 // be a well-formed integral constant expression
-namespace dr2386 { // dr2386: 9
+namespace cwg2386 { // cwg2386: 9
 struct Bad1 { int a, b; };
 struct Bad2 { int a, b; };
-} // namespace dr2386
+} // namespace cwg2386
 namespace std {
 template <typename T> struct tuple_size;
-template <> struct tuple_size<dr2386::Bad1> {};
-template <> struct tuple_size<dr2386::Bad2> {
+template <> struct tuple_size<cwg2386::Bad1> {};
+template <> struct tuple_size<cwg2386::Bad2> {
   static const int value = 42;
 };
 } // namespace std
-namespace dr2386 {
+namespace cwg2386 {
 void no_value() { auto [x, y] = Bad1(); }
 void wrong_value() { auto [x, y] = Bad2(); }
 // since-cxx17-error@-1 {{type 'Bad2' decomposes into 42 elements, but only 2 names were provided}}
-} // namespace dr2386
+} // namespace cwg2386
 #endif
 
-// dr2385: na
+// cwg2385: na
 
-namespace dr2387 { // dr2387: 9
+namespace cwg2387 { // cwg2387: 9
 #if __cplusplus >= 201402L
   template<int> int a = 0;
   extern template int a<0>; // ok
@@ -273,9 +273,9 @@ namespace dr2387 { // dr2387: 9
 #endif
 }
 
-// dr2390 is in dr2390.cpp
+// cwg2390 is in cwg2390.cpp
 
-namespace dr2394 { // dr2394: 15
+namespace cwg2394 { // cwg2394: 15
 
 struct A {};
 const A a;
@@ -286,7 +286,7 @@ B b;
 
 }
 
-namespace dr2396 { // dr2396: no
+namespace cwg2396 { // cwg2396: no
   struct A {
     struct B;
     operator B B::*();
@@ -295,7 +295,7 @@ namespace dr2396 { // dr2396: no
 
   // FIXME: per P1787 "Calling a conversion function" example, all of the
   // examples below are well-formed, with B resolving to A::B, but currently
-  // it's been resolved to dr2396::B.
+  // it's been resolved to cwg2396::B.
 
   // void f(A a) { a.operator B B::*(); }
   // void g(A a) { a.operator decltype(B()) B::*(); }
@@ -303,13 +303,13 @@ namespace dr2396 { // dr2396: no
 }
 
 #if __cplusplus >= 201103L
-namespace dr2397 { // dr2397: 17
+namespace cwg2397 { // cwg2397: 17
   void foo() {
     int a[5];
 
     auto (&b)[5] = a;
     auto (*c)[5] = &a;
   }
-} // namespace dr2397
+} // namespace cwg2397
 
 #endif
diff --git a/clang/test/CXX/drs/dr24xx.cpp b/clang/test/CXX/drs/dr24xx.cpp
index 4534ed26e56d0..5ffaebda68c13 100644
--- a/clang/test/CXX/drs/dr24xx.cpp
+++ b/clang/test/CXX/drs/dr24xx.cpp
@@ -10,7 +10,7 @@
 // expected-no-diagnostics
 #endif
 
-namespace dr2406 { // dr2406: 5
+namespace cwg2406 { // cwg2406: 5
 #if __cplusplus >= 201703L
 void fallthrough(int n) {
   void g(), h(), i();
@@ -45,7 +45,7 @@ void fallthrough(int n) {
 #endif
 }
 
-namespace dr2450 { // dr2450: 18 review P2308R1
+namespace cwg2450 { // cwg2450: 18 review P2308R1
 #if __cplusplus >= 202302L
 struct S {int a;};
 template <S s>
@@ -59,7 +59,7 @@ f<{.a= 0}>();
 #endif
 }
 
-namespace dr2459 { // dr2459: 18 drafting P2308R1
+namespace cwg2459 { // cwg2459: 18 drafting P2308R1
 #if __cplusplus >= 202302L
 struct A {
   constexpr A(float) {}
@@ -69,7 +69,7 @@ X<1> x;
 #endif
 }
 
-namespace dr2445 { // dr2445: 19
+namespace cwg2445 { // cwg2445: 19
 #if __cplusplus >= 202002L
   template <typename> constexpr bool F = false;
   template <typename T> struct A { };
diff --git a/clang/test/CXX/drs/dr25xx.cpp b/clang/test/CXX/drs/dr25xx.cpp
index 46532486e50e5..62b2a0a088cc1 100644
--- a/clang/test/CXX/drs/dr25xx.cpp
+++ b/clang/test/CXX/drs/dr25xx.cpp
@@ -10,9 +10,9 @@
 // expected-no-diagnostics
 #endif
 
-// dr2504 is in dr2504.cpp
+// cwg2504 is in cwg2504.cpp
 
-namespace dr2516 { // dr2516: 3.0
+namespace cwg2516 { // cwg2516: 3.0
                    // NB: reusing 1482 test
 #if __cplusplus >= 201103L
 template <typename T> struct S {
@@ -21,9 +21,9 @@ template <typename T> struct S {
 enum E2 : S<E2>::I { e };
 // since-cxx11-error@-1 {{use of undeclared identifier 'E2'}}
 #endif
-} // namespace dr2516
+} // namespace cwg2516
 
-namespace dr2518 { // dr2518: 17
+namespace cwg2518 { // cwg2518: 17
 
 #if __cplusplus >= 201103L
 template <class T>
@@ -32,13 +32,13 @@ void f(T t) {
   // cxx11-14-error@-1 {{constexpr if is a C++17 extension}}
     static_assert(false, "must be int-sized");
     // since-cxx11-error@-1 {{static assertion failed: must be int-sized}}
-    //   since-cxx11-note@#dr2518-f-c {{in instantiation of function template specialization 'dr2518::f<char>' requested here}}
+    //   since-cxx11-note@#cwg2518-f-c {{in instantiation of function template specialization 'cwg2518::f<char>' requested here}}
   }
 }
 
 void g(char c) {
   f(0);
-  f(c); // #dr2518-f-c
+  f(c); // #cwg2518-f-c
 }
 
 template <typename Ty>
@@ -46,7 +46,7 @@ struct S {
   static_assert(false);
   // cxx11-14-error@-1 {{'static_assert' with no message is a C++17 extension}}
   // since-cxx11-error@-2 {{static assertion failed}}
-  //   since-cxx11-note@#dr2518-S-double {{in instantiation of template class 'dr2518::S<double>' requested here}}
+  //   since-cxx11-note@#cwg2518-S-double {{in instantiation of template class 'cwg2518::S<double>' requested here}}
 };
 
 template <>
@@ -58,13 +58,13 @@ struct S<float> {};
 int test_specialization() {
   S<int> s1;
   S<float> s2;
-  S<double> s3; // #dr2518-S-double
+  S<double> s3; // #cwg2518-S-double
 }
 #endif
 
 }
 
-namespace dr2521 { // dr2521: 17
+namespace cwg2521 { // cwg2521: 17
 #if __cplusplus >= 201103L
 #pragma clang diagnostic push
 #pragma clang diagnostic warning "-Wdeprecated-literal-operator"
@@ -76,61 +76,61 @@ template <char... Chars> decltype(sizeof 0)
 operator""  _div();
 // since-cxx11-warning@-1 {{identifier '_div' preceded by whitespace in a literal operator declaration is deprecated}}
 
-using ::dr2521::operator"" _\u03C0___;
-using ::dr2521::operator""_div;
+using ::cwg2521::operator"" _\u03C0___;
+using ::cwg2521::operator""_div;
 // since-cxx11-warning@-2 {{identifier '_π___' preceded by whitespace in a literal operator declaration is deprecated}}
 #pragma clang diagnostic pop
 #endif
-} // namespace dr2521
+} // namespace cwg2521
 
 
 #if __cplusplus >= 202302L
-namespace dr2553 { // dr2553: 18 review 2023-07-14
+namespace cwg2553 { // cwg2553: 18 review 2023-07-14
 struct B {
   virtual void f(this B&); 
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a virtual function}}
   static void f(this B&);
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a static function}}
-  virtual void g(); // #dr2553-g
+  virtual void g(); // #cwg2553-g
 };
 struct D : B {
   void g(this D&);
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a virtual function}}
-  //   since-cxx23-note@#dr2553-g {{overridden virtual function is here}}
+  //   since-cxx23-note@#cwg2553-g {{overridden virtual function is here}}
 };
 
 }
 #endif
 
 #if __cplusplus >= 202302L
-namespace dr2554 { // dr2554: 18 review 2021-12-10
+namespace cwg2554 { // cwg2554: 18 review 2021-12-10
 struct B {
-  virtual void f(); // #dr2554-g
+  virtual void f(); // #cwg2554-g
 };
 
 struct D : B {
   void f(this D&);
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a virtual function}}
-  //   since-cxx23-note@#dr2554-g {{overridden virtual function is here}}
+  //   since-cxx23-note@#cwg2554-g {{overridden virtual function is here}}
 };
 
 struct D2 : B {
   void f(this B&);
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a virtual function}}
-  //   since-cxx23-note@#dr2554-g {{overridden virtual function is here}}
+  //   since-cxx23-note@#cwg2554-g {{overridden virtual function is here}}
 };
 struct T {};
 struct D3 : B {
   void f(this T&);
   // since-cxx23-error@-1 {{an explicit object parameter cannot appear in a virtual function}}
-  //   since-cxx23-note@#dr2554-g {{overridden virtual function is here}}
+  //   since-cxx23-note@#cwg2554-g {{overridden virtual function is here}}
 };
 
 }
 #endif
 
 #if __cplusplus >= 202302L
-namespace dr2561 { // dr2561: 18 review 2023-11-09
+namespace cwg2561 { // cwg2561: 18 review 2023-11-09
 struct C {
     constexpr C(auto) { }
 };
@@ -145,7 +145,7 @@ void foo() {
 #endif
 
 
-namespace dr2565 { // dr2565: 16 open 2023-06-07
+namespace cwg2565 { // cwg2565: 16 open 2023-06-07
 #if __cplusplus >= 202002L
   template<typename T>
     concept C = requires (typename T::type x) {
@@ -165,28 +165,28 @@ namespace dr2565 { // dr2565: 16 open 2023-06-07
   static_assert(is_referenceable<int>::value);
 
   template<typename T, typename U>
-  concept TwoParams = requires (T *a, U b){ true;}; // #dr2565-TPC
+  concept TwoParams = requires (T *a, U b){ true;}; // #cwg2565-TPC
 
   template<typename T, typename U>
-    requires TwoParams<T, U> // #dr2565-TPSREQ
+    requires TwoParams<T, U> // #cwg2565-TPSREQ
   struct TwoParamsStruct{};
 
   using TPSU = TwoParamsStruct<void, void>;
   // since-cxx20-error@-1 {{constraints not satisfied for class template 'TwoParamsStruct'}}
-  //   since-cxx20-note@#dr2565-TPSREQ {{because 'TwoParams<void, void>' evaluated to false}}
-  //   since-cxx20-note@#dr2565-TPC {{because 'b' would be invalid: argument may not have 'void' type}}
+  //   since-cxx20-note@#cwg2565-TPSREQ {{because 'TwoParams<void, void>' evaluated to false}}
+  //   since-cxx20-note@#cwg2565-TPC {{because 'b' would be invalid: argument may not have 'void' type}}
 
   template<typename T, typename ...U>
-  concept Variadic = requires (U* ... a, T b){ true;}; // #dr2565-VC
+  concept Variadic = requires (U* ... a, T b){ true;}; // #cwg2565-VC
 
   template<typename T, typename ...U>
-    requires Variadic<T, U...> // #dr2565-VSREQ
+    requires Variadic<T, U...> // #cwg2565-VSREQ
   struct VariadicStruct{};
 
   using VSU = VariadicStruct<void, int, char, double>;
   // since-cxx20-error@-1 {{constraints not satisfied for class template 'VariadicStruct'}}
-  //   since-cxx20-note@#dr2565-VSREQ {{because 'Variadic<void, int, char, double>' evaluated to false}}
-  //   since-cxx20-note@#dr2565-VC {{because 'b' would be invalid: argument may not have 'void' type}}
+  //   since-cxx20-note@#cwg2565-VSREQ {{because 'Variadic<void, int, char, double>' evaluated to false}}
+  //   since-cxx20-note@#cwg2565-VC {{because 'b' would be invalid: argument may not have 'void' type}}
 
   template<typename T>
   concept ErrorRequires = requires (ErrorRequires auto x) {
@@ -211,7 +211,7 @@ namespace dr2565 { // dr2565: 16 open 2023-06-07
 #endif
 }
 
-namespace dr2583 { // dr2583: 19
+namespace cwg2583 { // cwg2583: 19
 #if __cplusplus >= 201103L
 struct A {
   int i;
@@ -236,9 +236,9 @@ union V {
 static_assert(!__is_layout_compatible(A, B), "");
 static_assert(__is_layout_compatible(U, V), "");
 #endif
-} // namespace dr2583
+} // namespace cwg2583
 
-namespace dr2598 { // dr2598: 18
+namespace cwg2598 { // cwg2598: 18
 #if __cplusplus >= 201103L
 struct NonLiteral {
     NonLiteral();
diff --git a/clang/test/CXX/drs/dr26xx.cpp b/clang/test/CXX/drs/dr26xx.cpp
index f151c9eea051a..f7a05b9827a23 100644
--- a/clang/test/CXX/drs/dr26xx.cpp
+++ b/clang/test/CXX/drs/dr26xx.cpp
@@ -7,7 +7,7 @@
 // RUN: %clang_cc1 -std=c++2c -triple x86_64-unknown-unknown %s -verify=expected,since-cxx11,since-cxx20,since-cxx23
 
 
-namespace dr2621 { // dr2621: 16
+namespace cwg2621 { // cwg2621: 16
 #if __cplusplus >= 202002L
 enum class E { a };
 namespace One {
@@ -16,7 +16,7 @@ using enum E_t; // typedef ok
 auto v = a;
 }
 namespace Two {
-using dr2621::E;
+using cwg2621::E;
 int E; // we see this
 using enum E;
 // since-cxx20-error@-1 {{unknown type name E}}
@@ -24,32 +24,32 @@ using enum E;
 #endif
 }
 
-namespace dr2628 { // dr2628: no
+namespace cwg2628 { // cwg2628: no
                    // this was reverted for the 16.x release
                    // due to regressions, see the issue for more details:
                    // https://github.com/llvm/llvm-project/issues/60777
 #if __cplusplus >= 202002L
 template <bool A = false, bool B = false>
 struct foo {
-  // The expected notes below should be removed when dr2628 is fully implemented again
-  constexpr foo() requires (!A && !B) = delete; // #dr2628-ctor-1
-  constexpr foo() requires (A || B) = delete; //  #dr2628-ctor-2
+  // The expected notes below should be removed when cwg2628 is fully implemented again
+  constexpr foo() requires (!A && !B) = delete; // #cwg2628-ctor-1
+  constexpr foo() requires (A || B) = delete; //  #cwg2628-ctor-2
 };
 
 void f() {
-  // The FIXME's below should be the expected errors when dr2628 is
+  // The FIXME's below should be the expected errors when cwg2628 is
   // fully implemented again.
-  foo fooable; // #dr2628-fooable
+  foo fooable; // #cwg2628-fooable
   // since-cxx20-error@-1 {{ambiguous deduction for template arguments of 'foo'}}
-  //   since-cxx20-note@#dr2628-ctor-1 {{candidate function [with A = false, B = false]}}
-  //   since-cxx20-note@#dr2628-ctor-2 {{candidate function [with A = false, B = false]}}
-  // FIXME-since-cxx20-error@#dr2628-fooable {{call to deleted}} 
-  //   FIXME-since-cxx20-note@#dr2628-ctor {{marked deleted here}} 
+  //   since-cxx20-note@#cwg2628-ctor-1 {{candidate function [with A = false, B = false]}}
+  //   since-cxx20-note@#cwg2628-ctor-2 {{candidate function [with A = false, B = false]}}
+  // FIXME-since-cxx20-error@#cwg2628-fooable {{call to deleted}} 
+  //   FIXME-since-cxx20-note@#cwg2628-ctor {{marked deleted here}} 
 }
 #endif
 }
 
-namespace dr2631 { // dr2631: 16
+namespace cwg2631 { // cwg2631: 16
 #if __cplusplus >= 202002L
   constexpr int g();
   consteval int f() {
@@ -67,7 +67,7 @@ namespace dr2631 { // dr2631: 16
 #endif
 }
 
-namespace dr2635 { // dr2635: 16
+namespace cwg2635 { // cwg2635: 16
 #if __cplusplus >= 202002L
 template<typename T>
 concept UnaryC = true;
@@ -97,9 +97,9 @@ void TemplUse() {
 #endif
 }
 
-// dr2636: na
+// cwg2636: na
 
-namespace dr2640 { // dr2640: 16
+namespace cwg2640 { // cwg2640: 16
 
 int \N{Λ} = 0;
 // expected-error@-1 {{'Λ' is not a valid Unicode character name}}
@@ -113,18 +113,18 @@ const char* emoji = "\N{🤡}";
 //   expected-note@-6 {{did you mean BAT ('🦇' U+1F987)?}}
 
 #define z(x) 0
-#define dr2640_a z(
-int x = dr2640_a\N{abc});
+#define cwg2640_a z(
+int x = cwg2640_a\N{abc});
 // expected-error@-1 {{'abc' is not a valid Unicode character name}}
-int y = dr2640_a\N{LOTUS});
+int y = cwg2640_a\N{LOTUS});
 // expected-error@-1 {{character <U+1FAB7> not allowed in an identifier}}
-// expected-error@-2 {{use of undeclared identifier 'dr2640_a🪷'}}
+// expected-error@-2 {{use of undeclared identifier 'cwg2640_a🪷'}}
 // expected-error@-3 {{extraneous ')' before ';'}}
 }
 
-// dr2642: na
+// cwg2642: na
 
-namespace dr2644 { // dr2644: 8
+namespace cwg2644 { // cwg2644: 8
 #if __cplusplus >= 201103L
 auto z = [a = 42](int a) {
 // cxx11-warning@-1 {{initialized lambda captures are a C++14 extension}}
@@ -136,20 +136,20 @@ auto z = [a = 42](int a) {
 }
 
 #if __cplusplus >= 202302L
-namespace dr2650 { // dr2650: 17
+namespace cwg2650 { // cwg2650: 17
 template <class T, T> struct S {};
-template <class T> int f(S<T, T{}>*); // #dr2650-f
+template <class T> int f(S<T, T{}>*); // #cwg2650-f
 class X {
   int m;
 };
 int i0 = f<X>(0);
 // since-cxx23-error@-1 {{no matching function for call to 'f'}}
-//   since-cxx23-note@#dr2650-f {{type 'X' of non-type template parameter is not a structural type}}
+//   since-cxx23-note@#cwg2650-f {{type 'X' of non-type template parameter is not a structural type}}
 }
 #endif
 
 #if __cplusplus >= 202302L
-namespace dr2653 { // dr2653: 18
+namespace cwg2653 { // cwg2653: 18
   struct Test { void f(this const auto& = Test{}); };
   // since-cxx23-error@-1 {{the explicit object parameter cannot have a default argument}}
   auto L = [](this const auto& = Test{}){};
@@ -157,7 +157,7 @@ namespace dr2653 { // dr2653: 18
 }
 #endif
 
-namespace dr2654 { // dr2654: 16
+namespace cwg2654 { // cwg2654: 16
 void f() {
     int neck, tail;
     volatile int brachiosaur;
@@ -167,7 +167,7 @@ void f() {
 }
 }
 
-namespace dr2681 { // dr2681: 17
+namespace cwg2681 { // cwg2681: 17
 #if __cplusplus >= 202002L
 using size_t = decltype(sizeof(int));
 
@@ -180,7 +180,7 @@ struct I {
   volatile T array[N];
 };
 template<size_t N>
-struct J { // #dr2681-J
+struct J { // #cwg2681-J
   unsigned char array[N];
 };
 
@@ -191,13 +191,13 @@ static_assert(__is_same(decltype(i), I<char, 4>));
 
 J j = { "ghi" };
 // since-cxx20-error@-1 {{no viable constructor or deduction guide}}
-//   since-cxx20-note@#dr2681-J {{candidate template ignored: could not match 'J<N>' against 'const char *'}}
-//   since-cxx20-note@#dr2681-J {{candidate template ignored: could not match 'const unsigned char' against 'const char'}}
-//   since-cxx20-note@#dr2681-J {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
+//   since-cxx20-note@#cwg2681-J {{candidate template ignored: could not match 'J<N>' against 'const char *'}}
+//   since-cxx20-note@#cwg2681-J {{candidate template ignored: could not match 'const unsigned char' against 'const char'}}
+//   since-cxx20-note@#cwg2681-J {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
 #endif
 }
 
-namespace dr2672 { // dr2672: 18
+namespace cwg2672 { // cwg2672: 18
 #if __cplusplus >= 202002L
 template <class T>
 void f(T) requires requires { []() { T::invalid; } (); };
@@ -205,8 +205,8 @@ void f(T) requires requires { []() { T::invalid; } (); };
 //   since-cxx20-note@-2 {{while substituting into a lambda expression here}}
 //   since-cxx20-note@-3 {{in instantiation of requirement here}}
 //   since-cxx20-note@-4 {{while substituting template arguments into constraint expression here}}
-//   since-cxx20-note@#dr2672-f-0 {{while checking constraint satisfaction for template 'f<int>' required here}}
-//   since-cxx20-note@#dr2672-f-0 {{in instantiation of function template specialization 'dr2672::f<int>' requested here}}
+//   since-cxx20-note@#cwg2672-f-0 {{while checking constraint satisfaction for template 'f<int>' required here}}
+//   since-cxx20-note@#cwg2672-f-0 {{in instantiation of function template specialization 'cwg2672::f<int>' requested here}}
 void f(...);
 
 template <class T>
@@ -216,14 +216,14 @@ void bar(T) requires requires {
 void bar(...);
 
 void m() {
-  f(0); // #dr2672-f-0
+  f(0); // #cwg2672-f-0
   bar(0);
 }
 #endif
 }
 
 #if __cplusplus >= 202302L
-namespace dr2687 { // dr2687: 18
+namespace cwg2687 { // cwg2687: 18
 struct S{
     void f(int);
     static void g(int);
@@ -232,7 +232,7 @@ struct S{
 
 void test() {
     (&S::f)(1);
-    // since-cxx23-error@-1 {{called object type 'void (dr2687::S::*)(int)' is not a function or function pointer}}
+    // since-cxx23-error@-1 {{called object type 'void (cwg2687::S::*)(int)' is not a function or function pointer}}
     (&S::g)(1);
     (&S::h)(S(), 1);
 }
diff --git a/clang/test/CXX/drs/dr27xx.cpp b/clang/test/CXX/drs/dr27xx.cpp
index c956c4355abd3..0434427d6c92a 100644
--- a/clang/test/CXX/drs/dr27xx.cpp
+++ b/clang/test/CXX/drs/dr27xx.cpp
@@ -10,7 +10,7 @@
 // expected-no-diagnostics
 #endif
 
-namespace dr2759 { // dr2759: 19
+namespace cwg2759 { // cwg2759: 19
 #if __cplusplus >= 201103L
 
 struct CStruct {
@@ -91,18 +91,18 @@ static_assert(!__is_layout_compatible(UnionLayout, UnionLayout3), "");
 static_assert(!__is_layout_compatible(StructWithAnonUnion, StructWithAnonUnion2), "");
 static_assert(!__is_layout_compatible(StructWithAnonUnion, StructWithAnonUnion3), "");
 #endif
-} // namespace dr2759
+} // namespace cwg2759
 
-namespace dr2789 { // dr2789: 18
+namespace cwg2789 { // cwg2789: 18
 #if __cplusplus >= 202302L
 template <typename T = int>
 struct Base {
-    constexpr void g(); // #dr2789-g1
+    constexpr void g(); // #cwg2789-g1
 };
 
 template <typename T = int>
 struct Base2 {
-    constexpr void g() requires true;  // #dr2789-g2
+    constexpr void g() requires true;  // #cwg2789-g2
 };
 
 template <typename T = int>
@@ -119,13 +119,13 @@ void test() {
     s.f();
     s.g();
     // since-cxx23-error@-1 {{call to member function 'g' is ambiguous}}
-    //   since-cxx23-note@#dr2789-g1 {{candidate function}}
-    //   since-cxx23-note@#dr2789-g2 {{candidate function}}
+    //   since-cxx23-note@#cwg2789-g1 {{candidate function}}
+    //   since-cxx23-note@#cwg2789-g2 {{candidate function}}
 }
 #endif
 }
 
-namespace dr2798 { // dr2798: 17
+namespace cwg2798 { // cwg2798: 17
 #if __cpp_static_assert >= 202306
 struct string {
   constexpr string() {
@@ -147,5 +147,5 @@ consteval X f() { return {}; }
 static_assert(false, f().s);
 // since-cxx26-error@-1 {{static assertion failed: Hello}}
 #endif
-} // namespace dr2798
+} // namespace cwg2798
 
diff --git a/clang/test/CXX/drs/dr28xx.cpp b/clang/test/CXX/drs/dr28xx.cpp
index 9b21d3410a049..4d9b0c76758d5 100644
--- a/clang/test/CXX/drs/dr28xx.cpp
+++ b/clang/test/CXX/drs/dr28xx.cpp
@@ -10,7 +10,7 @@
 // expected-no-diagnostics
 #endif
 
-namespace dr2847 { // dr2847: 19
+namespace cwg2847 { // cwg2847: 19
 
 #if __cplusplus >= 202002L
 
@@ -57,9 +57,9 @@ void B<int>::g() requires true;
 
 #endif
 
-} // namespace dr2847
+} // namespace cwg2847
 
-namespace dr2858 { // dr2858: 19
+namespace cwg2858 { // cwg2858: 19
 
 #if __cplusplus > 202302L
 
@@ -80,4 +80,4 @@ struct A {
 
 #endif
 
-} // namespace dr2858
+} // namespace cwg2858
diff --git a/clang/test/CXX/drs/dr2xx.cpp b/clang/test/CXX/drs/dr2xx.cpp
index cbb8734e10c64..e655e7226d51d 100644
--- a/clang/test/CXX/drs/dr2xx.cpp
+++ b/clang/test/CXX/drs/dr2xx.cpp
@@ -16,7 +16,7 @@ typedef __SIZE_TYPE__ size_t;
 #define fold
 #endif
 
-namespace dr200 { // dr200: dup 214
+namespace cwg200 { // cwg200: dup 214
   template <class T> T f(int);
   template <class T, class U> T f(U) = delete;
   // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
@@ -26,9 +26,9 @@ namespace dr200 { // dr200: dup 214
   }
 }
 
-// dr201 is in dr201.cpp
+// cwg201 is in cwg201.cpp
 
-namespace dr202 { // dr202: 3.1
+namespace cwg202 { // cwg202: 3.1
   template<typename T> T f();
   template<int (*g)()> struct X {
     int arr[fold(g == &f<int>) ? 1 : -1];
@@ -36,19 +36,19 @@ namespace dr202 { // dr202: 3.1
   template struct X<f>;
 }
 
-// FIXME (export) dr204: no
+// FIXME (export) cwg204: no
 
-namespace dr206 { // dr206: yes
-  struct S; // #dr206-S
+namespace cwg206 { // cwg206: yes
+  struct S; // #cwg206-S
   template<typename T> struct Q { S s; };
   // expected-error@-1 {{field has incomplete type 'S'}}
-  //   expected-note@#dr206-S {{forward declaration of 'dr206::S'}}
+  //   expected-note@#cwg206-S {{forward declaration of 'cwg206::S'}}
   template<typename T> void f() { S s; }
   // expected-error@-1 {{variable has incomplete type 'S'}}
-  //   expected-note@#dr206-S {{forward declaration of 'dr206::S'}}
+  //   expected-note@#cwg206-S {{forward declaration of 'cwg206::S'}}
 }
 
-namespace dr207 { // dr207: yes
+namespace cwg207 { // cwg207: yes
   class A {
   protected:
     static void f() {}
@@ -63,22 +63,22 @@ namespace dr207 { // dr207: yes
   };
 }
 
-// dr208 FIXME: write codegen test
+// cwg208 FIXME: write codegen test
 
-namespace dr209 { // dr209: 3.2
+namespace cwg209 { // cwg209: 3.2
   class A {
-    void f(); // #dr209-A-f
+    void f(); // #cwg209-A-f
   };
   class B {
     friend void A::f();
-    // expected-error@-1 {{friend function 'f' is a private member of 'dr209::A'}}
-    //   expected-note@#dr209-A-f {{implicitly declared private here}}
+    // expected-error@-1 {{friend function 'f' is a private member of 'cwg209::A'}}
+    //   expected-note@#cwg209-A-f {{implicitly declared private here}}
   };
 }
 
-// dr210 is in dr210.cpp
+// cwg210 is in cwg210.cpp
 
-namespace dr211 { // dr211: yes
+namespace cwg211 { // cwg211: yes
   struct A {
     A() try {
       throw 0;
@@ -89,26 +89,26 @@ namespace dr211 { // dr211: yes
   };
 }
 
-namespace dr213 { // dr213: yes
+namespace cwg213 { // cwg213: yes
   template <class T> struct A : T {
     void h(T t) {
       char &r1 = f(t);
       int &r2 = g(t);
       // expected-error@-1 {{explicit qualification required to use member 'g' from dependent base class}}
-      //   expected-note@#dr213-instantiation {{in instantiation of member function 'dr213::A<dr213::B>::h' requested here}}
-      //   expected-note@#dr213-B-g {{member is declared here}}
+      //   expected-note@#cwg213-instantiation {{in instantiation of member function 'cwg213::A<cwg213::B>::h' requested here}}
+      //   expected-note@#cwg213-B-g {{member is declared here}}
     }
   };
   struct B {
     int &f(B);
-    int &g(B); // #dr213-B-g
+    int &g(B); // #cwg213-B-g
   };
   char &f(B);
 
-  template void A<B>::h(B); // #dr213-instantiation
+  template void A<B>::h(B); // #cwg213-instantiation
 }
 
-namespace dr214 { // dr214: yes
+namespace cwg214 { // cwg214: yes
   template<typename T, typename U> T checked_cast(U from) { U::error; }
   template<typename T, typename U> T checked_cast(U *from);
   class C {};
@@ -121,7 +121,7 @@ namespace dr214 { // dr214: yes
   }
 }
 
-namespace dr215 { // dr215: 2.9
+namespace cwg215 { // cwg215: 2.9
   template<typename T> class X {
     friend void T::foo();
     int n;
@@ -131,7 +131,7 @@ namespace dr215 { // dr215: 2.9
   };
 }
 
-namespace dr216 { // dr216: no
+namespace cwg216 { // cwg216: no
   // FIXME: Should reject this: 'f' has linkage but its type does not,
   // and 'f' is odr-used but not defined in this TU.
   typedef enum { e } *E;
@@ -147,7 +147,7 @@ namespace dr216 { // dr216: no
   void g(S s, S::E e) { s.f(e); }
 }
 
-namespace dr217 { // dr217: yes
+namespace cwg217 { // cwg217: yes
   template<typename T> struct S {
     void f(int);
   };
@@ -155,7 +155,7 @@ namespace dr217 { // dr217: yes
   // expected-error@-1 {{default arguments cannot be added to an out-of-line definition of a member of a class template}}
 }
 
-namespace dr218 { // dr218: yes
+namespace cwg218 { // cwg218: yes
                   // NB: also dup 405
   namespace A {
     struct S {};
@@ -219,12 +219,12 @@ namespace dr218 { // dr218: yes
   void testG(G::C<G::X::A, G::Y::B> gc) { f(gc); }
 }
 
-// dr219: na
-// dr220: na
+// cwg219: na
+// cwg220: na
 
-namespace dr221 { // dr221: 3.6
-  struct A { // #dr221-S
-    A &operator=(int&); // #dr221-S-copy-assign
+namespace cwg221 { // cwg221: 3.6
+  struct A { // #cwg221-S
+    A &operator=(int&); // #cwg221-S-copy-assign
     A &operator+=(int&);
     static A &operator=(A&, double&);
     // expected-error@-1 {{overloaded 'operator=' cannot be a static member function}}
@@ -243,20 +243,20 @@ namespace dr221 { // dr221: 3.6
     a += n;
     a = c;
     // expected-error@-1 {{no viable overloaded '='}}
-    //   expected-note@#dr221-S-copy-assign {{candidate function not viable: no known conversion from 'char' to 'int &' for 1st argument}}
-    //   since-cxx11-note@#dr221-S {{candidate function (the implicit move assignment operator) not viable: no known conversion from 'char' to 'A' for 1st argument}}
-    //   expected-note@#dr221-S {{candidate function (the implicit copy assignment operator) not viable: no known conversion from 'char' to 'const A' for 1st argument}}
+    //   expected-note@#cwg221-S-copy-assign {{candidate function not viable: no known conversion from 'char' to 'int &' for 1st argument}}
+    //   since-cxx11-note@#cwg221-S {{candidate function (the implicit move assignment operator) not viable: no known conversion from 'char' to 'A' for 1st argument}}
+    //   expected-note@#cwg221-S {{candidate function (the implicit copy assignment operator) not viable: no known conversion from 'char' to 'const A' for 1st argument}}
     a += c;
     a = f;
     // expected-error@-1 {{no viable overloaded '='}}
-    //   expected-note@#dr221-S-copy-assign {{candidate function not viable: no known conversion from 'float' to 'int &' for 1st argument}}
-    //   since-cxx11-note@#dr221-S {{candidate function (the implicit move assignment operator) not viable: no known conversion from 'float' to 'A' for 1st argument}}
-    //   expected-note@#dr221-S {{candidate function (the implicit copy assignment operator) not viable: no known conversion from 'float' to 'const A' for 1st argument}}
+    //   expected-note@#cwg221-S-copy-assign {{candidate function not viable: no known conversion from 'float' to 'int &' for 1st argument}}
+    //   since-cxx11-note@#cwg221-S {{candidate function (the implicit move assignment operator) not viable: no known conversion from 'float' to 'A' for 1st argument}}
+    //   expected-note@#cwg221-S {{candidate function (the implicit copy assignment operator) not viable: no known conversion from 'float' to 'const A' for 1st argument}}
     a += f;
   }
 }
 
-namespace dr222 { // dr222: dup 637
+namespace cwg222 { // cwg222: dup 637
   void f(int a, int b, int c, int *x) {
 #pragma clang diagnostic push
 #pragma clang diagnostic warning "-Wunsequenced"
@@ -276,9 +276,9 @@ namespace dr222 { // dr222: dup 637
   }
 }
 
-// dr223: na
+// cwg223: na
 
-namespace dr224 { // dr224: 16
+namespace cwg224 { // cwg224: 16
   namespace example1 {
     template <class T> class A {
       typedef int type;
@@ -286,7 +286,7 @@ namespace dr224 { // dr224: 16
       A<T>::type b;
       A<T*>::type c;
       // cxx98-17-error@-1 {{missing 'typename' prior to dependent type name A<T *>::type; implicit 'typename' is a C++20 extension}}
-      ::dr224::example1::A<T>::type d;
+      ::cwg224::example1::A<T>::type d;
 
       class B {
         typedef int type;
@@ -295,7 +295,7 @@ namespace dr224 { // dr224: 16
         A<T>::type b;
         A<T*>::type c;
         // cxx98-17-error@-1 {{missing 'typename' prior to dependent type name A<T *>::type; implicit 'typename' is a C++20 extension}}
-        ::dr224::example1::A<T>::type d;
+        ::cwg224::example1::A<T>::type d;
 
         B::type e;
         A<T>::B::type f;
@@ -348,15 +348,15 @@ namespace dr224 { // dr224: 16
   }
 }
 
-// dr225: yes
-template<typename T> void dr225_f(T t) { dr225_g(t); }
-// expected-error@-1 {{call to function 'dr225_g' that is neither visible in the template definition nor found by argument-dependent lookup}}
-//   expected-note@#dr225-f {{in instantiation of function template specialization 'dr225_f<int>' requested here}}
-//   expected-note@#dr225-g {{'dr225_g' should be declared prior to the call site}}
-void dr225_g(int); // #dr225-g
-template void dr225_f(int); // #dr225-f
+// cwg225: yes
+template<typename T> void cwg225_f(T t) { cwg225_g(t); }
+// expected-error@-1 {{call to function 'cwg225_g' that is neither visible in the template definition nor found by argument-dependent lookup}}
+//   expected-note@#cwg225-f {{in instantiation of function template specialization 'cwg225_f<int>' requested here}}
+//   expected-note@#cwg225-g {{'cwg225_g' should be declared prior to the call site}}
+void cwg225_g(int); // #cwg225-g
+template void cwg225_f(int); // #cwg225-f
 
-namespace dr226 { // dr226: no
+namespace cwg226 { // cwg226: no
   // FIXME: This appears to be wrong: default arguments for function templates
   // are listed as a defect (in c++98) not an extension. EDG accepts them in
   // strict c++98 mode.
@@ -406,14 +406,14 @@ namespace dr226 { // dr226: no
   int x = foo(0, 0);
 }
 
-void dr227(bool b) { // dr227: yes
+void cwg227(bool b) { // cwg227: yes
   if (b)
     int n;
   else
     int n;
 }
 
-namespace dr228 { // dr228: yes
+namespace cwg228 { // cwg228: yes
   template <class T> struct X {
     void f();
   };
@@ -422,50 +422,50 @@ namespace dr228 { // dr228: yes
   };
 }
 
-namespace dr229 { // dr229: 2.9
+namespace cwg229 { // cwg229: 2.9
   template<typename T> void f();
   template<typename T> void f<T*>() {}
   // expected-error@-1 {{function template partial specialization is not allowed}}
   template<> void f<int>() {}
 }
 
-namespace dr230 { // dr230: 3.0
+namespace cwg230 { // cwg230: 3.0
   struct S {
     S() { f(); }
-    // expected-warning@-1 {{call to pure virtual member function 'f' has undefined behavior; overrides of 'f' in subclasses are not available in the constructor of 'dr230::S'}}
-    //   expected-note@#dr230-f {{'f' declared here}}
-    virtual void f() = 0; // #dr230-f
+    // expected-warning@-1 {{call to pure virtual member function 'f' has undefined behavior; overrides of 'f' in subclasses are not available in the constructor of 'cwg230::S'}}
+    //   expected-note@#cwg230-f {{'f' declared here}}
+    virtual void f() = 0; // #cwg230-f
   };
 }
 
-namespace dr231 { // dr231: yes
+namespace cwg231 { // cwg231: yes
   namespace outer {
     namespace inner {
-      int i; // #dr231-i
+      int i; // #cwg231-i
     }
     void f() { using namespace inner; }
     int j = i;
     // expected-error@-1 {{use of undeclared identifier 'i'; did you mean 'inner::i'?}}
-    //   expected-note@#dr231-i {{'inner::i' declared here}}
+    //   expected-note@#cwg231-i {{'inner::i' declared here}}
   }
 }
 
-// dr234: na
-// dr235: na
+// cwg234: na
+// cwg235: na
 
-namespace dr236 { // dr236: 3.2
+namespace cwg236 { // cwg236: 3.2
   void *p = int();
   // cxx98-warning@-1 {{expression which evaluates to zero treated as a null pointer constant of type 'void *'}}
   // since-cxx11-error@-2 {{cannot initialize a variable of type 'void *' with an rvalue of type 'int'}}
 }
 
-namespace dr237 { // dr237: dup 470
+namespace cwg237 { // cwg237: dup 470
   template<typename T> struct A { void f() { T::error; } };
   template<typename T> struct B : A<T> {};
   template struct B<int>; // ok
 }
 
-namespace dr239 { // dr239: yes
+namespace cwg239 { // cwg239: yes
   namespace NS {
     class T {};
     void f(T);
@@ -481,64 +481,64 @@ namespace dr239 { // dr239: yes
   }
 }
 
-// dr240: dup 616
+// cwg240: dup 616
 
-namespace dr241 { // dr241: yes
+namespace cwg241 { // cwg241: yes
   namespace A {
     struct B {};
-    template <int X> void f(); // #dr241-A-f
+    template <int X> void f(); // #cwg241-A-f
     template <int X> void g(B);
   }
   namespace C {
-    template <class T> void f(T t); // #dr241-C-f
-    template <class T> void g(T t); // #dr241-C-g
+    template <class T> void f(T t); // #cwg241-C-f
+    template <class T> void g(T t); // #cwg241-C-g
   }
   void h(A::B b) {
     f<3>(b);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
+    //   expected-note@#cwg241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
     // cxx98-17-error@-3 {{use of function template name with no prior declaration in function call with explicit template arguments is a C++20 extension}}
     g<3>(b);
     // cxx98-17-error@-1 {{use of function template name with no prior declaration in function call with explicit template arguments is a C++20 extension}}
     A::f<3>(b);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
+    //   expected-note@#cwg241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
     A::g<3>(b);
     C::f<3>(b);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr241-C-f {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
+    //   expected-note@#cwg241-C-f {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
     C::g<3>(b);
     // expected-error@-1 {{no matching function for call to 'g'}}
-    //   expected-note@#dr241-C-g {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
+    //   expected-note@#cwg241-C-g {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
     using C::f;
     using C::g;
     f<3>(b);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr241-C-f {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
-    //   expected-note@#dr241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
+    //   expected-note@#cwg241-C-f {{candidate template ignored: invalid explicitly-specified argument for template parameter 'T'}}
+    //   expected-note@#cwg241-A-f {{candidate function template not viable: requires 0 arguments, but 1 was provided}}
     g<3>(b);
   }
 }
 
-namespace dr243 { // dr243: yes
+namespace cwg243 { // cwg243: yes
   struct B;
   struct A {
-    A(B); // #dr243-A
+    A(B); // #cwg243-A
   };
   struct B {
-    operator A() = delete; // #dr243-B
+    operator A() = delete; // #cwg243-B
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   } b;
   A a1(b);
   A a2 = b;
   // expected-error@-1 {{conversion from 'struct B' to 'A' is ambiguous}}
-  //   expected-note@#dr243-A {{candidate constructor}}
-  //   expected-note@#dr243-B {{candidate function has been explicitly deleted}}
+  //   expected-note@#cwg243-A {{candidate constructor}}
+  //   expected-note@#cwg243-B {{candidate function has been explicitly deleted}}
 }
 
-namespace dr244 { // dr244: 11
-                  // NB: this test is reused by dr399
-  struct B {}; // #dr244-B
+namespace cwg244 { // cwg244: 11
+                  // NB: this test is reused by cwg399
+  struct B {}; // #cwg244-B
   struct D : B {};
 
   D D_object;
@@ -547,18 +547,18 @@ namespace dr244 { // dr244: 11
 
   void f() {
     D_object.~B();
-    // expected-error@-1 {{destructor type 'dr244::B' in object destruction expression does not match the type 'D' of the object being destroyed}}
-    //   expected-note@#dr244-B {{type 'dr244::B' found by destructor name lookup}}
+    // expected-error@-1 {{destructor type 'cwg244::B' in object destruction expression does not match the type 'D' of the object being destroyed}}
+    //   expected-note@#cwg244-B {{type 'cwg244::B' found by destructor name lookup}}
     D_object.B::~B();
     D_object.D::~B(); // FIXME: Missing diagnostic for this.
     B_ptr->~B();
     B_ptr->~B_alias();
     B_ptr->B_alias::~B();
     B_ptr->B_alias::~B_alias();
-    B_ptr->dr244::~B();
-    // expected-error@-1 {{qualified member access refers to a member in namespace 'dr244'}}
-    B_ptr->dr244::~B_alias();
-    // expected-error@-1 {{qualified member access refers to a member in namespace 'dr244'}}
+    B_ptr->cwg244::~B();
+    // expected-error@-1 {{qualified member access refers to a member in namespace 'cwg244'}}
+    B_ptr->cwg244::~B_alias();
+    // expected-error@-1 {{qualified member access refers to a member in namespace 'cwg244'}}
   }
 
   template<typename T, typename U>
@@ -570,10 +570,10 @@ namespace dr244 { // dr244: 11
     B_ptr->~B_alias();
     B_ptr->B_alias::~B();
     B_ptr->B_alias::~B_alias();
-    B_ptr->dr244::~B();
-    // expected-error@-1 {{'dr244' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
-    B_ptr->dr244::~B_alias();
-    // expected-error@-1 {{'dr244' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
+    B_ptr->cwg244::~B();
+    // expected-error@-1 {{'cwg244' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
+    B_ptr->cwg244::~B_alias();
+    // expected-error@-1 {{'cwg244' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
   }
   template void f<B, D>(B*, D);
 
@@ -582,7 +582,7 @@ namespace dr244 { // dr244: 11
     typedef E<int> F;
   }
   void g(N::F f) {
-    typedef N::F G; // #dr244-G
+    typedef N::F G; // #cwg244-G
     f.~G();
     f.G::~E();
     // expected-error@-1 {{ISO C++ requires the name after '::~' to be found in the same scope as the name before '::~'}}
@@ -601,7 +601,7 @@ namespace dr244 { // dr244: 11
     // Rejecting this seems correct, but most compilers accept, so we do also.
     f.N::F::~G();
     // expected-error@-1 {{qualified destructor name only found in lexical scope; omit the qualifier to find this type name by unqualified lookup}}
-    //   expected-note@#dr244-G {{type 'G' (aka 'E<int>') found by destructor name lookup}}
+    //   expected-note@#cwg244-G {{type 'G' (aka 'E<int>') found by destructor name lookup}}
   }
 
   // Bizarrely, compilers perform lookup in the scope for qualified destructor
@@ -613,46 +613,46 @@ namespace dr244 { // dr244: 11
     }
     template <typename U> void f(typename N::S<U>::Inner *p) {
       typedef typename N::S<U>::Inner T;
-      p->::dr244::QualifiedLookupInScope::N::S<U>::Inner::~T();
-      // expected-error@-1 {{no type named 'T' in 'dr244::QualifiedLookupInScope::N::S<int>'}}
-      //   expected-note@#dr244-f {{in instantiation of function template specialization 'dr244::QualifiedLookupInScope::f<int>' requested here}}
+      p->::cwg244::QualifiedLookupInScope::N::S<U>::Inner::~T();
+      // expected-error@-1 {{no type named 'T' in 'cwg244::QualifiedLookupInScope::N::S<int>'}}
+      //   expected-note@#cwg244-f {{in instantiation of function template specialization 'cwg244::QualifiedLookupInScope::f<int>' requested here}}
     }
-    template void f<int>(N::S<int>::Inner *); // #dr244-f
+    template void f<int>(N::S<int>::Inner *); // #cwg244-f
 
     template <typename U> void g(U *p) {
       typedef U T;
       p->T::~T();
       p->U::~T();
-      p->::dr244::QualifiedLookupInScope::N::S<int>::Inner::~T();
+      p->::cwg244::QualifiedLookupInScope::N::S<int>::Inner::~T();
       // expected-error@-1 {{'T' does not refer to a type name in pseudo-destructor expression; expected the name of type 'U'}}
     }
     template void g(N::S<int>::Inner *);
   }
 }
 
-namespace dr245 { // dr245: yes
+namespace cwg245 { // cwg245: yes
   struct S {
-    enum E {}; // #dr245-E
+    enum E {}; // #cwg245-E
     class E *p;
     // expected-error@-1 {{use of 'E' with tag type that does not match previous declaration}}
-    //   expected-note@#dr245-E {{previous use is here}}
+    //   expected-note@#cwg245-E {{previous use is here}}
   };
 }
 
-namespace dr246 { // dr246: 3.2
+namespace cwg246 { // cwg246: 3.2
   struct S {
-    S() try { // #dr246-try
+    S() try { // #cwg246-try
       throw 0;
 X: ;
     } catch (int) {
       goto X;
       // expected-error@-1 {{cannot jump from this goto statement to its label}}
-      //   expected-note@#dr246-try {{jump bypasses initialization of try block}}
+      //   expected-note@#cwg246-try {{jump bypasses initialization of try block}}
     }
   };
 }
 
-namespace dr247 { // dr247: yes
+namespace cwg247 { // cwg247: yes
   struct A {};
   struct B : A {
     void f();
@@ -678,16 +678,16 @@ namespace dr247 { // dr247: yes
   void (F::*i)() = &F::f;
 }
 
-namespace dr248 { // dr248: sup P1949
+namespace cwg248 { // cwg248: sup P1949
   int \u040d\u040e = 0;
 }
 
-namespace dr249 { // dr249: yes
+namespace cwg249 { // cwg249: yes
   template<typename T> struct X { void f(); };
   template<typename T> void X<T>::f() {}
 }
 
-namespace dr250 { // dr250: yes
+namespace cwg250 { // cwg250: yes
   typedef void (*FPtr)(double x[]);
 
   template<int I> void f(double x[]);
@@ -698,49 +698,49 @@ namespace dr250 { // dr250: yes
   FPtr gp = &g<>;
 }
 
-namespace dr252 { // dr252: 3.1
+namespace cwg252 { // cwg252: 3.1
   struct A {
-    void operator delete(void*); // #dr252-A
+    void operator delete(void*); // #cwg252-A
   };
   struct B {
-    void operator delete(void*); // #dr252-B
+    void operator delete(void*); // #cwg252-B
   };
   struct C : A, B {
     virtual ~C();
   };
   C::~C() {}
   // expected-error@-1 {{member 'operator delete' found in multiple base classes of different types}}
-  //   expected-note@#dr252-A {{member found by ambiguous name lookup}}
-  //   expected-note@#dr252-B {{member found by ambiguous name lookup}}
+  //   expected-note@#cwg252-A {{member found by ambiguous name lookup}}
+  //   expected-note@#cwg252-B {{member found by ambiguous name lookup}}
 
   struct D {
-    void operator delete(void*, int); // #dr252-D
+    void operator delete(void*, int); // #cwg252-D
     virtual ~D();
   };
   D::~D() {}
   // expected-error@-1 {{no suitable member 'operator delete' in 'D'}}
-  //   expected-note@#dr252-D {{member 'operator delete' declared here}}
+  //   expected-note@#cwg252-D {{member 'operator delete' declared here}}
 
   struct E {
     void operator delete(void*, int);
-    void operator delete(void*) = delete; // #dr252-E
+    void operator delete(void*) = delete; // #cwg252-E
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}} 
     virtual ~E();
   };
   E::~E() {}
   // expected-error@-1 {{attempt to use a deleted function}}
-  //   expected-note@#dr252-E {{'operator delete' has been explicitly marked deleted here}}
+  //   expected-note@#cwg252-E {{'operator delete' has been explicitly marked deleted here}}
 
   struct F {
     // If both functions are available, the first one is a placement delete.
     void operator delete(void*, size_t);
-    void operator delete(void*) = delete; // #dr252-F
+    void operator delete(void*) = delete; // #cwg252-F
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
     virtual ~F();
   };
   F::~F() {}
   // expected-error@-1 {{attempt to use a deleted function}}
-  //   expected-note@#dr252-F {{'operator delete' has been explicitly marked deleted here}}
+  //   expected-note@#cwg252-F {{'operator delete' has been explicitly marked deleted here}}
 
   struct G {
     void operator delete(void*, size_t);
@@ -749,33 +749,33 @@ namespace dr252 { // dr252: 3.1
   G::~G() {}
 }
 
-namespace dr254 { // dr254: 2.9
+namespace cwg254 { // cwg254: 2.9
   template<typename T> struct A {
     typedef typename T::type type; // ok even if this is a typedef-name, because
                                    // it's not an elaborated-type-specifier
     typedef struct T::type foo;
     // expected-error@-1 {{typedef 'type' cannot be referenced with a struct specifier}}
-    //   expected-note@#dr254-instantiation {{in instantiation of template class 'dr254::A<dr254::C>' requested here}}
-    //   expected-note@#dr254-C {{declared here}}
+    //   expected-note@#cwg254-instantiation {{in instantiation of template class 'cwg254::A<cwg254::C>' requested here}}
+    //   expected-note@#cwg254-C {{declared here}}
   };
   struct B { struct type {}; };
-  struct C { typedef struct {} type; }; // #dr254-C
+  struct C { typedef struct {} type; }; // #cwg254-C
   A<B>::type n;
-  A<C>::type n; // #dr254-instantiation
+  A<C>::type n; // #cwg254-instantiation
 }
 
-namespace dr255 { // dr255: yes
+namespace cwg255 { // cwg255: yes
 struct S {
   void operator delete(void *){};
   void operator delete(void *, int){};
 };
 void f(S *p) { delete p; }
-} // namespace dr255
+} // namespace cwg255
 
-// dr256: dup 624
+// cwg256: dup 624
 
-namespace dr257 { // dr257: 3.4
-  struct A { A(int); }; // #dr257-A
+namespace cwg257 { // cwg257: 3.4
+  struct A { A(int); }; // #cwg257-A
   struct B : virtual A {
     B() {}
     virtual void f() = 0;
@@ -785,13 +785,13 @@ namespace dr257 { // dr257: 3.4
   };
   struct D : B {
     D() {}
-    // expected-error@-1 {{constructor for 'dr257::D' must explicitly initialize the base class 'A' which does not have a default constructor}}
-    //   expected-note@#dr257-A {{'dr257::A' declared here}}
+    // expected-error@-1 {{constructor for 'cwg257::D' must explicitly initialize the base class 'A' which does not have a default constructor}}
+    //   expected-note@#cwg257-A {{'cwg257::A' declared here}}
     void f();
   };
 }
 
-namespace dr258 { // dr258: 2.8
+namespace cwg258 { // cwg258: 2.8
   struct A {
     void f(const int);
     template<typename> void g(int);
@@ -802,13 +802,13 @@ namespace dr258 { // dr258: 2.8
     using A::g;
     using A::h;
     int &f(int);
-    template<int> int &g(int); // #dr258-B-g
+    template<int> int &g(int); // #cwg258-B-g
     int &h();
   } b;
   int &w = b.f(0);
   int &x = b.g<int>(0);
   // expected-error@-1 {{no matching member function for call to 'g'}}
-  //   expected-note@#dr258-B-g {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
+  //   expected-note@#cwg258-B-g {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
   int &y = b.h();
   float &z = const_cast<const B&>(b).h();
 
@@ -820,53 +820,53 @@ namespace dr258 { // dr258: 2.8
   } d;
 
   struct E {
-    virtual void f() = 0; // #dr258-E-f
+    virtual void f() = 0; // #cwg258-E-f
   };
   struct F : E {
     void f() const {}
   } f;
   // expected-error@-1 {{variable type 'struct F' is an abstract class}}
-  //   expected-note@#dr258-E-f {{unimplemented pure virtual method 'f' in 'F'}}
+  //   expected-note@#cwg258-E-f {{unimplemented pure virtual method 'f' in 'F'}}
 }
 
-namespace dr259 { // dr259: 4
+namespace cwg259 { // cwg259: 4
   template<typename T> struct A {};
-  template struct A<int>; // #dr259-A-int 
+  template struct A<int>; // #cwg259-A-int 
   template struct A<int>;
   // expected-error@-1 {{duplicate explicit instantiation of 'A<int>'}}
-  //   expected-note@#dr259-A-int {{previous explicit instantiation is here}}
+  //   expected-note@#cwg259-A-int {{previous explicit instantiation is here}}
 
-  template<> struct A<float>; // #dr259-A-float
+  template<> struct A<float>; // #cwg259-A-float
   template struct A<float>;
   // expected-warning@-1 {{explicit instantiation of 'A<float>' that occurs after an explicit specialization has no effect}}
-  //   expected-note@#dr259-A-float {{previous template specialization is here}}
+  //   expected-note@#cwg259-A-float {{previous template specialization is here}}
 
-  template struct A<char>; // #dr259-A-char
+  template struct A<char>; // #cwg259-A-char
   template<> struct A<char>;
-  // expected-error@-1 {{explicit specialization of 'dr259::A<char>' after instantiation}}
-  //   expected-note@#dr259-A-char {{explicit instantiation first required here}}
+  // expected-error@-1 {{explicit specialization of 'cwg259::A<char>' after instantiation}}
+  //   expected-note@#cwg259-A-char {{explicit instantiation first required here}}
 
   template<> struct A<double>;
   template<> struct A<double>;
-  template<> struct A<double> {}; // #dr259-A-double
+  template<> struct A<double> {}; // #cwg259-A-double
   template<> struct A<double> {};
   // expected-error@-1 {{redefinition of 'A<double>'}}
-  //   expected-note@#dr259-A-double {{previous definition is here}}
+  //   expected-note@#cwg259-A-double {{previous definition is here}}
 
-  template<typename T> struct B; // #dr259-B
+  template<typename T> struct B; // #cwg259-B
   template struct B<int>;
-  // expected-error@-1 {{explicit instantiation of undefined template 'dr259::B<int>'}}
-  //   expected-note@#dr259-B {{template is declared here}}
+  // expected-error@-1 {{explicit instantiation of undefined template 'cwg259::B<int>'}}
+  //   expected-note@#cwg259-B {{template is declared here}}
 
-  template<> struct B<float>; // #dr259-B-float
+  template<> struct B<float>; // #cwg259-B-float
   template struct B<float>;
   // expected-warning@-1 {{explicit instantiation of 'B<float>' that occurs after an explicit specialization has no effect}}
-  //   expected-note@#dr259-B-float {{previous template specialization is here}}
+  //   expected-note@#cwg259-B-float {{previous template specialization is here}}
 }
 
-// FIXME: When dr260 is resolved, also add tests for DR507.
+// FIXME: When cwg260 is resolved, also add tests for CWG507.
 
-namespace dr261 { // dr261: no
+namespace cwg261 { // cwg261: no
 #pragma clang diagnostic push
 #pragma clang diagnostic warning "-Wused-but-marked-unused"
 
@@ -901,14 +901,14 @@ namespace dr261 { // dr261: no
 #pragma clang diagnostic pop
 }
 
-namespace dr262 { // dr262: yes
+namespace cwg262 { // cwg262: yes
   int f(int = 0, ...);
   int k = f();
   int l = f(0);
   int m = f(0, 0);
 }
 
-namespace dr263 { // dr263: 3.3
+namespace cwg263 { // cwg263: 3.3
   struct X {};
   struct Y {
 #if __cplusplus < 201103L
@@ -928,12 +928,12 @@ namespace dr263 { // dr263: 3.3
   };
 }
 
-// dr265: dup 353
-// dr266: na
-// dr269: na
-// dr270: na
+// cwg265: dup 353
+// cwg266: na
+// cwg269: na
+// cwg270: na
 
-namespace dr272 { // dr272: yes
+namespace cwg272 { // cwg272: yes
   struct X {
     void f() {
       this->~X();
@@ -946,7 +946,7 @@ namespace dr272 { // dr272: yes
 
 #include <stdarg.h>
 #include <stddef.h>
-namespace dr273 { // dr273: yes
+namespace cwg273 { // cwg273: yes
   struct A {
     int n;
   };
@@ -959,12 +959,12 @@ namespace dr273 { // dr273: yes
   }
 }
 
-// dr274: na
+// cwg274: na
 
-namespace dr275 { // dr275: no
+namespace cwg275 { // cwg275: no
   namespace N {
-    template <class T> void f(T) {} // #dr275-N-f
-    template <class T> void g(T) {} // #dr275-N-g
+    template <class T> void f(T) {} // #cwg275-N-f
+    template <class T> void g(T) {} // #cwg275-N-g
     template <> void f(int);
     template <> void f(char);
     template <> void f(double);
@@ -980,19 +980,19 @@ namespace dr275 { // dr275: no
     template <> void g(char) {}
     // FIXME: this should be rejected in c++98 too
     template void f(long);
-    // since-cxx11-error@-1 {{explicit instantiation of 'dr275::N::f' must occur in namespace 'N'}}
-    //   since-cxx11-note@#dr275-N-f {{explicit instantiation refers here}}
+    // since-cxx11-error@-1 {{explicit instantiation of 'cwg275::N::f' must occur in namespace 'N'}}
+    //   since-cxx11-note@#cwg275-N-f {{explicit instantiation refers here}}
     // FIXME: this should be rejected in c++98 too
     template void N::f(unsigned long);
     // since-cxx11-error@-1 {{explicit instantiation of 'f' not in a namespace enclosing 'N'}}
-    //   since-cxx11-note@#dr275-N-f {{explicit instantiation refers here}}
+    //   since-cxx11-note@#cwg275-N-f {{explicit instantiation refers here}}
     template void h(long);
     // expected-error@-1 {{explicit instantiation of 'h' does not refer to a function template, variable template, member function, member class, or static data member}}
     template <> void f(double) {}
     // expected-error@-1 {{no function template matches function template specialization 'f'}}
   }
 
-  template <class T> void g(T) {} // #dr275-g 
+  template <class T> void g(T) {} // #cwg275-g 
 
   template <> void N::f(char) {}
   template <> void f(int) {}
@@ -1000,87 +1000,87 @@ namespace dr275 { // dr275: no
 
   // FIXME: this should be rejected in c++98 too
   template void f(short);
-  // since-cxx11-error@-1 {{explicit instantiation of 'dr275::N::f' must occur in namespace 'N'}}
-  //   since-cxx11-note@#dr275-N-f {{explicit instantiation refers here}}
+  // since-cxx11-error@-1 {{explicit instantiation of 'cwg275::N::f' must occur in namespace 'N'}}
+  //   since-cxx11-note@#cwg275-N-f {{explicit instantiation refers here}}
   template void N::f(unsigned short);
 
   // FIXME: this should probably be valid. the wording from the issue
   // doesn't clarify this, but it follows from the usual rules.
   template void g(int);
   // expected-error@-1 {{partial ordering for explicit instantiation of 'g' is ambiguous}}
-  //   expected-note@#dr275-g {{explicit instantiation candidate function 'dr275::g<int>' template here [with T = int]}}
-  //   expected-note@#dr275-N-g {{explicit instantiation candidate function 'dr275::N::g<int>' template here [with T = int]}}
+  //   expected-note@#cwg275-g {{explicit instantiation candidate function 'cwg275::g<int>' template here [with T = int]}}
+  //   expected-note@#cwg275-N-g {{explicit instantiation candidate function 'cwg275::N::g<int>' template here [with T = int]}}
 
   // FIXME: likewise, this should also be valid.
-  template<typename T> void f(T) {} // #dr275-f
+  template<typename T> void f(T) {} // #cwg275-f
   template void f(short);
   // expected-error@-1 {{partial ordering for explicit instantiation of 'f' is ambiguous}}
-  //   expected-note@#dr275-f {{explicit instantiation candidate function 'dr275::f<short>' template here [with T = short]}}
-  //   expected-note@#dr275-N-f {{explicit instantiation candidate function 'dr275::N::f<short>' template here [with T = short]}}
+  //   expected-note@#cwg275-f {{explicit instantiation candidate function 'cwg275::f<short>' template here [with T = short]}}
+  //   expected-note@#cwg275-N-f {{explicit instantiation candidate function 'cwg275::N::f<short>' template here [with T = short]}}
 }
 
-// dr276: na
+// cwg276: na
 
-namespace dr277 { // dr277: 3.1
+namespace cwg277 { // cwg277: 3.1
   typedef int *intp;
   int *p = intp();
   int a[fold(intp() ? -1 : 1)];
 }
 
-namespace dr280 { // dr280: 2.9
+namespace cwg280 { // cwg280: 2.9
   typedef void f0();
   typedef void f1(int);
   typedef void f2(int, int);
   typedef void f3(int, int, int);
   struct A {
-    operator f1*(); // #dr280-A-f1
+    operator f1*(); // #cwg280-A-f1
     operator f2*();
   };
   struct B {
-    operator f0*(); // #dr280-B-f0
+    operator f0*(); // #cwg280-B-f0
   private:
-    operator f3*(); // #dr280-B-f3
+    operator f3*(); // #cwg280-B-f3
   };
   struct C {
-    operator f0*(); // #dr280-C-f0
-    operator f1*(); // #dr280-C-f1
-    operator f2*(); // #dr280-C-f2
-    operator f3*(); // #dr280-C-f3
+    operator f0*(); // #cwg280-C-f0
+    operator f1*(); // #cwg280-C-f1
+    operator f2*(); // #cwg280-C-f2
+    operator f3*(); // #cwg280-C-f3
   };
-  struct D : private A, B { // #dr280-D
-    operator f2*(); // #dr280-D-f2
+  struct D : private A, B { // #cwg280-D
+    operator f2*(); // #cwg280-D-f2
   } d;
   struct E : C, D {} e;
   void g() {
     d(); // ok, public
     d(0);
-    // expected-error@-1 {{'operator void (*)(int)' is a private member of 'dr280::A'}}
-    //   expected-note@#dr280-D {{constrained by private inheritance here}}
-    //   expected-note@#dr280-A-f1 {{member is declared here}}
+    // expected-error@-1 {{'operator void (*)(int)' is a private member of 'cwg280::A'}}
+    //   expected-note@#cwg280-D {{constrained by private inheritance here}}
+    //   expected-note@#cwg280-A-f1 {{member is declared here}}
     d(0, 0); // ok, suppressed by member in D
     d(0, 0, 0);
-    // expected-error@-1 {{'operator void (*)(int, int, int)' is a private member of 'dr280::B'}}
-    //   expected-note@#dr280-B-f3 {{declared private here}}
+    // expected-error@-1 {{'operator void (*)(int, int, int)' is a private member of 'cwg280::B'}}
+    //   expected-note@#cwg280-B-f3 {{declared private here}}
     e();
     // expected-error@-1 {{call to object of type 'struct E' is ambiguous}}
-    //   expected-note@#dr280-B-f0 {{conversion candidate of type 'void (*)()'}}
-    //   expected-note@#dr280-C-f0 {{conversion candidate of type 'void (*)()'}}
+    //   expected-note@#cwg280-B-f0 {{conversion candidate of type 'void (*)()'}}
+    //   expected-note@#cwg280-C-f0 {{conversion candidate of type 'void (*)()'}}
     e(0);
     // expected-error@-1 {{call to object of type 'struct E' is ambiguous}}
-    //   expected-note@#dr280-A-f1 {{conversion candidate of type 'void (*)(int)'}}
-    //   expected-note@#dr280-C-f1 {{conversion candidate of type 'void (*)(int)'}}
+    //   expected-note@#cwg280-A-f1 {{conversion candidate of type 'void (*)(int)'}}
+    //   expected-note@#cwg280-C-f1 {{conversion candidate of type 'void (*)(int)'}}
     e(0, 0);
     // expected-error@-1 {{call to object of type 'struct E' is ambiguous}}
-    //   expected-note@#dr280-C-f2 {{conversion candidate of type 'void (*)(int, int)'}}
-    //   expected-note@#dr280-D-f2 {{conversion candidate of type 'void (*)(int, int)'}}
+    //   expected-note@#cwg280-C-f2 {{conversion candidate of type 'void (*)(int, int)'}}
+    //   expected-note@#cwg280-D-f2 {{conversion candidate of type 'void (*)(int, int)'}}
     e(0, 0, 0);
     // expected-error@-1 {{call to object of type 'struct E' is ambiguous}}
-    //   expected-note@#dr280-B-f3 {{conversion candidate of type 'void (*)(int, int, int)'}}
-    //   expected-note@#dr280-C-f3 {{conversion candidate of type 'void (*)(int, int, int)'}}
+    //   expected-note@#cwg280-B-f3 {{conversion candidate of type 'void (*)(int, int, int)'}}
+    //   expected-note@#cwg280-C-f3 {{conversion candidate of type 'void (*)(int, int, int)'}}
   }
 }
 
-namespace dr281 { // dr281: no
+namespace cwg281 { // cwg281: no
   void a();
   inline void b();
 
@@ -1097,19 +1097,19 @@ namespace dr281 { // dr281: no
   };
 }
 
-namespace dr283 { // dr283: yes
-  template<typename T> // #dr283-template
+namespace cwg283 { // cwg283: yes
+  template<typename T> // #cwg283-template
   struct S {
     friend class T;
     // expected-error@-1 {{declaration of 'T' shadows template parameter}}
-    //   expected-note@#dr283-template {{template parameter is declared here}}
+    //   expected-note@#cwg283-template {{template parameter is declared here}}
     class T;
     // expected-error@-1 {{declaration of 'T' shadows template parameter}}
-    //   expected-note@#dr283-template {{template parameter is declared here}}
+    //   expected-note@#cwg283-template {{template parameter is declared here}}
   };
 }
 
-namespace dr284 { // dr284: no
+namespace cwg284 { // cwg284: no
   namespace A {
     struct X;
     enum Y {};
@@ -1122,11 +1122,11 @@ namespace dr284 { // dr284: no
     using A::Z;
   }
   struct B::V {};
-  // expected-error@-1 {{no struct named 'V' in namespace 'dr284::B'}}
+  // expected-error@-1 {{no struct named 'V' in namespace 'cwg284::B'}}
   struct B::W {};
   struct B::X {}; // FIXME: ill-formed
-  enum B::Y e; // ok per dr417
-  class B::Z z; // ok per dr417
+  enum B::Y e; // ok per cwg417
+  class B::Z z; // ok per cwg417
 
   struct C {
     struct X;
@@ -1140,26 +1140,26 @@ namespace dr284 { // dr284: no
     using C::Z;
   };
   struct D::V {};
-  // expected-error@-1 {{no struct named 'V' in 'dr284::D'}}
+  // expected-error@-1 {{no struct named 'V' in 'cwg284::D'}}
   struct D::W {};
   struct D::X {}; // FIXME: ill-formed
-  enum D::Y e2; // ok per dr417
-  class D::Z z2; // ok per dr417
+  enum D::Y e2; // ok per cwg417
+  class D::Z z2; // ok per cwg417
 }
 
-namespace dr285 { // dr285: yes
-  template<typename T> void f(T, int); // #dr285-f-T-int
-  template<typename T> void f(int, T); // #dr285-f-int-T
+namespace cwg285 { // cwg285: yes
+  template<typename T> void f(T, int); // #cwg285-f-T-int
+  template<typename T> void f(int, T); // #cwg285-f-int-T
   template<> void f<int>(int, int) {}
   // expected-error@-1 {{function template specialization 'f' ambiguously refers to more than one function template; explicitly specify additional template arguments to identify a particular function template}}
-  //   expected-note@#dr285-f-int-T {{function template 'dr285::f<int>' matches specialization [with T = int]}}
-  //   expected-note@#dr285-f-T-int {{function template 'dr285::f<int>' matches specialization [with T = int]}}
+  //   expected-note@#cwg285-f-int-T {{function template 'cwg285::f<int>' matches specialization [with T = int]}}
+  //   expected-note@#cwg285-f-T-int {{function template 'cwg285::f<int>' matches specialization [with T = int]}}
 }
 
-namespace dr286 { // dr286: 2.8
+namespace cwg286 { // cwg286: 2.8
   template<class T> struct A {
     class C {
-      template<class T2> struct B {}; // #dr286-B 
+      template<class T2> struct B {}; // #cwg286-B 
     };
   };
 
@@ -1168,29 +1168,29 @@ namespace dr286 { // dr286: 2.8
   struct A<T>::C::B<T2*> { };
 
   A<short>::C::B<int*> absip;
-  // expected-error@-1 {{'B' is a private member of 'dr286::A<short>::C'}}
-  //   expected-note@#dr286-B {{implicitly declared private here}}
+  // expected-error@-1 {{'B' is a private member of 'cwg286::A<short>::C'}}
+  //   expected-note@#cwg286-B {{implicitly declared private here}}
 }
 
-// dr288: na
+// cwg288: na
 
-namespace dr289 { // dr289: yes
-  struct A; // #dr289-A
+namespace cwg289 { // cwg289: yes
+  struct A; // #cwg289-A
   struct B : A {};
   // expected-error@-1 {{base class has incomplete type}}
-  //   expected-note@#dr289-A {{forward declaration of 'dr289::A'}}
+  //   expected-note@#cwg289-A {{forward declaration of 'cwg289::A'}}
 
   template<typename T> struct C { typename T::error error; };
   // expected-error@-1 {{type 'int' cannot be used prior to '::' because it has no members}}
-  //   expected-note@#dr289-C-int {{in instantiation of template class 'dr289::C<int>' requested here}}
-  struct D : C<int> {}; // #dr289-C-int
+  //   expected-note@#cwg289-C-int {{in instantiation of template class 'cwg289::C<int>' requested here}}
+  struct D : C<int> {}; // #cwg289-C-int
 }
 
-// dr290: na
-// dr291: dup 391
-// dr292 is in dr292.cpp
+// cwg290: na
+// cwg291: dup 391
+// cwg292 is in cwg292.cpp
 
-namespace dr294 { // dr294: no
+namespace cwg294 { // cwg294: no
   void f() throw(int);
   // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
   //   since-cxx17-note@-2 {{use 'noexcept(false)' instead}}
@@ -1218,7 +1218,7 @@ namespace dr294 { // dr294: no
   }
 }
 
-namespace dr295 { // dr295: 3.7
+namespace cwg295 { // cwg295: 3.7
   typedef int f();
   const f g;
   // expected-warning@-1 {{'const' qualifier on function type 'f' (aka 'int ()') has no effect}}
@@ -1237,21 +1237,21 @@ namespace dr295 { // dr295: 3.7
   // expected-warning@-1 {{'volatile' qualifier on function type 'U' (aka 'int ()') has no effect}}
 }
 
-namespace dr296 { // dr296: yes
+namespace cwg296 { // cwg296: yes
   struct A {
     static operator int() { return 0; }
     // expected-error@-1 {{conversion function must be a non-static member function}}
   };
 }
 
-namespace dr298 { // dr298: 3.1
+namespace cwg298 { // cwg298: 3.1
   struct A {
     typedef int type;
     A();
     ~A();
   };
-  typedef A B; // #dr298-B
-  typedef const A C; // #dr298-C
+  typedef A B; // #cwg298-B
+  typedef const A C; // #cwg298-C
 
   A::type i1;
   B::type i2;
@@ -1260,21 +1260,21 @@ namespace dr298 { // dr298: 3.1
   struct A a;
   struct B b;
   // expected-error@-1 {{typedef 'B' cannot be referenced with a struct specifier}}
-  //   expected-note@#dr298-B {{declared here}}
+  //   expected-note@#cwg298-B {{declared here}}
   struct C c;
   // expected-error@-1 {{typedef 'C' cannot be referenced with a struct specifier}}
-  //   expected-note@#dr298-C {{declared here}}
+  //   expected-note@#cwg298-C {{declared here}}
 
   B::B() {}
   // expected-error@-1 {{a type specifier is required for all declarations}}
   B::A() {} // ok
   C::~C() {}
-  // expected-error@-1 {{destructor cannot be declared using a typedef 'C' (aka 'const dr298::A') of the class name}}
+  // expected-error@-1 {{destructor cannot be declared using a typedef 'C' (aka 'const cwg298::A') of the class name}}
 
-  typedef struct D E; // #dr298-E
+  typedef struct D E; // #cwg298-E
   struct E {};
   // expected-error@-1 {{definition of type 'E' conflicts with typedef of the same name}}
-  //   expected-note@#dr298-E {{'E' declared here}}
+  //   expected-note@#cwg298-E {{'E' declared here}}
 
   struct F {
     ~F();
@@ -1283,22 +1283,22 @@ namespace dr298 { // dr298: 3.1
   G::~F() {} // ok
 }
 
-namespace dr299 { // dr299: 2.8 c++11
+namespace cwg299 { // cwg299: 2.8 c++11
   struct S {
     operator int();
   };
   struct T {
-    operator int(); // #dr299-int
-    operator unsigned short(); // #dr299-ushort
+    operator int(); // #cwg299-int
+    operator unsigned short(); // #cwg299-ushort
   };
   // FIXME: should this apply to c++98 mode?
   int *p = new int[S()];
   // cxx98-error@-1 {{implicit conversion from array size expression of type 'S' to integral type 'int' is a C++11 extension}}
-  int *q = new int[T()]; // #dr299-q
-  // cxx98-11-error@#dr299-q {{ambiguous conversion of array size expression of type 'T' to an integral or enumeration type}}
-  //  cxx98-11-note@#dr299-int {{conversion to integral type 'int' declared here}}
-  //  cxx98-11-note@#dr299-ushort {{conversion to integral type 'unsigned short' declared here}}
-  // since-cxx14-error-re@#dr299-q {{{{conversion from 'T' to 'unsigned (long long|long|int)' is ambiguous}}}}
-  //  since-cxx14-note@#dr299-int {{candidate function}}
-  //  since-cxx14-note@#dr299-ushort {{candidate function}}
+  int *q = new int[T()]; // #cwg299-q
+  // cxx98-11-error@#cwg299-q {{ambiguous conversion of array size expression of type 'T' to an integral or enumeration type}}
+  //  cxx98-11-note@#cwg299-int {{conversion to integral type 'int' declared here}}
+  //  cxx98-11-note@#cwg299-ushort {{conversion to integral type 'unsigned short' declared here}}
+  // since-cxx14-error-re@#cwg299-q {{{{conversion from 'T' to 'unsigned (long long|long|int)' is ambiguous}}}}
+  //  since-cxx14-note@#cwg299-int {{candidate function}}
+  //  since-cxx14-note@#cwg299-ushort {{candidate function}}
 }
diff --git a/clang/test/CXX/drs/dr3xx.cpp b/clang/test/CXX/drs/dr3xx.cpp
index d5cfebccf8134..6d1c6958ac8eb 100644
--- a/clang/test/CXX/drs/dr3xx.cpp
+++ b/clang/test/CXX/drs/dr3xx.cpp
@@ -5,14 +5,14 @@
 // RUN: %clang_cc1 -std=c++11 -verify=expected,cxx98-14,cxx98-17,cxx98-20,cxx11-14,since-cxx11 -triple %itanium_abi_triple %s -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++98 -verify=expected,cxx98-14,cxx98-17,cxx98-20,cxx98 -triple %itanium_abi_triple %s -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr300 { // dr300: yes
+namespace cwg300 { // cwg300: yes
   template<typename R, typename A> void f(R (&)(A)) {}
   int g(int);
   void h() { f(g); }
 }
 
-namespace dr301 { // dr301: 3.5
-  // see also dr38
+namespace cwg301 { // cwg301: 3.5
+  // see also cwg38
   struct S;
   template<typename T> void operator+(T, T);
   void operator-(S, S);
@@ -66,55 +66,55 @@ namespace dr301 { // dr301: 3.5
   // expected-error@-3 {{declaration does not declare anything}}
 }
 
-namespace dr302 { // dr302: 3.0
+namespace cwg302 { // cwg302: 3.0
   struct A { A(); ~A(); };
 #if __cplusplus < 201103L
   struct B {
-  // expected-error@-1 {{implicit default constructor for 'dr302::B' must explicitly initialize the const member 'n'}}
-  //   expected-note@#dr302-b {{in implicit default constructor for 'dr302::B' first required here}}
-  //   expected-note@#dr302-B-n {{declared here}}
-    const int n; // #dr302-B-n
+  // expected-error@-1 {{implicit default constructor for 'cwg302::B' must explicitly initialize the const member 'n'}}
+  //   expected-note@#cwg302-b {{in implicit default constructor for 'cwg302::B' first required here}}
+  //   expected-note@#cwg302-B-n {{declared here}}
+    const int n; // #cwg302-B-n
     A a;
-  } b = B(); // #dr302-b
+  } b = B(); // #cwg302-b
   // Trivial default constructor C::C() is not called here.
   struct C {
     const int n;
   } c = C();
 #else
   struct B {
-    const int n; // #dr302-B-n
+    const int n; // #cwg302-B-n
     A a;
   } b = B();
   // expected-error@-1 {{call to implicitly-deleted default constructor of 'B'}}
-  //   expected-note@#dr302-B-n {{default constructor of 'B' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
+  //   expected-note@#cwg302-B-n {{default constructor of 'B' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
   // C::C() is called here, because even though it's trivial, it's deleted.
   struct C {
-    const int n; // #dr302-C-n
+    const int n; // #cwg302-C-n
   } c = C();
   // expected-error@-1 {{call to implicitly-deleted default constructor of 'C'}}
-  //   expected-note@#dr302-C-n {{default constructor of 'C' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
+  //   expected-note@#cwg302-C-n {{default constructor of 'C' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
   struct D {
     const int n = 0;
   } d = D();
 #endif
 }
 
-// dr303: na
+// cwg303: na
 
-namespace dr304 { // dr304: 2.9
+namespace cwg304 { // cwg304: 2.9
   typedef int &a;
   int n = a();
   // expected-error@-1 {{reference to type 'int' requires an initializer}}
 
-  struct S { int &b; }; // #dr304-S
+  struct S { int &b; }; // #cwg304-S
   // cxx98-error@-1 {{reference to type 'int' requires an initializer}}
-  //   cxx98-note@#dr304-m {{in value-initialization of type 'S' here}}
-  int m = S().b; // #dr304-m
+  //   cxx98-note@#cwg304-m {{in value-initialization of type 'S' here}}
+  int m = S().b; // #cwg304-m
   // since-cxx11-error@-1 {{call to implicitly-deleted default constructor of 'S'}}
-  //   since-cxx11-note@#dr304-S {{default constructor of 'S' is implicitly deleted because field 'b' of reference type 'int &' would not be initialized}}
+  //   since-cxx11-note@#cwg304-S {{default constructor of 'S' is implicitly deleted because field 'b' of reference type 'int &' would not be initialized}}
 }
 
-namespace dr305 { // dr305: no
+namespace cwg305 { // cwg305: no
   struct A {
     typedef A C;
   };
@@ -129,10 +129,10 @@ namespace dr305 { // dr305: no
     b->~C();
   }
   void h(B *b) {
-    struct B {}; // #dr305-h-B
+    struct B {}; // #cwg305-h-B
     b->~B();
-    // expected-error@-1 {{destructor type 'B' in object destruction expression does not match the type 'B' (aka 'dr305::A') of the object being destroyed}}
-    //   expected-note@#dr305-h-B {{type 'B' found by destructor name lookup}}
+    // expected-error@-1 {{destructor type 'B' in object destruction expression does not match the type 'B' (aka 'cwg305::A') of the object being destroyed}}
+    //   expected-note@#cwg305-h-B {{type 'B' found by destructor name lookup}}
   }
 
   template<typename T> struct X {};
@@ -141,7 +141,7 @@ namespace dr305 { // dr305: no
     x->~X<int>();
     x->~X();
     x->~X<char>();
-    // expected-error@-1 {{no member named '~X' in 'dr305::X<int>'}}
+    // expected-error@-1 {{no member named '~X' in 'cwg305::X<int>'}}
   }
 
 #if __cplusplus >= 201103L
@@ -158,9 +158,9 @@ namespace dr305 { // dr305: no
   };
   void k(Z *z) {
     z->~T1<int>();
-    // expected-error@-1 {{no member named 'T1' in 'dr305::Z'}}
+    // expected-error@-1 {{no member named 'T1' in 'cwg305::Z'}}
     z->~T2<int>();
-    // expected-error@-1 {{no member named '~int' in 'dr305::Z'}}
+    // expected-error@-1 {{no member named '~int' in 'cwg305::Z'}}
     z->~T2<Z>();
   }
 
@@ -170,29 +170,29 @@ namespace dr305 { // dr305: no
   }
   template<typename A> using R = Q::R<int>;
   void qr(Q::R<int> x) { x.~R<char>(); }
-  // expected-error@-1 {{no member named '~R' in 'dr305::Q::R<int>'}}
+  // expected-error@-1 {{no member named '~R' in 'cwg305::Q::R<int>'}}
 #endif
 }
 
-namespace dr306 { // dr306: dup 39
+namespace cwg306 { // cwg306: dup 39
   struct A { struct B {}; };
   struct C { typedef A::B B; };
   struct D : A, A::B, C {};
   D::B b;
 
-  struct X {}; // #dr306-X
-  template<typename T> struct Y { typedef T X; }; // #dr306-typedef-X
+  struct X {}; // #cwg306-X
+  template<typename T> struct Y { typedef T X; }; // #cwg306-typedef-X
   template<typename T> struct Z : X, Y<T> {};
   Z<X>::X zx;
   Z<const X>::X zcx;
   // expected-error@-1 {{member 'X' found in multiple base classes of different types}}
-  //   expected-note@#dr306-X {{member type 'dr306::X' found}}
-  //   expected-note@#dr306-typedef-X {{member type 'const dr306::X' found}}
+  //   expected-note@#cwg306-X {{member type 'cwg306::X' found}}
+  //   expected-note@#cwg306-typedef-X {{member type 'const cwg306::X' found}}
 }
 
-// dr307: na
+// cwg307: na
 
-namespace dr308 { // dr308: 3.7
+namespace cwg308 { // cwg308: 3.7
   // This is mostly an ABI library issue.
   struct A {};
   struct B : A {};
@@ -210,19 +210,19 @@ namespace dr308 { // dr308: 3.7
     // positive rate should be acceptably low.
     try {
       throw D();
-    } catch (const A&) { // #dr308-catch-A
+    } catch (const A&) { // #cwg308-catch-A
       // unreachable
     } catch (const B&) {
       // expected-warning@-1 {{exception of type 'const B &' will be caught by earlier handler}}
-      //   expected-note@#dr308-catch-A {{for type 'const A &'}}
+      //   expected-note@#cwg308-catch-A {{for type 'const A &'}}
       // get here instead
     }
   }
 }
 
-// dr309: dup 485
+// cwg309: dup 485
 
-namespace dr311 { // dr311: 3.0
+namespace cwg311 { // cwg311: 3.0
   namespace X { namespace Y {} }
   namespace X::Y {}
   // cxx98-14-error@-1 {{nested namespace definition is a C++17 extension; define each namespace separately}}
@@ -231,23 +231,23 @@ namespace dr311 { // dr311: 3.0
     // cxx98-14-error@-1 {{nested namespace definition is a C++17 extension; define each namespace separately}}
   }
   // FIXME: The diagnostics here are not very good.
-  namespace ::dr311::X {}
+  namespace ::cwg311::X {}
   // expected-error@-1 {{expected identifier or '{'}}
   // expected-warning@-2 {{extra qualification on member 'X'}}
   // expected-error@-3 {{a type specifier is required for all declarations}}
   // expected-error@-4 {{expected ';' after top level declarator}}
 }
 
-// dr312: dup 616
+// cwg312: dup 616
 
-namespace dr313 { // dr313: dup 299 c++11
+namespace cwg313 { // cwg313: dup 299 c++11
   struct A { operator int() const; };
   // FIXME: should this be available in c++98 mode?
   int *p = new int[A()];
   // cxx98-error@-1 {{implicit conversion from array size expression of type 'A' to integral type 'int' is a C++11 extension}}
 }
 
-namespace dr314 { // dr314: no
+namespace cwg314 { // cwg314: no
                   // NB: dup 1710
 template <typename T> struct A {
   template <typename U> struct B {};
@@ -260,16 +260,16 @@ template <typename T> struct C2 : public A<T>::B<T> {
   C2() : A<T>::B<T>() {}
   // expected-error@-1 {{use 'template' keyword to treat 'B' as a dependent template name}}
 };
-} // namespace dr314
+} // namespace cwg314
 
-// dr315: na
-// dr316: sup 1004
+// cwg315: na
+// cwg316: sup 1004
 
-namespace dr317 { // dr317: 3.5
-  void f() {} // #dr317-f
+namespace cwg317 { // cwg317: 3.5
+  void f() {} // #cwg317-f
   inline void f();
   // expected-error@-1 {{inline declaration of 'f' follows non-inline definition}}
-  //   expected-note@#dr317-f {{previous definition is here}}
+  //   expected-note@#cwg317-f {{previous definition is here}}
 
   int g();
   int n = g();
@@ -277,19 +277,19 @@ namespace dr317 { // dr317: 3.5
 
   int h();
   int m = h();
-  int h() { return 0; } // #dr317-h
+  int h() { return 0; } // #cwg317-h
   inline int h();
   // expected-error@-1 {{inline declaration of 'h' follows non-inline definition}}
-  //   expected-note@#dr317-h {{previous definition is here}}
+  //   expected-note@#cwg317-h {{previous definition is here}}
 }
 
-namespace dr318 { // dr318: sup 1310
+namespace cwg318 { // cwg318: sup 1310
   struct A {};
   struct A::A a;
 }
 
-namespace dr319 { // dr319: no
-  // FIXME: dup dr389
+namespace cwg319 { // cwg319: no
+  // FIXME: dup cwg389
   // FIXME: We don't have a diagnostic for a name with linkage
   //        having a type without linkage.
   typedef struct {
@@ -325,7 +325,7 @@ namespace dr319 { // dr319: no
   }
 }
 
-namespace dr320 { // dr320: yes
+namespace cwg320 { // cwg320: yes
 #if __cplusplus >= 201103L
   struct X {
     constexpr X() {}
@@ -338,7 +338,7 @@ namespace dr320 { // dr320: yes
 #endif
 }
 
-namespace dr321 { // dr321: dup 557
+namespace cwg321 { // cwg321: dup 557
   namespace N {
     template<int> struct A {
       template<int> struct B;
@@ -358,7 +358,7 @@ namespace dr321 { // dr321: dup 557
   bool x = i == j;
 }
 
-namespace dr322 { // dr322: 2.8
+namespace cwg322 { // cwg322: 2.8
   struct A {
     template<typename T> operator T&();
   } a;
@@ -366,18 +366,18 @@ namespace dr322 { // dr322: 2.8
   int &s = a;
 }
 
-// dr323: no
+// cwg323: no
 
-namespace dr324 { // dr324: 3.6
-  struct S { int n : 1; } s; // #dr324-n
+namespace cwg324 { // cwg324: 3.6
+  struct S { int n : 1; } s; // #cwg324-n
   int &a = s.n;
   // expected-error@-1 {{non-const reference cannot bind to bit-field 'n'}}
-  //   expected-note@#dr324-n {{bit-field is declared here}}
+  //   expected-note@#cwg324-n {{bit-field is declared here}}
   int *b = &s.n;
   // expected-error@-1 {{address of bit-field requested}}
   int &c = (s.n = 0);
   // expected-error@-1 {{non-const reference cannot bind to bit-field 'n'}}
-  //   expected-note@#dr324-n {{bit-field is declared here}}
+  //   expected-note@#cwg324-n {{bit-field is declared here}}
   int *d = &(s.n = 0);
   // expected-error@-1 {{address of bit-field requested}}
   // FIXME: why don't we emit a note here, as for the rest of this type of diagnostic in this test?
@@ -387,19 +387,19 @@ namespace dr324 { // dr324: 3.6
   // expected-error@-1 {{address of bit-field requested}}
   int &g = (void(), s.n);
   // expected-error@-1 {{non-const reference cannot bind to bit-field 'n'}}
-  //   expected-note@#dr324-n {{bit-field is declared here}}
+  //   expected-note@#cwg324-n {{bit-field is declared here}}
   int *h = &(void(), s.n);
   // expected-error@-1 {{address of bit-field requested}}
   int *i = &++s.n;
   // expected-error@-1 {{address of bit-field requested}}
 }
 
-namespace dr326 { // dr326: 3.1
+namespace cwg326 { // cwg326: 3.1
   struct S {};
   int test[__is_trivially_constructible(S, const S&) ? 1 : -1];
 }
 
-namespace dr327 { // dr327: dup 538
+namespace cwg327 { // cwg327: dup 538
   struct A;
   class A {};
 
@@ -407,40 +407,40 @@ namespace dr327 { // dr327: dup 538
   struct B {};
 }
 
-namespace dr328 { // dr328: yes
-  struct A; // #dr328-A
+namespace cwg328 { // cwg328: yes
+  struct A; // #cwg328-A
   struct B { A a; };
   // expected-error@-1 {{field has incomplete type 'A'}}
-  //   expected-note@#dr328-A {{forward declaration of 'dr328::A'}}
+  //   expected-note@#cwg328-A {{forward declaration of 'cwg328::A'}}
   template<typename> struct C { A a; };
   // expected-error@-1 {{field has incomplete type 'A'}}
-  //   expected-note@#dr328-A {{forward declaration of 'dr328::A'}}
+  //   expected-note@#cwg328-A {{forward declaration of 'cwg328::A'}}
   A *p = new A[0];
   // expected-error@-1 {{allocation of incomplete type 'A'}}
-  //   expected-note@#dr328-A {{forward declaration of 'dr328::A'}}
+  //   expected-note@#cwg328-A {{forward declaration of 'cwg328::A'}}
 }
 
-namespace dr329 { // dr329: 3.5
+namespace cwg329 { // cwg329: 3.5
   struct B {};
   template<typename T> struct A : B {
     friend void f(A a) { g(a); }
     friend void h(A a) { g(a); }
     // expected-error@-1 {{use of undeclared identifier 'g'}}
-    //   expected-note@#dr329-h-call {{in instantiation of member function 'dr329::h' requested here}}
-    friend void i(B b) {} // #dr329-i
+    //   expected-note@#cwg329-h-call {{in instantiation of member function 'cwg329::h' requested here}}
+    friend void i(B b) {} // #cwg329-i
     // expected-error@-1 {{redefinition of 'i'}}
-    //   expected-note@#dr329-b {{in instantiation of template class 'dr329::A<char>' requested here}}
-    //   expected-note@#dr329-i {{previous definition is here}}
+    //   expected-note@#cwg329-b {{in instantiation of template class 'cwg329::A<char>' requested here}}
+    //   expected-note@#cwg329-i {{previous definition is here}}
   };
   A<int> a; 
-  A<char> b; // #dr329-b
+  A<char> b; // #cwg329-b
 
   void test() {
-    h(a); // #dr329-h-call
+    h(a); // #cwg329-h-call
   }
 }
 
-namespace dr330 { // dr330: 7
+namespace cwg330 { // cwg330: 7
   // Conversions between P and Q will be allowed by P0388.
   typedef int *(*P)[3];
   typedef const int *const (*Q)[3];
@@ -526,27 +526,27 @@ namespace dr330 { // dr330: 7
     void f(A *a) {
       (void) reinterpret_cast<B1*>(a);
       (void) reinterpret_cast<B2*>(a);
-      // expected-error@-1 {{ISO C++ does not allow reinterpret_cast from 'A *' (aka 'const volatile int (*)[1][2][3]') to 'B2 *' (aka 'int *const dr330::swift_17882::X::***') because it casts away qualifiers, even though the source and destination types are unrelated}}
+      // expected-error@-1 {{ISO C++ does not allow reinterpret_cast from 'A *' (aka 'const volatile int (*)[1][2][3]') to 'B2 *' (aka 'int *const cwg330::swift_17882::X::***') because it casts away qualifiers, even though the source and destination types are unrelated}}
       (void) reinterpret_cast<B3*>(a);
-      // expected-error@-1 {{ISO C++ does not allow reinterpret_cast from 'A *' (aka 'const volatile int (*)[1][2][3]') to 'B3 *' (aka 'int *dr330::swift_17882::X::*volatile **') because it casts away qualifiers, even though the source and destination types are unrelated}}
+      // expected-error@-1 {{ISO C++ does not allow reinterpret_cast from 'A *' (aka 'const volatile int (*)[1][2][3]') to 'B3 *' (aka 'int *cwg330::swift_17882::X::*volatile **') because it casts away qualifiers, even though the source and destination types are unrelated}}
       (void) reinterpret_cast<B4*>(a);
     }
   }
 }
 
-namespace dr331 { // dr331: 11
+namespace cwg331 { // cwg331: 11
   struct A {
-    A(volatile A&); // #dr331-A-ctor
+    A(volatile A&); // #cwg331-A-ctor
   };
   const A a;
   // expected-error@-1 {{no matching constructor for initialization of 'const A'}}
-  //   expected-note@#dr331-A-ctor {{candidate constructor not viable: requires 1 argument, but 0 were provided}}
+  //   expected-note@#cwg331-A-ctor {{candidate constructor not viable: requires 1 argument, but 0 were provided}}
   const A b(a);
   // expected-error@-1 {{no matching constructor for initialization of 'const A'}}
-  //   expected-note@#dr331-A-ctor {{candidate constructor not viable: 1st argument ('const A') would lose const qualifier}}
+  //   expected-note@#cwg331-A-ctor {{candidate constructor not viable: 1st argument ('const A') would lose const qualifier}}
 }
 
-namespace dr332 { // dr332: dup 577
+namespace cwg332 { // cwg332: dup 577
   void f(volatile void);
   // expected-error@-1 {{'void' as parameter must not have type qualifiers}}
   // cxx20-23-warning@-2 {{volatile-qualified parameter type 'volatile void' is deprecated}}
@@ -557,14 +557,14 @@ namespace dr332 { // dr332: dup 577
   // cxx20-23-warning@-2 {{volatile-qualified parameter type 'volatile void' is deprecated}}
 }
 
-namespace dr333 { // dr333: yes
+namespace cwg333 { // cwg333: yes
   int n = 0;
   int f(int(n));
   int g((int(n)));
   int h = f(g);
 }
 
-namespace dr334 { // dr334: yes
+namespace cwg334 { // cwg334: yes
   template<typename T> void f() {
     T x;
     f((x, 123));
@@ -576,9 +576,9 @@ namespace dr334 { // dr334: yes
   template void f<S>();
 }
 
-// dr335: no
+// cwg335: no
 
-namespace dr336 { // dr336: yes
+namespace cwg336 { // cwg336: yes
   namespace Pre {
     template<class T1> class A {
       template<class T2> class B {
@@ -588,7 +588,7 @@ namespace dr336 { // dr336: yes
     };
     template<> template<class X> class A<int>::B {};
     template<> template<> template<class T> void A<int>::B<double>::mf1(T t) {}
-    // expected-error@-1 {{out-of-line definition of 'mf1' does not match any declaration in 'dr336::Pre::A<int>::B<double>'}}
+    // expected-error@-1 {{out-of-line definition of 'mf1' does not match any declaration in 'cwg336::Pre::A<int>::B<double>'}}
     template<class Y> template<> void A<Y>::B<double>::mf2() {}
     // expected-error@-1 {{nested name specifier 'A<Y>::B<double>::' for declaration does not refer into a class, class template or class template partial specialization}}
   }
@@ -609,7 +609,7 @@ namespace dr336 { // dr336: yes
   }
 }
 
-namespace dr337 { // dr337: yes
+namespace cwg337 { // cwg337: yes
   template<typename T> void f(T (*)[1]);
   template<typename T> int &f(...);
 
@@ -617,7 +617,7 @@ namespace dr337 { // dr337: yes
   int &r = f<A>(0);
 
   // FIXME: The language rules here are completely broken. We cannot determine
-  // whether an incomplete type is abstract. See DR1640, which will probably
+  // whether an incomplete type is abstract. See CWG1640, which will probably
   // supersede this one and remove this rule.
   struct B;
   int &s = f<B>(0);
@@ -625,20 +625,20 @@ namespace dr337 { // dr337: yes
   struct B { virtual ~B() = 0; };
 }
 
-namespace dr339 { // dr339: 2.8
+namespace cwg339 { // cwg339: 2.8
   template <int I> struct A { static const int value = I; };
 
   char xxx(int);
   char (&xxx(float))[2];
 
-  template<class T> A<sizeof(xxx((T)0))> f(T) {} // #dr339-f 
+  template<class T> A<sizeof(xxx((T)0))> f(T) {} // #cwg339-f 
 
   void test() {
     A<1> a = f(0);
     A<2> b = f(0.0f);
     A<3> c = f("foo");
     // expected-error@-1 {{no matching function}}
-    //   expected-note@#dr339-f {{candidate}}
+    //   expected-note@#cwg339-f {{candidate}}
   }
 
 
@@ -658,50 +658,50 @@ namespace dr339 { // dr339: 2.8
   A<1> c = make_A<char>();
 }
 
-namespace dr340 { // dr340: yes
+namespace cwg340 { // cwg340: yes
   struct A { A(int); };
   struct B { B(A, A, int); };
   int x, y;
   B b(A(x), A(y), 3);
 }
 
-namespace dr341 { // dr341: sup 1708
+namespace cwg341 { // cwg341: sup 1708
   namespace A {
     int n;
-    extern "C" int &dr341_a = n; // #dr341_a
+    extern "C" int &cwg341_a = n; // #cwg341_a
   }
   namespace B {
-    extern "C" int &dr341_a = dr341_a;
-    // expected-error@-1 {{redefinition of 'dr341_a'}}
-    //   expected-note@#dr341_a {{previous definition is here}} 
+    extern "C" int &cwg341_a = cwg341_a;
+    // expected-error@-1 {{redefinition of 'cwg341_a'}}
+    //   expected-note@#cwg341_a {{previous definition is here}} 
   }
-  extern "C" void dr341_b(); // #dr341_b 
+  extern "C" void cwg341_b(); // #cwg341_b 
 }
-int dr341_a;
-// expected-error@-1 {{declaration of 'dr341_a' in global scope conflicts with declaration with C language linkage}}
-//   expected-note@#dr341_a {{declared with C language linkage here}}
-int dr341_b;
-// expected-error@-1 {{declaration of 'dr341_b' in global scope conflicts with declaration with C language linkage}}
-//   expected-note@#dr341_b {{declared with C language linkage here}}
-int dr341_c; // #dr341_c
-int dr341_d; // #dr341_d
-namespace dr341 {
-  extern "C" int dr341_c;
-  // expected-error@-1 {{declaration of 'dr341_c' with C language linkage conflicts with declaration in global scope}}
-  //   expected-note@#dr341_c {{declared in global scope here}}
-  extern "C" void dr341_d();
-  // expected-error@-1 {{declaration of 'dr341_d' with C language linkage conflicts with declaration in global scope}}
-  //   expected-note@#dr341_d {{declared in global scope here}}
-
-  namespace A { extern "C" int dr341_e; } // #dr341_e 
-  namespace B { extern "C" void dr341_e(); }
-  // expected-error@-1 {{redefinition of 'dr341_e' as different kind of symbol}}
-  //   expected-note@#dr341_e {{previous definition is here}}
+int cwg341_a;
+// expected-error@-1 {{declaration of 'cwg341_a' in global scope conflicts with declaration with C language linkage}}
+//   expected-note@#cwg341_a {{declared with C language linkage here}}
+int cwg341_b;
+// expected-error@-1 {{declaration of 'cwg341_b' in global scope conflicts with declaration with C language linkage}}
+//   expected-note@#cwg341_b {{declared with C language linkage here}}
+int cwg341_c; // #cwg341_c
+int cwg341_d; // #cwg341_d
+namespace cwg341 {
+  extern "C" int cwg341_c;
+  // expected-error@-1 {{declaration of 'cwg341_c' with C language linkage conflicts with declaration in global scope}}
+  //   expected-note@#cwg341_c {{declared in global scope here}}
+  extern "C" void cwg341_d();
+  // expected-error@-1 {{declaration of 'cwg341_d' with C language linkage conflicts with declaration in global scope}}
+  //   expected-note@#cwg341_d {{declared in global scope here}}
+
+  namespace A { extern "C" int cwg341_e; } // #cwg341_e 
+  namespace B { extern "C" void cwg341_e(); }
+  // expected-error@-1 {{redefinition of 'cwg341_e' as different kind of symbol}}
+  //   expected-note@#cwg341_e {{previous definition is here}}
 }
 
-// dr342: na
+// cwg342: na
 
-namespace dr343 { // dr343: no
+namespace cwg343 { // cwg343: no
   // FIXME: dup 1710
   template<typename T> struct A {
     template<typename U> struct B {};
@@ -714,32 +714,32 @@ namespace dr343 { // dr343: no
   };
 }
 
-namespace dr344 { // dr344: dup 1435
+namespace cwg344 { // cwg344: dup 1435
   struct A { inline virtual ~A(); };
   struct B { friend A::~A(); };
 }
 
-namespace dr345 { // dr345: yes
+namespace cwg345 { // cwg345: yes
   struct A {
     struct X {};
-    int X; // #dr345-int-X
+    int X; // #cwg345-int-X
   };
   struct B {
     struct X {};
   };
   template <class T> void f(T t) { typename T::X x; }
-  // expected-error@-1 {{typename specifier refers to non-type member 'X' in 'dr345::A'}}
-  //   expected-note@#dr345-f-a {{in instantiation of function template specialization 'dr345::f<dr345::A>' requested here}}
-  //   expected-note@#dr345-int-X {{referenced member 'X' is declared here}}
+  // expected-error@-1 {{typename specifier refers to non-type member 'X' in 'cwg345::A'}}
+  //   expected-note@#cwg345-f-a {{in instantiation of function template specialization 'cwg345::f<cwg345::A>' requested here}}
+  //   expected-note@#cwg345-int-X {{referenced member 'X' is declared here}}
   void f(A a, B b) {
     f(b);
-    f(a); // #dr345-f-a
+    f(a); // #cwg345-f-a
   }
 }
 
-// dr346: na
+// cwg346: na
 
-namespace dr347 { // dr347: yes
+namespace cwg347 { // cwg347: yes
   struct base {
     struct nested;
     static int n;
@@ -750,31 +750,31 @@ namespace dr347 { // dr347: yes
   struct derived : base {};
 
   struct derived::nested {};
-  // expected-error@-1 {{no struct named 'nested' in 'dr347::derived'}}
+  // expected-error@-1 {{no struct named 'nested' in 'cwg347::derived'}}
   int derived::n;
-  // expected-error@-1 {{no member named 'n' in 'dr347::derived'}}
+  // expected-error@-1 {{no member named 'n' in 'cwg347::derived'}}
   void derived::f() {}
-  // expected-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'dr347::derived'}}
+  // expected-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'cwg347::derived'}}
   void derived::g() {}
-  // expected-error@-1 {{out-of-line definition of 'g' does not match any declaration in 'dr347::derived'}}
+  // expected-error@-1 {{out-of-line definition of 'g' does not match any declaration in 'cwg347::derived'}}
 }
 
-// dr348: na
+// cwg348: na
 
-namespace dr349 { // dr349: no
+namespace cwg349 { // cwg349: no
   struct A {
     template <class T> operator T ***() {
       int ***p = 0;
       return p;
       // cxx98-20-error@-1 {{cannot initialize return object of type 'const int ***' with an lvalue of type 'int ***'}}
       // since-cxx23-error@-2 {{cannot initialize return object of type 'const int ***' with an rvalue of type 'int ***'}}
-      //   expected-note@#dr349-p1 {{in instantiation of function template specialization 'dr349::A::operator const int ***<const int>' requested here}}
+      //   expected-note@#cwg349-p1 {{in instantiation of function template specialization 'cwg349::A::operator const int ***<const int>' requested here}}
     }
   };
 
   // FIXME: This is valid.
   A a;
-  const int *const *const *p1 = a; // #dr349-p1
+  const int *const *const *p1 = a; // #cwg349-p1
 
   struct B {
     template <class T> operator T ***() {
@@ -788,28 +788,28 @@ namespace dr349 { // dr349: no
   const int *const *const *p2 = b;
 }
 
-// dr351: na
+// cwg351: na
 
-namespace dr352 { // dr352: 2.8
+namespace cwg352 { // cwg352: 2.8
   namespace example1 {
     namespace A {
       enum E {};
-      template<typename R, typename A> void foo(E, R (*)(A)); // #dr352-foo
+      template<typename R, typename A> void foo(E, R (*)(A)); // #cwg352-foo
     }
 
     template<typename T> void arg(T);
-    template<typename T> int arg(T) = delete; // #dr352-deleted
+    template<typename T> int arg(T) = delete; // #cwg352-deleted
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
 
     void f(A::E e) {
       foo(e, &arg);
       // expected-error@-1 {{no matching function for call to 'foo'}}
-      //   expected-note@#dr352-foo {{candidate template ignored: couldn't infer template argument 'R'}}
+      //   expected-note@#cwg352-foo {{candidate template ignored: couldn't infer template argument 'R'}}
 
       using A::foo;
       foo<int, int>(e, &arg);
       // expected-error@-1 {{attempt to use a deleted function}}
-      //   expected-note@#dr352-deleted {{'arg<int>' has been explicitly marked deleted here}}
+      //   expected-note@#cwg352-deleted {{'arg<int>' has been explicitly marked deleted here}}
     }
 
     int arg(int);
@@ -817,7 +817,7 @@ namespace dr352 { // dr352: 2.8
     void g(A::E e) {
       foo(e, &arg);
       // expected-error@-1 {{no matching function for call to 'foo'}}
-      //   expected-note@#dr352-foo {{candidate template ignored: couldn't infer template argument 'R'}} 
+      //   expected-note@#cwg352-foo {{candidate template ignored: couldn't infer template argument 'R'}} 
 
       using A::foo;
       foo<int, int>(e, &arg); // ok, uses non-template
@@ -826,7 +826,7 @@ namespace dr352 { // dr352: 2.8
 
   namespace contexts {
     template<int I> void f1(int (&)[I]);
-    template<int I> void f2(int (&)[I+1]); // #dr352-f2
+    template<int I> void f2(int (&)[I+1]); // #cwg352-f2
     template<int I> void f3(int (&)[I+1], int (&)[I]);
     void f() {
       int a[4];
@@ -834,13 +834,13 @@ namespace dr352 { // dr352: 2.8
       f1(a);
       f2(a);
       // expected-error@-1 {{no matching function for call to 'f2'}}
-      //   expected-note@#dr352-f2 {{candidate template ignored: couldn't infer template argument 'I'}}
+      //   expected-note@#cwg352-f2 {{candidate template ignored: couldn't infer template argument 'I'}}
       f3(a, b);
     }
 
     template<int I> struct S {};
     template<int I> void g1(S<I>);
-    template<int I> void g2(S<I+1>); // #dr352-g2
+    template<int I> void g2(S<I+1>); // #cwg352-g2
     template<int I> void g3(S<I+1>, S<I>);
     void g() {
       S<4> a;
@@ -848,24 +848,24 @@ namespace dr352 { // dr352: 2.8
       g1(a);
       g2(a);
       // expected-error@-1 {{no matching function for call to 'g2'}}
-      //   expected-note@#dr352-g2 {{candidate template ignored: couldn't infer template argument 'I'}}
+      //   expected-note@#cwg352-g2 {{candidate template ignored: couldn't infer template argument 'I'}}
       g3(a, b);
     }
 
-    template<typename T> void h1(T = 0); // #dr352-h1
+    template<typename T> void h1(T = 0); // #cwg352-h1
     template<typename T> void h2(T, T = 0);
     void h() {
       h1();
       // expected-error@-1 {{no matching function for call to 'h1'}}
-      //   expected-note@#dr352-h1 {{candidate template ignored: couldn't infer template argument 'T'}}
+      //   expected-note@#cwg352-h1 {{candidate template ignored: couldn't infer template argument 'T'}}
       h1(0);
       h1<int>();
       h2(0);
     }
 
     template<typename T> int tmpl(T);
-    template<typename R, typename A> void i1(R (*)(A)); // #dr352-i1
-    template<typename R, typename A> void i2(R, A, R (*)(A)); // #dr352-i2
+    template<typename R, typename A> void i1(R (*)(A)); // #cwg352-i1
+    template<typename R, typename A> void i2(R, A, R (*)(A)); // #cwg352-i2
     void i() {
       extern int single(int);
       i1(single);
@@ -874,20 +874,20 @@ namespace dr352 { // dr352: 2.8
       extern int ambig(float), ambig(int);
       i1(ambig);
       // expected-error@-1 {{no matching function for call to 'i1'}}
-      //   expected-note@#dr352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
+      //   expected-note@#cwg352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
       i2(0, 0, ambig);
 
       extern void no_match(float), no_match(int);
       i1(no_match);
       // expected-error@-1 {{no matching function for call to 'i1'}}
-      //   expected-note@#dr352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
+      //   expected-note@#cwg352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
       i2(0, 0, no_match);
       // expected-error@-1 {{no matching function for call to 'i2'}}
-      //   expected-note@#dr352-i2 {{candidate function [with R = int, A = int] not viable: no overload of 'no_match' matching 'int (*)(int)' for 3rd argument}}
+      //   expected-note@#cwg352-i2 {{candidate function [with R = int, A = int] not viable: no overload of 'no_match' matching 'int (*)(int)' for 3rd argument}}
 
       i1(tmpl);
       // expected-error@-1 {{no matching function for call to 'i1'}}
-      //   expected-note@#dr352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
+      //   expected-note@#cwg352-i1 {{candidate template ignored: couldn't infer template argument 'R'}}
       i2(0, 0, tmpl);
     }
   }
@@ -918,109 +918,109 @@ namespace dr352 { // dr352: 2.8
 
   namespace example5 {
     template<int I> class A {};
-    template<int I> void g(A<I+1>); // #dr352-g 
+    template<int I> void g(A<I+1>); // #cwg352-g 
     template<int I> void f(A<I>, A<I+1>);
     void h(A<1> a1, A<2> a2) {
       g(a1);
       // expected-error@-1 {{no matching function for call to 'g'}}
-      //   expected-note@#dr352-g {{candidate template ignored: couldn't infer template argument 'I'}}
+      //   expected-note@#cwg352-g {{candidate template ignored: couldn't infer template argument 'I'}}
       g<0>(a1);
       f(a1, a2);
     }
   }
 }
 
-// dr353 needs an IRGen test.
+// cwg353 needs an IRGen test.
 
-namespace dr354 { // dr354: yes c++11
+namespace cwg354 { // cwg354: yes c++11
   // FIXME: Should we allow this in C++98 too?
   struct S {};
 
-  template<int*> struct ptr {}; // #dr354-ptr
-  ptr<0> p0; // #dr354-p0
-  // cxx98-error@#dr354-p0 {{non-type template argument does not refer to any declaration}}
-  //   cxx98-note@#dr354-ptr {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-p0 {{null non-type template argument must be cast to template parameter type 'int *'}}
-  //   cxx11-14-note@#dr354-ptr {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-p0 {{conversion from 'int' to 'int *' is not allowed in a converted constant expression}}
+  template<int*> struct ptr {}; // #cwg354-ptr
+  ptr<0> p0; // #cwg354-p0
+  // cxx98-error@#cwg354-p0 {{non-type template argument does not refer to any declaration}}
+  //   cxx98-note@#cwg354-ptr {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-p0 {{null non-type template argument must be cast to template parameter type 'int *'}}
+  //   cxx11-14-note@#cwg354-ptr {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-p0 {{conversion from 'int' to 'int *' is not allowed in a converted constant expression}}
   ptr<(int*)0> p1;
   // cxx98-error@-1 {{non-type template argument does not refer to any declaration}}
-  //   cxx98-note@#dr354-ptr {{template parameter is declared here}}
-  ptr<(float*)0> p2; // #dr354-p2
-  // cxx98-error@#dr354-p2 {{non-type template argument does not refer to any declaration}}
-  //   cxx98-note@#dr354-ptr {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-p2 {{null non-type template argument of type 'float *' does not match template parameter of type 'int *'}}
-  //   cxx11-14-note@#dr354-ptr {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-p2 {{value of type 'float *' is not implicitly convertible to 'int *'}}
-  ptr<(int S::*)0> p3; // #dr354-p3
-  // cxx98-error@#dr354-p3 {{non-type template argument does not refer to any declaration}}
-  //   cxx98-note@#dr354-ptr {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-p3 {{null non-type template argument of type 'int dr354::S::*' does not match template parameter of type 'int *'}}
-  //   cxx11-14-note@#dr354-ptr {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-p3 {{value of type 'int dr354::S::*' is not implicitly convertible to 'int *'}}
-
-  template<int*> int both(); // #dr354-both-int-ptr
-  template<int> int both(); // #dr354-both-int
+  //   cxx98-note@#cwg354-ptr {{template parameter is declared here}}
+  ptr<(float*)0> p2; // #cwg354-p2
+  // cxx98-error@#cwg354-p2 {{non-type template argument does not refer to any declaration}}
+  //   cxx98-note@#cwg354-ptr {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-p2 {{null non-type template argument of type 'float *' does not match template parameter of type 'int *'}}
+  //   cxx11-14-note@#cwg354-ptr {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-p2 {{value of type 'float *' is not implicitly convertible to 'int *'}}
+  ptr<(int S::*)0> p3; // #cwg354-p3
+  // cxx98-error@#cwg354-p3 {{non-type template argument does not refer to any declaration}}
+  //   cxx98-note@#cwg354-ptr {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-p3 {{null non-type template argument of type 'int cwg354::S::*' does not match template parameter of type 'int *'}}
+  //   cxx11-14-note@#cwg354-ptr {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-p3 {{value of type 'int cwg354::S::*' is not implicitly convertible to 'int *'}}
+
+  template<int*> int both(); // #cwg354-both-int-ptr
+  template<int> int both(); // #cwg354-both-int
   int b0 = both<0>();
   int b1 = both<(int*)0>();
   // cxx98-error@-1 {{no matching function for call to 'both'}}
-  //   cxx98-note@#dr354-both-int-ptr {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
-  //   cxx98-note@#dr354-both-int {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
-
-  template<int S::*> struct ptr_mem {}; // #dr354-ptr_mem
-  ptr_mem<0> m0; // #dr354-m0
-  // cxx98-error@#dr354-m0 {{non-type template argument of type 'int' cannot be converted to a value of type 'int dr354::S::*'}}
-  //   cxx98-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-m0 {{null non-type template argument must be cast to template parameter type 'int dr354::S::*'}}
-  //   cxx11-14-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-m0 {{conversion from 'int' to 'int dr354::S::*' is not allowed in a converted constant expression}}
+  //   cxx98-note@#cwg354-both-int-ptr {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
+  //   cxx98-note@#cwg354-both-int {{candidate template ignored: invalid explicitly-specified argument for 1st template parameter}}
+
+  template<int S::*> struct ptr_mem {}; // #cwg354-ptr_mem
+  ptr_mem<0> m0; // #cwg354-m0
+  // cxx98-error@#cwg354-m0 {{non-type template argument of type 'int' cannot be converted to a value of type 'int cwg354::S::*'}}
+  //   cxx98-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-m0 {{null non-type template argument must be cast to template parameter type 'int cwg354::S::*'}}
+  //   cxx11-14-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-m0 {{conversion from 'int' to 'int cwg354::S::*' is not allowed in a converted constant expression}}
   ptr_mem<(int S::*)0> m1;
   // cxx98-error@-1 {{non-type template argument is not a pointer to member constant}}
-  ptr_mem<(float S::*)0> m2; // #dr354-m2
-  // cxx98-error@#dr354-m2 {{non-type template argument of type 'float dr354::S::*' cannot be converted to a value of type 'int dr354::S::*'}}
-  //   cxx98-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-m2 {{null non-type template argument of type 'float dr354::S::*' does not match template parameter of type 'int dr354::S::*'}}
-  //   cxx11-14-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-m2 {{value of type 'float dr354::S::*' is not implicitly convertible to 'int dr354::S::*'}}
-  ptr_mem<(int *)0> m3; // #dr354-m3
-  // cxx98-error@#dr354-m3 {{non-type template argument of type 'int *' cannot be converted to a value of type 'int dr354::S::*'}}
-  //   cxx98-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // cxx11-14-error@#dr354-m3 {{null non-type template argument of type 'int *' does not match template parameter of type 'int dr354::S::*'}}
-  //   cxx11-14-note@#dr354-ptr_mem {{template parameter is declared here}}
-  // since-cxx17-error@#dr354-m3 {{value of type 'int *' is not implicitly convertible to 'int dr354::S::*'}}
+  ptr_mem<(float S::*)0> m2; // #cwg354-m2
+  // cxx98-error@#cwg354-m2 {{non-type template argument of type 'float cwg354::S::*' cannot be converted to a value of type 'int cwg354::S::*'}}
+  //   cxx98-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-m2 {{null non-type template argument of type 'float cwg354::S::*' does not match template parameter of type 'int cwg354::S::*'}}
+  //   cxx11-14-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-m2 {{value of type 'float cwg354::S::*' is not implicitly convertible to 'int cwg354::S::*'}}
+  ptr_mem<(int *)0> m3; // #cwg354-m3
+  // cxx98-error@#cwg354-m3 {{non-type template argument of type 'int *' cannot be converted to a value of type 'int cwg354::S::*'}}
+  //   cxx98-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // cxx11-14-error@#cwg354-m3 {{null non-type template argument of type 'int *' does not match template parameter of type 'int cwg354::S::*'}}
+  //   cxx11-14-note@#cwg354-ptr_mem {{template parameter is declared here}}
+  // since-cxx17-error@#cwg354-m3 {{value of type 'int *' is not implicitly convertible to 'int cwg354::S::*'}}
 }
 
-struct dr355_S; // dr355: yes
-struct ::dr355_S {};
-// expected-warning@-1 {{extra qualification on member 'dr355_S'}}
-namespace dr355 { struct ::dr355_S s; }
+struct cwg355_S; // cwg355: yes
+struct ::cwg355_S {};
+// expected-warning@-1 {{extra qualification on member 'cwg355_S'}}
+namespace cwg355 { struct ::cwg355_S s; }
 
-// dr356: na
+// cwg356: na
 
-namespace dr357 { // dr357: yes
+namespace cwg357 { // cwg357: yes
   template<typename T> struct A {
-    void f() const; // #dr357-f
+    void f() const; // #cwg357-f
   };
   template<typename T> void A<T>::f() {}
   // expected-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'A<T>'}}
-  //   expected-note@#dr357-f {{member declaration does not match because it is const qualified}}
+  //   expected-note@#cwg357-f {{member declaration does not match because it is const qualified}}
 
   struct B {
     template<typename T> void f();
   };
   template<typename T> void B::f() const {}
-  // expected-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'dr357::B'}}
+  // expected-error@-1 {{out-of-line definition of 'f' does not match any declaration in 'cwg357::B'}}
 }
 
-namespace dr358 { // dr358: yes
-  extern "C" void dr358_f();
+namespace cwg358 { // cwg358: yes
+  extern "C" void cwg358_f();
   namespace N {
     int var;
-    extern "C" void dr358_f() { var = 10; }
+    extern "C" void cwg358_f() { var = 10; }
   }
 }
 
-namespace dr359 { // dr359: yes
+namespace cwg359 { // cwg359: yes
   // Note, the example in the DR is wrong; it doesn't contain an anonymous
   // union.
   struct E {
@@ -1049,7 +1049,7 @@ namespace dr359 { // dr359: yes
   };
 }
 
-namespace dr360 { // dr360: yes
+namespace cwg360 { // cwg360: yes
 struct A {
   int foo();
   int bar();
@@ -1060,28 +1060,28 @@ struct A {
 
 struct B : A {
 private:
-  using A::foo; // #dr360-using-foo
+  using A::foo; // #cwg360-using-foo
 protected:
-  using A::bar; // #dr360-using-bar
+  using A::bar; // #cwg360-using-bar
 public:
   using A::baz;
 };
 
 int main() {
   int foo = B().foo();
-  // expected-error@-1 {{'foo' is a private member of 'dr360::B'}}
-  //   expected-note@#dr360-using-foo {{declared private here}}
+  // expected-error@-1 {{'foo' is a private member of 'cwg360::B'}}
+  //   expected-note@#cwg360-using-foo {{declared private here}}
   int bar = B().bar();
-  // expected-error@-1 {{'bar' is a protected member of 'dr360::B'}}
-  //   expected-note@#dr360-using-bar {{declared protected here}}
+  // expected-error@-1 {{'bar' is a protected member of 'cwg360::B'}}
+  //   expected-note@#cwg360-using-bar {{declared protected here}}
   int baz = B().baz();
 }
-} // namespace dr360
+} // namespace cwg360
 
-// dr362: na
-// dr363: na
+// cwg362: na
+// cwg363: na
 
-namespace dr364 { // dr364: yes
+namespace cwg364 { // cwg364: yes
   struct S {
     static void f(int);
     void f(char);
@@ -1094,11 +1094,11 @@ namespace dr364 { // dr364: yes
   }
 }
 
-// dr366: yes
+// cwg366: yes
 #if "foo" // expected-error {{invalid token at start of a preprocessor expression}} 
 #endif
 
-namespace dr367 { // dr367: yes
+namespace cwg367 { // cwg367: yes
   // FIXME: These diagnostics are terrible. Don't diagnose an ill-formed global
   // array as being a VLA!
   int a[true ? throw 0 : 4];
@@ -1116,30 +1116,30 @@ namespace dr367 { // dr367: yes
   // cxx98-error@-2 {{variable length array folded to constant array as an extension}}
 }
 
-namespace dr368 { // dr368: 3.6
-  template<typename T, T> struct S {}; // #dr368-S
+namespace cwg368 { // cwg368: 3.6
+  template<typename T, T> struct S {}; // #cwg368-S
   template<typename T> int f(S<T, T()> *);
   // expected-error@-1 {{template argument for non-type template parameter is treated as function type 'T ()'}}
-  //   expected-note@#dr368-S {{template parameter is declared here}}
-  template<typename T> int g(S<T, (T())> *); // #dr368-g
-  template<typename T> int g(S<T, true ? T() : T()> *); // #dr368-g-2
+  //   expected-note@#cwg368-S {{template parameter is declared here}}
+  template<typename T> int g(S<T, (T())> *); // #cwg368-g
+  template<typename T> int g(S<T, true ? T() : T()> *); // #cwg368-g-2
   struct X {};
-  int n = g<X>(0); // #dr368-g-call
-  // cxx98-17-error@#dr368-g-call {{no matching function for call to 'g'}}
-  //   cxx98-17-note@#dr368-g {{candidate template ignored: substitution failure [with T = X]: a non-type template parameter cannot have type 'X' before C++20}}
-  //   cxx98-17-note@#dr368-g-2 {{candidate template ignored: substitution failure [with T = X]: a non-type template parameter cannot have type 'X' before C++20}}
-  // cxx20-23-error@#dr368-g-call {{call to 'g' is ambiguous}}
-  //   cxx20-23-note@#dr368-g {{candidate function [with T = dr368::X]}}
-  //   cxx20-23-note@#dr368-g-2 {{candidate function [with T = dr368::X]}}
+  int n = g<X>(0); // #cwg368-g-call
+  // cxx98-17-error@#cwg368-g-call {{no matching function for call to 'g'}}
+  //   cxx98-17-note@#cwg368-g {{candidate template ignored: substitution failure [with T = X]: a non-type template parameter cannot have type 'X' before C++20}}
+  //   cxx98-17-note@#cwg368-g-2 {{candidate template ignored: substitution failure [with T = X]: a non-type template parameter cannot have type 'X' before C++20}}
+  // cxx20-23-error@#cwg368-g-call {{call to 'g' is ambiguous}}
+  //   cxx20-23-note@#cwg368-g {{candidate function [with T = cwg368::X]}}
+  //   cxx20-23-note@#cwg368-g-2 {{candidate function [with T = cwg368::X]}}
 }
 
-// dr370: na
+// cwg370: na
 
-namespace dr372 { // dr372: no
+namespace cwg372 { // cwg372: no
   namespace example1 {
     template<typename T> struct X {
     protected:
-      typedef T Type; // #dr372-ex1-Type
+      typedef T Type; // #cwg372-ex1-Type
     };
     template<typename T> struct Y {};
 
@@ -1150,55 +1150,55 @@ namespace dr372 { // dr372: no
              template<typename> class T2> struct Z1 :
       T1<T>,
       T2<typename T1<T>::Type> {};
-      // expected-error@-1 {{'Type' is a protected member of 'dr372::example1::X<int>'}}
-      //   expected-note@#dr372-z1 {{in instantiation of template class 'dr372::example1::Z1<int, dr372::example1::X, dr372::example1::Y>' requested here}}
-      //   expected-note@#dr372-ex1-Type {{declared protected here}}
+      // expected-error@-1 {{'Type' is a protected member of 'cwg372::example1::X<int>'}}
+      //   expected-note@#cwg372-z1 {{in instantiation of template class 'cwg372::example1::Z1<int, cwg372::example1::X, cwg372::example1::Y>' requested here}}
+      //   expected-note@#cwg372-ex1-Type {{declared protected here}}
 
     template<typename T,
              template<typename> class T1,
              template<typename> class T2> struct Z2 :
       T2<typename T1<T>::Type>,
-      // expected-error@-1 {{'Type' is a protected member of 'dr372::example1::X<int>'}}
-      //   expected-note@#dr372-z2 {{in instantiation of template class 'dr372::example1::Z2<int, dr372::example1::X, dr372::example1::Y>' requested here}}
-      //   expected-note@#dr372-ex1-Type {{declared protected here}}
+      // expected-error@-1 {{'Type' is a protected member of 'cwg372::example1::X<int>'}}
+      //   expected-note@#cwg372-z2 {{in instantiation of template class 'cwg372::example1::Z2<int, cwg372::example1::X, cwg372::example1::Y>' requested here}}
+      //   expected-note@#cwg372-ex1-Type {{declared protected here}}
       T1<T> {};
 
-    Z1<int, X, Y> z1; // #dr372-z1
-    Z2<int, X, Y> z2; // #dr372-z2
+    Z1<int, X, Y> z1; // #cwg372-z1
+    Z2<int, X, Y> z2; // #cwg372-z2
   }
 
   namespace example2 {
     struct X {
     private:
-      typedef int Type; // #dr372-ex2-Type
+      typedef int Type; // #cwg372-ex2-Type
     };
     template<typename T> struct A {
       typename T::Type t;
-      // expected-error@-1 {{'Type' is a private member of 'dr372::example2::X'}}
-      //   expected-note@#dr372-ax {{in instantiation of template class 'dr372::example2::A<dr372::example2::X>' requested here}}
-      //   expected-note@#dr372-ex2-Type {{declared private here}}
+      // expected-error@-1 {{'Type' is a private member of 'cwg372::example2::X'}}
+      //   expected-note@#cwg372-ax {{in instantiation of template class 'cwg372::example2::A<cwg372::example2::X>' requested here}}
+      //   expected-note@#cwg372-ex2-Type {{declared private here}}
     };
-    A<X> ax; // #dr372-ax
+    A<X> ax; // #cwg372-ax
   }
 
   namespace example3 {
     struct A {
     protected:
-      typedef int N; // #dr372-N
+      typedef int N; // #cwg372-N
     };
 
     template<typename T> struct B {};
     template<typename U> struct C : U, B<typename U::N> {};
-    // expected-error@-1 {{'N' is a protected member of 'dr372::example3::A'}}
-    //   expected-note@#dr372-x {{in instantiation of template class 'dr372::example3::C<dr372::example3::A>' requested here}}
-    //   expected-note@#dr372-N {{declared protected here}}
+    // expected-error@-1 {{'N' is a protected member of 'cwg372::example3::A'}}
+    //   expected-note@#cwg372-x {{in instantiation of template class 'cwg372::example3::C<cwg372::example3::A>' requested here}}
+    //   expected-note@#cwg372-N {{declared protected here}}
     template<typename U> struct D : B<typename U::N>, U {};
-    // expected-error@-1 {{'N' is a protected member of 'dr372::example3::A'}}
-    //   expected-note@#dr372-y {{in instantiation of template class 'dr372::example3::D<dr372::example3::A>' requested here}}
-    //   expected-note@#dr372-N {{declared protected here}}
+    // expected-error@-1 {{'N' is a protected member of 'cwg372::example3::A'}}
+    //   expected-note@#cwg372-y {{in instantiation of template class 'cwg372::example3::D<cwg372::example3::A>' requested here}}
+    //   expected-note@#cwg372-N {{declared protected here}}
 
-    C<A> x; // #dr372-x
-    D<A> y; // #dr372-y
+    C<A> x; // #cwg372-x
+    D<A> y; // #cwg372-y
   }
 
   namespace example4 {
@@ -1219,49 +1219,49 @@ namespace dr372 { // dr372: no
   namespace std_example {
     class A {
     protected:
-      struct B {}; // #dr372-B-std
+      struct B {}; // #cwg372-B-std
     };
     struct D : A::B, A {};
-    // expected-error@-1 {{'B' is a protected member of 'dr372::std_example::A'}}
-    //   expected-note@#dr372-B-std {{declared protected here}}
+    // expected-error@-1 {{'B' is a protected member of 'cwg372::std_example::A'}}
+    //   expected-note@#cwg372-B-std {{declared protected here}}
   }
 
   // FIXME: This is valid: deriving from A::B gives access to A::B!
   namespace badwolf {
     class A {
     protected:
-      struct B; // #dr372-B
+      struct B; // #cwg372-B
     };
     struct A::B : A {};
     struct C : A::B {};
-    // expected-error@-1 {{'B' is a protected member of 'dr372::badwolf::A'}}
-    //   expected-note@#dr372-B {{declared protected here}}
+    // expected-error@-1 {{'B' is a protected member of 'cwg372::badwolf::A'}}
+    //   expected-note@#cwg372-B {{declared protected here}}
   }
 }
 
-namespace dr373 { // dr373: 5
-  namespace X { int dr373; }
-  struct dr373 { // #dr373-struct
+namespace cwg373 { // cwg373: 5
+  namespace X { int cwg373; }
+  struct cwg373 { // #cwg373-struct
     void f() {
-      using namespace dr373::X;
-      int k = dr373;
-      // expected-error@-1 {{'dr373' does not refer to a value}}
-      //   expected-note@#dr373-struct {{declared here}}
-      namespace Y = dr373::X;
-      k = Y::dr373;
+      using namespace cwg373::X;
+      int k = cwg373;
+      // expected-error@-1 {{'cwg373' does not refer to a value}}
+      //   expected-note@#cwg373-struct {{declared here}}
+      namespace Y = cwg373::X;
+      k = Y::cwg373;
     }
   };
 
-  struct A { struct B {}; }; // #dr373-A 
+  struct A { struct B {}; }; // #cwg373-A 
   namespace X = A::B;
   // expected-error@-1 {{expected namespace name}}
-  //   expected-note@#dr373-A {{'A' declared here}}
+  //   expected-note@#cwg373-A {{'A' declared here}}
   using namespace A::B;
   // expected-error@-1 {{expected namespace name}}
-  //   expected-note@#dr373-A {{'A' declared here}}
+  //   expected-note@#cwg373-A {{'A' declared here}}
 }
 
-namespace dr374 { // dr374: 7
+namespace cwg374 { // cwg374: 7
                   // NB 2.9 c++11
   namespace N {
     template<typename T> void f();
@@ -1272,10 +1272,10 @@ namespace dr374 { // dr374: 7
   template<> struct N::A<int> {};
 }
 
-// dr375: dup 345
-// dr376: na
+// cwg375: dup 345
+// cwg376: na
 
-namespace dr377 { // dr377: yes
+namespace cwg377 { // cwg377: yes
   enum E {
   // expected-error@-1 {{enumeration values exceed range of largest integer}}
     a = -__LONG_LONG_MAX__ - 1,
@@ -1286,10 +1286,10 @@ namespace dr377 { // dr377: yes
   };
 }
 
-// dr378: dup 276
-// dr379: na
+// cwg378: dup 276
+// cwg379: na
 
-namespace dr381 { // dr381: yes
+namespace cwg381 { // cwg381: yes
   struct A {
     int a;
   };
@@ -1301,26 +1301,26 @@ namespace dr381 { // dr381: yes
   void f() {
     E e;
     e.B::a = 0;
-    /* expected-error@-1 {{ambiguous conversion from derived class 'E' to base class 'dr381::B':
-    struct dr381::E -> C -> B
-    struct dr381::E -> D -> B}} */
+    /* expected-error@-1 {{ambiguous conversion from derived class 'E' to base class 'cwg381::B':
+    struct cwg381::E -> C -> B
+    struct cwg381::E -> D -> B}} */
     F f;
     f.A::a = 1;
   }
 }
 
-namespace dr382 { // dr382: yes c++11
+namespace cwg382 { // cwg382: yes c++11
   // FIXME: Should we allow this in C++98 mode?
   struct A { typedef int T; };
   typename A::T t;
   // cxx98-error@-1 {{'typename' occurs outside of a template}}
-  typename dr382::A a;
+  typename cwg382::A a;
   // cxx98-error@-1 {{'typename' occurs outside of a template}}
   typename A b;
   // expected-error@-1 {{expected a qualified name after 'typename'}}
 }
 
-namespace dr383 { // dr383: yes
+namespace cwg383 { // cwg383: yes
   struct A { A &operator=(const A&); };
   struct B { ~B(); };
   union C { C &operator=(const C&); };
@@ -1328,7 +1328,7 @@ namespace dr383 { // dr383: yes
   int check[(__is_pod(A) || __is_pod(B) || __is_pod(C) || __is_pod(D)) ? -1 : 1];
 }
 
-namespace dr384 { // dr384: yes
+namespace cwg384 { // cwg384: yes
   namespace N1 {
     template<typename T> struct Base {};
     template<typename T> struct X {
@@ -1350,35 +1350,35 @@ namespace dr384 { // dr384: yes
   }
 }
 
-namespace dr385 { // dr385: 2.8
+namespace cwg385 { // cwg385: 2.8
   struct A { protected: void f(); };
   struct B : A { using A::f; };
   struct C : A { void g(B b) { b.f(); } };
   void h(B b) { b.f(); }
 
-  struct D { int n; }; // #dr385-n
-  struct E : protected D {}; // #dr385-E
+  struct D { int n; }; // #cwg385-n
+  struct E : protected D {}; // #cwg385-E
   struct F : E { friend int i(E); };
   int i(E e) { return e.n; }
-  // expected-error@-1 {{'n' is a protected member of 'dr385::D'}}
-  //   expected-note@#dr385-E {{constrained by protected inheritance here}}
-  //   expected-note@#dr385-n {{member is declared here}}
+  // expected-error@-1 {{'n' is a protected member of 'cwg385::D'}}
+  //   expected-note@#cwg385-E {{constrained by protected inheritance here}}
+  //   expected-note@#cwg385-n {{member is declared here}}
 }
 
-namespace dr387 { // dr387: 2.8
+namespace cwg387 { // cwg387: 2.8
   namespace old {
     template<typename T> class number {
-      number(int); // #dr387-number-ctor
+      number(int); // #cwg387-number-ctor
       friend number gcd(number &x, number &y) {}
     };
 
     void g() {
       number<double> a(3);
-      // expected-error@-1 {{calling a private constructor of class 'dr387::old::number<double>'}}
-      //   expected-note@#dr387-number-ctor {{implicitly declared private here}}
+      // expected-error@-1 {{calling a private constructor of class 'cwg387::old::number<double>'}}
+      //   expected-note@#cwg387-number-ctor {{implicitly declared private here}}
       number<double> b(4);
-      // expected-error@-1 {{calling a private constructor of class 'dr387::old::number<double>'}}
-      //   expected-note@#dr387-number-ctor {{implicitly declared private here}}
+      // expected-error@-1 {{calling a private constructor of class 'cwg387::old::number<double>'}}
+      //   expected-note@#cwg387-number-ctor {{implicitly declared private here}}
       a = gcd(a, b);
       b = gcd(3, 4);
       // expected-error@-1 {{use of undeclared identifier 'gcd'}}
@@ -1401,14 +1401,14 @@ namespace dr387 { // dr387: 2.8
   }
 }
 
-// FIXME: dr388 needs libc++abi test
+// FIXME: cwg388 needs libc++abi test
 
-namespace dr389 { // dr389: no
+namespace cwg389 { // cwg389: no
   struct S {
     typedef struct {} A;
     typedef enum {} B;
-    typedef struct {} const C; // #dr389-C
-    typedef enum {} const D; // #dr389-D
+    typedef struct {} const C; // #cwg389-C
+    typedef enum {} const D; // #cwg389-D
   };
   template<typename> struct T {};
 
@@ -1422,8 +1422,8 @@ namespace dr389 { // dr389: no
   typedef void (*WithLinkage8)(WithLinkage2 WithLinkage1::*, WithLinkage5 *);
   typedef T<WithLinkage5> WithLinkage9;
 
-  typedef struct {} *WithoutLinkage1; // #dr389-no-link-1
-  typedef enum {} const WithoutLinkage2; // #dr389-no-link-2
+  typedef struct {} *WithoutLinkage1; // #cwg389-no-link-1
+  typedef enum {} const WithoutLinkage2; // #cwg389-no-link-2
   // These two types don't have linkage even though they are externally visible
   // and the ODR requires them to be merged across TUs.
   typedef S::C WithoutLinkage3;
@@ -1451,19 +1451,19 @@ namespace dr389 { // dr389: no
 
   typedef T<WithoutLinkage1> BadArg1;
   // expected-error@-1 {{template argument uses unnamed type}}
-  //   expected-note@#dr389-no-link-1 {{unnamed type used in template argument was declared here}}
+  //   expected-note@#cwg389-no-link-1 {{unnamed type used in template argument was declared here}}
   typedef T<WithoutLinkage2> BadArg2;
   // expected-error@-1 {{template argument uses unnamed type}}
-  //   expected-note@#dr389-no-link-2 {{unnamed type used in template argument was declared here}}
+  //   expected-note@#cwg389-no-link-2 {{unnamed type used in template argument was declared here}}
   typedef T<WithoutLinkage3> BadArg3;
   // expected-error@-1 {{template argument uses unnamed type}}
-  //   expected-note@#dr389-C {{unnamed type used in template argument was declared here}}
+  //   expected-note@#cwg389-C {{unnamed type used in template argument was declared here}}
   typedef T<WithoutLinkage4> BadArg4;
   // expected-error@-1 {{template argument uses unnamed type}}
-  //   expected-note@#dr389-D {{unnamed type used in template argument was declared here}}
+  //   expected-note@#cwg389-D {{unnamed type used in template argument was declared here}}
   typedef T<WithoutLinkage5> BadArg5;
   // expected-error@-1 {{template argument uses unnamed type}}
-  //   expected-note@#dr389-C {{unnamed type used in template argument was declared here}}
+  //   expected-note@#cwg389-C {{unnamed type used in template argument was declared here}}
 #endif
 
   extern WithLinkage1 withLinkage1;
@@ -1488,11 +1488,11 @@ namespace dr389 { // dr389: no
 
   // OK, extern "C".
   extern "C" {
-    extern WithoutLinkage1 dr389_withoutLinkage1;
-    extern WithoutLinkage2 dr389_withoutLinkage2;
-    extern WithoutLinkage3 dr389_withoutLinkage3;
-    extern WithoutLinkage4 dr389_withoutLinkage4;
-    extern WithoutLinkage5 dr389_withoutLinkage5;
+    extern WithoutLinkage1 cwg389_withoutLinkage1;
+    extern WithoutLinkage2 cwg389_withoutLinkage2;
+    extern WithoutLinkage3 cwg389_withoutLinkage3;
+    extern WithoutLinkage4 cwg389_withoutLinkage4;
+    extern WithoutLinkage5 cwg389_withoutLinkage5;
   }
 
   // OK, defined.
@@ -1511,9 +1511,9 @@ namespace dr389 { // dr389: no
     use(&withoutLinkage1); use(&withoutLinkage2); use(&withoutLinkage3);
     use(&withoutLinkage4); use(&withoutLinkage5);
 
-    use(&dr389_withoutLinkage1); use(&dr389_withoutLinkage2);
-    use(&dr389_withoutLinkage3); use(&dr389_withoutLinkage4);
-    use(&dr389_withoutLinkage5);
+    use(&cwg389_withoutLinkage1); use(&cwg389_withoutLinkage2);
+    use(&cwg389_withoutLinkage3); use(&cwg389_withoutLinkage4);
+    use(&cwg389_withoutLinkage5);
 
     use(&withoutLinkageDef1); use(&withoutLinkageDef2); use(&withoutLinkageDef3);
     use(&withoutLinkageDef4); use(&withoutLinkageDef5);
@@ -1525,39 +1525,39 @@ namespace dr389 { // dr389: no
   }
 }
 
-namespace dr390 { // dr390: 3.3
+namespace cwg390 { // cwg390: 3.3
   template<typename T>
   struct A {
     A() { f(); }
-    // expected-warning@-1 {{call to pure virtual member function 'f' has undefined behavior; overrides of 'f' in subclasses are not available in the constructor of 'dr390::A<int>'}}
-    //   expected-note@#dr390-A-int {{in instantiation of member function 'dr390::A<int>::A' requested here}}
-    //   expected-note@#dr390-f {{'f' declared here}}
-    virtual void f() = 0; // #dr390-f
+    // expected-warning@-1 {{call to pure virtual member function 'f' has undefined behavior; overrides of 'f' in subclasses are not available in the constructor of 'cwg390::A<int>'}}
+    //   expected-note@#cwg390-A-int {{in instantiation of member function 'cwg390::A<int>::A' requested here}}
+    //   expected-note@#cwg390-f {{'f' declared here}}
+    virtual void f() = 0; // #cwg390-f
     virtual ~A() = 0;
   };
   template<typename T> A<T>::~A() { T::error; }
   // expected-error@-1 {{type 'int' cannot be used prior to '::' because it has no members}}
-  //   expected-note@#dr390-A-int {{in instantiation of member function 'dr390::A<int>::~A' requested here}}
+  //   expected-note@#cwg390-A-int {{in instantiation of member function 'cwg390::A<int>::~A' requested here}}
   template<typename T> void A<T>::f() { T::error; } // ok, not odr-used
-  struct B : A<int> { // #dr390-A-int
+  struct B : A<int> { // #cwg390-A-int
     void f() {}
   } b;
 }
 
-namespace dr391 { // dr391: 2.8 c++11
+namespace cwg391 { // cwg391: 2.8 c++11
   // FIXME: Should this apply to C++98 too?
-  class A { A(const A&); }; // #dr391-A
+  class A { A(const A&); }; // #cwg391-A
   A fa();
   const A &a = fa();
-  // cxx98-error@-1 {{C++98 requires an accessible copy constructor for class 'dr391::A' when binding a reference to a temporary; was private}}
-  //   cxx98-note@#dr391-A {{implicitly declared private here}}
+  // cxx98-error@-1 {{C++98 requires an accessible copy constructor for class 'cwg391::A' when binding a reference to a temporary; was private}}
+  //   cxx98-note@#cwg391-A {{implicitly declared private here}}
 
-  struct B { B(const B&) = delete; }; // #dr391-B
+  struct B { B(const B&) = delete; }; // #cwg391-B
   // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   B fb();
   const B &b = fb();
   // cxx98-error@-1 {{copying variable of type 'B' invokes deleted constructor}}
-  //   cxx98-note@#dr391-B {{'B' has been explicitly marked deleted here}}
+  //   cxx98-note@#cwg391-B {{'B' has been explicitly marked deleted here}}
 
   template<typename T>
   struct C {
@@ -1567,9 +1567,9 @@ namespace dr391 { // dr391: 2.8 c++11
   const C<int> &c = fc();
 }
 
-// dr392 is in dr392.cpp
+// cwg392 is in cwg392.cpp
 
-namespace dr393 { // dr393: 2.7
+namespace cwg393 { // cwg393: 2.7
 template <typename T>
 struct S {};
 
@@ -1585,11 +1585,11 @@ void h() {
   T (*p) [] = (U(*)[])0;
   sp_assert_convertible<T[]>( (U(*)[])0 );
 }
-} // namespace dr393
+} // namespace cwg393
 
-// dr394: na
+// cwg394: na
 
-namespace dr395 { // dr395: 3.0
+namespace cwg395 { // cwg395: 3.0
   struct S {
     template <typename T, int N>(&operator T())[N];
     // expected-error@-1 {{cannot specify any part of a return type in the declaration of a conversion function}}
@@ -1613,13 +1613,13 @@ namespace dr395 { // dr395: 3.0
     };
 
     template <class T, class U>
-    operator typename ptr_mem_fun_t<T, U>::type() const { // #dr395-conv-func
+    operator typename ptr_mem_fun_t<T, U>::type() const { // #cwg395-conv-func
       return 0;
     }
   } null1;
   int (S::*p)() = null1;
-  // expected-error@-1 {{no viable conversion from 'struct null1_t' to 'int (dr395::S::*)()'}}
-  //   expected-note@#dr395-conv-func {{candidate template ignored: couldn't infer template argument 'T'}}
+  // expected-error@-1 {{no viable conversion from 'struct null1_t' to 'int (cwg395::S::*)()'}}
+  //   expected-note@#cwg395-conv-func {{candidate template ignored: couldn't infer template argument 'T'}}
 
   template <typename T> using id = T;
   // cxx98-error@-1 {{alias declarations are a C++11 extension}}
@@ -1637,22 +1637,22 @@ namespace dr395 { // dr395: 3.0
   int (S::*q)() = null2;
 }
 
-namespace dr396 { // dr396: yes
+namespace cwg396 { // cwg396: yes
   void f() {
     auto int a();
     // since-cxx11-error@-1 {{'auto' storage class specifier is not permitted in C++11, and will not be supported in future releases}}
     // expected-error@-2 {{illegal storage class on function}}
-    int (i); // #dr396-i
+    int (i); // #cwg396-i
     auto int (i);
     // since-cxx11-error@-1 {{'auto' storage class specifier is not permitted in C++11, and will not be supported in future releases}}
     // expected-error@-2 {{redefinition of 'i'}}
-    //   expected-note@#dr396-i {{previous definition is here}}
+    //   expected-note@#cwg396-i {{previous definition is here}}
   }
 }
 
-// dr397: sup 1823
+// cwg397: sup 1823
 
-namespace dr398 { // dr398: yes
+namespace cwg398 { // cwg398: yes
   namespace example1 {
     struct S {
       static int const I = 42;
@@ -1666,9 +1666,9 @@ namespace dr398 { // dr398: yes
   namespace example2 {
     template <int I> struct X {};
     template <template <class T> class> struct Z {};
-    template <class T> void f(typename T::Y *) {} // #dr398-f
-    template <class T> void g(X<T::N> *) {} // #dr398-g
-    template <class T> void h(Z<T::template TT> *) {} // #dr398-h
+    template <class T> void f(typename T::Y *) {} // #cwg398-f
+    template <class T> void g(X<T::N> *) {} // #cwg398-g
+    template <class T> void h(Z<T::template TT> *) {} // #cwg398-h
     struct A {};
     struct B {
       int Y;
@@ -1683,23 +1683,23 @@ namespace dr398 { // dr398: yes
     void test() {
       f<A>(0);
       // expected-error@-1 {{no matching function for call to 'f'}}
-      //   expected-note@#dr398-f {{candidate template ignored: substitution failure [with T = A]: no type named 'Y' in 'dr398::example2::A'}}
+      //   expected-note@#cwg398-f {{candidate template ignored: substitution failure [with T = A]: no type named 'Y' in 'cwg398::example2::A'}}
       f<B>(0);
       // expected-error@-1 {{no matching function for call to 'f'}}
-      //   expected-note@#dr398-f {{candidate template ignored: substitution failure [with T = B]: typename specifier refers to non-type member 'Y' in 'dr398::example2::B'}}
+      //   expected-note@#cwg398-f {{candidate template ignored: substitution failure [with T = B]: typename specifier refers to non-type member 'Y' in 'cwg398::example2::B'}}
       g<C>(0);
       // expected-error@-1 {{no matching function for call to 'g'}}
-      //   expected-note@#dr398-g {{candidate template ignored: substitution failure [with T = C]: missing 'typename' prior to dependent type name 'C::N'}}
+      //   expected-note@#cwg398-g {{candidate template ignored: substitution failure [with T = C]: missing 'typename' prior to dependent type name 'C::N'}}
       h<D>(0);
       // expected-error@-1 {{no matching function for call to 'h'}}
-      //   expected-note@#dr398-h {{candidate template ignored: substitution failure [with T = D]: 'TT' following the 'template' keyword does not refer to a template}}
+      //   expected-note@#cwg398-h {{candidate template ignored: substitution failure [with T = D]: 'TT' following the 'template' keyword does not refer to a template}}
     }
   }
 }
 
-namespace dr399 { // dr399: 11
-                  // NB: reuse dr244 test
-  struct B {}; // #dr399-B
+namespace cwg399 { // cwg399: 11
+                  // NB: reuse cwg244 test
+  struct B {}; // #cwg399-B
   struct D : B {};
 
   D D_object;
@@ -1708,18 +1708,18 @@ namespace dr399 { // dr399: 11
 
   void f() {
     D_object.~B();
-    // expected-error@-1 {{destructor type 'dr399::B' in object destruction expression does not match the type 'D' of the object being destroyed}}
-    //   expected-note@#dr399-B {{type 'dr399::B' found by destructor name lookup}}
+    // expected-error@-1 {{destructor type 'cwg399::B' in object destruction expression does not match the type 'D' of the object being destroyed}}
+    //   expected-note@#cwg399-B {{type 'cwg399::B' found by destructor name lookup}}
     D_object.B::~B();
     D_object.D::~B(); // FIXME: Missing diagnostic for this.
     B_ptr->~B();
     B_ptr->~B_alias();
     B_ptr->B_alias::~B();
     B_ptr->B_alias::~B_alias();
-    B_ptr->dr399::~B();
-    // expected-error@-1 {{qualified member access refers to a member in namespace 'dr399'}}
-    B_ptr->dr399::~B_alias();
-    // expected-error@-1 {{qualified member access refers to a member in namespace 'dr399'}}
+    B_ptr->cwg399::~B();
+    // expected-error@-1 {{qualified member access refers to a member in namespace 'cwg399'}}
+    B_ptr->cwg399::~B_alias();
+    // expected-error@-1 {{qualified member access refers to a member in namespace 'cwg399'}}
   }
 
   template<typename T, typename U>
@@ -1731,10 +1731,10 @@ namespace dr399 { // dr399: 11
     B_ptr->~B_alias();
     B_ptr->B_alias::~B();
     B_ptr->B_alias::~B_alias();
-    B_ptr->dr399::~B();
-    // expected-error@-1 {{'dr399' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
-    B_ptr->dr399::~B_alias();
-    // expected-error@-1 {{'dr399' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
+    B_ptr->cwg399::~B();
+    // expected-error@-1 {{'cwg399' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
+    B_ptr->cwg399::~B_alias();
+    // expected-error@-1 {{'cwg399' does not refer to a type name in pseudo-destructor expression; expected the name of type 'T'}}
   }
   template void f<B, D>(B*, D);
 
@@ -1743,7 +1743,7 @@ namespace dr399 { // dr399: 11
     typedef E<int> F;
   }
   void g(N::F f) {
-    typedef N::F G; // #dr399-G
+    typedef N::F G; // #cwg399-G
     f.~G();
     f.G::~E();
     // expected-error@-1 {{ISO C++ requires the name after '::~' to be found in the same scope as the name before '::~'}}
@@ -1762,7 +1762,7 @@ namespace dr399 { // dr399: 11
     // Rejecting this seems correct, but most compilers accept, so we do also.
     f.N::F::~G();
     // expected-error@-1 {{qualified destructor name only found in lexical scope; omit the qualifier to find this type name by unqualified lookup}}
-    //   expected-note@#dr399-G {{type 'G' (aka 'E<int>') found by destructor name lookup}}
+    //   expected-note@#cwg399-G {{type 'G' (aka 'E<int>') found by destructor name lookup}}
   }
 
   // Bizarrely, compilers perform lookup in the scope for qualified destructor
@@ -1774,17 +1774,17 @@ namespace dr399 { // dr399: 11
     }
     template <typename U> void f(typename N::S<U>::Inner *p) {
       typedef typename N::S<U>::Inner T;
-      p->::dr399::QualifiedLookupInScope::N::S<U>::Inner::~T();
-      // expected-error@-1 {{no type named 'T' in 'dr399::QualifiedLookupInScope::N::S<int>'}}
-      //   expected-note@#dr399-f {{in instantiation of function template specialization 'dr399::QualifiedLookupInScope::f<int>' requested here}}
+      p->::cwg399::QualifiedLookupInScope::N::S<U>::Inner::~T();
+      // expected-error@-1 {{no type named 'T' in 'cwg399::QualifiedLookupInScope::N::S<int>'}}
+      //   expected-note@#cwg399-f {{in instantiation of function template specialization 'cwg399::QualifiedLookupInScope::f<int>' requested here}}
     }
-    template void f<int>(N::S<int>::Inner *); // #dr399-f
+    template void f<int>(N::S<int>::Inner *); // #cwg399-f
 
     template <typename U> void g(U *p) {
       typedef U T;
       p->T::~T();
       p->U::~T();
-      p->::dr399::QualifiedLookupInScope::N::S<int>::Inner::~T();
+      p->::cwg399::QualifiedLookupInScope::N::S<int>::Inner::~T();
       // expected-error@-1 {{'T' does not refer to a type name in pseudo-destructor expression; expected the name of type 'U'}}
     }
     template void g(N::S<int>::Inner *);
diff --git a/clang/test/CXX/drs/dr4xx.cpp b/clang/test/CXX/drs/dr4xx.cpp
index 9b7a9777202f2..611b791470785 100644
--- a/clang/test/CXX/drs/dr4xx.cpp
+++ b/clang/test/CXX/drs/dr4xx.cpp
@@ -11,56 +11,56 @@ __extension__ typedef __SIZE_TYPE__ size_t;
 
 namespace std { struct type_info; }
 
-namespace dr400 { // dr400: yes
-  struct A { int a; struct a {}; }; // #dr400-A
-  struct B { int a; struct a {}; }; // #dr400-B
+namespace cwg400 { // cwg400: yes
+  struct A { int a; struct a {}; }; // #cwg400-A
+  struct B { int a; struct a {}; }; // #cwg400-B
   struct C : A, B { using A::a; struct a b; };
   struct D : A, B { 
     using A::a;
     // FIXME: we should issue a single diagnostic
-    using B::a; // #dr400-using-B-a
-    // expected-error@#dr400-using-B-a {{target of using declaration conflicts with declaration already in scope}}
-    //   expected-note@#dr400-B {{target of using declaration}}
-    //   expected-note@#dr400-A {{conflicting declaration}}
-    // expected-error@#dr400-using-B-a {{target of using declaration conflicts with declaration already in scope}}
-    //   expected-note@#dr400-B {{target of using declaration}}
-    //   expected-note@#dr400-A {{conflicting declaration}}
+    using B::a; // #cwg400-using-B-a
+    // expected-error@#cwg400-using-B-a {{target of using declaration conflicts with declaration already in scope}}
+    //   expected-note@#cwg400-B {{target of using declaration}}
+    //   expected-note@#cwg400-A {{conflicting declaration}}
+    // expected-error@#cwg400-using-B-a {{target of using declaration conflicts with declaration already in scope}}
+    //   expected-note@#cwg400-B {{target of using declaration}}
+    //   expected-note@#cwg400-A {{conflicting declaration}}
     struct a b;
   };
   struct E : A, B { struct a b; };
   // expected-error@-1 {{member 'a' found in multiple base classes of different types}}
-  //   expected-note@#dr400-A {{member type 'dr400::A::a' found by ambiguous name lookup}}
-  //   expected-note@#dr400-B {{member type 'dr400::B::a' found by ambiguous name lookup}}
-}
-
-namespace dr401 { // dr401: 2.8
-  template<class T, class U = typename T::type> class A : public T {}; // #dr401-A
-  // expected-error@#dr401-A {{'type' is a private member of 'dr401::C'}}
-  //   expected-note@#dr402-friend-A-C {{in instantiation of default argument for 'A<C>' required here}}
-  //   expected-note@#dr402-C-type {{implicitly declared private here}}
-  // expected-error@#dr401-A {{'type' is a protected member of 'dr401::B'}}
-  //   expected-note@#dr402-b {{in instantiation of default argument for 'A<B>' required here}}
-  //   expected-note@#dr402-B-type {{declared protected here}}
-  // expected-error@#dr401-A {{'type' is a private member of 'dr401::D'}}
-  //   expected-note@#dr402-d {{in instantiation of default argument for 'A<D>' required here}}
-  //   expected-note@#dr402-D-type {{implicitly declared private here}}
+  //   expected-note@#cwg400-A {{member type 'cwg400::A::a' found by ambiguous name lookup}}
+  //   expected-note@#cwg400-B {{member type 'cwg400::B::a' found by ambiguous name lookup}}
+}
+
+namespace cwg401 { // cwg401: 2.8
+  template<class T, class U = typename T::type> class A : public T {}; // #cwg401-A
+  // expected-error@#cwg401-A {{'type' is a private member of 'cwg401::C'}}
+  //   expected-note@#cwg402-friend-A-C {{in instantiation of default argument for 'A<C>' required here}}
+  //   expected-note@#cwg402-C-type {{implicitly declared private here}}
+  // expected-error@#cwg401-A {{'type' is a protected member of 'cwg401::B'}}
+  //   expected-note@#cwg402-b {{in instantiation of default argument for 'A<B>' required here}}
+  //   expected-note@#cwg402-B-type {{declared protected here}}
+  // expected-error@#cwg401-A {{'type' is a private member of 'cwg401::D'}}
+  //   expected-note@#cwg402-d {{in instantiation of default argument for 'A<D>' required here}}
+  //   expected-note@#cwg402-D-type {{implicitly declared private here}}
   class B {
   protected:
-    typedef int type; // #dr402-B-type
+    typedef int type; // #cwg402-B-type
   };
 
   class C {
-    typedef int type; // #dr402-C-type
-    friend class A<C>; // #dr402-friend-A-C
+    typedef int type; // #cwg402-C-type
+    friend class A<C>; // #cwg402-friend-A-C
   };
 
   class D {
-    typedef int type; // #dr402-D-type
+    typedef int type; // #cwg402-D-type
     friend class A<D, int>;
   };
 
-  A<B> *b; // #dr402-b
-  A<D> *d; // #dr402-d
+  A<B> *b; // #cwg402-b
+  A<D> *d; // #cwg402-d
 
   struct E {
     template<class T, class U = typename T::type> class A : public T {};
@@ -73,18 +73,18 @@ namespace dr401 { // dr401: 2.8
 
   // FIXME: Why do we get different diagnostics in C++11 onwards here? We seem
   // to not treat the default template argument as a SFINAE context in C++98.
-  template<class T, class U = typename T::type> void f(T) {} // #dr402-f
+  template<class T, class U = typename T::type> void f(T) {} // #cwg402-f
   // cxx98-error@-1 {{default template arguments for a function template are a C++11 extension}}
-  // cxx98-error@-2 {{'type' is a protected member of 'dr401::B'}}
+  // cxx98-error@-2 {{'type' is a protected member of 'cwg401::B'}}
   //   cxx98-note@-3 {{in instantiation of default argument for 'f<B>' required here}}
-  //   cxx98-note@#dr402-f-b {{while substituting deduced template arguments into function template 'f' [with T = B, U = (no value)]}}
-  //   cxx98-note@#dr402-B-type {{declared protected here}}
-  void g(B b) { f(b); } // #dr402-f-b
+  //   cxx98-note@#cwg402-f-b {{while substituting deduced template arguments into function template 'f' [with T = B, U = (no value)]}}
+  //   cxx98-note@#cwg402-B-type {{declared protected here}}
+  void g(B b) { f(b); } // #cwg402-f-b
   // since-cxx11-error@-1 {{no matching function for call to 'f'}}
-  //   since-cxx11-note@#dr402-f {{candidate template ignored: substitution failure [with T = B, U = typename B::type]: 'type' is a protected member of 'dr401::B'}}
+  //   since-cxx11-note@#cwg402-f {{candidate template ignored: substitution failure [with T = B, U = typename B::type]: 'type' is a protected member of 'cwg401::B'}}
 }
 
-namespace dr403 { // dr403: yes
+namespace cwg403 { // cwg403: yes
   namespace A {
     struct S {};
     int f(void*);
@@ -98,10 +98,10 @@ namespace dr403 { // dr403: yes
                 // template-id as we can make it.
 }
 
-// dr404: na
+// cwg404: na
 // (NB: also sup 594)
 
-namespace dr405 { // dr405: yes
+namespace cwg405 { // cwg405: yes
                   // NB: also dup 218
   namespace A {
     struct S {};
@@ -145,7 +145,7 @@ namespace dr405 { // dr405: yes
   // expected-error@-1 {{use of undeclared identifier 'f'}}
 }
 
-namespace dr406 { // dr406: 2.9
+namespace cwg406 { // cwg406: 2.9
   typedef struct {
     static int n;
     // expected-error@-1 {{static data member 'n' not allowed in anonymous struct}}
@@ -156,35 +156,35 @@ namespace dr406 { // dr406: 2.9
   } B;
 }
 
-namespace dr407 { // dr407: 3.8
-                  // NB: reused by dr1894 and dr2199
+namespace cwg407 { // cwg407: 3.8
+                  // NB: reused by cwg1894 and cwg2199
   struct S;
   typedef struct S S;
   void f() {
     struct S *p;
     {
-      typedef struct S S; // #dr407-typedef-S
+      typedef struct S S; // #cwg407-typedef-S
       struct S *p;
       // expected-error@-1 {{typedef 'S' cannot be referenced with a struct specifier}}
-      //   expected-note@#dr407-typedef-S {{declared here}}
+      //   expected-note@#cwg407-typedef-S {{declared here}}
     }
   }
   struct S {};
 
   namespace UsingDir {
     namespace A {
-      struct S {}; // #dr407-A-S
+      struct S {}; // #cwg407-A-S
     }
     namespace B {
-      typedef int S; // #dr407-B-S
+      typedef int S; // #cwg407-B-S
     }
     namespace C {
       using namespace A;
       using namespace B;
       struct S s;
       // expected-error@-1 {{ambiguous}}
-      //   expected-note@#dr407-A-S {{candidate found by name lookup is 'dr407::UsingDir::A::S'}}
-      //   expected-note@#dr407-B-S {{candidate found by name lookup is 'dr407::UsingDir::B::S'}}
+      //   expected-note@#cwg407-A-S {{candidate found by name lookup is 'cwg407::UsingDir::A::S'}}
+      //   expected-note@#cwg407-B-S {{candidate found by name lookup is 'cwg407::UsingDir::B::S'}}
     }
     namespace D {
       using A::S;
@@ -193,7 +193,7 @@ namespace dr407 { // dr407: 3.8
     }
     namespace E {
       // The standard doesn't say whether this is valid. We interpret
-      // DR407 as meaning "if lookup finds both a tag and a typedef with the
+      // CWG407 as meaning "if lookup finds both a tag and a typedef with the
       // same type, then it's OK in an elaborated-type-specifier".
       typedef A::S S;
       using A::S;
@@ -216,7 +216,7 @@ namespace dr407 { // dr407: 3.8
   }
 }
 
-namespace dr408 { // dr408: 3.4
+namespace cwg408 { // cwg408: 3.4
   template<int N> void g() { int arr[N != 1 ? 1 : -1]; }
   template<> void g<2>() { }
 
@@ -245,7 +245,7 @@ namespace dr408 { // dr408: 3.4
   template void R<int>::f();
 }
 
-namespace dr409 { // dr409: yes
+namespace cwg409 { // cwg409: yes
   template<typename T> struct A {
     typedef int B;
     B b1;
@@ -256,7 +256,7 @@ namespace dr409 { // dr409: yes
   };
 }
 
-namespace dr410 { // dr410: no
+namespace cwg410 { // cwg410: no
   template<class T> void f(T);
   void g(int);
   namespace M {
@@ -269,7 +269,7 @@ namespace dr410 { // dr410: no
       template<class T> void i(T);
       friend void i<>(int);
     private:
-      static void z(); // #dr410-z
+      static void z(); // #cwg410-z
     };
 
     template<> void h(int) { A::z(); }
@@ -279,13 +279,13 @@ namespace dr410 { // dr410: no
   }
   template<> void f(int) { M::A::z(); }
   void g(int) { M::A::z(); }
-  // expected-error@-1 {{'z' is a private member of 'dr410::M::A'}}
-  //   expected-note@#dr410-z {{declared private here}}
+  // expected-error@-1 {{'z' is a private member of 'cwg410::M::A'}}
+  //   expected-note@#cwg410-z {{declared private here}}
 }
 
-// dr412 is in dr412.cpp
+// cwg412 is in cwg412.cpp
 
-namespace dr413 { // dr413: yes
+namespace cwg413 { // cwg413: yes
   struct S {
     int a;
     int : 17;
@@ -295,7 +295,7 @@ namespace dr413 { // dr413: yes
   // expected-error@-1 {{excess elements in struct initializer}}
 
   struct E {};
-  struct T { // #dr413-T
+  struct T { // #cwg413-T
     int a;
     E e;
     int b;
@@ -303,10 +303,10 @@ namespace dr413 { // dr413: yes
   T t1 = { 1, {}, 2 };
   T t2 = { 1, 2 };
   // expected-error@-1 {{initializer for aggregate with no elements requires explicit braces}}
-  //   expected-note@#dr413-T {{'dr413::T' declared here}}
+  //   expected-note@#cwg413-T {{'cwg413::T' declared here}}
 }
 
-namespace dr414 { // dr414: dup 305
+namespace cwg414 { // cwg414: dup 305
   struct X {};
   void f() {
     X x;
@@ -315,13 +315,13 @@ namespace dr414 { // dr414: dup 305
   }
 }
 
-namespace dr415 { // dr415: yes
+namespace cwg415 { // cwg415: yes
   template<typename T> void f(T, ...) { T::error; }
   void f(int, int);
   void g() { f(0, 0); } // ok
 }
 
-namespace dr416 { // dr416: yes
+namespace cwg416 { // cwg416: yes
   extern struct A a;
   int &operator+(const A&, const A&);
   int &k = a + a;
@@ -329,14 +329,14 @@ namespace dr416 { // dr416: yes
   float &f = a + a;
 }
 
-namespace dr417 { // dr417: no
+namespace cwg417 { // cwg417: no
   struct A;
-  struct dr417::A {};
+  struct cwg417::A {};
   // expected-warning@-1 {{extra qualification on member 'A'}}
   struct B { struct X; };
   struct C : B {};
   struct C::X {};
-  // expected-error@-1 {{no struct named 'X' in 'dr417::C'}}
+  // expected-error@-1 {{no struct named 'X' in 'cwg417::C'}}
   struct B::X { struct Y; };
   struct C::X::Y {}; // ok!
   namespace N {
@@ -347,24 +347,24 @@ namespace dr417 { // dr417: no
   }
   // FIXME: This is ill-formed.
   using N::D;
-  struct dr417::D {};
+  struct cwg417::D {};
   // expected-warning@-1 {{extra qualification on member 'D'}}
   using namespace N;
-  struct dr417::E {};
-  // expected-error@-1 {{no struct named 'E' in namespace 'dr417'}}
+  struct cwg417::E {};
+  // expected-error@-1 {{no struct named 'E' in namespace 'cwg417'}}
   // expected-warning@-2 {{extra qualification on member 'E'}}
   struct N::F {};
   struct G;
   using N::H;
   namespace M {
-    struct dr417::G {};
-    // expected-error@-1 {{cannot define or redeclare 'G' here because namespace 'M' does not enclose namespace 'dr417'}}
-    struct dr417::H {};
-    // expected-error@-1 {{cannot define or redeclare 'H' here because namespace 'M' does not enclose namespace 'dr417'}}
+    struct cwg417::G {};
+    // expected-error@-1 {{cannot define or redeclare 'G' here because namespace 'M' does not enclose namespace 'cwg417'}}
+    struct cwg417::H {};
+    // expected-error@-1 {{cannot define or redeclare 'H' here because namespace 'M' does not enclose namespace 'cwg417'}}
   }
 }
 
-namespace dr418 { // dr418: no
+namespace cwg418 { // cwg418: no
 namespace example1 {
 void f1(int, int = 0);
 void f1(int = 0, int);
@@ -374,7 +374,7 @@ void g() { f1(); }
 
 namespace example2 {
 namespace A {
-void f2(int); // #dr418-f2
+void f2(int); // #cwg418-f2
 }
 namespace B {
 using A::f2;
@@ -386,7 +386,7 @@ void g2() {
   using B::f2;
   f2();
   // expected-error@-1 {{no matching function for call to 'f2'}}
-  //   expected-note@#dr418-f2 {{candidate function not viable: requires 1 argument, but 0 were provided}}
+  //   expected-note@#cwg418-f2 {{candidate function not viable: requires 1 argument, but 0 were provided}}
 }
 } // namespace example2
 
@@ -407,9 +407,9 @@ void use() {
   f(); // FIXME: this should fail
 }
 } // namespace example3
-} // namespace dr418
+} // namespace cwg418
 
-namespace dr420 { // dr420: 9
+namespace cwg420 { // cwg420: 9
   template<typename T> struct ptr {
     T *operator->() const;
     T &operator*() const;
@@ -453,15 +453,15 @@ namespace dr420 { // dr420: 9
     q->~id<int>();
     p->id<int>::~id<int>();
     q->id<int>::~id<int>();
-    p->template id<int>::~id<int>(); // OK since dr2292
-    q->template id<int>::~id<int>(); // OK since dr2292
+    p->template id<int>::~id<int>(); // OK since cwg2292
+    q->template id<int>::~id<int>(); // OK since cwg2292
     p->A::template id<int>::~id<int>();
     q->A::template id<int>::~id<int>();
   }
 #endif
 }
 
-namespace dr421 { // dr421: yes
+namespace cwg421 { // cwg421: yes
   struct X { X(); int n; int &r; };
   int *p = &X().n;
   // cxx98-error@-1 {{taking the address of a temporary object of type 'int'}}
@@ -469,32 +469,32 @@ namespace dr421 { // dr421: yes
   int *q = &X().r;
 }
 
-namespace dr422 { // dr422: yes
+namespace cwg422 { // cwg422: yes
   template<typename T, typename U> void f() {
-    typedef T type; // #dr422-typedef-T
+    typedef T type; // #cwg422-typedef-T
     typedef U type;
     // expected-error@-1 {{typedef redefinition with different types ('char' vs 'int')}}
-    //   expected-note@#dr422-f-int-char {{in instantiation of function template specialization 'dr422::f<int, char>' requested here}}
-    //   expected-note@#dr422-typedef-T {{previous definition is here}}
+    //   expected-note@#cwg422-f-int-char {{in instantiation of function template specialization 'cwg422::f<int, char>' requested here}}
+    //   expected-note@#cwg422-typedef-T {{previous definition is here}}
   }
   template void f<int, int>();
-  template void f<int, char>(); // #dr422-f-int-char
+  template void f<int, char>(); // #cwg422-f-int-char
 }
 
-namespace dr423 { // dr423: yes
+namespace cwg423 { // cwg423: yes
   template<typename T> struct X { operator T&(); };
   void f(X<int> x) { x += 1; }
 }
 
-namespace dr424 { // dr424: yes
+namespace cwg424 { // cwg424: yes
   struct A {
-    typedef int N; // #dr424-N
+    typedef int N; // #cwg424-N
     typedef int N;
     // expected-error@-1 {{redefinition of 'N'}}
-    //   expected-note@#dr424-N {{previous definition is here}}
+    //   expected-note@#cwg424-N {{previous definition is here}}
 
     struct X;
-    typedef X X; // #dr424-X
+    typedef X X; // #cwg424-X
     struct X {};
 
     struct X *p;
@@ -503,20 +503,20 @@ namespace dr424 { // dr424: yes
 
     typedef X X;
     // expected-error@-1 {{redefinition of 'X'}}
-    //   expected-note@#dr424-X {{previous definition is here}}
+    //   expected-note@#cwg424-X {{previous definition is here}}
   };
   struct B {
     typedef int M;
   };
   struct C : B {
-    typedef int M; // #dr424-M
+    typedef int M; // #cwg424-M
     typedef int M;
     // expected-error@-1 {{redefinition of 'M'}}
-    //   expected-note@#dr424-M {{previous definition is here}}
+    //   expected-note@#cwg424-M {{previous definition is here}}
   };
 }
 
-namespace dr425 { // dr425: yes
+namespace cwg425 { // cwg425: yes
   struct A { template<typename T> operator T() const; } a;
   float f = 1.0f * a;
   // expected-error@-1 {{use of overloaded operator '*' is ambiguous (with operand types 'float' and 'struct A')}}
@@ -532,10 +532,10 @@ namespace dr425 { // dr425: yes
   float g = 1.0f * b; // ok
 }
 
-namespace dr427 { // dr427: yes
+namespace cwg427 { // cwg427: yes
   struct B {};
   struct D : public B {
-    D(B &) = delete; // #dr427-D
+    D(B &) = delete; // #cwg427-D
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   };
 
@@ -545,12 +545,12 @@ namespace dr427 { // dr427: yes
   const D &d3 = (const D&)b;
   const D &d4(b);
   // expected-error@-1 {{conversion function from 'B' to 'const D' invokes a deleted function}}
-  //   expected-note@#dr427-D {{'D' has been explicitly marked deleted here}}
+  //   expected-note@#cwg427-D {{'D' has been explicitly marked deleted here}}
 }
 
-namespace dr428 { // dr428: yes
+namespace cwg428 { // cwg428: yes
   template<typename T> T make();
-  extern struct X x; // #dr428-X
+  extern struct X x; // #cwg428-X
   void f() {
     throw void();
     // expected-error@-1 {{cannot throw}}
@@ -558,30 +558,30 @@ namespace dr428 { // dr428: yes
     throw make<const volatile void*>();
     throw x;
     // expected-error@-1 {{cannot throw}}
-    //   expected-note@#dr428-X {{forward declaration of 'dr428::X'}}
+    //   expected-note@#cwg428-X {{forward declaration of 'cwg428::X'}}
     throw make<X&>();
     // expected-error@-1 {{cannot throw}}
-    //   expected-note@#dr428-X {{forward declaration of 'dr428::X'}}
+    //   expected-note@#cwg428-X {{forward declaration of 'cwg428::X'}}
     throw make<X*>();
     // expected-error@-1 {{cannot throw}}
-    //   expected-note@#dr428-X {{forward declaration of 'dr428::X'}}
+    //   expected-note@#cwg428-X {{forward declaration of 'cwg428::X'}}
     throw make<const volatile X&>();
     // expected-error@-1 {{cannot throw}}
-    //   expected-note@#dr428-X {{forward declaration of 'dr428::X'}}
+    //   expected-note@#cwg428-X {{forward declaration of 'cwg428::X'}}
     throw make<const volatile X*>();
     // expected-error@-1 {{cannot throw}}
-    //   expected-note@#dr428-X {{forward declaration of 'dr428::X'}}
+    //   expected-note@#cwg428-X {{forward declaration of 'cwg428::X'}}
   }
 }
 
-namespace dr429 { // dr429: 2.8 c++11
+namespace cwg429 { // cwg429: 2.8 c++11
   // FIXME: This rule is obviously intended to apply to C++98 as well.
   struct A {
     static void *operator new(size_t, size_t);
-    static void operator delete(void*, size_t); // #dr429-delete
+    static void operator delete(void*, size_t); // #cwg429-delete
   } *a = new (0) A;
   // since-cxx11-error@-1 {{'new' expression with placement arguments refers to non-placement 'operator delete'}}
-  //   since-cxx11-note@#dr429-delete {{here}}
+  //   since-cxx11-note@#cwg429-delete {{here}}
   struct B {
     static void *operator new(size_t, size_t);
     static void operator delete(void*);
@@ -589,7 +589,7 @@ namespace dr429 { // dr429: 2.8 c++11
   } *b = new (0) B; // ok, second delete is not a non-placement deallocation function
 }
 
-namespace dr430 { // dr430: yes c++11
+namespace cwg430 { // cwg430: yes c++11
   // resolved by n2239
   // FIXME: This should apply in C++98 too.
   void f(int n) {
@@ -598,7 +598,7 @@ namespace dr430 { // dr430: yes c++11
   }
 }
 
-namespace dr431 { // dr431: yes
+namespace cwg431 { // cwg431: yes
   struct A {
     template<typename T> T *get();
     template<typename T> struct B {
@@ -630,7 +630,7 @@ namespace dr431 { // dr431: yes
   }
 }
 
-namespace dr432 { // dr432: 3.0
+namespace cwg432 { // cwg432: 3.0
   template<typename T> struct A {};
   template<typename T> struct B : A<B> {};
   // expected-error@-1 {{use of class template 'B' requires template arguments}}
@@ -643,7 +643,7 @@ namespace dr432 { // dr432: 3.0
 #endif
 }
 
-namespace dr433 { // dr433: yes
+namespace cwg433 { // cwg433: yes
   template<class T> struct S {
     void f(union U*);
   };
@@ -653,7 +653,7 @@ namespace dr433 { // dr433: yes
   S<int> s;
 }
 
-namespace dr434 { // dr434: sup 2352
+namespace cwg434 { // cwg434: sup 2352
   void f() {
     const int ci = 0;
     int *pi = 0;
@@ -671,16 +671,16 @@ namespace dr434 { // dr434: sup 2352
 #endif
 }
 
-// dr435: na
+// cwg435: na
 
-namespace dr436 { // dr436: yes
-  enum E { f }; // #dr436-f
+namespace cwg436 { // cwg436: yes
+  enum E { f }; // #cwg436-f
   void f();
   // expected-error@-1 {{redefinition of 'f' as different kind of symbol}}
-  //   expected-note@#dr436-f {{previous definition is here}}
+  //   expected-note@#cwg436-f {{previous definition is here}}
 }
 
-namespace dr437 { // dr437: sup 1308
+namespace cwg437 { // cwg437: sup 1308
   // This is superseded by 1308, which is in turn superseded by 1330,
   // which restores this rule.
   template<typename U> struct T : U {};
@@ -699,46 +699,46 @@ namespace dr437 { // dr437: sup 1308
   };
 }
 
-// dr438 is in dr438.cpp
-// dr439 is in dr439.cpp
-// dr441 is in dr441.cpp
-// dr442: sup 348
-// dr443: na
+// cwg438 is in cwg438.cpp
+// cwg439 is in cwg439.cpp
+// cwg441 is in cwg441.cpp
+// cwg442: sup 348
+// cwg443: na
 
-namespace dr444 { // dr444: yes
+namespace cwg444 { // cwg444: yes
   struct D;
-  struct B { // #dr444-B
-    D &operator=(D &) = delete; // #dr444-deleted
+  struct B { // #cwg444-B
+    D &operator=(D &) = delete; // #cwg444-deleted
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   };
-  struct D : B { // #dr444-D
+  struct D : B { // #cwg444-D
     using B::operator=;
   } extern d;
   void f() {
     d = d;
     // expected-error@-1 {{overload resolution selected deleted operator '='}}
-    //   expected-note@#dr444-deleted {{candidate function has been explicitly deleted}}
-    //   expected-note@#dr444-D {{candidate function (the implicit copy assignment operator)}}
-    //   expected-note@#dr444-B {{candidate function (the implicit copy assignment operator)}}
-    //   since-cxx11-note@#dr444-B {{candidate function (the implicit move assignment operator) not viable: expects an rvalue for 1st argument}}
-    //   since-cxx11-note@#dr444-D {{candidate function (the implicit move assignment operator) not viable: expects an rvalue for 1st argument}}
+    //   expected-note@#cwg444-deleted {{candidate function has been explicitly deleted}}
+    //   expected-note@#cwg444-D {{candidate function (the implicit copy assignment operator)}}
+    //   expected-note@#cwg444-B {{candidate function (the implicit copy assignment operator)}}
+    //   since-cxx11-note@#cwg444-B {{candidate function (the implicit move assignment operator) not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg444-D {{candidate function (the implicit move assignment operator) not viable: expects an rvalue for 1st argument}}
   }
 }
 
-namespace dr445 { // dr445: 3.2
-  class A { void f(); }; // #dr445-f
+namespace cwg445 { // cwg445: 3.2
+  class A { void f(); }; // #cwg445-f
   struct B {
     friend void A::f();
-    // expected-error@-1 {{friend function 'f' is a private member of 'dr445::A'}}
-    //   expected-note@#dr445-f {{implicitly declared private here}}
+    // expected-error@-1 {{friend function 'f' is a private member of 'cwg445::A'}}
+    //   expected-note@#cwg445-f {{implicitly declared private here}}
   };
 }
 
-namespace dr446 { // dr446: 2.8
+namespace cwg446 { // cwg446: 2.8
   struct C;
   struct A {
     A();
-    A(const A&) = delete; // #dr446-deleted
+    A(const A&) = delete; // #cwg446-deleted
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
     A(const C&);
   };
@@ -747,28 +747,28 @@ namespace dr446 { // dr446: 2.8
     void(b ? a : a);
     b ? A() : a;
     // expected-error@-1 {{call to deleted constructor of 'A'}}
-    //   expected-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   expected-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
     b ? a : A();
     // expected-error@-1 {{call to deleted constructor of 'A'}}
-    //   expected-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   expected-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
     b ? A() : A();
     // cxx98-14-error@-1 {{call to deleted constructor of 'A'}}
-    //   expected-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   expected-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
 
     void(b ? a : c);
     b ? a : C();
     // expected-error@-1 {{call to deleted constructor of 'A'}}
-    //   cxx98-14-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   cxx98-14-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
     b ? c : A();
     // cxx98-14-error@-1 {{call to deleted constructor of 'A'}}
-    //   cxx98-14-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   cxx98-14-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
     b ? A() : C();
     // cxx98-14-error@-1 {{call to deleted constructor of 'A'}}
-    //   cxx98-14-note@#dr446-deleted {{'A' has been explicitly marked deleted here}}
+    //   cxx98-14-note@#cwg446-deleted {{'A' has been explicitly marked deleted here}}
   }
 }
 
-namespace dr447 { // dr447: yes
+namespace cwg447 { // cwg447: yes
   struct A { int n; int a[4]; };
   template<int> struct U {
     typedef int type;
@@ -796,60 +796,60 @@ namespace dr447 { // dr447: yes
   }
 }
 
-namespace dr448 { // dr448: 2.8
-  template<typename T = int> void f(int); // #dr448-f-int
+namespace cwg448 { // cwg448: 2.8
+  template<typename T = int> void f(int); // #cwg448-f-int
   // cxx98-error@-1 {{default template arguments for a function template are a C++11 extension}}
   template<typename T> void g(T t) {
     f<T>(t);
     // expected-error@-1 {{call to function 'f' that is neither visible in the template definition nor found by argument-dependent lookup}}
-    //   expected-note@#dr448-g {{in instantiation of function template specialization 'dr448::g<dr448::HideFromADL::X>' requested here}}
-    //   expected-note@#dr448-f-T {{'f' should be declared prior to the call site or in namespace 'dr448::HideFromADL'}}
-    dr448::f(t);
+    //   expected-note@#cwg448-g {{in instantiation of function template specialization 'cwg448::g<cwg448::HideFromADL::X>' requested here}}
+    //   expected-note@#cwg448-f-T {{'f' should be declared prior to the call site or in namespace 'cwg448::HideFromADL'}}
+    cwg448::f(t);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr448-f-int {{candidate function template not viable: no known conversion from 'dr448::HideFromADL::X' to 'int' for 1st argument}}
+    //   expected-note@#cwg448-f-int {{candidate function template not viable: no known conversion from 'cwg448::HideFromADL::X' to 'int' for 1st argument}}
   }
-  template<typename T> void f(T); // #dr448-f-T
+  template<typename T> void f(T); // #cwg448-f-T
   namespace HideFromADL { struct X {}; }
   template void g(int); // ok
-  template void g(HideFromADL::X); // #dr448-g
+  template void g(HideFromADL::X); // #cwg448-g
 }
 
-// dr449: na
+// cwg449: na
 
-namespace dr450 { // dr450: yes
+namespace cwg450 { // cwg450: yes
   typedef int A[3];
   void f1(const A &);
-  void f2(A &); // #dr450-f2
+  void f2(A &); // #cwg450-f2
   struct S { A n; };
   void g() {
     f1(S().n);
     f2(S().n);
     // expected-error@-1 {{no matching function for call to 'f2'}}}
-    //   expected-note@#dr450-f2 {{candidate function not viable: expects an lvalue for 1st argument}}
+    //   expected-note@#cwg450-f2 {{candidate function not viable: expects an lvalue for 1st argument}}
   }
 #if __cplusplus >= 201103L
   void h() {
     f1(A{});
     f2(A{});
     // expected-error@-1 {{no matching function for call to 'f2'}}}
-    //   expected-note@#dr450-f2 {{candidate function not viable: expects an lvalue for 1st argument}}
+    //   expected-note@#cwg450-f2 {{candidate function not viable: expects an lvalue for 1st argument}}
   }
 #endif
 }
 
-namespace dr451 { // dr451: yes
+namespace cwg451 { // cwg451: yes
   const int a = 1 / 0;
   // expected-warning@-1 {{division by zero is undefined}}
-  const int b = 1 / 0; // #dr451-b
+  const int b = 1 / 0; // #cwg451-b
   // expected-warning@-1 {{division by zero is undefined}}
-  int arr[b]; // #dr451-arr
+  int arr[b]; // #cwg451-arr
   // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
   //   expected-note@-2 {{initializer of 'b' is not a constant expression}}
-  //   expected-note@#dr451-b {{declared here}}
-  // expected-error@#dr451-arr {{variable length array declaration not allowed at file scope}}
+  //   expected-note@#cwg451-b {{declared here}}
+  // expected-error@#cwg451-arr {{variable length array declaration not allowed at file scope}}
 }
 
-namespace dr452 { // dr452: yes
+namespace cwg452 { // cwg452: yes
   struct A {
     int a, b, c;
     A *p;
@@ -858,9 +858,9 @@ namespace dr452 { // dr452: yes
   };
 }
 
-// dr454 FIXME write a codegen test
+// cwg454 FIXME write a codegen test
 
-namespace dr456 { // dr456: yes
+namespace cwg456 { // cwg456: yes
   // sup 903 c++11
   const int null = 0;
   void *p = null;
@@ -873,7 +873,7 @@ namespace dr456 { // dr456: yes
   // since-cxx11-error@-2 {{cannot initialize a variable of type 'void *' with an lvalue of type 'const bool'}}
 }
 
-namespace dr457 { // dr457: yes
+namespace cwg457 { // cwg457: yes
   const int a = 1;
   const volatile int b = 1;
   int ax[a];
@@ -890,7 +890,7 @@ namespace dr457 { // dr457: yes
   };
 }
 
-namespace dr458 { // dr458: 11
+namespace cwg458 { // cwg458: 11
   struct A {
     int T;
     int f();
@@ -906,11 +906,11 @@ namespace dr458 { // dr458: 11
   int A::f() {
     return T;
   }
-  template<typename T> // #dr458-g-T
+  template<typename T> // #cwg458-g-T
   int A::g() {
     return T;
     // expected-error@-1 {{'T' does not refer to a value}}
-    //   expected-note@#dr458-g-T {{declared here}}
+    //   expected-note@#cwg458-g-T {{declared here}}
   }
 
   template<typename T>
@@ -923,33 +923,33 @@ namespace dr458 { // dr458: 11
     return T;
   }
   template<typename U>
-  template<typename T> // #dr458-h-T
+  template<typename T> // #cwg458-h-T
   int B<U>::h() {
     return T;
     // expected-error@-1 {{'T' does not refer to a value}}
-    //   expected-note@#dr458-h-T {{declared here}}
+    //   expected-note@#cwg458-h-T {{declared here}}
   }
 }
 
-namespace dr460 { // dr460: yes
+namespace cwg460 { // cwg460: yes
   namespace X { namespace Q { int n; } }
   namespace Y {
     using X;
     // expected-error@-1 {{using declaration requires a qualified name}}
-    using dr460::X;
+    using cwg460::X;
     // expected-error@-1 {{using declaration cannot refer to a namespace}}
     using X::Q;
     // expected-error@-1 {{using declaration cannot refer to a namespace}}
   }
 }
 
-// dr461: na
-// dr462 is in dr462.cpp
-// dr463: na
-// dr464: na
-// dr465: na
+// cwg461: na
+// cwg462 is in cwg462.cpp
+// cwg463: na
+// cwg464: na
+// cwg465: na
 
-namespace dr466 { // dr466: 2.8
+namespace cwg466 { // cwg466: 2.8
 typedef int I;
 typedef const int CI;
 typedef volatile int VI;
@@ -980,7 +980,7 @@ void g(int a, CI b, VI c) {
 }
 }
 
-namespace dr467 { // dr467: yes
+namespace cwg467 { // cwg467: yes
   int stuff();
 
   int f() {
@@ -995,51 +995,51 @@ namespace dr467 { // dr467: yes
   int g() {
     goto later;
     // expected-error@-1 {{cannot jump from this goto statement to its label}}
-    //   expected-note@#dr467-k {{jump bypasses variable initialization}}
-    int k = stuff(); // #dr467-k
+    //   expected-note@#cwg467-k {{jump bypasses variable initialization}}
+    int k = stuff(); // #cwg467-k
   later:
     return k;
   }
 }
 
-namespace dr468 { // dr468: yes c++11
+namespace cwg468 { // cwg468: yes c++11
   // FIXME: Should we allow this in C++98 too?
   template<typename> struct A {
     template<typename> struct B {
       static int C;
     };
   };
-  int k = dr468::template A<int>::template B<char>::C;
+  int k = cwg468::template A<int>::template B<char>::C;
   // cxx98-error@-1 {{'template' keyword outside of a template}}
   // cxx98-error@-2 {{'template' keyword outside of a template}}
 }
 
-namespace dr469 { // dr469: no
-  template<typename T> struct X; // #dr469-X
+namespace cwg469 { // cwg469: no
+  template<typename T> struct X; // #cwg469-X
   template<typename T> struct X<const T> {};
   X<int&> x;
-  // expected-error@-1 {{implicit instantiation of undefined template 'dr469::X<int &>'}}
-  //   expected-note@#dr469-X {{template is declared here}}
+  // expected-error@-1 {{implicit instantiation of undefined template 'cwg469::X<int &>'}}
+  //   expected-note@#cwg469-X {{template is declared here}}
 }
 
-namespace dr470 { // dr470: yes
+namespace cwg470 { // cwg470: yes
   template<typename T> struct A {
     struct B {};
   };
   template<typename T> struct C {
   };
 
-  template struct A<int>; // #dr470-A-int
+  template struct A<int>; // #cwg470-A-int
   template struct A<int>::B;
   // expected-error@-1 {{duplicate explicit instantiation of 'B'}}
-  //   expected-note@#dr470-A-int {{previous explicit instantiation is here}}
+  //   expected-note@#cwg470-A-int {{previous explicit instantiation is here}}
 
   // ok, instantiating C<char> doesn't instantiate base class members.
   template struct A<char>;
   template struct C<char>;
 }
 
-namespace dr471 { // dr471: 2.8
+namespace cwg471 { // cwg471: 2.8
   struct A { int n; };
   struct B : private virtual A {};
   struct C : protected virtual A {};
@@ -1050,50 +1050,50 @@ namespace dr471 { // dr471: 2.8
   struct F : E, B { int f() { return n; } };
   struct G : virtual A {
   private:
-    using A::n; // #dr471-G-using
+    using A::n; // #cwg471-G-using
   };
   struct H : B, G { int f() { return n; } };
-  // expected-error@-1 {{'n' is a private member of 'dr471::G'}}
-  //   expected-note@#dr471-G-using {{declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg471::G'}}
+  //   expected-note@#cwg471-G-using {{declared private here}}
 }
 
-namespace dr472 { // dr472: no drafting 2011-04
+namespace cwg472 { // cwg472: no drafting 2011-04
 struct B {
-  int i; // #dr472-i
+  int i; // #cwg472-i
 };
-struct I : protected B {}; // #dr472-struct-I
+struct I : protected B {}; // #cwg472-struct-I
 struct D : public I {
   void f(I *ip) {
     ip->i = 0;
-    // expected-error@-1 {{'i' is a protected member of 'dr472::B'}}
-    //   expected-note@#dr472-struct-I {{constrained by protected inheritance here}}
-    //   expected-note@#dr472-i {{member is declared here}}
+    // expected-error@-1 {{'i' is a protected member of 'cwg472::B'}}
+    //   expected-note@#cwg472-struct-I {{constrained by protected inheritance here}}
+    //   expected-note@#cwg472-i {{member is declared here}}
     B *bp = ip;
     bp->i = 5;
   }
 };
 }
 
-namespace dr474 { // dr474: 3.4
+namespace cwg474 { // cwg474: 3.4
   namespace N {
     struct S {
       void f();
     };
   }
   void N::S::f() {
-    void g(); // #dr474-g
+    void g(); // #cwg474-g
   }
   int g();
   namespace N {
     int g();
     // expected-error@-1 {{functions that differ only in their return type cannot be overloaded}}
-    //   expected-note@#dr474-g {{previous declaration is here}}
+    //   expected-note@#cwg474-g {{previous declaration is here}}
   }
 }
 
-// dr475 FIXME write a libc++abi test
+// cwg475 FIXME write a libc++abi test
 
-namespace dr477 { // dr477: 3.5
+namespace cwg477 { // cwg477: 3.5
   struct A {
     explicit A();
     virtual void f();
@@ -1110,56 +1110,56 @@ namespace dr477 { // dr477: 3.5
   // expected-error@-1 {{can only be specified inside the class definition}}
 }
 
-namespace dr478 { // dr478: yes
-  struct A { virtual void f() = 0; }; // #dr478-f
+namespace cwg478 { // cwg478: yes
+  struct A { virtual void f() = 0; }; // #cwg478-f
   void f(A *a);
   void f(A a[10]);
   // expected-error@-1 {{array of abstract class type 'A'}}
-  //   expected-note@#dr478-f {{unimplemented pure virtual method 'f' in 'A'}}
+  //   expected-note@#cwg478-f {{unimplemented pure virtual method 'f' in 'A'}}
 }
 
-namespace dr479 { // dr479: 2.8
+namespace cwg479 { // cwg479: 2.8
   struct S {
     S();
   private:
-    S(const S&); // #dr479-S-copy-ctor
-    ~S(); // #dr479-S-dtor
+    S(const S&); // #cwg479-S-copy-ctor
+    ~S(); // #cwg479-S-dtor
   };
   void f() {
     throw S();
     // expected-error@-1 {{temporary of type 'S' has private destructor}}
-    //   expected-note@#dr479-S-dtor {{declared private here}}
+    //   expected-note@#cwg479-S-dtor {{declared private here}}
     // expected-error@-3 {{exception object of type 'S' has private destructor}}
-    //   expected-note@#dr479-S-dtor {{declared private here}}
-    // cxx98-error@-5 {{C++98 requires an accessible copy constructor for class 'dr479::S' when binding a reference to a temporary; was private}}
-    //   cxx98-note@#dr479-S-copy-ctor {{declared private here}}
-    // cxx98-14-error@-7 {{calling a private constructor of class 'dr479::S'}}
-    //   cxx98-14-note@#dr479-S-copy-ctor {{declared private here}}
+    //   expected-note@#cwg479-S-dtor {{declared private here}}
+    // cxx98-error@-5 {{C++98 requires an accessible copy constructor for class 'cwg479::S' when binding a reference to a temporary; was private}}
+    //   cxx98-note@#cwg479-S-copy-ctor {{declared private here}}
+    // cxx98-14-error@-7 {{calling a private constructor of class 'cwg479::S'}}
+    //   cxx98-14-note@#cwg479-S-copy-ctor {{declared private here}}
   }
   void g() {
     S s;
     // expected-error@-1 {{variable of type 'S' has private destructor}}
-    //   expected-note@#dr479-S-dtor {{declared private here}}
+    //   expected-note@#cwg479-S-dtor {{declared private here}}
     throw s;
     // expected-error@-1 {{exception object of type 'S' has private destructor}}
-    //   expected-note@#dr479-S-dtor {{declared private here}}
-    // expected-error@-3 {{calling a private constructor of class 'dr479::S'}}
-    //   expected-note@#dr479-S-copy-ctor {{declared private here}}
+    //   expected-note@#cwg479-S-dtor {{declared private here}}
+    // expected-error@-3 {{calling a private constructor of class 'cwg479::S'}}
+    //   expected-note@#cwg479-S-copy-ctor {{declared private here}}
   }
   void h() {
     try {
       f();
       g();
     } catch (S s) {
-      // expected-error@-1 {{calling a private constructor of class 'dr479::S'}}
-      //   expected-note@#dr479-S-copy-ctor {{declared private here}}
+      // expected-error@-1 {{calling a private constructor of class 'cwg479::S'}}
+      //   expected-note@#cwg479-S-copy-ctor {{declared private here}}
       // expected-error@-3 {{variable of type 'S' has private destructor}}
-      //   expected-note@#dr479-S-dtor {{declared private here}}
+      //   expected-note@#cwg479-S-dtor {{declared private here}}
     }
   }
 }
 
-namespace dr480 { // dr480: yes
+namespace cwg480 { // cwg480: yes
   struct A { int n; };
   struct B : A {};
   struct C : virtual B {};
@@ -1167,24 +1167,24 @@ namespace dr480 { // dr480: yes
 
   int A::*a = &A::n;
   int D::*b = a;
-  // expected-error@-1 {{conversion from pointer to member of class 'dr480::A' to pointer to member of class 'dr480::D' via virtual base 'dr480::B' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg480::A' to pointer to member of class 'cwg480::D' via virtual base 'cwg480::B' is not allowed}}
 
   extern int D::*c;
   int A::*d = static_cast<int A::*>(c);
-  // expected-error@-1 {{conversion from pointer to member of class 'dr480::D' to pointer to member of class 'dr480::A' via virtual base 'dr480::B' is not allowed}}
+  // expected-error@-1 {{conversion from pointer to member of class 'cwg480::D' to pointer to member of class 'cwg480::A' via virtual base 'cwg480::B' is not allowed}}
 
   D *e;
   A *f = e;
   D *g = static_cast<D*>(f);
-  // expected-error@-1 {{cannot cast 'dr480::A *' to 'D *' via virtual base 'dr480::B'}}
+  // expected-error@-1 {{cannot cast 'cwg480::A *' to 'D *' via virtual base 'cwg480::B'}}
 
   extern D &i;
   A &j = i;
   D &k = static_cast<D&>(j);
-  // expected-error@-1 {{cannot cast 'A' to 'D &' via virtual base 'dr480::B'}}
+  // expected-error@-1 {{cannot cast 'A' to 'D &' via virtual base 'cwg480::B'}}
 }
 
-namespace dr481 { // dr481: 2.8
+namespace cwg481 { // cwg481: 2.8
   template<class T, T U> class A { T *x; };
   T *x;
   // expected-error@-1 {{unknown type name 'T'}}
@@ -1199,15 +1199,15 @@ namespace dr481 { // dr481: 2.8
 
   template<typename A = C, typename C = A> struct E {
     void f() {
-      typedef ::dr481::C c; // #dr481-c
+      typedef ::cwg481::C c; // #cwg481-c
       typedef C c;
-      // expected-error@-1 {{typedef redefinition with different types ('int' vs '::dr481::C')}}
-      //   expected-note@#dr481-E-int {{in instantiation of member function 'dr481::E<int>::f' requested here}}
-      //   expected-note@#dr481-c {{previous definition is here}}
+      // expected-error@-1 {{typedef redefinition with different types ('int' vs '::cwg481::C')}}
+      //   expected-note@#cwg481-E-int {{in instantiation of member function 'cwg481::E<int>::f' requested here}}
+      //   expected-note@#cwg481-c {{previous definition is here}}
     }
   };
   template struct E<>; // ok
-  template struct E<int>; // #dr481-E-int
+  template struct E<int>; // #cwg481-E-int
 
   template<template<typename U_no_typo_correction> class A,
            A<int> *B,
@@ -1229,13 +1229,13 @@ namespace dr481 { // dr481: 2.8
   I<123, char*, J> *j;
 }
 
-namespace dr482 { // dr482: 3.5
+namespace cwg482 { // cwg482: 3.5
   extern int a;
   void f();
 
-  int dr482::a = 0;
+  int cwg482::a = 0;
   // expected-warning@-1 {{extra qualification on member 'a'}}
-  void dr482::f() {}
+  void cwg482::f() {}
   // expected-warning@-1 {{extra qualification on member 'f'}}
 
   inline namespace X {
@@ -1244,19 +1244,19 @@ namespace dr482 { // dr482: 3.5
     void g();
     struct S;
   }
-  int dr482::b = 0;
+  int cwg482::b = 0;
   // expected-warning@-1 {{extra qualification on member 'b'}}
-  void dr482::g() {}
+  void cwg482::g() {}
   // expected-warning@-1 {{extra qualification on member 'g'}}
-  struct dr482::S {};
+  struct cwg482::S {};
   // expected-warning@-1 {{extra qualification on member 'S'}}
 
-  void dr482::f();
+  void cwg482::f();
   // expected-warning@-1 {{extra qualification on member 'f'}}
-  void dr482::g();
+  void cwg482::g();
   // expected-warning@-1 {{extra qualification on member 'g'}}
 
-  // FIXME: The following are valid in DR482's wording, but these are bugs in
+  // FIXME: The following are valid in CWG482's wording, but these are bugs in
   // the wording which we deliberately don't implement.
   namespace N { typedef int type; }
   typedef int N::type;
@@ -1274,7 +1274,7 @@ namespace dr482 { // dr482: 3.5
   };
 }
 
-namespace dr483 { // dr483: yes
+namespace cwg483 { // cwg483: yes
   namespace climits {
     int check1[__SCHAR_MAX__ >= 127 ? 1 : -1];
     int check2[__SHRT_MAX__ >= 32767 ? 1 : -1];
@@ -1293,7 +1293,7 @@ namespace dr483 { // dr483: yes
   }
 }
 
-namespace dr484 { // dr484: yes
+namespace cwg484 { // cwg484: yes
   struct A {
     A();
     void f();
@@ -1312,18 +1312,18 @@ namespace dr484 { // dr484: yes
   };
 
   struct C;
-  typedef C CT; // #dr484-typedef-CT
+  typedef C CT; // #cwg484-typedef-CT
   struct CT {};
   // expected-error@-1 {{definition of type 'CT' conflicts with typedef of the same name}}
-  //   expected-note@#dr484-typedef-CT {{'CT' declared here}}
+  //   expected-note@#cwg484-typedef-CT {{'CT' declared here}}
 
   namespace N {
     struct D;
-    typedef D DT; // #dr484-typedef-DT
+    typedef D DT; // #cwg484-typedef-DT
   }
   struct N::DT {};
   // expected-error@-1 {{definition of type 'DT' conflicts with typedef of the same name}}
-  //   expected-note@#dr484-typedef-DT {{'DT' declared here}}
+  //   expected-note@#cwg484-typedef-DT {{'DT' declared here}}
 
   typedef struct {
     S();
@@ -1331,7 +1331,7 @@ namespace dr484 { // dr484: yes
   } S;
 }
 
-namespace dr485 { // dr485: yes
+namespace cwg485 { // cwg485: yes
   namespace N {
     struct S {};
     int operator+(S, S);
@@ -1344,8 +1344,8 @@ namespace dr485 { // dr485: yes
   int b = f<int>(s);
 }
 
-namespace dr486 { // dr486: yes
-  template<typename T> T f(T *); // #dr486-f
+namespace cwg486 { // cwg486: yes
+  template<typename T> T f(T *); // #cwg486-f
   int &f(...);
 
   void g();
@@ -1356,24 +1356,24 @@ namespace dr486 { // dr486: yes
     int &b = f(&n);
     f<void()>(&g);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr486-f {{candidate template ignored: substitution failure [with T = void ()]: function cannot return function type 'void ()'}}
+    //   expected-note@#cwg486-f {{candidate template ignored: substitution failure [with T = void ()]: function cannot return function type 'void ()'}}
     f<int[10]>(&n);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr486-f {{candidate template ignored: substitution failure [with T = int[10]]: function cannot return array type 'int[10]'}}
+    //   expected-note@#cwg486-f {{candidate template ignored: substitution failure [with T = int[10]]: function cannot return array type 'int[10]'}}
   }
 }
 
-namespace dr487 { // dr487: yes
+namespace cwg487 { // cwg487: yes
   enum E { e };
-  int operator+(int, E); // #dr487-operator-plus
-  int i[4 + e]; // #dr487-i
+  int operator+(int, E); // #cwg487-operator-plus
+  int i[4 + e]; // #cwg487-i
   // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
   //   since-cxx11-note@-2 {{non-constexpr function 'operator+' cannot be used in a constant expression}}
-  //   since-cxx11-note@#dr487-operator-plus {{declared here}}
-  // expected-error@#dr487-i {{variable length array declaration not allowed at file scope}}
+  //   since-cxx11-note@#cwg487-operator-plus {{declared here}}
+  // expected-error@#cwg487-i {{variable length array declaration not allowed at file scope}}
 }
 
-namespace dr488 { // dr488: yes c++11
+namespace cwg488 { // cwg488: yes c++11
   template <typename T> void f(T);
   void f(int);
   void g() {
@@ -1387,19 +1387,19 @@ namespace dr488 { // dr488: yes c++11
   }
 }
 
-// dr489: na
+// cwg489: na
 
-namespace dr490 { // dr490: 2.8
+namespace cwg490 { // cwg490: 2.8
   template<typename T> struct X {};
 
   struct A {
     typedef int T;
-    struct K {}; // #dr490-k
+    struct K {}; // #cwg490-k
 
     int f(T);
     int g(T);
     int h(X<T>);
-    int X<T>::*i(); // #dr490-i
+    int X<T>::*i(); // #cwg490-i
     int K::*j();
 
     template<typename T> T k();
@@ -1417,11 +1417,11 @@ namespace dr490 { // dr490: 2.8
     // FIXME: Per this DR, these two are valid! That is another defect
     // (no number yet...) which will eventually supersede this one.
     friend int X<T>::*A::i();
-    // expected-error@-1 {{return type of out-of-line definition of 'dr490::A::i' differs from that in the declaration}}
-    //   expected-note@#dr490-i {{previous declaration is here}}
+    // expected-error@-1 {{return type of out-of-line definition of 'cwg490::A::i' differs from that in the declaration}}
+    //   expected-note@#cwg490-i {{previous declaration is here}}
     friend int K::*A::j();
     // expected-error@-1 {{use of undeclared identifier 'K'; did you mean 'A::K'?}}
-    //   expected-note@#dr490-k {{'A::K' declared here}}
+    //   expected-note@#cwg490-k {{'A::K' declared here}}
 
     // ok, lookup finds B::T, not A::T, so return type matches
     friend char A::k<T>();
@@ -1433,15 +1433,15 @@ namespace dr490 { // dr490: 2.8
   };
 }
 
-namespace dr491 { // dr491: dup 413
+namespace cwg491 { // cwg491: dup 413
   struct A {} a, b[3] = { a, {} };
   A c[2] = { a, {}, b[1] };
   // expected-error@-1 {{excess elements in array initializer}}
 }
 
-// dr492 is in dr492.cpp
+// cwg492 is in cwg492.cpp
 
-namespace dr493 { // dr493: dup 976
+namespace cwg493 { // cwg493: dup 976
   struct X {
     template <class T> operator const T &() const;
   };
@@ -1451,7 +1451,7 @@ namespace dr493 { // dr493: dup 976
   }
 }
 
-namespace dr494 { // dr494: dup 372
+namespace cwg494 { // cwg494: dup 372
   class A {
     class B {};
     friend class C;
@@ -1464,7 +1464,7 @@ namespace dr494 { // dr494: dup 372
   };
 }
 
-namespace dr495 { // dr495: 3.5
+namespace cwg495 { // cwg495: 3.5
   template<typename T>
   struct S {
     operator int() { return T::error; }
@@ -1482,11 +1482,11 @@ namespace dr495 { // dr495: 3.5
   long n2 = s2;
 }
 
-namespace dr496 { // dr496: sup 2094
+namespace cwg496 { // cwg496: sup 2094
   struct A { int n; };
   struct B { volatile int n; };
   int check1[ __is_trivially_copyable(const int) ? 1 : -1];
-  // This checks the dr2094 behavior, not dr496
+  // This checks the cwg2094 behavior, not cwg496
   int check2[ __is_trivially_copyable(volatile int) ? 1 : -1];
   int check3[ __is_trivially_constructible(A, const A&) ? 1 : -1];
   int check4[ __is_trivially_constructible(B, const B&) ? 1 : -1];
@@ -1494,7 +1494,7 @@ namespace dr496 { // dr496: sup 2094
   int check6[ __is_trivially_assignable(B, const B&) ? 1 : -1];
 }
 
-namespace dr497 { // dr497: sup 253
+namespace cwg497 { // cwg497: sup 253
   void before() {
     struct S {
       mutable int i;
@@ -1517,7 +1517,7 @@ namespace dr497 { // dr497: sup 253
   }
 }
 
-namespace dr499 { // dr499: yes
+namespace cwg499 { // cwg499: yes
   extern char str[];
   void f() { throw str; }
 }
diff --git a/clang/test/CXX/drs/dr5xx.cpp b/clang/test/CXX/drs/dr5xx.cpp
index d473f4827455c..0fe64102d70b0 100644
--- a/clang/test/CXX/drs/dr5xx.cpp
+++ b/clang/test/CXX/drs/dr5xx.cpp
@@ -9,20 +9,20 @@
 // pointing at the implicit operator new. We can't match such a diagnostic
 // with -verify.
 __extension__ typedef __SIZE_TYPE__ size_t;
-void *operator new(size_t); // #dr5xx-global-operator-new
+void *operator new(size_t); // #cwg5xx-global-operator-new
 // cxx98-error@-1 {{'operator new' is missing exception specification 'throw(std::bad_alloc)'}}
 #if __cplusplus > 201402L
 namespace std {
   enum class align_val_t : size_t {};
 }
-void *operator new(size_t, std::align_val_t); // #dr5xx-global-operator-new-aligned
+void *operator new(size_t, std::align_val_t); // #cwg5xx-global-operator-new-aligned
 #endif
 
 namespace std {
   struct type_info;
 }
 
-namespace dr500 { // dr500: dup 372
+namespace cwg500 { // cwg500: dup 372
   class D;
   class A {
     class B;
@@ -34,7 +34,7 @@ namespace dr500 { // dr500: dup 372
   class D : public A::B {};
 }
 
-namespace dr501 { // dr501: yes
+namespace cwg501 { // cwg501: yes
   struct A {
     friend void f() {}
     void g() {
@@ -44,7 +44,7 @@ namespace dr501 { // dr501: yes
   };
 }
 
-namespace dr502 { // dr502: yes
+namespace cwg502 { // cwg502: yes
   struct Q {};
   template<typename T> struct A {
     enum E { e = 1 };
@@ -59,7 +59,7 @@ namespace dr502 { // dr502: yes
   template struct A<int>;
 }
 
-namespace dr505 { // dr505: yes
+namespace cwg505 { // cwg505: yes
   const char *exts = "\e\(\{\[\%";
   // expected-error@-1 {{use of non-standard escape character '\e'}}
   // expected-error@-2 {{use of non-standard escape character '\('}}
@@ -70,7 +70,7 @@ namespace dr505 { // dr505: yes
   // expected-error@-1 {{unknown escape sequence '\Q'}}
 }
 
-namespace dr506 { // dr506: yes
+namespace cwg506 { // cwg506: yes
   struct NonPod { ~NonPod(); };
   void f(...);
   void g(NonPod np) { f(np); }
@@ -78,33 +78,33 @@ namespace dr506 { // dr506: yes
   // since-cxx11-error@-2 {{cannot pass object of non-trivial type 'NonPod' through variadic function; call will abort at runtime}}
 }
 
-// FIXME: Add tests here once DR260 is resolved.
-// dr507: dup 260
+// FIXME: Add tests here once CWG260 is resolved.
+// cwg507: dup 260
 
-// dr508: na
-// dr509: na
-// dr510: na
+// cwg508: na
+// cwg509: na
+// cwg510: na
 
-namespace dr512 { // dr512: yes
-  struct A { // #dr512-A
-    A(int); // #dr512-A-ctor
+namespace cwg512 { // cwg512: yes
+  struct A { // #cwg512-A
+    A(int); // #cwg512-A-ctor
   };
   union U { A a; };
   // cxx98-error@-1 {{union member 'a' has a non-trivial default constructor}}
-  //   cxx98-note@#dr512-A {{because type 'dr512::A' has no default constructor}}
-  //   cxx98-note@#dr512-A-ctor {{implicit default constructor suppressed by user-declared constructor}}
+  //   cxx98-note@#cwg512-A {{because type 'cwg512::A' has no default constructor}}
+  //   cxx98-note@#cwg512-A-ctor {{implicit default constructor suppressed by user-declared constructor}}
 }
 
-// dr513: na
+// cwg513: na
 
-namespace dr514 { // dr514: yes
+namespace cwg514 { // cwg514: yes
   namespace A { extern int x, y; }
   int A::x = y;
 }
 
-namespace dr515 { // dr515: sup 1017
-  // FIXME: dr1017 reverses the wording of dr515, but the current draft has
-  // dr515's wording, with a different fix for dr1017.
+namespace cwg515 { // cwg515: sup 1017
+  // FIXME: cwg1017 reverses the wording of cwg515, but the current draft has
+  // cwg515's wording, with a different fix for cwg1017.
 
   struct X { int n; };
   template<typename T> struct Y : T {
@@ -117,9 +117,9 @@ namespace dr515 { // dr515: sup 1017
   // cxx98-error@-1 {{invalid use of non-static data member 'a'}}
 }
 
-// dr516: na
+// cwg516: na
 
-namespace dr517 { // dr517: no
+namespace cwg517 { // cwg517: no
   // This is NDR, but we should diagnose it anyway.
   template<typename T> struct S {};
   template<typename T> int v = 0;
@@ -140,25 +140,25 @@ namespace dr517 { // dr517: no
   template<typename T> int v<T&> = 0;
 }
 
-namespace dr518 { // dr518: yes c++11
+namespace cwg518 { // cwg518: yes c++11
   enum E { e, };
   // cxx98-error@-1 {{commas at the end of enumerator lists are a C++11 extension}}
 }
 
-// dr519 is in dr519.cpp
-// dr520: na
+// cwg519 is in cwg519.cpp
+// cwg520: na
 
-// dr521: no
+// cwg521: no
 // FIXME: The wording here is broken. It's not reasonable to expect a
 // diagnostic here. Once the relevant DR gets a number, mark this as a dup.
 
-namespace dr522 { // dr522: yes
+namespace cwg522 { // cwg522: yes
   struct S {};
   template<typename T> void b1(volatile T &);
   template<typename T> void b2(volatile T * const *);
   template<typename T> void b2(volatile T * const S::*);
   template<typename T> void b2(volatile T * const S::* const *);
-  template<typename T> void b2a(volatile T *S::* const *); // #dr522-b2a
+  template<typename T> void b2a(volatile T *S::* const *); // #cwg522-b2a
 
   template<typename T> struct Base {};
   struct Derived : Base<int> {};
@@ -178,28 +178,28 @@ namespace dr522 { // dr522: yes
     b2(am);
     b2a(am);
     // expected-error@-1 {{no matching function for call to 'b2a'}}
-    //   expected-note@#dr522-b2a {{candidate template ignored: deduced type 'volatile int *dr522::S::*const *' of 1st parameter does not match adjusted type 'int *dr522::S::**' of argument}}
+    //   expected-note@#cwg522-b2a {{candidate template ignored: deduced type 'volatile int *cwg522::S::*const *' of 1st parameter does not match adjusted type 'int *cwg522::S::**' of argument}}
     b3(d);
     b3(cd);
   }
 }
 
-namespace dr524 { // dr524: yes
+namespace cwg524 { // cwg524: yes
   template<typename T> void f(T a, T b) { operator+(a, b); }
   // expected-error@-1 {{call to function 'operator+' that is neither visible in the template definition nor found by argument-dependent lookup}}
-  //   expected-note@#dr524-f-N-S {{in instantiation of function template specialization 'dr524::f<dr524::N::S>' requested here}}
-  //   expected-note@#dr524-operator-plus {{'operator+' should be declared prior to the call site or in namespace 'dr524::N'}}
+  //   expected-note@#cwg524-f-N-S {{in instantiation of function template specialization 'cwg524::f<cwg524::N::S>' requested here}}
+  //   expected-note@#cwg524-operator-plus {{'operator+' should be declared prior to the call site or in namespace 'cwg524::N'}}
 
   struct S {};
   void operator+(S, S);
   template void f(S, S);
 
   namespace N { struct S {}; }
-  void operator+(N::S, N::S); // #dr524-operator-plus
-  template void f(N::S, N::S); // #dr524-f-N-S
+  void operator+(N::S, N::S); // #cwg524-operator-plus
+  template void f(N::S, N::S); // #cwg524-f-N-S
 }
 
-namespace dr525 { // dr525: yes
+namespace cwg525 { // cwg525: yes
   namespace before {
     // Note, the example was correct prior to the change; instantiation is
     // required for cases like this:
@@ -211,37 +211,37 @@ namespace dr525 { // dr525: yes
   namespace after {
     template <class T> struct D { typename T::error e; };
     // expected-error@-1 {{type 'double' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr525-ppp {{in instantiation of template class 'dr525::after::D<double>' requested here}}
+    //   expected-note@#cwg525-ppp {{in instantiation of template class 'cwg525::after::D<double>' requested here}}
     void g(D<double> *ppp) {
-      delete ppp; // #dr525-ppp
+      delete ppp; // #cwg525-ppp
     }
   }
 }
 
-namespace dr526 { // dr526: yes
+namespace cwg526 { // cwg526: yes
   template<int> struct S {};
   template<int N> void f1(S<N> s);
-  template<int N> void f2(S<(N)> s); // #dr526-f2
-  template<int N> void f3(S<+N> s); // #dr526-f3
+  template<int N> void f2(S<(N)> s); // #cwg526-f2
+  template<int N> void f3(S<+N> s); // #cwg526-f3
   template<int N> void g1(int (&)[N]);
-  template<int N> void g2(int (&)[(N)]); // #dr526-g2
-  template<int N> void g3(int (&)[+N]); // #dr526-g3
+  template<int N> void g2(int (&)[(N)]); // #cwg526-g2
+  template<int N> void g3(int (&)[+N]); // #cwg526-g3
 
   void test(int (&a)[3], S<3> s) {
     f1(s);
     f2(s);
     // expected-error@-1 {{no matching function for call to 'f2'}}
-    //   expected-note@#dr526-f2 {{candidate template ignored: couldn't infer template argument 'N'}}
+    //   expected-note@#cwg526-f2 {{candidate template ignored: couldn't infer template argument 'N'}}
     f3(s);
     // expected-error@-1 {{no matching function for call to 'f3'}}
-    //   expected-note@#dr526-f3 {{candidate template ignored: couldn't infer template argument 'N'}}
+    //   expected-note@#cwg526-f3 {{candidate template ignored: couldn't infer template argument 'N'}}
     g1(a);
     g2(a);
     // expected-error@-1 {{no matching function for call to 'g2'}}
-    //   expected-note@#dr526-g2 {{candidate template ignored: couldn't infer template argument 'N'}}
+    //   expected-note@#cwg526-g2 {{candidate template ignored: couldn't infer template argument 'N'}}
     g3(a);
     // expected-error@-1 {{no matching function for call to 'g3'}}
-    //   expected-note@#dr526-g3 {{candidate template ignored: couldn't infer template argument 'N'}}
+    //   expected-note@#cwg526-g3 {{candidate template ignored: couldn't infer template argument 'N'}}
   }
 
   template<int N> struct X {
@@ -254,34 +254,34 @@ namespace dr526 { // dr526: yes
   };
 }
 
-namespace dr527 { // dr527: na
+namespace cwg527 { // cwg527: na
   // This DR is meaningless. It removes a required diagnostic from the case
   // where a not-externally-visible object is odr-used but not defined, which
   // requires a diagnostic for a different reason.
-  extern struct { int x; } a; // FIXME: We should reject this, per dr389.
+  extern struct { int x; } a; // FIXME: We should reject this, per cwg389.
   static struct { int x; } b;
   extern "C" struct { int x; } c;
   namespace { extern struct { int x; } d; }
   typedef struct { int x; } *P;
-  struct E { static P e; }; // FIXME: We should reject this, per dr389.
+  struct E { static P e; }; // FIXME: We should reject this, per cwg389.
   namespace { struct F { static P f; }; }
 
   int ax = a.x, bx = b.x, cx = c.x, dx = d.x, ex = E::e->x, fx = F::f->x;
 }
 
-namespace dr528 { // dr528: 2.7
+namespace cwg528 { // cwg528: 2.7
 
-struct S; // #dr528-S
+struct S; // #cwg528-S
 
 void f() {
   typeid(S);
   // expected-error@-1 {{'typeid' of incomplete type 'S'}}
-  //   expected-note@#dr528-S {{forward declaration of 'dr528::S'}}
+  //   expected-note@#cwg528-S {{forward declaration of 'cwg528::S'}}
 }
 
-} // namespace dr528
+} // namespace cwg528
 
-namespace dr530 { // dr530: yes
+namespace cwg530 { // cwg530: yes
   template<int*> struct S { enum { N = 1 }; };
   template<void(*)()> struct T { enum { N = 1 }; };
   int n;
@@ -290,7 +290,7 @@ namespace dr530 { // dr530: yes
   int b[T<&f>::N];
 }
 
-namespace dr531 { // dr531: partial
+namespace cwg531 { // cwg531: partial
   namespace good {
     template<typename T> struct A {
       void f(T) { T::error; }
@@ -334,7 +334,7 @@ namespace dr531 { // dr531: partial
       void f(T) { T::error; }
       template<typename U> void g(T, U) { T::error; }
       struct B { typename T::error error; };
-      template<typename U> struct C { typename T::error error; }; // #dr531-C
+      template<typename U> struct C { typename T::error error; }; // #cwg531-C
       static T n;
     };
     template<typename T> T A<T>::n = T::error;
@@ -342,20 +342,20 @@ namespace dr531 { // dr531: partial
     void A<int>::f(int) {}
     // expected-error@-1 {{template specialization requires 'template<>'}}
     template<typename U> void A<int>::g(int, U) {}
-    // expected-error@-1 {{template parameter list matching the non-templated nested type 'dr531::bad::A<int>' should be empty}}
+    // expected-error@-1 {{template parameter list matching the non-templated nested type 'cwg531::bad::A<int>' should be empty}}
     struct A<int>::B {};
     // expected-error@-1 {{template specialization requires 'template<>'}}
     template<typename U> struct A<int>::C {};
-    // expected-error@-1 {{template parameter list matching the non-templated nested type 'dr531::bad::A<int>' should be empty}}
+    // expected-error@-1 {{template parameter list matching the non-templated nested type 'cwg531::bad::A<int>' should be empty}}
     // expected-error@-2 {{redefinition of 'C' as different kind of symbol}}
-    //   expected-note@#dr531-C {{previous definition is here}}
+    //   expected-note@#cwg531-C {{previous definition is here}}
     int A<int>::n = 0;
     // expected-error@-1 {{template specialization requires 'template<>'}}
 
-    template<> struct A<char> { // #dr531-A-char
+    template<> struct A<char> { // #cwg531-A-char
       void f(char);
       template<typename U> void g(char, U);
-      struct B; // #dr531-B
+      struct B; // #cwg531-B
       template<typename U> struct C;
       static char n;
     };
@@ -364,14 +364,14 @@ namespace dr531 { // dr531: partial
     // expected-error@-1 {{no function template matches function template specialization 'f'}}
     template<> template<typename U> void A<char>::g(char, U) {}
     // expected-error@-1 {{extraneous template parameter list in template specialization}}
-    //   expected-note@#dr531-A-char {{'template<>' header not required for explicitly-specialized class 'dr531::bad::A<char>' declared here}}
+    //   expected-note@#cwg531-A-char {{'template<>' header not required for explicitly-specialized class 'cwg531::bad::A<char>' declared here}}
     template<> struct A<char>::B {};
     // expected-error@-1 {{extraneous 'template<>' in declaration of struct 'B'}}
-    // expected-error@-2 {{specialization of member 'dr531::bad::A<char>::B' does not specialize an instantiated member}}
-    //  expected-note@#dr531-B {{attempt to specialize declaration here}}
+    // expected-error@-2 {{specialization of member 'cwg531::bad::A<char>::B' does not specialize an instantiated member}}
+    //  expected-note@#cwg531-B {{attempt to specialize declaration here}}
     template<> template<typename U> struct A<char>::C {};
     // expected-error@-1 {{extraneous template parameter list in template specialization}}
-    //   expected-note@#dr531-A-char {{'template<>' header not required for explicitly-specialized class 'dr531::bad::A<char>' declared here}}
+    //   expected-note@#cwg531-A-char {{'template<>' header not required for explicitly-specialized class 'cwg531::bad::A<char>' declared here}}
     template<> char A<char>::n = 0;
     // expected-error@-1 {{extraneous 'template<>' in declaration of variable 'n'}}
   }
@@ -388,11 +388,11 @@ namespace dr531 { // dr531: partial
     };
     template<> template<typename U> void A<int>::B<U>::f() {}
     template<typename U> void A<int>::B<U>::g() {}
-    // expected-error@-1 {{template parameter list matching the non-templated nested type 'dr531::nested::A<int>' should be empty ('template<>')}}
+    // expected-error@-1 {{template parameter list matching the non-templated nested type 'cwg531::nested::A<int>' should be empty ('template<>')}}
 
     template<> template<typename U> template<typename V> void A<int>::B<U>::h() {}
     template<typename U> template<typename V> void A<int>::B<U>::i() {}
-    // expected-error@-1 {{template parameter list matching the non-templated nested type 'dr531::nested::A<int>' should be empty ('template<>')}}
+    // expected-error@-1 {{template parameter list matching the non-templated nested type 'cwg531::nested::A<int>' should be empty ('template<>')}}
 
 #if __cplusplus <= 201703L
     // FIXME: All of those declarations shouldn't crash in C++20 mode.
@@ -403,13 +403,13 @@ namespace dr531 { // dr531: partial
     template<> void A<int>::B<char>::f() {}
     // cxx98-17-error@-1 {{template specialization requires 'template<>'}}
     template<> template<typename V> void A<int>::B<char>::h() {}
-    // cxx98-17-error@-1 {{template parameter list matching the non-templated nested type 'dr531::nested::A<int>::B<char>' should be empty ('template<>')}}
+    // cxx98-17-error@-1 {{template parameter list matching the non-templated nested type 'cwg531::nested::A<int>::B<char>' should be empty ('template<>')}}
 #endif
   }
 }
 
 // PR8130
-namespace dr532 { // dr532: 3.5
+namespace cwg532 { // cwg532: 3.5
   struct A { };
 
   template<class T> struct B {
@@ -424,16 +424,16 @@ namespace dr532 { // dr532: 3.5
   }
 }
 
-// dr533: na
+// cwg533: na
 
-namespace dr534 { // dr534: 2.9
+namespace cwg534 { // cwg534: 2.9
   struct S {};
   template<typename T> void operator+(S, T);
   template<typename T> void operator+<T*>(S, T*) {}
   // expected-error@-1 {{function template partial specialization is not allowed}}
 }
 
-namespace dr535 { // dr535: yes
+namespace cwg535 { // cwg535: yes
   class X { private: X(const X&); };
   struct A {
     X x;
@@ -464,12 +464,12 @@ namespace dr535 { // dr535: yes
 #endif
 }
 
-// dr536: na
-// dr537: na
-// dr538: na
+// cwg536: na
+// cwg537: na
+// cwg538: na
 
-// dr539: yes
-const dr539(
+// cwg539: yes
+const cwg539(
 // expected-error@-1 {{a type specifier is required for all declarations}}
     const a) {
     // expected-error@-1 {{unknown type name 'a'}}
@@ -521,21 +521,21 @@ const dr539(
 #endif
 }
 
-namespace dr540 { // dr540: yes
+namespace cwg540 { // cwg540: yes
   typedef int &a;
   typedef const a &a;
   // expected-warning@-1 {{'const' qualifier on reference type 'a' (aka 'int &') has no effect}}
   typedef const int &b;
   typedef b &b;
-  typedef const a &c; // #dr540-typedef-a-c
+  typedef const a &c; // #cwg540-typedef-a-c
   // expected-warning@-1 {{'const' qualifier on reference type 'a' (aka 'int &') has no effect}}
-  typedef const b &c; // #dr540-typedef-b-c
-  // expected-error@#dr540-typedef-b-c {{typedef redefinition with different types ('const int &' vs 'int &')}}
-  //   expected-note@#dr540-typedef-a-c {{previous definition is here}}
-  // expected-warning@#dr540-typedef-b-c {{'const' qualifier on reference type 'b' (aka 'const int &') has no effect}}
+  typedef const b &c; // #cwg540-typedef-b-c
+  // expected-error@#cwg540-typedef-b-c {{typedef redefinition with different types ('const int &' vs 'int &')}}
+  //   expected-note@#cwg540-typedef-a-c {{previous definition is here}}
+  // expected-warning@#cwg540-typedef-b-c {{'const' qualifier on reference type 'b' (aka 'const int &') has no effect}}
 }
 
-namespace dr541 { // dr541: yes
+namespace cwg541 { // cwg541: yes
   template<int> struct X { typedef int type; };
   template<typename T> struct S {
     int f(T);
@@ -565,45 +565,45 @@ namespace dr541 { // dr541: yes
   };
 }
 
-namespace dr542 { // dr542: yes
+namespace cwg542 { // cwg542: yes
 #if __cplusplus >= 201103L
   // In C++20 A and B are no longer aggregates and thus the constructor is
   // called, which fails.
-  struct A { A() = delete; int n; }; // #dr542-A
+  struct A { A() = delete; int n; }; // #cwg542-A
   // ok, constructor not called
-  A a[32] = {}; // #dr542-a
+  A a[32] = {}; // #cwg542-a
   // since-cxx20-error@-1 {{call to deleted constructor of 'A'}}
-  //   since-cxx20-note@#dr542-A {{'A' has been explicitly marked deleted here}}
-  //   since-cxx20-note@#dr542-a {{in implicit initialization of array element 0 with omitted initializer}}
+  //   since-cxx20-note@#cwg542-A {{'A' has been explicitly marked deleted here}}
+  //   since-cxx20-note@#cwg542-a {{in implicit initialization of array element 0 with omitted initializer}}
 
   struct B {
     int n;
   private:
-    B() = default; // #dr542-B-ctor
+    B() = default; // #cwg542-B-ctor
   };
   B b[32] = {}; // ok, constructor not called
-  // since-cxx20-error@-1 {{calling a private constructor of class 'dr542::B'}}
-  //   since-cxx20-note@#dr542-B-ctor {{declared private here}}
+  // since-cxx20-error@-1 {{calling a private constructor of class 'cwg542::B'}}
+  //   since-cxx20-note@#cwg542-B-ctor {{declared private here}}
 #endif
 }
 
-namespace dr543 { // dr543: 3.0
-  // In C++98+DR543, this is valid because value-initialization doesn't call a
+namespace cwg543 { // cwg543: 3.0
+  // In C++98+CWG543, this is valid because value-initialization doesn't call a
   // trivial default constructor, so we never notice that defining the
   // constructor would be ill-formed.
   //
-  // In C++11+DR543, this is ill-formed, because the default constructor is
+  // In C++11+CWG543, this is ill-formed, because the default constructor is
   // deleted, and value-initialization *does* call a deleted default
   // constructor, even if it is trivial.
   struct A {
-    const int n; // #dr543-A-n
+    const int n; // #cwg543-A-n
   };
   A a = A();
   // since-cxx11-error@-1 {{call to implicitly-deleted default constructor of 'A'}}
-  //   since-cxx11-note@#dr543-A-n {{default constructor of 'A' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
+  //   since-cxx11-note@#cwg543-A-n {{default constructor of 'A' is implicitly deleted because field 'n' of const-qualified type 'const int' would not be initialized}}
 }
 
-namespace dr544 { // dr544: yes
+namespace cwg544 { // cwg544: yes
   int *n;
 
   template<class T> struct A { int n; };
@@ -612,13 +612,13 @@ namespace dr544 { // dr544: yes
   int k = B<int>().get();
 }
 
-namespace dr546 { // dr546: yes
+namespace cwg546 { // cwg546: yes
   template<typename T> struct A { void f(); };
   template struct A<int>;
   template<typename T> void A<T>::f() { T::error; }
 }
 
-namespace dr547 { // dr547: 3.2
+namespace cwg547 { // cwg547: 3.2
   template<typename T> struct X;
   template<typename T> struct X<T() const> {};
   template<typename T, typename C> X<T> f(T C::*) { return X<T>(); }
@@ -627,16 +627,16 @@ namespace dr547 { // dr547: 3.2
   X<void() const> x = f(&S::f);
 }
 
-namespace dr548 { // dr548: dup 482
+namespace cwg548 { // cwg548: dup 482
   template<typename T> struct S {};
   template<typename T> void f() {}
-  template struct dr548::S<int>;
-  template void dr548::f<int>();
+  template struct cwg548::S<int>;
+  template void cwg548::f<int>();
 }
 
-// dr550: dup 393
+// cwg550: dup 393
 
-namespace dr551 { // dr551: yes c++11
+namespace cwg551 { // cwg551: yes c++11
   // FIXME: This obviously should apply in C++98 mode too.
   template<typename T> void f() {}
   template inline void f<int>();
@@ -653,25 +653,25 @@ namespace dr551 { // dr551: yes c++11
   // since-cxx11-error@-1 {{explicit instantiation cannot be 'inline'}}
 }
 
-namespace dr552 { // dr552: yes
+namespace cwg552 { // cwg552: yes
   template<typename T, typename T::U> struct X {};
   struct Y { typedef int U; };
   X<Y, 0> x;
 }
 
-// dr553: 2.7
-struct dr553_class {
-  friend void *operator new(size_t, dr553_class);
+// cwg553: 2.7
+struct cwg553_class {
+  friend void *operator new(size_t, cwg553_class);
 };
-namespace dr553 {
-  dr553_class c;
+namespace cwg553 {
+  cwg553_class c;
   // Contrary to the apparent intention of the DR, operator new is not actually
   // looked up with a lookup mechanism that performs ADL; the standard says it
   // "is looked up in global scope", where it is not visible.
   void *p = new (c) int;
   // expected-error@-1 {{no matching function for call to 'operator new'}}
-  //   since-cxx17-note@#dr5xx-global-operator-new-aligned {{candidate function not viable: no known conversion from 'dr553_class' to 'std::align_val_t' for 2nd argument}}
-  //   expected-note@#dr5xx-global-operator-new {{candidate function not viable: requires 1 argument, but 2 were provided}}
+  //   since-cxx17-note@#cwg5xx-global-operator-new-aligned {{candidate function not viable: no known conversion from 'cwg553_class' to 'std::align_val_t' for 2nd argument}}
+  //   expected-note@#cwg5xx-global-operator-new {{candidate function not viable: requires 1 argument, but 2 were provided}}
 
   struct namespace_scope {
     friend void *operator new(size_t, namespace_scope);
@@ -679,9 +679,9 @@ namespace dr553 {
   };
 }
 
-// dr554: na
+// cwg554: na
 
-namespace dr555 { // dr555: 2.8
+namespace cwg555 { // cwg555: 2.8
 typedef int I;
 typedef const int CI;
 typedef volatile int VI;
@@ -734,11 +734,11 @@ void g(int &a, CI &b, VI &c) {
   c.CI::~CI();
   c.VI::~VI();
 }
-} // namespace dr555
+} // namespace cwg555
 
-// dr556: na
+// cwg556: na
 
-namespace dr557 { // dr557: 3.1
+namespace cwg557 { // cwg557: 3.1
   template<typename T> struct S {
     friend void f(S<T> *);
     friend void g(S<S<T> > *);
@@ -749,7 +749,7 @@ namespace dr557 { // dr557: 3.1
   }
 }
 
-namespace dr558 { // dr558: 2.9
+namespace cwg558 { // cwg558: 2.9
   wchar_t a = L'\uD7FF';
   wchar_t b = L'\xD7FF';
   wchar_t c = L'\uD800';
@@ -762,9 +762,9 @@ namespace dr558 { // dr558: 2.9
   wchar_t h = L'\xE000';
 }
 
-template<typename> struct dr559 { typedef int T; dr559::T u; }; // dr559: yes
+template<typename> struct cwg559 { typedef int T; cwg559::T u; }; // cwg559: yes
 
-namespace dr560 { // dr560: 16
+namespace cwg560 { // cwg560: 16
 
 template <class T>
 struct Outer {
@@ -776,9 +776,9 @@ template <class T>
 Outer<T>::Inner* Outer<T>::Inner::self() { return this; }
 // cxx98-17-error@-1 {{missing 'typename' prior to dependent type name Outer<T>::Inner; implicit 'typename' is a C++20 extension}}
 
-} // namespace dr560
+} // namespace cwg560
 
-namespace dr561 { // dr561: yes
+namespace cwg561 { // cwg561: yes
   template<typename T> void f(int);
   template<typename T> void g(T t) {
     f<T>(t);
@@ -792,20 +792,20 @@ namespace dr561 { // dr561: yes
   }
 }
 
-// dr562: na
+// cwg562: na
 
-namespace dr564 { // dr564: yes
+namespace cwg564 { // cwg564: yes
   extern "C++" void f(int);
   void f(int); // ok
   extern "C++" { extern int n; }
   int n; // ok
 }
 
-namespace dr565 { // dr565: yes
+namespace cwg565 { // cwg565: yes
   namespace N {
-    template<typename T> int f(T); // #dr565-f
+    template<typename T> int f(T); // #cwg565-f
   }
-  using N::f; // #dr565-using
+  using N::f; // #cwg565-using
   template<typename T> int f(T*);
   template<typename T> void f(T);
   template<typename T, int = 0> int f(T);
@@ -813,19 +813,19 @@ namespace dr565 { // dr565: yes
   template<typename T> int f(T, int = 0);
   template<typename T> int f(T);
   // expected-error@-1 {{declaration conflicts with target of using declaration already in scope}}
-  //   expected-note@#dr565-f {{target of using declaration}}
-  //   expected-note@#dr565-using {{using declaration}}
+  //   expected-note@#cwg565-f {{target of using declaration}}
+  //   expected-note@#cwg565-using {{using declaration}}
 }
 
-namespace dr566 { // dr566: yes
+namespace cwg566 { // cwg566: yes
 #if __cplusplus >= 201103L
   int check[int(-3.99) == -3 ? 1 : -1];
 #endif
 }
 
-// dr567: na
+// cwg567: na
 
-namespace dr568 { // dr568: 3.0 c++11
+namespace cwg568 { // cwg568: 3.0 c++11
   // FIXME: This is a DR issue against C++98, so should probably apply there
   // too.
   struct x { int y; };
@@ -860,35 +860,35 @@ namespace dr568 { // dr568: 3.0 c++11
   void jump() {
     goto x;
     // cxx98-error@-1 {{cannot jump from this goto statement to its label}}
-    //   cxx98-note@#dr568-t {{jump bypasses initialization of non-POD variable}}
-    trivial t; // #dr568-t
+    //   cxx98-note@#cwg568-t {{jump bypasses initialization of non-POD variable}}
+    trivial t; // #cwg568-t
   x: ;
   }
 }
 
-namespace dr569 { // dr569: yes c++11
+namespace cwg569 { // cwg569: yes c++11
   // FIXME: This is a DR issue against C++98, so should probably apply there
   // too.
   ;;;;;
   // cxx98-error@-1 {{C++11 extension}}
 }
 
-namespace dr570 { // dr570: dup 633
+namespace cwg570 { // cwg570: dup 633
   int n;
-  int &r = n; // #dr570-r
+  int &r = n; // #cwg570-r
   int &r = n;
   // expected-error@-1 {{redefinition of 'r'}}
-  //   expected-note@#dr570-r {{previous definition is here}}
+  //   expected-note@#cwg570-r {{previous definition is here}}
 }
 
-// dr571 is in dr571.cpp
+// cwg571 is in cwg571.cpp
 
-namespace dr572 { // dr572: yes
+namespace cwg572 { // cwg572: yes
   enum E { a = 1, b = 2 };
   int check[a + b == 3 ? 1 : -1];
 }
 
-namespace dr573 { // dr573: no
+namespace cwg573 { // cwg573: no
   void *a;
   int *b = reinterpret_cast<int*>(a);
   void (*c)() = reinterpret_cast<void(*)()>(a);
@@ -904,62 +904,62 @@ namespace dr573 { // dr573: no
   template<int*> struct T;
 }
 
-namespace dr574 { // dr574: 3.0
+namespace cwg574 { // cwg574: 3.0
   struct A {
-    A &operator=(const A&) const; // #dr574-A-copy-assign
+    A &operator=(const A&) const; // #cwg574-A-copy-assign
   };
   struct B {
-    B &operator=(const B&) volatile; // #dr574-B-copy-assign
+    B &operator=(const B&) volatile; // #cwg574-B-copy-assign
   };
 #if __cplusplus >= 201103L
   struct C {
-    C &operator=(const C&) &; // #dr574-C-copy-assign
+    C &operator=(const C&) &; // #cwg574-C-copy-assign
   };
   struct D {
-    D &operator=(const D&) &&; // #dr574-D-copy-assign
+    D &operator=(const D&) &&; // #cwg574-D-copy-assign
   };
   void test(C c, D d) {
     c = c;
     C() = c;
     // since-cxx11-error@-1 {{no viable overloaded '='}}
-    //   since-cxx11-note@#dr574-C-copy-assign {{candidate function not viable: expects an lvalue for object argument}}
+    //   since-cxx11-note@#cwg574-C-copy-assign {{candidate function not viable: expects an lvalue for object argument}}
     d = d;
     // since-cxx11-error@-1 {{no viable overloaded '='}}
-    //   since-cxx11-note@#dr574-D-copy-assign {{candidate function not viable: expects an rvalue for object argument}}
+    //   since-cxx11-note@#cwg574-D-copy-assign {{candidate function not viable: expects an rvalue for object argument}}
     D() = d;
   }
 #endif
   struct Test {
     friend A &A::operator=(const A&);
-    // expected-error@-1 {{friend declaration of 'operator=' does not match any declaration in 'dr574::A'}}
-    //   expected-note@#dr574-A-copy-assign {{candidate function has different qualifiers (expected unqualified but found 'const')}}
+    // expected-error@-1 {{friend declaration of 'operator=' does not match any declaration in 'cwg574::A'}}
+    //   expected-note@#cwg574-A-copy-assign {{candidate function has different qualifiers (expected unqualified but found 'const')}}
     friend B &B::operator=(const B&);
-    // expected-error@-1 {{friend declaration of 'operator=' does not match any declaration in 'dr574::B'}}
-    //   expected-note@#dr574-B-copy-assign {{candidate function has different qualifiers (expected unqualified but found 'volatile')}}
+    // expected-error@-1 {{friend declaration of 'operator=' does not match any declaration in 'cwg574::B'}}
+    //   expected-note@#cwg574-B-copy-assign {{candidate function has different qualifiers (expected unqualified but found 'volatile')}}
 #if __cplusplus >= 202302L
     friend C &C::operator=(const C&);
     // since-cxx23-error@-1 {{conflicting types for 'operator='}}
-    //   since-cxx23-note@#dr574-C-copy-assign {{previous declaration is here}}
+    //   since-cxx23-note@#cwg574-C-copy-assign {{previous declaration is here}}
     friend D &D::operator=(const D&);
     // since-cxx23-error@-1 {{conflicting types for 'operator='}}
-    //   since-cxx23-note@#dr574-D-copy-assign {{previous declaration is here}}
+    //   since-cxx23-note@#cwg574-D-copy-assign {{previous declaration is here}}
 #elif __cplusplus >= 201103L
     // FIXME: We shouldn't produce the 'cannot overload' diagnostics here.
-    friend C &C::operator=(const C&); // #dr574-test-C
-    // since-cxx11-error@#dr574-test-C {{cannot overload}}
-    //   since-cxx11-note@#dr574-C-copy-assign {{previous declaration is here}}
-    // since-cxx11-error@#dr574-test-C {{friend declaration of 'operator=' does not match any declaration in 'dr574::C'}}
-    //   since-cxx11-note@#dr574-C-copy-assign {{candidate function}}
-    friend D &D::operator=(const D&); // #dr574-test-D
-    // since-cxx11-error@#dr574-test-D {{cannot overload a member function without a ref-qualifier with a member function with ref-qualifier '&&'}}
-    //   since-cxx11-note@#dr574-D-copy-assign {{previous declaration is here}}
-    // since-cxx11-error@#dr574-test-D {{friend declaration of 'operator=' does not match any declaration in 'dr574::D'}}
-    //   since-cxx11-note@#dr574-D-copy-assign {{candidate function}}
+    friend C &C::operator=(const C&); // #cwg574-test-C
+    // since-cxx11-error@#cwg574-test-C {{cannot overload}}
+    //   since-cxx11-note@#cwg574-C-copy-assign {{previous declaration is here}}
+    // since-cxx11-error@#cwg574-test-C {{friend declaration of 'operator=' does not match any declaration in 'cwg574::C'}}
+    //   since-cxx11-note@#cwg574-C-copy-assign {{candidate function}}
+    friend D &D::operator=(const D&); // #cwg574-test-D
+    // since-cxx11-error@#cwg574-test-D {{cannot overload a member function without a ref-qualifier with a member function with ref-qualifier '&&'}}
+    //   since-cxx11-note@#cwg574-D-copy-assign {{previous declaration is here}}
+    // since-cxx11-error@#cwg574-test-D {{friend declaration of 'operator=' does not match any declaration in 'cwg574::D'}}
+    //   since-cxx11-note@#cwg574-D-copy-assign {{candidate function}}
 #endif
   };
 }
 
-namespace dr575 { // dr575: yes
+namespace cwg575 { // cwg575: yes
   template<typename T, typename U = typename T::type> void a(T); void a(...);
   // cxx98-error@-1 {{default template arguments for a function template are a C++11 extension}}
   template<typename T, typename T::type U = 0> void b(T); void b(...);
@@ -968,12 +968,12 @@ namespace dr575 { // dr575: yes
   // cxx98-error@-1 {{default template arguments for a function template are a C++11 extension}}
   template<typename T> void d(T, int = T::value); void d(...);
   // expected-error@-1 {{type 'int' cannot be used prior to '::' because it has no members}}
-  //   expected-note@#dr575-d {{in instantiation of default function argument expression for 'd<int>' required here}}
+  //   expected-note@#cwg575-d {{in instantiation of default function argument expression for 'd<int>' required here}}
   void x() {
     a(0);
     b(0);
     c(0);
-    d(0); // #dr575-d
+    d(0); // #cwg575-d
   }
 
   template<typename T = int&> void f(T* = 0);
@@ -987,7 +987,7 @@ namespace dr575 { // dr575: yes
   void *p = h((void*)0);
 }
 
-namespace dr576 { // dr576: 3.5
+namespace cwg576 { // cwg576: 3.5
   typedef void f() {}
   // expected-error@-1 {{function definition declared 'typedef'}}
   void f(typedef int n);
@@ -995,7 +995,7 @@ namespace dr576 { // dr576: 3.5
   void f(char c) { typedef int n; }
 }
 
-namespace dr577 { // dr577: 3.5
+namespace cwg577 { // cwg577: 3.5
   typedef void V;
   typedef const void CV;
   void a(void);
@@ -1010,35 +1010,35 @@ namespace dr577 { // dr577: 3.5
   void (*g)(V) = a;
   void (*h)(CV);
   // expected-error@-1 {{'void' as parameter must not have type qualifiers}}
-  template<typename T> void i(T); // #dr577-i
-  template<typename T> void j(void (*)(T)); // #dr577-j
+  template<typename T> void i(T); // #cwg577-i
+  template<typename T> void j(void (*)(T)); // #cwg577-j
   void k() {
     a();
     c();
     i<void>();
     // expected-error@-1 {{no matching function for call to 'i'}}
-    //   expected-note@#dr577-i {{candidate function template not viable: requires 1 argument, but 0 were provided}}
+    //   expected-note@#cwg577-i {{candidate function template not viable: requires 1 argument, but 0 were provided}}
     i<const void>();
     // expected-error@-1 {{no matching function for call to 'i'}}
-    //   expected-note@#dr577-i {{candidate function template not viable: requires 1 argument, but 0 were provided}}
+    //   expected-note@#cwg577-i {{candidate function template not viable: requires 1 argument, but 0 were provided}}
     j<void>(0);
     // expected-error@-1 {{no matching function for call to 'j'}}
-    //   expected-note@#dr577-j {{candidate template ignored: substitution failure [with T = void]: argument may not have 'void' type}}
+    //   expected-note@#cwg577-j {{candidate template ignored: substitution failure [with T = void]: argument may not have 'void' type}}
     j<const void>(0);
     // expected-error@-1 {{no matching function for call to 'j'}}
-    //   expected-note@#dr577-j {{candidate template ignored: substitution failure [with T = const void]: argument may not have 'void' type}}
+    //   expected-note@#cwg577-j {{candidate template ignored: substitution failure [with T = const void]: argument may not have 'void' type}}
   }
 }
 
-namespace dr580 { // dr580: partial
+namespace cwg580 { // cwg580: partial
   class C;
   struct A { static C c; };
   struct B { static C c; };
   class C {
-    C(); // #dr580-C-ctor
-    ~C(); // #dr580-C-dtor
+    C(); // #cwg580-C-ctor
+    ~C(); // #cwg580-C-dtor
 
-    typedef int I; // #dr580-I
+    typedef int I; // #cwg580-I
     template<int> struct X;
     template<int> friend struct Y;
     template<int> void f();
@@ -1049,8 +1049,8 @@ namespace dr580 { // dr580: partial
   template<C::I> struct C::X {};
   template<C::I> struct Y {};
   template<C::I> struct Z {};
-  // expected-error@-1 {{'I' is a private member of 'dr580::C'}}
-  //   expected-note@#dr580-I {{implicitly declared private here}}
+  // expected-error@-1 {{'I' is a private member of 'cwg580::C'}}
+  //   expected-note@#cwg580-I {{implicitly declared private here}}
 
   struct C2 {
     class X {
@@ -1069,20 +1069,20 @@ namespace dr580 { // dr580: partial
   template<C::I> void C::f() {}
   template<C::I> void g() {}
   template<C::I> void h() {}
-  // expected-error@-1 {{'I' is a private member of 'dr580::C'}}
-  //   expected-note@#dr580-I {{implicitly declared private here}}
+  // expected-error@-1 {{'I' is a private member of 'cwg580::C'}}
+  //   expected-note@#cwg580-I {{implicitly declared private here}}
 
   C A::c;
-  C B::c; // #dr580-c
-  // expected-error@#dr580-c {{calling a private constructor of class 'dr580::C'}}
-  //   expected-note@#dr580-C-ctor {{implicitly declared private here}}
-  // expected-error@#dr580-c {{variable of type 'C' has private destructor}}
-  //   expected-note@#dr580-C-dtor {{implicitly declared private here}}
+  C B::c; // #cwg580-c
+  // expected-error@#cwg580-c {{calling a private constructor of class 'cwg580::C'}}
+  //   expected-note@#cwg580-C-ctor {{implicitly declared private here}}
+  // expected-error@#cwg580-c {{variable of type 'C' has private destructor}}
+  //   expected-note@#cwg580-C-dtor {{implicitly declared private here}}
 }
 
-// dr582: na
+// cwg582: na
 
-namespace dr583 { // dr583: 4
+namespace cwg583 { // cwg583: 4
   // see n3624
   int *p;
   bool b1 = p < 0;
@@ -1095,35 +1095,35 @@ namespace dr583 { // dr583: 4
   // expected-error@-1 {{ordered comparison between pointer and zero ('int *' and 'int')}}
 }
 
-// dr584: na
+// cwg584: na
 
-namespace dr585 { // dr585: 3.0
-  template<typename> struct T; // #dr585-struct-T
+namespace cwg585 { // cwg585: 3.0
+  template<typename> struct T; // #cwg585-struct-T
   struct A {
     friend T;
     // cxx98-14-error@-1 {{a type specifier is required for all declarations}}
     // cxx98-14-error@-2 {{friends can only be classes or functions}}
     // since-cxx17-error@-3 {{use of class template 'T' requires template arguments; argument deduction not allowed in friend declaration}}
-    //   since-cxx17-note@#dr585-struct-T {{template is declared here}}
+    //   since-cxx17-note@#cwg585-struct-T {{template is declared here}}
     // FIXME: It's not clear whether the standard allows this or what it means,
-    // but the DR585 writeup suggests it as an alternative.
+    // but the CWG585 writeup suggests it as an alternative.
     template<typename U> friend T<U>;
     // expected-error@-1 {{friend type templates must use an elaborated type}}
   };
-  template<template<typename> class T> struct B { // #dr585-template-T
+  template<template<typename> class T> struct B { // #cwg585-template-T
     friend T;
     // cxx98-14-error@-1 {{a type specifier is required for all declarations}}
     // cxx98-14-error@-2 {{friends can only be classes or functions}}
     // since-cxx17-error@-3 {{use of template template parameter 'T' requires template arguments; argument deduction not allowed in friend declaration}}
-    //   since-cxx17-note@#dr585-template-T {{template is declared here}}
+    //   since-cxx17-note@#cwg585-template-T {{template is declared here}}
     template<typename U> friend T<U>;
     // expected-error@-1 {{friend type templates must use an elaborated type}}
   };
 }
 
-// dr586: na
+// cwg586: na
 
-namespace dr587 { // dr587: 3.2
+namespace cwg587 { // cwg587: 3.2
   template<typename T> void f(bool b, const T x, T y) {
     const T *p = &(b ? x : y);
   }
@@ -1132,22 +1132,22 @@ namespace dr587 { // dr587: 3.2
   template void f(bool, const S, S);
 }
 
-namespace dr588 { // dr588: yes
-  struct A { int n; }; // #dr588-A
+namespace cwg588 { // cwg588: yes
+  struct A { int n; }; // #cwg588-A
   template<typename T> int f() {
     struct S : A, T { int f() { return n; } } s;
     int a = s.f();
     int b = s.n;
     // expected-error@-1 {{member 'n' found in multiple base classes of different types}}
-    //   expected-note@#dr588-k {{in instantiation of function template specialization 'dr588::f<dr588::B>' requested here}}
-    //   expected-note@#dr588-A {{member found by ambiguous name lookup}}
-    //   expected-note@#dr588-B {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg588-k {{in instantiation of function template specialization 'cwg588::f<cwg588::B>' requested here}}
+    //   expected-note@#cwg588-A {{member found by ambiguous name lookup}}
+    //   expected-note@#cwg588-B {{member found by ambiguous name lookup}}
   }
-  struct B { int n; }; // #dr588-B
-  int k = f<B>(); // #dr588-k
+  struct B { int n; }; // #cwg588-B
+  int k = f<B>(); // #cwg588-k
 }
 
-namespace dr589 { // dr589: yes
+namespace cwg589 { // cwg589: yes
   struct B { };
   struct D : B { };
   D f();
@@ -1159,7 +1159,7 @@ namespace dr589 { // dr589: yes
   // expected-error@-1 {{taking the address of a temporary object of type 'const B'}}
 }
 
-namespace dr590 { // dr590: yes
+namespace cwg590 { // cwg590: yes
   template<typename T> struct A {
     struct B {
       struct C {
@@ -1170,7 +1170,7 @@ namespace dr590 { // dr590: yes
   template<typename T> typename A<T>::B::C A<T>::B::C::f(A<T>::B::C) {}
 }
 
-namespace dr591 { // dr591: no
+namespace cwg591 { // cwg591: no
   template<typename T> struct A {
     typedef int M;
     struct B {
@@ -1186,11 +1186,11 @@ namespace dr591 { // dr591: no
   };
 }
 
-// dr592: na
-// dr593 is in dr593.cpp
-// dr594: na
+// cwg592: na
+// cwg593 is in cwg593.cpp
+// cwg594: na
 
-namespace dr595 { // dr595: dup 1330
+namespace cwg595 { // cwg595: dup 1330
   template<class T> struct X {
     void f() throw(T) {}
     // since-cxx17-error@-1 {{ISO C++17 does not allow dynamic exception specifications}}
@@ -1201,9 +1201,9 @@ namespace dr595 { // dr595: dup 1330
   };
 }
 
-// dr597: na
+// cwg597: na
 
-namespace dr598 { // dr598: yes
+namespace cwg598 { // cwg598: yes
   namespace N {
     void f(int);
     void f(char);
@@ -1225,11 +1225,11 @@ namespace dr598 { // dr598: yes
   int &t = h(N::i);
 }
 
-namespace dr599 { // dr599: partial
+namespace cwg599 { // cwg599: partial
   typedef int Fn();
   struct S { operator void*(); };
   struct T { operator Fn*(); };
-  struct U { operator int*(); operator void*(); }; // #dr599-U
+  struct U { operator int*(); operator void*(); }; // #cwg599-U
   struct V { operator int*(); operator Fn*(); };
   void f(void *p, void (*q)(), S s, T t, U u, V v) {
     delete p;
@@ -1244,8 +1244,8 @@ namespace dr599 { // dr599: partial
     // extension allowing deletion of pointers to void.
     delete u;
     // expected-error@-1 {{ambiguous conversion of delete expression of type 'U' to a pointer}}
-    //   expected-note@#dr599-U {{conversion to pointer type 'int *'}}
-    //   expected-note@#dr599-U {{conversion to pointer type 'void *'}}
+    //   expected-note@#cwg599-U {{conversion to pointer type 'int *'}}
+    //   expected-note@#cwg599-U {{conversion to pointer type 'void *'}}
     delete v;
   }
 }
diff --git a/clang/test/CXX/drs/dr6xx.cpp b/clang/test/CXX/drs/dr6xx.cpp
index eeb41eee9c30f..d004d5aa4c791 100644
--- a/clang/test/CXX/drs/dr6xx.cpp
+++ b/clang/test/CXX/drs/dr6xx.cpp
@@ -5,29 +5,29 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,cxx11-20,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors -fno-spell-checking
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors -fno-spell-checking
 
-namespace dr600 { // dr600: 2.8
+namespace cwg600 { // cwg600: 2.8
 struct S {
   void f(int);
 
 private:
-  void f(double); // #dr600-f-double
+  void f(double); // #cwg600-f-double
 };
 
 void g(S *sp) {
   sp->f(2);
   // access control is applied after overload resolution
   sp->f(2.2);
-  // expected-error@-1 {{'f' is a private member of 'dr600::S'}}
-  //   expected-note@#dr600-f-double {{declared private here}}
+  // expected-error@-1 {{'f' is a private member of 'cwg600::S'}}
+  //   expected-note@#cwg600-f-double {{declared private here}}
 }
-} // namespace dr600
+} // namespace cwg600
 
 namespace std {
   struct type_info {};
   __extension__ typedef __SIZE_TYPE__ size_t;
 } // namespace std
 
-namespace dr601 { // dr601: yes
+namespace cwg601 { // cwg601: yes
 #if __cplusplus >= 201103L
 #define MAX __LLONG_MAX__
 #else
@@ -58,7 +58,7 @@ static_assert(0x8000000000000000 < -1, "0x8000000000000000 should be unsigned");
 #undef MAX
 }
 
-namespace dr602 { // dr602: yes
+namespace cwg602 { // cwg602: yes
   template<class T> struct A {
     template<class U> friend struct A;
   };
@@ -73,31 +73,31 @@ namespace dr602 { // dr602: yes
   B<int> b;
 }
 
-namespace dr603 { // dr603: yes
+namespace cwg603 { // cwg603: yes
   template<unsigned char> struct S {};
   typedef S<'\001'> S1;
   typedef S<(1ul << __CHAR_BIT__) + 1> S1;
   // since-cxx11-error@-1 {{cannot be narrowed}}
 }
 
-// dr604: na
-// dr605 is in dr605.cpp
+// cwg604: na
+// cwg605 is in cwg605.cpp
 
-namespace dr606 { // dr606: 3.0
+namespace cwg606 { // cwg606: 3.0
 #if __cplusplus >= 201103L
   template<typename T> struct S {};
-  template<typename T> void f(S<T> &&); // #dr606-f
+  template<typename T> void f(S<T> &&); // #cwg606-f
   template<typename T> void g(T &&);
-  template<typename T> void h(const T &&); // #dr606-h
+  template<typename T> void h(const T &&); // #cwg606-h
 
   void test(S<int> s) {
     f(s);
     // since-cxx11-error@-1 {{no matching function for call to 'f'}}
-    //   since-cxx11-note@#dr606-f {{candidate function [with T = int] not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg606-f {{candidate function [with T = int] not viable: expects an rvalue for 1st argument}}
     g(s);
     h(s);
     // since-cxx11-error@-1 {{no matching function for call to 'h'}}
-    //   since-cxx11-note@#dr606-h {{candidate function [with T = dr606::S<int>] not viable: expects an rvalue for 1st argument}}
+    //   since-cxx11-note@#cwg606-h {{candidate function [with T = cwg606::S<int>] not viable: expects an rvalue for 1st argument}}
 
     g(test);
     h(test); // ok, an rvalue reference can bind to a function lvalue
@@ -105,7 +105,7 @@ namespace dr606 { // dr606: 3.0
 #endif
 }
 
-namespace dr607 { // dr607: yes
+namespace cwg607 { // cwg607: yes
 namespace example1 {
 struct Y {};
 
@@ -131,25 +131,25 @@ struct D : B {
 
 N::D::D() : typedef_B(0) {}
 } // namespace example2
-} // namespace dr607
+} // namespace cwg607
 
-namespace dr608 { // dr608: yes
+namespace cwg608 { // cwg608: yes
   struct A { virtual void f(); };
   struct B : A {};
   struct C : A { void f(); };
   struct D : B, C {};
 }
 
-int dr610[-0u == 0u ? 1 : -1]; // dr610: yes
+int cwg610[-0u == 0u ? 1 : -1]; // cwg610: yes
 
-namespace dr611 { // dr611: yes
+namespace cwg611 { // cwg611: yes
   int k;
   struct S { int &r; } s = { k ? k : k };
 }
 
-// dr612: na
+// cwg612: na
 
-namespace dr613 { // dr613: yes c++11
+namespace cwg613 { // cwg613: yes c++11
   // see also n2253
   struct A { int n; static void f(); };
   int f(int);
@@ -158,7 +158,7 @@ namespace dr613 { // dr613: yes c++11
 
   int an1 = sizeof(A::n);
   // cxx98-error@-1 {{invalid use of non-static data member 'n'}}
-  int an2 = sizeof(A::n + 1); // valid per dr850
+  int an2 = sizeof(A::n + 1); // valid per cwg850
   // cxx98-error@-1 {{invalid use of non-static data member 'n'}}
   int an3 = sizeof A::n;
   // cxx98-error@-1 {{invalid use of non-static data member 'n'}}
@@ -186,15 +186,15 @@ namespace dr613 { // dr613: yes c++11
   }
 }
 
-int dr614_a[(-1) / 2 == 0 ? 1 : -1]; // dr614: yes
-int dr614_b[(-1) % 2 == -1 ? 1 : -1];
+int cwg614_a[(-1) / 2 == 0 ? 1 : -1]; // cwg614: yes
+int cwg614_b[(-1) % 2 == -1 ? 1 : -1];
 
-namespace dr615 { // dr615: yes
+namespace cwg615 { // cwg615: yes
   int f();
   static int n = f();
 }
 
-namespace dr616 { // dr616: 4
+namespace cwg616 { // cwg616: 4
 #if __cplusplus >= 201103L
   struct S { int n; } s;
   S f();
@@ -212,13 +212,13 @@ namespace dr616 { // dr616: 4
 #endif
 }
 
-namespace dr618 { // dr618: yes
+namespace cwg618 { // cwg618: yes
 #if (unsigned)-1 > 0
 #error wrong
 #endif
 }
 
-namespace dr619 { // dr619: yes
+namespace cwg619 { // cwg619: yes
   extern int x[10];
   struct S { static int x[10]; };
 
@@ -240,22 +240,22 @@ namespace dr619 { // dr619: yes
   }
 }
 
-// dr620: dup 568
+// cwg620: dup 568
 
-namespace dr621 { // dr621: yes
+namespace cwg621 { // cwg621: yes
   template<typename T> T f();
-  template<> int f() {} // #dr621-f
+  template<> int f() {} // #cwg621-f
   template<> int f<int>() {}
   // expected-error@-1 {{redefinition of 'f<int>'}}
-  //   expected-note@#dr621-f {{previous definition is here}}
+  //   expected-note@#cwg621-f {{previous definition is here}}
 }
 
-// dr623: na
+// cwg623: na
 // FIXME: Add documentation saying we allow invalid pointer values.
 
-// dr624 needs a libc++abi test.
+// cwg624 needs a libc++abi test.
 
-namespace dr625 { // dr625: yes
+namespace cwg625 { // cwg625: yes
   template<typename T> struct A {};
   A<auto> x = A<int>();
   // cxx98-error@-1 {{'auto' type specifier is a C++11 extension}}
@@ -266,14 +266,14 @@ namespace dr625 { // dr625: yes
   // expected-error@-2 {{'auto' not allowed in function prototype}}
 }
 
-namespace dr626 { // dr626: yes
+namespace cwg626 { // cwg626: yes
 #define STR(x) #x
   char c[2] = STR(c); // ok, type matches
   wchar_t w[2] = STR(w);
   // expected-error@-1 {{initializing wide char array with non-wide string literal}}
 }
 
-namespace dr627 { // dr627: yes
+namespace cwg627 { // cwg627: yes
   void f() {
     // FIXME: emitted diagnostic have a room for improvement
     true a = 0;
@@ -283,22 +283,22 @@ namespace dr627 { // dr627: yes
   }
 }
 
-// dr628: na
+// cwg628: na
 
-namespace dr629 { // dr629: 2.9
+namespace cwg629 { // cwg629: 2.9
   typedef int T;
   int n = 1;
   void f() {
-    auto T = 2; // #dr629-T
+    auto T = 2; // #cwg629-T
     // cxx98-error@-1 {{expected unqualified-id}}
 
     auto T(n);
     // since-cxx11-error@-1 {{redefinition of 'T'}}
-    //   since-cxx11-note@#dr629-T {{previous definition is here}}
+    //   since-cxx11-note@#cwg629-T {{previous definition is here}}
   }
 }
 
-namespace dr630 { // dr630: yes
+namespace cwg630 { // cwg630: yes
 const bool MB_EQ_WC =
     ' ' == L' ' && '\t' == L'\t' && '\v' == L'\v' && '\r' == L'\r' &&
     '\n' == L'\n' && //
@@ -334,17 +334,17 @@ _Static_assert(MB_EQ_WC, "!__STDC_MB_MIGHT_NEQ_WC__ but some character differs")
 #endif
 }
 
-// dr631: na
+// cwg631: na
 
-namespace dr632 { // dr632: yes
+namespace cwg632 { // cwg632: yes
   struct S { int n; } s = {{5}};
   // expected-warning@-1 {{braces around scalar initializer}}
 }
 
-// dr633: na
+// cwg633: na
 // see also n2993
 
-namespace dr634 { // dr634: yes
+namespace cwg634 { // cwg634: yes
   struct S { S(); S(const S&); virtual void f(); ~S(); };
   int f(...);
   char f(int);
@@ -356,7 +356,7 @@ namespace dr634 { // dr634: yes
   // since-cxx11-error@-2 {{cannot pass object of non-trivial type 'S' through variadic function; call will abort at runtime}}
 }
 
-namespace dr635 { // dr635: yes
+namespace cwg635 { // cwg635: yes
   template<typename T> struct A { A(); ~A(); };
   template<typename T> A<T>::A<T>() {}
   // expected-error@-1 {{out-of-line constructor for 'A' cannot have template arguments}}
@@ -375,15 +375,15 @@ namespace dr635 { // dr635: yes
   /*FIXME: needed for error recovery:*/;
 
   template<typename T> struct D { template<typename U> D(); D(); };
-  template<typename T> D<T>::D() {} // #dr635-D
+  template<typename T> D<T>::D() {} // #cwg635-D
   template<typename T> template<typename U> D<T>::D() {}
-  template<typename T> D<T>::D<T>() {} // #dr635-D-T
-  // expected-error@#dr635-D-T {{out-of-line constructor for 'D' cannot have template arguments}}
-  // expected-error@#dr635-D-T {{redefinition of 'D<T>'}}
-  //   expected-note@#dr635-D {{previous definition is here}}
+  template<typename T> D<T>::D<T>() {} // #cwg635-D-T
+  // expected-error@#cwg635-D-T {{out-of-line constructor for 'D' cannot have template arguments}}
+  // expected-error@#cwg635-D-T {{redefinition of 'D<T>'}}
+  //   expected-note@#cwg635-D {{previous definition is here}}
 }
 
-namespace dr637 { // dr637: yes
+namespace cwg637 { // cwg637: yes
   void f(int i) {
     i = ++i + 1;
     i = i++ + 1;
@@ -391,7 +391,7 @@ namespace dr637 { // dr637: yes
   }
 }
 
-namespace dr638 { // dr638: no
+namespace cwg638 { // cwg638: no
   template<typename T> struct A {
     struct B;
     void f();
@@ -426,23 +426,23 @@ namespace dr638 { // dr638: no
   };
 }
 
-namespace dr639 { // dr639: 3.3
+namespace cwg639 { // cwg639: 3.3
   void f(int i) {
     void((i = 0) + (i = 0));
     // expected-warning@-1 {{multiple unsequenced modifications to 'i'}}
   }
 }
 
-namespace dr641 { // dr641: yes
+namespace cwg641 { // cwg641: yes
   namespace std_example {
     struct abc;
 
     struct xyz {
-      xyz(); // #dr641-xyz-ctor
-      xyz(xyz &); // #dr641-xyz-copy-ctor
+      xyz(); // #cwg641-xyz-ctor
+      xyz(xyz &); // #cwg641-xyz-copy-ctor
 
       operator xyz &() = delete;
-      // expected-warning@-1 {{conversion function converting 'dr641::std_example::xyz' to itself will never be used}}
+      // expected-warning@-1 {{conversion function converting 'cwg641::std_example::xyz' to itself will never be used}}
       // cxx98-error@-2 {{deleted function definitions are a C++11 extension}}
       operator abc &() = delete;
       // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
@@ -451,15 +451,15 @@ namespace dr641 { // dr641: yes
     struct abc : xyz {};
 
     template<typename T>
-    void use(T &); // #dr641-use
+    void use(T &); // #cwg641-use
     void test() {
       use<xyz>(xyz());
       // expected-error@-1 {{no matching function for call to 'use'}}
-      //   expected-note@#dr641-use {{candidate function template not viable: expects an lvalue for 1st argument}}
+      //   expected-note@#cwg641-use {{candidate function template not viable: expects an lvalue for 1st argument}}
       use<const xyz>(xyz());
       // cxx98-error@-1 {{no viable constructor copying parameter of type 'xyz'; C++98 requires a copy constructor when binding a reference to a temporary}}
-      //   cxx98-note@#dr641-xyz-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
-      //   cxx98-note@#dr641-xyz-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
+      //   cxx98-note@#cwg641-xyz-copy-ctor {{candidate constructor not viable: expects an lvalue for 1st argument}}
+      //   cxx98-note@#cwg641-xyz-ctor {{candidate constructor not viable: requires 0 arguments, but 1 was provided}}
     }
   }
 
@@ -470,20 +470,20 @@ namespace dr641 { // dr641: yes
     // cxx98-error@-1 {{default template arguments for a function template are a C++11 extension}}
   };
   A a;
-  void f(A&); // #dr641-f
+  void f(A&); // #cwg641-f
   void g(const A ca) {
     f(A());
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr641-f {{candidate function not viable: expects an lvalue for 1st argument}}
+    //   expected-note@#cwg641-f {{candidate function not viable: expects an lvalue for 1st argument}}
     f(ca);
     // expected-error@-1 {{no matching function for call to 'f'}}
-    //   expected-note@#dr641-f {{candidate function not viable: 1st argument ('const A') would lose const qualifier}}
+    //   expected-note@#cwg641-f {{candidate function not viable: 1st argument ('const A') would lose const qualifier}}
     (void)A();
     (void)ca;
   }
 }
 
-namespace dr642 { // dr642: yes
+namespace cwg642 { // cwg642: yes
   void f() {
     const int i = 2;
     {
@@ -501,16 +501,16 @@ namespace dr642 { // dr642: yes
 }
 
 #if __cplusplus >= 201103L
-namespace dr643 { // dr643: 3.2
+namespace cwg643 { // cwg643: 3.2
   struct A {
     int x;
     auto f() -> decltype(this->x);
     auto f(A &a) -> decltype(a.x);
     auto g() -> decltype(x);
     auto h() -> decltype(this->y);
-    // since-cxx11-error@-1 {{no member named 'y' in 'dr643::A'}}
+    // since-cxx11-error@-1 {{no member named 'y' in 'cwg643::A'}}
     auto h(A &a) -> decltype(a.y);
-    // since-cxx11-error@-1 {{no member named 'y' in 'dr643::A'}}
+    // since-cxx11-error@-1 {{no member named 'y' in 'cwg643::A'}}
     auto i() -> decltype(y);
     // since-cxx11-error@-1 {{use of undeclared identifier 'y'}}
     int y;
@@ -519,7 +519,7 @@ namespace dr643 { // dr643: 3.2
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr644 { // dr644: partial
+namespace cwg644 { // cwg644: partial
   struct A {
     A() = default;
     int x, y;
@@ -535,7 +535,7 @@ namespace dr644 { // dr644: partial
   struct D { C c; };
   static_assert(!__is_literal_type(D), "");
 
-  // FIXME: According to DR644, E<C> is a literal type despite having virtual
+  // FIXME: According to CWG644, E<C> is a literal type despite having virtual
   // base classes. This appears to be a wording defect.
   template<typename T>
   struct E : T {
@@ -545,12 +545,12 @@ namespace dr644 { // dr644: partial
 }
 #endif
 
-// dr645 increases permission to optimize; it's not clear that it's possible to
+// cwg645 increases permission to optimize; it's not clear that it's possible to
 // test for this.
-// dr645: na
+// cwg645: na
 
 #if __cplusplus >= 201103L
-namespace dr646 { // dr646: sup 981
+namespace cwg646 { // cwg646: sup 981
   struct A {
     constexpr A(const A&) = default; // ok
   };
@@ -564,28 +564,28 @@ namespace dr646 { // dr646: sup 981
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr647 { // dr647: 3.1
-  // This is partially superseded by dr1358.
+namespace cwg647 { // cwg647: 3.1
+  // This is partially superseded by cwg1358.
   struct A {
     constexpr virtual void f() const;
     constexpr virtual void g() const {}
     // cxx11-17-error@-1 {{virtual function cannot be constexpr}}
   };
 
-  struct X { virtual void f() const; }; // #dr647-f
+  struct X { virtual void f() const; }; // #cwg647-f
   struct B : X {
     constexpr void f() const {}
     // cxx11-17-error@-1 {{virtual function cannot be constexpr}}
-    //   cxx11-17-note@#dr647-f {{overridden virtual function is here}}
+    //   cxx11-17-note@#cwg647-f {{overridden virtual function is here}}
   };
 
-  struct NonLiteral { NonLiteral() {} }; // #dr647-NonLiteral
+  struct NonLiteral { NonLiteral() {} }; // #cwg647-NonLiteral
 
   struct C {
     constexpr C(NonLiteral);
     constexpr C(NonLiteral, int) {}
     // cxx11-20-error@-1 {{constexpr constructor's 1st parameter type 'NonLiteral' is not a literal type}}
-    //   cxx11-20-note@#dr647-NonLiteral {{'NonLiteral' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
+    //   cxx11-20-note@#cwg647-NonLiteral {{'NonLiteral' is not literal because it is not an aggregate and has no constexpr constructors other than copy or move constructors}}
     constexpr C() try {} catch (...) {}
     // cxx11-17-error@-1 {{function try block in constexpr constructor is a C++20 extension}}
     // cxx11-error@-2 {{use of this statement in a constexpr constructor is a C++14 extension}}
@@ -594,7 +594,7 @@ namespace dr647 { // dr647: 3.1
   struct D {
     operator int() const;
     constexpr D(int) {}
-    D(float); // #dr647-D-float-ctor
+    D(float); // #cwg647-D-float-ctor
   };
   constexpr int get();
   struct E {
@@ -602,7 +602,7 @@ namespace dr647 { // dr647: 3.1
     D d;
 
     // FIXME: We should diagnose this, as the conversion function is not
-    // constexpr. However, that part of this issue is supreseded by dr1364 and
+    // constexpr. However, that part of this issue is supreseded by cwg1364 and
     // others; no diagnostic is required for this any more.
     constexpr E()
         : n(D(0)),
@@ -610,22 +610,22 @@ namespace dr647 { // dr647: 3.1
 
     constexpr E(int)
     // cxx11-20-error@-1 {{constexpr constructor never produces a constant expression}}
-    //   cxx11-20-note@#dr647-int-d {{non-constexpr constructor 'D' cannot be used in a constant expression}}
-    //   cxx11-20-note@#dr647-D-float-ctor {{declared here}}
+    //   cxx11-20-note@#cwg647-int-d {{non-constexpr constructor 'D' cannot be used in a constant expression}}
+    //   cxx11-20-note@#cwg647-D-float-ctor {{declared here}}
         : n(0),
-          d(0.0f) {} // #dr647-int-d
+          d(0.0f) {} // #cwg647-int-d
     constexpr E(float f)
     // cxx11-20-error@-1 {{never produces a constant expression}}
-    //   cxx11-20-note@#dr647-float-d {{non-constexpr constructor}}
-    //   cxx11-20-note@#dr647-D-float-ctor {{declared here}}
+    //   cxx11-20-note@#cwg647-float-d {{non-constexpr constructor}}
+    //   cxx11-20-note@#cwg647-D-float-ctor {{declared here}}
         : n(get()),
-          d(D(0) + f) {} // #dr647-float-d
+          d(D(0) + f) {} // #cwg647-float-d
   };
 }
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr648 { // dr648: yes
+namespace cwg648 { // cwg648: yes
   int f();
   constexpr int a = (true ? 1 : f());
   constexpr int b = false && f();
@@ -634,7 +634,7 @@ namespace dr648 { // dr648: yes
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr649 { // dr649: 3.5
+namespace cwg649 { // cwg649: 3.5
 // Maximum alignment is 8192 bytes for Windows, and 4 GB for Linux
 alignas(0x200000000) int n;
 // since-cxx11-error-re@-1 {{{{requested alignment must be (8192|4294967296) bytes or smaller}}}}
@@ -645,15 +645,15 @@ struct Y {
   // since-cxx11-error-re@-1 {{{{requested alignment must be (8192|4294967296) bytes or smaller}}}}
 };
   struct alignas(256) Z {};
-  // This part is superseded by dr2130 and eventually by aligned allocation support.
+  // This part is superseded by cwg2130 and eventually by aligned allocation support.
   auto *p = new Z;
 }
 #endif
 
-// dr650 is in dr650.cpp
+// cwg650 is in cwg650.cpp
 
 #if __cplusplus >= 201103L
-namespace dr651 { // dr651: yes
+namespace cwg651 { // cwg651: yes
   struct X {
     virtual X &f();
   };
@@ -666,16 +666,16 @@ namespace dr651 { // dr651: yes
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr652 { // dr652: yes
+namespace cwg652 { // cwg652: yes
   constexpr int n = 1.2 * 3.4;
   static_assert(n == 4, "");
 }
 #endif
 
-// dr653 is in dr653.cpp
+// cwg653 is in cwg653.cpp
 
 #if __cplusplus >= 201103L
-namespace dr654 { // dr654: sup 1423
+namespace cwg654 { // cwg654: sup 1423
   void f() {
     if (nullptr) {}
     // since-cxx11-warning@-1 {{implicit conversion of nullptr constant to 'bool'}}
@@ -701,31 +701,31 @@ namespace dr654 { // dr654: sup 1423
 }
 #endif
 
-namespace dr655 { // dr655: yes
-  struct A { A(int); }; // #dr655-A
+namespace cwg655 { // cwg655: yes
+  struct A { A(int); }; // #cwg655-A
   struct B : A {
-    A a; // #dr655-a
+    A a; // #cwg655-a
     B();
     B(int) : B() {}
     // cxx98-error@-1 {{delegating constructors are permitted only in C++11}}
-    B(int*) : A() {} // #dr655-delegating-to-A
+    B(int*) : A() {} // #cwg655-delegating-to-A
     // expected-error@-1 {{no matching constructor for initialization of 'A'}}
-    //   expected-note@#dr655-A {{candidate constructor not viable: requires 1 argument, but 0 were provided}}
-    //   expected-note@#dr655-A {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
-    //   since-cxx11-note@#dr655-A {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
-    // expected-error@#dr655-delegating-to-A {{constructor for 'dr655::B' must explicitly initialize the member 'a' which does not have a default constructor}}
-    //   expected-note@#dr655-a {{member is declared here}}
-    //   expected-note@#dr655-A {{'dr655::A' declared here}}
+    //   expected-note@#cwg655-A {{candidate constructor not viable: requires 1 argument, but 0 were provided}}
+    //   expected-note@#cwg655-A {{candidate constructor (the implicit copy constructor) not viable: requires 1 argument, but 0 were provided}}
+    //   since-cxx11-note@#cwg655-A {{candidate constructor (the implicit move constructor) not viable: requires 1 argument, but 0 were provided}}
+    // expected-error@#cwg655-delegating-to-A {{constructor for 'cwg655::B' must explicitly initialize the member 'a' which does not have a default constructor}}
+    //   expected-note@#cwg655-a {{member is declared here}}
+    //   expected-note@#cwg655-A {{'cwg655::A' declared here}}
   };
 }
 
-namespace dr656 { // dr656: yes
+namespace cwg656 { // cwg656: yes
   struct A { A(const A&) = delete; };
   // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   struct B : A {};
   struct X { operator B(); } x;
   const A &r = x;
-  struct Y : private A { // #dr656-Y
+  struct Y : private A { // #cwg656-Y
     operator B() volatile;
   };
   extern Y y;
@@ -733,7 +733,7 @@ namespace dr656 { // dr656: yes
   // Conversion not considered due to reference-related types.
   const A &s = y;
   // expected-error@-1 {{cannot cast 'const Y' to its private base class 'const A'}}
-  //   expected-note@#dr656-Y {{declared private here}}
+  //   expected-note@#cwg656-Y {{declared private here}}
   const A &t = vy;
   // expected-error@-1 {{binding reference of type 'const A' to value of type 'volatile Y' drops 'volatile' qualifier}}
 
@@ -741,27 +741,27 @@ namespace dr656 { // dr656: yes
   struct D : C {};
   const D &d = c; // ok, D not reference-related to C
 
-  template<typename T> void accept(T); // #dr656-accept-T
-  template<typename T> void accept(...) = delete; // #dr656-accept-var
+  template<typename T> void accept(T); // #cwg656-accept-T
+  template<typename T> void accept(...) = delete; // #cwg656-accept-var
   // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
   void f() {
     accept<const A&>(x);
     accept<const A&>(y);
-    // expected-error@-1 {{cannot cast 'const Y' to its private base class 'const dr656::A'}}
-    //   expected-note@#dr656-Y {{declared private here}}
-    accept<const A&>(vy); // #dr656-vy
+    // expected-error@-1 {{cannot cast 'const Y' to its private base class 'const cwg656::A'}}
+    //   expected-note@#cwg656-Y {{declared private here}}
+    accept<const A&>(vy); // #cwg656-vy
     // expected-error@-1 {{call to deleted function 'accept'}}
-    //   expected-note@#dr656-accept-var {{candidate function [with T = const dr656::A &] has been explicitly deleted}}
-    //   expected-note@#dr656-accept-T {{candidate function template not viable: no known conversion from 'volatile Y' to 'const A &' for 1st argument}}
-    // expected-error@#dr656-vy {{no matching constructor for initialization of 'volatile Y'}}
-    //   expected-note@#dr656-Y {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('volatile Y') would lose volatile qualifier}}
-    //   expected-note@#dr656-Y {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
+    //   expected-note@#cwg656-accept-var {{candidate function [with T = const cwg656::A &] has been explicitly deleted}}
+    //   expected-note@#cwg656-accept-T {{candidate function template not viable: no known conversion from 'volatile Y' to 'const A &' for 1st argument}}
+    // expected-error@#cwg656-vy {{no matching constructor for initialization of 'volatile Y'}}
+    //   expected-note@#cwg656-Y {{candidate constructor (the implicit copy constructor) not viable: 1st argument ('volatile Y') would lose volatile qualifier}}
+    //   expected-note@#cwg656-Y {{candidate constructor (the implicit default constructor) not viable: requires 0 arguments, but 1 was provided}}
     accept<const D&>(c);
   }
 }
 
-namespace dr657 { // dr657: partial
-  struct Abs { virtual void x() = 0; }; // #dr657-Abs
+namespace cwg657 { // cwg657: partial
+  struct Abs { virtual void x() = 0; }; // #cwg657-Abs
   struct Der : public Abs { virtual void x(); };
 
   struct Cnvt { template<typename F> Cnvt(F); };
@@ -779,7 +779,7 @@ namespace dr657 { // dr657: partial
 
   struct C { C(Abs) {} };
   // expected-error@-1 {{parameter type 'Abs' is an abstract class}}
-  //   expected-note@#dr657-Abs {{unimplemented pure virtual method 'x' in 'Abs'}}
+  //   expected-note@#cwg657-Abs {{unimplemented pure virtual method 'x' in 'Abs'}}
   struct Q { operator Abs() { __builtin_unreachable(); } } q;
   // expected-error@-1 {{return type 'Abs' is an abstract class}}
 #if __cplusplus >= 201703L
@@ -798,23 +798,23 @@ namespace dr657 { // dr657: partial
   Cnvt2<Abs>::type err;
 }
 
-// dr658 is in dr658.cpp
+// cwg658 is in cwg658.cpp
 
 #if __cplusplus >= 201103L
-namespace dr659 { // dr659: 3.0
+namespace cwg659 { // cwg659: 3.0
   static_assert(alignof(char) == alignof(char&), "");
   static_assert(alignof(int) == alignof(int&), "");
   int n = alignof(int(&)());
   // since-cxx11-error@-1 {{invalid application of 'alignof' to a function type}}
-  struct A; // #dr659-A
+  struct A; // #cwg659-A
   int m = alignof(A&);
   // since-cxx11-error@-1 {{invalid application of 'alignof' to an incomplete type 'A'}}
-  //   since-cxx11-note@#dr659-A {{forward declaration of 'dr659::A'}}
+  //   since-cxx11-note@#cwg659-A {{forward declaration of 'cwg659::A'}}
 }
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr660 { // dr660: 3.0
+namespace cwg660 { // cwg660: 3.0
   enum : int { a };
   enum class { b };
   // since-cxx11-error@-1 {{scoped enumeration requires a name}}
@@ -829,27 +829,27 @@ namespace dr660 { // dr660: 3.0
 }
 #endif
 
-// dr661 is in dr661.cpp
+// cwg661 is in cwg661.cpp
 
-namespace dr662 { // dr662: yes
+namespace cwg662 { // cwg662: yes
   template <typename T> void f(T t) {
     T &tr = t;
     T *tp = &t;
     // expected-error@-1 {{'tp' declared as a pointer to a reference of type 'int &'}}
-    //   expected-note@#dr662-f-call {{in instantiation of function template specialization 'dr662::f<int &>' requested here}}
+    //   expected-note@#cwg662-f-call {{in instantiation of function template specialization 'cwg662::f<int &>' requested here}}
 #if __cplusplus >= 201103L
     auto *ap = &t;
 #endif
   }
-  void g(int n) { f<int&>(n); } // #dr662-f-call
+  void g(int n) { f<int&>(n); } // #cwg662-f-call
 }
 
-namespace dr663 { // dr663: sup P1949
+namespace cwg663 { // cwg663: sup P1949
   int ЍЎ = 123;
 }
 
 #if __cplusplus >= 201103L
-namespace dr664 { // dr664: yes
+namespace cwg664 { // cwg664: yes
   struct A { A(const A&) = delete; };
   A &&f(A &&a, int n) {
     if (n)
@@ -859,41 +859,41 @@ namespace dr664 { // dr664: yes
 }
 #endif
 
-namespace dr665 { // dr665: 2.8
+namespace cwg665 { // cwg665: 2.8
   struct A { virtual ~A(); };
   struct B : A {} *b;
-  struct C : private A {} *c; // #dr665-C
+  struct C : private A {} *c; // #cwg665-C
   struct D : B, C {} *d;
 
   struct VB : virtual A {} *vb;
-  struct VC : private virtual A {} *vc; // #dr665-VC
+  struct VC : private virtual A {} *vc; // #cwg665-VC
   struct VD : VB, VC {} *vd;
 
   void f() {
     (void)dynamic_cast<A*>(b);
     (void)dynamic_cast<A*>(c);
-    // expected-error@-1 {{cannot cast 'dr665::C' to its private base class 'dr665::A'}}
-    //   expected-note@#dr665-C {{declared private here}}
+    // expected-error@-1 {{cannot cast 'cwg665::C' to its private base class 'cwg665::A'}}
+    //   expected-note@#cwg665-C {{declared private here}}
     (void)dynamic_cast<A*>(d);
-    /* expected-error@-1 {{ambiguous conversion from derived class 'dr665::D' to base class 'dr665::A':
-    struct dr665::D -> B -> A
-    struct dr665::D -> C -> A}} */
+    /* expected-error@-1 {{ambiguous conversion from derived class 'cwg665::D' to base class 'cwg665::A':
+    struct cwg665::D -> B -> A
+    struct cwg665::D -> C -> A}} */
     (void)dynamic_cast<A*>(vb);
     (void)dynamic_cast<A*>(vc); // emitting diagnostic, even though it could be valid at runtime
-    // expected-error@-1 {{cannot cast 'dr665::VC' to its private base class 'dr665::A'}}
-    //   expected-note@#dr665-VC {{declared private here}}
+    // expected-error@-1 {{cannot cast 'cwg665::VC' to its private base class 'cwg665::A'}}
+    //   expected-note@#cwg665-VC {{declared private here}}
     (void)dynamic_cast<A*>(vd);
   }
 }
 
-namespace dr666 { // dr666: 2.8
+namespace cwg666 { // cwg666: 2.8
   struct P { friend P operator*(P, P); P(int); } p(0);
 
   template<int> int f();
   template<typename T> int f() {
     T::type *p = 0;
     // expected-error@-1 {{missing 'typename' prior to dependent type name 'Y::type'}}
-    //   expected-note@#dr666-f-Y {{in instantiation of function template specialization 'dr666::f<dr666::Y>' requested here}}
+    //   expected-note@#cwg666-f-Y {{in instantiation of function template specialization 'cwg666::f<cwg666::Y>' requested here}}
     int a(T::type);
     // expected-error@-1 {{missing 'typename' prior to dependent type name 'Y::type'}}
     return f<T::type>();
@@ -902,18 +902,18 @@ namespace dr666 { // dr666: 2.8
   struct X { static const int type = 0; };
   struct Y { typedef int type; };
   int a = f<X>();
-  int b = f<Y>(); // #dr666-f-Y
+  int b = f<Y>(); // #cwg666-f-Y
 }
 
 // Triviality is entirely different in C++98.
 #if __cplusplus >= 201103L
-namespace dr667 { // dr667: 8
+namespace cwg667 { // cwg667: 8
   struct A {
-    A() = default; // #dr667-A-ctor
+    A() = default; // #cwg667-A-ctor
     // since-cxx11-warning@-1 {{explicitly defaulted default constructor is implicitly deleted}}
-    //   since-cxx11-note@#dr667-r {{default constructor of 'A' is implicitly deleted because field 'r' of reference type 'int &' would not be initialized}}
-    //   since-cxx11-note@#dr667-A-ctor {{replace 'default' with 'delete'}}
-    int &r; // #dr667-r
+    //   since-cxx11-note@#cwg667-r {{default constructor of 'A' is implicitly deleted because field 'r' of reference type 'int &' would not be initialized}}
+    //   since-cxx11-note@#cwg667-A-ctor {{replace 'default' with 'delete'}}
+    int &r; // #cwg667-r
   };
   static_assert(!__is_trivially_constructible(A), "");
 
@@ -931,10 +931,10 @@ namespace dr667 { // dr667: 8
 }
 #endif
 
-// dr668 needs an libc++abi test
+// cwg668 needs an libc++abi test
 
 #if __cplusplus >= 201103L
-namespace dr669 { // dr669: yes
+namespace cwg669 { // cwg669: yes
   void f() {
     int n;
     using T = decltype(n);
@@ -962,7 +962,7 @@ namespace dr669 { // dr669: yes
 }
 #endif
 
-namespace dr671 { // dr671: 2.9
+namespace cwg671 { // cwg671: 2.9
   enum class E { e };
   // cxx98-error@-1 {{scoped enumerations are a C++11 extension}}
   E e = static_cast<E>(0);
@@ -971,9 +971,9 @@ namespace dr671 { // dr671: 2.9
   int m = static_cast<int>(e);
 }
 
-// dr672 is in dr672.cpp
+// cwg672 is in cwg672.cpp
 
-namespace dr673 { // dr673: yes
+namespace cwg673 { // cwg673: yes
   template<typename> struct X { static const int n = 0; };
 
   class A {
@@ -991,7 +991,7 @@ namespace dr673 { // dr673: yes
   // expected-error@-1 {{unknown type name 'F'}}
 }
 
-namespace dr674 { // dr674: 8
+namespace cwg674 { // cwg674: 8
   template<typename T> int f(T);
 
   int g(int);
@@ -1001,25 +1001,25 @@ namespace dr674 { // dr674: 8
   template<typename T> int h(T);
 
   class X {
-    friend int dr674::f(int);
-    friend int dr674::g(int);
-    friend int dr674::h<>(int);
-    int n; // #dr674-X-n
+    friend int cwg674::f(int);
+    friend int cwg674::g(int);
+    friend int cwg674::h<>(int);
+    int n; // #cwg674-X-n
   };
 
   template<typename T> int f(T) { return X().n; }
   int g(int) { return X().n; }
   template<typename T> int g(T) { return X().n; }
-  // expected-error@-1 {{'n' is a private member of 'dr674::X'}}
-  //   expected-note@#dr674-g-int {{in instantiation of function template specialization 'dr674::g<int>' requested here}}
-  //   expected-note@#dr674-X-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg674::X'}}
+  //   expected-note@#cwg674-g-int {{in instantiation of function template specialization 'cwg674::g<int>' requested here}}
+  //   expected-note@#cwg674-X-n {{implicitly declared private here}}
   int h(int) { return X().n; }
-  // expected-error@-1 {{'n' is a private member of 'dr674::X'}}
-  //   expected-note@#dr674-X-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg674::X'}}
+  //   expected-note@#cwg674-X-n {{implicitly declared private here}}
   template<typename T> int h(T) { return X().n; }
 
   template int f(int);
-  template int g(int); // #dr674-g-int
+  template int g(int); // #cwg674-g-int
   template int h(int);
 
 
@@ -1037,27 +1037,27 @@ namespace dr674 { // dr674: 8
     friend int Y::f(int);
     friend int Y::g(int);
     friend int Y::h<>(int);
-    int n; // #dr674-Z-n
+    int n; // #cwg674-Z-n
   };
 
   template<typename T> int Y::f(T) { return Z().n; }
   int Y::g(int) { return Z().n; }
   template<typename T> int Y::g(T) { return Z().n; }
-  // expected-error@-1 {{'n' is a private member of 'dr674::Z'}}
-  //   expected-note@#dr674-Y-g-int {{in instantiation of function template specialization 'dr674::Y::g<int>' requested here}}
-  //   expected-note@#dr674-Z-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg674::Z'}}
+  //   expected-note@#cwg674-Y-g-int {{in instantiation of function template specialization 'cwg674::Y::g<int>' requested here}}
+  //   expected-note@#cwg674-Z-n {{implicitly declared private here}}
   int Y::h(int) { return Z().n; }
-  // expected-error@-1 {{'n' is a private member of 'dr674::Z'}}
-  //   expected-note@#dr674-Z-n {{implicitly declared private here}}
+  // expected-error@-1 {{'n' is a private member of 'cwg674::Z'}}
+  //   expected-note@#cwg674-Z-n {{implicitly declared private here}}
   template<typename T> int Y::h(T) { return Z().n; }
 
   // FIXME: Should the <> be required here?
   template int Y::f<>(int);
-  template int Y::g<>(int); // #dr674-Y-g-int
+  template int Y::g<>(int); // #cwg674-Y-g-int
   template int Y::h<>(int);
 }
 
-namespace dr675 { // dr675: dup 739
+namespace cwg675 { // cwg675: dup 739
   template<typename T> struct A { T n : 1; };
 #if __cplusplus >= 201103L
   static_assert(A<char>{1}.n < 0, "");
@@ -1069,47 +1069,47 @@ namespace dr675 { // dr675: dup 739
 #endif
 }
 
-// dr676: na
+// cwg676: na
 
-namespace dr677 { // dr677: no
+namespace cwg677 { // cwg677: no
   struct A {
     void *operator new(std::size_t);
-    void operator delete(void*) = delete; // #dr677-A-delete
+    void operator delete(void*) = delete; // #cwg677-A-delete
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}} 
   };
   struct B {
     void *operator new(std::size_t);
-    void operator delete(void*) = delete; // #dr677-B-delete
+    void operator delete(void*) = delete; // #cwg677-B-delete
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
     virtual ~B();
   };
   void f(A *p) { delete p; }
   // expected-error@-1 {{attempt to use a deleted function}}
-  //   expected-note@#dr677-A-delete {{'operator delete' has been explicitly marked deleted here}}
+  //   expected-note@#cwg677-A-delete {{'operator delete' has been explicitly marked deleted here}}
   // FIXME: This appears to be valid; we shouldn't even be looking up the 'operator delete' here.
   void f(B *p) { delete p; }
   // expected-error@-1 {{attempt to use a deleted function}}
-  //   expected-note@#dr677-B-delete {{'operator delete' has been explicitly marked deleted here}}
+  //   expected-note@#cwg677-B-delete {{'operator delete' has been explicitly marked deleted here}}
   B::~B() {}
   // expected-error@-1 {{attempt to use a deleted function}}
-  //   expected-note@#dr677-B-delete {{'operator delete' has been explicitly marked deleted here}}
+  //   expected-note@#cwg677-B-delete {{'operator delete' has been explicitly marked deleted here}}
 }
 
-// dr678 FIXME: check that the modules ODR check catches this
+// cwg678 FIXME: check that the modules ODR check catches this
 
-namespace dr679 { // dr679: yes
+namespace cwg679 { // cwg679: yes
   struct X {};
   template<int> void operator+(X, X);
-  template<> void operator+<0>(X, X) {} // #dr679-def
+  template<> void operator+<0>(X, X) {} // #cwg679-def
   template<> void operator+<0>(X, X) {}
   // expected-error@-1 {{redefinition of 'operator+<0>'}}
-  //   expected-note@#dr679-def {{previous definition is here}}
+  //   expected-note@#cwg679-def {{previous definition is here}}
 }
 
-// dr680: na
+// cwg680: na
 
 #if __cplusplus >= 201103L
-namespace dr681 { // dr681: partial
+namespace cwg681 { // cwg681: partial
   auto *a() -> int;
   // since-cxx11-error@-1 {{function with trailing return type must specify return type 'auto', not 'auto *'}}
   auto (*b)() -> int;
@@ -1132,7 +1132,7 @@ namespace dr681 { // dr681: partial
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr683 { // dr683: yes
+namespace cwg683 { // cwg683: yes
   struct A {
     A() = default;
     A(const A&) = default;
@@ -1150,18 +1150,18 @@ namespace dr683 { // dr683: yes
 #endif
 
 #if __cplusplus >= 201103L
-namespace dr684 { // dr684: sup 1454
+namespace cwg684 { // cwg684: sup 1454
   void f() {
-    int a;  // #dr684-a
+    int a;  // #cwg684-a
     constexpr int *p = &a;
     // expected-error@-1 {{constexpr variable 'p' must be initialized by a constant expression}}
     //   expected-note@-2 {{pointer to 'a' is not a constant expression}}
-    //   expected-note@#dr684-a {{here}}
+    //   expected-note@#cwg684-a {{here}}
   }
 }
 #endif
 
-namespace dr685 { // dr685: yes
+namespace cwg685 { // cwg685: yes
   enum E : long { e };
   // cxx98-error@-1 {{enumeration types with a fixed underlying type are a C++11 extension}}
   void f(int);
@@ -1178,20 +1178,20 @@ namespace dr685 { // dr685: yes
   void i(long);
   int c = i(g);
 
-  int j(unsigned int); // #dr685-j-uint
-  void j(long); // #dr685-j-long
+  int j(unsigned int); // #cwg685-j-uint
+  void j(long); // #cwg685-j-long
   int d = j(g);
   // expected-error@-1 {{call to 'j' is ambiguous}}
-  //   expected-note@#dr685-j-uint {{candidate function}}
-  //   expected-note@#dr685-j-long {{candidate function}}
+  //   expected-note@#cwg685-j-uint {{candidate function}}
+  //   expected-note@#cwg685-j-long {{candidate function}}
 
-  // Valid per dr1601
+  // Valid per cwg1601
   int k(short);
   void k(int);
   int x = k(g);
 }
 
-namespace dr686 { // dr686: 3.0
+namespace cwg686 { // cwg686: 3.0
   void f() {
     (void)dynamic_cast<struct A*>(0);
     // expected-error@-1 {{'A' is an incomplete type}}
@@ -1214,11 +1214,11 @@ namespace dr686 { // dr686: 3.0
     (void)sizeof(struct F*);
     (void)sizeof(struct F{}*);
     // expected-error@-1 {{'F' cannot be defined in a type specifier}}
-    (void)new struct G*; // #dr686-G
-    (void)new struct G{}*; // #dr686-G-def
+    (void)new struct G*; // #cwg686-G
+    (void)new struct G{}*; // #cwg686-G-def
     // expected-error@-1 {{allocation of incomplete type 'struct G'}}
-    //   expected-note@#dr686-G {{forward declaration of 'G'}}
-    // since-cxx11-error@#dr686-G-def {{expected expression}}
+    //   expected-note@#cwg686-G {{forward declaration of 'G'}}
+    // since-cxx11-error@#cwg686-G-def {{expected expression}}
 #if __cplusplus >= 201103L
     (void)alignof(struct H*);
     (void)alignof(struct H{}*);
@@ -1256,10 +1256,10 @@ namespace dr686 { // dr686: 3.0
   }
   template<struct R *> struct X;
   template<struct R {} *> struct Y;
-  // expected-error@-1 {{'dr686::R' cannot be defined in a type specifier}}
+  // expected-error@-1 {{'cwg686::R' cannot be defined in a type specifier}}
 }
 
-namespace dr687 { // dr687 (9 c++20, but the issue is still considered open)
+namespace cwg687 { // cwg687 (9 c++20, but the issue is still considered open)
   template<typename T> void f(T a) {
     // This is valid in C++20.
     g<int>(a);
@@ -1271,8 +1271,8 @@ namespace dr687 { // dr687 (9 c++20, but the issue is still considered open)
   }
 }
 
-namespace dr692 { // dr692: 16
-  // Also see dr1395.
+namespace cwg692 { // cwg692: 16
+  // Also see cwg1395.
 
   namespace temp_func_order_example2 {
     template <typename... T> struct A1 {};
@@ -1287,11 +1287,11 @@ namespace dr692 { // dr692: 16
     // cxx98-error@-1 {{variadic templates are a C++11 extension}}
     // cxx98-error@-2 {{deleted function definitions are a C++11 extension}}
     template <typename T1> void e2(A2<T1>);
-    template <typename T, typename U> void f(U, A1<U, T> *p = 0) = delete; // #dr692-f-deleted
+    template <typename T, typename U> void f(U, A1<U, T> *p = 0) = delete; // #cwg692-f-deleted
     // cxx98-error@-1 {{deleted function definitions are a C++11 extension}}
-    template <typename U> int &f(U, A1<U, U> *p = 0); // #dr692-f
-    template <typename T> void g(T, T = T()); // #dr692-g
-    template <typename T, typename... U> void g(T, U...); // #dr692-g-variadic
+    template <typename U> int &f(U, A1<U, U> *p = 0); // #cwg692-f
+    template <typename T> void g(T, T = T()); // #cwg692-g
+    template <typename T, typename... U> void g(T, U...); // #cwg692-g-variadic
     // cxx98-error@-1 {{variadic templates are a C++11 extension}}
     void h() {
       A1<int, int> a;
@@ -1302,12 +1302,12 @@ namespace dr692 { // dr692: 16
       e2(b2);
       f<int>(42);
       // expected-error@-1 {{call to 'f' is ambiguous}}
-      //   expected-note@#dr692-f-deleted {{candidate function [with T = int, U = int] has been explicitly deleted}}
-      //   expected-note@#dr692-f {{candidate function [with U = int]}}
+      //   expected-note@#cwg692-f-deleted {{candidate function [with T = int, U = int] has been explicitly deleted}}
+      //   expected-note@#cwg692-f {{candidate function [with U = int]}}
       g(42);
       // expected-error@-1 {{ambiguous}}
-      //   expected-note@#dr692-g {{candidate function [with T = int]}}
-      //   expected-note@#dr692-g-variadic {{candidate function [with T = int, U = <>]}}
+      //   expected-note@#cwg692-g {{candidate function [with T = int]}}
+      //   expected-note@#cwg692-g-variadic {{candidate function [with T = int, U = <>]}}
     }
   }
 
@@ -1319,7 +1319,7 @@ namespace dr692 { // dr692: 16
     // cxx98-error@-1 {{variadic templates are a C++11 extension}}
     template <typename T> void g(T);
     void h(int i) {
-      // This is made ambiguous by dr692, but made valid again by dr1395.
+      // This is made ambiguous by cwg692, but made valid again by cwg1395.
       f(&i);
       int &r = g(&i);
     }
@@ -1362,23 +1362,23 @@ namespace dr692 { // dr692: 16
   }
 }
 
-namespace dr696 { // dr696: 3.1
+namespace cwg696 { // cwg696: 3.1
   void f(const int*);
   void g() {
-    const int N = 10; // #dr696-N
+    const int N = 10; // #cwg696-N
     struct A {
       void h() {
         int arr[N]; (void)arr;
         f(&N);
-        // expected-error@-1 {{reference to local variable 'N' declared in enclosing function 'dr696::g'}}
-        //   expected-note@#dr696-N {{'N' declared here}}
+        // expected-error@-1 {{reference to local variable 'N' declared in enclosing function 'cwg696::g'}}
+        //   expected-note@#cwg696-N {{'N' declared here}}
       }
     };
 #if __cplusplus >= 201103L
     (void) [] { int arr[N]; (void)arr; };
     (void)[] { f(&N); };
     // since-cxx11-error@-1 {{variable 'N' cannot be implicitly captured in a lambda with no capture-default specified}}
-    //   since-cxx11-note@#dr696-N {{'N' declared here}}
+    //   since-cxx11-note@#cwg696-N {{'N' declared here}}
     //   since-cxx11-note@-3 {{lambda expression begins here}}
     //   since-cxx11-note@-4 {{capture 'N' by value}}
     //   since-cxx11-note@-5 {{capture 'N' by reference}}
diff --git a/clang/test/CXX/drs/dr7xx.cpp b/clang/test/CXX/drs/dr7xx.cpp
index 2cbdc218ab7b5..69ee6d6d4e6ae 100644
--- a/clang/test/CXX/drs/dr7xx.cpp
+++ b/clang/test/CXX/drs/dr7xx.cpp
@@ -4,10 +4,10 @@
 // RUN: %clang_cc1 -triple %itanium_abi_triple -std=c++17 %s -verify=expected,since-cxx14,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -triple %itanium_abi_triple -std=c++2a %s -verify=expected,since-cxx14,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 
-namespace dr705 { // dr705: yes
+namespace cwg705 { // cwg705: yes
   namespace N {
     struct S {};
-    void f(S); // #dr705-f
+    void f(S); // #cwg705-f
   }
 
   void g() {
@@ -15,14 +15,14 @@ namespace dr705 { // dr705: yes
     f(s);      // ok
     (f)(s);
     // expected-error@-1 {{use of undeclared identifier 'f'}}
-    //   expected-note@#dr705-f {{'N::f' declared here}}
+    //   expected-note@#cwg705-f {{'N::f' declared here}}
   }
 }
 
-namespace dr712 { // dr712: partial
+namespace cwg712 { // cwg712: partial
   void use(int);
   void f() {
-    const int a = 0; // #dr712-f-a
+    const int a = 0; // #cwg712-f-a
     struct X {
       void g(bool cond) {
         use(a);
@@ -34,20 +34,20 @@ namespace dr712 { // dr712: partial
         // expected-warning@-2 {{left operand of comma operator has no effect}}
 
         (void)a;
-        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr712::f'}} FIXME
-        //   expected-note@#dr712-f-a {{'a' declared here}}
+        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg712::f'}} FIXME
+        //   expected-note@#cwg712-f-a {{'a' declared here}}
         (void)(a);
-        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr712::f'}} FIXME
-        //   expected-note@#dr712-f-a {{'a' declared here}}
-        (void)(cond ? a : a); // #dr712-ternary
-        // expected-error@#dr712-ternary {{reference to local variable 'a' declared in enclosing function 'dr712::f'}} FIXME
-        //   expected-note@#dr712-f-a {{'a' declared here}}
-        // expected-error@#dr712-ternary {{reference to local variable 'a' declared in enclosing function 'dr712::f'}} FIXME
-        //   expected-note@#dr712-f-a {{'a' declared here}}
-        (void)(cond, a); // #dr712-comma
-        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr712::f'}} FIXME
-        //   expected-note@#dr712-f-a {{'a' declared here}}
-        // expected-warning@#dr712-comma {{left operand of comma operator has no effect}}
+        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg712::f'}} FIXME
+        //   expected-note@#cwg712-f-a {{'a' declared here}}
+        (void)(cond ? a : a); // #cwg712-ternary
+        // expected-error@#cwg712-ternary {{reference to local variable 'a' declared in enclosing function 'cwg712::f'}} FIXME
+        //   expected-note@#cwg712-f-a {{'a' declared here}}
+        // expected-error@#cwg712-ternary {{reference to local variable 'a' declared in enclosing function 'cwg712::f'}} FIXME
+        //   expected-note@#cwg712-f-a {{'a' declared here}}
+        (void)(cond, a); // #cwg712-comma
+        // expected-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg712::f'}} FIXME
+        //   expected-note@#cwg712-f-a {{'a' declared here}}
+        // expected-warning@#cwg712-comma {{left operand of comma operator has no effect}}
       }
     };
   }
@@ -55,29 +55,29 @@ namespace dr712 { // dr712: partial
 #if __cplusplus >= 201103L
   void g() {
     struct A { int n; };
-    constexpr A a = {0};  // #dr712-g-a
+    constexpr A a = {0};  // #cwg712-g-a
     struct X {
       void g(bool cond) {
         use(a.n);
         use(a.*&A::n);
 
         (void)a.n;
-        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr712::g'}} FIXME
-        //   since-cxx11-note@#dr712-g-a {{'a' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg712::g'}} FIXME
+        //   since-cxx11-note@#cwg712-g-a {{'a' declared here}}
         (void)(a.*&A::n);
-        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'dr712::g'}} FIXME
-        //   since-cxx11-note@#dr712-g-a {{'a' declared here}}
+        // since-cxx11-error@-1 {{reference to local variable 'a' declared in enclosing function 'cwg712::g'}} FIXME
+        //   since-cxx11-note@#cwg712-g-a {{'a' declared here}}
       }
     };
   }
 #endif
 }
 
-namespace dr727 { // dr727: partial
+namespace cwg727 { // cwg727: partial
   struct A {
-    template<typename T> struct C; // #dr727-C
-    template<typename T> void f(); // #dr727-f
-    template<typename T> static int N; // #dr727-N
+    template<typename T> struct C; // #cwg727-C
+    template<typename T> void f(); // #cwg727-f
+    template<typename T> static int N; // #cwg727-N
     // cxx98-11-error@-1 {{variable templates are a C++14 extension}}
 
     template<> struct C<int>;
@@ -90,40 +90,40 @@ namespace dr727 { // dr727: partial
     struct B {
       template<> struct C<float>;
       // expected-error@-1 {{class template specialization of 'C' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
       template<> void f<float>();
       // expected-error@-1 {{no function template matches function template specialization 'f'}}
       template<> static int N<float>;
       // expected-error@-1 {{variable template specialization of 'N' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
 
       template<typename T> struct C<T**>;
       // expected-error@-1 {{class template partial specialization of 'C' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
       template<typename T> static int N<T**>;
       // expected-error@-1 {{variable template partial specialization of 'N' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
 
       template<> struct A::C<double>;
       // expected-error@-1 {{non-friend class member 'C' cannot have a qualified name}}
       // expected-error@-2 {{class template specialization of 'C' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
       template<> void A::f<double>();
       // expected-error@-1 {{o function template matches function template specialization 'f'}}
       // expected-error@-2 {{non-friend class member 'f' cannot have a qualified name}}
       template<> static int A::N<double>;
       // expected-error@-1 {{non-friend class member 'N' cannot have a qualified name}}
       // expected-error@-2 {{variable template specialization of 'N' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
 
       template<typename T> struct A::C<T***>;
       // expected-error@-1 {{non-friend class member 'C' cannot have a qualified name}}
       // expected-error@-2 {{class template partial specialization of 'C' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
       template<typename T> static int A::N<T***>;
       // expected-error@-1 {{non-friend class member 'N' cannot have a qualified name}}
       // expected-error@-2 {{variable template partial specialization of 'N' not in class 'A' or an enclosing namespace}}
-      //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+      //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
     };
   };
 
@@ -137,34 +137,34 @@ namespace dr727 { // dr727: partial
   namespace C {
     template<> struct A::C<long>;
     // expected-error@-1 {{class template specialization of 'C' not in class 'A' or an enclosing namespace}}
-    //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+    //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
     template<> void A::f<long>();
     // expected-error@-1 {{function template specialization of 'f' not in class 'A' or an enclosing namespace}}
-    //   expected-note@#dr727-f {{explicitly specialized declaration is here}}
+    //   expected-note@#cwg727-f {{explicitly specialized declaration is here}}
     template<> int A::N<long>;
     // expected-error@-1 {{variable template specialization of 'N' not in class 'A' or an enclosing namespace}}
-    //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+    //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
 
     template<typename T> struct A::C<T*****>;
     // expected-error@-1 {{class template partial specialization of 'C' not in class 'A' or an enclosing namespace}}
-    //   expected-note@#dr727-C {{explicitly specialized declaration is here}}
+    //   expected-note@#cwg727-C {{explicitly specialized declaration is here}}
     template<typename T> int A::N<T*****>;
     // expected-error@-1 {{variable template partial specialization of 'N' not in class 'A' or an enclosing namespace}}
-    //   expected-note@#dr727-N {{explicitly specialized declaration is here}}
+    //   expected-note@#cwg727-N {{explicitly specialized declaration is here}}
   }
 
   template<typename>
   struct D {
     template<typename T> struct C { typename T::error e; };
     // expected-error@-1 {{type 'float' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr727-C-float {{in instantiation of template class 'dr727::D<int>::C<float>' requested here}}
+    //   expected-note@#cwg727-C-float {{in instantiation of template class 'cwg727::D<int>::C<float>' requested here}}
     template<typename T> void f() { T::error; }
     // expected-error@-1 {{type 'float' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr727-f-float {{in instantiation of function template specialization 'dr727::D<int>::f<float>' requested here}}
+    //   expected-note@#cwg727-f-float {{in instantiation of function template specialization 'cwg727::D<int>::f<float>' requested here}}
     template<typename T> static const int N = T::error;
     // cxx98-11-error@-1 {{variable templates are a C++14 extension}}
     // expected-error@-2 {{type 'float' cannot be used prior to '::' because it has no members}}
-    //   expected-note@#dr727-N-float {{in instantiation of static data member 'dr727::D<int>::N<float>' requested here}}
+    //   expected-note@#cwg727-N-float {{in instantiation of static data member 'cwg727::D<int>::N<float>' requested here}}
 
     template<> struct C<int> {};
     template<> void f<int>() {}
@@ -188,9 +188,9 @@ namespace dr727 { // dr727: partial
     D<int>::C<int*>();
     int b = D<int>::N<int*>;
 
-    D<int>::C<float>(); // #dr727-C-float
-    di.f<float>(); // #dr727-f-float
-    int c = D<int>::N<float>; // #dr727-N-float
+    D<int>::C<float>(); // #cwg727-C-float
+    di.f<float>(); // #cwg727-f-float
+    int c = D<int>::N<float>; // #cwg727-N-float
   }
 
   namespace mixed_inner_outer_specialization {
@@ -210,14 +210,14 @@ namespace dr727 { // dr727: partial
 #if __cplusplus >= 201402L
     template<int> struct B {
       template<int> static const int u = 1;
-      template<> static const int u<0> = 2; // #dr727-u0
+      template<> static const int u<0> = 2; // #cwg727-u0
 
       // Note that in C++17 onwards, these are implicitly inline, and so the
       // initializer of v<0> is not instantiated with the declaration. In
       // C++14, v<0> is a non-defining declaration and its initializer is
       // instantiated with the class.
       template<int> static constexpr int v = 1;
-      template<> static constexpr int v<0> = 2; // #dr727-v0
+      template<> static constexpr int v<0> = 2; // #cwg727-v0
 
       template<int> static const inline int w = 1;
       // cxx14-error@-1 {{inline variables are a C++17 extension}}
@@ -228,12 +228,12 @@ namespace dr727 { // dr727: partial
     template<> template<int> constexpr int B<0>::u = 3;
     template<> template<> constexpr int B<0>::u<0> = 4;
     // since-cxx14-error@-1 {{static data member 'u' already has an initializer}}
-    //   since-cxx14-note@#dr727-u0 {{previous initialization is here}}
+    //   since-cxx14-note@#cwg727-u0 {{previous initialization is here}}
 
     template<> template<int> constexpr int B<0>::v = 3;
     template<> template<> constexpr int B<0>::v<0> = 4;
     // cxx14-error@-1 {{static data member 'v' already has an initializer}}
-    //   cxx14-note@#dr727-v0 {{previous initialization is here}}
+    //   cxx14-note@#cwg727-v0 {{previous initialization is here}}
 
     template<> template<int> constexpr int B<0>::w = 3;
     template<> template<> constexpr int B<0>::w<0> = 4;
@@ -246,7 +246,7 @@ namespace dr727 { // dr727: partial
     static_assert(B<1>().v<0> == 2, "");
     static_assert(B<0>().v<1> == 3, "");
     static_assert(B<0>().v<0> == 4, "");
-    // cxx14-error@-1 {{static assertion failed due to requirement 'dr727::mixed_inner_outer_specialization::B<0>().v<0> == 4'}}
+    // cxx14-error@-1 {{static assertion failed due to requirement 'cwg727::mixed_inner_outer_specialization::B<0>().v<0> == 4'}}
     //   cxx14-note@-2 {{expression evaluates to '2 == 4'}}
 
     static_assert(B<1>().w<1> == 1, "");
@@ -269,21 +269,21 @@ namespace dr727 { // dr727: partial
 
     template<typename> static int v1;
     // cxx98-11-error@-1 {{variable templates are a C++14 extension}}
-    template<> static int v1<T>; // #dr727-v1-T
+    template<> static int v1<T>; // #cwg727-v1-T
     template<> static int v1<U>;
     // expected-error@-1 {{duplicate member 'v1'}}
-    //   expected-note@#dr727-Collision-int-int {{in instantiation of template class 'dr727::Collision<int, int>' requested here}}
-    //   expected-note@#dr727-v1-T {{previous}}
+    //   expected-note@#cwg727-Collision-int-int {{in instantiation of template class 'cwg727::Collision<int, int>' requested here}}
+    //   expected-note@#cwg727-v1-T {{previous}}
 
     template<typename> static inline int v2;
     // cxx98-11-error@-1 {{variable templates are a C++14 extension}}
     // cxx98-14-error@-2 {{inline variables are a C++17 extension}}
-    template<> static inline int v2<T>; // #dr727-v2-T
+    template<> static inline int v2<T>; // #cwg727-v2-T
     // cxx98-14-error@-1 {{inline variables are a C++17 extension}} 
     template<> static inline int v2<U>;
     // cxx98-14-error@-1 {{inline variables are a C++17 extension}}
     // expected-error@-2 {{duplicate member 'v2'}}
-    //   expected-note@#dr727-v2-T {{previous declaration is here}}
+    //   expected-note@#cwg727-v2-T {{previous declaration is here}}
 
     // FIXME: Missing diagnostic for duplicate class explicit specialization.
     template<typename> struct S1;
@@ -291,15 +291,15 @@ namespace dr727 { // dr727: partial
     template<> struct S1<U>;
 
     template<typename> struct S2;
-    template<> struct S2<T> {}; // #dr727-S2-T
+    template<> struct S2<T> {}; // #cwg727-S2-T
     template<> struct S2<U> {};
     // expected-error@-1 {{redefinition of 'S2<int>'}}
-    //   expected-note@#dr727-S2-T {{previous}}
+    //   expected-note@#cwg727-S2-T {{previous}}
   };
-  Collision<int, int> c; // #dr727-Collision-int-int
+  Collision<int, int> c; // #cwg727-Collision-int-int
 }
 
-namespace dr777 { // dr777: 3.7
+namespace cwg777 { // cwg777: 3.7
 #if __cplusplus >= 201103L
 template <typename... T>
 void f(int i = 0, T ...args) {}
diff --git a/clang/test/CXX/drs/dr8xx.cpp b/clang/test/CXX/drs/dr8xx.cpp
index b031e65095cdb..eba601300584d 100644
--- a/clang/test/CXX/drs/dr8xx.cpp
+++ b/clang/test/CXX/drs/dr8xx.cpp
@@ -10,18 +10,18 @@
 // expected-no-diagnostics
 #endif
 
-namespace dr873 { // dr873: 3.0
+namespace cwg873 { // cwg873: 3.0
 #if __cplusplus >= 201103L
 template <typename T> void f(T &&);
-template <> void f(int &) = delete;  // #dr873-lvalue-ref
-template <> void f(int &&) = delete; // #dr873-rvalue-ref
+template <> void f(int &) = delete;  // #cwg873-lvalue-ref
+template <> void f(int &&) = delete; // #cwg873-rvalue-ref
 void g(int i) {
   f(i); // calls f<int&>(int&)
   // since-cxx11-error@-1 {{call to deleted function 'f'}}
-  //   since-cxx11-note@#dr873-lvalue-ref {{candidate function [with T = int &] has been implicitly deleted}}
+  //   since-cxx11-note@#cwg873-lvalue-ref {{candidate function [with T = int &] has been implicitly deleted}}
   f(0); // calls f<int>(int&&)
   // since-cxx11-error@-1 {{call to deleted function 'f'}}
-  //   since-cxx11-note@#dr873-rvalue-ref {{candidate function [with T = int] has been implicitly deleted}}
+  //   since-cxx11-note@#cwg873-rvalue-ref {{candidate function [with T = int] has been implicitly deleted}}
 }
 #endif
-} // namespace dr873
+} // namespace cwg873
diff --git a/clang/test/CXX/drs/dr9xx.cpp b/clang/test/CXX/drs/dr9xx.cpp
index 7ea295a684b7a..8ecb149c355f6 100644
--- a/clang/test/CXX/drs/dr9xx.cpp
+++ b/clang/test/CXX/drs/dr9xx.cpp
@@ -14,7 +14,7 @@ namespace std {
   };
 }
 
-namespace dr948 { // dr948: 3.7
+namespace cwg948 { // cwg948: 3.7
 #if __cplusplus >= 201103L
   class A {
   public:
@@ -43,34 +43,34 @@ namespace dr948 { // dr948: 3.7
 #endif
 }
 
-namespace dr952 { // dr952: 2.8
+namespace cwg952 { // cwg952: 2.8
 namespace example1 {
 struct A {
-  typedef int I; // #dr952-I
+  typedef int I; // #cwg952-I
 };
-struct B : private A { // #dr952-B
+struct B : private A { // #cwg952-B
 };
 struct C : B {
   void f() {
     I i1;
-    // expected-error@-1 {{'I' is a private member of 'dr952::example1::A'}}
-    //   expected-note@#dr952-B {{constrained by private inheritance here}}
-    //   expected-note@#dr952-I {{member is declared here}}
+    // expected-error@-1 {{'I' is a private member of 'cwg952::example1::A'}}
+    //   expected-note@#cwg952-B {{constrained by private inheritance here}}
+    //   expected-note@#cwg952-I {{member is declared here}}
   }
   I i2;
-  // expected-error@-1 {{'I' is a private member of 'dr952::example1::A'}}
-  //   expected-note@#dr952-B {{constrained by private inheritance here}}
-  //   expected-note@#dr952-I {{member is declared here}}
+  // expected-error@-1 {{'I' is a private member of 'cwg952::example1::A'}}
+  //   expected-note@#cwg952-B {{constrained by private inheritance here}}
+  //   expected-note@#cwg952-I {{member is declared here}}
   struct D {
     I i3;
-    // expected-error@-1 {{'I' is a private member of 'dr952::example1::A'}}
-    //   expected-note@#dr952-B {{constrained by private inheritance here}}
-    //   expected-note@#dr952-I {{member is declared here}}
+    // expected-error@-1 {{'I' is a private member of 'cwg952::example1::A'}}
+    //   expected-note@#cwg952-B {{constrained by private inheritance here}}
+    //   expected-note@#cwg952-I {{member is declared here}}
     void g() {
       I i4;
-      // expected-error@-1 {{'I' is a private member of 'dr952::example1::A'}}
-      //   expected-note@#dr952-B {{constrained by private inheritance here}}
-      //   expected-note@#dr952-I {{member is declared here}}
+      // expected-error@-1 {{'I' is a private member of 'cwg952::example1::A'}}
+      //   expected-note@#cwg952-B {{constrained by private inheritance here}}
+      //   expected-note@#cwg952-I {{member is declared here}}
     }
   };
 };
@@ -84,9 +84,9 @@ struct B : A {
   friend int get(B) { return x; }
 };
 } // namespace example2
-} // namespace dr952
+} // namespace cwg952
 
-namespace dr974 { // dr974: yes
+namespace cwg974 { // cwg974: yes
 #if __cplusplus >= 201103L
   void test() {
     auto lam = [](int x = 42) { return x; };
@@ -94,23 +94,23 @@ namespace dr974 { // dr974: yes
 #endif
 }
 
-namespace dr977 { // dr977: yes
-enum E { e = E() }; // #dr977-E
+namespace cwg977 { // cwg977: yes
+enum E { e = E() }; // #cwg977-E
 #if !defined(_WIN32) || defined(__MINGW32__)
-// expected-error@#dr977-E {{invalid use of incomplete type 'E'}}
-//   expected-note@#dr977-E {{definition of 'dr977::E' is not complete until the closing '}'}}
+// expected-error@#cwg977-E {{invalid use of incomplete type 'E'}}
+//   expected-note@#cwg977-E {{definition of 'cwg977::E' is not complete until the closing '}'}}
 #endif
 #if __cplusplus >= 201103L
 enum E2 : int { e2 = E2() };
 enum struct E3 { e = static_cast<int>(E3()) };
 enum struct E4 : int { e = static_cast<int>(E4()) };
 #endif
-} // namespace dr977
+} // namespace cwg977
 
-namespace dr990 { // dr990: 3.5
+namespace cwg990 { // cwg990: 3.5
 #if __cplusplus >= 201103L
-  struct A { // #dr990-A
-    A(std::initializer_list<int>); // #dr990-A-init-list
+  struct A { // #cwg990-A
+    A(std::initializer_list<int>); // #cwg990-A-init-list
   };
   struct B {
     A a;
@@ -118,20 +118,20 @@ namespace dr990 { // dr990: 3.5
   B b1 { };
   B b2 { 1 };
   // since-cxx11-error@-1 {{no viable conversion from 'int' to 'A'}}
-  //   since-cxx11-note@#dr990-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int' to 'const A &' for 1st argument}}
-  //   since-cxx11-note@#dr990-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int' to 'A &&' for 1st argument}}
-  //   since-cxx11-note@#dr990-A-init-list {{candidate constructor not viable: no known conversion from 'int' to 'std::initializer_list<int>' for 1st argument}}
+  //   since-cxx11-note@#cwg990-A {{candidate constructor (the implicit copy constructor) not viable: no known conversion from 'int' to 'const A &' for 1st argument}}
+  //   since-cxx11-note@#cwg990-A {{candidate constructor (the implicit move constructor) not viable: no known conversion from 'int' to 'A &&' for 1st argument}}
+  //   since-cxx11-note@#cwg990-A-init-list {{candidate constructor not viable: no known conversion from 'int' to 'std::initializer_list<int>' for 1st argument}}
   B b3 { { 1 } };
 
   struct C {
     C();
     C(int);
-    C(std::initializer_list<int>) = delete; // #dr990-deleted
+    C(std::initializer_list<int>) = delete; // #cwg990-deleted
   };
   C c1[3] { 1 }; // ok
   C c2[3] { 1, {2} };
   // since-cxx11-error@-1 {{call to deleted constructor of 'C'}}
-  //   since-cxx11-note@#dr990-deleted {{'C' has been explicitly marked deleted here}}
+  //   since-cxx11-note@#cwg990-deleted {{'C' has been explicitly marked deleted here}}
 
   struct D {
     D();
diff --git a/clang/www/make_cxx_dr_status b/clang/www/make_cxx_dr_status
index 7183e1a6d2be1..7c0cf77a1524f 100755
--- a/clang/www/make_cxx_dr_status
+++ b/clang/www/make_cxx_dr_status
@@ -32,7 +32,7 @@ def parse(dr):
   return DR(section, issue, url, status, title)
 
 def collect_tests():
-  status_re = re.compile(r'\bdr([0-9]+): (.*)')
+  status_re = re.compile(r'\bcwg([0-9]+): (.*)')
   status_map = {}
   for test_cpp in os.listdir(dr_test_dir):
     if not test_cpp.endswith('.cpp'):
@@ -42,12 +42,12 @@ def collect_tests():
     for match in re.finditer(status_re, open(test_cpp, 'r').read()):
       dr_number = int(match.group(1))
       if dr_number in status_map:
-        print("error: Comment for dr{} encountered more than once. Duplicate found in {}".format(dr_number, test_cpp))
+        print("error: Comment for cwg{} encountered more than once. Duplicate found in {}".format(dr_number, test_cpp))
         sys.exit(1)
       status_map[dr_number] = match.group(2)
       found_any = True
     if not found_any:
-      print("warning:%s: no '// dr123: foo' comments in this file" % test_cpp, file=sys.stderr)
+      print("warning:%s: no '// cwg123: foo' comments in this file" % test_cpp, file=sys.stderr)
   return status_map
 
 def get_issues(path):

From 4138367444a2e3431cdfc7ab6e16c9cab5c76ad8 Mon Sep 17 00:00:00 2001
From: Ian Anderson <iana@apple.com>
Date: Fri, 12 Apr 2024 22:36:43 -0700
Subject: [PATCH 323/886] [modules] allow use of ptrauth module from
 no_undeclared_includes system modules (#88432)

Co-authored-by: Alex Lorenz <arphaman@gmail.com>
---
 clang/lib/Basic/Module.cpp                                | 4 ++++
 .../Inputs/ptrauth-include-from-darwin/module.modulemap   | 8 ++++++++
 .../Modules/Inputs/ptrauth-include-from-darwin/ptrauth.h  | 1 +
 .../Modules/Inputs/ptrauth-include-from-darwin/stddef.h   | 1 +
 clang/test/Modules/ptrauth-include-from-darwin.m          | 6 ++++++
 5 files changed, 20 insertions(+)
 create mode 100644 clang/test/Modules/Inputs/ptrauth-include-from-darwin/module.modulemap
 create mode 100644 clang/test/Modules/Inputs/ptrauth-include-from-darwin/ptrauth.h
 create mode 100644 clang/test/Modules/Inputs/ptrauth-include-from-darwin/stddef.h
 create mode 100644 clang/test/Modules/ptrauth-include-from-darwin.m

diff --git a/clang/lib/Basic/Module.cpp b/clang/lib/Basic/Module.cpp
index 256365d66bb90..bb212cde87882 100644
--- a/clang/lib/Basic/Module.cpp
+++ b/clang/lib/Basic/Module.cpp
@@ -305,6 +305,10 @@ bool Module::directlyUses(const Module *Requested) {
   if (Requested->fullModuleNameIs({"_Builtin_stddef", "max_align_t"}) ||
       Requested->fullModuleNameIs({"_Builtin_stddef_wint_t"}))
     return true;
+  // Darwin is allowed is to use our builtin 'ptrauth.h' and its accompanying
+  // module.
+  if (!Requested->Parent && Requested->Name == "ptrauth")
+    return true;
 
   if (NoUndeclaredIncludes)
     UndeclaredUses.insert(Requested);
diff --git a/clang/test/Modules/Inputs/ptrauth-include-from-darwin/module.modulemap b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/module.modulemap
new file mode 100644
index 0000000000000..741b9bb1efc54
--- /dev/null
+++ b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/module.modulemap
@@ -0,0 +1,8 @@
+module libc [no_undeclared_includes] {
+  module stddef { header "stddef.h" export * }
+}
+
+module ptrauth {
+  header "ptrauth.h"
+  export *
+}
diff --git a/clang/test/Modules/Inputs/ptrauth-include-from-darwin/ptrauth.h b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/ptrauth.h
new file mode 100644
index 0000000000000..c8620b64b2cee
--- /dev/null
+++ b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/ptrauth.h
@@ -0,0 +1 @@
+void foo();
diff --git a/clang/test/Modules/Inputs/ptrauth-include-from-darwin/stddef.h b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/stddef.h
new file mode 100644
index 0000000000000..777a524fc6711
--- /dev/null
+++ b/clang/test/Modules/Inputs/ptrauth-include-from-darwin/stddef.h
@@ -0,0 +1 @@
+@import ptrauth;
diff --git a/clang/test/Modules/ptrauth-include-from-darwin.m b/clang/test/Modules/ptrauth-include-from-darwin.m
new file mode 100644
index 0000000000000..72b0c36e7cb7d
--- /dev/null
+++ b/clang/test/Modules/ptrauth-include-from-darwin.m
@@ -0,0 +1,6 @@
+// RUN: rm -rf %t
+// RUN: %clang_cc1 -fmodules-cache-path=%t -fmodules -fimplicit-module-maps -I %S/Inputs/ptrauth-include-from-darwin %s -verify
+// expected-no-diagnostics
+
+@import libc;
+void bar() { foo(); }

From 5b36bf549c8c900d0f01451e02abd821f38d4538 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 08:38:07 +0300
Subject: [PATCH 324/886] [clang][NFC] Tidy up extension warnings about
 `_Static_assert`

Some tests in `dr6xx.cpp` were using C11 `_Static_assert`, and were expecting extension warnings in C++98 mode because of that. This is noise, and we can do better than that.
---
 clang/test/CXX/drs/dr6xx.cpp | 26 ++++++++++++--------------
 1 file changed, 12 insertions(+), 14 deletions(-)

diff --git a/clang/test/CXX/drs/dr6xx.cpp b/clang/test/CXX/drs/dr6xx.cpp
index d004d5aa4c791..9d3613ae8589e 100644
--- a/clang/test/CXX/drs/dr6xx.cpp
+++ b/clang/test/CXX/drs/dr6xx.cpp
@@ -5,6 +5,11 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,cxx11-20,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors -fno-spell-checking
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors -fno-spell-checking
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
 namespace cwg600 { // cwg600: 2.8
 struct S {
   void f(int);
@@ -43,8 +48,7 @@ namespace cwg601 { // cwg601: yes
 #endif
 
 #if __INT_MAX__ == 0x7FFFFFFF
-_Static_assert(0x80000000 < -1, "0x80000000 should be unsigned");
-// expected-error@-1 {{'_Static_assert' is a C11 extension}}
+static_assert(0x80000000 < -1, "0x80000000 should be unsigned");
 #endif
 
 #if MAX > 0xFFFFFFFFFFFFFFFF && 0x8000000000000000 < -1
@@ -223,15 +227,12 @@ namespace cwg619 { // cwg619: yes
   struct S { static int x[10]; };
 
   int x[];
-  _Static_assert(sizeof(x) == sizeof(int) * 10, "");
-  // expected-error@-1 {{'_Static_assert' is a C11 extension}}
+  static_assert(sizeof(x) == sizeof(int) * 10, "");
   extern int x[];
-  _Static_assert(sizeof(x) == sizeof(int) * 10, "");
-  // expected-error@-1 {{'_Static_assert' is a C11 extension}}
+  static_assert(sizeof(x) == sizeof(int) * 10, "");
 
   int S::x[];
-  _Static_assert(sizeof(S::x) == sizeof(int) * 10, "");
-  // expected-error@-1 {{'_Static_assert' is a C11 extension}}
+  static_assert(sizeof(S::x) == sizeof(int) * 10, "");
 
   void f() {
     extern int x[];
@@ -325,12 +326,10 @@ const bool MB_EQ_WC =
     ',' == L',' && '\\' == L'\\' && '"' == L'"' && '\'' == L'\'';
 #if __STDC_MB_MIGHT_NEQ_WC__
 #ifndef __FreeBSD__ // PR22208, FreeBSD expects us to give a bad (but conforming) answer here.
-_Static_assert(!MB_EQ_WC, "__STDC_MB_MIGHT_NEQ_WC__ but all basic source characters have same representation");
-// expected-error@-1 {{'_Static_assert' is a C11 extension}}
+static_assert(!MB_EQ_WC, "__STDC_MB_MIGHT_NEQ_WC__ but all basic source characters have same representation");
 #endif
 #else
-_Static_assert(MB_EQ_WC, "!__STDC_MB_MIGHT_NEQ_WC__ but some character differs");
-// expected-error@-1 {{'_Static_assert' is a C11 extension}}
+static_assert(MB_EQ_WC, "!__STDC_MB_MIGHT_NEQ_WC__ but some character differs");
 #endif
 }
 
@@ -488,8 +487,7 @@ namespace cwg642 { // cwg642: yes
     const int i = 2;
     {
       char i[i];
-      _Static_assert(sizeof(i) == 2, "");
-      // expected-error@-1 {{'_Static_assert' is a C11 extension}}
+      static_assert(sizeof(i) == 2, "");
     }
   }
 

From 8d8502a1b6e4e2b775d16be456f6b5ee544dcc55 Mon Sep 17 00:00:00 2001
From: Hsiangkai Wang <hsiangkai.wang@arm.com>
Date: Sat, 13 Apr 2024 07:00:46 +0100
Subject: [PATCH 325/886] [mlir][spirv] Add op decoration NoContraction
 (#88578)

---
 mlir/lib/Target/SPIRV/Deserialization/Deserializer.cpp |  1 +
 mlir/lib/Target/SPIRV/Serialization/Serializer.cpp     |  1 +
 mlir/test/Target/SPIRV/decorations.mlir                | 10 ++++++++++
 3 files changed, 12 insertions(+)

diff --git a/mlir/lib/Target/SPIRV/Deserialization/Deserializer.cpp b/mlir/lib/Target/SPIRV/Deserialization/Deserializer.cpp
index 57e8d9a772962..faaa42023a803 100644
--- a/mlir/lib/Target/SPIRV/Deserialization/Deserializer.cpp
+++ b/mlir/lib/Target/SPIRV/Deserialization/Deserializer.cpp
@@ -310,6 +310,7 @@ LogicalResult spirv::Deserializer::processDecoration(ArrayRef<uint32_t> words) {
   case spirv::Decoration::RelaxedPrecision:
   case spirv::Decoration::Restrict:
   case spirv::Decoration::RestrictPointer:
+  case spirv::Decoration::NoContraction:
     if (words.size() != 2) {
       return emitError(unknownLoc, "OpDecoration with ")
              << decorationName << "needs a single target <id>";
diff --git a/mlir/lib/Target/SPIRV/Serialization/Serializer.cpp b/mlir/lib/Target/SPIRV/Serialization/Serializer.cpp
index 9a74ac115e955..200abdf993ce8 100644
--- a/mlir/lib/Target/SPIRV/Serialization/Serializer.cpp
+++ b/mlir/lib/Target/SPIRV/Serialization/Serializer.cpp
@@ -276,6 +276,7 @@ LogicalResult Serializer::processDecorationAttr(Location loc, uint32_t resultID,
   case spirv::Decoration::RelaxedPrecision:
   case spirv::Decoration::Restrict:
   case spirv::Decoration::RestrictPointer:
+  case spirv::Decoration::NoContraction:
     // For unit attributes and decoration attributes, the args list
     // has no values so we do nothing.
     if (isa<UnitAttr, DecorationAttr>(attr))
diff --git a/mlir/test/Target/SPIRV/decorations.mlir b/mlir/test/Target/SPIRV/decorations.mlir
index e8f75a1edbd08..195773735431e 100644
--- a/mlir/test/Target/SPIRV/decorations.mlir
+++ b/mlir/test/Target/SPIRV/decorations.mlir
@@ -87,3 +87,13 @@ spirv.func @fadd_decorations(%arg: f32) -> f32 "None" {
   spirv.ReturnValue %0 : f32
 }
 }
+
+// -----
+
+spirv.module Logical GLSL450 requires #spirv.vce<v1.0, [Kernel], []> {
+spirv.func @fmul_decorations(%arg: f32) -> f32 "None" {
+  // CHECK: spirv.FMul %{{.*}}, %{{.*}} {no_contraction}
+  %0 = spirv.FMul %arg, %arg {no_contraction} : f32
+  spirv.ReturnValue %0 : f32
+}
+}

From 37ebf2ae748f8cbfab756321e04fc7e8c86996a4 Mon Sep 17 00:00:00 2001
From: Lang Hames <lhames@gmail.com>
Date: Fri, 12 Apr 2024 21:28:55 -0600
Subject: [PATCH 326/886] [ORC] Replace more KV loop variables with structured
 bindings.

Coding my way home: 4.42841S, 102.96190W
---
 llvm/lib/ExecutionEngine/Orc/Core.cpp | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/llvm/lib/ExecutionEngine/Orc/Core.cpp b/llvm/lib/ExecutionEngine/Orc/Core.cpp
index 4841a2d8c4fdb..f03dd434b704b 100644
--- a/llvm/lib/ExecutionEngine/Orc/Core.cpp
+++ b/llvm/lib/ExecutionEngine/Orc/Core.cpp
@@ -187,8 +187,8 @@ AsynchronousSymbolQuery::AsynchronousSymbolQuery(
 
   OutstandingSymbolsCount = Symbols.size();
 
-  for (auto &KV : Symbols)
-    ResolvedSymbols[KV.first] = ExecutorSymbolDef();
+  for (auto &[Name, Flags] : Symbols)
+    ResolvedSymbols[Name] = ExecutorSymbolDef();
 }
 
 void AsynchronousSymbolQuery::notifySymbolMetRequiredState(
@@ -271,8 +271,8 @@ void AsynchronousSymbolQuery::dropSymbol(const SymbolStringPtr &Name) {
 void AsynchronousSymbolQuery::detach() {
   ResolvedSymbols.clear();
   OutstandingSymbolsCount = 0;
-  for (auto &KV : QueryRegistrations)
-    KV.first->detachQueryHelper(*this, KV.second);
+  for (auto &[JD, Syms] : QueryRegistrations)
+    JD->detachQueryHelper(*this, Syms);
   QueryRegistrations.clear();
 }
 
@@ -312,8 +312,8 @@ void AbsoluteSymbolsMaterializationUnit::discard(const JITDylib &JD,
 MaterializationUnit::Interface
 AbsoluteSymbolsMaterializationUnit::extractFlags(const SymbolMap &Symbols) {
   SymbolFlagsMap Flags;
-  for (const auto &KV : Symbols)
-    Flags[KV.first] = KV.second.getFlags();
+  for (const auto &[Name, Def] : Symbols)
+    Flags[Name] = Def.getFlags();
   return MaterializationUnit::Interface(std::move(Flags), nullptr);
 }
 

From 3cf8535dbf0bf5fafa99ea1f300e2384a7254fba Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang@intel.com>
Date: Sat, 13 Apr 2024 14:51:29 +0800
Subject: [PATCH 327/886] [X86][BF16] Improve vectorization of BF16 (#88486)

1. Move expansion to combineFP_EXTEND to help with small vectors;
2. Combine FP_ROUND to reduce fptrunc then fpextend after promotion;
---
 llvm/lib/Target/X86/X86ISelLowering.cpp      |  53 +++--
 llvm/test/CodeGen/X86/bfloat.ll              | 234 +++----------------
 llvm/test/CodeGen/X86/concat-fpext-v2bf16.ll |   8 +-
 3 files changed, 66 insertions(+), 229 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index b7cb4b7dafeb6..edfa05a66288e 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -21433,25 +21433,9 @@ SDValue X86TargetLowering::LowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const {
     return Res;
   }
 
-  if (!SVT.isVector())
+  if (!SVT.isVector() || SVT.getVectorElementType() == MVT::bf16)
     return Op;
 
-  if (SVT.getVectorElementType() == MVT::bf16) {
-    // FIXME: Do we need to support strict FP?
-    assert(!IsStrict && "Strict FP doesn't support BF16");
-    if (VT.getVectorElementType() == MVT::f64) {
-      MVT TmpVT = VT.changeVectorElementType(MVT::f32);
-      return DAG.getNode(ISD::FP_EXTEND, DL, VT,
-                         DAG.getNode(ISD::FP_EXTEND, DL, TmpVT, In));
-    }
-    assert(VT.getVectorElementType() == MVT::f32 && "Unexpected fpext");
-    MVT NVT = SVT.changeVectorElementType(MVT::i32);
-    In = DAG.getBitcast(SVT.changeTypeToInteger(), In);
-    In = DAG.getNode(ISD::ZERO_EXTEND, DL, NVT, In);
-    In = DAG.getNode(ISD::SHL, DL, NVT, In, DAG.getConstant(16, DL, NVT));
-    return DAG.getBitcast(VT, In);
-  }
-
   if (SVT.getVectorElementType() == MVT::f16) {
     if (Subtarget.hasFP16() && isTypeLegal(SVT))
       return Op;
@@ -56517,17 +56501,40 @@ static SDValue combineFP16_TO_FP(SDNode *N, SelectionDAG &DAG,
 
 static SDValue combineFP_EXTEND(SDNode *N, SelectionDAG &DAG,
                                 const X86Subtarget &Subtarget) {
+  EVT VT = N->getValueType(0);
+  bool IsStrict = N->isStrictFPOpcode();
+  SDValue Src = N->getOperand(IsStrict ? 1 : 0);
+  EVT SrcVT = Src.getValueType();
+
+  SDLoc dl(N);
+  if (SrcVT.getScalarType() == MVT::bf16) {
+    if (!IsStrict && Src.getOpcode() == ISD::FP_ROUND &&
+        Src.getOperand(0).getValueType() == VT)
+      return Src.getOperand(0);
+
+    if (!SrcVT.isVector())
+      return SDValue();
+
+    assert(!IsStrict && "Strict FP doesn't support BF16");
+    if (VT.getVectorElementType() == MVT::f64) {
+      MVT TmpVT = VT.getSimpleVT().changeVectorElementType(MVT::f32);
+      return DAG.getNode(ISD::FP_EXTEND, dl, VT,
+                         DAG.getNode(ISD::FP_EXTEND, dl, TmpVT, Src));
+    }
+    assert(VT.getVectorElementType() == MVT::f32 && "Unexpected fpext");
+    MVT NVT = SrcVT.getSimpleVT().changeVectorElementType(MVT::i32);
+    Src = DAG.getBitcast(SrcVT.changeTypeToInteger(), Src);
+    Src = DAG.getNode(ISD::ZERO_EXTEND, dl, NVT, Src);
+    Src = DAG.getNode(ISD::SHL, dl, NVT, Src, DAG.getConstant(16, dl, NVT));
+    return DAG.getBitcast(VT, Src);
+  }
+
   if (!Subtarget.hasF16C() || Subtarget.useSoftFloat())
     return SDValue();
 
   if (Subtarget.hasFP16())
     return SDValue();
 
-  bool IsStrict = N->isStrictFPOpcode();
-  EVT VT = N->getValueType(0);
-  SDValue Src = N->getOperand(IsStrict ? 1 : 0);
-  EVT SrcVT = Src.getValueType();
-
   if (!SrcVT.isVector() || SrcVT.getVectorElementType() != MVT::f16)
     return SDValue();
 
@@ -56539,8 +56546,6 @@ static SDValue combineFP_EXTEND(SDNode *N, SelectionDAG &DAG,
   if (NumElts == 1 || !isPowerOf2_32(NumElts))
     return SDValue();
 
-  SDLoc dl(N);
-
   // Convert the input to vXi16.
   EVT IntVT = SrcVT.changeVectorElementTypeToInteger();
   Src = DAG.getBitcast(IntVT, Src);
diff --git a/llvm/test/CodeGen/X86/bfloat.ll b/llvm/test/CodeGen/X86/bfloat.ll
index 8a2109a1c78df..39d8e2d50c91e 100644
--- a/llvm/test/CodeGen/X86/bfloat.ll
+++ b/llvm/test/CodeGen/X86/bfloat.ll
@@ -1629,22 +1629,8 @@ define <4 x float> @pr64460_1(<4 x bfloat> %a) {
 ;
 ; SSE2-LABEL: pr64460_1:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pextrw $1, %xmm0, %eax
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm2
-; SSE2-NEXT:    movd %xmm0, %eax
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm1
-; SSE2-NEXT:    pextrw $3, %xmm0, %eax
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm2
-; SSE2-NEXT:    movd %xmm0, %eax
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm0
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
@@ -1666,41 +1652,11 @@ define <8 x float> @pr64460_2(<8 x bfloat> %a) {
 ;
 ; SSE2-LABEL: pr64460_2:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq %xmm0, %rdx
-; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
-; SSE2-NEXT:    movq %xmm0, %rcx
-; SSE2-NEXT:    movq %rcx, %rax
-; SSE2-NEXT:    shrq $32, %rax
-; SSE2-NEXT:    movq %rdx, %rsi
-; SSE2-NEXT:    shrq $32, %rsi
-; SSE2-NEXT:    movl %edx, %edi
-; SSE2-NEXT:    andl $-65536, %edi # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    movl %edx, %edi
-; SSE2-NEXT:    shll $16, %edi
-; SSE2-NEXT:    movd %edi, %xmm0
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT:    shrq $48, %rdx
-; SSE2-NEXT:    shll $16, %edx
-; SSE2-NEXT:    movd %edx, %xmm1
-; SSE2-NEXT:    shll $16, %esi
-; SSE2-NEXT:    movd %esi, %xmm2
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE2-NEXT:    movl %ecx, %edx
-; SSE2-NEXT:    andl $-65536, %edx # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %edx, %xmm2
-; SSE2-NEXT:    movl %ecx, %edx
-; SSE2-NEXT:    shll $16, %edx
-; SSE2-NEXT:    movd %edx, %xmm1
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE2-NEXT:    shrq $48, %rcx
-; SSE2-NEXT:    shll $16, %ecx
-; SSE2-NEXT:    movd %ecx, %xmm2
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; SSE2-NEXT:    movdqa %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: pr64460_2:
@@ -1721,76 +1677,16 @@ define <16 x float> @pr64460_3(<16 x bfloat> %a) {
 ;
 ; SSE2-LABEL: pr64460_3:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq %xmm1, %rdi
-; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm1 = xmm1[1,1]
-; SSE2-NEXT:    movq %xmm1, %rcx
-; SSE2-NEXT:    movq %rcx, %rax
-; SSE2-NEXT:    shrq $32, %rax
-; SSE2-NEXT:    movq %xmm0, %r9
-; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
-; SSE2-NEXT:    movq %xmm0, %rsi
-; SSE2-NEXT:    movq %rsi, %rdx
-; SSE2-NEXT:    shrq $32, %rdx
-; SSE2-NEXT:    movq %rdi, %r8
-; SSE2-NEXT:    shrq $32, %r8
-; SSE2-NEXT:    movq %r9, %r10
-; SSE2-NEXT:    shrq $32, %r10
-; SSE2-NEXT:    movl %r9d, %r11d
-; SSE2-NEXT:    andl $-65536, %r11d # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %r11d, %xmm1
-; SSE2-NEXT:    movl %r9d, %r11d
-; SSE2-NEXT:    shll $16, %r11d
-; SSE2-NEXT:    movd %r11d, %xmm0
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT:    shrq $48, %r9
-; SSE2-NEXT:    shll $16, %r9d
-; SSE2-NEXT:    movd %r9d, %xmm1
-; SSE2-NEXT:    shll $16, %r10d
-; SSE2-NEXT:    movd %r10d, %xmm2
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE2-NEXT:    movl %edi, %r9d
-; SSE2-NEXT:    andl $-65536, %r9d # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %r9d, %xmm1
-; SSE2-NEXT:    movl %edi, %r9d
-; SSE2-NEXT:    shll $16, %r9d
-; SSE2-NEXT:    movd %r9d, %xmm2
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE2-NEXT:    shrq $48, %rdi
-; SSE2-NEXT:    shll $16, %edi
-; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    shll $16, %r8d
-; SSE2-NEXT:    movd %r8d, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
-; SSE2-NEXT:    movl %esi, %edi
-; SSE2-NEXT:    andl $-65536, %edi # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %edi, %xmm3
-; SSE2-NEXT:    movl %esi, %edi
-; SSE2-NEXT:    shll $16, %edi
-; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE2-NEXT:    shrq $48, %rsi
-; SSE2-NEXT:    shll $16, %esi
-; SSE2-NEXT:    movd %esi, %xmm3
-; SSE2-NEXT:    shll $16, %edx
-; SSE2-NEXT:    movd %edx, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
-; SSE2-NEXT:    movl %ecx, %edx
-; SSE2-NEXT:    andl $-65536, %edx # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %edx, %xmm4
-; SSE2-NEXT:    movl %ecx, %edx
-; SSE2-NEXT:    shll $16, %edx
-; SSE2-NEXT:    movd %edx, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE2-NEXT:    shrq $48, %rcx
-; SSE2-NEXT:    shll $16, %ecx
-; SSE2-NEXT:    movd %ecx, %xmm4
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm5
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
+; SSE2-NEXT:    pxor %xmm3, %xmm3
+; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE2-NEXT:    pxor %xmm4, %xmm4
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE2-NEXT:    movdqa %xmm5, %xmm0
+; SSE2-NEXT:    movdqa %xmm4, %xmm1
 ; SSE2-NEXT:    retq
 ;
 ; F16-LABEL: pr64460_3:
@@ -1822,47 +1718,17 @@ define <8 x double> @pr64460_4(<8 x bfloat> %a) {
 ;
 ; SSE2-LABEL: pr64460_4:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq %xmm0, %rsi
-; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
-; SSE2-NEXT:    movq %xmm0, %rdx
-; SSE2-NEXT:    movq %rdx, %rax
-; SSE2-NEXT:    shrq $32, %rax
-; SSE2-NEXT:    movq %rdx, %rcx
-; SSE2-NEXT:    shrq $48, %rcx
-; SSE2-NEXT:    movq %rsi, %rdi
-; SSE2-NEXT:    shrq $32, %rdi
-; SSE2-NEXT:    movq %rsi, %r8
-; SSE2-NEXT:    shrq $48, %r8
-; SSE2-NEXT:    movl %esi, %r9d
-; SSE2-NEXT:    andl $-65536, %r9d # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %r9d, %xmm0
-; SSE2-NEXT:    cvtss2sd %xmm0, %xmm1
-; SSE2-NEXT:    shll $16, %esi
-; SSE2-NEXT:    movd %esi, %xmm0
-; SSE2-NEXT:    cvtss2sd %xmm0, %xmm0
-; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE2-NEXT:    shll $16, %r8d
-; SSE2-NEXT:    movd %r8d, %xmm1
-; SSE2-NEXT:    cvtss2sd %xmm1, %xmm2
-; SSE2-NEXT:    shll $16, %edi
-; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    cvtss2sd %xmm1, %xmm1
-; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; SSE2-NEXT:    movl %edx, %esi
-; SSE2-NEXT:    andl $-65536, %esi # imm = 0xFFFF0000
-; SSE2-NEXT:    movd %esi, %xmm2
-; SSE2-NEXT:    cvtss2sd %xmm2, %xmm3
-; SSE2-NEXT:    shll $16, %edx
-; SSE2-NEXT:    movd %edx, %xmm2
-; SSE2-NEXT:    cvtss2sd %xmm2, %xmm2
-; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
-; SSE2-NEXT:    shll $16, %ecx
-; SSE2-NEXT:    movd %ecx, %xmm3
-; SSE2-NEXT:    cvtss2sd %xmm3, %xmm4
-; SSE2-NEXT:    shll $16, %eax
-; SSE2-NEXT:    movd %eax, %xmm3
-; SSE2-NEXT:    cvtss2sd %xmm3, %xmm3
-; SSE2-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; SSE2-NEXT:    pxor %xmm3, %xmm3
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE2-NEXT:    cvtps2pd %xmm1, %xmm4
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; SSE2-NEXT:    cvtps2pd %xmm3, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE2-NEXT:    cvtps2pd %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; SSE2-NEXT:    cvtps2pd %xmm0, %xmm3
+; SSE2-NEXT:    movaps %xmm4, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; F16-LABEL: pr64460_4:
@@ -1874,45 +1740,11 @@ define <8 x double> @pr64460_4(<8 x bfloat> %a) {
 ;
 ; AVXNC-LABEL: pr64460_4:
 ; AVXNC:       # %bb.0:
-; AVXNC-NEXT:    vpextrw $3, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm1
-; AVXNC-NEXT:    vcvtss2sd %xmm1, %xmm1, %xmm1
-; AVXNC-NEXT:    vpextrw $2, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm2
-; AVXNC-NEXT:    vcvtss2sd %xmm2, %xmm2, %xmm2
-; AVXNC-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVXNC-NEXT:    vpextrw $1, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm2
-; AVXNC-NEXT:    vcvtss2sd %xmm2, %xmm2, %xmm2
-; AVXNC-NEXT:    vmovd %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm3
-; AVXNC-NEXT:    vcvtss2sd %xmm3, %xmm3, %xmm3
-; AVXNC-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVXNC-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm2
-; AVXNC-NEXT:    vpextrw $7, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm1
-; AVXNC-NEXT:    vcvtss2sd %xmm1, %xmm1, %xmm1
-; AVXNC-NEXT:    vpextrw $6, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm3
-; AVXNC-NEXT:    vcvtss2sd %xmm3, %xmm3, %xmm3
-; AVXNC-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm1[0]
-; AVXNC-NEXT:    vpextrw $5, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm3
-; AVXNC-NEXT:    vcvtss2sd %xmm3, %xmm3, %xmm3
-; AVXNC-NEXT:    vpextrw $4, %xmm0, %eax
-; AVXNC-NEXT:    shll $16, %eax
-; AVXNC-NEXT:    vmovd %eax, %xmm0
-; AVXNC-NEXT:    vcvtss2sd %xmm0, %xmm0, %xmm0
-; AVXNC-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
-; AVXNC-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVXNC-NEXT:    vmovaps %ymm2, %ymm0
+; AVXNC-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVXNC-NEXT:    vpslld $16, %ymm0, %ymm1
+; AVXNC-NEXT:    vcvtps2pd %xmm1, %ymm0
+; AVXNC-NEXT:    vextracti128 $1, %ymm1, %xmm1
+; AVXNC-NEXT:    vcvtps2pd %xmm1, %ymm1
 ; AVXNC-NEXT:    retq
   %b = fpext <8 x bfloat> %a to <8 x double>
   ret <8 x double> %b
diff --git a/llvm/test/CodeGen/X86/concat-fpext-v2bf16.ll b/llvm/test/CodeGen/X86/concat-fpext-v2bf16.ll
index eff1937b59343..c079a44bc5efd 100644
--- a/llvm/test/CodeGen/X86/concat-fpext-v2bf16.ll
+++ b/llvm/test/CodeGen/X86/concat-fpext-v2bf16.ll
@@ -10,11 +10,11 @@ define void @test(<2 x ptr> %ptr) {
 ; CHECK-NEXT:  # %bb.2: # %loop.127.preheader
 ; CHECK-NEXT:    retq
 ; CHECK-NEXT:  .LBB0_1: # %ifmerge.89
-; CHECK-NEXT:    movzwl (%rax), %eax
-; CHECK-NEXT:    shll $16, %eax
-; CHECK-NEXT:    vmovd %eax, %xmm0
-; CHECK-NEXT:    vmulss %xmm0, %xmm0, %xmm0
 ; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
+; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpbroadcastw (%rax), %xmm2
+; CHECK-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; CHECK-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vmovlps %xmm0, (%rax)
 entry:
   br label %then.13

From df9c00beea5be08fc47aaea58eae333fc6cbef33 Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Sat, 13 Apr 2024 15:07:55 +0800
Subject: [PATCH 328/886]  [CVP] Use at-use info in `processBinOp` (#88523)

This patch uses `getConstantRangeAtUse` to infer nsw/nuw flags with
at-use info. It will enables more optimizations in InstCombine.

Compile-time impact:
http://llvm-compile-time-tracker.com/compare.php?from=a5ed14bc8e122fa5ac0aa81f8d8390931bd6b4e4&to=a83d3402b663439b91cb37a046fb7ac0220ba5c7&stat=instructions%3Au

Related issue: #87854
---
 .../Scalar/CorrelatedValuePropagation.cpp     | 11 +++---
 .../CorrelatedValuePropagation/cond-at-use.ll | 34 +++++++++++++++++++
 .../phi-common-val.ll                         |  3 +-
 3 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Transforms/Scalar/CorrelatedValuePropagation.cpp b/llvm/lib/Transforms/Scalar/CorrelatedValuePropagation.cpp
index fb334e8292eb1..715cdaff97279 100644
--- a/llvm/lib/Transforms/Scalar/CorrelatedValuePropagation.cpp
+++ b/llvm/lib/Transforms/Scalar/CorrelatedValuePropagation.cpp
@@ -1105,13 +1105,10 @@ static bool processBinOp(BinaryOperator *BinOp, LazyValueInfo *LVI) {
     return false;
 
   Instruction::BinaryOps Opcode = BinOp->getOpcode();
-  Value *LHS = BinOp->getOperand(0);
-  Value *RHS = BinOp->getOperand(1);
-
-  ConstantRange LRange =
-      LVI->getConstantRange(LHS, BinOp, /*UndefAllowed*/ false);
-  ConstantRange RRange =
-      LVI->getConstantRange(RHS, BinOp, /*UndefAllowed*/ false);
+  ConstantRange LRange = LVI->getConstantRangeAtUse(BinOp->getOperandUse(0),
+                                                    /*UndefAllowed=*/false);
+  ConstantRange RRange = LVI->getConstantRangeAtUse(BinOp->getOperandUse(1),
+                                                    /*UndefAllowed=*/false);
 
   bool Changed = false;
   bool NewNUW = false, NewNSW = false;
diff --git a/llvm/test/Transforms/CorrelatedValuePropagation/cond-at-use.ll b/llvm/test/Transforms/CorrelatedValuePropagation/cond-at-use.ll
index 0b95139f3dcba..3af4c70a5621c 100644
--- a/llvm/test/Transforms/CorrelatedValuePropagation/cond-at-use.ll
+++ b/llvm/test/Transforms/CorrelatedValuePropagation/cond-at-use.ll
@@ -596,3 +596,37 @@ define i16 @and_elide_poison_flags_missing_noundef(i16 %a) {
   %sel = select i1 %cmp, i16 %and, i16 24
   ret i16 %sel
 }
+
+define i32 @pr87854(i32 noundef %x.1, i32 noundef %i) {
+; CHECK-LABEL: @pr87854(
+; CHECK-NEXT:    [[COND:%.*]] = icmp sgt i32 [[X_1:%.*]], -1
+; CHECK-NEXT:    tail call void @llvm.assume(i1 [[COND]])
+; CHECK-NEXT:    [[INBOUNDS:%.*]] = icmp ult i32 [[I:%.*]], [[X_1]]
+; CHECK-NEXT:    [[NEXT:%.*]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = select i1 [[INBOUNDS]], i32 [[NEXT]], i32 -1
+; CHECK-NEXT:    ret i32 [[SPEC_SELECT]]
+;
+  %cond = icmp sgt i32 %x.1, -1
+  tail call void @llvm.assume(i1 %cond)
+  %inbounds = icmp ult i32 %i, %x.1
+  %next = add i32 %i, 1
+  %spec.select = select i1 %inbounds, i32 %next, i32 -1
+  ret i32 %spec.select
+}
+
+define i64 @test_shl_nsw_at_use(i64 noundef %x) {
+; CHECK-LABEL: @test_shl_nsw_at_use(
+; CHECK-NEXT:    [[ADD:%.*]] = add i64 [[X:%.*]], 2147483648
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i64 [[ADD]], 4294967296
+; CHECK-NEXT:    [[SHL:%.*]] = shl nsw i64 [[X]], 32
+; CHECK-NEXT:    [[SHR:%.*]] = ashr exact i64 [[SHL]], 32
+; CHECK-NEXT:    [[RES:%.*]] = select i1 [[CMP]], i64 [[SHR]], i64 0
+; CHECK-NEXT:    ret i64 [[RES]]
+;
+  %add = add i64 %x, 2147483648
+  %cmp = icmp ult i64 %add, 4294967296
+  %shl = shl i64 %x, 32
+  %shr = ashr exact i64 %shl, 32
+  %res = select i1 %cmp, i64 %shr, i64 0
+  ret i64 %res
+}
diff --git a/llvm/test/Transforms/CorrelatedValuePropagation/phi-common-val.ll b/llvm/test/Transforms/CorrelatedValuePropagation/phi-common-val.ll
index 7f102bc1e4fe2..7e712947e37ad 100644
--- a/llvm/test/Transforms/CorrelatedValuePropagation/phi-common-val.ll
+++ b/llvm/test/Transforms/CorrelatedValuePropagation/phi-common-val.ll
@@ -156,12 +156,13 @@ bb3:
 define i32 @PR43802_without_nowrap(i32 %arg) {
 ; CHECK-LABEL: @PR43802_without_nowrap(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[SUB:%.*]] = sub i32 0, [[ARG:%.*]]
+; CHECK-NEXT:    [[SUB1:%.*]] = sub nsw i32 0, [[ARG:%.*]]
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[ARG]], -2147483648
 ; CHECK-NEXT:    br i1 [[CMP]], label [[BB2:%.*]], label [[BB3:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    br label [[BB3]]
 ; CHECK:       bb3:
+; CHECK-NEXT:    [[SUB:%.*]] = phi i32 [ -2147483648, [[BB2]] ], [ [[SUB1]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    ret i32 [[SUB]]
 ;
 entry:

From 6d66db3890a18e3926a49cbfeb28e99c464cfcd5 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Sat, 13 Apr 2024 08:14:40 +0100
Subject: [PATCH 329/886] [SLP] Initial vectorization of non-power-of-2 ops.
 (#77790)

This patch enables vectorization for non-power-of-2 VFs. Initially only
VFs where adding 1 makes the VF a power-of-2, i.e. we can still make
relatively effective use of the vectors.

It relies on the existing target cost-models to return accurate costs
for
non-power-of-2 vectors. I checked mostly AArch64 and X86 and
there the costs seem reasonable for the costs I checked, although
I expect there will be a need to refine both the cost-models and
lowering
to make most effective use of non-power-of-2 SLP vectorization.

Note that re-ordering and shuffling is not implemented for nodes
requiring padding yet to keep the initial implementation simpler.

The feature is guarded by a new flag, off by defaul for now.

PR: https://github.com/llvm/llvm-project/pull/77790
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    |  83 ++++-
 .../SLPVectorizer/AArch64/vec15-base.ll       |  70 ++--
 .../SLPVectorizer/AArch64/vec3-base.ll        | 247 +++++++++-----
 .../SLPVectorizer/AArch64/vec3-calls.ll       |   3 +-
 .../AArch64/vec3-reorder-reshuffle.ll         | 266 ++++++++++-----
 .../Transforms/SLPVectorizer/X86/odd_store.ll |  84 +++--
 .../Transforms/SLPVectorizer/X86/vec3-base.ll |  35 +-
 .../SLPVectorizer/X86/vec3-calls.ll           |  33 +-
 .../X86/vec3-gather-some-loads.ll             |  74 +++--
 .../X86/vec3-reorder-reshuffle.ll             | 248 ++++++++------
 .../X86/vect_copyable_in_binops.ll            | 303 ++++++++++--------
 11 files changed, 947 insertions(+), 499 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 7b5b0efca8d3f..09db50132c892 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -190,6 +190,10 @@ static cl::opt<bool>
     ViewSLPTree("view-slp-tree", cl::Hidden,
                 cl::desc("Display the SLP trees with Graphviz"));
 
+static cl::opt<bool> VectorizeNonPowerOf2(
+    "slp-vectorize-non-power-of-2", cl::init(false), cl::Hidden,
+    cl::desc("Try to vectorize with non-power-of-2 number of elements."));
+
 // Limit the number of alias checks. The limit is chosen so that
 // it has no negative effect on the llvm benchmarks.
 static const unsigned AliasedCheckLimit = 10;
@@ -2829,6 +2833,14 @@ class BoUpSLP {
                           SmallVectorImpl<Value *> *OpScalars = nullptr,
                           SmallVectorImpl<Value *> *AltScalars = nullptr) const;
 
+    /// Return true if this is a non-power-of-2 node.
+    bool isNonPowOf2Vec() const {
+      bool IsNonPowerOf2 = !isPowerOf2_32(Scalars.size());
+      assert((!IsNonPowerOf2 || ReuseShuffleIndices.empty()) &&
+             "Reshuffling not supported with non-power-of-2 vectors yet.");
+      return IsNonPowerOf2;
+    }
+
 #ifndef NDEBUG
     /// Debug printer.
     LLVM_DUMP_METHOD void dump() const {
@@ -2994,9 +3006,11 @@ class BoUpSLP {
       MustGather.insert(VL.begin(), VL.end());
     }
 
-    if (UserTreeIdx.UserTE)
+    if (UserTreeIdx.UserTE) {
       Last->UserTreeIndices.push_back(UserTreeIdx);
-
+      assert((!Last->isNonPowOf2Vec() || Last->ReorderIndices.empty()) &&
+             "Reordering isn't implemented for non-power-of-2 nodes yet");
+    }
     return Last;
   }
 
@@ -4256,6 +4270,13 @@ BoUpSLP::LoadsState BoUpSLP::canVectorizeLoads(
   auto *VecTy = FixedVectorType::get(ScalarTy, Sz);
   // Check the order of pointer operands or that all pointers are the same.
   bool IsSorted = sortPtrAccesses(PointerOps, ScalarTy, *DL, *SE, Order);
+  // FIXME: Reordering isn't implemented for non-power-of-2 nodes yet.
+  if (!Order.empty() && !isPowerOf2_32(VL.size())) {
+    assert(VectorizeNonPowerOf2 && "non-power-of-2 number of loads only "
+                                   "supported with VectorizeNonPowerOf2");
+    return LoadsState::Gather;
+  }
+
   Align CommonAlignment = computeCommonAlignment<LoadInst>(VL);
   if (!IsSorted && Sz > MinProfitableStridedLoads && TTI->isTypeLegal(VecTy) &&
       TTI->isLegalStridedLoadStore(VecTy, CommonAlignment) &&
@@ -4575,6 +4596,10 @@ static bool areTwoInsertFromSameBuildVector(
 
 std::optional<BoUpSLP::OrdersType>
 BoUpSLP::getReorderingData(const TreeEntry &TE, bool TopToBottom) {
+  // FIXME: Vectorizing is not supported yet for non-power-of-2 ops.
+  if (TE.isNonPowOf2Vec())
+    return std::nullopt;
+
   // No need to reorder if need to shuffle reuses, still need to shuffle the
   // node.
   if (!TE.ReuseShuffleIndices.empty()) {
@@ -5145,6 +5170,10 @@ bool BoUpSLP::canReorderOperands(
     TreeEntry *UserTE, SmallVectorImpl<std::pair<unsigned, TreeEntry *>> &Edges,
     ArrayRef<TreeEntry *> ReorderableGathers,
     SmallVectorImpl<TreeEntry *> &GatherOps) {
+  // FIXME: Reordering isn't implemented for non-power-of-2 nodes yet.
+  if (UserTE->isNonPowOf2Vec())
+    return false;
+
   for (unsigned I = 0, E = UserTE->getNumOperands(); I < E; ++I) {
     if (any_of(Edges, [I](const std::pair<unsigned, TreeEntry *> &OpData) {
           return OpData.first == I &&
@@ -5318,6 +5347,9 @@ void BoUpSLP::reorderBottomToTop(bool IgnoreReorder) {
         }
         auto Res = OrdersUses.insert(std::make_pair(OrdersType(), 0));
         const auto AllowsReordering = [&](const TreeEntry *TE) {
+          // FIXME: Reordering isn't implemented for non-power-of-2 nodes yet.
+          if (TE->isNonPowOf2Vec())
+            return false;
           if (!TE->ReorderIndices.empty() || !TE->ReuseShuffleIndices.empty() ||
               (TE->State == TreeEntry::Vectorize && TE->isAltShuffle()) ||
               (IgnoreReorder && TE->Idx == 0))
@@ -5944,6 +5976,9 @@ BoUpSLP::TreeEntry::EntryState BoUpSLP::getScalarsVectorizationState(
   case Instruction::ExtractValue:
   case Instruction::ExtractElement: {
     bool Reuse = canReuseExtract(VL, VL0, CurrentOrder);
+    // FIXME: Vectorizing is not supported yet for non-power-of-2 ops.
+    if (!isPowerOf2_32(VL.size()))
+      return TreeEntry::NeedToGather;
     if (Reuse || !CurrentOrder.empty())
       return TreeEntry::Vectorize;
     LLVM_DEBUG(dbgs() << "SLP: Gather extract sequence.\n");
@@ -6258,6 +6293,13 @@ void BoUpSLP::buildTree_rec(ArrayRef<Value *> VL, unsigned Depth,
     if (NumUniqueScalarValues == VL.size()) {
       ReuseShuffleIndicies.clear();
     } else {
+      // FIXME: Reshuffing scalars is not supported yet for non-power-of-2 ops.
+      if (UserTreeIdx.UserTE && UserTreeIdx.UserTE->isNonPowOf2Vec()) {
+        LLVM_DEBUG(dbgs() << "SLP: Reshuffling scalars not yet supported "
+                             "for nodes with padding.\n");
+        newTreeEntry(VL, std::nullopt /*not vectorized*/, S, UserTreeIdx);
+        return false;
+      }
       LLVM_DEBUG(dbgs() << "SLP: Shuffle for reused scalars.\n");
       if (NumUniqueScalarValues <= 1 ||
           (UniquePositions.size() == 1 && all_of(UniqueValues,
@@ -7868,7 +7910,8 @@ class BoUpSLP::ShuffleCostEstimator : public BaseShuffleAnalysis {
         for (unsigned I = 0, End = VL.size(); I < End; I += VF) {
           if (VectorizedLoads.contains(VL[I]))
             continue;
-          GatherCost += getBuildVectorCost(VL.slice(I, VF), Root);
+          GatherCost +=
+              getBuildVectorCost(VL.slice(I, std::min(End - I, VF)), Root);
         }
         // Exclude potentially vectorized loads from list of gathered
         // scalars.
@@ -10678,6 +10721,9 @@ BoUpSLP::isGatherShuffledEntry(
   // No need to check for the topmost gather node.
   if (TE == VectorizableTree.front().get())
     return {};
+  // FIXME: Gathering for non-power-of-2 nodes not implemented yet.
+  if (TE->isNonPowOf2Vec())
+    return {};
   Mask.assign(VL.size(), PoisonMaskElem);
   assert(TE->UserTreeIndices.size() == 1 &&
          "Expected only single user of the gather node.");
@@ -14995,8 +15041,13 @@ bool SLPVectorizerPass::vectorizeStoreChain(ArrayRef<Value *> Chain, BoUpSLP &R,
   const unsigned Sz = R.getVectorElementSize(Chain[0]);
   unsigned VF = Chain.size();
 
-  if (!isPowerOf2_32(Sz) || !isPowerOf2_32(VF) || VF < 2 || VF < MinVF)
-    return false;
+  if (!isPowerOf2_32(Sz) || !isPowerOf2_32(VF) || VF < 2 || VF < MinVF) {
+    // Check if vectorizing with a non-power-of-2 VF should be considered. At
+    // the moment, only consider cases where VF + 1 is a power-of-2, i.e. almost
+    // all vector lanes are used.
+    if (!VectorizeNonPowerOf2 || (VF < MinVF && VF + 1 != MinVF))
+      return false;
+  }
 
   LLVM_DEBUG(dbgs() << "SLP: Analyzing " << VF << " stores at offset " << Idx
                     << "\n");
@@ -15095,14 +15146,22 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
         continue;
       }
 
+      unsigned NonPowerOf2VF = 0;
+      if (VectorizeNonPowerOf2) {
+        // First try vectorizing with a non-power-of-2 VF. At the moment, only
+        // consider cases where VF + 1 is a power-of-2, i.e. almost all vector
+        // lanes are used.
+        unsigned CandVF = Operands.size();
+        if (isPowerOf2_32(CandVF + 1) && CandVF <= MaxVF)
+          NonPowerOf2VF = CandVF;
+      }
+
       unsigned Sz = 1 + Log2_32(MaxVF) - Log2_32(MinVF);
-      SmallVector<unsigned> CandidateVFs(Sz);
-      // FIXME: Is division-by-2 the correct step? Should we assert that the
-      // register size is a power-of-2?
-      unsigned Size = MaxVF;
-      for_each(CandidateVFs, [&](unsigned &VF) {
-        VF = Size;
-        Size /= 2;
+      SmallVector<unsigned> CandidateVFs(Sz + (NonPowerOf2VF > 0 ? 1 : 0));
+      unsigned Size = MinVF;
+      for_each(reverse(CandidateVFs), [&](unsigned &VF) {
+        VF = Size > MaxVF ? NonPowerOf2VF : Size;
+        Size *= 2;
       });
       unsigned StartIdx = 0;
       for (unsigned Size : CandidateVFs) {
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/vec15-base.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/vec15-base.ll
index b9e959d50befd..7b27489782fc4 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/vec15-base.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/vec15-base.ll
@@ -1,35 +1,45 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=slp-vectorizer -mtriple=arm64-apple-ios -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=POW2-ONLY %s
 
 define void @v15_load_i8_mul_by_constant_store(ptr %src, ptr noalias %dst) {
-; CHECK-LABEL: define void @v15_load_i8_mul_by_constant_store(
-; CHECK-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 0
-; CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i8>, ptr [[GEP_SRC_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = mul nsw <8 x i8> [[TMP0]], <i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10>
-; CHECK-NEXT:    store <8 x i8> [[TMP1]], ptr [[DST]], align 1
-; CHECK-NEXT:    [[GEP_SRC_8:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 8
-; CHECK-NEXT:    [[DST_8:%.*]] = getelementptr i8, ptr [[DST]], i8 8
-; CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i8>, ptr [[GEP_SRC_8]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = mul nsw <4 x i8> [[TMP2]], <i8 10, i8 10, i8 10, i8 10>
-; CHECK-NEXT:    store <4 x i8> [[TMP3]], ptr [[DST_8]], align 1
-; CHECK-NEXT:    [[GEP_SRC_12:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 12
-; CHECK-NEXT:    [[L_SRC_12:%.*]] = load i8, ptr [[GEP_SRC_12]], align 4
-; CHECK-NEXT:    [[MUL_12:%.*]] = mul nsw i8 [[L_SRC_12]], 10
-; CHECK-NEXT:    [[DST_12:%.*]] = getelementptr i8, ptr [[DST]], i8 12
-; CHECK-NEXT:    store i8 [[MUL_12]], ptr [[DST_12]], align 1
-; CHECK-NEXT:    [[GEP_SRC_13:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 13
-; CHECK-NEXT:    [[L_SRC_13:%.*]] = load i8, ptr [[GEP_SRC_13]], align 4
-; CHECK-NEXT:    [[MUL_13:%.*]] = mul nsw i8 [[L_SRC_13]], 10
-; CHECK-NEXT:    [[DST_13:%.*]] = getelementptr i8, ptr [[DST]], i8 13
-; CHECK-NEXT:    store i8 [[MUL_13]], ptr [[DST_13]], align 1
-; CHECK-NEXT:    [[GEP_SRC_14:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 14
-; CHECK-NEXT:    [[L_SRC_14:%.*]] = load i8, ptr [[GEP_SRC_14]], align 4
-; CHECK-NEXT:    [[MUL_14:%.*]] = mul nsw i8 [[L_SRC_14]], 10
-; CHECK-NEXT:    [[DST_14:%.*]] = getelementptr i8, ptr [[DST]], i8 14
-; CHECK-NEXT:    store i8 [[MUL_14]], ptr [[DST_14]], align 1
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @v15_load_i8_mul_by_constant_store(
+; NON-POW2-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <15 x i8>, ptr [[GEP_SRC_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = mul nsw <15 x i8> [[TMP0]], <i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10>
+; NON-POW2-NEXT:    store <15 x i8> [[TMP1]], ptr [[DST]], align 1
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @v15_load_i8_mul_by_constant_store(
+; POW2-ONLY-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 0
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <8 x i8>, ptr [[GEP_SRC_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = mul nsw <8 x i8> [[TMP0]], <i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10, i8 10>
+; POW2-ONLY-NEXT:    store <8 x i8> [[TMP1]], ptr [[DST]], align 1
+; POW2-ONLY-NEXT:    [[GEP_SRC_8:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 8
+; POW2-ONLY-NEXT:    [[DST_8:%.*]] = getelementptr i8, ptr [[DST]], i8 8
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = load <4 x i8>, ptr [[GEP_SRC_8]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = mul nsw <4 x i8> [[TMP2]], <i8 10, i8 10, i8 10, i8 10>
+; POW2-ONLY-NEXT:    store <4 x i8> [[TMP3]], ptr [[DST_8]], align 1
+; POW2-ONLY-NEXT:    [[GEP_SRC_12:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 12
+; POW2-ONLY-NEXT:    [[L_SRC_12:%.*]] = load i8, ptr [[GEP_SRC_12]], align 4
+; POW2-ONLY-NEXT:    [[MUL_12:%.*]] = mul nsw i8 [[L_SRC_12]], 10
+; POW2-ONLY-NEXT:    [[DST_12:%.*]] = getelementptr i8, ptr [[DST]], i8 12
+; POW2-ONLY-NEXT:    store i8 [[MUL_12]], ptr [[DST_12]], align 1
+; POW2-ONLY-NEXT:    [[GEP_SRC_13:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 13
+; POW2-ONLY-NEXT:    [[L_SRC_13:%.*]] = load i8, ptr [[GEP_SRC_13]], align 4
+; POW2-ONLY-NEXT:    [[MUL_13:%.*]] = mul nsw i8 [[L_SRC_13]], 10
+; POW2-ONLY-NEXT:    [[DST_13:%.*]] = getelementptr i8, ptr [[DST]], i8 13
+; POW2-ONLY-NEXT:    store i8 [[MUL_13]], ptr [[DST_13]], align 1
+; POW2-ONLY-NEXT:    [[GEP_SRC_14:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i8 14
+; POW2-ONLY-NEXT:    [[L_SRC_14:%.*]] = load i8, ptr [[GEP_SRC_14]], align 4
+; POW2-ONLY-NEXT:    [[MUL_14:%.*]] = mul nsw i8 [[L_SRC_14]], 10
+; POW2-ONLY-NEXT:    [[DST_14:%.*]] = getelementptr i8, ptr [[DST]], i8 14
+; POW2-ONLY-NEXT:    store i8 [[MUL_14]], ptr [[DST_14]], align 1
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %gep.src.0 = getelementptr inbounds i8, ptr %src, i8 0
@@ -123,5 +133,3 @@ entry:
 
   ret void
 }
-
-
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-base.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-base.ll
index 59ffbf7ef9b24..c18811a35c1ee 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-base.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-base.ll
@@ -1,16 +1,69 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=slp-vectorizer -mtriple=arm64-apple-ios -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 define void @v3_load_i32_mul_by_constant_store(ptr %src, ptr %dst) {
-; CHECK-LABEL: @v3_load_i32_mul_by_constant_store(
+; NON-POW2-LABEL: @v3_load_i32_mul_by_constant_store(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <3 x i32>, ptr [[GEP_SRC_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = mul nsw <3 x i32> [[TMP0]], <i32 10, i32 10, i32 10>
+; NON-POW2-NEXT:    store <3 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @v3_load_i32_mul_by_constant_store(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_2:%.*]] = load i32, ptr [[GEP_SRC_2]], align 4
+; POW2-ONLY-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_2]], 10
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = mul nsw <2 x i32> [[TMP0]], <i32 10, i32 10>
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
+; POW2-ONLY-NEXT:    store i32 [[MUL_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
+;
+entry:
+  %gep.src.0 = getelementptr inbounds i32, ptr %src, i32 0
+  %l.src.0 = load i32, ptr %gep.src.0, align 4
+  %mul.0 = mul nsw i32 %l.src.0, 10
+
+  %gep.src.1 = getelementptr inbounds i32, ptr %src, i32 1
+  %l.src.1 = load i32, ptr %gep.src.1, align 4
+  %mul.1 = mul nsw i32 %l.src.1, 10
+
+  %gep.src.2 = getelementptr inbounds i32, ptr %src, i32 2
+  %l.src.2 = load i32, ptr %gep.src.2, align 4
+  %mul.2 = mul nsw i32 %l.src.2, 10
+
+  store i32 %mul.0, ptr %dst
+
+  %dst.1 = getelementptr i32, ptr %dst, i32 1
+  store i32 %mul.1, ptr %dst.1
+
+  %dst.2 = getelementptr i32, ptr %dst, i32 2
+  store i32 %mul.2, ptr %dst.2
+
+  ret void
+}
+
+; Should no be vectorized with a undef/poison element as padding, as division by undef/poison may cause UB.
+define void @v3_load_i32_udiv_by_constant_store(ptr %src, ptr %dst) {
+; CHECK-LABEL: @v3_load_i32_udiv_by_constant_store(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 0
+; CHECK-NEXT:    [[L_SRC_0:%.*]] = load i32, ptr [[GEP_SRC_0]], align 4
+; CHECK-NEXT:    [[MUL_0:%.*]] = udiv i32 10, [[L_SRC_0]]
+; CHECK-NEXT:    [[GEP_SRC_1:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 1
+; CHECK-NEXT:    [[L_SRC_1:%.*]] = load i32, ptr [[GEP_SRC_1]], align 4
+; CHECK-NEXT:    [[MUL_1:%.*]] = udiv i32 10, [[L_SRC_1]]
 ; CHECK-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 2
 ; CHECK-NEXT:    [[L_SRC_2:%.*]] = load i32, ptr [[GEP_SRC_2]], align 4
-; CHECK-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_2]], 10
-; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = mul nsw <2 x i32> [[TMP0]], <i32 10, i32 10>
-; CHECK-NEXT:    store <2 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[MUL_2:%.*]] = udiv i32 10, [[L_SRC_2]]
+; CHECK-NEXT:    store i32 [[MUL_0]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[DST_1:%.*]] = getelementptr i32, ptr [[DST]], i32 1
+; CHECK-NEXT:    store i32 [[MUL_1]], ptr [[DST_1]], align 4
 ; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
 ; CHECK-NEXT:    store i32 [[MUL_2]], ptr [[DST_2]], align 4
 ; CHECK-NEXT:    ret void
@@ -18,15 +71,15 @@ define void @v3_load_i32_mul_by_constant_store(ptr %src, ptr %dst) {
 entry:
   %gep.src.0 = getelementptr inbounds i32, ptr %src, i32 0
   %l.src.0 = load i32, ptr %gep.src.0, align 4
-  %mul.0 = mul nsw i32 %l.src.0, 10
+  %mul.0 = udiv i32 10, %l.src.0
 
   %gep.src.1 = getelementptr inbounds i32, ptr %src, i32 1
   %l.src.1 = load i32, ptr %gep.src.1, align 4
-  %mul.1 = mul nsw i32 %l.src.1, 10
+  %mul.1 = udiv i32 10, %l.src.1
 
   %gep.src.2 = getelementptr inbounds i32, ptr %src, i32 2
   %l.src.2 = load i32, ptr %gep.src.2, align 4
-  %mul.2 = mul nsw i32 %l.src.2, 10
+  %mul.2 = udiv i32 10, %l.src.2
 
   store i32 %mul.0, ptr %dst
 
@@ -39,23 +92,35 @@ entry:
   ret void
 }
 
+
+
 define void @v3_load_i32_mul_store(ptr %src.1, ptr %src.2, ptr %dst) {
-; CHECK-LABEL: @v3_load_i32_mul_store(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_1_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_1]], i32 2
-; CHECK-NEXT:    [[L_SRC_1_2:%.*]] = load i32, ptr [[GEP_SRC_1_2]], align 4
-; CHECK-NEXT:    [[GEP_SRC_2_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_2]], i32 2
-; CHECK-NEXT:    [[L_SRC_2_2:%.*]] = load i32, ptr [[GEP_SRC_2_2]], align 4
-; CHECK-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_1_2]], [[L_SRC_2_2]]
-; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_1_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[GEP_SRC_2_0]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP0]], [[TMP1]]
-; CHECK-NEXT:    store <2 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
-; CHECK-NEXT:    store i32 [[MUL_2]], ptr [[DST_2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @v3_load_i32_mul_store(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
+; NON-POW2-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <3 x i32>, ptr [[GEP_SRC_1_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x i32>, ptr [[GEP_SRC_2_0]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = mul nsw <3 x i32> [[TMP0]], [[TMP1]]
+; NON-POW2-NEXT:    store <3 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @v3_load_i32_mul_store(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_1_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_1]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_1_2:%.*]] = load i32, ptr [[GEP_SRC_1_2]], align 4
+; POW2-ONLY-NEXT:    [[GEP_SRC_2_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_2]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_2_2:%.*]] = load i32, ptr [[GEP_SRC_2_2]], align 4
+; POW2-ONLY-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_1_2]], [[L_SRC_2_2]]
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_1_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[GEP_SRC_2_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP0]], [[TMP1]]
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
+; POW2-ONLY-NEXT:    store i32 [[MUL_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %gep.src.1.0 = getelementptr inbounds i32, ptr %src.1, i32 0
@@ -88,24 +153,35 @@ entry:
 }
 
 define void @v3_load_i32_mul_add_const_store(ptr %src.1, ptr %src.2, ptr %dst) {
-; CHECK-LABEL: @v3_load_i32_mul_add_const_store(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_1_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_1]], i32 2
-; CHECK-NEXT:    [[L_SRC_1_2:%.*]] = load i32, ptr [[GEP_SRC_1_2]], align 4
-; CHECK-NEXT:    [[GEP_SRC_2_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_2]], i32 2
-; CHECK-NEXT:    [[L_SRC_2_2:%.*]] = load i32, ptr [[GEP_SRC_2_2]], align 4
-; CHECK-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_1_2]], [[L_SRC_2_2]]
-; CHECK-NEXT:    [[ADD_2:%.*]] = add i32 [[MUL_2]], 9
-; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_1_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[GEP_SRC_2_0]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP0]], [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = add <2 x i32> [[TMP2]], <i32 9, i32 9>
-; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
-; CHECK-NEXT:    store i32 [[ADD_2]], ptr [[DST_2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @v3_load_i32_mul_add_const_store(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
+; NON-POW2-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <3 x i32>, ptr [[GEP_SRC_1_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x i32>, ptr [[GEP_SRC_2_0]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = mul nsw <3 x i32> [[TMP0]], [[TMP1]]
+; NON-POW2-NEXT:    [[TMP3:%.*]] = add <3 x i32> [[TMP2]], <i32 9, i32 9, i32 9>
+; NON-POW2-NEXT:    store <3 x i32> [[TMP3]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @v3_load_i32_mul_add_const_store(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_1_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_1:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_2_0:%.*]] = getelementptr inbounds i32, ptr [[SRC_2:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_1_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_1]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_1_2:%.*]] = load i32, ptr [[GEP_SRC_1_2]], align 4
+; POW2-ONLY-NEXT:    [[GEP_SRC_2_2:%.*]] = getelementptr inbounds i32, ptr [[SRC_2]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_2_2:%.*]] = load i32, ptr [[GEP_SRC_2_2]], align 4
+; POW2-ONLY-NEXT:    [[MUL_2:%.*]] = mul nsw i32 [[L_SRC_1_2]], [[L_SRC_2_2]]
+; POW2-ONLY-NEXT:    [[ADD_2:%.*]] = add i32 [[MUL_2]], 9
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[GEP_SRC_1_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[GEP_SRC_2_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP0]], [[TMP1]]
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = add <2 x i32> [[TMP2]], <i32 9, i32 9>
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP3]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST]], i32 2
+; POW2-ONLY-NEXT:    store i32 [[ADD_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %gep.src.1.0 = getelementptr inbounds i32, ptr %src.1, i32 0
@@ -141,18 +217,26 @@ entry:
 }
 
 define void @v3_load_f32_fadd_fadd_by_constant_store(ptr %src, ptr %dst) {
-; CHECK-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds float, ptr [[SRC]], i32 2
-; CHECK-NEXT:    [[L_SRC_2:%.*]] = load float, ptr [[GEP_SRC_2]], align 4
-; CHECK-NEXT:    [[FADD_2:%.*]] = fadd float [[L_SRC_2]], 1.000000e+01
-; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[GEP_SRC_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01>
-; CHECK-NEXT:    store <2 x float> [[TMP1]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr float, ptr [[DST]], i32 2
-; CHECK-NEXT:    store float [[FADD_2]], ptr [[DST_2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <3 x float>, ptr [[GEP_SRC_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = fadd <3 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>
+; NON-POW2-NEXT:    store <3 x float> [[TMP1]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds float, ptr [[SRC]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_2:%.*]] = load float, ptr [[GEP_SRC_2]], align 4
+; POW2-ONLY-NEXT:    [[FADD_2:%.*]] = fadd float [[L_SRC_2]], 1.000000e+01
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[GEP_SRC_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = fadd <2 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP1]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr float, ptr [[DST]], i32 2
+; POW2-ONLY-NEXT:    store float [[FADD_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %gep.src.0 = getelementptr inbounds float, ptr %src, i32 0
@@ -179,18 +263,28 @@ entry:
 }
 
 define void @phi_store3(ptr %dst) {
-; CHECK-LABEL: @phi_store3(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br label [[EXIT:%.*]]
-; CHECK:       invoke.cont8.loopexit:
-; CHECK-NEXT:    br label [[EXIT]]
-; CHECK:       exit:
-; CHECK-NEXT:    [[P_2:%.*]] = phi i32 [ 3, [[ENTRY:%.*]] ], [ 0, [[INVOKE_CONT8_LOOPEXIT:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[ENTRY]] ], [ poison, [[INVOKE_CONT8_LOOPEXIT]] ]
-; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 2
-; CHECK-NEXT:    store <2 x i32> [[TMP0]], ptr [[DST]], align 4
-; CHECK-NEXT:    store i32 [[P_2]], ptr [[DST_2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @phi_store3(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    br label [[EXIT:%.*]]
+; NON-POW2:       invoke.cont8.loopexit:
+; NON-POW2-NEXT:    br label [[EXIT]]
+; NON-POW2:       exit:
+; NON-POW2-NEXT:    [[TMP0:%.*]] = phi <3 x i32> [ <i32 1, i32 2, i32 3>, [[ENTRY:%.*]] ], [ poison, [[INVOKE_CONT8_LOOPEXIT:%.*]] ]
+; NON-POW2-NEXT:    store <3 x i32> [[TMP0]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @phi_store3(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    br label [[EXIT:%.*]]
+; POW2-ONLY:       invoke.cont8.loopexit:
+; POW2-ONLY-NEXT:    br label [[EXIT]]
+; POW2-ONLY:       exit:
+; POW2-ONLY-NEXT:    [[P_2:%.*]] = phi i32 [ 3, [[ENTRY:%.*]] ], [ 0, [[INVOKE_CONT8_LOOPEXIT:%.*]] ]
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[ENTRY]] ], [ poison, [[INVOKE_CONT8_LOOPEXIT]] ]
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 2
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP0]], ptr [[DST]], align 4
+; POW2-ONLY-NEXT:    store i32 [[P_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   br label %exit
@@ -213,13 +307,18 @@ exit:
 }
 
 define void @store_try_reorder(ptr %dst) {
-; CHECK-LABEL: @store_try_reorder(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[ADD:%.*]] = add i32 0, 0
-; CHECK-NEXT:    store i32 [[ADD]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    [[ARRAYIDX_I1887:%.*]] = getelementptr i32, ptr [[DST]], i64 1
-; CHECK-NEXT:    store <2 x i32> zeroinitializer, ptr [[ARRAYIDX_I1887]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @store_try_reorder(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    store <3 x i32> zeroinitializer, ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @store_try_reorder(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[ADD:%.*]] = add i32 0, 0
+; POW2-ONLY-NEXT:    store i32 [[ADD]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX_I1887:%.*]] = getelementptr i32, ptr [[DST]], i64 1
+; POW2-ONLY-NEXT:    store <2 x i32> zeroinitializer, ptr [[ARRAYIDX_I1887]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %add = add i32 0, 0
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-calls.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-calls.ll
index 2cb84eeb7fc8f..67746f2cbf5d2 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-calls.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-calls.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=slp-vectorizer -mtriple=arm64-apple-ios -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK %s
 
 define void @vec3_vectorize_call(ptr %Colour, float %0) {
 ; CHECK-LABEL: @vec3_vectorize_call(
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-reorder-reshuffle.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-reorder-reshuffle.ll
index e4925961f284f..47d918eabdfe2 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-reorder-reshuffle.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/vec3-reorder-reshuffle.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=slp-vectorizer -mtriple=arm64-apple-ios -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=arm64-apple-ios -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 %struct.zot = type { i32, i32, i32 }
 
@@ -139,21 +140,34 @@ if.end668:                                        ; preds = %if.then665, %entry
 }
 
 define void @gather_2(ptr %mat1, float %0, float %1) {
-; CHECK-LABEL: define void @gather_2(
-; CHECK-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x float> <float 0.000000e+00, float poison>, float [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP6]], <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP4:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP1]], float 0.000000e+00)
-; CHECK-NEXT:    [[TMP7:%.*]] = fmul float [[TMP4]], 0.000000e+00
-; CHECK-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
-; CHECK-NEXT:    [[ARRAYIDX5_I_I_I280:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1, i64 2
-; CHECK-NEXT:    [[TMP8:%.*]] = fmul <2 x float> [[TMP5]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
-; CHECK-NEXT:    store float [[TMP7]], ptr [[ARRAYIDX5_I_I_I280]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @gather_2(
+; NON-POW2-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP3:%.*]] = shufflevector <3 x float> [[TMP2]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP4:%.*]] = insertelement <3 x float> <float 0.000000e+00, float poison, float poison>, float [[TMP1]], i32 1
+; NON-POW2-NEXT:    [[TMP5:%.*]] = shufflevector <3 x float> [[TMP4]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
+; NON-POW2-NEXT:    [[TMP6:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP3]], <3 x float> [[TMP5]], <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
+; NON-POW2-NEXT:    [[TMP7:%.*]] = fmul <3 x float> [[TMP6]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP7]], ptr [[ARRAYIDX163]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @gather_2(
+; POW2-ONLY-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = insertelement <2 x float> <float 0.000000e+00, float poison>, float [[TMP1]], i32 1
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP1]], float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = fmul float [[TMP6]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
+; POW2-ONLY-NEXT:    [[ARRAYIDX5_I_I_I280:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1, i64 2
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = fmul <2 x float> [[TMP5]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
+; POW2-ONLY-NEXT:    store float [[TMP7]], ptr [[ARRAYIDX5_I_I_I280]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %2 = call float @llvm.fmuladd.f32(float %0, float 0.000000e+00, float 0.000000e+00)
@@ -172,31 +186,47 @@ entry:
 }
 
 define i32 @reorder_indices_1(float %0) {
-; CHECK-LABEL: define i32 @reorder_indices_1(
-; CHECK-SAME: float [[TMP0:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
-; CHECK-NEXT:    [[ARRAYIDX2_I265:%.*]] = getelementptr float, ptr [[NOR1]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2_I265]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[NOR1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = fneg float [[TMP3]]
-; CHECK-NEXT:    [[NEG11_I:%.*]] = fmul float [[TMP4]], [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = call float @llvm.fmuladd.f32(float [[TMP1]], float 0.000000e+00, float [[NEG11_I]])
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 poison>
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP8:%.*]] = fneg <2 x float> [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]]
-; CHECK-NEXT:    [[TMP12:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> zeroinitializer, <2 x float> [[TMP11]])
-; CHECK-NEXT:    [[TMP13:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP10]], <2 x float> [[TMP12]], <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP14:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP5]], float 0.000000e+00)
-; CHECK-NEXT:    [[TMP15:%.*]] = fmul <2 x float> [[TMP13]], zeroinitializer
-; CHECK-NEXT:    [[MUL6_I_I_I:%.*]] = fmul float [[TMP14]], 0.000000e+00
-; CHECK-NEXT:    store <2 x float> [[TMP15]], ptr [[NOR1]], align 4
-; CHECK-NEXT:    store float [[MUL6_I_I_I]], ptr [[ARRAYIDX2_I265]], align 4
-; CHECK-NEXT:    ret i32 0
+; NON-POW2-LABEL: define i32 @reorder_indices_1(
+; NON-POW2-SAME: float [[TMP0:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x float>, ptr [[NOR1]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = shufflevector <3 x float> [[TMP1]], <3 x float> poison, <3 x i32> <i32 1, i32 2, i32 0>
+; NON-POW2-NEXT:    [[TMP3:%.*]] = fneg <3 x float> [[TMP2]]
+; NON-POW2-NEXT:    [[TMP4:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP5:%.*]] = shufflevector <3 x float> [[TMP4]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP6:%.*]] = fmul <3 x float> [[TMP3]], [[TMP5]]
+; NON-POW2-NEXT:    [[TMP7:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP1]], <3 x float> zeroinitializer, <3 x float> [[TMP6]])
+; NON-POW2-NEXT:    [[TMP8:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP5]], <3 x float> [[TMP7]], <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[TMP9:%.*]] = fmul <3 x float> [[TMP8]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP9]], ptr [[NOR1]], align 4
+; NON-POW2-NEXT:    ret i32 0
+;
+; POW2-ONLY-LABEL: define i32 @reorder_indices_1(
+; POW2-ONLY-SAME: float [[TMP0:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX2_I265:%.*]] = getelementptr float, ptr [[NOR1]], i64 2
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2_I265]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[NOR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = fneg float [[TMP3]]
+; POW2-ONLY-NEXT:    [[NEG11_I:%.*]] = fmul float [[TMP4]], [[TMP0]]
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = call float @llvm.fmuladd.f32(float [[TMP1]], float 0.000000e+00, float [[NEG11_I]])
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 poison>
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = fneg <2 x float> [[TMP7]]
+; POW2-ONLY-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]]
+; POW2-ONLY-NEXT:    [[TMP12:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> zeroinitializer, <2 x float> [[TMP11]])
+; POW2-ONLY-NEXT:    [[TMP13:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP10]], <2 x float> [[TMP12]], <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP14:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP5]], float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[TMP15:%.*]] = fmul <2 x float> [[TMP13]], zeroinitializer
+; POW2-ONLY-NEXT:    [[MUL6_I_I_I:%.*]] = fmul float [[TMP14]], 0.000000e+00
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP15]], ptr [[NOR1]], align 4
+; POW2-ONLY-NEXT:    store float [[MUL6_I_I_I]], ptr [[ARRAYIDX2_I265]], align 4
+; POW2-ONLY-NEXT:    ret i32 0
 ;
 entry:
   %nor1 = alloca [0 x [3 x float]], i32 0, align 4
@@ -227,19 +257,28 @@ entry:
 }
 
 define void @reorder_indices_2(ptr %spoint) {
-; CHECK-LABEL: define void @reorder_indices_2(
-; CHECK-SAME: ptr [[SPOINT:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> zeroinitializer, i64 0
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 0.000000e+00, float 0.000000e+00)
-; CHECK-NEXT:    [[MUL4_I461:%.*]] = fmul float [[TMP1]], 0.000000e+00
-; CHECK-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer, <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP3:%.*]] = fmul <2 x float> [[TMP2]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[DSCO]], align 4
-; CHECK-NEXT:    [[ARRAYIDX5_I476:%.*]] = getelementptr float, ptr [[SPOINT]], i64 2
-; CHECK-NEXT:    store float [[MUL4_I461]], ptr [[ARRAYIDX5_I476]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @reorder_indices_2(
+; NON-POW2-SAME: ptr [[SPOINT:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> zeroinitializer, <3 x float> zeroinitializer, <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[TMP1:%.*]] = fmul <3 x float> [[TMP0]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP1]], ptr [[DSCO]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @reorder_indices_2(
+; POW2-ONLY-SAME: ptr [[SPOINT:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> zeroinitializer, i64 0
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 0.000000e+00, float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[MUL4_I461:%.*]] = fmul float [[TMP1]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer, <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = fmul <2 x float> [[TMP2]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP3]], ptr [[DSCO]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX5_I476:%.*]] = getelementptr float, ptr [[SPOINT]], i64 2
+; POW2-ONLY-NEXT:    store float [[MUL4_I461]], ptr [[ARRAYIDX5_I476]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %0 = extractelement <3 x float> zeroinitializer, i64 1
@@ -291,19 +330,30 @@ entry:
 }
 
 define void @reuse_shuffle_indidces_1(ptr %col, float %0, float %1) {
-; CHECK-LABEL: define void @reuse_shuffle_indidces_1(
-; CHECK-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP0]], i32 1
-; CHECK-NEXT:    [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd <2 x float> [[TMP4]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP5]], ptr [[COL]], align 4
-; CHECK-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr float, ptr [[COL]], i64 2
-; CHECK-NEXT:    [[MUL38:%.*]] = fmul float [[TMP0]], 0.000000e+00
-; CHECK-NEXT:    [[TMP6:%.*]] = fadd float [[MUL38]], 0.000000e+00
-; CHECK-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX33]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @reuse_shuffle_indidces_1(
+; NON-POW2-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP1]], i32 0
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <3 x float> [[TMP2]], float [[TMP0]], i32 1
+; NON-POW2-NEXT:    [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
+; NON-POW2-NEXT:    [[TMP5:%.*]] = fmul <3 x float> [[TMP4]], zeroinitializer
+; NON-POW2-NEXT:    [[TMP6:%.*]] = fadd <3 x float> [[TMP5]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP6]], ptr [[COL]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @reuse_shuffle_indidces_1(
+; POW2-ONLY-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP0]], i32 1
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = fadd <2 x float> [[TMP4]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP5]], ptr [[COL]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr float, ptr [[COL]], i64 2
+; POW2-ONLY-NEXT:    [[MUL38:%.*]] = fmul float [[TMP0]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = fadd float [[MUL38]], 0.000000e+00
+; POW2-ONLY-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX33]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %mul24 = fmul float %1, 0.000000e+00
@@ -487,18 +537,24 @@ entry:
 }
 
 define void @vec3_extract(<3 x i16> %pixel.sroa.0.4.vec.insert606, ptr %call3.i536) {
-; CHECK-LABEL: define void @vec3_extract(
-; CHECK-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[PIXEL_SROA_0_4_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 2
-; CHECK-NEXT:    [[RED668:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 2
-; CHECK-NEXT:    store i16 [[PIXEL_SROA_0_4_VEC_EXTRACT]], ptr [[RED668]], align 2
-; CHECK-NEXT:    [[PIXEL_SROA_0_2_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 1
-; CHECK-NEXT:    [[GREEN670:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 1
-; CHECK-NEXT:    store i16 [[PIXEL_SROA_0_2_VEC_EXTRACT]], ptr [[GREEN670]], align 2
-; CHECK-NEXT:    [[PIXEL_SROA_0_0_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 0
-; CHECK-NEXT:    store i16 [[PIXEL_SROA_0_0_VEC_EXTRACT]], ptr [[CALL3_I536]], align 2
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @vec3_extract(
+; NON-POW2-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    store <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], ptr [[CALL3_I536]], align 2
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @vec3_extract(
+; POW2-ONLY-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[PIXEL_SROA_0_4_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 2
+; POW2-ONLY-NEXT:    [[RED668:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 2
+; POW2-ONLY-NEXT:    store i16 [[PIXEL_SROA_0_4_VEC_EXTRACT]], ptr [[RED668]], align 2
+; POW2-ONLY-NEXT:    [[PIXEL_SROA_0_2_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 1
+; POW2-ONLY-NEXT:    [[GREEN670:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 1
+; POW2-ONLY-NEXT:    store i16 [[PIXEL_SROA_0_2_VEC_EXTRACT]], ptr [[GREEN670]], align 2
+; POW2-ONLY-NEXT:    [[PIXEL_SROA_0_0_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 0
+; POW2-ONLY-NEXT:    store i16 [[PIXEL_SROA_0_0_VEC_EXTRACT]], ptr [[CALL3_I536]], align 2
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %pixel.sroa.0.4.vec.extract = extractelement <3 x i16> %pixel.sroa.0.4.vec.insert606, i64 2
@@ -512,4 +568,56 @@ entry:
   ret void
 }
 
+define void @can_reorder_vec3_op_with_padding(ptr %A, <3 x float> %in) {
+; NON-POW2-LABEL: define void @can_reorder_vec3_op_with_padding(
+; NON-POW2-SAME: ptr [[A:%.*]], <3 x float> [[IN:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP0:%.*]] = shufflevector <3 x float> [[IN]], <3 x float> poison, <3 x i32> <i32 1, i32 2, i32 0>
+; NON-POW2-NEXT:    [[TMP1:%.*]] = fsub <3 x float> [[TMP0]], [[TMP0]]
+; NON-POW2-NEXT:    [[TMP2:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP1]], <3 x float> <float 2.000000e+00, float 2.000000e+00, float 2.000000e+00>, <3 x float> <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>)
+; NON-POW2-NEXT:    [[TMP3:%.*]] = fmul <3 x float> [[TMP2]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>
+; NON-POW2-NEXT:    store <3 x float> [[TMP3]], ptr [[A]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @can_reorder_vec3_op_with_padding(
+; POW2-ONLY-SAME: ptr [[A:%.*]], <3 x float> [[IN:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[ARRAYIDX42_I:%.*]] = getelementptr float, ptr [[A]], i64 2
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> [[IN]], i64 0
+; POW2-ONLY-NEXT:    [[SUB_I362:%.*]] = fsub float [[TMP0]], [[TMP0]]
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = call float @llvm.fmuladd.f32(float [[SUB_I362]], float 2.000000e+00, float 3.000000e+00)
+; POW2-ONLY-NEXT:    [[MUL6_I_I_I_I:%.*]] = fmul float [[TMP1]], 3.000000e+00
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = shufflevector <3 x float> [[IN]], <3 x float> poison, <2 x i32> <i32 1, i32 2>
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = fsub <2 x float> [[TMP2]], [[TMP2]]
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> <float 2.000000e+00, float 2.000000e+00>, <2 x float> <float 3.000000e+00, float 3.000000e+00>)
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = fmul <2 x float> [[TMP4]], <float 3.000000e+00, float 3.000000e+00>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP5]], ptr [[A]], align 4
+; POW2-ONLY-NEXT:    store float [[MUL6_I_I_I_I]], ptr [[ARRAYIDX42_I]], align 4
+; POW2-ONLY-NEXT:    ret void
+;
+entry:
+  %arrayidx42.i = getelementptr float, ptr %A, i64 2
+  %arrayidx35.i = getelementptr float, ptr %A, i64 1
+  %0 = extractelement <3 x float> %in, i64 0
+  %1 = extractelement <3 x float> %in, i64 0
+  %sub.i362 = fsub float %0, %1
+  %2 = extractelement <3 x float> %in, i64 1
+  %3 = extractelement <3 x float> %in, i64 1
+  %sub5.i = fsub float %2, %3
+  %4 = extractelement <3 x float> %in, i64 2
+  %5 = extractelement <3 x float> %in, i64 2
+  %sub9.i = fsub float %4, %5
+  %6 = call float @llvm.fmuladd.f32(float %sub5.i, float 2.000000e+00, float 3.000000e+00)
+  %7 = call float @llvm.fmuladd.f32(float %sub9.i, float 2.000000e+00, float 3.000000e+00)
+  %8 = call float @llvm.fmuladd.f32(float %sub.i362, float 2.000000e+00, float 3.000000e+00)
+  %mul.i.i.i.i373 = fmul float %6, 3.000000e+00
+  %mul3.i.i.i.i = fmul float %7, 3.000000e+00
+  %mul6.i.i.i.i = fmul float %8, 3.000000e+00
+  store float %mul.i.i.i.i373, ptr %A, align 4
+  store float %mul3.i.i.i.i, ptr %arrayidx35.i, align 4
+  store float %mul6.i.i.i.i, ptr %arrayidx42.i, align 4
+  ret void
+}
+
 declare float @llvm.fmuladd.f32(float, float, float)
+declare double @llvm.fmuladd.f64(double, double, double)
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/odd_store.ll b/llvm/test/Transforms/SLPVectorizer/X86/odd_store.ll
index 4795ac6559203..5f2c42d5c2dec 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/odd_store.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/odd_store.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -passes=slp-vectorizer,dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+; RUN: opt < %s -passes=slp-vectorizer,dce -slp-vectorize-non-power-of-2 -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt < %s -passes=slp-vectorizer,dce -slp-vectorize-non-power-of-2=false -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 ;int foo(char * restrict A, ptr restrict B, float T) {
 ;  A[0] = (T * B[10] + 4.0);
@@ -8,31 +9,43 @@
 ;}
 
 define i32 @foo(ptr noalias nocapture %A, ptr noalias nocapture %B, float %T) {
-; CHECK-LABEL: @foo(
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 10
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = fmul float [[TMP2]], [[T:%.*]]
-; CHECK-NEXT:    [[TMP4:%.*]] = fpext float [[TMP3]] to double
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd double [[TMP4]], 4.000000e+00
-; CHECK-NEXT:    [[TMP6:%.*]] = fptosi double [[TMP5]] to i8
-; CHECK-NEXT:    store i8 [[TMP6]], ptr [[A:%.*]], align 1
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[B]], i64 11
-; CHECK-NEXT:    [[TMP8:%.*]] = load float, ptr [[TMP7]], align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = fmul float [[TMP8]], [[T]]
-; CHECK-NEXT:    [[TMP10:%.*]] = fpext float [[TMP9]] to double
-; CHECK-NEXT:    [[TMP11:%.*]] = fadd double [[TMP10]], 5.000000e+00
-; CHECK-NEXT:    [[TMP12:%.*]] = fptosi double [[TMP11]] to i8
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 1
-; CHECK-NEXT:    store i8 [[TMP12]], ptr [[TMP13]], align 1
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 12
-; CHECK-NEXT:    [[TMP15:%.*]] = load float, ptr [[TMP14]], align 4
-; CHECK-NEXT:    [[TMP16:%.*]] = fmul float [[TMP15]], [[T]]
-; CHECK-NEXT:    [[TMP17:%.*]] = fpext float [[TMP16]] to double
-; CHECK-NEXT:    [[TMP18:%.*]] = fadd double [[TMP17]], 6.000000e+00
-; CHECK-NEXT:    [[TMP19:%.*]] = fptosi double [[TMP18]] to i8
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 2
-; CHECK-NEXT:    store i8 [[TMP19]], ptr [[TMP20]], align 1
-; CHECK-NEXT:    ret i32 undef
+; NON-POW2-LABEL: @foo(
+; NON-POW2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 10
+; NON-POW2-NEXT:    [[TMP2:%.*]] = load <3 x float>, ptr [[TMP1]], align 4
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <3 x float> poison, float [[T:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP5:%.*]] = fmul <3 x float> [[TMP2]], [[TMP4]]
+; NON-POW2-NEXT:    [[TMP6:%.*]] = fpext <3 x float> [[TMP5]] to <3 x double>
+; NON-POW2-NEXT:    [[TMP7:%.*]] = fadd <3 x double> [[TMP6]], <double 4.000000e+00, double 5.000000e+00, double 6.000000e+00>
+; NON-POW2-NEXT:    [[TMP8:%.*]] = fptosi <3 x double> [[TMP7]] to <3 x i8>
+; NON-POW2-NEXT:    store <3 x i8> [[TMP8]], ptr [[A:%.*]], align 1
+; NON-POW2-NEXT:    ret i32 undef
+;
+; POW2-ONLY-LABEL: @foo(
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 10
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = load float, ptr [[TMP1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = fmul float [[TMP2]], [[T:%.*]]
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = fpext float [[TMP3]] to double
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = fadd double [[TMP4]], 4.000000e+00
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = fptosi double [[TMP5]] to i8
+; POW2-ONLY-NEXT:    store i8 [[TMP6]], ptr [[A:%.*]], align 1
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[B]], i64 11
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = load float, ptr [[TMP7]], align 4
+; POW2-ONLY-NEXT:    [[TMP9:%.*]] = fmul float [[TMP8]], [[T]]
+; POW2-ONLY-NEXT:    [[TMP10:%.*]] = fpext float [[TMP9]] to double
+; POW2-ONLY-NEXT:    [[TMP11:%.*]] = fadd double [[TMP10]], 5.000000e+00
+; POW2-ONLY-NEXT:    [[TMP12:%.*]] = fptosi double [[TMP11]] to i8
+; POW2-ONLY-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 1
+; POW2-ONLY-NEXT:    store i8 [[TMP12]], ptr [[TMP13]], align 1
+; POW2-ONLY-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 12
+; POW2-ONLY-NEXT:    [[TMP15:%.*]] = load float, ptr [[TMP14]], align 4
+; POW2-ONLY-NEXT:    [[TMP16:%.*]] = fmul float [[TMP15]], [[T]]
+; POW2-ONLY-NEXT:    [[TMP17:%.*]] = fpext float [[TMP16]] to double
+; POW2-ONLY-NEXT:    [[TMP18:%.*]] = fadd double [[TMP17]], 6.000000e+00
+; POW2-ONLY-NEXT:    [[TMP19:%.*]] = fptosi double [[TMP18]] to i8
+; POW2-ONLY-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 2
+; POW2-ONLY-NEXT:    store i8 [[TMP19]], ptr [[TMP20]], align 1
+; POW2-ONLY-NEXT:    ret i32 undef
 ;
   %1 = getelementptr inbounds float, ptr %B, i64 10
   %2 = load float, ptr %1, align 4
@@ -91,13 +104,18 @@ define void @test_v4f32_v2f32_splat_store(<4 x float> %f, ptr %p){
 }
 
 define void @test_v4f32_v3f32_store(<4 x float> %f, ptr %p){
-; CHECK-LABEL: @test_v4f32_v3f32_store(
-; CHECK-NEXT:    [[X2:%.*]] = extractelement <4 x float> [[F:%.*]], i64 2
-; CHECK-NEXT:    [[P2:%.*]] = getelementptr inbounds float, ptr [[P:%.*]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[F]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT:    store <2 x float> [[TMP1]], ptr [[P]], align 4
-; CHECK-NEXT:    store float [[X2]], ptr [[P2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @test_v4f32_v3f32_store(
+; NON-POW2-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[F:%.*]], <4 x float> poison, <3 x i32> <i32 0, i32 1, i32 2>
+; NON-POW2-NEXT:    store <3 x float> [[TMP1]], ptr [[P:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @test_v4f32_v3f32_store(
+; POW2-ONLY-NEXT:    [[X2:%.*]] = extractelement <4 x float> [[F:%.*]], i64 2
+; POW2-ONLY-NEXT:    [[P2:%.*]] = getelementptr inbounds float, ptr [[P:%.*]], i64 2
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[F]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP1]], ptr [[P]], align 4
+; POW2-ONLY-NEXT:    store float [[X2]], ptr [[P2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
   %x0 = extractelement <4 x float> %f, i64 0
   %x1 = extractelement <4 x float> %f, i64 1
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vec3-base.ll b/llvm/test/Transforms/SLPVectorizer/X86/vec3-base.ll
index 6560fc6a14526..96d4b84e03691 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/vec3-base.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/vec3-base.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=slp-vectorizer -mtriple=x86_64-apple-macosx -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 define void @v3_load_i32_mul_by_constant_store(ptr %src, ptr %dst) {
 ; CHECK-LABEL: @v3_load_i32_mul_by_constant_store(
@@ -161,18 +162,26 @@ entry:
 }
 
 define void @v3_load_f32_fadd_fadd_by_constant_store(ptr %src, ptr %dst) {
-; CHECK-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
-; CHECK-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds float, ptr [[SRC]], i32 2
-; CHECK-NEXT:    [[L_SRC_2:%.*]] = load float, ptr [[GEP_SRC_2]], align 4
-; CHECK-NEXT:    [[FADD_2:%.*]] = fadd float [[L_SRC_2]], 1.000000e+01
-; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[GEP_SRC_0]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = fadd <2 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01>
-; CHECK-NEXT:    store <2 x float> [[TMP1]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    [[DST_2:%.*]] = getelementptr float, ptr [[DST]], i32 2
-; CHECK-NEXT:    store float [[FADD_2]], ptr [[DST_2]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load <3 x float>, ptr [[GEP_SRC_0]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = fadd <3 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01, float 1.000000e+01>
+; NON-POW2-NEXT:    store <3 x float> [[TMP1]], ptr [[DST:%.*]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @v3_load_f32_fadd_fadd_by_constant_store(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i32 0
+; POW2-ONLY-NEXT:    [[GEP_SRC_2:%.*]] = getelementptr inbounds float, ptr [[SRC]], i32 2
+; POW2-ONLY-NEXT:    [[L_SRC_2:%.*]] = load float, ptr [[GEP_SRC_2]], align 4
+; POW2-ONLY-NEXT:    [[FADD_2:%.*]] = fadd float [[L_SRC_2]], 1.000000e+01
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[GEP_SRC_0]], align 4
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = fadd <2 x float> [[TMP0]], <float 1.000000e+01, float 1.000000e+01>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP1]], ptr [[DST:%.*]], align 4
+; POW2-ONLY-NEXT:    [[DST_2:%.*]] = getelementptr float, ptr [[DST]], i32 2
+; POW2-ONLY-NEXT:    store float [[FADD_2]], ptr [[DST_2]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %gep.src.0 = getelementptr inbounds float, ptr %src, i32 0
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vec3-calls.ll b/llvm/test/Transforms/SLPVectorizer/X86/vec3-calls.ll
index 71b9315839ecf..243087c6d8d95 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/vec3-calls.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/vec3-calls.ll
@@ -1,16 +1,29 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=slp-vectorizer -mtriple=x86_64-apple-macosx -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 define void @vec3_vectorize_call(ptr %Colour, float %0) {
-; CHECK-LABEL: @vec3_vectorize_call(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[COLOUR:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP1]], <2 x float> zeroinitializer, <2 x float> zeroinitializer)
-; CHECK-NEXT:    store <2 x float> [[TMP2]], ptr [[COLOUR]], align 4
-; CHECK-NEXT:    [[ARRAYIDX99_I1:%.*]] = getelementptr float, ptr [[COLOUR]], i64 2
-; CHECK-NEXT:    [[TMP3:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0:%.*]], float 0.000000e+00, float 0.000000e+00)
-; CHECK-NEXT:    store float [[TMP3]], ptr [[ARRAYIDX99_I1]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @vec3_vectorize_call(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load float, ptr [[COLOUR:%.*]], align 4
+; NON-POW2-NEXT:    [[ARRAYIDX91_I:%.*]] = getelementptr float, ptr [[COLOUR]], i64 1
+; NON-POW2-NEXT:    [[TMP2:%.*]] = load float, ptr [[ARRAYIDX91_I]], align 4
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <3 x float> poison, float [[TMP0:%.*]], i32 2
+; NON-POW2-NEXT:    [[TMP4:%.*]] = insertelement <3 x float> [[TMP3]], float [[TMP1]], i32 0
+; NON-POW2-NEXT:    [[TMP5:%.*]] = insertelement <3 x float> [[TMP4]], float [[TMP2]], i32 1
+; NON-POW2-NEXT:    [[TMP6:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP5]], <3 x float> zeroinitializer, <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    store <3 x float> [[TMP6]], ptr [[COLOUR]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @vec3_vectorize_call(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[COLOUR:%.*]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP1]], <2 x float> zeroinitializer, <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP2]], ptr [[COLOUR]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX99_I1:%.*]] = getelementptr float, ptr [[COLOUR]], i64 2
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0:%.*]], float 0.000000e+00, float 0.000000e+00)
+; POW2-ONLY-NEXT:    store float [[TMP3]], ptr [[ARRAYIDX99_I1]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %1 = load float, ptr %Colour, align 4
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vec3-gather-some-loads.ll b/llvm/test/Transforms/SLPVectorizer/X86/vec3-gather-some-loads.ll
index 1411f9416f69d..e8adda0bdc703 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/vec3-gather-some-loads.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/vec3-gather-some-loads.ll
@@ -1,35 +1,55 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=slp-vectorizer -mtriple=x86_64-apple-macosx -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=POW2-ONLY %s
 
 target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
 
 define void @test_insert_loads(ptr %A, ptr noalias %B, float %0) #0 {
-; CHECK-LABEL: define void @test_insert_loads(
-; CHECK-SAME: ptr [[A:%.*]], ptr noalias [[B:%.*]], float [[TMP0:%.*]]) #[[ATTR0:[0-9]+]] {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[MULADD_0:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 1.000000e+00, float 1.000000e+00)
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> <float 3.000000e+00, float 2.000000e+00>, <2 x float> <float 3.000000e+00, float 2.000000e+00>)
-; CHECK-NEXT:    [[A_28:%.*]] = getelementptr i8, ptr [[A]], i64 28
-; CHECK-NEXT:    [[L_A_28:%.*]] = load float, ptr [[A_28]], align 4
-; CHECK-NEXT:    [[A_12:%.*]] = getelementptr i8, ptr [[A]], i64 12
-; CHECK-NEXT:    [[L_A_12:%.*]] = load float, ptr [[A_12]], align 4
-; CHECK-NEXT:    [[GEP_4:%.*]] = getelementptr i8, ptr [[B]], i64 4
-; CHECK-NEXT:    [[L_B_0:%.*]] = load float, ptr [[B]], align 4
-; CHECK-NEXT:    [[GEP_28:%.*]] = getelementptr i8, ptr [[B]], i64 28
-; CHECK-NEXT:    [[GEP_20:%.*]] = getelementptr i8, ptr [[B]], i64 20
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <4 x float> <float poison, float poison, float poison, float 4.000000e+00>, float [[L_A_12]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> [[TMP6]], float [[L_A_28]], i32 1
-; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 3>
-; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <4 x float> <float poison, float 0.000000e+00, float 0.000000e+00, float 4.000000e+00>, float [[L_B_0]], i32 0
-; CHECK-NEXT:    [[TMP10:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP5]], <4 x float> [[TMP8]], <4 x float> [[TMP9]])
-; CHECK-NEXT:    store <4 x float> [[TMP10]], ptr [[GEP_4]], align 4
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[GEP_20]], align 4
-; CHECK-NEXT:    store float [[MULADD_0]], ptr [[GEP_28]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @test_insert_loads(
+; NON-POW2-SAME: ptr [[A:%.*]], ptr noalias [[B:%.*]], float [[TMP0:%.*]]) #[[ATTR0:[0-9]+]] {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[A_28:%.*]] = getelementptr i8, ptr [[A]], i64 28
+; NON-POW2-NEXT:    [[L_A_28:%.*]] = load float, ptr [[A_28]], align 4
+; NON-POW2-NEXT:    [[A_12:%.*]] = getelementptr i8, ptr [[A]], i64 12
+; NON-POW2-NEXT:    [[L_A_12:%.*]] = load float, ptr [[A_12]], align 4
+; NON-POW2-NEXT:    [[GEP_4:%.*]] = getelementptr i8, ptr [[B]], i64 4
+; NON-POW2-NEXT:    [[L_B_0:%.*]] = load float, ptr [[B]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = insertelement <7 x float> poison, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP2:%.*]] = shufflevector <7 x float> [[TMP1]], <7 x float> poison, <7 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <7 x float> <float poison, float poison, float poison, float 4.000000e+00, float 3.000000e+00, float 2.000000e+00, float 1.000000e+00>, float [[L_A_12]], i32 0
+; NON-POW2-NEXT:    [[TMP4:%.*]] = insertelement <7 x float> [[TMP3]], float [[L_A_28]], i32 1
+; NON-POW2-NEXT:    [[TMP5:%.*]] = shufflevector <7 x float> [[TMP4]], <7 x float> poison, <7 x i32> <i32 0, i32 1, i32 1, i32 3, i32 4, i32 5, i32 6>
+; NON-POW2-NEXT:    [[TMP6:%.*]] = insertelement <7 x float> <float poison, float 0.000000e+00, float 0.000000e+00, float 4.000000e+00, float 3.000000e+00, float 2.000000e+00, float 1.000000e+00>, float [[L_B_0]], i32 0
+; NON-POW2-NEXT:    [[TMP7:%.*]] = call <7 x float> @llvm.fmuladd.v7f32(<7 x float> [[TMP2]], <7 x float> [[TMP5]], <7 x float> [[TMP6]])
+; NON-POW2-NEXT:    store <7 x float> [[TMP7]], ptr [[GEP_4]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @test_insert_loads(
+; POW2-ONLY-SAME: ptr [[A:%.*]], ptr noalias [[B:%.*]], float [[TMP0:%.*]]) #[[ATTR0:[0-9]+]] {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[MULADD_0:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 1.000000e+00, float 1.000000e+00)
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> <float 3.000000e+00, float 2.000000e+00>, <2 x float> <float 3.000000e+00, float 2.000000e+00>)
+; POW2-ONLY-NEXT:    [[A_28:%.*]] = getelementptr i8, ptr [[A]], i64 28
+; POW2-ONLY-NEXT:    [[L_A_28:%.*]] = load float, ptr [[A_28]], align 4
+; POW2-ONLY-NEXT:    [[A_12:%.*]] = getelementptr i8, ptr [[A]], i64 12
+; POW2-ONLY-NEXT:    [[L_A_12:%.*]] = load float, ptr [[A_12]], align 4
+; POW2-ONLY-NEXT:    [[GEP_4:%.*]] = getelementptr i8, ptr [[B]], i64 4
+; POW2-ONLY-NEXT:    [[L_B_0:%.*]] = load float, ptr [[B]], align 4
+; POW2-ONLY-NEXT:    [[GEP_28:%.*]] = getelementptr i8, ptr [[B]], i64 28
+; POW2-ONLY-NEXT:    [[GEP_20:%.*]] = getelementptr i8, ptr [[B]], i64 20
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> poison, <4 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = insertelement <4 x float> <float poison, float poison, float poison, float 4.000000e+00>, float [[L_A_12]], i32 0
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = insertelement <4 x float> [[TMP6]], float [[L_A_28]], i32 1
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 3>
+; POW2-ONLY-NEXT:    [[TMP9:%.*]] = insertelement <4 x float> <float poison, float 0.000000e+00, float 0.000000e+00, float 4.000000e+00>, float [[L_B_0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP10:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP5]], <4 x float> [[TMP8]], <4 x float> [[TMP9]])
+; POW2-ONLY-NEXT:    store <4 x float> [[TMP10]], ptr [[GEP_4]], align 4
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP3]], ptr [[GEP_20]], align 4
+; POW2-ONLY-NEXT:    store float [[MULADD_0]], ptr [[GEP_28]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %muladd.0 = tail call float @llvm.fmuladd.f32(float %0, float 1.000000e+00, float 1.000000e+00)
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vec3-reorder-reshuffle.ll b/llvm/test/Transforms/SLPVectorizer/X86/vec3-reorder-reshuffle.ll
index 1faeea716b176..1399b4c35c781 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/vec3-reorder-reshuffle.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/vec3-reorder-reshuffle.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=slp-vectorizer -mtriple=x86_64-apple-macosx -S %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -mtriple=x86_64-apple-macosx -S %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 %struct.zot = type { i32, i32, i32 }
 
@@ -138,21 +139,34 @@ if.end668:                                        ; preds = %if.then665, %entry
 }
 
 define void @gather_2(ptr %mat1, float %0, float %1) {
-; CHECK-LABEL: define void @gather_2(
-; CHECK-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x float> <float 0.000000e+00, float poison>, float [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP6:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP1]], float 0.000000e+00)
-; CHECK-NEXT:    [[TMP7:%.*]] = fmul float [[TMP6]], 0.000000e+00
-; CHECK-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
-; CHECK-NEXT:    [[ARRAYIDX5_I_I_I280:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1, i64 2
-; CHECK-NEXT:    [[TMP8:%.*]] = fmul <2 x float> [[TMP5]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
-; CHECK-NEXT:    store float [[TMP7]], ptr [[ARRAYIDX5_I_I_I280]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @gather_2(
+; NON-POW2-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP4:%.*]] = shufflevector <3 x float> [[TMP2]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP5:%.*]] = insertelement <3 x float> <float 0.000000e+00, float poison, float poison>, float [[TMP1]], i32 1
+; NON-POW2-NEXT:    [[TMP6:%.*]] = shufflevector <3 x float> [[TMP5]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
+; NON-POW2-NEXT:    [[TMP7:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP4]], <3 x float> [[TMP6]], <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
+; NON-POW2-NEXT:    [[TMP8:%.*]] = fmul <3 x float> [[TMP7]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @gather_2(
+; POW2-ONLY-SAME: ptr [[MAT1:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = insertelement <2 x float> <float 0.000000e+00, float poison>, float [[TMP1]], i32 1
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP1]], float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = fmul float [[TMP6]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[ARRAYIDX163:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1
+; POW2-ONLY-NEXT:    [[ARRAYIDX5_I_I_I280:%.*]] = getelementptr [4 x [4 x float]], ptr [[MAT1]], i64 0, i64 1, i64 2
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = fmul <2 x float> [[TMP5]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP8]], ptr [[ARRAYIDX163]], align 4
+; POW2-ONLY-NEXT:    store float [[TMP7]], ptr [[ARRAYIDX5_I_I_I280]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %2 = call float @llvm.fmuladd.f32(float %0, float 0.000000e+00, float 0.000000e+00)
@@ -171,31 +185,47 @@ entry:
 }
 
 define i32 @reorder_indices_1(float %0) {
-; CHECK-LABEL: define i32 @reorder_indices_1(
-; CHECK-SAME: float [[TMP0:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
-; CHECK-NEXT:    [[ARRAYIDX2_I265:%.*]] = getelementptr float, ptr [[NOR1]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2_I265]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[NOR1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = fneg float [[TMP3]]
-; CHECK-NEXT:    [[NEG11_I:%.*]] = fmul float [[TMP4]], [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = call float @llvm.fmuladd.f32(float [[TMP1]], float 0.000000e+00, float [[NEG11_I]])
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 poison>
-; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP8:%.*]] = fneg <2 x float> [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]]
-; CHECK-NEXT:    [[TMP12:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> zeroinitializer, <2 x float> [[TMP11]])
-; CHECK-NEXT:    [[TMP13:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP10]], <2 x float> [[TMP12]], <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP14:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP5]], float 0.000000e+00)
-; CHECK-NEXT:    [[TMP15:%.*]] = fmul <2 x float> [[TMP13]], zeroinitializer
-; CHECK-NEXT:    [[MUL6_I_I_I:%.*]] = fmul float [[TMP14]], 0.000000e+00
-; CHECK-NEXT:    store <2 x float> [[TMP15]], ptr [[NOR1]], align 4
-; CHECK-NEXT:    store float [[MUL6_I_I_I]], ptr [[ARRAYIDX2_I265]], align 4
-; CHECK-NEXT:    ret i32 0
+; NON-POW2-LABEL: define i32 @reorder_indices_1(
+; NON-POW2-SAME: float [[TMP0:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x float>, ptr [[NOR1]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = shufflevector <3 x float> [[TMP1]], <3 x float> poison, <3 x i32> <i32 1, i32 2, i32 0>
+; NON-POW2-NEXT:    [[TMP3:%.*]] = fneg <3 x float> [[TMP2]]
+; NON-POW2-NEXT:    [[TMP4:%.*]] = insertelement <3 x float> poison, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP5:%.*]] = shufflevector <3 x float> [[TMP4]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP6:%.*]] = fmul <3 x float> [[TMP3]], [[TMP5]]
+; NON-POW2-NEXT:    [[TMP7:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP1]], <3 x float> zeroinitializer, <3 x float> [[TMP6]])
+; NON-POW2-NEXT:    [[TMP8:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP5]], <3 x float> [[TMP7]], <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[TMP9:%.*]] = fmul <3 x float> [[TMP8]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP9]], ptr [[NOR1]], align 4
+; NON-POW2-NEXT:    ret i32 0
+;
+; POW2-ONLY-LABEL: define i32 @reorder_indices_1(
+; POW2-ONLY-SAME: float [[TMP0:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[NOR1:%.*]] = alloca [0 x [3 x float]], i32 0, align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX2_I265:%.*]] = getelementptr float, ptr [[NOR1]], i64 2
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2_I265]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[NOR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = fneg float [[TMP3]]
+; POW2-ONLY-NEXT:    [[NEG11_I:%.*]] = fmul float [[TMP4]], [[TMP0]]
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = call float @llvm.fmuladd.f32(float [[TMP1]], float 0.000000e+00, float [[NEG11_I]])
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 poison>
+; POW2-ONLY-NEXT:    [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
+; POW2-ONLY-NEXT:    [[TMP8:%.*]] = fneg <2 x float> [[TMP7]]
+; POW2-ONLY-NEXT:    [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP10:%.*]] = shufflevector <2 x float> [[TMP9]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP11:%.*]] = fmul <2 x float> [[TMP8]], [[TMP10]]
+; POW2-ONLY-NEXT:    [[TMP12:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> zeroinitializer, <2 x float> [[TMP11]])
+; POW2-ONLY-NEXT:    [[TMP13:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP10]], <2 x float> [[TMP12]], <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP14:%.*]] = call float @llvm.fmuladd.f32(float [[TMP0]], float [[TMP5]], float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[TMP15:%.*]] = fmul <2 x float> [[TMP13]], zeroinitializer
+; POW2-ONLY-NEXT:    [[MUL6_I_I_I:%.*]] = fmul float [[TMP14]], 0.000000e+00
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP15]], ptr [[NOR1]], align 4
+; POW2-ONLY-NEXT:    store float [[MUL6_I_I_I]], ptr [[ARRAYIDX2_I265]], align 4
+; POW2-ONLY-NEXT:    ret i32 0
 ;
 entry:
   %nor1 = alloca [0 x [3 x float]], i32 0, align 4
@@ -226,19 +256,28 @@ entry:
 }
 
 define void @reorder_indices_2(ptr %spoint) {
-; CHECK-LABEL: define void @reorder_indices_2(
-; CHECK-SAME: ptr [[SPOINT:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> zeroinitializer, i64 0
-; CHECK-NEXT:    [[TMP1:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 0.000000e+00, float 0.000000e+00)
-; CHECK-NEXT:    [[MUL4_I461:%.*]] = fmul float [[TMP1]], 0.000000e+00
-; CHECK-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer, <2 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP3:%.*]] = fmul <2 x float> [[TMP2]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[DSCO]], align 4
-; CHECK-NEXT:    [[ARRAYIDX5_I476:%.*]] = getelementptr float, ptr [[SPOINT]], i64 2
-; CHECK-NEXT:    store float [[MUL4_I461]], ptr [[ARRAYIDX5_I476]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @reorder_indices_2(
+; NON-POW2-SAME: ptr [[SPOINT:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
+; NON-POW2-NEXT:    [[TMP0:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> zeroinitializer, <3 x float> zeroinitializer, <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    [[TMP1:%.*]] = fmul <3 x float> [[TMP0]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP1]], ptr [[DSCO]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @reorder_indices_2(
+; POW2-ONLY-SAME: ptr [[SPOINT:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = extractelement <3 x float> zeroinitializer, i64 0
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP0]], float 0.000000e+00, float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[MUL4_I461:%.*]] = fmul float [[TMP1]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[DSCO:%.*]] = getelementptr float, ptr [[SPOINT]], i64 0
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> zeroinitializer, <2 x float> zeroinitializer, <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = fmul <2 x float> [[TMP2]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP3]], ptr [[DSCO]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX5_I476:%.*]] = getelementptr float, ptr [[SPOINT]], i64 2
+; POW2-ONLY-NEXT:    store float [[MUL4_I461]], ptr [[ARRAYIDX5_I476]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %0 = extractelement <3 x float> zeroinitializer, i64 1
@@ -290,19 +329,30 @@ entry:
 }
 
 define void @reuse_shuffle_indidces_1(ptr %col, float %0, float %1) {
-; CHECK-LABEL: define void @reuse_shuffle_indidces_1(
-; CHECK-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP0]], i32 1
-; CHECK-NEXT:    [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd <2 x float> [[TMP4]], zeroinitializer
-; CHECK-NEXT:    store <2 x float> [[TMP5]], ptr [[COL]], align 4
-; CHECK-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr float, ptr [[COL]], i64 2
-; CHECK-NEXT:    [[MUL38:%.*]] = fmul float [[TMP0]], 0.000000e+00
-; CHECK-NEXT:    [[TMP6:%.*]] = fadd float [[MUL38]], 0.000000e+00
-; CHECK-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX33]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @reuse_shuffle_indidces_1(
+; NON-POW2-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[TMP2:%.*]] = insertelement <3 x float> poison, float [[TMP1]], i32 0
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <3 x float> [[TMP2]], float [[TMP0]], i32 1
+; NON-POW2-NEXT:    [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> <i32 0, i32 1, i32 1>
+; NON-POW2-NEXT:    [[TMP5:%.*]] = fmul <3 x float> [[TMP4]], zeroinitializer
+; NON-POW2-NEXT:    [[TMP6:%.*]] = fadd <3 x float> [[TMP5]], zeroinitializer
+; NON-POW2-NEXT:    store <3 x float> [[TMP6]], ptr [[COL]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @reuse_shuffle_indidces_1(
+; POW2-ONLY-SAME: ptr [[COL:%.*]], float [[TMP0:%.*]], float [[TMP1:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP0]], i32 1
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = fmul <2 x float> [[TMP3]], zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = fadd <2 x float> [[TMP4]], zeroinitializer
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP5]], ptr [[COL]], align 4
+; POW2-ONLY-NEXT:    [[ARRAYIDX33:%.*]] = getelementptr float, ptr [[COL]], i64 2
+; POW2-ONLY-NEXT:    [[MUL38:%.*]] = fmul float [[TMP0]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = fadd float [[MUL38]], 0.000000e+00
+; POW2-ONLY-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX33]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %mul24 = fmul float %1, 0.000000e+00
@@ -350,20 +400,32 @@ entry:
 }
 
 define void @reuse_shuffle_indices_cost_crash_2(ptr %bezt, float %0) {
-; CHECK-LABEL: define void @reuse_shuffle_indices_cost_crash_2(
-; CHECK-SAME: ptr [[BEZT:%.*]], float [[TMP0:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[FNEG:%.*]] = fmul float [[TMP0]], 0.000000e+00
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> poison, float [[FNEG]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
-; CHECK-NEXT:    store <2 x float> [[TMP5]], ptr [[BEZT]], align 4
-; CHECK-NEXT:    [[TMP6:%.*]] = tail call float @llvm.fmuladd.f32(float [[FNEG]], float 0.000000e+00, float 0.000000e+00)
-; CHECK-NEXT:    [[ARRAYIDX8_I831:%.*]] = getelementptr float, ptr [[BEZT]], i64 2
-; CHECK-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX8_I831]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @reuse_shuffle_indices_cost_crash_2(
+; NON-POW2-SAME: ptr [[BEZT:%.*]], float [[TMP0:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[FNEG:%.*]] = fmul float [[TMP0]], 0.000000e+00
+; NON-POW2-NEXT:    [[TMP1:%.*]] = insertelement <3 x float> poison, float [[FNEG]], i32 0
+; NON-POW2-NEXT:    [[TMP2:%.*]] = shufflevector <3 x float> [[TMP1]], <3 x float> poison, <3 x i32> zeroinitializer
+; NON-POW2-NEXT:    [[TMP3:%.*]] = insertelement <3 x float> <float poison, float poison, float 0.000000e+00>, float [[TMP0]], i32 0
+; NON-POW2-NEXT:    [[TMP4:%.*]] = shufflevector <3 x float> [[TMP3]], <3 x float> poison, <3 x i32> <i32 0, i32 0, i32 2>
+; NON-POW2-NEXT:    [[TMP5:%.*]] = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> [[TMP2]], <3 x float> [[TMP4]], <3 x float> zeroinitializer)
+; NON-POW2-NEXT:    store <3 x float> [[TMP5]], ptr [[BEZT]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @reuse_shuffle_indices_cost_crash_2(
+; POW2-ONLY-SAME: ptr [[BEZT:%.*]], float [[TMP0:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[FNEG:%.*]] = fmul float [[TMP0]], 0.000000e+00
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = insertelement <2 x float> poison, float [[TMP0]], i32 0
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> poison, float [[FNEG]], i32 0
+; POW2-ONLY-NEXT:    [[TMP4:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> poison, <2 x i32> zeroinitializer
+; POW2-ONLY-NEXT:    [[TMP5:%.*]] = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> [[TMP2]], <2 x float> [[TMP4]], <2 x float> zeroinitializer)
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP5]], ptr [[BEZT]], align 4
+; POW2-ONLY-NEXT:    [[TMP6:%.*]] = tail call float @llvm.fmuladd.f32(float [[FNEG]], float 0.000000e+00, float 0.000000e+00)
+; POW2-ONLY-NEXT:    [[ARRAYIDX8_I831:%.*]] = getelementptr float, ptr [[BEZT]], i64 2
+; POW2-ONLY-NEXT:    store float [[TMP6]], ptr [[ARRAYIDX8_I831]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %fneg = fmul float %0, 0.000000e+00
@@ -487,15 +549,21 @@ entry:
 }
 
 define void @vec3_extract(<3 x i16> %pixel.sroa.0.4.vec.insert606, ptr %call3.i536) {
-; CHECK-LABEL: define void @vec3_extract(
-; CHECK-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[PIXEL_SROA_0_4_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 2
-; CHECK-NEXT:    [[RED668:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 2
-; CHECK-NEXT:    store i16 [[PIXEL_SROA_0_4_VEC_EXTRACT]], ptr [[RED668]], align 2
-; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
-; CHECK-NEXT:    store <2 x i16> [[TMP0]], ptr [[CALL3_I536]], align 2
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: define void @vec3_extract(
+; NON-POW2-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    store <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], ptr [[CALL3_I536]], align 2
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: define void @vec3_extract(
+; POW2-ONLY-SAME: <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606:%.*]], ptr [[CALL3_I536:%.*]]) {
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[PIXEL_SROA_0_4_VEC_EXTRACT:%.*]] = extractelement <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], i64 2
+; POW2-ONLY-NEXT:    [[RED668:%.*]] = getelementptr i16, ptr [[CALL3_I536]], i64 2
+; POW2-ONLY-NEXT:    store i16 [[PIXEL_SROA_0_4_VEC_EXTRACT]], ptr [[RED668]], align 2
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = shufflevector <3 x i16> [[PIXEL_SROA_0_4_VEC_INSERT606]], <3 x i16> poison, <2 x i32> <i32 0, i32 1>
+; POW2-ONLY-NEXT:    store <2 x i16> [[TMP0]], ptr [[CALL3_I536]], align 2
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %pixel.sroa.0.4.vec.extract = extractelement <3 x i16> %pixel.sroa.0.4.vec.insert606, i64 2
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll b/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll
index e8395fe69ea6b..e30cb76d53d92 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll
@@ -1,12 +1,13 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux-gnu -mcpu=bdver2 < %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2 -S -mtriple=x86_64-unknown-linux-gnu -mcpu=bdver2 < %s | FileCheck --check-prefixes=CHECK,NON-POW2 %s
+; RUN: opt -passes=slp-vectorizer -slp-vectorize-non-power-of-2=false -S -mtriple=x86_64-unknown-linux-gnu -mcpu=bdver2 < %s | FileCheck --check-prefixes=CHECK,POW2-ONLY %s
 
 define void @add0(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @add0(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[TMP1]], <i32 1, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = add nsw <4 x i32> [[TMP0]], <i32 1, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    store <4 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -32,21 +33,32 @@ entry:
 }
 
 define void @add1(ptr noalias %dst, ptr noalias %src) {
-; CHECK-LABEL: @add1(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
-; CHECK-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i32> [[TMP2]], <i32 1, i32 2>
-; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP5]], 3
-; CHECK-NEXT:    store i32 [[ADD9]], ptr [[INCDEC_PTR7]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @add1(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
+; NON-POW2-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
+; NON-POW2-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x i32>, ptr [[INCDEC_PTR]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = add nsw <3 x i32> [[TMP1]], <i32 1, i32 2, i32 3>
+; NON-POW2-NEXT:    store <3 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @add1(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
+; POW2-ONLY-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
+; POW2-ONLY-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = add nsw <2 x i32> [[TMP1]], <i32 1, i32 2>
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = load i32, ptr [[INCDEC_PTR5]], align 4
+; POW2-ONLY-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP3]], 3
+; POW2-ONLY-NEXT:    store i32 [[ADD9]], ptr [[INCDEC_PTR7]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, ptr %src, i64 1
@@ -81,9 +93,9 @@ define void @sub0(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 2
 ; CHECK-NEXT:    store i32 [[TMP1]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i32> [[TMP3]], <i32 -2, i32 -3>
-; CHECK-NEXT:    store <2 x i32> [[TMP4]], ptr [[INCDEC_PTR3]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i32> [[TMP2]], <i32 -2, i32 -3>
+; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[INCDEC_PTR3]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -110,9 +122,9 @@ entry:
 define void @sub1(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub1(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[TMP1]], <i32 4, i32 -1, i32 -2, i32 -3>
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = add nsw <4 x i32> [[TMP0]], <i32 4, i32 -1, i32 -2, i32 -3>
+; CHECK-NEXT:    store <4 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -140,9 +152,9 @@ entry:
 define void @sub2(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub2(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[TMP1]], <i32 -1, i32 -1, i32 -2, i32 -3>
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = add nsw <4 x i32> [[TMP0]], <i32 -1, i32 -1, i32 -2, i32 -3>
+; CHECK-NEXT:    store <4 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -218,11 +230,11 @@ define void @addsub1(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[TMP6]], ptr [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP7]], -3
+; CHECK-NEXT:    store i32 [[TMP4]], ptr [[INCDEC_PTR3]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP5]], -3
 ; CHECK-NEXT:    store i32 [[SUB8]], ptr [[INCDEC_PTR6]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -252,15 +264,15 @@ define void @mul(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 2
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP1]], <i32 257, i32 -3>
-; CHECK-NEXT:    store <2 x i32> [[TMP2]], ptr [[DST]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr [[SRC]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nsw <2 x i32> [[TMP0]], <i32 257, i32 -3>
+; CHECK-NEXT:    store <2 x i32> [[TMP1]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[TMP4]], ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP5]], -9
+; CHECK-NEXT:    store i32 [[TMP2]], ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP3]], -9
 ; CHECK-NEXT:    store i32 [[MUL9]], ptr [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -286,21 +298,32 @@ entry:
 }
 
 define void @shl0(ptr noalias %dst, ptr noalias %src) {
-; CHECK-LABEL: @shl0(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
-; CHECK-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = shl <2 x i32> [[TMP2]], <i32 1, i32 2>
-; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SHL8:%.*]] = shl i32 [[TMP5]], 3
-; CHECK-NEXT:    store i32 [[SHL8]], ptr [[INCDEC_PTR6]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @shl0(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
+; NON-POW2-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
+; NON-POW2-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x i32>, ptr [[INCDEC_PTR]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = shl <3 x i32> [[TMP1]], <i32 1, i32 2, i32 3>
+; NON-POW2-NEXT:    store <3 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @shl0(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 1
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 1
+; POW2-ONLY-NEXT:    store i32 [[TMP0]], ptr [[DST]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 3
+; POW2-ONLY-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 3
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[INCDEC_PTR]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = shl <2 x i32> [[TMP1]], <i32 1, i32 2>
+; POW2-ONLY-NEXT:    store <2 x i32> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = load i32, ptr [[INCDEC_PTR4]], align 4
+; POW2-ONLY-NEXT:    [[SHL8:%.*]] = shl i32 [[TMP3]], 3
+; POW2-ONLY-NEXT:    store i32 [[SHL8]], ptr [[INCDEC_PTR6]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, ptr %src, i64 1
@@ -326,9 +349,9 @@ entry:
 define void @shl1(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @shl1(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = shl <4 x i32> [[TMP1]], <i32 7, i32 1, i32 2, i32 3>
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <4 x i32> [[TMP0]], <i32 7, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    store <4 x i32> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -356,9 +379,9 @@ entry:
 define void @add0f(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @add0f(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd fast <4 x float> [[TMP1]], <float 1.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd fast <4 x float> [[TMP0]], <float 1.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -384,21 +407,32 @@ entry:
 }
 
 define void @add1f(ptr noalias %dst, ptr noalias %src) {
-; CHECK-LABEL: @add1f(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
-; CHECK-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = fadd fast <2 x float> [[TMP2]], <float 1.000000e+00, float 2.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd fast float [[TMP5]], 3.000000e+00
-; CHECK-NEXT:    store float [[ADD9]], ptr [[INCDEC_PTR7]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @add1f(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
+; NON-POW2-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
+; NON-POW2-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x float>, ptr [[INCDEC_PTR]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = fadd fast <3 x float> [[TMP1]], <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; NON-POW2-NEXT:    store <3 x float> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @add1f(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
+; POW2-ONLY-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
+; POW2-ONLY-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[INCDEC_PTR]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = fadd fast <2 x float> [[TMP1]], <float 1.000000e+00, float 2.000000e+00>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
+; POW2-ONLY-NEXT:    [[ADD9:%.*]] = fadd fast float [[TMP3]], 3.000000e+00
+; POW2-ONLY-NEXT:    store float [[ADD9]], ptr [[INCDEC_PTR7]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, ptr %src, i64 1
@@ -433,9 +467,9 @@ define void @sub0f(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 2
 ; CHECK-NEXT:    store float [[TMP1]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, ptr [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP4]], ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd fast <2 x float> [[TMP2]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[INCDEC_PTR4]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -462,9 +496,9 @@ entry:
 define void @sub1f(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub1f(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd fast <4 x float> [[TMP1]], <float 4.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd fast <4 x float> [[TMP0]], <float 4.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -492,9 +526,9 @@ entry:
 define void @sub2f(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub2f(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd fast <4 x float> [[TMP1]], <float -1.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd fast <4 x float> [[TMP0]], <float -1.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -570,11 +604,11 @@ define void @addsub1f(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP1]], <2 x float> [[TMP2]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP6:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP6]], ptr [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP7:%.*]] = load float, ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = fsub fast float [[TMP7]], -3.000000e+00
+; CHECK-NEXT:    store float [[TMP4]], ptr [[INCDEC_PTR3]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load float, ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[SUB8:%.*]] = fsub fast float [[TMP5]], -3.000000e+00
 ; CHECK-NEXT:    store float [[SUB8]], ptr [[INCDEC_PTR6]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -604,15 +638,15 @@ define void @mulf(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 2
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <2 x float> [[TMP1]], <float 2.570000e+02, float -3.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP2]], ptr [[DST]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[SRC]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul fast <2 x float> [[TMP0]], <float 2.570000e+02, float -3.000000e+00>
+; CHECK-NEXT:    store <2 x float> [[TMP1]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP4:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP4]], ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP5]], -9.000000e+00
+; CHECK-NEXT:    store float [[TMP2]], ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP3]], -9.000000e+00
 ; CHECK-NEXT:    store float [[SUB9]], ptr [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
@@ -640,9 +674,9 @@ entry:
 define void @add0fn(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @add0fn(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd <4 x float> [[TMP1]], <float 1.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x float> [[TMP0]], <float 1.000000e+00, float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -668,21 +702,32 @@ entry:
 }
 
 define void @add1fn(ptr noalias %dst, ptr noalias %src) {
-; CHECK-LABEL: @add1fn(
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
-; CHECK-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x float> [[TMP2]], <float 1.000000e+00, float 2.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd float [[TMP5]], 3.000000e+00
-; CHECK-NEXT:    store float [[ADD9]], ptr [[INCDEC_PTR7]], align 4
-; CHECK-NEXT:    ret void
+; NON-POW2-LABEL: @add1fn(
+; NON-POW2-NEXT:  entry:
+; NON-POW2-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
+; NON-POW2-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
+; NON-POW2-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
+; NON-POW2-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
+; NON-POW2-NEXT:    [[TMP1:%.*]] = load <3 x float>, ptr [[INCDEC_PTR]], align 4
+; NON-POW2-NEXT:    [[TMP2:%.*]] = fadd <3 x float> [[TMP1]], <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>
+; NON-POW2-NEXT:    store <3 x float> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; NON-POW2-NEXT:    ret void
+;
+; POW2-ONLY-LABEL: @add1fn(
+; POW2-ONLY-NEXT:  entry:
+; POW2-ONLY-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 1
+; POW2-ONLY-NEXT:    [[TMP0:%.*]] = load float, ptr [[SRC]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 1
+; POW2-ONLY-NEXT:    store float [[TMP0]], ptr [[DST]], align 4
+; POW2-ONLY-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
+; POW2-ONLY-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
+; POW2-ONLY-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[INCDEC_PTR]], align 4
+; POW2-ONLY-NEXT:    [[TMP2:%.*]] = fadd <2 x float> [[TMP1]], <float 1.000000e+00, float 2.000000e+00>
+; POW2-ONLY-NEXT:    store <2 x float> [[TMP2]], ptr [[INCDEC_PTR1]], align 4
+; POW2-ONLY-NEXT:    [[TMP3:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
+; POW2-ONLY-NEXT:    [[ADD9:%.*]] = fadd float [[TMP3]], 3.000000e+00
+; POW2-ONLY-NEXT:    store float [[ADD9]], ptr [[INCDEC_PTR7]], align 4
+; POW2-ONLY-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, ptr %src, i64 1
@@ -717,9 +762,9 @@ define void @sub0fn(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 2
 ; CHECK-NEXT:    store float [[TMP1]], ptr [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, ptr [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = fadd <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP4]], ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x float> [[TMP2]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[INCDEC_PTR4]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -746,9 +791,9 @@ entry:
 define void @sub1fn(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub1fn(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd <4 x float> [[TMP1]], <float 4.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x float> [[TMP0]], <float 4.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -776,9 +821,9 @@ entry:
 define void @sub2fn(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-LABEL: @sub2fn(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fadd <4 x float> [[TMP1]], <float -1.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[DST:%.*]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, ptr [[SRC:%.*]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fadd <4 x float> [[TMP0]], <float -1.000000e+00, float -1.000000e+00, float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[DST:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
@@ -808,15 +853,15 @@ define void @mulfn(ptr noalias %dst, ptr noalias %src) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 2
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x float> [[TMP1]], <float 2.570000e+02, float -3.000000e+00>
-; CHECK-NEXT:    store <2 x float> [[TMP2]], ptr [[DST]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = load <2 x float>, ptr [[SRC]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <2 x float> [[TMP0]], <float 2.570000e+02, float -3.000000e+00>
+; CHECK-NEXT:    store <2 x float> [[TMP1]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, ptr [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP4:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, ptr [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP4]], ptr [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP5]], -9.000000e+00
+; CHECK-NEXT:    store float [[TMP2]], ptr [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = load float, ptr [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP3]], -9.000000e+00
 ; CHECK-NEXT:    store float [[SUB9]], ptr [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;

From 11f22f1a963ab3c8949cb723a63c07d7a409c8a8 Mon Sep 17 00:00:00 2001
From: Eric <eric@efcs.ca>
Date: Sat, 13 Apr 2024 03:16:11 -0400
Subject: [PATCH 330/886] [tzdb] Replace shared_mutex with mutex. (#87929)

The overhead of taking a std::mutex is much lower than taking a reader
lock on a shared mutex, even under heavy contention.

The benefit of shared_mutex only occurs as the amount of
time spent in the critical sections grows large enough.

In our case all we do is read a pointer and return the lock.
As a result, using a shared lock can be ~50%-100% slower

Here are the results for the provided benchmark on my machine:

```
2024-04-07T12:48:51-04:00
Running ./libcxx/benchmarks/shared_mutex_vs_mutex.libcxx.out
Run on (12 X 400 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x6)
  L1 Instruction 32 KiB (x6)
  L2 Unified 1024 KiB (x6)
  L3 Unified 32768 KiB (x1)
Load Average: 2.70, 2.70, 1.63
---------------------------------------------------------------------
Benchmark                           Time             CPU   Iterations
---------------------------------------------------------------------
BM_shared_mutex/threads:1        13.9 ns         13.9 ns     50533700
BM_shared_mutex/threads:2        34.5 ns         68.9 ns      9957784
BM_shared_mutex/threads:4        38.4 ns          137 ns      4987772
BM_shared_mutex/threads:8        51.1 ns          358 ns      1974160
BM_shared_mutex/threads:32       57.1 ns          682 ns      1043648
BM_mutex/threads:1               5.54 ns         5.53 ns    125867422
BM_mutex/threads:2               15.5 ns         30.9 ns     21830116
BM_mutex/threads:4               15.4 ns         57.2 ns     12136920
BM_mutex/threads:8               19.3 ns          140 ns      4997080
BM_mutex/threads:32              20.8 ns          252 ns      2859808
```
---
 libcxx/benchmarks/CMakeLists.txt              |  1 +
 .../shared_mutex_vs_mutex.bench.cpp           | 41 +++++++++++++++++++
 libcxx/src/include/tzdb/tzdb_list_private.h   | 11 +++--
 3 files changed, 49 insertions(+), 4 deletions(-)
 create mode 100644 libcxx/benchmarks/shared_mutex_vs_mutex.bench.cpp

diff --git a/libcxx/benchmarks/CMakeLists.txt b/libcxx/benchmarks/CMakeLists.txt
index 928238c1ac69b..527a2acf2d3b3 100644
--- a/libcxx/benchmarks/CMakeLists.txt
+++ b/libcxx/benchmarks/CMakeLists.txt
@@ -221,6 +221,7 @@ set(BENCHMARK_TESTS
     map.bench.cpp
     monotonic_buffer.bench.cpp
     ordered_set.bench.cpp
+    shared_mutex_vs_mutex.bench.cpp
     stop_token.bench.cpp
     std_format_spec_string_unicode.bench.cpp
     string.bench.cpp
diff --git a/libcxx/benchmarks/shared_mutex_vs_mutex.bench.cpp b/libcxx/benchmarks/shared_mutex_vs_mutex.bench.cpp
new file mode 100644
index 0000000000000..19d13b799cc13
--- /dev/null
+++ b/libcxx/benchmarks/shared_mutex_vs_mutex.bench.cpp
@@ -0,0 +1,41 @@
+//===----------------------------------------------------------------------===//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// This benchmark compares the performance of std::mutex and std::shared_mutex in contended scenarios.
+// it's meant to establish a baseline overhead for std::shared_mutex and std::mutex, and to help inform decisions about
+// which mutex to use when selecting a mutex type for a given use case.
+
+#include <atomic>
+#include <mutex>
+#include <numeric>
+#include <shared_mutex>
+#include <thread>
+
+#include "benchmark/benchmark.h"
+
+int global_value = 42;
+std::mutex m;
+std::shared_mutex sm;
+
+static void BM_shared_mutex(benchmark::State& state) {
+  for (auto _ : state) {
+    std::shared_lock<std::shared_mutex> lock(sm);
+    benchmark::DoNotOptimize(global_value);
+  }
+}
+
+static void BM_mutex(benchmark::State& state) {
+  for (auto _ : state) {
+    std::lock_guard<std::mutex> lock(m);
+    benchmark::DoNotOptimize(global_value);
+  }
+}
+
+BENCHMARK(BM_shared_mutex)->Threads(1)->Threads(2)->Threads(4)->Threads(8)->Threads(32);
+BENCHMARK(BM_mutex)->Threads(1)->Threads(2)->Threads(4)->Threads(8)->Threads(32);
+
+BENCHMARK_MAIN();
diff --git a/libcxx/src/include/tzdb/tzdb_list_private.h b/libcxx/src/include/tzdb/tzdb_list_private.h
index 969b2b9f8a9f6..0be7b8e5a1005 100644
--- a/libcxx/src/include/tzdb/tzdb_list_private.h
+++ b/libcxx/src/include/tzdb/tzdb_list_private.h
@@ -15,8 +15,11 @@
 #include <forward_list>
 
 // When threads are not available the locking is not required.
+// When threads are available, we use std::mutex over std::shared_mutex
+// due to the increased overhead of std::shared_mutex.
+// See shared_mutex_vs_mutex.bench.cpp
 #ifndef _LIBCPP_HAS_NO_THREADS
-#  include <shared_mutex>
+#  include <mutex>
 #endif
 
 #include "types_private.h"
@@ -56,7 +59,7 @@ class tzdb_list::__impl {
 
   const tzdb& __front() const noexcept {
 #ifndef _LIBCPP_HAS_NO_THREADS
-    shared_lock __lock{__mutex_};
+    unique_lock __lock{__mutex_};
 #endif
     return __tzdb_.front();
   }
@@ -72,7 +75,7 @@ class tzdb_list::__impl {
 
   const_iterator __begin() const noexcept {
 #ifndef _LIBCPP_HAS_NO_THREADS
-    shared_lock __lock{__mutex_};
+    unique_lock __lock{__mutex_};
 #endif
     return __tzdb_.begin();
   }
@@ -87,7 +90,7 @@ class tzdb_list::__impl {
   void __load_no_lock() { chrono::__init_tzdb(__tzdb_.emplace_front(), __rules_.emplace_front()); }
 
 #ifndef _LIBCPP_HAS_NO_THREADS
-  mutable shared_mutex __mutex_;
+  mutable mutex __mutex_;
 #endif
   forward_list<tzdb> __tzdb_;
 

From 9931e7efc6505002e866ee553af5d812123c22f5 Mon Sep 17 00:00:00 2001
From: Hristo Hristov <hghristov.rmm@gmail.com>
Date: Sat, 13 Apr 2024 11:39:38 +0300
Subject: [PATCH 331/886] [libc++][spaceship][NFC] P0905R1: Symmetry for
 spaceship - mark as implemented (#87871)

Implemented: https://wg21.link/P0905R1 in
0913ca197855626c93fdae4184ae80ff5a26503a
- https://eel.is/c++draft/cmp.weakeq (Removed in
https://wg21.link/P1959R0)
- https://eel.is/c++draft/cmp.strongeq (Removed in
https://wg21.link/P1959R0)
- https://eel.is/c++draft/cmp.partialord
-
https://github.com/llvm/llvm-project/blob/4abb722ffa7fcf809faa4a479fdf2f78c685b351/libcxx/include/__compare/ordering.h#L101
- https://eel.is/c++draft/cmp.weakord
-
https://github.com/llvm/llvm-project/blob/4abb722ffa7fcf809faa4a479fdf2f78c685b351/libcxx/include/__compare/ordering.h#L173
- https://eel.is/c++draft/cmp.strongord
-
https://github.com/llvm/llvm-project/blob/4abb722ffa7fcf809faa4a479fdf2f78c685b351/libcxx/include/__compare/ordering.h#L250

References:
- https://wg21.link/P1959R0 - Remove `std::weak_equality` and
`std::strong_equality`
---
 libcxx/docs/Status/Cxx20Papers.csv     | 2 +-
 libcxx/docs/Status/SpaceshipPapers.csv | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/libcxx/docs/Status/Cxx20Papers.csv b/libcxx/docs/Status/Cxx20Papers.csv
index 77078b11a7278..6e4678da28887 100644
--- a/libcxx/docs/Status/Cxx20Papers.csv
+++ b/libcxx/docs/Status/Cxx20Papers.csv
@@ -23,7 +23,7 @@
 "`P0754R2 <https://wg21.link/P0754R2>`__","LWG","<version>","Jacksonville","|Complete|","7.0"
 "`P0809R0 <https://wg21.link/P0809R0>`__","LWG","Comparing Unordered Containers","Jacksonville","|Nothing To Do|",""
 "`P0858R0 <https://wg21.link/P0858R0>`__","LWG","Constexpr iterator requirements","Jacksonville","|Complete|","12.0"
-"`P0905R1 <https://wg21.link/P0905R1>`__","CWG","Symmetry for spaceship","Jacksonville","","","|spaceship|"
+"`P0905R1 <https://wg21.link/P0905R1>`__","CWG","Symmetry for spaceship","Jacksonville","|Complete|","7.0","|spaceship|"
 "`P0966R1 <https://wg21.link/P0966R1>`__","LWG","``string::reserve``\  Should Not Shrink","Jacksonville","|Complete| [#note-P0966]_","12.0"
 "","","","","","",""
 "`P0019R8 <https://wg21.link/P0019R8>`__","LWG","Atomic Ref","Rapperswil","",""
diff --git a/libcxx/docs/Status/SpaceshipPapers.csv b/libcxx/docs/Status/SpaceshipPapers.csv
index 21b788ee42921..39e1f968c1754 100644
--- a/libcxx/docs/Status/SpaceshipPapers.csv
+++ b/libcxx/docs/Status/SpaceshipPapers.csv
@@ -7,6 +7,6 @@
 `LWG3360 <https://wg21.link/LWG3360>`_,``three_way_comparable_with`` is inconsistent with similar concepts,|Nothing To Do|,
 `LWG3380 <https://wg21.link/LWG3380>`_,``common_type`` and comparison categories,|Nothing To Do|,
 `LWG3395 <https://wg21.link/LWG3395>`_,Definition for *three-way* comparison needs to be updated,|Nothing To Do|,
-`P0905R1 <https://wg21.link/P0905R1>`_,Symmetry for spaceship,,
+`P0905R1 <https://wg21.link/P0905R1>`_,Symmetry for spaceship,|Complete|,7.0
 `P1120R0 <https://wg21.link/P1120R0>`_,Consistency improvements for ``<=>`` and other comparison operators,,
 `LWG3431 <https://wg21.link/LWG3431>`_,``<=>`` for containers should require ``three_way_comparable<T>`` instead of ``<=>``,,

From e481f565286e7fbf5dda9c4a0d00ca1daae8abeb Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 11:50:02 +0200
Subject: [PATCH 332/886] [lldb][libc++] Adds local_t clock data formatters.
 (#88178)

---
 .../Language/CPlusPlus/CPlusPlusLanguage.cpp  |  25 ++++-
 .../Plugins/Language/CPlusPlus/LibCxx.cpp     |  38 ++++++-
 .../Plugins/Language/CPlusPlus/LibCxx.h       |   8 ++
 .../chrono/TestDataFormatterLibcxxChrono.py   | 104 ++++++++++++++++++
 .../data-formatter-stl/libcxx/chrono/main.cpp |  52 +++++++++
 5 files changed, 221 insertions(+), 6 deletions(-)

diff --git a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
index 4a536096a066f..afb683f7d846a 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
@@ -1050,7 +1050,7 @@ static void LoadLibCxxFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
                 "libc++ std::chrono::sys_seconds summary provider",
                 "^std::__[[:alnum:]]+::chrono::time_point<"
                 "std::__[[:alnum:]]+::chrono::system_clock, "
-                "std::__[[:alnum:]]+::chrono::duration<long long, "
+                "std::__[[:alnum:]]+::chrono::duration<.*, "
                 "std::__[[:alnum:]]+::ratio<1, 1> "
                 "> >$",
                 eTypeOptionHideChildren | eTypeOptionHideValue |
@@ -1068,6 +1068,29 @@ static void LoadLibCxxFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
                     eTypeOptionCascade,
                 true);
 
+  AddCXXSummary(
+      cpp_category_sp,
+      lldb_private::formatters::LibcxxChronoLocalSecondsSummaryProvider,
+      "libc++ std::chrono::local_seconds summary provider",
+      "^std::__[[:alnum:]]+::chrono::time_point<"
+      "std::__[[:alnum:]]+::chrono::local_t, "
+      "std::__[[:alnum:]]+::chrono::duration<.*, "
+      "std::__[[:alnum:]]+::ratio<1, 1> "
+      "> >$",
+      eTypeOptionHideChildren | eTypeOptionHideValue | eTypeOptionCascade,
+      true);
+  AddCXXSummary(cpp_category_sp,
+                lldb_private::formatters::LibcxxChronoLocalDaysSummaryProvider,
+                "libc++ std::chrono::local_seconds summary provider",
+                "^std::__[[:alnum:]]+::chrono::time_point<"
+                "std::__[[:alnum:]]+::chrono::local_t, "
+                "std::__[[:alnum:]]+::chrono::duration<int, "
+                "std::__[[:alnum:]]+::ratio<86400, 1> "
+                "> >$",
+                eTypeOptionHideChildren | eTypeOptionHideValue |
+                    eTypeOptionCascade,
+                true);
+
   // Chrono calendar types
 
   cpp_category_sp->AddTypeSummary(
diff --git a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.cpp b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.cpp
index d2d50152c07cf..e160fd0763939 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.cpp
@@ -1087,8 +1087,10 @@ bool lldb_private::formatters::LibcxxWStringViewSummaryProvider(
                                         dataobj, size);
 }
 
-bool lldb_private::formatters::LibcxxChronoSysSecondsSummaryProvider(
-    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
+static bool
+LibcxxChronoTimePointSecondsSummaryProvider(ValueObject &valobj, Stream &stream,
+                                            const TypeSummaryOptions &options,
+                                            const char *fmt) {
   ValueObjectSP ptr_sp = valobj.GetChildMemberWithName("__d_");
   if (!ptr_sp)
     return false;
@@ -1112,7 +1114,7 @@ bool lldb_private::formatters::LibcxxChronoSysSecondsSummaryProvider(
   else {
     std::array<char, 128> str;
     std::size_t size =
-        std::strftime(str.data(), str.size(), "%FT%H:%M:%SZ", gmtime(&seconds));
+        std::strftime(str.data(), str.size(), fmt, gmtime(&seconds));
     if (size == 0)
       return false;
 
@@ -1123,8 +1125,22 @@ bool lldb_private::formatters::LibcxxChronoSysSecondsSummaryProvider(
   return true;
 }
 
-bool lldb_private::formatters::LibcxxChronoSysDaysSummaryProvider(
+bool lldb_private::formatters::LibcxxChronoSysSecondsSummaryProvider(
     ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
+  return LibcxxChronoTimePointSecondsSummaryProvider(valobj, stream, options,
+                                                     "%FT%H:%M:%SZ");
+}
+
+bool lldb_private::formatters::LibcxxChronoLocalSecondsSummaryProvider(
+    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
+  return LibcxxChronoTimePointSecondsSummaryProvider(valobj, stream, options,
+                                                     "%FT%H:%M:%S");
+}
+
+static bool
+LibcxxChronoTimepointDaysSummaryProvider(ValueObject &valobj, Stream &stream,
+                                         const TypeSummaryOptions &options,
+                                         const char *fmt) {
   ValueObjectSP ptr_sp = valobj.GetChildMemberWithName("__d_");
   if (!ptr_sp)
     return false;
@@ -1148,7 +1164,7 @@ bool lldb_private::formatters::LibcxxChronoSysDaysSummaryProvider(
 
     std::array<char, 128> str;
     std::size_t size =
-        std::strftime(str.data(), str.size(), "%FZ", gmtime(&seconds));
+        std::strftime(str.data(), str.size(), fmt, gmtime(&seconds));
     if (size == 0)
       return false;
 
@@ -1158,6 +1174,18 @@ bool lldb_private::formatters::LibcxxChronoSysDaysSummaryProvider(
   return true;
 }
 
+bool lldb_private::formatters::LibcxxChronoSysDaysSummaryProvider(
+    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
+  return LibcxxChronoTimepointDaysSummaryProvider(valobj, stream, options,
+                                                  "%FZ");
+}
+
+bool lldb_private::formatters::LibcxxChronoLocalDaysSummaryProvider(
+    ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
+  return LibcxxChronoTimepointDaysSummaryProvider(valobj, stream, options,
+                                                  "%F");
+}
+
 bool lldb_private::formatters::LibcxxChronoMonthSummaryProvider(
     ValueObject &valobj, Stream &stream, const TypeSummaryOptions &options) {
   // FIXME: These are the names used in the C++20 ostream operator. Since LLVM
diff --git a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
index d8b807d180e06..8e97174dc3075 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
+++ b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
@@ -281,6 +281,14 @@ bool LibcxxChronoSysDaysSummaryProvider(
     ValueObject &valobj, Stream &stream,
     const TypeSummaryOptions &options); // libc++ std::chrono::sys_days
 
+bool LibcxxChronoLocalSecondsSummaryProvider(
+    ValueObject &valobj, Stream &stream,
+    const TypeSummaryOptions &options); // libc++ std::chrono::local_seconds
+
+bool LibcxxChronoLocalDaysSummaryProvider(
+    ValueObject &valobj, Stream &stream,
+    const TypeSummaryOptions &options); // libc++ std::chrono::local_days
+
 bool LibcxxChronoMonthSummaryProvider(
     ValueObject &valobj, Stream &stream,
     const TypeSummaryOptions &options); // libc++ std::chrono::month
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/TestDataFormatterLibcxxChrono.py b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/TestDataFormatterLibcxxChrono.py
index c306315895960..fb35481d55514 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/TestDataFormatterLibcxxChrono.py
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/TestDataFormatterLibcxxChrono.py
@@ -128,6 +128,110 @@ def test_with_run_command(self):
             substrs=["sd_max = timestamp=2147483647 days"],
         )
 
+        # local_seconds aliasses
+
+        self.expect(
+            "frame variable ls_tp",
+            substrs=["ls_tp = date/time=1970-01-01T00:00:00 timestamp=0 s"],
+        )
+        self.expect(
+            "frame variable ls_tp_d",
+            substrs=["ls_tp_d = date/time=1970-01-01T00:00:00 timestamp=0 s"],
+        )
+        self.expect(
+            "frame variable ls_tp_d_r",
+            substrs=["ls_tp_d_r = date/time=1970-01-01T00:00:00 timestamp=0 s"],
+        )
+        self.expect(
+            "frame variable ls_tp_d_r2",
+            substrs=["ls_tp_d_r2 = date/time=1970-01-01T00:00:00 timestamp=0 s"],
+        )
+
+        # local_seconds
+
+        self.expect(
+            "frame variable ls_0",
+            substrs=["ls_0 = date/time=1970-01-01T00:00:00 timestamp=0 s"],
+        )
+
+        self.expect(
+            "frame variable ls_neg_date_time",
+            substrs=[
+                "ls_neg_date_time = date/time=-32767-01-01T00:00:00 timestamp=-1096193779200 s"
+            ],
+        )
+        self.expect(
+            "frame variable ls_neg_seconds",
+            substrs=["ls_neg_seconds = timestamp=-1096193779201 s"],
+        )
+
+        self.expect(
+            "frame variable ls_pos_date_time",
+            substrs=[
+                "ls_pos_date_time = date/time=32767-12-31T23:59:59 timestamp=971890963199 s"
+            ],
+        )
+        self.expect(
+            "frame variable ls_pos_seconds",
+            substrs=["ls_pos_seconds = timestamp=971890963200 s"],
+        )
+
+        self.expect(
+            "frame variable ls_min",
+            substrs=["ls_min = timestamp=-9223372036854775808 s"],
+        )
+        self.expect(
+            "frame variable ls_max",
+            substrs=["ls_max = timestamp=9223372036854775807 s"],
+        )
+
+        # local_days aliasses
+
+        self.expect(
+            "frame variable ld_tp",
+            substrs=["ld_tp = date=1970-01-01 timestamp=0 days"],
+        )
+        self.expect(
+            "frame variable ld_tp_d_r",
+            substrs=["ld_tp_d_r = date=1970-01-01 timestamp=0 days"],
+        )
+        self.expect(
+            "frame variable ld_tp_d_r2",
+            substrs=["ld_tp_d_r2 = date=1970-01-01 timestamp=0 days"],
+        )
+
+        # local_days
+
+        self.expect(
+            "frame variable ld_0", substrs=["ld_0 = date=1970-01-01 timestamp=0 days"]
+        )
+        self.expect(
+            "frame variable ld_neg_date",
+            substrs=["ld_neg_date = date=-32767-01-01 timestamp=-12687428 days"],
+        )
+        self.expect(
+            "frame variable ld_neg_days",
+            substrs=["ld_neg_days = timestamp=-12687429 days"],
+        )
+
+        self.expect(
+            "frame variable ld_pos_date",
+            substrs=["ld_pos_date = date=32767-12-31 timestamp=11248737 days"],
+        )
+        self.expect(
+            "frame variable ld_pos_days",
+            substrs=["ld_pos_days = timestamp=11248738 days"],
+        )
+
+        self.expect(
+            "frame variable ld_min",
+            substrs=["ld_min = timestamp=-2147483648 days"],
+        )
+        self.expect(
+            "frame variable ld_max",
+            substrs=["ld_max = timestamp=2147483647 days"],
+        )
+
         self.expect("frame variable d_0", substrs=["d_0 = day=0"])
         self.expect("frame variable d_1", substrs=["d_1 = day=1"])
         self.expect("frame variable d_31", substrs=["d_31 = day=31"])
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/main.cpp b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/main.cpp
index 315c88a787d82..0ea42d1a107e1 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/main.cpp
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/chrono/main.cpp
@@ -67,6 +67,58 @@ int main() {
   std::chrono::sys_days sd_max{
       std::chrono::days{std::numeric_limits<int>::max()}};
 
+  // local_seconds aliasses
+  std::chrono::time_point<std::chrono::local_t, std::chrono::seconds> ls_tp{
+      std::chrono::seconds{0}};
+  std::chrono::time_point<std::chrono::local_t,
+                          std::chrono::duration<long long>>
+      ls_tp_d{std::chrono::seconds{0}};
+  std::chrono::time_point<std::chrono::local_t,
+                          std::chrono::duration<long long, std::ratio<1>>>
+      ls_tp_d_r{std::chrono::seconds{0}};
+  std::chrono::time_point<std::chrono::local_t,
+                          std::chrono::duration<long long, std::ratio<1>>>
+      ls_tp_d_r2{std::chrono::seconds{0}};
+
+  // local_seconds
+  std::chrono::local_seconds ls_0{std::chrono::seconds{0}};
+  std::chrono::local_seconds ls_neg_date_time{
+      std::chrono::seconds{-1'096'193'779'200}};
+  std::chrono::local_seconds ls_neg_seconds{
+      std::chrono::seconds{-1'096'193'779'201}};
+  std::chrono::local_seconds ls_pos_date_time{
+      std::chrono::seconds{971'890'963'199}};
+  std::chrono::local_seconds ls_pos_seconds{
+      std::chrono::seconds{971'890'963'200}};
+  std::chrono::local_seconds ls_min{
+      std::chrono::seconds{std::numeric_limits<long long>::min()}};
+  std::chrono::local_seconds ls_max{
+      std::chrono::seconds{std::numeric_limits<long long>::max()}};
+
+  // local_days aliasses
+  std::chrono::time_point<std::chrono::local_t, std::chrono::days> ld_tp{
+      std::chrono::days{0}};
+  std::chrono::time_point<std::chrono::local_t,
+                          std::chrono::duration<int, std::ratio<86400>>>
+      ld_tp_d_r{std::chrono::days{0}};
+  std::chrono::time_point<std::chrono::local_t,
+                          std::chrono::duration<int, std::ratio<86400, 1>>>
+      ld_tp_d_r2{std::chrono::days{0}};
+
+  // local_days
+  std::chrono::local_days ld_0{std::chrono::days{0}};
+
+  std::chrono::local_days ld_neg_date{std::chrono::days{-12'687'428}};
+  std::chrono::local_days ld_neg_days{std::chrono::days{-12'687'429}};
+
+  std::chrono::local_days ld_pos_date{std::chrono::days{11'248'737}};
+  std::chrono::local_days ld_pos_days{std::chrono::days{11'248'738}};
+
+  std::chrono::local_days ld_min{
+      std::chrono::days{std::numeric_limits<int>::min()}};
+  std::chrono::local_days ld_max{
+      std::chrono::days{std::numeric_limits<int>::max()}};
+
   std::chrono::day d_0{0};
   std::chrono::day d_1{1};
   std::chrono::day d_31{31};

From a33a754cff02fda7520fa63fbc6c85c274a44baa Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 12:03:18 +0200
Subject: [PATCH 333/886] [lldb] Fixes comment typos.

---
 lldb/source/Plugins/Language/CPlusPlus/LibCxxSliceArray.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lldb/source/Plugins/Language/CPlusPlus/LibCxxSliceArray.cpp b/lldb/source/Plugins/Language/CPlusPlus/LibCxxSliceArray.cpp
index 32e67d2e38c5d..c33d8e91af709 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/LibCxxSliceArray.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/LibCxxSliceArray.cpp
@@ -73,9 +73,9 @@ class LibcxxStdSliceArraySyntheticFrontEnd : public SyntheticChildrenFrontEnd {
   size_t m_size = 0;
   /// slice_array.__stride_.
   size_t m_stride = 0;
-  /// The type of slize_array's template argument T.
+  /// The type of slice_array's template argument T.
   CompilerType m_element_type;
-  /// The sizeof slize_array's template argument T.
+  /// The sizeof slice_array's template argument T.
   uint32_t m_element_size = 0;
 };
 

From 0412a8651aa6cbdd697e904a758e0f95e6635cee Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 13:10:09 +0300
Subject: [PATCH 334/886] [clang] Add missing documentation entry for
 `__is_pointer_interconvertible_base_of()`

---
 clang/docs/LanguageExtensions.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/clang/docs/LanguageExtensions.rst b/clang/docs/LanguageExtensions.rst
index 7b23e4d1c2f30..96691b45d63a3 100644
--- a/clang/docs/LanguageExtensions.rst
+++ b/clang/docs/LanguageExtensions.rst
@@ -1610,6 +1610,7 @@ The following type trait primitives are supported by Clang. Those traits marked
 * ``__is_pod`` (C++, GNU, Microsoft, Embarcadero):
   Note, the corresponding standard trait was deprecated in C++20.
 * ``__is_pointer`` (C++, Embarcadero)
+* ``__is_pointer_interconvertible_base_of`` (C++, GNU, Microsoft)
 * ``__is_polymorphic`` (C++, GNU, Microsoft, Embarcadero)
 * ``__is_reference`` (C++, Embarcadero)
 * ``__is_referenceable`` (C++, GNU, Microsoft, Embarcadero):

From 6704faf6f8a26c21c71fd0efad6af6a8d6ee52b4 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Sat, 13 Apr 2024 11:12:33 +0100
Subject: [PATCH 335/886] [SLP] Use StoreTy to compute min VF.

This ensures that MinVF is a power-of-2, even if ValueTy's width is
not a power-of-2.

This should fix a number of buildbot failures with X86 bootstrapping.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    |   2 +-
 .../trunc-store-value-ty-not-power-of-2.ll    | 109 ++++++++++++++++++
 2 files changed, 110 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 09db50132c892..b031b40a978f5 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15137,7 +15137,7 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
       if (auto *Trunc = dyn_cast<TruncInst>(Store->getValueOperand()))
         ValueTy = Trunc->getSrcTy();
       unsigned MinVF = TTI->getStoreMinimumVF(
-          R.getMinVF(DL->getTypeSizeInBits(ValueTy)), StoreTy, ValueTy);
+          R.getMinVF(DL->getTypeSizeInBits(StoreTy)), StoreTy, ValueTy);
 
       if (MaxVF < MinVF) {
         LLVM_DEBUG(dbgs() << "SLP: Vectorization infeasible as MaxVF (" << MaxVF
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll b/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll
new file mode 100644
index 0000000000000..81b4ee40e7fdf
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll
@@ -0,0 +1,109 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -p slp-vectorizer -S %s | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @test_2_i24_stores(ptr %A) {
+; CHECK-LABEL: define void @test_2_i24_stores(
+; CHECK-SAME: ptr [[A:%.*]]) {
+; CHECK-NEXT:    [[L:%.*]] = load i24, ptr [[A]], align 4
+; CHECK-NEXT:    store i24 [[L]], ptr [[A]], align 1
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i24, ptr [[A]], i64 1
+; CHECK-NEXT:    store i24 0, ptr [[GEP]], align 1
+; CHECK-NEXT:    ret void
+;
+  %l = load i24, ptr %A
+  store i24 %l, ptr %A, align 1
+  %gep = getelementptr i24, ptr %A, i64 1
+  store i24 0, ptr %gep, align 1
+  ret void
+}
+
+define void @test_2_trunc_i24_to_i8(i24 %x, ptr %A) {
+; CHECK-LABEL: define void @test_2_trunc_i24_to_i8(
+; CHECK-SAME: i24 [[X:%.*]], ptr [[A:%.*]]) {
+; CHECK-NEXT:    [[T:%.*]] = trunc i24 [[X]] to i8
+; CHECK-NEXT:    store i8 [[T]], ptr [[A]], align 1
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i8, ptr [[A]], i64 1
+; CHECK-NEXT:    store i8 0, ptr [[GEP]], align 1
+; CHECK-NEXT:    ret void
+;
+  %t = trunc i24 %x to i8
+  store i8 %t, ptr %A, align 1
+  %gep = getelementptr i8, ptr %A, i64 1
+  store i8 0, ptr %gep, align 1
+  ret void
+}
+
+define void @test_4_trunc_i24_to_i8(i24 %x, ptr %A) {
+; CHECK-LABEL: define void @test_4_trunc_i24_to_i8(
+; CHECK-SAME: i24 [[X:%.*]], ptr [[A:%.*]]) {
+; CHECK-NEXT:    [[T:%.*]] = trunc i24 [[X]] to i8
+; CHECK-NEXT:    store i8 [[T]], ptr [[A]], align 1
+; CHECK-NEXT:    [[GEP_1:%.*]] = getelementptr i8, ptr [[A]], i64 1
+; CHECK-NEXT:    store i8 [[T]], ptr [[GEP_1]], align 1
+; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr i8, ptr [[A]], i64 2
+; CHECK-NEXT:    store i8 [[T]], ptr [[GEP_2]], align 1
+; CHECK-NEXT:    [[GEP_3:%.*]] = getelementptr i8, ptr [[A]], i64 3
+; CHECK-NEXT:    store i8 [[T]], ptr [[GEP_3]], align 1
+; CHECK-NEXT:    ret void
+;
+  %t = trunc i24 %x to i8
+  store i8 %t, ptr %A, align 1
+  %gep.1 = getelementptr i8, ptr %A, i64 1
+  store i8 %t, ptr %gep.1, align 1
+  %gep.2 = getelementptr i8, ptr %A, i64 2
+  store i8 %t, ptr %gep.2, align 1
+  %gep.3 = getelementptr i8, ptr %A, i64 3
+  store i8 %t, ptr %gep.3, align 1
+  ret void
+}
+
+define void @test_8_trunc_i24_to_i8(i24 %x, ptr %A) {
+; CHECK-LABEL: define void @test_8_trunc_i24_to_i8(
+; CHECK-SAME: i24 [[X:%.*]], ptr [[A:%.*]]) {
+; CHECK-NEXT:    [[T:%.*]] = trunc i24 [[X]] to i8
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> poison, i8 [[T]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> poison, <8 x i32> zeroinitializer
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], ptr [[A]], align 1
+; CHECK-NEXT:    ret void
+;
+  %t = trunc i24 %x to i8
+  store i8 %t, ptr %A, align 1
+  %gep.1 = getelementptr i8, ptr %A, i64 1
+  store i8 %t, ptr %gep.1, align 1
+  %gep.2 = getelementptr i8, ptr %A, i64 2
+  store i8 %t, ptr %gep.2, align 1
+  %gep.3 = getelementptr i8, ptr %A, i64 3
+  store i8 %t, ptr %gep.3, align 1
+  %gep.4 = getelementptr i8, ptr %A, i64 4
+  store i8 %t, ptr %gep.4, align 1
+  %gep.5 = getelementptr i8, ptr %A, i64 5
+  store i8 %t, ptr %gep.5, align 1
+  %gep.6 = getelementptr i8, ptr %A, i64 6
+  store i8 %t, ptr %gep.6, align 1
+  %gep.7 = getelementptr i8, ptr %A, i64 7
+  store i8 %t, ptr %gep.7, align 1
+  ret void
+}
+
+define void @test_4_trunc_i24_to_i16(i24 %x, ptr %A) {
+; CHECK-LABEL: define void @test_4_trunc_i24_to_i16(
+; CHECK-SAME: i24 [[X:%.*]], ptr [[A:%.*]]) {
+; CHECK-NEXT:    [[T:%.*]] = trunc i24 [[X]] to i16
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[T]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    store <4 x i16> [[TMP2]], ptr [[A]], align 1
+; CHECK-NEXT:    ret void
+;
+  %t = trunc i24 %x to i16
+  store i16 %t, ptr %A, align 1
+  %gep.1 = getelementptr i16, ptr %A, i64 1
+  store i16 %t, ptr %gep.1, align 1
+  %gep.2 = getelementptr i16, ptr %A, i64 2
+  store i16 %t, ptr %gep.2, align 1
+  %gep.3 = getelementptr i16, ptr %A, i64 3
+  store i16 %t, ptr %gep.3, align 1
+  ret void
+}

From 6775285e7695f2d45cf455f5d31b2c9fa9362d3d Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 12:32:46 +0200
Subject: [PATCH 336/886] [libc++][CMake] Removes LIBCXX_ENABLE_CLANG_TIDY.
 (#85794)

The clang-tidy selection in CMake was refactored in
https://github.com/llvm/llvm-project/pull/81362. During review it was
suggested to remove this CMake option.
---
 libcxx/CMakeLists.txt                         |  5 --
 libcxx/docs/ReleaseNotes/19.rst               |  3 +
 libcxx/test/tools/CMakeLists.txt              | 12 ++--
 .../tools/clang_tidy_checks/CMakeLists.txt    | 56 ++++++++++++++-----
 libcxx/utils/ci/buildkite-pipeline.yml        |  1 -
 libcxx/utils/ci/run-buildbot                  |  8 ---
 6 files changed, 48 insertions(+), 37 deletions(-)

diff --git a/libcxx/CMakeLists.txt b/libcxx/CMakeLists.txt
index e565c47c76687..043d5a8295c1a 100644
--- a/libcxx/CMakeLists.txt
+++ b/libcxx/CMakeLists.txt
@@ -123,7 +123,6 @@ option(LIBCXX_ENABLE_VENDOR_AVAILABILITY_ANNOTATIONS
    to provide compile-time errors when using features unavailable on some version of
    the shared library they shipped should turn this on and see `include/__availability`
    for more details." OFF)
-option(LIBCXX_ENABLE_CLANG_TIDY "Whether to compile and run clang-tidy checks" OFF)
 
 if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
   set(LIBCXX_DEFAULT_TEST_CONFIG "llvm-libc++-shared-gcc.cfg.in")
@@ -863,10 +862,6 @@ add_subdirectory(modules)
 
 set(LIBCXX_TEST_DEPS "cxx_experimental")
 
-if (LIBCXX_ENABLE_CLANG_TIDY)
-  list(APPEND LIBCXX_TEST_DEPS cxx-tidy)
-endif()
-
 list(APPEND LIBCXX_TEST_DEPS generate-cxx-modules)
 
 if (LIBCXX_INCLUDE_BENCHMARKS)
diff --git a/libcxx/docs/ReleaseNotes/19.rst b/libcxx/docs/ReleaseNotes/19.rst
index 7bc0148c9ff0a..390de789ddb64 100644
--- a/libcxx/docs/ReleaseNotes/19.rst
+++ b/libcxx/docs/ReleaseNotes/19.rst
@@ -132,3 +132,6 @@ Build System Changes
 
 - The ``LIBCXX_EXECUTOR`` and ``LIBCXXABI_EXECUTOR`` CMake variables have been removed. Please
   set ``LIBCXX_TEST_PARAMS`` to ``executor=<...>`` instead.
+
+- The Cmake variable ``LIBCXX_ENABLE_CLANG_TIDY`` has been removed. The build system has been changed
+  to automatically detect the presence of ``clang-tidy`` and the required ``Clang`` libraries.
diff --git a/libcxx/test/tools/CMakeLists.txt b/libcxx/test/tools/CMakeLists.txt
index e30ad6cdd8201..6d99c53ad46d9 100644
--- a/libcxx/test/tools/CMakeLists.txt
+++ b/libcxx/test/tools/CMakeLists.txt
@@ -1,12 +1,8 @@
 
 set(LIBCXX_TEST_TOOLS_PATH ${CMAKE_CURRENT_BINARY_DIR} PARENT_SCOPE)
 
-# TODO: Remove LIBCXX_ENABLE_CLANG_TIDY
-if(LIBCXX_ENABLE_CLANG_TIDY)
-  if(NOT CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
-    message(STATUS "Clang-tidy can only be used when building libc++ with "
-                   "a clang compiler.")
-    return()
-  endif()
-  add_subdirectory(clang_tidy_checks)
+if(NOT CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+  message(STATUS "Clang-tidy tests are disabled due to non-clang based compiler.")
+  return()
 endif()
+add_subdirectory(clang_tidy_checks)
diff --git a/libcxx/test/tools/clang_tidy_checks/CMakeLists.txt b/libcxx/test/tools/clang_tidy_checks/CMakeLists.txt
index 74905a0c3ed1c..28eed61445833 100644
--- a/libcxx/test/tools/clang_tidy_checks/CMakeLists.txt
+++ b/libcxx/test/tools/clang_tidy_checks/CMakeLists.txt
@@ -9,22 +9,14 @@ set(Clang_DIR_SAVE ${Clang_DIR})
 # versions must match. Otherwise there likely will be ODR-violations. This had
 # led to crashes and incorrect output of the clang-tidy based checks.
 find_package(Clang ${CMAKE_CXX_COMPILER_VERSION})
-
-set(SOURCES
-    abi_tag_on_virtual.cpp
-    header_exportable_declarations.cpp
-    hide_from_abi.cpp
-    proper_version_checks.cpp
-    qualify_declval.cpp
-    robust_against_adl.cpp
-    uglify_attributes.cpp
-
-    libcpp_module.cpp
-   )
-
 if(NOT Clang_FOUND)
-  message(STATUS "Could not find a suitable version of the Clang development package;
-                  custom libc++ clang-tidy checks will not be available.")
+  message(STATUS "Clang-tidy tests are disabled since the "
+                 "Clang development package is unavailable.")
+  return()
+endif()
+if(NOT TARGET clangTidy)
+  message(STATUS "Clang-tidy tests are disabled since the "
+                 "Clang development package has no clangTidy target.")
   return()
 endif()
 
@@ -54,6 +46,38 @@ if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
                         )
 endif()
 
+# In some cases even with the clangTidy target present the headers appear not to
+# be on the system. Run a short test to see whether the header is present.
+file(WRITE "${CMAKE_CURRENT_BINARY_DIR}/test.cpp" "
+#if !__has_include(\"clang-tidy/ClangTidyCheck.h\")
+  # error No clang-tidy headers
+#endif
+int main(){}
+")
+try_compile(HAS_CLANG_TIDY_HEADERS
+  "${CMAKE_CURRENT_BINARY_DIR}"
+  "${CMAKE_CURRENT_BINARY_DIR}/test.cpp"
+   LINK_LIBRARIES clangTidy)
+
+if(NOT HAS_CLANG_TIDY_HEADERS)
+  message(STATUS "Clang-tidy tests are disabled since the "
+                 "clang-tidy headers are not present.")
+  return()
+endif()
+message(STATUS "Clang-tidy tests are enabled.")
+
+set(SOURCES
+    abi_tag_on_virtual.cpp
+    header_exportable_declarations.cpp
+    hide_from_abi.cpp
+    proper_version_checks.cpp
+    qualify_declval.cpp
+    robust_against_adl.cpp
+    uglify_attributes.cpp
+
+    libcpp_module.cpp
+   )
+
 add_library(cxx-tidy MODULE ${SOURCES})
 target_link_libraries(cxx-tidy clangTidy)
 
@@ -64,3 +88,5 @@ set_target_properties(cxx-tidy PROPERTIES
 
 set_target_properties(cxx-tidy PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
 set(CMAKE_SHARED_MODULE_SUFFIX_CXX .plugin) # Use a portable suffix to simplify how we can find it from Lit
+
+list(APPEND LIBCXX_TEST_DEPS cxx-tidy)
diff --git a/libcxx/utils/ci/buildkite-pipeline.yml b/libcxx/utils/ci/buildkite-pipeline.yml
index c43e414418729..4bacdec8f8d6b 100644
--- a/libcxx/utils/ci/buildkite-pipeline.yml
+++ b/libcxx/utils/ci/buildkite-pipeline.yml
@@ -43,7 +43,6 @@ definitions:
 
 environment_definitions:
   _common_env: &common_env
-      ENABLE_CLANG_TIDY: "On"
       LLVM_SYMBOLIZER_PATH: "/usr/bin/llvm-symbolizer-${LLVM_HEAD_VERSION}"
       CLANG_CRASH_DIAGNOSTICS_DIR: "crash_diagnostics"
       CC: clang-${LLVM_HEAD_VERSION}
diff --git a/libcxx/utils/ci/run-buildbot b/libcxx/utils/ci/run-buildbot
index a6f3eb174308b..cc72f4639b1a9 100755
--- a/libcxx/utils/ci/run-buildbot
+++ b/libcxx/utils/ci/run-buildbot
@@ -44,9 +44,6 @@ CMAKE               The CMake binary to use. This variable is optional.
 CLANG_FORMAT        The clang-format binary to use when generating the format
                     ignore list.
 
-ENABLE_CLANG_TIDY   Whether to compile and run clang-tidy checks. This variable
-                    is optional.
-
 EOF
 }
 
@@ -111,10 +108,6 @@ function clean() {
     rm -rf "${BUILD_DIR}"
 }
 
-if [ -z "${ENABLE_CLANG_TIDY}" ]; then
-    ENABLE_CLANG_TIDY=Off
-fi
-
 function generate-cmake-base() {
     echo "--- Generating CMake"
     ${CMAKE} \
@@ -126,7 +119,6 @@ function generate-cmake-base() {
           -DLIBCXX_ENABLE_WERROR=YES \
           -DLIBCXXABI_ENABLE_WERROR=YES \
           -DLIBUNWIND_ENABLE_WERROR=YES \
-          -DLIBCXX_ENABLE_CLANG_TIDY=${ENABLE_CLANG_TIDY} \
           -DLLVM_LIT_ARGS="-sv --xunit-xml-output test-results.xml --timeout=1500 --time-tests" \
           "${@}"
 }

From f6ca37bf1809b2fa8f6615d4a30eadf8f479c700 Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97@outlook.com>
Date: Sat, 13 Apr 2024 18:45:37 +0800
Subject: [PATCH 337/886] [NewPM] Set diagnostic handler in
 `MachineModuleAnalysis` (#88229)

`setDiagnosticHandler` is idempotent.
---
 llvm/lib/CodeGen/MachineModuleInfo.cpp     | 12 +++++-
 llvm/unittests/CodeGen/PassManagerTest.cpp | 50 +++++++++++++++++++++-
 2 files changed, 60 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/CodeGen/MachineModuleInfo.cpp b/llvm/lib/CodeGen/MachineModuleInfo.cpp
index f24288bb69de0..1dba591d02b6c 100644
--- a/llvm/lib/CodeGen/MachineModuleInfo.cpp
+++ b/llvm/lib/CodeGen/MachineModuleInfo.cpp
@@ -213,7 +213,6 @@ static unsigned getLocCookie(const SMDiagnostic &SMD, const SourceMgr &SrcMgr,
 bool MachineModuleInfoWrapperPass::doInitialization(Module &M) {
   MMI.initialize();
   MMI.TheModule = &M;
-  // FIXME: Do this for new pass manager.
   LLVMContext &Ctx = M.getContext();
   MMI.getContext().setDiagnosticHandler(
       [&Ctx, &M](const SMDiagnostic &SMD, bool IsInlineAsm,
@@ -240,6 +239,17 @@ AnalysisKey MachineModuleAnalysis::Key;
 MachineModuleAnalysis::Result
 MachineModuleAnalysis::run(Module &M, ModuleAnalysisManager &) {
   MMI.TheModule = &M;
+  LLVMContext &Ctx = M.getContext();
+  MMI.getContext().setDiagnosticHandler(
+      [&Ctx, &M](const SMDiagnostic &SMD, bool IsInlineAsm,
+                 const SourceMgr &SrcMgr,
+                 std::vector<const MDNode *> &LocInfos) {
+        unsigned LocCookie = 0;
+        if (IsInlineAsm)
+          LocCookie = getLocCookie(SMD, SrcMgr, LocInfos);
+        Ctx.diagnose(
+            DiagnosticInfoSrcMgr(SMD, M.getName(), IsInlineAsm, LocCookie));
+      });
   MMI.DbgInfoAvailable =
       !DisableDebugInfoPrinting && !M.debug_compile_units().empty();
   return Result(MMI);
diff --git a/llvm/unittests/CodeGen/PassManagerTest.cpp b/llvm/unittests/CodeGen/PassManagerTest.cpp
index 4283eb01a9c8f..71f8832d43656 100644
--- a/llvm/unittests/CodeGen/PassManagerTest.cpp
+++ b/llvm/unittests/CodeGen/PassManagerTest.cpp
@@ -125,6 +125,15 @@ struct TestMachineModulePass : public PassInfoMixin<TestMachineModulePass> {
   std::vector<int> &Counts;
 };
 
+struct ReportWarningPass : public PassInfoMixin<ReportWarningPass> {
+  PreservedAnalyses run(MachineFunction &MF,
+                        MachineFunctionAnalysisManager &MFAM) {
+    auto &Ctx = MF.getContext();
+    Ctx.reportWarning(SMLoc(), "Test warning message.");
+    return PreservedAnalyses::all();
+  }
+};
+
 std::unique_ptr<Module> parseIR(LLVMContext &Context, const char *IR) {
   SMDiagnostic Err;
   return parseAssemblyString(IR, Err, Context);
@@ -176,10 +185,10 @@ TEST_F(PassManagerTest, Basic) {
   MachineModuleInfo MMI(LLVMTM);
 
   LoopAnalysisManager LAM;
+  MachineFunctionAnalysisManager MFAM;
   FunctionAnalysisManager FAM;
   CGSCCAnalysisManager CGAM;
   ModuleAnalysisManager MAM;
-  MachineFunctionAnalysisManager MFAM;
   PassBuilder PB(TM.get());
   PB.registerModuleAnalyses(MAM);
   PB.registerCGSCCAnalyses(CGAM);
@@ -204,10 +213,49 @@ TEST_F(PassManagerTest, Basic) {
   MFPM.addPass(TestMachineFunctionPass(Count, Counts));
   MPM.addPass(createModuleToMachineFunctionPassAdaptor(std::move(MFPM)));
 
+  testing::internal::CaptureStderr();
   MPM.run(*M, MAM);
+  std::string Output = testing::internal::GetCapturedStderr();
 
   EXPECT_EQ((std::vector<int>{10, 16, 18, 20, 30, 36, 38, 40}), Counts);
   EXPECT_EQ(40, Count);
 }
 
+TEST_F(PassManagerTest, DiagnosticHandler) {
+  if (!TM)
+    GTEST_SKIP();
+
+  LLVMTargetMachine *LLVMTM = static_cast<LLVMTargetMachine *>(TM.get());
+  M->setDataLayout(TM->createDataLayout());
+
+  MachineModuleInfo MMI(LLVMTM);
+
+  LoopAnalysisManager LAM;
+  MachineFunctionAnalysisManager MFAM;
+  FunctionAnalysisManager FAM;
+  CGSCCAnalysisManager CGAM;
+  ModuleAnalysisManager MAM;
+  PassBuilder PB(TM.get());
+  PB.registerModuleAnalyses(MAM);
+  PB.registerCGSCCAnalyses(CGAM);
+  PB.registerFunctionAnalyses(FAM);
+  PB.registerLoopAnalyses(LAM);
+  PB.registerMachineFunctionAnalyses(MFAM);
+  PB.crossRegisterProxies(LAM, FAM, CGAM, MAM, &MFAM);
+
+  MAM.registerPass([&] { return MachineModuleAnalysis(MMI); });
+
+  ModulePassManager MPM;
+  FunctionPassManager FPM;
+  MachineFunctionPassManager MFPM;
+  MFPM.addPass(ReportWarningPass());
+  MPM.addPass(createModuleToMachineFunctionPassAdaptor(std::move(MFPM)));
+  testing::internal::CaptureStderr();
+  MPM.run(*M, MAM);
+  std::string Output = testing::internal::GetCapturedStderr();
+
+  EXPECT_TRUE(Output.find("warning: <unknown>:0: Test warning message.") !=
+              std::string::npos);
+}
+
 } // namespace

From 844b532713986999aa1ffed0883eff2d1339ec7a Mon Sep 17 00:00:00 2001
From: Andreas Jonson <andjo403@hotmail.com>
Date: Sat, 13 Apr 2024 13:01:36 +0200
Subject: [PATCH 338/886] [MergeFunc]  Handle ConstantRange attributes (#88584)

It is possible to update
[AttributeImpl::operator<](https://github.com/andjo403/llvm-project/blob/a9da350aadfb5c86d36ae18398471558b22c1309/llvm/lib/IR/Attributes.cpp#L744)
instead but feels strange to say that a range is less then an other
range.
---
 .../Transforms/Utils/FunctionComparator.cpp   |  12 ++
 .../call-and-invoke-with-ranges-attr.ll       | 111 ++++++++++++++++++
 2 files changed, 123 insertions(+)
 create mode 100644 llvm/test/Transforms/MergeFunc/call-and-invoke-with-ranges-attr.ll

diff --git a/llvm/lib/Transforms/Utils/FunctionComparator.cpp b/llvm/lib/Transforms/Utils/FunctionComparator.cpp
index 8cfb8ed07d240..67aeba7048f86 100644
--- a/llvm/lib/Transforms/Utils/FunctionComparator.cpp
+++ b/llvm/lib/Transforms/Utils/FunctionComparator.cpp
@@ -143,6 +143,18 @@ int FunctionComparator::cmpAttrs(const AttributeList L,
         if (int Res = cmpNumbers((uint64_t)TyL, (uint64_t)TyR))
           return Res;
         continue;
+      } else if (LA.isConstantRangeAttribute() &&
+                 RA.isConstantRangeAttribute()) {
+        if (LA.getKindAsEnum() != RA.getKindAsEnum())
+          return cmpNumbers(LA.getKindAsEnum(), RA.getKindAsEnum());
+
+        ConstantRange LCR = LA.getRange();
+        ConstantRange RCR = RA.getRange();
+        if (int Res = cmpAPInts(LCR.getLower(), RCR.getLower()))
+          return Res;
+        if (int Res = cmpAPInts(LCR.getUpper(), RCR.getUpper()))
+          return Res;
+        continue;
       }
       if (LA < RA)
         return -1;
diff --git a/llvm/test/Transforms/MergeFunc/call-and-invoke-with-ranges-attr.ll b/llvm/test/Transforms/MergeFunc/call-and-invoke-with-ranges-attr.ll
new file mode 100644
index 0000000000000..e5d62319bf9db
--- /dev/null
+++ b/llvm/test/Transforms/MergeFunc/call-and-invoke-with-ranges-attr.ll
@@ -0,0 +1,111 @@
+; RUN: opt -passes=mergefunc -S < %s | FileCheck %s
+
+define i8 @call_with_range_attr(i8 range(i8 0, 2) %v) {
+  %out = call i8 @dummy2(i8 %v)
+  ret i8 %out
+}
+
+define i8 @call_no_range_attr(i8 %v) {
+; CHECK-LABEL: @call_no_range_attr
+; CHECK-NEXT: %out = call i8 @dummy2(i8 %v)
+; CHECK-NEXT: ret i8 %out
+  %out = call i8 @dummy2(i8 %v)
+  ret i8 %out
+}
+
+define i8 @call_different_range_attr(i8 range(i8 5, 7) %v) {
+; CHECK-LABEL: @call_different_range_attr
+; CHECK-NEXT: %out = call i8 @dummy2(i8 %v)
+; CHECK-NEXT: ret i8 %out
+  %out = call i8 @dummy2(i8 %v)
+  ret i8 %out
+}
+
+define i8 @call_with_range() {
+  %out = call range(i8 0, 2) i8 @dummy()
+  ret i8 %out
+}
+
+define i8 @call_no_range() {
+; CHECK-LABEL: @call_no_range
+; CHECK-NEXT: %out = call i8 @dummy()
+; CHECK-NEXT: ret i8 %out
+  %out = call i8 @dummy()
+  ret i8 %out
+}
+
+define i8 @call_different_range() {
+; CHECK-LABEL: @call_different_range
+; CHECK-NEXT: %out = call range(i8 5, 7) i8 @dummy()
+; CHECK-NEXT: ret i8 %out
+  %out = call range(i8 5, 7) i8 @dummy()
+  ret i8 %out
+}
+
+define i8 @invoke_with_range() personality ptr undef {
+  %out = invoke range(i8 0, 2) i8 @dummy() to label %next unwind label %lpad
+
+next:
+  ret i8 %out
+
+lpad:
+  %pad = landingpad { ptr, i32 } cleanup
+  resume { ptr, i32 } zeroinitializer
+}
+
+define i8 @invoke_no_range() personality ptr undef {
+; CHECK-LABEL: @invoke_no_range()
+; CHECK-NEXT: invoke i8 @dummy
+  %out = invoke i8 @dummy() to label %next unwind label %lpad
+
+next:
+  ret i8 %out
+
+lpad:
+  %pad = landingpad { ptr, i32 } cleanup
+  resume { ptr, i32 } zeroinitializer
+}
+
+define i8 @invoke_different_range() personality ptr undef {
+; CHECK-LABEL: @invoke_different_range()
+; CHECK-NEXT: invoke range(i8 5, 7) i8 @dummy
+  %out = invoke range(i8 5, 7) i8 @dummy() to label %next unwind label %lpad
+
+next:
+  ret i8 %out
+
+lpad:
+  %pad = landingpad { ptr, i32 } cleanup
+  resume { ptr, i32 } zeroinitializer
+}
+
+define i8 @invoke_with_same_range() personality ptr undef {
+; CHECK-LABEL: @invoke_with_same_range()
+; CHECK: tail call i8 @invoke_with_range()
+  %out = invoke range(i8 0, 2) i8 @dummy() to label %next unwind label %lpad
+
+next:
+  ret i8 %out
+
+lpad:
+  %pad = landingpad { ptr, i32 } cleanup
+  resume { ptr, i32 } zeroinitializer
+}
+
+define i8 @call_with_same_range() {
+; CHECK-LABEL: @call_with_same_range
+; CHECK: tail call i8 @call_with_range
+  %out = call range(i8 0, 2) i8 @dummy()
+  ret i8 %out
+}
+
+define i8 @call_with_same_range_attr(i8 range(i8 0, 2) %v) {
+; CHECK-LABEL: @call_with_same_range_attr
+; CHECK: tail call i8 @call_with_range_attr
+  %out = call i8 @dummy2(i8 %v)
+  ret i8 %out
+}
+
+declare i8 @dummy();
+declare i8 @dummy2(i8);
+declare i32 @__gxx_personality_v0(...)

From 1a895bd95ea741dd4756121244f7c2b59e693f1c Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 13:45:59 +0200
Subject: [PATCH 339/886] [libc++] Marks a variable const. (#88562)

This removes a TODO from the code base.
---
 libcxx/include/__algorithm/sort.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/libcxx/include/__algorithm/sort.h b/libcxx/include/__algorithm/sort.h
index 8a5e0211cdf4c..07b5814639e9e 100644
--- a/libcxx/include/__algorithm/sort.h
+++ b/libcxx/include/__algorithm/sort.h
@@ -696,9 +696,8 @@ __partition_with_equals_on_left(_RandomAccessIterator __first, _RandomAccessIter
   using _Ops = _IterOps<_AlgPolicy>;
   typedef typename iterator_traits<_RandomAccessIterator>::difference_type difference_type;
   typedef typename std::iterator_traits<_RandomAccessIterator>::value_type value_type;
-  // TODO(LLVM18): Make __begin const, see https://reviews.llvm.org/D147089#4349748
-  _RandomAccessIterator __begin     = __first; // used for bounds checking, those are not moved around
-  const _RandomAccessIterator __end = __last;
+  const _RandomAccessIterator __begin = __first; // used for bounds checking, those are not moved around
+  const _RandomAccessIterator __end   = __last;
   (void)__end; //
   value_type __pivot(_Ops::__iter_move(__first));
   if (__comp(__pivot, *(__last - difference_type(1)))) {

From 910ec6ff6dd9ed031e31800c70740fdd17cc1c2a Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 13:46:34 +0200
Subject: [PATCH 340/886] [libc++] Undeprecate POSIX STREAM macros. (#88296)

LWG3869 Deprecate std::errc constants related to UNIX STREAMS

deprecates the POSIX macros ENODATA, ENOSR, ENOSTR, and ETIME. These
were deprecated in libc++ in
https://github.com/llvm/llvm-project/pull/80542. Based on the post
commit feedback the macro are no longer deprecated. Instead libc++
leaves the deprecation to the provider of errno.h.

---------

Co-authored-by: Hristo Hristov <zingam@outlook.com>
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
---
 libcxx/include/__system_error/errc.h          |  5 +++
 libcxx/include/cerrno                         | 18 +++------
 .../test/std/depr.cerro/cerrno.syn.verify.cpp | 37 -------------------
 3 files changed, 11 insertions(+), 49 deletions(-)
 delete mode 100644 libcxx/test/std/depr.cerro/cerrno.syn.verify.cpp

diff --git a/libcxx/include/__system_error/errc.h b/libcxx/include/__system_error/errc.h
index e9f3656b7b9c2..0004c46e2279b 100644
--- a/libcxx/include/__system_error/errc.h
+++ b/libcxx/include/__system_error/errc.h
@@ -112,6 +112,11 @@ enum class errc
 // macros. So GCC does not need the pushing and popping.
 //
 // TODO Remove this when the deprecated constants are removed.
+//
+// Note based on the post-review comments in
+// https://github.com/llvm/llvm-project/pull/80542 libc++ no longer deprecates
+// the macros. Since C libraries may start to deprecate these POSIX macros the
+// deprecation warning avoidance is kept.
 #if defined(_LIBCPP_COMPILER_CLANG_BASED)
 #  define _LIBCPP_SUPPRESS_DEPRECATED_ERRC_PUSH _LIBCPP_SUPPRESS_DEPRECATED_PUSH
 #  define _LIBCPP_SUPPRESS_DEPRECATED_ERRC_POP _LIBCPP_SUPPRESS_DEPRECATED_POP
diff --git a/libcxx/include/cerrno b/libcxx/include/cerrno
index 6171ae31f1847..f1295680fed6c 100644
--- a/libcxx/include/cerrno
+++ b/libcxx/include/cerrno
@@ -38,17 +38,11 @@ Macros:
 #  pragma GCC system_header
 #endif
 
-#ifdef ENODATA
-#  pragma clang deprecated(ENODATA, "ENODATA is deprecated in ISO C++")
-#endif
-#ifdef ENOSR
-#  pragma clang deprecated(ENOSR, "ENOSR is deprecated in ISO C++")
-#endif
-#ifdef ENOSTR
-#  pragma clang deprecated(ENOSTR, "ENOSTR is deprecated in ISO C++")
-#endif
-#ifdef ETIME
-#  pragma clang deprecated(ETIME, "ETIME is deprecated in ISO C++")
-#endif
+// LWG3869 Deprecate std::errc constants related to UNIX STREAMS
+//
+// This LWG issue deprecates the POSIX macros ENODATA, ENOSR, ENOSTR, and ETIME. These were
+// deprecated in libc++ in https://github.com/llvm/llvm-project/pull/80542.
+// Based on the post commit feedback the macro are no longer deprecated.
+// Instead libc++ leaves the deprecation to the provider of errno.h.
 
 #endif // _LIBCPP_CERRNO
diff --git a/libcxx/test/std/depr.cerro/cerrno.syn.verify.cpp b/libcxx/test/std/depr.cerro/cerrno.syn.verify.cpp
deleted file mode 100644
index 3a38605570daf..0000000000000
--- a/libcxx/test/std/depr.cerro/cerrno.syn.verify.cpp
+++ /dev/null
@@ -1,37 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// UNSUPPORTED: clang-modules-build
-// UNSUPPORTED: apple-clang && c++03
-
-// <cerrno>
-
-// tests LWG 3869 deprecated macros.
-//
-// Note the macros may not be defined. When they are not defined the
-// ifdef XXX does not trigger a deprecated message. So use them in the
-// ifdef and test for 2 deprecated messages.
-
-#include <cerrno>
-
-#ifdef ENODATA
-[[maybe_unused]] int nodata =
-    ENODATA; // expected-warning@cerrno.syn.verify.cpp:* 2 {{macro 'ENODATA' has been marked as deprecated}}
-#endif
-#ifdef ENOSR
-[[maybe_unused]] int nosr =
-    ENOSR; // expected-warning@cerrno.syn.verify.cpp:* 2 {{macro 'ENOSR' has been marked as deprecated}}
-#endif
-#ifdef ENOSTR
-[[maybe_unused]] int nostr =
-    ENOSTR; // expected-warning@cerrno.syn.verify.cpp:* 2 {{macro 'ENOSTR' has been marked as deprecated}}
-#endif
-#ifdef ETIME
-[[maybe_unused]] int timeout =
-    ETIME; // expected-warning@cerrno.syn.verify.cpp:* 2 {{macro 'ETIME' has been marked as deprecated}}
-#endif

From cbfcfdf75e9939bc47ac7a7c11d2122a6ad426ed Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sat, 13 Apr 2024 15:16:35 +0300
Subject: [PATCH 341/886] [clang][NFC] Add a test for CWG2254 to
 `is_pointer_interconvertible_base_of` tests

Resolution of that issue makes _any_ base class subobject interconvertible with the containing object, not just the first one.
---
 clang/test/SemaCXX/type-traits.cpp | 1 +
 1 file changed, 1 insertion(+)

diff --git a/clang/test/SemaCXX/type-traits.cpp b/clang/test/SemaCXX/type-traits.cpp
index d43701c3d976e..dee4a29bd2bff 100644
--- a/clang/test/SemaCXX/type-traits.cpp
+++ b/clang/test/SemaCXX/type-traits.cpp
@@ -1864,6 +1864,7 @@ void is_pointer_interconvertible_base_of(int n)
   static_assert(!__is_pointer_interconvertible_base_of(Base2, Derived));
   static_assert(__is_pointer_interconvertible_base_of(Base, DerivedIndirect));
   static_assert(__is_pointer_interconvertible_base_of(Base, DerivedMultiple));
+  static_assert(__is_pointer_interconvertible_base_of(Base2, DerivedMultiple));
   static_assert(!__is_pointer_interconvertible_base_of(Base3, DerivedMultiple));
   static_assert(!__is_pointer_interconvertible_base_of(Base, DerivedAmbiguous));
   static_assert(__is_pointer_interconvertible_base_of(Base, DerivedPrivate));

From 26852565a5f609e6b466f43c2f690ce3047d04c7 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 16:08:07 +0200
Subject: [PATCH 342/886] [libc++][TZDZ] Validates the database version.

It seems one of the tests requires a very recent timezone database. Skip
the test when the database is older.
---
 .../time.zone.members/get_info.sys_time.pass.cpp              | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
index 2ad408968589e..b710f3df23352 100644
--- a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
@@ -175,6 +175,10 @@ static void test_indian_kerguelen() {
   // Z Indian/Kerguelen 0 - -00 1950
   // 5 - +05
 
+  //  Make sure the entry is present in the database.
+  if (std::chrono::get_tzdb().version < "2024a")
+    return;
+
   const std::chrono::time_zone* tz = std::chrono::locate_zone("Indian/Kerguelen");
 
   std::chrono::sys_seconds transition =

From 61f1f1300254e91334404f0d22e75e96f0d98a85 Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Sat, 13 Apr 2024 18:24:12 +0200
Subject: [PATCH 343/886] [libc++][NFC] Move basic ASan annotation functions
 into a utility header (#87220)

---
 libcxx/include/CMakeLists.txt              |   1 +
 libcxx/include/__config                    |   9 --
 libcxx/include/__debug_utils/sanitizers.h  | 104 +++++++++++++++++++++
 libcxx/include/__memory/allocator_traits.h |  12 ---
 libcxx/include/deque                       |  39 +-------
 libcxx/include/module.modulemap            |   1 +
 libcxx/include/string                      |   6 +-
 libcxx/include/vector                      |  11 +--
 8 files changed, 115 insertions(+), 68 deletions(-)
 create mode 100644 libcxx/include/__debug_utils/sanitizers.h

diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index d4e8c196a9a88..1b6c6de082da9 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -345,6 +345,7 @@ set(files
   __coroutine/noop_coroutine_handle.h
   __coroutine/trivial_awaitables.h
   __debug_utils/randomize_range.h
+  __debug_utils/sanitizers.h
   __debug_utils/strict_weak_ordering_check.h
   __exception/exception.h
   __exception/exception_ptr.h
diff --git a/libcxx/include/__config b/libcxx/include/__config
index d98b54926bbe8..82782b31c557b 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -1061,15 +1061,6 @@ typedef __char32_t char32_t;
 #    define _LIBCPP_CONSTEXPR_SINCE_CXX23
 #  endif
 
-#  ifndef _LIBCPP_HAS_NO_ASAN
-extern "C" _LIBCPP_EXPORTED_FROM_ABI void
-__sanitizer_annotate_contiguous_container(const void*, const void*, const void*, const void*);
-extern "C" _LIBCPP_EXPORTED_FROM_ABI void __sanitizer_annotate_double_ended_contiguous_container(
-    const void*, const void*, const void*, const void*, const void*, const void*);
-extern "C" _LIBCPP_EXPORTED_FROM_ABI int
-__sanitizer_verify_double_ended_contiguous_container(const void*, const void*, const void*, const void*);
-#  endif
-
 // Try to find out if RTTI is disabled.
 #  if !defined(__cpp_rtti) || __cpp_rtti < 199711L
 #    define _LIBCPP_HAS_NO_RTTI
diff --git a/libcxx/include/__debug_utils/sanitizers.h b/libcxx/include/__debug_utils/sanitizers.h
new file mode 100644
index 0000000000000..d8547e3249330
--- /dev/null
+++ b/libcxx/include/__debug_utils/sanitizers.h
@@ -0,0 +1,104 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___LIBCXX_DEBUG_UTILS_SANITIZERS_H
+#define _LIBCPP___LIBCXX_DEBUG_UTILS_SANITIZERS_H
+
+#include <__config>
+#include <__type_traits/integral_constant.h>
+#include <__type_traits/is_constant_evaluated.h>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+#ifndef _LIBCPP_HAS_NO_ASAN
+
+extern "C" {
+_LIBCPP_EXPORTED_FROM_ABI void
+__sanitizer_annotate_contiguous_container(const void*, const void*, const void*, const void*);
+_LIBCPP_EXPORTED_FROM_ABI void __sanitizer_annotate_double_ended_contiguous_container(
+    const void*, const void*, const void*, const void*, const void*, const void*);
+_LIBCPP_EXPORTED_FROM_ABI int
+__sanitizer_verify_double_ended_contiguous_container(const void*, const void*, const void*, const void*);
+}
+
+#endif // _LIBCPP_HAS_NO_ASAN
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+
+// ASan choices
+#ifndef _LIBCPP_HAS_NO_ASAN
+#  define _LIBCPP_HAS_ASAN_CONTAINER_ANNOTATIONS_FOR_ALL_ALLOCATORS 1
+#endif
+
+#ifdef _LIBCPP_HAS_ASAN_CONTAINER_ANNOTATIONS_FOR_ALL_ALLOCATORS
+// __asan_annotate_container_with_allocator determines whether containers with custom allocators are annotated. This is
+// a public customization point to disable annotations if the custom allocator assumes that the memory isn't poisoned.
+// See the https://libcxx.llvm.org/UsingLibcxx.html#turning-off-asan-annotation-in-containers for more information.
+template <class _Alloc>
+struct __asan_annotate_container_with_allocator : true_type {};
+#endif
+
+// Annotate a double-ended contiguous range.
+// - [__first_storage, __last_storage) is the allocated memory region,
+// - [__first_old_contained, __last_old_contained) is the previously allowed (unpoisoned) range, and
+// - [__first_new_contained, __last_new_contained) is the new allowed (unpoisoned) range.
+template <class _Allocator>
+_LIBCPP_HIDE_FROM_ABI void __annotate_double_ended_contiguous_container(
+    const void* __first_storage,
+    const void* __last_storage,
+    const void* __first_old_contained,
+    const void* __last_old_contained,
+    const void* __first_new_contained,
+    const void* __last_new_contained) {
+#ifdef _LIBCPP_HAS_NO_ASAN
+  (void)__first_storage;
+  (void)__last_storage;
+  (void)__first_old_contained;
+  (void)__last_old_contained;
+  (void)__first_new_contained;
+  (void)__last_new_contained;
+#else
+  if (__asan_annotate_container_with_allocator<_Allocator>::value && __first_storage != nullptr)
+    __sanitizer_annotate_double_ended_contiguous_container(
+        __first_storage,
+        __last_storage,
+        __first_old_contained,
+        __last_old_contained,
+        __first_new_contained,
+        __last_new_contained);
+#endif
+}
+
+// Annotate a contiguous range.
+// [__first_storage, __last_storage) is the allocated memory region,
+// __old_last_contained is the previously last allowed (unpoisoned) element, and
+// __new_last_contained is the new last allowed (unpoisoned) element.
+template <class _Allocator>
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 void __annotate_contiguous_container(
+    const void* __first_storage,
+    const void* __last_storage,
+    const void* __old_last_contained,
+    const void* __new_last_contained) {
+#ifdef _LIBCPP_HAS_NO_ASAN
+  (void)__first_storage;
+  (void)__last_storage;
+  (void)__old_last_contained;
+  (void)__new_last_contained;
+#else
+  if (!__libcpp_is_constant_evaluated() && __asan_annotate_container_with_allocator<_Allocator>::value &&
+      __first_storage != nullptr)
+    __sanitizer_annotate_contiguous_container(
+        __first_storage, __last_storage, __old_last_contained, __new_last_contained);
+#endif
+}
+
+_LIBCPP_END_NAMESPACE_STD
+
+#endif // _LIBCPP___LIBCXX_DEBUG_UTILS_SANITIZERS_H
diff --git a/libcxx/include/__memory/allocator_traits.h b/libcxx/include/__memory/allocator_traits.h
index efdaa5e4ed8a5..7b3deb0f58e95 100644
--- a/libcxx/include/__memory/allocator_traits.h
+++ b/libcxx/include/__memory/allocator_traits.h
@@ -407,18 +407,6 @@ struct __is_cpp17_copy_insertable<
                    __has_construct<_Alloc, typename _Alloc::value_type*, const typename _Alloc::value_type&>::value > >
     : __is_cpp17_move_insertable<_Alloc> {};
 
-// ASan choices
-#ifndef _LIBCPP_HAS_NO_ASAN
-#  define _LIBCPP_HAS_ASAN_CONTAINER_ANNOTATIONS_FOR_ALL_ALLOCATORS 1
-#endif
-
-#ifdef _LIBCPP_HAS_ASAN_CONTAINER_ANNOTATIONS_FOR_ALL_ALLOCATORS
-template <class _Alloc>
-struct __asan_annotate_container_with_allocator : true_type {};
-template <class _Tp>
-struct __asan_annotate_container_with_allocator<allocator<_Tp> > : true_type {};
-#endif
-
 #undef _LIBCPP_ALLOCATOR_TRAITS_HAS_XXX
 
 _LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/include/deque b/libcxx/include/deque
index a6472e46d426c..d42669dd6dc0e 100644
--- a/libcxx/include/deque
+++ b/libcxx/include/deque
@@ -191,6 +191,7 @@ template <class T, class Allocator, class Predicate>
 #include <__assert>
 #include <__availability>
 #include <__config>
+#include <__debug_utils/sanitizers.h>
 #include <__format/enable_insertable.h>
 #include <__fwd/deque.h>
 #include <__iterator/distance.h>
@@ -864,33 +865,6 @@ private:
     __asan_back_moved,
   };
 
-  // The following functions are no-ops outside of AddressSanitizer mode.
-  // We call annotations for every allocator, unless explicitly disabled.
-  //
-  // To disable annotations for a particular allocator, change value of
-  // __asan_annotate_container_with_allocator to false.
-  // For more details, see the "Using libc++" documentation page or
-  // the documentation for __sanitizer_annotate_contiguous_container.
-  _LIBCPP_HIDE_FROM_ABI void __annotate_double_ended_contiguous_container(
-      const void* __beg,
-      const void* __end,
-      const void* __old_con_beg,
-      const void* __old_con_end,
-      const void* __new_con_beg,
-      const void* __new_con_end) const {
-    (void)__beg;
-    (void)__end;
-    (void)__old_con_beg;
-    (void)__old_con_end;
-    (void)__new_con_beg;
-    (void)__new_con_end;
-#ifndef _LIBCPP_HAS_NO_ASAN
-    if (__beg != nullptr && __asan_annotate_container_with_allocator<_Allocator>::value)
-      __sanitizer_annotate_double_ended_contiguous_container(
-          __beg, __end, __old_con_beg, __old_con_end, __new_con_beg, __new_con_end);
-#endif
-  }
-
   _LIBCPP_HIDE_FROM_ABI void __annotate_from_to(
       size_type __beg,
       size_type __end,
@@ -990,7 +964,8 @@ private:
       const void* __new_beg = __front ? __new_edge : __old_edge;
       const void* __new_end = __front ? __old_edge : __new_edge;
 
-      __annotate_double_ended_contiguous_container(__mem_beg, __mem_end, __old_beg, __old_end, __new_beg, __new_end);
+      std::__annotate_double_ended_contiguous_container<_Allocator>(
+          __mem_beg, __mem_end, __old_beg, __old_end, __new_beg, __new_end);
     }
 #endif // !_LIBCPP_HAS_NO_ASAN
   }
@@ -1051,11 +1026,7 @@ private:
   }
 
   _LIBCPP_HIDE_FROM_ABI void __annotate_poison_block(const void* __beginning, const void* __end) const _NOEXCEPT {
-    (void)__beginning;
-    (void)__end;
-#ifndef _LIBCPP_HAS_NO_ASAN
-    __annotate_double_ended_contiguous_container(__beginning, __end, __beginning, __end, __end, __end);
-#endif
+    std::__annotate_double_ended_contiguous_container<_Allocator>(__beginning, __end, __beginning, __end, __end, __end);
   }
 
   _LIBCPP_HIDE_FROM_ABI void
@@ -1070,7 +1041,7 @@ private:
     if (__annotation_type == __asan_poison)
       __annotate_poison_block(__block_start, __block_end);
     else {
-      __annotate_double_ended_contiguous_container(
+      std::__annotate_double_ended_contiguous_container<_Allocator>(
           __block_start, __block_end, __block_start, __block_start, __block_start, __block_end);
     }
 #endif
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index 9c0f0ddd20c96..372219e8f8363 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -1254,6 +1254,7 @@ module std_private_coroutine_noop_coroutine_handle [system] { header "__coroutin
 module std_private_coroutine_trivial_awaitables    [system] { header "__coroutine/trivial_awaitables.h" }
 
 module std_private_debug_utils_randomize_range            [system] { header "__debug_utils/randomize_range.h" }
+module std_private_debug_utils_sanitizers                 [system] { header "__debug_utils/sanitizers.h" }
 module std_private_debug_utils_strict_weak_ordering_check [system] {
   header "__debug_utils/strict_weak_ordering_check.h"
   export std_private_type_traits_is_constant_evaluated
diff --git a/libcxx/include/string b/libcxx/include/string
index a456f8cb80ee3..4ea6df00bb7e3 100644
--- a/libcxx/include/string
+++ b/libcxx/include/string
@@ -574,6 +574,7 @@ basic_string<char32_t> operator""s( const char32_t *str, size_t len );
 #include <__algorithm/remove_if.h>
 #include <__assert>
 #include <__config>
+#include <__debug_utils/sanitizers.h>
 #include <__format/enable_insertable.h>
 #include <__functional/hash.h>
 #include <__functional/unary_function.h>
@@ -1909,10 +1910,7 @@ private:
     (void)__old_mid;
     (void)__new_mid;
 #if !defined(_LIBCPP_HAS_NO_ASAN) && defined(_LIBCPP_INSTRUMENTED_WITH_ASAN)
-    const void* __begin = data();
-    const void* __end   = data() + capacity() + 1;
-    if (__asan_annotate_container_with_allocator<allocator_type>::value && !__libcpp_is_constant_evaluated())
-      __sanitizer_annotate_contiguous_container(__begin, __end, __old_mid, __new_mid);
+    std::__annotate_contiguous_container<_Allocator>(data(), data() + capacity() + 1, __old_mid, __new_mid);
 #endif
   }
 
diff --git a/libcxx/include/vector b/libcxx/include/vector
index 1defc43a52478..7303d7ec0b24a 100644
--- a/libcxx/include/vector
+++ b/libcxx/include/vector
@@ -320,6 +320,7 @@ template<class T, class charT> requires is-vector-bool-reference<T> // Since C++
 #include <__bit_reference>
 #include <__concepts/same_as.h>
 #include <__config>
+#include <__debug_utils/sanitizers.h>
 #include <__format/enable_insertable.h>
 #include <__format/formatter.h>
 #include <__format/formatter_bool.h>
@@ -834,15 +835,7 @@ private:
 
   _LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI void
   __annotate_contiguous_container(const void* __old_mid, const void* __new_mid) const {
-    (void)__old_mid;
-    (void)__new_mid;
-#ifndef _LIBCPP_HAS_NO_ASAN
-    const void* __beg = data();
-    const void* __end = data() + capacity();
-    if (!__libcpp_is_constant_evaluated() && __beg != nullptr &&
-        __asan_annotate_container_with_allocator<_Allocator>::value)
-      __sanitizer_annotate_contiguous_container(__beg, __end, __old_mid, __new_mid);
-#endif
+    std::__annotate_contiguous_container<_Allocator>(data(), data() + capacity(), __old_mid, __new_mid);
   }
 
   _LIBCPP_CONSTEXPR_SINCE_CXX20 _LIBCPP_HIDE_FROM_ABI void __annotate_new(size_type __current_size) const _NOEXCEPT {

From ecf44b4b719c33465d9ebfbc5910b1f761856e0b Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Sat, 13 Apr 2024 20:35:59 +0200
Subject: [PATCH 344/886] [libc++][TZDB] Removes test_indian_kerguelen test.

The work-around in 26852565a5f609e6b466f43c2f690ce3047d04c7 didn't fix
the CI. Since the entire local database is compared with the libc++
implementation in a separate this this change does not remove coverage.
---
 .../get_info.sys_time.pass.cpp                | 44 -------------------
 1 file changed, 44 deletions(-)

diff --git a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
index b710f3df23352..a751a2fb6347b 100644
--- a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
@@ -169,49 +169,6 @@ static void test_durations() {
   assert_equal(expected, tz->get_info(std::chrono::sys_time<std::chrono::years>{}));
 }
 
-static void test_indian_kerguelen() {
-  // One change, no rules, no dst changes.
-
-  // Z Indian/Kerguelen 0 - -00 1950
-  // 5 - +05
-
-  //  Make sure the entry is present in the database.
-  if (std::chrono::get_tzdb().version < "2024a")
-    return;
-
-  const std::chrono::time_zone* tz = std::chrono::locate_zone("Indian/Kerguelen");
-
-  std::chrono::sys_seconds transition =
-      to_sys_seconds(std::chrono::year(1950), std::chrono::January, std::chrono::day(1));
-
-  assert_equal(
-      std::chrono::sys_info(
-          std::chrono::sys_seconds::min(), //
-          transition,                      //
-          std::chrono::seconds(0),         //
-          std::chrono::minutes(0),         //
-          "-00"),                          //
-      tz->get_info(std::chrono::sys_seconds::min()));
-
-  assert_equal(
-      std::chrono::sys_info(
-          std::chrono::sys_seconds::min(), //
-          transition,                      //
-          std::chrono::seconds(0),         //
-          std::chrono::minutes(0),         //
-          "-00"),                          //
-      tz->get_info(transition - std::chrono::seconds{1}));
-
-  assert_equal(
-      std::chrono::sys_info(
-          transition,                      //
-          std::chrono::sys_seconds::max(), //
-          std::chrono::hours(5),           //
-          std::chrono::minutes(0),         //
-          "+05"),                          //
-      tz->get_info(transition));
-}
-
 static void test_antarctica_syowa() {
   // One change, no rules, no dst changes
   // This change uses an ON field with a day number
@@ -1343,7 +1300,6 @@ int main(int, const char**) {
   // Basic tests
   test_gmt();
   test_durations();
-  test_indian_kerguelen();
   test_antarctica_syowa();
   test_asia_hong_kong();
   test_europe_berlin();

From fad37526a3ea7d669af621342968029085862281 Mon Sep 17 00:00:00 2001
From: Edwin Vane <revane@google.com>
Date: Sat, 13 Apr 2024 15:01:06 -0400
Subject: [PATCH 345/886] [clang-tidy] Export fixes from check_clang_tidy.py
 (#88186)

Makes it possible to export fixes from running llvm-lit on a clang-tidy
test. To enable, modify the RUN invocation directly in the test with the
new -export flag. llvm-lit will report the test passed and fixes can be
found in the file specified to the -export flag.
---
 clang-tools-extra/docs/ReleaseNotes.rst       |  2 +
 .../test/clang-tidy/check_clang_tidy.py       | 74 ++++++++++++++-----
 2 files changed, 57 insertions(+), 19 deletions(-)

diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index b66be44e9f8a6..1405fb0df1f8d 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -100,6 +100,8 @@ Improvements to clang-tidy
 - Improved :program:`run-clang-tidy.py` script. Added argument `-source-filter`
   to filter source files from the compilation database, via a RegEx. In a
   similar fashion to what `-header-filter` does for header files.
+- Improved :program:`check_clang_tidy.py` script. Added argument `-export-fixes`
+  to aid in clang-tidy and test development.
 
 New checks
 ^^^^^^^^^^
diff --git a/clang-tools-extra/test/clang-tidy/check_clang_tidy.py b/clang-tools-extra/test/clang-tidy/check_clang_tidy.py
index 53ffca0bad8d0..6d4b466afa691 100755
--- a/clang-tools-extra/test/clang-tidy/check_clang_tidy.py
+++ b/clang-tools-extra/test/clang-tidy/check_clang_tidy.py
@@ -8,25 +8,35 @@
 #
 # ===------------------------------------------------------------------------===#
 
-r"""
+"""
 ClangTidy Test Helper
 =====================
 
-This script runs clang-tidy in fix mode and verify fixes, messages or both.
+This script is used to simplify writing, running, and debugging tests compatible
+with llvm-lit. By default it runs clang-tidy in fix mode and uses FileCheck to
+verify messages and/or fixes.
+
+For debugging, with --export-fixes, the tool simply exports fixes to a provided
+file and does not run FileCheck.
 
-Usage:
-  check_clang_tidy.py [-resource-dir=<resource-dir>] \
-    [-assume-filename=<file-with-source-extension>] \
-    [-check-suffix=<comma-separated-file-check-suffixes>] \
-    [-check-suffixes=<comma-separated-file-check-suffixes>] \
-    [-std=c++(98|11|14|17|20)[-or-later]] \
-    <source-file> <check-name> <temp-file> \
-    -- [optional clang-tidy arguments]
+Extra arguments, those after the first -- if any, are passed to either
+clang-tidy or clang:
+* Arguments between the first -- and second -- are clang-tidy arguments.
+  * May be only whitespace if there are no clang-tidy arguments.
+  * clang-tidy's --config would go here.
+* Arguments after the second -- are clang arguments
+
+Examples
+--------
 
-Example:
   // RUN: %check_clang_tidy %s llvm-include-order %t -- -- -isystem %S/Inputs
 
-Notes:
+or
+
+  // RUN: %check_clang_tidy %s llvm-include-order --export-fixes=fixes.yaml %t -std=c++20
+
+Notes
+-----
   -std=c++(98|11|14|17|20)-or-later:
     This flag will cause multiple runs within the same check_clang_tidy
     execution. Make sure you don't have shared state across these runs.
@@ -34,6 +44,7 @@
 
 import argparse
 import os
+import pathlib
 import re
 import subprocess
 import sys
@@ -88,6 +99,7 @@ def __init__(self, args, extra_args):
         self.has_check_fixes = False
         self.has_check_messages = False
         self.has_check_notes = False
+        self.export_fixes = args.export_fixes
         self.fixes = MessagePrefix("CHECK-FIXES")
         self.messages = MessagePrefix("CHECK-MESSAGES")
         self.notes = MessagePrefix("CHECK-NOTES")
@@ -181,7 +193,13 @@ def run_clang_tidy(self):
             [
                 "clang-tidy",
                 self.temp_file_name,
-                "-fix",
+            ]
+            + [
+                "-fix"
+                if self.export_fixes is None
+                else "--export-fixes=" + self.export_fixes
+            ]
+            + [
                 "--checks=-*," + self.check_name,
             ]
             + self.clang_tidy_extra_args
@@ -255,12 +273,14 @@ def check_notes(self, clang_tidy_output):
 
     def run(self):
         self.read_input()
-        self.get_prefixes()
+        if self.export_fixes is None:
+            self.get_prefixes()
         self.prepare_test_inputs()
         clang_tidy_output = self.run_clang_tidy()
-        self.check_fixes()
-        self.check_messages(clang_tidy_output)
-        self.check_notes(clang_tidy_output)
+        if self.export_fixes is None:
+            self.check_fixes()
+            self.check_messages(clang_tidy_output)
+            self.check_notes(clang_tidy_output)
 
 
 def expand_std(std):
@@ -284,7 +304,11 @@ def csv(string):
 
 
 def parse_arguments():
-    parser = argparse.ArgumentParser()
+    parser = argparse.ArgumentParser(
+        prog=pathlib.Path(__file__).stem,
+        description=__doc__,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
     parser.add_argument("-expect-clang-tidy-error", action="store_true")
     parser.add_argument("-resource-dir")
     parser.add_argument("-assume-filename")
@@ -298,7 +322,19 @@ def parse_arguments():
         type=csv,
         help="comma-separated list of FileCheck suffixes",
     )
-    parser.add_argument("-std", type=csv, default=["c++11-or-later"])
+    parser.add_argument(
+        "-export-fixes",
+        default=None,
+        type=str,
+        metavar="file",
+        help="A file to export fixes into instead of fixing.",
+    )
+    parser.add_argument(
+        "-std",
+        type=csv,
+        default=["c++11-or-later"],
+        help="Passed to clang. Special -or-later values are expanded.",
+    )
     return parser.parse_known_args()
 
 

From ef9446bd2d362ec90cd681ae59463d16bf671fe8 Mon Sep 17 00:00:00 2001
From: Enna1 <xumingjie.enna1@bytedance.com>
Date: Sun, 14 Apr 2024 11:01:02 +0800
Subject: [PATCH 346/886] [GVN] lazily update dominator tree when merge
 unconditional branches in function (#88477)

With this change, the wall time for of GVN pass decreased from
873,745.492 ms to 367,375.304 ms in an our internal testcase.
---
 llvm/lib/Transforms/Scalar/GVN.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/Scalar/GVN.cpp b/llvm/lib/Transforms/Scalar/GVN.cpp
index 67fb2a5da3bb7..86d5c9909f3dc 100644
--- a/llvm/lib/Transforms/Scalar/GVN.cpp
+++ b/llvm/lib/Transforms/Scalar/GVN.cpp
@@ -2708,7 +2708,7 @@ bool GVNPass::runImpl(Function &F, AssumptionCache &RunAC, DominatorTree &RunDT,
   bool Changed = false;
   bool ShouldContinue = true;
 
-  DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Eager);
+  DomTreeUpdater DTU(DT, DomTreeUpdater::UpdateStrategy::Lazy);
   // Merge unconditional branches, allowing PRE to catch more
   // optimization opportunities.
   for (BasicBlock &BB : llvm::make_early_inc_range(F)) {
@@ -2718,6 +2718,7 @@ bool GVNPass::runImpl(Function &F, AssumptionCache &RunAC, DominatorTree &RunDT,
 
     Changed |= removedBlock;
   }
+  DTU.flush();
 
   unsigned Iteration = 0;
   while (ShouldContinue) {

From 342aa3eb23992ad742d22ef0e0d8821a9b29de3e Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Sat, 13 Apr 2024 20:54:09 -0700
Subject: [PATCH 347/886] [LegalizeTypes][RISCV] Support scalable vectors in
 SplitVecRes_BITCAST/SplitVecOp_BITCAST. (#88590)

Instead of bitcasting to a scalar integer, try to split the other vector
type that is legal. We already handle the case where both types need to
be split.

Fixes #88576
---
 .../SelectionDAG/LegalizeVectorTypes.cpp      | 20 ++++++-
 llvm/test/CodeGen/RISCV/rvv/pr88576.ll        | 60 +++++++++++++++++++
 2 files changed, 78 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/rvv/pr88576.ll

diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
index 1f6e0097f31ab..8776d89f4c5bd 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp
@@ -1388,6 +1388,13 @@ void DAGTypeLegalizer::SplitVecRes_BITCAST(SDNode *N, SDValue &Lo,
     report_fatal_error("Scalarization of scalable vectors is not supported.");
   }
 
+  if (LoVT.isScalableVector()) {
+    auto [InLo, InHi] = DAG.SplitVectorOperand(N, 0);
+    Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, InLo);
+    Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, InHi);
+    return;
+  }
+
   // In the general case, convert the input to an integer and split it by hand.
   EVT LoIntVT = EVT::getIntegerVT(*DAG.getContext(), LoVT.getSizeInBits());
   EVT HiIntVT = EVT::getIntegerVT(*DAG.getContext(), HiVT.getSizeInBits());
@@ -3259,16 +3266,25 @@ SDValue DAGTypeLegalizer::SplitVecOp_BITCAST(SDNode *N) {
   // For example, i64 = BITCAST v4i16 on alpha.  Typically the vector will
   // end up being split all the way down to individual components.  Convert the
   // split pieces into integers and reassemble.
+  EVT ResVT = N->getValueType(0);
   SDValue Lo, Hi;
   GetSplitVector(N->getOperand(0), Lo, Hi);
+  SDLoc dl(N);
+
+  if (ResVT.isScalableVector()) {
+    auto [LoVT, HiVT] = DAG.GetSplitDestVTs(ResVT);
+    Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo);
+    Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi);
+    return DAG.getNode(ISD::CONCAT_VECTORS, dl, ResVT, Lo, Hi);
+  }
+
   Lo = BitConvertToInteger(Lo);
   Hi = BitConvertToInteger(Hi);
 
   if (DAG.getDataLayout().isBigEndian())
     std::swap(Lo, Hi);
 
-  return DAG.getNode(ISD::BITCAST, SDLoc(N), N->getValueType(0),
-                     JoinIntegers(Lo, Hi));
+  return DAG.getNode(ISD::BITCAST, dl, ResVT, JoinIntegers(Lo, Hi));
 }
 
 SDValue DAGTypeLegalizer::SplitVecOp_INSERT_SUBVECTOR(SDNode *N,
diff --git a/llvm/test/CodeGen/RISCV/rvv/pr88576.ll b/llvm/test/CodeGen/RISCV/rvv/pr88576.ll
new file mode 100644
index 0000000000000..b6e0d1e2ff4ae
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/pr88576.ll
@@ -0,0 +1,60 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=riscv64 -mattr=+v | FileCheck %s
+
+define i1 @foo(<vscale x 16 x i8> %x, i64 %y) {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    slli a2, a1, 4
+; CHECK-NEXT:    addi a2, a2, -1
+; CHECK-NEXT:    bltu a0, a2, .LBB0_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    mv a0, a2
+; CHECK-NEXT:  .LBB0_2:
+; CHECK-NEXT:    addi sp, sp, -80
+; CHECK-NEXT:    .cfi_def_cfa_offset 80
+; CHECK-NEXT:    sd ra, 72(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    sd s0, 64(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_offset ra, -8
+; CHECK-NEXT:    .cfi_offset s0, -16
+; CHECK-NEXT:    addi s0, sp, 80
+; CHECK-NEXT:    .cfi_def_cfa s0, 0
+; CHECK-NEXT:    csrr a2, vlenb
+; CHECK-NEXT:    slli a2, a2, 4
+; CHECK-NEXT:    sub sp, sp, a2
+; CHECK-NEXT:    andi sp, sp, -64
+; CHECK-NEXT:    addi a2, sp, 64
+; CHECK-NEXT:    add a0, a2, a0
+; CHECK-NEXT:    slli a1, a1, 3
+; CHECK-NEXT:    add a1, a2, a1
+; CHECK-NEXT:    vsetvli a3, zero, e8, m8, ta, ma
+; CHECK-NEXT:    vmv.v.i v16, 0
+; CHECK-NEXT:    vmv1r.v v0, v9
+; CHECK-NEXT:    vmerge.vim v24, v16, 1, v0
+; CHECK-NEXT:    vs8r.v v24, (a1)
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vmerge.vim v8, v16, 1, v0
+; CHECK-NEXT:    vs8r.v v8, (a2)
+; CHECK-NEXT:    lbu a0, 0(a0)
+; CHECK-NEXT:    addi sp, s0, -80
+; CHECK-NEXT:    ld ra, 72(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    ld s0, 64(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    addi sp, sp, 80
+; CHECK-NEXT:    ret
+  %a = bitcast <vscale x 16 x i8> %x to <vscale x 128 x i1>
+  %b = extractelement <vscale x 128 x i1> %a, i64 %y
+  ret i1 %b
+}
+
+define i8 @bar(<vscale x 128 x i1> %x, i64 %y) {
+; CHECK-LABEL: bar:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmv1r.v v1, v8
+; CHECK-NEXT:    vsetivli zero, 1, e8, m2, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v0, a0
+; CHECK-NEXT:    vmv.x.s a0, v8
+; CHECK-NEXT:    ret
+  %a = bitcast <vscale x 128 x i1> %x to <vscale x 16 x i8>
+  %b = extractelement <vscale x 16 x i8> %a, i64 %y
+  ret i8 %b
+}

From 09327efdf0f02c4f865a4536db96cac539bb1c01 Mon Sep 17 00:00:00 2001
From: Chris Copeland <chris@chrisnc.net>
Date: Sat, 13 Apr 2024 23:37:58 -0700
Subject: [PATCH 348/886] [clang][docs] fix whitespace in AttrDocs.td (#88631)

Noticed this failure while working on a different change:
https://buildkite.com/llvm-project/clang-ci/builds/15553#018ed1bb-923a-48b5-9788-e8b6d3827aba
Seems to be introduced by #88448.
---
 clang/include/clang/Basic/AttrDocs.td | 32 +++++++++++++--------------
 1 file changed, 16 insertions(+), 16 deletions(-)

diff --git a/clang/include/clang/Basic/AttrDocs.td b/clang/include/clang/Basic/AttrDocs.td
index 8687c4f57d3f8..a0bbe5861c572 100644
--- a/clang/include/clang/Basic/AttrDocs.td
+++ b/clang/include/clang/Basic/AttrDocs.td
@@ -1604,39 +1604,39 @@ specifies availability for the current target platform, the availability
 attributes are ignored. Supported platforms are:
 
 ``ios``
-  Apple's iOS operating system. The minimum deployment target is specified 
-  as part of the ``-target *arch*-apple-ios*version*`` command line argument. 
-  Alternatively, it can be specified by the ``-mtargetos=ios*version*`` 
+  Apple's iOS operating system. The minimum deployment target is specified
+  as part of the ``-target *arch*-apple-ios*version*`` command line argument.
+  Alternatively, it can be specified by the ``-mtargetos=ios*version*``
   command-line argument.
 
 ``macos``
-  Apple's macOS operating system. The minimum deployment target is specified 
-  as part of the ``-target *arch*-apple-macos*version*`` command line argument. 
-  Alternatively, it can be specified by the ``-mtargetos=macos*version*`` 
-  command-line argument. ``macosx`` is supported for 
+  Apple's macOS operating system. The minimum deployment target is specified
+  as part of the ``-target *arch*-apple-macos*version*`` command line argument.
+  Alternatively, it can be specified by the ``-mtargetos=macos*version*``
+  command-line argument. ``macosx`` is supported for
   backward-compatibility reasons, but it is deprecated.
 
 ``tvos``
-  Apple's tvOS operating system. The minimum deployment target is specified 
-  as part of the ``-target *arch*-apple-tvos*version*`` command line argument. 
-  Alternatively, it can be specified by the ``-mtargetos=tvos*version*`` 
+  Apple's tvOS operating system. The minimum deployment target is specified
+  as part of the ``-target *arch*-apple-tvos*version*`` command line argument.
+  Alternatively, it can be specified by the ``-mtargetos=tvos*version*``
   command-line argument.
 
 ``watchos``
   Apple's watchOS operating system. The minimum deployment target is specified
-  as part of the ``-target *arch*-apple-watchos*version*`` command line argument. 
-  Alternatively, it can be specified by the ``-mtargetos=watchos*version*`` 
+  as part of the ``-target *arch*-apple-watchos*version*`` command line argument.
+  Alternatively, it can be specified by the ``-mtargetos=watchos*version*``
   command-line argument.
 
 ``visionos``
   Apple's visionOS operating system. The minimum deployment target is specified
-  as part of the ``-target *arch*-apple-visionos*version*`` command line argument. 
-  Alternatively, it can be specified by the ``-mtargetos=visionos*version*`` 
+  as part of the ``-target *arch*-apple-visionos*version*`` command line argument.
+  Alternatively, it can be specified by the ``-mtargetos=visionos*version*``
   command-line argument.
 
 ``driverkit``
   Apple's DriverKit userspace kernel extensions. The minimum deployment target
-  is specified as part of the ``-target *arch*-apple-driverkit*version*`` 
+  is specified as part of the ``-target *arch*-apple-driverkit*version*``
   command line argument.
 
 A declaration can typically be used even when deploying back to a platform
@@ -7522,7 +7522,7 @@ means that it can e.g no longer be part of an initializer expression.
 
       /* This may print something else than "6 * 7 = 42",
          if there is a non-weak definition of "ANSWER" in
-	 an object linked in */
+         an object linked in */
       printf("6 * 7 = %d\n", ANSWER);
 
       return 0;

From c6f9c84e498ee05a812511ae969773ff166fd25e Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Sun, 14 Apr 2024 12:08:30 +0200
Subject: [PATCH 349/886] [Clang] Reduce the size of Decl and classes derived
 from it (#87361)

Class | Old size (in bytes) | New size (in bytes)

----------------------------------|---------------------|--------------------
Decl                              | 40                  | 32
AccessSpecDecl                    | 40                  | 40
BlockDecl                         | 128                 | 120
CapturedDecl                      | 88                  | 80
EmptyDecl                         | 40                  | 32
ExportDecl                        | 80                  | 72
ExternCContextDecl                | 72                  | 64
FileScopeAsmDecl                  | 56                  | 48
FriendDecl                        | 64                  | 56
FriendTemplateDecl                | 64                  | 64
ImplicitConceptSpecializationDecl | 40                  | 40
ImportDecl                        | 56                  | 48
LifetimeExtendedTemporaryDecl     | 72                  | 64
LinkageSpecDecl                   | 80                  | 72
NamedDecl                         | 48                  | 40
ObjCPropertyImplDecl              | 96                  | 88
PragmaCommentDecl                 | 40                  | 40
PragmaDetectMismatchDecl          | 48                  | 40
RequiresExprBodyDecl              | 72                  | 64
StaticAssertDecl                  | 64                  | 56
TopLevelStmtDecl                  | 88                  | 80
TranslationUnitDecl               | 104                 | 96
BaseUsingDecl                     | 56                  | 48
UsingDecl                         | 88                  | 80
UsingEnumDecl                     | 72                  | 64
HLSLBufferDecl                    | 96                  | 88
LabelDecl                         | 80                  | 72
NamespaceAliasDecl                | 96                  | 88
NamespaceDecl                     | 112                 | 104
ObjCCompatibleAliasDecl           | 56                  | 48
ObjCContainerDecl                 | 88                  | 80
ObjCMethodDecl                    | 136                 | 128
ObjCPropertyDecl                  | 128                 | 120
TemplateDecl                      | 64                  | 56
BuiltinTemplateDecl               | 72                  | 64
TypeDecl                          | 64                  | 56
UnresolvedUsingIfExistsDecl       | 48                  | 40
UsingDirectiveDecl                | 88                  | 80
UsingPackDecl                     | 64                  | 56
UsingShadowDecl                   | 80                  | 72
ValueDecl                         | 56                  | 48

When parsing libc++'s `<string>` header the used memory is reduced from
42.8MB to 42.5MB.
---
 clang/include/clang/AST/DeclBase.h        | 72 +++++++++++++++--------
 clang/lib/AST/DeclBase.cpp                | 31 +++++++---
 clang/lib/Serialization/ASTReaderDecl.cpp |  2 +-
 3 files changed, 70 insertions(+), 35 deletions(-)

diff --git a/clang/include/clang/AST/DeclBase.h b/clang/include/clang/AST/DeclBase.h
index 858450926455c..4bee18767dd11 100644
--- a/clang/include/clang/AST/DeclBase.h
+++ b/clang/include/clang/AST/DeclBase.h
@@ -268,17 +268,37 @@ class alignas(8) Decl {
   ///   }
   ///   void A::f(); // SemanticDC == namespace 'A'
   ///                // LexicalDC == global namespace
-  llvm::PointerUnion<DeclContext*, MultipleDC*> DeclCtx;
 
-  bool isInSemaDC() const { return DeclCtx.is<DeclContext*>(); }
-  bool isOutOfSemaDC() const { return DeclCtx.is<MultipleDC*>(); }
+  // Compress the InvalidDecl and HasAttrs bits into DeclCtx to keep Decl below
+  // 32 bytes in size
+  llvm::PointerIntPair<
+      llvm::PointerIntPair<llvm::PointerUnion<DeclContext *, MultipleDC *>, 1,
+                           bool>,
+      1, bool>
+      DeclCtxWithInvalidDeclAndHasAttrs;
+
+  bool isInSemaDC() const {
+    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
+        .getPointer()
+        .is<DeclContext *>();
+  }
+
+  bool isOutOfSemaDC() const {
+    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
+        .getPointer()
+        .is<MultipleDC *>();
+  }
 
   MultipleDC *getMultipleDC() const {
-    return DeclCtx.get<MultipleDC*>();
+    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
+        .getPointer()
+        .get<MultipleDC *>();
   }
 
   DeclContext *getSemanticDC() const {
-    return DeclCtx.get<DeclContext*>();
+    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
+        .getPointer()
+        .get<DeclContext *>();
   }
 
   /// Loc - The location of this decl.
@@ -288,14 +308,6 @@ class alignas(8) Decl {
   LLVM_PREFERRED_TYPE(Kind)
   unsigned DeclKind : 7;
 
-  /// InvalidDecl - This indicates a semantic error occurred.
-  LLVM_PREFERRED_TYPE(bool)
-  unsigned InvalidDecl :  1;
-
-  /// HasAttrs - This indicates whether the decl has attributes or not.
-  LLVM_PREFERRED_TYPE(bool)
-  unsigned HasAttrs : 1;
-
   /// Implicit - Whether this declaration was implicitly generated by
   /// the implementation rather than explicitly written by the user.
   LLVM_PREFERRED_TYPE(bool)
@@ -393,21 +405,22 @@ class alignas(8) Decl {
 protected:
   Decl(Kind DK, DeclContext *DC, SourceLocation L)
       : NextInContextAndBits(nullptr, getModuleOwnershipKindForChildOf(DC)),
-        DeclCtx(DC), Loc(L), DeclKind(DK), InvalidDecl(false), HasAttrs(false),
-        Implicit(false), Used(false), Referenced(false),
+        DeclCtxWithInvalidDeclAndHasAttrs({DC, false}, false), Loc(L),
+        DeclKind(DK), Implicit(false), Used(false), Referenced(false),
         TopLevelDeclInObjCContainer(false), Access(AS_none), FromASTFile(0),
         IdentifierNamespace(getIdentifierNamespaceForKind(DK)),
         CacheValidAndLinkage(llvm::to_underlying(Linkage::Invalid)) {
-    if (StatisticsEnabled) add(DK);
+    if (StatisticsEnabled)
+      add(DK);
   }
 
   Decl(Kind DK, EmptyShell Empty)
-      : DeclKind(DK), InvalidDecl(false), HasAttrs(false), Implicit(false),
-        Used(false), Referenced(false), TopLevelDeclInObjCContainer(false),
-        Access(AS_none), FromASTFile(0),
+      : DeclKind(DK), Implicit(false), Used(false), Referenced(false),
+        TopLevelDeclInObjCContainer(false), Access(AS_none), FromASTFile(0),
         IdentifierNamespace(getIdentifierNamespaceForKind(DK)),
         CacheValidAndLinkage(llvm::to_underlying(Linkage::Invalid)) {
-    if (StatisticsEnabled) add(DK);
+    if (StatisticsEnabled)
+      add(DK);
   }
 
   virtual ~Decl();
@@ -520,7 +533,9 @@ class alignas(8) Decl {
     return AccessSpecifier(Access);
   }
 
-  bool hasAttrs() const { return HasAttrs; }
+  bool hasAttrs() const {
+    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer().getInt();
+  }
 
   void setAttrs(const AttrVec& Attrs) {
     return setAttrsImpl(Attrs, getASTContext());
@@ -549,13 +564,17 @@ class alignas(8) Decl {
   }
 
   template <typename... Ts> void dropAttrs() {
-    if (!HasAttrs) return;
+    if (!hasAttrs())
+      return;
 
     AttrVec &Vec = getAttrs();
     llvm::erase_if(Vec, [](Attr *A) { return isa<Ts...>(A); });
 
-    if (Vec.empty())
-      HasAttrs = false;
+    if (Vec.empty()) {
+      auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+      InnerPtr.setInt(false);
+      DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+    }
   }
 
   template <typename T> void dropAttr() { dropAttrs<T>(); }
@@ -590,7 +609,10 @@ class alignas(8) Decl {
   /// setInvalidDecl - Indicates the Decl had a semantic error. This
   /// allows for graceful error recovery.
   void setInvalidDecl(bool Invalid = true);
-  bool isInvalidDecl() const { return (bool) InvalidDecl; }
+
+  bool isInvalidDecl() const {
+    return DeclCtxWithInvalidDeclAndHasAttrs.getInt();
+  }
 
   /// isImplicit - Indicates whether the declaration was implicitly
   /// generated by the implementation. If false, this declaration
diff --git a/clang/lib/AST/DeclBase.cpp b/clang/lib/AST/DeclBase.cpp
index 66a727d9dd0c3..8c83c711d882d 100644
--- a/clang/lib/AST/DeclBase.cpp
+++ b/clang/lib/AST/DeclBase.cpp
@@ -52,6 +52,8 @@
 
 using namespace clang;
 
+static_assert(sizeof(Decl) <= 32, "Decl grew beyond 32 bytes!");
+
 //===----------------------------------------------------------------------===//
 //  Statistics
 //===----------------------------------------------------------------------===//
@@ -130,7 +132,7 @@ const char *Decl::getDeclKindName() const {
 }
 
 void Decl::setInvalidDecl(bool Invalid) {
-  InvalidDecl = Invalid;
+  DeclCtxWithInvalidDeclAndHasAttrs.setInt(Invalid);
   assert(!isa<TagDecl>(this) || !cast<TagDecl>(this)->isCompleteDefinition());
   if (!Invalid) {
     return;
@@ -334,7 +336,9 @@ void PrettyStackTraceDecl::print(raw_ostream &OS) const {
 Decl::~Decl() = default;
 
 void Decl::setDeclContext(DeclContext *DC) {
-  DeclCtx = DC;
+  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+  InnerPtr.setPointer(DC);
+  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
 }
 
 void Decl::setLexicalDeclContext(DeclContext *DC) {
@@ -364,12 +368,16 @@ void Decl::setLexicalDeclContext(DeclContext *DC) {
 void Decl::setDeclContextsImpl(DeclContext *SemaDC, DeclContext *LexicalDC,
                                ASTContext &Ctx) {
   if (SemaDC == LexicalDC) {
-    DeclCtx = SemaDC;
+    auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+    InnerPtr.setPointer(SemaDC);
+    DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
   } else {
     auto *MDC = new (Ctx) Decl::MultipleDC();
     MDC->SemanticDC = SemaDC;
     MDC->LexicalDC = LexicalDC;
-    DeclCtx = MDC;
+    auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+    InnerPtr.setPointer(MDC);
+    DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
   }
 }
 
@@ -956,19 +964,24 @@ unsigned Decl::getIdentifierNamespaceForKind(Kind DeclKind) {
 }
 
 void Decl::setAttrsImpl(const AttrVec &attrs, ASTContext &Ctx) {
-  assert(!HasAttrs && "Decl already contains attrs.");
+  assert(!hasAttrs() && "Decl already contains attrs.");
 
   AttrVec &AttrBlank = Ctx.getDeclAttrs(this);
   assert(AttrBlank.empty() && "HasAttrs was wrong?");
 
   AttrBlank = attrs;
-  HasAttrs = true;
+  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+  InnerPtr.setInt(true);
+  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
 }
 
 void Decl::dropAttrs() {
-  if (!HasAttrs) return;
+  if (!hasAttrs())
+    return;
 
-  HasAttrs = false;
+  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
+  InnerPtr.setInt(false);
+  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
   getASTContext().eraseDeclAttrs(this);
 }
 
@@ -996,7 +1009,7 @@ void Decl::addAttr(Attr *A) {
 }
 
 const AttrVec &Decl::getAttrs() const {
-  assert(HasAttrs && "No attrs to get!");
+  assert(hasAttrs() && "No attrs to get!");
   return getASTContext().getDeclAttrs(this);
 }
 
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index ff72fc30c718f..c5dd89e1c8a4d 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -588,7 +588,7 @@ void ASTDeclReader::VisitDecl(Decl *D) {
   bool HasStandaloneLexicalDC = DeclBits.getNextBit();
   bool HasAttrs = DeclBits.getNextBit();
   D->setTopLevelDeclInObjCContainer(DeclBits.getNextBit());
-  D->InvalidDecl = DeclBits.getNextBit();
+  D->DeclCtxWithInvalidDeclAndHasAttrs.setInt(DeclBits.getNextBit());
   D->FromASTFile = true;
 
   if (D->isTemplateParameter() || D->isTemplateParameterPack() ||

From d48d6ba9477aa380cd5a71f899d3cb6d629f175b Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sun, 14 Apr 2024 14:14:37 +0400
Subject: [PATCH 350/886] [clang][NFC] Factor out VLA checks in type traits
 (#88646)

This is a follow-up to #88473 suggested by @cor3ntin in
https://github.com/llvm/llvm-project/pull/88473#discussion_r1562198117.
---
 clang/lib/Sema/SemaExprCXX.cpp | 59 ++++++++++++++++++----------------
 1 file changed, 31 insertions(+), 28 deletions(-)

diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 1cfd3e56a583e..d2f18d4aa9b98 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -5012,6 +5012,20 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
   return From;
 }
 
+/// Checks that type T is not a VLA.
+///
+/// @returns @c true if @p T is VLA and a diagnostic was emitted,
+/// @c false otherwise.
+static bool DiagnoseVLAInCXXTypeTrait(Sema &S, const TypeSourceInfo *T,
+                                      clang::tok::TokenKind TypeTraitID) {
+  if (!T->getType()->isVariableArrayType())
+    return false;
+
+  S.Diag(T->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
+      << 1 << TypeTraitID;
+  return true;
+}
+
 /// Check the completeness of a type in a unary type trait.
 ///
 /// If the particular type trait requires a complete type, tries to complete
@@ -5188,7 +5202,9 @@ static bool HasNoThrowOperator(const RecordType *RT, OverloadedOperatorKind Op,
 }
 
 static bool EvaluateUnaryTypeTrait(Sema &Self, TypeTrait UTT,
-                                   SourceLocation KeyLoc, QualType T) {
+                                   SourceLocation KeyLoc,
+                                   TypeSourceInfo *TInfo) {
+  QualType T = TInfo->getType();
   assert(!T->isDependentType() && "Cannot evaluate traits of dependent type");
 
   ASTContext &C = Self.Context;
@@ -5205,21 +5221,13 @@ static bool EvaluateUnaryTypeTrait(Sema &Self, TypeTrait UTT,
   case UTT_IsArray:
     return T->isArrayType();
   case UTT_IsBoundedArray:
-    if (!T->isVariableArrayType()) {
-      return T->isArrayType() && !T->isIncompleteArrayType();
-    }
-
-    Self.Diag(KeyLoc, diag::err_vla_unsupported)
-        << 1 << tok::kw___is_bounded_array;
-    return false;
+    if (DiagnoseVLAInCXXTypeTrait(Self, TInfo, tok::kw___is_bounded_array))
+      return false;
+    return T->isArrayType() && !T->isIncompleteArrayType();
   case UTT_IsUnboundedArray:
-    if (!T->isVariableArrayType()) {
-      return T->isIncompleteArrayType();
-    }
-
-    Self.Diag(KeyLoc, diag::err_vla_unsupported)
-        << 1 << tok::kw___is_unbounded_array;
-    return false;
+    if (DiagnoseVLAInCXXTypeTrait(Self, TInfo, tok::kw___is_unbounded_array))
+      return false;
+    return T->isIncompleteArrayType();
   case UTT_IsPointer:
     return T->isAnyPointerType();
   case UTT_IsNullPointer:
@@ -5631,7 +5639,7 @@ static bool EvaluateBooleanTypeTrait(Sema &S, TypeTrait Kind,
     return false;
 
   if (Kind <= UTT_Last)
-    return EvaluateUnaryTypeTrait(S, Kind, KWLoc, Args[0]->getType());
+    return EvaluateUnaryTypeTrait(S, Kind, KWLoc, Args[0]);
 
   // Evaluate ReferenceBindsToTemporary and ReferenceConstructsFromTemporary
   // alongside the IsConstructible traits to avoid duplication.
@@ -6093,12 +6101,9 @@ static bool EvaluateBinaryTypeTrait(Sema &Self, TypeTrait BTT, const TypeSourceI
       Self.RequireCompleteType(Rhs->getTypeLoc().getBeginLoc(), RhsT,
                                diag::err_incomplete_type);
 
-    if (LhsT->isVariableArrayType())
-      Self.Diag(Lhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
-          << 1 << tok::kw___is_layout_compatible;
-    if (RhsT->isVariableArrayType())
-      Self.Diag(Rhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
-          << 1 << tok::kw___is_layout_compatible;
+    DiagnoseVLAInCXXTypeTrait(Self, Lhs, tok::kw___is_layout_compatible);
+    DiagnoseVLAInCXXTypeTrait(Self, Rhs, tok::kw___is_layout_compatible);
+
     return Self.IsLayoutCompatible(LhsT, RhsT);
   }
   case BTT_IsPointerInterconvertibleBaseOf: {
@@ -6108,12 +6113,10 @@ static bool EvaluateBinaryTypeTrait(Sema &Self, TypeTrait BTT, const TypeSourceI
                                diag::err_incomplete_type);
     }
 
-    if (LhsT->isVariableArrayType())
-      Self.Diag(Lhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
-          << 1 << tok::kw___is_pointer_interconvertible_base_of;
-    if (RhsT->isVariableArrayType())
-      Self.Diag(Rhs->getTypeLoc().getBeginLoc(), diag::err_vla_unsupported)
-          << 1 << tok::kw___is_pointer_interconvertible_base_of;
+    DiagnoseVLAInCXXTypeTrait(Self, Lhs,
+                              tok::kw___is_pointer_interconvertible_base_of);
+    DiagnoseVLAInCXXTypeTrait(Self, Rhs,
+                              tok::kw___is_pointer_interconvertible_base_of);
 
     return Self.IsPointerInterconvertibleBaseOf(Lhs, Rhs);
   }

From ed06b847d4e77d0b81fa6b095366bb070db57846 Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Sun, 14 Apr 2024 12:25:49 +0200
Subject: [PATCH 351/886] Revert "[Clang] Reduce the size of Decl and classes
 derived from it" (#88654)

Reverts llvm/llvm-project#87361

On 32 bit platforms there is only a single bit available in the
`DeclCtx`, resulting in an assertion failure.
---
 clang/include/clang/AST/DeclBase.h        | 72 ++++++++---------------
 clang/lib/AST/DeclBase.cpp                | 31 +++-------
 clang/lib/Serialization/ASTReaderDecl.cpp |  2 +-
 3 files changed, 35 insertions(+), 70 deletions(-)

diff --git a/clang/include/clang/AST/DeclBase.h b/clang/include/clang/AST/DeclBase.h
index 4bee18767dd11..858450926455c 100644
--- a/clang/include/clang/AST/DeclBase.h
+++ b/clang/include/clang/AST/DeclBase.h
@@ -268,37 +268,17 @@ class alignas(8) Decl {
   ///   }
   ///   void A::f(); // SemanticDC == namespace 'A'
   ///                // LexicalDC == global namespace
+  llvm::PointerUnion<DeclContext*, MultipleDC*> DeclCtx;
 
-  // Compress the InvalidDecl and HasAttrs bits into DeclCtx to keep Decl below
-  // 32 bytes in size
-  llvm::PointerIntPair<
-      llvm::PointerIntPair<llvm::PointerUnion<DeclContext *, MultipleDC *>, 1,
-                           bool>,
-      1, bool>
-      DeclCtxWithInvalidDeclAndHasAttrs;
-
-  bool isInSemaDC() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
-        .getPointer()
-        .is<DeclContext *>();
-  }
-
-  bool isOutOfSemaDC() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
-        .getPointer()
-        .is<MultipleDC *>();
-  }
+  bool isInSemaDC() const { return DeclCtx.is<DeclContext*>(); }
+  bool isOutOfSemaDC() const { return DeclCtx.is<MultipleDC*>(); }
 
   MultipleDC *getMultipleDC() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
-        .getPointer()
-        .get<MultipleDC *>();
+    return DeclCtx.get<MultipleDC*>();
   }
 
   DeclContext *getSemanticDC() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer()
-        .getPointer()
-        .get<DeclContext *>();
+    return DeclCtx.get<DeclContext*>();
   }
 
   /// Loc - The location of this decl.
@@ -308,6 +288,14 @@ class alignas(8) Decl {
   LLVM_PREFERRED_TYPE(Kind)
   unsigned DeclKind : 7;
 
+  /// InvalidDecl - This indicates a semantic error occurred.
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned InvalidDecl :  1;
+
+  /// HasAttrs - This indicates whether the decl has attributes or not.
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned HasAttrs : 1;
+
   /// Implicit - Whether this declaration was implicitly generated by
   /// the implementation rather than explicitly written by the user.
   LLVM_PREFERRED_TYPE(bool)
@@ -405,22 +393,21 @@ class alignas(8) Decl {
 protected:
   Decl(Kind DK, DeclContext *DC, SourceLocation L)
       : NextInContextAndBits(nullptr, getModuleOwnershipKindForChildOf(DC)),
-        DeclCtxWithInvalidDeclAndHasAttrs({DC, false}, false), Loc(L),
-        DeclKind(DK), Implicit(false), Used(false), Referenced(false),
+        DeclCtx(DC), Loc(L), DeclKind(DK), InvalidDecl(false), HasAttrs(false),
+        Implicit(false), Used(false), Referenced(false),
         TopLevelDeclInObjCContainer(false), Access(AS_none), FromASTFile(0),
         IdentifierNamespace(getIdentifierNamespaceForKind(DK)),
         CacheValidAndLinkage(llvm::to_underlying(Linkage::Invalid)) {
-    if (StatisticsEnabled)
-      add(DK);
+    if (StatisticsEnabled) add(DK);
   }
 
   Decl(Kind DK, EmptyShell Empty)
-      : DeclKind(DK), Implicit(false), Used(false), Referenced(false),
-        TopLevelDeclInObjCContainer(false), Access(AS_none), FromASTFile(0),
+      : DeclKind(DK), InvalidDecl(false), HasAttrs(false), Implicit(false),
+        Used(false), Referenced(false), TopLevelDeclInObjCContainer(false),
+        Access(AS_none), FromASTFile(0),
         IdentifierNamespace(getIdentifierNamespaceForKind(DK)),
         CacheValidAndLinkage(llvm::to_underlying(Linkage::Invalid)) {
-    if (StatisticsEnabled)
-      add(DK);
+    if (StatisticsEnabled) add(DK);
   }
 
   virtual ~Decl();
@@ -533,9 +520,7 @@ class alignas(8) Decl {
     return AccessSpecifier(Access);
   }
 
-  bool hasAttrs() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getPointer().getInt();
-  }
+  bool hasAttrs() const { return HasAttrs; }
 
   void setAttrs(const AttrVec& Attrs) {
     return setAttrsImpl(Attrs, getASTContext());
@@ -564,17 +549,13 @@ class alignas(8) Decl {
   }
 
   template <typename... Ts> void dropAttrs() {
-    if (!hasAttrs())
-      return;
+    if (!HasAttrs) return;
 
     AttrVec &Vec = getAttrs();
     llvm::erase_if(Vec, [](Attr *A) { return isa<Ts...>(A); });
 
-    if (Vec.empty()) {
-      auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-      InnerPtr.setInt(false);
-      DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
-    }
+    if (Vec.empty())
+      HasAttrs = false;
   }
 
   template <typename T> void dropAttr() { dropAttrs<T>(); }
@@ -609,10 +590,7 @@ class alignas(8) Decl {
   /// setInvalidDecl - Indicates the Decl had a semantic error. This
   /// allows for graceful error recovery.
   void setInvalidDecl(bool Invalid = true);
-
-  bool isInvalidDecl() const {
-    return DeclCtxWithInvalidDeclAndHasAttrs.getInt();
-  }
+  bool isInvalidDecl() const { return (bool) InvalidDecl; }
 
   /// isImplicit - Indicates whether the declaration was implicitly
   /// generated by the implementation. If false, this declaration
diff --git a/clang/lib/AST/DeclBase.cpp b/clang/lib/AST/DeclBase.cpp
index 8c83c711d882d..66a727d9dd0c3 100644
--- a/clang/lib/AST/DeclBase.cpp
+++ b/clang/lib/AST/DeclBase.cpp
@@ -52,8 +52,6 @@
 
 using namespace clang;
 
-static_assert(sizeof(Decl) <= 32, "Decl grew beyond 32 bytes!");
-
 //===----------------------------------------------------------------------===//
 //  Statistics
 //===----------------------------------------------------------------------===//
@@ -132,7 +130,7 @@ const char *Decl::getDeclKindName() const {
 }
 
 void Decl::setInvalidDecl(bool Invalid) {
-  DeclCtxWithInvalidDeclAndHasAttrs.setInt(Invalid);
+  InvalidDecl = Invalid;
   assert(!isa<TagDecl>(this) || !cast<TagDecl>(this)->isCompleteDefinition());
   if (!Invalid) {
     return;
@@ -336,9 +334,7 @@ void PrettyStackTraceDecl::print(raw_ostream &OS) const {
 Decl::~Decl() = default;
 
 void Decl::setDeclContext(DeclContext *DC) {
-  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-  InnerPtr.setPointer(DC);
-  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+  DeclCtx = DC;
 }
 
 void Decl::setLexicalDeclContext(DeclContext *DC) {
@@ -368,16 +364,12 @@ void Decl::setLexicalDeclContext(DeclContext *DC) {
 void Decl::setDeclContextsImpl(DeclContext *SemaDC, DeclContext *LexicalDC,
                                ASTContext &Ctx) {
   if (SemaDC == LexicalDC) {
-    auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-    InnerPtr.setPointer(SemaDC);
-    DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+    DeclCtx = SemaDC;
   } else {
     auto *MDC = new (Ctx) Decl::MultipleDC();
     MDC->SemanticDC = SemaDC;
     MDC->LexicalDC = LexicalDC;
-    auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-    InnerPtr.setPointer(MDC);
-    DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+    DeclCtx = MDC;
   }
 }
 
@@ -964,24 +956,19 @@ unsigned Decl::getIdentifierNamespaceForKind(Kind DeclKind) {
 }
 
 void Decl::setAttrsImpl(const AttrVec &attrs, ASTContext &Ctx) {
-  assert(!hasAttrs() && "Decl already contains attrs.");
+  assert(!HasAttrs && "Decl already contains attrs.");
 
   AttrVec &AttrBlank = Ctx.getDeclAttrs(this);
   assert(AttrBlank.empty() && "HasAttrs was wrong?");
 
   AttrBlank = attrs;
-  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-  InnerPtr.setInt(true);
-  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+  HasAttrs = true;
 }
 
 void Decl::dropAttrs() {
-  if (!hasAttrs())
-    return;
+  if (!HasAttrs) return;
 
-  auto InnerPtr = DeclCtxWithInvalidDeclAndHasAttrs.getPointer();
-  InnerPtr.setInt(false);
-  DeclCtxWithInvalidDeclAndHasAttrs.setPointer(InnerPtr);
+  HasAttrs = false;
   getASTContext().eraseDeclAttrs(this);
 }
 
@@ -1009,7 +996,7 @@ void Decl::addAttr(Attr *A) {
 }
 
 const AttrVec &Decl::getAttrs() const {
-  assert(hasAttrs() && "No attrs to get!");
+  assert(HasAttrs && "No attrs to get!");
   return getASTContext().getDeclAttrs(this);
 }
 
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index c5dd89e1c8a4d..ff72fc30c718f 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -588,7 +588,7 @@ void ASTDeclReader::VisitDecl(Decl *D) {
   bool HasStandaloneLexicalDC = DeclBits.getNextBit();
   bool HasAttrs = DeclBits.getNextBit();
   D->setTopLevelDeclInObjCContainer(DeclBits.getNextBit());
-  D->DeclCtxWithInvalidDeclAndHasAttrs.setInt(DeclBits.getNextBit());
+  D->InvalidDecl = DeclBits.getNextBit();
   D->FromASTFile = true;
 
   if (D->isTemplateParameter() || D->isTemplateParameterPack() ||

From ef164cee90477e294ff692209b4cf97a0e1958ed Mon Sep 17 00:00:00 2001
From: Sirraide <aeternalmail@gmail.com>
Date: Sun, 14 Apr 2024 12:30:01 +0200
Subject: [PATCH 352/886] [Clang] [C++26] Implement P2573R2: `= delete("should
 have a reason");` (#86526)

This implements support for the `= delete("message")` syntax that was
only just added to C++26
([P2573R2](https://isocpp.org/files/papers/P2573R2.html#proposal-scope)).
---
 clang/docs/LanguageExtensions.rst             |   1 +
 clang/docs/ReleaseNotes.rst                   |   2 +
 clang/include/clang/AST/Decl.h                |  47 ++-
 clang/include/clang/AST/DeclBase.h            |   2 +-
 .../clang/Basic/DiagnosticParseKinds.td       |   6 +
 .../clang/Basic/DiagnosticSemaKinds.td        |  17 +-
 clang/include/clang/Parse/Parser.h            |   2 +
 clang/include/clang/Sema/Sema.h               |  17 +-
 clang/lib/AST/ASTImporter.cpp                 |  13 +
 clang/lib/AST/Decl.cpp                        |  69 +++--
 clang/lib/AST/DeclPrinter.cpp                 |   9 +-
 clang/lib/AST/JSONNodeDumper.cpp              |   3 +
 clang/lib/AST/ODRHash.cpp                     |   6 +
 clang/lib/AST/TextNodeDumper.cpp              |   3 +
 clang/lib/Frontend/InitPreprocessor.cpp       |   3 +
 clang/lib/Parse/ParseCXXInlineMethods.cpp     |  46 ++-
 clang/lib/Parse/ParseDecl.cpp                 |   1 +
 clang/lib/Parse/ParseDeclCXX.cpp              |   1 +
 clang/lib/Parse/Parser.cpp                    |   4 +-
 clang/lib/Sema/SemaCast.cpp                   |  20 +-
 clang/lib/Sema/SemaDecl.cpp                   |  12 +-
 clang/lib/Sema/SemaDeclCXX.cpp                |  18 +-
 clang/lib/Sema/SemaExpr.cpp                   |   7 +-
 clang/lib/Sema/SemaExprCXX.cpp                |  23 +-
 clang/lib/Sema/SemaInit.cpp                   |  19 +-
 clang/lib/Sema/SemaOverload.cpp               |  84 ++++--
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |  11 +-
 clang/lib/Serialization/ASTReaderDecl.cpp     |  18 +-
 clang/lib/Serialization/ASTWriterDecl.cpp     |  11 +-
 .../ast-dump-cxx2c-delete-with-message.cpp    |  23 ++
 .../ast-print-cxx2c-delete-with-message.cpp   |  18 ++
 clang/test/Lexer/cxx-features.cpp             |   4 +
 clang/test/PCH/cxx2a-defaulted-comparison.cpp |   2 +-
 .../test/Parser/cxx2c-delete-with-message.cpp |  51 ++++
 .../SemaCXX/cxx2c-delete-with-message.cpp     | 273 ++++++++++++++++++
 clang/www/cxx_status.html                     |   2 +-
 36 files changed, 728 insertions(+), 120 deletions(-)
 create mode 100644 clang/test/AST/ast-dump-cxx2c-delete-with-message.cpp
 create mode 100644 clang/test/AST/ast-print-cxx2c-delete-with-message.cpp
 create mode 100644 clang/test/Parser/cxx2c-delete-with-message.cpp
 create mode 100644 clang/test/SemaCXX/cxx2c-delete-with-message.cpp

diff --git a/clang/docs/LanguageExtensions.rst b/clang/docs/LanguageExtensions.rst
index 96691b45d63a3..05c8f765b5569 100644
--- a/clang/docs/LanguageExtensions.rst
+++ b/clang/docs/LanguageExtensions.rst
@@ -1493,6 +1493,7 @@ Conditional ``explicit``                     __cpp_conditional_explicit       C+
 ``if consteval``                             __cpp_if_consteval               C++23         C++20
 ``static operator()``                        __cpp_static_call_operator       C++23         C++03
 Attributes on Lambda-Expressions                                              C++23         C++11
+``= delete ("should have a reason");``       __cpp_deleted_function           C++26         C++03
 -------------------------------------------- -------------------------------- ------------- -------------
 Designated initializers (N494)                                                C99           C89
 Array & element qualification (N2607)                                         C23           C89
diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 399d5ba796d15..ade8f4e93d5a0 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -129,6 +129,8 @@ C++2c Feature Support
 
 - Implemented `P2662R3 Pack Indexing <https://wg21.link/P2662R3>`_.
 
+- Implemented `P2573R2: = delete("should have a reason"); <https://wg21.link/P2573R2>`_
+
 
 Resolutions to C++ Defect Reports
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/include/clang/AST/Decl.h b/clang/include/clang/AST/Decl.h
index ed6790acdfc7c..01af50ca694fd 100644
--- a/clang/include/clang/AST/Decl.h
+++ b/clang/include/clang/AST/Decl.h
@@ -1993,21 +1993,35 @@ class FunctionDecl : public DeclaratorDecl,
 
   };
 
-  /// Stashed information about a defaulted function definition whose body has
-  /// not yet been lazily generated.
-  class DefaultedFunctionInfo final
-      : llvm::TrailingObjects<DefaultedFunctionInfo, DeclAccessPair> {
+  /// Stashed information about a defaulted/deleted function body.
+  class DefaultedOrDeletedFunctionInfo final
+      : llvm::TrailingObjects<DefaultedOrDeletedFunctionInfo, DeclAccessPair,
+                              StringLiteral *> {
     friend TrailingObjects;
     unsigned NumLookups;
+    bool HasDeletedMessage;
+
+    size_t numTrailingObjects(OverloadToken<DeclAccessPair>) const {
+      return NumLookups;
+    }
 
   public:
-    static DefaultedFunctionInfo *Create(ASTContext &Context,
-                                         ArrayRef<DeclAccessPair> Lookups);
+    static DefaultedOrDeletedFunctionInfo *
+    Create(ASTContext &Context, ArrayRef<DeclAccessPair> Lookups,
+           StringLiteral *DeletedMessage = nullptr);
+
     /// Get the unqualified lookup results that should be used in this
     /// defaulted function definition.
     ArrayRef<DeclAccessPair> getUnqualifiedLookups() const {
       return {getTrailingObjects<DeclAccessPair>(), NumLookups};
     }
+
+    StringLiteral *getDeletedMessage() const {
+      return HasDeletedMessage ? *getTrailingObjects<StringLiteral *>()
+                               : nullptr;
+    }
+
+    void setDeletedMessage(StringLiteral *Message);
   };
 
 private:
@@ -2017,12 +2031,12 @@ class FunctionDecl : public DeclaratorDecl,
   ParmVarDecl **ParamInfo = nullptr;
 
   /// The active member of this union is determined by
-  /// FunctionDeclBits.HasDefaultedFunctionInfo.
+  /// FunctionDeclBits.HasDefaultedOrDeletedInfo.
   union {
     /// The body of the function.
     LazyDeclStmtPtr Body;
     /// Information about a future defaulted function definition.
-    DefaultedFunctionInfo *DefaultedInfo;
+    DefaultedOrDeletedFunctionInfo *DefaultedOrDeletedInfo;
   };
 
   unsigned ODRHash;
@@ -2280,18 +2294,18 @@ class FunctionDecl : public DeclaratorDecl,
 
   /// Returns whether this specific declaration of the function has a body.
   bool doesThisDeclarationHaveABody() const {
-    return (!FunctionDeclBits.HasDefaultedFunctionInfo && Body) ||
+    return (!FunctionDeclBits.HasDefaultedOrDeletedInfo && Body) ||
            isLateTemplateParsed();
   }
 
   void setBody(Stmt *B);
   void setLazyBody(uint64_t Offset) {
-    FunctionDeclBits.HasDefaultedFunctionInfo = false;
+    FunctionDeclBits.HasDefaultedOrDeletedInfo = false;
     Body = LazyDeclStmtPtr(Offset);
   }
 
-  void setDefaultedFunctionInfo(DefaultedFunctionInfo *Info);
-  DefaultedFunctionInfo *getDefaultedFunctionInfo() const;
+  void setDefaultedOrDeletedInfo(DefaultedOrDeletedFunctionInfo *Info);
+  DefaultedOrDeletedFunctionInfo *getDefalutedOrDeletedInfo() const;
 
   /// Whether this function is variadic.
   bool isVariadic() const;
@@ -2494,7 +2508,7 @@ class FunctionDecl : public DeclaratorDecl,
     return FunctionDeclBits.IsDeleted && !isDefaulted();
   }
 
-  void setDeletedAsWritten(bool D = true) { FunctionDeclBits.IsDeleted = D; }
+  void setDeletedAsWritten(bool D = true, StringLiteral *Message = nullptr);
 
   /// Determines whether this function is "main", which is the
   /// entry point into an executable program.
@@ -2650,6 +2664,13 @@ class FunctionDecl : public DeclaratorDecl,
       AC.push_back(TRC);
   }
 
+  /// Get the message that indicates why this function was deleted.
+  StringLiteral *getDeletedMessage() const {
+    return FunctionDeclBits.HasDefaultedOrDeletedInfo
+               ? DefaultedOrDeletedInfo->getDeletedMessage()
+               : nullptr;
+  }
+
   void setPreviousDeclaration(FunctionDecl * PrevDecl);
 
   FunctionDecl *getCanonicalDecl() override;
diff --git a/clang/include/clang/AST/DeclBase.h b/clang/include/clang/AST/DeclBase.h
index 858450926455c..2194d268fa86f 100644
--- a/clang/include/clang/AST/DeclBase.h
+++ b/clang/include/clang/AST/DeclBase.h
@@ -1739,7 +1739,7 @@ class DeclContext {
     LLVM_PREFERRED_TYPE(bool)
     uint64_t IsExplicitlyDefaulted : 1;
     LLVM_PREFERRED_TYPE(bool)
-    uint64_t HasDefaultedFunctionInfo : 1;
+    uint64_t HasDefaultedOrDeletedInfo : 1;
 
     /// For member functions of complete types, whether this is an ineligible
     /// special member function or an unselected destructor. See
diff --git a/clang/include/clang/Basic/DiagnosticParseKinds.td b/clang/include/clang/Basic/DiagnosticParseKinds.td
index 46a44418a3153..bb9ca2a50cc06 100644
--- a/clang/include/clang/Basic/DiagnosticParseKinds.td
+++ b/clang/include/clang/Basic/DiagnosticParseKinds.td
@@ -941,6 +941,12 @@ def warn_cxx98_compat_defaulted_deleted_function : Warning<
   "%select{defaulted|deleted}0 function definitions are incompatible with C++98">,
   InGroup<CXX98Compat>, DefaultIgnore;
 
+def ext_delete_with_message : ExtWarn<
+  "'= delete' with a message is a C++2c extension">, InGroup<CXX26>;
+def warn_cxx23_delete_with_message : Warning<
+  "'= delete' with a message is incompatible with C++ standards before C++2c">,
+  DefaultIgnore, InGroup<CXXPre26Compat>;
+
 // C++11 default member initialization
 def ext_nonstatic_member_init : ExtWarn<
   "default member initializer for non-static data member is a C++11 "
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 774d2b53a3825..5ec0218aedfe8 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -4683,11 +4683,10 @@ def err_ovl_no_viable_member_function_in_call : Error<
   "no matching member function for call to %0">;
 def err_ovl_ambiguous_call : Error<
   "call to %0 is ambiguous">;
-def err_ovl_deleted_call : Error<"call to deleted function %0">;
+def err_ovl_deleted_call : Error<"call to deleted"
+  "%select{| member}0 function %1%select{|: %3}2">;
 def err_ovl_ambiguous_member_call : Error<
   "call to member function %0 is ambiguous">;
-def err_ovl_deleted_member_call : Error<
-  "call to deleted member function %0">;
 def note_ovl_too_many_candidates : Note<
     "remaining %0 candidate%s0 omitted; "
     "pass -fshow-overloads=all to show them">;
@@ -4915,12 +4914,12 @@ def err_ovl_ambiguous_conversion_in_cast : Error<
   "dynamic_cast|C-style cast|functional-style cast|}0 from %1 to %2">;
 def err_ovl_deleted_conversion_in_cast : Error<
   "%select{|static_cast|reinterpret_cast|dynamic_cast|C-style cast|"
-  "functional-style cast|}0 from %1 to %2 uses deleted function">;
+  "functional-style cast|}0 from %1 to %2 uses deleted function%select{|: %4}3">;
 def err_ovl_ambiguous_init : Error<"call to constructor of %0 is ambiguous">;
 def err_ref_init_ambiguous : Error<
   "reference initialization of type %0 with initializer of type %1 is ambiguous">;
 def err_ovl_deleted_init : Error<
-  "call to deleted constructor of %0">;
+  "call to deleted constructor of %0%select{|: %2}1">;
 def err_ovl_deleted_special_init : Error<
   "call to implicitly-deleted %select{default constructor|copy constructor|"
   "move constructor|copy assignment operator|move assignment operator|"
@@ -4946,7 +4945,7 @@ def note_ovl_ambiguous_oper_binary_reversed_candidate : Note<
 def err_ovl_no_viable_oper : Error<"no viable overloaded '%0'">;
 def note_assign_lhs_incomplete : Note<"type %0 is incomplete">;
 def err_ovl_deleted_oper : Error<
-  "overload resolution selected deleted operator '%0'">;
+  "overload resolution selected deleted operator '%0'%select{|: %2}1">;
 def err_ovl_deleted_special_oper : Error<
   "object of type %0 cannot be %select{constructed|copied|moved|assigned|"
   "assigned|destroyed}1 because its %sub{select_special_member_kind}1 is "
@@ -4983,7 +4982,7 @@ def err_ovl_ambiguous_object_call : Error<
 def err_ovl_ambiguous_subscript_call : Error<
   "call to subscript operator of type %0 is ambiguous">;
 def err_ovl_deleted_object_call : Error<
-  "call to deleted function call operator in type %0">;
+  "call to deleted function call operator in type %0%select{|: %2}1">;
 def note_ovl_surrogate_cand : Note<"conversion candidate of type %0">;
 def err_member_call_without_object : Error<
   "call to %select{non-static|explicit}0 member function without an object argument">;
@@ -8284,7 +8283,7 @@ def err_typecheck_nonviable_condition_incomplete : Error<
   "no viable conversion%diff{ from $ to incomplete type $|}0,1">;
 def err_typecheck_deleted_function : Error<
   "conversion function %diff{from $ to $|between types}0,1 "
-  "invokes a deleted function">;
+  "invokes a deleted function%select{|: %3}2">;
 
 def err_expected_class_or_namespace : Error<"%0 is not a class"
   "%select{ or namespace|, namespace, or enumeration}1">;
@@ -8884,7 +8883,7 @@ def err_nontemporal_builtin_must_be_pointer_intfltptr_or_vector : Error<
   "address argument to nontemporal builtin must be a pointer to integer, float, "
   "pointer, or a vector of such types (%0 invalid)">;
 
-def err_deleted_function_use : Error<"attempt to use a deleted function">;
+def err_deleted_function_use : Error<"attempt to use a deleted function%select{|: %1}0">;
 def err_deleted_inherited_ctor_use : Error<
   "constructor inherited by %0 from base class %1 is implicitly deleted">;
 
diff --git a/clang/include/clang/Parse/Parser.h b/clang/include/clang/Parse/Parser.h
index c719218731c35..5950dd74cfe83 100644
--- a/clang/include/clang/Parse/Parser.h
+++ b/clang/include/clang/Parse/Parser.h
@@ -1601,6 +1601,8 @@ class Parser : public CodeCompletionHandler {
                                      const ParsedTemplateInfo &TemplateInfo,
                                      const VirtSpecifiers &VS,
                                      SourceLocation PureSpecLoc);
+  StringLiteral *ParseCXXDeletedFunctionMessage();
+  void SkipDeletedFunctionBody();
   void ParseCXXNonStaticMemberInitializer(Decl *VarD);
   void ParseLexedAttributes(ParsingClass &Class);
   void ParseLexedAttributeList(LateParsedAttrList &LAs, Decl *D,
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 3b252f94dcbee..c6e0332c3176b 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -3045,14 +3045,18 @@ class Sema final : public SemaBase {
   void ActOnDocumentableDecls(ArrayRef<Decl *> Group);
 
   enum class FnBodyKind {
-    /// C++ [dcl.fct.def.general]p1
+    /// C++26 [dcl.fct.def.general]p1
     /// function-body:
     ///   ctor-initializer[opt] compound-statement
     ///   function-try-block
     Other,
     ///   = default ;
     Default,
+    ///   deleted-function-body
+    ///
+    /// deleted-function-body:
     ///   = delete ;
+    ///   = delete ( unevaluated-string ) ;
     Delete
   };
 
@@ -4751,10 +4755,12 @@ class Sema final : public SemaBase {
                                SourceLocation EqualLoc);
 
   void ActOnPureSpecifier(Decl *D, SourceLocation PureSpecLoc);
-  void SetDeclDeleted(Decl *dcl, SourceLocation DelLoc);
+  void SetDeclDeleted(Decl *dcl, SourceLocation DelLoc,
+                      StringLiteral *Message = nullptr);
   void SetDeclDefaulted(Decl *dcl, SourceLocation DefaultLoc);
 
-  void SetFunctionBodyKind(Decl *D, SourceLocation Loc, FnBodyKind BodyKind);
+  void SetFunctionBodyKind(Decl *D, SourceLocation Loc, FnBodyKind BodyKind,
+                           StringLiteral *DeletedMessage = nullptr);
   void ActOnStartTrailingRequiresClause(Scope *S, Declarator &D);
   ExprResult ActOnFinishTrailingRequiresClause(ExprResult ConstraintExpr);
   ExprResult ActOnRequiresClause(ExprResult ConstraintExpr);
@@ -8093,6 +8099,11 @@ class Sema final : public SemaBase {
   bool IsFunctionConversion(QualType FromType, QualType ToType,
                             QualType &ResultTy);
   bool DiagnoseMultipleUserDefinedConversion(Expr *From, QualType ToType);
+  void DiagnoseUseOfDeletedFunction(SourceLocation Loc, SourceRange Range,
+                                    DeclarationName Name,
+                                    OverloadCandidateSet &CandidateSet,
+                                    FunctionDecl *Fn, MultiExprArg Args,
+                                    bool IsMember = false);
 
   ExprResult InitializeExplicitObjectArgument(Sema &S, Expr *Obj,
                                               FunctionDecl *Fun);
diff --git a/clang/lib/AST/ASTImporter.cpp b/clang/lib/AST/ASTImporter.cpp
index a5e43fc631667..6aaa34c55ce30 100644
--- a/clang/lib/AST/ASTImporter.cpp
+++ b/clang/lib/AST/ASTImporter.cpp
@@ -3947,6 +3947,14 @@ ExpectedDecl ASTNodeImporter::VisitFunctionDecl(FunctionDecl *D) {
     // decl and its redeclarations may be required.
   }
 
+  StringLiteral *Msg = D->getDeletedMessage();
+  if (Msg) {
+    auto Imported = import(Msg);
+    if (!Imported)
+      return Imported.takeError();
+    Msg = *Imported;
+  }
+
   ToFunction->setQualifierInfo(ToQualifierLoc);
   ToFunction->setAccess(D->getAccess());
   ToFunction->setLexicalDeclContext(LexicalDC);
@@ -3961,6 +3969,11 @@ ExpectedDecl ASTNodeImporter::VisitFunctionDecl(FunctionDecl *D) {
   ToFunction->setRangeEnd(ToEndLoc);
   ToFunction->setDefaultLoc(ToDefaultLoc);
 
+  if (Msg)
+    ToFunction->setDefaultedOrDeletedInfo(
+        FunctionDecl::DefaultedOrDeletedFunctionInfo::Create(
+            Importer.getToContext(), {}, Msg));
+
   // Set the parameters.
   for (auto *Param : Parameters) {
     Param->setOwningFunction(ToFunction);
diff --git a/clang/lib/AST/Decl.cpp b/clang/lib/AST/Decl.cpp
index 60e0a3aecf6c8..2b2d5a2663a18 100644
--- a/clang/lib/AST/Decl.cpp
+++ b/clang/lib/AST/Decl.cpp
@@ -3058,7 +3058,7 @@ FunctionDecl::FunctionDecl(Kind DK, ASTContext &C, DeclContext *DC,
   FunctionDeclBits.IsTrivialForCall = false;
   FunctionDeclBits.IsDefaulted = false;
   FunctionDeclBits.IsExplicitlyDefaulted = false;
-  FunctionDeclBits.HasDefaultedFunctionInfo = false;
+  FunctionDeclBits.HasDefaultedOrDeletedInfo = false;
   FunctionDeclBits.IsIneligibleOrNotSelected = false;
   FunctionDeclBits.HasImplicitReturnZero = false;
   FunctionDeclBits.IsLateTemplateParsed = false;
@@ -3092,30 +3092,65 @@ bool FunctionDecl::isVariadic() const {
   return false;
 }
 
-FunctionDecl::DefaultedFunctionInfo *
-FunctionDecl::DefaultedFunctionInfo::Create(ASTContext &Context,
-                                            ArrayRef<DeclAccessPair> Lookups) {
-  DefaultedFunctionInfo *Info = new (Context.Allocate(
-      totalSizeToAlloc<DeclAccessPair>(Lookups.size()),
-      std::max(alignof(DefaultedFunctionInfo), alignof(DeclAccessPair))))
-      DefaultedFunctionInfo;
+FunctionDecl::DefaultedOrDeletedFunctionInfo *
+FunctionDecl::DefaultedOrDeletedFunctionInfo::Create(
+    ASTContext &Context, ArrayRef<DeclAccessPair> Lookups,
+    StringLiteral *DeletedMessage) {
+  static constexpr size_t Alignment =
+      std::max({alignof(DefaultedOrDeletedFunctionInfo),
+                alignof(DeclAccessPair), alignof(StringLiteral *)});
+  size_t Size = totalSizeToAlloc<DeclAccessPair, StringLiteral *>(
+      Lookups.size(), DeletedMessage != nullptr);
+
+  DefaultedOrDeletedFunctionInfo *Info =
+      new (Context.Allocate(Size, Alignment)) DefaultedOrDeletedFunctionInfo;
   Info->NumLookups = Lookups.size();
+  Info->HasDeletedMessage = DeletedMessage != nullptr;
+
   std::uninitialized_copy(Lookups.begin(), Lookups.end(),
                           Info->getTrailingObjects<DeclAccessPair>());
+  if (DeletedMessage)
+    *Info->getTrailingObjects<StringLiteral *>() = DeletedMessage;
   return Info;
 }
 
-void FunctionDecl::setDefaultedFunctionInfo(DefaultedFunctionInfo *Info) {
-  assert(!FunctionDeclBits.HasDefaultedFunctionInfo && "already have this");
+void FunctionDecl::setDefaultedOrDeletedInfo(
+    DefaultedOrDeletedFunctionInfo *Info) {
+  assert(!FunctionDeclBits.HasDefaultedOrDeletedInfo && "already have this");
   assert(!Body && "can't replace function body with defaulted function info");
 
-  FunctionDeclBits.HasDefaultedFunctionInfo = true;
-  DefaultedInfo = Info;
+  FunctionDeclBits.HasDefaultedOrDeletedInfo = true;
+  DefaultedOrDeletedInfo = Info;
+}
+
+void FunctionDecl::setDeletedAsWritten(bool D, StringLiteral *Message) {
+  FunctionDeclBits.IsDeleted = D;
+
+  if (Message) {
+    assert(isDeletedAsWritten() && "Function must be deleted");
+    if (FunctionDeclBits.HasDefaultedOrDeletedInfo)
+      DefaultedOrDeletedInfo->setDeletedMessage(Message);
+    else
+      setDefaultedOrDeletedInfo(DefaultedOrDeletedFunctionInfo::Create(
+          getASTContext(), /*Lookups=*/{}, Message));
+  }
+}
+
+void FunctionDecl::DefaultedOrDeletedFunctionInfo::setDeletedMessage(
+    StringLiteral *Message) {
+  // We should never get here with the DefaultedOrDeletedInfo populated, but
+  // no space allocated for the deleted message, since that would require
+  // recreating this, but setDefaultedOrDeletedInfo() disallows overwriting
+  // an already existing DefaultedOrDeletedFunctionInfo.
+  assert(HasDeletedMessage &&
+         "No space to store a delete message in this DefaultedOrDeletedInfo");
+  *getTrailingObjects<StringLiteral *>() = Message;
 }
 
-FunctionDecl::DefaultedFunctionInfo *
-FunctionDecl::getDefaultedFunctionInfo() const {
-  return FunctionDeclBits.HasDefaultedFunctionInfo ? DefaultedInfo : nullptr;
+FunctionDecl::DefaultedOrDeletedFunctionInfo *
+FunctionDecl::getDefalutedOrDeletedInfo() const {
+  return FunctionDeclBits.HasDefaultedOrDeletedInfo ? DefaultedOrDeletedInfo
+                                                    : nullptr;
 }
 
 bool FunctionDecl::hasBody(const FunctionDecl *&Definition) const {
@@ -3202,7 +3237,7 @@ Stmt *FunctionDecl::getBody(const FunctionDecl *&Definition) const {
   if (!hasBody(Definition))
     return nullptr;
 
-  assert(!Definition->FunctionDeclBits.HasDefaultedFunctionInfo &&
+  assert(!Definition->FunctionDeclBits.HasDefaultedOrDeletedInfo &&
          "definition should not have a body");
   if (Definition->Body)
     return Definition->Body.get(getASTContext().getExternalSource());
@@ -3211,7 +3246,7 @@ Stmt *FunctionDecl::getBody(const FunctionDecl *&Definition) const {
 }
 
 void FunctionDecl::setBody(Stmt *B) {
-  FunctionDeclBits.HasDefaultedFunctionInfo = false;
+  FunctionDeclBits.HasDefaultedOrDeletedInfo = false;
   Body = LazyDeclStmtPtr(B);
   if (B)
     EndRangeLoc = B->getEndLoc();
diff --git a/clang/lib/AST/DeclPrinter.cpp b/clang/lib/AST/DeclPrinter.cpp
index c66774dd1df15..93857adb990bf 100644
--- a/clang/lib/AST/DeclPrinter.cpp
+++ b/clang/lib/AST/DeclPrinter.cpp
@@ -822,9 +822,14 @@ void DeclPrinter::VisitFunctionDecl(FunctionDecl *D) {
 
   if (D->isPureVirtual())
     Out << " = 0";
-  else if (D->isDeletedAsWritten())
+  else if (D->isDeletedAsWritten()) {
     Out << " = delete";
-  else if (D->isExplicitlyDefaulted())
+    if (const StringLiteral *M = D->getDeletedMessage()) {
+      Out << "(";
+      M->outputString(Out);
+      Out << ")";
+    }
+  } else if (D->isExplicitlyDefaulted())
     Out << " = default";
   else if (D->doesThisDeclarationHaveABody()) {
     if (!Policy.TerseOutput) {
diff --git a/clang/lib/AST/JSONNodeDumper.cpp b/clang/lib/AST/JSONNodeDumper.cpp
index 78115200f4f0d..42608476b1c19 100644
--- a/clang/lib/AST/JSONNodeDumper.cpp
+++ b/clang/lib/AST/JSONNodeDumper.cpp
@@ -975,6 +975,9 @@ void JSONNodeDumper::VisitFunctionDecl(const FunctionDecl *FD) {
   if (FD->isDefaulted())
     JOS.attribute("explicitlyDefaulted",
                   FD->isDeleted() ? "deleted" : "default");
+
+  if (StringLiteral *Msg = FD->getDeletedMessage())
+    JOS.attribute("deletedMessage", Msg->getString());
 }
 
 void JSONNodeDumper::VisitEnumDecl(const EnumDecl *ED) {
diff --git a/clang/lib/AST/ODRHash.cpp b/clang/lib/AST/ODRHash.cpp
index e159a1b00be55..6f04739cf6693 100644
--- a/clang/lib/AST/ODRHash.cpp
+++ b/clang/lib/AST/ODRHash.cpp
@@ -696,6 +696,12 @@ void ODRHash::AddFunctionDecl(const FunctionDecl *Function,
   AddBoolean(Function->isDeletedAsWritten());
   AddBoolean(Function->isExplicitlyDefaulted());
 
+  StringLiteral *DeletedMessage = Function->getDeletedMessage();
+  AddBoolean(DeletedMessage);
+
+  if (DeletedMessage)
+    ID.AddString(DeletedMessage->getBytes());
+
   AddDecl(Function);
 
   AddQualType(Function->getReturnType());
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 0973e63183bd7..688daa64d6197 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -1961,6 +1961,9 @@ void TextNodeDumper::VisitFunctionDecl(const FunctionDecl *D) {
   if (D->isTrivial())
     OS << " trivial";
 
+  if (const StringLiteral *M = D->getDeletedMessage())
+    AddChild("delete message", [=] { Visit(M); });
+
   if (D->isIneligibleOrNotSelected())
     OS << (isa<CXXDestructorDecl>(D) ? " not_selected" : " ineligible");
 
diff --git a/clang/lib/Frontend/InitPreprocessor.cpp b/clang/lib/Frontend/InitPreprocessor.cpp
index 84069e96f4146..4f44c3b7b89d4 100644
--- a/clang/lib/Frontend/InitPreprocessor.cpp
+++ b/clang/lib/Frontend/InitPreprocessor.cpp
@@ -747,6 +747,9 @@ static void InitializeCPlusPlusFeatureTestMacros(const LangOptions &LangOpts,
   Builder.defineMacro("__cpp_named_character_escapes", "202207L");
   Builder.defineMacro("__cpp_placeholder_variables", "202306L");
 
+  // C++26 features supported in earlier language modes.
+  Builder.defineMacro("__cpp_deleted_function", "202403L");
+
   if (LangOpts.Char8)
     Builder.defineMacro("__cpp_char8_t", "202207L");
   Builder.defineMacro("__cpp_impl_destroying_delete", "201806L");
diff --git a/clang/lib/Parse/ParseCXXInlineMethods.cpp b/clang/lib/Parse/ParseCXXInlineMethods.cpp
index d790060c17c04..d054eda279b8c 100644
--- a/clang/lib/Parse/ParseCXXInlineMethods.cpp
+++ b/clang/lib/Parse/ParseCXXInlineMethods.cpp
@@ -20,6 +20,49 @@
 
 using namespace clang;
 
+/// Parse the optional ("message") part of a deleted-function-body.
+StringLiteral *Parser::ParseCXXDeletedFunctionMessage() {
+  if (!Tok.is(tok::l_paren))
+    return nullptr;
+  StringLiteral *Message = nullptr;
+  BalancedDelimiterTracker BT{*this, tok::l_paren};
+  BT.consumeOpen();
+
+  if (isTokenStringLiteral()) {
+    ExprResult Res = ParseUnevaluatedStringLiteralExpression();
+    if (Res.isUsable()) {
+      Message = Res.getAs<StringLiteral>();
+      Diag(Message->getBeginLoc(), getLangOpts().CPlusPlus26
+                                       ? diag::warn_cxx23_delete_with_message
+                                       : diag::ext_delete_with_message)
+          << Message->getSourceRange();
+    }
+  } else {
+    Diag(Tok.getLocation(), diag::err_expected_string_literal)
+        << /*Source='in'*/ 0 << "'delete'";
+    SkipUntil(tok::r_paren, StopAtSemi | StopBeforeMatch);
+  }
+
+  BT.consumeClose();
+  return Message;
+}
+
+/// If we've encountered '= delete' in a context where it is ill-formed, such
+/// as in the declaration of a non-function, also skip the ("message") part if
+/// it is present to avoid issuing further diagnostics.
+void Parser::SkipDeletedFunctionBody() {
+  if (!Tok.is(tok::l_paren))
+    return;
+
+  BalancedDelimiterTracker BT{*this, tok::l_paren};
+  BT.consumeOpen();
+
+  // Just skip to the end of the current declaration.
+  SkipUntil(tok::r_paren, tok::comma, StopAtSemi | StopBeforeMatch);
+  if (Tok.is(tok::r_paren))
+    BT.consumeClose();
+}
+
 /// ParseCXXInlineMethodDef - We parsed and verified that the specified
 /// Declarator is a well formed C++ inline method definition. Now lex its body
 /// and store its tokens for parsing after the C++ class is complete.
@@ -70,7 +113,8 @@ NamedDecl *Parser::ParseCXXInlineMethodDef(
                       ? diag::warn_cxx98_compat_defaulted_deleted_function
                       : diag::ext_defaulted_deleted_function)
         << 1 /* deleted */;
-      Actions.SetDeclDeleted(FnD, KWLoc);
+      StringLiteral *Message = ParseCXXDeletedFunctionMessage();
+      Actions.SetDeclDeleted(FnD, KWLoc, Message);
       Delete = true;
       if (auto *DeclAsFunction = dyn_cast<FunctionDecl>(FnD)) {
         DeclAsFunction->setRangeEnd(KWEndLoc);
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index 951e2210031a1..a990ef398c823 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -2678,6 +2678,7 @@ Decl *Parser::ParseDeclarationAfterDeclaratorAndAttributes(
           << 1 /* delete */;
       else
         Diag(ConsumeToken(), diag::err_deleted_non_function);
+      SkipDeletedFunctionBody();
     } else if (Tok.is(tok::kw_default)) {
       if (D.isFunctionDeclarator())
         Diag(ConsumeToken(), diag::err_default_delete_in_multiple_declaration)
diff --git a/clang/lib/Parse/ParseDeclCXX.cpp b/clang/lib/Parse/ParseDeclCXX.cpp
index 477d81cdc2c23..cd4803d51bc1d 100644
--- a/clang/lib/Parse/ParseDeclCXX.cpp
+++ b/clang/lib/Parse/ParseDeclCXX.cpp
@@ -3397,6 +3397,7 @@ ExprResult Parser::ParseCXXMemberInitializer(Decl *D, bool IsFunction,
               << 1 /* delete */;
         else
           Diag(ConsumeToken(), diag::err_deleted_non_function);
+        SkipDeletedFunctionBody();
         return ExprError();
       }
     } else if (Tok.is(tok::kw_default)) {
diff --git a/clang/lib/Parse/Parser.cpp b/clang/lib/Parse/Parser.cpp
index cc0e41ed221c4..d6f2b9f448cd5 100644
--- a/clang/lib/Parse/Parser.cpp
+++ b/clang/lib/Parse/Parser.cpp
@@ -1404,6 +1404,7 @@ Decl *Parser::ParseFunctionDefinition(ParsingDeclarator &D,
 
   // Parse function body eagerly if it is either '= delete;' or '= default;' as
   // ActOnStartOfFunctionDef needs to know whether the function is deleted.
+  StringLiteral *DeletedMessage = nullptr;
   Sema::FnBodyKind BodyKind = Sema::FnBodyKind::Other;
   SourceLocation KWLoc;
   if (TryConsumeToken(tok::equal)) {
@@ -1415,6 +1416,7 @@ Decl *Parser::ParseFunctionDefinition(ParsingDeclarator &D,
                       : diag::ext_defaulted_deleted_function)
           << 1 /* deleted */;
       BodyKind = Sema::FnBodyKind::Delete;
+      DeletedMessage = ParseCXXDeletedFunctionMessage();
     } else if (TryConsumeToken(tok::kw_default, KWLoc)) {
       Diag(KWLoc, getLangOpts().CPlusPlus11
                       ? diag::warn_cxx98_compat_defaulted_deleted_function
@@ -1473,7 +1475,7 @@ Decl *Parser::ParseFunctionDefinition(ParsingDeclarator &D,
   D.getMutableDeclSpec().abort();
 
   if (BodyKind != Sema::FnBodyKind::Other) {
-    Actions.SetFunctionBodyKind(Res, KWLoc, BodyKind);
+    Actions.SetFunctionBodyKind(Res, KWLoc, BodyKind, DeletedMessage);
     Stmt *GeneratedBody = Res ? Res->getBody() : nullptr;
     Actions.ActOnFinishFunctionBody(Res, GeneratedBody, false);
     return Res;
diff --git a/clang/lib/Sema/SemaCast.cpp b/clang/lib/Sema/SemaCast.cpp
index 9d85568d97b2d..aac43e2cad603 100644
--- a/clang/lib/Sema/SemaCast.cpp
+++ b/clang/lib/Sema/SemaCast.cpp
@@ -498,10 +498,22 @@ static bool tryDiagnoseOverloadedCast(Sema &S, CastType CT,
     howManyCandidates = OCD_AmbiguousCandidates;
     break;
 
-  case OR_Deleted:
-    msg = diag::err_ovl_deleted_conversion_in_cast;
-    howManyCandidates = OCD_ViableCandidates;
-    break;
+  case OR_Deleted: {
+    OverloadCandidateSet::iterator Best;
+    OverloadingResult Res =
+        candidates.BestViableFunction(S, range.getBegin(), Best);
+    assert(Res == OR_Deleted && "Inconsistent overload resolution");
+
+    StringLiteral *Msg = Best->Function->getDeletedMessage();
+    candidates.NoteCandidates(
+        PartialDiagnosticAt(range.getBegin(),
+                            S.PDiag(diag::err_ovl_deleted_conversion_in_cast)
+                                << CT << srcType << destType << (Msg != nullptr)
+                                << (Msg ? Msg->getString() : StringRef())
+                                << range << src->getSourceRange()),
+        S, OCD_ViableCandidates, src);
+    return true;
+  }
   }
 
   candidates.NoteCandidates(
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 17032d1370521..8b3b9d020db57 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -16083,7 +16083,17 @@ Decl *Sema::ActOnFinishFunctionBody(Decl *dcl, Stmt *Body,
     // This is meant to pop the context added in ActOnStartOfFunctionDef().
     ExitFunctionBodyRAII ExitRAII(*this, isLambdaCallOperator(FD));
     if (FD) {
-      FD->setBody(Body);
+      // If this is called by Parser::ParseFunctionDefinition() after marking
+      // the declaration as deleted, and if the deleted-function-body contains
+      // a message (C++26), then a DefaultedOrDeletedInfo will have already been
+      // added to store that message; do not overwrite it in that case.
+      //
+      // Since this would always set the body to 'nullptr' in that case anyway,
+      // which is already done when the function decl is initially created,
+      // always skipping this irrespective of whether there is a delete message
+      // should not be a problem.
+      if (!FD->isDeletedAsWritten())
+        FD->setBody(Body);
       FD->setWillHaveBody(false);
       CheckImmediateEscalatingFunctionDefinition(FD, FSI);
 
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 1fe10375222c5..7669171fea56f 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -7982,7 +7982,7 @@ class DefaultedComparisonVisitor {
   DefaultedComparisonVisitor(Sema &S, CXXRecordDecl *RD, FunctionDecl *FD,
                              DefaultedComparisonKind DCK)
       : S(S), RD(RD), FD(FD), DCK(DCK) {
-    if (auto *Info = FD->getDefaultedFunctionInfo()) {
+    if (auto *Info = FD->getDefalutedOrDeletedInfo()) {
       // FIXME: Change CreateOverloadedBinOp to take an ArrayRef instead of an
       // UnresolvedSet to avoid this copy.
       Fns.assign(Info->getUnqualifiedLookups().begin(),
@@ -8850,8 +8850,9 @@ bool Sema::CheckExplicitlyDefaultedComparison(Scope *S, FunctionDecl *FD,
     UnresolvedSet<32> Operators;
     lookupOperatorsForDefaultedComparison(*this, S, Operators,
                                           FD->getOverloadedOperator());
-    FD->setDefaultedFunctionInfo(FunctionDecl::DefaultedFunctionInfo::Create(
-        Context, Operators.pairs()));
+    FD->setDefaultedOrDeletedInfo(
+        FunctionDecl::DefaultedOrDeletedFunctionInfo::Create(
+            Context, Operators.pairs()));
   }
 
   // C++2a [class.compare.default]p1:
@@ -18159,7 +18160,8 @@ NamedDecl *Sema::ActOnFriendFunctionDecl(Scope *S, Declarator &D,
   return ND;
 }
 
-void Sema::SetDeclDeleted(Decl *Dcl, SourceLocation DelLoc) {
+void Sema::SetDeclDeleted(Decl *Dcl, SourceLocation DelLoc,
+                          StringLiteral *Message) {
   AdjustDeclIfTemplate(Dcl);
 
   FunctionDecl *Fn = dyn_cast_or_null<FunctionDecl>(Dcl);
@@ -18208,7 +18210,7 @@ void Sema::SetDeclDeleted(Decl *Dcl, SourceLocation DelLoc) {
   // C++11 [dcl.fct.def.delete]p4:
   //  A deleted function is implicitly inline.
   Fn->setImplicitlyInline();
-  Fn->setDeletedAsWritten();
+  Fn->setDeletedAsWritten(true, Message);
 }
 
 void Sema::SetDeclDefaulted(Decl *Dcl, SourceLocation DefaultLoc) {
@@ -18321,11 +18323,11 @@ void Sema::DiagnoseReturnInConstructorExceptionHandler(CXXTryStmt *TryBlock) {
   }
 }
 
-void Sema::SetFunctionBodyKind(Decl *D, SourceLocation Loc,
-                               FnBodyKind BodyKind) {
+void Sema::SetFunctionBodyKind(Decl *D, SourceLocation Loc, FnBodyKind BodyKind,
+                               StringLiteral *DeletedMessage) {
   switch (BodyKind) {
   case FnBodyKind::Delete:
-    SetDeclDeleted(D, Loc);
+    SetDeclDeleted(D, Loc, DeletedMessage);
     break;
   case FnBodyKind::Default:
     SetDeclDefaulted(D, Loc);
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 823bf36d88bc9..505d068ac42eb 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -273,8 +273,11 @@ bool Sema::DiagnoseUseOfDecl(NamedDecl *D, ArrayRef<SourceLocation> Locs,
         Diag(Loc, diag::err_deleted_inherited_ctor_use)
             << Ctor->getParent()
             << Ctor->getInheritedConstructor().getConstructor()->getParent();
-      else
-        Diag(Loc, diag::err_deleted_function_use);
+      else {
+        StringLiteral *Msg = FD->getDeletedMessage();
+        Diag(Loc, diag::err_deleted_function_use)
+            << (Msg != nullptr) << (Msg ? Msg->getString() : StringRef());
+      }
       NoteDeletedFunction(FD);
       return true;
     }
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index d2f18d4aa9b98..25f23a3abf171 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -2716,13 +2716,9 @@ static bool resolveAllocationOverload(
     return true;
 
   case OR_Deleted: {
-    if (Diagnose) {
-      Candidates.NoteCandidates(
-          PartialDiagnosticAt(R.getNameLoc(),
-                              S.PDiag(diag::err_ovl_deleted_call)
-                                  << R.getLookupName() << Range),
-          S, OCD_AllCandidates, Args);
-    }
+    if (Diagnose)
+      S.DiagnoseUseOfDeletedFunction(R.getNameLoc(), Range, R.getLookupName(),
+                                     Candidates, Best->Function, Args);
     return true;
   }
   }
@@ -3376,7 +3372,9 @@ bool Sema::FindDeallocationFunction(SourceLocation StartLoc, CXXRecordDecl *RD,
     // FIXME: DiagnoseUseOfDecl?
     if (Operator->isDeleted()) {
       if (Diagnose) {
-        Diag(StartLoc, diag::err_deleted_function_use);
+        StringLiteral *Msg = Operator->getDeletedMessage();
+        Diag(StartLoc, diag::err_deleted_function_use)
+            << (Msg != nullptr) << (Msg ? Msg->getString() : StringRef());
         NoteDeletedFunction(Operator);
       }
       return true;
@@ -3980,14 +3978,11 @@ static bool resolveBuiltinNewDeleteOverload(Sema &S, CallExpr *TheCall,
         S, OCD_AmbiguousCandidates, Args);
     return true;
 
-  case OR_Deleted: {
-    Candidates.NoteCandidates(
-        PartialDiagnosticAt(R.getNameLoc(), S.PDiag(diag::err_ovl_deleted_call)
-                                                << R.getLookupName() << Range),
-        S, OCD_AllCandidates, Args);
+  case OR_Deleted:
+    S.DiagnoseUseOfDeletedFunction(R.getNameLoc(), Range, R.getLookupName(),
+                                   Candidates, Best->Function, Args);
     return true;
   }
-  }
   llvm_unreachable("Unreachable, bad result from BestViableFunction");
 }
 
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index 6b8ce0f633a3e..fb7a80ab02846 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -9763,12 +9763,15 @@ bool InitializationSequence::Diagnose(Sema &S,
       break;
     }
     case OR_Deleted: {
-      S.Diag(Kind.getLocation(), diag::err_typecheck_deleted_function)
-        << OnlyArg->getType() << DestType.getNonReferenceType()
-        << Args[0]->getSourceRange();
       OverloadCandidateSet::iterator Best;
       OverloadingResult Ovl
         = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
+
+      StringLiteral *Msg = Best->Function->getDeletedMessage();
+      S.Diag(Kind.getLocation(), diag::err_typecheck_deleted_function)
+          << OnlyArg->getType() << DestType.getNonReferenceType()
+          << (Msg != nullptr) << (Msg ? Msg->getString() : StringRef())
+          << Args[0]->getSourceRange();
       if (Ovl == OR_Deleted) {
         S.NoteDeletedFunction(Best->Function);
       } else {
@@ -10027,9 +10030,12 @@ bool InitializationSequence::Diagnose(Sema &S,
               << llvm::to_underlying(
                      S.getSpecialMember(cast<CXXMethodDecl>(Best->Function)))
               << DestType << ArgsRange;
-        else
+        else {
+          StringLiteral *Msg = Best->Function->getDeletedMessage();
           S.Diag(Kind.getLocation(), diag::err_ovl_deleted_init)
-              << DestType << ArgsRange;
+              << DestType << (Msg != nullptr)
+              << (Msg ? Msg->getString() : StringRef()) << ArgsRange;
+        }
 
         S.NoteDeletedFunction(Best->Function);
         break;
@@ -11061,6 +11067,9 @@ QualType Sema::DeduceTemplateSpecializationFromInitializer(
   }
 
   case OR_Deleted: {
+    // FIXME: There are no tests for this diagnostic, and it doesn't seem
+    // like we ever get here; attempts to trigger this seem to yield a
+    // generic c'all to deleted function' diagnostic instead.
     Diag(Kind.getLocation(), diag::err_deduced_class_template_deleted)
       << TemplateName;
     NoteDeletedFunction(Best->Function);
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 397e7681828f3..675eb6ec05e78 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -14171,15 +14171,13 @@ static ExprResult FinishOverloadedCallExpr(Sema &SemaRef, Scope *S, Expr *Fn,
     break;
 
   case OR_Deleted: {
-    CandidateSet->NoteCandidates(
-        PartialDiagnosticAt(Fn->getBeginLoc(),
-                            SemaRef.PDiag(diag::err_ovl_deleted_call)
-                                << ULE->getName() << Fn->getSourceRange()),
-        SemaRef, OCD_AllCandidates, Args);
+    FunctionDecl *FDecl = (*Best)->Function;
+    SemaRef.DiagnoseUseOfDeletedFunction(Fn->getBeginLoc(),
+                                         Fn->getSourceRange(), ULE->getName(),
+                                         *CandidateSet, FDecl, Args);
 
     // We emitted an error for the unavailable/deleted function call but keep
     // the call in the AST.
-    FunctionDecl *FDecl = (*Best)->Function;
     ExprResult Res =
         SemaRef.FixOverloadedFunctionReference(Fn, (*Best)->FoundDecl, FDecl);
     if (Res.isInvalid())
@@ -14531,20 +14529,24 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
         UnaryOperator::getOpcodeStr(Opc), OpLoc);
     return ExprError();
 
-  case OR_Deleted:
+  case OR_Deleted: {
     // CreateOverloadedUnaryOp fills the first element of ArgsArray with the
     // object whose method was called. Later in NoteCandidates size of ArgsArray
     // is passed further and it eventually ends up compared to number of
     // function candidate parameters which never includes the object parameter,
     // so slice ArgsArray to make sure apples are compared to apples.
+    StringLiteral *Msg = Best->Function->getDeletedMessage();
     CandidateSet.NoteCandidates(
         PartialDiagnosticAt(OpLoc, PDiag(diag::err_ovl_deleted_oper)
                                        << UnaryOperator::getOpcodeStr(Opc)
+                                       << (Msg != nullptr)
+                                       << (Msg ? Msg->getString() : StringRef())
                                        << Input->getSourceRange()),
         *this, OCD_AllCandidates, ArgsArray.drop_front(),
         UnaryOperator::getOpcodeStr(Opc), OpLoc);
     return ExprError();
   }
+  }
 
   // Either we found no viable overloaded operator or we matched a
   // built-in operator. In either case, fall through to trying to
@@ -15061,7 +15063,7 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
           OpLoc);
       return ExprError();
 
-    case OR_Deleted:
+    case OR_Deleted: {
       if (isImplicitlyDeleted(Best->Function)) {
         FunctionDecl *DeletedFD = Best->Function;
         DefaultedFunctionKind DFK = getDefaultedFunctionKind(DeletedFD);
@@ -15080,16 +15082,20 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
         NoteDeletedFunction(DeletedFD);
         return ExprError();
       }
+
+      StringLiteral *Msg = Best->Function->getDeletedMessage();
       CandidateSet.NoteCandidates(
           PartialDiagnosticAt(
-              OpLoc, PDiag(diag::err_ovl_deleted_oper)
-                         << getOperatorSpelling(Best->Function->getDeclName()
-                                                    .getCXXOverloadedOperator())
-                         << Args[0]->getSourceRange()
-                         << Args[1]->getSourceRange()),
+              OpLoc,
+              PDiag(diag::err_ovl_deleted_oper)
+                  << getOperatorSpelling(Best->Function->getDeclName()
+                                             .getCXXOverloadedOperator())
+                  << (Msg != nullptr) << (Msg ? Msg->getString() : StringRef())
+                  << Args[0]->getSourceRange() << Args[1]->getSourceRange()),
           *this, OCD_AllCandidates, Args, BinaryOperator::getOpcodeStr(Opc),
           OpLoc);
       return ExprError();
+    }
   }
 
   // We matched a built-in operator; build it.
@@ -15401,14 +15407,18 @@ ExprResult Sema::CreateOverloadedArraySubscriptExpr(SourceLocation LLoc,
       }
       return ExprError();
 
-    case OR_Deleted:
+    case OR_Deleted: {
+      StringLiteral *Msg = Best->Function->getDeletedMessage();
       CandidateSet.NoteCandidates(
-          PartialDiagnosticAt(LLoc, PDiag(diag::err_ovl_deleted_oper)
-                                        << "[]" << Args[0]->getSourceRange()
-                                        << Range),
+          PartialDiagnosticAt(LLoc,
+                              PDiag(diag::err_ovl_deleted_oper)
+                                  << "[]" << (Msg != nullptr)
+                                  << (Msg ? Msg->getString() : StringRef())
+                                  << Args[0]->getSourceRange() << Range),
           *this, OCD_AllCandidates, Args, "[]", LLoc);
       return ExprError();
     }
+    }
 
   // We matched a built-in operator; build it.
   return CreateBuiltinArraySubscriptExpr(Args[0], LLoc, Args[1], RLoc);
@@ -15622,11 +15632,9 @@ ExprResult Sema::BuildCallToMemberFunction(Scope *S, Expr *MemExprE,
           *this, OCD_AmbiguousCandidates, Args);
       break;
     case OR_Deleted:
-      CandidateSet.NoteCandidates(
-          PartialDiagnosticAt(UnresExpr->getMemberLoc(),
-                              PDiag(diag::err_ovl_deleted_member_call)
-                                  << DeclName << MemExprE->getSourceRange()),
-          *this, OCD_AllCandidates, Args);
+      DiagnoseUseOfDeletedFunction(
+          UnresExpr->getMemberLoc(), MemExprE->getSourceRange(), DeclName,
+          CandidateSet, Best->Function, Args, /*IsMember=*/true);
       break;
     }
     // Overload resolution fails, try to recover.
@@ -15890,15 +15898,21 @@ Sema::BuildCallToObjectOfClassType(Scope *S, Expr *Obj,
           *this, OCD_AmbiguousCandidates, Args);
     break;
 
-  case OR_Deleted:
+  case OR_Deleted: {
+    // FIXME: Is this diagnostic here really necessary? It seems that
+    //   1. we don't have any tests for this diagnostic, and
+    //   2. we already issue err_deleted_function_use for this later on anyway.
+    StringLiteral *Msg = Best->Function->getDeletedMessage();
     CandidateSet.NoteCandidates(
         PartialDiagnosticAt(Object.get()->getBeginLoc(),
                             PDiag(diag::err_ovl_deleted_object_call)
-                                << Object.get()->getType()
+                                << Object.get()->getType() << (Msg != nullptr)
+                                << (Msg ? Msg->getString() : StringRef())
                                 << Object.get()->getSourceRange()),
         *this, OCD_AllCandidates, Args);
     break;
   }
+  }
 
   if (Best == CandidateSet.end())
     return true;
@@ -16097,13 +16111,17 @@ Sema::BuildOverloadedArrowExpr(Scope *S, Expr *Base, SourceLocation OpLoc,
           *this, OCD_AmbiguousCandidates, Base);
     return ExprError();
 
-  case OR_Deleted:
+  case OR_Deleted: {
+    StringLiteral *Msg = Best->Function->getDeletedMessage();
     CandidateSet.NoteCandidates(
         PartialDiagnosticAt(OpLoc, PDiag(diag::err_ovl_deleted_oper)
-                                       << "->" << Base->getSourceRange()),
+                                       << "->" << (Msg != nullptr)
+                                       << (Msg ? Msg->getString() : StringRef())
+                                       << Base->getSourceRange()),
         *this, OCD_AllCandidates, Base);
     return ExprError();
   }
+  }
 
   CheckMemberOperatorAccess(OpLoc, Base, nullptr, Best->FoundDecl);
 
@@ -16517,3 +16535,17 @@ bool clang::shouldEnforceArgLimit(bool PartialOverloading,
         return false;
   return true;
 }
+
+void Sema::DiagnoseUseOfDeletedFunction(SourceLocation Loc, SourceRange Range,
+                                        DeclarationName Name,
+                                        OverloadCandidateSet &CandidateSet,
+                                        FunctionDecl *Fn, MultiExprArg Args,
+                                        bool IsMember) {
+  StringLiteral *Msg = Fn->getDeletedMessage();
+  CandidateSet.NoteCandidates(
+      PartialDiagnosticAt(Loc, PDiag(diag::err_ovl_deleted_call)
+                                   << IsMember << Name << (Msg != nullptr)
+                                   << (Msg ? Msg->getString() : StringRef())
+                                   << Range),
+      *this, OCD_AllCandidates, Args);
+}
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index c0469a47ab8b6..21a139e434cb7 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -2439,7 +2439,7 @@ Decl *TemplateDeclInstantiator::VisitFunctionDecl(
       return nullptr;
   }
   if (D->isDeleted())
-    SemaRef.SetDeclDeleted(Function, D->getLocation());
+    SemaRef.SetDeclDeleted(Function, D->getLocation(), D->getDeletedMessage());
 
   NamedDecl *PrincipalDecl =
       (TemplateParams ? cast<NamedDecl>(FunctionTemplate) : Function);
@@ -2815,7 +2815,8 @@ Decl *TemplateDeclInstantiator::VisitCXXMethodDecl(
       return nullptr;
   }
   if (D->isDeletedAsWritten())
-    SemaRef.SetDeclDeleted(Method, Method->getLocation());
+    SemaRef.SetDeclDeleted(Method, Method->getLocation(),
+                           D->getDeletedMessage());
 
   // If this is an explicit specialization, mark the implicitly-instantiated
   // template specialization as being an explicit specialization too.
@@ -4867,7 +4868,7 @@ TemplateDeclInstantiator::InitMethodInstantiation(CXXMethodDecl *New,
 bool TemplateDeclInstantiator::SubstDefaultedFunction(FunctionDecl *New,
                                                       FunctionDecl *Tmpl) {
   // Transfer across any unqualified lookups.
-  if (auto *DFI = Tmpl->getDefaultedFunctionInfo()) {
+  if (auto *DFI = Tmpl->getDefalutedOrDeletedInfo()) {
     SmallVector<DeclAccessPair, 32> Lookups;
     Lookups.reserve(DFI->getUnqualifiedLookups().size());
     bool AnyChanged = false;
@@ -4882,8 +4883,8 @@ bool TemplateDeclInstantiator::SubstDefaultedFunction(FunctionDecl *New,
 
     // It's unlikely that substitution will change any declarations. Don't
     // store an unnecessary copy in that case.
-    New->setDefaultedFunctionInfo(
-        AnyChanged ? FunctionDecl::DefaultedFunctionInfo::Create(
+    New->setDefaultedOrDeletedInfo(
+        AnyChanged ? FunctionDecl::DefaultedOrDeletedFunctionInfo::Create(
                          SemaRef.Context, Lookups)
                    : DFI);
   }
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index ff72fc30c718f..e4b6a75c118ba 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -1103,16 +1103,26 @@ void ASTDeclReader::VisitFunctionDecl(FunctionDecl *FD) {
     FD->setHasODRHash(true);
   }
 
-  if (FD->isDefaulted()) {
-    if (unsigned NumLookups = Record.readInt()) {
+  if (FD->isDefaulted() || FD->isDeletedAsWritten()) {
+    // If 'Info' is nonzero, we need to read an DefaultedOrDeletedInfo; if,
+    // additionally, the second bit is also set, we also need to read
+    // a DeletedMessage for the DefaultedOrDeletedInfo.
+    if (auto Info = Record.readInt()) {
+      bool HasMessage = Info & 2;
+      StringLiteral *DeletedMessage =
+          HasMessage ? cast<StringLiteral>(Record.readExpr()) : nullptr;
+
+      unsigned NumLookups = Record.readInt();
       SmallVector<DeclAccessPair, 8> Lookups;
       for (unsigned I = 0; I != NumLookups; ++I) {
         NamedDecl *ND = Record.readDeclAs<NamedDecl>();
         AccessSpecifier AS = (AccessSpecifier)Record.readInt();
         Lookups.push_back(DeclAccessPair::make(ND, AS));
       }
-      FD->setDefaultedFunctionInfo(FunctionDecl::DefaultedFunctionInfo::Create(
-          Reader.getContext(), Lookups));
+
+      FD->setDefaultedOrDeletedInfo(
+          FunctionDecl::DefaultedOrDeletedFunctionInfo::Create(
+              Reader.getContext(), Lookups, DeletedMessage));
     }
   }
 
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index 4b417909751fb..276b6257f1d84 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -749,8 +749,15 @@ void ASTDeclWriter::VisitFunctionDecl(FunctionDecl *D) {
   if (!ShouldSkipCheckingODR)
     Record.push_back(D->getODRHash());
 
-  if (D->isDefaulted()) {
-    if (auto *FDI = D->getDefaultedFunctionInfo()) {
+  if (D->isDefaulted() || D->isDeletedAsWritten()) {
+    if (auto *FDI = D->getDefalutedOrDeletedInfo()) {
+      // Store both that there is an DefaultedOrDeletedInfo and whether it
+      // contains a DeletedMessage.
+      StringLiteral *DeletedMessage = FDI->getDeletedMessage();
+      Record.push_back(1 | (DeletedMessage ? 2 : 0));
+      if (DeletedMessage)
+        Record.AddStmt(DeletedMessage);
+
       Record.push_back(FDI->getUnqualifiedLookups().size());
       for (DeclAccessPair P : FDI->getUnqualifiedLookups()) {
         Record.AddDeclRef(P.getDecl());
diff --git a/clang/test/AST/ast-dump-cxx2c-delete-with-message.cpp b/clang/test/AST/ast-dump-cxx2c-delete-with-message.cpp
new file mode 100644
index 0000000000000..ea16b97da23e4
--- /dev/null
+++ b/clang/test/AST/ast-dump-cxx2c-delete-with-message.cpp
@@ -0,0 +1,23 @@
+// Without serialization:
+// RUN: %clang_cc1 -ast-dump %s | FileCheck %s
+//
+// With serialization:
+// RUN: %clang_cc1 -emit-pch -o %t %s
+// RUN: %clang_cc1 -x c++ -include-pch %t -ast-dump-all  /dev/null | FileCheck %s
+
+struct S {
+  // CHECK:      CXXMethodDecl {{.*}} a 'void ()' delete
+  // CHECK-NEXT:   delete message: StringLiteral {{.*}} "foo"
+  void a() = delete("foo");
+
+  // CHECK:      FunctionTemplateDecl {{.*}} b
+  // CHECK-NEXT:   TemplateTypeParmDecl
+  // CHECK-NEXT:   CXXMethodDecl {{.*}} b 'void ()' delete
+  // CHECK-NEXT:     delete message: StringLiteral {{.*}} "bar"
+  template <typename>
+  void b() = delete("bar");
+};
+
+// CHECK:      FunctionDecl {{.*}} c 'void ()' delete
+// CHECK-NEXT:   delete message: StringLiteral {{.*}} "baz"
+void c() = delete("baz");
diff --git a/clang/test/AST/ast-print-cxx2c-delete-with-message.cpp b/clang/test/AST/ast-print-cxx2c-delete-with-message.cpp
new file mode 100644
index 0000000000000..11e037e4d7443
--- /dev/null
+++ b/clang/test/AST/ast-print-cxx2c-delete-with-message.cpp
@@ -0,0 +1,18 @@
+// Without serialization:
+// RUN: %clang_cc1 -ast-print %s | FileCheck %s
+//
+// With serialization:
+// RUN: %clang_cc1 -emit-pch -o %t %s
+// RUN: %clang_cc1 -x c++ -include-pch %t -ast-print  /dev/null | FileCheck %s
+
+// CHECK: struct S {
+struct S {
+  // CHECK-NEXT: void a() = delete("foo");
+  void a() = delete("foo");
+
+  // CHECK-NEXT: template <typename T> T b() = delete("bar");
+  template <typename T> T b() = delete("bar");
+};
+
+// CHECK: void c() = delete("baz");
+void c() = delete("baz");
diff --git a/clang/test/Lexer/cxx-features.cpp b/clang/test/Lexer/cxx-features.cpp
index 1de32498cd345..baaa9d4434e9b 100644
--- a/clang/test/Lexer/cxx-features.cpp
+++ b/clang/test/Lexer/cxx-features.cpp
@@ -34,6 +34,10 @@
 
 // --- C++26 features ---
 
+#if check(deleted_function, 202403, 202403, 202403, 202403, 202403, 202403, 202403)
+#error "wrong value for __cpp_deleted_function"
+#endif
+
 #if check(placeholder_variables, 202306, 202306, 202306, 202306, 202306, 202306, 202306)
 #error "wrong value for __cpp_placeholder_variables"
 #endif
diff --git a/clang/test/PCH/cxx2a-defaulted-comparison.cpp b/clang/test/PCH/cxx2a-defaulted-comparison.cpp
index 8aeb1683961af..bdf649919659d 100644
--- a/clang/test/PCH/cxx2a-defaulted-comparison.cpp
+++ b/clang/test/PCH/cxx2a-defaulted-comparison.cpp
@@ -22,7 +22,7 @@ namespace std {
   constexpr strong_ordering strong_ordering::less = {-1};
 }
 
-// Ensure that we can round-trip DefaultedFunctionInfo through an AST file.
+// Ensure that we can round-trip DefaultedOrDeletedInfo through an AST file.
 namespace LookupContext {
   struct A {};
 
diff --git a/clang/test/Parser/cxx2c-delete-with-message.cpp b/clang/test/Parser/cxx2c-delete-with-message.cpp
new file mode 100644
index 0000000000000..1767a080a7dcd
--- /dev/null
+++ b/clang/test/Parser/cxx2c-delete-with-message.cpp
@@ -0,0 +1,51 @@
+// RUN: %clang_cc1 -std=c++23 -fsyntax-only -verify=expected,pre26 -pedantic %s
+// RUN: %clang_cc1 -std=c++2c -fsyntax-only -verify=expected,compat -Wpre-c++26-compat %s
+// RUN: %clang_cc1 -std=c++2c -fsyntax-only -verify %s
+
+struct S {
+  void a() = delete;
+  void b() = delete(; // expected-error {{expected string literal}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  void c() = delete(); // expected-error {{expected string literal}}
+  void d() = delete(42); // expected-error {{expected string literal}}
+  void e() = delete("foo"[0]); // expected-error {{expected ')'}} expected-note {{to match this '('}} // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  void f() = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+
+  S() = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  ~S() = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  S(const S&) = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  S(S&&) = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  S& operator=(const S&) = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+  S& operator=(S&&) = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+};
+
+struct T {
+  T() = delete(); // expected-error {{expected string literal}}
+  ~T() = delete(); // expected-error {{expected string literal}}
+  T(const T&) = delete(); // expected-error {{expected string literal}}
+  T(T&&) = delete(); // expected-error {{expected string literal}}
+  T& operator=(const T&) = delete(); // expected-error {{expected string literal}}
+  T& operator=(T&&) = delete(); // expected-error {{expected string literal}}
+};
+
+void a() = delete;
+void b() = delete(; // expected-error {{expected string literal}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+void c() = delete(); // expected-error {{expected string literal}}
+void d() = delete(42); // expected-error {{expected string literal}}
+void e() = delete("foo"[0]); // expected-error {{expected ')'}} expected-note {{to match this '('}} // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+void f() = delete("foo"); // pre26-warning {{'= delete' with a message is a C++2c extension}} compat-warning {{'= delete' with a message is incompatible with C++ standards before C++2c}}
+
+constexpr const char *getMsg() { return "this is a message"; }
+void func() = delete(getMsg()); // expected-error {{expected string literal}}
+
+namespace CWG2876 {
+using T = void ();
+using U = int;
+
+T a = delete ("hello"); // expected-error {{only functions can have deleted definitions}}
+U b = delete ("hello"), c, d = delete ("hello"); // expected-error 2 {{only functions can have deleted definitions}}
+
+struct C {
+  T e = delete ("hello"); // expected-error {{'= delete' is a function definition and must occur in a standalone declaration}}
+  U f = delete ("hello"); // expected-error {{cannot delete expression of type 'const char[6]'}}
+};
+}
diff --git a/clang/test/SemaCXX/cxx2c-delete-with-message.cpp b/clang/test/SemaCXX/cxx2c-delete-with-message.cpp
new file mode 100644
index 0000000000000..22e65d902ecd4
--- /dev/null
+++ b/clang/test/SemaCXX/cxx2c-delete-with-message.cpp
@@ -0,0 +1,273 @@
+// RUN: %clang_cc1 -std=c++2c -fsyntax-only -verify %s
+
+struct S {
+  void f() = delete("deleted (1)"); // expected-note {{explicitly marked deleted}}
+
+  template <typename T>
+  T g() = delete("deleted (2)"); // expected-note {{explicitly deleted}}
+};
+
+template <typename T>
+struct TS {
+  T f() = delete("deleted (3)"); // expected-note {{explicitly marked deleted}}
+
+  template <typename U>
+  T g(U) = delete("deleted (4)"); // expected-note {{explicitly deleted}}
+};
+
+void f() = delete("deleted (5)"); // expected-note {{explicitly deleted}}
+
+template <typename T>
+T g() = delete("deleted (6)"); // expected-note {{explicitly deleted}}
+
+template <typename T, typename U>
+struct TSp {
+  T f() = delete("deleted (7)"); // expected-note 2 {{explicitly marked deleted}}
+};
+
+template <typename T>
+struct TSp<T, int> {
+  T f() = delete("deleted (8)"); // expected-note {{explicitly marked deleted}}
+};
+
+template <>
+struct TSp<int, int> {
+  int f() = delete("deleted (9)"); // expected-note {{explicitly marked deleted}}
+};
+
+void u1() = delete(L"\xFFFFFFFF"); // expected-error {{an unevaluated string literal cannot have an encoding prefix}} \
+                                   // expected-error {{invalid escape sequence '\xFFFFFFFF' in an unevaluated string literal}}
+void u2() = delete(u"\U000317FF"); // expected-error {{an unevaluated string literal cannot have an encoding prefix}}
+
+void u3() = delete("Ω"); // expected-note {{explicitly deleted}}
+void u4() = delete("\u1234"); // expected-note {{explicitly deleted}}
+
+void u5() = delete("\x1ff"       // expected-error {{hex escape sequence out of range}} \
+                                 // expected-error {{invalid escape sequence '\x1ff' in an unevaluated string literal}}
+                     "0\x123"    // expected-error {{invalid escape sequence '\x123' in an unevaluated string literal}}
+                     "fx\xfffff" // expected-error {{invalid escape sequence '\xfffff' in an unevaluated string literal}}
+                     "goop");
+
+void u6() = delete("\'\"\?\\\a\b\f\n\r\t\v"); // expected-note {{explicitly deleted}}
+void u7() = delete("\xFF"); // expected-error {{invalid escape sequence '\xFF' in an unevaluated string literal}}
+void u8() = delete("\123"); // expected-error {{invalid escape sequence '\123' in an unevaluated string literal}}
+void u9() = delete("\pOh no, a Pascal string!"); // expected-warning {{unknown escape sequence '\p'}} \
+                                                 // expected-error {{invalid escape sequence '\p' in an unevaluated string literal}}
+// expected-note@+1 {{explicitly deleted}}
+void u10() = delete(R"(a
+\tb
+c
+)");
+
+void u11() = delete("\u0080\u0081\u0082\u0083\u0099\u009A\u009B\u009C\u009D\u009E\u009F"); // expected-note {{explicitly deleted}}
+
+
+//! Contains RTL/LTR marks
+void u12() = delete("\u200Eabc\u200Fdef\u200Fgh"); // expected-note {{explicitly deleted}}
+
+//! Contains ZWJ/regional indicators
+void u13() = delete("🏳️‍🌈 🏴󠁧󠁢󠁥󠁮󠁧󠁿 🇪🇺"); // expected-note {{explicitly deleted}}
+
+void h() {
+  S{}.f(); // expected-error {{attempt to use a deleted function: deleted (1)}}
+  S{}.g<int>(); // expected-error {{call to deleted member function 'g': deleted (2)}}
+  TS<int>{}.f(); // expected-error {{attempt to use a deleted function: deleted (3)}}
+  TS<int>{}.g<int>(0); // expected-error {{call to deleted member function 'g': deleted (4)}}
+  f(); // expected-error {{call to deleted function 'f': deleted (5)}}
+  g<int>(); // expected-error {{call to deleted function 'g': deleted (6)}}
+  TSp<double, double>{}.f(); // expected-error {{attempt to use a deleted function: deleted (7)}}
+  TSp<int, double>{}.f(); // expected-error {{attempt to use a deleted function: deleted (7)}}
+  TSp<double, int>{}.f(); // expected-error {{attempt to use a deleted function: deleted (8)}}
+  TSp<int, int>{}.f(); // expected-error {{attempt to use a deleted function: deleted (9)}}
+  u3(); // expected-error {{call to deleted function 'u3': Ω}}
+  u4(); // expected-error {{call to deleted function 'u4': ሴ}}
+  u6(); // expected-error {{call to deleted function 'u6': '"?\<U+0007><U+0008>}}
+  u10(); // expected-error {{call to deleted function 'u10': a\n\tb\nc\n}}
+  u11(); // expected-error {{call to deleted function 'u11': <U+0080><U+0081><U+0082><U+0083><U+0099><U+009A><U+009B><U+009C><U+009D><U+009E><U+009F>}}
+  u12(); // expected-error {{call to deleted function 'u12': ‎abc‏def‏gh}}
+  u13(); // expected-error {{call to deleted function 'u13': 🏳️‍🌈 🏴󠁧󠁢󠁥󠁮󠁧󠁿 🇪🇺}}
+}
+
+struct C {
+  C() = delete("deleted (C, Constructor)"); // expected-note {{explicitly marked deleted}}
+  C(int) = delete("deleted (C, C(int))"); // expected-note {{explicitly marked deleted}}
+  C(const C&) = delete("deleted (C, Copy Constructor)"); // expected-note {{explicitly marked deleted}}
+  C(C&&) = delete("deleted (C, Move Constructor)"); // expected-note {{explicitly marked deleted}}
+  C& operator=(const C&) = delete("deleted (C, Copy Assignment)"); // expected-note 2 {{explicitly deleted}}
+  C& operator=(C&&) = delete("deleted (C, Move Assignment)"); // expected-note {{explicitly deleted}} expected-note {{not viable}}
+  ~C() = delete("deleted (C, Destructor)"); // expected-note {{explicitly marked deleted}}
+  void* operator new(__SIZE_TYPE__) = delete("deleted (C, New)"); // expected-note {{explicitly deleted}}
+  void operator delete(void*) = delete("deleted (C, Delete)"); // expected-note {{explicitly marked deleted}}
+};
+
+template <typename T, typename U>
+struct TC {
+  TC() = delete("deleted (TC, Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(int) = delete("deleted (TC, TC(int))"); // expected-note {{explicitly marked deleted}}
+  TC(const TC&) = delete("deleted (TC, Copy Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(TC&&) = delete("deleted (TC, Move Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC& operator=(const TC&) = delete("deleted (TC, Copy Assignment)"); // expected-note 2 {{explicitly deleted}}
+  TC& operator=(TC&&) = delete("deleted (TC, Move Assignment)");  // expected-note {{explicitly deleted}} expected-note {{not viable}}
+  ~TC() = delete("deleted (TC, Destructor)"); // expected-note {{explicitly marked deleted}}
+  void* operator new(__SIZE_TYPE__) = delete("deleted (TC, New)"); // expected-note {{explicitly deleted}}
+  void operator delete(void*) = delete("deleted (TC, Delete)"); // expected-note {{explicitly marked deleted}}
+};
+
+template <typename T>
+struct TC<T, int> {
+  TC() = delete("deleted (TC<T, int>, Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(int) = delete("deleted (TC<T, int>, TC(int))"); // expected-note {{explicitly marked deleted}}
+  TC(const TC&) = delete("deleted (TC<T, int>, Copy Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(TC&&) = delete("deleted (TC<T, int>, Move Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC& operator=(const TC&) = delete("deleted (TC<T, int>, Copy Assignment)"); // expected-note 2 {{explicitly deleted}}
+  TC& operator=(TC&&) = delete("deleted (TC<T, int>, Move Assignment)"); // expected-note {{explicitly deleted}} expected-note {{not viable}}
+  ~TC() = delete("deleted (TC<T, int>, Destructor)"); // expected-note {{explicitly marked deleted}}
+  void* operator new(__SIZE_TYPE__) = delete("deleted (TC<T, int>, New)"); // expected-note {{explicitly deleted}}
+  void operator delete(void*) = delete("deleted (TC<T, int>, Delete)"); // expected-note {{explicitly marked deleted}}
+};
+
+template <>
+struct TC<int, int> {
+  TC() = delete("deleted (TC<int, int>, Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(int) = delete("deleted (TC<int, int>, TC(int))"); // expected-note {{explicitly marked deleted}}
+  TC(const TC&) = delete("deleted (TC<int, int>, Copy Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC(TC&&) = delete("deleted (TC<int, int>, Move Constructor)"); // expected-note {{explicitly marked deleted}}
+  TC& operator=(const TC&) = delete("deleted (TC<int, int>, Copy Assignment)"); // expected-note 2 {{explicitly deleted}}
+  TC& operator=(TC&&) = delete("deleted (TC<int, int>, Move Assignment)"); // expected-note {{explicitly deleted}} expected-note {{not viable}}
+  ~TC() = delete("deleted (TC<int, int>, Destructor)"); // expected-note {{explicitly marked deleted}}
+  void* operator new(__SIZE_TYPE__) = delete("deleted (TC<int, int>, New)"); // expected-note {{explicitly deleted}}
+  void operator delete(void*) = delete("deleted (TC<int, int>, Delete)"); // expected-note {{explicitly marked deleted}}
+};
+
+void special_members(
+  C& c1,
+  C& c2,
+  TC<double, double>& tc1,
+  TC<double, double>& tc2,
+  TC<double, int>& tc_int1,
+  TC<double, int>& tc_int2,
+  TC<int, int>& tc_int_int1,
+  TC<int, int>& tc_int_int2
+) {
+  C{}; // expected-error {{call to deleted constructor of 'C': deleted (C, Constructor)}}
+  C{c1}; // expected-error {{call to deleted constructor of 'C': deleted (C, Copy Constructor)}}
+  C{static_cast<C&&>(c1)}; // expected-error {{call to deleted constructor of 'C': deleted (C, Move Constructor)}}
+  c1 = c2; // expected-error {{overload resolution selected deleted operator '=': deleted (C, Copy Assignment)}}
+  c1 = static_cast<C&&>(c2); // expected-error {{overload resolution selected deleted operator '=': deleted (C, Move Assignment)}}
+  c1.~C(); // expected-error {{attempt to use a deleted function: deleted (C, Destructor)}}
+  new C{}; // expected-error {{call to deleted function 'operator new': deleted (C, New)}}
+  delete &c2; // expected-error {{attempt to use a deleted function: deleted (C, Delete)}}
+
+  TC<double, double>{}; // expected-error {{call to deleted constructor of 'TC<double, double>': deleted (TC, Constructor)}}
+  TC<double, double>{tc1}; // expected-error {{call to deleted constructor of 'TC<double, double>': deleted (TC, Copy Constructor)}}
+  TC<double, double>{static_cast<TC<double, double>&&>(tc1)}; // expected-error {{call to deleted constructor of 'TC<double, double>': deleted (TC, Move Constructor)}}
+  tc1 = tc2; // expected-error {{overload resolution selected deleted operator '=': deleted (TC, Copy Assignment)}}
+  tc1 = static_cast<TC<double, double>&&>(tc2); // expected-error {{overload resolution selected deleted operator '=': deleted (TC, Move Assignment)}}
+  tc1.~TC(); // expected-error {{attempt to use a deleted function: deleted (TC, Destructor)}}
+  new TC<double, double>{}; // expected-error {{call to deleted function 'operator new': deleted (TC, New)}}
+  delete &tc2; // expected-error {{attempt to use a deleted function: deleted (TC, Delete)}}
+
+  TC<double, int>{}; // expected-error {{call to deleted constructor of 'TC<double, int>': deleted (TC<T, int>, Constructor)}}
+  TC<double, int>{tc_int1}; // expected-error {{call to deleted constructor of 'TC<double, int>': deleted (TC<T, int>, Copy Constructor)}}
+  TC<double, int>{static_cast<TC<double, int>&&>(tc_int1)}; // expected-error {{call to deleted constructor of 'TC<double, int>': deleted (TC<T, int>, Move Constructor)}}
+  tc_int1 = tc_int2; // expected-error {{overload resolution selected deleted operator '=': deleted (TC<T, int>, Copy Assignment)}}
+  tc_int1 = static_cast<TC<double, int>&&>(tc_int2); // expected-error {{overload resolution selected deleted operator '=': deleted (TC<T, int>, Move Assignment)}}
+  tc_int1.~TC(); // expected-error {{attempt to use a deleted function: deleted (TC<T, int>, Destructor)}}
+  new TC<double, int>{}; // expected-error {{call to deleted function 'operator new': deleted (TC<T, int>, New)}}
+  delete &tc_int2; // expected-error {{attempt to use a deleted function: deleted (TC<T, int>, Delete)}}
+
+  TC<int, int>{}; // expected-error {{call to deleted constructor of 'TC<int, int>': deleted (TC<int, int>, Constructor)}}
+  TC<int, int>{tc_int_int1}; // expected-error {{call to deleted constructor of 'TC<int, int>': deleted (TC<int, int>, Copy Constructor)}}
+  TC<int, int>{static_cast<TC<int, int>&&>(tc_int_int1)}; // expected-error {{call to deleted constructor of 'TC<int, int>': deleted (TC<int, int>, Move Constructor)}}
+  tc_int_int1 = tc_int_int2; // expected-error {{overload resolution selected deleted operator '=': deleted (TC<int, int>, Copy Assignment)}}
+  tc_int_int1 = static_cast<TC<int, int>&&>(tc_int_int2); // expected-error {{overload resolution selected deleted operator '=': deleted (TC<int, int>, Move Assignment)}}
+  tc_int_int1.~TC(); // expected-error {{attempt to use a deleted function: deleted (TC<int, int>, Destructor)}}
+  new TC<int, int>{}; // expected-error {{call to deleted function 'operator new': deleted (TC<int, int>, New)}}
+  delete &tc_int_int2; // expected-error {{attempt to use a deleted function: deleted (TC<int, int>, Delete)}}
+}
+
+C conv1() { return 1; } // expected-error {{conversion function from 'int' to 'C' invokes a deleted function: deleted (C, C(int))}}
+TC<double, double> conv2() { return 1; } // expected-error {{conversion function from 'int' to 'TC<double, double>' invokes a deleted function: deleted (TC, TC(int))}}
+TC<double, int> conv3() { return 1; } // expected-error {{conversion function from 'int' to 'TC<double, int>' invokes a deleted function: deleted (TC<T, int>, TC(int))}}
+TC<int, int> conv4() { return 1; } // expected-error {{conversion function from 'int' to 'TC<int, int>' invokes a deleted function: deleted (TC<int, int>, TC(int))}}
+
+struct O {
+  int x;
+  int operator+() = delete("deleted (O, +)"); // expected-note {{explicitly deleted}}
+  O* operator->() = delete("deleted (O, ->)"); // expected-note {{explicitly deleted}}
+  int operator-(O) = delete("deleted (O, -)"); // expected-note {{explicitly deleted}}
+  int operator[](O) = delete("deleted (O, [])"); // expected-note {{explicitly deleted}}
+  int operator()(O) = delete("deleted (O, ())"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+  explicit operator bool() = delete("deleted (O, operator bool)"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+};
+
+template <typename T, typename U>
+struct TO {
+  T x;
+  T operator+() = delete("deleted (TO, +)"); // expected-note {{explicitly deleted}}
+  T* operator->() = delete("deleted (TO, ->)"); // expected-note {{explicitly deleted}}
+  T operator-(TO) = delete("deleted (TO, -)"); // expected-note {{explicitly deleted}}
+  T operator[](TO) = delete("deleted (TO, [])"); // expected-note {{explicitly deleted}}
+  T operator()(TO) = delete("deleted (TO, ())"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+  explicit operator bool() = delete("deleted (TO, operator bool)"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+};
+
+template <typename T>
+struct TO<T, int> {
+  T x;
+  T operator+() = delete("deleted (TO<T, int>, +)"); // expected-note {{explicitly deleted}}
+  T* operator->() = delete("deleted (TO<T, int>, ->)"); // expected-note {{explicitly deleted}}
+  T operator-(TO) = delete("deleted (TO<T, int>, -)"); // expected-note {{explicitly deleted}}
+  T operator[](TO) = delete("deleted (TO<T, int>, [])"); // expected-note {{explicitly deleted}}
+  T operator()(TO) = delete("deleted (TO<T, int>, ())"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+  explicit operator bool() = delete("deleted (TO<T, int>, operator bool)"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+};
+
+template <>
+struct TO<int, int> {
+  int x;
+  int operator+() = delete("deleted (TO<int, int>, +)"); // expected-note {{explicitly deleted}}
+  int* operator->() = delete("deleted (TO<int, int>, ->)"); // expected-note {{explicitly deleted}}
+  int operator-(TO) = delete("deleted (TO<int, int>, -)"); // expected-note {{explicitly deleted}}
+  int operator[](TO) = delete("deleted (TO<int, int>, [])"); // expected-note {{explicitly deleted}}
+  int operator()(TO) = delete("deleted (TO<int, int>, ())"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+  explicit operator bool() = delete("deleted (TO<int, int>, operator bool)"); // expected-note {{explicitly marked deleted}} expected-note {{explicitly deleted}}
+};
+
+void operators() {
+  O o;
+  +o; // expected-error {{overload resolution selected deleted operator '+': deleted (O, +)}}
+  o->x; // expected-error {{overload resolution selected deleted operator '->': deleted (O, ->)}}
+  o - o; // expected-error {{overload resolution selected deleted operator '-': deleted (O, -)}}
+  o[o]; // expected-error {{overload resolution selected deleted operator '[]': deleted (O, [])}}
+  o(o); // expected-error {{call to deleted function call operator in type 'O': deleted (O, ())}} expected-error {{attempt to use a deleted function: deleted (O, ())}}
+  if (o) {} // expected-error {{attempt to use a deleted function: deleted (O, operator bool)}}
+  static_cast<bool>(o); // expected-error {{static_cast from 'O' to 'bool' uses deleted function: deleted (O, operator bool)}}
+
+  TO<double, double> to;
+  +to; // expected-error {{overload resolution selected deleted operator '+': deleted (TO, +)}}
+  to->x; // expected-error {{overload resolution selected deleted operator '->': deleted (TO, ->)}}
+  to - to; // expected-error {{overload resolution selected deleted operator '-': deleted (TO, -)}}
+  to[to]; // expected-error {{overload resolution selected deleted operator '[]': deleted (TO, [])}}
+  to(to); // expected-error {{call to deleted function call operator in type 'TO<double, double>': deleted (TO, ())}} expected-error {{attempt to use a deleted function: deleted (TO, ())}}
+  if (to) {} // expected-error {{attempt to use a deleted function: deleted (TO, operator bool)}}
+  static_cast<bool>(to); // expected-error {{static_cast from 'TO<double, double>' to 'bool' uses deleted function: deleted (TO, operator bool)}}
+
+  TO<double, int> to_int;
+  +to_int; // expected-error {{overload resolution selected deleted operator '+': deleted (TO<T, int>, +)}}
+  to_int->x; // expected-error {{overload resolution selected deleted operator '->': deleted (TO<T, int>, ->)}}
+  to_int - to_int; // expected-error {{overload resolution selected deleted operator '-': deleted (TO<T, int>, -)}}
+  to_int[to_int]; // expected-error {{overload resolution selected deleted operator '[]': deleted (TO<T, int>, [])}}
+  to_int(to_int); // expected-error {{call to deleted function call operator in type 'TO<double, int>': deleted (TO<T, int>, ())}} expected-error {{attempt to use a deleted function: deleted (TO<T, int>, ())}}
+  if (to_int) {} // expected-error {{attempt to use a deleted function: deleted (TO<T, int>, operator bool)}}
+  static_cast<bool>(to_int); // expected-error {{static_cast from 'TO<double, int>' to 'bool' uses deleted function: deleted (TO<T, int>, operator bool)}}
+
+  TO<int, int> to_int_int;
+  +to_int_int; // expected-error {{overload resolution selected deleted operator '+': deleted (TO<int, int>, +)}}
+  to_int_int->x; // expected-error {{overload resolution selected deleted operator '->': deleted (TO<int, int>, ->)}}
+  to_int_int - to_int_int; // expected-error {{overload resolution selected deleted operator '-': deleted (TO<int, int>, -)}}
+  to_int_int[to_int_int]; // expected-error {{overload resolution selected deleted operator '[]': deleted (TO<int, int>, [])}}
+  to_int_int(to_int_int); // expected-error {{call to deleted function call operator in type 'TO<int, int>': deleted (TO<int, int>, ())}} expected-error {{attempt to use a deleted function: deleted (TO<int, int>, ())}}
+  if (to_int_int) {} // expected-error {{attempt to use a deleted function: deleted (TO<int, int>, operator bool)}}
+  static_cast<bool>(to_int_int); // expected-error {{static_cast from 'TO<int, int>' to 'bool' uses deleted function: deleted (TO<int, int>, operator bool)}}
+};
diff --git a/clang/www/cxx_status.html b/clang/www/cxx_status.html
index 130148c7420fa..c233171e63c81 100755
--- a/clang/www/cxx_status.html
+++ b/clang/www/cxx_status.html
@@ -197,7 +197,7 @@ <h2 id="cxx26">C++2c implementation status</h2>
  <tr>
   <td><tt>= delete("should have a reason");</tt></td>
   <td><a href="https://wg21.link/P2573R2">P2573R2</a></td>
-  <td class="none" align="center">No</td>
+  <td class="unreleased" align="center">Clang 19</td>
  </tr>
  <tr>
   <td>Variadic friends</td>

From 8891fd5acbe441d24a1734aa144f3f3dca075620 Mon Sep 17 00:00:00 2001
From: Kai Sasaki <lewuathe@gmail.com>
Date: Sun, 14 Apr 2024 19:52:03 +0900
Subject: [PATCH 353/886] [mlir][complex] Fastmath flag support for
 complex.tanh (#88571)

---
 .../ComplexToStandard/ComplexToStandard.cpp   | 14 ++++++++------
 .../convert-to-standard.mlir                  | 19 +++++++++++++++++++
 2 files changed, 27 insertions(+), 6 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 0664b053fc9e6..49eb575212ffc 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -973,6 +973,7 @@ struct TanhOpConversion : public OpConversionPattern<complex::TanhOp> {
     auto loc = op.getLoc();
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
+    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
 
     // The hyperbolic tangent for complex number can be calculated as follows.
     // tanh(x + i * y) = (tanh(x) + i * tan(y)) / (1 + tanh(x) * tan(y))
@@ -981,17 +982,18 @@ struct TanhOpConversion : public OpConversionPattern<complex::TanhOp> {
         rewriter.create<complex::ReOp>(loc, elementType, adaptor.getComplex());
     Value imag =
         rewriter.create<complex::ImOp>(loc, elementType, adaptor.getComplex());
-    Value tanhA = rewriter.create<math::TanhOp>(loc, real);
-    Value cosB = rewriter.create<math::CosOp>(loc, imag);
-    Value sinB = rewriter.create<math::SinOp>(loc, imag);
-    Value tanB = rewriter.create<arith::DivFOp>(loc, sinB, cosB);
+    Value tanhA = rewriter.create<math::TanhOp>(loc, real, fmf);
+    Value cosB = rewriter.create<math::CosOp>(loc, imag, fmf);
+    Value sinB = rewriter.create<math::SinOp>(loc, imag, fmf);
+    Value tanB = rewriter.create<arith::DivFOp>(loc, sinB, cosB, fmf);
     Value numerator =
         rewriter.create<complex::CreateOp>(loc, type, tanhA, tanB);
     Value one = rewriter.create<arith::ConstantOp>(
         loc, elementType, rewriter.getFloatAttr(elementType, 1));
-    Value mul = rewriter.create<arith::MulFOp>(loc, tanhA, tanB);
+    Value mul = rewriter.create<arith::MulFOp>(loc, tanhA, tanB, fmf);
     Value denominator = rewriter.create<complex::CreateOp>(loc, type, one, mul);
-    rewriter.replaceOpWithNewOp<complex::DivOp>(op, numerator, denominator);
+    rewriter.replaceOpWithNewOp<complex::DivOp>(op, numerator, denominator,
+                                                fmf);
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index b22c1acacaea1..e0e7cdadd317d 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -2085,3 +2085,22 @@ func.func @complex_tan_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL_WITH_SPECIAL_CASES]], %[[RESULT_IMAG_WITH_SPECIAL_CASES]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
 
+
+// -----
+
+// CHECK-LABEL: func @complex_tanh_with_fmf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_tanh_with_fmf(%arg: complex<f32>) -> complex<f32> {
+  %tanh = complex.tanh %arg fastmath<nnan,contract> : complex<f32>
+  return %tanh : complex<f32>
+}
+// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[TANH_A:.*]] = math.tanh %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[COS_B:.*]] = math.cos %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN_B:.*]] = math.sin %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[TAN_B:.*]] = arith.divf %[[SIN_B]], %[[COS_B]] fastmath<nnan,contract> : f32
+// CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
+// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
+// CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] fastmath<nnan,contract> : f32
+// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
\ No newline at end of file

From 6c2cc8240e11721cac466cfce89bc0f87a5019be Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Sun, 14 Apr 2024 12:09:14 +0100
Subject: [PATCH 354/886] [AArch64] Improve cost of non-zero lane splats

This adds a cost for non-zero lane splats, which is not included by default in
SK_Broadcast but can be handled by aarch64 dup lane instruction.
---
 llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp | 5 ++++-
 llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll  | 6 +++---
 2 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index bd943de06b4b2..e80931a03f30b 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -3939,7 +3939,10 @@ InstructionCost AArch64TTIImpl::getShuffleCost(
       LT.second.getVectorNumElements() == Mask.size() &&
       (Kind == TTI::SK_PermuteTwoSrc || Kind == TTI::SK_PermuteSingleSrc) &&
       (isZIPMask(Mask, LT.second, Unused) ||
-       isUZPMask(Mask, LT.second, Unused)))
+       isUZPMask(Mask, LT.second, Unused) ||
+       // Check for non-zero lane splats
+       all_of(drop_begin(Mask),
+              [&Mask](int M) { return M < 0 || M == Mask[0]; })))
     return 1;
 
   if (Kind == TTI::SK_Broadcast || Kind == TTI::SK_Transpose ||
diff --git a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
index 6c45ebcb69f40..d67f056366104 100644
--- a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
@@ -361,9 +361,9 @@ define void @uzp() {
 define void @multipart() {
 ; CHECK-LABEL: 'multipart'
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16a = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16b = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16c = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16d = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16b = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16c = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16d = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32a = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32a4 = shufflevector <16 x i32> undef, <16 x i32> undef, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32idrev = shufflevector <16 x i32> undef, <16 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 31, i32 30, i32 29, i32 28>

From 9b832b726c9c9bb0672c5f0912f6f131e3e27a10 Mon Sep 17 00:00:00 2001
From: Hristo Hristov <hghristov.rmm@gmail.com>
Date: Sun, 14 Apr 2024 14:37:51 +0300
Subject: [PATCH 355/886] [libc++] Deprecated `shared_ptr` Atomic Access APIs
 as per P0718R2 & Implemented P2869R3: Remove Deprecated `shared_ptr` Atomic
 Access APIs from C++26 (#87111)

Implements https://wg21.link/P2869R4
Implements deprecations as per https://wg21.link/P0718R2
---
 libcxx/docs/ReleaseNotes/19.rst               |  4 +++
 libcxx/docs/Status/Cxx20.rst                  |  2 ++
 libcxx/docs/Status/Cxx20Papers.csv            |  2 +-
 libcxx/docs/Status/Cxx2cPapers.csv            |  2 +-
 libcxx/docs/UsingLibcxx.rst                   |  4 +++
 libcxx/include/__memory/shared_ptr.h          | 29 ++++++++++------
 libcxx/include/memory                         | 22 ++++++------
 libcxx/modules/std/memory.inc                 |  6 ++++
 ..._exchange_strong.depr_in_cxx20..verify.cpp | 32 +++++++++++++++++
 ...e_strong_explicit.depr_in_cxx20.verify.cpp | 34 +++++++++++++++++++
 ...re_exchange_weak.depr_in_cxx20..verify.cpp | 31 +++++++++++++++++
 ...ge_weak_explicit.depr_in_cxx20..verify.cpp | 34 +++++++++++++++++++
 .../atomic_exchange.depr_in_cxx20..verify.cpp | 29 ++++++++++++++++
 ...xchange_explicit.verify.depr_in_cxx20..cpp | 30 ++++++++++++++++
 ...mic_is_lock_free.depr_in_cxx20..verify.cpp | 28 +++++++++++++++
 .../atomic_load.depr_in_cxx20..verify.cpp     | 28 +++++++++++++++
 ...ic_load_explicit.depr_in_cxx20..verify.cpp | 29 ++++++++++++++++
 .../atomic_store.depr_in_cxx20..verify.cpp    | 28 +++++++++++++++
 ...c_store_explicit.depr_in_cxx20..verify.cpp | 29 ++++++++++++++++
 .../atomic_compare_exchange_strong.pass.cpp   |  6 ++--
 ..._compare_exchange_strong_explicit.pass.cpp |  6 ++--
 .../atomic_compare_exchange_weak.pass.cpp     |  6 ++--
 ...ic_compare_exchange_weak_explicit.pass.cpp |  6 ++--
 .../atomic_exchange.pass.cpp                  |  6 ++--
 .../atomic_exchange_explicit.pass.cpp         |  6 ++--
 .../atomic_is_lock_free.pass.cpp              |  6 ++--
 .../atomic_load.pass.cpp                      |  6 ++--
 .../atomic_load_explicit.pass.cpp             |  6 ++--
 .../atomic_store.pass.cpp                     |  6 ++--
 .../atomic_store_explicit.pass.cpp            |  6 ++--
 30 files changed, 423 insertions(+), 46 deletions(-)
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.depr_in_cxx20.verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange_explicit.verify.depr_in_cxx20..cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_is_lock_free.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load_explicit.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store.depr_in_cxx20..verify.cpp
 create mode 100644 libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store_explicit.depr_in_cxx20..verify.cpp

diff --git a/libcxx/docs/ReleaseNotes/19.rst b/libcxx/docs/ReleaseNotes/19.rst
index 390de789ddb64..e5db17daa4823 100644
--- a/libcxx/docs/ReleaseNotes/19.rst
+++ b/libcxx/docs/ReleaseNotes/19.rst
@@ -43,6 +43,7 @@ Implemented Papers
 - P2819R2 - Add ``tuple`` protocol to ``complex``
 - P2495R3 - Interfacing ``stringstream``\s with ``string_view``
 - P2867R2 - Remove Deprecated ``strstream``\s From C++26
+- P2869R4 - Remove Deprecated ``shared_ptr`` Atomic Access APIs from C++26
 - P2872R3 - Remove ``wstring_convert`` From C++26
 - P3142R0 - Printing Blank Lines with ``println`` (as DR against C++23)
 - P2302R4 - ``std::ranges::contains``
@@ -63,6 +64,9 @@ Improvements and New Features
 
 - The ``_LIBCPP_ENABLE_CXX26_REMOVED_STRSTREAM`` macro has been added to make the declarations in ``<strstream>`` available.
 
+- The ``_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS`` macro has been added to make the declarations in ``<memory>``
+  available.
+
 - The ``_LIBCPP_ENABLE_CXX26_REMOVED_WSTRING_CONVERT`` macro has been added to make the declarations in ``<locale>``
   available.
 
diff --git a/libcxx/docs/Status/Cxx20.rst b/libcxx/docs/Status/Cxx20.rst
index 2deb82547d631..23289dc6e5961 100644
--- a/libcxx/docs/Status/Cxx20.rst
+++ b/libcxx/docs/Status/Cxx20.rst
@@ -59,6 +59,8 @@ Paper Status
       * ``GPS clock`` not done
       * ``UTC clock`` not done
 
+   .. [#note-P0718] P0718: Implemented deprecation of ``shared_ptr`` atomic access APIs only.
+
 .. _issues-status-cxx20:
 
 Library Working Group Issues Status
diff --git a/libcxx/docs/Status/Cxx20Papers.csv b/libcxx/docs/Status/Cxx20Papers.csv
index 6e4678da28887..d31720b7576d7 100644
--- a/libcxx/docs/Status/Cxx20Papers.csv
+++ b/libcxx/docs/Status/Cxx20Papers.csv
@@ -12,7 +12,7 @@
 "`P0600R1 <https://wg21.link/P0600R1>`__","LWG","nodiscard in the Library","Albuquerque","|Complete|","16.0"
 "`P0616R0 <https://wg21.link/P0616R0>`__","LWG","de-pessimize legacy <numeric> algorithms with std::move","Albuquerque","|Complete|","12.0"
 "`P0653R2 <https://wg21.link/P0653R2>`__","LWG","Utility to convert a pointer to a raw pointer","Albuquerque","|Complete|","6.0"
-"`P0718R2 <https://wg21.link/P0718R2>`__","LWG","Atomic shared_ptr","Albuquerque","",""
+"`P0718R2 <https://wg21.link/P0718R2>`__","LWG","Atomic shared_ptr","Albuquerque","|Partial| [#note-P0718]_",""
 "`P0767R1 <https://wg21.link/P0767R1>`__","CWG","Deprecate POD","Albuquerque","|Complete|","7.0"
 "`P0768R1 <https://wg21.link/P0768R1>`__","CWG","Library Support for the Spaceship (Comparison) Operator","Albuquerque","|Complete|",""
 "`P0777R1 <https://wg21.link/P0777R1>`__","LWG","Treating Unnecessary ``decay``\ ","Albuquerque","|Complete|","7.0"
diff --git a/libcxx/docs/Status/Cxx2cPapers.csv b/libcxx/docs/Status/Cxx2cPapers.csv
index 8516070c72b8f..1b00a05ad3d6f 100644
--- a/libcxx/docs/Status/Cxx2cPapers.csv
+++ b/libcxx/docs/Status/Cxx2cPapers.csv
@@ -48,7 +48,7 @@
 "","","","","","",""
 "`P2875R4 <https://wg21.link/P2875R4>`__","LWG","Undeprecate ``polymorphic_allocator::destroy`` for C++26","Tokyo March 2024","|Complete|","15.0",""
 "`P2867R2 <https://wg21.link/P2867R2>`__","LWG","Remove Deprecated ``strstreams`` From C++26","Tokyo March 2024","|Complete|","19.0",""
-"`P2869R4 <https://wg21.link/P2869R4>`__","LWG","Remove Deprecated ``shared_ptr`` Atomic Access APIs from C++26","Tokyo March 2024","","",""
+"`P2869R4 <https://wg21.link/P2869R4>`__","LWG","Remove Deprecated ``shared_ptr`` Atomic Access APIs from C++26","Tokyo March 2024","|Complete|","19.0",""
 "`P2872R3 <https://wg21.link/P2872R3>`__","LWG","Remove ``wstring_convert`` From C++26","Tokyo March 2024","|Complete|","19.0",""
 "`P3107R5 <https://wg21.link/P3107R5>`__","LWG","Permit an efficient implementation of ``std::print``","Tokyo March 2024","","","|format| |DR|"
 "`P3142R0 <https://wg21.link/P3142R0>`__","LWG","Printing Blank Lines with ``println``","Tokyo March 2024","|Complete|","19.0","|format|"
diff --git a/libcxx/docs/UsingLibcxx.rst b/libcxx/docs/UsingLibcxx.rst
index bc7817d14d04d..c0e85ad4d5e24 100644
--- a/libcxx/docs/UsingLibcxx.rst
+++ b/libcxx/docs/UsingLibcxx.rst
@@ -277,10 +277,14 @@ C++26 Specific Configuration Macros
 **_LIBCPP_ENABLE_CXX26_REMOVED_STRSTREAM**:
   This macro is used to re-enable all named declarations in ``<strstream>``.
 
+**_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS**:
+  This macro is used to re-enable all ``shared_ptr`` atomic access APIs in ``<memory>``.
+
 **_LIBCPP_ENABLE_CXX26_REMOVED_WSTRING_CONVERT**:
   This macro is used to re-enable the ``wstring_convert`` and ``wbuffer_convert``
   in ``<locale>``.
 
+
 Libc++ Extensions
 =================
 
diff --git a/libcxx/include/__memory/shared_ptr.h b/libcxx/include/__memory/shared_ptr.h
index a8ff189df2aa5..992b1ba43f100 100644
--- a/libcxx/include/__memory/shared_ptr.h
+++ b/libcxx/include/__memory/shared_ptr.h
@@ -1580,13 +1580,15 @@ class _LIBCPP_EXPORTED_FROM_ABI __sp_mut {
 
 _LIBCPP_EXPORTED_FROM_ABI __sp_mut& __get_sp_mut(const void*);
 
+#  if _LIBCPP_STD_VER < 26 || defined(_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS)
+
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI bool atomic_is_lock_free(const shared_ptr<_Tp>*) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI bool atomic_is_lock_free(const shared_ptr<_Tp>*) {
   return false;
 }
 
 template <class _Tp>
-_LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_load(const shared_ptr<_Tp>* __p) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_load(const shared_ptr<_Tp>* __p) {
   __sp_mut& __m = std::__get_sp_mut(__p);
   __m.lock();
   shared_ptr<_Tp> __q = *__p;
@@ -1595,12 +1597,13 @@ _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_load(const shared_ptr<_Tp>* __p) {
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_load_explicit(const shared_ptr<_Tp>* __p, memory_order) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp>
+atomic_load_explicit(const shared_ptr<_Tp>* __p, memory_order) {
   return std::atomic_load(__p);
 }
 
 template <class _Tp>
-_LIBCPP_HIDE_FROM_ABI void atomic_store(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI void atomic_store(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r) {
   __sp_mut& __m = std::__get_sp_mut(__p);
   __m.lock();
   __p->swap(__r);
@@ -1608,12 +1611,14 @@ _LIBCPP_HIDE_FROM_ABI void atomic_store(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI void atomic_store_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r, memory_order) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI void
+atomic_store_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r, memory_order) {
   std::atomic_store(__p, __r);
 }
 
 template <class _Tp>
-_LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_exchange(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r) {
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp>
+atomic_exchange(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r) {
   __sp_mut& __m = std::__get_sp_mut(__p);
   __m.lock();
   __p->swap(__r);
@@ -1622,13 +1627,13 @@ _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp> atomic_exchange(shared_ptr<_Tp>* __p, shar
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp>
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI shared_ptr<_Tp>
 atomic_exchange_explicit(shared_ptr<_Tp>* __p, shared_ptr<_Tp> __r, memory_order) {
   return std::atomic_exchange(__p, __r);
 }
 
 template <class _Tp>
-_LIBCPP_HIDE_FROM_ABI bool
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI bool
 atomic_compare_exchange_strong(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w) {
   shared_ptr<_Tp> __temp;
   __sp_mut& __m = std::__get_sp_mut(__p);
@@ -1646,23 +1651,25 @@ atomic_compare_exchange_strong(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, share
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI bool
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI bool
 atomic_compare_exchange_weak(shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w) {
   return std::atomic_compare_exchange_strong(__p, __v, __w);
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI bool atomic_compare_exchange_strong_explicit(
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI bool atomic_compare_exchange_strong_explicit(
     shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w, memory_order, memory_order) {
   return std::atomic_compare_exchange_strong(__p, __v, __w);
 }
 
 template <class _Tp>
-inline _LIBCPP_HIDE_FROM_ABI bool atomic_compare_exchange_weak_explicit(
+_LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_HIDE_FROM_ABI bool atomic_compare_exchange_weak_explicit(
     shared_ptr<_Tp>* __p, shared_ptr<_Tp>* __v, shared_ptr<_Tp> __w, memory_order, memory_order) {
   return std::atomic_compare_exchange_weak(__p, __v, __w);
 }
 
+#  endif // _LIBCPP_STD_VER < 26 || defined(_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS)
+
 #endif // !defined(_LIBCPP_HAS_NO_THREADS)
 
 _LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/include/memory b/libcxx/include/memory
index a8c0264eb9eb7..30fadce760d75 100644
--- a/libcxx/include/memory
+++ b/libcxx/include/memory
@@ -830,34 +830,34 @@ public:
 };
 
 template<class T>
-    bool atomic_is_lock_free(const shared_ptr<T>* p);
+    bool atomic_is_lock_free(const shared_ptr<T>* p);                                     // Deprecated in C++20, removed in C++26
 template<class T>
-    shared_ptr<T> atomic_load(const shared_ptr<T>* p);
+    shared_ptr<T> atomic_load(const shared_ptr<T>* p);                                    // Deprecated in C++20, removed in C++26
 template<class T>
-    shared_ptr<T> atomic_load_explicit(const shared_ptr<T>* p, memory_order mo);
+    shared_ptr<T> atomic_load_explicit(const shared_ptr<T>* p, memory_order mo);          // Deprecated in C++20, removed in C++26
 template<class T>
-    void atomic_store(shared_ptr<T>* p, shared_ptr<T> r);
+    void atomic_store(shared_ptr<T>* p, shared_ptr<T> r);                                 // Deprecated in C++20, removed in C++26
 template<class T>
-    void atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);
+    void atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);       // Deprecated in C++20, removed in C++26
 template<class T>
-    shared_ptr<T> atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r);
+    shared_ptr<T> atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r);                     // Deprecated in C++20, removed in C++26
 template<class T>
     shared_ptr<T>
-    atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);
+    atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo);         // Deprecated in C++20, removed in C++26
 template<class T>
     bool
-    atomic_compare_exchange_weak(shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w);
+    atomic_compare_exchange_weak(shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w);    // Deprecated in C++20, removed in C++26
 template<class T>
     bool
-    atomic_compare_exchange_strong( shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w);
+    atomic_compare_exchange_strong( shared_ptr<T>* p, shared_ptr<T>* v, shared_ptr<T> w); // Deprecated in C++20, removed in C++26
 template<class T>
     bool
-    atomic_compare_exchange_weak_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+    atomic_compare_exchange_weak_explicit(shared_ptr<T>* p, shared_ptr<T>* v,             // Deprecated in C++20, removed in C++26
                                           shared_ptr<T> w, memory_order success,
                                           memory_order failure);
 template<class T>
     bool
-    atomic_compare_exchange_strong_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+    atomic_compare_exchange_strong_explicit(shared_ptr<T>* p, shared_ptr<T>* v,           // Deprecated in C++20, removed in C++26
                                             shared_ptr<T> w, memory_order success,
                                             memory_order failure);
 // Hash support
diff --git a/libcxx/modules/std/memory.inc b/libcxx/modules/std/memory.inc
index 56c621c0cf17f..b23c27707afd6 100644
--- a/libcxx/modules/std/memory.inc
+++ b/libcxx/modules/std/memory.inc
@@ -190,6 +190,9 @@ export namespace std {
   //  using std::inout_ptr;
 
 #ifndef _LIBCPP_HAS_NO_THREADS
+
+#  if _LIBCPP_STD_VER < 26 || defined(_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS)
+
   // [depr.util.smartptr.shared.atomic]
   using std::atomic_is_lock_free;
 
@@ -206,5 +209,8 @@ export namespace std {
   using std::atomic_compare_exchange_strong_explicit;
   using std::atomic_compare_exchange_weak;
   using std::atomic_compare_exchange_weak_explicit;
+
+#  endif // _LIBCPP_STD_VER < 26 || defined(_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS)
+
 #endif // _LIBCPP_HAS_NO_THREADS
 } // namespace std
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..3ca01cad451a2
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong.depr_in_cxx20..verify.cpp
@@ -0,0 +1,32 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// bool
+// atomic_compare_exchange_strong(shared_ptr<T>* p, shared_ptr<T>* v,
+//                                shared_ptr<T> w);                     // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> v(new int(3));
+  std::shared_ptr<int> w(new int(2));
+  // expected-warning@+1 {{'atomic_compare_exchange_strong<int>' is deprecated}}
+  std::ignore = std::atomic_compare_exchange_strong(&p, &v, w);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.depr_in_cxx20.verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.depr_in_cxx20.verify.cpp
new file mode 100644
index 0000000000000..93384772c8afc
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.depr_in_cxx20.verify.cpp
@@ -0,0 +1,34 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// bool
+// atomic_compare_exchange_strong_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+//                                         shared_ptr<T> w, memory_order success,
+//                                         memory_order failure);                 // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> v(new int(3));
+  std::shared_ptr<int> w(new int(2));
+  // expected-warning@+2 {{'atomic_compare_exchange_strong_explicit<int>' is deprecated}}
+  std::ignore =
+      std::atomic_compare_exchange_strong_explicit(&p, &v, w, std::memory_order_seq_cst, std::memory_order_seq_cst);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..30ca70523fb64
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak.depr_in_cxx20..verify.cpp
@@ -0,0 +1,31 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// bool
+// atomic_compare_exchange_weak(shared_ptr<T>* p, shared_ptr<T>* v,
+//                              shared_ptr<T> w);                    // Deprecated in C++20, removed in C++26
+
+#include <memory>
+#include <tuple>
+
+int main(int, char**) {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> v(new int(3));
+  std::shared_ptr<int> w(new int(2));
+  // expected-warning@+1 {{'atomic_compare_exchange_weak<int>' is deprecated}}
+  std::ignore = std::atomic_compare_exchange_weak(&p, &v, w);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..ae6e98d0692d2
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.depr_in_cxx20..verify.cpp
@@ -0,0 +1,34 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// bool
+// atomic_compare_exchange_weak_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
+//                                       shared_ptr<T> w, memory_order success,
+//                                       memory_order failure);                 // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> v(new int(3));
+  std::shared_ptr<int> w(new int(2));
+  // expected-warning@+2 {{'atomic_compare_exchange_weak_explicit<int>' is deprecated}}
+  std::ignore =
+      std::atomic_compare_exchange_weak_explicit(&p, &v, w, std::memory_order_seq_cst, std::memory_order_seq_cst);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..0e4bc1bc50b18
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange.depr_in_cxx20..verify.cpp
@@ -0,0 +1,29 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// shared_ptr<T>
+// atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r)   // Deprecated in C++20, removed in C++26
+
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> r(new int(3));
+  // expected-warning@+1 {{'atomic_exchange<int>' is deprecated}}
+  std::ignore = std::atomic_exchange(&p, r);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange_explicit.verify.depr_in_cxx20..cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange_explicit.verify.depr_in_cxx20..cpp
new file mode 100644
index 0000000000000..5a6a2bfe4c1e2
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_exchange_explicit.verify.depr_in_cxx20..cpp
@@ -0,0 +1,30 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// shared_ptr<T>
+// atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r)  // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(4));
+  std::shared_ptr<int> r(new int(3));
+  // expected-warning@+1 {{'atomic_exchange_explicit<int>' is deprecated}}
+  std::ignore = std::atomic_exchange_explicit(&p, r, std::memory_order_seq_cst);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_is_lock_free.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_is_lock_free.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..6799d31ce1fa9
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_is_lock_free.depr_in_cxx20..verify.cpp
@@ -0,0 +1,28 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template<class T>
+// bool
+// atomic_is_lock_free(const shared_ptr<T>* p);    // Deprecated in C++20, removed in C++26
+
+#include <memory>
+#include <tuple>
+
+void test() {
+  const std::shared_ptr<int> p(new int(3));
+  // expected-warning@+1 {{'atomic_is_lock_free<int>' is deprecated}}
+  std::ignore = std::atomic_is_lock_free(&p);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..768ad0ded983e
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load.depr_in_cxx20..verify.cpp
@@ -0,0 +1,28 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// shared_ptr<T>
+// atomic_load(const shared_ptr<T>* p)      // Deprecated in C++20, removed in C++26
+
+#include <memory>
+#include <tuple>
+
+void test() {
+  std::shared_ptr<int> p(new int(3));
+  // expected-warning@+1 {{'atomic_load<int>' is deprecated}}
+  std::ignore = std::atomic_load(&p);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load_explicit.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load_explicit.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..dc56193324f04
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_load_explicit.depr_in_cxx20..verify.cpp
@@ -0,0 +1,29 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// shared_ptr<T>
+// atomic_load_explicit(const shared_ptr<T>* p, memory_order mo)     // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+#include <tuple>
+
+void test() {
+  const std::shared_ptr<int> p(new int(3));
+  // expected-warning@+1 {{'atomic_load_explicit<int>' is deprecated}}
+  std::ignore = std::atomic_load_explicit(&p, std::memory_order_relaxed);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..38b8df6f24393
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store.depr_in_cxx20..verify.cpp
@@ -0,0 +1,28 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// void
+// atomic_store(shared_ptr<T>* p, shared_ptr<T> r)   // Deprecated in C++20, removed in C++26
+
+#include <memory>
+
+void test() {
+  std::shared_ptr<int> p;
+  std::shared_ptr<int> r(new int(3));
+  // expected-warning@+1 {{'atomic_store<int>' is deprecated}}
+  std::atomic_store(&p, r);
+}
diff --git a/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store_explicit.depr_in_cxx20..verify.cpp b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store_explicit.depr_in_cxx20..verify.cpp
new file mode 100644
index 0000000000000..18f59c4618699
--- /dev/null
+++ b/libcxx/test/libcxx/depr/depr.util.smartptr.shared.atomic/atomic_store_explicit.depr_in_cxx20..verify.cpp
@@ -0,0 +1,29 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// UNSUPPORTED: no-threads
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+
+// <memory>
+
+// shared_ptr
+
+// template <class T>
+// void
+// atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo)  // Deprecated in C++20, removed in C++26
+
+#include <atomic>
+#include <memory>
+
+void test() {
+  std::shared_ptr<int> p;
+  std::shared_ptr<int> r(new int(3));
+  // expected-warning@+1 {{'atomic_store_explicit<int>' is deprecated}}
+  std::atomic_store_explicit(&p, r, std::memory_order_seq_cst);
+}
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong.pass.cpp
index 37be6ceea3e09..ad9d2a9d9ccc0 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -15,7 +17,7 @@
 // template <class T>
 // bool
 // atomic_compare_exchange_strong(shared_ptr<T>* p, shared_ptr<T>* v,
-//                                shared_ptr<T> w);
+//                                shared_ptr<T> w);                     // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.pass.cpp
index 3965863b86ccf..26496df96d472 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_strong_explicit.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -16,7 +18,7 @@
 // bool
 // atomic_compare_exchange_strong_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
 //                                         shared_ptr<T> w, memory_order success,
-//                                         memory_order failure);
+//                                         memory_order failure);                 // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak.pass.cpp
index 6dd04f924a105..5cdc0ea3312ff 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -15,7 +17,7 @@
 // template <class T>
 // bool
 // atomic_compare_exchange_weak(shared_ptr<T>* p, shared_ptr<T>* v,
-//                              shared_ptr<T> w);
+//                              shared_ptr<T> w);                    // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.pass.cpp
index 4837fa9793a5e..e4cc4cf77a64e 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_compare_exchange_weak_explicit.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -16,7 +18,7 @@
 // bool
 // atomic_compare_exchange_weak_explicit(shared_ptr<T>* p, shared_ptr<T>* v,
 //                                       shared_ptr<T> w, memory_order success,
-//                                       memory_order failure);
+//                                       memory_order failure);                 // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange.pass.cpp
index f488e0ed1d14e..db02c9d9f75b4 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // shared_ptr<T>
-// atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r)
+// atomic_exchange(shared_ptr<T>* p, shared_ptr<T> r)   // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange_explicit.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange_explicit.pass.cpp
index 1945f7bba6dc0..704232e71fa4a 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange_explicit.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_exchange_explicit.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // shared_ptr<T>
-// atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r)
+// atomic_exchange_explicit(shared_ptr<T>* p, shared_ptr<T> r)  // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_is_lock_free.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_is_lock_free.pass.cpp
index 37f7d12eda2c6..8f3b5bbb51558 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_is_lock_free.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_is_lock_free.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template<class T>
 // bool
-// atomic_is_lock_free(const shared_ptr<T>* p);
+// atomic_is_lock_free(const shared_ptr<T>* p);    // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load.pass.cpp
index 1b9a15ac92ac1..9d81e81b4392b 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // shared_ptr<T>
-// atomic_load(const shared_ptr<T>* p)
+// atomic_load(const shared_ptr<T>* p)      // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load_explicit.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load_explicit.pass.cpp
index 5c2970133328f..f4875dce38a81 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load_explicit.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_load_explicit.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // shared_ptr<T>
-// atomic_load_explicit(const shared_ptr<T>* p, memory_order mo)
+// atomic_load_explicit(const shared_ptr<T>* p, memory_order mo)     // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store.pass.cpp
index 5b7bd5fad69c4..836dcb16ff32e 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // void
-// atomic_store(shared_ptr<T>* p, shared_ptr<T> r)
+// atomic_store(shared_ptr<T>* p, shared_ptr<T> r)   // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 
diff --git a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store_explicit.pass.cpp b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store_explicit.pass.cpp
index 5712190421308..85b31c017d736 100644
--- a/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store_explicit.pass.cpp
+++ b/libcxx/test/std/utilities/memory/util.smartptr/util.smartptr.shared.atomic/atomic_store_explicit.pass.cpp
@@ -5,7 +5,9 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-//
+
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX26_REMOVED_SHARED_PTR_ATOMICS
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
 // UNSUPPORTED: no-threads
 
 // <memory>
@@ -14,7 +16,7 @@
 
 // template <class T>
 // void
-// atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo)
+// atomic_store_explicit(shared_ptr<T>* p, shared_ptr<T> r, memory_order mo)  // Deprecated in C++20, removed in C++26
 
 // UNSUPPORTED: c++03
 

From 6e934b7cfba8e0c167d2b287a30dfc1b720397c1 Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Sun, 14 Apr 2024 20:34:26 +0800
Subject: [PATCH 356/886] [clang] Fix -Wunused-variable in SemaCast.cpp (NFC)

llvm-project/clang/lib/Sema/SemaCast.cpp:503:23:
error: unused variable 'Res' [-Werror,-Wunused-variable]
    OverloadingResult Res =
                      ^
1 error generated.
---
 clang/lib/Sema/SemaCast.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/lib/Sema/SemaCast.cpp b/clang/lib/Sema/SemaCast.cpp
index aac43e2cad603..b0c28531fe873 100644
--- a/clang/lib/Sema/SemaCast.cpp
+++ b/clang/lib/Sema/SemaCast.cpp
@@ -500,7 +500,7 @@ static bool tryDiagnoseOverloadedCast(Sema &S, CastType CT,
 
   case OR_Deleted: {
     OverloadCandidateSet::iterator Best;
-    OverloadingResult Res =
+    [[maybe_unused]] OverloadingResult Res =
         candidates.BestViableFunction(S, range.getBegin(), Best);
     assert(Res == OR_Deleted && "Inconsistent overload resolution");
 

From 3c4d9559495e60dcbb431ab80d806096559a7486 Mon Sep 17 00:00:00 2001
From: Jacques Pienaar <jpienaar@google.com>
Date: Sun, 14 Apr 2024 06:36:40 -0700
Subject: [PATCH 357/886] [mlir] Add lib to tests for shared build (#88574)

These resulted in link failures:

```
/usr/bin/ld:
tools/mlir/test/CAPI/CMakeFiles/mlir-capi-translation-test.dir/translation.c.o:
in function `main':
translation.c:(.text.main+0x58): undefined reference to
`LLVMContextCreate'
/usr/bin/ld: translation.c:(.text.main+0x9b): undefined reference to
`LLVMDumpModule'
/usr/bin/ld: translation.c:(.text.main+0xa3): undefined reference to
`LLVMDisposeModule'
/usr/bin/ld: translation.c:(.text.main+0xb3): undefined reference to
`LLVMContextDispose'
```

Found in mlir-hs. Not sure why this hasn't been flagged elsewhere.
---
 mlir/test/CAPI/CMakeLists.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mlir/test/CAPI/CMakeLists.txt b/mlir/test/CAPI/CMakeLists.txt
index 79b61fdef38b4..b9cd63ef7c673 100644
--- a/mlir/test/CAPI/CMakeLists.txt
+++ b/mlir/test/CAPI/CMakeLists.txt
@@ -12,6 +12,7 @@ function(_add_capi_test_executable name)
   llvm_update_compile_flags(${name})
   if(MLIR_BUILD_MLIR_C_DYLIB)
     target_link_libraries(${name} PRIVATE
+      LLVMCore
       MLIR-C)
   else()
     target_link_libraries(${name} PRIVATE

From 7cfe73624780010ec81ca11c41ebbf214400abdd Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Sun, 14 Apr 2024 17:51:15 +0400
Subject: [PATCH 358/886] [clang] Fix name conflict with `sys/mac.h` on AIX
 (#88644)

Fixes clang-ppc64-aix bot failure after #88559
(0a6f6df5b0c3d0f2a42f013bf5cafb9b5020dcac)
https://lab.llvm.org/buildbot/#/builders/214/builds/11887

---------

Co-authored-by: Joseph Huber <huberjn@outlook.com>
---
 clang/include/clang/Basic/Cuda.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index acc6bb6581d85..38f30543a0f66 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -50,6 +50,10 @@ const char *CudaVersionToString(CudaVersion V);
 // Input is "Major.Minor"
 CudaVersion CudaStringToVersion(const llvm::Twine &S);
 
+// We have a name conflict with sys/mac.h on AIX
+#ifdef SM_32
+#undef SM_32
+#endif
 enum class CudaArch {
   UNUSED,
   UNKNOWN,

From d5c654b5b7c8bb81affdc69976ef9bc5ad5b4302 Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Sun, 14 Apr 2024 15:52:56 +0200
Subject: [PATCH 359/886] [libc++][RFC] Only include what is required
 by-version in the umbrella headers (#83740)

This is a relatively low cost way of reducing the include sizes in older
language modes compared to the effect. For example, in C++14 mode the
include time of `<algorithm>` is reduced from 198ms to 127ms.
---
 libcxx/include/algorithm                      | 259 +++++++++---------
 libcxx/include/bit                            |  36 ++-
 libcxx/include/charconv                       |  32 ++-
 libcxx/include/chrono                         |  75 ++---
 libcxx/include/compare                        |  36 ++-
 libcxx/include/concepts                       |  52 ++--
 libcxx/include/coroutine                      |   4 +
 libcxx/include/module.modulemap               |   5 +-
 libcxx/include/streambuf                      |   1 +
 .../libcxx/algorithms/half_positive.pass.cpp  |   4 +-
 libcxx/test/libcxx/numerics/bit.ops.pass.cpp  |   4 +-
 .../test/libcxx/time/convert_to_tm.pass.cpp   |   1 +
 .../test/libcxx/transitive_includes/cxx03.csv |  12 +-
 .../test/libcxx/transitive_includes/cxx11.csv |  13 +-
 .../test/libcxx/transitive_includes/cxx14.csv |  12 +
 .../test/libcxx/transitive_includes/cxx17.csv |  12 +
 .../test/libcxx/transitive_includes/cxx20.csv |  12 +
 .../test/libcxx/transitive_includes/cxx23.csv |  13 +
 .../test/libcxx/transitive_includes/cxx26.csv |  13 +
 19 files changed, 374 insertions(+), 222 deletions(-)

diff --git a/libcxx/include/algorithm b/libcxx/include/algorithm
index 0f62de7fa83f9..869fc19737b57 100644
--- a/libcxx/include/algorithm
+++ b/libcxx/include/algorithm
@@ -1794,15 +1794,11 @@ template <class BidirectionalIterator, class Compare>
 */
 
 #include <__config>
-#include <version>
 
 #include <__algorithm/adjacent_find.h>
 #include <__algorithm/all_of.h>
 #include <__algorithm/any_of.h>
 #include <__algorithm/binary_search.h>
-#include <__algorithm/clamp.h>
-#include <__algorithm/comp.h>
-#include <__algorithm/comp_ref_type.h>
 #include <__algorithm/copy.h>
 #include <__algorithm/copy_backward.h>
 #include <__algorithm/copy_if.h>
@@ -1818,18 +1814,9 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/find_first_of.h>
 #include <__algorithm/find_if.h>
 #include <__algorithm/find_if_not.h>
-#include <__algorithm/fold.h>
 #include <__algorithm/for_each.h>
-#include <__algorithm/for_each_n.h>
 #include <__algorithm/generate.h>
 #include <__algorithm/generate_n.h>
-#include <__algorithm/half_positive.h>
-#include <__algorithm/in_found_result.h>
-#include <__algorithm/in_fun_result.h>
-#include <__algorithm/in_in_out_result.h>
-#include <__algorithm/in_in_result.h>
-#include <__algorithm/in_out_out_result.h>
-#include <__algorithm/in_out_result.h>
 #include <__algorithm/includes.h>
 #include <__algorithm/inplace_merge.h>
 #include <__algorithm/is_heap.h>
@@ -1840,7 +1827,6 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/is_sorted_until.h>
 #include <__algorithm/iter_swap.h>
 #include <__algorithm/lexicographical_compare.h>
-#include <__algorithm/lexicographical_compare_three_way.h>
 #include <__algorithm/lower_bound.h>
 #include <__algorithm/make_heap.h>
 #include <__algorithm/max.h>
@@ -1848,7 +1834,6 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/merge.h>
 #include <__algorithm/min.h>
 #include <__algorithm/min_element.h>
-#include <__algorithm/min_max_result.h>
 #include <__algorithm/minmax.h>
 #include <__algorithm/minmax_element.h>
 #include <__algorithm/mismatch.h>
@@ -1864,112 +1849,7 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/partition_point.h>
 #include <__algorithm/pop_heap.h>
 #include <__algorithm/prev_permutation.h>
-#include <__algorithm/pstl_any_all_none_of.h>
-#include <__algorithm/pstl_copy.h>
-#include <__algorithm/pstl_count.h>
-#include <__algorithm/pstl_equal.h>
-#include <__algorithm/pstl_fill.h>
-#include <__algorithm/pstl_find.h>
-#include <__algorithm/pstl_for_each.h>
-#include <__algorithm/pstl_generate.h>
-#include <__algorithm/pstl_is_partitioned.h>
-#include <__algorithm/pstl_merge.h>
-#include <__algorithm/pstl_move.h>
-#include <__algorithm/pstl_replace.h>
-#include <__algorithm/pstl_rotate_copy.h>
-#include <__algorithm/pstl_sort.h>
-#include <__algorithm/pstl_stable_sort.h>
-#include <__algorithm/pstl_transform.h>
 #include <__algorithm/push_heap.h>
-#include <__algorithm/ranges_adjacent_find.h>
-#include <__algorithm/ranges_all_of.h>
-#include <__algorithm/ranges_any_of.h>
-#include <__algorithm/ranges_binary_search.h>
-#include <__algorithm/ranges_clamp.h>
-#include <__algorithm/ranges_contains.h>
-#include <__algorithm/ranges_contains_subrange.h>
-#include <__algorithm/ranges_copy.h>
-#include <__algorithm/ranges_copy_backward.h>
-#include <__algorithm/ranges_copy_if.h>
-#include <__algorithm/ranges_copy_n.h>
-#include <__algorithm/ranges_count.h>
-#include <__algorithm/ranges_count_if.h>
-#include <__algorithm/ranges_ends_with.h>
-#include <__algorithm/ranges_equal.h>
-#include <__algorithm/ranges_equal_range.h>
-#include <__algorithm/ranges_fill.h>
-#include <__algorithm/ranges_fill_n.h>
-#include <__algorithm/ranges_find.h>
-#include <__algorithm/ranges_find_end.h>
-#include <__algorithm/ranges_find_first_of.h>
-#include <__algorithm/ranges_find_if.h>
-#include <__algorithm/ranges_find_if_not.h>
-#include <__algorithm/ranges_for_each.h>
-#include <__algorithm/ranges_for_each_n.h>
-#include <__algorithm/ranges_generate.h>
-#include <__algorithm/ranges_generate_n.h>
-#include <__algorithm/ranges_includes.h>
-#include <__algorithm/ranges_inplace_merge.h>
-#include <__algorithm/ranges_is_heap.h>
-#include <__algorithm/ranges_is_heap_until.h>
-#include <__algorithm/ranges_is_partitioned.h>
-#include <__algorithm/ranges_is_permutation.h>
-#include <__algorithm/ranges_is_sorted.h>
-#include <__algorithm/ranges_is_sorted_until.h>
-#include <__algorithm/ranges_lexicographical_compare.h>
-#include <__algorithm/ranges_lower_bound.h>
-#include <__algorithm/ranges_make_heap.h>
-#include <__algorithm/ranges_max.h>
-#include <__algorithm/ranges_max_element.h>
-#include <__algorithm/ranges_merge.h>
-#include <__algorithm/ranges_min.h>
-#include <__algorithm/ranges_min_element.h>
-#include <__algorithm/ranges_minmax.h>
-#include <__algorithm/ranges_minmax_element.h>
-#include <__algorithm/ranges_mismatch.h>
-#include <__algorithm/ranges_move.h>
-#include <__algorithm/ranges_move_backward.h>
-#include <__algorithm/ranges_next_permutation.h>
-#include <__algorithm/ranges_none_of.h>
-#include <__algorithm/ranges_nth_element.h>
-#include <__algorithm/ranges_partial_sort.h>
-#include <__algorithm/ranges_partial_sort_copy.h>
-#include <__algorithm/ranges_partition.h>
-#include <__algorithm/ranges_partition_copy.h>
-#include <__algorithm/ranges_partition_point.h>
-#include <__algorithm/ranges_pop_heap.h>
-#include <__algorithm/ranges_prev_permutation.h>
-#include <__algorithm/ranges_push_heap.h>
-#include <__algorithm/ranges_remove.h>
-#include <__algorithm/ranges_remove_copy.h>
-#include <__algorithm/ranges_remove_copy_if.h>
-#include <__algorithm/ranges_remove_if.h>
-#include <__algorithm/ranges_replace.h>
-#include <__algorithm/ranges_replace_copy.h>
-#include <__algorithm/ranges_replace_copy_if.h>
-#include <__algorithm/ranges_replace_if.h>
-#include <__algorithm/ranges_reverse.h>
-#include <__algorithm/ranges_reverse_copy.h>
-#include <__algorithm/ranges_rotate.h>
-#include <__algorithm/ranges_rotate_copy.h>
-#include <__algorithm/ranges_sample.h>
-#include <__algorithm/ranges_search.h>
-#include <__algorithm/ranges_search_n.h>
-#include <__algorithm/ranges_set_difference.h>
-#include <__algorithm/ranges_set_intersection.h>
-#include <__algorithm/ranges_set_symmetric_difference.h>
-#include <__algorithm/ranges_set_union.h>
-#include <__algorithm/ranges_shuffle.h>
-#include <__algorithm/ranges_sort.h>
-#include <__algorithm/ranges_sort_heap.h>
-#include <__algorithm/ranges_stable_partition.h>
-#include <__algorithm/ranges_stable_sort.h>
-#include <__algorithm/ranges_starts_with.h>
-#include <__algorithm/ranges_swap_ranges.h>
-#include <__algorithm/ranges_transform.h>
-#include <__algorithm/ranges_unique.h>
-#include <__algorithm/ranges_unique_copy.h>
-#include <__algorithm/ranges_upper_bound.h>
 #include <__algorithm/remove.h>
 #include <__algorithm/remove_copy.h>
 #include <__algorithm/remove_copy_if.h>
@@ -1982,17 +1862,13 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/reverse_copy.h>
 #include <__algorithm/rotate.h>
 #include <__algorithm/rotate_copy.h>
-#include <__algorithm/sample.h>
 #include <__algorithm/search.h>
 #include <__algorithm/search_n.h>
 #include <__algorithm/set_difference.h>
 #include <__algorithm/set_intersection.h>
 #include <__algorithm/set_symmetric_difference.h>
 #include <__algorithm/set_union.h>
-#include <__algorithm/shift_left.h>
-#include <__algorithm/shift_right.h>
 #include <__algorithm/shuffle.h>
-#include <__algorithm/sift_down.h>
 #include <__algorithm/sort.h>
 #include <__algorithm/sort_heap.h>
 #include <__algorithm/stable_partition.h>
@@ -2001,9 +1877,138 @@ template <class BidirectionalIterator, class Compare>
 #include <__algorithm/transform.h>
 #include <__algorithm/unique.h>
 #include <__algorithm/unique_copy.h>
-#include <__algorithm/unwrap_iter.h>
 #include <__algorithm/upper_bound.h>
 
+#if _LIBCPP_STD_VER >= 17
+#  include <__algorithm/clamp.h>
+#  include <__algorithm/for_each_n.h>
+#  include <__algorithm/pstl_any_all_none_of.h>
+#  include <__algorithm/pstl_copy.h>
+#  include <__algorithm/pstl_count.h>
+#  include <__algorithm/pstl_equal.h>
+#  include <__algorithm/pstl_fill.h>
+#  include <__algorithm/pstl_find.h>
+#  include <__algorithm/pstl_for_each.h>
+#  include <__algorithm/pstl_generate.h>
+#  include <__algorithm/pstl_is_partitioned.h>
+#  include <__algorithm/pstl_merge.h>
+#  include <__algorithm/pstl_move.h>
+#  include <__algorithm/pstl_replace.h>
+#  include <__algorithm/pstl_rotate_copy.h>
+#  include <__algorithm/pstl_sort.h>
+#  include <__algorithm/pstl_stable_sort.h>
+#  include <__algorithm/pstl_transform.h>
+#  include <__algorithm/sample.h>
+#endif // _LIBCPP_STD_VER >= 17
+
+#if _LIBCPP_STD_VER >= 20
+#  include <__algorithm/in_found_result.h>
+#  include <__algorithm/in_fun_result.h>
+#  include <__algorithm/in_in_out_result.h>
+#  include <__algorithm/in_in_result.h>
+#  include <__algorithm/in_out_out_result.h>
+#  include <__algorithm/in_out_result.h>
+#  include <__algorithm/lexicographical_compare_three_way.h>
+#  include <__algorithm/min_max_result.h>
+#  include <__algorithm/ranges_adjacent_find.h>
+#  include <__algorithm/ranges_all_of.h>
+#  include <__algorithm/ranges_any_of.h>
+#  include <__algorithm/ranges_binary_search.h>
+#  include <__algorithm/ranges_clamp.h>
+#  include <__algorithm/ranges_contains.h>
+#  include <__algorithm/ranges_copy.h>
+#  include <__algorithm/ranges_copy_backward.h>
+#  include <__algorithm/ranges_copy_if.h>
+#  include <__algorithm/ranges_copy_n.h>
+#  include <__algorithm/ranges_count.h>
+#  include <__algorithm/ranges_count_if.h>
+#  include <__algorithm/ranges_equal.h>
+#  include <__algorithm/ranges_equal_range.h>
+#  include <__algorithm/ranges_fill.h>
+#  include <__algorithm/ranges_fill_n.h>
+#  include <__algorithm/ranges_find.h>
+#  include <__algorithm/ranges_find_end.h>
+#  include <__algorithm/ranges_find_first_of.h>
+#  include <__algorithm/ranges_find_if.h>
+#  include <__algorithm/ranges_find_if_not.h>
+#  include <__algorithm/ranges_for_each.h>
+#  include <__algorithm/ranges_for_each_n.h>
+#  include <__algorithm/ranges_generate.h>
+#  include <__algorithm/ranges_generate_n.h>
+#  include <__algorithm/ranges_includes.h>
+#  include <__algorithm/ranges_inplace_merge.h>
+#  include <__algorithm/ranges_is_heap.h>
+#  include <__algorithm/ranges_is_heap_until.h>
+#  include <__algorithm/ranges_is_partitioned.h>
+#  include <__algorithm/ranges_is_permutation.h>
+#  include <__algorithm/ranges_is_sorted.h>
+#  include <__algorithm/ranges_is_sorted_until.h>
+#  include <__algorithm/ranges_lexicographical_compare.h>
+#  include <__algorithm/ranges_lower_bound.h>
+#  include <__algorithm/ranges_make_heap.h>
+#  include <__algorithm/ranges_max.h>
+#  include <__algorithm/ranges_max_element.h>
+#  include <__algorithm/ranges_merge.h>
+#  include <__algorithm/ranges_min.h>
+#  include <__algorithm/ranges_min_element.h>
+#  include <__algorithm/ranges_minmax.h>
+#  include <__algorithm/ranges_minmax_element.h>
+#  include <__algorithm/ranges_mismatch.h>
+#  include <__algorithm/ranges_move.h>
+#  include <__algorithm/ranges_move_backward.h>
+#  include <__algorithm/ranges_next_permutation.h>
+#  include <__algorithm/ranges_none_of.h>
+#  include <__algorithm/ranges_nth_element.h>
+#  include <__algorithm/ranges_partial_sort.h>
+#  include <__algorithm/ranges_partial_sort_copy.h>
+#  include <__algorithm/ranges_partition.h>
+#  include <__algorithm/ranges_partition_copy.h>
+#  include <__algorithm/ranges_partition_point.h>
+#  include <__algorithm/ranges_pop_heap.h>
+#  include <__algorithm/ranges_prev_permutation.h>
+#  include <__algorithm/ranges_push_heap.h>
+#  include <__algorithm/ranges_remove.h>
+#  include <__algorithm/ranges_remove_copy.h>
+#  include <__algorithm/ranges_remove_copy_if.h>
+#  include <__algorithm/ranges_remove_if.h>
+#  include <__algorithm/ranges_replace.h>
+#  include <__algorithm/ranges_replace_copy.h>
+#  include <__algorithm/ranges_replace_copy_if.h>
+#  include <__algorithm/ranges_replace_if.h>
+#  include <__algorithm/ranges_reverse.h>
+#  include <__algorithm/ranges_reverse_copy.h>
+#  include <__algorithm/ranges_rotate.h>
+#  include <__algorithm/ranges_rotate_copy.h>
+#  include <__algorithm/ranges_sample.h>
+#  include <__algorithm/ranges_search.h>
+#  include <__algorithm/ranges_search_n.h>
+#  include <__algorithm/ranges_set_difference.h>
+#  include <__algorithm/ranges_set_intersection.h>
+#  include <__algorithm/ranges_set_symmetric_difference.h>
+#  include <__algorithm/ranges_set_union.h>
+#  include <__algorithm/ranges_shuffle.h>
+#  include <__algorithm/ranges_sort.h>
+#  include <__algorithm/ranges_sort_heap.h>
+#  include <__algorithm/ranges_stable_partition.h>
+#  include <__algorithm/ranges_stable_sort.h>
+#  include <__algorithm/ranges_swap_ranges.h>
+#  include <__algorithm/ranges_transform.h>
+#  include <__algorithm/ranges_unique.h>
+#  include <__algorithm/ranges_unique_copy.h>
+#  include <__algorithm/ranges_upper_bound.h>
+#  include <__algorithm/shift_left.h>
+#  include <__algorithm/shift_right.h>
+#endif
+
+#if _LIBCPP_STD_VER >= 23
+#  include <__algorithm/fold.h>
+#  include <__algorithm/ranges_contains_subrange.h>
+#  include <__algorithm/ranges_ends_with.h>
+#  include <__algorithm/ranges_starts_with.h>
+#endif // _LIBCPP_STD_VER >= 23
+
+#include <version>
+
 // standard-mandated includes
 
 // [algorithm.syn]
@@ -2013,6 +2018,10 @@ template <class BidirectionalIterator, class Compare>
 #  pragma GCC system_header
 #endif
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER == 14
+#  include <execution>
+#endif
+
 #if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 20
 #  include <atomic>
 #  include <bit>
diff --git a/libcxx/include/bit b/libcxx/include/bit
index b8e4bdc2dfe20..94387d101a398 100644
--- a/libcxx/include/bit
+++ b/libcxx/include/bit
@@ -61,26 +61,36 @@ namespace std {
 
 */
 
-#include <__bit/bit_cast.h>
-#include <__bit/bit_ceil.h>
-#include <__bit/bit_floor.h>
-#include <__bit/bit_log2.h>
-#include <__bit/bit_width.h>
-#include <__bit/blsr.h>
-#include <__bit/byteswap.h>
-#include <__bit/countl.h>
-#include <__bit/countr.h>
-#include <__bit/endian.h>
-#include <__bit/has_single_bit.h>
-#include <__bit/popcount.h>
-#include <__bit/rotate.h>
 #include <__config>
+
+#if _LIBCPP_STD_VER >= 20
+#  include <__bit/bit_cast.h>
+#  include <__bit/bit_ceil.h>
+#  include <__bit/bit_floor.h>
+#  include <__bit/bit_log2.h>
+#  include <__bit/bit_width.h>
+#  include <__bit/countl.h>
+#  include <__bit/countr.h>
+#  include <__bit/endian.h>
+#  include <__bit/has_single_bit.h>
+#  include <__bit/popcount.h>
+#  include <__bit/rotate.h>
+#endif
+
+#if _LIBCPP_STD_VER >= 23
+#  include <__bit/byteswap.h>
+#endif
+
 #include <version>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
 #endif
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 17
+#  include <cstdint>
+#endif
+
 #if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 20
 #  include <cstdlib>
 #  include <iosfwd>
diff --git a/libcxx/include/charconv b/libcxx/include/charconv
index 5bc7b9011be02..a2e270e9316dc 100644
--- a/libcxx/include/charconv
+++ b/libcxx/include/charconv
@@ -69,18 +69,21 @@ namespace std {
 
 */
 
-#include <__charconv/chars_format.h>
-#include <__charconv/from_chars_integral.h>
-#include <__charconv/from_chars_result.h>
-#include <__charconv/tables.h>
-#include <__charconv/to_chars.h>
-#include <__charconv/to_chars_base_10.h>
-#include <__charconv/to_chars_floating_point.h>
-#include <__charconv/to_chars_integral.h>
-#include <__charconv/to_chars_result.h>
-#include <__charconv/traits.h>
 #include <__config>
-#include <__system_error/errc.h>
+
+#if _LIBCPP_STD_VER >= 17
+#  include <__charconv/chars_format.h>
+#  include <__charconv/from_chars_integral.h>
+#  include <__charconv/from_chars_result.h>
+#  include <__charconv/tables.h>
+#  include <__charconv/to_chars.h>
+#  include <__charconv/to_chars_base_10.h>
+#  include <__charconv/to_chars_floating_point.h>
+#  include <__charconv/to_chars_integral.h>
+#  include <__charconv/to_chars_result.h>
+#  include <__charconv/traits.h>
+#endif // _LIBCPP_STD_VER >= 17
+
 #include <version>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
@@ -91,6 +94,13 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 
 _LIBCPP_END_NAMESPACE_STD
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 14
+#  include <cerrno>
+#  include <cstddef>
+#  include <initializer_list>
+#  include <new>
+#endif
+
 #if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 20
 #  include <cmath>
 #  include <concepts>
diff --git a/libcxx/include/chrono b/libcxx/include/chrono
index 5eddd050196de..513ae52006e89 100644
--- a/libcxx/include/chrono
+++ b/libcxx/include/chrono
@@ -880,28 +880,48 @@ constexpr chrono::year                                  operator ""y(unsigned lo
 
 // clang-format on
 
-#include <__chrono/calendar.h>
-#include <__chrono/convert_to_timespec.h>
-#include <__chrono/convert_to_tm.h>
-#include <__chrono/day.h>
+#include <__config>
+
 #include <__chrono/duration.h>
 #include <__chrono/file_clock.h>
-#include <__chrono/hh_mm_ss.h>
 #include <__chrono/high_resolution_clock.h>
-#include <__chrono/literals.h>
-#include <__chrono/month.h>
-#include <__chrono/month_weekday.h>
-#include <__chrono/monthday.h>
 #include <__chrono/steady_clock.h>
-#include <__chrono/sys_info.h>
 #include <__chrono/system_clock.h>
 #include <__chrono/time_point.h>
-#include <__chrono/weekday.h>
-#include <__chrono/year.h>
-#include <__chrono/year_month.h>
-#include <__chrono/year_month_day.h>
-#include <__chrono/year_month_weekday.h>
-#include <__config>
+
+#if _LIBCPP_STD_VER >= 20
+#  include <__chrono/calendar.h>
+#  include <__chrono/day.h>
+#  include <__chrono/hh_mm_ss.h>
+#  include <__chrono/literals.h>
+#  include <__chrono/month.h>
+#  include <__chrono/month_weekday.h>
+#  include <__chrono/monthday.h>
+#  include <__chrono/sys_info.h>
+#  include <__chrono/weekday.h>
+#  include <__chrono/year.h>
+#  include <__chrono/year_month.h>
+#  include <__chrono/year_month_day.h>
+#  include <__chrono/year_month_weekday.h>
+
+#  if !defined(_LIBCPP_HAS_NO_LOCALIZATION)
+#    include <__chrono/formatter.h>
+#    include <__chrono/ostream.h>
+#    include <__chrono/parser_std_format_spec.h>
+#    include <__chrono/statically_widen.h>
+#  endif
+
+#  if !defined(_LIBCPP_HAS_NO_TIME_ZONE_DATABASE) && !defined(_LIBCPP_HAS_NO_FILESYSTEM) &&                            \
+      !defined(_LIBCPP_HAS_NO_LOCALIZATION)
+#    include <__chrono/leap_second.h>
+#    include <__chrono/time_zone.h>
+#    include <__chrono/time_zone_link.h>
+#    include <__chrono/tzdb.h>
+#    include <__chrono/tzdb_list.h>
+#  endif
+
+#endif
+
 #include <version>
 
 // standard-mandated includes
@@ -909,26 +929,17 @@ constexpr chrono::year                                  operator ""y(unsigned lo
 // [time.syn]
 #include <compare>
 
-#if !defined(_LIBCPP_HAS_NO_LOCALIZATION) && _LIBCPP_STD_VER >= 20
-#  include <__chrono/formatter.h>
-#  include <__chrono/ostream.h>
-#  include <__chrono/parser_std_format_spec.h>
-#  include <__chrono/statically_widen.h>
-#endif
-
-#if !defined(_LIBCPP_HAS_NO_TIME_ZONE_DATABASE) && !defined(_LIBCPP_HAS_NO_FILESYSTEM) &&                              \
-    !defined(_LIBCPP_HAS_NO_LOCALIZATION)
-#  include <__chrono/leap_second.h>
-#  include <__chrono/time_zone.h>
-#  include <__chrono/time_zone_link.h>
-#  include <__chrono/tzdb.h>
-#  include <__chrono/tzdb_list.h>
-#endif
-
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
 #endif
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 17
+#  include <cstdint>
+#  include <stdexcept>
+#  include <string_view>
+#  include <vector>
+#endif
+
 #if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 20
 #  include <bit>
 #  include <concepts>
diff --git a/libcxx/include/compare b/libcxx/include/compare
index 93953254b7843..8a41835b14897 100644
--- a/libcxx/include/compare
+++ b/libcxx/include/compare
@@ -140,26 +140,36 @@ namespace std {
 }
 */
 
-#include <__compare/common_comparison_category.h>
-#include <__compare/compare_partial_order_fallback.h>
-#include <__compare/compare_strong_order_fallback.h>
-#include <__compare/compare_three_way.h>
-#include <__compare/compare_three_way_result.h>
-#include <__compare/compare_weak_order_fallback.h>
-#include <__compare/is_eq.h>
-#include <__compare/ordering.h>
-#include <__compare/partial_order.h>
-#include <__compare/strong_order.h>
-#include <__compare/synth_three_way.h>
-#include <__compare/three_way_comparable.h>
-#include <__compare/weak_order.h>
 #include <__config>
+
+#if _LIBCPP_STD_VER >= 20
+#  include <__compare/common_comparison_category.h>
+#  include <__compare/compare_partial_order_fallback.h>
+#  include <__compare/compare_strong_order_fallback.h>
+#  include <__compare/compare_three_way.h>
+#  include <__compare/compare_three_way_result.h>
+#  include <__compare/compare_weak_order_fallback.h>
+#  include <__compare/is_eq.h>
+#  include <__compare/ordering.h>
+#  include <__compare/partial_order.h>
+#  include <__compare/strong_order.h>
+#  include <__compare/synth_three_way.h>
+#  include <__compare/three_way_comparable.h>
+#  include <__compare/weak_order.h>
+#endif // _LIBCPP_STD_VER >= 20
+
 #include <version>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
 #endif
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 17
+#  include <cstddef>
+#  include <cstdint>
+#  include <limits>
+#endif
+
 #if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 20
 #  include <cmath>
 #  include <type_traits>
diff --git a/libcxx/include/concepts b/libcxx/include/concepts
index e10f5ab5ad8a1..e89d216a59372 100644
--- a/libcxx/include/concepts
+++ b/libcxx/include/concepts
@@ -129,31 +129,39 @@ namespace std {
 
 */
 
-#include <__concepts/arithmetic.h>
-#include <__concepts/assignable.h>
-#include <__concepts/boolean_testable.h>
-#include <__concepts/class_or_enum.h>
-#include <__concepts/common_reference_with.h>
-#include <__concepts/common_with.h>
-#include <__concepts/constructible.h>
-#include <__concepts/convertible_to.h>
-#include <__concepts/copyable.h>
-#include <__concepts/derived_from.h>
-#include <__concepts/destructible.h>
-#include <__concepts/different_from.h>
-#include <__concepts/equality_comparable.h>
-#include <__concepts/invocable.h>
-#include <__concepts/movable.h>
-#include <__concepts/predicate.h>
-#include <__concepts/regular.h>
-#include <__concepts/relation.h>
-#include <__concepts/same_as.h>
-#include <__concepts/semiregular.h>
-#include <__concepts/swappable.h>
-#include <__concepts/totally_ordered.h>
 #include <__config>
+
+#if _LIBCPP_STD_VER >= 20
+#  include <__concepts/arithmetic.h>
+#  include <__concepts/assignable.h>
+#  include <__concepts/boolean_testable.h>
+#  include <__concepts/class_or_enum.h>
+#  include <__concepts/common_reference_with.h>
+#  include <__concepts/common_with.h>
+#  include <__concepts/constructible.h>
+#  include <__concepts/convertible_to.h>
+#  include <__concepts/copyable.h>
+#  include <__concepts/derived_from.h>
+#  include <__concepts/destructible.h>
+#  include <__concepts/different_from.h>
+#  include <__concepts/equality_comparable.h>
+#  include <__concepts/invocable.h>
+#  include <__concepts/movable.h>
+#  include <__concepts/predicate.h>
+#  include <__concepts/regular.h>
+#  include <__concepts/relation.h>
+#  include <__concepts/same_as.h>
+#  include <__concepts/semiregular.h>
+#  include <__concepts/swappable.h>
+#  include <__concepts/totally_ordered.h>
+#endif // _LIBCPP_STD_VER >= 20
+
 #include <version>
 
+#if !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES) && _LIBCPP_STD_VER <= 17
+#  include <cstddef>
+#endif
+
 #if _LIBCPP_STD_VER <= 20 && !defined(_LIBCPP_REMOVE_TRANSITIVE_INCLUDES)
 #  include <type_traits>
 #endif
diff --git a/libcxx/include/coroutine b/libcxx/include/coroutine
index b1ba83b541b4b..d1a44e2e12dd4 100644
--- a/libcxx/include/coroutine
+++ b/libcxx/include/coroutine
@@ -39,10 +39,14 @@ struct suspend_always;
  */
 
 #include <__config>
+
+#if _LIBCPP_STD_VER >= 20
 #include <__coroutine/coroutine_handle.h>
 #include <__coroutine/coroutine_traits.h>
 #include <__coroutine/noop_coroutine_handle.h>
 #include <__coroutine/trivial_awaitables.h>
+#endif // _LIBCPP_STD_VER >= 20
+
 #include <version>
 
 // standard-mandated includes
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index 372219e8f8363..ccee7029824e3 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -1123,7 +1123,10 @@ module std_private_charconv_to_chars                [system] { header "__charcon
 module std_private_charconv_to_chars_base_10        [system] { header "__charconv/to_chars_base_10.h" }
 module std_private_charconv_to_chars_floating_point [system] { header "__charconv/to_chars_floating_point.h" }
 module std_private_charconv_to_chars_integral       [system] { header "__charconv/to_chars_integral.h" }
-module std_private_charconv_to_chars_result         [system] { header "__charconv/to_chars_result.h" }
+module std_private_charconv_to_chars_result         [system] {
+  header "__charconv/to_chars_result.h"
+  export *
+}
 module std_private_charconv_traits                  [system] { header "__charconv/traits.h" }
 
 module std_private_chrono_calendar               [system] { header "__chrono/calendar.h" }
diff --git a/libcxx/include/streambuf b/libcxx/include/streambuf
index aec537866c203..7964758c908f4 100644
--- a/libcxx/include/streambuf
+++ b/libcxx/include/streambuf
@@ -109,6 +109,7 @@ protected:
 
 #include <__config>
 #include <__fwd/streambuf.h>
+#include <__locale>
 #include <__type_traits/is_same.h>
 #include <climits>
 #include <ios>
diff --git a/libcxx/test/libcxx/algorithms/half_positive.pass.cpp b/libcxx/test/libcxx/algorithms/half_positive.pass.cpp
index 82d18cba37f34..40f4697419572 100644
--- a/libcxx/test/libcxx/algorithms/half_positive.pass.cpp
+++ b/libcxx/test/libcxx/algorithms/half_positive.pass.cpp
@@ -11,10 +11,10 @@
 // __half_positive divides an integer number by 2 as unsigned number for known types.
 // It can be an important optimization for lower bound, for example.
 
-#include <algorithm>
+#include <__algorithm/half_positive.h>
 #include <cassert>
+#include <cstddef>
 #include <limits>
-#include <type_traits>
 
 #include "test_macros.h"
 #include "user_defined_integral.h"
diff --git a/libcxx/test/libcxx/numerics/bit.ops.pass.cpp b/libcxx/test/libcxx/numerics/bit.ops.pass.cpp
index d3ca8b2f8030b..7f502d6e01d1e 100644
--- a/libcxx/test/libcxx/numerics/bit.ops.pass.cpp
+++ b/libcxx/test/libcxx/numerics/bit.ops.pass.cpp
@@ -9,7 +9,9 @@
 // Test the __XXXX routines in the <bit> header.
 // These are not supposed to be exhaustive tests, just sanity checks.
 
-#include <bit>
+#include <__bit/bit_log2.h>
+#include <__bit/countl.h>
+#include <__bit/rotate.h>
 #include <cassert>
 
 #include "test_macros.h"
diff --git a/libcxx/test/libcxx/time/convert_to_tm.pass.cpp b/libcxx/test/libcxx/time/convert_to_tm.pass.cpp
index 708d2c3ff7c29..908a38dec83d2 100644
--- a/libcxx/test/libcxx/time/convert_to_tm.pass.cpp
+++ b/libcxx/test/libcxx/time/convert_to_tm.pass.cpp
@@ -15,6 +15,7 @@
 // Most of the code is tested indirectly in the chrono formatters. This only
 // tests the hour overflow.
 
+#include <__chrono/convert_to_tm.h>
 #include <chrono>
 #include <cassert>
 #include <format>
diff --git a/libcxx/test/libcxx/transitive_includes/cxx03.csv b/libcxx/test/libcxx/transitive_includes/cxx03.csv
index 2e246644f626c..c2250899a8002 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx03.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx03.csv
@@ -7,7 +7,6 @@ algorithm cstdint
 algorithm cstdlib
 algorithm cstring
 algorithm cwchar
-algorithm execution
 algorithm initializer_list
 algorithm iosfwd
 algorithm iterator
@@ -777,10 +776,21 @@ stop_token limits
 stop_token ratio
 stop_token type_traits
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
 streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf typeinfo
 streambuf version
 string algorithm
 string climits
diff --git a/libcxx/test/libcxx/transitive_includes/cxx11.csv b/libcxx/test/libcxx/transitive_includes/cxx11.csv
index e074bf1f7dcc8..3e929e8f94096 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx11.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx11.csv
@@ -7,7 +7,6 @@ algorithm cstdint
 algorithm cstdlib
 algorithm cstring
 algorithm cwchar
-algorithm execution
 algorithm initializer_list
 algorithm iosfwd
 algorithm iterator
@@ -783,10 +782,22 @@ stop_token limits
 stop_token ratio
 stop_token type_traits
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
 streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string algorithm
 string climits
diff --git a/libcxx/test/libcxx/transitive_includes/cxx14.csv b/libcxx/test/libcxx/transitive_includes/cxx14.csv
index 88f9c24f08646..422db19b6bb8a 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx14.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx14.csv
@@ -785,10 +785,22 @@ stop_token limits
 stop_token ratio
 stop_token type_traits
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
 streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string algorithm
 string climits
diff --git a/libcxx/test/libcxx/transitive_includes/cxx17.csv b/libcxx/test/libcxx/transitive_includes/cxx17.csv
index 88f9c24f08646..422db19b6bb8a 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx17.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx17.csv
@@ -785,10 +785,22 @@ stop_token limits
 stop_token ratio
 stop_token type_traits
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
 streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string algorithm
 string climits
diff --git a/libcxx/test/libcxx/transitive_includes/cxx20.csv b/libcxx/test/libcxx/transitive_includes/cxx20.csv
index 27f59660fb98d..6b80790a9d19b 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx20.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx20.csv
@@ -795,10 +795,22 @@ stop_token limits
 stop_token ratio
 stop_token type_traits
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
 streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string algorithm
 string climits
diff --git a/libcxx/test/libcxx/transitive_includes/cxx23.csv b/libcxx/test/libcxx/transitive_includes/cxx23.csv
index 9ae422a31f074..ea01e41345850 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx23.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx23.csv
@@ -541,9 +541,22 @@ stop_token ctime
 stop_token limits
 stop_token ratio
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
+streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string climits
 string compare
diff --git a/libcxx/test/libcxx/transitive_includes/cxx26.csv b/libcxx/test/libcxx/transitive_includes/cxx26.csv
index 9ae422a31f074..ea01e41345850 100644
--- a/libcxx/test/libcxx/transitive_includes/cxx26.csv
+++ b/libcxx/test/libcxx/transitive_includes/cxx26.csv
@@ -541,9 +541,22 @@ stop_token ctime
 stop_token limits
 stop_token ratio
 stop_token version
+streambuf cctype
 streambuf climits
+streambuf clocale
+streambuf cstddef
+streambuf cstdint
+streambuf cstdlib
+streambuf cstring
+streambuf cwchar
+streambuf initializer_list
 streambuf ios
 streambuf iosfwd
+streambuf limits
+streambuf new
+streambuf string
+streambuf tuple
+streambuf typeinfo
 streambuf version
 string climits
 string compare

From 00162162dd853795c532afa5dec4dc4e798d4a4b Mon Sep 17 00:00:00 2001
From: YunQiang Su <syq@debian.org>
Date: Sun, 14 Apr 2024 23:16:52 +0800
Subject: [PATCH 360/886] CompilerRT: Normalize
 COMPILER_RT_DEFAULT_TARGET_TRIPLE (#88407)

If LLVM is configured with -DLLVM_DEFAULT_TARGET_TRIPLE, and the
argument is not normalized, such as Debian-style vendor-less triple,
clang will try to find libclang_rt in lib/<normalized_triple>, while
libclang_rt is placed into lib/<triple_arg>.

Let's also place libclang_rt into lib/<normalized_triple>.
---
 .../cmake/Modules/CompilerRTUtils.cmake        | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
index e8e5f612d5b03..75f111efeb102 100644
--- a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
+++ b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
@@ -362,10 +362,22 @@ macro(construct_compiler_rt_default_triple)
       message(FATAL_ERROR "CMAKE_C_COMPILER_TARGET must also be set when COMPILER_RT_DEFAULT_TARGET_ONLY is ON")
     endif()
     message(STATUS "cmake c compiler target: ${CMAKE_C_COMPILER_TARGET}")
-    set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${CMAKE_C_COMPILER_TARGET})
+    if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
+      execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${CMAKE_C_COMPILER_TARGET} -print-effective-triple
+                      OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
+                      OUTPUT_STRIP_TRAILING_WHITESPACE)
+    else()
+      set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${CMAKE_C_COMPILER_TARGET})
+    endif()
   else()
-    set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${LLVM_TARGET_TRIPLE} CACHE STRING
-          "Default triple for which compiler-rt runtimes will be built.")
+    if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
+      execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${LLVM_TARGET_TRIPLE} -print-effective-triple
+                      OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
+                      OUTPUT_STRIP_TRAILING_WHITESPACE)
+    else()
+      set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${LLVM_TARGET_TRIPLE} CACHE STRING
+            "Default triple for which compiler-rt runtimes will be built.")
+    endif()
   endif()
 
   string(REPLACE "-" ";" LLVM_TARGET_TRIPLE_LIST ${COMPILER_RT_DEFAULT_TARGET_TRIPLE})

From 5927492e8e68c1ee5a0d84cf6c402a900aac4a3c Mon Sep 17 00:00:00 2001
From: YunQiang Su <syq@debian.org>
Date: Mon, 15 Apr 2024 00:12:42 +0800
Subject: [PATCH 361/886] Revert "CompilerRT: Normalize
 COMPILER_RT_DEFAULT_TARGET_TRIPLE" (#88663)

Reverts llvm/llvm-project#88407
---
 .../cmake/Modules/CompilerRTUtils.cmake        | 18 +++---------------
 1 file changed, 3 insertions(+), 15 deletions(-)

diff --git a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
index 75f111efeb102..e8e5f612d5b03 100644
--- a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
+++ b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
@@ -362,22 +362,10 @@ macro(construct_compiler_rt_default_triple)
       message(FATAL_ERROR "CMAKE_C_COMPILER_TARGET must also be set when COMPILER_RT_DEFAULT_TARGET_ONLY is ON")
     endif()
     message(STATUS "cmake c compiler target: ${CMAKE_C_COMPILER_TARGET}")
-    if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
-      execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${CMAKE_C_COMPILER_TARGET} -print-effective-triple
-                      OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
-                      OUTPUT_STRIP_TRAILING_WHITESPACE)
-    else()
-      set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${CMAKE_C_COMPILER_TARGET})
-    endif()
+    set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${CMAKE_C_COMPILER_TARGET})
   else()
-    if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
-      execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${LLVM_TARGET_TRIPLE} -print-effective-triple
-                      OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
-                      OUTPUT_STRIP_TRAILING_WHITESPACE)
-    else()
-      set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${LLVM_TARGET_TRIPLE} CACHE STRING
-            "Default triple for which compiler-rt runtimes will be built.")
-    endif()
+    set(COMPILER_RT_DEFAULT_TARGET_TRIPLE ${LLVM_TARGET_TRIPLE} CACHE STRING
+          "Default triple for which compiler-rt runtimes will be built.")
   endif()
 
   string(REPLACE "-" ";" LLVM_TARGET_TRIPLE_LIST ${COMPILER_RT_DEFAULT_TARGET_TRIPLE})

From b8d0cba14bcfc5c1c2f7a878ad9eaa12b6a590b6 Mon Sep 17 00:00:00 2001
From: Cyndy Ishida <cyndy_ishida@apple.com>
Date: Sun, 14 Apr 2024 09:23:42 -0700
Subject: [PATCH 362/886] [TextAPI] Apply NFC code fixups (#88639)

* Remove unnecessary cast
* Annotate `#endif`
---
 llvm/include/llvm/TextAPI/Utils.h             | 2 +-
 llvm/lib/TextAPI/BinaryReader/DylibReader.cpp | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/llvm/include/llvm/TextAPI/Utils.h b/llvm/include/llvm/TextAPI/Utils.h
index 319e0abce25b0..ebfe88984f803 100644
--- a/llvm/include/llvm/TextAPI/Utils.h
+++ b/llvm/include/llvm/TextAPI/Utils.h
@@ -21,7 +21,7 @@
 
 #if !defined(PATH_MAX)
 #define PATH_MAX 1024
-#endif
+#endif // !defined(PATH_MAX)
 
 #define MACCATALYST_PREFIX_PATH "/System/iOSSupport"
 #define DRIVERKIT_PREFIX_PATH "/System/DriverKit"
diff --git a/llvm/lib/TextAPI/BinaryReader/DylibReader.cpp b/llvm/lib/TextAPI/BinaryReader/DylibReader.cpp
index f92a2d19a63fc..ec0271f774ecc 100644
--- a/llvm/lib/TextAPI/BinaryReader/DylibReader.cpp
+++ b/llvm/lib/TextAPI/BinaryReader/DylibReader.cpp
@@ -484,7 +484,7 @@ accumulateLocs(MachOObjectFile &Obj,
     auto Sym = parseSymbol(Name);
 
     if (!File.empty() && Line != 0)
-      LocMap.insert({Sym.Name.str(), RecordLoc(File, Line)});
+      LocMap.insert({Sym.Name, RecordLoc(File, Line)});
   }
 
   return LocMap;

From 48e955496bee035121ebfe5a837ebcbc2ba8846c Mon Sep 17 00:00:00 2001
From: Marc Auberer <marc.auberer@chillibits.com>
Date: Sun, 14 Apr 2024 21:37:39 +0200
Subject: [PATCH 363/886] [Docs] Update python dependencies for documentation
 (#87877)

This has no particular reason, but imo we should update dependencies
every now and then to reduce the gap if breaking changes happen
somewhere.
---
 llvm/docs/requirements-hashed.txt | 24 ++++++++++++------------
 llvm/docs/requirements.txt        |  8 ++++----
 2 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/llvm/docs/requirements-hashed.txt b/llvm/docs/requirements-hashed.txt
index 19d1cfdfc572d..64cfc8fc79799 100644
--- a/llvm/docs/requirements-hashed.txt
+++ b/llvm/docs/requirements-hashed.txt
@@ -124,9 +124,9 @@ docutils==0.20.1 \
     #   myst-parser
     #   recommonmark
     #   sphinx
-furo==2023.8.19 \
-    --hash=sha256:12f99f87a1873b6746228cfde18f77244e6c1ffb85d7fed95e638aae70d80590 \
-    --hash=sha256:e671ee638ab3f1b472f4033b0167f502ab407830e0db0f843b1c1028119c9cd1
+furo==2024.1.29 \
+    --hash=sha256:3548be2cef45a32f8cdc0272d415fcb3e5fa6a0eb4ddfe21df3ecf1fe45a13cf \
+    --hash=sha256:4d6b2fe3f10a6e36eb9cc24c1e7beb38d7a23fc7b3c382867503b7fcac8a1e02
     # via -r requirements.txt
 idna==3.6 \
     --hash=sha256:9ecdbbd083b06798ae1e86adcbfe8ab1479cf864e4ee30fe4e46a003d12491ca \
@@ -304,9 +304,9 @@ soupsieve==2.5 \
     --hash=sha256:5663d5a7b3bfaeee0bc4372e7fc48f9cff4940b3eec54a6451cc5299f1097690 \
     --hash=sha256:eaa337ff55a1579b6549dc679565eac1e3d000563bcb1c8ab0d0fefbc0c2cdc7
     # via beautifulsoup4
-sphinx==7.1.2 \
-    --hash=sha256:780f4d32f1d7d1126576e0e5ecc19dc32ab76cd24e950228dcf7b1f6d3d9e22f \
-    --hash=sha256:d170a81825b2fcacb6dfd5a0d7f578a053e45d3f2b153fecc948c37344eb4cbe
+sphinx==7.2.6 \
+    --hash=sha256:1e09160a40b956dc623c910118fa636da93bd3ca0b9876a7b3df90f07d691560 \
+    --hash=sha256:9a5160e1ea90688d5963ba09a2dcd8bdd526620edbb65c328728f1b2228d5ab5
     # via
     #   -r requirements.txt
     #   furo
@@ -318,9 +318,9 @@ sphinx==7.1.2 \
     #   sphinxcontrib-htmlhelp
     #   sphinxcontrib-qthelp
     #   sphinxcontrib-serializinghtml
-sphinx-automodapi==0.16.0 \
-    --hash=sha256:68fc47064804604b90aa27c047016e86aaf970981d90a0082d5b5dd2e9d38afd \
-    --hash=sha256:6c673ef93066408e5ad3e2fa3533044d432a47fe6a826212b9ebf5f52a872554
+sphinx-automodapi==0.17.0 \
+    --hash=sha256:4d029cb79eef29413e94ab01bb0177ebd2d5ba86e9789b73575afe9c06ae1501 \
+    --hash=sha256:7ccdadad57add4aa9149d9f2bb5cf28c8f8b590280b4735b1156ea8355c423a1
     # via -r requirements.txt
 sphinx-basic-ng==1.0.0b2 \
     --hash=sha256:9ec55a47c90c8c002b5960c57492ec3021f5193cb26cebc2dc4ea226848651c9 \
@@ -334,9 +334,9 @@ sphinx-markdown-tables==0.0.17 \
     --hash=sha256:2bd0c30779653e4dd120300cbd9ca412c480738cc2241f6dea477a883f299e04 \
     --hash=sha256:6bc6d3d400eaccfeebd288446bc08dd83083367c58b85d40fe6c12d77ef592f1
     # via -r requirements.txt
-sphinxcontrib-applehelp==1.0.4 \
-    --hash=sha256:29d341f67fb0f6f586b23ad80e072c8e6ad0b48417db2bde114a4c9746feb228 \
-    --hash=sha256:828f867945bbe39817c210a1abfd1bc4895c8b73fcaade56d45357a348a07d7e
+sphinxcontrib-applehelp==1.0.8 \
+    --hash=sha256:c40a4f96f3776c4393d933412053962fac2b84f4c99a7982ba42e09576a70619 \
+    --hash=sha256:cb61eb0ec1b61f349e5cc36b2028e9e7ca765be05e49641c97241274753067b4
     # via
     #   -r requirements.txt
     #   sphinx
diff --git a/llvm/docs/requirements.txt b/llvm/docs/requirements.txt
index d66569c1b3e79..960a6eb644e14 100644
--- a/llvm/docs/requirements.txt
+++ b/llvm/docs/requirements.txt
@@ -1,9 +1,9 @@
-sphinx==7.1.2
+sphinx==7.2.6
 docutils==0.20.1
 sphinx-markdown-tables==0.0.17
 recommonmark==0.7.1
-sphinx-automodapi==0.16.0
+sphinx-automodapi==0.17.0
 sphinx-bootstrap-theme==0.8.1
-sphinxcontrib-applehelp==1.0.4
-furo==2023.8.19
+sphinxcontrib-applehelp==1.0.8
+furo==2024.1.29
 myst-parser==2.0.0

From 84ec2d3b38635dead7690f97f215ff4c473184f0 Mon Sep 17 00:00:00 2001
From: Maksim Levental <maksim.levental@gmail.com>
Date: Sun, 14 Apr 2024 17:20:05 -0500
Subject: [PATCH 364/886] [mlir][python] add to CODEOWNERS (#88561)

---
 .github/CODEOWNERS | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
index 76b8266cae87c..c4727a0c267d3 100644
--- a/.github/CODEOWNERS
+++ b/.github/CODEOWNERS
@@ -115,6 +115,10 @@ clang/test/AST/Interp/ @tbaederr
 /mlir/**/LLVMIR/**/BasicPtxBuilderInterface* @grypp
 /mlir/**/NVVM*/ @grypp
 
+# MLIR Python Bindings
+/mlir/test/python/ @makslevental @stellaraccident
+/mlir/python/ @makslevental @stellaraccident
+
 # BOLT
 /bolt/ @aaupov @maksfb @rafaelauler @ayermolo @dcci
 

From a5f54175dcf120180c3d91bbc13062bbf8f42f61 Mon Sep 17 00:00:00 2001
From: Zentrik <Zentrik@users.noreply.github.com>
Date: Sun, 14 Apr 2024 23:21:06 +0100
Subject: [PATCH 365/886] Include cmath to fix build error on mac os 10.14
 (#88665)

This fixes #88664.
---
 clang-tools-extra/clang-tidy/modernize/UseStdNumbersCheck.cpp | 1 +
 1 file changed, 1 insertion(+)

diff --git a/clang-tools-extra/clang-tidy/modernize/UseStdNumbersCheck.cpp b/clang-tools-extra/clang-tidy/modernize/UseStdNumbersCheck.cpp
index b299afd540b9a..1548fc454cfb3 100644
--- a/clang-tools-extra/clang-tidy/modernize/UseStdNumbersCheck.cpp
+++ b/clang-tools-extra/clang-tidy/modernize/UseStdNumbersCheck.cpp
@@ -29,6 +29,7 @@
 #include "llvm/Support/FormatVariadic.h"
 #include "llvm/Support/MathExtras.h"
 #include <array>
+#include <cmath>
 #include <cstdint>
 #include <cstdlib>
 #include <initializer_list>

From 9f43a41db37253685c5ed428c215528eb92bbd43 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Sun, 14 Apr 2024 17:04:03 -0700
Subject: [PATCH 366/886] [Driver,test] Make Android runtime tests resisent to
 #87866 change

Suggested by YunQiang Su <syq@debian.org>
---
 clang/test/Driver/linux-ld.c     | 2 ++
 clang/test/Driver/sanitizer-ld.c | 6 ++++++
 2 files changed, 8 insertions(+)

diff --git a/clang/test/Driver/linux-ld.c b/clang/test/Driver/linux-ld.c
index 4020b138dc8fd..f4e40439a5f31 100644
--- a/clang/test/Driver/linux-ld.c
+++ b/clang/test/Driver/linux-ld.c
@@ -92,6 +92,7 @@
 // RUN: %clang -### %s -no-pie 2>&1 \
 // RUN:     --target=arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:     --rtlib=compiler-rt \
 // RUN:   | FileCheck --check-prefix=CHECK-LD-RT-ANDROID %s
 // CHECK-LD-RT-ANDROID-NOT: warning:
@@ -262,6 +263,7 @@
 // RUN: %clang -static -### %s -no-pie 2>&1 \
 // RUN:     --target=aarch64-linux-android -rtlib=platform --unwindlib=platform \
 // RUN:     --sysroot=%S/Inputs/basic_linux_tree \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANG-ANDROID-STATIC %s
 // CHECK-CLANG-ANDROID-STATIC: "{{.*}}ld{{(.exe)?}}" "--sysroot=[[SYSROOT:[^"]+]]"
 // CHECK-CLANG-ANDROID-STATIC: "--start-group" "{{[^"]*}}{{/|\\\\}}libclang_rt.builtins-aarch64-android.a" "-l:libunwind.a" "-lc" "--end-group"
diff --git a/clang/test/Driver/sanitizer-ld.c b/clang/test/Driver/sanitizer-ld.c
index 53e536d772924..f4ac484d6e4cb 100644
--- a/clang/test/Driver/sanitizer-ld.c
+++ b/clang/test/Driver/sanitizer-ld.c
@@ -177,6 +177,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=address \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID %s
 //
 // CHECK-ASAN-ANDROID: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
@@ -191,6 +192,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=address \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:     -static-libsan \
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID-STATICLIBASAN %s
 //
@@ -203,6 +205,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=undefined \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:   | FileCheck --check-prefix=CHECK-UBSAN-ANDROID %s
 //
 // CHECK-UBSAN-ANDROID: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
@@ -217,6 +220,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=undefined \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:     -static-libsan \
 // RUN:   | FileCheck --check-prefix=CHECK-UBSAN-ANDROID-STATICLIBASAN %s
 //
@@ -230,6 +234,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=i686-linux-android -fuse-ld=ld -fsanitize=address \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID-X86 %s
 //
 // CHECK-ASAN-ANDROID-X86: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
@@ -252,6 +257,7 @@
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=address \
 // RUN:     --sysroot=%S/Inputs/basic_android_tree/sysroot \
+// RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:     -shared \
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID-SHARED %s
 //

From 1693009679313282afbed38778dd3fad62641e1b Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Sun, 14 Apr 2024 17:05:55 -0700
Subject: [PATCH 367/886] [Driver] Ensure ToolChain::LibraryPaths is not empty
 for non-Darwin

Follow-up to #81037.

ToolChain::LibraryPaths holds the new compiler-rt library directory
(e.g. `/tmp/Debug/lib/clang/19/lib/x86_64-unknown-linux-gnu`). However,
it might be empty when the directory does not exist (due to the `if
(getVFS().exists(P))` change in https://reviews.llvm.org/D158475).

If neither the old/new compiler-rt library directories exists, we would
suggest the undesired old compiler-rt file name:

```
% /tmp/Debug/bin/clang++ a.cc -fsanitize=memory -o a
ld.lld: error: cannot open /tmp/Debug/lib/clang/19/lib/linux/libclang_rt.msan-x86_64.a: No such file or directory
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
```

With this change, we will correctly suggest the new compiler-rt file name.

Fix #87150

Pull Request: https://github.com/llvm/llvm-project/pull/87866
---
 clang/lib/Driver/ToolChain.cpp                |  8 ++++-
 clang/test/Driver/arm-compiler-rt.c           | 14 ++++-----
 clang/test/Driver/cl-link.c                   | 16 +++++-----
 clang/test/Driver/compiler-rt-unwind.c        |  6 ++--
 clang/test/Driver/coverage-ld.c               |  8 ++---
 clang/test/Driver/instrprof-ld.c              | 16 +++++-----
 clang/test/Driver/linux-ld.c                  |  6 ++--
 clang/test/Driver/mingw-sanitizers.c          | 16 +++++-----
 clang/test/Driver/msp430-toolchain.c          |  4 +--
 .../Driver/print-libgcc-file-name-clangrt.c   | 12 ++++----
 clang/test/Driver/print-runtime-dir.c         |  6 ----
 clang/test/Driver/riscv32-toolchain-extra.c   |  6 ++--
 clang/test/Driver/riscv32-toolchain.c         |  6 ++--
 clang/test/Driver/riscv64-toolchain-extra.c   |  6 ++--
 clang/test/Driver/riscv64-toolchain.c         |  6 ++--
 clang/test/Driver/sanitizer-ld.c              | 30 +++++++++----------
 clang/test/Driver/wasm-toolchain.c            | 18 +++++------
 clang/test/Driver/wasm-toolchain.cpp          | 16 +++++-----
 clang/test/Driver/windows-cross.c             | 18 +++++------
 clang/test/Driver/zos-ld.c                    | 12 ++++----
 .../test/Driver/msvc-dependent-lib-flags.f90  |  8 ++---
 21 files changed, 119 insertions(+), 119 deletions(-)

diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 03450fc0f57b9..237092ed07e5d 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -796,7 +796,13 @@ ToolChain::getTargetSubDirPath(StringRef BaseDir) const {
 std::optional<std::string> ToolChain::getRuntimePath() const {
   SmallString<128> P(D.ResourceDir);
   llvm::sys::path::append(P, "lib");
-  return getTargetSubDirPath(P);
+  if (auto Ret = getTargetSubDirPath(P))
+    return Ret;
+  // Darwin does not use per-target runtime directory.
+  if (Triple.isOSDarwin())
+    return {};
+  llvm::sys::path::append(P, Triple.str());
+  return std::string(P);
 }
 
 std::optional<std::string> ToolChain::getStdlibPath() const {
diff --git a/clang/test/Driver/arm-compiler-rt.c b/clang/test/Driver/arm-compiler-rt.c
index 5e9e528400d08..cb6c29f48a781 100644
--- a/clang/test/Driver/arm-compiler-rt.c
+++ b/clang/test/Driver/arm-compiler-rt.c
@@ -10,47 +10,47 @@
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABI
-// ARM-GNUEABI: "{{.*[/\\]}}libclang_rt.builtins-arm.a"
+// ARM-GNUEABI: "{{.*[/\\]}}libclang_rt.builtins.a"
 
 // RUN: %clang -target arm-linux-gnueabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=hard -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABI-ABI
-// ARM-GNUEABI-ABI: "{{.*[/\\]}}libclang_rt.builtins-armhf.a"
+// ARM-GNUEABI-ABI: "{{.*[/\\]}}libclang_rt.builtins.a"
 
 // RUN: %clang -target arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABIHF
-// ARM-GNUEABIHF: "{{.*[/\\]}}libclang_rt.builtins-armhf.a"
+// ARM-GNUEABIHF: "{{.*[/\\]}}libclang_rt.builtins.a"
 
 // RUN: %clang -target arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=soft -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-GNUEABIHF-ABI
-// ARM-GNUEABIHF-ABI: "{{.*[/\\]}}libclang_rt.builtins-arm.a"
+// ARM-GNUEABIHF-ABI: "{{.*[/\\]}}libclang_rt.builtins.a"
 
 // RUN: %clang -target arm-windows-itanium \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-WINDOWS
-// ARM-WINDOWS: "{{.*[/\\]}}clang_rt.builtins-arm.lib"
+// ARM-WINDOWS: "{{.*[/\\]}}clang_rt.builtins.lib"
 
 // RUN: %clang -target arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-ANDROID
-// ARM-ANDROID: "{{.*[/\\]}}libclang_rt.builtins-arm-android.a"
+// ARM-ANDROID: "{{.*[/\\]}}libclang_rt.builtins.a"
 
 // RUN: not %clang --target=arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -rtlib=compiler-rt -mfloat-abi=hard -### %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix ARM-ANDROIDHF
-// ARM-ANDROIDHF: "{{.*[/\\]}}libclang_rt.builtins-armhf-android.a"
+// ARM-ANDROIDHF: "{{.*[/\\]}}libclang_rt.builtins.a"
 
diff --git a/clang/test/Driver/cl-link.c b/clang/test/Driver/cl-link.c
index 444f0c01b3f99..ffd0b5ac4bade 100644
--- a/clang/test/Driver/cl-link.c
+++ b/clang/test/Driver/cl-link.c
@@ -13,20 +13,20 @@
 // ASAN: link.exe
 // ASAN: "-debug"
 // ASAN: "-incremental:no"
-// ASAN: "{{[^"]*}}clang_rt.asan-i386.lib"
-// ASAN: "-wholearchive:{{.*}}clang_rt.asan-i386.lib"
-// ASAN: "{{[^"]*}}clang_rt.asan_cxx-i386.lib"
-// ASAN: "-wholearchive:{{.*}}clang_rt.asan_cxx-i386.lib"
+// ASAN: "{{[^"]*}}clang_rt.asan.lib"
+// ASAN: "-wholearchive:{{.*}}clang_rt.asan.lib"
+// ASAN: "{{[^"]*}}clang_rt.asan_cxx.lib"
+// ASAN: "-wholearchive:{{.*}}clang_rt.asan_cxx.lib"
 // ASAN: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl -m32 -arch:IA32 --target=i386-pc-win32 /MD /Tc%s -fuse-ld=link -### -fsanitize=address 2>&1 | FileCheck --check-prefix=ASAN-MD %s
 // ASAN-MD: link.exe
 // ASAN-MD: "-debug"
 // ASAN-MD: "-incremental:no"
-// ASAN-MD: "{{.*}}clang_rt.asan_dynamic-i386.lib"
-// ASAN-MD: "{{[^"]*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib"
+// ASAN-MD: "{{.*}}clang_rt.asan_dynamic.lib"
+// ASAN-MD: "{{[^"]*}}clang_rt.asan_dynamic_runtime_thunk.lib"
 // ASAN-MD: "-include:___asan_seh_interceptor"
-// ASAN-MD: "-wholearchive:{{.*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib"
+// ASAN-MD: "-wholearchive:{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib"
 // ASAN-MD: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl /LD -fuse-ld=link -### /Tc%s 2>&1 | FileCheck --check-prefix=DLL %s
@@ -40,7 +40,7 @@
 // ASAN-DLL: "-dll"
 // ASAN-DLL: "-debug"
 // ASAN-DLL: "-incremental:no"
-// ASAN-DLL: "{{.*}}clang_rt.asan_dll_thunk-i386.lib"
+// ASAN-DLL: "{{.*}}clang_rt.asan_dll_thunk.lib"
 // ASAN-DLL: "{{.*}}cl-link{{.*}}.obj"
 
 // RUN: %clang_cl /Zi /Tc%s -fuse-ld=link -### 2>&1 | FileCheck --check-prefix=DEBUG %s
diff --git a/clang/test/Driver/compiler-rt-unwind.c b/clang/test/Driver/compiler-rt-unwind.c
index 7f4e3f22ab19a..c5040d7fd900b 100644
--- a/clang/test/Driver/compiler-rt-unwind.c
+++ b/clang/test/Driver/compiler-rt-unwind.c
@@ -98,14 +98,14 @@
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:     -shared-libgcc \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
+// MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
 // MINGW-RTLIB-COMPILER-RT-SHARED-UNWINDLIB-COMPILER-RT-SAME: "-l:libunwind.dll.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:     -static-libgcc \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
+// MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
 // MINGW-RTLIB-COMPILER-RT-STATIC-UNWINDLIB-COMPILER-RT-SAME: "-l:libunwind.a"
 //
 // RUN: %clang -### %s 2>&1 \
@@ -114,5 +114,5 @@
 // RUN: %clangxx -### %s 2>&1 \
 // RUN:     --target=x86_64-w64-mingw32 -rtlib=compiler-rt --unwindlib=libunwind \
 // RUN:   | FileCheck --check-prefix=MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT %s
-// MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins-x86_64.a"
+// MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT: "{{.*}}libclang_rt.builtins.a"
 // MINGW-RTLIB-COMPILER-RT-UNWINDLIB-COMPILER-RT-SAME: "-lunwind"
diff --git a/clang/test/Driver/coverage-ld.c b/clang/test/Driver/coverage-ld.c
index acb08eb5db59a..be1d8320ab8be 100644
--- a/clang/test/Driver/coverage-ld.c
+++ b/clang/test/Driver/coverage-ld.c
@@ -33,7 +33,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64 %s
 //
 // CHECK-FREEBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-netbsd --coverage -fuse-ld=ld \
@@ -42,7 +42,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64 %s
 
 // CHECK-NETBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-openbsd --coverage -fuse-ld=ld \
@@ -51,7 +51,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64 %s
 
 // CHECK-OPENBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi --coverage -fuse-ld=ld \
@@ -60,4 +60,4 @@
 // RUN:   | FileCheck --check-prefix=CHECK-ANDROID-ARM %s
 //
 // CHECK-ANDROID-ARM: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-ANDROID-ARM: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}linux{{/|\\\\}}libclang_rt.profile-arm-android.a"
+// CHECK-ANDROID-ARM: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}arm-unknown-linux-android{{/|\\\\}}libclang_rt.profile.a"
diff --git a/clang/test/Driver/instrprof-ld.c b/clang/test/Driver/instrprof-ld.c
index 674580b349d42..a96bba4a1e763 100644
--- a/clang/test/Driver/instrprof-ld.c
+++ b/clang/test/Driver/instrprof-ld.c
@@ -34,7 +34,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64 %s
 //
 // CHECK-FREEBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-FREEBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-netbsd -fprofile-instr-generate -fuse-ld=ld \
@@ -43,7 +43,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64 %s
 
 // CHECK-NETBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-NETBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-unknown-openbsd -fprofile-instr-generate -fuse-ld=ld \
@@ -52,7 +52,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64 %s
 
 // CHECK-OPENBSD-X86-64: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-OPENBSD-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -72,7 +72,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-LINUX-X86-64-SHARED %s
 //
 // CHECK-LINUX-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-LINUX-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{.*}}linux{{.*}}libclang_rt.profile.a" {{.*}} "-lc"
+// CHECK-LINUX-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{.*}}x86_64-unknown-linux{{.*}}libclang_rt.profile.a" {{.*}} "-lc"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -82,7 +82,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-FREEBSD-X86-64-SHARED %s
 //
 // CHECK-FREEBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-FREEBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}freebsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-FREEBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-freebsd{{/|\\\\}}libclang_rt.profile.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -92,7 +92,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-NETBSD-X86-64-SHARED %s
 
 // CHECK-NETBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-NETBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}netbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-NETBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-netbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     -shared \
@@ -102,7 +102,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-OPENBSD-X86-64-SHARED %s
 
 // CHECK-OPENBSD-X86-64-SHARED: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-OPENBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}openbsd{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-OPENBSD-X86-64-SHARED: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-openbsd{{/|\\\\}}libclang_rt.profile.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=x86_64-apple-darwin14 -fprofile-instr-generate -fuse-ld=ld \
@@ -174,7 +174,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-MINGW-X86-64 %s
 //
 // CHECK-MINGW-X86-64: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
-// CHECK-MINGW-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}windows{{/|\\\\}}libclang_rt.profile-x86_64.a"
+// CHECK-MINGW-X86-64: "{{.*}}/Inputs/resource_dir{{/|\\\\}}lib{{/|\\\\}}x86_64-unknown-windows-gnu{{/|\\\\}}libclang_rt.profile.a"
 
 // Test instrumented profiling dependent-lib flags
 //
diff --git a/clang/test/Driver/linux-ld.c b/clang/test/Driver/linux-ld.c
index f4e40439a5f31..d918f4f2d7dbd 100644
--- a/clang/test/Driver/linux-ld.c
+++ b/clang/test/Driver/linux-ld.c
@@ -100,9 +100,9 @@
 // CHECK-LD-RT-ANDROID: "--eh-frame-hdr"
 // CHECK-LD-RT-ANDROID: "-m" "armelf_linux_eabi"
 // CHECK-LD-RT-ANDROID: "-dynamic-linker"
-// CHECK-LD-RT-ANDROID: libclang_rt.builtins-arm-android.a"
+// CHECK-LD-RT-ANDROID: libclang_rt.builtins.a"
 // CHECK-LD-RT-ANDROID: "-lc"
-// CHECK-LD-RT-ANDROID: libclang_rt.builtins-arm-android.a"
+// CHECK-LD-RT-ANDROID: libclang_rt.builtins.a"
 //
 // RUN: %clang -### %s -no-pie 2>&1 \
 // RUN:     --target=x86_64-unknown-linux -rtlib=platform --unwindlib=platform \
@@ -266,7 +266,7 @@
 // RUN:     -resource-dir=%S/Inputs/resource_dir \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANG-ANDROID-STATIC %s
 // CHECK-CLANG-ANDROID-STATIC: "{{.*}}ld{{(.exe)?}}" "--sysroot=[[SYSROOT:[^"]+]]"
-// CHECK-CLANG-ANDROID-STATIC: "--start-group" "{{[^"]*}}{{/|\\\\}}libclang_rt.builtins-aarch64-android.a" "-l:libunwind.a" "-lc" "--end-group"
+// CHECK-CLANG-ANDROID-STATIC: "--start-group" "{{[^"]*}}{{/|\\\\}}libclang_rt.builtins.a" "-l:libunwind.a" "-lc" "--end-group"
 //
 // RUN: %clang -### %s 2>&1      \
 // RUN:     --target=x86_64-unknown-linux -rtlib=platform --unwindlib=platform \
diff --git a/clang/test/Driver/mingw-sanitizers.c b/clang/test/Driver/mingw-sanitizers.c
index d165648a8fdf6..2325f8f0f1f23 100644
--- a/clang/test/Driver/mingw-sanitizers.c
+++ b/clang/test/Driver/mingw-sanitizers.c
@@ -4,17 +4,17 @@
 //
 // ASAN-ALL-NOT:"-l{{[^"]+"]}}"
 // ASAN-ALL-NOT:"[[INPUT]]"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic-i386.dll.a"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic-x86_64.dll.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
 // ASAN-ALL:    "-lcomponent"
 // ASAN-ALL:    "[[INPUT]]"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic-i386.dll.a"
-// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-i386.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
+// ASAN-I686:   "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a"
 // ASAN-I686:   "--require-defined" "___asan_seh_interceptor"
-// ASAN-I686:   "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-i386.a" "--no-whole-archive"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic-x86_64.dll.a"
-// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-x86_64.a"
+// ASAN-I686:   "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a" "--no-whole-archive"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic.dll.a"
+// ASAN-X86_64: "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a"
 // ASAN-X86_64: "--require-defined" "__asan_seh_interceptor"
-// ASAN-X86_64: "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk-x86_64.a" "--no-whole-archive"
+// ASAN-X86_64: "--whole-archive" "{{[^"]*}}libclang_rt.asan_dynamic_runtime_thunk.a" "--no-whole-archive"
 
 // RUN: %clang -target x86_64-windows-gnu %s -### -fsanitize=vptr
diff --git a/clang/test/Driver/msp430-toolchain.c b/clang/test/Driver/msp430-toolchain.c
index ef6780c38f2ee..3c3042b482ef2 100644
--- a/clang/test/Driver/msp430-toolchain.c
+++ b/clang/test/Driver/msp430-toolchain.c
@@ -103,8 +103,8 @@
 // LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtbegin_no_eh.o"
 // LIBS-COMPILER-RT-POS: "-L{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430"
 // LIBS-COMPILER-RT-POS: "-L{{.*}}/Inputs/basic_msp430_tree{{/|\\\\}}msp430-elf{{/|\\\\}}lib/430"
-// LIBS-COMPILER-RT-POS: "{{[^"]*}}libclang_rt.builtins-msp430.a" "--start-group" "-lmul_none" "-lc" "{{[^"]*}}libclang_rt.builtins-msp430.a" "-lcrt" "-lnosys" "--end-group" "{{[^"]*}}libclang_rt.builtins-msp430.a"
-// LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtend_no_eh.o" "{{[^"]*}}libclang_rt.builtins-msp430.a"
+// LIBS-COMPILER-RT-POS: "{{[^"]*}}libclang_rt.builtins.a" "--start-group" "-lmul_none" "-lc" "{{[^"]*}}libclang_rt.builtins.a" "-lcrt" "-lnosys" "--end-group" "{{[^"]*}}libclang_rt.builtins.a"
+// LIBS-COMPILER-RT-POS: "{{.*}}/Inputs/basic_msp430_tree/lib/gcc/msp430-elf/8.3.1/430{{/|\\\\}}crtend_no_eh.o" "{{[^"]*}}libclang_rt.builtins.a"
 // LIBS-COMPILER-RT-NEG-NOT: crtbegin.o
 // LIBS-COMPILER-RT-NEG-NOT: -lssp_nonshared
 // LIBS-COMPILER-RT-NEG-NOT: -lssp
diff --git a/clang/test/Driver/print-libgcc-file-name-clangrt.c b/clang/test/Driver/print-libgcc-file-name-clangrt.c
index ed740e0d2917d..a902eedc85209 100644
--- a/clang/test/Driver/print-libgcc-file-name-clangrt.c
+++ b/clang/test/Driver/print-libgcc-file-name-clangrt.c
@@ -5,14 +5,14 @@
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-X8664 %s
-// CHECK-CLANGRT-X8664: libclang_rt.builtins-x86_64.a
+// CHECK-CLANGRT-X8664: libclang_rt.builtins.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=i386-pc-linux \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-I386 %s
-// CHECK-CLANGRT-I386: libclang_rt.builtins-i386.a
+// CHECK-CLANGRT-I386: libclang_rt.builtins.a
 
 // Check whether alternate arch values map to the correct library.
 //
@@ -27,28 +27,28 @@
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM %s
-// CHECK-CLANGRT-ARM: libclang_rt.builtins-arm.a
+// CHECK-CLANGRT-ARM: libclang_rt.builtins.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-androideabi \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM-ANDROID %s
-// CHECK-CLANGRT-ARM-ANDROID: libclang_rt.builtins-arm-android.a
+// CHECK-CLANGRT-ARM-ANDROID: libclang_rt.builtins.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-gnueabihf \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARMHF %s
-// CHECK-CLANGRT-ARMHF: libclang_rt.builtins-armhf.a
+// CHECK-CLANGRT-ARMHF: libclang_rt.builtins.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=arm-linux-gnueabi -mfloat-abi=hard \
 // RUN:     --sysroot=%S/Inputs/resource_dir_with_arch_subdir \
 // RUN:     -resource-dir=%S/Inputs/resource_dir_with_arch_subdir 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-CLANGRT-ARM-ABI %s
-// CHECK-CLANGRT-ARM-ABI: libclang_rt.builtins-armhf.a
+// CHECK-CLANGRT-ARM-ABI: libclang_rt.builtins.a
 
 // RUN: %clang -rtlib=compiler-rt -print-libgcc-file-name \
 // RUN:     --target=armv7m-none-eabi \
diff --git a/clang/test/Driver/print-runtime-dir.c b/clang/test/Driver/print-runtime-dir.c
index 550ffef1aaf6e..ac1ff7e634b81 100644
--- a/clang/test/Driver/print-runtime-dir.c
+++ b/clang/test/Driver/print-runtime-dir.c
@@ -1,9 +1,3 @@
-// Default directory layout
-// RUN: %clang -print-runtime-dir --target=x86_64-pc-windows-msvc \
-// RUN:   -resource-dir=%S/Inputs/resource_dir \
-// RUN:   | FileCheck --check-prefix=PRINT-RUNTIME-DIR -DFILE=%S/Inputs/resource_dir %s
-// PRINT-RUNTIME-DIR: [[FILE]]{{/|\\}}lib{{/|\\}}windows
-
 // Per-target directory layout
 // RUN: %clang -print-runtime-dir --target=x86_64-pc-windows-msvc \
 // RUN:   -resource-dir=%S/Inputs/resource_dir_with_per_target_subdir \
diff --git a/clang/test/Driver/riscv32-toolchain-extra.c b/clang/test/Driver/riscv32-toolchain-extra.c
index 2d38aa3b545fe..aab6b36f3cfca 100644
--- a/clang/test/Driver/riscv32-toolchain-extra.c
+++ b/clang/test/Driver/riscv32-toolchain-extra.c
@@ -29,8 +29,8 @@
 // C-RV32-BAREMETAL-ILP32-NOGCC: "-internal-isystem" "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/include"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/riscv32-unknown-elf-ld"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/lib/crt0.o"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/clang_rt.crtbegin-riscv32.o"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/clang_rt.crtbegin.o"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/bin/../riscv32-unknown-elf/lib"
 // C-RV32-BAREMETAL-ILP32-NOGCC: "--start-group" "-lc" "-lgloss" "--end-group"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/libclang_rt.builtins-riscv32.a"
-// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/lib/clang_rt.crtend-riscv32.o"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/libclang_rt.builtins.a"
+// C-RV32-BAREMETAL-ILP32-NOGCC: "{{.*}}/riscv32-nogcc/{{.*}}/riscv32-unknown-unknown-elf/clang_rt.crtend.o"
diff --git a/clang/test/Driver/riscv32-toolchain.c b/clang/test/Driver/riscv32-toolchain.c
index bb2533cdf1bce..322a6ca2840fb 100644
--- a/clang/test/Driver/riscv32-toolchain.c
+++ b/clang/test/Driver/riscv32-toolchain.c
@@ -195,9 +195,9 @@
 // RUN:   --target=riscv32-unknown-elf --rtlib=compiler-rt --unwindlib=compiler-rt 2>&1 \
 // RUN:   | FileCheck -check-prefix=C-RV32-RTLIB-COMPILERRT-ILP32 %s
 // C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}crt0.o"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtbegin-riscv32.o"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins-riscv32.a"
-// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtend-riscv32.o"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtbegin.o"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins.a"
+// C-RV32-RTLIB-COMPILERRT-ILP32: "{{.*}}clang_rt.crtend.o"
 
 // RUN: %clang -### %s --target=riscv32 \
 // RUN:   --gcc-toolchain=%S/Inputs/basic_riscv32_tree --sysroot= \
diff --git a/clang/test/Driver/riscv64-toolchain-extra.c b/clang/test/Driver/riscv64-toolchain-extra.c
index a6ec9b16cc5ca..d8d9b58441676 100644
--- a/clang/test/Driver/riscv64-toolchain-extra.c
+++ b/clang/test/Driver/riscv64-toolchain-extra.c
@@ -29,8 +29,8 @@
 // C-RV64-BAREMETAL-LP64-NOGCC: "-internal-isystem" "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/include"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/riscv64-unknown-elf-ld"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/lib/crt0.o"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/clang_rt.crtbegin-riscv64.o"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/clang_rt.crtbegin.o"
 // C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/bin/../riscv64-unknown-elf/lib"
 // C-RV64-BAREMETAL-LP64-NOGCC: "--start-group" "-lc" "-lgloss" "--end-group"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/libclang_rt.builtins-riscv64.a"
-// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/lib/clang_rt.crtend-riscv64.o"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/libclang_rt.builtins.a"
+// C-RV64-BAREMETAL-LP64-NOGCC: "{{.*}}/riscv64-nogcc/{{.*}}/riscv64-unknown-unknown-elf/clang_rt.crtend.o"
diff --git a/clang/test/Driver/riscv64-toolchain.c b/clang/test/Driver/riscv64-toolchain.c
index 381ee58c470c5..b3216de307540 100644
--- a/clang/test/Driver/riscv64-toolchain.c
+++ b/clang/test/Driver/riscv64-toolchain.c
@@ -151,9 +151,9 @@
 // RUN:   --target=riscv64-unknown-elf --rtlib=compiler-rt --unwindlib=compiler-rt 2>&1 \
 // RUN:   | FileCheck -check-prefix=C-RV64-RTLIB-COMPILERRT-LP64 %s
 // C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}crt0.o"
-// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtbegin-riscv64.o"
-// C-RV64-RTLIB-COMPILERRT-LP64: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins-riscv64.a"
-// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtend-riscv64.o"
+// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtbegin.o"
+// C-RV64-RTLIB-COMPILERRT-LP64: "--start-group" "-lc" "-lgloss" "--end-group" "{{.*}}libclang_rt.builtins.a"
+// C-RV64-RTLIB-COMPILERRT-LP64: "{{.*}}clang_rt.crtend.o"
 
 // RUN: %clang -### %s --target=riscv64 \
 // RUN:   --gcc-toolchain=%S/Inputs/basic_riscv64_tree --sysroot= \
diff --git a/clang/test/Driver/sanitizer-ld.c b/clang/test/Driver/sanitizer-ld.c
index f4ac484d6e4cb..f5657e47626e1 100644
--- a/clang/test/Driver/sanitizer-ld.c
+++ b/clang/test/Driver/sanitizer-ld.c
@@ -111,7 +111,7 @@
 // CHECK-ASAN-FREEBSD: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-FREEBSD-NOT: "-lc"
 // CHECK-ASAN-FREEBSD-NOT: libclang_rt.asan_cxx
-// CHECK-ASAN-FREEBSD: freebsd{{/|\\+}}libclang_rt.asan-i386.a"
+// CHECK-ASAN-FREEBSD: freebsd{{/|\\+}}libclang_rt.asan.a"
 // CHECK-ASAN-FREEBSD-NOT: libclang_rt.asan_cxx
 // CHECK-ASAN-FREEBSD-NOT: "--dynamic-list"
 // CHECK-ASAN-FREEBSD: "--export-dynamic"
@@ -135,8 +135,8 @@
 //
 // CHECK-ASAN-LINUX-CXX: "{{(.*[^-.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-LINUX-CXX-NOT: "-lc"
-// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan-i386.a" "--no-whole-archive"
-// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan_cxx-i386.a" "--no-whole-archive"
+// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan.a" "--no-whole-archive"
+// CHECK-ASAN-LINUX-CXX: "--whole-archive" "{{.*}}libclang_rt.asan_cxx.a" "--no-whole-archive"
 // CHECK-ASAN-LINUX-CXX-NOT: "--dynamic-list"
 // CHECK-ASAN-LINUX-CXX: "--export-dynamic"
 // CHECK-ASAN-LINUX-CXX: stdc++
@@ -163,7 +163,7 @@
 //
 // CHECK-ASAN-ARM: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-ARM-NOT: "-lc"
-// CHECK-ASAN-ARM: libclang_rt.asan-arm.a"
+// CHECK-ASAN-ARM: libclang_rt.asan.a"
 //
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=armv7l-linux-gnueabi -fuse-ld=ld -fsanitize=address \
@@ -172,7 +172,7 @@
 //
 // CHECK-ASAN-ARMv7: "{{(.*[^.0-9A-Z_a-z])?}}ld{{(.exe)?}}"
 // CHECK-ASAN-ARMv7-NOT: "-lc"
-// CHECK-ASAN-ARMv7: libclang_rt.asan-arm.a"
+// CHECK-ASAN-ARMv7: libclang_rt.asan.a"
 
 // RUN: %clang -### %s 2>&1 \
 // RUN:     --target=arm-linux-androideabi -fuse-ld=ld -fsanitize=address \
@@ -185,7 +185,7 @@
 // CHECK-ASAN-ANDROID-NOT: "-lc"
 // CHECK-ASAN-ANDROID-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-NOT: "-lresolv"
-// CHECK-ASAN-ANDROID: libclang_rt.asan-arm-android.so"
+// CHECK-ASAN-ANDROID: libclang_rt.asan.so"
 // CHECK-ASAN-ANDROID-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-NOT: "-lresolv"
 
@@ -197,7 +197,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-ASAN-ANDROID-STATICLIBASAN %s
 //
 // CHECK-ASAN-ANDROID-STATICLIBASAN: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-ASAN-ANDROID-STATICLIBASAN: libclang_rt.asan-arm-android.a"
+// CHECK-ASAN-ANDROID-STATICLIBASAN: libclang_rt.asan.a"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lrt"
 // CHECK-ASAN-ANDROID-STATICLIBASAN-NOT: "-lresolv"
@@ -213,7 +213,7 @@
 // CHECK-UBSAN-ANDROID-NOT: "-lc"
 // CHECK-UBSAN-ANDROID-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-NOT: "-lresolv"
-// CHECK-UBSAN-ANDROID: libclang_rt.ubsan_standalone-arm-android.so"
+// CHECK-UBSAN-ANDROID: libclang_rt.ubsan_standalone.so"
 // CHECK-UBSAN-ANDROID-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-NOT: "-lresolv"
 
@@ -225,7 +225,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-UBSAN-ANDROID-STATICLIBASAN %s
 //
 // CHECK-UBSAN-ANDROID-STATICLIBASAN: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
-// CHECK-UBSAN-ANDROID-STATICLIBASAN: libclang_rt.ubsan_standalone-arm-android.a"
+// CHECK-UBSAN-ANDROID-STATICLIBASAN: libclang_rt.ubsan_standalone.a"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lpthread"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lrt"
 // CHECK-UBSAN-ANDROID-STATICLIBASAN-NOT: "-lresolv"
@@ -242,7 +242,7 @@
 // CHECK-ASAN-ANDROID-X86-NOT: "-lc"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lresolv"
-// CHECK-ASAN-ANDROID-X86: libclang_rt.asan-i686-android.so"
+// CHECK-ASAN-ANDROID-X86: libclang_rt.asan.so"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-X86-NOT: "-lresolv"
 //
@@ -263,7 +263,7 @@
 //
 // CHECK-ASAN-ANDROID-SHARED: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lc"
-// CHECK-ASAN-ANDROID-SHARED: libclang_rt.asan-arm-android.so"
+// CHECK-ASAN-ANDROID-SHARED: libclang_rt.asan.so"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lpthread"
 // CHECK-ASAN-ANDROID-SHARED-NOT: "-lresolv"
 
@@ -353,7 +353,7 @@
 // CHECK-UBSAN-LINUX: "{{.*}}ld{{(.exe)?}}"
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.asan
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.ubsan_standalone_cxx
-// CHECK-UBSAN-LINUX: "--whole-archive" "{{.*}}libclang_rt.ubsan_standalone-x32.a" "--no-whole-archive"
+// CHECK-UBSAN-LINUX: "--whole-archive" "{{.*}}libclang_rt.ubsan_standalone.a" "--no-whole-archive"
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.asan
 // CHECK-UBSAN-LINUX-NOT: libclang_rt.ubsan_standalone_cxx
 // CHECK-UBSAN-LINUX-NOT: "-lstdc++"
@@ -684,7 +684,7 @@
 // RUN:     --sysroot=%S/Inputs/basic_android_tree \
 // RUN:   | FileCheck --check-prefix=CHECK-CFI-CROSS-DSO-DIAG-ANDROID %s
 // CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{.*}}ld{{(.exe)?}}"
-// CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{[^"]*}}libclang_rt.ubsan_standalone-aarch64-android.so"
+// CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "{{[^"]*}}libclang_rt.ubsan_standalone.so"
 // CHECK-CFI-CROSS-DSO-DIAG-ANDROID: "--export-dynamic-symbol=__cfi_check"
 
 // RUN: %clangxx -fsanitize=address -### %s 2>&1 \
@@ -935,7 +935,7 @@
 // CHECK-SCUDO-ANDROID: "-pie"
 // CHECK-SCUDO-ANDROID-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-NOT: "-lresolv"
-// CHECK-SCUDO-ANDROID: libclang_rt.scudo_standalone-arm-android.so"
+// CHECK-SCUDO-ANDROID: libclang_rt.scudo_standalone.so"
 // CHECK-SCUDO-ANDROID-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-NOT: "-lresolv"
 
@@ -946,7 +946,7 @@
 // RUN:   | FileCheck --check-prefix=CHECK-SCUDO-ANDROID-STATIC %s
 // CHECK-SCUDO-ANDROID-STATIC: "{{(.*[^.0-9A-Z_a-z])?}}ld.lld{{(.exe)?}}"
 // CHECK-SCUDO-ANDROID-STATIC: "-pie"
-// CHECK-SCUDO-ANDROID-STATIC: "--whole-archive" "{{.*}}libclang_rt.scudo_standalone-arm-android.a" "--no-whole-archive"
+// CHECK-SCUDO-ANDROID-STATIC: "--whole-archive" "{{.*}}libclang_rt.scudo_standalone.a" "--no-whole-archive"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lstdc++"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lpthread"
 // CHECK-SCUDO-ANDROID-STATIC-NOT: "-lrt"
diff --git a/clang/test/Driver/wasm-toolchain.c b/clang/test/Driver/wasm-toolchain.c
index 88590a3ba4c45..dabf0ac2433bb 100644
--- a/clang/test/Driver/wasm-toolchain.c
+++ b/clang/test/Driver/wasm-toolchain.c
@@ -17,42 +17,42 @@
 // RUN: %clang -### --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK %s
 // LINK: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C link command-line with optimization with unknown OS.
 
 // RUN: %clang -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT %s
 // LINK_OPT: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C link command-line with known OS.
 
 // RUN: %clang -### --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN %s
 // LINK_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // -shared should be passed through to `wasm-ld` and include crt1-reactor.o with a known OS.
 
 // RUN: %clang -### -shared -mexec-model=reactor --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN_SHARED %s
 // LINK_KNOWN_SHARED: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN_SHARED: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_KNOWN_SHARED: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // -shared should be passed through to `wasm-ld` and include crt1-reactor.o with an unknown OS.
 
 // RUN: %clang -### -shared -mexec-model=reactor --target=wasm32-unknown-unknown --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_UNKNOWN_SHARED %s
 // LINK_UNKNOWN_SHARED: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_UNKNOWN_SHARED: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_UNKNOWN_SHARED: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "-shared" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C link command-line with optimization with known OS.
 
 // RUN: %clang -### -O2 --target=wasm32-wasi --sysroot=/foo %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN %s
 // LINK_OPT_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C compile command-line with known OS.
 
@@ -180,12 +180,12 @@
 // RUN: %clang -### %s --target=wasm32-unknown-unknown --sysroot=%s/no-sysroot-there -mexec-model=command 2>&1 \
 // RUN:   | FileCheck -check-prefix=CHECK-COMMAND %s
 // CHECK-COMMAND: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// CHECK-COMMAND: wasm-ld{{.*}}" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// CHECK-COMMAND: wasm-ld{{.*}}" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // RUN: %clang -### %s --target=wasm32-unknown-unknown --sysroot=%s/no-sysroot-there -mexec-model=reactor 2>&1 \
 // RUN:   | FileCheck -check-prefix=CHECK-REACTOR %s
 // CHECK-REACTOR: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// CHECK-REACTOR: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// CHECK-REACTOR: wasm-ld{{.*}}" "crt1-reactor.o" "--entry" "_initialize" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // -fPIC implies +mutable-globals
 
@@ -204,7 +204,7 @@
 // RUN: %clang -### -O2 --target=wasm32-wasip2 %s --sysroot /foo 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_WASIP2 %s
 // LINK_WASIP2: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_WASIP2: wasm-component-ld{{.*}}" "-L/foo/lib/wasm32-wasip2" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_WASIP2: wasm-component-ld{{.*}}" "-L/foo/lib/wasm32-wasip2" "crt1.o" "[[temp]]" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // Test that on `wasm32-wasip2` the `wasm-component-ld` programs is told where
 // to find `wasm-ld` by default.
diff --git a/clang/test/Driver/wasm-toolchain.cpp b/clang/test/Driver/wasm-toolchain.cpp
index 4af011097021f..ba1c55b33edca 100644
--- a/clang/test/Driver/wasm-toolchain.cpp
+++ b/clang/test/Driver/wasm-toolchain.cpp
@@ -17,48 +17,48 @@
 // RUN: %clangxx -### --target=wasm32-unknown-unknown --sysroot=/foo --stdlib=libc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK %s
 // LINK: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // RUN: %clangxx -### --target=wasm32-unknown-unknown --sysroot=/foo --stdlib=libstdc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_STDCXX %s
 // LINK_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C++ link command-line with optimization with unknown OS.
 
 // RUN: %clangxx -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s --stdlib=libc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT %s
 // LINK_OPT: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // RUN: %clangxx -### -O2 --target=wasm32-unknown-unknown --sysroot=/foo %s --stdlib=libstdc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_STDCXX %s
 // LINK_OPT_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT_STDCXX: wasm-ld{{.*}}" "-L/foo/lib" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C++ link command-line with known OS.
 
 // RUN: %clangxx -### --target=wasm32-wasi --sysroot=/foo --stdlib=libc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN %s
 // LINK_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // RUN: %clangxx -### --target=wasm32-wasi --sysroot=/foo --stdlib=libstdc++ %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_KNOWN_STDCXX %s
 // LINK_KNOWN_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C++ link command-line with optimization with known OS.
 
 // RUN: %clangxx -### -O2 --target=wasm32-wasi --sysroot=/foo %s --stdlib=libc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN %s
 // LINK_OPT_KNOWN: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT_KNOWN: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lc++" "-lc++abi" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // RUN: %clangxx -### -O2 --target=wasm32-wasi --sysroot=/foo %s --stdlib=libstdc++ 2>&1 \
 // RUN:   | FileCheck -check-prefix=LINK_OPT_KNOWN_STDCXX %s
 // LINK_OPT_KNOWN_STDCXX: "-cc1" {{.*}} "-o" "[[temp:[^"]*]]"
-// LINK_OPT_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins-wasm32.a" "-o" "a.out"
+// LINK_OPT_KNOWN_STDCXX: wasm-ld{{.*}}" "-L/foo/lib/wasm32-wasi" "crt1.o" "[[temp]]" "-lstdc++" "-lc" "{{.*[/\\]}}libclang_rt.builtins.a" "-o" "a.out"
 
 // A basic C++ compile command-line with known OS.
 
diff --git a/clang/test/Driver/windows-cross.c b/clang/test/Driver/windows-cross.c
index 75490b992d78d..f6e831f00e13a 100644
--- a/clang/test/Driver/windows-cross.c
+++ b/clang/test/Driver/windows-cross.c
@@ -11,32 +11,32 @@
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %s/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libstdc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-RTLIB
 
-// CHECK-RTLIB: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-RTLIB: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-C-LIBCXX
 
-// CHECK-C-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-C-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clangxx -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -rtlib=compiler-rt -stdlib=libc++ -o /dev/null %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-LIBCXX
 
-// CHECK-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lc++" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-LIBCXX: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-Bdynamic" "--entry" "mainCRTStartup" "--allow-multiple-definition" "-o" "{{[^"]*}}" "{{.*}}.o" "-lc++" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-SHARED
 
-// CHECK-SHARED: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-SHARED: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -static -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-SHARED-STATIC
 
-// CHECK-SHARED-STATIC: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bstatic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-SHARED-STATIC: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bstatic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %s/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -nostartfiles -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-NOSTARTFILES
 
-// CHECK-NOSTARTFILES: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins-arm.lib"
+// CHECK-NOSTARTFILES: {{[/\\]}}ld" "--sysroot={{.*}}/Inputs/Windows/ARM/8.1" "-m" "thumb2pe" "-shared" "-Bdynamic" "--enable-auto-image-base" "--entry" "_DllMainCRTStartup" "--allow-multiple-definition" "-o" "shared.dll" "--out-implib" "shared.lib" "{{.*}}.o" "-lmsvcrt" "{{.*[\\/]}}clang_rt.builtins.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=ld -shared -rtlib=compiler-rt -stdlib=libc++ -nostartfiles -nodefaultlibs -o shared.dll %s 2>&1 \
 // RUN:   | FileCheck %s --check-prefix CHECK-STANDALONE
@@ -52,19 +52,19 @@
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS
 
 // CHECK-SANITIZE-ADDRESS: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS: "{{.*}}clang_rt.asan_dll_thunk-arm.lib"
+// CHECK-SANITIZE-ADDRESS: "{{.*}}clang_rt.asan_dll_thunk.lib"
 
 // RUN: %clang -### -target armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -o test.exe -fsanitize=address -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS-EXE
 
 // CHECK-SANITIZE-ADDRESS-EXE: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS-EXE: "{{.*}}clang_rt.asan_dynamic-arm.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk-arm.lib" "--undefined" "__asan_seh_interceptor"
+// CHECK-SANITIZE-ADDRESS-EXE: "{{.*}}clang_rt.asan_dynamic.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib" "--undefined" "__asan_seh_interceptor"
 
 // RUN: %clang -### -target i686-windows-itanium -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -o test.exe -fsanitize=address -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-ADDRESS-EXE-X86
 
 // CHECK-SANITIZE-ADDRESS-EXE-X86: "-fsanitize=address"
-// CHECK-SANITIZE-ADDRESS-EXE-X86: "{{.*}}clang_rt.asan_dynamic-i386.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk-i386.lib" "--undefined" "___asan_seh_interceptor"
+// CHECK-SANITIZE-ADDRESS-EXE-X86: "{{.*}}clang_rt.asan_dynamic.lib" "{{.*}}clang_rt.asan_dynamic_runtime_thunk.lib" "--undefined" "___asan_seh_interceptor"
 
 // RUN: not %clang -### --target=armv7-windows-itanium --sysroot %S/Inputs/Windows/ARM/8.1 -B %S/Inputs/Windows/ARM/8.1/usr/bin -fuse-ld=lld-link2 -shared -o shared.dll -fsanitize=tsan -x c++ %s 2>&1 \
 // RUN:    | FileCheck %s --check-prefix CHECK-SANITIZE-TSAN
diff --git a/clang/test/Driver/zos-ld.c b/clang/test/Driver/zos-ld.c
index 4d4decdd0e65b..87d169936e129 100644
--- a/clang/test/Driver/zos-ld.c
+++ b/clang/test/Driver/zos-ld.c
@@ -14,7 +14,7 @@
 // C-LD-SAME: "-S" "//'SYS1.CSSLIB'"
 // C-LD-SAME: "//'CEE.SCEELIB(CELQS001)'"
 // C-LD-SAME: "//'CEE.SCEELIB(CELQS003)'"
-// C-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// C-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
 
 // 2. General C link for dll
 // RUN: %clang -### --shared --target=s390x-ibm-zos %s 2>&1 \
@@ -30,7 +30,7 @@
 // C-LD-DLL-SAME: "-S" "//'SYS1.CSSLIB'"
 // C-LD-DLL-SAME: "//'CEE.SCEELIB(CELQS001)'"
 // C-LD-DLL-SAME: "//'CEE.SCEELIB(CELQS003)'"
-// C-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// C-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
 
 // 3. General C++ link for executable
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -52,7 +52,7 @@
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQCXA)'"
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQXLA)'"
 // CXX-LD-SAME: "//'CEE.SCEELIB(CRTDQUNW)'"
-// CXX-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// CXX-LD-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
 
 // 4. General C++ link for dll
 // RUN: %clangxx -### --shared --target=s390x-ibm-zos %s 2>&1 \
@@ -74,7 +74,7 @@
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQCXA)'"
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQXLA)'"
 // CXX-LD-DLL-SAME: "//'CEE.SCEELIB(CRTDQUNW)'"
-// CXX-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// CXX-LD-DLL-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
 
 // 5. C++ link for executable w/ -mzos-hlq-le=, -mzos-hlq-csslib=
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -97,7 +97,7 @@
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQCXA)'"
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQXLA)'"
 // CXX-LD5-SAME: "//'AAAA.SCEELIB(CRTDQUNW)'"
-// CXX-LD5-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// CXX-LD5-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
 
 // 6. C++ link for executable w/ -mzos-hlq-clang=
 // RUN: %clangxx -### --target=s390x-ibm-zos %s 2>&1 \
@@ -120,4 +120,4 @@
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQCXA)'"
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQXLA)'"
 // CXX-LD6-SAME: "//'AAAA.SCEELIB(CRTDQUNW)'"
-// CXX-LD6-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}zos{{/|\\\\}}libclang_rt.builtins-s390x.a"
+// CXX-LD6-SAME: "[[RESOURCE_DIR]]{{/|\\\\}}lib{{/|\\\\}}s390x-ibm-zos{{/|\\\\}}libclang_rt.builtins.a"
diff --git a/flang/test/Driver/msvc-dependent-lib-flags.f90 b/flang/test/Driver/msvc-dependent-lib-flags.f90
index 7c1f962e339f9..643dbe9e949cb 100644
--- a/flang/test/Driver/msvc-dependent-lib-flags.f90
+++ b/flang/test/Driver/msvc-dependent-lib-flags.f90
@@ -4,7 +4,7 @@
 ! RUN: %flang -### --target=aarch64-windows-msvc -fms-runtime-lib=dll_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL-DEBUG
 
 ! MSVC: -fc1
-! MSVC-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
+! MSVC-SAME: --dependent-lib=clang_rt.builtins.lib
 ! MSVC-SAME: -D_MT
 ! MSVC-SAME: --dependent-lib=libcmt
 ! MSVC-SAME: --dependent-lib=Fortran_main.static.lib
@@ -12,7 +12,7 @@
 ! MSVC-SAME: --dependent-lib=FortranDecimal.static.lib
 
 ! MSVC-DEBUG: -fc1
-! MSVC-DEBUG-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
+! MSVC-DEBUG-SAME: --dependent-lib=clang_rt.builtins.lib
 ! MSVC-DEBUG-SAME: -D_MT
 ! MSVC-DEBUG-SAME: -D_DEBUG
 ! MSVC-DEBUG-SAME: --dependent-lib=libcmtd
@@ -21,7 +21,7 @@
 ! MSVC-DEBUG-SAME: --dependent-lib=FortranDecimal.static_dbg.lib
 
 ! MSVC-DLL: -fc1
-! MSVC-DLL-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
+! MSVC-DLL-SAME: --dependent-lib=clang_rt.builtins.lib
 ! MSVC-DLL-SAME: -D_MT
 ! MSVC-DLL-SAME: -D_DLL
 ! MSVC-DLL-SAME: --dependent-lib=msvcrt
@@ -30,7 +30,7 @@
 ! MSVC-DLL-SAME: --dependent-lib=FortranDecimal.dynamic.lib
 
 ! MSVC-DLL-DEBUG: -fc1
-! MSVC-DLL-DEBUG-SAME: --dependent-lib=clang_rt.builtins-aarch64.lib
+! MSVC-DLL-DEBUG-SAME: --dependent-lib=clang_rt.builtins.lib
 ! MSVC-DLL-DEBUG-SAME: -D_MT
 ! MSVC-DLL-DEBUG-SAME: -D_DEBUG
 ! MSVC-DLL-DEBUG-SAME: -D_DLL

From e8232c325f7c56c175b27a3538699c31519c8f5e Mon Sep 17 00:00:00 2001
From: Andreas Jonson <andjo403@hotmail.com>
Date: Mon, 15 Apr 2024 02:54:37 +0200
Subject: [PATCH 368/886] [ADT] In FoldingSet calulation for APInt add bit
 width to avoid crash (#88635)

Multiple ranges attributes with the same lower and upper limit but
different bit width caused a crash.
---
 llvm/include/llvm/ADT/FoldingSet.h | 1 +
 llvm/test/Bitcode/attributes.ll    | 5 +++++
 2 files changed, 6 insertions(+)

diff --git a/llvm/include/llvm/ADT/FoldingSet.h b/llvm/include/llvm/ADT/FoldingSet.h
index ddc3e52255d6c..28a84d9658838 100644
--- a/llvm/include/llvm/ADT/FoldingSet.h
+++ b/llvm/include/llvm/ADT/FoldingSet.h
@@ -356,6 +356,7 @@ class FoldingSetNodeID {
     AddInteger(unsigned(I >> 32));
   }
   void AddInteger(const APInt &Int) {
+    AddInteger(Int.getBitWidth());
     const auto *Parts = Int.getRawData();
     for (int i = 0, N = Int.getNumWords(); i < N; ++i) {
       AddInteger(Parts[i]);
diff --git a/llvm/test/Bitcode/attributes.ll b/llvm/test/Bitcode/attributes.ll
index 26163b4d38c89..0c753e58d4358 100644
--- a/llvm/test/Bitcode/attributes.ll
+++ b/llvm/test/Bitcode/attributes.ll
@@ -531,6 +531,11 @@ define range(i32 -1, 42) i32 @range_attribute(<4 x i32> range(i32 -1, 42) %a) {
   ret i32 0
 }
 
+; CHECK: define range(i32 0, 42) i32 @range_attribute_same_range_other_bitwidth(i8 range(i8 0, 42) %a)
+define range(i32 0, 42) i32 @range_attribute_same_range_other_bitwidth(i8 range(i8 0, 42) %a) {
+  ret i32 0
+}
+
 ; CHECK: define void @wide_range_attribute(i128 range(i128 618970019642690137449562111, 618970019642690137449562114) %a)
 define void @wide_range_attribute(i128 range(i128 618970019642690137449562111, 618970019642690137449562114) %a) {
   ret void

From 8e4b0890a61088ae55b4bccbb59e5c5e10d28385 Mon Sep 17 00:00:00 2001
From: wanglei <wanglei@loongson.cn>
Date: Mon, 15 Apr 2024 09:01:04 +0800
Subject: [PATCH 369/886] [LoongArch] Return true from
 shouldConsiderGEPOffsetSplit (#88371)

If not performing gep splits can prevent important optimizations, such
as preventing the element indices / member offsets from being
(partially) folded into load/store instruction immediates.
---
 llvm/lib/Target/LoongArch/LoongArchISelLowering.h         | 1 +
 llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll | 6 +++---
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index 9e9ac0b826929..4bb6c049f8d75 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -238,6 +238,7 @@ class LoongArchTargetLowering : public TargetLowering {
   bool isShuffleMaskLegal(ArrayRef<int> Mask, EVT VT) const override {
     return false;
   }
+  bool shouldConsiderGEPOffsetSplit() const override { return true; }
 
 private:
   /// Target-specific function used to lower LoongArch calling conventions.
diff --git a/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll b/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll
index ca2a16a0ca7ec..20cc25e95adf9 100644
--- a/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll
+++ b/llvm/test/Transforms/CodeGenPrepare/LoongArch/splitgep.ll
@@ -4,9 +4,9 @@
 define void @test(ptr %sp, ptr %t, i32 %n) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    %s = load ptr, ptr %sp, align 8
-; CHECK-NEXT:    br label %while_cond
-
+; CHECK-NEXT:    %splitgep1 = getelementptr i8, ptr %t, i64 80000
+; CHECK-NEXT:    %s = load ptr, ptr %sp
+; CHECK-NEXT:    %splitgep = getelementptr i8, ptr %s, i64 80000
 entry:
   %s = load ptr, ptr %sp
   br label %while_cond

From dd4b40ffe3bd9da2e3d998b674305ae1f612ad51 Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97@outlook.com>
Date: Mon, 15 Apr 2024 09:14:30 +0800
Subject: [PATCH 370/886] [CodeGen] Let `CodeGenPassBuilder` know concrete
 target machine (#88614)

Many backends passes accept only `<Target>TargetMachine`, this can avoid
cast to derived type.
---
 llvm/include/llvm/Passes/CodeGenPassBuilder.h | 107 +++++++++---------
 llvm/lib/Target/X86/X86CodeGenPassBuilder.cpp |   5 +-
 2 files changed, 59 insertions(+), 53 deletions(-)

diff --git a/llvm/include/llvm/Passes/CodeGenPassBuilder.h b/llvm/include/llvm/Passes/CodeGenPassBuilder.h
index ed5d8affa7938..2e94a19502131 100644
--- a/llvm/include/llvm/Passes/CodeGenPassBuilder.h
+++ b/llvm/include/llvm/Passes/CodeGenPassBuilder.h
@@ -109,9 +109,9 @@ namespace llvm {
 /// construction. The \c MachinePassRegistry.def file specifies how to construct
 /// all of the built-in passes, and those may reference these members during
 /// construction.
-template <typename DerivedT> class CodeGenPassBuilder {
+template <typename DerivedT, typename TargetMachineT> class CodeGenPassBuilder {
 public:
-  explicit CodeGenPassBuilder(LLVMTargetMachine &TM,
+  explicit CodeGenPassBuilder(TargetMachineT &TM,
                               const CGPassBuilderOption &Opts,
                               PassInstrumentationCallbacks *PIC)
       : TM(TM), Opt(Opts), PIC(PIC) {
@@ -237,7 +237,7 @@ template <typename DerivedT> class CodeGenPassBuilder {
     const DerivedT &PB;
   };
 
-  LLVMTargetMachine &TM;
+  TargetMachineT &TM;
   CGPassBuilderOption Opt;
   PassInstrumentationCallbacks *PIC;
 
@@ -498,8 +498,8 @@ template <typename DerivedT> class CodeGenPassBuilder {
   mutable bool Stopped = true;
 };
 
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::buildPipeline(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::buildPipeline(
     ModulePassManager &MPM, raw_pwrite_stream &Out, raw_pwrite_stream *DwoOut,
     CodeGenFileType FileType) const {
   auto StartStopInfo = TargetPassConfig::getStartStopInfo(*PIC);
@@ -542,8 +542,8 @@ Error CodeGenPassBuilder<Derived>::buildPipeline(
   return verifyStartStop(*StartStopInfo);
 }
 
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::setStartStopPasses(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::setStartStopPasses(
     const TargetPassConfig::StartStopInfo &Info) const {
   if (!Info.StartPass.empty()) {
     Started = false;
@@ -585,8 +585,8 @@ void CodeGenPassBuilder<Derived>::setStartStopPasses(
   }
 }
 
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::verifyStartStop(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::verifyStartStop(
     const TargetPassConfig::StartStopInfo &Info) const {
   if (Started && Stopped)
     return Error::success();
@@ -604,8 +604,9 @@ Error CodeGenPassBuilder<Derived>::verifyStartStop(
   return Error::success();
 }
 
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addISelPasses(AddIRPass &addPass) const {
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addISelPasses(
+    AddIRPass &addPass) const {
   derived().addGlobalMergePass(addPass);
   if (TM.useEmulatedTLS())
     addPass(LowerEmuTLSPass());
@@ -620,8 +621,9 @@ void CodeGenPassBuilder<Derived>::addISelPasses(AddIRPass &addPass) const {
 
 /// Add common target configurable passes that perform LLVM IR to IR transforms
 /// following machine independent optimization.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addIRPasses(AddIRPass &addPass) const {
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addIRPasses(
+    AddIRPass &addPass) const {
   // Before running any passes, run the verifier to determine if the input
   // coming from the front-end and/or optimizer is valid.
   if (!Opt.DisableVerify)
@@ -686,8 +688,8 @@ void CodeGenPassBuilder<Derived>::addIRPasses(AddIRPass &addPass) const {
 
 /// Turn exception handling constructs into something the code generators can
 /// handle.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addPassesToHandleExceptions(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addPassesToHandleExceptions(
     AddIRPass &addPass) const {
   const MCAsmInfo *MCAI = TM.getMCAsmInfo();
   assert(MCAI && "No MCAsmInfo");
@@ -733,8 +735,9 @@ void CodeGenPassBuilder<Derived>::addPassesToHandleExceptions(
 
 /// Add pass to prepare the LLVM IR for code generation. This should be done
 /// before exception handling preparation passes.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addCodeGenPrepare(AddIRPass &addPass) const {
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addCodeGenPrepare(
+    AddIRPass &addPass) const {
   if (getOptLevel() != CodeGenOptLevel::None && !Opt.DisableCGP)
     addPass(CodeGenPreparePass(&TM));
   // TODO: Default ctor'd RewriteSymbolPass is no-op.
@@ -743,8 +746,9 @@ void CodeGenPassBuilder<Derived>::addCodeGenPrepare(AddIRPass &addPass) const {
 
 /// Add common passes that perform LLVM IR to IR transforms in preparation for
 /// instruction selection.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addISelPrepare(AddIRPass &addPass) const {
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addISelPrepare(
+    AddIRPass &addPass) const {
   derived().addPreISel(addPass);
 
   addPass(CallBrPreparePass());
@@ -763,8 +767,8 @@ void CodeGenPassBuilder<Derived>::addISelPrepare(AddIRPass &addPass) const {
     addPass(VerifierPass());
 }
 
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::addCoreISelPasses(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::addCoreISelPasses(
     AddMachinePass &addPass) const {
   // Enable FastISel with -fast-isel, but allow that to be overridden.
   TM.setO0WantsFastISel(Opt.EnableFastISelOption.value_or(true));
@@ -846,17 +850,18 @@ Error CodeGenPassBuilder<Derived>::addCoreISelPasses(
 /// with nontrivial configuration or multiple passes are broken out below in
 /// add%Stage routines.
 ///
-/// Any CodeGenPassBuilder<Derived>::addXX routine may be overriden by the
-/// Target. The addPre/Post methods with empty header implementations allow
-/// injecting target-specific fixups just before or after major stages.
-/// Additionally, targets have the flexibility to change pass order within a
-/// stage by overriding default implementation of add%Stage routines below. Each
-/// technique has maintainability tradeoffs because alternate pass orders are
-/// not well supported. addPre/Post works better if the target pass is easily
-/// tied to a common pass. But if it has subtle dependencies on multiple passes,
-/// the target should override the stage instead.
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::addMachinePasses(
+/// Any CodeGenPassBuilder<Derived, TargetMachine>::addXX routine may be
+/// overriden by the Target. The addPre/Post methods with empty header
+/// implementations allow injecting target-specific fixups just before or after
+/// major stages. Additionally, targets have the flexibility to change pass
+/// order within a stage by overriding default implementation of add%Stage
+/// routines below. Each technique has maintainability tradeoffs because
+/// alternate pass orders are not well supported. addPre/Post works better if
+/// the target pass is easily tied to a common pass. But if it has subtle
+/// dependencies on multiple passes, the target should override the stage
+/// instead.
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::addMachinePasses(
     AddMachinePass &addPass) const {
   // Add passes that optimize machine instructions in SSA form.
   if (getOptLevel() != CodeGenOptLevel::None) {
@@ -962,8 +967,8 @@ Error CodeGenPassBuilder<Derived>::addMachinePasses(
 }
 
 /// Add passes that optimize machine instructions in SSA form.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addMachineSSAOptimization(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addMachineSSAOptimization(
     AddMachinePass &addPass) const {
   // Pre-ra tail duplication.
   addPass(EarlyTailDuplicatePass());
@@ -1014,8 +1019,8 @@ void CodeGenPassBuilder<Derived>::addMachineSSAOptimization(
 /// A target that uses the standard regalloc pass order for fast or optimized
 /// allocation may still override this for per-target regalloc
 /// selection. But -regalloc=... always takes precedence.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addTargetRegisterAllocator(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addTargetRegisterAllocator(
     AddMachinePass &addPass, bool Optimized) const {
   if (Optimized)
     addPass(RAGreedyPass());
@@ -1026,22 +1031,22 @@ void CodeGenPassBuilder<Derived>::addTargetRegisterAllocator(
 /// Find and instantiate the register allocation pass requested by this target
 /// at the current optimization level.  Different register allocators are
 /// defined as separate passes because they may require different analysis.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addRegAllocPass(AddMachinePass &addPass,
-                                                  bool Optimized) const {
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addRegAllocPass(
+    AddMachinePass &addPass, bool Optimized) const {
   // TODO: Parse Opt.RegAlloc to add register allocator.
 }
 
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::addRegAssignmentFast(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::addRegAssignmentFast(
     AddMachinePass &addPass) const {
   // TODO: Ensure allocator is default or fast.
   addRegAllocPass(addPass, false);
   return Error::success();
 }
 
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::addRegAssignmentOptimized(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::addRegAssignmentOptimized(
     AddMachinePass &addPass) const {
   // Add the selected register allocation pass.
   addRegAllocPass(addPass, true);
@@ -1062,8 +1067,8 @@ Error CodeGenPassBuilder<Derived>::addRegAssignmentOptimized(
 
 /// Add the minimum set of target-independent passes that are required for
 /// register allocation. No coalescing or scheduling.
-template <typename Derived>
-Error CodeGenPassBuilder<Derived>::addFastRegAlloc(
+template <typename Derived, typename TargetMachineT>
+Error CodeGenPassBuilder<Derived, TargetMachineT>::addFastRegAlloc(
     AddMachinePass &addPass) const {
   addPass(PHIEliminationPass());
   addPass(TwoAddressInstructionPass());
@@ -1073,8 +1078,8 @@ Error CodeGenPassBuilder<Derived>::addFastRegAlloc(
 /// Add standard target-independent passes that are tightly coupled with
 /// optimized register allocation, including coalescing, machine instruction
 /// scheduling, and register allocation itself.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addOptimizedRegAlloc(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addOptimizedRegAlloc(
     AddMachinePass &addPass) const {
   addPass(DetectDeadLanesPass());
 
@@ -1121,8 +1126,8 @@ void CodeGenPassBuilder<Derived>::addOptimizedRegAlloc(
 //===---------------------------------------------------------------------===//
 
 /// Add passes that optimize machine instructions after register allocation.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addMachineLateOptimization(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addMachineLateOptimization(
     AddMachinePass &addPass) const {
   // Branch folding must be run after regalloc and prolog/epilog insertion.
   addPass(BranchFolderPass());
@@ -1142,8 +1147,8 @@ void CodeGenPassBuilder<Derived>::addMachineLateOptimization(
 }
 
 /// Add standard basic block placement passes.
-template <typename Derived>
-void CodeGenPassBuilder<Derived>::addBlockPlacement(
+template <typename Derived, typename TargetMachineT>
+void CodeGenPassBuilder<Derived, TargetMachineT>::addBlockPlacement(
     AddMachinePass &addPass) const {
   addPass(MachineBlockPlacementPass());
   // Run a separate pass to collect block placement statistics.
diff --git a/llvm/lib/Target/X86/X86CodeGenPassBuilder.cpp b/llvm/lib/Target/X86/X86CodeGenPassBuilder.cpp
index 453754381034e..7c1fb0b99f413 100644
--- a/llvm/lib/Target/X86/X86CodeGenPassBuilder.cpp
+++ b/llvm/lib/Target/X86/X86CodeGenPassBuilder.cpp
@@ -19,9 +19,10 @@ using namespace llvm;
 
 namespace {
 
-class X86CodeGenPassBuilder : public CodeGenPassBuilder<X86CodeGenPassBuilder> {
+class X86CodeGenPassBuilder
+    : public CodeGenPassBuilder<X86CodeGenPassBuilder, X86TargetMachine> {
 public:
-  explicit X86CodeGenPassBuilder(LLVMTargetMachine &TM,
+  explicit X86CodeGenPassBuilder(X86TargetMachine &TM,
                                  const CGPassBuilderOption &Opts,
                                  PassInstrumentationCallbacks *PIC)
       : CodeGenPassBuilder(TM, Opts, PIC) {}

From 7476b412d46322f52a7f58ad1d94a4398df941e2 Mon Sep 17 00:00:00 2001
From: Chen Zheng <czhengsz@cn.ibm.com>
Date: Sun, 14 Apr 2024 21:22:48 -0400
Subject: [PATCH 371/886] [NFC] add myself as the code owner of the PowerPC
 target

Norminated by Nemanja in https://discourse.llvm.org/t/powerpc-code-owner/78142
---
 llvm/CODE_OWNERS.TXT | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/llvm/CODE_OWNERS.TXT b/llvm/CODE_OWNERS.TXT
index 48026bd2f74eb..d1620d1cbf870 100644
--- a/llvm/CODE_OWNERS.TXT
+++ b/llvm/CODE_OWNERS.TXT
@@ -79,8 +79,8 @@ N: Hal Finkel
 E: hfinkel@anl.gov
 D: The loop reroller and alias analysis
 
-N: Nemanja Ivanovic
-E: nemanja.i.ibm@gmail.com
+N: Zheng Chen
+E: czhengsz@cn.ibm.com
 D: PowerPC Backend
 
 N: Dan Gohman

From a9d7ad23fa625e7f52f2641dbb0f90e8ea12bebf Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Mon, 15 Apr 2024 10:22:11 +0900
Subject: [PATCH 372/886] [InstCombine] Relax shamt assertion in fsh fold

Allow the result of the comparison to contain poison elements,
which happens if one of the elements in the input vector is
poison.
---
 llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp |  4 ++--
 llvm/test/Transforms/InstCombine/rotate.ll           | 11 +++++++++++
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index b35ab8751a211..20f51c8af617d 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -1979,8 +1979,8 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
       if (ModuloC != ShAmtC)
         return replaceOperand(*II, 2, ModuloC);
 
-      assert(ConstantExpr::getICmp(ICmpInst::ICMP_UGT, WidthC, ShAmtC) ==
-                 ConstantInt::getTrue(CmpInst::makeCmpResultType(Ty)) &&
+      assert(match(ConstantExpr::getICmp(ICmpInst::ICMP_UGT, WidthC, ShAmtC),
+                   m_One()) &&
              "Shift amount expected to be modulo bitwidth");
 
       // Canonicalize funnel shift right by constant to funnel shift left. This
diff --git a/llvm/test/Transforms/InstCombine/rotate.ll b/llvm/test/Transforms/InstCombine/rotate.ll
index edd9bf39ca076..6c70c791fd881 100644
--- a/llvm/test/Transforms/InstCombine/rotate.ll
+++ b/llvm/test/Transforms/InstCombine/rotate.ll
@@ -147,6 +147,17 @@ define <2 x i32> @rotr_v2i32_constant_nonsplat_undef0(<2 x i32> %x) {
   ret <2 x i32> %r
 }
 
+define <2 x i32> @rotr_v2i32_constant_nonsplat_poison0(<2 x i32> %x) {
+; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_poison0(
+; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 poison, i32 19>)
+; CHECK-NEXT:    ret <2 x i32> [[R]]
+;
+  %shl = shl <2 x i32> %x, <i32 poison, i32 19>
+  %shr = lshr <2 x i32> %x, <i32 15, i32 13>
+  %r = or <2 x i32> %shl, %shr
+  ret <2 x i32> %r
+}
+
 define <2 x i32> @rotr_v2i32_constant_nonsplat_undef1(<2 x i32> %x) {
 ; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_undef1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 17, i32 0>)

From 9c3475acc9b11d08f0e0b3e230429e9f552d5f80 Mon Sep 17 00:00:00 2001
From: Billy Zhu <billyzhu@modular.com>
Date: Sun, 14 Apr 2024 19:17:20 -0700
Subject: [PATCH 373/886] [MLIR][LLVM] Fix recursive DI type export memory leak
 (#88122)

Followup to discussion
https://github.com/llvm/llvm-project/pull/87295#discussion_r1556016667.

The export cache should not cache temporary nodes.
---
 mlir/lib/Target/LLVMIR/DebugTranslation.cpp | 3 ++-
 mlir/test/Target/LLVMIR/llvmir-debug.mlir   | 8 ++++++--
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/mlir/lib/Target/LLVMIR/DebugTranslation.cpp b/mlir/lib/Target/LLVMIR/DebugTranslation.cpp
index 46e2e7f2ba5dc..2de5e372d88c0 100644
--- a/mlir/lib/Target/LLVMIR/DebugTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/DebugTranslation.cpp
@@ -352,7 +352,8 @@ llvm::DINode *DebugTranslation::translate(DINodeAttr attr) {
                      DISubroutineTypeAttr>(
                    [&](auto attr) { return translateImpl(attr); });
 
-  attrToNode.insert({attr, node});
+  if (node && !node->isTemporary())
+    attrToNode.insert({attr, node});
   return node;
 }
 
diff --git a/mlir/test/Target/LLVMIR/llvmir-debug.mlir b/mlir/test/Target/LLVMIR/llvmir-debug.mlir
index f4c18bf6bd53c..8ab1a1b290dad 100644
--- a/mlir/test/Target/LLVMIR/llvmir-debug.mlir
+++ b/mlir/test/Target/LLVMIR/llvmir-debug.mlir
@@ -442,12 +442,16 @@ llvm.mlir.global @global_variable() {dbg_expr = #di_global_variable_expression}
 #di_subprogram = #llvm.di_subprogram<scope = #di_file, file = #di_file, subprogramFlags = Optimized, type = #di_subroutine_type>
 #di_composite_type = #llvm.di_composite_type<tag = DW_TAG_class_type, recId = distinct[0]<>, scope = #di_subprogram>
 
-#di_global_variable = #llvm.di_global_variable<file = #di_file, line = 1, type = #di_composite_type>
+// Use the inner type standalone outside too. Ensures it's not cached wrong.
+#di_var_type = #llvm.di_subroutine_type<types = #di_composite_type, #di_composite_type_inner>
+#di_global_variable = #llvm.di_global_variable<file = #di_file, line = 1, type = #di_var_type>
 #di_global_variable_expression = #llvm.di_global_variable_expression<var = #di_global_variable>
 
 llvm.mlir.global @global_variable() {dbg_expr = #di_global_variable_expression} : !llvm.struct<()>
 
-// CHECK: distinct !DIGlobalVariable({{.*}}type: ![[COMP:[0-9]+]],
+// CHECK: distinct !DIGlobalVariable({{.*}}type: ![[VAR:[0-9]+]],
+// CHECK: ![[VAR]] = !DISubroutineType(types: ![[COMPS:[0-9]+]])
+// CHECK: ![[COMPS]] = !{![[COMP:[0-9]+]],
 // CHECK: ![[COMP]] = distinct !DICompositeType({{.*}}scope: ![[SCOPE:[0-9]+]],
 // CHECK: ![[SCOPE]] = !DISubprogram({{.*}}type: ![[SUBROUTINE:[0-9]+]],
 // CHECK: ![[SUBROUTINE]] = !DISubroutineType(types: ![[SR_TYPES:[0-9]+]])

From 37b7207651b44743909a427b5509bed5e6c21b59 Mon Sep 17 00:00:00 2001
From: Allen <zhongyunde@huawei.com>
Date: Mon, 15 Apr 2024 10:19:49 +0800
Subject: [PATCH 374/886] [SimplifyCFG] Fix crash when there is unreachable
 large index (#88616)

The large case index out of scope is dead code, but it is still be
created for TableContents in SwitchLookupTable::SwitchLookupTable,
so make sure the table size after growing should not get smaller.

Fix https://github.com/llvm/llvm-project/issues/88607
---
 llvm/lib/Transforms/Utils/SimplifyCFG.cpp     |  4 ++-
 .../Transforms/SimplifyCFG/switch_mask.ll     | 25 +++++++++++++++++++
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
index 55bbffb18879f..da3eb229c041f 100644
--- a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -6783,9 +6783,11 @@ static bool SwitchToLookupTable(SwitchInst *SI, IRBuilder<> &Builder,
           return SwitchLookupTable::WouldFitInRegister(
               DL, UpperBound, KV.second /* ResultType */);
         })) {
+      // There may be some case index larger than the UpperBound (unreachable
+      // case), so make sure the table size does not get smaller.
+      TableSize = std::max(UpperBound, TableSize);
       // The default branch is unreachable after we enlarge the lookup table.
       // Adjust DefaultIsReachable to reuse code path.
-      TableSize = UpperBound;
       DefaultIsReachable = false;
     }
   }
diff --git a/llvm/test/Transforms/SimplifyCFG/switch_mask.ll b/llvm/test/Transforms/SimplifyCFG/switch_mask.ll
index 17f7b583ee840..51c15774d4762 100644
--- a/llvm/test/Transforms/SimplifyCFG/switch_mask.ll
+++ b/llvm/test/Transforms/SimplifyCFG/switch_mask.ll
@@ -214,3 +214,28 @@ lor.end:                                          ; preds = %default, %for.end,
   %retval.0.i.i = phi i32 [ 1, %default ], [ 0, %for.end ], [ 0, %for.end ], [ 0, %for.end ]
   ret void
 }
+
+define i1 @pr88607() {
+; CHECK-LABEL: @pr88607(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[COND:%.*]] = select i1 false, i32 4, i32 1
+; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = select i1 false, i32 2, i32 [[COND]]
+; CHECK-NEXT:    ret i1 false
+;
+entry:
+  %cond = select i1 false, i32 4, i32 1
+  %spec.select = select i1 false, i32 2, i32 %cond
+  switch i32 %spec.select, label %lor.rhs [
+  i32 0, label %exit
+  i32 5, label %exit ; unreachable large case index
+  i32 1, label %exit
+  ]
+
+lor.rhs:                                        ; preds = %entry
+  br label %exit
+
+exit: ; preds = %lor.rhs, %entry, %entry, %entry, %entry
+  %res.ph = phi i1 [ false, %entry ], [ false, %lor.rhs ], [ false, %entry ], [ false, %entry ]
+  ret i1 %res.ph
+}
+

From ccea9f2ae81d140a2367a82069cd00d6b08fe855 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 15:29:58 -0500
Subject: [PATCH 375/886] [ValueTracking] Add tests for `computeKnownBits` of
 `llvm.vector.reduce.{or,and}`; NFC

---
 .../test/Transforms/InstCombine/known-bits.ll | 52 +++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index fea2f647ee2a8..cdc3e991a3fb5 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -1195,5 +1195,57 @@ define i1 @extract_value_smul_fail(i8 %xx, i8 %yy) {
   ret i1 %r
 }
 
+define i8 @known_reduce_or(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_or(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @known_reduce_or_fail(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_or_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 4
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = and i8 %v, 4
+  ret i8 %r
+}
+
+define i8 @known_reduce_and(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_and(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @known_reduce_and_fail(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_and_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = and i8 %v, 2
+  ret i8 %r
+}
+
 declare void @use(i1)
 declare void @sink(i8)

From 6063e3c40867e9b9865aee326ff36cb2ebbcf939 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 14:12:50 -0500
Subject: [PATCH 376/886] [ValueTracking] Implement `computeKnownBits` for
 `llvm.vector.reduce.{or,and}`

---
 llvm/lib/Analysis/ValueTracking.cpp            |  6 ++++--
 llvm/test/Transforms/InstCombine/known-bits.ll | 10 ++--------
 2 files changed, 6 insertions(+), 10 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index b83e7e6769c20..359a9dc7907b7 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1624,8 +1624,10 @@ static void computeKnownBitsFromOperator(const Operator *I,
         computeKnownBits(I->getOperand(1), Known2, Depth + 1, Q);
         Known = KnownBits::ssub_sat(Known, Known2);
         break;
-        // for min/max reduce, any bit common to each element in the input vec
-        // is set in the output.
+        // for min/max/and/or reduce, any bit common to each element in the
+        // input vec is set in the output.
+      case Intrinsic::vector_reduce_and:
+      case Intrinsic::vector_reduce_or:
       case Intrinsic::vector_reduce_umax:
       case Intrinsic::vector_reduce_umin:
       case Intrinsic::vector_reduce_smax:
diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index cdc3e991a3fb5..ce415dd406c05 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -1197,10 +1197,7 @@ define i1 @extract_value_smul_fail(i8 %xx, i8 %yy) {
 
 define i8 @known_reduce_or(<2 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_or(
-; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 1
 ;
   %x = or <2 x i8> %xx, <i8 5, i8 3>
   %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
@@ -1223,10 +1220,7 @@ define i8 @known_reduce_or_fail(<2 x i8> %xx) {
 
 define i8 @known_reduce_and(<2 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_and(
-; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 1
 ;
   %x = or <2 x i8> %xx, <i8 5, i8 3>
   %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)

From 44b1523b9a4e7e21677395e75352d9a62efe2790 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 15:30:11 -0500
Subject: [PATCH 377/886] [ValueTracking] Add tests for `computeKnownBits` of
 `llvm.vector.reduce.xor`; NFC

---
 .../test/Transforms/InstCombine/known-bits.ll | 151 ++++++++++++++++++
 1 file changed, 151 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index ce415dd406c05..fa19adbd5580c 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -1241,5 +1241,156 @@ define i8 @known_reduce_and_fail(<2 x i8> %xx) {
   ret i8 %r
 }
 
+define i8 @known_reduce_xor_even(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_even(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.xor(<2 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_even2(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_even2(
+; CHECK-NEXT:    [[X:%.*]] = and <2 x i8> [[XX:%.*]], <i8 15, i8 15>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 16
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = and <2 x i8> %xx, <i8 15, i8 15>
+  %v = call i8 @llvm.vector.reduce.xor(<2 x i8> %x)
+  %r = and i8 %v, 16
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_even_fail(<2 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_even_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <2 x i8> %xx, <i8 5, i8 3>
+  %v = call i8 @llvm.vector.reduce.xor(<2 x i8> %x)
+  %r = and i8 %v, 2
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_odd(<3 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_odd(
+; CHECK-NEXT:    [[X:%.*]] = or <3 x i8> [[XX:%.*]], <i8 5, i8 3, i8 9>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <3 x i8> %xx, <i8 5, i8 3, i8 9>
+  %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_odd2(<3 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_odd2(
+; CHECK-NEXT:    [[X:%.*]] = and <3 x i8> [[XX:%.*]], <i8 15, i8 15, i8 31>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 32
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = and <3 x i8> %xx, <i8 15, i8 15, i8 31>
+  %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
+  %r = and i8 %v, 32
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_odd2_fail(<3 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_odd2_fail(
+; CHECK-NEXT:    [[X:%.*]] = and <3 x i8> [[XX:%.*]], <i8 15, i8 15, i8 31>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 16
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = and <3 x i8> %xx, <i8 15, i8 15, i8 31>
+  %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
+  %r = and i8 %v, 16
+  ret i8 %r
+}
+
+define i8 @known_reduce_xor_odd_fail(<3 x i8> %xx) {
+; CHECK-LABEL: @known_reduce_xor_odd_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <3 x i8> [[XX:%.*]], <i8 5, i8 3, i8 9>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %x = or <3 x i8> %xx, <i8 5, i8 3, i8 9>
+  %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
+  %r = and i8 %v, 2
+  ret i8 %r
+}
+
+define i8 @nonzero_reduce_xor_vscale_even(<vscale x 2 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_xor_vscale_even(
+; CHECK-NEXT:    [[X:%.*]] = or <vscale x 2 x i8> [[XX:%.*]], shufflevector (<vscale x 2 x i8> insertelement (<vscale x 2 x i8> poison, i8 1, i64 0), <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %one = insertelement <vscale x 2 x i8> poison, i8 1, i64 0
+  %ones = shufflevector <vscale x 2 x i8> %one, <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer
+  %x = or <vscale x 2 x i8> %xx, %ones
+  %v = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @nonzero_reduce_xor_vscale_odd_fail(<vscale x 3 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_xor_vscale_odd_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <vscale x 3 x i8> [[XX:%.*]], shufflevector (<vscale x 3 x i8> insertelement (<vscale x 3 x i8> poison, i8 1, i64 0), <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer)
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv3i8(<vscale x 3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %one = insertelement <vscale x 3 x i8> poison, i8 1, i64 0
+  %ones = shufflevector <vscale x 3 x i8> %one, <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer
+  %x = or <vscale x 3 x i8> %xx, %ones
+  %v = call i8 @llvm.vector.reduce.xor.nxv3i8(<vscale x 3 x i8> %x)
+  %r = and i8 %v, 1
+  ret i8 %r
+}
+
+define i8 @nonzero_reduce_xor_vscale_even_fail(<vscale x 2 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_xor_vscale_even_fail(
+; CHECK-NEXT:    [[X:%.*]] = or <vscale x 2 x i8> [[XX:%.*]], shufflevector (<vscale x 2 x i8> insertelement (<vscale x 2 x i8> poison, i8 1, i64 0), <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %one = insertelement <vscale x 2 x i8> poison, i8 1, i64 0
+  %ones = shufflevector <vscale x 2 x i8> %one, <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer
+  %x = or <vscale x 2 x i8> %xx, %ones
+  %v = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> %x)
+  %r = and i8 %v, 2
+  ret i8 %r
+}
+
+define i8 @nonzero_reduce_xor_vscale_odd(<vscale x 3 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_xor_vscale_odd(
+; CHECK-NEXT:    [[X:%.*]] = and <vscale x 3 x i8> [[XX:%.*]], shufflevector (<vscale x 3 x i8> insertelement (<vscale x 3 x i8> poison, i8 1, i64 0), <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer)
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv3i8(<vscale x 3 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
+; CHECK-NEXT:    ret i8 [[R]]
+;
+  %one = insertelement <vscale x 3 x i8> poison, i8 1, i64 0
+  %ones = shufflevector <vscale x 3 x i8> %one, <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer
+  %x = and <vscale x 3 x i8> %xx, %ones
+  %v = call i8 @llvm.vector.reduce.xor.nxv3i8(<vscale x 3 x i8> %x)
+  %r = and i8 %v, 2
+  ret i8 %r
+}
+
 declare void @use(i1)
 declare void @sink(i8)

From 6c71707872e4d2f9aef19d0dbd502752b7fdfba0 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 14:15:54 -0500
Subject: [PATCH 378/886] [ValueTracking] Implement `computeKnownBits` for
 `llvm.vector.reduce.xor`

---
 llvm/lib/Analysis/ValueTracking.cpp           | 15 ++++++++++
 .../test/Transforms/InstCombine/known-bits.ll | 30 ++++---------------
 2 files changed, 21 insertions(+), 24 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 359a9dc7907b7..378e4aa9974e5 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1634,6 +1634,21 @@ static void computeKnownBitsFromOperator(const Operator *I,
       case Intrinsic::vector_reduce_smin:
         computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
         break;
+      case Intrinsic::vector_reduce_xor: {
+        computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
+        // The zeros common to all vecs are zero in the output.
+        // If the number of elements is odd, then the common ones remain. If the
+        // number of elements is even, then the common ones becomes zeros.
+        auto *VecTy = cast<VectorType>(I->getOperand(0)->getType());
+        // Even, so the ones become zeros.
+        bool EvenCnt = VecTy->getElementCount().isKnownEven();
+        if (EvenCnt)
+          Known.Zero |= Known.One;
+        // Maybe even element count so need to clear ones.
+        if (VecTy->isScalableTy() || EvenCnt)
+          Known.One.clearAllBits();
+        break;
+      }
       case Intrinsic::umin:
         computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
         computeKnownBits(I->getOperand(1), Known2, Depth + 1, Q);
diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index fa19adbd5580c..85a21332b0788 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -1243,10 +1243,7 @@ define i8 @known_reduce_and_fail(<2 x i8> %xx) {
 
 define i8 @known_reduce_xor_even(<2 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_xor_even(
-; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v2i8(<2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
   %x = or <2 x i8> %xx, <i8 5, i8 3>
   %v = call i8 @llvm.vector.reduce.xor(<2 x i8> %x)
@@ -1256,10 +1253,7 @@ define i8 @known_reduce_xor_even(<2 x i8> %xx) {
 
 define i8 @known_reduce_xor_even2(<2 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_xor_even2(
-; CHECK-NEXT:    [[X:%.*]] = and <2 x i8> [[XX:%.*]], <i8 15, i8 15>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v2i8(<2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 16
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
   %x = and <2 x i8> %xx, <i8 15, i8 15>
   %v = call i8 @llvm.vector.reduce.xor(<2 x i8> %x)
@@ -1282,10 +1276,7 @@ define i8 @known_reduce_xor_even_fail(<2 x i8> %xx) {
 
 define i8 @known_reduce_xor_odd(<3 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_xor_odd(
-; CHECK-NEXT:    [[X:%.*]] = or <3 x i8> [[XX:%.*]], <i8 5, i8 3, i8 9>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 1
 ;
   %x = or <3 x i8> %xx, <i8 5, i8 3, i8 9>
   %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
@@ -1295,10 +1286,7 @@ define i8 @known_reduce_xor_odd(<3 x i8> %xx) {
 
 define i8 @known_reduce_xor_odd2(<3 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_xor_odd2(
-; CHECK-NEXT:    [[X:%.*]] = and <3 x i8> [[XX:%.*]], <i8 15, i8 15, i8 31>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 32
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
   %x = and <3 x i8> %xx, <i8 15, i8 15, i8 31>
   %v = call i8 @llvm.vector.reduce.xor.v3i8(<3 x i8> %x)
@@ -1334,10 +1322,7 @@ define i8 @known_reduce_xor_odd_fail(<3 x i8> %xx) {
 
 define i8 @nonzero_reduce_xor_vscale_even(<vscale x 2 x i8> %xx) {
 ; CHECK-LABEL: @nonzero_reduce_xor_vscale_even(
-; CHECK-NEXT:    [[X:%.*]] = or <vscale x 2 x i8> [[XX:%.*]], shufflevector (<vscale x 2 x i8> insertelement (<vscale x 2 x i8> poison, i8 1, i64 0), <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer)
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
   %one = insertelement <vscale x 2 x i8> poison, i8 1, i64 0
   %ones = shufflevector <vscale x 2 x i8> %one, <vscale x 2 x i8> poison, <vscale x 2 x i32> zeroinitializer
@@ -1379,10 +1364,7 @@ define i8 @nonzero_reduce_xor_vscale_even_fail(<vscale x 2 x i8> %xx) {
 
 define i8 @nonzero_reduce_xor_vscale_odd(<vscale x 3 x i8> %xx) {
 ; CHECK-LABEL: @nonzero_reduce_xor_vscale_odd(
-; CHECK-NEXT:    [[X:%.*]] = and <vscale x 3 x i8> [[XX:%.*]], shufflevector (<vscale x 3 x i8> insertelement (<vscale x 3 x i8> poison, i8 1, i64 0), <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer)
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.xor.nxv3i8(<vscale x 3 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 0
 ;
   %one = insertelement <vscale x 3 x i8> poison, i8 1, i64 0
   %ones = shufflevector <vscale x 3 x i8> %one, <vscale x 3 x i8> poison, <vscale x 3 x i32> zeroinitializer

From 3921dffac29d8ea1c07bc6a7253876247a6eb1a3 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Thu, 11 Apr 2024 11:18:33 -0500
Subject: [PATCH 379/886] [ValueTracking] Add tests for `isKnownNonZero` of
 `llvm.vector.reduce.or`; NFC

---
 .../Transforms/InstSimplify/known-non-zero.ll | 26 +++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index d9b8f5eed3239..b0b825efaeee2 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -377,3 +377,29 @@ define <2 x i1> @insert_nonzero_any_idx_fail(<2 x i8> %xx, i8 %yy, i32 %idx) {
   %r = icmp eq <2 x i8> %ins, zeroinitializer
   ret <2 x i1> %r
 }
+
+define i1 @nonzero_reduce_or(<2 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_or(
+; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nuw <2 x i8> %xx, <i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}
+
+define i1 @nonzero_reduce_or_fail(<2 x i8> %xx) {
+; CHECK-LABEL: @nonzero_reduce_or_fail(
+; CHECK-NEXT:    [[X:%.*]] = add nsw <2 x i8> [[XX:%.*]], <i8 1, i8 1>
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
+; CHECK-NEXT:    ret i1 [[R]]
+;
+  %x = add nsw <2 x i8> %xx, <i8 1, i8 1>
+  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %r = icmp eq i8 %v, 0
+  ret i1 %r
+}

From 1e16a35fbcc6b21d1bee597c7e76f2444b7000e5 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Thu, 11 Apr 2024 11:19:15 -0500
Subject: [PATCH 380/886] [ValueTracking] Implement `isKnownNonZero` for
 `llvm.vector.reduce.or`

Closes #88320
---
 llvm/lib/Analysis/ValueTracking.cpp                 | 3 ++-
 llvm/test/Transforms/InstSimplify/known-non-zero.ll | 5 +----
 2 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 378e4aa9974e5..5c2806078261d 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -2918,7 +2918,8 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
         return isNonZeroAdd(DemandedElts, Depth, Q, BitWidth,
                             II->getArgOperand(0), II->getArgOperand(1),
                             /*NSW=*/true, /* NUW=*/false);
-        // umin/smin/smax/smin of all non-zero elements is always non-zero.
+        // umin/smin/smax/smin/or of all non-zero elements is always non-zero.
+      case Intrinsic::vector_reduce_or:
       case Intrinsic::vector_reduce_umax:
       case Intrinsic::vector_reduce_umin:
       case Intrinsic::vector_reduce_smax:
diff --git a/llvm/test/Transforms/InstSimplify/known-non-zero.ll b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
index b0b825efaeee2..fd2862eb04a24 100644
--- a/llvm/test/Transforms/InstSimplify/known-non-zero.ll
+++ b/llvm/test/Transforms/InstSimplify/known-non-zero.ll
@@ -380,10 +380,7 @@ define <2 x i1> @insert_nonzero_any_idx_fail(<2 x i8> %xx, i8 %yy, i32 %idx) {
 
 define i1 @nonzero_reduce_or(<2 x i8> %xx) {
 ; CHECK-LABEL: @nonzero_reduce_or(
-; CHECK-NEXT:    [[X:%.*]] = add nuw <2 x i8> [[XX:%.*]], <i8 1, i8 1>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i8 [[V]], 0
-; CHECK-NEXT:    ret i1 [[R]]
+; CHECK-NEXT:    ret i1 false
 ;
   %x = add nuw <2 x i8> %xx, <i8 1, i8 1>
   %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)

From f811d7b50957b801788d7b171ddeb25b1fda415a Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Mon, 15 Apr 2024 11:55:45 +0800
Subject: [PATCH 381/886] [C++20] [Modules] Introduce
 -fexperimental-modules-reduced-bmi (#85050)

This is the driver part of
https://github.com/llvm/llvm-project/pull/75894.

This patch introduces '-fexperimental-modules-reduced-bmi' to enable
generating the reduced BMI.

This patch did:
- When `-fexperimental-modules-reduced-bmi` is specified but
`--precompile` is not specified for a module unit, we'll skip the
precompile phase to avoid unnecessary two-phase compilation phases. Then
if `-c` is specified, we will generate the reduced BMI in CodeGenAction
as a by-product.
- When `-fexperimental-modules-reduced-bmi` is specified and
`--precompile` is specified, we will generate the reduced BMI in
GenerateModuleInterfaceAction as a by-product.
- When `-fexperimental-modules-reduced-bmi` is specified for a
non-module unit. We don't do anything nor try to give a warn. This is
more user friendly so that the end users can try to test and experiment
with the feature without asking help from the build systems.

The core design idea is that users should be able to enable this easily
with the existing cmake mechanisms.

The future plan for the flag is:
- Add this to clang19 and make it opt-in for 1~2 releases. It depends on
the testing feedback to decide how long we like to make it opt-in.
- Then we can announce the existing BMI generating may be deprecated and
suggesting people (end users or build systems) to enable this for 1~2
releases.
- Finally we will enable this by default. When that time comes, the term
`BMI` will refer to the reduced BMI today and the existing BMI will only
be meaningful to build systems which loves to support two phase
compilations.

I'll send release notes and document in seperate commits after this get
landed.
---
 clang/include/clang/CodeGen/CodeGenAction.h   |  2 +
 clang/include/clang/Driver/Options.td         |  6 +++
 .../include/clang/Frontend/FrontendOptions.h  | 10 +++-
 clang/lib/CodeGen/CodeGenAction.cpp           | 19 +++++++
 clang/lib/Driver/Driver.cpp                   | 12 ++++-
 clang/lib/Driver/ToolChains/Clang.cpp         | 18 +++++++
 clang/lib/Frontend/FrontendActions.cpp        |  7 +++
 .../test/Driver/module-fgen-reduced-bmi.cppm  | 51 +++++++++++++++++++
 clang/test/Modules/modules-reduced-bmi.cppm   | 36 +++++++++++++
 9 files changed, 159 insertions(+), 2 deletions(-)
 create mode 100644 clang/test/Driver/module-fgen-reduced-bmi.cppm
 create mode 100644 clang/test/Modules/modules-reduced-bmi.cppm

diff --git a/clang/include/clang/CodeGen/CodeGenAction.h b/clang/include/clang/CodeGen/CodeGenAction.h
index 7ad2988e589eb..186dbb43f01ef 100644
--- a/clang/include/clang/CodeGen/CodeGenAction.h
+++ b/clang/include/clang/CodeGen/CodeGenAction.h
@@ -57,6 +57,8 @@ class CodeGenAction : public ASTFrontendAction {
   bool loadLinkModules(CompilerInstance &CI);
 
 protected:
+  bool BeginSourceFileAction(CompilerInstance &CI) override;
+
   /// Create a new code generation action.  If the optional \p _VMContext
   /// parameter is supplied, the action uses it without taking ownership,
   /// otherwise it creates a fresh LLVM context and takes ownership.
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 9a0b5d3304ca6..e24626913add7 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -3036,6 +3036,7 @@ defm prebuilt_implicit_modules : BoolFOption<"prebuilt-implicit-modules",
 
 def fmodule_output_EQ : Joined<["-"], "fmodule-output=">,
   Flags<[NoXarchOption]>, Visibility<[ClangOption, CC1Option]>,
+  MarshallingInfoString<FrontendOpts<"ModuleOutputPath">>,
   HelpText<"Save intermediate module file results when compiling a standard C++ module unit.">;
 def fmodule_output : Flag<["-"], "fmodule-output">, Flags<[NoXarchOption]>,
   Visibility<[ClangOption, CC1Option]>,
@@ -3049,6 +3050,11 @@ defm skip_odr_check_in_gmf : BoolOption<"f", "skip-odr-check-in-gmf",
           "Perform ODR checks for decls in the global module fragment.">>,
   Group<f_Group>;
 
+def modules_reduced_bmi : Flag<["-"], "fexperimental-modules-reduced-bmi">,
+  Group<f_Group>, Visibility<[ClangOption, CC1Option]>,
+  HelpText<"Generate the reduced BMI">,
+  MarshallingInfoFlag<FrontendOpts<"GenReducedBMI">>;
+
 def fmodules_prune_interval : Joined<["-"], "fmodules-prune-interval=">, Group<i_Group>,
   Visibility<[ClangOption, CC1Option]>, MetaVarName<"<seconds>">,
   HelpText<"Specify the interval (in seconds) between attempts to prune the module cache">,
diff --git a/clang/include/clang/Frontend/FrontendOptions.h b/clang/include/clang/Frontend/FrontendOptions.h
index 5ee4d471670f4..a738c1f375768 100644
--- a/clang/include/clang/Frontend/FrontendOptions.h
+++ b/clang/include/clang/Frontend/FrontendOptions.h
@@ -404,6 +404,10 @@ class FrontendOptions {
   LLVM_PREFERRED_TYPE(bool)
   unsigned EmitPrettySymbolGraphs : 1;
 
+  /// Whether to generate reduced BMI for C++20 named modules.
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned GenReducedBMI : 1;
+
   CodeCompleteOptions CodeCompleteOpts;
 
   /// Specifies the output format of the AST.
@@ -568,6 +572,9 @@ class FrontendOptions {
   /// Path which stores the output files for -ftime-trace
   std::string TimeTracePath;
 
+  /// Output Path for module output file.
+  std::string ModuleOutputPath;
+
 public:
   FrontendOptions()
       : DisableFree(false), RelocatablePCH(false), ShowHelp(false),
@@ -582,7 +589,8 @@ class FrontendOptions {
         AllowPCMWithCompilerErrors(false), ModulesShareFileManager(true),
         EmitSymbolGraph(false), EmitExtensionSymbolGraphs(false),
         EmitSymbolGraphSymbolLabelsForTesting(false),
-        EmitPrettySymbolGraphs(false), TimeTraceGranularity(500) {}
+        EmitPrettySymbolGraphs(false), GenReducedBMI(false),
+        TimeTraceGranularity(500) {}
 
   /// getInputKindForExtension - Return the appropriate input kind for a file
   /// extension. For example, "c" would return Language::C.
diff --git a/clang/lib/CodeGen/CodeGenAction.cpp b/clang/lib/CodeGen/CodeGenAction.cpp
index bb9aaba025fa5..1a6b628016f74 100644
--- a/clang/lib/CodeGen/CodeGenAction.cpp
+++ b/clang/lib/CodeGen/CodeGenAction.cpp
@@ -25,8 +25,11 @@
 #include "clang/CodeGen/ModuleBuilder.h"
 #include "clang/Driver/DriverDiagnostic.h"
 #include "clang/Frontend/CompilerInstance.h"
+#include "clang/Frontend/FrontendActions.h"
 #include "clang/Frontend/FrontendDiagnostic.h"
+#include "clang/Frontend/MultiplexConsumer.h"
 #include "clang/Lex/Preprocessor.h"
+#include "clang/Serialization/ASTWriter.h"
 #include "llvm/ADT/Hashing.h"
 #include "llvm/Bitcode/BitcodeReader.h"
 #include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
@@ -1003,6 +1006,12 @@ CodeGenerator *CodeGenAction::getCodeGenerator() const {
   return BEConsumer->getCodeGenerator();
 }
 
+bool CodeGenAction::BeginSourceFileAction(CompilerInstance &CI) {
+  if (CI.getFrontendOpts().GenReducedBMI)
+    CI.getLangOpts().setCompilingModule(LangOptions::CMK_ModuleInterface);
+  return true;
+}
+
 static std::unique_ptr<raw_pwrite_stream>
 GetOutputStream(CompilerInstance &CI, StringRef InFile, BackendAction Action) {
   switch (Action) {
@@ -1061,6 +1070,16 @@ CodeGenAction::CreateASTConsumer(CompilerInstance &CI, StringRef InFile) {
     CI.getPreprocessor().addPPCallbacks(std::move(Callbacks));
   }
 
+  if (CI.getFrontendOpts().GenReducedBMI &&
+      !CI.getFrontendOpts().ModuleOutputPath.empty()) {
+    std::vector<std::unique_ptr<ASTConsumer>> Consumers(2);
+    Consumers[0] = std::make_unique<ReducedBMIGenerator>(
+        CI.getPreprocessor(), CI.getModuleCache(),
+        CI.getFrontendOpts().ModuleOutputPath);
+    Consumers[1] = std::move(Result);
+    return std::make_unique<MultiplexConsumer>(std::move(Consumers));
+  }
+
   return std::move(Result);
 }
 
diff --git a/clang/lib/Driver/Driver.cpp b/clang/lib/Driver/Driver.cpp
index e7335a61b10c5..0da92001e08c2 100644
--- a/clang/lib/Driver/Driver.cpp
+++ b/clang/lib/Driver/Driver.cpp
@@ -4756,6 +4756,14 @@ Action *Driver::ConstructPhaseAction(
     if (Args.hasArg(options::OPT_extract_api))
       return C.MakeAction<ExtractAPIJobAction>(Input, types::TY_API_INFO);
 
+    // With 'fexperimental-modules-reduced-bmi', we don't want to run the
+    // precompile phase unless the user specified '--precompile'. In the case
+    // the '--precompile' flag is enabled, we will try to emit the reduced BMI
+    // as a by product in GenerateModuleInterfaceAction.
+    if (Args.hasArg(options::OPT_modules_reduced_bmi) &&
+        !Args.getLastArg(options::OPT__precompile))
+      return Input;
+
     types::ID OutputTy = getPrecompiledType(Input->getType());
     assert(OutputTy != types::TY_INVALID &&
            "Cannot precompile this input type!");
@@ -5916,8 +5924,10 @@ const char *Driver::GetNamedOutputPath(Compilation &C, const JobAction &JA,
   // If we're emitting a module output with the specified option
   // `-fmodule-output`.
   if (!AtTopLevel && isa<PrecompileJobAction>(JA) &&
-      JA.getType() == types::TY_ModuleFile && SpecifiedModuleOutput)
+      JA.getType() == types::TY_ModuleFile && SpecifiedModuleOutput) {
+    assert(!C.getArgs().hasArg(options::OPT_modules_reduced_bmi));
     return GetModuleOutputPath(C, JA, BaseInput);
+  }
 
   // Output to a temporary file?
   if ((!AtTopLevel && !isSaveTempsEnabled() &&
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 766a9b91e3c0a..6d52eced10429 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -4045,6 +4045,24 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
   // module fragment.
   CmdArgs.push_back("-fskip-odr-check-in-gmf");
 
+  if (Args.hasArg(options::OPT_modules_reduced_bmi) &&
+      (Input.getType() == driver::types::TY_CXXModule ||
+       Input.getType() == driver::types::TY_PP_CXXModule)) {
+    CmdArgs.push_back("-fexperimental-modules-reduced-bmi");
+
+    if (Args.hasArg(options::OPT_fmodule_output_EQ))
+      Args.AddLastArg(CmdArgs, options::OPT_fmodule_output_EQ);
+    else
+      CmdArgs.push_back(Args.MakeArgString(
+          "-fmodule-output=" +
+          getCXX20NamedModuleOutputPath(Args, Input.getBaseInput())));
+  }
+
+  // Noop if we see '-fexperimental-modules-reduced-bmi' with other translation
+  // units than module units. This is more user friendly to allow end uers to
+  // enable this feature without asking for help from build systems.
+  Args.ClaimAllArgs(options::OPT_modules_reduced_bmi);
+
   // We need to include the case the input file is a module file here.
   // Since the default compilation model for C++ module interface unit will
   // create temporary module file and compile the temporary module file
diff --git a/clang/lib/Frontend/FrontendActions.cpp b/clang/lib/Frontend/FrontendActions.cpp
index 642b14d8b09d9..04eb104132671 100644
--- a/clang/lib/Frontend/FrontendActions.cpp
+++ b/clang/lib/Frontend/FrontendActions.cpp
@@ -281,6 +281,13 @@ GenerateModuleInterfaceAction::CreateASTConsumer(CompilerInstance &CI,
   if (Consumers.empty())
     return nullptr;
 
+  if (CI.getFrontendOpts().GenReducedBMI &&
+      !CI.getFrontendOpts().ModuleOutputPath.empty()) {
+    Consumers.push_back(std::make_unique<ReducedBMIGenerator>(
+        CI.getPreprocessor(), CI.getModuleCache(),
+        CI.getFrontendOpts().ModuleOutputPath));
+  }
+
   return std::make_unique<MultiplexConsumer>(std::move(Consumers));
 }
 
diff --git a/clang/test/Driver/module-fgen-reduced-bmi.cppm b/clang/test/Driver/module-fgen-reduced-bmi.cppm
new file mode 100644
index 0000000000000..1223189fb49b7
--- /dev/null
+++ b/clang/test/Driver/module-fgen-reduced-bmi.cppm
@@ -0,0 +1,51 @@
+// It is annoying to handle different slash direction
+// in Windows and Linux. So we disable the test on Windows
+// here.
+// REQUIRES: !system-windows
+// On AIX, the default output for `-c` may be `.s` instead of `.o`,
+// which makes the test fail. So disable the test on AIX.
+// UNSUPPORTED: system-aix
+//
+// RUN: rm -rf %t && split-file %s %t && cd %t
+//
+// RUN: %clang -std=c++20 Hello.cppm -fmodule-output=Hello.pcm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o Hello.o -### 2>&1 | FileCheck Hello.cppm
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o Hello.o -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-UNSPECIFIED
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-NO-O
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o AnotherName.o -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-ANOTHER-NAME
+//
+// RUN: %clang -std=c++20 Hello.cppm --precompile -fexperimental-modules-reduced-bmi \
+// RUN:     -o Hello.full.pcm -### 2>&1 | FileCheck Hello.cppm \
+// RUN:     --check-prefix=CHECK-EMIT-MODULE-INTERFACE
+//
+// RUN: %clang -std=c++20 Hello.cc -fexperimental-modules-reduced-bmi -Wall -Werror \
+// RUN:     -c -o Hello.o -### 2>&1 | FileCheck Hello.cc
+
+//--- Hello.cppm
+export module Hello;
+
+// Test that we won't generate the emit-module-interface as 2 phase compilation model.
+// CHECK-NOT: -emit-module-interface
+// CHECK: "-fexperimental-modules-reduced-bmi"
+
+// CHECK-UNSPECIFIED: -fmodule-output=Hello.pcm
+
+// CHECK-NO-O: -fmodule-output=Hello.pcm
+// CHECK-ANOTHER-NAME: -fmodule-output=AnotherName.pcm
+
+// With `-emit-module-interface` specified, we should still see the `-emit-module-interface`
+// flag.
+// CHECK-EMIT-MODULE-INTERFACE: -emit-module-interface
+
+//--- Hello.cc
+
+// CHECK-NOT: "-fexperimental-modules-reduced-bmi"
diff --git a/clang/test/Modules/modules-reduced-bmi.cppm b/clang/test/Modules/modules-reduced-bmi.cppm
new file mode 100644
index 0000000000000..9b84220ae0303
--- /dev/null
+++ b/clang/test/Modules/modules-reduced-bmi.cppm
@@ -0,0 +1,36 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-reduced-module-interface -o %t/a.reduced.pcm
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -fexperimental-modules-reduced-bmi -fmodule-output=%t/a.pcm \
+// RUN:     -S -emit-llvm -o %t/a.ll
+//
+// Test that the generated BMI from `-fexperimental-modules-reduced-bmi -fmodule-output=` is same with
+// `-emit-reduced-module-interface`.
+// RUN: diff %t/a.reduced.pcm %t/a.pcm
+//
+// Test that we can consume the produced BMI correctly.
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
+//
+// RUN: rm -f %t/a.pcm
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -fexperimental-modules-reduced-bmi -fmodule-output=%t/a.pcm \
+// RUN:     -emit-module-interface -o %t/a.full.pcm
+// RUN: diff %t/a.reduced.pcm %t/a.pcm
+// RUN: not diff %t/a.pcm %t/a.full.pcm
+//
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.full.pcm -fsyntax-only -verify
+
+//--- a.cppm
+export module a;
+export int a() {
+    return 43;
+}
+
+//--- b.cppm
+// Test that we can consume the produced BMI correctly as a smocking test.
+// expected-no-diagnostics
+export module b;
+import a;
+export int b() { return a(); }

From c7902d87a5a050b816edfe99e7e093ae63f4e564 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Mon, 15 Apr 2024 13:33:29 +0900
Subject: [PATCH 382/886] [InstCombine] Add test for isKnownNonZero() undef
 miscompile (NFC)

---
 llvm/test/Transforms/InstCombine/add.ll | 26 +++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/add.ll b/llvm/test/Transforms/InstCombine/add.ll
index 23eee8547597e..515b80d9d6ea0 100644
--- a/llvm/test/Transforms/InstCombine/add.ll
+++ b/llvm/test/Transforms/InstCombine/add.ll
@@ -3093,6 +3093,32 @@ define <2 x i32> @dec_zext_add_nonzero_vec(<2 x i8> %x) {
   ret <2 x i32> %c
 }
 
+define <2 x i32> @dec_zext_add_nonzero_vec_undef0(<2 x i8> %x) {
+; CHECK-LABEL: @dec_zext_add_nonzero_vec_undef0(
+; CHECK-NEXT:    [[O:%.*]] = or <2 x i8> [[X:%.*]], <i8 8, i8 undef>
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i8> [[O]] to <2 x i32>
+; CHECK-NEXT:    ret <2 x i32> [[C]]
+;
+  %o = or <2 x i8> %x, <i8 8, i8 undef>
+  %a = add <2 x i8> %o, <i8 -1, i8 -1>
+  %b = zext <2 x i8> %a to <2 x i32>
+  %c = add <2 x i32> %b, <i32 1, i32 1>
+  ret <2 x i32> %c
+}
+
+define <2 x i32> @dec_zext_add_nonzero_poison0(<2 x i8> %x) {
+; CHECK-LABEL: @dec_zext_add_nonzero_poison0(
+; CHECK-NEXT:    [[O:%.*]] = or <2 x i8> [[X:%.*]], <i8 8, i8 poison>
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i8> [[O]] to <2 x i32>
+; CHECK-NEXT:    ret <2 x i32> [[C]]
+;
+  %o = or <2 x i8> %x, <i8 8, i8 poison>
+  %a = add <2 x i8> %o, <i8 -1, i8 -1>
+  %b = zext <2 x i8> %a to <2 x i32>
+  %c = add <2 x i32> %b, <i32 1, i32 1>
+  ret <2 x i32> %c
+}
+
 define <2 x i32> @dec_zext_add_nonzero_vec_poison1(<2 x i8> %x) {
 ; CHECK-LABEL: @dec_zext_add_nonzero_vec_poison1(
 ; CHECK-NEXT:    [[O:%.*]] = or <2 x i8> [[X:%.*]], <i8 8, i8 8>

From 52a1998f15ab0e5b9ff7afa8b92cc714463d5dd8 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Mon, 15 Apr 2024 13:36:57 +0900
Subject: [PATCH 383/886] [ValueTracking] Don't accept undef in
 isKnownNonZero()

As the undef can be replaced with a zero value, this is not legal
in the general case. We can only allow poison values. This matches
what the other ValueTracking helpers like computeKnownBits() do.
---
 llvm/lib/Analysis/ValueTracking.cpp          |  4 ++--
 llvm/test/Transforms/InstCombine/add.ll      |  6 +++++-
 llvm/test/Transforms/InstSimplify/vec-cmp.ll | 17 ++++++++++++++---
 3 files changed, 21 insertions(+), 6 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 5c2806078261d..5beea614e332e 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -3034,7 +3034,7 @@ bool isKnownNonZero(const Value *V, const APInt &DemandedElts, unsigned Depth,
       // Must be non-zero due to null test above.
       return true;
 
-    // For constant vectors, check that all elements are undefined or known
+    // For constant vectors, check that all elements are poison or known
     // non-zero to determine that the whole vector is known non-zero.
     if (auto *VecTy = dyn_cast<FixedVectorType>(Ty)) {
       for (unsigned i = 0, e = VecTy->getNumElements(); i != e; ++i) {
@@ -3043,7 +3043,7 @@ bool isKnownNonZero(const Value *V, const APInt &DemandedElts, unsigned Depth,
         Constant *Elt = C->getAggregateElement(i);
         if (!Elt || Elt->isNullValue())
           return false;
-        if (!isa<UndefValue>(Elt) && !isa<ConstantInt>(Elt))
+        if (!isa<PoisonValue>(Elt) && !isa<ConstantInt>(Elt))
           return false;
       }
       return true;
diff --git a/llvm/test/Transforms/InstCombine/add.ll b/llvm/test/Transforms/InstCombine/add.ll
index 515b80d9d6ea0..408b0c6559b00 100644
--- a/llvm/test/Transforms/InstCombine/add.ll
+++ b/llvm/test/Transforms/InstCombine/add.ll
@@ -3093,10 +3093,14 @@ define <2 x i32> @dec_zext_add_nonzero_vec(<2 x i8> %x) {
   ret <2 x i32> %c
 }
 
+; Negative test: Folding this with undef is not safe.
+
 define <2 x i32> @dec_zext_add_nonzero_vec_undef0(<2 x i8> %x) {
 ; CHECK-LABEL: @dec_zext_add_nonzero_vec_undef0(
 ; CHECK-NEXT:    [[O:%.*]] = or <2 x i8> [[X:%.*]], <i8 8, i8 undef>
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i8> [[O]] to <2 x i32>
+; CHECK-NEXT:    [[A:%.*]] = add <2 x i8> [[O]], <i8 -1, i8 -1>
+; CHECK-NEXT:    [[B:%.*]] = zext <2 x i8> [[A]] to <2 x i32>
+; CHECK-NEXT:    [[C:%.*]] = add nuw nsw <2 x i32> [[B]], <i32 1, i32 1>
 ; CHECK-NEXT:    ret <2 x i32> [[C]]
 ;
   %o = or <2 x i8> %x, <i8 8, i8 undef>
diff --git a/llvm/test/Transforms/InstSimplify/vec-cmp.ll b/llvm/test/Transforms/InstSimplify/vec-cmp.ll
index 6391e9aaa0469..dce1261d7031c 100644
--- a/llvm/test/Transforms/InstSimplify/vec-cmp.ll
+++ b/llvm/test/Transforms/InstSimplify/vec-cmp.ll
@@ -21,13 +21,24 @@ define <2 x i1> @nonzero_vec_nonsplat(<2 x i32> %x) {
 
 define <2 x i1> @nonzero_vec_undef_elt(<2 x i32> %x) {
 ; CHECK-LABEL: @nonzero_vec_undef_elt(
-; CHECK-NEXT:    ret <2 x i1> zeroinitializer
+; CHECK-NEXT:    [[Y:%.*]] = or <2 x i32> [[X:%.*]], <i32 undef, i32 1>
+; CHECK-NEXT:    [[C:%.*]] = icmp eq <2 x i32> [[Y]], zeroinitializer
+; CHECK-NEXT:    ret <2 x i1> [[C]]
 ;
   %y = or <2 x i32> %x, <i32 undef, i32 1>
   %c = icmp eq <2 x i32> %y, zeroinitializer
   ret <2 x i1> %c
 }
 
+define <2 x i1> @nonzero_vec_poison_elt(<2 x i32> %x) {
+; CHECK-LABEL: @nonzero_vec_poison_elt(
+; CHECK-NEXT:    ret <2 x i1> zeroinitializer
+;
+  %y = or <2 x i32> %x, <i32 poison, i32 1>
+  %c = icmp eq <2 x i32> %y, zeroinitializer
+  ret <2 x i1> %c
+}
+
 define <2 x i1> @may_be_zero_vec(<2 x i32> %x) {
 ; CHECK-LABEL: @may_be_zero_vec(
 ; CHECK-NEXT:    [[Y:%.*]] = or <2 x i32> [[X:%.*]], <i32 0, i32 1>
@@ -45,7 +56,7 @@ define <2 x i1> @nonzero_vec_mul_nuw(<2 x i32> %x, <2 x i32> %y) {
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %xnz = or <2 x i32> %x, <i32 1, i32 2>
-  %ynz = or <2 x i32> %y, <i32 3, i32 undef>
+  %ynz = or <2 x i32> %y, <i32 3, i32 poison>
   %m = mul nuw <2 x i32> %xnz, %ynz
   %c = icmp eq <2 x i32> %m, zeroinitializer
   ret <2 x i1> %c
@@ -56,7 +67,7 @@ define <2 x i1> @nonzero_vec_mul_nsw(<2 x i32> %x, <2 x i32> %y) {
 ; CHECK-LABEL: @nonzero_vec_mul_nsw(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %xnz = or <2 x i32> %x, <i32 undef, i32 2>
+  %xnz = or <2 x i32> %x, <i32 poison, i32 2>
   %ynz = or <2 x i32> %y, <i32 3, i32 4>
   %m = mul nsw <2 x i32> %xnz, %ynz
   %c = icmp ne <2 x i32> %m, zeroinitializer

From 7177dc2ef7f3a50a1d8b892d7bd298f3d52a1aab Mon Sep 17 00:00:00 2001
From: fengfeng <153487255+fengfeng09@users.noreply.github.com>
Date: Mon, 15 Apr 2024 12:59:48 +0800
Subject: [PATCH 384/886] [SDAG] Apply or-disjoint in
 SelectionDAG::isBaseWithConstantOffset (#88493)

Signed-off-by: feng.feng <feng.feng@iluvatar.com>
---
 llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 11 ++---------
 .../AVR/base-with-add-like-constant-offset.ll  | 18 ++++++++++++++++++
 2 files changed, 20 insertions(+), 9 deletions(-)
 create mode 100644 llvm/test/CodeGen/AVR/base-with-add-like-constant-offset.ll

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 412e1de9fc41c..fc972664e5f01 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -5203,15 +5203,8 @@ bool SelectionDAG::isADDLike(SDValue Op) const {
 }
 
 bool SelectionDAG::isBaseWithConstantOffset(SDValue Op) const {
-  if ((Op.getOpcode() != ISD::ADD && Op.getOpcode() != ISD::OR) ||
-      !isa<ConstantSDNode>(Op.getOperand(1)))
-    return false;
-
-  if (Op.getOpcode() == ISD::OR &&
-      !MaskedValueIsZero(Op.getOperand(0), Op.getConstantOperandAPInt(1)))
-    return false;
-
-  return true;
+  return Op.getNumOperands() == 2 && isa<ConstantSDNode>(Op.getOperand(1)) &&
+         (Op.getOpcode() == ISD::ADD || isADDLike(Op));
 }
 
 bool SelectionDAG::isKnownNeverNaN(SDValue Op, bool SNaN, unsigned Depth) const {
diff --git a/llvm/test/CodeGen/AVR/base-with-add-like-constant-offset.ll b/llvm/test/CodeGen/AVR/base-with-add-like-constant-offset.ll
new file mode 100644
index 0000000000000..278dc4893a68d
--- /dev/null
+++ b/llvm/test/CodeGen/AVR/base-with-add-like-constant-offset.ll
@@ -0,0 +1,18 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=avr %s -start-before=avr-isel -o - | FileCheck %s
+
+define void @test(i16 %x, ptr addrspace(1) %o) {
+; CHECK-LABEL: test:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    mov r30, r22
+; CHECK-NEXT:    mov r31, r23
+; CHECK-NEXT:    std Z+11, r25
+; CHECK-NEXT:    std Z+10, r24
+; CHECK-NEXT:    ret
+  %int = ptrtoint ptr addrspace(1) %o to i16
+  %or = or disjoint i16 %int, 10
+  %addr = inttoptr i16 %or to ptr addrspace(1)
+  store i16 %x, ptr addrspace(1) %addr
+  ret void
+}
+

From 6b80e2fef5b0e99f81d1f4a7322b110f9ead3521 Mon Sep 17 00:00:00 2001
From: laichunfeng <laichunfeng@tencent.com>
Date: Mon, 15 Apr 2024 13:11:45 +0800
Subject: [PATCH 385/886] [mi-sched] Suppress register pressure with i64.
 (#88256)

Machine scheduler will suppress register pressure when the scheduling
window is too small, but now it doesn't consider i64 register type,
and this MR extends it into i64 register type, so architecture like
RISCV64 that only supports i64 interger register will have the same
behavior like RISCV32.
---
 llvm/lib/CodeGen/MachineScheduler.cpp         |   6 +-
 .../LoongArch/atomicrmw-uinc-udec-wrap.ll     |  16 +-
 llvm/test/CodeGen/LoongArch/vector-fp-imm.ll  |  18 +-
 .../CodeGen/RISCV/atomicrmw-uinc-udec-wrap.ll |   8 +-
 llvm/test/CodeGen/RISCV/bfloat-convert.ll     |   6 +-
 .../calling-conv-lp64-lp64f-lp64d-common.ll   |   6 +-
 llvm/test/CodeGen/RISCV/double-convert.ll     |  28 ++--
 llvm/test/CodeGen/RISCV/float-convert.ll      |  16 +-
 llvm/test/CodeGen/RISCV/half-convert.ll       | 132 +++++++--------
 .../RISCV/rvv/fixed-vectors-fp-shuffles.ll    |  46 ++---
 .../rvv/fixed-vectors-insert-subvector.ll     |   4 +-
 .../RISCV/rvv/fixed-vectors-setcc-fp-vp.ll    | 157 ++++++------------
 .../RISCV/rvv/fixed-vectors-setcc-int-vp.ll   | 154 ++++++-----------
 .../CodeGen/RISCV/rvv/vpscatter-sdnode.ll     |  24 +--
 14 files changed, 255 insertions(+), 366 deletions(-)

diff --git a/llvm/lib/CodeGen/MachineScheduler.cpp b/llvm/lib/CodeGen/MachineScheduler.cpp
index 203bdffc49ae1..78d581c8ceadf 100644
--- a/llvm/lib/CodeGen/MachineScheduler.cpp
+++ b/llvm/lib/CodeGen/MachineScheduler.cpp
@@ -3281,14 +3281,16 @@ void GenericScheduler::initPolicy(MachineBasicBlock::iterator Begin,
 
   // Avoid setting up the register pressure tracker for small regions to save
   // compile time. As a rough heuristic, only track pressure when the number of
-  // schedulable instructions exceeds half the integer register file.
+  // schedulable instructions exceeds half the allocatable integer register file
+  // that is the largest legal integer regiser type.
   RegionPolicy.ShouldTrackPressure = true;
-  for (unsigned VT = MVT::i32; VT > (unsigned)MVT::i1; --VT) {
+  for (unsigned VT = MVT::i64; VT > (unsigned)MVT::i1; --VT) {
     MVT::SimpleValueType LegalIntVT = (MVT::SimpleValueType)VT;
     if (TLI->isTypeLegal(LegalIntVT)) {
       unsigned NIntRegs = Context->RegClassInfo->getNumAllocatableRegs(
         TLI->getRegClassFor(LegalIntVT));
       RegionPolicy.ShouldTrackPressure = NumRegionInstrs > (NIntRegs / 2);
+      break;
     }
   }
 
diff --git a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
index bf48c0df3e496..7cde034726e0b 100644
--- a/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
+++ b/llvm/test/CodeGen/LoongArch/atomicrmw-uinc-udec-wrap.ll
@@ -4,12 +4,12 @@
 define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; LA64-LABEL: atomicrmw_uinc_wrap_i8:
 ; LA64:       # %bb.0:
-; LA64-NEXT:    slli.d $a2, $a0, 3
+; LA64-NEXT:    slli.d $a4, $a0, 3
 ; LA64-NEXT:    bstrins.d $a0, $zero, 1, 0
-; LA64-NEXT:    ori $a3, $zero, 255
-; LA64-NEXT:    sll.w $a4, $a3, $a2
+; LA64-NEXT:    andi $a2, $a4, 24
+; LA64-NEXT:    ori $a5, $zero, 255
 ; LA64-NEXT:    ld.w $a3, $a0, 0
-; LA64-NEXT:    andi $a2, $a2, 24
+; LA64-NEXT:    sll.w $a4, $a5, $a4
 ; LA64-NEXT:    nor $a4, $a4, $zero
 ; LA64-NEXT:    andi $a1, $a1, 255
 ; LA64-NEXT:    .p2align 4, , 16
@@ -54,13 +54,13 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 define i16 @atomicrmw_uinc_wrap_i16(ptr %ptr, i16 %val) {
 ; LA64-LABEL: atomicrmw_uinc_wrap_i16:
 ; LA64:       # %bb.0:
-; LA64-NEXT:    slli.d $a2, $a0, 3
+; LA64-NEXT:    slli.d $a4, $a0, 3
 ; LA64-NEXT:    bstrins.d $a0, $zero, 1, 0
+; LA64-NEXT:    andi $a2, $a4, 24
 ; LA64-NEXT:    lu12i.w $a3, 15
-; LA64-NEXT:    ori $a3, $a3, 4095
-; LA64-NEXT:    sll.w $a4, $a3, $a2
+; LA64-NEXT:    ori $a5, $a3, 4095
 ; LA64-NEXT:    ld.w $a3, $a0, 0
-; LA64-NEXT:    andi $a2, $a2, 24
+; LA64-NEXT:    sll.w $a4, $a5, $a4
 ; LA64-NEXT:    nor $a4, $a4, $zero
 ; LA64-NEXT:    bstrpick.d $a1, $a1, 15, 0
 ; LA64-NEXT:    .p2align 4, , 16
diff --git a/llvm/test/CodeGen/LoongArch/vector-fp-imm.ll b/llvm/test/CodeGen/LoongArch/vector-fp-imm.ll
index d03af114bceef..18d17751a7719 100644
--- a/llvm/test/CodeGen/LoongArch/vector-fp-imm.ll
+++ b/llvm/test/CodeGen/LoongArch/vector-fp-imm.ll
@@ -124,10 +124,10 @@ define void @test_f2(ptr %P, ptr %S) nounwind {
 ; LA64F:       # %bb.0:
 ; LA64F-NEXT:    fld.s $fa0, $a0, 4
 ; LA64F-NEXT:    fld.s $fa1, $a0, 0
-; LA64F-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_0)
-; LA64F-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI1_0)
-; LA64F-NEXT:    fld.s $fa2, $a0, 0
 ; LA64F-NEXT:    addi.w $a0, $zero, 1
+; LA64F-NEXT:    pcalau12i $a2, %pc_hi20(.LCPI1_0)
+; LA64F-NEXT:    addi.d $a2, $a2, %pc_lo12(.LCPI1_0)
+; LA64F-NEXT:    fld.s $fa2, $a2, 0
 ; LA64F-NEXT:    movgr2fr.w $fa3, $a0
 ; LA64F-NEXT:    ffint.s.w $fa3, $fa3
 ; LA64F-NEXT:    fadd.s $fa1, $fa1, $fa3
@@ -140,10 +140,10 @@ define void @test_f2(ptr %P, ptr %S) nounwind {
 ; LA64D:       # %bb.0:
 ; LA64D-NEXT:    fld.s $fa0, $a0, 4
 ; LA64D-NEXT:    fld.s $fa1, $a0, 0
-; LA64D-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI1_0)
-; LA64D-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI1_0)
-; LA64D-NEXT:    fld.s $fa2, $a0, 0
 ; LA64D-NEXT:    addi.w $a0, $zero, 1
+; LA64D-NEXT:    pcalau12i $a2, %pc_hi20(.LCPI1_0)
+; LA64D-NEXT:    addi.d $a2, $a2, %pc_lo12(.LCPI1_0)
+; LA64D-NEXT:    fld.s $fa2, $a2, 0
 ; LA64D-NEXT:    movgr2fr.w $fa3, $a0
 ; LA64D-NEXT:    ffint.s.w $fa3, $fa3
 ; LA64D-NEXT:    fadd.s $fa1, $fa1, $fa3
@@ -527,10 +527,10 @@ define void @test_d2(ptr %P, ptr %S) nounwind {
 ; LA64D:       # %bb.0:
 ; LA64D-NEXT:    fld.d $fa0, $a0, 8
 ; LA64D-NEXT:    fld.d $fa1, $a0, 0
-; LA64D-NEXT:    pcalau12i $a0, %pc_hi20(.LCPI4_0)
-; LA64D-NEXT:    addi.d $a0, $a0, %pc_lo12(.LCPI4_0)
-; LA64D-NEXT:    fld.d $fa2, $a0, 0
 ; LA64D-NEXT:    addi.d $a0, $zero, 1
+; LA64D-NEXT:    pcalau12i $a2, %pc_hi20(.LCPI4_0)
+; LA64D-NEXT:    addi.d $a2, $a2, %pc_lo12(.LCPI4_0)
+; LA64D-NEXT:    fld.d $fa2, $a2, 0
 ; LA64D-NEXT:    movgr2fr.d $fa3, $a0
 ; LA64D-NEXT:    ffint.d.l $fa3, $fa3
 ; LA64D-NEXT:    fadd.d $fa1, $fa1, $fa3
diff --git a/llvm/test/CodeGen/RISCV/atomicrmw-uinc-udec-wrap.ll b/llvm/test/CodeGen/RISCV/atomicrmw-uinc-udec-wrap.ll
index 5914e45a15330..f96e1bad2e389 100644
--- a/llvm/test/CodeGen/RISCV/atomicrmw-uinc-udec-wrap.ll
+++ b/llvm/test/CodeGen/RISCV/atomicrmw-uinc-udec-wrap.ll
@@ -127,11 +127,11 @@ define i8 @atomicrmw_uinc_wrap_i8(ptr %ptr, i8 %val) {
 ; RV64IA-LABEL: atomicrmw_uinc_wrap_i8:
 ; RV64IA:       # %bb.0:
 ; RV64IA-NEXT:    andi a2, a0, -4
-; RV64IA-NEXT:    slli a0, a0, 3
-; RV64IA-NEXT:    li a3, 255
-; RV64IA-NEXT:    sllw a4, a3, a0
+; RV64IA-NEXT:    slli a4, a0, 3
+; RV64IA-NEXT:    andi a0, a4, 24
+; RV64IA-NEXT:    li a5, 255
 ; RV64IA-NEXT:    lw a3, 0(a2)
-; RV64IA-NEXT:    andi a0, a0, 24
+; RV64IA-NEXT:    sllw a4, a5, a4
 ; RV64IA-NEXT:    not a4, a4
 ; RV64IA-NEXT:    andi a1, a1, 255
 ; RV64IA-NEXT:  .LBB0_1: # %atomicrmw.start
diff --git a/llvm/test/CodeGen/RISCV/bfloat-convert.ll b/llvm/test/CodeGen/RISCV/bfloat-convert.ll
index d5041c2a7ca78..9e2b0b5c3cbb4 100644
--- a/llvm/test/CodeGen/RISCV/bfloat-convert.ll
+++ b/llvm/test/CodeGen/RISCV/bfloat-convert.ll
@@ -84,12 +84,12 @@ define i16 @fcvt_si_bf16_sat(bfloat %a) nounwind {
 ; CHECK64ZFBFMIN:       # %bb.0: # %start
 ; CHECK64ZFBFMIN-NEXT:    fcvt.s.bf16 fa5, fa0
 ; CHECK64ZFBFMIN-NEXT:    feq.s a0, fa5, fa5
+; CHECK64ZFBFMIN-NEXT:    neg a0, a0
 ; CHECK64ZFBFMIN-NEXT:    lui a1, %hi(.LCPI1_0)
 ; CHECK64ZFBFMIN-NEXT:    flw fa4, %lo(.LCPI1_0)(a1)
 ; CHECK64ZFBFMIN-NEXT:    lui a1, 815104
 ; CHECK64ZFBFMIN-NEXT:    fmv.w.x fa3, a1
 ; CHECK64ZFBFMIN-NEXT:    fmax.s fa5, fa5, fa3
-; CHECK64ZFBFMIN-NEXT:    neg a0, a0
 ; CHECK64ZFBFMIN-NEXT:    fmin.s fa5, fa5, fa4
 ; CHECK64ZFBFMIN-NEXT:    fcvt.l.s a1, fa5, rtz
 ; CHECK64ZFBFMIN-NEXT:    and a0, a0, a1
@@ -187,10 +187,10 @@ define i16 @fcvt_ui_bf16_sat(bfloat %a) nounwind {
 ;
 ; RV64ID-LABEL: fcvt_ui_bf16_sat:
 ; RV64ID:       # %bb.0: # %start
-; RV64ID-NEXT:    lui a0, %hi(.LCPI3_0)
-; RV64ID-NEXT:    flw fa5, %lo(.LCPI3_0)(a0)
 ; RV64ID-NEXT:    fmv.x.w a0, fa0
 ; RV64ID-NEXT:    slli a0, a0, 16
+; RV64ID-NEXT:    lui a1, %hi(.LCPI3_0)
+; RV64ID-NEXT:    flw fa5, %lo(.LCPI3_0)(a1)
 ; RV64ID-NEXT:    fmv.w.x fa4, a0
 ; RV64ID-NEXT:    fmv.w.x fa3, zero
 ; RV64ID-NEXT:    fmax.s fa4, fa4, fa3
diff --git a/llvm/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll b/llvm/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll
index d847112943302..67123466354c4 100644
--- a/llvm/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll
+++ b/llvm/test/CodeGen/RISCV/calling-conv-lp64-lp64f-lp64d-common.ll
@@ -140,11 +140,11 @@ define i64 @caller_large_scalars() nounwind {
 ; RV64I-NEXT:    sd a0, 0(sp)
 ; RV64I-NEXT:    sd zero, 56(sp)
 ; RV64I-NEXT:    sd zero, 48(sp)
-; RV64I-NEXT:    li a0, 1
-; RV64I-NEXT:    sd a0, 32(sp)
+; RV64I-NEXT:    sd zero, 40(sp)
+; RV64I-NEXT:    li a2, 1
 ; RV64I-NEXT:    addi a0, sp, 32
 ; RV64I-NEXT:    mv a1, sp
-; RV64I-NEXT:    sd zero, 40(sp)
+; RV64I-NEXT:    sd a2, 32(sp)
 ; RV64I-NEXT:    call callee_large_scalars
 ; RV64I-NEXT:    ld ra, 72(sp) # 8-byte Folded Reload
 ; RV64I-NEXT:    addi sp, sp, 80
diff --git a/llvm/test/CodeGen/RISCV/double-convert.ll b/llvm/test/CodeGen/RISCV/double-convert.ll
index da882cafd9971..c147d6ec6d9b1 100644
--- a/llvm/test/CodeGen/RISCV/double-convert.ll
+++ b/llvm/test/CodeGen/RISCV/double-convert.ll
@@ -1651,8 +1651,8 @@ define signext i16 @fcvt_w_s_sat_i16(double %a) nounwind {
 ; RV64IFD-NEXT:    lui a0, %hi(.LCPI26_1)
 ; RV64IFD-NEXT:    fld fa4, %lo(.LCPI26_1)(a0)
 ; RV64IFD-NEXT:    feq.d a0, fa0, fa0
-; RV64IFD-NEXT:    fmax.d fa5, fa0, fa5
 ; RV64IFD-NEXT:    neg a0, a0
+; RV64IFD-NEXT:    fmax.d fa5, fa0, fa5
 ; RV64IFD-NEXT:    fmin.d fa5, fa5, fa4
 ; RV64IFD-NEXT:    fcvt.l.d a1, fa5, rtz
 ; RV64IFD-NEXT:    and a0, a0, a1
@@ -1680,12 +1680,12 @@ define signext i16 @fcvt_w_s_sat_i16(double %a) nounwind {
 ; RV64IZFINXZDINX-NEXT:    ld a1, %lo(.LCPI26_0)(a1)
 ; RV64IZFINXZDINX-NEXT:    lui a2, %hi(.LCPI26_1)
 ; RV64IZFINXZDINX-NEXT:    ld a2, %lo(.LCPI26_1)(a2)
-; RV64IZFINXZDINX-NEXT:    fmax.d a1, a0, a1
-; RV64IZFINXZDINX-NEXT:    feq.d a0, a0, a0
-; RV64IZFINXZDINX-NEXT:    neg a0, a0
-; RV64IZFINXZDINX-NEXT:    fmin.d a1, a1, a2
-; RV64IZFINXZDINX-NEXT:    fcvt.l.d a1, a1, rtz
-; RV64IZFINXZDINX-NEXT:    and a0, a0, a1
+; RV64IZFINXZDINX-NEXT:    feq.d a3, a0, a0
+; RV64IZFINXZDINX-NEXT:    neg a3, a3
+; RV64IZFINXZDINX-NEXT:    fmax.d a0, a0, a1
+; RV64IZFINXZDINX-NEXT:    fmin.d a0, a0, a2
+; RV64IZFINXZDINX-NEXT:    fcvt.l.d a0, a0, rtz
+; RV64IZFINXZDINX-NEXT:    and a0, a3, a0
 ; RV64IZFINXZDINX-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i16:
@@ -2026,8 +2026,8 @@ define signext i8 @fcvt_w_s_sat_i8(double %a) nounwind {
 ; RV64IFD-NEXT:    lui a0, %hi(.LCPI30_1)
 ; RV64IFD-NEXT:    fld fa4, %lo(.LCPI30_1)(a0)
 ; RV64IFD-NEXT:    feq.d a0, fa0, fa0
-; RV64IFD-NEXT:    fmax.d fa5, fa0, fa5
 ; RV64IFD-NEXT:    neg a0, a0
+; RV64IFD-NEXT:    fmax.d fa5, fa0, fa5
 ; RV64IFD-NEXT:    fmin.d fa5, fa5, fa4
 ; RV64IFD-NEXT:    fcvt.l.d a1, fa5, rtz
 ; RV64IFD-NEXT:    and a0, a0, a1
@@ -2055,12 +2055,12 @@ define signext i8 @fcvt_w_s_sat_i8(double %a) nounwind {
 ; RV64IZFINXZDINX-NEXT:    ld a1, %lo(.LCPI30_0)(a1)
 ; RV64IZFINXZDINX-NEXT:    lui a2, %hi(.LCPI30_1)
 ; RV64IZFINXZDINX-NEXT:    ld a2, %lo(.LCPI30_1)(a2)
-; RV64IZFINXZDINX-NEXT:    fmax.d a1, a0, a1
-; RV64IZFINXZDINX-NEXT:    feq.d a0, a0, a0
-; RV64IZFINXZDINX-NEXT:    neg a0, a0
-; RV64IZFINXZDINX-NEXT:    fmin.d a1, a1, a2
-; RV64IZFINXZDINX-NEXT:    fcvt.l.d a1, a1, rtz
-; RV64IZFINXZDINX-NEXT:    and a0, a0, a1
+; RV64IZFINXZDINX-NEXT:    feq.d a3, a0, a0
+; RV64IZFINXZDINX-NEXT:    neg a3, a3
+; RV64IZFINXZDINX-NEXT:    fmax.d a0, a0, a1
+; RV64IZFINXZDINX-NEXT:    fmin.d a0, a0, a2
+; RV64IZFINXZDINX-NEXT:    fcvt.l.d a0, a0, rtz
+; RV64IZFINXZDINX-NEXT:    and a0, a3, a0
 ; RV64IZFINXZDINX-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i8:
diff --git a/llvm/test/CodeGen/RISCV/float-convert.ll b/llvm/test/CodeGen/RISCV/float-convert.ll
index 2c7315fbe59f6..653b64ec73049 100644
--- a/llvm/test/CodeGen/RISCV/float-convert.ll
+++ b/llvm/test/CodeGen/RISCV/float-convert.ll
@@ -1424,12 +1424,12 @@ define signext i16 @fcvt_w_s_sat_i16(float %a) nounwind {
 ; RV64IF-LABEL: fcvt_w_s_sat_i16:
 ; RV64IF:       # %bb.0: # %start
 ; RV64IF-NEXT:    feq.s a0, fa0, fa0
+; RV64IF-NEXT:    neg a0, a0
 ; RV64IF-NEXT:    lui a1, %hi(.LCPI24_0)
 ; RV64IF-NEXT:    flw fa5, %lo(.LCPI24_0)(a1)
 ; RV64IF-NEXT:    lui a1, 815104
 ; RV64IF-NEXT:    fmv.w.x fa4, a1
 ; RV64IF-NEXT:    fmax.s fa4, fa0, fa4
-; RV64IF-NEXT:    neg a0, a0
 ; RV64IF-NEXT:    fmin.s fa5, fa4, fa5
 ; RV64IF-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64IF-NEXT:    and a0, a0, a1
@@ -1450,15 +1450,15 @@ define signext i16 @fcvt_w_s_sat_i16(float %a) nounwind {
 ;
 ; RV64IZFINX-LABEL: fcvt_w_s_sat_i16:
 ; RV64IZFINX:       # %bb.0: # %start
-; RV64IZFINX-NEXT:    lui a1, 815104
+; RV64IZFINX-NEXT:    feq.s a1, a0, a0
 ; RV64IZFINX-NEXT:    lui a2, %hi(.LCPI24_0)
 ; RV64IZFINX-NEXT:    lw a2, %lo(.LCPI24_0)(a2)
-; RV64IZFINX-NEXT:    fmax.s a1, a0, a1
-; RV64IZFINX-NEXT:    feq.s a0, a0, a0
-; RV64IZFINX-NEXT:    neg a0, a0
-; RV64IZFINX-NEXT:    fmin.s a1, a1, a2
-; RV64IZFINX-NEXT:    fcvt.l.s a1, a1, rtz
-; RV64IZFINX-NEXT:    and a0, a0, a1
+; RV64IZFINX-NEXT:    neg a1, a1
+; RV64IZFINX-NEXT:    lui a3, 815104
+; RV64IZFINX-NEXT:    fmax.s a0, a0, a3
+; RV64IZFINX-NEXT:    fmin.s a0, a0, a2
+; RV64IZFINX-NEXT:    fcvt.l.s a0, a0, rtz
+; RV64IZFINX-NEXT:    and a0, a1, a0
 ; RV64IZFINX-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i16:
diff --git a/llvm/test/CodeGen/RISCV/half-convert.ll b/llvm/test/CodeGen/RISCV/half-convert.ll
index 16c096290720d..277749c75bbbf 100644
--- a/llvm/test/CodeGen/RISCV/half-convert.ll
+++ b/llvm/test/CodeGen/RISCV/half-convert.ll
@@ -210,12 +210,12 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64IZFH:       # %bb.0: # %start
 ; RV64IZFH-NEXT:    fcvt.s.h fa5, fa0
 ; RV64IZFH-NEXT:    feq.s a0, fa5, fa5
+; RV64IZFH-NEXT:    neg a0, a0
 ; RV64IZFH-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV64IZFH-NEXT:    flw fa4, %lo(.LCPI1_0)(a1)
 ; RV64IZFH-NEXT:    lui a1, 815104
 ; RV64IZFH-NEXT:    fmv.w.x fa3, a1
 ; RV64IZFH-NEXT:    fmax.s fa5, fa5, fa3
-; RV64IZFH-NEXT:    neg a0, a0
 ; RV64IZFH-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64IZFH-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64IZFH-NEXT:    and a0, a0, a1
@@ -240,12 +240,12 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64IDZFH:       # %bb.0: # %start
 ; RV64IDZFH-NEXT:    fcvt.s.h fa5, fa0
 ; RV64IDZFH-NEXT:    feq.s a0, fa5, fa5
+; RV64IDZFH-NEXT:    neg a0, a0
 ; RV64IDZFH-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV64IDZFH-NEXT:    flw fa4, %lo(.LCPI1_0)(a1)
 ; RV64IDZFH-NEXT:    lui a1, 815104
 ; RV64IDZFH-NEXT:    fmv.w.x fa3, a1
 ; RV64IDZFH-NEXT:    fmax.s fa5, fa5, fa3
-; RV64IDZFH-NEXT:    neg a0, a0
 ; RV64IDZFH-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64IDZFH-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64IDZFH-NEXT:    and a0, a0, a1
@@ -268,15 +268,15 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64IZHINX-LABEL: fcvt_si_h_sat:
 ; RV64IZHINX:       # %bb.0: # %start
 ; RV64IZHINX-NEXT:    fcvt.s.h a0, a0
-; RV64IZHINX-NEXT:    lui a1, 815104
+; RV64IZHINX-NEXT:    feq.s a1, a0, a0
 ; RV64IZHINX-NEXT:    lui a2, %hi(.LCPI1_0)
 ; RV64IZHINX-NEXT:    lw a2, %lo(.LCPI1_0)(a2)
-; RV64IZHINX-NEXT:    fmax.s a1, a0, a1
-; RV64IZHINX-NEXT:    feq.s a0, a0, a0
-; RV64IZHINX-NEXT:    neg a0, a0
-; RV64IZHINX-NEXT:    fmin.s a1, a1, a2
-; RV64IZHINX-NEXT:    fcvt.l.s a1, a1, rtz
-; RV64IZHINX-NEXT:    and a0, a0, a1
+; RV64IZHINX-NEXT:    neg a1, a1
+; RV64IZHINX-NEXT:    lui a3, 815104
+; RV64IZHINX-NEXT:    fmax.s a0, a0, a3
+; RV64IZHINX-NEXT:    fmin.s a0, a0, a2
+; RV64IZHINX-NEXT:    fcvt.l.s a0, a0, rtz
+; RV64IZHINX-NEXT:    and a0, a1, a0
 ; RV64IZHINX-NEXT:    ret
 ;
 ; RV32IZDINXZHINX-LABEL: fcvt_si_h_sat:
@@ -296,15 +296,15 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64IZDINXZHINX-LABEL: fcvt_si_h_sat:
 ; RV64IZDINXZHINX:       # %bb.0: # %start
 ; RV64IZDINXZHINX-NEXT:    fcvt.s.h a0, a0
-; RV64IZDINXZHINX-NEXT:    lui a1, 815104
+; RV64IZDINXZHINX-NEXT:    feq.s a1, a0, a0
 ; RV64IZDINXZHINX-NEXT:    lui a2, %hi(.LCPI1_0)
 ; RV64IZDINXZHINX-NEXT:    lw a2, %lo(.LCPI1_0)(a2)
-; RV64IZDINXZHINX-NEXT:    fmax.s a1, a0, a1
-; RV64IZDINXZHINX-NEXT:    feq.s a0, a0, a0
-; RV64IZDINXZHINX-NEXT:    neg a0, a0
-; RV64IZDINXZHINX-NEXT:    fmin.s a1, a1, a2
-; RV64IZDINXZHINX-NEXT:    fcvt.l.s a1, a1, rtz
-; RV64IZDINXZHINX-NEXT:    and a0, a0, a1
+; RV64IZDINXZHINX-NEXT:    neg a1, a1
+; RV64IZDINXZHINX-NEXT:    lui a3, 815104
+; RV64IZDINXZHINX-NEXT:    fmax.s a0, a0, a3
+; RV64IZDINXZHINX-NEXT:    fmin.s a0, a0, a2
+; RV64IZDINXZHINX-NEXT:    fcvt.l.s a0, a0, rtz
+; RV64IZDINXZHINX-NEXT:    and a0, a1, a0
 ; RV64IZDINXZHINX-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_si_h_sat:
@@ -420,12 +420,12 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64ID-LP64-NEXT:    call __extendhfsf2
 ; RV64ID-LP64-NEXT:    fmv.w.x fa5, a0
 ; RV64ID-LP64-NEXT:    feq.s a0, fa5, fa5
+; RV64ID-LP64-NEXT:    neg a0, a0
 ; RV64ID-LP64-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV64ID-LP64-NEXT:    flw fa4, %lo(.LCPI1_0)(a1)
 ; RV64ID-LP64-NEXT:    lui a1, 815104
 ; RV64ID-LP64-NEXT:    fmv.w.x fa3, a1
 ; RV64ID-LP64-NEXT:    fmax.s fa5, fa5, fa3
-; RV64ID-LP64-NEXT:    neg a0, a0
 ; RV64ID-LP64-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64ID-LP64-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64ID-LP64-NEXT:    and a0, a0, a1
@@ -458,12 +458,12 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; RV64ID-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
 ; RV64ID-NEXT:    call __extendhfsf2
 ; RV64ID-NEXT:    feq.s a0, fa0, fa0
+; RV64ID-NEXT:    neg a0, a0
 ; RV64ID-NEXT:    lui a1, %hi(.LCPI1_0)
 ; RV64ID-NEXT:    flw fa5, %lo(.LCPI1_0)(a1)
 ; RV64ID-NEXT:    lui a1, 815104
 ; RV64ID-NEXT:    fmv.w.x fa4, a1
 ; RV64ID-NEXT:    fmax.s fa4, fa0, fa4
-; RV64ID-NEXT:    neg a0, a0
 ; RV64ID-NEXT:    fmin.s fa5, fa4, fa5
 ; RV64ID-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64ID-NEXT:    and a0, a0, a1
@@ -490,12 +490,12 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; CHECK64-IZFHMIN:       # %bb.0: # %start
 ; CHECK64-IZFHMIN-NEXT:    fcvt.s.h fa5, fa0
 ; CHECK64-IZFHMIN-NEXT:    feq.s a0, fa5, fa5
+; CHECK64-IZFHMIN-NEXT:    neg a0, a0
 ; CHECK64-IZFHMIN-NEXT:    lui a1, %hi(.LCPI1_0)
 ; CHECK64-IZFHMIN-NEXT:    flw fa4, %lo(.LCPI1_0)(a1)
 ; CHECK64-IZFHMIN-NEXT:    lui a1, 815104
 ; CHECK64-IZFHMIN-NEXT:    fmv.w.x fa3, a1
 ; CHECK64-IZFHMIN-NEXT:    fmax.s fa5, fa5, fa3
-; CHECK64-IZFHMIN-NEXT:    neg a0, a0
 ; CHECK64-IZFHMIN-NEXT:    fmin.s fa5, fa5, fa4
 ; CHECK64-IZFHMIN-NEXT:    fcvt.l.s a1, fa5, rtz
 ; CHECK64-IZFHMIN-NEXT:    and a0, a0, a1
@@ -518,15 +518,15 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; CHECK64-IZHINXMIN-LABEL: fcvt_si_h_sat:
 ; CHECK64-IZHINXMIN:       # %bb.0: # %start
 ; CHECK64-IZHINXMIN-NEXT:    fcvt.s.h a0, a0
-; CHECK64-IZHINXMIN-NEXT:    lui a1, 815104
+; CHECK64-IZHINXMIN-NEXT:    feq.s a1, a0, a0
 ; CHECK64-IZHINXMIN-NEXT:    lui a2, %hi(.LCPI1_0)
 ; CHECK64-IZHINXMIN-NEXT:    lw a2, %lo(.LCPI1_0)(a2)
-; CHECK64-IZHINXMIN-NEXT:    fmax.s a1, a0, a1
-; CHECK64-IZHINXMIN-NEXT:    feq.s a0, a0, a0
-; CHECK64-IZHINXMIN-NEXT:    neg a0, a0
-; CHECK64-IZHINXMIN-NEXT:    fmin.s a1, a1, a2
-; CHECK64-IZHINXMIN-NEXT:    fcvt.l.s a1, a1, rtz
-; CHECK64-IZHINXMIN-NEXT:    and a0, a0, a1
+; CHECK64-IZHINXMIN-NEXT:    neg a1, a1
+; CHECK64-IZHINXMIN-NEXT:    lui a3, 815104
+; CHECK64-IZHINXMIN-NEXT:    fmax.s a0, a0, a3
+; CHECK64-IZHINXMIN-NEXT:    fmin.s a0, a0, a2
+; CHECK64-IZHINXMIN-NEXT:    fcvt.l.s a0, a0, rtz
+; CHECK64-IZHINXMIN-NEXT:    and a0, a1, a0
 ; CHECK64-IZHINXMIN-NEXT:    ret
 ;
 ; CHECK32-IZDINXZHINXMIN-LABEL: fcvt_si_h_sat:
@@ -546,15 +546,15 @@ define i16 @fcvt_si_h_sat(half %a) nounwind {
 ; CHECK64-IZDINXZHINXMIN-LABEL: fcvt_si_h_sat:
 ; CHECK64-IZDINXZHINXMIN:       # %bb.0: # %start
 ; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.s.h a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    lui a1, 815104
+; CHECK64-IZDINXZHINXMIN-NEXT:    feq.s a1, a0, a0
 ; CHECK64-IZDINXZHINXMIN-NEXT:    lui a2, %hi(.LCPI1_0)
 ; CHECK64-IZDINXZHINXMIN-NEXT:    lw a2, %lo(.LCPI1_0)(a2)
-; CHECK64-IZDINXZHINXMIN-NEXT:    fmax.s a1, a0, a1
-; CHECK64-IZDINXZHINXMIN-NEXT:    feq.s a0, a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    neg a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    fmin.s a1, a1, a2
-; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.l.s a1, a1, rtz
-; CHECK64-IZDINXZHINXMIN-NEXT:    and a0, a0, a1
+; CHECK64-IZDINXZHINXMIN-NEXT:    neg a1, a1
+; CHECK64-IZDINXZHINXMIN-NEXT:    lui a3, 815104
+; CHECK64-IZDINXZHINXMIN-NEXT:    fmax.s a0, a0, a3
+; CHECK64-IZDINXZHINXMIN-NEXT:    fmin.s a0, a0, a2
+; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.l.s a0, a0, rtz
+; CHECK64-IZDINXZHINXMIN-NEXT:    and a0, a1, a0
 ; CHECK64-IZDINXZHINXMIN-NEXT:    ret
 start:
   %0 = tail call i16 @llvm.fptosi.sat.i16.f16(half %a)
@@ -6377,12 +6377,12 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64IZFH:       # %bb.0: # %start
 ; RV64IZFH-NEXT:    fcvt.s.h fa5, fa0
 ; RV64IZFH-NEXT:    feq.s a0, fa5, fa5
+; RV64IZFH-NEXT:    neg a0, a0
 ; RV64IZFH-NEXT:    lui a1, %hi(.LCPI32_0)
 ; RV64IZFH-NEXT:    flw fa4, %lo(.LCPI32_0)(a1)
 ; RV64IZFH-NEXT:    lui a1, 815104
 ; RV64IZFH-NEXT:    fmv.w.x fa3, a1
 ; RV64IZFH-NEXT:    fmax.s fa5, fa5, fa3
-; RV64IZFH-NEXT:    neg a0, a0
 ; RV64IZFH-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64IZFH-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64IZFH-NEXT:    and a0, a0, a1
@@ -6407,12 +6407,12 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64IDZFH:       # %bb.0: # %start
 ; RV64IDZFH-NEXT:    fcvt.s.h fa5, fa0
 ; RV64IDZFH-NEXT:    feq.s a0, fa5, fa5
+; RV64IDZFH-NEXT:    neg a0, a0
 ; RV64IDZFH-NEXT:    lui a1, %hi(.LCPI32_0)
 ; RV64IDZFH-NEXT:    flw fa4, %lo(.LCPI32_0)(a1)
 ; RV64IDZFH-NEXT:    lui a1, 815104
 ; RV64IDZFH-NEXT:    fmv.w.x fa3, a1
 ; RV64IDZFH-NEXT:    fmax.s fa5, fa5, fa3
-; RV64IDZFH-NEXT:    neg a0, a0
 ; RV64IDZFH-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64IDZFH-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64IDZFH-NEXT:    and a0, a0, a1
@@ -6435,15 +6435,15 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64IZHINX-LABEL: fcvt_w_s_sat_i16:
 ; RV64IZHINX:       # %bb.0: # %start
 ; RV64IZHINX-NEXT:    fcvt.s.h a0, a0
-; RV64IZHINX-NEXT:    lui a1, 815104
+; RV64IZHINX-NEXT:    feq.s a1, a0, a0
 ; RV64IZHINX-NEXT:    lui a2, %hi(.LCPI32_0)
 ; RV64IZHINX-NEXT:    lw a2, %lo(.LCPI32_0)(a2)
-; RV64IZHINX-NEXT:    fmax.s a1, a0, a1
-; RV64IZHINX-NEXT:    feq.s a0, a0, a0
-; RV64IZHINX-NEXT:    neg a0, a0
-; RV64IZHINX-NEXT:    fmin.s a1, a1, a2
-; RV64IZHINX-NEXT:    fcvt.l.s a1, a1, rtz
-; RV64IZHINX-NEXT:    and a0, a0, a1
+; RV64IZHINX-NEXT:    neg a1, a1
+; RV64IZHINX-NEXT:    lui a3, 815104
+; RV64IZHINX-NEXT:    fmax.s a0, a0, a3
+; RV64IZHINX-NEXT:    fmin.s a0, a0, a2
+; RV64IZHINX-NEXT:    fcvt.l.s a0, a0, rtz
+; RV64IZHINX-NEXT:    and a0, a1, a0
 ; RV64IZHINX-NEXT:    ret
 ;
 ; RV32IZDINXZHINX-LABEL: fcvt_w_s_sat_i16:
@@ -6463,15 +6463,15 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64IZDINXZHINX-LABEL: fcvt_w_s_sat_i16:
 ; RV64IZDINXZHINX:       # %bb.0: # %start
 ; RV64IZDINXZHINX-NEXT:    fcvt.s.h a0, a0
-; RV64IZDINXZHINX-NEXT:    lui a1, 815104
+; RV64IZDINXZHINX-NEXT:    feq.s a1, a0, a0
 ; RV64IZDINXZHINX-NEXT:    lui a2, %hi(.LCPI32_0)
 ; RV64IZDINXZHINX-NEXT:    lw a2, %lo(.LCPI32_0)(a2)
-; RV64IZDINXZHINX-NEXT:    fmax.s a1, a0, a1
-; RV64IZDINXZHINX-NEXT:    feq.s a0, a0, a0
-; RV64IZDINXZHINX-NEXT:    neg a0, a0
-; RV64IZDINXZHINX-NEXT:    fmin.s a1, a1, a2
-; RV64IZDINXZHINX-NEXT:    fcvt.l.s a1, a1, rtz
-; RV64IZDINXZHINX-NEXT:    and a0, a0, a1
+; RV64IZDINXZHINX-NEXT:    neg a1, a1
+; RV64IZDINXZHINX-NEXT:    lui a3, 815104
+; RV64IZDINXZHINX-NEXT:    fmax.s a0, a0, a3
+; RV64IZDINXZHINX-NEXT:    fmin.s a0, a0, a2
+; RV64IZDINXZHINX-NEXT:    fcvt.l.s a0, a0, rtz
+; RV64IZDINXZHINX-NEXT:    and a0, a1, a0
 ; RV64IZDINXZHINX-NEXT:    ret
 ;
 ; RV32I-LABEL: fcvt_w_s_sat_i16:
@@ -6591,12 +6591,12 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64ID-LP64-NEXT:    call __extendhfsf2
 ; RV64ID-LP64-NEXT:    fmv.w.x fa5, a0
 ; RV64ID-LP64-NEXT:    feq.s a0, fa5, fa5
+; RV64ID-LP64-NEXT:    neg a0, a0
 ; RV64ID-LP64-NEXT:    lui a1, %hi(.LCPI32_0)
 ; RV64ID-LP64-NEXT:    flw fa4, %lo(.LCPI32_0)(a1)
 ; RV64ID-LP64-NEXT:    lui a1, 815104
 ; RV64ID-LP64-NEXT:    fmv.w.x fa3, a1
 ; RV64ID-LP64-NEXT:    fmax.s fa5, fa5, fa3
-; RV64ID-LP64-NEXT:    neg a0, a0
 ; RV64ID-LP64-NEXT:    fmin.s fa5, fa5, fa4
 ; RV64ID-LP64-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64ID-LP64-NEXT:    and a0, a0, a1
@@ -6629,12 +6629,12 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; RV64ID-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
 ; RV64ID-NEXT:    call __extendhfsf2
 ; RV64ID-NEXT:    feq.s a0, fa0, fa0
+; RV64ID-NEXT:    neg a0, a0
 ; RV64ID-NEXT:    lui a1, %hi(.LCPI32_0)
 ; RV64ID-NEXT:    flw fa5, %lo(.LCPI32_0)(a1)
 ; RV64ID-NEXT:    lui a1, 815104
 ; RV64ID-NEXT:    fmv.w.x fa4, a1
 ; RV64ID-NEXT:    fmax.s fa4, fa0, fa4
-; RV64ID-NEXT:    neg a0, a0
 ; RV64ID-NEXT:    fmin.s fa5, fa4, fa5
 ; RV64ID-NEXT:    fcvt.l.s a1, fa5, rtz
 ; RV64ID-NEXT:    and a0, a0, a1
@@ -6661,12 +6661,12 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; CHECK64-IZFHMIN:       # %bb.0: # %start
 ; CHECK64-IZFHMIN-NEXT:    fcvt.s.h fa5, fa0
 ; CHECK64-IZFHMIN-NEXT:    feq.s a0, fa5, fa5
+; CHECK64-IZFHMIN-NEXT:    neg a0, a0
 ; CHECK64-IZFHMIN-NEXT:    lui a1, %hi(.LCPI32_0)
 ; CHECK64-IZFHMIN-NEXT:    flw fa4, %lo(.LCPI32_0)(a1)
 ; CHECK64-IZFHMIN-NEXT:    lui a1, 815104
 ; CHECK64-IZFHMIN-NEXT:    fmv.w.x fa3, a1
 ; CHECK64-IZFHMIN-NEXT:    fmax.s fa5, fa5, fa3
-; CHECK64-IZFHMIN-NEXT:    neg a0, a0
 ; CHECK64-IZFHMIN-NEXT:    fmin.s fa5, fa5, fa4
 ; CHECK64-IZFHMIN-NEXT:    fcvt.l.s a1, fa5, rtz
 ; CHECK64-IZFHMIN-NEXT:    and a0, a0, a1
@@ -6689,15 +6689,15 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; CHECK64-IZHINXMIN-LABEL: fcvt_w_s_sat_i16:
 ; CHECK64-IZHINXMIN:       # %bb.0: # %start
 ; CHECK64-IZHINXMIN-NEXT:    fcvt.s.h a0, a0
-; CHECK64-IZHINXMIN-NEXT:    lui a1, 815104
+; CHECK64-IZHINXMIN-NEXT:    feq.s a1, a0, a0
 ; CHECK64-IZHINXMIN-NEXT:    lui a2, %hi(.LCPI32_0)
 ; CHECK64-IZHINXMIN-NEXT:    lw a2, %lo(.LCPI32_0)(a2)
-; CHECK64-IZHINXMIN-NEXT:    fmax.s a1, a0, a1
-; CHECK64-IZHINXMIN-NEXT:    feq.s a0, a0, a0
-; CHECK64-IZHINXMIN-NEXT:    neg a0, a0
-; CHECK64-IZHINXMIN-NEXT:    fmin.s a1, a1, a2
-; CHECK64-IZHINXMIN-NEXT:    fcvt.l.s a1, a1, rtz
-; CHECK64-IZHINXMIN-NEXT:    and a0, a0, a1
+; CHECK64-IZHINXMIN-NEXT:    neg a1, a1
+; CHECK64-IZHINXMIN-NEXT:    lui a3, 815104
+; CHECK64-IZHINXMIN-NEXT:    fmax.s a0, a0, a3
+; CHECK64-IZHINXMIN-NEXT:    fmin.s a0, a0, a2
+; CHECK64-IZHINXMIN-NEXT:    fcvt.l.s a0, a0, rtz
+; CHECK64-IZHINXMIN-NEXT:    and a0, a1, a0
 ; CHECK64-IZHINXMIN-NEXT:    ret
 ;
 ; CHECK32-IZDINXZHINXMIN-LABEL: fcvt_w_s_sat_i16:
@@ -6717,15 +6717,15 @@ define signext i16 @fcvt_w_s_sat_i16(half %a) nounwind {
 ; CHECK64-IZDINXZHINXMIN-LABEL: fcvt_w_s_sat_i16:
 ; CHECK64-IZDINXZHINXMIN:       # %bb.0: # %start
 ; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.s.h a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    lui a1, 815104
+; CHECK64-IZDINXZHINXMIN-NEXT:    feq.s a1, a0, a0
 ; CHECK64-IZDINXZHINXMIN-NEXT:    lui a2, %hi(.LCPI32_0)
 ; CHECK64-IZDINXZHINXMIN-NEXT:    lw a2, %lo(.LCPI32_0)(a2)
-; CHECK64-IZDINXZHINXMIN-NEXT:    fmax.s a1, a0, a1
-; CHECK64-IZDINXZHINXMIN-NEXT:    feq.s a0, a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    neg a0, a0
-; CHECK64-IZDINXZHINXMIN-NEXT:    fmin.s a1, a1, a2
-; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.l.s a1, a1, rtz
-; CHECK64-IZDINXZHINXMIN-NEXT:    and a0, a0, a1
+; CHECK64-IZDINXZHINXMIN-NEXT:    neg a1, a1
+; CHECK64-IZDINXZHINXMIN-NEXT:    lui a3, 815104
+; CHECK64-IZDINXZHINXMIN-NEXT:    fmax.s a0, a0, a3
+; CHECK64-IZDINXZHINXMIN-NEXT:    fmin.s a0, a0, a2
+; CHECK64-IZDINXZHINXMIN-NEXT:    fcvt.l.s a0, a0, rtz
+; CHECK64-IZDINXZHINXMIN-NEXT:    and a0, a1, a0
 ; CHECK64-IZDINXZHINXMIN-NEXT:    ret
 start:
   %0 = tail call i16 @llvm.fptosi.sat.i16.f16(half %a)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp-shuffles.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp-shuffles.ll
index 59eb4b89a2f56..b0f6bebea0381 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp-shuffles.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp-shuffles.ll
@@ -125,35 +125,20 @@ define <4 x double> @vrgather_shuffle_xv_v4f64(<4 x double> %x) {
 }
 
 define <4 x double> @vrgather_shuffle_vx_v4f64(<4 x double> %x) {
-; RV32-LABEL: vrgather_shuffle_vx_v4f64:
-; RV32:       # %bb.0:
-; RV32-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
-; RV32-NEXT:    vid.v v12
-; RV32-NEXT:    li a0, 3
-; RV32-NEXT:    lui a1, %hi(.LCPI8_0)
-; RV32-NEXT:    addi a1, a1, %lo(.LCPI8_0)
-; RV32-NEXT:    vlse64.v v10, (a1), zero
-; RV32-NEXT:    vmul.vx v12, v12, a0
-; RV32-NEXT:    vmv.v.i v0, 3
-; RV32-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
-; RV32-NEXT:    vrgatherei16.vv v10, v8, v12, v0.t
-; RV32-NEXT:    vmv.v.v v8, v10
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: vrgather_shuffle_vx_v4f64:
-; RV64:       # %bb.0:
-; RV64-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
-; RV64-NEXT:    vid.v v12
-; RV64-NEXT:    lui a0, %hi(.LCPI8_0)
-; RV64-NEXT:    addi a0, a0, %lo(.LCPI8_0)
-; RV64-NEXT:    vlse64.v v10, (a0), zero
-; RV64-NEXT:    li a0, 3
-; RV64-NEXT:    vmul.vx v12, v12, a0
-; RV64-NEXT:    vmv.v.i v0, 3
-; RV64-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
-; RV64-NEXT:    vrgatherei16.vv v10, v8, v12, v0.t
-; RV64-NEXT:    vmv.v.v v8, v10
-; RV64-NEXT:    ret
+; CHECK-LABEL: vrgather_shuffle_vx_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
+; CHECK-NEXT:    vid.v v12
+; CHECK-NEXT:    li a0, 3
+; CHECK-NEXT:    lui a1, %hi(.LCPI8_0)
+; CHECK-NEXT:    addi a1, a1, %lo(.LCPI8_0)
+; CHECK-NEXT:    vlse64.v v10, (a1), zero
+; CHECK-NEXT:    vmul.vx v12, v12, a0
+; CHECK-NEXT:    vmv.v.i v0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vrgatherei16.vv v10, v8, v12, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
   %s = shufflevector <4 x double> %x, <4 x double> <double 2.0, double 2.0, double 2.0, double 2.0>, <4 x i32> <i32 0, i32 3, i32 6, i32 5>
   ret <4 x double> %s
 }
@@ -279,3 +264,6 @@ define <8 x double> @splice_binary2(<8 x double> %x, <8 x double> %y) {
   %s = shufflevector <8 x double> %x, <8 x double> %y, <8 x i32> <i32 13, i32 14, i32 15, i32 0, i32 1, i32 2, i32 3, i32 4>
   ret <8 x double> %s
 }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; RV32: {{.*}}
+; RV64: {{.*}}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
index 9f0240c53b219..1d3c22a02efc0 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
@@ -774,9 +774,9 @@ define void @insert_v2i64_nxv16i64_hi(ptr %psv, ptr %out) {
 ; RV64VLS-NEXT:    vl1re64.v v8, (a0)
 ; RV64VLS-NEXT:    addi a0, sp, 128
 ; RV64VLS-NEXT:    vs1r.v v8, (a0)
-; RV64VLS-NEXT:    addi a0, sp, 192
-; RV64VLS-NEXT:    vl8re64.v v8, (a0)
 ; RV64VLS-NEXT:    addi a0, sp, 64
+; RV64VLS-NEXT:    addi a2, sp, 192
+; RV64VLS-NEXT:    vl8re64.v v8, (a2)
 ; RV64VLS-NEXT:    vl8re64.v v16, (a0)
 ; RV64VLS-NEXT:    addi a0, a1, 128
 ; RV64VLS-NEXT:    vs8r.v v8, (a0)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-fp-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-fp-vp.ll
index 6ba90b00fdba5..4598bf67a2363 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-fp-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-fp-vp.ll
@@ -1142,109 +1142,57 @@ define <8 x i1> @fcmp_uno_vf_swap_v8f16(<8 x half> %va, half %b, <8 x i1> %m, i3
 declare <128 x i1> @llvm.vp.fcmp.v128f16(<128 x half>, <128 x half>, metadata, <128 x i1>, i32)
 
 define <128 x i1> @fcmp_oeq_vv_v128f16(<128 x half> %va, <128 x half> %vb, <128 x i1> %m, i32 zeroext %evl) {
-; ZVFH32-LABEL: fcmp_oeq_vv_v128f16:
-; ZVFH32:       # %bb.0:
-; ZVFH32-NEXT:    addi sp, sp, -16
-; ZVFH32-NEXT:    .cfi_def_cfa_offset 16
-; ZVFH32-NEXT:    csrr a1, vlenb
-; ZVFH32-NEXT:    slli a1, a1, 4
-; ZVFH32-NEXT:    sub sp, sp, a1
-; ZVFH32-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
-; ZVFH32-NEXT:    addi a1, a0, 128
-; ZVFH32-NEXT:    li a3, 64
-; ZVFH32-NEXT:    vsetvli zero, a3, e16, m8, ta, ma
-; ZVFH32-NEXT:    vle16.v v24, (a1)
-; ZVFH32-NEXT:    csrr a1, vlenb
-; ZVFH32-NEXT:    slli a1, a1, 3
-; ZVFH32-NEXT:    add a1, sp, a1
-; ZVFH32-NEXT:    addi a1, a1, 16
-; ZVFH32-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
-; ZVFH32-NEXT:    vle16.v v24, (a0)
-; ZVFH32-NEXT:    addi a0, sp, 16
-; ZVFH32-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
-; ZVFH32-NEXT:    vsetivli zero, 8, e8, m1, ta, ma
-; ZVFH32-NEXT:    vslidedown.vi v7, v0, 8
-; ZVFH32-NEXT:    mv a0, a2
-; ZVFH32-NEXT:    bltu a2, a3, .LBB43_2
-; ZVFH32-NEXT:  # %bb.1:
-; ZVFH32-NEXT:    li a0, 64
-; ZVFH32-NEXT:  .LBB43_2:
-; ZVFH32-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
-; ZVFH32-NEXT:    addi a0, sp, 16
-; ZVFH32-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
-; ZVFH32-NEXT:    vmfeq.vv v6, v8, v24, v0.t
-; ZVFH32-NEXT:    addi a0, a2, -64
-; ZVFH32-NEXT:    sltu a1, a2, a0
-; ZVFH32-NEXT:    addi a1, a1, -1
-; ZVFH32-NEXT:    and a0, a1, a0
-; ZVFH32-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
-; ZVFH32-NEXT:    vmv1r.v v0, v7
-; ZVFH32-NEXT:    csrr a0, vlenb
-; ZVFH32-NEXT:    slli a0, a0, 3
-; ZVFH32-NEXT:    add a0, sp, a0
-; ZVFH32-NEXT:    addi a0, a0, 16
-; ZVFH32-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
-; ZVFH32-NEXT:    vmfeq.vv v24, v16, v8, v0.t
-; ZVFH32-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
-; ZVFH32-NEXT:    vslideup.vi v6, v24, 8
-; ZVFH32-NEXT:    vmv.v.v v0, v6
-; ZVFH32-NEXT:    csrr a0, vlenb
-; ZVFH32-NEXT:    slli a0, a0, 4
-; ZVFH32-NEXT:    add sp, sp, a0
-; ZVFH32-NEXT:    addi sp, sp, 16
-; ZVFH32-NEXT:    ret
-;
-; ZVFH64-LABEL: fcmp_oeq_vv_v128f16:
-; ZVFH64:       # %bb.0:
-; ZVFH64-NEXT:    addi sp, sp, -16
-; ZVFH64-NEXT:    .cfi_def_cfa_offset 16
-; ZVFH64-NEXT:    csrr a1, vlenb
-; ZVFH64-NEXT:    slli a1, a1, 4
-; ZVFH64-NEXT:    sub sp, sp, a1
-; ZVFH64-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
-; ZVFH64-NEXT:    addi a1, a0, 128
-; ZVFH64-NEXT:    li a3, 64
-; ZVFH64-NEXT:    vsetvli zero, a3, e16, m8, ta, ma
-; ZVFH64-NEXT:    vle16.v v24, (a1)
-; ZVFH64-NEXT:    csrr a1, vlenb
-; ZVFH64-NEXT:    slli a1, a1, 3
-; ZVFH64-NEXT:    add a1, sp, a1
-; ZVFH64-NEXT:    addi a1, a1, 16
-; ZVFH64-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
-; ZVFH64-NEXT:    vle16.v v24, (a0)
-; ZVFH64-NEXT:    addi a0, sp, 16
-; ZVFH64-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
-; ZVFH64-NEXT:    vsetivli zero, 8, e8, m1, ta, ma
-; ZVFH64-NEXT:    mv a0, a2
-; ZVFH64-NEXT:    vslidedown.vi v7, v0, 8
-; ZVFH64-NEXT:    bltu a2, a3, .LBB43_2
-; ZVFH64-NEXT:  # %bb.1:
-; ZVFH64-NEXT:    li a0, 64
-; ZVFH64-NEXT:  .LBB43_2:
-; ZVFH64-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
-; ZVFH64-NEXT:    addi a0, sp, 16
-; ZVFH64-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
-; ZVFH64-NEXT:    vmfeq.vv v6, v8, v24, v0.t
-; ZVFH64-NEXT:    addi a0, a2, -64
-; ZVFH64-NEXT:    sltu a1, a2, a0
-; ZVFH64-NEXT:    addi a1, a1, -1
-; ZVFH64-NEXT:    and a0, a1, a0
-; ZVFH64-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
-; ZVFH64-NEXT:    vmv1r.v v0, v7
-; ZVFH64-NEXT:    csrr a0, vlenb
-; ZVFH64-NEXT:    slli a0, a0, 3
-; ZVFH64-NEXT:    add a0, sp, a0
-; ZVFH64-NEXT:    addi a0, a0, 16
-; ZVFH64-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
-; ZVFH64-NEXT:    vmfeq.vv v24, v16, v8, v0.t
-; ZVFH64-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
-; ZVFH64-NEXT:    vslideup.vi v6, v24, 8
-; ZVFH64-NEXT:    vmv.v.v v0, v6
-; ZVFH64-NEXT:    csrr a0, vlenb
-; ZVFH64-NEXT:    slli a0, a0, 4
-; ZVFH64-NEXT:    add sp, sp, a0
-; ZVFH64-NEXT:    addi sp, sp, 16
-; ZVFH64-NEXT:    ret
+; ZVFH-LABEL: fcmp_oeq_vv_v128f16:
+; ZVFH:       # %bb.0:
+; ZVFH-NEXT:    addi sp, sp, -16
+; ZVFH-NEXT:    .cfi_def_cfa_offset 16
+; ZVFH-NEXT:    csrr a1, vlenb
+; ZVFH-NEXT:    slli a1, a1, 4
+; ZVFH-NEXT:    sub sp, sp, a1
+; ZVFH-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
+; ZVFH-NEXT:    addi a1, a0, 128
+; ZVFH-NEXT:    li a3, 64
+; ZVFH-NEXT:    vsetvli zero, a3, e16, m8, ta, ma
+; ZVFH-NEXT:    vle16.v v24, (a1)
+; ZVFH-NEXT:    csrr a1, vlenb
+; ZVFH-NEXT:    slli a1, a1, 3
+; ZVFH-NEXT:    add a1, sp, a1
+; ZVFH-NEXT:    addi a1, a1, 16
+; ZVFH-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
+; ZVFH-NEXT:    vle16.v v24, (a0)
+; ZVFH-NEXT:    addi a0, sp, 16
+; ZVFH-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
+; ZVFH-NEXT:    vsetivli zero, 8, e8, m1, ta, ma
+; ZVFH-NEXT:    vslidedown.vi v7, v0, 8
+; ZVFH-NEXT:    mv a0, a2
+; ZVFH-NEXT:    bltu a2, a3, .LBB43_2
+; ZVFH-NEXT:  # %bb.1:
+; ZVFH-NEXT:    li a0, 64
+; ZVFH-NEXT:  .LBB43_2:
+; ZVFH-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
+; ZVFH-NEXT:    addi a0, sp, 16
+; ZVFH-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
+; ZVFH-NEXT:    vmfeq.vv v6, v8, v24, v0.t
+; ZVFH-NEXT:    addi a0, a2, -64
+; ZVFH-NEXT:    sltu a1, a2, a0
+; ZVFH-NEXT:    addi a1, a1, -1
+; ZVFH-NEXT:    and a0, a1, a0
+; ZVFH-NEXT:    vsetvli zero, a0, e16, m8, ta, ma
+; ZVFH-NEXT:    vmv1r.v v0, v7
+; ZVFH-NEXT:    csrr a0, vlenb
+; ZVFH-NEXT:    slli a0, a0, 3
+; ZVFH-NEXT:    add a0, sp, a0
+; ZVFH-NEXT:    addi a0, a0, 16
+; ZVFH-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
+; ZVFH-NEXT:    vmfeq.vv v24, v16, v8, v0.t
+; ZVFH-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; ZVFH-NEXT:    vslideup.vi v6, v24, 8
+; ZVFH-NEXT:    vmv.v.v v0, v6
+; ZVFH-NEXT:    csrr a0, vlenb
+; ZVFH-NEXT:    slli a0, a0, 4
+; ZVFH-NEXT:    add sp, sp, a0
+; ZVFH-NEXT:    addi sp, sp, 16
+; ZVFH-NEXT:    ret
 ;
 ; ZVFHMIN32-LABEL: fcmp_oeq_vv_v128f16:
 ; ZVFHMIN32:       # %bb.0:
@@ -2956,3 +2904,6 @@ define <32 x i1> @fcmp_oeq_vv_v32f64(<32 x double> %va, <32 x double> %vb, <32 x
   %v = call <32 x i1> @llvm.vp.fcmp.v32f64(<32 x double> %va, <32 x double> %vb, metadata !"oeq", <32 x i1> %m, i32 %evl)
   ret <32 x i1> %v
 }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; ZVFH32: {{.*}}
+; ZVFH64: {{.*}}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-int-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-int-vp.ll
index 981715bd2b998..21bbca00921d6 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-int-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-setcc-int-vp.ll
@@ -1243,109 +1243,57 @@ define <8 x i1> @icmp_sle_vi_swap_v8i32(<8 x i32> %va, <8 x i1> %m, i32 zeroext
 declare <64 x i1> @llvm.vp.icmp.v64i32(<64 x i32>, <64 x i32>, metadata, <64 x i1>, i32)
 
 define <64 x i1> @icmp_eq_vv_v64i32(<64 x i32> %va, <64 x i32> %vb, <64 x i1> %m, i32 zeroext %evl) {
-; RV32-LABEL: icmp_eq_vv_v64i32:
-; RV32:       # %bb.0:
-; RV32-NEXT:    addi sp, sp, -16
-; RV32-NEXT:    .cfi_def_cfa_offset 16
-; RV32-NEXT:    csrr a1, vlenb
-; RV32-NEXT:    slli a1, a1, 4
-; RV32-NEXT:    sub sp, sp, a1
-; RV32-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
-; RV32-NEXT:    addi a1, a0, 128
-; RV32-NEXT:    li a3, 32
-; RV32-NEXT:    vsetvli zero, a3, e32, m8, ta, ma
-; RV32-NEXT:    vle32.v v24, (a1)
-; RV32-NEXT:    csrr a1, vlenb
-; RV32-NEXT:    slli a1, a1, 3
-; RV32-NEXT:    add a1, sp, a1
-; RV32-NEXT:    addi a1, a1, 16
-; RV32-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
-; RV32-NEXT:    vle32.v v24, (a0)
-; RV32-NEXT:    addi a0, sp, 16
-; RV32-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
-; RV32-NEXT:    vsetivli zero, 4, e8, mf2, ta, ma
-; RV32-NEXT:    vslidedown.vi v7, v0, 4
-; RV32-NEXT:    mv a0, a2
-; RV32-NEXT:    bltu a2, a3, .LBB99_2
-; RV32-NEXT:  # %bb.1:
-; RV32-NEXT:    li a0, 32
-; RV32-NEXT:  .LBB99_2:
-; RV32-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
-; RV32-NEXT:    addi a0, sp, 16
-; RV32-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
-; RV32-NEXT:    vmseq.vv v6, v8, v24, v0.t
-; RV32-NEXT:    addi a0, a2, -32
-; RV32-NEXT:    sltu a1, a2, a0
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    and a0, a1, a0
-; RV32-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
-; RV32-NEXT:    vmv1r.v v0, v7
-; RV32-NEXT:    csrr a0, vlenb
-; RV32-NEXT:    slli a0, a0, 3
-; RV32-NEXT:    add a0, sp, a0
-; RV32-NEXT:    addi a0, a0, 16
-; RV32-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
-; RV32-NEXT:    vmseq.vv v24, v16, v8, v0.t
-; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; RV32-NEXT:    vslideup.vi v6, v24, 4
-; RV32-NEXT:    vmv1r.v v0, v6
-; RV32-NEXT:    csrr a0, vlenb
-; RV32-NEXT:    slli a0, a0, 4
-; RV32-NEXT:    add sp, sp, a0
-; RV32-NEXT:    addi sp, sp, 16
-; RV32-NEXT:    ret
-;
-; RV64-LABEL: icmp_eq_vv_v64i32:
-; RV64:       # %bb.0:
-; RV64-NEXT:    addi sp, sp, -16
-; RV64-NEXT:    .cfi_def_cfa_offset 16
-; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 4
-; RV64-NEXT:    sub sp, sp, a1
-; RV64-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
-; RV64-NEXT:    addi a1, a0, 128
-; RV64-NEXT:    li a3, 32
-; RV64-NEXT:    vsetvli zero, a3, e32, m8, ta, ma
-; RV64-NEXT:    vle32.v v24, (a1)
-; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 3
-; RV64-NEXT:    add a1, sp, a1
-; RV64-NEXT:    addi a1, a1, 16
-; RV64-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
-; RV64-NEXT:    vle32.v v24, (a0)
-; RV64-NEXT:    addi a0, sp, 16
-; RV64-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
-; RV64-NEXT:    vsetivli zero, 4, e8, mf2, ta, ma
-; RV64-NEXT:    mv a0, a2
-; RV64-NEXT:    vslidedown.vi v7, v0, 4
-; RV64-NEXT:    bltu a2, a3, .LBB99_2
-; RV64-NEXT:  # %bb.1:
-; RV64-NEXT:    li a0, 32
-; RV64-NEXT:  .LBB99_2:
-; RV64-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
-; RV64-NEXT:    addi a0, sp, 16
-; RV64-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
-; RV64-NEXT:    vmseq.vv v6, v8, v24, v0.t
-; RV64-NEXT:    addi a0, a2, -32
-; RV64-NEXT:    sltu a1, a2, a0
-; RV64-NEXT:    addi a1, a1, -1
-; RV64-NEXT:    and a0, a1, a0
-; RV64-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
-; RV64-NEXT:    vmv1r.v v0, v7
-; RV64-NEXT:    csrr a0, vlenb
-; RV64-NEXT:    slli a0, a0, 3
-; RV64-NEXT:    add a0, sp, a0
-; RV64-NEXT:    addi a0, a0, 16
-; RV64-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
-; RV64-NEXT:    vmseq.vv v24, v16, v8, v0.t
-; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; RV64-NEXT:    vslideup.vi v6, v24, 4
-; RV64-NEXT:    vmv1r.v v0, v6
-; RV64-NEXT:    csrr a0, vlenb
-; RV64-NEXT:    slli a0, a0, 4
-; RV64-NEXT:    add sp, sp, a0
-; RV64-NEXT:    addi sp, sp, 16
-; RV64-NEXT:    ret
+; CHECK-LABEL: icmp_eq_vv_v64i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    slli a1, a1, 4
+; CHECK-NEXT:    sub sp, sp, a1
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
+; CHECK-NEXT:    addi a1, a0, 128
+; CHECK-NEXT:    li a3, 32
+; CHECK-NEXT:    vsetvli zero, a3, e32, m8, ta, ma
+; CHECK-NEXT:    vle32.v v24, (a1)
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    slli a1, a1, 3
+; CHECK-NEXT:    add a1, sp, a1
+; CHECK-NEXT:    addi a1, a1, 16
+; CHECK-NEXT:    vs8r.v v24, (a1) # Unknown-size Folded Spill
+; CHECK-NEXT:    vle32.v v24, (a0)
+; CHECK-NEXT:    addi a0, sp, 16
+; CHECK-NEXT:    vs8r.v v24, (a0) # Unknown-size Folded Spill
+; CHECK-NEXT:    vsetivli zero, 4, e8, mf2, ta, ma
+; CHECK-NEXT:    vslidedown.vi v7, v0, 4
+; CHECK-NEXT:    mv a0, a2
+; CHECK-NEXT:    bltu a2, a3, .LBB99_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:  .LBB99_2:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
+; CHECK-NEXT:    addi a0, sp, 16
+; CHECK-NEXT:    vl8r.v v24, (a0) # Unknown-size Folded Reload
+; CHECK-NEXT:    vmseq.vv v6, v8, v24, v0.t
+; CHECK-NEXT:    addi a0, a2, -32
+; CHECK-NEXT:    sltu a1, a2, a0
+; CHECK-NEXT:    addi a1, a1, -1
+; CHECK-NEXT:    and a0, a1, a0
+; CHECK-NEXT:    vsetvli zero, a0, e32, m8, ta, ma
+; CHECK-NEXT:    vmv1r.v v0, v7
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    slli a0, a0, 3
+; CHECK-NEXT:    add a0, sp, a0
+; CHECK-NEXT:    addi a0, a0, 16
+; CHECK-NEXT:    vl8r.v v8, (a0) # Unknown-size Folded Reload
+; CHECK-NEXT:    vmseq.vv v24, v16, v8, v0.t
+; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; CHECK-NEXT:    vslideup.vi v6, v24, 4
+; CHECK-NEXT:    vmv1r.v v0, v6
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    slli a0, a0, 4
+; CHECK-NEXT:    add sp, sp, a0
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    ret
   %v = call <64 x i1> @llvm.vp.icmp.v64i32(<64 x i32> %va, <64 x i32> %vb, metadata !"eq", <64 x i1> %m, i32 %evl)
   ret <64 x i1> %v
 }
diff --git a/llvm/test/CodeGen/RISCV/rvv/vpscatter-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/vpscatter-sdnode.ll
index 38e4aab4deb34..351fc500145ea 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vpscatter-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vpscatter-sdnode.ll
@@ -2219,8 +2219,8 @@ define void @vpscatter_baseidx_nxv16i16_nxv16f64(<vscale x 16 x double> %val, pt
 ; RV64-NEXT:    vsll.vi v16, v16, 3
 ; RV64-NEXT:    addi a1, sp, 16
 ; RV64-NEXT:    vs8r.v v16, (a1) # Unknown-size Folded Spill
-; RV64-NEXT:    csrr a1, vlenb
 ; RV64-NEXT:    vsext.vf4 v16, v24
+; RV64-NEXT:    csrr a1, vlenb
 ; RV64-NEXT:    vsll.vi v24, v16, 3
 ; RV64-NEXT:    mv a3, a2
 ; RV64-NEXT:    bltu a2, a1, .LBB100_2
@@ -2286,29 +2286,31 @@ define void @vpscatter_baseidx_sext_nxv16i16_nxv16f64(<vscale x 16 x double> %va
 ; RV64-NEXT:    addi sp, sp, -16
 ; RV64-NEXT:    .cfi_def_cfa_offset 16
 ; RV64-NEXT:    csrr a3, vlenb
-; RV64-NEXT:    slli a3, a3, 4
+; RV64-NEXT:    li a4, 10
+; RV64-NEXT:    mul a3, a3, a4
 ; RV64-NEXT:    sub sp, sp, a3
-; RV64-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 16 * vlenb
+; RV64-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x0a, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 10 * vlenb
 ; RV64-NEXT:    vl4re16.v v24, (a1)
+; RV64-NEXT:    addi a1, sp, 16
+; RV64-NEXT:    vs1r.v v0, (a1) # Unknown-size Folded Spill
 ; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 3
 ; RV64-NEXT:    add a1, sp, a1
 ; RV64-NEXT:    addi a1, a1, 16
 ; RV64-NEXT:    vs8r.v v16, (a1) # Unknown-size Folded Spill
 ; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, ma
+; RV64-NEXT:    vsext.vf4 v0, v24
 ; RV64-NEXT:    vsext.vf4 v16, v26
 ; RV64-NEXT:    vsll.vi v16, v16, 3
-; RV64-NEXT:    addi a1, sp, 16
-; RV64-NEXT:    vs8r.v v16, (a1) # Unknown-size Folded Spill
 ; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    vsext.vf4 v16, v24
-; RV64-NEXT:    vsll.vi v24, v16, 3
+; RV64-NEXT:    vsll.vi v24, v0, 3
 ; RV64-NEXT:    mv a3, a2
 ; RV64-NEXT:    bltu a2, a1, .LBB101_2
 ; RV64-NEXT:  # %bb.1:
 ; RV64-NEXT:    mv a3, a1
 ; RV64-NEXT:  .LBB101_2:
 ; RV64-NEXT:    vsetvli zero, a3, e64, m8, ta, ma
+; RV64-NEXT:    addi a3, sp, 16
+; RV64-NEXT:    vl1r.v v0, (a3) # Unknown-size Folded Reload
 ; RV64-NEXT:    vsoxei64.v v8, (a0), v24, v0.t
 ; RV64-NEXT:    sub a3, a2, a1
 ; RV64-NEXT:    sltu a2, a2, a3
@@ -2319,15 +2321,13 @@ define void @vpscatter_baseidx_sext_nxv16i16_nxv16f64(<vscale x 16 x double> %va
 ; RV64-NEXT:    vslidedown.vx v0, v0, a1
 ; RV64-NEXT:    vsetvli zero, a2, e64, m8, ta, ma
 ; RV64-NEXT:    csrr a1, vlenb
-; RV64-NEXT:    slli a1, a1, 3
 ; RV64-NEXT:    add a1, sp, a1
 ; RV64-NEXT:    addi a1, a1, 16
 ; RV64-NEXT:    vl8r.v v8, (a1) # Unknown-size Folded Reload
-; RV64-NEXT:    addi a1, sp, 16
-; RV64-NEXT:    vl8r.v v16, (a1) # Unknown-size Folded Reload
 ; RV64-NEXT:    vsoxei64.v v8, (a0), v16, v0.t
 ; RV64-NEXT:    csrr a0, vlenb
-; RV64-NEXT:    slli a0, a0, 4
+; RV64-NEXT:    li a1, 10
+; RV64-NEXT:    mul a0, a0, a1
 ; RV64-NEXT:    add sp, sp, a0
 ; RV64-NEXT:    addi sp, sp, 16
 ; RV64-NEXT:    ret

From 2cc0c2104909558680409f8a8f39755936305e72 Mon Sep 17 00:00:00 2001
From: Christian Sigg <csigg@google.com>
Date: Mon, 15 Apr 2024 07:39:02 +0200
Subject: [PATCH 386/886] [clang][bazel] Fix BUILD after
 f811d7b50957b801788d7b171ddeb25b1fda415a.

---
 utils/bazel/llvm-project-overlay/clang/BUILD.bazel | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
index 2c3c39e5b53df..c2f77e3abca0e 100644
--- a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
@@ -1928,6 +1928,7 @@ cc_library(
         ":frontend",
         ":lex",
         ":sema",
+        ":serialization",
         ":type_nodes_gen",
         "//llvm:AllTargetsAsmParsers",
         "//llvm:Analysis",

From 46131aaf616c5cd97df0ec376a7e6ba475e1913c Mon Sep 17 00:00:00 2001
From: Mikhail Goncharov <goncharov.mikhail@gmail.com>
Date: Mon, 15 Apr 2024 08:41:32 +0200
Subject: [PATCH 387/886] Revert "Reapply "[Clang][Sema] Fix crash when 'this'
 is used in a dependent class scope function template specialization that
 instantiates to a static member function (#87541)" (#88311)"

This reverts commit aa80f3ec48419a73aafcc2ff947c6dd1e3734481.

See
https://github.com/llvm/llvm-project/pull/88311#issuecomment-2052291140.

There is a fix forward proposed but I am reverting this commit to fix
trunk.
---
 clang/docs/ReleaseNotes.rst                   |   2 -
 clang/include/clang/Sema/Sema.h               |  16 +--
 clang/lib/Sema/SemaExpr.cpp                   |  28 ++++-
 clang/lib/Sema/SemaExprCXX.cpp                |  61 +++++-----
 clang/lib/Sema/SemaExprMember.cpp             |  63 +++--------
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |   8 --
 clang/lib/Sema/TreeTransform.h                |  65 +++++------
 .../SemaTemplate/instantiate-using-decl.cpp   |   2 +-
 ...ms-function-specialization-class-scope.cpp | 106 +-----------------
 9 files changed, 99 insertions(+), 252 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index ade8f4e93d5a0..de5fa99bff5db 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -529,8 +529,6 @@ Bug Fixes to C++ Support
 - Fix an issue caused by not handling invalid cases when substituting into the parameter mapping of a constraint. Fixes (#GH86757).
 - Fixed a bug that prevented member function templates of class templates declared with a deduced return type
   from being explicitly specialized for a given implicit instantiation of the class template.
-- Fixed a crash when ``this`` is used in a dependent class scope function template specialization
-  that instantiates to a static member function.
 
 - Fix crash when inheriting from a cv-qualified type. Fixes:
   (`#35603 <https://github.com/llvm/llvm-project/issues/35603>`_)
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index c6e0332c3176b..c6035445e9019 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -5452,8 +5452,7 @@ class Sema final : public SemaBase {
 
   ExprResult BuildDeclarationNameExpr(const CXXScopeSpec &SS, LookupResult &R,
                                       bool NeedsADL,
-                                      bool AcceptInvalidDecl = false,
-                                      bool NeedUnresolved = false);
+                                      bool AcceptInvalidDecl = false);
   ExprResult BuildDeclarationNameExpr(
       const CXXScopeSpec &SS, const DeclarationNameInfo &NameInfo, NamedDecl *D,
       NamedDecl *FoundD = nullptr,
@@ -6596,10 +6595,7 @@ class Sema final : public SemaBase {
                             SourceLocation RParenLoc);
 
   //// ActOnCXXThis -  Parse 'this' pointer.
-  ExprResult ActOnCXXThis(SourceLocation Loc);
-
-  /// Check whether the type of 'this' is valid in the current context.
-  bool CheckCXXThisType(SourceLocation Loc, QualType Type);
+  ExprResult ActOnCXXThis(SourceLocation loc);
 
   /// Build a CXXThisExpr and mark it referenced in the current context.
   Expr *BuildCXXThisExpr(SourceLocation Loc, QualType Type, bool IsImplicit);
@@ -7022,14 +7018,10 @@ class Sema final : public SemaBase {
   ///@{
 
 public:
-  /// Check whether an expression might be an implicit class member access.
-  bool isPotentialImplicitMemberAccess(const CXXScopeSpec &SS, LookupResult &R,
-                                       bool IsAddressOfOperand);
-
   ExprResult BuildPossibleImplicitMemberExpr(
       const CXXScopeSpec &SS, SourceLocation TemplateKWLoc, LookupResult &R,
-      const TemplateArgumentListInfo *TemplateArgs, const Scope *S);
-
+      const TemplateArgumentListInfo *TemplateArgs, const Scope *S,
+      UnresolvedLookupExpr *AsULE = nullptr);
   ExprResult
   BuildImplicitMemberExpr(const CXXScopeSpec &SS, SourceLocation TemplateKWLoc,
                           LookupResult &R,
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 505d068ac42eb..c5395fb2068a1 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -2917,9 +2917,26 @@ Sema::ActOnIdExpression(Scope *S, CXXScopeSpec &SS,
   // to get this right here so that we don't end up making a
   // spuriously dependent expression if we're inside a dependent
   // instance method.
-  if (isPotentialImplicitMemberAccess(SS, R, IsAddressOfOperand))
-    return BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs,
-                                           S);
+  if (getLangOpts().CPlusPlus && !R.empty() &&
+      (*R.begin())->isCXXClassMember()) {
+    bool MightBeImplicitMember;
+    if (!IsAddressOfOperand)
+      MightBeImplicitMember = true;
+    else if (!SS.isEmpty())
+      MightBeImplicitMember = false;
+    else if (R.isOverloadedResult())
+      MightBeImplicitMember = false;
+    else if (R.isUnresolvableResult())
+      MightBeImplicitMember = true;
+    else
+      MightBeImplicitMember = isa<FieldDecl>(R.getFoundDecl()) ||
+                              isa<IndirectFieldDecl>(R.getFoundDecl()) ||
+                              isa<MSPropertyDecl>(R.getFoundDecl());
+
+    if (MightBeImplicitMember)
+      return BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc,
+                                             R, TemplateArgs, S);
+  }
 
   if (TemplateArgs || TemplateKWLoc.isValid()) {
 
@@ -3430,11 +3447,10 @@ static bool ShouldLookupResultBeMultiVersionOverload(const LookupResult &R) {
 
 ExprResult Sema::BuildDeclarationNameExpr(const CXXScopeSpec &SS,
                                           LookupResult &R, bool NeedsADL,
-                                          bool AcceptInvalidDecl,
-                                          bool NeedUnresolved) {
+                                          bool AcceptInvalidDecl) {
   // If this is a single, fully-resolved result and we don't need ADL,
   // just build an ordinary singleton decl ref.
-  if (!NeedUnresolved && !NeedsADL && R.isSingleResult() &&
+  if (!NeedsADL && R.isSingleResult() &&
       !R.getAsSingle<FunctionTemplateDecl>() &&
       !ShouldLookupResultBeMultiVersionOverload(R))
     return BuildDeclarationNameExpr(SS, R.getLookupNameInfo(), R.getFoundDecl(),
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 25f23a3abf171..fa7b92a43346b 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -1416,42 +1416,26 @@ bool Sema::CheckCXXThisCapture(SourceLocation Loc, const bool Explicit,
 }
 
 ExprResult Sema::ActOnCXXThis(SourceLocation Loc) {
-  // C++20 [expr.prim.this]p1:
-  //   The keyword this names a pointer to the object for which an
-  //   implicit object member function is invoked or a non-static
-  //   data member's initializer is evaluated.
+  /// C++ 9.3.2: In the body of a non-static member function, the keyword this
+  /// is a non-lvalue expression whose value is the address of the object for
+  /// which the function is called.
   QualType ThisTy = getCurrentThisType();
 
-  if (CheckCXXThisType(Loc, ThisTy))
-    return ExprError();
+  if (ThisTy.isNull()) {
+    DeclContext *DC = getFunctionLevelDeclContext();
 
-  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
-}
+    if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
+        Method && Method->isExplicitObjectMemberFunction()) {
+      return Diag(Loc, diag::err_invalid_this_use) << 1;
+    }
 
-bool Sema::CheckCXXThisType(SourceLocation Loc, QualType Type) {
-  if (!Type.isNull())
-    return false;
+    if (isLambdaCallWithExplicitObjectParameter(CurContext))
+      return Diag(Loc, diag::err_invalid_this_use) << 1;
 
-  // C++20 [expr.prim.this]p3:
-  //   If a declaration declares a member function or member function template
-  //   of a class X, the expression this is a prvalue of type
-  //   "pointer to cv-qualifier-seq X" wherever X is the current class between
-  //   the optional cv-qualifier-seq and the end of the function-definition,
-  //   member-declarator, or declarator. It shall not appear within the
-  //   declaration of either a static member function or an explicit object
-  //   member function of the current class (although its type and value
-  //   category are defined within such member functions as they are within
-  //   an implicit object member function).
-  DeclContext *DC = getFunctionLevelDeclContext();
-  if (const auto *Method = dyn_cast<CXXMethodDecl>(DC);
-      Method && Method->isExplicitObjectMemberFunction()) {
-    Diag(Loc, diag::err_invalid_this_use) << 1;
-  } else if (isLambdaCallWithExplicitObjectParameter(CurContext)) {
-    Diag(Loc, diag::err_invalid_this_use) << 1;
-  } else {
-    Diag(Loc, diag::err_invalid_this_use) << 0;
+    return Diag(Loc, diag::err_invalid_this_use) << 0;
   }
-  return true;
+
+  return BuildCXXThisExpr(Loc, ThisTy, /*IsImplicit=*/false);
 }
 
 Expr *Sema::BuildCXXThisExpr(SourceLocation Loc, QualType Type,
@@ -8658,8 +8642,21 @@ static ExprResult attemptRecovery(Sema &SemaRef,
 
       // Detect and handle the case where the decl might be an implicit
       // member.
-      if (SemaRef.isPotentialImplicitMemberAccess(
-              NewSS, R, Consumer.isAddressOfOperand()))
+      bool MightBeImplicitMember;
+      if (!Consumer.isAddressOfOperand())
+        MightBeImplicitMember = true;
+      else if (!NewSS.isEmpty())
+        MightBeImplicitMember = false;
+      else if (R.isOverloadedResult())
+        MightBeImplicitMember = false;
+      else if (R.isUnresolvableResult())
+        MightBeImplicitMember = true;
+      else
+        MightBeImplicitMember = isa<FieldDecl>(ND) ||
+                                isa<IndirectFieldDecl>(ND) ||
+                                isa<MSPropertyDecl>(ND);
+
+      if (MightBeImplicitMember)
         return SemaRef.BuildPossibleImplicitMemberExpr(
             NewSS, /*TemplateKWLoc*/ SourceLocation(), R,
             /*TemplateArgs*/ nullptr, /*S*/ nullptr);
diff --git a/clang/lib/Sema/SemaExprMember.cpp b/clang/lib/Sema/SemaExprMember.cpp
index eeac753a34897..32998ae60eafe 100644
--- a/clang/lib/Sema/SemaExprMember.cpp
+++ b/clang/lib/Sema/SemaExprMember.cpp
@@ -61,10 +61,6 @@ enum IMAKind {
   /// The reference is a contextually-permitted abstract member reference.
   IMA_Abstract,
 
-  /// Whether the context is static is dependent on the enclosing template (i.e.
-  /// in a dependent class scope explicit specialization).
-  IMA_Dependent,
-
   /// The reference may be to an unresolved using declaration and the
   /// context is not an instance method.
   IMA_Unresolved_StaticOrExplicitContext,
@@ -95,18 +91,10 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
 
   DeclContext *DC = SemaRef.getFunctionLevelDeclContext();
 
-  bool couldInstantiateToStatic = false;
-  bool isStaticOrExplicitContext = SemaRef.CXXThisTypeOverride.isNull();
-
-  if (auto *MD = dyn_cast<CXXMethodDecl>(DC)) {
-    if (MD->isImplicitObjectMemberFunction()) {
-      isStaticOrExplicitContext = false;
-      // A dependent class scope function template explicit specialization
-      // that is neither declared 'static' nor with an explicit object
-      // parameter could instantiate to a static or non-static member function.
-      couldInstantiateToStatic = MD->getDependentSpecializationInfo();
-    }
-  }
+  bool isStaticOrExplicitContext =
+      SemaRef.CXXThisTypeOverride.isNull() &&
+      (!isa<CXXMethodDecl>(DC) || cast<CXXMethodDecl>(DC)->isStatic() ||
+       cast<CXXMethodDecl>(DC)->isExplicitObjectMemberFunction());
 
   if (R.isUnresolvableResult())
     return isStaticOrExplicitContext ? IMA_Unresolved_StaticOrExplicitContext
@@ -135,9 +123,6 @@ static IMAKind ClassifyImplicitMemberAccess(Sema &SemaRef,
   if (Classes.empty())
     return IMA_Static;
 
-  if (couldInstantiateToStatic)
-    return IMA_Dependent;
-
   // C++11 [expr.prim.general]p12:
   //   An id-expression that denotes a non-static data member or non-static
   //   member function of a class can only be used:
@@ -278,52 +263,32 @@ static void diagnoseInstanceReference(Sema &SemaRef,
   }
 }
 
-bool Sema::isPotentialImplicitMemberAccess(const CXXScopeSpec &SS,
-                                           LookupResult &R,
-                                           bool IsAddressOfOperand) {
-  if (!getLangOpts().CPlusPlus)
-    return false;
-  else if (R.empty() || !R.begin()->isCXXClassMember())
-    return false;
-  else if (!IsAddressOfOperand)
-    return true;
-  else if (!SS.isEmpty())
-    return false;
-  else if (R.isOverloadedResult())
-    return false;
-  else if (R.isUnresolvableResult())
-    return true;
-  else
-    return isa<FieldDecl, IndirectFieldDecl, MSPropertyDecl>(R.getFoundDecl());
-}
-
 /// Builds an expression which might be an implicit member expression.
 ExprResult Sema::BuildPossibleImplicitMemberExpr(
     const CXXScopeSpec &SS, SourceLocation TemplateKWLoc, LookupResult &R,
-    const TemplateArgumentListInfo *TemplateArgs, const Scope *S) {
-  switch (IMAKind Classification = ClassifyImplicitMemberAccess(*this, R)) {
+    const TemplateArgumentListInfo *TemplateArgs, const Scope *S,
+    UnresolvedLookupExpr *AsULE) {
+  switch (ClassifyImplicitMemberAccess(*this, R)) {
   case IMA_Instance:
+    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, true, S);
+
   case IMA_Mixed:
   case IMA_Mixed_Unrelated:
   case IMA_Unresolved:
-    return BuildImplicitMemberExpr(
-        SS, TemplateKWLoc, R, TemplateArgs,
-        /*IsKnownInstance=*/Classification == IMA_Instance, S);
+    return BuildImplicitMemberExpr(SS, TemplateKWLoc, R, TemplateArgs, false,
+                                   S);
+
   case IMA_Field_Uneval_Context:
     Diag(R.getNameLoc(), diag::warn_cxx98_compat_non_static_member_use)
       << R.getLookupNameInfo().getName();
     [[fallthrough]];
   case IMA_Static:
   case IMA_Abstract:
-  case IMA_Dependent:
   case IMA_Mixed_StaticOrExplicitContext:
   case IMA_Unresolved_StaticOrExplicitContext:
     if (TemplateArgs || TemplateKWLoc.isValid())
-      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, /*RequiresADL=*/false,
-                                 TemplateArgs);
-    return BuildDeclarationNameExpr(
-        SS, R, /*NeedsADL=*/false, /*AcceptInvalidDecl=*/false,
-        /*NeedUnresolved=*/Classification == IMA_Dependent);
+      return BuildTemplateIdExpr(SS, TemplateKWLoc, R, false, TemplateArgs);
+    return AsULE ? AsULE : BuildDeclarationNameExpr(SS, R, false);
 
   case IMA_Error_StaticOrExplicitContext:
   case IMA_Error_Unrelated:
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 21a139e434cb7..02d0b35ce940f 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5097,14 +5097,6 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
   EnterExpressionEvaluationContext EvalContext(
       *this, Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 
-  Qualifiers ThisTypeQuals;
-  CXXRecordDecl *ThisContext = nullptr;
-  if (CXXMethodDecl *Method = dyn_cast<CXXMethodDecl>(Function)) {
-    ThisContext = Method->getParent();
-    ThisTypeQuals = Method->getMethodQualifiers();
-  }
-  CXXThisScopeRAII ThisScope(*this, ThisContext, ThisTypeQuals);
-
   // Introduce a new scope where local variable instantiations will be
   // recorded, unless we're actually a member function within a local
   // class, in which case we need to merge our results with the parent
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index ee6bd228b459e..fdf84c512c493 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -795,9 +795,6 @@ class TreeTransform {
       ParenExpr *PE, DependentScopeDeclRefExpr *DRE, bool IsAddressOfOperand,
       TypeSourceInfo **RecoveryTSI);
 
-  ExprResult TransformUnresolvedLookupExpr(UnresolvedLookupExpr *E,
-                                           bool IsAddressOfOperand);
-
   StmtResult TransformOMPExecutableDirective(OMPExecutableDirective *S);
 
 // FIXME: We use LLVM_ATTRIBUTE_NOINLINE because inlining causes a ridiculous
@@ -3312,13 +3309,12 @@ class TreeTransform {
 
   /// Build a new C++ "this" expression.
   ///
-  /// By default, performs semantic analysis to build a new "this" expression.
-  /// Subclasses may override this routine to provide different behavior.
+  /// By default, builds a new "this" expression without performing any
+  /// semantic analysis. Subclasses may override this routine to provide
+  /// different behavior.
   ExprResult RebuildCXXThisExpr(SourceLocation ThisLoc,
                                 QualType ThisType,
                                 bool isImplicit) {
-    if (getSema().CheckCXXThisType(ThisLoc, ThisType))
-      return ExprError();
     return getSema().BuildCXXThisExpr(ThisLoc, ThisType, isImplicit);
   }
 
@@ -11369,11 +11365,7 @@ template<typename Derived>
 ExprResult
 TreeTransform<Derived>::TransformAddressOfOperand(Expr *E) {
   if (DependentScopeDeclRefExpr *DRE = dyn_cast<DependentScopeDeclRefExpr>(E))
-    return getDerived().TransformDependentScopeDeclRefExpr(
-        DRE, /*IsAddressOfOperand=*/true, nullptr);
-  else if (UnresolvedLookupExpr *ULE = dyn_cast<UnresolvedLookupExpr>(E))
-    return getDerived().TransformUnresolvedLookupExpr(
-        ULE, /*IsAddressOfOperand=*/true);
+    return getDerived().TransformDependentScopeDeclRefExpr(DRE, true, nullptr);
   else
     return getDerived().TransformExpr(E);
 }
@@ -13079,16 +13071,10 @@ bool TreeTransform<Derived>::TransformOverloadExprDecls(OverloadExpr *Old,
   return false;
 }
 
-template <typename Derived>
-ExprResult TreeTransform<Derived>::TransformUnresolvedLookupExpr(
-    UnresolvedLookupExpr *Old) {
-  return TransformUnresolvedLookupExpr(Old, /*IsAddressOfOperand=*/false);
-}
-
-template <typename Derived>
+template<typename Derived>
 ExprResult
-TreeTransform<Derived>::TransformUnresolvedLookupExpr(UnresolvedLookupExpr *Old,
-                                                      bool IsAddressOfOperand) {
+TreeTransform<Derived>::TransformUnresolvedLookupExpr(
+                                                  UnresolvedLookupExpr *Old) {
   LookupResult R(SemaRef, Old->getName(), Old->getNameLoc(),
                  Sema::LookupOrdinaryName);
 
@@ -13120,8 +13106,26 @@ TreeTransform<Derived>::TransformUnresolvedLookupExpr(UnresolvedLookupExpr *Old,
     R.setNamingClass(NamingClass);
   }
 
-  // Rebuild the template arguments, if any.
   SourceLocation TemplateKWLoc = Old->getTemplateKeywordLoc();
+
+  // If we have neither explicit template arguments, nor the template keyword,
+  // it's a normal declaration name or member reference.
+  if (!Old->hasExplicitTemplateArgs() && !TemplateKWLoc.isValid()) {
+    NamedDecl *D = R.getAsSingle<NamedDecl>();
+    // In a C++11 unevaluated context, an UnresolvedLookupExpr might refer to an
+    // instance member. In other contexts, BuildPossibleImplicitMemberExpr will
+    // give a good diagnostic.
+    if (D && D->isCXXInstanceMember()) {
+      return SemaRef.BuildPossibleImplicitMemberExpr(SS, TemplateKWLoc, R,
+                                                     /*TemplateArgs=*/nullptr,
+                                                     /*Scope=*/nullptr);
+    }
+
+    return getDerived().RebuildDeclarationNameExpr(SS, R, Old->requiresADL());
+  }
+
+  // If we have template arguments, rebuild them, then rebuild the
+  // templateid expression.
   TemplateArgumentListInfo TransArgs(Old->getLAngleLoc(), Old->getRAngleLoc());
   if (Old->hasExplicitTemplateArgs() &&
       getDerived().TransformTemplateArguments(Old->getTemplateArgs(),
@@ -13131,23 +13135,6 @@ TreeTransform<Derived>::TransformUnresolvedLookupExpr(UnresolvedLookupExpr *Old,
     return ExprError();
   }
 
-  // An UnresolvedLookupExpr can refer to a class member. This occurs e.g. when
-  // a non-static data member is named in an unevaluated operand, or when
-  // a member is named in a dependent class scope function template explicit
-  // specialization that is neither declared static nor with an explicit object
-  // parameter.
-  if (SemaRef.isPotentialImplicitMemberAccess(SS, R, IsAddressOfOperand))
-    return SemaRef.BuildPossibleImplicitMemberExpr(
-        SS, TemplateKWLoc, R,
-        Old->hasExplicitTemplateArgs() ? &TransArgs : nullptr,
-        /*S=*/nullptr);
-
-  // If we have neither explicit template arguments, nor the template keyword,
-  // it's a normal declaration name or member reference.
-  if (!Old->hasExplicitTemplateArgs() && !TemplateKWLoc.isValid())
-    return getDerived().RebuildDeclarationNameExpr(SS, R, Old->requiresADL());
-
-  // If we have template arguments, then rebuild the template-id expression.
   return getDerived().RebuildTemplateIdExpr(SS, TemplateKWLoc, R,
                                             Old->requiresADL(), &TransArgs);
 }
diff --git a/clang/test/SemaTemplate/instantiate-using-decl.cpp b/clang/test/SemaTemplate/instantiate-using-decl.cpp
index 28d8376438513..0bbb3ca9c88c8 100644
--- a/clang/test/SemaTemplate/instantiate-using-decl.cpp
+++ b/clang/test/SemaTemplate/instantiate-using-decl.cpp
@@ -121,7 +121,7 @@ template <typename Scalar> struct Derived : Base<Scalar> {
     (void)&field;
     // expected-error@+1 {{call to non-static member function without an object argument}}
     (void)method;
-    // expected-error@+1 {{must explicitly qualify name of member function when taking its address}}
+    // expected-error@+1 {{call to non-static member function without an object argument}}
     (void)&method;
     // expected-error@+1 {{call to non-static member function without an object argument}}
     method();
diff --git a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
index f7b59f2ac7158..dcab9bfaeabcb 100644
--- a/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
+++ b/clang/test/SemaTemplate/ms-function-specialization-class-scope.cpp
@@ -1,6 +1,7 @@
-// RUN: %clang_cc1 -fms-extensions -fsyntax-only -Wno-unused-value -verify %s
-// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -Wno-unused-value -verify %s
+// RUN: %clang_cc1 -fms-extensions -fsyntax-only -verify %s
+// RUN: %clang_cc1 -fms-extensions -fdelayed-template-parsing -fsyntax-only -verify %s
 
+// expected-no-diagnostics
 class A {
 public:
   template<class U> A(U p) {}
@@ -75,104 +76,3 @@ struct S {
   int f<0>(int);
 };
 }
-
-namespace UsesThis {
-  template<typename T>
-  struct A {
-    int x;
-
-    static inline int y;
-
-    template<typename U = void>
-    static void f();
-
-    template<typename U = void>
-    void g();
-
-    template<typename U>
-    static auto h() -> A*;
-
-    void i();
-
-    static void j();
-
-    template<>
-    void f<int>() {
-      this->x; // expected-error {{invalid use of 'this' outside of a non-static member function}}
-      x; // expected-error {{invalid use of member 'x' in static member function}}
-      A::x; // expected-error {{invalid use of member 'x' in static member function}}
-      +x; // expected-error {{invalid use of member 'x' in static member function}}
-      +A::x; // expected-error {{invalid use of member 'x' in static member function}}
-      &x; // expected-error {{invalid use of member 'x' in static member function}}
-      &A::x;
-      this->y; // expected-error {{invalid use of 'this' outside of a non-static member function}}
-      y;
-      A::y;
-      +y;
-      +A::y;
-      &y;
-      &A::y;
-      f();
-      f<void>();
-      g(); // expected-error {{call to non-static member function without an object argument}}
-      g<void>(); // expected-error {{call to non-static member function without an object argument}}
-      i(); // expected-error {{call to non-static member function without an object argument}}
-      j();
-      &i; // expected-error 2{{must explicitly qualify name of member function when taking its address}}
-      &j;
-      &A::i;
-      &A::j;
-    }
-
-    template<>
-    void g<int>() {
-      this->x;
-      x;
-      A::x;
-      +x;
-      +A::x;
-      &x;
-      &A::x;
-      this->y;
-      y;
-      A::y;
-      +y;
-      +A::y;
-      &y;
-      &A::y;
-      f();
-      f<void>();
-      g();
-      g<void>();
-      i();
-      j();
-      &i; // expected-error 2{{must explicitly qualify name of member function when taking its address}}
-      &j;
-      &A::i;
-      &A::j;
-    }
-
-    template<>
-    auto h<int>() -> decltype(this); // expected-error {{'this' cannot be used in a static member function declaration}}
-  };
-
-  template struct A<int>; // expected-note 3{{in instantiation of}}
-
-  template <typename T>
-  struct Foo {
-    template <typename X>
-    int bar(X x) {
-      return 0;
-    }
-
-    template <>
-    int bar(int x) {
-      return bar(5.0); // ok
-    }
-  };
-
-  void call() {
-    Foo<double> f;
-    f.bar(1);
-  }
-}

From a71565d75e23fc28076aa5bf1c5cf4432623afc5 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Mon, 15 Apr 2024 15:42:38 +0900
Subject: [PATCH 388/886] [ARM] Don't include IRBuilder.h in ARMISelLowering.h
 (NFC)

Just a forward declaration is sufficient.
---
 llvm/lib/Target/ARM/ARMISelLowering.h  | 2 +-
 llvm/lib/Target/ARM/ARMParallelDSP.cpp | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/ARM/ARMISelLowering.h b/llvm/lib/Target/ARM/ARMISelLowering.h
index b13ddf697cb80..26ef295e3d3fc 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.h
+++ b/llvm/lib/Target/ARM/ARMISelLowering.h
@@ -27,7 +27,6 @@
 #include "llvm/IR/Attributes.h"
 #include "llvm/IR/CallingConv.h"
 #include "llvm/IR/Function.h"
-#include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/InlineAsm.h"
 #include "llvm/Support/CodeGen.h"
 #include <optional>
@@ -42,6 +41,7 @@ class FunctionLoweringInfo;
 class GlobalValue;
 class InstrItineraryData;
 class Instruction;
+class IRBuilderBase;
 class MachineBasicBlock;
 class MachineInstr;
 class SelectionDAG;
diff --git a/llvm/lib/Target/ARM/ARMParallelDSP.cpp b/llvm/lib/Target/ARM/ARMParallelDSP.cpp
index 1efda5d1c9375..8c1b3325f155f 100644
--- a/llvm/lib/Target/ARM/ARMParallelDSP.cpp
+++ b/llvm/lib/Target/ARM/ARMParallelDSP.cpp
@@ -26,6 +26,7 @@
 #include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/IntrinsicsARM.h"
+#include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/NoFolder.h"
 #include "llvm/IR/PatternMatch.h"
 #include "llvm/Pass.h"

From 302d0f3476c0fc68d796e7b8a0d4237121ec33fe Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Mon, 15 Apr 2024 08:33:14 +0100
Subject: [PATCH 389/886] [VectorCombine] Add some tests for converting
 shuffles to identities. NFC

---
 .../AArch64/shuffletoidentity.ll              | 647 ++++++++++++++++++
 1 file changed, 647 insertions(+)
 create mode 100644 llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll

diff --git a/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
new file mode 100644
index 0000000000000..d96dfec849167
--- /dev/null
+++ b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
@@ -0,0 +1,647 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -passes='vector-combine' -S %s | FileCheck %s
+
+target triple = "aarch64"
+
+define <8 x i8> @trivial(<8 x i8> %a) {
+; CHECK-LABEL: @trivial(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[AT]], <4 x i8> [[AB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %r = shufflevector <4 x i8> %at, <4 x i8> %ab, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @add(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @add(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, %bt
+  %abb = add <4 x i8> %ab, %bb
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+; Different opcodes
+define <8 x i8> @wrong_addsub(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @wrong_addsub(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = sub <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, %bt
+  %abb = sub <4 x i8> %ab, %bb
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+; Different lanes that do not make an identity
+define <8 x i8> @wrong_lanes(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @wrong_lanes(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 6, i32 7, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, %bt
+  %abb = add <4 x i8> %ab, %bb
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 6, i32 7, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x half> @fadd(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @fadd(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x half> [[B]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = fadd <4 x half> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = fadd <4 x half> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = fadd <4 x half> %at, %bt
+  %abb = fadd <4 x half> %ab, %bb
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x half> @fneg(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @fneg(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = fneg <4 x half> [[AT]]
+; CHECK-NEXT:    [[ABB:%.*]] = fneg <4 x half> [[AB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = fneg <4 x half> %at
+  %abb = fneg <4 x half> %ab
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x i8> @abs(<8 x i8> %a) {
+; CHECK-LABEL: @abs(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = call <4 x i8> @llvm.abs.v4i8(<4 x i8> [[AT]], i1 true)
+; CHECK-NEXT:    [[ABB:%.*]] = call <4 x i8> @llvm.abs.v4i8(<4 x i8> [[AB]], i1 false)
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %at, i1 true)
+  %abb = call <4 x i8> @llvm.abs.v4i8(<4 x i8> %ab, i1 false)
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x half> @splat0(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @splat0(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BS:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[ABT:%.*]] = fadd <4 x half> [[AT]], [[BS]]
+; CHECK-NEXT:    [[ABB:%.*]] = fadd <4 x half> [[AB]], [[BS]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> zeroinitializer
+  %abt = fadd <4 x half> %at, %bs
+  %abb = fadd <4 x half> %ab, %bs
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x half> @splat2(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @splat2(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BS:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT:    [[ABT:%.*]] = fadd <4 x half> [[AT]], [[BS]]
+; CHECK-NEXT:    [[ABB:%.*]] = fadd <4 x half> [[AB]], [[BS]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+  %abt = fadd <4 x half> %at, %bs
+  %abb = fadd <4 x half> %ab, %bs
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x half> @splatandidentity(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @splatandidentity(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BS:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[ABT:%.*]] = fadd <4 x half> [[AT]], [[BS]]
+; CHECK-NEXT:    [[ABB:%.*]] = fadd <4 x half> [[AB]], [[BS]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bs = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> zeroinitializer
+  %abt = fadd <4 x half> %at, %bs
+  %abb = fadd <4 x half> %ab, %bs
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x half> @splattwice(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: @splattwice(
+; CHECK-NEXT:    [[AS:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[BS:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[AB1:%.*]] = fadd <4 x half> [[AS]], [[BS]]
+; CHECK-NEXT:    [[AB2:%.*]] = fadd <4 x half> [[AS]], [[BS]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[AB1]], <4 x half> [[AB2]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %as = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> zeroinitializer
+  %bs = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> zeroinitializer
+  %ab1 = fadd <4 x half> %as, %bs
+  %ab2 = fadd <4 x half> %as, %bs
+  %r = shufflevector <4 x half> %ab1, <4 x half> %ab2, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define <8 x i8> @undeflane(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @undeflane(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 poison, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, %bt
+  %abb = add <4 x i8> %ab, %bb
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 poison, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @constantsplat(<8 x i8> %a) {
+; CHECK-LABEL: @constantsplat(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], <i8 10, i8 10, i8 10, i8 10>
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], <i8 10, i8 10, i8 10, i8 10>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, <i8 10, i8 10, i8 10, i8 10>
+  %abb = add <4 x i8> %ab, <i8 10, i8 10, i8 10, i8 10>
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @constantdiff(<8 x i8> %a) {
+; CHECK-LABEL: @constantdiff(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], <i8 1, i8 2, i8 3, i8 4>
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], <i8 5, i8 6, i8 7, i8 8>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, <i8 1, i8 2, i8 3, i8 4>
+  %abb = add <4 x i8> %ab, <i8 5, i8 6, i8 7, i8 8>
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @constantdiff2(<8 x i8> %a) {
+; CHECK-LABEL: @constantdiff2(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], <i8 1, i8 2, i8 3, i8 4>
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], <i8 1, i8 2, i8 3, i8 4>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, <i8 1, i8 2, i8 3, i8 4>
+  %abb = add <4 x i8> %ab, <i8 1, i8 2, i8 3, i8 4>
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @inner_shuffle(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c) {
+; CHECK-LABEL: @inner_shuffle(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[CS:%.*]] = shufflevector <8 x i8> [[C:%.*]], <8 x i8> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[ABT:%.*]] = mul <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = mul <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[ABT2:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[ABB2:%.*]] = shufflevector <4 x i8> [[ABB]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[ABT3:%.*]] = add <4 x i8> [[ABT2]], [[CS]]
+; CHECK-NEXT:    [[ABB3:%.*]] = add <4 x i8> [[ABB2]], [[CS]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %cs = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> zeroinitializer
+  %abt = mul <4 x i8> %at, %bt
+  %abb = mul <4 x i8> %ab, %bb
+  %abt2 = shufflevector <4 x i8> %abt, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %abb2 = shufflevector <4 x i8> %abb, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %abt3 = add <4 x i8> %abt2, %cs
+  %abb3 = add <4 x i8> %abb2, %cs
+  %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @extrause_add(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @extrause_add(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    call void @use(<4 x i8> [[ABB]])
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = add <4 x i8> %at, %bt
+  %abb = add <4 x i8> %ab, %bb
+  call void @use(<4 x i8> %abb)
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @extrause_shuffle(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @extrause_shuffle(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B1:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT1:%.*]] = shufflevector <8 x i8> [[B1]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    call void @use(<4 x i8> [[BT1]])
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[BT]], [[BT1]]
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  call void @use(<4 x i8> %bt)
+  %abt = add <4 x i8> %at, %bt
+  %abb = add <4 x i8> %ab, %bb
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @icmpsel(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d) {
+; CHECK-LABEL: @icmpsel(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[CB:%.*]] = shufflevector <8 x i8> [[C:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[CT:%.*]] = shufflevector <8 x i8> [[C]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[DB:%.*]] = shufflevector <8 x i8> [[D:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[DT:%.*]] = shufflevector <8 x i8> [[D]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT1:%.*]] = icmp slt <4 x i8> [[AT]], [[BT]]
+; CHECK-NEXT:    [[ABB1:%.*]] = icmp slt <4 x i8> [[AB]], [[BB]]
+; CHECK-NEXT:    [[ABT:%.*]] = select <4 x i1> [[ABT1]], <4 x i8> [[CT]], <4 x i8> [[DT]]
+; CHECK-NEXT:    [[ABB:%.*]] = select <4 x i1> [[ABB1]], <4 x i8> [[CB]], <4 x i8> [[DB]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %cb = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %ct = shufflevector <8 x i8> %c, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %db = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %dt = shufflevector <8 x i8> %d, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt1 = icmp slt <4 x i8> %at, %bt
+  %abb1 = icmp slt <4 x i8> %ab, %bb
+  %abt = select <4 x i1> %abt1, <4 x i8> %ct, <4 x i8> %dt
+  %abb = select <4 x i1> %abb1, <4 x i8> %cb, <4 x i8> %db
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x half> @fma(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
+; CHECK-LABEL: @fma(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x half> [[A:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x half> [[A]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x half> [[B:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x half> [[B]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[CB:%.*]] = shufflevector <8 x half> [[C:%.*]], <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[CT:%.*]] = shufflevector <8 x half> [[C]], <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABB:%.*]] = call <4 x half> @llvm.fma.v4f16(<4 x half> [[AB]], <4 x half> [[BB]], <4 x half> [[CB]])
+; CHECK-NEXT:    [[ABT:%.*]] = call <4 x half> @llvm.fma.v4f16(<4 x half> [[AT]], <4 x half> [[BT]], <4 x half> [[CT]])
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x half> [[ABT]], <4 x half> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x half> [[R]]
+;
+  %ab = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x half> %a, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x half> %b, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %cb = shufflevector <8 x half> %c, <8 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %ct = shufflevector <8 x half> %c, <8 x half> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abb = call <4 x half> @llvm.fma.v4f16(<4 x half> %ab, <4 x half> %bb, <4 x half> %cb)
+  %abt = call <4 x half> @llvm.fma.v4f16(<4 x half> %at, <4 x half> %bt, <4 x half> %ct)
+  %r = shufflevector <4 x half> %abt, <4 x half> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x half> %r
+}
+
+define void @exttrunc(<8 x i32> %a, <8 x i32> %b, ptr %p) {
+; CHECK-LABEL: @exttrunc(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i32> [[A:%.*]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i32> [[A]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i32> [[B:%.*]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i32> [[B]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[AB1:%.*]] = zext <4 x i32> [[AB]] to <4 x i64>
+; CHECK-NEXT:    [[AT1:%.*]] = zext <4 x i32> [[AT]] to <4 x i64>
+; CHECK-NEXT:    [[BB1:%.*]] = sext <4 x i32> [[BB]] to <4 x i64>
+; CHECK-NEXT:    [[BT1:%.*]] = sext <4 x i32> [[BT]] to <4 x i64>
+; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i64> [[AB1]], [[BB1]]
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i64> [[AT1]], [[BT1]]
+; CHECK-NEXT:    [[ABB1:%.*]] = trunc <4 x i64> [[ABB]] to <4 x i32>
+; CHECK-NEXT:    [[ABT1:%.*]] = trunc <4 x i64> [[ABT]] to <4 x i32>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[ABB1]], <4 x i32> [[ABT1]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:    store <8 x i32> [[TMP1]], ptr [[P:%.*]], align 32
+; CHECK-NEXT:    ret void
+;
+  %ab = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %at = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %bb = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %bt = shufflevector <8 x i32> %b, <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %ab1 = zext <4 x i32> %ab to <4 x i64>
+  %at1 = zext <4 x i32> %at to <4 x i64>
+  %bb1 = sext <4 x i32> %bb to <4 x i64>
+  %bt1 = sext <4 x i32> %bt to <4 x i64>
+  %abb = add <4 x i64> %ab1, %bb1
+  %abt = add <4 x i64> %at1, %bt1
+  %abb1 = trunc <4 x i64> %abb to <4 x i32>
+  %abt1 = trunc <4 x i64> %abt to <4 x i32>
+  %r = shufflevector <4 x i32> %abb1, <4 x i32> %abt1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+  store <8 x i32> %r, ptr %p
+  ret void
+}
+
+define <8 x i8> @intrinsics_minmax(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @intrinsics_minmax(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[AT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[AB]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT1:%.*]] = call <4 x i8> @llvm.smix.v4i8(<4 x i8> [[ABT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB1:%.*]] = call <4 x i8> @llvm.smax.v4i8(<4 x i8> [[ABB]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT2:%.*]] = call <4 x i8> @llvm.umin.v4i8(<4 x i8> [[ABT1]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB2:%.*]] = call <4 x i8> @llvm.umin.v4i8(<4 x i8> [[ABB1]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT3:%.*]] = call <4 x i8> @llvm.umax.v4i8(<4 x i8> [[ABT2]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB3:%.*]] = call <4 x i8> @llvm.umax.v4i8(<4 x i8> [[ABB2]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R1]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %at, <4 x i8> %bt)
+  %abb = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %ab, <4 x i8> %bb)
+  %abt1 = call <4 x i8> @llvm.smix.v4i8(<4 x i8> %abt, <4 x i8> %bt)
+  %abb1 = call <4 x i8> @llvm.smax.v4i8(<4 x i8> %abb, <4 x i8> %bb)
+  %abt2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abt1, <4 x i8> %bt)
+  %abb2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abb1, <4 x i8> %bb)
+  %abt3 = call <4 x i8> @llvm.umax.v4i8(<4 x i8> %abt2, <4 x i8> %bt)
+  %abb3 = call <4 x i8> @llvm.umax.v4i8(<4 x i8> %abb2, <4 x i8> %bb)
+  %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @intrinsics_addsat(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @intrinsics_addsat(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> [[AT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB:%.*]] = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> [[AB]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT1:%.*]] = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> [[ABT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB1:%.*]] = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> [[ABB]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT2:%.*]] = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> [[ABT1]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB2:%.*]] = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> [[ABB1]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[ABT3:%.*]] = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> [[ABT2]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB3:%.*]] = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> [[ABB2]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R1]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> %at, <4 x i8> %bt)
+  %abb = call <4 x i8> @llvm.sadd.sat.v4i8(<4 x i8> %ab, <4 x i8> %bb)
+  %abt1 = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> %abt, <4 x i8> %bt)
+  %abb1 = call <4 x i8> @llvm.ssub.sat.v4i8(<4 x i8> %abb, <4 x i8> %bb)
+  %abt2 = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %abt1, <4 x i8> %bt)
+  %abb2 = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %abb1, <4 x i8> %bb)
+  %abt3 = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %abt2, <4 x i8> %bt)
+  %abb3 = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %abb2, <4 x i8> %bb)
+  %r = shufflevector <4 x i8> %abt3, <4 x i8> %abb3, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define <8 x i8> @intrinsics_different(<8 x i8> %a, <8 x i8> %b) {
+; CHECK-LABEL: @intrinsics_different(
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[ABT:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[AT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABB:%.*]] = call <4 x i8> @llvm.umin.v4i8(<4 x i8> [[AB]], <4 x i8> [[BB]])
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
+;
+  %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %bb = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+  %abt = call <4 x i8> @llvm.smin(<4 x i8> %at, <4 x i8> %bt)
+  %abb = call <4 x i8> @llvm.umin(<4 x i8> %ab, <4 x i8> %bb)
+  %r = shufflevector <4 x i8> %abt, <4 x i8> %abb, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  ret <8 x i8> %r
+}
+
+define void @v8f64interleave(i64 %0, ptr %1, ptr %x, double %z) {
+; CHECK-LABEL: @v8f64interleave(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x double> poison, double [[Z:%.*]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x double> [[BROADCAST_SPLATINSERT]], <2 x double> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <16 x double>, ptr [[TMP1:%.*]], align 8
+; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 0, i32 8>
+; CHECK-NEXT:    [[STRIDED_VEC27:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 1, i32 9>
+; CHECK-NEXT:    [[STRIDED_VEC28:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 2, i32 10>
+; CHECK-NEXT:    [[STRIDED_VEC29:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 3, i32 11>
+; CHECK-NEXT:    [[STRIDED_VEC30:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 4, i32 12>
+; CHECK-NEXT:    [[STRIDED_VEC31:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 5, i32 13>
+; CHECK-NEXT:    [[STRIDED_VEC32:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 6, i32 14>
+; CHECK-NEXT:    [[STRIDED_VEC33:%.*]] = shufflevector <16 x double> [[WIDE_VEC]], <16 x double> poison, <2 x i32> <i32 7, i32 15>
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <2 x double> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[X:%.*]], i64 [[TMP0:%.*]]
+; CHECK-NEXT:    [[WIDE_VEC34:%.*]] = load <16 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[STRIDED_VEC35:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 0, i32 8>
+; CHECK-NEXT:    [[STRIDED_VEC36:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 1, i32 9>
+; CHECK-NEXT:    [[STRIDED_VEC37:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 2, i32 10>
+; CHECK-NEXT:    [[STRIDED_VEC38:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 3, i32 11>
+; CHECK-NEXT:    [[STRIDED_VEC39:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 4, i32 12>
+; CHECK-NEXT:    [[STRIDED_VEC40:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 5, i32 13>
+; CHECK-NEXT:    [[STRIDED_VEC41:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 6, i32 14>
+; CHECK-NEXT:    [[STRIDED_VEC42:%.*]] = shufflevector <16 x double> [[WIDE_VEC34]], <16 x double> poison, <2 x i32> <i32 7, i32 15>
+; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <2 x double> [[STRIDED_VEC35]], [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = fmul fast <2 x double> [[STRIDED_VEC27]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP6:%.*]] = fadd fast <2 x double> [[STRIDED_VEC36]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = fmul fast <2 x double> [[STRIDED_VEC28]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP8:%.*]] = fadd fast <2 x double> [[STRIDED_VEC37]], [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = fmul fast <2 x double> [[STRIDED_VEC29]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP10:%.*]] = fadd fast <2 x double> [[STRIDED_VEC38]], [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = fmul fast <2 x double> [[STRIDED_VEC30]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP12:%.*]] = fadd fast <2 x double> [[STRIDED_VEC39]], [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = fmul fast <2 x double> [[STRIDED_VEC31]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP14:%.*]] = fadd fast <2 x double> [[STRIDED_VEC40]], [[TMP13]]
+; CHECK-NEXT:    [[TMP15:%.*]] = fmul fast <2 x double> [[STRIDED_VEC32]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP16:%.*]] = fadd fast <2 x double> [[STRIDED_VEC41]], [[TMP15]]
+; CHECK-NEXT:    [[TMP17:%.*]] = or disjoint i64 [[TMP0]], 7
+; CHECK-NEXT:    [[TMP18:%.*]] = fmul fast <2 x double> [[STRIDED_VEC33]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[TMP17]]
+; CHECK-NEXT:    [[TMP20:%.*]] = fadd fast <2 x double> [[STRIDED_VEC42]], [[TMP18]]
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i64 -56
+; CHECK-NEXT:    [[TMP22:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP23:%.*]] = shufflevector <2 x double> [[TMP8]], <2 x double> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP24:%.*]] = shufflevector <2 x double> [[TMP12]], <2 x double> [[TMP14]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP25:%.*]] = shufflevector <2 x double> [[TMP16]], <2 x double> [[TMP20]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <4 x double> [[TMP22]], <4 x double> [[TMP23]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[TMP27:%.*]] = shufflevector <4 x double> [[TMP24]], <4 x double> [[TMP25]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x double> [[TMP26]], <8 x double> [[TMP27]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    store <16 x double> [[INTERLEAVED_VEC]], ptr [[TMP21]], align 8
+; CHECK-NEXT:    ret void
+;
+entry:
+  %broadcast.splatinsert = insertelement <2 x double> poison, double %z, i64 0
+  %broadcast.splat = shufflevector <2 x double> %broadcast.splatinsert, <2 x double> poison, <2 x i32> zeroinitializer
+  %wide.vec = load <16 x double>, ptr %1, align 8
+  %strided.vec = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 0, i32 8>
+  %strided.vec27 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 1, i32 9>
+  %strided.vec28 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 2, i32 10>
+  %strided.vec29 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 3, i32 11>
+  %strided.vec30 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 4, i32 12>
+  %strided.vec31 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 5, i32 13>
+  %strided.vec32 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 6, i32 14>
+  %strided.vec33 = shufflevector <16 x double> %wide.vec, <16 x double> poison, <2 x i32> <i32 7, i32 15>
+  %2 = fmul fast <2 x double> %strided.vec, %broadcast.splat
+  %3 = getelementptr inbounds double, ptr %x, i64 %0
+  %wide.vec34 = load <16 x double>, ptr %3, align 8
+  %strided.vec35 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 0, i32 8>
+  %strided.vec36 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 1, i32 9>
+  %strided.vec37 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 2, i32 10>
+  %strided.vec38 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 3, i32 11>
+  %strided.vec39 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 4, i32 12>
+  %strided.vec40 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 5, i32 13>
+  %strided.vec41 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 6, i32 14>
+  %strided.vec42 = shufflevector <16 x double> %wide.vec34, <16 x double> poison, <2 x i32> <i32 7, i32 15>
+  %4 = fadd fast <2 x double> %strided.vec35, %2
+  %5 = fmul fast <2 x double> %strided.vec27, %broadcast.splat
+  %6 = fadd fast <2 x double> %strided.vec36, %5
+  %7 = fmul fast <2 x double> %strided.vec28, %broadcast.splat
+  %8 = fadd fast <2 x double> %strided.vec37, %7
+  %9 = fmul fast <2 x double> %strided.vec29, %broadcast.splat
+  %10 = fadd fast <2 x double> %strided.vec38, %9
+  %11 = fmul fast <2 x double> %strided.vec30, %broadcast.splat
+  %12 = fadd fast <2 x double> %strided.vec39, %11
+  %13 = fmul fast <2 x double> %strided.vec31, %broadcast.splat
+  %14 = fadd fast <2 x double> %strided.vec40, %13
+  %15 = fmul fast <2 x double> %strided.vec32, %broadcast.splat
+  %16 = fadd fast <2 x double> %strided.vec41, %15
+  %17 = or disjoint i64 %0, 7
+  %18 = fmul fast <2 x double> %strided.vec33, %broadcast.splat
+  %19 = getelementptr inbounds double, ptr %x, i64 %17
+  %20 = fadd fast <2 x double> %strided.vec42, %18
+  %21 = getelementptr inbounds i8, ptr %19, i64 -56
+  %22 = shufflevector <2 x double> %4, <2 x double> %6, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %23 = shufflevector <2 x double> %8, <2 x double> %10, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %24 = shufflevector <2 x double> %12, <2 x double> %14, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %25 = shufflevector <2 x double> %16, <2 x double> %20, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %26 = shufflevector <4 x double> %22, <4 x double> %23, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %27 = shufflevector <4 x double> %24, <4 x double> %25, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  %interleaved.vec = shufflevector <8 x double> %26, <8 x double> %27, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  store <16 x double> %interleaved.vec, ptr %21, align 8
+  ret void
+}
+
+declare void @use(<4 x i8>)

From 07942987b58558a8bcb2cfb7d8bc3d46b1cb1d2b Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Mon, 15 Apr 2024 00:48:51 -0700
Subject: [PATCH 390/886] [llvm-objcopy] Add --compress-sections

--compress-sections is similar to --compress-debug-sections but applies
to arbitrary sections.

* `--compress-sections <section>=none`: decompress sections
* `--compress-sections <section>=[zlib|zstd]`: compress sections with zlib/zstd

Like `--remove-section`, the pattern is by default a glob, but a regex
when --regex is specified.

For `--remove-section` like options, `!` prevents matches and is not
dependent on ordering (see `ELF/wildcard-syntax.test`). Since
`--compress-sections a=zlib --compress-sections a=none` naturally allows
overriding, having an order-independent `!` would be confusing.
Therefore, `!` is disallowed.

Sections within a segment are effectively immutable. Report an error for
an attempt to (de)compress them. `SHF_ALLOC` sections in a relocatable
file can be compressed, but linkers usually reject them.

Note: Before this patch, a compressed relocation section is recognized
as a `RelocationSectionBase` as well and `removeSections` `!ToRemove(*ToRelSec)`
may incorrectly interpret a `CompressedSections` as `RelocationSectionBase`,
leading to ubsan failure for the new test. Fix this by setting
`OriginalFlags` in CompressedSection::CompressedSection.

Link: https://discourse.llvm.org/t/rfc-compress-arbitrary-sections-with-ld-lld-compress-sections/71674

Pull Request: https://github.com/llvm/llvm-project/pull/85036
---
 llvm/docs/CommandGuide/llvm-objcopy.rst       |   8 ++
 llvm/docs/ReleaseNotes.rst                    |   4 +
 llvm/include/llvm/ObjCopy/CommonConfig.h      |   3 +
 llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp           |  34 +++--
 llvm/lib/ObjCopy/ELF/ELFObject.cpp            |   5 +
 .../ELF/compress-sections-within-segment.s    |  38 ++++++
 .../llvm-objcopy/ELF/compress-sections.s      | 128 ++++++++++++++++++
 .../llvm-objcopy/ELF/decompress-sections.test |  29 ++++
 llvm/tools/llvm-objcopy/ObjcopyOptions.cpp    |  36 +++++
 llvm/tools/llvm-objcopy/ObjcopyOpts.td        |   6 +
 10 files changed, 283 insertions(+), 8 deletions(-)
 create mode 100644 llvm/test/tools/llvm-objcopy/ELF/compress-sections-within-segment.s
 create mode 100644 llvm/test/tools/llvm-objcopy/ELF/compress-sections.s

diff --git a/llvm/docs/CommandGuide/llvm-objcopy.rst b/llvm/docs/CommandGuide/llvm-objcopy.rst
index 985d16eb11cfb..57d6280d57c8b 100644
--- a/llvm/docs/CommandGuide/llvm-objcopy.rst
+++ b/llvm/docs/CommandGuide/llvm-objcopy.rst
@@ -309,6 +309,14 @@ them.
  Compress DWARF debug sections in the output, using the specified format.
  Supported formats are ``zlib`` and ``zstd``. Use ``zlib`` if ``<format>`` is omitted.
 
+.. option:: --compress-sections <section>=<format>
+
+ Compress or decompress sections matched by ``<section>`` using the specified
+ format. Supported formats are ``zlib`` and ``zstd``. Specify ``none`` for
+ decompression. When a section is matched by multiple options, the last one
+ wins. A wildcard ``<section>`` starting with '!' is disallowed.
+ Sections within a segment cannot be (de)compressed.
+
 .. option:: --decompress-debug-sections
 
  Decompress any compressed DWARF debug sections in the output.
diff --git a/llvm/docs/ReleaseNotes.rst b/llvm/docs/ReleaseNotes.rst
index d2d542752b555..11add9274a7a0 100644
--- a/llvm/docs/ReleaseNotes.rst
+++ b/llvm/docs/ReleaseNotes.rst
@@ -186,6 +186,10 @@ Changes to the LLVM tools
   for ELF input to skip the specified symbols when executing other options
   that can change a symbol's name, binding or visibility.
 
+* llvm-objcopy now supports ``--compress-sections`` to compress or decompress
+  arbitrary sections not within a segment.
+  (`#85036 <https://github.com/llvm/llvm-project/pull/85036>`_.)
+
 * llvm-profgen now supports COFF+DWARF binaries. This enables Sample-based PGO
   on Windows using Intel VTune's SEP. For details on usage, see the `end-user
   documentation for SPGO
diff --git a/llvm/include/llvm/ObjCopy/CommonConfig.h b/llvm/include/llvm/ObjCopy/CommonConfig.h
index 9d6d5fb23b18e..ae08d4032736e 100644
--- a/llvm/include/llvm/ObjCopy/CommonConfig.h
+++ b/llvm/include/llvm/ObjCopy/CommonConfig.h
@@ -262,6 +262,9 @@ struct CommonConfig {
   bool DecompressDebugSections = false;
 
   DebugCompressionType CompressionType = DebugCompressionType::None;
+
+  SmallVector<std::pair<NameMatcher, llvm::DebugCompressionType>, 0>
+      compressSections;
 };
 
 } // namespace objcopy
diff --git a/llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp b/llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp
index 205bc1ef5b1a1..f343d1447e055 100644
--- a/llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp
+++ b/llvm/lib/ObjCopy/ELF/ELFObjcopy.cpp
@@ -215,23 +215,41 @@ static Error dumpSectionToFile(StringRef SecName, StringRef Filename,
 }
 
 Error Object::compressOrDecompressSections(const CommonConfig &Config) {
-  // Build a list of the debug sections we are going to replace.
-  // We can't call `AddSection` while iterating over sections,
+  // Build a list of sections we are going to replace.
+  // We can't call `addSection` while iterating over sections,
   // because it would mutate the sections array.
   SmallVector<std::pair<SectionBase *, std::function<SectionBase *()>>, 0>
       ToReplace;
   for (SectionBase &Sec : sections()) {
-    if ((Sec.Flags & SHF_ALLOC) || !StringRef(Sec.Name).starts_with(".debug"))
+    std::optional<DebugCompressionType> CType;
+    for (auto &[Matcher, T] : Config.compressSections)
+      if (Matcher.matches(Sec.Name))
+        CType = T;
+    // Handle --compress-debug-sections and --decompress-debug-sections, which
+    // apply to non-ALLOC debug sections.
+    if (!(Sec.Flags & SHF_ALLOC) && StringRef(Sec.Name).starts_with(".debug")) {
+      if (Config.CompressionType != DebugCompressionType::None)
+        CType = Config.CompressionType;
+      else if (Config.DecompressDebugSections)
+        CType = DebugCompressionType::None;
+    }
+    if (!CType)
       continue;
+
+    if (Sec.ParentSegment)
+      return createStringError(
+          errc::invalid_argument,
+          "section '" + Sec.Name +
+              "' within a segment cannot be (de)compressed");
+
     if (auto *CS = dyn_cast<CompressedSection>(&Sec)) {
-      if (Config.DecompressDebugSections) {
+      if (*CType == DebugCompressionType::None)
         ToReplace.emplace_back(
             &Sec, [=] { return &addSection<DecompressedSection>(*CS); });
-      }
-    } else if (Config.CompressionType != DebugCompressionType::None) {
-      ToReplace.emplace_back(&Sec, [&, S = &Sec] {
+    } else if (*CType != DebugCompressionType::None) {
+      ToReplace.emplace_back(&Sec, [=, S = &Sec] {
         return &addSection<CompressedSection>(
-            CompressedSection(*S, Config.CompressionType, Is64Bits));
+            CompressedSection(*S, *CType, Is64Bits));
       });
     }
   }
diff --git a/llvm/lib/ObjCopy/ELF/ELFObject.cpp b/llvm/lib/ObjCopy/ELF/ELFObject.cpp
index 8b6a0035dae35..02591e6f987c2 100644
--- a/llvm/lib/ObjCopy/ELF/ELFObject.cpp
+++ b/llvm/lib/ObjCopy/ELF/ELFObject.cpp
@@ -548,6 +548,7 @@ CompressedSection::CompressedSection(const SectionBase &Sec,
                         CompressedData);
 
   Flags |= ELF::SHF_COMPRESSED;
+  OriginalFlags |= ELF::SHF_COMPRESSED;
   size_t ChdrSize = Is64Bits ? sizeof(object::Elf_Chdr_Impl<object::ELF64LE>)
                              : sizeof(object::Elf_Chdr_Impl<object::ELF32LE>);
   Size = ChdrSize + CompressedData.size();
@@ -2161,6 +2162,10 @@ Error Object::removeSections(
       std::begin(Sections), std::end(Sections), [=](const SecPtr &Sec) {
         if (ToRemove(*Sec))
           return false;
+        // TODO: A compressed relocation section may be recognized as
+        // RelocationSectionBase. We don't want such a section to be removed.
+        if (isa<CompressedSection>(Sec))
+          return true;
         if (auto RelSec = dyn_cast<RelocationSectionBase>(Sec.get())) {
           if (auto ToRelSec = RelSec->getSection())
             return !ToRemove(*ToRelSec);
diff --git a/llvm/test/tools/llvm-objcopy/ELF/compress-sections-within-segment.s b/llvm/test/tools/llvm-objcopy/ELF/compress-sections-within-segment.s
new file mode 100644
index 0000000000000..064ffcadc12e3
--- /dev/null
+++ b/llvm/test/tools/llvm-objcopy/ELF/compress-sections-within-segment.s
@@ -0,0 +1,38 @@
+## Disallow (de)compression for sections within a segment as they are
+## effectively immutable.
+# RUN: rm -rf %t && mkdir %t && cd %t
+# RUN: yaml2obj %s -o a
+# RUN: not llvm-objcopy a /dev/null --compress-sections .text=zlib 2>&1 | FileCheck %s --implicit-check-not=error:
+
+# CHECK: error: 'a': section '.text' within a segment cannot be (de)compressed
+
+# RUN: not llvm-objcopy a /dev/null --compress-sections foo=none 2>&1 | FileCheck %s --check-prefix=CHECK2 --implicit-check-not=error:
+
+# CHECK2: error: 'a': section 'foo' within a segment cannot be (de)compressed
+
+## There is an error even if 'foo' is already compressed with zlib.
+# RUN: not llvm-objcopy a /dev/null --compress-sections foo=zlib 2>&1 | FileCheck %s --check-prefix=CHECK3 --implicit-check-not=error:
+
+# CHECK3: error: 'a': section 'foo' within a segment cannot be (de)compressed
+
+--- !ELF
+FileHeader:
+  Class:      ELFCLASS64
+  Data:       ELFDATA2LSB
+  Type:       ET_EXEC
+  Machine:    EM_X86_64
+ProgramHeaders:
+  - Type:     PT_LOAD
+    FirstSec: .text
+    LastSec:  foo
+    Align:    0x1000
+    Offset:   0x1000
+Sections:
+  - Name:     .text
+    Type:     SHT_PROGBITS
+    Offset:   0x1000
+    Content:  C3
+  - Name:     foo
+    Type:     SHT_PROGBITS
+    Flags:    [ SHF_COMPRESSED ]
+    Content:  010000000000000040000000000000000100000000000000789cd36280002d3269002f800151
diff --git a/llvm/test/tools/llvm-objcopy/ELF/compress-sections.s b/llvm/test/tools/llvm-objcopy/ELF/compress-sections.s
new file mode 100644
index 0000000000000..e6fa86068a1ae
--- /dev/null
+++ b/llvm/test/tools/llvm-objcopy/ELF/compress-sections.s
@@ -0,0 +1,128 @@
+# REQUIRES: x86-registered-target, zlib, zstd
+
+# RUN: rm -rf %t && mkdir %t && cd %t
+# RUN: llvm-mc -filetype=obj -triple=x86_64 %s -o a.o
+## '*0=none' wins because it is the last. '*0' sections are decompressed (if originally compressed) or kept unchanged (if uncompressed).
+## No section is named 'nomatch'. The third option is a no-op.
+# RUN: llvm-objcopy a.o out --compress-sections='*0=zlib' --compress-sections '*0=none' --compress-sections 'nomatch=none' 2>&1 | count 0
+# RUN: llvm-readelf -S out | FileCheck %s --check-prefix=CHECK1
+
+# CHECK1:      Name           Type          Address     Off      Size     ES Flg Lk Inf Al
+# CHECK1:      .text          PROGBITS [[#%x,TEXT:]]    [[#%x,]] [[#%x,]] 00 AX   0   0  4
+# CHECK1:      foo0           PROGBITS [[#%x,FOO0:]]    [[#%x,]] [[#%x,]] 00 A    0   0  8
+# CHECK1-NEXT: .relafoo0      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  I  11   3  8
+# CHECK1-NEXT: foo1           PROGBITS [[#%x,FOO1:]]    [[#%x,]] [[#%x,]] 00 A    0   0  8
+# CHECK1-NEXT: .relafoo1      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  I  11   5  8
+# CHECK1:      nonalloc0      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  8
+# CHECK1-NEXT: .relanonalloc0 RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  I  11   7  8
+# CHECK1-NEXT: nonalloc1      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  8
+# CHECK1-NEXT: .debug_str     PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 01 MS   0   0  1
+
+## Mixing zlib and zstd.
+# RUN: llvm-objcopy a.o out2 --compress-sections '*c0=zlib' --compress-sections .debug_str=zstd
+# RUN: llvm-readelf -Sr -x nonalloc0 -x .debug_str out2 2>&1 | FileCheck %s --check-prefix=CHECK2
+# RUN: llvm-readelf -z -x nonalloc0 -x .debug_str out2 | FileCheck %s --check-prefix=CHECK2DE
+
+# CHECK2:      Name           Type          Address     Off      Size     ES Flg Lk Inf Al
+# CHECK2:      .text          PROGBITS [[#%x,TEXT:]]    [[#%x,]] [[#%x,]] 00 AX   0   0  4
+# CHECK2:      foo0           PROGBITS [[#%x,FOO0:]]    [[#%x,]] [[#%x,]] 00 A    0   0  8
+# CHECK2-NEXT: .relafoo0      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  I  11   3  8
+# CHECK2-NEXT: foo1           PROGBITS [[#%x,FOO1:]]    [[#%x,]] [[#%x,]] 00 A    0   0  8
+# CHECK2-NEXT: .relafoo1      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  I  11   5  8
+# CHECK2:      nonalloc0      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00   C  0   0  8
+# CHECK2-NEXT: .relanonalloc0 RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  IC 11   7  8
+# CHECK2-NEXT: nonalloc1      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  8
+# CHECK2-NEXT: .debug_str     PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 01 MSC  0   0  8
+
+## llvm-readelf -r doesn't support SHF_COMPRESSED SHT_RELA.
+# CHECK2: warning: {{.*}}: unable to read relocations from SHT_RELA section with index 8: section [index 8] has an invalid sh_size ([[#]]) which is not a multiple of its sh_entsize (24)
+
+# CHECK2:      Hex dump of section 'nonalloc0':
+## zlib with ch_size=0x10
+# CHECK2-NEXT: 01000000 00000000 10000000 00000000
+# CHECK2-NEXT: 08000000 00000000 {{.*}}
+# CHECK2:      Hex dump of section '.debug_str':
+## zstd with ch_size=0x38
+# CHECK2-NEXT: 02000000 00000000 38000000 00000000
+# CHECK2-NEXT: 01000000 00000000 {{.*}}
+
+# CHECK2DE:       Hex dump of section 'nonalloc0':
+# CHECK2DE-NEXT:  0x00000000 00000000 00000000 00000000 00000000 ................
+# CHECK2DE-EMPTY:
+# CHECK2DE-NEXT:  Hex dump of section '.debug_str':
+# CHECK2DE-NEXT:  0x00000000 41414141 41414141 41414141 41414141 AAAAAAAAAAAAAAAA
+
+## --decompress-debug-sections takes precedence, even if it is before --compress-sections.
+# RUN: llvm-objcopy a.o out3 --decompress-debug-sections --compress-sections .debug_str=zstd
+# RUN: llvm-readelf -S out3 | FileCheck %s --check-prefix=CHECK3
+
+# CHECK3:      .debug_str PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 01 MS   0   0  1
+
+# RUN: llvm-objcopy a.o out4 --compress-sections '*0=zlib'
+# RUN: llvm-readelf -S out4 | FileCheck %s --check-prefix=CHECK4
+
+# CHECK4:      Name           Type          Address     Off      Size     ES Flg Lk Inf Al
+# CHECK4:      .text          PROGBITS [[#%x,TEXT:]]    [[#%x,]] [[#%x,]] 00  AX  0   0  4
+# CHECK4:      foo0           PROGBITS [[#%x,FOO0:]]    [[#%x,]] [[#%x,]] 00  AC  0   0  8
+# CHECK4-NEXT: .relafoo0      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  IC 11   3  8
+# CHECK4-NEXT: foo1           PROGBITS [[#%x,FOO1:]]    [[#%x,]] [[#%x,]] 00   A  0   0  8
+# CHECK4-NEXT: .relafoo1      RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18   I 11   5  8
+# CHECK4:      nonalloc0      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00   C  0   0  8
+# CHECK4-NEXT: .relanonalloc0 RELA     [[#%x,]]         [[#%x,]] [[#%x,]] 18  IC 11   7  8
+# CHECK4-NEXT: nonalloc1      PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  8
+# CHECK4-NEXT: .debug_str     PROGBITS 0000000000000000 [[#%x,]] [[#%x,]] 01  MS  0   0  1
+
+## If a section is already compressed, compression request for another format is ignored.
+# RUN: llvm-objcopy a.o out5 --compress-sections 'nonalloc0=zlib'
+# RUN: llvm-readelf -x nonalloc0 out5 | FileCheck %s --check-prefix=CHECK5
+# RUN: llvm-objcopy out5 out5a --compress-sections 'nonalloc0=zstd'
+# RUN: cmp out5 out5a
+
+# CHECK5:      Hex dump of section 'nonalloc0':
+## zlib with ch_size=0x10
+# CHECK5-NEXT: 01000000 00000000 10000000 00000000
+# CHECK5-NEXT: 08000000 00000000 {{.*}}
+
+# RUN: not llvm-objcopy --compress-sections=foo a.o out 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=ERR1 --implicit-check-not=error:
+# ERR1:      error: --compress-sections: parse error, not 'section-glob=[none|zlib|zstd]'
+
+# RUN: llvm-objcopy --compress-sections 'a[=zlib' a.o out 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=ERR2 --implicit-check-not=error:
+# ERR2:      warning: invalid glob pattern, unmatched '['
+
+# RUN: not llvm-objcopy a.o out --compress-sections='.debug*=zlib-gabi' --compress-sections='.debug*=' 2>&1 | \
+# RUN:   FileCheck -check-prefix=ERR3 %s
+# ERR3:      error: invalid or unsupported --compress-sections format: .debug*=zlib-gabi
+
+# RUN: not llvm-objcopy a.o out --compress-sections='!.debug*=zlib' 2>&1 | \
+# RUN:   FileCheck -check-prefix=ERR4 %s
+# ERR4:      error: --compress-sections: negative pattern is unsupported
+
+.globl _start
+_start:
+  ret
+
+.section foo0,"a"
+.balign 8
+.quad .text-.
+.quad .text-.
+.section foo1,"a"
+.balign 8
+.quad .text-.
+.quad .text-.
+.section nonalloc0,""
+.balign 8
+.quad .text+1
+.quad .text+2
+sym0:
+.section nonalloc1,""
+.balign 8
+.quad 42
+sym1:
+
+.section .debug_str,"MS",@progbits,1
+.Linfo_string0:
+  .asciz "AAAAAAAAAAAAAAAAAAAAAAAAAAA"
+.Linfo_string1:
+  .asciz "BBBBBBBBBBBBBBBBBBBBBBBBBBB"
diff --git a/llvm/test/tools/llvm-objcopy/ELF/decompress-sections.test b/llvm/test/tools/llvm-objcopy/ELF/decompress-sections.test
index 4258ddbe66a3e..d9f4f3809c4dd 100644
--- a/llvm/test/tools/llvm-objcopy/ELF/decompress-sections.test
+++ b/llvm/test/tools/llvm-objcopy/ELF/decompress-sections.test
@@ -4,6 +4,8 @@
 # RUN: yaml2obj %s -o %t
 # RUN: llvm-objcopy --decompress-debug-sections %t %t.de
 # RUN: llvm-readelf -S %t.de | FileCheck %s
+# RUN: llvm-objcopy --compress-sections '*nonalloc=none' --compress-sections .debugx=none %t %t.1.de
+# RUN: cmp %t.de %t.1.de
 
 # CHECK:        Name              Type            Address          Off      Size     ES Flg Lk Inf Al
 # CHECK:        .debug_alloc      PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00  AC  0   0  0
@@ -11,6 +13,33 @@
 # CHECK-NEXT:   .debugx           PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  1
 # CHECK-NEXT:   nodebug           PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00   C  0   0  0
 
+# RUN: llvm-objcopy --compress-sections '.debug*=none' %t %t2.de
+# RUN: llvm-readelf -S -x .debug_alloc -x .debug_nonalloc -x .debugx %t2.de | FileCheck %s --check-prefix=CHECK2
+
+# CHECK2:        Name              Type            Address          Off      Size     ES Flg Lk Inf Al
+# CHECK2:        .debug_alloc      PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00  A   0   0  1
+# CHECK2-NEXT:   .debug_nonalloc   PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  1
+# CHECK2-NEXT:   .debugx           PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00      0   0  1
+# CHECK2-NEXT:   nodebug           PROGBITS        0000000000000000 [[#%x,]] [[#%x,]] 00   C  0   0  0
+
+# CHECK2:       Hex dump of section '.debug_alloc':
+# CHECK2-NEXT:  0x00000000 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000010 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000020 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000030 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-EMPTY:
+# CHECK2:       Hex dump of section '.debug_nonalloc':
+# CHECK2-NEXT:  0x00000000 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000010 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000020 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000030 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-EMPTY:
+# CHECK2-NEXT:  Hex dump of section '.debugx':
+# CHECK2-NEXT:  0x00000000 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000010 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000020 2a000000 00000000 2a000000 00000000 *.......*.......
+# CHECK2-NEXT:  0x00000030 2a000000 00000000 2a000000 00000000 *.......*.......
+
 --- !ELF
 FileHeader:
   Class:   ELFCLASS64
diff --git a/llvm/tools/llvm-objcopy/ObjcopyOptions.cpp b/llvm/tools/llvm-objcopy/ObjcopyOptions.cpp
index 7269c51a08d6b..70e85460d3df0 100644
--- a/llvm/tools/llvm-objcopy/ObjcopyOptions.cpp
+++ b/llvm/tools/llvm-objcopy/ObjcopyOptions.cpp
@@ -736,6 +736,42 @@ objcopy::parseObjcopyOptions(ArrayRef<const char *> RawArgsArr,
       return createStringError(errc::invalid_argument, Reason);
   }
 
+  for (const auto *A : InputArgs.filtered(OBJCOPY_compress_sections)) {
+    SmallVector<StringRef, 0> Fields;
+    StringRef(A->getValue()).split(Fields, '=');
+    if (Fields.size() != 2 || Fields[1].empty()) {
+      return createStringError(
+          errc::invalid_argument,
+          A->getSpelling() +
+              ": parse error, not 'section-glob=[none|zlib|zstd]'");
+    }
+
+    auto Type = StringSwitch<DebugCompressionType>(Fields[1])
+                    .Case("zlib", DebugCompressionType::Zlib)
+                    .Case("zstd", DebugCompressionType::Zstd)
+                    .Default(DebugCompressionType::None);
+    if (Type == DebugCompressionType::None && Fields[1] != "none") {
+      return createStringError(
+          errc::invalid_argument,
+          "invalid or unsupported --compress-sections format: %s",
+          A->getValue());
+    }
+
+    auto &P = Config.compressSections.emplace_back();
+    P.second = Type;
+    auto Matcher =
+        NameOrPattern::create(Fields[0], SectionMatchStyle, ErrorCallback);
+    // =none allows overriding a previous =zlib or =zstd. Reject negative
+    // patterns, which would be confusing.
+    if (Matcher && !Matcher->isPositiveMatch()) {
+      return createStringError(
+          errc::invalid_argument,
+          "--compress-sections: negative pattern is unsupported");
+    }
+    if (Error E = P.first.addMatcher(std::move(Matcher)))
+      return std::move(E);
+  }
+
   Config.AddGnuDebugLink = InputArgs.getLastArgValue(OBJCOPY_add_gnu_debuglink);
   // The gnu_debuglink's target is expected to not change or else its CRC would
   // become invalidated and get rejected. We can avoid recalculating the
diff --git a/llvm/tools/llvm-objcopy/ObjcopyOpts.td b/llvm/tools/llvm-objcopy/ObjcopyOpts.td
index be02616e8c68e..4bc80eba05f8e 100644
--- a/llvm/tools/llvm-objcopy/ObjcopyOpts.td
+++ b/llvm/tools/llvm-objcopy/ObjcopyOpts.td
@@ -35,6 +35,12 @@ def : Flag<["--"], "compress-debug-sections">, Alias<compress_debug_sections>,
       AliasArgs<["zlib"]>;
 def decompress_debug_sections : Flag<["--"], "decompress-debug-sections">,
                                 HelpText<"Decompress DWARF debug sections">;
+defm compress_sections
+    : Eq<"compress-sections",
+         "Compress or decompress sections using specified format. Supported "
+         "formats: zlib, zstd. Specify 'none' for decompression">,
+      MetaVarName<"<section-glob>=<format>">;
+
 defm split_dwo
     : Eq<"split-dwo", "Equivalent to --extract-dwo and <dwo-file> as the output file and no other options, "
                       "and then --strip-dwo on the input file">,

From b79b6f9cf00e72b21f5629aa3c5d9c1786611ec9 Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov@fb.com>
Date: Mon, 15 Apr 2024 09:50:12 +0200
Subject: [PATCH 391/886] [BOLT] Use offset deduplication for cold fragments

Apply deduplication for uniformity and BAT section size reduction.

Changes BAT section size to:
- large binary: 39541552 bytes (1.02x original),
- medium binary: 3828996 bytes (0.64x),
- small binary: 928 bytes (0.65x).

Test Plan: Updated bolt-address-translation.test

Reviewers: rafaelauler, dcci, ayermolo, JDevlieghere, maksfb

Reviewed By: maksfb

Pull Request: https://github.com/llvm/llvm-project/pull/87853
---
 bolt/docs/BAT.md                              |  9 +-
 .../bolt/Profile/BoltAddressTranslation.h     |  6 +-
 bolt/lib/Profile/BoltAddressTranslation.cpp   | 86 ++++++++++---------
 bolt/test/X86/bolt-address-translation.test   |  2 +-
 4 files changed, 53 insertions(+), 50 deletions(-)

diff --git a/bolt/docs/BAT.md b/bolt/docs/BAT.md
index f23ef1abf8761..7ffb5d7c00816 100644
--- a/bolt/docs/BAT.md
+++ b/bolt/docs/BAT.md
@@ -81,9 +81,10 @@ Hot indices are delta encoded, implicitly starting at zero.
 | `FuncHash` | 8b | Function hash for input function | Hot |
 | `NumBlocks` | ULEB128 | Number of basic blocks in the original function | Hot |
 | `NumSecEntryPoints` | ULEB128 | Number of secondary entry points in the original function | Hot |
+| `ColdInputSkew` | ULEB128 | Skew to apply to all input offsets | Cold |
 | `NumEntries` | ULEB128 | Number of address translation entries for a function | Both |
-| `EqualElems` | ULEB128 | Number of equal offsets in the beginning of a function | Hot |
-| `BranchEntries` | Bitmask, `alignTo(EqualElems, 8)` bits | If `EqualElems` is non-zero, bitmask denoting entries with `BRANCHENTRY` bit | Hot |
+| `EqualElems` | ULEB128 | Number of equal offsets in the beginning of a function | Both |
+| `BranchEntries` | Bitmask, `alignTo(EqualElems, 8)` bits | If `EqualElems` is non-zero, bitmask denoting entries with `BRANCHENTRY` bit | Both |
 
 Function header is followed by *Address Translation Table* with `NumEntries`
 total entries, and *Secondary Entry Points* table with `NumSecEntryPoints`
@@ -99,8 +100,8 @@ entry is encoded. Input offsets implicitly start at zero.
 | `BBHash` | Optional, 8b | Basic block hash in input binary | BB |
 | `BBIdx`  | Optional, Delta, ULEB128 | Basic block index in input binary | BB |
 
-For hot fragments, the table omits the first `EqualElems` input offsets
-where the input offset equals output offset.
+The table omits the first `EqualElems` input offsets where the input offset
+equals output offset.
 
 `BRANCHENTRY` bit denotes whether a given offset pair is a control flow source
 (branch or call instruction). If not set, it signifies a control flow target
diff --git a/bolt/include/bolt/Profile/BoltAddressTranslation.h b/bolt/include/bolt/Profile/BoltAddressTranslation.h
index eef05e8a0e681..68b993ee363cc 100644
--- a/bolt/include/bolt/Profile/BoltAddressTranslation.h
+++ b/bolt/include/bolt/Profile/BoltAddressTranslation.h
@@ -149,9 +149,9 @@ class BoltAddressTranslation {
   /// entries in function address translation map.
   APInt calculateBranchEntriesBitMask(MapTy &Map, size_t EqualElems);
 
-  /// Calculate the number of equal offsets (output = input) in the beginning
-  /// of the function.
-  size_t getNumEqualOffsets(const MapTy &Map) const;
+  /// Calculate the number of equal offsets (output = input - skew) in the
+  /// beginning of the function.
+  size_t getNumEqualOffsets(const MapTy &Map, uint32_t Skew) const;
 
   std::map<uint64_t, MapTy> Maps;
 
diff --git a/bolt/lib/Profile/BoltAddressTranslation.cpp b/bolt/lib/Profile/BoltAddressTranslation.cpp
index 0141ce189acda..ac6e3d01f2718 100644
--- a/bolt/lib/Profile/BoltAddressTranslation.cpp
+++ b/bolt/lib/Profile/BoltAddressTranslation.cpp
@@ -153,12 +153,13 @@ APInt BoltAddressTranslation::calculateBranchEntriesBitMask(MapTy &Map,
   return BitMask;
 }
 
-size_t BoltAddressTranslation::getNumEqualOffsets(const MapTy &Map) const {
+size_t BoltAddressTranslation::getNumEqualOffsets(const MapTy &Map,
+                                                  uint32_t Skew) const {
   size_t EqualOffsets = 0;
   for (const std::pair<const uint32_t, uint32_t> &KeyVal : Map) {
     const uint32_t OutputOffset = KeyVal.first;
     const uint32_t InputOffset = KeyVal.second >> 1;
-    if (OutputOffset == InputOffset)
+    if (OutputOffset == InputOffset - Skew)
       ++EqualOffsets;
     else
       break;
@@ -196,12 +197,17 @@ void BoltAddressTranslation::writeMaps(std::map<uint64_t, MapTy> &Maps,
         SecondaryEntryPointsMap.count(Address)
             ? SecondaryEntryPointsMap[Address].size()
             : 0;
+    uint32_t Skew = 0;
     if (Cold) {
       auto HotEntryIt = Maps.find(ColdPartSource[Address]);
       assert(HotEntryIt != Maps.end());
       size_t HotIndex = std::distance(Maps.begin(), HotEntryIt);
       encodeULEB128(HotIndex - PrevIndex, OS);
       PrevIndex = HotIndex;
+      // Skew of all input offsets for cold fragments is simply the first input
+      // offset.
+      Skew = Map.begin()->second >> 1;
+      encodeULEB128(Skew, OS);
     } else {
       // Function hash
       size_t BFHash = getBFHash(HotInputAddress);
@@ -217,24 +223,21 @@ void BoltAddressTranslation::writeMaps(std::map<uint64_t, MapTy> &Maps,
                         << '\n');
     }
     encodeULEB128(NumEntries, OS);
-    // For hot fragments only: encode the number of equal offsets
-    // (output = input) in the beginning of the function. Only encode one offset
-    // in these cases.
-    const size_t EqualElems = Cold ? 0 : getNumEqualOffsets(Map);
-    if (!Cold) {
-      encodeULEB128(EqualElems, OS);
-      if (EqualElems) {
-        const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
-        APInt BranchEntries = calculateBranchEntriesBitMask(Map, EqualElems);
-        OS.write(reinterpret_cast<const char *>(BranchEntries.getRawData()),
-                 BranchEntriesBytes);
-        LLVM_DEBUG({
-          dbgs() << "BranchEntries: ";
-          SmallString<8> BitMaskStr;
-          BranchEntries.toString(BitMaskStr, 2, false);
-          dbgs() << BitMaskStr << '\n';
-        });
-      }
+    // Encode the number of equal offsets (output = input - skew) in the
+    // beginning of the function. Only encode one offset in these cases.
+    const size_t EqualElems = getNumEqualOffsets(Map, Skew);
+    encodeULEB128(EqualElems, OS);
+    if (EqualElems) {
+      const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
+      APInt BranchEntries = calculateBranchEntriesBitMask(Map, EqualElems);
+      OS.write(reinterpret_cast<const char *>(BranchEntries.getRawData()),
+               BranchEntriesBytes);
+      LLVM_DEBUG({
+        dbgs() << "BranchEntries: ";
+        SmallString<8> BitMaskStr;
+        BranchEntries.toString(BitMaskStr, 2, false);
+        dbgs() << BitMaskStr << '\n';
+      });
     }
     const BBHashMapTy &BBHashMap = getBBHashMap(HotInputAddress);
     size_t Index = 0;
@@ -315,10 +318,12 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
     uint64_t HotAddress = Cold ? 0 : Address;
     PrevAddress = Address;
     uint32_t SecondaryEntryPoints = 0;
+    uint64_t ColdInputSkew = 0;
     if (Cold) {
       HotIndex += DE.getULEB128(&Offset, &Err);
       HotAddress = HotFuncs[HotIndex];
       ColdPartSource.emplace(Address, HotAddress);
+      ColdInputSkew = DE.getULEB128(&Offset, &Err);
     } else {
       HotFuncs.push_back(Address);
       // Function hash
@@ -339,28 +344,25 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
                             getULEB128Size(SecondaryEntryPoints)));
     }
     const uint32_t NumEntries = DE.getULEB128(&Offset, &Err);
-    // Equal offsets, hot fragments only.
-    size_t EqualElems = 0;
+    // Equal offsets.
+    const size_t EqualElems = DE.getULEB128(&Offset, &Err);
     APInt BEBitMask;
-    if (!Cold) {
-      EqualElems = DE.getULEB128(&Offset, &Err);
-      LLVM_DEBUG(dbgs() << formatv("Equal offsets: {0}, {1} bytes\n",
-                                   EqualElems, getULEB128Size(EqualElems)));
-      if (EqualElems) {
-        const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
-        BEBitMask = APInt(alignTo(EqualElems, 8), 0);
-        LoadIntFromMemory(
-            BEBitMask,
-            reinterpret_cast<const uint8_t *>(
-                DE.getBytes(&Offset, BranchEntriesBytes, &Err).data()),
-            BranchEntriesBytes);
-        LLVM_DEBUG({
-          dbgs() << "BEBitMask: ";
-          SmallString<8> BitMaskStr;
-          BEBitMask.toString(BitMaskStr, 2, false);
-          dbgs() << BitMaskStr << ", " << BranchEntriesBytes << " bytes\n";
-        });
-      }
+    LLVM_DEBUG(dbgs() << formatv("Equal offsets: {0}, {1} bytes\n", EqualElems,
+                                 getULEB128Size(EqualElems)));
+    if (EqualElems) {
+      const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
+      BEBitMask = APInt(alignTo(EqualElems, 8), 0);
+      LoadIntFromMemory(
+          BEBitMask,
+          reinterpret_cast<const uint8_t *>(
+              DE.getBytes(&Offset, BranchEntriesBytes, &Err).data()),
+          BranchEntriesBytes);
+      LLVM_DEBUG({
+        dbgs() << "BEBitMask: ";
+        SmallString<8> BitMaskStr;
+        BEBitMask.toString(BitMaskStr, 2, false);
+        dbgs() << BitMaskStr << ", " << BranchEntriesBytes << " bytes\n";
+      });
     }
     MapTy Map;
 
@@ -375,7 +377,7 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
       PrevAddress = OutputAddress;
       int64_t InputDelta = 0;
       if (J < EqualElems) {
-        InputOffset = (OutputOffset << 1) | BEBitMask[J];
+        InputOffset = (OutputOffset + ColdInputSkew << 1) | BEBitMask[J];
       } else {
         InputDelta = DE.getSLEB128(&Offset, &Err);
         InputOffset += InputDelta;
diff --git a/bolt/test/X86/bolt-address-translation.test b/bolt/test/X86/bolt-address-translation.test
index 63234b4c1d218..e6b21c14077b4 100644
--- a/bolt/test/X86/bolt-address-translation.test
+++ b/bolt/test/X86/bolt-address-translation.test
@@ -37,7 +37,7 @@
 # CHECK:      BOLT: 3 out of 7 functions were overwritten.
 # CHECK:      BOLT-INFO: Wrote 6 BAT maps
 # CHECK:      BOLT-INFO: Wrote 3 function and 58 basic block hashes
-# CHECK:      BOLT-INFO: BAT section size (bytes): 924
+# CHECK:      BOLT-INFO: BAT section size (bytes): 928
 #
 # usqrt mappings (hot part). We match against any key (left side containing
 # the bolted binary offsets) because BOLT may change where it puts instructions

From f76808351677f4e361f2acd9c4b3f385901d37ef Mon Sep 17 00:00:00 2001
From: Vyacheslav Levytskyy <vyacheslav.levytskyy@intel.com>
Date: Mon, 15 Apr 2024 09:59:47 +0200
Subject: [PATCH 392/886] [SPIR-V] Update type inference and instruction
 selection (#88254)

This PR contains a series of fixes which are to improve type inference
and instruction selection.

Namely, it includes:
* fix OpSelect to support operands of a pointer type, according to the
SPIR-V specification (previously only integer/float/vectors of integer
or float were supported) -- a new test case is added and existing test
case is updated;
* fix TableGen typo's in definition of register classes and introduce a
new reg class that is a vector of pointers;
* fix usage of a machine function context when there is a need to switch
between different machine functions to infer/validate correct types;
* add usage of TypedPointerType instead of PointerType so that later
stages of type inference are able to distinguish pointer types by their
element types, effectively supporting hierarchy of pointer/pointee types
and avoiding more complicated recursive type matching on level of
machine instructions in favor of direct pointer comparison using LLVM's
`Type *` values;
* extracting detailed information about operand types using known type
rules for some llvm instructions (for instance, by deducing PHI's
operand pointee types if PHI's results type was deducted on previous
stages of type inference), and adding correspondent
`Intrinsic::spv_assign_ptr_type` to keep type info along consequent
passes,
* ensure that OpConstantComposite reuses a constant when it's already
created and available in the same machine function -- otherwise there is
a crash while building a dependency graph, the corresponding test case
is attached,
* implement deduction of function's return type for opaque pointers, a
new test case is attached,
* make 'emit intrinsics' a module pass to resolve function return types
over the module -- first types for all functions of the module must be
calculated, and only after that it's feasible to deduct function return
types on this earlier stage of translation.
---
 llvm/lib/Target/SPIRV/SPIRV.h                 |   2 +-
 llvm/lib/Target/SPIRV/SPIRVCallLowering.cpp   |  18 +-
 llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp | 167 +++++++++++++++++-
 llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp |  11 +-
 llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h   |  22 ++-
 llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp   |  23 ++-
 llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp      |   2 +-
 llvm/lib/Target/SPIRV/SPIRVInstrInfo.td       |  13 +-
 llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp  |   2 +-
 llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp   |  13 +-
 .../Target/SPIRV/SPIRVRegisterBankInfo.cpp    |   2 +
 llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td   |   1 +
 llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td    |  19 +-
 llvm/lib/Target/SPIRV/SPIRVUtils.cpp          |   3 +-
 llvm/lib/Target/SPIRV/SPIRVUtils.h            |   7 +
 llvm/test/CodeGen/SPIRV/const-composite.ll    |  26 +++
 .../CodeGen/SPIRV/instructions/ret-type.ll    |  82 +++++++++
 .../CodeGen/SPIRV/instructions/select-phi.ll  |  58 ++++++
 .../test/CodeGen/SPIRV/instructions/select.ll |  15 ++
 19 files changed, 448 insertions(+), 38 deletions(-)
 create mode 100644 llvm/test/CodeGen/SPIRV/const-composite.ll
 create mode 100644 llvm/test/CodeGen/SPIRV/instructions/ret-type.ll
 create mode 100644 llvm/test/CodeGen/SPIRV/instructions/select-phi.ll

diff --git a/llvm/lib/Target/SPIRV/SPIRV.h b/llvm/lib/Target/SPIRV/SPIRV.h
index 6979107349d96..fb8580cd47c01 100644
--- a/llvm/lib/Target/SPIRV/SPIRV.h
+++ b/llvm/lib/Target/SPIRV/SPIRV.h
@@ -24,7 +24,7 @@ FunctionPass *createSPIRVStripConvergenceIntrinsicsPass();
 FunctionPass *createSPIRVRegularizerPass();
 FunctionPass *createSPIRVPreLegalizerPass();
 FunctionPass *createSPIRVPostLegalizerPass();
-FunctionPass *createSPIRVEmitIntrinsicsPass(SPIRVTargetMachine *TM);
+ModulePass *createSPIRVEmitIntrinsicsPass(SPIRVTargetMachine *TM);
 InstructionSelector *
 createSPIRVInstructionSelector(const SPIRVTargetMachine &TM,
                                const SPIRVSubtarget &Subtarget,
diff --git a/llvm/lib/Target/SPIRV/SPIRVCallLowering.cpp b/llvm/lib/Target/SPIRV/SPIRVCallLowering.cpp
index 9e4ba2191366b..c107b99cf4cb6 100644
--- a/llvm/lib/Target/SPIRV/SPIRVCallLowering.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVCallLowering.cpp
@@ -383,7 +383,16 @@ bool SPIRVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
   if (F.isDeclaration())
     GR->add(&F, &MIRBuilder.getMF(), FuncVReg);
   FunctionType *FTy = getOriginalFunctionType(F);
-  SPIRVType *RetTy = GR->getOrCreateSPIRVType(FTy->getReturnType(), MIRBuilder);
+  Type *FRetTy = FTy->getReturnType();
+  if (isUntypedPointerTy(FRetTy)) {
+    if (Type *FRetElemTy = GR->findDeducedElementType(&F)) {
+      TypedPointerType *DerivedTy =
+          TypedPointerType::get(FRetElemTy, getPointerAddressSpace(FRetTy));
+      GR->addReturnType(&F, DerivedTy);
+      FRetTy = DerivedTy;
+    }
+  }
+  SPIRVType *RetTy = GR->getOrCreateSPIRVType(FRetTy, MIRBuilder);
   FTy = fixFunctionTypeIfPtrArgs(GR, F, FTy, RetTy, ArgTypeVRegs);
   SPIRVType *FuncTy = GR->getOrCreateOpTypeFunctionWithArgs(
       FTy, RetTy, ArgTypeVRegs, MIRBuilder);
@@ -505,8 +514,13 @@ bool SPIRVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
     // TODO: support constexpr casts and indirect calls.
     if (CF == nullptr)
       return false;
-    if (FunctionType *FTy = getOriginalFunctionType(*CF))
+    if (FunctionType *FTy = getOriginalFunctionType(*CF)) {
       OrigRetTy = FTy->getReturnType();
+      if (isUntypedPointerTy(OrigRetTy)) {
+        if (auto *DerivedRetTy = GR->findReturnType(CF))
+          OrigRetTy = DerivedRetTy;
+      }
+    }
   }
 
   MachineRegisterInfo *MRI = MIRBuilder.getMRI();
diff --git a/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp b/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp
index e8ce5a35b457d..472bc8638c9af 100644
--- a/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp
@@ -51,7 +51,7 @@ void initializeSPIRVEmitIntrinsicsPass(PassRegistry &);
 
 namespace {
 class SPIRVEmitIntrinsics
-    : public FunctionPass,
+    : public ModulePass,
       public InstVisitor<SPIRVEmitIntrinsics, Instruction *> {
   SPIRVTargetMachine *TM = nullptr;
   SPIRVGlobalRegistry *GR = nullptr;
@@ -61,6 +61,9 @@ class SPIRVEmitIntrinsics
   DenseMap<Instruction *, Type *> AggrConstTypes;
   DenseSet<Instruction *> AggrStores;
 
+  // a registry of created Intrinsic::spv_assign_ptr_type instructions
+  DenseMap<Value *, CallInst *> AssignPtrTypeInstr;
+
   // deduce element type of untyped pointers
   Type *deduceElementType(Value *I);
   Type *deduceElementTypeHelper(Value *I);
@@ -75,6 +78,9 @@ class SPIRVEmitIntrinsics
   Type *deduceNestedTypeHelper(User *U, Type *Ty,
                                std::unordered_set<Value *> &Visited);
 
+  // deduce Types of operands of the Instruction if possible
+  void deduceOperandElementType(Instruction *I);
+
   void preprocessCompositeConstants(IRBuilder<> &B);
   void preprocessUndefs(IRBuilder<> &B);
 
@@ -111,10 +117,10 @@ class SPIRVEmitIntrinsics
 
 public:
   static char ID;
-  SPIRVEmitIntrinsics() : FunctionPass(ID) {
+  SPIRVEmitIntrinsics() : ModulePass(ID) {
     initializeSPIRVEmitIntrinsicsPass(*PassRegistry::getPassRegistry());
   }
-  SPIRVEmitIntrinsics(SPIRVTargetMachine *_TM) : FunctionPass(ID), TM(_TM) {
+  SPIRVEmitIntrinsics(SPIRVTargetMachine *_TM) : ModulePass(ID), TM(_TM) {
     initializeSPIRVEmitIntrinsicsPass(*PassRegistry::getPassRegistry());
   }
   Instruction *visitInstruction(Instruction &I) { return &I; }
@@ -130,7 +136,15 @@ class SPIRVEmitIntrinsics
   Instruction *visitAllocaInst(AllocaInst &I);
   Instruction *visitAtomicCmpXchgInst(AtomicCmpXchgInst &I);
   Instruction *visitUnreachableInst(UnreachableInst &I);
-  bool runOnFunction(Function &F) override;
+
+  StringRef getPassName() const override { return "SPIRV emit intrinsics"; }
+
+  bool runOnModule(Module &M) override;
+  bool runOnFunction(Function &F);
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    ModulePass::getAnalysisUsage(AU);
+  }
 };
 } // namespace
 
@@ -269,6 +283,12 @@ Type *SPIRVEmitIntrinsics::deduceElementTypeHelper(
       if (Ty)
         break;
     }
+  } else if (auto *Ref = dyn_cast<SelectInst>(I)) {
+    for (Value *Op : {Ref->getTrueValue(), Ref->getFalseValue()}) {
+      Ty = deduceElementTypeByUsersDeep(Op, Visited);
+      if (Ty)
+        break;
+    }
   }
 
   // remember the found relationship
@@ -368,6 +388,112 @@ Type *SPIRVEmitIntrinsics::deduceElementType(Value *I) {
   return IntegerType::getInt8Ty(I->getContext());
 }
 
+// If the Instruction has Pointer operands with unresolved types, this function
+// tries to deduce them. If the Instruction has Pointer operands with known
+// types which differ from expected, this function tries to insert a bitcast to
+// resolve the issue.
+void SPIRVEmitIntrinsics::deduceOperandElementType(Instruction *I) {
+  SmallVector<std::pair<Value *, unsigned>> Ops;
+  Type *KnownElemTy = nullptr;
+  // look for known basic patterns of type inference
+  if (auto *Ref = dyn_cast<PHINode>(I)) {
+    if (!isPointerTy(I->getType()) ||
+        !(KnownElemTy = GR->findDeducedElementType(I)))
+      return;
+    for (unsigned i = 0; i < Ref->getNumIncomingValues(); i++) {
+      Value *Op = Ref->getIncomingValue(i);
+      if (isPointerTy(Op->getType()))
+        Ops.push_back(std::make_pair(Op, i));
+    }
+  } else if (auto *Ref = dyn_cast<SelectInst>(I)) {
+    if (!isPointerTy(I->getType()) ||
+        !(KnownElemTy = GR->findDeducedElementType(I)))
+      return;
+    for (unsigned i = 0; i < Ref->getNumOperands(); i++) {
+      Value *Op = Ref->getOperand(i);
+      if (isPointerTy(Op->getType()))
+        Ops.push_back(std::make_pair(Op, i));
+    }
+  } else if (auto *Ref = dyn_cast<ReturnInst>(I)) {
+    Type *RetTy = F->getReturnType();
+    if (!isPointerTy(RetTy))
+      return;
+    Value *Op = Ref->getReturnValue();
+    if (!Op)
+      return;
+    if (!(KnownElemTy = GR->findDeducedElementType(F))) {
+      if (Type *OpElemTy = GR->findDeducedElementType(Op)) {
+        GR->addDeducedElementType(F, OpElemTy);
+        TypedPointerType *DerivedTy =
+            TypedPointerType::get(OpElemTy, getPointerAddressSpace(RetTy));
+        GR->addReturnType(F, DerivedTy);
+      }
+      return;
+    }
+    Ops.push_back(std::make_pair(Op, 0));
+  } else if (auto *Ref = dyn_cast<ICmpInst>(I)) {
+    if (!isPointerTy(Ref->getOperand(0)->getType()))
+      return;
+    Value *Op0 = Ref->getOperand(0);
+    Value *Op1 = Ref->getOperand(1);
+    Type *ElemTy0 = GR->findDeducedElementType(Op0);
+    Type *ElemTy1 = GR->findDeducedElementType(Op1);
+    if (ElemTy0) {
+      KnownElemTy = ElemTy0;
+      Ops.push_back(std::make_pair(Op1, 1));
+    } else if (ElemTy1) {
+      KnownElemTy = ElemTy1;
+      Ops.push_back(std::make_pair(Op0, 0));
+    }
+  }
+
+  // There is no enough info to deduce types or all is valid.
+  if (!KnownElemTy || Ops.size() == 0)
+    return;
+
+  LLVMContext &Ctx = F->getContext();
+  IRBuilder<> B(Ctx);
+  for (auto &OpIt : Ops) {
+    Value *Op = OpIt.first;
+    if (Op->use_empty())
+      continue;
+    Type *Ty = GR->findDeducedElementType(Op);
+    if (Ty == KnownElemTy)
+      continue;
+    if (Instruction *User = dyn_cast<Instruction>(Op->use_begin()->get()))
+      setInsertPointSkippingPhis(B, User->getNextNode());
+    else
+      B.SetInsertPoint(I);
+    Value *OpTyVal = Constant::getNullValue(KnownElemTy);
+    Type *OpTy = Op->getType();
+    if (!Ty) {
+      GR->addDeducedElementType(Op, KnownElemTy);
+      // check if there is existing Intrinsic::spv_assign_ptr_type instruction
+      auto It = AssignPtrTypeInstr.find(Op);
+      if (It == AssignPtrTypeInstr.end()) {
+        CallInst *CI =
+            buildIntrWithMD(Intrinsic::spv_assign_ptr_type, {OpTy}, OpTyVal, Op,
+                            {B.getInt32(getPointerAddressSpace(OpTy))}, B);
+        AssignPtrTypeInstr[Op] = CI;
+      } else {
+        It->second->setArgOperand(
+            1,
+            MetadataAsValue::get(
+                Ctx, MDNode::get(Ctx, ValueAsMetadata::getConstant(OpTyVal))));
+      }
+    } else {
+      SmallVector<Type *, 2> Types = {OpTy, OpTy};
+      MetadataAsValue *VMD = MetadataAsValue::get(
+          Ctx, MDNode::get(Ctx, ValueAsMetadata::getConstant(OpTyVal)));
+      SmallVector<Value *, 2> Args = {Op, VMD,
+                                      B.getInt32(getPointerAddressSpace(OpTy))};
+      CallInst *PtrCastI =
+          B.CreateIntrinsic(Intrinsic::spv_ptrcast, {Types}, Args);
+      I->setOperand(OpIt.second, PtrCastI);
+    }
+  }
+}
+
 void SPIRVEmitIntrinsics::replaceMemInstrUses(Instruction *Old,
                                               Instruction *New,
                                               IRBuilder<> &B) {
@@ -630,6 +756,7 @@ void SPIRVEmitIntrinsics::replacePointerOperandWithPtrCast(
         ExpectedElementTypeConst, Pointer, {B.getInt32(AddressSpace)}, B);
     GR->addDeducedElementType(CI, ExpectedElementType);
     GR->addDeducedElementType(Pointer, ExpectedElementType);
+    AssignPtrTypeInstr[Pointer] = CI;
     return;
   }
 
@@ -914,6 +1041,7 @@ void SPIRVEmitIntrinsics::insertAssignPtrTypeIntrs(Instruction *I,
   CallInst *CI = buildIntrWithMD(Intrinsic::spv_assign_ptr_type, {I->getType()},
                                  EltTyConst, I, {B.getInt32(AddressSpace)}, B);
   GR->addDeducedElementType(CI, ElemTy);
+  AssignPtrTypeInstr[I] = CI;
 }
 
 void SPIRVEmitIntrinsics::insertAssignTypeIntrs(Instruction *I,
@@ -1070,6 +1198,7 @@ void SPIRVEmitIntrinsics::processParamTypes(Function *F, IRBuilder<> &B) {
             {B.getInt32(getPointerAddressSpace(Arg->getType()))}, B);
         GR->addDeducedElementType(AssignPtrTyCI, ElemTy);
         GR->addDeducedElementType(Arg, ElemTy);
+        AssignPtrTypeInstr[Arg] = AssignPtrTyCI;
       }
     }
   }
@@ -1114,6 +1243,10 @@ bool SPIRVEmitIntrinsics::runOnFunction(Function &Func) {
     insertAssignTypeIntrs(I, B);
     insertPtrCastOrAssignTypeInstr(I, B);
   }
+
+  for (auto &I : instructions(Func))
+    deduceOperandElementType(&I);
+
   for (auto *I : Worklist) {
     TrackConstants = true;
     if (!I->getType()->isVoidTy() || isa<StoreInst>(I))
@@ -1126,13 +1259,29 @@ bool SPIRVEmitIntrinsics::runOnFunction(Function &Func) {
     processInstrAfterVisit(I, B);
   }
 
-  // check if function parameter types are set
-  if (!F->isIntrinsic())
-    processParamTypes(F, B);
-
   return true;
 }
 
-FunctionPass *llvm::createSPIRVEmitIntrinsicsPass(SPIRVTargetMachine *TM) {
+bool SPIRVEmitIntrinsics::runOnModule(Module &M) {
+  bool Changed = false;
+
+  for (auto &F : M) {
+    Changed |= runOnFunction(F);
+  }
+
+  for (auto &F : M) {
+    // check if function parameter types are set
+    if (!F.isDeclaration() && !F.isIntrinsic()) {
+      const SPIRVSubtarget &ST = TM->getSubtarget<SPIRVSubtarget>(F);
+      GR = ST.getSPIRVGlobalRegistry();
+      IRBuilder<> B(F.getContext());
+      processParamTypes(&F, B);
+    }
+  }
+
+  return Changed;
+}
+
+ModulePass *llvm::createSPIRVEmitIntrinsicsPass(SPIRVTargetMachine *TM) {
   return new SPIRVEmitIntrinsics(TM);
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
index 70197e948c658..05e41e06248e3 100644
--- a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
@@ -23,7 +23,6 @@
 #include "llvm/ADT/APInt.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Type.h"
-#include "llvm/IR/TypedPointerType.h"
 #include "llvm/Support/Casting.h"
 #include <cassert>
 
@@ -61,7 +60,6 @@ SPIRVType *SPIRVGlobalRegistry::assignVectTypeToVReg(
 SPIRVType *SPIRVGlobalRegistry::assignTypeToVReg(
     const Type *Type, Register VReg, MachineIRBuilder &MIRBuilder,
     SPIRV::AccessQualifier::AccessQualifier AccessQual, bool EmitIR) {
-
   SPIRVType *SpirvType =
       getOrCreateSPIRVType(Type, MIRBuilder, AccessQual, EmitIR);
   assignSPIRVTypeToVReg(SpirvType, VReg, MIRBuilder.getMF());
@@ -726,7 +724,8 @@ SPIRVType *SPIRVGlobalRegistry::getOpTypeStruct(const StructType *Ty,
                                                 bool EmitIR) {
   SmallVector<Register, 4> FieldTypes;
   for (const auto &Elem : Ty->elements()) {
-    SPIRVType *ElemTy = findSPIRVType(Elem, MIRBuilder);
+    SPIRVType *ElemTy =
+        findSPIRVType(toTypedPointer(Elem, Ty->getContext()), MIRBuilder);
     assert(ElemTy && ElemTy->getOpcode() != SPIRV::OpTypeVoid &&
            "Invalid struct element type");
     FieldTypes.push_back(getSPIRVTypeID(ElemTy));
@@ -919,8 +918,10 @@ SPIRVType *SPIRVGlobalRegistry::restOfCreateSPIRVType(
   return SpirvType;
 }
 
-SPIRVType *SPIRVGlobalRegistry::getSPIRVTypeForVReg(Register VReg) const {
-  auto t = VRegToTypeMap.find(CurMF);
+SPIRVType *
+SPIRVGlobalRegistry::getSPIRVTypeForVReg(Register VReg,
+                                         const MachineFunction *MF) const {
+  auto t = VRegToTypeMap.find(MF ? MF : CurMF);
   if (t != VRegToTypeMap.end()) {
     auto tt = t->second.find(VReg);
     if (tt != t->second.end())
diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
index 2e3e69456ac26..55979ba403a0e 100644
--- a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
+++ b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.h
@@ -21,6 +21,7 @@
 #include "SPIRVInstrInfo.h"
 #include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h"
 #include "llvm/IR/Constant.h"
+#include "llvm/IR/TypedPointerType.h"
 
 namespace llvm {
 using SPIRVType = const MachineInstr;
@@ -58,6 +59,9 @@ class SPIRVGlobalRegistry {
   SmallPtrSet<const Type *, 4> TypesInProcessing;
   DenseMap<const Type *, SPIRVType *> ForwardPointerTypes;
 
+  // if a function returns a pointer, this is to map it into TypedPointerType
+  DenseMap<const Function *, TypedPointerType *> FunResPointerTypes;
+
   // Number of bits pointers and size_t integers require.
   const unsigned PointerSize;
 
@@ -134,6 +138,16 @@ class SPIRVGlobalRegistry {
   void setBound(unsigned V) { Bound = V; }
   unsigned getBound() { return Bound; }
 
+  // Add a record to the map of function return pointer types.
+  void addReturnType(const Function *ArgF, TypedPointerType *DerivedTy) {
+    FunResPointerTypes[ArgF] = DerivedTy;
+  }
+  // Find a record in the map of function return pointer types.
+  const TypedPointerType *findReturnType(const Function *ArgF) {
+    auto It = FunResPointerTypes.find(ArgF);
+    return It == FunResPointerTypes.end() ? nullptr : It->second;
+  }
+
   // Deduced element types of untyped pointers and composites:
   // - Add a record to the map of deduced element types.
   void addDeducedElementType(Value *Val, Type *Ty) { DeducedElTys[Val] = Ty; }
@@ -276,8 +290,12 @@ class SPIRVGlobalRegistry {
           SPIRV::AccessQualifier::ReadWrite);
 
   // Return the SPIR-V type instruction corresponding to the given VReg, or
-  // nullptr if no such type instruction exists.
-  SPIRVType *getSPIRVTypeForVReg(Register VReg) const;
+  // nullptr if no such type instruction exists. The second argument MF
+  // allows to search for the association in a context of the machine functions
+  // than the current one, without switching between different "current" machine
+  // functions.
+  SPIRVType *getSPIRVTypeForVReg(Register VReg,
+                                 const MachineFunction *MF = nullptr) const;
 
   // Whether the given VReg has a SPIR-V type mapped to it yet.
   bool hasSPIRVTypeForVReg(Register VReg) const {
diff --git a/llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp b/llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp
index 8db54c74f2369..b8296c3f6eeae 100644
--- a/llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp
@@ -88,19 +88,24 @@ static void validatePtrTypes(const SPIRVSubtarget &STI,
                              MachineRegisterInfo *MRI, SPIRVGlobalRegistry &GR,
                              MachineInstr &I, unsigned OpIdx,
                              SPIRVType *ResType, const Type *ResTy = nullptr) {
+  // Get operand type
+  MachineFunction *MF = I.getParent()->getParent();
   Register OpReg = I.getOperand(OpIdx).getReg();
   SPIRVType *TypeInst = MRI->getVRegDef(OpReg);
-  SPIRVType *OpType = GR.getSPIRVTypeForVReg(
+  Register OpTypeReg =
       TypeInst && TypeInst->getOpcode() == SPIRV::OpFunctionParameter
           ? TypeInst->getOperand(1).getReg()
-          : OpReg);
+          : OpReg;
+  SPIRVType *OpType = GR.getSPIRVTypeForVReg(OpTypeReg, MF);
   if (!ResType || !OpType || OpType->getOpcode() != SPIRV::OpTypePointer)
     return;
-  SPIRVType *ElemType = GR.getSPIRVTypeForVReg(OpType->getOperand(2).getReg());
+  // Get operand's pointee type
+  Register ElemTypeReg = OpType->getOperand(2).getReg();
+  SPIRVType *ElemType = GR.getSPIRVTypeForVReg(ElemTypeReg, MF);
   if (!ElemType)
     return;
-  bool IsSameMF =
-      ElemType->getParent()->getParent() == ResType->getParent()->getParent();
+  // Check if we need a bitcast to make a statement valid
+  bool IsSameMF = MF == ResType->getParent()->getParent();
   bool IsEqualTypes = IsSameMF ? ElemType == ResType
                                : GR.getTypeForSPIRVType(ElemType) == ResTy;
   if (IsEqualTypes)
@@ -156,7 +161,8 @@ void validateFunCallMachineDef(const SPIRVSubtarget &STI,
     SPIRVType *DefPtrType = DefMRI->getVRegDef(FunDef->getOperand(1).getReg());
     SPIRVType *DefElemType =
         DefPtrType && DefPtrType->getOpcode() == SPIRV::OpTypePointer
-            ? GR.getSPIRVTypeForVReg(DefPtrType->getOperand(2).getReg())
+            ? GR.getSPIRVTypeForVReg(DefPtrType->getOperand(2).getReg(),
+                                     DefPtrType->getParent()->getParent())
             : nullptr;
     if (DefElemType) {
       const Type *DefElemTy = GR.getTypeForSPIRVType(DefElemType);
@@ -177,7 +183,7 @@ void validateFunCallMachineDef(const SPIRVSubtarget &STI,
 // with a processed definition. Return Function pointer if it's a forward
 // call (ahead of definition), and nullptr otherwise.
 const Function *validateFunCall(const SPIRVSubtarget &STI,
-                                MachineRegisterInfo *MRI,
+                                MachineRegisterInfo *CallMRI,
                                 SPIRVGlobalRegistry &GR,
                                 MachineInstr &FunCall) {
   const GlobalValue *GV = FunCall.getOperand(2).getGlobal();
@@ -186,7 +192,8 @@ const Function *validateFunCall(const SPIRVSubtarget &STI,
       const_cast<MachineInstr *>(GR.getFunctionDefinition(F));
   if (!FunDef)
     return F;
-  validateFunCallMachineDef(STI, MRI, MRI, GR, FunCall, FunDef);
+  MachineRegisterInfo *DefMRI = &FunDef->getParent()->getParent()->getRegInfo();
+  validateFunCallMachineDef(STI, DefMRI, CallMRI, GR, FunCall, FunDef);
   return nullptr;
 }
 
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
index e3f76419f1313..aacfecc1e313f 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
@@ -248,7 +248,7 @@ void SPIRVInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 bool SPIRVInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
   if (MI.getOpcode() == SPIRV::GET_ID || MI.getOpcode() == SPIRV::GET_fID ||
       MI.getOpcode() == SPIRV::GET_pID || MI.getOpcode() == SPIRV::GET_vfID ||
-      MI.getOpcode() == SPIRV::GET_vID) {
+      MI.getOpcode() == SPIRV::GET_vID || MI.getOpcode() == SPIRV::GET_vpID) {
     auto &MRI = MI.getMF()->getRegInfo();
     MRI.replaceRegWith(MI.getOperand(0).getReg(), MI.getOperand(1).getReg());
     MI.eraseFromParent();
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
index 99c57dac4141d..a3f981457c8da 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
+++ b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
@@ -22,6 +22,7 @@ let isCodeGenOnly=1 in {
   def GET_pID: Pseudo<(outs pID:$dst_id), (ins ANYID:$src)>;
   def GET_vID: Pseudo<(outs vID:$dst_id), (ins ANYID:$src)>;
   def GET_vfID: Pseudo<(outs vfID:$dst_id), (ins ANYID:$src)>;
+  def GET_vpID: Pseudo<(outs vpID:$dst_id), (ins ANYID:$src)>;
 }
 
 def SPVTypeBin : SDTypeProfile<1, 2, []>;
@@ -55,7 +56,7 @@ multiclass BinOpTypedGen<string name, bits<16> opCode, SDNode node, bit genF = 0
   }
 }
 
-multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genI = 1, bit genF = 0, bit genV = 0> {
+multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genP = 1, bit genI = 1, bit genF = 0, bit genV = 0> {
   if genF then {
     def SFSCond: TernOpTyped<name, opCode, ID, fID, node>;
     def SFVCond: TernOpTyped<name, opCode, vID, fID, node>;
@@ -64,6 +65,10 @@ multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genI =
     def SISCond: TernOpTyped<name, opCode, ID, ID, node>;
     def SIVCond: TernOpTyped<name, opCode, vID, ID, node>;
   }
+  if genP then {
+    def SPSCond: TernOpTyped<name, opCode, ID, pID, node>;
+    def SPVCond: TernOpTyped<name, opCode, vID, pID, node>;
+  }
   if genV then {
     if genF then {
       def VFSCond: TernOpTyped<name, opCode, ID, vfID, node>;
@@ -73,6 +78,10 @@ multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genI =
       def VISCond: TernOpTyped<name, opCode, ID, vID, node>;
       def VIVCond: TernOpTyped<name, opCode, vID, vID, node>;
     }
+    if genP then {
+      def VPSCond: TernOpTyped<name, opCode, ID, vpID, node>;
+      def VPVCond: TernOpTyped<name, opCode, vID, vpID, node>;
+    }
   }
 }
 
@@ -552,7 +561,7 @@ def OpLogicalOr: BinOp<"OpLogicalOr", 166>;
 def OpLogicalAnd: BinOp<"OpLogicalAnd", 167>;
 def OpLogicalNot: UnOp<"OpLogicalNot", 168>;
 
-defm OpSelect: TernOpTypedGen<"OpSelect", 169, select, 1, 1, 1>;
+defm OpSelect: TernOpTypedGen<"OpSelect", 169, select, 1, 1, 1, 1>;
 
 def OpIEqual: BinOp<"OpIEqual", 170>;
 def OpINotEqual: BinOp<"OpINotEqual", 171>;
diff --git a/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
index b9d66de9555b1..f069a92ac6868 100644
--- a/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
@@ -56,7 +56,7 @@ extern void processInstr(MachineInstr &MI, MachineIRBuilder &MIB,
 static bool isMetaInstrGET(unsigned Opcode) {
   return Opcode == SPIRV::GET_ID || Opcode == SPIRV::GET_fID ||
          Opcode == SPIRV::GET_pID || Opcode == SPIRV::GET_vID ||
-         Opcode == SPIRV::GET_vfID;
+         Opcode == SPIRV::GET_vfID || Opcode == SPIRV::GET_vpID;
 }
 
 static bool mayBeInserted(unsigned Opcode) {
diff --git a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
index 7e155a36aadbc..2c964595fc39e 100644
--- a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
@@ -64,9 +64,16 @@ static void addConstantsToTrack(MachineFunction &MF, SPIRVGlobalRegistry *GR) {
             auto *BuildVec = MRI.getVRegDef(MI.getOperand(2).getReg());
             assert(BuildVec &&
                    BuildVec->getOpcode() == TargetOpcode::G_BUILD_VECTOR);
-            for (unsigned i = 0; i < ConstVec->getNumElements(); ++i)
-              GR->add(ConstVec->getElementAsConstant(i), &MF,
-                      BuildVec->getOperand(1 + i).getReg());
+            for (unsigned i = 0; i < ConstVec->getNumElements(); ++i) {
+              // Ensure that OpConstantComposite reuses a constant when it's
+              // already created and available in the same machine function.
+              Constant *ElemConst = ConstVec->getElementAsConstant(i);
+              Register ElemReg = GR->find(ElemConst, &MF);
+              if (!ElemReg.isValid())
+                GR->add(ElemConst, &MF, BuildVec->getOperand(1 + i).getReg());
+              else
+                BuildVec->getOperand(1 + i).setReg(ElemReg);
+            }
           }
           GR->add(Const, &MF, MI.getOperand(2).getReg());
         } else {
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
index 9bf9d7fe5b39e..5983c9229cb3c 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
@@ -39,6 +39,8 @@ SPIRVRegisterBankInfo::getRegBankFromRegClass(const TargetRegisterClass &RC,
     return SPIRV::vIDRegBank;
   case SPIRV::vfIDRegClassID:
     return SPIRV::vfIDRegBank;
+  case SPIRV::vpIDRegClassID:
+    return SPIRV::vpIDRegBank;
   case SPIRV::ANYIDRegClassID:
   case SPIRV::ANYRegClassID:
     return SPIRV::IDRegBank;
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td b/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
index 90c7f3a6e6726..c7f1e172f3d4f 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
@@ -12,4 +12,5 @@ def IDRegBank : RegisterBank<"IDBank", [ID]>;
 def fIDRegBank : RegisterBank<"fIDBank", [fID]>;
 def vIDRegBank : RegisterBank<"vIDBank", [vID]>;
 def vfIDRegBank : RegisterBank<"vfIDBank", [vfID]>;
+def vpIDRegBank : RegisterBank<"vpIDBank", [vpID]>;
 def TYPERegBank : RegisterBank<"TYPEBank", [TYPE]>;
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td b/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
index d0b64b6895d03..6d2bfb91a97f1 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
@@ -12,6 +12,17 @@
 
 let Namespace = "SPIRV" in {
   def p0 : PtrValueType <i32, 0>;
+
+  class P0Vec<ValueType scalar>
+      : PtrValueType <scalar, 0> {
+    let nElem = 2;
+    let ElementType = p0;
+    let isInteger = false;
+    let isFP = false;
+    let isVector = true;
+  }
+
+  def v2p0 : P0Vec<i32>;
   // All registers are for 32-bit identifiers, so have a single dummy register
 
   // Class for registers that are the result of OpTypeXXX instructions
@@ -21,14 +32,16 @@ let Namespace = "SPIRV" in {
   // Class for every other non-type ID
   def ID0 : Register<"ID0">;
   def ID : RegisterClass<"SPIRV", [i32], 32, (add ID0)>;
-  def fID0 : Register<"FID0">;
+  def fID0 : Register<"fID0">;
   def fID : RegisterClass<"SPIRV", [f32], 32, (add fID0)>;
   def pID0 : Register<"pID0">;
   def pID : RegisterClass<"SPIRV", [p0], 32, (add pID0)>;
-  def vID0 : Register<"pID0">;
+  def vID0 : Register<"vID0">;
   def vID : RegisterClass<"SPIRV", [v2i32], 32, (add vID0)>;
-  def vfID0 : Register<"pID0">;
+  def vfID0 : Register<"vfID0">;
   def vfID : RegisterClass<"SPIRV", [v2f32], 32, (add vfID0)>;
+  def vpID0 : Register<"vpID0">;
+  def vpID : RegisterClass<"SPIRV", [v2p0], 32, (add vpID0)>;
 
   def ANYID : RegisterClass<"SPIRV", [i32, f32, p0, v2i32, v2f32], 32, (add ID, fID, pID, vID, vfID)>;
 
diff --git a/llvm/lib/Target/SPIRV/SPIRVUtils.cpp b/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
index 299a4341193bf..2e44c208ed8e0 100644
--- a/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVUtils.cpp
@@ -251,7 +251,8 @@ bool isSpvIntrinsic(const MachineInstr &MI, Intrinsic::ID IntrinsicID) {
 }
 
 Type *getMDOperandAsType(const MDNode *N, unsigned I) {
-  return cast<ValueAsMetadata>(N->getOperand(I))->getType();
+  Type *ElementTy = cast<ValueAsMetadata>(N->getOperand(I))->getType();
+  return toTypedPointer(ElementTy, N->getContext());
 }
 
 // The set of names is borrowed from the SPIR-V translator.
diff --git a/llvm/lib/Target/SPIRV/SPIRVUtils.h b/llvm/lib/Target/SPIRV/SPIRVUtils.h
index c2c3475e1a936..cd1a2af09147e 100644
--- a/llvm/lib/Target/SPIRV/SPIRVUtils.h
+++ b/llvm/lib/Target/SPIRV/SPIRVUtils.h
@@ -149,5 +149,12 @@ inline Type *reconstructFunctionType(Function *F) {
   return FunctionType::get(F->getReturnType(), ArgTys, F->isVarArg());
 }
 
+inline Type *toTypedPointer(Type *Ty, LLVMContext &Ctx) {
+  return isUntypedPointerTy(Ty)
+             ? TypedPointerType::get(IntegerType::getInt8Ty(Ctx),
+                                     getPointerAddressSpace(Ty))
+             : Ty;
+}
+
 } // namespace llvm
 #endif // LLVM_LIB_TARGET_SPIRV_SPIRVUTILS_H
diff --git a/llvm/test/CodeGen/SPIRV/const-composite.ll b/llvm/test/CodeGen/SPIRV/const-composite.ll
new file mode 100644
index 0000000000000..4e304bb951670
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/const-composite.ll
@@ -0,0 +1,26 @@
+; This test is to ensure that OpConstantComposite reuses a constant when it's
+; already created and available in the same machine function. In this test case
+; it's `1` that is passed implicitly as a part of the `foo` function argument
+; and also takes part in a composite constant creation.
+
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-SPIRV: %[[#type_int32:]] = OpTypeInt 32 0
+; CHECK-SPIRV: %[[#const1:]] = OpConstant %[[#type_int32]] 1
+; CHECK-SPIRV: OpTypeArray %[[#]] %[[#const1:]]
+; CHECK-SPIRV: %[[#const0:]] = OpConstant %[[#type_int32]] 0
+; CHECK-SPIRV: OpConstantComposite %[[#]] %[[#const0]] %[[#const1]]
+
+%struct = type { [1 x i64] }
+
+define spir_kernel void @foo(ptr noundef byval(%struct) %arg) {
+entry:
+  call spir_func void @bar(<2 x i32> noundef <i32 0, i32 1>)
+  ret void
+}
+
+define spir_func void @bar(<2 x i32> noundef) {
+entry:
+  ret void
+}
diff --git a/llvm/test/CodeGen/SPIRV/instructions/ret-type.ll b/llvm/test/CodeGen/SPIRV/instructions/ret-type.ll
new file mode 100644
index 0000000000000..bf71eb5628e21
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/instructions/ret-type.ll
@@ -0,0 +1,82 @@
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown --translator-compatibility-mode %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-DAG: OpName %[[Test1:.*]] "test1"
+; CHECK-DAG: OpName %[[Foo:.*]] "foo"
+; CHECK-DAG: OpName %[[Bar:.*]] "bar"
+; CHECK-DAG: OpName %[[Test2:.*]] "test2"
+
+; CHECK-DAG: %[[Long:.*]] = OpTypeInt 64 0
+; CHECK-DAG: %[[Array:.*]] = OpTypeArray %[[Long]] %[[#]]
+; CHECK-DAG: %[[Struct1:.*]] = OpTypeStruct %[[Array]]
+; CHECK-DAG: %[[Struct2:.*]] = OpTypeStruct %[[Struct1]]
+; CHECK-DAG: %[[StructPtr:.*]] = OpTypePointer Function %[[Struct2]]
+; CHECK-DAG: %[[Bool:.*]] = OpTypeBool
+; CHECK-DAG: %[[FooType:.*]] = OpTypeFunction %[[StructPtr:.*]] %[[StructPtr]] %[[StructPtr]] %[[Bool]]
+; CHECK-DAG: %[[Char:.*]] = OpTypeInt 8 0
+; CHECK-DAG: %[[CharPtr:.*]] = OpTypePointer Function %[[Char]]
+
+; CHECK: %[[Test1]] = OpFunction
+; CHECK: OpFunctionCall %[[StructPtr:.*]] %[[Foo]]
+; CHECK: OpFunctionCall %[[StructPtr:.*]] %[[Bar]]
+; CHECK: OpFunctionEnd
+
+; CHECK: %[[Foo]] = OpFunction %[[StructPtr:.*]] None %[[FooType]]
+; CHECK: %[[Arg1:.*]] = OpFunctionParameter %[[StructPtr]]
+; CHECK: %[[Arg2:.*]] = OpFunctionParameter
+; CHECK: %[[Sw:.*]] = OpFunctionParameter
+; CHECK: %[[Res:.*]] = OpInBoundsPtrAccessChain %[[StructPtr]] %[[Arg1]] %[[#]]
+; CHECK: OpReturnValue %[[Res]]
+; CHECK: OpReturnValue %[[Arg2]]
+
+; CHECK: %[[Bar]] = OpFunction %[[StructPtr:.*]] None %[[#]]
+; CHECK: %[[BarArg:.*]] = OpFunctionParameter
+; CHECK: %[[BarRes:.*]] = OpInBoundsPtrAccessChain %[[CharPtr]] %[[BarArg]] %[[#]]
+; CHECK: %[[BarResCasted:.*]] = OpBitcast %[[StructPtr]] %[[BarRes]]
+; CHECK: %[[BarResStruct:.*]] = OpInBoundsPtrAccessChain %[[StructPtr]] %[[#]] %[[#]]
+; CHECK: OpReturnValue %[[BarResStruct]]
+; CHECK: OpReturnValue %[[BarResCasted]]
+
+; CHECK: %[[Test2]] = OpFunction
+; CHECK: OpFunctionCall %[[StructPtr:.*]] %[[Foo]]
+; CHECK: OpFunctionCall %[[StructPtr:.*]] %[[Bar]]
+; CHECK: OpFunctionEnd
+
+%struct = type { %array }
+%array = type { [1 x i64] }
+
+define spir_func void @test1(ptr %arg1, ptr %arg2, i1 %sw) {
+entry:
+  %r1 = call ptr @foo(ptr %arg1, ptr %arg2, i1 %sw)
+  %r2 = call ptr @bar(ptr %arg1, i1 %sw)
+  ret void
+}
+
+define spir_func ptr @foo(ptr %arg1, ptr %arg2, i1 %sw) {
+entry:
+  br i1 %sw, label %exit, label %sw1
+sw1:
+  %result = getelementptr inbounds %struct, ptr %arg1, i64 100
+  ret ptr %result
+exit:
+  ret ptr %arg2
+}
+
+define spir_func ptr @bar(ptr %arg1, i1 %sw) {
+entry:
+  %charptr = getelementptr inbounds i8, ptr %arg1, i64 0
+  br i1 %sw, label %exit, label %sw1
+sw1:
+  %result = getelementptr inbounds %struct, ptr %arg1, i64 100
+  ret ptr %result
+exit:
+  ret ptr %charptr
+}
+
+define spir_func void @test2(ptr %arg1, ptr %arg2, i1 %sw) {
+entry:
+  %r1 = call ptr @foo(ptr %arg1, ptr %arg2, i1 %sw)
+  %r2 = call ptr @bar(ptr %arg1, i1 %sw)
+  ret void
+}
diff --git a/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll b/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll
new file mode 100644
index 0000000000000..afc75c616f023
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll
@@ -0,0 +1,58 @@
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown --translator-compatibility-mode %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-DAG: %[[Char:.*]] = OpTypeInt 8 0
+; CHECK-DAG: %[[Long:.*]] = OpTypeInt 32 0
+; CHECK-DAG: %[[Array:.*]] = OpTypeArray %[[Long]] %[[#]]
+; CHECK-DAG: %[[Struct:.*]] = OpTypeStruct %[[Array]]
+; CHECK-DAG: %[[StructPtr:.*]] = OpTypePointer Function %[[Struct]]
+; CHECK-DAG: %[[CharPtr:.*]] = OpTypePointer Function %[[Char]]
+
+; CHECK: %[[Branch1:.*]] = OpLabel
+; CHECK: %[[Res1:.*]] = OpVariable %[[StructPtr]] Function
+; CHECK: OpBranchConditional %[[#]] %[[#]] %[[Branch2:.*]]
+; CHECK: %[[Res2:.*]] = OpInBoundsPtrAccessChain %[[CharPtr]] %[[#]] %[[#]]
+; CHECK: %[[Res2Casted:.*]] = OpBitcast %[[StructPtr]] %[[Res2]]
+; CHECK: OpBranchConditional %[[#]] %[[#]] %[[BranchSelect:.*]]
+; CHECK: %[[SelectRes:.*]] = OpSelect %[[CharPtr]] %[[#]] %[[#]] %[[#]]
+; CHECK: %[[SelectResCasted:.*]] = OpBitcast %[[StructPtr]] %[[SelectRes]]
+; CHECK: OpLabel
+; CHECK: OpPhi %[[StructPtr]] %[[Res1]] %[[Branch1]] %[[Res2Casted]] %[[Branch2]] %[[SelectResCasted]] %[[BranchSelect]]
+
+%struct = type { %array }
+%array = type { [1 x i64] }
+%array3 = type { [3 x i32] }
+
+define spir_kernel void @foo(ptr addrspace(1) noundef align 1 %arg1, ptr noundef byval(%struct) align 8 %arg2, i1 noundef zeroext %expected) {
+entry:
+  %agg = alloca %array3, align 8
+  %r0 = load i64, ptr %arg2, align 8
+  %add.ptr = getelementptr inbounds i8, ptr %agg, i64 12
+  %r1 = load i32, ptr %agg, align 4
+  %tobool0 = icmp slt i32 %r1, 0
+  br i1 %tobool0, label %exit, label %sw1
+
+sw1:                            ; preds = %entry
+  %incdec1 = getelementptr inbounds i8, ptr %agg, i64 4
+  %r2 = load i32, ptr %incdec1, align 4
+  %tobool1 = icmp slt i32 %r2, 0
+  br i1 %tobool1, label %exit, label %sw2
+
+sw2:                            ; preds = %sw1
+  %incdec2 = getelementptr inbounds i8, ptr %agg, i64 8
+  %r3 = load i32, ptr %incdec2, align 4
+  %tobool2 = icmp slt i32 %r3, 0
+  %spec.select = select i1 %tobool2, ptr %incdec2, ptr %add.ptr
+  br label %exit
+
+exit: ; preds = %sw2, %sw1, %entry
+  %retval.0 = phi ptr [ %agg, %entry ], [ %incdec1, %sw1 ], [ %spec.select, %sw2 ]
+  %add.ptr.i = getelementptr inbounds i8, ptr addrspace(1) %arg1, i64 %r0
+  %r4 = icmp eq ptr %retval.0, %add.ptr
+  %cmp = xor i1 %r4, %expected
+  %frombool6.i = zext i1 %cmp to i8
+  store i8 %frombool6.i, ptr addrspace(1) %add.ptr.i, align 1
+  %r5 = icmp eq ptr %add.ptr, %retval.0
+  ret void
+}
diff --git a/llvm/test/CodeGen/SPIRV/instructions/select.ll b/llvm/test/CodeGen/SPIRV/instructions/select.ll
index f54ef21f20859..c4176b17abb44 100644
--- a/llvm/test/CodeGen/SPIRV/instructions/select.ll
+++ b/llvm/test/CodeGen/SPIRV/instructions/select.ll
@@ -1,6 +1,8 @@
 ; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
 
 ; CHECK-DAG:  OpName [[SCALARi32:%.+]] "select_i32"
+; CHECK-DAG:  OpName [[SCALARPTR:%.+]] "select_ptr"
 ; CHECK-DAG:  OpName [[VEC2i32:%.+]] "select_i32v2"
 ; CHECK-DAG:  OpName [[VEC2i32v2:%.+]] "select_v2i32v2"
 
@@ -17,6 +19,19 @@ define i32 @select_i32(i1 %c, i32 %t, i32 %f) {
   ret i32 %r
 }
 
+; CHECK:      [[SCALARPTR]] = OpFunction
+; CHECK-NEXT: [[C:%.+]] = OpFunctionParameter
+; CHECK-NEXT: [[T:%.+]] = OpFunctionParameter
+; CHECK-NEXT: [[F:%.+]] = OpFunctionParameter
+; CHECK:      OpLabel
+; CHECK:      [[R:%.+]] = OpSelect {{%.+}} [[C]] [[T]] [[F]]
+; CHECK:      OpReturnValue [[R]]
+; CHECK-NEXT: OpFunctionEnd
+define ptr @select_ptr(i1 %c, ptr %t, ptr %f) {
+  %r = select i1 %c, ptr %t, ptr %f
+  ret ptr %r
+}
+
 ; CHECK:      [[VEC2i32]] = OpFunction
 ; CHECK-NEXT: [[C:%.+]] = OpFunctionParameter
 ; CHECK-NEXT: [[T:%.+]] = OpFunctionParameter

From f7bf11cd8db29a178463f2a12b9f1589ecab64e2 Mon Sep 17 00:00:00 2001
From: wanglei <wanglei@loongson.cn>
Date: Mon, 15 Apr 2024 14:38:28 +0800
Subject: [PATCH 393/886] [LoongArch] Pre-commit test for #88694. NFC

---
 llvm/test/CodeGen/LoongArch/gep-imm.ll | 52 ++++++++++++++++++++++++++
 1 file changed, 52 insertions(+)
 create mode 100644 llvm/test/CodeGen/LoongArch/gep-imm.ll

diff --git a/llvm/test/CodeGen/LoongArch/gep-imm.ll b/llvm/test/CodeGen/LoongArch/gep-imm.ll
new file mode 100644
index 0000000000000..52d25f9d63894
--- /dev/null
+++ b/llvm/test/CodeGen/LoongArch/gep-imm.ll
@@ -0,0 +1,52 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc --mtriple=loongarch64 < %s | FileCheck %s
+
+define void @test(ptr %sp, ptr %t, i32 %n) {
+; CHECK-LABEL: test:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    move $a3, $zero
+; CHECK-NEXT:    ld.d $a4, $a0, 0
+; CHECK-NEXT:    lu12i.w $a0, 1
+; CHECK-NEXT:    ori $a5, $a0, 3904
+; CHECK-NEXT:    add.d $a0, $a1, $a5
+; CHECK-NEXT:    add.d $a1, $a4, $a5
+; CHECK-NEXT:    addi.w $a2, $a2, 0
+; CHECK-NEXT:    addi.w $a4, $a3, 0
+; CHECK-NEXT:    bge $a4, $a2, .LBB0_2
+; CHECK-NEXT:    .p2align 4, , 16
+; CHECK-NEXT:  .LBB0_1: # %while_body
+; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT:    addi.d $a4, $a3, 1
+; CHECK-NEXT:    st.w $a4, $a1, 0
+; CHECK-NEXT:    st.w $a3, $a1, 4
+; CHECK-NEXT:    st.w $a4, $a0, 0
+; CHECK-NEXT:    st.w $a3, $a0, 4
+; CHECK-NEXT:    move $a3, $a4
+; CHECK-NEXT:    addi.w $a4, $a3, 0
+; CHECK-NEXT:    blt $a4, $a2, .LBB0_1
+; CHECK-NEXT:  .LBB0_2: # %while_end
+; CHECK-NEXT:    ret
+entry:
+  %s = load ptr, ptr %sp
+  br label %while_cond
+
+while_cond:
+  %phi = phi i32 [ 0, %entry ], [ %i, %while_body ]
+  %gep0 = getelementptr [65536 x i32], ptr %s, i64 0, i64 2000
+  %gep1 = getelementptr [65536 x i32], ptr %s, i64 0, i64 2001
+  %gep2 = getelementptr [65536 x i32], ptr %t, i64 0, i64 2000
+  %gep3 = getelementptr [65536 x i32], ptr %t, i64 0, i64 2001
+  %cmp = icmp slt i32 %phi, %n
+  br i1 %cmp, label %while_body, label %while_end
+
+while_body:
+  %i = add i32 %phi, 1
+  store i32 %i, ptr %gep0
+  store i32 %phi, ptr %gep1
+  store i32 %i, ptr %gep2
+  store i32 %phi, ptr %gep3
+  br label %while_cond
+
+while_end:
+  ret void
+}

From 06714e1bfe8999c0acb5c9b94b60690f7b9d2904 Mon Sep 17 00:00:00 2001
From: Ramkumar Ramachandra <ram.ramachandra@arm.com>
Date: Mon, 15 Apr 2024 09:36:40 +0100
Subject: [PATCH 394/886] llvm-readobj/ARM: activate big-endian attribute
 printing (#87806)

Activate attribute printing for big-endian encoding for ARM. The
supporting code already exists, and the patch is trivial.

Fixes #62400.
---
 .../ELF/ARM/attribute-big-endian.test            | 16 +++++++++++-----
 llvm/tools/llvm-readobj/ELFDumper.cpp            | 10 +++-------
 2 files changed, 14 insertions(+), 12 deletions(-)

diff --git a/llvm/test/tools/llvm-readobj/ELF/ARM/attribute-big-endian.test b/llvm/test/tools/llvm-readobj/ELF/ARM/attribute-big-endian.test
index 7d20b310f9d90..3b94c7994cf3c 100644
--- a/llvm/test/tools/llvm-readobj/ELF/ARM/attribute-big-endian.test
+++ b/llvm/test/tools/llvm-readobj/ELF/ARM/attribute-big-endian.test
@@ -1,17 +1,23 @@
-## We only implement attribute section printing for little-endian encoding.
-
 # RUN: yaml2obj %s -o %t.o
-# RUN: llvm-readobj -A %t.o 2>&1 | FileCheck %s -DFILE=%t.o
+# RUN: llvm-readelf -A %t.o 2>&1 | FileCheck %s
 
-# CHECK: warning: '[[FILE]]': attribute printing not implemented for big-endian ARM objects
+# CHECK:      BuildAttributes {
+# CHECK-NEXT:   FormatVersion: 0x41
+# CHECK-NEXT:   Section 1 {
+# CHECK-NEXT:     SectionLength: 22
+# CHECK-NEXT:     Vendor: armabi
+# CHECK-NEXT:   }
+# CHECK-NEXT: }
 
 --- !ELF
 FileHeader:
   Class:   ELFCLASS32
-## Test big-endian encoding.
   Data:    ELFDATA2MSB
   Type:    ET_REL
   Machine: EM_ARM
 Sections:
   - Name: .ARM.attributes
     Type: SHT_ARM_ATTRIBUTES
+    ContentArray: [ 0x41, 0x00, 0x00, 0x00, 0x16, 0x61, 0x72, 0x6D, 0x61, 0x62,
+    0x69, 0x00, 0x01, 0x0b, 0x00, 0x00, 0x00, 0x04, 0x01, 0x06, 0x01, 0x08,
+    0x01 ]
diff --git a/llvm/tools/llvm-readobj/ELFDumper.cpp b/llvm/tools/llvm-readobj/ELFDumper.cpp
index d3534820c1172..1108672003fc7 100644
--- a/llvm/tools/llvm-readobj/ELFDumper.cpp
+++ b/llvm/tools/llvm-readobj/ELFDumper.cpp
@@ -2833,13 +2833,9 @@ template <class ELFT> void ELFDumper<ELFT>::printArchSpecificInfo() {
                     llvm::endianness::little);
     break;
   case EM_ARM:
-    if (Obj.isLE())
-      printAttributes(ELF::SHT_ARM_ATTRIBUTES,
-                      std::make_unique<ARMAttributeParser>(&W),
-                      llvm::endianness::little);
-    else
-      reportUniqueWarning("attribute printing not implemented for big-endian "
-                          "ARM objects");
+    printAttributes(
+        ELF::SHT_ARM_ATTRIBUTES, std::make_unique<ARMAttributeParser>(&W),
+        Obj.isLE() ? llvm::endianness::little : llvm::endianness::big);
     break;
   case EM_RISCV:
     if (Obj.isLE())

From 5fe146672d2b1c9f257a6ee045e0bc13fed1e504 Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Mon, 15 Apr 2024 16:40:16 +0800
Subject: [PATCH 395/886] [InstCombine] Simplify switch with selects (#84143)

An example from https://github.com/image-rs/image:
```
define void @test_ult_rhsc(i8 %x) {
  %val = add nsw i8 %x, -2
  %cmp = icmp ult i8 %val, 11
  %cond = select i1 %cmp, i8 %val, i8 6
  switch i8 %cond, label %bb1 [
  i8 0, label %bb2
  i8 10, label %bb3
  ]

bb1:
  call void @func1()
  unreachable
bb2:
  call void @func2()
  unreachable
bb3:
  call void @func3()
  unreachable
}
```

When `%cmp` evaluates to false, we can prove that the range of `%val` is
[11, umax]. Thus we can safely replace `%cond` with `%val` since both
`switch 6` and `switch %val` go to the default dest `%bb1`.

Alive2: https://alive2.llvm.org/ce/z/uSTj6w
Godbolt: https://godbolt.org/z/MGrG84bzr

This patch will benefit many rust applications and some C/C++
applications (e.g., cvc5).
---
 .../InstCombine/InstructionCombining.cpp      |  42 +++++
 .../Transforms/InstCombine/switch-select.ll   | 159 ++++++++++++++++++
 2 files changed, 201 insertions(+)
 create mode 100644 llvm/test/Transforms/InstCombine/switch-select.ll

diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 4cfafa7ac8006..4c00f2a0ea176 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -3572,6 +3572,38 @@ Instruction *InstCombinerImpl::visitBranchInst(BranchInst &BI) {
   return nullptr;
 }
 
+// Replaces (switch (select cond, X, C)/(select cond, C, X)) with (switch X) if
+// we can prove that both (switch C) and (switch X) go to the default when cond
+// is false/true.
+static Value *simplifySwitchOnSelectUsingRanges(SwitchInst &SI,
+                                                SelectInst *Select,
+                                                bool IsTrueArm) {
+  unsigned CstOpIdx = IsTrueArm ? 1 : 2;
+  auto *C = dyn_cast<ConstantInt>(Select->getOperand(CstOpIdx));
+  if (!C)
+    return nullptr;
+
+  BasicBlock *CstBB = SI.findCaseValue(C)->getCaseSuccessor();
+  if (CstBB != SI.getDefaultDest())
+    return nullptr;
+  Value *X = Select->getOperand(3 - CstOpIdx);
+  ICmpInst::Predicate Pred;
+  const APInt *RHSC;
+  if (!match(Select->getCondition(),
+             m_ICmp(Pred, m_Specific(X), m_APInt(RHSC))))
+    return nullptr;
+  if (IsTrueArm)
+    Pred = ICmpInst::getInversePredicate(Pred);
+
+  // See whether we can replace the select with X
+  ConstantRange CR = ConstantRange::makeExactICmpRegion(Pred, *RHSC);
+  for (auto Case : SI.cases())
+    if (!CR.contains(Case.getCaseValue()->getValue()))
+      return nullptr;
+
+  return X;
+}
+
 Instruction *InstCombinerImpl::visitSwitchInst(SwitchInst &SI) {
   Value *Cond = SI.getCondition();
   Value *Op0;
@@ -3645,6 +3677,16 @@ Instruction *InstCombinerImpl::visitSwitchInst(SwitchInst &SI) {
     }
   }
 
+  // Fold switch(select cond, X, Y) into switch(X/Y) if possible
+  if (auto *Select = dyn_cast<SelectInst>(Cond)) {
+    if (Value *V =
+            simplifySwitchOnSelectUsingRanges(SI, Select, /*IsTrueArm=*/true))
+      return replaceOperand(SI, 0, V);
+    if (Value *V =
+            simplifySwitchOnSelectUsingRanges(SI, Select, /*IsTrueArm=*/false))
+      return replaceOperand(SI, 0, V);
+  }
+
   KnownBits Known = computeKnownBits(Cond, 0, &SI);
   unsigned LeadingKnownZeros = Known.countMinLeadingZeros();
   unsigned LeadingKnownOnes = Known.countMinLeadingOnes();
diff --git a/llvm/test/Transforms/InstCombine/switch-select.ll b/llvm/test/Transforms/InstCombine/switch-select.ll
new file mode 100644
index 0000000000000..60757c5d22527
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/switch-select.ll
@@ -0,0 +1,159 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+
+define void @test_ult_rhsc(i8 %x) {
+; CHECK-LABEL: define void @test_ult_rhsc(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT:    switch i8 [[X]], label [[BB1:%.*]] [
+; CHECK-NEXT:      i8 2, label [[BB2:%.*]]
+; CHECK-NEXT:      i8 12, label [[BB3:%.*]]
+; CHECK-NEXT:    ]
+; CHECK:       bb1:
+; CHECK-NEXT:    call void @func1()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb2:
+; CHECK-NEXT:    call void @func2()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb3:
+; CHECK-NEXT:    call void @func3()
+; CHECK-NEXT:    unreachable
+;
+  %val = add nsw i8 %x, -2
+  %cmp = icmp ult i8 %val, 11
+  %cond = select i1 %cmp, i8 %val, i8 6
+  switch i8 %cond, label %bb1 [
+  i8 0, label %bb2
+  i8 10, label %bb3
+  ]
+
+bb1:
+  call void @func1()
+  unreachable
+bb2:
+  call void @func2()
+  unreachable
+bb3:
+  call void @func3()
+  unreachable
+}
+
+define void @test_eq_lhsc(i8 %x) {
+; CHECK-LABEL: define void @test_eq_lhsc(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT:    switch i8 [[X]], label [[BB1:%.*]] [
+; CHECK-NEXT:      i8 0, label [[BB2:%.*]]
+; CHECK-NEXT:      i8 10, label [[BB3:%.*]]
+; CHECK-NEXT:    ]
+; CHECK:       bb1:
+; CHECK-NEXT:    call void @func1()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb2:
+; CHECK-NEXT:    call void @func2()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb3:
+; CHECK-NEXT:    call void @func3()
+; CHECK-NEXT:    unreachable
+;
+  %cmp = icmp eq i8 %x, 4
+  %cond = select i1 %cmp, i8 6, i8 %x
+  switch i8 %cond, label %bb1 [
+  i8 0, label %bb2
+  i8 10, label %bb3
+  ]
+
+bb1:
+  call void @func1()
+  unreachable
+bb2:
+  call void @func2()
+  unreachable
+bb3:
+  call void @func3()
+  unreachable
+}
+
+define void @test_ult_rhsc_invalid_cond(i8 %x, i8 %y) {
+; CHECK-LABEL: define void @test_ult_rhsc_invalid_cond(
+; CHECK-SAME: i8 [[X:%.*]], i8 [[Y:%.*]]) {
+; CHECK-NEXT:    [[VAL:%.*]] = add nsw i8 [[X]], -2
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[Y]], 11
+; CHECK-NEXT:    [[COND:%.*]] = select i1 [[CMP]], i8 [[VAL]], i8 6
+; CHECK-NEXT:    switch i8 [[COND]], label [[BB1:%.*]] [
+; CHECK-NEXT:      i8 0, label [[BB2:%.*]]
+; CHECK-NEXT:      i8 10, label [[BB3:%.*]]
+; CHECK-NEXT:      i8 13, label [[BB3]]
+; CHECK-NEXT:    ]
+; CHECK:       bb1:
+; CHECK-NEXT:    call void @func1()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb2:
+; CHECK-NEXT:    call void @func2()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb3:
+; CHECK-NEXT:    call void @func3()
+; CHECK-NEXT:    unreachable
+;
+  %val = add nsw i8 %x, -2
+  %cmp = icmp ult i8 %y, 11
+  %cond = select i1 %cmp, i8 %val, i8 6
+  switch i8 %cond, label %bb1 [
+  i8 0, label %bb2
+  i8 10, label %bb3
+  i8 13, label %bb3
+  ]
+
+bb1:
+  call void @func1()
+  unreachable
+bb2:
+  call void @func2()
+  unreachable
+bb3:
+  call void @func3()
+  unreachable
+}
+
+define void @test_ult_rhsc_fail(i8 %x) {
+; CHECK-LABEL: define void @test_ult_rhsc_fail(
+; CHECK-SAME: i8 [[X:%.*]]) {
+; CHECK-NEXT:    [[VAL:%.*]] = add nsw i8 [[X]], -2
+; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i8 [[VAL]], 11
+; CHECK-NEXT:    [[COND:%.*]] = select i1 [[CMP]], i8 [[VAL]], i8 6
+; CHECK-NEXT:    switch i8 [[COND]], label [[BB1:%.*]] [
+; CHECK-NEXT:      i8 0, label [[BB2:%.*]]
+; CHECK-NEXT:      i8 10, label [[BB3:%.*]]
+; CHECK-NEXT:      i8 13, label [[BB3]]
+; CHECK-NEXT:    ]
+; CHECK:       bb1:
+; CHECK-NEXT:    call void @func1()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb2:
+; CHECK-NEXT:    call void @func2()
+; CHECK-NEXT:    unreachable
+; CHECK:       bb3:
+; CHECK-NEXT:    call void @func3()
+; CHECK-NEXT:    unreachable
+;
+  %val = add nsw i8 %x, -2
+  %cmp = icmp ult i8 %val, 11
+  %cond = select i1 %cmp, i8 %val, i8 6
+  switch i8 %cond, label %bb1 [
+  i8 0, label %bb2
+  i8 10, label %bb3
+  i8 13, label %bb3
+  ]
+
+bb1:
+  call void @func1()
+  unreachable
+bb2:
+  call void @func2()
+  unreachable
+bb3:
+  call void @func3()
+  unreachable
+}
+
+declare void @func1()
+declare void @func2()
+declare void @func3()

From 28d28d58bedf413671859b959b7ba06bccc5066d Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Mon, 15 Apr 2024 09:41:08 +0100
Subject: [PATCH 396/886] [AArch64][GlobalISel] Extend scalar lrint
 legalization. (#88360)

This extends the legalization of lrint, adding libcall support for
fp128. The old vector legal types were removed as they were not being
properly handled (vector lrint is a fairly new concept as far as I
understand). They can be re-added properly in a followup.
---
 .../CodeGen/GlobalISel/LegalizerHelper.cpp    | 22 +++++++++++++++++++
 .../AArch64/GISel/AArch64LegalizerInfo.cpp    | 20 +++--------------
 .../AArch64/GISel/AArch64RegisterBankInfo.cpp |  1 +
 .../GlobalISel/legalizer-info-validation.mir  |  4 ++--
 llvm/test/CodeGen/AArch64/lrint-conv-fp16.ll  |  2 ++
 llvm/test/CodeGen/AArch64/lrint-conv.ll       |  5 +----
 llvm/test/CodeGen/AArch64/vector-lrint.ll     |  3 +--
 7 files changed, 32 insertions(+), 25 deletions(-)

diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 95c6a359e52ec..20a7fa2535199 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -472,6 +472,8 @@ static RTLIB::Libcall getRTLibDesc(unsigned Opcode, unsigned Size) {
     RTLIBCASE(NEARBYINT_F);
   case TargetOpcode::G_INTRINSIC_ROUNDEVEN:
     RTLIBCASE(ROUNDEVEN_F);
+  case TargetOpcode::G_INTRINSIC_LRINT:
+    RTLIBCASE(LRINT_F);
   }
   llvm_unreachable("Unknown libcall function");
 }
@@ -1059,6 +1061,25 @@ LegalizerHelper::libcall(MachineInstr &MI, LostDebugLocObserver &LocObserver) {
       return Status;
     break;
   }
+  case TargetOpcode::G_INTRINSIC_LRINT: {
+    LLT LLTy = MRI.getType(MI.getOperand(1).getReg());
+    unsigned Size = LLTy.getSizeInBits();
+    Type *HLTy = getFloatTypeForLLT(Ctx, LLTy);
+    Type *ITy = IntegerType::get(
+        Ctx, MRI.getType(MI.getOperand(0).getReg()).getSizeInBits());
+    if (!HLTy || (Size != 32 && Size != 64 && Size != 80 && Size != 128)) {
+      LLVM_DEBUG(dbgs() << "No libcall available for type " << LLTy << ".\n");
+      return UnableToLegalize;
+    }
+    auto Libcall = getRTLibDesc(MI.getOpcode(), Size);
+    LegalizeResult Status =
+        createLibcall(MIRBuilder, Libcall, {MI.getOperand(0).getReg(), ITy, 0},
+                      {{MI.getOperand(1).getReg(), HLTy, 0}}, LocObserver, &MI);
+    if (Status != Legalized)
+      return Status;
+    MI.eraseFromParent();
+    return Legalized;
+  }
   case TargetOpcode::G_FPOWI: {
     LLT LLTy = MRI.getType(MI.getOperand(0).getReg());
     unsigned Size = LLTy.getSizeInBits();
@@ -2639,6 +2660,7 @@ LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
 
   case TargetOpcode::G_FPTOSI:
   case TargetOpcode::G_FPTOUI:
+  case TargetOpcode::G_INTRINSIC_LRINT:
   case TargetOpcode::G_IS_FPCLASS:
     Observer.changingInstr(MI);
 
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
index 96ded69905f7c..661ea151d1a0c 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
@@ -263,23 +263,9 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
       .scalarize(0);
 
   getActionDefinitionsBuilder(G_INTRINSIC_LRINT)
-      // If we don't have full FP16 support, then scalarize the elements of
-      // vectors containing fp16 types.
-      .fewerElementsIf(
-          [=, &ST](const LegalityQuery &Query) {
-            const auto &Ty = Query.Types[0];
-            return Ty.isVector() && Ty.getElementType() == s16 &&
-                   !ST.hasFullFP16();
-          },
-          [=](const LegalityQuery &Query) { return std::make_pair(0, s16); })
-      // If we don't have full FP16 support, then widen s16 to s32 if we
-      // encounter it.
-      .widenScalarIf(
-          [=, &ST](const LegalityQuery &Query) {
-            return Query.Types[0] == s16 && !ST.hasFullFP16();
-          },
-          [=](const LegalityQuery &Query) { return std::make_pair(0, s32); })
-      .legalFor({s16, s32, s64, v2s32, v4s32, v2s64, v2s16, v4s16, v8s16});
+      .legalFor({{s64, MinFPScalar}, {s64, s32}, {s64, s64}})
+      .libcallFor({{s64, s128}})
+      .minScalarOrElt(1, MinFPScalar);
 
   getActionDefinitionsBuilder(
       {G_FCOS, G_FSIN, G_FPOW, G_FLOG, G_FLOG2, G_FLOG10,
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
index 58d000b6b2a9e..d39de770eaf16 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
@@ -829,6 +829,7 @@ AArch64RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   }
   case TargetOpcode::G_FPTOSI:
   case TargetOpcode::G_FPTOUI:
+  case TargetOpcode::G_INTRINSIC_LRINT:
     if (MRI.getType(MI.getOperand(0).getReg()).isVector())
       break;
     OpRegBankIdx = {PMI_FirstGPR, PMI_FirstFPR};
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir b/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
index 200e9d19d58d2..0793f3983c8e5 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
@@ -154,8 +154,8 @@
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
 # DEBUG-NEXT: G_INTRINSIC_LRINT (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
-# DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
-# DEBUG-NEXT: .. imm index coverage check SKIPPED: user-defined predicate detected
+# DEBUG-NEXT: .. the first uncovered type index: 2, OK
+# DEBUG-NEXT: .. the first uncovered imm index: 0, OK
 # DEBUG-NEXT: G_INTRINSIC_ROUNDEVEN (opcode {{[0-9]+}}): 1 type index, 0 imm indices
 # DEBUG-NEXT: .. opcode {{[0-9]+}} is aliased to {{[0-9]+}}
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
diff --git a/llvm/test/CodeGen/AArch64/lrint-conv-fp16.ll b/llvm/test/CodeGen/AArch64/lrint-conv-fp16.ll
index 7557ceac1212f..6771a862c386f 100644
--- a/llvm/test/CodeGen/AArch64/lrint-conv-fp16.ll
+++ b/llvm/test/CodeGen/AArch64/lrint-conv-fp16.ll
@@ -1,6 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
 ; RUN: llc < %s -mtriple=aarch64 | FileCheck %s --check-prefixes=CHECK-NOFP16
 ; RUN: llc < %s -mtriple=aarch64 -mattr=+fullfp16 | FileCheck %s --check-prefixes=CHECK-FP16
+; RUN: llc < %s -mtriple=aarch64 -global-isel | FileCheck %s --check-prefixes=CHECK-NOFP16
+; RUN: llc < %s -mtriple=aarch64 -mattr=+fullfp16 -global-isel | FileCheck %s --check-prefixes=CHECK-FP16
 
 define i16 @testmhhs(half %x) {
 ; CHECK-NOFP16-LABEL: testmhhs:
diff --git a/llvm/test/CodeGen/AArch64/lrint-conv.ll b/llvm/test/CodeGen/AArch64/lrint-conv.ll
index b61d6f04b400e..60393b4f18d9d 100644
--- a/llvm/test/CodeGen/AArch64/lrint-conv.ll
+++ b/llvm/test/CodeGen/AArch64/lrint-conv.ll
@@ -1,9 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
 ; RUN: llc < %s -mtriple=aarch64 | FileCheck %s
-; RUN: llc < %s -mtriple=aarch64 -global-isel -global-isel-abort=2 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:       warning: Instruction selection used fallback path for testmswl
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for testmsll
+; RUN: llc < %s -mtriple=aarch64 -global-isel | FileCheck %s
 
 define i32 @testmsws(float %x) {
 ; CHECK-LABEL: testmsws:
diff --git a/llvm/test/CodeGen/AArch64/vector-lrint.ll b/llvm/test/CodeGen/AArch64/vector-lrint.ll
index b7fcd11ba8d16..db85b23428216 100644
--- a/llvm/test/CodeGen/AArch64/vector-lrint.ll
+++ b/llvm/test/CodeGen/AArch64/vector-lrint.ll
@@ -2,8 +2,7 @@
 ; RUN: llc < %s -mtriple=aarch64 | FileCheck %s --check-prefixes=CHECK,CHECK-SD
 ; RUN: llc < %s -mtriple=aarch64 -global-isel -global-isel-abort=2 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
-; CHECK-GI:       warning: Instruction selection used fallback path for lrint_v1f16
-; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v2f16
+; CHECK-GI:       warning: Instruction selection used fallback path for lrint_v2f16
 ; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v4f16
 ; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v8f16
 ; CHECK-GI-NEXT:  warning: Instruction selection used fallback path for lrint_v16i64_v16f16

From 36230f90eef455b6af867ca7ad4ed25d00ad5d3b Mon Sep 17 00:00:00 2001
From: fengfeng <153487255+fengfeng09@users.noreply.github.com>
Date: Mon, 15 Apr 2024 17:01:15 +0800
Subject: [PATCH 397/886] [SelectionDAG] Propagate Disjoint flag. (#88370)

Signed-off-by: feng.feng <feng.feng@iluvatar.com>
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp     |  3 ++-
 llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp  |  3 ++-
 .../CodeGen/X86/propagate-disjoint-in-shl-or.ll   | 15 +++++++++++++++
 3 files changed, 19 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/CodeGen/X86/propagate-disjoint-in-shl-or.ll

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index d9c6c28d5dac9..dbbe8d9193c25 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -9530,7 +9530,8 @@ static SDValue combineShiftOfShiftedLogic(SDNode *Shift, SelectionDAG &DAG) {
   SDValue ShiftSumC = DAG.getConstant(*C0Val + C1Val, DL, ShiftAmtVT);
   SDValue NewShift1 = DAG.getNode(ShiftOpcode, DL, VT, X, ShiftSumC);
   SDValue NewShift2 = DAG.getNode(ShiftOpcode, DL, VT, Y, C1);
-  return DAG.getNode(LogicOpcode, DL, VT, NewShift1, NewShift2);
+  return DAG.getNode(LogicOpcode, DL, VT, NewShift1, NewShift2,
+                     LogicOp->getFlags());
 }
 
 /// Handle transforms common to the three shifts, when the shift amount is a
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 4939976b8e287..c938b3996be39 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -544,7 +544,8 @@ bool TargetLowering::ShrinkDemandedConstant(SDValue Op,
     if (!C.isSubsetOf(DemandedBits)) {
       EVT VT = Op.getValueType();
       SDValue NewC = TLO.DAG.getConstant(DemandedBits & C, DL, VT);
-      SDValue NewOp = TLO.DAG.getNode(Opcode, DL, VT, Op.getOperand(0), NewC);
+      SDValue NewOp = TLO.DAG.getNode(Opcode, DL, VT, Op.getOperand(0), NewC,
+                                      Op->getFlags());
       return TLO.CombineTo(Op, NewOp);
     }
 
diff --git a/llvm/test/CodeGen/X86/propagate-disjoint-in-shl-or.ll b/llvm/test/CodeGen/X86/propagate-disjoint-in-shl-or.ll
new file mode 100644
index 0000000000000..e38840f3e4610
--- /dev/null
+++ b/llvm/test/CodeGen/X86/propagate-disjoint-in-shl-or.ll
@@ -0,0 +1,15 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=x86_64 %s -start-before=x86-isel -o - | FileCheck %s
+
+define i32 @add_shl_or_disjoint(i32 %x) {
+; CHECK-LABEL: add_shl_or_disjoint:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    # kill: def $edi killed $edi def $rdi
+; CHECK-NEXT:    leal 165036822(,%rdi,4), %eax
+; CHECK-NEXT:    retq
+  %or = or disjoint i32 %x, 4027584529
+  %shl = shl i32 %or, 2
+  %add = add i32 %shl, 1234567890
+  ret i32 %add
+}
+

From 2347020e4ce62ee7c7378824dc6ad2b94917a638 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?D=C3=A1vid=20Ferenc=20Szab=C3=B3?=
 <30732159+dfszabo@users.noreply.github.com>
Date: Mon, 15 Apr 2024 11:01:55 +0200
Subject: [PATCH 398/886] [GlobalISel] Fix fewerElementsVectorPhi to insert
 after G_PHIs (#87927)

Currently the inserted mergelike instructions will be inserted at the
location of the G_PHI. Seems like the behaviour was correct before, but
the rework done in https://reviews.llvm.org/D114198 forgot to include
the part which makes sure the instructions will be inserted after all
the G_PHIs.
---
 llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp           | 4 ++++
 llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp | 6 +++---
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 20a7fa2535199..156353296cfc1 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -4287,6 +4287,10 @@ LegalizerHelper::fewerElementsVectorPhi(GenericMachineInstr &MI,
     }
   }
 
+  // Set the insert point after the existing PHIs
+  MachineBasicBlock &MBB = *MI.getParent();
+  MIRBuilder.setInsertPt(MBB, MBB.getFirstNonPHI());
+
   // Merge small outputs into MI's def.
   if (NumLeftovers) {
     mergeMixedSubvectors(MI.getReg(0), OutputRegs);
diff --git a/llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp b/llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp
index 33155d2c9a964..831d7e6292e33 100644
--- a/llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp
+++ b/llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp
@@ -1556,12 +1556,12 @@ TEST_F(AArch64GISelMITest, FewerElementsPhi) {
   CHECK: [[PHI0:%[0-9]+]]:_(<2 x s32>) = G_PHI [[INITVAL_E01]]:_(<2 x s32>), %bb.0, [[MIDVAL_E01]]:_(<2 x s32>), %bb.1
   CHECK: [[PHI1:%[0-9]+]]:_(<2 x s32>) = G_PHI [[INITVAL_E23]]:_(<2 x s32>), %bb.0, [[MIDVAL_E23]]:_(<2 x s32>), %bb.1
   CHECK: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[INITVAL_E4]]:_(s32), %bb.0, [[MIDVAL_E4]]:_(s32), %bb.1
-  CHECK: [[UNMERGE0:%[0-9]+]]:_(s32), [[UNMERGE1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI0]]:_(<2 x s32>)
-  CHECK: [[UNMERGE2:%[0-9]+]]:_(s32), [[UNMERGE3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI1]]:_(<2 x s32>)
-  CHECK: [[BV:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UNMERGE0]]:_(s32), [[UNMERGE1]]:_(s32), [[UNMERGE2]]:_(s32), [[UNMERGE3]]:_(s32), [[PHI2]]:_(s32)
 
   CHECK: [[OTHER_PHI:%[0-9]+]]:_(s64) = G_PHI
 
+  CHECK: [[UNMERGE0:%[0-9]+]]:_(s32), [[UNMERGE1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI0]]:_(<2 x s32>)
+  CHECK: [[UNMERGE2:%[0-9]+]]:_(s32), [[UNMERGE3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI1]]:_(<2 x s32>)
+  CHECK: [[BV:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UNMERGE0]]:_(s32), [[UNMERGE1]]:_(s32), [[UNMERGE2]]:_(s32), [[UNMERGE3]]:_(s32), [[PHI2]]:_(s32)
   CHECK: [[USE_OP:%[0-9]+]]:_(<5 x s32>) = G_AND [[BV]]:_, [[BV]]:_
   )";
 

From aa2741449c3609b2ae244d8d3f3e14ad16de72e4 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Mon, 15 Apr 2024 16:56:44 +0800
Subject: [PATCH 399/886] Revert "[C++20] [Modules] Don't import non-inline
 function bodies even if it is marked as always_inline"

This reverts commit 1ecbab56dcbb78268c8d19af34a50591f90b12a0.

See the discussion in https://github.com/llvm/llvm-project/issues/86893.

The original commit receives too many complaints. Let's try to
workaround the issue to give better user experiences.
---
 clang/lib/CodeGen/CodeGenModule.cpp                  | 3 ++-
 clang/test/CodeGenCXX/module-funcs-from-imports.cppm | 6 +++---
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 73a9cb9d6e042..e44749672d582 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -3952,7 +3952,8 @@ bool CodeGenModule::shouldEmitFunction(GlobalDecl GD) {
   // behavior may break ABI compatibility of the current unit.
   if (const Module *M = F->getOwningModule();
       M && M->getTopLevelModule()->isNamedModule() &&
-      getContext().getCurrentNamedModule() != M->getTopLevelModule())
+      getContext().getCurrentNamedModule() != M->getTopLevelModule() &&
+      !F->hasAttr<AlwaysInlineAttr>())
     return false;
 
   if (F->hasAttr<NoInlineAttr>())
diff --git a/clang/test/CodeGenCXX/module-funcs-from-imports.cppm b/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
index 8d04328eaf3fe..33cdf437110a9 100644
--- a/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
+++ b/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
@@ -53,11 +53,11 @@ int use() {
     return exported_func() + always_inline_func();
 }
 
-// Checks that none of the function in the importees
+// Checks that none of the function (except the always_inline_func) in the importees
 // are generated in the importer's code.
 // CHECK-O0: define{{.*}}_Z3usev(
 // CHECK-O0: declare{{.*}}_ZW1M13exported_funcv(
-// CHECK-O0: declare{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O0: define{{.*}}available_externally{{.*}}_ZW1M18always_inline_funcv(
 // CHECK-O0-NOT: func_in_gmf
 // CHECK-O0-NOT: func_in_gmf_not_called
 // CHECK-O0-NOT: non_exported_func
@@ -68,7 +68,7 @@ int use() {
 // O0 to keep consistent ABI.
 // CHECK-O1: define{{.*}}_Z3usev(
 // CHECK-O1: declare{{.*}}_ZW1M13exported_funcv(
-// CHECK-O1: declare{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O1: define{{.*}}available_externally{{.*}}_ZW1M18always_inline_funcv(
 // CHECK-O1-NOT: func_in_gmf
 // CHECK-O1-NOT: func_in_gmf_not_called
 // CHECK-O1-NOT: non_exported_func

From 7740e933c41f07444ead12fd0417c5e53c6a4c1a Mon Sep 17 00:00:00 2001
From: abidh <haqadeer@amd.com>
Date: Mon, 15 Apr 2024 10:18:31 +0100
Subject: [PATCH 400/886] [flang] Rename AddDebugFoundation to AddDebugInfo
 (NFC) (#88526)

As discussed in PR 86939, this PR renames the pass and updates the
references. The actual changes for debug info will come in separate PRs.
---
 flang/include/flang/Optimizer/Transforms/Passes.h |  2 +-
 .../include/flang/Optimizer/Transforms/Passes.td  |  6 +++---
 flang/include/flang/Tools/CLOptions.inc           |  8 ++++----
 .../{AddDebugFoundation.cpp => AddDebugInfo.cpp}  | 15 +++++++--------
 flang/lib/Optimizer/Transforms/CMakeLists.txt     |  2 +-
 flang/test/Driver/mlir-debug-pass-pipeline.f90    |  4 ++--
 .../test/Transforms/debug-line-table-existing.fir |  2 +-
 .../test/Transforms/debug-line-table-inc-file.fir |  2 +-
 .../Transforms/debug-line-table-inc-same-file.fir |  2 +-
 flang/test/Transforms/debug-line-table.fir        |  2 +-
 10 files changed, 22 insertions(+), 23 deletions(-)
 rename flang/lib/Optimizer/Transforms/{AddDebugFoundation.cpp => AddDebugInfo.cpp} (90%)

diff --git a/flang/include/flang/Optimizer/Transforms/Passes.h b/flang/include/flang/Optimizer/Transforms/Passes.h
index a60d39c8df08d..d8840d9e967b4 100644
--- a/flang/include/flang/Optimizer/Transforms/Passes.h
+++ b/flang/include/flang/Optimizer/Transforms/Passes.h
@@ -67,7 +67,7 @@ std::unique_ptr<mlir::Pass> createMemoryAllocationPass();
 std::unique_ptr<mlir::Pass> createStackArraysPass();
 std::unique_ptr<mlir::Pass> createAliasTagsPass();
 std::unique_ptr<mlir::Pass> createSimplifyIntrinsicsPass();
-std::unique_ptr<mlir::Pass> createAddDebugFoundationPass();
+std::unique_ptr<mlir::Pass> createAddDebugInfoPass();
 std::unique_ptr<mlir::Pass> createLoopVersioningPass();
 
 std::unique_ptr<mlir::Pass>
diff --git a/flang/include/flang/Optimizer/Transforms/Passes.td b/flang/include/flang/Optimizer/Transforms/Passes.td
index 978911c21794c..187796d77cf5c 100644
--- a/flang/include/flang/Optimizer/Transforms/Passes.td
+++ b/flang/include/flang/Optimizer/Transforms/Passes.td
@@ -201,12 +201,12 @@ def MemRefDataFlowOpt : Pass<"fir-memref-dataflow-opt", "::mlir::func::FuncOp">
 
 // This needs to be a "mlir::ModuleOp" pass, because we are creating debug for
 // the module in this pass.
-def AddDebugFoundation : Pass<"add-debug-foundation", "mlir::ModuleOp"> {
-  let summary = "Add the foundation for debug info";
+def AddDebugInfo : Pass<"add-debug-info", "mlir::ModuleOp"> {
+  let summary = "Add the debug info";
   let description = [{
     Add the foundation for emitting debug info that can be understood by llvm.
   }];
-  let constructor = "::fir::createAddDebugFoundationPass()";
+  let constructor = "::fir::createAddDebugInfoPass()";
   let dependentDialects = [
     "fir::FIROpsDialect", "mlir::func::FuncDialect", "mlir::LLVM::LLVMDialect"
   ];
diff --git a/flang/include/flang/Tools/CLOptions.inc b/flang/include/flang/Tools/CLOptions.inc
index bc6df2c830696..268d00b5a6053 100644
--- a/flang/include/flang/Tools/CLOptions.inc
+++ b/flang/include/flang/Tools/CLOptions.inc
@@ -155,9 +155,9 @@ inline void addTargetRewritePass(mlir::PassManager &pm) {
   });
 }
 
-inline void addDebugFoundationPass(mlir::PassManager &pm) {
+inline void addDebugInfoPass(mlir::PassManager &pm) {
   addPassConditionally(pm, disableDebugFoundation,
-      [&]() { return fir::createAddDebugFoundationPass(); });
+      [&]() { return fir::createAddDebugInfoPass(); });
 }
 
 inline void addFIRToLLVMPass(
@@ -288,7 +288,7 @@ inline void createDebugPasses(
   // Currently only -g1, -g, -gline-tables-only supported
   switch (debugLevel) {
   case llvm::codegenoptions::DebugLineTablesOnly:
-    addDebugFoundationPass(pm);
+    addDebugInfoPass(pm);
     return;
   case llvm::codegenoptions::NoDebugInfo:
     return;
@@ -296,7 +296,7 @@ inline void createDebugPasses(
     // TODO: Add cases and passes for other debug options.
     // All other debug options not implemented yet, currently emits warning
     // and generates as much debug information as possible.
-    addDebugFoundationPass(pm);
+    addDebugInfoPass(pm);
     return;
   }
 }
diff --git a/flang/lib/Optimizer/Transforms/AddDebugFoundation.cpp b/flang/lib/Optimizer/Transforms/AddDebugInfo.cpp
similarity index 90%
rename from flang/lib/Optimizer/Transforms/AddDebugFoundation.cpp
rename to flang/lib/Optimizer/Transforms/AddDebugInfo.cpp
index 678fbf6a7d23b..b719f677230ff 100644
--- a/flang/lib/Optimizer/Transforms/AddDebugFoundation.cpp
+++ b/flang/lib/Optimizer/Transforms/AddDebugInfo.cpp
@@ -1,4 +1,4 @@
-//===- AddDebugFoundation.cpp -- add basic debug linetable info -----------===//
+//===-------------- AddDebugInfo.cpp -- add debug info -------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -32,8 +32,8 @@
 #include "llvm/Support/raw_ostream.h"
 
 namespace fir {
-#define GEN_PASS_DEF_ADDDEBUGFOUNDATION
-#define GEN_PASS_DECL_ADDDEBUGFOUNDATION
+#define GEN_PASS_DEF_ADDDEBUGINFO
+#define GEN_PASS_DECL_ADDDEBUGINFO
 #include "flang/Optimizer/Transforms/Passes.h.inc"
 } // namespace fir
 
@@ -41,15 +41,14 @@ namespace fir {
 
 namespace {
 
-class AddDebugFoundationPass
-    : public fir::impl::AddDebugFoundationBase<AddDebugFoundationPass> {
+class AddDebugInfoPass : public fir::impl::AddDebugInfoBase<AddDebugInfoPass> {
 public:
   void runOnOperation() override;
 };
 
 } // namespace
 
-void AddDebugFoundationPass::runOnOperation() {
+void AddDebugInfoPass::runOnOperation() {
   mlir::ModuleOp module = getOperation();
   mlir::MLIRContext *context = &getContext();
   mlir::OpBuilder builder(context);
@@ -113,6 +112,6 @@ void AddDebugFoundationPass::runOnOperation() {
   });
 }
 
-std::unique_ptr<mlir::Pass> fir::createAddDebugFoundationPass() {
-  return std::make_unique<AddDebugFoundationPass>();
+std::unique_ptr<mlir::Pass> fir::createAddDebugInfoPass() {
+  return std::make_unique<AddDebugInfoPass>();
 }
diff --git a/flang/lib/Optimizer/Transforms/CMakeLists.txt b/flang/lib/Optimizer/Transforms/CMakeLists.txt
index ba2e267996150..d55655c53906e 100644
--- a/flang/lib/Optimizer/Transforms/CMakeLists.txt
+++ b/flang/lib/Optimizer/Transforms/CMakeLists.txt
@@ -14,7 +14,7 @@ add_flang_library(FIRTransforms
   SimplifyRegionLite.cpp
   AlgebraicSimplification.cpp
   SimplifyIntrinsics.cpp
-  AddDebugFoundation.cpp
+  AddDebugInfo.cpp
   PolymorphicOpConversion.cpp
   LoopVersioning.cpp
   OMPDescriptorMapInfoGen.cpp
diff --git a/flang/test/Driver/mlir-debug-pass-pipeline.f90 b/flang/test/Driver/mlir-debug-pass-pipeline.f90
index e714a66b85f33..04d432f854ca3 100644
--- a/flang/test/Driver/mlir-debug-pass-pipeline.f90
+++ b/flang/test/Driver/mlir-debug-pass-pipeline.f90
@@ -82,7 +82,7 @@
 ! ALL-NEXT:   (S) 0 num-dce'd - Number of operations eliminated
 ! ALL-NEXT: TargetRewrite
 ! ALL-NEXT: ExternalNameConversion
-! DEBUG-NEXT: AddDebugFoundation
-! NO-DEBUG-NOT: AddDebugFoundation
+! DEBUG-NEXT: AddDebugInfo
+! NO-DEBUG-NOT: AddDebugInfo
 ! ALL: FIRToLLVMLowering
 ! ALL-NOT: LLVMIRLoweringPass
diff --git a/flang/test/Transforms/debug-line-table-existing.fir b/flang/test/Transforms/debug-line-table-existing.fir
index 3c81d75dbd663..534278ebc972d 100644
--- a/flang/test/Transforms/debug-line-table-existing.fir
+++ b/flang/test/Transforms/debug-line-table-existing.fir
@@ -1,5 +1,5 @@
 
-// RUN: fir-opt --add-debug-foundation --mlir-print-debuginfo %s | FileCheck %s
+// RUN: fir-opt --add-debug-info --mlir-print-debuginfo %s | FileCheck %s
 // REQUIRES: system-linux
 
 // Test that there are no changes to a function with existed fused loc debug
diff --git a/flang/test/Transforms/debug-line-table-inc-file.fir b/flang/test/Transforms/debug-line-table-inc-file.fir
index f809ab99b4727..be4f005bf664a 100644
--- a/flang/test/Transforms/debug-line-table-inc-file.fir
+++ b/flang/test/Transforms/debug-line-table-inc-file.fir
@@ -1,5 +1,5 @@
 
-// RUN: fir-opt --add-debug-foundation --mlir-print-debuginfo %s | FileCheck %s
+// RUN: fir-opt --add-debug-info --mlir-print-debuginfo %s | FileCheck %s
 // REQUIRES: system-linux
 
 // Test for included functions that have a different debug location than the current file
diff --git a/flang/test/Transforms/debug-line-table-inc-same-file.fir b/flang/test/Transforms/debug-line-table-inc-same-file.fir
index 3ca00e68768c5..4836f2e21dd9d 100644
--- a/flang/test/Transforms/debug-line-table-inc-same-file.fir
+++ b/flang/test/Transforms/debug-line-table-inc-same-file.fir
@@ -1,5 +1,5 @@
 
-// RUN: fir-opt --add-debug-foundation --mlir-print-debuginfo %s | FileCheck %s
+// RUN: fir-opt --add-debug-info --mlir-print-debuginfo %s | FileCheck %s
 // REQUIRES: system-linux
 
 // Test that there is only one FileAttribute generated for multiple functions
diff --git a/flang/test/Transforms/debug-line-table.fir b/flang/test/Transforms/debug-line-table.fir
index 8e66fc1ab3988..0ba88d3d9f7fa 100644
--- a/flang/test/Transforms/debug-line-table.fir
+++ b/flang/test/Transforms/debug-line-table.fir
@@ -1,5 +1,5 @@
 
-// RUN: fir-opt --add-debug-foundation --mlir-print-debuginfo %s | FileCheck %s
+// RUN: fir-opt --add-debug-info --mlir-print-debuginfo %s | FileCheck %s
 
 module attributes { fir.defaultkind = "a1c4d8i4l4r4", fir.kindmap = "", llvm.data_layout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128", llvm.target_triple = "aarch64-unknown-linux-gnu"} {
   func.func @_QPsb() {

From 8ddaf750746d7f9b5f7e878870b086edc0f55326 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Mon, 15 Apr 2024 11:24:00 +0200
Subject: [PATCH 401/886] Fix rsqrt inaccuracies. (#88691)

The current lowering has issues with large/subnormal values. This ports
XLA's lowering and was verified using XLA's test suite and the
MLIR-based emitters.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 88 ++++++++++++++++---
 .../convert-to-standard.mlir                  | 17 +++-
 2 files changed, 92 insertions(+), 13 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 49eb575212ffc..3ebee9baff31b 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -27,9 +27,11 @@ using namespace mlir;
 
 namespace {
 
-// Returns the absolute value or its square root.
+enum class AbsFn { abs, sqrt, rsqrt };
+
+// Returns the absolute value, its square root or its reciprocal square root.
 Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
-                 ImplicitLocOpBuilder &b, bool returnSqrt = false) {
+                 ImplicitLocOpBuilder &b, AbsFn fn = AbsFn::abs) {
   Value one = b.create<arith::ConstantOp>(real.getType(),
                                           b.getFloatAttr(real.getType(), 1.0));
 
@@ -43,7 +45,13 @@ Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
   Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
   Value result;
 
-  if (returnSqrt) {
+  if (fn == AbsFn::rsqrt) {
+    ratioSqPlusOne = b.create<math::RsqrtOp>(ratioSqPlusOne, fmf);
+    min = b.create<math::RsqrtOp>(min, fmf);
+    max = b.create<math::RsqrtOp>(max, fmf);
+  }
+
+  if (fn == AbsFn::sqrt) {
     Value quarter = b.create<arith::ConstantOp>(
         real.getType(), b.getFloatAttr(real.getType(), 0.25));
     // sqrt(sqrt(a*b)) would avoid the pow, but will overflow more easily.
@@ -863,7 +871,7 @@ struct SqrtOpConversion : public OpConversionPattern<complex::SqrtOp> {
 
     Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
     Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
-    Value absSqrt = computeAbs(real, imag, fmf, b, /*returnSqrt=*/true);
+    Value absSqrt = computeAbs(real, imag, fmf, b, AbsFn::sqrt);
     Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
     Value sqrtArg = b.create<arith::MulFOp>(argArg, half, fmf);
     Value cos = b.create<math::CosOp>(sqrtArg, fmf);
@@ -1147,18 +1155,74 @@ struct RsqrtOpConversion : public OpConversionPattern<complex::RsqrtOp> {
   LogicalResult
   matchAndRewrite(complex::RsqrtOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    mlir::ImplicitLocOpBuilder builder(op.getLoc(), rewriter);
+    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
 
-    Value c = builder.create<arith::ConstantOp>(
-        elementType, builder.getFloatAttr(elementType, -0.5));
-    Value d = builder.create<arith::ConstantOp>(
-        elementType, builder.getFloatAttr(elementType, 0));
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
+
+    auto cst = [&](APFloat v) {
+      return b.create<arith::ConstantOp>(elementType,
+                                         b.getFloatAttr(elementType, v));
+    };
+    const auto &floatSemantics = elementType.getFloatSemantics();
+    Value zero = cst(APFloat::getZero(floatSemantics));
+    Value inf = cst(APFloat::getInf(floatSemantics));
+    Value negHalf = b.create<arith::ConstantOp>(
+        elementType, b.getFloatAttr(elementType, -0.5));
+    Value nan = cst(APFloat::getNaN(floatSemantics));
+
+    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
+    Value absRsqrt = computeAbs(real, imag, fmf, b, AbsFn::rsqrt);
+    Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
+    Value rsqrtArg = b.create<arith::MulFOp>(argArg, negHalf, fmf);
+    Value cos = b.create<math::CosOp>(rsqrtArg, fmf);
+    Value sin = b.create<math::SinOp>(rsqrtArg, fmf);
+
+    Value resultReal = b.create<arith::MulFOp>(absRsqrt, cos, fmf);
+    Value resultImag = b.create<arith::MulFOp>(absRsqrt, sin, fmf);
+
+    if (!arith::bitEnumContainsAll(fmf, arith::FastMathFlags::nnan |
+                                            arith::FastMathFlags::ninf)) {
+      Value negOne = b.create<arith::ConstantOp>(
+          elementType, b.getFloatAttr(elementType, -1));
+
+      Value realSignedZero = b.create<math::CopySignOp>(zero, real, fmf);
+      Value imagSignedZero = b.create<math::CopySignOp>(zero, imag, fmf);
+      Value negImagSignedZero =
+          b.create<arith::MulFOp>(negOne, imagSignedZero, fmf);
 
-    rewriter.replaceOp(op,
-                       {powOpConversionImpl(builder, type, adaptor.getComplex(),
-                                            c, d, op.getFastmath())});
+      Value absReal = b.create<math::AbsFOp>(real, fmf);
+      Value absImag = b.create<math::AbsFOp>(imag, fmf);
+
+      Value absImagIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absImag, inf, fmf);
+      Value realIsNan =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, real, real, fmf);
+      Value realIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absReal, inf, fmf);
+      Value inIsNanInf = b.create<arith::AndIOp>(absImagIsInf, realIsNan);
+
+      Value resultIsZero = b.create<arith::OrIOp>(inIsNanInf, realIsInf);
+
+      resultReal =
+          b.create<arith::SelectOp>(resultIsZero, realSignedZero, resultReal);
+      resultImag = b.create<arith::SelectOp>(resultIsZero, negImagSignedZero,
+                                             resultImag);
+    }
+
+    Value isRealZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero, fmf);
+    Value isImagZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero, fmf);
+    Value isZero = b.create<arith::AndIOp>(isRealZero, isImagZero);
+
+    resultReal = b.create<arith::SelectOp>(isZero, inf, resultReal);
+    resultImag = b.create<arith::SelectOp>(isZero, nan, resultImag);
+
+    rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
+                                                   resultImag);
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index e0e7cdadd317d..8b4ea9777f797 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -837,6 +837,21 @@ func.func @complex_rsqrt(%arg: complex<f32>) -> complex<f32> {
   return %rsqrt : complex<f32>
 }
 
+// CHECK-COUNT-5: arith.select
+// CHECK-NOT: arith.select
+
+// -----
+
+// CHECK-LABEL: func @complex_rsqrt_nnan_ninf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_rsqrt_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
+  %sqrt = complex.rsqrt %arg fastmath<nnan,ninf> : complex<f32>
+  return %sqrt : complex<f32>
+}
+
+// CHECK-COUNT-3: arith.select
+// CHECK-NOT: arith.select
+
 // -----
 
 // CHECK-LABEL:   func.func @complex_angle
@@ -2103,4 +2118,4 @@ func.func @complex_tanh_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] fastmath<nnan,contract> : f32
-// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
\ No newline at end of file
+// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>

From f2923e31b64907df44ae641114228ad6e283627f Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Mon, 15 Apr 2024 10:33:54 +0100
Subject: [PATCH 402/886] [MLIR][OpenMP] Introduce the LoopWrapperInterface
 (#87232)

This patch defines a common interface to be shared by all OpenMP loop
wrapper operations. The main restrictions these operations must meet in
order to be considered a wrapper are:

- They contain a single region.
- Their region contains a single block.
- Their block only contains another loop wrapper or `omp.loop_nest` and
a terminator.

The new interface is attached to the `omp.parallel`, `omp.wsloop`,
`omp.simdloop`, `omp.distribute` and `omp.taskloop` operations. It is
not currently enforced that these operations meet the wrapper
restrictions, which would break existing OpenMP loop-generating code.
Rather, this will be introduced progressively in subsequent patches.
---
 .../mlir/Dialect/OpenMP/OpenMPInterfaces.h    |  3 +
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td | 16 +++--
 .../Dialect/OpenMP/OpenMPOpsInterfaces.td     | 67 +++++++++++++++++++
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  | 18 +++++
 mlir/test/Dialect/OpenMP/invalid.mlir         | 16 ++++-
 5 files changed, 115 insertions(+), 5 deletions(-)

diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPInterfaces.h b/mlir/include/mlir/Dialect/OpenMP/OpenMPInterfaces.h
index b3184db885216..787c48b05c5c5 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPInterfaces.h
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPInterfaces.h
@@ -21,6 +21,9 @@
 #include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 
+#define GET_OP_FWD_DEFINES
+#include "mlir/Dialect/OpenMP/OpenMPOps.h.inc"
+
 #include "mlir/Dialect/OpenMP/OpenMPOpsInterfaces.h.inc"
 
 namespace mlir::omp {
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 596b2f5e4444f..3abdbe3adfd0b 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -236,6 +236,7 @@ def PrivateClauseOp : OpenMP_Op<"private", [IsolatedFromAbove]> {
 
 def ParallelOp : OpenMP_Op<"parallel", [
                  AutomaticAllocationScope, AttrSizedOperandSegments,
+                 DeclareOpInterfaceMethods<LoopWrapperInterface>,
                  DeclareOpInterfaceMethods<OutlineableOpenMPOpInterface>,
                  RecursiveMemoryEffects, ReductionClauseInterface]> {
   let summary = "parallel construct";
@@ -530,8 +531,6 @@ def SingleOp : OpenMP_Op<"single", [AttrSizedOperandSegments]> {
 
 def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
                         AllTypesMatch<["lowerBound", "upperBound", "step"]>,
-                        ParentOneOf<["DistributeOp", "SimdLoopOp", "TaskloopOp",
-                                     "WsloopOp"]>,
                         RecursiveMemoryEffects]> {
   let summary = "rectangular loop nest";
   let description = [{
@@ -586,6 +585,10 @@ def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
 
     /// Returns the induction variables of the loop nest.
     ArrayRef<BlockArgument> getIVs() { return getRegion().getArguments(); }
+
+    /// Fills a list of wrapper operations around this loop nest. Wrappers
+    /// in the resulting vector will be sorted from innermost to outermost.
+    void gatherWrappers(SmallVectorImpl<LoopWrapperInterface> &wrappers);
   }];
 
   let hasCustomAssemblyFormat = 1;
@@ -598,6 +601,7 @@ def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
 
 def WsloopOp : OpenMP_Op<"wsloop", [AttrSizedOperandSegments,
                          AllTypesMatch<["lowerBound", "upperBound", "step"]>,
+                         DeclareOpInterfaceMethods<LoopWrapperInterface>,
                          RecursiveMemoryEffects, ReductionClauseInterface]> {
   let summary = "worksharing-loop construct";
   let description = [{
@@ -719,7 +723,9 @@ def WsloopOp : OpenMP_Op<"wsloop", [AttrSizedOperandSegments,
 //===----------------------------------------------------------------------===//
 
 def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
-                         AllTypesMatch<["lowerBound", "upperBound", "step"]>]> {
+                         AllTypesMatch<["lowerBound", "upperBound", "step"]>,
+                         DeclareOpInterfaceMethods<LoopWrapperInterface>,
+                         RecursiveMemoryEffects]> {
  let summary = "simd loop construct";
   let description = [{
     The simd construct can be applied to a loop to indicate that the loop can be
@@ -833,7 +839,8 @@ def YieldOp : OpenMP_Op<"yield",
 // Distribute construct [2.9.4.1]
 //===----------------------------------------------------------------------===//
 def DistributeOp : OpenMP_Op<"distribute", [AttrSizedOperandSegments,
-                              MemoryEffects<[MemWrite]>]> {
+                             DeclareOpInterfaceMethods<LoopWrapperInterface>,
+                             RecursiveMemoryEffects]> {
   let summary = "distribute construct";
   let description = [{
     The distribute construct specifies that the iterations of one or more loops
@@ -1011,6 +1018,7 @@ def TaskOp : OpenMP_Op<"task", [AttrSizedOperandSegments,
 def TaskloopOp : OpenMP_Op<"taskloop", [AttrSizedOperandSegments,
                            AutomaticAllocationScope, RecursiveMemoryEffects,
                            AllTypesMatch<["lowerBound", "upperBound", "step"]>,
+                           DeclareOpInterfaceMethods<LoopWrapperInterface>,
                            ReductionClauseInterface]> {
   let summary = "taskloop construct";
   let description = [{
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOpsInterfaces.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOpsInterfaces.td
index 2e37384ce3eb7..ab9b78e755d9d 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOpsInterfaces.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOpsInterfaces.td
@@ -69,6 +69,73 @@ def ReductionClauseInterface : OpInterface<"ReductionClauseInterface"> {
   ];
 }
 
+def LoopWrapperInterface : OpInterface<"LoopWrapperInterface"> {
+  let description = [{
+    OpenMP operations that can wrap a single loop nest. When taking a wrapper
+    role, these operations must only contain a single region with a single block
+    in which there's a single operation and a terminator. That nested operation
+    must be another loop wrapper or an `omp.loop_nest`.
+  }];
+
+  let cppNamespace = "::mlir::omp";
+
+  let methods = [
+    InterfaceMethod<
+      /*description=*/[{
+        Tell whether the operation could be taking the role of a loop wrapper.
+        That is, it has a single region with a single block in which there are
+        two operations: another wrapper or `omp.loop_nest` operation and a
+        terminator.
+      }],
+      /*retTy=*/"bool",
+      /*methodName=*/"isWrapper",
+      (ins ), [{}], [{
+        if ($_op->getNumRegions() != 1)
+          return false;
+
+        Region &r = $_op->getRegion(0);
+        if (!r.hasOneBlock())
+          return false;
+
+        if (::llvm::range_size(r.getOps()) != 2)
+          return false;
+
+        Operation &firstOp = *r.op_begin();
+        Operation &secondOp = *(std::next(r.op_begin()));
+        return ::llvm::isa<LoopNestOp, LoopWrapperInterface>(firstOp) &&
+               secondOp.hasTrait<OpTrait::IsTerminator>();
+      }]
+    >,
+    InterfaceMethod<
+      /*description=*/[{
+        If there is another loop wrapper immediately nested inside, return that
+        operation. Assumes this operation is taking a loop wrapper role.
+      }],
+      /*retTy=*/"::mlir::omp::LoopWrapperInterface",
+      /*methodName=*/"getNestedWrapper",
+      (ins), [{}], [{
+        assert($_op.isWrapper() && "Unexpected non-wrapper op");
+        Operation *nested = &*$_op->getRegion(0).op_begin();
+        return ::llvm::dyn_cast<LoopWrapperInterface>(nested);
+      }]
+    >,
+    InterfaceMethod<
+      /*description=*/[{
+        Return the loop nest nested directly or indirectly inside of this loop
+        wrapper. Assumes this operation is taking a loop wrapper role.
+      }],
+      /*retTy=*/"::mlir::Operation *",
+      /*methodName=*/"getWrappedLoop",
+      (ins), [{}], [{
+        assert($_op.isWrapper() && "Unexpected non-wrapper op");
+        if (LoopWrapperInterface nested = $_op.getNestedWrapper())
+          return nested.getWrappedLoop();
+        return &*$_op->getRegion(0).op_begin();
+      }]
+    >
+  ];
+}
+
 def DeclareTargetInterface : OpInterface<"DeclareTargetInterface"> {
   let description = [{
     OpenMP operations that support declare target have this interface.
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 35fb174046a3a..90b49b2528b79 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1936,9 +1936,27 @@ LogicalResult LoopNestOp::verify() {
              << "range argument type does not match corresponding IV type";
   }
 
+  auto wrapper =
+      llvm::dyn_cast_if_present<LoopWrapperInterface>((*this)->getParentOp());
+
+  if (!wrapper || !wrapper.isWrapper())
+    return emitOpError() << "expects parent op to be a valid loop wrapper";
+
   return success();
 }
 
+void LoopNestOp::gatherWrappers(
+    SmallVectorImpl<LoopWrapperInterface> &wrappers) {
+  Operation *parent = (*this)->getParentOp();
+  while (auto wrapper =
+             llvm::dyn_cast_if_present<LoopWrapperInterface>(parent)) {
+    if (!wrapper.isWrapper())
+      break;
+    wrappers.push_back(wrapper);
+    parent = parent->getParentOp();
+  }
+}
+
 //===----------------------------------------------------------------------===//
 // WsloopOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index 3802fbde534d6..88dca1b85ee5f 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -88,7 +88,7 @@ func.func @proc_bind_once() {
 // -----
 
 func.func @invalid_parent(%lb : index, %ub : index, %step : index) {
-  // expected-error@+1 {{op expects parent op to be one of 'omp.distribute, omp.simdloop, omp.taskloop, omp.wsloop'}}
+  // expected-error@+1 {{op expects parent op to be a valid loop wrapper}}
   omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
     omp.yield
   }
@@ -96,6 +96,20 @@ func.func @invalid_parent(%lb : index, %ub : index, %step : index) {
 
 // -----
 
+func.func @invalid_wrapper(%lb : index, %ub : index, %step : index) {
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    %0 = arith.constant 0 : i32
+    // expected-error@+1 {{op expects parent op to be a valid loop wrapper}}
+    omp.loop_nest (%iv2) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+    omp.yield
+  }
+}
+
+// -----
+
 func.func @type_mismatch(%lb : index, %ub : index, %step : index) {
   // TODO Remove induction variables from omp.wsloop.
   omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {

From c7bd284dc3ff21a5a14b4ccbbfa1988de8723c33 Mon Sep 17 00:00:00 2001
From: shamithoke <152091883+shamithoke@users.noreply.github.com>
Date: Mon, 15 Apr 2024 15:08:26 +0530
Subject: [PATCH 403/886] Support for i8/i16 for bitreverse using GFNI.
 (#88625)

In continuation to the PR #81764, this change extends the GFNI support to i8 and i16.

---------

Co-authored-by: shami <shami_thoke@yahoo.com>
---
 llvm/lib/Target/X86/X86ISelLowering.cpp    | 14 ++--
 llvm/test/CodeGen/X86/bitreverse.ll        | 57 ++++------------
 llvm/test/CodeGen/X86/vector-bitreverse.ll | 78 +++++-----------------
 3 files changed, 38 insertions(+), 111 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index edfa05a66288e..87f2eaddedf86 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1287,8 +1287,10 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     }
 
     if (Subtarget.hasGFNI()) {
-      setOperationAction(ISD::BITREVERSE,       MVT::i32, Custom);
-      setOperationAction(ISD::BITREVERSE,       MVT::i64, Custom);
+      setOperationAction(ISD::BITREVERSE, MVT::i8, Custom);
+      setOperationAction(ISD::BITREVERSE, MVT::i16, Custom);
+      setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
+      setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
     }
 
     // These might be better off as horizontal vector ops.
@@ -31317,16 +31319,18 @@ static SDValue LowerBITREVERSE(SDValue Op, const X86Subtarget &Subtarget,
   if (VT.is256BitVector() && !Subtarget.hasInt256())
     return splitVectorIntUnary(Op, DAG, DL);
 
-  // Lower i32/i64 as vXi8 BITREVERSE + BSWAP
+  // Lower i8/i16/i32/i64 as vXi8 BITREVERSE + BSWAP
   if (!VT.isVector()) {
-    assert((VT == MVT::i32 || VT == MVT::i64) && "Only tested for i32/i64");
+    assert(
+        (VT == MVT::i32 || VT == MVT::i64 || VT == MVT::i16 || VT == MVT::i8) &&
+        "Only tested for i8/i16/i32/i64");
     MVT VecVT = MVT::getVectorVT(VT, 128 / VT.getSizeInBits());
     SDValue Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VecVT, In);
     Res = DAG.getNode(ISD::BITREVERSE, DL, MVT::v16i8,
                       DAG.getBitcast(MVT::v16i8, Res));
     Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT,
                       DAG.getBitcast(VecVT, Res), DAG.getIntPtrConstant(0, DL));
-    return DAG.getNode(ISD::BSWAP, DL, VT, Res);
+    return (VT == MVT::i8) ? Res : DAG.getNode(ISD::BSWAP, DL, VT, Res);
   }
 
   assert(VT.isVector() && VT.getSizeInBits() >= 128);
diff --git a/llvm/test/CodeGen/X86/bitreverse.ll b/llvm/test/CodeGen/X86/bitreverse.ll
index 704563ab1bbf7..4f2654843728f 100644
--- a/llvm/test/CodeGen/X86/bitreverse.ll
+++ b/llvm/test/CodeGen/X86/bitreverse.ll
@@ -374,24 +374,10 @@ define i16 @test_bitreverse_i16(i16 %a) nounwind {
 ;
 ; GFNI-LABEL: test_bitreverse_i16:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNI-NEXT:    rolw $8, %di
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $3855, %eax # imm = 0xF0F
-; GFNI-NEXT:    shll $4, %eax
-; GFNI-NEXT:    shrl $4, %edi
-; GFNI-NEXT:    andl $3855, %edi # imm = 0xF0F
-; GFNI-NEXT:    orl %eax, %edi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andl $13107, %eax # imm = 0x3333
-; GFNI-NEXT:    shrl $2, %edi
-; GFNI-NEXT:    andl $13107, %edi # imm = 0x3333
-; GFNI-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNI-NEXT:    movl %eax, %ecx
-; GFNI-NEXT:    andl $21845, %ecx # imm = 0x5555
-; GFNI-NEXT:    shrl %eax
-; GFNI-NEXT:    andl $21845, %eax # imm = 0x5555
-; GFNI-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNI-NEXT:    vmovd %edi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovd %xmm0, %eax
+; GFNI-NEXT:    rolw $8, %ax
 ; GFNI-NEXT:    # kill: def $ax killed $ax killed $eax
 ; GFNI-NEXT:    retq
   %b = call i16 @llvm.bitreverse.i16(i16 %a)
@@ -446,19 +432,10 @@ define i8 @test_bitreverse_i8(i8 %a) {
 ;
 ; GFNI-LABEL: test_bitreverse_i8:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    rolb $4, %dil
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andb $51, %al
-; GFNI-NEXT:    shlb $2, %al
-; GFNI-NEXT:    shrb $2, %dil
-; GFNI-NEXT:    andb $51, %dil
-; GFNI-NEXT:    orb %dil, %al
-; GFNI-NEXT:    movl %eax, %ecx
-; GFNI-NEXT:    andb $85, %cl
-; GFNI-NEXT:    addb %cl, %cl
-; GFNI-NEXT:    shrb %al
-; GFNI-NEXT:    andb $85, %al
-; GFNI-NEXT:    orb %cl, %al
+; GFNI-NEXT:    vmovd %edi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovd %xmm0, %eax
+; GFNI-NEXT:    # kill: def $al killed $al killed $eax
 ; GFNI-NEXT:    retq
   %b = call i8 @llvm.bitreverse.i8(i8 %a)
   ret i8 %b
@@ -514,19 +491,11 @@ define i4 @test_bitreverse_i4(i4 %a) {
 ;
 ; GFNI-LABEL: test_bitreverse_i4:
 ; GFNI:       # %bb.0:
-; GFNI-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNI-NEXT:    movl %edi, %eax
-; GFNI-NEXT:    andb $8, %al
-; GFNI-NEXT:    leal (%rdi,%rdi), %ecx
-; GFNI-NEXT:    andb $4, %cl
-; GFNI-NEXT:    leal (,%rdi,8), %edx
-; GFNI-NEXT:    andb $8, %dl
-; GFNI-NEXT:    orb %cl, %dl
-; GFNI-NEXT:    shrb %dil
-; GFNI-NEXT:    andb $2, %dil
-; GFNI-NEXT:    orb %dil, %dl
-; GFNI-NEXT:    shrb $3, %al
-; GFNI-NEXT:    orb %dl, %al
+; GFNI-NEXT:    vmovd %edi, %xmm0
+; GFNI-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; GFNI-NEXT:    vmovd %xmm0, %eax
+; GFNI-NEXT:    shrb $4, %al
+; GFNI-NEXT:    # kill: def $al killed $al killed $eax
 ; GFNI-NEXT:    retq
   %b = call i4 @llvm.bitreverse.i4(i4 %a)
   ret i4 %b
diff --git a/llvm/test/CodeGen/X86/vector-bitreverse.ll b/llvm/test/CodeGen/X86/vector-bitreverse.ll
index b22b508db8b28..90cc3d5fdde82 100644
--- a/llvm/test/CodeGen/X86/vector-bitreverse.ll
+++ b/llvm/test/CodeGen/X86/vector-bitreverse.ll
@@ -61,36 +61,18 @@ define i8 @test_bitreverse_i8(i8 %a) nounwind {
 ;
 ; GFNISSE-LABEL: test_bitreverse_i8:
 ; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    rolb $4, %dil
-; GFNISSE-NEXT:    movl %edi, %eax
-; GFNISSE-NEXT:    andb $51, %al
-; GFNISSE-NEXT:    shlb $2, %al
-; GFNISSE-NEXT:    shrb $2, %dil
-; GFNISSE-NEXT:    andb $51, %dil
-; GFNISSE-NEXT:    orb %dil, %al
-; GFNISSE-NEXT:    movl %eax, %ecx
-; GFNISSE-NEXT:    andb $85, %cl
-; GFNISSE-NEXT:    addb %cl, %cl
-; GFNISSE-NEXT:    shrb %al
-; GFNISSE-NEXT:    andb $85, %al
-; GFNISSE-NEXT:    orb %cl, %al
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    # kill: def $al killed $al killed $eax
 ; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i8:
 ; GFNIAVX:       # %bb.0:
-; GFNIAVX-NEXT:    rolb $4, %dil
-; GFNIAVX-NEXT:    movl %edi, %eax
-; GFNIAVX-NEXT:    andb $51, %al
-; GFNIAVX-NEXT:    shlb $2, %al
-; GFNIAVX-NEXT:    shrb $2, %dil
-; GFNIAVX-NEXT:    andb $51, %dil
-; GFNIAVX-NEXT:    orb %dil, %al
-; GFNIAVX-NEXT:    movl %eax, %ecx
-; GFNIAVX-NEXT:    andb $85, %cl
-; GFNIAVX-NEXT:    addb %cl, %cl
-; GFNIAVX-NEXT:    shrb %al
-; GFNIAVX-NEXT:    andb $85, %al
-; GFNIAVX-NEXT:    orb %cl, %al
+; GFNIAVX-NEXT:    vmovd %edi, %xmm0
+; GFNIAVX-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; GFNIAVX-NEXT:    vmovd %xmm0, %eax
+; GFNIAVX-NEXT:    # kill: def $al killed $al killed $eax
 ; GFNIAVX-NEXT:    retq
   %b = call i8 @llvm.bitreverse.i8(i8 %a)
   ret i8 %b
@@ -153,47 +135,19 @@ define i16 @test_bitreverse_i16(i16 %a) nounwind {
 ;
 ; GFNISSE-LABEL: test_bitreverse_i16:
 ; GFNISSE:       # %bb.0:
-; GFNISSE-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNISSE-NEXT:    rolw $8, %di
-; GFNISSE-NEXT:    movl %edi, %eax
-; GFNISSE-NEXT:    andl $3855, %eax # imm = 0xF0F
-; GFNISSE-NEXT:    shll $4, %eax
-; GFNISSE-NEXT:    shrl $4, %edi
-; GFNISSE-NEXT:    andl $3855, %edi # imm = 0xF0F
-; GFNISSE-NEXT:    orl %eax, %edi
-; GFNISSE-NEXT:    movl %edi, %eax
-; GFNISSE-NEXT:    andl $13107, %eax # imm = 0x3333
-; GFNISSE-NEXT:    shrl $2, %edi
-; GFNISSE-NEXT:    andl $13107, %edi # imm = 0x3333
-; GFNISSE-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNISSE-NEXT:    movl %eax, %ecx
-; GFNISSE-NEXT:    andl $21845, %ecx # imm = 0x5555
-; GFNISSE-NEXT:    shrl %eax
-; GFNISSE-NEXT:    andl $21845, %eax # imm = 0x5555
-; GFNISSE-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNISSE-NEXT:    movd %edi, %xmm0
+; GFNISSE-NEXT:    gf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; GFNISSE-NEXT:    movd %xmm0, %eax
+; GFNISSE-NEXT:    rolw $8, %ax
 ; GFNISSE-NEXT:    # kill: def $ax killed $ax killed $eax
 ; GFNISSE-NEXT:    retq
 ;
 ; GFNIAVX-LABEL: test_bitreverse_i16:
 ; GFNIAVX:       # %bb.0:
-; GFNIAVX-NEXT:    # kill: def $edi killed $edi def $rdi
-; GFNIAVX-NEXT:    rolw $8, %di
-; GFNIAVX-NEXT:    movl %edi, %eax
-; GFNIAVX-NEXT:    andl $3855, %eax # imm = 0xF0F
-; GFNIAVX-NEXT:    shll $4, %eax
-; GFNIAVX-NEXT:    shrl $4, %edi
-; GFNIAVX-NEXT:    andl $3855, %edi # imm = 0xF0F
-; GFNIAVX-NEXT:    orl %eax, %edi
-; GFNIAVX-NEXT:    movl %edi, %eax
-; GFNIAVX-NEXT:    andl $13107, %eax # imm = 0x3333
-; GFNIAVX-NEXT:    shrl $2, %edi
-; GFNIAVX-NEXT:    andl $13107, %edi # imm = 0x3333
-; GFNIAVX-NEXT:    leal (%rdi,%rax,4), %eax
-; GFNIAVX-NEXT:    movl %eax, %ecx
-; GFNIAVX-NEXT:    andl $21845, %ecx # imm = 0x5555
-; GFNIAVX-NEXT:    shrl %eax
-; GFNIAVX-NEXT:    andl $21845, %eax # imm = 0x5555
-; GFNIAVX-NEXT:    leal (%rax,%rcx,2), %eax
+; GFNIAVX-NEXT:    vmovd %edi, %xmm0
+; GFNIAVX-NEXT:    vgf2p8affineqb $0, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; GFNIAVX-NEXT:    vmovd %xmm0, %eax
+; GFNIAVX-NEXT:    rolw $8, %ax
 ; GFNIAVX-NEXT:    # kill: def $ax killed $ax killed $eax
 ; GFNIAVX-NEXT:    retq
   %b = call i16 @llvm.bitreverse.i16(i16 %a)

From b4e7b56403387630d4c390ea4f53921312326e34 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Mon, 15 Apr 2024 11:48:22 +0200
Subject: [PATCH 404/886] Revert "Fix rsqrt inaccuracies." (#88705)

Reverts llvm/llvm-project#88691
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 88 +++----------------
 .../convert-to-standard.mlir                  | 17 +---
 2 files changed, 13 insertions(+), 92 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 3ebee9baff31b..49eb575212ffc 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -27,11 +27,9 @@ using namespace mlir;
 
 namespace {
 
-enum class AbsFn { abs, sqrt, rsqrt };
-
-// Returns the absolute value, its square root or its reciprocal square root.
+// Returns the absolute value or its square root.
 Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
-                 ImplicitLocOpBuilder &b, AbsFn fn = AbsFn::abs) {
+                 ImplicitLocOpBuilder &b, bool returnSqrt = false) {
   Value one = b.create<arith::ConstantOp>(real.getType(),
                                           b.getFloatAttr(real.getType(), 1.0));
 
@@ -45,13 +43,7 @@ Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
   Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
   Value result;
 
-  if (fn == AbsFn::rsqrt) {
-    ratioSqPlusOne = b.create<math::RsqrtOp>(ratioSqPlusOne, fmf);
-    min = b.create<math::RsqrtOp>(min, fmf);
-    max = b.create<math::RsqrtOp>(max, fmf);
-  }
-
-  if (fn == AbsFn::sqrt) {
+  if (returnSqrt) {
     Value quarter = b.create<arith::ConstantOp>(
         real.getType(), b.getFloatAttr(real.getType(), 0.25));
     // sqrt(sqrt(a*b)) would avoid the pow, but will overflow more easily.
@@ -871,7 +863,7 @@ struct SqrtOpConversion : public OpConversionPattern<complex::SqrtOp> {
 
     Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
     Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
-    Value absSqrt = computeAbs(real, imag, fmf, b, AbsFn::sqrt);
+    Value absSqrt = computeAbs(real, imag, fmf, b, /*returnSqrt=*/true);
     Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
     Value sqrtArg = b.create<arith::MulFOp>(argArg, half, fmf);
     Value cos = b.create<math::CosOp>(sqrtArg, fmf);
@@ -1155,74 +1147,18 @@ struct RsqrtOpConversion : public OpConversionPattern<complex::RsqrtOp> {
   LogicalResult
   matchAndRewrite(complex::RsqrtOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
+    mlir::ImplicitLocOpBuilder builder(op.getLoc(), rewriter);
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
 
-    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
-
-    auto cst = [&](APFloat v) {
-      return b.create<arith::ConstantOp>(elementType,
-                                         b.getFloatAttr(elementType, v));
-    };
-    const auto &floatSemantics = elementType.getFloatSemantics();
-    Value zero = cst(APFloat::getZero(floatSemantics));
-    Value inf = cst(APFloat::getInf(floatSemantics));
-    Value negHalf = b.create<arith::ConstantOp>(
-        elementType, b.getFloatAttr(elementType, -0.5));
-    Value nan = cst(APFloat::getNaN(floatSemantics));
-
-    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
-    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
-    Value absRsqrt = computeAbs(real, imag, fmf, b, AbsFn::rsqrt);
-    Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
-    Value rsqrtArg = b.create<arith::MulFOp>(argArg, negHalf, fmf);
-    Value cos = b.create<math::CosOp>(rsqrtArg, fmf);
-    Value sin = b.create<math::SinOp>(rsqrtArg, fmf);
-
-    Value resultReal = b.create<arith::MulFOp>(absRsqrt, cos, fmf);
-    Value resultImag = b.create<arith::MulFOp>(absRsqrt, sin, fmf);
-
-    if (!arith::bitEnumContainsAll(fmf, arith::FastMathFlags::nnan |
-                                            arith::FastMathFlags::ninf)) {
-      Value negOne = b.create<arith::ConstantOp>(
-          elementType, b.getFloatAttr(elementType, -1));
-
-      Value realSignedZero = b.create<math::CopySignOp>(zero, real, fmf);
-      Value imagSignedZero = b.create<math::CopySignOp>(zero, imag, fmf);
-      Value negImagSignedZero =
-          b.create<arith::MulFOp>(negOne, imagSignedZero, fmf);
+    Value c = builder.create<arith::ConstantOp>(
+        elementType, builder.getFloatAttr(elementType, -0.5));
+    Value d = builder.create<arith::ConstantOp>(
+        elementType, builder.getFloatAttr(elementType, 0));
 
-      Value absReal = b.create<math::AbsFOp>(real, fmf);
-      Value absImag = b.create<math::AbsFOp>(imag, fmf);
-
-      Value absImagIsInf =
-          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absImag, inf, fmf);
-      Value realIsNan =
-          b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, real, real, fmf);
-      Value realIsInf =
-          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absReal, inf, fmf);
-      Value inIsNanInf = b.create<arith::AndIOp>(absImagIsInf, realIsNan);
-
-      Value resultIsZero = b.create<arith::OrIOp>(inIsNanInf, realIsInf);
-
-      resultReal =
-          b.create<arith::SelectOp>(resultIsZero, realSignedZero, resultReal);
-      resultImag = b.create<arith::SelectOp>(resultIsZero, negImagSignedZero,
-                                             resultImag);
-    }
-
-    Value isRealZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero, fmf);
-    Value isImagZero =
-        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero, fmf);
-    Value isZero = b.create<arith::AndIOp>(isRealZero, isImagZero);
-
-    resultReal = b.create<arith::SelectOp>(isZero, inf, resultReal);
-    resultImag = b.create<arith::SelectOp>(isZero, nan, resultImag);
-
-    rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
-                                                   resultImag);
+    rewriter.replaceOp(op,
+                       {powOpConversionImpl(builder, type, adaptor.getComplex(),
+                                            c, d, op.getFastmath())});
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 8b4ea9777f797..e0e7cdadd317d 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -837,21 +837,6 @@ func.func @complex_rsqrt(%arg: complex<f32>) -> complex<f32> {
   return %rsqrt : complex<f32>
 }
 
-// CHECK-COUNT-5: arith.select
-// CHECK-NOT: arith.select
-
-// -----
-
-// CHECK-LABEL: func @complex_rsqrt_nnan_ninf
-// CHECK-SAME: %[[ARG:.*]]: complex<f32>
-func.func @complex_rsqrt_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
-  %sqrt = complex.rsqrt %arg fastmath<nnan,ninf> : complex<f32>
-  return %sqrt : complex<f32>
-}
-
-// CHECK-COUNT-3: arith.select
-// CHECK-NOT: arith.select
-
 // -----
 
 // CHECK-LABEL:   func.func @complex_angle
@@ -2118,4 +2103,4 @@ func.func @complex_tanh_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] fastmath<nnan,contract> : f32
-// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
+// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
\ No newline at end of file

From 45137766ca6be16b9bc6082eb9bedae7bc621437 Mon Sep 17 00:00:00 2001
From: Victor Campos <victor.campos@arm.com>
Date: Mon, 15 Apr 2024 10:58:30 +0100
Subject: [PATCH 405/886] [ARM][Thumb2] Mark BTI-clearing instructions as
 scheduling region boundaries (#87982)

Following https://github.com/llvm/llvm-project/pull/68313 this patch
extends the idea to M-profile PACBTI.

The Machine Scheduler can reorder instructions within a scheduling
region depending on the scheduling policy set. If a BTI-clearing
instruction happens to partake in one such region, it might be moved
around, therefore ending up where it shouldn't.

The solution is to mark all BTI-clearing instructions as scheduling
region boundaries. This essentially means that they must not be part of
any scheduling region, and as consequence never get moved:

 - PAC
 - PACBTI
 - BTI
 - SG

Note that PAC isn't BTI-clearing, but it's replaced by PACBTI late in
the compilation pipeline.

As far as I know, currently it isn't possible to organically obtain code
that's susceptible to the bug:

- Instructions that write to SP are region boundaries. PAC seems to
always be followed by the pushing of r12 to the stack, so essentially
PAC is always by itself in a scheduling region.
- CALL_BTI is expanded into a machine instruction bundle. Bundles are
unpacked only after the last machine scheduler run. Thus setjmp and BTI
can be separated only if someone deliberately run the scheduler once
more.
- The BTI insertion pass is run late in the pipeline, only after the
last machine scheduling has run. So once again it can be reordered only
if someone deliberately runs the scheduler again.

Nevertheless, one can reasonably argue that we should prevent the bug in
spite of the compiler not being able to produce the required conditions
for it. If things change, the compiler will be robust against this
issue.

The tests written for this are contrived: bogus MIR instructions have
been added adjacent to the BTI-clearing instructions in order to have
them inside non-trivial scheduling regions.
---
 llvm/lib/Target/ARM/Thumb2InstrInfo.cpp       |  19 ++
 llvm/lib/Target/ARM/Thumb2InstrInfo.h         |   4 +
 .../CodeGen/ARM/misched-branch-targets.mir    | 169 ++++++++++++++++++
 3 files changed, 192 insertions(+)
 create mode 100644 llvm/test/CodeGen/ARM/misched-branch-targets.mir

diff --git a/llvm/lib/Target/ARM/Thumb2InstrInfo.cpp b/llvm/lib/Target/ARM/Thumb2InstrInfo.cpp
index 083f25f49dec4..fc2834cb0b45c 100644
--- a/llvm/lib/Target/ARM/Thumb2InstrInfo.cpp
+++ b/llvm/lib/Target/ARM/Thumb2InstrInfo.cpp
@@ -286,6 +286,25 @@ MachineInstr *Thumb2InstrInfo::commuteInstructionImpl(MachineInstr &MI,
   return ARMBaseInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
 }
 
+bool Thumb2InstrInfo::isSchedulingBoundary(const MachineInstr &MI,
+                                           const MachineBasicBlock *MBB,
+                                           const MachineFunction &MF) const {
+  // BTI clearing instructions shall not take part in scheduling regions as
+  // they must stay in their intended place. Although PAC isn't BTI clearing,
+  // it can be transformed into PACBTI after the pre-RA Machine Scheduling
+  // has taken place, so its movement must also be restricted.
+  switch (MI.getOpcode()) {
+  case ARM::t2BTI:
+  case ARM::t2PAC:
+  case ARM::t2PACBTI:
+  case ARM::t2SG:
+    return true;
+  default:
+    break;
+  }
+  return ARMBaseInstrInfo::isSchedulingBoundary(MI, MBB, MF);
+}
+
 void llvm::emitT2RegPlusImmediate(MachineBasicBlock &MBB,
                                   MachineBasicBlock::iterator &MBBI,
                                   const DebugLoc &dl, Register DestReg,
diff --git a/llvm/lib/Target/ARM/Thumb2InstrInfo.h b/llvm/lib/Target/ARM/Thumb2InstrInfo.h
index 4bb412f09dcbe..8915da8c5bf3c 100644
--- a/llvm/lib/Target/ARM/Thumb2InstrInfo.h
+++ b/llvm/lib/Target/ARM/Thumb2InstrInfo.h
@@ -68,6 +68,10 @@ class Thumb2InstrInfo : public ARMBaseInstrInfo {
                                        unsigned OpIdx1,
                                        unsigned OpIdx2) const override;
 
+  bool isSchedulingBoundary(const MachineInstr &MI,
+                            const MachineBasicBlock *MBB,
+                            const MachineFunction &MF) const override;
+
 private:
   void expandLoadStackGuard(MachineBasicBlock::iterator MI) const override;
 };
diff --git a/llvm/test/CodeGen/ARM/misched-branch-targets.mir b/llvm/test/CodeGen/ARM/misched-branch-targets.mir
new file mode 100644
index 0000000000000..d828d9e516273
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/misched-branch-targets.mir
@@ -0,0 +1,169 @@
+# RUN: llc -o - -run-pass=machine-scheduler -misched=shuffle %s | FileCheck %s
+# RUN: llc -o - -run-pass=postmisched %s | FileCheck %s
+
+# REQUIRES: asserts
+# -misched=shuffle is only available with assertions enabled
+
+--- |
+  target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+  target triple = "thumbv8.1m.main-arm-none-eabi"
+
+  define i32 @foo_bti() #0 {
+  entry:
+    ret i32 0
+  }
+
+  define i32 @foo_pac() #0 {
+  entry:
+    ret i32 0
+  }
+
+  define i32 @foo_pacbti() #0 {
+  entry:
+    ret i32 0
+  }
+
+  define i32 @foo_setjmp() #0 {
+  entry:
+    ret i32 0
+  if.then:
+    ret i32 0
+  }
+
+  define i32 @foo_sg() #0 {
+  entry:
+    ret i32 0
+  }
+
+  declare i32 @setjmp(ptr noundef) #1
+  declare void @longjmp(ptr noundef, i32 noundef) #2
+
+  attributes #0 = { "frame-pointer"="all" "target-cpu"="cortex-m55" "target-features"="+armv8.1-m.main" }
+  attributes #1 = { nounwind returns_twice "frame-pointer"="all" "target-cpu"="cortex-m55" "target-features"="+armv8.1-m.main" }
+  attributes #2 = { noreturn nounwind "frame-pointer"="all" "target-cpu"="cortex-m55" "target-features"="+armv8.1-m.main" }
+
+...
+---
+name:            foo_bti
+tracksRegLiveness: true
+body:             |
+  bb.0.entry:
+    liveins: $r0
+
+    t2BTI
+    renamable $r0, dead $cpsr = nsw tADDi8 killed renamable $r0, 1, 14 /* CC::al */, $noreg
+    tBX_RET 14 /* CC::al */, $noreg, implicit killed $r0
+
+...
+
+# CHECK-LABEL: name:            foo_bti
+# CHECK:       body:
+# CHECK-NEXT:   bb.0.entry:
+# CHECK-NEXT:     liveins: $r0
+# CHECK-NEXT:     {{^ +$}}
+# CHECK-NEXT:     t2BTI
+
+---
+name:            foo_pac
+tracksRegLiveness: true
+body:             |
+  bb.0.entry:
+    liveins: $r0, $lr, $r12
+
+    frame-setup t2PAC implicit-def $r12, implicit $lr, implicit $sp
+    renamable $r2 = nsw t2ADDri $r0, 3, 14 /* CC::al */, $noreg, $noreg
+    $sp = frame-setup t2STMDB_UPD $sp, 14 /* CC::al */, $noreg, killed $r7, killed $lr
+    $r7 = frame-setup tMOVr killed $sp, 14 /* CC::al */, $noreg
+    early-clobber $sp = frame-setup t2STR_PRE killed $r12, $sp, -4, 14 /* CC::al */, $noreg
+    $r12, $sp = frame-destroy t2LDR_POST $sp, 4, 14 /* CC::al */, $noreg
+    $sp = frame-destroy t2LDMIA_UPD $sp, 14 /* CC::al */, $noreg, def $r7, def $lr
+    t2AUT implicit $r12, implicit $lr, implicit $sp
+    tBX_RET 14 /* CC::al */, $noreg, implicit $r0
+
+...
+
+# CHECK-LABEL: name:            foo_pac
+# CHECK:       body:
+# CHECK-NEXT:    bb.0.entry:
+# CHECK-NEXT:      liveins: $r0, $lr, $r12
+# CHECK-NEXT:      {{^ +$}}
+# CHECK-NEXT:      frame-setup t2PAC implicit-def $r12, implicit $lr, implicit $sp
+
+---
+name:            foo_pacbti
+tracksRegLiveness: true
+body:             |
+  bb.0.entry:
+    liveins: $r0, $lr, $r12
+
+    frame-setup t2PACBTI implicit-def $r12, implicit $lr, implicit $sp
+    renamable $r2 = nsw t2ADDri $r0, 3, 14 /* CC::al */, $noreg, $noreg
+    $sp = frame-setup t2STMDB_UPD $sp, 14 /* CC::al */, $noreg, killed $r7, killed $lr
+    $r7 = frame-setup tMOVr killed $sp, 14 /* CC::al */, $noreg
+    early-clobber $sp = frame-setup t2STR_PRE killed $r12, $sp, -4, 14 /* CC::al */, $noreg
+    $r12, $sp = frame-destroy t2LDR_POST $sp, 4, 14 /* CC::al */, $noreg
+    $sp = frame-destroy t2LDMIA_UPD $sp, 14 /* CC::al */, $noreg, def $r7, def $lr
+    t2AUT implicit $r12, implicit $lr, implicit $sp
+    tBX_RET 14 /* CC::al */, $noreg, implicit $r0
+
+...
+
+# CHECK-LABEL: name:            foo_pacbti
+# CHECK:       body:
+# CHECK-NEXT:    bb.0.entry:
+# CHECK-NEXT:      liveins: $r0, $lr, $r12
+# CHECK-NEXT:      {{^ +$}}
+# CHECK-NEXT:      frame-setup t2PACBTI implicit-def $r12, implicit $lr, implicit $sp
+
+---
+name:            foo_setjmp
+tracksRegLiveness: true
+body:             |
+  bb.0.entry:
+    successors: %bb.1
+    liveins: $lr
+
+    frame-setup tPUSH 14 /* CC::al */, $noreg, $r7, killed $lr, implicit-def $sp, implicit $sp
+    $r7 = frame-setup tMOVr $sp, 14 /* CC::al */, $noreg
+    $sp = frame-setup tSUBspi $sp, 40, 14 /* CC::al */, $noreg
+    renamable $r0 = tMOVr $sp, 14 /* CC::al */, $noreg
+    tBL 14 /* CC::al */, $noreg, @setjmp, csr_aapcs, implicit-def dead $lr, implicit $sp, implicit killed $r0, implicit-def $sp, implicit-def $r0
+    t2BTI
+    renamable $r2 = nsw t2ADDri $r0, 3, 14 /* CC::al */, $noreg, $noreg
+    tCMPi8 killed renamable $r0, 0, 14 /* CC::al */, $noreg, implicit-def $cpsr
+    t2IT 0, 2, implicit-def $itstate
+    renamable $r0 = tMOVi8 $noreg, 0, 0 /* CC::eq */, $cpsr, implicit $itstate
+    $sp = frame-destroy tADDspi $sp, 40, 0 /* CC::eq */, $cpsr, implicit $itstate
+    frame-destroy tPOP_RET 0 /* CC::eq */, killed $cpsr, def $r7, def $pc, implicit killed $r0, implicit $sp, implicit killed $itstate
+
+  bb.1.if.then:
+    renamable $r0 = tMOVr $sp, 14 /* CC::al */, $noreg
+    renamable $r1, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
+    tBL 14 /* CC::al */, $noreg, @longjmp, csr_aapcs, implicit-def dead $lr, implicit $sp, implicit killed $r0, implicit killed $r1, implicit-def $sp
+
+...
+
+# CHECK-LABEL: name:            foo_setjmp
+# CHECK:       body:
+# CHECK:         tBL 14 /* CC::al */, $noreg, @setjmp, csr_aapcs, implicit-def dead $lr, implicit $sp, implicit killed $r0, implicit-def $sp, implicit-def $r0
+# CHECK-NEXT:    t2BTI
+
+---
+name:            foo_sg
+tracksRegLiveness: true
+body:             |
+  bb.0.entry:
+    liveins: $r0
+
+    t2SG 14 /* CC::al */, $noreg
+    renamable $r0, dead $cpsr = nsw tADDi8 killed renamable $r0, 1, 14 /* CC::al */, $noreg
+    tBX_RET 14 /* CC::al */, $noreg, implicit killed $r0
+
+...
+
+# CHECK-LABEL: name:            foo_sg
+# CHECK:       body:
+# CHECK-NEXT:   bb.0.entry:
+# CHECK-NEXT:     liveins: $r0
+# CHECK-NEXT:     {{^ +$}}
+# CHECK-NEXT:     t2SG

From 0822780b7ffa552b351364be350a997a47f22250 Mon Sep 17 00:00:00 2001
From: wanglei <wanglei@loongson.cn>
Date: Mon, 15 Apr 2024 18:01:09 +0800
Subject: [PATCH 406/886] [LoongArch] Fix incorrect logic in
 isLegalAddressingMode() (#88694)

This will adress issue:
https://github.com/ClangBuiltLinux/linux/issues/2014
---
 .../Target/LoongArch/LoongArchISelLowering.cpp    | 15 +++++++--------
 llvm/test/CodeGen/LoongArch/gep-imm.ll            | 14 +++++---------
 2 files changed, 12 insertions(+), 17 deletions(-)

diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index c13b10a320f83..285d5c2a63b2d 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -4876,20 +4876,19 @@ bool LoongArchTargetLowering::isLegalAddressingMode(const DataLayout &DL,
   if (AM.BaseGV)
     return false;
 
-  // Require a 12 or 14 bit signed offset.
-  if (!isInt<12>(AM.BaseOffs) || !isShiftedInt<14, 2>(AM.BaseOffs))
+  // Require a 12-bit signed offset or 14-bit signed offset left-shifted by 2
+  // with `UAL` feature.
+  if (!isInt<12>(AM.BaseOffs) &&
+      !(isShiftedInt<14, 2>(AM.BaseOffs) && Subtarget.hasUAL()))
     return false;
 
   switch (AM.Scale) {
   case 0:
-    // "i" is not allowed.
-    if (!AM.HasBaseReg)
-      return false;
-    // Otherwise we have "r+i".
+    // "r+i" or just "i", depending on HasBaseReg.
     break;
   case 1:
     // "r+r+i" is not allowed.
-    if (AM.HasBaseReg && AM.BaseOffs != 0)
+    if (AM.HasBaseReg && AM.BaseOffs)
       return false;
     // Otherwise we have "r+r" or "r+i".
     break;
@@ -4897,7 +4896,7 @@ bool LoongArchTargetLowering::isLegalAddressingMode(const DataLayout &DL,
     // "2*r+r" or "2*r+i" is not allowed.
     if (AM.HasBaseReg || AM.BaseOffs)
       return false;
-    // Otherwise we have "r+r".
+    // Allow "2*r" as "r+r".
     break;
   default:
     return false;
diff --git a/llvm/test/CodeGen/LoongArch/gep-imm.ll b/llvm/test/CodeGen/LoongArch/gep-imm.ll
index 52d25f9d63894..0eef7e4517f3d 100644
--- a/llvm/test/CodeGen/LoongArch/gep-imm.ll
+++ b/llvm/test/CodeGen/LoongArch/gep-imm.ll
@@ -4,12 +4,8 @@
 define void @test(ptr %sp, ptr %t, i32 %n) {
 ; CHECK-LABEL: test:
 ; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    ld.d $a0, $a0, 0
 ; CHECK-NEXT:    move $a3, $zero
-; CHECK-NEXT:    ld.d $a4, $a0, 0
-; CHECK-NEXT:    lu12i.w $a0, 1
-; CHECK-NEXT:    ori $a5, $a0, 3904
-; CHECK-NEXT:    add.d $a0, $a1, $a5
-; CHECK-NEXT:    add.d $a1, $a4, $a5
 ; CHECK-NEXT:    addi.w $a2, $a2, 0
 ; CHECK-NEXT:    addi.w $a4, $a3, 0
 ; CHECK-NEXT:    bge $a4, $a2, .LBB0_2
@@ -17,10 +13,10 @@ define void @test(ptr %sp, ptr %t, i32 %n) {
 ; CHECK-NEXT:  .LBB0_1: # %while_body
 ; CHECK-NEXT:    # =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    addi.d $a4, $a3, 1
-; CHECK-NEXT:    st.w $a4, $a1, 0
-; CHECK-NEXT:    st.w $a3, $a1, 4
-; CHECK-NEXT:    st.w $a4, $a0, 0
-; CHECK-NEXT:    st.w $a3, $a0, 4
+; CHECK-NEXT:    stptr.w $a4, $a0, 8000
+; CHECK-NEXT:    stptr.w $a3, $a0, 8004
+; CHECK-NEXT:    stptr.w $a4, $a1, 8000
+; CHECK-NEXT:    stptr.w $a3, $a1, 8004
 ; CHECK-NEXT:    move $a3, $a4
 ; CHECK-NEXT:    addi.w $a4, $a3, 0
 ; CHECK-NEXT:    blt $a4, $a2, .LBB0_1

From 33e60f358cc0c564e771c0234dd8a60f3bddce5d Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Mon, 15 Apr 2024 12:04:36 +0200
Subject: [PATCH 407/886] Fix rsqrt inaccuracies. (#88707)

The current lowering has issues with large/subnormal values. This po
XLA's lowering and was verified using XLA's test suite and the
MLIR-based emitters.

This updates https://github.com/llvm/llvm-project/pull/88691 to also
update the correctness test for rsqrt(0). I checked C++ and Python, they
both agree the result should be (inf, nan). Updated the correctness test
to match this.
---
 .../ComplexToStandard/ComplexToStandard.cpp   | 88 ++++++++++++++++---
 .../convert-to-standard.mlir                  | 17 +++-
 .../Dialect/Complex/CPU/correctness.mlir      |  2 +-
 3 files changed, 93 insertions(+), 14 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 49eb575212ffc..3ebee9baff31b 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -27,9 +27,11 @@ using namespace mlir;
 
 namespace {
 
-// Returns the absolute value or its square root.
+enum class AbsFn { abs, sqrt, rsqrt };
+
+// Returns the absolute value, its square root or its reciprocal square root.
 Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
-                 ImplicitLocOpBuilder &b, bool returnSqrt = false) {
+                 ImplicitLocOpBuilder &b, AbsFn fn = AbsFn::abs) {
   Value one = b.create<arith::ConstantOp>(real.getType(),
                                           b.getFloatAttr(real.getType(), 1.0));
 
@@ -43,7 +45,13 @@ Value computeAbs(Value real, Value imag, arith::FastMathFlags fmf,
   Value ratioSqPlusOne = b.create<arith::AddFOp>(ratioSq, one, fmf);
   Value result;
 
-  if (returnSqrt) {
+  if (fn == AbsFn::rsqrt) {
+    ratioSqPlusOne = b.create<math::RsqrtOp>(ratioSqPlusOne, fmf);
+    min = b.create<math::RsqrtOp>(min, fmf);
+    max = b.create<math::RsqrtOp>(max, fmf);
+  }
+
+  if (fn == AbsFn::sqrt) {
     Value quarter = b.create<arith::ConstantOp>(
         real.getType(), b.getFloatAttr(real.getType(), 0.25));
     // sqrt(sqrt(a*b)) would avoid the pow, but will overflow more easily.
@@ -863,7 +871,7 @@ struct SqrtOpConversion : public OpConversionPattern<complex::SqrtOp> {
 
     Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
     Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
-    Value absSqrt = computeAbs(real, imag, fmf, b, /*returnSqrt=*/true);
+    Value absSqrt = computeAbs(real, imag, fmf, b, AbsFn::sqrt);
     Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
     Value sqrtArg = b.create<arith::MulFOp>(argArg, half, fmf);
     Value cos = b.create<math::CosOp>(sqrtArg, fmf);
@@ -1147,18 +1155,74 @@ struct RsqrtOpConversion : public OpConversionPattern<complex::RsqrtOp> {
   LogicalResult
   matchAndRewrite(complex::RsqrtOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
-    mlir::ImplicitLocOpBuilder builder(op.getLoc(), rewriter);
+    mlir::ImplicitLocOpBuilder b(op.getLoc(), rewriter);
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
 
-    Value c = builder.create<arith::ConstantOp>(
-        elementType, builder.getFloatAttr(elementType, -0.5));
-    Value d = builder.create<arith::ConstantOp>(
-        elementType, builder.getFloatAttr(elementType, 0));
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
+
+    auto cst = [&](APFloat v) {
+      return b.create<arith::ConstantOp>(elementType,
+                                         b.getFloatAttr(elementType, v));
+    };
+    const auto &floatSemantics = elementType.getFloatSemantics();
+    Value zero = cst(APFloat::getZero(floatSemantics));
+    Value inf = cst(APFloat::getInf(floatSemantics));
+    Value negHalf = b.create<arith::ConstantOp>(
+        elementType, b.getFloatAttr(elementType, -0.5));
+    Value nan = cst(APFloat::getNaN(floatSemantics));
+
+    Value real = b.create<complex::ReOp>(elementType, adaptor.getComplex());
+    Value imag = b.create<complex::ImOp>(elementType, adaptor.getComplex());
+    Value absRsqrt = computeAbs(real, imag, fmf, b, AbsFn::rsqrt);
+    Value argArg = b.create<math::Atan2Op>(imag, real, fmf);
+    Value rsqrtArg = b.create<arith::MulFOp>(argArg, negHalf, fmf);
+    Value cos = b.create<math::CosOp>(rsqrtArg, fmf);
+    Value sin = b.create<math::SinOp>(rsqrtArg, fmf);
+
+    Value resultReal = b.create<arith::MulFOp>(absRsqrt, cos, fmf);
+    Value resultImag = b.create<arith::MulFOp>(absRsqrt, sin, fmf);
+
+    if (!arith::bitEnumContainsAll(fmf, arith::FastMathFlags::nnan |
+                                            arith::FastMathFlags::ninf)) {
+      Value negOne = b.create<arith::ConstantOp>(
+          elementType, b.getFloatAttr(elementType, -1));
+
+      Value realSignedZero = b.create<math::CopySignOp>(zero, real, fmf);
+      Value imagSignedZero = b.create<math::CopySignOp>(zero, imag, fmf);
+      Value negImagSignedZero =
+          b.create<arith::MulFOp>(negOne, imagSignedZero, fmf);
 
-    rewriter.replaceOp(op,
-                       {powOpConversionImpl(builder, type, adaptor.getComplex(),
-                                            c, d, op.getFastmath())});
+      Value absReal = b.create<math::AbsFOp>(real, fmf);
+      Value absImag = b.create<math::AbsFOp>(imag, fmf);
+
+      Value absImagIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absImag, inf, fmf);
+      Value realIsNan =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO, real, real, fmf);
+      Value realIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absReal, inf, fmf);
+      Value inIsNanInf = b.create<arith::AndIOp>(absImagIsInf, realIsNan);
+
+      Value resultIsZero = b.create<arith::OrIOp>(inIsNanInf, realIsInf);
+
+      resultReal =
+          b.create<arith::SelectOp>(resultIsZero, realSignedZero, resultReal);
+      resultImag = b.create<arith::SelectOp>(resultIsZero, negImagSignedZero,
+                                             resultImag);
+    }
+
+    Value isRealZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, real, zero, fmf);
+    Value isImagZero =
+        b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero, fmf);
+    Value isZero = b.create<arith::AndIOp>(isRealZero, isImagZero);
+
+    resultReal = b.create<arith::SelectOp>(isZero, inf, resultReal);
+    resultImag = b.create<arith::SelectOp>(isZero, nan, resultImag);
+
+    rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
+                                                   resultImag);
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index e0e7cdadd317d..8b4ea9777f797 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -837,6 +837,21 @@ func.func @complex_rsqrt(%arg: complex<f32>) -> complex<f32> {
   return %rsqrt : complex<f32>
 }
 
+// CHECK-COUNT-5: arith.select
+// CHECK-NOT: arith.select
+
+// -----
+
+// CHECK-LABEL: func @complex_rsqrt_nnan_ninf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_rsqrt_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
+  %sqrt = complex.rsqrt %arg fastmath<nnan,ninf> : complex<f32>
+  return %sqrt : complex<f32>
+}
+
+// CHECK-COUNT-3: arith.select
+// CHECK-NOT: arith.select
+
 // -----
 
 // CHECK-LABEL:   func.func @complex_angle
@@ -2103,4 +2118,4 @@ func.func @complex_tanh_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
 // CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
 // CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] fastmath<nnan,contract> : f32
-// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
\ No newline at end of file
+// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
diff --git a/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir b/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
index 441f7500538f7..b0e414d157268 100644
--- a/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
+++ b/mlir/test/Integration/Dialect/Complex/CPU/correctness.mlir
@@ -242,7 +242,7 @@ func.func @entry() {
     // CHECK-NEXT:  0.321
     // CHECK-NEXT:  -0.776
     (0.0, 0.0),
-    // CHECK-NEXT:  nan
+    // CHECK-NEXT:  inf
     // CHECK-NEXT:  nan
     (0.0, 1.0),
     // CHECK-NEXT:  0.707

From 5b95c9e0cad189cc609c31029f6bc3a89bc7b612 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Markus=20B=C3=B6ck?= <markus.boeck02@gmail.com>
Date: Mon, 15 Apr 2024 12:13:06 +0200
Subject: [PATCH 408/886] [mlir][ODS] Make `prop-dict` behave closer to
 `attr-dict` (#88659)

`attr-dict` currently prints any attribute (inherent or discardable)
that does not occur elsewhere in the assembly format. `prop-dict` on the
other hand, always prints and parses all properties (including inherent
attributes stored as properties) as part of the property dictionary,
regardless of whether the properties or attributes are used elsewhere.
(with the exception of default-valued attributes implemented recently in
https://github.com/llvm/llvm-project/pull/87970).

This PR changes the behavior of `prop-dict` to only print and parse
attributes and properties that do not occur elsewhere in the assembly
format. This is achieved by 1) adding used attributes and properties to
the elision list when printing and 2) using a custom version of
`setPropertiesFromAttr` called `setPropertiesFromParsedAttr` that is
sensitive to the assembly format and auto-generated by ODS.

The current and new behavior of `prop-dict` and `attr-dict` were also
documented.

Happens to also fix https://github.com/llvm/llvm-project/issues/88506
---
 mlir/docs/DefiningDialects/Operations.md   |  12 ++-
 mlir/include/mlir/IR/OpDefinition.h        |  32 +++++-
 mlir/test/lib/Dialect/Test/TestDialect.cpp |  11 ++
 mlir/test/lib/Dialect/Test/TestOps.td      |  26 +++++
 mlir/test/mlir-tblgen/op-format.mlir       |  11 ++
 mlir/tools/mlir-tblgen/OpFormatGen.cpp     | 118 ++++++++++++++++++++-
 6 files changed, 202 insertions(+), 8 deletions(-)

diff --git a/mlir/docs/DefiningDialects/Operations.md b/mlir/docs/DefiningDialects/Operations.md
index b27330319f659..729393d536267 100644
--- a/mlir/docs/DefiningDialects/Operations.md
+++ b/mlir/docs/DefiningDialects/Operations.md
@@ -640,13 +640,23 @@ The available directives are as follows:
 
 *   `attr-dict`
 
-    -   Represents the attribute dictionary of the operation.
+    -   Represents the attribute dictionary of the operation. Any inherent 
+    -   attributes that are not used elsewhere in the format are printed as
+    -   part of the attribute dictionary unless a `prop-dict` is present.
+    -   Discardable attributes are always part of the `attr-dict`.  
 
 *   `attr-dict-with-keyword`
 
     -   Represents the attribute dictionary of the operation, but prefixes the
         dictionary with an `attributes` keyword.
 
+*   `prop-dict`
+
+    -   Represents the properties of the operation converted to a dictionary.
+    -   Any property or inherent attribute that are not used elsewhere in the
+    -   format are parsed and printed as part of this dictionary.
+    -   If present, the `attr-dict` will not contain any inherent attributes.
+
 *   `custom` < UserDirective > ( Params )
 
     -   Represents a custom directive implemented by the user in C++.
diff --git a/mlir/include/mlir/IR/OpDefinition.h b/mlir/include/mlir/IR/OpDefinition.h
index f4355c8ce26ac..59f094d669099 100644
--- a/mlir/include/mlir/IR/OpDefinition.h
+++ b/mlir/include/mlir/IR/OpDefinition.h
@@ -1993,8 +1993,10 @@ class Op : public OpState, public Traits<ConcreteType>... {
         p, ConcreteType::getPropertiesAsAttr(ctx, properties), elidedProps);
   }
 
-  /// Parser the properties. Unless overridden, this method will print by
-  /// converting the properties to an Attribute.
+  /// Parses 'prop-dict' for the operation. Unless overridden, the method will
+  /// parse the properties using the generic property dictionary using the
+  /// '<{ ... }>' syntax. The resulting properties are stored within the
+  /// property structure of 'result', accessible via 'getOrAddProperties'.
   template <typename T = ConcreteType>
   static ParseResult parseProperties(OpAsmParser &parser,
                                      OperationState &result) {
@@ -2002,7 +2004,31 @@ class Op : public OpState, public Traits<ConcreteType>... {
       return parseProperties(
           parser, result.getOrAddProperties<InferredProperties<T>>());
     }
-    return genericParseProperties(parser, result.propertiesAttr);
+
+    Attribute propertyDictionary;
+    if (genericParseProperties(parser, propertyDictionary))
+      return failure();
+
+    // The generated 'setPropertiesFromParsedAttr', like
+    // 'setPropertiesFromAttr', expects a 'DictionaryAttr' that is not null.
+    // Use an empty dictionary in the case that the whole dictionary is
+    // optional.
+    if (!propertyDictionary)
+      propertyDictionary = DictionaryAttr::get(result.getContext());
+
+    auto emitError = [&]() {
+      return mlir::emitError(result.location, "invalid properties ")
+             << propertyDictionary << " for op " << result.name.getStringRef()
+             << ": ";
+    };
+
+    // Copy the data from the dictionary attribute into the property struct of
+    // the operation. This method is generated by ODS by default if there are
+    // any occurrences of 'prop-dict' in the assembly format and should set
+    // any properties that aren't parsed elsewhere.
+    return ConcreteOpType::setPropertiesFromParsedAttr(
+        result.getOrAddProperties<InferredProperties<T>>(), propertyDictionary,
+        emitError);
   }
 
 private:
diff --git a/mlir/test/lib/Dialect/Test/TestDialect.cpp b/mlir/test/lib/Dialect/Test/TestDialect.cpp
index 380c74a47e509..74378633032ce 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialect.cpp
@@ -740,6 +740,17 @@ LogicalResult OpWithResultShapePerDimInterfaceOp::reifyResultShapes(
   return success();
 }
 
+LogicalResult TestOpWithPropertiesAndInferredType::inferReturnTypes(
+    MLIRContext *context, std::optional<Location>, ValueRange operands,
+    DictionaryAttr attributes, OpaqueProperties properties, RegionRange regions,
+    SmallVectorImpl<Type> &inferredReturnTypes) {
+
+  Adaptor adaptor(operands, attributes, properties, regions);
+  inferredReturnTypes.push_back(IntegerType::get(
+      context, adaptor.getLhs() + adaptor.getProperties().rhs));
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // Test SideEffect interfaces
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/lib/Dialect/Test/TestOps.td b/mlir/test/lib/Dialect/Test/TestOps.td
index edca05fde5a52..c88f85b8b6b36 100644
--- a/mlir/test/lib/Dialect/Test/TestOps.td
+++ b/mlir/test/lib/Dialect/Test/TestOps.td
@@ -2852,6 +2852,23 @@ def TestOpWithProperties : TEST_Op<"with_properties"> {
   );
 }
 
+def TestOpWithPropertiesAndAttr
+  : TEST_Op<"with_properties_and_attr"> {
+  let assemblyFormat = "$lhs prop-dict attr-dict";
+
+  let arguments = (ins I32Attr:$lhs, IntProperty<"int64_t">:$rhs);
+}
+
+def TestOpWithPropertiesAndInferredType
+  : TEST_Op<"with_properties_and_inferred_type", [
+    DeclareOpInterfaceMethods<InferTypeOpInterface>
+  ]> {
+  let assemblyFormat = "$lhs prop-dict attr-dict";
+
+  let arguments = (ins I32Attr:$lhs, IntProperty<"int64_t">:$rhs);
+  let results = (outs AnyType:$result);
+}
+
 // Demonstrate how to wrap an existing C++ class named MyPropStruct.
 def MyStructProperty : Property<"MyPropStruct"> {
   let convertToAttribute = "$_storage.asAttribute($_ctxt)";
@@ -2871,6 +2888,15 @@ def TestOpUsingPropertyInCustom : TEST_Op<"using_property_in_custom"> {
   let arguments = (ins ArrayProperty<"int64_t", 3>:$prop);
 }
 
+def TestOpUsingPropertyInCustomAndOther
+  : TEST_Op<"using_property_in_custom_and_other"> {
+  let assemblyFormat = "custom<UsingPropertyInCustom>($prop) prop-dict attr-dict";
+  let arguments = (ins
+    ArrayProperty<"int64_t", 3>:$prop,
+    IntProperty<"int64_t">:$other
+  );
+}
+
 def TestOpUsingPropertyRefInCustom : TEST_Op<"using_property_ref_in_custom"> {
   let assemblyFormat = "custom<IntProperty>($first) `+` custom<SumProperty>($second, ref($first)) attr-dict";
   let arguments = (ins IntProperty<"int64_t">:$first, IntProperty<"int64_t">:$second);
diff --git a/mlir/test/mlir-tblgen/op-format.mlir b/mlir/test/mlir-tblgen/op-format.mlir
index 14e1cdb07db39..46d272649caed 100644
--- a/mlir/test/mlir-tblgen/op-format.mlir
+++ b/mlir/test/mlir-tblgen/op-format.mlir
@@ -480,6 +480,17 @@ test.format_infer_variadic_type_from_non_variadic %i64, %i64 : i64
 // CHECK: test.format_infer_type_variadic_operands(%[[I32]], %[[I32]] : i32, i32) (%[[I64]], %[[I64]] : i64, i64)
 %ignored_res13:4 = test.format_infer_type_variadic_operands(%i32, %i32 : i32, i32) (%i64, %i64 : i64, i64)
 
+// CHECK: test.with_properties_and_attr 16 < {rhs = 16 : i64}>
+test.with_properties_and_attr 16 <{rhs = 16 : i64}>
+
+// CHECK: test.with_properties_and_inferred_type 16 < {rhs = 16 : i64}>
+%should_be_i32 = test.with_properties_and_inferred_type 16 <{rhs = 16 : i64}>
+// Assert through the verifier that its inferred as i32.
+test.format_all_types_match_var %should_be_i32, %i32 : i32
+
+// CHECK: test.using_property_in_custom_and_other [1, 4, 20] < {other = 16 : i64}>
+test.using_property_in_custom_and_other [1, 4, 20] <{other = 16 : i64}>
+
 //===----------------------------------------------------------------------===//
 // Check DefaultValuedStrAttr
 //===----------------------------------------------------------------------===//
diff --git a/mlir/tools/mlir-tblgen/OpFormatGen.cpp b/mlir/tools/mlir-tblgen/OpFormatGen.cpp
index 5963b5e689da7..806991035e668 100644
--- a/mlir/tools/mlir-tblgen/OpFormatGen.cpp
+++ b/mlir/tools/mlir-tblgen/OpFormatGen.cpp
@@ -379,8 +379,11 @@ struct OperationFormat {
   std::vector<TypeResolution> operandTypes, resultTypes;
 
   /// The set of attributes explicitly used within the format.
-  SmallVector<const NamedAttribute *, 8> usedAttributes;
+  llvm::SmallSetVector<const NamedAttribute *, 8> usedAttributes;
   llvm::StringSet<> inferredAttributes;
+
+  /// The set of properties explicitly used within the format.
+  llvm::SmallSetVector<const NamedProperty *, 8> usedProperties;
 };
 } // namespace
 
@@ -1183,6 +1186,105 @@ static void genAttrParser(AttributeVariable *attr, MethodBody &body,
   }
 }
 
+// Generates the 'setPropertiesFromParsedAttr' used to set properties from a
+// 'prop-dict' dictionary attr.
+static void genParsedAttrPropertiesSetter(OperationFormat &fmt, Operator &op,
+                                          OpClass &opClass) {
+  // Not required unless 'prop-dict' is present.
+  if (!fmt.hasPropDict)
+    return;
+
+  SmallVector<MethodParameter> paramList;
+  paramList.emplace_back("Properties &", "prop");
+  paramList.emplace_back("::mlir::Attribute", "attr");
+  paramList.emplace_back("::llvm::function_ref<::mlir::InFlightDiagnostic()>",
+                         "emitError");
+
+  Method *method = opClass.addStaticMethod("::mlir::LogicalResult",
+                                           "setPropertiesFromParsedAttr",
+                                           std::move(paramList));
+  MethodBody &body = method->body().indent();
+
+  body << R"decl(
+::mlir::DictionaryAttr dict = ::llvm::dyn_cast<::mlir::DictionaryAttr>(attr);
+if (!dict) {
+  emitError() << "expected DictionaryAttr to set properties";
+  return ::mlir::failure();
+}
+)decl";
+
+  // TODO: properties might be optional as well.
+  const char *propFromAttrFmt = R"decl(
+auto setFromAttr = [] (auto &propStorage, ::mlir::Attribute propAttr,
+         ::llvm::function_ref<::mlir::InFlightDiagnostic()> emitError) {{
+  {0};
+};
+auto attr = dict.get("{1}");
+if (!attr) {{
+  emitError() << "expected key entry for {1} in DictionaryAttr to set "
+             "Properties.";
+  return ::mlir::failure();
+}
+if (::mlir::failed(setFromAttr(prop.{1}, attr, emitError)))
+  return ::mlir::failure();
+)decl";
+
+  // Generate the setter for any property not parsed elsewhere.
+  for (const NamedProperty &namedProperty : op.getProperties()) {
+    if (fmt.usedProperties.contains(&namedProperty))
+      continue;
+
+    auto scope = body.scope("{\n", "}\n", /*indent=*/true);
+
+    StringRef name = namedProperty.name;
+    const Property &prop = namedProperty.prop;
+    FmtContext fctx;
+    body << formatv(propFromAttrFmt,
+                    tgfmt(prop.getConvertFromAttributeCall(),
+                          &fctx.addSubst("_attr", "propAttr")
+                               .addSubst("_storage", "propStorage")
+                               .addSubst("_diag", "emitError")),
+                    name);
+  }
+
+  // Generate the setter for any attribute not parsed elsewhere.
+  for (const NamedAttribute &namedAttr : op.getAttributes()) {
+    if (fmt.usedAttributes.contains(&namedAttr))
+      continue;
+
+    const Attribute &attr = namedAttr.attr;
+    // Derived attributes do not need to be parsed.
+    if (attr.isDerivedAttr())
+      continue;
+
+    auto scope = body.scope("{\n", "}\n", /*indent=*/true);
+
+    // If the attribute has a default value or is optional, it does not need to
+    // be present in the parsed dictionary attribute.
+    bool isRequired = !attr.isOptional() && !attr.hasDefaultValue();
+    body << formatv(R"decl(
+auto &propStorage = prop.{0};
+auto attr = dict.get("{0}");
+if (attr || /*isRequired=*/{1}) {{
+  if (!attr) {{
+    emitError() << "expected key entry for {0} in DictionaryAttr to set "
+               "Properties.";
+    return ::mlir::failure();
+  }
+  auto convertedAttr = ::llvm::dyn_cast<std::remove_reference_t<decltype(propStorage)>>(attr);
+  if (convertedAttr) {{
+    propStorage = convertedAttr;
+  } else {{
+    emitError() << "Invalid attribute `{0}` in property conversion: " << attr;
+    return ::mlir::failure();
+  }
+}
+)decl",
+                    namedAttr.name, isRequired);
+  }
+  body << "return ::mlir::success();\n";
+}
+
 void OperationFormat::genParser(Operator &op, OpClass &opClass) {
   SmallVector<MethodParameter> paramList;
   paramList.emplace_back("::mlir::OpAsmParser &", "parser");
@@ -1214,6 +1316,8 @@ void OperationFormat::genParser(Operator &op, OpClass &opClass) {
   genParserTypeResolution(op, body);
 
   body << "  return ::mlir::success();\n";
+
+  genParsedAttrPropertiesSetter(*this, op, opClass);
 }
 
 void OperationFormat::genElementParser(FormatElement *element, MethodBody &body,
@@ -1776,6 +1880,11 @@ const char *enumAttrBeginPrinterCode = R"(
 static void genPropDictPrinter(OperationFormat &fmt, Operator &op,
                                MethodBody &body) {
   body << "  ::llvm::SmallVector<::llvm::StringRef, 2> elidedProps;\n";
+  for (const NamedProperty *namedProperty : fmt.usedProperties)
+    body << "  elidedProps.push_back(\"" << namedProperty->name << "\");\n";
+  for (const NamedAttribute *namedAttr : fmt.usedAttributes)
+    body << "  elidedProps.push_back(\"" << namedAttr->name << "\");\n";
+
   // Add code to check attributes for equality with the default value
   // for attributes with the elidePrintingDefaultValue bit set.
   for (const NamedAttribute &namedAttr : op.getAttributes()) {
@@ -2543,7 +2652,7 @@ class OpFormatParser : public FormatParser {
   llvm::DenseSet<const NamedTypeConstraint *> seenOperands;
   llvm::DenseSet<const NamedRegion *> seenRegions;
   llvm::DenseSet<const NamedSuccessor *> seenSuccessors;
-  llvm::DenseSet<const NamedProperty *> seenProperties;
+  llvm::SmallSetVector<const NamedProperty *, 8> seenProperties;
 };
 } // namespace
 
@@ -2589,7 +2698,8 @@ LogicalResult OpFormatParser::verify(SMLoc loc,
     return failure();
 
   // Collect the set of used attributes in the format.
-  fmt.usedAttributes = seenAttrs.takeVector();
+  fmt.usedAttributes = std::move(seenAttrs);
+  fmt.usedProperties = std::move(seenProperties);
 
   // Set whether prop-dict is used in the format
   fmt.hasPropDict = hasPropDict;
@@ -3042,7 +3152,7 @@ OpFormatParser::parseVariableImpl(SMLoc loc, StringRef name, Context ctx) {
         return emitError(loc, "property '" + name +
                                   "' must be bound before it is referenced");
     } else {
-      if (!seenProperties.insert(property).second)
+      if (!seenProperties.insert(property))
         return emitError(loc, "property '" + name + "' is already bound");
     }
 

From 7dbba39e583a3fd64e7e6b947251c035e483f054 Mon Sep 17 00:00:00 2001
From: Julian Nagele <j.nagele@apple.com>
Date: Mon, 15 Apr 2024 11:25:05 +0100
Subject: [PATCH 409/886] Reapply "[TBAA] Add verifier for tbaa.struct metadata
 (#86709)"

This reverts commit b9cd48f96acdd07c627ccafbf4386a1f3dcd6c51.

-------------------------------------------------------------
Original commit message:

Adds logic to the IR verifier that checks whether !tbaa.struct nodes are
well-formed. That is, it checks that the operands of !tbaa.struct nodes
are in groups of three, that each group of three operands consists of
two integers and a valid tbaa node, and that the regions described by
the offset and size operands are non-overlapping.

PR: https://github.com/llvm/llvm-project/pull/86709
---
 llvm/include/llvm/IR/Verifier.h               |  1 +
 llvm/lib/IR/Verifier.cpp                      | 32 +++++++++++++++++++
 llvm/test/CodeGen/AArch64/arm64-abi_align.ll  |  4 ++-
 .../AMDGPU/mem-intrinsics.ll                  |  2 +-
 .../InstCombine/struct-assign-tbaa.ll         |  2 +-
 llvm/test/Transforms/SROA/tbaa-struct3.ll     |  2 +-
 .../Scalarizer/basic-inseltpoison.ll          |  3 +-
 llvm/test/Transforms/Scalarizer/basic.ll      |  3 +-
 llvm/test/Verifier/tbaa-struct.ll             | 14 ++++++--
 9 files changed, 54 insertions(+), 9 deletions(-)

diff --git a/llvm/include/llvm/IR/Verifier.h b/llvm/include/llvm/IR/Verifier.h
index b25f8eb77ee38..b7db6e0bbfb71 100644
--- a/llvm/include/llvm/IR/Verifier.h
+++ b/llvm/include/llvm/IR/Verifier.h
@@ -77,6 +77,7 @@ class TBAAVerifier {
   /// Visit an instruction and return true if it is valid, return false if an
   /// invalid TBAA is attached.
   bool visitTBAAMetadata(Instruction &I, const MDNode *MD);
+  bool visitTBAAStructMetadata(Instruction &I, const MDNode *MD);
 };
 
 /// Check a function for errors, useful for use when debugging a
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 25cb99f0f9bd8..516d4a0515569 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -5122,6 +5122,9 @@ void Verifier::visitInstruction(Instruction &I) {
   if (MDNode *TBAA = I.getMetadata(LLVMContext::MD_tbaa))
     TBAAVerifyHelper.visitTBAAMetadata(I, TBAA);
 
+  if (MDNode *TBAA = I.getMetadata(LLVMContext::MD_tbaa_struct))
+    TBAAVerifyHelper.visitTBAAStructMetadata(I, TBAA);
+
   if (MDNode *MD = I.getMetadata(LLVMContext::MD_noalias))
     visitAliasScopeListMetadata(MD);
   if (MDNode *MD = I.getMetadata(LLVMContext::MD_alias_scope))
@@ -7453,6 +7456,35 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
   return true;
 }
 
+bool TBAAVerifier::visitTBAAStructMetadata(Instruction &I, const MDNode *MD) {
+  CheckTBAA(MD->getNumOperands() % 3 == 0,
+            "tbaa.struct operands must occur in groups of three", &I, MD);
+
+  // Each group of three operands must consist of two integers and a
+  // tbaa node. Moreover, the regions described by the offset and size
+  // operands must be non-overlapping.
+  std::optional<APInt> NextFree;
+  for (unsigned int Idx = 0; Idx < MD->getNumOperands(); Idx += 3) {
+    auto *OffsetCI =
+        mdconst::dyn_extract_or_null<ConstantInt>(MD->getOperand(Idx));
+    CheckTBAA(OffsetCI, "Offset must be a constant integer", &I, MD);
+
+    auto *SizeCI =
+        mdconst::dyn_extract_or_null<ConstantInt>(MD->getOperand(Idx + 1));
+    CheckTBAA(SizeCI, "Size must be a constant integer", &I, MD);
+
+    MDNode *TBAA = dyn_cast_or_null<MDNode>(MD->getOperand(Idx + 2));
+    CheckTBAA(TBAA, "TBAA tag missing", &I, MD);
+    visitTBAAMetadata(I, TBAA);
+
+    bool NonOverlapping = !NextFree || NextFree->ule(OffsetCI->getValue());
+    CheckTBAA(NonOverlapping, "Overlapping tbaa.struct regions", &I, MD);
+
+    NextFree = OffsetCI->getValue() + SizeCI->getValue();
+  }
+  return true;
+}
+
 char VerifierLegacyPass::ID = 0;
 INITIALIZE_PASS(VerifierLegacyPass, "verify", "Module Verifier", false, false)
 
diff --git a/llvm/test/CodeGen/AArch64/arm64-abi_align.ll b/llvm/test/CodeGen/AArch64/arm64-abi_align.ll
index 089e171e5a4a7..c9fd2d38e27ac 100644
--- a/llvm/test/CodeGen/AArch64/arm64-abi_align.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-abi_align.ll
@@ -518,4 +518,6 @@ attributes #5 = { nobuiltin }
 !1 = !{!"omnipotent char", !2}
 !2 = !{!"Simple C/C++ TBAA"}
 !3 = !{!"short", !1}
-!4 = !{i64 0, i64 4, !0, i64 4, i64 2, !3, i64 8, i64 4, !0, i64 12, i64 2, !3, i64 16, i64 4, !0, i64 20, i64 2, !3}
+!4 = !{i64 0, i64 4, !5, i64 4, i64 2, !6, i64 8, i64 4, !5, i64 12, i64 2, !6, i64 16, i64 4, !5, i64 20, i64 2, !6}
+!5 = !{!0, !0, i64 0}
+!6 = !{!3, !3, i64 0}
diff --git a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll
index 50b0e7a0f5471..2f264a2432fc3 100644
--- a/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll
+++ b/llvm/test/Transforms/InferAddressSpaces/AMDGPU/mem-intrinsics.ll
@@ -141,4 +141,4 @@ attributes #1 = { argmemonly nounwind }
 !5 = distinct !{!5, !"some domain"}
 !6 = !{!7}
 !7 = distinct !{!7, !5, !"some scope 2"}
-!8 = !{i64 0, i64 8, null}
+!8 = !{i64 0, i64 8, !0}
diff --git a/llvm/test/Transforms/InstCombine/struct-assign-tbaa.ll b/llvm/test/Transforms/InstCombine/struct-assign-tbaa.ll
index 996d2c0e67e16..d079c03f1dcb9 100644
--- a/llvm/test/Transforms/InstCombine/struct-assign-tbaa.ll
+++ b/llvm/test/Transforms/InstCombine/struct-assign-tbaa.ll
@@ -75,7 +75,7 @@ entry:
 !1 = !{!"omnipotent char", !0}
 !2 = !{!5, !5, i64 0}
 !3 = !{i64 0, i64 4, !2}
-!4 = !{i64 0, i64 8, null}
+!4 = !{i64 0, i64 8, !2}
 !5 = !{!"float", !0}
 !6 = !{i64 0, i64 4, !2, i64 4, i64 4, !2}
 !7 = !{i64 0, i64 2, !2, i64 4, i64 6, !2}
diff --git a/llvm/test/Transforms/SROA/tbaa-struct3.ll b/llvm/test/Transforms/SROA/tbaa-struct3.ll
index 0fcd787fef976..61034de81e4b2 100644
--- a/llvm/test/Transforms/SROA/tbaa-struct3.ll
+++ b/llvm/test/Transforms/SROA/tbaa-struct3.ll
@@ -539,7 +539,7 @@ declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias
 !6 = !{!5, !5, i64 0}
 !7 = !{i64 0, i64 8, !6, i64 8, i64 4, !1}
 !8 = !{i64 0, i64 4, !1, i64 4, i64 8, !6}
-!9 = !{i64 0, i64 8, !6, i64 4, i64 8, !1}
+!9 = !{i64 0, i64 8, !6, i64 8, i64 8, !1}
 !10 = !{i64 0, i64 2, !1, i64 2, i64 2, !1}
 !11 = !{i64 0, i64 1, !1, i64 1, i64 3, !1}
 !12 = !{i64 0, i64 2, !1, i64 2, i64 6, !1}
diff --git a/llvm/test/Transforms/Scalarizer/basic-inseltpoison.ll b/llvm/test/Transforms/Scalarizer/basic-inseltpoison.ll
index bbcdcb6f58674..73ae66dd76c66 100644
--- a/llvm/test/Transforms/Scalarizer/basic-inseltpoison.ll
+++ b/llvm/test/Transforms/Scalarizer/basic-inseltpoison.ll
@@ -836,5 +836,6 @@ define <2 x i32> @f23_crash(<2 x i32> %srcvec, i32 %v1) {
 !2 = !{ !"set2", !0 }
 !3 = !{ !3, !{!"llvm.loop.parallel_accesses", !13} }
 !4 = !{ float 4.0 }
-!5 = !{ i64 0, i64 8, null }
+!5 = !{ i64 0, i64 8, !6 }
+!6 = !{ !1, !1, i64 0 }
 !13 = distinct !{}
diff --git a/llvm/test/Transforms/Scalarizer/basic.ll b/llvm/test/Transforms/Scalarizer/basic.ll
index db7c5f535f7e9..87a70ccd3fc7c 100644
--- a/llvm/test/Transforms/Scalarizer/basic.ll
+++ b/llvm/test/Transforms/Scalarizer/basic.ll
@@ -870,5 +870,6 @@ define <2 x float> @f25(<2 x float> %src) {
 !2 = !{ !"set2", !0 }
 !3 = !{ !3, !{!"llvm.loop.parallel_accesses", !13} }
 !4 = !{ float 4.0 }
-!5 = !{ i64 0, i64 8, null }
+!5 = !{ i64 0, i64 8, !6 }
+!6 = !{ !1, !1, i64 0 }
 !13 = distinct !{}
diff --git a/llvm/test/Verifier/tbaa-struct.ll b/llvm/test/Verifier/tbaa-struct.ll
index b8ddc7cee496a..14c19a19d5ae8 100644
--- a/llvm/test/Verifier/tbaa-struct.ll
+++ b/llvm/test/Verifier/tbaa-struct.ll
@@ -1,28 +1,36 @@
-; RUN: llvm-as < %s 2>&1
-
-; FIXME: The verifer should reject the invalid !tbaa.struct nodes below.
+; RUN: not llvm-as < %s 2>&1 | FileCheck %s
 
 define void @test_overlapping_regions(ptr %a1) {
+; CHECK: Overlapping tbaa.struct regions
+; CHECK-NEXT:  %ld = load i8, ptr %a1, align 1, !tbaa.struct !0
   %ld = load i8, ptr %a1, align 1, !tbaa.struct !0
   ret void
 }
 
 define void @test_size_not_integer(ptr %a1) {
+; CHECK: Size must be a constant integer
+; CHECK-NEXT:  store i8 1, ptr %a1, align 1, !tbaa.struct !5
   store i8 1, ptr %a1, align 1, !tbaa.struct !5
   ret void
 }
 
 define void @test_offset_not_integer(ptr %a1, ptr %a2) {
+; CHECK: Offset must be a constant integer
+; CHECK-NEXT:  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 %a1, ptr align 8 %a2, i64 16, i1 false), !tbaa.struct !6
   tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 %a1, ptr align 8 %a2, i64 16, i1 false), !tbaa.struct !6
   ret void
 }
 
 define void @test_tbaa_missing(ptr %a1, ptr %a2) {
+; CHECK: TBAA tag missing
+; CHECK-NEXT:  tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 %a1, ptr align 8 %a2, i64 16, i1 false), !tbaa.struct !7
   tail call void @llvm.memcpy.p0.p0.i64(ptr align 8 %a1, ptr align 8 %a2, i64 16, i1 false), !tbaa.struct !7
   ret void
 }
 
 define void @test_tbaa_invalid(ptr %a1) {
+; CHECK: Old-style TBAA is no longer allowed, use struct-path TBAA instead
+; CHECK-NEXT:  store i8 1, ptr %a1, align 1, !tbaa.struct !8
   store i8 1, ptr %a1, align 1, !tbaa.struct !8
   ret void
 }

From c006b9088fa564fe343550a8dd5f9b0ad0c15b75 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Mon, 15 Apr 2024 11:30:36 +0100
Subject: [PATCH 410/886] [VPlan] Add test for dump of recipe not inserted in
 VPlan. (NFC)

Add extra tests for printing recipes not inserted in a VPlan yet, e.g.
when using a debugger.

Guard against regressions in changes to printing, i.e.
https://github.com/llvm/llvm-project/pull/81411.
---
 .../Transforms/Vectorize/VPlanTest.cpp        | 163 +++++++++++++++++-
 1 file changed, 162 insertions(+), 1 deletion(-)

diff --git a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
index 73687846e281a..ff0f482b833c8 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
@@ -1227,7 +1227,7 @@ TEST(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-TEST(VPRecipeTest, dump) {
+TEST(VPRecipeTest, dumpRecipeInPlan) {
   VPBasicBlock *VPBB0 = new VPBasicBlock("preheader");
   VPBasicBlock *VPBB1 = new VPBasicBlock();
   VPlan Plan(VPBB0, VPBB1);
@@ -1280,6 +1280,167 @@ TEST(VPRecipeTest, dump) {
 
   delete AI;
 }
+
+TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
+  VPBasicBlock *VPBB0 = new VPBasicBlock("preheader");
+  VPBasicBlock *VPBB1 = new VPBasicBlock();
+  VPlan Plan(VPBB0, VPBB1);
+
+  LLVMContext C;
+
+  IntegerType *Int32 = IntegerType::get(C, 32);
+  SmallVector<VPValue *, 2> Args;
+  VPValue *ExtVPV1 = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 1));
+  VPValue *ExtVPV2 = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 2));
+  Args.push_back(ExtVPV1);
+  Args.push_back(ExtVPV2);
+  VPInstruction *I1 = new VPInstruction(Instruction::Add, {ExtVPV1, ExtVPV2});
+  VPInstruction *I2 = new VPInstruction(Instruction::Mul, {I1, I1});
+  VPBB1->appendRecipe(I1);
+  VPBB1->appendRecipe(I2);
+
+  // Check printing I1.
+  {
+    // Use EXPECT_EXIT to capture stderr and compare against expected output.
+    //
+    // Test VPValue::dump().
+    VPValue *VPV = I1;
+    EXPECT_EXIT(
+        {
+          VPV->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+
+    // Test VPRecipeBase::dump().
+    VPRecipeBase *R = I1;
+    EXPECT_EXIT(
+        {
+          R->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+
+    // Test VPDef::dump().
+    VPDef *D = I1;
+    EXPECT_EXIT(
+        {
+          D->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+  }
+  // Check printing I2.
+  {
+    // Use EXPECT_EXIT to capture stderr and compare against expected output.
+    //
+    // Test VPValue::dump().
+    VPValue *VPV = I2;
+    EXPECT_EXIT(
+        {
+          VPV->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%2> = mul vp<%1>, vp<%1>");
+
+    // Test VPRecipeBase::dump().
+    VPRecipeBase *R = I2;
+    EXPECT_EXIT(
+        {
+          R->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%2> = mul vp<%1>, vp<%1>");
+
+    // Test VPDef::dump().
+    VPDef *D = I2;
+    EXPECT_EXIT(
+        {
+          D->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT vp<%2> = mul vp<%1>, vp<%1>");
+  }
+}
+
+TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
+  LLVMContext C;
+  IntegerType *Int32 = IntegerType::get(C, 32);
+  VPValue *ExtVPV1 = new VPValue(ConstantInt::get(Int32, 1));
+  VPValue *ExtVPV2 = new VPValue(ConstantInt::get(Int32, 2));
+
+  VPInstruction *I1 = new VPInstruction(Instruction::Add, {ExtVPV1, ExtVPV2});
+  VPInstruction *I2 = new VPInstruction(Instruction::Mul, {I1, I1});
+
+  // Check printing I1.
+  {
+    // Use EXPECT_EXIT to capture stderr and compare against expected output.
+    //
+    // Test VPValue::dump().
+    VPValue *VPV = I1;
+    EXPECT_EXIT(
+        {
+          VPV->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+
+    // Test VPRecipeBase::dump().
+    VPRecipeBase *R = I1;
+    EXPECT_EXIT(
+        {
+          R->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+
+    // Test VPDef::dump().
+    VPDef *D = I1;
+    EXPECT_EXIT(
+        {
+          D->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+  }
+  // Check printing I2.
+  {
+    // Use EXPECT_EXIT to capture stderr and compare against expected output.
+    //
+    // Test VPValue::dump().
+    VPValue *VPV = I2;
+    EXPECT_EXIT(
+        {
+          VPV->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = mul <badref>, <badref>");
+
+    // Test VPRecipeBase::dump().
+    VPRecipeBase *R = I2;
+    EXPECT_EXIT(
+        {
+          R->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = mul <badref>, <badref>");
+
+    // Test VPDef::dump().
+    VPDef *D = I2;
+    EXPECT_EXIT(
+        {
+          D->dump();
+          exit(0);
+        },
+        testing::ExitedWithCode(0), "EMIT <badref> = mul <badref>, <badref>");
+  }
+
+  delete I2;
+  delete I1;
+  delete ExtVPV2;
+  delete ExtVPV1;
+}
+
 #endif
 
 TEST(VPRecipeTest, CastVPReductionRecipeToVPUser) {

From e060cb26d9b503ea5138909ec5c7099e7d4499c2 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Mon, 15 Apr 2024 11:47:44 +0100
Subject: [PATCH 411/886] [X86] Add isCompletePermute helper for single input
 shuffles where we don't lose/reuse any vector element. NFC

Matches purely reorder shuffles.
---
 llvm/lib/Target/X86/X86ISelLowering.cpp | 34 +++++++++++++------------
 1 file changed, 18 insertions(+), 16 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 87f2eaddedf86..97a79d4b2bcb4 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -3596,6 +3596,17 @@ static bool isUndefOrZeroInRange(ArrayRef<int> Mask, unsigned Pos,
   return llvm::all_of(Mask.slice(Pos, Size), isUndefOrZero);
 }
 
+/// Return true if every element of a single input is referenced by the shuffle
+/// mask. i.e. it just permutes them all.
+static bool isCompletePermute(ArrayRef<int> Mask) {
+  unsigned NumElts = Mask.size();
+  APInt DemandedElts = APInt::getZero(NumElts);
+  for (int M : Mask)
+    if (isInRange(M, 0, NumElts))
+      DemandedElts.setBit(M);
+  return DemandedElts.isAllOnes();
+}
+
 /// Helper function to test whether a shuffle mask could be
 /// simplified by widening the elements being shuffled.
 ///
@@ -46469,24 +46480,15 @@ static SDValue combineSetCCMOVMSK(SDValue EFLAGS, X86::CondCode &CC,
   if (NumElts <= CmpBits &&
       getTargetShuffleInputs(peekThroughBitcasts(Vec), ShuffleInputs,
                              ShuffleMask, DAG) &&
-      ShuffleInputs.size() == 1 && !isAnyZeroOrUndef(ShuffleMask) &&
+      ShuffleInputs.size() == 1 && isCompletePermute(ShuffleMask) &&
       ShuffleInputs[0].getValueSizeInBits() == VecVT.getSizeInBits() &&
       scaleShuffleElements(ShuffleMask, NumElts, ScaledMaskUnused)) {
-    unsigned NumShuffleElts = ShuffleMask.size();
-    APInt DemandedElts = APInt::getZero(NumShuffleElts);
-    for (int M : ShuffleMask) {
-      assert(0 <= M && M < (int)NumShuffleElts && "Bad unary shuffle index");
-      DemandedElts.setBit(M);
-    }
-    if (DemandedElts.isAllOnes()) {
-      SDLoc DL(EFLAGS);
-      SDValue Result = DAG.getBitcast(VecVT, ShuffleInputs[0]);
-      Result = DAG.getNode(X86ISD::MOVMSK, DL, MVT::i32, Result);
-      Result =
-          DAG.getZExtOrTrunc(Result, DL, EFLAGS.getOperand(0).getValueType());
-      return DAG.getNode(X86ISD::CMP, DL, MVT::i32, Result,
-                         EFLAGS.getOperand(1));
-    }
+    SDLoc DL(EFLAGS);
+    SDValue Result = DAG.getBitcast(VecVT, ShuffleInputs[0]);
+    Result = DAG.getNode(X86ISD::MOVMSK, DL, MVT::i32, Result);
+    Result =
+        DAG.getZExtOrTrunc(Result, DL, EFLAGS.getOperand(0).getValueType());
+    return DAG.getNode(X86ISD::CMP, DL, MVT::i32, Result, EFLAGS.getOperand(1));
   }
 
   // MOVMSKPS(V) !=/== 0 -> TESTPS(V,V)

From 4859a593f4016f0a78e2fadf21722e370560c5e9 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Mon, 15 Apr 2024 11:50:56 +0100
Subject: [PATCH 412/886] [VPlan] Extend recipe printing tests with underlying
 IR values. (NFC)

Make tests added in c006b9088fa56 more interesting by using an
underlying IR value as well.
---
 .../Transforms/Vectorize/VPlanTest.cpp        | 24 ++++++++++++-------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
index ff0f482b833c8..cb8737a9e64d2 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
@@ -1289,9 +1289,12 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
   LLVMContext C;
 
   IntegerType *Int32 = IntegerType::get(C, 32);
+  auto *AI =
+      BinaryOperator::CreateAdd(UndefValue::get(Int32), UndefValue::get(Int32));
+  AI->setName("a");
   SmallVector<VPValue *, 2> Args;
   VPValue *ExtVPV1 = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 1));
-  VPValue *ExtVPV2 = Plan.getOrAddLiveIn(ConstantInt::get(Int32, 2));
+  VPValue *ExtVPV2 = Plan.getOrAddLiveIn(AI);
   Args.push_back(ExtVPV1);
   Args.push_back(ExtVPV2);
   VPInstruction *I1 = new VPInstruction(Instruction::Add, {ExtVPV1, ExtVPV2});
@@ -1310,7 +1313,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
           VPV->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<%a>");
 
     // Test VPRecipeBase::dump().
     VPRecipeBase *R = I1;
@@ -1319,7 +1322,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
           R->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<%a>");
 
     // Test VPDef::dump().
     VPDef *D = I1;
@@ -1328,7 +1331,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
           D->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT vp<%1> = add ir<1>, ir<%a>");
   }
   // Check printing I2.
   {
@@ -1361,13 +1364,17 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesInPlan) {
         },
         testing::ExitedWithCode(0), "EMIT vp<%2> = mul vp<%1>, vp<%1>");
   }
+  delete AI;
 }
 
 TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
   LLVMContext C;
   IntegerType *Int32 = IntegerType::get(C, 32);
+  auto *AI =
+      BinaryOperator::CreateAdd(UndefValue::get(Int32), UndefValue::get(Int32));
+  AI->setName("a");
   VPValue *ExtVPV1 = new VPValue(ConstantInt::get(Int32, 1));
-  VPValue *ExtVPV2 = new VPValue(ConstantInt::get(Int32, 2));
+  VPValue *ExtVPV2 = new VPValue(AI);
 
   VPInstruction *I1 = new VPInstruction(Instruction::Add, {ExtVPV1, ExtVPV2});
   VPInstruction *I2 = new VPInstruction(Instruction::Mul, {I1, I1});
@@ -1383,7 +1390,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
           VPV->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<%a>");
 
     // Test VPRecipeBase::dump().
     VPRecipeBase *R = I1;
@@ -1392,7 +1399,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
           R->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<%a>");
 
     // Test VPDef::dump().
     VPDef *D = I1;
@@ -1401,7 +1408,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
           D->dump();
           exit(0);
         },
-        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<2>");
+        testing::ExitedWithCode(0), "EMIT <badref> = add ir<1>, ir<%a>");
   }
   // Check printing I2.
   {
@@ -1439,6 +1446,7 @@ TEST(VPRecipeTest, dumpRecipeUnnamedVPValuesNotInPlanOrBlock) {
   delete I1;
   delete ExtVPV2;
   delete ExtVPV1;
+  delete AI;
 }
 
 #endif

From f78b3466caa9296b32ec235dee87ace2dea94507 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Mon, 15 Apr 2024 12:51:29 +0200
Subject: [PATCH 413/886] ValueTracking: Treat poison more aggressively in
 computeKnownFPClass (#87990)

Assume no valid values, and the sign bit is 0.
---
 llvm/lib/Analysis/ValueTracking.cpp           |  6 ++
 llvm/test/CodeGen/AMDGPU/fold-fabs.ll         |  6 --
 llvm/test/Transforms/Attributor/nofpclass.ll  | 12 ++++
 llvm/unittests/Analysis/ValueTrackingTest.cpp | 55 +++++++++++++++++++
 4 files changed, 73 insertions(+), 6 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 5beea614e332e..b3abf016cfb93 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -4664,6 +4664,12 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,
     return;
   }
 
+  if (isa<PoisonValue>(V)) {
+    Known.KnownFPClasses = fcNone;
+    Known.SignBit = false;
+    return;
+  }
+
   // Try to handle fixed width vector constants
   auto *VFVTy = dyn_cast<FixedVectorType>(V->getType());
   const Constant *CV = dyn_cast<Constant>(V);
diff --git a/llvm/test/CodeGen/AMDGPU/fold-fabs.ll b/llvm/test/CodeGen/AMDGPU/fold-fabs.ll
index bb2bad9d3521a..a04bf44549325 100644
--- a/llvm/test/CodeGen/AMDGPU/fold-fabs.ll
+++ b/llvm/test/CodeGen/AMDGPU/fold-fabs.ll
@@ -99,12 +99,6 @@ define float @fold_abs_in_branch_poison(float %arg1, float %arg2) {
 ; GFX10-LABEL: fold_abs_in_branch_poison:
 ; GFX10:       ; %bb.0: ; %entry
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10-NEXT:    v_add_f32_e64 v0, |s4|, |s4|
-; GFX10-NEXT:    v_cmp_lt_f32_e32 vcc_lo, 1.0, v0
-; GFX10-NEXT:    s_cbranch_vccnz .LBB3_2
-; GFX10-NEXT:  ; %bb.1: ; %if
-; GFX10-NEXT:    v_mul_f32_e64 v0, 0x3e4ccccd, |s4|
-; GFX10-NEXT:  .LBB3_2: ; %exit
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 entry:
   %0 = fadd reassoc nnan nsz arcp contract afn float %arg1, %arg2
diff --git a/llvm/test/Transforms/Attributor/nofpclass.ll b/llvm/test/Transforms/Attributor/nofpclass.ll
index 7828629f9fc01..d2d11e0276c4d 100644
--- a/llvm/test/Transforms/Attributor/nofpclass.ll
+++ b/llvm/test/Transforms/Attributor/nofpclass.ll
@@ -54,6 +54,18 @@ define float @returned_poison() {
   ret float poison
 }
 
+; Know nothing
+define float @returned_freeze_poison() {
+; CHECK-LABEL: define noundef float @returned_freeze_poison() {
+; CHECK-NEXT:    call void @unknown()
+; CHECK-NEXT:    [[FREEZE_POISON:%.*]] = freeze float poison
+; CHECK-NEXT:    ret float [[FREEZE_POISON]]
+;
+  call void @unknown()
+  %freeze.poison = freeze float poison
+  ret float %freeze.poison
+}
+
 define double @returned_snan() {
 ; CHECK-LABEL: define noundef nofpclass(qnan inf zero sub norm) double @returned_snan() {
 ; CHECK-NEXT:    call void @unknown()
diff --git a/llvm/unittests/Analysis/ValueTrackingTest.cpp b/llvm/unittests/Analysis/ValueTrackingTest.cpp
index 6b1044876d366..8ebd9b511f39f 100644
--- a/llvm/unittests/Analysis/ValueTrackingTest.cpp
+++ b/llvm/unittests/Analysis/ValueTrackingTest.cpp
@@ -2035,6 +2035,61 @@ TEST_F(ComputeKnownFPClassTest, Constants) {
     ASSERT_TRUE(ConstAggZero.SignBit);
     EXPECT_FALSE(*ConstAggZero.SignBit);
   }
+
+  {
+    KnownFPClass Undef =
+        computeKnownFPClass(UndefValue::get(F32), M->getDataLayout(),
+                            fcAllFlags, 0, nullptr, nullptr, nullptr, nullptr);
+    EXPECT_EQ(fcAllFlags, Undef.KnownFPClasses);
+    EXPECT_FALSE(Undef.SignBit);
+  }
+
+  {
+    KnownFPClass Poison =
+        computeKnownFPClass(PoisonValue::get(F32), M->getDataLayout(),
+                            fcAllFlags, 0, nullptr, nullptr, nullptr, nullptr);
+    EXPECT_EQ(fcNone, Poison.KnownFPClasses);
+    ASSERT_TRUE(Poison.SignBit);
+    EXPECT_FALSE(*Poison.SignBit);
+  }
+
+  {
+    // Assume the poison element should be 0.
+    Constant *ZeroF32 = ConstantFP::getZero(F32);
+    Constant *PoisonF32 = PoisonValue::get(F32);
+
+    KnownFPClass PartiallyPoison = computeKnownFPClass(
+        ConstantVector::get({ZeroF32, PoisonF32}), M->getDataLayout(),
+        fcAllFlags, 0, nullptr, nullptr, nullptr, nullptr);
+    EXPECT_EQ(fcPosZero, PartiallyPoison.KnownFPClasses);
+    ASSERT_TRUE(PartiallyPoison.SignBit);
+    EXPECT_FALSE(*PartiallyPoison.SignBit);
+  }
+
+  {
+    // Assume the poison element should be 1.
+    Constant *NegZeroF32 = ConstantFP::getZero(F32, true);
+    Constant *PoisonF32 = PoisonValue::get(F32);
+
+    KnownFPClass PartiallyPoison = computeKnownFPClass(
+        ConstantVector::get({NegZeroF32, PoisonF32}), M->getDataLayout(),
+        fcAllFlags, 0, nullptr, nullptr, nullptr, nullptr);
+    EXPECT_EQ(fcNegZero, PartiallyPoison.KnownFPClasses);
+    ASSERT_TRUE(PartiallyPoison.SignBit);
+    EXPECT_TRUE(*PartiallyPoison.SignBit);
+  }
+
+  {
+    // Assume the poison element should be 1.
+    Constant *NegZeroF32 = ConstantFP::getZero(F32, true);
+    Constant *PoisonF32 = PoisonValue::get(F32);
+
+    KnownFPClass PartiallyPoison = computeKnownFPClass(
+        ConstantVector::get({PoisonF32, NegZeroF32}), M->getDataLayout(),
+        fcAllFlags, 0, nullptr, nullptr, nullptr, nullptr);
+    EXPECT_EQ(fcNegZero, PartiallyPoison.KnownFPClasses);
+    EXPECT_TRUE(PartiallyPoison.SignBit);
+  }
 }
 
 TEST_F(ValueTrackingTest, isNonZeroRecurrence) {

From 6acc50ae5b5971af563a0f40b5f0279870171541 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Mon, 15 Apr 2024 11:50:11 +0100
Subject: [PATCH 414/886] Fix typo in
 computeKnownBits_AVGFLOORU_AVGFLOORS_AVGCEILU_AVGCEILS.

ISD::AVGFLOORU -> ISD::AVGFLOORS
---
 llvm/unittests/CodeGen/AArch64SelectionDAGTest.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/unittests/CodeGen/AArch64SelectionDAGTest.cpp b/llvm/unittests/CodeGen/AArch64SelectionDAGTest.cpp
index 27bcad7c24c4d..e498a2f60bad3 100644
--- a/llvm/unittests/CodeGen/AArch64SelectionDAGTest.cpp
+++ b/llvm/unittests/CodeGen/AArch64SelectionDAGTest.cpp
@@ -826,7 +826,7 @@ TEST_F(AArch64SelectionDAGTest,
   EXPECT_EQ(KnownAVGFLOORU.One, Ones);
 
   SDValue AVGFLOORS =
-      DAG->getNode(ISD::AVGFLOORU, Loc, Int16Vec8VT, ZextOp0, ZextOp1);
+      DAG->getNode(ISD::AVGFLOORS, Loc, Int16Vec8VT, ZextOp0, ZextOp1);
   KnownBits KnownAVGFLOORS = DAG->computeKnownBits(AVGFLOORS);
   EXPECT_EQ(KnownAVGFLOORS.Zero, Zeroes);
   EXPECT_EQ(KnownAVGFLOORS.One, Ones);

From 14774ad59da4aaa8eb3da21c8c53b59834a1aeb7 Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Mon, 15 Apr 2024 19:01:28 +0800
Subject: [PATCH 415/886] [X86] Remove unused function (NFC)

llvm-project/llvm/lib/Target/X86/X86ISelLowering.cpp:3537:13:
error: unused function 'isAnyZeroOrUndef' [-Werror,-Wunused-function]
static bool isAnyZeroOrUndef(ArrayRef<int> Mask) {
            ^
1 error generated.
---
 llvm/lib/Target/X86/X86ISelLowering.cpp | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 97a79d4b2bcb4..0a91e3f554b3d 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -3532,14 +3532,6 @@ static bool isAnyZero(ArrayRef<int> Mask) {
   return llvm::any_of(Mask, [](int M) { return M == SM_SentinelZero; });
 }
 
-/// Return true if the value of any element in Mask is the zero or undef
-/// sentinel values.
-static bool isAnyZeroOrUndef(ArrayRef<int> Mask) {
-  return llvm::any_of(Mask, [](int M) {
-    return M == SM_SentinelZero || M == SM_SentinelUndef;
-  });
-}
-
 /// Return true if Val is undef or if its value falls within the
 /// specified range (L, H].
 static bool isUndefOrInRange(int Val, int Low, int Hi) {

From 9da0ef165efd91b862b8c3cff1604472476839d6 Mon Sep 17 00:00:00 2001
From: Johannes Reifferscheid <jreiffers@google.com>
Date: Mon, 15 Apr 2024 13:13:01 +0200
Subject: [PATCH 416/886] Fix complex tanh overflows. (#88708)

This ports the XLA lowering and was verified using XLA's
exhaustive_unary_test_complex test.
---
 .../ComplexToStandard/ComplexToStandard.cpp   |  90 ++++++++++++----
 .../convert-to-standard.mlir                  | 102 +++++++++++++++---
 2 files changed, 157 insertions(+), 35 deletions(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 3ebee9baff31b..03e578136e590 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -978,30 +978,84 @@ struct TanhOpConversion : public OpConversionPattern<complex::TanhOp> {
   LogicalResult
   matchAndRewrite(complex::TanhOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
+    ImplicitLocOpBuilder b(op.getLoc(), rewriter);
     auto loc = op.getLoc();
     auto type = cast<ComplexType>(adaptor.getComplex().getType());
     auto elementType = cast<FloatType>(type.getElementType());
-    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
+    arith::FastMathFlags fmf = op.getFastMathFlagsAttr().getValue();
+    const auto &floatSemantics = elementType.getFloatSemantics();
 
-    // The hyperbolic tangent for complex number can be calculated as follows.
-    // tanh(x + i * y) = (tanh(x) + i * tan(y)) / (1 + tanh(x) * tan(y))
-    // See: https://proofwiki.org/wiki/Hyperbolic_Tangent_of_Complex_Number
     Value real =
-        rewriter.create<complex::ReOp>(loc, elementType, adaptor.getComplex());
+        b.create<complex::ReOp>(loc, elementType, adaptor.getComplex());
     Value imag =
-        rewriter.create<complex::ImOp>(loc, elementType, adaptor.getComplex());
-    Value tanhA = rewriter.create<math::TanhOp>(loc, real, fmf);
-    Value cosB = rewriter.create<math::CosOp>(loc, imag, fmf);
-    Value sinB = rewriter.create<math::SinOp>(loc, imag, fmf);
-    Value tanB = rewriter.create<arith::DivFOp>(loc, sinB, cosB, fmf);
-    Value numerator =
-        rewriter.create<complex::CreateOp>(loc, type, tanhA, tanB);
-    Value one = rewriter.create<arith::ConstantOp>(
-        loc, elementType, rewriter.getFloatAttr(elementType, 1));
-    Value mul = rewriter.create<arith::MulFOp>(loc, tanhA, tanB, fmf);
-    Value denominator = rewriter.create<complex::CreateOp>(loc, type, one, mul);
-    rewriter.replaceOpWithNewOp<complex::DivOp>(op, numerator, denominator,
-                                                fmf);
+        b.create<complex::ImOp>(loc, elementType, adaptor.getComplex());
+
+    auto cst = [&](APFloat v) {
+      return b.create<arith::ConstantOp>(elementType,
+                                         b.getFloatAttr(elementType, v));
+    };
+    Value inf = cst(APFloat::getInf(floatSemantics));
+    Value negOne = b.create<arith::ConstantOp>(
+        elementType, b.getFloatAttr(elementType, -1.0));
+    Value four = b.create<arith::ConstantOp>(elementType,
+                                             b.getFloatAttr(elementType, 4.0));
+    Value twoReal = b.create<arith::AddFOp>(real, real, fmf);
+    Value negTwoReal = b.create<arith::MulFOp>(negOne, twoReal, fmf);
+
+    Value expTwoRealMinusOne = b.create<math::ExpM1Op>(twoReal, fmf);
+    Value expNegTwoRealMinusOne = b.create<math::ExpM1Op>(negTwoReal, fmf);
+    Value realNum =
+        b.create<arith::SubFOp>(expTwoRealMinusOne, expNegTwoRealMinusOne, fmf);
+
+    Value cosImag = b.create<math::CosOp>(imag, fmf);
+    Value cosImagSq = b.create<arith::MulFOp>(cosImag, cosImag, fmf);
+    Value twoCosTwoImagPlusOne = b.create<arith::MulFOp>(cosImagSq, four, fmf);
+    Value sinImag = b.create<math::SinOp>(imag, fmf);
+
+    Value imagNum = b.create<arith::MulFOp>(
+        four, b.create<arith::MulFOp>(cosImag, sinImag, fmf), fmf);
+
+    Value expSumMinusTwo =
+        b.create<arith::AddFOp>(expTwoRealMinusOne, expNegTwoRealMinusOne, fmf);
+    Value denom =
+        b.create<arith::AddFOp>(expSumMinusTwo, twoCosTwoImagPlusOne, fmf);
+
+    Value isInf = b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ,
+                                          expSumMinusTwo, inf, fmf);
+    Value realLimit = b.create<math::CopySignOp>(negOne, real, fmf);
+
+    Value resultReal = b.create<arith::SelectOp>(
+        isInf, realLimit, b.create<arith::DivFOp>(realNum, denom, fmf));
+    Value resultImag = b.create<arith::DivFOp>(imagNum, denom, fmf);
+
+    if (!arith::bitEnumContainsAll(fmf, arith::FastMathFlags::nnan |
+                                            arith::FastMathFlags::ninf)) {
+      Value absReal = b.create<math::AbsFOp>(real, fmf);
+      Value zero = b.create<arith::ConstantOp>(
+          elementType, b.getFloatAttr(elementType, 0.0));
+      Value nan = cst(APFloat::getNaN(floatSemantics));
+
+      Value absRealIsInf =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, absReal, inf, fmf);
+      Value imagIsZero =
+          b.create<arith::CmpFOp>(arith::CmpFPredicate::OEQ, imag, zero, fmf);
+      Value absRealIsNotInf = b.create<arith::XOrIOp>(
+          absRealIsInf, b.create<arith::ConstantIntOp>(true, /*width=*/1));
+
+      Value imagNumIsNaN = b.create<arith::CmpFOp>(arith::CmpFPredicate::UNO,
+                                                   imagNum, imagNum, fmf);
+      Value resultRealIsNaN =
+          b.create<arith::AndIOp>(imagNumIsNaN, absRealIsNotInf);
+      Value resultImagIsZero = b.create<arith::OrIOp>(
+          imagIsZero, b.create<arith::AndIOp>(absRealIsInf, imagNumIsNaN));
+
+      resultReal = b.create<arith::SelectOp>(resultRealIsNaN, nan, resultReal);
+      resultImag =
+          b.create<arith::SelectOp>(resultImagIsZero, zero, resultImag);
+    }
+
+    rewriter.replaceOpWithNewOp<complex::CreateOp>(op, type, resultReal,
+                                                   resultImag);
     return success();
   }
 };
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index 8b4ea9777f797..fa1d564d6ad35 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -679,14 +679,42 @@ func.func @complex_tanh(%arg: complex<f32>) -> complex<f32> {
 }
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[TANH_A:.*]] = math.tanh %[[REAL]] : f32
-// CHECK: %[[COS_B:.*]] = math.cos %[[IMAG]] : f32
-// CHECK: %[[SIN_B:.*]] = math.sin %[[IMAG]] : f32
-// CHECK: %[[TAN_B:.*]] = arith.divf %[[SIN_B]], %[[COS_B]] : f32
-// CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] : f32
-// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
+// CHECK: %[[INF:.*]] = arith.constant 0x7F800000 : f32
+// CHECK: %[[NEG_ONE:.*]] = arith.constant -1.000000e+00 : f32
+// CHECK: %[[FOUR:.*]] = arith.constant 4.000000e+00 : f32
+// CHECK: %[[TWO_REAL:.*]] = arith.addf %[[REAL]], %[[REAL]] : f32
+// CHECK: %[[NEG_TWO_REAL:.*]] = arith.mulf %[[NEG_ONE]], %[[TWO_REAL]] : f32
+// CHECK: %[[EXPM1:.*]] = math.expm1 %[[TWO_REAL]] : f32
+// CHECK: %[[EXPM1_2:.*]] = math.expm1 %[[NEG_TWO_REAL]] : f32
+// CHECK: %[[REAL_NUM:.*]] = arith.subf %[[EXPM1]], %[[EXPM1_2]] : f32
+// CHECK: %[[COS:.*]] = math.cos %[[IMAG]] : f32
+// CHECK: %[[COS_SQ:.*]] = arith.mulf %[[COS]], %[[COS]] : f32
+// CHECK: %[[FOUR_COS_SQ:.*]] = arith.mulf %[[COS_SQ]], %[[FOUR]] : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[IMAG]] : f32
+// CHECK: %[[MUL:.*]] = arith.mulf %[[COS]], %[[SIN]] : f32
+// CHECK: %[[IMAG_NUM:.*]] = arith.mulf %[[FOUR]], %[[MUL]] : f32
+// CHECK: %[[ADD:.*]] = arith.addf %[[EXPM1]], %[[EXPM1_2]] : f32
+// CHECK: %[[DENOM:.*]] = arith.addf %[[ADD]], %[[FOUR_COS_SQ]] : f32
+// CHECK: %[[IS_INF:.*]] = arith.cmpf oeq, %[[ADD]], %[[INF]] : f32
+// CHECK: %[[LIMIT:.*]] = math.copysign %[[NEG_ONE]], %[[REAL]] : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.divf %[[REAL_NUM]], %[[DENOM]] : f32
+// CHECK: %[[RESULT_REAL2:.*]] = arith.select %[[IS_INF]], %[[LIMIT]], %[[RESULT_REAL]] : f32
+// CHECK: %[[RESULT_IMAG:.*]] = arith.divf %[[IMAG_NUM]], %[[DENOM]] : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] : f32
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[NAN:.*]] = arith.constant 0x7FC00000 : f32
+// CHECK: %[[ABS_REAL_INF:.*]] = arith.cmpf oeq, %[[ABS_REAL]], %[[INF]] : f32
+// CHECK: %[[IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] : f32
+// CHECK: %true = arith.constant true
+// CHECK: %[[ABS_REAL_NOT_INF:.*]] = arith.xori %[[ABS_REAL_INF]], %true : i1
+// CHECK: %[[IMAG_IS_NAN:.*]] = arith.cmpf uno, %[[IMAG_NUM]], %[[IMAG_NUM]] : f32
+// CHECK: %[[REAL_IS_NAN:.*]] = arith.andi %[[IMAG_IS_NAN]], %[[ABS_REAL_NOT_INF]] : i1
+// CHECK: %[[AND:.*]] = arith.andi %[[ABS_REAL_INF]], %[[IMAG_IS_NAN]] : i1
+// CHECK: %[[IMAG_IS_NAN2:.*]] = arith.ori %[[IMAG_ZERO]], %[[AND]] : i1
+// CHECK: %[[RESULT_REAL3:.*]] = arith.select %[[REAL_IS_NAN]], %[[NAN]], %[[RESULT_REAL2]] : f32
+// CHECK: %[[RESULT_IMAG2:.*]] = arith.select %[[IMAG_IS_NAN2]], %[[ZERO]], %[[RESULT_IMAG]] : f32
+// CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL3]], %[[RESULT_IMAG2]] : complex<f32>
+// CHECK: return %[[RESULT]] : complex<f32>
 
 // -----
 
@@ -2100,7 +2128,6 @@ func.func @complex_tan_with_fmf(%arg: complex<f32>) -> complex<f32> {
 // CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL_WITH_SPECIAL_CASES]], %[[RESULT_IMAG_WITH_SPECIAL_CASES]] : complex<f32>
 // CHECK: return %[[RESULT]] : complex<f32>
 
-
 // -----
 
 // CHECK-LABEL: func @complex_tanh_with_fmf
@@ -2109,13 +2136,54 @@ func.func @complex_tanh_with_fmf(%arg: complex<f32>) -> complex<f32> {
   %tanh = complex.tanh %arg fastmath<nnan,contract> : complex<f32>
   return %tanh : complex<f32>
 }
+
 // CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
 // CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
-// CHECK: %[[TANH_A:.*]] = math.tanh %[[REAL]] fastmath<nnan,contract> : f32
-// CHECK: %[[COS_B:.*]] = math.cos %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[SIN_B:.*]] = math.sin %[[IMAG]] fastmath<nnan,contract> : f32
-// CHECK: %[[TAN_B:.*]] = arith.divf %[[SIN_B]], %[[COS_B]] fastmath<nnan,contract> : f32
-// CHECK: %[[NUM:.*]] = complex.create %[[TANH_A]], %[[TAN_B]] : complex<f32>
-// CHECK: %[[ONE:.*]] = arith.constant 1.000000e+00 : f32
-// CHECK: %[[MUL:.*]] = arith.mulf %[[TANH_A]], %[[TAN_B]] fastmath<nnan,contract> : f32
-// CHECK: %[[DENOM:.*]] = complex.create %[[ONE]], %[[MUL]] : complex<f32>
+// CHECK: %[[INF:.*]] = arith.constant 0x7F800000 : f32
+// CHECK: %[[NEG_ONE:.*]] = arith.constant -1.000000e+00 : f32
+// CHECK: %[[FOUR:.*]] = arith.constant 4.000000e+00 : f32
+// CHECK: %[[TWO_REAL:.*]] = arith.addf %[[REAL]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[NEG_TWO_REAL:.*]] = arith.mulf %[[NEG_ONE]], %[[TWO_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[EXPM1:.*]] = math.expm1 %[[TWO_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[EXPM1_2:.*]] = math.expm1 %[[NEG_TWO_REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_NUM:.*]] = arith.subf %[[EXPM1]], %[[EXPM1_2]] fastmath<nnan,contract> : f32
+// CHECK: %[[COS:.*]] = math.cos %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[COS_SQ:.*]] = arith.mulf %[[COS]], %[[COS]] fastmath<nnan,contract> : f32
+// CHECK: %[[FOUR_COS_SQ:.*]] = arith.mulf %[[COS_SQ]], %[[FOUR]] fastmath<nnan,contract> : f32
+// CHECK: %[[SIN:.*]] = math.sin %[[IMAG]] fastmath<nnan,contract> : f32
+// CHECK: %[[MUL:.*]] = arith.mulf %[[COS]], %[[SIN]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_NUM:.*]] = arith.mulf %[[FOUR]], %[[MUL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ADD:.*]] = arith.addf %[[EXPM1]], %[[EXPM1_2]] fastmath<nnan,contract> : f32
+// CHECK: %[[DENOM:.*]] = arith.addf %[[ADD]], %[[FOUR_COS_SQ]] fastmath<nnan,contract> : f32
+// CHECK: %[[IS_INF:.*]] = arith.cmpf oeq, %[[ADD]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[LIMIT:.*]] = math.copysign %[[NEG_ONE]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_REAL:.*]] = arith.divf %[[REAL_NUM]], %[[DENOM]] fastmath<nnan,contract> : f32
+// CHECK: %[[RESULT_REAL2:.*]] = arith.select %[[IS_INF]], %[[LIMIT]], %[[RESULT_REAL]] : f32
+// CHECK: %[[RESULT_IMAG:.*]] = arith.divf %[[IMAG_NUM]], %[[DENOM]] fastmath<nnan,contract> : f32
+// CHECK: %[[ABS_REAL:.*]] = math.absf %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: %[[ZERO:.*]] = arith.constant 0.000000e+00 : f32
+// CHECK: %[[NAN:.*]] = arith.constant 0x7FC00000 : f32
+// CHECK: %[[ABS_REAL_INF:.*]] = arith.cmpf oeq, %[[ABS_REAL]], %[[INF]] fastmath<nnan,contract> : f32
+// CHECK: %[[IMAG_ZERO:.*]] = arith.cmpf oeq, %[[IMAG]], %[[ZERO]] fastmath<nnan,contract> : f32
+// CHECK: %true = arith.constant true
+// CHECK: %[[ABS_REAL_NOT_INF:.*]] = arith.xori %[[ABS_REAL_INF]], %true : i1
+// CHECK: %[[IMAG_IS_NAN:.*]] = arith.cmpf uno, %[[IMAG_NUM]], %[[IMAG_NUM]] fastmath<nnan,contract> : f32
+// CHECK: %[[REAL_IS_NAN:.*]] = arith.andi %[[IMAG_IS_NAN]], %[[ABS_REAL_NOT_INF]] : i1
+// CHECK: %[[AND:.*]] = arith.andi %[[ABS_REAL_INF]], %[[IMAG_IS_NAN]] : i1
+// CHECK: %[[IMAG_IS_NAN2:.*]] = arith.ori %[[IMAG_ZERO]], %[[AND]] : i1
+// CHECK: %[[RESULT_REAL3:.*]] = arith.select %[[REAL_IS_NAN]], %[[NAN]], %[[RESULT_REAL2]] : f32
+// CHECK: %[[RESULT_IMAG2:.*]] = arith.select %[[IMAG_IS_NAN2]], %[[ZERO]], %[[RESULT_IMAG]] : f32
+// CHECK: %[[RESULT:.*]] = complex.create %[[RESULT_REAL3]], %[[RESULT_IMAG2]] : complex<f32>
+// CHECK: return %[[RESULT]] : complex<f32>
+
+// -----
+
+// CHECK-LABEL: func @complex_tanh_nnan_ninf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_tanh_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
+  %tanh = complex.tanh %arg fastmath<nnan,ninf> : complex<f32>
+  return %tanh : complex<f32>
+}
+
+// CHECK-COUNT-1: arith.select
+// CHECK-NOT: arith.select

From d48ebac0389ba75defce5bd83d8539d51008b28d Mon Sep 17 00:00:00 2001
From: Paul Walker <paul.walker@arm.com>
Date: Mon, 15 Apr 2024 12:24:42 +0100
Subject: [PATCH 417/886] [LLVM][SVE][CodeGen] Fix incorrect isel for signed
 saturating instructions. (#88136)

The immediate forms of SQADD and SQSUB interpret their immediate
operand as unsigned and thus effectively only support positive
immediate operands.

The original code is only wrong for the i8 cases because they
previously accepted all values, however, the new patterns enable more
uses and this I've added tests for the larger element types as well.
---
 .../Target/AArch64/AArch64ISelDAGToDAG.cpp    |  57 ++++++++++
 .../lib/Target/AArch64/AArch64SVEInstrInfo.td |   4 +-
 llvm/lib/Target/AArch64/SVEInstrFormats.td    |  28 +++++
 .../AArch64/sve-intrinsics-int-arith-imm.ll   | 104 ++++++++++++++++++
 4 files changed, 191 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp b/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
index 51bec3604026b..80272213dd389 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp
@@ -258,6 +258,11 @@ class AArch64DAGToDAGISel : public SelectionDAGISel {
     return SelectSVEAddSubImm(N, VT, Imm, Shift);
   }
 
+  template <MVT::SimpleValueType VT, bool Negate>
+  bool SelectSVEAddSubSSatImm(SDValue N, SDValue &Imm, SDValue &Shift) {
+    return SelectSVEAddSubSSatImm(N, VT, Imm, Shift, Negate);
+  }
+
   template <MVT::SimpleValueType VT>
   bool SelectSVECpyDupImm(SDValue N, SDValue &Imm, SDValue &Shift) {
     return SelectSVECpyDupImm(N, VT, Imm, Shift);
@@ -484,6 +489,8 @@ class AArch64DAGToDAGISel : public SelectionDAGISel {
   bool SelectCMP_SWAP(SDNode *N);
 
   bool SelectSVEAddSubImm(SDValue N, MVT VT, SDValue &Imm, SDValue &Shift);
+  bool SelectSVEAddSubSSatImm(SDValue N, MVT VT, SDValue &Imm, SDValue &Shift,
+                              bool Negate);
   bool SelectSVECpyDupImm(SDValue N, MVT VT, SDValue &Imm, SDValue &Shift);
   bool SelectSVELogicalImm(SDValue N, MVT VT, SDValue &Imm, bool Invert);
 
@@ -4014,6 +4021,56 @@ bool AArch64DAGToDAGISel::SelectSVEAddSubImm(SDValue N, MVT VT, SDValue &Imm,
   return false;
 }
 
+bool AArch64DAGToDAGISel::SelectSVEAddSubSSatImm(SDValue N, MVT VT,
+                                                 SDValue &Imm, SDValue &Shift,
+                                                 bool Negate) {
+  if (!isa<ConstantSDNode>(N))
+    return false;
+
+  SDLoc DL(N);
+  int64_t Val = cast<ConstantSDNode>(N)
+                    ->getAPIntValue()
+                    .trunc(VT.getFixedSizeInBits())
+                    .getSExtValue();
+
+  if (Negate)
+    Val = -Val;
+
+  // Signed saturating instructions treat their immediate operand as unsigned,
+  // whereas the related intrinsics define their operands to be signed. This
+  // means we can only use the immediate form when the operand is non-negative.
+  if (Val < 0)
+    return false;
+
+  switch (VT.SimpleTy) {
+  case MVT::i8:
+    // All positive immediates are supported.
+    Shift = CurDAG->getTargetConstant(0, DL, MVT::i32);
+    Imm = CurDAG->getTargetConstant(Val, DL, MVT::i32);
+    return true;
+  case MVT::i16:
+  case MVT::i32:
+  case MVT::i64:
+    // Support 8bit positive immediates.
+    if (Val <= 255) {
+      Shift = CurDAG->getTargetConstant(0, DL, MVT::i32);
+      Imm = CurDAG->getTargetConstant(Val, DL, MVT::i32);
+      return true;
+    }
+    // Support 16bit positive immediates that are a multiple of 256.
+    if (Val <= 65280 && Val % 256 == 0) {
+      Shift = CurDAG->getTargetConstant(8, DL, MVT::i32);
+      Imm = CurDAG->getTargetConstant(Val >> 8, DL, MVT::i32);
+      return true;
+    }
+    break;
+  default:
+    break;
+  }
+
+  return false;
+}
+
 bool AArch64DAGToDAGISel::SelectSVECpyDupImm(SDValue N, MVT VT, SDValue &Imm,
                                              SDValue &Shift) {
   if (!isa<ConstantSDNode>(N))
diff --git a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
index 9c747198c12d8..6972acd985cb9 100644
--- a/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td
@@ -554,9 +554,9 @@ let Predicates = [HasSVEorSME] in {
   defm ADD_ZI   : sve_int_arith_imm0<0b000, "add", add>;
   defm SUB_ZI   : sve_int_arith_imm0<0b001, "sub", sub>;
   defm SUBR_ZI  : sve_int_arith_imm0<0b011, "subr", AArch64subr>;
-  defm SQADD_ZI : sve_int_arith_imm0<0b100, "sqadd", saddsat>;
+  defm SQADD_ZI : sve_int_arith_imm0_ssat<0b100, "sqadd", saddsat, ssubsat>;
   defm UQADD_ZI : sve_int_arith_imm0<0b101, "uqadd", uaddsat>;
-  defm SQSUB_ZI : sve_int_arith_imm0<0b110, "sqsub", ssubsat>;
+  defm SQSUB_ZI : sve_int_arith_imm0_ssat<0b110, "sqsub", ssubsat, saddsat>;
   defm UQSUB_ZI : sve_int_arith_imm0<0b111, "uqsub", usubsat>;
 
   defm MAD_ZPmZZ : sve_int_mladdsub_vvv_pred<0b0, "mad", AArch64mad_m1, "MLA_ZPmZZ", /*isReverseInstr*/ 1>;
diff --git a/llvm/lib/Target/AArch64/SVEInstrFormats.td b/llvm/lib/Target/AArch64/SVEInstrFormats.td
index fb0c6188edb34..3317cf84aa6fa 100644
--- a/llvm/lib/Target/AArch64/SVEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SVEInstrFormats.td
@@ -249,6 +249,16 @@ def SVEAddSubImm16Pat : ComplexPattern<i32, 2, "SelectSVEAddSubImm<MVT::i16>", [
 def SVEAddSubImm32Pat : ComplexPattern<i32, 2, "SelectSVEAddSubImm<MVT::i32>", []>;
 def SVEAddSubImm64Pat : ComplexPattern<i64, 2, "SelectSVEAddSubImm<MVT::i64>", []>;
 
+def SVEAddSubSSatNegImm8Pat  : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i8, true>", []>;
+def SVEAddSubSSatNegImm16Pat : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i16, true>", []>;
+def SVEAddSubSSatNegImm32Pat : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i32, true>", []>;
+def SVEAddSubSSatNegImm64Pat : ComplexPattern<i64, 2, "SelectSVEAddSubSSatImm<MVT::i64, true>", []>;
+
+def SVEAddSubSSatPosImm8Pat  : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i8, false>", []>;
+def SVEAddSubSSatPosImm16Pat : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i16, false>", []>;
+def SVEAddSubSSatPosImm32Pat : ComplexPattern<i32, 2, "SelectSVEAddSubSSatImm<MVT::i32, false>", []>;
+def SVEAddSubSSatPosImm64Pat : ComplexPattern<i64, 2, "SelectSVEAddSubSSatImm<MVT::i64, false>", []>;
+
 def SVECpyDupImm8Pat  : ComplexPattern<i32, 2, "SelectSVECpyDupImm<MVT::i8>", []>;
 def SVECpyDupImm16Pat : ComplexPattern<i32, 2, "SelectSVECpyDupImm<MVT::i16>", []>;
 def SVECpyDupImm32Pat : ComplexPattern<i32, 2, "SelectSVECpyDupImm<MVT::i32>", []>;
@@ -4775,6 +4785,24 @@ multiclass sve_int_arith_imm0<bits<3> opc, string asm, SDPatternOperator op> {
   def : SVE_1_Op_Imm_OptLsl_Pat<nxv2i64, op, ZPR64, i64, SVEAddSubImm64Pat, !cast<Instruction>(NAME # _D)>;
 }
 
+multiclass sve_int_arith_imm0_ssat<bits<3> opc, string asm, SDPatternOperator op,
+                                   SDPatternOperator inv_op> {
+  def _B : sve_int_arith_imm0<0b00, opc, asm, ZPR8,  addsub_imm8_opt_lsl_i8>;
+  def _H : sve_int_arith_imm0<0b01, opc, asm, ZPR16, addsub_imm8_opt_lsl_i16>;
+  def _S : sve_int_arith_imm0<0b10, opc, asm, ZPR32, addsub_imm8_opt_lsl_i32>;
+  def _D : sve_int_arith_imm0<0b11, opc, asm, ZPR64, addsub_imm8_opt_lsl_i64>;
+
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv16i8, op, ZPR8,  i32, SVEAddSubSSatPosImm8Pat,  !cast<Instruction>(NAME # _B)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv8i16, op, ZPR16, i32, SVEAddSubSSatPosImm16Pat, !cast<Instruction>(NAME # _H)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv4i32, op, ZPR32, i32, SVEAddSubSSatPosImm32Pat, !cast<Instruction>(NAME # _S)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv2i64, op, ZPR64, i64, SVEAddSubSSatPosImm64Pat, !cast<Instruction>(NAME # _D)>;
+
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv16i8, inv_op, ZPR8,  i32, SVEAddSubSSatNegImm8Pat,  !cast<Instruction>(NAME # _B)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv8i16, inv_op, ZPR16, i32, SVEAddSubSSatNegImm16Pat, !cast<Instruction>(NAME # _H)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv4i32, inv_op, ZPR32, i32, SVEAddSubSSatNegImm32Pat, !cast<Instruction>(NAME # _S)>;
+  def : SVE_1_Op_Imm_OptLsl_Pat<nxv2i64, inv_op, ZPR64, i64, SVEAddSubSSatNegImm64Pat, !cast<Instruction>(NAME # _D)>;
+}
+
 class sve_int_arith_imm<bits<2> sz8_64, bits<6> opc, string asm,
                         ZPRRegOp zprty, Operand immtype>
 : I<(outs zprty:$Zdn), (ins zprty:$_Zdn, immtype:$imm),
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
index c70006d988c19..3e453a6b78179 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
@@ -1059,6 +1059,19 @@ define <vscale x 16 x i8> @sqadd_b_lowimm(<vscale x 16 x i8> %a) {
   ret <vscale x 16 x i8> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 16 x i8> @sqadd_b_negimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: sqadd_b_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.b, z0.b, #128 // =0x80
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 16 x i8> undef, i8 -128, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqadd.x.nxv16i8(<vscale x 16 x i8> %a,
+                                                                   <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
 define <vscale x 8 x i16> @sqadd_h_lowimm(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: sqadd_h_lowimm:
 ; CHECK:       // %bb.0:
@@ -1083,6 +1096,19 @@ define <vscale x 8 x i16> @sqadd_h_highimm(<vscale x 8 x i16> %a) {
   ret <vscale x 8 x i16> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 8 x i16> @sqadd_h_negimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqadd_h_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.h, z0.h, #1 // =0x1
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 -1, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqadd.x.nxv8i16(<vscale x 8 x i16> %a,
+                                                                   <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @sqadd_s_lowimm(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: sqadd_s_lowimm:
 ; CHECK:       // %bb.0:
@@ -1107,6 +1133,19 @@ define <vscale x 4 x i32> @sqadd_s_highimm(<vscale x 4 x i32> %a) {
   ret <vscale x 4 x i32> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 4 x i32> @sqadd_s_negimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqadd_s_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.s, z0.s, #65280 // =0xff00
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 -65280, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqadd.x.nxv4i32(<vscale x 4 x i32> %a,
+                                                                   <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
 define <vscale x 2 x i64> @sqadd_d_lowimm(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: sqadd_d_lowimm:
 ; CHECK:       // %bb.0:
@@ -1131,6 +1170,19 @@ define <vscale x 2 x i64> @sqadd_d_highimm(<vscale x 2 x i64> %a) {
   ret <vscale x 2 x i64> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 2 x i64> @sqadd_d_negimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqadd_d_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqsub z0.d, z0.d, #3840 // =0xf00
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 -3840, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqadd.x.nxv2i64(<vscale x 2 x i64> %a,
+                                                                   <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
 ; SQSUB
 
 define <vscale x 16 x i8> @sqsub_b_lowimm(<vscale x 16 x i8> %a) {
@@ -1145,6 +1197,19 @@ define <vscale x 16 x i8> @sqsub_b_lowimm(<vscale x 16 x i8> %a) {
   ret <vscale x 16 x i8> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 16 x i8> @sqsub_b_negimm(<vscale x 16 x i8> %a) {
+; CHECK-LABEL: sqsub_b_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.b, z0.b, #1 // =0x1
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 16 x i8> undef, i8 -1, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %elt, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %out = call <vscale x 16 x i8> @llvm.aarch64.sve.sqsub.x.nxv16i8(<vscale x 16 x i8> %a,
+                                                                   <vscale x 16 x i8> %splat)
+  ret <vscale x 16 x i8> %out
+}
+
 define <vscale x 8 x i16> @sqsub_h_lowimm(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: sqsub_h_lowimm:
 ; CHECK:       // %bb.0:
@@ -1169,6 +1234,19 @@ define <vscale x 8 x i16> @sqsub_h_highimm(<vscale x 8 x i16> %a) {
   ret <vscale x 8 x i16> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 8 x i16> @sqsub_h_negimm(<vscale x 8 x i16> %a) {
+; CHECK-LABEL: sqsub_h_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.h, z0.h, #128 // =0x80
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 8 x i16> undef, i16 -128, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %elt, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %out = call <vscale x 8 x i16> @llvm.aarch64.sve.sqsub.x.nxv8i16(<vscale x 8 x i16> %a,
+                                                                   <vscale x 8 x i16> %splat)
+  ret <vscale x 8 x i16> %out
+}
+
 define <vscale x 4 x i32> @sqsub_s_lowimm(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: sqsub_s_lowimm:
 ; CHECK:       // %bb.0:
@@ -1193,6 +1271,19 @@ define <vscale x 4 x i32> @sqsub_s_highimm(<vscale x 4 x i32> %a) {
   ret <vscale x 4 x i32> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 4 x i32> @sqsub_s_negimm(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: sqsub_s_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.s, z0.s, #32768 // =0x8000
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 4 x i32> undef, i32 -32768, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %elt, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %out = call <vscale x 4 x i32> @llvm.aarch64.sve.sqsub.x.nxv4i32(<vscale x 4 x i32> %a,
+                                                                   <vscale x 4 x i32> %splat)
+  ret <vscale x 4 x i32> %out
+}
+
 define <vscale x 2 x i64> @sqsub_d_lowimm(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: sqsub_d_lowimm:
 ; CHECK:       // %bb.0:
@@ -1217,6 +1308,19 @@ define <vscale x 2 x i64> @sqsub_d_highimm(<vscale x 2 x i64> %a) {
   ret <vscale x 2 x i64> %out
 }
 
+; Immediate instruction form only supports positive values.
+define <vscale x 2 x i64> @sqsub_d_negimm(<vscale x 2 x i64> %a) {
+; CHECK-LABEL: sqsub_d_negimm:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    sqadd z0.d, z0.d, #57344 // =0xe000
+; CHECK-NEXT:    ret
+  %elt = insertelement <vscale x 2 x i64> undef, i64 -57344, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %elt, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %out = call <vscale x 2 x i64> @llvm.aarch64.sve.sqsub.x.nxv2i64(<vscale x 2 x i64> %a,
+                                                                   <vscale x 2 x i64> %splat)
+  ret <vscale x 2 x i64> %out
+}
+
 ; UQADD
 
 define <vscale x 16 x i8> @uqadd_b_lowimm(<vscale x 16 x i8> %a) {

From 6254b6dd892728306ddfe657f10eb43a3799864d Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Mon, 15 Apr 2024 12:27:45 +0100
Subject: [PATCH 418/886] [VPlan] Version VPValue names in VPSlotTracker. 
 (#81411)

This patch restructures the way names for printing VPValues are handled.
It moves the logic to generate names for printing to VPSlotTracker.

VPSlotTracker will now version names of the same underlying value if it
is used by multiple VPValues, by adding a .V suffix to the name.

This fixes cases where at the moment the same name is printed for
different VPValues.

PR: https://github.com/llvm/llvm-project/pull/81411
---
 llvm/lib/Transforms/Vectorize/VPlan.cpp       | 93 ++++++++++++++-----
 llvm/lib/Transforms/Vectorize/VPlanValue.h    | 34 ++++---
 .../vplan-sink-scalars-and-merge.ll           |  8 +-
 3 files changed, 94 insertions(+), 41 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index 3e1069d82ddad..999236ae84898 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -1300,18 +1300,7 @@ void VPValue::replaceUsesWithIf(
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
 void VPValue::printAsOperand(raw_ostream &OS, VPSlotTracker &Tracker) const {
-  if (const Value *UV = getUnderlyingValue()) {
-    OS << "ir<";
-    UV->printAsOperand(OS, false);
-    OS << ">";
-    return;
-  }
-
-  unsigned Slot = Tracker.getSlot(this);
-  if (Slot == unsigned(-1))
-    OS << "<badref>";
-  else
-    OS << "vp<%" << Tracker.getSlot(this) << ">";
+  OS << Tracker.getOrCreateName(this);
 }
 
 void VPUser::printOperands(raw_ostream &O, VPSlotTracker &SlotTracker) const {
@@ -1373,32 +1362,88 @@ VPInterleavedAccessInfo::VPInterleavedAccessInfo(VPlan &Plan,
   visitRegion(Plan.getVectorLoopRegion(), Old2New, IAI);
 }
 
-void VPSlotTracker::assignSlot(const VPValue *V) {
-  if (V->getUnderlyingValue())
+void VPSlotTracker::assignName(const VPValue *V) {
+  assert(!VPValue2Name.contains(V) && "VPValue already has a name!");
+  auto *UV = V->getUnderlyingValue();
+  if (!UV) {
+    VPValue2Name[V] = (Twine("vp<%") + Twine(NextSlot) + ">").str();
+    NextSlot++;
     return;
-  assert(!Slots.contains(V) && "VPValue already has a slot!");
-  Slots[V] = NextSlot++;
+  }
+
+  // Use the name of the underlying Value, wrapped in "ir<>", and versioned by
+  // appending ".Number" to the name if there are multiple uses.
+  std::string Name;
+  raw_string_ostream S(Name);
+  UV->printAsOperand(S, false);
+  assert(!Name.empty() && "Name cannot be empty.");
+  std::string BaseName = (Twine("ir<") + Name + Twine(">")).str();
+
+  // First assign the base name for V.
+  const auto &[A, _] = VPValue2Name.insert({V, BaseName});
+  // Integer or FP constants with different types will result in he same string
+  // due to stripping types.
+  if (V->isLiveIn() && isa<ConstantInt, ConstantFP>(UV))
+    return;
+
+  // If it is already used by C > 0 other VPValues, increase the version counter
+  // C and use it for V.
+  const auto &[C, UseInserted] = BaseName2Version.insert({BaseName, 0});
+  if (!UseInserted) {
+    C->second++;
+    A->second = (BaseName + Twine(".") + Twine(C->second)).str();
+  }
 }
 
-void VPSlotTracker::assignSlots(const VPlan &Plan) {
+void VPSlotTracker::assignNames(const VPlan &Plan) {
   if (Plan.VFxUF.getNumUsers() > 0)
-    assignSlot(&Plan.VFxUF);
-  assignSlot(&Plan.VectorTripCount);
+    assignName(&Plan.VFxUF);
+  assignName(&Plan.VectorTripCount);
   if (Plan.BackedgeTakenCount)
-    assignSlot(Plan.BackedgeTakenCount);
-  assignSlots(Plan.getPreheader());
+    assignName(Plan.BackedgeTakenCount);
+  for (VPValue *LI : Plan.VPLiveInsToFree)
+    assignName(LI);
+  assignNames(Plan.getPreheader());
 
   ReversePostOrderTraversal<VPBlockDeepTraversalWrapper<const VPBlockBase *>>
       RPOT(VPBlockDeepTraversalWrapper<const VPBlockBase *>(Plan.getEntry()));
   for (const VPBasicBlock *VPBB :
        VPBlockUtils::blocksOnly<const VPBasicBlock>(RPOT))
-    assignSlots(VPBB);
+    assignNames(VPBB);
 }
 
-void VPSlotTracker::assignSlots(const VPBasicBlock *VPBB) {
+void VPSlotTracker::assignNames(const VPBasicBlock *VPBB) {
   for (const VPRecipeBase &Recipe : *VPBB)
     for (VPValue *Def : Recipe.definedValues())
-      assignSlot(Def);
+      assignName(Def);
+}
+
+std::string VPSlotTracker::getOrCreateName(const VPValue *V) const {
+  std::string Name = VPValue2Name.lookup(V);
+  if (!Name.empty())
+    return Name;
+
+  // If no name was assigned, no VPlan was provided when creating the slot
+  // tracker or it is not reachable from the provided VPlan. This can happen,
+  // e.g. when trying to print a recipe that has not been inserted into a VPlan
+  // in a debugger.
+  // TODO: Update VPSlotTracker constructor to assign names to recipes &
+  // VPValues not associated with a VPlan, instead of constructing names ad-hoc
+  // here.
+  const VPRecipeBase *DefR = V->getDefiningRecipe();
+  (void)DefR;
+  assert((!DefR || !DefR->getParent() || !DefR->getParent()->getPlan()) &&
+         "VPValue defined by a recipe in a VPlan?");
+
+  // Use the underlying value's name, if there is one.
+  if (auto *UV = V->getUnderlyingValue()) {
+    std::string Name;
+    raw_string_ostream S(Name);
+    UV->printAsOperand(S, false);
+    return (Twine("ir<") + Name + ">").str();
+  }
+
+  return "<badref>";
 }
 
 bool vputils::onlyFirstLaneUsed(const VPValue *Def) {
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 8b221d30e5254..da3a768552fc5 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -23,6 +23,7 @@
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/TinyPtrVector.h"
 #include "llvm/ADT/iterator_range.h"
 
@@ -443,29 +444,36 @@ class VPDef {
 class VPlan;
 class VPBasicBlock;
 
-/// This class can be used to assign consecutive numbers to all VPValues in a
-/// VPlan and allows querying the numbering for printing, similar to the
+/// This class can be used to assign names to VPValues. For VPValues without
+/// underlying value, assign consecutive numbers and use those as names (wrapped
+/// in vp<>). Otherwise, use the name from the underlying value (wrapped in
+/// ir<>), appending a .V version number if there are multiple uses of the same
+/// name. Allows querying names for VPValues for printing, similar to the
 /// ModuleSlotTracker for IR values.
 class VPSlotTracker {
-  DenseMap<const VPValue *, unsigned> Slots;
+  /// Keep track of versioned names assigned to VPValues with underlying IR
+  /// values.
+  DenseMap<const VPValue *, std::string> VPValue2Name;
+  /// Keep track of the next number to use to version the base name.
+  StringMap<unsigned> BaseName2Version;
+
+  /// Number to assign to the next VPValue without underlying value.
   unsigned NextSlot = 0;
 
-  void assignSlot(const VPValue *V);
-  void assignSlots(const VPlan &Plan);
-  void assignSlots(const VPBasicBlock *VPBB);
+  void assignName(const VPValue *V);
+  void assignNames(const VPlan &Plan);
+  void assignNames(const VPBasicBlock *VPBB);
 
 public:
   VPSlotTracker(const VPlan *Plan = nullptr) {
     if (Plan)
-      assignSlots(*Plan);
+      assignNames(*Plan);
   }
 
-  unsigned getSlot(const VPValue *V) const {
-    auto I = Slots.find(V);
-    if (I == Slots.end())
-      return -1;
-    return I->second;
-  }
+  /// Returns the name assigned to \p V, if there is one, otherwise try to
+  /// construct one from the underlying value, if there's one; else return
+  /// <badref>.
+  std::string getOrCreateName(const VPValue *V) const;
 };
 
 } // namespace llvm
diff --git a/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll b/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
index 0cacb02dc4891..108b78a70fa1a 100644
--- a/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
+++ b/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
@@ -986,8 +986,8 @@ define void @sinking_requires_duplication(ptr %addr) {
 ; CHECK-NEXT:   Successor(s): pred.store.if, pred.store.continue
 ; CHECK-EMPTY:
 ; CHECK-NEXT:   pred.store.if:
-; CHECK-NEXT:     REPLICATE ir<%gep> = getelementptr ir<%addr>, vp<[[STEPS]]>
-; CHECK-NEXT:     REPLICATE store ir<1.000000e+01>, ir<%gep>
+; CHECK-NEXT:     REPLICATE ir<%gep>.1 = getelementptr ir<%addr>, vp<[[STEPS]]>
+; CHECK-NEXT:     REPLICATE store ir<1.000000e+01>, ir<%gep>.1
 ; CHECK-NEXT:   Successor(s): pred.store.continue
 ; CHECK-EMPTY:
 ; CHECK-NEXT:   pred.store.continue:
@@ -1129,8 +1129,8 @@ define void @ptr_induction_remove_dead_recipe(ptr %start, ptr %end) {
 ; CHECK-NEXT:     Successor(s): pred.store.if, pred.store.continue
 ; CHECK-EMPTY:
 ; CHECK-NEXT:     pred.store.if:
-; CHECK-NEXT:       REPLICATE ir<%ptr.iv.next> = getelementptr inbounds vp<[[PTR_IV]]>, ir<-1>
-; CHECK-NEXT:       REPLICATE store ir<95>, ir<%ptr.iv.next>
+; CHECK-NEXT:       REPLICATE ir<%ptr.iv.next>.1 = getelementptr inbounds vp<[[PTR_IV]]>, ir<-1>
+; CHECK-NEXT:       REPLICATE store ir<95>, ir<%ptr.iv.next>.1
 ; CHECK-NEXT:     Successor(s): pred.store.continue
 ; CHECK-EMPTY:
 ; CHECK-NEXT:     pred.store.continue:

From 8095b9ce6bf5831a14c72028920708f38d13d0c3 Mon Sep 17 00:00:00 2001
From: Congcong Cai <congcongcai0907@163.com>
Date: Mon, 15 Apr 2024 19:38:50 +0800
Subject: [PATCH 419/886] [clang analysis] ExprMutationAnalyzer avoid infinite
 recursion for recursive forwarding reference (#87954)

---
 clang-tools-extra/docs/ReleaseNotes.rst       |  4 +++
 .../misc/const-correctness-templates.cpp      | 15 ++++++++++
 .../Analysis/Analyses/ExprMutationAnalyzer.h  | 28 ++++++++++++-----
 clang/lib/Analysis/ExprMutationAnalyzer.cpp   | 22 +++++++++-----
 .../Analysis/ExprMutationAnalyzerTest.cpp     | 30 +++++++++++++++++++
 5 files changed, 84 insertions(+), 15 deletions(-)

diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index 1405fb0df1f8d..de0a7b79bb2cf 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -221,6 +221,10 @@ Changes in existing checks
   <clang-tidy/checks/llvm/header-guard>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
 
+- Improved :doc:`misc-const-correctness
+  <clang-tidy/checks/misc/const-correctness>` check by avoiding infinite recursion
+  for recursive forwarding reference.
+
 - Improved :doc:`misc-definitions-in-headers
   <clang-tidy/checks/misc/definitions-in-headers>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
diff --git a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
index 9da468128743e..248374a71dd40 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
@@ -58,3 +58,18 @@ void concatenate3(Args... args)
     (..., (stream << args));
 }
 } // namespace gh70323
+
+namespace gh60895 {
+
+template <class T> void f1(T &&a);
+template <class T> void f2(T &&a);
+template <class T> void f1(T &&a) { f2<T>(a); }
+template <class T> void f2(T &&a) { f1<T>(a); }
+void f() {
+  int x = 0;
+  // CHECK-MESSAGES:[[@LINE-1]]:3: warning: variable 'x' of type 'int' can be declared 'const'
+  // CHECK-FIXES: int const x = 0;
+  f1(x);
+}
+
+} // namespace gh60895
diff --git a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
index 1ceef944fbc34..c4e5d0badb8e5 100644
--- a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
+++ b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
@@ -8,11 +8,10 @@
 #ifndef LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 #define LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 
-#include <type_traits>
-
 #include "clang/AST/AST.h"
 #include "clang/ASTMatchers/ASTMatchers.h"
 #include "llvm/ADT/DenseMap.h"
+#include <variant>
 
 namespace clang {
 
@@ -22,8 +21,15 @@ class FunctionParmMutationAnalyzer;
 /// a given statement.
 class ExprMutationAnalyzer {
 public:
+  friend class FunctionParmMutationAnalyzer;
+  struct Cache {
+    llvm::SmallDenseMap<const FunctionDecl *,
+                        std::unique_ptr<FunctionParmMutationAnalyzer>>
+        FuncParmAnalyzer;
+  };
+
   ExprMutationAnalyzer(const Stmt &Stm, ASTContext &Context)
-      : Stm(Stm), Context(Context) {}
+      : ExprMutationAnalyzer(Stm, Context, std::make_shared<Cache>()) {}
 
   bool isMutated(const Expr *Exp) { return findMutation(Exp) != nullptr; }
   bool isMutated(const Decl *Dec) { return findMutation(Dec) != nullptr; }
@@ -45,6 +51,11 @@ class ExprMutationAnalyzer {
   using MutationFinder = const Stmt *(ExprMutationAnalyzer::*)(const Expr *);
   using ResultMap = llvm::DenseMap<const Expr *, const Stmt *>;
 
+  ExprMutationAnalyzer(const Stmt &Stm, ASTContext &Context,
+                       std::shared_ptr<Cache> CrossAnalysisCache)
+      : Stm(Stm), Context(Context),
+        CrossAnalysisCache(std::move(CrossAnalysisCache)) {}
+
   const Stmt *findMutationMemoized(const Expr *Exp,
                                    llvm::ArrayRef<MutationFinder> Finders,
                                    ResultMap &MemoizedResults);
@@ -69,9 +80,7 @@ class ExprMutationAnalyzer {
 
   const Stmt &Stm;
   ASTContext &Context;
-  llvm::DenseMap<const FunctionDecl *,
-                 std::unique_ptr<FunctionParmMutationAnalyzer>>
-      FuncParmAnalyzer;
+  std::shared_ptr<Cache> CrossAnalysisCache;
   ResultMap Results;
   ResultMap PointeeResults;
 };
@@ -80,7 +89,12 @@ class ExprMutationAnalyzer {
 // params.
 class FunctionParmMutationAnalyzer {
 public:
-  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context);
+  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context)
+      : FunctionParmMutationAnalyzer(
+            Func, Context, std::make_shared<ExprMutationAnalyzer::Cache>()) {}
+  FunctionParmMutationAnalyzer(
+      const FunctionDecl &Func, ASTContext &Context,
+      std::shared_ptr<ExprMutationAnalyzer::Cache> CrossAnalysisCache);
 
   bool isMutated(const ParmVarDecl *Parm) {
     return findMutation(Parm) != nullptr;
diff --git a/clang/lib/Analysis/ExprMutationAnalyzer.cpp b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
index bb042760d297a..503ca4bea99e5 100644
--- a/clang/lib/Analysis/ExprMutationAnalyzer.cpp
+++ b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
@@ -638,9 +638,10 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
       if (!RefType->getPointeeType().getQualifiers() &&
           RefType->getPointeeType()->getAs<TemplateTypeParmType>()) {
         std::unique_ptr<FunctionParmMutationAnalyzer> &Analyzer =
-            FuncParmAnalyzer[Func];
+            CrossAnalysisCache->FuncParmAnalyzer[Func];
         if (!Analyzer)
-          Analyzer.reset(new FunctionParmMutationAnalyzer(*Func, Context));
+          Analyzer.reset(new FunctionParmMutationAnalyzer(*Func, Context,
+                                                          CrossAnalysisCache));
         if (Analyzer->findMutation(Parm))
           return Exp;
         continue;
@@ -653,13 +654,15 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
 }
 
 FunctionParmMutationAnalyzer::FunctionParmMutationAnalyzer(
-    const FunctionDecl &Func, ASTContext &Context)
-    : BodyAnalyzer(*Func.getBody(), Context) {
+    const FunctionDecl &Func, ASTContext &Context,
+    std::shared_ptr<ExprMutationAnalyzer::Cache> CrossAnalysisCache)
+    : BodyAnalyzer(*Func.getBody(), Context, CrossAnalysisCache) {
   if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(&Func)) {
     // CXXCtorInitializer might also mutate Param but they're not part of
     // function body, check them eagerly here since they're typically trivial.
     for (const CXXCtorInitializer *Init : Ctor->inits()) {
-      ExprMutationAnalyzer InitAnalyzer(*Init->getInit(), Context);
+      ExprMutationAnalyzer InitAnalyzer(*Init->getInit(), Context,
+                                        CrossAnalysisCache);
       for (const ParmVarDecl *Parm : Ctor->parameters()) {
         if (Results.contains(Parm))
           continue;
@@ -675,11 +678,14 @@ FunctionParmMutationAnalyzer::findMutation(const ParmVarDecl *Parm) {
   const auto Memoized = Results.find(Parm);
   if (Memoized != Results.end())
     return Memoized->second;
-
+  // To handle call A -> call B -> call A. Assume parameters of A is not mutated
+  // before analyzing parameters of A. Then when analyzing the second "call A",
+  // FunctionParmMutationAnalyzer can use this memoized value to avoid infinite
+  // recursion.
+  Results[Parm] = nullptr;
   if (const Stmt *S = BodyAnalyzer.findMutation(Parm))
     return Results[Parm] = S;
-
-  return Results[Parm] = nullptr;
+  return Results[Parm];
 }
 
 } // namespace clang
diff --git a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
index f58ce4aebcbfc..9c1dc1a76db63 100644
--- a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
+++ b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
@@ -977,6 +977,36 @@ TEST(ExprMutationAnalyzerTest, FollowFuncArgModified) {
                          "void f() { int x; g(x); }");
   Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
   EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("g(x)"));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_FALSE(isMutated(Results, AST.get()));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); a++; }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); a++; }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
 }
 
 TEST(ExprMutationAnalyzerTest, FollowFuncArgNotModified) {

From 673da8c83413c4d1e7e76e1b52a2924e837e7221 Mon Sep 17 00:00:00 2001
From: Paul Walker <paul.walker@arm.com>
Date: Mon, 15 Apr 2024 12:51:06 +0100
Subject: [PATCH 420/886] [LLVM][CodeGen][AArch64] Remove bogus lowering of
 sve_while{gt/ge/hi/hs} intrinsics. (#88126)

When fed constant operands we try to lower WHILE intrinsics to PTRUE
using a fixed length pattern. This is not valid for the decrementing
variants of WHILE because they construct their result predicate vector
by traversing from high->low lanes whereas the incrementing variants and
PTRUE traverse from low->high lanes.

Whilst we can still utilise PTRUE by reversing its result I figure
replacing a single WHILE with multiple SVE instructions is likely
counterproductive.
---
 .../Target/AArch64/AArch64ISelLowering.cpp    |  39 +---
 .../CodeGen/AArch64/sve2-intrinsics-while.ll  | 212 ++----------------
 2 files changed, 28 insertions(+), 223 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 80181a77c9d23..27dc79f3f8d0c 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -5035,8 +5035,8 @@ static inline SDValue getPTrue(SelectionDAG &DAG, SDLoc DL, EVT VT,
                      DAG.getTargetConstant(Pattern, DL, MVT::i32));
 }
 
-static SDValue optimizeWhile(SDValue Op, SelectionDAG &DAG, bool IsSigned,
-                             bool IsLess, bool IsEqual) {
+static SDValue optimizeIncrementingWhile(SDValue Op, SelectionDAG &DAG,
+                                         bool IsSigned, bool IsEqual) {
   if (!isa<ConstantSDNode>(Op.getOperand(1)) ||
       !isa<ConstantSDNode>(Op.getOperand(2)))
     return SDValue();
@@ -5044,12 +5044,9 @@ static SDValue optimizeWhile(SDValue Op, SelectionDAG &DAG, bool IsSigned,
   SDLoc dl(Op);
   APInt X = Op.getConstantOperandAPInt(1);
   APInt Y = Op.getConstantOperandAPInt(2);
-  APInt NumActiveElems;
   bool Overflow;
-  if (IsLess)
-    NumActiveElems = IsSigned ? Y.ssub_ov(X, Overflow) : Y.usub_ov(X, Overflow);
-  else
-    NumActiveElems = IsSigned ? X.ssub_ov(Y, Overflow) : X.usub_ov(Y, Overflow);
+  APInt NumActiveElems =
+      IsSigned ? Y.ssub_ov(X, Overflow) : Y.usub_ov(X, Overflow);
 
   if (Overflow)
     return SDValue();
@@ -5396,29 +5393,17 @@ SDValue AArch64TargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
     return SDValue();
   }
   case Intrinsic::aarch64_sve_whilelo:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/false, /*IsLess=*/true,
-                         /*IsEqual=*/false);
+    return optimizeIncrementingWhile(Op, DAG, /*IsSigned=*/false,
+                                     /*IsEqual=*/false);
   case Intrinsic::aarch64_sve_whilelt:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/true, /*IsLess=*/true,
-                         /*IsEqual=*/false);
+    return optimizeIncrementingWhile(Op, DAG, /*IsSigned=*/true,
+                                     /*IsEqual=*/false);
   case Intrinsic::aarch64_sve_whilels:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/false, /*IsLess=*/true,
-                         /*IsEqual=*/true);
+    return optimizeIncrementingWhile(Op, DAG, /*IsSigned=*/false,
+                                     /*IsEqual=*/true);
   case Intrinsic::aarch64_sve_whilele:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/true, /*IsLess=*/true,
-                         /*IsEqual=*/true);
-  case Intrinsic::aarch64_sve_whilege:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/true, /*IsLess=*/false,
-                         /*IsEqual=*/true);
-  case Intrinsic::aarch64_sve_whilegt:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/true, /*IsLess=*/false,
-                         /*IsEqual=*/false);
-  case Intrinsic::aarch64_sve_whilehs:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/false, /*IsLess=*/false,
-                         /*IsEqual=*/true);
-  case Intrinsic::aarch64_sve_whilehi:
-    return optimizeWhile(Op, DAG, /*IsSigned=*/false, /*IsLess=*/false,
-                         /*IsEqual=*/false);
+    return optimizeIncrementingWhile(Op, DAG, /*IsSigned=*/true,
+                                     /*IsEqual=*/true);
   case Intrinsic::aarch64_sve_sunpkhi:
     return DAG.getNode(AArch64ISD::SUNPKHI, dl, Op.getValueType(),
                        Op.getOperand(1));
diff --git a/llvm/test/CodeGen/AArch64/sve2-intrinsics-while.ll b/llvm/test/CodeGen/AArch64/sve2-intrinsics-while.ll
index 9b82d79cfa962..2052ebf4c8bb0 100644
--- a/llvm/test/CodeGen/AArch64/sve2-intrinsics-while.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-intrinsics-while.ll
@@ -78,68 +78,16 @@ define <vscale x 2 x i1> @whilege_d_xx(i64 %a, i64 %b) {
   ret <vscale x 2 x i1> %out
 }
 
-define <vscale x 2 x i1> @whilege_d_ii_dont_fold_to_ptrue_larger_than_minvec() {
-; CHECK-LABEL: whilege_d_ii_dont_fold_to_ptrue_larger_than_minvec:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #3
-; CHECK-NEXT:    whilege p0.d, x8, xzr
-; CHECK-NEXT:    ret
-  %out = call <vscale x 2 x i1> @llvm.aarch64.sve.whilege.nxv2i1.i64(i64 3, i64 0)
-  ret <vscale x 2 x i1> %out
-}
-
+; Ensure we don't convert constant decrementing while instructions to ptrue.
 define <vscale x 16 x i1> @whilege_b_ii() {
 ; CHECK-LABEL: whilege_b_ii:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b, vl6
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i32(i32 3, i32 -2)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilege_b_ii_dont_fold_to_ptrue_nonexistent_vl9() {
-; CHECK-LABEL: whilege_b_ii_dont_fold_to_ptrue_nonexistent_vl9:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #9
-; CHECK-NEXT:    whilege p0.b, x8, xzr
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i64(i64 9, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilege_b_ii_vl_maximum() vscale_range(16, 16) {
-; CHECK-LABEL: whilege_b_ii_vl_maximum:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
-; CHECK-NEXT:    ret
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i64(i64 255, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilege_b_ii_dont_fold_to_ptrue_overflow() {
-; CHECK-LABEL: whilege_b_ii_dont_fold_to_ptrue_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #2
-; CHECK-NEXT:    mov w9, #2147483647
-; CHECK-NEXT:    movk w8, #32768, lsl #16
+; CHECK-NEXT:    mov w8, #-2 // =0xfffffffe
+; CHECK-NEXT:    mov w9, #3 // =0x3
 ; CHECK-NEXT:    whilege p0.b, w9, w8
 ; CHECK-NEXT:    ret
 entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i32(i32 2147483647, i32 -2147483646)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilege_b_ii_dont_fold_to_ptrue_increment_overflow() {
-; CHECK-LABEL: whilege_b_ii_dont_fold_to_ptrue_increment_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #2147483647
-; CHECK-NEXT:    mov w9, #-2147483641
-; CHECK-NEXT:    whilege p0.b, w9, w8
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i32(i32 -2147483641, i32 2147483647)
+  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i32(i32 3, i32 -2)
   ret <vscale x 16 x i1> %out
 }
 
@@ -219,69 +167,19 @@ define <vscale x 2 x i1> @whilehs_d_xx(i64 %a, i64 %b) {
   ret <vscale x 2 x i1> %out
 }
 
-define <vscale x 2 x i1> @whilehs_d_ii_dont_fold_to_ptrue_larger_than_minvec() {
-; CHECK-LABEL: whilehs_d_ii_dont_fold_to_ptrue_larger_than_minvec:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #3
-; CHECK-NEXT:    whilehs p0.d, x8, xzr
-; CHECK-NEXT:    ret
-  %out = call <vscale x 2 x i1> @llvm.aarch64.sve.whilehs.nxv2i1.i64(i64 3, i64 0)
-  ret <vscale x 2 x i1> %out
-}
-
+; Ensure we don't convert constant decrementing while instructions to ptrue.
 define <vscale x 16 x i1> @whilehs_b_ii() {
 ; CHECK-LABEL: whilehs_b_ii:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b, vl7
+; CHECK-NEXT:    mov w8, #2 // =0x2
+; CHECK-NEXT:    mov w9, #8 // =0x8
+; CHECK-NEXT:    whilehs p0.b, x9, x8
 ; CHECK-NEXT:    ret
 entry:
   %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehs.nxv16i1.i64(i64 8, i64 2)
   ret <vscale x 16 x i1> %out
 }
 
-define <vscale x 16 x i1> @whilehs_b_ii_dont_fold_to_ptrue_nonexistent_vl9() {
-; CHECK-LABEL: whilehs_b_ii_dont_fold_to_ptrue_nonexistent_vl9:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #9
-; CHECK-NEXT:    whilehs p0.b, x8, xzr
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehs.nxv16i1.i64(i64 9, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilehs_b_ii_vl_maximum() vscale_range(16, 16) {
-; CHECK-LABEL: whilehs_b_ii_vl_maximum:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
-; CHECK-NEXT:    ret
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehs.nxv16i1.i64(i64 255, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilehs_b_ii_dont_fold_to_ptrue_overflow() {
-; CHECK-LABEL: whilehs_b_ii_dont_fold_to_ptrue_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    mov w9, #6
-; CHECK-NEXT:    whilehs p0.b, w9, w8
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehs.nxv16i1.i32(i32 6, i32 4294967295)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilehs_b_ii_dont_fold_to_ptrue_increment_overflow() {
-; CHECK-LABEL: whilehs_b_ii_dont_fold_to_ptrue_increment_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    whilehs p0.b, w8, wzr
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehs.nxv16i1.i32(i32 4294967295, i32 0)
-  ret <vscale x 16 x i1> %out
-}
-
 ;
 ; WHILEGT
 ;
@@ -358,55 +256,16 @@ define <vscale x 2 x i1> @whilegt_d_xx(i64 %a, i64 %b) {
   ret <vscale x 2 x i1> %out
 }
 
-define <vscale x 2 x i1> @whilegt_d_ii_dont_fold_to_ptrue_larger_than_minvec() {
-; CHECK-LABEL: whilegt_d_ii_dont_fold_to_ptrue_larger_than_minvec:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #3
-; CHECK-NEXT:    whilegt p0.d, x8, xzr
-; CHECK-NEXT:    ret
-  %out = call <vscale x 2 x i1> @llvm.aarch64.sve.whilegt.nxv2i1.i64(i64 3, i64 0)
-  ret <vscale x 2 x i1> %out
-}
-
+; Ensure we don't convert constant decrementing while instructions to ptrue.
 define <vscale x 16 x i1> @whilegt_b_ii() {
 ; CHECK-LABEL: whilegt_b_ii:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b, vl5
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilegt.nxv16i1.i32(i32 3, i32 -2)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilegt_b_ii_fold_to_ptrue_nonexistent_vl9() {
-; CHECK-LABEL: whilegt_b_ii_fold_to_ptrue_nonexistent_vl9:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #9
-; CHECK-NEXT:    whilegt p0.b, x8, xzr
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilegt.nxv16i1.i64(i64 9, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilegt_b_ii_vl_maximum() vscale_range(16, 16) {
-; CHECK-LABEL: whilegt_b_ii_vl_maximum:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
-; CHECK-NEXT:    ret
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilegt.nxv16i1.i64(i64 256, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilegt_b_ii_dont_fold_to_ptrue_overflow() {
-; CHECK-LABEL: whilegt_b_ii_dont_fold_to_ptrue_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #2147483647
-; CHECK-NEXT:    mov w9, #-2147483641
+; CHECK-NEXT:    mov w8, #-2 // =0xfffffffe
+; CHECK-NEXT:    mov w9, #3 // =0x3
 ; CHECK-NEXT:    whilegt p0.b, w9, w8
 ; CHECK-NEXT:    ret
 entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilegt.nxv16i1.i32(i32 -2147483641, i32 2147483647)
+  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilegt.nxv16i1.i32(i32 3, i32 -2)
   ret <vscale x 16 x i1> %out
 }
 
@@ -486,58 +345,19 @@ define <vscale x 2 x i1> @whilehi_d_xx(i64 %a, i64 %b) {
   ret <vscale x 2 x i1> %out
 }
 
-define <vscale x 2 x i1> @whilehi_d_ii_dont_fold_to_ptrue_larger_than_minvec() {
-; CHECK-LABEL: whilehi_d_ii_dont_fold_to_ptrue_larger_than_minvec:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #3
-; CHECK-NEXT:    whilehi p0.d, x8, xzr
-; CHECK-NEXT:    ret
-  %out = call <vscale x 2 x i1> @llvm.aarch64.sve.whilehi.nxv2i1.i64(i64 3, i64 0)
-  ret <vscale x 2 x i1> %out
-}
-
+; Ensure we don't convert constant decrementing while instructions to ptrue.
 define <vscale x 16 x i1> @whilehi_b_ii() {
 ; CHECK-LABEL: whilehi_b_ii:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b, vl6
+; CHECK-NEXT:    mov w8, #2 // =0x2
+; CHECK-NEXT:    mov w9, #8 // =0x8
+; CHECK-NEXT:    whilehi p0.b, x9, x8
 ; CHECK-NEXT:    ret
 entry:
   %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehi.nxv16i1.i64(i64 8, i64 2)
   ret <vscale x 16 x i1> %out
 }
 
-define <vscale x 16 x i1> @whilehi_b_ii_dont_fold_to_ptrue_nonexistent_vl9() {
-; CHECK-LABEL: whilehi_b_ii_dont_fold_to_ptrue_nonexistent_vl9:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #9
-; CHECK-NEXT:    whilehi p0.b, x8, xzr
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehi.nxv16i1.i64(i64 9, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilehi_b_ii_vl_maximum() vscale_range(16, 16) {
-; CHECK-LABEL: whilehi_b_ii_vl_maximum:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
-; CHECK-NEXT:    ret
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehi.nxv16i1.i64(i64 256, i64 0)
-  ret <vscale x 16 x i1> %out
-}
-
-define <vscale x 16 x i1> @whilelhi_b_ii_dont_fold_to_ptrue_overflow() {
-; CHECK-LABEL: whilelhi_b_ii_dont_fold_to_ptrue_overflow:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    mov w9, #7
-; CHECK-NEXT:    whilehi p0.b, w9, w8
-; CHECK-NEXT:    ret
-entry:
-  %out = call <vscale x 16 x i1> @llvm.aarch64.sve.whilehi.nxv16i1.i32(i32 7, i32 4294967295)
-  ret <vscale x 16 x i1> %out
-}
-
 declare <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i32(i32, i32)
 declare <vscale x 16 x i1> @llvm.aarch64.sve.whilege.nxv16i1.i64(i64, i64)
 declare <vscale x 8 x i1> @llvm.aarch64.sve.whilege.nxv8i1.i32(i32, i32)

From 7b039c0cf0045d81da7e1575232a46266cff1e88 Mon Sep 17 00:00:00 2001
From: Stephen Tozer <stephen.tozer@sony.com>
Date: Mon, 15 Apr 2024 13:15:27 +0100
Subject: [PATCH 421/886] [DebugInfo][RemoveDIs] Fix error from not tracking
 seen debug labels (#88718)

Fixes the reported errors on:
https://github.com/llvm/llvm-project/pull/87379

A previous patch updated the bitcode reading for debug
intrinsics/records to not perform the expensive debug info format
conversion from records to intrinsics in cases where no records were
present, but the patch did not actually track when debug labels had been
seen, resulting in errors when parsing bitcode where functions contained
debug label records but no other debug records. This patch fixes that
case and adds a test for it.
---
 llvm/lib/Bitcode/Reader/BitcodeReader.cpp |  1 +
 llvm/test/Bitcode/dbg-label-record-bc.ll  | 50 +++++++++++++++++++++++
 2 files changed, 51 insertions(+)
 create mode 100644 llvm/test/Bitcode/dbg-label-record-bc.ll

diff --git a/llvm/lib/Bitcode/Reader/BitcodeReader.cpp b/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
index fe4f0d6dca6c0..0b7fcd8841889 100644
--- a/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
+++ b/llvm/lib/Bitcode/Reader/BitcodeReader.cpp
@@ -6454,6 +6454,7 @@ Error BitcodeReader::parseFunctionBody(Function *F) {
     case bitc::FUNC_CODE_DEBUG_RECORD_LABEL: {
       // DbgLabelRecords are placed after the Instructions that they are
       // attached to.
+      SeenDebugRecord = true;
       Instruction *Inst = getLastInstruction();
       if (!Inst)
         return error("Invalid dbg record: missing instruction");
diff --git a/llvm/test/Bitcode/dbg-label-record-bc.ll b/llvm/test/Bitcode/dbg-label-record-bc.ll
new file mode 100644
index 0000000000000..e151f7f6cc157
--- /dev/null
+++ b/llvm/test/Bitcode/dbg-label-record-bc.ll
@@ -0,0 +1,50 @@
+;; Tests that we can parse and print a function containing a debug label record
+;; and no other debug record kinds.
+
+; RUN: llvm-as --write-experimental-debuginfo-iterators-to-bitcode=true %s -o - \
+; RUN: | opt -S | FileCheck %s --check-prefixes=CHECK,INTRINSIC
+
+; RUN: llvm-as --write-experimental-debuginfo-iterators-to-bitcode=true %s -o - \
+; RUN: | opt -S --preserve-input-debuginfo-format=true \
+; RUN: | FileCheck %s --check-prefixes=CHECK,RECORD
+
+source_filename = "bbi-94196.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; CHECK-LABEL: void @foo()
+; CHECK: bar:
+; INTRINSIC-NEXT: call void @llvm.dbg.label(metadata ![[LABEL:[0-9]+]]), !dbg ![[LOC:[0-9]+]]
+; RECORD-NEXT: #dbg_label(![[LABEL:[0-9]+]], ![[LOC:[0-9]+]])
+
+; CHECK-DAG: ![[LABEL]] = !DILabel({{.*}}name: "bar"
+; CHECK-DAG: ![[LOC]] = !DILocation(line: 5, column: 1
+
+define dso_local void @foo() !dbg !5 {
+entry:
+  br label %bar, !dbg !9
+
+bar:                                              ; preds = %entry
+  tail call void @llvm.dbg.label(metadata !10), !dbg !11
+  ret void, !dbg !12
+}
+
+declare void @llvm.dbg.label(metadata)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!2, !3}
+!llvm.ident = !{!4}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C11, file: !1, producer: "clang version 19.0.0git", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "<stdin>", directory: "/home/gbtozers/dev/llvm-project-ddd-textual-ir")
+!2 = !{i32 2, !"Debug Info Version", i32 3}
+!3 = !{i32 1, !"wchar_size", i32 4}
+!4 = !{!"clang version 19.0.0git"}
+!5 = distinct !DISubprogram(name: "foo", scope: !6, file: !6, line: 1, type: !7, scopeLine: 2, spFlags: DISPFlagDefinition, unit: !0)
+!6 = !DIFile(filename: "bbi-94196.c", directory: "/home/gbtozers/dev/llvm-project-ddd-textual-ir")
+!7 = !DISubroutineType(types: !8)
+!8 = !{null}
+!9 = !DILocation(line: 3, column: 3, scope: !5)
+!10 = !DILabel(scope: !5, name: "bar", file: !6, line: 5)
+!11 = !DILocation(line: 5, column: 1, scope: !5)
+!12 = !DILocation(line: 6, column: 3, scope: !5)

From 58d4470fa49443da1477a7d2e43685e91bbd6630 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Mon, 15 Apr 2024 13:02:18 +0100
Subject: [PATCH 422/886] [X86] Remove scaleVectorShuffleBlendMask and use
 APIntOps::ScaleBitMask directly.

---
 llvm/lib/Target/X86/X86ISelLowering.cpp | 18 +++++-------------
 1 file changed, 5 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 0a91e3f554b3d..f16a751a166d6 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -10589,15 +10589,6 @@ static bool matchShuffleAsBlend(MVT VT, SDValue V1, SDValue V2,
   return true;
 }
 
-static uint64_t scaleVectorShuffleBlendMask(uint64_t BlendMask, int Size,
-                                            int Scale) {
-  uint64_t ScaledMask = 0;
-  for (int i = 0; i != Size; ++i)
-    if (BlendMask & (1ull << i))
-      ScaledMask |= ((1ull << Scale) - 1) << (i * Scale);
-  return ScaledMask;
-}
-
 /// Try to emit a blend instruction for a shuffle.
 ///
 /// This doesn't do any checks for the availability of instructions for blending
@@ -40539,14 +40530,15 @@ static SDValue combineTargetShuffle(SDValue N, const SDLoc &DL,
       MVT SrcVT = N0.getOperand(0).getSimpleValueType();
       if ((VT.getScalarSizeInBits() % SrcVT.getScalarSizeInBits()) == 0 &&
           SrcVT.getScalarSizeInBits() >= 32) {
-        unsigned BlendMask = N.getConstantOperandVal(2);
         unsigned Size = VT.getVectorNumElements();
-        unsigned Scale = VT.getScalarSizeInBits() / SrcVT.getScalarSizeInBits();
-        BlendMask = scaleVectorShuffleBlendMask(BlendMask, Size, Scale);
+        unsigned NewSize = SrcVT.getVectorNumElements();
+        APInt BlendMask = N.getConstantOperandAPInt(2).zextOrTrunc(Size);
+        APInt NewBlendMask = APIntOps::ScaleBitMask(BlendMask, NewSize);
         return DAG.getBitcast(
             VT, DAG.getNode(X86ISD::BLENDI, DL, SrcVT, N0.getOperand(0),
                             N1.getOperand(0),
-                            DAG.getTargetConstant(BlendMask, DL, MVT::i8)));
+                            DAG.getTargetConstant(NewBlendMask.getZExtValue(),
+                                                  DL, MVT::i8)));
       }
     }
     return SDValue();

From 7089c359a3845323f6f30c44a47dd901f2edfe63 Mon Sep 17 00:00:00 2001
From: Paul Walker <paul.walker@arm.com>
Date: Mon, 15 Apr 2024 13:50:09 +0100
Subject: [PATCH 423/886] [LLVM][SelectionDAG] Allow verification of target ISD
 nodes. (#88121)

Patch includes an initial implementation for AArch64 that covers a
handful of nodes where I've observed bogus nodes within the DAG.
---
 llvm/include/llvm/CodeGen/TargetLowering.h    |  5 +++
 .../lib/CodeGen/SelectionDAG/SelectionDAG.cpp |  6 ++-
 .../Target/AArch64/AArch64ISelLowering.cpp    | 43 +++++++++++++++++++
 llvm/lib/Target/AArch64/AArch64ISelLowering.h |  4 ++
 4 files changed, 56 insertions(+), 2 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index 980d75ad91a0d..e0ade02959025 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -4873,6 +4873,11 @@ class TargetLowering : public TargetLoweringBase {
   bool verifyReturnAddressArgumentIsConstant(SDValue Op,
                                              SelectionDAG &DAG) const;
 
+#ifndef NDEBUG
+  /// Check the given SDNode.  Aborts if it is invalid.
+  virtual void verifyTargetSDNode(const SDNode *N) const {};
+#endif
+
   //===--------------------------------------------------------------------===//
   // Inline Asm Support hooks
   //
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index fc972664e5f01..ca0a95750ba8d 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -1111,9 +1111,11 @@ void SelectionDAG::DeallocateNode(SDNode *N) {
 
 #ifndef NDEBUG
 /// VerifySDNode - Check the given SDNode.  Aborts if it is invalid.
-static void VerifySDNode(SDNode *N) {
+static void VerifySDNode(SDNode *N, const TargetLowering *TLI) {
   switch (N->getOpcode()) {
   default:
+    if (N->getOpcode() > ISD::BUILTIN_OP_END)
+      TLI->verifyTargetSDNode(N);
     break;
   case ISD::BUILD_PAIR: {
     EVT VT = N->getValueType(0);
@@ -1157,7 +1159,7 @@ void SelectionDAG::InsertNode(SDNode *N) {
   AllNodes.push_back(N);
 #ifndef NDEBUG
   N->PersistentId = NextPersistentId++;
-  VerifySDNode(N);
+  VerifySDNode(N, TLI);
 #endif
   for (DAGUpdateListener *DUL = UpdateListeners; DUL; DUL = DUL->Next)
     DUL->NodeInserted(N);
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 27dc79f3f8d0c..eee67a0f823c1 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -27868,3 +27868,46 @@ bool AArch64TargetLowering::hasInlineStackProbe(
   return !Subtarget->isTargetWindows() &&
          MF.getInfo<AArch64FunctionInfo>()->hasStackProbing();
 }
+
+#ifndef NDEBUG
+void AArch64TargetLowering::verifyTargetSDNode(const SDNode *N) const {
+  switch (N->getOpcode()) {
+  default:
+    break;
+  case AArch64ISD::SUNPKLO:
+  case AArch64ISD::SUNPKHI:
+  case AArch64ISD::UUNPKLO:
+  case AArch64ISD::UUNPKHI: {
+    assert(N->getNumValues() == 1 && "Expected one result!");
+    assert(N->getNumOperands() == 1 && "Expected one operand!");
+    EVT VT = N->getValueType(0);
+    EVT OpVT = N->getOperand(0).getValueType();
+    assert(OpVT.isVector() && VT.isVector() && OpVT.isInteger() &&
+           VT.isInteger() && "Expected integer vectors!");
+    assert(OpVT.getSizeInBits() == VT.getSizeInBits() &&
+           "Expected vectors of equal size!");
+    // TODO: Enable assert once bogus creations have been fixed.
+    // assert(OpVT.getVectorElementCount() == VT.getVectorElementCount()*2 &&
+    //       "Expected result vector with half the lanes of its input!");
+    break;
+  }
+  case AArch64ISD::TRN1:
+  case AArch64ISD::TRN2:
+  case AArch64ISD::UZP1:
+  case AArch64ISD::UZP2:
+  case AArch64ISD::ZIP1:
+  case AArch64ISD::ZIP2: {
+    assert(N->getNumValues() == 1 && "Expected one result!");
+    assert(N->getNumOperands() == 2 && "Expected two operands!");
+    EVT VT = N->getValueType(0);
+    EVT Op0VT = N->getOperand(0).getValueType();
+    EVT Op1VT = N->getOperand(1).getValueType();
+    assert(VT.isVector() && Op0VT.isVector() && Op1VT.isVector() &&
+           "Expected vectors!");
+    // TODO: Enable assert once bogus creations have been fixed.
+    // assert(VT == Op0VT && VT == Op1VT && "Expected matching vectors!");
+    break;
+  }
+  }
+}
+#endif
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
index 18439dc7f0102..db6e8a00d2fb5 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -998,6 +998,10 @@ class AArch64TargetLowering : public TargetLowering {
   /// True if stack clash protection is enabled for this functions.
   bool hasInlineStackProbe(const MachineFunction &MF) const override;
 
+#ifndef NDEBUG
+  void verifyTargetSDNode(const SDNode *N) const override;
+#endif
+
 private:
   /// Keep a pointer to the AArch64Subtarget around so that we can
   /// make the right decision when generating code for different targets.

From 2e0e3b0f10b5c433504e31ca73e55f86e0b27660 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Fri, 12 Apr 2024 07:50:59 -0700
Subject: [PATCH 424/886] [RISCV] Split single width convert to FP pseudos by
 SEW

---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 61 +++++++++++--------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |  2 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  4 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |  4 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 26 ++------
 llvm/lib/Target/RISCV/RISCVScheduleV.td       |  8 +--
 6 files changed, 51 insertions(+), 54 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index a5f56a1c209c6..e9cb027592706 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -3593,12 +3593,15 @@ multiclass VPseudoConversionRoundingMode<VReg RetClass,
                              VReg Op1Class,
                              LMULInfo MInfo,
                              string Constraint = "",
+                             int sew = 0,
                              int TargetConstraintType = 1> {
   let VLMul = MInfo.value in {
-    def "_" # MInfo.MX : VPseudoUnaryNoMaskRoundingMode<RetClass, Op1Class, Constraint, TargetConstraintType>;
-    def "_" # MInfo.MX # "_MASK" : VPseudoUnaryMaskRoundingMode<RetClass, Op1Class,
-                                                                Constraint, TargetConstraintType>,
-                                   RISCVMaskedPseudo<MaskIdx=2>;
+    defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
+    def suffix : VPseudoUnaryNoMaskRoundingMode<RetClass, Op1Class, Constraint, TargetConstraintType>;
+    def suffix # "_MASK" : VPseudoUnaryMaskRoundingMode<RetClass, Op1Class,
+                                                        Constraint,
+                                                        TargetConstraintType>,
+                           RISCVMaskedPseudo<MaskIdx=2>;
   }
 }
 
@@ -3607,13 +3610,15 @@ multiclass VPseudoConversionRM<VReg RetClass,
                                VReg Op1Class,
                                LMULInfo MInfo,
                                string Constraint = "",
+                               int sew = 0,
                                int TargetConstraintType = 1> {
   let VLMul = MInfo.value in {
-    def "_" # MInfo.MX : VPseudoUnaryNoMask_FRM<RetClass, Op1Class,
-                                                        Constraint, TargetConstraintType>;
-    def "_" # MInfo.MX # "_MASK" : VPseudoUnaryMask_FRM<RetClass, Op1Class,
-                                                        Constraint, TargetConstraintType>,
-                                   RISCVMaskedPseudo<MaskIdx=2>;
+    defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
+    def suffix : VPseudoUnaryNoMask_FRM<RetClass, Op1Class,
+                                        Constraint, TargetConstraintType>;
+    def suffix # "_MASK" : VPseudoUnaryMask_FRM<RetClass, Op1Class,
+                                                Constraint, TargetConstraintType>,
+                           RISCVMaskedPseudo<MaskIdx=2>;
   }
 }
 
@@ -3660,17 +3665,19 @@ multiclass VPseudoVFROUND_NOEXCEPT_V {
 
 multiclass VPseudoVCVTF_V_RM {
   foreach m = MxListF in {
-    defm _V : VPseudoConversionRoundingMode<m.vrclass, m.vrclass, m>,
-              SchedUnary<"WriteVFCvtIToFV", "ReadVFCvtIToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm _V : VPseudoConversionRoundingMode<m.vrclass, m.vrclass, m, sew=e>,
+                SchedUnary<"WriteVFCvtIToFV", "ReadVFCvtIToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
 multiclass VPseudoVCVTF_RM_V {
   foreach m = MxListF in {
-    defm _V : VPseudoConversionRM<m.vrclass, m.vrclass, m>,
-              SchedUnary<"WriteVFCvtIToFV", "ReadVFCvtIToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm _V : VPseudoConversionRM<m.vrclass, m.vrclass, m, sew=e>,
+                SchedUnary<"WriteVFCvtIToFV", "ReadVFCvtIToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
@@ -4905,14 +4912,17 @@ multiclass VPatConversionTARoundingMode<string intrinsic,
                                         ValueType result_type,
                                         ValueType op1_type,
                                         ValueType mask_type,
-                                        int sew,
+                                        int log2sew,
                                         LMULInfo vlmul,
                                         VReg result_reg_class,
-                                        VReg op1_reg_class> {
+                                        VReg op1_reg_class,
+                                        bit isSEWAware = 0> {
   def : VPatUnaryNoMaskRoundingMode<intrinsic, inst, kind, result_type, op1_type,
-                                    sew, vlmul, result_reg_class, op1_reg_class>;
+                                    log2sew, vlmul, result_reg_class,
+                                    op1_reg_class, isSEWAware>;
   def : VPatUnaryMaskRoundingMode<intrinsic, inst, kind, result_type, op1_type,
-                                  mask_type, sew, vlmul, result_reg_class, op1_reg_class>;
+                                  mask_type, log2sew, vlmul, result_reg_class,
+                                  op1_reg_class, isSEWAware>;
 }
 
 multiclass VPatBinaryV_VV<string intrinsic, string instruction,
@@ -5905,15 +5915,16 @@ multiclass VPatConversionVI_VF_RM<string intrinsic,
   }
 }
 
-multiclass VPatConversionVF_VI_RM<string intrinsic,
-                                  string instruction> {
+multiclass VPatConversionVF_VI_RM<string intrinsic, string instruction,
+                                  bit isSEWAware = 0> {
   foreach fvti = AllFloatVectors in {
     defvar ivti = GetIntVTypeInfo<fvti>.Vti;
     let Predicates = !listconcat(GetVTypePredicates<fvti>.Predicates,
                                  GetVTypePredicates<ivti>.Predicates) in
     defm : VPatConversionTARoundingMode<intrinsic, instruction, "V",
                                         fvti.Vector, ivti.Vector, fvti.Mask, ivti.Log2SEW,
-                                        ivti.LMul, fvti.RegClass, ivti.RegClass>;
+                                        ivti.LMul, fvti.RegClass, ivti.RegClass,
+                                        isSEWAware>;
   }
 }
 
@@ -7269,8 +7280,10 @@ defm : VPatConversionVI_VF_RM<"int_riscv_vfcvt_x_f_v", "PseudoVFCVT_X_F">;
 defm : VPatConversionVI_VF_RM<"int_riscv_vfcvt_xu_f_v", "PseudoVFCVT_XU_F">;
 defm : VPatConversionVI_VF<"int_riscv_vfcvt_rtz_xu_f_v", "PseudoVFCVT_RTZ_XU_F">;
 defm : VPatConversionVI_VF<"int_riscv_vfcvt_rtz_x_f_v", "PseudoVFCVT_RTZ_X_F">;
-defm : VPatConversionVF_VI_RM<"int_riscv_vfcvt_f_x_v", "PseudoVFCVT_F_X">;
-defm : VPatConversionVF_VI_RM<"int_riscv_vfcvt_f_xu_v", "PseudoVFCVT_F_XU">;
+defm : VPatConversionVF_VI_RM<"int_riscv_vfcvt_f_x_v", "PseudoVFCVT_F_X",
+                              isSEWAware=1>;
+defm : VPatConversionVF_VI_RM<"int_riscv_vfcvt_f_xu_v", "PseudoVFCVT_F_XU",
+                              isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.18. Widening Floating-Point/Integer Type-Convert Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 7c77449b4f6e1..788db6f447a18 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -410,7 +410,7 @@ multiclass VPatConvertI2FPSDNode_V_RM<SDPatternOperator vop,
     let Predicates = !listconcat(GetVTypePredicates<fvti>.Predicates,
                                  GetVTypePredicates<ivti>.Predicates) in
     def : Pat<(fvti.Vector (vop (ivti.Vector ivti.RegClass:$rs1))),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX)
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW)
                   (fvti.Vector (IMPLICIT_DEF)),
                   ivti.RegClass:$rs1,
                   // Value to indicate no rounding mode change in
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index 322c055306e86..2b56eea29cfcc 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1229,7 +1229,7 @@ multiclass VPatConvertI2FPVL_V_RM<SDPatternOperator vop, string instruction_name
     def : Pat<(fvti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
                                 (ivti.Mask V0),
                                 VLOpFrag)),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                   (fvti.Vector (IMPLICIT_DEF)), ivti.RegClass:$rs1,
                   (ivti.Mask V0),
                   // Value to indicate no rounding mode change in
@@ -1247,7 +1247,7 @@ multiclass VPatConvertI2FP_RM_VL_V<SDNode vop, string instruction_name> {
     def : Pat<(fvti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
                                 (ivti.Mask V0), (XLenVT timm:$frm),
                                 VLOpFrag)),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                   (fvti.Vector (IMPLICIT_DEF)), ivti.RegClass:$rs1,
                   (ivti.Mask V0), timm:$frm, GPR:$vl, fvti.Log2SEW, TA_MA)>;
   }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 2a13cb4beac14..354fe7c27e68d 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -744,6 +744,7 @@ foreach mx = SchedMxListF in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV",   [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -751,7 +752,6 @@ foreach mx = SchedMxList in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
   let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFCvtIToFV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCvtFToIV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
   let Latency = 4, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
@@ -1179,7 +1179,7 @@ defm "" : LMULReadAdvance<"ReadVFClassV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
-defm "" : LMULReadAdvance<"ReadVFCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 80090a0027149..863e245c32256 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -495,36 +495,20 @@ foreach mx = SchedMxListF in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFALUF",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulV",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulF",  [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
-
-    }
-  }
-}
-
-foreach mx = SchedMxListF in {
-  foreach sew = SchedSEWSet<mx, isF=1>.val in {
-    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
-    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
-    let Latency = 6, ReleaseAtCycles = [LMulLat] in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFMulAddF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
     }
-  }
-}
-foreach mx = SchedMxListF in {
-  foreach sew = SchedSEWSet<mx, isF=1>.val in {
-    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
-    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListF, isF=1>.c;
     let Latency = 2, ReleaseAtCycles = [LMulLat] in
     defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    let Latency = 3, ReleaseAtCycles = [LMulLat] in
+    defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
   }
 }
 foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
-  let Latency = 3, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFCvtIToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFCvtFToIV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-  }
+  let Latency = 3, ReleaseAtCycles = [LMulLat] in
+  defm "" : LMULWriteResMX<"WriteVFCvtFToIV", [SiFiveP600VectorArith], mx, IsWorstCase>;
   let Latency = 2, ReleaseAtCycles = [LMulLat] in {
     defm "" : LMULWriteResMX<"WriteVFCmpV",  [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFCmpF",  [SiFiveP600VectorArith], mx, IsWorstCase>;
@@ -976,7 +960,7 @@ defm "" : LMULReadAdvance<"ReadVFClassV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
-defm "" : LMULReadAdvance<"ReadVFCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 6070482ea1757..e7acc0c87d704 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -449,7 +449,7 @@ defm "" : LMULSchedWrites<"WriteVFMergeV">;
 // 13.16. Vector Floating-Point Move Instruction
 defm "" : LMULSchedWrites<"WriteVFMovV">;
 // 13.17. Single-Width Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedWrites<"WriteVFCvtIToFV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFCvtIToFV">;
 defm "" : LMULSchedWrites<"WriteVFCvtFToIV">;
 // 13.18. Widening Floating-Point/Integer Type-Convert Instructions
 defm "" : LMULSchedWritesW<"WriteVFWCvtIToFV">;
@@ -675,7 +675,7 @@ defm "" : LMULSchedReads<"ReadVFMergeF">;
 // 13.16. Vector Floating-Point Move Instruction
 defm "" : LMULSchedReads<"ReadVFMovF">;
 // 13.17. Single-Width Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedReads<"ReadVFCvtIToFV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFCvtIToFV">;
 defm "" : LMULSchedReads<"ReadVFCvtFToIV">;
 // 13.18. Widening Floating-Point/Integer Type-Convert Instructions
 defm "" : LMULSchedReadsW<"ReadVFWCvtIToFV">;
@@ -905,7 +905,7 @@ defm "" : LMULWriteRes<"WriteVFCmpF", []>;
 defm "" : LMULWriteRes<"WriteVFClassV", []>;
 defm "" : LMULWriteRes<"WriteVFMergeV", []>;
 defm "" : LMULWriteRes<"WriteVFMovV", []>;
-defm "" : LMULWriteRes<"WriteVFCvtIToFV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFCvtIToFV", []>;
 defm "" : LMULWriteRes<"WriteVFCvtFToIV", []>;
 defm "" : LMULWriteResW<"WriteVFWCvtIToFV", []>;
 defm "" : LMULWriteResFW<"WriteVFWCvtFToIV", []>;
@@ -1062,7 +1062,7 @@ defm "" : LMULReadAdvance<"ReadVFClassV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
-defm "" : LMULReadAdvance<"ReadVFCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;

From 60a1158f31bb99a5d049d5cac5a704d906a1277e Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Fri, 12 Apr 2024 12:16:20 -0700
Subject: [PATCH 425/886] [RISCV] Split Widening convert to FP pseudos by SEW

---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 100 +++++++++++++++---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    |  63 ++++++-----
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |   2 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |   4 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    |  22 ++--
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td |  20 ++--
 llvm/lib/Target/RISCV/RISCVScheduleV.td       |  28 +++--
 7 files changed, 168 insertions(+), 71 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 5a572002091ff..27387595164a4 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -17642,8 +17642,7 @@ static MachineBasicBlock *emitSelectPseudo(MachineInstr &MI,
 
 static MachineBasicBlock *emitVFROUND_NOEXCEPT_MASK(MachineInstr &MI,
                                                     MachineBasicBlock *BB,
-                                                    unsigned CVTXOpc,
-                                                    unsigned CVTFOpc) {
+                                                    unsigned CVTXOpc) {
   DebugLoc DL = MI.getDebugLoc();
 
   const TargetInstrInfo &TII = *BB->getParent()->getSubtarget().getInstrInfo();
@@ -17674,6 +17673,85 @@ static MachineBasicBlock *emitVFROUND_NOEXCEPT_MASK(MachineInstr &MI,
                                      /*IsImp*/ true));
 
   // Emit a VFCVT_F_X
+  RISCVII::VLMUL LMul = RISCVII::getLMul(MI.getDesc().TSFlags);
+  unsigned Log2SEW = MI.getOperand(RISCVII::getSEWOpNum(MI.getDesc())).getImm();
+  // There is no E8 variant for VFCVT_F_X.
+  assert(Log2SEW >= 4);
+  // Since MI (VFROUND) isn't SEW specific, we cannot use a macro to make
+  // handling of different (LMUL, SEW) pairs easier because we need to pull the
+  // SEW immediate from MI, and that information is not avaliable during macro
+  // expansion.
+  unsigned CVTFOpc;
+  if (Log2SEW == 4) {
+    switch (LMul) {
+    case RISCVII::LMUL_1:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M1_E16_MASK;
+      break;
+    case RISCVII::LMUL_2:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M2_E16_MASK;
+      break;
+    case RISCVII::LMUL_4:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M4_E16_MASK;
+      break;
+    case RISCVII::LMUL_8:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M8_E16_MASK;
+      break;
+    case RISCVII::LMUL_F2:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_MF2_E16_MASK;
+      break;
+    case RISCVII::LMUL_F4:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_MF4_E16_MASK;
+      break;
+    case RISCVII::LMUL_F8:
+    case RISCVII::LMUL_RESERVED:
+      llvm_unreachable("Unexpected LMUL and SEW combination value for MI.");
+    }
+  } else if (Log2SEW == 5) {
+    switch (LMul) {
+    case RISCVII::LMUL_1:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M1_E32_MASK;
+      break;
+    case RISCVII::LMUL_2:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M2_E32_MASK;
+      break;
+    case RISCVII::LMUL_4:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M4_E32_MASK;
+      break;
+    case RISCVII::LMUL_8:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M8_E32_MASK;
+      break;
+    case RISCVII::LMUL_F2:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_MF2_E32_MASK;
+      break;
+    case RISCVII::LMUL_F4:
+    case RISCVII::LMUL_F8:
+    case RISCVII::LMUL_RESERVED:
+      llvm_unreachable("Unexpected LMUL and SEW combination value for MI.");
+    }
+  } else if (Log2SEW == 6) {
+    switch (LMul) {
+    case RISCVII::LMUL_1:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M1_E64_MASK;
+      break;
+    case RISCVII::LMUL_2:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M2_E64_MASK;
+      break;
+    case RISCVII::LMUL_4:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M4_E64_MASK;
+      break;
+    case RISCVII::LMUL_8:
+      CVTFOpc = RISCV::PseudoVFCVT_F_X_V_M8_E64_MASK;
+      break;
+    case RISCVII::LMUL_F2:
+    case RISCVII::LMUL_F4:
+    case RISCVII::LMUL_F8:
+    case RISCVII::LMUL_RESERVED:
+      llvm_unreachable("Unexpected LMUL and SEW combination value for MI.");
+    }
+  } else {
+    llvm_unreachable("Unexpected LMUL and SEW combination value for MI.");
+  }
+
   BuildMI(*BB, MI, DL, TII.get(CVTFOpc))
       .add(MI.getOperand(0))
       .add(MI.getOperand(1))
@@ -17883,23 +17961,17 @@ RISCVTargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
                          Subtarget);
 
   case RISCV::PseudoVFROUND_NOEXCEPT_V_M1_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M1_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_M1_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M1_MASK);
   case RISCV::PseudoVFROUND_NOEXCEPT_V_M2_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M2_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_M2_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M2_MASK);
   case RISCV::PseudoVFROUND_NOEXCEPT_V_M4_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M4_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_M4_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M4_MASK);
   case RISCV::PseudoVFROUND_NOEXCEPT_V_M8_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M8_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_M8_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_M8_MASK);
   case RISCV::PseudoVFROUND_NOEXCEPT_V_MF2_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_MF2_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_MF2_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_MF2_MASK);
   case RISCV::PseudoVFROUND_NOEXCEPT_V_MF4_MASK:
-    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_MF4_MASK,
-                                     RISCV::PseudoVFCVT_F_X_V_MF4_MASK);
+    return emitVFROUND_NOEXCEPT_MASK(MI, BB, RISCV::PseudoVFCVT_X_F_V_MF4_MASK);
   case RISCV::PseudoFROUND_H:
   case RISCV::PseudoFROUND_H_INX:
   case RISCV::PseudoFROUND_S:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index e9cb027592706..9906049e9aaf9 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -3580,12 +3580,14 @@ multiclass VPseudoConversion<VReg RetClass,
                              VReg Op1Class,
                              LMULInfo MInfo,
                              string Constraint = "",
+                             int sew = 0,
                              int TargetConstraintType = 1> {
+  defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
   let VLMul = MInfo.value in {
-    def "_" # MInfo.MX : VPseudoUnaryNoMask<RetClass, Op1Class, Constraint, TargetConstraintType>;
-    def "_" # MInfo.MX # "_MASK" : VPseudoUnaryMask<RetClass, Op1Class,
-                                                    Constraint, TargetConstraintType>,
-                                   RISCVMaskedPseudo<MaskIdx=2>;
+    def suffix : VPseudoUnaryNoMask<RetClass, Op1Class, Constraint, TargetConstraintType>;
+    def suffix # "_MASK" : VPseudoUnaryMask<RetClass, Op1Class,
+                                            Constraint, TargetConstraintType>,
+                           RISCVMaskedPseudo<MaskIdx=2>;
   }
 }
 
@@ -3711,18 +3713,22 @@ multiclass VPseudoVWCVTI_RM_V {
 multiclass VPseudoVWCVTF_V {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListW in {
-    defm _V : VPseudoConversion<m.wvrclass, m.vrclass, m, constraint, TargetConstraintType=3>,
-              SchedUnary<"WriteVFWCvtIToFV", "ReadVFWCvtIToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=0, isWidening=1>.val in
+      defm _V : VPseudoConversion<m.wvrclass, m.vrclass, m, constraint, sew=e,
+                                  TargetConstraintType=3>,
+                SchedUnary<"WriteVFWCvtIToFV", "ReadVFWCvtIToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
 multiclass VPseudoVWCVTD_V {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListFW in {
-    defm _V : VPseudoConversion<m.wvrclass, m.vrclass, m, constraint, TargetConstraintType=3>,
-              SchedUnary<"WriteVFWCvtFToFV", "ReadVFWCvtFToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm _V : VPseudoConversion<m.wvrclass, m.vrclass, m, constraint, sew=e,
+                                  TargetConstraintType=3>,
+                SchedUnary<"WriteVFWCvtFToFV", "ReadVFWCvtFToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
@@ -4896,14 +4902,17 @@ multiclass VPatConversionTA<string intrinsic,
                             ValueType result_type,
                             ValueType op1_type,
                             ValueType mask_type,
-                            int sew,
+                            int log2sew,
                             LMULInfo vlmul,
                             VReg result_reg_class,
-                            VReg op1_reg_class> {
+                            VReg op1_reg_class,
+                            bit isSEWAware = 0> {
   def : VPatUnaryNoMask<intrinsic, inst, kind, result_type, op1_type,
-                        sew, vlmul, result_reg_class, op1_reg_class>;
+                        log2sew, vlmul, result_reg_class, op1_reg_class,
+                        isSEWAware>;
   def : VPatUnaryMask<intrinsic, inst, kind, result_type, op1_type,
-                      mask_type, sew, vlmul, result_reg_class, op1_reg_class>;
+                      mask_type, log2sew, vlmul, result_reg_class, op1_reg_class,
+                      isSEWAware>;
 }
 
 multiclass VPatConversionTARoundingMode<string intrinsic,
@@ -5952,7 +5961,8 @@ multiclass VPatConversionWI_VF_RM<string intrinsic, string instruction> {
   }
 }
 
-multiclass VPatConversionWF_VI<string intrinsic, string instruction> {
+multiclass VPatConversionWF_VI<string intrinsic, string instruction,
+                               bit isSEWAware = 0> {
   foreach vtiToWti = AllWidenableIntToFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar fwti = vtiToWti.Wti;
@@ -5960,11 +5970,12 @@ multiclass VPatConversionWF_VI<string intrinsic, string instruction> {
                                  GetVTypePredicates<fwti>.Predicates) in
     defm : VPatConversionTA<intrinsic, instruction, "V",
                             fwti.Vector, vti.Vector, fwti.Mask, vti.Log2SEW,
-                            vti.LMul, fwti.RegClass, vti.RegClass>;
+                            vti.LMul, fwti.RegClass, vti.RegClass, isSEWAware>;
   }
 }
 
-multiclass VPatConversionWF_VF<string intrinsic, string instruction> {
+multiclass VPatConversionWF_VF<string intrinsic, string instruction,
+                               bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
@@ -5974,11 +5985,12 @@ multiclass VPatConversionWF_VF<string intrinsic, string instruction> {
                                      GetVTypePredicates<fwti>.Predicates)) in
       defm : VPatConversionTA<intrinsic, instruction, "V",
                               fwti.Vector, fvti.Vector, fwti.Mask, fvti.Log2SEW,
-                              fvti.LMul, fwti.RegClass, fvti.RegClass>;
+                              fvti.LMul, fwti.RegClass, fvti.RegClass, isSEWAware>;
   }
 }
 
-multiclass VPatConversionWF_VF_BF <string intrinsic, string instruction> {
+multiclass VPatConversionWF_VF_BF <string intrinsic, string instruction,
+                                   bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableBFloatToFloatVectors in
   {
     defvar fvti = fvtiToFWti.Vti;
@@ -5987,7 +5999,7 @@ multiclass VPatConversionWF_VF_BF <string intrinsic, string instruction> {
                                  GetVTypePredicates<fwti>.Predicates) in
     defm : VPatConversionTA<intrinsic, instruction, "V",
                             fwti.Vector, fvti.Vector, fwti.Mask, fvti.Log2SEW,
-                            fvti.LMul, fwti.RegClass, fvti.RegClass>;
+                            fvti.LMul, fwti.RegClass, fvti.RegClass, isSEWAware>;
   }
 }
 
@@ -7292,11 +7304,14 @@ defm : VPatConversionWI_VF_RM<"int_riscv_vfwcvt_xu_f_v", "PseudoVFWCVT_XU_F">;
 defm : VPatConversionWI_VF_RM<"int_riscv_vfwcvt_x_f_v", "PseudoVFWCVT_X_F">;
 defm : VPatConversionWI_VF<"int_riscv_vfwcvt_rtz_xu_f_v", "PseudoVFWCVT_RTZ_XU_F">;
 defm : VPatConversionWI_VF<"int_riscv_vfwcvt_rtz_x_f_v", "PseudoVFWCVT_RTZ_X_F">;
-defm : VPatConversionWF_VI<"int_riscv_vfwcvt_f_xu_v", "PseudoVFWCVT_F_XU">;
-defm : VPatConversionWF_VI<"int_riscv_vfwcvt_f_x_v", "PseudoVFWCVT_F_X">;
-defm : VPatConversionWF_VF<"int_riscv_vfwcvt_f_f_v", "PseudoVFWCVT_F_F">;
+defm : VPatConversionWF_VI<"int_riscv_vfwcvt_f_xu_v", "PseudoVFWCVT_F_XU",
+                           isSEWAware=1>;
+defm : VPatConversionWF_VI<"int_riscv_vfwcvt_f_x_v", "PseudoVFWCVT_F_X",
+                           isSEWAware=1>;
+defm : VPatConversionWF_VF<"int_riscv_vfwcvt_f_f_v", "PseudoVFWCVT_F_F",
+                           isSEWAware=1>;
 defm : VPatConversionWF_VF_BF<"int_riscv_vfwcvtbf16_f_f_v",
-                              "PseudoVFWCVTBF16_F_F">;
+                              "PseudoVFWCVTBF16_F_F", isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.19. Narrowing Floating-Point/Integer Type-Convert Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 788db6f447a18..23852414cf56b 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -441,7 +441,7 @@ multiclass VPatWConvertI2FPSDNode_V<SDPatternOperator vop,
     let Predicates = !listconcat(GetVTypePredicates<ivti>.Predicates,
                                  GetVTypePredicates<fwti>.Predicates) in
     def : Pat<(fwti.Vector (vop (ivti.Vector ivti.RegClass:$rs1))),
-              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX)
+              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX#"_E"#ivti.SEW)
                   (fwti.Vector (IMPLICIT_DEF)),
                   ivti.RegClass:$rs1,
                   ivti.AVL, ivti.Log2SEW, TA_MA)>;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index 2b56eea29cfcc..eb46d47079b23 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1315,7 +1315,7 @@ multiclass VPatWConvertI2FPVL_V<SDPatternOperator vop,
     def : Pat<(fwti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
                                 (ivti.Mask V0),
                                 VLOpFrag)),
-              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX#"_MASK")
+              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX#"_E"#ivti.SEW#"_MASK")
                   (fwti.Vector (IMPLICIT_DEF)), ivti.RegClass:$rs1,
                   (ivti.Mask V0),
                   GPR:$vl, ivti.Log2SEW, TA_MA)>;
@@ -2672,7 +2672,7 @@ foreach fvtiToFWti = AllWidenableFloatVectors in {
                              (fvti.Vector fvti.RegClass:$rs1),
                              (fvti.Mask V0),
                              VLOpFrag)),
-            (!cast<Instruction>("PseudoVFWCVT_F_F_V_"#fvti.LMul.MX#"_MASK")
+            (!cast<Instruction>("PseudoVFWCVT_F_F_V_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                 (fwti.Vector (IMPLICIT_DEF)), fvti.RegClass:$rs1,
                 (fvti.Mask V0),
                 GPR:$vl, fvti.Log2SEW, TA_MA)>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 354fe7c27e68d..f1246749ac433 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -784,10 +784,11 @@ foreach mx = SchedMxListF in {
 
 // Widening
 foreach mx = SchedMxListW in {
-  defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
-  defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxListW>.c;
-  let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFWCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+  foreach sew = SchedSEWSet<mx, isF=0, isWidening=1>.val in {
+    defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
+    defvar IsWorstCase = SiFive7IsWorstCaseMXSEW<mx, sew, SchedMxListW>.c;
+    let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in
+    defm "" : LMULSEWWriteResMXSEW<"WriteVFWCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
   }
 }
 foreach mx = SchedMxListFW in {
@@ -801,16 +802,13 @@ foreach mx = SchedMxListFW in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
   }
-}
-foreach mx = SchedMxListFW in {
   defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
   defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxListFW>.c;
-  let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-  }
+  let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in
+  defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
 }
 // Narrowing
 foreach mx = SchedMxListW in {
@@ -1181,9 +1179,9 @@ defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
-defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 863e245c32256..0884cea396932 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -526,19 +526,18 @@ foreach mx = SchedMxList in {
 
 // Widening
 foreach mx = SchedMxListW in {
-  defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
-  defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxListW>.c;
-  let Latency = 3, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFWCvtIToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
+  foreach sew = SchedSEWSet<mx, isF=0, isWidening=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListW>.c;
+    let Latency = 3, ReleaseAtCycles = [LMulLat] in
+    defm "" : LMULSEWWriteResMXSEW<"WriteVFWCvtIToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
   }
 }
 foreach mx = SchedMxListFW in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxListFW>.c;
-  let Latency = 6, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFWCvtFToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-  }
+  let Latency = 6, ReleaseAtCycles = [LMulLat] in
+  defm "" : LMULWriteResMX<"WriteVFWCvtFToIV", [SiFiveP600VectorArith], mx, IsWorstCase>;
 }
 foreach mx = SchedMxListFW in {
   foreach sew = SchedSEWSet<mx, isF=1, isWidening=1>.val in {
@@ -551,6 +550,7 @@ foreach mx = SchedMxListFW in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFWMulAddF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFWCvtFToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
     }
   }
 }
@@ -962,9 +962,9 @@ defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
-defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index e7acc0c87d704..42c6dde89eac3 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -253,6 +253,18 @@ multiclass LMULReadAdvanceFW<string name, int val, list<SchedWrite> writes = []>
   : LMULReadAdvanceImpl<name, val, writes>;
 class LMULSchedWriteListFW<list<string> names> : LMULSchedWriteListImpl<names, SchedMxListFW>;
 
+multiclass LMULSEWSchedWritesW<string name>
+    : LMULSEWSchedWritesImpl<name, SchedMxListW, isF = 0, isWidening = 1>;
+multiclass LMULSEWSchedReadsW<string name>
+    : LMULSEWSchedReadsImpl<name, SchedMxListW, isF = 0, isWidening = 1>;
+multiclass LMULSEWWriteResW<string name, list<ProcResourceKind> resources>
+    : LMULSEWWriteResImpl<name, resources, SchedMxListW, isF = 0,
+                          isWidening = 1>;
+multiclass
+    LMULSEWReadAdvanceW<string name, int val, list<SchedWrite> writes = []>
+    : LMULSEWReadAdvanceImpl<name, val, writes, SchedMxListW, isF = 0,
+                             isWidening = 1>;
+
 multiclass LMULSEWSchedWritesFW<string name>
     : LMULSEWSchedWritesImpl<name, SchedMxListFW, isF = 1, isWidening = 1>;
 multiclass LMULSEWSchedReadsFW<string name>
@@ -452,9 +464,9 @@ defm "" : LMULSchedWrites<"WriteVFMovV">;
 defm "" : LMULSEWSchedWritesF<"WriteVFCvtIToFV">;
 defm "" : LMULSchedWrites<"WriteVFCvtFToIV">;
 // 13.18. Widening Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedWritesW<"WriteVFWCvtIToFV">;
+defm "" : LMULSEWSchedWritesW<"WriteVFWCvtIToFV">;
 defm "" : LMULSchedWritesFW<"WriteVFWCvtFToIV">;
-defm "" : LMULSchedWritesFW<"WriteVFWCvtFToFV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFWCvtFToFV">;
 // 13.19. Narrowing Floating-Point/Integer Type-Convert Instructions
 defm "" : LMULSchedWritesFW<"WriteVFNCvtIToFV">;
 defm "" : LMULSchedWritesW<"WriteVFNCvtFToIV">;
@@ -678,9 +690,9 @@ defm "" : LMULSchedReads<"ReadVFMovF">;
 defm "" : LMULSEWSchedReadsF<"ReadVFCvtIToFV">;
 defm "" : LMULSchedReads<"ReadVFCvtFToIV">;
 // 13.18. Widening Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedReadsW<"ReadVFWCvtIToFV">;
+defm "" : LMULSEWSchedReadsW<"ReadVFWCvtIToFV">;
 defm "" : LMULSchedReadsFW<"ReadVFWCvtFToIV">;
-defm "" : LMULSchedReadsFW<"ReadVFWCvtFToFV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFWCvtFToFV">;
 // 13.19. Narrowing Floating-Point/Integer Type-Convert Instructions
 defm "" : LMULSchedReadsFW<"ReadVFNCvtIToFV">;
 defm "" : LMULSchedReadsW<"ReadVFNCvtFToIV">;
@@ -907,9 +919,9 @@ defm "" : LMULWriteRes<"WriteVFMergeV", []>;
 defm "" : LMULWriteRes<"WriteVFMovV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFCvtIToFV", []>;
 defm "" : LMULWriteRes<"WriteVFCvtFToIV", []>;
-defm "" : LMULWriteResW<"WriteVFWCvtIToFV", []>;
+defm "" : LMULSEWWriteResW<"WriteVFWCvtIToFV", []>;
 defm "" : LMULWriteResFW<"WriteVFWCvtFToIV", []>;
-defm "" : LMULWriteResFW<"WriteVFWCvtFToFV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFWCvtFToFV", []>;
 defm "" : LMULWriteResFW<"WriteVFNCvtIToFV", []>;
 defm "" : LMULWriteResW<"WriteVFNCvtFToIV", []>;
 defm "" : LMULWriteResFW<"WriteVFNCvtFToFV", []>;
@@ -1064,9 +1076,9 @@ defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
 defm "" : LMULReadAdvance<"ReadVFMovF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFCvtIToFV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
-defm "" : LMULReadAdvanceW<"ReadVFWCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;

From 469493f55639a5cce2fc69300222be609fcc5238 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Fri, 12 Apr 2024 12:43:41 -0700
Subject: [PATCH 426/886] [RISCV] Split narrowing convert to FP pseudos by SEW

---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 73 ++++++++++++-------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |  4 +-
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  8 +-
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    | 16 ++--
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 16 ++--
 llvm/lib/Target/RISCV/RISCVScheduleV.td       | 16 ++--
 6 files changed, 78 insertions(+), 55 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index 9906049e9aaf9..5d8b091392bea 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -3762,36 +3762,45 @@ multiclass VPseudoVNCVTI_RM_W {
 multiclass VPseudoVNCVTF_W_RM {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListFW in {
-    defm _W : VPseudoConversionRoundingMode<m.vrclass, m.wvrclass, m, constraint, TargetConstraintType=2>,
-              SchedUnary<"WriteVFNCvtIToFV", "ReadVFNCvtIToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm _W : VPseudoConversionRoundingMode<m.vrclass, m.wvrclass, m,
+                                              constraint, sew=e,
+                                              TargetConstraintType=2>,
+                SchedUnary<"WriteVFNCvtIToFV", "ReadVFNCvtIToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
 multiclass VPseudoVNCVTF_RM_W {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListFW in {
-    defm _W : VPseudoConversionRM<m.vrclass, m.wvrclass, m, constraint>,
-              SchedUnary<"WriteVFNCvtIToFV", "ReadVFNCvtIToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm _W : VPseudoConversionRM<m.vrclass, m.wvrclass, m, constraint, sew=e>,
+                SchedUnary<"WriteVFNCvtIToFV", "ReadVFNCvtIToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
 multiclass VPseudoVNCVTD_W {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListFW in {
-    defm _W : VPseudoConversion<m.vrclass, m.wvrclass, m, constraint, TargetConstraintType=2>,
-              SchedUnary<"WriteVFNCvtFToFV", "ReadVFNCvtFToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm _W : VPseudoConversion<m.vrclass, m.wvrclass, m, constraint, sew=e,
+                                  TargetConstraintType=2>,
+                SchedUnary<"WriteVFNCvtFToFV", "ReadVFNCvtFToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
 multiclass VPseudoVNCVTD_W_RM {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxListFW in {
-    defm _W : VPseudoConversionRoundingMode<m.vrclass, m.wvrclass, m, constraint, TargetConstraintType=2>,
-              SchedUnary<"WriteVFNCvtFToFV", "ReadVFNCvtFToFV", m.MX,
-                         forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1, isWidening=1>.val in
+      defm _W : VPseudoConversionRoundingMode<m.vrclass, m.wvrclass, m,
+                                              constraint, sew=e,
+                                              TargetConstraintType=2>,
+                SchedUnary<"WriteVFNCvtFToFV", "ReadVFNCvtFToFV", m.MX, e,
+                           forceMergeOpRead=true>;
   }
 }
 
@@ -6027,7 +6036,8 @@ multiclass VPatConversionVI_WF_RM <string intrinsic, string instruction> {
   }
 }
 
-multiclass VPatConversionVF_WI_RM <string intrinsic, string instruction> {
+multiclass VPatConversionVF_WI_RM <string intrinsic, string instruction,
+                                   bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar iwti = GetIntVTypeInfo<fvtiToFWti.Wti>.Vti;
@@ -6035,11 +6045,13 @@ multiclass VPatConversionVF_WI_RM <string intrinsic, string instruction> {
                                  GetVTypePredicates<iwti>.Predicates) in
     defm : VPatConversionTARoundingMode<intrinsic, instruction, "W",
                                         fvti.Vector, iwti.Vector, fvti.Mask, fvti.Log2SEW,
-                                        fvti.LMul, fvti.RegClass, iwti.RegClass>;
+                                        fvti.LMul, fvti.RegClass, iwti.RegClass,
+                                        isSEWAware>;
   }
 }
 
-multiclass VPatConversionVF_WF <string intrinsic, string instruction> {
+multiclass VPatConversionVF_WF<string intrinsic, string instruction,
+                               bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
@@ -6047,12 +6059,13 @@ multiclass VPatConversionVF_WF <string intrinsic, string instruction> {
                                  GetVTypePredicates<fwti>.Predicates) in
     defm : VPatConversionTA<intrinsic, instruction, "W",
                             fvti.Vector, fwti.Vector, fvti.Mask, fvti.Log2SEW,
-                            fvti.LMul, fvti.RegClass, fwti.RegClass>;
+                            fvti.LMul, fvti.RegClass, fwti.RegClass, isSEWAware>;
   }
 }
 
-multiclass VPatConversionVF_WF_RM <string intrinsic, string instruction,
-                                   list<VTypeInfoToWide> wlist = AllWidenableFloatVectors> {
+multiclass VPatConversionVF_WF_RM<string intrinsic, string instruction,
+                                   list<VTypeInfoToWide> wlist = AllWidenableFloatVectors,
+                                   bit isSEWAware = 0> {
   foreach fvtiToFWti = wlist in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
@@ -6060,11 +6073,13 @@ multiclass VPatConversionVF_WF_RM <string intrinsic, string instruction,
                                  GetVTypePredicates<fwti>.Predicates) in
     defm : VPatConversionTARoundingMode<intrinsic, instruction, "W",
                                         fvti.Vector, fwti.Vector, fvti.Mask, fvti.Log2SEW,
-                                        fvti.LMul, fvti.RegClass, fwti.RegClass>;
+                                        fvti.LMul, fvti.RegClass, fwti.RegClass,
+                                        isSEWAware>;
   }
 }
 
-multiclass VPatConversionVF_WF_BF_RM <string intrinsic, string instruction> {
+multiclass VPatConversionVF_WF_BF_RM <string intrinsic, string instruction,
+                                      bit isSEWAware = 0> {
   foreach fvtiToFWti = AllWidenableBFloatToFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
@@ -6072,7 +6087,8 @@ multiclass VPatConversionVF_WF_BF_RM <string intrinsic, string instruction> {
                                  GetVTypePredicates<fwti>.Predicates) in
     defm : VPatConversionTARoundingMode<intrinsic, instruction, "W",
                                         fvti.Vector, fwti.Vector, fvti.Mask, fvti.Log2SEW,
-                                        fvti.LMul, fvti.RegClass, fwti.RegClass>;
+                                        fvti.LMul, fvti.RegClass, fwti.RegClass,
+                                        isSEWAware>;
   }
 }
 
@@ -7320,21 +7336,24 @@ defm : VPatConversionVI_WF_RM<"int_riscv_vfncvt_xu_f_w", "PseudoVFNCVT_XU_F">;
 defm : VPatConversionVI_WF_RM<"int_riscv_vfncvt_x_f_w", "PseudoVFNCVT_X_F">;
 defm : VPatConversionVI_WF<"int_riscv_vfncvt_rtz_xu_f_w", "PseudoVFNCVT_RTZ_XU_F">;
 defm : VPatConversionVI_WF<"int_riscv_vfncvt_rtz_x_f_w", "PseudoVFNCVT_RTZ_X_F">;
-defm : VPatConversionVF_WI_RM <"int_riscv_vfncvt_f_xu_w", "PseudoVFNCVT_F_XU">;
-defm : VPatConversionVF_WI_RM <"int_riscv_vfncvt_f_x_w", "PseudoVFNCVT_F_X">;
+defm : VPatConversionVF_WI_RM<"int_riscv_vfncvt_f_xu_w", "PseudoVFNCVT_F_XU",
+                              isSEWAware=1>;
+defm : VPatConversionVF_WI_RM<"int_riscv_vfncvt_f_x_w", "PseudoVFNCVT_F_X",
+                              isSEWAware=1>;
 defvar WidenableFloatVectorsExceptF16 = !filter(fvtiToFWti, AllWidenableFloatVectors,
                                                 !ne(fvtiToFWti.Vti.Scalar, f16));
 defm : VPatConversionVF_WF_RM<"int_riscv_vfncvt_f_f_w", "PseudoVFNCVT_F_F",
-                           WidenableFloatVectorsExceptF16>;
+                           WidenableFloatVectorsExceptF16, isSEWAware=1>;
 // Define vfncvt.f.f.w for f16 when Zvfhmin is enable.
 defvar F16WidenableFloatVectors = !filter(fvtiToFWti, AllWidenableFloatVectors,
                                           !eq(fvtiToFWti.Vti.Scalar, f16));
 let Predicates = [HasVInstructionsF16Minimal] in
 defm : VPatConversionVF_WF_RM<"int_riscv_vfncvt_f_f_w", "PseudoVFNCVT_F_F",
-                           F16WidenableFloatVectors>;
+                           F16WidenableFloatVectors, isSEWAware=1>;
 defm : VPatConversionVF_WF_BF_RM<"int_riscv_vfncvtbf16_f_f_w", 
-                                 "PseudoVFNCVTBF16_F_F">;
-defm : VPatConversionVF_WF<"int_riscv_vfncvt_rod_f_f_w", "PseudoVFNCVT_ROD_F_F">;
+                                 "PseudoVFNCVTBF16_F_F", isSEWAware=1>;
+defm : VPatConversionVF_WF<"int_riscv_vfncvt_rod_f_f_w", "PseudoVFNCVT_ROD_F_F",
+                           isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 14. Vector Reduction Operations
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 23852414cf56b..9b324c40cc5be 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -470,7 +470,7 @@ multiclass VPatNConvertI2FPSDNode_W_RM<SDPatternOperator vop,
     let Predicates = !listconcat(GetVTypePredicates<fvti>.Predicates,
                                  GetVTypePredicates<iwti>.Predicates) in
     def : Pat<(fvti.Vector (vop (iwti.Vector iwti.RegClass:$rs1))),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX)
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW)
                   (fvti.Vector (IMPLICIT_DEF)),
                   iwti.RegClass:$rs1,
                   // Value to indicate no rounding mode change in
@@ -1445,7 +1445,7 @@ foreach fvtiToFWti = AllWidenableFloatVectors in {
                        !listconcat(GetVTypePredicates<fvti>.Predicates,
                                    GetVTypePredicates<fwti>.Predicates)) in
   def : Pat<(fvti.Vector (fpround (fwti.Vector fwti.RegClass:$rs1))),
-            (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX)
+            (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX#"_E"#fvti.SEW)
                 (fvti.Vector (IMPLICIT_DEF)),
                 fwti.RegClass:$rs1,
                 // Value to indicate no rounding mode change in
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index eb46d47079b23..8e143b74a6663 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -1389,7 +1389,7 @@ multiclass VPatNConvertI2FPVL_W_RM<SDPatternOperator vop,
     def : Pat<(fvti.Vector (vop (iwti.Vector iwti.RegClass:$rs1),
                                 (iwti.Mask V0),
                                 VLOpFrag)),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK") 
                   (fvti.Vector (IMPLICIT_DEF)), iwti.RegClass:$rs1,
                   (iwti.Mask V0),
                   // Value to indicate no rounding mode change in
@@ -1408,7 +1408,7 @@ multiclass VPatNConvertI2FP_RM_VL_W<SDNode vop, string instruction_name> {
     def : Pat<(fvti.Vector (vop (iwti.Vector iwti.RegClass:$rs1),
                                 (iwti.Mask V0),  (XLenVT timm:$frm),
                                 VLOpFrag)),
-              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                   (fvti.Vector (IMPLICIT_DEF)), iwti.RegClass:$rs1,
                   (iwti.Mask V0), timm:$frm, GPR:$vl, fvti.Log2SEW, TA_MA)>;
   }
@@ -2703,7 +2703,7 @@ foreach fvtiToFWti = AllWidenableFloatVectors in {
     def : Pat<(fvti.Vector (any_riscv_fpround_vl
                                (fwti.Vector fwti.RegClass:$rs1),
                                (fwti.Mask V0), VLOpFrag)),
-              (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                   (fvti.Vector (IMPLICIT_DEF)), fwti.RegClass:$rs1,
                   (fwti.Mask V0),
                   // Value to indicate no rounding mode change in
@@ -2716,7 +2716,7 @@ foreach fvtiToFWti = AllWidenableFloatVectors in {
     def : Pat<(fvti.Vector (any_riscv_fncvt_rod_vl
                                (fwti.Vector fwti.RegClass:$rs1),
                                (fwti.Mask V0), VLOpFrag)),
-              (!cast<Instruction>("PseudoVFNCVT_ROD_F_F_W_"#fvti.LMul.MX#"_MASK")
+              (!cast<Instruction>("PseudoVFNCVT_ROD_F_F_W_"#fvti.LMul.MX#"_E"#fvti.SEW#"_MASK")
                   (fvti.Vector (IMPLICIT_DEF)), fwti.RegClass:$rs1,
                   (fwti.Mask V0), GPR:$vl, fvti.Log2SEW, TA_MA)>;
   }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index f1246749ac433..feeefa7c42c49 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -819,11 +819,13 @@ foreach mx = SchedMxListW in {
   }
 }
 foreach mx = SchedMxListFW in {
-  defvar Cycles = SiFive7GetCyclesNarrowing<mx>.c;
-  defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxListFW>.c;
-  let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFNCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFNCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+  foreach sew = SchedSEWSet<mx, isF=1, isWidening=1>.val in {
+    defvar Cycles = SiFive7GetCyclesNarrowing<mx>.c;
+    defvar IsWorstCase = SiFive7IsWorstCaseMXSEW<mx, sew, SchedMxListFW, isF=1>.c;
+    let Latency = 8, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFNCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFNCvtFToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+    }
   }
 }
 
@@ -1182,9 +1184,9 @@ defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
 
 // 15. Vector Reduction Operations
 def : ReadAdvance<ReadVIRedV, 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 0884cea396932..9dda9c500bd04 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -563,11 +563,13 @@ foreach mx = SchedMxListW in {
   }
 }
 foreach mx = SchedMxListFW in {
-  defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
-  defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxListFW>.c;
-  let Latency = 3, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFNCvtIToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFNCvtFToFV", [SiFiveP600VectorArith], mx, IsWorstCase>;
+  foreach sew = SchedSEWSet<mx, isF=1, isWidening=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxListFW, isF=1>.c;
+    let Latency = 3, ReleaseAtCycles = [LMulLat] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFNCvtIToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFNCvtFToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    }
   }
 }
 
@@ -965,9 +967,9 @@ defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
 
 // 15. Vector Reduction Operations
 def : ReadAdvance<ReadVIRedV, 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 42c6dde89eac3..14a4e4cbc59f3 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -468,9 +468,9 @@ defm "" : LMULSEWSchedWritesW<"WriteVFWCvtIToFV">;
 defm "" : LMULSchedWritesFW<"WriteVFWCvtFToIV">;
 defm "" : LMULSEWSchedWritesFW<"WriteVFWCvtFToFV">;
 // 13.19. Narrowing Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedWritesFW<"WriteVFNCvtIToFV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFNCvtIToFV">;
 defm "" : LMULSchedWritesW<"WriteVFNCvtFToIV">;
-defm "" : LMULSchedWritesFW<"WriteVFNCvtFToFV">;
+defm "" : LMULSEWSchedWritesFW<"WriteVFNCvtFToFV">;
 
 // 14. Vector Reduction Operations
 // The latency of reduction is determined by the size of the read resource.
@@ -694,9 +694,9 @@ defm "" : LMULSEWSchedReadsW<"ReadVFWCvtIToFV">;
 defm "" : LMULSchedReadsFW<"ReadVFWCvtFToIV">;
 defm "" : LMULSEWSchedReadsFW<"ReadVFWCvtFToFV">;
 // 13.19. Narrowing Floating-Point/Integer Type-Convert Instructions
-defm "" : LMULSchedReadsFW<"ReadVFNCvtIToFV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFNCvtIToFV">;
 defm "" : LMULSchedReadsW<"ReadVFNCvtFToIV">;
-defm "" : LMULSchedReadsFW<"ReadVFNCvtFToFV">;
+defm "" : LMULSEWSchedReadsFW<"ReadVFNCvtFToFV">;
 
 // 14. Vector Reduction Operations
 // 14.1. Vector Single-Width Integer Reduction Instructions
@@ -922,9 +922,9 @@ defm "" : LMULWriteRes<"WriteVFCvtFToIV", []>;
 defm "" : LMULSEWWriteResW<"WriteVFWCvtIToFV", []>;
 defm "" : LMULWriteResFW<"WriteVFWCvtFToIV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWCvtFToFV", []>;
-defm "" : LMULWriteResFW<"WriteVFNCvtIToFV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFNCvtIToFV", []>;
 defm "" : LMULWriteResW<"WriteVFNCvtFToIV", []>;
-defm "" : LMULWriteResFW<"WriteVFNCvtFToFV", []>;
+defm "" : LMULSEWWriteResFW<"WriteVFNCvtFToFV", []>;
 
 // 14. Vector Reduction Operations
 defm "" : LMULSEWWriteRes<"WriteVIRedV_From", []>;
@@ -1079,9 +1079,9 @@ defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"SEWReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
-defm "" : LMULReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
 
 // 14. Vector Reduction Operations
 def : ReadAdvance<ReadVIRedV, 0>;

From 1b310c45e914f0a76fa594bebdff44e7a386d4f9 Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Wed, 10 Apr 2024 12:38:06 -0700
Subject: [PATCH 427/886] [RISCV] Split PseudoVFMIN, PseudoVFMAX PseudoVFSGNJ,
 PseudoVFSGNJN, and PseudoVFSGNJX by SEW

---
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 37 +++++++++++--------
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td | 16 ++++----
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td | 14 +++----
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    | 18 +++++----
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 24 ++++++++----
 llvm/lib/Target/RISCV/RISCVScheduleV.td       | 32 ++++++++--------
 6 files changed, 79 insertions(+), 62 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index 5d8b091392bea..ad1821d57256b 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2894,32 +2894,34 @@ multiclass VPseudoVALU_VV_VX {
 
 multiclass VPseudoVSGNJ_VV_VF {
   foreach m = MxListF in {
-    defm "" : VPseudoBinaryFV_VV<m>,
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+    defm "" : VPseudoBinaryFV_VV<m, sew=e>,
               SchedBinary<"WriteVFSgnjV", "ReadVFSgnjV", "ReadVFSgnjV", m.MX,
-                          forceMergeOpRead=true>;
+                          e, forceMergeOpRead=true>;
   }
 
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoBinaryV_VF<m, f>,
+      defm "" : VPseudoBinaryV_VF<m, f, sew=f.SEW>,
                 SchedBinary<"WriteVFSgnjF", "ReadVFSgnjV", "ReadVFSgnjF", m.MX,
-                            forceMergeOpRead=true>;
+                            f.SEW, forceMergeOpRead=true>;
     }
   }
 }
 
 multiclass VPseudoVMAX_VV_VF {
   foreach m = MxListF in {
-    defm "" : VPseudoBinaryFV_VV<m>,
-              SchedBinary<"WriteVFMinMaxV", "ReadVFMinMaxV", "ReadVFMinMaxV", m.MX,
-                          forceMergeOpRead=true>;
+    foreach e = SchedSEWSet<m.MX, isF=1>.val in
+      defm "" : VPseudoBinaryFV_VV<m, sew=e>,
+                SchedBinary<"WriteVFMinMaxV", "ReadVFMinMaxV", "ReadVFMinMaxV",
+                            m.MX, e, forceMergeOpRead=true>;
   }
 
   foreach f = FPList in {
     foreach m = f.MxList in {
-      defm "" : VPseudoBinaryV_VF<m, f>,
-                SchedBinary<"WriteVFMinMaxF", "ReadVFMinMaxV", "ReadVFMinMaxF", m.MX,
-                            forceMergeOpRead=true>;
+      defm "" : VPseudoBinaryV_VF<m, f, sew=f.SEW>,
+                SchedBinary<"WriteVFMinMaxF", "ReadVFMinMaxV", "ReadVFMinMaxF",
+                            m.MX, f.SEW, forceMergeOpRead=true>;
     }
   }
 }
@@ -7236,15 +7238,20 @@ defm : VPatUnaryV_V_RM<"int_riscv_vfrec7", "PseudoVFREC7", AllFloatVectors, isSE
 //===----------------------------------------------------------------------===//
 // 13.11. Vector Floating-Point Min/Max Instructions
 //===----------------------------------------------------------------------===//
-defm : VPatBinaryV_VV_VX<"int_riscv_vfmin", "PseudoVFMIN", AllFloatVectors>;
-defm : VPatBinaryV_VV_VX<"int_riscv_vfmax", "PseudoVFMAX", AllFloatVectors>;
+defm : VPatBinaryV_VV_VX<"int_riscv_vfmin", "PseudoVFMIN", AllFloatVectors,
+                         isSEWAware=1>;
+defm : VPatBinaryV_VV_VX<"int_riscv_vfmax", "PseudoVFMAX", AllFloatVectors,
+                         isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.12. Vector Floating-Point Sign-Injection Instructions
 //===----------------------------------------------------------------------===//
-defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnj", "PseudoVFSGNJ", AllFloatVectors>;
-defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnjn", "PseudoVFSGNJN", AllFloatVectors>;
-defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnjx", "PseudoVFSGNJX", AllFloatVectors>;
+defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnj", "PseudoVFSGNJ", AllFloatVectors,
+                         isSEWAware=1>;
+defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnjn", "PseudoVFSGNJN", AllFloatVectors,
+                         isSEWAware=1>;
+defm : VPatBinaryV_VV_VX<"int_riscv_vfsgnjx", "PseudoVFSGNJX", AllFloatVectors,
+                         isSEWAware=1>;
 
 //===----------------------------------------------------------------------===//
 // 13.13. Vector Floating-Point Compare Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 9b324c40cc5be..82ee4b0cbce90 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -1339,42 +1339,42 @@ foreach vti = AllFloatVectors in {
 
     // 13.12. Vector Floating-Point Sign-Injection Instructions
     def : Pat<(fabs (vti.Vector vti.RegClass:$rs)),
-              (!cast<Instruction>("PseudoVFSGNJX_VV_"# vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJX_VV_"# vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs, vti.RegClass:$rs, vti.AVL, vti.Log2SEW, TA_MA)>;
     // Handle fneg with VFSGNJN using the same input for both operands.
     def : Pat<(fneg (vti.Vector vti.RegClass:$rs)),
-              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs, vti.RegClass:$rs, vti.AVL, vti.Log2SEW, TA_MA)>;
 
     def : Pat<(vti.Vector (fcopysign (vti.Vector vti.RegClass:$rs1),
                                      (vti.Vector vti.RegClass:$rs2))),
-              (!cast<Instruction>("PseudoVFSGNJ_VV_"# vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJ_VV_"# vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs1, vti.RegClass:$rs2, vti.AVL, vti.Log2SEW, TA_MA)>;
     def : Pat<(vti.Vector (fcopysign (vti.Vector vti.RegClass:$rs1),
                                      (vti.Vector (SplatFPOp vti.ScalarRegClass:$rs2)))),
-              (!cast<Instruction>("PseudoVFSGNJ_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJ_V"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs1, vti.ScalarRegClass:$rs2, vti.AVL, vti.Log2SEW, TA_MA)>;
 
     def : Pat<(vti.Vector (fcopysign (vti.Vector vti.RegClass:$rs1),
                                      (vti.Vector (fneg vti.RegClass:$rs2)))),
-              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs1, vti.RegClass:$rs2, vti.AVL, vti.Log2SEW, TA_MA)>;
     def : Pat<(vti.Vector (fcopysign (vti.Vector vti.RegClass:$rs1),
                                      (vti.Vector (fneg (SplatFPOp vti.ScalarRegClass:$rs2))))),
-              (!cast<Instruction>("PseudoVFSGNJN_V"#vti.ScalarSuffix#"_"#vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJN_V"#vti.ScalarSuffix#"_"#vti.LMul.MX#"_E"#vti.SEW)
                    (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs1, vti.ScalarRegClass:$rs2, vti.AVL, vti.Log2SEW, TA_MA)>;
   }
 }
 
 // 13.11. Vector Floating-Point MIN/MAX Instructions
-defm : VPatBinaryFPSDNode_VV_VF<fminnum, "PseudoVFMIN">;
-defm : VPatBinaryFPSDNode_VV_VF<fmaxnum, "PseudoVFMAX">;
+defm : VPatBinaryFPSDNode_VV_VF<fminnum, "PseudoVFMIN", isSEWAware=1>;
+defm : VPatBinaryFPSDNode_VV_VF<fmaxnum, "PseudoVFMAX", isSEWAware=1>;
 
 // 13.13. Vector Floating-Point Compare Instructions
 defm : VPatFPSetCCSDNode_VV_VF_FV<SETEQ,  "PseudoVMFEQ", "PseudoVMFEQ">;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index 8e143b74a6663..f2d97ba147254 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -2468,8 +2468,8 @@ defm : VPatWidenFPMulAccVL_VV_VF_RM<riscv_vfwmsub_vl, "PseudoVFWMSAC">;
 defm : VPatWidenFPMulAccVL_VV_VF_RM<riscv_vfwnmsub_vl, "PseudoVFWNMSAC">;
 
 // 13.11. Vector Floating-Point MIN/MAX Instructions
-defm : VPatBinaryFPVL_VV_VF<riscv_vfmin_vl, "PseudoVFMIN">;
-defm : VPatBinaryFPVL_VV_VF<riscv_vfmax_vl, "PseudoVFMAX">;
+defm : VPatBinaryFPVL_VV_VF<riscv_vfmin_vl, "PseudoVFMIN", isSEWAware=1>;
+defm : VPatBinaryFPVL_VV_VF<riscv_vfmax_vl, "PseudoVFMAX", isSEWAware=1>;
 
 // 13.13. Vector Floating-Point Compare Instructions
 defm : VPatFPSetCCVL_VV_VF_FV<any_riscv_fsetcc_vl, SETEQ,
@@ -2505,14 +2505,14 @@ foreach vti = AllFloatVectors in {
     // 13.12. Vector Floating-Point Sign-Injection Instructions
     def : Pat<(riscv_fabs_vl (vti.Vector vti.RegClass:$rs), (vti.Mask V0),
                              VLOpFrag),
-              (!cast<Instruction>("PseudoVFSGNJX_VV_"# vti.LMul.MX #"_MASK")
+              (!cast<Instruction>("PseudoVFSGNJX_VV_"# vti.LMul.MX #"_E"#vti.SEW#"_MASK")
                    (vti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs,
                    vti.RegClass:$rs, (vti.Mask V0), GPR:$vl, vti.Log2SEW,
                    TA_MA)>;
     // Handle fneg with VFSGNJN using the same input for both operands.
     def : Pat<(riscv_fneg_vl (vti.Vector vti.RegClass:$rs), (vti.Mask V0),
                              VLOpFrag),
-              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX #"_MASK")
+              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX#"_E"#vti.SEW #"_MASK")
                    (vti.Vector (IMPLICIT_DEF)), vti.RegClass:$rs,
                    vti.RegClass:$rs, (vti.Mask V0), GPR:$vl, vti.Log2SEW,
                    TA_MA)>;
@@ -2522,7 +2522,7 @@ foreach vti = AllFloatVectors in {
                                   vti.RegClass:$merge,
                                   (vti.Mask V0),
                                   VLOpFrag),
-              (!cast<Instruction>("PseudoVFSGNJ_VV_"# vti.LMul.MX#"_MASK")
+              (!cast<Instruction>("PseudoVFSGNJ_VV_"# vti.LMul.MX#"_E"#vti.SEW#"_MASK")
                    vti.RegClass:$merge, vti.RegClass:$rs1,
                    vti.RegClass:$rs2, (vti.Mask V0), GPR:$vl, vti.Log2SEW,
                    TAIL_AGNOSTIC)>;
@@ -2534,7 +2534,7 @@ foreach vti = AllFloatVectors in {
                                   srcvalue,
                                   (vti.Mask true_mask),
                                   VLOpFrag),
-              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX)
+              (!cast<Instruction>("PseudoVFSGNJN_VV_"# vti.LMul.MX#"_E"#vti.SEW)
         (vti.Vector (IMPLICIT_DEF)),
                    vti.RegClass:$rs1, vti.RegClass:$rs2, GPR:$vl, vti.Log2SEW, TA_MA)>;
 
@@ -2543,7 +2543,7 @@ foreach vti = AllFloatVectors in {
                                   vti.RegClass:$merge,
                                   (vti.Mask V0),
                                   VLOpFrag),
-              (!cast<Instruction>("PseudoVFSGNJ_V"#vti.ScalarSuffix#"_"# vti.LMul.MX#"_MASK")
+              (!cast<Instruction>("PseudoVFSGNJ_V"#vti.ScalarSuffix#"_"# vti.LMul.MX#"_E"#vti.SEW#"_MASK")
                    vti.RegClass:$merge, vti.RegClass:$rs1,
                    vti.ScalarRegClass:$rs2, (vti.Mask V0), GPR:$vl, vti.Log2SEW,
                    TAIL_AGNOSTIC)>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index feeefa7c42c49..4dcec96df9d57 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -746,6 +746,12 @@ foreach mx = SchedMxListF in {
       defm "" : LMULSEWWriteResMXSEW<"WriteVFRecpV",   [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
       defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
     }
+    let Latency = 4, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxV", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxF", [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjV",   [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjF",   [SiFive7VCQ, SiFive7VA], mx, sew, IsWorstCase>;
+    }
   }
 }
 foreach mx = SchedMxList in {
@@ -755,10 +761,6 @@ foreach mx = SchedMxList in {
     defm "" : LMULWriteResMX<"WriteVFCvtFToIV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
   }
   let Latency = 4, AcquireAtCycles = [0, 1], ReleaseAtCycles = [1, !add(1, Cycles)] in {
-    defm "" : LMULWriteResMX<"WriteVFSgnjV",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFSgnjF",     [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMinMaxV",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMinMaxF",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFClassV",    [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMergeV",    [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMovV",      [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
@@ -1169,10 +1171,10 @@ defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFRecpV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxF", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjV", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjF", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpF", 0>;
 defm "" : LMULReadAdvance<"ReadVFClassV", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 9dda9c500bd04..0ced2efa3f7ab 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -504,6 +504,18 @@ foreach mx = SchedMxListF in {
     defm "" : LMULSEWWriteResMXSEW<"WriteVFCvtIToFV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
   }
 }
+foreach mx = SchedMxListF in {
+  foreach sew = SchedSEWSet<mx, isF=1>.val in {
+    defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
+    defvar IsWorstCase = SiFiveP600IsWorstCaseMXSEW<mx, sew, SchedMxList, isF=1>.c;
+    let Latency = 1, ReleaseAtCycles = [LMulLat] in {
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxV", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFMinMaxF", [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjV",   [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+      defm "" : LMULSEWWriteResMXSEW<"WriteVFSgnjF",   [SiFiveP600VectorArith], mx, sew, IsWorstCase>;
+    }
+  }
+}
 foreach mx = SchedMxList in {
   defvar LMulLat = SiFiveP600GetLMulCycles<mx>.c;
   defvar IsWorstCase = SiFiveP600IsWorstCaseMX<mx, SchedMxList>.c;
@@ -514,10 +526,6 @@ foreach mx = SchedMxList in {
     defm "" : LMULWriteResMX<"WriteVFCmpF",  [SiFiveP600VectorArith], mx, IsWorstCase>;
   }
   let Latency = 1, ReleaseAtCycles = [LMulLat] in {
-    defm "" : LMULWriteResMX<"WriteVFSgnjV",   [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFSgnjF",   [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMinMaxV", [SiFiveP600VectorArith], mx, IsWorstCase>;
-    defm "" : LMULWriteResMX<"WriteVFMinMaxF", [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFClassV",  [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMergeV",  [SiFiveP600VectorArith], mx, IsWorstCase>;
     defm "" : LMULWriteResMX<"WriteVFMovV",    [SiFiveP600VectorArith], mx, IsWorstCase>;
@@ -954,10 +962,10 @@ defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULSEWReadAdvance<"ReadVFRecpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpF", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxF", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjV", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjF", 0>;
 defm "" : LMULReadAdvance<"ReadVFClassV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeV", 0>;
 defm "" : LMULReadAdvance<"ReadVFMergeF", 0>;
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 14a4e4cbc59f3..5993884bc2c1c 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -446,11 +446,11 @@ defm "" : LMULSEWSchedWritesF<"WriteVFSqrtV">;
 // 13.10. Vector Floating-Point Reciprocal Estimate Instruction
 defm "" : LMULSEWSchedWritesF<"WriteVFRecpV">;
 // 13.11. Vector Floating-Point MIN/MAX Instructions
-defm "" : LMULSchedWrites<"WriteVFMinMaxV">;
-defm "" : LMULSchedWrites<"WriteVFMinMaxF">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMinMaxV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFMinMaxF">;
 // 13.12. Vector Floating-Point Sign-Injection Instructions
-defm "" : LMULSchedWrites<"WriteVFSgnjV">;
-defm "" : LMULSchedWrites<"WriteVFSgnjF">;
+defm "" : LMULSEWSchedWritesF<"WriteVFSgnjV">;
+defm "" : LMULSEWSchedWritesF<"WriteVFSgnjF">;
 // 13.13. Vector Floating-Point Compare Instructions
 defm "" : LMULSchedWrites<"WriteVFCmpV">;
 defm "" : LMULSchedWrites<"WriteVFCmpF">;
@@ -671,11 +671,11 @@ defm "" : LMULSEWSchedReadsF<"ReadVFSqrtV">;
 // 13.10. Vector Floating-Point Reciprocal Estimate Instruction
 defm "" : LMULSEWSchedReadsF<"ReadVFRecpV">;
 // 13.11. Vector Floating-Point MIN/MAX Instructions
-defm "" : LMULSchedReads<"ReadVFMinMaxV">;
-defm "" : LMULSchedReads<"ReadVFMinMaxF">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMinMaxV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFMinMaxF">;
 // 13.12. Vector Floating-Point Sign-Injection Instructions
-defm "" : LMULSchedReads<"ReadVFSgnjV">;
-defm "" : LMULSchedReads<"ReadVFSgnjF">;
+defm "" : LMULSEWSchedReadsF<"ReadVFSgnjV">;
+defm "" : LMULSEWSchedReadsF<"ReadVFSgnjF">;
 // 13.13. Vector Floating-Point Compare Instructions
 defm "" : LMULSchedReads<"ReadVFCmpV">;
 defm "" : LMULSchedReads<"ReadVFCmpF">;
@@ -908,10 +908,10 @@ defm "" : LMULSEWWriteResFW<"WriteVFWMulAddV", []>;
 defm "" : LMULSEWWriteResFW<"WriteVFWMulAddF", []>;
 defm "" : LMULSEWWriteResF<"WriteVFSqrtV", []>;
 defm "" : LMULSEWWriteResF<"WriteVFRecpV", []>;
-defm "" : LMULWriteRes<"WriteVFMinMaxV", []>;
-defm "" : LMULWriteRes<"WriteVFMinMaxF", []>;
-defm "" : LMULWriteRes<"WriteVFSgnjV", []>;
-defm "" : LMULWriteRes<"WriteVFSgnjF", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMinMaxV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFMinMaxF", []>;
+defm "" : LMULSEWWriteResF<"WriteVFSgnjV", []>;
+defm "" : LMULSEWWriteResF<"WriteVFSgnjF", []>;
 defm "" : LMULWriteRes<"WriteVFCmpV", []>;
 defm "" : LMULWriteRes<"WriteVFCmpF", []>;
 defm "" : LMULWriteRes<"WriteVFClassV", []>;
@@ -1064,10 +1064,10 @@ defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWMulAddF", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFSqrtV", 0>;
 defm "" : LMULSEWReadAdvanceF<"ReadVFRecpV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxV", 0>;
-defm "" : LMULReadAdvance<"ReadVFMinMaxF", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjV", 0>;
-defm "" : LMULReadAdvance<"ReadVFSgnjF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFMinMaxF", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjV", 0>;
+defm "" : LMULSEWReadAdvanceF<"ReadVFSgnjF", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpV", 0>;
 defm "" : LMULReadAdvance<"ReadVFCmpF", 0>;
 defm "" : LMULReadAdvance<"ReadVFClassV", 0>;

From 17cb8a537b688ac6e4187f28f9e784491bb8abc5 Mon Sep 17 00:00:00 2001
From: Kiran Chandramohan <kiran.chandramohan@arm.com>
Date: Mon, 15 Apr 2024 14:24:42 +0100
Subject: [PATCH 428/886] [Flang][OpenMP] Accept the reduction modifier
 (#86492)

Accept the reduction modifier in the Flang parser. Issue a TODO message
during lowering.

OpenMP 5.0 introduced the reduction modifier. Details can be seen in
2.19.5.4 reductionClause.
OpenMP 5.2 relevant section is 5.5.8reductionClause.

This will help remove some of the parser errors highlighted in the
following post and also bring it to a well defined behaviour (produce
TODO errors for unsupported features, do not crash).
https://discourse.llvm.org/t/proposal-rename-flang-new-to-flang/69462/60
---
 flang/include/flang/Parser/dump-parse-tree.h  |  1 +
 flang/include/flang/Parser/parse-tree.h       |  5 +++-
 flang/lib/Lower/OpenMP/Clauses.cpp            | 26 +++++++++++++++----
 flang/lib/Lower/OpenMP/ReductionProcessor.cpp |  5 ++++
 flang/lib/Parser/openmp-parsers.cpp           |  5 ++++
 flang/lib/Parser/unparse.cpp                  |  4 +++
 flang/lib/Semantics/check-omp-structure.cpp   |  2 +-
 .../Lower/OpenMP/Todo/reduction-modifiers.f90 | 13 ++++++++++
 .../test/Parser/OpenMP/reduction-modifier.f90 | 20 ++++++++++++++
 llvm/include/llvm/Frontend/OpenMP/ClauseT.h   |  2 +-
 10 files changed, 75 insertions(+), 8 deletions(-)
 create mode 100644 flang/test/Lower/OpenMP/Todo/reduction-modifiers.f90
 create mode 100644 flang/test/Parser/OpenMP/reduction-modifier.f90

diff --git a/flang/include/flang/Parser/dump-parse-tree.h b/flang/include/flang/Parser/dump-parse-tree.h
index 06c168a5de612..477d391277ee2 100644
--- a/flang/include/flang/Parser/dump-parse-tree.h
+++ b/flang/include/flang/Parser/dump-parse-tree.h
@@ -542,6 +542,7 @@ class ParseTreeDumper {
   NODE_ENUM(OmpOrderModifier, Kind)
   NODE(parser, OmpProcBindClause)
   NODE_ENUM(OmpProcBindClause, Type)
+  NODE_ENUM(OmpReductionClause, ReductionModifier)
   NODE(parser, OmpReductionClause)
   NODE(parser, OmpInReductionClause)
   NODE(parser, OmpReductionCombiner)
diff --git a/flang/include/flang/Parser/parse-tree.h b/flang/include/flang/Parser/parse-tree.h
index 574a95cf22afd..d7c23755c57b2 100644
--- a/flang/include/flang/Parser/parse-tree.h
+++ b/flang/include/flang/Parser/parse-tree.h
@@ -3552,7 +3552,10 @@ struct OmpReductionOperator {
 //                                         variable-name-list)
 struct OmpReductionClause {
   TUPLE_CLASS_BOILERPLATE(OmpReductionClause);
-  std::tuple<OmpReductionOperator, OmpObjectList> t;
+  ENUM_CLASS(ReductionModifier, Inscan, Task, Default)
+  std::tuple<std::optional<ReductionModifier>, OmpReductionOperator,
+      OmpObjectList>
+      t;
 };
 
 // OMP 5.0 2.19.5.6 in_reduction-clause -> IN_REDUCTION (reduction-identifier:
diff --git a/flang/lib/Lower/OpenMP/Clauses.cpp b/flang/lib/Lower/OpenMP/Clauses.cpp
index 40da71c8b55f8..97337cfc08c72 100644
--- a/flang/lib/Lower/OpenMP/Clauses.cpp
+++ b/flang/lib/Lower/OpenMP/Clauses.cpp
@@ -1005,12 +1005,28 @@ ProcBind make(const parser::OmpClause::ProcBind &inp,
 Reduction make(const parser::OmpClause::Reduction &inp,
                semantics::SemanticsContext &semaCtx) {
   // inp.v -> parser::OmpReductionClause
-  auto &t0 = std::get<parser::OmpReductionOperator>(inp.v.t);
-  auto &t1 = std::get<parser::OmpObjectList>(inp.v.t);
+  using wrapped = parser::OmpReductionClause;
+
+  CLAUSET_ENUM_CONVERT( //
+      convert, wrapped::ReductionModifier, Reduction::ReductionModifier,
+      // clang-format off
+      MS(Inscan,  Inscan)
+      MS(Task,    Task)
+      MS(Default, Default)
+      // clang-format on
+  );
+
+  auto &t0 =
+      std::get<std::optional<parser::OmpReductionClause::ReductionModifier>>(
+          inp.v.t);
+  auto &t1 = std::get<parser::OmpReductionOperator>(inp.v.t);
+  auto &t2 = std::get<parser::OmpObjectList>(inp.v.t);
   return Reduction{
-      {/*ReductionIdentifiers=*/{makeReductionOperator(t0, semaCtx)},
-       /*ReductionModifier=*/std::nullopt,
-       /*List=*/makeObjects(t1, semaCtx)}};
+      {/*ReductionModifier=*/t0
+           ? std::make_optional<Reduction::ReductionModifier>(convert(*t0))
+           : std::nullopt,
+       /*ReductionIdentifiers=*/{makeReductionOperator(t1, semaCtx)},
+       /*List=*/makeObjects(t2, semaCtx)}};
 }
 
 // Relaxed: empty
diff --git a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
index 918edf27baf66..f42386fe2736d 100644
--- a/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ReductionProcessor.cpp
@@ -574,6 +574,11 @@ void ReductionProcessor::addDeclareReduction(
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
         *reductionSymbols) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+
+  if (std::get<std::optional<omp::clause::Reduction::ReductionModifier>>(
+          reduction.t))
+    TODO(currentLocation, "Reduction modifiers are not supported");
+
   mlir::omp::DeclareReductionOp decl;
   const auto &redOperatorList{
       std::get<omp::clause::Reduction::ReductionIdentifiers>(reduction.t)};
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index bba1be27158ce..eae4784169146 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -136,6 +136,11 @@ TYPE_PARSER(construct<OmpReductionOperator>(Parser<DefinedOperator>{}) ||
     construct<OmpReductionOperator>(Parser<ProcedureDesignator>{}))
 
 TYPE_PARSER(construct<OmpReductionClause>(
+    maybe(
+        ("INSCAN" >> pure(OmpReductionClause::ReductionModifier::Inscan) ||
+            "TASK" >> pure(OmpReductionClause::ReductionModifier::Task) ||
+            "DEFAULT" >> pure(OmpReductionClause::ReductionModifier::Default)) /
+        ","),
     Parser<OmpReductionOperator>{} / ":", Parser<OmpObjectList>{}))
 
 // OMP 5.0 2.19.5.6 IN_REDUCTION (reduction-identifier: variable-name-list)
diff --git a/flang/lib/Parser/unparse.cpp b/flang/lib/Parser/unparse.cpp
index c06458833f072..3398b395f198f 100644
--- a/flang/lib/Parser/unparse.cpp
+++ b/flang/lib/Parser/unparse.cpp
@@ -2090,6 +2090,8 @@ class UnparseVisitor {
     Walk(":", x.step);
   }
   void Unparse(const OmpReductionClause &x) {
+    Walk(std::get<std::optional<OmpReductionClause::ReductionModifier>>(x.t),
+        ",");
     Walk(std::get<OmpReductionOperator>(x.t));
     Put(":");
     Walk(std::get<OmpObjectList>(x.t));
@@ -2727,6 +2729,8 @@ class UnparseVisitor {
   WALK_NESTED_ENUM(OmpScheduleClause, ScheduleType) // OMP schedule-type
   WALK_NESTED_ENUM(OmpDeviceClause, DeviceModifier) // OMP device modifier
   WALK_NESTED_ENUM(OmpDeviceTypeClause, Type) // OMP DEVICE_TYPE
+  WALK_NESTED_ENUM(
+      OmpReductionClause, ReductionModifier) // OMP reduction-modifier
   WALK_NESTED_ENUM(OmpIfClause, DirectiveNameModifier) // OMP directive-modifier
   WALK_NESTED_ENUM(OmpCancelType, Type) // OMP cancel-type
   WALK_NESTED_ENUM(OmpOrderClause, Type) // OMP order-type
diff --git a/flang/lib/Semantics/check-omp-structure.cpp b/flang/lib/Semantics/check-omp-structure.cpp
index bf4debee1df34..e85d8d1f7ab53 100644
--- a/flang/lib/Semantics/check-omp-structure.cpp
+++ b/flang/lib/Semantics/check-omp-structure.cpp
@@ -2289,7 +2289,7 @@ void OmpStructureChecker::Enter(const parser::OmpClause::Reduction &x) {
 bool OmpStructureChecker::CheckReductionOperators(
     const parser::OmpClause::Reduction &x) {
 
-  const auto &definedOp{std::get<0>(x.v.t)};
+  const auto &definedOp{std::get<parser::OmpReductionOperator>(x.v.t)};
   bool ok = false;
   common::visit(
       common::visitors{
diff --git a/flang/test/Lower/OpenMP/Todo/reduction-modifiers.f90 b/flang/test/Lower/OpenMP/Todo/reduction-modifiers.f90
new file mode 100644
index 0000000000000..5e566466492ce
--- /dev/null
+++ b/flang/test/Lower/OpenMP/Todo/reduction-modifiers.f90
@@ -0,0 +1,13 @@
+! RUN: %not_todo_cmd bbc -emit-fir -fopenmp -o - %s 2>&1 | FileCheck %s
+! RUN: %not_todo_cmd %flang_fc1 -emit-fir -fopenmp -o - %s 2>&1 | FileCheck %s
+
+! CHECK: not yet implemented: Reduction modifiers are not supported
+
+subroutine foo()
+  integer :: i, j
+  j = 0
+  !$omp do reduction (inscan, *: j)
+  do i = 1, 10
+    j = j + 1
+  end do
+end subroutine
diff --git a/flang/test/Parser/OpenMP/reduction-modifier.f90 b/flang/test/Parser/OpenMP/reduction-modifier.f90
new file mode 100644
index 0000000000000..d46aa70959592
--- /dev/null
+++ b/flang/test/Parser/OpenMP/reduction-modifier.f90
@@ -0,0 +1,20 @@
+! RUN: %flang_fc1 -fdebug-unparse-no-sema -fopenmp %s | FileCheck --ignore-case %s
+! RUN: %flang_fc1 -fdebug-dump-parse-tree-no-sema -fopenmp %s | FileCheck --check-prefix="PARSE-TREE" %s
+
+subroutine foo()
+  integer :: i, j
+  j = 0
+! CHECK: !$OMP DO  REDUCTION(TASK,*:j)
+! PARSE-TREE: | | ExecutionPartConstruct -> ExecutableConstruct -> OpenMPConstruct -> OpenMPLoopConstruct
+! PARSE-TREE: | | | OmpBeginLoopDirective
+! PARSE-TREE: | | | | OmpLoopDirective -> llvm::omp::Directive = do
+! PARSE-TREE: | | | | OmpClauseList -> OmpClause -> Reduction -> OmpReductionClause
+! PARSE-TREE: | | | | | ReductionModifier = Task
+! PARSE-TREE: | | | | | OmpReductionOperator -> DefinedOperator -> IntrinsicOperator = Multiply
+! PARSE-TREE: | | | | | OmpObjectList -> OmpObject -> Designator -> DataRef -> Name = 'j
+  !$omp do reduction (task, *: j)
+  do i = 1, 10
+    j = j + 1
+  end do
+  !$omp end do
+end
diff --git a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h
index 6ce972adcf0fb..daef02bcfc9a3 100644
--- a/llvm/include/llvm/Frontend/OpenMP/ClauseT.h
+++ b/llvm/include/llvm/Frontend/OpenMP/ClauseT.h
@@ -949,7 +949,7 @@ struct ReductionT {
   using ReductionIdentifiers = ListT<type::ReductionIdentifierT<I, E>>;
   ENUM(ReductionModifier, Default, Inscan, Task);
   using TupleTrait = std::true_type;
-  std::tuple<ReductionIdentifiers, OPT(ReductionModifier), List> t;
+  std::tuple<OPT(ReductionModifier), ReductionIdentifiers, List> t;
 };
 
 // V5.2: [15.8.1] `memory-order` clauses

From 67e733d36ef0c5ec9fab899ecf9f191d83c7d418 Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Mon, 15 Apr 2024 09:46:58 -0400
Subject: [PATCH 429/886] [BOLT][NFC] Refactor relocation printing (#88180)

Some refactoring.

* Make the arrays themselves const -- not just the strings being pointed to.
* Move each relocation array to the case using it.
* Add a suitable assert, to make sure there's no out-of-bounds indexing.
---
 bolt/lib/Core/Relocation.cpp | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/bolt/lib/Core/Relocation.cpp b/bolt/lib/Core/Relocation.cpp
index d16b7a94787c6..4e888a5b147ac 100644
--- a/bolt/lib/Core/Relocation.cpp
+++ b/bolt/lib/Core/Relocation.cpp
@@ -1064,21 +1064,19 @@ MCBinaryExpr::Opcode Relocation::getComposeOpcodeFor(uint64_t Type) {
   }
 }
 
-#define ELF_RELOC(name, value) #name,
-
 void Relocation::print(raw_ostream &OS) const {
-  static const char *X86RelocNames[] = {
-#include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
-  };
-  static const char *AArch64RelocNames[] = {
-#include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
-  };
   switch (Arch) {
   default:
     OS << "RType:" << Twine::utohexstr(Type);
     break;
 
   case Triple::aarch64:
+    static const char *const AArch64RelocNames[] = {
+#define ELF_RELOC(name, value) #name,
+#include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
+#undef ELF_RELOC
+    };
+    assert(Type < ArrayRef(AArch64RelocNames).size());
     OS << AArch64RelocNames[Type];
     break;
 
@@ -1088,16 +1086,22 @@ void Relocation::print(raw_ostream &OS) const {
     switch (Type) {
     default:
       llvm_unreachable("illegal RISC-V relocation");
-#undef ELF_RELOC
 #define ELF_RELOC(name, value)                                                 \
   case value:                                                                  \
     OS << #name;                                                               \
     break;
 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
+#undef ELF_RELOC
     }
     break;
 
   case Triple::x86_64:
+    static const char *const X86RelocNames[] = {
+#define ELF_RELOC(name, value) #name,
+#include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
+#undef ELF_RELOC
+    };
+    assert(Type < ArrayRef(X86RelocNames).size());
     OS << X86RelocNames[Type];
     break;
   }

From 105dcc882cf0152baeaa02ac0e50e2527b6940db Mon Sep 17 00:00:00 2001
From: Farzon Lotfi <1802579+farzonl@users.noreply.github.com>
Date: Mon, 15 Apr 2024 06:52:47 -0700
Subject: [PATCH 430/886] [HLSL][SPIRV] Add any intrinsic lowering (#88325)

- `CGBuiltin.cpp` - Switch to using
`CGM.getHLSLRuntime().get##NAME##Intrinsic()`
- `CGHLSLRuntime.h` - Add any to backend intrinsic abstraction
-  `IntrinsicsSPIRV.td` - Add any intrinsic to SPIR-V.
- `SPIRVInstructionSelector.cpp` - Add means of selecting any intrinsic.
Any and All share the same behavior up to the opCode. They are only
different in vector cases.

Completes #88045
---
 clang/lib/CodeGen/CGBuiltin.cpp               |   3 +-
 clang/lib/CodeGen/CGHLSLRuntime.h             |   1 +
 clang/test/CodeGenHLSL/builtins/any.hlsl      | 370 ++++++++++++------
 llvm/include/llvm/IR/IntrinsicsSPIRV.td       |   1 +
 .../Target/SPIRV/SPIRVInstructionSelector.cpp |  29 +-
 .../test/CodeGen/SPIRV/hlsl-intrinsics/any.ll | 187 +++++++++
 6 files changed, 460 insertions(+), 131 deletions(-)
 create mode 100644 llvm/test/CodeGen/SPIRV/hlsl-intrinsics/any.ll

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index c7b219dcfcec5..9f95697f284c4 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18194,7 +18194,8 @@ Value *CodeGenFunction::EmitHLSLBuiltinExpr(unsigned BuiltinID,
     Value *Op0 = EmitScalarExpr(E->getArg(0));
     return Builder.CreateIntrinsic(
         /*ReturnType=*/llvm::Type::getInt1Ty(getLLVMContext()),
-        Intrinsic::dx_any, ArrayRef<Value *>{Op0}, nullptr, "dx.any");
+        CGM.getHLSLRuntime().getAnyIntrinsic(), ArrayRef<Value *>{Op0}, nullptr,
+        "hlsl.any");
   }
   case Builtin::BI__builtin_hlsl_elementwise_clamp: {
     Value *OpX = EmitScalarExpr(E->getArg(0));
diff --git a/clang/lib/CodeGen/CGHLSLRuntime.h b/clang/lib/CodeGen/CGHLSLRuntime.h
index 2b8073aef973f..506b364f5b2ec 100644
--- a/clang/lib/CodeGen/CGHLSLRuntime.h
+++ b/clang/lib/CodeGen/CGHLSLRuntime.h
@@ -73,6 +73,7 @@ class CGHLSLRuntime {
   //===----------------------------------------------------------------------===//
 
   GENERATE_HLSL_INTRINSIC_FUNCTION(All, all)
+  GENERATE_HLSL_INTRINSIC_FUNCTION(Any, any)
   GENERATE_HLSL_INTRINSIC_FUNCTION(ThreadId, thread_id)
 
   //===----------------------------------------------------------------------===//
diff --git a/clang/test/CodeGenHLSL/builtins/any.hlsl b/clang/test/CodeGenHLSL/builtins/any.hlsl
index ae348fec756b3..84584281a3b7d 100644
--- a/clang/test/CodeGenHLSL/builtins/any.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/any.hlsl
@@ -1,186 +1,304 @@
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-unknown-vulkan-compute %s -fnative-half-type \
+// RUN:   -emit-llvm -disable-llvm-passes -o - | FileCheck %s \ 
+// RUN:   --check-prefixes=CHECK,NATIVE_HALF,SPIR_NATIVE_HALF,SPIR_CHECK
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
+// RUN:   spirv-unknown-vulkan-compute %s -emit-llvm -disable-llvm-passes \
+// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,SPIR_NO_HALF,SPIR_CHECK
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -fnative-half-type \
 // RUN:   -emit-llvm -disable-llvm-passes -o - | FileCheck %s \ 
-// RUN:   --check-prefixes=CHECK,NATIVE_HALF
+// RUN:   --check-prefixes=CHECK,NATIVE_HALF,DXIL_NATIVE_HALF,DXIL_CHECK
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
 // RUN:   dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes \
-// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,NO_HALF
+// RUN:   -o - | FileCheck %s --check-prefixes=CHECK,DXIL_NO_HALF,DXIL_CHECK
 
 #ifdef __HLSL_ENABLE_16_BIT
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.i16
-// NATIVE_HALF: ret i1 %dx.any
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_int16_t(int16_t p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v2i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v2i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v2i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_int16_t2(int16_t2 p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v3i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v3i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v3i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_int16_t3(int16_t3 p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v4i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v4i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v4i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_int16_t4(int16_t4 p0) { return any(p0); }
 
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.i16
-// NATIVE_HALF: ret i1 %dx.any
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_uint16_t(uint16_t p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v2i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v2i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v2i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_uint16_t2(uint16_t2 p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v3i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v3i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v3i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_uint16_t3(uint16_t3 p0) { return any(p0); }
-// NATIVE_HALF: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v4i16
-// NATIVE_HALF: ret i1 %dx.any
+
+// DXIL_NATIVE_HALF: define noundef i1 @
+// SPIR_NATIVE_HALF: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v4i16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v4i16
+// NATIVE_HALF: ret i1 %hlsl.any
 bool test_any_uint16_t4(uint16_t4 p0) { return any(p0); }
 #endif // __HLSL_ENABLE_16_BIT
 
-// CHECK: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.f16
-// NO_HALF: %dx.any = call i1 @llvm.dx.any.f32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.f16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.f16
+// DXIL_NO_HALF: %hlsl.any = call i1 @llvm.dx.any.f32
+// SPIR_NO_HALF: %hlsl.any = call i1 @llvm.spv.any.f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_half(half p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v2f16
-// NO_HALF: %dx.any = call i1 @llvm.dx.any.v2f32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v2f16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v2f16
+// DXIL_NO_HALF: %hlsl.any = call i1 @llvm.dx.any.v2f32
+// SPIR_NO_HALF: %hlsl.any = call i1 @llvm.spv.any.v2f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_half2(half2 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v3f16
-// NO_HALF: %dx.any = call i1 @llvm.dx.any.v3f32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v3f16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v3f16
+// DXIL_NO_HALF: %hlsl.any = call i1 @llvm.dx.any.v3f32
+// SPIR_NO_HALF: %hlsl.any = call i1 @llvm.spv.any.v3f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_half3(half3 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// NATIVE_HALF: %dx.any = call i1 @llvm.dx.any.v4f16
-// NO_HALF: %dx.any = call i1 @llvm.dx.any.v4f32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_NATIVE_HALF: %hlsl.any = call i1 @llvm.dx.any.v4f16
+// SPIR_NATIVE_HALF: %hlsl.any = call i1 @llvm.spv.any.v4f16
+// DXIL_NO_HALF: %hlsl.any = call i1 @llvm.dx.any.v4f32
+// SPIR_NO_HALF: %hlsl.any = call i1 @llvm.spv.any.v4f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_half4(half4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.f32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.f32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_float(float p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2f32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2f32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_float2(float2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3f32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3f32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_float3(float3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4f32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4f32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4f32
+// CHECK: ret i1 %hlsl.any
 bool test_any_float4(float4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.f64
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.f64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.f64
+// CHECK: ret i1 %hlsl.any
 bool test_any_double(double p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2f64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2f64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2f64
+// CHECK: ret i1 %hlsl.any
 bool test_any_double2(double2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3f64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3f64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3f64
+// CHECK: ret i1 %hlsl.any
 bool test_any_double3(double3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4f64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4f64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4f64
+// CHECK: ret i1 %hlsl.any
 bool test_any_double4(double4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.i32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_int(int p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_int2(int2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_int3(int3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_int4(int4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.i32
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint(uint p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint2(uint2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint3(uint3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4i32
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4i32
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4i32
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint4(uint4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.i64
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_int64_t(int64_t p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_int64_t2(int64_t2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_int64_t3(int64_t3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_int64_t4(int64_t4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.i64
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint64_t(uint64_t p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint64_t2(uint64_t2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint64_t3(uint64_t3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4i64
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4i64
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4i64
+// CHECK: ret i1 %hlsl.any
 bool test_any_uint64_t4(uint64_t4 p0) { return any(p0); }
 
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.i1
-// CHECK: ret i1 %dx.any
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.i1
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.i1
+// CHECK: ret i1 %hlsl.any
 bool test_any_bool(bool p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v2i1
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v2i1
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v2i1
+// CHECK: ret i1 %hlsl.any
 bool test_any_bool2(bool2 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v3i1
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v3i1
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v3i1
+// CHECK: ret i1 %hlsl.any
 bool test_any_bool3(bool3 p0) { return any(p0); }
-// CHECK: define noundef i1 @
-// CHECK: %dx.any = call i1 @llvm.dx.any.v4i1
-// CHECK: ret i1 %dx.any
+
+// DXIL_CHECK: define noundef i1 @
+// SPIR_CHECK: define spir_func noundef i1 @
+// DXIL_CHECK: %hlsl.any = call i1 @llvm.dx.any.v4i1
+// SPIR_CHECK: %hlsl.any = call i1 @llvm.spv.any.v4i1
+// CHECK: ret i1 %hlsl.any
 bool test_any_bool4(bool4 p0) { return any(p0); }
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index f843383f0b00e..b6618baceb560 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -57,4 +57,5 @@ let TargetPrefix = "spv" in {
   def int_spv_create_handle : ClangBuiltin<"__builtin_hlsl_create_handle">,
       Intrinsic<[ llvm_ptr_ty ], [llvm_i8_ty], [IntrWillReturn]>;
   def int_spv_all : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_any_ty]>;
+  def int_spv_any : DefaultAttrsIntrinsic<[llvm_i1_ty], [llvm_any_ty]>;
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index c1c0fc4b7dd48..200fe38298c02 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -145,9 +145,15 @@ class SPIRVInstructionSelector : public InstructionSelector {
   bool selectAddrSpaceCast(Register ResVReg, const SPIRVType *ResType,
                            MachineInstr &I) const;
 
+  bool selectAnyOrAll(Register ResVReg, const SPIRVType *ResType,
+                      MachineInstr &I, unsigned OpType) const;
+
   bool selectAll(Register ResVReg, const SPIRVType *ResType,
                  MachineInstr &I) const;
 
+  bool selectAny(Register ResVReg, const SPIRVType *ResType,
+                 MachineInstr &I) const;
+
   bool selectBitreverse(Register ResVReg, const SPIRVType *ResType,
                         MachineInstr &I) const;
 
@@ -1160,9 +1166,10 @@ static unsigned getBoolCmpOpcode(unsigned PredNum) {
   }
 }
 
-bool SPIRVInstructionSelector::selectAll(Register ResVReg,
-                                         const SPIRVType *ResType,
-                                         MachineInstr &I) const {
+bool SPIRVInstructionSelector::selectAnyOrAll(Register ResVReg,
+                                              const SPIRVType *ResType,
+                                              MachineInstr &I,
+                                              unsigned OpAnyOrAll) const {
   assert(I.getNumOperands() == 3);
   assert(I.getOperand(2).isReg());
   MachineBasicBlock &BB = *I.getParent();
@@ -1212,13 +1219,25 @@ bool SPIRVInstructionSelector::selectAll(Register ResVReg,
   if (!IsVectorTy)
     return true;
 
-  return BuildMI(BB, I, I.getDebugLoc(), TII.get(SPIRV::OpAll))
+  return BuildMI(BB, I, I.getDebugLoc(), TII.get(OpAnyOrAll))
       .addDef(ResVReg)
       .addUse(GR.getSPIRVTypeID(SpvBoolScalarTy))
       .addUse(NotEqualReg)
       .constrainAllUses(TII, TRI, RBI);
 }
 
+bool SPIRVInstructionSelector::selectAll(Register ResVReg,
+                                         const SPIRVType *ResType,
+                                         MachineInstr &I) const {
+  return selectAnyOrAll(ResVReg, ResType, I, SPIRV::OpAll);
+}
+
+bool SPIRVInstructionSelector::selectAny(Register ResVReg,
+                                         const SPIRVType *ResType,
+                                         MachineInstr &I) const {
+  return selectAnyOrAll(ResVReg, ResType, I, SPIRV::OpAny);
+}
+
 bool SPIRVInstructionSelector::selectBitreverse(Register ResVReg,
                                                 const SPIRVType *ResType,
                                                 MachineInstr &I) const {
@@ -1877,6 +1896,8 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
     return selectSpvThreadId(ResVReg, ResType, I);
   case Intrinsic::spv_all:
     return selectAll(ResVReg, ResType, I);
+  case Intrinsic::spv_any:
+    return selectAny(ResVReg, ResType, I);
   case Intrinsic::spv_lifetime_start:
   case Intrinsic::spv_lifetime_end: {
     unsigned Op = IID == Intrinsic::spv_lifetime_start ? SPIRV::OpLifetimeStart
diff --git a/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/any.ll b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/any.ll
new file mode 100644
index 0000000000000..b1dd388f5c6e3
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/any.ll
@@ -0,0 +1,187 @@
+; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-HLSL
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-OCL
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; Make sure spirv operation function calls for any are generated.
+
+; CHECK-HLSL-DAG: OpMemoryModel Logical GLSL450
+; CHECK-OCL-DAG: OpMemoryModel Physical32 OpenCL
+; CHECK-DAG: OpName %[[#any_bool_arg:]] "a"
+; CHECK-DAG: %[[#int_64:]] = OpTypeInt 64 0
+; CHECK-DAG: %[[#bool:]] = OpTypeBool
+; CHECK-DAG: %[[#int_32:]] = OpTypeInt 32 0
+; CHECK-DAG: %[[#int_16:]] = OpTypeInt 16 0
+; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
+; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
+; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
+; CHECK-DAG: %[[#vec4_bool:]] = OpTypeVector %[[#bool]] 4
+; CHECK-DAG: %[[#vec4_16:]] = OpTypeVector %[[#int_16]] 4
+; CHECK-DAG: %[[#vec4_32:]] = OpTypeVector %[[#int_32]] 4
+; CHECK-DAG: %[[#vec4_64:]] = OpTypeVector %[[#int_64]] 4
+; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
+; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
+; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
+
+; CHECK-HLSL-DAG: %[[#const_i64_0:]] = OpConstant %[[#int_64]] 0
+; CHECK-HLSL-DAG: %[[#const_i32_0:]] = OpConstant %[[#int_32]] 0
+; CHECK-HLSL-DAG: %[[#const_i16_0:]] = OpConstant %[[#int_16]] 0
+; CHECK-HLSL-DAG: %[[#const_f64_0:]] = OpConstant %[[#float_64]] 0
+; CHECK-HLSL-DAG: %[[#const_f32_0:]] = OpConstant %[[#float_32:]] 0
+; CHECK-HLSL-DAG: %[[#const_f16_0:]] = OpConstant %[[#float_16:]] 0
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i16:]] = OpConstantComposite %[[#vec4_16:]] %[[#const_i16_0:]] %[[#const_i16_0:]] %[[#const_i16_0:]] %[[#const_i16_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i32:]] = OpConstantComposite %[[#vec4_32:]] %[[#const_i32_0:]] %[[#const_i32_0:]] %[[#const_i32_0:]] %[[#const_i32_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_i64:]] = OpConstantComposite %[[#vec4_64:]] %[[#const_i64_0:]] %[[#const_i64_0:]] %[[#const_i64_0:]] %[[#const_i64_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f16:]] = OpConstantComposite %[[#vec4_float_16:]] %[[#const_f16_0:]] %[[#const_f16_0:]] %[[#const_f16_0:]] %[[#const_f16_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f32:]] = OpConstantComposite %[[#vec4_float_32:]] %[[#const_f32_0:]] %[[#const_f32_0:]] %[[#const_f32_0:]] %[[#const_f32_0:]]
+; CHECK-HLSL-DAG: %[[#vec4_const_zeros_f64:]] = OpConstantComposite %[[#vec4_float_64:]] %[[#const_f64_0:]] %[[#const_f64_0:]] %[[#const_f64_0:]] %[[#const_f64_0:]]
+
+; CHECK-OCL-DAG: %[[#const_i64_0:]] = OpConstantNull %[[#int_64]]
+; CHECK-OCL-DAG: %[[#const_i32_0:]] = OpConstantNull %[[#int_32]]
+; CHECK-OCL-DAG: %[[#const_i16_0:]] = OpConstantNull %[[#int_16]]
+; CHECK-OCL-DAG: %[[#const_f64_0:]] = OpConstantNull %[[#float_64]] 
+; CHECK-OCL-DAG: %[[#const_f32_0:]] = OpConstantNull %[[#float_32:]]
+; CHECK-OCL-DAG: %[[#const_f16_0:]] = OpConstantNull %[[#float_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i16:]] = OpConstantNull %[[#vec4_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i32:]] = OpConstantNull %[[#vec4_32:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_i64:]] = OpConstantNull %[[#vec4_64:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f16:]] = OpConstantNull %[[#vec4_float_16:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f32:]] = OpConstantNull %[[#vec4_float_32:]]
+; CHECK-OCL-DAG: %[[#vec4_const_zeros_f64:]] = OpConstantNull %[[#vec4_float_64:]]
+
+define noundef i1 @any_int64_t(i64 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i64_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.i64(i64 %p0)
+  ret i1 %hlsl.any
+}
+
+
+define noundef i1 @any_int(i32 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i32_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.i32(i32 %p0)
+  ret i1 %hlsl.any
+}
+
+
+define noundef i1 @any_int16_t(i16 noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpINotEqual %[[#bool:]] %[[#arg0:]] %[[#const_i16_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.i16(i16 %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_double(double noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f64_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.f64(double %p0)
+  ret i1 %hlsl.any
+}
+
+
+define noundef i1 @any_float(float noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f32_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.f32(float %p0)
+  ret i1 %hlsl.any
+}
+
+
+define noundef i1 @any_half(half noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpFOrdNotEqual %[[#bool:]] %[[#arg0:]] %[[#const_f16_0:]]
+  %hlsl.any = call i1 @llvm.spv.any.f16(half %p0)
+  ret i1 %hlsl.any
+}
+
+
+define noundef i1 @any_bool4(<4 x i1> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#]] = OpAny %[[#vec4_bool:]] %[[#arg0:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4i1(<4 x i1> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_short4(<4 x i16> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#shortVecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i16:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool:]] %[[#shortVecNotEq:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4i16(<4 x i16> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_int4(<4 x i32> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#i32VecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i32:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool:]] %[[#i32VecNotEq:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4i32(<4 x i32> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_int64_t4(<4 x i64> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#i64VecNotEq:]] = OpINotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_i64:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool:]] %[[#i64VecNotEq]]
+  %hlsl.any = call i1 @llvm.spv.any.v4i64(<4 x i64> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_half4(<4 x half> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f16VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f16:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool]] %[[#f16VecNotEq:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4f16(<4 x half> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_float4(<4 x float> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f32VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f32:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool:]] %[[#f32VecNotEq:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4f32(<4 x float> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_double4(<4 x double> noundef %p0) {
+entry:
+  ; CHECK: %[[#arg0:]] = OpFunctionParameter %[[#]]
+  ; CHECK: %[[#f64VecNotEq:]] = OpFOrdNotEqual %[[#vec4_bool:]] %[[#arg0:]] %[[#vec4_const_zeros_f64:]]
+  ; CHECK: %[[#]] = OpAny %[[#bool:]] %[[#f64VecNotEq:]]
+  %hlsl.any = call i1 @llvm.spv.any.v4f64(<4 x double> %p0)
+  ret i1 %hlsl.any
+}
+
+define noundef i1 @any_bool(i1 noundef %a) {
+entry:
+  ; CHECK: %[[#any_bool_arg:]] = OpFunctionParameter %[[#bool:]]
+  ; CHECK: OpReturnValue %[[#any_bool_arg:]]
+  %hlsl.any = call i1 @llvm.spv.any.i1(i1 %a)
+  ret i1 %hlsl.any
+}
+
+declare i1 @llvm.spv.any.v4f16(<4 x half>)
+declare i1 @llvm.spv.any.v4f32(<4 x float>)
+declare i1 @llvm.spv.any.v4f64(<4 x double>)
+declare i1 @llvm.spv.any.v4i1(<4 x i1>)
+declare i1 @llvm.spv.any.v4i16(<4 x i16>)
+declare i1 @llvm.spv.any.v4i32(<4 x i32>)
+declare i1 @llvm.spv.any.v4i64(<4 x i64>)
+declare i1 @llvm.spv.any.i1(i1)
+declare i1 @llvm.spv.any.i16(i16)
+declare i1 @llvm.spv.any.i32(i32)
+declare i1 @llvm.spv.any.i64(i64)
+declare i1 @llvm.spv.any.f16(half)
+declare i1 @llvm.spv.any.f32(float)
+declare i1 @llvm.spv.any.f64(double)

From 32cb3c55080eec70abf282fbe75e6a58cf76f92a Mon Sep 17 00:00:00 2001
From: Piyou Chen <piyou.chen@sifive.com>
Date: Mon, 15 Apr 2024 21:58:57 +0800
Subject: [PATCH 431/886] [NFC][LLVM][CodeGen] Move LiveDebugVariables.h into
 llvm/include/llvm/CodeGen (#88374)

This patch make `LiveDebugVariables` can be used by passes outside of
`lib/CodeGen`.

If we run a pass that occurs between the split register allocation pass
without preserving this pass, it will be freed and recomputed until it
encounters the next pass that needs LiveDebugVariables.

However, `LiveDebugVariables` will raise an assertion due to the pass
being freed without emitting a debug value.

This is reason we need `LiveDebugVariables` to be available for passes
outside of lib/Codegen.
---
 llvm/{lib => include/llvm}/CodeGen/LiveDebugVariables.h | 8 ++++----
 llvm/lib/CodeGen/LiveDebugVariables.cpp                 | 2 +-
 llvm/lib/CodeGen/RegAllocBasic.cpp                      | 2 +-
 llvm/lib/CodeGen/RegAllocGreedy.cpp                     | 2 +-
 llvm/lib/CodeGen/VirtRegMap.cpp                         | 2 +-
 5 files changed, 8 insertions(+), 8 deletions(-)
 rename llvm/{lib => include/llvm}/CodeGen/LiveDebugVariables.h (90%)

diff --git a/llvm/lib/CodeGen/LiveDebugVariables.h b/llvm/include/llvm/CodeGen/LiveDebugVariables.h
similarity index 90%
rename from llvm/lib/CodeGen/LiveDebugVariables.h
rename to llvm/include/llvm/CodeGen/LiveDebugVariables.h
index 9998ce9e8dad8..a4b5a87fd3887 100644
--- a/llvm/lib/CodeGen/LiveDebugVariables.h
+++ b/llvm/include/llvm/CodeGen/LiveDebugVariables.h
@@ -17,8 +17,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_LIB_CODEGEN_LIVEDEBUGVARIABLES_H
-#define LLVM_LIB_CODEGEN_LIVEDEBUGVARIABLES_H
+#ifndef LLVM_CODEGEN_LIVEDEBUGVARIABLES_H
+#define LLVM_CODEGEN_LIVEDEBUGVARIABLES_H
 
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/Support/Compiler.h"
@@ -29,7 +29,7 @@ template <typename T> class ArrayRef;
 class LiveIntervals;
 class VirtRegMap;
 
-class LLVM_LIBRARY_VISIBILITY LiveDebugVariables : public MachineFunctionPass {
+class LiveDebugVariables : public MachineFunctionPass {
   void *pImpl = nullptr;
 
 public:
@@ -65,4 +65,4 @@ class LLVM_LIBRARY_VISIBILITY LiveDebugVariables : public MachineFunctionPass {
 
 } // end namespace llvm
 
-#endif // LLVM_LIB_CODEGEN_LIVEDEBUGVARIABLES_H
+#endif // LLVM_CODEGEN_LIVEDEBUGVARIABLES_H
diff --git a/llvm/lib/CodeGen/LiveDebugVariables.cpp b/llvm/lib/CodeGen/LiveDebugVariables.cpp
index 7cb90af5ff173..3a59ae7ab0664 100644
--- a/llvm/lib/CodeGen/LiveDebugVariables.cpp
+++ b/llvm/lib/CodeGen/LiveDebugVariables.cpp
@@ -18,7 +18,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "LiveDebugVariables.h"
+#include "llvm/CodeGen/LiveDebugVariables.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/IntervalMap.h"
diff --git a/llvm/lib/CodeGen/RegAllocBasic.cpp b/llvm/lib/CodeGen/RegAllocBasic.cpp
index 6661991396302..5bd3b126aa166 100644
--- a/llvm/lib/CodeGen/RegAllocBasic.cpp
+++ b/llvm/lib/CodeGen/RegAllocBasic.cpp
@@ -12,10 +12,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "AllocationOrder.h"
-#include "LiveDebugVariables.h"
 #include "RegAllocBase.h"
 #include "llvm/Analysis/AliasAnalysis.h"
 #include "llvm/CodeGen/CalcSpillWeights.h"
+#include "llvm/CodeGen/LiveDebugVariables.h"
 #include "llvm/CodeGen/LiveIntervals.h"
 #include "llvm/CodeGen/LiveRangeEdit.h"
 #include "llvm/CodeGen/LiveRegMatrix.h"
diff --git a/llvm/lib/CodeGen/RegAllocGreedy.cpp b/llvm/lib/CodeGen/RegAllocGreedy.cpp
index a208bf89fadf2..348277224c7ae 100644
--- a/llvm/lib/CodeGen/RegAllocGreedy.cpp
+++ b/llvm/lib/CodeGen/RegAllocGreedy.cpp
@@ -14,7 +14,6 @@
 #include "RegAllocGreedy.h"
 #include "AllocationOrder.h"
 #include "InterferenceCache.h"
-#include "LiveDebugVariables.h"
 #include "RegAllocBase.h"
 #include "RegAllocEvictionAdvisor.h"
 #include "RegAllocPriorityAdvisor.h"
@@ -31,6 +30,7 @@
 #include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/CodeGen/CalcSpillWeights.h"
 #include "llvm/CodeGen/EdgeBundles.h"
+#include "llvm/CodeGen/LiveDebugVariables.h"
 #include "llvm/CodeGen/LiveInterval.h"
 #include "llvm/CodeGen/LiveIntervalUnion.h"
 #include "llvm/CodeGen/LiveIntervals.h"
diff --git a/llvm/lib/CodeGen/VirtRegMap.cpp b/llvm/lib/CodeGen/VirtRegMap.cpp
index 48f4ee29fbe95..2c778980f5d1e 100644
--- a/llvm/lib/CodeGen/VirtRegMap.cpp
+++ b/llvm/lib/CodeGen/VirtRegMap.cpp
@@ -16,9 +16,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/CodeGen/VirtRegMap.h"
-#include "LiveDebugVariables.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
+#include "llvm/CodeGen/LiveDebugVariables.h"
 #include "llvm/CodeGen/LiveInterval.h"
 #include "llvm/CodeGen/LiveIntervals.h"
 #include "llvm/CodeGen/LiveStacks.h"

From c3d58867bd3da85958dbbb46c4a22c0d3c60af66 Mon Sep 17 00:00:00 2001
From: Sander de Smalen <sander.desmalen@arm.com>
Date: Mon, 15 Apr 2024 15:07:20 +0100
Subject: [PATCH 432/886] [AArch64][SME] Create new pass to remove
 COALESCER_BARRIER early. (#85386)

The purpose of the COALESCER_BARRIER pseudo node is to prevent the
register coalescer from coalescing certain COPY instructions around
smstart/smstop instructions, so that we spill only the (required) FPR
register rather than the encompassing ZPR register.

The pseudos are removed in the AArch64ExpandPseudo pass. However,
because the node itself is a _use_ of a register, this occassionally
leads to redundant spills/fills, because the register allocator thinks
the virtual register is actually used before an smstart/smstop
instruction, causing it to be filled, at which points it requires
immediate spilling again to ensure it stays live over the smstart/smstop
instruction.

We can avoid that by removing the pseudo nodes right after coalescing,
but before register allocation.
---
 llvm/lib/Target/AArch64/AArch64.h             |   2 +
 .../AArch64/AArch64PostCoalescerPass.cpp      | 101 ++++++++++++++++++
 .../Target/AArch64/AArch64TargetMachine.cpp   |   7 ++
 llvm/lib/Target/AArch64/CMakeLists.txt        |   1 +
 llvm/test/CodeGen/AArch64/O3-pipeline.ll      |   1 +
 .../sme-avoid-coalescing-locally-streaming.ll |  10 +-
 ...compatible-to-normal-fn-wihout-sme-attr.ll |   2 -
 .../AArch64/sme-disable-gisel-fisel.ll        |  10 +-
 ...ate-sm-changing-call-disable-coalescing.ll |   8 --
 ...ing-body-streaming-compatible-interface.ll |  26 ++---
 .../CodeGen/AArch64/sme-streaming-body.ll     |   4 -
 .../sme-streaming-compatible-interface.ll     |   4 -
 .../AArch64/sme-streaming-interface.ll        |   8 +-
 ...nging-call-disable-stackslot-scavenging.ll |   2 +-
 14 files changed, 134 insertions(+), 52 deletions(-)
 create mode 100644 llvm/lib/Target/AArch64/AArch64PostCoalescerPass.cpp

diff --git a/llvm/lib/Target/AArch64/AArch64.h b/llvm/lib/Target/AArch64/AArch64.h
index f7d81f42ef5d8..b70fbe42fe5fc 100644
--- a/llvm/lib/Target/AArch64/AArch64.h
+++ b/llvm/lib/Target/AArch64/AArch64.h
@@ -54,6 +54,7 @@ FunctionPass *createFalkorMarkStridedAccessesPass();
 FunctionPass *createAArch64PointerAuthPass();
 FunctionPass *createAArch64BranchTargetsPass();
 FunctionPass *createAArch64MIPeepholeOptPass();
+FunctionPass *createAArch64PostCoalescerPass();
 
 FunctionPass *createAArch64CleanupLocalDynamicTLSPass();
 
@@ -93,6 +94,7 @@ void initializeAArch64LoopIdiomTransformLegacyPassPass(PassRegistry &);
 void initializeAArch64LowerHomogeneousPrologEpilogPass(PassRegistry &);
 void initializeAArch64MIPeepholeOptPass(PassRegistry &);
 void initializeAArch64O0PreLegalizerCombinerPass(PassRegistry &);
+void initializeAArch64PostCoalescerPass(PassRegistry &);
 void initializeAArch64PostLegalizerCombinerPass(PassRegistry &);
 void initializeAArch64PostLegalizerLoweringPass(PassRegistry &);
 void initializeAArch64PostSelectOptimizePass(PassRegistry &);
diff --git a/llvm/lib/Target/AArch64/AArch64PostCoalescerPass.cpp b/llvm/lib/Target/AArch64/AArch64PostCoalescerPass.cpp
new file mode 100644
index 0000000000000..dd5234c4504d3
--- /dev/null
+++ b/llvm/lib/Target/AArch64/AArch64PostCoalescerPass.cpp
@@ -0,0 +1,101 @@
+//===- AArch64PostCoalescerPass.cpp - AArch64 Post Coalescer pass ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//===----------------------------------------------------------------------===//
+
+#include "AArch64InstrInfo.h"
+#include "AArch64MachineFunctionInfo.h"
+#include "llvm/CodeGen/LiveIntervals.h"
+#include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/InitializePasses.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "aarch64-post-coalescer-pass"
+
+namespace {
+
+struct AArch64PostCoalescer : public MachineFunctionPass {
+  static char ID;
+
+  AArch64PostCoalescer() : MachineFunctionPass(ID) {
+    initializeAArch64PostCoalescerPass(*PassRegistry::getPassRegistry());
+  }
+
+  LiveIntervals *LIS;
+  MachineRegisterInfo *MRI;
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+
+  StringRef getPassName() const override {
+    return "AArch64 Post Coalescer pass";
+  }
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.setPreservesAll();
+    AU.addRequired<LiveIntervals>();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+};
+
+char AArch64PostCoalescer::ID = 0;
+
+} // end anonymous namespace
+
+INITIALIZE_PASS_BEGIN(AArch64PostCoalescer, "aarch64-post-coalescer-pass",
+                      "AArch64 Post Coalescer Pass", false, false)
+INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
+INITIALIZE_PASS_END(AArch64PostCoalescer, "aarch64-post-coalescer-pass",
+                    "AArch64 Post Coalescer Pass", false, false)
+
+bool AArch64PostCoalescer::runOnMachineFunction(MachineFunction &MF) {
+  if (skipFunction(MF.getFunction()))
+    return false;
+
+  AArch64FunctionInfo *FuncInfo = MF.getInfo<AArch64FunctionInfo>();
+  if (!FuncInfo->hasStreamingModeChanges())
+    return false;
+
+  MRI = &MF.getRegInfo();
+  LIS = &getAnalysis<LiveIntervals>();
+  bool Changed = false;
+
+  for (MachineBasicBlock &MBB : MF) {
+    for (MachineInstr &MI : make_early_inc_range(MBB)) {
+      switch (MI.getOpcode()) {
+      default:
+        break;
+      case AArch64::COALESCER_BARRIER_FPR16:
+      case AArch64::COALESCER_BARRIER_FPR32:
+      case AArch64::COALESCER_BARRIER_FPR64:
+      case AArch64::COALESCER_BARRIER_FPR128: {
+        Register Src = MI.getOperand(1).getReg();
+        Register Dst = MI.getOperand(0).getReg();
+        if (Src != Dst)
+          MRI->replaceRegWith(Dst, Src);
+
+        // MI must be erased from the basic block before recalculating the live
+        // interval.
+        LIS->RemoveMachineInstrFromMaps(MI);
+        MI.eraseFromParent();
+
+        LIS->removeInterval(Src);
+        LIS->createAndComputeVirtRegInterval(Src);
+
+        Changed = true;
+        break;
+      }
+      }
+    }
+  }
+
+  return Changed;
+}
+
+FunctionPass *llvm::createAArch64PostCoalescerPass() {
+  return new AArch64PostCoalescer();
+}
diff --git a/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp b/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
index 7d4a57d792a19..7ef78cbba352a 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetMachine.cpp
@@ -235,6 +235,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAArch64Target() {
   initializeAArch64O0PreLegalizerCombinerPass(*PR);
   initializeAArch64PreLegalizerCombinerPass(*PR);
   initializeAArch64PointerAuthPass(*PR);
+  initializeAArch64PostCoalescerPass(*PR);
   initializeAArch64PostLegalizerCombinerPass(*PR);
   initializeAArch64PostLegalizerLoweringPass(*PR);
   initializeAArch64PostSelectOptimizePass(*PR);
@@ -539,6 +540,7 @@ class AArch64PassConfig : public TargetPassConfig {
   void addPreEmitPass() override;
   void addPostBBSections() override;
   void addPreEmitPass2() override;
+  bool addRegAssignAndRewriteOptimized() override;
 
   std::unique_ptr<CSEConfigBase> getCSEConfig() const override;
 };
@@ -876,6 +878,11 @@ void AArch64PassConfig::addPreEmitPass2() {
   addPass(createUnpackMachineBundles(nullptr));
 }
 
+bool AArch64PassConfig::addRegAssignAndRewriteOptimized() {
+  addPass(createAArch64PostCoalescerPass());
+  return TargetPassConfig::addRegAssignAndRewriteOptimized();
+}
+
 MachineFunctionInfo *AArch64TargetMachine::createMachineFunctionInfo(
     BumpPtrAllocator &Allocator, const Function &F,
     const TargetSubtargetInfo *STI) const {
diff --git a/llvm/lib/Target/AArch64/CMakeLists.txt b/llvm/lib/Target/AArch64/CMakeLists.txt
index 95b228f293204..8e76f6c9279e7 100644
--- a/llvm/lib/Target/AArch64/CMakeLists.txt
+++ b/llvm/lib/Target/AArch64/CMakeLists.txt
@@ -73,6 +73,7 @@ add_llvm_target(AArch64CodeGen
   AArch64MIPeepholeOpt.cpp
   AArch64MCInstLower.cpp
   AArch64PointerAuth.cpp
+  AArch64PostCoalescerPass.cpp
   AArch64PromoteConstant.cpp
   AArch64PBQPRegAlloc.cpp
   AArch64RegisterInfo.cpp
diff --git a/llvm/test/CodeGen/AArch64/O3-pipeline.ll b/llvm/test/CodeGen/AArch64/O3-pipeline.ll
index eee9a27c90c19..d3c8e3b7e805c 100644
--- a/llvm/test/CodeGen/AArch64/O3-pipeline.ll
+++ b/llvm/test/CodeGen/AArch64/O3-pipeline.ll
@@ -167,6 +167,7 @@
 ; CHECK-NEXT:       Register Coalescer
 ; CHECK-NEXT:       Rename Disconnected Subregister Components
 ; CHECK-NEXT:       Machine Instruction Scheduler
+; CHECK-NEXT:       AArch64 Post Coalescer pass
 ; CHECK-NEXT:       Machine Block Frequency Analysis
 ; CHECK-NEXT:       Debug Variable Analysis
 ; CHECK-NEXT:       Live Stack Slot Analysis
diff --git a/llvm/test/CodeGen/AArch64/sme-avoid-coalescing-locally-streaming.ll b/llvm/test/CodeGen/AArch64/sme-avoid-coalescing-locally-streaming.ll
index cd5046a9a6475..8e3866fcec89a 100644
--- a/llvm/test/CodeGen/AArch64/sme-avoid-coalescing-locally-streaming.ll
+++ b/llvm/test/CodeGen/AArch64/sme-avoid-coalescing-locally-streaming.ll
@@ -25,8 +25,7 @@ define void @dont_coalesce_args(<2 x i64> %a) "aarch64_pstate_sm_body" nounwind
   ; CHECK-REGALLOC: bb.0 (%ir-block.0):
   ; CHECK-REGALLOC-NEXT:   liveins: $q0
   ; CHECK-REGALLOC-NEXT: {{  $}}
-  ; CHECK-REGALLOC-NEXT:   renamable $q0 = COALESCER_BARRIER_FPR128 killed renamable $q0
-  ; CHECK-REGALLOC-NEXT:   STRQui killed renamable $q0, %stack.0, 0 :: (store (s128) into %stack.0)
+  ; CHECK-REGALLOC-NEXT:   STRQui $q0, %stack.0, 0 :: (store (s128) into %stack.0)
   ; CHECK-REGALLOC-NEXT:   MSRpstatesvcrImm1 1, 1, csr_aarch64_smstartstop, implicit-def dead $nzcv, implicit $vg, implicit-def $vg
   ; CHECK-REGALLOC-NEXT:   renamable $q0 = LDRQui %stack.0, 0 :: (load (s128) from %stack.0)
   ; CHECK-REGALLOC-NEXT:   renamable $q0 = KILL killed renamable $q0, implicit-def $z0
@@ -61,7 +60,6 @@ define <2 x i64> @dont_coalesce_res() "aarch64_pstate_sm_body" nounwind {
   ; CHECK-REGALLOC-NEXT:   BL @scalable_res, csr_aarch64_sve_aapcs, implicit-def dead $lr, implicit $sp, implicit-def $sp, implicit-def $z0
   ; CHECK-REGALLOC-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
   ; CHECK-REGALLOC-NEXT:   renamable $q0 = KILL renamable $q0, implicit killed $z0
-  ; CHECK-REGALLOC-NEXT:   renamable $q0 = COALESCER_BARRIER_FPR128 killed renamable $q0
   ; CHECK-REGALLOC-NEXT:   STRQui killed renamable $q0, %stack.0, 0 :: (store (s128) into %stack.0)
   ; CHECK-REGALLOC-NEXT:   MSRpstatesvcrImm1 1, 0, csr_aarch64_smstartstop, implicit-def dead $nzcv, implicit-def dead $q0, implicit $vg, implicit-def $vg
   ; CHECK-REGALLOC-NEXT:   $q0 = LDRQui %stack.0, 0 :: (load (s128) from %stack.0)
@@ -94,17 +92,13 @@ define <2 x i64> @dont_coalesce_arg_that_is_also_res(<2 x i64> %a) "aarch64_psta
   ; CHECK-REGALLOC: bb.0 (%ir-block.0):
   ; CHECK-REGALLOC-NEXT:   liveins: $q0
   ; CHECK-REGALLOC-NEXT: {{  $}}
-  ; CHECK-REGALLOC-NEXT:   renamable $q0 = COALESCER_BARRIER_FPR128 killed renamable $q0
-  ; CHECK-REGALLOC-NEXT:   STRQui killed renamable $q0, %stack.0, 0 :: (store (s128) into %stack.0)
+  ; CHECK-REGALLOC-NEXT:   STRQui $q0, %stack.0, 0 :: (store (s128) into %stack.0)
   ; CHECK-REGALLOC-NEXT:   MSRpstatesvcrImm1 1, 1, csr_aarch64_smstartstop, implicit-def dead $nzcv, implicit $vg, implicit-def $vg
   ; CHECK-REGALLOC-NEXT:   renamable $q0 = LDRQui %stack.0, 0 :: (load (s128) from %stack.0)
   ; CHECK-REGALLOC-NEXT:   renamable $q0 = KILL killed renamable $q0, implicit-def $z0
   ; CHECK-REGALLOC-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
   ; CHECK-REGALLOC-NEXT:   BL @scalable_args, csr_aarch64_sve_aapcs, implicit-def dead $lr, implicit $sp, implicit $z0, implicit-def $sp
   ; CHECK-REGALLOC-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
-  ; CHECK-REGALLOC-NEXT:   renamable $q0 = LDRQui %stack.0, 0 :: (load (s128) from %stack.0)
-  ; CHECK-REGALLOC-NEXT:   renamable $q0 = COALESCER_BARRIER_FPR128 killed renamable $q0
-  ; CHECK-REGALLOC-NEXT:   STRQui killed renamable $q0, %stack.0, 0 :: (store (s128) into %stack.0)
   ; CHECK-REGALLOC-NEXT:   MSRpstatesvcrImm1 1, 0, csr_aarch64_smstartstop, implicit-def dead $nzcv, implicit-def dead $q0, implicit $vg, implicit-def $vg
   ; CHECK-REGALLOC-NEXT:   $q0 = LDRQui %stack.0, 0 :: (load (s128) from %stack.0)
   ; CHECK-REGALLOC-NEXT:   RET_ReallyLR implicit $q0
diff --git a/llvm/test/CodeGen/AArch64/sme-call-streaming-compatible-to-normal-fn-wihout-sme-attr.ll b/llvm/test/CodeGen/AArch64/sme-call-streaming-compatible-to-normal-fn-wihout-sme-attr.ll
index dba3227459b90..07377195d62a0 100644
--- a/llvm/test/CodeGen/AArch64/sme-call-streaming-compatible-to-normal-fn-wihout-sme-attr.ll
+++ b/llvm/test/CodeGen/AArch64/sme-call-streaming-compatible-to-normal-fn-wihout-sme-attr.ll
@@ -52,9 +52,7 @@ define void @streaming_compatible_arg(float %f) #0 {
 ; CHECK-NEXT:    stp x30, x19, [sp, #80] // 16-byte Folded Spill
 ; CHECK-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-NEXT:    bl __arm_sme_state
-; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
 ; CHECK-NEXT:    and x19, x0, #0x1
-; CHECK-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-NEXT:    tbz w19, #0, .LBB1_2
 ; CHECK-NEXT:  // %bb.1:
 ; CHECK-NEXT:    smstop sm
diff --git a/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll b/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
index cd348be5d771d..254e37e836cbb 100644
--- a/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
+++ b/llvm/test/CodeGen/AArch64/sme-disable-gisel-fisel.ll
@@ -331,9 +331,9 @@ define fp128 @f128_call_sm(fp128 %a, fp128 %b) "aarch64_pstate_sm_enabled" nounw
 ; CHECK-COMMON-NEXT:    stp d11, d10, [sp, #64] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    stp d9, d8, [sp, #80] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    str x30, [sp, #96] // 8-byte Folded Spill
-; CHECK-COMMON-NEXT:    stp q1, q0, [sp] // 32-byte Folded Spill
+; CHECK-COMMON-NEXT:    stp q0, q1, [sp] // 32-byte Folded Spill
 ; CHECK-COMMON-NEXT:    smstop sm
-; CHECK-COMMON-NEXT:    ldp q1, q0, [sp] // 32-byte Folded Reload
+; CHECK-COMMON-NEXT:    ldp q0, q1, [sp] // 32-byte Folded Reload
 ; CHECK-COMMON-NEXT:    bl __addtf3
 ; CHECK-COMMON-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    smstart sm
@@ -392,9 +392,9 @@ define float @frem_call_sm(float %a, float %b) "aarch64_pstate_sm_enabled" nounw
 ; CHECK-COMMON-NEXT:    stp d11, d10, [sp, #48] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    stp d9, d8, [sp, #64] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    str x30, [sp, #80] // 8-byte Folded Spill
-; CHECK-COMMON-NEXT:    stp s1, s0, [sp, #8] // 8-byte Folded Spill
+; CHECK-COMMON-NEXT:    stp s0, s1, [sp, #8] // 8-byte Folded Spill
 ; CHECK-COMMON-NEXT:    smstop sm
-; CHECK-COMMON-NEXT:    ldp s1, s0, [sp, #8] // 8-byte Folded Reload
+; CHECK-COMMON-NEXT:    ldp s0, s1, [sp, #8] // 8-byte Folded Reload
 ; CHECK-COMMON-NEXT:    bl fmodf
 ; CHECK-COMMON-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-COMMON-NEXT:    smstart sm
@@ -422,9 +422,7 @@ define float @frem_call_sm_compat(float %a, float %b) "aarch64_pstate_sm_compati
 ; CHECK-COMMON-NEXT:    stp x30, x19, [sp, #80] // 16-byte Folded Spill
 ; CHECK-COMMON-NEXT:    stp s0, s1, [sp, #8] // 8-byte Folded Spill
 ; CHECK-COMMON-NEXT:    bl __arm_sme_state
-; CHECK-COMMON-NEXT:    ldp s2, s0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-COMMON-NEXT:    and x19, x0, #0x1
-; CHECK-COMMON-NEXT:    stp s2, s0, [sp, #8] // 8-byte Folded Spill
 ; CHECK-COMMON-NEXT:    tbz w19, #0, .LBB12_2
 ; CHECK-COMMON-NEXT:  // %bb.1:
 ; CHECK-COMMON-NEXT:    smstop sm
diff --git a/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll b/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
index d5bea725b6d14..0c674c5685e00 100644
--- a/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
+++ b/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
@@ -1085,7 +1085,6 @@ define void @dont_coalesce_res_f32(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
-; CHECK-NEXT:    // kill: def $s0 killed $s0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1116,7 +1115,6 @@ define void @dont_coalesce_res_f64(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1151,7 +1149,6 @@ define void @dont_coalesce_res_v1i8(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1183,7 +1180,6 @@ define void @dont_coalesce_res_v1i16(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1215,7 +1211,6 @@ define void @dont_coalesce_res_v1i32(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1247,7 +1242,6 @@ define void @dont_coalesce_res_v1i64(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1311,7 +1305,6 @@ define void @dont_coalesce_res_v1f32(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
@@ -1343,7 +1336,6 @@ define void @dont_coalesce_res_v1f64(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-body-streaming-compatible-interface.ll b/llvm/test/CodeGen/AArch64/sme-streaming-body-streaming-compatible-interface.ll
index 6e262cc0786e4..d67573384ca95 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-body-streaming-compatible-interface.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-body-streaming-compatible-interface.ll
@@ -8,31 +8,27 @@ declare void @streaming_compatible_callee() "aarch64_pstate_sm_compatible";
 define float @sm_body_sm_compatible_simple() "aarch64_pstate_sm_compatible" "aarch64_pstate_sm_body" nounwind {
 ; CHECK-LABEL: sm_body_sm_compatible_simple:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub sp, sp, #96
-; CHECK-NEXT:    stp d15, d14, [sp, #16] // 16-byte Folded Spill
-; CHECK-NEXT:    stp d13, d12, [sp, #32] // 16-byte Folded Spill
-; CHECK-NEXT:    stp d11, d10, [sp, #48] // 16-byte Folded Spill
-; CHECK-NEXT:    stp d9, d8, [sp, #64] // 16-byte Folded Spill
-; CHECK-NEXT:    str x30, [sp, #80] // 8-byte Folded Spill
+; CHECK-NEXT:    stp d15, d14, [sp, #-80]! // 16-byte Folded Spill
+; CHECK-NEXT:    stp d13, d12, [sp, #16] // 16-byte Folded Spill
+; CHECK-NEXT:    stp d11, d10, [sp, #32] // 16-byte Folded Spill
+; CHECK-NEXT:    stp d9, d8, [sp, #48] // 16-byte Folded Spill
+; CHECK-NEXT:    str x30, [sp, #64] // 8-byte Folded Spill
 ; CHECK-NEXT:    bl __arm_sme_state
 ; CHECK-NEXT:    and x8, x0, #0x1
 ; CHECK-NEXT:    tbnz w8, #0, .LBB0_2
 ; CHECK-NEXT:  // %bb.1:
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:  .LBB0_2:
-; CHECK-NEXT:    fmov s0, wzr
-; CHECK-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-NEXT:    tbnz w8, #0, .LBB0_4
 ; CHECK-NEXT:  // %bb.3:
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:  .LBB0_4:
-; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
-; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    ldr x30, [sp, #80] // 8-byte Folded Reload
-; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    add sp, sp, #96
+; CHECK-NEXT:    ldp d9, d8, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT:    fmov s0, wzr
+; CHECK-NEXT:    ldp d11, d10, [sp, #32] // 16-byte Folded Reload
+; CHECK-NEXT:    ldr x30, [sp, #64] // 8-byte Folded Reload
+; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp d15, d14, [sp], #80 // 16-byte Folded Reload
 ; CHECK-NEXT:    ret
   ret float zeroinitializer
 }
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-body.ll b/llvm/test/CodeGen/AArch64/sme-streaming-body.ll
index 08dec228d2f75..cd6d45f54eb9a 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-body.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-body.ll
@@ -247,15 +247,11 @@ define double @call_to_intrinsic_without_chain(double %x) nounwind "aarch64_psta
 ; CHECK-NEXT:    str x30, [sp, #80] // 8-byte Folded Spill
 ; CHECK-NEXT:    str d0, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    str d0, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl cos
 ; CHECK-NEXT:    str d0, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
-; CHECK-NEXT:    str d0, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll b/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
index 6d2abf7e18419..47b24290d3c85 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
@@ -469,11 +469,7 @@ define void @call_to_non_streaming_pass_args(ptr nocapture noundef readnone %ptr
 ; CHECK-NEXT:    mov x9, x0
 ; CHECK-NEXT:    stp s0, s1, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    bl __arm_sme_state
-; CHECK-NEXT:    ldp s4, s0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    and x19, x0, #0x1
-; CHECK-NEXT:    stp s4, s0, [sp, #8] // 8-byte Folded Spill
-; CHECK-NEXT:    ldp d4, d0, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    stp d4, d0, [sp, #16] // 16-byte Folded Spill
 ; CHECK-NEXT:    tbz w19, #0, .LBB10_2
 ; CHECK-NEXT:  // %bb.1: // %entry
 ; CHECK-NEXT:    smstop sm
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-interface.ll b/llvm/test/CodeGen/AArch64/sme-streaming-interface.ll
index de676ac5e0d2e..465fb4667af69 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-interface.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-interface.ll
@@ -405,11 +405,11 @@ define void @call_to_non_streaming_pass_args(ptr nocapture noundef readnone %ptr
 ; CHECK-NEXT:    stp d11, d10, [sp, #64] // 16-byte Folded Spill
 ; CHECK-NEXT:    stp d9, d8, [sp, #80] // 16-byte Folded Spill
 ; CHECK-NEXT:    str x30, [sp, #96] // 8-byte Folded Spill
-; CHECK-NEXT:    stp s1, s0, [sp, #24] // 8-byte Folded Spill
-; CHECK-NEXT:    stp d3, d2, [sp, #8] // 16-byte Folded Spill
+; CHECK-NEXT:    stp d2, d3, [sp, #16] // 16-byte Folded Spill
+; CHECK-NEXT:    stp s0, s1, [sp, #8] // 8-byte Folded Spill
 ; CHECK-NEXT:    smstop sm
-; CHECK-NEXT:    ldp s1, s0, [sp, #24] // 8-byte Folded Reload
-; CHECK-NEXT:    ldp d3, d2, [sp, #8] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp s0, s1, [sp, #8] // 8-byte Folded Reload
+; CHECK-NEXT:    ldp d2, d3, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl bar
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:    ldp d9, d8, [sp, #80] // 16-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-mode-changing-call-disable-stackslot-scavenging.ll b/llvm/test/CodeGen/AArch64/sme-streaming-mode-changing-call-disable-stackslot-scavenging.ll
index cf171f8ef5ed3..45ca7844b0655 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-mode-changing-call-disable-stackslot-scavenging.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-mode-changing-call-disable-stackslot-scavenging.ll
@@ -22,9 +22,9 @@ define void @test_no_stackslot_scavenging(float %f) #0 {
 ; CHECK-NEXT:    stp x30, x24, [sp, #80] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-NEXT:    //APP
 ; CHECK-NEXT:    //NO_APP
-; CHECK-NEXT:    str s0, [sp, #12] // 4-byte Folded Spill
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
 ; CHECK-NEXT:    bl use_f

From b68ff060ab70c44fb617e34451003f34ff062d2b Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov@fb.com>
Date: Mon, 15 Apr 2024 16:16:12 +0200
Subject: [PATCH 433/886] [BOLT][NFC] Remove unused BinaryData::hasNameRegex
 (#88618)

---
 bolt/include/bolt/Core/BinaryData.h | 1 -
 bolt/lib/Core/BinaryData.cpp        | 8 --------
 2 files changed, 9 deletions(-)

diff --git a/bolt/include/bolt/Core/BinaryData.h b/bolt/include/bolt/Core/BinaryData.h
index 495163f1b61aa..8a67b3e73b802 100644
--- a/bolt/include/bolt/Core/BinaryData.h
+++ b/bolt/include/bolt/Core/BinaryData.h
@@ -107,7 +107,6 @@ class BinaryData {
   std::vector<MCSymbol *> &getSymbols() { return Symbols; }
 
   bool hasName(StringRef Name) const;
-  bool hasNameRegex(StringRef Name) const;
   bool nameStartsWith(StringRef Prefix) const;
 
   bool hasSymbol(const MCSymbol *Symbol) const {
diff --git a/bolt/lib/Core/BinaryData.cpp b/bolt/lib/Core/BinaryData.cpp
index 0068a93580042..e9ddf08d8695f 100644
--- a/bolt/lib/Core/BinaryData.cpp
+++ b/bolt/lib/Core/BinaryData.cpp
@@ -55,14 +55,6 @@ bool BinaryData::hasName(StringRef Name) const {
   return false;
 }
 
-bool BinaryData::hasNameRegex(StringRef NameRegex) const {
-  Regex MatchName(NameRegex);
-  for (const MCSymbol *Symbol : Symbols)
-    if (MatchName.match(Symbol->getName()))
-      return true;
-  return false;
-}
-
 bool BinaryData::nameStartsWith(StringRef Prefix) const {
   for (const MCSymbol *Symbol : Symbols)
     if (Symbol->getName().starts_with(Prefix))

From a3ce29f7bb5510131971ac5ccc63132dd48c8dd2 Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Mon, 15 Apr 2024 10:30:00 -0400
Subject: [PATCH 434/886] [libc++][PSTL] Introduce cpu traits (#88134)

Currently, CPU backends in the PSTL are created by defining functions
in the __par_backend namespace. Then, the PSTL includes the CPU backend
that gets configured via CMake and gets those definitions.

This prevents CPU backends from easily co-existing and is a bit
confusing.
To solve this problem, this patch introduces the notion of __cpu_traits,
which is a cheap encapsulation of the basis operations required to
implement a CPU-based PSTL. Different backends can now define their own
tag and coexist, and the CPU-based PSTL will simply use __cpu_traits to
dispatch to the right implementation of e.g. __for_each.

Note that this patch doesn't change the actual implementation of the
backends in any way, it only modifies how that implementation is
accessed
to implement PSTL algorithms.

This patch is a step towards #88131.
---
 libcxx/include/CMakeLists.txt                 |   1 +
 .../__algorithm/pstl_backends/cpu_backend.h   |  47 +-
 .../pstl_backends/cpu_backends/any_of.h       |   9 +-
 .../pstl_backends/cpu_backends/backend.h      |  14 +-
 .../pstl_backends/cpu_backends/fill.h         |   3 +-
 .../pstl_backends/cpu_backends/find_if.h      |  22 +-
 .../pstl_backends/cpu_backends/for_each.h     |   3 +-
 .../pstl_backends/cpu_backends/libdispatch.h  | 457 +++++++++---------
 .../pstl_backends/cpu_backends/merge.h        |   3 +-
 .../pstl_backends/cpu_backends/serial.h       |  98 ++--
 .../pstl_backends/cpu_backends/stable_sort.h  |   3 +-
 .../pstl_backends/cpu_backends/thread.h       |  96 ++--
 .../pstl_backends/cpu_backends/transform.h    |   5 +-
 .../cpu_backends/transform_reduce.h           |  18 +-
 libcxx/include/__pstl/cpu_algos/cpu_traits.h  |  87 ++++
 libcxx/include/module.modulemap               |   2 +
 libcxx/src/pstl/libdispatch.cpp               |   7 +-
 ...pstl.libdispatch.chunk_partitions.pass.cpp |   6 +-
 libcxx/utils/generate_iwyu_mapping.py         |   1 +
 19 files changed, 476 insertions(+), 406 deletions(-)
 create mode 100644 libcxx/include/__pstl/cpu_algos/cpu_traits.h

diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index 1b6c6de082da9..148a51a216ed4 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -594,6 +594,7 @@ set(files
   __numeric/transform_exclusive_scan.h
   __numeric/transform_inclusive_scan.h
   __numeric/transform_reduce.h
+  __pstl/cpu_algos/cpu_traits.h
   __random/bernoulli_distribution.h
   __random/binomial_distribution.h
   __random/cauchy_distribution.h
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backend.h b/libcxx/include/__algorithm/pstl_backends/cpu_backend.h
index 6980ded189ea2..53eae58f96095 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backend.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backend.h
@@ -9,52 +9,6 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
 
-#include <__config>
-
-/*
-
-  // _Functor takes a subrange for [__first, __last) that should be executed in serial
-  template <class _RandomAccessIterator, class _Functor>
-  optional<__empty> __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func);
-
-  template <class _Iterator, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduction>
-  optional<_Tp>
-  __parallel_transform_reduce(_Iterator __first, _Iterator __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduction);
-
-  // Cancel the execution of other jobs - they aren't needed anymore
-  void __cancel_execution();
-
-  template <class _RandomAccessIterator1,
-            class _RandomAccessIterator2,
-            class _RandomAccessIterator3,
-            class _Compare,
-            class _LeafMerge>
-  optional<void> __parallel_merge(
-      _RandomAccessIterator1 __first1,
-      _RandomAccessIterator1 __last1,
-      _RandomAccessIterator2 __first2,
-      _RandomAccessIterator2 __last2,
-      _RandomAccessIterator3 __outit,
-      _Compare __comp,
-      _LeafMerge __leaf_merge);
-
-  template <class _RandomAccessIterator, class _Comp, class _LeafSort>
-  void __parallel_stable_sort(_RandomAccessIterator __first,
-                              _RandomAccessIterator __last,
-                              _Comp __comp,
-                              _LeafSort __leaf_sort);
-
-  TODO: Document the parallel backend
-
-Exception handling
-==================
-
-CPU backends are expected to report errors (i.e. failure to allocate) by returning a disengaged `optional` from their
-implementation. Exceptions shouldn't be used to report an internal failure-to-allocate, since all exceptions are turned
-into a program termination at the front-end level. When a backend returns a disengaged `optional` to the frontend, the
-frontend will turn that into a call to `std::__throw_bad_alloc();` to report the internal failure to the user.
-*/
-
 #include <__algorithm/pstl_backends/cpu_backends/any_of.h>
 #include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__algorithm/pstl_backends/cpu_backends/fill.h>
@@ -64,5 +18,6 @@ frontend will turn that into a call to `std::__throw_bad_alloc();` to report the
 #include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
 #include <__algorithm/pstl_backends/cpu_backends/transform.h>
 #include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+#include <__config>
 
 #endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
index 13dff80086e72..be5e54f3fa5c8 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
@@ -17,6 +17,7 @@
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/concepts.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
 #include <__utility/pair.h>
@@ -30,13 +31,13 @@ _LIBCPP_PUSH_MACROS
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-template <class _Index, class _Brick>
+template <class _Backend, class _Index, class _Brick>
 _LIBCPP_HIDE_FROM_ABI optional<bool> __parallel_or(_Index __first, _Index __last, _Brick __f) {
   std::atomic<bool> __found(false);
-  auto __ret = __par_backend::__parallel_for(__first, __last, [__f, &__found](_Index __i, _Index __j) {
+  auto __ret = __pstl::__cpu_traits<_Backend>::__parallel_for(__first, __last, [__f, &__found](_Index __i, _Index __j) {
     if (!__found.load(std::memory_order_relaxed) && __f(__i, __j)) {
       __found.store(true, std::memory_order_relaxed);
-      __par_backend::__cancel_execution();
+      __pstl::__cpu_traits<_Backend>::__cancel_execution();
     }
   });
   if (!__ret)
@@ -74,7 +75,7 @@ _LIBCPP_HIDE_FROM_ABI optional<bool>
 __pstl_any_of(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return std::__parallel_or(
+    return std::__parallel_or<__cpu_backend_tag>(
         __first, __last, [&__pred](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           auto __res = std::__pstl_any_of<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __pred);
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
index ea2210a4a7adb..cb9425862a2b0 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
@@ -26,16 +26,20 @@
 #  pragma GCC system_header
 #endif
 
-#if _LIBCPP_STD_VER >= 17
+#if !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-struct __cpu_backend_tag {};
-
-inline constexpr size_t __lane_size = 64;
+#  if defined(_LIBCPP_PSTL_CPU_BACKEND_SERIAL)
+using __cpu_backend_tag = __pstl::__serial_backend_tag;
+#  elif defined(_LIBCPP_PSTL_CPU_BACKEND_THREAD)
+using __cpu_backend_tag = __pstl::__std_thread_backend_tag;
+#  elif defined(_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH)
+using __cpu_backend_tag = __pstl::__libdispatch_backend_tag;
+#  endif
 
 _LIBCPP_END_NAMESPACE_STD
 
-#endif // _LIBCPP_STD_VER >= 17
+#endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && && _LIBCPP_STD_VER >= 17
 
 #endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_BACKEND_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
index 64babe9fd2bda..49a32f6c5ce55 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
@@ -13,6 +13,7 @@
 #include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
 #include <optional>
@@ -39,7 +40,7 @@ _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_fill(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return __par_backend::__parallel_for(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
         __first, __last, [&__value](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           [[maybe_unused]] auto __res = std::__pstl_fill<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __value);
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
index 170470e4fb7ed..11a5668bf25af 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
@@ -16,6 +16,7 @@
 #include <__functional/operations.h>
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
 #include <__utility/pair.h>
@@ -33,7 +34,7 @@ _LIBCPP_PUSH_MACROS
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-template <class _Index, class _Brick, class _Compare>
+template <class _Backend, class _Index, class _Brick, class _Compare>
 _LIBCPP_HIDE_FROM_ABI optional<_Index>
 __parallel_find(_Index __first, _Index __last, _Brick __f, _Compare __comp, bool __b_first) {
   typedef typename std::iterator_traits<_Index>::difference_type _DifferenceType;
@@ -41,8 +42,8 @@ __parallel_find(_Index __first, _Index __last, _Brick __f, _Compare __comp, bool
   _DifferenceType __initial_dist = __b_first ? __n : -1;
   std::atomic<_DifferenceType> __extremum(__initial_dist);
   // TODO: find out what is better here: parallel_for or parallel_reduce
-  auto __res =
-      __par_backend::__parallel_for(__first, __last, [__comp, __f, __first, &__extremum](_Index __i, _Index __j) {
+  auto __res = __pstl::__cpu_traits<_Backend>::__parallel_for(
+      __first, __last, [__comp, __f, __first, &__extremum](_Index __i, _Index __j) {
         // See "Reducing Contention Through Priority Updates", PPoPP '13, for discussion of
         // why using a shared variable scales fairly well in this situation.
         if (__comp(__i - __first, __extremum)) {
@@ -61,12 +62,12 @@ __parallel_find(_Index __first, _Index __last, _Brick __f, _Compare __comp, bool
   return __extremum.load() != __initial_dist ? __first + __extremum.load() : __last;
 }
 
-template <class _Index, class _DifferenceType, class _Compare>
+template <class _Backend, class _Index, class _DifferenceType, class _Compare>
 _LIBCPP_HIDE_FROM_ABI _Index
 __simd_first(_Index __first, _DifferenceType __begin, _DifferenceType __end, _Compare __comp) noexcept {
   // Experiments show good block sizes like this
-  const _DifferenceType __block_size                        = 8;
-  alignas(__lane_size) _DifferenceType __lane[__block_size] = {0};
+  const _DifferenceType __block_size                                                        = 8;
+  alignas(__pstl::__cpu_traits<_Backend>::__lane_size) _DifferenceType __lane[__block_size] = {0};
   while (__end - __begin >= __block_size) {
     _DifferenceType __found = 0;
     _PSTL_PRAGMA_SIMD_REDUCTION(| : __found) for (_DifferenceType __i = __begin; __i < __begin + __block_size; ++__i) {
@@ -102,7 +103,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardIterator>
 __pstl_find_if(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return std::__parallel_find(
+    return std::__parallel_find<__cpu_backend_tag>(
         __first,
         __last,
         [&__pred](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
@@ -116,9 +117,10 @@ __pstl_find_if(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __l
   } else if constexpr (__is_unsequenced_execution_policy_v<_ExecutionPolicy> &&
                        __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
     using __diff_t = __iter_diff_t<_ForwardIterator>;
-    return std::__simd_first(__first, __diff_t(0), __last - __first, [&__pred](_ForwardIterator __iter, __diff_t __i) {
-      return __pred(__iter[__i]);
-    });
+    return std::__simd_first<__cpu_backend_tag>(
+        __first, __diff_t(0), __last - __first, [&__pred](_ForwardIterator __iter, __diff_t __i) {
+          return __pred(__iter[__i]);
+        });
   } else {
     return std::find_if(__first, __last, __pred);
   }
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
index 81fd4526b8dbf..1667ec0f0c4f4 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
@@ -13,6 +13,7 @@
 #include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
 #include <optional>
@@ -39,7 +40,7 @@ _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_for_each(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, _Functor __func) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return std::__par_backend::__parallel_for(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
         __first, __last, [__func](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           [[maybe_unused]] auto __res = std::__pstl_for_each<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __func);
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
index e885e7f225172..8757f24968037 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
@@ -23,6 +23,7 @@
 #include <__memory/construct_at.h>
 #include <__memory/unique_ptr.h>
 #include <__numeric/reduce.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/exception_guard.h>
 #include <__utility/move.h>
@@ -37,10 +38,11 @@ _LIBCPP_PUSH_MACROS
 #if !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
 _LIBCPP_BEGIN_NAMESPACE_STD
+namespace __pstl {
 
-namespace __par_backend {
-inline namespace __libdispatch {
+struct __libdispatch_backend_tag {};
 
+namespace __libdispatch {
 // ::dispatch_apply is marked as __attribute__((nothrow)) because it doesn't let exceptions propagate, and neither do
 // we.
 // TODO: Do we want to add [[_Clang::__callback__(__func, __context, __)]]?
@@ -77,267 +79,270 @@ __dispatch_parallel_for(__chunk_partitions __partitions, _RandomAccessIterator _
 
   return __empty{};
 }
+} // namespace __libdispatch
 
-template <class _RandomAccessIterator, class _Functor>
-_LIBCPP_HIDE_FROM_ABI optional<__empty>
-__parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func) {
-  return __libdispatch::__dispatch_parallel_for(
-      __libdispatch::__partition_chunks(__last - __first), std::move(__first), std::move(__func));
-}
-
-template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIteratorOut>
-struct __merge_range {
-  __merge_range(_RandomAccessIterator1 __mid1, _RandomAccessIterator2 __mid2, _RandomAccessIteratorOut __result)
-      : __mid1_(__mid1), __mid2_(__mid2), __result_(__result) {}
+template <>
+struct __cpu_traits<__libdispatch_backend_tag> {
+  template <class _RandomAccessIterator, class _Functor>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func) {
+    return __libdispatch::__dispatch_parallel_for(
+        __libdispatch::__partition_chunks(__last - __first), std::move(__first), std::move(__func));
+  }
 
-  _RandomAccessIterator1 __mid1_;
-  _RandomAccessIterator2 __mid2_;
-  _RandomAccessIteratorOut __result_;
-};
+  template <class _RandomAccessIterator1, class _RandomAccessIterator2, class _RandomAccessIteratorOut>
+  struct __merge_range {
+    __merge_range(_RandomAccessIterator1 __mid1, _RandomAccessIterator2 __mid2, _RandomAccessIteratorOut __result)
+        : __mid1_(__mid1), __mid2_(__mid2), __result_(__result) {}
 
-template <typename _RandomAccessIterator1,
-          typename _RandomAccessIterator2,
-          typename _RandomAccessIterator3,
-          typename _Compare,
-          typename _LeafMerge>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_merge(
-    _RandomAccessIterator1 __first1,
-    _RandomAccessIterator1 __last1,
-    _RandomAccessIterator2 __first2,
-    _RandomAccessIterator2 __last2,
-    _RandomAccessIterator3 __result,
-    _Compare __comp,
-    _LeafMerge __leaf_merge) noexcept {
-  __chunk_partitions __partitions =
-      __libdispatch::__partition_chunks(std::max<ptrdiff_t>(__last1 - __first1, __last2 - __first2));
-
-  if (__partitions.__chunk_count_ == 0)
-    return __empty{};
+    _RandomAccessIterator1 __mid1_;
+    _RandomAccessIterator2 __mid2_;
+    _RandomAccessIteratorOut __result_;
+  };
 
-  if (__partitions.__chunk_count_ == 1) {
-    __leaf_merge(__first1, __last1, __first2, __last2, __result, __comp);
-    return __empty{};
-  }
+  template <typename _RandomAccessIterator1,
+            typename _RandomAccessIterator2,
+            typename _RandomAccessIterator3,
+            typename _Compare,
+            typename _LeafMerge>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
+      _RandomAccessIterator1 __first1,
+      _RandomAccessIterator1 __last1,
+      _RandomAccessIterator2 __first2,
+      _RandomAccessIterator2 __last2,
+      _RandomAccessIterator3 __result,
+      _Compare __comp,
+      _LeafMerge __leaf_merge) noexcept {
+    __libdispatch::__chunk_partitions __partitions =
+        __libdispatch::__partition_chunks(std::max<ptrdiff_t>(__last1 - __first1, __last2 - __first2));
+
+    if (__partitions.__chunk_count_ == 0)
+      return __empty{};
+
+    if (__partitions.__chunk_count_ == 1) {
+      __leaf_merge(__first1, __last1, __first2, __last2, __result, __comp);
+      return __empty{};
+    }
 
-  using __merge_range_t = __merge_range<_RandomAccessIterator1, _RandomAccessIterator2, _RandomAccessIterator3>;
-  auto const __n_ranges = __partitions.__chunk_count_ + 1;
+    using __merge_range_t = __merge_range<_RandomAccessIterator1, _RandomAccessIterator2, _RandomAccessIterator3>;
+    auto const __n_ranges = __partitions.__chunk_count_ + 1;
 
-  // TODO: use __uninitialized_buffer
-  auto __destroy = [=](__merge_range_t* __ptr) {
-    std::destroy_n(__ptr, __n_ranges);
-    std::allocator<__merge_range_t>().deallocate(__ptr, __n_ranges);
-  };
+    // TODO: use __uninitialized_buffer
+    auto __destroy = [=](__merge_range_t* __ptr) {
+      std::destroy_n(__ptr, __n_ranges);
+      std::allocator<__merge_range_t>().deallocate(__ptr, __n_ranges);
+    };
 
-  unique_ptr<__merge_range_t[], decltype(__destroy)> __ranges(
-      [&]() -> __merge_range_t* {
+    unique_ptr<__merge_range_t[], decltype(__destroy)> __ranges(
+        [&]() -> __merge_range_t* {
 #  ifndef _LIBCPP_HAS_NO_EXCEPTIONS
-        try {
+          try {
 #  endif
-          return std::allocator<__merge_range_t>().allocate(__n_ranges);
+            return std::allocator<__merge_range_t>().allocate(__n_ranges);
 #  ifndef _LIBCPP_HAS_NO_EXCEPTIONS
-        } catch (const std::bad_alloc&) {
-          return nullptr;
-        }
+          } catch (const std::bad_alloc&) {
+            return nullptr;
+          }
 #  endif
-      }(),
-      __destroy);
-
-  if (!__ranges)
-    return nullopt;
+        }(),
+        __destroy);
+
+    if (!__ranges)
+      return nullopt;
+
+    // TODO: Improve the case where the smaller range is merged into just a few (or even one) chunks of the larger case
+    __merge_range_t* __r = __ranges.get();
+    std::__construct_at(__r++, __first1, __first2, __result);
+
+    bool __iterate_first_range = __last1 - __first1 > __last2 - __first2;
+
+    auto __compute_chunk = [&](size_t __chunk_size) -> __merge_range_t {
+      auto [__mid1, __mid2] = [&] {
+        if (__iterate_first_range) {
+          auto __m1 = __first1 + __chunk_size;
+          auto __m2 = std::lower_bound(__first2, __last2, __m1[-1], __comp);
+          return std::make_pair(__m1, __m2);
+        } else {
+          auto __m2 = __first2 + __chunk_size;
+          auto __m1 = std::lower_bound(__first1, __last1, __m2[-1], __comp);
+          return std::make_pair(__m1, __m2);
+        }
+      }();
 
-  // TODO: Improve the case where the smaller range is merged into just a few (or even one) chunks of the larger case
-  __merge_range_t* __r = __ranges.get();
-  std::__construct_at(__r++, __first1, __first2, __result);
+      __result += (__mid1 - __first1) + (__mid2 - __first2);
+      __first1 = __mid1;
+      __first2 = __mid2;
+      return {std::move(__mid1), std::move(__mid2), __result};
+    };
 
-  bool __iterate_first_range = __last1 - __first1 > __last2 - __first2;
+    // handle first chunk
+    std::__construct_at(__r++, __compute_chunk(__partitions.__first_chunk_size_));
 
-  auto __compute_chunk = [&](size_t __chunk_size) -> __merge_range_t {
-    auto [__mid1, __mid2] = [&] {
-      if (__iterate_first_range) {
-        auto __m1 = __first1 + __chunk_size;
-        auto __m2 = std::lower_bound(__first2, __last2, __m1[-1], __comp);
-        return std::make_pair(__m1, __m2);
-      } else {
-        auto __m2 = __first2 + __chunk_size;
-        auto __m1 = std::lower_bound(__first1, __last1, __m2[-1], __comp);
-        return std::make_pair(__m1, __m2);
-      }
-    }();
+    // handle 2 -> N - 1 chunks
+    for (ptrdiff_t __i = 0; __i != __partitions.__chunk_count_ - 2; ++__i)
+      std::__construct_at(__r++, __compute_chunk(__partitions.__chunk_size_));
 
-    __result += (__mid1 - __first1) + (__mid2 - __first2);
-    __first1 = __mid1;
-    __first2 = __mid2;
-    return {std::move(__mid1), std::move(__mid2), __result};
-  };
+    // handle last chunk
+    std::__construct_at(__r, __last1, __last2, __result);
 
-  // handle first chunk
-  std::__construct_at(__r++, __compute_chunk(__partitions.__first_chunk_size_));
-
-  // handle 2 -> N - 1 chunks
-  for (ptrdiff_t __i = 0; __i != __partitions.__chunk_count_ - 2; ++__i)
-    std::__construct_at(__r++, __compute_chunk(__partitions.__chunk_size_));
-
-  // handle last chunk
-  std::__construct_at(__r, __last1, __last2, __result);
-
-  __libdispatch::__dispatch_apply(__partitions.__chunk_count_, [&](size_t __index) {
-    auto __first_iters = __ranges[__index];
-    auto __last_iters  = __ranges[__index + 1];
-    __leaf_merge(
-        __first_iters.__mid1_,
-        __last_iters.__mid1_,
-        __first_iters.__mid2_,
-        __last_iters.__mid2_,
-        __first_iters.__result_,
-        __comp);
-  });
+    __libdispatch::__dispatch_apply(__partitions.__chunk_count_, [&](size_t __index) {
+      auto __first_iters = __ranges[__index];
+      auto __last_iters  = __ranges[__index + 1];
+      __leaf_merge(
+          __first_iters.__mid1_,
+          __last_iters.__mid1_,
+          __first_iters.__mid2_,
+          __last_iters.__mid2_,
+          __first_iters.__result_,
+          __comp);
+    });
 
-  return __empty{};
-}
+    return __empty{};
+  }
 
-template <class _RandomAccessIterator, class _Transform, class _Value, class _Combiner, class _Reduction>
-_LIBCPP_HIDE_FROM_ABI optional<_Value> __parallel_transform_reduce(
-    _RandomAccessIterator __first,
-    _RandomAccessIterator __last,
-    _Transform __transform,
-    _Value __init,
-    _Combiner __combiner,
-    _Reduction __reduction) {
-  if (__first == __last)
-    return __init;
-
-  auto __partitions = __libdispatch::__partition_chunks(__last - __first);
-
-  auto __destroy = [__count = __partitions.__chunk_count_](_Value* __ptr) {
-    std::destroy_n(__ptr, __count);
-    std::allocator<_Value>().deallocate(__ptr, __count);
-  };
+  template <class _RandomAccessIterator, class _Transform, class _Value, class _Combiner, class _Reduction>
+  _LIBCPP_HIDE_FROM_ABI static optional<_Value> __parallel_transform_reduce(
+      _RandomAccessIterator __first,
+      _RandomAccessIterator __last,
+      _Transform __transform,
+      _Value __init,
+      _Combiner __combiner,
+      _Reduction __reduction) {
+    if (__first == __last)
+      return __init;
+
+    auto __partitions = __libdispatch::__partition_chunks(__last - __first);
+
+    auto __destroy = [__count = __partitions.__chunk_count_](_Value* __ptr) {
+      std::destroy_n(__ptr, __count);
+      std::allocator<_Value>().deallocate(__ptr, __count);
+    };
 
-  // TODO: use __uninitialized_buffer
-  // TODO: allocate one element per worker instead of one element per chunk
-  unique_ptr<_Value[], decltype(__destroy)> __values(
-      std::allocator<_Value>().allocate(__partitions.__chunk_count_), __destroy);
+    // TODO: use __uninitialized_buffer
+    // TODO: allocate one element per worker instead of one element per chunk
+    unique_ptr<_Value[], decltype(__destroy)> __values(
+        std::allocator<_Value>().allocate(__partitions.__chunk_count_), __destroy);
+
+    // __dispatch_apply is noexcept
+    __libdispatch::__dispatch_apply(__partitions.__chunk_count_, [&](size_t __chunk) {
+      auto __this_chunk_size = __chunk == 0 ? __partitions.__first_chunk_size_ : __partitions.__chunk_size_;
+      auto __index           = __chunk == 0 ? 0
+                                            : (__chunk * __partitions.__chunk_size_) +
+                                        (__partitions.__first_chunk_size_ - __partitions.__chunk_size_);
+      if (__this_chunk_size != 1) {
+        std::__construct_at(
+            __values.get() + __chunk,
+            __reduction(__first + __index + 2,
+                        __first + __index + __this_chunk_size,
+                        __combiner(__transform(__first + __index), __transform(__first + __index + 1))));
+      } else {
+        std::__construct_at(__values.get() + __chunk, __transform(__first + __index));
+      }
+    });
 
-  // __dispatch_apply is noexcept
-  __libdispatch::__dispatch_apply(__partitions.__chunk_count_, [&](size_t __chunk) {
-    auto __this_chunk_size = __chunk == 0 ? __partitions.__first_chunk_size_ : __partitions.__chunk_size_;
-    auto __index =
-        __chunk == 0
-            ? 0
-            : (__chunk * __partitions.__chunk_size_) + (__partitions.__first_chunk_size_ - __partitions.__chunk_size_);
-    if (__this_chunk_size != 1) {
-      std::__construct_at(
-          __values.get() + __chunk,
-          __reduction(__first + __index + 2,
-                      __first + __index + __this_chunk_size,
-                      __combiner(__transform(__first + __index), __transform(__first + __index + 1))));
-    } else {
-      std::__construct_at(__values.get() + __chunk, __transform(__first + __index));
-    }
-  });
+    return std::reduce(
+        std::make_move_iterator(__values.get()),
+        std::make_move_iterator(__values.get() + __partitions.__chunk_count_),
+        std::move(__init),
+        __combiner);
+  }
 
-  return std::reduce(
-      std::make_move_iterator(__values.get()),
-      std::make_move_iterator(__values.get() + __partitions.__chunk_count_),
-      std::move(__init),
-      __combiner);
-}
+  template <class _RandomAccessIterator, class _Comp, class _LeafSort>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
+      _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp, _LeafSort __leaf_sort) {
+    const auto __size = __last - __first;
+    auto __partitions = __libdispatch::__partition_chunks(__size);
 
-template <class _RandomAccessIterator, class _Comp, class _LeafSort>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_stable_sort(
-    _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp, _LeafSort __leaf_sort) {
-  const auto __size = __last - __first;
-  auto __partitions = __libdispatch::__partition_chunks(__size);
+    if (__partitions.__chunk_count_ == 0)
+      return __empty{};
 
-  if (__partitions.__chunk_count_ == 0)
-    return __empty{};
+    if (__partitions.__chunk_count_ == 1) {
+      __leaf_sort(__first, __last, __comp);
+      return __empty{};
+    }
 
-  if (__partitions.__chunk_count_ == 1) {
-    __leaf_sort(__first, __last, __comp);
-    return __empty{};
-  }
+    using _Value = __iter_value_type<_RandomAccessIterator>;
 
-  using _Value = __iter_value_type<_RandomAccessIterator>;
+    auto __destroy = [__size](_Value* __ptr) {
+      std::destroy_n(__ptr, __size);
+      std::allocator<_Value>().deallocate(__ptr, __size);
+    };
 
-  auto __destroy = [__size](_Value* __ptr) {
-    std::destroy_n(__ptr, __size);
-    std::allocator<_Value>().deallocate(__ptr, __size);
-  };
+    // TODO: use __uninitialized_buffer
+    unique_ptr<_Value[], decltype(__destroy)> __values(std::allocator<_Value>().allocate(__size), __destroy);
 
-  // TODO: use __uninitialized_buffer
-  unique_ptr<_Value[], decltype(__destroy)> __values(std::allocator<_Value>().allocate(__size), __destroy);
+    // Initialize all elements to a moved-from state
+    // TODO: Don't do this - this can be done in the first merge - see https://llvm.org/PR63928
+    std::__construct_at(__values.get(), std::move(*__first));
+    for (__iter_diff_t<_RandomAccessIterator> __i = 1; __i != __size; ++__i) {
+      std::__construct_at(__values.get() + __i, std::move(__values.get()[__i - 1]));
+    }
+    *__first = std::move(__values.get()[__size - 1]);
+
+    __libdispatch::__dispatch_parallel_for(
+        __partitions,
+        __first,
+        [&__leaf_sort, &__comp](_RandomAccessIterator __chunk_first, _RandomAccessIterator __chunk_last) {
+          __leaf_sort(std::move(__chunk_first), std::move(__chunk_last), __comp);
+        });
+
+    bool __objects_are_in_buffer = false;
+    do {
+      const auto __old_chunk_size = __partitions.__chunk_size_;
+      if (__partitions.__chunk_count_ % 2 == 1) {
+        auto __inplace_merge_chunks = [&__comp, &__partitions](auto __first_chunk_begin) {
+          std::inplace_merge(
+              __first_chunk_begin,
+              __first_chunk_begin + __partitions.__first_chunk_size_,
+              __first_chunk_begin + __partitions.__first_chunk_size_ + __partitions.__chunk_size_,
+              __comp);
+        };
+        if (__objects_are_in_buffer)
+          __inplace_merge_chunks(__values.get());
+        else
+          __inplace_merge_chunks(__first);
+        __partitions.__first_chunk_size_ += 2 * __partitions.__chunk_size_;
+      } else {
+        __partitions.__first_chunk_size_ += __partitions.__chunk_size_;
+      }
 
-  // Initialize all elements to a moved-from state
-  // TODO: Don't do this - this can be done in the first merge - see https://llvm.org/PR63928
-  std::__construct_at(__values.get(), std::move(*__first));
-  for (__iter_diff_t<_RandomAccessIterator> __i = 1; __i != __size; ++__i) {
-    std::__construct_at(__values.get() + __i, std::move(__values.get()[__i - 1]));
-  }
-  *__first = std::move(__values.get()[__size - 1]);
-
-  __libdispatch::__dispatch_parallel_for(
-      __partitions,
-      __first,
-      [&__leaf_sort, &__comp](_RandomAccessIterator __chunk_first, _RandomAccessIterator __chunk_last) {
-        __leaf_sort(std::move(__chunk_first), std::move(__chunk_last), __comp);
-      });
-
-  bool __objects_are_in_buffer = false;
-  do {
-    const auto __old_chunk_size = __partitions.__chunk_size_;
-    if (__partitions.__chunk_count_ % 2 == 1) {
-      auto __inplace_merge_chunks = [&__comp, &__partitions](auto __first_chunk_begin) {
-        std::inplace_merge(
-            __first_chunk_begin,
-            __first_chunk_begin + __partitions.__first_chunk_size_,
-            __first_chunk_begin + __partitions.__first_chunk_size_ + __partitions.__chunk_size_,
-            __comp);
+      __partitions.__chunk_size_ *= 2;
+      __partitions.__chunk_count_ /= 2;
+
+      auto __merge_chunks = [__partitions, __old_chunk_size, &__comp](auto __from_first, auto __to_first) {
+        __libdispatch::__dispatch_parallel_for(
+            __partitions,
+            __from_first,
+            [__old_chunk_size, &__from_first, &__to_first, &__comp](auto __chunk_first, auto __chunk_last) {
+              std::merge(std::make_move_iterator(__chunk_first),
+                         std::make_move_iterator(__chunk_last - __old_chunk_size),
+                         std::make_move_iterator(__chunk_last - __old_chunk_size),
+                         std::make_move_iterator(__chunk_last),
+                         __to_first + (__chunk_first - __from_first),
+                         __comp);
+            });
       };
+
       if (__objects_are_in_buffer)
-        __inplace_merge_chunks(__values.get());
+        __merge_chunks(__values.get(), __first);
       else
-        __inplace_merge_chunks(__first);
-      __partitions.__first_chunk_size_ += 2 * __partitions.__chunk_size_;
-    } else {
-      __partitions.__first_chunk_size_ += __partitions.__chunk_size_;
-    }
-
-    __partitions.__chunk_size_ *= 2;
-    __partitions.__chunk_count_ /= 2;
-
-    auto __merge_chunks = [__partitions, __old_chunk_size, &__comp](auto __from_first, auto __to_first) {
-      __libdispatch::__dispatch_parallel_for(
-          __partitions,
-          __from_first,
-          [__old_chunk_size, &__from_first, &__to_first, &__comp](auto __chunk_first, auto __chunk_last) {
-            std::merge(std::make_move_iterator(__chunk_first),
-                       std::make_move_iterator(__chunk_last - __old_chunk_size),
-                       std::make_move_iterator(__chunk_last - __old_chunk_size),
-                       std::make_move_iterator(__chunk_last),
-                       __to_first + (__chunk_first - __from_first),
-                       __comp);
-          });
-    };
+        __merge_chunks(__first, __values.get());
+      __objects_are_in_buffer = !__objects_are_in_buffer;
+    } while (__partitions.__chunk_count_ > 1);
 
-    if (__objects_are_in_buffer)
-      __merge_chunks(__values.get(), __first);
-    else
-      __merge_chunks(__first, __values.get());
-    __objects_are_in_buffer = !__objects_are_in_buffer;
-  } while (__partitions.__chunk_count_ > 1);
+    if (__objects_are_in_buffer) {
+      std::move(__values.get(), __values.get() + __size, __first);
+    }
 
-  if (__objects_are_in_buffer) {
-    std::move(__values.get(), __values.get() + __size, __first);
+    return __empty{};
   }
 
-  return __empty{};
-}
+  _LIBCPP_HIDE_FROM_ABI static void __cancel_execution() {}
 
-_LIBCPP_HIDE_FROM_ABI inline void __cancel_execution() {}
-
-} // namespace __libdispatch
-} // namespace __par_backend
+  static constexpr size_t __lane_size = 64;
+};
 
+} // namespace __pstl
 _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
index b0db70f58b2ef..d034447904872 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
@@ -13,6 +13,7 @@
 #include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
 #include <optional>
@@ -45,7 +46,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_merge(
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    auto __res = __par_backend::__parallel_merge(
+    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_merge(
         __first1,
         __last1,
         __first2,
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
index afcc7ffb26613..c3d2905daed17 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
@@ -11,6 +11,7 @@
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_SERIAL_H
 
 #include <__config>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/move.h>
 #include <cstddef>
@@ -26,54 +27,55 @@ _LIBCPP_PUSH_MACROS
 #  include <__undef_macros>
 
 _LIBCPP_BEGIN_NAMESPACE_STD
-
-namespace __par_backend {
-inline namespace __serial_cpu_backend {
-
-template <class _RandomAccessIterator, class _Fp>
-_LIBCPP_HIDE_FROM_ABI optional<__empty>
-__parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
-  __f(__first, __last);
-  return __empty{};
-}
-
-template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
-_LIBCPP_HIDE_FROM_ABI optional<_Tp>
-__parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
-  return __reduce(std::move(__first), std::move(__last), std::move(__init));
-}
-
-template <class _RandomAccessIterator, class _Compare, class _LeafSort>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_stable_sort(
-    _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
-  __leaf_sort(__first, __last, __comp);
-  return __empty{};
-}
-
-_LIBCPP_HIDE_FROM_ABI inline void __cancel_execution() {}
-
-template <class _RandomAccessIterator1,
-          class _RandomAccessIterator2,
-          class _RandomAccessIterator3,
-          class _Compare,
-          class _LeafMerge>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_merge(
-    _RandomAccessIterator1 __first1,
-    _RandomAccessIterator1 __last1,
-    _RandomAccessIterator2 __first2,
-    _RandomAccessIterator2 __last2,
-    _RandomAccessIterator3 __outit,
-    _Compare __comp,
-    _LeafMerge __leaf_merge) {
-  __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
-  return __empty{};
-}
-
-// TODO: Complete this list
-
-} // namespace __serial_cpu_backend
-} // namespace __par_backend
-
+namespace __pstl {
+
+struct __serial_backend_tag {};
+
+template <>
+struct __cpu_traits<__serial_backend_tag> {
+  template <class _RandomAccessIterator, class _Fp>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
+    __f(__first, __last);
+    return __empty{};
+  }
+
+  template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
+  _LIBCPP_HIDE_FROM_ABI static optional<_Tp>
+  __parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
+    return __reduce(std::move(__first), std::move(__last), std::move(__init));
+  }
+
+  template <class _RandomAccessIterator, class _Compare, class _LeafSort>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
+      _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
+    __leaf_sort(__first, __last, __comp);
+    return __empty{};
+  }
+
+  _LIBCPP_HIDE_FROM_ABI static void __cancel_execution() {}
+
+  template <class _RandomAccessIterator1,
+            class _RandomAccessIterator2,
+            class _RandomAccessIterator3,
+            class _Compare,
+            class _LeafMerge>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
+      _RandomAccessIterator1 __first1,
+      _RandomAccessIterator1 __last1,
+      _RandomAccessIterator2 __first2,
+      _RandomAccessIterator2 __last2,
+      _RandomAccessIterator3 __outit,
+      _Compare __comp,
+      _LeafMerge __leaf_merge) {
+    __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
+    return __empty{};
+  }
+
+  static constexpr size_t __lane_size = 64;
+};
+
+} // namespace __pstl
 _LIBCPP_END_NAMESPACE_STD
 
 _LIBCPP_POP_MACROS
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
index 34c423586c4b7..ebfa0fc69147d 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
@@ -12,6 +12,7 @@
 #include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__algorithm/stable_sort.h>
 #include <__config>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
 #include <optional>
@@ -28,7 +29,7 @@ template <class _ExecutionPolicy, class _RandomAccessIterator, class _Comp>
 _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_stable_sort(__cpu_backend_tag, _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy>) {
-    return __par_backend::__parallel_stable_sort(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_stable_sort(
         __first, __last, __comp, [](_RandomAccessIterator __g_first, _RandomAccessIterator __g_last, _Comp __g_comp) {
           std::stable_sort(__g_first, __g_last, __g_comp);
         });
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
index eb11a961b760c..8d1cb221c3d82 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
@@ -11,6 +11,7 @@
 
 #include <__assert>
 #include <__config>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/move.h>
 #include <cstddef>
@@ -29,52 +30,55 @@ _LIBCPP_PUSH_MACROS
 // by a proper implementation once the PSTL implementation is somewhat stable.
 
 _LIBCPP_BEGIN_NAMESPACE_STD
-
-namespace __par_backend {
-inline namespace __thread_cpu_backend {
-
-template <class _RandomAccessIterator, class _Fp>
-_LIBCPP_HIDE_FROM_ABI optional<__empty>
-__parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
-  __f(__first, __last);
-  return __empty{};
-}
-
-template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
-_LIBCPP_HIDE_FROM_ABI optional<_Tp>
-__parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
-  return __reduce(std::move(__first), std::move(__last), std::move(__init));
-}
-
-template <class _RandomAccessIterator, class _Compare, class _LeafSort>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_stable_sort(
-    _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
-  __leaf_sort(__first, __last, __comp);
-  return __empty{};
-}
-
-_LIBCPP_HIDE_FROM_ABI inline void __cancel_execution() {}
-
-template <class _RandomAccessIterator1,
-          class _RandomAccessIterator2,
-          class _RandomAccessIterator3,
-          class _Compare,
-          class _LeafMerge>
-_LIBCPP_HIDE_FROM_ABI optional<__empty> __parallel_merge(
-    _RandomAccessIterator1 __first1,
-    _RandomAccessIterator1 __last1,
-    _RandomAccessIterator2 __first2,
-    _RandomAccessIterator2 __last2,
-    _RandomAccessIterator3 __outit,
-    _Compare __comp,
-    _LeafMerge __leaf_merge) {
-  __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
-  return __empty{};
-}
-
-} // namespace __thread_cpu_backend
-} // namespace __par_backend
-
+namespace __pstl {
+
+struct __std_thread_backend_tag {};
+
+template <>
+struct __cpu_traits<__std_thread_backend_tag> {
+  template <class _RandomAccessIterator, class _Fp>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
+    __f(__first, __last);
+    return __empty{};
+  }
+
+  template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
+  _LIBCPP_HIDE_FROM_ABI static optional<_Tp>
+  __parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
+    return __reduce(std::move(__first), std::move(__last), std::move(__init));
+  }
+
+  template <class _RandomAccessIterator, class _Compare, class _LeafSort>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
+      _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
+    __leaf_sort(__first, __last, __comp);
+    return __empty{};
+  }
+
+  _LIBCPP_HIDE_FROM_ABI static void __cancel_execution() {}
+
+  template <class _RandomAccessIterator1,
+            class _RandomAccessIterator2,
+            class _RandomAccessIterator3,
+            class _Compare,
+            class _LeafMerge>
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
+      _RandomAccessIterator1 __first1,
+      _RandomAccessIterator1 __last1,
+      _RandomAccessIterator2 __first2,
+      _RandomAccessIterator2 __last2,
+      _RandomAccessIterator3 __outit,
+      _Compare __comp,
+      _LeafMerge __leaf_merge) {
+    __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
+    return __empty{};
+  }
+
+  static constexpr size_t __lane_size = 64;
+};
+
+} // namespace __pstl
 _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && && _LIBCPP_STD_VER >= 17
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
index fdf1a2e78dad9..d4c383997a67a 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
@@ -14,6 +14,7 @@
 #include <__config>
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
@@ -49,7 +50,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_transform(
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    std::__par_backend::__parallel_for(
+    __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
         __first, __last, [__op, __first, __result](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           auto __res = std::__pstl_transform<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __result + (__brick_first - __first), __op);
@@ -97,7 +98,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_transform(
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    auto __res = std::__par_backend::__parallel_for(
+    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
         __first1,
         __last1,
         [__op, __first1, __first2, __result](_ForwardIterator1 __brick_first, _ForwardIterator1 __brick_last) {
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
index 376abd39fa36e..956c7d6a88ce2 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
@@ -14,6 +14,7 @@
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/transform_reduce.h>
+#include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/desugars_to.h>
 #include <__type_traits/is_arithmetic.h>
 #include <__type_traits/is_execution_policy.h>
@@ -32,7 +33,8 @@ _LIBCPP_PUSH_MACROS
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-template <typename _DifferenceType,
+template <typename _Backend,
+          typename _DifferenceType,
           typename _Tp,
           typename _BinaryOperation,
           typename _UnaryOperation,
@@ -48,7 +50,8 @@ __simd_transform_reduce(_DifferenceType __n, _Tp __init, _BinaryOperation, _Unar
   return __init;
 }
 
-template <typename _Size,
+template <typename _Backend,
+          typename _Size,
           typename _Tp,
           typename _BinaryOperation,
           typename _UnaryOperation,
@@ -58,7 +61,8 @@ template <typename _Size,
                         int>    = 0>
 _LIBCPP_HIDE_FROM_ABI _Tp
 __simd_transform_reduce(_Size __n, _Tp __init, _BinaryOperation __binary_op, _UnaryOperation __f) noexcept {
-  const _Size __block_size = __lane_size / sizeof(_Tp);
+  constexpr size_t __lane_size = __pstl::__cpu_traits<_Backend>::__lane_size;
+  const _Size __block_size     = __lane_size / sizeof(_Tp);
   if (__n > 2 * __block_size && __block_size > 1) {
     alignas(__lane_size) char __lane_buffer[__lane_size];
     _Tp* __lane = reinterpret_cast<_Tp*>(__lane_buffer);
@@ -116,7 +120,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value) {
-    return __par_backend::__parallel_transform_reduce(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_transform_reduce(
         __first1,
         std::move(__last1),
         [__first1, __first2, __transform](_ForwardIterator1 __iter) {
@@ -138,7 +142,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
   } else if constexpr (__is_unsequenced_execution_policy_v<_ExecutionPolicy> &&
                        __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                        __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value) {
-    return std::__simd_transform_reduce(
+    return std::__simd_transform_reduce<__cpu_backend_tag>(
         __last1 - __first1, std::move(__init), std::move(__reduce), [&](__iter_diff_t<_ForwardIterator1> __i) {
           return __transform(__first1[__i], __first2[__i]);
         });
@@ -163,7 +167,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
     _UnaryOperation __transform) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return __par_backend::__parallel_transform_reduce(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_transform_reduce(
         std::move(__first),
         std::move(__last),
         [__transform](_ForwardIterator __iter) { return __transform(*__iter); },
@@ -182,7 +186,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
         });
   } else if constexpr (__is_unsequenced_execution_policy_v<_ExecutionPolicy> &&
                        __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return std::__simd_transform_reduce(
+    return std::__simd_transform_reduce<__cpu_backend_tag>(
         __last - __first,
         std::move(__init),
         std::move(__reduce),
diff --git a/libcxx/include/__pstl/cpu_algos/cpu_traits.h b/libcxx/include/__pstl/cpu_algos/cpu_traits.h
new file mode 100644
index 0000000000000..2f0db46e9be83
--- /dev/null
+++ b/libcxx/include/__pstl/cpu_algos/cpu_traits.h
@@ -0,0 +1,87 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_CPU_TRAITS_H
+#define _LIBCPP___PSTL_CPU_ALGOS_CPU_TRAITS_H
+
+#include <__config>
+#include <cstddef>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+_LIBCPP_PUSH_MACROS
+#include <__undef_macros>
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+namespace __pstl {
+
+// __cpu_traits
+//
+// This traits class encapsulates the basis operations for a CPU-based implementation of the PSTL.
+// All the operations in the PSTL can be implemented from these basis operations, so a pure CPU backend
+// only needs to customize these traits in order to get an implementation of the whole PSTL.
+//
+// Basis operations
+// ================
+//
+//  template <class _RandomAccessIterator, class _Functor>
+//  optional<__empty> __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func);
+//    - __func must take a subrange of [__first, __last) that should be executed in serial
+//
+//  template <class _Iterator, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduction>
+//  optional<_Tp> __parallel_transform_reduce(_Iterator __first, _Iterator __last, _UnaryOp, _Tp __init, _BinaryOp,
+//  _Reduction);
+//
+//  template <class _RandomAccessIterator1,
+//            class _RandomAccessIterator2,
+//            class _RandomAccessIterator3,
+//            class _Compare,
+//            class _LeafMerge>
+//  optional<_RandomAccessIterator3> __parallel_merge(_RandomAccessIterator1 __first1,
+//                                                    _RandomAccessIterator1 __last1,
+//                                                    _RandomAccessIterator2 __first2,
+//                                                    _RandomAccessIterator2 __last2,
+//                                                    _RandomAccessIterator3 __outit,
+//                                                    _Compare __comp,
+//                                                    _LeafMerge __leaf_merge);
+//
+//  template <class _RandomAccessIterator, class _Comp, class _LeafSort>
+//  optional<__empty> __parallel_stable_sort(_RandomAccessIterator __first,
+//                                           _RandomAccessIterator __last,
+//                                           _Comp __comp,
+//                                           _LeafSort __leaf_sort);
+//
+//   void __cancel_execution();
+//      Cancel the execution of other jobs - they aren't needed anymore. This is not a binding request,
+//      some backends may not actually be able to cancel jobs.
+//
+//   constexpr size_t __lane_size;
+//      Size of SIMD lanes.
+//      TODO: Merge this with __native_vector_size from __algorithm/simd_utils.h
+//
+//
+// Exception handling
+// ==================
+//
+// CPU backends are expected to report errors (i.e. failure to allocate) by returning a disengaged `optional` from their
+// implementation. Exceptions shouldn't be used to report an internal failure-to-allocate, since all exceptions are
+// turned into a program termination at the front-end level. When a backend returns a disengaged `optional` to the
+// frontend, the frontend will turn that into a call to `std::__throw_bad_alloc();` to report the internal failure to
+// the user.
+
+template <class _Backend>
+struct __cpu_traits;
+
+} // namespace __pstl
+_LIBCPP_END_NAMESPACE_STD
+
+_LIBCPP_POP_MACROS
+
+#endif // _LIBCPP___PSTL_CPU_ALGOS_CPU_TRAITS_H
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index ccee7029824e3..7312a744ef679 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -1613,6 +1613,8 @@ module std_private_numeric_transform_exclusive_scan [system] { header "__numeric
 module std_private_numeric_transform_inclusive_scan [system] { header "__numeric/transform_inclusive_scan.h" }
 module std_private_numeric_transform_reduce         [system] { header "__numeric/transform_reduce.h" }
 
+module std_private_pstl_cpu_algos_cpu_traits [system] { header "__pstl/cpu_algos/cpu_traits.h" }
+
 module std_private_queue_fwd [system] { header "__fwd/queue.h" }
 
 module std_private_random_bernoulli_distribution          [system] { header "__random/bernoulli_distribution.h" }
diff --git a/libcxx/src/pstl/libdispatch.cpp b/libcxx/src/pstl/libdispatch.cpp
index 52d4afbcce6e0..d997a9c73463d 100644
--- a/libcxx/src/pstl/libdispatch.cpp
+++ b/libcxx/src/pstl/libdispatch.cpp
@@ -12,8 +12,7 @@
 #include <dispatch/dispatch.h>
 
 _LIBCPP_BEGIN_NAMESPACE_STD
-
-namespace __par_backend::inline __libdispatch {
+namespace __pstl::__libdispatch {
 
 void __dispatch_apply(size_t chunk_count, void* context, void (*func)(void* context, size_t chunk)) noexcept {
   ::dispatch_apply_f(chunk_count, DISPATCH_APPLY_AUTO, context, func);
@@ -29,7 +28,5 @@ __chunk_partitions __partition_chunks(ptrdiff_t element_count) noexcept {
   return partitions;
 }
 
-// NOLINTNEXTLINE(llvm-namespace-comment) // This is https://llvm.org/PR56804
-} // namespace __par_backend::inline __libdispatch
-
+} // namespace __pstl::__libdispatch
 _LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp b/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
index 554924a0179d5..8c7016a80b811 100644
--- a/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
+++ b/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
@@ -18,21 +18,21 @@
 
 int main(int, char**) {
   {
-    auto chunks = std::__par_backend::__libdispatch::__partition_chunks(0);
+    auto chunks = std::__pstl::__libdispatch::__partition_chunks(0);
     assert(chunks.__chunk_count_ == 1);
     assert(chunks.__first_chunk_size_ == 0);
     assert(chunks.__chunk_size_ == 0);
   }
 
   {
-    auto chunks = std::__par_backend::__libdispatch::__partition_chunks(1);
+    auto chunks = std::__pstl::__libdispatch::__partition_chunks(1);
     assert(chunks.__chunk_count_ == 1);
     assert(chunks.__first_chunk_size_ == 1);
     assert(chunks.__chunk_size_ == 1);
   }
 
   for (std::ptrdiff_t i = 2; i != 2ll << 20; ++i) {
-    auto chunks = std::__par_backend::__libdispatch::__partition_chunks(i);
+    auto chunks = std::__pstl::__libdispatch::__partition_chunks(i);
     assert(chunks.__chunk_count_ >= 1);
     assert(chunks.__chunk_count_ <= i);
     assert((chunks.__chunk_count_ - 1) * chunks.__chunk_size_ + chunks.__first_chunk_size_ == i);
diff --git a/libcxx/utils/generate_iwyu_mapping.py b/libcxx/utils/generate_iwyu_mapping.py
index 2265438ab49cc..33607f1996841 100644
--- a/libcxx/utils/generate_iwyu_mapping.py
+++ b/libcxx/utils/generate_iwyu_mapping.py
@@ -10,6 +10,7 @@ def IWYU_mapping(header: str) -> typing.Optional[typing.List[str]]:
     ignore = [
         "__debug_utils/.+",
         "__fwd/get[.]h",
+        "__pstl/.+",
         "__support/.+",
         "__utility/private_constructor_tag.h",
     ]

From 7e4c6e98fa05f5c3bf14f96365ae74a8d12c6257 Mon Sep 17 00:00:00 2001
From: Jonas Paulsson <paulson1@linux.ibm.com>
Date: Mon, 15 Apr 2024 16:32:14 +0200
Subject: [PATCH 435/886] [SystemZ] Bugfix in getDemandedSrcElements().
 (#88623)

For the intrinsic s390_vperm, all of the elements are demanded, so use
an APInt with the value of '-1' for them (not '1').

Fixes https://github.com/llvm/llvm-project/issues/88397
---
 .../Target/SystemZ/SystemZISelLowering.cpp    |  2 +-
 .../SystemZ/knownbits-intrinsics-binop.ll     | 19 +++++++++++++++++++
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
index 3b3057f8fe964..5c2579f3bf187 100644
--- a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
@@ -7818,7 +7818,7 @@ static APInt getDemandedSrcElements(SDValue Op, const APInt &DemandedElts,
       break;
     }
     case Intrinsic::s390_vperm:
-      SrcDemE = APInt(NumElts, 1);
+      SrcDemE = APInt(NumElts, -1);
       break;
     default:
       llvm_unreachable("Unhandled intrinsic.");
diff --git a/llvm/test/CodeGen/SystemZ/knownbits-intrinsics-binop.ll b/llvm/test/CodeGen/SystemZ/knownbits-intrinsics-binop.ll
index 3bcbbb45581f9..b855d01934782 100644
--- a/llvm/test/CodeGen/SystemZ/knownbits-intrinsics-binop.ll
+++ b/llvm/test/CodeGen/SystemZ/knownbits-intrinsics-binop.ll
@@ -458,3 +458,22 @@ define <16 x i8> @f30() {
                                i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
   ret <16 x i8> %res
 }
+
+; Test VPERM with various constant operands.
+define i32 @f31() {
+; CHECK-LABEL: f31:
+; CHECK-LABEL: # %bb.0:
+; CHECK-NEXT: larl %r1, .LCPI31_0
+; CHECK-NEXT: vl %v0, 0(%r1), 3
+; CHECK-NEXT: larl %r1, .LCPI31_1
+; CHECK-NEXT: vl %v1, 0(%r1), 3
+; CHECK-NEXT: vperm %v0, %v1, %v1, %v0
+; CHECK-NEXT: vlgvb %r2, %v0, 0
+; CHECK-NEXT: nilf %r2, 7
+; CHECK-NEXT:                           # kill: def $r2l killed $r2l killed $r2d
+; CHECK-NEXT: br %r14
+  %P = tail call <16 x i8> @llvm.s390.vperm(<16 x i8> <i8 0, i8 1, i8 1, i8 2, i8 1, i8 2, i8 2, i8 3, i8 1, i8 2, i8 2, i8 3, i8 2, i8 3, i8 3, i8 4>, <16 x i8> <i8 0, i8 1, i8 1, i8 2, i8 1, i8 2, i8 2, i8 3, i8 1, i8 2, i8 2, i8 3, i8 2, i8 3, i8 3, i8 4>, <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0>)
+  %E = extractelement <16 x i8> %P, i64 0
+  %res = zext i8 %E to i32
+  ret i32 %res
+}

From a2691959cd23de9505243e9396897a3415606689 Mon Sep 17 00:00:00 2001
From: whisperity <whisperity@gmail.com>
Date: Mon, 15 Apr 2024 16:32:52 +0200
Subject: [PATCH 436/886] [clang-tidy][NFC] Fix `linuxkernel-must-check-errs`
 documentation file name (#88655)

The check was originally introduced in commit
fc8c65b2e11d21b1c0ce070cc0a4ee031a542dae with conflicting names, likely as a
result of a partial rename, with the name in the documentation
(`linuxkernel-must-use-errs`) not matching the name of the check as configured
by users.
---
 .../clang-tidy/linuxkernel/MustCheckErrsCheck.h          | 9 +--------
 clang-tools-extra/docs/ReleaseNotes.rst                  | 4 ++++
 .../{must-use-errs.rst => must-check-errs.rst}           | 6 +++---
 clang-tools-extra/docs/clang-tidy/checks/list.rst        | 2 +-
 4 files changed, 9 insertions(+), 12 deletions(-)
 rename clang-tools-extra/docs/clang-tidy/checks/linuxkernel/{must-use-errs.rst => must-check-errs.rst} (88%)

diff --git a/clang-tools-extra/clang-tidy/linuxkernel/MustCheckErrsCheck.h b/clang-tools-extra/clang-tidy/linuxkernel/MustCheckErrsCheck.h
index f08fed4798392..7406aaead836e 100644
--- a/clang-tools-extra/clang-tidy/linuxkernel/MustCheckErrsCheck.h
+++ b/clang-tools-extra/clang-tidy/linuxkernel/MustCheckErrsCheck.h
@@ -17,15 +17,8 @@ namespace clang::tidy::linuxkernel {
 /// linux/err.h. Also checks to see if code uses the results from functions that
 /// directly return a value from one of these error functions.
 ///
-/// This is important in the Linux kernel because ERR_PTR, PTR_ERR, IS_ERR,
-/// IS_ERR_OR_NULL, ERR_CAST, and PTR_ERR_OR_ZERO return values must be checked,
-/// since positive pointers and negative error codes are being used in the same
-/// context. These functions are marked with
-/// __attribute__((warn_unused_result)), but some kernel versions do not have
-/// this warning enabled for clang.
-///
 /// For the user-facing documentation see:
-/// http://clang.llvm.org/extra/clang-tidy/checks/linuxkernel/must-use-errs.html
+/// http://clang.llvm.org/extra/clang-tidy/checks/linuxkernel/must-check-errs.html
 class MustCheckErrsCheck : public ClangTidyCheck {
 public:
   MustCheckErrsCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index de0a7b79bb2cf..7095c564444fe 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -303,6 +303,10 @@ Miscellaneous
   ``--format`` option is specified. Now :program:`clang-apply-replacements`
   applies formatting only with the option.
 
+- Fixed the :doc:`linuxkernel-must-check-errs
+  <clang-tidy/checks/linuxkernel/must-check-errs>` documentation to consistently
+  use the check's proper name.
+
 Improvements to include-fixer
 -----------------------------
 
diff --git a/clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-use-errs.rst b/clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-check-errs.rst
similarity index 88%
rename from clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-use-errs.rst
rename to clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-check-errs.rst
index 8a85426880987..cef5a70db309e 100644
--- a/clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-use-errs.rst
+++ b/clang-tools-extra/docs/clang-tidy/checks/linuxkernel/must-check-errs.rst
@@ -1,7 +1,7 @@
-.. title:: clang-tidy - linuxkernel-must-use-errs
+.. title:: clang-tidy - linuxkernel-must-check-errs
 
-linuxkernel-must-use-errs
-=========================
+linuxkernel-must-check-errs
+===========================
 
 Checks Linux kernel code to see if it uses the results from the functions in
 ``linux/err.h``. Also checks to see if code uses the results from functions that
diff --git a/clang-tools-extra/docs/clang-tidy/checks/list.rst b/clang-tools-extra/docs/clang-tidy/checks/list.rst
index 188a42bfddd38..8bc46acad56c8 100644
--- a/clang-tools-extra/docs/clang-tidy/checks/list.rst
+++ b/clang-tools-extra/docs/clang-tidy/checks/list.rst
@@ -233,7 +233,7 @@ Clang-Tidy Checks
    :doc:`hicpp-multiway-paths-covered <hicpp/multiway-paths-covered>`,
    :doc:`hicpp-no-assembler <hicpp/no-assembler>`,
    :doc:`hicpp-signed-bitwise <hicpp/signed-bitwise>`,
-   :doc:`linuxkernel-must-use-errs <linuxkernel/must-use-errs>`,
+   :doc:`linuxkernel-must-check-errs <linuxkernel/must-check-errs>`,
    :doc:`llvm-header-guard <llvm/header-guard>`,
    :doc:`llvm-include-order <llvm/include-order>`, "Yes"
    :doc:`llvm-namespace-comment <llvm/namespace-comment>`,

From a06073f91e7bbbb532e68bbf6b903c2f5051f4c2 Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Mon, 15 Apr 2024 10:45:26 -0400
Subject: [PATCH 437/886] [libc++] Add a utility to check whether a range is
 valid (#87665)

In the future, this utility could be made to also work with iterators,
including bounded iterators. We could also query the ASAN runtime for
this information when it's around.
---
 libcxx/include/CMakeLists.txt                 |  1 +
 .../include/__utility/is_pointer_in_range.h   |  6 +-
 libcxx/include/__utility/is_valid_range.h     | 37 ++++++++++
 libcxx/include/libcxx.imp                     |  1 +
 libcxx/include/module.modulemap               |  1 +
 .../libcxx/utilities/is_valid_range.pass.cpp  | 68 +++++++++++++++++++
 6 files changed, 111 insertions(+), 3 deletions(-)
 create mode 100644 libcxx/include/__utility/is_valid_range.h
 create mode 100644 libcxx/test/libcxx/utilities/is_valid_range.pass.cpp

diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index 148a51a216ed4..a2af1d9915be4 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -860,6 +860,7 @@ set(files
   __utility/in_place.h
   __utility/integer_sequence.h
   __utility/is_pointer_in_range.h
+  __utility/is_valid_range.h
   __utility/move.h
   __utility/no_destroy.h
   __utility/pair.h
diff --git a/libcxx/include/__utility/is_pointer_in_range.h b/libcxx/include/__utility/is_pointer_in_range.h
index 68cdfea6f9452..9eee8bf811c60 100644
--- a/libcxx/include/__utility/is_pointer_in_range.h
+++ b/libcxx/include/__utility/is_pointer_in_range.h
@@ -17,6 +17,7 @@
 #include <__type_traits/is_constant_evaluated.h>
 #include <__type_traits/void_t.h>
 #include <__utility/declval.h>
+#include <__utility/is_valid_range.h>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
@@ -34,16 +35,15 @@ struct __is_less_than_comparable<_Tp, _Up, __void_t<decltype(std::declval<_Tp>()
 template <class _Tp, class _Up, __enable_if_t<__is_less_than_comparable<const _Tp*, const _Up*>::value, int> = 0>
 _LIBCPP_CONSTEXPR_SINCE_CXX14 _LIBCPP_HIDE_FROM_ABI _LIBCPP_NO_SANITIZE("address") bool __is_pointer_in_range(
     const _Tp* __begin, const _Tp* __end, const _Up* __ptr) {
-  if (__libcpp_is_constant_evaluated()) {
-    _LIBCPP_ASSERT_VALID_INPUT_RANGE(__builtin_constant_p(__begin <= __end), "__begin and __end do not form a range");
+  _LIBCPP_ASSERT_VALID_INPUT_RANGE(std::__is_valid_range(__begin, __end), "[__begin, __end) is not a valid range");
 
+  if (__libcpp_is_constant_evaluated()) {
     // If this is not a constant during constant evaluation we know that __ptr is not part of the allocation where
     // [__begin, __end) is.
     if (!__builtin_constant_p(__begin <= __ptr && __ptr < __end))
       return false;
   }
 
-  // Checking this for unrelated pointers is technically UB, but no compiler optimizes based on it (currently).
   return !__less<>()(__ptr, __begin) && __less<>()(__ptr, __end);
 }
 
diff --git a/libcxx/include/__utility/is_valid_range.h b/libcxx/include/__utility/is_valid_range.h
new file mode 100644
index 0000000000000..7286662dbf309
--- /dev/null
+++ b/libcxx/include/__utility/is_valid_range.h
@@ -0,0 +1,37 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___UTILITY_IS_VALID_RANGE_H
+#define _LIBCPP___UTILITY_IS_VALID_RANGE_H
+
+#include <__algorithm/comp.h>
+#include <__config>
+#include <__type_traits/is_constant_evaluated.h>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+
+template <class _Tp>
+_LIBCPP_CONSTEXPR_SINCE_CXX14 _LIBCPP_HIDE_FROM_ABI _LIBCPP_NO_SANITIZE("address") bool
+__is_valid_range(const _Tp* __first, const _Tp* __last) {
+  if (__libcpp_is_constant_evaluated()) {
+    // If this is not a constant during constant evaluation, that is because __first and __last are not
+    // part of the same allocation. If they are part of the same allocation, we must still make sure they
+    // are ordered properly.
+    return __builtin_constant_p(__first <= __last) && __first <= __last;
+  }
+
+  return !__less<>()(__last, __first);
+}
+
+_LIBCPP_END_NAMESPACE_STD
+
+#endif // _LIBCPP___UTILITY_IS_VALID_RANGE_H
diff --git a/libcxx/include/libcxx.imp b/libcxx/include/libcxx.imp
index 6c77ba8343c60..8820fb8c0936f 100644
--- a/libcxx/include/libcxx.imp
+++ b/libcxx/include/libcxx.imp
@@ -853,6 +853,7 @@
   { include: [ "<__utility/in_place.h>", "private", "<utility>", "public" ] },
   { include: [ "<__utility/integer_sequence.h>", "private", "<utility>", "public" ] },
   { include: [ "<__utility/is_pointer_in_range.h>", "private", "<utility>", "public" ] },
+  { include: [ "<__utility/is_valid_range.h>", "private", "<utility>", "public" ] },
   { include: [ "<__utility/move.h>", "private", "<utility>", "public" ] },
   { include: [ "<__utility/no_destroy.h>", "private", "<utility>", "public" ] },
   { include: [ "<__utility/pair.h>", "private", "<utility>", "public" ] },
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index 7312a744ef679..ce133e471deb7 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -2075,6 +2075,7 @@ module std_private_utility_forward_like           [system] { header "__utility/f
 module std_private_utility_in_place               [system] { header "__utility/in_place.h" }
 module std_private_utility_integer_sequence       [system] { header "__utility/integer_sequence.h" }
 module std_private_utility_is_pointer_in_range    [system] { header "__utility/is_pointer_in_range.h" }
+module std_private_utility_is_valid_range         [system] { header "__utility/is_valid_range.h" }
 module std_private_utility_move                   [system] {
   header "__utility/move.h"
   export std_private_type_traits_is_copy_constructible
diff --git a/libcxx/test/libcxx/utilities/is_valid_range.pass.cpp b/libcxx/test/libcxx/utilities/is_valid_range.pass.cpp
new file mode 100644
index 0000000000000..345e2feeda81c
--- /dev/null
+++ b/libcxx/test/libcxx/utilities/is_valid_range.pass.cpp
@@ -0,0 +1,68 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <__utility/is_valid_range.h>
+#include <cassert>
+
+#include "test_macros.h"
+
+template <class T, class TQualified>
+TEST_CONSTEXPR_CXX14 void check_type() {
+  {
+    // We need to ensure that the addresses of i and j are ordered as &i < &j for
+    // the test below to work portably, so we define them in a struct.
+    struct {
+      T i = 0;
+      T j = 0;
+    } storage;
+    assert(std::__is_valid_range(static_cast<TQualified*>(&storage.i), static_cast<TQualified*>(&storage.i)));
+    assert(std::__is_valid_range(static_cast<TQualified*>(&storage.i), static_cast<TQualified*>(&storage.i + 1)));
+
+    assert(!std::__is_valid_range(static_cast<TQualified*>(&storage.j), static_cast<TQualified*>(&storage.i)));
+    assert(!std::__is_valid_range(static_cast<TQualified*>(&storage.i + 1), static_cast<TQualified*>(&storage.i)));
+
+    // We detect this as being a valid range even though it is not really valid.
+    assert(std::__is_valid_range(static_cast<TQualified*>(&storage.i), static_cast<TQualified*>(&storage.j)));
+  }
+
+  {
+    T arr[3] = {1, 2, 3};
+    assert(std::__is_valid_range(static_cast<TQualified*>(&arr[0]), static_cast<TQualified*>(&arr[0])));
+    assert(std::__is_valid_range(static_cast<TQualified*>(&arr[0]), static_cast<TQualified*>(&arr[1])));
+    assert(std::__is_valid_range(static_cast<TQualified*>(&arr[0]), static_cast<TQualified*>(&arr[2])));
+
+    assert(!std::__is_valid_range(static_cast<TQualified*>(&arr[1]), static_cast<TQualified*>(&arr[0])));
+    assert(!std::__is_valid_range(static_cast<TQualified*>(&arr[2]), static_cast<TQualified*>(&arr[0])));
+  }
+
+#if TEST_STD_VER >= 20
+  {
+    T* arr = new int[4]{1, 2, 3, 4};
+    assert(std::__is_valid_range(static_cast<TQualified*>(arr), static_cast<TQualified*>(arr + 4)));
+    delete[] arr;
+  }
+#endif
+}
+
+TEST_CONSTEXPR_CXX14 bool test() {
+  check_type<int, int>();
+  check_type<int, int const>();
+  check_type<int, int volatile>();
+  check_type<int, int const volatile>();
+
+  return true;
+}
+
+int main(int, char**) {
+  test();
+#if TEST_STD_VER >= 14
+  static_assert(test(), "");
+#endif
+
+  return 0;
+}

From 7b9b28dbd2938877a94737e28b7eb9e3cdd82755 Mon Sep 17 00:00:00 2001
From: yronglin <yronglin777@gmail.com>
Date: Mon, 15 Apr 2024 23:52:26 +0800
Subject: [PATCH 438/886] [Clang] Refactor implementation of "Lifetime
 extension in range-based for loops" (#87930)

This PR remove `InMaterializeTemporaryObjectContext` , because it's
redundant, materialize non-cv void prvalue temporaries in discarded
expressions can only appear under lifetime-extension context.

Signed-off-by: yronglin <yronglin777@gmail.com>
---
 clang/include/clang/Sema/Sema.h               | 47 -------------------
 clang/lib/Parse/ParseDecl.cpp                 |  4 --
 clang/lib/Sema/SemaExpr.cpp                   |  7 ++-
 clang/lib/Sema/SemaExprCXX.cpp                |  2 +-
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |  1 -
 clang/lib/Sema/TreeTransform.h                |  5 --
 6 files changed, 4 insertions(+), 62 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index c6035445e9019..d93ac7863b721 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -5099,34 +5099,6 @@ class Sema final : public SemaBase {
     /// example, in a for-range initializer).
     bool InLifetimeExtendingContext = false;
 
-    /// Whether we are currently in a context in which all temporaries must be
-    /// materialized.
-    ///
-    /// [class.temporary]/p2:
-    /// The materialization of a temporary object is generally delayed as long
-    /// as possible in order to avoid creating unnecessary temporary objects.
-    ///
-    /// Temporary objects are materialized:
-    ///   (2.1) when binding a reference to a prvalue ([dcl.init.ref],
-    ///   [expr.type.conv], [expr.dynamic.cast], [expr.static.cast],
-    ///   [expr.const.cast], [expr.cast]),
-    ///
-    ///   (2.2) when performing member access on a class prvalue ([expr.ref],
-    ///   [expr.mptr.oper]),
-    ///
-    ///   (2.3) when performing an array-to-pointer conversion or subscripting
-    ///   on an array prvalue ([conv.array], [expr.sub]),
-    ///
-    ///   (2.4) when initializing an object of type
-    ///   std​::​initializer_list<T> from a braced-init-list
-    ///   ([dcl.init.list]),
-    ///
-    ///   (2.5) for certain unevaluated operands ([expr.typeid], [expr.sizeof])
-    ///
-    ///   (2.6) when a prvalue that has type other than cv void appears as a
-    ///   discarded-value expression ([expr.context]).
-    bool InMaterializeTemporaryObjectContext = false;
-
     // When evaluating immediate functions in the initializer of a default
     // argument or default member initializer, this is the declaration whose
     // default initializer is being evaluated and the location of the call
@@ -6398,19 +6370,6 @@ class Sema final : public SemaBase {
     }
   }
 
-  /// keepInMaterializeTemporaryObjectContext - Pull down
-  /// InMaterializeTemporaryObjectContext flag from previous context.
-  void keepInMaterializeTemporaryObjectContext() {
-    if (ExprEvalContexts.size() > 2 &&
-        ExprEvalContexts[ExprEvalContexts.size() - 2]
-            .InMaterializeTemporaryObjectContext) {
-      auto &LastRecord = ExprEvalContexts.back();
-      auto &PrevRecord = ExprEvalContexts[ExprEvalContexts.size() - 2];
-      LastRecord.InMaterializeTemporaryObjectContext =
-          PrevRecord.InMaterializeTemporaryObjectContext;
-    }
-  }
-
   DefaultedComparisonKind getDefaultedComparisonKind(const FunctionDecl *FD) {
     return getDefaultedFunctionKind(FD).asComparison();
   }
@@ -6554,12 +6513,6 @@ class Sema final : public SemaBase {
   /// used in initializer of the field.
   llvm::MapVector<FieldDecl *, DeleteLocs> DeleteExprs;
 
-  bool isInMaterializeTemporaryObjectContext() const {
-    assert(!ExprEvalContexts.empty() &&
-           "Must be in an expression evaluation context");
-    return ExprEvalContexts.back().InMaterializeTemporaryObjectContext;
-  }
-
   ParsedType getInheritingConstructorName(CXXScopeSpec &SS,
                                           SourceLocation NameLoc,
                                           const IdentifierInfo &Name);
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index a990ef398c823..2b934234b7cf5 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -2380,10 +2380,6 @@ Parser::DeclGroupPtrTy Parser::ParseDeclGroup(ParsingDeclSpec &DS,
       if (getLangOpts().CPlusPlus23) {
         auto &LastRecord = Actions.ExprEvalContexts.back();
         LastRecord.InLifetimeExtendingContext = true;
-
-        // Materialize non-`cv void` prvalue temporaries in discarded
-        // expressions. These materialized temporaries may be lifetime-extented.
-        LastRecord.InMaterializeTemporaryObjectContext = true;
       }
 
       if (getLangOpts().OpenMP)
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index c5395fb2068a1..24f354f1c7249 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -6314,7 +6314,6 @@ ExprResult Sema::BuildCXXDefaultArgExpr(SourceLocation CallLoc,
       // Pass down lifetime extending flag, and collect temporaries in
       // CreateMaterializeTemporaryExpr when we rewrite the call argument.
       keepInLifetimeExtendingContext();
-      keepInMaterializeTemporaryObjectContext();
       EnsureImmediateInvocationInDefaultArgs Immediate(*this);
       ExprResult Res;
       runWithSufficientStackSpace(CallLoc, [&] {
@@ -18679,9 +18678,9 @@ void Sema::PopExpressionEvaluationContext() {
   // Append the collected materialized temporaries into previous context before
   // exit if the previous also is a lifetime extending context.
   auto &PrevRecord = ExprEvalContexts[ExprEvalContexts.size() - 2];
-  if (getLangOpts().CPlusPlus23 && isInLifetimeExtendingContext() &&
-      PrevRecord.InLifetimeExtendingContext && !ExprEvalContexts.empty()) {
-    auto &PrevRecord = ExprEvalContexts[ExprEvalContexts.size() - 2];
+  if (getLangOpts().CPlusPlus23 && Rec.InLifetimeExtendingContext &&
+      PrevRecord.InLifetimeExtendingContext &&
+      !Rec.ForRangeLifetimeExtendTemps.empty()) {
     PrevRecord.ForRangeLifetimeExtendTemps.append(
         Rec.ForRangeLifetimeExtendTemps);
   }
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index fa7b92a43346b..74ed3fe7bd520 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -8429,7 +8429,7 @@ ExprResult Sema::IgnoredValueConversions(Expr *E) {
     // unnecessary temporary objects. If we skip this step, IR generation is
     // able to synthesize the storage for itself in the aggregate case, and
     // adding the extra node to the AST is just clutter.
-    if (isInMaterializeTemporaryObjectContext() && getLangOpts().CPlusPlus17 &&
+    if (isInLifetimeExtendingContext() && getLangOpts().CPlusPlus17 &&
         E->isPRValue() && !E->getType()->isVoidType()) {
       ExprResult Res = TemporaryMaterializationConversion(E);
       if (Res.isInvalid())
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 02d0b35ce940f..c45a8d1408fff 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5483,7 +5483,6 @@ void Sema::InstantiateVariableInitializer(
         *this, Sema::ExpressionEvaluationContext::PotentiallyEvaluated, Var);
 
     keepInLifetimeExtendingContext();
-    keepInMaterializeTemporaryObjectContext();
     // Instantiate the initializer.
     ExprResult Init;
 
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index fdf84c512c493..8c96134af7c8f 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -4177,7 +4177,6 @@ ExprResult TreeTransform<Derived>::TransformInitializer(Expr *Init,
       getSema(), EnterExpressionEvaluationContext::InitList,
       Construct->isListInitialization());
 
-  getSema().keepInLifetimeExtendingContext();
   getSema().keepInLifetimeExtendingContext();
   SmallVector<Expr*, 8> NewArgs;
   bool ArgChanged = false;
@@ -8752,10 +8751,6 @@ TreeTransform<Derived>::TransformCXXForRangeStmt(CXXForRangeStmt *S) {
   if (getSema().getLangOpts().CPlusPlus23) {
     auto &LastRecord = getSema().ExprEvalContexts.back();
     LastRecord.InLifetimeExtendingContext = true;
-
-    // Materialize non-`cv void` prvalue temporaries in discarded
-    // expressions. These materialized temporaries may be lifetime-extented.
-    LastRecord.InMaterializeTemporaryObjectContext = true;
   }
   StmtResult Init =
       S->getInit() ? getDerived().TransformStmt(S->getInit()) : StmtResult();

From 8927ac865755d7e98411500a7a2da70e8a185cc2 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Mon, 15 Apr 2024 08:56:57 -0700
Subject: [PATCH 439/886] [BOLT] Fix a warning

This patch fixes:

  bolt/lib/Profile/BoltAddressTranslation.cpp:380:37: error: operator
  '<<' has lower precedence than '+'; '+' will be evaluated first
  [-Werror,-Wshift-op-parentheses]
---
 bolt/lib/Profile/BoltAddressTranslation.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/bolt/lib/Profile/BoltAddressTranslation.cpp b/bolt/lib/Profile/BoltAddressTranslation.cpp
index ac6e3d01f2718..7cfb9c132c2c6 100644
--- a/bolt/lib/Profile/BoltAddressTranslation.cpp
+++ b/bolt/lib/Profile/BoltAddressTranslation.cpp
@@ -377,7 +377,7 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
       PrevAddress = OutputAddress;
       int64_t InputDelta = 0;
       if (J < EqualElems) {
-        InputOffset = (OutputOffset + ColdInputSkew << 1) | BEBitMask[J];
+        InputOffset = ((OutputOffset + ColdInputSkew) << 1) | BEBitMask[J];
       } else {
         InputDelta = DE.getSLEB128(&Offset, &Err);
         InputOffset += InputDelta;

From b4776b8d8ea742a46039002fac4c280e619ac48d Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Mon, 15 Apr 2024 17:58:37 +0200
Subject: [PATCH 440/886] [libc++][CI] Tests LLDB libc++ data formatters.
 (#88312)

This enables testing of the LLDB libc++ specific data formatters.
This is enabled in the bootstrap build since building LLDB requires
Clang and this
is quite expensive. Adding this test changes the build time from 31 to
34 minutes.
---
 libcxx/utils/ci/run-buildbot                    | 10 +++++++---
 lldb/packages/Python/lldbsuite/test/lldbtest.py |  2 ++
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/libcxx/utils/ci/run-buildbot b/libcxx/utils/ci/run-buildbot
index cc72f4639b1a9..23a2a1bbbc63f 100755
--- a/libcxx/utils/ci/run-buildbot
+++ b/libcxx/utils/ci/run-buildbot
@@ -368,18 +368,22 @@ bootstrapping-build)
           -DCMAKE_CXX_COMPILER_LAUNCHER="ccache" \
           -DCMAKE_BUILD_TYPE=Release \
           -DCMAKE_INSTALL_PREFIX="${INSTALL_DIR}" \
-          -DLLVM_ENABLE_PROJECTS="clang" \
+          -DLLVM_ENABLE_PROJECTS="clang;lldb" \
           -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi;libunwind" \
           -DLLVM_RUNTIME_TARGETS="$(${CXX} --print-target-triple)" \
+          -DLLVM_HOST_TRIPLE="$(${CXX} --print-target-triple)" \
           -DLLVM_TARGETS_TO_BUILD="host" \
           -DRUNTIMES_BUILD_ALLOW_DARWIN=ON \
           -DLLVM_ENABLE_ASSERTIONS=ON \
           -DLLVM_LIT_ARGS="-sv --xunit-xml-output test-results.xml --timeout=1500 --time-tests"
 
-    echo "+++ Running the libc++ and libc++abi tests"
+    echo "+++ Running the LLDB libc++ data formatter tests"
+    ${NINJA} -vC "${BUILD_DIR}" check-lldb-api-functionalities-data-formatter-data-formatter-stl-libcxx
+
+    echo "--- Running the libc++ and libc++abi tests"
     ${NINJA} -vC "${BUILD_DIR}" check-runtimes
 
-    echo "--- Installing libc++ and libc++abi to a fake location"
+    echo "+++ Installing libc++ and libc++abi to a fake location"
     ${NINJA} -vC "${BUILD_DIR}" install-runtimes
 
     ccache -s
diff --git a/lldb/packages/Python/lldbsuite/test/lldbtest.py b/lldb/packages/Python/lldbsuite/test/lldbtest.py
index c28a78a2c4a27..7a7afec734570 100644
--- a/lldb/packages/Python/lldbsuite/test/lldbtest.py
+++ b/lldb/packages/Python/lldbsuite/test/lldbtest.py
@@ -751,6 +751,8 @@ def setUpCommands(cls):
             "settings set symbols.enable-external-lookup false",
             # Inherit the TCC permissions from the inferior's parent.
             "settings set target.inherit-tcc true",
+            # Based on https://discourse.llvm.org/t/running-lldb-in-a-container/76801/4
+            "settings set target.disable-aslr false",
             # Kill rather than detach from the inferior if something goes wrong.
             "settings set target.detach-on-error false",
             # Disable fix-its by default so that incorrect expressions in tests don't

From 1c63a3e0cdb3f02098f8ec525ca31a3b44d9d31b Mon Sep 17 00:00:00 2001
From: mmoadeli <mahmoud.moadeli@codeplay.com>
Date: Mon, 15 Apr 2024 17:05:40 +0100
Subject: [PATCH 441/886] Resolve static analyser report on pointer
 dereferencing after null check (#88278)

- Resolve Static Analyzer Check Failure: Pointer Dereferencing After
Null Check.
- Minor naming and style improvement
---
 .../AMDGPU/AMDGPULowerModuleLDSPass.cpp       | 33 +++++++++----------
 1 file changed, 15 insertions(+), 18 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp b/llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
index 595f09664c55e..c8bf9dd39e389 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULowerModuleLDSPass.cpp
@@ -1052,21 +1052,18 @@ class AMDGPULowerModuleLDS {
   void removeNoLdsKernelIdFromReachable(CallGraph &CG, Function *KernelRoot) {
     KernelRoot->removeFnAttr("amdgpu-no-lds-kernel-id");
 
-    SmallVector<Function *> Tmp({CG[KernelRoot]->getFunction()});
-    if (!Tmp.back())
-      return;
-
+    SmallVector<Function *> WorkList({CG[KernelRoot]->getFunction()});
     SmallPtrSet<Function *, 8> Visited;
     bool SeenUnknownCall = false;
 
-    do {
-      Function *F = Tmp.pop_back_val();
+    while (!WorkList.empty()) {
+      Function *F = WorkList.pop_back_val();
 
-      for (auto &N : *CG[F]) {
-        if (!N.second)
+      for (auto &CallRecord : *CG[F]) {
+        if (!CallRecord.second)
           continue;
 
-        Function *Callee = N.second->getFunction();
+        Function *Callee = CallRecord.second->getFunction();
         if (!Callee) {
           if (!SeenUnknownCall) {
             SeenUnknownCall = true;
@@ -1074,21 +1071,21 @@ class AMDGPULowerModuleLDS {
             // If we see any indirect calls, assume nothing about potential
             // targets.
             // TODO: This could be refined to possible LDS global users.
-            for (auto &N : *CG.getExternalCallingNode()) {
-              Function *PotentialCallee = N.second->getFunction();
+            for (auto &ExternalCallRecord : *CG.getExternalCallingNode()) {
+              Function *PotentialCallee =
+                  ExternalCallRecord.second->getFunction();
+              assert(PotentialCallee);
               if (!isKernelLDS(PotentialCallee))
                 PotentialCallee->removeFnAttr("amdgpu-no-lds-kernel-id");
             }
-
-            continue;
           }
+        } else {
+          Callee->removeFnAttr("amdgpu-no-lds-kernel-id");
+          if (Visited.insert(Callee).second)
+            WorkList.push_back(Callee);
         }
-
-        Callee->removeFnAttr("amdgpu-no-lds-kernel-id");
-        if (Visited.insert(Callee).second)
-          Tmp.push_back(Callee);
       }
-    } while (!Tmp.empty());
+    }
   }
 
   DenseMap<Function *, GlobalVariable *> lowerDynamicLDSVariables(

From 9abb1ffc5cb75465340cb604988d1e386415bd72 Mon Sep 17 00:00:00 2001
From: Valery Dmitriev <valery.n.dmitriev@intel.com>
Date: Mon, 15 Apr 2024 09:09:39 -0700
Subject: [PATCH 442/886] [SLP][NFC] Add option to bypass early profitability
 check. (#88594)

The option intended primarily for LIT tests to suppress heuristic based
profitability check and proceed vectorization of a seemingly
unprofitable alternate operation pattern. This allows the vectorizer to
execute path that was the original intent of a test.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    |  7 +++++-
 .../X86/ext-int-reduced-not-operand.ll        | 19 +++++++++++++++
 .../X86/gather-move-out-of-loop.ll            | 16 +++++++++++++
 ...gathered-delayed-nodes-with-reused-user.ll | 23 +++++++++++++++++++
 .../non-scheduled-inst-reused-as-last-inst.ll | 22 ++++++++++++++++++
 5 files changed, 86 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index b031b40a978f5..9a0eaca2f5980 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -118,6 +118,11 @@ static cl::opt<int>
                      cl::desc("Only vectorize if you gain more than this "
                               "number "));
 
+static cl::opt<bool> SLPSkipEarlyProfitabilityCheck(
+    "slp-skip-early-profitability-check", cl::init(false), cl::Hidden,
+    cl::desc("When true, SLP vectorizer bypasses profitability checks based on "
+             "heuristics and makes vectorization decision via cost modeling."));
+
 static cl::opt<bool>
 ShouldVectorizeHor("slp-vectorize-hor", cl::init(true), cl::Hidden,
                    cl::desc("Attempt to vectorize horizontal reductions"));
@@ -6250,7 +6255,7 @@ BoUpSLP::TreeEntry::EntryState BoUpSLP::getScalarsVectorizationState(
       LLVM_DEBUG(dbgs() << "SLP: ShuffleVector are not vectorized.\n");
       return TreeEntry::NeedToGather;
     }
-    if (!areAltOperandsProfitable(S, VL)) {
+    if (!SLPSkipEarlyProfitabilityCheck && !areAltOperandsProfitable(S, VL)) {
       LLVM_DEBUG(
           dbgs()
           << "SLP: ShuffleVector not vectorized, operands are buildvector and "
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll b/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
index b76e26e0fd571..2ff6785c0da64 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/ext-int-reduced-not-operand.ll
@@ -1,7 +1,26 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-99999 < %s | FileCheck %s
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -slp-threshold=-99999\
+; RUN: -slp-skip-early-profitability-check < %s | FileCheck %s --check-prefixes=FORCED
 
 define i64 @wombat() {
+; FORCED-LABEL: define i64 @wombat() {
+; FORCED-NEXT:  bb:
+; FORCED-NEXT:    br label [[BB2:%.*]]
+; FORCED:       bb1:
+; FORCED-NEXT:    br label [[BB2]]
+; FORCED:       bb2:
+; FORCED-NEXT:    [[PHI:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ 0, [[BB1:%.*]] ]
+; FORCED-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[PHI]], i32 0
+; FORCED-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i32> [[TMP0]], <2 x i32> poison, <2 x i32> zeroinitializer
+; FORCED-NEXT:    [[TMP2:%.*]] = trunc <2 x i32> [[TMP1]] to <2 x i1>
+; FORCED-NEXT:    [[TMP3:%.*]] = extractelement <2 x i1> [[TMP2]], i32 0
+; FORCED-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP3]] to i64
+; FORCED-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP2]], i32 1
+; FORCED-NEXT:    [[TMP6:%.*]] = zext i1 [[TMP5]] to i64
+; FORCED-NEXT:    [[OR:%.*]] = or i64 [[TMP4]], [[TMP6]]
+; FORCED-NEXT:    ret i64 [[OR]]
+;
 ; CHECK-LABEL: define i64 @wombat() {
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    br label [[BB2:%.*]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll b/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
index 3c3dea3f1ea88..f2ea2df7cc982 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
@@ -1,7 +1,23 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -passes=slp-vectorizer -slp-threshold=-100 -mtriple=x86_64-w64-windows-gnu < %s | FileCheck %s
+; RUN: opt -S -passes=slp-vectorizer -slp-threshold=-100 -mtriple=x86_64-w64-windows-gnu\
+; RUN: -slp-skip-early-profitability-check < %s | FileCheck %s --check-prefixes=FORCED
 
 define void @test(i16 %0) {
+; FORCED-LABEL: @test(
+; FORCED-NEXT:  for.body92.preheader:
+; FORCED-NEXT:    [[TMP1:%.*]] = insertelement <2 x i16> <i16 0, i16 poison>, i16 [[TMP0:%.*]], i32 1
+; FORCED-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i32>
+; FORCED-NEXT:    [[TMP3:%.*]] = zext <2 x i16> [[TMP1]] to <2 x i32>
+; FORCED-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 0, i32 3>
+; FORCED-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> poison, <4 x i32> <i32 0, i32 poison, i32 1, i32 poison>
+; FORCED-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> <i32 poison, i32 0, i32 poison, i32 0>, <4 x i32> [[TMP5]], <4 x i32> <i32 4, i32 1, i32 6, i32 3>
+; FORCED-NEXT:    br label [[FOR_BODY92:%.*]]
+; FORCED:       for.body92:
+; FORCED-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> zeroinitializer, [[TMP6]]
+; FORCED-NEXT:    store <4 x i32> [[TMP7]], ptr undef, align 8
+; FORCED-NEXT:    br label [[FOR_BODY92]]
+;
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:  for.body92.preheader:
 ; CHECK-NEXT:    br label [[FOR_BODY92:%.*]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll b/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
index 19a8aa9b61815..d7144d750321f 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
@@ -1,7 +1,30 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -passes=slp-vectorizer -S -slp-threshold=-9999 < %s | FileCheck %s
+; RUN: opt -passes=slp-vectorizer -S -slp-threshold=-9999\
+; RUN: -slp-skip-early-profitability-check < %s | FileCheck %s --check-prefixes=FORCED
 
 define i64 @foo() {
+; FORCED-LABEL: define i64 @foo() {
+; FORCED-NEXT:  bb:
+; FORCED-NEXT:    br label [[BB3:%.*]]
+; FORCED:       bb1:
+; FORCED-NEXT:    [[TMP0:%.*]] = phi <2 x i64> [ [[TMP5:%.*]], [[BB3]] ]
+; FORCED-NEXT:    ret i64 0
+; FORCED:       bb3:
+; FORCED-NEXT:    [[PHI5:%.*]] = phi i64 [ 0, [[BB:%.*]] ], [ 0, [[BB3]] ]
+; FORCED-NEXT:    [[TMP1:%.*]] = phi <2 x i64> [ zeroinitializer, [[BB]] ], [ [[TMP7:%.*]], [[BB3]] ]
+; FORCED-NEXT:    [[TMP2:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[PHI5]], i32 0
+; FORCED-NEXT:    [[TMP3:%.*]] = add <2 x i64> [[TMP1]], [[TMP2]]
+; FORCED-NEXT:    [[TMP4:%.*]] = or <2 x i64> [[TMP1]], [[TMP2]]
+; FORCED-NEXT:    [[TMP5]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP4]], <2 x i32> <i32 0, i32 3>
+; FORCED-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 poison, i64 0>, <2 x i32> <i32 0, i32 3>
+; FORCED-NEXT:    [[TMP7]] = add <2 x i64> [[TMP6]], [[TMP2]]
+; FORCED-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP7]], i32 1
+; FORCED-NEXT:    [[GETELEMENTPTR:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[TMP8]]
+; FORCED-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
+; FORCED-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[TMP9]], 0
+; FORCED-NEXT:    br i1 false, label [[BB3]], label [[BB1:%.*]]
+;
 ; CHECK-LABEL: define i64 @foo() {
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    br label [[BB3:%.*]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll b/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
index 59cd1c0ccddf8..1163c8219dabe 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/non-scheduled-inst-reused-as-last-inst.ll
@@ -1,7 +1,29 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
 ; RUN: opt -S -passes=slp-vectorizer -slp-threshold=-9999 -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+; RUN: opt -S -passes=slp-vectorizer -slp-threshold=-9999 -mtriple=x86_64-unknown-linux-gnu\
+; RUN: -slp-skip-early-profitability-check < %s | FileCheck %s --check-prefixes=FORCED
 
 define void @foo() {
+; FORCED-LABEL: define void @foo() {
+; FORCED-NEXT:  bb:
+; FORCED-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 0, i32 0
+; FORCED-NEXT:    br label [[BB1:%.*]]
+; FORCED:       bb1:
+; FORCED-NEXT:    [[TMP1:%.*]] = phi <2 x i32> [ zeroinitializer, [[BB:%.*]] ], [ [[TMP6:%.*]], [[BB4:%.*]] ]
+; FORCED-NEXT:    [[TMP2:%.*]] = shl <2 x i32> [[TMP1]], [[TMP0]]
+; FORCED-NEXT:    [[TMP3:%.*]] = or <2 x i32> [[TMP1]], [[TMP0]]
+; FORCED-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 0, i32 3>
+; FORCED-NEXT:    [[TMP5:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> [[TMP1]], <2 x i32> <i32 0, i32 3>
+; FORCED-NEXT:    [[TMP6]] = or <2 x i32> [[TMP5]], zeroinitializer
+; FORCED-NEXT:    [[TMP7:%.*]] = extractelement <2 x i32> [[TMP6]], i32 0
+; FORCED-NEXT:    [[CALL:%.*]] = call i64 null(i32 [[TMP7]])
+; FORCED-NEXT:    br label [[BB4]]
+; FORCED:       bb4:
+; FORCED-NEXT:    br i1 false, label [[BB5:%.*]], label [[BB1]]
+; FORCED:       bb5:
+; FORCED-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ [[TMP4]], [[BB4]] ]
+; FORCED-NEXT:    ret void
+;
 ; CHECK-LABEL: define void @foo() {
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    br label [[BB1:%.*]]

From 55b6f17071d25b77fcdc910ca9b15f89305137e0 Mon Sep 17 00:00:00 2001
From: Jeremy Kun <jkun@google.com>
Date: Mon, 15 Apr 2024 09:12:00 -0700
Subject: [PATCH 443/886] Add a polynomial dialect shell, attributes, and types
 (#72081)

RFC:
https://discourse.llvm.org/t/rfc-a-poly-dialect-for-polynomial-arithmetic/73891

This PR implements the minimal work needed to represent the polynomial
type such that it can be tested with `lit`.

In this PR:

- Dialect shell
- `Polynomial` data structure needed for folding
- Polynomial attributes (`PolynomialAttr` and `RingAttr` which store a polynomial)
- `polynomial.polynomial` type
- Basic lit tests

---------

Co-authored-by: Jeremy Kun <j2kun@users.noreply.github.com>
---
 mlir/include/mlir/Dialect/CMakeLists.txt      |   1 +
 .../mlir/Dialect/Polynomial/CMakeLists.txt    |   1 +
 .../mlir/Dialect/Polynomial/IR/CMakeLists.txt |  10 +
 .../mlir/Dialect/Polynomial/IR/Polynomial.h   | 130 +++++++++++
 .../mlir/Dialect/Polynomial/IR/Polynomial.td  | 153 +++++++++++++
 .../Polynomial/IR/PolynomialAttributes.h      |  17 ++
 .../Dialect/Polynomial/IR/PolynomialDialect.h |  19 ++
 .../Dialect/Polynomial/IR/PolynomialOps.h     |  21 ++
 .../Dialect/Polynomial/IR/PolynomialTypes.h   |  17 ++
 mlir/include/mlir/InitAllDialects.h           |   2 +
 mlir/lib/Dialect/CMakeLists.txt               |   1 +
 mlir/lib/Dialect/Polynomial/CMakeLists.txt    |   1 +
 mlir/lib/Dialect/Polynomial/IR/CMakeLists.txt |  19 ++
 mlir/lib/Dialect/Polynomial/IR/Polynomial.cpp |  96 ++++++++
 .../Polynomial/IR/PolynomialAttributes.cpp    | 213 ++++++++++++++++++
 .../Polynomial/IR/PolynomialDialect.cpp       |  41 ++++
 .../Dialect/Polynomial/IR/PolynomialOps.cpp   |  15 ++
 mlir/test/Dialect/Polynomial/attributes.mlir  |  45 ++++
 mlir/test/Dialect/Polynomial/types.mlir       |  42 ++++
 19 files changed, 844 insertions(+)
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/CMakeLists.txt
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.h
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.td
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.h
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/PolynomialDialect.h
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/PolynomialOps.h
 create mode 100644 mlir/include/mlir/Dialect/Polynomial/IR/PolynomialTypes.h
 create mode 100644 mlir/lib/Dialect/Polynomial/CMakeLists.txt
 create mode 100644 mlir/lib/Dialect/Polynomial/IR/CMakeLists.txt
 create mode 100644 mlir/lib/Dialect/Polynomial/IR/Polynomial.cpp
 create mode 100644 mlir/lib/Dialect/Polynomial/IR/PolynomialAttributes.cpp
 create mode 100644 mlir/lib/Dialect/Polynomial/IR/PolynomialDialect.cpp
 create mode 100644 mlir/lib/Dialect/Polynomial/IR/PolynomialOps.cpp
 create mode 100644 mlir/test/Dialect/Polynomial/attributes.mlir
 create mode 100644 mlir/test/Dialect/Polynomial/types.mlir

diff --git a/mlir/include/mlir/Dialect/CMakeLists.txt b/mlir/include/mlir/Dialect/CMakeLists.txt
index 2da79011fa26a..4bd7f12fabf7b 100644
--- a/mlir/include/mlir/Dialect/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/CMakeLists.txt
@@ -28,6 +28,7 @@ add_subdirectory(OpenACCMPCommon)
 add_subdirectory(OpenMP)
 add_subdirectory(PDL)
 add_subdirectory(PDLInterp)
+add_subdirectory(Polynomial)
 add_subdirectory(Quant)
 add_subdirectory(SCF)
 add_subdirectory(Shape)
diff --git a/mlir/include/mlir/Dialect/Polynomial/CMakeLists.txt b/mlir/include/mlir/Dialect/Polynomial/CMakeLists.txt
new file mode 100644
index 0000000000000..f33061b2d87cf
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/CMakeLists.txt
@@ -0,0 +1 @@
+add_subdirectory(IR)
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
new file mode 100644
index 0000000000000..d8039deb5ee21
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
@@ -0,0 +1,10 @@
+add_mlir_dialect(Polynomial polynomial)
+add_mlir_doc(PolynomialDialect PolynomialDialect Polynomial/ -gen-dialect-doc)
+add_mlir_doc(PolynomialOps PolynomialOps Polynomial/ -gen-op-doc)
+add_mlir_doc(PolynomialAttributes PolynomialAttributes Dialects/ -gen-attrdef-doc)
+add_mlir_doc(PolynomialTypes PolynomialTypes Dialects/ -gen-typedef-doc)
+
+set(LLVM_TARGET_DEFINITIONS Polynomial.td)
+mlir_tablegen(PolynomialAttributes.cpp.inc -gen-attrdef-defs -attrdefs-dialect=polynomial)
+mlir_tablegen(PolynomialAttributes.h.inc -gen-attrdef-decls -attrdefs-dialect=polynomial)
+add_public_tablegen_target(MLIRPolynomialAttributesIncGen)
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.h b/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.h
new file mode 100644
index 0000000000000..39b05b9d3ad14
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.h
@@ -0,0 +1,130 @@
+//===- Polynomial.h - A data class for polynomials --------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIAL_H_
+#define MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIAL_H_
+
+#include "mlir/Support/LLVM.h"
+#include "mlir/Support/LogicalResult.h"
+#include "llvm/ADT/APInt.h"
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/Hashing.h"
+#include "llvm/ADT/SmallVector.h"
+
+namespace mlir {
+
+class MLIRContext;
+
+namespace polynomial {
+
+/// This restricts statically defined polynomials to have at most 64-bit
+/// coefficients. This may be relaxed in the future, but it seems unlikely one
+/// would want to specify 128-bit polynomials statically in the source code.
+constexpr unsigned apintBitWidth = 64;
+
+/// A class representing a monomial of a single-variable polynomial with integer
+/// coefficients.
+class Monomial {
+public:
+  Monomial(int64_t coeff, uint64_t expo)
+      : coefficient(apintBitWidth, coeff), exponent(apintBitWidth, expo) {}
+
+  Monomial(const APInt &coeff, const APInt &expo)
+      : coefficient(coeff), exponent(expo) {}
+
+  Monomial() : coefficient(apintBitWidth, 0), exponent(apintBitWidth, 0) {}
+
+  bool operator==(const Monomial &other) const {
+    return other.coefficient == coefficient && other.exponent == exponent;
+  }
+  bool operator!=(const Monomial &other) const {
+    return other.coefficient != coefficient || other.exponent != exponent;
+  }
+
+  /// Monomials are ordered by exponent.
+  bool operator<(const Monomial &other) const {
+    return (exponent.ult(other.exponent));
+  }
+
+  // Prints polynomial to 'os'.
+  void print(raw_ostream &os) const;
+
+  friend ::llvm::hash_code hash_value(const Monomial &arg);
+
+public:
+  APInt coefficient;
+
+  // Always unsigned
+  APInt exponent;
+};
+
+/// A single-variable polynomial with integer coefficients.
+///
+/// Eg: x^1024 + x + 1
+///
+/// The symbols used as the polynomial's indeterminate don't matter, so long as
+/// it is used consistently throughout the polynomial.
+class Polynomial {
+public:
+  Polynomial() = delete;
+
+  explicit Polynomial(ArrayRef<Monomial> terms) : terms(terms){};
+
+  // Returns a Polynomial from a list of monomials.
+  // Fails if two monomials have the same exponent.
+  static FailureOr<Polynomial> fromMonomials(ArrayRef<Monomial> monomials);
+
+  /// Returns a polynomial with coefficients given by `coeffs`. The value
+  /// coeffs[i] is converted to a monomial with exponent i.
+  static Polynomial fromCoefficients(ArrayRef<int64_t> coeffs);
+
+  explicit operator bool() const { return !terms.empty(); }
+  bool operator==(const Polynomial &other) const {
+    return other.terms == terms;
+  }
+  bool operator!=(const Polynomial &other) const {
+    return !(other.terms == terms);
+  }
+
+  // Prints polynomial to 'os'.
+  void print(raw_ostream &os) const;
+  void print(raw_ostream &os, ::llvm::StringRef separator,
+             ::llvm::StringRef exponentiation) const;
+  void dump() const;
+
+  // Prints polynomial so that it can be used as a valid identifier
+  std::string toIdentifier() const;
+
+  unsigned getDegree() const;
+
+  friend ::llvm::hash_code hash_value(const Polynomial &arg);
+
+private:
+  // The monomial terms for this polynomial.
+  SmallVector<Monomial> terms;
+};
+
+// Make Polynomial hashable.
+inline ::llvm::hash_code hash_value(const Polynomial &arg) {
+  return ::llvm::hash_combine_range(arg.terms.begin(), arg.terms.end());
+}
+
+inline ::llvm::hash_code hash_value(const Monomial &arg) {
+  return llvm::hash_combine(::llvm::hash_value(arg.coefficient),
+                            ::llvm::hash_value(arg.exponent));
+}
+
+inline raw_ostream &operator<<(raw_ostream &os, const Polynomial &polynomial) {
+  polynomial.print(os);
+  return os;
+}
+
+} // namespace polynomial
+} // namespace mlir
+
+#endif // MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIAL_H_
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.td b/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.td
new file mode 100644
index 0000000000000..5d8da8399b01b
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/Polynomial.td
@@ -0,0 +1,153 @@
+//===- PolynomialOps.td - Polynomial dialect ---------------*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef POLYNOMIAL_OPS
+#define POLYNOMIAL_OPS
+
+include "mlir/IR/BuiltinAttributes.td"
+include "mlir/IR/OpBase.td"
+include "mlir/Interfaces/InferTypeOpInterface.td"
+include "mlir/Interfaces/SideEffectInterfaces.td"
+
+def Polynomial_Dialect : Dialect {
+  let name = "polynomial";
+  let cppNamespace = "::mlir::polynomial";
+  let description = [{
+    The Polynomial dialect defines single-variable polynomial types and
+    operations.
+
+    The simplest use of `polynomial` is to represent mathematical operations in
+    a polynomial ring `R[x]`, where `R` is another MLIR type like `i32`.
+
+    More generally, this dialect supports representing polynomial operations in a
+    quotient ring `R[X]/(f(x))` for some statically fixed polynomial `f(x)`.
+    Two polyomials `p(x), q(x)` are considered equal in this ring if they have the
+    same remainder when dividing by `f(x)`. When a modulus is given, ring operations
+    are performed with reductions modulo `f(x)` and relative to the coefficient ring
+    `R`.
+
+    Examples:
+
+    ```mlir
+    // A constant polynomial in a ring with i32 coefficients and no polynomial modulus
+    #ring = #polynomial.ring<ctype=i32>
+    %a = polynomial.constant <1 + x**2 - 3x**3> : polynomial.polynomial<#ring>
+
+    // A constant polynomial in a ring with i32 coefficients, modulo (x^1024 + 1)
+    #modulus = #polynomial.polynomial<1 + x**1024>
+    #ring = #polynomial.ring<ctype=i32, ideal=#modulus>
+    %a = polynomial.constant <1 + x**2 - 3x**3> : polynomial.polynomial<#ring>
+
+    // A constant polynomial in a ring with i32 coefficients, with a polynomial
+    // modulus of (x^1024 + 1) and a coefficient modulus of 17.
+    #modulus = #polynomial.polynomial<1 + x**1024>
+    #ring = #polynomial.ring<ctype=i32, cmod=17, ideal=#modulus>
+    %a = polynomial.constant <1 + x**2 - 3x**3> : polynomial.polynomial<#ring>
+    ```
+  }];
+
+  let useDefaultTypePrinterParser = 1;
+  let useDefaultAttributePrinterParser = 1;
+}
+
+class Polynomial_Attr<string name, string attrMnemonic, list<Trait> traits = []>
+    : AttrDef<Polynomial_Dialect, name, traits> {
+  let mnemonic = attrMnemonic;
+}
+
+def Polynomial_PolynomialAttr : Polynomial_Attr<"Polynomial", "polynomial"> {
+  let summary = "An attribute containing a single-variable polynomial.";
+  let description = [{
+     #poly = #polynomial.poly<x**1024 + 1>
+  }];
+  let parameters = (ins "Polynomial":$polynomial);
+  let hasCustomAssemblyFormat = 1;
+}
+
+def Polynomial_RingAttr : Polynomial_Attr<"Ring", "ring"> {
+  let summary = "An attribute specifying a polynomial ring.";
+  let description = [{
+    A ring describes the domain in which polynomial arithmetic occurs. The ring
+    attribute in `polynomial` represents the more specific case of polynomials
+    with a single indeterminate; whose coefficients can be represented by
+    another MLIR type (`coefficientType`); and, if the coefficient type is
+    integral, whose coefficients are taken modulo some statically known modulus
+    (`coefficientModulus`).
+
+    Additionally, a polynomial ring can specify an _ideal_, which converts
+    polynomial arithmetic to the analogue of modular integer arithmetic, where
+    each polynomial is represented as its remainder when dividing by the
+    modulus. For single-variable polynomials, an "ideal" is always specificed
+    via a single polynomial, which we call `polynomialModulus`.
+
+    An expressive example is polynomials with i32 coefficients, whose
+    coefficients are taken modulo `2**32 - 5`, with a polynomial modulus of
+    `x**1024 - 1`.
+
+    ```mlir
+    #poly_mod = #polynomial.polynomial<-1 + x**1024>
+    #ring = #polynomial.ring<coefficientType=i32,
+                             coefficientModulus=4294967291,
+                             polynomialModulus=#poly_mod>
+
+    %0 = ... : polynomial.polynomial<#ring>
+    ```
+
+    In this case, the value of a polynomial is always "converted" to a
+    canonical form by applying repeated reductions by setting `x**1024 = 1`
+    and simplifying.
+
+    The coefficient and polynomial modulus parameters are optional, and the
+    coefficient modulus is only allowed if the coefficient type is integral.
+  }];
+
+  let parameters = (ins
+    "Type": $coefficientType,
+    OptionalParameter<"IntegerAttr">: $coefficientModulus,
+    OptionalParameter<"PolynomialAttr">: $polynomialModulus
+  );
+
+  let hasCustomAssemblyFormat = 1;
+}
+
+class Polynomial_Type<string name, string typeMnemonic>
+    : TypeDef<Polynomial_Dialect, name> {
+  let mnemonic = typeMnemonic;
+}
+
+def Polynomial_PolynomialType : Polynomial_Type<"Polynomial", "polynomial"> {
+  let summary = "An element of a polynomial ring.";
+
+  let description = [{
+    A type for polynomials in a polynomial quotient ring.
+  }];
+
+  let parameters = (ins Polynomial_RingAttr:$ring);
+  let assemblyFormat = "`<` $ring `>`";
+}
+
+class Polynomial_Op<string mnemonic, list<Trait> traits = []> :
+    Op<Polynomial_Dialect, mnemonic, traits # [Pure]>;
+
+class Polynomial_UnaryOp<string mnemonic, list<Trait> traits = []> :
+    Polynomial_Op<mnemonic, traits # [SameOperandsAndResultType]> {
+  let arguments = (ins Polynomial_PolynomialType:$operand);
+  let results = (outs Polynomial_PolynomialType:$result);
+
+  let assemblyFormat = "$operand attr-dict `:` qualified(type($result))";
+}
+
+class Polynomial_BinaryOp<string mnemonic, list<Trait> traits = []> :
+    Polynomial_Op<mnemonic, traits # [SameOperandsAndResultType]> {
+  let arguments = (ins Polynomial_PolynomialType:$lhs, Polynomial_PolynomialType:$rhs);
+  let results = (outs Polynomial_PolynomialType:$result);
+
+  let assemblyFormat = "$lhs `,` $rhs attr-dict `:` qualified(type($result))";
+}
+
+#endif // POLYNOMIAL_OPS
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.h b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.h
new file mode 100644
index 0000000000000..b37d17bb89fb2
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.h
@@ -0,0 +1,17 @@
+//===- PolynomialAttributes.h - polynomial dialect attributes ---*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALATTRIBUTES_H_
+#define MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALATTRIBUTES_H_
+
+#include "Polynomial.h"
+#include "PolynomialDialect.h"
+
+#define GET_ATTRDEF_CLASSES
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.h.inc"
+
+#endif // MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALATTRIBUTES_H_
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialDialect.h b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialDialect.h
new file mode 100644
index 0000000000000..7b7acebe7a93b
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialDialect.h
@@ -0,0 +1,19 @@
+//===- PolynomialDialect.h - The Polynomial dialect -------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALDIALECT_H_
+#define MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALDIALECT_H_
+
+#include "mlir/IR/Builders.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/Dialect.h"
+#include "mlir/IR/DialectImplementation.h"
+
+// Generated headers (block clang-format from messing up order)
+#include "mlir/Dialect/Polynomial/IR/PolynomialDialect.h.inc"
+
+#endif // MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALDIALECT_H_
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialOps.h b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialOps.h
new file mode 100644
index 0000000000000..bacaad81ce8e5
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialOps.h
@@ -0,0 +1,21 @@
+//===- PolynomialOps.h - Ops for the Polynomial dialect ---------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALOPS_H_
+#define MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALOPS_H_
+
+#include "PolynomialDialect.h"
+#include "PolynomialTypes.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/Dialect.h"
+#include "mlir/Interfaces/InferTypeOpInterface.h"
+
+#define GET_OP_CLASSES
+#include "mlir/Dialect/Polynomial/IR/Polynomial.h.inc"
+
+#endif // MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALOPS_H_
diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialTypes.h b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialTypes.h
new file mode 100644
index 0000000000000..2fc6877452547
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/PolynomialTypes.h
@@ -0,0 +1,17 @@
+//===- PolynomialTypes.h - Types for the Polynomial dialect -----*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALTYPES_H_
+#define MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALTYPES_H_
+
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.h"
+#include "mlir/Dialect/Polynomial/IR/PolynomialDialect.h"
+
+#define GET_TYPEDEF_CLASSES
+#include "mlir/Dialect/Polynomial/IR/PolynomialTypes.h.inc"
+
+#endif // MLIR_INCLUDE_MLIR_DIALECT_POLYNOMIAL_IR_POLYNOMIALTYPES_H_
diff --git a/mlir/include/mlir/InitAllDialects.h b/mlir/include/mlir/InitAllDialects.h
index c558dc53cc7fa..c4d788cf8ed31 100644
--- a/mlir/include/mlir/InitAllDialects.h
+++ b/mlir/include/mlir/InitAllDialects.h
@@ -61,6 +61,7 @@
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
 #include "mlir/Dialect/PDL/IR/PDL.h"
 #include "mlir/Dialect/PDLInterp/IR/PDLInterp.h"
+#include "mlir/Dialect/Polynomial/IR/PolynomialDialect.h"
 #include "mlir/Dialect/Quant/QuantOps.h"
 #include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.h"
@@ -131,6 +132,7 @@ inline void registerAllDialects(DialectRegistry &registry) {
                   omp::OpenMPDialect,
                   pdl::PDLDialect,
                   pdl_interp::PDLInterpDialect,
+                  polynomial::PolynomialDialect,
                   quant::QuantizationDialect,
                   ROCDL::ROCDLDialect,
                   scf::SCFDialect,
diff --git a/mlir/lib/Dialect/CMakeLists.txt b/mlir/lib/Dialect/CMakeLists.txt
index b1ba5a3bc8817..a324ce7f9b19f 100644
--- a/mlir/lib/Dialect/CMakeLists.txt
+++ b/mlir/lib/Dialect/CMakeLists.txt
@@ -28,6 +28,7 @@ add_subdirectory(OpenACCMPCommon)
 add_subdirectory(OpenMP)
 add_subdirectory(PDL)
 add_subdirectory(PDLInterp)
+add_subdirectory(Polynomial)
 add_subdirectory(Quant)
 add_subdirectory(SCF)
 add_subdirectory(Shape)
diff --git a/mlir/lib/Dialect/Polynomial/CMakeLists.txt b/mlir/lib/Dialect/Polynomial/CMakeLists.txt
new file mode 100644
index 0000000000000..f33061b2d87cf
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/CMakeLists.txt
@@ -0,0 +1 @@
+add_subdirectory(IR)
diff --git a/mlir/lib/Dialect/Polynomial/IR/CMakeLists.txt b/mlir/lib/Dialect/Polynomial/IR/CMakeLists.txt
new file mode 100644
index 0000000000000..7f5b3255d5d90
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/IR/CMakeLists.txt
@@ -0,0 +1,19 @@
+add_mlir_dialect_library(MLIRPolynomialDialect
+  Polynomial.cpp
+  PolynomialAttributes.cpp
+  PolynomialDialect.cpp
+  PolynomialOps.cpp
+
+  ADDITIONAL_HEADER_DIRS
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Polynomial
+
+  DEPENDS
+  MLIRPolynomialIncGen
+  MLIRPolynomialAttributesIncGen
+  MLIRBuiltinAttributesIncGen
+
+  LINK_LIBS PUBLIC
+  MLIRSupport
+  MLIRDialect
+  MLIRIR
+  )
diff --git a/mlir/lib/Dialect/Polynomial/IR/Polynomial.cpp b/mlir/lib/Dialect/Polynomial/IR/Polynomial.cpp
new file mode 100644
index 0000000000000..5916ffba78e24
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/IR/Polynomial.cpp
@@ -0,0 +1,96 @@
+//===- Polynomial.cpp - MLIR storage type for static Polynomial -*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Polynomial/IR/Polynomial.h"
+
+#include "mlir/Support/LogicalResult.h"
+#include "llvm/ADT/APInt.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/Twine.h"
+#include "llvm/Support/raw_ostream.h"
+
+namespace mlir {
+namespace polynomial {
+
+FailureOr<Polynomial> Polynomial::fromMonomials(ArrayRef<Monomial> monomials) {
+  // A polynomial's terms are canonically stored in order of increasing degree.
+  auto monomialsCopy = llvm::SmallVector<Monomial>(monomials);
+  std::sort(monomialsCopy.begin(), monomialsCopy.end());
+
+  // Ensure non-unique exponents are not present. Since we sorted the list by
+  // exponent, a linear scan of adjancent monomials suffices.
+  if (std::adjacent_find(monomialsCopy.begin(), monomialsCopy.end(),
+                         [](const Monomial &lhs, const Monomial &rhs) {
+                           return lhs.exponent == rhs.exponent;
+                         }) != monomialsCopy.end()) {
+    return failure();
+  }
+
+  return Polynomial(monomialsCopy);
+}
+
+Polynomial Polynomial::fromCoefficients(ArrayRef<int64_t> coeffs) {
+  llvm::SmallVector<Monomial> monomials;
+  auto size = coeffs.size();
+  monomials.reserve(size);
+  for (size_t i = 0; i < size; i++) {
+    monomials.emplace_back(coeffs[i], i);
+  }
+  auto result = Polynomial::fromMonomials(monomials);
+  // Construction guarantees unique exponents, so the failure mode of
+  // fromMonomials can be bypassed.
+  assert(succeeded(result));
+  return result.value();
+}
+
+void Polynomial::print(raw_ostream &os, ::llvm::StringRef separator,
+                       ::llvm::StringRef exponentiation) const {
+  bool first = true;
+  for (const Monomial &term : terms) {
+    if (first) {
+      first = false;
+    } else {
+      os << separator;
+    }
+    std::string coeffToPrint;
+    if (term.coefficient == 1 && term.exponent.uge(1)) {
+      coeffToPrint = "";
+    } else {
+      llvm::SmallString<16> coeffString;
+      term.coefficient.toStringSigned(coeffString);
+      coeffToPrint = coeffString.str();
+    }
+
+    if (term.exponent == 0) {
+      os << coeffToPrint;
+    } else if (term.exponent == 1) {
+      os << coeffToPrint << "x";
+    } else {
+      llvm::SmallString<16> expString;
+      term.exponent.toStringSigned(expString);
+      os << coeffToPrint << "x" << exponentiation << expString;
+    }
+  }
+}
+
+void Polynomial::print(raw_ostream &os) const { print(os, " + ", "**"); }
+
+std::string Polynomial::toIdentifier() const {
+  std::string result;
+  llvm::raw_string_ostream os(result);
+  print(os, "_", "");
+  return os.str();
+}
+
+unsigned Polynomial::getDegree() const {
+  return terms.back().exponent.getZExtValue();
+}
+
+} // namespace polynomial
+} // namespace mlir
diff --git a/mlir/lib/Dialect/Polynomial/IR/PolynomialAttributes.cpp b/mlir/lib/Dialect/Polynomial/IR/PolynomialAttributes.cpp
new file mode 100644
index 0000000000000..ee09c73bb3c4a
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/IR/PolynomialAttributes.cpp
@@ -0,0 +1,213 @@
+//===- PolynomialAttributes.cpp - Polynomial dialect attrs ------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.h"
+
+#include "mlir/Dialect/Polynomial/IR/Polynomial.h"
+#include "mlir/Support/LLVM.h"
+#include "mlir/Support/LogicalResult.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/StringSet.h"
+
+namespace mlir {
+namespace polynomial {
+
+void PolynomialAttr::print(AsmPrinter &p) const {
+  p << '<';
+  p << getPolynomial();
+  p << '>';
+}
+
+/// Try to parse a monomial. If successful, populate the fields of the outparam
+/// `monomial` with the results, and the `variable` outparam with the parsed
+/// variable name. Sets shouldParseMore to true if the monomial is followed by
+/// a '+'.
+ParseResult parseMonomial(AsmParser &parser, Monomial &monomial,
+                          llvm::StringRef &variable, bool &isConstantTerm,
+                          bool &shouldParseMore) {
+  APInt parsedCoeff(apintBitWidth, 1);
+  auto parsedCoeffResult = parser.parseOptionalInteger(parsedCoeff);
+  monomial.coefficient = parsedCoeff;
+
+  isConstantTerm = false;
+  shouldParseMore = false;
+
+  // A + indicates it's a constant term with more to go, as in `1 + x`.
+  if (succeeded(parser.parseOptionalPlus())) {
+    // If no coefficient was parsed, and there's a +, then it's effectively
+    // parsing an empty string.
+    if (!parsedCoeffResult.has_value()) {
+      return failure();
+    }
+    monomial.exponent = APInt(apintBitWidth, 0);
+    isConstantTerm = true;
+    shouldParseMore = true;
+    return success();
+  }
+
+  // A monomial can be a trailing constant term, as in `x + 1`.
+  if (failed(parser.parseOptionalKeyword(&variable))) {
+    // If neither a coefficient nor a variable was found, then it's effectively
+    // parsing an empty string.
+    if (!parsedCoeffResult.has_value()) {
+      return failure();
+    }
+
+    monomial.exponent = APInt(apintBitWidth, 0);
+    isConstantTerm = true;
+    return success();
+  }
+
+  // Parse exponentiation symbol as `**`. We can't use caret because it's
+  // reserved for basic block identifiers If no star is present, it's treated
+  // as a polynomial with exponent 1.
+  if (succeeded(parser.parseOptionalStar())) {
+    // If there's one * there must be two.
+    if (failed(parser.parseStar())) {
+      return failure();
+    }
+
+    // If there's a **, then the integer exponent is required.
+    APInt parsedExponent(apintBitWidth, 0);
+    if (failed(parser.parseInteger(parsedExponent))) {
+      parser.emitError(parser.getCurrentLocation(),
+                       "found invalid integer exponent");
+      return failure();
+    }
+
+    monomial.exponent = parsedExponent;
+  } else {
+    monomial.exponent = APInt(apintBitWidth, 1);
+  }
+
+  if (succeeded(parser.parseOptionalPlus())) {
+    shouldParseMore = true;
+  }
+  return success();
+}
+
+Attribute PolynomialAttr::parse(AsmParser &parser, Type type) {
+  if (failed(parser.parseLess()))
+    return {};
+
+  llvm::SmallVector<Monomial> monomials;
+  llvm::StringSet<> variables;
+
+  while (true) {
+    Monomial parsedMonomial;
+    llvm::StringRef parsedVariableRef;
+    bool isConstantTerm;
+    bool shouldParseMore;
+    if (failed(parseMonomial(parser, parsedMonomial, parsedVariableRef,
+                             isConstantTerm, shouldParseMore))) {
+      parser.emitError(parser.getCurrentLocation(), "expected a monomial");
+      return {};
+    }
+
+    if (!isConstantTerm) {
+      std::string parsedVariable = parsedVariableRef.str();
+      variables.insert(parsedVariable);
+    }
+    monomials.push_back(parsedMonomial);
+
+    if (shouldParseMore)
+      continue;
+
+    if (succeeded(parser.parseOptionalGreater())) {
+      break;
+    }
+    parser.emitError(
+        parser.getCurrentLocation(),
+        "expected + and more monomials, or > to end polynomial attribute");
+    return {};
+  }
+
+  if (variables.size() > 1) {
+    std::string vars = llvm::join(variables.keys(), ", ");
+    parser.emitError(
+        parser.getCurrentLocation(),
+        "polynomials must have one indeterminate, but there were multiple: " +
+            vars);
+  }
+
+  auto result = Polynomial::fromMonomials(monomials);
+  if (failed(result)) {
+    parser.emitError(parser.getCurrentLocation())
+        << "parsed polynomial must have unique exponents among monomials";
+    return {};
+  }
+  return PolynomialAttr::get(parser.getContext(), result.value());
+}
+
+void RingAttr::print(AsmPrinter &p) const {
+  p << "#polynomial.ring<coefficientType=" << getCoefficientType()
+    << ", coefficientModulus=" << getCoefficientModulus()
+    << ", polynomialModulus=" << getPolynomialModulus() << '>';
+}
+
+Attribute RingAttr::parse(AsmParser &parser, Type type) {
+  if (failed(parser.parseLess()))
+    return {};
+
+  if (failed(parser.parseKeyword("coefficientType")))
+    return {};
+
+  if (failed(parser.parseEqual()))
+    return {};
+
+  Type ty;
+  if (failed(parser.parseType(ty)))
+    return {};
+
+  if (failed(parser.parseComma()))
+    return {};
+
+  IntegerAttr coefficientModulusAttr = nullptr;
+  if (succeeded(parser.parseKeyword("coefficientModulus"))) {
+    if (failed(parser.parseEqual()))
+      return {};
+
+    IntegerType iType = ty.dyn_cast<IntegerType>();
+    if (!iType) {
+      parser.emitError(parser.getCurrentLocation(),
+                       "coefficientType must specify an integer type");
+      return {};
+    }
+    APInt coefficientModulus(iType.getWidth(), 0);
+    auto result = parser.parseInteger(coefficientModulus);
+    if (failed(result)) {
+      parser.emitError(parser.getCurrentLocation(),
+                       "invalid coefficient modulus");
+      return {};
+    }
+    coefficientModulusAttr = IntegerAttr::get(iType, coefficientModulus);
+
+    if (failed(parser.parseComma()))
+      return {};
+  }
+
+  PolynomialAttr polyAttr = nullptr;
+  if (succeeded(parser.parseKeyword("polynomialModulus"))) {
+    if (failed(parser.parseEqual()))
+      return {};
+
+    PolynomialAttr attr;
+    if (failed(parser.parseAttribute<PolynomialAttr>(attr)))
+      return {};
+    polyAttr = attr;
+  }
+
+  if (failed(parser.parseGreater()))
+    return {};
+
+  return RingAttr::get(parser.getContext(), ty, coefficientModulusAttr,
+                       polyAttr);
+}
+
+} // namespace polynomial
+} // namespace mlir
diff --git a/mlir/lib/Dialect/Polynomial/IR/PolynomialDialect.cpp b/mlir/lib/Dialect/Polynomial/IR/PolynomialDialect.cpp
new file mode 100644
index 0000000000000..a672a59b8a465
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/IR/PolynomialDialect.cpp
@@ -0,0 +1,41 @@
+//===- PolynomialDialect.cpp - Polynomial dialect ---------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Polynomial/IR/Polynomial.h"
+
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.h"
+#include "mlir/Dialect/Polynomial/IR/PolynomialOps.h"
+#include "mlir/Dialect/Polynomial/IR/PolynomialTypes.h"
+#include "llvm/ADT/TypeSwitch.h"
+
+using namespace mlir;
+using namespace mlir::polynomial;
+
+#include "mlir/Dialect/Polynomial/IR/PolynomialDialect.cpp.inc"
+
+#define GET_ATTRDEF_CLASSES
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.cpp.inc"
+#define GET_TYPEDEF_CLASSES
+#include "mlir/Dialect/Polynomial/IR/PolynomialTypes.cpp.inc"
+#define GET_OP_CLASSES
+#include "mlir/Dialect/Polynomial/IR/Polynomial.cpp.inc"
+
+void PolynomialDialect::initialize() {
+  addAttributes<
+#define GET_ATTRDEF_LIST
+#include "mlir/Dialect/Polynomial/IR/PolynomialAttributes.cpp.inc"
+      >();
+  addTypes<
+#define GET_TYPEDEF_LIST
+#include "mlir/Dialect/Polynomial/IR/PolynomialTypes.cpp.inc"
+      >();
+  addOperations<
+#define GET_OP_LIST
+#include "mlir/Dialect/Polynomial/IR/Polynomial.cpp.inc"
+      >();
+}
diff --git a/mlir/lib/Dialect/Polynomial/IR/PolynomialOps.cpp b/mlir/lib/Dialect/Polynomial/IR/PolynomialOps.cpp
new file mode 100644
index 0000000000000..96c59a28b8fdc
--- /dev/null
+++ b/mlir/lib/Dialect/Polynomial/IR/PolynomialOps.cpp
@@ -0,0 +1,15 @@
+//===- PolynomialOps.cpp - Polynomial dialect ops ---------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Polynomial/IR/Polynomial.h"
+
+using namespace mlir;
+using namespace mlir::polynomial;
+
+#define GET_OP_CLASSES
+#include "mlir/Dialect/Polynomial/IR/Polynomial.cpp.inc"
diff --git a/mlir/test/Dialect/Polynomial/attributes.mlir b/mlir/test/Dialect/Polynomial/attributes.mlir
new file mode 100644
index 0000000000000..3973ae3944335
--- /dev/null
+++ b/mlir/test/Dialect/Polynomial/attributes.mlir
@@ -0,0 +1,45 @@
+// RUN: mlir-opt %s --split-input-file --verify-diagnostics
+
+#my_poly = #polynomial.polynomial<y + x**1024>
+// expected-error@below {{polynomials must have one indeterminate, but there were multiple: x, y}}
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+// -----
+
+// expected-error@below {{expected integer value}}
+// expected-error@below {{expected a monomial}}
+// expected-error@below {{found invalid integer exponent}}
+#my_poly = #polynomial.polynomial<5 + x**f>
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+// -----
+
+#my_poly = #polynomial.polynomial<5 + x**2 + 3x**2>
+// expected-error@below {{parsed polynomial must have unique exponents among monomials}}
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+// -----
+
+// expected-error@below {{expected + and more monomials, or > to end polynomial attribute}}
+#my_poly = #polynomial.polynomial<5 + x**2 7>
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+// -----
+
+// expected-error@below {{expected a monomial}}
+#my_poly = #polynomial.polynomial<5 + x**2 +>
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+
+// -----
+
+#my_poly = #polynomial.polynomial<5 + x**2>
+// expected-error@below {{coefficientType must specify an integer type}}
+#ring1 = #polynomial.ring<coefficientType=f64, coefficientModulus=2837465, polynomialModulus=#my_poly>
+
+// -----
+
+#my_poly = #polynomial.polynomial<5 + x**2>
+// expected-error@below {{expected integer value}}
+// expected-error@below {{invalid coefficient modulus}}
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=x, polynomialModulus=#my_poly>
diff --git a/mlir/test/Dialect/Polynomial/types.mlir b/mlir/test/Dialect/Polynomial/types.mlir
new file mode 100644
index 0000000000000..64b74d9d36bb1
--- /dev/null
+++ b/mlir/test/Dialect/Polynomial/types.mlir
@@ -0,0 +1,42 @@
+// RUN: mlir-opt %s | FileCheck %s
+
+// CHECK-LABEL: func @test_types
+// CHECK-SAME:  !polynomial.polynomial<
+// CHECK-SAME:    #polynomial.ring<
+// CHECK-SAME:       coefficientType=i32,
+// CHECK-SAME:       coefficientModulus=2837465 : i32,
+// CHECK-SAME:       polynomialModulus=#polynomial.polynomial<1 + x**1024>>>
+#my_poly = #polynomial.polynomial<1 + x**1024>
+#ring1 = #polynomial.ring<coefficientType=i32, coefficientModulus=2837465, polynomialModulus=#my_poly>
+!ty = !polynomial.polynomial<#ring1>
+func.func @test_types(%0: !ty) -> !ty {
+  return %0 : !ty
+}
+
+
+// CHECK-LABEL: func @test_non_x_variable_64_bit
+// CHECK-SAME:  !polynomial.polynomial<
+// CHECK-SAME:    #polynomial.ring<
+// CHECK-SAME:       coefficientType=i64,
+// CHECK-SAME:       coefficientModulus=2837465 : i64,
+// CHECK-SAME:       polynomialModulus=#polynomial.polynomial<2 + 4x + x**3>>>
+#my_poly_2 = #polynomial.polynomial<t**3 + 4t + 2>
+#ring2 = #polynomial.ring<coefficientType=i64, coefficientModulus=2837465, polynomialModulus=#my_poly_2>
+!ty2 = !polynomial.polynomial<#ring2>
+func.func @test_non_x_variable_64_bit(%0: !ty2) -> !ty2 {
+  return %0 : !ty2
+}
+
+
+// CHECK-LABEL: func @test_linear_poly
+// CHECK-SAME:  !polynomial.polynomial<
+// CHECK-SAME:    #polynomial.ring<
+// CHECK-SAME:       coefficientType=i32,
+// CHECK-SAME:       coefficientModulus=12 : i32,
+// CHECK-SAME:       polynomialModulus=#polynomial.polynomial<4x>>
+#my_poly_3 = #polynomial.polynomial<4x>
+#ring3 = #polynomial.ring<coefficientType=i32, coefficientModulus=12, polynomialModulus=#my_poly_3>
+!ty3 = !polynomial.polynomial<#ring3>
+func.func @test_linear_poly(%0: !ty3) -> !ty3 {
+  return %0 : !ty3
+}

From f8d314f0ee5e750e1e47b47fb55d14f6d1e991e1 Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Mon, 15 Apr 2024 18:14:18 +0200
Subject: [PATCH 444/886] [mlir][Interfaces][NFC] Add TableGen test op for
 value bounds tests (#88717)

This commit is a code cleanup. It defines the test ops the are used for
the `ValueBoundsOpInterface` tests in TableGen, along with proper
verifiers.

---------

Co-authored-by: Benjamin Maxwell <benjamin.maxwell@arm.com>
---
 .../value-bounds-op-interface-impl.mlir       |   2 +-
 .../Affine/value-bounds-reification.mlir      |   6 +-
 .../value-bounds-op-interface-impl.mlir       |   4 +-
 .../Dialect/Vector/test-scalable-bounds.mlir  |  18 +-
 mlir/test/lib/Dialect/Affine/CMakeLists.txt   |   8 +
 .../Dialect/Affine/TestReifyValueBounds.cpp   | 305 ++++++------------
 mlir/test/lib/Dialect/Test/CMakeLists.txt     |   1 +
 mlir/test/lib/Dialect/Test/TestDialect.cpp    |  48 +++
 mlir/test/lib/Dialect/Test/TestDialect.h      |   1 +
 mlir/test/lib/Dialect/Test/TestOps.td         |  47 +++
 .../mlir/test/BUILD.bazel                     |   2 +
 11 files changed, 226 insertions(+), 216 deletions(-)

diff --git a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
index 10da91870f49d..23c6872dcebe9 100644
--- a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
@@ -74,7 +74,7 @@ func.func @composed_affine_apply(%i1 : index) -> (index) {
   %i2 = affine.apply affine_map<(d0) -> ((d0 floordiv 32) * 16)>(%i1)
   %i3 = affine.apply affine_map<(d0) -> ((d0 floordiv 32) * 16 + 8)>(%i1)
   %s = affine.apply affine_map<()[s0, s1] -> (s0 - s1)>()[%i2, %i3]
-  %reified = "test.reify_constant_bound"(%s) {type = "EQ"} : (index) -> (index)
+  %reified = "test.reify_bound"(%s) {type = "EQ", constant} : (index) -> (index)
   return %reified : index
 }
 
diff --git a/mlir/test/Dialect/Affine/value-bounds-reification.mlir b/mlir/test/Dialect/Affine/value-bounds-reification.mlir
index 909c9098c5160..75622f59af83b 100644
--- a/mlir/test/Dialect/Affine/value-bounds-reification.mlir
+++ b/mlir/test/Dialect/Affine/value-bounds-reification.mlir
@@ -47,7 +47,7 @@ func.func @reify_slice_bound(%t: tensor<?x?xi32>, %idx: index, %ub: index, %f: f
     %bound = "test.reify_bound"(%filled) {dim = 1, type = "UB"} : (tensor<1x?xi32>) -> (index)
     "test.some_use"(%bound) : (index) -> ()
 
-    %bound_const = "test.reify_constant_bound"(%filled) {dim = 1, type = "UB"} : (tensor<1x?xi32>) -> (index)
+    %bound_const = "test.reify_bound"(%filled) {dim = 1, type = "UB", constant} : (tensor<1x?xi32>) -> (index)
     "test.some_use"(%bound_const) : (index) -> ()
   }
   return
@@ -93,7 +93,7 @@ func.func @reify_slice_bound2(%lb0: index, %ub0: index, %step0: index,
 
     // CHECK: %[[c129:.*]] = arith.constant 129 : index
     // CHECK: "test.some_use"(%[[c129]])
-    %lb1_ub_const = "test.reify_constant_bound"(%lb1) {type = "UB"} : (index) -> (index)
+    %lb1_ub_const = "test.reify_bound"(%lb1) {type = "UB", constant} : (index) -> (index)
     "test.some_use"(%lb1_ub_const) : (index) -> ()
 
     scf.for %iv1 = %lb1 to %ub1 step %c32 {
@@ -116,7 +116,7 @@ func.func @reify_slice_bound2(%lb0: index, %ub0: index, %step0: index,
 
         // CHECK: %[[c32:.*]] = arith.constant 32 : index
         // CHECK: "test.some_use"(%[[c32]])
-        %matmul_ub_const = "test.reify_constant_bound"(%matmul) {dim = 1, type = "UB"} : (tensor<1x?xi32>) -> (index)
+        %matmul_ub_const = "test.reify_bound"(%matmul) {dim = 1, type = "UB", constant} : (tensor<1x?xi32>) -> (index)
         "test.some_use"(%matmul_ub_const) : (index) -> ()
       }
     }
diff --git a/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
index 0c90bcdb42028..0ba9983723a0a 100644
--- a/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/Tensor/value-bounds-op-interface-impl.mlir
@@ -83,7 +83,7 @@ func.func @extract_slice_static(%t: tensor<?xf32>) -> index {
 func.func @extract_slice_dynamic_constant(%t: tensor<?xf32>, %sz: index) -> index {
   %0 = tensor.extract_slice %t[2][%sz][1] : tensor<?xf32> to tensor<?xf32>
   // expected-error @below{{could not reify bound}}
-  %1 = "test.reify_constant_bound"(%0) {dim = 0} : (tensor<?xf32>) -> (index)
+  %1 = "test.reify_bound"(%0) {dim = 0, constant} : (tensor<?xf32>) -> (index)
   return %1 : index
 }
 
@@ -95,7 +95,7 @@ func.func @extract_slice_dynamic_constant(%t: tensor<?xf32>, %sz: index) -> inde
 //       CHECK:   return %[[c5]]
 func.func @extract_slice_static_constant(%t: tensor<?xf32>) -> index {
   %0 = tensor.extract_slice %t[2][5][1] : tensor<?xf32> to tensor<5xf32>
-  %1 = "test.reify_constant_bound"(%0) {dim = 0} : (tensor<5xf32>) -> (index)
+  %1 = "test.reify_bound"(%0) {dim = 0, constant} : (tensor<5xf32>) -> (index)
   return %1 : index
 }
 
diff --git a/mlir/test/Dialect/Vector/test-scalable-bounds.mlir b/mlir/test/Dialect/Vector/test-scalable-bounds.mlir
index 245a6f5c13ac3..d549c5bd1c378 100644
--- a/mlir/test/Dialect/Vector/test-scalable-bounds.mlir
+++ b/mlir/test/Dialect/Vector/test-scalable-bounds.mlir
@@ -26,8 +26,8 @@ func.func @fixed_size_loop_nest() {
     %min_i = affine.min #map_dim_i(%i)[%c4_vscale]
     scf.for %j = %c0 to %c16 step %c4_vscale {
       %min_j = affine.min #map_dim_j(%j)[%c4_vscale]
-      %bound_i = "test.reify_scalable_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
-      %bound_j = "test.reify_scalable_bound"(%min_j) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
+      %bound_i = "test.reify_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
+      %bound_j = "test.reify_bound"(%min_j) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
       "test.some_use"(%bound_i, %bound_j) : (index, index) -> ()
     }
   }
@@ -58,8 +58,8 @@ func.func @dynamic_size_loop_nest(%dim0: index, %dim1: index) {
     %min_i = affine.min #map_dynamic_dim(%i)[%c4_vscale, %dim0]
     scf.for %j = %c0 to %dim1 step %c4_vscale {
       %min_j = affine.min #map_dynamic_dim(%j)[%c4_vscale, %dim1]
-      %bound_i = "test.reify_scalable_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
-      %bound_j = "test.reify_scalable_bound"(%min_j) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
+      %bound_i = "test.reify_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
+      %bound_j = "test.reify_bound"(%min_j) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
       "test.some_use"(%bound_i, %bound_j) : (index, index) -> ()
     }
   }
@@ -80,7 +80,7 @@ func.func @add_to_vscale() {
   %vscale = vector.vscale
   %c8 = arith.constant 8 : index
   %vscale_plus_c8 = arith.addi %vscale, %c8 : index
-  %bound = "test.reify_scalable_bound"(%vscale_plus_c8) {type = "EQ", vscale_min = 1, vscale_max = 16} : (index) -> index
+  %bound = "test.reify_bound"(%vscale_plus_c8) {type = "EQ", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
   "test.some_use"(%bound) : (index) -> ()
   return
 }
@@ -94,7 +94,7 @@ func.func @add_to_vscale() {
 //       CHECK:   "test.some_use"(%[[C2]]) : (index) -> ()
 func.func @vscale_fixed_size() {
   %vscale = vector.vscale
-  %bound = "test.reify_scalable_bound"(%vscale) {type = "EQ", vscale_min = 2, vscale_max = 2} : (index) -> index
+  %bound = "test.reify_bound"(%vscale) {type = "EQ", vscale_min = 2, vscale_max = 2, scalable} : (index) -> index
   "test.some_use"(%bound) : (index) -> ()
   return
 }
@@ -107,7 +107,7 @@ func.func @unknown_bound(%a: index) {
   %vscale = vector.vscale
   %vscale_plus_a = arith.muli %vscale, %a : index
   // expected-error @below{{could not reify bound}}
-  %bound = "test.reify_scalable_bound"(%vscale_plus_a) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
+  %bound = "test.reify_bound"(%vscale_plus_a) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
   "test.some_use"(%bound) : (index) -> ()
   return
 }
@@ -134,7 +134,7 @@ func.func @duplicate_vscale_values() {
   %c2_vscale = arith.muli %vscale_1, %c2 : index
   %add = arith.addi %c2_vscale, %c4_vscale : index
 
-  %bound = "test.reify_scalable_bound"(%add) {type = "EQ", vscale_min = 1, vscale_max = 16} : (index) -> index
+  %bound = "test.reify_bound"(%add) {type = "EQ", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
   "test.some_use"(%bound) : (index) -> ()
   return
 }
@@ -154,7 +154,7 @@ func.func @non_scalable_code() {
   %c0 = arith.constant 0 : index
   scf.for %i = %c0 to %c1024 step %c4 {
     %min_i = affine.min #map_dim_i(%i)[%c4]
-    %bound_i = "test.reify_scalable_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16} : (index) -> index
+    %bound_i = "test.reify_bound"(%min_i) {type = "UB", vscale_min = 1, vscale_max = 16, scalable} : (index) -> index
     "test.some_use"(%bound_i) : (index) -> ()
   }
   return
diff --git a/mlir/test/lib/Dialect/Affine/CMakeLists.txt b/mlir/test/lib/Dialect/Affine/CMakeLists.txt
index 14960a45d39ba..33cefab9fa2ed 100644
--- a/mlir/test/lib/Dialect/Affine/CMakeLists.txt
+++ b/mlir/test/lib/Dialect/Affine/CMakeLists.txt
@@ -30,5 +30,13 @@ add_mlir_library(MLIRAffineTransformsTestPasses
   MLIRSupport
   MLIRMemRefDialect
   MLIRTensorDialect
+  MLIRTestDialect
   MLIRVectorUtils
   )
+
+target_include_directories(MLIRAffineTransformsTestPasses
+  PRIVATE
+  ${CMAKE_CURRENT_SOURCE_DIR}/../Test
+  ${CMAKE_CURRENT_BINARY_DIR}/../Test
+  )
+
diff --git a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
index f38631054fb3c..6730f9b292ad9 100644
--- a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
+++ b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
@@ -6,6 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "TestDialect.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Affine/IR/ValueBoundsOpInterfaceImpl.h"
 #include "mlir/Dialect/Affine/Transforms/Transforms.h"
@@ -57,31 +58,6 @@ struct TestReifyValueBounds
 
 } // namespace
 
-static FailureOr<BoundType> parseBoundType(const std::string &type) {
-  if (type == "EQ")
-    return BoundType::EQ;
-  if (type == "LB")
-    return BoundType::LB;
-  if (type == "UB")
-    return BoundType::UB;
-  return failure();
-}
-
-static FailureOr<ValueBoundsConstraintSet::ComparisonOperator>
-parseComparisonOperator(const std::string &type) {
-  if (type == "EQ")
-    return ValueBoundsConstraintSet::ComparisonOperator::EQ;
-  if (type == "LT")
-    return ValueBoundsConstraintSet::ComparisonOperator::LT;
-  if (type == "LE")
-    return ValueBoundsConstraintSet::ComparisonOperator::LE;
-  if (type == "GT")
-    return ValueBoundsConstraintSet::ComparisonOperator::GT;
-  if (type == "GE")
-    return ValueBoundsConstraintSet::ComparisonOperator::GE;
-  return failure();
-}
-
 static ValueBoundsConstraintSet::ComparisonOperator
 invertComparisonOperator(ValueBoundsConstraintSet::ComparisonOperator cmp) {
   if (cmp == ValueBoundsConstraintSet::ComparisonOperator::LT)
@@ -101,144 +77,89 @@ static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
                                           bool reifyToFuncArgs,
                                           bool useArithOps) {
   IRRewriter rewriter(funcOp.getContext());
-  WalkResult result = funcOp.walk([&](Operation *op) {
-    // Look for test.reify_bound ops.
-    if (op->getName().getStringRef() == "test.reify_bound" ||
-        op->getName().getStringRef() == "test.reify_constant_bound" ||
-        op->getName().getStringRef() == "test.reify_scalable_bound") {
-      if (op->getNumOperands() != 1 || op->getNumResults() != 1 ||
-          !op->getResultTypes()[0].isIndex()) {
-        op->emitOpError("invalid op");
-        return WalkResult::skip();
-      }
-      Value value = op->getOperand(0);
-      if (isa<IndexType>(value.getType()) !=
-          !op->hasAttrOfType<IntegerAttr>("dim")) {
-        // Op should have "dim" attribute if and only if the operand is an
-        // index-typed value.
-        op->emitOpError("invalid op");
-        return WalkResult::skip();
-      }
-
-      // Get bound type.
-      std::string boundTypeStr = "EQ";
-      if (auto boundTypeAttr = op->getAttrOfType<StringAttr>("type"))
-        boundTypeStr = boundTypeAttr.str();
-      auto boundType = parseBoundType(boundTypeStr);
-      if (failed(boundType)) {
-        op->emitOpError("invalid op");
-        return WalkResult::interrupt();
-      }
-
-      // Get shape dimension (if any).
-      auto dim = value.getType().isIndex()
-                     ? std::nullopt
-                     : std::make_optional<int64_t>(
-                           op->getAttrOfType<IntegerAttr>("dim").getInt());
-
-      // Check if a constant was requested.
-      bool constant =
-          op->getName().getStringRef() == "test.reify_constant_bound";
-
-      bool scalable = !constant && op->getName().getStringRef() ==
-                                       "test.reify_scalable_bound";
-
-      // Prepare stop condition. By default, reify in terms of the op's
-      // operands. No stop condition is used when a constant was requested.
-      std::function<bool(Value, std::optional<int64_t>,
-                         ValueBoundsConstraintSet & cstr)>
-          stopCondition = [&](Value v, std::optional<int64_t> d,
-                              ValueBoundsConstraintSet &cstr) {
-            // Reify in terms of SSA values that are different from `value`.
-            return v != value;
-          };
-      if (reifyToFuncArgs) {
-        // Reify in terms of function block arguments.
-        stopCondition = [](Value v, std::optional<int64_t> d,
-                           ValueBoundsConstraintSet &cstr) {
-          auto bbArg = dyn_cast<BlockArgument>(v);
-          if (!bbArg)
-            return false;
-          return isa<FunctionOpInterface>(
-              bbArg.getParentBlock()->getParentOp());
+  WalkResult result = funcOp.walk([&](test::ReifyBoundOp op) {
+    auto boundType = op.getBoundType();
+    Value value = op.getVar();
+    std::optional<int64_t> dim = op.getDim();
+    bool constant = op.getConstant();
+    bool scalable = op.getScalable();
+
+    // Prepare stop condition. By default, reify in terms of the op's
+    // operands. No stop condition is used when a constant was requested.
+    std::function<bool(Value, std::optional<int64_t>,
+                       ValueBoundsConstraintSet & cstr)>
+        stopCondition = [&](Value v, std::optional<int64_t> d,
+                            ValueBoundsConstraintSet &cstr) {
+          // Reify in terms of SSA values that are different from `value`.
+          return v != value;
         };
-      }
+    if (reifyToFuncArgs) {
+      // Reify in terms of function block arguments.
+      stopCondition = [](Value v, std::optional<int64_t> d,
+                         ValueBoundsConstraintSet &cstr) {
+        auto bbArg = dyn_cast<BlockArgument>(v);
+        if (!bbArg)
+          return false;
+        return isa<FunctionOpInterface>(bbArg.getParentBlock()->getParentOp());
+      };
+    }
 
-      // Reify value bound
-      rewriter.setInsertionPointAfter(op);
-      FailureOr<OpFoldResult> reified = failure();
-      if (constant) {
-        auto reifiedConst = ValueBoundsConstraintSet::computeConstantBound(
-            *boundType, value, dim, /*stopCondition=*/nullptr);
-        if (succeeded(reifiedConst))
-          reified =
-              FailureOr<OpFoldResult>(rewriter.getIndexAttr(*reifiedConst));
-      } else if (scalable) {
-        unsigned vscaleMin = 0;
-        unsigned vscaleMax = 0;
-        if (auto attr = "vscale_min"; op->hasAttrOfType<IntegerAttr>(attr)) {
-          vscaleMin = unsigned(op->getAttrOfType<IntegerAttr>(attr).getInt());
-        } else {
-          op->emitOpError("expected `vscale_min` to be provided");
-          return WalkResult::skip();
+    // Reify value bound
+    rewriter.setInsertionPointAfter(op);
+    FailureOr<OpFoldResult> reified = failure();
+    if (constant) {
+      auto reifiedConst = ValueBoundsConstraintSet::computeConstantBound(
+          boundType, value, dim, /*stopCondition=*/nullptr);
+      if (succeeded(reifiedConst))
+        reified = FailureOr<OpFoldResult>(rewriter.getIndexAttr(*reifiedConst));
+    } else if (scalable) {
+      auto loc = op->getLoc();
+      auto reifiedScalable =
+          vector::ScalableValueBoundsConstraintSet::computeScalableBound(
+              value, dim, *op.getVscaleMin(), *op.getVscaleMax(), boundType);
+      if (succeeded(reifiedScalable)) {
+        SmallVector<std::pair<Value, std::optional<int64_t>>, 1> vscaleOperand;
+        if (reifiedScalable->map.getNumInputs() == 1) {
+          // The only possible input to the bound is vscale.
+          vscaleOperand.push_back(std::make_pair(
+              rewriter.create<vector::VectorScaleOp>(loc), std::nullopt));
         }
-        if (auto attr = "vscale_max"; op->hasAttrOfType<IntegerAttr>(attr)) {
-          vscaleMax = unsigned(op->getAttrOfType<IntegerAttr>(attr).getInt());
+        reified = affine::materializeComputedBound(
+            rewriter, loc, reifiedScalable->map, vscaleOperand);
+      }
+    } else {
+      if (dim) {
+        if (useArithOps) {
+          reified = arith::reifyShapedValueDimBound(
+              rewriter, op->getLoc(), boundType, value, *dim, stopCondition);
         } else {
-          op->emitOpError("expected `vscale_max` to be provided");
-          return WalkResult::skip();
-        }
-
-        auto loc = op->getLoc();
-        auto reifiedScalable =
-            vector::ScalableValueBoundsConstraintSet::computeScalableBound(
-                value, dim, vscaleMin, vscaleMax, *boundType);
-        if (succeeded(reifiedScalable)) {
-          SmallVector<std::pair<Value, std::optional<int64_t>>, 1>
-              vscaleOperand;
-          if (reifiedScalable->map.getNumInputs() == 1) {
-            // The only possible input to the bound is vscale.
-            vscaleOperand.push_back(std::make_pair(
-                rewriter.create<vector::VectorScaleOp>(loc), std::nullopt));
-          }
-          reified = affine::materializeComputedBound(
-              rewriter, loc, reifiedScalable->map, vscaleOperand);
+          reified = reifyShapedValueDimBound(rewriter, op->getLoc(), boundType,
+                                             value, *dim, stopCondition);
         }
       } else {
-        if (dim) {
-          if (useArithOps) {
-            reified = arith::reifyShapedValueDimBound(
-                rewriter, op->getLoc(), *boundType, value, *dim, stopCondition);
-          } else {
-            reified = reifyShapedValueDimBound(
-                rewriter, op->getLoc(), *boundType, value, *dim, stopCondition);
-          }
+        if (useArithOps) {
+          reified = arith::reifyIndexValueBound(
+              rewriter, op->getLoc(), boundType, value, stopCondition);
         } else {
-          if (useArithOps) {
-            reified = arith::reifyIndexValueBound(
-                rewriter, op->getLoc(), *boundType, value, stopCondition);
-          } else {
-            reified = reifyIndexValueBound(rewriter, op->getLoc(), *boundType,
-                                           value, stopCondition);
-          }
+          reified = reifyIndexValueBound(rewriter, op->getLoc(), boundType,
+                                         value, stopCondition);
         }
       }
-      if (failed(reified)) {
-        op->emitOpError("could not reify bound");
-        return WalkResult::interrupt();
-      }
+    }
+    if (failed(reified)) {
+      op->emitOpError("could not reify bound");
+      return WalkResult::interrupt();
+    }
 
-      // Replace the op with the reified bound.
-      if (auto val = llvm::dyn_cast_if_present<Value>(*reified)) {
-        rewriter.replaceOp(op, val);
-        return WalkResult::skip();
-      }
-      Value constOp = rewriter.create<arith::ConstantIndexOp>(
-          op->getLoc(), cast<IntegerAttr>(reified->get<Attribute>()).getInt());
-      rewriter.replaceOp(op, constOp);
+    // Replace the op with the reified bound.
+    if (auto val = llvm::dyn_cast_if_present<Value>(*reified)) {
+      rewriter.replaceOp(op, val);
       return WalkResult::skip();
     }
-    return WalkResult::advance();
+    Value constOp = rewriter.create<arith::ConstantIndexOp>(
+        op->getLoc(), cast<IntegerAttr>(reified->get<Attribute>()).getInt());
+    rewriter.replaceOp(op, constOp);
+    return WalkResult::skip();
   });
   return failure(result.wasInterrupted());
 }
@@ -246,60 +167,42 @@ static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
 /// Look for "test.compare" ops and emit errors/remarks.
 static LogicalResult testEquality(func::FuncOp funcOp) {
   IRRewriter rewriter(funcOp.getContext());
-  WalkResult result = funcOp.walk([&](Operation *op) {
-    // Look for test.compare ops.
-    if (op->getName().getStringRef() == "test.compare") {
-      if (op->getNumOperands() != 2 || !op->getOperand(0).getType().isIndex() ||
-          !op->getOperand(1).getType().isIndex()) {
-        op->emitOpError("invalid op");
-        return WalkResult::skip();
-      }
-
-      // Get comparison operator.
-      std::string cmpStr = "EQ";
-      if (auto cmpAttr = op->getAttrOfType<StringAttr>("cmp"))
-        cmpStr = cmpAttr.str();
-      auto cmpType = parseComparisonOperator(cmpStr);
-      if (failed(cmpType)) {
-        op->emitOpError("invalid comparison operator");
+  WalkResult result = funcOp.walk([&](test::CompareOp op) {
+    auto cmpType = op.getComparisonOperator();
+    if (op.getCompose()) {
+      if (cmpType != ValueBoundsConstraintSet::EQ) {
+        op->emitOpError(
+            "comparison operator must be EQ when 'composed' is specified");
         return WalkResult::interrupt();
       }
-
-      if (op->hasAttr("compose")) {
-        if (cmpType != ValueBoundsConstraintSet::EQ) {
-          op->emitOpError(
-              "comparison operator must be EQ when 'composed' is specified");
-          return WalkResult::interrupt();
-        }
-        FailureOr<int64_t> delta = affine::fullyComposeAndComputeConstantDelta(
-            op->getOperand(0), op->getOperand(1));
-        if (failed(delta)) {
-          op->emitError("could not determine equality");
-        } else if (*delta == 0) {
-          op->emitRemark("equal");
-        } else {
-          op->emitRemark("different");
-        }
-        return WalkResult::advance();
-      }
-
-      auto compare = [&](ValueBoundsConstraintSet::ComparisonOperator cmp) {
-        return ValueBoundsConstraintSet::compare(
-            /*lhs=*/op->getOperand(0), /*lhsDim=*/std::nullopt, cmp,
-            /*rhs=*/op->getOperand(1), /*rhsDim=*/std::nullopt);
-      };
-      if (compare(*cmpType)) {
-        op->emitRemark("true");
-      } else if (*cmpType != ValueBoundsConstraintSet::EQ &&
-                 compare(invertComparisonOperator(*cmpType))) {
-        op->emitRemark("false");
-      } else if (*cmpType == ValueBoundsConstraintSet::EQ &&
-                 (compare(ValueBoundsConstraintSet::ComparisonOperator::LT) ||
-                  compare(ValueBoundsConstraintSet::ComparisonOperator::GT))) {
-        op->emitRemark("false");
+      FailureOr<int64_t> delta = affine::fullyComposeAndComputeConstantDelta(
+          op->getOperand(0), op->getOperand(1));
+      if (failed(delta)) {
+        op->emitError("could not determine equality");
+      } else if (*delta == 0) {
+        op->emitRemark("equal");
       } else {
-        op->emitError("unknown");
+        op->emitRemark("different");
       }
+      return WalkResult::advance();
+    }
+
+    auto compare = [&](ValueBoundsConstraintSet::ComparisonOperator cmp) {
+      return ValueBoundsConstraintSet::compare(
+          /*lhs=*/op.getLhs(), /*lhsDim=*/std::nullopt, cmp,
+          /*rhs=*/op.getRhs(), /*rhsDim=*/std::nullopt);
+    };
+    if (compare(cmpType)) {
+      op->emitRemark("true");
+    } else if (cmpType != ValueBoundsConstraintSet::EQ &&
+               compare(invertComparisonOperator(cmpType))) {
+      op->emitRemark("false");
+    } else if (cmpType == ValueBoundsConstraintSet::EQ &&
+               (compare(ValueBoundsConstraintSet::ComparisonOperator::LT) ||
+                compare(ValueBoundsConstraintSet::ComparisonOperator::GT))) {
+      op->emitRemark("false");
+    } else {
+      op->emitError("unknown");
     }
     return WalkResult::advance();
   });
diff --git a/mlir/test/lib/Dialect/Test/CMakeLists.txt b/mlir/test/lib/Dialect/Test/CMakeLists.txt
index 47ddcf6524748..d246c0492a3bd 100644
--- a/mlir/test/lib/Dialect/Test/CMakeLists.txt
+++ b/mlir/test/lib/Dialect/Test/CMakeLists.txt
@@ -85,6 +85,7 @@ add_mlir_library(MLIRTestDialect
   MLIRTensorDialect
   MLIRTransformUtils
   MLIRTransforms
+  MLIRValueBoundsOpInterface
 )
 
 add_mlir_translation_library(MLIRTestFromLLVMIRTranslation
diff --git a/mlir/test/lib/Dialect/Test/TestDialect.cpp b/mlir/test/lib/Dialect/Test/TestDialect.cpp
index 74378633032ce..25c5190ca0ef3 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialect.cpp
@@ -516,6 +516,54 @@ static void printOptionalCustomParser(AsmPrinter &p, Operation *,
   p.printAttribute(result);
 }
 
+//===----------------------------------------------------------------------===//
+// ReifyBoundOp
+//===----------------------------------------------------------------------===//
+
+::mlir::presburger::BoundType ReifyBoundOp::getBoundType() {
+  if (getType() == "EQ")
+    return ::mlir::presburger::BoundType::EQ;
+  if (getType() == "LB")
+    return ::mlir::presburger::BoundType::LB;
+  if (getType() == "UB")
+    return ::mlir::presburger::BoundType::UB;
+  llvm_unreachable("invalid bound type");
+}
+
+LogicalResult ReifyBoundOp::verify() {
+  if (isa<ShapedType>(getVar().getType())) {
+    if (!getDim().has_value())
+      return emitOpError("expected 'dim' attribute for shaped type variable");
+  } else if (getVar().getType().isIndex()) {
+    if (getDim().has_value())
+      return emitOpError("unexpected 'dim' attribute for index variable");
+  } else {
+    return emitOpError("expected index-typed variable or shape type variable");
+  }
+  if (getConstant() && getScalable())
+    return emitOpError("'scalable' and 'constant' are mutually exlusive");
+  if (getScalable() != getVscaleMin().has_value())
+    return emitOpError("expected 'vscale_min' if and only if 'scalable'");
+  if (getScalable() != getVscaleMax().has_value())
+    return emitOpError("expected 'vscale_min' if and only if 'scalable'");
+  return success();
+}
+
+::mlir::ValueBoundsConstraintSet::ComparisonOperator
+CompareOp::getComparisonOperator() {
+  if (getCmp() == "EQ")
+    return ValueBoundsConstraintSet::ComparisonOperator::EQ;
+  if (getCmp() == "LT")
+    return ValueBoundsConstraintSet::ComparisonOperator::LT;
+  if (getCmp() == "LE")
+    return ValueBoundsConstraintSet::ComparisonOperator::LE;
+  if (getCmp() == "GT")
+    return ValueBoundsConstraintSet::ComparisonOperator::GT;
+  if (getCmp() == "GE")
+    return ValueBoundsConstraintSet::ComparisonOperator::GE;
+  llvm_unreachable("invalid comparison operator");
+}
+
 //===----------------------------------------------------------------------===//
 // Test removing op with inner ops.
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/lib/Dialect/Test/TestDialect.h b/mlir/test/lib/Dialect/Test/TestDialect.h
index 4ba28c47ed1c3..d5b2fbeafc410 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.h
+++ b/mlir/test/lib/Dialect/Test/TestDialect.h
@@ -41,6 +41,7 @@
 #include "mlir/Interfaces/InferTypeOpInterface.h"
 #include "mlir/Interfaces/LoopLikeInterface.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
+#include "mlir/Interfaces/ValueBoundsOpInterface.h"
 #include "mlir/Interfaces/ViewLikeInterface.h"
 
 #include <memory>
diff --git a/mlir/test/lib/Dialect/Test/TestOps.td b/mlir/test/lib/Dialect/Test/TestOps.td
index c88f85b8b6b36..ebf158b8bb820 100644
--- a/mlir/test/lib/Dialect/Test/TestOps.td
+++ b/mlir/test/lib/Dialect/Test/TestOps.td
@@ -2184,6 +2184,53 @@ def ForwardBufferOp : TEST_Op<"forward_buffer", [Pure]> {
   let results = (outs AnyRankedOrUnrankedMemRef:$result);
 }
 
+//===----------------------------------------------------------------------===//
+// Test ValueBoundsOpInterface
+//===----------------------------------------------------------------------===//
+
+def ReifyBoundOp : TEST_Op<"reify_bound", [Pure]> {
+  let description = [{
+    Reify a bound for the given index-typed value or dimension size of a shaped
+    value. "LB", "EQ" and "UB" bounds are supported. If `scalable` is set, 
+    `vscale_min` and `vscale_max` must be provided, which allows computing
+    a bound in terms of "vector.vscale" for a given range of vscale.
+  }];
+
+  let arguments = (ins AnyType:$var,
+                       OptionalAttr<I64Attr>:$dim,
+                       DefaultValuedAttr<StrAttr, "\"EQ\"">:$type,
+                       UnitAttr:$constant,
+                       UnitAttr:$scalable,
+                       OptionalAttr<I64Attr>:$vscale_min,
+                       OptionalAttr<I64Attr>:$vscale_max);
+  let results = (outs Index:$result);
+
+  let extraClassDeclaration = [{
+    ::mlir::presburger::BoundType getBoundType();
+  }];
+
+  let hasVerifier = 1;
+}
+
+def CompareOp : TEST_Op<"compare"> {
+  let description = [{
+    Compare `lhs` and `rhs`. A remark is emitted which indicates whether the
+    specified comparison operator was proven to hold. The remark also indicates
+    whether the opposite comparison operator was proven to hold.
+  }];
+
+  let arguments = (ins Index:$lhs,
+                       Index:$rhs,
+                       DefaultValuedAttr<StrAttr, "\"EQ\"">:$cmp,
+                       UnitAttr:$compose);
+  let results = (outs);
+
+  let extraClassDeclaration = [{
+    ::mlir::ValueBoundsConstraintSet::ComparisonOperator
+        getComparisonOperator();
+  }];
+}
+
 //===----------------------------------------------------------------------===//
 // Test RegionBranchOpInterface
 //===----------------------------------------------------------------------===//
diff --git a/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
index 684b59e7f62f6..dc5f4047c286d 100644
--- a/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
@@ -421,6 +421,7 @@ cc_library(
         "//mlir:TranslateLib",
         "//mlir:TransformUtils",
         "//mlir:Transforms",
+        "//mlir:ValueBoundsOpInterface",
         "//mlir:ViewLikeInterface",
     ],
 )
@@ -644,6 +645,7 @@ cc_library(
         "lib/Dialect/Affine/*.cpp",
     ]),
     deps = [
+        ":TestDialect",
         "//llvm:Support",
         "//mlir:AffineAnalysis",
         "//mlir:AffineDialect",

From a34887550b7f2926ea335d4aedf8b72811f9c945 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Mon, 15 Apr 2024 18:46:58 +0200
Subject: [PATCH 445/886] [LLDB][libc++] Adds valarray proxy data formatters.
 (#88613)

These proxies are returned by operator[](...). These proxies all
"behave" the same. They store a pointer to the data of the valarray they
are a proxy for and they have an internal array of indices. This
internal array is considered its contents.
---
 .../Plugins/Language/CPlusPlus/CMakeLists.txt |   1 +
 .../Language/CPlusPlus/CPlusPlusLanguage.cpp  |  11 +
 .../Plugins/Language/CPlusPlus/LibCxx.h       |   4 +
 .../Language/CPlusPlus/LibCxxProxyArray.cpp   | 194 ++++++++++++++++++
 .../TestDataFormatterLibcxxValarray.py        |  88 +++++++-
 .../libcxx/valarray/main.cpp                  |   6 +-
 6 files changed, 295 insertions(+), 9 deletions(-)
 create mode 100644 lldb/source/Plugins/Language/CPlusPlus/LibCxxProxyArray.cpp

diff --git a/lldb/source/Plugins/Language/CPlusPlus/CMakeLists.txt b/lldb/source/Plugins/Language/CPlusPlus/CMakeLists.txt
index 0c6fdb2b95731..f59032c423880 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/CMakeLists.txt
+++ b/lldb/source/Plugins/Language/CPlusPlus/CMakeLists.txt
@@ -14,6 +14,7 @@ add_lldb_library(lldbPluginCPlusPlusLanguage PLUGIN
   LibCxxQueue.cpp
   LibCxxRangesRefView.cpp
   LibCxxSliceArray.cpp
+  LibCxxProxyArray.cpp
   LibCxxSpan.cpp
   LibCxxTuple.cpp
   LibCxxUnorderedMap.cpp
diff --git a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
index afb683f7d846a..5f0684163328f 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
@@ -760,6 +760,12 @@ static void LoadLibCxxFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
       lldb_private::formatters::LibcxxStdSliceArraySyntheticFrontEndCreator,
       "libc++ std::slice_array synthetic children",
       "^std::__[[:alnum:]]+::slice_array<.+>$", stl_deref_flags, true);
+  AddCXXSynthetic(
+      cpp_category_sp,
+      lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEndCreator,
+      "libc++ synthetic children for the valarray proxy arrays",
+      "^std::__[[:alnum:]]+::(gslice|mask|indirect)_array<.+>$",
+      stl_deref_flags, true);
   AddCXXSynthetic(
       cpp_category_sp,
       lldb_private::formatters::LibcxxStdForwardListSyntheticFrontEndCreator,
@@ -890,6 +896,11 @@ static void LoadLibCxxFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
                 "libc++ std::slice_array summary provider",
                 "^std::__[[:alnum:]]+::slice_array<.+>$", stl_summary_flags,
                 true);
+  AddCXXSummary(cpp_category_sp,
+                lldb_private::formatters::LibcxxContainerSummaryProvider,
+                "libc++ summary provider for the valarray proxy arrays",
+                "^std::__[[:alnum:]]+::(gslice|mask|indirect)_array<.+>$",
+                stl_summary_flags, true);
   AddCXXSummary(
       cpp_category_sp, lldb_private::formatters::LibcxxContainerSummaryProvider,
       "libc++ std::list summary provider",
diff --git a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
index 8e97174dc3075..7fe15d1bf3f7a 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
+++ b/lldb/source/Plugins/Language/CPlusPlus/LibCxx.h
@@ -231,6 +231,10 @@ SyntheticChildrenFrontEnd *
 LibcxxStdSliceArraySyntheticFrontEndCreator(CXXSyntheticChildren *,
                                             lldb::ValueObjectSP);
 
+SyntheticChildrenFrontEnd *
+LibcxxStdProxyArraySyntheticFrontEndCreator(CXXSyntheticChildren *,
+                                            lldb::ValueObjectSP);
+
 SyntheticChildrenFrontEnd *
 LibcxxStdListSyntheticFrontEndCreator(CXXSyntheticChildren *,
                                       lldb::ValueObjectSP);
diff --git a/lldb/source/Plugins/Language/CPlusPlus/LibCxxProxyArray.cpp b/lldb/source/Plugins/Language/CPlusPlus/LibCxxProxyArray.cpp
new file mode 100644
index 0000000000000..726f06523b97b
--- /dev/null
+++ b/lldb/source/Plugins/Language/CPlusPlus/LibCxxProxyArray.cpp
@@ -0,0 +1,194 @@
+//===-- LibCxxProxyArray.cpp-----------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "LibCxx.h"
+
+#include "lldb/Core/ValueObject.h"
+#include "lldb/DataFormatters/FormattersHelpers.h"
+#include <optional>
+
+using namespace lldb;
+using namespace lldb_private;
+using namespace lldb_private::formatters;
+
+namespace lldb_private {
+namespace formatters {
+
+/// Data formatter for libc++'s std::"proxy_array".
+///
+/// A proxy_array's are created by using:
+///   std::gslice_array   operator[](const std::gslice& gslicearr);
+///   std::mask_array     operator[](const std::valarray<bool>& boolarr);
+///   std::indirect_array operator[](const std::valarray<std::size_t>& indarr);
+///
+/// These arrays have the following members:
+/// - __vp_ points to std::valarray::__begin_
+/// - __1d_ an array of offsets of the elements from @a __vp_
+class LibcxxStdProxyArraySyntheticFrontEnd : public SyntheticChildrenFrontEnd {
+public:
+  LibcxxStdProxyArraySyntheticFrontEnd(lldb::ValueObjectSP valobj_sp);
+
+  ~LibcxxStdProxyArraySyntheticFrontEnd() override;
+
+  llvm::Expected<uint32_t> CalculateNumChildren() override;
+
+  lldb::ValueObjectSP GetChildAtIndex(uint32_t idx) override;
+
+  lldb::ChildCacheState Update() override;
+
+  bool MightHaveChildren() override;
+
+  size_t GetIndexOfChildWithName(ConstString name) override;
+
+private:
+  /// A non-owning pointer to the array's __vp_.
+  ValueObject *m_base = nullptr;
+  /// The type of the array's template argument T.
+  CompilerType m_element_type;
+  /// The sizeof the array's template argument T.
+  uint32_t m_element_size = 0;
+
+  /// A non-owning pointer to the array's __1d_.__begin_.
+  ValueObject *m_start = nullptr;
+  /// A non-owning pointer to the array's __1d_.__end_.
+  ValueObject *m_finish = nullptr;
+  /// The type of the __1d_ array's template argument T (size_t).
+  CompilerType m_element_type_size_t;
+  /// The sizeof the __1d_ array's template argument T (size_t)
+  uint32_t m_element_size_size_t = 0;
+};
+
+} // namespace formatters
+} // namespace lldb_private
+
+lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::
+    LibcxxStdProxyArraySyntheticFrontEnd(lldb::ValueObjectSP valobj_sp)
+    : SyntheticChildrenFrontEnd(*valobj_sp), m_element_type() {
+  if (valobj_sp)
+    Update();
+}
+
+lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::
+    ~LibcxxStdProxyArraySyntheticFrontEnd() {
+  // these need to stay around because they are child objects who will follow
+  // their parent's life cycle
+  // delete m_base;
+}
+
+llvm::Expected<uint32_t> lldb_private::formatters::
+    LibcxxStdProxyArraySyntheticFrontEnd::CalculateNumChildren() {
+
+  if (!m_start || !m_finish)
+    return 0;
+  uint64_t start_val = m_start->GetValueAsUnsigned(0);
+  uint64_t finish_val = m_finish->GetValueAsUnsigned(0);
+
+  if (start_val == 0 || finish_val == 0)
+    return 0;
+
+  if (start_val >= finish_val)
+    return 0;
+
+  size_t num_children = (finish_val - start_val);
+  if (num_children % m_element_size_size_t)
+    return 0;
+  return num_children / m_element_size_size_t;
+}
+
+lldb::ValueObjectSP
+lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::GetChildAtIndex(
+    uint32_t idx) {
+  if (!m_base)
+    return lldb::ValueObjectSP();
+
+  uint64_t offset = idx * m_element_size_size_t;
+  offset = offset + m_start->GetValueAsUnsigned(0);
+
+  lldb::ValueObjectSP indirect = CreateValueObjectFromAddress(
+      "", offset, m_backend.GetExecutionContextRef(), m_element_type_size_t);
+  if (!indirect)
+    return lldb::ValueObjectSP();
+
+  const size_t value = indirect->GetValueAsUnsigned(0);
+  if (!value)
+    return lldb::ValueObjectSP();
+
+  offset = value * m_element_size;
+  offset = offset + m_base->GetValueAsUnsigned(0);
+
+  StreamString name;
+  name.Printf("[%" PRIu64 "] -> [%zu]", (uint64_t)idx, value);
+  return CreateValueObjectFromAddress(name.GetString(), offset,
+                                      m_backend.GetExecutionContextRef(),
+                                      m_element_type);
+}
+
+lldb::ChildCacheState
+lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::Update() {
+  m_base = nullptr;
+  m_start = nullptr;
+  m_finish = nullptr;
+
+  CompilerType type = m_backend.GetCompilerType();
+  if (type.GetNumTemplateArguments() == 0)
+    return ChildCacheState::eRefetch;
+
+  m_element_type = type.GetTypeTemplateArgument(0);
+  if (std::optional<uint64_t> size = m_element_type.GetByteSize(nullptr))
+    m_element_size = *size;
+
+  if (m_element_size == 0)
+    return ChildCacheState::eRefetch;
+
+  ValueObjectSP vector = m_backend.GetChildMemberWithName("__1d_");
+  if (!vector)
+    return ChildCacheState::eRefetch;
+
+  type = vector->GetCompilerType();
+  if (type.GetNumTemplateArguments() == 0)
+    return ChildCacheState::eRefetch;
+
+  m_element_type_size_t = type.GetTypeTemplateArgument(0);
+  if (std::optional<uint64_t> size = m_element_type_size_t.GetByteSize(nullptr))
+    m_element_size_size_t = *size;
+
+  if (m_element_size_size_t == 0)
+    return ChildCacheState::eRefetch;
+
+  ValueObjectSP base = m_backend.GetChildMemberWithName("__vp_");
+  ValueObjectSP start = vector->GetChildMemberWithName("__begin_");
+  ValueObjectSP finish = vector->GetChildMemberWithName("__end_");
+  if (!base || !start || !finish)
+    return ChildCacheState::eRefetch;
+
+  m_base = base.get();
+  m_start = start.get();
+  m_finish = finish.get();
+
+  return ChildCacheState::eRefetch;
+}
+
+bool lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::
+    MightHaveChildren() {
+  return true;
+}
+
+size_t lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEnd::
+    GetIndexOfChildWithName(ConstString name) {
+  if (!m_base)
+    return std::numeric_limits<size_t>::max();
+  return ExtractIndexFromString(name.GetCString());
+}
+
+lldb_private::SyntheticChildrenFrontEnd *
+lldb_private::formatters::LibcxxStdProxyArraySyntheticFrontEndCreator(
+    CXXSyntheticChildren *, lldb::ValueObjectSP valobj_sp) {
+  if (!valobj_sp)
+    return nullptr;
+  return new LibcxxStdProxyArraySyntheticFrontEnd(valobj_sp);
+}
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/TestDataFormatterLibcxxValarray.py b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/TestDataFormatterLibcxxValarray.py
index b59b770ed6790..613546b50a770 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/TestDataFormatterLibcxxValarray.py
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/TestDataFormatterLibcxxValarray.py
@@ -89,21 +89,93 @@ def test_with_run_command(self):
             "frame variable sa",
             substrs=[
                 "sa = stride=2 size=4",
-                "[0] = 1",
-                "[1] = 3",
-                "[2] = 5",
-                "[3] = 7",
+                "[0] = 11",
+                "[1] = 13",
+                "[2] = 15",
+                "[3] = 17",
                 "}",
             ],
         )
 
         # check access-by-index
-        self.expect("frame variable sa[0]", substrs=["1"])
-        self.expect("frame variable sa[1]", substrs=["3"])
-        self.expect("frame variable sa[2]", substrs=["5"])
-        self.expect("frame variable sa[3]", substrs=["7"])
+        self.expect("frame variable sa[0]", substrs=["11"])
+        self.expect("frame variable sa[1]", substrs=["13"])
+        self.expect("frame variable sa[2]", substrs=["15"])
+        self.expect("frame variable sa[3]", substrs=["17"])
         self.expect(
             "frame variable sa[4]",
             error=True,
             substrs=['array index 4 is not valid for "(slice_array<int>) sa"'],
         )
+
+        #
+        # std::gslice_array
+        #
+
+        self.expect(
+            "frame variable ga",
+            substrs=[
+                "ga = size=3",
+                "[0] -> [3] = 13",
+                "[1] -> [4] = 14",
+                "[2] -> [5] = 15",
+                "}",
+            ],
+        )
+
+        # check access-by-index
+        self.expect("frame variable ga[0]", substrs=["13"])
+        self.expect("frame variable ga[1]", substrs=["14"])
+        self.expect("frame variable ga[2]", substrs=["15"])
+        self.expect(
+            "frame variable ga[3]",
+            error=True,
+            substrs=['array index 3 is not valid for "(gslice_array<int>) ga"'],
+        )
+        #
+        # std::mask_array
+        #
+
+        self.expect(
+            "frame variable ma",
+            substrs=[
+                "ma = size=2",
+                "[0] -> [1] = 11",
+                "[1] -> [2] = 12",
+                "}",
+            ],
+        )
+
+        # check access-by-index
+        self.expect("frame variable ma[0]", substrs=["11"])
+        self.expect("frame variable ma[1]", substrs=["12"])
+        self.expect(
+            "frame variable ma[2]",
+            error=True,
+            substrs=['array index 2 is not valid for "(mask_array<int>) ma"'],
+        )
+
+        #
+        # std::indirect_array
+        #
+
+        self.expect(
+            "frame variable ia",
+            substrs=[
+                "ia = size=3",
+                "[0] -> [3] = 13",
+                "[1] -> [6] = 16",
+                "[2] -> [9] = 19",
+                "}",
+            ],
+        )
+
+        # check access-by-index
+        self.expect("frame variable ia[0]", substrs=["13"])
+        self.expect("frame variable ia[1]", substrs=["16"])
+        self.expect("frame variable ia[2]", substrs=["19"])
+        self.expect(
+            "frame variable ia[3]",
+            error=True,
+            substrs=['array index 3 is not valid for "(indirect_array<int>) ia"'],
+        )
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/main.cpp b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/main.cpp
index 1481d8b403292..d31951c755eac 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/main.cpp
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/libcxx/valarray/main.cpp
@@ -13,8 +13,12 @@ int main() {
 
   std::valarray<double> va_double({1.0, 0.5, 0.25, 0.125});
 
-  std::valarray<int> va({0, 1, 2, 3, 4, 5, 6, 7, 8, 9});
+  std::valarray<int> va({10, 11, 12, 13, 14, 15, 16, 17, 18, 19});
   std::slice_array<int> sa = va[std::slice(1, 4, 2)];
+  std::gslice_array<int> ga = va[std::gslice(
+      3, std::valarray<std::size_t>(3, 1), std::valarray<std::size_t>(1, 1))];
+  std::mask_array<int> ma = va[std::valarray<bool>{false, true, true}];
+  std::indirect_array<int> ia = va[std::valarray<size_t>{3, 6, 9}];
 
   std::cout << "break here\n";
 }

From 9dbb6e1978d3d2d61ef65c2dac1fd8add5a4c7a2 Mon Sep 17 00:00:00 2001
From: Hsiangkai Wang <hsiangkai.wang@arm.com>
Date: Mon, 15 Apr 2024 17:59:17 +0100
Subject: [PATCH 446/886] [mlir][spirv] Add target width to SPIR-V ABI (#88555)

There are execution modes need target width as their extra operands.
SignedZeroInfNanPreserve is one of them. This patch adds `target width`
as one of SPIR-V ABI attributes.
---
 .../mlir/Dialect/SPIRV/IR/SPIRVAttributes.td  |  4 +++-
 .../mlir/Dialect/SPIRV/IR/TargetAndABI.h      |  7 ++++++-
 mlir/lib/Dialect/SPIRV/IR/TargetAndABI.cpp    | 11 +++++-----
 .../Transforms/LowerABIAttributesPass.cpp     | 20 ++++++++++++++++---
 .../Conversion/GPUToSPIRV/entry-point.mlir    | 12 +++++++++++
 .../lib/Dialect/SPIRV/TestEntryPointAbi.cpp   | 17 +++++++++++++++-
 6 files changed, 59 insertions(+), 12 deletions(-)

diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVAttributes.td b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVAttributes.td
index 74d36445e3113..3a11284da0512 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVAttributes.td
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/SPIRVAttributes.td
@@ -32,10 +32,12 @@ class SPIRV_Attr<string attrName, string attrMnemonic>
 // points in the generated SPIR-V module:
 // 1) [optional] Requested workgroup size.
 // 2) [optional] Requested subgroup size.
+// 3) [optional] Requested target width.
 def SPIRV_EntryPointABIAttr : SPIRV_Attr<"EntryPointABI", "entry_point_abi"> {
   let parameters = (ins
     OptionalParameter<"DenseI32ArrayAttr">:$workgroup_size,
-    OptionalParameter<"std::optional<int>">:$subgroup_size
+    OptionalParameter<"std::optional<int>">:$subgroup_size,
+    OptionalParameter<"std::optional<int>">:$target_width
   );
   let assemblyFormat = "`<` struct(params) `>`";
 }
diff --git a/mlir/include/mlir/Dialect/SPIRV/IR/TargetAndABI.h b/mlir/include/mlir/Dialect/SPIRV/IR/TargetAndABI.h
index c35a8c26c2bc9..24574bfaf6199 100644
--- a/mlir/include/mlir/Dialect/SPIRV/IR/TargetAndABI.h
+++ b/mlir/include/mlir/Dialect/SPIRV/IR/TargetAndABI.h
@@ -87,9 +87,14 @@ bool needsInterfaceVarABIAttrs(TargetEnvAttr targetAttr);
 StringRef getEntryPointABIAttrName();
 
 /// Gets the EntryPointABIAttr given its fields.
+/// targetWidth is used by several execution modes. It is the element width
+/// of floating-point operations.
+/// Refer to Execution Mode in SPIR-V specification.
+/// https://registry.khronos.org/SPIR-V/specs/unified1/SPIRV.html#_execution_mode
 EntryPointABIAttr getEntryPointABIAttr(MLIRContext *context,
                                        ArrayRef<int32_t> workgroupSize = {},
-                                       std::optional<int> subgroupSize = {});
+                                       std::optional<int> subgroupSize = {},
+                                       std::optional<int> targetWidth = {});
 
 /// Queries the entry point ABI on the nearest function-like op containing the
 /// given `op`. Returns null attribute if not found.
diff --git a/mlir/lib/Dialect/SPIRV/IR/TargetAndABI.cpp b/mlir/lib/Dialect/SPIRV/IR/TargetAndABI.cpp
index 5b7c0a59ba420..bbc318e17300a 100644
--- a/mlir/lib/Dialect/SPIRV/IR/TargetAndABI.cpp
+++ b/mlir/lib/Dialect/SPIRV/IR/TargetAndABI.cpp
@@ -120,17 +120,16 @@ bool spirv::needsInterfaceVarABIAttrs(spirv::TargetEnvAttr targetAttr) {
 
 StringRef spirv::getEntryPointABIAttrName() { return "spirv.entry_point_abi"; }
 
-spirv::EntryPointABIAttr
-spirv::getEntryPointABIAttr(MLIRContext *context,
-                            ArrayRef<int32_t> workgroupSize,
-                            std::optional<int> subgroupSize) {
+spirv::EntryPointABIAttr spirv::getEntryPointABIAttr(
+    MLIRContext *context, ArrayRef<int32_t> workgroupSize,
+    std::optional<int> subgroupSize, std::optional<int> targetWidth) {
   DenseI32ArrayAttr workgroupSizeAttr;
   if (!workgroupSize.empty()) {
     assert(workgroupSize.size() == 3);
     workgroupSizeAttr = DenseI32ArrayAttr::get(context, workgroupSize);
   }
-  return spirv::EntryPointABIAttr::get(context, workgroupSizeAttr,
-                                       subgroupSize);
+  return spirv::EntryPointABIAttr::get(context, workgroupSizeAttr, subgroupSize,
+                                       targetWidth);
 }
 
 spirv::EntryPointABIAttr spirv::lookupEntryPointABI(Operation *op) {
diff --git a/mlir/lib/Dialect/SPIRV/Transforms/LowerABIAttributesPass.cpp b/mlir/lib/Dialect/SPIRV/Transforms/LowerABIAttributesPass.cpp
index 6150b5ee17851..2024a2e5279ff 100644
--- a/mlir/lib/Dialect/SPIRV/Transforms/LowerABIAttributesPass.cpp
+++ b/mlir/lib/Dialect/SPIRV/Transforms/LowerABIAttributesPass.cpp
@@ -157,7 +157,7 @@ static LogicalResult lowerEntryPointABIAttr(spirv::FuncOp funcOp,
       // Erase workgroup size.
       entryPointAttr = spirv::EntryPointABIAttr::get(
           entryPointAttr.getContext(), DenseI32ArrayAttr(),
-          entryPointAttr.getSubgroupSize());
+          entryPointAttr.getSubgroupSize(), entryPointAttr.getTargetWidth());
     }
   }
   if (std::optional<int> subgroupSize = entryPointAttr.getSubgroupSize()) {
@@ -170,10 +170,24 @@ static LogicalResult lowerEntryPointABIAttr(spirv::FuncOp funcOp,
       // Erase subgroup size.
       entryPointAttr = spirv::EntryPointABIAttr::get(
           entryPointAttr.getContext(), entryPointAttr.getWorkgroupSize(),
-          std::nullopt);
+          std::nullopt, entryPointAttr.getTargetWidth());
     }
   }
-  if (entryPointAttr.getWorkgroupSize() || entryPointAttr.getSubgroupSize())
+  if (std::optional<int> targetWidth = entryPointAttr.getTargetWidth()) {
+    std::optional<ArrayRef<spirv::Capability>> caps =
+        spirv::getCapabilities(spirv::ExecutionMode::SignedZeroInfNanPreserve);
+    if (!caps || targetEnv.allows(*caps)) {
+      builder.create<spirv::ExecutionModeOp>(
+          funcOp.getLoc(), funcOp,
+          spirv::ExecutionMode::SignedZeroInfNanPreserve, *targetWidth);
+      // Erase target width.
+      entryPointAttr = spirv::EntryPointABIAttr::get(
+          entryPointAttr.getContext(), entryPointAttr.getWorkgroupSize(),
+          entryPointAttr.getSubgroupSize(), std::nullopt);
+    }
+  }
+  if (entryPointAttr.getWorkgroupSize() || entryPointAttr.getSubgroupSize() ||
+      entryPointAttr.getTargetWidth())
     funcOp->setAttr(entryPointAttrName, entryPointAttr);
   else
     funcOp->removeAttr(entryPointAttrName);
diff --git a/mlir/test/Conversion/GPUToSPIRV/entry-point.mlir b/mlir/test/Conversion/GPUToSPIRV/entry-point.mlir
index 99369d11a4ba3..8b14d6569f486 100644
--- a/mlir/test/Conversion/GPUToSPIRV/entry-point.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/entry-point.mlir
@@ -1,5 +1,8 @@
 // RUN: mlir-opt -test-spirv-entry-point-abi %s | FileCheck %s -check-prefix=DEFAULT
 // RUN: mlir-opt -test-spirv-entry-point-abi="workgroup-size=32" %s | FileCheck %s -check-prefix=WG32
+// RUN: mlir-opt -test-spirv-entry-point-abi="subgroup-size=4" %s | FileCheck %s -check-prefix=SG4
+// RUN: mlir-opt -test-spirv-entry-point-abi="target-width=32" %s | FileCheck %s -check-prefix=TW32
+// RUN: mlir-opt -test-spirv-entry-point-abi="workgroup-size=32,8 subgroup-size=4 target-width=32" %s | FileCheck %s -check-prefix=WG32_8-SG4-TW32
 
 //      DEFAULT: gpu.func @foo()
 // DEFAULT-SAME: spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [1, 1, 1]>
@@ -7,6 +10,15 @@
 //      WG32: gpu.func @foo()
 // WG32-SAME:  spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [32, 1, 1]>
 
+//      SG4: gpu.func @foo()
+// SG4-SAME:  spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [1, 1, 1], subgroup_size = 4>
+
+//      TW32: gpu.func @foo()
+// TW32-SAME:  spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [1, 1, 1], target_width = 32>
+
+//      WG32_8-SG4-TW32: gpu.func @foo()
+// WG32_8-SG4-TW32-SAME:  spirv.entry_point_abi = #spirv.entry_point_abi<workgroup_size = [32, 8, 1], subgroup_size = 4, target_width = 32>
+
 gpu.module @kernels {
   gpu.func @foo() kernel {
     gpu.return
diff --git a/mlir/test/lib/Dialect/SPIRV/TestEntryPointAbi.cpp b/mlir/test/lib/Dialect/SPIRV/TestEntryPointAbi.cpp
index 47b06e4531152..0e8dfb8e3ee9a 100644
--- a/mlir/test/lib/Dialect/SPIRV/TestEntryPointAbi.cpp
+++ b/mlir/test/lib/Dialect/SPIRV/TestEntryPointAbi.cpp
@@ -45,6 +45,16 @@ struct TestSpirvEntryPointABIPass
           "Workgroup size to use for all gpu.func kernels in the module, "
           "specified with x-dimension first, y-dimension next and z-dimension "
           "last. Unspecified dimensions will be set to 1")};
+  Pass::Option<int> subgroupSize{
+      *this, "subgroup-size",
+      llvm::cl::desc(
+          "Subgroup size to use for all gpu.func kernels in the module"),
+      llvm::cl::init(0)};
+  Pass::Option<int> targetWidth{
+      *this, "target-width",
+      llvm::cl::desc(
+          "Specify the component width of floating-point instructions"),
+      llvm::cl::init(0)};
 };
 } // namespace
 
@@ -60,7 +70,12 @@ void TestSpirvEntryPointABIPass::runOnOperation() {
                                              workgroupSize.end());
     workgroupSizeVec.resize(3, 1);
     gpuFunc->setAttr(attrName,
-                     spirv::getEntryPointABIAttr(context, workgroupSizeVec));
+                     spirv::getEntryPointABIAttr(
+                         context, workgroupSizeVec,
+                         (subgroupSize == 0) ? std::nullopt
+                                             : std::optional<int>(subgroupSize),
+                         (targetWidth == 0) ? std::nullopt
+                                            : std::optional<int>(targetWidth)));
   }
 }
 

From e09245b3b16fe78e30e79e92f8ce16d1edf7787b Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Mon, 15 Apr 2024 10:00:56 -0700
Subject: [PATCH 447/886] [memprof] Fix typos in serializedSizeV0 and
 serializedSizeV2 (#88629)

The first field to serialize is the size of
IndexedMemProfRecord::AllocSites.  It has nothing to do with
GlobalValue::GUID.  This happens to work because of:

  using GUID = uint64_t;
---
 llvm/lib/ProfileData/MemProf.cpp | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/ProfileData/MemProf.cpp b/llvm/lib/ProfileData/MemProf.cpp
index 96aeedf2e6913..97414505f1c13 100644
--- a/llvm/lib/ProfileData/MemProf.cpp
+++ b/llvm/lib/ProfileData/MemProf.cpp
@@ -42,7 +42,8 @@ size_t IndexedAllocationInfo::serializedSize(IndexedVersion Version) const {
 }
 
 static size_t serializedSizeV0(const IndexedMemProfRecord &Record) {
-  size_t Result = sizeof(GlobalValue::GUID);
+  // The number of alloc sites to serialize.
+  size_t Result = sizeof(uint64_t);
   for (const IndexedAllocationInfo &N : Record.AllocSites)
     Result += N.serializedSize(Version0);
 
@@ -57,7 +58,8 @@ static size_t serializedSizeV0(const IndexedMemProfRecord &Record) {
 }
 
 static size_t serializedSizeV2(const IndexedMemProfRecord &Record) {
-  size_t Result = sizeof(GlobalValue::GUID);
+  // The number of alloc sites to serialize.
+  size_t Result = sizeof(uint64_t);
   for (const IndexedAllocationInfo &N : Record.AllocSites)
     Result += N.serializedSize(Version2);
 

From 51d85b7b107e9508b90cec274f3f122444fe7ee6 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Mon, 15 Apr 2024 10:02:23 -0700
Subject: [PATCH 448/886] [Support] Add llvm::support::endian::writeNext
 (#88685)

`writeNext` overloads increase the pointer argument like `readNext`.
Code like the following
```
endian::write32<ELFT::Endianness>(p, 42);
p += 4;
endian::write32<ELFT::Endianness>(p, 43);
p += 4;
```

can be simplified to:
```
endian::writeNext<uint32_t, ELFT::Endianness>(p, 42);
endian::writeNext<uint32_t, ELFT::Endianness>(p, 43);
```
---
 llvm/include/llvm/Support/Endian.h    | 15 +++++++++++++++
 llvm/unittests/Support/EndianTest.cpp | 23 +++++++++++++++++++++++
 2 files changed, 38 insertions(+)

diff --git a/llvm/include/llvm/Support/Endian.h b/llvm/include/llvm/Support/Endian.h
index 4c0405cf1e2f6..1cdb5ca0d5eaa 100644
--- a/llvm/include/llvm/Support/Endian.h
+++ b/llvm/include/llvm/Support/Endian.h
@@ -102,6 +102,21 @@ inline void write(void *memory, value_type value) {
   write<value_type, alignment>(memory, value, endian);
 }
 
+/// Write a value of a particular endianness, and increment the buffer past that
+/// value.
+template <typename value_type, std::size_t alignment = unaligned,
+          typename CharT>
+inline void writeNext(CharT *&memory, value_type value, endianness endian) {
+  write(memory, value, endian);
+  memory += sizeof(value_type);
+}
+
+template <typename value_type, endianness endian,
+          std::size_t alignment = unaligned, typename CharT>
+inline void writeNext(CharT *&memory, value_type value) {
+  writeNext<value_type, alignment, CharT>(memory, value, endian);
+}
+
 template <typename value_type>
 using make_unsigned_t = std::make_unsigned_t<value_type>;
 
diff --git a/llvm/unittests/Support/EndianTest.cpp b/llvm/unittests/Support/EndianTest.cpp
index ab7dfc3800691..bba1a56168f70 100644
--- a/llvm/unittests/Support/EndianTest.cpp
+++ b/llvm/unittests/Support/EndianTest.cpp
@@ -36,6 +36,29 @@ TEST(Endian, Read) {
                                                                   1)));
 }
 
+TEST(Endian, WriteNext) {
+  unsigned char bigval[] = {0x00, 0x00}, *p = bigval;
+  endian::writeNext<int16_t, llvm::endianness::big>(p, short(0xaabb));
+  EXPECT_EQ(bigval[0], 0xaa);
+  EXPECT_EQ(bigval[1], 0xbb);
+  EXPECT_EQ(p, bigval + 2);
+
+  char littleval[8] = {}, *q = littleval;
+  endian::writeNext<uint32_t, llvm::endianness::little>(q, 0x44556677);
+  EXPECT_EQ(littleval[0], 0x77);
+  EXPECT_EQ(littleval[1], 0x66);
+  EXPECT_EQ(littleval[2], 0x55);
+  EXPECT_EQ(littleval[3], 0x44);
+  EXPECT_EQ(q, littleval + 4);
+
+  endian::writeNext<uint32_t>(q, 0x11223344, llvm::endianness::little);
+  EXPECT_EQ(littleval[4], 0x44);
+  EXPECT_EQ(littleval[5], 0x33);
+  EXPECT_EQ(littleval[6], 0x22);
+  EXPECT_EQ(littleval[7], 0x11);
+  EXPECT_EQ(q, littleval + 8);
+}
+
 TEST(Endian, ReadBitAligned) {
   // Simple test to make sure we properly pull out the 0x0 word.
   unsigned char littleval[] = {0x3f, 0x00, 0x00, 0x00, 0xc0, 0xff, 0xff, 0xff};

From 603fa4c6b9a540206ec6f6b2115274878b65bfbb Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Mon, 15 Apr 2024 13:11:29 -0400
Subject: [PATCH 449/886] [BOLT][NFC] Be more obvious about selecting X86
 (#88527)

Use `isX86()` rather than `!isAArch64() && !isRISCV()`, and similar.
---
 bolt/lib/Core/BinaryContext.cpp      | 2 +-
 bolt/lib/Core/BinaryEmitter.cpp      | 2 +-
 bolt/lib/Rewrite/RewriteInstance.cpp | 6 ++++--
 3 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/bolt/lib/Core/BinaryContext.cpp b/bolt/lib/Core/BinaryContext.cpp
index 7c2d8c52287be..ad2eb18caf109 100644
--- a/bolt/lib/Core/BinaryContext.cpp
+++ b/bolt/lib/Core/BinaryContext.cpp
@@ -1880,7 +1880,7 @@ MarkerSymType BinaryContext::getMarkerType(const SymbolRef &Symbol) const {
   // For aarch64 and riscv, the ABI defines mapping symbols so we identify data
   // in the code section (see IHI0056B). $x identifies a symbol starting code or
   // the end of a data chunk inside code, $d identifies start of data.
-  if ((!isAArch64() && !isRISCV()) || ELFSymbolRef(Symbol).getSize())
+  if (isX86() || ELFSymbolRef(Symbol).getSize())
     return MarkerSymType::NONE;
 
   Expected<StringRef> NameOrError = Symbol.getName();
diff --git a/bolt/lib/Core/BinaryEmitter.cpp b/bolt/lib/Core/BinaryEmitter.cpp
index 97d19b75200f5..6f86ddc774544 100644
--- a/bolt/lib/Core/BinaryEmitter.cpp
+++ b/bolt/lib/Core/BinaryEmitter.cpp
@@ -512,7 +512,7 @@ void BinaryEmitter::emitFunctionBody(BinaryFunction &BF, FunctionFragment &FF,
 
       // Emit sized NOPs via MCAsmBackend::writeNopData() interface on x86.
       // This is a workaround for invalid NOPs handling by asm/disasm layer.
-      if (BC.MIB->isNoop(Instr) && BC.isX86()) {
+      if (BC.isX86() && BC.MIB->isNoop(Instr)) {
         if (std::optional<uint32_t> Size = BC.MIB->getSize(Instr)) {
           SmallString<15> Code;
           raw_svector_ostream VecOS(Code);
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index f22bede002da5..bb76abd2f843b 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -1670,7 +1670,9 @@ void RewriteInstance::disassemblePLT() {
       return disassemblePLTSectionAArch64(Section);
     if (BC->isRISCV())
       return disassemblePLTSectionRISCV(Section);
-    return disassemblePLTSectionX86(Section, EntrySize);
+    if (BC->isX86())
+      return disassemblePLTSectionX86(Section, EntrySize);
+    llvm_unreachable("Unmplemented PLT");
   };
 
   for (BinarySection &Section : BC->allocatableSections()) {
@@ -2605,7 +2607,7 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
   const bool IsToCode = ReferencedSection && ReferencedSection->isText();
 
   // Special handling of PC-relative relocations.
-  if (!IsAArch64 && !BC->isRISCV() && Relocation::isPCRelative(RType)) {
+  if (BC->isX86() && Relocation::isPCRelative(RType)) {
     if (!IsFromCode && IsToCode) {
       // PC-relative relocations from data to code are tricky since the
       // original information is typically lost after linking, even with

From bd7b170e97e314cb786dd33a750ef79412fc9682 Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Mon, 15 Apr 2024 13:12:53 -0400
Subject: [PATCH 450/886] [BOLT][NFC] Remove extraneous braces (#88620)

A small cleanup -- no braces needed here.
---
 bolt/lib/Rewrite/RewriteInstance.cpp | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index bb76abd2f843b..fd2477231142e 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -2861,15 +2861,14 @@ void RewriteInstance::handleRelocation(const SectionRef &RelocatedSection,
       BC->isRISCV())
     ForceRelocation = true;
 
-  if (IsFromCode) {
+  if (IsFromCode)
     ContainingBF->addRelocation(Rel.getOffset(), ReferencedSymbol, RType,
                                 Addend, ExtractedValue);
-  } else if (IsToCode || ForceRelocation) {
+  else if (IsToCode || ForceRelocation)
     BC->addRelocation(Rel.getOffset(), ReferencedSymbol, RType, Addend,
                       ExtractedValue);
-  } else {
+  else
     LLVM_DEBUG(dbgs() << "BOLT-DEBUG: ignoring relocation from data to data\n");
-  }
 }
 
 void RewriteInstance::selectFunctionsToProcess() {

From 9c970d5ecd6a85188cd2b0a941fcd4d60063ef81 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Mon, 15 Apr 2024 20:14:07 +0300
Subject: [PATCH 451/886] [LLD] [COFF] Don't add pseudo relocs for dangling
 references (#88487)

When doing GC, we normally won't have dangling references, because such
a reference would keep the other section alive, keeping it from being
eliminated.

However, references within DWARF sections are ignored for the purposes
of GC (because otherwise, they would essentially keep everything alive,
defeating the point of the GC), see
c579a5b1d92a9bc2046d00ee2d427832e0f5ddec for more context.

Therefore, dangling relocations against discarded symbols are ignored
within DWARF sections (see maybeReportRelocationToDiscarded in
Chunks.cpp). Consequently, we also shouldn't create any pseudo
relocations for these cases, as we run into a null pointer dereference
when trying to generate the pseudo relocation info for it.

This fixes the downstream bug
https://github.com/mstorsjo/llvm-mingw/issues/418, fixing crashes on
combinations with -ffunction-sections, -fdata-sections,
-Wl,--gc-sections and debug info.
---
 lld/COFF/Chunks.cpp           |  7 ++++++
 lld/test/COFF/autoimport-gc.s | 41 +++++++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+)
 create mode 100644 lld/test/COFF/autoimport-gc.s

diff --git a/lld/COFF/Chunks.cpp b/lld/COFF/Chunks.cpp
index 004d71097387a..0cae52785a000 100644
--- a/lld/COFF/Chunks.cpp
+++ b/lld/COFF/Chunks.cpp
@@ -648,6 +648,13 @@ void SectionChunk::getRuntimePseudoRelocs(
         dyn_cast_or_null<Defined>(file->getSymbol(rel.SymbolTableIndex));
     if (!target || !target->isRuntimePseudoReloc)
       continue;
+    // If the target doesn't have a chunk allocated, it may be a
+    // DefinedImportData symbol which ended up unnecessary after GC.
+    // Normally we wouldn't eliminate section chunks that are referenced, but
+    // references within DWARF sections don't count for keeping section chunks
+    // alive. Thus such dangling references in DWARF sections are expected.
+    if (!target->getChunk())
+      continue;
     int sizeInBits =
         getRuntimePseudoRelocSize(rel.Type, file->ctx.config.machine);
     if (sizeInBits == 0) {
diff --git a/lld/test/COFF/autoimport-gc.s b/lld/test/COFF/autoimport-gc.s
new file mode 100644
index 0000000000000..fef6c02eba82f
--- /dev/null
+++ b/lld/test/COFF/autoimport-gc.s
@@ -0,0 +1,41 @@
+# REQUIRES: x86
+# RUN: split-file %s %t.dir
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/lib.s -filetype=obj -o %t.dir/lib.obj
+# RUN: lld-link -out:%t.dir/lib.dll -dll -entry:DllMainCRTStartup %t.dir/lib.obj -lldmingw -implib:%t.dir/lib.lib
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/main.s -filetype=obj -o %t.dir/main.obj
+# RUN: lld-link -lldmingw -out:%t.dir/main.exe -entry:main %t.dir/main.obj %t.dir/lib.lib -opt:ref -debug:dwarf
+
+#--- main.s
+    .global main
+    .section .text$main,"xr",one_only,main
+main:
+    ret
+
+    .global other
+    .section .text$other,"xr",one_only,other
+other:
+    movq .refptr.variable(%rip), %rax
+    movl (%rax), %eax
+    ret
+
+    .section .rdata$.refptr.variable,"dr",discard,.refptr.variable
+    .global .refptr.variable
+.refptr.variable:
+    .quad   variable
+
+    .section .debug_info
+    .long 1
+    .quad variable
+    .long 2
+
+#--- lib.s
+    .global variable
+    .global DllMainCRTStartup
+    .text
+DllMainCRTStartup:
+    ret
+    .data
+variable:
+    .long 42

From 64dc55835a1552dbe74eef79460094014edf659e Mon Sep 17 00:00:00 2001
From: David Tenty <daltenty@ibm.com>
Date: Mon, 15 Apr 2024 13:18:01 -0400
Subject: [PATCH 452/886] [NFC][libunwind][AIX] Add diagnostic push and TODO

clang treats the cast here as an error since 999d4f840777bf8de26d45947192aa0728edc0fb landed
with -Xextra -Werror. We believe this cast to be safe for the reasons noted in comment, but
we should do some further cleanup at some point.
---
 libcxxabi/src/aix_state_tab_eh.inc | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/libcxxabi/src/aix_state_tab_eh.inc b/libcxxabi/src/aix_state_tab_eh.inc
index 0cd94834b37ad..9f46001b02090 100644
--- a/libcxxabi/src/aix_state_tab_eh.inc
+++ b/libcxxabi/src/aix_state_tab_eh.inc
@@ -195,8 +195,13 @@ static void invoke_destructor(FSMEntry* fsmEntry, void* addr) {
       _LIBCXXABI_TRACE_STATETAB0("returned from scalar destructor\n");
     } else {
       _LIBCXXABI_TRACE_STATETAB0("calling vector destructor\n");
+      // TODO: in the legacy ABI, destructors had a second argument. We don't expect to encounter
+      // destructors of this type in the itanium-based ABI, so this should be safe, but this could use some cleanup.
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wcast-function-type"
       __cxa_vec_cleanup(addr, reinterpret_cast<size_t>(fsmEntry->elementCount), fsmEntry->elemSize,
                         reinterpret_cast<destruct_f>(fsmEntry->destructor));
+#pragma GCC diagnostic pop
       _LIBCXXABI_TRACE_STATETAB0("returned from vector destructor\n");
     }
   } catch (...) {

From 4d28d3f93bfee3cecbeb234d1ce306250e12e912 Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Tue, 16 Apr 2024 01:28:28 +0800
Subject: [PATCH 453/886] [SDAG] Turn umin into smin if the saturation pattern
 is broken (#88505)

As we canonicalizes smin with non-negative operands into umin in the
middle-end, the saturation pattern will be broken.
This patch reverts the transform in DAGCombine to fix the regression on
ARM.

Fixes https://github.com/llvm/llvm-project/issues/85706.
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 11 ++-
 llvm/test/CodeGen/AMDGPU/umed3.ll             |  3 +-
 llvm/test/CodeGen/ARM/usat.ll                 | 80 ++++++++-----------
 3 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index dbbe8d9193c25..cbba3a294b3d6 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -5577,9 +5577,12 @@ SDValue DAGCombiner::visitIMINMAX(SDNode *N) {
     return RMINMAX;
 
   // Is sign bits are zero, flip between UMIN/UMAX and SMIN/SMAX.
-  // Only do this if the current op isn't legal and the flipped is.
-  if (!TLI.isOperationLegal(Opcode, VT) &&
-      (N0.isUndef() || DAG.SignBitIsZero(N0)) &&
+  // Only do this if:
+  // 1. The current op isn't legal and the flipped is.
+  // 2. The saturation pattern is broken by canonicalization in InstCombine.
+  bool IsOpIllegal = !TLI.isOperationLegal(Opcode, VT);
+  bool IsSatBroken = Opcode == ISD::UMIN && N0.getOpcode() == ISD::SMAX;
+  if ((IsSatBroken || IsOpIllegal) && (N0.isUndef() || DAG.SignBitIsZero(N0)) &&
       (N1.isUndef() || DAG.SignBitIsZero(N1))) {
     unsigned AltOpcode;
     switch (Opcode) {
@@ -5589,7 +5592,7 @@ SDValue DAGCombiner::visitIMINMAX(SDNode *N) {
     case ISD::UMAX: AltOpcode = ISD::SMAX; break;
     default: llvm_unreachable("Unknown MINMAX opcode");
     }
-    if (TLI.isOperationLegal(AltOpcode, VT))
+    if ((IsSatBroken && IsOpIllegal) || TLI.isOperationLegal(AltOpcode, VT))
       return DAG.getNode(AltOpcode, DL, VT, N0, N1);
   }
 
diff --git a/llvm/test/CodeGen/AMDGPU/umed3.ll b/llvm/test/CodeGen/AMDGPU/umed3.ll
index a2d99f1f8c2c2..557d023c45f9d 100644
--- a/llvm/test/CodeGen/AMDGPU/umed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/umed3.ll
@@ -43,8 +43,7 @@ define amdgpu_kernel void @v_test_umed3_multi_use_r_i_i_i32(ptr addrspace(1) %ou
 }
 
 ; GCN-LABEL: {{^}}v_test_umed3_r_i_i_sign_mismatch_i32:
-; GCN: v_max_i32_e32 v{{[0-9]+}}, 12, v{{[0-9]+}}
-; GCN: v_min_u32_e32 v{{[0-9]+}}, 17, v{{[0-9]+}}
+; GCN: v_med3_i32 v{{[0-9]+}}, v{{[0-9]+}}, 12, 17
 define amdgpu_kernel void @v_test_umed3_r_i_i_sign_mismatch_i32(ptr addrspace(1) %out, ptr addrspace(1) %aptr) #1 {
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr i32, ptr addrspace(1) %aptr, i32 %tid
diff --git a/llvm/test/CodeGen/ARM/usat.ll b/llvm/test/CodeGen/ARM/usat.ll
index 024a98dd29346..d01aa1520b326 100644
--- a/llvm/test/CodeGen/ARM/usat.ll
+++ b/llvm/test/CodeGen/ARM/usat.ll
@@ -756,7 +756,7 @@ define i32 @mm_unsigned_sat_upper_lower_1(i32 %x) {
 ; V4T-NEXT:    bic r1, r0, r0, asr #31
 ; V4T-NEXT:    ldr r0, .LCPI20_0
 ; V4T-NEXT:    cmp r1, r0
-; V4T-NEXT:    movlo r0, r1
+; V4T-NEXT:    movlt r0, r1
 ; V4T-NEXT:    bx lr
 ; V4T-NEXT:    .p2align 2
 ; V4T-NEXT:  @ %bb.1:
@@ -765,23 +765,12 @@ define i32 @mm_unsigned_sat_upper_lower_1(i32 %x) {
 ;
 ; V6-LABEL: mm_unsigned_sat_upper_lower_1:
 ; V6:       @ %bb.0: @ %entry
-; V6-NEXT:    bic r1, r0, r0, asr #31
-; V6-NEXT:    ldr r0, .LCPI20_0
-; V6-NEXT:    cmp r1, r0
-; V6-NEXT:    movlo r0, r1
+; V6-NEXT:    usat r0, #23, r0
 ; V6-NEXT:    bx lr
-; V6-NEXT:    .p2align 2
-; V6-NEXT:  @ %bb.1:
-; V6-NEXT:  .LCPI20_0:
-; V6-NEXT:    .long 8388607 @ 0x7fffff
 ;
 ; V6T2-LABEL: mm_unsigned_sat_upper_lower_1:
 ; V6T2:       @ %bb.0: @ %entry
-; V6T2-NEXT:    bic r1, r0, r0, asr #31
-; V6T2-NEXT:    movw r0, #65535
-; V6T2-NEXT:    movt r0, #127
-; V6T2-NEXT:    cmp r1, r0
-; V6T2-NEXT:    movlo r0, r1
+; V6T2-NEXT:    usat r0, #23, r0
 ; V6T2-NEXT:    bx lr
 entry:
   %0 = call i32 @llvm.smax.i32(i32 %x, i32 0)
@@ -795,7 +784,7 @@ define i32 @mm_unsigned_sat_upper_lower_2(i32 %x) {
 ; V4T-NEXT:    bic r1, r0, r0, asr #31
 ; V4T-NEXT:    ldr r0, .LCPI21_0
 ; V4T-NEXT:    cmp r1, r0
-; V4T-NEXT:    movlo r0, r1
+; V4T-NEXT:    movlt r0, r1
 ; V4T-NEXT:    bx lr
 ; V4T-NEXT:    .p2align 2
 ; V4T-NEXT:  @ %bb.1:
@@ -804,23 +793,12 @@ define i32 @mm_unsigned_sat_upper_lower_2(i32 %x) {
 ;
 ; V6-LABEL: mm_unsigned_sat_upper_lower_2:
 ; V6:       @ %bb.0: @ %entry
-; V6-NEXT:    bic r1, r0, r0, asr #31
-; V6-NEXT:    ldr r0, .LCPI21_0
-; V6-NEXT:    cmp r1, r0
-; V6-NEXT:    movlo r0, r1
+; V6-NEXT:    usat r0, #23, r0
 ; V6-NEXT:    bx lr
-; V6-NEXT:    .p2align 2
-; V6-NEXT:  @ %bb.1:
-; V6-NEXT:  .LCPI21_0:
-; V6-NEXT:    .long 8388607 @ 0x7fffff
 ;
 ; V6T2-LABEL: mm_unsigned_sat_upper_lower_2:
 ; V6T2:       @ %bb.0: @ %entry
-; V6T2-NEXT:    bic r1, r0, r0, asr #31
-; V6T2-NEXT:    movw r0, #65535
-; V6T2-NEXT:    movt r0, #127
-; V6T2-NEXT:    cmp r1, r0
-; V6T2-NEXT:    movlo r0, r1
+; V6T2-NEXT:    usat r0, #23, r0
 ; V6T2-NEXT:    bx lr
 entry:
   %0 = call i32 @llvm.smax.i32(i32 %x, i32 0)
@@ -834,7 +812,7 @@ define i32 @mm_unsigned_sat_upper_lower_3(i32 %x) {
 ; V4T-NEXT:    bic r1, r0, r0, asr #31
 ; V4T-NEXT:    ldr r0, .LCPI22_0
 ; V4T-NEXT:    cmp r1, r0
-; V4T-NEXT:    movlo r0, r1
+; V4T-NEXT:    movlt r0, r1
 ; V4T-NEXT:    bx lr
 ; V4T-NEXT:    .p2align 2
 ; V4T-NEXT:  @ %bb.1:
@@ -843,23 +821,12 @@ define i32 @mm_unsigned_sat_upper_lower_3(i32 %x) {
 ;
 ; V6-LABEL: mm_unsigned_sat_upper_lower_3:
 ; V6:       @ %bb.0: @ %entry
-; V6-NEXT:    bic r1, r0, r0, asr #31
-; V6-NEXT:    ldr r0, .LCPI22_0
-; V6-NEXT:    cmp r1, r0
-; V6-NEXT:    movlo r0, r1
+; V6-NEXT:    usat r0, #23, r0
 ; V6-NEXT:    bx lr
-; V6-NEXT:    .p2align 2
-; V6-NEXT:  @ %bb.1:
-; V6-NEXT:  .LCPI22_0:
-; V6-NEXT:    .long 8388607 @ 0x7fffff
 ;
 ; V6T2-LABEL: mm_unsigned_sat_upper_lower_3:
 ; V6T2:       @ %bb.0: @ %entry
-; V6T2-NEXT:    bic r1, r0, r0, asr #31
-; V6T2-NEXT:    movw r0, #65535
-; V6T2-NEXT:    movt r0, #127
-; V6T2-NEXT:    cmp r1, r0
-; V6T2-NEXT:    movlo r0, r1
+; V6T2-NEXT:    usat r0, #23, r0
 ; V6T2-NEXT:    bx lr
 entry:
   %0 = call i32 @llvm.smax.i32(i32 %x, i32 0)
@@ -913,7 +880,7 @@ define i32 @mm_no_unsigned_sat_incorrect_constant2(i32 %x) {
 ; V4T-NEXT:    mov r0, #1
 ; V4T-NEXT:    orr r0, r0, #8388608
 ; V4T-NEXT:    cmp r1, #8388608
-; V4T-NEXT:    movls r0, r1
+; V4T-NEXT:    movle r0, r1
 ; V4T-NEXT:    bx lr
 ;
 ; V6-LABEL: mm_no_unsigned_sat_incorrect_constant2:
@@ -922,7 +889,7 @@ define i32 @mm_no_unsigned_sat_incorrect_constant2(i32 %x) {
 ; V6-NEXT:    mov r0, #1
 ; V6-NEXT:    orr r0, r0, #8388608
 ; V6-NEXT:    cmp r1, #8388608
-; V6-NEXT:    movls r0, r1
+; V6-NEXT:    movle r0, r1
 ; V6-NEXT:    bx lr
 ;
 ; V6T2-LABEL: mm_no_unsigned_sat_incorrect_constant2:
@@ -931,7 +898,7 @@ define i32 @mm_no_unsigned_sat_incorrect_constant2(i32 %x) {
 ; V6T2-NEXT:    movw r0, #1
 ; V6T2-NEXT:    movt r0, #128
 ; V6T2-NEXT:    cmp r1, #8388608
-; V6T2-NEXT:    movls r0, r1
+; V6T2-NEXT:    movle r0, r1
 ; V6T2-NEXT:    bx lr
 entry:
   %0 = call i32 @llvm.smax.i32(i32 %x, i32 0)
@@ -981,6 +948,29 @@ entry:
   ret i32 %1
 }
 
+define i32 @test_umin_smax_usat(i32 %x) {
+; V4T-LABEL: test_umin_smax_usat:
+; V4T:       @ %bb.0: @ %entry
+; V4T-NEXT:    bic r0, r0, r0, asr #31
+; V4T-NEXT:    cmp r0, #255
+; V4T-NEXT:    movge r0, #255
+; V4T-NEXT:    bx lr
+;
+; V6-LABEL: test_umin_smax_usat:
+; V6:       @ %bb.0: @ %entry
+; V6-NEXT:    usat r0, #8, r0
+; V6-NEXT:    bx lr
+;
+; V6T2-LABEL: test_umin_smax_usat:
+; V6T2:       @ %bb.0: @ %entry
+; V6T2-NEXT:    usat r0, #8, r0
+; V6T2-NEXT:    bx lr
+entry:
+  %v1 = tail call i32 @llvm.smax.i32(i32 %x, i32 0)
+  %v2 = tail call i32 @llvm.umin.i32(i32 %v1, i32 255)
+  ret i32 %v2
+}
+
 declare i32 @llvm.smin.i32(i32, i32)
 declare i32 @llvm.smax.i32(i32, i32)
 declare i16 @llvm.smin.i16(i16, i16)

From b2f07a91192596a3584040208b7fac4f5ef0dd9c Mon Sep 17 00:00:00 2001
From: Florian Mayer <fmayer@google.com>
Date: Mon, 15 Apr 2024 10:47:13 -0700
Subject: [PATCH 454/886] Revert "[clang analysis] ExprMutationAnalyzer avoid
 infinite recursion for recursive forwarding reference" (#88765)

Reverts llvm/llvm-project#87954

Broke sanitizer bots, e.g.
https://lab.llvm.org/buildbot/#/builders/239/builds/6587/steps/10/logs/stdio
---
 clang-tools-extra/docs/ReleaseNotes.rst       |  4 ---
 .../misc/const-correctness-templates.cpp      | 15 ----------
 .../Analysis/Analyses/ExprMutationAnalyzer.h  | 28 +++++------------
 clang/lib/Analysis/ExprMutationAnalyzer.cpp   | 22 +++++---------
 .../Analysis/ExprMutationAnalyzerTest.cpp     | 30 -------------------
 5 files changed, 15 insertions(+), 84 deletions(-)

diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index 7095c564444fe..4dfbd8ca49ab9 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -221,10 +221,6 @@ Changes in existing checks
   <clang-tidy/checks/llvm/header-guard>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
 
-- Improved :doc:`misc-const-correctness
-  <clang-tidy/checks/misc/const-correctness>` check by avoiding infinite recursion
-  for recursive forwarding reference.
-
 - Improved :doc:`misc-definitions-in-headers
   <clang-tidy/checks/misc/definitions-in-headers>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
diff --git a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
index 248374a71dd40..9da468128743e 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
@@ -58,18 +58,3 @@ void concatenate3(Args... args)
     (..., (stream << args));
 }
 } // namespace gh70323
-
-namespace gh60895 {
-
-template <class T> void f1(T &&a);
-template <class T> void f2(T &&a);
-template <class T> void f1(T &&a) { f2<T>(a); }
-template <class T> void f2(T &&a) { f1<T>(a); }
-void f() {
-  int x = 0;
-  // CHECK-MESSAGES:[[@LINE-1]]:3: warning: variable 'x' of type 'int' can be declared 'const'
-  // CHECK-FIXES: int const x = 0;
-  f1(x);
-}
-
-} // namespace gh60895
diff --git a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
index c4e5d0badb8e5..1ceef944fbc34 100644
--- a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
+++ b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
@@ -8,10 +8,11 @@
 #ifndef LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 #define LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 
+#include <type_traits>
+
 #include "clang/AST/AST.h"
 #include "clang/ASTMatchers/ASTMatchers.h"
 #include "llvm/ADT/DenseMap.h"
-#include <variant>
 
 namespace clang {
 
@@ -21,15 +22,8 @@ class FunctionParmMutationAnalyzer;
 /// a given statement.
 class ExprMutationAnalyzer {
 public:
-  friend class FunctionParmMutationAnalyzer;
-  struct Cache {
-    llvm::SmallDenseMap<const FunctionDecl *,
-                        std::unique_ptr<FunctionParmMutationAnalyzer>>
-        FuncParmAnalyzer;
-  };
-
   ExprMutationAnalyzer(const Stmt &Stm, ASTContext &Context)
-      : ExprMutationAnalyzer(Stm, Context, std::make_shared<Cache>()) {}
+      : Stm(Stm), Context(Context) {}
 
   bool isMutated(const Expr *Exp) { return findMutation(Exp) != nullptr; }
   bool isMutated(const Decl *Dec) { return findMutation(Dec) != nullptr; }
@@ -51,11 +45,6 @@ class ExprMutationAnalyzer {
   using MutationFinder = const Stmt *(ExprMutationAnalyzer::*)(const Expr *);
   using ResultMap = llvm::DenseMap<const Expr *, const Stmt *>;
 
-  ExprMutationAnalyzer(const Stmt &Stm, ASTContext &Context,
-                       std::shared_ptr<Cache> CrossAnalysisCache)
-      : Stm(Stm), Context(Context),
-        CrossAnalysisCache(std::move(CrossAnalysisCache)) {}
-
   const Stmt *findMutationMemoized(const Expr *Exp,
                                    llvm::ArrayRef<MutationFinder> Finders,
                                    ResultMap &MemoizedResults);
@@ -80,7 +69,9 @@ class ExprMutationAnalyzer {
 
   const Stmt &Stm;
   ASTContext &Context;
-  std::shared_ptr<Cache> CrossAnalysisCache;
+  llvm::DenseMap<const FunctionDecl *,
+                 std::unique_ptr<FunctionParmMutationAnalyzer>>
+      FuncParmAnalyzer;
   ResultMap Results;
   ResultMap PointeeResults;
 };
@@ -89,12 +80,7 @@ class ExprMutationAnalyzer {
 // params.
 class FunctionParmMutationAnalyzer {
 public:
-  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context)
-      : FunctionParmMutationAnalyzer(
-            Func, Context, std::make_shared<ExprMutationAnalyzer::Cache>()) {}
-  FunctionParmMutationAnalyzer(
-      const FunctionDecl &Func, ASTContext &Context,
-      std::shared_ptr<ExprMutationAnalyzer::Cache> CrossAnalysisCache);
+  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context);
 
   bool isMutated(const ParmVarDecl *Parm) {
     return findMutation(Parm) != nullptr;
diff --git a/clang/lib/Analysis/ExprMutationAnalyzer.cpp b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
index 503ca4bea99e5..bb042760d297a 100644
--- a/clang/lib/Analysis/ExprMutationAnalyzer.cpp
+++ b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
@@ -638,10 +638,9 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
       if (!RefType->getPointeeType().getQualifiers() &&
           RefType->getPointeeType()->getAs<TemplateTypeParmType>()) {
         std::unique_ptr<FunctionParmMutationAnalyzer> &Analyzer =
-            CrossAnalysisCache->FuncParmAnalyzer[Func];
+            FuncParmAnalyzer[Func];
         if (!Analyzer)
-          Analyzer.reset(new FunctionParmMutationAnalyzer(*Func, Context,
-                                                          CrossAnalysisCache));
+          Analyzer.reset(new FunctionParmMutationAnalyzer(*Func, Context));
         if (Analyzer->findMutation(Parm))
           return Exp;
         continue;
@@ -654,15 +653,13 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
 }
 
 FunctionParmMutationAnalyzer::FunctionParmMutationAnalyzer(
-    const FunctionDecl &Func, ASTContext &Context,
-    std::shared_ptr<ExprMutationAnalyzer::Cache> CrossAnalysisCache)
-    : BodyAnalyzer(*Func.getBody(), Context, CrossAnalysisCache) {
+    const FunctionDecl &Func, ASTContext &Context)
+    : BodyAnalyzer(*Func.getBody(), Context) {
   if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(&Func)) {
     // CXXCtorInitializer might also mutate Param but they're not part of
     // function body, check them eagerly here since they're typically trivial.
     for (const CXXCtorInitializer *Init : Ctor->inits()) {
-      ExprMutationAnalyzer InitAnalyzer(*Init->getInit(), Context,
-                                        CrossAnalysisCache);
+      ExprMutationAnalyzer InitAnalyzer(*Init->getInit(), Context);
       for (const ParmVarDecl *Parm : Ctor->parameters()) {
         if (Results.contains(Parm))
           continue;
@@ -678,14 +675,11 @@ FunctionParmMutationAnalyzer::findMutation(const ParmVarDecl *Parm) {
   const auto Memoized = Results.find(Parm);
   if (Memoized != Results.end())
     return Memoized->second;
-  // To handle call A -> call B -> call A. Assume parameters of A is not mutated
-  // before analyzing parameters of A. Then when analyzing the second "call A",
-  // FunctionParmMutationAnalyzer can use this memoized value to avoid infinite
-  // recursion.
-  Results[Parm] = nullptr;
+
   if (const Stmt *S = BodyAnalyzer.findMutation(Parm))
     return Results[Parm] = S;
-  return Results[Parm];
+
+  return Results[Parm] = nullptr;
 }
 
 } // namespace clang
diff --git a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
index 9c1dc1a76db63..f58ce4aebcbfc 100644
--- a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
+++ b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
@@ -977,36 +977,6 @@ TEST(ExprMutationAnalyzerTest, FollowFuncArgModified) {
                          "void f() { int x; g(x); }");
   Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
   EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("g(x)"));
-
-  AST = buildASTFromCode(
-      StdRemoveReference + StdForward +
-      "template <class T> void f1(T &&a);"
-      "template <class T> void f2(T &&a);"
-      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
-      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
-      "void f() { int x; f1(x); }");
-  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
-  EXPECT_FALSE(isMutated(Results, AST.get()));
-
-  AST = buildASTFromCode(
-      StdRemoveReference + StdForward +
-      "template <class T> void f1(T &&a);"
-      "template <class T> void f2(T &&a);"
-      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
-      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); a++; }"
-      "void f() { int x; f1(x); }");
-  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
-  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
-
-  AST = buildASTFromCode(
-      StdRemoveReference + StdForward +
-      "template <class T> void f1(T &&a);"
-      "template <class T> void f2(T &&a);"
-      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); a++; }"
-      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
-      "void f() { int x; f1(x); }");
-  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
-  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
 }
 
 TEST(ExprMutationAnalyzerTest, FollowFuncArgNotModified) {

From b5b17bf613010b5ea900d2981365b9d2c846a20f Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Tue, 16 Apr 2024 01:50:22 +0800
Subject: [PATCH 455/886] [RISCV] Fix assertion failure in `genShXAddAddShift`
 (#88757)

Fix assertion failure in our downstream CI
https://github.com/dtcxzyw/llvm-codegen-benchmark/issues/1.
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp |  2 +-
 llvm/test/CodeGen/RISCV/rv64zba.ll       | 23 +++++++++++++++++++++++
 2 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 508f607fab20f..b0fda040519a5 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -1986,7 +1986,7 @@ genShXAddAddShift(MachineInstr &Root, unsigned AddOpIdx,
       MRI.getUniqueVRegDef(AddMI->getOperand(AddOpIdx).getReg());
 
   unsigned InnerShiftAmt = ShiftMI->getOperand(2).getImm();
-  assert(InnerShiftAmt > OuterShiftAmt && "Unexpected shift amount");
+  assert(InnerShiftAmt >= OuterShiftAmt && "Unexpected shift amount");
 
   unsigned InnerOpc;
   switch (InnerShiftAmt - OuterShiftAmt) {
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index 067addc819f7e..bb4be323ecb2e 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -2389,3 +2389,26 @@ define i64 @array_index_sh4_sh3(ptr %p, i64 %idx1, i64 %idx2) {
   %b = load i64, ptr %a, align 8
   ret i64 %b
 }
+
+define ptr @test_gep_gep_dont_crash(ptr %p, i64 %a1, i64 %a2) {
+; RV64I-LABEL: test_gep_gep_dont_crash:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    srliw a2, a2, 6
+; RV64I-NEXT:    slli a2, a2, 3
+; RV64I-NEXT:    slli a1, a1, 3
+; RV64I-NEXT:    add a0, a0, a1
+; RV64I-NEXT:    add a0, a0, a2
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: test_gep_gep_dont_crash:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    srliw a2, a2, 6
+; RV64ZBA-NEXT:    add a1, a2, a1
+; RV64ZBA-NEXT:    sh3add a0, a1, a0
+; RV64ZBA-NEXT:    ret
+  %lshr = lshr i64 %a2, 6
+  %and = and i64 %lshr, 67108863
+  %gep1 = getelementptr i64, ptr %p, i64 %and
+  %gep2 = getelementptr i64, ptr %gep1, i64 %a1
+  ret ptr %gep2
+}

From 09c3bfe9b3eb47a2af0c10531b25f90cfb5fa9f4 Mon Sep 17 00:00:00 2001
From: Leonard Chan <leonardchan@google.com>
Date: Thu, 11 Apr 2024 14:51:58 -0700
Subject: [PATCH 456/886] Reapply "[llvm] Teach whole program devirtualization
 about relative vtables"

This reverts commit 474f5efebed24547e76d022f0c5ffcc9db97ce6f.
---
 .../include/llvm/Analysis/TypeMetadataUtils.h |   2 +-
 llvm/lib/Analysis/ModuleSummaryAnalysis.cpp   |  37 ++++-
 llvm/lib/Analysis/TypeMetadataUtils.cpp       |  16 +-
 llvm/test/ThinLTO/X86/devirt.ll               |  95 ++++++++++-
 .../WholeProgramDevirt/Inputs/export.yaml     |   8 +
 .../WholeProgramDevirt/branch-funnel.ll       | 157 ++++++++++++++++++
 .../WholeProgramDevirt/constant-arg.ll        |  45 +++++
 .../devirt-single-impl-check.ll               |  25 +++
 .../devirt-single-impl-multiple-assumes.ll    |  20 +++
 .../WholeProgramDevirt/devirt-single-impl.ll  |  47 +++++-
 .../WholeProgramDevirt/pointer-vtable.ll      |  16 ++
 11 files changed, 458 insertions(+), 10 deletions(-)

diff --git a/llvm/include/llvm/Analysis/TypeMetadataUtils.h b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
index 8894945c28d94..2c7ce0858b9d3 100644
--- a/llvm/include/llvm/Analysis/TypeMetadataUtils.h
+++ b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
@@ -66,7 +66,7 @@ void findDevirtualizableCallsForTypeCheckedLoad(
 /// Used for example from GlobalDCE to find an entry in a C++ vtable that
 /// matches a vcall offset.
 ///
-/// To support Swift vtables, getPointerAtOffset can see through "relative
+/// To support relative vtables, getPointerAtOffset can see through "relative
 /// pointers", i.e. (sub-)expressions of the form of:
 ///
 /// @symbol = ... {
diff --git a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
index deda1eebb3b57..c3d15afe6d9cb 100644
--- a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
+++ b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
@@ -22,6 +22,7 @@
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Analysis/BlockFrequencyInfo.h"
 #include "llvm/Analysis/BranchProbabilityInfo.h"
+#include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/IndirectCallPromotionAnalysis.h"
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/MemoryProfileInfo.h"
@@ -668,7 +669,8 @@ static void computeFunctionSummary(
 /// within the initializer.
 static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
                              const Module &M, ModuleSummaryIndex &Index,
-                             VTableFuncList &VTableFuncs) {
+                             VTableFuncList &VTableFuncs,
+                             const GlobalVariable &OrigGV) {
   // First check if this is a function pointer.
   if (I->getType()->isPointerTy()) {
     auto C = I->stripPointerCasts();
@@ -696,7 +698,7 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
       auto Offset = SL->getElementOffset(EI.index());
       unsigned Op = SL->getElementContainingOffset(Offset);
       findFuncPointers(cast<Constant>(I->getOperand(Op)),
-                       StartingOffset + Offset, M, Index, VTableFuncs);
+                       StartingOffset + Offset, M, Index, VTableFuncs, OrigGV);
     }
   } else if (auto *C = dyn_cast<ConstantArray>(I)) {
     ArrayType *ATy = C->getType();
@@ -704,7 +706,34 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
     uint64_t EltSize = DL.getTypeAllocSize(EltTy);
     for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i) {
       findFuncPointers(cast<Constant>(I->getOperand(i)),
-                       StartingOffset + i * EltSize, M, Index, VTableFuncs);
+                       StartingOffset + i * EltSize, M, Index, VTableFuncs,
+                       OrigGV);
+    }
+  } else if (const auto *CE = dyn_cast<ConstantExpr>(I)) {
+    // For relative vtables, the next sub-component should be a trunc.
+    if (CE->getOpcode() != Instruction::Trunc ||
+        !(CE = dyn_cast<ConstantExpr>(CE->getOperand(0))))
+      return;
+
+    // If this constant can be reduced to the offset between a function and a
+    // global, then we know this is a valid virtual function if the RHS is the
+    // original vtable we're scanning through.
+    if (CE->getOpcode() == Instruction::Sub) {
+      GlobalValue *LHS, *RHS;
+      APSInt LHSOffset, RHSOffset;
+      if (IsConstantOffsetFromGlobal(CE->getOperand(0), LHS, LHSOffset, DL) &&
+          IsConstantOffsetFromGlobal(CE->getOperand(1), RHS, RHSOffset, DL) &&
+          RHS == &OrigGV &&
+
+          // For relative vtables, this component should point to the callable
+          // function without any offsets.
+          LHSOffset == 0 &&
+
+          // Also, the RHS should always point to somewhere within the vtable.
+          RHSOffset <=
+              static_cast<uint64_t>(DL.getTypeAllocSize(OrigGV.getInitializer()->getType()))) {
+        findFuncPointers(LHS, StartingOffset, M, Index, VTableFuncs, OrigGV);
+      }
     }
   }
 }
@@ -717,7 +746,7 @@ static void computeVTableFuncs(ModuleSummaryIndex &Index,
     return;
 
   findFuncPointers(V.getInitializer(), /*StartingOffset=*/0, M, Index,
-                   VTableFuncs);
+                   VTableFuncs, V);
 
 #ifndef NDEBUG
   // Validate that the VTableFuncs list is ordered by offset.
diff --git a/llvm/lib/Analysis/TypeMetadataUtils.cpp b/llvm/lib/Analysis/TypeMetadataUtils.cpp
index b8dcc39e9223c..ee400175ef11c 100644
--- a/llvm/lib/Analysis/TypeMetadataUtils.cpp
+++ b/llvm/lib/Analysis/TypeMetadataUtils.cpp
@@ -67,6 +67,14 @@ static void findLoadCallsAtConstantOffset(
         findLoadCallsAtConstantOffset(M, DevirtCalls, User, Offset + GEPOffset,
                                       CI, DT);
       }
+    } else if (auto *Call = dyn_cast<CallInst>(User)) {
+      if (Call->getIntrinsicID() == llvm::Intrinsic::load_relative) {
+        if (auto *LoadOffset = dyn_cast<ConstantInt>(Call->getOperand(1))) {
+          findCallsAtConstantOffset(DevirtCalls, nullptr, User,
+                                    Offset + LoadOffset->getSExtValue(), CI,
+                                    DT);
+        }
+      }
     }
   }
 }
@@ -131,6 +139,12 @@ void llvm::findDevirtualizableCallsForTypeCheckedLoad(
 
 Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                                    Constant *TopLevelGlobal) {
+  // TODO: Ideally it would be the caller who knows if it's appropriate to strip
+  // the DSOLocalEquicalent. More generally, it would feel more appropriate to
+  // have two functions that handle absolute and relative pointers separately.
+  if (auto *Equiv = dyn_cast<DSOLocalEquivalent>(I))
+    I = Equiv->getGlobalValue();
+
   if (I->getType()->isPointerTy()) {
     if (Offset == 0)
       return I;
@@ -161,7 +175,7 @@ Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                               Offset % ElemSize, M, TopLevelGlobal);
   }
 
-  // (Swift-specific) relative-pointer support starts here.
+  // Relative-pointer support starts here.
   if (auto *CI = dyn_cast<ConstantInt>(I)) {
     if (Offset == 0 && CI->isZero()) {
       return I;
diff --git a/llvm/test/ThinLTO/X86/devirt.ll b/llvm/test/ThinLTO/X86/devirt.ll
index 472e43d776803..a40f89e644f00 100644
--- a/llvm/test/ThinLTO/X86/devirt.ll
+++ b/llvm/test/ThinLTO/X86/devirt.ll
@@ -27,24 +27,36 @@
 ; NOENABLESPLITFLAG-DAG: [[B:\^[0-9]+]] = gv: (name: "_ZTV1B", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[Bf]], [[An]])
 ; NOENABLESPLITFLAG-DAG: [[C:\^[0-9]+]] = gv: (name: "_ZTV1C", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[An]], [[Cf]])
 ; NOENABLESPLITFLAG-DAG: [[D:\^[0-9]+]] = gv: (name: "_ZTV1D", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 16)), refs: ([[Dm]])
+; NOENABLESPLITFLAG-DAG: [[B_RV:\^[0-9]+]] = gv: (name: "_ZTV1B_RV", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[B_RV]], [[Bf]], [[An]])
+; NOENABLESPLITFLAG-DAG: [[C_RV:\^[0-9]+]] = gv: (name: "_ZTV1C_RV", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[C_RV]], [[An]], [[Cf]])
+; NOENABLESPLITFLAG-DAG: [[D_RV:\^[0-9]+]] = gv: (name: "_ZTV1D_RV", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 8)), refs: ([[D_RV]], [[Dm]])
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A", summary: ((offset: 16, [[B]]), (offset: 16, [[C]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B", summary: ((offset: 16, [[B]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C", summary: ((offset: 16, [[C]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A_RV", summary: ((offset: 8, [[B_RV]]), (offset: 8, [[C_RV]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B_RV", summary: ((offset: 8, [[B_RV]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C_RV", summary: ((offset: 8, [[C_RV]])))
 ; Type Id on _ZTV1D should have been promoted
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "1.{{.*}}", summary: ((offset: 16, [[D]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "2.{{.*}}", summary: ((offset: 8, [[D_RV]])))
 
 ; Index based WPD
 ; RUN: llvm-lto2 run %t2.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
+; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK
+; RUN:   -r=%t2.o,_ZTV1D,px \
+; RUN:   -r=%t2.o,_ZTV1B_RV,px \
+; RUN:   -r=%t2.o,_ZTV1C_RV,px \
+; RUN:   -r=%t2.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
 ; Check that we're able to prevent specific function from being
@@ -54,18 +66,24 @@
 ; RUN:   -wholeprogramdevirt-skip=_ZN1A1nEi \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
+; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=SKIP
+; RUN:   -r=%t2.o,_ZTV1D,px \
+; RUN:   -r=%t2.o,_ZTV1B_RV,px \
+; RUN:   -r=%t2.o,_ZTV1C_RV,px \
+; RUN:   -r=%t2.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=SKIP
 
 ; RUN: llvm-lto2 run %t.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t.o,test,px \
+; RUN:   -r=%t.o,test_rv,px \
 ; RUN:   -r=%t.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t.o,_ZN1C1fEi,p \
@@ -73,15 +91,24 @@
 ; RUN:   -r=%t.o,_ZTV1B, \
 ; RUN:   -r=%t.o,_ZTV1C, \
 ; RUN:   -r=%t.o,_ZTV1D, \
+; RUN:   -r=%t.o,_ZTV1B_RV, \
+; RUN:   -r=%t.o,_ZTV1C_RV, \
+; RUN:   -r=%t.o,_ZTV1D_RV, \
 ; RUN:   -r=%t.o,_ZN1A1nEi, \
 ; RUN:   -r=%t.o,_ZN1B1fEi, \
 ; RUN:   -r=%t.o,_ZN1C1fEi, \
 ; RUN:   -r=%t.o,_ZN1D1mEi, \
 ; RUN:   -r=%t.o,_ZTV1B,px \
 ; RUN:   -r=%t.o,_ZTV1C,px \
-; RUN:   -r=%t.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
+; RUN:   -r=%t.o,_ZTV1D,px \
+; RUN:   -r=%t.o,_ZTV1B_RV,px \
+; RUN:   -r=%t.o,_ZTV1C_RV,px \
+; RUN:   -r=%t.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
+; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
+; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 
@@ -99,6 +126,25 @@ target triple = "x86_64-grtev4-linux-gnu"
 @_ZTV1C = constant { [4 x ptr] } { [4 x ptr] [ptr null, ptr undef, ptr @_ZN1C1fEi, ptr @_ZN1A1nEi] }, !type !0, !type !2
 @_ZTV1D = constant { [3 x ptr] } { [3 x ptr] [ptr null, ptr undef, ptr @_ZN1D1mEi] }, !type !3
 
+@_ZTV1B_RV = constant { [4 x i32] } { [4 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1B1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 3) to i64)) to i32)
+] }, !type !7, !type !8
+
+@_ZTV1C_RV = constant { [4 x i32] } { [4 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1C1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 3) to i64)) to i32)
+] }, !type !7, !type !9
+
+@_ZTV1D_RV = constant { [3 x i32] } { [3 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1D1mEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1D_RV, i32 0, i32 0, i32 2) to i64)) to i32)
+] }, !type !10
 
 ; CHECK-IR-LABEL: define {{(noundef )?}}i32 @test
 define i32 @test(ptr %obj, ptr %obj2, i32 %a) {
@@ -136,6 +182,43 @@ entry:
 ; CHECK-IR-LABEL: ret i32
 ; CHECK-IR-LABEL: }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+; CHECK-IR-LABEL: define i32 @test_rv
+define i32 @test_rv(ptr %obj, ptr %obj2, i32 %a) {
+entry:
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"_ZTS1A_RV")
+  call void @llvm.assume(i1 %p)
+  %fptr1_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 4)
+
+  ; Check that the call was devirtualized.
+  ; CHECK-IR: %call = tail call i32 @_ZN1A1nEi
+  ; Ensure !prof and !callees metadata for indirect call promotion removed.
+  ; CHECK-IR-NOT: prof
+  ; CHECK-IR-NOT: callees
+  %call = tail call i32 %fptr1_rv(ptr nonnull %obj, i32 %a), !prof !5, !callees !6
+
+  %fptr22_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+
+  ; We still have to call it as virtual.
+  ; CHECK-IR: %call3 = tail call i32 %fptr22
+  %call3 = tail call i32 %fptr22_rv(ptr nonnull %obj, i32 %call)
+
+  %vtable2 = load ptr, ptr %obj2
+  %p2 = call i1 @llvm.type.test(ptr %vtable2, metadata !11)
+  call void @llvm.assume(i1 %p2)
+
+  %fptr33_rv = call ptr @llvm.load.relative.i32(ptr %vtable2, i32 0)
+
+  ; Check that the call was devirtualized.
+  ; CHECK-IR: %call4 = tail call i32 @_ZN1D1mEi
+  %call4 = tail call i32 %fptr33_rv(ptr nonnull %obj2, i32 %call3)
+  ret i32 %call4
+}
+; CHECK-IR-LABEL: ret i32
+; CHECK-IR-LABEL: }
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -165,3 +248,9 @@ attributes #0 = { noinline optnone }
 !4 = distinct !{}
 !5 = !{!"VP", i32 0, i64 1, i64 1621563287929432257, i64 1}
 !6 = !{ptr @_ZN1A1nEi}
+
+!7 = !{i64 8, !"_ZTS1A_RV"}
+!8 = !{i64 8, !"_ZTS1B_RV"}
+!9 = !{i64 8, !"_ZTS1C_RV"}
+!10 = !{i64 8, !11}
+!11 = distinct !{}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
index 71cf38b216c71..dd0c90d7ae8c3 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
+++ b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
@@ -5,14 +5,22 @@ GlobalValueMap:
       TypeTestAssumeVCalls:
         - GUID: 14276520915468743435  # typeid1
           Offset: 0
+        - GUID: 271751036925422857  # typeid1_rv
+          Offset: 0
       TypeCheckedLoadVCalls:
         - GUID: 15427464259790519041  # typeid2
           Offset: 0
+        - GUID: 1146149264729288256  # typeid2_rv
+          Offset: 0
       TypeTestAssumeConstVCalls:
         - VFunc:
             GUID: 3515965990081467659  # typeid3
             Offset: 0
           Args: [12, 24]
+        - VFunc:
+            GUID: 2777626534618191571  # typeid3_rv
+            Offset: 0
+          Args: [12, 24]
       TypeCheckedLoadConstVCalls:
         - VFunc:
             GUID: 17525413373118030901  # typeid4
diff --git a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
index b55713fe4b296..0b1023eee2732 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
@@ -8,6 +8,45 @@
 ; RUN: FileCheck --check-prefix=SUMMARY %s < %t
 
 ; SUMMARY:      TypeIdMap:       
+; SUMMARY-NEXT:  typeid1_rv:
+; SUMMARY-NEXT:    TTRes:
+; SUMMARY-NEXT:      Kind:            Unknown
+; SUMMARY-NEXT:      SizeM1BitWidth:  0
+; SUMMARY-NEXT:      AlignLog2:       0
+; SUMMARY-NEXT:      SizeM1:          0
+; SUMMARY-NEXT:      BitMask:         0
+; SUMMARY-NEXT:      InlineBits:      0
+; SUMMARY-NEXT:    WPDRes:
+; SUMMARY-NEXT:      0:
+; SUMMARY-NEXT:        Kind:            BranchFunnel
+; SUMMARY-NEXT:        SingleImplName:  ''
+; SUMMARY-NEXT:        ResByArg:
+; SUMMARY-NEXT:  typeid2_rv:
+; SUMMARY-NEXT:    TTRes:
+; SUMMARY-NEXT:      Kind:            Unknown
+; SUMMARY-NEXT:      SizeM1BitWidth:  0
+; SUMMARY-NEXT:      AlignLog2:       0
+; SUMMARY-NEXT:      SizeM1:          0
+; SUMMARY-NEXT:      BitMask:         0
+; SUMMARY-NEXT:      InlineBits:      0
+; SUMMARY-NEXT:    WPDRes:
+; SUMMARY-NEXT:      0:
+; SUMMARY-NEXT:        Kind:            Indir
+; SUMMARY-NEXT:        SingleImplName:  ''
+; SUMMARY-NEXT:        ResByArg:
+; SUMMARY-NEXT:   typeid3_rv:
+; SUMMARY-NEXT:     TTRes:           
+; SUMMARY-NEXT:       Kind:            Unknown
+; SUMMARY-NEXT:       SizeM1BitWidth:  0
+; SUMMARY-NEXT:       AlignLog2:       0
+; SUMMARY-NEXT:       SizeM1:          0
+; SUMMARY-NEXT:       BitMask:         0
+; SUMMARY-NEXT:       InlineBits:      0
+; SUMMARY-NEXT:     WPDRes:          
+; SUMMARY-NEXT:       0:               
+; SUMMARY-NEXT:         Kind:            BranchFunnel
+; SUMMARY-NEXT:         SingleImplName:  ''
+; SUMMARY-NEXT:         ResByArg:        
 ; SUMMARY-NEXT:   typeid3:
 ; SUMMARY-NEXT:     TTRes:           
 ; SUMMARY-NEXT:       Kind:            Unknown
@@ -93,6 +132,29 @@ declare i32 @vf3_2(ptr %this, i32 %arg)
 declare i32 @vf4_1(ptr %this, i32 %arg)
 declare i32 @vf4_2(ptr %this, i32 %arg)
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+;; These are relative vtables equivalent to the ones above.
+@vt1_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_1 to i64), i64 ptrtoint (ptr @vt1_1_rv to i64)) to i32)], !type !5
+@vt1_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_2 to i64), i64 ptrtoint (ptr @vt1_2_rv to i64)) to i32)], !type !5
+
+@vt2_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_1 to i64), i64 ptrtoint (ptr @vt2_1_rv to i64)) to i32)], !type !6
+@vt2_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_2 to i64), i64 ptrtoint (ptr @vt2_2_rv to i64)) to i32)], !type !6
+@vt2_3_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_3 to i64), i64 ptrtoint (ptr @vt2_3_rv to i64)) to i32)], !type !6
+@vt2_4_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_4 to i64), i64 ptrtoint (ptr @vt2_4_rv to i64)) to i32)], !type !6
+@vt2_5_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_5 to i64), i64 ptrtoint (ptr @vt2_5_rv to i64)) to i32)], !type !6
+@vt2_6_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_6 to i64), i64 ptrtoint (ptr @vt2_6_rv to i64)) to i32)], !type !6
+@vt2_7_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_7 to i64), i64 ptrtoint (ptr @vt2_7_rv to i64)) to i32)], !type !6
+@vt2_8_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_8 to i64), i64 ptrtoint (ptr @vt2_8_rv to i64)) to i32)], !type !6
+@vt2_9_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_9 to i64), i64 ptrtoint (ptr @vt2_9_rv to i64)) to i32)], !type !6
+@vt2_10_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_10 to i64), i64 ptrtoint (ptr @vt2_10_rv to i64)) to i32)], !type !6
+@vt2_11_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_11 to i64), i64 ptrtoint (ptr @vt2_11_rv to i64)) to i32)], !type !6
+
+@vt3_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_1 to i64), i64 ptrtoint (ptr @vt3_1_rv to i64)) to i32)], !type !7
+@vt3_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_2 to i64), i64 ptrtoint (ptr @vt3_2_rv to i64)) to i32)], !type !7
+
+@vt4_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_1 to i64), i64 ptrtoint (ptr @vt4_1_rv to i64)) to i32)], !type !8
+@vt4_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_2 to i64), i64 ptrtoint (ptr @vt4_2_rv to i64)) to i32)], !type !8
 
 
 ; CHECK-LABEL: define i32 @fn1
@@ -108,6 +170,19 @@ define i32 @fn1(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn1_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn1_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest %vtable, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define i32 @fn2
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn2(ptr %obj) #0 {
@@ -120,6 +195,18 @@ define i32 @fn2(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn2_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn2_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call i32 %
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define i32 @fn3
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn3(ptr %obj) #0 {
@@ -133,10 +220,75 @@ define i32 @fn3(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn3_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn3_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !9)
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; RETP: call i32 @branch_funnel.1(ptr
+  ; NORETP: call i32 %
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
+  call void @llvm.assume(i1 %p)
+  %fptr = load ptr, ptr @vt1_1
+  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_cpy
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_cpy(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
+  call void @llvm.assume(i1 %p)
+  %fptr = load ptr, ptr @vt1_1
+  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_rv(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_rv_cpy
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_rv_cpy(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define hidden void @__typeid_typeid1_0_branch_funnel(ptr nest %0, ...)
 ; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2, ptr {{(nonnull )?}}@vf1_2, ...)
 
+; CHECK-LABEL: define hidden void @__typeid_typeid1_rv_0_branch_funnel(ptr nest %0, ...)
+; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1_rv, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2_rv, ptr {{(nonnull )?}}@vf1_2, ...)
+
 ; CHECK: define internal void @branch_funnel(ptr
+; CHECK: define internal void @branch_funnel.1(ptr
 
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
@@ -146,5 +298,10 @@ declare void @llvm.assume(i1)
 !2 = !{i32 0, !"typeid3"}
 !3 = !{i32 0, !4}
 !4 = distinct !{}
+!5 = !{i32 0, !"typeid1_rv"}
+!6 = !{i32 0, !"typeid2_rv"}
+!7 = !{i32 0, !"typeid3_rv"}
+!8 = !{i32 0, !9}
+!9 = distinct !{}
 
 attributes #0 = { "target-features"="+retpoline" }
diff --git a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
index 91bae4af9a211..e685a1a54c520 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
@@ -7,6 +7,10 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf2], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\01", [1 x ptr] [ptr @vf4], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf8], [0 x i8] zeroinitializer }, !type [[T8]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4:![0-9]+]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
 
 @vt1 = constant [1 x ptr] [ptr @vf1], !type !0
 @vt2 = constant [1 x ptr] [ptr @vf2], !type !0
@@ -61,8 +65,49 @@ define i1 @call2(ptr %obj) {
   ret i1 %result
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt1_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)
+], align 4, !type !1
+@vt2_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)
+], align 4, !type !1
+@vt4_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)
+], align 4, !type !1
+@vt8_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define i1 @call3
+define i1 @call3(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: getelementptr {{.*}} -1
+  ; CHECK: and {{.*}}, 1
+  %result = call i1 %fptr(ptr %obj, i32 5)
+  ret i1 %result
+}
+
+; CHECK: define i1 @call4
+define i1 @call4(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: getelementptr {{.*}} -1
+  ; CHECK: and {{.*}}, 2
+  %result = call i1 %fptr(ptr %obj, i32 10)
+  ret i1 %result
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 ; CHECK: [[T8]] = !{i32 8, !"typeid"}
+; CHECK: [[T4]] = !{i32 4, !"typeid2"}
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
index dc7b2024bc591..5a9e6c12c3582 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
@@ -3,6 +3,7 @@
 target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
+; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: devirtualized vf
 ; CHECK-NOT: devirtualized
@@ -33,7 +34,31 @@ trap:
   unreachable
 }
 
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %pair = call {ptr, i1} @llvm.type.checked.load(ptr %vtable, i32 0, metadata !"typeid2")
+  %fptr = extractvalue {ptr, i1} %pair, 0
+  %p = extractvalue {ptr, i1} %pair, 1
+  ; CHECK: br i1 true,
+  br i1 %p, label %cont, label %trap
+
+cont:
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+
+trap:
+  call void @llvm.trap()
+  unreachable
+}
+
 declare {ptr, i1} @llvm.type.checked.load(ptr, i32, metadata)
 declare void @llvm.trap()
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
index ed144c23302dd..bfbbeaf8a0656 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
@@ -23,7 +23,27 @@ define void @call(ptr %obj) {
   ret void
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %p2 = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p2)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
index 1ce96f70d905f..d7f33df594956 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
@@ -7,6 +7,8 @@ target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
 ; CHECK: remark: devirt-single.cc:30:32: single-impl: devirtualized a call to vf
+; CHECK: remark: devirt-single.cc:41:32: single-impl: devirtualized a call to vf
+; CHECK: remark: devirt-single.cc:51:32: single-impl: devirtualized a call to vf
 ; CHECK: remark: devirt-single.cc:13:0: devirtualized vf
 ; CHECK-NOT: devirtualized
 
@@ -28,6 +30,41 @@ define void @call(ptr %obj) #1 !dbg !5 {
   ret void
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !11
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) #1 !dbg !9 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj), !dbg !10
+  ret void
+}
+
+@_ZTV1A.local = private unnamed_addr constant { [3 x i32] } { [3 x i32] [
+  i32 0,  ; offset to top
+  i32 0,  ; rtti
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1A.local, i32 0, i32 0, i32 2) to i64)) to i32)  ; vfunc offset
+] }, align 4, !type !14
+
+; CHECK: define void @call3
+define void @call3(ptr %obj) #1 !dbg !12 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid3")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 8)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj), !dbg !13
+  ret void
+}
+
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -45,5 +82,13 @@ declare void @llvm.assume(i1)
 !7 = distinct !DISubprogram(name: "vf", linkageName: "_ZN3vt12vfEv", scope: !1, file: !1, line: 13, isLocal: false, isDefinition: true, scopeLine: 13, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
 !8 = !{i32 0, !"typeid"}
 
+!9 = distinct !DISubprogram(name: "call2", linkageName: "_Z5call2Pv", scope: !1, file: !1, line: 40, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
+!10 = !DILocation(line: 41, column: 32, scope: !9)
+!11 = !{i32 0, !"typeid2"}
+
+!12 = distinct !DISubprogram(name: "call3", linkageName: "_Z5call3Pv", scope: !1, file: !1, line: 50, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
+!13 = !DILocation(line: 51, column: 32, scope: !12)
+!14 = !{i32 0, !"typeid3"}
+
 ; CHECK: 1 wholeprogramdevirt - Number of whole program devirtualization targets
-; CHECK: 1 wholeprogramdevirt - Number of single implementation devirtualizations
+; CHECK: 3 wholeprogramdevirt - Number of single implementation devirtualizations
diff --git a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
index 3c5d9b839bc4d..062cef9b88fe4 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
@@ -20,7 +20,23 @@ define void @call(ptr %obj) {
   ret void
 }
 
+@vt2 = constant i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt2 to i64)) to i32), !type !1
+
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}

From b8b2f9902cb23af0811ec218a4f43226c0691510 Mon Sep 17 00:00:00 2001
From: Leonard Chan <leonardchan@google.com>
Date: Thu, 11 Apr 2024 14:55:17 -0700
Subject: [PATCH 457/886] Reapply "[llvm] Teach GlobalDCE about
 dso_local_equivalent"

This reverts commit 1c604a9780fcfe92a99d539913553f0835b81de3.
---
 .../include/llvm/Analysis/TypeMetadataUtils.h |  4 ++--
 ...virtual-functions-relative-pointers-bad.ll | 18 ++++++++++++++
 ...virtual-functions-relative-pointers-gep.ll | 24 +++++++++++++++++++
 .../virtual-functions-relative-pointers.ll    | 24 +++++++++++++++++++
 4 files changed, 68 insertions(+), 2 deletions(-)

diff --git a/llvm/include/llvm/Analysis/TypeMetadataUtils.h b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
index 2c7ce0858b9d3..bdb477b54b532 100644
--- a/llvm/include/llvm/Analysis/TypeMetadataUtils.h
+++ b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
@@ -87,8 +87,8 @@ std::pair<Function *, Constant *>
 getFunctionAtVTableOffset(GlobalVariable *GV, uint64_t Offset, Module &M);
 
 /// Finds the same "relative pointer" pattern as described above, where the
-/// target is `F`, and replaces the entire pattern with a constant zero.
-void replaceRelativePointerUsersWithZero(Function *F);
+/// target is `C`, and replaces the entire pattern with a constant zero.
+void replaceRelativePointerUsersWithZero(Constant *C);
 
 } // namespace llvm
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
index 9d87fdb084a12..9a4bff80383cb 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
@@ -16,14 +16,32 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 
 ; CHECK:      @vtable = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 8, !type !0, !type !1, !vcall_visibility !2
 
+@vtable2 = internal unnamed_addr constant { [3 x i32] } { [3 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+
+  ; a "bad" relative pointer because it's base is not the @vtable symbol
+  i32 trunc (i64 sub (i64 ptrtoint (ptr @weird_ref_3         to i64), i64 ptrtoint (ptr @weird_ref_4 to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 0, !"vfunc3.type"}
+!4 = !{i64 4, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 4, !type !3, !type !4, !vcall_visibility !2
+
 define internal void @vfunc1() { ret void }
 define internal void @vfunc2() { ret void }
 define internal void @weird_ref_1() { ret void }
 define internal void @weird_ref_2() { ret void }
+declare void @vfunc3()
+declare void @vfunc4()
+declare void @weird_ref_3()
+declare void @weird_ref_4()
 
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   call void @weird_ref_2()
+  %2 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  call void @weird_ref_4()
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
index fb45d37dde6e4..d89b024e7a6d0 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
@@ -19,6 +19,20 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
+@vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
+  i32 42,
+  i32 1337,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 8, !"vfunc3.type"}
+!4 = !{i64 12, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
+; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
+; CHECK-SAME:   i32 0
+; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
+
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -31,9 +45,19 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
+; (3) vfunc3_live_extern is referenced from @main, stays alive
+; CHECK: declare void @vfunc3_live_extern
+declare void @vfunc3_live_extern()
+
+; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
+; CHECK-NOT: declare void @vfunc4_dead_extern
+declare void @vfunc4_dead_extern()
+
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
+  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
index 76a617ea99e6f..b652765296bc5 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
@@ -17,6 +17,20 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
+; Similar to above, but the vtable is more aligned to how C++ relative vtables look.
+; That is, the functions may not be dso-local.
+@vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 0, !"vfunc3.type"}
+!4 = !{i64 4, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
+; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+; CHECK-SAME:   i32 0
+; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
+
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -29,9 +43,19 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
+; (3) vfunc3_live_extern is referenced from @main, stays alive
+; CHECK: declare void @vfunc3_live_extern
+declare void @vfunc3_live_extern()
+
+; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
+; CHECK-NOT: declare void @vfunc4_dead_extern
+declare void @vfunc4_dead_extern()
+
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
+  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 

From 472ccf61a39387098ad3ed66c1353f82c7245ba3 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:04 +0000
Subject: [PATCH 458/886] [gn build] Port 61f1f1300254

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 4270bae57ff2c..2819eacaa8574 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -415,6 +415,7 @@ if (current_toolchain == default_toolchain) {
       "__coroutine/noop_coroutine_handle.h",
       "__coroutine/trivial_awaitables.h",
       "__debug_utils/randomize_range.h",
+      "__debug_utils/sanitizers.h",
       "__debug_utils/strict_weak_ordering_check.h",
       "__exception/exception.h",
       "__exception/exception_ptr.h",

From c71be83638917f089f8f154525cc6d6f61591f21 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:05 +0000
Subject: [PATCH 459/886] [gn build] Port a06073f91e7b

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 2819eacaa8574..20116606fa645 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -928,6 +928,7 @@ if (current_toolchain == default_toolchain) {
       "__utility/in_place.h",
       "__utility/integer_sequence.h",
       "__utility/is_pointer_in_range.h",
+      "__utility/is_valid_range.h",
       "__utility/move.h",
       "__utility/no_destroy.h",
       "__utility/pair.h",

From 8969bf3ea670dfeafbac53624883dc5c7ce23f0c Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:06 +0000
Subject: [PATCH 460/886] [gn build] Port a11a4324bb27

---
 llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn b/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
index 43f8b886f9b97..5b9fed07186c4 100644
--- a/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
@@ -31,6 +31,7 @@ unittest("ToolingTests") {
     "CommentHandlerTest.cpp",
     "CompilationDatabaseTest.cpp",
     "DependencyScannerTest.cpp",
+    "DependencyScanning/DependencyScanningFilesystemTest.cpp",
     "DiagnosticsYamlTest.cpp",
     "ExecutionTest.cpp",
     "FixItTest.cpp",

From 2f720a1c65e0b9669e67d7c319d77d6c221d423d Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:07 +0000
Subject: [PATCH 461/886] [gn build] Port a3ce29f7bb55

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 20116606fa645..4383f1d6d18ff 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -664,6 +664,7 @@ if (current_toolchain == default_toolchain) {
       "__numeric/transform_exclusive_scan.h",
       "__numeric/transform_inclusive_scan.h",
       "__numeric/transform_reduce.h",
+      "__pstl/cpu_algos/cpu_traits.h",
       "__random/bernoulli_distribution.h",
       "__random/binomial_distribution.h",
       "__random/cauchy_distribution.h",

From 2601ce41d6139761a4267b7cc4add4fea499d172 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:08 +0000
Subject: [PATCH 462/886] [gn build] Port c11976f525f3

---
 llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn b/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
index 5b9fed07186c4..258fb61038df5 100644
--- a/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/unittests/Tooling/BUILD.gn
@@ -30,7 +30,7 @@ unittest("ToolingTests") {
     "CastExprTest.cpp",
     "CommentHandlerTest.cpp",
     "CompilationDatabaseTest.cpp",
-    "DependencyScannerTest.cpp",
+    "DependencyScanning/DependencyScannerTest.cpp",
     "DependencyScanning/DependencyScanningFilesystemTest.cpp",
     "DiagnosticsYamlTest.cpp",
     "ExecutionTest.cpp",

From 831515147c551f2c2dc7289d4bc0be2215e6b445 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 17:59:09 +0000
Subject: [PATCH 463/886] [gn build] Port c3d58867bd3d

---
 llvm/utils/gn/secondary/llvm/lib/Target/AArch64/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/llvm/lib/Target/AArch64/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/Target/AArch64/BUILD.gn
index a61e6058053d8..879b7f0960a90 100644
--- a/llvm/utils/gn/secondary/llvm/lib/Target/AArch64/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/Target/AArch64/BUILD.gn
@@ -139,6 +139,7 @@ static_library("LLVMAArch64CodeGen") {
     "AArch64MacroFusion.cpp",
     "AArch64PBQPRegAlloc.cpp",
     "AArch64PointerAuth.cpp",
+    "AArch64PostCoalescerPass.cpp",
     "AArch64PromoteConstant.cpp",
     "AArch64RedundantCopyElimination.cpp",
     "AArch64RegisterInfo.cpp",

From b570eb5e99567c92636279d014f376938fe4e79a Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 18:05:13 +0000
Subject: [PATCH 464/886] [gn build] Manually port 44de2bb

---
 llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn | 1 +
 llvm/utils/gn/secondary/clang/test/BUILD.gn                 | 1 +
 2 files changed, 2 insertions(+)

diff --git a/llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn b/llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
index 7273803dd5165..a3fb952b1112b 100644
--- a/llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/include/clang/Config/BUILD.gn
@@ -27,6 +27,7 @@ write_cmake_config("Config") {
     "ENABLE_LINKER_BUILD_ID=",
     "ENABLE_X86_RELAX_RELOCATIONS=1",
     "CLANG_ENABLE_OBJC_REWRITER=1",  # FIXME: flag?
+    "CLANG_ENABLE_CIR=",
     "CLANG_SYSTEMZ_DEFAULT_ARCH=z10",
     "PPC_LINUX_DEFAULT_IEEELONGDOUBLE=",
   ]
diff --git a/llvm/utils/gn/secondary/clang/test/BUILD.gn b/llvm/utils/gn/secondary/clang/test/BUILD.gn
index bcf7c86ab9d08..4ed9352da9c98 100644
--- a/llvm/utils/gn/secondary/clang/test/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/test/BUILD.gn
@@ -48,6 +48,7 @@ write_lit_config("lit_site_cfg") {
     "CLANG_BUILT_STANDALONE=0",
     "CLANG_DEFAULT_CXX_STDLIB=",  # Empty string means "default value" here.
     "CLANG_DEFAULT_PIE_ON_LINUX=0",
+    "CLANG_ENABLE_CIR=0",
     "CLANG_PLUGIN_SUPPORT=0",  # FIXME: Analysis/plugins need global -fPIC
     "CURRENT_TOOLS_DIR=" + rebase_path("$root_out_dir/bin", dir),
     "CLANG_VENDOR_UTI=org.llvm.clang",

From fa5fdfb126f9beeaa2f719930ef951d339f41e36 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Mon, 15 Apr 2024 11:06:35 -0700
Subject: [PATCH 465/886] [CodeGen] Remove extraneous ArrayRef (NFC) (#88601)

We don't need to create these instances of ArrayRef because
StructType::create and ConstantStruct::get take ArrayRef, and ArrayRef
can be implicitly constructed from C arrays.
---
 llvm/lib/CodeGen/LowerEmuTLS.cpp | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/CodeGen/LowerEmuTLS.cpp b/llvm/lib/CodeGen/LowerEmuTLS.cpp
index af0b0a20c8566..ec36b669ac018 100644
--- a/llvm/lib/CodeGen/LowerEmuTLS.cpp
+++ b/llvm/lib/CodeGen/LowerEmuTLS.cpp
@@ -139,8 +139,7 @@ bool addEmuTlsVar(Module &M, const GlobalVariable *GV) {
   IntegerType *WordType = DL.getIntPtrType(C);
   PointerType *InitPtrType = PointerType::getUnqual(C);
   Type *ElementTypes[4] = {WordType, WordType, VoidPtrType, InitPtrType};
-  ArrayRef<Type*> ElementTypeArray(ElementTypes, 4);
-  StructType *EmuTlsVarType = StructType::create(ElementTypeArray);
+  StructType *EmuTlsVarType = StructType::create(ElementTypes);
   EmuTlsVar = cast<GlobalVariable>(
       M.getOrInsertGlobal(EmuTlsVarName, EmuTlsVarType));
   copyLinkageVisibility(M, GV, EmuTlsVar);
@@ -170,9 +169,7 @@ bool addEmuTlsVar(Module &M, const GlobalVariable *GV) {
       ConstantInt::get(WordType, DL.getTypeStoreSize(GVType)),
       ConstantInt::get(WordType, GVAlignment.value()), NullPtr,
       EmuTlsTmplVar ? EmuTlsTmplVar : NullPtr};
-  ArrayRef<Constant*> ElementValueArray(ElementValues, 4);
-  EmuTlsVar->setInitializer(
-      ConstantStruct::get(EmuTlsVarType, ElementValueArray));
+  EmuTlsVar->setInitializer(ConstantStruct::get(EmuTlsVarType, ElementValues));
   Align MaxAlignment =
       std::max(DL.getABITypeAlign(WordType), DL.getABITypeAlign(VoidPtrType));
   EmuTlsVar->setAlignment(MaxAlignment);

From e39c653dbab3b4fdc5118219c7b5179244d0e1b6 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 18:07:33 +0000
Subject: [PATCH 466/886] [gn build] Port a34887550b7f

---
 .../gn/secondary/lldb/source/Plugins/Language/CPlusPlus/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/lldb/source/Plugins/Language/CPlusPlus/BUILD.gn b/llvm/utils/gn/secondary/lldb/source/Plugins/Language/CPlusPlus/BUILD.gn
index 9d57f2f0877e3..91e5be0c8e27c 100644
--- a/llvm/utils/gn/secondary/lldb/source/Plugins/Language/CPlusPlus/BUILD.gn
+++ b/llvm/utils/gn/secondary/lldb/source/Plugins/Language/CPlusPlus/BUILD.gn
@@ -32,6 +32,7 @@ static_library("CPlusPlus") {
     "LibCxxInitializerList.cpp",
     "LibCxxList.cpp",
     "LibCxxMap.cpp",
+    "LibCxxProxyArray.cpp",
     "LibCxxQueue.cpp",
     "LibCxxRangesRefView.cpp",
     "LibCxxSliceArray.cpp",

From 1c5d5478f7f35d7f212fcbf2af042fc3c720e9ed Mon Sep 17 00:00:00 2001
From: Leonard Chan <leonardchan@google.com>
Date: Mon, 15 Apr 2024 11:11:20 -0700
Subject: [PATCH 467/886] Revert "Reapply "[llvm] Teach GlobalDCE about
 dso_local_equivalent""

This reverts commit b8b2f9902cb23af0811ec218a4f43226c0691510.

This was failing on a local build.
---
 .../include/llvm/Analysis/TypeMetadataUtils.h |  4 ++--
 ...virtual-functions-relative-pointers-bad.ll | 18 --------------
 ...virtual-functions-relative-pointers-gep.ll | 24 -------------------
 .../virtual-functions-relative-pointers.ll    | 24 -------------------
 4 files changed, 2 insertions(+), 68 deletions(-)

diff --git a/llvm/include/llvm/Analysis/TypeMetadataUtils.h b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
index bdb477b54b532..2c7ce0858b9d3 100644
--- a/llvm/include/llvm/Analysis/TypeMetadataUtils.h
+++ b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
@@ -87,8 +87,8 @@ std::pair<Function *, Constant *>
 getFunctionAtVTableOffset(GlobalVariable *GV, uint64_t Offset, Module &M);
 
 /// Finds the same "relative pointer" pattern as described above, where the
-/// target is `C`, and replaces the entire pattern with a constant zero.
-void replaceRelativePointerUsersWithZero(Constant *C);
+/// target is `F`, and replaces the entire pattern with a constant zero.
+void replaceRelativePointerUsersWithZero(Function *F);
 
 } // namespace llvm
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
index 9a4bff80383cb..9d87fdb084a12 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
@@ -16,32 +16,14 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 
 ; CHECK:      @vtable = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 8, !type !0, !type !1, !vcall_visibility !2
 
-@vtable2 = internal unnamed_addr constant { [3 x i32] } { [3 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
-
-  ; a "bad" relative pointer because it's base is not the @vtable symbol
-  i32 trunc (i64 sub (i64 ptrtoint (ptr @weird_ref_3         to i64), i64 ptrtoint (ptr @weird_ref_4 to i64)) to i32)
-]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
-!3 = !{i64 0, !"vfunc3.type"}
-!4 = !{i64 4, !"vfunc4.type"}
-
-; CHECK:      @vtable2 = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 4, !type !3, !type !4, !vcall_visibility !2
-
 define internal void @vfunc1() { ret void }
 define internal void @vfunc2() { ret void }
 define internal void @weird_ref_1() { ret void }
 define internal void @weird_ref_2() { ret void }
-declare void @vfunc3()
-declare void @vfunc4()
-declare void @weird_ref_3()
-declare void @weird_ref_4()
 
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   call void @weird_ref_2()
-  %2 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
-  call void @weird_ref_4()
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
index d89b024e7a6d0..fb45d37dde6e4 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
@@ -19,20 +19,6 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
-@vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
-  i32 42,
-  i32 1337,
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32)
-]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
-!3 = !{i64 8, !"vfunc3.type"}
-!4 = !{i64 12, !"vfunc4.type"}
-
-; CHECK:      @vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
-; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
-; CHECK-SAME:   i32 0
-; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
-
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -45,19 +31,9 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
-; (3) vfunc3_live_extern is referenced from @main, stays alive
-; CHECK: declare void @vfunc3_live_extern
-declare void @vfunc3_live_extern()
-
-; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
-; CHECK-NOT: declare void @vfunc4_dead_extern
-declare void @vfunc4_dead_extern()
-
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
-  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
-  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
index b652765296bc5..76a617ea99e6f 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
@@ -17,20 +17,6 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
-; Similar to above, but the vtable is more aligned to how C++ relative vtables look.
-; That is, the functions may not be dso-local.
-@vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32)
-]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
-!3 = !{i64 0, !"vfunc3.type"}
-!4 = !{i64 4, !"vfunc4.type"}
-
-; CHECK:      @vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
-; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
-; CHECK-SAME:   i32 0
-; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
-
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -43,19 +29,9 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
-; (3) vfunc3_live_extern is referenced from @main, stays alive
-; CHECK: declare void @vfunc3_live_extern
-declare void @vfunc3_live_extern()
-
-; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
-; CHECK-NOT: declare void @vfunc4_dead_extern
-declare void @vfunc4_dead_extern()
-
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
-  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
-  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 

From c0b77e0a4a9bb090e5ad3a28adcd0aa98715cfe5 Mon Sep 17 00:00:00 2001
From: Leonard Chan <leonardchan@google.com>
Date: Mon, 15 Apr 2024 11:14:50 -0700
Subject: [PATCH 468/886] Revert "Reapply "[llvm] Teach whole program
 devirtualization about relative vtables""

This reverts commit 09c3bfe9b3eb47a2af0c10531b25f90cfb5fa9f4.
---
 .../include/llvm/Analysis/TypeMetadataUtils.h |   2 +-
 llvm/lib/Analysis/ModuleSummaryAnalysis.cpp   |  37 +----
 llvm/lib/Analysis/TypeMetadataUtils.cpp       |  16 +-
 llvm/test/ThinLTO/X86/devirt.ll               |  95 +----------
 .../WholeProgramDevirt/Inputs/export.yaml     |   8 -
 .../WholeProgramDevirt/branch-funnel.ll       | 157 ------------------
 .../WholeProgramDevirt/constant-arg.ll        |  45 -----
 .../devirt-single-impl-check.ll               |  25 ---
 .../devirt-single-impl-multiple-assumes.ll    |  20 ---
 .../WholeProgramDevirt/devirt-single-impl.ll  |  47 +-----
 .../WholeProgramDevirt/pointer-vtable.ll      |  16 --
 11 files changed, 10 insertions(+), 458 deletions(-)

diff --git a/llvm/include/llvm/Analysis/TypeMetadataUtils.h b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
index 2c7ce0858b9d3..8894945c28d94 100644
--- a/llvm/include/llvm/Analysis/TypeMetadataUtils.h
+++ b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
@@ -66,7 +66,7 @@ void findDevirtualizableCallsForTypeCheckedLoad(
 /// Used for example from GlobalDCE to find an entry in a C++ vtable that
 /// matches a vcall offset.
 ///
-/// To support relative vtables, getPointerAtOffset can see through "relative
+/// To support Swift vtables, getPointerAtOffset can see through "relative
 /// pointers", i.e. (sub-)expressions of the form of:
 ///
 /// @symbol = ... {
diff --git a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
index c3d15afe6d9cb..deda1eebb3b57 100644
--- a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
+++ b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
@@ -22,7 +22,6 @@
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Analysis/BlockFrequencyInfo.h"
 #include "llvm/Analysis/BranchProbabilityInfo.h"
-#include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/IndirectCallPromotionAnalysis.h"
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/MemoryProfileInfo.h"
@@ -669,8 +668,7 @@ static void computeFunctionSummary(
 /// within the initializer.
 static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
                              const Module &M, ModuleSummaryIndex &Index,
-                             VTableFuncList &VTableFuncs,
-                             const GlobalVariable &OrigGV) {
+                             VTableFuncList &VTableFuncs) {
   // First check if this is a function pointer.
   if (I->getType()->isPointerTy()) {
     auto C = I->stripPointerCasts();
@@ -698,7 +696,7 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
       auto Offset = SL->getElementOffset(EI.index());
       unsigned Op = SL->getElementContainingOffset(Offset);
       findFuncPointers(cast<Constant>(I->getOperand(Op)),
-                       StartingOffset + Offset, M, Index, VTableFuncs, OrigGV);
+                       StartingOffset + Offset, M, Index, VTableFuncs);
     }
   } else if (auto *C = dyn_cast<ConstantArray>(I)) {
     ArrayType *ATy = C->getType();
@@ -706,34 +704,7 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
     uint64_t EltSize = DL.getTypeAllocSize(EltTy);
     for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i) {
       findFuncPointers(cast<Constant>(I->getOperand(i)),
-                       StartingOffset + i * EltSize, M, Index, VTableFuncs,
-                       OrigGV);
-    }
-  } else if (const auto *CE = dyn_cast<ConstantExpr>(I)) {
-    // For relative vtables, the next sub-component should be a trunc.
-    if (CE->getOpcode() != Instruction::Trunc ||
-        !(CE = dyn_cast<ConstantExpr>(CE->getOperand(0))))
-      return;
-
-    // If this constant can be reduced to the offset between a function and a
-    // global, then we know this is a valid virtual function if the RHS is the
-    // original vtable we're scanning through.
-    if (CE->getOpcode() == Instruction::Sub) {
-      GlobalValue *LHS, *RHS;
-      APSInt LHSOffset, RHSOffset;
-      if (IsConstantOffsetFromGlobal(CE->getOperand(0), LHS, LHSOffset, DL) &&
-          IsConstantOffsetFromGlobal(CE->getOperand(1), RHS, RHSOffset, DL) &&
-          RHS == &OrigGV &&
-
-          // For relative vtables, this component should point to the callable
-          // function without any offsets.
-          LHSOffset == 0 &&
-
-          // Also, the RHS should always point to somewhere within the vtable.
-          RHSOffset <=
-              static_cast<uint64_t>(DL.getTypeAllocSize(OrigGV.getInitializer()->getType()))) {
-        findFuncPointers(LHS, StartingOffset, M, Index, VTableFuncs, OrigGV);
-      }
+                       StartingOffset + i * EltSize, M, Index, VTableFuncs);
     }
   }
 }
@@ -746,7 +717,7 @@ static void computeVTableFuncs(ModuleSummaryIndex &Index,
     return;
 
   findFuncPointers(V.getInitializer(), /*StartingOffset=*/0, M, Index,
-                   VTableFuncs, V);
+                   VTableFuncs);
 
 #ifndef NDEBUG
   // Validate that the VTableFuncs list is ordered by offset.
diff --git a/llvm/lib/Analysis/TypeMetadataUtils.cpp b/llvm/lib/Analysis/TypeMetadataUtils.cpp
index ee400175ef11c..b8dcc39e9223c 100644
--- a/llvm/lib/Analysis/TypeMetadataUtils.cpp
+++ b/llvm/lib/Analysis/TypeMetadataUtils.cpp
@@ -67,14 +67,6 @@ static void findLoadCallsAtConstantOffset(
         findLoadCallsAtConstantOffset(M, DevirtCalls, User, Offset + GEPOffset,
                                       CI, DT);
       }
-    } else if (auto *Call = dyn_cast<CallInst>(User)) {
-      if (Call->getIntrinsicID() == llvm::Intrinsic::load_relative) {
-        if (auto *LoadOffset = dyn_cast<ConstantInt>(Call->getOperand(1))) {
-          findCallsAtConstantOffset(DevirtCalls, nullptr, User,
-                                    Offset + LoadOffset->getSExtValue(), CI,
-                                    DT);
-        }
-      }
     }
   }
 }
@@ -139,12 +131,6 @@ void llvm::findDevirtualizableCallsForTypeCheckedLoad(
 
 Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                                    Constant *TopLevelGlobal) {
-  // TODO: Ideally it would be the caller who knows if it's appropriate to strip
-  // the DSOLocalEquicalent. More generally, it would feel more appropriate to
-  // have two functions that handle absolute and relative pointers separately.
-  if (auto *Equiv = dyn_cast<DSOLocalEquivalent>(I))
-    I = Equiv->getGlobalValue();
-
   if (I->getType()->isPointerTy()) {
     if (Offset == 0)
       return I;
@@ -175,7 +161,7 @@ Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                               Offset % ElemSize, M, TopLevelGlobal);
   }
 
-  // Relative-pointer support starts here.
+  // (Swift-specific) relative-pointer support starts here.
   if (auto *CI = dyn_cast<ConstantInt>(I)) {
     if (Offset == 0 && CI->isZero()) {
       return I;
diff --git a/llvm/test/ThinLTO/X86/devirt.ll b/llvm/test/ThinLTO/X86/devirt.ll
index a40f89e644f00..472e43d776803 100644
--- a/llvm/test/ThinLTO/X86/devirt.ll
+++ b/llvm/test/ThinLTO/X86/devirt.ll
@@ -27,36 +27,24 @@
 ; NOENABLESPLITFLAG-DAG: [[B:\^[0-9]+]] = gv: (name: "_ZTV1B", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[Bf]], [[An]])
 ; NOENABLESPLITFLAG-DAG: [[C:\^[0-9]+]] = gv: (name: "_ZTV1C", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[An]], [[Cf]])
 ; NOENABLESPLITFLAG-DAG: [[D:\^[0-9]+]] = gv: (name: "_ZTV1D", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 16)), refs: ([[Dm]])
-; NOENABLESPLITFLAG-DAG: [[B_RV:\^[0-9]+]] = gv: (name: "_ZTV1B_RV", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[B_RV]], [[Bf]], [[An]])
-; NOENABLESPLITFLAG-DAG: [[C_RV:\^[0-9]+]] = gv: (name: "_ZTV1C_RV", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[C_RV]], [[An]], [[Cf]])
-; NOENABLESPLITFLAG-DAG: [[D_RV:\^[0-9]+]] = gv: (name: "_ZTV1D_RV", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 8)), refs: ([[D_RV]], [[Dm]])
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A", summary: ((offset: 16, [[B]]), (offset: 16, [[C]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B", summary: ((offset: 16, [[B]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C", summary: ((offset: 16, [[C]])))
-; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A_RV", summary: ((offset: 8, [[B_RV]]), (offset: 8, [[C_RV]])))
-; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B_RV", summary: ((offset: 8, [[B_RV]])))
-; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C_RV", summary: ((offset: 8, [[C_RV]])))
 ; Type Id on _ZTV1D should have been promoted
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "1.{{.*}}", summary: ((offset: 16, [[D]])))
-; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "2.{{.*}}", summary: ((offset: 8, [[D_RV]])))
 
 ; Index based WPD
 ; RUN: llvm-lto2 run %t2.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
-; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px \
-; RUN:   -r=%t2.o,_ZTV1B_RV,px \
-; RUN:   -r=%t2.o,_ZTV1C_RV,px \
-; RUN:   -r=%t2.o,_ZTV1D_RV,px \
-; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK
+; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
 ; Check that we're able to prevent specific function from being
@@ -66,24 +54,18 @@
 ; RUN:   -wholeprogramdevirt-skip=_ZN1A1nEi \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
-; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px \
-; RUN:   -r=%t2.o,_ZTV1B_RV,px \
-; RUN:   -r=%t2.o,_ZTV1C_RV,px \
-; RUN:   -r=%t2.o,_ZTV1D_RV,px \
-; RUN:   2>&1 | FileCheck %s --check-prefix=SKIP
+; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=SKIP
 
 ; RUN: llvm-lto2 run %t.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t.o,test,px \
-; RUN:   -r=%t.o,test_rv,px \
 ; RUN:   -r=%t.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t.o,_ZN1C1fEi,p \
@@ -91,24 +73,15 @@
 ; RUN:   -r=%t.o,_ZTV1B, \
 ; RUN:   -r=%t.o,_ZTV1C, \
 ; RUN:   -r=%t.o,_ZTV1D, \
-; RUN:   -r=%t.o,_ZTV1B_RV, \
-; RUN:   -r=%t.o,_ZTV1C_RV, \
-; RUN:   -r=%t.o,_ZTV1D_RV, \
 ; RUN:   -r=%t.o,_ZN1A1nEi, \
 ; RUN:   -r=%t.o,_ZN1B1fEi, \
 ; RUN:   -r=%t.o,_ZN1C1fEi, \
 ; RUN:   -r=%t.o,_ZN1D1mEi, \
 ; RUN:   -r=%t.o,_ZTV1B,px \
 ; RUN:   -r=%t.o,_ZTV1C,px \
-; RUN:   -r=%t.o,_ZTV1D,px \
-; RUN:   -r=%t.o,_ZTV1B_RV,px \
-; RUN:   -r=%t.o,_ZTV1C_RV,px \
-; RUN:   -r=%t.o,_ZTV1D_RV,px \
-; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
+; RUN:   -r=%t.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
-; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
-; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 
@@ -126,25 +99,6 @@ target triple = "x86_64-grtev4-linux-gnu"
 @_ZTV1C = constant { [4 x ptr] } { [4 x ptr] [ptr null, ptr undef, ptr @_ZN1C1fEi, ptr @_ZN1A1nEi] }, !type !0, !type !2
 @_ZTV1D = constant { [3 x ptr] } { [3 x ptr] [ptr null, ptr undef, ptr @_ZN1D1mEi] }, !type !3
 
-@_ZTV1B_RV = constant { [4 x i32] } { [4 x i32] [
-  i32 0,
-  i32 undef,
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1B1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 2) to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 3) to i64)) to i32)
-] }, !type !7, !type !8
-
-@_ZTV1C_RV = constant { [4 x i32] } { [4 x i32] [
-  i32 0,
-  i32 undef,
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1C1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 2) to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 3) to i64)) to i32)
-] }, !type !7, !type !9
-
-@_ZTV1D_RV = constant { [3 x i32] } { [3 x i32] [
-  i32 0,
-  i32 undef,
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1D1mEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1D_RV, i32 0, i32 0, i32 2) to i64)) to i32)
-] }, !type !10
 
 ; CHECK-IR-LABEL: define {{(noundef )?}}i32 @test
 define i32 @test(ptr %obj, ptr %obj2, i32 %a) {
@@ -182,43 +136,6 @@ entry:
 ; CHECK-IR-LABEL: ret i32
 ; CHECK-IR-LABEL: }
 
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-; CHECK-IR-LABEL: define i32 @test_rv
-define i32 @test_rv(ptr %obj, ptr %obj2, i32 %a) {
-entry:
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"_ZTS1A_RV")
-  call void @llvm.assume(i1 %p)
-  %fptr1_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 4)
-
-  ; Check that the call was devirtualized.
-  ; CHECK-IR: %call = tail call i32 @_ZN1A1nEi
-  ; Ensure !prof and !callees metadata for indirect call promotion removed.
-  ; CHECK-IR-NOT: prof
-  ; CHECK-IR-NOT: callees
-  %call = tail call i32 %fptr1_rv(ptr nonnull %obj, i32 %a), !prof !5, !callees !6
-
-  %fptr22_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-
-  ; We still have to call it as virtual.
-  ; CHECK-IR: %call3 = tail call i32 %fptr22
-  %call3 = tail call i32 %fptr22_rv(ptr nonnull %obj, i32 %call)
-
-  %vtable2 = load ptr, ptr %obj2
-  %p2 = call i1 @llvm.type.test(ptr %vtable2, metadata !11)
-  call void @llvm.assume(i1 %p2)
-
-  %fptr33_rv = call ptr @llvm.load.relative.i32(ptr %vtable2, i32 0)
-
-  ; Check that the call was devirtualized.
-  ; CHECK-IR: %call4 = tail call i32 @_ZN1D1mEi
-  %call4 = tail call i32 %fptr33_rv(ptr nonnull %obj2, i32 %call3)
-  ret i32 %call4
-}
-; CHECK-IR-LABEL: ret i32
-; CHECK-IR-LABEL: }
-
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -248,9 +165,3 @@ attributes #0 = { noinline optnone }
 !4 = distinct !{}
 !5 = !{!"VP", i32 0, i64 1, i64 1621563287929432257, i64 1}
 !6 = !{ptr @_ZN1A1nEi}
-
-!7 = !{i64 8, !"_ZTS1A_RV"}
-!8 = !{i64 8, !"_ZTS1B_RV"}
-!9 = !{i64 8, !"_ZTS1C_RV"}
-!10 = !{i64 8, !11}
-!11 = distinct !{}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
index dd0c90d7ae8c3..71cf38b216c71 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
+++ b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
@@ -5,22 +5,14 @@ GlobalValueMap:
       TypeTestAssumeVCalls:
         - GUID: 14276520915468743435  # typeid1
           Offset: 0
-        - GUID: 271751036925422857  # typeid1_rv
-          Offset: 0
       TypeCheckedLoadVCalls:
         - GUID: 15427464259790519041  # typeid2
           Offset: 0
-        - GUID: 1146149264729288256  # typeid2_rv
-          Offset: 0
       TypeTestAssumeConstVCalls:
         - VFunc:
             GUID: 3515965990081467659  # typeid3
             Offset: 0
           Args: [12, 24]
-        - VFunc:
-            GUID: 2777626534618191571  # typeid3_rv
-            Offset: 0
-          Args: [12, 24]
       TypeCheckedLoadConstVCalls:
         - VFunc:
             GUID: 17525413373118030901  # typeid4
diff --git a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
index 0b1023eee2732..b55713fe4b296 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
@@ -8,45 +8,6 @@
 ; RUN: FileCheck --check-prefix=SUMMARY %s < %t
 
 ; SUMMARY:      TypeIdMap:       
-; SUMMARY-NEXT:  typeid1_rv:
-; SUMMARY-NEXT:    TTRes:
-; SUMMARY-NEXT:      Kind:            Unknown
-; SUMMARY-NEXT:      SizeM1BitWidth:  0
-; SUMMARY-NEXT:      AlignLog2:       0
-; SUMMARY-NEXT:      SizeM1:          0
-; SUMMARY-NEXT:      BitMask:         0
-; SUMMARY-NEXT:      InlineBits:      0
-; SUMMARY-NEXT:    WPDRes:
-; SUMMARY-NEXT:      0:
-; SUMMARY-NEXT:        Kind:            BranchFunnel
-; SUMMARY-NEXT:        SingleImplName:  ''
-; SUMMARY-NEXT:        ResByArg:
-; SUMMARY-NEXT:  typeid2_rv:
-; SUMMARY-NEXT:    TTRes:
-; SUMMARY-NEXT:      Kind:            Unknown
-; SUMMARY-NEXT:      SizeM1BitWidth:  0
-; SUMMARY-NEXT:      AlignLog2:       0
-; SUMMARY-NEXT:      SizeM1:          0
-; SUMMARY-NEXT:      BitMask:         0
-; SUMMARY-NEXT:      InlineBits:      0
-; SUMMARY-NEXT:    WPDRes:
-; SUMMARY-NEXT:      0:
-; SUMMARY-NEXT:        Kind:            Indir
-; SUMMARY-NEXT:        SingleImplName:  ''
-; SUMMARY-NEXT:        ResByArg:
-; SUMMARY-NEXT:   typeid3_rv:
-; SUMMARY-NEXT:     TTRes:           
-; SUMMARY-NEXT:       Kind:            Unknown
-; SUMMARY-NEXT:       SizeM1BitWidth:  0
-; SUMMARY-NEXT:       AlignLog2:       0
-; SUMMARY-NEXT:       SizeM1:          0
-; SUMMARY-NEXT:       BitMask:         0
-; SUMMARY-NEXT:       InlineBits:      0
-; SUMMARY-NEXT:     WPDRes:          
-; SUMMARY-NEXT:       0:               
-; SUMMARY-NEXT:         Kind:            BranchFunnel
-; SUMMARY-NEXT:         SingleImplName:  ''
-; SUMMARY-NEXT:         ResByArg:        
 ; SUMMARY-NEXT:   typeid3:
 ; SUMMARY-NEXT:     TTRes:           
 ; SUMMARY-NEXT:       Kind:            Unknown
@@ -132,29 +93,6 @@ declare i32 @vf3_2(ptr %this, i32 %arg)
 declare i32 @vf4_1(ptr %this, i32 %arg)
 declare i32 @vf4_2(ptr %this, i32 %arg)
 
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-;; These are relative vtables equivalent to the ones above.
-@vt1_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_1 to i64), i64 ptrtoint (ptr @vt1_1_rv to i64)) to i32)], !type !5
-@vt1_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_2 to i64), i64 ptrtoint (ptr @vt1_2_rv to i64)) to i32)], !type !5
-
-@vt2_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_1 to i64), i64 ptrtoint (ptr @vt2_1_rv to i64)) to i32)], !type !6
-@vt2_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_2 to i64), i64 ptrtoint (ptr @vt2_2_rv to i64)) to i32)], !type !6
-@vt2_3_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_3 to i64), i64 ptrtoint (ptr @vt2_3_rv to i64)) to i32)], !type !6
-@vt2_4_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_4 to i64), i64 ptrtoint (ptr @vt2_4_rv to i64)) to i32)], !type !6
-@vt2_5_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_5 to i64), i64 ptrtoint (ptr @vt2_5_rv to i64)) to i32)], !type !6
-@vt2_6_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_6 to i64), i64 ptrtoint (ptr @vt2_6_rv to i64)) to i32)], !type !6
-@vt2_7_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_7 to i64), i64 ptrtoint (ptr @vt2_7_rv to i64)) to i32)], !type !6
-@vt2_8_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_8 to i64), i64 ptrtoint (ptr @vt2_8_rv to i64)) to i32)], !type !6
-@vt2_9_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_9 to i64), i64 ptrtoint (ptr @vt2_9_rv to i64)) to i32)], !type !6
-@vt2_10_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_10 to i64), i64 ptrtoint (ptr @vt2_10_rv to i64)) to i32)], !type !6
-@vt2_11_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_11 to i64), i64 ptrtoint (ptr @vt2_11_rv to i64)) to i32)], !type !6
-
-@vt3_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_1 to i64), i64 ptrtoint (ptr @vt3_1_rv to i64)) to i32)], !type !7
-@vt3_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_2 to i64), i64 ptrtoint (ptr @vt3_2_rv to i64)) to i32)], !type !7
-
-@vt4_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_1 to i64), i64 ptrtoint (ptr @vt4_1_rv to i64)) to i32)], !type !8
-@vt4_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_2 to i64), i64 ptrtoint (ptr @vt4_2_rv to i64)) to i32)], !type !8
 
 
 ; CHECK-LABEL: define i32 @fn1
@@ -170,19 +108,6 @@ define i32 @fn1(ptr %obj) #0 {
   ret i32 %result
 }
 
-; CHECK-LABEL: define i32 @fn1_rv
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn1_rv(ptr %obj) #0 {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid1_rv")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest %vtable, ptr %obj, i32 1)
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ; NORETP: call i32 %
-  ret i32 %result
-}
-
 ; CHECK-LABEL: define i32 @fn2
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn2(ptr %obj) #0 {
@@ -195,18 +120,6 @@ define i32 @fn2(ptr %obj) #0 {
   ret i32 %result
 }
 
-; CHECK-LABEL: define i32 @fn2_rv
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn2_rv(ptr %obj) #0 {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2_rv")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: call i32 %
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ret i32 %result
-}
-
 ; CHECK-LABEL: define i32 @fn3
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn3(ptr %obj) #0 {
@@ -220,75 +133,10 @@ define i32 @fn3(ptr %obj) #0 {
   ret i32 %result
 }
 
-; CHECK-LABEL: define i32 @fn3_rv
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn3_rv(ptr %obj) #0 {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !9)
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; RETP: call i32 @branch_funnel.1(ptr
-  ; NORETP: call i32 %
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ret i32 %result
-}
-
-; CHECK-LABEL: define i32 @fn4
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn4(ptr %obj) #0 {
-  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
-  call void @llvm.assume(i1 %p)
-  %fptr = load ptr, ptr @vt1_1
-  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ; NORETP: call i32 %
-  ret i32 %result
-}
-
-; CHECK-LABEL: define i32 @fn4_cpy
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn4_cpy(ptr %obj) #0 {
-  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
-  call void @llvm.assume(i1 %p)
-  %fptr = load ptr, ptr @vt1_1
-  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ; NORETP: call i32 %
-  ret i32 %result
-}
-
-; CHECK-LABEL: define i32 @fn4_rv
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn4_rv(ptr %obj) #0 {
-  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
-  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ; NORETP: call i32 %
-  ret i32 %result
-}
-
-; CHECK-LABEL: define i32 @fn4_rv_cpy
-; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
-define i32 @fn4_rv_cpy(ptr %obj) #0 {
-  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
-  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
-  %result = call i32 %fptr(ptr %obj, i32 1)
-  ; NORETP: call i32 %
-  ret i32 %result
-}
-
 ; CHECK-LABEL: define hidden void @__typeid_typeid1_0_branch_funnel(ptr nest %0, ...)
 ; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2, ptr {{(nonnull )?}}@vf1_2, ...)
 
-; CHECK-LABEL: define hidden void @__typeid_typeid1_rv_0_branch_funnel(ptr nest %0, ...)
-; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1_rv, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2_rv, ptr {{(nonnull )?}}@vf1_2, ...)
-
 ; CHECK: define internal void @branch_funnel(ptr
-; CHECK: define internal void @branch_funnel.1(ptr
 
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
@@ -298,10 +146,5 @@ declare void @llvm.assume(i1)
 !2 = !{i32 0, !"typeid3"}
 !3 = !{i32 0, !4}
 !4 = distinct !{}
-!5 = !{i32 0, !"typeid1_rv"}
-!6 = !{i32 0, !"typeid2_rv"}
-!7 = !{i32 0, !"typeid3_rv"}
-!8 = !{i32 0, !9}
-!9 = distinct !{}
 
 attributes #0 = { "target-features"="+retpoline" }
diff --git a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
index e685a1a54c520..91bae4af9a211 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
@@ -7,10 +7,6 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf2], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\01", [1 x ptr] [ptr @vf4], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf8], [0 x i8] zeroinitializer }, !type [[T8]]
-; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4:![0-9]+]]
-; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
-; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
-; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
 
 @vt1 = constant [1 x ptr] [ptr @vf1], !type !0
 @vt2 = constant [1 x ptr] [ptr @vf2], !type !0
@@ -65,49 +61,8 @@ define i1 @call2(ptr %obj) {
   ret i1 %result
 }
 
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-@vt1_rv = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)
-], align 4, !type !1
-@vt2_rv = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)
-], align 4, !type !1
-@vt4_rv = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)
-], align 4, !type !1
-@vt8_rv = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)
-], align 4, !type !1
-
-; CHECK: define i1 @call3
-define i1 @call3(ptr %obj) {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: getelementptr {{.*}} -1
-  ; CHECK: and {{.*}}, 1
-  %result = call i1 %fptr(ptr %obj, i32 5)
-  ret i1 %result
-}
-
-; CHECK: define i1 @call4
-define i1 @call4(ptr %obj) {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: getelementptr {{.*}} -1
-  ; CHECK: and {{.*}}, 2
-  %result = call i1 %fptr(ptr %obj, i32 10)
-  ret i1 %result
-}
-
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 ; CHECK: [[T8]] = !{i32 8, !"typeid"}
-; CHECK: [[T4]] = !{i32 4, !"typeid2"}
 !0 = !{i32 0, !"typeid"}
-!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
index 5a9e6c12c3582..dc7b2024bc591 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
@@ -3,7 +3,6 @@
 target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
-; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: devirtualized vf
 ; CHECK-NOT: devirtualized
@@ -34,31 +33,7 @@ trap:
   unreachable
 }
 
-@vt3 = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
-], align 4, !type !1
-
-; CHECK: define void @call2
-define void @call2(ptr %obj) {
-  %vtable = load ptr, ptr %obj
-  %pair = call {ptr, i1} @llvm.type.checked.load(ptr %vtable, i32 0, metadata !"typeid2")
-  %fptr = extractvalue {ptr, i1} %pair, 0
-  %p = extractvalue {ptr, i1} %pair, 1
-  ; CHECK: br i1 true,
-  br i1 %p, label %cont, label %trap
-
-cont:
-  ; CHECK: call void @vf(
-  call void %fptr(ptr %obj)
-  ret void
-
-trap:
-  call void @llvm.trap()
-  unreachable
-}
-
 declare {ptr, i1} @llvm.type.checked.load(ptr, i32, metadata)
 declare void @llvm.trap()
 
 !0 = !{i32 0, !"typeid"}
-!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
index bfbbeaf8a0656..ed144c23302dd 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
@@ -23,27 +23,7 @@ define void @call(ptr %obj) {
   ret void
 }
 
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-@vt3 = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
-], align 4, !type !1
-
-; CHECK: define void @call2
-define void @call2(ptr %obj) {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p)
-  %p2 = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p2)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: call void @vf(
-  call void %fptr(ptr %obj)
-  ret void
-}
-
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
-!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
index d7f33df594956..1ce96f70d905f 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
@@ -7,8 +7,6 @@ target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
 ; CHECK: remark: devirt-single.cc:30:32: single-impl: devirtualized a call to vf
-; CHECK: remark: devirt-single.cc:41:32: single-impl: devirtualized a call to vf
-; CHECK: remark: devirt-single.cc:51:32: single-impl: devirtualized a call to vf
 ; CHECK: remark: devirt-single.cc:13:0: devirtualized vf
 ; CHECK-NOT: devirtualized
 
@@ -30,41 +28,6 @@ define void @call(ptr %obj) #1 !dbg !5 {
   ret void
 }
 
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-@vt3 = private unnamed_addr constant [1 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
-], align 4, !type !11
-
-; CHECK: define void @call2
-define void @call2(ptr %obj) #1 !dbg !9 {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: call void @vf(
-  call void %fptr(ptr %obj), !dbg !10
-  ret void
-}
-
-@_ZTV1A.local = private unnamed_addr constant { [3 x i32] } { [3 x i32] [
-  i32 0,  ; offset to top
-  i32 0,  ; rtti
-  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1A.local, i32 0, i32 0, i32 2) to i64)) to i32)  ; vfunc offset
-] }, align 4, !type !14
-
-; CHECK: define void @call3
-define void @call3(ptr %obj) #1 !dbg !12 {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid3")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 8)
-  ; CHECK: call void @vf(
-  call void %fptr(ptr %obj), !dbg !13
-  ret void
-}
-
-
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -82,13 +45,5 @@ declare void @llvm.assume(i1)
 !7 = distinct !DISubprogram(name: "vf", linkageName: "_ZN3vt12vfEv", scope: !1, file: !1, line: 13, isLocal: false, isDefinition: true, scopeLine: 13, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
 !8 = !{i32 0, !"typeid"}
 
-!9 = distinct !DISubprogram(name: "call2", linkageName: "_Z5call2Pv", scope: !1, file: !1, line: 40, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
-!10 = !DILocation(line: 41, column: 32, scope: !9)
-!11 = !{i32 0, !"typeid2"}
-
-!12 = distinct !DISubprogram(name: "call3", linkageName: "_Z5call3Pv", scope: !1, file: !1, line: 50, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
-!13 = !DILocation(line: 51, column: 32, scope: !12)
-!14 = !{i32 0, !"typeid3"}
-
 ; CHECK: 1 wholeprogramdevirt - Number of whole program devirtualization targets
-; CHECK: 3 wholeprogramdevirt - Number of single implementation devirtualizations
+; CHECK: 1 wholeprogramdevirt - Number of single implementation devirtualizations
diff --git a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
index 062cef9b88fe4..3c5d9b839bc4d 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
@@ -20,23 +20,7 @@ define void @call(ptr %obj) {
   ret void
 }
 
-@vt2 = constant i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt2 to i64)) to i32), !type !1
-
-declare ptr @llvm.load.relative.i32(ptr, i32)
-
-; CHECK: define void @call2
-define void @call2(ptr %obj) {
-  %vtable = load ptr, ptr %obj
-  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
-  call void @llvm.assume(i1 %p)
-  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
-  ; CHECK: call void @vf(
-  call void %fptr(ptr %obj)
-  ret void
-}
-
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
-!1 = !{i32 0, !"typeid2"}

From b7b183371b54a2c4b5d2a39c594d3967a7384cb8 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Mon, 15 Apr 2024 10:55:41 -0700
Subject: [PATCH 469/886] [SLP][NFC]Improve perf of BoUpSLP::buildExternalUses
 function, NFC.

Perform required operations, only when the result is required + check if
value was not inserted already into the list of the externally used
scalars and early exit, if it was.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 20 ++++++++++++++++---
 1 file changed, 17 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 9a0eaca2f5980..c63b500f546f3 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -5514,6 +5514,7 @@ void BoUpSLP::reorderBottomToTop(bool IgnoreReorder) {
 
 void BoUpSLP::buildExternalUses(
     const ExtraValueToDebugLocsMap &ExternallyUsedValues) {
+  DenseMap<Value *, unsigned> ScalarToExtUses;
   // Collect the values that we need to extract from the tree.
   for (auto &TEPtr : VectorizableTree) {
     TreeEntry *Entry = TEPtr.get();
@@ -5527,13 +5528,18 @@ void BoUpSLP::buildExternalUses(
       Value *Scalar = Entry->Scalars[Lane];
       if (!isa<Instruction>(Scalar))
         continue;
-      int FoundLane = Entry->findLaneForValue(Scalar);
+      // All uses must be replaced already? No need to do it again.
+      auto It = ScalarToExtUses.find(Scalar);
+      if (It != ScalarToExtUses.end() && !ExternalUses[It->second].User)
+        continue;
 
       // Check if the scalar is externally used as an extra arg.
       const auto *ExtI = ExternallyUsedValues.find(Scalar);
       if (ExtI != ExternallyUsedValues.end()) {
+        int FoundLane = Entry->findLaneForValue(Scalar);
         LLVM_DEBUG(dbgs() << "SLP: Need to extract: Extra arg from lane "
-                          << Lane << " from " << *Scalar << ".\n");
+                          << FoundLane << " from " << *Scalar << ".\n");
+        ScalarToExtUses.try_emplace(Scalar, ExternalUses.size());
         ExternalUses.emplace_back(Scalar, nullptr, FoundLane);
       }
       for (User *U : Scalar->users()) {
@@ -5561,12 +5567,20 @@ void BoUpSLP::buildExternalUses(
             continue;
           }
           U = nullptr;
+          if (It != ScalarToExtUses.end()) {
+            ExternalUses[It->second].User = nullptr;
+            break;
+          }
         }
 
+        int FoundLane = Entry->findLaneForValue(Scalar);
         LLVM_DEBUG(dbgs() << "SLP: Need to extract:" << *UserInst
-                          << " from lane " << Lane << " from " << *Scalar
+                          << " from lane " << FoundLane << " from " << *Scalar
                           << ".\n");
+        It = ScalarToExtUses.try_emplace(Scalar, ExternalUses.size()).first;
         ExternalUses.emplace_back(Scalar, U, FoundLane);
+        if (!U)
+          break;
       }
     }
   }

From 3eb097339e5fb78e91182674e891445ebe5a845b Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Mon, 15 Apr 2024 14:33:53 -0400
Subject: [PATCH 470/886] [lld/mac] Fix (adorable!) typo to cycle bots

---
 lld/MachO/ObjC.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/lld/MachO/ObjC.cpp b/lld/MachO/ObjC.cpp
index 5902b82d30f55..c28f2b42a72ce 100644
--- a/lld/MachO/ObjC.cpp
+++ b/lld/MachO/ObjC.cpp
@@ -458,7 +458,7 @@ class ObjcCategoryMerger {
                                const PointerListInfo &ptrList);
 
   Defined *emitCategory(const ClassExtensionInfo &extInfo);
-  Defined *emitCatListEntrySec(const std::string &forCateogryName,
+  Defined *emitCatListEntrySec(const std::string &forCategoryName,
                                const std::string &forBaseClassName,
                                ObjFile *objFile);
   Defined *emitCategoryBody(const std::string &name, const Defined *nameSym,
@@ -878,7 +878,7 @@ void ObjcCategoryMerger::emitAndLinkPointerList(
 
 // This method creates an __objc_catlist ConcatInputSection with a single slot
 Defined *
-ObjcCategoryMerger::emitCatListEntrySec(const std::string &forCateogryName,
+ObjcCategoryMerger::emitCatListEntrySec(const std::string &forCategoryName,
                                         const std::string &forBaseClassName,
                                         ObjFile *objFile) {
   uint32_t sectionSize = target->wordSize;
@@ -894,7 +894,7 @@ ObjcCategoryMerger::emitCatListEntrySec(const std::string &forCateogryName,
   newCatList->parent = infoCategoryWriter.catListInfo.outputSection;
 
   std::string catSymName = "<__objc_catlist slot for merged category ";
-  catSymName += forBaseClassName + "(" + forCateogryName + ")>";
+  catSymName += forBaseClassName + "(" + forCategoryName + ")>";
 
   Defined *catListSym = make<Defined>(
       newStringData(catSymName.c_str()), /*file=*/objFile, newCatList,
@@ -1069,7 +1069,7 @@ void ObjcCategoryMerger::collectAndValidateCategoriesData() {
          off += target->wordSize) {
       Defined *categorySym = tryGetDefinedAtIsecOffset(catListCisec, off);
       assert(categorySym &&
-             "Failed to get a valid cateogry at __objc_catlit offset");
+             "Failed to get a valid category at __objc_catlit offset");
 
       // We only support ObjC categories (no swift + @objc)
       // TODO: Support swift + @objc categories also

From 58c5f50f4c73945fdc5440dee2fba03267a460d4 Mon Sep 17 00:00:00 2001
From: Leonard Chan <leonardchan@google.com>
Date: Thu, 11 Apr 2024 14:55:17 -0700
Subject: [PATCH 471/886] Reapply "[llvm] Teach GlobalDCE about
 dso_local_equivalent"

Also reapply "[llvm] Teach whole program devirtualization about
relative vtables"

This reverts commit 1c604a9780fcfe92a99d539913553f0835b81de3 and
474f5efebed24547e76d022f0c5ffcc9db97ce6f.
---
 .../include/llvm/Analysis/TypeMetadataUtils.h |   6 +-
 llvm/lib/Analysis/ModuleSummaryAnalysis.cpp   |  37 ++++-
 llvm/lib/Analysis/TypeMetadataUtils.cpp       |  47 ++++--
 llvm/test/ThinLTO/X86/devirt.ll               |  95 ++++++++++-
 ...virtual-functions-relative-pointers-bad.ll |  18 ++
 ...virtual-functions-relative-pointers-gep.ll |  24 +++
 .../virtual-functions-relative-pointers.ll    |  30 +++-
 .../WholeProgramDevirt/Inputs/export.yaml     |   8 +
 .../WholeProgramDevirt/branch-funnel.ll       | 157 ++++++++++++++++++
 .../WholeProgramDevirt/constant-arg.ll        |  45 +++++
 .../devirt-single-impl-check.ll               |  25 +++
 .../devirt-single-impl-multiple-assumes.ll    |  20 +++
 .../WholeProgramDevirt/devirt-single-impl.ll  |  47 +++++-
 .../WholeProgramDevirt/pointer-vtable.ll      |  16 ++
 14 files changed, 548 insertions(+), 27 deletions(-)

diff --git a/llvm/include/llvm/Analysis/TypeMetadataUtils.h b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
index 8894945c28d94..bdb477b54b532 100644
--- a/llvm/include/llvm/Analysis/TypeMetadataUtils.h
+++ b/llvm/include/llvm/Analysis/TypeMetadataUtils.h
@@ -66,7 +66,7 @@ void findDevirtualizableCallsForTypeCheckedLoad(
 /// Used for example from GlobalDCE to find an entry in a C++ vtable that
 /// matches a vcall offset.
 ///
-/// To support Swift vtables, getPointerAtOffset can see through "relative
+/// To support relative vtables, getPointerAtOffset can see through "relative
 /// pointers", i.e. (sub-)expressions of the form of:
 ///
 /// @symbol = ... {
@@ -87,8 +87,8 @@ std::pair<Function *, Constant *>
 getFunctionAtVTableOffset(GlobalVariable *GV, uint64_t Offset, Module &M);
 
 /// Finds the same "relative pointer" pattern as described above, where the
-/// target is `F`, and replaces the entire pattern with a constant zero.
-void replaceRelativePointerUsersWithZero(Function *F);
+/// target is `C`, and replaces the entire pattern with a constant zero.
+void replaceRelativePointerUsersWithZero(Constant *C);
 
 } // namespace llvm
 
diff --git a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
index deda1eebb3b57..c3d15afe6d9cb 100644
--- a/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
+++ b/llvm/lib/Analysis/ModuleSummaryAnalysis.cpp
@@ -22,6 +22,7 @@
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Analysis/BlockFrequencyInfo.h"
 #include "llvm/Analysis/BranchProbabilityInfo.h"
+#include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/IndirectCallPromotionAnalysis.h"
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/MemoryProfileInfo.h"
@@ -668,7 +669,8 @@ static void computeFunctionSummary(
 /// within the initializer.
 static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
                              const Module &M, ModuleSummaryIndex &Index,
-                             VTableFuncList &VTableFuncs) {
+                             VTableFuncList &VTableFuncs,
+                             const GlobalVariable &OrigGV) {
   // First check if this is a function pointer.
   if (I->getType()->isPointerTy()) {
     auto C = I->stripPointerCasts();
@@ -696,7 +698,7 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
       auto Offset = SL->getElementOffset(EI.index());
       unsigned Op = SL->getElementContainingOffset(Offset);
       findFuncPointers(cast<Constant>(I->getOperand(Op)),
-                       StartingOffset + Offset, M, Index, VTableFuncs);
+                       StartingOffset + Offset, M, Index, VTableFuncs, OrigGV);
     }
   } else if (auto *C = dyn_cast<ConstantArray>(I)) {
     ArrayType *ATy = C->getType();
@@ -704,7 +706,34 @@ static void findFuncPointers(const Constant *I, uint64_t StartingOffset,
     uint64_t EltSize = DL.getTypeAllocSize(EltTy);
     for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i) {
       findFuncPointers(cast<Constant>(I->getOperand(i)),
-                       StartingOffset + i * EltSize, M, Index, VTableFuncs);
+                       StartingOffset + i * EltSize, M, Index, VTableFuncs,
+                       OrigGV);
+    }
+  } else if (const auto *CE = dyn_cast<ConstantExpr>(I)) {
+    // For relative vtables, the next sub-component should be a trunc.
+    if (CE->getOpcode() != Instruction::Trunc ||
+        !(CE = dyn_cast<ConstantExpr>(CE->getOperand(0))))
+      return;
+
+    // If this constant can be reduced to the offset between a function and a
+    // global, then we know this is a valid virtual function if the RHS is the
+    // original vtable we're scanning through.
+    if (CE->getOpcode() == Instruction::Sub) {
+      GlobalValue *LHS, *RHS;
+      APSInt LHSOffset, RHSOffset;
+      if (IsConstantOffsetFromGlobal(CE->getOperand(0), LHS, LHSOffset, DL) &&
+          IsConstantOffsetFromGlobal(CE->getOperand(1), RHS, RHSOffset, DL) &&
+          RHS == &OrigGV &&
+
+          // For relative vtables, this component should point to the callable
+          // function without any offsets.
+          LHSOffset == 0 &&
+
+          // Also, the RHS should always point to somewhere within the vtable.
+          RHSOffset <=
+              static_cast<uint64_t>(DL.getTypeAllocSize(OrigGV.getInitializer()->getType()))) {
+        findFuncPointers(LHS, StartingOffset, M, Index, VTableFuncs, OrigGV);
+      }
     }
   }
 }
@@ -717,7 +746,7 @@ static void computeVTableFuncs(ModuleSummaryIndex &Index,
     return;
 
   findFuncPointers(V.getInitializer(), /*StartingOffset=*/0, M, Index,
-                   VTableFuncs);
+                   VTableFuncs, V);
 
 #ifndef NDEBUG
   // Validate that the VTableFuncs list is ordered by offset.
diff --git a/llvm/lib/Analysis/TypeMetadataUtils.cpp b/llvm/lib/Analysis/TypeMetadataUtils.cpp
index b8dcc39e9223c..67ce1540112bb 100644
--- a/llvm/lib/Analysis/TypeMetadataUtils.cpp
+++ b/llvm/lib/Analysis/TypeMetadataUtils.cpp
@@ -67,6 +67,14 @@ static void findLoadCallsAtConstantOffset(
         findLoadCallsAtConstantOffset(M, DevirtCalls, User, Offset + GEPOffset,
                                       CI, DT);
       }
+    } else if (auto *Call = dyn_cast<CallInst>(User)) {
+      if (Call->getIntrinsicID() == llvm::Intrinsic::load_relative) {
+        if (auto *LoadOffset = dyn_cast<ConstantInt>(Call->getOperand(1))) {
+          findCallsAtConstantOffset(DevirtCalls, nullptr, User,
+                                    Offset + LoadOffset->getSExtValue(), CI,
+                                    DT);
+        }
+      }
     }
   }
 }
@@ -131,6 +139,12 @@ void llvm::findDevirtualizableCallsForTypeCheckedLoad(
 
 Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                                    Constant *TopLevelGlobal) {
+  // TODO: Ideally it would be the caller who knows if it's appropriate to strip
+  // the DSOLocalEquicalent. More generally, it would feel more appropriate to
+  // have two functions that handle absolute and relative pointers separately.
+  if (auto *Equiv = dyn_cast<DSOLocalEquivalent>(I))
+    I = Equiv->getGlobalValue();
+
   if (I->getType()->isPointerTy()) {
     if (Offset == 0)
       return I;
@@ -161,7 +175,7 @@ Constant *llvm::getPointerAtOffset(Constant *I, uint64_t Offset, Module &M,
                               Offset % ElemSize, M, TopLevelGlobal);
   }
 
-  // (Swift-specific) relative-pointer support starts here.
+  // Relative-pointer support starts here.
   if (auto *CI = dyn_cast<ConstantInt>(I)) {
     if (Offset == 0 && CI->isZero()) {
       return I;
@@ -221,19 +235,26 @@ llvm::getFunctionAtVTableOffset(GlobalVariable *GV, uint64_t Offset,
   return std::pair<Function *, Constant *>(Fn, C);
 }
 
-void llvm::replaceRelativePointerUsersWithZero(Function *F) {
-  for (auto *U : F->users()) {
-    auto *PtrExpr = dyn_cast<ConstantExpr>(U);
-    if (!PtrExpr || PtrExpr->getOpcode() != Instruction::PtrToInt)
-      continue;
+static void replaceRelativePointerUserWithZero(User *U) {
+  auto *PtrExpr = dyn_cast<ConstantExpr>(U);
+  if (!PtrExpr || PtrExpr->getOpcode() != Instruction::PtrToInt)
+    return;
 
-    for (auto *PtrToIntUser : PtrExpr->users()) {
-      auto *SubExpr = dyn_cast<ConstantExpr>(PtrToIntUser);
-      if (!SubExpr || SubExpr->getOpcode() != Instruction::Sub)
-        continue;
+  for (auto *PtrToIntUser : PtrExpr->users()) {
+    auto *SubExpr = dyn_cast<ConstantExpr>(PtrToIntUser);
+    if (!SubExpr || SubExpr->getOpcode() != Instruction::Sub)
+      return;
 
-      SubExpr->replaceNonMetadataUsesWith(
-          ConstantInt::get(SubExpr->getType(), 0));
-    }
+    SubExpr->replaceNonMetadataUsesWith(
+        ConstantInt::get(SubExpr->getType(), 0));
+  }
+}
+
+void llvm::replaceRelativePointerUsersWithZero(Constant *C) {
+  for (auto *U : C->users()) {
+    if (auto *Equiv = dyn_cast<DSOLocalEquivalent>(U))
+      replaceRelativePointerUsersWithZero(Equiv);
+    else
+      replaceRelativePointerUserWithZero(U);
   }
 }
diff --git a/llvm/test/ThinLTO/X86/devirt.ll b/llvm/test/ThinLTO/X86/devirt.ll
index 472e43d776803..c4c2f861dc1f4 100644
--- a/llvm/test/ThinLTO/X86/devirt.ll
+++ b/llvm/test/ThinLTO/X86/devirt.ll
@@ -27,24 +27,36 @@
 ; NOENABLESPLITFLAG-DAG: [[B:\^[0-9]+]] = gv: (name: "_ZTV1B", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[Bf]], [[An]])
 ; NOENABLESPLITFLAG-DAG: [[C:\^[0-9]+]] = gv: (name: "_ZTV1C", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 16), (virtFunc: [[An]], offset: 24)), refs: ([[An]], [[Cf]])
 ; NOENABLESPLITFLAG-DAG: [[D:\^[0-9]+]] = gv: (name: "_ZTV1D", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 16)), refs: ([[Dm]])
+; NOENABLESPLITFLAG-DAG: [[B_RV:\^[0-9]+]] = gv: (name: "_ZTV1B_RV", {{.*}} vTableFuncs: ((virtFunc: [[Bf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[B_RV]], [[Bf]], [[An]])
+; NOENABLESPLITFLAG-DAG: [[C_RV:\^[0-9]+]] = gv: (name: "_ZTV1C_RV", {{.*}} vTableFuncs: ((virtFunc: [[Cf]], offset: 8), (virtFunc: [[An]], offset: 12)), refs: ([[C_RV]], [[An]], [[Cf]])
+; NOENABLESPLITFLAG-DAG: [[D_RV:\^[0-9]+]] = gv: (name: "_ZTV1D_RV", {{.*}} vTableFuncs: ((virtFunc: [[Dm]], offset: 8)), refs: ([[D_RV]], [[Dm]])
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A", summary: ((offset: 16, [[B]]), (offset: 16, [[C]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B", summary: ((offset: 16, [[B]])))
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C", summary: ((offset: 16, [[C]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1A_RV", summary: ((offset: 8, [[B_RV]]), (offset: 8, [[C_RV]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1B_RV", summary: ((offset: 8, [[B_RV]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "_ZTS1C_RV", summary: ((offset: 8, [[C_RV]])))
 ; Type Id on _ZTV1D should have been promoted
 ; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "1.{{.*}}", summary: ((offset: 16, [[D]])))
+; NOENABLESPLITFLAG-DAG: typeidCompatibleVTable: (name: "2.{{.*}}", summary: ((offset: 8, [[D_RV]])))
 
 ; Index based WPD
 ; RUN: llvm-lto2 run %t2.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
+; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK
+; RUN:   -r=%t2.o,_ZTV1D,px \
+; RUN:   -r=%t2.o,_ZTV1B_RV,px \
+; RUN:   -r=%t2.o,_ZTV1C_RV,px \
+; RUN:   -r=%t2.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
 ; Check that we're able to prevent specific function from being
@@ -54,18 +66,24 @@
 ; RUN:   -wholeprogramdevirt-skip=_ZN1A1nEi \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t2.o,test,px \
+; RUN:   -r=%t2.o,test_rv,px \
 ; RUN:   -r=%t2.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t2.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1C1fEi,p \
 ; RUN:   -r=%t2.o,_ZN1D1mEi,p \
 ; RUN:   -r=%t2.o,_ZTV1B,px \
 ; RUN:   -r=%t2.o,_ZTV1C,px \
-; RUN:   -r=%t2.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=SKIP
+; RUN:   -r=%t2.o,_ZTV1D,px \
+; RUN:   -r=%t2.o,_ZTV1B_RV,px \
+; RUN:   -r=%t2.o,_ZTV1C_RV,px \
+; RUN:   -r=%t2.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=SKIP
 
 ; RUN: llvm-lto2 run %t.o -save-temps -pass-remarks=. \
 ; RUN:   -whole-program-visibility \
 ; RUN:   -o %t3 \
 ; RUN:   -r=%t.o,test,px \
+; RUN:   -r=%t.o,test_rv,px \
 ; RUN:   -r=%t.o,_ZN1A1nEi,p \
 ; RUN:   -r=%t.o,_ZN1B1fEi,p \
 ; RUN:   -r=%t.o,_ZN1C1fEi,p \
@@ -73,15 +91,24 @@
 ; RUN:   -r=%t.o,_ZTV1B, \
 ; RUN:   -r=%t.o,_ZTV1C, \
 ; RUN:   -r=%t.o,_ZTV1D, \
+; RUN:   -r=%t.o,_ZTV1B_RV, \
+; RUN:   -r=%t.o,_ZTV1C_RV, \
+; RUN:   -r=%t.o,_ZTV1D_RV, \
 ; RUN:   -r=%t.o,_ZN1A1nEi, \
 ; RUN:   -r=%t.o,_ZN1B1fEi, \
 ; RUN:   -r=%t.o,_ZN1C1fEi, \
 ; RUN:   -r=%t.o,_ZN1D1mEi, \
 ; RUN:   -r=%t.o,_ZTV1B,px \
 ; RUN:   -r=%t.o,_ZTV1C,px \
-; RUN:   -r=%t.o,_ZTV1D,px 2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
+; RUN:   -r=%t.o,_ZTV1D,px \
+; RUN:   -r=%t.o,_ZTV1B_RV,px \
+; RUN:   -r=%t.o,_ZTV1C_RV,px \
+; RUN:   -r=%t.o,_ZTV1D_RV,px \
+; RUN:   2>&1 | FileCheck %s --check-prefix=REMARK --dump-input=fail
 ; RUN: llvm-dis %t3.1.4.opt.bc -o - | FileCheck %s --check-prefix=CHECK-IR
 
+; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
+; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1A1nEi
 ; REMARK-DAG: single-impl: devirtualized a call to _ZN1D1mEi
 
@@ -99,6 +126,25 @@ target triple = "x86_64-grtev4-linux-gnu"
 @_ZTV1C = constant { [4 x ptr] } { [4 x ptr] [ptr null, ptr undef, ptr @_ZN1C1fEi, ptr @_ZN1A1nEi] }, !type !0, !type !2
 @_ZTV1D = constant { [3 x ptr] } { [3 x ptr] [ptr null, ptr undef, ptr @_ZN1D1mEi] }, !type !3
 
+@_ZTV1B_RV = constant { [4 x i32] } { [4 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1B1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1B_RV, i32 0, i32 0, i32 3) to i64)) to i32)
+] }, !type !7, !type !8
+
+@_ZTV1C_RV = constant { [4 x i32] } { [4 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1C1fEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1A1nEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @_ZTV1C_RV, i32 0, i32 0, i32 3) to i64)) to i32)
+] }, !type !7, !type !9
+
+@_ZTV1D_RV = constant { [3 x i32] } { [3 x i32] [
+  i32 0,
+  i32 undef,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @_ZN1D1mEi to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1D_RV, i32 0, i32 0, i32 2) to i64)) to i32)
+] }, !type !10
 
 ; CHECK-IR-LABEL: define {{(noundef )?}}i32 @test
 define i32 @test(ptr %obj, ptr %obj2, i32 %a) {
@@ -136,6 +182,43 @@ entry:
 ; CHECK-IR-LABEL: ret i32
 ; CHECK-IR-LABEL: }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+; CHECK-IR-LABEL: define {{.*}}i32 @test_rv
+define i32 @test_rv(ptr %obj, ptr %obj2, i32 %a) {
+entry:
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"_ZTS1A_RV")
+  call void @llvm.assume(i1 %p)
+  %fptr1_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 4)
+
+  ; Check that the call was devirtualized.
+  ; CHECK-IR: %call = tail call i32 @_ZN1A1nEi
+  ; Ensure !prof and !callees metadata for indirect call promotion removed.
+  ; CHECK-IR-NOT: prof
+  ; CHECK-IR-NOT: callees
+  %call = tail call i32 %fptr1_rv(ptr nonnull %obj, i32 %a), !prof !5, !callees !6
+
+  %fptr22_rv = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+
+  ; We still have to call it as virtual.
+  ; CHECK-IR: %call3 = tail call i32 %fptr22
+  %call3 = tail call i32 %fptr22_rv(ptr nonnull %obj, i32 %call)
+
+  %vtable2 = load ptr, ptr %obj2
+  %p2 = call i1 @llvm.type.test(ptr %vtable2, metadata !11)
+  call void @llvm.assume(i1 %p2)
+
+  %fptr33_rv = call ptr @llvm.load.relative.i32(ptr %vtable2, i32 0)
+
+  ; Check that the call was devirtualized.
+  ; CHECK-IR: %call4 = tail call i32 @_ZN1D1mEi
+  %call4 = tail call i32 %fptr33_rv(ptr nonnull %obj2, i32 %call3)
+  ret i32 %call4
+}
+; CHECK-IR-LABEL: ret i32
+; CHECK-IR-LABEL: }
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -165,3 +248,9 @@ attributes #0 = { noinline optnone }
 !4 = distinct !{}
 !5 = !{!"VP", i32 0, i64 1, i64 1621563287929432257, i64 1}
 !6 = !{ptr @_ZN1A1nEi}
+
+!7 = !{i64 8, !"_ZTS1A_RV"}
+!8 = !{i64 8, !"_ZTS1B_RV"}
+!9 = !{i64 8, !"_ZTS1C_RV"}
+!10 = !{i64 8, !11}
+!11 = distinct !{}
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
index 9d87fdb084a12..9a4bff80383cb 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-bad.ll
@@ -16,14 +16,32 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 
 ; CHECK:      @vtable = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 8, !type !0, !type !1, !vcall_visibility !2
 
+@vtable2 = internal unnamed_addr constant { [3 x i32] } { [3 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4 to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+
+  ; a "bad" relative pointer because it's base is not the @vtable symbol
+  i32 trunc (i64 sub (i64 ptrtoint (ptr @weird_ref_3         to i64), i64 ptrtoint (ptr @weird_ref_4 to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 0, !"vfunc3.type"}
+!4 = !{i64 4, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [3 x i32] } zeroinitializer, align 4, !type !3, !type !4, !vcall_visibility !2
+
 define internal void @vfunc1() { ret void }
 define internal void @vfunc2() { ret void }
 define internal void @weird_ref_1() { ret void }
 define internal void @weird_ref_2() { ret void }
+declare void @vfunc3()
+declare void @vfunc4()
+declare void @weird_ref_3()
+declare void @weird_ref_4()
 
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   call void @weird_ref_2()
+  %2 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  call void @weird_ref_4()
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
index fb45d37dde6e4..d89b024e7a6d0 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers-gep.ll
@@ -19,6 +19,20 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
+@vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
+  i32 42,
+  i32 1337,
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 8, !"vfunc3.type"}
+!4 = !{i64 12, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [4 x i32] } { [4 x i32] [
+; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [4 x i32] }, ptr @vtable2, i32 0, i32 0, i32 2) to i64)) to i32),
+; CHECK-SAME:   i32 0
+; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
+
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -31,9 +45,19 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
+; (3) vfunc3_live_extern is referenced from @main, stays alive
+; CHECK: declare void @vfunc3_live_extern
+declare void @vfunc3_live_extern()
+
+; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
+; CHECK-NOT: declare void @vfunc4_dead_extern
+declare void @vfunc4_dead_extern()
+
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
+  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 
diff --git a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
index 76a617ea99e6f..0b70148aa0eec 100644
--- a/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
+++ b/llvm/test/Transforms/GlobalDCE/virtual-functions-relative-pointers.ll
@@ -6,17 +6,31 @@ declare { ptr, i1 } @llvm.type.checked.load(ptr, i32, metadata)
 
 ; A vtable with "relative pointers", slots don't contain pointers to implementations, but instead have an i32 offset from the vtable itself to the implementation.
 @vtable = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
-  i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc1_live              to i64), i64 ptrtoint (ptr @vtable to i64)) to i32),
-  i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc2_dead              to i64), i64 ptrtoint (ptr @vtable to i64)) to i32)
+  i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc1_live to i64), i64 ptrtoint (ptr @vtable to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc2_dead to i64), i64 ptrtoint (ptr @vtable to i64)) to i32)
 ]}, align 8, !type !0, !type !1, !vcall_visibility !{i64 2}
 !0 = !{i64 0, !"vfunc1.type"}
 !1 = !{i64 4, !"vfunc2.type"}
 
 ; CHECK:      @vtable = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
-; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc1_live              to i64), i64 ptrtoint (ptr @vtable to i64)) to i32),
+; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr @vfunc1_live to i64), i64 ptrtoint (ptr @vtable to i64)) to i32),
 ; CHECK-SAME:   i32 0
 ; CHECK-SAME: ] }, align 8, !type !0, !type !1, !vcall_visibility !2
 
+; Similar to above, but the vtable is more aligned to how C++ relative vtables look.
+; That is, the functions may not be dso-local.
+@vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc4_dead_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32)
+]}, align 4, !type !3, !type !4, !vcall_visibility !{i64 2}
+!3 = !{i64 0, !"vfunc3.type"}
+!4 = !{i64 4, !"vfunc4.type"}
+
+; CHECK:      @vtable2 = internal unnamed_addr constant { [2 x i32] } { [2 x i32] [
+; CHECK-SAME:   i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vfunc3_live_extern to i64), i64 ptrtoint (ptr @vtable2 to i64)) to i32),
+; CHECK-SAME:   i32 0
+; CHECK-SAME: ] }, align 4, !type !3, !type !4, !vcall_visibility !2
+
 ; (1) vfunc1_live is referenced from @main, stays alive
 define internal void @vfunc1_live() {
   ; CHECK: define internal void @vfunc1_live(
@@ -29,9 +43,19 @@ define internal void @vfunc2_dead() {
   ret void
 }
 
+; (3) vfunc3_live_extern is referenced from @main, stays alive
+; CHECK: declare void @vfunc3_live_extern
+declare void @vfunc3_live_extern()
+
+; (4) vfunc4_dead_extern is never referenced, gets removed and vtable slot is null'd
+; CHECK-NOT: declare void @vfunc4_dead_extern
+declare void @vfunc4_dead_extern()
+
 define void @main() {
   %1 = ptrtoint ptr @vtable to i64 ; to keep @vtable alive
   %2 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc1.type")
+  %3 = ptrtoint ptr @vtable2 to i64 ; to keep @vtable2 alive
+  %4 = tail call { ptr, i1 } @llvm.type.checked.load(ptr null, i32 0, metadata !"vfunc3.type")
   ret void
 }
 
diff --git a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
index 71cf38b216c71..dd0c90d7ae8c3 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
+++ b/llvm/test/Transforms/WholeProgramDevirt/Inputs/export.yaml
@@ -5,14 +5,22 @@ GlobalValueMap:
       TypeTestAssumeVCalls:
         - GUID: 14276520915468743435  # typeid1
           Offset: 0
+        - GUID: 271751036925422857  # typeid1_rv
+          Offset: 0
       TypeCheckedLoadVCalls:
         - GUID: 15427464259790519041  # typeid2
           Offset: 0
+        - GUID: 1146149264729288256  # typeid2_rv
+          Offset: 0
       TypeTestAssumeConstVCalls:
         - VFunc:
             GUID: 3515965990081467659  # typeid3
             Offset: 0
           Args: [12, 24]
+        - VFunc:
+            GUID: 2777626534618191571  # typeid3_rv
+            Offset: 0
+          Args: [12, 24]
       TypeCheckedLoadConstVCalls:
         - VFunc:
             GUID: 17525413373118030901  # typeid4
diff --git a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
index b55713fe4b296..0b1023eee2732 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/branch-funnel.ll
@@ -8,6 +8,45 @@
 ; RUN: FileCheck --check-prefix=SUMMARY %s < %t
 
 ; SUMMARY:      TypeIdMap:       
+; SUMMARY-NEXT:  typeid1_rv:
+; SUMMARY-NEXT:    TTRes:
+; SUMMARY-NEXT:      Kind:            Unknown
+; SUMMARY-NEXT:      SizeM1BitWidth:  0
+; SUMMARY-NEXT:      AlignLog2:       0
+; SUMMARY-NEXT:      SizeM1:          0
+; SUMMARY-NEXT:      BitMask:         0
+; SUMMARY-NEXT:      InlineBits:      0
+; SUMMARY-NEXT:    WPDRes:
+; SUMMARY-NEXT:      0:
+; SUMMARY-NEXT:        Kind:            BranchFunnel
+; SUMMARY-NEXT:        SingleImplName:  ''
+; SUMMARY-NEXT:        ResByArg:
+; SUMMARY-NEXT:  typeid2_rv:
+; SUMMARY-NEXT:    TTRes:
+; SUMMARY-NEXT:      Kind:            Unknown
+; SUMMARY-NEXT:      SizeM1BitWidth:  0
+; SUMMARY-NEXT:      AlignLog2:       0
+; SUMMARY-NEXT:      SizeM1:          0
+; SUMMARY-NEXT:      BitMask:         0
+; SUMMARY-NEXT:      InlineBits:      0
+; SUMMARY-NEXT:    WPDRes:
+; SUMMARY-NEXT:      0:
+; SUMMARY-NEXT:        Kind:            Indir
+; SUMMARY-NEXT:        SingleImplName:  ''
+; SUMMARY-NEXT:        ResByArg:
+; SUMMARY-NEXT:   typeid3_rv:
+; SUMMARY-NEXT:     TTRes:           
+; SUMMARY-NEXT:       Kind:            Unknown
+; SUMMARY-NEXT:       SizeM1BitWidth:  0
+; SUMMARY-NEXT:       AlignLog2:       0
+; SUMMARY-NEXT:       SizeM1:          0
+; SUMMARY-NEXT:       BitMask:         0
+; SUMMARY-NEXT:       InlineBits:      0
+; SUMMARY-NEXT:     WPDRes:          
+; SUMMARY-NEXT:       0:               
+; SUMMARY-NEXT:         Kind:            BranchFunnel
+; SUMMARY-NEXT:         SingleImplName:  ''
+; SUMMARY-NEXT:         ResByArg:        
 ; SUMMARY-NEXT:   typeid3:
 ; SUMMARY-NEXT:     TTRes:           
 ; SUMMARY-NEXT:       Kind:            Unknown
@@ -93,6 +132,29 @@ declare i32 @vf3_2(ptr %this, i32 %arg)
 declare i32 @vf4_1(ptr %this, i32 %arg)
 declare i32 @vf4_2(ptr %this, i32 %arg)
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+;; These are relative vtables equivalent to the ones above.
+@vt1_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_1 to i64), i64 ptrtoint (ptr @vt1_1_rv to i64)) to i32)], !type !5
+@vt1_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1_2 to i64), i64 ptrtoint (ptr @vt1_2_rv to i64)) to i32)], !type !5
+
+@vt2_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_1 to i64), i64 ptrtoint (ptr @vt2_1_rv to i64)) to i32)], !type !6
+@vt2_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_2 to i64), i64 ptrtoint (ptr @vt2_2_rv to i64)) to i32)], !type !6
+@vt2_3_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_3 to i64), i64 ptrtoint (ptr @vt2_3_rv to i64)) to i32)], !type !6
+@vt2_4_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_4 to i64), i64 ptrtoint (ptr @vt2_4_rv to i64)) to i32)], !type !6
+@vt2_5_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_5 to i64), i64 ptrtoint (ptr @vt2_5_rv to i64)) to i32)], !type !6
+@vt2_6_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_6 to i64), i64 ptrtoint (ptr @vt2_6_rv to i64)) to i32)], !type !6
+@vt2_7_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_7 to i64), i64 ptrtoint (ptr @vt2_7_rv to i64)) to i32)], !type !6
+@vt2_8_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_8 to i64), i64 ptrtoint (ptr @vt2_8_rv to i64)) to i32)], !type !6
+@vt2_9_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_9 to i64), i64 ptrtoint (ptr @vt2_9_rv to i64)) to i32)], !type !6
+@vt2_10_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_10 to i64), i64 ptrtoint (ptr @vt2_10_rv to i64)) to i32)], !type !6
+@vt2_11_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2_11 to i64), i64 ptrtoint (ptr @vt2_11_rv to i64)) to i32)], !type !6
+
+@vt3_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_1 to i64), i64 ptrtoint (ptr @vt3_1_rv to i64)) to i32)], !type !7
+@vt3_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf3_2 to i64), i64 ptrtoint (ptr @vt3_2_rv to i64)) to i32)], !type !7
+
+@vt4_1_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_1 to i64), i64 ptrtoint (ptr @vt4_1_rv to i64)) to i32)], !type !8
+@vt4_2_rv = constant [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4_2 to i64), i64 ptrtoint (ptr @vt4_2_rv to i64)) to i32)], !type !8
 
 
 ; CHECK-LABEL: define i32 @fn1
@@ -108,6 +170,19 @@ define i32 @fn1(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn1_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn1_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest %vtable, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define i32 @fn2
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn2(ptr %obj) #0 {
@@ -120,6 +195,18 @@ define i32 @fn2(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn2_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn2_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call i32 %
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define i32 @fn3
 ; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
 define i32 @fn3(ptr %obj) #0 {
@@ -133,10 +220,75 @@ define i32 @fn3(ptr %obj) #0 {
   ret i32 %result
 }
 
+; CHECK-LABEL: define i32 @fn3_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn3_rv(ptr %obj) #0 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !9)
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; RETP: call i32 @branch_funnel.1(ptr
+  ; NORETP: call i32 %
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
+  call void @llvm.assume(i1 %p)
+  %fptr = load ptr, ptr @vt1_1
+  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_cpy
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_cpy(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1, metadata !"typeid1")
+  call void @llvm.assume(i1 %p)
+  %fptr = load ptr, ptr @vt1_1
+  ; RETP: call i32 @__typeid_typeid1_0_branch_funnel(ptr nest @vt1_1, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_rv
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_rv(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
+; CHECK-LABEL: define i32 @fn4_rv_cpy
+; CHECK-NOT: call void (...) @llvm.icall.branch.funnel
+define i32 @fn4_rv_cpy(ptr %obj) #0 {
+  %p = call i1 @llvm.type.test(ptr @vt1_1_rv, metadata !"typeid1_rv")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr @vt1_1_rv, i32 0)
+  ; RETP: call i32 @__typeid_typeid1_rv_0_branch_funnel(ptr nest @vt1_1_rv, ptr %obj, i32 1)
+  %result = call i32 %fptr(ptr %obj, i32 1)
+  ; NORETP: call i32 %
+  ret i32 %result
+}
+
 ; CHECK-LABEL: define hidden void @__typeid_typeid1_0_branch_funnel(ptr nest %0, ...)
 ; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2, ptr {{(nonnull )?}}@vf1_2, ...)
 
+; CHECK-LABEL: define hidden void @__typeid_typeid1_rv_0_branch_funnel(ptr nest %0, ...)
+; CHECK-NEXT: musttail call void (...) @llvm.icall.branch.funnel(ptr %0, ptr {{(nonnull )?}}@vt1_1_rv, ptr {{(nonnull )?}}@vf1_1, ptr {{(nonnull )?}}@vt1_2_rv, ptr {{(nonnull )?}}@vf1_2, ...)
+
 ; CHECK: define internal void @branch_funnel(ptr
+; CHECK: define internal void @branch_funnel.1(ptr
 
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
@@ -146,5 +298,10 @@ declare void @llvm.assume(i1)
 !2 = !{i32 0, !"typeid3"}
 !3 = !{i32 0, !4}
 !4 = distinct !{}
+!5 = !{i32 0, !"typeid1_rv"}
+!6 = !{i32 0, !"typeid2_rv"}
+!7 = !{i32 0, !"typeid3_rv"}
+!8 = !{i32 0, !9}
+!9 = distinct !{}
 
 attributes #0 = { "target-features"="+retpoline" }
diff --git a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
index 91bae4af9a211..e685a1a54c520 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/constant-arg.ll
@@ -7,6 +7,10 @@ target triple = "x86_64-unknown-linux-gnu"
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf2], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\01", [1 x ptr] [ptr @vf4], [0 x i8] zeroinitializer }, !type [[T8]]
 ; CHECK: private constant { [8 x i8], [1 x ptr], [0 x i8] } { [8 x i8] c"\00\00\00\00\00\00\00\02", [1 x ptr] [ptr @vf8], [0 x i8] zeroinitializer }, !type [[T8]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4:![0-9]+]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\01", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
+; CHECK: private constant { [4 x i8], [1 x i32], [0 x i8] } { [4 x i8] c"\00\00\00\02", [1 x i32] [i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)], [0 x i8] zeroinitializer }, align 4, !type [[T4]]
 
 @vt1 = constant [1 x ptr] [ptr @vf1], !type !0
 @vt2 = constant [1 x ptr] [ptr @vf2], !type !0
@@ -61,8 +65,49 @@ define i1 @call2(ptr %obj) {
   ret i1 %result
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt1_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf1 to i64), i64 ptrtoint (ptr @vt1_rv to i64)) to i32)
+], align 4, !type !1
+@vt2_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf2 to i64), i64 ptrtoint (ptr @vt2_rv to i64)) to i32)
+], align 4, !type !1
+@vt4_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf4 to i64), i64 ptrtoint (ptr @vt4_rv to i64)) to i32)
+], align 4, !type !1
+@vt8_rv = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf8 to i64), i64 ptrtoint (ptr @vt8_rv to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define i1 @call3
+define i1 @call3(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: getelementptr {{.*}} -1
+  ; CHECK: and {{.*}}, 1
+  %result = call i1 %fptr(ptr %obj, i32 5)
+  ret i1 %result
+}
+
+; CHECK: define i1 @call4
+define i1 @call4(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: getelementptr {{.*}} -1
+  ; CHECK: and {{.*}}, 2
+  %result = call i1 %fptr(ptr %obj, i32 10)
+  ret i1 %result
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 ; CHECK: [[T8]] = !{i32 8, !"typeid"}
+; CHECK: [[T4]] = !{i32 4, !"typeid2"}
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
index dc7b2024bc591..5a9e6c12c3582 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-check.ll
@@ -3,6 +3,7 @@
 target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
+; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: single-impl: devirtualized a call to vf
 ; CHECK: remark: <unknown>:0:0: devirtualized vf
 ; CHECK-NOT: devirtualized
@@ -33,7 +34,31 @@ trap:
   unreachable
 }
 
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %pair = call {ptr, i1} @llvm.type.checked.load(ptr %vtable, i32 0, metadata !"typeid2")
+  %fptr = extractvalue {ptr, i1} %pair, 0
+  %p = extractvalue {ptr, i1} %pair, 1
+  ; CHECK: br i1 true,
+  br i1 %p, label %cont, label %trap
+
+cont:
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+
+trap:
+  call void @llvm.trap()
+  unreachable
+}
+
 declare {ptr, i1} @llvm.type.checked.load(ptr, i32, metadata)
 declare void @llvm.trap()
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
index ed144c23302dd..bfbbeaf8a0656 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl-multiple-assumes.ll
@@ -23,7 +23,27 @@ define void @call(ptr %obj) {
   ret void
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !1
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %p2 = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p2)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}
diff --git a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
index 1ce96f70d905f..d7f33df594956 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/devirt-single-impl.ll
@@ -7,6 +7,8 @@ target datalayout = "e-p:64:64"
 target triple = "x86_64-unknown-linux-gnu"
 
 ; CHECK: remark: devirt-single.cc:30:32: single-impl: devirtualized a call to vf
+; CHECK: remark: devirt-single.cc:41:32: single-impl: devirtualized a call to vf
+; CHECK: remark: devirt-single.cc:51:32: single-impl: devirtualized a call to vf
 ; CHECK: remark: devirt-single.cc:13:0: devirtualized vf
 ; CHECK-NOT: devirtualized
 
@@ -28,6 +30,41 @@ define void @call(ptr %obj) #1 !dbg !5 {
   ret void
 }
 
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+@vt3 = private unnamed_addr constant [1 x i32] [
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt3 to i64)) to i32)
+], align 4, !type !11
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) #1 !dbg !9 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj), !dbg !10
+  ret void
+}
+
+@_ZTV1A.local = private unnamed_addr constant { [3 x i32] } { [3 x i32] [
+  i32 0,  ; offset to top
+  i32 0,  ; rtti
+  i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr getelementptr inbounds ({ [3 x i32] }, ptr @_ZTV1A.local, i32 0, i32 0, i32 2) to i64)) to i32)  ; vfunc offset
+] }, align 4, !type !14
+
+; CHECK: define void @call3
+define void @call3(ptr %obj) #1 !dbg !12 {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid3")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 8)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj), !dbg !13
+  ret void
+}
+
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
@@ -45,5 +82,13 @@ declare void @llvm.assume(i1)
 !7 = distinct !DISubprogram(name: "vf", linkageName: "_ZN3vt12vfEv", scope: !1, file: !1, line: 13, isLocal: false, isDefinition: true, scopeLine: 13, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
 !8 = !{i32 0, !"typeid"}
 
+!9 = distinct !DISubprogram(name: "call2", linkageName: "_Z5call2Pv", scope: !1, file: !1, line: 40, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
+!10 = !DILocation(line: 41, column: 32, scope: !9)
+!11 = !{i32 0, !"typeid2"}
+
+!12 = distinct !DISubprogram(name: "call3", linkageName: "_Z5call3Pv", scope: !1, file: !1, line: 50, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: false, unit: !0)
+!13 = !DILocation(line: 51, column: 32, scope: !12)
+!14 = !{i32 0, !"typeid3"}
+
 ; CHECK: 1 wholeprogramdevirt - Number of whole program devirtualization targets
-; CHECK: 1 wholeprogramdevirt - Number of single implementation devirtualizations
+; CHECK: 3 wholeprogramdevirt - Number of single implementation devirtualizations
diff --git a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
index 3c5d9b839bc4d..062cef9b88fe4 100644
--- a/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
+++ b/llvm/test/Transforms/WholeProgramDevirt/pointer-vtable.ll
@@ -20,7 +20,23 @@ define void @call(ptr %obj) {
   ret void
 }
 
+@vt2 = constant i32 trunc (i64 sub (i64 ptrtoint (ptr dso_local_equivalent @vf to i64), i64 ptrtoint (ptr @vt2 to i64)) to i32), !type !1
+
+declare ptr @llvm.load.relative.i32(ptr, i32)
+
+; CHECK: define void @call2
+define void @call2(ptr %obj) {
+  %vtable = load ptr, ptr %obj
+  %p = call i1 @llvm.type.test(ptr %vtable, metadata !"typeid2")
+  call void @llvm.assume(i1 %p)
+  %fptr = call ptr @llvm.load.relative.i32(ptr %vtable, i32 0)
+  ; CHECK: call void @vf(
+  call void %fptr(ptr %obj)
+  ret void
+}
+
 declare i1 @llvm.type.test(ptr, metadata)
 declare void @llvm.assume(i1)
 
 !0 = !{i32 0, !"typeid"}
+!1 = !{i32 0, !"typeid2"}

From 75bbf4dd7cfda744d4a2487cfd06a58194d02db3 Mon Sep 17 00:00:00 2001
From: Job Henandez Lara <hj93@protonmail.com>
Date: Mon, 15 Apr 2024 11:45:16 -0700
Subject: [PATCH 472/886] [libc] Add proxy headers for fenv types. (#88467)

Fixes #88187
---
 libc/hdr/types/CMakeLists.txt                 | 18 +++++++++++++++
 libc/hdr/types/fenv_t.h                       | 22 +++++++++++++++++++
 libc/hdr/types/fexcept_t.h                    | 22 +++++++++++++++++++
 libc/src/__support/FPUtil/CMakeLists.txt      |  2 +-
 libc/src/__support/FPUtil/aarch64/FEnvImpl.h  |  2 +-
 .../FPUtil/aarch64/fenv_darwin_impl.h         |  2 +-
 libc/src/__support/FPUtil/arm/FEnvImpl.h      |  2 +-
 libc/src/__support/FPUtil/riscv/FEnvImpl.h    |  1 +
 libc/src/__support/FPUtil/x86_64/FEnvImpl.h   |  2 +-
 libc/src/fenv/CMakeLists.txt                  |  7 +++++-
 libc/src/fenv/fegetenv.h                      |  2 +-
 libc/src/fenv/fegetexceptflag.cpp             |  3 +--
 libc/src/fenv/feholdexcept.cpp                |  2 +-
 libc/src/fenv/feholdexcept.h                  |  2 +-
 libc/src/fenv/fesetenv.h                      |  2 +-
 libc/src/fenv/fesetexceptflag.cpp             |  2 +-
 libc/src/fenv/feupdateenv.h                   |  2 +-
 libc/test/UnitTest/CMakeLists.txt             |  1 +
 libc/test/UnitTest/FPExceptMatcher.cpp        |  7 +++---
 libc/test/src/fenv/exception_flags_test.cpp   |  3 +--
 libc/test/src/fenv/feholdexcept_test.cpp      |  3 +--
 libc/test/src/fenv/feupdateenv_test.cpp       |  2 +-
 libc/test/src/fenv/getenv_and_setenv_test.cpp |  3 +--
 .../llvm-project-overlay/libc/BUILD.bazel     | 13 +++++++++++
 .../libc/test/UnitTest/BUILD.bazel            |  1 +
 .../libc/test/src/fenv/BUILD.bazel            |  4 ++++
 26 files changed, 108 insertions(+), 24 deletions(-)
 create mode 100644 libc/hdr/types/fenv_t.h
 create mode 100644 libc/hdr/types/fexcept_t.h

diff --git a/libc/hdr/types/CMakeLists.txt b/libc/hdr/types/CMakeLists.txt
index b685d82fd8cc8..ecb952b60cc06 100644
--- a/libc/hdr/types/CMakeLists.txt
+++ b/libc/hdr/types/CMakeLists.txt
@@ -21,3 +21,21 @@ add_proxy_header_library(
   FULL_BUILD_DEPENDS
     libc.include.llvm-libc-types.struct_timespec
 )
+
+add_proxy_header_library(
+  fenv_t
+  HDRS
+    fenv_t.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-types.fenv_t
+    libc.incude.fenv
+)
+
+add_proxy_header_library(
+  fexcept_t
+  HDRS
+    fexcept_t.h
+  FULL_BUILD_DEPENDS
+    libc.include.llvm-libc-types.fexcept_t
+    libc.incude.fenv
+)
diff --git a/libc/hdr/types/fenv_t.h b/libc/hdr/types/fenv_t.h
new file mode 100644
index 0000000000000..dc2dd0217a16f
--- /dev/null
+++ b/libc/hdr/types/fenv_t.h
@@ -0,0 +1,22 @@
+//===-- Definition of macros from fenv_t.h --------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_HDR_FENV_T_H
+#define LLVM_LIBC_HDR_FENV_T_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-types/fenv_t.h"
+
+#else // Overlay mode
+
+#include <fenv.h>
+
+#endif // LLVM_LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_FENV_T_H
diff --git a/libc/hdr/types/fexcept_t.h b/libc/hdr/types/fexcept_t.h
new file mode 100644
index 0000000000000..7a6702516f0a3
--- /dev/null
+++ b/libc/hdr/types/fexcept_t.h
@@ -0,0 +1,22 @@
+//===-- Definition of macros from fexcept_t.h -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_HDR_FEXCEPT_T_H
+#define LLVM_LIBC_HDR_FEXCEPT_T_H
+
+#ifdef LIBC_FULL_BUILD
+
+#include "include/llvm-libc-types/fexcept_t.h"
+
+#else // Overlay mode
+
+#include <fenv.h>
+
+#endif // LLVM_LIBC_FULL_BUILD
+
+#endif // LLVM_LIBC_HDR_FENV_T_H
diff --git a/libc/src/__support/FPUtil/CMakeLists.txt b/libc/src/__support/FPUtil/CMakeLists.txt
index 0b5ea83689430..01ca4254c7996 100644
--- a/libc/src/__support/FPUtil/CMakeLists.txt
+++ b/libc/src/__support/FPUtil/CMakeLists.txt
@@ -3,7 +3,7 @@ add_header_library(
   HDRS
     FEnvImpl.h
   DEPENDS
-    libc.include.fenv
+    libc.hdr.types.fenv_t
     libc.hdr.fenv_macros
     libc.hdr.math_macros
     libc.src.__support.macros.attributes
diff --git a/libc/src/__support/FPUtil/aarch64/FEnvImpl.h b/libc/src/__support/FPUtil/aarch64/FEnvImpl.h
index 4b593cdd8cc4e..d1d92169475d1 100644
--- a/libc/src/__support/FPUtil/aarch64/FEnvImpl.h
+++ b/libc/src/__support/FPUtil/aarch64/FEnvImpl.h
@@ -17,10 +17,10 @@
 #endif
 
 #include <arm_acle.h>
-#include <fenv.h>
 #include <stdint.h>
 
 #include "hdr/fenv_macros.h"
+#include "hdr/types/fenv_t.h"
 #include "src/__support/FPUtil/FPBits.h"
 
 namespace LIBC_NAMESPACE {
diff --git a/libc/src/__support/FPUtil/aarch64/fenv_darwin_impl.h b/libc/src/__support/FPUtil/aarch64/fenv_darwin_impl.h
index 773d6bfe9f892..5b59ba38d67bb 100644
--- a/libc/src/__support/FPUtil/aarch64/fenv_darwin_impl.h
+++ b/libc/src/__support/FPUtil/aarch64/fenv_darwin_impl.h
@@ -17,10 +17,10 @@
 #endif
 
 #include <arm_acle.h>
-#include <fenv.h>
 #include <stdint.h>
 
 #include "hdr/fenv_macros.h"
+#include "hdr/types/fenv_t.h"
 #include "src/__support/FPUtil/FPBits.h"
 
 namespace LIBC_NAMESPACE {
diff --git a/libc/src/__support/FPUtil/arm/FEnvImpl.h b/libc/src/__support/FPUtil/arm/FEnvImpl.h
index ddb0edcf82780..78fbda4f7afff 100644
--- a/libc/src/__support/FPUtil/arm/FEnvImpl.h
+++ b/libc/src/__support/FPUtil/arm/FEnvImpl.h
@@ -10,9 +10,9 @@
 #define LLVM_LIBC_SRC___SUPPORT_FPUTIL_ARM_FENVIMPL_H
 
 #include "hdr/fenv_macros.h"
+#include "hdr/types/fenv_t.h"
 #include "src/__support/FPUtil/FPBits.h"
 #include "src/__support/macros/attributes.h" // For LIBC_INLINE
-#include <fenv.h>
 #include <stdint.h>
 
 namespace LIBC_NAMESPACE {
diff --git a/libc/src/__support/FPUtil/riscv/FEnvImpl.h b/libc/src/__support/FPUtil/riscv/FEnvImpl.h
index a5224330f339a..6e940453f7a94 100644
--- a/libc/src/__support/FPUtil/riscv/FEnvImpl.h
+++ b/libc/src/__support/FPUtil/riscv/FEnvImpl.h
@@ -10,6 +10,7 @@
 #define LLVM_LIBC_SRC___SUPPORT_FPUTIL_RISCV_FENVIMPL_H
 
 #include "hdr/fenv_macros.h"
+#include "hdr/types/fenv_t.h"
 #include "src/__support/FPUtil/FPBits.h"
 #include "src/__support/macros/attributes.h" // For LIBC_INLINE_ASM
 #include "src/__support/macros/config.h"     // For LIBC_INLINE
diff --git a/libc/src/__support/FPUtil/x86_64/FEnvImpl.h b/libc/src/__support/FPUtil/x86_64/FEnvImpl.h
index b73b70e907790..0595658d7df32 100644
--- a/libc/src/__support/FPUtil/x86_64/FEnvImpl.h
+++ b/libc/src/__support/FPUtil/x86_64/FEnvImpl.h
@@ -16,9 +16,9 @@
 #error "Invalid include"
 #endif
 
-#include <fenv.h>
 #include <stdint.h>
 
+#include "hdr/types/fenv_t.h"
 #include "src/__support/macros/sanitizer.h"
 
 namespace LIBC_NAMESPACE {
diff --git a/libc/src/fenv/CMakeLists.txt b/libc/src/fenv/CMakeLists.txt
index 5dcf21de04f1a..a28a7ca4c2d82 100644
--- a/libc/src/fenv/CMakeLists.txt
+++ b/libc/src/fenv/CMakeLists.txt
@@ -5,7 +5,6 @@ add_entrypoint_object(
   HDRS
     fegetround.h
   DEPENDS
-    libc.include.fenv
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -71,6 +70,7 @@ add_entrypoint_object(
     fegetenv.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -84,6 +84,7 @@ add_entrypoint_object(
     fesetenv.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -97,6 +98,7 @@ add_entrypoint_object(
     fegetexceptflag.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fexcept_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -123,6 +125,7 @@ add_entrypoint_object(
     fesetexceptflag.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fexcept_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -136,6 +139,7 @@ add_entrypoint_object(
     feholdexcept.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -149,6 +153,7 @@ add_entrypoint_object(
     feupdateenv.h
   DEPENDS
     libc.hdr.fenv_macros
+    libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
diff --git a/libc/src/fenv/fegetenv.h b/libc/src/fenv/fegetenv.h
index 658316482984e..8d33029647402 100644
--- a/libc/src/fenv/fegetenv.h
+++ b/libc/src/fenv/fegetenv.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FEGETENV_H
 #define LLVM_LIBC_SRC_FENV_FEGETENV_H
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/fegetexceptflag.cpp b/libc/src/fenv/fegetexceptflag.cpp
index c6160da7afbde..72f31bf7abd52 100644
--- a/libc/src/fenv/fegetexceptflag.cpp
+++ b/libc/src/fenv/fegetexceptflag.cpp
@@ -7,11 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/fenv/fegetexceptflag.h"
+#include "hdr/types/fexcept_t.h"
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "src/__support/common.h"
 
-#include <fenv.h>
-
 namespace LIBC_NAMESPACE {
 
 LLVM_LIBC_FUNCTION(int, fegetexceptflag, (fexcept_t * flagp, int excepts)) {
diff --git a/libc/src/fenv/feholdexcept.cpp b/libc/src/fenv/feholdexcept.cpp
index f264c5ae251d3..e5ca257e2be5d 100644
--- a/libc/src/fenv/feholdexcept.cpp
+++ b/libc/src/fenv/feholdexcept.cpp
@@ -7,9 +7,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/fenv/feholdexcept.h"
+#include "hdr/types/fenv_t.h"
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "src/__support/common.h"
-#include <fenv.h>
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/feholdexcept.h b/libc/src/fenv/feholdexcept.h
index bbefc4ecbd41b..e53619e12feaf 100644
--- a/libc/src/fenv/feholdexcept.h
+++ b/libc/src/fenv/feholdexcept.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FEHOLDEXCEPT_H
 #define LLVM_LIBC_SRC_FENV_FEHOLDEXCEPT_H
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/fesetenv.h b/libc/src/fenv/fesetenv.h
index 8b56bebc2e36d..8636711cb1fee 100644
--- a/libc/src/fenv/fesetenv.h
+++ b/libc/src/fenv/fesetenv.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FESETENV_H
 #define LLVM_LIBC_SRC_FENV_FESETENV_H
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/fesetexceptflag.cpp b/libc/src/fenv/fesetexceptflag.cpp
index 3ff8e270dc0a7..628f33dcb9c4b 100644
--- a/libc/src/fenv/fesetexceptflag.cpp
+++ b/libc/src/fenv/fesetexceptflag.cpp
@@ -7,9 +7,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/fenv/fesetexceptflag.h"
+#include "hdr/types/fexcept_t.h"
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "src/__support/common.h"
-#include <fenv.h>
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/feupdateenv.h b/libc/src/fenv/feupdateenv.h
index 294c041ddeae7..3c9c88b36e728 100644
--- a/libc/src/fenv/feupdateenv.h
+++ b/libc/src/fenv/feupdateenv.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FEUPDATEENV_H
 #define LLVM_LIBC_SRC_FENV_FEUPDATEENV_H
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/test/UnitTest/CMakeLists.txt b/libc/test/UnitTest/CMakeLists.txt
index 4411170502ed6..9113eca388e05 100644
--- a/libc/test/UnitTest/CMakeLists.txt
+++ b/libc/test/UnitTest/CMakeLists.txt
@@ -135,6 +135,7 @@ add_unittest_framework_library(
     LibcTest
     libc.src.__support.FPUtil.fp_bits
     libc.src.__support.FPUtil.fenv_impl
+    libc.hdr.types.fenv_t
 )
 
 add_unittest_framework_library(
diff --git a/libc/test/UnitTest/FPExceptMatcher.cpp b/libc/test/UnitTest/FPExceptMatcher.cpp
index 1601b7e53f2be..53ea72ad9ddd8 100644
--- a/libc/test/UnitTest/FPExceptMatcher.cpp
+++ b/libc/test/UnitTest/FPExceptMatcher.cpp
@@ -8,7 +8,8 @@
 
 #include "FPExceptMatcher.h"
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
 #include <memory>
 #include <setjmp.h>
 #include <signal.h>
@@ -36,12 +37,12 @@ FPExceptMatcher::FPExceptMatcher(FunctionCaller *func) {
 
   caughtExcept = false;
   fenv_t oldEnv;
-  fegetenv(&oldEnv);
+  fputil::get_env(&oldEnv);
   if (sigsetjmp(jumpBuffer, 1) == 0)
     funcUP->call();
   // We restore the previous floating point environment after
   // the call to the function which can potentially raise SIGFPE.
-  fesetenv(&oldEnv);
+  fputil::set_env(&oldEnv);
   signal(SIGFPE, oldSIGFPEHandler);
   exceptionRaised = caughtExcept;
 }
diff --git a/libc/test/src/fenv/exception_flags_test.cpp b/libc/test/src/fenv/exception_flags_test.cpp
index 434adc06b1a36..d1d8bfcc53db5 100644
--- a/libc/test/src/fenv/exception_flags_test.cpp
+++ b/libc/test/src/fenv/exception_flags_test.cpp
@@ -6,14 +6,13 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "hdr/types/fexcept_t.h"
 #include "src/fenv/fegetexceptflag.h"
 #include "src/fenv/fesetexceptflag.h"
 
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "test/UnitTest/Test.h"
 
-#include <fenv.h>
-
 TEST(LlvmLibcFenvTest, GetExceptFlagAndSetExceptFlag) {
   // We will disable all exceptions to prevent invocation of the exception
   // handler.
diff --git a/libc/test/src/fenv/feholdexcept_test.cpp b/libc/test/src/fenv/feholdexcept_test.cpp
index 735c7705ff49c..0689d89ab233a 100644
--- a/libc/test/src/fenv/feholdexcept_test.cpp
+++ b/libc/test/src/fenv/feholdexcept_test.cpp
@@ -6,6 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "hdr/types/fenv_t.h"
 #include "src/fenv/feholdexcept.h"
 
 #include "src/__support/FPUtil/FEnvImpl.h"
@@ -13,8 +14,6 @@
 #include "test/UnitTest/FPExceptMatcher.h"
 #include "test/UnitTest/Test.h"
 
-#include <fenv.h>
-
 TEST(LlvmLibcFEnvTest, RaiseAndCrash) {
 #if defined(LIBC_TARGET_ARCH_IS_ANY_ARM) ||                                    \
     defined(LIBC_TARGET_ARCH_IS_ANY_RISCV)
diff --git a/libc/test/src/fenv/feupdateenv_test.cpp b/libc/test/src/fenv/feupdateenv_test.cpp
index 96f253f76077a..251b8566aac3d 100644
--- a/libc/test/src/fenv/feupdateenv_test.cpp
+++ b/libc/test/src/fenv/feupdateenv_test.cpp
@@ -6,12 +6,12 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "hdr/types/fenv_t.h"
 #include "src/fenv/feupdateenv.h"
 
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "test/UnitTest/Test.h"
 
-#include <fenv.h>
 #include <signal.h>
 
 TEST(LlvmLibcFEnvTest, UpdateEnvTest) {
diff --git a/libc/test/src/fenv/getenv_and_setenv_test.cpp b/libc/test/src/fenv/getenv_and_setenv_test.cpp
index 8184a5c3bb999..f767e8ab9b2fb 100644
--- a/libc/test/src/fenv/getenv_and_setenv_test.cpp
+++ b/libc/test/src/fenv/getenv_and_setenv_test.cpp
@@ -6,6 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "hdr/types/fenv_t.h"
 #include "src/fenv/fegetenv.h"
 #include "src/fenv/fegetround.h"
 #include "src/fenv/fesetenv.h"
@@ -14,8 +15,6 @@
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "test/UnitTest/Test.h"
 
-#include <fenv.h>
-
 TEST(LlvmLibcFenvTest, GetEnvAndSetEnv) {
   // We will disable all exceptions to prevent invocation of the exception
   // handler.
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index 9270a9d75b01c..fb37f113b310a 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -130,6 +130,16 @@ libc_support_library(
 
 ############################ Type Proxy Header Files ###########################
 
+libc_support_library(
+    name = "types_fenv_t",
+    hdrs = ["hdr/types/fenv_t.h"],
+)
+
+libc_support_library(
+    name = "types_fexcept_t",
+    hdrs = ["hdr/types/fexcept_t.h"],
+)
+
 libc_support_library(
     name = "types_sigset_t",
     hdrs = ["hdr/types/sigset_t.h"],
@@ -776,6 +786,7 @@ libc_support_library(
         ":errno",
         ":hdr_fenv_macros",
         ":hdr_math_macros",
+	":types_fenv_t"
     ],
 )
 
@@ -1251,6 +1262,7 @@ libc_function(
     deps = [
         ":__support_common",
         ":__support_fputil_fenv_impl",
+	":types_fexcept_t"
     ],
 )
 
@@ -1261,6 +1273,7 @@ libc_function(
     deps = [
         ":__support_common",
         ":__support_fputil_fenv_impl",
+	":types_fexcept_t",
     ],
 )
 
diff --git a/utils/bazel/llvm-project-overlay/libc/test/UnitTest/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/UnitTest/BUILD.bazel
index 82c015a7eeda0..b1140a5a16090 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/UnitTest/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/UnitTest/BUILD.bazel
@@ -87,6 +87,7 @@ libc_support_library(
         "//libc:__support_fputil_rounding_mode",
         "//libc:hdr_math_macros",
 	"//libc:hdr_fenv_macros",
+	"//libc:types_fenv_t",
     ],
 )
 
diff --git a/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
index bce1dd786a850..359db0723dfd3 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
@@ -65,6 +65,7 @@ libc_test(
         "//libc:__support_fputil_fenv_impl",
         "//libc:__support_macros_properties_architectures",
         "//libc/test/UnitTest:fp_test_helpers",
+	"//libc:types_fenv_t",
     ],
 )
 
@@ -77,6 +78,7 @@ libc_test(
     ],
     deps = [
         "//libc:__support_fputil_fenv_impl",
+	"//libc:types_fexcept_t",
     ],
 )
 
@@ -115,6 +117,7 @@ libc_test(
     ],
     deps = [
         "//libc:__support_fputil_fenv_impl",
+	"//libc:types_fenv_t",
     ],
 )
 
@@ -129,5 +132,6 @@ libc_test(
     ],
     deps = [
         "//libc:__support_fputil_fenv_impl",
+	"//libc:types_fenv_t",
     ],
 )

From 45339568486cafaa414a1e40d73a251f7d15e284 Mon Sep 17 00:00:00 2001
From: Yijia Gu <yijiagu@google.com>
Date: Mon, 15 Apr 2024 11:46:40 -0700
Subject: [PATCH 473/886] [mlir] add polynomial dialect in bazel BUILD

---
 .../llvm-project-overlay/mlir/BUILD.bazel     | 104 ++++++++++++++++++
 1 file changed, 104 insertions(+)

diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 67052fcd39930..522e3efe4e0ab 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -6602,6 +6602,109 @@ gentbl_cc_library(
     deps = [":PDLInterpOpsTdFiles"],
 )
 
+cc_library(
+    name = "PolynomialDialect",
+    srcs = glob([
+        "lib/Dialect/Polynomial/IR/*.cpp",
+    ]),
+    hdrs = glob([
+        "include/mlir/Dialect/Polynomial/IR/*.h",
+    ]),
+    includes = ["include"],
+    deps = [
+        ":IR",
+        ":InferTypeOpInterface",
+        ":PolynomialAttributesIncGen",
+        ":PolynomialIncGen",
+        ":Support",
+        "//third_party/llvm/llvm-project/llvm:Support",
+    ],
+)
+
+td_library(
+    name = "PolynomialTdFiles",
+    srcs = ["include/mlir/Dialect/Polynomial/IR/Polynomial.td"],
+    includes = ["include"],
+    deps = [
+        ":BuiltinDialectTdFiles",
+        ":InferTypeOpInterfaceTdFiles",
+        ":OpBaseTdFiles",
+        ":SideEffectInterfacesTdFiles",
+    ],
+)
+
+gentbl_cc_library(
+    name = "PolynomialIncGen",
+    tbl_outs = [
+        (
+            ["-gen-op-decls"],
+            "include/mlir/Dialect/Polynomial/IR/Polynomial.h.inc",
+        ),
+        (
+            ["-gen-op-defs"],
+            "include/mlir/Dialect/Polynomial/IR/Polynomial.cpp.inc",
+        ),
+        (
+            [
+                "-gen-dialect-decls",
+                "-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialDialect.h.inc",
+        ),
+        (
+            [
+                "-gen-dialect-defs",
+                "-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialDialect.cpp.inc",
+        ),
+        (
+            [
+                "--gen-typedef-decls",
+                "-typedefs-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialTypes.h.inc",
+        ),
+        (
+            [
+                "--gen-typedef-defs",
+                "-typedefs-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialTypes.cpp.inc",
+        ),
+        (
+            ["-gen-op-doc"],
+            "g3doc/Dialects/Polynomial/Polynomial.md",
+        ),
+    ],
+    tblgen = ":mlir-tblgen",
+    td_file = "include/mlir/Dialect/Polynomial/IR/Polynomial.td",
+    deps = [":PolynomialTdFiles"],
+)
+
+gentbl_cc_library(
+    name = "PolynomialAttributesIncGen",
+    tbl_outs = [
+        (
+            [
+                "-gen-attrdef-decls",
+                "-attrdefs-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.h.inc",
+        ),
+        (
+            [
+                "-gen-attrdef-defs",
+                "-attrdefs-dialect=polynomial",
+            ],
+            "include/mlir/Dialect/Polynomial/IR/PolynomialAttributes.cpp.inc",
+        ),
+    ],
+    tblgen = ":mlir-tblgen",
+    td_file = "include/mlir/Dialect/Polynomial/IR/Polynomial.td",
+    deps = [":PolynomialTdFiles"],
+)
+
 td_library(
     name = "SPIRVOpsTdFiles",
     srcs = glob(["include/mlir/Dialect/SPIRV/IR/*.td"]),
@@ -9076,6 +9179,7 @@ cc_library(
         ":PDLDialect",
         ":PDLInterpDialect",
         ":PDLToPDLInterp",
+        ":PolynomialDialect",
         ":QuantOps",
         ":ROCDLDialect",
         ":ROCDLTarget",

From f9d761df50d12147fc7caa19927be81816d60e2d Mon Sep 17 00:00:00 2001
From: Justin Fargnoli <justinfargnoli@gmail.com>
Date: Mon, 15 Apr 2024 11:53:15 -0700
Subject: [PATCH 474/886] [tools] Fix `LLVM_USE_SANITIZE_COVERAGE` build
 (#87722)

---
 llvm/tools/llvm-dis-fuzzer/CMakeLists.txt                   | 2 ++
 llvm/tools/vfabi-demangle-fuzzer/vfabi-demangler-fuzzer.cpp | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/llvm/tools/llvm-dis-fuzzer/CMakeLists.txt b/llvm/tools/llvm-dis-fuzzer/CMakeLists.txt
index 4b0b80e2f7582..b424a14374f37 100644
--- a/llvm/tools/llvm-dis-fuzzer/CMakeLists.txt
+++ b/llvm/tools/llvm-dis-fuzzer/CMakeLists.txt
@@ -1,5 +1,7 @@
 set(LLVM_LINK_COMPONENTS
   BitReader
+  Core
+  Support
 )
 add_llvm_fuzzer(llvm-dis-fuzzer
   llvm-dis-fuzzer.cpp
diff --git a/llvm/tools/vfabi-demangle-fuzzer/vfabi-demangler-fuzzer.cpp b/llvm/tools/vfabi-demangle-fuzzer/vfabi-demangler-fuzzer.cpp
index e1041bd4eef99..309fe2b9957a4 100644
--- a/llvm/tools/vfabi-demangle-fuzzer/vfabi-demangler-fuzzer.cpp
+++ b/llvm/tools/vfabi-demangle-fuzzer/vfabi-demangler-fuzzer.cpp
@@ -11,7 +11,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/AsmParser/Parser.h"
+#include "llvm/IR/Module.h"
 #include "llvm/IR/VFABIDemangler.h"
+#include "llvm/Support/SourceMgr.h"
 
 using namespace llvm;
 

From c31a8104d1173d79f8b71518829046b441d98d59 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 12:01:41 -0700
Subject: [PATCH 475/886] [RISCV] Add -mattr=+v to intrinsic-cttz-elts.ll. NFC

---
 .../test/CodeGen/RISCV/intrinsic-cttz-elts.ll | 108 ++++++------------
 1 file changed, 38 insertions(+), 70 deletions(-)

diff --git a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
index 15abc9b75883c..49d4760a2e9ab 100644
--- a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
+++ b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
@@ -1,38 +1,22 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
-; RUN: llc -mtriple=riscv32 < %s | FileCheck %s -check-prefix=RV32
-; RUN: llc -mtriple=riscv64 < %s | FileCheck %s -check-prefix=RV64
+; RUN: llc -mtriple=riscv32 -mattr=+v < %s | FileCheck %s -check-prefix=RV32
+; RUN: llc -mtriple=riscv64 -mattr=+v < %s | FileCheck %s -check-prefix=RV64
 
 ; FIXED WIDTH
 
 define i16 @ctz_v4i32(<4 x i32> %a) {
 ; RV32-LABEL: ctz_v4i32:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    lw a3, 0(a0)
-; RV32-NEXT:    lw a1, 4(a0)
-; RV32-NEXT:    lw a2, 12(a0)
-; RV32-NEXT:    lw a4, 8(a0)
-; RV32-NEXT:    seqz a0, a3
-; RV32-NEXT:    addi a0, a0, -1
-; RV32-NEXT:    andi a0, a0, 4
-; RV32-NEXT:    seqz a3, a4
-; RV32-NEXT:    addi a3, a3, -1
-; RV32-NEXT:    andi a3, a3, 2
-; RV32-NEXT:    bltu a3, a0, .LBB0_2
-; RV32-NEXT:  # %bb.1:
-; RV32-NEXT:    mv a0, a3
-; RV32-NEXT:  .LBB0_2:
-; RV32-NEXT:    snez a2, a2
-; RV32-NEXT:    seqz a1, a1
-; RV32-NEXT:    addi a1, a1, -1
-; RV32-NEXT:    andi a1, a1, 3
-; RV32-NEXT:    bltu a2, a1, .LBB0_4
-; RV32-NEXT:  # %bb.3:
-; RV32-NEXT:    mv a1, a2
-; RV32-NEXT:  .LBB0_4:
-; RV32-NEXT:    bltu a1, a0, .LBB0_6
-; RV32-NEXT:  # %bb.5:
-; RV32-NEXT:    mv a0, a1
-; RV32-NEXT:  .LBB0_6:
+; RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
+; RV32-NEXT:    vmsne.vi v0, v8, 0
+; RV32-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
+; RV32-NEXT:    vmv.v.i v8, 0
+; RV32-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV32-NEXT:    vid.v v9
+; RV32-NEXT:    vrsub.vi v9, v9, 4
+; RV32-NEXT:    vand.vv v8, v8, v9
+; RV32-NEXT:    vredmaxu.vs v8, v8, v8
+; RV32-NEXT:    vmv.x.s a0, v8
 ; RV32-NEXT:    li a1, 4
 ; RV32-NEXT:    sub a1, a1, a0
 ; RV32-NEXT:    andi a0, a1, 255
@@ -40,32 +24,16 @@ define i16 @ctz_v4i32(<4 x i32> %a) {
 ;
 ; RV64-LABEL: ctz_v4i32:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    lw a3, 0(a0)
-; RV64-NEXT:    lw a1, 8(a0)
-; RV64-NEXT:    lw a2, 24(a0)
-; RV64-NEXT:    lw a4, 16(a0)
-; RV64-NEXT:    seqz a0, a3
-; RV64-NEXT:    addi a0, a0, -1
-; RV64-NEXT:    andi a0, a0, 4
-; RV64-NEXT:    seqz a3, a4
-; RV64-NEXT:    addi a3, a3, -1
-; RV64-NEXT:    andi a3, a3, 2
-; RV64-NEXT:    bltu a3, a0, .LBB0_2
-; RV64-NEXT:  # %bb.1:
-; RV64-NEXT:    mv a0, a3
-; RV64-NEXT:  .LBB0_2:
-; RV64-NEXT:    snez a2, a2
-; RV64-NEXT:    seqz a1, a1
-; RV64-NEXT:    addi a1, a1, -1
-; RV64-NEXT:    andi a1, a1, 3
-; RV64-NEXT:    bltu a2, a1, .LBB0_4
-; RV64-NEXT:  # %bb.3:
-; RV64-NEXT:    mv a1, a2
-; RV64-NEXT:  .LBB0_4:
-; RV64-NEXT:    bltu a1, a0, .LBB0_6
-; RV64-NEXT:  # %bb.5:
-; RV64-NEXT:    mv a0, a1
-; RV64-NEXT:  .LBB0_6:
+; RV64-NEXT:    vsetivli zero, 4, e32, m1, ta, ma
+; RV64-NEXT:    vmsne.vi v0, v8, 0
+; RV64-NEXT:    vsetvli zero, zero, e8, mf4, ta, ma
+; RV64-NEXT:    vmv.v.i v8, 0
+; RV64-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV64-NEXT:    vid.v v9
+; RV64-NEXT:    vrsub.vi v9, v9, 4
+; RV64-NEXT:    vand.vv v8, v8, v9
+; RV64-NEXT:    vredmaxu.vs v8, v8, v8
+; RV64-NEXT:    vmv.x.s a0, v8
 ; RV64-NEXT:    li a1, 4
 ; RV64-NEXT:    subw a1, a1, a0
 ; RV64-NEXT:    andi a0, a1, 255
@@ -79,14 +47,14 @@ define i16 @ctz_v4i32(<4 x i32> %a) {
 define i32 @ctz_v2i1_poison(<2 x i1> %a) {
 ; RV32-LABEL: ctz_v2i1_poison:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    andi a1, a1, 1
-; RV32-NEXT:    slli a0, a0, 31
-; RV32-NEXT:    srai a0, a0, 31
-; RV32-NEXT:    andi a0, a0, 2
-; RV32-NEXT:    bltu a1, a0, .LBB1_2
-; RV32-NEXT:  # %bb.1:
-; RV32-NEXT:    mv a0, a1
-; RV32-NEXT:  .LBB1_2:
+; RV32-NEXT:    vsetivli zero, 2, e8, mf8, ta, ma
+; RV32-NEXT:    vmv.v.i v8, 0
+; RV32-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV32-NEXT:    vid.v v9
+; RV32-NEXT:    vrsub.vi v9, v9, 2
+; RV32-NEXT:    vand.vv v8, v8, v9
+; RV32-NEXT:    vredmaxu.vs v8, v8, v8
+; RV32-NEXT:    vmv.x.s a0, v8
 ; RV32-NEXT:    li a1, 2
 ; RV32-NEXT:    sub a1, a1, a0
 ; RV32-NEXT:    andi a0, a1, 255
@@ -94,14 +62,14 @@ define i32 @ctz_v2i1_poison(<2 x i1> %a) {
 ;
 ; RV64-LABEL: ctz_v2i1_poison:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    andi a1, a1, 1
-; RV64-NEXT:    slli a0, a0, 63
-; RV64-NEXT:    srai a0, a0, 63
-; RV64-NEXT:    andi a0, a0, 2
-; RV64-NEXT:    bltu a1, a0, .LBB1_2
-; RV64-NEXT:  # %bb.1:
-; RV64-NEXT:    mv a0, a1
-; RV64-NEXT:  .LBB1_2:
+; RV64-NEXT:    vsetivli zero, 2, e8, mf8, ta, ma
+; RV64-NEXT:    vmv.v.i v8, 0
+; RV64-NEXT:    vmerge.vim v8, v8, -1, v0
+; RV64-NEXT:    vid.v v9
+; RV64-NEXT:    vrsub.vi v9, v9, 2
+; RV64-NEXT:    vand.vv v8, v8, v9
+; RV64-NEXT:    vredmaxu.vs v8, v8, v8
+; RV64-NEXT:    vmv.x.s a0, v8
 ; RV64-NEXT:    li a1, 2
 ; RV64-NEXT:    subw a1, a1, a0
 ; RV64-NEXT:    andi a0, a1, 255

From b153c05cba9be7f009b8ad8413c5840baf7d278c Mon Sep 17 00:00:00 2001
From: Ivan Butygin <ivan.butygin@gmail.com>
Date: Mon, 15 Apr 2024 22:16:59 +0300
Subject: [PATCH 476/886] [mlir][scf] Uplift `scf.while` to `scf.for` (#76108)

Add uplifting from `scf.while` to `scf.for`.

This uplifting expects a very specific ops pattern:
* `before` block consisting of single `arith.cmp` op
* `after` block containing `arith.addi`

We also have a set of patterns to cleanup `scf.while` loops to get them
close to the desired form, they will be added in separate PRs.

This is part of upstreaming `numba-mlir` scf uplifting pipeline: `cf ->
scf.while -> scf.for -> scf.parallel`

Original code:
https://github.com/numba/numba-mlir/blob/main/mlir/lib/Transforms/PromoteToParallel.cpp
---
 .../mlir/Dialect/SCF/Transforms/Patterns.h    |   6 +
 .../mlir/Dialect/SCF/Transforms/Transforms.h  |   6 +
 .../lib/Dialect/SCF/Transforms/CMakeLists.txt |   1 +
 .../SCF/Transforms/UpliftWhileToFor.cpp       | 222 ++++++++++++++++++
 mlir/test/Dialect/SCF/uplift-while.mlir       | 157 +++++++++++++
 mlir/test/lib/Dialect/SCF/CMakeLists.txt      |   1 +
 .../lib/Dialect/SCF/TestUpliftWhileToFor.cpp  |  50 ++++
 mlir/tools/mlir-opt/mlir-opt.cpp              |   2 +
 8 files changed, 445 insertions(+)
 create mode 100644 mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
 create mode 100644 mlir/test/Dialect/SCF/uplift-while.mlir
 create mode 100644 mlir/test/lib/Dialect/SCF/TestUpliftWhileToFor.cpp

diff --git a/mlir/include/mlir/Dialect/SCF/Transforms/Patterns.h b/mlir/include/mlir/Dialect/SCF/Transforms/Patterns.h
index 5c0d5643c0198..fdf2570626980 100644
--- a/mlir/include/mlir/Dialect/SCF/Transforms/Patterns.h
+++ b/mlir/include/mlir/Dialect/SCF/Transforms/Patterns.h
@@ -79,6 +79,12 @@ void populateSCFLoopPipeliningPatterns(RewritePatternSet &patterns,
 /// loop bounds and loop steps are canonicalized.
 void populateSCFForLoopCanonicalizationPatterns(RewritePatternSet &patterns);
 
+/// Populate patterns to uplift `scf.while` ops to `scf.for`.
+/// Uplifitng expects a specific ops pattern:
+///  * `before` block consisting of single arith.cmp op
+///  * `after` block containing arith.addi
+void populateUpliftWhileToForPatterns(RewritePatternSet &patterns);
+
 } // namespace scf
 } // namespace mlir
 
diff --git a/mlir/include/mlir/Dialect/SCF/Transforms/Transforms.h b/mlir/include/mlir/Dialect/SCF/Transforms/Transforms.h
index 690cd146c606e..220dcb35571d2 100644
--- a/mlir/include/mlir/Dialect/SCF/Transforms/Transforms.h
+++ b/mlir/include/mlir/Dialect/SCF/Transforms/Transforms.h
@@ -222,6 +222,12 @@ FailureOr<WhileOp> wrapWhileLoopInZeroTripCheck(WhileOp whileOp,
                                                 RewriterBase &rewriter,
                                                 bool forceCreateCheck = false);
 
+/// Try to uplift `scf.while` op to `scf.for`.
+/// Uplifitng expects a specific ops pattern:
+///  * `before` block consisting of single arith.cmp op
+///  * `after` block containing arith.addi
+FailureOr<ForOp> upliftWhileToForLoop(RewriterBase &rewriter, WhileOp loop);
+
 } // namespace scf
 } // namespace mlir
 
diff --git a/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt b/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
index e5494205e086a..a2925aef17ca7 100644
--- a/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
@@ -14,6 +14,7 @@ add_mlir_dialect_library(MLIRSCFTransforms
   StructuralTypeConversions.cpp
   TileUsingInterface.cpp
   WrapInZeroTripCheck.cpp
+  UpliftWhileToFor.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/SCF
diff --git a/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp b/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
new file mode 100644
index 0000000000000..fea2f659535bb
--- /dev/null
+++ b/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
@@ -0,0 +1,222 @@
+//===- UpliftWhileToFor.cpp - scf.while to scf.for loop uplifting ---------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Transforms SCF.WhileOp's into SCF.ForOp's.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/SCF/Transforms/Passes.h"
+
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/SCF/Transforms/Patterns.h"
+#include "mlir/IR/Dominance.h"
+#include "mlir/IR/PatternMatch.h"
+
+using namespace mlir;
+
+namespace {
+struct UpliftWhileOp : public OpRewritePattern<scf::WhileOp> {
+  using OpRewritePattern::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(scf::WhileOp loop,
+                                PatternRewriter &rewriter) const override {
+    return upliftWhileToForLoop(rewriter, loop);
+  }
+};
+} // namespace
+
+FailureOr<scf::ForOp> mlir::scf::upliftWhileToForLoop(RewriterBase &rewriter,
+                                                      scf::WhileOp loop) {
+  Block *beforeBody = loop.getBeforeBody();
+  if (!llvm::hasSingleElement(beforeBody->without_terminator()))
+    return rewriter.notifyMatchFailure(loop, "Loop body must have single op");
+
+  auto cmp = dyn_cast<arith::CmpIOp>(beforeBody->front());
+  if (!cmp)
+    return rewriter.notifyMatchFailure(loop,
+                                       "Loop body must have single cmp op");
+
+  scf::ConditionOp beforeTerm = loop.getConditionOp();
+  if (!cmp->hasOneUse() || beforeTerm.getCondition() != cmp.getResult())
+    return rewriter.notifyMatchFailure(loop, [&](Diagnostic &diag) {
+      diag << "Expected single condition use: " << *cmp;
+    });
+
+  // All `before` block args must be directly forwarded to ConditionOp.
+  // They will be converted to `scf.for` `iter_vars` except induction var.
+  if (ValueRange(beforeBody->getArguments()) != beforeTerm.getArgs())
+    return rewriter.notifyMatchFailure(loop, "Invalid args order");
+
+  using Pred = arith::CmpIPredicate;
+  Pred predicate = cmp.getPredicate();
+  if (predicate != Pred::slt && predicate != Pred::sgt)
+    return rewriter.notifyMatchFailure(loop, [&](Diagnostic &diag) {
+      diag << "Expected 'slt' or 'sgt' predicate: " << *cmp;
+    });
+
+  BlockArgument inductionVar;
+  Value ub;
+  DominanceInfo dom;
+
+  // Check if cmp has a suitable form. One of the arguments must be a `before`
+  // block arg, other must be defined outside `scf.while` and will be treated
+  // as upper bound.
+  for (bool reverse : {false, true}) {
+    auto expectedPred = reverse ? Pred::sgt : Pred::slt;
+    if (cmp.getPredicate() != expectedPred)
+      continue;
+
+    auto arg1 = reverse ? cmp.getRhs() : cmp.getLhs();
+    auto arg2 = reverse ? cmp.getLhs() : cmp.getRhs();
+
+    auto blockArg = dyn_cast<BlockArgument>(arg1);
+    if (!blockArg || blockArg.getOwner() != beforeBody)
+      continue;
+
+    if (!dom.properlyDominates(arg2, loop))
+      continue;
+
+    inductionVar = blockArg;
+    ub = arg2;
+    break;
+  }
+
+  if (!inductionVar)
+    return rewriter.notifyMatchFailure(loop, [&](Diagnostic &diag) {
+      diag << "Unrecognized cmp form: " << *cmp;
+    });
+
+  // inductionVar must have 2 uses: one is in `cmp` and other is `condition`
+  // arg.
+  if (!llvm::hasNItems(inductionVar.getUses(), 2))
+    return rewriter.notifyMatchFailure(loop, [&](Diagnostic &diag) {
+      diag << "Unrecognized induction var: " << inductionVar;
+    });
+
+  Block *afterBody = loop.getAfterBody();
+  scf::YieldOp afterTerm = loop.getYieldOp();
+  auto argNumber = inductionVar.getArgNumber();
+  auto afterTermIndArg = afterTerm.getResults()[argNumber];
+
+  auto inductionVarAfter = afterBody->getArgument(argNumber);
+
+  Value step;
+
+  // Find suitable `addi` op inside `after` block, one of the args must be an
+  // Induction var passed from `before` block and second arg must be defined
+  // outside of the loop and will be considered step value.
+  // TODO: Add `subi` support?
+  for (auto &use : inductionVarAfter.getUses()) {
+    auto owner = dyn_cast<arith::AddIOp>(use.getOwner());
+    if (!owner)
+      continue;
+
+    auto other =
+        (inductionVarAfter == owner.getLhs() ? owner.getRhs() : owner.getLhs());
+    if (!dom.properlyDominates(other, loop))
+      continue;
+
+    if (afterTermIndArg != owner.getResult())
+      continue;
+
+    step = other;
+    break;
+  }
+
+  if (!step)
+    return rewriter.notifyMatchFailure(loop, "Didn't found suitable 'addi' op");
+
+  auto lb = loop.getInits()[argNumber];
+
+  assert(lb.getType().isIntOrIndex());
+  assert(lb.getType() == ub.getType());
+  assert(lb.getType() == step.getType());
+
+  llvm::SmallVector<Value> newArgs;
+
+  // Populate inits for new `scf.for`, skip induction var.
+  newArgs.reserve(loop.getInits().size());
+  for (auto &&[i, init] : llvm::enumerate(loop.getInits())) {
+    if (i == argNumber)
+      continue;
+
+    newArgs.emplace_back(init);
+  }
+
+  Location loc = loop.getLoc();
+
+  // With `builder == nullptr`, ForOp::build will try to insert terminator at
+  // the end of newly created block and we don't want it. Provide empty
+  // dummy builder instead.
+  auto emptyBuilder = [](OpBuilder &, Location, Value, ValueRange) {};
+  auto newLoop =
+      rewriter.create<scf::ForOp>(loc, lb, ub, step, newArgs, emptyBuilder);
+
+  Block *newBody = newLoop.getBody();
+
+  // Populate block args for `scf.for` body, move induction var to the front.
+  newArgs.clear();
+  ValueRange newBodyArgs = newBody->getArguments();
+  for (auto i : llvm::seq<size_t>(0, newBodyArgs.size())) {
+    if (i < argNumber) {
+      newArgs.emplace_back(newBodyArgs[i + 1]);
+    } else if (i == argNumber) {
+      newArgs.emplace_back(newBodyArgs.front());
+    } else {
+      newArgs.emplace_back(newBodyArgs[i]);
+    }
+  }
+
+  rewriter.inlineBlockBefore(loop.getAfterBody(), newBody, newBody->end(),
+                             newArgs);
+
+  auto term = cast<scf::YieldOp>(newBody->getTerminator());
+
+  // Populate new yield args, skipping the induction var.
+  newArgs.clear();
+  for (auto &&[i, arg] : llvm::enumerate(term.getResults())) {
+    if (i == argNumber)
+      continue;
+
+    newArgs.emplace_back(arg);
+  }
+
+  OpBuilder::InsertionGuard g(rewriter);
+  rewriter.setInsertionPoint(term);
+  rewriter.replaceOpWithNewOp<scf::YieldOp>(term, newArgs);
+
+  // Compute induction var value after loop execution.
+  rewriter.setInsertionPointAfter(newLoop);
+  Value one;
+  if (isa<IndexType>(step.getType())) {
+    one = rewriter.create<arith::ConstantIndexOp>(loc, 1);
+  } else {
+    one = rewriter.create<arith::ConstantIntOp>(loc, 1, step.getType());
+  }
+
+  Value stepDec = rewriter.create<arith::SubIOp>(loc, step, one);
+  Value len = rewriter.create<arith::SubIOp>(loc, ub, lb);
+  len = rewriter.create<arith::AddIOp>(loc, len, stepDec);
+  len = rewriter.create<arith::DivSIOp>(loc, len, step);
+  len = rewriter.create<arith::SubIOp>(loc, len, one);
+  Value res = rewriter.create<arith::MulIOp>(loc, len, step);
+  res = rewriter.create<arith::AddIOp>(loc, lb, res);
+
+  // Reconstruct `scf.while` results, inserting final induction var value
+  // into proper place.
+  newArgs.clear();
+  llvm::append_range(newArgs, newLoop.getResults());
+  newArgs.insert(newArgs.begin() + argNumber, res);
+  rewriter.replaceOp(loop, newArgs);
+  return newLoop;
+}
+
+void mlir::scf::populateUpliftWhileToForPatterns(RewritePatternSet &patterns) {
+  patterns.add<UpliftWhileOp>(patterns.getContext());
+}
diff --git a/mlir/test/Dialect/SCF/uplift-while.mlir b/mlir/test/Dialect/SCF/uplift-while.mlir
new file mode 100644
index 0000000000000..25ea6142a332d
--- /dev/null
+++ b/mlir/test/Dialect/SCF/uplift-while.mlir
@@ -0,0 +1,157 @@
+// RUN: mlir-opt %s -pass-pipeline='builtin.module(func.func(test-scf-uplift-while-to-for))' -split-input-file -allow-unregistered-dialect | FileCheck %s
+
+func.func @uplift_while(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %0 = scf.while (%arg3 = %arg0) : (index) -> (index) {
+    %1 = arith.cmpi slt, %arg3, %arg1 : index
+    scf.condition(%1) %arg3 : index
+  } do {
+  ^bb0(%arg3: index):
+    "test.test1"(%arg3) : (index) -> ()
+    %added = arith.addi %arg3, %arg2 : index
+    "test.test2"(%added) : (index) -> ()
+    scf.yield %added : index
+  }
+  return %0 : index
+}
+
+// CHECK-LABEL: func @uplift_while
+//  CHECK-SAME:     (%[[BEGIN:.*]]: index, %[[END:.*]]: index, %[[STEP:.*]]: index) -> index
+//       CHECK:     %[[C1:.*]] = arith.constant 1 : index
+//       CHECK:     scf.for %[[I:.*]] = %[[BEGIN]] to %[[END]] step %[[STEP]] {
+//       CHECK:     "test.test1"(%[[I]]) : (index) -> ()
+//       CHECK:     %[[INC:.*]] = arith.addi %[[I]], %[[STEP]] : index
+//       CHECK:     "test.test2"(%[[INC]]) : (index) -> ()
+//       CHECK:     %[[R1:.*]] = arith.subi %[[STEP]], %[[C1]] : index
+//       CHECK:     %[[R2:.*]] = arith.subi %[[END]], %[[BEGIN]] : index
+//       CHECK:     %[[R3:.*]] = arith.addi %[[R2]], %[[R1]] : index
+//       CHECK:     %[[R4:.*]] = arith.divsi %[[R3]], %[[STEP]] : index
+//       CHECK:     %[[R5:.*]] = arith.subi %[[R4]], %[[C1]] : index
+//       CHECK:     %[[R6:.*]] = arith.muli %[[R5]], %[[STEP]] : index
+//       CHECK:     %[[R7:.*]] = arith.addi %[[BEGIN]], %[[R6]] : index
+//       CHECK:     return %[[R7]] : index
+
+// -----
+
+func.func @uplift_while(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %0 = scf.while (%arg3 = %arg0) : (index) -> (index) {
+    %1 = arith.cmpi sgt, %arg1, %arg3 : index
+    scf.condition(%1) %arg3 : index
+  } do {
+  ^bb0(%arg3: index):
+    "test.test1"(%arg3) : (index) -> ()
+    %added = arith.addi %arg3, %arg2 : index
+    "test.test2"(%added) : (index) -> ()
+    scf.yield %added : index
+  }
+  return %0 : index
+}
+
+// CHECK-LABEL: func @uplift_while
+//  CHECK-SAME:     (%[[BEGIN:.*]]: index, %[[END:.*]]: index, %[[STEP:.*]]: index) -> index
+//       CHECK:     %[[C1:.*]] = arith.constant 1 : index
+//       CHECK:     scf.for %[[I:.*]] = %[[BEGIN]] to %[[END]] step %[[STEP]] {
+//       CHECK:     "test.test1"(%[[I]]) : (index) -> ()
+//       CHECK:     %[[INC:.*]] = arith.addi %[[I]], %[[STEP]] : index
+//       CHECK:     "test.test2"(%[[INC]]) : (index) -> ()
+//       CHECK:     %[[R1:.*]] = arith.subi %[[STEP]], %[[C1]] : index
+//       CHECK:     %[[R2:.*]] = arith.subi %[[END]], %[[BEGIN]] : index
+//       CHECK:     %[[R3:.*]] = arith.addi %[[R2]], %[[R1]] : index
+//       CHECK:     %[[R4:.*]] = arith.divsi %[[R3]], %[[STEP]] : index
+//       CHECK:     %[[R5:.*]] = arith.subi %[[R4]], %[[C1]] : index
+//       CHECK:     %[[R6:.*]] = arith.muli %[[R5]], %[[STEP]] : index
+//       CHECK:     %[[R7:.*]] = arith.addi %[[BEGIN]], %[[R6]] : index
+//       CHECK:     return %[[R7]] : index
+
+// -----
+
+func.func @uplift_while(%arg0: index, %arg1: index, %arg2: index) -> index {
+  %0 = scf.while (%arg3 = %arg0) : (index) -> (index) {
+    %1 = arith.cmpi slt, %arg3, %arg1 : index
+    scf.condition(%1) %arg3 : index
+  } do {
+  ^bb0(%arg3: index):
+    "test.test1"(%arg3) : (index) -> ()
+    %added = arith.addi %arg2, %arg3 : index
+    "test.test2"(%added) : (index) -> ()
+    scf.yield %added : index
+  }
+  return %0 : index
+}
+
+// CHECK-LABEL: func @uplift_while
+//  CHECK-SAME:     (%[[BEGIN:.*]]: index, %[[END:.*]]: index, %[[STEP:.*]]: index) -> index
+//       CHECK:     %[[C1:.*]] = arith.constant 1 : index
+//       CHECK:     scf.for %[[I:.*]] = %[[BEGIN]] to %[[END]] step %[[STEP]] {
+//       CHECK:     "test.test1"(%[[I]]) : (index) -> ()
+//       CHECK:     %[[INC:.*]] = arith.addi %[[STEP]], %[[I]] : index
+//       CHECK:     "test.test2"(%[[INC]]) : (index) -> ()
+//       CHECK:     %[[R1:.*]] = arith.subi %[[STEP]], %[[C1]] : index
+//       CHECK:     %[[R2:.*]] = arith.subi %[[END]], %[[BEGIN]] : index
+//       CHECK:     %[[R3:.*]] = arith.addi %[[R2]], %[[R1]] : index
+//       CHECK:     %[[R4:.*]] = arith.divsi %[[R3]], %[[STEP]] : index
+//       CHECK:     %[[R5:.*]] = arith.subi %[[R4]], %[[C1]] : index
+//       CHECK:     %[[R6:.*]] = arith.muli %[[R5]], %[[STEP]] : index
+//       CHECK:     %[[R7:.*]] = arith.addi %[[BEGIN]], %[[R6]] : index
+//       CHECK:     return %[[R7]] : index
+
+
+// -----
+
+func.func @uplift_while(%arg0: index, %arg1: index, %arg2: index) -> (i32, f32) {
+  %c1 = arith.constant 1 : i32
+  %c2 = arith.constant 2.0 : f32
+  %0:3 = scf.while (%arg4 = %c1, %arg3 = %arg0, %arg5 = %c2) : (i32, index, f32) -> (i32, index, f32) {
+    %1 = arith.cmpi slt, %arg3, %arg1 : index
+    scf.condition(%1) %arg4, %arg3, %arg5 : i32, index, f32
+  } do {
+  ^bb0(%arg4: i32, %arg3: index, %arg5: f32):
+    %1 = "test.test1"(%arg4) : (i32) -> i32
+    %added = arith.addi %arg3, %arg2 : index
+    %2 = "test.test2"(%arg5) : (f32) -> f32
+    scf.yield %1, %added, %2 : i32, index, f32
+  }
+  return %0#0, %0#2 : i32, f32
+}
+
+// CHECK-LABEL: func @uplift_while
+//  CHECK-SAME:     (%[[BEGIN:.*]]: index, %[[END:.*]]: index, %[[STEP:.*]]: index) -> (i32, f32)
+//   CHECK-DAG:     %[[C1:.*]] = arith.constant 1 : i32
+//   CHECK-DAG:     %[[C2:.*]] = arith.constant 2.000000e+00 : f32
+//       CHECK:     %[[RES:.*]]:2 = scf.for %[[I:.*]] = %[[BEGIN]] to %[[END]] step %[[STEP]]
+//  CHECK-SAME:     iter_args(%[[ARG1:.*]] = %[[C1]], %[[ARG2:.*]] = %[[C2]]) -> (i32, f32) {
+//       CHECK:     %[[T1:.*]] = "test.test1"(%[[ARG1]]) : (i32) -> i32
+//       CHECK:     %[[T2:.*]] = "test.test2"(%[[ARG2]]) : (f32) -> f32
+//       CHECK:     scf.yield %[[T1]], %[[T2]] : i32, f32
+//       CHECK:     return %[[RES]]#0, %[[RES]]#1 : i32, f32
+
+// -----
+
+func.func @uplift_while(%arg0: i64, %arg1: i64, %arg2: i64) -> i64 {
+  %0 = scf.while (%arg3 = %arg0) : (i64) -> (i64) {
+    %1 = arith.cmpi slt, %arg3, %arg1 : i64
+    scf.condition(%1) %arg3 : i64
+  } do {
+  ^bb0(%arg3: i64):
+    "test.test1"(%arg3) : (i64) -> ()
+    %added = arith.addi %arg3, %arg2 : i64
+    "test.test2"(%added) : (i64) -> ()
+    scf.yield %added : i64
+  }
+  return %0 : i64
+}
+
+// CHECK-LABEL: func @uplift_while
+//  CHECK-SAME:     (%[[BEGIN:.*]]: i64, %[[END:.*]]: i64, %[[STEP:.*]]: i64) -> i64
+//       CHECK:     %[[C1:.*]] = arith.constant 1 : i64
+//       CHECK:     scf.for %[[I:.*]] = %[[BEGIN]] to %[[END]] step %[[STEP]] : i64 {
+//       CHECK:     "test.test1"(%[[I]]) : (i64) -> ()
+//       CHECK:     %[[INC:.*]] = arith.addi %[[I]], %[[STEP]] : i64
+//       CHECK:     "test.test2"(%[[INC]]) : (i64) -> ()
+//       CHECK:     %[[R1:.*]] = arith.subi %[[STEP]], %[[C1]] : i64
+//       CHECK:     %[[R2:.*]] = arith.subi %[[END]], %[[BEGIN]] : i64
+//       CHECK:     %[[R3:.*]] = arith.addi %[[R2]], %[[R1]] : i64
+//       CHECK:     %[[R4:.*]] = arith.divsi %[[R3]], %[[STEP]] : i64
+//       CHECK:     %[[R5:.*]] = arith.subi %[[R4]], %[[C1]] : i64
+//       CHECK:     %[[R6:.*]] = arith.muli %[[R5]], %[[STEP]] : i64
+//       CHECK:     %[[R7:.*]] = arith.addi %[[BEGIN]], %[[R6]] : i64
+//       CHECK:     return %[[R7]] : i64
diff --git a/mlir/test/lib/Dialect/SCF/CMakeLists.txt b/mlir/test/lib/Dialect/SCF/CMakeLists.txt
index d93bd55915182..792430cc84b65 100644
--- a/mlir/test/lib/Dialect/SCF/CMakeLists.txt
+++ b/mlir/test/lib/Dialect/SCF/CMakeLists.txt
@@ -4,6 +4,7 @@ add_mlir_library(MLIRSCFTestPasses
   TestLoopUnrolling.cpp
   TestSCFUtils.cpp
   TestSCFWrapInZeroTripCheck.cpp
+  TestUpliftWhileToFor.cpp
   TestWhileOpBuilder.cpp
 
   EXCLUDE_FROM_LIBMLIR
diff --git a/mlir/test/lib/Dialect/SCF/TestUpliftWhileToFor.cpp b/mlir/test/lib/Dialect/SCF/TestUpliftWhileToFor.cpp
new file mode 100644
index 0000000000000..468bc0ca78489
--- /dev/null
+++ b/mlir/test/lib/Dialect/SCF/TestUpliftWhileToFor.cpp
@@ -0,0 +1,50 @@
+//===- TestUpliftWhileToFor.cpp - while to for loop uplifting test pass ---===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Pass to test transforms SCF.WhileOp's into SCF.ForOp's.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/SCF/Transforms/Patterns.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/Pass/Pass.h"
+#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
+
+using namespace mlir;
+
+namespace {
+
+struct TestSCFUpliftWhileToFor
+    : public PassWrapper<TestSCFUpliftWhileToFor, OperationPass<void>> {
+  MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestSCFUpliftWhileToFor)
+
+  StringRef getArgument() const final { return "test-scf-uplift-while-to-for"; }
+
+  StringRef getDescription() const final {
+    return "test scf while to for uplifting";
+  }
+
+  void runOnOperation() override {
+    Operation *op = getOperation();
+    MLIRContext *ctx = op->getContext();
+    RewritePatternSet patterns(ctx);
+    scf::populateUpliftWhileToForPatterns(patterns);
+    if (failed(applyPatternsAndFoldGreedily(op, std::move(patterns))))
+      signalPassFailure();
+  }
+};
+
+} // namespace
+
+namespace mlir {
+namespace test {
+void registerTestSCFUpliftWhileToFor() {
+  PassRegistration<TestSCFUpliftWhileToFor>();
+}
+} // namespace test
+} // namespace mlir
diff --git a/mlir/tools/mlir-opt/mlir-opt.cpp b/mlir/tools/mlir-opt/mlir-opt.cpp
index 6ce9f3041d6f4..237ebeb166dc9 100644
--- a/mlir/tools/mlir-opt/mlir-opt.cpp
+++ b/mlir/tools/mlir-opt/mlir-opt.cpp
@@ -130,6 +130,7 @@ void registerTestOneToNTypeConversionPass();
 void registerTestOpaqueLoc();
 void registerTestPadFusion();
 void registerTestRecursiveTypesPass();
+void registerTestSCFUpliftWhileToFor();
 void registerTestSCFUtilsPass();
 void registerTestSCFWhileOpBuilderPass();
 void registerTestSCFWrapInZeroTripCheckPasses();
@@ -258,6 +259,7 @@ void registerTestPasses() {
   mlir::test::registerTestOpaqueLoc();
   mlir::test::registerTestPadFusion();
   mlir::test::registerTestRecursiveTypesPass();
+  mlir::test::registerTestSCFUpliftWhileToFor();
   mlir::test::registerTestSCFUtilsPass();
   mlir::test::registerTestSCFWhileOpBuilderPass();
   mlir::test::registerTestSCFWrapInZeroTripCheckPasses();

From 43b4e5b9bff1719a6d3628a3a805cfeb59adaf16 Mon Sep 17 00:00:00 2001
From: Utkarsh Saxena <usx@google.com>
Date: Mon, 15 Apr 2024 21:30:18 +0200
Subject: [PATCH 477/886] Use pushFullExprCleanup for deferred destroy (#88670)

Instead of directly pushing the `Destroy`, we should use
`pushFullExprCleanup` which handles conditional branches.
This fixes a backend crash due to
89ba7e183e6e2c64370ed1b963e54c06352211db.

Added minimized crash reproducer.
---
 clang/lib/CodeGen/CGDecl.cpp                  |  7 +++--
 .../CodeGenCXX/control-flow-in-stmt-expr.cpp  | 31 ++++++++++++++++++-
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 8bdafa7c569b0..3f05ebb561da5 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -2216,8 +2216,11 @@ void CodeGenFunction::pushDestroyAndDeferDeactivation(
 void CodeGenFunction::pushDestroyAndDeferDeactivation(
     CleanupKind cleanupKind, Address addr, QualType type, Destroyer *destroyer,
     bool useEHCleanupForArray) {
-  pushCleanupAndDeferDeactivation<DestroyObject>(
-      cleanupKind, addr, type, destroyer, useEHCleanupForArray);
+  llvm::Instruction *DominatingIP =
+      Builder.CreateFlagLoad(llvm::Constant::getNullValue(Int8PtrTy));
+  pushDestroy(cleanupKind, addr, type, destroyer, useEHCleanupForArray);
+  DeferredDeactivationCleanupStack.push_back(
+      {EHStack.stable_begin(), DominatingIP});
 }
 
 void CodeGenFunction::pushStackRestore(CleanupKind Kind, Address SPMem) {
diff --git a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp b/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
index ffde1bd6a724d..95deee8bb1f1f 100644
--- a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
+++ b/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
@@ -1,4 +1,5 @@
-// RUN: %clang_cc1 --std=c++20 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 --std=c++20 -fexceptions -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefixes=EH %s
+// RUN: %clang_cc1 --std=c++20 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefixes=NOEH,CHECK %s
 
 struct Printy {
   Printy(const char *name) : name(name) {}
@@ -349,6 +350,34 @@ void NewArrayInit() {
   // CHECK-NEXT:    br label %return
 }
 
+void DestroyInConditionalCleanup() {
+  // EH-LABEL: DestroyInConditionalCleanupv()
+  // NOEH-LABEL: DestroyInConditionalCleanupv()
+  struct A {
+    A() {}
+    ~A() {}
+  };
+
+  struct Value {
+    Value(A) {}
+    ~Value() {}
+  };
+
+  struct V2 {
+    Value K;
+    Value V;
+  };
+  // Verify we use conditional cleanups.
+  (void)(foo() ? V2{A(), A()} : V2{A(), A()});
+  // NOEH:   cond.true:
+  // NOEH:      call void @_ZZ27DestroyInConditionalCleanupvEN1AC1Ev
+  // NOEH:      store ptr %{{.*}}, ptr %cond-cleanup.save
+
+  // EH:   cond.true:
+  // EH:        invoke void @_ZZ27DestroyInConditionalCleanupvEN1AC1Ev
+  // EH:        store ptr %{{.*}}, ptr %cond-cleanup.save
+}
+
 void ArrayInitWithContinue() {
   // CHECK-LABEL: @_Z21ArrayInitWithContinuev
   // Verify that we start to emit the array destructor.

From b03dc7d16251bf249b67e4db545353551727f646 Mon Sep 17 00:00:00 2001
From: cor3ntin <corentinjabot@gmail.com>
Date: Mon, 15 Apr 2024 21:38:10 +0200
Subject: [PATCH 478/886] [Clang] Properly  set the value category of dependent
 unary operators (#88740)

This fixes an assertion in Expr::Classify when a
trying to deduce a dependent dereference operator.

Fixes #88329
---
 clang/docs/ReleaseNotes.rst                |  1 +
 clang/lib/Sema/SemaOverload.cpp            | 13 ++++++++++---
 clang/test/CXX/over/over.built/ast.cpp     |  4 ++--
 clang/test/SemaCXX/overloaded-operator.cpp | 10 ++++++++++
 4 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index de5fa99bff5db..dc108785f6cc9 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -537,6 +537,7 @@ Bug Fixes to C++ Support
 - Clang now correctly tracks type dependence of by-value captures in lambdas with an explicit
   object parameter.
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
+- Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 675eb6ec05e78..227ef564ba3e0 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -14395,9 +14395,16 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
   ArrayRef<Expr *> ArgsArray(Args, NumArgs);
 
   if (Input->isTypeDependent()) {
+    ExprValueKind VK = ExprValueKind::VK_PRValue;
+    // [C++26][expr.unary.op][expr.pre.incr]
+    // The * operator yields an lvalue of type
+    // The pre/post increment operators yied an lvalue.
+    if (Opc == UO_PreDec || Opc == UO_PreInc || Opc == UO_Deref)
+      VK = VK_LValue;
+
     if (Fns.empty())
-      return UnaryOperator::Create(Context, Input, Opc, Context.DependentTy,
-                                   VK_PRValue, OK_Ordinary, OpLoc, false,
+      return UnaryOperator::Create(Context, Input, Opc, Context.DependentTy, VK,
+                                   OK_Ordinary, OpLoc, false,
                                    CurFPFeatureOverrides());
 
     CXXRecordDecl *NamingClass = nullptr; // lookup ignores member operators
@@ -14406,7 +14413,7 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
     if (Fn.isInvalid())
       return ExprError();
     return CXXOperatorCallExpr::Create(Context, Op, Fn.get(), ArgsArray,
-                                       Context.DependentTy, VK_PRValue, OpLoc,
+                                       Context.DependentTy, VK, OpLoc,
                                        CurFPFeatureOverrides());
   }
 
diff --git a/clang/test/CXX/over/over.built/ast.cpp b/clang/test/CXX/over/over.built/ast.cpp
index f76606b1f9869..56a63431269f3 100644
--- a/clang/test/CXX/over/over.built/ast.cpp
+++ b/clang/test/CXX/over/over.built/ast.cpp
@@ -4,11 +4,11 @@ struct A{};
 
 template <typename T, typename U>
 auto Test(T* pt, U* pu) {
-  // CHECK: UnaryOperator {{.*}} '<dependent type>' prefix '*'
+  // CHECK: UnaryOperator {{.*}} '<dependent type>' lvalue prefix '*'
   // CHECK-NEXT: DeclRefExpr {{.*}} 'T *' lvalue ParmVar {{.*}} 'pt' 'T *'
   (void)*pt;
 
-  // CHECK: UnaryOperator {{.*}} '<dependent type>' prefix '++'
+  // CHECK: UnaryOperator {{.*}} '<dependent type>' lvalue prefix '++'
   // CHECK-NEXT: DeclRefExpr {{.*}} 'T *' lvalue ParmVar {{.*}} 'pt' 'T *'
   (void)(++pt);
 
diff --git a/clang/test/SemaCXX/overloaded-operator.cpp b/clang/test/SemaCXX/overloaded-operator.cpp
index 49311625d7ab2..cab21d67a002f 100644
--- a/clang/test/SemaCXX/overloaded-operator.cpp
+++ b/clang/test/SemaCXX/overloaded-operator.cpp
@@ -682,3 +682,13 @@ namespace nw{
   }
 }
 #endif
+
+#if __cplusplus >= 201703L
+namespace GH88329 {
+
+template <auto T> struct A {};
+template <auto T> A<*T> operator *() { return {}; }
+// expected-error@-1 {{overloaded 'operator*' must have at least one parameter of class or enumeration type}}
+}
+
+#endif

From 187c035c047616860d31fc399702538ad1f42755 Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Mon, 15 Apr 2024 15:38:25 -0400
Subject: [PATCH 479/886] [gn] port dfafe3822b1

---
 llvm/utils/gn/secondary/compiler-rt/test/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/compiler-rt/test/BUILD.gn b/llvm/utils/gn/secondary/compiler-rt/test/BUILD.gn
index d533e79e6374d..2b358f74ae6c3 100644
--- a/llvm/utils/gn/secondary/compiler-rt/test/BUILD.gn
+++ b/llvm/utils/gn/secondary/compiler-rt/test/BUILD.gn
@@ -56,6 +56,7 @@ write_cmake_config("lit_common_configured") {
     "COMPILER_RT_INTERCEPT_LIBDISPATCH_PYBOOL=False",
     "COMPILER_RT_RESOLVED_LIBRARY_OUTPUT_DIR=" +
         rebase_path(crt_current_out_dir),
+    "COMPILER_RT_RESOLVED_OUTPUT_DIR=" + rebase_path(crt_current_out_dir),
     "COMPILER_RT_EMULATOR=",
     "COMPILER_RT_ASAN_SHADOW_SCALE=",
     "COMPILER_RT_MEMPROF_SHADOW_SCALE=",

From 51bc8cdf7bbc98cd691fc480b333907e665b8eae Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 12:45:02 -0700
Subject: [PATCH 480/886] [RISCV] Add missing space to RISCVInstrInfoZvk.td.
 NFC

---
 llvm/lib/Target/RISCV/RISCVInstrInfoZvk.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZvk.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZvk.td
index 575f9b41accb5..aac7dc444a2de 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZvk.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZvk.td
@@ -152,7 +152,7 @@ let Predicates = [HasStdExtZvknhaOrZvknhb], RVVConstraint = Sha2Constraint in {
   def VSHA2MS_VV : PALUVVNoVmTernary<0b101101, OPMVV, "vsha2ms.vv">;
 } // Predicates = [HasStdExtZvknhaOrZvknhb]
 
-let Predicates = [HasStdExtZvkned]in {
+let Predicates = [HasStdExtZvkned] in {
   defm VAESDF     : VAES_MV_V_S<0b101000, 0b101001, 0b00001, OPMVV, "vaesdf">;
   defm VAESDM     : VAES_MV_V_S<0b101000, 0b101001, 0b00000, OPMVV, "vaesdm">;
   defm VAESEF     : VAES_MV_V_S<0b101000, 0b101001, 0b00011, OPMVV, "vaesef">;

From 52a88d3f5af6bed08a47f1c2351f320e0e1f4a88 Mon Sep 17 00:00:00 2001
From: Yijia Gu <yijiagu@google.com>
Date: Mon, 15 Apr 2024 12:47:47 -0700
Subject: [PATCH 481/886] [mlir] fix llvm:support dependency error

---
 utils/bazel/llvm-project-overlay/mlir/BUILD.bazel | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 522e3efe4e0ab..653c4bd30600a 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -6617,7 +6617,7 @@ cc_library(
         ":PolynomialAttributesIncGen",
         ":PolynomialIncGen",
         ":Support",
-        "//third_party/llvm/llvm-project/llvm:Support",
+        "//llvm:Support",
     ],
 )
 

From faad4e3fa81d10b91a9128ca84abfe381ac7c2e0 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Mon, 15 Apr 2024 15:03:37 -0500
Subject: [PATCH 482/886] [Libomptarget] Split PowerPC into separate triples
 (#88773)

Summary:
The previous patch mistakenly merged these when they indeed need to be
treated like separate triples because it's what's passed to the test
suite.
---
 .../libomptarget/plugins-nextgen/host/CMakeLists.txt   | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/openmp/libomptarget/plugins-nextgen/host/CMakeLists.txt b/openmp/libomptarget/plugins-nextgen/host/CMakeLists.txt
index ccbf7d033fd66..c1493d293d30e 100644
--- a/openmp/libomptarget/plugins-nextgen/host/CMakeLists.txt
+++ b/openmp/libomptarget/plugins-nextgen/host/CMakeLists.txt
@@ -51,8 +51,14 @@ else()
 endif()
 
 # Define the target specific triples and ELF machine values.
-if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le$" OR
-   CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64$")
+if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le$")
+  target_compile_definitions(omptarget.rtl.${machine} PRIVATE TARGET_ELF_ID=EM_PPC64)
+  target_compile_definitions(omptarget.rtl.${machine} PRIVATE
+      LIBOMPTARGET_NEXTGEN_GENERIC_PLUGIN_TRIPLE="powerpc64le-ibm-linux-gnu")
+  list(APPEND LIBOMPTARGET_SYSTEM_TARGETS 
+       "powerpc64le-ibm-linux-gnu" "powerpc64le-ibm-linux-gnu-LTO")
+  set(LIBOMPTARGET_SYSTEM_TARGETS "${LIBOMPTARGET_SYSTEM_TARGETS}" PARENT_SCOPE)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64$")
   target_compile_definitions(omptarget.rtl.${machine} PRIVATE TARGET_ELF_ID=EM_PPC64)
   target_compile_definitions(omptarget.rtl.${machine} PRIVATE
       LIBOMPTARGET_NEXTGEN_GENERIC_PLUGIN_TRIPLE="powerpc64-ibm-linux-gnu")

From fb9c7dd6f1f3463b539d1f62f6ec7b5c7e061cf8 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 13:04:14 -0700
Subject: [PATCH 483/886] [RISCV] Minor cleanups to
 RISCVISAInfo::parseArchString. NFC

Remove unneeded an unneeded variable and rename another.
---
 llvm/lib/Support/RISCVISAInfo.cpp | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Support/RISCVISAInfo.cpp b/llvm/lib/Support/RISCVISAInfo.cpp
index 7a19d24d1ff48..cbdc64bc7a97b 100644
--- a/llvm/lib/Support/RISCVISAInfo.cpp
+++ b/llvm/lib/Support/RISCVISAInfo.cpp
@@ -935,7 +935,6 @@ RISCVISAInfo::parseArchString(StringRef Arch, bool EnableExperimentalExtension,
 
   // The canonical order specified in ISA manual.
   // Ref: Table 22.1 in RISC-V User-Level ISA V2.2
-  StringRef StdExts = AllStdExts;
   char Baseline = Arch[4];
 
   // First letter should be 'e', 'i' or 'g'.
@@ -951,7 +950,6 @@ RISCVISAInfo::parseArchString(StringRef Arch, bool EnableExperimentalExtension,
     if (Arch.size() > 5 && isDigit(Arch[5]))
       return createStringError(errc::invalid_argument,
                                "version not supported for 'g'");
-    StdExts = StdExts.drop_front(4);
     break;
   }
 
@@ -1001,11 +999,11 @@ RISCVISAInfo::parseArchString(StringRef Arch, bool EnableExperimentalExtension,
   Exts = Exts.drop_front(ConsumeLength);
   Exts.consume_front("_");
 
-  std::vector<std::string> SplittedExts;
-  if (auto E = splitExtsByUnderscore(Exts, SplittedExts))
+  std::vector<std::string> SplitExts;
+  if (auto E = splitExtsByUnderscore(Exts, SplitExts))
     return std::move(E);
 
-  for (auto &Ext : SplittedExts) {
+  for (auto &Ext : SplitExts) {
     StringRef CurrExt = Ext;
     while (!CurrExt.empty()) {
       if (AllStdExts.contains(CurrExt.front())) {

From 9e485f48228114ab6c34151503740ea15cddb351 Mon Sep 17 00:00:00 2001
From: cmtice <cmtice@google.com>
Date: Mon, 15 Apr 2024 13:11:24 -0700
Subject: [PATCH 484/886] [DWARF] Add accessor for NameIndex Offsets field.
 (#88783)

Add an accessor function to DWARFDebugNames::NameIndex to allow other
classes to access the Offsets field within NameIndex. This will be used
by LLD to create a merged .debug_names index
(https://github.com/llvm/llvm-project/pull/86508)
---
 llvm/include/llvm/DebugInfo/DWARF/DWARFAcceleratorTable.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/llvm/include/llvm/DebugInfo/DWARF/DWARFAcceleratorTable.h b/llvm/include/llvm/DebugInfo/DWARF/DWARFAcceleratorTable.h
index 9543b78ea6130..0d447a78cdc61 100644
--- a/llvm/include/llvm/DebugInfo/DWARF/DWARFAcceleratorTable.h
+++ b/llvm/include/llvm/DebugInfo/DWARF/DWARFAcceleratorTable.h
@@ -608,6 +608,9 @@ class DWARFDebugNames : public DWARFAcceleratorTable {
     /// Returns Hdr field
     Header getHeader() const { return Hdr; }
 
+    /// Returns Offsets field
+    DWARFDebugNamesOffsets getOffsets() const { return Offsets; }
+
     /// Reads offset of compilation unit CU. CU is 0-based.
     uint64_t getCUOffset(uint32_t CU) const;
     uint32_t getCUCount() const { return Hdr.CompUnitCount; }

From 071ac0ae3029594167d2bb4d28859bdc36367034 Mon Sep 17 00:00:00 2001
From: Jonas Devlieghere <jonas@devlieghere.com>
Date: Mon, 15 Apr 2024 13:18:26 -0700
Subject: [PATCH 485/886] [lldb] Skip lldb-server unit tests when building with
 ASan

The lldb-server unit tests are timing out on GreenDragon and swift-ci.
We haven't been able to reproduce this locally or figure out why. We've
made several attempts to address potential issues, but we've reached a
point where we need to skip them to get signal out of this bot again.
---
 lldb/unittests/tools/CMakeLists.txt | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/lldb/unittests/tools/CMakeLists.txt b/lldb/unittests/tools/CMakeLists.txt
index 055fc6e6f5df4..42b0c25dd1fc9 100644
--- a/lldb/unittests/tools/CMakeLists.txt
+++ b/lldb/unittests/tools/CMakeLists.txt
@@ -1,3 +1,5 @@
 if(LLDB_TOOL_LLDB_SERVER_BUILD)
-  add_subdirectory(lldb-server)
+  if (NOT LLVM_USE_SANITIZER MATCHES ".*Address.*")
+    add_subdirectory(lldb-server)
+  endif()
 endif()

From 0287a5cc4e2a5ded1ae2e4079f91052e6a6b8d9b Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Mon, 15 Apr 2024 15:28:05 -0500
Subject: [PATCH 486/886] [OpenMP] Remove 'minncta' attributes from NVPTX
 kernels (#88398)

Summary:
Currently we treat this attribute as a minimum number for the amount of
blocks scheduled on the kernel. However, the doucmentation states that
this applies to CTA's mapped onto a *single* SM. Currently we just set
it to the total number of blocks, which will almost always result in a
warning that the value is out of range and will be ignored. We don't
have a good way to automatically know how many CTAs can be put on a
single SM nor if we should do this, so we should probably leave this up
to users manually adding it.


https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#performance-tuning-directives-minnctapersm
---
 clang/test/OpenMP/ompx_attributes_codegen.cpp | 3 +--
 llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp     | 4 +---
 2 files changed, 2 insertions(+), 5 deletions(-)

diff --git a/clang/test/OpenMP/ompx_attributes_codegen.cpp b/clang/test/OpenMP/ompx_attributes_codegen.cpp
index 6735972c6b107..87eb2913537ba 100644
--- a/clang/test/OpenMP/ompx_attributes_codegen.cpp
+++ b/clang/test/OpenMP/ompx_attributes_codegen.cpp
@@ -36,6 +36,5 @@ void func() {
 // NVIDIA: "omp_target_thread_limit"="45"
 // NVIDIA: "omp_target_thread_limit"="17"
 // NVIDIA: !{ptr @__omp_offloading[[HASH1:.*]]_l16, !"maxntidx", i32 20}
-// NVIDIA: !{ptr @__omp_offloading[[HASH2:.*]]_l18, !"minctasm", i32 90}
-// NVIDIA: !{ptr @__omp_offloading[[HASH2]]_l18, !"maxntidx", i32 45}
+// NVIDIA: !{ptr @__omp_offloading[[HASH2:.*]]_l18, !"maxntidx", i32 45}
 // NVIDIA: !{ptr @__omp_offloading[[HASH3:.*]]_l20, !"maxntidx", i32 17}
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 7fd8474c2ec89..4d2d352f7520b 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -4786,11 +4786,9 @@ OpenMPIRBuilder::readTeamBoundsForKernel(const Triple &, Function &Kernel) {
 
 void OpenMPIRBuilder::writeTeamsForKernel(const Triple &T, Function &Kernel,
                                           int32_t LB, int32_t UB) {
-  if (T.isNVPTX()) {
+  if (T.isNVPTX())
     if (UB > 0)
       updateNVPTXMetadata(Kernel, "maxclusterrank", UB, true);
-    updateNVPTXMetadata(Kernel, "minctasm", LB, false);
-  }
   if (T.isAMDGPU())
     Kernel.addFnAttr("amdgpu-max-num-workgroups", llvm::utostr(LB) + ",1,1");
 

From 7eb5022cbb11fdf5e74d707b4c93d3669a46eccf Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Mon, 15 Apr 2024 21:45:00 +0100
Subject: [PATCH 487/886] [lldb][NFC] Turn ChildCacheState into an unscoped
 enum (#88703)

Done for consistency with the rest of the enums in
`lldb-enumerations.h`. See
https://github.com/llvm/llvm-project/pull/80167#issuecomment-2043721298
---
 lldb/include/lldb/lldb-enumerations.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lldb/include/lldb/lldb-enumerations.h b/lldb/include/lldb/lldb-enumerations.h
index f3b07ea6d2039..15e4585718609 100644
--- a/lldb/include/lldb/lldb-enumerations.h
+++ b/lldb/include/lldb/lldb-enumerations.h
@@ -1310,7 +1310,7 @@ enum CompletionType {
 
 /// Specifies if children need to be re-computed
 /// after a call to \ref SyntheticChildrenFrontEnd::Update.
-enum class ChildCacheState {
+enum ChildCacheState {
   eRefetch = 0, ///< Children need to be recomputed dynamically.
 
   eReuse = 1, ///< Children did not change and don't need to be recomputed;

From b8c16db938327efdabc63f21fdc770069ac4406b Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Mon, 15 Apr 2024 21:45:25 +0100
Subject: [PATCH 488/886] [lldb][TypeSystemClang][NFCI] Use
 LangOptions::setLangDefaults when creating new LangOptions (#88721)

This logic was originally copied from `CompilerInstance::parseLangArgs`.
Since then, the `CompilerInstance` uses `LangOptions::setLangDefaults`
to set up new `LangOptions` instances. In our case, we only ever passed
`Language::ObjCXX` into LLDB's `ParseLangArgs`, so most of this function
was dead code.

This patch replaces the duplicated logic with a call to
`LangOptions::setLangDefaults`.
---
 .../TypeSystem/Clang/TypeSystemClang.cpp      | 79 ++-----------------
 1 file changed, 6 insertions(+), 73 deletions(-)

diff --git a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
index 44bd02bd4b367..be0ddb06f82c1 100644
--- a/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
+++ b/lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp
@@ -459,85 +459,19 @@ TypeSystemClang::ConvertAccessTypeToAccessSpecifier(AccessType access) {
   return AS_none;
 }
 
-static void ParseLangArgs(LangOptions &Opts, InputKind IK, const char *triple) {
+static void ParseLangArgs(LangOptions &Opts, ArchSpec arch) {
   // FIXME: Cleanup per-file based stuff.
 
-  // Set some properties which depend solely on the input kind; it would be
-  // nice to move these to the language standard, and have the driver resolve
-  // the input kind + language standard.
-  if (IK.getLanguage() == clang::Language::Asm) {
-    Opts.AsmPreprocessor = 1;
-  } else if (IK.isObjectiveC()) {
-    Opts.ObjC = 1;
-  }
-
-  LangStandard::Kind LangStd = LangStandard::lang_unspecified;
-
-  if (LangStd == LangStandard::lang_unspecified) {
-    // Based on the base language, pick one.
-    switch (IK.getLanguage()) {
-    case clang::Language::Unknown:
-    case clang::Language::CIR:
-    case clang::Language::LLVM_IR:
-    case clang::Language::RenderScript:
-      llvm_unreachable("Invalid input kind!");
-    case clang::Language::OpenCL:
-      LangStd = LangStandard::lang_opencl10;
-      break;
-    case clang::Language::OpenCLCXX:
-      LangStd = LangStandard::lang_openclcpp10;
-      break;
-    case clang::Language::Asm:
-    case clang::Language::C:
-    case clang::Language::ObjC:
-      LangStd = LangStandard::lang_gnu99;
-      break;
-    case clang::Language::CXX:
-    case clang::Language::ObjCXX:
-      LangStd = LangStandard::lang_gnucxx98;
-      break;
-    case clang::Language::CUDA:
-    case clang::Language::HIP:
-      LangStd = LangStandard::lang_gnucxx17;
-      break;
-    case clang::Language::HLSL:
-      LangStd = LangStandard::lang_hlsl;
-      break;
-    }
-  }
-
-  const LangStandard &Std = LangStandard::getLangStandardForKind(LangStd);
-  Opts.LineComment = Std.hasLineComments();
-  Opts.C99 = Std.isC99();
-  Opts.CPlusPlus = Std.isCPlusPlus();
-  Opts.CPlusPlus11 = Std.isCPlusPlus11();
-  Opts.CPlusPlus14 = Std.isCPlusPlus14();
-  Opts.CPlusPlus17 = Std.isCPlusPlus17();
-  Opts.CPlusPlus20 = Std.isCPlusPlus20();
-  Opts.Digraphs = Std.hasDigraphs();
-  Opts.GNUMode = Std.isGNUMode();
-  Opts.GNUInline = !Std.isC99();
-  Opts.HexFloats = Std.hasHexFloats();
-
-  Opts.WChar = true;
-
-  // OpenCL has some additional defaults.
-  if (LangStd == LangStandard::lang_opencl10) {
-    Opts.OpenCL = 1;
-    Opts.AltiVec = 1;
-    Opts.CXXOperatorNames = 1;
-    Opts.setLaxVectorConversions(LangOptions::LaxVectorConversionKind::All);
-  }
-
-  // OpenCL and C++ both have bool, true, false keywords.
-  Opts.Bool = Opts.OpenCL || Opts.CPlusPlus;
+  std::vector<std::string> Includes;
+  LangOptions::setLangDefaults(Opts, clang::Language::ObjCXX, arch.GetTriple(),
+                               Includes, clang::LangStandard::lang_gnucxx98);
 
   Opts.setValueVisibilityMode(DefaultVisibility);
 
   // Mimicing gcc's behavior, trigraphs are only enabled if -trigraphs is
   // specified, or -std is set to a conforming mode.
   Opts.Trigraphs = !Opts.GNUMode;
-  Opts.CharIsSigned = ArchSpec(triple).CharIsSignedByDefault();
+  Opts.CharIsSigned = arch.CharIsSignedByDefault();
   Opts.OptimizeSize = 0;
 
   // FIXME: Eliminate this dependency.
@@ -727,8 +661,7 @@ void TypeSystemClang::CreateASTContext() {
   m_ast_owned = true;
 
   m_language_options_up = std::make_unique<LangOptions>();
-  ParseLangArgs(*m_language_options_up, clang::Language::ObjCXX,
-                GetTargetTriple());
+  ParseLangArgs(*m_language_options_up, ArchSpec(GetTargetTriple()));
 
   m_identifier_table_up =
       std::make_unique<IdentifierTable>(*m_language_options_up, nullptr);

From 905d2ecbb63f4087b50d633b012c9915ab3a6829 Mon Sep 17 00:00:00 2001
From: Michael Buch <michaelbuch12@gmail.com>
Date: Mon, 15 Apr 2024 21:45:41 +0100
Subject: [PATCH 489/886] [lldb][ClangExpressionDeclMap][NFC] Remove unused
 NameSearchContext::m_found_function (#88724)

This member was never actually used, ever since its introduction in
`ca4e0fd7e63b90e6f68044af47248c64f250ee8f`.
---
 .../Plugins/ExpressionParser/Clang/ClangExpressionDeclMap.cpp | 4 ----
 .../source/Plugins/ExpressionParser/Clang/NameSearchContext.h | 1 -
 2 files changed, 5 deletions(-)

diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionDeclMap.cpp b/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionDeclMap.cpp
index 2d306b42760b1..31f6447d66f64 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionDeclMap.cpp
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangExpressionDeclMap.cpp
@@ -1049,7 +1049,6 @@ void ClangExpressionDeclMap::LookupInModulesDeclVendor(
     context.AddNamedDecl(copied_function);
 
     context.m_found_function_with_type_info = true;
-    context.m_found_function = true;
   } else if (auto copied_var = dyn_cast<clang::VarDecl>(copied_decl)) {
     context.AddNamedDecl(copied_var);
     context.m_found_variable = true;
@@ -1299,7 +1298,6 @@ void ClangExpressionDeclMap::LookupFunction(
 
         AddOneFunction(context, sym_ctx.function, nullptr);
         context.m_found_function_with_type_info = true;
-        context.m_found_function = true;
       } else if (sym_ctx.symbol) {
         Symbol *symbol = sym_ctx.symbol;
         if (target && symbol->GetType() == eSymbolTypeReExported) {
@@ -1331,10 +1329,8 @@ void ClangExpressionDeclMap::LookupFunction(
     if (!context.m_found_function_with_type_info) {
       if (extern_symbol) {
         AddOneFunction(context, nullptr, extern_symbol);
-        context.m_found_function = true;
       } else if (non_extern_symbol) {
         AddOneFunction(context, nullptr, non_extern_symbol);
-        context.m_found_function = true;
       }
     }
   }
diff --git a/lldb/source/Plugins/ExpressionParser/Clang/NameSearchContext.h b/lldb/source/Plugins/ExpressionParser/Clang/NameSearchContext.h
index dc8621dd6aba5..9a3320636081b 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/NameSearchContext.h
+++ b/lldb/source/Plugins/ExpressionParser/Clang/NameSearchContext.h
@@ -41,7 +41,6 @@ struct NameSearchContext {
 
   bool m_found_variable = false;
   bool m_found_function_with_type_info = false;
-  bool m_found_function = false;
   bool m_found_local_vars_nsp = false;
   bool m_found_type = false;
 

From 69d861e1320119e9a02907155ca626b1db90ad93 Mon Sep 17 00:00:00 2001
From: Nathan Sidwell <nathan@acm.org>
Date: Mon, 15 Apr 2024 16:55:58 -0400
Subject: [PATCH 490/886] [clang] Move tailclipping to bitfield allocation
 (#87090)

Move bitfield access clipping to bitfield access computation.
---
 clang/lib/CodeGen/CGRecordLayoutBuilder.cpp | 84 +++++++++++----------
 clang/test/CodeGen/bitfield-access-unit.c   | 18 +++++
 2 files changed, 62 insertions(+), 40 deletions(-)

diff --git a/clang/lib/CodeGen/CGRecordLayoutBuilder.cpp b/clang/lib/CodeGen/CGRecordLayoutBuilder.cpp
index 634a55fec5182..868b1ab98e048 100644
--- a/clang/lib/CodeGen/CGRecordLayoutBuilder.cpp
+++ b/clang/lib/CodeGen/CGRecordLayoutBuilder.cpp
@@ -41,10 +41,11 @@ namespace {
 ///   contains enough information to determine where the runs break.  Microsoft
 ///   and Itanium follow different rules and use different codepaths.
 /// * It is desired that, when possible, bitfields use the appropriate iN type
-///   when lowered to llvm types.  For example unsigned x : 24 gets lowered to
+///   when lowered to llvm types. For example unsigned x : 24 gets lowered to
 ///   i24.  This isn't always possible because i24 has storage size of 32 bit
-///   and if it is possible to use that extra byte of padding we must use
-///   [i8 x 3] instead of i24.  The function clipTailPadding does this.
+///   and if it is possible to use that extra byte of padding we must use [i8 x
+///   3] instead of i24. This is computed when accumulating bitfields in
+///   accumulateBitfields.
 ///   C++ examples that require clipping:
 ///   struct { int a : 24; char b; }; // a must be clipped, b goes at offset 3
 ///   struct A { int a : 24; ~A(); }; // a must be clipped because:
@@ -62,11 +63,7 @@ namespace {
 ///   that the tail padding is not used in the complete class.) However,
 ///   because LLVM reads from the complete type it can generate incorrect code
 ///   if we do not clip the tail padding off of the bitfield in the complete
-///   layout.  This introduces a somewhat awkward extra unnecessary clip stage.
-///   The location of the clip is stored internally as a sentinel of type
-///   SCISSOR.  If LLVM were updated to read base types (which it probably
-///   should because locations of things such as VBases are bogus in the llvm
-///   type anyway) then we could eliminate the SCISSOR.
+///   layout.
 /// * Itanium allows nearly empty primary virtual bases.  These bases don't get
 ///   get their own storage because they're laid out as part of another base
 ///   or at the beginning of the structure.  Determining if a VBase actually
@@ -200,9 +197,7 @@ struct CGRecordLowering {
                      const CXXRecordDecl *Query) const;
   void calculateZeroInit();
   CharUnits calculateTailClippingOffset(bool isNonVirtualBaseType) const;
-  /// Lowers bitfield storage types to I8 arrays for bitfields with tail
-  /// padding that is or can potentially be used.
-  void clipTailPadding();
+  void checkBitfieldClipping() const;
   /// Determines if we need a packed llvm struct.
   void determinePacked(bool NVBaseType);
   /// Inserts padding everywhere it's needed.
@@ -305,7 +300,7 @@ void CGRecordLowering::lower(bool NVBaseType) {
   }
   llvm::stable_sort(Members);
   Members.push_back(StorageInfo(Size, getIntNType(8)));
-  clipTailPadding();
+  checkBitfieldClipping();
   determinePacked(NVBaseType);
   insertPadding();
   Members.pop_back();
@@ -531,6 +526,7 @@ CGRecordLowering::accumulateBitFields(bool isNonVirtualBaseType,
   // available padding characters.
   RecordDecl::field_iterator BestEnd = Begin;
   CharUnits BestEndOffset;
+  bool BestClipped; // Whether the representation must be in a byte array.
 
   for (;;) {
     // AtAlignedBoundary is true iff Field is the (potential) start of a new
@@ -593,10 +589,9 @@ CGRecordLowering::accumulateBitFields(bool isNonVirtualBaseType,
         // this is the best seen so far.
         BestEnd = Field;
         BestEndOffset = BeginOffset + AccessSize;
-        if (Types.getCodeGenOpts().FineGrainedBitfieldAccesses)
-          // Fine-grained access, so no merging of spans.
-          InstallBest = true;
-        else if (!BitSizeSinceBegin)
+        // Assume clipped until proven not below.
+        BestClipped = true;
+        if (!BitSizeSinceBegin)
           // A zero-sized initial span -- this will install nothing and reset
           // for another.
           InstallBest = true;
@@ -624,6 +619,12 @@ CGRecordLowering::accumulateBitFields(bool isNonVirtualBaseType,
             // The access unit is not at a naturally aligned offset within the
             // structure.
             InstallBest = true;
+
+          if (InstallBest && BestEnd == Field)
+            // We're installing the first span, whose clipping was presumed
+            // above. Compute it correctly.
+            if (getSize(Type) == AccessSize)
+              BestClipped = false;
         }
 
         if (!InstallBest) {
@@ -656,11 +657,15 @@ CGRecordLowering::accumulateBitFields(bool isNonVirtualBaseType,
             // access unit.
             BestEndOffset = BeginOffset + TypeSize;
             BestEnd = Field;
+            BestClipped = false;
           }
 
           if (Barrier)
             // The next field is a barrier that we cannot merge across.
             InstallBest = true;
+          else if (Types.getCodeGenOpts().FineGrainedBitfieldAccesses)
+            // Fine-grained access, so no merging of spans.
+            InstallBest = true;
           else
             // Otherwise, we're not installing. Update the bit size
             // of the current span to go all the way to LimitOffset, which is
@@ -679,7 +684,17 @@ CGRecordLowering::accumulateBitFields(bool isNonVirtualBaseType,
         // Add the storage member for the access unit to the record. The
         // bitfields get the offset of their storage but come afterward and
         // remain there after a stable sort.
-        llvm::Type *Type = getIntNType(Context.toBits(AccessSize));
+        llvm::Type *Type;
+        if (BestClipped) {
+          assert(getSize(getIntNType(Context.toBits(AccessSize))) >
+                     AccessSize &&
+                 "Clipped access need not be clipped");
+          Type = getByteArrayType(AccessSize);
+        } else {
+          Type = getIntNType(Context.toBits(AccessSize));
+          assert(getSize(Type) == AccessSize &&
+                 "Unclipped access must be clipped");
+        }
         Members.push_back(StorageInfo(BeginOffset, Type));
         for (; Begin != BestEnd; ++Begin)
           if (!Begin->isZeroLengthBitField(Context))
@@ -934,32 +949,21 @@ void CGRecordLowering::calculateZeroInit() {
   }
 }
 
-void CGRecordLowering::clipTailPadding() {
-  std::vector<MemberInfo>::iterator Prior = Members.begin();
-  CharUnits Tail = getSize(Prior->Data);
-  for (std::vector<MemberInfo>::iterator Member = Prior + 1,
-                                         MemberEnd = Members.end();
-       Member != MemberEnd; ++Member) {
+// Verify accumulateBitfields computed the correct storage representations.
+void CGRecordLowering::checkBitfieldClipping() const {
+#ifndef NDEBUG
+  auto Tail = CharUnits::Zero();
+  for (const auto &M : Members) {
     // Only members with data and the scissor can cut into tail padding.
-    if (!Member->Data && Member->Kind != MemberInfo::Scissor)
+    if (!M.Data && M.Kind != MemberInfo::Scissor)
       continue;
-    if (Member->Offset < Tail) {
-      assert(Prior->Kind == MemberInfo::Field &&
-             "Only storage fields have tail padding!");
-      if (!Prior->FD || Prior->FD->isBitField())
-        Prior->Data = getByteArrayType(bitsToCharUnits(llvm::alignTo(
-            cast<llvm::IntegerType>(Prior->Data)->getIntegerBitWidth(), 8)));
-      else {
-        assert(Prior->FD->hasAttr<NoUniqueAddressAttr>() &&
-               "should not have reused this field's tail padding");
-        Prior->Data = getByteArrayType(
-            Context.getTypeInfoDataSizeInChars(Prior->FD->getType()).Width);
-      }
-    }
-    if (Member->Data)
-      Prior = Member;
-    Tail = Prior->Offset + getSize(Prior->Data);
+
+    assert(M.Offset >= Tail && "Bitfield access unit is not clipped");
+    Tail = M.Offset;
+    if (M.Data)
+      Tail += getSize(M.Data);
   }
+#endif
 }
 
 void CGRecordLowering::determinePacked(bool NVBaseType) {
diff --git a/clang/test/CodeGen/bitfield-access-unit.c b/clang/test/CodeGen/bitfield-access-unit.c
index 1aed2e7202fc6..d0553c5183eef 100644
--- a/clang/test/CodeGen/bitfield-access-unit.c
+++ b/clang/test/CodeGen/bitfield-access-unit.c
@@ -222,6 +222,24 @@ struct G {
 // LAYOUT-DWN32-NEXT: <CGBitFieldInfo Offset:{{[0-9]+}} Size:7 IsSigned:1 StorageSize:8 StorageOffset:4
 // CHECK-NEXT: ]>
 
+struct __attribute__((aligned(8))) H {
+  char a;
+  unsigned b : 24; // on expensive alignment we want this to stay 24
+  unsigned c __attribute__((aligned(8))); // Think 'long long' or lp64 ptr
+} h;
+// CHECK-LABEL: LLVMType:%struct.H =
+// LAYOUT-FLEX-SAME: type <{ i8, i32, [3 x i8], i32, [4 x i8] }>
+// LAYOUT-STRICT-SAME: type { i8, [3 x i8], [4 x i8], i32, [4 x i8] }
+// LAYOUT-DWN32-FLEX-SAME: type <{ i8, i32, [3 x i8], i32, [4 x i8] }>
+// LAYOUT-DWN32-STRICT-SAME: type { i8, [3 x i8], [4 x i8], i32, [4 x i8] }
+// CHECK: BitFields:[
+// LAYOUT-FLEX-NEXT: <CGBitFieldInfo Offset:{{[0-9]+}} Size:24 IsSigned:0 StorageSize:32 StorageOffset:1
+// LAYOUT-STRICT-NEXT: <CGBitFieldInfo Offset:{{[0-9]+}} Size:24 IsSigned:0 StorageSize:24 StorageOffset:1
+
+// LAYOUT-DWN32-FLEX-NEXT: <CGBitFieldInfo Offset:{{[0-9]+}} Size:24 IsSigned:0 StorageSize:32 StorageOffset:1
+// LAYOUT-DWN32-STRICT-NEXT: <CGBitFieldInfo Offset:{{[0-9]+}} Size:24 IsSigned:0 StorageSize:24 StorageOffset:1
+// CHECK-NEXT: ]>
+
 #if _LP64
 struct A64 {
   int a : 16;

From c303945409a740c8fdb4103a4f21df55187aa84f Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Mon, 15 Apr 2024 17:01:28 -0400
Subject: [PATCH 491/886] [gn] port e356f68 more

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 4383f1d6d18ff..865a79b63cd84 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -988,6 +988,7 @@ if (current_toolchain == default_toolchain) {
       "errno.h",
       "exception",
       "execution",
+      "expected",
       "experimental/__config",
       "experimental/__simd/aligned_tag.h",
       "experimental/__simd/declaration.h",

From a855eea7fe86ef09a87f6251b3b711b821ae32bf Mon Sep 17 00:00:00 2001
From: Jonas Devlieghere <jonas@devlieghere.com>
Date: Mon, 15 Apr 2024 14:05:01 -0700
Subject: [PATCH 492/886] [lldb] Fix the standalone Xcode build after #88317

In #88317, the clang resource headers was converted to an interface
library. Update LLDB and fix the Xcode standalone build. Thanks Evan for
the help!
---
 lldb/cmake/modules/LLDBFramework.cmake | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lldb/cmake/modules/LLDBFramework.cmake b/lldb/cmake/modules/LLDBFramework.cmake
index 81fc596ef4244..f915839f6b45a 100644
--- a/lldb/cmake/modules/LLDBFramework.cmake
+++ b/lldb/cmake/modules/LLDBFramework.cmake
@@ -119,7 +119,7 @@ add_custom_command(TARGET liblldb POST_BUILD
 if(NOT APPLE_EMBEDDED)
   if (TARGET clang-resource-headers)
     add_dependencies(liblldb clang-resource-headers)
-    set(clang_resource_headers_dir $<TARGET_PROPERTY:clang-resource-headers,RUNTIME_OUTPUT_DIRECTORY>)
+    set(clang_resource_headers_dir $<TARGET_PROPERTY:clang-resource-headers,INTERFACE_INCLUDE_DIRECTORIES>)
   else()
     set(clang_resource_headers_dir ${LLDB_EXTERNAL_CLANG_RESOURCE_DIR}/include)
     if(NOT EXISTS ${clang_resource_headers_dir})

From 21d177096f84c38cf434c21bd3ff0dbd2ca163d0 Mon Sep 17 00:00:00 2001
From: Kai Nacke <kai.peter.nacke@ibm.com>
Date: Mon, 15 Apr 2024 17:12:25 -0400
Subject: [PATCH 493/886] [NFC] Refactor looping over recomputeLiveIns into
 function (#88040)

https://github.com/llvm/llvm-project/pull/79940 put calls to
recomputeLiveIns into
a loop, to repeatedly call the function until the computation converges.
However,
this repeats a lot of code. This changes moves the loop into a function
to simplify
the handling.

Note that this changes the order in which recomputeLiveIns is called.
For example,

```
  bool anyChange = false;
  do {
    anyChange = recomputeLiveIns(*ExitMBB) || recomputeLiveIns(*LoopMBB);
  } while (anyChange);
```

only begins to recompute the live-ins for LoopMBB after the computation
for ExitMBB
has converged. With this change, all basic blocks have a recomputation
of the live-ins
for each loop iteration. This can result in less or more calls,
depending on the
situation.
---
 llvm/include/llvm/CodeGen/LivePhysRegs.h       | 18 ++++++++++++++++++
 llvm/lib/CodeGen/BranchFolding.cpp             |  8 ++------
 .../Target/AArch64/AArch64FrameLowering.cpp    |  5 +----
 llvm/lib/Target/AArch64/AArch64InstrInfo.cpp   | 11 ++---------
 llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp    |  8 +-------
 .../PowerPC/PPCExpandAtomicPseudoInsts.cpp     | 11 ++---------
 llvm/lib/Target/PowerPC/PPCFrameLowering.cpp   | 11 ++---------
 .../Target/SystemZ/SystemZFrameLowering.cpp    | 10 ++--------
 llvm/lib/Target/X86/X86FrameLowering.cpp       | 11 ++---------
 9 files changed, 32 insertions(+), 61 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/LivePhysRegs.h b/llvm/include/llvm/CodeGen/LivePhysRegs.h
index 1d40b1cbb0eaa..9574a6f0c7c00 100644
--- a/llvm/include/llvm/CodeGen/LivePhysRegs.h
+++ b/llvm/include/llvm/CodeGen/LivePhysRegs.h
@@ -39,6 +39,8 @@
 
 namespace llvm {
 
+template <typename T> class ArrayRef;
+
 class MachineInstr;
 class MachineFunction;
 class MachineOperand;
@@ -207,6 +209,22 @@ static inline bool recomputeLiveIns(MachineBasicBlock &MBB) {
   return oldLiveIns != newLiveIns;
 }
 
+/// Convenience function for recomputing live-in's for a set of MBBs until the
+/// computation converges.
+inline void fullyRecomputeLiveIns(ArrayRef<MachineBasicBlock *> MBBs) {
+  MachineBasicBlock *const *Data = MBBs.data();
+  const size_t Len = MBBs.size();
+  while (true) {
+    bool AnyChange = false;
+    for (size_t I = 0; I < Len; ++I)
+      if (recomputeLiveIns(*Data[I]))
+        AnyChange = true;
+    if (!AnyChange)
+      return;
+  }
+}
+
+
 } // end namespace llvm
 
 #endif // LLVM_CODEGEN_LIVEPHYSREGS_H
diff --git a/llvm/lib/CodeGen/BranchFolding.cpp b/llvm/lib/CodeGen/BranchFolding.cpp
index ecf7bc30913f5..55aa1d438b2a6 100644
--- a/llvm/lib/CodeGen/BranchFolding.cpp
+++ b/llvm/lib/CodeGen/BranchFolding.cpp
@@ -2047,12 +2047,8 @@ bool BranchFolder::HoistCommonCodeInSuccs(MachineBasicBlock *MBB) {
   MBB->splice(Loc, TBB, TBB->begin(), TIB);
   FBB->erase(FBB->begin(), FIB);
 
-  if (UpdateLiveIns) {
-    bool anyChange = false;
-    do {
-      anyChange = recomputeLiveIns(*TBB) || recomputeLiveIns(*FBB);
-    } while (anyChange);
-  }
+  if (UpdateLiveIns)
+    fullyRecomputeLiveIns({TBB, FBB});
 
   ++NumHoist;
   return true;
diff --git a/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp b/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
index 5cc612e89162a..419c141121c32 100644
--- a/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
@@ -4325,10 +4325,7 @@ AArch64FrameLowering::inlineStackProbeLoopExactMultiple(
   ExitMBB->transferSuccessorsAndUpdatePHIs(&MBB);
   MBB.addSuccessor(LoopMBB);
   // Update liveins.
-  bool anyChange = false;
-  do {
-    anyChange = recomputeLiveIns(*ExitMBB) || recomputeLiveIns(*LoopMBB);
-  } while (anyChange);
+  fullyRecomputeLiveIns({ExitMBB, LoopMBB});
 
   return ExitMBB->begin();
 }
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index 92647cb405252..9518d573bccdd 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -9556,15 +9556,8 @@ AArch64InstrInfo::probedStackAlloc(MachineBasicBlock::iterator MBBI,
   MBB.addSuccessor(LoopTestMBB);
 
   // Update liveins.
-  if (MF.getRegInfo().reservedRegsFrozen()) {
-    bool anyChange = false;
-    do {
-      anyChange = recomputeLiveIns(*ExitMBB) ||
-                  recomputeLiveIns(*LoopBodyMBB) ||
-                  recomputeLiveIns(*LoopTestMBB);
-    } while (anyChange);
-    ;
-  }
+  if (MF.getRegInfo().reservedRegsFrozen())
+    fullyRecomputeLiveIns({ExitMBB, LoopBodyMBB, LoopTestMBB});
 
   return ExitMBB->begin();
 }
diff --git a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
index 8629551152cb6..ea5dd5427ce72 100644
--- a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
+++ b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
@@ -1806,13 +1806,7 @@ void ARMLowOverheadLoops::Expand(LowOverheadLoop &LoLoop) {
   PostOrderLoopTraversal DFS(LoLoop.ML, *MLI);
   DFS.ProcessLoop();
   const SmallVectorImpl<MachineBasicBlock*> &PostOrder = DFS.getOrder();
-  bool anyChange = false;
-  do {
-    anyChange = false;
-    for (auto *MBB : PostOrder) {
-      anyChange = recomputeLiveIns(*MBB) || anyChange;
-    }
-  } while (anyChange);
+  fullyRecomputeLiveIns(PostOrder);
 
   for (auto *MBB : reverse(PostOrder))
     recomputeLivenessFlags(*MBB);
diff --git a/llvm/lib/Target/PowerPC/PPCExpandAtomicPseudoInsts.cpp b/llvm/lib/Target/PowerPC/PPCExpandAtomicPseudoInsts.cpp
index b43eee8fdd8c0..b3cfcb2aa1440 100644
--- a/llvm/lib/Target/PowerPC/PPCExpandAtomicPseudoInsts.cpp
+++ b/llvm/lib/Target/PowerPC/PPCExpandAtomicPseudoInsts.cpp
@@ -208,10 +208,7 @@ bool PPCExpandAtomicPseudo::expandAtomicRMW128(
       .addMBB(LoopMBB);
   CurrentMBB->addSuccessor(LoopMBB);
   CurrentMBB->addSuccessor(ExitMBB);
-  bool anyChange = false;
-  do {
-    anyChange = recomputeLiveIns(*ExitMBB) || recomputeLiveIns(*LoopMBB);
-  } while (anyChange);
+  fullyRecomputeLiveIns({ExitMBB, LoopMBB});
   NMBBI = MBB.end();
   MI.eraseFromParent();
   return true;
@@ -288,11 +285,7 @@ bool PPCExpandAtomicPseudo::expandAtomicCmpSwap128(
   CurrentMBB->addSuccessor(LoopCmpMBB);
   CurrentMBB->addSuccessor(ExitMBB);
 
-  bool anyChange = false;
-  do {
-    anyChange = recomputeLiveIns(*ExitMBB) || recomputeLiveIns(*CmpSuccMBB) ||
-                recomputeLiveIns(*LoopCmpMBB);
-  } while (anyChange);
+  fullyRecomputeLiveIns({ExitMBB, CmpSuccMBB, LoopCmpMBB});
   NMBBI = MBB.end();
   MI.eraseFromParent();
   return true;
diff --git a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
index 6dcb59a3a57f8..04e9f9e2366ed 100644
--- a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
@@ -1435,11 +1435,7 @@ void PPCFrameLowering::inlineStackProbe(MachineFunction &MF,
       ProbeLoopBodyMBB->addSuccessor(ProbeLoopBodyMBB);
     }
     // Update liveins.
-    bool anyChange = false;
-    do {
-      anyChange = recomputeLiveIns(*ProbeExitMBB) ||
-                  recomputeLiveIns(*ProbeLoopBodyMBB);
-    } while (anyChange);
+    fullyRecomputeLiveIns({ProbeExitMBB, ProbeLoopBodyMBB});
     return ProbeExitMBB;
   };
   // For case HasBP && MaxAlign > 1, we have to realign the SP by performing
@@ -1531,10 +1527,7 @@ void PPCFrameLowering::inlineStackProbe(MachineFunction &MF,
         buildDefCFAReg(*ExitMBB, ExitMBB->begin(), SPReg);
       }
       // Update liveins.
-      bool anyChange = false;
-      do {
-        anyChange = recomputeLiveIns(*ExitMBB) || recomputeLiveIns(*LoopMBB);
-      } while (anyChange);
+      fullyRecomputeLiveIns({ExitMBB, LoopMBB});
     }
   }
   ++NumPrologProbed;
diff --git a/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp b/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
index 4897b37d8eb1e..50ecd6e074414 100644
--- a/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
@@ -824,10 +824,7 @@ void SystemZELFFrameLowering::inlineStackProbe(
   StackAllocMI->eraseFromParent();
   if (DoneMBB != nullptr) {
     // Compute the live-in lists for the new blocks.
-    bool anyChange = false;
-    do {
-      anyChange = recomputeLiveIns(*DoneMBB) || recomputeLiveIns(*LoopMBB);
-    } while (anyChange);
+    fullyRecomputeLiveIns({DoneMBB, LoopMBB});
   }
 }
 
@@ -1425,10 +1422,7 @@ void SystemZXPLINKFrameLowering::inlineStackProbe(
   StackAllocMI->eraseFromParent();
 
   // Compute the live-in lists for the new blocks.
-  bool anyChange = false;
-  do {
-    anyChange = recomputeLiveIns(*StackExtMBB) || recomputeLiveIns(*NextMBB);
-  } while (anyChange);
+  fullyRecomputeLiveIns({StackExtMBB, NextMBB});
 }
 
 bool SystemZXPLINKFrameLowering::hasFP(const MachineFunction &MF) const {
diff --git a/llvm/lib/Target/X86/X86FrameLowering.cpp b/llvm/lib/Target/X86/X86FrameLowering.cpp
index d914e1b61ab07..4521401d8741c 100644
--- a/llvm/lib/Target/X86/X86FrameLowering.cpp
+++ b/llvm/lib/Target/X86/X86FrameLowering.cpp
@@ -885,10 +885,7 @@ void X86FrameLowering::emitStackProbeInlineGenericLoop(
   }
 
   // Update Live In information
-  bool anyChange = false;
-  do {
-    anyChange = recomputeLiveIns(*tailMBB) || recomputeLiveIns(*testMBB);
-  } while (anyChange);
+  fullyRecomputeLiveIns({tailMBB, testMBB});
 }
 
 void X86FrameLowering::emitStackProbeInlineWindowsCoreCLR64(
@@ -1380,11 +1377,7 @@ void X86FrameLowering::BuildStackAlignAND(MachineBasicBlock &MBB,
         footMBB->addSuccessor(&MBB);
       }
 
-      bool anyChange = false;
-      do {
-        anyChange = recomputeLiveIns(*footMBB) || recomputeLiveIns(*bodyMBB) ||
-                    recomputeLiveIns(*headMBB) || recomputeLiveIns(MBB);
-      } while (anyChange);
+      fullyRecomputeLiveIns({footMBB, bodyMBB, headMBB, &MBB});
     }
   } else {
     MachineInstr *MI = BuildMI(MBB, MBBI, DL, TII.get(AndOp), Reg)

From 67571ffd2c51a72a23d57fb5ef746a6fadd6b09c Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Mon, 15 Apr 2024 15:00:30 -0700
Subject: [PATCH 494/886] [test][sanitizer] Compile .c file as C

---
 .../test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c     | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c b/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
index b14ac7bcf1924..a6aed77d5691b 100644
--- a/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
+++ b/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
@@ -1,4 +1,4 @@
-// RUN: %clangxx %s -pie -fPIE -o %t && %run %t
+// RUN: %clang %s -pie -fPIE -o %t && %run %t
 // REQUIRES: x86_64-target-arch
 
 #include <assert.h>

From a1ed652fb3589670180c08c2c6d5ef1ff337c658 Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Mon, 15 Apr 2024 15:01:09 -0700
Subject: [PATCH 495/886] [test][sanitizer] Temporarily disable test

Test, as expected, fails with Asan on system with 5lvl page tables.
Disabling the test to migrate buildbot.
---
 .../test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c    | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c b/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
index a6aed77d5691b..02220cb78e6c9 100644
--- a/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
+++ b/compiler-rt/test/sanitizer_common/TestCases/Linux/mmap_56bit_test.c
@@ -1,6 +1,9 @@
 // RUN: %clang %s -pie -fPIE -o %t && %run %t
 // REQUIRES: x86_64-target-arch
 
+// FIXME: Fails Asan, as expected, with 5lvl page tables.
+// UNSUPPORTED: x86_64-target-arch
+
 #include <assert.h>
 #include <stdio.h>
 #include <sys/mman.h>

From 63310243537ba8830f3533a5d93e7b04b10d6c9e Mon Sep 17 00:00:00 2001
From: Argyrios Kyrtzidis <kyrtzidis@apple.com>
Date: Mon, 15 Apr 2024 15:05:55 -0700
Subject: [PATCH 496/886] [clang/DependencyScanning/ModuleDepCollector]
 Refactor part of `makeCommonInvocationForModuleBuild` into its own function
 (#88447)

The new function is about clearing out benign codegen options and can be
applied for PCH invocations as well.
---
 .../DependencyScanning/ModuleDepCollector.h   |  5 +++
 .../DependencyScanning/ModuleDepCollector.cpp | 36 ++++++++++++-------
 clang/test/ClangScanDeps/removed-args.c       | 28 +++++++++++++++
 3 files changed, 57 insertions(+), 12 deletions(-)

diff --git a/clang/include/clang/Tooling/DependencyScanning/ModuleDepCollector.h b/clang/include/clang/Tooling/DependencyScanning/ModuleDepCollector.h
index 081899cc2c850..da51292296a90 100644
--- a/clang/include/clang/Tooling/DependencyScanning/ModuleDepCollector.h
+++ b/clang/include/clang/Tooling/DependencyScanning/ModuleDepCollector.h
@@ -308,6 +308,11 @@ class ModuleDepCollector final : public DependencyCollector {
                                 ModuleDeps &Deps);
 };
 
+/// Resets codegen options that don't affect modules/PCH.
+void resetBenignCodeGenOptions(frontend::ActionKind ProgramAction,
+                               const LangOptions &LangOpts,
+                               CodeGenOptions &CGOpts);
+
 } // end namespace dependencies
 } // end namespace tooling
 } // end namespace clang
diff --git a/clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp b/clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp
index 94ccbd3351b09..e19f19b2528c1 100644
--- a/clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp
+++ b/clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp
@@ -154,6 +154,26 @@ void ModuleDepCollector::addOutputPaths(CowCompilerInvocation &CI,
   }
 }
 
+void dependencies::resetBenignCodeGenOptions(frontend::ActionKind ProgramAction,
+                                             const LangOptions &LangOpts,
+                                             CodeGenOptions &CGOpts) {
+  // TODO: Figure out better way to set options to their default value.
+  if (ProgramAction == frontend::GenerateModule) {
+    CGOpts.MainFileName.clear();
+    CGOpts.DwarfDebugFlags.clear();
+  }
+  if (ProgramAction == frontend::GeneratePCH ||
+      (ProgramAction == frontend::GenerateModule && !LangOpts.ModulesCodegen)) {
+    CGOpts.DebugCompilationDir.clear();
+    CGOpts.CoverageCompilationDir.clear();
+    CGOpts.CoverageDataFile.clear();
+    CGOpts.CoverageNotesFile.clear();
+    CGOpts.ProfileInstrumentUsePath.clear();
+    CGOpts.SampleProfileFile.clear();
+    CGOpts.ProfileRemappingFile.clear();
+  }
+}
+
 static CowCompilerInvocation
 makeCommonInvocationForModuleBuild(CompilerInvocation CI) {
   CI.resetNonModularOptions();
@@ -167,18 +187,8 @@ makeCommonInvocationForModuleBuild(CompilerInvocation CI) {
   // LLVM options are not going to affect the AST
   CI.getFrontendOpts().LLVMArgs.clear();
 
-  // TODO: Figure out better way to set options to their default value.
-  CI.getCodeGenOpts().MainFileName.clear();
-  CI.getCodeGenOpts().DwarfDebugFlags.clear();
-  if (!CI.getLangOpts().ModulesCodegen) {
-    CI.getCodeGenOpts().DebugCompilationDir.clear();
-    CI.getCodeGenOpts().CoverageCompilationDir.clear();
-    CI.getCodeGenOpts().CoverageDataFile.clear();
-    CI.getCodeGenOpts().CoverageNotesFile.clear();
-    CI.getCodeGenOpts().ProfileInstrumentUsePath.clear();
-    CI.getCodeGenOpts().SampleProfileFile.clear();
-    CI.getCodeGenOpts().ProfileRemappingFile.clear();
-  }
+  resetBenignCodeGenOptions(frontend::GenerateModule, CI.getLangOpts(),
+                            CI.getCodeGenOpts());
 
   // Map output paths that affect behaviour to "-" so their existence is in the
   // context hash. The final path will be computed in addOutputPaths.
@@ -342,6 +352,8 @@ static bool needsModules(FrontendInputFile FIF) {
 
 void ModuleDepCollector::applyDiscoveredDependencies(CompilerInvocation &CI) {
   CI.clearImplicitModuleBuildOptions();
+  resetBenignCodeGenOptions(CI.getFrontendOpts().ProgramAction,
+                            CI.getLangOpts(), CI.getCodeGenOpts());
 
   if (llvm::any_of(CI.getFrontendOpts().Inputs, needsModules)) {
     Preprocessor &PP = ScanInstance.getPreprocessor();
diff --git a/clang/test/ClangScanDeps/removed-args.c b/clang/test/ClangScanDeps/removed-args.c
index f49e4ead82f7b..3e108f0549450 100644
--- a/clang/test/ClangScanDeps/removed-args.c
+++ b/clang/test/ClangScanDeps/removed-args.c
@@ -93,3 +93,31 @@
 // CHECK-NOT:          "-fmodules-prune-interval=
 // CHECK-NOT:          "-fmodules-prune-after=
 // CHECK:            ],
+
+// Check for removed args for PCH invocations.
+
+// RUN: split-file %s %t
+// RUN: sed "s|DIR|%/t|g" %t/cdb-pch.json.template > %t/cdb-pch.json
+// RUN: clang-scan-deps -compilation-database %t/cdb-pch.json -format experimental-full > %t/result-pch.json
+// RUN: cat %t/result-pch.json | sed 's:\\\\\?:/:g' | FileCheck %s -DPREFIX=%/t -check-prefix=PCH
+//
+// PCH-NOT:          "-fdebug-compilation-dir="
+// PCH-NOT:          "-fcoverage-compilation-dir="
+// PCH-NOT:          "-coverage-notes-file
+// PCH-NOT:          "-coverage-data-file
+// PCH-NOT:          "-fprofile-instrument-use-path
+// PCH-NOT:          "-include"
+// PCH-NOT:          "-fmodules-cache-path=
+// PCH-NOT:          "-fmodules-validate-once-per-build-session"
+// PCH-NOT:          "-fbuild-session-timestamp=
+// PCH-NOT:          "-fmodules-prune-interval=
+// PCH-NOT:          "-fmodules-prune-after=
+
+//--- cdb-pch.json.template
+[
+  {
+    "directory": "DIR",
+    "command": "clang -x c-header DIR/header.h -fmodules -fimplicit-module-maps -fmodules-cache-path=DIR/cache -fdebug-compilation-dir=DIR/debug -fcoverage-compilation-dir=DIR/coverage -ftest-coverage -fprofile-instr-use=DIR/tu.profdata -o DIR/header.h.pch -serialize-diagnostics DIR/header.h.pch.diag ",
+    "file": "DIR/header.h.pch"
+  }
+]

From 6d234638f90bd422078c93745bdee73d6de201bf Mon Sep 17 00:00:00 2001
From: Fr4nk1in <sh.fu@outlook.com>
Date: Tue, 16 Apr 2024 06:13:38 +0800
Subject: [PATCH 497/886] [docs][mlir] Fix broken links in 'llvm' dialects.
 (#88704)

Links to `llvm.mlir.global` and `llvm.mlir.addressof` in the ["Globals"
section of LLVM dialect
documentation](https://mlir.llvm.org/docs/Dialects/LLVM/#globals) are
broken.
---
 mlir/docs/Dialects/LLVM.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mlir/docs/Dialects/LLVM.md b/mlir/docs/Dialects/LLVM.md
index a49ba35db9a68..ba466aa6bc401 100644
--- a/mlir/docs/Dialects/LLVM.md
+++ b/mlir/docs/Dialects/LLVM.md
@@ -139,12 +139,12 @@ will be reevaluated after considering composite constants.
 ### Globals
 
 Global variables are also defined using a special operation,
-[`llvm.mlir.global`](#llvmmlirglobal-mlirllvmglobalop), located at the module
+[`llvm.mlir.global`](#llvmmlirglobal-llvmglobalop), located at the module
 level. Globals are MLIR symbols and are identified by their name.
 
 Since functions need to be isolated-from-above, i.e. values defined outside the
 function cannot be directly used inside the function, an additional operation,
-[`llvm.mlir.addressof`](#llvmmliraddressof-mlirllvmaddressofop), is provided to
+[`llvm.mlir.addressof`](#llvmmliraddressof-llvmaddressofop), is provided to
 locally define a value containing the _address_ of a global. The actual value
 can then be loaded from that pointer, or a new value can be stored into it if
 the global is not declared constant. This is similar to LLVM IR where globals

From 206acf72c3b6c23e77716ccfc55ff94a4e7a7e3e Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Mon, 15 Apr 2024 18:48:49 -0400
Subject: [PATCH 498/886] [gn] port 8a7846fe86f95 (C++23 for libcxx, libcxxabi)

---
 llvm/utils/gn/secondary/libcxx/src/BUILD.gn    | 2 +-
 llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
index 5da8db4574a0c..1f6879358f22b 100644
--- a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
@@ -49,7 +49,7 @@ config("cxx_config") {
     "-Wno-covered-switch-default",
   ]
   cflags_cc = [
-    "-std=c++20",
+    "-std=c++23",
     "-nostdinc++",
   ]
   defines = [
diff --git a/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn b/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
index c82634e2bb064..7a923c5c854d7 100644
--- a/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
@@ -66,7 +66,7 @@ config("cxxabi_config") {
     "//libcxx/src",
   ]
   cflags_cc = [
-    "-std=c++20",
+    "-std=c++23",
     "-nostdinc++",
   ]
   defines = [

From 466017c8dab74f66ce513c8752f0c1dcd16a8a63 Mon Sep 17 00:00:00 2001
From: Adrian Prantl <aprantl@apple.com>
Date: Mon, 15 Apr 2024 15:50:07 -0700
Subject: [PATCH 499/886] Work around test failure due to new aslr default

---
 .../TestDiagnoseDereferenceFunctionReturn.py                   | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/lldb/test/API/commands/frame/diagnose/dereference-function-return/TestDiagnoseDereferenceFunctionReturn.py b/lldb/test/API/commands/frame/diagnose/dereference-function-return/TestDiagnoseDereferenceFunctionReturn.py
index d8f45161378b0..4d9b036f5102c 100644
--- a/lldb/test/API/commands/frame/diagnose/dereference-function-return/TestDiagnoseDereferenceFunctionReturn.py
+++ b/lldb/test/API/commands/frame/diagnose/dereference-function-return/TestDiagnoseDereferenceFunctionReturn.py
@@ -19,6 +19,9 @@ def test_diagnose_dereference_function_return(self):
         TestBase.setUp(self)
         self.build()
         exe = self.getBuildArtifact("a.out")
+        # FIXME: This default changed in lldbtest.py and this test
+        # seems to rely on having it turned off.
+        self.runCmd("settings set target.disable-aslr true")
         self.runCmd("file " + exe, CURRENT_EXECUTABLE_SET)
         self.runCmd("run", RUN_SUCCEEDED)
         self.expect("thread list", "Thread should be stopped", substrs=["stopped"])

From 9cb755cf5b2f48117fa34f257f386acf59ad6397 Mon Sep 17 00:00:00 2001
From: Wu Yingcong <yingcong.wu@intel.com>
Date: Mon, 15 Apr 2024 16:37:35 -0700
Subject: [PATCH 500/886] [Test][JITLink] Save rbx in
 ExecutionEngine/JITLink/x86-64/ELF_vtune.s (#86472)

The callee should preserve rbx according to the calling convention, but
it is not in the test case `ExecutionEngine/JITLink/x86-64/ELF_vtune.s`.
Not preserving the rbx register may result in some random error to the
caller function. This patch adds the missing command to preserve the
rbx.
---
 llvm/test/ExecutionEngine/JITLink/x86-64/ELF_vtune.s | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/llvm/test/ExecutionEngine/JITLink/x86-64/ELF_vtune.s b/llvm/test/ExecutionEngine/JITLink/x86-64/ELF_vtune.s
index 1c95bde51e121..936486b8a319c 100644
--- a/llvm/test/ExecutionEngine/JITLink/x86-64/ELF_vtune.s
+++ b/llvm/test/ExecutionEngine/JITLink/x86-64/ELF_vtune.s
@@ -20,12 +20,14 @@ main:
         .cfi_def_cfa_offset 16
         .cfi_offset 6, -16
         movq    %rsp, %rbp
+        pushq   %rbx
         .cfi_def_cfa_register 6
-        movl    %edi, -4(%rbp)
-        movq    %rsi, -16(%rbp)
-        movl    -4(%rbp), %ebx
+        movl    %edi, -16(%rbp)
+        movq    %rsi, -24(%rbp)
+        movl    -16(%rbp), %ebx
         addl    $1, %ebx
-	movl   $0, %eax
+        movl    $0, %eax
+        popq    %rbx
         popq    %rbp
         .cfi_def_cfa 7, 8
         ret

From c50f7e9a425bfa4ab8655c79a715c88ed3b1e830 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Tue, 16 Apr 2024 09:17:52 +0900
Subject: [PATCH 501/886] [InstCombine] Remove mul of SPF abs fold (#88675)

Remove the fold working on abs in SPF representation now that we
canonicalize SPF to intrinsics.

This is not strictly NFC because the SPF fold might fire for
non-canonical IR due to multi-use, but given the lack of test coverage,
I assume this is not important.
---
 .../InstCombine/InstCombineMulDivRem.cpp        | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 4dc1319f1c437..48372381a0d1c 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -319,19 +319,12 @@ Instruction *InstCombinerImpl::visitMul(BinaryOperator &I) {
   }
 
   // abs(X) * abs(X) -> X * X
-  // nabs(X) * nabs(X) -> X * X
-  if (Op0 == Op1) {
-    Value *X, *Y;
-    SelectPatternFlavor SPF = matchSelectPattern(Op0, X, Y).Flavor;
-    if (SPF == SPF_ABS || SPF == SPF_NABS)
-      return BinaryOperator::CreateMul(X, X);
-
-    if (match(Op0, m_Intrinsic<Intrinsic::abs>(m_Value(X))))
-      return BinaryOperator::CreateMul(X, X);
-  }
+  Value *X;
+  if (Op0 == Op1 && match(Op0, m_Intrinsic<Intrinsic::abs>(m_Value(X))))
+    return BinaryOperator::CreateMul(X, X);
 
   {
-    Value *X, *Y;
+    Value *Y;
     // abs(X) * abs(Y) -> abs(X * Y)
     if (I.hasNoSignedWrap() &&
         match(Op0,
@@ -344,7 +337,7 @@ Instruction *InstCombinerImpl::visitMul(BinaryOperator &I) {
   }
 
   // -X * C --> X * -C
-  Value *X, *Y;
+  Value *Y;
   Constant *Op1C;
   if (match(Op0, m_Neg(m_Value(X))) && match(Op1, m_Constant(Op1C)))
     return BinaryOperator::CreateMul(X, ConstantExpr::getNeg(Op1C));

From 2b06ff555aa32b316710b4708fbc16f36d6eab15 Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Mon, 15 Apr 2024 17:38:39 -0700
Subject: [PATCH 502/886] [RISCV] Expand mul to shNadd x, (slli x, c) in
 DAGCombine (#88524)

This expansion is directly inspired by the analogous code in the x86
backend for LEA. shXadd and (this sub-case of) LEA are largely
equivalent.

This is an alternative to
https://github.com/llvm/llvm-project/pull/87105.

This expansion is also supported via the decomposeMulByConstant
callback, but restricted because of interactions with other combines
since that code runs before legalization. As discussed in the other
review, my original plan had been to support post legalization expansion
through the same interface, but that ended up being more complicated
than seems justified.

Instead, lets go ahead and do the general expansion post-legalize. Other
targets use the combine approach, and matching that structure makes it
easier for us to adapt ideas from other targets to RISCV.
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 52 +++++++++++++++++--
 llvm/test/CodeGen/RISCV/addimm-mulimm.ll      |  9 ++--
 llvm/test/CodeGen/RISCV/rv32zba.ll            | 48 +++++++++++------
 .../CodeGen/RISCV/rv64-legal-i32/rv64zba.ll   | 48 +++++++++++------
 llvm/test/CodeGen/RISCV/rv64zba.ll            | 48 +++++++++++------
 5 files changed, 153 insertions(+), 52 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 27387595164a4..259cc388276c6 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -13363,10 +13363,56 @@ static SDValue performXORCombine(SDNode *N, SelectionDAG &DAG,
   return combineSelectAndUseCommutative(N, DAG, /*AllOnes*/ false, Subtarget);
 }
 
-static SDValue performMULCombine(SDNode *N, SelectionDAG &DAG) {
+// Try to expand a scalar multiply to a faster sequence.
+static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
+                         TargetLowering::DAGCombinerInfo &DCI,
+                         const RISCVSubtarget &Subtarget) {
+
   EVT VT = N->getValueType(0);
-  if (!VT.isVector())
+
+  // LI + MUL is usually smaller than the alternative sequence.
+  if (DAG.getMachineFunction().getFunction().hasMinSize())
+    return SDValue();
+
+  if (DCI.isBeforeLegalize() || DCI.isCalledByLegalizer())
+    return SDValue();
+
+  if (VT != Subtarget.getXLenVT())
+    return SDValue();
+
+  if (!Subtarget.hasStdExtZba())
+    return SDValue();
+
+  ConstantSDNode *CNode = dyn_cast<ConstantSDNode>(N->getOperand(1));
+  if (!CNode)
     return SDValue();
+  uint64_t MulAmt = CNode->getZExtValue();
+
+  // If this is a power 2 + 2/4/8, we can use a shift followed by a single
+  // shXadd. First check if this a sum of two power of 2s because that's
+  // easy. Then count how many zeros are up to the first bit.
+  if (isPowerOf2_64(MulAmt & (MulAmt - 1))) {
+    unsigned ScaleShift = llvm::countr_zero(MulAmt);
+    if (ScaleShift >= 1 && ScaleShift < 4) {
+      unsigned ShiftAmt = Log2_64((MulAmt & (MulAmt - 1)));
+      SDLoc DL(N);
+      SDValue Shift1 = DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0),
+                                   DAG.getConstant(ShiftAmt, DL, VT));
+      SDValue Shift2 = DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0),
+                                   DAG.getConstant(ScaleShift, DL, VT));
+      return DAG.getNode(ISD::ADD, DL, VT, Shift1, Shift2);
+    }
+  }
+  return SDValue();
+}
+
+
+static SDValue performMULCombine(SDNode *N, SelectionDAG &DAG,
+                                 TargetLowering::DAGCombinerInfo &DCI,
+                                 const RISCVSubtarget &Subtarget) {
+  EVT VT = N->getValueType(0);
+  if (!VT.isVector())
+    return expandMul(N, DAG, DCI, Subtarget);
 
   SDLoc DL(N);
   SDValue N0 = N->getOperand(0);
@@ -15913,7 +15959,7 @@ SDValue RISCVTargetLowering::PerformDAGCombine(SDNode *N,
   case ISD::MUL:
     if (SDValue V = combineBinOp_VLToVWBinOp_VL(N, DCI, Subtarget))
       return V;
-    return performMULCombine(N, DAG);
+    return performMULCombine(N, DAG, DCI, Subtarget);
   case ISD::SDIV:
   case ISD::UDIV:
   case ISD::SREM:
diff --git a/llvm/test/CodeGen/RISCV/addimm-mulimm.ll b/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
index 48fa69e104565..10103f071462c 100644
--- a/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
+++ b/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
@@ -551,8 +551,9 @@ define i64 @add_mul_combine_infinite_loop(i64 %x) {
 ; RV32IMB-NEXT:    sh3add a1, a1, a2
 ; RV32IMB-NEXT:    sh1add a0, a0, a0
 ; RV32IMB-NEXT:    slli a2, a0, 3
-; RV32IMB-NEXT:    addi a0, a2, 2047
-; RV32IMB-NEXT:    addi a0, a0, 1
+; RV32IMB-NEXT:    li a3, 1
+; RV32IMB-NEXT:    slli a3, a3, 11
+; RV32IMB-NEXT:    sh3add a0, a0, a3
 ; RV32IMB-NEXT:    sltu a2, a0, a2
 ; RV32IMB-NEXT:    add a1, a1, a2
 ; RV32IMB-NEXT:    ret
@@ -561,8 +562,8 @@ define i64 @add_mul_combine_infinite_loop(i64 %x) {
 ; RV64IMB:       # %bb.0:
 ; RV64IMB-NEXT:    addi a0, a0, 86
 ; RV64IMB-NEXT:    sh1add a0, a0, a0
-; RV64IMB-NEXT:    li a1, -16
-; RV64IMB-NEXT:    sh3add a0, a0, a1
+; RV64IMB-NEXT:    slli a0, a0, 3
+; RV64IMB-NEXT:    addi a0, a0, -16
 ; RV64IMB-NEXT:    ret
   %tmp0 = mul i64 %x, 24
   %tmp1 = add i64 %tmp0, 2048
diff --git a/llvm/test/CodeGen/RISCV/rv32zba.ll b/llvm/test/CodeGen/RISCV/rv32zba.ll
index 0908a393338c5..cc632a09c8054 100644
--- a/llvm/test/CodeGen/RISCV/rv32zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv32zba.ll
@@ -271,31 +271,49 @@ define i32 @mul288(i32 %a) {
 }
 
 define i32 @mul258(i32 %a) {
-; CHECK-LABEL: mul258:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 258
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV32I-LABEL: mul258:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 258
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32ZBA-LABEL: mul258:
+; RV32ZBA:       # %bb.0:
+; RV32ZBA-NEXT:    slli a1, a0, 8
+; RV32ZBA-NEXT:    sh1add a0, a0, a1
+; RV32ZBA-NEXT:    ret
   %c = mul i32 %a, 258
   ret i32 %c
 }
 
 define i32 @mul260(i32 %a) {
-; CHECK-LABEL: mul260:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 260
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV32I-LABEL: mul260:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 260
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32ZBA-LABEL: mul260:
+; RV32ZBA:       # %bb.0:
+; RV32ZBA-NEXT:    slli a1, a0, 8
+; RV32ZBA-NEXT:    sh2add a0, a0, a1
+; RV32ZBA-NEXT:    ret
   %c = mul i32 %a, 260
   ret i32 %c
 }
 
 define i32 @mul264(i32 %a) {
-; CHECK-LABEL: mul264:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 264
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV32I-LABEL: mul264:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 264
+; RV32I-NEXT:    mul a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32ZBA-LABEL: mul264:
+; RV32ZBA:       # %bb.0:
+; RV32ZBA-NEXT:    slli a1, a0, 8
+; RV32ZBA-NEXT:    sh3add a0, a0, a1
+; RV32ZBA-NEXT:    ret
   %c = mul i32 %a, 264
   ret i32 %c
 }
diff --git a/llvm/test/CodeGen/RISCV/rv64-legal-i32/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64-legal-i32/rv64zba.ll
index 90cfb1fdcb779..ee9b73ca82f21 100644
--- a/llvm/test/CodeGen/RISCV/rv64-legal-i32/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64-legal-i32/rv64zba.ll
@@ -811,31 +811,49 @@ define i64 @adduw_imm(i32 signext %0) nounwind {
 }
 
 define i64 @mul258(i64 %a) {
-; CHECK-LABEL: mul258:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 258
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul258:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 258
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul258:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh1add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 258
   ret i64 %c
 }
 
 define i64 @mul260(i64 %a) {
-; CHECK-LABEL: mul260:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 260
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul260:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 260
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul260:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh2add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 260
   ret i64 %c
 }
 
 define i64 @mul264(i64 %a) {
-; CHECK-LABEL: mul264:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 264
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul264:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 264
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul264:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh3add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 264
   ret i64 %c
 }
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index bb4be323ecb2e..0d1d4838c6113 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -834,31 +834,49 @@ define i64 @adduw_imm(i32 signext %0) nounwind {
 }
 
 define i64 @mul258(i64 %a) {
-; CHECK-LABEL: mul258:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 258
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul258:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 258
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul258:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh1add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 258
   ret i64 %c
 }
 
 define i64 @mul260(i64 %a) {
-; CHECK-LABEL: mul260:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 260
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul260:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 260
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul260:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh2add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 260
   ret i64 %c
 }
 
 define i64 @mul264(i64 %a) {
-; CHECK-LABEL: mul264:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 264
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul264:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 264
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul264:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    slli a1, a0, 8
+; RV64ZBA-NEXT:    sh3add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 264
   ret i64 %c
 }

From 40bbdb609f58d6cbbae1ca525832d7a21641a347 Mon Sep 17 00:00:00 2001
From: Alina Sbirlea <asbirlea@google.com>
Date: Mon, 15 Apr 2024 17:36:15 -0700
Subject: [PATCH 503/886] Revert "[DAG] Fold
 extract_subvector(insert_subvector(x,y,c1),c2) --> extract_subvector(y,c2-c1)
 (#87925)"

This reverts commit 8c0f52e9d5a99bf96bb64ac23b5893482c292527.

Reverting to green, reproducer attached in the PR/revision comments.
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |   17 -
 .../any_extend_vector_inreg_of_broadcast.ll   |   42 +-
 llvm/test/CodeGen/X86/dpbusd.ll               |    2 +-
 llvm/test/CodeGen/X86/dpbusd_i4.ll            |    2 +-
 .../vector-interleaved-load-i16-stride-3.ll   | 1300 +++++++++--------
 .../vector-interleaved-store-i8-stride-7.ll   |  602 ++++----
 .../zero_extend_vector_inreg_of_broadcast.ll  |   28 +-
 7 files changed, 1006 insertions(+), 987 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index cbba3a294b3d6..0fa0bf2609bb3 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -24467,23 +24467,6 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
       if (!LegalOperations || TLI.isOperationLegal(ISD::SPLAT_VECTOR, NVT))
         return DAG.getSplatVector(NVT, DL, V.getOperand(0));
 
-  // extract_subvector(insert_subvector(x,y,c1),c2)
-  //  --> extract_subvector(y,c2-c1)
-  // iff we're just extracting from the inserted subvector.
-  if (V.getOpcode() == ISD::INSERT_SUBVECTOR) {
-    SDValue InsSub = V.getOperand(1);
-    EVT InsSubVT = InsSub.getValueType();
-    unsigned NumInsElts = InsSubVT.getVectorMinNumElements();
-    unsigned InsIdx = V.getConstantOperandVal(2);
-    unsigned NumSubElts = NVT.getVectorMinNumElements();
-    if (InsIdx <= ExtIdx && (ExtIdx + NumSubElts) <= (InsIdx + NumInsElts) &&
-        TLI.isExtractSubvectorCheap(NVT, InsSubVT, ExtIdx - InsIdx)) {
-      SDLoc DL(N);
-      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NVT, InsSub,
-                         DAG.getVectorIdxConstant(ExtIdx - InsIdx, DL));
-    }
-  }
-
   // Try to move vector bitcast after extract_subv by scaling extraction index:
   // extract_subv (bitcast X), Index --> bitcast (extract_subv X, Index')
   if (V.getOpcode() == ISD::BITCAST &&
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
index 39c7ce1413d1b..4242d8483e723 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -3507,12 +3507,13 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm2
+; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
-; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512F-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3522,12 +3523,13 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512DQ-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3766,10 +3768,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -3782,10 +3784,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4145,9 +4147,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %xmm0
@@ -4159,9 +4161,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/dpbusd.ll b/llvm/test/CodeGen/X86/dpbusd.ll
index 04d7a9691b645..fbea08eb1e550 100644
--- a/llvm/test/CodeGen/X86/dpbusd.ll
+++ b/llvm/test/CodeGen/X86/dpbusd.ll
@@ -26,7 +26,7 @@ define i32 @no_dpbusd(ptr%a, ptr%b, i32 %c, i32 %n) {
 ; AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
 ; AVX512-NEXT:    vpmaddwd %ymm0, %ymm1, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/dpbusd_i4.ll b/llvm/test/CodeGen/X86/dpbusd_i4.ll
index a212f99680ef4..906fead7f8db5 100644
--- a/llvm/test/CodeGen/X86/dpbusd_i4.ll
+++ b/llvm/test/CodeGen/X86/dpbusd_i4.ll
@@ -86,7 +86,7 @@ define i32 @mul_sext_i4i4(<16 x i4> %a, <16 x i4> %b, i32 %c) {
 ; CHECK-NEXT:    vpsraw $12, %ymm0, %ymm0
 ; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 6d5fc9ed0ab5b..1436922f9dd11 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -1828,22 +1828,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-LABEL: load_i16_stride3_vf32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm1
+; AVX512-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm0
 ; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
 ; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm3
+; AVX512-NEXT:    vmovdqa %ymm1, %ymm3
 ; AVX512-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1857,14 +1857,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
+; AVX512-NEXT:    vmovdqa %ymm1, %ymm10
 ; AVX512-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
 ; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1885,19 +1885,21 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
-; AVX512-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
-; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
+; AVX512-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1906,22 +1908,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512-FCP:       # %bb.0:
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
+; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
 ; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm3
+; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm3
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1935,14 +1937,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
+; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm10
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1963,19 +1965,21 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1984,22 +1988,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm1
+; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm3
+; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm3
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2013,14 +2017,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
+; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm10
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2041,19 +2045,21 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
-; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
+; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -2062,22 +2068,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ-FCP:       # %bb.0:
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2091,14 +2097,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2119,19 +2125,21 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -3492,668 +3500,688 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-LABEL: load_i16_stride3_vf64:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm18
-; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm20
+; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm20
+; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm21
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
-; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm1
+; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
+; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm8
 ; AVX512-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512-NEXT:    vmovdqa %xmm2, %xmm3
-; AVX512-NEXT:    vmovdqa64 %xmm1, %xmm19
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
+; AVX512-NEXT:    vmovdqa %xmm2, %xmm14
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
 ; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm21
-; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm22
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm8
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
-; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
+; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm22
+; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm23
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm6
+; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
 ; AVX512-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-NEXT:    vmovdqa %xmm1, %xmm8
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
-; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm23
-; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm11
+; AVX512-NEXT:    vmovdqa %xmm1, %xmm6
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
+; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm24
+; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm13
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
+; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm15
-; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
-; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm24
-; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm12
+; AVX512-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm11
+; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm12
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
+; AVX512-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm5
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
-; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
-; AVX512-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
+; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
+; AVX512-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
+; AVX512-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
 ; AVX512-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm1
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
-; AVX512-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
-; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm25
+; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm15
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
+; AVX512-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
+; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX512-NEXT:    vmovdqa64 %ymm2, %ymm28
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm6, %xmm25
 ; AVX512-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-NEXT:    vmovdqa %ymm13, %ymm2
-; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-NEXT:    vmovdqa %ymm9, %ymm1
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-NEXT:    vmovdqa64 %xmm19, %xmm8
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm3, %xmm27
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
+; AVX512-NEXT:    vmovdqa %xmm14, %xmm7
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm27
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
 ; AVX512-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
-; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-NEXT:    vmovdqa %ymm13, %ymm6
-; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
-; AVX512-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
-; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512-NEXT:    vmovdqa64 %ymm28, %ymm3
+; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512-NEXT:    vmovdqa %ymm9, %ymm3
+; AVX512-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
+; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
+; AVX512-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
+; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
+; AVX512-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
-; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
-; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm7
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
-; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
+; AVX512-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
+; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
+; AVX512-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
+; AVX512-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
 ; AVX512-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm5
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
+; AVX512-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
+; AVX512-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
+; AVX512-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-NEXT:    vmovdqa64 %zmm17, (%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm18, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-NEXT:    vmovdqa64 %zmm2, (%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm1, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-NEXT:    vmovdqa64 %zmm1, (%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm2, (%rcx)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
-; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
+; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
+; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
-; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
+; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
 ; AVX512-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm3
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm14
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
-; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm8
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
+; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
+; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm6
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
 ; AVX512-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm8
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
-; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
-; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
+; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm6
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
+; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
+; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
-; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
-; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
+; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
 ; AVX512-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
+; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm2
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vmovdqa %xmm14, %xmm7
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
-; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm6
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
+; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm3
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
+; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
 ; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
+; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm18
-; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm20
+; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm20
+; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm21
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
-; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
+; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm8
 ; AVX512DQ-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm3
-; AVX512DQ-NEXT:    vmovdqa64 %xmm1, %xmm19
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm14
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm21
-; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm22
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm8
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
+; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm22
+; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm23
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm6
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
 ; AVX512DQ-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm8
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
-; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm23
-; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm11
+; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm6
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
+; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm24
+; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm13
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm15
-; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
-; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm24
-; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm12
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm11
+; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm12
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
+; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm5
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
-; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
+; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
 ; AVX512DQ-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm1
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm25
+; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm15
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vmovdqa64 %ymm2, %ymm28
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm6, %xmm25
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm2
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vmovdqa64 %xmm19, %xmm8
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm3, %xmm27
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vmovdqa %xmm14, %xmm7
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm27
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
-; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm6
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512DQ-NEXT:    vmovdqa64 %ymm28, %ymm3
+; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm3
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm7
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
+; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
 ; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm5
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
+; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vmovdqa64 %zmm17, (%rsi)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm18, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rdx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rcx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
-; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
+; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
+; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
-; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
+; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
 ; AVX512DQ-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm3
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm14
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
-; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm8
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
+; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
+; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm6
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
 ; AVX512DQ-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm8
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
-; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
-; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm6
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
+; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
+; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
-; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
-; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
+; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
+; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
 ; AVX512DQ-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
+; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm14, %xmm7
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm6
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm3
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
+; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
+; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 8091afbbfd70c..8b6ba51506ab7 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -1246,28 +1246,29 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
+; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
+; AVX512BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
+; AVX512BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
 ; AVX512BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
-; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
-; AVX512BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
-; AVX512BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
-; AVX512BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
+; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
+; AVX512BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
+; AVX512BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
+; AVX512BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
 ; AVX512BW-FCP-NEXT:    vzeroupper
 ; AVX512BW-FCP-NEXT:    retq
 ;
@@ -1325,28 +1326,29 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512DQ-BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
+; AVX512DQ-BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
 ; AVX512DQ-BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
-; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
+; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
+; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
 ; AVX512DQ-BW-FCP-NEXT:    vzeroupper
 ; AVX512DQ-BW-FCP-NEXT:    retq
   %in.vec0 = load <8 x i8>, ptr %in.vecptr0, align 64
@@ -2051,76 +2053,77 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512-NEXT:    vmovdqa (%rsi), %xmm1
-; AVX512-NEXT:    vmovdqa (%rdx), %xmm5
-; AVX512-NEXT:    vmovdqa (%rcx), %xmm6
-; AVX512-NEXT:    vmovdqa (%r8), %xmm3
-; AVX512-NEXT:    vmovdqa (%r9), %xmm4
-; AVX512-NEXT:    vmovdqa (%r10), %xmm2
-; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
-; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
-; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
-; AVX512-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
-; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
-; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
-; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
-; AVX512-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
-; AVX512-NEXT:    vporq %zmm10, %zmm11, %zmm10
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
-; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
-; AVX512-NEXT:    vpandn %ymm12, %ymm13, %ymm12
-; AVX512-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
-; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
-; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
-; AVX512-NEXT:    vporq %zmm12, %zmm11, %zmm11
-; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
-; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
-; AVX512-NEXT:    vpor %ymm8, %ymm9, %ymm8
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
+; AVX512-NEXT:    vmovdqa (%rdi), %xmm4
+; AVX512-NEXT:    vmovdqa (%rsi), %xmm5
+; AVX512-NEXT:    vmovdqa (%rdx), %xmm6
+; AVX512-NEXT:    vmovdqa (%rcx), %xmm7
+; AVX512-NEXT:    vmovdqa (%r8), %xmm0
+; AVX512-NEXT:    vmovdqa (%r10), %xmm1
+; AVX512-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
+; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
+; AVX512-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
+; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
+; AVX512-NEXT:    vpor %xmm6, %xmm4, %xmm4
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
+; AVX512-NEXT:    vpor %xmm6, %xmm5, %xmm5
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
+; AVX512-NEXT:    vpor %ymm4, %ymm6, %ymm4
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
 ; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
-; AVX512-NEXT:    vpor %xmm5, %xmm0, %xmm0
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
-; AVX512-NEXT:    vpor %xmm2, %xmm1, %xmm1
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
-; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
-; AVX512-NEXT:    vmovdqa %xmm1, 96(%rax)
-; AVX512-NEXT:    vmovdqa %ymm0, 64(%rax)
-; AVX512-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
+; AVX512-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
+; AVX512-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
+; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
+; AVX512-NEXT:    vporq %zmm3, %zmm2, %zmm2
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512-NEXT:    vpandn %ymm3, %ymm6, %ymm3
+; AVX512-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
+; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
+; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
+; AVX512-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
+; AVX512-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
+; AVX512-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
+; AVX512-NEXT:    vmovdqa %xmm5, 96(%rax)
+; AVX512-NEXT:    vmovdqa64 %zmm0, (%rax)
+; AVX512-NEXT:    vmovdqa %ymm4, 64(%rax)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
@@ -2128,69 +2131,70 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm1
-; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm5
-; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm6
-; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm3
-; AVX512-FCP-NEXT:    vmovdqa (%r9), %xmm4
-; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm2
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
-; AVX512-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
-; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
-; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
-; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
-; AVX512-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
-; AVX512-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
-; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
-; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
-; AVX512-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
-; AVX512-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
-; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
-; AVX512-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
-; AVX512-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
-; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
-; AVX512-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm2
+; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm3
+; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm4
+; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm5
+; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm1
+; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm0
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
+; AVX512-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
+; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
+; AVX512-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
+; AVX512-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
+; AVX512-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
+; AVX512-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
+; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
+; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
+; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
+; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
+; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
+; AVX512-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
+; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
+; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
+; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
+; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
+; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
+; AVX512-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
+; AVX512-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
@@ -2198,76 +2202,77 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm1
-; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm5
-; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm6
-; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm3
-; AVX512DQ-NEXT:    vmovdqa (%r9), %xmm4
-; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm2
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512DQ-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
-; AVX512DQ-NEXT:    vporq %zmm10, %zmm11, %zmm10
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
-; AVX512DQ-NEXT:    vpandn %ymm12, %ymm13, %ymm12
-; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
-; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
-; AVX512DQ-NEXT:    vporq %zmm12, %zmm11, %zmm11
-; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
-; AVX512DQ-NEXT:    vpor %ymm8, %ymm9, %ymm8
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm4
+; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm5
+; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm6
+; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm7
+; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm1
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
+; AVX512DQ-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
+; AVX512DQ-NEXT:    vpor %xmm6, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
+; AVX512DQ-NEXT:    vpor %xmm6, %xmm5, %xmm5
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
+; AVX512DQ-NEXT:    vpor %ymm4, %ymm6, %ymm4
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
-; AVX512DQ-NEXT:    vpor %xmm5, %xmm0, %xmm0
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
-; AVX512DQ-NEXT:    vpor %xmm2, %xmm1, %xmm1
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
-; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
-; AVX512DQ-NEXT:    vmovdqa %xmm1, 96(%rax)
-; AVX512DQ-NEXT:    vmovdqa %ymm0, 64(%rax)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
+; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512DQ-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
+; AVX512DQ-NEXT:    vporq %zmm3, %zmm2, %zmm2
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512DQ-NEXT:    vpandn %ymm3, %ymm6, %ymm3
+; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
+; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vporq %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
+; AVX512DQ-NEXT:    vmovdqa %xmm5, 96(%rax)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rax)
+; AVX512DQ-NEXT:    vmovdqa %ymm4, 64(%rax)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
@@ -2275,69 +2280,70 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm1
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm5
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm6
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm3
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r9), %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm2
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
-; AVX512DQ-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
-; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512DQ-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
-; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
-; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
-; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
-; AVX512DQ-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
-; AVX512DQ-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
-; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
-; AVX512DQ-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
-; AVX512DQ-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
-; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm2
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm3
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm5
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm1
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm0
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
+; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
+; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
+; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
+; AVX512DQ-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
+; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
+; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
+; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
+; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
+; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
+; AVX512DQ-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
+; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
+; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index 99e8cdb179c8d..11f422d671541 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -4026,10 +4026,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F-FAST:       # %bb.0:
-; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512F-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4062,10 +4062,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ-FAST:       # %bb.0:
-; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4541,9 +4541,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
-; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4559,9 +4559,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
-; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0

From 694c444b5bbb56dcba8978d283fe5385237c309a Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Mon, 15 Apr 2024 20:58:08 -0400
Subject: [PATCH 504/886] [gn] port 311ff3917827 more

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index 865a79b63cd84..ee44558a4e994 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -982,6 +982,7 @@ if (current_toolchain == default_toolchain) {
       "ctgmath",
       "ctime",
       "ctype.h",
+      "cuchar",
       "cwchar",
       "cwctype",
       "deque",

From 2ac562ab784c6bf04e6d3026c567d3552a735668 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Mon, 15 Apr 2024 17:58:39 -0700
Subject: [PATCH 505/886] [Sema] Mark alias/ifunc targets used and consider
 mangled names

https://reviews.llvm.org/D54188 marked "alias" targets as used in C to
fix -Wunused false positives. This patch extends the approach to handle
mangled names to support global scope names in C++ and the
`overloadable` attribute in C.

In addition, we mark ifunc targets as used to fix #63957.

While our approach has false negatives for namespace scope names, the
majority of alias/ifunc C++ uses (global scope with no overloads) are
handled.

Note: The following function with internal linkage but C language
linkage type is mangled in Clang but not in GCC. This inconsistency
makes alias/ifunc difficult to use in C++ with portability (#88593).
```
extern "C" {
static void f0() {}
// GCC: void g0() __attribute__((alias("_ZL2f0v")));
// Clang: void g0() __attribute__((alias("f0")));
}
```

Pull Request: https://github.com/llvm/llvm-project/pull/87130
---
 clang/lib/Sema/CMakeLists.txt                 |  1 +
 clang/lib/Sema/SemaDeclAttr.cpp               | 44 ++++++++++++++-----
 clang/test/AST/ast-dump-attr-json.cpp         |  1 +
 clang/test/Sema/alias-unused-win.cpp          |  2 +-
 clang/test/Sema/alias-unused.cpp              | 16 ++++---
 .../llvm-project-overlay/clang/BUILD.bazel    |  1 +
 6 files changed, 46 insertions(+), 19 deletions(-)

diff --git a/clang/lib/Sema/CMakeLists.txt b/clang/lib/Sema/CMakeLists.txt
index ab3b813a9ccd9..a96439df66422 100644
--- a/clang/lib/Sema/CMakeLists.txt
+++ b/clang/lib/Sema/CMakeLists.txt
@@ -1,5 +1,6 @@
 set(LLVM_LINK_COMPONENTS
   Core
+  Demangle
   FrontendHLSL
   FrontendOpenMP
   MC
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index b7b1fbc625a15..d26f130b5774c 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -45,6 +45,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/Demangle/Demangle.h"
 #include "llvm/IR/Assumptions.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Support/Error.h"
@@ -1983,6 +1984,36 @@ static void handleWeakRefAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   D->addAttr(::new (S.Context) WeakRefAttr(S.Context, AL));
 }
 
+// Mark alias/ifunc target as used. Due to name mangling, we look up the
+// demangled name ignoring parameters (not supported by microsoftDemangle
+// https://github.com/llvm/llvm-project/issues/88825). This should handle the
+// majority of use cases while leaving namespace scope names unmarked.
+static void markUsedForAliasOrIfunc(Sema &S, Decl *D, const ParsedAttr &AL,
+                                    StringRef Str) {
+  std::unique_ptr<char, llvm::FreeDeleter> Demangled;
+  if (S.getASTContext().getCXXABIKind() != TargetCXXABI::Microsoft)
+    Demangled.reset(llvm::itaniumDemangle(Str, /*ParseParams=*/false));
+  std::unique_ptr<MangleContext> MC(S.Context.createMangleContext());
+  SmallString<256> Name;
+
+  const DeclarationNameInfo Target(
+      &S.Context.Idents.get(Demangled ? Demangled.get() : Str), AL.getLoc());
+  LookupResult LR(S, Target, Sema::LookupOrdinaryName);
+  if (S.LookupName(LR, S.TUScope)) {
+    for (NamedDecl *ND : LR) {
+      if (MC->shouldMangleDeclName(ND)) {
+        llvm::raw_svector_ostream Out(Name);
+        Name.clear();
+        MC->mangleName(GlobalDecl(ND), Out);
+      } else {
+        Name = ND->getIdentifier()->getName();
+      }
+      if (Name == Str)
+        ND->markUsed(S.Context);
+    }
+  }
+}
+
 static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   StringRef Str;
   if (!S.checkStringLiteralArgumentAttr(AL, 0, Str))
@@ -1995,6 +2026,7 @@ static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     return;
   }
 
+  markUsedForAliasOrIfunc(S, D, AL, Str);
   D->addAttr(::new (S.Context) IFuncAttr(S.Context, AL, Str));
 }
 
@@ -2029,17 +2061,7 @@ static void handleAliasAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     }
   }
 
-  // Mark target used to prevent unneeded-internal-declaration warnings.
-  if (!S.LangOpts.CPlusPlus) {
-    // FIXME: demangle Str for C++, as the attribute refers to the mangled
-    // linkage name, not the pre-mangled identifier.
-    const DeclarationNameInfo target(&S.Context.Idents.get(Str), AL.getLoc());
-    LookupResult LR(S, target, Sema::LookupOrdinaryName);
-    if (S.LookupQualifiedName(LR, S.getCurLexicalContext()))
-      for (NamedDecl *ND : LR)
-        ND->markUsed(S.Context);
-  }
-
+  markUsedForAliasOrIfunc(S, D, AL, Str);
   D->addAttr(::new (S.Context) AliasAttr(S.Context, AL, Str));
 }
 
diff --git a/clang/test/AST/ast-dump-attr-json.cpp b/clang/test/AST/ast-dump-attr-json.cpp
index 051c2956abfdf..883e584bfedf0 100644
--- a/clang/test/AST/ast-dump-attr-json.cpp
+++ b/clang/test/AST/ast-dump-attr-json.cpp
@@ -46,6 +46,7 @@ __thread __attribute__ ((tls_model ("local-exec"))) int tls_model_var;
 // CHECK-NEXT:    "tokLen": 11
 // CHECK-NEXT:   }
 // CHECK-NEXT:  },
+// CHECK-NEXT:  "isUsed": true,
 // CHECK-NEXT:  "name": "global_decl",
 // CHECK-NEXT:  "mangledName": "global_decl",
 // CHECK-NEXT:  "type": {
diff --git a/clang/test/Sema/alias-unused-win.cpp b/clang/test/Sema/alias-unused-win.cpp
index 47c96d4117517..97d57a3bbd1e3 100644
--- a/clang/test/Sema/alias-unused-win.cpp
+++ b/clang/test/Sema/alias-unused-win.cpp
@@ -7,7 +7,7 @@ extern "C" {
 static int f(void) { return 42; } // cxx-warning{{unused function 'f'}}
 int g(void) __attribute__((alias("f")));
 
-static int foo [] = { 42, 0xDEAD }; // cxx-warning{{variable 'foo' is not needed and will not be emitted}}
+static int foo [] = { 42, 0xDEAD };
 extern typeof(foo) bar __attribute__((unused, alias("foo")));
 
 static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
diff --git a/clang/test/Sema/alias-unused.cpp b/clang/test/Sema/alias-unused.cpp
index dc8e46f072d74..c0b541c880e52 100644
--- a/clang/test/Sema/alias-unused.cpp
+++ b/clang/test/Sema/alias-unused.cpp
@@ -14,24 +14,26 @@ extern typeof(foo) bar __attribute__((unused, alias("foo")));
 /// We report a warning in C++ mode because the internal linkage `resolver` gets
 /// mangled as it does not have a language linkage. GCC does not mangle
 /// `resolver` or report a warning.
-static int (*resolver(void))(void) { return f; } // expected-warning{{unused function 'resolver'}}
+static int (*resolver(void))(void) { return f; } // cxx-warning{{unused function 'resolver'}}
 int ifunc(void) __attribute__((ifunc("resolver")));
 
-static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
+static int __attribute__((overloadable)) f0(int x) { return x; }
 static float __attribute__((overloadable)) f0(float x) { return x; } // expected-warning{{unused function 'f0'}}
 int g0(void) __attribute__((alias("_ZL2f0i")));
 
 #ifdef __cplusplus
-static int f1() { return 42; } // expected-warning{{unused function 'f1'}}
+static int f1() { return 42; }
 int g1(void) __attribute__((alias("_ZL2f1v")));
 }
 
-static int f2(int) { return 42; } // expected-warning{{unused function 'f2'}}
-static int f2() { return 42; } // expected-warning{{unused function 'f2'}}
+/// We demangle alias/ifunc target and mark all found functions as used.
+
+static int f2(int) { return 42; } // cxx-warning{{unused function 'f2'}}
+static int f2() { return 42; }
 int g2() __attribute__((alias("_ZL2f2v")));
 
-static int (*resolver1())() { return f; } // expected-warning{{unused function 'resolver1'}}
-static int (*resolver1(int))() { return f; } // expected-warning{{unused function 'resolver1'}}
+static int (*resolver1())() { return f; } // cxx-warning{{unused function 'resolver1'}}
+static int (*resolver1(int))() { return f; }
 int ifunc1() __attribute__((ifunc("_ZL9resolver1i")));
 
 /// TODO: We should report "unused function" for f3(int).
diff --git a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
index c2f77e3abca0e..725ac6bb38120 100644
--- a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
@@ -1136,6 +1136,7 @@ cc_library(
         "//llvm:AllTargetsAsmParsers",
         "//llvm:AllTargetsCodeGens",
         "//llvm:Core",
+        "//llvm:Demangle",
         "//llvm:FrontendHLSL",
         "//llvm:FrontendOpenMP",
         "//llvm:MC",

From 8aa7e378dee27ec81959ef6750a7dd07cefdc77d Mon Sep 17 00:00:00 2001
From: Allen <zhongyunde@huawei.com>
Date: Tue, 16 Apr 2024 09:00:03 +0800
Subject: [PATCH 506/886] [InterleavedAccessPass] Get round the unsupported
 large scalarize vectors (#88643)

When build with option -msve-vector-bits=512, the return vaule of
Subtarget->getMinSVEVectorSizeInBits() is 512;
While the MinElts is still 4 for <vscale x 4 x double> in
getNumInterleavedAccesses, so it creates invalid
llvm.aarch64.sve.ld2.sret.nxv4f64, which need be splited.
Unlikely, the related custom spilting is not supported now.

Fix https://github.com/llvm/llvm-project/issues/88247
---
 .../Target/AArch64/AArch64ISelLowering.cpp    |  2 +-
 .../AArch64/sve-interleaved-accesses.ll       | 24 +++++++++++++++++++
 2 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index eee67a0f823c1..7947d73f9a4dd 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -15888,7 +15888,7 @@ unsigned AArch64TargetLowering::getNumInterleavedAccesses(
   unsigned VecSize = 128;
   unsigned ElSize = DL.getTypeSizeInBits(VecTy->getElementType());
   unsigned MinElts = VecTy->getElementCount().getKnownMinValue();
-  if (UseScalable)
+  if (UseScalable && isa<FixedVectorType>(VecTy))
     VecSize = std::max(Subtarget->getMinSVEVectorSizeInBits(), 128u);
   return std::max<unsigned>(1, (MinElts * ElSize + 127) / VecSize);
 }
diff --git a/llvm/test/Transforms/InterleavedAccess/AArch64/sve-interleaved-accesses.ll b/llvm/test/Transforms/InterleavedAccess/AArch64/sve-interleaved-accesses.ll
index feb22aa1a3763..45e2c36836ffc 100644
--- a/llvm/test/Transforms/InterleavedAccess/AArch64/sve-interleaved-accesses.ll
+++ b/llvm/test/Transforms/InterleavedAccess/AArch64/sve-interleaved-accesses.ll
@@ -491,6 +491,30 @@ define void @store_bfloat_factor2(ptr %ptr, <16 x bfloat> %v0, <16 x bfloat> %v1
   ret void
 }
 
+; Ensure vscale_range property does not affect scalable vector types.
+define { <vscale x 4 x double>, <vscale x 4 x double> } @deinterleave_nxptr_factor2(ptr %ptr) #2 {
+; CHECK-LABEL: define { <vscale x 4 x double>, <vscale x 4 x double> } @deinterleave_nxptr_factor2(
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 2 x double>, ptr [[PTR]], i64 0
+; CHECK-NEXT:    [[LDN1:%.*]] = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), ptr [[TMP1]])
+; CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[LDN1]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+; CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[LDN1]], 1
+; CHECK-NEXT:    [[TMP5:%.*]] = call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr <vscale x 2 x double>, ptr [[PTR]], i64 2
+; CHECK-NEXT:    [[LDN2:%.*]] = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), ptr [[TMP6]])
+; CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[LDN2]], 0
+; CHECK-NEXT:    [[TMP8:%.*]] = call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP3]], <vscale x 2 x double> [[TMP7]], i64 2)
+; CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[LDN2]], 1
+; CHECK-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP5]], <vscale x 2 x double> [[TMP9]], i64 2)
+; CHECK-NEXT:    [[TMP11:%.*]] = insertvalue { <vscale x 4 x double>, <vscale x 4 x double> } poison, <vscale x 4 x double> [[TMP8]], 0
+; CHECK-NEXT:    [[TMP12:%.*]] = insertvalue { <vscale x 4 x double>, <vscale x 4 x double> } [[TMP11]], <vscale x 4 x double> [[TMP10]], 1
+; CHECK-NEXT:    ret { <vscale x 4 x double>, <vscale x 4 x double> } [[TMP12]]
+;
+  %wide.vec = load <vscale x 8 x double>, ptr %ptr, align 8
+  %ldN = tail call { <vscale x 4 x double>, <vscale x 4 x double> } @llvm.experimental.vector.deinterleave2.nxv8f64(<vscale x 8 x double> %wide.vec)
+  ret { <vscale x 4 x double>, <vscale x 4 x double> } %ldN
+}
+
 attributes #0 = { vscale_range(2,2) "target-features"="+sve" }
 attributes #1 = { vscale_range(2,4) "target-features"="+sve" }
 attributes #2 = { vscale_range(4,4) "target-features"="+sve" }

From d23a85066b5d89a47fde7b6777e0fc5a711b99d4 Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze@braunis.de>
Date: Mon, 15 Apr 2024 18:19:06 -0700
Subject: [PATCH 507/886] InstCombine: Increase threadlocal.address alignment
 if pointee is more aligned (#88435)

Increase alignment of `llvm.threadlocal.address` if the pointed to
global has higher alignment.
---
 .../InstCombine/InstCombineCalls.cpp          |  9 ++++
 .../InstCombine/threadlocal_address.ll        | 41 +++++++++++++++++++
 2 files changed, 50 insertions(+)
 create mode 100644 llvm/test/Transforms/InstCombine/threadlocal_address.ll

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 20f51c8af617d..bae8579fc3650 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -3408,6 +3408,15 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
       return I;
     break;
   }
+  case Intrinsic::threadlocal_address: {
+    Align MinAlign = getKnownAlignment(II->getArgOperand(0), DL, II, &AC, &DT);
+    MaybeAlign Align = II->getRetAlign();
+    if (MinAlign > Align.valueOrOne()) {
+      II->addRetAttr(Attribute::getWithAlignment(II->getContext(), MinAlign));
+      return II;
+    }
+    break;
+  }
   default: {
     // Handle target specific intrinsics
     std::optional<Instruction *> V = targetInstCombineIntrinsic(*II);
diff --git a/llvm/test/Transforms/InstCombine/threadlocal_address.ll b/llvm/test/Transforms/InstCombine/threadlocal_address.ll
new file mode 100644
index 0000000000000..0c220d996839e
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/threadlocal_address.ll
@@ -0,0 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -o - -S %s -passes=instcombine | FileCheck %s
+
+@tlsvar_a4 = thread_local global i32 4, align 4
+
+define void @func_increase_alignment() {
+; CHECK-LABEL: define void @func_increase_alignment() {
+; CHECK-NEXT:    [[P:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr @tlsvar_a4)
+; CHECK-NEXT:    store i32 42, ptr [[P]], align 2
+; CHECK-NEXT:    ret void
+;
+  %p = call align 2 ptr @llvm.threadlocal.address(ptr @tlsvar_a4)
+  store i32 42, ptr %p, align 2
+  ret void
+}
+
+@tlsvar_a32 = thread_local global i32 5, align 32
+
+define i1 @func_add_alignment() {
+; CHECK-LABEL: define i1 @func_add_alignment() {
+; CHECK-NEXT:    ret i1 true
+;
+  %p = call ptr @llvm.threadlocal.address(ptr @tlsvar_a32)
+  %p_int = ptrtoint ptr %p to i32
+  %lowbits = and i32 %p_int, 31
+  %zero = icmp eq i32 %lowbits, 0
+  ret i1 %zero
+}
+
+@tlsvar_a1 = thread_local global i8 6, align 1
+
+define i1 @func_dont_reduce_alignment() {
+; CHECK-LABEL: define i1 @func_dont_reduce_alignment() {
+; CHECK-NEXT:    ret i1 true
+;
+  %p = call align 4 ptr @llvm.threadlocal.address(ptr @tlsvar_a1)
+  %p_int = ptrtoint ptr %p to i32
+  %lowbits = and i32 %p_int, 3
+  %zero = icmp eq i32 %lowbits, 0
+  ret i1 %zero
+}

From 2e26ee9dce32d12ffa9bfb7f7d3e97778f0b7a75 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Mon, 15 Apr 2024 18:22:15 -0700
Subject: [PATCH 508/886] [DWARF] Clarify a variable name. NFC (#88814)

The parameter of `findDebugNamesOffsets` has been renamed to
`EndOfHeaderOffset` in #88064 to make it clear it is a section offset
instead of an offset relative to the current name index. Rename the call
site variable as well.
---
 llvm/lib/DebugInfo/DWARF/DWARFAcceleratorTable.cpp | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/llvm/lib/DebugInfo/DWARF/DWARFAcceleratorTable.cpp b/llvm/lib/DebugInfo/DWARF/DWARFAcceleratorTable.cpp
index 22c9e8cd143c2..ac19ac7932971 100644
--- a/llvm/lib/DebugInfo/DWARF/DWARFAcceleratorTable.cpp
+++ b/llvm/lib/DebugInfo/DWARF/DWARFAcceleratorTable.cpp
@@ -572,12 +572,12 @@ dwarf::findDebugNamesOffsets(uint64_t EndOfHeaderOffset,
 
 Error DWARFDebugNames::NameIndex::extract() {
   const DWARFDataExtractor &AS = Section.AccelSection;
-  uint64_t hdrSize = Base;
-  if (Error E = Hdr.extract(AS, &hdrSize))
+  uint64_t EndOfHeaderOffset = Base;
+  if (Error E = Hdr.extract(AS, &EndOfHeaderOffset))
     return E;
 
   const unsigned SectionOffsetSize = dwarf::getDwarfOffsetByteSize(Hdr.Format);
-  Offsets = dwarf::findDebugNamesOffsets(hdrSize, Hdr);
+  Offsets = dwarf::findDebugNamesOffsets(EndOfHeaderOffset, Hdr);
 
   uint64_t Offset =
       Offsets.EntryOffsetsBase + (Hdr.NameCount * SectionOffsetSize);

From f3a8112d9839a166f7eb240c6c72d7ecd47d3560 Mon Sep 17 00:00:00 2001
From: Shih-Po Hung <shihpo.hung@sifive.com>
Date: Tue, 16 Apr 2024 09:37:32 +0800
Subject: [PATCH 509/886] [RISCV][TTI] Scale the cost of ICmp with LMUL
 (#88235)

Use the Val type to estimate the instruction cost for ICmp.
---
 .../Target/RISCV/RISCVTargetTransformInfo.cpp |  17 +-
 llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll | 320 +++++++++---------
 .../LoopVectorize/RISCV/illegal-type.ll       |  13 +-
 3 files changed, 174 insertions(+), 176 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index bc9756c5e6dda..56f5bd8794ae7 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -1335,8 +1335,8 @@ InstructionCost RISCVTTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
     return BaseT::getCmpSelInstrCost(Opcode, ValTy, CondTy, VecPred, CostKind,
                                      I);
 
+  std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(ValTy);
   if (Opcode == Instruction::Select && ValTy->isVectorTy()) {
-    std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(ValTy);
     if (CondTy->isVectorTy()) {
       if (ValTy->getScalarSizeInBits() == 1) {
         // vmandn.mm v8, v8, v9
@@ -1375,14 +1375,15 @@ InstructionCost RISCVTTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
                           LT.second, CostKind);
   }
 
-  if ((Opcode == Instruction::ICmp || Opcode == Instruction::FCmp) &&
-      ValTy->isVectorTy()) {
-    std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(ValTy);
-
-    // Support natively.
-    if (CmpInst::isIntPredicate(VecPred))
-      return LT.first * 1;
+  if ((Opcode == Instruction::ICmp) && ValTy->isVectorTy() &&
+      CmpInst::isIntPredicate(VecPred)) {
+    // Use VMSLT_VV to represent VMSEQ, VMSNE, VMSLTU, VMSLEU, VMSLT, VMSLE
+    // provided they incur the same cost across all implementations
+    return LT.first *
+           getRISCVInstructionCost(RISCV::VMSLT_VV, LT.second, CostKind);
+  }
 
+  if ((Opcode == Instruction::FCmp) && ValTy->isVectorTy()) {
     // If we do not support the input floating point vector type, use the base
     // one which will calculate as:
     // ScalarizeCost + Num * Cost for fixed vector,
diff --git a/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll b/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
index 27d24faf0a8da..caa6d6f483a24 100644
--- a/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
@@ -9,38 +9,38 @@ define void @icmp_eq() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp eq <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp eq <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp eq <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp eq <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp eq <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp eq <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp eq <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp eq <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp eq <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp eq <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp eq <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp eq <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp eq <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp eq <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp eq <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp eq <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp eq <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp eq <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp eq <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp eq <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp eq <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp eq <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp eq <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp eq <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp eq <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp eq <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp eq <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp eq <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp eq <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp eq <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp eq <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp eq <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp eq <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp eq <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp eq <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp eq <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp eq <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp eq <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp eq <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp eq <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp eq <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp eq <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp eq <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp eq <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp eq <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp eq <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp eq <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp eq <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp eq <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp eq <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp eq <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp eq <2 x i8> undef, undef
@@ -96,38 +96,38 @@ define void @icmp_ne() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp ne <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp ne <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp ne <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp ne <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp ne <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp ne <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp ne <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp ne <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp ne <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp ne <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp ne <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp ne <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp ne <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp ne <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp ne <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp ne <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp ne <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp ne <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp ne <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp ne <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp ne <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp ne <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp ne <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp ne <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp ne <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp ne <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp ne <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp ne <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp ne <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp ne <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp ne <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp ne <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp ne <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp ne <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp ne <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp ne <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp ne <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp ne <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp ne <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp ne <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp ne <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp ne <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp ne <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp ne <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp ne <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp ne <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp ne <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp ne <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp ne <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp ne <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp ne <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp ne <2 x i8> undef, undef
@@ -183,38 +183,38 @@ define void @icmp_ugt() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp ugt <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp ugt <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp ugt <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp ugt <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp ugt <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp ugt <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp ugt <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp ugt <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp ugt <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp ugt <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp ugt <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp ugt <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp ugt <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp ugt <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp ugt <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp ugt <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp ugt <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp ugt <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp ugt <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp ugt <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp ugt <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp ugt <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp ugt <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp ugt <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp ugt <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp ugt <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp ugt <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp ugt <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp ugt <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp ugt <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp ugt <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp ugt <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp ugt <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp ugt <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp ugt <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp ugt <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp ugt <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp ugt <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp ugt <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp ugt <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp ugt <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp ugt <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp ugt <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp ugt <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp ugt <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp ugt <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp ugt <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp ugt <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp ugt <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp ugt <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp ugt <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp ugt <2 x i8> undef, undef
@@ -270,38 +270,38 @@ define void @icmp_uge() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp uge <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp uge <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp uge <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp uge <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp uge <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp uge <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp uge <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp uge <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp uge <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp uge <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp uge <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp uge <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp uge <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp uge <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp uge <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp uge <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp uge <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp uge <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp uge <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp uge <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp uge <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp uge <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp uge <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp uge <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp uge <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp uge <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp uge <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp uge <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp uge <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp uge <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp uge <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp uge <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp uge <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp uge <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp uge <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp uge <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp uge <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp uge <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp uge <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp uge <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp uge <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp uge <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp uge <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp uge <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp uge <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp uge <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp uge <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp uge <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp uge <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp uge <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp uge <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp uge <2 x i8> undef, undef
@@ -357,38 +357,38 @@ define void @icmp_ult() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp ult <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp ult <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp ult <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp ult <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp ult <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp ult <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp ult <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp ult <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp ult <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp ult <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp ult <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp ult <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp ult <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp ult <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp ult <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp ult <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp ult <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp ult <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp ult <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp ult <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp ult <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp ult <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp ult <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp ult <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp ult <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp ult <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp ult <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp ult <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp ult <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp ult <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp ult <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp ult <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp ult <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp ult <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp ult <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp ult <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp ult <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp ult <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp ult <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp ult <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp ult <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp ult <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp ult <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp ult <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp ult <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp ult <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp ult <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp ult <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp ult <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp ult <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp ult <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp ult <2 x i8> undef, undef
@@ -444,38 +444,38 @@ define void @icmp_ule() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp ule <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp ule <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp ule <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp ule <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp ule <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp ule <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp ule <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp ule <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp ule <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp ule <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp ule <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp ule <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp ule <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp ule <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp ule <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp ule <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp ule <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp ule <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp ule <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp ule <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp ule <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp ule <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp ule <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp ule <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp ule <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp ule <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp ule <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp ule <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp ule <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp ule <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp ule <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp ule <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp ule <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp ule <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp ule <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp ule <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp ule <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp ule <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp ule <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp ule <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp ule <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp ule <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp ule <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp ule <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp ule <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp ule <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp ule <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp ule <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp ule <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp ule <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp ule <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp ule <2 x i8> undef, undef
@@ -531,38 +531,38 @@ define void @icmp_sgt() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp sgt <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp sgt <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp sgt <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp sgt <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp sgt <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp sgt <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp sgt <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp sgt <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp sgt <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp sgt <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp sgt <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp sgt <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp sgt <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp sgt <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp sgt <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp sgt <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp sgt <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp sgt <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp sgt <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp sgt <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp sgt <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp sgt <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp sgt <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp sgt <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp sgt <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp sgt <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp sgt <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp sgt <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp sgt <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp sgt <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp sgt <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp sgt <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp sgt <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp sgt <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp sgt <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp sgt <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp sgt <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp sgt <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp sgt <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp sgt <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp sgt <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp sgt <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp sgt <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp sgt <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp sgt <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp sgt <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp sgt <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp sgt <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp sgt <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp sgt <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp sgt <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp sgt <2 x i8> undef, undef
@@ -618,38 +618,38 @@ define void @icmp_sge() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp sge <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp sge <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp sge <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp sge <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp sge <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp sge <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp sge <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp sge <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp sge <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp sge <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp sge <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp sge <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp sge <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp sge <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp sge <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp sge <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp sge <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp sge <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp sge <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp sge <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp sge <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp sge <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp sge <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp sge <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp sge <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp sge <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp sge <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp sge <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp sge <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp sge <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp sge <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp sge <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp sge <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp sge <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp sge <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp sge <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp sge <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp sge <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp sge <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp sge <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp sge <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp sge <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp sge <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp sge <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp sge <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp sge <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp sge <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp sge <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp sge <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp sge <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp sge <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp sge <2 x i8> undef, undef
@@ -705,38 +705,38 @@ define void @icmp_slt() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp slt <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp slt <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp slt <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp slt <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp slt <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp slt <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp slt <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp slt <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp slt <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp slt <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp slt <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp slt <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp slt <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp slt <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp slt <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp slt <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp slt <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp slt <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp slt <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp slt <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp slt <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp slt <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp slt <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp slt <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp slt <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp slt <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp slt <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp slt <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp slt <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp slt <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp slt <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp slt <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp slt <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp slt <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp slt <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp slt <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp slt <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp slt <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp slt <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp slt <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp slt <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp slt <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp slt <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp slt <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp slt <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp slt <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp slt <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp slt <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp slt <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp slt <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp slt <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp slt <2 x i8> undef, undef
@@ -792,38 +792,38 @@ define void @icmp_sle() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = icmp sle <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = icmp sle <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = icmp sle <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8 = icmp sle <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = icmp sle <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i8 = icmp sle <vscale x 1 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i8 = icmp sle <vscale x 2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i8 = icmp sle <vscale x 4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i8 = icmp sle <vscale x 8 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i8 = icmp sle <vscale x 16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv32i8 = icmp sle <vscale x 32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv16i8 = icmp sle <vscale x 16 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv32i8 = icmp sle <vscale x 32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = icmp sle <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = icmp sle <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = icmp sle <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16 = icmp sle <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = icmp sle <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i16 = icmp sle <vscale x 1 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i16 = icmp sle <vscale x 2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i16 = icmp sle <vscale x 4 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i16 = icmp sle <vscale x 8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i16 = icmp sle <vscale x 16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8i16 = icmp sle <vscale x 8 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16i16 = icmp sle <vscale x 16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = icmp sle <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = icmp sle <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32 = icmp sle <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i32 = icmp sle <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = icmp sle <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = icmp sle <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i32 = icmp sle <vscale x 1 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i32 = icmp sle <vscale x 2 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i32 = icmp sle <vscale x 4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i32 = icmp sle <vscale x 8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16i32 = icmp sle <vscale x 16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4i32 = icmp sle <vscale x 4 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8i32 = icmp sle <vscale x 8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16i32 = icmp sle <vscale x 16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = icmp sle <2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i64 = icmp sle <4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i64 = icmp sle <8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = icmp sle <4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64 = icmp sle <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1i64 = icmp sle <vscale x 1 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2i64 = icmp sle <vscale x 2 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4i64 = icmp sle <vscale x 4 x i64> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8i64 = icmp sle <vscale x 8 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2i64 = icmp sle <vscale x 2 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4i64 = icmp sle <vscale x 4 x i64> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8i64 = icmp sle <vscale x 8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp sle <2 x i8> undef, undef
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/illegal-type.ll b/llvm/test/Transforms/LoopVectorize/RISCV/illegal-type.ll
index c49f7d4a5d5f1..eeef8f199353b 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/illegal-type.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/illegal-type.ll
@@ -102,10 +102,10 @@ define void @uniform_store_i1(ptr noalias %dst, ptr noalias %start, i64 %N) {
 ; CHECK-LABEL: @uniform_store_i1(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], 1
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 64
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 32
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 64
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 32
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[N_VEC]], 8
 ; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[START:%.*]], i64 [[TMP1]]
@@ -116,15 +116,12 @@ define void @uniform_store_i1(ptr noalias %dst, ptr noalias %start, i64 %N) {
 ; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[START]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <32 x i64> <i64 0, i64 8, i64 16, i64 24, i64 32, i64 40, i64 48, i64 56, i64 64, i64 72, i64 80, i64 88, i64 96, i64 104, i64 112, i64 120, i64 128, i64 136, i64 144, i64 152, i64 160, i64 168, i64 176, i64 184, i64 192, i64 200, i64 208, i64 216, i64 224, i64 232, i64 240, i64 248>
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <32 x i64> <i64 256, i64 264, i64 272, i64 280, i64 288, i64 296, i64 304, i64 312, i64 320, i64 328, i64 336, i64 344, i64 352, i64 360, i64 368, i64 376, i64 384, i64 392, i64 400, i64 408, i64 416, i64 424, i64 432, i64 440, i64 448, i64 456, i64 464, i64 472, i64 480, i64 488, i64 496, i64 504>
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, <32 x ptr> [[TMP2]], i64 1
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, <32 x ptr> [[TMP3]], i64 1
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq <32 x ptr> [[TMP4]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, <32 x ptr> [[TMP2]], i64 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq <32 x ptr> [[TMP5]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <32 x i1> [[TMP7]], i32 31
 ; CHECK-NEXT:    store i1 [[TMP8]], ptr [[DST:%.*]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 64
-; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 512
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
+; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 256
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK:       middle.block:

From dfe12b3fd193318403622a8ae51e0362c27502d1 Mon Sep 17 00:00:00 2001
From: Benji Smith <6193112+Benjins@users.noreply.github.com>
Date: Mon, 15 Apr 2024 21:37:55 -0400
Subject: [PATCH 510/886] [C API] Support uinc_wrap/udec_wrap in atomicrmw when
 accessing the bin op (#87163)

These previously were added in the C++ API in
778cf5431cafc243f81dd5c8cbd27701ff7f9120, but without updating the enum
in the C API or mapping functions.

Corresponding tests for all current atomicrmw bin ops have been added as
well.
---
 llvm/docs/ReleaseNotes.rst           |  3 ++
 llvm/include/llvm-c/Core.h           | 60 +++++++++++++++-------------
 llvm/lib/IR/Core.cpp                 |  8 ++++
 llvm/test/Bindings/llvm-c/atomics.ll | 25 ++++++++++++
 4 files changed, 68 insertions(+), 28 deletions(-)

diff --git a/llvm/docs/ReleaseNotes.rst b/llvm/docs/ReleaseNotes.rst
index 11add9274a7a0..76ef6ceb94078 100644
--- a/llvm/docs/ReleaseNotes.rst
+++ b/llvm/docs/ReleaseNotes.rst
@@ -149,6 +149,9 @@ Changes to the C API
 
 * Deprecated ``LLVMConstNUWNeg`` and ``LLVMBuildNUWNeg``.
 
+* Added ``LLVMAtomicRMWBinOpUIncWrap`` and ``LLVMAtomicRMWBinOpUDecWrap`` to
+  ``LLVMAtomicRMWBinOp`` enum for AtomicRMW instructions.
+
 Changes to the CodeGen infrastructure
 -------------------------------------
 
diff --git a/llvm/include/llvm-c/Core.h b/llvm/include/llvm-c/Core.h
index 6be5957ce6103..0b03f3b36fcdd 100644
--- a/llvm/include/llvm-c/Core.h
+++ b/llvm/include/llvm-c/Core.h
@@ -361,35 +361,39 @@ typedef enum {
 } LLVMAtomicOrdering;
 
 typedef enum {
-    LLVMAtomicRMWBinOpXchg, /**< Set the new value and return the one old */
-    LLVMAtomicRMWBinOpAdd, /**< Add a value and return the old one */
-    LLVMAtomicRMWBinOpSub, /**< Subtract a value and return the old one */
-    LLVMAtomicRMWBinOpAnd, /**< And a value and return the old one */
-    LLVMAtomicRMWBinOpNand, /**< Not-And a value and return the old one */
-    LLVMAtomicRMWBinOpOr, /**< OR a value and return the old one */
-    LLVMAtomicRMWBinOpXor, /**< Xor a value and return the old one */
-    LLVMAtomicRMWBinOpMax, /**< Sets the value if it's greater than the
-                             original using a signed comparison and return
-                             the old one */
-    LLVMAtomicRMWBinOpMin, /**< Sets the value if it's Smaller than the
-                             original using a signed comparison and return
-                             the old one */
-    LLVMAtomicRMWBinOpUMax, /**< Sets the value if it's greater than the
-                             original using an unsigned comparison and return
-                             the old one */
-    LLVMAtomicRMWBinOpUMin, /**< Sets the value if it's greater than the
-                              original using an unsigned comparison and return
-                              the old one */
-    LLVMAtomicRMWBinOpFAdd, /**< Add a floating point value and return the
-                              old one */
-    LLVMAtomicRMWBinOpFSub, /**< Subtract a floating point value and return the
+  LLVMAtomicRMWBinOpXchg, /**< Set the new value and return the one old */
+  LLVMAtomicRMWBinOpAdd,  /**< Add a value and return the old one */
+  LLVMAtomicRMWBinOpSub,  /**< Subtract a value and return the old one */
+  LLVMAtomicRMWBinOpAnd,  /**< And a value and return the old one */
+  LLVMAtomicRMWBinOpNand, /**< Not-And a value and return the old one */
+  LLVMAtomicRMWBinOpOr,   /**< OR a value and return the old one */
+  LLVMAtomicRMWBinOpXor,  /**< Xor a value and return the old one */
+  LLVMAtomicRMWBinOpMax,  /**< Sets the value if it's greater than the
+                            original using a signed comparison and return
+                            the old one */
+  LLVMAtomicRMWBinOpMin,  /**< Sets the value if it's Smaller than the
+                            original using a signed comparison and return
+                            the old one */
+  LLVMAtomicRMWBinOpUMax, /**< Sets the value if it's greater than the
+                           original using an unsigned comparison and return
+                           the old one */
+  LLVMAtomicRMWBinOpUMin, /**< Sets the value if it's greater than the
+                            original using an unsigned comparison and return
+                            the old one */
+  LLVMAtomicRMWBinOpFAdd, /**< Add a floating point value and return the
                             old one */
-    LLVMAtomicRMWBinOpFMax, /**< Sets the value if it's greater than the
-                             original using an floating point comparison and
-                             return the old one */
-    LLVMAtomicRMWBinOpFMin, /**< Sets the value if it's smaller than the
-                             original using an floating point comparison and
-                             return the old one */
+  LLVMAtomicRMWBinOpFSub, /**< Subtract a floating point value and return the
+                          old one */
+  LLVMAtomicRMWBinOpFMax, /**< Sets the value if it's greater than the
+                           original using an floating point comparison and
+                           return the old one */
+  LLVMAtomicRMWBinOpFMin, /**< Sets the value if it's smaller than the
+                           original using an floating point comparison and
+                           return the old one */
+  LLVMAtomicRMWBinOpUIncWrap, /**< Increments the value, wrapping back to zero
+                               when incremented above input value */
+  LLVMAtomicRMWBinOpUDecWrap, /**< Decrements the value, wrapping back to
+                               the input value when decremented below zero */
 } LLVMAtomicRMWBinOp;
 
 typedef enum {
diff --git a/llvm/lib/IR/Core.cpp b/llvm/lib/IR/Core.cpp
index 8ce9c5ca63bed..6aff94f39d9c0 100644
--- a/llvm/lib/IR/Core.cpp
+++ b/llvm/lib/IR/Core.cpp
@@ -3769,6 +3769,10 @@ static AtomicRMWInst::BinOp mapFromLLVMRMWBinOp(LLVMAtomicRMWBinOp BinOp) {
     case LLVMAtomicRMWBinOpFSub: return AtomicRMWInst::FSub;
     case LLVMAtomicRMWBinOpFMax: return AtomicRMWInst::FMax;
     case LLVMAtomicRMWBinOpFMin: return AtomicRMWInst::FMin;
+    case LLVMAtomicRMWBinOpUIncWrap:
+      return AtomicRMWInst::UIncWrap;
+    case LLVMAtomicRMWBinOpUDecWrap:
+      return AtomicRMWInst::UDecWrap;
   }
 
   llvm_unreachable("Invalid LLVMAtomicRMWBinOp value!");
@@ -3791,6 +3795,10 @@ static LLVMAtomicRMWBinOp mapToLLVMRMWBinOp(AtomicRMWInst::BinOp BinOp) {
     case AtomicRMWInst::FSub: return LLVMAtomicRMWBinOpFSub;
     case AtomicRMWInst::FMax: return LLVMAtomicRMWBinOpFMax;
     case AtomicRMWInst::FMin: return LLVMAtomicRMWBinOpFMin;
+    case AtomicRMWInst::UIncWrap:
+      return LLVMAtomicRMWBinOpUIncWrap;
+    case AtomicRMWInst::UDecWrap:
+      return LLVMAtomicRMWBinOpUDecWrap;
     default: break;
   }
 
diff --git a/llvm/test/Bindings/llvm-c/atomics.ll b/llvm/test/Bindings/llvm-c/atomics.ll
index e64a29944ef9d..162368c9d98d0 100644
--- a/llvm/test/Bindings/llvm-c/atomics.ll
+++ b/llvm/test/Bindings/llvm-c/atomics.ll
@@ -36,6 +36,31 @@ define void @atomic_load_store(ptr %word) {
   ret void
 }
 
+define void @atomic_rmw_ops(ptr %p, i32 %i, float %f) {
+  ; Test all atomicrmw operations
+  %a.xchg      = atomicrmw xchg      ptr %p, i32 %i acq_rel, align 8
+  %a.add       = atomicrmw add       ptr %p, i32 %i acq_rel, align 8
+  %a.sub       = atomicrmw sub       ptr %p, i32 %i acq_rel, align 8
+  %a.and       = atomicrmw and       ptr %p, i32 %i acq_rel, align 8
+  %a.nand      = atomicrmw nand      ptr %p, i32 %i acq_rel, align 8
+  %a.or        = atomicrmw or        ptr %p, i32 %i acq_rel, align 8
+  %a.xor       = atomicrmw xor       ptr %p, i32 %i acq_rel, align 8
+  %a.max       = atomicrmw max       ptr %p, i32 %i acq_rel, align 8
+  %a.min       = atomicrmw min       ptr %p, i32 %i acq_rel, align 8
+  %a.umax      = atomicrmw umax      ptr %p, i32 %i acq_rel, align 8
+  %a.umin      = atomicrmw umin      ptr %p, i32 %i acq_rel, align 8
+
+  %a.fadd      = atomicrmw fadd      ptr %p, float %f acq_rel, align 8
+  %a.fsub      = atomicrmw fsub      ptr %p, float %f acq_rel, align 8
+  %a.fmax      = atomicrmw fmax      ptr %p, float %f acq_rel, align 8
+  %a.fmin      = atomicrmw fmin      ptr %p, float %f acq_rel, align 8
+
+  %a.uinc_wrap = atomicrmw uinc_wrap ptr %p, i32 %i acq_rel, align 8
+  %a.udec_wrap = atomicrmw udec_wrap ptr %p, i32 %i acq_rel, align 8
+
+  ret void
+}
+
 define i32 @main() {
   %1 = alloca i32, align 4
   %2 = cmpxchg ptr %1, i32 2, i32 3 seq_cst acquire

From 5b9af38a03c8119cc2a42ae80d4a25e6f454c721 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 18:38:54 -0700
Subject: [PATCH 511/886] [RISCV] Provide a more efficient lowering for
 experimental.cttz.elts. (#88552)

For experimental.cttz.elts, we can use a vfirst instruction, but we need
to correct the result if input vector can be 0. cttz.elts returns the
vector length while vfirst returns -1.
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   |  36 ++++++
 llvm/lib/Target/RISCV/RISCVISelLowering.h     |   2 +
 .../RISCV/intrinsic-cttz-elts-vscale.ll       | 114 ++++++++++++++----
 .../test/CodeGen/RISCV/intrinsic-cttz-elts.ll |  22 +---
 4 files changed, 132 insertions(+), 42 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 259cc388276c6..f6ed6420c9e1f 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1484,6 +1484,11 @@ bool RISCVTargetLowering::shouldExpandGetVectorLength(EVT TripCountVT,
   return VF > MaxVF || !isPowerOf2_32(VF);
 }
 
+bool RISCVTargetLowering::shouldExpandCttzElements(EVT VT) const {
+  return !Subtarget.hasVInstructions() ||
+         VT.getVectorElementType() != MVT::i1 || !isTypeLegal(VT);
+}
+
 bool RISCVTargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
                                              const CallInst &I,
                                              MachineFunction &MF,
@@ -8718,6 +8723,29 @@ static SDValue lowerGetVectorLength(SDNode *N, SelectionDAG &DAG,
   return DAG.getNode(ISD::TRUNCATE, DL, N->getValueType(0), Res);
 }
 
+static SDValue lowerCttzElts(SDNode *N, SelectionDAG &DAG,
+                             const RISCVSubtarget &Subtarget) {
+  SDValue Op0 = N->getOperand(1);
+  MVT OpVT = Op0.getSimpleValueType();
+  MVT ContainerVT = OpVT;
+  if (OpVT.isFixedLengthVector()) {
+    ContainerVT = getContainerForFixedLengthVector(DAG, OpVT, Subtarget);
+    Op0 = convertToScalableVector(ContainerVT, Op0, DAG, Subtarget);
+  }
+  MVT XLenVT = Subtarget.getXLenVT();
+  SDLoc DL(N);
+  auto [Mask, VL] = getDefaultVLOps(OpVT, ContainerVT, DL, DAG, Subtarget);
+  SDValue Res = DAG.getNode(RISCVISD::VFIRST_VL, DL, XLenVT, Op0, Mask, VL);
+  if (isOneConstant(N->getOperand(2)))
+    return Res;
+
+  // Convert -1 to VL.
+  SDValue Setcc =
+      DAG.getSetCC(DL, XLenVT, Res, DAG.getConstant(0, DL, XLenVT), ISD::SETLT);
+  VL = DAG.getElementCount(DL, XLenVT, OpVT.getVectorElementCount());
+  return DAG.getSelect(DL, XLenVT, Setcc, VL, Res);
+}
+
 static inline void promoteVCIXScalar(const SDValue &Op,
                                      SmallVectorImpl<SDValue> &Operands,
                                      SelectionDAG &DAG) {
@@ -8913,6 +8941,8 @@ SDValue RISCVTargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
   }
   case Intrinsic::experimental_get_vector_length:
     return lowerGetVectorLength(Op.getNode(), DAG, Subtarget);
+  case Intrinsic::experimental_cttz_elts:
+    return lowerCttzElts(Op.getNode(), DAG, Subtarget);
   case Intrinsic::riscv_vmv_x_s: {
     SDValue Res = DAG.getNode(RISCVISD::VMV_X_S, DL, XLenVT, Op.getOperand(1));
     return DAG.getNode(ISD::TRUNCATE, DL, Op.getValueType(), Res);
@@ -12336,6 +12366,12 @@ void RISCVTargetLowering::ReplaceNodeResults(SDNode *N,
       Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Res));
       return;
     }
+    case Intrinsic::experimental_cttz_elts: {
+      SDValue Res = lowerCttzElts(N, DAG, Subtarget);
+      Results.push_back(
+          DAG.getNode(ISD::TRUNCATE, DL, N->getValueType(0), Res));
+      return;
+    }
     case Intrinsic::riscv_orc_b:
     case Intrinsic::riscv_brev8:
     case Intrinsic::riscv_sha256sig0:
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.h b/llvm/lib/Target/RISCV/RISCVISelLowering.h
index ace5b3fd2b95b..e2633733c31b1 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.h
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.h
@@ -986,6 +986,8 @@ class RISCVTargetLowering : public TargetLowering {
   bool shouldExpandGetVectorLength(EVT TripCountVT, unsigned VF,
                                    bool IsScalable) const override;
 
+  bool shouldExpandCttzElements(EVT VT) const override;
+
   /// RVV code generation for fixed length vectors does not lower all
   /// BUILD_VECTORs. This makes BUILD_VECTOR legalisation a source of stores to
   /// merge. However, merging them creates a BUILD_VECTOR that is just as
diff --git a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts-vscale.ll b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts-vscale.ll
index 65d0768c60885..ea8feef332984 100644
--- a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts-vscale.ll
+++ b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts-vscale.ll
@@ -128,43 +128,113 @@ define i64 @ctz_nxv8i1_no_range(<vscale x 8 x i16> %a) {
 define i32 @ctz_nxv16i1(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %a) {
 ; RV32-LABEL: ctz_nxv16i1:
 ; RV32:       # %bb.0:
-; RV32-NEXT:    vmv1r.v v0, v8
+; RV32-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
+; RV32-NEXT:    vfirst.m a0, v8
+; RV32-NEXT:    bgez a0, .LBB2_2
+; RV32-NEXT:  # %bb.1:
 ; RV32-NEXT:    csrr a0, vlenb
 ; RV32-NEXT:    slli a0, a0, 1
-; RV32-NEXT:    vsetvli a1, zero, e32, m8, ta, ma
-; RV32-NEXT:    vmv.v.x v8, a0
-; RV32-NEXT:    vid.v v16
-; RV32-NEXT:    li a1, -1
-; RV32-NEXT:    vmadd.vx v16, a1, v8
-; RV32-NEXT:    vmv.v.i v8, 0
-; RV32-NEXT:    vmerge.vvm v8, v8, v16, v0
-; RV32-NEXT:    vredmaxu.vs v8, v8, v8
-; RV32-NEXT:    vmv.x.s a1, v8
-; RV32-NEXT:    sub a0, a0, a1
+; RV32-NEXT:  .LBB2_2:
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: ctz_nxv16i1:
 ; RV64:       # %bb.0:
-; RV64-NEXT:    vmv1r.v v0, v8
+; RV64-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
+; RV64-NEXT:    vfirst.m a0, v8
+; RV64-NEXT:    bgez a0, .LBB2_2
+; RV64-NEXT:  # %bb.1:
 ; RV64-NEXT:    csrr a0, vlenb
 ; RV64-NEXT:    slli a0, a0, 1
-; RV64-NEXT:    vsetvli a1, zero, e32, m8, ta, ma
-; RV64-NEXT:    vmv.v.x v8, a0
-; RV64-NEXT:    vid.v v16
-; RV64-NEXT:    li a1, -1
-; RV64-NEXT:    vmadd.vx v16, a1, v8
-; RV64-NEXT:    vmv.v.i v8, 0
-; RV64-NEXT:    vmerge.vvm v8, v8, v16, v0
-; RV64-NEXT:    vredmaxu.vs v8, v8, v8
-; RV64-NEXT:    vmv.x.s a1, v8
-; RV64-NEXT:    subw a0, a0, a1
+; RV64-NEXT:  .LBB2_2:
 ; RV64-NEXT:    ret
   %res = call i32 @llvm.experimental.cttz.elts.i32.nxv16i1(<vscale x 16 x i1> %a, i1 0)
   ret i32 %res
 }
 
+define i32 @ctz_nxv16i1_poison(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %a) {
+; RV32-LABEL: ctz_nxv16i1_poison:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
+; RV32-NEXT:    vfirst.m a0, v8
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: ctz_nxv16i1_poison:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
+; RV64-NEXT:    vfirst.m a0, v8
+; RV64-NEXT:    ret
+  %res = call i32 @llvm.experimental.cttz.elts.i32.nxv16i1(<vscale x 16 x i1> %a, i1 1)
+  ret i32 %res
+}
+
+define i32 @ctz_v16i1(<16 x i1> %pg, <16 x i1> %a) {
+; RV32-LABEL: ctz_v16i1:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; RV32-NEXT:    vfirst.m a0, v8
+; RV32-NEXT:    bgez a0, .LBB4_2
+; RV32-NEXT:  # %bb.1:
+; RV32-NEXT:    li a0, 16
+; RV32-NEXT:  .LBB4_2:
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: ctz_v16i1:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; RV64-NEXT:    vfirst.m a0, v8
+; RV64-NEXT:    bgez a0, .LBB4_2
+; RV64-NEXT:  # %bb.1:
+; RV64-NEXT:    li a0, 16
+; RV64-NEXT:  .LBB4_2:
+; RV64-NEXT:    ret
+  %res = call i32 @llvm.experimental.cttz.elts.i32.v16i1(<16 x i1> %a, i1 0)
+  ret i32 %res
+}
+
+define i32 @ctz_v16i1_poison(<16 x i1> %pg, <16 x i1> %a) {
+; RV32-LABEL: ctz_v16i1_poison:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; RV32-NEXT:    vfirst.m a0, v8
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: ctz_v16i1_poison:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 16, e8, m1, ta, ma
+; RV64-NEXT:    vfirst.m a0, v8
+; RV64-NEXT:    ret
+  %res = call i32 @llvm.experimental.cttz.elts.i32.v16i1(<16 x i1> %a, i1 1)
+  ret i32 %res
+}
+
+define i16 @ctz_v8i1_i16_ret(<8 x i1> %a) {
+; RV32-LABEL: ctz_v8i1_i16_ret:
+; RV32:       # %bb.0:
+; RV32-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV32-NEXT:    vfirst.m a0, v0
+; RV32-NEXT:    bgez a0, .LBB6_2
+; RV32-NEXT:  # %bb.1:
+; RV32-NEXT:    li a0, 8
+; RV32-NEXT:  .LBB6_2:
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: ctz_v8i1_i16_ret:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
+; RV64-NEXT:    vfirst.m a0, v0
+; RV64-NEXT:    bgez a0, .LBB6_2
+; RV64-NEXT:  # %bb.1:
+; RV64-NEXT:    li a0, 8
+; RV64-NEXT:  .LBB6_2:
+; RV64-NEXT:    ret
+  %res = call i16 @llvm.experimental.cttz.elts.i16.v8i1(<8 x i1> %a, i1 0)
+  ret i16 %res
+}
+
 declare i64 @llvm.experimental.cttz.elts.i64.nxv8i16(<vscale x 8 x i16>, i1)
 declare i32 @llvm.experimental.cttz.elts.i32.nxv16i1(<vscale x 16 x i1>, i1)
 declare i32 @llvm.experimental.cttz.elts.i32.nxv4i32(<vscale x 4 x i32>, i1)
+declare i32 @llvm.experimental.cttz.elts.i32.v16i1(<16 x i1>, i1)
+declare i16 @llvm.experimental.cttz.elts.i16.v16i1(<8 x i1>, i1)
 
 attributes #0 = { vscale_range(2,1024) }
diff --git a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
index 49d4760a2e9ab..94b717b42e92b 100644
--- a/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
+++ b/llvm/test/CodeGen/RISCV/intrinsic-cttz-elts.ll
@@ -48,31 +48,13 @@ define i32 @ctz_v2i1_poison(<2 x i1> %a) {
 ; RV32-LABEL: ctz_v2i1_poison:
 ; RV32:       # %bb.0:
 ; RV32-NEXT:    vsetivli zero, 2, e8, mf8, ta, ma
-; RV32-NEXT:    vmv.v.i v8, 0
-; RV32-NEXT:    vmerge.vim v8, v8, -1, v0
-; RV32-NEXT:    vid.v v9
-; RV32-NEXT:    vrsub.vi v9, v9, 2
-; RV32-NEXT:    vand.vv v8, v8, v9
-; RV32-NEXT:    vredmaxu.vs v8, v8, v8
-; RV32-NEXT:    vmv.x.s a0, v8
-; RV32-NEXT:    li a1, 2
-; RV32-NEXT:    sub a1, a1, a0
-; RV32-NEXT:    andi a0, a1, 255
+; RV32-NEXT:    vfirst.m a0, v0
 ; RV32-NEXT:    ret
 ;
 ; RV64-LABEL: ctz_v2i1_poison:
 ; RV64:       # %bb.0:
 ; RV64-NEXT:    vsetivli zero, 2, e8, mf8, ta, ma
-; RV64-NEXT:    vmv.v.i v8, 0
-; RV64-NEXT:    vmerge.vim v8, v8, -1, v0
-; RV64-NEXT:    vid.v v9
-; RV64-NEXT:    vrsub.vi v9, v9, 2
-; RV64-NEXT:    vand.vv v8, v8, v9
-; RV64-NEXT:    vredmaxu.vs v8, v8, v8
-; RV64-NEXT:    vmv.x.s a0, v8
-; RV64-NEXT:    li a1, 2
-; RV64-NEXT:    subw a1, a1, a0
-; RV64-NEXT:    andi a0, a1, 255
+; RV64-NEXT:    vfirst.m a0, v0
 ; RV64-NEXT:    ret
   %res = call i32 @llvm.experimental.cttz.elts.i32.v2i1(<2 x i1> %a, i1 1)
   ret i32 %res

From 4b22a923c4bfd0aa1d483149f84b6787263c2d76 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Tue, 16 Apr 2024 10:42:32 +0900
Subject: [PATCH 512/886] [IndVars] Mark truncs as nuw/nsw (#88686)

When inserting truncs during IV widening, mark the trunc as either nuw
or nsw depending on whether zext or sext widening was used. For
non-negative IVs both nuw and nsw apply.
---
 llvm/include/llvm/IR/IRBuilder.h              | 14 ++++++-
 llvm/lib/Transforms/Utils/SimplifyIndVar.cpp  | 40 ++++++++++---------
 .../IndVarSimplify/AArch64/widen-loop-comp.ll |  8 ++--
 .../Transforms/IndVarSimplify/X86/iv-widen.ll | 16 ++++----
 .../Transforms/IndVarSimplify/elim-extend.ll  |  2 +-
 ...wide-inc-for-narrow-use-recompute-flags.ll |  2 +-
 .../test/Transforms/IndVarSimplify/iv-sext.ll |  2 +-
 .../IndVarSimplify/iv-widen-elim-ext.ll       |  8 ++--
 llvm/test/Transforms/IndVarSimplify/lftr.ll   |  4 +-
 .../IndVarSimplify/no-iv-rewrite.ll           |  2 +-
 .../IndVarSimplify/post-inc-range.ll          |  2 +-
 .../test/Transforms/IndVarSimplify/pr25578.ll |  2 +-
 .../test/Transforms/IndVarSimplify/pr55925.ll |  8 ++--
 .../widen-nonnegative-countdown.ll            | 22 +++++-----
 .../IndVarSimplify/widen-nonnegative.ll       | 20 +++++-----
 llvm/test/Transforms/LoopFlatten/widen-iv3.ll |  2 +-
 16 files changed, 84 insertions(+), 70 deletions(-)

diff --git a/llvm/include/llvm/IR/IRBuilder.h b/llvm/include/llvm/IR/IRBuilder.h
index f381273c46cfb..b6534a1962a2f 100644
--- a/llvm/include/llvm/IR/IRBuilder.h
+++ b/llvm/include/llvm/IR/IRBuilder.h
@@ -2004,8 +2004,18 @@ class IRBuilderBase {
   // Instruction creation methods: Cast/Conversion Operators
   //===--------------------------------------------------------------------===//
 
-  Value *CreateTrunc(Value *V, Type *DestTy, const Twine &Name = "") {
-    return CreateCast(Instruction::Trunc, V, DestTy, Name);
+  Value *CreateTrunc(Value *V, Type *DestTy, const Twine &Name = "",
+                     bool IsNUW = false, bool IsNSW = false) {
+    if (V->getType() == DestTy)
+      return V;
+    if (Value *Folded = Folder.FoldCast(Instruction::Trunc, V, DestTy))
+      return Folded;
+    Instruction *I = CastInst::Create(Instruction::Trunc, V, DestTy);
+    if (IsNUW)
+      I->setHasNoUnsignedWrap();
+    if (IsNSW)
+      I->setHasNoSignedWrap();
+    return Insert(I, Name);
   }
 
   Value *CreateZExt(Value *V, Type *DestTy, const Twine &Name = "",
diff --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
index 440fe0790d795..31be7d62c8d1d 100644
--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
@@ -1153,6 +1153,7 @@ class WidenIV {
 
   Instruction *widenIVUse(NarrowIVDefUse DU, SCEVExpander &Rewriter,
                           PHINode *OrigPhi, PHINode *WidePhi);
+  void truncateIVUse(NarrowIVDefUse DU);
 
   bool widenLoopCompare(NarrowIVDefUse DU);
   bool widenWithVariantUse(NarrowIVDefUse DU);
@@ -1569,15 +1570,18 @@ WidenIV::WidenedRecTy WidenIV::getWideRecurrence(WidenIV::NarrowIVDefUse DU) {
 
 /// This IV user cannot be widened. Replace this use of the original narrow IV
 /// with a truncation of the new wide IV to isolate and eliminate the narrow IV.
-static void truncateIVUse(WidenIV::NarrowIVDefUse DU, DominatorTree *DT,
-                          LoopInfo *LI) {
+void WidenIV::truncateIVUse(NarrowIVDefUse DU) {
   auto *InsertPt = getInsertPointForUses(DU.NarrowUse, DU.NarrowDef, DT, LI);
   if (!InsertPt)
     return;
   LLVM_DEBUG(dbgs() << "INDVARS: Truncate IV " << *DU.WideDef << " for user "
                     << *DU.NarrowUse << "\n");
+  ExtendKind ExtKind = getExtendKind(DU.NarrowDef);
   IRBuilder<> Builder(InsertPt);
-  Value *Trunc = Builder.CreateTrunc(DU.WideDef, DU.NarrowDef->getType());
+  Value *Trunc =
+      Builder.CreateTrunc(DU.WideDef, DU.NarrowDef->getType(), "",
+                          DU.NeverNegative || ExtKind == ExtendKind::Zero,
+                          DU.NeverNegative || ExtKind == ExtendKind::Sign);
   DU.NarrowUse->replaceUsesOfWith(DU.NarrowDef, Trunc);
 }
 
@@ -1826,6 +1830,13 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
   assert(ExtendKindMap.count(DU.NarrowDef) &&
          "Should already know the kind of extension used to widen NarrowDef");
 
+  // This narrow use can be widened by a sext if it's non-negative or its narrow
+  // def was widened by a sext. Same for zext.
+  bool CanWidenBySExt =
+      DU.NeverNegative || getExtendKind(DU.NarrowDef) == ExtendKind::Sign;
+  bool CanWidenByZExt =
+      DU.NeverNegative || getExtendKind(DU.NarrowDef) == ExtendKind::Zero;
+
   // Stop traversing the def-use chain at inner-loop phis or post-loop phis.
   if (PHINode *UsePhi = dyn_cast<PHINode>(DU.NarrowUse)) {
     if (LI->getLoopFor(UsePhi->getParent()) != L) {
@@ -1833,7 +1844,7 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
       // After SimplifyCFG most loop exit targets have a single predecessor.
       // Otherwise fall back to a truncate within the loop.
       if (UsePhi->getNumOperands() != 1)
-        truncateIVUse(DU, DT, LI);
+        truncateIVUse(DU);
       else {
         // Widening the PHI requires us to insert a trunc.  The logical place
         // for this trunc is in the same BB as the PHI.  This is not possible if
@@ -1847,7 +1858,8 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
         WidePhi->addIncoming(DU.WideDef, UsePhi->getIncomingBlock(0));
         BasicBlock *WidePhiBB = WidePhi->getParent();
         IRBuilder<> Builder(WidePhiBB, WidePhiBB->getFirstInsertionPt());
-        Value *Trunc = Builder.CreateTrunc(WidePhi, DU.NarrowDef->getType());
+        Value *Trunc = Builder.CreateTrunc(WidePhi, DU.NarrowDef->getType(), "",
+                                           CanWidenByZExt, CanWidenBySExt);
         UsePhi->replaceAllUsesWith(Trunc);
         DeadInsts.emplace_back(UsePhi);
         LLVM_DEBUG(dbgs() << "INDVARS: Widen lcssa phi " << *UsePhi << " to "
@@ -1857,18 +1869,9 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
     }
   }
 
-  // This narrow use can be widened by a sext if it's non-negative or its narrow
-  // def was widened by a sext. Same for zext.
-  auto canWidenBySExt = [&]() {
-    return DU.NeverNegative || getExtendKind(DU.NarrowDef) == ExtendKind::Sign;
-  };
-  auto canWidenByZExt = [&]() {
-    return DU.NeverNegative || getExtendKind(DU.NarrowDef) == ExtendKind::Zero;
-  };
-
   // Our raison d'etre! Eliminate sign and zero extension.
-  if ((match(DU.NarrowUse, m_SExtLike(m_Value())) && canWidenBySExt()) ||
-      (isa<ZExtInst>(DU.NarrowUse) && canWidenByZExt())) {
+  if ((match(DU.NarrowUse, m_SExtLike(m_Value())) && CanWidenBySExt) ||
+      (isa<ZExtInst>(DU.NarrowUse) && CanWidenByZExt)) {
     Value *NewDef = DU.WideDef;
     if (DU.NarrowUse->getType() != WideType) {
       unsigned CastWidth = SE->getTypeSizeInBits(DU.NarrowUse->getType());
@@ -1876,7 +1879,8 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
       if (CastWidth < IVWidth) {
         // The cast isn't as wide as the IV, so insert a Trunc.
         IRBuilder<> Builder(DU.NarrowUse);
-        NewDef = Builder.CreateTrunc(DU.WideDef, DU.NarrowUse->getType());
+        NewDef = Builder.CreateTrunc(DU.WideDef, DU.NarrowUse->getType(), "",
+                                     CanWidenByZExt, CanWidenBySExt);
       }
       else {
         // A wider extend was hidden behind a narrower one. This may induce
@@ -1975,7 +1979,7 @@ Instruction *WidenIV::widenIVUse(WidenIV::NarrowIVDefUse DU,
   // This user does not evaluate to a recurrence after widening, so don't
   // follow it. Instead insert a Trunc to kill off the original use,
   // eventually isolating the original narrow IV so it can be removed.
-  truncateIVUse(DU, DT, LI);
+  truncateIVUse(DU);
   return nullptr;
 }
 
diff --git a/llvm/test/Transforms/IndVarSimplify/AArch64/widen-loop-comp.ll b/llvm/test/Transforms/IndVarSimplify/AArch64/widen-loop-comp.ll
index 6f659a88da2e2..c5f656c870a23 100644
--- a/llvm/test/Transforms/IndVarSimplify/AArch64/widen-loop-comp.ll
+++ b/llvm/test/Transforms/IndVarSimplify/AArch64/widen-loop-comp.ll
@@ -41,7 +41,7 @@ define i32 @test1() {
 ; CHECK-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_COND]]
 ; CHECK:       if.then:
 ; CHECK-NEXT:    [[I_05_LCSSA_WIDE:%.*]] = phi i64 [ [[INDVARS_IV]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[I_05_LCSSA_WIDE]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc nuw nsw i64 [[I_05_LCSSA_WIDE]] to i32
 ; CHECK-NEXT:    store i32 [[TMP5]], ptr @idx, align 4
 ; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       for.cond.for.end.loopexit_crit_edge:
@@ -237,7 +237,7 @@ define i32 @test4(i32 %a) {
 ; CHECK-NEXT:    [[CONV3:%.*]] = trunc i32 [[OR]] to i8
 ; CHECK-NEXT:    [[CALL:%.*]] = call i32 @fn1(i8 signext [[CONV3]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i32 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i32 [[INDVARS_IV_NEXT]] to i8
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw i32 [[INDVARS_IV_NEXT]] to i8
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i8 [[TMP0]], -14
 ; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END:%.*]]
 ; CHECK:       for.end:
@@ -466,7 +466,7 @@ define i32 @test9(ptr %a, i32 %b, i32 %init) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 ; CHECK-NEXT:    [[ADD]] = add nsw i32 [[SUM_0]], [[TMP1]]
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 0, [[TMP2]]
 ; CHECK-NEXT:    br i1 [[CMP2]], label [[FOR_COND]], label [[FOR_END]]
 ; CHECK:       for.end:
@@ -997,7 +997,7 @@ define i32 @test16_unsigned_neg(i32 %start, ptr %p, ptr %q, i32 %x) {
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[BACKEDGE:%.*]] ], [ [[TMP0]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[COND:%.*]] = icmp eq i64 [[INDVARS_IV]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[FOO:%.*]] = add i32 [[TMP1]], -1
 ; CHECK-NEXT:    br i1 [[COND]], label [[EXIT:%.*]], label [[GUARDED:%.*]]
 ; CHECK:       guarded:
diff --git a/llvm/test/Transforms/IndVarSimplify/X86/iv-widen.ll b/llvm/test/Transforms/IndVarSimplify/X86/iv-widen.ll
index d05755bea0ddd..4e0c503794bfe 100644
--- a/llvm/test/Transforms/IndVarSimplify/X86/iv-widen.ll
+++ b/llvm/test/Transforms/IndVarSimplify/X86/iv-widen.ll
@@ -23,7 +23,7 @@ define void @loop_0(ptr %a) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[B18_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[B24:%.*]] ]
 ; CHECK-NEXT:    call void @use(i64 [[INDVARS_IV]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[O:%.*]] = getelementptr i32, ptr [[A:%.*]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[V:%.*]] = load i32, ptr [[O]], align 4
 ; CHECK-NEXT:    [[T:%.*]] = icmp eq i32 [[V]], 0
@@ -37,7 +37,7 @@ define void @loop_0(ptr %a) {
 ; CHECK-NEXT:    ret void
 ; CHECK:       exit24:
 ; CHECK-NEXT:    [[DOT02_LCSSA_WIDE:%.*]] = phi i64 [ [[INDVARS_IV]], [[B18]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[DOT02_LCSSA_WIDE]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[DOT02_LCSSA_WIDE]] to i32
 ; CHECK-NEXT:    call void @dummy(i32 [[TMP1]])
 ; CHECK-NEXT:    unreachable
 ;
@@ -159,7 +159,7 @@ declare void @dummy(i32)
 declare void @dummy.i64(i64)
 
 
-define void @loop_2(i32 %size, i32 %nsteps, i32 %hsize, ptr %lined, i8 %tmp1) {
+define void @loop_2(i32 %size, i32 %nsteps, i32 %hsize, ptr %lined, i8 %arg) {
 ; CHECK-LABEL: @loop_2(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[CMP215:%.*]] = icmp sgt i32 [[SIZE:%.*]], 1
@@ -180,12 +180,12 @@ define void @loop_2(i32 %size, i32 %nsteps, i32 %hsize, ptr %lined, i8 %tmp1) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 1, [[FOR_BODY2_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY2]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw i64 [[TMP3]], [[INDVARS_IV]]
 ; CHECK-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i8, ptr [[LINED:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    store i8 [[TMP1:%.*]], ptr [[ADD_PTR]], align 1
+; CHECK-NEXT:    store i8 [[ARG:%.*]], ptr [[ADD_PTR]], align 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY2]], label [[FOR_BODY3_PREHEADER:%.*]]
 ; CHECK:       for.body3.preheader:
-; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[TMP3]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc nsw i64 [[TMP3]] to i32
 ; CHECK-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
 ; CHECK-NEXT:    [[WIDE_TRIP_COUNT7:%.*]] = zext i32 [[SIZE]] to i64
 ; CHECK-NEXT:    br label [[FOR_BODY3:%.*]]
@@ -193,7 +193,7 @@ define void @loop_2(i32 %size, i32 %nsteps, i32 %hsize, ptr %lined, i8 %tmp1) {
 ; CHECK-NEXT:    [[INDVARS_IV3:%.*]] = phi i64 [ 1, [[FOR_BODY3_PREHEADER]] ], [ [[INDVARS_IV_NEXT4:%.*]], [[FOR_BODY3]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], [[INDVARS_IV3]]
 ; CHECK-NEXT:    [[ADD_PTR2:%.*]] = getelementptr inbounds i8, ptr [[LINED]], i64 [[TMP7]]
-; CHECK-NEXT:    store i8 [[TMP1]], ptr [[ADD_PTR2]], align 1
+; CHECK-NEXT:    store i8 [[ARG]], ptr [[ADD_PTR2]], align 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT4]] = add nuw nsw i64 [[INDVARS_IV3]], 1
 ; CHECK-NEXT:    [[EXITCOND8:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT4]], [[WIDE_TRIP_COUNT7]]
 ; CHECK-NEXT:    br i1 [[EXITCOND8]], label [[FOR_BODY3]], label [[FOR_INC_LOOPEXIT:%.*]]
@@ -222,7 +222,7 @@ for.body2:
   %add4 = add nsw i32 %add, %k
   %idx.ext = sext i32 %add4 to i64
   %add.ptr = getelementptr inbounds i8, ptr %lined, i64 %idx.ext
-  store i8 %tmp1, ptr %add.ptr, align 1
+  store i8 %arg, ptr %add.ptr, align 1
   %inc = add nsw i32 %k, 1
   %cmp2 = icmp slt i32 %inc, %size
   br i1 %cmp2, label %for.body2, label %for.body3
@@ -233,7 +233,7 @@ for.body3:
   %add5 = add nuw i32 %add, %l
   %idx.ext2 = zext i32 %add5 to i64
   %add.ptr2 = getelementptr inbounds i8, ptr %lined, i64 %idx.ext2
-  store i8 %tmp1, ptr %add.ptr2, align 1
+  store i8 %arg, ptr %add.ptr2, align 1
   %inc2 = add nsw i32 %l, 1
   %cmp3 = icmp slt i32 %inc2, %size
   br i1 %cmp3, label %for.body3, label %for.inc
diff --git a/llvm/test/Transforms/IndVarSimplify/elim-extend.ll b/llvm/test/Transforms/IndVarSimplify/elim-extend.ll
index 54bb9951ff66a..01c95dadd1626 100644
--- a/llvm/test/Transforms/IndVarSimplify/elim-extend.ll
+++ b/llvm/test/Transforms/IndVarSimplify/elim-extend.ll
@@ -142,7 +142,7 @@ define void @nestedIV(ptr %address, i32 %limit) nounwind {
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[INNERLOOP]], label [[INNEREXIT:%.*]]
 ; CHECK:       innerexit:
 ; CHECK-NEXT:    [[INNERCOUNT_LCSSA_WIDE:%.*]] = phi i64 [ [[INDVARS_IV_NEXT]], [[INNERLOOP]] ]
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[INNERCOUNT_LCSSA_WIDE]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nsw i64 [[INNERCOUNT_LCSSA_WIDE]] to i32
 ; CHECK-NEXT:    br label [[OUTERMERGE]]
 ; CHECK:       outermerge:
 ; CHECK-NEXT:    [[INNERCOUNT_MERGE]] = phi i32 [ [[TMP3]], [[INNEREXIT]] ], [ [[INNERCOUNT]], [[INNERPREHEADER]] ]
diff --git a/llvm/test/Transforms/IndVarSimplify/hoist-wide-inc-for-narrow-use-recompute-flags.ll b/llvm/test/Transforms/IndVarSimplify/hoist-wide-inc-for-narrow-use-recompute-flags.ll
index cc99ee312ccb7..1135ca9dbf00d 100644
--- a/llvm/test/Transforms/IndVarSimplify/hoist-wide-inc-for-narrow-use-recompute-flags.ll
+++ b/llvm/test/Transforms/IndVarSimplify/hoist-wide-inc-for-narrow-use-recompute-flags.ll
@@ -15,7 +15,7 @@ define void @test_pr82243(ptr %f) {
 ; CHECK-NEXT:    [[GEP_IV_EXT:%.*]] = getelementptr i32, ptr [[F]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store i32 1, ptr [[GEP_IV_EXT]], align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[SHL:%.*]] = shl i32 123, [[TMP0]]
 ; CHECK-NEXT:    [[GEP_SHL:%.*]] = getelementptr i32, ptr [[F]], i32 [[SHL]]
 ; CHECK-NEXT:    br label [[INNER_HEADER:%.*]]
diff --git a/llvm/test/Transforms/IndVarSimplify/iv-sext.ll b/llvm/test/Transforms/IndVarSimplify/iv-sext.ll
index 450913f16baa2..95a036f0e54c7 100644
--- a/llvm/test/Transforms/IndVarSimplify/iv-sext.ll
+++ b/llvm/test/Transforms/IndVarSimplify/iv-sext.ll
@@ -99,7 +99,7 @@ define void @t(ptr %pval1, ptr %peakWeight, ptr %nrgReducePeakrate, i32 %bandEdg
 ; CHECK-NEXT:    [[VAL35_LCSSA:%.*]] = phi float [ [[VAL35]], [[BB5]] ]
 ; CHECK-NEXT:    [[VAL31_LCSSA_WIDE:%.*]] = phi i64 [ [[INDVARS_IV_NEXT]], [[BB5]] ]
 ; CHECK-NEXT:    [[VAL30_LCSSA:%.*]] = phi float [ [[VAL30]], [[BB5]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[VAL31_LCSSA_WIDE]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc nsw i64 [[VAL31_LCSSA_WIDE]] to i32
 ; CHECK-NEXT:    br label [[BB7]]
 ; CHECK:       bb7:
 ; CHECK-NEXT:    [[DISTERBHI_2_LCSSA]] = phi float [ [[VAL30_LCSSA]], [[BB5_BB7_CRIT_EDGE]] ], [ [[DISTERBHI_0_PH]], [[BB5_PREHEADER]] ]
diff --git a/llvm/test/Transforms/IndVarSimplify/iv-widen-elim-ext.ll b/llvm/test/Transforms/IndVarSimplify/iv-widen-elim-ext.ll
index 59a0241bfe9fd..a83e9ce74b12a 100644
--- a/llvm/test/Transforms/IndVarSimplify/iv-widen-elim-ext.ll
+++ b/llvm/test/Transforms/IndVarSimplify/iv-widen-elim-ext.ll
@@ -22,7 +22,7 @@ define void @foo(ptr %A, ptr %B, ptr %C, i32 %N) {
 ; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[TMP1]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
 ; CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP0]], [[TMP2]]
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw nsw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    [[DIV0:%.*]] = udiv i32 5, [[TMP3]]
 ; CHECK-NEXT:    [[ADD4:%.*]] = add nsw i32 [[ADD3]], [[DIV0]]
 ; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
@@ -224,7 +224,7 @@ define i32 @foo3(i32 %M) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = add nsw i64 [[INDVARS_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc nsw i64 [[TMP3]] to i32
 ; CHECK-NEXT:    [[IDXPROM4:%.*]] = zext i32 [[TMP4]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x i32], ptr @a, i64 0, i64 [[IDXPROM4]]
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX5]], align 4
@@ -365,7 +365,7 @@ define i32 @foo5(ptr %input, i32 %length, ptr %in) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 1, [[FOR_BODY_LR_PH]] ]
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[INPUT]], align 8
-; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc nuw nsw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP4]], [[TMP5]]
 ; CHECK-NEXT:    [[IDX_EXT:%.*]] = sext i32 [[MUL]] to i64
 ; CHECK-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i32, ptr [[IN:%.*]], i64 [[IDX_EXT]]
@@ -514,7 +514,7 @@ define void @foo7(i32 %n, ptr %a, i32 %x) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = or disjoint i64 [[TMP2]], 1
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], [[TMP0]]
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i64 [[INDVARS_IV_NEXT]], [[TMP1]]
diff --git a/llvm/test/Transforms/IndVarSimplify/lftr.ll b/llvm/test/Transforms/IndVarSimplify/lftr.ll
index 41db925de577e..7f4820f093e55 100644
--- a/llvm/test/Transforms/IndVarSimplify/lftr.ll
+++ b/llvm/test/Transforms/IndVarSimplify/lftr.ll
@@ -525,7 +525,7 @@ define float @wide_trip_count_test3(ptr %b,
 ; CHECK-NEXT:    [[TMP0:%.*]] = add nsw i64 [[INDVARS_IV]], 20
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TEMP:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul float [[CONV]], [[TEMP]]
 ; CHECK-NEXT:    [[ADD1]] = fadd float [[SUM_07]], [[MUL]]
@@ -584,7 +584,7 @@ define float @wide_trip_count_test4(ptr %b,
 ; CHECK-NEXT:    [[TMP0:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 20
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TEMP:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul float [[CONV]], [[TEMP]]
 ; CHECK-NEXT:    [[ADD1]] = fadd float [[SUM_07]], [[MUL]]
diff --git a/llvm/test/Transforms/IndVarSimplify/no-iv-rewrite.ll b/llvm/test/Transforms/IndVarSimplify/no-iv-rewrite.ll
index c35c5bacf68ca..579b8536cedf0 100644
--- a/llvm/test/Transforms/IndVarSimplify/no-iv-rewrite.ll
+++ b/llvm/test/Transforms/IndVarSimplify/no-iv-rewrite.ll
@@ -213,7 +213,7 @@ define void @maxvisitor(i32 %limit, ptr %base) nounwind {
 ; CHECK-NEXT:    [[CMP19:%.*]] = icmp sgt i32 [[VAL]], [[MAX]]
 ; CHECK-NEXT:    br i1 [[CMP19]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    br label [[LOOP_INC]]
 ; CHECK:       if.else:
 ; CHECK-NEXT:    br label [[LOOP_INC]]
diff --git a/llvm/test/Transforms/IndVarSimplify/post-inc-range.ll b/llvm/test/Transforms/IndVarSimplify/post-inc-range.ll
index 5c22ba1044b60..bbdee0267effb 100644
--- a/llvm/test/Transforms/IndVarSimplify/post-inc-range.ll
+++ b/llvm/test/Transforms/IndVarSimplify/post-inc-range.ll
@@ -180,7 +180,7 @@ define void @test_neg(ptr %array_length_ptr, ptr %base,
 ; CHECK-NEXT:    br label [[FOR_INC]]
 ; CHECK:       for.inc:
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP2]], [[LIMIT:%.*]]
 ; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END]]
 ; CHECK:       for.end:
diff --git a/llvm/test/Transforms/IndVarSimplify/pr25578.ll b/llvm/test/Transforms/IndVarSimplify/pr25578.ll
index d8adc178474c0..380e8171798b0 100644
--- a/llvm/test/Transforms/IndVarSimplify/pr25578.ll
+++ b/llvm/test/Transforms/IndVarSimplify/pr25578.ll
@@ -13,7 +13,7 @@ L1_header:
 
 ; CHECK: L2_header:
 ; CHECK: %[[INDVAR:.*]] = phi i64
-; CHECK: %[[TRUNC:.*]] = trunc i64 %[[INDVAR]] to i32
+; CHECK: %[[TRUNC:.*]] = trunc nuw nsw i64 %[[INDVAR]] to i32
 L2_header:
   %i = phi i32 [ 0, %L1_header ], [ %i_next, %L2_latch ]
   %i_prom = sext i32 %i to i64
diff --git a/llvm/test/Transforms/IndVarSimplify/pr55925.ll b/llvm/test/Transforms/IndVarSimplify/pr55925.ll
index 312a8295ccdc9..2ad187add4e10 100644
--- a/llvm/test/Transforms/IndVarSimplify/pr55925.ll
+++ b/llvm/test/Transforms/IndVarSimplify/pr55925.ll
@@ -14,11 +14,11 @@ define void @test(ptr %p) personality ptr undef {
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[RES:%.*]] = invoke i32 @foo(i32 returned [[TMP0]])
 ; CHECK-NEXT:            to label [[LOOP_LATCH]] unwind label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @foo(i32 [[TMP1]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    br label [[LOOP]]
@@ -56,8 +56,8 @@ define void @test_critedge(i1 %c, ptr %p) personality ptr undef {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br i1 [[C:%.*]], label [[LOOP_INVOKE:%.*]], label [[LOOP_OTHER:%.*]]
 ; CHECK:       loop.invoke:
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[RES:%.*]] = invoke i32 @foo(i32 returned [[TMP0]])
 ; CHECK-NEXT:            to label [[LOOP_LATCH]] unwind label [[EXIT:%.*]]
 ; CHECK:       loop.other:
diff --git a/llvm/test/Transforms/IndVarSimplify/widen-nonnegative-countdown.ll b/llvm/test/Transforms/IndVarSimplify/widen-nonnegative-countdown.ll
index d473103f5824e..9c8983421029f 100644
--- a/llvm/test/Transforms/IndVarSimplify/widen-nonnegative-countdown.ll
+++ b/llvm/test/Transforms/IndVarSimplify/widen-nonnegative-countdown.ll
@@ -223,7 +223,7 @@ define void @sext_postinc(ptr %A, i32 %start) {
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP1]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
@@ -262,7 +262,7 @@ define void @sext_preinc(ptr %A, i32 %start) {
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP1]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
@@ -366,7 +366,7 @@ define void @zext_postinc_offset_constant_one(ptr %A, i32 %start) {
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[TMP0]], [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[J_016_US:%.*]] = phi i32 [ [[INC_US:%.*]], [[FOR_BODY]] ], [ [[START]], [[FOR_BODY_PREHEADER]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nuw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP1]], 1
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -513,13 +513,13 @@ define void @sext_postinc_offset_constant_one(ptr %A, i32 %start) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[TMP0]], [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP1]], 1
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP2]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
@@ -556,13 +556,13 @@ define void @sext_preinc_offset_constant_one(ptr %A, i32 %start) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[TMP0]], [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add nuw i32 [[TMP1]], 1
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP2]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
@@ -808,13 +808,13 @@ define void @sext_postinc_offset_constant_minus_one(ptr %A, i32 %start) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[TMP0]], [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP1]], -1
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[INDVARS_IV_NEXT]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP2]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
@@ -851,13 +851,13 @@ define void @sext_preinc_offset_constant_minus_one(ptr %A, i32 %start) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[TMP0]], [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP1]], -1
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[CMP2_US:%.*]] = icmp ugt i32 [[TMP2]], 6
 ; CHECK-NEXT:    br i1 [[CMP2_US]], label [[FOR_BODY]], label [[EXIT_LOOPEXIT:%.*]]
 ; CHECK:       exit.loopexit:
diff --git a/llvm/test/Transforms/IndVarSimplify/widen-nonnegative.ll b/llvm/test/Transforms/IndVarSimplify/widen-nonnegative.ll
index 739db26311f4a..e00eaafa3f192 100644
--- a/llvm/test/Transforms/IndVarSimplify/widen-nonnegative.ll
+++ b/llvm/test/Transforms/IndVarSimplify/widen-nonnegative.ll
@@ -150,7 +150,7 @@ define void @sext_add_nuw(ptr %A, i32 %offset, i32 %M) {
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDVARS_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[TMP2]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
@@ -185,7 +185,7 @@ define void @sext_add_noflags(ptr %A, i32 %offset, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP0]], [[OFFSET:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -223,7 +223,7 @@ define void @zext_add_nsw(ptr %A, i32 %offset, i32 %M) {
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nsw i64 [[INDVARS_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext i32 [[TMP2]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
@@ -293,7 +293,7 @@ define void @zext_add_noflags(ptr %A, i32 %offset, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP0]], [[OFFSET:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -399,7 +399,7 @@ define void @zext_nneg_add_noflags(ptr %A, i32 %offset, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[ADD_US:%.*]] = add i32 [[TMP0]], [[OFFSET:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext nneg i32 [[ADD_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -475,7 +475,7 @@ define void @sext_mul_nuw(ptr %A, i32 %multiple, i32 %M) {
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i64 [[INDVARS_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nuw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[TMP2]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
@@ -510,7 +510,7 @@ define void @sext_mul_noflags(ptr %A, i32 %multiple, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[MUL_US:%.*]] = mul i32 [[TMP0]], [[MULTIPLE:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = sext i32 [[MUL_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -548,7 +548,7 @@ define void @zext_mul_nsw(ptr %A, i32 %multiple, i32 %M) {
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nsw i64 [[INDVARS_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc nsw i64 [[TMP1]] to i32
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext i32 [[TMP2]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
 ; CHECK-NEXT:    tail call void @use_ptr(ptr [[ARRAYIDX_US]])
@@ -618,7 +618,7 @@ define void @zext_mul_noflags(ptr %A, i32 %multiple, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[MUL_US:%.*]] = mul i32 [[TMP0]], [[MULTIPLE:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext i32 [[MUL_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
@@ -724,7 +724,7 @@ define void @zext_nneg_mul_noflags(ptr %A, i32 %multiple, i32 %M) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDVARS_IV]] to i32
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc nuw nsw i64 [[INDVARS_IV]] to i32
 ; CHECK-NEXT:    [[MUL_US:%.*]] = mul i32 [[TMP0]], [[MULTIPLE:%.*]]
 ; CHECK-NEXT:    [[IDXPROM_US:%.*]] = zext nneg i32 [[MUL_US]] to i64
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IDXPROM_US]]
diff --git a/llvm/test/Transforms/LoopFlatten/widen-iv3.ll b/llvm/test/Transforms/LoopFlatten/widen-iv3.ll
index 6e6c045661c24..3ac5a69a496ff 100644
--- a/llvm/test/Transforms/LoopFlatten/widen-iv3.ll
+++ b/llvm/test/Transforms/LoopFlatten/widen-iv3.ll
@@ -35,7 +35,7 @@ define i16 @foo() {
 ; CHECK-NEXT:    [[SUM_110:%.*]] = phi i16 [ [[SUM_012]], [[FOR_COND1_PREHEADER]] ], [ [[ADD5]], [[FOR_BODY4]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[INDVAR]], [[TMP0]]
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i16 [[J_011]], [[MUL]]
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc i32 [[TMP2]] to i16
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc nuw nsw i32 [[TMP2]] to i16
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [64 x i16], ptr @v, i16 0, i16 [[TMP3]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i16, ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[ADD5]] = add nsw i16 [[TMP4]], [[SUM_110]]

From 5f680724838188f516d349bd9459710308d721e0 Mon Sep 17 00:00:00 2001
From: Qizhi Hu <836744285@qq.com>
Date: Tue, 16 Apr 2024 09:57:23 +0800
Subject: [PATCH 513/886] [Clang][Sema] Fix issue on requires expression with
 templated base class member function (#85198)

Fix https://github.com/llvm/llvm-project/issues/84020
Skip checking implicit object parameter in the context of
`RequiresExprBodyDecl`.

Co-authored-by: huqizhi <836744285@qq.com>
---
 clang/docs/ReleaseNotes.rst    |  1 +
 clang/lib/Sema/SemaExpr.cpp    |  3 ++-
 clang/test/SemaCXX/PR84020.cpp | 23 +++++++++++++++++++++++
 3 files changed, 26 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/SemaCXX/PR84020.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index dc108785f6cc9..76701dc723b6c 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -538,6 +538,7 @@ Bug Fixes to C++ Support
   object parameter.
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
 - Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
+- Fix a crash in requires expression with templated base class member function. Fixes (#GH84020).
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 24f354f1c7249..189764cb4b6b0 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -7739,7 +7739,8 @@ ExprResult Sema::BuildResolvedCallExpr(Expr *Fn, NamedDecl *NDecl,
   }
 
   if (CXXMethodDecl *Method = dyn_cast_or_null<CXXMethodDecl>(FDecl))
-    if (Method->isImplicitObjectMemberFunction())
+    if (!isa<RequiresExprBodyDecl>(CurContext) &&
+        Method->isImplicitObjectMemberFunction())
       return ExprError(Diag(LParenLoc, diag::err_member_call_without_object)
                        << Fn->getSourceRange() << 0);
 
diff --git a/clang/test/SemaCXX/PR84020.cpp b/clang/test/SemaCXX/PR84020.cpp
new file mode 100644
index 0000000000000..8ea5dcc4527ae
--- /dev/null
+++ b/clang/test/SemaCXX/PR84020.cpp
@@ -0,0 +1,23 @@
+// RUN: %clang_cc1 -std=c++20 -verify %s
+// RUN: %clang_cc1 -std=c++23 -verify %s
+// expected-no-diagnostics
+
+struct B {
+    template <typename S>
+    void foo();
+
+    void bar();
+};
+
+template <typename T, typename S>
+struct A : T {
+    auto foo() {
+        static_assert(requires { T::template foo<S>(); });
+        static_assert(requires { T::bar(); });
+    }
+};
+
+int main() {
+    A<B, double> a;
+    a.foo();
+}

From 568368a43e5b4adb3c5d105a0eff3e0c13c0af8c Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Mon, 15 Apr 2024 19:05:30 -0700
Subject: [PATCH 514/886] [Support] Make readNext default to unaligned (#88808)

Without this patch, you would typically use readNext as:

  readNext<uint32_t, llvm::endianness::little, unaligned>(Ptr)

which is quite mouthful.  Since most serialization/deserialization
operations are unaligned accesses, this patch makes the alignment
template parameter default to unaligned, allowing us to say:

  readNext<uint32_t, llvm::endianness::little>(Ptr)

I'm including a few examples of migration in this patch.  I'll do the
rest in a separate patch.

Note that writeNext already has the same trick for the alignment
template parameter.
---
 llvm/include/llvm/Support/Endian.h          | 4 ++--
 llvm/include/llvm/Support/OnDiskHashTable.h | 9 +++------
 2 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/llvm/include/llvm/Support/Endian.h b/llvm/include/llvm/Support/Endian.h
index 1cdb5ca0d5eaa..30e0852b972c5 100644
--- a/llvm/include/llvm/Support/Endian.h
+++ b/llvm/include/llvm/Support/Endian.h
@@ -80,8 +80,8 @@ template <typename value_type, std::size_t alignment, typename CharT>
   return ret;
 }
 
-template <typename value_type, endianness endian, std::size_t alignment,
-          typename CharT>
+template <typename value_type, endianness endian,
+          std::size_t alignment = unaligned, typename CharT>
 [[nodiscard]] inline value_type readNext(const CharT *&memory) {
   return readNext<value_type, alignment, CharT>(memory, endian);
 }
diff --git a/llvm/include/llvm/Support/OnDiskHashTable.h b/llvm/include/llvm/Support/OnDiskHashTable.h
index 0a8cbbd8b1883..f6b4055e74de7 100644
--- a/llvm/include/llvm/Support/OnDiskHashTable.h
+++ b/llvm/include/llvm/Support/OnDiskHashTable.h
@@ -368,14 +368,12 @@ template <typename Info> class OnDiskChainedHashTable {
 
     // 'Items' starts with a 16-bit unsigned integer representing the
     // number of items in this bucket.
-    unsigned Len =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Items);
+    unsigned Len = endian::readNext<uint16_t, llvm::endianness::little>(Items);
 
     for (unsigned i = 0; i < Len; ++i) {
       // Read the hash.
       hash_value_type ItemHash =
-          endian::readNext<hash_value_type, llvm::endianness::little,
-                           unaligned>(Items);
+          endian::readNext<hash_value_type, llvm::endianness::little>(Items);
 
       // Determine the length of the key and the data.
       const std::pair<offset_type, offset_type> &L =
@@ -473,8 +471,7 @@ class OnDiskIterableChainedHashTable : public OnDiskChainedHashTable<Info> {
         // 'Items' starts with a 16-bit unsigned integer representing the
         // number of items in this bucket.
         NumItemsInBucketLeft =
-            endian::readNext<uint16_t, llvm::endianness::little, unaligned>(
-                Ptr);
+            endian::readNext<uint16_t, llvm::endianness::little>(Ptr);
       }
       Ptr += sizeof(hash_value_type); // Skip the hash.
       // Determine the length of the key and the data.

From b4cf63d26f4c41dd9403c4e62500d82a6d31d692 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Mon, 15 Apr 2024 20:27:04 -0600
Subject: [PATCH 515/886] [X86] Remove obsolete tablegen rules for near data in
 small static code model (#84523)

These should be already handled by other code.

Removing the kernel code model rules right above it cause
bss_pagealigned.ll to fail by using a movabsq to get the address of a
global, haven't figured out where that code is yet.
---
 llvm/lib/Target/X86/X86InstrCompiler.td | 22 ----------------------
 1 file changed, 22 deletions(-)

diff --git a/llvm/lib/Target/X86/X86InstrCompiler.td b/llvm/lib/Target/X86/X86InstrCompiler.td
index 270dd32c7235a..2350636be10ef 100644
--- a/llvm/lib/Target/X86/X86InstrCompiler.td
+++ b/llvm/lib/Target/X86/X86InstrCompiler.td
@@ -1247,28 +1247,6 @@ def : Pat<(i64 (X86Wrapper mcsym:$dst)),
 def : Pat<(i64 (X86Wrapper tblockaddress:$dst)),
           (MOV64ri32 tblockaddress:$dst)>, Requires<[KernelCode]>;
 
-// If we have small model and -static mode, it is safe to store global addresses
-// directly as immediates.  FIXME: This is really a hack, the 'imm' predicate
-// for MOV64mi32 should handle this sort of thing.
-def : Pat<(store (i64 (X86Wrapper tconstpool:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, tconstpool:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-def : Pat<(store (i64 (X86Wrapper tjumptable:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, tjumptable:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-def : Pat<(store (i64 (X86Wrapper tglobaladdr:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, tglobaladdr:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-def : Pat<(store (i64 (X86Wrapper texternalsym:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, texternalsym:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-def : Pat<(store (i64 (X86Wrapper mcsym:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, mcsym:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-def : Pat<(store (i64 (X86Wrapper tblockaddress:$src)), addr:$dst),
-          (MOV64mi32 addr:$dst, tblockaddress:$src)>,
-          Requires<[NearData, IsNotPIC]>;
-
 def : Pat<(i32 (X86RecoverFrameAlloc mcsym:$dst)), (MOV32ri mcsym:$dst)>;
 def : Pat<(i64 (X86RecoverFrameAlloc mcsym:$dst)), (MOV64ri mcsym:$dst)>;
 

From 82f479ba315a417b6cd01a8c2efdc15c26689f2e Mon Sep 17 00:00:00 2001
From: Usama Hameed <u_hameed@apple.com>
Date: Mon, 15 Apr 2024 19:42:45 -0700
Subject: [PATCH 516/886] Add asan tests for libsanitizers. (#88349)

This patch tests LLDB integration with libsanitizers for ASan.

rdar://111856681
---
 lldb/test/API/functionalities/asan/Makefile   |  6 +-
 .../functionalities/asan/TestMemoryHistory.py | 73 ++++++++++++++++++-
 .../functionalities/asan/TestReportData.py    | 20 ++++-
 .../API/functionalities/libsanitizers/util.py |  3 +
 4 files changed, 97 insertions(+), 5 deletions(-)
 create mode 100644 lldb/test/API/functionalities/libsanitizers/util.py

diff --git a/lldb/test/API/functionalities/asan/Makefile b/lldb/test/API/functionalities/asan/Makefile
index 4913a18d8cc6f..d66696fed7078 100644
--- a/lldb/test/API/functionalities/asan/Makefile
+++ b/lldb/test/API/functionalities/asan/Makefile
@@ -1,4 +1,8 @@
 C_SOURCES := main.c
-CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
+asan: CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
+asan: all
+
+libsanitizers: CFLAGS_EXTRAS := -fsanitize=address -fsanitize-stable-abi -g -gcolumn-info
+libsanitizers: all
 
 include Makefile.rules
diff --git a/lldb/test/API/functionalities/asan/TestMemoryHistory.py b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
index 00162ae8822c7..ee7939203ead1 100644
--- a/lldb/test/API/functionalities/asan/TestMemoryHistory.py
+++ b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
@@ -9,15 +9,21 @@
 from lldbsuite.test import lldbplatform
 from lldbsuite.test import lldbutil
 
+from functionalities.libsanitizers.util import no_libsanitizers
 
 class AsanTestCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
     @expectedFailureNetBSD
     @skipUnlessAddressSanitizer
     def test(self):
-        self.build()
+        self.build(make_targets=["asan"])
         self.asan_tests()
 
+    @skipIf(oslist=no_match(["macosx"]))
+    def test_libsanitizers_asan(self):
+        self.build(make_targets=["libsanitizers"])
+        self.libsanitizer_tests()
+
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -26,6 +32,71 @@ def setUp(self):
         self.line_free = line_number("main.c", "// free line")
         self.line_breakpoint = line_number("main.c", "// break line")
 
+    # Test line numbers: rdar://126237493
+    def libsanitizer_tests(self):
+        target = self.createTestTarget()
+
+        if no_libsanitizers(self):
+            self.skipTest("libsanitizers not found")
+
+        self.runCmd(
+            "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
+        )
+
+        self.runCmd("run")
+
+        # In libsanitizers, memory history is not supported until a report has been generated
+        self.expect(
+            "thread list",
+            "Process should be stopped due to ASan report",
+            substrs=["stopped", "stop reason = Use of deallocated memory"],
+        )
+
+        # test the 'memory history' command
+        self.expect(
+            "memory history 'pointer'",
+            substrs=[
+                "Memory deallocated by Thread",
+                "a.out`f2",
+                "main.c",
+                "Memory allocated by Thread",
+                "a.out`f1",
+                "main.c",
+            ],
+        )
+
+        # do the same using SB API
+        process = self.dbg.GetSelectedTarget().process
+        val = (
+            process.GetSelectedThread().GetSelectedFrame().EvaluateExpression("pointer")
+        )
+        addr = val.GetValueAsUnsigned()
+        threads = process.GetHistoryThreads(addr)
+        self.assertEqual(threads.GetSize(), 2)
+
+        history_thread = threads.GetThreadAtIndex(0)
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
+        history_thread = threads.GetThreadAtIndex(1)
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
+        # let's free the container (SBThreadCollection) and see if the
+        # SBThreads still live
+        threads = None
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
     def asan_tests(self):
         target = self.createTestTarget()
 
diff --git a/lldb/test/API/functionalities/asan/TestReportData.py b/lldb/test/API/functionalities/asan/TestReportData.py
index 543c5fe66a208..de0c1206a57ad 100644
--- a/lldb/test/API/functionalities/asan/TestReportData.py
+++ b/lldb/test/API/functionalities/asan/TestReportData.py
@@ -9,6 +9,7 @@
 from lldbsuite.test.lldbtest import *
 from lldbsuite.test import lldbutil
 
+from functionalities.libsanitizers.util import no_libsanitizers
 
 class AsanTestReportDataCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
@@ -16,9 +17,14 @@ class AsanTestReportDataCase(TestBase):
     @skipUnlessAddressSanitizer
     @skipIf(archs=["i386"], bugnumber="llvm.org/PR36710")
     def test(self):
-        self.build()
+        self.build(make_targets=["asan"])
         self.asan_tests()
 
+    @skipIf(oslist=no_match(["macosx"]))
+    def test_libsanitizers_asan(self):
+        self.build(make_targets=["libsanitizers"])
+        self.asan_tests(libsanitizers=True)
+
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -29,10 +35,18 @@ def setUp(self):
         self.line_crash = line_number("main.c", "// BOOM line")
         self.col_crash = 16
 
-    def asan_tests(self):
+    def asan_tests(self, libsanitizers=False):
         target = self.createTestTarget()
 
-        self.registerSanitizerLibrariesWithTarget(target)
+        if libsanitizers and no_libsanitizers(self):
+            self.skipTest("libsanitizers not found")
+
+        if libsanitizers:
+            self.runCmd(
+                "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
+            )
+        else:
+            self.registerSanitizerLibrariesWithTarget(target)
 
         self.runCmd("run")
 
diff --git a/lldb/test/API/functionalities/libsanitizers/util.py b/lldb/test/API/functionalities/libsanitizers/util.py
new file mode 100644
index 0000000000000..ad68541aba8d0
--- /dev/null
+++ b/lldb/test/API/functionalities/libsanitizers/util.py
@@ -0,0 +1,3 @@
+def no_libsanitizers(testbase):
+    testbase.runCmd("image list libsystem_sanitizers.dylib", check=False)
+    return not "libsystem_sanitizers.dylib" in testbase.res.GetOutput()

From 00ae4b738be6e840cfdb3e7461409c3e49403e50 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Tue, 16 Apr 2024 03:49:28 +0000
Subject: [PATCH 517/886] Revert "[X86] Remove obsolete tablegen rules for near
 data in small static code model (#84523)"

This reverts commit b4cf63d26f4c41dd9403c4e62500d82a6d31d692.

Breaks indirect-branch-tracking-eh2.ll.
---
 llvm/lib/Target/X86/X86InstrCompiler.td | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/llvm/lib/Target/X86/X86InstrCompiler.td b/llvm/lib/Target/X86/X86InstrCompiler.td
index 2350636be10ef..270dd32c7235a 100644
--- a/llvm/lib/Target/X86/X86InstrCompiler.td
+++ b/llvm/lib/Target/X86/X86InstrCompiler.td
@@ -1247,6 +1247,28 @@ def : Pat<(i64 (X86Wrapper mcsym:$dst)),
 def : Pat<(i64 (X86Wrapper tblockaddress:$dst)),
           (MOV64ri32 tblockaddress:$dst)>, Requires<[KernelCode]>;
 
+// If we have small model and -static mode, it is safe to store global addresses
+// directly as immediates.  FIXME: This is really a hack, the 'imm' predicate
+// for MOV64mi32 should handle this sort of thing.
+def : Pat<(store (i64 (X86Wrapper tconstpool:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, tconstpool:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+def : Pat<(store (i64 (X86Wrapper tjumptable:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, tjumptable:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+def : Pat<(store (i64 (X86Wrapper tglobaladdr:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, tglobaladdr:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+def : Pat<(store (i64 (X86Wrapper texternalsym:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, texternalsym:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+def : Pat<(store (i64 (X86Wrapper mcsym:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, mcsym:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+def : Pat<(store (i64 (X86Wrapper tblockaddress:$src)), addr:$dst),
+          (MOV64mi32 addr:$dst, tblockaddress:$src)>,
+          Requires<[NearData, IsNotPIC]>;
+
 def : Pat<(i32 (X86RecoverFrameAlloc mcsym:$dst)), (MOV32ri mcsym:$dst)>;
 def : Pat<(i64 (X86RecoverFrameAlloc mcsym:$dst)), (MOV64ri mcsym:$dst)>;
 

From 8ee7d9732204a7415967bce2c1d2c0fa63af264f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Mon, 15 Apr 2024 20:51:31 -0700
Subject: [PATCH 518/886] [flang][cuda] Add fir.cuda_allocate operation
 (#88586)

Allocatable with cuda device attribute have special semantic for the
allocate statement. In flang the allocate statement is lowered to a
sequence of runtime call initializing the descriptor and then allocating
the descriptor data. This new operation will replace the last runtime
call and abstract all the device memory allocation needed.
The lowering patch will follow.
---
 .../include/flang/Optimizer/Dialect/FIROps.td | 32 +++++++++
 .../flang/Optimizer/Dialect/FIRTypes.td       |  1 +
 flang/lib/Optimizer/Dialect/FIROps.cpp        | 19 +++++
 flang/test/Fir/cuf-invalid.fir                | 50 +++++++++++++
 flang/test/Fir/cuf.mlir                       | 70 +++++++++++++++++++
 5 files changed, 172 insertions(+)
 create mode 100644 flang/test/Fir/cuf-invalid.fir
 create mode 100644 flang/test/Fir/cuf.mlir

diff --git a/flang/include/flang/Optimizer/Dialect/FIROps.td b/flang/include/flang/Optimizer/Dialect/FIROps.td
index dff1cdb20cbfe..c181c7ed62dff 100644
--- a/flang/include/flang/Optimizer/Dialect/FIROps.td
+++ b/flang/include/flang/Optimizer/Dialect/FIROps.td
@@ -3190,4 +3190,36 @@ def fir_CUDADataTransferOp : fir_Op<"cuda_data_transfer", []> {
   }];
 }
 
+def fir_CUDAAllocateOp : fir_Op<"cuda_allocate", [AttrSizedOperandSegments,
+    MemoryEffects<[MemAlloc<DefaultResource>]>]> {
+  let summary = "Perform the device allocation of data of an allocatable";
+
+  let description = [{
+    The fir.cuda_allocate operation performs the allocation on the device
+    of the data of an allocatable. The descriptor passed to the operation
+    is initialized before with the standard flang runtime calls.
+  }];
+
+  let arguments = (ins Arg<AnyRefOrBoxType, "", [MemWrite]>:$box,
+                       Arg<Optional<AnyRefOrBoxType>, "", [MemWrite]>:$errmsg,
+                       Optional<AnyIntegerType>:$stream,
+                       Arg<Optional<AnyRefOrBoxType>, "", [MemWrite]>:$pinned,
+                       Arg<Optional<AnyRefOrBoxType>, "", [MemRead]>:$source,
+                       fir_CUDADataAttributeAttr:$cuda_attr,
+                       UnitAttr:$hasStat);
+
+  let results = (outs AnyIntegerType:$stat);
+
+  let assemblyFormat = [{
+    $box `:` qualified(type($box))
+    ( `source` `(` $source^ `:` qualified(type($source) )`)` )?
+    ( `errmsg` `(` $errmsg^ `:` type($errmsg) `)` )?
+    ( `stream` `(` $stream^ `:` type($stream) `)` )?
+    ( `pinned` `(` $pinned^ `:` type($pinned) `)` )?
+    attr-dict `->` type($stat)
+  }];
+
+  let hasVerifier = 1;
+}
+
 #endif
diff --git a/flang/include/flang/Optimizer/Dialect/FIRTypes.td b/flang/include/flang/Optimizer/Dialect/FIRTypes.td
index 4c6a8064991ab..3b876e4642da9 100644
--- a/flang/include/flang/Optimizer/Dialect/FIRTypes.td
+++ b/flang/include/flang/Optimizer/Dialect/FIRTypes.td
@@ -625,6 +625,7 @@ def AnyRefOrBoxLike : TypeConstraint<Or<[AnyReferenceLike.predicate,
 def AnyRefOrBox : TypeConstraint<Or<[fir_ReferenceType.predicate,
     fir_HeapType.predicate, fir_PointerType.predicate,
     IsBaseBoxTypePred]>, "any reference or box">;
+def AnyRefOrBoxType : Type<AnyRefOrBox.predicate, "any legal ref or box type">;
 
 def AnyShapeLike : TypeConstraint<Or<[fir_ShapeType.predicate,
     fir_ShapeShiftType.predicate]>, "any legal shape type">;
diff --git a/flang/lib/Optimizer/Dialect/FIROps.cpp b/flang/lib/Optimizer/Dialect/FIROps.cpp
index 8ab74103cb6a8..88710880174d2 100644
--- a/flang/lib/Optimizer/Dialect/FIROps.cpp
+++ b/flang/lib/Optimizer/Dialect/FIROps.cpp
@@ -3993,6 +3993,25 @@ mlir::LogicalResult fir::CUDAKernelOp::verify() {
   return mlir::success();
 }
 
+mlir::LogicalResult fir::CUDAAllocateOp::verify() {
+  if (getPinned() && getStream())
+    return emitOpError("pinned and stream cannot appears at the same time");
+  if (!fir::unwrapRefType(getBox().getType()).isa<fir::BaseBoxType>())
+    return emitOpError(
+        "expect box to be a reference to/or a class or box type value");
+  if (getSource() &&
+      !fir::unwrapRefType(getSource().getType()).isa<fir::BaseBoxType>())
+    return emitOpError(
+        "expect source to be a reference to/or a class or box type value");
+  if (getErrmsg() &&
+      !fir::unwrapRefType(getErrmsg().getType()).isa<fir::BoxType>())
+    return emitOpError(
+        "expect errmsg to be a reference to/or a box type value");
+  if (getErrmsg() && !getHasStat())
+    return emitOpError("expect stat attribute when errmsg is provided");
+  return mlir::success();
+}
+
 //===----------------------------------------------------------------------===//
 // FIROpsDialect
 //===----------------------------------------------------------------------===//
diff --git a/flang/test/Fir/cuf-invalid.fir b/flang/test/Fir/cuf-invalid.fir
new file mode 100644
index 0000000000000..9c5ffe7176a3b
--- /dev/null
+++ b/flang/test/Fir/cuf-invalid.fir
@@ -0,0 +1,50 @@
+// RUN: fir-opt -split-input-file -verify-diagnostics %s
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %1 = fir.alloca i32
+  %pinned = fir.alloca i1
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %s = fir.load %1 : !fir.ref<i32>
+  // expected-error@+1{{'fir.cuda_allocate' op pinned and stream cannot appears at the same time}}
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> stream(%s : i32) pinned(%pinned : !fir.ref<i1>) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// -----
+
+func.func @_QPsub1() {
+  %1 = fir.alloca i32
+  // expected-error@+1{{'fir.cuda_allocate' op expect box to be a reference to/or a class or box type value}}
+  %2 = fir.cuda_allocate %1 : !fir.ref<i32> {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %c100 = arith.constant 100 : index
+  %7 = fir.alloca !fir.char<1,100> {bindc_name = "msg", uniq_name = "_QFsub1Emsg"}
+  %8:2 = hlfir.declare %7 typeparams %c100 {uniq_name = "_QFsub1Emsg"} : (!fir.ref<!fir.char<1,100>>, index) -> (!fir.ref<!fir.char<1,100>>, !fir.ref<!fir.char<1,100>>)
+  %9 = fir.embox %8#1 : (!fir.ref<!fir.char<1,100>>) -> !fir.box<!fir.char<1,100>>
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %16 = fir.convert %9 : (!fir.box<!fir.char<1,100>>) -> !fir.box<none>
+  // expected-error@+1{{'fir.cuda_allocate' op expect stat attribute when errmsg is provided}}
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> errmsg(%16 : !fir.box<none>) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %1 = fir.alloca i32
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  // expected-error@+1{{'fir.cuda_allocate' op expect errmsg to be a reference to/or a box type value}}
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> errmsg(%1 : !fir.ref<i32>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+  return
+}
diff --git a/flang/test/Fir/cuf.mlir b/flang/test/Fir/cuf.mlir
new file mode 100644
index 0000000000000..67eff31b35b2b
--- /dev/null
+++ b/flang/test/Fir/cuf.mlir
@@ -0,0 +1,70 @@
+// RUN: fir-opt --split-input-file %s | fir-opt --split-input-file | FileCheck %s
+
+// Simple round trip test of operations.
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %1 = fir.alloca i32
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %s = fir.load %1 : !fir.ref<i32>
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> stream(%s : i32) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> stream(%{{.*}} : i32) {cuda_attr = #fir.cuda<device>} -> i32
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %1 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "b", uniq_name = "_QFsub1Eb"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %5:2 = hlfir.declare %1 {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %12 = fir.convert %5#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> source(%12 : !fir.ref<!fir.box<none>>) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> source(%{{.*}} : !fir.ref<!fir.box<none>>) {cuda_attr = #fir.cuda<device>} -> i32
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %pinned = fir.alloca i1
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> pinned(%pinned : !fir.ref<i1>) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> pinned(%{{.*}} : !fir.ref<i1>) {cuda_attr = #fir.cuda<device>} -> i32
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %c100 = arith.constant 100 : index
+  %7 = fir.alloca !fir.char<1,100> {bindc_name = "msg", uniq_name = "_QFsub1Emsg"}
+  %8:2 = hlfir.declare %7 typeparams %c100 {uniq_name = "_QFsub1Emsg"} : (!fir.ref<!fir.char<1,100>>, index) -> (!fir.ref<!fir.char<1,100>>, !fir.ref<!fir.char<1,100>>)
+  %9 = fir.embox %8#1 : (!fir.ref<!fir.char<1,100>>) -> !fir.box<!fir.char<1,100>>
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %16 = fir.convert %9 : (!fir.box<!fir.char<1,100>>) -> !fir.box<none>
+  %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> errmsg(%16 : !fir.box<none>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+  return
+}

From fe48bf672e1ab293368a3212203db94a4e21c533 Mon Sep 17 00:00:00 2001
From: Chelsea Cassanova <chelsea_cassanova@apple.com>
Date: Mon, 15 Apr 2024 21:26:18 -0700
Subject: [PATCH 519/886] [lldb][lit] Guard MallocNanoZone envvar in shell
 tests (#88824)

Previously the MallocNanoZone envvar would be set to 0 on Darwin for the
LLDB shell tests, but this should guarded behind ASan being enabled as
opposed to simply running the test suite behind Darwin. This required
that the LLVM_USE_SANITIZER option be added as an attribute to the lit
config for shell tests.
---
 lldb/test/Shell/lit.cfg.py         | 12 ++++++++----
 lldb/test/Shell/lit.site.cfg.py.in |  1 +
 2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/lldb/test/Shell/lit.cfg.py b/lldb/test/Shell/lit.cfg.py
index 290569576ac80..e24f3fbb4d931 100644
--- a/lldb/test/Shell/lit.cfg.py
+++ b/lldb/test/Shell/lit.cfg.py
@@ -50,10 +50,14 @@
 )
 
 # Enable sanitizer runtime flags.
-config.environment["ASAN_OPTIONS"] = "detect_stack_use_after_return=1"
-config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
-if platform.system() == "Darwin":
-    config.environment["MallocNanoZone"] = "0"
+if "Address" in config.llvm_use_sanitizer:
+    config.environment["ASAN_OPTIONS"] = "detect_stack_use_after_return=1"
+    if platform.system() == "Darwin":
+        config.environment["MallocNanoZone"] = "0"
+
+if "Thread" in config.llvm_use_sanitizer:
+    config.environment["TSAN_OPTIONS"] = "halt_on_error=1"
+
 
 # Support running the test suite under the lldb-repro wrapper. This makes it
 # possible to capture a test suite run and then rerun all the test from the
diff --git a/lldb/test/Shell/lit.site.cfg.py.in b/lldb/test/Shell/lit.site.cfg.py.in
index 736dfc335732b..b69e7bce1bc0b 100644
--- a/lldb/test/Shell/lit.site.cfg.py.in
+++ b/lldb/test/Shell/lit.site.cfg.py.in
@@ -26,6 +26,7 @@ config.lldb_enable_lua = @LLDB_ENABLE_LUA@
 config.lldb_build_directory = "@LLDB_TEST_BUILD_DIRECTORY@"
 config.have_lldb_server = @LLDB_TOOL_LLDB_SERVER_BUILD@
 config.lldb_system_debugserver = @LLDB_USE_SYSTEM_DEBUGSERVER@
+config.llvm_use_sanitizer = "@LLVM_USE_SANITIZER@"
 # The shell tests use their own module caches.
 config.lldb_module_cache = os.path.join("@LLDB_TEST_MODULE_CACHE_LLDB@", "lldb-shell")
 config.clang_module_cache = os.path.join("@LLDB_TEST_MODULE_CACHE_CLANG@", "lldb-shell")

From 65b0cc610f80d9b9724a98cf7c5bcfd38e1cf799 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 21:54:26 -0700
Subject: [PATCH 520/886] [RISCV] Add FeatureStdExtI to all CPUs in
 RISCVProcessors.td. NFC (#88805)

This is currently being implied in RISCVISAInfo.cpp. Make it explicit.

I'm planning to move all extension information to RISCVFeatures.td and
have tablegen create the tables for RISCVISAInfo.cpp. This requires
making the creation of RISCVTargetParserDef.inc in tablegen independent
of RISCVISAInfo.cpp. So we need an accurate extension list for CPUs in
tablegen.
---
 llvm/lib/Target/RISCV/RISCVProcessors.td | 27 ++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVProcessors.td b/llvm/lib/Target/RISCV/RISCVProcessors.td
index fd6d6078ec238..739b50749e132 100644
--- a/llvm/lib/Target/RISCV/RISCVProcessors.td
+++ b/llvm/lib/Target/RISCV/RISCVProcessors.td
@@ -56,11 +56,13 @@ class RISCVTuneProcessorModel<string n,
 
 def GENERIC_RV32 : RISCVProcessorModel<"generic-rv32",
                                        NoSchedModel,
-                                       [Feature32Bit]>,
+                                       [Feature32Bit,
+                                        FeatureStdExtI]>,
                    GenericTuneInfo;
 def GENERIC_RV64 : RISCVProcessorModel<"generic-rv64",
                                        NoSchedModel,
-                                       [Feature64Bit]>,
+                                       [Feature64Bit,
+                                        FeatureStdExtI]>,
                    GenericTuneInfo;
 // Support generic for compatibility with other targets. The triple will be used
 // to change to the appropriate rv32/rv64 version.
@@ -69,11 +71,13 @@ def : ProcessorModel<"generic", NoSchedModel, []>, GenericTuneInfo;
 def ROCKET_RV32 : RISCVProcessorModel<"rocket-rv32",
                                       RocketModel,
                                       [Feature32Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtZicsr]>;
 def ROCKET_RV64 : RISCVProcessorModel<"rocket-rv64",
                                       RocketModel,
                                       [Feature64Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtZicsr]>;
 def ROCKET : RISCVTuneProcessorModel<"rocket",
@@ -86,6 +90,7 @@ def SIFIVE_7 : RISCVTuneProcessorModel<"sifive-7-series",
 def SIFIVE_E20 : RISCVProcessorModel<"sifive-e20",
                                      RocketModel,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZicsr,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
@@ -94,6 +99,7 @@ def SIFIVE_E20 : RISCVProcessorModel<"sifive-e20",
 def SIFIVE_E21 : RISCVProcessorModel<"sifive-e21",
                                      RocketModel,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZicsr,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
@@ -103,6 +109,7 @@ def SIFIVE_E21 : RISCVProcessorModel<"sifive-e21",
 def SIFIVE_E24 : RISCVProcessorModel<"sifive-e24",
                                      RocketModel,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -112,6 +119,7 @@ def SIFIVE_E24 : RISCVProcessorModel<"sifive-e24",
 def SIFIVE_E31 : RISCVProcessorModel<"sifive-e31",
                                      RocketModel,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtZicsr,
                                       FeatureStdExtM,
@@ -121,6 +129,7 @@ def SIFIVE_E31 : RISCVProcessorModel<"sifive-e31",
 def SIFIVE_E34 : RISCVProcessorModel<"sifive-e34",
                                      RocketModel,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -130,6 +139,7 @@ def SIFIVE_E34 : RISCVProcessorModel<"sifive-e34",
 def SIFIVE_E76 : RISCVProcessorModel<"sifive-e76",
                                      SiFive7Model,
                                      [Feature32Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -140,6 +150,7 @@ def SIFIVE_E76 : RISCVProcessorModel<"sifive-e76",
 def SIFIVE_S21 : RISCVProcessorModel<"sifive-s21",
                                      RocketModel,
                                      [Feature64Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZicsr,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
@@ -149,6 +160,7 @@ def SIFIVE_S21 : RISCVProcessorModel<"sifive-s21",
 def SIFIVE_S51 : RISCVProcessorModel<"sifive-s51",
                                      RocketModel,
                                      [Feature64Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZicsr,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
@@ -158,6 +170,7 @@ def SIFIVE_S51 : RISCVProcessorModel<"sifive-s51",
 def SIFIVE_S54 : RISCVProcessorModel<"sifive-s54",
                                       RocketModel,
                                       [Feature64Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtM,
                                        FeatureStdExtA,
@@ -168,6 +181,7 @@ def SIFIVE_S54 : RISCVProcessorModel<"sifive-s54",
 def SIFIVE_S76 : RISCVProcessorModel<"sifive-s76",
                                      SiFive7Model,
                                      [Feature64Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -180,6 +194,7 @@ def SIFIVE_S76 : RISCVProcessorModel<"sifive-s76",
 def SIFIVE_U54 : RISCVProcessorModel<"sifive-u54",
                                      RocketModel,
                                      [Feature64Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -190,6 +205,7 @@ def SIFIVE_U54 : RISCVProcessorModel<"sifive-u54",
 def SIFIVE_U74 : RISCVProcessorModel<"sifive-u74",
                                      SiFive7Model,
                                      [Feature64Bit,
+                                      FeatureStdExtI,
                                       FeatureStdExtZifencei,
                                       FeatureStdExtM,
                                       FeatureStdExtA,
@@ -200,6 +216,7 @@ def SIFIVE_U74 : RISCVProcessorModel<"sifive-u74",
 
 def SIFIVE_X280 : RISCVProcessorModel<"sifive-x280", SiFive7Model,
                                       [Feature64Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtM,
                                        FeatureStdExtA,
@@ -217,6 +234,7 @@ def SIFIVE_X280 : RISCVProcessorModel<"sifive-x280", SiFive7Model,
 
 def SIFIVE_P450 : RISCVProcessorModel<"sifive-p450", SiFiveP400Model,
                                       [Feature64Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtM,
                                        FeatureStdExtA,
@@ -247,6 +265,7 @@ def SIFIVE_P450 : RISCVProcessorModel<"sifive-p450", SiFiveP400Model,
 
 def SIFIVE_P670 : RISCVProcessorModel<"sifive-p670", SiFiveP600Model,
                                       [Feature64Bit,
+                                       FeatureStdExtI,
                                        FeatureStdExtZifencei,
                                        FeatureStdExtM,
                                        FeatureStdExtA,
@@ -286,6 +305,7 @@ def SIFIVE_P670 : RISCVProcessorModel<"sifive-p670", SiFiveP600Model,
 def SYNTACORE_SCR1_BASE : RISCVProcessorModel<"syntacore-scr1-base",
                                               SyntacoreSCR1Model,
                                               [Feature32Bit,
+                                               FeatureStdExtI,
                                                FeatureStdExtZicsr,
                                                FeatureStdExtZifencei,
                                                FeatureStdExtC],
@@ -294,6 +314,7 @@ def SYNTACORE_SCR1_BASE : RISCVProcessorModel<"syntacore-scr1-base",
 def SYNTACORE_SCR1_MAX : RISCVProcessorModel<"syntacore-scr1-max",
                                              SyntacoreSCR1Model,
                                              [Feature32Bit,
+                                              FeatureStdExtI,
                                               FeatureStdExtZicsr,
                                               FeatureStdExtZifencei,
                                               FeatureStdExtM,
@@ -303,6 +324,7 @@ def SYNTACORE_SCR1_MAX : RISCVProcessorModel<"syntacore-scr1-max",
 def VENTANA_VEYRON_V1 : RISCVProcessorModel<"veyron-v1",
                                             NoSchedModel,
                                             [Feature64Bit,
+                                             FeatureStdExtI,
                                              FeatureStdExtZifencei,
                                              FeatureStdExtZicsr,
                                              FeatureStdExtZicntr,
@@ -332,6 +354,7 @@ def VENTANA_VEYRON_V1 : RISCVProcessorModel<"veyron-v1",
 def XIANGSHAN_NANHU : RISCVProcessorModel<"xiangshan-nanhu",
                                           XiangShanNanHuModel,
                                           [Feature64Bit,
+                                           FeatureStdExtI,
                                            FeatureStdExtZicsr,
                                            FeatureStdExtZifencei,
                                            FeatureStdExtM,

From 4e2d11f87a595d658f25ec5838b002e4bf5a26fc Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 21:55:50 -0700
Subject: [PATCH 521/886] [RISCV] Remove unnecessary NoHasStdExtZicfiss
 Predicate from cmop.1 and cmop.5. NFC (#88822)

I'm not sure what this was supposed do. Maybe it was for the
disassembler, but that should be managed through DecoderNamespace.

So let's remove it and simplify the code.
---
 llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
index dd13a07d606d0..65091aad91003 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
@@ -20,13 +20,7 @@ class CMOPInst<bits<3> imm3, string opcodestr>
   let Inst{12-11} = 0;
 }
 
-// CMOP1, CMOP5 is used by Zicfiss.
-let Predicates = [HasStdExtZcmop, NoHasStdExtZicfiss] in {
-  def CMOP1 : CMOPInst<0, "cmop.1">, Sched<[]>;
-  def CMOP5 : CMOPInst<2, "cmop.5">, Sched<[]>;
-}
-
-foreach n = [3, 7, 9, 11, 13, 15] in {
+foreach n = [1, 3, 5, 7, 9, 11, 13, 15] in {
   let Predicates = [HasStdExtZcmop] in
   def CMOP # n : CMOPInst<!srl(n, 1), "cmop." # n>, Sched<[]>;
 }

From edb0708dc1ceeaeb3356311a4ddf72a0dc9b224f Mon Sep 17 00:00:00 2001
From: Vlad Mishel <43666597+vmishelcs@users.noreply.github.com>
Date: Mon, 15 Apr 2024 22:05:29 -0700
Subject: [PATCH 522/886] [InstCombine] Implement `fcmp (fadd x, 0.0), y` =>
 `fcmp x, y` optimization (#88476)

This PR addresses issue #88168. It implements an optimization for
the case of

```
define i1 @fcmp_fadd_zero_ugt(float %x, float %y) {
  %add = fadd float %x, 0.000000e+00
  %cmp = fcmp ugt float %add, %y
  ret i1 %cmp
}
```
`=>`
```
define i1 @fcmp_fadd_zero_ugt(float %x, float %y) {
  %cmp = fcmp ugt float %x, %y
  ret i1 %cmp
}
```
and all other types of `fcmp` instructions (`uge`, `ogt`, etc).

Proofs:
`fadd x, 0.0` https://alive2.llvm.org/ce/z/7FzNnM
`fsub x, 0.0` https://alive2.llvm.org/ce/z/puUxLK
---
 .../InstCombine/InstCombineCompares.cpp       |   8 +
 llvm/test/Transforms/InstCombine/fcmp.ll      | 202 ++++++++++++++++++
 2 files changed, 210 insertions(+)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 90550cdbdf891..ee783eed190a7 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -8097,6 +8097,14 @@ Instruction *InstCombinerImpl::visitFCmpInst(FCmpInst &I) {
         return new FCmpInst(I.getSwappedPredicate(), X, NegC, "", &I);
   }
 
+  // fcmp (fadd X, 0.0), Y --> fcmp X, Y
+  if (match(Op0, m_FAdd(m_Value(X), m_AnyZeroFP())))
+    return new FCmpInst(Pred, X, Op1, "", &I);
+
+  // fcmp X, (fadd Y, 0.0) --> fcmp X, Y
+  if (match(Op1, m_FAdd(m_Value(Y), m_AnyZeroFP())))
+    return new FCmpInst(Pred, Op0, Y, "", &I);
+
   if (match(Op0, m_FPExt(m_Value(X)))) {
     // fcmp (fpext X), (fpext Y) -> fcmp X, Y
     if (match(Op1, m_FPExt(m_Value(Y))) && X->getType() == Y->getType())
diff --git a/llvm/test/Transforms/InstCombine/fcmp.ll b/llvm/test/Transforms/InstCombine/fcmp.ll
index f2701d16d0f3d..069512b0f2d8e 100644
--- a/llvm/test/Transforms/InstCombine/fcmp.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp.ll
@@ -1284,3 +1284,205 @@ define <1 x i1> @bitcast_1vec_eq0(i32 %x) {
   %cmp = fcmp oeq <1 x float> %f, zeroinitializer
   ret <1 x i1> %cmp
 }
+
+; Simplify fcmp (x + 0.0), y => fcmp x, y
+
+define i1 @fcmp_fadd_zero_ugt(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ugt(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ugt float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ugt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_uge(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_uge(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp uge float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp uge float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ogt(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ogt(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ogt float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ogt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_oge(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_oge(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp oge float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ult(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ult(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ult float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ult float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ule(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ule(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ule float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ule float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_olt(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_olt(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp olt float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp olt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ole(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ole(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ole float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ole float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_oeq(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_oeq(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oeq float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp oeq float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_one(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_one(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp one float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp one float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ueq(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ueq(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ueq float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ueq float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_une(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_une(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp une float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_ord(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_ord(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ord float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp ord float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_uno(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_uno(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp uno float [[ADD:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp uno float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_neg_zero(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_neg_zero(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ugt float [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, -0.000000e+00
+  %cmp = fcmp ugt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_zero_switched(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_switched(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ult float [[Y:%.*]], [[X:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %y, 0.000000e+00
+  %cmp = fcmp ugt float %x, %add
+  ret i1 %cmp
+}
+
+define <2 x i1> @fcmp_fadd_zero_vec(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @fcmp_fadd_zero_vec(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ugt <2 x float> [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret <2 x i1> [[CMP]]
+;
+  %add = fadd <2 x float> %x, <float 0.0, float -0.0>
+  %cmp = fcmp ugt <2 x float> %add, %y
+  ret <2 x i1> %cmp
+}
+
+define i1 @fcmp_fast_fadd_fast_zero(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fast_fadd_fast_zero(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp fast ugt float [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd fast float %x, 0.000000e+00
+  %cmp = fcmp fast ugt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fast_fadd_zero(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fast_fadd_zero(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp fast ugt float [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd float %x, 0.000000e+00
+  %cmp = fcmp fast ugt float %add, %y
+  ret i1 %cmp
+}
+
+define i1 @fcmp_fadd_fast_zero(float %x, float %y) {
+; CHECK-LABEL: @fcmp_fadd_fast_zero(
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp ugt float [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = fadd fast float %x, 0.000000e+00
+  %cmp = fcmp ugt float %add, %y
+  ret i1 %cmp
+}

From 39016e33b0fe78ddb1f11822f71a8a233af4dca9 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Tue, 16 Apr 2024 13:00:06 +0800
Subject: [PATCH 523/886] [C++20] [Modules] Don't import non-inline function
 bodies even if it is always-inline

Recommit
https://github.com/llvm/llvm-project/commit/1ecbab56dcbb78268c8d19af34a50591f90b12a0

Close https://github.com/llvm/llvm-project/issues/80949

The new thing in this commit is to allow to import the function body
from instantiations if it is marked with always-inline. See the
discussion in https://github.com/llvm/llvm-project/issues/86893 for
details.
---
 clang/lib/CodeGen/CodeGenModule.cpp           | 17 ++++++++--
 .../CodeGenCXX/module-funcs-from-imports.cppm | 31 ++++++++++++++++---
 2 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index e44749672d582..0c447b20cef40 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -3952,9 +3952,20 @@ bool CodeGenModule::shouldEmitFunction(GlobalDecl GD) {
   // behavior may break ABI compatibility of the current unit.
   if (const Module *M = F->getOwningModule();
       M && M->getTopLevelModule()->isNamedModule() &&
-      getContext().getCurrentNamedModule() != M->getTopLevelModule() &&
-      !F->hasAttr<AlwaysInlineAttr>())
-    return false;
+      getContext().getCurrentNamedModule() != M->getTopLevelModule()) {
+    // There are practices to mark template member function as always-inline
+    // and mark the template as extern explicit instantiation but not give
+    // the definition for member function. So we have to emit the function
+    // from explicitly instantiation with always-inline.
+    //
+    // See https://github.com/llvm/llvm-project/issues/86893 for details.
+    //
+    // TODO: Maybe it is better to give it a warning if we call a non-inline
+    // function from other module units which is marked as always-inline.
+    if (!F->isTemplateInstantiation() || !F->hasAttr<AlwaysInlineAttr>()) {
+      return false;
+    }
+  }
 
   if (F->hasAttr<NoInlineAttr>())
     return false;
diff --git a/clang/test/CodeGenCXX/module-funcs-from-imports.cppm b/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
index 33cdf437110a9..a2a9122fc3913 100644
--- a/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
+++ b/clang/test/CodeGenCXX/module-funcs-from-imports.cppm
@@ -23,6 +23,21 @@ int func_in_gmf_not_called() {
     return 44;
 }
 
+template <class T>
+class A {
+public:
+    __attribute__((always_inline))
+    inline constexpr int getValue() {
+        return 43;
+    }
+
+    inline constexpr int getValue2() {
+        return 43;
+    }
+};
+
+extern template class A<char>;
+
 //--- M.cppm
 module;
 #include "foo.h"
@@ -47,17 +62,21 @@ int always_inline_func() {
     return 45;
 }
 
+export using ::A;
+
 //--- Use.cpp
 import M;
 int use() {
-    return exported_func() + always_inline_func();
+    A<char> a;
+    return exported_func() + always_inline_func() +
+           a.getValue() + a.getValue2();
 }
 
-// Checks that none of the function (except the always_inline_func) in the importees
-// are generated in the importer's code.
 // CHECK-O0: define{{.*}}_Z3usev(
 // CHECK-O0: declare{{.*}}_ZW1M13exported_funcv(
-// CHECK-O0: define{{.*}}available_externally{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O0: declare{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O0: define{{.*}}@_ZN1AIcE8getValueEv(
+// CHECK-O0: declare{{.*}}@_ZN1AIcE9getValue2Ev(
 // CHECK-O0-NOT: func_in_gmf
 // CHECK-O0-NOT: func_in_gmf_not_called
 // CHECK-O0-NOT: non_exported_func
@@ -68,7 +87,9 @@ int use() {
 // O0 to keep consistent ABI.
 // CHECK-O1: define{{.*}}_Z3usev(
 // CHECK-O1: declare{{.*}}_ZW1M13exported_funcv(
-// CHECK-O1: define{{.*}}available_externally{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O1: declare{{.*}}_ZW1M18always_inline_funcv(
+// CHECK-O1: define{{.*}}@_ZN1AIcE8getValueEv(
+// CHECK-O1: declare{{.*}}@_ZN1AIcE9getValue2Ev(
 // CHECK-O1-NOT: func_in_gmf
 // CHECK-O1-NOT: func_in_gmf_not_called
 // CHECK-O1-NOT: non_exported_func

From f14146fd46dd8bf7ef1e64d3c13af73dc5668045 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 22:33:14 -0700
Subject: [PATCH 524/886] [RISCV] Correct the spelling of the Zcmop mnemonics.
 (#88826)

The instruction names should be c.mop.1 instead of cmop.1.
---
 llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td |  2 +-
 llvm/test/MC/RISCV/rv32zcmop-invalid.s       |  6 ++--
 llvm/test/MC/RISCV/rvzcmop-valid.s           | 32 ++++++++++----------
 3 files changed, 20 insertions(+), 20 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
index 65091aad91003..32e7f962aa2ab 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZcmop.td
@@ -22,5 +22,5 @@ class CMOPInst<bits<3> imm3, string opcodestr>
 
 foreach n = [1, 3, 5, 7, 9, 11, 13, 15] in {
   let Predicates = [HasStdExtZcmop] in
-  def CMOP # n : CMOPInst<!srl(n, 1), "cmop." # n>, Sched<[]>;
+  def C_MOP # n : CMOPInst<!srl(n, 1), "c.mop." # n>, Sched<[]>;
 }
diff --git a/llvm/test/MC/RISCV/rv32zcmop-invalid.s b/llvm/test/MC/RISCV/rv32zcmop-invalid.s
index 71d72d59b0209..fb6252f7f0760 100644
--- a/llvm/test/MC/RISCV/rv32zcmop-invalid.s
+++ b/llvm/test/MC/RISCV/rv32zcmop-invalid.s
@@ -1,7 +1,7 @@
 # RUN: not llvm-mc -triple riscv32 -mattr=+zcmop < %s 2>&1 | FileCheck %s
 
-cmop.0 # CHECK: :[[@LINE]]:1: error: unrecognized instruction mnemonic
+c.mop.0 # CHECK: :[[@LINE]]:1: error: unrecognized instruction mnemonic
 
-cmop.1 t0 # CHECK: :[[@LINE]]:8: error: invalid operand for instruction
+c.mop.1 t0 # CHECK: :[[@LINE]]:9: error: invalid operand for instruction
 
-cmop.1 0x0 # CHECK: :[[@LINE]]:8: error: invalid operand for instruction
+c.mop.1 0x0 # CHECK: :[[@LINE]]:9: error: invalid operand for instruction
diff --git a/llvm/test/MC/RISCV/rvzcmop-valid.s b/llvm/test/MC/RISCV/rvzcmop-valid.s
index c6bb4a1580825..dd5d26ac5dd0c 100644
--- a/llvm/test/MC/RISCV/rvzcmop-valid.s
+++ b/llvm/test/MC/RISCV/rvzcmop-valid.s
@@ -9,34 +9,34 @@
 # RUN:     | llvm-objdump --mattr=+zcmop -d -r - \
 # RUN:     | FileCheck --check-prefix=CHECK-ASM-AND-OBJ %s
 
-# CHECK-ASM-AND-OBJ: cmop.1
+# CHECK-ASM-AND-OBJ: c.mop.1
 # CHECK-ASM: encoding: [0x81,0x60]
-cmop.1
+c.mop.1
 
-# CHECK-ASM-AND-OBJ: cmop.3
+# CHECK-ASM-AND-OBJ: c.mop.3
 # CHECK-ASM: encoding: [0x81,0x61]
-cmop.3
+c.mop.3
 
-# CHECK-ASM-AND-OBJ: cmop.5
+# CHECK-ASM-AND-OBJ: c.mop.5
 # CHECK-ASM: encoding: [0x81,0x62]
-cmop.5
+c.mop.5
 
-# CHECK-ASM-AND-OBJ: cmop.7
+# CHECK-ASM-AND-OBJ: c.mop.7
 # CHECK-ASM: encoding: [0x81,0x63]
-cmop.7
+c.mop.7
 
-# CHECK-ASM-AND-OBJ: cmop.9
+# CHECK-ASM-AND-OBJ: c.mop.9
 # CHECK-ASM: encoding: [0x81,0x64]
-cmop.9
+c.mop.9
 
-# CHECK-ASM-AND-OBJ: cmop.11
+# CHECK-ASM-AND-OBJ: c.mop.11
 # CHECK-ASM: encoding: [0x81,0x65]
-cmop.11
+c.mop.11
 
-# CHECK-ASM-AND-OBJ: cmop.13
+# CHECK-ASM-AND-OBJ: c.mop.13
 # CHECK-ASM: encoding: [0x81,0x66]
-cmop.13
+c.mop.13
 
-# CHECK-ASM-AND-OBJ: cmop.15
+# CHECK-ASM-AND-OBJ: c.mop.15
 # CHECK-ASM: encoding: [0x81,0x67]
-cmop.15
+c.mop.15

From 17d6bf046cea381413895f91e24d26d65763b59a Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Mon, 15 Apr 2024 22:39:09 -0700
Subject: [PATCH 525/886] [RISCV] Change how MMO is rebuilt in
 lowerFixedLengthVectorLoadToRVV/lowerFixedLengthVectorStoreToRVV (#88811)

Copy the pointer info, flags, alignment, AAInfo, and ranges, but let
getLoad rebuild the MMO using the scalable type used for the the new
load/store. This makes sure the LLT minimum size matches the ContainerVT
minimum size. This is important since vscale_range may have been used to
determine that the fixed vector was the exact size of a scalable vector.

Fixes #88799
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 19 +++++--------------
 llvm/test/CodeGen/RISCV/rvv/pr88799.ll      | 19 +++++++++++++++++++
 2 files changed, 24 insertions(+), 14 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/rvv/pr88799.ll

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index f6ed6420c9e1f..1d1ea6bae6c10 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -10433,14 +10433,10 @@ RISCVTargetLowering::lowerFixedLengthVectorLoadToRVV(SDValue Op,
   if (MinVLMAX == MaxVLMAX && MinVLMAX == VT.getVectorNumElements() &&
       getLMUL1VT(ContainerVT).bitsLE(ContainerVT)) {
     MachineMemOperand *MMO = Load->getMemOperand();
-    MachineFunction &MF = DAG.getMachineFunction();
-    MMO = MF.getMachineMemOperand(
-        MMO, MMO->getPointerInfo(),
-        MMO->getMemoryType().isValid()
-            ? LLT::scalable_vector(1, MMO->getMemoryType().getSizeInBits())
-            : MMO->getMemoryType());
     SDValue NewLoad =
-        DAG.getLoad(ContainerVT, DL, Load->getChain(), Load->getBasePtr(), MMO);
+        DAG.getLoad(ContainerVT, DL, Load->getChain(), Load->getBasePtr(),
+                    MMO->getPointerInfo(), MMO->getBaseAlign(), MMO->getFlags(),
+                    MMO->getAAInfo(), MMO->getRanges());
     SDValue Result = convertFromScalableVector(VT, NewLoad, DAG, Subtarget);
     return DAG.getMergeValues({Result, NewLoad.getValue(1)}, DL);
   }
@@ -10500,14 +10496,9 @@ RISCVTargetLowering::lowerFixedLengthVectorStoreToRVV(SDValue Op,
   if (MinVLMAX == MaxVLMAX && MinVLMAX == VT.getVectorNumElements() &&
       getLMUL1VT(ContainerVT).bitsLE(ContainerVT)) {
     MachineMemOperand *MMO = Store->getMemOperand();
-    MachineFunction &MF = DAG.getMachineFunction();
-    MMO = MF.getMachineMemOperand(
-        MMO, MMO->getPointerInfo(),
-        MMO->getMemoryType().isValid()
-            ? LLT::scalable_vector(1, MMO->getMemoryType().getSizeInBits())
-            : MMO->getMemoryType());
     return DAG.getStore(Store->getChain(), DL, NewValue, Store->getBasePtr(),
-                        MMO);
+                        MMO->getPointerInfo(), MMO->getBaseAlign(),
+                        MMO->getFlags(), MMO->getAAInfo());
   }
 
   SDValue VL = getVLOp(VT.getVectorNumElements(), ContainerVT, DL, DAG,
diff --git a/llvm/test/CodeGen/RISCV/rvv/pr88799.ll b/llvm/test/CodeGen/RISCV/rvv/pr88799.ll
new file mode 100644
index 0000000000000..7212a789f9e7e
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/pr88799.ll
@@ -0,0 +1,19 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=riscv64-unknown-linux-gnu -mattr=+v | FileCheck %s
+
+define i32 @main() vscale_range(2,2) {
+; CHECK-LABEL: main:
+; CHECK:       # %bb.0: # %vector.body
+; CHECK-NEXT:    lui a0, 1040368
+; CHECK-NEXT:    addiw a0, a0, -144
+; CHECK-NEXT:    vl2re16.v v8, (a0)
+; CHECK-NEXT:    vs2r.v v8, (zero)
+; CHECK-NEXT:    li a0, 0
+; CHECK-NEXT:    ret
+vector.body:
+  %0 = load <16 x i16>, ptr getelementptr ([3 x [23 x [23 x i16]]], ptr null, i64 -10593, i64 1, i64 22, i64 0), align 16
+  store <16 x i16> %0, ptr null, align 2
+  %wide.load = load <vscale x 8 x i16>, ptr getelementptr ([3 x [23 x [23 x i16]]], ptr null, i64 -10593, i64 1, i64 22, i64 0), align 16
+  store <vscale x 8 x i16> %wide.load, ptr null, align 2
+  ret i32 0
+}

From 6da1966bc503e1ce44ef36e7107c9db482fac6ab Mon Sep 17 00:00:00 2001
From: Michal Terepeta <michalt@google.com>
Date: Tue, 16 Apr 2024 07:55:45 +0200
Subject: [PATCH 526/886] [RISCV] Add scheduling information for SiFive VCIX
 (#86093)

This adds `RISCVScheduleXSf.td` with `SchedWrite` definitions for all
VCIX instructions and uses it in `RISCVSchedSiFive7.td` to set default
latencies for these instructions, helping with issue
https://github.com/llvm/llvm-project/issues/83391. Of course these
default latencies cannot be accurate (since each coprocessor will have
different latencies), but this seems to be enough to avoid some of the
problematic behavior described in the bug.

In any case, this seems to be enough to help with #83391 in our internal
testing.

A subsequent discussion is how to structure the code such that it's
easier for downstream consumers of this to use `SiFive7` scheduling
model with accurate VCIX latencies. But we can probably have a separate
issue to discuss that.
---
 llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td    | 48 +++++++++++----
 llvm/lib/Target/RISCV/RISCVSchedRocket.td     |  1 +
 llvm/lib/Target/RISCV/RISCVSchedSiFive7.td    | 48 +++++++++++++++
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP400.td |  1 +
 llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td |  1 +
 .../Target/RISCV/RISCVSchedSyntacoreSCR1.td   |  1 +
 .../Target/RISCV/RISCVSchedXiangShanNanHu.td  |  1 +
 llvm/lib/Target/RISCV/RISCVSchedule.td        |  1 +
 llvm/lib/Target/RISCV/RISCVScheduleXSf.td     | 59 +++++++++++++++++++
 9 files changed, 149 insertions(+), 12 deletions(-)
 create mode 100644 llvm/lib/Target/RISCV/RISCVScheduleXSf.td

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
index 9a6818c99af20..71aa1f19e089a 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXSf.td
@@ -307,10 +307,16 @@ multiclass VPseudoVC_X<LMULInfo m, DAGOperand RS1Class,
                        Operand OpClass = payload2> {
   let VLMul = m.value in {
     let Defs = [VCIX_STATE], Uses = [VCIX_STATE] in {
-      def "PseudoVC_" # NAME # "_SE_" # m.MX : VPseudoVC_X<OpClass, RS1Class>;
-      def "PseudoVC_V_" # NAME # "_SE_" # m.MX : VPseudoVC_V_X<OpClass, m.vrclass, RS1Class>;
+      def "PseudoVC_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_X<OpClass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_" # NAME # "_" # m.MX)]>;
+      def "PseudoVC_V_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_V_X<OpClass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
     }
-    def "PseudoVC_V_" # NAME # "_" # m.MX : VPseudoVC_V_X<OpClass, m.vrclass, RS1Class>;
+    def "PseudoVC_V_" # NAME # "_" # m.MX
+      : VPseudoVC_V_X<OpClass, m.vrclass, RS1Class>,
+        Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
   }
 }
 
@@ -318,10 +324,16 @@ multiclass VPseudoVC_XV<LMULInfo m, DAGOperand RS1Class,
                         Operand OpClass = payload2> {
   let VLMul = m.value in {
     let Defs = [VCIX_STATE], Uses = [VCIX_STATE] in {
-      def "PseudoVC_" # NAME # "_SE_" # m.MX : VPseudoVC_XV<OpClass, m.vrclass, RS1Class>;
-      def "PseudoVC_V_" # NAME # "_SE_" # m.MX : VPseudoVC_V_XV<OpClass, m.vrclass, m.vrclass, RS1Class>;
+      def "PseudoVC_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_XV<OpClass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_" # NAME # "_" # m.MX)]>;
+      def "PseudoVC_V_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_V_XV<OpClass, m.vrclass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
     }
-    def "PseudoVC_V_" # NAME # "_" # m.MX : VPseudoVC_V_XV<OpClass, m.vrclass, m.vrclass, RS1Class>;
+    def "PseudoVC_V_" # NAME # "_" # m.MX
+      : VPseudoVC_V_XV<OpClass, m.vrclass, m.vrclass, RS1Class>,
+        Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
   }
 }
 
@@ -329,10 +341,16 @@ multiclass VPseudoVC_XVV<LMULInfo m, DAGOperand RS1Class,
                          Operand OpClass = payload2> {
   let VLMul = m.value in {
     let Defs = [VCIX_STATE], Uses = [VCIX_STATE] in {
-      def "PseudoVC_" # NAME # "_SE_" # m.MX : VPseudoVC_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>;
-      def "PseudoVC_V_" # NAME # "_SE_" # m.MX : VPseudoVC_V_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>;
+      def "PseudoVC_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_" # NAME # "_" # m.MX)]>;
+      def "PseudoVC_V_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_V_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
     }
-    def "PseudoVC_V_" # NAME # "_" # m.MX : VPseudoVC_V_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>;
+    def "PseudoVC_V_" # NAME # "_" # m.MX
+      : VPseudoVC_V_XVV<OpClass, m.vrclass, m.vrclass, RS1Class>,
+        Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
   }
 }
 
@@ -340,11 +358,17 @@ multiclass VPseudoVC_XVW<LMULInfo m, DAGOperand RS1Class,
                          Operand OpClass = payload2> {
   let VLMul = m.value in {
     let Defs = [VCIX_STATE], Uses = [VCIX_STATE] in
-    def "PseudoVC_" # NAME # "_SE_" # m.MX : VPseudoVC_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>;
+    def "PseudoVC_" # NAME # "_SE_" # m.MX
+      : VPseudoVC_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>,
+        Sched<[!cast<SchedWrite>("WriteVC_" # NAME # "_" # m.MX)]>;
     let Constraints = "@earlyclobber $rd, $rd = $rs3" in {
       let Defs = [VCIX_STATE], Uses = [VCIX_STATE] in
-      def "PseudoVC_V_" # NAME # "_SE_" # m.MX : VPseudoVC_V_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>;
-      def "PseudoVC_V_" # NAME # "_" # m.MX : VPseudoVC_V_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>;
+      def "PseudoVC_V_" # NAME # "_SE_" # m.MX
+        : VPseudoVC_V_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
+      def "PseudoVC_V_" # NAME # "_" # m.MX
+        : VPseudoVC_V_XVV<OpClass, m.wvrclass, m.vrclass, RS1Class>,
+          Sched<[!cast<SchedWrite>("WriteVC_V_" # NAME # "_" # m.MX)]>;
     }
   }
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedRocket.td b/llvm/lib/Target/RISCV/RISCVSchedRocket.td
index e74c7aab7474d..65494e73758d6 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedRocket.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedRocket.td
@@ -261,4 +261,5 @@ defm : UnsupportedSchedZbkx;
 defm : UnsupportedSchedZfa;
 defm : UnsupportedSchedZfh;
 defm : UnsupportedSchedSFB;
+defm : UnsupportedSchedXsfvcp;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index 4dcec96df9d57..a532066b3a1c8 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -962,6 +962,54 @@ let Latency = 3 in
 
 def : InstRW<[WriteIALU], (instrs COPY)>;
 
+// VCIX
+//
+// In principle we don't know the latency of any VCIX instructions. But instead
+// of taking the default of 1, which can lead to issues [1], we assume that they
+// have a fairly high latency.
+//
+// [1] https://github.com/llvm/llvm-project/issues/83391
+foreach mx = SchedMxList in {
+  defvar Cycles = SiFive7GetCyclesDefault<mx>.c;
+  defvar IsWorstCase = SiFive7IsWorstCaseMX<mx, SchedMxList>.c;
+  let Latency = !mul(Cycles, 10),
+      AcquireAtCycles = [0, 1],
+      ReleaseAtCycles = [1, !add(1, Cycles)] in {
+    defm "" : LMULWriteResMX<"WriteVC_V_I",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_X",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_IV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_VV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_XV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_IVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_IVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_VVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_VVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_XVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_V_XVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    foreach f = ["FPR16", "FPR32", "FPR64"] in {
+      defm "" : LMULWriteResMX<"WriteVC_V_" # f # "V",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+      defm "" : LMULWriteResMX<"WriteVC_V_" # f # "VV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+      defm "" : LMULWriteResMX<"WriteVC_V_" # f # "VW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    }
+    defm "" : LMULWriteResMX<"WriteVC_I",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_X",   [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_IV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_VV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_XV",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_IVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_IVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_VVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_VVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_XVV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    defm "" : LMULWriteResMX<"WriteVC_XVW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    foreach f = ["FPR16", "FPR32", "FPR64"] in {
+      defm "" : LMULWriteResMX<"WriteVC_" # f # "V",  [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+      defm "" : LMULWriteResMX<"WriteVC_" # f # "VV", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+      defm "" : LMULWriteResMX<"WriteVC_" # f # "VW", [SiFive7VCQ, SiFive7VA], mx, IsWorstCase>;
+    }
+  }
+}
+
 //===----------------------------------------------------------------------===//
 
 // Bypass and advance
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP400.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP400.td
index 8ec2e4ff885eb..fccdd7e4f3ec2 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP400.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP400.td
@@ -366,4 +366,5 @@ defm : UnsupportedSchedZbkx;
 defm : UnsupportedSchedSFB;
 defm : UnsupportedSchedZfa;
 defm : UnsupportedSchedV;
+defm : UnsupportedSchedXsfvcp;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 0ced2efa3f7ab..6e4fb19361f55 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -1040,4 +1040,5 @@ defm : UnsupportedSchedZbkb;
 defm : UnsupportedSchedZbkx;
 defm : UnsupportedSchedSFB;
 defm : UnsupportedSchedZfa;
+defm : UnsupportedSchedXsfvcp;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSyntacoreSCR1.td b/llvm/lib/Target/RISCV/RISCVSchedSyntacoreSCR1.td
index 9625d17e0b260..0885e325f24e6 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSyntacoreSCR1.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSyntacoreSCR1.td
@@ -212,4 +212,5 @@ defm : UnsupportedSchedZbkb;
 defm : UnsupportedSchedZbkx;
 defm : UnsupportedSchedZfa;
 defm : UnsupportedSchedZfh;
+defm : UnsupportedSchedXsfvcp;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedXiangShanNanHu.td b/llvm/lib/Target/RISCV/RISCVSchedXiangShanNanHu.td
index 4fc7b0335af53..e0f1fab1d6b40 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedXiangShanNanHu.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedXiangShanNanHu.td
@@ -311,4 +311,5 @@ defm : UnsupportedSchedZfa;
 defm : UnsupportedSchedZfh;
 defm : UnsupportedSchedSFB;
 defm : UnsupportedSchedZabha;
+defm : UnsupportedSchedXsfvcp;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVSchedule.td b/llvm/lib/Target/RISCV/RISCVSchedule.td
index 1d19624342d2b..0086557a41fe7 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedule.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedule.td
@@ -296,3 +296,4 @@ def : ReadAdvance<ReadAtomicHD, 0>;
 // Include the scheduler resources for other instruction extensions.
 include "RISCVScheduleZb.td"
 include "RISCVScheduleV.td"
+include "RISCVScheduleXSf.td"
diff --git a/llvm/lib/Target/RISCV/RISCVScheduleXSf.td b/llvm/lib/Target/RISCV/RISCVScheduleXSf.td
new file mode 100644
index 0000000000000..58d508460f019
--- /dev/null
+++ b/llvm/lib/Target/RISCV/RISCVScheduleXSf.td
@@ -0,0 +1,59 @@
+//===-- RISCVScheduleXSf.td - Scheduling Definitions XSf ---*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file describes the scheduling information for SiFive extensions.
+//
+//===----------------------------------------------------------------------===//
+
+multiclass LMULSchedWritesVCIX<string id>{
+defm "" : LMULSchedWrites<"WriteVC_" # id>;
+defm "" : LMULSchedWrites<"WriteVC_V_" # id>;
+}
+
+defm "" : LMULSchedWritesVCIX<"I">;
+defm "" : LMULSchedWritesVCIX<"X">;
+defm "" : LMULSchedWritesVCIX<"IV">;
+defm "" : LMULSchedWritesVCIX<"VV">;
+defm "" : LMULSchedWritesVCIX<"XV">;
+defm "" : LMULSchedWritesVCIX<"IVV">;
+defm "" : LMULSchedWritesVCIX<"IVW">;
+defm "" : LMULSchedWritesVCIX<"VVV">;
+defm "" : LMULSchedWritesVCIX<"VVW">;
+defm "" : LMULSchedWritesVCIX<"XVV">;
+defm "" : LMULSchedWritesVCIX<"XVW">;
+foreach f = ["FPR16", "FPR32", "FPR64"] in {
+  defm "" : LMULSchedWritesVCIX<f # "V">;
+  defm "" : LMULSchedWritesVCIX<f # "VV">;
+  defm "" : LMULSchedWritesVCIX<f # "VW">;
+}
+
+multiclass LMULWriteResVCIX<string id, list<ProcResourceKind> resources>{
+defm : LMULWriteRes<"WriteVC_" # id, resources>;
+defm : LMULWriteRes<"WriteVC_V_" # id, resources>;
+}
+
+multiclass UnsupportedSchedXsfvcp {
+let Unsupported = true in {
+defm : LMULWriteResVCIX<"I", []>;
+defm : LMULWriteResVCIX<"X", []>;
+defm : LMULWriteResVCIX<"IV", []>;
+defm : LMULWriteResVCIX<"VV", []>;
+defm : LMULWriteResVCIX<"XV", []>;
+defm : LMULWriteResVCIX<"IVV", []>;
+defm : LMULWriteResVCIX<"IVW", []>;
+defm : LMULWriteResVCIX<"VVV", []>;
+defm : LMULWriteResVCIX<"VVW", []>;
+defm : LMULWriteResVCIX<"XVV", []>;
+defm : LMULWriteResVCIX<"XVW", []>;
+foreach f = ["FPR16", "FPR32", "FPR64"] in {
+  defm : LMULWriteResVCIX<f # "V", []>;
+  defm : LMULWriteResVCIX<f # "VV", []>;
+  defm : LMULWriteResVCIX<f # "VW", []>;
+}
+}
+}

From e2c91091e537a54f1469610ab407e7c9561e7ffc Mon Sep 17 00:00:00 2001
From: darkbuck <michael.hliao@gmail.com>
Date: Tue, 16 Apr 2024 02:36:12 -0400
Subject: [PATCH 527/886] [GlobalISel] Handle more commutable instructions in
 `commute_constant_to_rhs`

Reviewers:
rupprecht, aartbik, cyndyishida, Pierre-vh, aemerson, ftynse, hanhanW, banach-space, jayfoad, nicolasvasilache, daniel-grumberg, arsenm, PeimingLiu, JDevlieghere, matthias-springer

Reviewed By: arsenm, Pierre-vh

Pull Request: https://github.com/llvm/llvm-project/pull/87424
---
 .../include/llvm/Target/GlobalISel/Combine.td |  10 +-
 .../lib/CodeGen/GlobalISel/CombinerHelper.cpp |  38 +-
 .../combine-commute-fp-const-lhs.mir          | 126 +++++
 .../combine-commute-int-const-lhs.mir         | 456 ++++++++++++++++++
 .../combine-const-fold-barrier-rhs.mir        | 160 ++++++
 ...relegalizer-combiner-select-to-fminmax.mir |  16 +-
 .../AArch64/GlobalISel/select-to-fmin-fmax.ll |  16 +-
 .../GlobalISel/combine-fcanonicalize.mir      |   6 +-
 .../GlobalISel/postlegalizercombiner-and.mir  |   2 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/smed3.ll  |   4 +-
 llvm/test/CodeGen/AMDGPU/GlobalISel/umed3.ll  |   4 +-
 .../CodeGen/AMDGPU/GlobalISel/urem.i64.ll     |  18 +-
 12 files changed, 815 insertions(+), 41 deletions(-)
 create mode 100644 llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-int-const-lhs.mir

diff --git a/llvm/include/llvm/Target/GlobalISel/Combine.td b/llvm/include/llvm/Target/GlobalISel/Combine.td
index 778ff7e437eb5..8568a7ae90e56 100644
--- a/llvm/include/llvm/Target/GlobalISel/Combine.td
+++ b/llvm/include/llvm/Target/GlobalISel/Combine.td
@@ -443,14 +443,20 @@ def select_constant_cmp: GICombineRule<
 // TODO: handle compares (currently not marked as isCommutable)
 def commute_int_constant_to_rhs : GICombineRule<
   (defs root:$root),
-  (match (wip_match_opcode G_ADD, G_MUL, G_AND, G_OR, G_XOR):$root,
+  (match (wip_match_opcode G_ADD, G_MUL, G_AND, G_OR, G_XOR,
+                           G_SMIN, G_SMAX, G_UMIN, G_UMAX, G_UADDO, G_SADDO,
+                           G_UMULO, G_SMULO, G_UMULH, G_SMULH,
+                           G_UADDSAT, G_SADDSAT, G_SMULFIX, G_UMULFIX,
+                           G_SMULFIXSAT, G_UMULFIXSAT):$root,
     [{ return Helper.matchCommuteConstantToRHS(*${root}); }]),
   (apply [{ Helper.applyCommuteBinOpOperands(*${root}); }])
 >;
 
 def commute_fp_constant_to_rhs : GICombineRule<
   (defs root:$root),
-  (match (wip_match_opcode G_FADD, G_FMUL):$root,
+  (match (wip_match_opcode G_FADD, G_FMUL, G_FMINNUM, G_FMAXNUM,
+                           G_FMINNUM_IEEE, G_FMAXNUM_IEEE,
+                           G_FMINIMUM, G_FMAXIMUM):$root,
     [{ return Helper.matchCommuteFPConstantToRHS(*${root}); }]),
   (apply [{ Helper.applyCommuteBinOpOperands(*${root}); }])
 >;
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
index 40c5119ee7fb3..3829c33369b27 100644
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -6273,8 +6273,21 @@ bool CombinerHelper::matchShiftsTooBig(MachineInstr &MI) {
 }
 
 bool CombinerHelper::matchCommuteConstantToRHS(MachineInstr &MI) {
-  Register LHS = MI.getOperand(1).getReg();
-  Register RHS = MI.getOperand(2).getReg();
+  unsigned LHSOpndIdx = 1;
+  unsigned RHSOpndIdx = 2;
+  switch (MI.getOpcode()) {
+  case TargetOpcode::G_UADDO:
+  case TargetOpcode::G_SADDO:
+  case TargetOpcode::G_UMULO:
+  case TargetOpcode::G_SMULO:
+    LHSOpndIdx = 2;
+    RHSOpndIdx = 3;
+    break;
+  default:
+    break;
+  }
+  Register LHS = MI.getOperand(LHSOpndIdx).getReg();
+  Register RHS = MI.getOperand(RHSOpndIdx).getReg();
   if (!getIConstantVRegVal(LHS, MRI)) {
     // Skip commuting if LHS is not a constant. But, LHS may be a
     // G_CONSTANT_FOLD_BARRIER. If so we commute as long as we don't already
@@ -6300,10 +6313,23 @@ bool CombinerHelper::matchCommuteFPConstantToRHS(MachineInstr &MI) {
 
 void CombinerHelper::applyCommuteBinOpOperands(MachineInstr &MI) {
   Observer.changingInstr(MI);
-  Register LHSReg = MI.getOperand(1).getReg();
-  Register RHSReg = MI.getOperand(2).getReg();
-  MI.getOperand(1).setReg(RHSReg);
-  MI.getOperand(2).setReg(LHSReg);
+  unsigned LHSOpndIdx = 1;
+  unsigned RHSOpndIdx = 2;
+  switch (MI.getOpcode()) {
+  case TargetOpcode::G_UADDO:
+  case TargetOpcode::G_SADDO:
+  case TargetOpcode::G_UMULO:
+  case TargetOpcode::G_SMULO:
+    LHSOpndIdx = 2;
+    RHSOpndIdx = 3;
+    break;
+  default:
+    break;
+  }
+  Register LHSReg = MI.getOperand(LHSOpndIdx).getReg();
+  Register RHSReg = MI.getOperand(RHSOpndIdx).getReg();
+  MI.getOperand(LHSOpndIdx).setReg(RHSReg);
+  MI.getOperand(RHSOpndIdx).setReg(LHSReg);
   Observer.changedInstr(MI);
 }
 
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-fp-const-lhs.mir b/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-fp-const-lhs.mir
index 76d82884a7b1f..d791660b7a5eb 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-fp-const-lhs.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-fp-const-lhs.mir
@@ -116,3 +116,129 @@ body:             |
     $q0 = COPY %mul
     RET_ReallyLR
 ...
+---
+name:            fminnum
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fminnum
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %min:_(s32) = G_FMINNUM [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %min(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %min:_(s32) = G_FMINNUM %cst, %0
+    $s0 = COPY %min
+    RET_ReallyLR
+...
+---
+name:            fmaxnum
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fmaxnum
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %max:_(s32) = G_FMAXNUM [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %max(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %max:_(s32) = G_FMAXNUM %cst, %0
+    $s0 = COPY %max
+    RET_ReallyLR
+...
+---
+name:            fminnum_ieee
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fminnum_ieee
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %min:_(s32) = G_FMINNUM_IEEE [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %min(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %min:_(s32) = G_FMINNUM_IEEE %cst, %0
+    $s0 = COPY %min
+    RET_ReallyLR
+...
+---
+name:            fmaxnum_ieee
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fmaxnum_ieee
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %max:_(s32) = G_FMAXNUM_IEEE [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %max(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %max:_(s32) = G_FMAXNUM_IEEE %cst, %0
+    $s0 = COPY %max
+    RET_ReallyLR
+...
+---
+name:            fminimum
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fminimum
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %min:_(s32) = G_FMINIMUM [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %min(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %min:_(s32) = G_FMINIMUM %cst, %0
+    $s0 = COPY %min
+    RET_ReallyLR
+...
+---
+name:            fmaximum
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: fmaximum
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    ; CHECK-NEXT: %max:_(s32) = G_FMAXIMUM [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %max(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_FCONSTANT float 2.000000e+00
+    %max:_(s32) = G_FMAXIMUM %cst, %0
+    $s0 = COPY %max
+    RET_ReallyLR
+...
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-int-const-lhs.mir b/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-int-const-lhs.mir
new file mode 100644
index 0000000000000..16365494f5f4e
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/combine-commute-int-const-lhs.mir
@@ -0,0 +1,456 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 2
+# RUN: llc -mtriple aarch64 -run-pass=aarch64-prelegalizer-combiner -verify-machineinstrs %s -o - | FileCheck %s
+---
+name:            add
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: add
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %add:_(s32) = G_ADD [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %add(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %add:_(s32) = G_ADD %cst, %0
+    $s0 = COPY %add
+    RET_ReallyLR
+
+...
+---
+name:            mul
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: mul
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_MUL [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_MUL %cst, %0
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            and
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: and
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 5
+    ; CHECK-NEXT: %and:_(s32) = G_AND [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %and(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 5
+    %and:_(s32) = G_AND %cst, %0
+    $s0 = COPY %and
+    RET_ReallyLR
+...
+---
+name:            or
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: or
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 5
+    ; CHECK-NEXT: %or:_(s32) = G_OR [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %or(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 5
+    %or:_(s32) = G_OR %cst, %0
+    $s0 = COPY %or
+    RET_ReallyLR
+...
+---
+name:            xor
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: xor
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 5
+    ; CHECK-NEXT: %xor:_(s32) = G_XOR [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %xor(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 5
+    %xor:_(s32) = G_XOR %cst, %0
+    $s0 = COPY %xor
+    RET_ReallyLR
+...
+---
+name:            smin
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smin
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 10
+    ; CHECK-NEXT: %min:_(s32) = G_SMIN [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %min(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 10
+    %min:_(s32) = G_SMIN %cst, %0
+    $s0 = COPY %min
+    RET_ReallyLR
+...
+---
+name:            smax
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smax
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 10
+    ; CHECK-NEXT: %max:_(s32) = G_SMAX [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %max(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 10
+    %max:_(s32) = G_SMAX %cst, %0
+    $s0 = COPY %max
+    RET_ReallyLR
+...
+---
+name:            umin
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umin
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 10
+    ; CHECK-NEXT: %min:_(s32) = G_UMIN [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %min(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 10
+    %min:_(s32) = G_UMIN %cst, %0
+    $s0 = COPY %min
+    RET_ReallyLR
+...
+---
+name:            umax
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umax
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 10
+    ; CHECK-NEXT: %max:_(s32) = G_UMAX [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %max(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 10
+    %max:_(s32) = G_UMAX %cst, %0
+    $s0 = COPY %max
+    RET_ReallyLR
+...
+---
+name:            uaddo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: uaddo
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %add:_(s32), %overflow:_(s1) = G_UADDO [[COPY]], %cst
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $s0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %add:_(s32), %overflow:_(s1) = G_UADDO %cst, %0
+    %ret:_(s32) = G_ANYEXT %overflow
+    $s0 = COPY %ret
+    RET_ReallyLR
+
+...
+---
+name:            saddo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: saddo
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %add:_(s32), %overflow:_(s1) = G_SADDO [[COPY]], %cst
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $s0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %add:_(s32), %overflow:_(s1) = G_SADDO %cst, %0
+    %ret:_(s32) = G_ANYEXT %overflow
+    $s0 = COPY %ret
+    RET_ReallyLR
+
+...
+---
+name:            umulo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umulo
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32), %overflow:_(s1) = G_UMULO [[COPY]], %cst
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $s0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32), %overflow:_(s1) = G_UMULO %cst, %0
+    %ret:_(s32) = G_ANYEXT %overflow
+    $s0 = COPY %ret
+    RET_ReallyLR
+...
+---
+name:            smulo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smulo
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32), %overflow:_(s1) = G_SMULO [[COPY]], %cst
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $s0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32), %overflow:_(s1) = G_SMULO %cst, %0
+    %ret:_(s32) = G_ANYEXT %overflow
+    $s0 = COPY %ret
+    RET_ReallyLR
+...
+---
+name:            umulh
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umulh
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_UMULH [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_UMULH %cst, %0
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            smulh
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smulh
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_UMULH [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_UMULH %cst, %0
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            uaddsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: uaddsat
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %add:_(s32) = G_UADDSAT [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %add(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %add:_(s32) = G_UADDSAT %cst, %0
+    $s0 = COPY %add
+    RET_ReallyLR
+
+...
+---
+name:            saddsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: saddsat
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %add:_(s32) = G_SADDSAT [[COPY]], %cst
+    ; CHECK-NEXT: $s0 = COPY %add(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %add:_(s32) = G_SADDSAT %cst, %0
+    $s0 = COPY %add
+    RET_ReallyLR
+
+...
+---
+name:            smulfix
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smulfix
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_SMULFIX [[COPY]], %cst, 7
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_SMULFIX %cst, %0, 7
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            umulfix
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umulfix
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_UMULFIX [[COPY]], %cst, 7
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_UMULFIX %cst, %0, 7
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            smulfixsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: smulfixsat
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_SMULFIXSAT [[COPY]], %cst, 7
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_SMULFIXSAT %cst, %0, 7
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
+---
+name:            umulfixsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: umulfixsat
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 3
+    ; CHECK-NEXT: %mul:_(s32) = G_UMULFIXSAT [[COPY]], %cst, 7
+    ; CHECK-NEXT: $s0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $s0
+    %cst:_(s32) = G_CONSTANT i32 3
+    %mul:_(s32) = G_UMULFIXSAT %cst, %0, 7
+    $s0 = COPY %mul
+    RET_ReallyLR
+...
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/combine-const-fold-barrier-rhs.mir b/llvm/test/CodeGen/AArch64/GlobalISel/combine-const-fold-barrier-rhs.mir
index 01e0dce5a661c..c967e4f2ea5e8 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/combine-const-fold-barrier-rhs.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/combine-const-fold-barrier-rhs.mir
@@ -78,3 +78,163 @@ body:             |
     RET_ReallyLR
 
 ...
+---
+name:            cfb_lhs_smulo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_smulo
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $w0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %mul:_(s32), %overflow:_(s1) = G_SMULO [[COPY]], %cfb
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $w0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %mul:_(s32), %overflow:_(s1) = G_SMULO %cfb, %0
+    %ret:_(s32) = G_ANYEXT %overflow
+    $w0 = COPY %ret
+    RET_ReallyLR
+
+...
+---
+name:            cfb_lhs_cfb_already_rhs_smulo
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_cfb_already_rhs_smulo
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %cst2:_(s32) = G_CONSTANT i32 6
+    ; CHECK-NEXT: %cfb2:_(s32) = G_CONSTANT_FOLD_BARRIER %cst2
+    ; CHECK-NEXT: %mul:_(s32), %overflow:_(s1) = G_SMULO %cfb, %cfb2
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $w0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %cst2:_(s32) = G_CONSTANT i32 6
+    %cfb2:_(s32) = G_CONSTANT_FOLD_BARRIER %cst2
+    %mul:_(s32), %overflow:_(s1) = G_SMULO %cfb, %cfb2
+    %ret:_(s32) = G_ANYEXT %overflow
+    $w0 = COPY %ret
+    RET_ReallyLR
+
+...
+---
+name:            cfb_lhs_cst_on_rhs_smulo
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_cst_on_rhs_smulo
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %cst2:_(s32) = G_CONSTANT i32 6
+    ; CHECK-NEXT: %mul:_(s32), %overflow:_(s1) = G_SMULO %cfb, %cst2
+    ; CHECK-NEXT: %ret:_(s32) = G_ANYEXT %overflow(s1)
+    ; CHECK-NEXT: $w0 = COPY %ret(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %cst2:_(s32) = G_CONSTANT i32 6
+    %mul:_(s32), %overflow:_(s1) = G_SMULO %cfb, %cst2
+    %ret:_(s32) = G_ANYEXT %overflow
+    $w0 = COPY %ret
+    RET_ReallyLR
+
+...
+---
+name:            cfb_lhs_umulfixsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_umulfixsat
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $w0
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %mul:_(s32) = G_UMULFIXSAT [[COPY]], %cfb, 7
+    ; CHECK-NEXT: $w0 = COPY %mul(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %mul:_(s32) = G_UMULFIXSAT %cfb, %0, 7
+    $w0 = COPY %mul
+    RET_ReallyLR
+
+...
+---
+name:            cfb_lhs_cfb_already_rhs_umulfixsat
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_cfb_already_rhs_umulfixsat
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %cst2:_(s32) = G_CONSTANT i32 2
+    ; CHECK-NEXT: %cfb2:_(s32) = G_CONSTANT_FOLD_BARRIER %cst2
+    ; CHECK-NEXT: %add:_(s32) = G_UMULFIXSAT %cfb, %cfb2, 7
+    ; CHECK-NEXT: $w0 = COPY %add(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %cst2:_(s32) = G_CONSTANT i32 2
+    %cfb2:_(s32) = G_CONSTANT_FOLD_BARRIER %cst2
+    %add:_(s32) = G_UMULFIXSAT %cfb, %cfb2, 7
+    $w0 = COPY %add
+    RET_ReallyLR
+
+...
+---
+name:            cfb_lhs_cst_on_rhs_umulfixsat
+alignment:       4
+tracksRegLiveness: true
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: cfb_lhs_cst_on_rhs_umulfixsat
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: %cst:_(s32) = G_CONSTANT i32 1
+    ; CHECK-NEXT: %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    ; CHECK-NEXT: %cst2:_(s32) = G_CONSTANT i32 2
+    ; CHECK-NEXT: %add:_(s32) = G_UMULFIXSAT %cfb, %cst2, 7
+    ; CHECK-NEXT: $w0 = COPY %add(s32)
+    ; CHECK-NEXT: RET_ReallyLR
+    %0:_(s32) = COPY $w0
+    %cst:_(s32) = G_CONSTANT i32 1
+    %cfb:_(s32) = G_CONSTANT_FOLD_BARRIER %cst
+    %cst2:_(s32) = G_CONSTANT i32 2
+    %add:_(s32) = G_UMULFIXSAT %cfb, %cst2, 7
+    $w0 = COPY %add
+    RET_ReallyLR
+
+...
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-select-to-fminmax.mir b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-select-to-fminmax.mir
index 8c4300d9e7329..03e507f5eaa7f 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-select-to-fminmax.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/prelegalizer-combiner-select-to-fminmax.mir
@@ -11,7 +11,7 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s16) = COPY $h0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH0000
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s16) = G_FMAXIMUM [[C]], [[COPY]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s16) = G_FMAXIMUM [[COPY]], [[C]]
     ; CHECK-NEXT: $h0 = COPY [[FMAXIMUM]](s16)
     ; CHECK-NEXT: RET_ReallyLR implicit $h0
     %0:_(s16) = COPY $h0
@@ -33,7 +33,7 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s32) = G_FMAXIMUM [[C]], [[COPY]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s32) = G_FMAXIMUM [[COPY]], [[C]]
     ; CHECK-NEXT: $s0 = COPY [[FMAXIMUM]](s32)
     ; CHECK-NEXT: RET_ReallyLR implicit $s0
     %0:_(s32) = COPY $s0
@@ -55,7 +55,7 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $d0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 0.000000e+00
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s64) = G_FMAXIMUM [[C]], [[COPY]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(s64) = G_FMAXIMUM [[COPY]], [[C]]
     ; CHECK-NEXT: $d0 = COPY [[FMAXIMUM]](s64)
     ; CHECK-NEXT: RET_ReallyLR implicit $d0
     %0:_(s64) = COPY $d0
@@ -77,7 +77,7 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $d0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 0.000000e+00
-    ; CHECK-NEXT: [[FMINIMUM:%[0-9]+]]:_(s64) = G_FMINIMUM [[C]], [[COPY]]
+    ; CHECK-NEXT: [[FMINIMUM:%[0-9]+]]:_(s64) = G_FMINIMUM [[COPY]], [[C]]
     ; CHECK-NEXT: $d0 = COPY [[FMINIMUM]](s64)
     ; CHECK-NEXT: RET_ReallyLR implicit $d0
     %0:_(s64) = COPY $d0
@@ -100,7 +100,7 @@ body:             |
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH0000
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s16>) = G_BUILD_VECTOR [[C]](s16), [[C]](s16), [[C]](s16), [[C]](s16), [[C]](s16), [[C]](s16), [[C]](s16), [[C]](s16)
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<8 x s16>) = G_FMAXIMUM [[BUILD_VECTOR]], [[COPY]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<8 x s16>) = G_FMAXIMUM [[COPY]], [[BUILD_VECTOR]]
     ; CHECK-NEXT: $q0 = COPY [[FMAXIMUM]](<8 x s16>)
     ; CHECK-NEXT: RET_ReallyLR implicit $q0
     %0:_(<8 x s16>) = COPY $q0
@@ -125,7 +125,7 @@ body:             |
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY]](<2 x s64>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<4 x s32>) = G_FMAXIMUM [[BUILD_VECTOR]], [[BITCAST]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<4 x s32>) = G_FMAXIMUM [[BITCAST]], [[BUILD_VECTOR]]
     ; CHECK-NEXT: $q0 = COPY [[FMAXIMUM]](<4 x s32>)
     ; CHECK-NEXT: RET_ReallyLR implicit $q0
     %1:_(<2 x s64>) = COPY $q0
@@ -150,7 +150,7 @@ body:             |
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 0.000000e+00
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C]](s64)
-    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<2 x s64>) = G_FMAXIMUM [[BUILD_VECTOR]], [[COPY]]
+    ; CHECK-NEXT: [[FMAXIMUM:%[0-9]+]]:_(<2 x s64>) = G_FMAXIMUM [[COPY]], [[BUILD_VECTOR]]
     ; CHECK-NEXT: $q0 = COPY [[FMAXIMUM]](<2 x s64>)
     ; CHECK-NEXT: RET_ReallyLR implicit $q0
     %0:_(<2 x s64>) = COPY $q0
@@ -174,7 +174,7 @@ body:             |
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 0.000000e+00
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C]](s64)
-    ; CHECK-NEXT: [[FMINIMUM:%[0-9]+]]:_(<2 x s64>) = G_FMINIMUM [[BUILD_VECTOR]], [[COPY]]
+    ; CHECK-NEXT: [[FMINIMUM:%[0-9]+]]:_(<2 x s64>) = G_FMINIMUM [[COPY]], [[BUILD_VECTOR]]
     ; CHECK-NEXT: $q0 = COPY [[FMINIMUM]](<2 x s64>)
     ; CHECK-NEXT: RET_ReallyLR implicit $q0
     %0:_(<2 x s64>) = COPY $q0
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/select-to-fmin-fmax.ll b/llvm/test/CodeGen/AArch64/GlobalISel/select-to-fmin-fmax.ll
index 7badf4732fd0d..ae0a9b1c7c4f1 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/select-to-fmin-fmax.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/select-to-fmin-fmax.ll
@@ -4,7 +4,7 @@ define half @test_s16(half %a) #0 {
 ; CHECK-LABEL: test_s16:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi d1, #0000000000000000
-; CHECK-NEXT:    fmax h0, h1, h0
+; CHECK-NEXT:    fmax h0, h0, h1
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt half %a, 0.0
@@ -16,7 +16,7 @@ define float @test_s32(float %a) #0 {
 ; CHECK-LABEL: test_s32:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi d1, #0000000000000000
-; CHECK-NEXT:    fmax s0, s1, s0
+; CHECK-NEXT:    fmax s0, s0, s1
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt float %a, 0.0
@@ -28,7 +28,7 @@ define double @test_s64(double %a) #0 {
 ; CHECK-LABEL: test_s64:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi d1, #0000000000000000
-; CHECK-NEXT:    fmax d0, d1, d0
+; CHECK-NEXT:    fmax d0, d0, d1
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt double %a, 0.0
@@ -40,7 +40,7 @@ define <4 x half> @test_v4s16(<4 x half> %a) #0 {
 ; CHECK-LABEL: test_v4s16:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    fmax v0.4h, v1.4h, v0.4h
+; CHECK-NEXT:    fmax v0.4h, v0.4h, v1.4h
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt <4 x half> %a, zeroinitializer
@@ -52,7 +52,7 @@ define <8 x half> @test_v8s16(<8 x half> %a) #0 {
 ; CHECK-LABEL: test_v8s16:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    fmax v0.8h, v1.8h, v0.8h
+; CHECK-NEXT:    fmax v0.8h, v0.8h, v1.8h
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt <8 x half> %a, zeroinitializer
@@ -64,7 +64,7 @@ define <2 x float> @test_v2s32(<2 x float> %a) #0 {
 ; CHECK-LABEL: test_v2s32:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    fmax v0.2s, v1.2s, v0.2s
+; CHECK-NEXT:    fmax v0.2s, v0.2s, v1.2s
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt <2 x float> %a, zeroinitializer
@@ -76,7 +76,7 @@ define <4 x float> @test_v4s32(<4 x float> %a) #0 {
 ; CHECK-LABEL: test_v4s32:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    fmax v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    fmax v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt <4 x float> %a, zeroinitializer
@@ -88,7 +88,7 @@ define <2 x double> @test_v2s64(<2 x double> %a) #0 {
 ; CHECK-LABEL: test_v2s64:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    fmax v0.2d, v1.2d, v0.2d
+; CHECK-NEXT:    fmax v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    ret
 entry:
   %fcmp = fcmp olt <2 x double> %a, zeroinitializer
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/combine-fcanonicalize.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/combine-fcanonicalize.mir
index ee0e83c5e0763..020761352148f 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/combine-fcanonicalize.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/combine-fcanonicalize.mir
@@ -254,8 +254,8 @@ body: |
     ; CHECK-NEXT: %one_s32:_(s32) = G_ANYEXT %one(s16)
     ; CHECK-NEXT: %one_undef:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC %one_s32(s32), %undef(s32)
     ; CHECK-NEXT: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[COPY]], %two_splat
-    ; CHECK-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE %zero_undef, [[FMUL]]
-    ; CHECK-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE %one_undef, [[FMAXNUM_IEEE]]
+    ; CHECK-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FMUL]], %zero_undef
+    ; CHECK-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FMAXNUM_IEEE]], %one_undef
     ; CHECK-NEXT: $vgpr0 = COPY [[FMINNUM_IEEE]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %two:_(s16) = G_FCONSTANT half 0xH4000
@@ -306,7 +306,7 @@ body: |
     ; CHECK-NEXT: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[COPY]], %two_splat
     ; CHECK-NEXT: %snan_undef_fcan:_(<2 x s16>) = G_FCANONICALIZE %snan_undef
     ; CHECK-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE %snan_undef_fcan, [[FMUL]]
-    ; CHECK-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE %qnan_undef, [[FMAXNUM_IEEE]]
+    ; CHECK-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FMAXNUM_IEEE]], %qnan_undef
     ; CHECK-NEXT: $vgpr0 = COPY [[FMINNUM_IEEE]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %two:_(s16) = G_FCONSTANT half 0xH4000
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/postlegalizercombiner-and.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/postlegalizercombiner-and.mir
index d6321dae3aa7e..67e6de1ce7644 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/postlegalizercombiner-and.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/postlegalizercombiner-and.mir
@@ -318,7 +318,7 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: %val:_(s32) = COPY $vgpr4
     ; CHECK-NEXT: %k255:_(s32) = G_CONSTANT i32 255
-    ; CHECK-NEXT: %umin0:_(s32) = G_UMIN %k255, %val
+    ; CHECK-NEXT: %umin0:_(s32) = G_UMIN %val, %k255
     ; CHECK-NEXT: $vgpr0 = COPY %umin0(s32)
     %ptr0:_(p1) = COPY $vgpr0_vgpr1
     %ptr1:_(p1) = COPY $vgpr2_vgpr3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/smed3.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/smed3.ll
index dc13dee4f148a..1d94d76da148f 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/smed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/smed3.ll
@@ -145,10 +145,10 @@ define <2 x i16> @test_max_K0min_K1Val__v2i16(<2 x i16> %a) {
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 17
 ; GFX8-NEXT:    v_min_i16_e32 v1, 17, v0
-; GFX8-NEXT:    v_min_i16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT:    v_min_i16_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX8-NEXT:    v_mov_b32_e32 v2, -12
 ; GFX8-NEXT:    v_max_i16_e32 v1, -12, v1
-; GFX8-NEXT:    v_max_i16_sdwa v0, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT:    v_max_i16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/umed3.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/umed3.ll
index 7e38762e7b559..a8233054db9bc 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/umed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/umed3.ll
@@ -145,10 +145,10 @@ define <2 x i16> @test_max_K0min_K1Val__v2u16(<2 x i16> %a) {
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 17
 ; GFX8-NEXT:    v_min_u16_e32 v1, 17, v0
-; GFX8-NEXT:    v_min_u16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT:    v_min_u16_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 12
 ; GFX8-NEXT:    v_max_u16_e32 v1, 12, v1
-; GFX8-NEXT:    v_max_u16_sdwa v0, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT:    v_max_u16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
 ; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll
index 07480a0ce0c2e..cc0f7e2ca5a54 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll
@@ -983,7 +983,7 @@ define i64 @v_urem_i64_oddk_denom(i64 %num) {
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v3, v3
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v7, v3, v5
-; CHECK-NEXT:    v_mul_hi_u32 v8, v5, v3
+; CHECK-NEXT:    v_mul_hi_u32 v8, v3, v5
 ; CHECK-NEXT:    v_sub_i32_e32 v6, vcc, v6, v3
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v4, v7
@@ -1010,7 +1010,7 @@ define i64 @v_urem_i64_oddk_denom(i64 %num) {
 ; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v6, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v3, v5
-; CHECK-NEXT:    v_mul_hi_u32 v7, v5, v3
+; CHECK-NEXT:    v_mul_hi_u32 v7, v3, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v5, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v4, v6
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v3, v6
@@ -1058,7 +1058,7 @@ define i64 @v_urem_i64_oddk_denom(i64 %num) {
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v3, v2
-; CHECK-NEXT:    v_mul_hi_u32 v3, v2, v3
+; CHECK-NEXT:    v_mul_hi_u32 v3, v3, v2
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v4, v4, v2
 ; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
@@ -1265,10 +1265,10 @@ define <2 x i64> @v_urem_v2i64_oddk_denom(<2 x i64> %num) {
 ; GISEL-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; GISEL-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; GISEL-NEXT:    v_mul_lo_u32 v11, v9, v4
-; GISEL-NEXT:    v_mul_hi_u32 v9, v4, v9
+; GISEL-NEXT:    v_mul_hi_u32 v9, v9, v4
 ; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v12, v8
 ; GISEL-NEXT:    v_mul_lo_u32 v12, v5, v4
-; GISEL-NEXT:    v_mul_hi_u32 v5, v4, v5
+; GISEL-NEXT:    v_mul_hi_u32 v5, v5, v4
 ; GISEL-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
 ; GISEL-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; GISEL-NEXT:    v_mul_lo_u32 v7, v7, v4
@@ -1339,7 +1339,7 @@ define <2 x i64> @v_urem_v2i64_oddk_denom(<2 x i64> %num) {
 ; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
 ; CGP-NEXT:    v_mul_lo_u32 v8, v6, v7
 ; CGP-NEXT:    v_mul_lo_u32 v9, v5, v7
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v5
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
 ; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v8, v5
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
 ; CGP-NEXT:    v_mul_lo_u32 v10, v6, v9
@@ -1366,7 +1366,7 @@ define <2 x i64> @v_urem_v2i64_oddk_denom(<2 x i64> %num) {
 ; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
 ; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v8, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v8, v5, v7
-; CGP-NEXT:    v_mul_hi_u32 v9, v7, v5
+; CGP-NEXT:    v_mul_hi_u32 v9, v5, v7
 ; CGP-NEXT:    v_mul_lo_u32 v7, v6, v7
 ; CGP-NEXT:    v_mul_lo_u32 v10, v6, v8
 ; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
@@ -1433,10 +1433,10 @@ define <2 x i64> @v_urem_v2i64_oddk_denom(<2 x i64> %num) {
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
 ; CGP-NEXT:    v_mul_lo_u32 v9, v7, v4
-; CGP-NEXT:    v_mul_hi_u32 v7, v4, v7
+; CGP-NEXT:    v_mul_hi_u32 v7, v7, v4
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_mul_lo_u32 v11, v5, v4
-; CGP-NEXT:    v_mul_hi_u32 v5, v4, v5
+; CGP-NEXT:    v_mul_hi_u32 v5, v5, v4
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v15, v8
 ; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_mul_lo_u32 v8, v8, v4

From 3c6f91e5b671321c95259dabecdbdfe4a6d69ce1 Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Tue, 16 Apr 2024 08:49:45 +0200
Subject: [PATCH 528/886] [clang][dataflow] Fix result object location for
 builtin `<=>`. (#88726)

The newly added test causes an assertion failure in
`PropagateResultObject()`
without the fix added here.
---
 .../FlowSensitive/DataflowEnvironment.cpp     |  5 ++
 .../Analysis/FlowSensitive/TransferTest.cpp   | 52 +++++++++++++++++++
 2 files changed, 57 insertions(+)

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index bea15ce9bd24d..ee2581143e114 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -508,6 +508,11 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
         isa<CXXStdInitializerListExpr>(E)) {
       return;
     }
+    if (auto *Op = dyn_cast<BinaryOperator>(E);
+        Op && Op->getOpcode() == BO_Cmp) {
+      // Builtin `<=>` returns a `std::strong_ordering` object.
+      return;
+    }
 
     if (auto *InitList = dyn_cast<InitListExpr>(E)) {
       if (!InitList->isSemanticForm())
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index 00dafb2988c69..d8bcc3da4b8b1 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -3098,6 +3098,58 @@ TEST(TransferTest, ResultObjectLocationForCXXOperatorCallExpr) {
       });
 }
 
+// Check that the `std::strong_ordering` object returned by builtin `<=>` has a
+// correctly modeled result object location.
+TEST(TransferTest, ResultObjectLocationForBuiltinSpaceshipOperator) {
+  std::string Code = R"(
+    namespace std {
+      // This is the minimal definition required to get
+      // `Sema::CheckComparisonCategoryType()` to accept this fake.
+      struct strong_ordering {
+        enum class ordering { less, equal, greater };
+        ordering o;
+        static const strong_ordering less;
+        static const strong_ordering equivalent;
+        static const strong_ordering equal;
+        static const strong_ordering greater;
+      };
+
+      inline constexpr strong_ordering strong_ordering::less =
+        { strong_ordering::ordering::less };
+      inline constexpr strong_ordering strong_ordering::equal =
+        { strong_ordering::ordering::equal };
+      inline constexpr strong_ordering strong_ordering::equivalent =
+        { strong_ordering::ordering::equal };
+      inline constexpr strong_ordering strong_ordering::greater =
+        { strong_ordering::ordering::greater };
+    }
+    void target(int i, int j) {
+      auto ordering = i <=> j;
+      // [[p]]
+    }
+  )";
+  using ast_matchers::binaryOperator;
+  using ast_matchers::hasOperatorName;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  using ast_matchers::traverse;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *Spaceship = selectFirst<BinaryOperator>(
+            "op",
+            match(binaryOperator(hasOperatorName("<=>")).bind("op"), ASTCtx));
+
+        EXPECT_EQ(
+            &Env.getResultObjectLocation(*Spaceship),
+            &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "ordering"));
+      },
+      LangStandard::lang_cxx20);
+}
+
 TEST(TransferTest, ResultObjectLocationForStdInitializerListExpr) {
   std::string Code = R"(
     namespace std {

From 7e49b0d5a67f212e84f8ec0ec2e39a6a8673bfaf Mon Sep 17 00:00:00 2001
From: Daniil Kovalev <dkovalev@accesssoftek.com>
Date: Tue, 16 Apr 2024 09:53:33 +0300
Subject: [PATCH 529/886] [lldb] Fix nullptr dereference on running x86 binary
 with x86-disabled llvm (#82603)

If `LLVM_TARGETS_TO_BUILD` does not contain `X86` and we try to run an
x86 binary in lldb, we get a `nullptr` dereference in
`LLVMDisasmInstruction(...)`. We try to call `getDisAsm()` method on a
`LLVMDisasmContext *DC` which is null. The pointer is passed from
`x86AssemblyInspectionEngine::instruction_length(...)` and is originally
`m_disasm_context` member of `x86AssemblyInspectionEngine`. This should
be filled by `LLVMCreateDisasm(...)` in the class constructor, but not
having X86 target enabled in llvm makes
`TargetRegistry::lookupTarget(...)` call return `nullptr`, which results
in `m_disasm_context` initialized with `nullptr` as well.

This patch adds if statements against `m_disasm_context` in
`x86AssemblyInspectionEngine::GetNonCallSiteUnwindPlanFromAssembly(...)`
and `x86AssemblyInspectionEngine::FindFirstNonPrologueInstruction(...)`
so subsequent calls to
`x86AssemblyInspectionEngine::instruction_length(...)` do not cause a
null pointer dereference.
---
 .../x86/x86AssemblyInspectionEngine.cpp       |   6 +
 lldb/unittests/UnwindAssembly/CMakeLists.txt  |   4 +
 .../x86-but-no-x86-target/CMakeLists.txt      |  10 ++
 .../Testx86AssemblyInspectionEngine.cpp       | 103 ++++++++++++++++++
 4 files changed, 123 insertions(+)
 create mode 100644 lldb/unittests/UnwindAssembly/x86-but-no-x86-target/CMakeLists.txt
 create mode 100644 lldb/unittests/UnwindAssembly/x86-but-no-x86-target/Testx86AssemblyInspectionEngine.cpp

diff --git a/lldb/source/Plugins/UnwindAssembly/x86/x86AssemblyInspectionEngine.cpp b/lldb/source/Plugins/UnwindAssembly/x86/x86AssemblyInspectionEngine.cpp
index 2032c5a68d054..6bfaa54135a95 100644
--- a/lldb/source/Plugins/UnwindAssembly/x86/x86AssemblyInspectionEngine.cpp
+++ b/lldb/source/Plugins/UnwindAssembly/x86/x86AssemblyInspectionEngine.cpp
@@ -909,6 +909,9 @@ bool x86AssemblyInspectionEngine::GetNonCallSiteUnwindPlanFromAssembly(
   if (!m_register_map_initialized)
     return false;
 
+  if (m_disasm_context == nullptr)
+    return false;
+
   addr_t current_func_text_offset = 0;
   int current_sp_bytes_offset_from_fa = 0;
   bool is_aligned = false;
@@ -1570,6 +1573,9 @@ bool x86AssemblyInspectionEngine::FindFirstNonPrologueInstruction(
   if (!m_register_map_initialized)
     return false;
 
+  if (m_disasm_context == nullptr)
+    return false;
+
   while (offset < size) {
     int regno;
     int insn_len;
diff --git a/lldb/unittests/UnwindAssembly/CMakeLists.txt b/lldb/unittests/UnwindAssembly/CMakeLists.txt
index 136fcd9ae9798..d6e4471af4ecb 100644
--- a/lldb/unittests/UnwindAssembly/CMakeLists.txt
+++ b/lldb/unittests/UnwindAssembly/CMakeLists.txt
@@ -9,3 +9,7 @@ endif()
 if ("X86" IN_LIST LLVM_TARGETS_TO_BUILD)
   add_subdirectory(x86)
 endif()
+
+if (NOT "X86" IN_LIST LLVM_TARGETS_TO_BUILD)
+  add_subdirectory(x86-but-no-x86-target)
+endif()
diff --git a/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/CMakeLists.txt b/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/CMakeLists.txt
new file mode 100644
index 0000000000000..d28e9629a64cf
--- /dev/null
+++ b/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/CMakeLists.txt
@@ -0,0 +1,10 @@
+add_lldb_unittest(UnwindAssemblyX86ButNoX86TargetTests
+    Testx86AssemblyInspectionEngine.cpp
+    LINK_LIBS
+      lldbCore
+      lldbSymbol
+      lldbPluginUnwindAssemblyX86
+    LINK_COMPONENTS
+      Support
+      ${LLVM_TARGETS_TO_BUILD}
+  )
diff --git a/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/Testx86AssemblyInspectionEngine.cpp b/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/Testx86AssemblyInspectionEngine.cpp
new file mode 100644
index 0000000000000..ed093d146440e
--- /dev/null
+++ b/lldb/unittests/UnwindAssembly/x86-but-no-x86-target/Testx86AssemblyInspectionEngine.cpp
@@ -0,0 +1,103 @@
+//===-- Testx86AssemblyInspectionEngine.cpp -------------------------------===//
+
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "gtest/gtest.h"
+
+#include "Plugins/UnwindAssembly/x86/x86AssemblyInspectionEngine.h"
+#include "lldb/Core/AddressRange.h"
+#include "lldb/Symbol/UnwindPlan.h"
+#include "lldb/Utility/ArchSpec.h"
+
+#include "llvm/Support/TargetSelect.h"
+
+#include <memory>
+#include <vector>
+
+using namespace lldb;
+using namespace lldb_private;
+
+class Testx86AssemblyInspectionEngine : public testing::Test {
+public:
+  static void SetUpTestCase();
+};
+
+void Testx86AssemblyInspectionEngine::SetUpTestCase() {
+  llvm::InitializeAllTargets();
+  llvm::InitializeAllAsmPrinters();
+  llvm::InitializeAllTargetMCs();
+  llvm::InitializeAllDisassemblers();
+}
+
+// only defining the register names / numbers that the unwinder is actually
+// using today
+
+// names should match the constants below.  These will be the eRegisterKindLLDB
+// register numbers.
+
+const char *x86_64_reg_names[] = {"rax", "rbx", "rcx", "rdx", "rsp", "rbp",
+                                  "rsi", "rdi", "r8",  "r9",  "r10", "r11",
+                                  "r12", "r13", "r14", "r15", "rip"};
+
+enum x86_64_regs {
+  k_rax = 0,
+  k_rbx = 1,
+  k_rcx = 2,
+  k_rdx = 3,
+  k_rsp = 4,
+  k_rbp = 5,
+  k_rsi = 6,
+  k_rdi = 7,
+  k_r8 = 8,
+  k_r9 = 9,
+  k_r10 = 10,
+  k_r11 = 11,
+  k_r12 = 12,
+  k_r13 = 13,
+  k_r14 = 14,
+  k_r15 = 15,
+  k_rip = 16
+};
+
+std::unique_ptr<x86AssemblyInspectionEngine> Getx86_64Inspector() {
+
+  ArchSpec arch("x86_64-apple-macosx");
+  std::unique_ptr<x86AssemblyInspectionEngine> engine(
+      new x86AssemblyInspectionEngine(arch));
+
+  std::vector<x86AssemblyInspectionEngine::lldb_reg_info> lldb_regnums;
+  int i = 0;
+  for (const auto &name : x86_64_reg_names) {
+    x86AssemblyInspectionEngine::lldb_reg_info ri;
+    ri.name = name;
+    ri.lldb_regnum = i++;
+    lldb_regnums.push_back(ri);
+  }
+
+  engine->Initialize(lldb_regnums);
+  return engine;
+}
+
+TEST_F(Testx86AssemblyInspectionEngine, TestSimple64bitFrameFunction) {
+  std::unique_ptr<x86AssemblyInspectionEngine> engine = Getx86_64Inspector();
+
+  // 'int main() { }' compiled for x86_64-apple-macosx with clang
+  uint8_t data[] = {
+      0x55,             // offset 0 -- pushq %rbp
+      0x48, 0x89, 0xe5, // offset 1 -- movq %rsp, %rbp
+      0x31, 0xc0,       // offset 4 -- xorl %eax, %eax
+      0x5d,             // offset 6 -- popq %rbp
+      0xc3              // offset 7 -- retq
+  };
+
+  AddressRange sample_range(0x1000, sizeof(data));
+
+  UnwindPlan unwind_plan(eRegisterKindLLDB);
+  EXPECT_FALSE(engine->GetNonCallSiteUnwindPlanFromAssembly(
+      data, sizeof(data), sample_range, unwind_plan));
+}

From 89071f35599ac58cde99923a376d6b75c0d49e4b Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 00:09:41 -0700
Subject: [PATCH 530/886] [clang] Drop unaligned from calls to readNext (NFC)
 (#88842)

Now readNext defaults to unaligned accesses.  This patch drops
unaligned to improve readability.
---
 clang/lib/APINotes/APINotesReader.cpp         | 137 +++++++-----------
 clang/lib/Serialization/ASTReader.cpp         |  87 +++++------
 clang/lib/Serialization/GlobalModuleIndex.cpp |  12 +-
 .../lib/Serialization/MultiOnDiskHashTable.h  |   4 +-
 4 files changed, 98 insertions(+), 142 deletions(-)

diff --git a/clang/lib/APINotes/APINotesReader.cpp b/clang/lib/APINotes/APINotesReader.cpp
index fbbe9c32ce125..dfc3beb6fa13e 100644
--- a/clang/lib/APINotes/APINotesReader.cpp
+++ b/clang/lib/APINotes/APINotesReader.cpp
@@ -30,23 +30,20 @@ namespace {
 llvm::VersionTuple ReadVersionTuple(const uint8_t *&Data) {
   uint8_t NumVersions = (*Data++) & 0x03;
 
-  unsigned Major =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+  unsigned Major = endian::readNext<uint32_t, llvm::endianness::little>(Data);
   if (NumVersions == 0)
     return llvm::VersionTuple(Major);
 
-  unsigned Minor =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+  unsigned Minor = endian::readNext<uint32_t, llvm::endianness::little>(Data);
   if (NumVersions == 1)
     return llvm::VersionTuple(Major, Minor);
 
   unsigned Subminor =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint32_t, llvm::endianness::little>(Data);
   if (NumVersions == 2)
     return llvm::VersionTuple(Major, Minor, Subminor);
 
-  unsigned Build =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+  unsigned Build = endian::readNext<uint32_t, llvm::endianness::little>(Data);
   return llvm::VersionTuple(Major, Minor, Subminor, Build);
 }
 
@@ -71,16 +68,16 @@ class VersionedTableInfo {
 
   static std::pair<unsigned, unsigned> ReadKeyDataLength(const uint8_t *&Data) {
     unsigned KeyLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     unsigned DataLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     return {KeyLength, DataLength};
   }
 
   static data_type ReadData(internal_key_type Key, const uint8_t *Data,
                             unsigned Length) {
     unsigned NumElements =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     data_type Result;
     Result.reserve(NumElements);
     for (unsigned i = 0; i != NumElements; ++i) {
@@ -105,14 +102,14 @@ void ReadCommonEntityInfo(const uint8_t *&Data, CommonEntityInfo &Info) {
     Info.setSwiftPrivate(static_cast<bool>((UnavailableBits >> 3) & 0x01));
 
   unsigned MsgLength =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   Info.UnavailableMsg =
       std::string(reinterpret_cast<const char *>(Data),
                   reinterpret_cast<const char *>(Data) + MsgLength);
   Data += MsgLength;
 
   unsigned SwiftNameLength =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   Info.SwiftName =
       std::string(reinterpret_cast<const char *>(Data),
                   reinterpret_cast<const char *>(Data) + SwiftNameLength);
@@ -124,7 +121,7 @@ void ReadCommonTypeInfo(const uint8_t *&Data, CommonTypeInfo &Info) {
   ReadCommonEntityInfo(Data, Info);
 
   unsigned SwiftBridgeLength =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   if (SwiftBridgeLength > 0) {
     Info.setSwiftBridge(std::string(reinterpret_cast<const char *>(Data),
                                     SwiftBridgeLength - 1));
@@ -132,7 +129,7 @@ void ReadCommonTypeInfo(const uint8_t *&Data, CommonTypeInfo &Info) {
   }
 
   unsigned ErrorDomainLength =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   if (ErrorDomainLength > 0) {
     Info.setNSErrorDomain(std::optional<std::string>(std::string(
         reinterpret_cast<const char *>(Data), ErrorDomainLength - 1)));
@@ -163,9 +160,9 @@ class IdentifierTableInfo {
 
   static std::pair<unsigned, unsigned> ReadKeyDataLength(const uint8_t *&Data) {
     unsigned KeyLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     unsigned DataLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     return {KeyLength, DataLength};
   }
 
@@ -175,8 +172,7 @@ class IdentifierTableInfo {
 
   static data_type ReadData(internal_key_type key, const uint8_t *Data,
                             unsigned Length) {
-    return endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-        Data);
+    return endian::readNext<uint32_t, llvm::endianness::little>(Data);
   }
 };
 
@@ -203,26 +199,24 @@ class ObjCContextIDTableInfo {
 
   static std::pair<unsigned, unsigned> ReadKeyDataLength(const uint8_t *&Data) {
     unsigned KeyLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     unsigned DataLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     return {KeyLength, DataLength};
   }
 
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
     auto ParentCtxID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto ContextKind =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
-    auto NameID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data);
+    auto NameID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     return {ParentCtxID, ContextKind, NameID};
   }
 
   static data_type ReadData(internal_key_type Key, const uint8_t *Data,
                             unsigned Length) {
-    return endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-        Data);
+    return endian::readNext<uint32_t, llvm::endianness::little>(Data);
   }
 };
 
@@ -232,8 +226,7 @@ class ObjCContextInfoTableInfo
                                 ObjCContextInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    return endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-        Data);
+    return endian::readNext<uint32_t, llvm::endianness::little>(Data);
   }
 
   hash_value_type ComputeHash(internal_key_type Key) {
@@ -273,8 +266,7 @@ void ReadVariableInfo(const uint8_t *&Data, VariableInfo &Info) {
   }
   ++Data;
 
-  auto TypeLen =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+  auto TypeLen = endian::readNext<uint16_t, llvm::endianness::little>(Data);
   Info.setType(std::string(Data, Data + TypeLen));
   Data += TypeLen;
 }
@@ -286,12 +278,9 @@ class ObjCPropertyTableInfo
                                 ObjCPropertyInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto ClassID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
-    auto NameID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
-    char IsInstance =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+    auto ClassID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
+    auto NameID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
+    char IsInstance = endian::readNext<uint8_t, llvm::endianness::little>(Data);
     return {ClassID, NameID, IsInstance};
   }
 
@@ -314,8 +303,7 @@ class ObjCPropertyTableInfo
 void ReadParamInfo(const uint8_t *&Data, ParamInfo &Info) {
   ReadVariableInfo(Data, Info);
 
-  uint8_t Payload =
-      endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+  uint8_t Payload = endian::readNext<uint8_t, llvm::endianness::little>(Data);
   if (auto RawConvention = Payload & 0x7) {
     auto Convention = static_cast<RetainCountConventionKind>(RawConvention - 1);
     Info.setRetainCountConvention(Convention);
@@ -331,8 +319,7 @@ void ReadParamInfo(const uint8_t *&Data, ParamInfo &Info) {
 void ReadFunctionInfo(const uint8_t *&Data, FunctionInfo &Info) {
   ReadCommonEntityInfo(Data, Info);
 
-  uint8_t Payload =
-      endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+  uint8_t Payload = endian::readNext<uint8_t, llvm::endianness::little>(Data);
   if (auto RawConvention = Payload & 0x7) {
     auto Convention = static_cast<RetainCountConventionKind>(RawConvention - 1);
     Info.setRetainCountConvention(Convention);
@@ -343,12 +330,12 @@ void ReadFunctionInfo(const uint8_t *&Data, FunctionInfo &Info) {
   assert(Payload == 0 && "Bad API notes");
 
   Info.NumAdjustedNullable =
-      endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint8_t, llvm::endianness::little>(Data);
   Info.NullabilityPayload =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint64_t, llvm::endianness::little>(Data);
 
   unsigned NumParams =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   while (NumParams > 0) {
     ParamInfo pi;
     ReadParamInfo(Data, pi);
@@ -357,7 +344,7 @@ void ReadFunctionInfo(const uint8_t *&Data, FunctionInfo &Info) {
   }
 
   unsigned ResultTypeLen =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+      endian::readNext<uint16_t, llvm::endianness::little>(Data);
   Info.ResultType = std::string(Data, Data + ResultTypeLen);
   Data += ResultTypeLen;
 }
@@ -369,12 +356,10 @@ class ObjCMethodTableInfo
                                 ObjCMethodInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto ClassID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto ClassID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto SelectorID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
-    auto IsInstance =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
+    auto IsInstance = endian::readNext<uint8_t, llvm::endianness::little>(Data);
     return {ClassID, SelectorID, IsInstance};
   }
 
@@ -419,29 +404,26 @@ class ObjCSelectorTableInfo {
 
   static std::pair<unsigned, unsigned> ReadKeyDataLength(const uint8_t *&Data) {
     unsigned KeyLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     unsigned DataLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     return {KeyLength, DataLength};
   }
 
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
     internal_key_type Key;
-    Key.NumArgs =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+    Key.NumArgs = endian::readNext<uint16_t, llvm::endianness::little>(Data);
     unsigned NumIdents = (Length - sizeof(uint16_t)) / sizeof(uint32_t);
     for (unsigned i = 0; i != NumIdents; ++i) {
       Key.Identifiers.push_back(
-          endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-              Data));
+          endian::readNext<uint32_t, llvm::endianness::little>(Data));
     }
     return Key;
   }
 
   static data_type ReadData(internal_key_type Key, const uint8_t *Data,
                             unsigned Length) {
-    return endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-        Data);
+    return endian::readNext<uint32_t, llvm::endianness::little>(Data);
   }
 };
 
@@ -451,12 +433,10 @@ class GlobalVariableTableInfo
                                 GlobalVariableInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto CtxID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto CtxID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto ContextKind =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
-    auto NameID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data);
+    auto NameID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     return {CtxID, ContextKind, NameID};
   }
 
@@ -478,12 +458,10 @@ class GlobalFunctionTableInfo
                                 GlobalFunctionInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto CtxID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto CtxID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto ContextKind =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
-    auto NameID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data);
+    auto NameID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     return {CtxID, ContextKind, NameID};
   }
 
@@ -505,8 +483,7 @@ class EnumConstantTableInfo
                                 EnumConstantInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto NameID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto NameID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     return NameID;
   }
 
@@ -527,13 +504,11 @@ class TagTableInfo
     : public VersionedTableInfo<TagTableInfo, ContextTableKey, TagInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto CtxID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto CtxID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto ContextKind =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data);
     auto NameID =
-        endian::readNext<IdentifierID, llvm::endianness::little, unaligned>(
-            Data);
+        endian::readNext<IdentifierID, llvm::endianness::little>(Data);
     return {CtxID, ContextKind, NameID};
   }
 
@@ -553,21 +528,21 @@ class TagTableInfo
           static_cast<EnumExtensibilityKind>((Payload & 0x3) - 1);
 
     unsigned ImportAsLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     if (ImportAsLength > 0) {
       Info.SwiftImportAs =
           std::string(reinterpret_cast<const char *>(Data), ImportAsLength - 1);
       Data += ImportAsLength - 1;
     }
     unsigned RetainOpLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     if (RetainOpLength > 0) {
       Info.SwiftRetainOp =
           std::string(reinterpret_cast<const char *>(Data), RetainOpLength - 1);
       Data += RetainOpLength - 1;
     }
     unsigned ReleaseOpLength =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint16_t, llvm::endianness::little>(Data);
     if (ReleaseOpLength > 0) {
       Info.SwiftReleaseOp = std::string(reinterpret_cast<const char *>(Data),
                                         ReleaseOpLength - 1);
@@ -585,13 +560,11 @@ class TypedefTableInfo
                                 TypedefInfo> {
 public:
   static internal_key_type ReadKey(const uint8_t *Data, unsigned Length) {
-    auto CtxID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+    auto CtxID = endian::readNext<uint32_t, llvm::endianness::little>(Data);
     auto ContextKind =
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data);
     auto nameID =
-        endian::readNext<IdentifierID, llvm::endianness::little, unaligned>(
-            Data);
+        endian::readNext<IdentifierID, llvm::endianness::little>(Data);
     return {CtxID, ContextKind, nameID};
   }
 
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 8c4b460970ad2..feb60bc54413a 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -915,10 +915,9 @@ ASTSelectorLookupTrait::ReadKey(const unsigned char* d, unsigned) {
   using namespace llvm::support;
 
   SelectorTable &SelTable = Reader.getContext().Selectors;
-  unsigned N =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
+  unsigned N = endian::readNext<uint16_t, llvm::endianness::little>(d);
   const IdentifierInfo *FirstII = Reader.getLocalIdentifier(
-      F, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d));
+      F, endian::readNext<uint32_t, llvm::endianness::little>(d));
   if (N == 0)
     return SelTable.getNullarySelector(FirstII);
   else if (N == 1)
@@ -928,7 +927,7 @@ ASTSelectorLookupTrait::ReadKey(const unsigned char* d, unsigned) {
   Args.push_back(FirstII);
   for (unsigned I = 1; I != N; ++I)
     Args.push_back(Reader.getLocalIdentifier(
-        F, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d)));
+        F, endian::readNext<uint32_t, llvm::endianness::little>(d)));
 
   return SelTable.getSelector(N, Args.data());
 }
@@ -941,11 +940,11 @@ ASTSelectorLookupTrait::ReadData(Selector, const unsigned char* d,
   data_type Result;
 
   Result.ID = Reader.getGlobalSelectorID(
-      F, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d));
+      F, endian::readNext<uint32_t, llvm::endianness::little>(d));
   unsigned FullInstanceBits =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
+      endian::readNext<uint16_t, llvm::endianness::little>(d);
   unsigned FullFactoryBits =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
+      endian::readNext<uint16_t, llvm::endianness::little>(d);
   Result.InstanceBits = FullInstanceBits & 0x3;
   Result.InstanceHasMoreThanOneDecl = (FullInstanceBits >> 2) & 0x1;
   Result.FactoryBits = FullFactoryBits & 0x3;
@@ -956,16 +955,14 @@ ASTSelectorLookupTrait::ReadData(Selector, const unsigned char* d,
   // Load instance methods
   for (unsigned I = 0; I != NumInstanceMethods; ++I) {
     if (ObjCMethodDecl *Method = Reader.GetLocalDeclAs<ObjCMethodDecl>(
-            F,
-            endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d)))
+            F, endian::readNext<uint32_t, llvm::endianness::little>(d)))
       Result.Instance.push_back(Method);
   }
 
   // Load factory methods
   for (unsigned I = 0; I != NumFactoryMethods; ++I) {
     if (ObjCMethodDecl *Method = Reader.GetLocalDeclAs<ObjCMethodDecl>(
-            F,
-            endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d)))
+            F, endian::readNext<uint32_t, llvm::endianness::little>(d)))
       Result.Factory.push_back(Method);
   }
 
@@ -1009,8 +1006,7 @@ static bool readBit(unsigned &Bits) {
 IdentID ASTIdentifierLookupTrait::ReadIdentifierID(const unsigned char *d) {
   using namespace llvm::support;
 
-  unsigned RawID =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+  unsigned RawID = endian::readNext<uint32_t, llvm::endianness::little>(d);
   return Reader.getGlobalIdentifierID(F, RawID >> 1);
 }
 
@@ -1028,8 +1024,7 @@ IdentifierInfo *ASTIdentifierLookupTrait::ReadData(const internal_key_type& k,
                                                    unsigned DataLen) {
   using namespace llvm::support;
 
-  unsigned RawID =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+  unsigned RawID = endian::readNext<uint32_t, llvm::endianness::little>(d);
   bool IsInteresting = RawID & 0x01;
 
   // Wipe out the "is interesting" bit.
@@ -1053,9 +1048,8 @@ IdentifierInfo *ASTIdentifierLookupTrait::ReadData(const internal_key_type& k,
   }
 
   unsigned ObjCOrBuiltinID =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
-  unsigned Bits =
-      endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
+      endian::readNext<uint16_t, llvm::endianness::little>(d);
+  unsigned Bits = endian::readNext<uint16_t, llvm::endianness::little>(d);
   bool CPlusPlusOperatorKeyword = readBit(Bits);
   bool HasRevertedTokenIDToIdentifier = readBit(Bits);
   bool Poisoned = readBit(Bits);
@@ -1084,7 +1078,7 @@ IdentifierInfo *ASTIdentifierLookupTrait::ReadData(const internal_key_type& k,
   // definition.
   if (HadMacroDefinition) {
     uint32_t MacroDirectivesOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+        endian::readNext<uint32_t, llvm::endianness::little>(d);
     DataLen -= 4;
 
     Reader.addPendingMacro(II, &F, MacroDirectivesOffset);
@@ -1098,8 +1092,7 @@ IdentifierInfo *ASTIdentifierLookupTrait::ReadData(const internal_key_type& k,
     SmallVector<uint32_t, 4> DeclIDs;
     for (; DataLen > 0; DataLen -= 4)
       DeclIDs.push_back(Reader.getGlobalDeclID(
-          F,
-          endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d)));
+          F, endian::readNext<uint32_t, llvm::endianness::little>(d)));
     Reader.SetGloballyVisibleDecls(II, DeclIDs);
   }
 
@@ -1169,7 +1162,7 @@ ASTDeclContextNameLookupTrait::ReadFileRef(const unsigned char *&d) {
   using namespace llvm::support;
 
   uint32_t ModuleFileID =
-      endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+      endian::readNext<uint32_t, llvm::endianness::little>(d);
   return Reader.getLocalModuleFile(F, ModuleFileID);
 }
 
@@ -1189,18 +1182,15 @@ ASTDeclContextNameLookupTrait::ReadKey(const unsigned char *d, unsigned) {
   case DeclarationName::CXXLiteralOperatorName:
   case DeclarationName::CXXDeductionGuideName:
     Data = (uint64_t)Reader.getLocalIdentifier(
-        F, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d));
+        F, endian::readNext<uint32_t, llvm::endianness::little>(d));
     break;
   case DeclarationName::ObjCZeroArgSelector:
   case DeclarationName::ObjCOneArgSelector:
   case DeclarationName::ObjCMultiArgSelector:
-    Data =
-        (uint64_t)Reader
-            .getLocalSelector(
-                F,
-                endian::readNext<uint32_t, llvm::endianness::little, unaligned>(
-                    d))
-            .getAsOpaquePtr();
+    Data = (uint64_t)Reader
+               .getLocalSelector(
+                   F, endian::readNext<uint32_t, llvm::endianness::little>(d))
+               .getAsOpaquePtr();
     break;
   case DeclarationName::CXXOperatorName:
     Data = *d++; // OverloadedOperatorKind
@@ -1223,8 +1213,7 @@ void ASTDeclContextNameLookupTrait::ReadDataInto(internal_key_type,
   using namespace llvm::support;
 
   for (unsigned NumDecls = DataLen / 4; NumDecls; --NumDecls) {
-    uint32_t LocalID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+    uint32_t LocalID = endian::readNext<uint32_t, llvm::endianness::little>(d);
     Val.insert(Reader.getGlobalDeclID(F, LocalID));
   }
 }
@@ -2033,10 +2022,9 @@ HeaderFileInfoTrait::ReadKey(const unsigned char *d, unsigned) {
   using namespace llvm::support;
 
   internal_key_type ikey;
-  ikey.Size =
-      off_t(endian::readNext<uint64_t, llvm::endianness::little, unaligned>(d));
-  ikey.ModTime = time_t(
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(d));
+  ikey.Size = off_t(endian::readNext<uint64_t, llvm::endianness::little>(d));
+  ikey.ModTime =
+      time_t(endian::readNext<uint64_t, llvm::endianness::little>(d));
   ikey.Filename = (const char *)d;
   ikey.Imported = true;
   return ikey;
@@ -2064,9 +2052,9 @@ HeaderFileInfoTrait::ReadData(internal_key_ref key, const unsigned char *d,
   HFI.DirInfo = (Flags >> 1) & 0x07;
   HFI.IndexHeaderMapHeader = Flags & 0x01;
   HFI.ControllingMacroID = Reader.getGlobalIdentifierID(
-      M, endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d));
+      M, endian::readNext<uint32_t, llvm::endianness::little>(d));
   if (unsigned FrameworkOffset =
-          endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d)) {
+          endian::readNext<uint32_t, llvm::endianness::little>(d)) {
     // The framework offset is 1 greater than the actual offset,
     // since 0 is used as an indicator for "no framework name".
     StringRef FrameworkName(FrameworkStrings + FrameworkOffset - 1);
@@ -2077,7 +2065,7 @@ HeaderFileInfoTrait::ReadData(internal_key_ref key, const unsigned char *d,
          "Wrong data length in HeaderFileInfo deserialization");
   while (d != End) {
     uint32_t LocalSMID =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+        endian::readNext<uint32_t, llvm::endianness::little>(d);
     auto HeaderRole = static_cast<ModuleMap::ModuleHeaderRole>(LocalSMID & 7);
     LocalSMID >>= 3;
 
@@ -4085,9 +4073,8 @@ void ASTReader::ReadModuleOffsetMap(ModuleFile &F) const {
     // how it goes...
     using namespace llvm::support;
     ModuleKind Kind = static_cast<ModuleKind>(
-        endian::readNext<uint8_t, llvm::endianness::little, unaligned>(Data));
-    uint16_t Len =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint8_t, llvm::endianness::little>(Data));
+    uint16_t Len = endian::readNext<uint16_t, llvm::endianness::little>(Data);
     StringRef Name = StringRef((const char*)Data, Len);
     Data += Len;
     ModuleFile *OM = (Kind == MK_PrebuiltModule || Kind == MK_ExplicitModule ||
@@ -4103,21 +4090,21 @@ void ASTReader::ReadModuleOffsetMap(ModuleFile &F) const {
     }
 
     SourceLocation::UIntTy SLocOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t IdentifierIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t MacroIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t PreprocessedEntityIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t SubmoduleIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t SelectorIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t DeclIDOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
     uint32_t TypeIndexOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Data);
+        endian::readNext<uint32_t, llvm::endianness::little>(Data);
 
     auto mapOffset = [&](uint32_t Offset, uint32_t BaseOffset,
                          RemapBuilder &Remap) {
diff --git a/clang/lib/Serialization/GlobalModuleIndex.cpp b/clang/lib/Serialization/GlobalModuleIndex.cpp
index dd4fc3e009050..8ff10f6a8621e 100644
--- a/clang/lib/Serialization/GlobalModuleIndex.cpp
+++ b/clang/lib/Serialization/GlobalModuleIndex.cpp
@@ -89,10 +89,8 @@ class IdentifierIndexReaderTrait {
   static std::pair<unsigned, unsigned>
   ReadKeyDataLength(const unsigned char*& d) {
     using namespace llvm::support;
-    unsigned KeyLen =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
-    unsigned DataLen =
-        endian::readNext<uint16_t, llvm::endianness::little, unaligned>(d);
+    unsigned KeyLen = endian::readNext<uint16_t, llvm::endianness::little>(d);
+    unsigned DataLen = endian::readNext<uint16_t, llvm::endianness::little>(d);
     return std::make_pair(KeyLen, DataLen);
   }
 
@@ -113,8 +111,7 @@ class IdentifierIndexReaderTrait {
 
     data_type Result;
     while (DataLen > 0) {
-      unsigned ID =
-          endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+      unsigned ID = endian::readNext<uint32_t, llvm::endianness::little>(d);
       Result.push_back(ID);
       DataLen -= 4;
     }
@@ -514,8 +511,7 @@ namespace {
       // The first bit indicates whether this identifier is interesting.
       // That's all we care about.
       using namespace llvm::support;
-      unsigned RawID =
-          endian::readNext<uint32_t, llvm::endianness::little, unaligned>(d);
+      unsigned RawID = endian::readNext<uint32_t, llvm::endianness::little>(d);
       bool IsInteresting = RawID & 0x01;
       return std::make_pair(k, IsInteresting);
     }
diff --git a/clang/lib/Serialization/MultiOnDiskHashTable.h b/clang/lib/Serialization/MultiOnDiskHashTable.h
index 2402a628b512f..a0d75ec3a9e76 100644
--- a/clang/lib/Serialization/MultiOnDiskHashTable.h
+++ b/clang/lib/Serialization/MultiOnDiskHashTable.h
@@ -200,11 +200,11 @@ template<typename Info> class MultiOnDiskHashTable {
     storage_type Ptr = Data;
 
     uint32_t BucketOffset =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint32_t, llvm::endianness::little>(Ptr);
 
     // Read the list of overridden files.
     uint32_t NumFiles =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint32_t, llvm::endianness::little>(Ptr);
     // FIXME: Add a reserve() to TinyPtrVector so that we don't need to make
     // an additional copy.
     llvm::SmallVector<file_type, 16> OverriddenFiles;

From a169d4c2e974ceb20b86faea3fa4ac286a1d44e8 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Tue, 16 Apr 2024 10:37:15 +0300
Subject: [PATCH 531/886] =?UTF-8?q?[LLD]=C2=A0[COFF]=20Error=20out=20if=20?=
 =?UTF-8?q?the=20runtime=20pseudo=20relocation=20function=20is=20missing?=
 =?UTF-8?q?=20(#88573)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

When then linker creates runtime pseudo relocations, it places them in a
list with the assumption that the runtime will fix these relocations
later, when the image gets loaded. If the relevant runtime function
doesn't seem to be present in the linked image, error out.

Normally when linking the mingw-w64 runtime libraries, this function
always is available. However, if linking without including the mingw-w64
CRT startup files, and the image needs runtime pseudo relocations, make
it clear that this won't work as expected at runtime.

With ld.bfd, this situation is a hard error too; ld.bfd adds an
undefined reference to this symbol if runtime pseudo relocations are
needed.

A later alternative would be to actually try to pull in the symbol (if
seen in a static library, but not included yet). This would allow
decoupling the function from the main mingw-w64 CRT startup code (making
it optional, only running if the linker actually produced runtime pseudo
relocations).

Doing that would require restructuring the lld code (gathering pseudo
relocations earlier, then loading the relocator function, then pulling in
more object files to satisfy the dependencies of the relocator) though.

Also, ld.bfd doesn't currently successfully pull in more object files to
satisfy the dependency on _pei386_runtime_relocator, so with that in
mind, there's not much extra value in making LLD do it currently either;
we can't make such a change in mingw-w64's CRT until both linkers
handle it.

This fixes one issue brought up in
https://github.com/llvm/llvm-project/issues/84424.
---
 lld/COFF/Writer.cpp                     | 10 ++++++-
 lld/test/COFF/autoimport-arm-data.s     |  3 +++
 lld/test/COFF/autoimport-arm64-data.s   |  3 +++
 lld/test/COFF/autoimport-gnu-implib.s   |  3 +++
 lld/test/COFF/autoimport-handler-func.s | 36 +++++++++++++++++++++++++
 lld/test/COFF/autoimport-warn.s         |  3 +++
 lld/test/COFF/autoimport-x86.s          |  3 +++
 7 files changed, 60 insertions(+), 1 deletion(-)
 create mode 100644 lld/test/COFF/autoimport-handler-func.s

diff --git a/lld/COFF/Writer.cpp b/lld/COFF/Writer.cpp
index 9c20bbb83d86d..7269d156752de 100644
--- a/lld/COFF/Writer.cpp
+++ b/lld/COFF/Writer.cpp
@@ -2072,8 +2072,16 @@ void Writer::createRuntimePseudoRelocs() {
     return;
   }
 
-  if (!rels.empty())
+  if (!rels.empty()) {
     log("Writing " + Twine(rels.size()) + " runtime pseudo relocations");
+    const char *symbolName = "_pei386_runtime_relocator";
+    Symbol *relocator = ctx.symtab.findUnderscore(symbolName);
+    if (!relocator)
+      error("output image has runtime pseudo relocations, but the function " +
+            Twine(symbolName) +
+            " is missing; it is needed for fixing the relocations at runtime");
+  }
+
   PseudoRelocTableChunk *table = make<PseudoRelocTableChunk>(rels);
   rdataSec->addChunk(table);
   EmptyChunk *endOfList = make<EmptyChunk>();
diff --git a/lld/test/COFF/autoimport-arm-data.s b/lld/test/COFF/autoimport-arm-data.s
index 74604aa5c8234..82c66f0989d49 100644
--- a/lld/test/COFF/autoimport-arm-data.s
+++ b/lld/test/COFF/autoimport-arm-data.s
@@ -33,6 +33,9 @@
     .text
     .thumb
 main:
+    bx lr
+    .global _pei386_runtime_relocator
+_pei386_runtime_relocator:
     bx lr
     .data
 ptr:
diff --git a/lld/test/COFF/autoimport-arm64-data.s b/lld/test/COFF/autoimport-arm64-data.s
index fa3654be3a71d..b49bd4f89c97c 100644
--- a/lld/test/COFF/autoimport-arm64-data.s
+++ b/lld/test/COFF/autoimport-arm64-data.s
@@ -33,6 +33,9 @@
     .global main
     .text
 main:
+    ret
+    .global _pei386_runtime_relocator
+_pei386_runtime_relocator:
     ret
     .data
 ptr:
diff --git a/lld/test/COFF/autoimport-gnu-implib.s b/lld/test/COFF/autoimport-gnu-implib.s
index d7d4ed626e83a..d9dc9d7a38fdc 100644
--- a/lld/test/COFF/autoimport-gnu-implib.s
+++ b/lld/test/COFF/autoimport-gnu-implib.s
@@ -27,5 +27,8 @@
     .text
 main:
     movl data(%rip), %eax
+    ret
+    .global _pei386_runtime_relocator
+_pei386_runtime_relocator:
     ret
     .data
diff --git a/lld/test/COFF/autoimport-handler-func.s b/lld/test/COFF/autoimport-handler-func.s
new file mode 100644
index 0000000000000..02d040bfa274c
--- /dev/null
+++ b/lld/test/COFF/autoimport-handler-func.s
@@ -0,0 +1,36 @@
+# REQUIRES: x86
+# RUN: split-file %s %t.dir
+
+# RUN: llvm-dlltool -m i386:x86-64 -d %t.dir/lib.def -D lib.dll -l %t.dir/lib.lib
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/main.s -filetype=obj -o %t.dir/main.obj
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/func.s -filetype=obj -o %t.dir/func.obj
+# RUN: env LLD_IN_TEST=1 not lld-link -lldmingw -out:%t.dir/main.exe -entry:main %t.dir/main.obj %t.dir/lib.lib 2>&1 | FileCheck %s --check-prefix=ERR
+
+# RUN: lld-link -lldmingw -out:%t.dir/main.exe -entry:main %t.dir/main.obj %t.dir/func.obj %t.dir/lib.lib 2>&1 | FileCheck %s --check-prefix=NOERR --allow-empty
+
+# ERR: error: output image has runtime pseudo relocations, but the function _pei386_runtime_relocator is missing; it is needed for fixing the relocations at runtime
+
+# NOERR-NOT: error
+
+#--- main.s
+    .global main
+    .text
+main:
+    ret
+
+    .data
+    .long 1
+    .quad variable
+    .long 2
+
+#--- func.s
+    .global _pei386_runtime_relocator
+    .text
+_pei386_runtime_relocator:
+    ret
+
+#--- lib.def
+EXPORTS
+variable DATA
+
diff --git a/lld/test/COFF/autoimport-warn.s b/lld/test/COFF/autoimport-warn.s
index 9c363ed30f245..eead0fed861f8 100644
--- a/lld/test/COFF/autoimport-warn.s
+++ b/lld/test/COFF/autoimport-warn.s
@@ -18,6 +18,9 @@ main:
     movl variable2(%rip), %ecx
     addl %ecx, %eax
     ret
+    .global _pei386_runtime_relocator
+_pei386_runtime_relocator:
+    ret
 
     .section .rdata$.refptr.variable1,"dr",discard,.refptr.variable1
     .global .refptr.variable1
diff --git a/lld/test/COFF/autoimport-x86.s b/lld/test/COFF/autoimport-x86.s
index fa36f10e9ca91..5d7c9c2c3fa58 100644
--- a/lld/test/COFF/autoimport-x86.s
+++ b/lld/test/COFF/autoimport-x86.s
@@ -55,6 +55,9 @@
     .text
 main:
     movl variable(%rip), %eax
+    ret
+    .global _pei386_runtime_relocator
+_pei386_runtime_relocator:
     ret
     .data
 ptr:

From dbaa1893c9afe6a245860efb8d68875ba4fd6794 Mon Sep 17 00:00:00 2001
From: Pengcheng Wang <wangpengcheng.pp@bytedance.com>
Date: Tue, 16 Apr 2024 15:37:31 +0800
Subject: [PATCH 532/886] [RISCV] Generate more W instructons

We rename `TuneNoStripWSuffix` to `TunePreferWInst`.

If all the users of an instruction just use the low 32 bits, we can
convert it to its W variant.

A quick test on Coremark (`-O3 -march=rv64gc`):

|        | W instructions | code size(.text) |
|--------|----------------|------------------|
| before | 302            | 12257            |
| after  | 343            | 12265            |
|        | +13.58%        | +0.065%          |

Reviewers: asb, dtcxzyw, preames, lukel97, michaelmaitland, topperc

Reviewed By: topperc, dtcxzyw

Pull Request: https://github.com/llvm/llvm-project/pull/87237
---
 llvm/lib/Target/RISCV/RISCVFeatures.td    |   6 +-
 llvm/lib/Target/RISCV/RISCVOptWInstrs.cpp |  89 +++++++-
 llvm/test/CodeGen/RISCV/prefer-w-inst.ll  | 105 +++++++++
 llvm/test/CodeGen/RISCV/prefer-w-inst.mir | 262 ++++++++++++++++++++++
 llvm/test/CodeGen/RISCV/strip-w-suffix.ll |  74 ------
 5 files changed, 447 insertions(+), 89 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/prefer-w-inst.ll
 create mode 100644 llvm/test/CodeGen/RISCV/prefer-w-inst.mir
 delete mode 100644 llvm/test/CodeGen/RISCV/strip-w-suffix.ll

diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index 794455aa73040..59962216e0c04 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -1226,9 +1226,9 @@ def TuneNoSinkSplatOperands
                        "false", "Disable sink splat operands to enable .vx, .vf,"
                        ".wx, and .wf instructions">;
 
-def TuneNoStripWSuffix
-    : SubtargetFeature<"no-strip-w-suffix", "EnableStripWSuffix", "false",
-                       "Disable strip W suffix">;
+def TunePreferWInst
+    : SubtargetFeature<"prefer-w-inst", "PreferWInst", "true",
+                       "Prefer instructions with W suffix">;
 
 def TuneConditionalCompressedMoveFusion
     : SubtargetFeature<"conditional-cmv-fusion", "HasConditionalCompressedMoveFusion",
diff --git a/llvm/lib/Target/RISCV/RISCVOptWInstrs.cpp b/llvm/lib/Target/RISCV/RISCVOptWInstrs.cpp
index 39d420c2fbf08..ead91c5656be8 100644
--- a/llvm/lib/Target/RISCV/RISCVOptWInstrs.cpp
+++ b/llvm/lib/Target/RISCV/RISCVOptWInstrs.cpp
@@ -12,15 +12,24 @@
 // extended bits aren't consumed or because the input was already sign extended
 // by an earlier instruction.
 //
-// Then it removes the -w suffix from opw instructions whenever all users are
-// dependent only on the lower word of the result of the instruction.
-// The cases handled are:
-// * addw because c.add has a larger register encoding than c.addw.
-// * addiw because it helps reduce test differences between RV32 and RV64
-//   w/o being a pessimization.
-// * mulw because c.mulw doesn't exist but c.mul does (w/ zcb)
-// * slliw because c.slliw doesn't exist and c.slli does
+// Then:
+// 1. Unless explicit disabled or the target prefers instructions with W suffix,
+//    it removes the -w suffix from opw instructions whenever all users are
+//    dependent only on the lower word of the result of the instruction.
+//    The cases handled are:
+//    * addw because c.add has a larger register encoding than c.addw.
+//    * addiw because it helps reduce test differences between RV32 and RV64
+//      w/o being a pessimization.
+//    * mulw because c.mulw doesn't exist but c.mul does (w/ zcb)
+//    * slliw because c.slliw doesn't exist and c.slli does
 //
+// 2. Or if explicit enabled or the target prefers instructions with W suffix,
+//    it adds the W suffix to the instruction whenever all users are dependent
+//    only on the lower word of the result of the instruction.
+//    The cases handled are:
+//    * add/addi/sub/mul.
+//    * slli with imm < 32.
+//    * ld/lwu.
 //===---------------------------------------------------------------------===//
 
 #include "RISCV.h"
@@ -60,6 +69,8 @@ class RISCVOptWInstrs : public MachineFunctionPass {
                          const RISCVSubtarget &ST, MachineRegisterInfo &MRI);
   bool stripWSuffixes(MachineFunction &MF, const RISCVInstrInfo &TII,
                       const RISCVSubtarget &ST, MachineRegisterInfo &MRI);
+  bool appendWSuffixes(MachineFunction &MF, const RISCVInstrInfo &TII,
+                       const RISCVSubtarget &ST, MachineRegisterInfo &MRI);
 
   void getAnalysisUsage(AnalysisUsage &AU) const override {
     AU.setPreservesCFG();
@@ -672,9 +683,6 @@ bool RISCVOptWInstrs::stripWSuffixes(MachineFunction &MF,
                                      const RISCVInstrInfo &TII,
                                      const RISCVSubtarget &ST,
                                      MachineRegisterInfo &MRI) {
-  if (DisableStripWSuffix || !ST.enableStripWSuffix())
-    return false;
-
   bool MadeChange = false;
   for (MachineBasicBlock &MBB : MF) {
     for (MachineInstr &MI : MBB) {
@@ -698,6 +706,58 @@ bool RISCVOptWInstrs::stripWSuffixes(MachineFunction &MF,
   return MadeChange;
 }
 
+bool RISCVOptWInstrs::appendWSuffixes(MachineFunction &MF,
+                                      const RISCVInstrInfo &TII,
+                                      const RISCVSubtarget &ST,
+                                      MachineRegisterInfo &MRI) {
+  bool MadeChange = false;
+  for (MachineBasicBlock &MBB : MF) {
+    for (MachineInstr &MI : MBB) {
+      unsigned WOpc;
+      // TODO: Add more?
+      switch (MI.getOpcode()) {
+      default:
+        continue;
+      case RISCV::ADD:
+        WOpc = RISCV::ADDW;
+        break;
+      case RISCV::ADDI:
+        WOpc = RISCV::ADDIW;
+        break;
+      case RISCV::SUB:
+        WOpc = RISCV::SUBW;
+        break;
+      case RISCV::MUL:
+        WOpc = RISCV::MULW;
+        break;
+      case RISCV::SLLI:
+        // SLLIW reads the lowest 5 bits, while SLLI reads lowest 6 bits
+        if (MI.getOperand(2).getImm() >= 32)
+          continue;
+        WOpc = RISCV::SLLIW;
+        break;
+      case RISCV::LD:
+      case RISCV::LWU:
+        WOpc = RISCV::LW;
+        break;
+      }
+
+      if (hasAllWUsers(MI, ST, MRI)) {
+        LLVM_DEBUG(dbgs() << "Replacing " << MI);
+        MI.setDesc(TII.get(WOpc));
+        MI.clearFlag(MachineInstr::MIFlag::NoSWrap);
+        MI.clearFlag(MachineInstr::MIFlag::NoUWrap);
+        MI.clearFlag(MachineInstr::MIFlag::IsExact);
+        LLVM_DEBUG(dbgs() << "     with " << MI);
+        ++NumTransformedToWInstrs;
+        MadeChange = true;
+      }
+    }
+  }
+
+  return MadeChange;
+}
+
 bool RISCVOptWInstrs::runOnMachineFunction(MachineFunction &MF) {
   if (skipFunction(MF.getFunction()))
     return false;
@@ -711,7 +771,12 @@ bool RISCVOptWInstrs::runOnMachineFunction(MachineFunction &MF) {
 
   bool MadeChange = false;
   MadeChange |= removeSExtWInstrs(MF, TII, ST, MRI);
-  MadeChange |= stripWSuffixes(MF, TII, ST, MRI);
+
+  if (!(DisableStripWSuffix || ST.preferWInst()))
+    MadeChange |= stripWSuffixes(MF, TII, ST, MRI);
+
+  if (ST.preferWInst())
+    MadeChange |= appendWSuffixes(MF, TII, ST, MRI);
 
   return MadeChange;
 }
diff --git a/llvm/test/CodeGen/RISCV/prefer-w-inst.ll b/llvm/test/CodeGen/RISCV/prefer-w-inst.ll
new file mode 100644
index 0000000000000..34ab74d78a76f
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/prefer-w-inst.ll
@@ -0,0 +1,105 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=NO-PREFER-W-INST %s
+; RUN: llc -mtriple=riscv64 -mattr=+m -riscv-disable-strip-w-suffix -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=NO-STRIP %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+prefer-w-inst -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefixes=PREFER-W-INST %s
+
+define i32 @addiw(i32 %a) {
+; NO-PREFER-W-INST-LABEL: addiw:
+; NO-PREFER-W-INST:       # %bb.0:
+; NO-PREFER-W-INST-NEXT:    lui a1, 1
+; NO-PREFER-W-INST-NEXT:    addi a1, a1, -1
+; NO-PREFER-W-INST-NEXT:    addw a0, a0, a1
+; NO-PREFER-W-INST-NEXT:    ret
+;
+; NO-STRIP-LABEL: addiw:
+; NO-STRIP:       # %bb.0:
+; NO-STRIP-NEXT:    lui a1, 1
+; NO-STRIP-NEXT:    addiw a1, a1, -1
+; NO-STRIP-NEXT:    addw a0, a0, a1
+; NO-STRIP-NEXT:    ret
+;
+; PREFER-W-INST-LABEL: addiw:
+; PREFER-W-INST:       # %bb.0:
+; PREFER-W-INST-NEXT:    lui a1, 1
+; PREFER-W-INST-NEXT:    addiw a1, a1, -1
+; PREFER-W-INST-NEXT:    addw a0, a0, a1
+; PREFER-W-INST-NEXT:    ret
+  %ret = add i32 %a, 4095
+  ret i32 %ret
+}
+
+define i32 @addw(i32 %a, i32 %b) {
+; NO-PREFER-W-INST-LABEL: addw:
+; NO-PREFER-W-INST:       # %bb.0:
+; NO-PREFER-W-INST-NEXT:    add a0, a0, a1
+; NO-PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; NO-PREFER-W-INST-NEXT:    ret
+;
+; NO-STRIP-LABEL: addw:
+; NO-STRIP:       # %bb.0:
+; NO-STRIP-NEXT:    addw a0, a0, a1
+; NO-STRIP-NEXT:    addiw a0, a0, 1024
+; NO-STRIP-NEXT:    ret
+;
+; PREFER-W-INST-LABEL: addw:
+; PREFER-W-INST:       # %bb.0:
+; PREFER-W-INST-NEXT:    addw a0, a0, a1
+; PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; PREFER-W-INST-NEXT:    ret
+  %add = add i32 %a, %b
+  %ret = add i32 %add, 1024
+  ret i32 %ret
+}
+
+define i32 @mulw(i32 %a, i32 %b) {
+; NO-PREFER-W-INST-LABEL: mulw:
+; NO-PREFER-W-INST:       # %bb.0:
+; NO-PREFER-W-INST-NEXT:    mul a1, a0, a1
+; NO-PREFER-W-INST-NEXT:    mul a0, a0, a1
+; NO-PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; NO-PREFER-W-INST-NEXT:    ret
+;
+; NO-STRIP-LABEL: mulw:
+; NO-STRIP:       # %bb.0:
+; NO-STRIP-NEXT:    mulw a1, a0, a1
+; NO-STRIP-NEXT:    mulw a0, a0, a1
+; NO-STRIP-NEXT:    addiw a0, a0, 1024
+; NO-STRIP-NEXT:    ret
+;
+; PREFER-W-INST-LABEL: mulw:
+; PREFER-W-INST:       # %bb.0:
+; PREFER-W-INST-NEXT:    mulw a1, a0, a1
+; PREFER-W-INST-NEXT:    mulw a0, a0, a1
+; PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; PREFER-W-INST-NEXT:    ret
+  %mul1 = mul i32 %a, %b
+  %mul = mul i32 %a, %mul1
+  %ret = add i32 %mul, 1024
+  ret i32 %ret
+}
+
+define i32 @slliw(i32 %a) {
+; NO-PREFER-W-INST-LABEL: slliw:
+; NO-PREFER-W-INST:       # %bb.0:
+; NO-PREFER-W-INST-NEXT:    slli a0, a0, 1
+; NO-PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; NO-PREFER-W-INST-NEXT:    ret
+;
+; NO-STRIP-LABEL: slliw:
+; NO-STRIP:       # %bb.0:
+; NO-STRIP-NEXT:    slliw a0, a0, 1
+; NO-STRIP-NEXT:    addiw a0, a0, 1024
+; NO-STRIP-NEXT:    ret
+;
+; PREFER-W-INST-LABEL: slliw:
+; PREFER-W-INST:       # %bb.0:
+; PREFER-W-INST-NEXT:    slliw a0, a0, 1
+; PREFER-W-INST-NEXT:    addiw a0, a0, 1024
+; PREFER-W-INST-NEXT:    ret
+  %shl = shl i32 %a, 1
+  %ret = add i32 %shl, 1024
+  ret i32 %ret
+}
diff --git a/llvm/test/CodeGen/RISCV/prefer-w-inst.mir b/llvm/test/CodeGen/RISCV/prefer-w-inst.mir
new file mode 100644
index 0000000000000..e05e27af4271c
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/prefer-w-inst.mir
@@ -0,0 +1,262 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
+# RUN: llc %s -mtriple=riscv64 -run-pass=riscv-opt-w-instrs -verify-machineinstrs \
+# RUN:   -mattr=+m -o - | FileCheck %s -check-prefixes=NO-PREFER-W-INST
+# RUN: llc %s -mtriple=riscv64 -run-pass=riscv-opt-w-instrs -verify-machineinstrs \
+# RUN:   -mattr=+m,+prefer-w-inst -o - | FileCheck %s -check-prefixes=PREFER-W-INST
+
+---
+name: addi
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: addi
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[ADDI:%[0-9]+]]:gpr = ADDI [[COPY]], 1
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[ADDI]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: addi
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[COPY]], 1
+    ; PREFER-W-INST-NEXT: [[ADDIW1:%[0-9]+]]:gpr = ADDIW [[ADDIW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW1]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = ADDI %1, 1
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: add
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: add
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[ADD:%[0-9]+]]:gpr = ADD [[COPY]], [[COPY1]]
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[ADD]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: add
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[ADDW:%[0-9]+]]:gpr = ADDW [[COPY]], [[COPY1]]
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[ADDW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = ADD %1, %2
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: sub
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: sub
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[SUB:%[0-9]+]]:gpr = SUB [[COPY]], [[COPY1]]
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SUB]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: sub
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[SUBW:%[0-9]+]]:gpr = SUBW [[COPY]], [[COPY1]]
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SUBW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = SUB %1, %2
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: mul
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: mul
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[MUL:%[0-9]+]]:gpr = MUL [[COPY]], [[COPY1]]
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[MUL]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: mul
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[MULW:%[0-9]+]]:gpr = MULW [[COPY]], [[COPY1]]
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[MULW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = MUL %1, %2
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+
+---
+name: slli_31
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: slli_31
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[SLLI:%[0-9]+]]:gpr = SLLI [[COPY]], 31
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SLLI]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: slli_31
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[SLLIW:%[0-9]+]]:gpr = SLLIW [[COPY]], 31
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SLLIW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = SLLI %1, 31
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: slli_32
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: slli_32
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[SLLI:%[0-9]+]]:gpr = SLLI [[COPY]], 32
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SLLI]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: slli_32
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[SLLI:%[0-9]+]]:gpr = SLLI [[COPY]], 32
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[SLLI]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = SLLI %1, 32
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: ld
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: ld
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[LD:%[0-9]+]]:gpr = LD [[COPY]], 0
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[LD]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: ld
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[LW:%[0-9]+]]:gpr = LW [[COPY]], 0
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[LW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = LD %1, 0
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
+
+---
+name: lwu
+body:             |
+  bb.0.entry:
+    liveins: $x10, $x11
+    ; NO-PREFER-W-INST-LABEL: name: lwu
+    ; NO-PREFER-W-INST: liveins: $x10, $x11
+    ; NO-PREFER-W-INST-NEXT: {{  $}}
+    ; NO-PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; NO-PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; NO-PREFER-W-INST-NEXT: [[LWU:%[0-9]+]]:gpr = LWU [[COPY]], 0
+    ; NO-PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[LWU]], 1
+    ; NO-PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; NO-PREFER-W-INST-NEXT: PseudoRET
+    ;
+    ; PREFER-W-INST-LABEL: name: lwu
+    ; PREFER-W-INST: liveins: $x10, $x11
+    ; PREFER-W-INST-NEXT: {{  $}}
+    ; PREFER-W-INST-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $x10
+    ; PREFER-W-INST-NEXT: [[COPY1:%[0-9]+]]:gpr = COPY $x11
+    ; PREFER-W-INST-NEXT: [[LW:%[0-9]+]]:gpr = LW [[COPY]], 0
+    ; PREFER-W-INST-NEXT: [[ADDIW:%[0-9]+]]:gpr = ADDIW [[LW]], 1
+    ; PREFER-W-INST-NEXT: $x10 = COPY [[ADDIW]]
+    ; PREFER-W-INST-NEXT: PseudoRET
+    %1:gpr = COPY $x10
+    %2:gpr = COPY $x11
+    %3:gpr = LWU %1, 0
+    %4:gpr = ADDIW %3, 1
+    $x10 = COPY %4
+    PseudoRET
+...
diff --git a/llvm/test/CodeGen/RISCV/strip-w-suffix.ll b/llvm/test/CodeGen/RISCV/strip-w-suffix.ll
deleted file mode 100644
index 4124b3d0d360d..0000000000000
--- a/llvm/test/CodeGen/RISCV/strip-w-suffix.ll
+++ /dev/null
@@ -1,74 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=riscv64 -mattr=+m -verify-machineinstrs < %s \
-; RUN:   | FileCheck -check-prefixes=STRIP %s
-; RUN: llc -mtriple=riscv64 -mattr=+m,+no-strip-w-suffix -verify-machineinstrs < %s \
-; RUN:   | FileCheck -check-prefixes=NO-STRIP %s
-
-define i32 @addiw(i32 %a) {
-; STRIP-LABEL: addiw:
-; STRIP:       # %bb.0:
-; STRIP-NEXT:    lui a1, 1
-; STRIP-NEXT:    addi a1, a1, -1
-; STRIP-NEXT:    addw a0, a0, a1
-; STRIP-NEXT:    ret
-;
-; NO-STRIP-LABEL: addiw:
-; NO-STRIP:       # %bb.0:
-; NO-STRIP-NEXT:    lui a1, 1
-; NO-STRIP-NEXT:    addiw a1, a1, -1
-; NO-STRIP-NEXT:    addw a0, a0, a1
-; NO-STRIP-NEXT:    ret
-  %ret = add i32 %a, 4095
-  ret i32 %ret
-}
-
-define i32 @addw(i32 %a, i32 %b) {
-; STRIP-LABEL: addw:
-; STRIP:       # %bb.0:
-; STRIP-NEXT:    add a0, a0, a1
-; STRIP-NEXT:    addiw a0, a0, 1024
-; STRIP-NEXT:    ret
-;
-; NO-STRIP-LABEL: addw:
-; NO-STRIP:       # %bb.0:
-; NO-STRIP-NEXT:    addw a0, a0, a1
-; NO-STRIP-NEXT:    addiw a0, a0, 1024
-; NO-STRIP-NEXT:    ret
-  %add = add i32 %a, %b
-  %ret = add i32 %add, 1024
-  ret i32 %ret
-}
-
-define i32 @mulw(i32 %a, i32 %b) {
-; STRIP-LABEL: mulw:
-; STRIP:       # %bb.0:
-; STRIP-NEXT:    mul a0, a0, a1
-; STRIP-NEXT:    addiw a0, a0, 1024
-; STRIP-NEXT:    ret
-;
-; NO-STRIP-LABEL: mulw:
-; NO-STRIP:       # %bb.0:
-; NO-STRIP-NEXT:    mulw a0, a0, a1
-; NO-STRIP-NEXT:    addiw a0, a0, 1024
-; NO-STRIP-NEXT:    ret
-  %mul = mul i32 %a, %b
-  %ret = add i32 %mul, 1024
-  ret i32 %ret
-}
-
-define i32 @slliw(i32 %a) {
-; STRIP-LABEL: slliw:
-; STRIP:       # %bb.0:
-; STRIP-NEXT:    slli a0, a0, 1
-; STRIP-NEXT:    addiw a0, a0, 1024
-; STRIP-NEXT:    ret
-;
-; NO-STRIP-LABEL: slliw:
-; NO-STRIP:       # %bb.0:
-; NO-STRIP-NEXT:    slliw a0, a0, 1
-; NO-STRIP-NEXT:    addiw a0, a0, 1024
-; NO-STRIP-NEXT:    ret
-  %shl = shl i32 %a, 1
-  %ret = add i32 %shl, 1024
-  ret i32 %ret
-}

From d26dd58ca5b59032eb371b8f51d9134acdd8d3ad Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Tue, 16 Apr 2024 14:37:38 +0800
Subject: [PATCH 533/886] [StmtProfile] Don't profile the body of lambda
 expressions

Close https://github.com/llvm/llvm-project/issues/87609

We tried to profile the body of the lambda expressions in
https://reviews.llvm.org/D153957. But as the original comments show,
it is indeed dangerous. After we tried to skip calculating the ODR
hash values recently, we have fall into this trap twice.

So in this patch, I choose to not profile the body of the lambda
expression. The signature of the lambda is still profiled.
---
 clang/include/clang/AST/DeclBase.h            | 10 -----
 clang/include/clang/Serialization/ASTReader.h |  6 +++
 clang/lib/AST/Decl.cpp                        |  2 +-
 clang/lib/AST/DeclBase.cpp                    |  5 ---
 clang/lib/AST/StmtProfile.cpp                 | 22 +++++++++-
 clang/lib/Serialization/ASTReader.cpp         |  2 +-
 clang/lib/Serialization/ASTReaderDecl.cpp     |  8 ++--
 clang/lib/Serialization/ASTWriter.cpp         |  2 +-
 clang/lib/Serialization/ASTWriterDecl.cpp     |  8 ++--
 .../hashing-decls-in-exprs-from-gmf-2.cppm    | 44 +++++++++++++++++++
 10 files changed, 81 insertions(+), 28 deletions(-)
 create mode 100644 clang/test/Modules/hashing-decls-in-exprs-from-gmf-2.cppm

diff --git a/clang/include/clang/AST/DeclBase.h b/clang/include/clang/AST/DeclBase.h
index 2194d268fa86f..1079993f49694 100644
--- a/clang/include/clang/AST/DeclBase.h
+++ b/clang/include/clang/AST/DeclBase.h
@@ -672,16 +672,6 @@ class alignas(8) Decl {
   /// Whether this declaration comes from explicit global module.
   bool isFromExplicitGlobalModule() const;
 
-  /// Check if we should skip checking ODRHash for declaration \param D.
-  ///
-  /// The existing ODRHash mechanism seems to be not stable enough and
-  /// the false positive ODR violation reports are annoying and we rarely see
-  /// true ODR violation reports. Also we learned that MSVC disabled ODR checks
-  /// for declarations in GMF. So we try to disable ODR checks in the GMF to
-  /// get better user experiences before we make the ODR violation checks stable
-  /// enough.
-  bool shouldSkipCheckingODR() const;
-
   /// Return true if this declaration has an attribute which acts as
   /// definition of the entity, such as 'alias' or 'ifunc'.
   bool hasDefiningAttr() const;
diff --git a/clang/include/clang/Serialization/ASTReader.h b/clang/include/clang/Serialization/ASTReader.h
index e3fde887f99cb..43ee06c524b3a 100644
--- a/clang/include/clang/Serialization/ASTReader.h
+++ b/clang/include/clang/Serialization/ASTReader.h
@@ -2457,6 +2457,12 @@ class BitsUnpacker {
   uint32_t Value;
   uint32_t CurrentBitsIndex = ~0;
 };
+
+inline bool shouldSkipCheckingODR(const Decl *D) {
+  return D->getASTContext().getLangOpts().SkipODRCheckInGMF &&
+         D->isFromExplicitGlobalModule();
+}
+
 } // namespace clang
 
 #endif // LLVM_CLANG_SERIALIZATION_ASTREADER_H
diff --git a/clang/lib/AST/Decl.cpp b/clang/lib/AST/Decl.cpp
index 2b2d5a2663a18..33b6f8611f216 100644
--- a/clang/lib/AST/Decl.cpp
+++ b/clang/lib/AST/Decl.cpp
@@ -4534,7 +4534,7 @@ unsigned FunctionDecl::getODRHash() {
   }
 
   class ODRHash Hash;
-  Hash.AddFunctionDecl(this, /*SkipBody=*/shouldSkipCheckingODR());
+  Hash.AddFunctionDecl(this);
   setHasODRHash(true);
   ODRHash = Hash.CalculateHash();
   return ODRHash;
diff --git a/clang/lib/AST/DeclBase.cpp b/clang/lib/AST/DeclBase.cpp
index 66a727d9dd0c3..434926324c96c 100644
--- a/clang/lib/AST/DeclBase.cpp
+++ b/clang/lib/AST/DeclBase.cpp
@@ -1106,11 +1106,6 @@ bool Decl::isFromExplicitGlobalModule() const {
   return getOwningModule() && getOwningModule()->isExplicitGlobalModule();
 }
 
-bool Decl::shouldSkipCheckingODR() const {
-  return getASTContext().getLangOpts().SkipODRCheckInGMF &&
-         isFromExplicitGlobalModule();
-}
-
 static Decl::Kind getKind(const Decl *D) { return D->getKind(); }
 static Decl::Kind getKind(const DeclContext *DC) { return DC->getDeclKind(); }
 
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index d2aac1e640380..789e4634bd293 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2071,13 +2071,31 @@ StmtProfiler::VisitLambdaExpr(const LambdaExpr *S) {
   }
 
   CXXRecordDecl *Lambda = S->getLambdaClass();
-  ID.AddInteger(Lambda->getODRHash());
-
   for (const auto &Capture : Lambda->captures()) {
     ID.AddInteger(Capture.getCaptureKind());
     if (Capture.capturesVariable())
       VisitDecl(Capture.getCapturedVar());
   }
+
+  // Profiling the body of the lambda may be dangerous during deserialization.
+  // So we'd like only to profile the signature here.
+  ODRHash Hasher;
+  // FIXME: We can't get the operator call easily by
+  // `CXXRecordDecl::getLambdaCallOperator()` if we're in deserialization.
+  // So we have to do something raw here.
+  for (auto *SubDecl : Lambda->decls()) {
+    FunctionDecl *Call = nullptr;
+    if (auto *FTD = dyn_cast<FunctionTemplateDecl>(SubDecl))
+      Call = FTD->getTemplatedDecl();
+    else if (auto *FD = dyn_cast<FunctionDecl>(SubDecl))
+      Call = FD;
+
+    if (!Call)
+      continue;
+
+    Hasher.AddFunctionDecl(Call, /*SkipBody=*/true);
+  }
+  ID.AddInteger(Hasher.CalculateHash());
 }
 
 void
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index feb60bc54413a..f47d540ea4b86 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -9785,7 +9785,7 @@ void ASTReader::finishPendingActions() {
             !NonConstDefn->isLateTemplateParsed() &&
             // We only perform ODR checks for decls not in the explicit
             // global module fragment.
-            !FD->shouldSkipCheckingODR() &&
+            !shouldSkipCheckingODR(FD) &&
             FD->getODRHash() != NonConstDefn->getODRHash()) {
           if (!isa<CXXMethodDecl>(FD)) {
             PendingFunctionOdrMergeFailures[FD].push_back(NonConstDefn);
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index e4b6a75c118ba..74d40f7da34ca 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -826,7 +826,7 @@ void ASTDeclReader::VisitEnumDecl(EnumDecl *ED) {
       Reader.mergeDefinitionVisibility(OldDef, ED);
       // We don't want to check the ODR hash value for declarations from global
       // module fragment.
-      if (!ED->shouldSkipCheckingODR() &&
+      if (!shouldSkipCheckingODR(ED) &&
           OldDef->getODRHash() != ED->getODRHash())
         Reader.PendingEnumOdrMergeFailures[OldDef].push_back(ED);
     } else {
@@ -868,7 +868,7 @@ void ASTDeclReader::VisitRecordDecl(RecordDecl *RD) {
   VisitRecordDeclImpl(RD);
   // We should only reach here if we're in C/Objective-C. There is no
   // global module fragment.
-  assert(!RD->shouldSkipCheckingODR());
+  assert(!shouldSkipCheckingODR(RD));
   RD->setODRHash(Record.readInt());
 
   // Maintain the invariant of a redeclaration chain containing only
@@ -2155,7 +2155,7 @@ void ASTDeclReader::MergeDefinitionData(
   }
 
   // We don't want to check ODR for decls in the global module fragment.
-  if (MergeDD.Definition->shouldSkipCheckingODR())
+  if (shouldSkipCheckingODR(MergeDD.Definition))
     return;
 
   if (D->getODRHash() != MergeDD.ODRHash) {
@@ -3530,7 +3530,7 @@ ASTDeclReader::FindExistingResult ASTDeclReader::findExisting(NamedDecl *D) {
   // same template specialization into the same CXXRecordDecl.
   auto MergedDCIt = Reader.MergedDeclContexts.find(D->getLexicalDeclContext());
   if (MergedDCIt != Reader.MergedDeclContexts.end() &&
-      !D->shouldSkipCheckingODR() && MergedDCIt->second == D->getDeclContext())
+      !shouldSkipCheckingODR(D) && MergedDCIt->second == D->getDeclContext())
     Reader.PendingOdrMergeChecks.push_back(D);
 
   return FindExistingResult(Reader, D, /*Existing=*/nullptr,
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 85b7fd5535a1b..ce6fa1feb1eeb 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -6188,7 +6188,7 @@ void ASTRecordWriter::AddCXXDefinitionData(const CXXRecordDecl *D) {
 
   BitsPacker DefinitionBits;
 
-  bool ShouldSkipCheckingODR = D->shouldSkipCheckingODR();
+  bool ShouldSkipCheckingODR = shouldSkipCheckingODR(D);
   DefinitionBits.addBit(ShouldSkipCheckingODR);
 
 #define FIELD(Name, Width, Merge)                                              \
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index 276b6257f1d84..d0d49bcdf991a 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -526,7 +526,7 @@ void ASTDeclWriter::VisitEnumDecl(EnumDecl *D) {
   BitsPacker EnumDeclBits;
   EnumDeclBits.addBits(D->getNumPositiveBits(), /*BitWidth=*/8);
   EnumDeclBits.addBits(D->getNumNegativeBits(), /*BitWidth=*/8);
-  bool ShouldSkipCheckingODR = D->shouldSkipCheckingODR();
+  bool ShouldSkipCheckingODR = shouldSkipCheckingODR(D);
   EnumDeclBits.addBit(ShouldSkipCheckingODR);
   EnumDeclBits.addBit(D->isScoped());
   EnumDeclBits.addBit(D->isScopedUsingClassTag());
@@ -552,7 +552,7 @@ void ASTDeclWriter::VisitEnumDecl(EnumDecl *D) {
       !D->isTopLevelDeclInObjCContainer() &&
       !CXXRecordDecl::classofKind(D->getKind()) &&
       !D->getIntegerTypeSourceInfo() && !D->getMemberSpecializationInfo() &&
-      !needsAnonymousDeclarationNumber(D) && !D->shouldSkipCheckingODR() &&
+      !needsAnonymousDeclarationNumber(D) && !shouldSkipCheckingODR(D) &&
       D->getDeclName().getNameKind() == DeclarationName::Identifier)
     AbbrevToUse = Writer.getDeclEnumAbbrev();
 
@@ -718,7 +718,7 @@ void ASTDeclWriter::VisitFunctionDecl(FunctionDecl *D) {
   // FIXME: stable encoding
   FunctionDeclBits.addBits(llvm::to_underlying(D->getLinkageInternal()), 3);
   FunctionDeclBits.addBits((uint32_t)D->getStorageClass(), /*BitWidth=*/3);
-  bool ShouldSkipCheckingODR = D->shouldSkipCheckingODR();
+  bool ShouldSkipCheckingODR = shouldSkipCheckingODR(D);
   FunctionDeclBits.addBit(ShouldSkipCheckingODR);
   FunctionDeclBits.addBit(D->isInlineSpecified());
   FunctionDeclBits.addBit(D->isInlined());
@@ -1559,7 +1559,7 @@ void ASTDeclWriter::VisitCXXMethodDecl(CXXMethodDecl *D) {
       D->getFirstDecl() == D->getMostRecentDecl() && !D->isInvalidDecl() &&
       !D->hasAttrs() && !D->isTopLevelDeclInObjCContainer() &&
       D->getDeclName().getNameKind() == DeclarationName::Identifier &&
-      !D->shouldSkipCheckingODR() && !D->hasExtInfo() &&
+      !shouldSkipCheckingODR(D) && !D->hasExtInfo() &&
       !D->isExplicitlyDefaulted()) {
     if (D->getTemplatedKind() == FunctionDecl::TK_NonTemplate ||
         D->getTemplatedKind() == FunctionDecl::TK_FunctionTemplate ||
diff --git a/clang/test/Modules/hashing-decls-in-exprs-from-gmf-2.cppm b/clang/test/Modules/hashing-decls-in-exprs-from-gmf-2.cppm
new file mode 100644
index 0000000000000..66143102cb9e4
--- /dev/null
+++ b/clang/test/Modules/hashing-decls-in-exprs-from-gmf-2.cppm
@@ -0,0 +1,44 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 -fskip-odr-check-in-gmf %t/A.cppm -emit-module-interface -o %t/A.pcm
+// RUN: %clang_cc1 -std=c++20 -fskip-odr-check-in-gmf %t/test.cpp -fprebuilt-module-path=%t -fsyntax-only -verify
+
+//--- header.h
+#pragma once
+template <class _Tp>
+class Optional {};
+
+template <class _Tp>
+concept C = requires(const _Tp& __t) {
+    []<class _Up>(const Optional<_Up>&) {}(__t);
+};
+
+//--- func.h
+#include "header.h"
+template <C T>
+void func() {}
+
+//--- test_func.h
+#include "func.h"
+
+inline void test_func() {
+    func<Optional<int>>();
+}
+
+//--- A.cppm
+module;
+#include "header.h"
+#include "test_func.h"
+export module A;
+export using ::test_func;
+
+//--- test.cpp
+// expected-no-diagnostics
+import A;
+#include "test_func.h"
+
+void test() {
+    test_func();
+}

From a29e85d6626aae7dba44c66ee5b703d3ab5de41c Mon Sep 17 00:00:00 2001
From: Aiden Grossman <aidengrossman@google.com>
Date: Tue, 16 Apr 2024 07:41:42 +0000
Subject: [PATCH 534/886] [llvm-exegesis] Change preprocessor directives for
 getCurrentTID

This patch changes the preprocessor directives surrounding
getCurrentTID, particularly moving it out of the block that is only
defined when not building for Android. The getCurrentTID function is
called in places that only require Linux definitions, so this function
should have the same preprocessor scoping around it to prevent link time
failures.
---
 llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp b/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
index 0a947f6e206fe..4699fbbea5def 100644
--- a/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
+++ b/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
@@ -22,7 +22,7 @@
 namespace llvm {
 namespace exegesis {
 
-#if defined(__linux__) && !defined(__ANDROID__)
+#if defined(__linux__)
 
 long SubprocessMemory::getCurrentTID() {
   // We're using the raw syscall here rather than the gettid() function provided
@@ -31,6 +31,8 @@ long SubprocessMemory::getCurrentTID() {
   return syscall(SYS_gettid);
 }
 
+#if !defined(__ANDROID__)
+
 Error SubprocessMemory::initializeSubprocessMemory(pid_t ProcessID) {
   // Add the PID to the shared memory name so that if we're running multiple
   // processes at the same time, they won't interfere with each other.
@@ -157,7 +159,8 @@ Expected<int> SubprocessMemory::setupAuxiliaryMemoryInSubprocess(
 
 SubprocessMemory::~SubprocessMemory() {}
 
-#endif // defined(__linux__) && !defined(__ANDROID__)
+#endif // !defined(__ANDROID__)
+#endif // defined(__linux__)
 
 } // namespace exegesis
 } // namespace llvm

From d34a2c2adb2a4f1dc262c5756d3725caa4ea2571 Mon Sep 17 00:00:00 2001
From: Pengcheng Wang <wangpengcheng.pp@bytedance.com>
Date: Tue, 16 Apr 2024 15:55:14 +0800
Subject: [PATCH 535/886] [RISCV] Make more vector pseudos commutable

This PR includes:
* vadd.vv/vand.vv/vor.vv/vxor.vv
* vmseq.vv/vmsne.vv
* vmin.vv/vminu.vv/vmax.vv/vmaxu.vv
* vmul.vv/vmulh.vv/vmulhu.vv
* vwadd.vv/vwaddu.vv
* vwmul.vv/vwmulu
* vwmacc.vv/vwmaccu.vv
* vadc.vvm

There is no test change, I may add it later.

Fixes part of #64422

Reviewers: michaelmaitland, preames, lukel97, topperc, asb

Reviewed By: topperc, lukel97

Pull Request: https://github.com/llvm/llvm-project/pull/88379
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp      | 66 +++++++++++++
 .../Target/RISCV/RISCVInstrInfoVPseudos.td    | 92 ++++++++++---------
 2 files changed, 116 insertions(+), 42 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index b0fda040519a5..668062c8d33f6 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -2718,6 +2718,50 @@ std::string RISCVInstrInfo::createMIROperandComment(
   return Comment;
 }
 
+// clang-format off
+#define CASE_RVV_OPCODE_UNMASK_LMUL(OP, LMUL)                                 \
+  RISCV::Pseudo##OP##_##LMUL
+
+#define CASE_RVV_OPCODE_MASK_LMUL(OP, LMUL)                                   \
+  RISCV::Pseudo##OP##_##LMUL##_MASK
+
+#define CASE_RVV_OPCODE_LMUL(OP, LMUL)                                        \
+  CASE_RVV_OPCODE_UNMASK_LMUL(OP, LMUL):                                      \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, LMUL)
+
+#define CASE_RVV_OPCODE_UNMASK_WIDEN(OP)                                      \
+  CASE_RVV_OPCODE_UNMASK_LMUL(OP, MF8):                                       \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, MF4):                                  \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, MF2):                                  \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, M1):                                   \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, M2):                                   \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, M4)
+
+#define CASE_RVV_OPCODE_UNMASK(OP)                                            \
+  CASE_RVV_OPCODE_UNMASK_WIDEN(OP):                                           \
+  case CASE_RVV_OPCODE_UNMASK_LMUL(OP, M8)
+
+#define CASE_RVV_OPCODE_MASK_WIDEN(OP)                                        \
+  CASE_RVV_OPCODE_MASK_LMUL(OP, MF8):                                         \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, MF4):                                    \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, MF2):                                    \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, M1):                                     \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, M2):                                     \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, M4)
+
+#define CASE_RVV_OPCODE_MASK(OP)                                              \
+  CASE_RVV_OPCODE_MASK_WIDEN(OP):                                             \
+  case CASE_RVV_OPCODE_MASK_LMUL(OP, M8)
+
+#define CASE_RVV_OPCODE_WIDEN(OP)                                             \
+  CASE_RVV_OPCODE_UNMASK_WIDEN(OP):                                           \
+  case CASE_RVV_OPCODE_MASK_WIDEN(OP)
+
+#define CASE_RVV_OPCODE(OP)                                                   \
+  CASE_RVV_OPCODE_UNMASK(OP):                                                 \
+  case CASE_RVV_OPCODE_MASK(OP)
+// clang-format on
+
 // clang-format off
 #define CASE_VMA_OPCODE_COMMON(OP, TYPE, LMUL)                                 \
   RISCV::PseudoV##OP##_##TYPE##_##LMUL
@@ -2798,6 +2842,28 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
   case RISCV::PseudoCCMOVGPR:
     // Operands 4 and 5 are commutable.
     return fixCommutedOpIndices(SrcOpIdx1, SrcOpIdx2, 4, 5);
+  case CASE_RVV_OPCODE(VADD_VV):
+  case CASE_RVV_OPCODE(VAND_VV):
+  case CASE_RVV_OPCODE(VOR_VV):
+  case CASE_RVV_OPCODE(VXOR_VV):
+  case CASE_RVV_OPCODE_MASK(VMSEQ_VV):
+  case CASE_RVV_OPCODE_MASK(VMSNE_VV):
+  case CASE_RVV_OPCODE(VMIN_VV):
+  case CASE_RVV_OPCODE(VMINU_VV):
+  case CASE_RVV_OPCODE(VMAX_VV):
+  case CASE_RVV_OPCODE(VMAXU_VV):
+  case CASE_RVV_OPCODE(VMUL_VV):
+  case CASE_RVV_OPCODE(VMULH_VV):
+  case CASE_RVV_OPCODE(VMULHU_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWADD_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWADDU_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWMUL_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWMULU_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWMACC_VV):
+  case CASE_RVV_OPCODE_WIDEN(VWMACCU_VV):
+  case CASE_RVV_OPCODE_UNMASK(VADC_VVM):
+    // Operands 2 and 3 are commutable.
+    return fixCommutedOpIndices(SrcOpIdx1, SrcOpIdx2, 2, 3);
   case CASE_VFMA_SPLATS(FMADD):
   case CASE_VFMA_SPLATS(FMSUB):
   case CASE_VFMA_SPLATS(FMACC):
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
index ad1821d57256b..435cd7f84c612 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -2127,8 +2127,9 @@ multiclass VPseudoBinary<VReg RetClass,
                          LMULInfo MInfo,
                          string Constraint = "",
                          int sew = 0,
-                         int TargetConstraintType = 1> {
-  let VLMul = MInfo.value, SEW=sew in {
+                         int TargetConstraintType = 1,
+                         bit Commutable = 0> {
+  let VLMul = MInfo.value, SEW=sew, isCommutable = Commutable in {
     defvar suffix = !if(sew, "_" # MInfo.MX # "_E" # sew, "_" # MInfo.MX);
     def suffix : VPseudoBinaryNoMaskTU<RetClass, Op1Class, Op2Class,
                                        Constraint, TargetConstraintType>;
@@ -2167,8 +2168,9 @@ multiclass VPseudoBinaryM<VReg RetClass,
                           DAGOperand Op2Class,
                           LMULInfo MInfo,
                           string Constraint = "",
-                          int TargetConstraintType = 1> {
-  let VLMul = MInfo.value in {
+                          int TargetConstraintType = 1,
+                          bit Commutable = 0> {
+  let VLMul = MInfo.value, isCommutable = Commutable in {
     def "_" # MInfo.MX : VPseudoBinaryMOutNoMask<RetClass, Op1Class, Op2Class,
                                                  Constraint, TargetConstraintType>;
     let ForceTailAgnostic = true in
@@ -2226,8 +2228,8 @@ multiclass VPseudoTiedBinaryRoundingMode<VReg RetClass,
 }
 
 
-multiclass VPseudoBinaryV_VV<LMULInfo m, string Constraint = "", int sew = 0> {
-  defm _VV : VPseudoBinary<m.vrclass, m.vrclass, m.vrclass, m, Constraint, sew>;
+multiclass VPseudoBinaryV_VV<LMULInfo m, string Constraint = "", int sew = 0, bit Commutable = 0> {
+  defm _VV : VPseudoBinary<m.vrclass, m.vrclass, m.vrclass, m, Constraint, sew, Commutable=Commutable>;
 }
 
 multiclass VPseudoBinaryV_VV_RM<LMULInfo m, string Constraint = ""> {
@@ -2331,9 +2333,10 @@ multiclass VPseudoVALU_MM<bit Commutable = 0> {
 // * The destination EEW is greater than the source EEW, the source EMUL is
 //   at least 1, and the overlap is in the highest-numbered part of the
 //   destination register group is legal. Otherwise, it is illegal.
-multiclass VPseudoBinaryW_VV<LMULInfo m> {
+multiclass VPseudoBinaryW_VV<LMULInfo m, bit Commutable = 0> {
   defm _VV : VPseudoBinary<m.wvrclass, m.vrclass, m.vrclass, m,
-                           "@earlyclobber $rd", TargetConstraintType=3>;
+                           "@earlyclobber $rd", TargetConstraintType=3,
+                           Commutable=Commutable>;
 }
 
 multiclass VPseudoBinaryW_VV_RM<LMULInfo m, int sew = 0> {
@@ -2453,7 +2456,9 @@ multiclass VPseudoBinaryV_VM<LMULInfo m, bit CarryOut = 0, bit CarryIn = 1,
                          m.vrclass, m.vrclass, m, CarryIn, Constraint, TargetConstraintType>;
 }
 
-multiclass VPseudoTiedBinaryV_VM<LMULInfo m, int TargetConstraintType = 1> {
+multiclass VPseudoTiedBinaryV_VM<LMULInfo m, int TargetConstraintType = 1,
+                                 bit Commutable = 0> {
+  let isCommutable = Commutable in
   def "_VVM" # "_" # m.MX:
     VPseudoTiedBinaryCarryIn<GetVRegNoV0<m.vrclass>.R,
                              m.vrclass, m.vrclass, m, 1, "",
@@ -2667,9 +2672,11 @@ multiclass PseudoVEXT_VF8 {
 //  lowest-numbered part of the source register group".
 // With LMUL<=1 the source and dest occupy a single register so any overlap
 // is in the lowest-numbered part.
-multiclass VPseudoBinaryM_VV<LMULInfo m, int TargetConstraintType = 1> {
+multiclass VPseudoBinaryM_VV<LMULInfo m, int TargetConstraintType = 1,
+                             bit Commutable = 0> {
   defm _VV : VPseudoBinaryM<VR, m.vrclass, m.vrclass, m,
-                            !if(!ge(m.octuple, 16), "@earlyclobber $rd", ""), TargetConstraintType>;
+                            !if(!ge(m.octuple, 16), "@earlyclobber $rd", ""),
+                            TargetConstraintType, Commutable=Commutable>;
 }
 
 multiclass VPseudoBinaryM_VX<LMULInfo m, int TargetConstraintType = 1> {
@@ -2751,10 +2758,11 @@ multiclass VPseudoVSSHT_VV_VX_VI_RM<Operand ImmType = simm5, string Constraint =
   }
 }
 
-multiclass VPseudoVALU_VV_VX_VI<Operand ImmType = simm5, string Constraint = ""> {
+multiclass VPseudoVALU_VV_VX_VI<Operand ImmType = simm5, string Constraint = "",
+                                bit Commutable = 0> {
   foreach m = MxList in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryV_VV<m, Constraint>,
+    defm "" : VPseudoBinaryV_VV<m, Constraint, Commutable=Commutable>,
             SchedBinary<"WriteVIALUV", "ReadVIALUV", "ReadVIALUV", mx,
                         forceMergeOpRead=true>;
     defm "" : VPseudoBinaryV_VX<m, Constraint>,
@@ -2804,17 +2812,17 @@ multiclass VPseudoVAALU_VV_VX_RM {
 multiclass VPseudoVMINMAX_VV_VX {
   foreach m = MxList in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryV_VV<m>,
+    defm "" : VPseudoBinaryV_VV<m, Commutable=1>,
               SchedBinary<"WriteVIMinMaxV", "ReadVIMinMaxV", "ReadVIMinMaxV", mx>;
     defm "" : VPseudoBinaryV_VX<m>,
               SchedBinary<"WriteVIMinMaxX", "ReadVIMinMaxV", "ReadVIMinMaxX", mx>;
   }
 }
 
-multiclass VPseudoVMUL_VV_VX {
+multiclass VPseudoVMUL_VV_VX<bit Commutable = 0> {
   foreach m = MxList in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryV_VV<m>,
+    defm "" : VPseudoBinaryV_VV<m, Commutable=Commutable>,
               SchedBinary<"WriteVIMulV", "ReadVIMulV", "ReadVIMulV", mx>;
     defm "" : VPseudoBinaryV_VX<m>,
               SchedBinary<"WriteVIMulX", "ReadVIMulV", "ReadVIMulX", mx>;
@@ -2964,10 +2972,10 @@ multiclass VPseudoVALU_VX_VI<Operand ImmType = simm5> {
   }
 }
 
-multiclass VPseudoVWALU_VV_VX {
+multiclass VPseudoVWALU_VV_VX<bit Commutable = 0> {
   foreach m = MxListW in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryW_VV<m>,
+    defm "" : VPseudoBinaryW_VV<m, Commutable=Commutable>,
               SchedBinary<"WriteVIWALUV", "ReadVIWALUV", "ReadVIWALUV", mx,
                           forceMergeOpRead=true>;
     defm "" : VPseudoBinaryW_VX<m>, 
@@ -2976,10 +2984,10 @@ multiclass VPseudoVWALU_VV_VX {
   }
 }
 
-multiclass VPseudoVWMUL_VV_VX {
+multiclass VPseudoVWMUL_VV_VX<bit Commutable = 0> {
   foreach m = MxListW in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryW_VV<m>,
+    defm "" : VPseudoBinaryW_VV<m, Commutable=Commutable>,
               SchedBinary<"WriteVIWMulV", "ReadVIWMulV", "ReadVIWMulV", mx,
                           forceMergeOpRead=true>;
     defm "" : VPseudoBinaryW_VX<m>,
@@ -3074,7 +3082,7 @@ multiclass VPseudoVMRG_VM_XM_IM {
 multiclass VPseudoVCALU_VM_XM_IM {
   foreach m = MxList in {
     defvar mx = m.MX;
-    defm "" : VPseudoTiedBinaryV_VM<m>,
+    defm "" : VPseudoTiedBinaryV_VM<m, Commutable=1>,
               SchedBinary<"WriteVICALUV", "ReadVICALUV", "ReadVICALUV", mx,
                           forceMergeOpRead=true>;
     defm "" : VPseudoTiedBinaryV_XM<m>,
@@ -3287,10 +3295,10 @@ multiclass VPseudoTernaryV_VF_AAXA_RM<LMULInfo m, FPR_Info f,
                                                           sew, Commutable=1>;
 }
 
-multiclass VPseudoTernaryW_VV<LMULInfo m> {
+multiclass VPseudoTernaryW_VV<LMULInfo m, bit Commutable = 0> {
   defvar constraint = "@earlyclobber $rd";
   defm _VV : VPseudoTernaryWithPolicy<m.wvrclass, m.vrclass, m.vrclass, m,
-                                      constraint, /*Commutable*/ 0, TargetConstraintType=3>;
+                                      constraint, Commutable=Commutable, TargetConstraintType=3>;
 }
 
 multiclass VPseudoTernaryW_VV_RM<LMULInfo m, int sew = 0> {
@@ -3380,10 +3388,10 @@ multiclass VPseudoVSLD_VX_VI<Operand ImmType = simm5, string Constraint = ""> {
   }
 }
 
-multiclass VPseudoVWMAC_VV_VX {
+multiclass VPseudoVWMAC_VV_VX<bit Commutable = 0> {
   foreach m = MxListW in {
     defvar mx = m.MX;
-    defm "" : VPseudoTernaryW_VV<m>,
+    defm "" : VPseudoTernaryW_VV<m, Commutable=Commutable>,
               SchedTernary<"WriteVIWMulAddV", "ReadVIWMulAddV", "ReadVIWMulAddV",
                            "ReadVIWMulAddV", mx>;
     defm "" : VPseudoTernaryW_VX<m>,
@@ -3436,10 +3444,10 @@ multiclass VPseudoVWMAC_VV_VF_BF_RM {
   }
 }
 
-multiclass VPseudoVCMPM_VV_VX_VI {
+multiclass VPseudoVCMPM_VV_VX_VI<bit Commutable = 0> {
   foreach m = MxList in {
     defvar mx = m.MX;
-    defm "" : VPseudoBinaryM_VV<m, TargetConstraintType=2>,
+    defm "" : VPseudoBinaryM_VV<m, TargetConstraintType=2, Commutable=Commutable>,
               SchedBinary<"WriteVICmpV", "ReadVICmpV", "ReadVICmpV", mx>;
     defm "" : VPseudoBinaryM_VX<m, TargetConstraintType=2>,
               SchedBinary<"WriteVICmpX", "ReadVICmpV", "ReadVICmpX", mx>;
@@ -6248,7 +6256,7 @@ defm PseudoVLSEG : VPseudoUSSegLoadFF;
 //===----------------------------------------------------------------------===//
 // 11.1. Vector Single-Width Integer Add and Subtract
 //===----------------------------------------------------------------------===//
-defm PseudoVADD   : VPseudoVALU_VV_VX_VI;
+defm PseudoVADD   : VPseudoVALU_VV_VX_VI<Commutable=1>;
 defm PseudoVSUB   : VPseudoVALU_VV_VX;
 defm PseudoVRSUB  : VPseudoVALU_VX_VI;
 
@@ -6313,9 +6321,9 @@ foreach vti = AllIntegerVectors in {
 //===----------------------------------------------------------------------===//
 // 11.2. Vector Widening Integer Add/Subtract
 //===----------------------------------------------------------------------===//
-defm PseudoVWADDU : VPseudoVWALU_VV_VX;
+defm PseudoVWADDU : VPseudoVWALU_VV_VX<Commutable=1>;
 defm PseudoVWSUBU : VPseudoVWALU_VV_VX;
-defm PseudoVWADD  : VPseudoVWALU_VV_VX;
+defm PseudoVWADD  : VPseudoVWALU_VV_VX<Commutable=1>;
 defm PseudoVWSUB  : VPseudoVWALU_VV_VX;
 defm PseudoVWADDU : VPseudoVWALU_WV_WX;
 defm PseudoVWSUBU : VPseudoVWALU_WV_WX;
@@ -6346,9 +6354,9 @@ defm PseudoVMSBC : VPseudoVCALUM_V_X<"@earlyclobber $rd">;
 //===----------------------------------------------------------------------===//
 // 11.5. Vector Bitwise Logical Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVAND : VPseudoVALU_VV_VX_VI;
-defm PseudoVOR  : VPseudoVALU_VV_VX_VI;
-defm PseudoVXOR : VPseudoVALU_VV_VX_VI;
+defm PseudoVAND : VPseudoVALU_VV_VX_VI<Commutable=1>;
+defm PseudoVOR  : VPseudoVALU_VV_VX_VI<Commutable=1>;
+defm PseudoVXOR : VPseudoVALU_VV_VX_VI<Commutable=1>;
 
 //===----------------------------------------------------------------------===//
 // 11.6. Vector Single-Width Bit Shift Instructions
@@ -6366,8 +6374,8 @@ defm PseudoVNSRA : VPseudoVNSHT_WV_WX_WI;
 //===----------------------------------------------------------------------===//
 // 11.8. Vector Integer Comparison Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVMSEQ  : VPseudoVCMPM_VV_VX_VI;
-defm PseudoVMSNE  : VPseudoVCMPM_VV_VX_VI;
+defm PseudoVMSEQ  : VPseudoVCMPM_VV_VX_VI<Commutable=1>;
+defm PseudoVMSNE  : VPseudoVCMPM_VV_VX_VI<Commutable=1>;
 defm PseudoVMSLTU : VPseudoVCMPM_VV_VX;
 defm PseudoVMSLT  : VPseudoVCMPM_VV_VX;
 defm PseudoVMSLEU : VPseudoVCMPM_VV_VX_VI;
@@ -6386,9 +6394,9 @@ defm PseudoVMAX  : VPseudoVMINMAX_VV_VX;
 //===----------------------------------------------------------------------===//
 // 11.10. Vector Single-Width Integer Multiply Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVMUL    : VPseudoVMUL_VV_VX;
-defm PseudoVMULH   : VPseudoVMUL_VV_VX;
-defm PseudoVMULHU  : VPseudoVMUL_VV_VX;
+defm PseudoVMUL    : VPseudoVMUL_VV_VX<Commutable=1>;
+defm PseudoVMULH   : VPseudoVMUL_VV_VX<Commutable=1>;
+defm PseudoVMULHU  : VPseudoVMUL_VV_VX<Commutable=1>;
 defm PseudoVMULHSU : VPseudoVMUL_VV_VX;
 
 //===----------------------------------------------------------------------===//
@@ -6402,8 +6410,8 @@ defm PseudoVREM  : VPseudoVDIV_VV_VX;
 //===----------------------------------------------------------------------===//
 // 11.12. Vector Widening Integer Multiply Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVWMUL   : VPseudoVWMUL_VV_VX;
-defm PseudoVWMULU  : VPseudoVWMUL_VV_VX;
+defm PseudoVWMUL   : VPseudoVWMUL_VV_VX<Commutable=1>;
+defm PseudoVWMULU  : VPseudoVWMUL_VV_VX<Commutable=1>;
 defm PseudoVWMULSU : VPseudoVWMUL_VV_VX;
 
 //===----------------------------------------------------------------------===//
@@ -6417,8 +6425,8 @@ defm PseudoVNMSUB : VPseudoVMAC_VV_VX_AAXA;
 //===----------------------------------------------------------------------===//
 // 11.14. Vector Widening Integer Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVWMACCU  : VPseudoVWMAC_VV_VX;
-defm PseudoVWMACC   : VPseudoVWMAC_VV_VX;
+defm PseudoVWMACCU  : VPseudoVWMAC_VV_VX<Commutable=1>;
+defm PseudoVWMACC   : VPseudoVWMAC_VV_VX<Commutable=1>;
 defm PseudoVWMACCSU : VPseudoVWMAC_VV_VX;
 defm PseudoVWMACCUS : VPseudoVWMAC_VX;
 

From 5b811562a520a8a3cd164897f24dee7da3115bbe Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Tue, 16 Apr 2024 04:33:39 -0400
Subject: [PATCH 536/886] [libc++] Rename __cpu_traits functions (#88741)

Functions inside __cpu_traits were needlessly prefixed with __parallel,
which doesn't serve a real purpose anymore now that they are inside a
traits class.
---
 .../pstl_backends/cpu_backends/any_of.h       |  2 +-
 .../pstl_backends/cpu_backends/fill.h         |  2 +-
 .../pstl_backends/cpu_backends/find_if.h      |  2 +-
 .../pstl_backends/cpu_backends/for_each.h     |  2 +-
 .../pstl_backends/cpu_backends/libdispatch.h  | 24 ++++++++---------
 .../pstl_backends/cpu_backends/merge.h        |  2 +-
 .../pstl_backends/cpu_backends/serial.h       | 24 ++++++++---------
 .../pstl_backends/cpu_backends/stable_sort.h  |  2 +-
 .../pstl_backends/cpu_backends/thread.h       | 24 ++++++++---------
 .../pstl_backends/cpu_backends/transform.h    |  4 +--
 .../cpu_backends/transform_reduce.h           |  4 +--
 libcxx/include/__pstl/cpu_algos/cpu_traits.h  | 27 +++++++++----------
 12 files changed, 59 insertions(+), 60 deletions(-)

diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
index be5e54f3fa5c8..3755d288047e0 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
@@ -34,7 +34,7 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 template <class _Backend, class _Index, class _Brick>
 _LIBCPP_HIDE_FROM_ABI optional<bool> __parallel_or(_Index __first, _Index __last, _Brick __f) {
   std::atomic<bool> __found(false);
-  auto __ret = __pstl::__cpu_traits<_Backend>::__parallel_for(__first, __last, [__f, &__found](_Index __i, _Index __j) {
+  auto __ret = __pstl::__cpu_traits<_Backend>::__for_each(__first, __last, [__f, &__found](_Index __i, _Index __j) {
     if (!__found.load(std::memory_order_relaxed) && __f(__i, __j)) {
       __found.store(true, std::memory_order_relaxed);
       __pstl::__cpu_traits<_Backend>::__cancel_execution();
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
index 49a32f6c5ce55..0c20bdff62675 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
@@ -40,7 +40,7 @@ _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_fill(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__for_each(
         __first, __last, [&__value](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           [[maybe_unused]] auto __res = std::__pstl_fill<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __value);
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
index 11a5668bf25af..626293faef692 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
@@ -42,7 +42,7 @@ __parallel_find(_Index __first, _Index __last, _Brick __f, _Compare __comp, bool
   _DifferenceType __initial_dist = __b_first ? __n : -1;
   std::atomic<_DifferenceType> __extremum(__initial_dist);
   // TODO: find out what is better here: parallel_for or parallel_reduce
-  auto __res = __pstl::__cpu_traits<_Backend>::__parallel_for(
+  auto __res = __pstl::__cpu_traits<_Backend>::__for_each(
       __first, __last, [__comp, __f, __first, &__extremum](_Index __i, _Index __j) {
         // See "Reducing Contention Through Priority Updates", PPoPP '13, for discussion of
         // why using a shared variable scales fairly well in this situation.
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
index 1667ec0f0c4f4..d637084e151d8 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
@@ -40,7 +40,7 @@ _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_for_each(__cpu_backend_tag, _ForwardIterator __first, _ForwardIterator __last, _Functor __func) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__for_each(
         __first, __last, [__func](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           [[maybe_unused]] auto __res = std::__pstl_for_each<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __func);
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
index 8757f24968037..17faadf55dd4f 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
@@ -85,7 +85,7 @@ template <>
 struct __cpu_traits<__libdispatch_backend_tag> {
   template <class _RandomAccessIterator, class _Functor>
   _LIBCPP_HIDE_FROM_ABI static optional<__empty>
-  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func) {
+  __for_each(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func) {
     return __libdispatch::__dispatch_parallel_for(
         __libdispatch::__partition_chunks(__last - __first), std::move(__first), std::move(__func));
   }
@@ -105,14 +105,14 @@ struct __cpu_traits<__libdispatch_backend_tag> {
             typename _RandomAccessIterator3,
             typename _Compare,
             typename _LeafMerge>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
-      _RandomAccessIterator1 __first1,
-      _RandomAccessIterator1 __last1,
-      _RandomAccessIterator2 __first2,
-      _RandomAccessIterator2 __last2,
-      _RandomAccessIterator3 __result,
-      _Compare __comp,
-      _LeafMerge __leaf_merge) noexcept {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __merge(_RandomAccessIterator1 __first1,
+          _RandomAccessIterator1 __last1,
+          _RandomAccessIterator2 __first2,
+          _RandomAccessIterator2 __last2,
+          _RandomAccessIterator3 __result,
+          _Compare __comp,
+          _LeafMerge __leaf_merge) noexcept {
     __libdispatch::__chunk_partitions __partitions =
         __libdispatch::__partition_chunks(std::max<ptrdiff_t>(__last1 - __first1, __last2 - __first2));
 
@@ -201,7 +201,7 @@ struct __cpu_traits<__libdispatch_backend_tag> {
   }
 
   template <class _RandomAccessIterator, class _Transform, class _Value, class _Combiner, class _Reduction>
-  _LIBCPP_HIDE_FROM_ABI static optional<_Value> __parallel_transform_reduce(
+  _LIBCPP_HIDE_FROM_ABI static optional<_Value> __transform_reduce(
       _RandomAccessIterator __first,
       _RandomAccessIterator __last,
       _Transform __transform,
@@ -248,8 +248,8 @@ struct __cpu_traits<__libdispatch_backend_tag> {
   }
 
   template <class _RandomAccessIterator, class _Comp, class _LeafSort>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
-      _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp, _LeafSort __leaf_sort) {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp, _LeafSort __leaf_sort) {
     const auto __size = __last - __first;
     auto __partitions = __libdispatch::__partition_chunks(__size);
 
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
index d034447904872..c93f4051c9d09 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
@@ -46,7 +46,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_merge(
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_merge(
+    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__merge(
         __first1,
         __last1,
         __first2,
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
index c3d2905daed17..7544619a8eefd 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
@@ -35,20 +35,20 @@ template <>
 struct __cpu_traits<__serial_backend_tag> {
   template <class _RandomAccessIterator, class _Fp>
   _LIBCPP_HIDE_FROM_ABI static optional<__empty>
-  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
+  __for_each(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
     __f(__first, __last);
     return __empty{};
   }
 
   template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
   _LIBCPP_HIDE_FROM_ABI static optional<_Tp>
-  __parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
+  __transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
     return __reduce(std::move(__first), std::move(__last), std::move(__init));
   }
 
   template <class _RandomAccessIterator, class _Compare, class _LeafSort>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
-      _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
     __leaf_sort(__first, __last, __comp);
     return __empty{};
   }
@@ -60,14 +60,14 @@ struct __cpu_traits<__serial_backend_tag> {
             class _RandomAccessIterator3,
             class _Compare,
             class _LeafMerge>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
-      _RandomAccessIterator1 __first1,
-      _RandomAccessIterator1 __last1,
-      _RandomAccessIterator2 __first2,
-      _RandomAccessIterator2 __last2,
-      _RandomAccessIterator3 __outit,
-      _Compare __comp,
-      _LeafMerge __leaf_merge) {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __merge(_RandomAccessIterator1 __first1,
+          _RandomAccessIterator1 __last1,
+          _RandomAccessIterator2 __first2,
+          _RandomAccessIterator2 __last2,
+          _RandomAccessIterator3 __outit,
+          _Compare __comp,
+          _LeafMerge __leaf_merge) {
     __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
     return __empty{};
   }
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
index ebfa0fc69147d..8c60cf897ff86 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
@@ -29,7 +29,7 @@ template <class _ExecutionPolicy, class _RandomAccessIterator, class _Comp>
 _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __pstl_stable_sort(__cpu_backend_tag, _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy>) {
-    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_stable_sort(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__stable_sort(
         __first, __last, __comp, [](_RandomAccessIterator __g_first, _RandomAccessIterator __g_last, _Comp __g_comp) {
           std::stable_sort(__g_first, __g_last, __g_comp);
         });
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
index 8d1cb221c3d82..2acf912264a00 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
@@ -38,20 +38,20 @@ template <>
 struct __cpu_traits<__std_thread_backend_tag> {
   template <class _RandomAccessIterator, class _Fp>
   _LIBCPP_HIDE_FROM_ABI static optional<__empty>
-  __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
+  __for_each(_RandomAccessIterator __first, _RandomAccessIterator __last, _Fp __f) {
     __f(__first, __last);
     return __empty{};
   }
 
   template <class _Index, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduce>
   _LIBCPP_HIDE_FROM_ABI static optional<_Tp>
-  __parallel_transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
+  __transform_reduce(_Index __first, _Index __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduce __reduce) {
     return __reduce(std::move(__first), std::move(__last), std::move(__init));
   }
 
   template <class _RandomAccessIterator, class _Compare, class _LeafSort>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_stable_sort(
-      _RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __stable_sort(_RandomAccessIterator __first, _RandomAccessIterator __last, _Compare __comp, _LeafSort __leaf_sort) {
     __leaf_sort(__first, __last, __comp);
     return __empty{};
   }
@@ -63,14 +63,14 @@ struct __cpu_traits<__std_thread_backend_tag> {
             class _RandomAccessIterator3,
             class _Compare,
             class _LeafMerge>
-  _LIBCPP_HIDE_FROM_ABI static optional<__empty> __parallel_merge(
-      _RandomAccessIterator1 __first1,
-      _RandomAccessIterator1 __last1,
-      _RandomAccessIterator2 __first2,
-      _RandomAccessIterator2 __last2,
-      _RandomAccessIterator3 __outit,
-      _Compare __comp,
-      _LeafMerge __leaf_merge) {
+  _LIBCPP_HIDE_FROM_ABI static optional<__empty>
+  __merge(_RandomAccessIterator1 __first1,
+          _RandomAccessIterator1 __last1,
+          _RandomAccessIterator2 __first2,
+          _RandomAccessIterator2 __last2,
+          _RandomAccessIterator3 __outit,
+          _Compare __comp,
+          _LeafMerge __leaf_merge) {
     __leaf_merge(__first1, __last1, __first2, __last2, __outit, __comp);
     return __empty{};
   }
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
index d4c383997a67a..4b9b296866832 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
@@ -50,7 +50,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_transform(
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
+    __pstl::__cpu_traits<__cpu_backend_tag>::__for_each(
         __first, __last, [__op, __first, __result](_ForwardIterator __brick_first, _ForwardIterator __brick_last) {
           auto __res = std::__pstl_transform<__remove_parallel_policy_t<_ExecutionPolicy>>(
               __cpu_backend_tag{}, __brick_first, __brick_last, __result + (__brick_first - __first), __op);
@@ -98,7 +98,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_ForwardOutIterator> __pstl_transform(
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardOutIterator>::value) {
-    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_for(
+    auto __res = __pstl::__cpu_traits<__cpu_backend_tag>::__for_each(
         __first1,
         __last1,
         [__op, __first1, __first2, __result](_ForwardIterator1 __brick_first, _ForwardIterator1 __brick_last) {
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
index 956c7d6a88ce2..c074eea9861c1 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
@@ -120,7 +120,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator1>::value &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator2>::value) {
-    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_transform_reduce(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__transform_reduce(
         __first1,
         std::move(__last1),
         [__first1, __first2, __transform](_ForwardIterator1 __iter) {
@@ -167,7 +167,7 @@ _LIBCPP_HIDE_FROM_ABI optional<_Tp> __pstl_transform_reduce(
     _UnaryOperation __transform) {
   if constexpr (__is_parallel_execution_policy_v<_ExecutionPolicy> &&
                 __has_random_access_iterator_category_or_concept<_ForwardIterator>::value) {
-    return __pstl::__cpu_traits<__cpu_backend_tag>::__parallel_transform_reduce(
+    return __pstl::__cpu_traits<__cpu_backend_tag>::__transform_reduce(
         std::move(__first),
         std::move(__last),
         [__transform](_ForwardIterator __iter) { return __transform(*__iter); },
diff --git a/libcxx/include/__pstl/cpu_algos/cpu_traits.h b/libcxx/include/__pstl/cpu_algos/cpu_traits.h
index 2f0db46e9be83..0483d6918fd01 100644
--- a/libcxx/include/__pstl/cpu_algos/cpu_traits.h
+++ b/libcxx/include/__pstl/cpu_algos/cpu_traits.h
@@ -32,31 +32,30 @@ namespace __pstl {
 // ================
 //
 //  template <class _RandomAccessIterator, class _Functor>
-//  optional<__empty> __parallel_for(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func);
+//  optional<__empty> __for_each(_RandomAccessIterator __first, _RandomAccessIterator __last, _Functor __func);
 //    - __func must take a subrange of [__first, __last) that should be executed in serial
 //
 //  template <class _Iterator, class _UnaryOp, class _Tp, class _BinaryOp, class _Reduction>
-//  optional<_Tp> __parallel_transform_reduce(_Iterator __first, _Iterator __last, _UnaryOp, _Tp __init, _BinaryOp,
-//  _Reduction);
+//  optional<_Tp> __transform_reduce(_Iterator __first, _Iterator __last, _UnaryOp, _Tp __init, _BinaryOp, _Reduction);
 //
 //  template <class _RandomAccessIterator1,
 //            class _RandomAccessIterator2,
 //            class _RandomAccessIterator3,
 //            class _Compare,
 //            class _LeafMerge>
-//  optional<_RandomAccessIterator3> __parallel_merge(_RandomAccessIterator1 __first1,
-//                                                    _RandomAccessIterator1 __last1,
-//                                                    _RandomAccessIterator2 __first2,
-//                                                    _RandomAccessIterator2 __last2,
-//                                                    _RandomAccessIterator3 __outit,
-//                                                    _Compare __comp,
-//                                                    _LeafMerge __leaf_merge);
+//  optional<_RandomAccessIterator3> __merge(_RandomAccessIterator1 __first1,
+//                                           _RandomAccessIterator1 __last1,
+//                                           _RandomAccessIterator2 __first2,
+//                                           _RandomAccessIterator2 __last2,
+//                                           _RandomAccessIterator3 __outit,
+//                                           _Compare __comp,
+//                                           _LeafMerge __leaf_merge);
 //
 //  template <class _RandomAccessIterator, class _Comp, class _LeafSort>
-//  optional<__empty> __parallel_stable_sort(_RandomAccessIterator __first,
-//                                           _RandomAccessIterator __last,
-//                                           _Comp __comp,
-//                                           _LeafSort __leaf_sort);
+//  optional<__empty> __stable_sort(_RandomAccessIterator __first,
+//                                  _RandomAccessIterator __last,
+//                                  _Comp __comp,
+//                                  _LeafSort __leaf_sort);
 //
 //   void __cancel_execution();
 //      Cancel the execution of other jobs - they aren't needed anymore. This is not a binding request,

From 66cf995da76b9da3cfdee2f29eff6ea4d84305ef Mon Sep 17 00:00:00 2001
From: Carlos Alberto Enciso <carlos.alberto.enciso@gmail.com>
Date: Tue, 16 Apr 2024 09:37:51 +0100
Subject: [PATCH 537/886] [IPSCCP] Variable not visible at Og: (#77901)

https://bugs.llvm.org/show_bug.cgi?id=51559
https://github.com/llvm/llvm-project/issues/50901

IPSCCP pass removes the global variable and does not create a constant
expression for the initializer value.

Extend test coverage to include:
- half, bfloat types.
- checks for undef (int32 and ptr).

There is no support for:
- fp128, x86_fp80, ppc_fp128  types.
https://github.com/llvm/llvm-project/issues/88102
---
 llvm/lib/Transforms/Utils/Local.cpp           |  8 ++--
 llvm/test/Transforms/SCCP/pr50901.ll          | 41 ++++++++++++++++++-
 llvm/unittests/Transforms/Utils/LocalTest.cpp | 21 ++++++----
 3 files changed, 57 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp
index 380bac9c61807..a42ef0c4e6ae9 100644
--- a/llvm/lib/Transforms/Utils/Local.cpp
+++ b/llvm/lib/Transforms/Utils/Local.cpp
@@ -3627,10 +3627,12 @@ DIExpression *llvm::getExpressionForConstant(DIBuilder &DIB, const Constant &C,
     return createIntegerExpression(C);
 
   auto *FP = dyn_cast<ConstantFP>(&C);
-  if (FP && (Ty.isFloatTy() || Ty.isDoubleTy())) {
+  if (FP && Ty.isFloatingPointTy() && Ty.getScalarSizeInBits() <= 64) {
     const APFloat &APF = FP->getValueAPF();
-    return DIB.createConstantValueExpression(
-        APF.bitcastToAPInt().getZExtValue());
+    APInt const &API = APF.bitcastToAPInt();
+    if (auto Temp = API.getZExtValue())
+      return DIB.createConstantValueExpression(static_cast<uint64_t>(Temp));
+    return DIB.createConstantValueExpression(*API.getRawData());
   }
 
   if (!Ty.isPointerTy())
diff --git a/llvm/test/Transforms/SCCP/pr50901.ll b/llvm/test/Transforms/SCCP/pr50901.ll
index 11d6bba6f6a93..d48d67532d88b 100644
--- a/llvm/test/Transforms/SCCP/pr50901.ll
+++ b/llvm/test/Transforms/SCCP/pr50901.ll
@@ -52,6 +52,16 @@
 ; CHECK:     = !DIGlobalVariableExpression(var: ![[DBG_FLOAT_UNDEF:.+]], expr: !DIExpression())
 ; CHECK-DAG: ![[DBG_FLOAT_UNDEF]]  = distinct !DIGlobalVariable(name: "g_float_undef"
 
+; CHECK: ![[G8:[0-9]+]] = !DIGlobalVariableExpression(var: ![[DBG8:[0-9]+]], expr: !DIExpression(DW_OP_constu, 22136, DW_OP_stack_value))
+; CHECK-DAG: ![[DBG8]] = distinct !DIGlobalVariable(name: "g_88", {{.*}}
+; CHECK: ![[G9:[0-9]+]] = !DIGlobalVariableExpression(var: ![[DBG9:[0-9]+]], expr: !DIExpression(DW_OP_constu, 23726, DW_OP_stack_value))
+; CHECK-DAG: ![[DBG9]] = distinct !DIGlobalVariable(name: "g_99", {{.*}}
+
+; CHECK-DAG: ![[DBGA:[0-9]+]] = distinct !DIGlobalVariable(name: "g_i32_undef"
+; CHECK-DAG: ![[GA:[0-9]+]] = !DIGlobalVariableExpression(var: ![[DBGA]], expr: !DIExpression())
+; CHECK-DAG: ![[DBGB:[0-9]+]] = distinct !DIGlobalVariable(name: "g_ptr_undef"
+; CHECK-DAG: ![[GB:[0-9]+]] = !DIGlobalVariableExpression(var: ![[DBGB]], expr: !DIExpression())
+
 @g_1 = dso_local global i32 -4, align 4, !dbg !0
 @g_2 = dso_local global float 0x4011C28F60000000, align 4, !dbg !8
 @g_3 = dso_local global i8 97, align 1, !dbg !10
@@ -59,6 +69,8 @@
 @g_5 = dso_local global i8 1, align 1, !dbg !16
 @g_6 = dso_local global ptr null, align 8, !dbg !19
 @g_7 = dso_local global ptr null, align 8, !dbg !23
+@g_8 = dso_local global half 0xH4321, align 4, !dbg !86
+@g_9 = dso_local global bfloat 0xR3F80, align 4, !dbg !90
 @_ZL4g_11 = internal global i32 -5, align 4, !dbg !25
 @_ZL4g_22 = internal global float 0x4016333340000000, align 4, !dbg !27
 @_ZL4g_33 = internal global i8 98, align 1, !dbg !29
@@ -67,6 +79,10 @@
 @_ZL4g_66 = internal global ptr null, align 8, !dbg !35
 @_ZL4g_77 = internal global ptr inttoptr (i64 70 to ptr), align 8, !dbg !37
 @g_float_undef = internal global float undef, align 4, !dbg !83
+@_ZL4g_88 = internal global half 0xH5678, align 4, !dbg !88
+@_ZL4g_99 = internal global bfloat 0xR5CAE, align 4, !dbg !92
+@g_i32_undef = internal global i32 undef, align 4, !dbg !95
+@g_ptr_undef = internal global ptr undef, align 8, !dbg !97
 
 define dso_local void @_Z3barv() !dbg !46 {
 entry:
@@ -88,6 +104,15 @@ entry:
   store ptr %6, ptr @g_7, align 8, !dbg !59
   %l = load float, ptr @g_float_undef, align 8, !dbg !59
   store float %l, ptr @g_2, align 8, !dbg !59
+  %7 = load half, ptr @_ZL4g_88, align 4, !dbg !59
+  store half %7, ptr @g_8, align 4, !dbg !59
+  %8 = load bfloat, ptr @_ZL4g_99, align 4, !dbg !59
+  store bfloat %8, ptr @g_9, align 4, !dbg !59
+  %9 = load i32, ptr @g_i32_undef, align 4, !dbg !59
+  store i32 %9, ptr @g_1, align 4, !dbg !59
+  %10 = load ptr, ptr @g_ptr_undef, align 8, !dbg !59
+  store ptr %10, ptr @g_6, align 8, !dbg !59
+
   ret void, !dbg !59
 }
 
@@ -108,7 +133,7 @@ entry:
 !4 = !{!5}
 !5 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !6, size: 64)
 !6 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
-!7 = !{!0, !8, !10, !13, !16, !19, !23, !25, !27, !29, !31, !33, !35, !37, !83}
+!7 = !{!0, !8, !10, !13, !16, !19, !23, !25, !27, !29, !31, !33, !35, !37, !83, !86, !88, !90, !92, !95, !97}
 !8 = !DIGlobalVariableExpression(var: !9, expr: !DIExpression())
 !9 = distinct !DIGlobalVariable(name: "g_2", scope: !2, file: !3, line: 2, type: !6, isLocal: false, isDefinition: true)
 !10 = !DIGlobalVariableExpression(var: !11, expr: !DIExpression())
@@ -159,3 +184,17 @@ entry:
 !82 = !DILocation(line: 31, column: 1, scope: !77)
 !83 = !DIGlobalVariableExpression(var: !84, expr: !DIExpression())
 !84 = distinct !DIGlobalVariable(name: "g_float_undef", linkageName: "g_float_undef", scope: !2, file: !3, line: 15, type: !6, isLocal: true, isDefinition: true)
+!85 = !DIBasicType(name: "float", size: 16, encoding: DW_ATE_float)
+!86 = !DIGlobalVariableExpression(var: !87, expr: !DIExpression())
+!87 = distinct !DIGlobalVariable(name: "g_8", scope: !2, file: !3, line: 2, type: !85, isLocal: false, isDefinition: true)
+!88 = !DIGlobalVariableExpression(var: !89, expr: !DIExpression())
+!89 = distinct !DIGlobalVariable(name: "g_88", linkageName: "_ZL4g_88", scope: !2, file: !3, line: 10, type: !85, isLocal: true, isDefinition: true)
+!90 = !DIGlobalVariableExpression(var: !91, expr: !DIExpression())
+!91 = distinct !DIGlobalVariable(name: "g_9", scope: !2, file: !3, line: 2, type: !85, isLocal: false, isDefinition: true)
+!92 = !DIGlobalVariableExpression(var: !93, expr: !DIExpression())
+!93 = distinct !DIGlobalVariable(name: "g_99", linkageName: "_ZL4g_99", scope: !2, file: !3, line: 10, type: !85, isLocal: true, isDefinition: true)
+
+!95 = !DIGlobalVariableExpression(var: !96, expr: !DIExpression())
+!96 = distinct !DIGlobalVariable(name: "g_i32_undef", linkageName: "g_i32_undef", scope: !2, file: !3, line: 9, type: !22, isLocal: true, isDefinition: true)
+!97 = !DIGlobalVariableExpression(var: !98, expr: !DIExpression())
+!98 = distinct !DIGlobalVariable(name: "g_ptr_undef", linkageName: "g_ptr_undef", scope: !2, file: !3, line: 14, type: !21, isLocal: true, isDefinition: true)
diff --git a/llvm/unittests/Transforms/Utils/LocalTest.cpp b/llvm/unittests/Transforms/Utils/LocalTest.cpp
index a86775a1366b0..d7d0ea2c6a6e7 100644
--- a/llvm/unittests/Transforms/Utils/LocalTest.cpp
+++ b/llvm/unittests/Transforms/Utils/LocalTest.cpp
@@ -1241,6 +1241,18 @@ TEST(Local, ExpressionForConstant) {
   EXPECT_NE(Expr, nullptr);
   EXPECT_EQ(Expr->getElement(1), 13841306799765140275U);
 
+  // Half.
+  Type *HalfTy = Type::getHalfTy(Context);
+  Expr = createExpression(ConstantFP::get(HalfTy, 5.55), HalfTy);
+  EXPECT_NE(Expr, nullptr);
+  EXPECT_EQ(Expr->getElement(1), 17805U);
+
+  // BFloat.
+  Type *BFloatTy = Type::getBFloatTy(Context);
+  Expr = createExpression(ConstantFP::get(BFloatTy, -5.55), BFloatTy);
+  EXPECT_NE(Expr, nullptr);
+  EXPECT_EQ(Expr->getElement(1), 49330U);
+
   // Pointer.
   PointerType *PtrTy = PointerType::get(Context, 0);
   Expr = createExpression(ConstantPointerNull::get(PtrTy), PtrTy);
@@ -1257,15 +1269,6 @@ TEST(Local, ExpressionForConstant) {
   EXPECT_NE(Expr, nullptr);
   EXPECT_EQ(Expr->getElement(1), 5678U);
 
-  // Others.
-  Type *HalfTy = Type::getHalfTy(Context);
-  Expr = createExpression(ConstantFP::get(HalfTy, 32), HalfTy);
-  EXPECT_EQ(Expr, nullptr);
-
-  Type *BFloatTy = Type::getBFloatTy(Context);
-  Expr = createExpression(ConstantFP::get(BFloatTy, 32), BFloatTy);
-  EXPECT_EQ(Expr, nullptr);
-
   Type *FP128Ty = Type::getFP128Ty(Context);
   Expr = createExpression(ConstantFP::get(FP128Ty, 32), FP128Ty);
   EXPECT_EQ(Expr, nullptr);

From d6d84b5d1448e4f2e24b467a0abcf42fe9d543e9 Mon Sep 17 00:00:00 2001
From: NagyDonat <donat.nagy@ericsson.com>
Date: Tue, 16 Apr 2024 10:41:26 +0200
Subject: [PATCH 538/886] [analyzer] Handle builtin functions in MallocChecker
 (#88416)

This commit ensures that the `CallDescription`s in `MallocChecker` are
matched with the mode `CDM::CLibrary`, so:
- they don't match methods or functions within user-defined namespaces;
- they also match builtin variants of these functions (if any), so the
checker can model `__builtin_alloca()` like `alloca()`.

This change fixes https://github.com/llvm/llvm-project/issues/81597. New
tests were added to verify that `std::malloc` and `std::free` (from
`<cstdlib>`) are modeled, but a method that's named e.g. `free` isn't
confused with the memory release function.

The responsibility for modeling `__builtin_alloca` and
`__builtin_alloca_with_align` was moved from `BuiltinFunctionChecker` to
`MallocChecker`, to avoid buggy interactions between the checkers and
ensure that the builtin and non-builtin variants are handled by exactly
the same logic.

This change might be a step backwards for the users who don't have
`unix.Malloc` enabled; but I suspect that `__builtin_alloca()` is so
rare that it would be a waste of time to implement backwards
compatibility for them.

There were several test files that relied on `__builtin_alloca()` calls
to get an `AllocaRegion`, these were modified to enable `unix.Malloc`.
One of these files (cxx-uninitialized-object-ptr-ref.cpp) had some tests
that relied on the fact that `malloc()` was treated as a "black box" in
them, these were updated to use `calloc()` (to get initialized memory)
and `free()` (to avoid memory leak reports).

While I was developing this change, I found a very suspicious assert in
`MallocChecker`. As it isn't blocking the goals of this commit, I just
marked it with a FIXME, but I'll try to investigate and fix it in a
follow-up change.
---
 .../Checkers/BuiltinFunctionChecker.cpp       | 25 ++----
 .../StaticAnalyzer/Checkers/MallocChecker.cpp | 89 ++++++++++---------
 .../Inputs/system-header-simulator-cxx.h      |  6 +-
 .../cxx-uninitialized-object-ptr-ref.cpp      | 24 +++--
 clang/test/Analysis/exercise-ps.c             |  2 +-
 clang/test/Analysis/explain-svals.cpp         |  2 +-
 clang/test/Analysis/malloc-std-namespace.cpp  | 24 +++++
 clang/test/Analysis/malloc.c                  | 11 +++
 clang/test/Analysis/malloc.cpp                | 11 +++
 clang/test/Analysis/stack-addr-ps.c           |  2 +-
 clang/test/Analysis/stackaddrleak.c           |  4 +-
 11 files changed, 125 insertions(+), 75 deletions(-)
 create mode 100644 clang/test/Analysis/malloc-std-namespace.cpp

diff --git a/clang/lib/StaticAnalyzer/Checkers/BuiltinFunctionChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/BuiltinFunctionChecker.cpp
index 01e46fa8591c0..1a75d7b52ad6e 100644
--- a/clang/lib/StaticAnalyzer/Checkers/BuiltinFunctionChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/BuiltinFunctionChecker.cpp
@@ -6,7 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This checker evaluates clang builtin functions.
+// This checker evaluates "standalone" clang builtin functions that are not
+// just special-cased variants of well-known non-builtin functions.
+// Builtin functions like __builtin_memcpy and __builtin_alloca should be
+// evaluated by the same checker that handles their non-builtin variant to
+// ensure that the two variants are handled consistently.
 //
 //===----------------------------------------------------------------------===//
 
@@ -80,25 +84,6 @@ bool BuiltinFunctionChecker::evalCall(const CallEvent &Call,
     return true;
   }
 
-  case Builtin::BI__builtin_alloca_with_align:
-  case Builtin::BI__builtin_alloca: {
-    SValBuilder &SVB = C.getSValBuilder();
-    const loc::MemRegionVal R =
-        SVB.getAllocaRegionVal(CE, C.getLocationContext(), C.blockCount());
-
-    // Set the extent of the region in bytes. This enables us to use the SVal
-    // of the argument directly. If we saved the extent in bits, it'd be more
-    // difficult to reason about values like symbol*8.
-    auto Size = Call.getArgSVal(0);
-    if (auto DefSize = Size.getAs<DefinedOrUnknownSVal>()) {
-      // This `getAs()` is mostly paranoia, because core.CallAndMessage reports
-      // undefined function arguments (unless it's disabled somehow).
-      state = setDynamicExtent(state, R.getRegion(), *DefSize, SVB);
-    }
-    C.addTransition(state->BindExpr(CE, LCtx, R));
-    return true;
-  }
-
   case Builtin::BI__builtin_dynamic_object_size:
   case Builtin::BI__builtin_object_size:
   case Builtin::BI__builtin_constant_p: {
diff --git a/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp
index 88fb42b6625aa..11651fd491f74 100644
--- a/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/MallocChecker.cpp
@@ -401,10 +401,11 @@ class MallocChecker
   };
 
   const CallDescriptionMap<CheckFn> FreeingMemFnMap{
-      {{{"free"}, 1}, &MallocChecker::checkFree},
-      {{{"if_freenameindex"}, 1}, &MallocChecker::checkIfFreeNameIndex},
-      {{{"kfree"}, 1}, &MallocChecker::checkFree},
-      {{{"g_free"}, 1}, &MallocChecker::checkFree},
+      {{CDM::CLibrary, {"free"}, 1}, &MallocChecker::checkFree},
+      {{CDM::CLibrary, {"if_freenameindex"}, 1},
+       &MallocChecker::checkIfFreeNameIndex},
+      {{CDM::CLibrary, {"kfree"}, 1}, &MallocChecker::checkFree},
+      {{CDM::CLibrary, {"g_free"}, 1}, &MallocChecker::checkFree},
   };
 
   bool isFreeingCall(const CallEvent &Call) const;
@@ -413,41 +414,46 @@ class MallocChecker
   friend class NoOwnershipChangeVisitor;
 
   CallDescriptionMap<CheckFn> AllocatingMemFnMap{
-      {{{"alloca"}, 1}, &MallocChecker::checkAlloca},
-      {{{"_alloca"}, 1}, &MallocChecker::checkAlloca},
-      {{{"malloc"}, 1}, &MallocChecker::checkBasicAlloc},
-      {{{"malloc"}, 3}, &MallocChecker::checkKernelMalloc},
-      {{{"calloc"}, 2}, &MallocChecker::checkCalloc},
-      {{{"valloc"}, 1}, &MallocChecker::checkBasicAlloc},
+      {{CDM::CLibrary, {"alloca"}, 1}, &MallocChecker::checkAlloca},
+      {{CDM::CLibrary, {"_alloca"}, 1}, &MallocChecker::checkAlloca},
+      // The line for "alloca" also covers "__builtin_alloca", but the
+      // _with_align variant must be listed separately because it takes an
+      // extra argument:
+      {{CDM::CLibrary, {"__builtin_alloca_with_align"}, 2},
+       &MallocChecker::checkAlloca},
+      {{CDM::CLibrary, {"malloc"}, 1}, &MallocChecker::checkBasicAlloc},
+      {{CDM::CLibrary, {"malloc"}, 3}, &MallocChecker::checkKernelMalloc},
+      {{CDM::CLibrary, {"calloc"}, 2}, &MallocChecker::checkCalloc},
+      {{CDM::CLibrary, {"valloc"}, 1}, &MallocChecker::checkBasicAlloc},
       {{CDM::CLibrary, {"strndup"}, 2}, &MallocChecker::checkStrdup},
       {{CDM::CLibrary, {"strdup"}, 1}, &MallocChecker::checkStrdup},
-      {{{"_strdup"}, 1}, &MallocChecker::checkStrdup},
-      {{{"kmalloc"}, 2}, &MallocChecker::checkKernelMalloc},
-      {{{"if_nameindex"}, 1}, &MallocChecker::checkIfNameIndex},
+      {{CDM::CLibrary, {"_strdup"}, 1}, &MallocChecker::checkStrdup},
+      {{CDM::CLibrary, {"kmalloc"}, 2}, &MallocChecker::checkKernelMalloc},
+      {{CDM::CLibrary, {"if_nameindex"}, 1}, &MallocChecker::checkIfNameIndex},
       {{CDM::CLibrary, {"wcsdup"}, 1}, &MallocChecker::checkStrdup},
       {{CDM::CLibrary, {"_wcsdup"}, 1}, &MallocChecker::checkStrdup},
-      {{{"g_malloc"}, 1}, &MallocChecker::checkBasicAlloc},
-      {{{"g_malloc0"}, 1}, &MallocChecker::checkGMalloc0},
-      {{{"g_try_malloc"}, 1}, &MallocChecker::checkBasicAlloc},
-      {{{"g_try_malloc0"}, 1}, &MallocChecker::checkGMalloc0},
-      {{{"g_memdup"}, 2}, &MallocChecker::checkGMemdup},
-      {{{"g_malloc_n"}, 2}, &MallocChecker::checkGMallocN},
-      {{{"g_malloc0_n"}, 2}, &MallocChecker::checkGMallocN0},
-      {{{"g_try_malloc_n"}, 2}, &MallocChecker::checkGMallocN},
-      {{{"g_try_malloc0_n"}, 2}, &MallocChecker::checkGMallocN0},
+      {{CDM::CLibrary, {"g_malloc"}, 1}, &MallocChecker::checkBasicAlloc},
+      {{CDM::CLibrary, {"g_malloc0"}, 1}, &MallocChecker::checkGMalloc0},
+      {{CDM::CLibrary, {"g_try_malloc"}, 1}, &MallocChecker::checkBasicAlloc},
+      {{CDM::CLibrary, {"g_try_malloc0"}, 1}, &MallocChecker::checkGMalloc0},
+      {{CDM::CLibrary, {"g_memdup"}, 2}, &MallocChecker::checkGMemdup},
+      {{CDM::CLibrary, {"g_malloc_n"}, 2}, &MallocChecker::checkGMallocN},
+      {{CDM::CLibrary, {"g_malloc0_n"}, 2}, &MallocChecker::checkGMallocN0},
+      {{CDM::CLibrary, {"g_try_malloc_n"}, 2}, &MallocChecker::checkGMallocN},
+      {{CDM::CLibrary, {"g_try_malloc0_n"}, 2}, &MallocChecker::checkGMallocN0},
   };
 
   CallDescriptionMap<CheckFn> ReallocatingMemFnMap{
-      {{{"realloc"}, 2},
+      {{CDM::CLibrary, {"realloc"}, 2},
        std::bind(&MallocChecker::checkRealloc, _1, _2, _3, false)},
-      {{{"reallocf"}, 2},
+      {{CDM::CLibrary, {"reallocf"}, 2},
        std::bind(&MallocChecker::checkRealloc, _1, _2, _3, true)},
-      {{{"g_realloc"}, 2},
+      {{CDM::CLibrary, {"g_realloc"}, 2},
        std::bind(&MallocChecker::checkRealloc, _1, _2, _3, false)},
-      {{{"g_try_realloc"}, 2},
+      {{CDM::CLibrary, {"g_try_realloc"}, 2},
        std::bind(&MallocChecker::checkRealloc, _1, _2, _3, false)},
-      {{{"g_realloc_n"}, 3}, &MallocChecker::checkReallocN},
-      {{{"g_try_realloc_n"}, 3}, &MallocChecker::checkReallocN},
+      {{CDM::CLibrary, {"g_realloc_n"}, 3}, &MallocChecker::checkReallocN},
+      {{CDM::CLibrary, {"g_try_realloc_n"}, 3}, &MallocChecker::checkReallocN},
 
       // NOTE: the following CallDescription also matches the C++ standard
       // library function std::getline(); the callback will filter it out.
@@ -1259,9 +1265,6 @@ static bool isStandardRealloc(const CallEvent &Call) {
   assert(FD);
   ASTContext &AC = FD->getASTContext();
 
-  if (isa<CXXMethodDecl>(FD))
-    return false;
-
   return FD->getDeclaredReturnType().getDesugaredType(AC) == AC.VoidPtrTy &&
          FD->getParamDecl(0)->getType().getDesugaredType(AC) == AC.VoidPtrTy &&
          FD->getParamDecl(1)->getType().getDesugaredType(AC) ==
@@ -1273,9 +1276,6 @@ static bool isGRealloc(const CallEvent &Call) {
   assert(FD);
   ASTContext &AC = FD->getASTContext();
 
-  if (isa<CXXMethodDecl>(FD))
-    return false;
-
   return FD->getDeclaredReturnType().getDesugaredType(AC) == AC.VoidPtrTy &&
          FD->getParamDecl(0)->getType().getDesugaredType(AC) == AC.VoidPtrTy &&
          FD->getParamDecl(1)->getType().getDesugaredType(AC) ==
@@ -1284,14 +1284,14 @@ static bool isGRealloc(const CallEvent &Call) {
 
 void MallocChecker::checkRealloc(const CallEvent &Call, CheckerContext &C,
                                  bool ShouldFreeOnFail) const {
-  // HACK: CallDescription currently recognizes non-standard realloc functions
-  // as standard because it doesn't check the type, or wether its a non-method
-  // function. This should be solved by making CallDescription smarter.
-  // Mind that this came from a bug report, and all other functions suffer from
-  // this.
-  // https://bugs.llvm.org/show_bug.cgi?id=46253
+  // Ignore calls to functions whose type does not match the expected type of
+  // either the standard realloc or g_realloc from GLib.
+  // FIXME: Should we perform this kind of checking consistently for each
+  // function? If yes, then perhaps extend the `CallDescription` interface to
+  // handle this.
   if (!isStandardRealloc(Call) && !isGRealloc(Call))
     return;
+
   ProgramStateRef State = C.getState();
   State = ReallocMemAux(C, Call, ShouldFreeOnFail, State, AF_Malloc);
   State = ProcessZeroAllocCheck(Call, 1, State);
@@ -1842,9 +1842,18 @@ static ProgramStateRef MallocUpdateRefState(CheckerContext &C, const Expr *E,
     return nullptr;
 
   SymbolRef Sym = RetVal->getAsLocSymbol();
+
   // This is a return value of a function that was not inlined, such as malloc()
   // or new(). We've checked that in the caller. Therefore, it must be a symbol.
   assert(Sym);
+  // FIXME: In theory this assertion should fail for `alloca()` calls (because
+  // `AllocaRegion`s are not symbolic); but in practice this does not happen.
+  // As the current code appears to work correctly, I'm not touching this issue
+  // now, but it would be good to investigate and clarify this.
+  // Also note that perhaps the special `AllocaRegion` should be replaced by
+  // `SymbolicRegion` (or turned into a subclass of `SymbolicRegion`) to enable
+  // proper tracking of memory allocated by `alloca()` -- and after that change
+  // this assertion would become valid again.
 
   // Set the symbol's state to Allocated.
   return State->set<RegionState>(Sym, RefState::getAllocated(Family, E));
diff --git a/clang/test/Analysis/Inputs/system-header-simulator-cxx.h b/clang/test/Analysis/Inputs/system-header-simulator-cxx.h
index 85db68d41a6c8..1c2be322f83c2 100644
--- a/clang/test/Analysis/Inputs/system-header-simulator-cxx.h
+++ b/clang/test/Analysis/Inputs/system-header-simulator-cxx.h
@@ -1106,6 +1106,7 @@ using ostream = basic_ostream<char>;
 extern std::ostream cout;
 
 ostream &operator<<(ostream &, const string &);
+
 #if __cplusplus >= 202002L
 template <class T>
 ostream &operator<<(ostream &, const std::unique_ptr<T> &);
@@ -1122,11 +1123,12 @@ istream &getline(istream &, string &, char);
 istream &getline(istream &, string &);
 } // namespace std
 
-#ifdef TEST_INLINABLE_ALLOCATORS
 namespace std {
   void *malloc(size_t);
   void free(void *);
-}
+} // namespace std
+
+#ifdef TEST_INLINABLE_ALLOCATORS
 void* operator new(std::size_t size, const std::nothrow_t&) throw() { return std::malloc(size); }
 void* operator new[](std::size_t size, const std::nothrow_t&) throw() { return std::malloc(size); }
 void operator delete(void* ptr, const std::nothrow_t&) throw() { std::free(ptr); }
diff --git a/clang/test/Analysis/cxx-uninitialized-object-ptr-ref.cpp b/clang/test/Analysis/cxx-uninitialized-object-ptr-ref.cpp
index fc067dd04428a..f46a2c9bc368f 100644
--- a/clang/test/Analysis/cxx-uninitialized-object-ptr-ref.cpp
+++ b/clang/test/Analysis/cxx-uninitialized-object-ptr-ref.cpp
@@ -1,9 +1,9 @@
-// RUN: %clang_analyze_cc1 -analyzer-checker=core,optin.cplusplus.UninitializedObject \
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc,optin.cplusplus.UninitializedObject \
 // RUN:   -analyzer-config optin.cplusplus.UninitializedObject:Pedantic=true -DPEDANTIC \
 // RUN:   -analyzer-config optin.cplusplus.UninitializedObject:CheckPointeeInitialization=true \
 // RUN:   -std=c++11 -verify  %s
 
-// RUN: %clang_analyze_cc1 -analyzer-checker=core,optin.cplusplus.UninitializedObject \
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc,optin.cplusplus.UninitializedObject \
 // RUN:   -analyzer-config optin.cplusplus.UninitializedObject:CheckPointeeInitialization=true \
 // RUN:   -std=c++11 -verify  %s
 
@@ -316,7 +316,10 @@ void fCyclicPointerTest2() {
 
 // Void pointer tests are mainly no-crash tests.
 
-void *malloc(int size);
+typedef __typeof(sizeof(int)) size_t;
+
+void *calloc(size_t nmemb, size_t size);
+void free(void *p);
 
 class VoidPointerTest1 {
   void *vptr;
@@ -328,8 +331,9 @@ class VoidPointerTest1 {
 };
 
 void fVoidPointerTest1() {
-  void *vptr = malloc(sizeof(int));
+  void *vptr = calloc(1, sizeof(int));
   VoidPointerTest1(vptr, char());
+  free(vptr);
 }
 
 class VoidPointerTest2 {
@@ -342,8 +346,9 @@ class VoidPointerTest2 {
 };
 
 void fVoidPointerTest2() {
-  void *vptr = malloc(sizeof(int));
+  void *vptr = calloc(1, sizeof(int));
   VoidPointerTest2(&vptr, char());
+  free(vptr);
 }
 
 class VoidPointerRRefTest1 {
@@ -359,8 +364,9 @@ upon returning to the caller.  This will be a dangling reference}}
 };
 
 void fVoidPointerRRefTest1() {
-  void *vptr = malloc(sizeof(int));
+  void *vptr = calloc(1, sizeof(int));
   VoidPointerRRefTest1(vptr, char());
+  free(vptr);
 }
 
 class VoidPointerRRefTest2 {
@@ -376,8 +382,9 @@ upon returning to the caller.  This will be a dangling reference}}
 };
 
 void fVoidPointerRRefTest2() {
-  void *vptr = malloc(sizeof(int));
+  void *vptr = calloc(1, sizeof(int));
   VoidPointerRRefTest2(&vptr, char());
+  free(vptr);
 }
 
 class VoidPointerLRefTest {
@@ -393,8 +400,9 @@ upon returning to the caller.  This will be a dangling reference}}
 };
 
 void fVoidPointerLRefTest() {
-  void *vptr = malloc(sizeof(int));
+  void *vptr = calloc(1, sizeof(int));
   VoidPointerLRefTest(vptr, char());
+  free(vptr);
 }
 
 struct CyclicVoidPointerTest {
diff --git a/clang/test/Analysis/exercise-ps.c b/clang/test/Analysis/exercise-ps.c
index d214c3959b207..d1e1771afddb5 100644
--- a/clang/test/Analysis/exercise-ps.c
+++ b/clang/test/Analysis/exercise-ps.c
@@ -1,5 +1,5 @@
 // RUN: %clang_analyze_cc1 %s -verify -Wno-error=implicit-function-declaration \
-// RUN:   -analyzer-checker=core \
+// RUN:   -analyzer-checker=core,unix.Malloc \
 // RUN:   -analyzer-config core.CallAndMessage:ArgPointeeInitializedness=true
 //
 // Just exercise the analyzer on code that has at one point caused issues
diff --git a/clang/test/Analysis/explain-svals.cpp b/clang/test/Analysis/explain-svals.cpp
index 30368b6976cc2..33fce10c4e2b2 100644
--- a/clang/test/Analysis/explain-svals.cpp
+++ b/clang/test/Analysis/explain-svals.cpp
@@ -1,7 +1,7 @@
 // RUN: %clang_analyze_cc1 -triple i386-apple-darwin10 -verify %s \
-// RUN:   -analyzer-checker=core.builtin \
 // RUN:   -analyzer-checker=debug.ExprInspection \
 // RUN:   -analyzer-checker=unix.cstring \
+// RUN:   -analyzer-checker=unix.Malloc \
 // RUN:   -analyzer-config display-checker-name=false
 
 typedef unsigned long size_t;
diff --git a/clang/test/Analysis/malloc-std-namespace.cpp b/clang/test/Analysis/malloc-std-namespace.cpp
new file mode 100644
index 0000000000000..d4e397bb812aa
--- /dev/null
+++ b/clang/test/Analysis/malloc-std-namespace.cpp
@@ -0,0 +1,24 @@
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc -verify -analyzer-output=text %s
+
+// This file tests that unix.Malloc can handle C++ code where e.g. malloc and
+// free are declared within the namespace 'std' by the header <cstdlib>.
+
+#include "Inputs/system-header-simulator-cxx.h"
+
+void leak() {
+  int *p = static_cast<int*>(std::malloc(sizeof(int))); // expected-note{{Memory is allocated}}
+} // expected-warning{{Potential leak of memory pointed to by 'p'}}
+  // expected-note@-1{{Potential leak of memory pointed to by 'p'}}
+
+void no_leak() {
+  int *p = static_cast<int*>(std::malloc(sizeof(int)));
+  std::free(p); // no-warning
+}
+
+void invalid_free() {
+  int i;
+  int *p = &i;
+  //expected-note@+2{{Argument to free() is the address of the local variable 'i', which is not memory allocated by malloc()}}
+  //expected-warning@+1{{Argument to free() is the address of the local variable 'i', which is not memory allocated by malloc()}}
+  std::free(p);
+}
diff --git a/clang/test/Analysis/malloc.c b/clang/test/Analysis/malloc.c
index 09cd4b0bfce63..e5cb45ba73352 100644
--- a/clang/test/Analysis/malloc.c
+++ b/clang/test/Analysis/malloc.c
@@ -740,6 +740,17 @@ void allocaFree(void) {
   free(p); // expected-warning {{Memory allocated by alloca() should not be deallocated}}
 }
 
+void allocaFreeBuiltin(void) {
+  int *p = __builtin_alloca(sizeof(int));
+  free(p); // expected-warning {{Memory allocated by alloca() should not be deallocated}}
+}
+
+void allocaFreeBuiltinAlign(void) {
+  int *p = __builtin_alloca_with_align(sizeof(int), 64);
+  free(p); // expected-warning {{Memory allocated by alloca() should not be deallocated}}
+}
+
+
 int* mallocEscapeRet(void) {
   int *p = malloc(12);
   return p; // no warning
diff --git a/clang/test/Analysis/malloc.cpp b/clang/test/Analysis/malloc.cpp
index 14b4c0576384f..300b344ab25d6 100644
--- a/clang/test/Analysis/malloc.cpp
+++ b/clang/test/Analysis/malloc.cpp
@@ -214,3 +214,14 @@ void *realloc(void **ptr, size_t size) { realloc(ptr, size); } // no-crash
 namespace pr46253_paramty2{
 void *realloc(void *ptr, int size) { realloc(ptr, size); } // no-crash
 } // namespace pr46253_paramty2
+
+namespace pr81597 {
+struct S {};
+struct T {
+  void free(const S& s);
+};
+void f(T& t) {
+  S s;
+  t.free(s); // no-warning: This is not the free you are looking for...
+}
+} // namespace pr81597
diff --git a/clang/test/Analysis/stack-addr-ps.c b/clang/test/Analysis/stack-addr-ps.c
index e469396e1bb22..e69ab4189b524 100644
--- a/clang/test/Analysis/stack-addr-ps.c
+++ b/clang/test/Analysis/stack-addr-ps.c
@@ -1,4 +1,4 @@
-// RUN: %clang_analyze_cc1 -analyzer-checker=core -fblocks -verify %s
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc -fblocks -verify %s
 
 int* f1(void) {
   int x = 0;
diff --git a/clang/test/Analysis/stackaddrleak.c b/clang/test/Analysis/stackaddrleak.c
index 0583bfc18711c..39c29f2a2635b 100644
--- a/clang/test/Analysis/stackaddrleak.c
+++ b/clang/test/Analysis/stackaddrleak.c
@@ -1,5 +1,5 @@
-// RUN: %clang_analyze_cc1 -analyzer-checker=core -verify -std=c99 -Dbool=_Bool -Wno-bool-conversion %s
-// RUN: %clang_analyze_cc1 -analyzer-checker=core -verify -x c++ -Wno-bool-conversion %s
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc -verify -std=c99 -Dbool=_Bool -Wno-bool-conversion %s
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,unix.Malloc -verify -x c++ -Wno-bool-conversion %s
 
 typedef __INTPTR_TYPE__ intptr_t;
 char const *p;

From 40dd3aa91d3f73184e34e45e597b84bec059c572 Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Tue, 16 Apr 2024 10:59:02 +0200
Subject: [PATCH 539/886] [mlir][Interfaces] `Variable` abstraction for
 `ValueBoundsOpInterface` (#87980)

This commit generalizes and cleans up the `ValueBoundsConstraintSet`
API. The API used to provide function overloads for comparing/computing
bounds of:
- index-typed SSA value
- dimension of shaped value
- affine map + operands

This commit removes all overloads. There is now a single entry point for
each `compare` variant and each `computeBound` variant. These functions
now take a `Variable`, which is internally represented as an affine map
and map operands.

This commit also adds support for computing bounds for an affine map +
operands. There was previously no public API for that.
---
 .../Dialect/Affine/Transforms/Transforms.h    |  11 +
 .../Dialect/Arith/Transforms/Transforms.h     |  11 +
 .../mlir/Interfaces/ValueBoundsOpInterface.h  | 119 +++---
 .../Affine/IR/ValueBoundsOpInterfaceImpl.cpp  |   6 +-
 .../Affine/Transforms/ReifyValueBounds.cpp    |  15 +-
 .../Arith/IR/ValueBoundsOpInterfaceImpl.cpp   |   8 +-
 .../Dialect/Arith/Transforms/IntNarrowing.cpp |   2 +-
 .../Arith/Transforms/ReifyValueBounds.cpp     |  15 +-
 .../lib/Dialect/Linalg/Transforms/Padding.cpp |   6 +-
 .../Dialect/Linalg/Transforms/Promotion.cpp   |   6 +-
 .../Transforms/IndependenceTransforms.cpp     |   5 +-
 .../SCF/IR/ValueBoundsOpInterfaceImpl.cpp     |  17 +-
 .../Tensor/IR/TensorTilingInterfaceImpl.cpp   |   3 +-
 .../Transforms/IndependenceTransforms.cpp     |   3 +-
 mlir/lib/Dialect/Tensor/Utils/Utils.cpp       |   4 +-
 .../lib/Interfaces/ValueBoundsOpInterface.cpp | 338 ++++++++----------
 .../value-bounds-op-interface-impl.mlir       |  24 ++
 .../Dialect/Affine/TestReifyValueBounds.cpp   |  26 +-
 mlir/test/lib/Dialect/Test/TestDialect.cpp    |  37 ++
 mlir/test/lib/Dialect/Test/TestOps.td         |  16 +-
 20 files changed, 361 insertions(+), 311 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Affine/Transforms/Transforms.h b/mlir/include/mlir/Dialect/Affine/Transforms/Transforms.h
index 8e840e744064d..1ea7375220815 100644
--- a/mlir/include/mlir/Dialect/Affine/Transforms/Transforms.h
+++ b/mlir/include/mlir/Dialect/Affine/Transforms/Transforms.h
@@ -53,6 +53,17 @@ void reorderOperandsByHoistability(RewriterBase &rewriter, AffineApplyOp op);
 /// maximally compose chains of AffineApplyOps.
 FailureOr<AffineApplyOp> decompose(RewriterBase &rewriter, AffineApplyOp op);
 
+/// Reify a bound for the given variable in terms of SSA values for which
+/// `stopCondition` is met.
+///
+/// By default, lower/equal bounds are closed and upper bounds are open. If
+/// `closedUB` is set to "true", upper bounds are also closed.
+FailureOr<OpFoldResult>
+reifyValueBound(OpBuilder &b, Location loc, presburger::BoundType type,
+                const ValueBoundsConstraintSet::Variable &var,
+                ValueBoundsConstraintSet::StopConditionFn stopCondition,
+                bool closedUB = false);
+
 /// Reify a bound for the given index-typed value in terms of SSA values for
 /// which `stopCondition` is met. If no stop condition is specified, reify in
 /// terms of the operands of the owner op.
diff --git a/mlir/include/mlir/Dialect/Arith/Transforms/Transforms.h b/mlir/include/mlir/Dialect/Arith/Transforms/Transforms.h
index 970a52a06a11a..bbc7e5d3e0dd7 100644
--- a/mlir/include/mlir/Dialect/Arith/Transforms/Transforms.h
+++ b/mlir/include/mlir/Dialect/Arith/Transforms/Transforms.h
@@ -24,6 +24,17 @@ enum class BoundType;
 
 namespace arith {
 
+/// Reify a bound for the given variable in terms of SSA values for which
+/// `stopCondition` is met.
+///
+/// By default, lower/equal bounds are closed and upper bounds are open. If
+/// `closedUB` is set to "true", upper bounds are also closed.
+FailureOr<OpFoldResult>
+reifyValueBound(OpBuilder &b, Location loc, presburger::BoundType type,
+                const ValueBoundsConstraintSet::Variable &var,
+                ValueBoundsConstraintSet::StopConditionFn stopCondition,
+                bool closedUB = false);
+
 /// Reify a bound for the given index-typed value in terms of SSA values for
 /// which `stopCondition` is met. If no stop condition is specified, reify in
 /// terms of the operands of the owner op.
diff --git a/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h b/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
index 1d7bc6ea961cc..ac17ace5a976d 100644
--- a/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
+++ b/mlir/include/mlir/Interfaces/ValueBoundsOpInterface.h
@@ -15,6 +15,7 @@
 #include "mlir/IR/Value.h"
 #include "mlir/Interfaces/DestinationStyleOpInterface.h"
 #include "llvm/ADT/SetVector.h"
+#include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/ExtensibleRTTI.h"
 
 #include <queue>
@@ -111,6 +112,39 @@ class ValueBoundsConstraintSet
 public:
   static char ID;
 
+  /// A variable that can be added to the constraint set as a "column". The
+  /// value bounds infrastructure can compute bounds for variables and compare
+  /// two variables.
+  ///
+  /// Internally, a variable is represented as an affine map and operands.
+  class Variable {
+  public:
+    /// Construct a variable for an index-typed attribute or SSA value.
+    Variable(OpFoldResult ofr);
+
+    /// Construct a variable for an index-typed SSA value.
+    Variable(Value indexValue);
+
+    /// Construct a variable for a dimension of a shaped value.
+    Variable(Value shapedValue, int64_t dim);
+
+    /// Construct a variable for an index-typed attribute/SSA value or for a
+    /// dimension of a shaped value. A non-null dimension must be provided if
+    /// and only if `ofr` is a shaped value.
+    Variable(OpFoldResult ofr, std::optional<int64_t> dim);
+
+    /// Construct a variable for a map and its operands.
+    Variable(AffineMap map, ArrayRef<Variable> mapOperands);
+    Variable(AffineMap map, ArrayRef<Value> mapOperands);
+
+    MLIRContext *getContext() const { return map.getContext(); }
+
+  private:
+    friend class ValueBoundsConstraintSet;
+    AffineMap map;
+    ValueDimList mapOperands;
+  };
+
   /// The stop condition when traversing the backward slice of a shaped value/
   /// index-type value. The traversal continues until the stop condition
   /// evaluates to "true" for a value.
@@ -121,35 +155,31 @@ class ValueBoundsConstraintSet
   using StopConditionFn = std::function<bool(
       Value, std::optional<int64_t> /*dim*/, ValueBoundsConstraintSet &cstr)>;
 
-  /// Compute a bound for the given index-typed value or shape dimension size.
-  /// The computed bound is stored in `resultMap`. The operands of the bound are
-  /// stored in `mapOperands`. An operand is either an index-type SSA value
-  /// or a shaped value and a dimension.
+  /// Compute a bound for the given variable. The computed bound is stored in
+  /// `resultMap`. The operands of the bound are stored in `mapOperands`. An
+  /// operand is either an index-type SSA value or a shaped value and a
+  /// dimension.
   ///
-  /// `dim` must be `nullopt` if and only if `value` is index-typed. The bound
-  /// is computed in terms of values/dimensions for which `stopCondition`
-  /// evaluates to "true". To that end, the backward slice (reverse use-def
-  /// chain) of the given value is visited in a worklist-driven manner and the
-  /// constraint set is populated according to `ValueBoundsOpInterface` for each
-  /// visited value.
+  /// The bound is computed in terms of values/dimensions for which
+  /// `stopCondition` evaluates to "true". To that end, the backward slice
+  /// (reverse use-def chain) of the given value is visited in a worklist-driven
+  /// manner and the constraint set is populated according to
+  /// `ValueBoundsOpInterface` for each visited value.
   ///
   /// By default, lower/equal bounds are closed and upper bounds are open. If
   /// `closedUB` is set to "true", upper bounds are also closed.
-  static LogicalResult computeBound(AffineMap &resultMap,
-                                    ValueDimList &mapOperands,
-                                    presburger::BoundType type, Value value,
-                                    std::optional<int64_t> dim,
-                                    StopConditionFn stopCondition,
-                                    bool closedUB = false);
+  static LogicalResult
+  computeBound(AffineMap &resultMap, ValueDimList &mapOperands,
+               presburger::BoundType type, const Variable &var,
+               StopConditionFn stopCondition, bool closedUB = false);
 
   /// Compute a bound in terms of the values/dimensions in `dependencies`. The
   /// computed bound consists of only constant terms and dependent values (or
   /// dimension sizes thereof).
   static LogicalResult
   computeDependentBound(AffineMap &resultMap, ValueDimList &mapOperands,
-                        presburger::BoundType type, Value value,
-                        std::optional<int64_t> dim, ValueDimList dependencies,
-                        bool closedUB = false);
+                        presburger::BoundType type, const Variable &var,
+                        ValueDimList dependencies, bool closedUB = false);
 
   /// Compute a bound in that is independent of all values in `independencies`.
   ///
@@ -161,13 +191,10 @@ class ValueBoundsConstraintSet
   /// appear in the computed bound.
   static LogicalResult
   computeIndependentBound(AffineMap &resultMap, ValueDimList &mapOperands,
-                          presburger::BoundType type, Value value,
-                          std::optional<int64_t> dim, ValueRange independencies,
-                          bool closedUB = false);
+                          presburger::BoundType type, const Variable &var,
+                          ValueRange independencies, bool closedUB = false);
 
-  /// Compute a constant bound for the given affine map, where dims and symbols
-  /// are bound to the given operands. The affine map must have exactly one
-  /// result.
+  /// Compute a constant bound for the given variable.
   ///
   /// This function traverses the backward slice of the given operands in a
   /// worklist-driven manner until `stopCondition` evaluates to "true". The
@@ -182,16 +209,9 @@ class ValueBoundsConstraintSet
   /// By default, lower/equal bounds are closed and upper bounds are open. If
   /// `closedUB` is set to "true", upper bounds are also closed.
   static FailureOr<int64_t>
-  computeConstantBound(presburger::BoundType type, Value value,
-                       std::optional<int64_t> dim = std::nullopt,
+  computeConstantBound(presburger::BoundType type, const Variable &var,
                        StopConditionFn stopCondition = nullptr,
                        bool closedUB = false);
-  static FailureOr<int64_t> computeConstantBound(
-      presburger::BoundType type, AffineMap map, ValueDimList mapOperands,
-      StopConditionFn stopCondition = nullptr, bool closedUB = false);
-  static FailureOr<int64_t> computeConstantBound(
-      presburger::BoundType type, AffineMap map, ArrayRef<Value> mapOperands,
-      StopConditionFn stopCondition = nullptr, bool closedUB = false);
 
   /// Compute a constant delta between the given two values. Return "failure"
   /// if a constant delta could not be determined.
@@ -221,9 +241,8 @@ class ValueBoundsConstraintSet
   /// proven. This could be because the specified relation does in fact not hold
   /// or because there is not enough information in the constraint set. In other
   /// words, if we do not know for sure, this function returns "false".
-  bool populateAndCompare(OpFoldResult lhs, std::optional<int64_t> lhsDim,
-                          ComparisonOperator cmp, OpFoldResult rhs,
-                          std::optional<int64_t> rhsDim);
+  bool populateAndCompare(const Variable &lhs, ComparisonOperator cmp,
+                          const Variable &rhs);
 
   /// Return "true" if "lhs cmp rhs" was proven to hold. Return "false" if the
   /// specified relation could not be proven. This could be because the
@@ -233,24 +252,12 @@ class ValueBoundsConstraintSet
   ///
   /// This function keeps traversing the backward slice of lhs/rhs until could
   /// prove the relation or until it ran out of IR.
-  static bool compare(OpFoldResult lhs, std::optional<int64_t> lhsDim,
-                      ComparisonOperator cmp, OpFoldResult rhs,
-                      std::optional<int64_t> rhsDim);
-  static bool compare(AffineMap lhs, ValueDimList lhsOperands,
-                      ComparisonOperator cmp, AffineMap rhs,
-                      ValueDimList rhsOperands);
-  static bool compare(AffineMap lhs, ArrayRef<Value> lhsOperands,
-                      ComparisonOperator cmp, AffineMap rhs,
-                      ArrayRef<Value> rhsOperands);
-
-  /// Compute whether the given values/dimensions are equal. Return "failure" if
+  static bool compare(const Variable &lhs, ComparisonOperator cmp,
+                      const Variable &rhs);
+
+  /// Compute whether the given variables are equal. Return "failure" if
   /// equality could not be determined.
-  ///
-  /// `dim1`/`dim2` must be `nullopt` if and only if `value1`/`value2` are
-  /// index-typed.
-  static FailureOr<bool> areEqual(OpFoldResult value1, OpFoldResult value2,
-                                  std::optional<int64_t> dim1 = std::nullopt,
-                                  std::optional<int64_t> dim2 = std::nullopt);
+  static FailureOr<bool> areEqual(const Variable &var1, const Variable &var2);
 
   /// Return "true" if the given slices are guaranteed to be overlapping.
   /// Return "false" if the given slices are guaranteed to be non-overlapping.
@@ -317,9 +324,6 @@ class ValueBoundsConstraintSet
   ///
   /// This function does not analyze any IR and does not populate any additional
   /// constraints.
-  bool compareValueDims(OpFoldResult lhs, std::optional<int64_t> lhsDim,
-                        ComparisonOperator cmp, OpFoldResult rhs,
-                        std::optional<int64_t> rhsDim);
   bool comparePos(int64_t lhsPos, ComparisonOperator cmp, int64_t rhsPos);
 
   /// Given an affine map with a single result (and map operands), add a new
@@ -374,6 +378,7 @@ class ValueBoundsConstraintSet
   /// constraint system. Return the position of the new column. Any operands
   /// that were not analyzed yet are put on the worklist.
   int64_t insert(AffineMap map, ValueDimList operands, bool isSymbol = true);
+  int64_t insert(const Variable &var, bool isSymbol = true);
 
   /// Project out the given column in the constraint set.
   void projectOut(int64_t pos);
@@ -381,6 +386,8 @@ class ValueBoundsConstraintSet
   /// Project out all columns for which the condition holds.
   void projectOut(function_ref<bool(ValueDim)> condition);
 
+  void projectOutAnonymous(std::optional<int64_t> except = std::nullopt);
+
   /// Mapping of columns to values/shape dimensions.
   SmallVector<std::optional<ValueDim>> positionToValueDim;
   /// Reverse mapping of values/shape dimensions to columns.
diff --git a/mlir/lib/Dialect/Affine/IR/ValueBoundsOpInterfaceImpl.cpp b/mlir/lib/Dialect/Affine/IR/ValueBoundsOpInterfaceImpl.cpp
index e0c3abe7a0f71..82a9fb0d49088 100644
--- a/mlir/lib/Dialect/Affine/IR/ValueBoundsOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Affine/IR/ValueBoundsOpInterfaceImpl.cpp
@@ -120,9 +120,7 @@ mlir::affine::fullyComposeAndComputeConstantDelta(Value value1, Value value2) {
   mapOperands.push_back(value1);
   mapOperands.push_back(value2);
   affine::fullyComposeAffineMapAndOperands(&map, &mapOperands);
-  ValueDimList valueDims;
-  for (Value v : mapOperands)
-    valueDims.push_back({v, std::nullopt});
   return ValueBoundsConstraintSet::computeConstantBound(
-      presburger::BoundType::EQ, map, valueDims);
+      presburger::BoundType::EQ,
+      ValueBoundsConstraintSet::Variable(map, mapOperands));
 }
diff --git a/mlir/lib/Dialect/Affine/Transforms/ReifyValueBounds.cpp b/mlir/lib/Dialect/Affine/Transforms/ReifyValueBounds.cpp
index 117ee8e8701ad..1a266b72d1f8d 100644
--- a/mlir/lib/Dialect/Affine/Transforms/ReifyValueBounds.cpp
+++ b/mlir/lib/Dialect/Affine/Transforms/ReifyValueBounds.cpp
@@ -16,16 +16,15 @@
 using namespace mlir;
 using namespace mlir::affine;
 
-static FailureOr<OpFoldResult>
-reifyValueBound(OpBuilder &b, Location loc, presburger::BoundType type,
-                Value value, std::optional<int64_t> dim,
-                ValueBoundsConstraintSet::StopConditionFn stopCondition,
-                bool closedUB) {
+FailureOr<OpFoldResult> mlir::affine::reifyValueBound(
+    OpBuilder &b, Location loc, presburger::BoundType type,
+    const ValueBoundsConstraintSet::Variable &var,
+    ValueBoundsConstraintSet::StopConditionFn stopCondition, bool closedUB) {
   // Compute bound.
   AffineMap boundMap;
   ValueDimList mapOperands;
   if (failed(ValueBoundsConstraintSet::computeBound(
-          boundMap, mapOperands, type, value, dim, stopCondition, closedUB)))
+          boundMap, mapOperands, type, var, stopCondition, closedUB)))
     return failure();
 
   // Reify bound.
@@ -93,7 +92,7 @@ FailureOr<OpFoldResult> mlir::affine::reifyShapedValueDimBound(
     // the owner of `value`.
     return v != value;
   };
-  return reifyValueBound(b, loc, type, value, dim,
+  return reifyValueBound(b, loc, type, {value, dim},
                          stopCondition ? stopCondition : reifyToOperands,
                          closedUB);
 }
@@ -105,7 +104,7 @@ FailureOr<OpFoldResult> mlir::affine::reifyIndexValueBound(
                              ValueBoundsConstraintSet &cstr) {
     return v != value;
   };
-  return reifyValueBound(b, loc, type, value, /*dim=*/std::nullopt,
+  return reifyValueBound(b, loc, type, value,
                          stopCondition ? stopCondition : reifyToOperands,
                          closedUB);
 }
diff --git a/mlir/lib/Dialect/Arith/IR/ValueBoundsOpInterfaceImpl.cpp b/mlir/lib/Dialect/Arith/IR/ValueBoundsOpInterfaceImpl.cpp
index f0d43808bc45d..7cfcc4180539c 100644
--- a/mlir/lib/Dialect/Arith/IR/ValueBoundsOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Arith/IR/ValueBoundsOpInterfaceImpl.cpp
@@ -107,9 +107,9 @@ struct SelectOpInterface
     // If trueValue <= falseValue:
     // * result <= falseValue
     // * result >= trueValue
-    if (cstr.compare(trueValue, dim,
+    if (cstr.compare(/*lhs=*/{trueValue, dim},
                      ValueBoundsConstraintSet::ComparisonOperator::LE,
-                     falseValue, dim)) {
+                     /*rhs=*/{falseValue, dim})) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(trueValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(falseValue, dim);
@@ -121,9 +121,9 @@ struct SelectOpInterface
     // If falseValue <= trueValue:
     // * result <= trueValue
     // * result >= falseValue
-    if (cstr.compare(falseValue, dim,
+    if (cstr.compare(/*lhs=*/{falseValue, dim},
                      ValueBoundsConstraintSet::ComparisonOperator::LE,
-                     trueValue, dim)) {
+                     /*rhs=*/{trueValue, dim})) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(falseValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(trueValue, dim);
diff --git a/mlir/lib/Dialect/Arith/Transforms/IntNarrowing.cpp b/mlir/lib/Dialect/Arith/Transforms/IntNarrowing.cpp
index 79fabd6ed2e99..f87f3d6350c02 100644
--- a/mlir/lib/Dialect/Arith/Transforms/IntNarrowing.cpp
+++ b/mlir/lib/Dialect/Arith/Transforms/IntNarrowing.cpp
@@ -449,7 +449,7 @@ struct IndexCastPattern final : NarrowingPattern<CastOp> {
       return failure();
 
     FailureOr<int64_t> ub = ValueBoundsConstraintSet::computeConstantBound(
-        presburger::BoundType::UB, in, /*dim=*/std::nullopt,
+        presburger::BoundType::UB, in,
         /*stopCondition=*/nullptr, /*closedUB=*/true);
     if (failed(ub))
       return failure();
diff --git a/mlir/lib/Dialect/Arith/Transforms/ReifyValueBounds.cpp b/mlir/lib/Dialect/Arith/Transforms/ReifyValueBounds.cpp
index fad221288f190..5fb7953f93700 100644
--- a/mlir/lib/Dialect/Arith/Transforms/ReifyValueBounds.cpp
+++ b/mlir/lib/Dialect/Arith/Transforms/ReifyValueBounds.cpp
@@ -61,16 +61,15 @@ static Value buildArithValue(OpBuilder &b, Location loc, AffineMap map,
   return buildExpr(map.getResult(0));
 }
 
-static FailureOr<OpFoldResult>
-reifyValueBound(OpBuilder &b, Location loc, presburger::BoundType type,
-                Value value, std::optional<int64_t> dim,
-                ValueBoundsConstraintSet::StopConditionFn stopCondition,
-                bool closedUB) {
+FailureOr<OpFoldResult> mlir::arith::reifyValueBound(
+    OpBuilder &b, Location loc, presburger::BoundType type,
+    const ValueBoundsConstraintSet::Variable &var,
+    ValueBoundsConstraintSet::StopConditionFn stopCondition, bool closedUB) {
   // Compute bound.
   AffineMap boundMap;
   ValueDimList mapOperands;
   if (failed(ValueBoundsConstraintSet::computeBound(
-          boundMap, mapOperands, type, value, dim, stopCondition, closedUB)))
+          boundMap, mapOperands, type, var, stopCondition, closedUB)))
     return failure();
 
   // Materialize tensor.dim/memref.dim ops.
@@ -128,7 +127,7 @@ FailureOr<OpFoldResult> mlir::arith::reifyShapedValueDimBound(
     // the owner of `value`.
     return v != value;
   };
-  return reifyValueBound(b, loc, type, value, dim,
+  return reifyValueBound(b, loc, type, {value, dim},
                          stopCondition ? stopCondition : reifyToOperands,
                          closedUB);
 }
@@ -140,7 +139,7 @@ FailureOr<OpFoldResult> mlir::arith::reifyIndexValueBound(
                              ValueBoundsConstraintSet &cstr) {
     return v != value;
   };
-  return reifyValueBound(b, loc, type, value, /*dim=*/std::nullopt,
+  return reifyValueBound(b, loc, type, value,
                          stopCondition ? stopCondition : reifyToOperands,
                          closedUB);
 }
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp b/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
index 8c4b70db24898..518d2e138c02a 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
@@ -72,8 +72,10 @@ static LogicalResult computePaddedShape(linalg::LinalgOp opToPad,
     // Otherwise, try to compute a constant upper bound for the size value.
     FailureOr<int64_t> upperBound =
         ValueBoundsConstraintSet::computeConstantBound(
-            presburger::BoundType::UB, opOperand->get(),
-            /*dim=*/i, /*stopCondition=*/nullptr, /*closedUB=*/true);
+            presburger::BoundType::UB,
+            {opOperand->get(),
+             /*dim=*/i},
+            /*stopCondition=*/nullptr, /*closedUB=*/true);
     if (failed(upperBound)) {
       LLVM_DEBUG(DBGS() << "----could not compute a bounding box for padding");
       return failure();
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp b/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
index ac896d6c30d04..71eb59d40836c 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Promotion.cpp
@@ -257,14 +257,12 @@ FailureOr<PromotionInfo> mlir::linalg::promoteSubviewAsNewBuffer(
     if (auto attr = llvm::dyn_cast_if_present<Attribute>(rangeValue.size)) {
       size = getValueOrCreateConstantIndexOp(b, loc, rangeValue.size);
     } else {
-      Value materializedSize =
-          getValueOrCreateConstantIndexOp(b, loc, rangeValue.size);
       FailureOr<int64_t> upperBound =
           ValueBoundsConstraintSet::computeConstantBound(
-              presburger::BoundType::UB, materializedSize, /*dim=*/std::nullopt,
+              presburger::BoundType::UB, rangeValue.size,
               /*stopCondition=*/nullptr, /*closedUB=*/true);
       size = failed(upperBound)
-                 ? materializedSize
+                 ? getValueOrCreateConstantIndexOp(b, loc, rangeValue.size)
                  : b.create<arith::ConstantIndexOp>(loc, *upperBound);
     }
     LLVM_DEBUG(llvm::dbgs() << "Extracted tightest: " << size << "\n");
diff --git a/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp b/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
index 10ba508265e7b..1f06318cbd60e 100644
--- a/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
+++ b/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
@@ -23,12 +23,11 @@ static FailureOr<OpFoldResult> makeIndependent(OpBuilder &b, Location loc,
                                                ValueRange independencies) {
   if (ofr.is<Attribute>())
     return ofr;
-  Value value = ofr.get<Value>();
   AffineMap boundMap;
   ValueDimList mapOperands;
   if (failed(ValueBoundsConstraintSet::computeIndependentBound(
-          boundMap, mapOperands, presburger::BoundType::UB, value,
-          /*dim=*/std::nullopt, independencies, /*closedUB=*/true)))
+          boundMap, mapOperands, presburger::BoundType::UB, ofr, independencies,
+          /*closedUB=*/true)))
     return failure();
   return affine::materializeComputedBound(b, loc, boundMap, mapOperands);
 }
diff --git a/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp b/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
index 087ffc438a830..17a1c016ea16d 100644
--- a/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/SCF/IR/ValueBoundsOpInterfaceImpl.cpp
@@ -61,12 +61,13 @@ struct ForOpInterface
     // An EQ constraint can be added if the yielded value (dimension size)
     // equals the corresponding block argument (dimension size).
     if (cstr.populateAndCompare(
-            yieldedValue, dim, ValueBoundsConstraintSet::ComparisonOperator::EQ,
-            iterArg, dim)) {
+            /*lhs=*/{yieldedValue, dim},
+            ValueBoundsConstraintSet::ComparisonOperator::EQ,
+            /*rhs=*/{iterArg, dim})) {
       if (dim.has_value()) {
         cstr.bound(value)[*dim] == cstr.getExpr(initArg, dim);
       } else {
-        cstr.bound(value) == initArg;
+        cstr.bound(value) == cstr.getExpr(initArg);
       }
     }
   }
@@ -113,8 +114,9 @@ struct IfOpInterface
     // * result <= elseValue
     // * result >= thenValue
     if (cstr.populateAndCompare(
-            thenValue, dim, ValueBoundsConstraintSet::ComparisonOperator::LE,
-            elseValue, dim)) {
+            /*lhs=*/{thenValue, dim},
+            ValueBoundsConstraintSet::ComparisonOperator::LE,
+            /*rhs=*/{elseValue, dim})) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(thenValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(elseValue, dim);
@@ -127,8 +129,9 @@ struct IfOpInterface
     // * result <= thenValue
     // * result >= elseValue
     if (cstr.populateAndCompare(
-            elseValue, dim, ValueBoundsConstraintSet::ComparisonOperator::LE,
-            thenValue, dim)) {
+            /*lhs=*/{elseValue, dim},
+            ValueBoundsConstraintSet::ComparisonOperator::LE,
+            /*rhs=*/{thenValue, dim})) {
       if (dim) {
         cstr.bound(value)[*dim] >= cstr.getExpr(elseValue, dim);
         cstr.bound(value)[*dim] <= cstr.getExpr(thenValue, dim);
diff --git a/mlir/lib/Dialect/Tensor/IR/TensorTilingInterfaceImpl.cpp b/mlir/lib/Dialect/Tensor/IR/TensorTilingInterfaceImpl.cpp
index 67080d8e301c1..d25efcf50ec56 100644
--- a/mlir/lib/Dialect/Tensor/IR/TensorTilingInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Tensor/IR/TensorTilingInterfaceImpl.cpp
@@ -289,8 +289,7 @@ static UnpackTileDimInfo getUnpackTileDimInfo(OpBuilder &b, UnPackOp unpackOp,
 
   info.isAlignedToInnerTileSize = false;
   FailureOr<int64_t> cstSize = ValueBoundsConstraintSet::computeConstantBound(
-      presburger::BoundType::UB,
-      getValueOrCreateConstantIndexOp(b, loc, tileSize), /*dim=*/std::nullopt,
+      presburger::BoundType::UB, tileSize,
       /*stopCondition=*/nullptr, /*closedUB=*/true);
   std::optional<int64_t> cstInnerSize = getConstantIntValue(innerTileSize);
   if (!failed(cstSize) && cstInnerSize) {
diff --git a/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp b/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
index 721730862d49b..a89ce20048dff 100644
--- a/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
+++ b/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
@@ -28,7 +28,8 @@ static FailureOr<OpFoldResult> makeIndependent(OpBuilder &b, Location loc,
   ValueDimList mapOperands;
   if (failed(ValueBoundsConstraintSet::computeIndependentBound(
           boundMap, mapOperands, presburger::BoundType::UB, value,
-          /*dim=*/std::nullopt, independencies, /*closedUB=*/true)))
+          independencies,
+          /*closedUB=*/true)))
     return failure();
   return mlir::affine::materializeComputedBound(b, loc, boundMap, mapOperands);
 }
diff --git a/mlir/lib/Dialect/Tensor/Utils/Utils.cpp b/mlir/lib/Dialect/Tensor/Utils/Utils.cpp
index 2dd91e2f7a170..15381ec520e21 100644
--- a/mlir/lib/Dialect/Tensor/Utils/Utils.cpp
+++ b/mlir/lib/Dialect/Tensor/Utils/Utils.cpp
@@ -154,7 +154,7 @@ bool mlir::tensor::isCastLikeInsertSliceOp(InsertSliceOp op) {
       continue;
     }
     FailureOr<bool> equalDimSize = ValueBoundsConstraintSet::areEqual(
-        op.getSource(), op.getResult(), srcDim, resultDim);
+        {op.getSource(), srcDim}, {op.getResult(), resultDim});
     if (failed(equalDimSize) || !*equalDimSize)
       return false;
     ++srcDim;
@@ -178,7 +178,7 @@ bool mlir::tensor::isCastLikeExtractSliceOp(ExtractSliceOp op) {
       continue;
     }
     FailureOr<bool> equalDimSize = ValueBoundsConstraintSet::areEqual(
-        op.getSource(), op.getResult(), dim, resultDim);
+        {op.getSource(), dim}, {op.getResult(), resultDim});
     if (failed(equalDimSize) || !*equalDimSize)
       return false;
     ++resultDim;
diff --git a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
index ffa4c0b55cad7..87937591e60ad 100644
--- a/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
+++ b/mlir/lib/Interfaces/ValueBoundsOpInterface.cpp
@@ -25,6 +25,12 @@ namespace mlir {
 #include "mlir/Interfaces/ValueBoundsOpInterface.cpp.inc"
 } // namespace mlir
 
+static Operation *getOwnerOfValue(Value value) {
+  if (auto bbArg = dyn_cast<BlockArgument>(value))
+    return bbArg.getOwner()->getParentOp();
+  return value.getDefiningOp();
+}
+
 HyperrectangularSlice::HyperrectangularSlice(ArrayRef<OpFoldResult> offsets,
                                              ArrayRef<OpFoldResult> sizes,
                                              ArrayRef<OpFoldResult> strides)
@@ -67,6 +73,83 @@ static std::optional<int64_t> getConstantIntValue(OpFoldResult ofr) {
   return std::nullopt;
 }
 
+ValueBoundsConstraintSet::Variable::Variable(OpFoldResult ofr)
+    : Variable(ofr, std::nullopt) {}
+
+ValueBoundsConstraintSet::Variable::Variable(Value indexValue)
+    : Variable(static_cast<OpFoldResult>(indexValue)) {}
+
+ValueBoundsConstraintSet::Variable::Variable(Value shapedValue, int64_t dim)
+    : Variable(static_cast<OpFoldResult>(shapedValue), std::optional(dim)) {}
+
+ValueBoundsConstraintSet::Variable::Variable(OpFoldResult ofr,
+                                             std::optional<int64_t> dim) {
+  Builder b(ofr.getContext());
+  if (auto constInt = ::getConstantIntValue(ofr)) {
+    assert(!dim && "expected no dim for index-typed values");
+    map = AffineMap::get(/*dimCount=*/0, /*symbolCount=*/0,
+                         b.getAffineConstantExpr(*constInt));
+    return;
+  }
+  Value value = cast<Value>(ofr);
+#ifndef NDEBUG
+  if (dim) {
+    assert(isa<ShapedType>(value.getType()) && "expected shaped type");
+  } else {
+    assert(value.getType().isIndex() && "expected index type");
+  }
+#endif // NDEBUG
+  map = AffineMap::get(/*dimCount=*/0, /*symbolCount=*/1,
+                       b.getAffineSymbolExpr(0));
+  mapOperands.emplace_back(value, dim);
+}
+
+ValueBoundsConstraintSet::Variable::Variable(AffineMap map,
+                                             ArrayRef<Variable> mapOperands) {
+  assert(map.getNumResults() == 1 && "expected single result");
+
+  // Turn all dims into symbols.
+  Builder b(map.getContext());
+  SmallVector<AffineExpr> dimReplacements, symReplacements;
+  for (int64_t i = 0, e = map.getNumDims(); i < e; ++i)
+    dimReplacements.push_back(b.getAffineSymbolExpr(i));
+  for (int64_t i = 0, e = map.getNumSymbols(); i < e; ++i)
+    symReplacements.push_back(b.getAffineSymbolExpr(i + map.getNumDims()));
+  AffineMap tmpMap = map.replaceDimsAndSymbols(
+      dimReplacements, symReplacements, /*numResultDims=*/0,
+      /*numResultSyms=*/map.getNumSymbols() + map.getNumDims());
+
+  // Inline operands.
+  DenseMap<AffineExpr, AffineExpr> replacements;
+  for (auto [index, var] : llvm::enumerate(mapOperands)) {
+    assert(var.map.getNumResults() == 1 && "expected single result");
+    assert(var.map.getNumDims() == 0 && "expected only symbols");
+    SmallVector<AffineExpr> symReplacements;
+    for (auto valueDim : var.mapOperands) {
+      auto it = llvm::find(this->mapOperands, valueDim);
+      if (it != this->mapOperands.end()) {
+        // There is already a symbol for this operand.
+        symReplacements.push_back(b.getAffineSymbolExpr(
+            std::distance(this->mapOperands.begin(), it)));
+      } else {
+        // This is a new operand: add a new symbol.
+        symReplacements.push_back(
+            b.getAffineSymbolExpr(this->mapOperands.size()));
+        this->mapOperands.push_back(valueDim);
+      }
+    }
+    replacements[b.getAffineSymbolExpr(index)] =
+        var.map.getResult(0).replaceSymbols(symReplacements);
+  }
+  this->map = tmpMap.replace(replacements, /*numResultDims=*/0,
+                             /*numResultSyms=*/this->mapOperands.size());
+}
+
+ValueBoundsConstraintSet::Variable::Variable(AffineMap map,
+                                             ArrayRef<Value> mapOperands)
+    : Variable(map, llvm::map_to_vector(mapOperands,
+                                        [](Value v) { return Variable(v); })) {}
+
 ValueBoundsConstraintSet::ValueBoundsConstraintSet(
     MLIRContext *ctx, StopConditionFn stopCondition)
     : builder(ctx), stopCondition(stopCondition) {
@@ -176,6 +259,11 @@ int64_t ValueBoundsConstraintSet::insert(Value value,
   assert(!valueDimToPosition.contains(valueDim) && "already mapped");
   int64_t pos = isSymbol ? cstr.appendVar(VarKind::Symbol)
                          : cstr.appendVar(VarKind::SetDim);
+  LLVM_DEBUG(llvm::dbgs() << "Inserting constraint set column " << pos
+                          << " for: " << value
+                          << " (dim: " << dim.value_or(kIndexValue)
+                          << ", owner: " << getOwnerOfValue(value)->getName()
+                          << ")\n");
   positionToValueDim.insert(positionToValueDim.begin() + pos, valueDim);
   // Update reverse mapping.
   for (int64_t i = pos, e = positionToValueDim.size(); i < e; ++i)
@@ -194,6 +282,8 @@ int64_t ValueBoundsConstraintSet::insert(Value value,
 int64_t ValueBoundsConstraintSet::insert(bool isSymbol) {
   int64_t pos = isSymbol ? cstr.appendVar(VarKind::Symbol)
                          : cstr.appendVar(VarKind::SetDim);
+  LLVM_DEBUG(llvm::dbgs() << "Inserting anonymous constraint set column " << pos
+                          << "\n");
   positionToValueDim.insert(positionToValueDim.begin() + pos, std::nullopt);
   // Update reverse mapping.
   for (int64_t i = pos, e = positionToValueDim.size(); i < e; ++i)
@@ -224,6 +314,10 @@ int64_t ValueBoundsConstraintSet::insert(AffineMap map, ValueDimList operands,
   return pos;
 }
 
+int64_t ValueBoundsConstraintSet::insert(const Variable &var, bool isSymbol) {
+  return insert(var.map, var.mapOperands, isSymbol);
+}
+
 int64_t ValueBoundsConstraintSet::getPos(Value value,
                                          std::optional<int64_t> dim) const {
 #ifndef NDEBUG
@@ -232,7 +326,10 @@ int64_t ValueBoundsConstraintSet::getPos(Value value,
           cast<BlockArgument>(value).getOwner()->isEntryBlock()) &&
          "unstructured control flow is not supported");
 #endif // NDEBUG
-
+  LLVM_DEBUG(llvm::dbgs() << "Getting pos for: " << value
+                          << " (dim: " << dim.value_or(kIndexValue)
+                          << ", owner: " << getOwnerOfValue(value)->getName()
+                          << ")\n");
   auto it =
       valueDimToPosition.find(std::make_pair(value, dim.value_or(kIndexValue)));
   assert(it != valueDimToPosition.end() && "expected mapped entry");
@@ -253,12 +350,6 @@ bool ValueBoundsConstraintSet::isMapped(Value value,
   return it != valueDimToPosition.end();
 }
 
-static Operation *getOwnerOfValue(Value value) {
-  if (auto bbArg = dyn_cast<BlockArgument>(value))
-    return bbArg.getOwner()->getParentOp();
-  return value.getDefiningOp();
-}
-
 void ValueBoundsConstraintSet::processWorklist() {
   LLVM_DEBUG(llvm::dbgs() << "Processing value bounds worklist...\n");
   while (!worklist.empty()) {
@@ -346,41 +437,47 @@ void ValueBoundsConstraintSet::projectOut(
   }
 }
 
+void ValueBoundsConstraintSet::projectOutAnonymous(
+    std::optional<int64_t> except) {
+  int64_t nextPos = 0;
+  while (nextPos < static_cast<int64_t>(positionToValueDim.size())) {
+    if (positionToValueDim[nextPos].has_value() || except == nextPos) {
+      ++nextPos;
+    } else {
+      projectOut(nextPos);
+      // The column was projected out so another column is now at that position.
+      // Do not increase the counter.
+    }
+  }
+}
+
 LogicalResult ValueBoundsConstraintSet::computeBound(
     AffineMap &resultMap, ValueDimList &mapOperands, presburger::BoundType type,
-    Value value, std::optional<int64_t> dim, StopConditionFn stopCondition,
-    bool closedUB) {
-#ifndef NDEBUG
-  assertValidValueDim(value, dim);
-#endif // NDEBUG
-
+    const Variable &var, StopConditionFn stopCondition, bool closedUB) {
+  MLIRContext *ctx = var.getContext();
   int64_t ubAdjustment = closedUB ? 0 : 1;
-  Builder b(value.getContext());
+  Builder b(ctx);
   mapOperands.clear();
 
   // Process the backward slice of `value` (i.e., reverse use-def chain) until
   // `stopCondition` is met.
-  ValueDim valueDim = std::make_pair(value, dim.value_or(kIndexValue));
-  ValueBoundsConstraintSet cstr(value.getContext(), stopCondition);
-  assert(!stopCondition(value, dim, cstr) &&
-         "stop condition should not be satisfied for starting point");
-  int64_t pos = cstr.insert(value, dim, /*isSymbol=*/false);
+  ValueBoundsConstraintSet cstr(ctx, stopCondition);
+  int64_t pos = cstr.insert(var, /*isSymbol=*/false);
+  assert(pos == 0 && "expected first column");
   cstr.processWorklist();
 
   // Project out all variables (apart from `valueDim`) that do not match the
   // stop condition.
   cstr.projectOut([&](ValueDim p) {
-    // Do not project out `valueDim`.
-    if (valueDim == p)
-      return false;
     auto maybeDim =
         p.second == kIndexValue ? std::nullopt : std::make_optional(p.second);
     return !stopCondition(p.first, maybeDim, cstr);
   });
+  cstr.projectOutAnonymous(/*except=*/pos);
 
   // Compute lower and upper bounds for `valueDim`.
   SmallVector<AffineMap> lb(1), ub(1);
-  cstr.cstr.getSliceBounds(pos, 1, value.getContext(), &lb, &ub,
+  cstr.cstr.getSliceBounds(pos, 1, ctx, &lb, &ub,
                            /*closedUB=*/true);
 
   // Note: There are TODOs in the implementation of `getSliceBounds`. In such a
@@ -477,10 +574,9 @@ LogicalResult ValueBoundsConstraintSet::computeBound(
 
 LogicalResult ValueBoundsConstraintSet::computeDependentBound(
     AffineMap &resultMap, ValueDimList &mapOperands, presburger::BoundType type,
-    Value value, std::optional<int64_t> dim, ValueDimList dependencies,
-    bool closedUB) {
+    const Variable &var, ValueDimList dependencies, bool closedUB) {
   return computeBound(
-      resultMap, mapOperands, type, value, dim,
+      resultMap, mapOperands, type, var,
       [&](Value v, std::optional<int64_t> d, ValueBoundsConstraintSet &cstr) {
         return llvm::is_contained(dependencies, std::make_pair(v, d));
       },
@@ -489,8 +585,7 @@ LogicalResult ValueBoundsConstraintSet::computeDependentBound(
 
 LogicalResult ValueBoundsConstraintSet::computeIndependentBound(
     AffineMap &resultMap, ValueDimList &mapOperands, presburger::BoundType type,
-    Value value, std::optional<int64_t> dim, ValueRange independencies,
-    bool closedUB) {
+    const Variable &var, ValueRange independencies, bool closedUB) {
   // Return "true" if the given value is independent of all values in
   // `independencies`. I.e., neither the value itself nor any value in the
   // backward slice (reverse use-def chain) is contained in `independencies`.
@@ -516,7 +611,7 @@ LogicalResult ValueBoundsConstraintSet::computeIndependentBound(
 
   // Reify bounds in terms of any independent values.
   return computeBound(
-      resultMap, mapOperands, type, value, dim,
+      resultMap, mapOperands, type, var,
       [&](Value v, std::optional<int64_t> d, ValueBoundsConstraintSet &cstr) {
         return isIndependent(v);
       },
@@ -524,35 +619,8 @@ LogicalResult ValueBoundsConstraintSet::computeIndependentBound(
 }
 
 FailureOr<int64_t> ValueBoundsConstraintSet::computeConstantBound(
-    presburger::BoundType type, Value value, std::optional<int64_t> dim,
-    StopConditionFn stopCondition, bool closedUB) {
-#ifndef NDEBUG
-  assertValidValueDim(value, dim);
-#endif // NDEBUG
-
-  AffineMap map =
-      AffineMap::get(/*dimCount=*/1, /*symbolCount=*/0,
-                     Builder(value.getContext()).getAffineDimExpr(0));
-  return computeConstantBound(type, map, {{value, dim}}, stopCondition,
-                              closedUB);
-}
-
-FailureOr<int64_t> ValueBoundsConstraintSet::computeConstantBound(
-    presburger::BoundType type, AffineMap map, ArrayRef<Value> operands,
+    presburger::BoundType type, const Variable &var,
     StopConditionFn stopCondition, bool closedUB) {
-  ValueDimList valueDims;
-  for (Value v : operands) {
-    assert(v.getType().isIndex() && "expected index type");
-    valueDims.emplace_back(v, std::nullopt);
-  }
-  return computeConstantBound(type, map, valueDims, stopCondition, closedUB);
-}
-
-FailureOr<int64_t> ValueBoundsConstraintSet::computeConstantBound(
-    presburger::BoundType type, AffineMap map, ValueDimList operands,
-    StopConditionFn stopCondition, bool closedUB) {
-  assert(map.getNumResults() == 1 && "expected affine map with one result");
-
   // Default stop condition if none was specified: Keep adding constraints until
   // a bound could be computed.
   int64_t pos = 0;
@@ -562,8 +630,8 @@ FailureOr<int64_t> ValueBoundsConstraintSet::computeConstantBound(
   };
 
   ValueBoundsConstraintSet cstr(
-      map.getContext(), stopCondition ? stopCondition : defaultStopCondition);
-  pos = cstr.populateConstraints(map, operands);
+      var.getContext(), stopCondition ? stopCondition : defaultStopCondition);
+  pos = cstr.populateConstraints(var.map, var.mapOperands);
   assert(pos == 0 && "expected `map` is the first column");
 
   // Compute constant bound for `valueDim`.
@@ -608,22 +676,13 @@ ValueBoundsConstraintSet::computeConstantDelta(Value value1, Value value2,
   Builder b(value1.getContext());
   AffineMap map = AffineMap::get(/*dimCount=*/2, /*symbolCount=*/0,
                                  b.getAffineDimExpr(0) - b.getAffineDimExpr(1));
-  return computeConstantBound(presburger::BoundType::EQ, map,
-                              {{value1, dim1}, {value2, dim2}});
+  return computeConstantBound(presburger::BoundType::EQ,
+                              Variable(map, {{value1, dim1}, {value2, dim2}}));
 }
 
-bool ValueBoundsConstraintSet::compareValueDims(OpFoldResult lhs,
-                                                std::optional<int64_t> lhsDim,
-                                                ComparisonOperator cmp,
-                                                OpFoldResult rhs,
-                                                std::optional<int64_t> rhsDim) {
-#ifndef NDEBUG
-  if (auto lhsVal = dyn_cast<Value>(lhs))
-    assertValidValueDim(lhsVal, lhsDim);
-  if (auto rhsVal = dyn_cast<Value>(rhs))
-    assertValidValueDim(rhsVal, rhsDim);
-#endif // NDEBUG
-
+bool ValueBoundsConstraintSet::comparePos(int64_t lhsPos,
+                                          ComparisonOperator cmp,
+                                          int64_t rhsPos) {
   // This function returns "true" if "lhs CMP rhs" is proven to hold.
   //
   // Example for ComparisonOperator::LE and index-typed values: We would like to
@@ -640,50 +699,6 @@ bool ValueBoundsConstraintSet::compareValueDims(OpFoldResult lhs,
     return false;
   }
 
-  // EQ can be expressed as LE and GE.
-  if (cmp == EQ)
-    return compareValueDims(lhs, lhsDim, ComparisonOperator::LE, rhs, rhsDim) &&
-           compareValueDims(lhs, lhsDim, ComparisonOperator::GE, rhs, rhsDim);
-
-  // Construct inequality. For the above example: lhs > rhs.
-  // `IntegerRelation` inequalities are expressed in the "flattened" form and
-  // with ">= 0". I.e., lhs - rhs - 1 >= 0.
-  SmallVector<int64_t> eq(cstr.getNumCols(), 0);
-  auto addToEq = [&](OpFoldResult ofr, std::optional<int64_t> dim,
-                     int64_t factor) {
-    if (auto constVal = ::getConstantIntValue(ofr)) {
-      eq[cstr.getNumCols() - 1] += *constVal * factor;
-    } else {
-      eq[getPos(cast<Value>(ofr), dim)] += factor;
-    }
-  };
-  if (cmp == LT || cmp == LE) {
-    addToEq(lhs, lhsDim, 1);
-    addToEq(rhs, rhsDim, -1);
-  } else if (cmp == GT || cmp == GE) {
-    addToEq(lhs, lhsDim, -1);
-    addToEq(rhs, rhsDim, 1);
-  } else {
-    llvm_unreachable("unsupported comparison operator");
-  }
-  if (cmp == LE || cmp == GE)
-    eq[cstr.getNumCols() - 1] -= 1;
-
-  // Add inequality to the constraint set and check if it made the constraint
-  // set empty.
-  int64_t ineqPos = cstr.getNumInequalities();
-  cstr.addInequality(eq);
-  bool isEmpty = cstr.isEmpty();
-  cstr.removeInequality(ineqPos);
-  return isEmpty;
-}
-
-bool ValueBoundsConstraintSet::comparePos(int64_t lhsPos,
-                                          ComparisonOperator cmp,
-                                          int64_t rhsPos) {
-  // This function returns "true" if "lhs CMP rhs" is proven to hold. For
-  // detailed documentation, see `compareValueDims`.
-
   // EQ can be expressed as LE and GE.
   if (cmp == EQ)
     return comparePos(lhsPos, ComparisonOperator::LE, rhsPos) &&
@@ -712,48 +727,17 @@ bool ValueBoundsConstraintSet::comparePos(int64_t lhsPos,
   return isEmpty;
 }
 
-bool ValueBoundsConstraintSet::populateAndCompare(
-    OpFoldResult lhs, std::optional<int64_t> lhsDim, ComparisonOperator cmp,
-    OpFoldResult rhs, std::optional<int64_t> rhsDim) {
-#ifndef NDEBUG
-  if (auto lhsVal = dyn_cast<Value>(lhs))
-    assertValidValueDim(lhsVal, lhsDim);
-  if (auto rhsVal = dyn_cast<Value>(rhs))
-    assertValidValueDim(rhsVal, rhsDim);
-#endif // NDEBUG
-
-  if (auto lhsVal = dyn_cast<Value>(lhs))
-    populateConstraints(lhsVal, lhsDim);
-  if (auto rhsVal = dyn_cast<Value>(rhs))
-    populateConstraints(rhsVal, rhsDim);
-
-  return compareValueDims(lhs, lhsDim, cmp, rhs, rhsDim);
+bool ValueBoundsConstraintSet::populateAndCompare(const Variable &lhs,
+                                                  ComparisonOperator cmp,
+                                                  const Variable &rhs) {
+  int64_t lhsPos = populateConstraints(lhs.map, lhs.mapOperands);
+  int64_t rhsPos = populateConstraints(rhs.map, rhs.mapOperands);
+  return comparePos(lhsPos, cmp, rhsPos);
 }
 
-bool ValueBoundsConstraintSet::compare(OpFoldResult lhs,
-                                       std::optional<int64_t> lhsDim,
-                                       ComparisonOperator cmp, OpFoldResult rhs,
-                                       std::optional<int64_t> rhsDim) {
-  auto stopCondition = [&](Value v, std::optional<int64_t> dim,
-                           ValueBoundsConstraintSet &cstr) {
-    // Keep processing as long as lhs/rhs are not mapped.
-    if (auto lhsVal = dyn_cast<Value>(lhs))
-      if (!cstr.isMapped(lhsVal, dim))
-        return false;
-    if (auto rhsVal = dyn_cast<Value>(rhs))
-      if (!cstr.isMapped(rhsVal, dim))
-        return false;
-    // Keep processing as long as the relation cannot be proven.
-    return cstr.compareValueDims(lhs, lhsDim, cmp, rhs, rhsDim);
-  };
-
-  ValueBoundsConstraintSet cstr(lhs.getContext(), stopCondition);
-  return cstr.populateAndCompare(lhs, lhsDim, cmp, rhs, rhsDim);
-}
-
-bool ValueBoundsConstraintSet::compare(AffineMap lhs, ValueDimList lhsOperands,
-                                       ComparisonOperator cmp, AffineMap rhs,
-                                       ValueDimList rhsOperands) {
+bool ValueBoundsConstraintSet::compare(const Variable &lhs,
+                                       ComparisonOperator cmp,
+                                       const Variable &rhs) {
   int64_t lhsPos = -1, rhsPos = -1;
   auto stopCondition = [&](Value v, std::optional<int64_t> dim,
                            ValueBoundsConstraintSet &cstr) {
@@ -765,39 +749,17 @@ bool ValueBoundsConstraintSet::compare(AffineMap lhs, ValueDimList lhsOperands,
     return cstr.comparePos(lhsPos, cmp, rhsPos);
   };
   ValueBoundsConstraintSet cstr(lhs.getContext(), stopCondition);
-  lhsPos = cstr.insert(lhs, lhsOperands);
-  rhsPos = cstr.insert(rhs, rhsOperands);
-  cstr.processWorklist();
+  lhsPos = cstr.populateConstraints(lhs.map, lhs.mapOperands);
+  rhsPos = cstr.populateConstraints(rhs.map, rhs.mapOperands);
   return cstr.comparePos(lhsPos, cmp, rhsPos);
 }
 
-bool ValueBoundsConstraintSet::compare(AffineMap lhs,
-                                       ArrayRef<Value> lhsOperands,
-                                       ComparisonOperator cmp, AffineMap rhs,
-                                       ArrayRef<Value> rhsOperands) {
-  ValueDimList lhsValueDimOperands =
-      llvm::map_to_vector(lhsOperands, [](Value v) {
-        return std::make_pair(v, std::optional<int64_t>());
-      });
-  ValueDimList rhsValueDimOperands =
-      llvm::map_to_vector(rhsOperands, [](Value v) {
-        return std::make_pair(v, std::optional<int64_t>());
-      });
-  return ValueBoundsConstraintSet::compare(lhs, lhsValueDimOperands, cmp, rhs,
-                                           rhsValueDimOperands);
-}
-
-FailureOr<bool>
-ValueBoundsConstraintSet::areEqual(OpFoldResult value1, OpFoldResult value2,
-                                   std::optional<int64_t> dim1,
-                                   std::optional<int64_t> dim2) {
-  if (ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::EQ,
-                                        value2, dim2))
+FailureOr<bool> ValueBoundsConstraintSet::areEqual(const Variable &var1,
+                                                   const Variable &var2) {
+  if (ValueBoundsConstraintSet::compare(var1, ComparisonOperator::EQ, var2))
     return true;
-  if (ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::LT,
-                                        value2, dim2) ||
-      ValueBoundsConstraintSet::compare(value1, dim1, ComparisonOperator::GT,
-                                        value2, dim2))
+  if (ValueBoundsConstraintSet::compare(var1, ComparisonOperator::LT, var2) ||
+      ValueBoundsConstraintSet::compare(var1, ComparisonOperator::GT, var2))
     return false;
   return failure();
 }
@@ -833,7 +795,7 @@ ValueBoundsConstraintSet::areOverlappingSlices(MLIRContext *ctx,
       AffineMap foldedMap =
           foldAttributesIntoMap(b, map, ofrOperands, valueOperands);
       FailureOr<int64_t> constBound = computeConstantBound(
-          presburger::BoundType::EQ, foldedMap, valueOperands);
+          presburger::BoundType::EQ, Variable(foldedMap, valueOperands));
       foundUnknownBound |= failed(constBound);
       if (succeeded(constBound) && *constBound <= 0)
         return false;
@@ -850,7 +812,7 @@ ValueBoundsConstraintSet::areOverlappingSlices(MLIRContext *ctx,
       AffineMap foldedMap =
           foldAttributesIntoMap(b, map, ofrOperands, valueOperands);
       FailureOr<int64_t> constBound = computeConstantBound(
-          presburger::BoundType::EQ, foldedMap, valueOperands);
+          presburger::BoundType::EQ, Variable(foldedMap, valueOperands));
       foundUnknownBound |= failed(constBound);
       if (succeeded(constBound) && *constBound <= 0)
         return false;
diff --git a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
index 23c6872dcebe9..935c08aceff54 100644
--- a/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
+++ b/mlir/test/Dialect/Affine/value-bounds-op-interface-impl.mlir
@@ -131,3 +131,27 @@ func.func @compare_affine_min(%a: index, %b: index) {
   "test.compare"(%0, %a) {cmp = "LE"} : (index, index) -> ()
   return
 }
+
+// -----
+
+func.func @compare_const_map() {
+  %c5 = arith.constant 5 : index
+  // expected-remark @below{{true}}
+  "test.compare"(%c5) {cmp = "GT", rhs_map = affine_map<() -> (4)>}
+      : (index) -> ()
+  // expected-remark @below{{true}}
+  "test.compare"(%c5) {cmp = "LT", lhs_map = affine_map<() -> (4)>}
+      : (index) -> ()
+  return
+}
+
+// -----
+
+func.func @compare_maps(%a: index, %b: index) {
+  // expected-remark @below{{true}}
+  "test.compare"(%a, %b, %b, %a)
+      {cmp = "GT", lhs_map = affine_map<(d0, d1) -> (1 + d0 + d1)>,
+       rhs_map = affine_map<(d0, d1) -> (d0 + d1)>}
+      : (index, index, index, index) -> ()
+  return
+}
diff --git a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
index 6730f9b292ad9..b098a5a23fd31 100644
--- a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
+++ b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
@@ -109,7 +109,7 @@ static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
     FailureOr<OpFoldResult> reified = failure();
     if (constant) {
       auto reifiedConst = ValueBoundsConstraintSet::computeConstantBound(
-          boundType, value, dim, /*stopCondition=*/nullptr);
+          boundType, {value, dim}, /*stopCondition=*/nullptr);
       if (succeeded(reifiedConst))
         reified = FailureOr<OpFoldResult>(rewriter.getIndexAttr(*reifiedConst));
     } else if (scalable) {
@@ -128,22 +128,12 @@ static LogicalResult testReifyValueBounds(func::FuncOp funcOp,
             rewriter, loc, reifiedScalable->map, vscaleOperand);
       }
     } else {
-      if (dim) {
-        if (useArithOps) {
-          reified = arith::reifyShapedValueDimBound(
-              rewriter, op->getLoc(), boundType, value, *dim, stopCondition);
-        } else {
-          reified = reifyShapedValueDimBound(rewriter, op->getLoc(), boundType,
-                                             value, *dim, stopCondition);
-        }
+      if (useArithOps) {
+        reified = arith::reifyValueBound(rewriter, op->getLoc(), boundType,
+                                         op.getVariable(), stopCondition);
       } else {
-        if (useArithOps) {
-          reified = arith::reifyIndexValueBound(
-              rewriter, op->getLoc(), boundType, value, stopCondition);
-        } else {
-          reified = reifyIndexValueBound(rewriter, op->getLoc(), boundType,
-                                         value, stopCondition);
-        }
+        reified = reifyValueBound(rewriter, op->getLoc(), boundType,
+                                  op.getVariable(), stopCondition);
       }
     }
     if (failed(reified)) {
@@ -188,9 +178,7 @@ static LogicalResult testEquality(func::FuncOp funcOp) {
     }
 
     auto compare = [&](ValueBoundsConstraintSet::ComparisonOperator cmp) {
-      return ValueBoundsConstraintSet::compare(
-          /*lhs=*/op.getLhs(), /*lhsDim=*/std::nullopt, cmp,
-          /*rhs=*/op.getRhs(), /*rhsDim=*/std::nullopt);
+      return ValueBoundsConstraintSet::compare(op.getLhs(), cmp, op.getRhs());
     };
     if (compare(cmpType)) {
       op->emitRemark("true");
diff --git a/mlir/test/lib/Dialect/Test/TestDialect.cpp b/mlir/test/lib/Dialect/Test/TestDialect.cpp
index 25c5190ca0ef3..36d7606fe1345 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialect.cpp
@@ -549,6 +549,12 @@ LogicalResult ReifyBoundOp::verify() {
   return success();
 }
 
+::mlir::ValueBoundsConstraintSet::Variable ReifyBoundOp::getVariable() {
+  if (getDim().has_value())
+    return ValueBoundsConstraintSet::Variable(getVar(), *getDim());
+  return ValueBoundsConstraintSet::Variable(getVar());
+}
+
 ::mlir::ValueBoundsConstraintSet::ComparisonOperator
 CompareOp::getComparisonOperator() {
   if (getCmp() == "EQ")
@@ -564,6 +570,37 @@ CompareOp::getComparisonOperator() {
   llvm_unreachable("invalid comparison operator");
 }
 
+::mlir::ValueBoundsConstraintSet::Variable CompareOp::getLhs() {
+  if (!getLhsMap())
+    return ValueBoundsConstraintSet::Variable(getVarOperands()[0]);
+  SmallVector<Value> mapOperands(
+      getVarOperands().slice(0, getLhsMap()->getNumInputs()));
+  return ValueBoundsConstraintSet::Variable(*getLhsMap(), mapOperands);
+}
+
+::mlir::ValueBoundsConstraintSet::Variable CompareOp::getRhs() {
+  int64_t rhsOperandsBegin = getLhsMap() ? getLhsMap()->getNumInputs() : 1;
+  if (!getRhsMap())
+    return ValueBoundsConstraintSet::Variable(
+        getVarOperands()[rhsOperandsBegin]);
+  SmallVector<Value> mapOperands(
+      getVarOperands().slice(rhsOperandsBegin, getRhsMap()->getNumInputs()));
+  return ValueBoundsConstraintSet::Variable(*getRhsMap(), mapOperands);
+}
+
+LogicalResult CompareOp::verify() {
+  if (getCompose() && (getLhsMap() || getRhsMap()))
+    return emitOpError(
+        "'compose' not supported when 'lhs_map' or 'rhs_map' is present");
+  int64_t expectedNumOperands = getLhsMap() ? getLhsMap()->getNumInputs() : 1;
+  expectedNumOperands += getRhsMap() ? getRhsMap()->getNumInputs() : 1;
+  if (getVarOperands().size() != expectedNumOperands)
+    return emitOpError("expected ")
+           << expectedNumOperands << " operands, but got "
+           << getVarOperands().size();
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // Test removing op with inner ops.
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/lib/Dialect/Test/TestOps.td b/mlir/test/lib/Dialect/Test/TestOps.td
index ebf158b8bb820..b641b3da719c7 100644
--- a/mlir/test/lib/Dialect/Test/TestOps.td
+++ b/mlir/test/lib/Dialect/Test/TestOps.td
@@ -2207,6 +2207,7 @@ def ReifyBoundOp : TEST_Op<"reify_bound", [Pure]> {
 
   let extraClassDeclaration = [{
     ::mlir::presburger::BoundType getBoundType();
+    ::mlir::ValueBoundsConstraintSet::Variable getVariable();
   }];
 
   let hasVerifier = 1;
@@ -2217,18 +2218,29 @@ def CompareOp : TEST_Op<"compare"> {
     Compare `lhs` and `rhs`. A remark is emitted which indicates whether the
     specified comparison operator was proven to hold. The remark also indicates
     whether the opposite comparison operator was proven to hold.
+
+    `var_operands` must have exactly two operands: one for the LHS operand and
+    one for the RHS operand. If `lhs_map` is specified, as many operands as
+    `lhs_map` has inputs are expected instead of the first operand. If `rhs_map`
+    is specified, as many operands as `rhs_map` has inputs are expected instead
+    of the second operand.
   }];
 
-  let arguments = (ins Index:$lhs,
-                       Index:$rhs,
+  let arguments = (ins Variadic<Index>:$var_operands,
                        DefaultValuedAttr<StrAttr, "\"EQ\"">:$cmp,
+                       OptionalAttr<AffineMapAttr>:$lhs_map,
+                       OptionalAttr<AffineMapAttr>:$rhs_map,
                        UnitAttr:$compose);
   let results = (outs);
 
   let extraClassDeclaration = [{
     ::mlir::ValueBoundsConstraintSet::ComparisonOperator
         getComparisonOperator();
+    ::mlir::ValueBoundsConstraintSet::Variable getLhs();
+    ::mlir::ValueBoundsConstraintSet::Variable getRhs();
   }];
+
+  let hasVerifier = 1;
 }
 
 //===----------------------------------------------------------------------===//

From 5a46123ddf62900d3dc73330f699c73038645198 Mon Sep 17 00:00:00 2001
From: Utkarsh Saxena <usx@google.com>
Date: Tue, 16 Apr 2024 11:01:03 +0200
Subject: [PATCH 540/886] Fix missing dtor in function calls accepting trivial
 ABI structs (#88751)

Fixes https://github.com/llvm/llvm-project/issues/88478

Promoting the `EHCleanup` to `NormalAndEHCleanup` in `EmitCallArgs`
surfaced another bug with deactivation of normal cleanups. Here we
missed emitting CPP scope ends for deactivated normal cleanups. This
patch also fixes that bug.

We missed emitting CPP scope ends because we remove the `fallthrough`
(clears the insertion point) before deactivating normal cleanups. This
is to make the emitted "normal" cleanup code unreachable. But we still
need to emit CPP scope ends in the original basic block even for a
deactivated normal cleanup.
(This worked correctly before we did not remove `fallthrough` for
`EHCleanup`s).
---
 clang/lib/CodeGen/CGCall.cpp                  | 13 ++++---
 clang/lib/CodeGen/CGCleanup.cpp               | 37 ++++++++++++-------
 clang/lib/CodeGen/CodeGenFunction.h           |  3 +-
 .../CodeGenCXX/control-flow-in-stmt-expr.cpp  | 16 ++++++++
 4 files changed, 48 insertions(+), 21 deletions(-)

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 7a0bc6fa77b88..0c860a3ccbd2f 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -4694,11 +4694,11 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
     AggValueSlot Slot = args.isUsingInAlloca()
         ? createPlaceholderSlot(*this, type) : CreateAggTemp(type, "agg.tmp");
 
-    bool DestroyedInCallee = true, NeedsEHCleanup = true;
+    bool DestroyedInCallee = true, NeedsCleanup = true;
     if (const auto *RD = type->getAsCXXRecordDecl())
       DestroyedInCallee = RD->hasNonTrivialDestructor();
     else
-      NeedsEHCleanup = needsEHCleanup(type.isDestructedType());
+      NeedsCleanup = type.isDestructedType();
 
     if (DestroyedInCallee)
       Slot.setExternallyDestructed();
@@ -4707,14 +4707,15 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
     RValue RV = Slot.asRValue();
     args.add(RV, type);
 
-    if (DestroyedInCallee && NeedsEHCleanup) {
+    if (DestroyedInCallee && NeedsCleanup) {
       // Create a no-op GEP between the placeholder and the cleanup so we can
       // RAUW it successfully.  It also serves as a marker of the first
       // instruction where the cleanup is active.
-      pushFullExprCleanup<DestroyUnpassedArg>(EHCleanup, Slot.getAddress(),
-                                              type);
+      pushFullExprCleanup<DestroyUnpassedArg>(NormalAndEHCleanup,
+                                              Slot.getAddress(), type);
       // This unreachable is a temporary marker which will be removed later.
-      llvm::Instruction *IsActive = Builder.CreateUnreachable();
+      llvm::Instruction *IsActive =
+          Builder.CreateFlagLoad(llvm::Constant::getNullValue(Int8PtrTy));
       args.addArgCleanupDeactivation(EHStack.stable_begin(), IsActive);
     }
     return;
diff --git a/clang/lib/CodeGen/CGCleanup.cpp b/clang/lib/CodeGen/CGCleanup.cpp
index 5bf48bc22a549..8683f19d9da28 100644
--- a/clang/lib/CodeGen/CGCleanup.cpp
+++ b/clang/lib/CodeGen/CGCleanup.cpp
@@ -634,12 +634,19 @@ static void destroyOptimisticNormalEntry(CodeGenFunction &CGF,
 /// Pops a cleanup block.  If the block includes a normal cleanup, the
 /// current insertion point is threaded through the cleanup, as are
 /// any branch fixups on the cleanup.
-void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
+void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
+                                      bool ForDeactivation) {
   assert(!EHStack.empty() && "cleanup stack is empty!");
   assert(isa<EHCleanupScope>(*EHStack.begin()) && "top not a cleanup!");
   EHCleanupScope &Scope = cast<EHCleanupScope>(*EHStack.begin());
   assert(Scope.getFixupDepth() <= EHStack.getNumBranchFixups());
 
+  // If we are deactivating a normal cleanup, we need to pretend that the
+  // fallthrough is unreachable. We restore this IP before returning.
+  CGBuilderTy::InsertPoint NormalDeactivateOrigIP;
+  if (ForDeactivation && (Scope.isNormalCleanup() || !getLangOpts().EHAsynch)) {
+    NormalDeactivateOrigIP = Builder.saveAndClearIP();
+  }
   // Remember activation information.
   bool IsActive = Scope.isActive();
   Address NormalActiveFlag =
@@ -729,6 +736,8 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
     EHStack.popCleanup(); // safe because there are no fixups
     assert(EHStack.getNumBranchFixups() == 0 ||
            EHStack.hasNormalCleanups());
+    if (NormalDeactivateOrigIP.isSet())
+      Builder.restoreIP(NormalDeactivateOrigIP);
     return;
   }
 
@@ -765,9 +774,16 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
   if (!RequiresNormalCleanup) {
     // Mark CPP scope end for passed-by-value Arg temp
     //   per Windows ABI which is "normally" Cleanup in callee
-    if (IsEHa && getInvokeDest() && Builder.GetInsertBlock()) {
-      if (Personality.isMSVCXXPersonality())
+    if (IsEHa && getInvokeDest()) {
+      // If we are deactivating a normal cleanup then we don't have a
+      // fallthrough. Restore original IP to emit CPP scope ends in the correct
+      // block.
+      if (NormalDeactivateOrigIP.isSet())
+        Builder.restoreIP(NormalDeactivateOrigIP);
+      if (Personality.isMSVCXXPersonality() && Builder.GetInsertBlock())
         EmitSehCppScopeEnd();
+      if (NormalDeactivateOrigIP.isSet())
+        NormalDeactivateOrigIP = Builder.saveAndClearIP();
     }
     destroyOptimisticNormalEntry(*this, Scope);
     Scope.MarkEmitted();
@@ -992,6 +1008,8 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
     }
   }
 
+  if (NormalDeactivateOrigIP.isSet())
+    Builder.restoreIP(NormalDeactivateOrigIP);
   assert(EHStack.hasNormalCleanups() || EHStack.getNumBranchFixups() == 0);
 
   // Emit the EH cleanup if required.
@@ -1281,17 +1299,8 @@ void CodeGenFunction::DeactivateCleanupBlock(EHScopeStack::stable_iterator C,
   // to the current RunCleanupsScope.
   if (C == EHStack.stable_begin() &&
       CurrentCleanupScopeDepth.strictlyEncloses(C)) {
-    // Per comment below, checking EHAsynch is not really necessary
-    // it's there to assure zero-impact w/o EHAsynch option
-    if (!Scope.isNormalCleanup() && getLangOpts().EHAsynch) {
-      PopCleanupBlock();
-    } else {
-      // If it's a normal cleanup, we need to pretend that the
-      // fallthrough is unreachable.
-      CGBuilderTy::InsertPoint SavedIP = Builder.saveAndClearIP();
-      PopCleanupBlock();
-      Builder.restoreIP(SavedIP);
-    }
+    PopCleanupBlock(/*FallthroughIsBranchThrough=*/false,
+                    /*ForDeactivation=*/true);
     return;
   }
 
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index c49e9fd00c8d3..d99188671f1f6 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -957,7 +957,8 @@ class CodeGenFunction : public CodeGenTypeCache {
 
   /// PopCleanupBlock - Will pop the cleanup entry on the stack and
   /// process all branch fixups.
-  void PopCleanupBlock(bool FallThroughIsBranchThrough = false);
+  void PopCleanupBlock(bool FallThroughIsBranchThrough = false,
+                       bool ForDeactivation = false);
 
   /// DeactivateCleanupBlock - Deactivates the given cleanup block.
   /// The block cannot be reactivated.  Pops it if it's the top of the
diff --git a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp b/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
index 95deee8bb1f1f..0a51b0e4121c3 100644
--- a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
+++ b/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
@@ -391,3 +391,19 @@ void ArrayInitWithContinue() {
                      })};
   }
 }
+
+struct [[clang::trivial_abi]] HasTrivialABI {
+  HasTrivialABI();
+  ~HasTrivialABI();
+};
+void AcceptTrivialABI(HasTrivialABI, int);
+void TrivialABI() {
+  // CHECK-LABEL: define dso_local void @_Z10TrivialABIv()
+  AcceptTrivialABI(HasTrivialABI(), ({
+                     if (foo()) return;
+                     // CHECK:      if.then:
+                     // CHECK-NEXT:   call void @_ZN13HasTrivialABID1Ev
+                     // CHECK-NEXT:   br label %return
+                     0;
+                   }));
+}

From 9141e1c24f87e5735bc4178a018eba4bdf2750aa Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Fri, 12 Apr 2024 10:43:14 +0200
Subject: [PATCH 541/886] [clang][Interp] Gracefully handle bitcasts to
 non-primitive types

We were calling classfiyPrim() instead of classify().
---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 2 +-
 clang/test/AST/Interp/vectors.cpp        | 8 +++++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 01ec31e4077f7..5866228663dca 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -262,7 +262,7 @@ bool ByteCodeExprGen<Emitter>::VisitCastExpr(const CastExpr *CE) {
       return this->discard(SubExpr);
 
     std::optional<PrimType> FromT = classify(SubExpr->getType());
-    std::optional<PrimType> ToT = classifyPrim(CE->getType());
+    std::optional<PrimType> ToT = classify(CE->getType());
     if (!FromT || !ToT)
       return false;
 
diff --git a/clang/test/AST/Interp/vectors.cpp b/clang/test/AST/Interp/vectors.cpp
index 8afef3c897bff..fb5787a9eda9a 100644
--- a/clang/test/AST/Interp/vectors.cpp
+++ b/clang/test/AST/Interp/vectors.cpp
@@ -1,7 +1,7 @@
 // RUN: %clang_cc1 -fexperimental-new-constant-interpreter -verify=expected,both %s
 // RUN: %clang_cc1 -verify=ref,both %s
 
-// both-no-diagnostics
+// ref-no-diagnostics
 
 typedef int __attribute__((vector_size(16))) VI4;
 constexpr VI4 A = {1,2,3,4};
@@ -20,3 +20,9 @@ namespace Vector {
   }
   constexpr auto v2 = g(4);
 }
+
+/// FIXME: We need to support BitCasts between vector types.
+namespace {
+  typedef float __attribute__((vector_size(16))) VI42;
+  constexpr VI42 A2 = A; // expected-error {{must be initialized by a constant expression}}
+}

From 4fc0a99b8f220b6b41648da491bcc81a067f1600 Mon Sep 17 00:00:00 2001
From: Jay Foad <jay.foad@amd.com>
Date: Tue, 16 Apr 2024 10:14:15 +0100
Subject: [PATCH 542/886] [AMDGPU] Fix implicit operands of VOPD cndmask
 instructions (#87788)

---
 llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp  |  1 +
 llvm/test/CodeGen/AMDGPU/vopd-combine.mir | 32 +++++++++++------------
 2 files changed, 17 insertions(+), 16 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp b/llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp
index 05e10a95b157c..1dda1b89b2d36 100644
--- a/llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp
@@ -101,6 +101,7 @@ class GCNCreateVOPD : public MachineFunctionPass {
       }
     }
 
+    SII->fixImplicitOperands(*VOPDInst);
     for (auto CompIdx : VOPD::COMPONENTS)
       VOPDInst.copyImplicitOps(*MI[CompIdx]);
 
diff --git a/llvm/test/CodeGen/AMDGPU/vopd-combine.mir b/llvm/test/CodeGen/AMDGPU/vopd-combine.mir
index 63bef40c34742..b8ac50c3aeb5e 100644
--- a/llvm/test/CodeGen/AMDGPU/vopd-combine.mir
+++ b/llvm/test/CodeGen/AMDGPU/vopd-combine.mir
@@ -160,7 +160,7 @@ body:             |
     ; PAIR-GFX11-NEXT: $vgpr3 = IMPLICIT_DEF
     ; PAIR-GFX11-NEXT: $sgpr20 = IMPLICIT_DEF
     ; PAIR-GFX11-NEXT: $vgpr4 = V_FMAMK_F32 $sgpr20, 12345, $vgpr3, implicit $mode, implicit $exec
-    ; PAIR-GFX11-NEXT: $vgpr2, $vgpr5 = V_DUAL_FMAC_F32_e32_X_CNDMASK_B32_e32_gfx11 $sgpr20, killed $vgpr1, killed $vgpr2, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX11-NEXT: $vgpr2, $vgpr5 = V_DUAL_FMAC_F32_e32_X_CNDMASK_B32_e32_gfx11 $sgpr20, killed $vgpr1, killed $vgpr2, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX11-NEXT: $vgpr7 = V_CNDMASK_B32_e32 killed $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX11-NEXT: $vgpr6 = V_ADD_F32_e32 $sgpr20, $vgpr3, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr9 = V_CNDMASK_B32_e32 killed $sgpr20, killed $vgpr3, implicit $mode, implicit $exec, implicit killed $vcc_lo
@@ -174,7 +174,7 @@ body:             |
     ; PAIR-GFX12-NEXT: $vgpr3 = IMPLICIT_DEF
     ; PAIR-GFX12-NEXT: $sgpr20 = IMPLICIT_DEF
     ; PAIR-GFX12-NEXT: $vgpr4 = V_FMAMK_F32 $sgpr20, 12345, $vgpr3, implicit $mode, implicit $exec
-    ; PAIR-GFX12-NEXT: $vgpr2, $vgpr5 = V_DUAL_FMAC_F32_e32_X_CNDMASK_B32_e32_gfx12 $sgpr20, killed $vgpr1, killed $vgpr2, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX12-NEXT: $vgpr2, $vgpr5 = V_DUAL_FMAC_F32_e32_X_CNDMASK_B32_e32_gfx12 $sgpr20, killed $vgpr1, killed $vgpr2, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX12-NEXT: $vgpr7 = V_CNDMASK_B32_e32 killed $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX12-NEXT: $vgpr6 = V_ADD_F32_e32 $sgpr20, $vgpr3, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr9 = V_CNDMASK_B32_e32 killed $sgpr20, killed $vgpr3, implicit $mode, implicit $exec, implicit killed $vcc_lo
@@ -458,9 +458,9 @@ body:             |
     ; PAIR-GFX11-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr2 = V_FMAC_F32_e32 10, $vgpr1, killed $vgpr2, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
-    ; PAIR-GFX11-NEXT: $vgpr12, $vgpr19 = V_DUAL_ADD_F32_e32_X_CNDMASK_B32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX11-NEXT: $vgpr12, $vgpr19 = V_DUAL_ADD_F32_e32_X_CNDMASK_B32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX11-NEXT: $vgpr11 = V_CNDMASK_B32_e32 $vgpr0, killed $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX11-NEXT: $vgpr17, $vgpr10 = V_DUAL_MUL_F32_e32_X_CNDMASK_B32_e32_gfx11 killed $vgpr0, $vgpr0, $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX11-NEXT: $vgpr17, $vgpr10 = V_DUAL_MUL_F32_e32_X_CNDMASK_B32_e32_gfx11 killed $vgpr0, $vgpr0, $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX11-NEXT: $vgpr15 = V_CNDMASK_B32_e32 $vgpr1, killed $vgpr2, implicit $mode, implicit $exec, implicit killed $vcc_lo
     ; PAIR-GFX11-NEXT: $vgpr16 = V_SUB_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr14 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
@@ -476,9 +476,9 @@ body:             |
     ; PAIR-GFX12-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr2 = V_FMAC_F32_e32 10, $vgpr1, killed $vgpr2, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
-    ; PAIR-GFX12-NEXT: $vgpr12, $vgpr19 = V_DUAL_ADD_F32_e32_X_CNDMASK_B32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX12-NEXT: $vgpr12, $vgpr19 = V_DUAL_ADD_F32_e32_X_CNDMASK_B32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX12-NEXT: $vgpr11 = V_CNDMASK_B32_e32 $vgpr0, killed $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX12-NEXT: $vgpr17, $vgpr10 = V_DUAL_MUL_F32_e32_X_CNDMASK_B32_e32_gfx12 killed $vgpr0, $vgpr0, $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX12-NEXT: $vgpr17, $vgpr10 = V_DUAL_MUL_F32_e32_X_CNDMASK_B32_e32_gfx12 killed $vgpr0, $vgpr0, $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
     ; PAIR-GFX12-NEXT: $vgpr15 = V_CNDMASK_B32_e32 $vgpr1, killed $vgpr2, implicit $mode, implicit $exec, implicit killed $vcc_lo
     ; PAIR-GFX12-NEXT: $vgpr16 = V_SUB_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr14 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
@@ -559,12 +559,12 @@ body:             |
     ; PAIR-GFX11-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr2 = V_FMAC_F32_e32 10, $vgpr1, killed $vgpr2, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
-    ; PAIR-GFX11-NEXT: $vgpr4, $vgpr29 = V_DUAL_SUB_F32_e32_X_CNDMASK_B32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX11-NEXT: $vgpr19, $vgpr20 = V_DUAL_CNDMASK_B32_e32_X_FMAC_F32_e32_gfx11 $vgpr0, $vgpr3, 10, $vgpr1, killed $vgpr20, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX11-NEXT: $vgpr4, $vgpr29 = V_DUAL_SUB_F32_e32_X_CNDMASK_B32_e32_gfx11 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX11-NEXT: $vgpr19, $vgpr20 = V_DUAL_CNDMASK_B32_e32_X_FMAC_F32_e32_gfx11 $vgpr0, $vgpr3, 10, $vgpr1, killed $vgpr20, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr15 = V_CNDMASK_B32_e32 $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX11-NEXT: $vgpr10, $vgpr17 = V_DUAL_CNDMASK_B32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr2, $vgpr0, $vgpr0, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
-    ; PAIR-GFX11-NEXT: $vgpr11, $vgpr12 = V_DUAL_CNDMASK_B32_e32_X_ADD_F32_e32_gfx11 $vgpr0, $vgpr3, $vgpr1, $vgpr1, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
-    ; PAIR-GFX11-NEXT: $vgpr37, $vgpr14 = V_DUAL_CNDMASK_B32_e32_X_SUB_F32_e32_gfx11 $vgpr0, killed $vgpr3, $vgpr1, $vgpr1, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX11-NEXT: $vgpr10, $vgpr17 = V_DUAL_CNDMASK_B32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr2, $vgpr0, $vgpr0, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX11-NEXT: $vgpr11, $vgpr12 = V_DUAL_CNDMASK_B32_e32_X_ADD_F32_e32_gfx11 $vgpr0, $vgpr3, $vgpr1, $vgpr1, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX11-NEXT: $vgpr37, $vgpr14 = V_DUAL_CNDMASK_B32_e32_X_SUB_F32_e32_gfx11 $vgpr0, killed $vgpr3, $vgpr1, $vgpr1, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr20 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr21, $vgpr24 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr1, killed $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX11-NEXT: $vgpr28 = V_CNDMASK_B32_e32 $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo
@@ -586,12 +586,12 @@ body:             |
     ; PAIR-GFX12-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr2 = V_FMAC_F32_e32 10, $vgpr1, killed $vgpr2, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
-    ; PAIR-GFX12-NEXT: $vgpr4, $vgpr29 = V_DUAL_SUB_F32_e32_X_CNDMASK_B32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX12-NEXT: $vgpr19, $vgpr20 = V_DUAL_CNDMASK_B32_e32_X_FMAC_F32_e32_gfx12 $vgpr0, $vgpr3, 10, $vgpr1, killed $vgpr20, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX12-NEXT: $vgpr4, $vgpr29 = V_DUAL_SUB_F32_e32_X_CNDMASK_B32_e32_gfx12 $vgpr1, $vgpr1, $vgpr0, $vgpr3, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $vcc_lo
+    ; PAIR-GFX12-NEXT: $vgpr19, $vgpr20 = V_DUAL_CNDMASK_B32_e32_X_FMAC_F32_e32_gfx12 $vgpr0, $vgpr3, 10, $vgpr1, killed $vgpr20, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr15 = V_CNDMASK_B32_e32 $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo
-    ; PAIR-GFX12-NEXT: $vgpr10, $vgpr17 = V_DUAL_CNDMASK_B32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr2, $vgpr0, $vgpr0, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
-    ; PAIR-GFX12-NEXT: $vgpr11, $vgpr12 = V_DUAL_CNDMASK_B32_e32_X_ADD_F32_e32_gfx12 $vgpr0, $vgpr3, $vgpr1, $vgpr1, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
-    ; PAIR-GFX12-NEXT: $vgpr37, $vgpr14 = V_DUAL_CNDMASK_B32_e32_X_SUB_F32_e32_gfx12 $vgpr0, killed $vgpr3, $vgpr1, $vgpr1, implicit $vcc, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX12-NEXT: $vgpr10, $vgpr17 = V_DUAL_CNDMASK_B32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr2, $vgpr0, $vgpr0, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX12-NEXT: $vgpr11, $vgpr12 = V_DUAL_CNDMASK_B32_e32_X_ADD_F32_e32_gfx12 $vgpr0, $vgpr3, $vgpr1, $vgpr1, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
+    ; PAIR-GFX12-NEXT: $vgpr37, $vgpr14 = V_DUAL_CNDMASK_B32_e32_X_SUB_F32_e32_gfx12 $vgpr0, killed $vgpr3, $vgpr1, $vgpr1, implicit $vcc_lo, implicit $exec, implicit $mode, implicit $mode, implicit $exec, implicit $vcc_lo, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr20 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr21, $vgpr24 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr1, killed $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
     ; PAIR-GFX12-NEXT: $vgpr28 = V_CNDMASK_B32_e32 $vgpr1, $vgpr2, implicit $mode, implicit $exec, implicit $vcc_lo

From f4f772ceef379bd434d266b6e0d2bbdf796f81cb Mon Sep 17 00:00:00 2001
From: Jay Foad <jay.foad@amd.com>
Date: Tue, 16 Apr 2024 10:15:00 +0100
Subject: [PATCH 543/886] [AMDGPU] Stop reserving $vcc_hi in wave32 mode
 (#87783)

This gives us one extra SGPR to play with. The comment suggested that it
could cause bugs, but I have tested it with Vulkan CTS with the default
wave size for compute shaders set to 32 and did not find any problems.
---
 llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp     |  7 --
 llvm/test/CodeGen/AMDGPU/bf16.ll              | 22 +++--
 .../CodeGen/AMDGPU/llvm.is.fpclass.f16.ll     | 20 ++---
 llvm/test/CodeGen/AMDGPU/load-constant-i16.ll | 22 ++---
 llvm/test/CodeGen/AMDGPU/load-constant-i8.ll  | 90 +++++++++----------
 .../AMDGPU/sgpr-spill-overlap-wwm-reserve.mir | 26 +++---
 6 files changed, 88 insertions(+), 99 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index 245731ad5fc7c..acb54fd10b90d 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -612,13 +612,6 @@ BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
   // Reserve null register - it shall never be allocated
   reserveRegisterTuples(Reserved, AMDGPU::SGPR_NULL64);
 
-  // Disallow vcc_hi allocation in wave32. It may be allocated but most likely
-  // will result in bugs.
-  if (isWave32) {
-    Reserved.set(AMDGPU::VCC);
-    Reserved.set(AMDGPU::VCC_HI);
-  }
-
   // Reserve SGPRs.
   //
   unsigned MaxNumSGPRs = ST.getMaxNumSGPRs(MF);
diff --git a/llvm/test/CodeGen/AMDGPU/bf16.ll b/llvm/test/CodeGen/AMDGPU/bf16.ll
index bf4302c156d83..4c9c34de7194c 100644
--- a/llvm/test/CodeGen/AMDGPU/bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/bf16.ll
@@ -38342,12 +38342,11 @@ define <32 x bfloat> @v_vselect_v32bf16(<32 x i1> %cond, <32 x bfloat> %a, <32 x
 ; GFX10-NEXT:    v_and_b32_e32 v2, 1, v2
 ; GFX10-NEXT:    v_and_b32_e32 v4, 1, v4
 ; GFX10-NEXT:    v_and_b32_e32 v6, 1, v6
-; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
 ; GFX10-NEXT:    v_and_b32_e32 v8, 1, v8
 ; GFX10-NEXT:    v_and_b32_e32 v10, 1, v10
+; GFX10-NEXT:    v_writelane_b32 v40, s31, 1
 ; GFX10-NEXT:    v_and_b32_e32 v1, 1, v1
 ; GFX10-NEXT:    v_and_b32_e32 v3, 1, v3
-; GFX10-NEXT:    v_writelane_b32 v40, s34, 2
 ; GFX10-NEXT:    v_and_b32_e32 v5, 1, v5
 ; GFX10-NEXT:    v_and_b32_e32 v7, 1, v7
 ; GFX10-NEXT:    v_and_b32_e32 v9, 1, v9
@@ -38366,7 +38365,7 @@ define <32 x bfloat> @v_vselect_v32bf16(<32 x i1> %cond, <32 x bfloat> %a, <32 x
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s17, 1, v4
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s18, 1, v2
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s19, 1, v0
-; GFX10-NEXT:    v_writelane_b32 v40, s35, 3
+; GFX10-NEXT:    v_writelane_b32 v40, s34, 2
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s20, 1, v27
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s21, 1, v25
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s22, 1, v23
@@ -38377,10 +38376,10 @@ define <32 x bfloat> @v_vselect_v32bf16(<32 x i1> %cond, <32 x bfloat> %a, <32 x
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s27, 1, v13
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s28, 1, v11
 ; GFX10-NEXT:    v_cmp_eq_u32_e64 s29, 1, v7
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s30, 1, v3
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s31, 1, v1
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s34, 1, v5
-; GFX10-NEXT:    v_cmp_eq_u32_e64 s35, 1, v9
+; GFX10-NEXT:    v_cmp_eq_u32_e64 vcc_hi, 1, v3
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s30, 1, v1
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s31, 1, v5
+; GFX10-NEXT:    v_cmp_eq_u32_e64 s34, 1, v9
 ; GFX10-NEXT:    s_waitcnt vmcnt(32)
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 16, v31
 ; GFX10-NEXT:    s_waitcnt vmcnt(31)
@@ -38460,10 +38459,10 @@ define <32 x bfloat> @v_vselect_v32bf16(<32 x i1> %cond, <32 x bfloat> %a, <32 x
 ; GFX10-NEXT:    v_cndmask_b32_e64 v6, v29, v39, s27
 ; GFX10-NEXT:    v_cndmask_b32_e64 v5, v28, v26, s28
 ; GFX10-NEXT:    v_cndmask_b32_e64 v20, v51, v20, s29
-; GFX10-NEXT:    v_cndmask_b32_e64 v0, v14, v12, s31
-; GFX10-NEXT:    v_cndmask_b32_e64 v1, v55, v16, s30
-; GFX10-NEXT:    v_cndmask_b32_e64 v2, v53, v18, s34
-; GFX10-NEXT:    v_cndmask_b32_e64 v12, v24, v22, s35
+; GFX10-NEXT:    v_cndmask_b32_e64 v0, v14, v12, s30
+; GFX10-NEXT:    v_cndmask_b32_e64 v1, v55, v16, vcc_hi
+; GFX10-NEXT:    v_cndmask_b32_e64 v2, v53, v18, s31
+; GFX10-NEXT:    v_cndmask_b32_e64 v12, v24, v22, s34
 ; GFX10-NEXT:    v_cndmask_b32_e64 v16, v4, v3, s4
 ; GFX10-NEXT:    v_perm_b32 v0, v0, v64, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v1, v1, v54, 0x5040100
@@ -38481,7 +38480,6 @@ define <32 x bfloat> @v_vselect_v32bf16(<32 x i1> %cond, <32 x bfloat> %a, <32 x
 ; GFX10-NEXT:    v_perm_b32 v13, v66, v13, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v14, v65, v17, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v15, v16, v15, 0x5040100
-; GFX10-NEXT:    v_readlane_b32 s35, v40, 3
 ; GFX10-NEXT:    v_readlane_b32 s34, v40, 2
 ; GFX10-NEXT:    v_readlane_b32 s31, v40, 1
 ; GFX10-NEXT:    v_readlane_b32 s30, v40, 0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
index ec3c08ec79523..da64c379672ef 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
@@ -1259,17 +1259,17 @@ define <4 x i1> @isnan_v4f16(<4 x half> %x) nounwind {
 ; GFX10SELDAG-LABEL: isnan_v4f16:
 ; GFX10SELDAG:       ; %bb.0:
 ; GFX10SELDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX10SELDAG-NEXT:    v_mov_b32_e32 v2, 3
-; GFX10SELDAG-NEXT:    v_cmp_class_f16_e64 s5, v0, 3
-; GFX10SELDAG-NEXT:    v_cmp_class_f16_sdwa s4, v1, v2 src0_sel:WORD_1 src1_sel:DWORD
-; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v4, 0, 1, s5
-; GFX10SELDAG-NEXT:    v_cmp_class_f16_sdwa s5, v0, v2 src0_sel:WORD_1 src1_sel:DWORD
+; GFX10SELDAG-NEXT:    v_cmp_class_f16_e64 s4, v0, 3
+; GFX10SELDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v5, 0, 1, s4
+; GFX10SELDAG-NEXT:    v_cmp_class_f16_e64 s4, v1, 3
+; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v2, 0, 1, s4
+; GFX10SELDAG-NEXT:    v_cmp_class_f16_sdwa s4, v0, v3 src0_sel:WORD_1 src1_sel:DWORD
+; GFX10SELDAG-NEXT:    v_mov_b32_e32 v0, v5
+; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v4, 0, 1, s4
+; GFX10SELDAG-NEXT:    v_cmp_class_f16_sdwa s4, v1, v3 src0_sel:WORD_1 src1_sel:DWORD
+; GFX10SELDAG-NEXT:    v_mov_b32_e32 v1, v4
 ; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v3, 0, 1, s4
-; GFX10SELDAG-NEXT:    v_mov_b32_e32 v0, v4
-; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v5, 0, 1, s5
-; GFX10SELDAG-NEXT:    v_cmp_class_f16_e64 s5, v1, 3
-; GFX10SELDAG-NEXT:    v_mov_b32_e32 v1, v5
-; GFX10SELDAG-NEXT:    v_cndmask_b32_e64 v2, 0, 1, s5
 ; GFX10SELDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10GLISEL-LABEL: isnan_v4f16:
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
index ab6a9dcf71ace..a87fa8bf36d9e 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i16.ll
@@ -7404,35 +7404,35 @@ define amdgpu_kernel void @constant_sextload_v16i16_to_v16i64(ptr addrspace(1) %
 ; GFX12-NEXT:    v_dual_mov_b32 v4, s22 :: v_dual_mov_b32 v9, s31
 ; GFX12-NEXT:    v_dual_mov_b32 v8, s30 :: v_dual_mov_b32 v11, s35
 ; GFX12-NEXT:    v_dual_mov_b32 v10, s34 :: v_dual_mov_b32 v3, s5
-; GFX12-NEXT:    s_bfe_i64 s[10:11], s[0:1], 0x100000
-; GFX12-NEXT:    s_lshr_b32 s12, s0, 16
-; GFX12-NEXT:    s_mov_b32 s14, s1
-; GFX12-NEXT:    s_lshr_b32 s16, s1, 16
-; GFX12-NEXT:    s_bfe_i64 s[0:1], s[2:3], 0x100000
+; GFX12-NEXT:    s_bfe_i64 s[16:17], s[2:3], 0x100000
 ; GFX12-NEXT:    s_lshr_b32 s2, s2, 16
 ; GFX12-NEXT:    s_bfe_i64 s[6:7], s[6:7], 0x100000
 ; GFX12-NEXT:    v_dual_mov_b32 v0, s28 :: v_dual_mov_b32 v5, s23
 ; GFX12-NEXT:    v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v13, s25
+; GFX12-NEXT:    s_mov_b32 s12, s1
+; GFX12-NEXT:    s_lshr_b32 s14, s1, 16
 ; GFX12-NEXT:    s_bfe_i64 s[18:19], s[18:19], 0x100000
 ; GFX12-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x100000
 ; GFX12-NEXT:    v_dual_mov_b32 v12, s24 :: v_dual_mov_b32 v15, s27
 ; GFX12-NEXT:    v_dual_mov_b32 v14, s26 :: v_dual_mov_b32 v7, s7
+; GFX12-NEXT:    s_bfe_i64 s[10:11], s[0:1], 0x100000
+; GFX12-NEXT:    s_lshr_b32 s0, s0, 16
 ; GFX12-NEXT:    s_bfe_i64 s[2:3], s[2:3], 0x100000
 ; GFX12-NEXT:    v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v17, s19
+; GFX12-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x100000
 ; GFX12-NEXT:    s_bfe_i64 s[14:15], s[14:15], 0x100000
-; GFX12-NEXT:    s_bfe_i64 s[16:17], s[16:17], 0x100000
 ; GFX12-NEXT:    v_dual_mov_b32 v16, s18 :: v_dual_mov_b32 v19, s21
 ; GFX12-NEXT:    v_mov_b32_e32 v18, s20
-; GFX12-NEXT:    s_bfe_i64 s[12:13], s[12:13], 0x100000
+; GFX12-NEXT:    s_bfe_i64 s[0:1], s[0:1], 0x100000
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    global_store_b128 v24, v[8:11], s[8:9] offset:80
 ; GFX12-NEXT:    global_store_b128 v24, v[0:3], s[8:9] offset:64
-; GFX12-NEXT:    v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v0, s0
+; GFX12-NEXT:    v_dual_mov_b32 v1, s17 :: v_dual_mov_b32 v0, s16
 ; GFX12-NEXT:    v_dual_mov_b32 v3, s3 :: v_dual_mov_b32 v2, s2
-; GFX12-NEXT:    v_dual_mov_b32 v9, s15 :: v_dual_mov_b32 v8, s14
-; GFX12-NEXT:    v_dual_mov_b32 v11, s17 :: v_dual_mov_b32 v10, s16
+; GFX12-NEXT:    v_dual_mov_b32 v9, s13 :: v_dual_mov_b32 v8, s12
+; GFX12-NEXT:    v_dual_mov_b32 v11, s15 :: v_dual_mov_b32 v10, s14
 ; GFX12-NEXT:    v_dual_mov_b32 v21, s11 :: v_dual_mov_b32 v20, s10
-; GFX12-NEXT:    v_dual_mov_b32 v23, s13 :: v_dual_mov_b32 v22, s12
+; GFX12-NEXT:    v_dual_mov_b32 v23, s1 :: v_dual_mov_b32 v22, s0
 ; GFX12-NEXT:    s_clause 0x5
 ; GFX12-NEXT:    global_store_b128 v24, v[12:15], s[8:9] offset:112
 ; GFX12-NEXT:    global_store_b128 v24, v[4:7], s[8:9] offset:96
diff --git a/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll b/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
index 952827b8cd0e7..889755c23bbc7 100644
--- a/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-constant-i8.ll
@@ -8808,73 +8808,73 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX12-NEXT:    v_lshrrev_b16 v2, 8, s6
 ; GFX12-NEXT:    v_lshrrev_b16 v4, 8, s5
 ; GFX12-NEXT:    v_lshrrev_b16 v8, 8, s2
-; GFX12-NEXT:    s_lshr_b32 s24, s7, 16
+; GFX12-NEXT:    s_lshr_b32 s22, s7, 16
 ; GFX12-NEXT:    v_bfe_i32 v31, v1, 0, 8
-; GFX12-NEXT:    s_lshr_b32 s42, s2, 24
-; GFX12-NEXT:    s_mov_b32 s48, s7
+; GFX12-NEXT:    s_lshr_b32 s40, s2, 24
+; GFX12-NEXT:    s_mov_b32 s46, s7
 ; GFX12-NEXT:    v_lshrrev_b16 v5, 8, s4
 ; GFX12-NEXT:    v_lshrrev_b16 v7, 8, s1
-; GFX12-NEXT:    s_lshr_b32 s26, s6, 16
-; GFX12-NEXT:    s_lshr_b32 s44, s1, 16
+; GFX12-NEXT:    s_lshr_b32 s24, s6, 16
+; GFX12-NEXT:    s_lshr_b32 s42, s1, 16
 ; GFX12-NEXT:    s_ashr_i64 s[58:59], s[6:7], 56
-; GFX12-NEXT:    s_bfe_i64 s[48:49], s[48:49], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[46:47], s[46:47], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[22:23], s[22:23], 0x80000
 ; GFX12-NEXT:    v_lshrrev_b16 v6, 8, s3
 ; GFX12-NEXT:    v_lshrrev_b16 v3, 8, s0
-; GFX12-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v33, s24
-; GFX12-NEXT:    s_lshr_b32 s28, s6, 24
-; GFX12-NEXT:    s_lshr_b32 s30, s5, 16
-; GFX12-NEXT:    s_lshr_b32 s40, s2, 16
+; GFX12-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v33, s22
+; GFX12-NEXT:    s_lshr_b32 s26, s6, 24
+; GFX12-NEXT:    s_lshr_b32 s28, s5, 16
+; GFX12-NEXT:    s_lshr_b32 s38, s2, 16
 ; GFX12-NEXT:    v_bfe_i32 v11, v8, 0, 8
 ; GFX12-NEXT:    v_bfe_i32 v23, v4, 0, 8
 ; GFX12-NEXT:    v_bfe_i32 v27, v2, 0, 8
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v32, 31, v31
-; GFX12-NEXT:    s_bfe_i64 s[44:45], s[44:45], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[26:27], s[26:27], 0x80000
-; GFX12-NEXT:    v_dual_mov_b32 v34, s25 :: v_dual_mov_b32 v35, s58
-; GFX12-NEXT:    v_dual_mov_b32 v36, s59 :: v_dual_mov_b32 v37, s26
-; GFX12-NEXT:    v_dual_mov_b32 v56, s43 :: v_dual_mov_b32 v29, s48
-; GFX12-NEXT:    v_mov_b32_e32 v30, s49
-; GFX12-NEXT:    s_lshr_b32 s46, s0, 24
-; GFX12-NEXT:    s_mov_b32 s50, s5
-; GFX12-NEXT:    s_mov_b32 s52, s3
-; GFX12-NEXT:    s_lshr_b32 s34, s4, 16
-; GFX12-NEXT:    s_lshr_b32 s36, s4, 24
-; GFX12-NEXT:    s_ashr_i64 s[22:23], s[2:3], 56
+; GFX12-NEXT:    s_bfe_i64 s[42:43], s[42:43], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[24:25], s[24:25], 0x80000
+; GFX12-NEXT:    v_dual_mov_b32 v34, s23 :: v_dual_mov_b32 v35, s58
+; GFX12-NEXT:    v_dual_mov_b32 v36, s59 :: v_dual_mov_b32 v37, s24
+; GFX12-NEXT:    v_dual_mov_b32 v56, s41 :: v_dual_mov_b32 v29, s46
+; GFX12-NEXT:    v_mov_b32_e32 v30, s47
+; GFX12-NEXT:    s_lshr_b32 s44, s0, 24
+; GFX12-NEXT:    s_mov_b32 s48, s5
+; GFX12-NEXT:    s_mov_b32 s50, s3
+; GFX12-NEXT:    s_lshr_b32 s30, s4, 16
+; GFX12-NEXT:    s_lshr_b32 s34, s4, 24
+; GFX12-NEXT:    s_ashr_i64 s[54:55], s[2:3], 56
 ; GFX12-NEXT:    s_ashr_i64 s[56:57], s[4:5], 56
 ; GFX12-NEXT:    v_bfe_i32 v7, v7, 0, 8
 ; GFX12-NEXT:    v_bfe_i32 v19, v5, 0, 8
-; GFX12-NEXT:    s_bfe_i64 s[40:41], s[40:41], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[30:31], s[30:31], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
 ; GFX12-NEXT:    s_bfe_i64 s[28:29], s[28:29], 0x80000
-; GFX12-NEXT:    s_lshr_b32 s38, s3, 16
-; GFX12-NEXT:    s_mov_b32 s54, s1
+; GFX12-NEXT:    s_bfe_i64 s[26:27], s[26:27], 0x80000
+; GFX12-NEXT:    s_lshr_b32 s36, s3, 16
+; GFX12-NEXT:    s_mov_b32 s52, s1
 ; GFX12-NEXT:    s_bfe_i64 s[12:13], s[2:3], 0x80000
 ; GFX12-NEXT:    s_bfe_i64 s[14:15], s[4:5], 0x80000
 ; GFX12-NEXT:    s_bfe_i64 s[16:17], s[6:7], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[2:3], s[52:53], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[4:5], s[50:51], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[6:7], s[46:47], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[2:3], s[50:51], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[4:5], s[48:49], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[6:7], s[44:45], 0x80000
 ; GFX12-NEXT:    s_lshr_b32 s20, s0, 16
 ; GFX12-NEXT:    s_ashr_i64 s[18:19], s[0:1], 56
 ; GFX12-NEXT:    v_bfe_i32 v3, v3, 0, 8
 ; GFX12-NEXT:    v_bfe_i32 v15, v6, 0, 8
-; GFX12-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x80000
 ; GFX12-NEXT:    s_bfe_i64 s[34:35], s[34:35], 0x80000
-; GFX12-NEXT:    v_dual_mov_b32 v38, s27 :: v_dual_mov_b32 v39, s28
-; GFX12-NEXT:    v_dual_mov_b32 v40, s29 :: v_dual_mov_b32 v41, s30
-; GFX12-NEXT:    v_dual_mov_b32 v42, s31 :: v_dual_mov_b32 v43, s56
-; GFX12-NEXT:    v_dual_mov_b32 v44, s57 :: v_dual_mov_b32 v45, s34
-; GFX12-NEXT:    v_dual_mov_b32 v52, s23 :: v_dual_mov_b32 v53, s40
-; GFX12-NEXT:    v_dual_mov_b32 v54, s41 :: v_dual_mov_b32 v55, s42
+; GFX12-NEXT:    s_bfe_i64 s[30:31], s[30:31], 0x80000
+; GFX12-NEXT:    v_dual_mov_b32 v38, s25 :: v_dual_mov_b32 v39, s26
+; GFX12-NEXT:    v_dual_mov_b32 v40, s27 :: v_dual_mov_b32 v41, s28
+; GFX12-NEXT:    v_dual_mov_b32 v42, s29 :: v_dual_mov_b32 v43, s56
+; GFX12-NEXT:    v_dual_mov_b32 v44, s57 :: v_dual_mov_b32 v45, s30
+; GFX12-NEXT:    v_dual_mov_b32 v52, s55 :: v_dual_mov_b32 v53, s38
+; GFX12-NEXT:    v_dual_mov_b32 v54, s39 :: v_dual_mov_b32 v55, s40
 ; GFX12-NEXT:    s_bfe_i64 s[10:11], s[0:1], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[0:1], s[54:55], 0x80000
+; GFX12-NEXT:    s_bfe_i64 s[0:1], s[52:53], 0x80000
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v12, 31, v11
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v24, 31, v23
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v28, 31, v27
 ; GFX12-NEXT:    global_store_b128 v0, v[33:36], s[8:9] offset:240
-; GFX12-NEXT:    v_mov_b32_e32 v33, s44
+; GFX12-NEXT:    v_mov_b32_e32 v33, s42
 ; GFX12-NEXT:    global_store_b128 v0, v[29:32], s[8:9] offset:224
 ; GFX12-NEXT:    v_dual_mov_b32 v25, s16 :: v_dual_mov_b32 v26, s17
 ; GFX12-NEXT:    v_dual_mov_b32 v32, s7 :: v_dual_mov_b32 v21, s4
@@ -8882,16 +8882,16 @@ define amdgpu_kernel void @constant_sextload_v32i8_to_v32i64(ptr addrspace(1) %o
 ; GFX12-NEXT:    v_dual_mov_b32 v14, s3 :: v_dual_mov_b32 v9, s12
 ; GFX12-NEXT:    v_dual_mov_b32 v10, s13 :: v_dual_mov_b32 v5, s0
 ; GFX12-NEXT:    s_bfe_i64 s[20:21], s[20:21], 0x80000
-; GFX12-NEXT:    s_bfe_i64 s[38:39], s[38:39], 0x80000
-; GFX12-NEXT:    v_dual_mov_b32 v46, s35 :: v_dual_mov_b32 v47, s36
-; GFX12-NEXT:    v_dual_mov_b32 v48, s37 :: v_dual_mov_b32 v49, s38
-; GFX12-NEXT:    v_dual_mov_b32 v34, s45 :: v_dual_mov_b32 v35, s18
+; GFX12-NEXT:    s_bfe_i64 s[36:37], s[36:37], 0x80000
+; GFX12-NEXT:    v_dual_mov_b32 v46, s31 :: v_dual_mov_b32 v47, s34
+; GFX12-NEXT:    v_dual_mov_b32 v48, s35 :: v_dual_mov_b32 v49, s36
+; GFX12-NEXT:    v_dual_mov_b32 v34, s43 :: v_dual_mov_b32 v35, s18
 ; GFX12-NEXT:    v_dual_mov_b32 v36, s19 :: v_dual_mov_b32 v29, s20
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v8, 31, v7
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v20, 31, v19
 ; GFX12-NEXT:    v_dual_mov_b32 v18, s15 :: v_dual_mov_b32 v13, s2
 ; GFX12-NEXT:    v_dual_mov_b32 v6, s1 :: v_dual_mov_b32 v1, s10
-; GFX12-NEXT:    v_dual_mov_b32 v50, s39 :: v_dual_mov_b32 v51, s22
+; GFX12-NEXT:    v_dual_mov_b32 v50, s37 :: v_dual_mov_b32 v51, s54
 ; GFX12-NEXT:    v_dual_mov_b32 v30, s21 :: v_dual_mov_b32 v31, s6
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v4, 31, v3
 ; GFX12-NEXT:    v_ashrrev_i32_e32 v16, 31, v15
diff --git a/llvm/test/CodeGen/AMDGPU/sgpr-spill-overlap-wwm-reserve.mir b/llvm/test/CodeGen/AMDGPU/sgpr-spill-overlap-wwm-reserve.mir
index f8e7cb397b475..8a5f75332557e 100644
--- a/llvm/test/CodeGen/AMDGPU/sgpr-spill-overlap-wwm-reserve.mir
+++ b/llvm/test/CodeGen/AMDGPU/sgpr-spill-overlap-wwm-reserve.mir
@@ -28,18 +28,17 @@ body:             |
   ; GCN-LABEL: name: test_main
   ; GCN: bb.0:
   ; GCN-NEXT:   successors: %bb.1(0x80000000)
-  ; GCN-NEXT:   liveins: $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr18, $sgpr19, $sgpr20, $sgpr21, $sgpr22, $sgpr23, $sgpr24, $sgpr25, $sgpr26, $sgpr27, $sgpr28, $sgpr29, $sgpr30, $sgpr31, $sgpr64, $sgpr65, $sgpr66, $sgpr67, $sgpr68, $sgpr69, $sgpr70, $sgpr71, $sgpr72, $sgpr73, $sgpr74, $sgpr75, $sgpr76, $sgpr77, $sgpr78, $sgpr79, $sgpr80, $sgpr81, $sgpr82, $sgpr83, $sgpr84, $sgpr85, $sgpr86, $sgpr87, $sgpr88, $sgpr89, $sgpr90, $sgpr91, $sgpr92, $sgpr93, $sgpr94, $sgpr95, $sgpr96, $sgpr97, $sgpr98, $sgpr99, $sgpr100, $sgpr101, $sgpr102, $sgpr103, $vgpr0, $vgpr2, $vgpr3, $vgpr4, $vgpr5
+  ; GCN-NEXT:   liveins: $vcc_hi, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $sgpr10, $sgpr11, $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr18, $sgpr19, $sgpr20, $sgpr21, $sgpr22, $sgpr23, $sgpr24, $sgpr25, $sgpr26, $sgpr27, $sgpr28, $sgpr29, $sgpr30, $sgpr31, $sgpr64, $sgpr65, $sgpr66, $sgpr67, $sgpr68, $sgpr69, $sgpr70, $sgpr71, $sgpr72, $sgpr73, $sgpr74, $sgpr75, $sgpr76, $sgpr77, $sgpr78, $sgpr79, $sgpr80, $sgpr81, $sgpr82, $sgpr83, $sgpr84, $sgpr85, $sgpr86, $sgpr87, $sgpr88, $sgpr89, $sgpr90, $sgpr91, $sgpr92, $sgpr93, $sgpr94, $sgpr95, $sgpr96, $sgpr97, $sgpr98, $sgpr99, $sgpr100, $sgpr101, $sgpr102, $sgpr103, $vgpr0, $vgpr2, $vgpr3, $vgpr4, $vgpr5
   ; GCN-NEXT: {{  $}}
-  ; GCN-NEXT:   $sgpr0 = COPY $sgpr33
+  ; GCN-NEXT:   $vcc_hi = frame-setup COPY $sgpr33
   ; GCN-NEXT:   $sgpr33 = frame-setup COPY $sgpr32
-  ; GCN-NEXT:   $sgpr1 = S_XOR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
+  ; GCN-NEXT:   $sgpr0 = S_XOR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR $vgpr3, $sgpr33, 0, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.69, addrspace 5)
   ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR $vgpr4, $sgpr33, 4, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.70, addrspace 5)
   ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR $vgpr5, $sgpr33, 8, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.71, addrspace 5)
   ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR $vgpr2, $sgpr33, 12, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.72, addrspace 5)
-  ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR killed $vgpr1, $sgpr33, 16, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.74, addrspace 5)
-  ; GCN-NEXT:   $exec_lo = S_MOV_B32 killed $sgpr1
-  ; GCN-NEXT:   $vgpr5 = SI_SPILL_S32_TO_VGPR $sgpr0, 4, undef $vgpr5
+  ; GCN-NEXT:   SCRATCH_STORE_DWORD_SADDR killed $vgpr1, $sgpr33, 16, 0, implicit $exec, implicit $flat_scr :: (store (s32) into %stack.73, addrspace 5)
+  ; GCN-NEXT:   $exec_lo = S_MOV_B32 killed $sgpr0
   ; GCN-NEXT:   $sgpr32 = frame-setup S_ADD_I32 $sgpr32, 24, implicit-def dead $scc
   ; GCN-NEXT:   renamable $vgpr2 = IMPLICIT_DEF
   ; GCN-NEXT:   $vgpr3 = SI_SPILL_S32_TO_VGPR $sgpr4, 0, $vgpr3
@@ -116,18 +115,18 @@ body:             |
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.1:
   ; GCN-NEXT:   successors: %bb.2(0x80000000)
-  ; GCN-NEXT:   liveins: $vgpr2, $vgpr3, $vgpr4, $vgpr5
+  ; GCN-NEXT:   liveins: $vcc_hi, $vgpr2, $vgpr3, $vgpr4, $vgpr5
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   KILL implicit-def $vcc_lo, implicit-def $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5_sgpr6_sgpr7_sgpr8_sgpr9_sgpr10_sgpr11_sgpr12_sgpr13_sgpr14_sgpr15_sgpr16_sgpr17_sgpr18_sgpr19_sgpr20_sgpr21_sgpr22_sgpr23_sgpr24_sgpr25_sgpr26_sgpr27_sgpr28_sgpr29_sgpr30_sgpr31, implicit-def $sgpr32_sgpr33_sgpr34_sgpr35_sgpr36_sgpr37_sgpr38_sgpr39_sgpr40_sgpr41_sgpr42_sgpr43_sgpr44_sgpr45_sgpr46_sgpr47_sgpr48_sgpr49_sgpr50_sgpr51_sgpr52_sgpr53_sgpr54_sgpr55_sgpr56_sgpr57_sgpr58_sgpr59_sgpr60_sgpr61_sgpr62_sgpr63, implicit-def $sgpr64_sgpr65_sgpr66_sgpr67_sgpr68_sgpr69_sgpr70_sgpr71_sgpr72_sgpr73_sgpr74_sgpr75_sgpr76_sgpr77_sgpr78_sgpr79_sgpr80_sgpr81_sgpr82_sgpr83_sgpr84_sgpr85_sgpr86_sgpr87_sgpr88_sgpr89_sgpr90_sgpr91_sgpr92_sgpr93_sgpr94_sgpr95, implicit-def $sgpr96_sgpr97_sgpr98_sgpr99_sgpr100_sgpr101_sgpr102_sgpr103
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.2:
   ; GCN-NEXT:   successors: %bb.3(0x80000000)
-  ; GCN-NEXT:   liveins: $vgpr2, $vgpr3, $vgpr4, $vgpr5
+  ; GCN-NEXT:   liveins: $vcc_hi, $vgpr2, $vgpr3, $vgpr4, $vgpr5
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   $sgpr22 = SI_RESTORE_S32_FROM_VGPR $vgpr2, 0
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.3:
-  ; GCN-NEXT:   liveins: $vgpr2, $vgpr3, $vgpr4, $vgpr5
+  ; GCN-NEXT:   liveins: $vcc_hi, $vgpr2, $vgpr3, $vgpr4, $vgpr5
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   $sgpr103 = SI_RESTORE_S32_FROM_VGPR $vgpr5, 3
   ; GCN-NEXT:   $sgpr102 = SI_RESTORE_S32_FROM_VGPR $vgpr5, 2
@@ -198,16 +197,15 @@ body:             |
   ; GCN-NEXT:   $sgpr5 = SI_RESTORE_S32_FROM_VGPR $vgpr3, 1
   ; GCN-NEXT:   $sgpr4 = SI_RESTORE_S32_FROM_VGPR $vgpr3, 0
   ; GCN-NEXT:   KILL killed renamable $vgpr2
-  ; GCN-NEXT:   $sgpr0 = SI_RESTORE_S32_FROM_VGPR $vgpr5, 4
-  ; GCN-NEXT:   $sgpr1 = S_XOR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
+  ; GCN-NEXT:   $sgpr0 = S_XOR_SAVEEXEC_B32 -1, implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; GCN-NEXT:   $vgpr3 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 0, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.69, addrspace 5)
   ; GCN-NEXT:   $vgpr4 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 4, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.70, addrspace 5)
   ; GCN-NEXT:   $vgpr5 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 8, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.71, addrspace 5)
   ; GCN-NEXT:   $vgpr2 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 12, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.72, addrspace 5)
-  ; GCN-NEXT:   $vgpr1 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 16, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.74, addrspace 5)
-  ; GCN-NEXT:   $exec_lo = S_MOV_B32 killed $sgpr1
+  ; GCN-NEXT:   $vgpr1 = SCRATCH_LOAD_DWORD_SADDR $sgpr33, 16, 0, implicit $exec, implicit $flat_scr :: (load (s32) from %stack.73, addrspace 5)
+  ; GCN-NEXT:   $exec_lo = S_MOV_B32 killed $sgpr0
   ; GCN-NEXT:   $sgpr32 = frame-destroy S_ADD_I32 $sgpr32, -24, implicit-def dead $scc
-  ; GCN-NEXT:   $sgpr33 = COPY $sgpr0
+  ; GCN-NEXT:   $sgpr33 = frame-destroy COPY $vcc_hi
   ; GCN-NEXT:   S_ENDPGM 0
   bb.0:
     liveins: $vgpr0

From 485d556d8c23b54da952e75c3cadc9db3050fd9e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Sat, 13 Apr 2024 06:10:38 +0200
Subject: [PATCH 544/886] [clang][Interp][NFC] Add Block::dump()

---
 clang/lib/AST/Interp/Disasm.cpp    | 16 ++++++++++++++++
 clang/lib/AST/Interp/InterpBlock.h |  3 +++
 2 files changed, 19 insertions(+)

diff --git a/clang/lib/AST/Interp/Disasm.cpp b/clang/lib/AST/Interp/Disasm.cpp
index 022b394e58e64..ebc4e4f195ba6 100644
--- a/clang/lib/AST/Interp/Disasm.cpp
+++ b/clang/lib/AST/Interp/Disasm.cpp
@@ -264,3 +264,19 @@ LLVM_DUMP_METHOD void Record::dump(llvm::raw_ostream &OS, unsigned Indentation,
     ++I;
   }
 }
+
+LLVM_DUMP_METHOD void Block::dump(llvm::raw_ostream &OS) const {
+  {
+    ColorScope SC(OS, true, {llvm::raw_ostream::BRIGHT_BLUE, true});
+    OS << "Block " << (void *)this << "\n";
+  }
+  unsigned NPointers = 0;
+  for (const Pointer *P = Pointers; P; P = P->Next) {
+    ++NPointers;
+  }
+  OS << "  Pointers: " << NPointers << "\n";
+  OS << "  Dead: " << IsDead << "\n";
+  OS << "  Static: " << IsStatic << "\n";
+  OS << "  Extern: " << IsExtern << "\n";
+  OS << "  Initialized: " << IsInitialized << "\n";
+}
diff --git a/clang/lib/AST/Interp/InterpBlock.h b/clang/lib/AST/Interp/InterpBlock.h
index 9db82567d2d5d..6d5856fbd4ea1 100644
--- a/clang/lib/AST/Interp/InterpBlock.h
+++ b/clang/lib/AST/Interp/InterpBlock.h
@@ -118,6 +118,9 @@ class Block final {
     IsInitialized = false;
   }
 
+  void dump() const { dump(llvm::errs()); }
+  void dump(llvm::raw_ostream &OS) const;
+
 protected:
   friend class Pointer;
   friend class DeadBlock;

From 80fce05f2104d1c42db814276130536b014fcca2 Mon Sep 17 00:00:00 2001
From: Yingwei Zheng <dtcxzyw2333@gmail.com>
Date: Tue, 16 Apr 2024 17:16:35 +0800
Subject: [PATCH 545/886] [InstCombine] Fold `minmax (X & NegPow2C, Y &
 NegPow2C) -> minmax(X, Y) & NegPow2C` (#88859)

Alive2: https://alive2.llvm.org/ce/z/NFtkSX

This optimization will be beneficial to jemalloc users.
---
 .../InstCombine/InstCombineCalls.cpp          |  8 +-
 .../InstCombine/minmax-intrinsics.ll          | 89 +++++++++++++++++++
 2 files changed, 96 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index bae8579fc3650..ba5db854647a4 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -1774,6 +1774,13 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
     if (Instruction *I = moveAddAfterMinMax(II, Builder))
       return I;
 
+    // minmax (X & NegPow2C, Y & NegPow2C) --> minmax(X, Y) & NegPow2C
+    const APInt *RHSC;
+    if (match(I0, m_OneUse(m_And(m_Value(X), m_NegatedPower2(RHSC)))) &&
+        match(I1, m_OneUse(m_And(m_Value(Y), m_SpecificInt(*RHSC)))))
+      return BinaryOperator::CreateAnd(Builder.CreateBinaryIntrinsic(IID, X, Y),
+                                       ConstantInt::get(II->getType(), *RHSC));
+
     // smax(X, -X) --> abs(X)
     // smin(X, -X) --> -abs(X)
     // umax(X, -X) --> -abs(X)
@@ -1815,7 +1822,6 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
        return NewMinMax;
 
     // Try to fold minmax with constant RHS based on range information
-    const APInt *RHSC;
     if (match(I1, m_APIntAllowUndef(RHSC))) {
       ICmpInst::Predicate Pred =
           ICmpInst::getNonStrictPredicate(MinMaxIntrinsic::getPredicate(IID));
diff --git a/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll b/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
index ae2e115b1dd9a..bd1a47bbfcc19 100644
--- a/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
@@ -2581,3 +2581,92 @@ entry:
   %val = call i8 @llvm.umin.i8(i8 %sub, i8 3)
   ret i8 %val
 }
+
+define i8 @test_umax_and(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_umax_and(
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.umax.i8(i8 [[X1:%.*]], i8 [[Y1:%.*]])
+; CHECK-NEXT:    [[RES1:%.*]] = and i8 [[RES]], -64
+; CHECK-NEXT:    ret i8 [[RES1]]
+;
+  %x1 = and i8 %x, -64
+  %y1 = and i8 %y, -64
+  %res = call i8 @llvm.umax.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_umin_and(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_umin_and(
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.umin.i8(i8 [[X1:%.*]], i8 [[Y1:%.*]])
+; CHECK-NEXT:    [[RES1:%.*]] = and i8 [[RES]], -64
+; CHECK-NEXT:    ret i8 [[RES1]]
+;
+  %x1 = and i8 %x, -64
+  %y1 = and i8 %y, -64
+  %res = call i8 @llvm.umin.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_smax_and(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_smax_and(
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.smax.i8(i8 [[X1:%.*]], i8 [[Y1:%.*]])
+; CHECK-NEXT:    [[RES1:%.*]] = and i8 [[RES]], -64
+; CHECK-NEXT:    ret i8 [[RES1]]
+;
+  %x1 = and i8 %x, -64
+  %y1 = and i8 %y, -64
+  %res = call i8 @llvm.smax.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_smin_and(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_smin_and(
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.smin.i8(i8 [[X1:%.*]], i8 [[Y1:%.*]])
+; CHECK-NEXT:    [[RES1:%.*]] = and i8 [[RES]], -64
+; CHECK-NEXT:    ret i8 [[RES1]]
+;
+  %x1 = and i8 %x, -64
+  %y1 = and i8 %y, -64
+  %res = call i8 @llvm.smin.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_smin_and_mismatch(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_smin_and_mismatch(
+; CHECK-NEXT:    [[X1:%.*]] = and i8 [[X:%.*]], -64
+; CHECK-NEXT:    [[Y1:%.*]] = and i8 [[Y:%.*]], -32
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.smin.i8(i8 [[X1]], i8 [[Y1]])
+; CHECK-NEXT:    ret i8 [[RES]]
+;
+  %x1 = and i8 %x, -64
+  %y1 = and i8 %y, -32
+  %res = call i8 @llvm.smin.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_smin_and_non_negated_pow2(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_smin_and_non_negated_pow2(
+; CHECK-NEXT:    [[X1:%.*]] = and i8 [[X:%.*]], 31
+; CHECK-NEXT:    [[Y1:%.*]] = and i8 [[Y:%.*]], 31
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.smin.i8(i8 [[X1]], i8 [[Y1]])
+; CHECK-NEXT:    ret i8 [[RES]]
+;
+  %x1 = and i8 %x, 31
+  %y1 = and i8 %y, 31
+  %res = call i8 @llvm.smin.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}
+
+define i8 @test_smin_and_multiuse(i8 %x, i8 %y) {
+; CHECK-LABEL: @test_smin_and_multiuse(
+; CHECK-NEXT:    [[X1:%.*]] = and i8 [[X:%.*]], 31
+; CHECK-NEXT:    [[Y1:%.*]] = and i8 [[Y:%.*]], 31
+; CHECK-NEXT:    call void @use(i8 [[Y1]])
+; CHECK-NEXT:    [[RES:%.*]] = call i8 @llvm.smin.i8(i8 [[X1]], i8 [[Y1]])
+; CHECK-NEXT:    ret i8 [[RES]]
+;
+  %x1 = and i8 %x, 31
+  %y1 = and i8 %y, 31
+  call void @use(i8 %y1)
+  %res = call i8 @llvm.smin.i8(i8 %x1, i8 %y1)
+  ret i8 %res
+}

From cce026bf8f7dcf5aa402a6da20f0d4da56aee8b5 Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Tue, 16 Apr 2024 17:19:45 +0800
Subject: [PATCH 546/886] [mlir][test] Fix -Wsign-compare in TestDialect.cpp
 (NFC)

llvm-project/mlir/test/lib/Dialect/Test/TestDialect.cpp:597:31:
error: comparison of integers of different signs: 'size_t' (aka 'unsigned long') and 'int64_t' (aka 'long') [-Werror,-Wsign-compare]
  if (getVarOperands().size() != expectedNumOperands)
      ~~~~~~~~~~~~~~~~~~~~~~~ ^  ~~~~~~~~~~~~~~~~~~~
1 error generated.
---
 mlir/test/lib/Dialect/Test/TestDialect.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/test/lib/Dialect/Test/TestDialect.cpp b/mlir/test/lib/Dialect/Test/TestDialect.cpp
index 36d7606fe1345..a23ed89c4b04d 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialect.cpp
@@ -594,7 +594,7 @@ LogicalResult CompareOp::verify() {
         "'compose' not supported when 'lhs_map' or 'rhs_map' is present");
   int64_t expectedNumOperands = getLhsMap() ? getLhsMap()->getNumInputs() : 1;
   expectedNumOperands += getRhsMap() ? getRhsMap()->getNumInputs() : 1;
-  if (getVarOperands().size() != expectedNumOperands)
+  if (getVarOperands().size() != size_t(expectedNumOperands))
     return emitOpError("expected ")
            << expectedNumOperands << " operands, but got "
            << getVarOperands().size();

From 58b49cef1d772a922a433fd4a42e41db3f18d34b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Sat, 13 Apr 2024 15:06:22 +0200
Subject: [PATCH 547/886] [clang][Interp] Support __builtin_vectorelements

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 9 +++++++++
 clang/test/AST/Interp/vectors.cpp        | 2 ++
 2 files changed, 11 insertions(+)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 5866228663dca..93059edc4622f 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1251,6 +1251,15 @@ bool ByteCodeExprGen<Emitter>::VisitUnaryExprOrTypeTraitExpr(
     return this->emitConst(Size.getQuantity(), E);
   }
 
+  if (Kind == UETT_VectorElements) {
+    if (const auto *VT = E->getTypeOfArgument()->getAs<VectorType>())
+      return this->emitConst(VT->getNumElements(), E);
+
+    // FIXME: Apparently we need to catch the fact that a sizeless vector type
+    // has been passed and diagnose that (at run time).
+    assert(E->getTypeOfArgument()->isSizelessVectorType());
+  }
+
   return false;
 }
 
diff --git a/clang/test/AST/Interp/vectors.cpp b/clang/test/AST/Interp/vectors.cpp
index fb5787a9eda9a..6c5d916f51f56 100644
--- a/clang/test/AST/Interp/vectors.cpp
+++ b/clang/test/AST/Interp/vectors.cpp
@@ -13,12 +13,14 @@ namespace Vector {
     return VI4 { n * 3, n + 4, n - 5, n / 6 };
   }
   constexpr auto v1 = f(10);
+  static_assert(__builtin_vectorelements(v1) == (16 / sizeof(int)), "");
 
   typedef double __attribute__((vector_size(32))) VD4;
   constexpr VD4 g(int n) {
     return (VD4) { n / 2.0, n + 1.5, n - 5.4, n * 0.9 };
   }
   constexpr auto v2 = g(4);
+  static_assert(__builtin_vectorelements(v2) == (32 / sizeof(double)), "");
 }
 
 /// FIXME: We need to support BitCasts between vector types.

From 70fe6ad535365de20c3a960dcfe97bd5cf68abb5 Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Tue, 16 Apr 2024 10:24:23 +0100
Subject: [PATCH 548/886] [MLIR][OpenMP] Make omp.distribute into a loop
 wrapper (#87239)

This patch updates the definition of `omp.distribute` to enforce the
restrictions of a wrapper operation.
---
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td | 22 +++++++++--
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  | 11 ++++++
 mlir/test/Dialect/OpenMP/invalid.mlir         | 34 ++++++++++++++++-
 mlir/test/Dialect/OpenMP/ops.mlir             | 38 +++++++++++++++----
 4 files changed, 93 insertions(+), 12 deletions(-)

diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 3abdbe3adfd0b..7e677c9839f60 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -840,7 +840,8 @@ def YieldOp : OpenMP_Op<"yield",
 //===----------------------------------------------------------------------===//
 def DistributeOp : OpenMP_Op<"distribute", [AttrSizedOperandSegments,
                              DeclareOpInterfaceMethods<LoopWrapperInterface>,
-                             RecursiveMemoryEffects]> {
+                             RecursiveMemoryEffects,
+                             SingleBlockImplicitTerminator<"TerminatorOp">]> {
   let summary = "distribute construct";
   let description = [{
     The distribute construct specifies that the iterations of one or more loops
@@ -855,15 +856,28 @@ def DistributeOp : OpenMP_Op<"distribute", [AttrSizedOperandSegments,
     The distribute loop construct specifies that the iterations of the loop(s)
     will be executed in parallel by threads in the current context. These
     iterations are spread across threads that already exist in the enclosing
-    region. The lower and upper bounds specify a half-open range: the
-    range includes the lower bound but does not include the upper bound. If the
-    `inclusive` attribute is specified then the upper bound is also included.
+    region.
+    
+    The body region can contain a single block which must contain a single
+    operation and a terminator. The operation must be another compatible loop
+    wrapper or an `omp.loop_nest`.
 
     The `dist_schedule_static` attribute specifies the  schedule for this
     loop, determining how the loop is distributed across the parallel threads.
     The optional `schedule_chunk` associated with this determines further
     controls this distribution.
 
+    ```mlir
+    omp.distribute <clauses> {
+      omp.loop_nest (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
+        %a = load %arrA[%i1, %i2] : memref<?x?xf32>
+        %b = load %arrB[%i1, %i2] : memref<?x?xf32>
+        %sum = arith.addf %a, %b : f32
+        store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
+        omp.yield
+      }
+    }
+    ```
     // TODO: private_var, firstprivate_var, lastprivate_var, collapse
   }];
   let arguments = (ins
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 90b49b2528b79..365b3c4246bfd 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1656,6 +1656,17 @@ LogicalResult DistributeOp::verify() {
     return emitError(
         "expected equal sizes for allocate and allocator variables");
 
+  if (!isWrapper())
+    return emitOpError() << "must be a loop wrapper";
+
+  if (LoopWrapperInterface nested = getNestedWrapper()) {
+    // Check for the allowed leaf constructs that may appear in a composite
+    // construct directly after DISTRIBUTE.
+    if (!isa<ParallelOp, SimdLoopOp>(nested))
+      return emitError() << "only supported nested wrappers are 'omp.parallel' "
+                            "and 'omp.simdloop'";
+  }
+
   return success();
 }
 
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index 88dca1b85ee5f..f8739887214bf 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -1866,7 +1866,16 @@ func.func @omp_target_depend(%data_var: memref<i32>) {
 
 // -----
 
-func.func @omp_distribute(%data_var : memref<i32>) -> () {
+func.func @omp_distribute_schedule(%chunk_size : i32) -> () {
+  // expected-error @below {{op chunk size set without dist_schedule_static being present}}
+  "omp.distribute"(%chunk_size) <{operandSegmentSizes = array<i32: 1, 0, 0>}> ({
+      "omp.terminator"() : () -> ()
+    }) : (i32) -> ()
+}
+
+// -----
+
+func.func @omp_distribute_allocate(%data_var : memref<i32>) -> () {
   // expected-error @below {{expected equal sizes for allocate and allocator variables}}
   "omp.distribute"(%data_var) <{operandSegmentSizes = array<i32: 0, 1, 0>}> ({
       "omp.terminator"() : () -> ()
@@ -1875,6 +1884,29 @@ func.func @omp_distribute(%data_var : memref<i32>) -> () {
 
 // -----
 
+func.func @omp_distribute_wrapper() -> () {
+  // expected-error @below {{op must be a loop wrapper}}
+  "omp.distribute"() ({
+      %0 = arith.constant 0 : i32
+      "omp.terminator"() : () -> ()
+    }) : () -> ()
+}
+
+// -----
+
+func.func @omp_distribute_nested_wrapper(%data_var : memref<i32>) -> () {
+  // expected-error @below {{only supported nested wrappers are 'omp.parallel' and 'omp.simdloop'}}
+  "omp.distribute"() ({
+      "omp.wsloop"() ({
+        %0 = arith.constant 0 : i32
+        "omp.terminator"() : () -> ()
+      }) : () -> ()
+      "omp.terminator"() : () -> ()
+    }) : () -> ()
+}
+
+// -----
+
 omp.private {type = private} @x.privatizer : i32 alloc {
 ^bb0(%arg0: i32):
   %0 = arith.constant 0.0 : f32
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index 851d44ad984ee..018b82e995d7d 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -559,30 +559,54 @@ func.func @omp_simdloop_pretty_multiple(%lb1 : index, %ub1 : index, %step1 : ind
 }
 
 // CHECK-LABEL: omp_distribute
-func.func @omp_distribute(%chunk_size : i32, %data_var : memref<i32>) -> () {
+func.func @omp_distribute(%chunk_size : i32, %data_var : memref<i32>, %arg0 : i32) -> () {
   // CHECK: omp.distribute
   "omp.distribute" () ({
-    omp.terminator
+    "omp.loop_nest" (%arg0, %arg0, %arg0) ({
+    ^bb0(%iv: i32):
+      "omp.yield"() : () -> ()
+    }) : (i32, i32, i32) -> ()
+    "omp.terminator"() : () -> ()
   }) {} : () -> ()
   // CHECK: omp.distribute
   omp.distribute {
-    omp.terminator
+    omp.loop_nest (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.yield
+    }
   }
   // CHECK: omp.distribute dist_schedule_static
   omp.distribute dist_schedule_static {
-    omp.terminator
+    omp.loop_nest (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.yield
+    }
   }
   // CHECK: omp.distribute dist_schedule_static chunk_size(%{{.+}} : i32)
   omp.distribute dist_schedule_static chunk_size(%chunk_size : i32) {
-    omp.terminator
+    omp.loop_nest (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.yield
+    }
   }
   // CHECK: omp.distribute order(concurrent)
   omp.distribute order(concurrent) {
-    omp.terminator
+    omp.loop_nest (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.yield
+    }
   }
   // CHECK: omp.distribute allocate(%{{.+}} : memref<i32> -> %{{.+}} : memref<i32>)
   omp.distribute allocate(%data_var : memref<i32> -> %data_var : memref<i32>) {
-    omp.terminator
+    omp.loop_nest (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.yield
+    }
+  }
+  // CHECK: omp.distribute
+  omp.distribute {
+    // TODO Remove induction variables from omp.simdloop.
+    omp.simdloop for (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+      omp.loop_nest (%iv2) : i32 = (%arg0) to (%arg0) step (%arg0) {
+        omp.yield
+      }
+      omp.yield
+    }
   }
 return
 }

From 668a58b8926473d731c41c55007f1fe4571ada86 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Tue, 16 Apr 2024 10:25:26 +0100
Subject: [PATCH 549/886] [flang][runtime] Add ACCESS library procedure
 (#88517)

Re-land https://github.com/llvm/llvm-project/pull/88395

Two build-bots were broken by the old version:
 - https://lab.llvm.org/buildbot/#/builders/285/builds/245
 - https://lab.llvm.org/buildbot/#/builders/21/builds/96988

The problem in both cases was that the compiler did not support
`std::filesystem` (which I use in the unit test).

I have removed the dependency upon std::filesystem because there isn't
an easy way to add the right linker options so that this is supported
correctly in all build environments [1]

[1] https://gitlab.kitware.com/cmake/cmake/-/issues/17834

---

This is a GNU extension:
https://gcc.gnu.org/onlinedocs/gfortran/ACCESS.html

Used in SALMON:
https://salmon-tddft.jp/download.html

Unfortunately the intrinsic takes a file path to operate on so there
isn't an easy way to make the test robust. The unit test expects to be
able to create, set read write and execute permissions, and delete files
called
std::filesystem::temp_directory_path() / <test_name>.<pid>

The test will fail if a file already exists with that name.

I have not implemented the intrinsic on Windows because this is wrapping
a POSIX system call and Windows doesn't support all of the permission
bits tested by the intrinsic. I don't have a Windows machine easily
available to check if Gfortran implements this intrinsic on Windows.
---
 flang/docs/Intrinsics.md                 |   8 +
 flang/include/flang/Runtime/extensions.h |   7 +
 flang/runtime/extensions.cpp             |  73 ++++
 flang/unittests/Runtime/AccessTest.cpp   | 422 +++++++++++++++++++++++
 flang/unittests/Runtime/CMakeLists.txt   |   1 +
 5 files changed, 511 insertions(+)
 create mode 100644 flang/unittests/Runtime/AccessTest.cpp

diff --git a/flang/docs/Intrinsics.md b/flang/docs/Intrinsics.md
index ccb93e104dab6..848619cb65d90 100644
--- a/flang/docs/Intrinsics.md
+++ b/flang/docs/Intrinsics.md
@@ -657,6 +657,14 @@ CALL CO_REDUCE
 CALL CO_SUM
 ```
 
+### Inquiry Functions
+ACCESS (GNU extension) is not supported on Windows. Otherwise:
+```
+CHARACTER(LEN=*) :: path = 'path/to/file'
+IF (ACCESS(path, 'rwx')) &
+  ...
+```
+
 ## Non-standard intrinsics
 ### PGI
 ```
diff --git a/flang/include/flang/Runtime/extensions.h b/flang/include/flang/Runtime/extensions.h
index 7d0952206fc19..fef651f3b2eed 100644
--- a/flang/include/flang/Runtime/extensions.h
+++ b/flang/include/flang/Runtime/extensions.h
@@ -44,5 +44,12 @@ std::int64_t RTNAME(Signal)(std::int64_t number, void (*handler)(int));
 // GNU extension subroutine SLEEP(SECONDS)
 void RTNAME(Sleep)(std::int64_t seconds);
 
+// GNU extension function ACCESS(NAME, MODE)
+// TODO: not supported on Windows
+#ifndef _WIN32
+std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
+    std::int64_t nameLength, const char *mode, std::int64_t modeLength);
+#endif
+
 } // extern "C"
 #endif // FORTRAN_RUNTIME_EXTENSIONS_H_
diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 3ac98000335d7..12498b502ae1c 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -17,6 +17,7 @@
 #include "flang/Runtime/entry-names.h"
 #include "flang/Runtime/io-api.h"
 #include <chrono>
+#include <cstring>
 #include <ctime>
 #include <signal.h>
 #include <thread>
@@ -138,5 +139,77 @@ void RTNAME(Sleep)(std::int64_t seconds) {
   std::this_thread::sleep_for(std::chrono::seconds(seconds));
 }
 
+// TODO: not supported on Windows
+#ifndef _WIN32
+std::int64_t FORTRAN_PROCEDURE_NAME(access)(const char *name,
+    std::int64_t nameLength, const char *mode, std::int64_t modeLength) {
+  std::int64_t ret{-1};
+  if (nameLength <= 0 || modeLength <= 0 || !name || !mode) {
+    return ret;
+  }
+
+  // ensure name is null terminated
+  char *newName{nullptr};
+  if (name[nameLength - 1] != '\0') {
+    newName = static_cast<char *>(std::malloc(nameLength + 1));
+    std::memcpy(newName, name, nameLength);
+    newName[nameLength] = '\0';
+    name = newName;
+  }
+
+  // calculate mode
+  bool read{false};
+  bool write{false};
+  bool execute{false};
+  bool exists{false};
+  int imode{0};
+
+  for (std::int64_t i = 0; i < modeLength; ++i) {
+    switch (mode[i]) {
+    case 'r':
+      read = true;
+      break;
+    case 'w':
+      write = true;
+      break;
+    case 'x':
+      execute = true;
+      break;
+    case ' ':
+      exists = true;
+      break;
+    default:
+      // invalid mode
+      goto cleanup;
+    }
+  }
+  if (!read && !write && !execute && !exists) {
+    // invalid mode
+    goto cleanup;
+  }
+
+  if (!read && !write && !execute) {
+    imode = F_OK;
+  } else {
+    if (read) {
+      imode |= R_OK;
+    }
+    if (write) {
+      imode |= W_OK;
+    }
+    if (execute) {
+      imode |= X_OK;
+    }
+  }
+  ret = access(name, imode);
+
+cleanup:
+  if (newName) {
+    free(newName);
+  }
+  return ret;
+}
+#endif
+
 } // namespace Fortran::runtime
 } // extern "C"
diff --git a/flang/unittests/Runtime/AccessTest.cpp b/flang/unittests/Runtime/AccessTest.cpp
new file mode 100644
index 0000000000000..66f19f78c7cfb
--- /dev/null
+++ b/flang/unittests/Runtime/AccessTest.cpp
@@ -0,0 +1,422 @@
+//===-- flang/unittests/Runtime/AccessTest.cpp ----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// TODO: ACCESS is not yet implemented on Windows
+#ifndef _WIN32
+
+#include "CrashHandlerFixture.h"
+#include "gtest/gtest.h"
+#include "flang/Runtime/extensions.h"
+#include "llvm/ADT/Twine.h"
+
+#include <fcntl.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+namespace {
+
+struct AccessTests : public CrashHandlerFixture {};
+
+struct AccessType {
+  bool read{false};
+  bool write{false};
+  bool execute{false};
+  bool exists{false};
+};
+
+} // namespace
+
+static std::string addPIDSuffix(const char *name) {
+  std::stringstream ss;
+  ss << name;
+  ss << '.';
+
+  ss << getpid();
+
+  return ss.str();
+}
+
+static bool exists(const std::string &path) {
+  return access(path.c_str(), F_OK) == 0;
+}
+
+// Implementation of std::filesystem::temp_directory_path adapted from libcxx
+// See llvm-project/libcxx/src/filesystem/operations.cpp
+// Using std::filesystem is inconvenient because the required flags are not
+// consistent accross compilers and CMake doesn't have built in support to
+// determine the correct flags.
+static const char *temp_directory_path() {
+  // TODO: Windows
+  const char *env_paths[] = {"TMPDIR", "TMP", "TEMP", "TEMPDIR"};
+  const char *ret = nullptr;
+
+  for (auto &ep : env_paths) {
+    if ((ret = getenv(ep))) {
+      break;
+    }
+  }
+
+  if (ret == nullptr) {
+#if defined(__ANDROID__)
+    ret = "/data/local/tmp";
+#else
+    ret = "/tmp";
+#endif
+  }
+
+  assert(exists(ret));
+  return ret;
+}
+
+static std::string createTemporaryFile(
+    const char *name, const AccessType &accessType) {
+  std::string path =
+      (llvm::Twine{temp_directory_path()} + "/" + addPIDSuffix(name)).str();
+
+  // O_CREAT | O_EXCL enforces that this file is newly created by this call.
+  // This feels risky. If we don't have permission to create files in the
+  // temporary directory or if the files already exist, the test will fail.
+  // But we can't use std::tmpfile() because we need a path to the file and
+  // to control the filesystem permissions
+  mode_t mode{0};
+  if (accessType.read) {
+    mode |= S_IRUSR;
+  }
+  if (accessType.write) {
+    mode |= S_IWUSR;
+  }
+  if (accessType.execute) {
+    mode |= S_IXUSR;
+  }
+
+  int file = open(path.c_str(), O_CREAT | O_EXCL, mode);
+  if (file == -1) {
+    return {};
+  }
+
+  close(file);
+
+  return path;
+}
+
+static std::int64_t callAccess(
+    const std::string &path, const AccessType &accessType) {
+  const char *cpath{path.c_str()};
+  std::int64_t pathlen = std::strlen(cpath);
+
+  std::string mode;
+  if (accessType.read) {
+    mode += 'r';
+  }
+  if (accessType.write) {
+    mode += 'w';
+  }
+  if (accessType.execute) {
+    mode += 'x';
+  }
+  if (accessType.exists) {
+    mode += ' ';
+  }
+
+  const char *cmode = mode.c_str();
+  std::int64_t modelen = std::strlen(cmode);
+
+  return FORTRAN_PROCEDURE_NAME(access)(cpath, pathlen, cmode, modelen);
+}
+
+TEST(AccessTests, TestExists) {
+  AccessType accessType;
+  accessType.exists = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotExists) {
+  std::string nonExistant{addPIDSuffix(__func__)};
+  ASSERT_FALSE(exists(nonExistant));
+
+  AccessType accessType;
+  accessType.exists = true;
+  std::int64_t res = callAccess(nonExistant, accessType);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestRead) {
+  AccessType accessType;
+  accessType.read = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotRead) {
+  AccessType accessType;
+  accessType.read = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestWrite) {
+  AccessType accessType;
+  accessType.write = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotWrite) {
+  AccessType accessType;
+  accessType.write = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestReadWrite) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite0) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite1) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotReadWrite2) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestExecute) {
+  AccessType accessType;
+  accessType.execute = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotExecute) {
+  AccessType accessType;
+  accessType.execute = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestRWX) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_EQ(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX0) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = false;
+  accessType.execute = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX1) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+  accessType.execute = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX2) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = false;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX3) {
+  AccessType accessType;
+  accessType.read = true;
+  accessType.write = false;
+  accessType.execute = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+TEST(AccessTests, TestNotRWX4) {
+  AccessType accessType;
+  accessType.read = false;
+  accessType.write = true;
+  accessType.execute = true;
+
+  std::string path = createTemporaryFile(__func__, accessType);
+  ASSERT_FALSE(path.empty());
+
+  accessType.read = true;
+  accessType.write = true;
+  accessType.execute = true;
+  std::int64_t res = callAccess(path, accessType);
+
+  ASSERT_EQ(unlink(path.c_str()), 0);
+
+  ASSERT_NE(res, 0);
+}
+
+#endif // !_WIN32
diff --git a/flang/unittests/Runtime/CMakeLists.txt b/flang/unittests/Runtime/CMakeLists.txt
index 23f02aa751246..f7caacad3a598 100644
--- a/flang/unittests/Runtime/CMakeLists.txt
+++ b/flang/unittests/Runtime/CMakeLists.txt
@@ -1,4 +1,5 @@
 add_flang_unittest(FlangRuntimeTests
+  AccessTest.cpp
   Allocatable.cpp
   ArrayConstructor.cpp
   BufferTest.cpp

From 76782e28869abf93716f72f195d55c28eaf263ed Mon Sep 17 00:00:00 2001
From: Kiran Chandramohan <kiran.chandramohan@arm.com>
Date: Tue, 16 Apr 2024 10:29:26 +0100
Subject: [PATCH 550/886] [Flang][OpenMP] NFC: Remove old reduction lowering
 code (#88798)

The old code was replaced by
https://github.com/llvm/llvm-project/pull/80019.
---
 flang/lib/Lower/OpenMP/OpenMP.cpp | 223 ------------------------------
 1 file changed, 223 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 3dcfe0fd775dc..352ca66e8735b 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -103,21 +103,6 @@ static fir::GlobalOp globalInitialization(
   return global;
 }
 
-static mlir::Operation *getCompareFromReductionOp(mlir::Operation *reductionOp,
-                                                  mlir::Value loadVal) {
-  for (mlir::Value reductionOperand : reductionOp->getOperands()) {
-    if (mlir::Operation *compareOp = reductionOperand.getDefiningOp()) {
-      if (compareOp->getOperand(0) == loadVal ||
-          compareOp->getOperand(1) == loadVal)
-        assert((mlir::isa<mlir::arith::CmpIOp>(compareOp) ||
-                mlir::isa<mlir::arith::CmpFOp>(compareOp)) &&
-               "Expected comparison not found in reduction intrinsic");
-      return compareOp;
-    }
-  }
-  return nullptr;
-}
-
 // Get the extended value for \p val by extracting additional variable
 // information from \p base.
 static fir::ExtendedValue getExtendedValue(fir::ExtendedValue base,
@@ -237,213 +222,6 @@ createAndSetPrivatizedLoopVar(Fortran::lower::AbstractConverter &converter,
   return storeOp;
 }
 
-static mlir::Operation *
-findReductionChain(mlir::Value loadVal, mlir::Value *reductionVal = nullptr) {
-  for (mlir::OpOperand &loadOperand : loadVal.getUses()) {
-    if (mlir::Operation *reductionOp = loadOperand.getOwner()) {
-      if (auto convertOp = mlir::dyn_cast<fir::ConvertOp>(reductionOp)) {
-        for (mlir::OpOperand &convertOperand : convertOp.getRes().getUses()) {
-          if (mlir::Operation *reductionOp = convertOperand.getOwner())
-            return reductionOp;
-        }
-      }
-      for (mlir::OpOperand &reductionOperand : reductionOp->getUses()) {
-        if (auto store =
-                mlir::dyn_cast<fir::StoreOp>(reductionOperand.getOwner())) {
-          if (store.getMemref() == *reductionVal) {
-            store.erase();
-            return reductionOp;
-          }
-        }
-        if (auto assign =
-                mlir::dyn_cast<hlfir::AssignOp>(reductionOperand.getOwner())) {
-          if (assign.getLhs() == *reductionVal) {
-            assign.erase();
-            return reductionOp;
-          }
-        }
-      }
-    }
-  }
-  return nullptr;
-}
-
-// for a logical operator 'op' reduction X = X op Y
-// This function returns the operation responsible for converting Y from
-// fir.logical<4> to i1
-static fir::ConvertOp getConvertFromReductionOp(mlir::Operation *reductionOp,
-                                                mlir::Value loadVal) {
-  for (mlir::Value reductionOperand : reductionOp->getOperands()) {
-    if (auto convertOp =
-            mlir::dyn_cast<fir::ConvertOp>(reductionOperand.getDefiningOp())) {
-      if (convertOp.getOperand() == loadVal)
-        continue;
-      return convertOp;
-    }
-  }
-  return nullptr;
-}
-
-static void updateReduction(mlir::Operation *op,
-                            fir::FirOpBuilder &firOpBuilder,
-                            mlir::Value loadVal, mlir::Value reductionVal,
-                            fir::ConvertOp *convertOp = nullptr) {
-  mlir::OpBuilder::InsertPoint insertPtDel = firOpBuilder.saveInsertionPoint();
-  firOpBuilder.setInsertionPoint(op);
-
-  mlir::Value reductionOp;
-  if (convertOp)
-    reductionOp = convertOp->getOperand();
-  else if (op->getOperand(0) == loadVal)
-    reductionOp = op->getOperand(1);
-  else
-    reductionOp = op->getOperand(0);
-
-  firOpBuilder.create<mlir::omp::ReductionOp>(op->getLoc(), reductionOp,
-                                              reductionVal);
-  firOpBuilder.restoreInsertionPoint(insertPtDel);
-}
-
-static void removeStoreOp(mlir::Operation *reductionOp, mlir::Value symVal) {
-  for (mlir::Operation *reductionOpUse : reductionOp->getUsers()) {
-    if (auto convertReduction =
-            mlir::dyn_cast<fir::ConvertOp>(reductionOpUse)) {
-      for (mlir::Operation *convertReductionUse :
-           convertReduction.getRes().getUsers()) {
-        if (auto storeOp = mlir::dyn_cast<fir::StoreOp>(convertReductionUse)) {
-          if (storeOp.getMemref() == symVal)
-            storeOp.erase();
-        }
-        if (auto assignOp =
-                mlir::dyn_cast<hlfir::AssignOp>(convertReductionUse)) {
-          if (assignOp.getLhs() == symVal)
-            assignOp.erase();
-        }
-      }
-    }
-  }
-}
-
-// Generate an OpenMP reduction operation.
-// TODO: Currently assumes it is either an integer addition/multiplication
-// reduction, or a logical and reduction. Generalize this for various reduction
-// operation types.
-// TODO: Generate the reduction operation during lowering instead of creating
-// and removing operations since this is not a robust approach. Also, removing
-// ops in the builder (instead of a rewriter) is probably not the best approach.
-static void
-genOpenMPReduction(Fortran::lower::AbstractConverter &converter,
-                   Fortran::semantics::SemanticsContext &semaCtx,
-                   const Fortran::parser::OmpClauseList &clauseList) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-
-  List<Clause> clauses{makeClauses(clauseList, semaCtx)};
-
-  for (const Clause &clause : clauses) {
-    if (const auto &reductionClause =
-            std::get_if<clause::Reduction>(&clause.u)) {
-      const auto &redOperatorList{
-          std::get<clause::Reduction::ReductionIdentifiers>(
-              reductionClause->t)};
-      assert(redOperatorList.size() == 1 && "Expecting single operator");
-      const auto &redOperator = redOperatorList.front();
-      const auto &objects{std::get<ObjectList>(reductionClause->t)};
-      if (const auto *reductionOp =
-              std::get_if<clause::DefinedOperator>(&redOperator.u)) {
-        const auto &intrinsicOp{
-            std::get<clause::DefinedOperator::IntrinsicOperator>(
-                reductionOp->u)};
-
-        switch (intrinsicOp) {
-        case clause::DefinedOperator::IntrinsicOperator::Add:
-        case clause::DefinedOperator::IntrinsicOperator::Multiply:
-        case clause::DefinedOperator::IntrinsicOperator::AND:
-        case clause::DefinedOperator::IntrinsicOperator::EQV:
-        case clause::DefinedOperator::IntrinsicOperator::OR:
-        case clause::DefinedOperator::IntrinsicOperator::NEQV:
-          break;
-        default:
-          continue;
-        }
-        for (const Object &object : objects) {
-          if (const Fortran::semantics::Symbol *symbol = object.id()) {
-            mlir::Value reductionVal = converter.getSymbolAddress(*symbol);
-            if (auto declOp = reductionVal.getDefiningOp<hlfir::DeclareOp>())
-              reductionVal = declOp.getBase();
-            mlir::Type reductionType =
-                reductionVal.getType().cast<fir::ReferenceType>().getEleTy();
-            if (!reductionType.isa<fir::LogicalType>()) {
-              if (!reductionType.isIntOrIndexOrFloat())
-                continue;
-            }
-            for (mlir::OpOperand &reductionValUse : reductionVal.getUses()) {
-              if (auto loadOp =
-                      mlir::dyn_cast<fir::LoadOp>(reductionValUse.getOwner())) {
-                mlir::Value loadVal = loadOp.getRes();
-                if (reductionType.isa<fir::LogicalType>()) {
-                  mlir::Operation *reductionOp = findReductionChain(loadVal);
-                  fir::ConvertOp convertOp =
-                      getConvertFromReductionOp(reductionOp, loadVal);
-                  updateReduction(reductionOp, firOpBuilder, loadVal,
-                                  reductionVal, &convertOp);
-                  removeStoreOp(reductionOp, reductionVal);
-                } else if (mlir::Operation *reductionOp =
-                               findReductionChain(loadVal, &reductionVal)) {
-                  updateReduction(reductionOp, firOpBuilder, loadVal,
-                                  reductionVal);
-                }
-              }
-            }
-          }
-        }
-      } else if (const auto *reductionIntrinsic =
-                     std::get_if<clause::ProcedureDesignator>(&redOperator.u)) {
-        if (!ReductionProcessor::supportedIntrinsicProcReduction(
-                *reductionIntrinsic))
-          continue;
-        ReductionProcessor::ReductionIdentifier redId =
-            ReductionProcessor::getReductionType(*reductionIntrinsic);
-        for (const Object &object : objects) {
-          if (const Fortran::semantics::Symbol *symbol = object.id()) {
-            mlir::Value reductionVal = converter.getSymbolAddress(*symbol);
-            if (auto declOp = reductionVal.getDefiningOp<hlfir::DeclareOp>())
-              reductionVal = declOp.getBase();
-            for (const mlir::OpOperand &reductionValUse :
-                 reductionVal.getUses()) {
-              if (auto loadOp =
-                      mlir::dyn_cast<fir::LoadOp>(reductionValUse.getOwner())) {
-                mlir::Value loadVal = loadOp.getRes();
-                // Max is lowered as a compare -> select.
-                // Match the pattern here.
-                mlir::Operation *reductionOp =
-                    findReductionChain(loadVal, &reductionVal);
-                if (reductionOp == nullptr)
-                  continue;
-
-                if (redId == ReductionProcessor::ReductionIdentifier::MAX ||
-                    redId == ReductionProcessor::ReductionIdentifier::MIN) {
-                  assert(mlir::isa<mlir::arith::SelectOp>(reductionOp) &&
-                         "Selection Op not found in reduction intrinsic");
-                  mlir::Operation *compareOp =
-                      getCompareFromReductionOp(reductionOp, loadVal);
-                  updateReduction(compareOp, firOpBuilder, loadVal,
-                                  reductionVal);
-                }
-                if (redId == ReductionProcessor::ReductionIdentifier::IOR ||
-                    redId == ReductionProcessor::ReductionIdentifier::IEOR ||
-                    redId == ReductionProcessor::ReductionIdentifier::IAND) {
-                  updateReduction(reductionOp, firOpBuilder, loadVal,
-                                  reductionVal);
-                }
-              }
-            }
-          }
-        }
-      }
-    }
-  }
-}
-
 struct OpWithBodyGenInfo {
   /// A type for a code-gen callback function. This takes as argument the op for
   /// which the code is being generated and returns the arguments of the op's
@@ -2197,7 +1975,6 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
     // 2.9.3.1 SIMD construct
     createSimdLoop(converter, semaCtx, eval, ompDirective, loopOpClauseList,
                    currentLocation);
-    genOpenMPReduction(converter, semaCtx, loopOpClauseList);
   } else {
     createWsloop(converter, semaCtx, eval, ompDirective, loopOpClauseList,
                  endClauseList, currentLocation);

From 61717c1aa1f08eb57839a21fb2d9004739022e0d Mon Sep 17 00:00:00 2001
From: Jon Chesterfield <jonathanchesterfield@gmail.com>
Date: Tue, 16 Apr 2024 10:33:54 +0100
Subject: [PATCH 551/886] [Verifier] Reject va_start in non-variadic function
 (#88809)

A va_start intrinsic lowers to something derived from the variadic
parameter to the function. If there is no such parameter, it can't lower
meaningfully. Clang sema rejects the same with `error: 'va_start' used
in function with fixed args`.

Moves the existing lint warning into a verifier error. Updates the one
lit test that had a va_start in a non-variadic function.
---
 llvm/lib/Analysis/Lint.cpp                      | 5 +----
 llvm/lib/IR/Verifier.cpp                        | 5 +++++
 llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll | 2 +-
 llvm/test/Other/lint.ll                         | 7 -------
 llvm/test/Verifier/variadic.ll                  | 8 ++++++++
 5 files changed, 15 insertions(+), 12 deletions(-)
 create mode 100644 llvm/test/Verifier/variadic.ll

diff --git a/llvm/lib/Analysis/Lint.cpp b/llvm/lib/Analysis/Lint.cpp
index 0694c2995dfcc..1ab856ac8830a 100644
--- a/llvm/lib/Analysis/Lint.cpp
+++ b/llvm/lib/Analysis/Lint.cpp
@@ -350,10 +350,7 @@ void Lint::visitCallBase(CallBase &I) {
     }
 
     case Intrinsic::vastart:
-      Check(I.getParent()->getParent()->isVarArg(),
-            "Undefined behavior: va_start called in a non-varargs function",
-            &I);
-
+      // vastart in non-varargs function is rejected by the verifier
       visitMemoryReference(I, MemoryLocation::getForArgument(&I, 0, TLI),
                            std::nullopt, nullptr, MemRef::Read | MemRef::Write);
       break;
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 516d4a0515569..4cd61e6e531bf 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -5798,6 +5798,11 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
 
     break;
   }
+  case Intrinsic::vastart: {
+    Check(Call.getFunction()->isVarArg(),
+          "va_start called in a non-varargs function");
+    break;
+  }
   case Intrinsic::vector_reduce_and:
   case Intrinsic::vector_reduce_or:
   case Intrinsic::vector_reduce_xor:
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
index bd576d0f70e9c..8c6e01d934c2d 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
@@ -3,7 +3,7 @@
 
 
 declare void @llvm.va_start(ptr)
-define void @test_va_start(ptr %list) {
+define void @test_va_start(ptr %list, ...) {
 ; CHECK-LABEL: name: test_va_start
 ; CHECK: [[LIST:%[0-9]+]]:_(p0) = COPY $x0
 ; CHECK-IOS: G_VASTART [[LIST]](p0) :: (store (s64) into %ir.list, align 1)
diff --git a/llvm/test/Other/lint.ll b/llvm/test/Other/lint.ll
index 6b31b31a78c98..6fd2d40cd2f29 100644
--- a/llvm/test/Other/lint.ll
+++ b/llvm/test/Other/lint.ll
@@ -124,13 +124,6 @@ define void @0() nounwind {
   ret void
 }
 
-; CHECK: va_start called in a non-varargs function
-declare void @llvm.va_start(ptr)
-define void @not_vararg(ptr %p) nounwind {
-  call void @llvm.va_start(ptr %p)
-  ret void
-}
-
 ; CHECK: Undefined behavior: Branch to non-blockaddress
 define void @use_indbr() {
   indirectbr ptr @foo, [label %block]
diff --git a/llvm/test/Verifier/variadic.ll b/llvm/test/Verifier/variadic.ll
new file mode 100644
index 0000000000000..55e4a4da0a920
--- /dev/null
+++ b/llvm/test/Verifier/variadic.ll
@@ -0,0 +1,8 @@
+; RUN: not opt -S -passes=verify 2>&1 < %s | FileCheck %s
+
+; CHECK: va_start called in a non-varargs function
+declare void @llvm.va_start(ptr)
+define void @not_vararg(ptr %p) nounwind {
+  call void @llvm.va_start(ptr %p)
+  ret void
+}

From 1ca6b4475c02e5d022ec6b35dbb65d0f11409a88 Mon Sep 17 00:00:00 2001
From: Ivan Butygin <ivan.butygin@gmail.com>
Date: Tue, 16 Apr 2024 12:39:57 +0300
Subject: [PATCH 552/886] [mlir][scf] `scf.while` uplifting: optimize op
 matching (#88813)

Instead of iterating over potential induction var uses looking for
suitable `arith.addi`, try to trace it back from yield argument.
---
 .../SCF/Transforms/UpliftWhileToFor.cpp       | 36 ++++++++-----------
 1 file changed, 14 insertions(+), 22 deletions(-)

diff --git a/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp b/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
index fea2f659535bb..7b4024b6861a7 100644
--- a/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
+++ b/mlir/lib/Dialect/SCF/Transforms/UpliftWhileToFor.cpp
@@ -101,38 +101,30 @@ FailureOr<scf::ForOp> mlir::scf::upliftWhileToForLoop(RewriterBase &rewriter,
 
   Block *afterBody = loop.getAfterBody();
   scf::YieldOp afterTerm = loop.getYieldOp();
-  auto argNumber = inductionVar.getArgNumber();
-  auto afterTermIndArg = afterTerm.getResults()[argNumber];
+  unsigned argNumber = inductionVar.getArgNumber();
+  Value afterTermIndArg = afterTerm.getResults()[argNumber];
 
-  auto inductionVarAfter = afterBody->getArgument(argNumber);
-
-  Value step;
+  Value inductionVarAfter = afterBody->getArgument(argNumber);
 
   // Find suitable `addi` op inside `after` block, one of the args must be an
   // Induction var passed from `before` block and second arg must be defined
   // outside of the loop and will be considered step value.
   // TODO: Add `subi` support?
-  for (auto &use : inductionVarAfter.getUses()) {
-    auto owner = dyn_cast<arith::AddIOp>(use.getOwner());
-    if (!owner)
-      continue;
-
-    auto other =
-        (inductionVarAfter == owner.getLhs() ? owner.getRhs() : owner.getLhs());
-    if (!dom.properlyDominates(other, loop))
-      continue;
-
-    if (afterTermIndArg != owner.getResult())
-      continue;
+  auto addOp = afterTermIndArg.getDefiningOp<arith::AddIOp>();
+  if (!addOp)
+    return rewriter.notifyMatchFailure(loop, "Didn't found suitable 'addi' op");
 
-    step = other;
-    break;
+  Value step;
+  if (addOp.getLhs() == inductionVarAfter) {
+    step = addOp.getRhs();
+  } else if (addOp.getRhs() == inductionVarAfter) {
+    step = addOp.getLhs();
   }
 
-  if (!step)
-    return rewriter.notifyMatchFailure(loop, "Didn't found suitable 'addi' op");
+  if (!step || !dom.properlyDominates(step, loop))
+    return rewriter.notifyMatchFailure(loop, "Invalid 'addi' form");
 
-  auto lb = loop.getInits()[argNumber];
+  Value lb = loop.getInits()[argNumber];
 
   assert(lb.getType().isIntOrIndex());
   assert(lb.getType() == ub.getType());

From aae08f4f8ef7bc9c35d263ed974679130c49f5fc Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Tue, 16 Apr 2024 10:40:46 +0100
Subject: [PATCH 553/886] [MLIR][OpenMP] Make omp.taskloop into a loop wrapper
 (#87253)

This patch updates the definition of `omp.taskloop` to enforce the
restrictions of a wrapper operation.
---
 .../Dialect/OpenMP/OpenMPClauseOperands.h     |   9 +-
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |  39 ++-
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  |  15 +-
 mlir/test/Dialect/OpenMP/invalid.mlir         |  92 +++++---
 mlir/test/Dialect/OpenMP/ops.mlir             | 222 ++++++++++--------
 5 files changed, 218 insertions(+), 159 deletions(-)

diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
index 304a9740d91ed..27a766aceb316 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
@@ -284,11 +284,10 @@ using TaskgroupClauseOps =
     detail::Clauses<AllocateClauseOps, TaskReductionClauseOps>;
 
 using TaskloopClauseOps =
-    detail::Clauses<AllocateClauseOps, CollapseClauseOps, FinalClauseOps,
-                    GrainsizeClauseOps, IfClauseOps, InReductionClauseOps,
-                    LoopRelatedOps, MergeableClauseOps, NogroupClauseOps,
-                    NumTasksClauseOps, PriorityClauseOps, PrivateClauseOps,
-                    ReductionClauseOps, UntiedClauseOps>;
+    detail::Clauses<AllocateClauseOps, FinalClauseOps, GrainsizeClauseOps,
+                    IfClauseOps, InReductionClauseOps, MergeableClauseOps,
+                    NogroupClauseOps, NumTasksClauseOps, PriorityClauseOps,
+                    PrivateClauseOps, ReductionClauseOps, UntiedClauseOps>;
 
 using TaskwaitClauseOps = detail::Clauses<DependClauseOps, NowaitClauseOps>;
 
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 7e677c9839f60..82be7ad31a158 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1030,10 +1030,10 @@ def TaskOp : OpenMP_Op<"task", [AttrSizedOperandSegments,
 }
 
 def TaskloopOp : OpenMP_Op<"taskloop", [AttrSizedOperandSegments,
-                           AutomaticAllocationScope, RecursiveMemoryEffects,
-                           AllTypesMatch<["lowerBound", "upperBound", "step"]>,
+                           AutomaticAllocationScope,
                            DeclareOpInterfaceMethods<LoopWrapperInterface>,
-                           ReductionClauseInterface]> {
+                           RecursiveMemoryEffects, ReductionClauseInterface,
+                           SingleBlockImplicitTerminator<"TerminatorOp">]> {
   let summary = "taskloop construct";
   let description = [{
     The taskloop construct specifies that the iterations of one or more
@@ -1041,21 +1041,19 @@ def TaskloopOp : OpenMP_Op<"taskloop", [AttrSizedOperandSegments,
     iterations are distributed across tasks generated by the construct and
     scheduled to be executed.
 
-    The `lowerBound` and `upperBound` specify a half-open range: the range
-    includes the lower bound but does not include the upper bound. If the
-    `inclusive` attribute is specified then the upper bound is also included.
-    The `step` specifies the loop step.
-
-    The body region can contain any number of blocks.
+    The body region can contain a single block which must contain a single
+    operation and a terminator. The operation must be another compatible loop
+    wrapper or an `omp.loop_nest`.
 
     ```
-    omp.taskloop <clauses>
-    for (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
-      %a = load %arrA[%i1, %i2] : memref<?x?xf32>
-      %b = load %arrB[%i1, %i2] : memref<?x?xf32>
-      %sum = arith.addf %a, %b : f32
-      store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
-      omp.terminator
+    omp.taskloop <clauses> {
+      omp.loop_nest (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
+        %a = load %arrA[%i1, %i2] : memref<?x?xf32>
+        %b = load %arrB[%i1, %i2] : memref<?x?xf32>
+        %sum = arith.addf %a, %b : f32
+        store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
+        omp.yield
+      }
     }
     ```
 
@@ -1132,11 +1130,7 @@ def TaskloopOp : OpenMP_Op<"taskloop", [AttrSizedOperandSegments,
     created.
   }];
 
-  let arguments = (ins Variadic<IntLikeType>:$lowerBound,
-                       Variadic<IntLikeType>:$upperBound,
-                       Variadic<IntLikeType>:$step,
-                       UnitAttr:$inclusive,
-                       Optional<I1>:$if_expr,
+  let arguments = (ins Optional<I1>:$if_expr,
                        Optional<I1>:$final_expr,
                        UnitAttr:$untied,
                        UnitAttr:$mergeable,
@@ -1179,8 +1173,7 @@ def TaskloopOp : OpenMP_Op<"taskloop", [AttrSizedOperandSegments,
           |`grain_size` `(` $grain_size `:` type($grain_size) `)`
           |`num_tasks` `(` $num_tasks `:` type($num_tasks) `)`
           |`nogroup` $nogroup
-    ) `for` custom<LoopControl>($region, $lowerBound, $upperBound, $step,
-                                  type($step), $inclusive) attr-dict
+    ) $region attr-dict
   }];
 
   let extraClassDeclaration = [{
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 365b3c4246bfd..e500d0fca741f 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1829,9 +1829,8 @@ void TaskloopOp::build(OpBuilder &builder, OperationState &state,
   MLIRContext *ctx = builder.getContext();
   // TODO Store clauses in op: reductionByRefAttr, privateVars, privatizers.
   TaskloopOp::build(
-      builder, state, clauses.loopLBVar, clauses.loopUBVar, clauses.loopStepVar,
-      clauses.loopInclusiveAttr, clauses.ifVar, clauses.finalVar,
-      clauses.untiedAttr, clauses.mergeableAttr, clauses.inReductionVars,
+      builder, state, clauses.ifVar, clauses.finalVar, clauses.untiedAttr,
+      clauses.mergeableAttr, clauses.inReductionVars,
       makeArrayAttr(ctx, clauses.inReductionDeclSymbols), clauses.reductionVars,
       makeArrayAttr(ctx, clauses.reductionDeclSymbols), clauses.priorityVar,
       clauses.allocateVars, clauses.allocatorVars, clauses.grainsizeVar,
@@ -1870,6 +1869,16 @@ LogicalResult TaskloopOp::verify() {
         "the grainsize clause and num_tasks clause are mutually exclusive and "
         "may not appear on the same taskloop directive");
   }
+
+  if (!isWrapper())
+    return emitOpError() << "must be a loop wrapper";
+
+  if (LoopWrapperInterface nested = getNestedWrapper()) {
+    // Check for the allowed leaf constructs that may appear in a composite
+    // construct directly after TASKLOOP.
+    if (!isa<SimdLoopOp>(nested))
+      return emitError() << "only supported nested wrapper is 'omp.simdloop'";
+  }
   return success();
 }
 
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index f8739887214bf..7f86a7f5b3182 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -1580,10 +1580,11 @@ func.func @omp_cancellationpoint2() {
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testmemref = "test.memref"() : () -> (memref<i32>)
   // expected-error @below {{expected equal sizes for allocate and allocator variables}}
-  "omp.taskloop"(%lb, %ub, %ub, %lb, %step, %step, %testmemref) ({
-  ^bb0(%arg3: i32, %arg4: i32):
-    "omp.terminator"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 2, 2, 2, 0, 0, 0, 0, 0, 1, 0, 0, 0>} : (i32, i32, i32, i32, i32, i32, memref<i32>) -> ()
+  "omp.taskloop"(%testmemref) ({
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }) {operandSegmentSizes = array<i32: 0, 0, 0, 0, 0, 1, 0, 0, 0>} : (memref<i32>) -> ()
   return
 }
 
@@ -1593,10 +1594,11 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
   %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{expected as many reduction symbol references as reduction variables}}
-  "omp.taskloop"(%lb, %ub, %ub, %lb, %step, %step, %testf32, %testf32_2) ({
-  ^bb0(%arg3: i32, %arg4: i32):
-    "omp.terminator"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 2, 2, 2, 0, 0, 0, 2, 0, 0, 0, 0, 0>, reductions = [@add_f32]} : (i32, i32, i32, i32, i32, i32, !llvm.ptr, !llvm.ptr) -> ()
+  "omp.taskloop"(%testf32, %testf32_2) ({
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }) {operandSegmentSizes = array<i32: 0, 0, 0, 2, 0, 0, 0, 0, 0>, reductions = [@add_f32]} : (!llvm.ptr, !llvm.ptr) -> ()
   return
 }
 
@@ -1604,12 +1606,12 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
 
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
-  %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{expected as many reduction symbol references as reduction variables}}
-  "omp.taskloop"(%lb, %ub, %ub, %lb, %step, %step, %testf32) ({
-  ^bb0(%arg3: i32, %arg4: i32):
-    "omp.terminator"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 2, 2, 2, 0, 0, 0, 1, 0, 0, 0, 0, 0>, reductions = [@add_f32, @add_f32]} : (i32, i32, i32, i32, i32, i32, !llvm.ptr) -> ()
+  "omp.taskloop"(%testf32) ({
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }) {operandSegmentSizes = array<i32: 0, 0, 0, 1, 0, 0, 0, 0, 0>, reductions = [@add_f32, @add_f32]} : (!llvm.ptr) -> ()
   return
 }
 
@@ -1619,10 +1621,11 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
   %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{expected as many reduction symbol references as reduction variables}}
-  "omp.taskloop"(%lb, %ub, %ub, %lb, %step, %step, %testf32, %testf32_2) ({
-  ^bb0(%arg3: i32, %arg4: i32):
-    "omp.terminator"() : () -> ()
-  }) {in_reductions = [@add_f32], operandSegmentSizes = array<i32: 2, 2, 2, 0, 0, 2, 0, 0, 0, 0, 0, 0>} : (i32, i32, i32, i32, i32, i32, !llvm.ptr, !llvm.ptr) -> ()
+  "omp.taskloop"(%testf32, %testf32_2) ({
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }) {in_reductions = [@add_f32], operandSegmentSizes = array<i32: 0, 0, 2, 0, 0, 0, 0, 0, 0>} : (!llvm.ptr, !llvm.ptr) -> ()
   return
 }
 
@@ -1630,12 +1633,12 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
 
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
-  %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{expected as many reduction symbol references as reduction variables}}
-  "omp.taskloop"(%lb, %ub, %ub, %lb, %step, %step, %testf32_2) ({
-  ^bb0(%arg3: i32, %arg4: i32):
-    "omp.terminator"() : () -> ()
-  }) {in_reductions = [@add_f32, @add_f32], operandSegmentSizes = array<i32: 2, 2, 2, 0, 0, 1, 0, 0, 0, 0, 0, 0>} : (i32, i32, i32, i32, i32, i32, !llvm.ptr) -> ()
+  "omp.taskloop"(%testf32) ({
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }) {in_reductions = [@add_f32, @add_f32], operandSegmentSizes = array<i32: 0, 0, 1, 0, 0, 0, 0, 0, 0>} : (!llvm.ptr) -> ()
   return
 }
 
@@ -1657,9 +1660,10 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
   %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{if a reduction clause is present on the taskloop directive, the nogroup clause must not be specified}}
-  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr) nogroup
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    omp.terminator
+  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr) nogroup {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
   }
   return
 }
@@ -1681,9 +1685,10 @@ combiner {
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
   // expected-error @below {{the same list item cannot appear in both a reduction and an in_reduction clause}}
-  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr) in_reduction(@add_f32 -> %testf32 : !llvm.ptr)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    omp.terminator
+  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr) in_reduction(@add_f32 -> %testf32 : !llvm.ptr) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
   }
   return
 }
@@ -1693,8 +1698,20 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
   %testi64 = "test.i64"() : () -> (i64)
   // expected-error @below {{the grainsize clause and num_tasks clause are mutually exclusive and may not appear on the same taskloop directive}}
-  omp.taskloop grain_size(%testi64: i64) num_tasks(%testi64: i64)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+  omp.taskloop grain_size(%testi64: i64) num_tasks(%testi64: i64) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      omp.yield
+    }
+  }
+  return
+}
+
+// -----
+
+func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
+  // expected-error @below {{op must be a loop wrapper}}
+  omp.taskloop {
+    %0 = arith.constant 0 : i32
     omp.terminator
   }
   return
@@ -1702,6 +1719,21 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
 
 // -----
 
+func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
+  // expected-error @below {{only supported nested wrapper is 'omp.simdloop'}}
+  omp.taskloop {
+    omp.distribute {
+      omp.loop_nest (%iv) : i32 = (%lb) to (%ub) step (%step) {
+        omp.yield
+      }
+      omp.terminator
+    }
+  }
+  return
+}
+
+// -----
+
 func.func @omp_threadprivate() {
   %1 = llvm.mlir.addressof @_QFsubEx : !llvm.ptr
   // expected-error @below {{op failed to verify that all of {sym_addr, tls_addr} have same type}}
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index 018b82e995d7d..802e1795b3fff 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -171,6 +171,23 @@ func.func @omp_loop_nest(%lb : index, %ub : index, %step : index) -> () {
     omp.yield
   }
 
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) step (%{{.*}})
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb0(%iv2: index):
+      // CHECK: test.op1
+      "test.op1"(%lb) : (index) -> ()
+      // CHECK: test.op2
+      "test.op2"() : () -> ()
+      // CHECK: omp.yield
+      omp.yield
+    }) : (index, index, index) -> ()
+    omp.yield
+  }
+
   return
 }
 
@@ -209,6 +226,22 @@ func.func @omp_loop_nest_pretty(%lb : index, %ub : index, %step : index) -> () {
     omp.yield
   }
 
+  // TODO Remove induction variables from omp.wsloop.
+  omp.wsloop for (%iv) : index = (%lb) to (%ub) step (%step) {
+    // CHECK: omp.loop_nest
+    // CHECK-SAME: (%{{.*}}) : index =
+    // CHECK-SAME: (%{{.*}}) to (%{{.*}}) step (%{{.*}})
+    omp.loop_nest (%iv2) : index = (%lb) to (%ub) step (%step)  {
+      // CHECK: test.op1
+      "test.op1"(%lb) : (index) -> ()
+      // CHECK: test.op2
+      "test.op2"() : () -> ()
+      // CHECK: omp.yield
+      omp.yield
+    }
+    omp.yield
+  }
+
   return
 }
 
@@ -2024,135 +2057,128 @@ func.func @omp_taskgroup_clauses() -> () {
 // CHECK-LABEL: @omp_taskloop
 func.func @omp_taskloop(%lb: i32, %ub: i32, %step: i32) -> () {
 
-  // CHECK: omp.taskloop for (%{{.+}}) : i32 = (%{{.+}}) to (%{{.+}}) step (%{{.+}}) {
-  omp.taskloop for (%i) : i32 = (%lb) to (%ub) step (%step)  {
-    // CHECK: omp.terminator
-    omp.terminator
-  }
-
-  // CHECK: omp.taskloop for (%{{.+}}) : i32 = (%{{.+}}) to (%{{.+}}) step (%{{.+}}) {
-  omp.taskloop for (%i) : i32 = (%lb) to (%ub) step (%step)  {
-    // CHECK: test.op1
-    "test.op1"(%lb) : (i32) -> ()
-    // CHECK: test.op2
-    "test.op2"() : () -> ()
-    // CHECK: omp.terminator
-    omp.terminator
-  }
-
-  // CHECK: omp.taskloop for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
-  }
-
-  // CHECK: omp.taskloop for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) inclusive step (%{{.+}}, %{{.+}}) {
-  omp.taskloop for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) inclusive step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop {
+  omp.taskloop {
+    omp.loop_nest (%i) : i32 = (%lb) to (%ub) step (%step)  {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   %testbool = "test.bool"() : () -> (i1)
 
-  // CHECK: omp.taskloop if(%{{[^)]+}})
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop if(%testbool)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop if(%{{[^)]+}}) {
+  omp.taskloop if(%testbool) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop final(%{{[^)]+}})
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop final(%testbool)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop final(%{{[^)]+}}) {
+  omp.taskloop final(%testbool) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop untied
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop untied
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop untied {
+  omp.taskloop untied {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop mergeable
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop mergeable
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop mergeable {
+  omp.taskloop mergeable {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   %testf32 = "test.f32"() : () -> (!llvm.ptr)
   %testf32_2 = "test.f32"() : () -> (!llvm.ptr)
-  // CHECK: omp.taskloop in_reduction(@add_f32 -> %{{.+}} : !llvm.ptr, @add_f32 -> %{{.+}} : !llvm.ptr)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop in_reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop in_reduction(@add_f32 -> %{{.+}} : !llvm.ptr, @add_f32 -> %{{.+}} : !llvm.ptr) {
+  omp.taskloop in_reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop reduction(@add_f32 -> %{{.+}} : !llvm.ptr, @add_f32 -> %{{.+}} : !llvm.ptr)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop reduction(@add_f32 -> %{{.+}} : !llvm.ptr, @add_f32 -> %{{.+}} : !llvm.ptr) {
+  omp.taskloop reduction(@add_f32 -> %testf32 : !llvm.ptr, @add_f32 -> %testf32_2 : !llvm.ptr) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop in_reduction(@add_f32 -> %{{.+}} : !llvm.ptr) reduction(@add_f32 -> %{{.+}} : !llvm.ptr)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop in_reduction(@add_f32 -> %testf32 : !llvm.ptr) reduction(@add_f32 -> %testf32_2 : !llvm.ptr)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop in_reduction(@add_f32 -> %{{.+}} : !llvm.ptr) reduction(@add_f32 -> %{{.+}} : !llvm.ptr) {
+  omp.taskloop in_reduction(@add_f32 -> %testf32 : !llvm.ptr) reduction(@add_f32 -> %testf32_2 : !llvm.ptr) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   %testi32 = "test.i32"() : () -> (i32)
-  // CHECK: omp.taskloop priority(%{{[^:]+}}: i32)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop priority(%testi32: i32)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop priority(%{{[^:]+}}: i32) {
+  omp.taskloop priority(%testi32: i32) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   %testmemref = "test.memref"() : () -> (memref<i32>)
-  // CHECK: omp.taskloop allocate(%{{.+}} : memref<i32> -> %{{.+}} : memref<i32>)
-  omp.taskloop allocate(%testmemref : memref<i32> -> %testmemref : memref<i32>)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop allocate(%{{.+}} : memref<i32> -> %{{.+}} : memref<i32>) {
+  omp.taskloop allocate(%testmemref : memref<i32> -> %testmemref : memref<i32>) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   %testi64 = "test.i64"() : () -> (i64)
-  // CHECK: omp.taskloop grain_size(%{{[^:]+}}: i64)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop grain_size(%testi64: i64)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop grain_size(%{{[^:]+}}: i64) {
+  omp.taskloop grain_size(%testi64: i64) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop num_tasks(%{{[^:]+}}: i64)
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop num_tasks(%testi64: i64)
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop num_tasks(%{{[^:]+}}: i64) {
+  omp.taskloop num_tasks(%testi64: i64) {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
-  // CHECK: omp.taskloop nogroup
-  // CHECK-SAME: for (%{{.+}}, %{{.+}}) : i32 = (%{{.+}}, %{{.+}}) to (%{{.+}}, %{{.+}}) step (%{{.+}}, %{{.+}}) {
-  omp.taskloop nogroup
-  for (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
-    // CHECK: omp.terminator
-    omp.terminator
+  // CHECK: omp.taskloop nogroup {
+  omp.taskloop nogroup {
+    omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+      // CHECK: omp.yield
+      omp.yield
+    }
+  }
+
+  // CHECK: omp.taskloop {
+  omp.taskloop {
+    // TODO Remove induction variables from omp.simdloop.
+    omp.simdloop for (%iv) : i32 = (%lb) to (%ub) step (%step) {
+      omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
+        // CHECK: omp.yield
+        omp.yield
+      }
+      // CHECK: omp.yield
+      omp.yield
+    }
   }
 
   // CHECK: return

From ca4cf973279a3991248056a73bcb2bac8b37d035 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Mon, 15 Apr 2024 12:16:55 +0200
Subject: [PATCH 554/886] [clang][Interp][NFC] Fix Pointer::isZero() for block
 pointers

We don't need to consider the offset here anymore since we now
have proper integral pointers.
---
 clang/lib/AST/Interp/Pointer.h | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/clang/lib/AST/Interp/Pointer.h b/clang/lib/AST/Interp/Pointer.h
index fcd00aac62f93..b4475577b7462 100644
--- a/clang/lib/AST/Interp/Pointer.h
+++ b/clang/lib/AST/Interp/Pointer.h
@@ -241,13 +241,10 @@ class Pointer {
 
   /// Checks if the pointer is null.
   bool isZero() const {
-    if (Offset != 0)
-      return false;
-
     if (isBlockPointer())
       return asBlockPointer().Pointee == nullptr;
     assert(isIntegralPointer());
-    return asIntPointer().Value == 0;
+    return asIntPointer().Value == 0 && Offset == 0;
   }
   /// Checks if the pointer is live.
   bool isLive() const {

From 422bf13f336923da89055f8e70e49e7e9ced2c70 Mon Sep 17 00:00:00 2001
From: choikwa <5455710+choikwa@users.noreply.github.com>
Date: Tue, 16 Apr 2024 06:04:37 -0400
Subject: [PATCH 555/886] =?UTF-8?q?[AMDGPU]=20In=20VectorLegalizer::Expand?=
 =?UTF-8?q?,=20if=20UnrollVectorOp=20returns=20Load,=20=E2=80=A6=20(#88475?=
 =?UTF-8?q?)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…return only Load since other output is chain.

Added testcase that showed mismatched expected arity when Load and chain
were returned as separate items after
003b58f65bdd5d9c7d0c1b355566c9ef430c0e7d
---
 .../SelectionDAG/LegalizeVectorOps.cpp        | 10 +-
 llvm/test/CodeGen/AMDGPU/build_vector.ll      | 97 +++++++++++++++++++
 2 files changed, 105 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
index 57a3f6a65e002..7a9cfdf5c3fda 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
@@ -1159,8 +1159,14 @@ void VectorLegalizer::Expand(SDNode *Node, SmallVectorImpl<SDValue> &Results) {
   }
 
   SDValue Unrolled = DAG.UnrollVectorOp(Node);
-  for (unsigned I = 0, E = Unrolled->getNumValues(); I != E; ++I)
-    Results.push_back(Unrolled.getValue(I));
+  if (Node->getNumValues() == 1) {
+    Results.push_back(Unrolled);
+  } else {
+    assert(Node->getNumValues() == Unrolled->getNumValues() &&
+      "VectorLegalizer Expand returned wrong number of results!");
+    for (unsigned I = 0, E = Unrolled->getNumValues(); I != E; ++I)
+      Results.push_back(Unrolled.getValue(I));
+  }
 }
 
 SDValue VectorLegalizer::ExpandSELECT(SDNode *Node) {
diff --git a/llvm/test/CodeGen/AMDGPU/build_vector.ll b/llvm/test/CodeGen/AMDGPU/build_vector.ll
index 37412ac3aa541..99755133f36d6 100644
--- a/llvm/test/CodeGen/AMDGPU/build_vector.ll
+++ b/llvm/test/CodeGen/AMDGPU/build_vector.ll
@@ -3,6 +3,7 @@
 ; RUN: llc < %s -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s --check-prefixes=GFX8,GFX678,ALL
 ; RUN: llc < %s -mtriple=amdgcn-amd-amdpal -mcpu=gfx1030 -verify-machineinstrs | FileCheck %s --check-prefixes=GFX10,GFX1011,ALL
 ; RUN: llc < %s -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs | FileCheck %s --check-prefixes=GFX11,GFX1011,ALL
+; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx940 | FileCheck %s --check-prefixes=GFX940,ALL
 
 ; ALL-LABEL: {{^}}build_vector2:
 ; R600: MOV
@@ -96,3 +97,99 @@ define amdgpu_kernel void @build_vector_v2i16_trunc (ptr addrspace(1) %out, i32
   store <2 x i16> %ins.1, ptr addrspace(1) %out
   ret void
 }
+
+; R600-LABEL: build_v2i32_from_v4i16_shuffle:
+; R600:       ; %bb.0: ; %entry
+; R600-NEXT:    ALU 0, @10, KC0[], KC1[]
+; R600-NEXT:    TEX 1 @6
+; R600-NEXT:    ALU 4, @11, KC0[CB0:0-32], KC1[]
+; R600-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; R600-NEXT:    CF_END
+; R600-NEXT:    PAD
+; R600-NEXT:    Fetch clause starting at 6:
+; R600-NEXT:     VTX_READ_16 T1.X, T0.X, 48, #3
+; R600-NEXT:     VTX_READ_16 T0.X, T0.X, 44, #3
+; R600-NEXT:    ALU clause starting at 10:
+; R600-NEXT:     MOV * T0.X, 0.0,
+; R600-NEXT:    ALU clause starting at 11:
+; R600-NEXT:     LSHL * T0.Y, T1.X, literal.x,
+; R600-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; R600-NEXT:     LSHL T0.X, T0.X, literal.x,
+; R600-NEXT:     LSHR * T1.X, KC0[2].Y, literal.y,
+; R600-NEXT:    16(2.242078e-44), 2(2.802597e-45)
+;
+; GFX6-LABEL: build_v2i32_from_v4i16_shuffle:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX6-NEXT:    s_mov_b32 s7, 0xf000
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX6-NEXT:    s_lshl_b32 s2, s2, 16
+; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    s_mov_b32 s4, s0
+; GFX6-NEXT:    s_mov_b32 s5, s1
+; GFX6-NEXT:    v_mov_b32_e32 v0, s2
+; GFX6-NEXT:    v_mov_b32_e32 v1, s3
+; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX8-LABEL: build_v2i32_from_v4i16_shuffle:
+; GFX8:       ; %bb.0: ; %entry
+; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX8-NEXT:    s_mov_b32 s7, 0xf000
+; GFX8-NEXT:    s_mov_b32 s6, -1
+; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX8-NEXT:    s_mov_b32 s4, s0
+; GFX8-NEXT:    s_mov_b32 s5, s1
+; GFX8-NEXT:    s_lshl_b32 s0, s3, 16
+; GFX8-NEXT:    s_lshl_b32 s1, s2, 16
+; GFX8-NEXT:    v_mov_b32_e32 v0, s1
+; GFX8-NEXT:    v_mov_b32_e32 v1, s0
+; GFX8-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX8-NEXT:    s_endpgm
+;
+; GFX10-LABEL: build_v2i32_from_v4i16_shuffle:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_lshl_b32 s2, s2, 16
+; GFX10-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX10-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: build_v2i32_from_v4i16_shuffle:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    s_lshl_b32 s2, s2, 16
+; GFX11-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX940-LABEL: build_v2i32_from_v4i16_shuffle:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX940-NEXT:    v_mov_b32_e32 v2, 0
+; GFX940-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-NEXT:    s_lshl_b32 s3, s3, 16
+; GFX940-NEXT:    s_lshl_b32 s2, s2, 16
+; GFX940-NEXT:    v_mov_b32_e32 v0, s2
+; GFX940-NEXT:    v_mov_b32_e32 v1, s3
+; GFX940-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1] sc0 sc1
+; GFX940-NEXT:    s_endpgm
+define amdgpu_kernel void @build_v2i32_from_v4i16_shuffle(ptr addrspace(1) %out, <4 x i16> %in) {
+entry:
+  %shuf = shufflevector <4 x i16> %in, <4 x i16> zeroinitializer, <2 x i32> <i32 0, i32 2>
+  %zextended = zext <2 x i16> %shuf to <2 x i32>
+  %shifted = shl <2 x i32> %zextended, <i32 16, i32 16>
+  store <2 x i32> %shifted, ptr addrspace(1) %out
+  ret void
+}

From 4dd5180a2d43b088d7637c30c2654f3c01c46987 Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Tue, 16 Apr 2024 11:08:25 +0100
Subject: [PATCH 556/886] [Flang][OpenMP][Lower] Split MLIR codegen for clauses
 and constructs (#86963)

This patch performs several cleanups with the main purpose of
normalizing the code patterns used to trigger codegen for MLIR OpenMP
operations and making the processing of clauses and constructs
independent. The following changes are made:

- Clean up unused `directive` argument to
`ClauseProcessor::processMap()`.
- Move general helper functions in OpenMP.cpp to the appropriate section
of the file.
- Create `gen<OpName>Clauses()` functions containing the clause
processing code specific for the associated OpenMP construct.
- Update `gen<OpName>Op()` functions to call the corresponding
`gen<OpName>Clauses()` function.
- Sort calls to `ClauseProcessor::process<ClauseName>()` alphabetically,
to avoid inadvertently relying on some arbitrary order. Update some
tests that broke due to the order change.
- Normalize `genOMP()` functions so they all delegate the generation of
MLIR to `gen<OpName>Op()` functions following the same pattern.
- Only process `nowait` clause on `TARGET` constructs if not compiling
for the target device.

A later patch can move the calls to `gen<OpName>Clauses()` out of
`gen<OpName>Op()` functions and passing completed clause structures
instead, in preparation to supporting composite constructs. That will
make it possible to reuse clause processing for a given leaf construct
when appearing alone or in a combined or composite construct, while
controlling where the associated code is produced.
---
 flang/lib/Lower/OpenMP/ClauseProcessor.cpp    |    4 +-
 flang/lib/Lower/OpenMP/ClauseProcessor.h      |    3 +-
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 2136 +++++++++--------
 flang/test/Lower/OpenMP/FIR/target.f90        |    2 +-
 flang/test/Lower/OpenMP/target.f90            |    2 +-
 .../use-device-ptr-to-use-device-addr.f90     |    4 +-
 6 files changed, 1198 insertions(+), 953 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index ae0d8bd37228d..4c51b61f6bf02 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -832,8 +832,8 @@ createMapInfoOp(fir::FirOpBuilder &builder, mlir::Location loc,
 }
 
 bool ClauseProcessor::processMap(
-    mlir::Location currentLocation, const llvm::omp::Directive &directive,
-    Fortran::lower::StatementContext &stmtCtx, mlir::omp::MapClauseOps &result,
+    mlir::Location currentLocation, Fortran::lower::StatementContext &stmtCtx,
+    mlir::omp::MapClauseOps &result,
     llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *mapSyms,
     llvm::SmallVectorImpl<mlir::Location> *mapSymLocs,
     llvm::SmallVectorImpl<mlir::Type> *mapSymTypes) const {
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index aa2c14b61e756..3f9701310ebae 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -114,8 +114,7 @@ class ClauseProcessor {
   // They may be used later on to create the block_arguments for some of the
   // target directives that require it.
   bool processMap(
-      mlir::Location currentLocation, const llvm::omp::Directive &directive,
-      Fortran::lower::StatementContext &stmtCtx,
+      mlir::Location currentLocation, Fortran::lower::StatementContext &stmtCtx,
       mlir::omp::MapClauseOps &result,
       llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> *mapSyms =
           nullptr,
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 352ca66e8735b..9b99752236662 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -222,6 +222,276 @@ createAndSetPrivatizedLoopVar(Fortran::lower::AbstractConverter &converter,
   return storeOp;
 }
 
+// This helper function implements the functionality of "promoting"
+// non-CPTR arguments of use_device_ptr to use_device_addr
+// arguments (automagic conversion of use_device_ptr ->
+// use_device_addr in these cases). The way we do so currently is
+// through the shuffling of operands from the devicePtrOperands to
+// deviceAddrOperands where neccesary and re-organizing the types,
+// locations and symbols to maintain the correct ordering of ptr/addr
+// input -> BlockArg.
+//
+// This effectively implements some deprecated OpenMP functionality
+// that some legacy applications unfortunately depend on
+// (deprecated in specification version 5.2):
+//
+// "If a list item in a use_device_ptr clause is not of type C_PTR,
+//  the behavior is as if the list item appeared in a use_device_addr
+//  clause. Support for such list items in a use_device_ptr clause
+//  is deprecated."
+static void promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
+    mlir::omp::UseDeviceClauseOps &clauseOps,
+    llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
+    llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
+        &useDeviceSymbols) {
+  auto moveElementToBack = [](size_t idx, auto &vector) {
+    auto *iter = std::next(vector.begin(), idx);
+    vector.push_back(*iter);
+    vector.erase(iter);
+  };
+
+  // Iterate over our use_device_ptr list and shift all non-cptr arguments into
+  // use_device_addr.
+  for (auto *it = clauseOps.useDevicePtrVars.begin();
+       it != clauseOps.useDevicePtrVars.end();) {
+    if (!fir::isa_builtin_cptr_type(fir::unwrapRefType(it->getType()))) {
+      clauseOps.useDeviceAddrVars.push_back(*it);
+      // We have to shuffle the symbols around as well, to maintain
+      // the correct Input -> BlockArg for use_device_ptr/use_device_addr.
+      // NOTE: However, as map's do not seem to be included currently
+      // this isn't as pertinent, but we must try to maintain for
+      // future alterations. I believe the reason they are not currently
+      // is that the BlockArg assign/lowering needs to be extended
+      // to a greater set of types.
+      auto idx = std::distance(clauseOps.useDevicePtrVars.begin(), it);
+      moveElementToBack(idx, useDeviceTypes);
+      moveElementToBack(idx, useDeviceLocs);
+      moveElementToBack(idx, useDeviceSymbols);
+      it = clauseOps.useDevicePtrVars.erase(it);
+      continue;
+    }
+    ++it;
+  }
+}
+
+/// Extract the list of function and variable symbols affected by the given
+/// 'declare target' directive and return the intended device type for them.
+static void getDeclareTargetInfo(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
+    mlir::omp::DeclareTargetClauseOps &clauseOps,
+    llvm::SmallVectorImpl<DeclareTargetCapturePair> &symbolAndClause) {
+  const auto &spec = std::get<Fortran::parser::OmpDeclareTargetSpecifier>(
+      declareTargetConstruct.t);
+  if (const auto *objectList{
+          Fortran::parser::Unwrap<Fortran::parser::OmpObjectList>(spec.u)}) {
+    ObjectList objects{makeObjects(*objectList, semaCtx)};
+    // Case: declare target(func, var1, var2)
+    gatherFuncAndVarSyms(objects, mlir::omp::DeclareTargetCaptureClause::to,
+                         symbolAndClause);
+  } else if (const auto *clauseList{
+                 Fortran::parser::Unwrap<Fortran::parser::OmpClauseList>(
+                     spec.u)}) {
+    if (clauseList->v.empty()) {
+      // Case: declare target, implicit capture of function
+      symbolAndClause.emplace_back(
+          mlir::omp::DeclareTargetCaptureClause::to,
+          eval.getOwningProcedure()->getSubprogramSymbol());
+    }
+
+    ClauseProcessor cp(converter, semaCtx, *clauseList);
+    cp.processDeviceType(clauseOps);
+    cp.processEnter(symbolAndClause);
+    cp.processLink(symbolAndClause);
+    cp.processTo(symbolAndClause);
+
+    cp.processTODO<clause::Indirect>(converter.getCurrentLocation(),
+                                     llvm::omp::Directive::OMPD_declare_target);
+  }
+}
+
+static void collectDeferredDeclareTargets(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
+    llvm::SmallVectorImpl<Fortran::lower::OMPDeferredDeclareTargetInfo>
+        &deferredDeclareTarget) {
+  mlir::omp::DeclareTargetClauseOps clauseOps;
+  llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
+  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
+                       clauseOps, symbolAndClause);
+  // Return the device type only if at least one of the targets for the
+  // directive is a function or subroutine
+  mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
+
+  for (const DeclareTargetCapturePair &symClause : symbolAndClause) {
+    mlir::Operation *op = mod.lookupSymbol(converter.mangleName(
+        std::get<const Fortran::semantics::Symbol &>(symClause)));
+
+    if (!op) {
+      deferredDeclareTarget.push_back({std::get<0>(symClause),
+                                       clauseOps.deviceType,
+                                       std::get<1>(symClause)});
+    }
+  }
+}
+
+static std::optional<mlir::omp::DeclareTargetDeviceType>
+getDeclareTargetFunctionDevice(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OpenMPDeclareTargetConstruct
+        &declareTargetConstruct) {
+  mlir::omp::DeclareTargetClauseOps clauseOps;
+  llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
+  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
+                       clauseOps, symbolAndClause);
+
+  // Return the device type only if at least one of the targets for the
+  // directive is a function or subroutine
+  mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
+  for (const DeclareTargetCapturePair &symClause : symbolAndClause) {
+    mlir::Operation *op = mod.lookupSymbol(converter.mangleName(
+        std::get<const Fortran::semantics::Symbol &>(symClause)));
+
+    if (mlir::isa_and_nonnull<mlir::func::FuncOp>(op))
+      return clauseOps.deviceType;
+  }
+
+  return std::nullopt;
+}
+
+static llvm::SmallVector<const Fortran::semantics::Symbol *>
+genLoopVars(mlir::Operation *op, Fortran::lower::AbstractConverter &converter,
+            mlir::Location &loc,
+            llvm::ArrayRef<const Fortran::semantics::Symbol *> args) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  auto &region = op->getRegion(0);
+
+  std::size_t loopVarTypeSize = 0;
+  for (const Fortran::semantics::Symbol *arg : args)
+    loopVarTypeSize = std::max(loopVarTypeSize, arg->GetUltimate().size());
+  mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize);
+  llvm::SmallVector<mlir::Type> tiv(args.size(), loopVarType);
+  llvm::SmallVector<mlir::Location> locs(args.size(), loc);
+  firOpBuilder.createBlock(&region, {}, tiv, locs);
+  // The argument is not currently in memory, so make a temporary for the
+  // argument, and store it there, then bind that location to the argument.
+  mlir::Operation *storeOp = nullptr;
+  for (auto [argIndex, argSymbol] : llvm::enumerate(args)) {
+    mlir::Value indexVal = fir::getBase(region.front().getArgument(argIndex));
+    storeOp =
+        createAndSetPrivatizedLoopVar(converter, loc, indexVal, argSymbol);
+  }
+  firOpBuilder.setInsertionPointAfter(storeOp);
+  return llvm::SmallVector<const Fortran::semantics::Symbol *>(args);
+}
+
+static void genReductionVars(
+    mlir::Operation *op, Fortran::lower::AbstractConverter &converter,
+    mlir::Location &loc,
+    llvm::ArrayRef<const Fortran::semantics::Symbol *> reductionArgs,
+    llvm::ArrayRef<mlir::Type> reductionTypes) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  llvm::SmallVector<mlir::Location> blockArgLocs(reductionArgs.size(), loc);
+
+  mlir::Block *entryBlock = firOpBuilder.createBlock(
+      &op->getRegion(0), {}, reductionTypes, blockArgLocs);
+
+  // Bind the reduction arguments to their block arguments.
+  for (auto [arg, prv] :
+       llvm::zip_equal(reductionArgs, entryBlock->getArguments())) {
+    converter.bindSymbol(*arg, prv);
+  }
+}
+
+static llvm::SmallVector<const Fortran::semantics::Symbol *>
+genLoopAndReductionVars(
+    mlir::Operation *op, Fortran::lower::AbstractConverter &converter,
+    mlir::Location &loc,
+    llvm::ArrayRef<const Fortran::semantics::Symbol *> loopArgs,
+    llvm::ArrayRef<const Fortran::semantics::Symbol *> reductionArgs,
+    llvm::ArrayRef<mlir::Type> reductionTypes) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+
+  llvm::SmallVector<mlir::Type> blockArgTypes;
+  llvm::SmallVector<mlir::Location> blockArgLocs;
+  blockArgTypes.reserve(loopArgs.size() + reductionArgs.size());
+  blockArgLocs.reserve(blockArgTypes.size());
+  mlir::Block *entryBlock;
+
+  if (loopArgs.size()) {
+    std::size_t loopVarTypeSize = 0;
+    for (const Fortran::semantics::Symbol *arg : loopArgs)
+      loopVarTypeSize = std::max(loopVarTypeSize, arg->GetUltimate().size());
+    mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize);
+    std::fill_n(std::back_inserter(blockArgTypes), loopArgs.size(),
+                loopVarType);
+    std::fill_n(std::back_inserter(blockArgLocs), loopArgs.size(), loc);
+  }
+  if (reductionArgs.size()) {
+    llvm::copy(reductionTypes, std::back_inserter(blockArgTypes));
+    std::fill_n(std::back_inserter(blockArgLocs), reductionArgs.size(), loc);
+  }
+  entryBlock = firOpBuilder.createBlock(&op->getRegion(0), {}, blockArgTypes,
+                                        blockArgLocs);
+  // The argument is not currently in memory, so make a temporary for the
+  // argument, and store it there, then bind that location to the argument.
+  if (loopArgs.size()) {
+    mlir::Operation *storeOp = nullptr;
+    for (auto [argIndex, argSymbol] : llvm::enumerate(loopArgs)) {
+      mlir::Value indexVal =
+          fir::getBase(op->getRegion(0).front().getArgument(argIndex));
+      storeOp =
+          createAndSetPrivatizedLoopVar(converter, loc, indexVal, argSymbol);
+    }
+    firOpBuilder.setInsertionPointAfter(storeOp);
+  }
+  // Bind the reduction arguments to their block arguments
+  for (auto [arg, prv] : llvm::zip_equal(
+           reductionArgs,
+           llvm::drop_begin(entryBlock->getArguments(), loopArgs.size()))) {
+    converter.bindSymbol(*arg, prv);
+  }
+
+  return llvm::SmallVector<const Fortran::semantics::Symbol *>(loopArgs);
+}
+
+static void
+markDeclareTarget(mlir::Operation *op,
+                  Fortran::lower::AbstractConverter &converter,
+                  mlir::omp::DeclareTargetCaptureClause captureClause,
+                  mlir::omp::DeclareTargetDeviceType deviceType) {
+  // TODO: Add support for program local variables with declare target applied
+  auto declareTargetOp = llvm::dyn_cast<mlir::omp::DeclareTargetInterface>(op);
+  if (!declareTargetOp)
+    fir::emitFatalError(
+        converter.getCurrentLocation(),
+        "Attempt to apply declare target on unsupported operation");
+
+  // The function or global already has a declare target applied to it, very
+  // likely through implicit capture (usage in another declare target
+  // function/subroutine). It should be marked as any if it has been assigned
+  // both host and nohost, else we skip, as there is no change
+  if (declareTargetOp.isDeclareTarget()) {
+    if (declareTargetOp.getDeclareTargetDeviceType() != deviceType)
+      declareTargetOp.setDeclareTarget(mlir::omp::DeclareTargetDeviceType::any,
+                                       captureClause);
+    return;
+  }
+
+  declareTargetOp.setDeclareTarget(deviceType, captureClause);
+}
+
+//===----------------------------------------------------------------------===//
+// Op body generation helper structures and functions
+//===----------------------------------------------------------------------===//
+
 struct OpWithBodyGenInfo {
   /// A type for a code-gen callback function. This takes as argument the op for
   /// which the code is being generated and returns the arguments of the op's
@@ -493,548 +763,737 @@ static void genBodyOfTargetDataOp(
     genNestedEvaluations(converter, eval);
 }
 
-template <typename OpTy, typename... Args>
-static OpTy genOpWithBody(OpWithBodyGenInfo &info, Args &&...args) {
-  auto op = info.converter.getFirOpBuilder().create<OpTy>(
-      info.loc, std::forward<Args>(args)...);
-  createBodyOfOp<OpTy>(op, info);
-  return op;
-}
-
-static mlir::omp::MasterOp
-genMasterOp(Fortran::lower::AbstractConverter &converter,
-            Fortran::semantics::SemanticsContext &semaCtx,
-            Fortran::lower::pft::Evaluation &eval, bool genNested,
-            mlir::Location currentLocation) {
-  return genOpWithBody<mlir::omp::MasterOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested));
-}
-
-static mlir::omp::OrderedRegionOp
-genOrderedRegionOp(Fortran::lower::AbstractConverter &converter,
-                   Fortran::semantics::SemanticsContext &semaCtx,
-                   Fortran::lower::pft::Evaluation &eval, bool genNested,
-                   mlir::Location currentLocation,
-                   const Fortran::parser::OmpClauseList &clauseList) {
-  mlir::omp::OrderedRegionClauseOps clauseOps;
-
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processTODO<clause::Simd>(currentLocation,
-                               llvm::omp::Directive::OMPD_ordered);
-
-  return genOpWithBody<mlir::omp::OrderedRegionOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested),
-      clauseOps);
-}
+// This functions creates a block for the body of the targetOp's region. It adds
+// all the symbols present in mapSymbols as block arguments to this block.
+static void
+genBodyOfTargetOp(Fortran::lower::AbstractConverter &converter,
+                  Fortran::semantics::SemanticsContext &semaCtx,
+                  Fortran::lower::pft::Evaluation &eval, bool genNested,
+                  mlir::omp::TargetOp &targetOp,
+                  llvm::ArrayRef<const Fortran::semantics::Symbol *> mapSyms,
+                  llvm::ArrayRef<mlir::Location> mapSymLocs,
+                  llvm::ArrayRef<mlir::Type> mapSymTypes,
+                  const mlir::Location &currentLocation) {
+  assert(mapSymTypes.size() == mapSymLocs.size());
 
-static mlir::omp::ParallelOp
-genParallelOp(Fortran::lower::AbstractConverter &converter,
-              Fortran::lower::SymMap &symTable,
-              Fortran::semantics::SemanticsContext &semaCtx,
-              Fortran::lower::pft::Evaluation &eval, bool genNested,
-              mlir::Location currentLocation,
-              const Fortran::parser::OmpClauseList &clauseList,
-              bool outerCombined = false) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::ParallelClauseOps clauseOps;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> privateSyms;
-  llvm::SmallVector<mlir::Type> reductionTypes;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
-
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_parallel, clauseOps);
-  cp.processNumThreads(stmtCtx, clauseOps);
-  cp.processProcBind(clauseOps);
-  cp.processDefault();
-  cp.processAllocate(clauseOps);
-
-  if (!outerCombined)
-    cp.processReduction(currentLocation, clauseOps, &reductionTypes,
-                        &reductionSyms);
+  mlir::Region &region = targetOp.getRegion();
 
-  if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
-    clauseOps.reductionByRefAttr = firOpBuilder.getUnitAttr();
+  auto *regionBlock =
+      firOpBuilder.createBlock(&region, {}, mapSymTypes, mapSymLocs);
 
-  auto reductionCallback = [&](mlir::Operation *op) {
-    llvm::SmallVector<mlir::Location> locs(clauseOps.reductionVars.size(),
-                                           currentLocation);
-    auto *block =
-        firOpBuilder.createBlock(&op->getRegion(0), {}, reductionTypes, locs);
-    for (auto [arg, prv] :
-         llvm::zip_equal(reductionSyms, block->getArguments())) {
-      converter.bindSymbol(*arg, prv);
+  // Clones the `bounds` placing them inside the target region and returns them.
+  auto cloneBound = [&](mlir::Value bound) {
+    if (mlir::isMemoryEffectFree(bound.getDefiningOp())) {
+      mlir::Operation *clonedOp = bound.getDefiningOp()->clone();
+      regionBlock->push_back(clonedOp);
+      return clonedOp->getResult(0);
     }
-    return reductionSyms;
+    TODO(converter.getCurrentLocation(),
+         "target map clause operand unsupported bound type");
   };
 
-  OpWithBodyGenInfo genInfo =
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setOuterCombined(outerCombined)
-          .setClauses(&clauseList)
-          .setReductions(&reductionSyms, &reductionTypes)
-          .setGenRegionEntryCb(reductionCallback);
+  auto cloneBounds = [cloneBound](llvm::ArrayRef<mlir::Value> bounds) {
+    llvm::SmallVector<mlir::Value> clonedBounds;
+    for (mlir::Value bound : bounds)
+      clonedBounds.emplace_back(cloneBound(bound));
+    return clonedBounds;
+  };
 
-  if (!enableDelayedPrivatization)
-    return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
+  // Bind the symbols to their corresponding block arguments.
+  for (auto [argIndex, argSymbol] : llvm::enumerate(mapSyms)) {
+    const mlir::BlockArgument &arg = region.getArgument(argIndex);
+    // Avoid capture of a reference to a structured binding.
+    const Fortran::semantics::Symbol *sym = argSymbol;
+    // Structure component symbols don't have bindings.
+    if (sym->owner().IsDerivedType())
+      continue;
+    fir::ExtendedValue extVal = converter.getSymbolExtendedValue(*sym);
+    extVal.match(
+        [&](const fir::BoxValue &v) {
+          converter.bindSymbol(*sym,
+                               fir::BoxValue(arg, cloneBounds(v.getLBounds()),
+                                             v.getExplicitParameters(),
+                                             v.getExplicitExtents()));
+        },
+        [&](const fir::MutableBoxValue &v) {
+          converter.bindSymbol(
+              *sym, fir::MutableBoxValue(arg, cloneBounds(v.getLBounds()),
+                                         v.getMutableProperties()));
+        },
+        [&](const fir::ArrayBoxValue &v) {
+          converter.bindSymbol(
+              *sym, fir::ArrayBoxValue(arg, cloneBounds(v.getExtents()),
+                                       cloneBounds(v.getLBounds()),
+                                       v.getSourceBox()));
+        },
+        [&](const fir::CharArrayBoxValue &v) {
+          converter.bindSymbol(
+              *sym, fir::CharArrayBoxValue(arg, cloneBound(v.getLen()),
+                                           cloneBounds(v.getExtents()),
+                                           cloneBounds(v.getLBounds())));
+        },
+        [&](const fir::CharBoxValue &v) {
+          converter.bindSymbol(*sym,
+                               fir::CharBoxValue(arg, cloneBound(v.getLen())));
+        },
+        [&](const fir::UnboxedValue &v) { converter.bindSymbol(*sym, arg); },
+        [&](const auto &) {
+          TODO(converter.getCurrentLocation(),
+               "target map clause operand unsupported type");
+        });
+  }
 
-  bool privatize = !outerCombined;
-  DataSharingProcessor dsp(converter, semaCtx, clauseList, eval,
-                           /*useDelayedPrivatization=*/true, &symTable);
+  // Check if cloning the bounds introduced any dependency on the outer region.
+  // If so, then either clone them as well if they are MemoryEffectFree, or else
+  // copy them to a new temporary and add them to the map and block_argument
+  // lists and replace their uses with the new temporary.
+  llvm::SetVector<mlir::Value> valuesDefinedAbove;
+  mlir::getUsedValuesDefinedAbove(region, valuesDefinedAbove);
+  while (!valuesDefinedAbove.empty()) {
+    for (mlir::Value val : valuesDefinedAbove) {
+      mlir::Operation *valOp = val.getDefiningOp();
+      if (mlir::isMemoryEffectFree(valOp)) {
+        mlir::Operation *clonedOp = valOp->clone();
+        regionBlock->push_front(clonedOp);
+        val.replaceUsesWithIf(
+            clonedOp->getResult(0), [regionBlock](mlir::OpOperand &use) {
+              return use.getOwner()->getBlock() == regionBlock;
+            });
+      } else {
+        auto savedIP = firOpBuilder.getInsertionPoint();
+        firOpBuilder.setInsertionPointAfter(valOp);
+        auto copyVal =
+            firOpBuilder.createTemporary(val.getLoc(), val.getType());
+        firOpBuilder.createStoreWithConvert(copyVal.getLoc(), val, copyVal);
 
-  if (privatize)
-    dsp.processStep1(&clauseOps, &privateSyms);
+        llvm::SmallVector<mlir::Value> bounds;
+        std::stringstream name;
+        firOpBuilder.setInsertionPoint(targetOp);
+        mlir::Value mapOp = createMapInfoOp(
+            firOpBuilder, copyVal.getLoc(), copyVal, mlir::Value{}, name.str(),
+            bounds, llvm::SmallVector<mlir::Value>{},
+            static_cast<
+                std::underlying_type_t<llvm::omp::OpenMPOffloadMappingFlags>>(
+                llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_IMPLICIT),
+            mlir::omp::VariableCaptureKind::ByCopy, copyVal.getType());
+        targetOp.getMapOperandsMutable().append(mapOp);
+        mlir::Value clonedValArg =
+            region.addArgument(copyVal.getType(), copyVal.getLoc());
+        firOpBuilder.setInsertionPointToStart(regionBlock);
+        auto loadOp = firOpBuilder.create<fir::LoadOp>(clonedValArg.getLoc(),
+                                                       clonedValArg);
+        val.replaceUsesWithIf(
+            loadOp->getResult(0), [regionBlock](mlir::OpOperand &use) {
+              return use.getOwner()->getBlock() == regionBlock;
+            });
+        firOpBuilder.setInsertionPoint(regionBlock, savedIP);
+      }
+    }
+    valuesDefinedAbove.clear();
+    mlir::getUsedValuesDefinedAbove(region, valuesDefinedAbove);
+  }
 
-  auto genRegionEntryCB = [&](mlir::Operation *op) {
-    auto parallelOp = llvm::cast<mlir::omp::ParallelOp>(op);
+  // Insert dummy instruction to remember the insertion position. The
+  // marker will be deleted since there are not uses.
+  // In the HLFIR flow there are hlfir.declares inserted above while
+  // setting block arguments.
+  mlir::Value undefMarker = firOpBuilder.create<fir::UndefOp>(
+      targetOp.getOperation()->getLoc(), firOpBuilder.getIndexType());
 
-    llvm::SmallVector<mlir::Location> reductionLocs(
-        clauseOps.reductionVars.size(), currentLocation);
+  // Create blocks for unstructured regions. This has to be done since
+  // blocks are initially allocated with the function as the parent region.
+  if (eval.lowerAsUnstructured()) {
+    Fortran::lower::createEmptyRegionBlocks<mlir::omp::TerminatorOp,
+                                            mlir::omp::YieldOp>(
+        firOpBuilder, eval.getNestedEvaluations());
+  }
 
-    mlir::OperandRange privateVars = parallelOp.getPrivateVars();
-    mlir::Region &region = parallelOp.getRegion();
+  firOpBuilder.create<mlir::omp::TerminatorOp>(currentLocation);
 
-    llvm::SmallVector<mlir::Type> privateVarTypes = reductionTypes;
-    privateVarTypes.reserve(privateVarTypes.size() + privateVars.size());
-    llvm::transform(privateVars, std::back_inserter(privateVarTypes),
-                    [](mlir::Value v) { return v.getType(); });
+  // Create the insertion point after the marker.
+  firOpBuilder.setInsertionPointAfter(undefMarker.getDefiningOp());
+  if (genNested)
+    genNestedEvaluations(converter, eval);
+}
 
-    llvm::SmallVector<mlir::Location> privateVarLocs = reductionLocs;
-    privateVarLocs.reserve(privateVarLocs.size() + privateVars.size());
-    llvm::transform(privateVars, std::back_inserter(privateVarLocs),
-                    [](mlir::Value v) { return v.getLoc(); });
+template <typename OpTy, typename... Args>
+static OpTy genOpWithBody(OpWithBodyGenInfo &info, Args &&...args) {
+  auto op = info.converter.getFirOpBuilder().create<OpTy>(
+      info.loc, std::forward<Args>(args)...);
+  createBodyOfOp<OpTy>(op, info);
+  return op;
+}
 
-    firOpBuilder.createBlock(&region, /*insertPt=*/{}, privateVarTypes,
-                             privateVarLocs);
+//===----------------------------------------------------------------------===//
+// Code generation functions for clauses
+//===----------------------------------------------------------------------===//
 
-    llvm::SmallVector<const Fortran::semantics::Symbol *> allSymbols =
-        reductionSyms;
-    allSymbols.append(privateSyms);
-    for (auto [arg, prv] : llvm::zip_equal(allSymbols, region.getArguments())) {
-      converter.bindSymbol(*arg, prv);
-    }
+static void genCriticalDeclareClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    mlir::omp::CriticalClauseOps &clauseOps, llvm::StringRef name) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processHint(clauseOps);
+  clauseOps.nameAttr =
+      mlir::StringAttr::get(converter.getFirOpBuilder().getContext(), name);
+}
 
-    return allSymbols;
-  };
+static void genFlushClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    const std::optional<Fortran::parser::OmpObjectList> &objects,
+    const std::optional<std::list<Fortran::parser::OmpMemoryOrderClause>>
+        &clauses,
+    mlir::Location loc, llvm::SmallVectorImpl<mlir::Value> &operandRange) {
+  if (objects)
+    genObjectList2(*objects, converter, operandRange);
+
+  if (clauses && clauses->size() > 0)
+    TODO(converter.getCurrentLocation(), "Handle OmpMemoryOrderClause");
+}
 
-  // TODO Merge with the reduction CB.
-  genInfo.setGenRegionEntryCb(genRegionEntryCB).setDataSharingProcessor(&dsp);
-  return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
+static void
+genOrderedRegionClauses(Fortran::lower::AbstractConverter &converter,
+                        Fortran::semantics::SemanticsContext &semaCtx,
+                        const Fortran::parser::OmpClauseList &clauses,
+                        mlir::Location loc,
+                        mlir::omp::OrderedRegionClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processTODO<clause::Simd>(loc, llvm::omp::Directive::OMPD_ordered);
+}
+
+static void genParallelClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    bool processReduction, mlir::omp::ParallelClauseOps &clauseOps,
+    llvm::SmallVectorImpl<mlir::Type> &reductionTypes,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &reductionSyms) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processAllocate(clauseOps);
+  cp.processDefault();
+  cp.processIf(llvm::omp::Directive::OMPD_parallel, clauseOps);
+  cp.processNumThreads(stmtCtx, clauseOps);
+  cp.processProcBind(clauseOps);
+
+  if (processReduction) {
+    cp.processReduction(loc, clauseOps, &reductionTypes, &reductionSyms);
+    if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
+      clauseOps.reductionByRefAttr = converter.getFirOpBuilder().getUnitAttr();
+  }
 }
 
-static mlir::omp::SectionOp
-genSectionOp(Fortran::lower::AbstractConverter &converter,
-             Fortran::semantics::SemanticsContext &semaCtx,
-             Fortran::lower::pft::Evaluation &eval, bool genNested,
-             mlir::Location currentLocation,
-             const Fortran::parser::OmpClauseList &sectionsClauseList) {
-  // Currently only private/firstprivate clause is handled, and
-  // all privatization is done within `omp.section` operations.
-  return genOpWithBody<mlir::omp::SectionOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setClauses(&sectionsClauseList));
+static void genSectionsClauses(Fortran::lower::AbstractConverter &converter,
+                               Fortran::semantics::SemanticsContext &semaCtx,
+                               const Fortran::parser::OmpClauseList &clauses,
+                               mlir::Location loc,
+                               bool clausesFromBeginSections,
+                               mlir::omp::SectionsClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  if (clausesFromBeginSections) {
+    cp.processAllocate(clauseOps);
+    cp.processSectionsReduction(loc, clauseOps);
+    // TODO Support delayed privatization.
+  } else {
+    cp.processNowait(clauseOps);
+  }
 }
 
-static mlir::omp::SingleOp
-genSingleOp(Fortran::lower::AbstractConverter &converter,
-            Fortran::semantics::SemanticsContext &semaCtx,
-            Fortran::lower::pft::Evaluation &eval, bool genNested,
-            mlir::Location currentLocation,
-            const Fortran::parser::OmpClauseList &beginClauseList,
-            const Fortran::parser::OmpClauseList &endClauseList) {
-  mlir::omp::SingleClauseOps clauseOps;
+static void genSimdLoopClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    mlir::omp::SimdLoopClauseOps &clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processCollapse(loc, eval, clauseOps, iv);
+  cp.processIf(llvm::omp::Directive::OMPD_simd, clauseOps);
+  cp.processReduction(loc, clauseOps);
+  cp.processSafelen(clauseOps);
+  cp.processSimdlen(clauseOps);
+  clauseOps.loopInclusiveAttr = converter.getFirOpBuilder().getUnitAttr();
+  // TODO Support delayed privatization.
 
-  ClauseProcessor cp(converter, semaCtx, beginClauseList);
-  cp.processAllocate(clauseOps);
+  cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
+                 clause::Nontemporal, clause::Order>(
+      loc, llvm::omp::Directive::OMPD_simd);
+}
+
+static void genSingleClauses(Fortran::lower::AbstractConverter &converter,
+                             Fortran::semantics::SemanticsContext &semaCtx,
+                             const Fortran::parser::OmpClauseList &beginClauses,
+                             const Fortran::parser::OmpClauseList &endClauses,
+                             mlir::Location loc,
+                             mlir::omp::SingleClauseOps &clauseOps) {
+  ClauseProcessor bcp(converter, semaCtx, beginClauses);
+  bcp.processAllocate(clauseOps);
   // TODO Support delayed privatization.
 
-  ClauseProcessor ecp(converter, semaCtx, endClauseList);
+  ClauseProcessor ecp(converter, semaCtx, endClauses);
+  ecp.processCopyprivate(loc, clauseOps);
   ecp.processNowait(clauseOps);
-  ecp.processCopyprivate(currentLocation, clauseOps);
+}
 
-  return genOpWithBody<mlir::omp::SingleOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setClauses(&beginClauseList),
-      clauseOps);
+static void genTargetClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    bool processHostOnlyClauses, bool processReduction,
+    mlir::omp::TargetClauseOps &clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &mapSyms,
+    llvm::SmallVectorImpl<mlir::Location> &mapLocs,
+    llvm::SmallVectorImpl<mlir::Type> &mapTypes,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &deviceAddrSyms,
+    llvm::SmallVectorImpl<mlir::Location> &deviceAddrLocs,
+    llvm::SmallVectorImpl<mlir::Type> &deviceAddrTypes,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &devicePtrSyms,
+    llvm::SmallVectorImpl<mlir::Location> &devicePtrLocs,
+    llvm::SmallVectorImpl<mlir::Type> &devicePtrTypes) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processDepend(clauseOps);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processHasDeviceAddr(clauseOps, deviceAddrTypes, deviceAddrLocs,
+                          deviceAddrSyms);
+  cp.processIf(llvm::omp::Directive::OMPD_target, clauseOps);
+  cp.processIsDevicePtr(clauseOps, devicePtrTypes, devicePtrLocs,
+                        devicePtrSyms);
+  cp.processMap(loc, stmtCtx, clauseOps, &mapSyms, &mapLocs, &mapTypes);
+  cp.processThreadLimit(stmtCtx, clauseOps);
+  // TODO Support delayed privatization.
+
+  if (processHostOnlyClauses)
+    cp.processNowait(clauseOps);
+
+  cp.processTODO<clause::Allocate, clause::Defaultmap, clause::Firstprivate,
+                 clause::InReduction, clause::Private, clause::Reduction,
+                 clause::UsesAllocators>(loc,
+                                         llvm::omp::Directive::OMPD_target);
 }
 
-static mlir::omp::TaskOp
-genTaskOp(Fortran::lower::AbstractConverter &converter,
-          Fortran::semantics::SemanticsContext &semaCtx,
-          Fortran::lower::pft::Evaluation &eval, bool genNested,
-          mlir::Location currentLocation,
-          const Fortran::parser::OmpClauseList &clauseList) {
-  Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::TaskClauseOps clauseOps;
+static void genTargetDataClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    mlir::omp::TargetDataClauseOps &clauseOps,
+    llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
+    llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &useDeviceSyms) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processIf(llvm::omp::Directive::OMPD_target_data, clauseOps);
+  cp.processMap(loc, stmtCtx, clauseOps);
+  cp.processUseDeviceAddr(clauseOps, useDeviceTypes, useDeviceLocs,
+                          useDeviceSyms);
+  cp.processUseDevicePtr(clauseOps, useDeviceTypes, useDeviceLocs,
+                         useDeviceSyms);
 
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_task, clauseOps);
+  // This function implements the deprecated functionality of use_device_ptr
+  // that allows users to provide non-CPTR arguments to it with the caveat
+  // that the compiler will treat them as use_device_addr. A lot of legacy
+  // code may still depend on this functionality, so we should support it
+  // in some manner. We do so currently by simply shifting non-cptr operands
+  // from the use_device_ptr list into the front of the use_device_addr list
+  // whilst maintaining the ordering of useDeviceLocs, useDeviceSyms and
+  // useDeviceTypes to use_device_ptr/use_device_addr input for BlockArg
+  // ordering.
+  // TODO: Perhaps create a user provideable compiler option that will
+  // re-introduce a hard-error rather than a warning in these cases.
+  promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(clauseOps, useDeviceTypes,
+                                                useDeviceLocs, useDeviceSyms);
+}
+
+static void genTargetEnterExitUpdateDataClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    llvm::omp::Directive directive,
+    mlir::omp::TargetEnterExitUpdateDataClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processDepend(clauseOps);
+  cp.processDevice(stmtCtx, clauseOps);
+  cp.processIf(directive, clauseOps);
+  cp.processNowait(clauseOps);
+
+  if (directive == llvm::omp::Directive::OMPD_target_update) {
+    cp.processMotionClauses<clause::To>(stmtCtx, clauseOps);
+    cp.processMotionClauses<clause::From>(stmtCtx, clauseOps);
+  } else {
+    cp.processMap(loc, stmtCtx, clauseOps);
+  }
+}
+
+static void genTaskClauses(Fortran::lower::AbstractConverter &converter,
+                           Fortran::semantics::SemanticsContext &semaCtx,
+                           Fortran::lower::StatementContext &stmtCtx,
+                           const Fortran::parser::OmpClauseList &clauses,
+                           mlir::Location loc,
+                           mlir::omp::TaskClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
   cp.processAllocate(clauseOps);
   cp.processDefault();
+  cp.processDepend(clauseOps);
   cp.processFinal(stmtCtx, clauseOps);
-  cp.processUntied(clauseOps);
+  cp.processIf(llvm::omp::Directive::OMPD_task, clauseOps);
   cp.processMergeable(clauseOps);
   cp.processPriority(stmtCtx, clauseOps);
-  cp.processDepend(clauseOps);
+  cp.processUntied(clauseOps);
   // TODO Support delayed privatization.
 
-  cp.processTODO<clause::InReduction, clause::Detach, clause::Affinity>(
-      currentLocation, llvm::omp::Directive::OMPD_task);
+  cp.processTODO<clause::Affinity, clause::Detach, clause::InReduction>(
+      loc, llvm::omp::Directive::OMPD_task);
+}
 
-  return genOpWithBody<mlir::omp::TaskOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setClauses(&clauseList),
-      clauseOps);
+static void genTaskgroupClauses(Fortran::lower::AbstractConverter &converter,
+                                Fortran::semantics::SemanticsContext &semaCtx,
+                                const Fortran::parser::OmpClauseList &clauses,
+                                mlir::Location loc,
+                                mlir::omp::TaskgroupClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processAllocate(clauseOps);
+
+  cp.processTODO<clause::TaskReduction>(loc,
+                                        llvm::omp::Directive::OMPD_taskgroup);
 }
 
-static mlir::omp::TaskgroupOp
-genTaskgroupOp(Fortran::lower::AbstractConverter &converter,
-               Fortran::semantics::SemanticsContext &semaCtx,
-               Fortran::lower::pft::Evaluation &eval, bool genNested,
-               mlir::Location currentLocation,
-               const Fortran::parser::OmpClauseList &clauseList) {
-  mlir::omp::TaskgroupClauseOps clauseOps;
+static void genTaskwaitClauses(Fortran::lower::AbstractConverter &converter,
+                               Fortran::semantics::SemanticsContext &semaCtx,
+                               const Fortran::parser::OmpClauseList &clauses,
+                               mlir::Location loc,
+                               mlir::omp::TaskwaitClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processTODO<clause::Depend, clause::Nowait>(
+      loc, llvm::omp::Directive::OMPD_taskwait);
+}
 
-  ClauseProcessor cp(converter, semaCtx, clauseList);
+static void genTeamsClauses(Fortran::lower::AbstractConverter &converter,
+                            Fortran::semantics::SemanticsContext &semaCtx,
+                            Fortran::lower::StatementContext &stmtCtx,
+                            const Fortran::parser::OmpClauseList &clauses,
+                            mlir::Location loc,
+                            mlir::omp::TeamsClauseOps &clauseOps) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
   cp.processAllocate(clauseOps);
-  cp.processTODO<clause::TaskReduction>(currentLocation,
-                                        llvm::omp::Directive::OMPD_taskgroup);
+  cp.processDefault();
+  cp.processIf(llvm::omp::Directive::OMPD_teams, clauseOps);
+  cp.processNumTeams(stmtCtx, clauseOps);
+  cp.processThreadLimit(stmtCtx, clauseOps);
+  // TODO Support delayed privatization.
 
-  return genOpWithBody<mlir::omp::TaskgroupOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setClauses(&clauseList),
-      clauseOps);
+  cp.processTODO<clause::Reduction>(loc, llvm::omp::Directive::OMPD_teams);
 }
 
-// This helper function implements the functionality of "promoting"
-// non-CPTR arguments of use_device_ptr to use_device_addr
-// arguments (automagic conversion of use_device_ptr ->
-// use_device_addr in these cases). The way we do so currently is
-// through the shuffling of operands from the devicePtrOperands to
-// deviceAddrOperands where neccesary and re-organizing the types,
-// locations and symbols to maintain the correct ordering of ptr/addr
-// input -> BlockArg.
-//
-// This effectively implements some deprecated OpenMP functionality
-// that some legacy applications unfortunately depend on
-// (deprecated in specification version 5.2):
-//
-// "If a list item in a use_device_ptr clause is not of type C_PTR,
-//  the behavior is as if the list item appeared in a use_device_addr
-//  clause. Support for such list items in a use_device_ptr clause
-//  is deprecated."
-static void promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(
-    mlir::omp::UseDeviceClauseOps &clauseOps,
-    llvm::SmallVectorImpl<mlir::Type> &useDeviceTypes,
-    llvm::SmallVectorImpl<mlir::Location> &useDeviceLocs,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *>
-        &useDeviceSymbols) {
-  auto moveElementToBack = [](size_t idx, auto &vector) {
-    auto *iter = std::next(vector.begin(), idx);
-    vector.push_back(*iter);
-    vector.erase(iter);
-  };
+static void genWsloopClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::StatementContext &stmtCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OmpClauseList &beginClauses,
+    const Fortran::parser::OmpClauseList *endClauses, mlir::Location loc,
+    mlir::omp::WsloopClauseOps &clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv,
+    llvm::SmallVectorImpl<mlir::Type> &reductionTypes,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &reductionSyms) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  ClauseProcessor bcp(converter, semaCtx, beginClauses);
+  bcp.processCollapse(loc, eval, clauseOps, iv);
+  bcp.processOrdered(clauseOps);
+  bcp.processReduction(loc, clauseOps, &reductionTypes, &reductionSyms);
+  bcp.processSchedule(stmtCtx, clauseOps);
+  clauseOps.loopInclusiveAttr = firOpBuilder.getUnitAttr();
+  // TODO Support delayed privatization.
 
-  // Iterate over our use_device_ptr list and shift all non-cptr arguments into
-  // use_device_addr.
-  for (auto *it = clauseOps.useDevicePtrVars.begin();
-       it != clauseOps.useDevicePtrVars.end();) {
-    if (!fir::isa_builtin_cptr_type(fir::unwrapRefType(it->getType()))) {
-      clauseOps.useDeviceAddrVars.push_back(*it);
-      // We have to shuffle the symbols around as well, to maintain
-      // the correct Input -> BlockArg for use_device_ptr/use_device_addr.
-      // NOTE: However, as map's do not seem to be included currently
-      // this isn't as pertinent, but we must try to maintain for
-      // future alterations. I believe the reason they are not currently
-      // is that the BlockArg assign/lowering needs to be extended
-      // to a greater set of types.
-      auto idx = std::distance(clauseOps.useDevicePtrVars.begin(), it);
-      moveElementToBack(idx, useDeviceTypes);
-      moveElementToBack(idx, useDeviceLocs);
-      moveElementToBack(idx, useDeviceSymbols);
-      it = clauseOps.useDevicePtrVars.erase(it);
-      continue;
+  if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
+    clauseOps.reductionByRefAttr = firOpBuilder.getUnitAttr();
+
+  if (endClauses) {
+    ClauseProcessor ecp(converter, semaCtx, *endClauses);
+    ecp.processNowait(clauseOps);
+  }
+
+  bcp.processTODO<clause::Allocate, clause::Linear, clause::Order>(
+      loc, llvm::omp::Directive::OMPD_do);
+}
+
+//===----------------------------------------------------------------------===//
+// Code generation functions for leaf constructs
+//===----------------------------------------------------------------------===//
+
+static mlir::omp::BarrierOp
+genBarrierOp(Fortran::lower::AbstractConverter &converter,
+             Fortran::semantics::SemanticsContext &semaCtx,
+             Fortran::lower::pft::Evaluation &eval, mlir::Location loc) {
+  return converter.getFirOpBuilder().create<mlir::omp::BarrierOp>(loc);
+}
+
+static mlir::omp::CriticalOp
+genCriticalOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, bool genNested,
+              mlir::Location loc,
+              const Fortran::parser::OmpClauseList &clauseList,
+              const std::optional<Fortran::parser::Name> &name) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  mlir::FlatSymbolRefAttr nameAttr;
+
+  if (name) {
+    std::string nameStr = name->ToString();
+    mlir::ModuleOp mod = firOpBuilder.getModule();
+    auto global = mod.lookupSymbol<mlir::omp::CriticalDeclareOp>(nameStr);
+    if (!global) {
+      mlir::omp::CriticalClauseOps clauseOps;
+      genCriticalDeclareClauses(converter, semaCtx, clauseList, loc, clauseOps,
+                                nameStr);
+
+      mlir::OpBuilder modBuilder(mod.getBodyRegion());
+      global = modBuilder.create<mlir::omp::CriticalDeclareOp>(loc, clauseOps);
     }
-    ++it;
+    nameAttr = mlir::FlatSymbolRefAttr::get(firOpBuilder.getContext(),
+                                            global.getSymName());
   }
+
+  return genOpWithBody<mlir::omp::CriticalOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested),
+      nameAttr);
 }
 
-static mlir::omp::TargetDataOp
-genTargetDataOp(Fortran::lower::AbstractConverter &converter,
+static mlir::omp::DistributeOp
+genDistributeOp(Fortran::lower::AbstractConverter &converter,
                 Fortran::semantics::SemanticsContext &semaCtx,
                 Fortran::lower::pft::Evaluation &eval, bool genNested,
-                mlir::Location currentLocation,
+                mlir::Location loc,
                 const Fortran::parser::OmpClauseList &clauseList) {
-  Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::TargetDataClauseOps clauseOps;
-  llvm::SmallVector<mlir::Type> useDeviceTypes;
-  llvm::SmallVector<mlir::Location> useDeviceLocs;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> useDeviceSyms;
+  TODO(loc, "Distribute construct");
+  return nullptr;
+}
 
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_target_data, clauseOps);
-  cp.processDevice(stmtCtx, clauseOps);
-  cp.processUseDevicePtr(clauseOps, useDeviceTypes, useDeviceLocs,
-                         useDeviceSyms);
-  cp.processUseDeviceAddr(clauseOps, useDeviceTypes, useDeviceLocs,
-                          useDeviceSyms);
+static mlir::omp::FlushOp
+genFlushOp(Fortran::lower::AbstractConverter &converter,
+           Fortran::semantics::SemanticsContext &semaCtx,
+           Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+           const std::optional<Fortran::parser::OmpObjectList> &objectList,
+           const std::optional<std::list<Fortran::parser::OmpMemoryOrderClause>>
+               &clauseList) {
+  llvm::SmallVector<mlir::Value> operandRange;
+  genFlushClauses(converter, semaCtx, objectList, clauseList, loc,
+                  operandRange);
+
+  return converter.getFirOpBuilder().create<mlir::omp::FlushOp>(
+      converter.getCurrentLocation(), operandRange);
+}
 
-  // This function implements the deprecated functionality of use_device_ptr
-  // that allows users to provide non-CPTR arguments to it with the caveat
-  // that the compiler will treat them as use_device_addr. A lot of legacy
-  // code may still depend on this functionality, so we should support it
-  // in some manner. We do so currently by simply shifting non-cptr operands
-  // from the use_device_ptr list into the front of the use_device_addr list
-  // whilst maintaining the ordering of useDeviceLocs, useDeviceSymbols and
-  // useDeviceTypes to use_device_ptr/use_device_addr input for BlockArg
-  // ordering.
-  // TODO: Perhaps create a user provideable compiler option that will
-  // re-introduce a hard-error rather than a warning in these cases.
-  promoteNonCPtrUseDevicePtrArgsToUseDeviceAddr(clauseOps, useDeviceTypes,
-                                                useDeviceLocs, useDeviceSyms);
-  cp.processMap(currentLocation, llvm::omp::Directive::OMPD_target_data,
-                stmtCtx, clauseOps);
+static mlir::omp::MasterOp
+genMasterOp(Fortran::lower::AbstractConverter &converter,
+            Fortran::semantics::SemanticsContext &semaCtx,
+            Fortran::lower::pft::Evaluation &eval, bool genNested,
+            mlir::Location loc) {
+  return genOpWithBody<mlir::omp::MasterOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested));
+}
+
+static mlir::omp::OrderedOp
+genOrderedOp(Fortran::lower::AbstractConverter &converter,
+             Fortran::semantics::SemanticsContext &semaCtx,
+             Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+             const Fortran::parser::OmpClauseList &clauseList) {
+  TODO(loc, "OMPD_ordered");
+  return nullptr;
+}
+
+static mlir::omp::OrderedRegionOp
+genOrderedRegionOp(Fortran::lower::AbstractConverter &converter,
+                   Fortran::semantics::SemanticsContext &semaCtx,
+                   Fortran::lower::pft::Evaluation &eval, bool genNested,
+                   mlir::Location loc,
+                   const Fortran::parser::OmpClauseList &clauseList) {
+  mlir::omp::OrderedRegionClauseOps clauseOps;
+  genOrderedRegionClauses(converter, semaCtx, clauseList, loc, clauseOps);
+
+  return genOpWithBody<mlir::omp::OrderedRegionOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested),
+      clauseOps);
+}
+
+static mlir::omp::ParallelOp
+genParallelOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::lower::SymMap &symTable,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, bool genNested,
+              mlir::Location loc,
+              const Fortran::parser::OmpClauseList &clauseList,
+              bool outerCombined = false) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  Fortran::lower::StatementContext stmtCtx;
+  mlir::omp::ParallelClauseOps clauseOps;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> privateSyms;
+  llvm::SmallVector<mlir::Type> reductionTypes;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
+  genParallelClauses(converter, semaCtx, stmtCtx, clauseList, loc,
+                     /*processReduction=*/!outerCombined, clauseOps,
+                     reductionTypes, reductionSyms);
 
-  auto dataOp = converter.getFirOpBuilder().create<mlir::omp::TargetDataOp>(
-      currentLocation, clauseOps);
+  auto reductionCallback = [&](mlir::Operation *op) {
+    genReductionVars(op, converter, loc, reductionSyms, reductionTypes);
+    return reductionSyms;
+  };
 
-  genBodyOfTargetDataOp(converter, semaCtx, eval, genNested, dataOp,
-                        useDeviceTypes, useDeviceLocs, useDeviceSyms,
-                        currentLocation);
-  return dataOp;
-}
+  OpWithBodyGenInfo genInfo =
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setOuterCombined(outerCombined)
+          .setClauses(&clauseList)
+          .setReductions(&reductionSyms, &reductionTypes)
+          .setGenRegionEntryCb(reductionCallback);
 
-template <typename OpTy>
-static OpTy genTargetEnterExitDataUpdateOp(
-    Fortran::lower::AbstractConverter &converter,
-    Fortran::semantics::SemanticsContext &semaCtx,
-    mlir::Location currentLocation,
-    const Fortran::parser::OmpClauseList &clauseList) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::TargetEnterExitUpdateDataClauseOps clauseOps;
+  if (!enableDelayedPrivatization)
+    return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
 
-  // GCC 9.3.0 emits a (probably) bogus warning about an unused variable.
-  [[maybe_unused]] llvm::omp::Directive directive;
-  if constexpr (std::is_same_v<OpTy, mlir::omp::TargetEnterDataOp>) {
-    directive = llvm::omp::Directive::OMPD_target_enter_data;
-  } else if constexpr (std::is_same_v<OpTy, mlir::omp::TargetExitDataOp>) {
-    directive = llvm::omp::Directive::OMPD_target_exit_data;
-  } else if constexpr (std::is_same_v<OpTy, mlir::omp::TargetUpdateOp>) {
-    directive = llvm::omp::Directive::OMPD_target_update;
-  } else {
-    return nullptr;
-  }
+  bool privatize = !outerCombined;
+  DataSharingProcessor dsp(converter, semaCtx, clauseList, eval,
+                           /*useDelayedPrivatization=*/true, &symTable);
 
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(directive, clauseOps);
-  cp.processDevice(stmtCtx, clauseOps);
-  cp.processDepend(clauseOps);
-  cp.processNowait(clauseOps);
+  if (privatize)
+    dsp.processStep1(&clauseOps, &privateSyms);
 
-  if constexpr (std::is_same_v<OpTy, mlir::omp::TargetUpdateOp>) {
-    cp.processMotionClauses<clause::To>(stmtCtx, clauseOps);
-    cp.processMotionClauses<clause::From>(stmtCtx, clauseOps);
-  } else {
-    cp.processMap(currentLocation, directive, stmtCtx, clauseOps);
-  }
+  auto genRegionEntryCB = [&](mlir::Operation *op) {
+    auto parallelOp = llvm::cast<mlir::omp::ParallelOp>(op);
 
-  return firOpBuilder.create<OpTy>(currentLocation, clauseOps);
-}
+    llvm::SmallVector<mlir::Location> reductionLocs(
+        clauseOps.reductionVars.size(), loc);
 
-// This functions creates a block for the body of the targetOp's region. It adds
-// all the symbols present in mapSymbols as block arguments to this block.
-static void
-genBodyOfTargetOp(Fortran::lower::AbstractConverter &converter,
-                  Fortran::semantics::SemanticsContext &semaCtx,
-                  Fortran::lower::pft::Evaluation &eval, bool genNested,
-                  mlir::omp::TargetOp &targetOp,
-                  llvm::ArrayRef<const Fortran::semantics::Symbol *> mapSyms,
-                  llvm::ArrayRef<mlir::Location> mapSymLocs,
-                  llvm::ArrayRef<mlir::Type> mapSymTypes,
-                  const mlir::Location &currentLocation) {
-  assert(mapSymTypes.size() == mapSymLocs.size());
+    mlir::OperandRange privateVars = parallelOp.getPrivateVars();
+    mlir::Region &region = parallelOp.getRegion();
 
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  mlir::Region &region = targetOp.getRegion();
+    llvm::SmallVector<mlir::Type> privateVarTypes = reductionTypes;
+    privateVarTypes.reserve(privateVarTypes.size() + privateVars.size());
+    llvm::transform(privateVars, std::back_inserter(privateVarTypes),
+                    [](mlir::Value v) { return v.getType(); });
 
-  auto *regionBlock =
-      firOpBuilder.createBlock(&region, {}, mapSymTypes, mapSymLocs);
+    llvm::SmallVector<mlir::Location> privateVarLocs = reductionLocs;
+    privateVarLocs.reserve(privateVarLocs.size() + privateVars.size());
+    llvm::transform(privateVars, std::back_inserter(privateVarLocs),
+                    [](mlir::Value v) { return v.getLoc(); });
 
-  // Clones the `bounds` placing them inside the target region and returns them.
-  auto cloneBound = [&](mlir::Value bound) {
-    if (mlir::isMemoryEffectFree(bound.getDefiningOp())) {
-      mlir::Operation *clonedOp = bound.getDefiningOp()->clone();
-      regionBlock->push_back(clonedOp);
-      return clonedOp->getResult(0);
+    firOpBuilder.createBlock(&region, /*insertPt=*/{}, privateVarTypes,
+                             privateVarLocs);
+
+    llvm::SmallVector<const Fortran::semantics::Symbol *> allSymbols =
+        reductionSyms;
+    allSymbols.append(privateSyms);
+    for (auto [arg, prv] : llvm::zip_equal(allSymbols, region.getArguments())) {
+      converter.bindSymbol(*arg, prv);
     }
-    TODO(converter.getCurrentLocation(),
-         "target map clause operand unsupported bound type");
-  };
 
-  auto cloneBounds = [cloneBound](llvm::ArrayRef<mlir::Value> bounds) {
-    llvm::SmallVector<mlir::Value> clonedBounds;
-    for (mlir::Value bound : bounds)
-      clonedBounds.emplace_back(cloneBound(bound));
-    return clonedBounds;
+    return allSymbols;
   };
 
-  // Bind the symbols to their corresponding block arguments.
-  for (auto [argIndex, argSymbol] : llvm::enumerate(mapSyms)) {
-    const mlir::BlockArgument &arg = region.getArgument(argIndex);
-    // Avoid capture of a reference to a structured binding.
-    const Fortran::semantics::Symbol *sym = argSymbol;
-    // Structure component symbols don't have bindings.
-    if (sym->owner().IsDerivedType())
-      continue;
-    fir::ExtendedValue extVal = converter.getSymbolExtendedValue(*sym);
-    extVal.match(
-        [&](const fir::BoxValue &v) {
-          converter.bindSymbol(*sym,
-                               fir::BoxValue(arg, cloneBounds(v.getLBounds()),
-                                             v.getExplicitParameters(),
-                                             v.getExplicitExtents()));
-        },
-        [&](const fir::MutableBoxValue &v) {
-          converter.bindSymbol(
-              *sym, fir::MutableBoxValue(arg, cloneBounds(v.getLBounds()),
-                                         v.getMutableProperties()));
-        },
-        [&](const fir::ArrayBoxValue &v) {
-          converter.bindSymbol(
-              *sym, fir::ArrayBoxValue(arg, cloneBounds(v.getExtents()),
-                                       cloneBounds(v.getLBounds()),
-                                       v.getSourceBox()));
-        },
-        [&](const fir::CharArrayBoxValue &v) {
-          converter.bindSymbol(
-              *sym, fir::CharArrayBoxValue(arg, cloneBound(v.getLen()),
-                                           cloneBounds(v.getExtents()),
-                                           cloneBounds(v.getLBounds())));
-        },
-        [&](const fir::CharBoxValue &v) {
-          converter.bindSymbol(*sym,
-                               fir::CharBoxValue(arg, cloneBound(v.getLen())));
-        },
-        [&](const fir::UnboxedValue &v) { converter.bindSymbol(*sym, arg); },
-        [&](const auto &) {
-          TODO(converter.getCurrentLocation(),
-               "target map clause operand unsupported type");
-        });
-  }
+  // TODO Merge with the reduction CB.
+  genInfo.setGenRegionEntryCb(genRegionEntryCB).setDataSharingProcessor(&dsp);
+  return genOpWithBody<mlir::omp::ParallelOp>(genInfo, clauseOps);
+}
 
-  // Check if cloning the bounds introduced any dependency on the outer region.
-  // If so, then either clone them as well if they are MemoryEffectFree, or else
-  // copy them to a new temporary and add them to the map and block_argument
-  // lists and replace their uses with the new temporary.
-  llvm::SetVector<mlir::Value> valuesDefinedAbove;
-  mlir::getUsedValuesDefinedAbove(region, valuesDefinedAbove);
-  while (!valuesDefinedAbove.empty()) {
-    for (mlir::Value val : valuesDefinedAbove) {
-      mlir::Operation *valOp = val.getDefiningOp();
-      if (mlir::isMemoryEffectFree(valOp)) {
-        mlir::Operation *clonedOp = valOp->clone();
-        regionBlock->push_front(clonedOp);
-        val.replaceUsesWithIf(
-            clonedOp->getResult(0), [regionBlock](mlir::OpOperand &use) {
-              return use.getOwner()->getBlock() == regionBlock;
-            });
-      } else {
-        auto savedIP = firOpBuilder.getInsertionPoint();
-        firOpBuilder.setInsertionPointAfter(valOp);
-        auto copyVal =
-            firOpBuilder.createTemporary(val.getLoc(), val.getType());
-        firOpBuilder.createStoreWithConvert(copyVal.getLoc(), val, copyVal);
+static mlir::omp::SectionOp
+genSectionOp(Fortran::lower::AbstractConverter &converter,
+             Fortran::semantics::SemanticsContext &semaCtx,
+             Fortran::lower::pft::Evaluation &eval, bool genNested,
+             mlir::Location loc,
+             const Fortran::parser::OmpClauseList &clauseList) {
+  // Currently only private/firstprivate clause is handled, and
+  // all privatization is done within `omp.section` operations.
+  return genOpWithBody<mlir::omp::SectionOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setClauses(&clauseList));
+}
 
-        llvm::SmallVector<mlir::Value> bounds;
-        std::stringstream name;
-        firOpBuilder.setInsertionPoint(targetOp);
-        mlir::Value mapOp = createMapInfoOp(
-            firOpBuilder, copyVal.getLoc(), copyVal, mlir::Value{}, name.str(),
-            bounds, llvm::SmallVector<mlir::Value>{},
-            static_cast<
-                std::underlying_type_t<llvm::omp::OpenMPOffloadMappingFlags>>(
-                llvm::omp::OpenMPOffloadMappingFlags::OMP_MAP_IMPLICIT),
-            mlir::omp::VariableCaptureKind::ByCopy, copyVal.getType());
-        targetOp.getMapOperandsMutable().append(mapOp);
-        mlir::Value clonedValArg =
-            region.addArgument(copyVal.getType(), copyVal.getLoc());
-        firOpBuilder.setInsertionPointToStart(regionBlock);
-        auto loadOp = firOpBuilder.create<fir::LoadOp>(clonedValArg.getLoc(),
-                                                       clonedValArg);
-        val.replaceUsesWithIf(
-            loadOp->getResult(0), [regionBlock](mlir::OpOperand &use) {
-              return use.getOwner()->getBlock() == regionBlock;
-            });
-        firOpBuilder.setInsertionPoint(regionBlock, savedIP);
-      }
-    }
-    valuesDefinedAbove.clear();
-    mlir::getUsedValuesDefinedAbove(region, valuesDefinedAbove);
-  }
+static mlir::omp::SectionsOp
+genSectionsOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+              const mlir::omp::SectionsClauseOps &clauseOps) {
+  return genOpWithBody<mlir::omp::SectionsOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(false),
+      clauseOps);
+}
 
-  // Insert dummy instruction to remember the insertion position. The
-  // marker will be deleted since there are not uses.
-  // In the HLFIR flow there are hlfir.declares inserted above while
-  // setting block arguments.
-  mlir::Value undefMarker = firOpBuilder.create<fir::UndefOp>(
-      targetOp.getOperation()->getLoc(), firOpBuilder.getIndexType());
+static mlir::omp::SimdLoopOp
+genSimdLoopOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+              const Fortran::parser::OmpClauseList &clauseList) {
+  DataSharingProcessor dsp(converter, semaCtx, clauseList, eval);
+  dsp.processStep1();
 
-  // Create blocks for unstructured regions. This has to be done since
-  // blocks are initially allocated with the function as the parent region.
-  if (eval.lowerAsUnstructured()) {
-    Fortran::lower::createEmptyRegionBlocks<mlir::omp::TerminatorOp,
-                                            mlir::omp::YieldOp>(
-        firOpBuilder, eval.getNestedEvaluations());
-  }
+  Fortran::lower::StatementContext stmtCtx;
+  mlir::omp::SimdLoopClauseOps clauseOps;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
+  genSimdLoopClauses(converter, semaCtx, stmtCtx, eval, clauseList, loc,
+                     clauseOps, iv);
 
-  firOpBuilder.create<mlir::omp::TerminatorOp>(currentLocation);
+  auto *nestedEval =
+      getCollapsedLoopEval(eval, Fortran::lower::getCollapseValue(clauseList));
 
-  // Create the insertion point after the marker.
-  firOpBuilder.setInsertionPointAfter(undefMarker.getDefiningOp());
-  if (genNested)
-    genNestedEvaluations(converter, eval);
+  auto ivCallback = [&](mlir::Operation *op) {
+    return genLoopVars(op, converter, loc, iv);
+  };
+
+  return genOpWithBody<mlir::omp::SimdLoopOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
+          .setClauses(&clauseList)
+          .setDataSharingProcessor(&dsp)
+          .setGenRegionEntryCb(ivCallback),
+      clauseOps);
+}
+
+static mlir::omp::SingleOp
+genSingleOp(Fortran::lower::AbstractConverter &converter,
+            Fortran::semantics::SemanticsContext &semaCtx,
+            Fortran::lower::pft::Evaluation &eval, bool genNested,
+            mlir::Location loc,
+            const Fortran::parser::OmpClauseList &beginClauseList,
+            const Fortran::parser::OmpClauseList &endClauseList) {
+  mlir::omp::SingleClauseOps clauseOps;
+  genSingleClauses(converter, semaCtx, beginClauseList, endClauseList, loc,
+                   clauseOps);
+
+  return genOpWithBody<mlir::omp::SingleOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setClauses(&beginClauseList),
+      clauseOps);
 }
 
 static mlir::omp::TargetOp
 genTargetOp(Fortran::lower::AbstractConverter &converter,
             Fortran::semantics::SemanticsContext &semaCtx,
             Fortran::lower::pft::Evaluation &eval, bool genNested,
-            mlir::Location currentLocation,
+            mlir::Location loc,
             const Fortran::parser::OmpClauseList &clauseList,
-            llvm::omp::Directive directive, bool outerCombined = false) {
+            bool outerCombined = false) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   Fortran::lower::StatementContext stmtCtx;
+
+  bool processHostOnlyClauses =
+      !llvm::cast<mlir::omp::OffloadModuleInterface>(*converter.getModuleOp())
+           .getIsTargetDevice();
+
   mlir::omp::TargetClauseOps clauseOps;
-  llvm::SmallVector<mlir::Type> mapTypes, devicePtrTypes, deviceAddrTypes;
-  llvm::SmallVector<mlir::Location> mapLocs, devicePtrLocs, deviceAddrLocs;
   llvm::SmallVector<const Fortran::semantics::Symbol *> mapSyms, devicePtrSyms,
       deviceAddrSyms;
-
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_target, clauseOps);
-  cp.processDevice(stmtCtx, clauseOps);
-  cp.processThreadLimit(stmtCtx, clauseOps);
-  cp.processDepend(clauseOps);
-  cp.processNowait(clauseOps);
-  cp.processMap(currentLocation, directive, stmtCtx, clauseOps, &mapSyms,
-                &mapLocs, &mapTypes);
-  cp.processIsDevicePtr(clauseOps, devicePtrTypes, devicePtrLocs,
-                        devicePtrSyms);
-  cp.processHasDeviceAddr(clauseOps, deviceAddrTypes, deviceAddrLocs,
-                          deviceAddrSyms);
-  // TODO Support delayed privatization.
-
-  cp.processTODO<clause::Private, clause::Firstprivate, clause::Reduction,
-                 clause::InReduction, clause::Allocate, clause::UsesAllocators,
-                 clause::Defaultmap>(currentLocation,
-                                     llvm::omp::Directive::OMPD_target);
+  llvm::SmallVector<mlir::Location> mapLocs, devicePtrLocs, deviceAddrLocs;
+  llvm::SmallVector<mlir::Type> mapTypes, devicePtrTypes, deviceAddrTypes;
+  genTargetClauses(converter, semaCtx, stmtCtx, clauseList, loc,
+                   processHostOnlyClauses, /*processReduction=*/outerCombined,
+                   clauseOps, mapSyms, mapLocs, mapTypes, deviceAddrSyms,
+                   deviceAddrLocs, deviceAddrTypes, devicePtrSyms,
+                   devicePtrLocs, devicePtrTypes);
 
   // 5.8.1 Implicit Data-Mapping Attribute Rules
   // The following code follows the implicit data-mapping rules to map all the
@@ -1056,22 +1515,21 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
         fir::ExtendedValue dataExv = converter.getSymbolExtendedValue(sym);
         name << sym.name().ToString();
 
-        Fortran::lower::AddrAndBoundsInfo info =
-            getDataOperandBaseAddr(converter, converter.getFirOpBuilder(), sym,
-                                   converter.getCurrentLocation());
+        Fortran::lower::AddrAndBoundsInfo info = getDataOperandBaseAddr(
+            converter, firOpBuilder, sym, converter.getCurrentLocation());
         if (fir::unwrapRefType(info.addr.getType()).isa<fir::BaseBoxType>())
           bounds =
               Fortran::lower::genBoundsOpsFromBox<mlir::omp::MapBoundsOp,
                                                   mlir::omp::MapBoundsType>(
-                  converter.getFirOpBuilder(), converter.getCurrentLocation(),
-                  converter, dataExv, info);
+                  firOpBuilder, converter.getCurrentLocation(), converter,
+                  dataExv, info);
         if (fir::unwrapRefType(info.addr.getType()).isa<fir::SequenceType>()) {
           bool dataExvIsAssumedSize =
               Fortran::semantics::IsAssumedSizeArray(sym.GetUltimate());
           bounds = Fortran::lower::genBaseBoundsOps<mlir::omp::MapBoundsOp,
                                                     mlir::omp::MapBoundsType>(
-              converter.getFirOpBuilder(), converter.getCurrentLocation(),
-              converter, dataExv, dataExvIsAssumedSize);
+              firOpBuilder, converter.getCurrentLocation(), converter, dataExv,
+              dataExvIsAssumedSize);
         }
 
         llvm::omp::OpenMPOffloadMappingFlags mapFlag =
@@ -1085,7 +1543,7 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
 
         // If a variable is specified in declare target link and if device
         // type is not specified as `nohost`, it needs to be mapped tofrom
-        mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
+        mlir::ModuleOp mod = firOpBuilder.getModule();
         mlir::Operation *op = mod.lookupSymbol(converter.mangleName(sym));
         auto declareTargetOp =
             llvm::dyn_cast_if_present<mlir::omp::DeclareTargetInterface>(op);
@@ -1105,8 +1563,8 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
         }
 
         mlir::Value mapOp = createMapInfoOp(
-            converter.getFirOpBuilder(), baseOp.getLoc(), baseOp, mlir::Value{},
-            name.str(), bounds, {},
+            firOpBuilder, baseOp.getLoc(), baseOp, mlir::Value{}, name.str(),
+            bounds, {},
             static_cast<
                 std::underlying_type_t<llvm::omp::OpenMPOffloadMappingFlags>>(
                 mapFlag),
@@ -1118,341 +1576,147 @@ genTargetOp(Fortran::lower::AbstractConverter &converter,
         mapTypes.push_back(baseOp.getType());
       }
     }
-  };
-  Fortran::lower::pft::visitAllSymbols(eval, captureImplicitMap);
-
-  auto targetOp = converter.getFirOpBuilder().create<mlir::omp::TargetOp>(
-      currentLocation, clauseOps);
-
-  genBodyOfTargetOp(converter, semaCtx, eval, genNested, targetOp, mapSyms,
-                    mapLocs, mapTypes, currentLocation);
-
-  return targetOp;
-}
-
-static mlir::omp::TeamsOp
-genTeamsOp(Fortran::lower::AbstractConverter &converter,
-           Fortran::semantics::SemanticsContext &semaCtx,
-           Fortran::lower::pft::Evaluation &eval, bool genNested,
-           mlir::Location currentLocation,
-           const Fortran::parser::OmpClauseList &clauseList,
-           bool outerCombined = false) {
-  Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::TeamsClauseOps clauseOps;
-
-  ClauseProcessor cp(converter, semaCtx, clauseList);
-  cp.processIf(llvm::omp::Directive::OMPD_teams, clauseOps);
-  cp.processAllocate(clauseOps);
-  cp.processDefault();
-  cp.processNumTeams(stmtCtx, clauseOps);
-  cp.processThreadLimit(stmtCtx, clauseOps);
-  // TODO Support delayed privatization.
-
-  cp.processTODO<clause::Reduction>(currentLocation,
-                                    llvm::omp::Directive::OMPD_teams);
-
-  return genOpWithBody<mlir::omp::TeamsOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(genNested)
-          .setOuterCombined(outerCombined)
-          .setClauses(&clauseList),
-      clauseOps);
-}
-
-/// Extract the list of function and variable symbols affected by the given
-/// 'declare target' directive and return the intended device type for them.
-static void getDeclareTargetInfo(
-    Fortran::lower::AbstractConverter &converter,
-    Fortran::semantics::SemanticsContext &semaCtx,
-    Fortran::lower::pft::Evaluation &eval,
-    const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
-    mlir::omp::DeclareTargetClauseOps &clauseOps,
-    llvm::SmallVectorImpl<DeclareTargetCapturePair> &symbolAndClause) {
-  const auto &spec = std::get<Fortran::parser::OmpDeclareTargetSpecifier>(
-      declareTargetConstruct.t);
-  if (const auto *objectList{
-          Fortran::parser::Unwrap<Fortran::parser::OmpObjectList>(spec.u)}) {
-    ObjectList objects{makeObjects(*objectList, semaCtx)};
-    // Case: declare target(func, var1, var2)
-    gatherFuncAndVarSyms(objects, mlir::omp::DeclareTargetCaptureClause::to,
-                         symbolAndClause);
-  } else if (const auto *clauseList{
-                 Fortran::parser::Unwrap<Fortran::parser::OmpClauseList>(
-                     spec.u)}) {
-    if (clauseList->v.empty()) {
-      // Case: declare target, implicit capture of function
-      symbolAndClause.emplace_back(
-          mlir::omp::DeclareTargetCaptureClause::to,
-          eval.getOwningProcedure()->getSubprogramSymbol());
-    }
-
-    ClauseProcessor cp(converter, semaCtx, *clauseList);
-    cp.processTo(symbolAndClause);
-    cp.processEnter(symbolAndClause);
-    cp.processLink(symbolAndClause);
-    cp.processDeviceType(clauseOps);
-    cp.processTODO<clause::Indirect>(converter.getCurrentLocation(),
-                                     llvm::omp::Directive::OMPD_declare_target);
-  }
-}
-
-static void collectDeferredDeclareTargets(
-    Fortran::lower::AbstractConverter &converter,
-    Fortran::semantics::SemanticsContext &semaCtx,
-    Fortran::lower::pft::Evaluation &eval,
-    const Fortran::parser::OpenMPDeclareTargetConstruct &declareTargetConstruct,
-    llvm::SmallVectorImpl<Fortran::lower::OMPDeferredDeclareTargetInfo>
-        &deferredDeclareTarget) {
-  mlir::omp::DeclareTargetClauseOps clauseOps;
-  llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
-  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
-                       clauseOps, symbolAndClause);
-  // Return the device type only if at least one of the targets for the
-  // directive is a function or subroutine
-  mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
-
-  for (const DeclareTargetCapturePair &symClause : symbolAndClause) {
-    mlir::Operation *op = mod.lookupSymbol(converter.mangleName(
-        std::get<const Fortran::semantics::Symbol &>(symClause)));
-
-    if (!op) {
-      deferredDeclareTarget.push_back({std::get<0>(symClause),
-                                       clauseOps.deviceType,
-                                       std::get<1>(symClause)});
-    }
-  }
-}
-
-static std::optional<mlir::omp::DeclareTargetDeviceType>
-getDeclareTargetFunctionDevice(
-    Fortran::lower::AbstractConverter &converter,
-    Fortran::semantics::SemanticsContext &semaCtx,
-    Fortran::lower::pft::Evaluation &eval,
-    const Fortran::parser::OpenMPDeclareTargetConstruct
-        &declareTargetConstruct) {
-  mlir::omp::DeclareTargetClauseOps clauseOps;
-  llvm::SmallVector<DeclareTargetCapturePair> symbolAndClause;
-  getDeclareTargetInfo(converter, semaCtx, eval, declareTargetConstruct,
-                       clauseOps, symbolAndClause);
-
-  // Return the device type only if at least one of the targets for the
-  // directive is a function or subroutine
-  mlir::ModuleOp mod = converter.getFirOpBuilder().getModule();
-  for (const DeclareTargetCapturePair &symClause : symbolAndClause) {
-    mlir::Operation *op = mod.lookupSymbol(converter.mangleName(
-        std::get<const Fortran::semantics::Symbol &>(symClause)));
-
-    if (mlir::isa_and_nonnull<mlir::func::FuncOp>(op))
-      return clauseOps.deviceType;
-  }
-
-  return std::nullopt;
-}
-
-//===----------------------------------------------------------------------===//
-// genOMP() Code generation helper functions
-//===----------------------------------------------------------------------===//
-
-static void
-genOmpSimpleStandalone(Fortran::lower::AbstractConverter &converter,
-                       Fortran::semantics::SemanticsContext &semaCtx,
-                       Fortran::lower::pft::Evaluation &eval, bool genNested,
-                       const Fortran::parser::OpenMPSimpleStandaloneConstruct
-                           &simpleStandaloneConstruct) {
-  const auto &directive =
-      std::get<Fortran::parser::OmpSimpleStandaloneDirective>(
-          simpleStandaloneConstruct.t);
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  const auto &opClauseList =
-      std::get<Fortran::parser::OmpClauseList>(simpleStandaloneConstruct.t);
-  mlir::Location currentLocation = converter.genLocation(directive.source);
-
-  switch (directive.v) {
-  default:
-    break;
-  case llvm::omp::Directive::OMPD_barrier:
-    firOpBuilder.create<mlir::omp::BarrierOp>(currentLocation);
-    break;
-  case llvm::omp::Directive::OMPD_taskwait: {
-    mlir::omp::TaskwaitClauseOps clauseOps;
-    ClauseProcessor cp(converter, semaCtx, opClauseList);
-    cp.processTODO<clause::Depend, clause::Nowait>(
-        currentLocation, llvm::omp::Directive::OMPD_taskwait);
-    firOpBuilder.create<mlir::omp::TaskwaitOp>(currentLocation, clauseOps);
-    break;
-  }
-  case llvm::omp::Directive::OMPD_taskyield:
-    firOpBuilder.create<mlir::omp::TaskyieldOp>(currentLocation);
-    break;
-  case llvm::omp::Directive::OMPD_target_data:
-    genTargetDataOp(converter, semaCtx, eval, genNested, currentLocation,
-                    opClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_target_enter_data:
-    genTargetEnterExitDataUpdateOp<mlir::omp::TargetEnterDataOp>(
-        converter, semaCtx, currentLocation, opClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_target_exit_data:
-    genTargetEnterExitDataUpdateOp<mlir::omp::TargetExitDataOp>(
-        converter, semaCtx, currentLocation, opClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_target_update:
-    genTargetEnterExitDataUpdateOp<mlir::omp::TargetUpdateOp>(
-        converter, semaCtx, currentLocation, opClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_ordered:
-    TODO(currentLocation, "OMPD_ordered");
-  }
-}
-
-static void
-genOmpFlush(Fortran::lower::AbstractConverter &converter,
-            Fortran::semantics::SemanticsContext &semaCtx,
-            Fortran::lower::pft::Evaluation &eval,
-            const Fortran::parser::OpenMPFlushConstruct &flushConstruct) {
-  llvm::SmallVector<mlir::Value, 4> operandRange;
-  if (const auto &ompObjectList =
-          std::get<std::optional<Fortran::parser::OmpObjectList>>(
-              flushConstruct.t))
-    genObjectList2(*ompObjectList, converter, operandRange);
-  const auto &memOrderClause =
-      std::get<std::optional<std::list<Fortran::parser::OmpMemoryOrderClause>>>(
-          flushConstruct.t);
-  if (memOrderClause && memOrderClause->size() > 0)
-    TODO(converter.getCurrentLocation(), "Handle OmpMemoryOrderClause");
-  converter.getFirOpBuilder().create<mlir::omp::FlushOp>(
-      converter.getCurrentLocation(), operandRange);
-}
-
-static llvm::SmallVector<const Fortran::semantics::Symbol *>
-genLoopVars(mlir::Operation *op, Fortran::lower::AbstractConverter &converter,
-            mlir::Location &loc,
-            llvm::ArrayRef<const Fortran::semantics::Symbol *> args) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  auto &region = op->getRegion(0);
-
-  std::size_t loopVarTypeSize = 0;
-  for (const Fortran::semantics::Symbol *arg : args)
-    loopVarTypeSize = std::max(loopVarTypeSize, arg->GetUltimate().size());
-  mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize);
-  llvm::SmallVector<mlir::Type> tiv(args.size(), loopVarType);
-  llvm::SmallVector<mlir::Location> locs(args.size(), loc);
-  firOpBuilder.createBlock(&region, {}, tiv, locs);
-  // The argument is not currently in memory, so make a temporary for the
-  // argument, and store it there, then bind that location to the argument.
-  mlir::Operation *storeOp = nullptr;
-  for (auto [argIndex, argSymbol] : llvm::enumerate(args)) {
-    mlir::Value indexVal = fir::getBase(region.front().getArgument(argIndex));
-    storeOp =
-        createAndSetPrivatizedLoopVar(converter, loc, indexVal, argSymbol);
-  }
-  firOpBuilder.setInsertionPointAfter(storeOp);
-
-  return llvm::SmallVector<const Fortran::semantics::Symbol *>(args);
-}
-
-static llvm::SmallVector<const Fortran::semantics::Symbol *>
-genLoopAndReductionVars(
-    mlir::Operation *op, Fortran::lower::AbstractConverter &converter,
-    mlir::Location &loc,
-    llvm::ArrayRef<const Fortran::semantics::Symbol *> loopArgs,
-    llvm::ArrayRef<const Fortran::semantics::Symbol *> reductionArgs,
-    llvm::ArrayRef<mlir::Type> reductionTypes) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-
-  llvm::SmallVector<mlir::Type> blockArgTypes;
-  llvm::SmallVector<mlir::Location> blockArgLocs;
-  blockArgTypes.reserve(loopArgs.size() + reductionArgs.size());
-  blockArgLocs.reserve(blockArgTypes.size());
-  mlir::Block *entryBlock;
-
-  if (loopArgs.size()) {
-    std::size_t loopVarTypeSize = 0;
-    for (const Fortran::semantics::Symbol *arg : loopArgs)
-      loopVarTypeSize = std::max(loopVarTypeSize, arg->GetUltimate().size());
-    mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize);
-    std::fill_n(std::back_inserter(blockArgTypes), loopArgs.size(),
-                loopVarType);
-    std::fill_n(std::back_inserter(blockArgLocs), loopArgs.size(), loc);
-  }
-  if (reductionArgs.size()) {
-    llvm::copy(reductionTypes, std::back_inserter(blockArgTypes));
-    std::fill_n(std::back_inserter(blockArgLocs), reductionArgs.size(), loc);
-  }
-  entryBlock = firOpBuilder.createBlock(&op->getRegion(0), {}, blockArgTypes,
-                                        blockArgLocs);
-  // The argument is not currently in memory, so make a temporary for the
-  // argument, and store it there, then bind that location to the argument.
-  if (loopArgs.size()) {
-    mlir::Operation *storeOp = nullptr;
-    for (auto [argIndex, argSymbol] : llvm::enumerate(loopArgs)) {
-      mlir::Value indexVal =
-          fir::getBase(op->getRegion(0).front().getArgument(argIndex));
-      storeOp =
-          createAndSetPrivatizedLoopVar(converter, loc, indexVal, argSymbol);
-    }
-    firOpBuilder.setInsertionPointAfter(storeOp);
-  }
-  // Bind the reduction arguments to their block arguments
-  for (auto [arg, prv] : llvm::zip_equal(
-           reductionArgs,
-           llvm::drop_begin(entryBlock->getArguments(), loopArgs.size()))) {
-    converter.bindSymbol(*arg, prv);
-  }
+  };
+  Fortran::lower::pft::visitAllSymbols(eval, captureImplicitMap);
 
-  return llvm::SmallVector<const Fortran::semantics::Symbol *>(loopArgs);
+  auto targetOp = firOpBuilder.create<mlir::omp::TargetOp>(loc, clauseOps);
+  genBodyOfTargetOp(converter, semaCtx, eval, genNested, targetOp, mapSyms,
+                    mapLocs, mapTypes, loc);
+  return targetOp;
 }
 
-static void
-createSimdLoop(Fortran::lower::AbstractConverter &converter,
-               Fortran::semantics::SemanticsContext &semaCtx,
-               Fortran::lower::pft::Evaluation &eval,
-               llvm::omp::Directive ompDirective,
-               const Fortran::parser::OmpClauseList &loopOpClauseList,
-               mlir::Location loc) {
+static mlir::omp::TargetDataOp
+genTargetDataOp(Fortran::lower::AbstractConverter &converter,
+                Fortran::semantics::SemanticsContext &semaCtx,
+                Fortran::lower::pft::Evaluation &eval, bool genNested,
+                mlir::Location loc,
+                const Fortran::parser::OmpClauseList &clauseList) {
+  Fortran::lower::StatementContext stmtCtx;
+  mlir::omp::TargetDataClauseOps clauseOps;
+  llvm::SmallVector<mlir::Type> useDeviceTypes;
+  llvm::SmallVector<mlir::Location> useDeviceLocs;
+  llvm::SmallVector<const Fortran::semantics::Symbol *> useDeviceSyms;
+  genTargetDataClauses(converter, semaCtx, stmtCtx, clauseList, loc, clauseOps,
+                       useDeviceTypes, useDeviceLocs, useDeviceSyms);
+
+  auto targetDataOp =
+      converter.getFirOpBuilder().create<mlir::omp::TargetDataOp>(loc,
+                                                                  clauseOps);
+  genBodyOfTargetDataOp(converter, semaCtx, eval, genNested, targetDataOp,
+                        useDeviceTypes, useDeviceLocs, useDeviceSyms, loc);
+  return targetDataOp;
+}
+
+template <typename OpTy>
+static OpTy genTargetEnterExitUpdateDataOp(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx, mlir::Location loc,
+    const Fortran::parser::OmpClauseList &clauseList) {
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  DataSharingProcessor dsp(converter, semaCtx, loopOpClauseList, eval);
-  dsp.processStep1();
+  Fortran::lower::StatementContext stmtCtx;
+
+  // GCC 9.3.0 emits a (probably) bogus warning about an unused variable.
+  [[maybe_unused]] llvm::omp::Directive directive;
+  if constexpr (std::is_same_v<OpTy, mlir::omp::TargetEnterDataOp>) {
+    directive = llvm::omp::Directive::OMPD_target_enter_data;
+  } else if constexpr (std::is_same_v<OpTy, mlir::omp::TargetExitDataOp>) {
+    directive = llvm::omp::Directive::OMPD_target_exit_data;
+  } else if constexpr (std::is_same_v<OpTy, mlir::omp::TargetUpdateOp>) {
+    directive = llvm::omp::Directive::OMPD_target_update;
+  } else {
+    llvm_unreachable("Unexpected TARGET DATA construct");
+  }
+
+  mlir::omp::TargetEnterExitUpdateDataClauseOps clauseOps;
+  genTargetEnterExitUpdateDataClauses(converter, semaCtx, stmtCtx, clauseList,
+                                      loc, directive, clauseOps);
+
+  return firOpBuilder.create<OpTy>(loc, clauseOps);
+}
 
+static mlir::omp::TaskOp
+genTaskOp(Fortran::lower::AbstractConverter &converter,
+          Fortran::semantics::SemanticsContext &semaCtx,
+          Fortran::lower::pft::Evaluation &eval, bool genNested,
+          mlir::Location loc,
+          const Fortran::parser::OmpClauseList &clauseList) {
   Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::SimdLoopClauseOps clauseOps;
-  llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
+  mlir::omp::TaskClauseOps clauseOps;
+  genTaskClauses(converter, semaCtx, stmtCtx, clauseList, loc, clauseOps);
 
-  ClauseProcessor cp(converter, semaCtx, loopOpClauseList);
-  cp.processCollapse(loc, eval, clauseOps, iv);
-  cp.processReduction(loc, clauseOps);
-  cp.processIf(llvm::omp::Directive::OMPD_simd, clauseOps);
-  cp.processSimdlen(clauseOps);
-  cp.processSafelen(clauseOps);
-  clauseOps.loopInclusiveAttr = firOpBuilder.getUnitAttr();
-  // TODO Support delayed privatization.
+  return genOpWithBody<mlir::omp::TaskOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setClauses(&clauseList),
+      clauseOps);
+}
 
-  cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
-                 clause::Nontemporal, clause::Order>(loc, ompDirective);
+static mlir::omp::TaskgroupOp
+genTaskgroupOp(Fortran::lower::AbstractConverter &converter,
+               Fortran::semantics::SemanticsContext &semaCtx,
+               Fortran::lower::pft::Evaluation &eval, bool genNested,
+               mlir::Location loc,
+               const Fortran::parser::OmpClauseList &clauseList) {
+  mlir::omp::TaskgroupClauseOps clauseOps;
+  genTaskgroupClauses(converter, semaCtx, clauseList, loc, clauseOps);
 
-  auto *nestedEval = getCollapsedLoopEval(
-      eval, Fortran::lower::getCollapseValue(loopOpClauseList));
+  return genOpWithBody<mlir::omp::TaskgroupOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setClauses(&clauseList),
+      clauseOps);
+}
 
-  auto ivCallback = [&](mlir::Operation *op) {
-    return genLoopVars(op, converter, loc, iv);
-  };
+static mlir::omp::TaskloopOp
+genTaskloopOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+              const Fortran::parser::OmpClauseList &clauseList) {
+  TODO(loc, "Taskloop construct");
+}
 
-  genOpWithBody<mlir::omp::SimdLoopOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
-          .setClauses(&loopOpClauseList)
-          .setDataSharingProcessor(&dsp)
-          .setGenRegionEntryCb(ivCallback),
+static mlir::omp::TaskwaitOp
+genTaskwaitOp(Fortran::lower::AbstractConverter &converter,
+              Fortran::semantics::SemanticsContext &semaCtx,
+              Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+              const Fortran::parser::OmpClauseList &clauseList) {
+  mlir::omp::TaskwaitClauseOps clauseOps;
+  genTaskwaitClauses(converter, semaCtx, clauseList, loc, clauseOps);
+  return converter.getFirOpBuilder().create<mlir::omp::TaskwaitOp>(loc,
+                                                                   clauseOps);
+}
+
+static mlir::omp::TaskyieldOp
+genTaskyieldOp(Fortran::lower::AbstractConverter &converter,
+               Fortran::semantics::SemanticsContext &semaCtx,
+               Fortran::lower::pft::Evaluation &eval, mlir::Location loc) {
+  return converter.getFirOpBuilder().create<mlir::omp::TaskyieldOp>(loc);
+}
+
+static mlir::omp::TeamsOp
+genTeamsOp(Fortran::lower::AbstractConverter &converter,
+           Fortran::semantics::SemanticsContext &semaCtx,
+           Fortran::lower::pft::Evaluation &eval, bool genNested,
+           mlir::Location loc, const Fortran::parser::OmpClauseList &clauseList,
+           bool outerCombined = false) {
+  Fortran::lower::StatementContext stmtCtx;
+  mlir::omp::TeamsClauseOps clauseOps;
+  genTeamsClauses(converter, semaCtx, stmtCtx, clauseList, loc, clauseOps);
+
+  return genOpWithBody<mlir::omp::TeamsOp>(
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+          .setGenNested(genNested)
+          .setOuterCombined(outerCombined)
+          .setClauses(&clauseList),
       clauseOps);
 }
 
-static void createWsloop(Fortran::lower::AbstractConverter &converter,
-                         Fortran::semantics::SemanticsContext &semaCtx,
-                         Fortran::lower::pft::Evaluation &eval,
-                         llvm::omp::Directive ompDirective,
-                         const Fortran::parser::OmpClauseList &beginClauseList,
-                         const Fortran::parser::OmpClauseList *endClauseList,
-                         mlir::Location loc) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+static mlir::omp::WsloopOp
+genWsloopOp(Fortran::lower::AbstractConverter &converter,
+            Fortran::semantics::SemanticsContext &semaCtx,
+            Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+            const Fortran::parser::OmpClauseList &beginClauseList,
+            const Fortran::parser::OmpClauseList *endClauseList) {
   DataSharingProcessor dsp(converter, semaCtx, beginClauseList, eval);
   dsp.processStep1();
 
@@ -1461,30 +1725,9 @@ static void createWsloop(Fortran::lower::AbstractConverter &converter,
   llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
   llvm::SmallVector<mlir::Type> reductionTypes;
   llvm::SmallVector<const Fortran::semantics::Symbol *> reductionSyms;
-
-  ClauseProcessor cp(converter, semaCtx, beginClauseList);
-  cp.processCollapse(loc, eval, clauseOps, iv);
-  cp.processSchedule(stmtCtx, clauseOps);
-  cp.processReduction(loc, clauseOps, &reductionTypes, &reductionSyms);
-  cp.processOrdered(clauseOps);
-  clauseOps.loopInclusiveAttr = firOpBuilder.getUnitAttr();
-  // TODO Support delayed privatization.
-
-  if (ReductionProcessor::doReductionByRef(clauseOps.reductionVars))
-    clauseOps.reductionByRefAttr = firOpBuilder.getUnitAttr();
-
-  cp.processTODO<clause::Allocate, clause::Linear, clause::Order>(loc,
-                                                                  ompDirective);
-
-  // In FORTRAN `nowait` clause occur at the end of `omp do` directive.
-  // i.e
-  // !$omp do
-  // <...>
-  // !$omp end do nowait
-  if (endClauseList) {
-    ClauseProcessor ecp(converter, semaCtx, *endClauseList);
-    ecp.processNowait(clauseOps);
-  }
+  genWsloopClauses(converter, semaCtx, stmtCtx, eval, beginClauseList,
+                   endClauseList, loc, clauseOps, iv, reductionTypes,
+                   reductionSyms);
 
   auto *nestedEval = getCollapsedLoopEval(
       eval, Fortran::lower::getCollapseValue(beginClauseList));
@@ -1494,7 +1737,7 @@ static void createWsloop(Fortran::lower::AbstractConverter &converter,
                                    reductionTypes);
   };
 
-  genOpWithBody<mlir::omp::WsloopOp>(
+  return genOpWithBody<mlir::omp::WsloopOp>(
       OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
           .setClauses(&beginClauseList)
           .setDataSharingProcessor(&dsp)
@@ -1503,7 +1746,11 @@ static void createWsloop(Fortran::lower::AbstractConverter &converter,
       clauseOps);
 }
 
-static void createSimdWsloop(
+//===----------------------------------------------------------------------===//
+// Code generation functions for composite constructs
+//===----------------------------------------------------------------------===//
+
+static void genCompositeDoSimd(
     Fortran::lower::AbstractConverter &converter,
     Fortran::semantics::SemanticsContext &semaCtx,
     Fortran::lower::pft::Evaluation &eval, llvm::omp::Directive ompDirective,
@@ -1511,7 +1758,7 @@ static void createSimdWsloop(
     const Fortran::parser::OmpClauseList *endClauseList, mlir::Location loc) {
   ClauseProcessor cp(converter, semaCtx, beginClauseList);
   cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
-                 clause::Safelen, clause::Simdlen, clause::Order>(loc,
+                 clause::Order, clause::Safelen, clause::Simdlen>(loc,
                                                                   ompDirective);
   // TODO: Add support for vectorization - add vectorization hints inside loop
   // body.
@@ -1521,34 +1768,7 @@ static void createSimdWsloop(
   // When support for vectorization is enabled, then we need to add handling of
   // if clause. Currently if clause can be skipped because we always assume
   // SIMD length = 1.
-  createWsloop(converter, semaCtx, eval, ompDirective, beginClauseList,
-               endClauseList, loc);
-}
-
-static void
-markDeclareTarget(mlir::Operation *op,
-                  Fortran::lower::AbstractConverter &converter,
-                  mlir::omp::DeclareTargetCaptureClause captureClause,
-                  mlir::omp::DeclareTargetDeviceType deviceType) {
-  // TODO: Add support for program local variables with declare target applied
-  auto declareTargetOp = llvm::dyn_cast<mlir::omp::DeclareTargetInterface>(op);
-  if (!declareTargetOp)
-    fir::emitFatalError(
-        converter.getCurrentLocation(),
-        "Attempt to apply declare target on unsupported operation");
-
-  // The function or global already has a declare target applied to it, very
-  // likely through implicit capture (usage in another declare target
-  // function/subroutine). It should be marked as any if it has been assigned
-  // both host and nohost, else we skip, as there is no change
-  if (declareTargetOp.isDeclareTarget()) {
-    if (declareTargetOp.getDeclareTargetDeviceType() != deviceType)
-      declareTargetOp.setDeclareTarget(mlir::omp::DeclareTargetDeviceType::any,
-                                       captureClause);
-    return;
-  }
-
-  declareTargetOp.setDeclareTarget(deviceType, captureClause);
+  genWsloopOp(converter, semaCtx, eval, loc, beginClauseList, endClauseList);
 }
 
 //===----------------------------------------------------------------------===//
@@ -1643,6 +1863,102 @@ genOMP(Fortran::lower::AbstractConverter &converter,
       ompDeclConstruct.u);
 }
 
+//===----------------------------------------------------------------------===//
+// OpenMPStandaloneConstruct visitors
+//===----------------------------------------------------------------------===//
+
+static void genOMP(Fortran::lower::AbstractConverter &converter,
+                   Fortran::lower::SymMap &symTable,
+                   Fortran::semantics::SemanticsContext &semaCtx,
+                   Fortran::lower::pft::Evaluation &eval,
+                   const Fortran::parser::OpenMPSimpleStandaloneConstruct
+                       &simpleStandaloneConstruct) {
+  const auto &directive =
+      std::get<Fortran::parser::OmpSimpleStandaloneDirective>(
+          simpleStandaloneConstruct.t);
+  const auto &clauseList =
+      std::get<Fortran::parser::OmpClauseList>(simpleStandaloneConstruct.t);
+  mlir::Location currentLocation = converter.genLocation(directive.source);
+
+  switch (directive.v) {
+  default:
+    break;
+  case llvm::omp::Directive::OMPD_barrier:
+    genBarrierOp(converter, semaCtx, eval, currentLocation);
+    break;
+  case llvm::omp::Directive::OMPD_taskwait:
+    genTaskwaitOp(converter, semaCtx, eval, currentLocation, clauseList);
+    break;
+  case llvm::omp::Directive::OMPD_taskyield:
+    genTaskyieldOp(converter, semaCtx, eval, currentLocation);
+    break;
+  case llvm::omp::Directive::OMPD_target_data:
+    genTargetDataOp(converter, semaCtx, eval, /*genNested=*/true,
+                    currentLocation, clauseList);
+    break;
+  case llvm::omp::Directive::OMPD_target_enter_data:
+    genTargetEnterExitUpdateDataOp<mlir::omp::TargetEnterDataOp>(
+        converter, semaCtx, currentLocation, clauseList);
+    break;
+  case llvm::omp::Directive::OMPD_target_exit_data:
+    genTargetEnterExitUpdateDataOp<mlir::omp::TargetExitDataOp>(
+        converter, semaCtx, currentLocation, clauseList);
+    break;
+  case llvm::omp::Directive::OMPD_target_update:
+    genTargetEnterExitUpdateDataOp<mlir::omp::TargetUpdateOp>(
+        converter, semaCtx, currentLocation, clauseList);
+    break;
+  case llvm::omp::Directive::OMPD_ordered:
+    genOrderedOp(converter, semaCtx, eval, currentLocation, clauseList);
+    break;
+  }
+}
+
+static void
+genOMP(Fortran::lower::AbstractConverter &converter,
+       Fortran::lower::SymMap &symTable,
+       Fortran::semantics::SemanticsContext &semaCtx,
+       Fortran::lower::pft::Evaluation &eval,
+       const Fortran::parser::OpenMPFlushConstruct &flushConstruct) {
+  const auto &verbatim = std::get<Fortran::parser::Verbatim>(flushConstruct.t);
+  const auto &objectList =
+      std::get<std::optional<Fortran::parser::OmpObjectList>>(flushConstruct.t);
+  const auto &clauseList =
+      std::get<std::optional<std::list<Fortran::parser::OmpMemoryOrderClause>>>(
+          flushConstruct.t);
+  mlir::Location currentLocation = converter.genLocation(verbatim.source);
+  genFlushOp(converter, semaCtx, eval, currentLocation, objectList, clauseList);
+}
+
+static void
+genOMP(Fortran::lower::AbstractConverter &converter,
+       Fortran::lower::SymMap &symTable,
+       Fortran::semantics::SemanticsContext &semaCtx,
+       Fortran::lower::pft::Evaluation &eval,
+       const Fortran::parser::OpenMPCancelConstruct &cancelConstruct) {
+  TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct");
+}
+
+static void genOMP(Fortran::lower::AbstractConverter &converter,
+                   Fortran::lower::SymMap &symTable,
+                   Fortran::semantics::SemanticsContext &semaCtx,
+                   Fortran::lower::pft::Evaluation &eval,
+                   const Fortran::parser::OpenMPCancellationPointConstruct
+                       &cancellationPointConstruct) {
+  TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct");
+}
+
+static void
+genOMP(Fortran::lower::AbstractConverter &converter,
+       Fortran::lower::SymMap &symTable,
+       Fortran::semantics::SemanticsContext &semaCtx,
+       Fortran::lower::pft::Evaluation &eval,
+       const Fortran::parser::OpenMPStandaloneConstruct &standaloneConstruct) {
+  std::visit(
+      [&](auto &&s) { return genOMP(converter, symTable, semaCtx, eval, s); },
+      standaloneConstruct.u);
+}
+
 //===----------------------------------------------------------------------===//
 // OpenMPConstruct visitors
 //===----------------------------------------------------------------------===//
@@ -1774,7 +2090,7 @@ genOMP(Fortran::lower::AbstractConverter &converter,
     break;
   case llvm::omp::Directive::OMPD_target:
     genTargetOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-                beginClauseList, directive.v);
+                beginClauseList);
     break;
   case llvm::omp::Directive::OMPD_target_data:
     genTargetDataOp(converter, semaCtx, eval, /*genNested=*/true,
@@ -1790,8 +2106,7 @@ genOMP(Fortran::lower::AbstractConverter &converter,
     break;
   case llvm::omp::Directive::OMPD_teams:
     genTeamsOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-               beginClauseList,
-               /*outerCombined=*/false);
+               beginClauseList);
     break;
   case llvm::omp::Directive::OMPD_workshare:
     // FIXME: Workshare is not a commonly used OpenMP construct, an
@@ -1813,8 +2128,7 @@ genOMP(Fortran::lower::AbstractConverter &converter,
   if ((llvm::omp::allTargetSet & llvm::omp::blockConstructSet)
           .test(directive.v)) {
     genTargetOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-                beginClauseList, directive.v,
-                /*outerCombined=*/true);
+                beginClauseList, /*outerCombined=*/true);
     combinedDirective = true;
   }
   if ((llvm::omp::allTeamsSet & llvm::omp::blockConstructSet)
@@ -1851,44 +2165,13 @@ genOMP(Fortran::lower::AbstractConverter &converter,
        Fortran::semantics::SemanticsContext &semaCtx,
        Fortran::lower::pft::Evaluation &eval,
        const Fortran::parser::OpenMPCriticalConstruct &criticalConstruct) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  mlir::Location currentLocation = converter.getCurrentLocation();
-  std::string name;
-  const Fortran::parser::OmpCriticalDirective &cd =
+  const auto &cd =
       std::get<Fortran::parser::OmpCriticalDirective>(criticalConstruct.t);
-  if (std::get<std::optional<Fortran::parser::Name>>(cd.t).has_value()) {
-    name =
-        std::get<std::optional<Fortran::parser::Name>>(cd.t).value().ToString();
-  }
-
-  mlir::omp::CriticalOp criticalOp = [&]() {
-    if (name.empty()) {
-      return firOpBuilder.create<mlir::omp::CriticalOp>(
-          currentLocation, mlir::FlatSymbolRefAttr());
-    }
-
-    mlir::ModuleOp module = firOpBuilder.getModule();
-    mlir::OpBuilder modBuilder(module.getBodyRegion());
-    auto global = module.lookupSymbol<mlir::omp::CriticalDeclareOp>(name);
-    if (!global) {
-      mlir::omp::CriticalClauseOps clauseOps;
-      const auto &clauseList = std::get<Fortran::parser::OmpClauseList>(cd.t);
-
-      ClauseProcessor cp(converter, semaCtx, clauseList);
-      cp.processHint(clauseOps);
-      clauseOps.nameAttr =
-          mlir::StringAttr::get(firOpBuilder.getContext(), name);
-
-      global = modBuilder.create<mlir::omp::CriticalDeclareOp>(currentLocation,
-                                                               clauseOps);
-    }
-
-    return firOpBuilder.create<mlir::omp::CriticalOp>(
-        currentLocation, mlir::FlatSymbolRefAttr::get(firOpBuilder.getContext(),
-                                                      global.getSymName()));
-  }();
-  auto genInfo = OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval);
-  createBodyOfOp<mlir::omp::CriticalOp>(criticalOp, genInfo);
+  const auto &clauseList = std::get<Fortran::parser::OmpClauseList>(cd.t);
+  const auto &name = std::get<std::optional<Fortran::parser::Name>>(cd.t);
+  mlir::Location currentLocation = converter.getCurrentLocation();
+  genCriticalOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
+                clauseList, name);
 }
 
 static void
@@ -1907,7 +2190,7 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
                    const Fortran::parser::OpenMPLoopConstruct &loopConstruct) {
   const auto &beginLoopDirective =
       std::get<Fortran::parser::OmpBeginLoopDirective>(loopConstruct.t);
-  const auto &loopOpClauseList =
+  const auto &beginClauseList =
       std::get<Fortran::parser::OmpClauseList>(beginLoopDirective.t);
   mlir::Location currentLocation =
       converter.genLocation(beginLoopDirective.source);
@@ -1928,33 +2211,31 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
   bool validDirective = false;
   if (llvm::omp::topTaskloopSet.test(ompDirective)) {
     validDirective = true;
-    TODO(currentLocation, "Taskloop construct");
+    genTaskloopOp(converter, semaCtx, eval, currentLocation, beginClauseList);
   } else {
     // Create omp.{target, teams, distribute, parallel} nested operations
     if ((llvm::omp::allTargetSet & llvm::omp::loopConstructSet)
             .test(ompDirective)) {
       validDirective = true;
       genTargetOp(converter, semaCtx, eval, /*genNested=*/false,
-                  currentLocation, loopOpClauseList, ompDirective,
-                  /*outerCombined=*/true);
+                  currentLocation, beginClauseList, /*outerCombined=*/true);
     }
     if ((llvm::omp::allTeamsSet & llvm::omp::loopConstructSet)
             .test(ompDirective)) {
       validDirective = true;
       genTeamsOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-                 loopOpClauseList,
-                 /*outerCombined=*/true);
+                 beginClauseList, /*outerCombined=*/true);
     }
     if (llvm::omp::allDistributeSet.test(ompDirective)) {
       validDirective = true;
-      TODO(currentLocation, "Distribute construct");
+      genDistributeOp(converter, semaCtx, eval, /*genNested=*/false,
+                      currentLocation, beginClauseList);
     }
     if ((llvm::omp::allParallelSet & llvm::omp::loopConstructSet)
             .test(ompDirective)) {
       validDirective = true;
       genParallelOp(converter, symTable, semaCtx, eval, /*genNested=*/false,
-                    currentLocation, loopOpClauseList,
-                    /*outerCombined=*/true);
+                    currentLocation, beginClauseList, /*outerCombined=*/true);
     }
   }
   if ((llvm::omp::allDoSet | llvm::omp::allSimdSet).test(ompDirective))
@@ -1968,16 +2249,14 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
 
   if (llvm::omp::allDoSimdSet.test(ompDirective)) {
     // 2.9.3.2 Workshare SIMD construct
-    createSimdWsloop(converter, semaCtx, eval, ompDirective, loopOpClauseList,
-                     endClauseList, currentLocation);
-
+    genCompositeDoSimd(converter, semaCtx, eval, ompDirective, beginClauseList,
+                       endClauseList, currentLocation);
   } else if (llvm::omp::allSimdSet.test(ompDirective)) {
     // 2.9.3.1 SIMD construct
-    createSimdLoop(converter, semaCtx, eval, ompDirective, loopOpClauseList,
-                   currentLocation);
+    genSimdLoopOp(converter, semaCtx, eval, currentLocation, beginClauseList);
   } else {
-    createWsloop(converter, semaCtx, eval, ompDirective, loopOpClauseList,
-                 endClauseList, currentLocation);
+    genWsloopOp(converter, semaCtx, eval, currentLocation, beginClauseList,
+                endClauseList);
   }
 }
 
@@ -1997,44 +2276,39 @@ genOMP(Fortran::lower::AbstractConverter &converter,
        Fortran::semantics::SemanticsContext &semaCtx,
        Fortran::lower::pft::Evaluation &eval,
        const Fortran::parser::OpenMPSectionsConstruct &sectionsConstruct) {
-  mlir::Location currentLocation = converter.getCurrentLocation();
-  mlir::omp::SectionsClauseOps clauseOps;
   const auto &beginSectionsDirective =
       std::get<Fortran::parser::OmpBeginSectionsDirective>(sectionsConstruct.t);
-  const auto &sectionsClauseList =
+  const auto &beginClauseList =
       std::get<Fortran::parser::OmpClauseList>(beginSectionsDirective.t);
 
   // Process clauses before optional omp.parallel, so that new variables are
   // allocated outside of the parallel region
-  ClauseProcessor cp(converter, semaCtx, sectionsClauseList);
-  cp.processSectionsReduction(currentLocation, clauseOps);
-  cp.processAllocate(clauseOps);
-  // TODO Support delayed privatization.
+  mlir::Location currentLocation = converter.getCurrentLocation();
+  mlir::omp::SectionsClauseOps clauseOps;
+  genSectionsClauses(converter, semaCtx, beginClauseList, currentLocation,
+                     /*clausesFromBeginSections=*/true, clauseOps);
 
+  // Parallel wrapper of PARALLEL SECTIONS construct
   llvm::omp::Directive dir =
       std::get<Fortran::parser::OmpSectionsDirective>(beginSectionsDirective.t)
           .v;
-
-  // Parallel wrapper of PARALLEL SECTIONS construct
   if (dir == llvm::omp::Directive::OMPD_parallel_sections) {
     genParallelOp(converter, symTable, semaCtx, eval,
-                  /*genNested=*/false, currentLocation, sectionsClauseList,
+                  /*genNested=*/false, currentLocation, beginClauseList,
                   /*outerCombined=*/true);
   } else {
     const auto &endSectionsDirective =
         std::get<Fortran::parser::OmpEndSectionsDirective>(sectionsConstruct.t);
-    const auto &endSectionsClauseList =
+    const auto &endClauseList =
         std::get<Fortran::parser::OmpClauseList>(endSectionsDirective.t);
-    ClauseProcessor(converter, semaCtx, endSectionsClauseList)
-        .processNowait(clauseOps);
+    genSectionsClauses(converter, semaCtx, endClauseList, currentLocation,
+                       /*clausesFromBeginSections=*/false, clauseOps);
   }
 
-  // SECTIONS construct
-  genOpWithBody<mlir::omp::SectionsOp>(
-      OpWithBodyGenInfo(converter, semaCtx, currentLocation, eval)
-          .setGenNested(false),
-      clauseOps);
+  // SECTIONS construct.
+  genSectionsOp(converter, semaCtx, eval, currentLocation, clauseOps);
 
+  // Generate nested SECTION operations recursively.
   const auto &sectionBlocks =
       std::get<Fortran::parser::OmpSectionBlocks>(sectionsConstruct.t);
   auto &firOpBuilder = converter.getFirOpBuilder();
@@ -2043,40 +2317,12 @@ genOMP(Fortran::lower::AbstractConverter &converter,
        llvm::zip(sectionBlocks.v, eval.getNestedEvaluations())) {
     symTable.pushScope();
     genSectionOp(converter, semaCtx, neval, /*genNested=*/true, currentLocation,
-                 sectionsClauseList);
+                 beginClauseList);
     symTable.popScope();
     firOpBuilder.restoreInsertionPoint(ip);
   }
 }
 
-static void
-genOMP(Fortran::lower::AbstractConverter &converter,
-       Fortran::lower::SymMap &symTable,
-       Fortran::semantics::SemanticsContext &semaCtx,
-       Fortran::lower::pft::Evaluation &eval,
-       const Fortran::parser::OpenMPStandaloneConstruct &standaloneConstruct) {
-  std::visit(
-      Fortran::common::visitors{
-          [&](const Fortran::parser::OpenMPSimpleStandaloneConstruct
-                  &simpleStandaloneConstruct) {
-            genOmpSimpleStandalone(converter, semaCtx, eval,
-                                   /*genNested=*/true,
-                                   simpleStandaloneConstruct);
-          },
-          [&](const Fortran::parser::OpenMPFlushConstruct &flushConstruct) {
-            genOmpFlush(converter, semaCtx, eval, flushConstruct);
-          },
-          [&](const Fortran::parser::OpenMPCancelConstruct &cancelConstruct) {
-            TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct");
-          },
-          [&](const Fortran::parser::OpenMPCancellationPointConstruct
-                  &cancellationPointConstruct) {
-            TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct");
-          },
-      },
-      standaloneConstruct.u);
-}
-
 static void genOMP(Fortran::lower::AbstractConverter &converter,
                    Fortran::lower::SymMap &symTable,
                    Fortran::semantics::SemanticsContext &semaCtx,
diff --git a/flang/test/Lower/OpenMP/FIR/target.f90 b/flang/test/Lower/OpenMP/FIR/target.f90
index 022327f9c25da..ca3162340d784 100644
--- a/flang/test/Lower/OpenMP/FIR/target.f90
+++ b/flang/test/Lower/OpenMP/FIR/target.f90
@@ -411,8 +411,8 @@ end subroutine omp_target_implicit_bounds
 !CHECK-LABEL: func.func @_QPomp_target_thread_limit() {
 subroutine omp_target_thread_limit
    integer :: a
-   !CHECK: %[[VAL_1:.*]] = arith.constant 64 : i32
    !CHECK: %[[MAP:.*]] = omp.map.info var_ptr({{.*}})   map_clauses(tofrom) capture(ByRef) -> !fir.ref<i32> {name = "a"}
+   !CHECK: %[[VAL_1:.*]] = arith.constant 64 : i32
    !CHECK: omp.target   thread_limit(%[[VAL_1]] : i32) map_entries(%[[MAP]] -> %[[ARG_0:.*]] : !fir.ref<i32>) {
    !CHECK: ^bb0(%[[ARG_0]]: !fir.ref<i32>):
    !$omp target map(tofrom: a) thread_limit(64)
diff --git a/flang/test/Lower/OpenMP/target.f90 b/flang/test/Lower/OpenMP/target.f90
index 6f72b5a34d069..51b66327dfb24 100644
--- a/flang/test/Lower/OpenMP/target.f90
+++ b/flang/test/Lower/OpenMP/target.f90
@@ -490,8 +490,8 @@ end subroutine omp_target_implicit_bounds
 !CHECK-LABEL: func.func @_QPomp_target_thread_limit() {
 subroutine omp_target_thread_limit
    integer :: a
-   !CHECK: %[[VAL_1:.*]] = arith.constant 64 : i32
    !CHECK: %[[MAP:.*]] = omp.map.info var_ptr({{.*}})   map_clauses(tofrom) capture(ByRef) -> !fir.ref<i32> {name = "a"}
+   !CHECK: %[[VAL_1:.*]] = arith.constant 64 : i32
    !CHECK: omp.target   thread_limit(%[[VAL_1]] : i32) map_entries(%[[MAP]] -> %{{.*}} : !fir.ref<i32>) {
    !CHECK: ^bb0(%{{.*}}: !fir.ref<i32>):
    !$omp target map(tofrom: a) thread_limit(64)
diff --git a/flang/test/Lower/OpenMP/use-device-ptr-to-use-device-addr.f90 b/flang/test/Lower/OpenMP/use-device-ptr-to-use-device-addr.f90
index 33b5971656010..d849dd206b943 100644
--- a/flang/test/Lower/OpenMP/use-device-ptr-to-use-device-addr.f90
+++ b/flang/test/Lower/OpenMP/use-device-ptr-to-use-device-addr.f90
@@ -21,7 +21,7 @@ subroutine only_use_device_ptr
 
 !CHECK: func.func @{{.*}}mix_use_device_ptr_and_addr()
 !CHECK: omp.target_data use_device_ptr({{.*}} : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) use_device_addr(%{{.*}}, %{{.*}} : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>) {
-!CHECK: ^bb0(%{{.*}}: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>):
+!CHECK: ^bb0(%{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, %{{.*}}: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>):
 subroutine mix_use_device_ptr_and_addr 
     use iso_c_binding
     integer, pointer, dimension(:) :: array
@@ -47,7 +47,7 @@ subroutine only_use_device_addr
 
 !CHECK: func.func @{{.*}}mix_use_device_ptr_and_addr_and_map()
 !CHECK: omp.target_data map_entries(%{{.*}}, %{{.*}} : !fir.ref<i32>, !fir.ref<i32>) use_device_ptr(%{{.*}} : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) use_device_addr(%{{.*}}, %{{.*}} : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>) {
-!CHECK: ^bb0(%{{.*}}: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>):
+!CHECK: ^bb0(%{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, %{{.*}}: !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, %{{.*}}: !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>):
 subroutine mix_use_device_ptr_and_addr_and_map
     use iso_c_binding
     integer :: i, j

From 31424be3aef4290dd84065b9371fcd0c5014e097 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Mon, 15 Apr 2024 13:14:41 +0200
Subject: [PATCH 557/886] [clang][Interp][NFC] Compare std::optionals directly

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 93059edc4622f..4a7b40440770e 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1624,7 +1624,7 @@ bool ByteCodeExprGen<Emitter>::VisitCompoundAssignOperator(
     return false;
   if (!this->emitLoad(*LT, E))
     return false;
-  if (*LT != *LHSComputationT) {
+  if (LT != LHSComputationT) {
     if (!this->emitCast(*LT, *LHSComputationT, E))
       return false;
   }
@@ -1680,7 +1680,7 @@ bool ByteCodeExprGen<Emitter>::VisitCompoundAssignOperator(
   }
 
   // And now cast from LHSComputationT to ResultT.
-  if (*ResultT != *LHSComputationT) {
+  if (ResultT != LHSComputationT) {
     if (!this->emitCast(*LHSComputationT, *ResultT, E))
       return false;
   }

From a831c54357c2bb7b8b457ccea22836c23e8b8625 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Mon, 15 Apr 2024 14:17:31 +0200
Subject: [PATCH 558/886] [clang][Interp] Avoid calling invalid functions

Check if the non-null function pointer is even valid before calling
the function.
---
 clang/lib/AST/Interp/FunctionPointer.h |  1 +
 clang/lib/AST/Interp/Interp.h          |  4 ++++
 clang/test/AST/Interp/functions.cpp    | 15 +++++++++++++--
 3 files changed, 18 insertions(+), 2 deletions(-)

diff --git a/clang/lib/AST/Interp/FunctionPointer.h b/clang/lib/AST/Interp/FunctionPointer.h
index c2ea295b82bdf..fc3d7a4214a72 100644
--- a/clang/lib/AST/Interp/FunctionPointer.h
+++ b/clang/lib/AST/Interp/FunctionPointer.h
@@ -32,6 +32,7 @@ class FunctionPointer final {
 
   const Function *getFunction() const { return Func; }
   bool isZero() const { return !Func; }
+  bool isValid() const { return Valid; }
   bool isWeak() const {
     if (!Func || !Valid)
       return false;
diff --git a/clang/lib/AST/Interp/Interp.h b/clang/lib/AST/Interp/Interp.h
index 4182254357eb9..dd0bacd73acb1 100644
--- a/clang/lib/AST/Interp/Interp.h
+++ b/clang/lib/AST/Interp/Interp.h
@@ -2236,6 +2236,10 @@ inline bool CallPtr(InterpState &S, CodePtr OpPC, uint32_t ArgSize,
         << const_cast<Expr *>(E) << E->getSourceRange();
     return false;
   }
+
+  if (!FuncPtr.isValid())
+    return false;
+
   assert(F);
 
   // Check argument nullability state.
diff --git a/clang/test/AST/Interp/functions.cpp b/clang/test/AST/Interp/functions.cpp
index 4fb3c816000ab..f9bb5d53634e0 100644
--- a/clang/test/AST/Interp/functions.cpp
+++ b/clang/test/AST/Interp/functions.cpp
@@ -584,9 +584,20 @@ namespace VariadicOperator {
 namespace WeakCompare {
   [[gnu::weak]]void weak_method();
   static_assert(weak_method != nullptr, ""); // both-error {{not an integral constant expression}} \
-                                         // both-note {{comparison against address of weak declaration '&weak_method' can only be performed at runtim}}
+                                             // both-note {{comparison against address of weak declaration '&weak_method' can only be performed at runtim}}
 
   constexpr auto A = &weak_method;
   static_assert(A != nullptr, ""); // both-error {{not an integral constant expression}} \
-                               // both-note {{comparison against address of weak declaration '&weak_method' can only be performed at runtim}}
+                                   // both-note {{comparison against address of weak declaration '&weak_method' can only be performed at runtim}}
+}
+
+namespace FromIntegral {
+#if __cplusplus >= 202002L
+  typedef double (*DoubleFn)();
+  int a[(int)DoubleFn((void*)-1)()]; // both-error {{not allowed at file scope}} \
+                                    // both-warning {{variable length arrays}}
+  int b[(int)DoubleFn((void*)(-1 + 1))()]; // both-error {{not allowed at file scope}} \
+                                           // expected-note {{evaluates to a null function pointer}} \
+                                           // both-warning {{variable length arrays}}
+#endif
 }

From b0194d2894db49d7cf4d36aed87952c3e0c6a390 Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang@intel.com>
Date: Tue, 16 Apr 2024 18:19:50 +0800
Subject: [PATCH 559/886] [SEH] Ignore async exception flag when the
 environment is not MSVC (#88101)

Fixes #62449
---
 clang/lib/Driver/ToolChains/Clang.cpp         | 28 +++++++++++++------
 .../test/Driver/windows-seh-async-verify.cpp  | 24 ++++++++++++++++
 2 files changed, 44 insertions(+), 8 deletions(-)
 create mode 100644 clang/test/Driver/windows-seh-async-verify.cpp

diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 6d52eced10429..096ed14f95704 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -346,11 +346,14 @@ static bool addExceptionArgs(const ArgList &Args, types::ID InputType,
   bool EH = Args.hasFlag(options::OPT_fexceptions, options::OPT_fno_exceptions,
                          false);
 
-  bool EHa = Args.hasFlag(options::OPT_fasync_exceptions,
-                          options::OPT_fno_async_exceptions, false);
-  if (EHa) {
-    CmdArgs.push_back("-fasync-exceptions");
-    EH = true;
+  // Async exceptions are Windows MSVC only.
+  if (Triple.isWindowsMSVCEnvironment()) {
+    bool EHa = Args.hasFlag(options::OPT_fasync_exceptions,
+                            options::OPT_fno_async_exceptions, false);
+    if (EHa) {
+      CmdArgs.push_back("-fasync-exceptions");
+      EH = true;
+    }
   }
 
   // Obj-C exceptions are enabled by default, regardless of -fexceptions. This
@@ -8102,7 +8105,8 @@ struct EHFlags {
 ///      The 'a' modifier is unimplemented and fundamentally hard in LLVM IR.
 /// - c: Assume that extern "C" functions are implicitly nounwind.
 /// The default is /EHs-c-, meaning cleanups are disabled.
-static EHFlags parseClangCLEHFlags(const Driver &D, const ArgList &Args) {
+static EHFlags parseClangCLEHFlags(const Driver &D, const ArgList &Args,
+                                   bool isWindowsMSVC) {
   EHFlags EH;
 
   std::vector<std::string> EHArgs =
@@ -8112,8 +8116,15 @@ static EHFlags parseClangCLEHFlags(const Driver &D, const ArgList &Args) {
       switch (EHVal[I]) {
       case 'a':
         EH.Asynch = maybeConsumeDash(EHVal, I);
-        if (EH.Asynch)
+        if (EH.Asynch) {
+          // Async exceptions are Windows MSVC only.
+          if (!isWindowsMSVC) {
+            EH.Asynch = false;
+            D.Diag(clang::diag::warn_drv_unused_argument) << "/EHa" << EHVal;
+            continue;
+          }
           EH.Synch = false;
+        }
         continue;
       case 'c':
         EH.NoUnwindC = maybeConsumeDash(EHVal, I);
@@ -8177,7 +8188,8 @@ void Clang::AddClangCLArgs(const ArgList &Args, types::ID InputType,
 
   const Driver &D = getToolChain().getDriver();
 
-  EHFlags EH = parseClangCLEHFlags(D, Args);
+  bool IsWindowsMSVC = getToolChain().getTriple().isWindowsMSVCEnvironment();
+  EHFlags EH = parseClangCLEHFlags(D, Args, IsWindowsMSVC);
   if (!isNVPTX && (EH.Synch || EH.Asynch)) {
     if (types::isCXX(InputType))
       CmdArgs.push_back("-fcxx-exceptions");
diff --git a/clang/test/Driver/windows-seh-async-verify.cpp b/clang/test/Driver/windows-seh-async-verify.cpp
new file mode 100644
index 0000000000000..5fda6a77dba04
--- /dev/null
+++ b/clang/test/Driver/windows-seh-async-verify.cpp
@@ -0,0 +1,24 @@
+// RUN: %clang --target=x86_64-pc-windows -fasync-exceptions -fsyntax-only %s -### 2>&1 | FileCheck %s
+// RUN: %clang_cl --target=x86_64-pc-windows /EHa -fsyntax-only %s -### 2>&1 | FileCheck %s
+// RUN: %clang --target=x86_64-pc-windows-gnu -fasync-exceptions -fsyntax-only %s -### 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,GNU
+// RUN: %clang_cl --target=x86_64-pc-windows-gnu /EHa -fsyntax-only %s -### 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,CL-GNU
+
+// CHECK-NOT: warning
+// GNU: warning: argument unused during compilation: '-fasync-exceptions' [-Wunused-command-line-argument]
+// CL-GNU: warning: argument unused during compilation: '/EHa' [-Wunused-command-line-argument]
+
+// CHECK: -fasync-exceptions
+// GNU-ALL-NOT: -fasync-exceptions
+struct S {
+    union _Un {
+        ~_Un() {}
+        char _Buf[12];
+    };
+    _Un _un;
+};
+
+struct Embed {
+    S v2;
+};
+
+void PR62449() { Embed v{}; }

From f4960da6023b8034ae68925c3223d51624621b37 Mon Sep 17 00:00:00 2001
From: Jon Chesterfield <jonathanchesterfield@gmail.com>
Date: Tue, 16 Apr 2024 11:24:45 +0100
Subject: [PATCH 560/886] Revert "[Verifier] Reject va_start in non-variadic
 function (#88809)"

This reverts commit 61717c1aa1f08eb57839a21fb2d9004739022e0d. Failed a MLIR test
---
 llvm/lib/Analysis/Lint.cpp                      | 5 ++++-
 llvm/lib/IR/Verifier.cpp                        | 5 -----
 llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll | 2 +-
 llvm/test/Other/lint.ll                         | 7 +++++++
 llvm/test/Verifier/variadic.ll                  | 8 --------
 5 files changed, 12 insertions(+), 15 deletions(-)
 delete mode 100644 llvm/test/Verifier/variadic.ll

diff --git a/llvm/lib/Analysis/Lint.cpp b/llvm/lib/Analysis/Lint.cpp
index 1ab856ac8830a..0694c2995dfcc 100644
--- a/llvm/lib/Analysis/Lint.cpp
+++ b/llvm/lib/Analysis/Lint.cpp
@@ -350,7 +350,10 @@ void Lint::visitCallBase(CallBase &I) {
     }
 
     case Intrinsic::vastart:
-      // vastart in non-varargs function is rejected by the verifier
+      Check(I.getParent()->getParent()->isVarArg(),
+            "Undefined behavior: va_start called in a non-varargs function",
+            &I);
+
       visitMemoryReference(I, MemoryLocation::getForArgument(&I, 0, TLI),
                            std::nullopt, nullptr, MemRef::Read | MemRef::Write);
       break;
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 4cd61e6e531bf..516d4a0515569 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -5798,11 +5798,6 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
 
     break;
   }
-  case Intrinsic::vastart: {
-    Check(Call.getFunction()->isVarArg(),
-          "va_start called in a non-varargs function");
-    break;
-  }
   case Intrinsic::vector_reduce_and:
   case Intrinsic::vector_reduce_or:
   case Intrinsic::vector_reduce_xor:
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
index 8c6e01d934c2d..bd576d0f70e9c 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
@@ -3,7 +3,7 @@
 
 
 declare void @llvm.va_start(ptr)
-define void @test_va_start(ptr %list, ...) {
+define void @test_va_start(ptr %list) {
 ; CHECK-LABEL: name: test_va_start
 ; CHECK: [[LIST:%[0-9]+]]:_(p0) = COPY $x0
 ; CHECK-IOS: G_VASTART [[LIST]](p0) :: (store (s64) into %ir.list, align 1)
diff --git a/llvm/test/Other/lint.ll b/llvm/test/Other/lint.ll
index 6fd2d40cd2f29..6b31b31a78c98 100644
--- a/llvm/test/Other/lint.ll
+++ b/llvm/test/Other/lint.ll
@@ -124,6 +124,13 @@ define void @0() nounwind {
   ret void
 }
 
+; CHECK: va_start called in a non-varargs function
+declare void @llvm.va_start(ptr)
+define void @not_vararg(ptr %p) nounwind {
+  call void @llvm.va_start(ptr %p)
+  ret void
+}
+
 ; CHECK: Undefined behavior: Branch to non-blockaddress
 define void @use_indbr() {
   indirectbr ptr @foo, [label %block]
diff --git a/llvm/test/Verifier/variadic.ll b/llvm/test/Verifier/variadic.ll
deleted file mode 100644
index 55e4a4da0a920..0000000000000
--- a/llvm/test/Verifier/variadic.ll
+++ /dev/null
@@ -1,8 +0,0 @@
-; RUN: not opt -S -passes=verify 2>&1 < %s | FileCheck %s
-
-; CHECK: va_start called in a non-varargs function
-declare void @llvm.va_start(ptr)
-define void @not_vararg(ptr %p) nounwind {
-  call void @llvm.va_start(ptr %p)
-  ret void
-}

From 01f79899ba349a0200586c8d05f5e22cca2ced31 Mon Sep 17 00:00:00 2001
From: Jinyang He <hejinyang@loongson.cn>
Date: Tue, 16 Apr 2024 18:31:03 +0800
Subject: [PATCH 561/886] [LoongArch] Use R_LARCH_ALIGN with section symbol
 (#84741)

In LoongArch psABI v2.30, the R_LARCH_ALIGN requires symbol index to
support the third parameter of alignment directive. Create symbol for
each section is redundant because they have section symbol which can
also be used as symbol index. So use section symbol directly for
R_LARCH_ALIGN.
---
 lld/ELF/InputSection.cpp                      |  6 +++-
 lld/test/ELF/loongarch-relax-align-ldr.s      | 28 +++++++++++++++++++
 lld/test/ELF/loongarch-relax-emit-relocs.s    |  5 ++--
 .../MCTargetDesc/LoongArchAsmBackend.cpp      |  7 ++---
 .../MC/LoongArch/Relocations/relax-addsub.s   |  2 +-
 .../MC/LoongArch/Relocations/relax-align.s    | 14 ++++++----
 6 files changed, 47 insertions(+), 15 deletions(-)
 create mode 100644 lld/test/ELF/loongarch-relax-align-ldr.s

diff --git a/lld/ELF/InputSection.cpp b/lld/ELF/InputSection.cpp
index c8350652e65a6..fa48552b8f7a1 100644
--- a/lld/ELF/InputSection.cpp
+++ b/lld/ELF/InputSection.cpp
@@ -464,7 +464,11 @@ void InputSection::copyRelocations(uint8_t *buf,
         addend += sec->getFile<ELFT>()->mipsGp0;
       }
 
-      if (RelTy::IsRela)
+      if (config->emachine == EM_LOONGARCH && type == R_LARCH_ALIGN)
+        // LoongArch psABI v2.30, the R_LARCH_ALIGN requires symbol index.
+        // If it use the section symbol, the addend should not be changed.
+        p->r_addend = addend;
+      else if (RelTy::IsRela)
         p->r_addend = sym.getVA(addend) - section->getOutputSection()->addr;
       // For SHF_ALLOC sections relocated by REL, append a relocation to
       // sec->relocations so that relocateAlloc transitively called by
diff --git a/lld/test/ELF/loongarch-relax-align-ldr.s b/lld/test/ELF/loongarch-relax-align-ldr.s
new file mode 100644
index 0000000000000..6534dc906cfd0
--- /dev/null
+++ b/lld/test/ELF/loongarch-relax-align-ldr.s
@@ -0,0 +1,28 @@
+# REQUIRES: loongarch
+## Test `ld -r` not changes the addend of R_LARCH_ALIGN.
+
+# RUN: llvm-mc --filetype=obj --triple=loongarch64 --mattr=+relax %s -o %t.64.o
+# RUN: ld.lld -r %t.64.o %t.64.o -o %t.64.r
+# RUN: llvm-objdump -dr --no-show-raw-insn %t.64.r | FileCheck %s
+
+# CHECK:      <.text>:
+# CHECK-NEXT:   break 1
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   {{0*}}04:  R_LARCH_ALIGN        .text+0x804
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   break 2
+# CHECK-NEXT:   break 0
+# CHECK-NEXT:   break 0
+# CHECK-NEXT:   break 0
+# CHECK-NEXT:   break 1
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   {{0*}}24:  R_LARCH_ALIGN        .text+0x804
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   nop
+# CHECK-NEXT:   break 2
+
+.text
+break 1
+.p2align 4, , 8
+break 2
diff --git a/lld/test/ELF/loongarch-relax-emit-relocs.s b/lld/test/ELF/loongarch-relax-emit-relocs.s
index 581fce8c95caa..9007f8fcc114f 100644
--- a/lld/test/ELF/loongarch-relax-emit-relocs.s
+++ b/lld/test/ELF/loongarch-relax-emit-relocs.s
@@ -25,7 +25,7 @@
 # CHECK-NEXT:     R_LARCH_PCALA_LO12 _start
 # CHECK-NEXT:     R_LARCH_RELAX *ABS*
 # CHECK-NEXT:   nop
-# CHECK-NEXT:     R_LARCH_ALIGN .Lla-relax-align0+0x4
+# CHECK-NEXT:     R_LARCH_ALIGN .text+0x4
 # CHECK-NEXT:   nop
 # CHECK-NEXT:   ret
 
@@ -37,11 +37,12 @@
 # CHECKR-NEXT:     R_LARCH_PCALA_LO12 _start
 # CHECKR-NEXT:     R_LARCH_RELAX *ABS*
 # CHECKR-NEXT:   nop
-# CHECKR-NEXT:     R_LARCH_ALIGN .Lla-relax-align0+0x4
+# CHECKR-NEXT:     R_LARCH_ALIGN .text+0x4
 # CHECKR-NEXT:   nop
 # CHECKR-NEXT:   nop
 # CHECKR-NEXT:   ret
 
+.text
 .global _start
 _start:
   la.pcrel $a0, _start
diff --git a/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp b/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
index de492f2b1f0a4..98f5014a34b1d 100644
--- a/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
+++ b/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
@@ -226,11 +226,8 @@ bool LoongArchAsmBackend::shouldInsertFixupForCodeAlign(
       MCFixup::create(0, Dummy, MCFixupKind(LoongArch::fixup_loongarch_align));
   const MCSymbolRefExpr *MCSym = getSecToAlignSym()[Sec];
   if (MCSym == nullptr) {
-    // Create a symbol and make the value of symbol is zero.
-    MCSymbol *Sym = Ctx.createNamedTempSymbol("la-relax-align");
-    Sym->setFragment(&*Sec->getBeginSymbol()->getFragment());
-    Asm.registerSymbol(*Sym);
-    MCSym = MCSymbolRefExpr::create(Sym, Ctx);
+    // Use section symbol directly.
+    MCSym = MCSymbolRefExpr::create(Sec->getBeginSymbol(), Ctx);
     getSecToAlignSym()[Sec] = MCSym;
   }
 
diff --git a/llvm/test/MC/LoongArch/Relocations/relax-addsub.s b/llvm/test/MC/LoongArch/Relocations/relax-addsub.s
index 18e0ede5e2937..0e27d6301bb3c 100644
--- a/llvm/test/MC/LoongArch/Relocations/relax-addsub.s
+++ b/llvm/test/MC/LoongArch/Relocations/relax-addsub.s
@@ -28,7 +28,7 @@
 
 # RELAX:       Relocations [
 # RELAX-NEXT:    Section ({{.*}}) .rela.text {
-# RELAX-NEXT:      0x4 R_LARCH_ALIGN {{.*}} 0x4
+# RELAX-NEXT:      0x4 R_LARCH_ALIGN .text 0x4
 # RELAX-NEXT:      0x10 R_LARCH_PCALA_HI20 .L1 0x0
 # RELAX-NEXT:      0x10 R_LARCH_RELAX - 0x0
 # RELAX-NEXT:      0x14 R_LARCH_PCALA_LO12 .L1 0x0
diff --git a/llvm/test/MC/LoongArch/Relocations/relax-align.s b/llvm/test/MC/LoongArch/Relocations/relax-align.s
index 294fd9fb916c7..0246d5b46431c 100644
--- a/llvm/test/MC/LoongArch/Relocations/relax-align.s
+++ b/llvm/test/MC/LoongArch/Relocations/relax-align.s
@@ -63,17 +63,19 @@ ret
 ## Test the symbol index is different from .text.
 .section .text2, "ax"
 .p2align 4
+.p2align 4, , 4
 break 7
 
 # RELOC:            Relocations [
 # RELAX-RELOC-NEXT:   Section ({{.*}}) .rela.text {
-# RELAX-RELOC-NEXT:     0x24 R_LARCH_ALIGN .Lla-relax-align0 0x4
-# RELAX-RELOC-NEXT:     0x34 R_LARCH_ALIGN .Lla-relax-align0 0x5
-# RELAX-RELOC-NEXT:     0x50 R_LARCH_ALIGN .Lla-relax-align0 0x4
-# RELAX-RELOC-NEXT:     0x60 R_LARCH_ALIGN .Lla-relax-align0 0xB04
-# RELAX-RELOC-NEXT:     0x70 R_LARCH_ALIGN .Lla-relax-align0 0x4
+# RELAX-RELOC-NEXT:     0x24 R_LARCH_ALIGN .text 0x4
+# RELAX-RELOC-NEXT:     0x34 R_LARCH_ALIGN .text 0x5
+# RELAX-RELOC-NEXT:     0x50 R_LARCH_ALIGN .text 0x4
+# RELAX-RELOC-NEXT:     0x60 R_LARCH_ALIGN .text 0xB04
+# RELAX-RELOC-NEXT:     0x70 R_LARCH_ALIGN .text 0x4
 # RELAX-RELOC-NEXT:   }
 # RELAX-RELOC-NEXT:   Section ({{.*}}) .rela.text2 {
-# RELAX-RELOC-NEXT:     0x0 R_LARCH_ALIGN .Lla-relax-align1 0x4
+# RELAX-RELOC-NEXT:     0x0 R_LARCH_ALIGN .text2 0x4
+# RELAX-RELOC-NEXT:     0xC R_LARCH_ALIGN .text2 0x404
 # RELAX-RELOC-NEXT:   }
 # RELOC-NEXT:       ]

From c09384e2b419c7b4e4167e0d0295d9018cc6169c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Mon, 15 Apr 2024 15:11:23 +0200
Subject: [PATCH 562/886] [clang][Interp] Support MemberExprs pointing to
 VarDecls

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 11 +++++++++--
 clang/test/AST/Interp/records.cpp        |  8 ++++++++
 2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 4a7b40440770e..a069f3ec27e72 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1267,10 +1267,19 @@ template <class Emitter>
 bool ByteCodeExprGen<Emitter>::VisitMemberExpr(const MemberExpr *E) {
   // 'Base.Member'
   const Expr *Base = E->getBase();
+  const ValueDecl *Member = E->getMemberDecl();
 
   if (DiscardResult)
     return this->discard(Base);
 
+  if (const auto *VD = dyn_cast<VarDecl>(Member)) {
+    // I am almost confident in saying that a var decl must be static
+    // and therefore registered as a global variable. But this will probably
+    // turn out to be wrong some time in the future, as always.
+    if (auto GlobalIndex = P.getGlobal(VD))
+      return this->emitGetPtrGlobal(*GlobalIndex, E);
+  }
+
   if (Initializing) {
     if (!this->delegate(Base))
       return false;
@@ -1280,8 +1289,6 @@ bool ByteCodeExprGen<Emitter>::VisitMemberExpr(const MemberExpr *E) {
   }
 
   // Base above gives us a pointer on the stack.
-  // TODO: Implement non-FieldDecl members.
-  const ValueDecl *Member = E->getMemberDecl();
   if (const auto *FD = dyn_cast<FieldDecl>(Member)) {
     const RecordDecl *RD = FD->getParent();
     const Record *R = getRecord(RD);
diff --git a/clang/test/AST/Interp/records.cpp b/clang/test/AST/Interp/records.cpp
index f251497ed7018..2c33fa1bf8843 100644
--- a/clang/test/AST/Interp/records.cpp
+++ b/clang/test/AST/Interp/records.cpp
@@ -1309,3 +1309,11 @@ namespace pr18633 {
     func2<int>();
   }
 }
+
+namespace {
+  struct F {
+    static constexpr int Z = 12;
+  };
+  F f;
+  static_assert(f.Z == 12, "");
+}

From 1120d8e6f799121b611aa23bdc128e40cf9c6c58 Mon Sep 17 00:00:00 2001
From: Alex Voicu <alexandru.voicu@amd.com>
Date: Tue, 16 Apr 2024 13:37:29 +0300
Subject: [PATCH 563/886] [clang][CodeGen] Add AS for Globals to SPIR & SPIRV
 datalayouts (#88455)

Currently neither the SPIR nor the SPIRV targets specify the AS for
globals in their datalayout strings. This is problematic because
CodeGen/LLVM will default to AS0 in this case, which produces Globals
that end up in the private address space for e.g. OCL, HIPSPV or SYCL.
This patch addresses it by completing the datalayout string.
---
 clang/lib/Basic/Targets/SPIR.h                |  8 +++---
 clang/test/CodeGen/target-data.c              |  4 +--
 llvm/lib/IR/AutoUpgrade.cpp                   |  9 ++++---
 llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp  |  4 +--
 .../Bitcode/DataLayoutUpgradeTest.cpp         | 27 +++++++++++++++++++
 5 files changed, 40 insertions(+), 12 deletions(-)

diff --git a/clang/lib/Basic/Targets/SPIR.h b/clang/lib/Basic/Targets/SPIR.h
index e25991e3dfe82..9a4a8b501460b 100644
--- a/clang/lib/Basic/Targets/SPIR.h
+++ b/clang/lib/Basic/Targets/SPIR.h
@@ -259,7 +259,7 @@ class LLVM_LIBRARY_VISIBILITY SPIR32TargetInfo : public SPIRTargetInfo {
     SizeType = TargetInfo::UnsignedInt;
     PtrDiffType = IntPtrType = TargetInfo::SignedInt;
     resetDataLayout("e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-"
-                    "v96:128-v192:256-v256:256-v512:512-v1024:1024");
+                    "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1");
   }
 
   void getTargetDefines(const LangOptions &Opts,
@@ -276,7 +276,7 @@ class LLVM_LIBRARY_VISIBILITY SPIR64TargetInfo : public SPIRTargetInfo {
     SizeType = TargetInfo::UnsignedLong;
     PtrDiffType = IntPtrType = TargetInfo::SignedLong;
     resetDataLayout("e-i64:64-v16:16-v24:32-v32:32-v48:64-"
-                    "v96:128-v192:256-v256:256-v512:512-v1024:1024");
+                    "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1");
   }
 
   void getTargetDefines(const LangOptions &Opts,
@@ -336,7 +336,7 @@ class LLVM_LIBRARY_VISIBILITY SPIRV32TargetInfo : public BaseSPIRVTargetInfo {
     SizeType = TargetInfo::UnsignedInt;
     PtrDiffType = IntPtrType = TargetInfo::SignedInt;
     resetDataLayout("e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-"
-                    "v96:128-v192:256-v256:256-v512:512-v1024:1024");
+                    "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1");
   }
 
   void getTargetDefines(const LangOptions &Opts,
@@ -357,7 +357,7 @@ class LLVM_LIBRARY_VISIBILITY SPIRV64TargetInfo : public BaseSPIRVTargetInfo {
     SizeType = TargetInfo::UnsignedLong;
     PtrDiffType = IntPtrType = TargetInfo::SignedLong;
     resetDataLayout("e-i64:64-v16:16-v24:32-v32:32-v48:64-"
-                    "v96:128-v192:256-v256:256-v512:512-v1024:1024");
+                    "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1");
   }
 
   void getTargetDefines(const LangOptions &Opts,
diff --git a/clang/test/CodeGen/target-data.c b/clang/test/CodeGen/target-data.c
index acff367d50eb9..c184f314f68f8 100644
--- a/clang/test/CodeGen/target-data.c
+++ b/clang/test/CodeGen/target-data.c
@@ -251,11 +251,11 @@
 
 // RUN: %clang_cc1 -triple spir-unknown -o - -emit-llvm %s | \
 // RUN: FileCheck %s -check-prefix=SPIR
-// SPIR: target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+// SPIR: target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-G1"
 
 // RUN: %clang_cc1 -triple spir64-unknown -o - -emit-llvm %s | \
 // RUN: FileCheck %s -check-prefix=SPIR64
-// SPIR64: target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+// SPIR64: target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-G1"
 
 // RUN: %clang_cc1 -triple bpfel -o - -emit-llvm %s | \
 // RUN: FileCheck %s -check-prefix=BPFEL
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index 2c480fb76ee4d..634b2dd5119e8 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -5341,10 +5341,11 @@ MDNode *llvm::upgradeInstructionLoopAttachment(MDNode &N) {
 
 std::string llvm::UpgradeDataLayoutString(StringRef DL, StringRef TT) {
   Triple T(TT);
-  // The only data layout upgrades needed for pre-GCN are setting the address
-  // space of globals to 1.
-  if (T.isAMDGPU() && !T.isAMDGCN() && !DL.contains("-G") &&
-      !DL.starts_with("G")) {
+  // The only data layout upgrades needed for pre-GCN, SPIR or SPIRV are setting
+  // the address space of globals to 1. This does not apply to SPIRV Logical.
+  if (((T.isAMDGPU() && !T.isAMDGCN()) ||
+       (T.isSPIR() || (T.isSPIRV() && !T.isSPIRVLogical()))) &&
+      !DL.contains("-G") && !DL.starts_with("G")) {
     return DL.empty() ? std::string("G1") : (DL + "-G1").str();
   }
 
diff --git a/llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp b/llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp
index fbf64f2b1dfb1..ae8baa3f11913 100644
--- a/llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp
@@ -55,9 +55,9 @@ static std::string computeDataLayout(const Triple &TT) {
   // mean anything.
   if (Arch == Triple::spirv32)
     return "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-"
-           "v96:128-v192:256-v256:256-v512:512-v1024:1024";
+           "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1";
   return "e-i64:64-v16:16-v24:32-v32:32-v48:64-"
-         "v96:128-v192:256-v256:256-v512:512-v1024:1024";
+         "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1";
 }
 
 static Reloc::Model getEffectiveRelocModel(std::optional<Reloc::Model> RM) {
diff --git a/llvm/unittests/Bitcode/DataLayoutUpgradeTest.cpp b/llvm/unittests/Bitcode/DataLayoutUpgradeTest.cpp
index 4865616e3e2ba..d7e4dba4ac170 100644
--- a/llvm/unittests/Bitcode/DataLayoutUpgradeTest.cpp
+++ b/llvm/unittests/Bitcode/DataLayoutUpgradeTest.cpp
@@ -59,6 +59,14 @@ TEST(DataLayoutUpgradeTest, ValidDataLayoutUpgrade) {
   EXPECT_EQ(UpgradeDataLayoutString("e-m:e-p:64:64-i64:64-i128:128-n64-S128",
                                     "riscv64"),
             "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128");
+
+  // Check that SPIR && SPIRV targets add -G1 if it's not present.
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32", "spir"), "e-p:32:32-G1");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32", "spir64"), "e-p:32:32-G1");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32", "spirv32"), "e-p:32:32-G1");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32", "spirv64"), "e-p:32:32-G1");
+  // but that SPIRV Logical does not.
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32", "spirv"), "e-p:32:32");
 }
 
 TEST(DataLayoutUpgradeTest, NoDataLayoutUpgrade) {
@@ -100,6 +108,17 @@ TEST(DataLayoutUpgradeTest, NoDataLayoutUpgrade) {
             "p7:64:64-G2-e-p:64:64-ni:7:8:9-p8:128:128-p9:192:256:256:32");
   EXPECT_EQ(UpgradeDataLayoutString("e-p:64:64-p7:64:64-G1", "amdgcn"),
             "e-p:64:64-p7:64:64-G1-ni:7:8:9-p8:128:128-p9:192:256:256:32");
+
+  // Check that SPIR & SPIRV targets don't add -G1 if there is already a -G
+  // flag.
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32-G2", "spir"), "e-p:32:32-G2");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32-G2", "spir64"), "e-p:32:32-G2");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32-G2", "spirv32"), "e-p:32:32-G2");
+  EXPECT_EQ(UpgradeDataLayoutString("e-p:32:32-G2", "spirv64"), "e-p:32:32-G2");
+  EXPECT_EQ(UpgradeDataLayoutString("G2", "spir"), "G2");
+  EXPECT_EQ(UpgradeDataLayoutString("G2", "spir64"), "G2");
+  EXPECT_EQ(UpgradeDataLayoutString("G2", "spirv32"), "G2");
+  EXPECT_EQ(UpgradeDataLayoutString("G2", "spirv64"), "G2");
 }
 
 TEST(DataLayoutUpgradeTest, EmptyDataLayout) {
@@ -113,6 +132,14 @@ TEST(DataLayoutUpgradeTest, EmptyDataLayout) {
   EXPECT_EQ(UpgradeDataLayoutString("", "r600"), "G1");
   EXPECT_EQ(UpgradeDataLayoutString("", "amdgcn"),
             "G1-ni:7:8:9-p7:160:256:256:32-p8:128:128-p9:192:256:256:32");
+
+  // Check that SPIR & SPIRV targets add G1 if it's not present.
+  EXPECT_EQ(UpgradeDataLayoutString("", "spir"), "G1");
+  EXPECT_EQ(UpgradeDataLayoutString("", "spir64"), "G1");
+  EXPECT_EQ(UpgradeDataLayoutString("", "spirv32"), "G1");
+  EXPECT_EQ(UpgradeDataLayoutString("", "spirv64"), "G1");
+  // but SPIRV Logical does not.
+  EXPECT_EQ(UpgradeDataLayoutString("", "spirv"), "");
 }
 
 } // end namespace

From 36b3c26451bf9a42f0b6b415993d3942bb73abdd Mon Sep 17 00:00:00 2001
From: XChy <xxs_chy@outlook.com>
Date: Tue, 16 Apr 2024 18:51:03 +0800
Subject: [PATCH 564/886] [JumpThreading] Thread over BB with only an
 unconditional branch (#86312)

Fixes #76609
This patch does:
- relax the phis constraint in `CanRedirectPredsOfEmptyBBToSucc`
- guarantee the BB has multiple different predecessors to redirect, so
that we can handle the case without phis in BB. Without this change and
phi constraint, we may redirect the CommonPred.

The motivation is consistent with JumpThreading. We always want the
branch to jump more direct to the destination, without passing the
middle block. In this way, we can expose more other optimization
opportunities.

An obivous example proposed by @dtcxzyw is like:
```llvm
define i32 @test(...) {
entry:
   br i1 %c, label %do.end, label %if.then

if.then:                                          ; preds = %entry
   %call2 = call i32 @dummy()
   %tobool3.not = icmp eq i32 %call2, 0
   br i1 %tobool3.not, label %do.end, label %return

do.end:                                           ; preds = %entry, %if.then
   br label %return

return:                                           ; preds = %if.then, %do.end
   %retval.0 = phi i32 [ 0, %do.end ], [ %call2, %if.then ]
   ret i32 %retval.0
}
```
`entry` can directly jump to return, without passing `do.end`, and then
the if-else pattern can be simplified further:
```llvm
define i32 @test(...) {
entry:
   br i1 %c, label %return, label %if.then

if.then:                                          ; preds = %entry
   %call2 = call i32 @dummy()
   br label %return

return:                                           ; preds = %if.then
   %retval.0 = phi i32 [ 0, %entry ], [ %call2, %if.then ]
   ret i32 %retval.0
}
```
---
 llvm/lib/Transforms/Utils/Local.cpp           |  10 +-
 llvm/test/CodeGen/AArch64/and-sink.ll         |   9 +-
 .../AArch64/combine-comparisons-by-cse.ll     | 122 +++++++----------
 llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll   |  18 +--
 llvm/test/Transforms/JumpThreading/pr79175.ll |   8 +-
 llvm/test/Transforms/JumpThreading/select.ll  |  50 +++----
 .../Transforms/JumpThreading/thread-prob-7.ll |   8 +-
 .../Transforms/JumpThreading/uncond-no-phi.ll | 123 ++++++++++++++++++
 .../PhaseOrdering/thread-uncond-bb.ll         |  62 +++++++++
 9 files changed, 284 insertions(+), 126 deletions(-)
 create mode 100644 llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
 create mode 100644 llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll

diff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp
index a42ef0c4e6ae9..baec51a07fcbf 100644
--- a/llvm/lib/Transforms/Utils/Local.cpp
+++ b/llvm/lib/Transforms/Utils/Local.cpp
@@ -1019,12 +1019,14 @@ CanRedirectPredsOfEmptyBBToSucc(BasicBlock *BB, BasicBlock *Succ,
                                 const SmallPtrSetImpl<BasicBlock *> &SuccPreds,
                                 BasicBlock *&CommonPred) {
 
-  // There must be phis in BB, otherwise BB will be merged into Succ directly
-  if (BB->phis().empty() || Succ->phis().empty())
+  // When Succ has no phis, BB may be merged into Succ directly. We don't need
+  // to redirect the predecessors of BB in this case.
+  if (Succ->phis().empty())
     return false;
 
-  // BB must have predecessors not shared that can be redirected to Succ
-  if (!BB->hasNPredecessorsOrMore(2))
+  // BB must have multiple different predecessors, so that at least one of
+  // predecessors can be redirected to Succ, except the common predecessor.
+  if (BB->getUniquePredecessor() || pred_empty(BB))
     return false;
 
   // Get single common predecessors of both BB and Succ
diff --git a/llvm/test/CodeGen/AArch64/and-sink.ll b/llvm/test/CodeGen/AArch64/and-sink.ll
index f298a55dab721..a57e9d54f3078 100644
--- a/llvm/test/CodeGen/AArch64/and-sink.ll
+++ b/llvm/test/CodeGen/AArch64/and-sink.ll
@@ -11,15 +11,14 @@
 define dso_local i32 @and_sink1(i32 %a, i1 %c) {
 ; CHECK-LABEL: and_sink1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    tbz w1, #0, .LBB0_3
+; CHECK-NEXT:    tbz w1, #0, .LBB0_2
 ; CHECK-NEXT:  // %bb.1: // %bb0
+; CHECK-NEXT:    tst w0, #0x4
 ; CHECK-NEXT:    adrp x8, A
+; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    str wzr, [x8, :lo12:A]
-; CHECK-NEXT:    tbnz w0, #2, .LBB0_3
-; CHECK-NEXT:  // %bb.2:
-; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB0_3: // %bb2
+; CHECK-NEXT:  .LBB0_2:
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 
diff --git a/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll b/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
index 6449c3e11d667..dde3e81833a63 100644
--- a/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
+++ b/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
@@ -13,10 +13,10 @@ define i32 @combine_gt_ge_10() #0 {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
-; CHECK-NEXT:    ldr w8, [x8]
-; CHECK-NEXT:    cmp w8, #10
+; CHECK-NEXT:    ldr w9, [x8]
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
+; CHECK-NEXT:    cmp w9, #10
 ; CHECK-NEXT:    b.le .LBB0_3
 ; CHECK-NEXT:  // %bb.1: // %land.lhs.true
 ; CHECK-NEXT:    adrp x9, :got:c
@@ -29,18 +29,17 @@ define i32 @combine_gt_ge_10() #0 {
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB0_3: // %lor.lhs.false
-; CHECK-NEXT:    b.lt .LBB0_6
+; CHECK-NEXT:    cmp w9, #10
+; CHECK-NEXT:    b.lt .LBB0_5
 ; CHECK-NEXT:  .LBB0_4: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB0_6
-; CHECK-NEXT:  // %bb.5:
-; CHECK-NEXT:    mov w0, #1 // =0x1
+; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB0_6: // %if.end
+; CHECK-NEXT:  .LBB0_5:
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
@@ -145,10 +144,10 @@ define i32 @combine_lt_ge_5() #0 {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
-; CHECK-NEXT:    ldr w8, [x8]
-; CHECK-NEXT:    cmp w8, #5
+; CHECK-NEXT:    ldr w9, [x8]
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
+; CHECK-NEXT:    cmp w9, #5
 ; CHECK-NEXT:    b.ge .LBB2_3
 ; CHECK-NEXT:  // %bb.1: // %land.lhs.true
 ; CHECK-NEXT:    adrp x9, :got:c
@@ -161,18 +160,17 @@ define i32 @combine_lt_ge_5() #0 {
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB2_3: // %lor.lhs.false
-; CHECK-NEXT:    b.gt .LBB2_6
+; CHECK-NEXT:    cmp w9, #5
+; CHECK-NEXT:    b.gt .LBB2_5
 ; CHECK-NEXT:  .LBB2_4: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB2_6
-; CHECK-NEXT:  // %bb.5:
-; CHECK-NEXT:    mov w0, #1 // =0x1
+; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB2_6: // %if.end
+; CHECK-NEXT:  .LBB2_5:
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
@@ -499,24 +497,17 @@ define i32 @do_nothing_if_resultant_opcodes_would_differ() #0 {
 ; CHECK-NEXT:  // %bb.3: // %while.cond.while.end_crit_edge
 ; CHECK-NEXT:    ldr w8, [x19]
 ; CHECK-NEXT:  .LBB7_4: // %while.end
-; CHECK-NEXT:    cmp w8, #1
-; CHECK-NEXT:    b.gt .LBB7_7
-; CHECK-NEXT:  // %bb.5: // %land.lhs.true
-; CHECK-NEXT:    adrp x8, :got:b
-; CHECK-NEXT:    adrp x9, :got:d
-; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
-; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
-; CHECK-NEXT:    ldr w8, [x8]
-; CHECK-NEXT:    ldr w9, [x9]
-; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB7_7
-; CHECK-NEXT:  // %bb.6:
-; CHECK-NEXT:    mov w0, #123 // =0x7b
-; CHECK-NEXT:    b .LBB7_8
-; CHECK-NEXT:  .LBB7_7: // %if.end
-; CHECK-NEXT:    mov w0, wzr
-; CHECK-NEXT:  .LBB7_8: // %return
+; CHECK-NEXT:    adrp x9, :got:b
+; CHECK-NEXT:    adrp x10, :got:d
+; CHECK-NEXT:    ldr x9, [x9, :got_lo12:b]
+; CHECK-NEXT:    ldr x10, [x10, :got_lo12:d]
 ; CHECK-NEXT:    ldp x20, x19, [sp, #16] // 16-byte Folded Reload
+; CHECK-NEXT:    ldr w9, [x9]
+; CHECK-NEXT:    ldr w10, [x10]
+; CHECK-NEXT:    cmp w9, w10
+; CHECK-NEXT:    ccmp w8, #2, #0, eq
+; CHECK-NEXT:    mov w8, #123 // =0x7b
+; CHECK-NEXT:    csel w0, w8, wzr, lt
 ; CHECK-NEXT:    ldr x30, [sp], #32 // 8-byte Folded Reload
 ; CHECK-NEXT:    .cfi_def_cfa_offset 0
 ; CHECK-NEXT:    .cfi_restore w19
@@ -564,52 +555,42 @@ return:                                           ; preds = %if.end, %land.lhs.t
 define i32 @do_nothing_if_compares_can_not_be_adjusted_to_each_other() #0 {
 ; CHECK-LABEL: do_nothing_if_compares_can_not_be_adjusted_to_each_other:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    stp x30, x19, [sp, #-16]! // 16-byte Folded Spill
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    .cfi_offset w19, -8
-; CHECK-NEXT:    .cfi_offset w30, -16
-; CHECK-NEXT:    .cfi_remember_state
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    cmp w8, #0
-; CHECK-NEXT:    b.gt .LBB8_3
+; CHECK-NEXT:    b.gt .LBB8_4
 ; CHECK-NEXT:  // %bb.1: // %while.body.preheader
+; CHECK-NEXT:    stp x30, x19, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w19, -8
+; CHECK-NEXT:    .cfi_offset w30, -16
 ; CHECK-NEXT:    sub w19, w8, #1
 ; CHECK-NEXT:  .LBB8_2: // %while.body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    bl do_something
 ; CHECK-NEXT:    adds w19, w19, #1
 ; CHECK-NEXT:    b.mi .LBB8_2
-; CHECK-NEXT:  .LBB8_3: // %while.end
-; CHECK-NEXT:    adrp x8, :got:c
-; CHECK-NEXT:    ldr x8, [x8, :got_lo12:c]
-; CHECK-NEXT:    ldr w8, [x8]
-; CHECK-NEXT:    cmn w8, #2
-; CHECK-NEXT:    b.lt .LBB8_6
-; CHECK-NEXT:  // %bb.4: // %land.lhs.true
+; CHECK-NEXT:  // %bb.3:
+; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-NEXT:    .cfi_restore w19
+; CHECK-NEXT:    .cfi_restore w30
+; CHECK-NEXT:  .LBB8_4: // %while.end
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:d
+; CHECK-NEXT:    adrp x10, :got:c
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
+; CHECK-NEXT:    ldr x10, [x10, :got_lo12:c]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
+; CHECK-NEXT:    ldr w10, [x10]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB8_6
-; CHECK-NEXT:  // %bb.5:
-; CHECK-NEXT:    mov w0, #123 // =0x7b
-; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
-; CHECK-NEXT:    .cfi_def_cfa_offset 0
-; CHECK-NEXT:    .cfi_restore w19
-; CHECK-NEXT:    .cfi_restore w30
-; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB8_6: // %if.end
-; CHECK-NEXT:    .cfi_restore_state
-; CHECK-NEXT:    mov w0, wzr
-; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
-; CHECK-NEXT:    .cfi_def_cfa_offset 0
-; CHECK-NEXT:    .cfi_restore w19
-; CHECK-NEXT:    .cfi_restore w30
+; CHECK-NEXT:    mov w8, #-3 // =0xfffffffd
+; CHECK-NEXT:    ccmp w10, w8, #4, eq
+; CHECK-NEXT:    mov w8, #123 // =0x7b
+; CHECK-NEXT:    csel w0, w8, wzr, gt
 ; CHECK-NEXT:    ret
 entry:
   %0 = load i32, ptr @a, align 4
@@ -782,12 +763,14 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    cmp w8, #0
 ; CHECK-NEXT:    csel x9, x0, xzr, gt
 ; CHECK-NEXT:    str x9, [x1]
-; CHECK-NEXT:    b.le .LBB11_2
+; CHECK-NEXT:    b.le .LBB11_3
 ; CHECK-NEXT:  // %bb.1: // %lor.lhs.false
 ; CHECK-NEXT:    cmp w8, #2
-; CHECK-NEXT:    b.ge .LBB11_4
-; CHECK-NEXT:    b .LBB11_6
-; CHECK-NEXT:  .LBB11_2: // %land.lhs.true
+; CHECK-NEXT:    b.ge .LBB11_5
+; CHECK-NEXT:  // %bb.2:
+; CHECK-NEXT:    mov w0, wzr
+; CHECK-NEXT:    ret
+; CHECK-NEXT:  .LBB11_3: // %land.lhs.true
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:c
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
@@ -795,11 +778,11 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB11_4
-; CHECK-NEXT:  // %bb.3:
+; CHECK-NEXT:    b.ne .LBB11_5
+; CHECK-NEXT:  // %bb.4:
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB11_4: // %land.lhs.true3
+; CHECK-NEXT:  .LBB11_5: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
@@ -807,12 +790,7 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB11_6
-; CHECK-NEXT:  // %bb.5:
-; CHECK-NEXT:    mov w0, #1 // =0x1
-; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB11_6: // %if.end
-; CHECK-NEXT:    mov w0, wzr
+; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    ret
 entry:
   %0 = load i32, ptr @a, align 4
diff --git a/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll b/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
index dddc4bd953d7a..c33c81841be65 100644
--- a/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
+++ b/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
@@ -10,12 +10,13 @@ define i32 @fred(ptr %a0) #0 {
 ; CHECK-LABEL: fred:
 ; CHECK:       // %bb.0: // %b0
 ; CHECK-NEXT:    {
-; CHECK-NEXT:     if (p0) jump:nt .LBB0_2
+; CHECK-NEXT:     r1:0 = combine(r0,#0)
+; CHECK-NEXT:     if (p0) jumpr r31
 ; CHECK-NEXT:    }
-; CHECK-NEXT:  // %bb.1: // %b2
+; CHECK-NEXT:  .LBB0_1: // %b2
 ; CHECK-NEXT:    {
 ; CHECK-NEXT:     r3:2 = combine(#0,#0)
-; CHECK-NEXT:     r1:0 = memd(r0+#0)
+; CHECK-NEXT:     r1:0 = memd(r1+#0)
 ; CHECK-NEXT:    }
 ; CHECK-NEXT:    {
 ; CHECK-NEXT:     p0 = vcmph.eq(r1:0,r3:2)
@@ -27,16 +28,7 @@ define i32 @fred(ptr %a0) #0 {
 ; CHECK-NEXT:     r0 = and(r0,#1)
 ; CHECK-NEXT:    }
 ; CHECK-NEXT:    {
-; CHECK-NEXT:     p0 = cmp.eq(r0,#11)
-; CHECK-NEXT:     r0 = #1
-; CHECK-NEXT:    }
-; CHECK-NEXT:    {
-; CHECK-NEXT:     if (p0) r0 = #0
-; CHECK-NEXT:     jumpr r31
-; CHECK-NEXT:    }
-; CHECK-NEXT:  .LBB0_2: // %b14
-; CHECK-NEXT:    {
-; CHECK-NEXT:     r0 = #0
+; CHECK-NEXT:     r0 = !cmp.eq(r0,#11)
 ; CHECK-NEXT:     jumpr r31
 ; CHECK-NEXT:    }
 b0:
diff --git a/llvm/test/Transforms/JumpThreading/pr79175.ll b/llvm/test/Transforms/JumpThreading/pr79175.ll
index 2c7ee0770cdc7..cce30ce079999 100644
--- a/llvm/test/Transforms/JumpThreading/pr79175.ll
+++ b/llvm/test/Transforms/JumpThreading/pr79175.ll
@@ -17,11 +17,11 @@ define i32 @test(i64 %idx, i32 %val) {
 ; CHECK:       cond.end:
 ; CHECK-NEXT:    [[CMP_I:%.*]] = icmp sgt i32 [[VAL]], 0
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[CMP_I]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[COND_END_THREAD]], label [[TMP0:%.*]]
-; CHECK:       cond.end.thread:
-; CHECK-NEXT:    br label [[TMP0]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[TMP0:%.*]], label [[COND_END_THREAD]]
 ; CHECK:       0:
-; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 0, [[COND_END_THREAD]] ], [ [[VAL]], [[COND_END]] ]
+; CHECK-NEXT:    br label [[COND_END_THREAD]]
+; CHECK:       cond.end.thread:
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ [[VAL]], [[COND_END]] ], [ 0, [[TMP0]] ], [ 0, [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[F_IDX:%.*]] = getelementptr inbounds i32, ptr @f, i64 [[IDX]]
 ; CHECK-NEXT:    store i32 [[TMP1]], ptr [[F_IDX]], align 4
 ; CHECK-NEXT:    [[F_RELOAD:%.*]] = load i32, ptr @f, align 4
diff --git a/llvm/test/Transforms/JumpThreading/select.ll b/llvm/test/Transforms/JumpThreading/select.ll
index 4ec55a66bb8ac..27ebf4c25da50 100644
--- a/llvm/test/Transforms/JumpThreading/select.ll
+++ b/llvm/test/Transforms/JumpThreading/select.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals
 ; RUN: opt -S -passes="jump-threading" -debug-only=branch-prob < %s 2>&1 | FileCheck %s
-; RUN: opt -S -passes="require<branch-prob>,jump-threading" -debug-only=branch-prob < %s 2>&1 | FileCheck -check-prefixes=CHECK,CHECK-BPI %s
+; RUN: opt -S -passes="require<branch-prob>,jump-threading" -debug-only=branch-prob  -disable-output < %s 2>&1 | FileCheck -check-prefix=CHECK-BPI %s
 ; REQUIRES: asserts
 
 ; CHECK-BPI-LABEL:  ---- Branch Probability Info : unfold1 ----
@@ -21,7 +21,7 @@ declare void @quux()
 ; booleans where at least one operand is true/false/undef.
 
 ;.
-; CHECK: @[[ANCHOR:[a-zA-Z0-9_$"\\.-]+]] = constant [3 x ptr] [ptr blockaddress(@test_indirectbr, [[L1:%.*]]), ptr inttoptr (i32 1 to ptr), ptr blockaddress(@test_indirectbr, [[L3:%.*]])]
+; CHECK: @anchor = constant [3 x ptr] [ptr blockaddress(@test_indirectbr, %L1), ptr inttoptr (i32 1 to ptr), ptr blockaddress(@test_indirectbr, %L3)]
 ;.
 define void @test_br(i1 %cond, i1 %value) nounwind {
 ; CHECK-LABEL: @test_br(
@@ -66,8 +66,8 @@ define void @test_switch(i1 %cond, i8 %value) nounwind {
 ; CHECK-NEXT:    call void @quux()
 ; CHECK-NEXT:    [[EXPR:%.*]] = select i1 [[COND]], i8 1, i8 [[VALUE:%.*]]
 ; CHECK-NEXT:    switch i8 [[EXPR]], label [[L3:%.*]] [
-; CHECK-NEXT:    i8 1, label [[L1]]
-; CHECK-NEXT:    i8 2, label [[L2:%.*]]
+; CHECK-NEXT:      i8 1, label [[L1]]
+; CHECK-NEXT:      i8 2, label [[L2:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L1:
 ; CHECK-NEXT:    call void @foo()
@@ -192,8 +192,8 @@ define void @test_switch_cmp(i1 %cond, i32 %val, i8 %value) nounwind {
 ; CHECK:       0:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi i8 [ [[VALUE:%.*]], [[L0]] ]
 ; CHECK-NEXT:    switch i8 [[TMP1]], label [[L3:%.*]] [
-; CHECK-NEXT:    i8 1, label [[L1]]
-; CHECK-NEXT:    i8 2, label [[L2:%.*]]
+; CHECK-NEXT:      i8 1, label [[L1]]
+; CHECK-NEXT:      i8 2, label [[L2:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L1:
 ; CHECK-NEXT:    call void @foo()
@@ -237,8 +237,8 @@ define void @test_switch_default(ptr nocapture %status) nounwind {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[STATUS:%.*]], align 4
 ; CHECK-NEXT:    switch i32 [[TMP0]], label [[L2:%.*]] [
-; CHECK-NEXT:    i32 5061, label [[L2_THREAD:%.*]]
-; CHECK-NEXT:    i32 0, label [[L2]]
+; CHECK-NEXT:      i32 5061, label [[L2_THREAD:%.*]]
+; CHECK-NEXT:      i32 0, label [[L2]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L2.thread:
 ; CHECK-NEXT:    store i32 10025, ptr [[STATUS]], align 4
@@ -377,21 +377,21 @@ define i32 @unfold3(i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z, i32 %j) noun
 ; CHECK-NEXT:    br i1 [[CMP_I]], label [[DOTEXIT_THREAD4:%.*]], label [[COND_FALSE_I:%.*]]
 ; CHECK:       cond.false.i:
 ; CHECK-NEXT:    [[CMP4_I:%.*]] = icmp sgt i32 [[U]], [[V]]
-; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD:%.*]], label [[COND_FALSE_6_I:%.*]]
+; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD4]], label [[COND_FALSE_6_I:%.*]]
 ; CHECK:       cond.false.6.i:
 ; CHECK-NEXT:    [[CMP8_I:%.*]] = icmp slt i32 [[W:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    br i1 [[CMP8_I]], label [[DOTEXIT_THREAD4]], label [[COND_FALSE_10_I:%.*]]
 ; CHECK:       cond.false.10.i:
 ; CHECK-NEXT:    [[CMP13_I:%.*]] = icmp sgt i32 [[W]], [[X]]
-; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD]], label [[DOTEXIT:%.*]]
+; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD4]], label [[DOTEXIT:%.*]]
 ; CHECK:       .exit:
 ; CHECK-NEXT:    [[PHITMP:%.*]] = icmp sge i32 [[Y:%.*]], [[Z:%.*]]
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[PHITMP]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD]], label [[DOTEXIT_THREAD4]]
-; CHECK:       .exit.thread:
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD:%.*]], label [[DOTEXIT_THREAD4]]
+; CHECK:       0:
 ; CHECK-NEXT:    br label [[DOTEXIT_THREAD4]]
-; CHECK:       .exit.thread4:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[J]], [[DOTEXIT_THREAD]] ], [ [[ADD3]], [[DOTEXIT]] ], [ [[ADD3]], [[ENTRY:%.*]] ], [ [[ADD3]], [[COND_FALSE_6_I]] ]
+; CHECK:       .exit.thread:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[ADD3]], [[DOTEXIT]] ], [ [[J]], [[DOTEXIT_THREAD]] ], [ [[J]], [[COND_FALSE_I]] ], [ [[J]], [[COND_FALSE_10_I]] ], [ [[ADD3]], [[ENTRY:%.*]] ], [ [[ADD3]], [[COND_FALSE_6_I]] ]
 ; CHECK-NEXT:    ret i32 [[TMP0]]
 ;
 entry:
@@ -430,23 +430,23 @@ define i32 @unfold4(i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z, i32 %j) noun
 ; CHECK-NEXT:    br i1 [[CMP_I]], label [[DOTEXIT_THREAD:%.*]], label [[COND_FALSE_I:%.*]]
 ; CHECK:       cond.false.i:
 ; CHECK-NEXT:    [[CMP4_I:%.*]] = icmp sgt i32 [[U]], [[V]]
-; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD5:%.*]], label [[COND_FALSE_6_I:%.*]]
+; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD]], label [[COND_FALSE_6_I:%.*]]
 ; CHECK:       cond.false.6.i:
 ; CHECK-NEXT:    [[CMP8_I:%.*]] = icmp slt i32 [[W:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    br i1 [[CMP8_I]], label [[DOTEXIT_THREAD]], label [[COND_FALSE_10_I:%.*]]
 ; CHECK:       cond.false.10.i:
 ; CHECK-NEXT:    [[CMP13_I:%.*]] = icmp sgt i32 [[W]], [[X]]
-; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD5]], label [[DOTEXIT:%.*]]
+; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD]], label [[DOTEXIT:%.*]]
 ; CHECK:       .exit:
 ; CHECK-NEXT:    [[CMP19_I:%.*]] = icmp sge i32 [[Y:%.*]], [[Z:%.*]]
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i1 [[CMP19_I]] to i32
 ; CHECK-NEXT:    [[LNOT_I18:%.*]] = icmp eq i32 [[CONV]], 1
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[LNOT_I18]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD]], label [[DOTEXIT_THREAD5]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[TMP1:%.*]], label [[DOTEXIT_THREAD]]
+; CHECK:       0:
+; CHECK-NEXT:    br label [[DOTEXIT_THREAD]]
 ; CHECK:       .exit.thread:
-; CHECK-NEXT:    br label [[DOTEXIT_THREAD5]]
-; CHECK:       .exit.thread5:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[J]], [[DOTEXIT_THREAD]] ], [ [[ADD3]], [[DOTEXIT]] ], [ [[ADD3]], [[COND_FALSE_I]] ], [ [[ADD3]], [[COND_FALSE_10_I]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[ADD3]], [[DOTEXIT]] ], [ [[J]], [[TMP1]] ], [ [[J]], [[ENTRY:%.*]] ], [ [[J]], [[COND_FALSE_6_I]] ], [ [[ADD3]], [[COND_FALSE_I]] ], [ [[ADD3]], [[COND_FALSE_10_I]] ]
 ; CHECK-NEXT:    ret i32 [[TMP0]]
 ;
 entry:
@@ -560,10 +560,10 @@ define void @test_func(ptr nocapture readonly %a, ptr nocapture readonly %b, ptr
 ; CHECK:       if.end:
 ; CHECK-NEXT:    [[LOCAL_VAR_0:%.*]] = phi i32 [ [[TMP1]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    switch i32 [[LOCAL_VAR_0]], label [[SW_DEFAULT]] [
-; CHECK-NEXT:    i32 2, label [[SW_BB]]
-; CHECK-NEXT:    i32 4, label [[SW_BB7]]
-; CHECK-NEXT:    i32 5, label [[SW_BB8:%.*]]
-; CHECK-NEXT:    i32 7, label [[SW_BB9:%.*]]
+; CHECK-NEXT:      i32 2, label [[SW_BB]]
+; CHECK-NEXT:      i32 4, label [[SW_BB7]]
+; CHECK-NEXT:      i32 5, label [[SW_BB8:%.*]]
+; CHECK-NEXT:      i32 7, label [[SW_BB9:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       sw.bb:
 ; CHECK-NEXT:    call void @foo()
@@ -674,3 +674,5 @@ if.end:
 ; CHECK: [[META0:![0-9]+]] = !{!"function_entry_count", i64 1984}
 ; CHECK: [[PROF1]] = !{!"branch_weights", i64 1073741824, i64 3221225472}
 ;.
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; CHECK-BPI: {{.*}}
diff --git a/llvm/test/Transforms/JumpThreading/thread-prob-7.ll b/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
index 8c9d89871d00b..4623a579be48f 100644
--- a/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
+++ b/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
@@ -14,15 +14,15 @@ define i32 @func0(i32 %a0, i32 %a1) !prof !0 {
 ; CHECK-NEXT:    br i1 [[CMP1]], label [[BB_JOIN_THREAD:%.*]], label [[TEST2_FALSE:%.*]], !prof [[PROF2:![0-9]+]]
 ; CHECK:       test2_false:
 ; CHECK-NEXT:    call void @foobar()
-; CHECK-NEXT:    br label [[TMP0:%.*]]
+; CHECK-NEXT:    br label [[BB_JOIN_THREAD]]
 ; CHECK:       bb_join:
 ; CHECK-NEXT:    [[C:%.*]] = phi i1 [ [[CX]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[C]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[BB_JOIN_THREAD]], label [[TMP0]], !prof [[PROF3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[BB_JOIN_THREAD1:%.*]], label [[BB_JOIN_THREAD]], !prof [[PROF3:![0-9]+]]
 ; CHECK:       bb_join.thread:
-; CHECK-NEXT:    br label [[TMP0]]
+; CHECK-NEXT:    br label [[BB_JOIN_THREAD]]
 ; CHECK:       0:
-; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 42, [[BB_JOIN_THREAD]] ], [ 7, [[BB_JOIN]] ], [ 7, [[TEST2_FALSE]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 7, [[BB_JOIN]] ], [ 7, [[TEST2_FALSE]] ], [ 42, [[TEST2]] ], [ 42, [[BB_JOIN_THREAD1]] ]
 ; CHECK-NEXT:    ret i32 [[TMP1]]
 ;
 entry:
diff --git a/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll b/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
new file mode 100644
index 0000000000000..6104e8f8778bc
--- /dev/null
+++ b/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
@@ -0,0 +1,123 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=jump-threading -S < %s | FileCheck %s
+
+define i1 @if_else(i1 %c, i1 %c1) {
+; CHECK-LABEL: define i1 @if_else(
+; CHECK-SAME: i1 [[C:%.*]], i1 [[C1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C]], label [[THEN:%.*]], label [[RETURN:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    call void @dummy()
+; CHECK-NEXT:    br i1 [[C1]], label [[ELSE:%.*]], label [[RETURN]]
+; CHECK:       else:
+; CHECK-NEXT:    br label [[RETURN]]
+; CHECK:       return:
+; CHECK-NEXT:    [[RETVAL_0:%.*]] = phi i1 [ false, [[THEN]] ], [ true, [[ENTRY:%.*]] ], [ true, [[ELSE]] ]
+; CHECK-NEXT:    ret i1 [[RETVAL_0]]
+;
+entry:
+  br i1 %c, label %then, label %else
+
+then:
+  call void @dummy()
+  br i1 %c1, label %else, label %return
+
+else:
+  br label %return
+
+return:
+  %retval.0 = phi i1 [ true, %else ], [ false, %then ]
+  ret i1 %retval.0
+}
+
+define i8 @switch_uncond(i8 %arg) {
+; CHECK-LABEL: define i8 @switch_uncond(
+; CHECK-SAME: i8 [[ARG:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    switch i8 [[ARG]], label [[DEFAULT:%.*]] [
+; CHECK-NEXT:      i8 0, label [[BB1:%.*]]
+; CHECK-NEXT:      i8 1, label [[BB3:%.*]]
+; CHECK-NEXT:      i8 2, label [[BB2:%.*]]
+; CHECK-NEXT:      i8 3, label [[END:%.*]]
+; CHECK-NEXT:    ]
+; CHECK:       default:
+; CHECK-NEXT:    unreachable
+; CHECK:       bb:
+; CHECK-NEXT:    call void @dummy()
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       bb1:
+; CHECK-NEXT:    call void @dummy()
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       bb2:
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i8 [ 1, [[ENTRY:%.*]] ], [ 0, [[BB3]] ], [ 0, [[BB1]] ], [ 0, [[BB2]] ]
+; CHECK-NEXT:    ret i8 [[PHI]]
+;
+entry:
+  switch i8 %arg, label %default [
+  i8 0, label %bb
+  i8 1, label %bb1
+  i8 2, label %bb2
+  i8 3, label %end
+  ]
+
+default:
+  unreachable
+
+bb:
+  call void @dummy()
+  br label %bb2
+
+bb1:
+  call void @dummy()
+  br label %bb2
+
+; Predecessors of %bb2 are %bb and %bb1, they are not identical.
+; So we can thread %bb2.
+bb2:
+  br label %end
+
+end:
+  %phi = phi i8 [ 0, %bb2 ], [ 1, %entry ]
+  ret i8 %phi
+}
+
+define i8 @switch_uncond_fail(i8 %arg) {
+; CHECK-LABEL: define i8 @switch_uncond_fail(
+; CHECK-SAME: i8 [[ARG:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    switch i8 [[ARG]], label [[DEFAULT:%.*]] [
+; CHECK-NEXT:      i8 0, label [[BB:%.*]]
+; CHECK-NEXT:      i8 1, label [[BB]]
+; CHECK-NEXT:      i8 2, label [[END:%.*]]
+; CHECK-NEXT:    ]
+; CHECK:       default:
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       bb:
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i8 [ 0, [[BB]] ], [ 1, [[ENTRY:%.*]] ], [ 2, [[DEFAULT]] ]
+; CHECK-NEXT:    ret i8 [[PHI]]
+;
+entry:
+  switch i8 %arg, label %default [
+  i8 0, label %bb
+  i8 1, label %bb
+  i8 2, label %end
+  ]
+
+default:
+  br label %end
+
+; Predecessor of %bb is only %entry (though there are two in predecessor list),
+; thus it's unthreadable.
+bb:
+  br label %end
+
+end:
+  %phi = phi i8 [ 0, %bb ], [ 1, %entry ], [ 2, %default ]
+  ret i8 %phi
+}
+
+declare void @dummy()
diff --git a/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll b/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll
new file mode 100644
index 0000000000000..17146d7d5987f
--- /dev/null
+++ b/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll
@@ -0,0 +1,62 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -O3 -S | FileCheck %s
+
+define i32 @thread_uncond_bb_cmp(i1 %c, i32 %v) {
+; CHECK-LABEL: define i32 @thread_uncond_bb_cmp(
+; CHECK-SAME: i1 [[C:%.*]], i32 [[V:%.*]]) local_unnamed_addr {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C]], label [[DO_END:%.*]], label [[IF_THEN:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    tail call void @dummy()
+; CHECK-NEXT:    br label [[DO_END]]
+; CHECK:       return:
+; CHECK-NEXT:    [[RETVAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[V]], [[IF_THEN]] ]
+; CHECK-NEXT:    ret i32 [[RETVAL]]
+;
+entry:
+  br i1 %c, label %do.end, label %if.then
+
+if.then:                                          ; preds = %entry
+  call void @dummy()
+  %tobool = icmp eq i32 %v, 0
+  br i1 %tobool, label %do.end, label %return
+
+do.end:                                           ; preds = %entry, %if.then
+  br label %return
+
+return:                                           ; preds = %if.then, %do.end
+  %retval = phi i32 [ 0, %do.end ], [ %v, %if.then ]
+  ret i32 %retval
+}
+
+define i32 @thread_uncond_bb_cmp_zext(i1 %c, i32 %v) {
+; CHECK-LABEL: define i32 @thread_uncond_bb_cmp_zext(
+; CHECK-SAME: i1 [[C:%.*]], i32 [[V:%.*]]) local_unnamed_addr {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[C]], label [[DO_END:%.*]], label [[IF_THEN:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    tail call void @dummy()
+; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[V]], 0
+; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = zext i1 [[TOBOOL]] to i32
+; CHECK-NEXT:    br label [[DO_END]]
+; CHECK:       return:
+; CHECK-NEXT:    [[RETVAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
+; CHECK-NEXT:    ret i32 [[RETVAL]]
+;
+entry:
+  br i1 %c, label %do.end, label %if.then
+
+if.then:                                          ; preds = %entry
+  call void @dummy()
+  %tobool = icmp eq i32 %v, 0
+  br i1 %tobool, label %do.end, label %return
+
+do.end:                                           ; preds = %entry, %if.then
+  br label %return
+
+return:                                           ; preds = %if.then, %do.end
+  %retval = phi i32 [ 0, %do.end ], [ 1, %if.then ]
+  ret i32 %retval
+}
+
+declare void @dummy()

From 32b74ca6e41768c91eee8b8ca26235b110a65deb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Mon, 15 Apr 2024 15:41:33 +0200
Subject: [PATCH 565/886] [clang][Interp] Load value from MemberExpr if
 required

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 17 ++++++++++++++---
 clang/test/AST/Interp/cxx03.cpp          | 14 ++++++++++++++
 2 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index a069f3ec27e72..00c4a9f161304 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1272,12 +1272,23 @@ bool ByteCodeExprGen<Emitter>::VisitMemberExpr(const MemberExpr *E) {
   if (DiscardResult)
     return this->discard(Base);
 
+  // MemberExprs are almost always lvalues, in which case we don't need to
+  // do the load. But sometimes they aren't.
+  const auto maybeLoadValue = [&]() -> bool {
+    if (E->isGLValue())
+      return true;
+    if (std::optional<PrimType> T = classify(E))
+      return this->emitLoadPop(*T, E);
+    return false;
+  };
+
   if (const auto *VD = dyn_cast<VarDecl>(Member)) {
     // I am almost confident in saying that a var decl must be static
     // and therefore registered as a global variable. But this will probably
     // turn out to be wrong some time in the future, as always.
     if (auto GlobalIndex = P.getGlobal(VD))
-      return this->emitGetPtrGlobal(*GlobalIndex, E);
+      return this->emitGetPtrGlobal(*GlobalIndex, E) && maybeLoadValue();
+    return false;
   }
 
   if (Initializing) {
@@ -1295,8 +1306,8 @@ bool ByteCodeExprGen<Emitter>::VisitMemberExpr(const MemberExpr *E) {
     const Record::Field *F = R->getField(FD);
     // Leave a pointer to the field on the stack.
     if (F->Decl->getType()->isReferenceType())
-      return this->emitGetFieldPop(PT_Ptr, F->Offset, E);
-    return this->emitGetPtrField(F->Offset, E);
+      return this->emitGetFieldPop(PT_Ptr, F->Offset, E) && maybeLoadValue();
+    return this->emitGetPtrField(F->Offset, E) && maybeLoadValue();
   }
 
   return false;
diff --git a/clang/test/AST/Interp/cxx03.cpp b/clang/test/AST/Interp/cxx03.cpp
index d30cbb2fd7a20..b6aaf0840cfb3 100644
--- a/clang/test/AST/Interp/cxx03.cpp
+++ b/clang/test/AST/Interp/cxx03.cpp
@@ -10,3 +10,17 @@ namespace NonInitializingMemberExpr {
                                                                                               // both-note {{required by}} \
                                                                                               // both-note {{subexpression not valid}}
 }
+
+
+namespace NonLValueMemberExpr {
+  struct PODType {
+    int value;
+  };
+
+#define ATTR __attribute__((require_constant_initialization))
+  struct TT1 {
+    ATTR static const int &subobj_init;
+  };
+
+  const int &TT1::subobj_init = PODType().value;
+}

From 92e96c7bbacbb477265c7e5ff6c49a6de5d4ee69 Mon Sep 17 00:00:00 2001
From: Malay Sanghi <malay.sanghi@intel.com>
Date: Tue, 16 Apr 2024 16:36:17 +0530
Subject: [PATCH 566/886] [X86][GISel] Add DU chain lookups for LOAD & STORE
 (#87453)

For G_LOAD and G_STORE we want this information during regbankselect.
Today we treat load dest as integer and insert converts.

---------

Co-authored-by: Evgenii Kudriashov <evgenii.kudriashov@intel.com>
---
 llvm/include/llvm/CodeGen/GlobalISel/Utils.h  |   4 +
 llvm/lib/CodeGen/GlobalISel/Utils.cpp         |  44 +++++
 .../AArch64/GISel/AArch64RegisterBankInfo.cpp |  37 -----
 llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp |  24 +--
 .../PowerPC/GISel/PPCRegisterBankInfo.cpp     |  39 +----
 .../RISCV/GISel/RISCVRegisterBankInfo.cpp     |  40 -----
 .../Target/X86/GISel/X86RegisterBankInfo.cpp  | 139 +++++++++++++++-
 .../Target/X86/GISel/X86RegisterBankInfo.h    |  16 ++
 llvm/test/CodeGen/X86/GlobalISel/fconstant.ll |  13 +-
 .../regbankselect-sse-intrinsics.ll           | 153 ++++++++++++++++++
 .../X86/GlobalISel/regbankselect-x87.ll       |  29 ++--
 11 files changed, 367 insertions(+), 171 deletions(-)
 create mode 100644 llvm/test/CodeGen/X86/GlobalISel/regbankselect-sse-intrinsics.ll

diff --git a/llvm/include/llvm/CodeGen/GlobalISel/Utils.h b/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
index 807cec3c177d9..c4174cee5e10c 100644
--- a/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
@@ -555,5 +555,9 @@ void eraseInstr(MachineInstr &MI, MachineRegisterInfo &MRI,
 /// debug users of \p MI by writing the effect of \p MI in a DIExpression.
 void salvageDebugInfo(const MachineRegisterInfo &MRI, MachineInstr &MI);
 
+/// Returns whether opcode \p Opc is a pre-isel generic floating-point opcode,
+/// having only floating-point operands.
+bool isPreISelGenericFloatingPointOpcode(unsigned Opc);
+
 } // End namespace llvm.
 #endif
diff --git a/llvm/lib/CodeGen/GlobalISel/Utils.cpp b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
index c3bc3203b6360..ae43e9ccf6112 100644
--- a/llvm/lib/CodeGen/GlobalISel/Utils.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
@@ -1665,3 +1665,47 @@ void llvm::salvageDebugInfo(const MachineRegisterInfo &MRI, MachineInstr &MI) {
     }
   }
 }
+
+bool llvm::isPreISelGenericFloatingPointOpcode(unsigned Opc) {
+  switch (Opc) {
+  case TargetOpcode::G_FABS:
+  case TargetOpcode::G_FADD:
+  case TargetOpcode::G_FCANONICALIZE:
+  case TargetOpcode::G_FCEIL:
+  case TargetOpcode::G_FCONSTANT:
+  case TargetOpcode::G_FCOPYSIGN:
+  case TargetOpcode::G_FCOS:
+  case TargetOpcode::G_FDIV:
+  case TargetOpcode::G_FEXP2:
+  case TargetOpcode::G_FEXP:
+  case TargetOpcode::G_FFLOOR:
+  case TargetOpcode::G_FLOG10:
+  case TargetOpcode::G_FLOG2:
+  case TargetOpcode::G_FLOG:
+  case TargetOpcode::G_FMA:
+  case TargetOpcode::G_FMAD:
+  case TargetOpcode::G_FMAXIMUM:
+  case TargetOpcode::G_FMAXNUM:
+  case TargetOpcode::G_FMAXNUM_IEEE:
+  case TargetOpcode::G_FMINIMUM:
+  case TargetOpcode::G_FMINNUM:
+  case TargetOpcode::G_FMINNUM_IEEE:
+  case TargetOpcode::G_FMUL:
+  case TargetOpcode::G_FNEARBYINT:
+  case TargetOpcode::G_FNEG:
+  case TargetOpcode::G_FPEXT:
+  case TargetOpcode::G_FPOW:
+  case TargetOpcode::G_FPTRUNC:
+  case TargetOpcode::G_FREM:
+  case TargetOpcode::G_FRINT:
+  case TargetOpcode::G_FSIN:
+  case TargetOpcode::G_FSQRT:
+  case TargetOpcode::G_FSUB:
+  case TargetOpcode::G_INTRINSIC_ROUND:
+  case TargetOpcode::G_INTRINSIC_ROUNDEVEN:
+  case TargetOpcode::G_INTRINSIC_TRUNC:
+    return true;
+  default:
+    return false;
+  }
+}
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
index d39de770eaf16..d5c4ce1888e78 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
@@ -424,43 +424,6 @@ void AArch64RegisterBankInfo::applyMappingImpl(
   }
 }
 
-/// Returns whether opcode \p Opc is a pre-isel generic floating-point opcode,
-/// having only floating-point operands.
-static bool isPreISelGenericFloatingPointOpcode(unsigned Opc) {
-  switch (Opc) {
-  case TargetOpcode::G_FADD:
-  case TargetOpcode::G_FSUB:
-  case TargetOpcode::G_FMUL:
-  case TargetOpcode::G_FMA:
-  case TargetOpcode::G_FDIV:
-  case TargetOpcode::G_FCONSTANT:
-  case TargetOpcode::G_FPEXT:
-  case TargetOpcode::G_FPTRUNC:
-  case TargetOpcode::G_FCEIL:
-  case TargetOpcode::G_FFLOOR:
-  case TargetOpcode::G_FNEARBYINT:
-  case TargetOpcode::G_FNEG:
-  case TargetOpcode::G_FCOS:
-  case TargetOpcode::G_FSIN:
-  case TargetOpcode::G_FLOG10:
-  case TargetOpcode::G_FLOG:
-  case TargetOpcode::G_FLOG2:
-  case TargetOpcode::G_FSQRT:
-  case TargetOpcode::G_FABS:
-  case TargetOpcode::G_FEXP:
-  case TargetOpcode::G_FRINT:
-  case TargetOpcode::G_INTRINSIC_TRUNC:
-  case TargetOpcode::G_INTRINSIC_ROUND:
-  case TargetOpcode::G_INTRINSIC_ROUNDEVEN:
-  case TargetOpcode::G_FMAXNUM:
-  case TargetOpcode::G_FMINNUM:
-  case TargetOpcode::G_FMAXIMUM:
-  case TargetOpcode::G_FMINIMUM:
-    return true;
-  }
-  return false;
-}
-
 const RegisterBankInfo::InstructionMapping &
 AArch64RegisterBankInfo::getSameKindOfOperandsMapping(
     const MachineInstr &MI) const {
diff --git a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
index 6af1fd8c88e57..62b58cba9f24a 100644
--- a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
+++ b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
@@ -104,26 +104,6 @@ MipsRegisterBankInfo::getRegBankFromRegClass(const TargetRegisterClass &RC,
   }
 }
 
-// Instructions where all register operands are floating point.
-static bool isFloatingPointOpcode(unsigned Opc) {
-  switch (Opc) {
-  case TargetOpcode::G_FCONSTANT:
-  case TargetOpcode::G_FADD:
-  case TargetOpcode::G_FSUB:
-  case TargetOpcode::G_FMUL:
-  case TargetOpcode::G_FDIV:
-  case TargetOpcode::G_FABS:
-  case TargetOpcode::G_FSQRT:
-  case TargetOpcode::G_FCEIL:
-  case TargetOpcode::G_FFLOOR:
-  case TargetOpcode::G_FPEXT:
-  case TargetOpcode::G_FPTRUNC:
-    return true;
-  default:
-    return false;
-  }
-}
-
 // Instructions where use operands are floating point registers.
 // Def operands are general purpose.
 static bool isFloatingPointOpcodeUse(unsigned Opc) {
@@ -133,7 +113,7 @@ static bool isFloatingPointOpcodeUse(unsigned Opc) {
   case TargetOpcode::G_FCMP:
     return true;
   default:
-    return isFloatingPointOpcode(Opc);
+    return isPreISelGenericFloatingPointOpcode(Opc);
   }
 }
 
@@ -145,7 +125,7 @@ static bool isFloatingPointOpcodeDef(unsigned Opc) {
   case TargetOpcode::G_UITOFP:
     return true;
   default:
-    return isFloatingPointOpcode(Opc);
+    return isPreISelGenericFloatingPointOpcode(Opc);
   }
 }
 
diff --git a/llvm/lib/Target/PowerPC/GISel/PPCRegisterBankInfo.cpp b/llvm/lib/Target/PowerPC/GISel/PPCRegisterBankInfo.cpp
index 6aeef145e3078..125a49de7b27d 100644
--- a/llvm/lib/Target/PowerPC/GISel/PPCRegisterBankInfo.cpp
+++ b/llvm/lib/Target/PowerPC/GISel/PPCRegisterBankInfo.cpp
@@ -13,6 +13,7 @@
 #include "PPCRegisterBankInfo.h"
 #include "PPCRegisterInfo.h"
 #include "llvm/CodeGen/GlobalISel/GenericMachineInstrs.h"
+#include "llvm/CodeGen/GlobalISel/Utils.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/Support/Debug.h"
@@ -239,44 +240,6 @@ PPCRegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   return getInstructionMapping(MappingID, Cost, OperandsMapping, NumOperands);
 }
 
-/// Returns whether opcode \p Opc is a pre-isel generic floating-point opcode,
-/// having only floating-point operands.
-/// FIXME: this is copied from target AArch64. Needs some code refactor here to
-/// put this function in GlobalISel/Utils.cpp.
-static bool isPreISelGenericFloatingPointOpcode(unsigned Opc) {
-  switch (Opc) {
-  case TargetOpcode::G_FADD:
-  case TargetOpcode::G_FSUB:
-  case TargetOpcode::G_FMUL:
-  case TargetOpcode::G_FMA:
-  case TargetOpcode::G_FDIV:
-  case TargetOpcode::G_FCONSTANT:
-  case TargetOpcode::G_FPEXT:
-  case TargetOpcode::G_FPTRUNC:
-  case TargetOpcode::G_FCEIL:
-  case TargetOpcode::G_FFLOOR:
-  case TargetOpcode::G_FNEARBYINT:
-  case TargetOpcode::G_FNEG:
-  case TargetOpcode::G_FCOS:
-  case TargetOpcode::G_FSIN:
-  case TargetOpcode::G_FLOG10:
-  case TargetOpcode::G_FLOG:
-  case TargetOpcode::G_FLOG2:
-  case TargetOpcode::G_FSQRT:
-  case TargetOpcode::G_FABS:
-  case TargetOpcode::G_FEXP:
-  case TargetOpcode::G_FRINT:
-  case TargetOpcode::G_INTRINSIC_TRUNC:
-  case TargetOpcode::G_INTRINSIC_ROUND:
-  case TargetOpcode::G_FMAXNUM:
-  case TargetOpcode::G_FMINNUM:
-  case TargetOpcode::G_FMAXIMUM:
-  case TargetOpcode::G_FMINIMUM:
-    return true;
-  }
-  return false;
-}
-
 /// \returns true if a given intrinsic \p ID only uses and defines FPRs.
 static bool isFPIntrinsic(unsigned ID) {
   // TODO: Add more intrinsics.
diff --git a/llvm/lib/Target/RISCV/GISel/RISCVRegisterBankInfo.cpp b/llvm/lib/Target/RISCV/GISel/RISCVRegisterBankInfo.cpp
index 86e44343b5086..cc534f29685f2 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVRegisterBankInfo.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVRegisterBankInfo.cpp
@@ -154,46 +154,6 @@ static const RegisterBankInfo::ValueMapping *getFPValueMapping(unsigned Size) {
   return &RISCV::ValueMappings[Idx];
 }
 
-/// Returns whether opcode \p Opc is a pre-isel generic floating-point opcode,
-/// having only floating-point operands.
-/// FIXME: this is copied from target AArch64. Needs some code refactor here to
-/// put this function in GlobalISel/Utils.cpp.
-static bool isPreISelGenericFloatingPointOpcode(unsigned Opc) {
-  switch (Opc) {
-  case TargetOpcode::G_FADD:
-  case TargetOpcode::G_FSUB:
-  case TargetOpcode::G_FMUL:
-  case TargetOpcode::G_FMA:
-  case TargetOpcode::G_FDIV:
-  case TargetOpcode::G_FCONSTANT:
-  case TargetOpcode::G_FPEXT:
-  case TargetOpcode::G_FPTRUNC:
-  case TargetOpcode::G_FCEIL:
-  case TargetOpcode::G_FFLOOR:
-  case TargetOpcode::G_FNEARBYINT:
-  case TargetOpcode::G_FNEG:
-  case TargetOpcode::G_FCOPYSIGN:
-  case TargetOpcode::G_FCOS:
-  case TargetOpcode::G_FSIN:
-  case TargetOpcode::G_FLOG10:
-  case TargetOpcode::G_FLOG:
-  case TargetOpcode::G_FLOG2:
-  case TargetOpcode::G_FSQRT:
-  case TargetOpcode::G_FABS:
-  case TargetOpcode::G_FEXP:
-  case TargetOpcode::G_FRINT:
-  case TargetOpcode::G_INTRINSIC_TRUNC:
-  case TargetOpcode::G_INTRINSIC_ROUND:
-  case TargetOpcode::G_INTRINSIC_ROUNDEVEN:
-  case TargetOpcode::G_FMAXNUM:
-  case TargetOpcode::G_FMINNUM:
-  case TargetOpcode::G_FMAXIMUM:
-  case TargetOpcode::G_FMINIMUM:
-    return true;
-  }
-  return false;
-}
-
 // TODO: Make this more like AArch64?
 bool RISCVRegisterBankInfo::hasFPConstraints(
     const MachineInstr &MI, const MachineRegisterInfo &MRI,
diff --git a/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.cpp b/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.cpp
index e7c9e60ba95f1..9e85424e76e62 100644
--- a/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.cpp
+++ b/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.cpp
@@ -13,10 +13,13 @@
 #include "X86RegisterBankInfo.h"
 #include "X86InstrInfo.h"
 #include "X86Subtarget.h"
+#include "llvm/CodeGen/GlobalISel/GenericMachineInstrs.h"
+#include "llvm/CodeGen/GlobalISel/Utils.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RegisterBank.h"
 #include "llvm/CodeGen/RegisterBankInfo.h"
 #include "llvm/CodeGen/TargetRegisterInfo.h"
+#include "llvm/IR/IntrinsicsX86.h"
 
 #define GET_TARGET_REGBANK_IMPL
 #include "X86GenRegisterBank.inc"
@@ -68,6 +71,98 @@ X86RegisterBankInfo::getRegBankFromRegClass(const TargetRegisterClass &RC,
   llvm_unreachable("Unsupported register kind yet.");
 }
 
+// \returns true if a given intrinsic only uses and defines FPRs.
+static bool isFPIntrinsic(const MachineRegisterInfo &MRI,
+                          const MachineInstr &MI) {
+  // TODO: Add more intrinsics.
+  switch (cast<GIntrinsic>(MI).getIntrinsicID()) {
+  default:
+    return false;
+  // SSE1
+  case Intrinsic::x86_sse_rcp_ss:
+  case Intrinsic::x86_sse_rcp_ps:
+  case Intrinsic::x86_sse_rsqrt_ss:
+  case Intrinsic::x86_sse_rsqrt_ps:
+  case Intrinsic::x86_sse_min_ss:
+  case Intrinsic::x86_sse_min_ps:
+  case Intrinsic::x86_sse_max_ss:
+  case Intrinsic::x86_sse_max_ps:
+    return true;
+  }
+  return false;
+}
+
+bool X86RegisterBankInfo::hasFPConstraints(const MachineInstr &MI,
+                                           const MachineRegisterInfo &MRI,
+                                           const TargetRegisterInfo &TRI,
+                                           unsigned Depth) const {
+  unsigned Op = MI.getOpcode();
+  if (Op == TargetOpcode::G_INTRINSIC && isFPIntrinsic(MRI, MI))
+    return true;
+
+  // Do we have an explicit floating point instruction?
+  if (isPreISelGenericFloatingPointOpcode(Op))
+    return true;
+
+  // No. Check if we have a copy-like instruction. If we do, then we could
+  // still be fed by floating point instructions.
+  if (Op != TargetOpcode::COPY && !MI.isPHI() &&
+      !isPreISelGenericOptimizationHint(Op))
+    return false;
+
+  // Check if we already know the register bank.
+  auto *RB = getRegBank(MI.getOperand(0).getReg(), MRI, TRI);
+  if (RB == &getRegBank(X86::PSRRegBankID))
+    return true;
+  if (RB == &getRegBank(X86::GPRRegBankID))
+    return false;
+
+  // We don't know anything.
+  //
+  // If we have a phi, we may be able to infer that it will be assigned a fp
+  // type based off of its inputs.
+  if (!MI.isPHI() || Depth > MaxFPRSearchDepth)
+    return false;
+
+  return any_of(MI.explicit_uses(), [&](const MachineOperand &Op) {
+    return Op.isReg() &&
+           onlyDefinesFP(*MRI.getVRegDef(Op.getReg()), MRI, TRI, Depth + 1);
+  });
+}
+
+bool X86RegisterBankInfo::onlyUsesFP(const MachineInstr &MI,
+                                     const MachineRegisterInfo &MRI,
+                                     const TargetRegisterInfo &TRI,
+                                     unsigned Depth) const {
+  switch (MI.getOpcode()) {
+  case TargetOpcode::G_FPTOSI:
+  case TargetOpcode::G_FPTOUI:
+  case TargetOpcode::G_FCMP:
+  case TargetOpcode::G_LROUND:
+  case TargetOpcode::G_LLROUND:
+  case TargetOpcode::G_INTRINSIC_TRUNC:
+  case TargetOpcode::G_INTRINSIC_ROUND:
+    return true;
+  default:
+    break;
+  }
+  return hasFPConstraints(MI, MRI, TRI, Depth);
+}
+
+bool X86RegisterBankInfo::onlyDefinesFP(const MachineInstr &MI,
+                                        const MachineRegisterInfo &MRI,
+                                        const TargetRegisterInfo &TRI,
+                                        unsigned Depth) const {
+  switch (MI.getOpcode()) {
+  case TargetOpcode::G_SITOFP:
+  case TargetOpcode::G_UITOFP:
+    return true;
+  default:
+    break;
+  }
+  return hasFPConstraints(MI, MRI, TRI, Depth);
+}
+
 X86GenRegisterBankInfo::PartialMappingIdx
 X86GenRegisterBankInfo::getPartialMappingIdx(const MachineInstr &MI,
                                              const LLT &Ty, bool isFP) {
@@ -180,11 +275,13 @@ X86RegisterBankInfo::getSameOperandsMapping(const MachineInstr &MI,
 const RegisterBankInfo::InstructionMapping &
 X86RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   const MachineFunction &MF = *MI.getParent()->getParent();
+  const TargetSubtargetInfo &STI = MF.getSubtarget();
+  const TargetRegisterInfo &TRI = *STI.getRegisterInfo();
   const MachineRegisterInfo &MRI = MF.getRegInfo();
   unsigned Opc = MI.getOpcode();
 
-  // Try the default logic for non-generic instructions that are either copies
-  // or already have some operands assigned to banks.
+  // Try the default logic for non-generic instructions that are either
+  // copies or already have some operands assigned to banks.
   if (!isPreISelGenericOpcode(Opc) || Opc == TargetOpcode::G_PHI) {
     const InstructionMapping &Mapping = getInstrMappingImpl(MI);
     if (Mapping.isValid())
@@ -221,13 +318,14 @@ X86RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   case TargetOpcode::G_FPEXT:
   case TargetOpcode::G_FPTRUNC:
   case TargetOpcode::G_FCONSTANT:
-    // Instruction having only floating-point operands (all scalars in VECRReg)
+    // Instruction having only floating-point operands (all scalars in
+    // VECRReg)
     getInstrPartialMappingIdxs(MI, MRI, /* isFP= */ true, OpRegBankIdx);
     break;
   case TargetOpcode::G_SITOFP:
   case TargetOpcode::G_FPTOSI: {
-    // Some of the floating-point instructions have mixed GPR and FP operands:
-    // fine-tune the computed mapping.
+    // Some of the floating-point instructions have mixed GPR and FP
+    // operands: fine-tune the computed mapping.
     auto &Op0 = MI.getOperand(0);
     auto &Op1 = MI.getOperand(1);
     const LLT Ty0 = MRI.getType(Op0.getReg());
@@ -271,9 +369,36 @@ X86RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
 
     getInstrPartialMappingIdxs(MI, MRI, /* isFP= */ isFPTrunc || isFPAnyExt,
                                OpRegBankIdx);
-  } break;
+    break;
+  }
+  case TargetOpcode::G_LOAD: {
+    // Check if that load feeds fp instructions.
+    // In that case, we want the default mapping to be on FPR
+    // instead of blind map every scalar to GPR.
+    bool IsFP = any_of(MRI.use_nodbg_instructions(cast<GLoad>(MI).getDstReg()),
+                       [&](const MachineInstr &UseMI) {
+                         // If we have at least one direct use in a FP
+                         // instruction, assume this was a floating point load
+                         // in the IR. If it was not, we would have had a
+                         // bitcast before reaching that instruction.
+                         return onlyUsesFP(UseMI, MRI, TRI);
+                       });
+    getInstrPartialMappingIdxs(MI, MRI, IsFP, OpRegBankIdx);
+    break;
+  }
+  case TargetOpcode::G_STORE: {
+    // Check if that store is fed by fp instructions.
+    Register VReg = cast<GStore>(MI).getValueReg();
+    if (!VReg)
+      break;
+    MachineInstr *DefMI = MRI.getVRegDef(VReg);
+    bool IsFP = onlyDefinesFP(*DefMI, MRI, TRI);
+    getInstrPartialMappingIdxs(MI, MRI, IsFP, OpRegBankIdx);
+    break;
+  }
   default:
-    // Track the bank of each register, use NotFP mapping (all scalars in GPRs)
+    // Track the bank of each register, use NotFP mapping (all scalars in
+    // GPRs)
     getInstrPartialMappingIdxs(MI, MRI, /* isFP= */ false, OpRegBankIdx);
     break;
   }
diff --git a/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.h b/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.h
index 989c5956ad591..8f38e717e36b0 100644
--- a/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.h
+++ b/llvm/lib/Target/X86/GISel/X86RegisterBankInfo.h
@@ -62,6 +62,22 @@ class X86RegisterBankInfo final : public X86GenRegisterBankInfo {
                        const SmallVectorImpl<PartialMappingIdx> &OpRegBankIdx,
                        SmallVectorImpl<const ValueMapping *> &OpdsMapping);
 
+  // Maximum recursion depth for hasFPConstraints.
+  const unsigned MaxFPRSearchDepth = 2;
+
+  /// \returns true if \p MI only uses and defines FPRs.
+  bool hasFPConstraints(const MachineInstr &MI, const MachineRegisterInfo &MRI,
+                        const TargetRegisterInfo &TRI,
+                        unsigned Depth = 0) const;
+
+  /// \returns true if \p MI only uses FPRs.
+  bool onlyUsesFP(const MachineInstr &MI, const MachineRegisterInfo &MRI,
+                  const TargetRegisterInfo &TRI, unsigned Depth = 0) const;
+
+  /// \returns true if \p MI only defines FPRs.
+  bool onlyDefinesFP(const MachineInstr &MI, const MachineRegisterInfo &MRI,
+                     const TargetRegisterInfo &TRI, unsigned Depth = 0) const;
+
 public:
   X86RegisterBankInfo(const TargetRegisterInfo &TRI);
 
diff --git a/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll b/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
index a9b2037e9947a..8d2ee3c50f215 100644
--- a/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
+++ b/llvm/test/CodeGen/X86/GlobalISel/fconstant.ll
@@ -10,27 +10,22 @@ define void @test_float(ptr %a , float %b) {
 ; CHECK64_SMALL:       # %bb.0: # %entry
 ; CHECK64_SMALL-NEXT:    movss {{.*#+}} xmm1 = [5.5E+0,0.0E+0,0.0E+0,0.0E+0]
 ; CHECK64_SMALL-NEXT:    addss %xmm0, %xmm1
-; CHECK64_SMALL-NEXT:    movd %xmm1, %eax
-; CHECK64_SMALL-NEXT:    movl %eax, (%rdi)
+; CHECK64_SMALL-NEXT:    movss %xmm1, (%rdi)
 ; CHECK64_SMALL-NEXT:    retq
 ;
 ; CHECK64_LARGE-LABEL: test_float:
 ; CHECK64_LARGE:       # %bb.0: # %entry
 ; CHECK64_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
 ; CHECK64_LARGE-NEXT:    addss (%rax), %xmm0
-; CHECK64_LARGE-NEXT:    movd %xmm0, %eax
-; CHECK64_LARGE-NEXT:    movl %eax, (%rdi)
+; CHECK64_LARGE-NEXT:    movss %xmm0, (%rdi)
 ; CHECK64_LARGE-NEXT:    retq
 ;
 ; CHECK32-LABEL: test_float:
 ; CHECK32:       # %bb.0: # %entry
 ; CHECK32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; CHECK32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; CHECK32-NEXT:    movss {{.*#+}} xmm0 = [5.5E+0,0.0E+0,0.0E+0,0.0E+0]
-; CHECK32-NEXT:    movd %ecx, %xmm1
-; CHECK32-NEXT:    addss %xmm0, %xmm1
-; CHECK32-NEXT:    movd %xmm1, %ecx
-; CHECK32-NEXT:    movl %ecx, (%eax)
+; CHECK32-NEXT:    addss {{[0-9]+}}(%esp), %xmm0
+; CHECK32-NEXT:    movss %xmm0, (%eax)
 ; CHECK32-NEXT:    retl
 entry:
   %aa = fadd float 5.500000e+00, %b
diff --git a/llvm/test/CodeGen/X86/GlobalISel/regbankselect-sse-intrinsics.ll b/llvm/test/CodeGen/X86/GlobalISel/regbankselect-sse-intrinsics.ll
new file mode 100644
index 0000000000000..3388af605d969
--- /dev/null
+++ b/llvm/test/CodeGen/X86/GlobalISel/regbankselect-sse-intrinsics.ll
@@ -0,0 +1,153 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
+; RUN: llc < %s -mtriple=i686-- -mattr=+sse -global-isel -stop-after=regbankselect | FileCheck %s
+
+define void @test_x86_sse_max_ps(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_max_ps
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (p0) from %fixed-stack.0)
+  ; CHECK-NEXT:   [[LOAD2:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[LOAD3:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD1]](p0) :: (load (<4 x s32>) from %ir.p2)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.max.ps), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %a1 = load <4 x float>, ptr %p2, align 16
+  %res = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_max_ss(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_max_ss
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (p0) from %fixed-stack.0)
+  ; CHECK-NEXT:   [[LOAD2:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[LOAD3:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD1]](p0) :: (load (<4 x s32>) from %ir.p2)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.max.ss), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %a1 = load <4 x float>, ptr %p2, align 16
+  %res = call <4 x float> @llvm.x86.sse.max.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_min_ps(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_min_ps
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (p0) from %fixed-stack.0)
+  ; CHECK-NEXT:   [[LOAD2:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[LOAD3:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD1]](p0) :: (load (<4 x s32>) from %ir.p2)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.min.ps), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %a1 = load <4 x float>, ptr %p2, align 16
+  %res = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_min_ss(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_min_ss
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (p0) from %fixed-stack.0)
+  ; CHECK-NEXT:   [[LOAD2:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[LOAD3:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD1]](p0) :: (load (<4 x s32>) from %ir.p2)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.min.ss), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %a1 = load <4 x float>, ptr %p2, align 16
+  %res = call <4 x float> @llvm.x86.sse.min.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_rcp_ps(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_rcp_ps
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.rcp.ps), [[LOAD1]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %res = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.rcp.ps(<4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_rcp_ss(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_rcp_ss
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.rcp.ss), [[LOAD1]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %res = call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.rcp.ss(<4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_rsqrt_ps(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_rsqrt_ps
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.rsqrt.ps), [[LOAD1]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %res = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_rsqrt_ss(ptr %p1, ptr %p2) {
+  ; CHECK-LABEL: name: test_x86_sse_rsqrt_ss
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.1
+  ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (p0) from %fixed-stack.1)
+  ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:vecr(<4 x s32>) = G_LOAD [[LOAD]](p0) :: (load (<4 x s32>) from %ir.p1)
+  ; CHECK-NEXT:   [[INT:%[0-9]+]]:vecr(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.x86.sse.rsqrt.ss), [[LOAD1]](<4 x s32>)
+  ; CHECK-NEXT:   G_STORE [[INT]](<4 x s32>), [[LOAD]](p0) :: (store (<4 x s32>) into %ir.p1)
+  ; CHECK-NEXT:   RET 0
+  %a0 = load <4 x float>, ptr %p1, align 16
+  %res = call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  store <4 x float> %res, ptr %p1
+  ret void
+}
+declare <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float>) nounwind readnone
diff --git a/llvm/test/CodeGen/X86/GlobalISel/regbankselect-x87.ll b/llvm/test/CodeGen/X86/GlobalISel/regbankselect-x87.ll
index d09db0f2474c9..99d458a183a9b 100644
--- a/llvm/test/CodeGen/X86/GlobalISel/regbankselect-x87.ll
+++ b/llvm/test/CodeGen/X86/GlobalISel/regbankselect-x87.ll
@@ -142,7 +142,7 @@ define float @f4(float %val) {
   ; X86-LABEL: name: f4
   ; X86: bb.1 (%ir-block.0):
   ; X86-NEXT:   [[FRAME_INDEX:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
-  ; X86-NEXT:   [[LOAD:%[0-9]+]]:gpr(s32) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (s32) from %fixed-stack.0)
+  ; X86-NEXT:   [[LOAD:%[0-9]+]]:psr(s32) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (s32) from %fixed-stack.0)
   ; X86-NEXT:   $fp0 = COPY [[LOAD]](s32)
   ; X86-NEXT:   RET 0, implicit $fp0
   ;
@@ -187,13 +187,10 @@ define void @f5(ptr %a, ptr %b) {
   ; X64-NEXT: {{  $}}
   ; X64-NEXT:   [[COPY:%[0-9]+]]:gpr(p0) = COPY $rdi
   ; X64-NEXT:   [[COPY1:%[0-9]+]]:gpr(p0) = COPY $rsi
-  ; X64-NEXT:   [[LOAD:%[0-9]+]]:gpr(s64) = G_LOAD [[COPY]](p0) :: (load (s64) from %ir.a)
-  ; X64-NEXT:   [[LOAD1:%[0-9]+]]:gpr(s64) = G_LOAD [[COPY1]](p0) :: (load (s64) from %ir.b)
-  ; X64-NEXT:   [[COPY2:%[0-9]+]]:psr(s64) = COPY [[LOAD]](s64)
-  ; X64-NEXT:   [[COPY3:%[0-9]+]]:psr(s64) = COPY [[LOAD1]](s64)
-  ; X64-NEXT:   [[FADD:%[0-9]+]]:psr(s64) = G_FADD [[COPY2]], [[COPY3]]
-  ; X64-NEXT:   [[COPY4:%[0-9]+]]:gpr(s64) = COPY [[FADD]](s64)
-  ; X64-NEXT:   G_STORE [[COPY4]](s64), [[COPY]](p0) :: (store (s64) into %ir.a)
+  ; X64-NEXT:   [[LOAD:%[0-9]+]]:psr(s64) = G_LOAD [[COPY]](p0) :: (load (s64) from %ir.a)
+  ; X64-NEXT:   [[LOAD1:%[0-9]+]]:psr(s64) = G_LOAD [[COPY1]](p0) :: (load (s64) from %ir.b)
+  ; X64-NEXT:   [[FADD:%[0-9]+]]:psr(s64) = G_FADD [[LOAD]], [[LOAD1]]
+  ; X64-NEXT:   G_STORE [[FADD]](s64), [[COPY]](p0) :: (store (s64) into %ir.a)
   ; X64-NEXT:   RET 0
   %load1 = load double, ptr %a, align 8
   %load2 = load double, ptr %b, align 8
@@ -210,11 +207,9 @@ define void @f6(ptr %0, ptr %1) {
   ; X86-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:gpr(p0) = G_FRAME_INDEX %fixed-stack.0
   ; X86-NEXT:   [[LOAD1:%[0-9]+]]:gpr(p0) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (p0) from %fixed-stack.0)
   ; X86-NEXT:   [[C:%[0-9]+]]:psr(s32) = G_FCONSTANT float 2.000000e+01
-  ; X86-NEXT:   [[LOAD2:%[0-9]+]]:gpr(s32) = G_LOAD [[LOAD]](p0) :: (load (s32) from %ir.0)
-  ; X86-NEXT:   [[COPY:%[0-9]+]]:psr(s32) = COPY [[LOAD2]](s32)
-  ; X86-NEXT:   [[FADD:%[0-9]+]]:psr(s32) = G_FADD [[COPY]], [[C]]
-  ; X86-NEXT:   [[COPY1:%[0-9]+]]:gpr(s32) = COPY [[FADD]](s32)
-  ; X86-NEXT:   G_STORE [[COPY1]](s32), [[LOAD1]](p0) :: (store (s32) into %ir.1)
+  ; X86-NEXT:   [[LOAD2:%[0-9]+]]:psr(s32) = G_LOAD [[LOAD]](p0) :: (load (s32) from %ir.0)
+  ; X86-NEXT:   [[FADD:%[0-9]+]]:psr(s32) = G_FADD [[LOAD2]], [[C]]
+  ; X86-NEXT:   G_STORE [[FADD]](s32), [[LOAD1]](p0) :: (store (s32) into %ir.1)
   ; X86-NEXT:   RET 0
   ;
   ; X64-LABEL: name: f6
@@ -224,11 +219,9 @@ define void @f6(ptr %0, ptr %1) {
   ; X64-NEXT:   [[COPY:%[0-9]+]]:gpr(p0) = COPY $rdi
   ; X64-NEXT:   [[COPY1:%[0-9]+]]:gpr(p0) = COPY $rsi
   ; X64-NEXT:   [[C:%[0-9]+]]:psr(s32) = G_FCONSTANT float 2.000000e+01
-  ; X64-NEXT:   [[LOAD:%[0-9]+]]:gpr(s32) = G_LOAD [[COPY]](p0) :: (load (s32) from %ir.0)
-  ; X64-NEXT:   [[COPY2:%[0-9]+]]:psr(s32) = COPY [[LOAD]](s32)
-  ; X64-NEXT:   [[FADD:%[0-9]+]]:psr(s32) = G_FADD [[COPY2]], [[C]]
-  ; X64-NEXT:   [[COPY3:%[0-9]+]]:gpr(s32) = COPY [[FADD]](s32)
-  ; X64-NEXT:   G_STORE [[COPY3]](s32), [[COPY1]](p0) :: (store (s32) into %ir.1)
+  ; X64-NEXT:   [[LOAD:%[0-9]+]]:psr(s32) = G_LOAD [[COPY]](p0) :: (load (s32) from %ir.0)
+  ; X64-NEXT:   [[FADD:%[0-9]+]]:psr(s32) = G_FADD [[LOAD]], [[C]]
+  ; X64-NEXT:   G_STORE [[FADD]](s32), [[COPY1]](p0) :: (store (s32) into %ir.1)
   ; X64-NEXT:   RET 0
   %load1 = load float, ptr %0
   %add = fadd float %load1, 20.0

From ac6b4c618a52c62cef9b143a767991dbba7453e1 Mon Sep 17 00:00:00 2001
From: Jon Chesterfield <jonathanchesterfield@gmail.com>
Date: Tue, 16 Apr 2024 12:25:14 +0100
Subject: [PATCH 567/886] Reapply "[Verifier] Reject va_start in non-variadic
 function (#88809)"

This reverts commit f4960da6023b8034ae68925c3223d51624621b37.
Includes a fix for the MLIR test case.
---
 llvm/lib/Analysis/Lint.cpp                      | 5 +----
 llvm/lib/IR/Verifier.cpp                        | 5 +++++
 llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll | 2 +-
 llvm/test/Other/lint.ll                         | 7 -------
 llvm/test/Verifier/variadic.ll                  | 8 ++++++++
 mlir/test/Target/LLVMIR/Import/intrinsic.ll     | 2 +-
 6 files changed, 16 insertions(+), 13 deletions(-)
 create mode 100644 llvm/test/Verifier/variadic.ll

diff --git a/llvm/lib/Analysis/Lint.cpp b/llvm/lib/Analysis/Lint.cpp
index 0694c2995dfcc..1ab856ac8830a 100644
--- a/llvm/lib/Analysis/Lint.cpp
+++ b/llvm/lib/Analysis/Lint.cpp
@@ -350,10 +350,7 @@ void Lint::visitCallBase(CallBase &I) {
     }
 
     case Intrinsic::vastart:
-      Check(I.getParent()->getParent()->isVarArg(),
-            "Undefined behavior: va_start called in a non-varargs function",
-            &I);
-
+      // vastart in non-varargs function is rejected by the verifier
       visitMemoryReference(I, MemoryLocation::getForArgument(&I, 0, TLI),
                            std::nullopt, nullptr, MemRef::Read | MemRef::Write);
       break;
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 516d4a0515569..4cd61e6e531bf 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -5798,6 +5798,11 @@ void Verifier::visitIntrinsicCall(Intrinsic::ID ID, CallBase &Call) {
 
     break;
   }
+  case Intrinsic::vastart: {
+    Check(Call.getFunction()->isVarArg(),
+          "va_start called in a non-varargs function");
+    break;
+  }
   case Intrinsic::vector_reduce_and:
   case Intrinsic::vector_reduce_or:
   case Intrinsic::vector_reduce_xor:
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
index bd576d0f70e9c..8c6e01d934c2d 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/vastart.ll
@@ -3,7 +3,7 @@
 
 
 declare void @llvm.va_start(ptr)
-define void @test_va_start(ptr %list) {
+define void @test_va_start(ptr %list, ...) {
 ; CHECK-LABEL: name: test_va_start
 ; CHECK: [[LIST:%[0-9]+]]:_(p0) = COPY $x0
 ; CHECK-IOS: G_VASTART [[LIST]](p0) :: (store (s64) into %ir.list, align 1)
diff --git a/llvm/test/Other/lint.ll b/llvm/test/Other/lint.ll
index 6b31b31a78c98..6fd2d40cd2f29 100644
--- a/llvm/test/Other/lint.ll
+++ b/llvm/test/Other/lint.ll
@@ -124,13 +124,6 @@ define void @0() nounwind {
   ret void
 }
 
-; CHECK: va_start called in a non-varargs function
-declare void @llvm.va_start(ptr)
-define void @not_vararg(ptr %p) nounwind {
-  call void @llvm.va_start(ptr %p)
-  ret void
-}
-
 ; CHECK: Undefined behavior: Branch to non-blockaddress
 define void @use_indbr() {
   indirectbr ptr @foo, [label %block]
diff --git a/llvm/test/Verifier/variadic.ll b/llvm/test/Verifier/variadic.ll
new file mode 100644
index 0000000000000..55e4a4da0a920
--- /dev/null
+++ b/llvm/test/Verifier/variadic.ll
@@ -0,0 +1,8 @@
+; RUN: not opt -S -passes=verify 2>&1 < %s | FileCheck %s
+
+; CHECK: va_start called in a non-varargs function
+declare void @llvm.va_start(ptr)
+define void @not_vararg(ptr %p) nounwind {
+  call void @llvm.va_start(ptr %p)
+  ret void
+}
diff --git a/mlir/test/Target/LLVMIR/Import/intrinsic.ll b/mlir/test/Target/LLVMIR/Import/intrinsic.ll
index 81a6eadbadd3f..bf6847a32ff4f 100644
--- a/mlir/test/Target/LLVMIR/Import/intrinsic.ll
+++ b/mlir/test/Target/LLVMIR/Import/intrinsic.ll
@@ -597,7 +597,7 @@ define void @ushl_sat_test(i32 %0, i32 %1, <8 x i32> %2, <8 x i32> %3) {
 }
 
 ; CHECK-LABEL: llvm.func @va_intrinsics_test
-define void @va_intrinsics_test(ptr %0, ptr %1) {
+define void @va_intrinsics_test(ptr %0, ptr %1, ...) {
 ; CHECK: llvm.intr.vastart %{{.*}}
   call void @llvm.va_start.p0(ptr %0)
 ; CHECK: llvm.intr.vacopy %{{.*}} to %{{.*}}

From 09e7d7585cf881fb598eb56738579b84d027318c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Tue, 16 Apr 2024 07:52:39 +0200
Subject: [PATCH 568/886] [clang][Interp] Don't add 'in call to' diagnostics
 for builtin frames

---
 clang/lib/AST/Interp/InterpFrame.cpp        |  7 ++++++
 clang/lib/AST/Interp/State.cpp              |  5 +++--
 clang/test/AST/Interp/builtin-functions.cpp | 24 +++++++--------------
 3 files changed, 18 insertions(+), 18 deletions(-)

diff --git a/clang/lib/AST/Interp/InterpFrame.cpp b/clang/lib/AST/Interp/InterpFrame.cpp
index 12e2e6ff9155b..ba957546473e9 100644
--- a/clang/lib/AST/Interp/InterpFrame.cpp
+++ b/clang/lib/AST/Interp/InterpFrame.cpp
@@ -152,6 +152,13 @@ void print(llvm::raw_ostream &OS, const Pointer &P, ASTContext &Ctx,
 }
 
 void InterpFrame::describe(llvm::raw_ostream &OS) const {
+  // We create frames for builtin functions as well, but we can't reliably
+  // diagnose them. The 'in call to' diagnostics for them add no value to the
+  // user _and_ it doesn't generally work since the argument types don't always
+  // match the function prototype. Just ignore them.
+  if (const auto *F = getFunction(); F && F->isBuiltin())
+    return;
+
   const FunctionDecl *F = getCallee();
   if (const auto *M = dyn_cast<CXXMethodDecl>(F);
       M && M->isInstance() && !isa<CXXConstructorDecl>(F)) {
diff --git a/clang/lib/AST/Interp/State.cpp b/clang/lib/AST/Interp/State.cpp
index 47fbf5145cd4e..0d9dadec4b958 100644
--- a/clang/lib/AST/Interp/State.cpp
+++ b/clang/lib/AST/Interp/State.cpp
@@ -155,7 +155,8 @@ void State::addCallStack(unsigned Limit) {
     SmallString<128> Buffer;
     llvm::raw_svector_ostream Out(Buffer);
     F->describe(Out);
-    addDiag(CallRange.getBegin(), diag::note_constexpr_call_here)
-        << Out.str() << CallRange;
+    if (!Buffer.empty())
+      addDiag(CallRange.getBegin(), diag::note_constexpr_call_here)
+          << Out.str() << CallRange;
   }
 }
diff --git a/clang/test/AST/Interp/builtin-functions.cpp b/clang/test/AST/Interp/builtin-functions.cpp
index a7adc92d3714f..1a29a664d7ce5 100644
--- a/clang/test/AST/Interp/builtin-functions.cpp
+++ b/clang/test/AST/Interp/builtin-functions.cpp
@@ -24,16 +24,13 @@ namespace strcmp {
   static_assert(__builtin_strcmp("abab", "abab\0banana") == 0, "");
   static_assert(__builtin_strcmp("abab\0banana", "abab\0canada") == 0, "");
   static_assert(__builtin_strcmp(0, "abab") == 0, ""); // both-error {{not an integral constant}} \
-                                                       // both-note {{dereferenced null}} \
-                                                       // expected-note {{in call to}}
+                                                       // both-note {{dereferenced null}}
   static_assert(__builtin_strcmp("abab", 0) == 0, ""); // both-error {{not an integral constant}} \
-                                                       // both-note {{dereferenced null}} \
-                                                       // expected-note {{in call to}}
+                                                       // both-note {{dereferenced null}}
 
   static_assert(__builtin_strcmp(kFoobar, kFoobazfoobar) == -1, "");
   static_assert(__builtin_strcmp(kFoobar, kFoobazfoobar + 6) == 0, ""); // both-error {{not an integral constant}} \
-                                                                        // both-note {{dereferenced one-past-the-end}} \
-                                                                        // expected-note {{in call to}}
+                                                                        // both-note {{dereferenced one-past-the-end}}
 
   /// Used to assert because we're passing a dummy pointer to
   /// __builtin_strcmp() when evaluating the return statement.
@@ -72,14 +69,11 @@ constexpr const char *a = "foo\0quux";
   static_assert(check(c), "");
 
   constexpr int over1 = __builtin_strlen(a + 9); // both-error {{constant expression}} \
-                                                 // both-note {{one-past-the-end}} \
-                                                 // expected-note {{in call to}}
+                                                 // both-note {{one-past-the-end}}
   constexpr int over2 = __builtin_strlen(b + 9); // both-error {{constant expression}} \
-                                                 // both-note {{one-past-the-end}} \
-                                                 // expected-note {{in call to}}
+                                                 // both-note {{one-past-the-end}}
   constexpr int over3 = __builtin_strlen(c + 9); // both-error {{constant expression}} \
-                                                 // both-note {{one-past-the-end}} \
-                                                 // expected-note {{in call to}}
+                                                 // both-note {{one-past-the-end}}
 
   constexpr int under1 = __builtin_strlen(a - 1); // both-error {{constant expression}} \
                                                   // both-note {{cannot refer to element -1}}
@@ -90,8 +84,7 @@ constexpr const char *a = "foo\0quux";
 
   constexpr char d[] = { 'f', 'o', 'o' }; // no nul terminator.
   constexpr int bad = __builtin_strlen(d); // both-error {{constant expression}} \
-                                           // both-note {{one-past-the-end}} \
-                                           // expected-note {{in call to}}
+                                           // both-note {{one-past-the-end}}
 }
 
 namespace nan {
@@ -114,8 +107,7 @@ namespace nan {
   /// FIXME: Current interpreter misses diagnostics.
   constexpr char f2[] = {'0', 'x', 'A', 'E'}; /// No trailing 0 byte.
   constexpr double NaN7 = __builtin_nan(f2); // both-error {{must be initialized by a constant expression}} \
-                                             // expected-note {{read of dereferenced one-past-the-end pointer}} \
-                                             // expected-note {{in call to}}
+                                             // expected-note {{read of dereferenced one-past-the-end pointer}}
   static_assert(!__builtin_issignaling(__builtin_nan("")), "");
   static_assert(__builtin_issignaling(__builtin_nans("")), "");
 }

From a8de3ee8994023ea7669397587f8118ae5bba9c9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Tue, 16 Apr 2024 13:29:10 +0200
Subject: [PATCH 569/886] [clang][Interp][NFC] Fix some build warnings

Fixes:
/buildbot/worker/arc-folder/llvm-project/clang/lib/AST/Interp/Disasm.cpp:143:25: warning: cast from type 'const clang::interp::Block*' to type 'void*' casts away qualifiers [-Wcast-qual]
/buildbot/worker/arc-folder/llvm-project/clang/lib/AST/Interp/Disasm.cpp:271:23: warning: cast from type 'const clang::interp::Block*' to type 'void*' casts away qualifiers [-Wcast-qual]
---
 clang/lib/AST/Interp/Disasm.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/clang/lib/AST/Interp/Disasm.cpp b/clang/lib/AST/Interp/Disasm.cpp
index ebc4e4f195ba6..d127f33223e80 100644
--- a/clang/lib/AST/Interp/Disasm.cpp
+++ b/clang/lib/AST/Interp/Disasm.cpp
@@ -140,7 +140,7 @@ LLVM_DUMP_METHOD void Program::dump(llvm::raw_ostream &OS) const {
     const Descriptor *Desc = G->block()->getDescriptor();
     Pointer GP = getPtrGlobal(GI);
 
-    OS << GI << ": " << (void *)G->block() << " ";
+    OS << GI << ": " << (const void *)G->block() << " ";
     {
       ColorScope SC(OS, true,
                     GP.isInitialized()
@@ -268,7 +268,7 @@ LLVM_DUMP_METHOD void Record::dump(llvm::raw_ostream &OS, unsigned Indentation,
 LLVM_DUMP_METHOD void Block::dump(llvm::raw_ostream &OS) const {
   {
     ColorScope SC(OS, true, {llvm::raw_ostream::BRIGHT_BLUE, true});
-    OS << "Block " << (void *)this << "\n";
+    OS << "Block " << (const void *)this << "\n";
   }
   unsigned NPointers = 0;
   for (const Pointer *P = Pointers; P; P = P->Next) {

From c18a3b6bd30456305cf1b3d78ad5a805577388c1 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 16 Apr 2024 12:30:13 +0100
Subject: [PATCH 570/886] [DAG] Fold
 extract_subvector(insert_subvector(x,y,c1),c2) --> extract_subvector(y,c2-c1)
 (#87925) (REAPPLIED)

If the extract_subvector is cheap, attempt to extract directly from an inserted subvector

Reapplied with a check to ensure we only attempt this for fixed vectors
---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |   16 +
 .../any_extend_vector_inreg_of_broadcast.ll   |   42 +-
 llvm/test/CodeGen/X86/dpbusd.ll               |    2 +-
 llvm/test/CodeGen/X86/dpbusd_i4.ll            |    2 +-
 .../vector-interleaved-load-i16-stride-3.ll   | 1300 ++++++++---------
 .../vector-interleaved-store-i8-stride-7.ll   |  602 ++++----
 .../zero_extend_vector_inreg_of_broadcast.ll  |   28 +-
 7 files changed, 986 insertions(+), 1006 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 0fa0bf2609bb3..c36b1cc9039c2 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -24467,6 +24467,22 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
       if (!LegalOperations || TLI.isOperationLegal(ISD::SPLAT_VECTOR, NVT))
         return DAG.getSplatVector(NVT, DL, V.getOperand(0));
 
+  // extract_subvector(insert_subvector(x,y,c1),c2)
+  //  --> extract_subvector(y,c2-c1)
+  // iff we're just extracting from the inserted subvector.
+  if (V.getOpcode() == ISD::INSERT_SUBVECTOR) {
+    SDValue InsSub = V.getOperand(1);
+    EVT InsSubVT = InsSub.getValueType();
+    unsigned NumInsElts = InsSubVT.getVectorMinNumElements();
+    unsigned InsIdx = V.getConstantOperandVal(2);
+    unsigned NumSubElts = NVT.getVectorMinNumElements();
+    if (InsIdx <= ExtIdx && (ExtIdx + NumSubElts) <= (InsIdx + NumInsElts) &&
+        TLI.isExtractSubvectorCheap(NVT, InsSubVT, ExtIdx - InsIdx) &&
+        InsSubVT.isFixedLengthVector() && NVT.isFixedLengthVector())
+      return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NVT, InsSub,
+                         DAG.getVectorIdxConstant(ExtIdx - InsIdx, DL));
+  }
+
   // Try to move vector bitcast after extract_subv by scaling extraction index:
   // extract_subv (bitcast X), Index --> bitcast (extract_subv X, Index')
   if (V.getOpcode() == ISD::BITCAST &&
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
index 4242d8483e723..39c7ce1413d1b 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -3507,13 +3507,12 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm2
-; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512F-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3523,13 +3522,12 @@ define void @vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12(ptr %i
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4],ymm1[5],ymm2[6],ymm1[7],ymm2[8],ymm1[9],ymm2[10],ymm1[11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm1, %ymm1
 ; AVX512DQ-NEXT:    vpaddb 32(%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, 32(%rcx)
@@ -3768,10 +3766,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -3784,10 +3782,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4147,9 +4145,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %xmm0
@@ -4161,9 +4159,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/dpbusd.ll b/llvm/test/CodeGen/X86/dpbusd.ll
index fbea08eb1e550..04d7a9691b645 100644
--- a/llvm/test/CodeGen/X86/dpbusd.ll
+++ b/llvm/test/CodeGen/X86/dpbusd.ll
@@ -26,7 +26,7 @@ define i32 @no_dpbusd(ptr%a, ptr%b, i32 %c, i32 %n) {
 ; AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
 ; AVX512-NEXT:    vpmaddwd %ymm0, %ymm1, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/dpbusd_i4.ll b/llvm/test/CodeGen/X86/dpbusd_i4.ll
index 906fead7f8db5..a212f99680ef4 100644
--- a/llvm/test/CodeGen/X86/dpbusd_i4.ll
+++ b/llvm/test/CodeGen/X86/dpbusd_i4.ll
@@ -86,7 +86,7 @@ define i32 @mul_sext_i4i4(<16 x i4> %a, <16 x i4> %b, i32 %c) {
 ; CHECK-NEXT:    vpsraw $12, %ymm0, %ymm0
 ; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm0
 ; CHECK-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; CHECK-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 1436922f9dd11..6d5fc9ed0ab5b 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -1828,22 +1828,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-LABEL: load_i16_stride3_vf32:
 ; AVX512:       # %bb.0:
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1857,14 +1857,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1885,21 +1885,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1908,22 +1906,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512-FCP:       # %bb.0:
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -1937,14 +1935,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512-FCP-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -1965,21 +1963,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -1988,22 +1984,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2017,14 +2013,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512DQ-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2045,21 +2041,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -2068,22 +2062,22 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf32:
 ; AVX512DQ-FCP:       # %bb.0:
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
 ; AVX512DQ-FCP-NEXT:    vmovdqa 128(%rdi), %ymm5
 ; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm6
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm0
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7],ymm0[8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13,14],ymm2[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm6, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6],xmm0[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm4[0,1,2],ymm3[3,4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %ymm8
 ; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm9
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm3
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm9, %ymm8, %ymm3
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13,14],ymm4[15]
@@ -2097,14 +2091,14 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm10, %zmm7
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, %ymm10
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm6, %ymm5, %ymm10
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7,8,9],ymm11[10],ymm10[11,12],ymm11[13],ymm10[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm10, %ymm10
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm0[2],xmm2[3,4],xmm0[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm11, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm12 = xmm12[2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4],xmm10[5,6,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm12 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
@@ -2125,21 +2119,19 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1,2],ymm5[3],ymm12[4,5],ymm5[6],ymm12[7],ymm5[8],ymm12[9,10],ymm5[11],ymm12[12,13],ymm5[14],ymm12[15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm6 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7],ymm8[8],ymm1[9,10],ymm8[11],ymm1[12,13],ymm8[14],ymm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm0, %xmm0
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm8, %ymm9, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7],ymm2[8],ymm0[9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6],xmm3[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
@@ -3500,688 +3492,668 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; AVX512-LABEL: load_i16_stride3_vf64:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
 ; AVX512-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512-FCP-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512-FCP-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512DQ-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512DQ-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512DQ-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512DQ-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512DQ-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512DQ-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512DQ-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512DQ-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512DQ-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512DQ-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512DQ-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512DQ-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512DQ-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512DQ-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512DQ-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512DQ-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512DQ-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512DQ-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512DQ-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512DQ-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512DQ-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
 ; AVX512DQ-FCP-LABEL: load_i16_stride3_vf64:
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm0 = [65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm20
-; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm21
+; AVX512DQ-FCP-NEXT:    vmovdqa64 224(%rdi), %ymm18
+; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %ymm20
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm21, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7],ymm1[8],ymm3[9],ymm1[10,11],ymm3[12],ymm1[13,14],ymm3[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm5
-; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm8
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm18, %ymm20, %ymm1
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13,14],ymm2[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,6,7,12,13,2,3,4,5,14,15,8,9,10,11,16,17,22,23,28,29,18,19,20,21,30,31,24,25,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm2, %ymm5
+; AVX512DQ-FCP-NEXT:    vmovdqa 272(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 256(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm8[2],xmm2[3,4],xmm8[5],xmm2[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm14
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm6, %xmm6
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm3
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm1, %xmm19
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm13 = [4,5,14,15,0,1,2,3,8,9,14,15,4,5,10,11]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm6, %xmm6
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2],ymm6[3,4,5,6,7],ymm5[8,9,10],ymm6[11,12,13,14,15]
 ; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm22
-; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm23
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm6
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm6
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7],ymm6[8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13,14],ymm7[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm6, %ymm12
+; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %ymm21
+; AVX512DQ-FCP-NEXT:    vmovdqa64 352(%rdi), %ymm22
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm8
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm8
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7],ymm8[8],ymm9[9],ymm8[10,11],ymm9[12],ymm8[13,14],ymm9[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11,16,17,22,23,28,29,18,19,24,25,30,31,20,21,26,27]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm11
 ; AVX512DQ-FCP-NEXT:    vmovdqa 304(%rdi), %xmm1
 ; AVX512DQ-FCP-NEXT:    vmovdqa 288(%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm13 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6],xmm1[7]
 ; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm6
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm15 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm13, %xmm13
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm5, %zmm16
-; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm24
-; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm13
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, %xmm8
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm14 = [0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm5, %zmm16
+; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %ymm23
+; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm11
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm13, %ymm5
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm11, %ymm5
 ; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm5[2,3,0,1]
 ; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm5 = ymm5[0],ymm12[1],ymm5[2,3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8],ymm12[9],ymm5[10,11],ymm12[12],ymm5[13,14],ymm12[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm5, %ymm5
-; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm11
-; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm12
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm10 = xmm12[0],xmm11[1],xmm12[2,3],xmm11[4],xmm12[5,6],xmm11[7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm15, %xmm10, %xmm10
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm5[3,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm17
-; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm5
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm10, %ymm5, %ymm10
+; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm15
+; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm5
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm12 = xmm5[0],xmm15[1],xmm5[2,3],xmm15[4],xmm5[5,6],xmm15[7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm14, %xmm12, %xmm12
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm10[3,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %ymm24
+; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm12
 ; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm10
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm5, %ymm17, %ymm10
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm15 = ymm10[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm10 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6],ymm15[7],ymm10[8],ymm15[9],ymm10[10,11],ymm15[12],ymm10[13,14],ymm15[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm10, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm12, %ymm24, %ymm10
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm1[1],ymm10[2,3],ymm1[4],ymm10[5,6],ymm1[7],ymm10[8],ymm1[9],ymm10[10,11],ymm1[12],ymm10[13,14],ymm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm7, %ymm1, %ymm7
 ; AVX512DQ-FCP-NEXT:    vmovdqa 80(%rdi), %xmm10
-; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm15
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1],xmm10[2],xmm15[3,4],xmm10[5],xmm15[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3,4,5,6,7],ymm2[8,9,10],ymm3[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm18
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm23, %ymm22, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm2, %ymm28
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm6, %xmm25
+; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %xmm1
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm10[2],xmm1[3,4],xmm10[5],xmm1[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm13, %xmm2, %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7],ymm7[8,9,10],ymm2[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm2, %zmm17
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm21, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7,8,9],ymm6[10],ymm2[11,12],ymm6[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13,18,19,24,25,30,31,20,21,26,27,16,17,22,23,28,29]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm4[0,1],xmm8[2],xmm4[3,4],xmm8[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm25
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm4, %xmm26
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,8,9,14,15,4,5,10,11,10,11,10,11,10,11]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm9 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm7
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm13 = [65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535,0,65535,65535]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
 ; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,3,8,9,14,15,4,5,12,13,10,11,0,1,6,7,18,19,24,25,30,31,20,21,28,29,26,27,16,17,22,23]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm14, %xmm7
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm14[2],xmm8[3,4],xmm14[5],xmm8[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm8, %xmm27
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm2 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm14, %xmm14
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm1[0,1,2],ymm14[3,4,5,6,7],ymm1[8,9,10],ymm14[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm19
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm13, %ymm24, %ymm1
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7,8,9],ymm3[10],ymm1[11,12],ymm3[13],ymm1[14,15]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %ymm28, %ymm3
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1],xmm11[2],xmm12[3,4],xmm11[5],xmm12[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm1[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm9, %ymm3
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm17, %ymm5, %ymm3
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm15[2],xmm10[3,4],xmm15[5],xmm10[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm2, %xmm4, %xmm2
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm9, %ymm13
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm13[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0],ymm13[1,2],ymm2[3],ymm13[4,5],ymm2[6],ymm13[7],ymm2[8],ymm13[9,10],ymm2[11],ymm13[12,13],ymm2[14],ymm13[15]
-; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm17, %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7],ymm3[8],ymm5[9,10],ymm3[11],ymm5[12,13],ymm3[14],ymm5[15]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm4 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0],xmm10[1],xmm15[2,3],xmm10[4],xmm15[5,6],xmm10[7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
 ; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm11[0,1],xmm12[2],xmm11[3,4],xmm12[5],xmm11[6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm8 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm5, %zmm3, %zmm5
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm5, %xmm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm2[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm19, %xmm8
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm14 = xmm8[0,1],xmm3[2],xmm8[3,4],xmm3[5],xmm8[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm3, %xmm27
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [4,5,4,5,4,5,4,5,10,11,0,1,6,7,12,13]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm14, %xmm14
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm14 = ymm2[0,1,2],ymm14[3,4,5,6,7],ymm2[8,9,10],ymm14[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm19
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, %ymm2
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm11, %ymm23, %ymm2
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7,8,9],ymm7[10],ymm2[11,12],ymm7[13],ymm2[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm9, %ymm2, %ymm2
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm7 = xmm5[0,1],xmm15[2],xmm5[3,4],xmm15[5],xmm5[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm7, %xmm6
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm2[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm13, %ymm6
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm24, %ymm12, %ymm6
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm6[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7,8,9],ymm7[10],ymm6[11,12],ymm7[13],ymm6[14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm6, %ymm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm6 = xmm10[0,1],xmm1[2],xmm10[3,4],xmm1[5],xmm10[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7],ymm4[8,9,10],ymm3[11,12,13,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm2
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm22, %ymm23, %ymm9
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm9[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm9[1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7],ymm3[8],ymm9[9,10],ymm3[11],ymm9[12,13],ymm3[14],ymm9[15]
-; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm20, %ymm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[2,3,0,1]
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7],ymm5[8],ymm0[9,10],ymm5[11],ymm0[12,13],ymm5[14],ymm0[15]
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm3, %ymm3
-; AVX512DQ-FCP-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm23, %ymm13, %ymm11
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm11[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm11[1,2],ymm3[3],ymm11[4,5],ymm3[6],ymm11[7],ymm3[8],ymm11[9,10],ymm3[11],ymm11[12,13],ymm3[14],ymm11[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15,20,21,26,27,16,17,22,23,28,29,18,19,24,25,30,31]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm15[0,1],xmm5[2],xmm15[3,4],xmm5[5],xmm15[6,7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm5 = [4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $226, %ymm24, %ymm0, %ymm12
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7],ymm4[8],ymm12[9,10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3],xmm10[4],xmm1[5,6],xmm10[7]
+; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,0,1,6,7,12,13,2,3,8,9,14,15]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm21, %ymm22, %ymm13
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm13[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm3 = ymm3[0],ymm13[1,2],ymm3[3],ymm13[4,5],ymm3[6],ymm13[7],ymm3[8],ymm13[9,10],ymm3[11],ymm13[12,13],ymm3[14],ymm13[15]
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm7
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
+; AVX512DQ-FCP-NEXT:    vpshufb %xmm5, %xmm4, %xmm4
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm3, %ymm3
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
+; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm20, %ymm18, %ymm0
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2],ymm4[3],ymm0[4,5],ymm4[6],ymm0[7],ymm4[8],ymm0[9,10],ymm4[11],ymm0[12,13],ymm4[14],ymm0[15]
+; AVX512DQ-FCP-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm27, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2,3],xmm4[4],xmm7[5,6],xmm4[7]
+; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm8[1],xmm4[2,3],xmm8[4],xmm4[5,6],xmm8[7]
 ; AVX512DQ-FCP-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
 ; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
 ; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm25, %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %xmm26, %xmm5
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
-; AVX512DQ-FCP-NEXT:    vpshufb %xmm8, %xmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm4
-; AVX512DQ-FCP-NEXT:    vextracti32x4 $2, %zmm4, %xmm4
-; AVX512DQ-FCP-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm3[5,6,7]
-; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
 ; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, (%rsi)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, (%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 64(%rsi)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rdx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
 ; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rcx)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rcx)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 8b6ba51506ab7..8091afbbfd70c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -1246,29 +1246,28 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
 ; AVX512BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
-; AVX512BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
 ; AVX512BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
-; AVX512BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
-; AVX512BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
+; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
+; AVX512BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
+; AVX512BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX512BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
+; AVX512BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
 ; AVX512BW-FCP-NEXT:    vzeroupper
 ; AVX512BW-FCP-NEXT:    retq
 ;
@@ -1326,29 +1325,28 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
 ; AVX512DQ-BW-FCP-NEXT:    vpermi2q %ymm3, %ymm0, %ymm1
-; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,3,5,7,1,3,5,7]
-; AVX512DQ-BW-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm0, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,ymm0[1,5,9,13],zero,zero,zero,ymm0[2,6,10,14],zero,zero,zero,ymm0[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,4,8],zero,zero,zero,zero,ymm3[1,5,9],zero,zero,zero,zero,ymm3[2,6,18],zero,zero,zero,zero,ymm3[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
+; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8],zero,zero,zero,zero,zero,ymm3[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm3[19,27],zero,zero,zero,zero,zero,ymm3[20,28],zero,zero
 ; AVX512DQ-BW-FCP-NEXT:    vpor %ymm0, %ymm3, %ymm0
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm3 = ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,ymm2[2,10,18,26],zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28]
-; AVX512DQ-BW-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,ymm2[0,8],zero,zero,zero,zero,zero,ymm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm2[19,27],zero,zero,zero,zero,zero,ymm2[20,28],zero,zero
-; AVX512DQ-BW-FCP-NEXT:    vpor %ymm3, %ymm2, %ymm2
 ; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
 ; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
-; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm3
 ; AVX512DQ-BW-FCP-NEXT:    movl $236730480, %ecx # imm = 0xE1C3870
 ; AVX512DQ-BW-FCP-NEXT:    kmovd %ecx, %k1
-; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 {%k1} = ymm1[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
-; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
-; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
-; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
-; AVX512DQ-BW-FCP-NEXT:    vextracti32x4 $3, %zmm0, %xmm0
-; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, 48(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm3[u,u,u,u,0,4,8,u,u,u,u,1,5,9,u,u,u,u,18,22,26,u,u,u,u,19,23,27,u,u,u,u]
+; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [1,3,5,7,1,3,5,7]
+; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm2, %ymm3, %ymm2
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = zero,zero,zero,ymm2[1,5,9,13],zero,zero,zero,ymm2[2,6,10,14],zero,zero,zero,ymm2[19,23,27,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [1,3,5,0,5,1,3,0]
+; AVX512DQ-BW-FCP-NEXT:    vpermd %ymm1, %ymm3, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4,8],zero,zero,zero,zero,ymm1[1,5,9],zero,zero,zero,zero,ymm1[2,6,18],zero,zero,zero,zero,ymm1[23,27,19],zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512DQ-BW-FCP-NEXT:    vpor %ymm2, %ymm1, %ymm1
+; AVX512DQ-BW-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm2, 48(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm1, 32(%rax)
+; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm0, (%rax)
 ; AVX512DQ-BW-FCP-NEXT:    vzeroupper
 ; AVX512DQ-BW-FCP-NEXT:    retq
   %in.vec0 = load <8 x i8>, ptr %in.vecptr0, align 64
@@ -2053,77 +2051,76 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-NEXT:    vmovdqa (%rdi), %xmm4
-; AVX512-NEXT:    vmovdqa (%rsi), %xmm5
-; AVX512-NEXT:    vmovdqa (%rdx), %xmm6
-; AVX512-NEXT:    vmovdqa (%rcx), %xmm7
-; AVX512-NEXT:    vmovdqa (%r8), %xmm0
-; AVX512-NEXT:    vmovdqa (%r10), %xmm1
-; AVX512-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
-; AVX512-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
-; AVX512-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
-; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
-; AVX512-NEXT:    vpor %xmm6, %xmm4, %xmm4
-; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm5
-; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
-; AVX512-NEXT:    vpor %xmm6, %xmm5, %xmm5
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
-; AVX512-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
-; AVX512-NEXT:    vpor %ymm4, %ymm6, %ymm4
-; AVX512-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
+; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
+; AVX512-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
+; AVX512-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
+; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
+; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
+; AVX512-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
+; AVX512-NEXT:    vporq %zmm10, %zmm11, %zmm10
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
+; AVX512-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
+; AVX512-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
+; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
+; AVX512-NEXT:    vpandn %ymm12, %ymm13, %ymm12
+; AVX512-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
+; AVX512-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
+; AVX512-NEXT:    vporq %zmm12, %zmm11, %zmm11
+; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
+; AVX512-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
+; AVX512-NEXT:    vpor %ymm8, %ymm9, %ymm8
+; AVX512-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
+; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
 ; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
-; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
-; AVX512-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
-; AVX512-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512-NEXT:    vporq %zmm3, %zmm2, %zmm2
-; AVX512-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
-; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512-NEXT:    vpandn %ymm3, %ymm6, %ymm3
-; AVX512-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
-; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; AVX512-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
-; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
-; AVX512-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
-; AVX512-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
-; AVX512-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
-; AVX512-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
-; AVX512-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
-; AVX512-NEXT:    vporq %zmm1, %zmm0, %zmm0
-; AVX512-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
-; AVX512-NEXT:    vmovdqa %xmm5, 96(%rax)
-; AVX512-NEXT:    vmovdqa64 %zmm0, (%rax)
-; AVX512-NEXT:    vmovdqa %ymm4, 64(%rax)
+; AVX512-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
+; AVX512-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512-NEXT:    vmovdqa64 %zmm11, (%rax)
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
 ;
@@ -2131,70 +2128,69 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512-FCP:       # %bb.0:
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm2
-; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm3
-; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm4
-; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm5
-; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm1
-; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm0
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
-; AVX512-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
-; AVX512-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
-; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
-; AVX512-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
-; AVX512-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
-; AVX512-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
-; AVX512-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
-; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
-; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
-; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
-; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
-; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
-; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
-; AVX512-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
-; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
-; AVX512-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
-; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
-; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
-; AVX512-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
-; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
-; AVX512-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
-; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
-; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
-; AVX512-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
-; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
-; AVX512-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
+; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512-FCP-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512-FCP-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512-FCP-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512-FCP-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512-FCP-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512-FCP-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
+; AVX512-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
+; AVX512-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
+; AVX512-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
+; AVX512-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
+; AVX512-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
+; AVX512-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
+; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
+; AVX512-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
+; AVX512-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
+; AVX512-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
+; AVX512-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
+; AVX512-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
+; AVX512-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
+; AVX512-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
+; AVX512-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
+; AVX512-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
 ; AVX512-FCP-NEXT:    vzeroupper
 ; AVX512-FCP-NEXT:    retq
 ;
@@ -2202,77 +2198,76 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ:       # %bb.0:
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm4
-; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm5
-; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm6
-; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm7
-; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm0
-; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm1
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm7, %ymm6, %ymm3
-; AVX512DQ-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm2
-; AVX512DQ-NEXT:    vinserti128 $1, (%r9), %ymm0, %ymm0
-; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[u,u],zero,zero,xmm6[12,13,u,u,u],zero,zero,xmm6[14,15,u,u,u]
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,12,13],zero,zero,xmm4[u,u,u,14,15],zero,zero,xmm4[u,u,u]
-; AVX512DQ-NEXT:    vpor %xmm6, %xmm4, %xmm4
-; AVX512DQ-NEXT:    vextracti128 $1, %ymm0, %xmm5
-; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[10],zero,xmm5[u,u,u,u,13,12],zero,xmm5[u,u,u,u,15,14],zero
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm6 = zero,xmm1[13,u,u,u,u],zero,zero,xmm1[14,u,u,u,u],zero,zero,xmm1[15]
-; AVX512DQ-NEXT:    vpor %xmm6, %xmm5, %xmm5
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm5
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[3,1,1,3]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[1],zero,zero,ymm4[u,u,u,10,2],zero,zero,ymm4[u,u,u,11,3],zero,zero,ymm4[u,u,u,20,28],zero,zero,ymm4[u,u,u,21,29],zero,zero,ymm4[u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[1,3,3,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = zero,ymm6[1,9,u,u,u],zero,zero,ymm6[2,10,u,u,u],zero,zero,ymm6[3,19,u,u,u],zero,zero,ymm6[28,20,u,u,u],zero,zero,ymm6[29,21,u]
-; AVX512DQ-NEXT:    vpor %ymm4, %ymm6, %ymm4
-; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,6]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[1,3,1,3]
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512DQ-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512DQ-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512DQ-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512DQ-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512DQ-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm8
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm9
+; AVX512DQ-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm7
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = ymm9[u,u,u,u,u,5],zero,ymm9[u,u,u,u,u,6],zero,ymm9[u,u,u,u,u],zero,ymm9[23,u,u,u,u,u],zero,ymm9[24,u,u,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm9[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,u],zero,ymm11[5,u,u,u,u,u],zero,ymm11[6,u,u,u,u,u,23],zero,ymm11[u,u,u,u,u,24],zero,ymm11[u,u,u,u]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; AVX512DQ-NEXT:    vpternlogq $50, %ymm10, %ymm12, %ymm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm10 = ymm9[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm11, %zmm10, %zmm10
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm8[u,u,u,5],zero,ymm8[u,u,u,u,u,6],zero,ymm8[u,u,u,u,u],zero,ymm8[23,u,u,u,u,u],zero,ymm8[24,u,u,u,u,u],zero
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm8[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u],zero,ymm13[5,u,u,u,u,u],zero,ymm13[6,u,u,u,u,u,23],zero,ymm13[u,u,u,u,u,24],zero,ymm13[u,u,u,u,u,25]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm12, %ymm13
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm8[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,8],zero,zero,ymm11[u,u,u,1,9],zero,zero,ymm11[u,u,u,2,10],zero,zero,ymm11[u,u,u,19,27],zero,zero,ymm11[u,u,u,20,28],zero,zero
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm13, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vporq %zmm10, %zmm11, %zmm10
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm7[4],zero,ymm7[u,u,u,u,u,5],zero,ymm7[u,u,u,u,u,6],zero,ymm7[u,u,u,u,u],zero,ymm7[23,u,u,u,u,u],zero,ymm7[24,u,u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[2,3,0,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm12 = zero,ymm12[4,u,u,u,u,u],zero,ymm12[5,u,u,u,u,u],zero,ymm12[6,u,u,u,u,u,23],zero,ymm12[u,u,u,u,u,24],zero,ymm12[u,u]
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; AVX512DQ-NEXT:    vpternlogq $200, %ymm11, %ymm13, %ymm12
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[0,2,0,2]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[u,u,u,u,0,8],zero,ymm11[u,u,u,u,1,9],zero,ymm11[u,u,u,u,18,26],zero,ymm11[u,u,u,u,19,27],zero,ymm11[u,u,u,u]
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
+; AVX512DQ-NEXT:    vpandn %ymm12, %ymm13, %ymm12
+; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm13, %ymm13
+; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm12, %zmm13, %zmm12
+; AVX512DQ-NEXT:    vporq %zmm12, %zmm11, %zmm11
+; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[3,1,1,3]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm8 = ymm8[1],zero,zero,ymm8[u,u,u,10,2],zero,zero,ymm8[u,u,u,11,3],zero,zero,ymm8[u,u,u,20,28],zero,zero,ymm8[u,u,u,21,29],zero,zero,ymm8[u]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,3,1]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm9 = zero,ymm9[1,9,u,u,u],zero,zero,ymm9[2,10,u,u,u],zero,zero,ymm9[3,19,u,u,u],zero,zero,ymm9[28,20,u,u,u],zero,zero,ymm9[29,21,u]
+; AVX512DQ-NEXT:    vpor %ymm8, %ymm9, %ymm8
+; AVX512DQ-NEXT:    vpshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[2,2,3,3]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[1,3,1,3]
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,9],zero,ymm7[u,u,u,u,2,10],zero,ymm7[u,u,u,u,19,27],zero,ymm7[u,u,u,u,20,28],zero,ymm7[u,u,u,u,21]
-; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm6, %ymm7
-; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm7
-; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm5, %zmm7, %zmm4
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,5],zero,ymm3[u,u,u,u,u,6],zero,ymm3[u,u,u,u,u],zero,ymm3[23,u,u,u,u,u],zero,ymm3[24,u,u,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; AVX512DQ-NEXT:    vpternlogq $50, %ymm6, %ymm8, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = zero,zero,ymm3[0,8,u,u,u],zero,zero,ymm3[1,9,u,u,u],zero,zero,ymm3[18,26,u,u,u],zero,zero,ymm3[19,27,u,u,u],zero,zero,ymm3[20,28]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm3, %zmm3
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[u,u,u,5],zero,ymm2[u,u,u,u,u,6],zero,ymm2[u,u,u,u,u],zero,ymm2[23,u,u,u,u,u],zero,ymm2[24,u,u,u,u,u],zero
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm2[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u,u,u,u,25]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm6, %ymm8, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[0,8],zero,zero,ymm2[u,u,u,1,9],zero,zero,ymm2[u,u,u,2,10],zero,zero,ymm2[u,u,u,19,27],zero,zero,ymm2[u,u,u,20,28],zero,zero
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm2, %zmm2
-; AVX512DQ-NEXT:    vporq %zmm3, %zmm2, %zmm2
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,1,0]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; AVX512DQ-NEXT:    vpandn %ymm3, %ymm6, %ymm3
-; AVX512DQ-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,1,0,0,4,5,6,7]
-; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
-; AVX512DQ-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm3 = ymm0[4],zero,ymm0[u,u,u,u,u,5],zero,ymm0[u,u,u,u,u,6],zero,ymm0[u,u,u,u,u],zero,ymm0[23,u,u,u,u,u],zero,ymm0[24,u,u]
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm7 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm7 = zero,ymm7[4,u,u,u,u,u],zero,ymm7[5,u,u,u,u,u],zero,ymm7[6,u,u,u,u,u,23],zero,ymm7[u,u,u,u,u,24],zero,ymm7[u,u]
-; AVX512DQ-NEXT:    vpternlogq $200, %ymm3, %ymm6, %ymm7
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,2]
-; AVX512DQ-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,0,8],zero,ymm0[u,u,u,u,1,9],zero,ymm0[u,u,u,u,18,26],zero,ymm0[u,u,u,u,19,27],zero,ymm0[u,u,u,u]
-; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vporq %zmm1, %zmm0, %zmm0
-; AVX512DQ-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm2, %zmm0
-; AVX512DQ-NEXT:    vmovdqa %xmm5, 96(%rax)
-; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rax)
-; AVX512DQ-NEXT:    vmovdqa %ymm4, 64(%rax)
+; AVX512DQ-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm7
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm7
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512DQ-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512DQ-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512DQ-NEXT:    vinserti32x4 $2, %xmm1, %zmm7, %zmm0
+; AVX512DQ-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512DQ-NEXT:    vmovdqa %ymm0, 64(%rax)
+; AVX512DQ-NEXT:    vmovdqa64 %zmm11, (%rax)
 ; AVX512DQ-NEXT:    vzeroupper
 ; AVX512DQ-NEXT:    retq
 ;
@@ -2280,70 +2275,69 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; AVX512DQ-FCP:       # %bb.0:
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm2
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm3
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm4
-; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm5
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm1
-; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm0
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm5, %ymm4, %ymm6
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm7
-; AVX512DQ-FCP-NEXT:    vinserti128 $1, (%r9), %ymm1, %ymm1
-; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm0, %zmm1, %zmm1
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u],zero,zero,xmm4[12,13,u,u,u],zero,zero,xmm4[14,15,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,12,13],zero,zero,xmm2[u,u,u,14,15],zero,zero,xmm2[u,u,u]
-; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm2, %xmm2
-; AVX512DQ-FCP-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[10],zero,xmm3[u,u,u,u,13,12],zero,xmm3[u,u,u,u,15,14],zero
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = zero,xmm0[13,u,u,u,u],zero,zero,xmm0[14,u,u,u,u],zero,zero,xmm0[15]
-; AVX512DQ-FCP-NEXT:    vpor %xmm4, %xmm3, %xmm3
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm3
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm2 = ymm7[3,1,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[1],zero,zero,ymm2[u,u,u,10,2],zero,zero,ymm2[u,u,u,11,3],zero,zero,ymm2[u,u,u,20,28],zero,zero,ymm2[u,u,u,21,29],zero,zero,ymm2[u]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[1,3,3,1]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,ymm4[1,9,u,u,u],zero,zero,ymm4[2,10,u,u,u],zero,zero,ymm4[3,19,u,u,u],zero,zero,ymm4[28,20,u,u,u],zero,zero,ymm4[29,21,u]
-; AVX512DQ-FCP-NEXT:    vpor %ymm2, %ymm4, %ymm2
-; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,5,6]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [2,2,3,3,2,2,3,3]
-; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm4, %ymm5, %ymm4
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm5 = ymm1[1,3,1,3]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,1,9],zero,ymm5[u,u,u,u,2,10],zero,ymm5[u,u,u,u,19,27],zero,ymm5[u,u,u,u,20,28],zero,ymm5[u,u,u,u,21]
-; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm4, %ymm5
-; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm2, %ymm5
-; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm3, %zmm5, %zmm2
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm4 = ymm6[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm4 = zero,zero,ymm4[0,8,u,u,u],zero,zero,ymm4[1,9,u,u,u],zero,zero,ymm4[18,26,u,u,u],zero,zero,ymm4[19,27,u,u,u],zero,zero,ymm4[20,28]
-; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,5,2,6,1,5,2,6]
-; AVX512DQ-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm6
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u],zero,zero,ymm6[1,5,u,u,u],zero,zero,ymm6[2,6,u,u,u],zero,zero,ymm6[19,23,u,u,u],zero,zero,ymm6[24,28,u,u,u],zero
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm6, %zmm4, %zmm4
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm7[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,8],zero,zero,ymm6[u,u,u,1,9],zero,zero,ymm6[u,u,u,2,10],zero,zero,ymm6[u,u,u,19,27],zero,zero,ymm6[u,u,u,20,28],zero,zero
-; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm5, %ymm7
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,1,5],zero,zero,ymm7[u,u,u,2,6],zero,zero,ymm7[u,u,u,19,23],zero,zero,ymm7[u,u,u,24,28],zero,zero,ymm7[u,u,u,25]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm7, %zmm6, %zmm6
-; AVX512DQ-FCP-NEXT:    vporq %zmm4, %zmm6, %zmm4
-; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[1,1,0,0,4,5,6,7]
-; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,0,1,0,0,0,0]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm6, %ymm7, %ymm6
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm0, %zmm6, %zmm0
-; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[0,2,0,2]
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,0,8],zero,ymm6[u,u,u,u,1,9],zero,ymm6[u,u,u,u,18,26],zero,ymm6[u,u,u,u,19,27],zero,ymm6[u,u,u,u]
-; AVX512DQ-FCP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
-; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,4],zero,ymm1[u,u,u,u,1,5],zero,ymm1[u,u,u,u,2,6],zero,ymm1[u,u,u,u,19,23],zero,ymm1[u,u,u,u,24,28],zero,ymm1[u]
-; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm1, %zmm6, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm1
-; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm1
-; AVX512DQ-FCP-NEXT:    vmovdqa %xmm3, 96(%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
-; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, 64(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rsi), %xmm1
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rdx), %xmm5
+; AVX512DQ-FCP-NEXT:    vmovdqa (%rcx), %xmm6
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r8), %xmm3
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r9), %xmm4
+; AVX512DQ-FCP-NEXT:    vmovdqa (%r10), %xmm2
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm7
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm6, %ymm5, %ymm8
+; AVX512DQ-FCP-NEXT:    vinserti128 $1, %xmm4, %ymm3, %ymm9
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm10 = ymm8[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm10 = zero,zero,ymm10[0,8,u,u,u],zero,zero,ymm10[1,9,u,u,u],zero,zero,ymm10[18,26,u,u,u],zero,zero,ymm10[19,27,u,u,u],zero,zero,ymm10[20,28]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [1,5,2,6,1,5,2,6]
+; AVX512DQ-FCP-NEXT:    # ymm11 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm11, %ymm12
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[u,u,u],zero,zero,ymm12[1,5,u,u,u],zero,zero,ymm12[2,6,u,u,u],zero,zero,ymm12[19,23,u,u,u],zero,zero,ymm12[24,28,u,u,u],zero
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm12, %zmm10, %zmm10
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm12 = ymm12[0,8],zero,zero,ymm12[u,u,u,1,9],zero,zero,ymm12[u,u,u,2,10],zero,zero,ymm12[u,u,u,19,27],zero,zero,ymm12[u,u,u,20,28],zero,zero
+; AVX512DQ-FCP-NEXT:    vpermd %ymm7, %ymm11, %ymm13
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,1,5],zero,zero,ymm13[u,u,u,2,6],zero,zero,ymm13[u,u,u,19,23],zero,zero,ymm13[u,u,u,24,28],zero,zero,ymm13[u,u,u,25]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512DQ-FCP-NEXT:    vporq %zmm10, %zmm12, %zmm10
+; AVX512DQ-FCP-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm2[1,1,0,0,4,5,6,7]
+; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [0,1,0,1,0,0,0,0]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm12, %ymm13, %ymm12
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm13 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,1,0]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm13, %zmm12, %zmm12
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm13 = ymm9[0,2,0,2]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm13 = ymm13[u,u,u,u,0,8],zero,ymm13[u,u,u,u,1,9],zero,ymm13[u,u,u,u,18,26],zero,ymm13[u,u,u,u,19,27],zero,ymm13[u,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm9, %ymm11, %ymm11
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm11 = ymm11[0,4],zero,ymm11[u,u,u,u,1,5],zero,ymm11[u,u,u,u,2,6],zero,ymm11[u,u,u,u,19,23],zero,ymm11[u,u,u,u,24,28],zero,ymm11[u]
+; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm11, %zmm13, %zmm11
+; AVX512DQ-FCP-NEXT:    vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm12, %zmm11
+; AVX512DQ-FCP-NEXT:    vpternlogd $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm10, %zmm11
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,ymm7[u,u,u,10,2],zero,zero,ymm7[u,u,u,11,3],zero,zero,ymm7[u,u,u,20,28],zero,zero,ymm7[u,u,u,21,29],zero,zero,ymm7[u]
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9,u,u,u],zero,zero,ymm8[2,10,u,u,u],zero,zero,ymm8[3,19,u,u,u],zero,zero,ymm8[28,20,u,u,u],zero,zero,ymm8[29,21,u]
+; AVX512DQ-FCP-NEXT:    vpor %ymm7, %ymm8, %ymm7
+; AVX512DQ-FCP-NEXT:    vpshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,4,5,5,6]
+; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [2,2,3,3,2,2,3,3]
+; AVX512DQ-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
+; AVX512DQ-FCP-NEXT:    vpermd %ymm8, %ymm10, %ymm8
+; AVX512DQ-FCP-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[1,3,1,3]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,1,9],zero,ymm9[u,u,u,u,2,10],zero,ymm9[u,u,u,u,19,27],zero,ymm9[u,u,u,u,20,28],zero,ymm9[u,u,u,u,21]
+; AVX512DQ-FCP-NEXT:    vpternlogq $244, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm9
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm7, %ymm9
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[u,u],zero,zero,xmm5[12,13,u,u,u],zero,zero,xmm5[14,15,u,u,u]
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,12,13],zero,zero,xmm0[u,u,u,14,15],zero,zero,xmm0[u,u,u]
+; AVX512DQ-FCP-NEXT:    vpor %xmm5, %xmm0, %xmm0
+; AVX512DQ-FCP-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[10],zero,xmm1[u,u,u,u,13,12],zero,xmm1[u,u,u,u,15,14],zero
+; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,xmm2[13,u,u,u,u],zero,zero,xmm2[14,u,u,u,u],zero,zero,xmm2[15]
+; AVX512DQ-FCP-NEXT:    vpor %xmm2, %xmm1, %xmm1
+; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
+; AVX512DQ-FCP-NEXT:    vinserti32x4 $2, %xmm1, %zmm9, %zmm0
+; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, 96(%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, (%rax)
+; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, 64(%rax)
 ; AVX512DQ-FCP-NEXT:    vzeroupper
 ; AVX512DQ-FCP-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index 11f422d671541..99e8cdb179c8d 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -314,8 +314,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512F-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -324,8 +324,8 @@ define void @vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2(ptr %in.v
 ;
 ; AVX512DQ-LABEL: vec64_i16_widen_to_i32_factor2_broadcast_to_v2i32_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,10,11,0,1,14,15,u,u,u,u,u,u,u,u]
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
@@ -981,7 +981,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512F-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512F-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512F-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512F-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512F-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512F-NEXT:    vzeroupper
@@ -992,7 +992,7 @@ define void @vec128_i32_widen_to_i64_factor2_broadcast_to_v2i64_factor2(ptr %in.
 ; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,5,0,7]
 ; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm1
 ; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermd %zmm1, %zmm0, %zmm0
+; AVX512DQ-NEXT:    vpermd %ymm1, %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vpaddb (%rdx), %ymm0, %ymm0
 ; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rcx)
 ; AVX512DQ-NEXT:    vzeroupper
@@ -4026,10 +4026,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512F-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512F-FAST:       # %bb.0:
-; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512F-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512F-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512F-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512F-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4062,10 +4062,10 @@ define void @vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6(ptr %in.
 ;
 ; AVX512DQ-FAST-LABEL: vec384_i16_widen_to_i64_factor4_broadcast_to_v6i64_factor6:
 ; AVX512DQ-FAST:       # %bb.0:
-; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-FAST-NEXT:    vmovdqa 48(%rdi), %xmm1
 ; AVX512DQ-FAST-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX512DQ-FAST-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-FAST-NEXT:    vpbroadcastq %xmm0, %ymm2
 ; AVX512DQ-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
 ; AVX512DQ-FAST-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm1
@@ -4541,9 +4541,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512F-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512F-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512F-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512F-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
@@ -4559,9 +4559,9 @@ define void @vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2(ptr %
 ;
 ; AVX512DQ-LABEL: vec384_i16_widen_to_i192_factor12_broadcast_to_v2i192_factor2:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
-; AVX512DQ-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512DQ-NEXT:    vmovdqa 48(%rdi), %xmm1
+; AVX512DQ-NEXT:    vpaddb (%rsi), %xmm0, %xmm0
 ; AVX512DQ-NEXT:    vpaddb 48(%rsi), %xmm1, %xmm1
 ; AVX512DQ-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5,6,7]
 ; AVX512DQ-NEXT:    vpbroadcastw %xmm0, %ymm0

From 71b9f6648222771470473431bc8ef2a2c25e872c Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Tue, 16 Apr 2024 07:34:27 -0400
Subject: [PATCH 571/886] [clang][Index] Use canonical function parameter types
 in USRs (#68222)

This is necessary to ensure that functions declared in different
translation units whose parameter types only differ in top-level
cv-qualification generate the same USR.

For example:
```
// A.cpp
void f(const int x); // c:@F@f#1I#

// B.cpp
void f(int x);       // c:@F@f#I#
```
With this patch, the USR for both functions will be
`c:@F@f#I#`.
---
 clang/lib/Index/USRGeneration.cpp  |  9 ++++++---
 clang/test/Index/USR/func-type.cpp | 12 ++++++++++++
 2 files changed, 18 insertions(+), 3 deletions(-)

diff --git a/clang/lib/Index/USRGeneration.cpp b/clang/lib/Index/USRGeneration.cpp
index 5acc86191f8f9..31c4a3345c09d 100644
--- a/clang/lib/Index/USRGeneration.cpp
+++ b/clang/lib/Index/USRGeneration.cpp
@@ -267,10 +267,13 @@ void USRGenerator::VisitFunctionDecl(const FunctionDecl *D) {
     Out << '>';
   }
 
+  QualType CanonicalType = D->getType().getCanonicalType();
   // Mangle in type information for the arguments.
-  for (auto *PD : D->parameters()) {
-    Out << '#';
-    VisitType(PD->getType());
+  if (const auto *FPT = CanonicalType->getAs<FunctionProtoType>()) {
+    for (QualType PT : FPT->param_types()) {
+      Out << '#';
+      VisitType(PT);
+    }
   }
   if (D->isVariadic())
     Out << '.';
diff --git a/clang/test/Index/USR/func-type.cpp b/clang/test/Index/USR/func-type.cpp
index ff1cd37a7fc42..459a8cd6da558 100644
--- a/clang/test/Index/USR/func-type.cpp
+++ b/clang/test/Index/USR/func-type.cpp
@@ -16,3 +16,15 @@ void Func( void (* (*)(int, int))(int, int) );
 // CHECK: {{[0-9]+}}:6 | function/C | Func | c:@F@Func#*F*Fv(#I#I)(#I#I)# |
 void Func( void (* (*)(int, int, int))(int) );
 // CHECK: {{[0-9]+}}:6 | function/C | Func | c:@F@Func#*F*Fv(#I)(#I#I#I)# |
+
+// Functions with parameter types that only differ in top-level cv-qualification should generate the same USR.
+
+void f( const int );
+// CHECK: {{[0-9]+}}:6 | function/C | f | c:@F@f#I# |
+void f( int );
+// CHECK: {{[0-9]+}}:6 | function/C | f | c:@F@f#I# |
+
+void g( int );
+// CHECK: {{[0-9]+}}:6 | function/C | g | c:@F@g#I# |
+void g( const int );
+// CHECK: {{[0-9]+}}:6 | function/C | g | c:@F@g#I# |

From dadcaf82274805456b7d85131cf94f921b5398b7 Mon Sep 17 00:00:00 2001
From: Benjamin Maxwell <benjamin.maxwell@arm.com>
Date: Tue, 16 Apr 2024 12:54:01 +0100
Subject: [PATCH 572/886] [mlir][ArmSME] Support decomposing constant splats
 into ArmSME tiles (#88762)

This adds a simple rewrite/legalization to decompose constant splats
larger than a single ArmSME tile into multiple SME virtual tile sized
splats. E.g. a constant splat to `vector<[8]x[8]xi32>` would decompose
into four `vector<[4]x[4]xi32>` splats.
---
 .../ArmSME/Transforms/VectorLegalization.cpp  | 32 ++++++++++++++++++-
 .../Dialect/ArmSME/vector-legalization.mlir   | 11 +++++++
 2 files changed, 42 insertions(+), 1 deletion(-)

diff --git a/mlir/lib/Dialect/ArmSME/Transforms/VectorLegalization.cpp b/mlir/lib/Dialect/ArmSME/Transforms/VectorLegalization.cpp
index 31500c62c0d60..b595c6dd8a684 100644
--- a/mlir/lib/Dialect/ArmSME/Transforms/VectorLegalization.cpp
+++ b/mlir/lib/Dialect/ArmSME/Transforms/VectorLegalization.cpp
@@ -165,6 +165,35 @@ int getNumberOfSMETilesForVectorType(VectorType type) {
   return (vectorRows * vectorCols) / (minNumElts * minNumElts);
 }
 
+/// Legalize `arith.constant dense<value>` splat operations to fit within SME
+/// tiles by decomposing them into tile-sized operations.
+struct LegalizeArithConstantOpsByDecomposition
+    : public OneToNOpConversionPattern<arith::ConstantOp> {
+  using OneToNOpConversionPattern::OneToNOpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(arith::ConstantOp constantOp, OpAdaptor adaptor,
+                  OneToNPatternRewriter &rewriter) const override {
+    auto vectorType = dyn_cast<VectorType>(constantOp.getType());
+    auto denseAttr = dyn_cast<DenseElementsAttr>(constantOp.getValueAttr());
+    if (!vectorType || !denseAttr || !denseAttr.isSplat())
+      return failure();
+
+    if (!isMultipleOfSMETileVectorType(vectorType))
+      return rewriter.notifyMatchFailure(constantOp,
+                                         kMatchFailureNotSMETileTypeMultiple);
+
+    auto smeTileType = getSMETileTypeForElement(vectorType.getElementType());
+    auto tileCount = getNumberOfSMETilesForVectorType(vectorType);
+    auto tileSplat = rewriter.create<arith::ConstantOp>(
+        constantOp.getLoc(), denseAttr.resizeSplat(smeTileType));
+    rewriter.replaceOp(constantOp, SmallVector<Value>(tileCount, tileSplat),
+                       adaptor.getResultMapping());
+
+    return success();
+  }
+};
+
 /// Legalize `vector.outerproduct` operations to fit within SME tiles by
 /// decomposing them into tile-sized operations.
 struct LegalizeVectorOuterProductOpsByDecomposition
@@ -637,7 +666,8 @@ struct VectorLegalizationPass
     // Note: High benefit to ensure masked outer products are lowered first.
     patterns.add<LegalizeMaskedVectorOuterProductOpsByDecomposition>(
         converter, context, 1024);
-    patterns.add<LegalizeVectorOuterProductOpsByDecomposition,
+    patterns.add<LegalizeArithConstantOpsByDecomposition,
+                 LegalizeVectorOuterProductOpsByDecomposition,
                  LegalizeTransferReadOpsByDecomposition,
                  LegalizeTransferWriteOpsByDecomposition>(converter, context);
     populateFuncTypeConversionPatterns(converter, patterns);
diff --git a/mlir/test/Dialect/ArmSME/vector-legalization.mlir b/mlir/test/Dialect/ArmSME/vector-legalization.mlir
index f8be697548c19..f43ef1cce787c 100644
--- a/mlir/test/Dialect/ArmSME/vector-legalization.mlir
+++ b/mlir/test/Dialect/ArmSME/vector-legalization.mlir
@@ -433,3 +433,14 @@ func.func @lift_illegal_1d_shape_cast_to_memory(%a: index, %b: index, %memref: m
   %cast = vector.shape_cast %illegalRead : vector<[4]x1xf32> to vector<[4]xf32>
   return %cast : vector<[4]xf32>
 }
+
+// -----
+
+// CHECK-LABEL: @multi_tile_splat
+func.func @multi_tile_splat() -> vector<[8]x[8]xi32>
+{
+  // CHECK: %[[SPLAT:.*]] = arith.constant dense<42> : vector<[4]x[4]xi32>
+  // CHECK-NEXT: return %[[SPLAT]], %[[SPLAT]], %[[SPLAT]], %[[SPLAT]] : vector<[4]x[4]xi32>, vector<[4]x[4]xi32>, vector<[4]x[4]xi32>, vector<[4]x[4]xi32>
+  %0 = arith.constant dense<42> : vector<[8]x[8]xi32>
+  return %0 : vector<[8]x[8]xi32>
+}

From 711df7b0ae4a9ea45e431d5c0ff4a0c8b2e732c1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Andrzej=20Warzy=C5=84ski?= <andrzej.warzynski@arm.com>
Date: Tue, 16 Apr 2024 12:57:59 +0100
Subject: [PATCH 573/886] [vector][mlir] Restrict vector.shuffle to fixed-width
 vectors (#88733)

At the moment there is no support for vector.shuffle for scalable
vectors - various hooks/helpers related to `vector.shuffle` simply
ignore the scalable flags (e.g. ` ShuffleOp::inferReturnTypes`).

This is unlikely to change any time soon (vector shuffles are known to
be tricky for scalable vectors), hence this patch restricts
`vector.shuffle` to fixed width vectors.
---
 mlir/include/mlir/Dialect/Vector/IR/VectorOps.td | 4 +++-
 mlir/test/Dialect/Vector/canonicalize.mlir       | 8 --------
 mlir/test/Dialect/Vector/invalid.mlir            | 7 +++++++
 3 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
index 147bc2354977d..332b5ad08ced9 100644
--- a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
+++ b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
@@ -420,7 +420,7 @@ def Vector_ShuffleOp :
      PredOpTrait<"second operand v2 and result have same element type",
                  TCresVTEtIsSameAsOpBase<0, 1>>,
      InferTypeOpAdaptor]>,
-     Arguments<(ins AnyVectorOfAnyRank:$v1, AnyVectorOfAnyRank:$v2,
+     Arguments<(ins AnyFixedVector:$v1, AnyFixedVector:$v2,
                     I64ArrayAttr:$mask)>,
      Results<(outs AnyVector:$vector)> {
   let summary = "shuffle operation";
@@ -444,6 +444,8 @@ def Vector_ShuffleOp :
       mask values must be within range, viz. given two k-D operands v1 and v2
       above, all mask values are in the range [0,s_1+t_1)
 
+    Note, scalable vectors are not supported.
+
     Example:
 
     ```mlir
diff --git a/mlir/test/Dialect/Vector/canonicalize.mlir b/mlir/test/Dialect/Vector/canonicalize.mlir
index 627ac54cf145b..61a5f2a96e1c1 100644
--- a/mlir/test/Dialect/Vector/canonicalize.mlir
+++ b/mlir/test/Dialect/Vector/canonicalize.mlir
@@ -1943,14 +1943,6 @@ func.func @shuffle_nofold1(%v0 : vector<4xi32>, %v1 : vector<2xi32>) -> vector<5
   return %shuffle : vector<5xi32>
 }
 
-// CHECK-LABEL: func @shuffle_nofold2
-//       CHECK:   %[[V:.+]] = vector.shuffle %arg0, %arg1 [0, 1, 2, 3] : vector<[4]xi32>, vector<[2]xi32>
-//       CHECK:   return %[[V]]
-func.func @shuffle_nofold2(%v0 : vector<[4]xi32>, %v1 : vector<[2]xi32>) -> vector<4xi32> {
-  %shuffle = vector.shuffle %v0, %v1 [0, 1, 2, 3] : vector<[4]xi32>, vector<[2]xi32>
-  return %shuffle : vector<4xi32>
-}
-
 // -----
 
 // CHECK-LABEL: func @transpose_scalar_broadcast1
diff --git a/mlir/test/Dialect/Vector/invalid.mlir b/mlir/test/Dialect/Vector/invalid.mlir
index c16f1cb2876db..c9f7e9c6e2fb0 100644
--- a/mlir/test/Dialect/Vector/invalid.mlir
+++ b/mlir/test/Dialect/Vector/invalid.mlir
@@ -84,6 +84,13 @@ func.func @shuffle_index_out_of_range(%arg0: vector<2xf32>, %arg1: vector<2xf32>
 
 // -----
 
+func.func @shuffle_scalable_vec(%arg0: vector<[2]xf32>, %arg1: vector<[2]xf32>) {
+  // expected-error@+1 {{'vector.shuffle' op operand #0 must be fixed-length vector of any type values}}
+  %1 = vector.shuffle %arg0, %arg1 [0, 1, 2, 3] : vector<[2]xf32>, vector<[2]xf32>
+}
+
+// -----
+
 func.func @shuffle_empty_mask(%arg0: vector<2xf32>, %arg1: vector<2xf32>) {
   // expected-error@+1 {{'vector.shuffle' op invalid mask length}}
   %1 = vector.shuffle %arg0, %arg1 [] : vector<2xf32>, vector<2xf32>

From 75244a1043d2be5003dea6914d5edc940c437cd5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Tue, 16 Apr 2024 09:44:54 +0200
Subject: [PATCH 574/886] [clang][Interp] Implement align builtins

__builtin_is_aligned
__builtin_is_align_up
__builtin_is_align_down
---
 clang/lib/AST/Interp/InterpBuiltin.cpp      | 118 +++++++++
 clang/test/AST/Interp/builtin-align-cxx.cpp | 258 ++++++++++++++++++++
 2 files changed, 376 insertions(+)
 create mode 100644 clang/test/AST/Interp/builtin-align-cxx.cpp

diff --git a/clang/lib/AST/Interp/InterpBuiltin.cpp b/clang/lib/AST/Interp/InterpBuiltin.cpp
index 984ba4f7f2689..f562f9e1cb19f 100644
--- a/clang/lib/AST/Interp/InterpBuiltin.cpp
+++ b/clang/lib/AST/Interp/InterpBuiltin.cpp
@@ -977,6 +977,117 @@ static bool interp__builtin_complex(InterpState &S, CodePtr OpPC,
   return true;
 }
 
+/// __builtin_is_aligned()
+/// __builtin_align_up()
+/// __builtin_align_down()
+/// The first parameter is either an integer or a pointer.
+/// The second parameter is the requested alignment as an integer.
+static bool interp__builtin_is_aligned_up_down(InterpState &S, CodePtr OpPC,
+                                               const InterpFrame *Frame,
+                                               const Function *Func,
+                                               const CallExpr *Call) {
+  unsigned BuiltinOp = Func->getBuiltinID();
+  unsigned CallSize = callArgSize(S, Call);
+
+  PrimType AlignmentT = *S.Ctx.classify(Call->getArg(1));
+  const APSInt &Alignment = peekToAPSInt(S.Stk, AlignmentT);
+
+  if (Alignment < 0 || !Alignment.isPowerOf2()) {
+    S.FFDiag(Call, diag::note_constexpr_invalid_alignment) << Alignment;
+    return false;
+  }
+  unsigned SrcWidth = S.getCtx().getIntWidth(Call->getArg(0)->getType());
+  APSInt MaxValue(APInt::getOneBitSet(SrcWidth, SrcWidth - 1));
+  if (APSInt::compareValues(Alignment, MaxValue) > 0) {
+    S.FFDiag(Call, diag::note_constexpr_alignment_too_big)
+        << MaxValue << Call->getArg(0)->getType() << Alignment;
+    return false;
+  }
+
+  // The first parameter is either an integer or a pointer (but not a function
+  // pointer).
+  PrimType FirstArgT = *S.Ctx.classify(Call->getArg(0));
+
+  if (isIntegralType(FirstArgT)) {
+    const APSInt &Src = peekToAPSInt(S.Stk, FirstArgT, CallSize);
+    APSInt Align = Alignment.extOrTrunc(Src.getBitWidth());
+    if (BuiltinOp == Builtin::BI__builtin_align_up) {
+      APSInt AlignedVal =
+          APSInt((Src + (Align - 1)) & ~(Align - 1), Src.isUnsigned());
+      pushInteger(S, AlignedVal, Call->getType());
+    } else if (BuiltinOp == Builtin::BI__builtin_align_down) {
+      APSInt AlignedVal = APSInt(Src & ~(Align - 1), Src.isUnsigned());
+      pushInteger(S, AlignedVal, Call->getType());
+    } else {
+      assert(*S.Ctx.classify(Call->getType()) == PT_Bool);
+      S.Stk.push<Boolean>((Src & (Align - 1)) == 0);
+    }
+    return true;
+  }
+
+  assert(FirstArgT == PT_Ptr);
+  const Pointer &Ptr = S.Stk.peek<Pointer>(CallSize);
+
+  unsigned PtrOffset = Ptr.getByteOffset();
+  PtrOffset = Ptr.getIndex();
+  CharUnits BaseAlignment =
+      S.getCtx().getDeclAlign(Ptr.getDeclDesc()->asValueDecl());
+  CharUnits PtrAlign =
+      BaseAlignment.alignmentAtOffset(CharUnits::fromQuantity(PtrOffset));
+
+  if (BuiltinOp == Builtin::BI__builtin_is_aligned) {
+    if (PtrAlign.getQuantity() >= Alignment) {
+      S.Stk.push<Boolean>(true);
+      return true;
+    }
+    // If the alignment is not known to be sufficient, some cases could still
+    // be aligned at run time. However, if the requested alignment is less or
+    // equal to the base alignment and the offset is not aligned, we know that
+    // the run-time value can never be aligned.
+    if (BaseAlignment.getQuantity() >= Alignment &&
+        PtrAlign.getQuantity() < Alignment) {
+      S.Stk.push<Boolean>(false);
+      return true;
+    }
+
+    S.FFDiag(Call->getArg(0), diag::note_constexpr_alignment_compute)
+        << Alignment;
+    return false;
+  }
+
+  assert(BuiltinOp == Builtin::BI__builtin_align_down ||
+         BuiltinOp == Builtin::BI__builtin_align_up);
+
+  // For align_up/align_down, we can return the same value if the alignment
+  // is known to be greater or equal to the requested value.
+  if (PtrAlign.getQuantity() >= Alignment) {
+    S.Stk.push<Pointer>(Ptr);
+    return true;
+  }
+
+  // The alignment could be greater than the minimum at run-time, so we cannot
+  // infer much about the resulting pointer value. One case is possible:
+  // For `_Alignas(32) char buf[N]; __builtin_align_down(&buf[idx], 32)` we
+  // can infer the correct index if the requested alignment is smaller than
+  // the base alignment so we can perform the computation on the offset.
+  if (BaseAlignment.getQuantity() >= Alignment) {
+    assert(Alignment.getBitWidth() <= 64 &&
+           "Cannot handle > 64-bit address-space");
+    uint64_t Alignment64 = Alignment.getZExtValue();
+    CharUnits NewOffset =
+        CharUnits::fromQuantity(BuiltinOp == Builtin::BI__builtin_align_down
+                                    ? llvm::alignDown(PtrOffset, Alignment64)
+                                    : llvm::alignTo(PtrOffset, Alignment64));
+
+    S.Stk.push<Pointer>(Ptr.atIndex(NewOffset.getQuantity()));
+    return true;
+  }
+
+  // Otherwise, we cannot constant-evaluate the result.
+  S.FFDiag(Call->getArg(0), diag::note_constexpr_alignment_adjust) << Alignment;
+  return false;
+}
+
 bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const Function *F,
                       const CallExpr *Call) {
   const InterpFrame *Frame = S.Current;
@@ -1291,6 +1402,13 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const Function *F,
       return false;
     break;
 
+  case Builtin::BI__builtin_is_aligned:
+  case Builtin::BI__builtin_align_up:
+  case Builtin::BI__builtin_align_down:
+    if (!interp__builtin_is_aligned_up_down(S, OpPC, Frame, F, Call))
+      return false;
+    break;
+
   default:
     S.FFDiag(S.Current->getLocation(OpPC),
              diag::note_invalid_subexpr_in_const_expr)
diff --git a/clang/test/AST/Interp/builtin-align-cxx.cpp b/clang/test/AST/Interp/builtin-align-cxx.cpp
new file mode 100644
index 0000000000000..62d73dba929b2
--- /dev/null
+++ b/clang/test/AST/Interp/builtin-align-cxx.cpp
@@ -0,0 +1,258 @@
+// C++-specific checks for the alignment builtins
+// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -std=c++11 %s -fsyntax-only -verify=expected,both -fexperimental-new-constant-interpreter
+// RUN: %clang_cc1 -triple=x86_64-unknown-unknown -std=c++11 %s -fsyntax-only -verify=ref,both
+
+
+/// This is just a copy of the one from test/SemaCXX/ with some of the
+/// diagnostic output adapted.
+/// Also, align32array has an initializer now, which means it's not just
+/// a dummy pointer for us and we do actually have type information for it.
+/// In the future, we need to retain type information for dummy pointers as
+/// well, so here is a test that will break once we do that:
+namespace {
+  _Alignas(32) char heh[4];
+  static_assert(!__builtin_is_aligned(&heh[1], 4), ""); // expected-error {{failed}}
+}
+
+
+// Check that we don't crash when using dependent types in __builtin_align:
+template <typename a, a b>
+void *c(void *d) { // both-note{{candidate template ignored}}
+  return __builtin_align_down(d, b);
+}
+
+struct x {};
+x foo;
+void test(void *value) {
+  c<int, 16>(value);
+  c<struct x, foo>(value); // both-error{{no matching function for call to 'c'}}
+}
+
+template <typename T, long Alignment, long ArraySize = 16>
+void test_templated_arguments() {
+  T array[ArraySize];                                                           // both-error{{variable has incomplete type 'fwddecl'}}
+  static_assert(__is_same(decltype(__builtin_align_up(array, Alignment)), T *), // both-error{{requested alignment is not a power of 2}}
+                "return type should be the decayed array type");
+  static_assert(__is_same(decltype(__builtin_align_down(array, Alignment)), T *),
+                "return type should be the decayed array type");
+  static_assert(__is_same(decltype(__builtin_is_aligned(array, Alignment)), bool),
+                "return type should be bool");
+  T *x1 = __builtin_align_up(array, Alignment);
+  T *x2 = __builtin_align_down(array, Alignment);
+  bool x3 = __builtin_align_up(array, Alignment);
+}
+
+void test() {
+  test_templated_arguments<int, 32>(); // fine
+  test_templated_arguments<struct fwddecl, 16>();
+  // both-note@-1{{in instantiation of function template specialization 'test_templated_arguments<fwddecl, 16L, 16L>'}}
+  // both-note@-2{{forward declaration of 'fwddecl'}}
+  test_templated_arguments<int, 7>(); // invalid alignment value
+  // both-note@-1{{in instantiation of function template specialization 'test_templated_arguments<int, 7L, 16L>'}}
+}
+
+template <typename T, long ArraySize>
+void test_incorrect_alignment_without_instatiation(T value) {
+  int array[32];
+  static_assert(__is_same(decltype(__builtin_align_up(array, 31)), int *), // both-error{{requested alignment is not a power of 2}}
+                "return type should be the decayed array type");
+  static_assert(__is_same(decltype(__builtin_align_down(array, 7)), int *), // both-error{{requested alignment is not a power of 2}}
+                "return type should be the decayed array type");
+  static_assert(__is_same(decltype(__builtin_is_aligned(array, -1)), bool), // both-error{{requested alignment must be 1 or greater}}
+                "return type should be bool");
+  __builtin_align_up(array);       // both-error{{too few arguments to function call, expected 2, have 1}}
+  __builtin_align_up(array, 31);   // both-error{{requested alignment is not a power of 2}}
+  __builtin_align_down(array, 31); // both-error{{requested alignment is not a power of 2}}
+  __builtin_align_up(array, 31);   // both-error{{requested alignment is not a power of 2}}
+  __builtin_align_up(value, 31);   // This shouldn't want since the type is dependent
+  __builtin_align_up(value);       // Same here
+
+  __builtin_align_up(array, sizeof(sizeof(value)) - 1); // both-error{{requested alignment is not a power of 2}}
+  __builtin_align_up(array, value); // no diagnostic as the alignment is value dependent.
+  (void)__builtin_align_up(array, ArraySize); // The same above here
+}
+
+// The original fix for the issue above broke some legitimate code.
+// Here is a regression test:
+typedef __SIZE_TYPE__ size_t;
+void *allocate_impl(size_t size);
+template <typename T>
+T *allocate() {
+  constexpr size_t allocation_size =
+      __builtin_align_up(sizeof(T), sizeof(void *));
+  return static_cast<T *>(
+      __builtin_assume_aligned(allocate_impl(allocation_size), sizeof(void *)));
+}
+struct Foo {
+  int value;
+};
+void *test2() {
+  return allocate<struct Foo>();
+}
+
+// Check that pointers-to-members cannot be used:
+class MemPtr {
+public:
+  int data;
+  void func();
+  virtual void vfunc();
+};
+void test_member_ptr() {
+  __builtin_align_up(&MemPtr::data, 64);    // both-error{{operand of type 'int MemPtr::*' where arithmetic or pointer type is required}}
+  __builtin_align_down(&MemPtr::func, 64);  // both-error{{operand of type 'void (MemPtr::*)()' where arithmetic or pointer type is required}}
+  __builtin_is_aligned(&MemPtr::vfunc, 64); // both-error{{operand of type 'void (MemPtr::*)()' where arithmetic or pointer type is required}}
+}
+
+void test_references(Foo &i) {
+  // Check that the builtins look at the referenced type rather than the reference itself.
+  (void)__builtin_align_up(i, 64);                            // both-error{{operand of type 'Foo' where arithmetic or pointer type is required}}
+  (void)__builtin_align_up(static_cast<Foo &>(i), 64);        // both-error{{operand of type 'Foo' where arithmetic or pointer type is required}}
+  (void)__builtin_align_up(static_cast<const Foo &>(i), 64);  // both-error{{operand of type 'const Foo' where arithmetic or pointer type is required}}
+  (void)__builtin_align_up(static_cast<Foo &&>(i), 64);       // both-error{{operand of type 'Foo' where arithmetic or pointer type is required}}
+  (void)__builtin_align_up(static_cast<const Foo &&>(i), 64); // both-error{{operand of type 'const Foo' where arithmetic or pointer type is required}}
+  (void)__builtin_align_up(&i, 64);
+}
+
+// Check that constexpr wrapper functions can be constant-evaluated.
+template <typename T>
+constexpr bool wrap_is_aligned(T ptr, long align) {
+  return __builtin_is_aligned(ptr, align);
+  // both-note@-1{{requested alignment -3 is not a positive power of two}}
+  // both-note@-2{{requested alignment 19 is not a positive power of two}}
+  // both-note@-3{{requested alignment must be 128 or less for type 'char'; 4194304 is invalid}}
+}
+template <typename T>
+constexpr T wrap_align_up(T ptr, long align) {
+  return __builtin_align_up(ptr, align);
+  // both-note@-1{{requested alignment -2 is not a positive power of two}}
+  // both-note@-2{{requested alignment 18 is not a positive power of two}}
+  // both-note@-3{{requested alignment must be 2147483648 or less for type 'int'; 8589934592 is invalid}}
+  // both-error@-4{{operand of type 'bool' where arithmetic or pointer type is required}}
+}
+
+template <typename T>
+constexpr T wrap_align_down(T ptr, long align) {
+  return __builtin_align_down(ptr, align);
+  // both-note@-1{{requested alignment -1 is not a positive power of two}}
+  // both-note@-2{{requested alignment 17 is not a positive power of two}}
+  // both-note@-3{{requested alignment must be 32768 or less for type 'short'; 1048576 is invalid}}
+}
+
+constexpr int a1 = wrap_align_up(22, 32);
+static_assert(a1 == 32, "");
+constexpr int a2 = wrap_align_down(22, 16);
+static_assert(a2 == 16, "");
+constexpr bool a3 = wrap_is_aligned(22, 32);
+static_assert(!a3, "");
+static_assert(wrap_align_down(wrap_align_up(22, 16), 32) == 32, "");
+static_assert(wrap_is_aligned(wrap_align_down(wrap_align_up(22, 16), 32), 32), "");
+static_assert(!wrap_is_aligned(wrap_align_down(wrap_align_up(22, 16), 32), 64), "");
+
+constexpr long const_value(long l) { return l; }
+// Check some invalid values during constant-evaluation
+static_assert(wrap_align_down(1, const_value(-1)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_align_up(1, const_value(-2)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_is_aligned(1, const_value(-3)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_align_down(1, const_value(17)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_align_up(1, const_value(18)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_is_aligned(1, const_value(19)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+
+// Check invalid values for smaller types:
+static_assert(wrap_align_down(static_cast<short>(1), const_value(1 << 20)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to }}
+// Check invalid boolean type
+static_assert(wrap_align_up(static_cast<int>(1), const_value(1ull << 33)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+static_assert(wrap_is_aligned(static_cast<char>(1), const_value(1 << 22)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in call to}}
+
+// Check invalid boolean type
+static_assert(wrap_align_up(static_cast<bool>(1), const_value(1 << 21)), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{in instantiation of function template specialization 'wrap_align_up<bool>' requested here}}
+
+// Check constant evaluation for pointers:
+_Alignas(32) char align32array[128] = {};
+static_assert(&align32array[0] == &align32array[0], "");
+// __builtin_align_up/down can be constant evaluated as a no-op for values
+// that are known to have greater alignment:
+static_assert(__builtin_align_up(&align32array[0], 32) == &align32array[0], "");
+static_assert(__builtin_align_up(&align32array[0], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[0], 4) == __builtin_align_up(&align32array[0], 8), "");
+// But it can not be evaluated if the alignment is greater than the minimum
+// known alignment, since in that case the value might be the same if it happens
+// to actually be aligned to 64 bytes at run time.
+static_assert(&align32array[0] == __builtin_align_up(&align32array[0], 64), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{cannot constant evaluate the result of adjusting alignment to 64}}
+static_assert(__builtin_align_up(&align32array[0], 64) == __builtin_align_up(&align32array[0], 64), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{cannot constant evaluate the result of adjusting alignment to 64}}
+
+// However, we can compute in case the requested alignment is less than the
+// base alignment:
+static_assert(__builtin_align_up(&align32array[0], 4) == &align32array[0], "");
+static_assert(__builtin_align_up(&align32array[1], 4) == &align32array[4], "");
+static_assert(__builtin_align_up(&align32array[2], 4) == &align32array[4], "");
+static_assert(__builtin_align_up(&align32array[3], 4) == &align32array[4], "");
+static_assert(__builtin_align_up(&align32array[4], 4) == &align32array[4], "");
+static_assert(__builtin_align_up(&align32array[5], 4) == &align32array[8], "");
+static_assert(__builtin_align_up(&align32array[6], 4) == &align32array[8], "");
+static_assert(__builtin_align_up(&align32array[7], 4) == &align32array[8], "");
+static_assert(__builtin_align_up(&align32array[8], 4) == &align32array[8], "");
+
+static_assert(__builtin_align_down(&align32array[0], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[1], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[2], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[3], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[4], 4) == &align32array[4], "");
+static_assert(__builtin_align_down(&align32array[5], 4) == &align32array[4], "");
+static_assert(__builtin_align_down(&align32array[6], 4) == &align32array[4], "");
+static_assert(__builtin_align_down(&align32array[7], 4) == &align32array[4], "");
+static_assert(__builtin_align_down(&align32array[8], 4) == &align32array[8], "");
+
+// Achieving the same thing using casts to uintptr_t is not allowed:
+static_assert((char *)((__UINTPTR_TYPE__)&align32array[7] & ~3) == &align32array[4], ""); // both-error{{not an integral constant expression}} \
+                                                                                          // expected-note {{cast that performs the conversions of a reinterpret_cast is not allowed in a constant expression}}
+
+static_assert(__builtin_align_down(&align32array[1], 4) == &align32array[0], "");
+static_assert(__builtin_align_down(&align32array[1], 64) == &align32array[0], ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{cannot constant evaluate the result of adjusting alignment to 64}}
+
+// Add some checks for __builtin_is_aligned:
+static_assert(__builtin_is_aligned(&align32array[0], 32), "");
+static_assert(__builtin_is_aligned(&align32array[4], 4), "");
+// We cannot constant evaluate whether the array is aligned to > 32 since this
+// may well be true at run time.
+static_assert(!__builtin_is_aligned(&align32array[0], 64), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{cannot constant evaluate whether run-time alignment is at least 64}}
+
+// However, if the alignment being checked is less than the minimum alignment of
+// the base object we can check the low bits of the alignment:
+static_assert(__builtin_is_aligned(&align32array[0], 4), "");
+static_assert(!__builtin_is_aligned(&align32array[1], 4), "");
+static_assert(!__builtin_is_aligned(&align32array[2], 4), "");
+static_assert(!__builtin_is_aligned(&align32array[3], 4), "");
+static_assert(__builtin_is_aligned(&align32array[4], 4), "");
+
+// TODO: this should evaluate to true even though we can't evaluate the result
+//  of __builtin_align_up() to a concrete value
+static_assert(__builtin_is_aligned(__builtin_align_up(&align32array[0], 64), 64), ""); // both-error{{not an integral constant expression}}
+// both-note@-1{{cannot constant evaluate the result of adjusting alignment to 64}}
+
+// Check different source and alignment type widths are handled correctly.
+static_assert(!__builtin_is_aligned(static_cast<signed long>(7), static_cast<signed short>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<signed short>(7), static_cast<signed long>(4)), "");
+// Also check signed -- unsigned mismatch.
+static_assert(!__builtin_is_aligned(static_cast<signed long>(7), static_cast<signed long>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<unsigned long>(7), static_cast<unsigned long>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<signed long>(7), static_cast<unsigned long>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<unsigned long>(7), static_cast<signed long>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<signed long>(7), static_cast<unsigned short>(4)), "");
+static_assert(!__builtin_is_aligned(static_cast<unsigned short>(7), static_cast<signed long>(4)), "");
+
+// Check the diagnostic message
+_Alignas(void) char align_void_array[1]; // both-error {{invalid application of '_Alignas' to an incomplete type 'void'}}

From 91dd844aa499d69c7ff75bf3156e2e3593a88057 Mon Sep 17 00:00:00 2001
From: Brandon Wu <brandon.wu@sifive.com>
Date: Tue, 16 Apr 2024 19:59:36 +0800
Subject: [PATCH 575/886] Recommit [RISCV] RISCV vector calling convention
 (2/2) (#79096) (#87736)

Bug fix: Handle RVV return type in calling convention correctly.
Return values are handled in a same way as function arguments.
One thing to mention is that if a type can be broken down into
homogeneous
vector types, e.g. {<vscale x 4 x i32>, {<vscale x 4 x i32>, <vscale x 4
x i32>}},
it is considered as a vector tuple type and need to be handled by tuple
type rule.
---
 llvm/lib/CodeGen/TargetLoweringBase.cpp       |  12 +-
 .../Target/RISCV/GISel/RISCVCallLowering.cpp  |  57 ++--
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 244 +++++++++++++++---
 llvm/lib/Target/RISCV/RISCVISelLowering.h     |  59 ++++-
 llvm/test/CodeGen/RISCV/rvv/calling-conv.ll   | 203 +++++++++++++++
 .../RISCV/rvv/vector-deinterleave-load.ll     |   6 +-
 .../CodeGen/RISCV/rvv/vector-deinterleave.ll  |  19 +-
 7 files changed, 515 insertions(+), 85 deletions(-)

diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index f64ded4f2cf96..6e7b67ded23c8 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -1809,8 +1809,16 @@ void llvm::GetReturnInfo(CallingConv::ID CC, Type *ReturnType,
     else if (attr.hasRetAttr(Attribute::ZExt))
       Flags.setZExt();
 
-    for (unsigned i = 0; i < NumParts; ++i)
-      Outs.push_back(ISD::OutputArg(Flags, PartVT, VT, /*isfixed=*/true, 0, 0));
+    for (unsigned i = 0; i < NumParts; ++i) {
+      ISD::ArgFlagsTy OutFlags = Flags;
+      if (NumParts > 1 && i == 0)
+        OutFlags.setSplit();
+      else if (i == NumParts - 1 && i != 0)
+        OutFlags.setSplitEnd();
+
+      Outs.push_back(
+          ISD::OutputArg(OutFlags, PartVT, VT, /*isfixed=*/true, 0, 0));
+    }
   }
 }
 
diff --git a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
index 45e19cdea300b..c18892ac62f24 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
@@ -34,14 +34,15 @@ struct RISCVOutgoingValueAssigner : public CallLowering::OutgoingValueAssigner {
   // Whether this is assigning args for a return.
   bool IsRet;
 
-  // true if assignArg has been called for a mask argument, false otherwise.
-  bool AssignedFirstMaskArg = false;
+  RVVArgDispatcher &RVVDispatcher;
 
 public:
   RISCVOutgoingValueAssigner(
-      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet)
+      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet,
+      RVVArgDispatcher &RVVDispatcher)
       : CallLowering::OutgoingValueAssigner(nullptr),
-        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet) {}
+        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet),
+        RVVDispatcher(RVVDispatcher) {}
 
   bool assignArg(unsigned ValNo, EVT OrigVT, MVT ValVT, MVT LocVT,
                  CCValAssign::LocInfo LocInfo,
@@ -51,16 +52,9 @@ struct RISCVOutgoingValueAssigner : public CallLowering::OutgoingValueAssigner {
     const DataLayout &DL = MF.getDataLayout();
     const RISCVSubtarget &Subtarget = MF.getSubtarget<RISCVSubtarget>();
 
-    std::optional<unsigned> FirstMaskArgument;
-    if (Subtarget.hasVInstructions() && !AssignedFirstMaskArg &&
-        ValVT.isVector() && ValVT.getVectorElementType() == MVT::i1) {
-      FirstMaskArgument = ValNo;
-      AssignedFirstMaskArg = true;
-    }
-
     if (RISCVAssignFn(DL, Subtarget.getTargetABI(), ValNo, ValVT, LocVT,
                       LocInfo, Flags, State, Info.IsFixed, IsRet, Info.Ty,
-                      *Subtarget.getTargetLowering(), FirstMaskArgument))
+                      *Subtarget.getTargetLowering(), RVVDispatcher))
       return true;
 
     StackSize = State.getStackSize();
@@ -181,14 +175,15 @@ struct RISCVIncomingValueAssigner : public CallLowering::IncomingValueAssigner {
   // Whether this is assigning args from a return.
   bool IsRet;
 
-  // true if assignArg has been called for a mask argument, false otherwise.
-  bool AssignedFirstMaskArg = false;
+  RVVArgDispatcher &RVVDispatcher;
 
 public:
   RISCVIncomingValueAssigner(
-      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet)
+      RISCVTargetLowering::RISCVCCAssignFn *RISCVAssignFn_, bool IsRet,
+      RVVArgDispatcher &RVVDispatcher)
       : CallLowering::IncomingValueAssigner(nullptr),
-        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet) {}
+        RISCVAssignFn(RISCVAssignFn_), IsRet(IsRet),
+        RVVDispatcher(RVVDispatcher) {}
 
   bool assignArg(unsigned ValNo, EVT OrigVT, MVT ValVT, MVT LocVT,
                  CCValAssign::LocInfo LocInfo,
@@ -201,16 +196,9 @@ struct RISCVIncomingValueAssigner : public CallLowering::IncomingValueAssigner {
     if (LocVT.isScalableVector())
       MF.getInfo<RISCVMachineFunctionInfo>()->setIsVectorCall();
 
-    std::optional<unsigned> FirstMaskArgument;
-    if (Subtarget.hasVInstructions() && !AssignedFirstMaskArg &&
-        ValVT.isVector() && ValVT.getVectorElementType() == MVT::i1) {
-      FirstMaskArgument = ValNo;
-      AssignedFirstMaskArg = true;
-    }
-
     if (RISCVAssignFn(DL, Subtarget.getTargetABI(), ValNo, ValVT, LocVT,
                       LocInfo, Flags, State, /*IsFixed=*/true, IsRet, Info.Ty,
-                      *Subtarget.getTargetLowering(), FirstMaskArgument))
+                      *Subtarget.getTargetLowering(), RVVDispatcher))
       return true;
 
     StackSize = State.getStackSize();
@@ -420,9 +408,11 @@ bool RISCVCallLowering::lowerReturnVal(MachineIRBuilder &MIRBuilder,
   SmallVector<ArgInfo, 4> SplitRetInfos;
   splitToValueTypes(OrigRetInfo, SplitRetInfos, DL, CC);
 
+  RVVArgDispatcher Dispatcher{&MF, getTLI<RISCVTargetLowering>(),
+                              ArrayRef(F.getReturnType())};
   RISCVOutgoingValueAssigner Assigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/true);
+      /*IsRet=*/true, Dispatcher);
   RISCVOutgoingValueHandler Handler(MIRBuilder, MF.getRegInfo(), Ret);
   return determineAndHandleAssignments(Handler, Assigner, SplitRetInfos,
                                        MIRBuilder, CC, F.isVarArg());
@@ -531,6 +521,7 @@ bool RISCVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
   CallingConv::ID CC = F.getCallingConv();
 
   SmallVector<ArgInfo, 32> SplitArgInfos;
+  SmallVector<Type *, 4> TypeList;
   unsigned Index = 0;
   for (auto &Arg : F.args()) {
     // Construct the ArgInfo object from destination register and argument type.
@@ -542,12 +533,16 @@ bool RISCVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
     // correspondingly and appended to SplitArgInfos.
     splitToValueTypes(AInfo, SplitArgInfos, DL, CC);
 
+    TypeList.push_back(Arg.getType());
+
     ++Index;
   }
 
+  RVVArgDispatcher Dispatcher{&MF, getTLI<RISCVTargetLowering>(),
+                              ArrayRef(TypeList)};
   RISCVIncomingValueAssigner Assigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/false);
+      /*IsRet=*/false, Dispatcher);
   RISCVFormalArgHandler Handler(MIRBuilder, MF.getRegInfo());
 
   SmallVector<CCValAssign, 16> ArgLocs;
@@ -585,11 +580,13 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
 
   SmallVector<ArgInfo, 32> SplitArgInfos;
   SmallVector<ISD::OutputArg, 8> Outs;
+  SmallVector<Type *, 4> TypeList;
   for (auto &AInfo : Info.OrigArgs) {
     // Handle any required unmerging of split value types from a given VReg into
     // physical registers. ArgInfo objects are constructed correspondingly and
     // appended to SplitArgInfos.
     splitToValueTypes(AInfo, SplitArgInfos, DL, CC);
+    TypeList.push_back(AInfo.Ty);
   }
 
   // TODO: Support tail calls.
@@ -607,9 +604,11 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   const TargetRegisterInfo *TRI = Subtarget.getRegisterInfo();
   Call.addRegMask(TRI->getCallPreservedMask(MF, Info.CallConv));
 
+  RVVArgDispatcher ArgDispatcher{&MF, getTLI<RISCVTargetLowering>(),
+                                 ArrayRef(TypeList)};
   RISCVOutgoingValueAssigner ArgAssigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/false);
+      /*IsRet=*/false, ArgDispatcher);
   RISCVOutgoingValueHandler ArgHandler(MIRBuilder, MF.getRegInfo(), Call);
   if (!determineAndHandleAssignments(ArgHandler, ArgAssigner, SplitArgInfos,
                                      MIRBuilder, CC, Info.IsVarArg))
@@ -637,9 +636,11 @@ bool RISCVCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   SmallVector<ArgInfo, 4> SplitRetInfos;
   splitToValueTypes(Info.OrigRet, SplitRetInfos, DL, CC);
 
+  RVVArgDispatcher RetDispatcher{&MF, getTLI<RISCVTargetLowering>(),
+                                 ArrayRef(F.getReturnType())};
   RISCVIncomingValueAssigner RetAssigner(
       CC == CallingConv::Fast ? RISCV::CC_RISCV_FastCC : RISCV::CC_RISCV,
-      /*IsRet=*/true);
+      /*IsRet=*/true, RetDispatcher);
   RISCVCallReturnHandler RetHandler(MIRBuilder, MF.getRegInfo(), Call);
   if (!determineAndHandleAssignments(RetHandler, RetAssigner, SplitRetInfos,
                                      MIRBuilder, CC, Info.IsVarArg))
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 1d1ea6bae6c10..765838aafb58d 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -22,6 +22,7 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/MemoryLocation.h"
 #include "llvm/Analysis/VectorUtils.h"
+#include "llvm/CodeGen/Analysis.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
@@ -18223,33 +18224,12 @@ static bool CC_RISCVAssign2XLen(unsigned XLen, CCState &State, CCValAssign VA1,
   return false;
 }
 
-static unsigned allocateRVVReg(MVT ValVT, unsigned ValNo,
-                               std::optional<unsigned> FirstMaskArgument,
-                               CCState &State, const RISCVTargetLowering &TLI) {
-  const TargetRegisterClass *RC = TLI.getRegClassFor(ValVT);
-  if (RC == &RISCV::VRRegClass) {
-    // Assign the first mask argument to V0.
-    // This is an interim calling convention and it may be changed in the
-    // future.
-    if (FirstMaskArgument && ValNo == *FirstMaskArgument)
-      return State.AllocateReg(RISCV::V0);
-    return State.AllocateReg(ArgVRs);
-  }
-  if (RC == &RISCV::VRM2RegClass)
-    return State.AllocateReg(ArgVRM2s);
-  if (RC == &RISCV::VRM4RegClass)
-    return State.AllocateReg(ArgVRM4s);
-  if (RC == &RISCV::VRM8RegClass)
-    return State.AllocateReg(ArgVRM8s);
-  llvm_unreachable("Unhandled register class for ValueType");
-}
-
 // Implements the RISC-V calling convention. Returns true upon failure.
 bool RISCV::CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
                      MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
                      ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
                      bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-                     std::optional<unsigned> FirstMaskArgument) {
+                     RVVArgDispatcher &RVVDispatcher) {
   unsigned XLen = DL.getLargestLegalIntTypeSizeInBits();
   assert(XLen == 32 || XLen == 64);
   MVT XLenVT = XLen == 32 ? MVT::i32 : MVT::i64;
@@ -18418,7 +18398,7 @@ bool RISCV::CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
   else if (ValVT == MVT::f64 && !UseGPRForF64)
     Reg = State.AllocateReg(ArgFPR64s);
   else if (ValVT.isVector()) {
-    Reg = allocateRVVReg(ValVT, ValNo, FirstMaskArgument, State, TLI);
+    Reg = RVVDispatcher.getNextPhysReg();
     if (!Reg) {
       // For return values, the vector must be passed fully via registers or
       // via the stack.
@@ -18504,9 +18484,15 @@ void RISCVTargetLowering::analyzeInputArgs(
   unsigned NumArgs = Ins.size();
   FunctionType *FType = MF.getFunction().getFunctionType();
 
-  std::optional<unsigned> FirstMaskArgument;
-  if (Subtarget.hasVInstructions())
-    FirstMaskArgument = preAssignMask(Ins);
+  RVVArgDispatcher Dispatcher;
+  if (IsRet) {
+    Dispatcher = RVVArgDispatcher{&MF, this, ArrayRef(Ins)};
+  } else {
+    SmallVector<Type *, 4> TypeList;
+    for (const Argument &Arg : MF.getFunction().args())
+      TypeList.push_back(Arg.getType());
+    Dispatcher = RVVArgDispatcher{&MF, this, ArrayRef(TypeList)};
+  }
 
   for (unsigned i = 0; i != NumArgs; ++i) {
     MVT ArgVT = Ins[i].VT;
@@ -18521,7 +18507,7 @@ void RISCVTargetLowering::analyzeInputArgs(
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (Fn(MF.getDataLayout(), ABI, i, ArgVT, ArgVT, CCValAssign::Full,
            ArgFlags, CCInfo, /*IsFixed=*/true, IsRet, ArgTy, *this,
-           FirstMaskArgument)) {
+           Dispatcher)) {
       LLVM_DEBUG(dbgs() << "InputArg #" << i << " has unhandled type "
                         << ArgVT << '\n');
       llvm_unreachable(nullptr);
@@ -18535,9 +18521,13 @@ void RISCVTargetLowering::analyzeOutputArgs(
     CallLoweringInfo *CLI, RISCVCCAssignFn Fn) const {
   unsigned NumArgs = Outs.size();
 
-  std::optional<unsigned> FirstMaskArgument;
-  if (Subtarget.hasVInstructions())
-    FirstMaskArgument = preAssignMask(Outs);
+  SmallVector<Type *, 4> TypeList;
+  if (IsRet)
+    TypeList.push_back(MF.getFunction().getReturnType());
+  else if (CLI)
+    for (const TargetLowering::ArgListEntry &Arg : CLI->getArgs())
+      TypeList.push_back(Arg.Ty);
+  RVVArgDispatcher Dispatcher{&MF, this, ArrayRef(TypeList)};
 
   for (unsigned i = 0; i != NumArgs; i++) {
     MVT ArgVT = Outs[i].VT;
@@ -18547,7 +18537,7 @@ void RISCVTargetLowering::analyzeOutputArgs(
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (Fn(MF.getDataLayout(), ABI, i, ArgVT, ArgVT, CCValAssign::Full,
            ArgFlags, CCInfo, Outs[i].IsFixed, IsRet, OrigTy, *this,
-           FirstMaskArgument)) {
+           Dispatcher)) {
       LLVM_DEBUG(dbgs() << "OutputArg #" << i << " has unhandled type "
                         << ArgVT << "\n");
       llvm_unreachable(nullptr);
@@ -18728,7 +18718,7 @@ bool RISCV::CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI,
                             ISD::ArgFlagsTy ArgFlags, CCState &State,
                             bool IsFixed, bool IsRet, Type *OrigTy,
                             const RISCVTargetLowering &TLI,
-                            std::optional<unsigned> FirstMaskArgument) {
+                            RVVArgDispatcher &RVVDispatcher) {
   if (LocVT == MVT::i32 || LocVT == MVT::i64) {
     if (unsigned Reg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
       State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
@@ -18806,13 +18796,14 @@ bool RISCV::CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI,
   }
 
   if (LocVT.isVector()) {
-    if (unsigned Reg =
-            allocateRVVReg(ValVT, ValNo, FirstMaskArgument, State, TLI)) {
+    MCPhysReg AllocatedVReg = RVVDispatcher.getNextPhysReg();
+    if (AllocatedVReg) {
       // Fixed-length vectors are located in the corresponding scalable-vector
       // container types.
       if (ValVT.isFixedLengthVector())
         LocVT = TLI.getContainerForFixedLengthVector(LocVT);
-      State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
+      State.addLoc(
+          CCValAssign::getReg(ValNo, ValVT, AllocatedVReg, LocVT, LocInfo));
     } else {
       // Try and pass the address via a "fast" GPR.
       if (unsigned GPRReg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
@@ -19440,17 +19431,15 @@ bool RISCVTargetLowering::CanLowerReturn(
   SmallVector<CCValAssign, 16> RVLocs;
   CCState CCInfo(CallConv, IsVarArg, MF, RVLocs, Context);
 
-  std::optional<unsigned> FirstMaskArgument;
-  if (Subtarget.hasVInstructions())
-    FirstMaskArgument = preAssignMask(Outs);
+  RVVArgDispatcher Dispatcher{&MF, this, ArrayRef(Outs)};
 
   for (unsigned i = 0, e = Outs.size(); i != e; ++i) {
     MVT VT = Outs[i].VT;
     ISD::ArgFlagsTy ArgFlags = Outs[i].Flags;
     RISCVABI::ABI ABI = MF.getSubtarget<RISCVSubtarget>().getTargetABI();
     if (RISCV::CC_RISCV(MF.getDataLayout(), ABI, i, VT, VT, CCValAssign::Full,
-                 ArgFlags, CCInfo, /*IsFixed=*/true, /*IsRet=*/true, nullptr,
-                 *this, FirstMaskArgument))
+                        ArgFlags, CCInfo, /*IsFixed=*/true, /*IsRet=*/true,
+                        nullptr, *this, Dispatcher))
       return false;
   }
   return true;
@@ -21247,6 +21236,181 @@ unsigned RISCVTargetLowering::getMinimumJumpTableEntries() const {
   return Subtarget.getMinimumJumpTableEntries();
 }
 
+// Handle single arg such as return value.
+template <typename Arg>
+void RVVArgDispatcher::constructArgInfos(ArrayRef<Arg> ArgList) {
+  // This lambda determines whether an array of types are constructed by
+  // homogeneous vector types.
+  auto isHomogeneousScalableVectorType = [](ArrayRef<Arg> ArgList) {
+    // First, extract the first element in the argument type.
+    auto It = ArgList.begin();
+    MVT FirstArgRegType = It->VT;
+
+    // Return if there is no return or the type needs split.
+    if (It == ArgList.end() || It->Flags.isSplit())
+      return false;
+
+    ++It;
+
+    // Return if this argument type contains only 1 element, or it's not a
+    // vector type.
+    if (It == ArgList.end() || !FirstArgRegType.isScalableVector())
+      return false;
+
+    // Second, check if the following elements in this argument type are all the
+    // same.
+    for (; It != ArgList.end(); ++It)
+      if (It->Flags.isSplit() || It->VT != FirstArgRegType)
+        return false;
+
+    return true;
+  };
+
+  if (isHomogeneousScalableVectorType(ArgList)) {
+    // Handle as tuple type
+    RVVArgInfos.push_back({(unsigned)ArgList.size(), ArgList[0].VT, false});
+  } else {
+    // Handle as normal vector type
+    bool FirstVMaskAssigned = false;
+    for (const auto &OutArg : ArgList) {
+      MVT RegisterVT = OutArg.VT;
+
+      // Skip non-RVV register type
+      if (!RegisterVT.isVector())
+        continue;
+
+      if (RegisterVT.isFixedLengthVector())
+        RegisterVT = TLI->getContainerForFixedLengthVector(RegisterVT);
+
+      if (!FirstVMaskAssigned && RegisterVT.getVectorElementType() == MVT::i1) {
+        RVVArgInfos.push_back({1, RegisterVT, true});
+        FirstVMaskAssigned = true;
+        continue;
+      }
+
+      RVVArgInfos.push_back({1, RegisterVT, false});
+    }
+  }
+}
+
+// Handle multiple args.
+template <>
+void RVVArgDispatcher::constructArgInfos<Type *>(ArrayRef<Type *> TypeList) {
+  const DataLayout &DL = MF->getDataLayout();
+  const Function &F = MF->getFunction();
+  LLVMContext &Context = F.getContext();
+
+  bool FirstVMaskAssigned = false;
+  for (Type *Ty : TypeList) {
+    StructType *STy = dyn_cast<StructType>(Ty);
+    if (STy && STy->containsHomogeneousScalableVectorTypes()) {
+      Type *ElemTy = STy->getTypeAtIndex(0U);
+      EVT VT = TLI->getValueType(DL, ElemTy);
+      MVT RegisterVT =
+          TLI->getRegisterTypeForCallingConv(Context, F.getCallingConv(), VT);
+      unsigned NumRegs =
+          TLI->getNumRegistersForCallingConv(Context, F.getCallingConv(), VT);
+
+      RVVArgInfos.push_back(
+          {NumRegs * STy->getNumElements(), RegisterVT, false});
+    } else {
+      SmallVector<EVT, 4> ValueVTs;
+      ComputeValueVTs(*TLI, DL, Ty, ValueVTs);
+
+      for (unsigned Value = 0, NumValues = ValueVTs.size(); Value != NumValues;
+           ++Value) {
+        EVT VT = ValueVTs[Value];
+        MVT RegisterVT =
+            TLI->getRegisterTypeForCallingConv(Context, F.getCallingConv(), VT);
+        unsigned NumRegs =
+            TLI->getNumRegistersForCallingConv(Context, F.getCallingConv(), VT);
+
+        // Skip non-RVV register type
+        if (!RegisterVT.isVector())
+          continue;
+
+        if (RegisterVT.isFixedLengthVector())
+          RegisterVT = TLI->getContainerForFixedLengthVector(RegisterVT);
+
+        if (!FirstVMaskAssigned &&
+            RegisterVT.getVectorElementType() == MVT::i1) {
+          RVVArgInfos.push_back({1, RegisterVT, true});
+          FirstVMaskAssigned = true;
+          --NumRegs;
+        }
+
+        RVVArgInfos.insert(RVVArgInfos.end(), NumRegs, {1, RegisterVT, false});
+      }
+    }
+  }
+}
+
+void RVVArgDispatcher::allocatePhysReg(unsigned NF, unsigned LMul,
+                                       unsigned StartReg) {
+  assert((StartReg % LMul) == 0 &&
+         "Start register number should be multiple of lmul");
+  const MCPhysReg *VRArrays;
+  switch (LMul) {
+  default:
+    report_fatal_error("Invalid lmul");
+  case 1:
+    VRArrays = ArgVRs;
+    break;
+  case 2:
+    VRArrays = ArgVRM2s;
+    break;
+  case 4:
+    VRArrays = ArgVRM4s;
+    break;
+  case 8:
+    VRArrays = ArgVRM8s;
+    break;
+  }
+
+  for (unsigned i = 0; i < NF; ++i)
+    if (StartReg)
+      AllocatedPhysRegs.push_back(VRArrays[(StartReg - 8) / LMul + i]);
+    else
+      AllocatedPhysRegs.push_back(MCPhysReg());
+}
+
+/// This function determines if each RVV argument is passed by register, if the
+/// argument can be assigned to a VR, then give it a specific register.
+/// Otherwise, assign the argument to 0 which is a invalid MCPhysReg.
+void RVVArgDispatcher::compute() {
+  uint32_t AssignedMap = 0;
+  auto allocate = [&](const RVVArgInfo &ArgInfo) {
+    // Allocate first vector mask argument to V0.
+    if (ArgInfo.FirstVMask) {
+      AllocatedPhysRegs.push_back(RISCV::V0);
+      return;
+    }
+
+    unsigned RegsNeeded = divideCeil(
+        ArgInfo.VT.getSizeInBits().getKnownMinValue(), RISCV::RVVBitsPerBlock);
+    unsigned TotalRegsNeeded = ArgInfo.NF * RegsNeeded;
+    for (unsigned StartReg = 0; StartReg + TotalRegsNeeded <= NumArgVRs;
+         StartReg += RegsNeeded) {
+      uint32_t Map = ((1 << TotalRegsNeeded) - 1) << StartReg;
+      if ((AssignedMap & Map) == 0) {
+        allocatePhysReg(ArgInfo.NF, RegsNeeded, StartReg + 8);
+        AssignedMap |= Map;
+        return;
+      }
+    }
+
+    allocatePhysReg(ArgInfo.NF, RegsNeeded, 0);
+  };
+
+  for (unsigned i = 0; i < RVVArgInfos.size(); ++i)
+    allocate(RVVArgInfos[i]);
+}
+
+MCPhysReg RVVArgDispatcher::getNextPhysReg() {
+  assert(CurIdx < AllocatedPhysRegs.size() && "Index out of range");
+  return AllocatedPhysRegs[CurIdx++];
+}
+
 namespace llvm::RISCVVIntrinsicsTable {
 
 #define GET_RISCVVIntrinsicsTable_IMPL
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.h b/llvm/lib/Target/RISCV/RISCVISelLowering.h
index e2633733c31b1..b10da3d40befb 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.h
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.h
@@ -24,6 +24,7 @@ namespace llvm {
 class InstructionCost;
 class RISCVSubtarget;
 struct RISCVRegisterInfo;
+class RVVArgDispatcher;
 
 namespace RISCVISD {
 // clang-format off
@@ -875,7 +876,7 @@ class RISCVTargetLowering : public TargetLowering {
                                ISD::ArgFlagsTy ArgFlags, CCState &State,
                                bool IsFixed, bool IsRet, Type *OrigTy,
                                const RISCVTargetLowering &TLI,
-                               std::optional<unsigned> FirstMaskArgument);
+                               RVVArgDispatcher &RVVDispatcher);
 
 private:
   void analyzeInputArgs(MachineFunction &MF, CCState &CCInfo,
@@ -1017,19 +1018,71 @@ class RISCVTargetLowering : public TargetLowering {
   unsigned getMinimumJumpTableEntries() const override;
 };
 
+/// As per the spec, the rules for passing vector arguments are as follows:
+///
+/// 1. For the first vector mask argument, use v0 to pass it.
+/// 2. For vector data arguments or rest vector mask arguments, starting from
+/// the v8 register, if a vector register group between v8-v23 that has not been
+/// allocated can be found and the first register number is a multiple of LMUL,
+/// then allocate this vector register group to the argument and mark these
+/// registers as allocated. Otherwise, pass it by reference and are replaced in
+/// the argument list with the address.
+/// 3. For tuple vector data arguments, starting from the v8 register, if
+/// NFIELDS consecutive vector register groups between v8-v23 that have not been
+/// allocated can be found and the first register number is a multiple of LMUL,
+/// then allocate these vector register groups to the argument and mark these
+/// registers as allocated. Otherwise, pass it by reference and are replaced in
+/// the argument list with the address.
+class RVVArgDispatcher {
+public:
+  static constexpr unsigned NumArgVRs = 16;
+
+  struct RVVArgInfo {
+    unsigned NF;
+    MVT VT;
+    bool FirstVMask = false;
+  };
+
+  template <typename Arg>
+  RVVArgDispatcher(const MachineFunction *MF, const RISCVTargetLowering *TLI,
+                   ArrayRef<Arg> ArgList)
+      : MF(MF), TLI(TLI) {
+    constructArgInfos(ArgList);
+    compute();
+  }
+
+  RVVArgDispatcher() = default;
+
+  MCPhysReg getNextPhysReg();
+
+private:
+  SmallVector<RVVArgInfo, 4> RVVArgInfos;
+  SmallVector<MCPhysReg, 4> AllocatedPhysRegs;
+
+  const MachineFunction *MF = nullptr;
+  const RISCVTargetLowering *TLI = nullptr;
+
+  unsigned CurIdx = 0;
+
+  template <typename Arg> void constructArgInfos(ArrayRef<Arg> Ret);
+  void compute();
+  void allocatePhysReg(unsigned NF = 1, unsigned LMul = 1,
+                       unsigned StartReg = 0);
+};
+
 namespace RISCV {
 
 bool CC_RISCV(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
               MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
               ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
               bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-              std::optional<unsigned> FirstMaskArgument);
+              RVVArgDispatcher &RVVDispatcher);
 
 bool CC_RISCV_FastCC(const DataLayout &DL, RISCVABI::ABI ABI, unsigned ValNo,
                      MVT ValVT, MVT LocVT, CCValAssign::LocInfo LocInfo,
                      ISD::ArgFlagsTy ArgFlags, CCState &State, bool IsFixed,
                      bool IsRet, Type *OrigTy, const RISCVTargetLowering &TLI,
-                     std::optional<unsigned> FirstMaskArgument);
+                     RVVArgDispatcher &RVVDispatcher);
 
 bool CC_RISCV_GHC(unsigned ValNo, MVT ValVT, MVT LocVT,
                   CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags,
diff --git a/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll b/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
index 78e8700a9feff..647d3158b6167 100644
--- a/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/calling-conv.ll
@@ -162,3 +162,206 @@ define void @caller_tuple_argument({<vscale x 4 x i32>, <vscale x 4 x i32>} %x)
 }
 
 declare void @callee_tuple_argument({<vscale x 4 x i32>, <vscale x 4 x i32>})
+
+; %0 -> v8
+; %1 -> v9
+define <vscale x 1 x i64> @case1(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1) {
+; CHECK-LABEL: case1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v8, v9
+; CHECK-NEXT:    ret
+  %a = add <vscale x 1 x i64> %0, %1
+  ret <vscale x 1 x i64> %a
+}
+
+; %0 -> v8
+; %1 -> v10-v11
+; %2 -> v9
+define <vscale x 1 x i64> @case2_1(<vscale x 1 x i64> %0, <vscale x 2 x i64> %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case2_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v8, v9
+; CHECK-NEXT:    ret
+  %a = add <vscale x 1 x i64> %0, %2
+  ret <vscale x 1 x i64> %a
+}
+define <vscale x 2 x i64> @case2_2(<vscale x 1 x i64> %0, <vscale x 2 x i64> %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case2_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v10, v10
+; CHECK-NEXT:    ret
+  %a = add <vscale x 2 x i64> %1, %1
+  ret <vscale x 2 x i64> %a
+}
+
+; %0 -> v8
+; %1 -> {v10-v11, v12-v13}
+; %2 -> v9
+define <vscale x 1 x i64> @case3_1(<vscale x 1 x i64> %0, {<vscale x 2 x i64>, <vscale x 2 x i64>} %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case3_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v8, v9
+; CHECK-NEXT:    ret
+  %add = add <vscale x 1 x i64> %0, %2
+  ret <vscale x 1 x i64> %add
+}
+define <vscale x 2 x i64> @case3_2(<vscale x 1 x i64> %0, {<vscale x 2 x i64>, <vscale x 2 x i64>} %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case3_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v10, v12
+; CHECK-NEXT:    ret
+  %a = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } %1, 0
+  %b = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } %1, 1
+  %add = add <vscale x 2 x i64> %a, %b
+  ret <vscale x 2 x i64> %add
+}
+
+; %0 -> v8
+; %1 -> {by-ref, by-ref}
+; %2 -> v9
+define <vscale x 8 x i64> @case4_1(<vscale x 1 x i64> %0, {<vscale x 8 x i64>, <vscale x 8 x i64>} %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case4_1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a1, vlenb
+; CHECK-NEXT:    slli a1, a1, 3
+; CHECK-NEXT:    add a1, a0, a1
+; CHECK-NEXT:    vl8re64.v v8, (a1)
+; CHECK-NEXT:    vl8re64.v v16, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v16, v8
+; CHECK-NEXT:    ret
+  %a = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i64> } %1, 0
+  %b = extractvalue { <vscale x 8 x i64>, <vscale x 8 x i64> } %1, 1
+  %add = add <vscale x 8 x i64> %a, %b
+  ret <vscale x 8 x i64> %add
+}
+define <vscale x 1 x i64> @case4_2(<vscale x 1 x i64> %0, {<vscale x 8 x i64>, <vscale x 8 x i64>} %1, <vscale x 1 x i64> %2) {
+; CHECK-LABEL: case4_2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vadd.vv v8, v8, v9
+; CHECK-NEXT:    ret
+  %add = add <vscale x 1 x i64> %0, %2
+  ret <vscale x 1 x i64> %add
+}
+
+declare <vscale x 1 x i64> @callee1()
+declare void @callee2(<vscale x 1 x i64>)
+declare void @callee3(<vscale x 4 x i32>)
+define void @caller() {
+; RV32-LABEL: caller:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    .cfi_offset ra, -4
+; RV32-NEXT:    call callee1
+; RV32-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; RV32-NEXT:    vadd.vv v8, v8, v8
+; RV32-NEXT:    call callee2
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: caller:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    .cfi_def_cfa_offset 16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    .cfi_offset ra, -8
+; RV64-NEXT:    call callee1
+; RV64-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; RV64-NEXT:    vadd.vv v8, v8, v8
+; RV64-NEXT:    call callee2
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+  %a = call <vscale x 1 x i64> @callee1()
+  %add = add <vscale x 1 x i64> %a, %a
+  call void @callee2(<vscale x 1 x i64> %add)
+  ret void
+}
+
+declare {<vscale x 4 x i32>, <vscale x 4 x i32>} @callee_tuple()
+define void @caller_tuple() {
+; RV32-LABEL: caller_tuple:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    .cfi_offset ra, -4
+; RV32-NEXT:    call callee_tuple
+; RV32-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; RV32-NEXT:    vadd.vv v8, v8, v10
+; RV32-NEXT:    call callee3
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: caller_tuple:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    .cfi_def_cfa_offset 16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    .cfi_offset ra, -8
+; RV64-NEXT:    call callee_tuple
+; RV64-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; RV64-NEXT:    vadd.vv v8, v8, v10
+; RV64-NEXT:    call callee3
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+  %a = call {<vscale x 4 x i32>, <vscale x 4 x i32>} @callee_tuple()
+  %b = extractvalue {<vscale x 4 x i32>, <vscale x 4 x i32>} %a, 0
+  %c = extractvalue {<vscale x 4 x i32>, <vscale x 4 x i32>} %a, 1
+  %add = add <vscale x 4 x i32> %b, %c
+  call void @callee3(<vscale x 4 x i32> %add)
+  ret void
+}
+
+declare {<vscale x 4 x i32>, {<vscale x 4 x i32>, <vscale x 4 x i32>}} @callee_nested()
+define void @caller_nested() {
+; RV32-LABEL: caller_nested:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw ra, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT:    .cfi_offset ra, -4
+; RV32-NEXT:    call callee_nested
+; RV32-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; RV32-NEXT:    vadd.vv v8, v8, v10
+; RV32-NEXT:    vadd.vv v8, v8, v12
+; RV32-NEXT:    call callee3
+; RV32-NEXT:    lw ra, 12(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: caller_nested:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    .cfi_def_cfa_offset 16
+; RV64-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64-NEXT:    .cfi_offset ra, -8
+; RV64-NEXT:    call callee_nested
+; RV64-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; RV64-NEXT:    vadd.vv v8, v8, v10
+; RV64-NEXT:    vadd.vv v8, v8, v12
+; RV64-NEXT:    call callee3
+; RV64-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    ret
+  %a = call {<vscale x 4 x i32>, {<vscale x 4 x i32>, <vscale x 4 x i32>}} @callee_nested()
+  %b = extractvalue {<vscale x 4 x i32>, {<vscale x 4 x i32>, <vscale x 4 x i32>}} %a, 0
+  %c = extractvalue {<vscale x 4 x i32>, {<vscale x 4 x i32>, <vscale x 4 x i32>}} %a, 1
+  %c0 = extractvalue {<vscale x 4 x i32>, <vscale x 4 x i32>} %c, 0
+  %c1 = extractvalue {<vscale x 4 x i32>, <vscale x 4 x i32>} %c, 1
+  %add0 = add <vscale x 4 x i32> %b, %c0
+  %add1 = add <vscale x 4 x i32> %add0, %c1
+  call void @callee3(<vscale x 4 x i32> %add1)
+  ret void
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
index a320aecc6fce4..6a712080fda74 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
@@ -18,10 +18,10 @@ define {<vscale x 16 x i1>, <vscale x 16 x i1>} @vector_deinterleave_load_nxv16i
 ; CHECK-NEXT:    vmerge.vim v14, v10, 1, v0
 ; CHECK-NEXT:    vmv1r.v v0, v8
 ; CHECK-NEXT:    vmerge.vim v12, v10, 1, v0
-; CHECK-NEXT:    vnsrl.wi v8, v12, 0
-; CHECK-NEXT:    vmsne.vi v0, v8, 0
-; CHECK-NEXT:    vnsrl.wi v10, v12, 8
+; CHECK-NEXT:    vnsrl.wi v10, v12, 0
 ; CHECK-NEXT:    vmsne.vi v8, v10, 0
+; CHECK-NEXT:    vnsrl.wi v10, v12, 8
+; CHECK-NEXT:    vmsne.vi v9, v10, 0
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 32 x i1>, ptr %p
   %retval = call {<vscale x 16 x i1>, <vscale x 16 x i1>} @llvm.experimental.vector.deinterleave2.nxv32i1(<vscale x 32 x i1> %vec)
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
index ef4baf34d23f0..d98597fabcd95 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
@@ -8,18 +8,18 @@ define {<vscale x 16 x i1>, <vscale x 16 x i1>} @vector_deinterleave_nxv16i1_nxv
 ; CHECK-LABEL: vector_deinterleave_nxv16i1_nxv32i1:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
-; CHECK-NEXT:    vmv.v.i v10, 0
-; CHECK-NEXT:    vmerge.vim v8, v10, 1, v0
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v12, v8, 1, v0
 ; CHECK-NEXT:    csrr a0, vlenb
 ; CHECK-NEXT:    srli a0, a0, 2
 ; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, ma
 ; CHECK-NEXT:    vslidedown.vx v0, v0, a0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, ma
-; CHECK-NEXT:    vmerge.vim v10, v10, 1, v0
-; CHECK-NEXT:    vnsrl.wi v12, v8, 0
-; CHECK-NEXT:    vmsne.vi v0, v12, 0
-; CHECK-NEXT:    vnsrl.wi v12, v8, 8
-; CHECK-NEXT:    vmsne.vi v8, v12, 0
+; CHECK-NEXT:    vmerge.vim v14, v8, 1, v0
+; CHECK-NEXT:    vnsrl.wi v10, v12, 0
+; CHECK-NEXT:    vmsne.vi v8, v10, 0
+; CHECK-NEXT:    vnsrl.wi v10, v12, 8
+; CHECK-NEXT:    vmsne.vi v9, v10, 0
 ; CHECK-NEXT:    ret
 %retval = call {<vscale x 16 x i1>, <vscale x 16 x i1>} @llvm.experimental.vector.deinterleave2.nxv32i1(<vscale x 32 x i1> %vec)
 ret {<vscale x 16 x i1>, <vscale x 16 x i1>} %retval
@@ -102,12 +102,13 @@ define {<vscale x 64 x i1>, <vscale x 64 x i1>} @vector_deinterleave_nxv64i1_nxv
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, ma
 ; CHECK-NEXT:    vnsrl.wi v28, v8, 0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, ma
-; CHECK-NEXT:    vmsne.vi v0, v24, 0
+; CHECK-NEXT:    vmsne.vi v7, v24, 0
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, ma
 ; CHECK-NEXT:    vnsrl.wi v24, v16, 8
 ; CHECK-NEXT:    vnsrl.wi v28, v8, 8
 ; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, ma
-; CHECK-NEXT:    vmsne.vi v8, v24, 0
+; CHECK-NEXT:    vmsne.vi v9, v24, 0
+; CHECK-NEXT:    vmv1r.v v8, v7
 ; CHECK-NEXT:    ret
 %retval = call {<vscale x 64 x i1>, <vscale x 64 x i1>} @llvm.experimental.vector.deinterleave2.nxv128i1(<vscale x 128 x i1> %vec)
 ret {<vscale x 64 x i1>, <vscale x 64 x i1>} %retval

From ac1f2de7b581c26a768c4d2a2aad36505cc63c31 Mon Sep 17 00:00:00 2001
From: Frederik Harwath <frederik.harwath@amd.com>
Date: Tue, 16 Apr 2024 14:09:25 +0200
Subject: [PATCH 576/886] [MLIR][docs] Mention declarePromisedInterface in
 Interfaces doc (#88689)

Co-authored-by: Frederik Harwath <fharwath@amd.com>
Co-authored-by: Mehdi Amini <joker.eph@gmail.com>
---
 mlir/docs/Interfaces.md | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/mlir/docs/Interfaces.md b/mlir/docs/Interfaces.md
index 536e7613e5093..51747db546bb7 100644
--- a/mlir/docs/Interfaces.md
+++ b/mlir/docs/Interfaces.md
@@ -299,6 +299,30 @@ owner of the dialect containing the object nor the owner of the interface are
 aware of an interface implementation, which can lead to duplicate or
 diverging implementations.
 
+Forgetting to register an external model can lead to bugs which are hard to
+track down. The `declarePromisedInterface` function can be used to declare that
+an external model implementation for an operation must eventually be provided.
+
+```
+  void MyDialect::initialize() {
+    declarePromisedInterface<SomeInterface, SomeOp>();
+     ...
+  }
+```
+
+Now attempting to use the interface, e.g in a cast, without a prior registration
+of the external model will lead to a runtime error that will look similar to
+this:
+
+```
+LLVM ERROR: checking for an interface (`SomeInterface`) that was promised by dialect 'mydialect' but never implemented. This is generally an indication that the dialect extension implementing the interface was never registered.
+```
+
+If you encounter this error for a dialect and an interface provided by MLIR, you
+may look for a method that will be named like
+`register<Dialect><Interface>ExternalModels(DialectRegistry &registry);` ; try
+to find it with `git grep 'register.*SomeInterface.*Model' mlir`.
+
 #### Dialect Fallback for OpInterface
 
 Some dialects have an open ecosystem and don't register all of the possible

From 1c076b43c294c7d29d99dd50f6853b33a5b99789 Mon Sep 17 00:00:00 2001
From: Spenser Bauman <sbauman@mathworks.com>
Date: Tue, 16 Apr 2024 08:14:22 -0400
Subject: [PATCH 577/886] [mlir][tosa] Implement dynamic shape support for
 tosa.max_pool2d lowering (#87538)

The existing lowering for tosa.max_pool2d only supports dynamic
dimensions when the dynamic dimension is the batch dimension. This
change updates the lowering to support arbitrary dynamic dimensions on
the inputs and outputs of the tosa.max_pool2d operation.

This change also fixes a bug in the implementation of implicit
broadcasting in the tosa-to-linalg pass, which was introducing uses of
constant ops that violated dominance requirements.
---
 .../mlir/Dialect/Tosa/IR/TosaTypesBase.td     |  10 +-
 .../Conversion/TosaToLinalg/TosaToLinalg.cpp  |  12 +-
 .../TosaToLinalg/TosaToLinalgNamed.cpp        |  88 ++++++++++----
 .../TosaToLinalg/tosa-to-linalg-named.mlir    |  54 +++++++++
 .../TosaToLinalg/tosa-to-linalg.mlir          |  12 +-
 .../Tosa/CPU/test-maxpool-dynamic.mlir        | 112 ++++++++++++++++++
 6 files changed, 251 insertions(+), 37 deletions(-)
 create mode 100644 mlir/test/Integration/Dialect/Tosa/CPU/test-maxpool-dynamic.mlir

diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td b/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
index cff3de0a69af9..3687891fe4b7c 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
@@ -130,11 +130,11 @@ def Tosa_ScalarTensor : TensorRankOf<[Tosa_AnyNumber], [0]>;
 // to not include any remaining unranked tensors.
 def Tosa_UnrankedTensor : UnrankedTensorOf<[Tosa_AnyNumber]>;
 
-def Tosa_Tensor1D : AnyTypeOf<[Tosa_UnrankedTensor, 1DTensorOf<[Tosa_AnyNumber]>]>;
-def Tosa_Tensor2D : AnyTypeOf<[Tosa_UnrankedTensor, 2DTensorOf<[Tosa_AnyNumber]>]>;
-def Tosa_Tensor3D : AnyTypeOf<[Tosa_UnrankedTensor, 3DTensorOf<[Tosa_AnyNumber]>]>;
-def Tosa_Tensor4D : AnyTypeOf<[Tosa_UnrankedTensor, 4DTensorOf<[Tosa_AnyNumber]>]>;
-def Tosa_Tensor5D : AnyTypeOf<[Tosa_UnrankedTensor, TensorRankOf<[Tosa_AnyNumber], [5]>]>;
+def Tosa_Tensor1D : AnyTypeOf<[Tosa_UnrankedTensor, 1DTensorOf<[Tosa_AnyNumber]>], "1-d tensor", "::mlir::TensorType">;
+def Tosa_Tensor2D : AnyTypeOf<[Tosa_UnrankedTensor, 2DTensorOf<[Tosa_AnyNumber]>], "2-d tensor", "::mlir::TensorType">;
+def Tosa_Tensor3D : AnyTypeOf<[Tosa_UnrankedTensor, 3DTensorOf<[Tosa_AnyNumber]>], "3-d tensor", "::mlir::TensorType">;
+def Tosa_Tensor4D : AnyTypeOf<[Tosa_UnrankedTensor, 4DTensorOf<[Tosa_AnyNumber]>], "4-d tensor", "::mlir::TensorType">;
+def Tosa_Tensor5D : AnyTypeOf<[Tosa_UnrankedTensor, TensorRankOf<[Tosa_AnyNumber], [5]>], "5-d tensor", "::mlir::TensorType">;
 
 // Ranked tensors up to given rank.
 def Tosa_Tensor1Dto4D : AnyTypeOf<[
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
index 7c477f2e1412b..d8dd1c93722b0 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -766,11 +766,15 @@ static Value broadcastDynamicDimension(PatternRewriter &rewriter, Location loc,
 
   // Emit 'then' region of 'scf.if'
   auto emitThenRegion = [&](OpBuilder &opBuilder, Location loc) {
+    // It is not safe to cache constants across regions.
+    // New constants could potentially violate dominance requirements.
+    IndexPool localPool;
+
     // Emit 'tensor.empty' op
     SmallVector<OpFoldResult> outputTensorShape;
     for (auto index : llvm::seq<int64_t>(0, rank)) {
       auto size = index == dim ? targetSize
-                               : getOrFoldTensorDim(rewriter, loc, indexPool,
+                               : getOrFoldTensorDim(rewriter, loc, localPool,
                                                     operand, index);
       outputTensorShape.push_back(size);
     }
@@ -812,9 +816,9 @@ static Value broadcastDynamicDimensions(PatternRewriter &rewriter, Location loc,
                                         IndexPool &indexPool, Value operand,
                                         ArrayRef<OpFoldResult> targetShape,
                                         ArrayRef<Value> masterOperands) {
-  size_t rank = operand.getType().cast<RankedTensorType>().getRank();
-  assert(targetShape.size() == rank);
-  assert(masterOperands.size() == rank);
+  int64_t rank = operand.getType().cast<RankedTensorType>().getRank();
+  assert((int64_t)targetShape.size() == rank);
+  assert((int64_t)masterOperands.size() == rank);
   for (auto index : llvm::seq<int64_t>(0, rank))
     operand =
         broadcastDynamicDimension(rewriter, loc, indexPool, operand, index,
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
index 3f39cbf03a9a8..8fb8d16486560 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
@@ -26,6 +26,8 @@
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
+#include "mlir/Interfaces/InferTypeOpInterface.h"
+
 #include <numeric>
 #include <type_traits>
 
@@ -34,7 +36,7 @@ using namespace mlir::tosa;
 
 static mlir::Value applyPad(Location loc, Value input, ArrayRef<int64_t> pad,
                             TypedAttr padAttr, OpBuilder &rewriter) {
-  // Input should be padded if necessary.
+  // Input should be padded only if necessary.
   if (llvm::all_of(pad, [](int64_t p) { return p == 0; }))
     return input;
 
@@ -47,7 +49,7 @@ static mlir::Value applyPad(Location loc, Value input, ArrayRef<int64_t> pad,
   SmallVector<int64_t, 4> paddedShape;
   SmallVector<OpFoldResult, 8> lowIndices;
   SmallVector<OpFoldResult, 8> highIndices;
-  for (int i = 0, s = inputShape.size(); i < s; i++) {
+  for (size_t i : llvm::seq(inputShape.size())) {
     auto lowPad = pad[i * 2];
     auto highPad = pad[i * 2 + 1];
     if (ShapedType::isDynamic(inputShape[i]))
@@ -131,20 +133,19 @@ static mlir::Value linalgBroadcastAndMaybeExtSI(PatternRewriter &rewriter,
 
 static mlir::Value reifyConstantDim(int64_t attr,
                                     ImplicitLocOpBuilder &builder) {
-  return builder.createOrFold<arith::IndexCastOp>(
-      builder.getIndexType(),
-      builder.create<arith::ConstantOp>(builder.getI64IntegerAttr(attr)));
+  return builder.create<arith::ConstantIndexOp>(attr);
 }
 
 // Calculating the output width/height using the formula:
 // H = ((IH+pad_top+pad_bottom-(dilation_y*(KH-1)+1))/stride_y)+1
 // W = ((IW+pad_left+pad_right-(dilation_x*(KW-1)+1))/stride_x)+1
 
-static mlir::Value getConvOutputDim(Location loc, Value inputDim,
-                                    int64_t padBeforeAttr, int64_t padAfterAttr,
-                                    Value kernelDim, int64_t strideAttr,
-                                    int64_t dilationAttr, Type inputETy,
-                                    OpBuilder &rewriter) {
+static mlir::Value getConvOrPoolOutputDim(Location loc, Value inputDim,
+                                          int64_t padBeforeAttr,
+                                          int64_t padAfterAttr, Value kernelDim,
+                                          int64_t strideAttr,
+                                          int64_t dilationAttr,
+                                          OpBuilder &rewriter) {
   ImplicitLocOpBuilder builder(loc, rewriter);
   auto one = rewriter.create<arith::ConstantOp>(
       loc, IntegerAttr::get(inputDim.getType(), 1));
@@ -171,7 +172,6 @@ static SmallVector<Value> inferDynamicDimsForConv(
     ArrayRef<int64_t> dilationAttr, ArrayRef<int64_t> inputSizeDims,
     ArrayRef<int64_t> kernelSizeDims, OpBuilder &rewriter) {
   ShapedType inputTy = cast<ShapedType>(input.getType());
-  Type inputETy = inputTy.getElementType();
   int64_t inputRank = inputTy.getRank();
 
   SmallVector<Value> dynDims;
@@ -190,8 +190,8 @@ static SmallVector<Value> inferDynamicDimsForConv(
           rewriter.create<tensor::DimOp>(loc, weight, kernelDim);
       // H = F(IH, pad_top, pad_bottom, dilation_y, KH, stride_y)
       dynDims[inputDim] =
-          getConvOutputDim(loc, initDynDim, padTop, padBottom, kernelDynDim,
-                           stride, dilation, inputETy, rewriter);
+          getConvOrPoolOutputDim(loc, initDynDim, padTop, padBottom,
+                                 kernelDynDim, stride, dilation, rewriter);
     }
   }
 
@@ -685,20 +685,61 @@ class MaxPool2dConverter : public OpRewritePattern<tosa::MaxPool2dOp> {
 public:
   using OpRewritePattern<tosa::MaxPool2dOp>::OpRewritePattern;
 
+  // Compute the dynamic output sizes of the maxpool operation.
+  static SmallVector<Value>
+  computeDynamicOutputSizes(tosa::MaxPool2dOp op, PatternRewriter &rewriter) {
+    TensorType resultTy = op.getType();
+    Location loc = op.getLoc();
+
+    TypedValue<TensorType> input = op.getInput();
+    ArrayRef<int64_t> kernel = op.getKernel();
+    ArrayRef<int64_t> pad = op.getPad();
+    ArrayRef<int64_t> stride = op.getStride();
+
+    SmallVector<Value> dynamicDims;
+
+    // Batch dimension
+    if (resultTy.isDynamicDim(0))
+      dynamicDims.push_back(rewriter.create<tensor::DimOp>(loc, input, 0));
+
+    // Height/width dimensions
+    for (int64_t dim : {1, 2}) {
+      if (!resultTy.isDynamicDim(dim))
+        continue;
+
+      // Index into the attribute arrays
+      int64_t index = dim - 1;
+
+      // Input height/width
+      Value ihw = rewriter.create<tensor::DimOp>(loc, input, dim);
+
+      // Kernel height/width
+      Value khw = rewriter.create<arith::ConstantIndexOp>(loc, kernel[index]);
+
+      // Output height/width
+      Value ohw = getConvOrPoolOutputDim(loc, ihw, pad[index * 2],
+                                         pad[index * 2 + 1], khw, stride[index],
+                                         /*dilationAttr=*/1, rewriter);
+      dynamicDims.push_back(ohw);
+    }
+
+    // Channel dimension
+    if (resultTy.isDynamicDim(3))
+      dynamicDims.push_back(rewriter.create<tensor::DimOp>(loc, input, 3));
+
+    return dynamicDims;
+  }
+
   LogicalResult matchAndRewrite(tosa::MaxPool2dOp op,
                                 PatternRewriter &rewriter) const final {
     Location loc = op.getLoc();
-    Value input = op.getInput();
-    ShapedType inputTy = cast<ShapedType>(input.getType());
+    TypedValue<TensorType> input = op.getInput();
+    ShapedType inputTy = input.getType();
 
-    ShapedType resultTy = cast<ShapedType>(op.getType());
+    ShapedType resultTy = op.getType();
     Type resultETy = inputTy.getElementType();
 
-    auto dynamicDimsOr =
-        checkHasDynamicBatchDims(rewriter, op, {input, op.getOutput()});
-    if (!dynamicDimsOr.has_value())
-      return failure();
-    SmallVector<Value> dynamicDims = *dynamicDimsOr;
+    SmallVector<Value> dynamicDims = computeDynamicOutputSizes(op, rewriter);
 
     // Determine what the initial value needs to be for the max pool op.
     TypedAttr initialAttr;
@@ -721,6 +762,7 @@ class MaxPool2dConverter : public OpRewritePattern<tosa::MaxPool2dOp> {
     pad.resize(2, 0);
     llvm::append_range(pad, op.getPad());
     pad.resize(pad.size() + 2, 0);
+
     Value paddedInput = applyPad(loc, input, pad, initialAttr, rewriter);
 
     Value initialValue = rewriter.create<arith::ConstantOp>(loc, initialAttr);
@@ -736,9 +778,7 @@ class MaxPool2dConverter : public OpRewritePattern<tosa::MaxPool2dOp> {
         loc, resultTy.getShape(), resultTy.getElementType(), dynamicDims);
 
     Value filledEmptyTensor =
-        rewriter
-            .create<linalg::FillOp>(loc, ValueRange{initialValue},
-                                    ValueRange{emptyTensor})
+        rewriter.create<linalg::FillOp>(loc, initialValue, emptyTensor)
             .result();
 
     Value fakeWindowDims =
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-named.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-named.mlir
index e64903671e599..b4049000c50dc 100644
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-named.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-named.mlir
@@ -1,5 +1,6 @@
 // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(func.func(tosa-to-linalg-named))" %s -verify-diagnostics -o -| FileCheck %s
 // RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(func.func(tosa-to-linalg-named{prefer-conv2d-kernel-layout-hwcf=true}))" %s -verify-diagnostics -o -| FileCheck --check-prefix="HWCF" %s
+// RUN: mlir-opt --split-input-file -pass-pipeline="builtin.module(func.func(tosa-to-linalg-named,cse))" %s -verify-diagnostics -o -| FileCheck --check-prefix="CHECK-CSE" %s
 
 // CHECK-LABEL: @matmul
 func.func @matmul(%arg0: tensor<1x5x3xf32>, %arg1: tensor<1x3x6xf32>) -> (tensor<1x5x6xf32>) {
@@ -215,6 +216,59 @@ func.func @max_pool_i32(%arg0: tensor<1x6x34x62xi32>) -> () {
   return
 }
 
+// CHECK-CSE-LABEL: @max_pool_all_dynamic
+func.func @max_pool_all_dynamic(%arg0: tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32> {
+  // Batch size
+  // CHECK-CSE: %[[C0:.+]] = arith.constant 0 : index
+  // CHECK-CSE: %[[BATCH:.+]] = tensor.dim %arg0, %[[C0]] : tensor<?x?x?x?xf32>
+
+  // Compute output height
+  // CHECK-CSE: %[[C1:.+]] = arith.constant 1 : index
+  // CHECK-CSE: %[[IH:.+]] = tensor.dim %arg0, %[[C1]] : tensor<?x?x?x?xf32>
+  // CHECK-CSE: %[[C2:.+]] = arith.constant 2 : index
+  // CHECK-CSE: %[[PADDED_BEFORE:.+]] = arith.addi %[[IH]], %[[C0]] : index
+  // CHECK-CSE: %[[PADDED_AFTER:.+]] = arith.addi %[[PADDED_BEFORE]], %[[C0]] : index
+  // CHECK-CSE: %[[SUB_ONE:.+]] = arith.subi %[[C2]], %[[C1]] : index
+  // CHECK-CSE: %[[DILATED:.+]] = arith.muli %[[C1]], %[[SUB_ONE]] : index
+  // CHECK-CSE: %[[ADD_ONE:.+]] = arith.addi %[[DILATED]], %[[C1]] : index
+  // CHECK-CSE: %[[SUBTRACT:.+]] = arith.subi %[[PADDED_AFTER]], %[[ADD_ONE]] : index
+  // CHECK-CSE: %[[DIVIDE:.+]] = arith.divui %[[SUBTRACT]], %[[C1]] : index
+  // CHECK-CSE: %[[HEIGHT:.+]] = arith.addi %[[DIVIDE]], %[[C1]] : index
+
+  // Compute output width
+  // CHECK-CSE: %[[IW:.+]] = tensor.dim %arg0, %[[C2]] : tensor<?x?x?x?xf32>
+  // CHECK-CSE: %[[C5:.+]] = arith.constant 5 : index
+  // CHECK-CSE: %[[PADDED_BEFORE:.+]] = arith.addi %[[IW]], %[[C2]] : index
+  // CHECK-CSE: %[[PADDED_AFTER:.+]] = arith.addi %[[PADDED_BEFORE]], %[[C2]] : index
+  // CHECK-CSE: %[[SUB_ONE:.+]] = arith.subi %[[C5]], %[[C1]] : index
+  // CHECK-CSE: %[[DILATED:.+]] = arith.muli %[[C1]], %[[SUB_ONE]] : index
+  // CHECK-CSE: %[[ADD_ONE:.+]] = arith.addi %[[DILATED]], %[[C1]] : index
+  // CHECK-CSE: %[[SUBTRACT:.+]] = arith.subi %[[PADDED_AFTER]], %[[ADD_ONE]] : index
+  // CHECK-CSE: %[[DIVIDE:.+]] = arith.divui %[[SUBTRACT]], %[[C1]] : index
+  // CHECK-CSE: %[[WIDTH:.+]] = arith.addi %14, %[[C1]] : index
+
+  // Channel size
+  // CHECK-CSE: %[[C3:.+]] = arith.constant 3 : index
+  // CHECK-CSE: %[[CHANNEL:.+]] = tensor.dim %arg0, %[[C3]] : tensor<?x?x?x?xf32>
+
+  // Pad the input
+  // CHECK-CSE: %[[FLOAT_MIN:.+]] = arith.constant -3.40282347E+38 : f32
+  // CHECK-CSE: %[[PADDED:.+]] = tensor.pad %arg0 low[0, 0, 2, 0] high[0, 0, 2, 0] {
+  // CHECK-CSE:   tensor.yield %[[FLOAT_MIN]] : f32
+
+  // Allocate the output and fill with minimum value
+  // CHECK-CSE: %[[INIT:.+]] = tensor.empty(%[[BATCH]], %[[HEIGHT]], %[[WIDTH]], %[[CHANNEL]]) : tensor<?x?x?x?xf32>
+  // CHECK-CSE: %[[FILL:.+]] = linalg.fill ins(%[[FLOAT_MIN]] : f32) outs(%[[INIT]] : tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
+  // CHECK-CSE: %[[FAKE_WINDOW:.+]] = tensor.empty() : tensor<2x5xf32>
+
+  // Compute max pool
+  // CHECK-CSE: %[[OUT:.+]] = linalg.pooling_nhwc_max {dilations = dense<1> : vector<2xi64>, strides = dense<1> : vector<2xi64>} ins(%[[PADDED]], %[[FAKE_WINDOW]] : tensor<?x?x?x?xf32>, tensor<2x5xf32>) outs(%[[FILL]] : tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
+  // CHECK-CSE: return %[[OUT]]
+
+  %0 = tosa.max_pool2d %arg0 {kernel = array<i64: 2, 5>, pad = array<i64: 0, 0, 2, 2>, stride = array<i64: 1, 1>} : (tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
+  return %0 : tensor<?x?x?x?xf32>
+}
+
 // -----
 
 // CHECK-LABEL: @avg_pool_f32
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
index 1fa783f05f04e..445e8be47678d 100644
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
@@ -270,7 +270,8 @@ func.func @test_add_2d_all_dynamic(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32
   // CHECK: %[[VAL_0:.*]] = tensor.dim %[[ARG0]], %[[CONST0]] : tensor<?x?xf32>
   // CHECK: %[[VAL_1:.*]] = arith.cmpi eq, %[[VAL_0]], %[[CONST1]] : index
   // CHECK: %[[ARG0_DIM0_BROADCAST:.*]] = scf.if %[[VAL_1]] -> (tensor<?x?xf32>) {
-  // CHECK:   %[[VAL_2:.*]] = tensor.dim %[[ARG0]], %[[CONST1]] : tensor<?x?xf32>
+  // CHECK:   %[[LOCAL_CONST1:.*]] = arith.constant 1 : index
+  // CHECK:   %[[VAL_2:.*]] = tensor.dim %[[ARG0]], %[[LOCAL_CONST1]] : tensor<?x?xf32>
   // CHECK:   %[[VAL_3:.*]] = tensor.empty(%[[MAX_DIM0]], %[[VAL_2]]) : tensor<?x?xf32>
   // CHECK:   %[[VAL_4:.*]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["parallel", "parallel"]} ins(%[[ARG0]] : tensor<?x?xf32>) outs(%[[VAL_3]] : tensor<?x?xf32>) {
   // CHECK:   ^bb0(%[[VAL_5:.*]]: f32, %[[VAL_6:.*]]: f32):
@@ -284,7 +285,8 @@ func.func @test_add_2d_all_dynamic(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32
   // CHECK: %[[VAL_7:.*]] = tensor.dim %[[ARG0_DIM0_BROADCAST]], %[[CONST1]] : tensor<?x?xf32>
   // CHECK: %[[VAL_8:.*]] = arith.cmpi eq, %[[VAL_7]], %[[CONST1]] : index
   // CHECK: %[[ARG0_DIM1_BROADCAST:.*]] = scf.if %[[VAL_8]] -> (tensor<?x?xf32>) {
-  // CHECK:   %[[VAL_9:.*]] = tensor.dim %[[ARG0_DIM0_BROADCAST]], %[[CONST0]] : tensor<?x?xf32>
+  // CHECK:   %[[LOCAL_CONST0:.*]] = arith.constant 0 : index
+  // CHECK:   %[[VAL_9:.*]] = tensor.dim %[[ARG0_DIM0_BROADCAST]], %[[LOCAL_CONST0]] : tensor<?x?xf32>
   // CHECK:   %[[VAL_10:.*]] = tensor.empty(%[[VAL_9]], %[[MAX_DIM1]]) : tensor<?x?xf32>
   // CHECK:   %[[VAL_11:.*]] = linalg.generic {indexing_maps = [#[[$MAP2]], #[[$MAP1]]], iterator_types = ["parallel", "parallel"]} ins(%[[ARG0_DIM0_BROADCAST]] : tensor<?x?xf32>) outs(%[[VAL_10]] : tensor<?x?xf32>) {
   // CHECK:   ^bb0(%[[VAL_12:.*]]: f32, %[[VAL_13:.*]]: f32):
@@ -298,7 +300,8 @@ func.func @test_add_2d_all_dynamic(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32
   // CHECK: %[[VAL_14:.*]] = tensor.dim %[[ARG1]], %[[CONST0]] : tensor<?x?xf32>
   // CHECK: %[[VAL_15:.*]] = arith.cmpi eq, %[[VAL_14]], %[[CONST1]] : index
   // CHECK: %[[ARG1_DIM0_BROADCAST:.*]] = scf.if %[[VAL_15]] -> (tensor<?x?xf32>) {
-  // CHECK:   %[[VAL_16:.*]] = tensor.dim %[[ARG1]], %[[CONST1]] : tensor<?x?xf32>
+  // CHECK:   %[[LOCAL_CONST1:.*]] = arith.constant 1 : index
+  // CHECK:   %[[VAL_16:.*]] = tensor.dim %[[ARG1]], %[[LOCAL_CONST1]] : tensor<?x?xf32>
   // CHECK:   %[[VAL_17:.*]] = tensor.empty(%[[MAX_DIM0]], %[[VAL_16]]) : tensor<?x?xf32>
   // CHECK:   %[[VAL_18:.*]] = linalg.generic {indexing_maps = [#[[$MAP0]], #[[$MAP1]]], iterator_types = ["parallel", "parallel"]} ins(%[[ARG1]] : tensor<?x?xf32>) outs(%[[VAL_17]] : tensor<?x?xf32>) {
   // CHECK:   ^bb0(%[[VAL_19:.*]]: f32, %[[VAL_20:.*]]: f32):
@@ -312,7 +315,8 @@ func.func @test_add_2d_all_dynamic(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?xf32
   // CHECK: %[[VAL_21:.*]] = tensor.dim %[[ARG1_DIM0_BROADCAST]], %[[CONST1]] : tensor<?x?xf32>
   // CHECK: %[[VAL_22:.*]] = arith.cmpi eq, %[[VAL_21]], %[[CONST1]] : index
   // CHECK: %[[ARG1_DIM1_BROADCAST:.*]] = scf.if %[[VAL_22]] -> (tensor<?x?xf32>) {
-  // CHECK:   %[[VAL_23:.*]] = tensor.dim %[[ARG1_DIM0_BROADCAST]], %[[CONST0]] : tensor<?x?xf32>
+  // CHECK:   %[[LOCAL_CONST0:.*]] = arith.constant 0 : index
+  // CHECK:   %[[VAL_23:.*]] = tensor.dim %[[ARG1_DIM0_BROADCAST]], %[[LOCAL_CONST0]] : tensor<?x?xf32>
   // CHECK:   %[[VAL_24:.*]] = tensor.empty(%[[VAL_23]], %[[MAX_DIM1]]) : tensor<?x?xf32>
   // CHECK:   %[[VAL_25:.*]] = linalg.generic {indexing_maps = [#[[$MAP2]], #[[$MAP1]]], iterator_types = ["parallel", "parallel"]} ins(%[[ARG1_DIM0_BROADCAST]] : tensor<?x?xf32>) outs(%[[VAL_24]] : tensor<?x?xf32>) {
   // CHECK:   ^bb0(%[[VAL_26:.*]]: f32, %[[VAL_27:.*]]: f32):
diff --git a/mlir/test/Integration/Dialect/Tosa/CPU/test-maxpool-dynamic.mlir b/mlir/test/Integration/Dialect/Tosa/CPU/test-maxpool-dynamic.mlir
new file mode 100644
index 0000000000000..05a78e32b9e11
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Tosa/CPU/test-maxpool-dynamic.mlir
@@ -0,0 +1,112 @@
+// DEFINE: %{tosa-to-linalg-pipeline} = -pass-pipeline="builtin.module(func.func(tosa-infer-shapes,tosa-to-linalg-named,tosa-to-linalg,tosa-to-arith))"
+
+// RUN:   mlir-opt %s \
+// RUN:     %{tosa-to-linalg-pipeline} \
+// RUN: | mlir-opt \
+// RUN:     -one-shot-bufferize="bufferize-function-boundaries" \
+// RUN:     -buffer-deallocation-pipeline \
+// RUN:     -test-lower-to-llvm \
+// RUN: | mlir-cpu-runner \
+// RUN:     -entry-point-result=void \
+// RUN:     -shared-libs=%mlir_runner_utils,%mlir_c_runner_utils \
+// RUN: | FileCheck %s
+
+// Validate that the TOSA lowering for tosa.max_pool2d produces the same results when
+// for fully static and fully dynamic inputs.
+
+!tensor_type = tensor<1x4x4x1xf32>
+!memref_type = memref<1x4x4x1xf32>
+
+// Utility functions
+func.func private @printMemrefF32(memref<*xf32>) attributes { llvm.emit_c_interface }
+
+func.func @max_pool_static(%arg0: !tensor_type) -> (!tensor_type) {
+  %0 = tosa.max_pool2d %arg0 {
+    pad = array<i64: 1, 1, 1, 1>,
+    kernel = array<i64: 3, 3>,
+    stride = array<i64: 1, 1>
+  } : (tensor<1x4x4x1xf32>) -> tensor<1x4x4x1xf32>
+  return %0 : tensor<1x4x4x1xf32>
+}
+
+func.func @max_pool_dynamic(%arg0: tensor<?x?x?x?xf32>) -> (tensor<?x?x?x?xf32>) {
+  %0 = tosa.max_pool2d %arg0 {
+    pad = array<i64: 1, 1, 1, 1>,
+    kernel = array<i64: 3, 3>,
+    stride = array<i64: 1, 1>
+  } : (tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
+  return %0 : tensor<?x?x?x?xf32>
+}
+
+// Test harness to compare the results of a fully statically shaped max_pool2d with
+// a fully dynamically shaped max_pool2d on the same inputs.
+func.func @main() {
+  %A = arith.constant dense<[[
+    [[0.0], [0.1], [0.2], [0.3]], // H = 0
+    [[1.0], [1.1], [1.2], [1.3]], // H = 1
+    [[2.0], [2.1], [2.2], [2.3]], // H = 2
+    [[3.0], [3.1], [3.2], [3.3]]  // H = 3
+  ]]> : tensor<1x4x4x1xf32>
+
+  %A_dynamic = tensor.cast %A : !tensor_type to tensor<?x?x?x?xf32>
+
+  // Call both static and dynamically sized variants
+  %result_static  = func.call @max_pool_static(%A) : (!tensor_type) -> !tensor_type
+  %result_dynamic = func.call @max_pool_dynamic(%A_dynamic) : (tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
+
+  %static_buffer = bufferization.to_memref %result_static : !memref_type
+  %unranked_static_buffer = memref.cast %static_buffer : !memref_type to memref<*xf32>
+
+  // CHECK: Unranked Memref base@ = {{.*}} rank = 4 offset = 0 sizes = [1, 4, 4, 1] strides = [16, 4, 1, 1] data =
+
+  // CHECK-NEXT: 1.1
+  // CHECK-NEXT: 1.2
+  // CHECK-NEXT: 1.3
+  // CHECK-NEXT: 1.3
+
+  // CHECK-NEXT: 2.1
+  // CHECK-NEXT: 2.2
+  // CHECK-NEXT: 2.3
+  // CHECK-NEXT: 2.3
+
+  // CHECK-NEXT: 3.1
+  // CHECK-NEXT: 3.2
+  // CHECK-NEXT: 3.3
+  // CHECK-NEXT: 3.3
+
+  // CHECK-NEXT: 3.1
+  // CHECK-NEXT: 3.2
+  // CHECK-NEXT: 3.3
+  // CHECK-NEXT: 3.3
+
+  func.call @printMemrefF32(%unranked_static_buffer) : (memref<*xf32>) -> ()
+
+  %dynamic_buffer = bufferization.to_memref %result_dynamic : memref<?x?x?x?xf32>
+  %unranked_dynamic_buffer = memref.cast %dynamic_buffer : memref<?x?x?x?xf32> to memref<*xf32>
+
+  // CHECK: Unranked Memref base@ = {{.*}} rank = 4 offset = 0 sizes = [1, 4, 4, 1] strides = [16, 4, 1, 1] data =
+  // CHECK-NEXT: 1.1
+  // CHECK-NEXT: 1.2
+  // CHECK-NEXT: 1.3
+  // CHECK-NEXT: 1.3
+
+  // CHECK-NEXT: 2.1
+  // CHECK-NEXT: 2.2
+  // CHECK-NEXT: 2.3
+  // CHECK-NEXT: 2.3
+
+  // CHECK-NEXT: 3.1
+  // CHECK-NEXT: 3.2
+  // CHECK-NEXT: 3.3
+  // CHECK-NEXT: 3.3
+
+  // CHECK-NEXT: 3.1
+  // CHECK-NEXT: 3.2
+  // CHECK-NEXT: 3.3
+  // CHECK-NEXT: 3.3
+
+  func.call @printMemrefF32(%unranked_dynamic_buffer) : (memref<*xf32>) -> ()
+
+  return
+}
+

From d4602a96b48b00a50c4d891673fc622ca2e37f0a Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Tue, 16 Apr 2024 08:14:41 -0400
Subject: [PATCH 578/886] [gn] port fe48bf672e1ab2

---
 llvm/utils/gn/secondary/lldb/test/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/lldb/test/BUILD.gn b/llvm/utils/gn/secondary/lldb/test/BUILD.gn
index 414ea4933c519..c8245739842d9 100644
--- a/llvm/utils/gn/secondary/lldb/test/BUILD.gn
+++ b/llvm/utils/gn/secondary/lldb/test/BUILD.gn
@@ -118,6 +118,7 @@ write_lit_cfg("lit_shell_site_cfg") {
     "LLDB_TOOLS_DIR=" + rebase_path("$root_out_dir/bin"),
     "LLDB_USE_SYSTEM_DEBUGSERVER=1",  # XXX port //lldb/tools/debugserver (?)
     "LLVM_HOST_TRIPLE=$llvm_current_triple",
+    "LLVM_USE_SANITIZER=",
   ]
 
   if (llvm_enable_zlib) {

From c309dc6d0759b23b570c563f611530ff1a49e1bd Mon Sep 17 00:00:00 2001
From: mahtohappy <Happy.Kumar@Windriver.com>
Date: Tue, 16 Apr 2024 17:48:45 +0530
Subject: [PATCH 579/886] [Clang][Sema] placement new initializes typedef array
 with correct size (#83124)

When in-place new-ing a local variable of an array of trivial type, the
generated code calls 'memset' with the correct size of the array,
earlier it was generating size (squared of the typedef array + size).

The cause: `typedef TYPE TArray[8]; TArray x;` The type of declarator is
Tarray[8] and in `SemaExprCXX.cpp::BuildCXXNew` we check if it's of
typedef and of constant size then we get the original type and it works
fine for non-dependent cases.
But in case of template we do `TreeTransform.h:TransformCXXNEWExpr` and
there we again check the allocated type which is TArray[8] and it stays
that way, so ArraySize=(Tarray[8] type, alloc Tarray[8*type]) so the
squared size allocation.

ArraySize gets calculated earlier in `TreeTransform.h` so that
`if(!ArraySize)` condition was failing.
fix: I changed that condition to `if(ArraySize)`.


Fixes #41441
---
 clang/docs/ReleaseNotes.rst                   |  2 ++
 clang/lib/Sema/TreeTransform.h                | 14 ++++++++++++-
 .../instantiate-new-placement-size.cpp        | 20 +++++++++++++++++++
 3 files changed, 35 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/SemaCXX/instantiate-new-placement-size.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 76701dc723b6c..255d2cc044043 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -539,6 +539,8 @@ Bug Fixes to C++ Support
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
 - Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
 - Fix a crash in requires expression with templated base class member function. Fixes (#GH84020).
+- placement new initializes typedef array with correct size
+  (`#GH41441 <https://github.com/llvm/llvm-project/issues/41441>`_)
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 8c96134af7c8f..9d15f3eacbb0f 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -12802,6 +12802,19 @@ TreeTransform<Derived>::TransformCXXNewExpr(CXXNewExpr *E) {
     ArraySize = NewArraySize.get();
   }
 
+  // Per C++0x [expr.new]p5, the type being constructed may be a
+  // typedef of an array type.
+  QualType AllocType = AllocTypeInfo->getType();
+  if (ArraySize) {
+    if (const ConstantArrayType *Array =
+            SemaRef.Context.getAsConstantArrayType(AllocType)) {
+      ArraySize = IntegerLiteral::Create(SemaRef.Context, Array->getSize(),
+                                         SemaRef.Context.getSizeType(),
+                                         E->getBeginLoc());
+      AllocType = Array->getElementType();
+    }
+  }
+
   // Transform the placement arguments (if any).
   bool ArgumentChanged = false;
   SmallVector<Expr*, 8> PlacementArgs;
@@ -12863,7 +12876,6 @@ TreeTransform<Derived>::TransformCXXNewExpr(CXXNewExpr *E) {
     return E;
   }
 
-  QualType AllocType = AllocTypeInfo->getType();
   if (!ArraySize) {
     // If no array size was specified, but the new expression was
     // instantiated with an array type (e.g., "new T" where T is
diff --git a/clang/test/SemaCXX/instantiate-new-placement-size.cpp b/clang/test/SemaCXX/instantiate-new-placement-size.cpp
new file mode 100644
index 0000000000000..7a29d3dee8491
--- /dev/null
+++ b/clang/test/SemaCXX/instantiate-new-placement-size.cpp
@@ -0,0 +1,20 @@
+// RUN: %clang -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
+// Issue no: 41441
+#include <new>
+
+// CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)
+// CHECK: call void @llvm.memset.p0.i64(ptr align 16 %x, i8 0, i64 32, i1 false)
+template <typename TYPE>
+void f()
+{
+    typedef TYPE TArray[8];
+
+    TArray x;
+    new(&x) TArray();
+}
+
+int main()
+{
+    f<char>();
+    f<int>();
+}

From 34013e7ce25868aa8ddea116f79184e8603af56c Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 16 Apr 2024 13:03:09 +0100
Subject: [PATCH 580/886] [X86] Add shuffle tests for
 BLEND(PERMUTE(X),PERMUTE(Y)) patterns

Some very basic tests for a case where we could fold BLEND(PERMUTE(X),PERMUTE(Y)) -> PERMUTE(BLEND(X,Y))

These assume the permute masks are the same, and "complete" (no undefs/duplicate elements) but we could relax that depending on the blend mask
---
 .../X86/vector-shuffle-combining-avx.ll       | 31 +++++++++++++
 .../X86/vector-shuffle-combining-avx512f.ll   | 44 +++++++++++++++++++
 .../X86/vector-shuffle-combining-sse41.ll     | 15 +++++++
 3 files changed, 90 insertions(+)

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
index 0c76c14afb0ae..4859a8e0eaaa5 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
@@ -305,6 +305,37 @@ define <4 x float> @combine_vpermilvar_4f32_as_insertps(<4 x float> %a0) {
   ret <4 x float> %2
 }
 
+define <8 x i32> @combine_blend_of_permutes_v8i32(<4 x i64> %a0, <4 x i64> %a1) {
+; AVX1-LABEL: combine_blend_of_permutes_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX1-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: combine_blend_of_permutes_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3,4],ymm0[5,6],ymm1[7]
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; AVX512-LABEL: combine_blend_of_permutes_v8i32:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    # kill: def $ymm1 killed $ymm1 def $zmm1
+; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [4,21,6,23,16,1,2,19]
+; AVX512-NEXT:    vpermt2d %zmm1, %zmm2, %zmm0
+; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512-NEXT:    ret{{[l|q]}}
+  %s0 = shufflevector <4 x i64> %a0, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+  %s1 = shufflevector <4 x i64> %a1, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 0, i32 1>
+  %x0 = bitcast <4 x i64> %s0 to <8 x i32>
+  %x1 = bitcast <4 x i64> %s1 to <8 x i32>
+  %r = shufflevector <8 x i32> %x0, <8 x i32> %x1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 12, i32 5, i32 6, i32 15>
+  ret <8 x i32> %r
+}
+
 define <2 x double> @constant_fold_vpermilvar_pd() {
 ; CHECK-LABEL: constant_fold_vpermilvar_pd:
 ; CHECK:       # %bb.0:
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
index f53b1eeaf8f54..e87e810971e11 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512f.ll
@@ -973,3 +973,47 @@ define <8 x i64> @combine_vpermvar_insertion_as_broadcast_v8i64(i64 %a0) {
   %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %1, <8 x i64> zeroinitializer)
   ret <8 x i64> %2
 }
+
+define <16 x i32> @blend_of_permutes_v16i32(<8 x i64> %a0, <8x i64> %a1) {
+; X86-AVX512F-LABEL: blend_of_permutes_v16i32:
+; X86-AVX512F:       # %bb.0:
+; X86-AVX512F-NEXT:    vpermq {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5]
+; X86-AVX512F-NEXT:    vpermq {{.*#+}} zmm1 = zmm1[2,3,0,1,6,7,4,5]
+; X86-AVX512F-NEXT:    movw $-25958, %ax # imm = 0x9A9A
+; X86-AVX512F-NEXT:    kmovw %eax, %k1
+; X86-AVX512F-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; X86-AVX512F-NEXT:    retl
+;
+; X86-AVX512BW-LABEL: blend_of_permutes_v16i32:
+; X86-AVX512BW:       # %bb.0:
+; X86-AVX512BW-NEXT:    vpermq {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5]
+; X86-AVX512BW-NEXT:    vpermq {{.*#+}} zmm1 = zmm1[2,3,0,1,6,7,4,5]
+; X86-AVX512BW-NEXT:    movw $-25958, %ax # imm = 0x9A9A
+; X86-AVX512BW-NEXT:    kmovd %eax, %k1
+; X86-AVX512BW-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; X86-AVX512BW-NEXT:    retl
+;
+; X64-AVX512F-LABEL: blend_of_permutes_v16i32:
+; X64-AVX512F:       # %bb.0:
+; X64-AVX512F-NEXT:    vpermq {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5]
+; X64-AVX512F-NEXT:    vpermq {{.*#+}} zmm1 = zmm1[2,3,0,1,6,7,4,5]
+; X64-AVX512F-NEXT:    movw $-25958, %ax # imm = 0x9A9A
+; X64-AVX512F-NEXT:    kmovw %eax, %k1
+; X64-AVX512F-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; X64-AVX512F-NEXT:    retq
+;
+; X64-AVX512BW-LABEL: blend_of_permutes_v16i32:
+; X64-AVX512BW:       # %bb.0:
+; X64-AVX512BW-NEXT:    vpermq {{.*#+}} zmm0 = zmm0[2,3,0,1,6,7,4,5]
+; X64-AVX512BW-NEXT:    vpermq {{.*#+}} zmm1 = zmm1[2,3,0,1,6,7,4,5]
+; X64-AVX512BW-NEXT:    movw $-25958, %ax # imm = 0x9A9A
+; X64-AVX512BW-NEXT:    kmovd %eax, %k1
+; X64-AVX512BW-NEXT:    vmovdqa32 %zmm1, %zmm0 {%k1}
+; X64-AVX512BW-NEXT:    retq
+  %s0 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5>
+  %s1 = shufflevector <8 x i64> %a1, <8 x i64> undef, <8 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5>
+  %x0 = bitcast <8 x i64> %s0 to <16 x i32>
+  %x1 = bitcast <8 x i64> %s1 to <16 x i32>
+  %r = shufflevector <16 x i32> %x0, <16 x i32> %x1, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 20, i32 5, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 28, i32 13, i32 14, i32 31>
+  ret <16 x i32> %r
+}
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
index 5eb017bc80ca5..33851f56fe8de 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
@@ -22,6 +22,21 @@ define <16 x i8> @combine_vpshufb_as_movzx(<16 x i8> %a0) {
   ret <16 x i8> %res0
 }
 
+define <4 x i32> @combine_blend_of_permutes_v4i32(<2 x i64> %a0, <2 x i64> %a1) {
+; SSE-LABEL: combine_blend_of_permutes_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3],xmm2[4,5],xmm0[6,7]
+; SSE-NEXT:    retq
+  %s0 = shufflevector <2 x i64> %a0, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
+  %s1 = shufflevector <2 x i64> %a1, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
+  %x0 = bitcast <2 x i64> %s0 to <4 x i32>
+  %x1 = bitcast <2 x i64> %s1 to <4 x i32>
+  %r = shufflevector <4 x i32> %x0, <4 x i32> %x1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+  ret <4 x i32> %r
+}
+
 define <16 x i8> @PR50049(ptr %p1, ptr %p2) {
 ; SSE-LABEL: PR50049:
 ; SSE:       # %bb.0:

From b73476c7843f21966acb2fb5cab8515d9ec02905 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Tue, 16 Apr 2024 13:29:34 +0100
Subject: [PATCH 581/886] [SLP] Make sure MinVF is a power-of-2 by using
 PowerOf2Ceil.

This should ensure we explore the same VFs as before 6d66db3890a18e39.

Fixes https://github.com/llvm/llvm-project/issues/88640.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    |  4 +--
 .../trunc-store-value-ty-not-power-of-2.ll    | 33 +++++++++++++++++++
 2 files changed, 35 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index c63b500f546f3..d0bcdceae392b 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15155,8 +15155,8 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
       Type *ValueTy = StoreTy;
       if (auto *Trunc = dyn_cast<TruncInst>(Store->getValueOperand()))
         ValueTy = Trunc->getSrcTy();
-      unsigned MinVF = TTI->getStoreMinimumVF(
-          R.getMinVF(DL->getTypeSizeInBits(StoreTy)), StoreTy, ValueTy);
+      unsigned MinVF = PowerOf2Ceil(TTI->getStoreMinimumVF(
+          R.getMinVF(DL->getTypeStoreSizeInBits(StoreTy)), StoreTy, ValueTy));
 
       if (MaxVF < MinVF) {
         LLVM_DEBUG(dbgs() << "SLP: Vectorization infeasible as MaxVF (" << MaxVF
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll b/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll
index 81b4ee40e7fdf..2f0fad70b593b 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/trunc-store-value-ty-not-power-of-2.ll
@@ -107,3 +107,36 @@ define void @test_4_trunc_i24_to_i16(i24 %x, ptr %A) {
   store i16 %t, ptr %gep.3, align 1
   ret void
 }
+
+%struct.d = type { [3 x i8], [3 x i8], [2 x i8] }
+
+; Test case for https://github.com/llvm/llvm-project/issues/88640.
+define void @test_access_i24_directly(ptr %src, ptr noalias %dst) "target-cpu"="btver2" {
+; CHECK-LABEL: define void @test_access_i24_directly(
+; CHECK-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr [[SRC]], align 8
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i24
+; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds [[STRUCT_D:%.*]], ptr [[SRC]], i64 0, i32 1
+; CHECK-NEXT:    [[BF_LOAD:%.*]] = load i24, ptr [[GEP_SRC]], align 1
+; CHECK-NEXT:    [[BF_VALUE:%.*]] = and i24 [[TMP1]], 8388607
+; CHECK-NEXT:    [[BF_CLEAR:%.*]] = and i24 [[BF_LOAD]], -8388608
+; CHECK-NEXT:    [[BF_SET:%.*]] = or disjoint i24 [[BF_CLEAR]], [[BF_VALUE]]
+; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds [[STRUCT_D]], ptr [[DST]], i64 0, i32 1
+; CHECK-NEXT:    store i24 [[BF_SET]], ptr [[GEP_DST]], align 1
+; CHECK-NEXT:    store i24 0, ptr [[DST]], align 8
+; CHECK-NEXT:    ret void
+;
+entry:
+  %0 = load i64, ptr %src, align 8
+  %1 = trunc i64 %0 to i24
+  %gep.src = getelementptr inbounds %struct.d, ptr %src, i64 0, i32 1
+  %bf.load = load i24, ptr %gep.src, align 1
+  %bf.value = and i24 %1, 8388607
+  %bf.clear = and i24 %bf.load, -8388608
+  %bf.set = or disjoint i24 %bf.clear, %bf.value
+  %gep.dst = getelementptr inbounds %struct.d, ptr %dst, i64 0, i32 1
+  store i24 %bf.set, ptr %gep.dst, align 1
+  store i24 0, ptr %dst, align 8
+  ret void
+}

From e272c37934a06cd80b9b072afc09afae5fd8c218 Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Tue, 16 Apr 2024 08:31:35 -0400
Subject: [PATCH 582/886] clang; Try to get windows-seh-async-verify.cpp to
 pass on mac

On macOS, file paths start with /Users/..., which clang-cl interptrets
as the /U switch followed by a preprocessor macro name to undefine.

Put the filename after `--` to prevent this. For consistency, move %s
to the end of the regular `clang` lines (where this isn't needed) as
well.
---
 clang/test/Driver/windows-seh-async-verify.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/clang/test/Driver/windows-seh-async-verify.cpp b/clang/test/Driver/windows-seh-async-verify.cpp
index 5fda6a77dba04..ace93cf44a31d 100644
--- a/clang/test/Driver/windows-seh-async-verify.cpp
+++ b/clang/test/Driver/windows-seh-async-verify.cpp
@@ -1,7 +1,7 @@
-// RUN: %clang --target=x86_64-pc-windows -fasync-exceptions -fsyntax-only %s -### 2>&1 | FileCheck %s
-// RUN: %clang_cl --target=x86_64-pc-windows /EHa -fsyntax-only %s -### 2>&1 | FileCheck %s
-// RUN: %clang --target=x86_64-pc-windows-gnu -fasync-exceptions -fsyntax-only %s -### 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,GNU
-// RUN: %clang_cl --target=x86_64-pc-windows-gnu /EHa -fsyntax-only %s -### 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,CL-GNU
+// RUN: %clang --target=x86_64-pc-windows -fasync-exceptions -fsyntax-only -### %s 2>&1 | FileCheck %s
+// RUN: %clang_cl --target=x86_64-pc-windows /EHa -fsyntax-only -### -- %s 2>&1 | FileCheck %s
+// RUN: %clang --target=x86_64-pc-windows-gnu -fasync-exceptions -fsyntax-only -### %s 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,GNU
+// RUN: %clang_cl --target=x86_64-pc-windows-gnu /EHa -fsyntax-only -### -- %s 2>&1 | FileCheck %s --check-prefixes=GNU-ALL,CL-GNU
 
 // CHECK-NOT: warning
 // GNU: warning: argument unused during compilation: '-fasync-exceptions' [-Wunused-command-line-argument]

From f69ded0d9965a6b2b76ce12db876c70f249d96d1 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Tue, 16 Apr 2024 16:36:53 +0400
Subject: [PATCH 583/886] [clang] Introduce `SemaOpenMP` (#88642)

This patch moves OpenMP-related entities out of `Sema` to a newly
created `SemaOpenMP` class. This is a part of the effort to split `Sema`
up, and follows the recent example of CUDA, OpenACC, SYCL, HLSL.
Additional context can be found in
https://github.com/llvm/llvm-project/pull/82217,
https://github.com/llvm/llvm-project/pull/84184,
https://github.com/llvm/llvm-project/pull/87634.
---
 clang/include/clang/Parse/Parser.h            |   15 +-
 clang/include/clang/Sema/Sema.h               | 1419 +-----
 clang/include/clang/Sema/SemaOpenMP.h         | 1447 ++++++
 clang/lib/Parse/ParseDecl.cpp                 |    3 +-
 clang/lib/Parse/ParseExpr.cpp                 |    5 +-
 clang/lib/Parse/ParseOpenMP.cpp               |  240 +-
 clang/lib/Parse/ParseStmt.cpp                 |    3 +-
 clang/lib/Sema/Sema.cpp                       |   25 +-
 clang/lib/Sema/SemaDecl.cpp                   |   32 +-
 clang/lib/Sema/SemaDeclCXX.cpp                |    9 +-
 clang/lib/Sema/SemaExpr.cpp                   |  610 +--
 clang/lib/Sema/SemaExprMember.cpp             |    9 +-
 clang/lib/Sema/SemaLambda.cpp                 |    3 +-
 clang/lib/Sema/SemaOpenMP.cpp                 | 4333 ++++++++++-------
 clang/lib/Sema/SemaStmt.cpp                   |    6 +-
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |   60 +-
 clang/lib/Sema/SemaType.cpp                   |    3 +-
 clang/lib/Sema/TreeTransform.h                |  684 +--
 18 files changed, 4565 insertions(+), 4341 deletions(-)
 create mode 100644 clang/include/clang/Sema/SemaOpenMP.h

diff --git a/clang/include/clang/Parse/Parser.h b/clang/include/clang/Parse/Parser.h
index 5950dd74cfe83..23b268126de4e 100644
--- a/clang/include/clang/Parse/Parser.h
+++ b/clang/include/clang/Parse/Parser.h
@@ -18,6 +18,7 @@
 #include "clang/Lex/CodeCompletionHandler.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/Sema.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Frontend/OpenMP/OMPContext.h"
 #include "llvm/Support/SaveAndRestore.h"
@@ -2537,7 +2538,7 @@ class Parser : public CodeCompletionHandler {
   /// Returns true for declaration, false for expression.
   bool isForInitDeclaration() {
     if (getLangOpts().OpenMP)
-      Actions.startOpenMPLoop();
+      Actions.OpenMP().startOpenMPLoop();
     if (getLangOpts().CPlusPlus)
       return Tok.is(tok::kw_using) ||
              isCXXSimpleDeclaration(/*AllowForRangeDecl=*/true);
@@ -3396,7 +3397,7 @@ class Parser : public CodeCompletionHandler {
                           SourceLocation Loc);
 
   /// Parse clauses for '#pragma omp [begin] declare target'.
-  void ParseOMPDeclareTargetClauses(Sema::DeclareTargetContextInfo &DTCI);
+  void ParseOMPDeclareTargetClauses(SemaOpenMP::DeclareTargetContextInfo &DTCI);
 
   /// Parse '#pragma omp end declare target'.
   void ParseOMPEndDeclareTargetDirective(OpenMPDirectiveKind BeginDKind,
@@ -3486,7 +3487,7 @@ class Parser : public CodeCompletionHandler {
   /// Parses indirect clause
   /// \param ParseOnly true to skip the clause's semantic actions and return
   // false;
-  bool ParseOpenMPIndirectClause(Sema::DeclareTargetContextInfo &DTCI,
+  bool ParseOpenMPIndirectClause(SemaOpenMP::DeclareTargetContextInfo &DTCI,
                                  bool ParseOnly);
   /// Parses clause with a single expression and an additional argument
   /// of a kind \a Kind.
@@ -3556,12 +3557,12 @@ class Parser : public CodeCompletionHandler {
 
   /// Parses a reserved locator like 'omp_all_memory'.
   bool ParseOpenMPReservedLocator(OpenMPClauseKind Kind,
-                                  Sema::OpenMPVarListDataTy &Data,
+                                  SemaOpenMP::OpenMPVarListDataTy &Data,
                                   const LangOptions &LangOpts);
   /// Parses clauses with list.
   bool ParseOpenMPVarList(OpenMPDirectiveKind DKind, OpenMPClauseKind Kind,
                           SmallVectorImpl<Expr *> &Vars,
-                          Sema::OpenMPVarListDataTy &Data);
+                          SemaOpenMP::OpenMPVarListDataTy &Data);
   bool ParseUnqualifiedId(CXXScopeSpec &SS, ParsedType ObjectType,
                           bool ObjectHadErrors, bool EnteringContext,
                           bool AllowDestructorName, bool AllowConstructorName,
@@ -3569,11 +3570,11 @@ class Parser : public CodeCompletionHandler {
                           SourceLocation *TemplateKWLoc, UnqualifiedId &Result);
 
   /// Parses the mapper modifier in map, to, and from clauses.
-  bool parseMapperModifier(Sema::OpenMPVarListDataTy &Data);
+  bool parseMapperModifier(SemaOpenMP::OpenMPVarListDataTy &Data);
   /// Parses map-type-modifiers in map clause.
   /// map([ [map-type-modifier[,] [map-type-modifier[,] ...] map-type : ] list)
   /// where, map-type-modifier ::= always | close | mapper(mapper-identifier)
-  bool parseMapTypeModifiers(Sema::OpenMPVarListDataTy &Data);
+  bool parseMapTypeModifiers(SemaOpenMP::OpenMPVarListDataTy &Data);
 
   //===--------------------------------------------------------------------===//
   // OpenACC Parsing.
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index d93ac7863b721..a5fe83a539aaf 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -26,14 +26,12 @@
 #include "clang/AST/ExprCXX.h"
 #include "clang/AST/ExprConcepts.h"
 #include "clang/AST/ExprObjC.h"
-#include "clang/AST/ExprOpenMP.h"
 #include "clang/AST/ExternalASTSource.h"
 #include "clang/AST/LocInfoType.h"
 #include "clang/AST/MangleNumberingContext.h"
 #include "clang/AST/NSAPI.h"
 #include "clang/AST/PrettyPrinter.h"
 #include "clang/AST/StmtCXX.h"
-#include "clang/AST/StmtOpenMP.h"
 #include "clang/AST/TypeLoc.h"
 #include "clang/AST/TypeOrdering.h"
 #include "clang/Basic/BitmaskEnum.h"
@@ -43,7 +41,6 @@
 #include "clang/Basic/ExpressionTraits.h"
 #include "clang/Basic/Module.h"
 #include "clang/Basic/OpenCLOptions.h"
-#include "clang/Basic/OpenMPKinds.h"
 #include "clang/Basic/PragmaKinds.h"
 #include "clang/Basic/Specifiers.h"
 #include "clang/Basic/TemplateKinds.h"
@@ -68,7 +65,6 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/TinyPtrVector.h"
-#include "llvm/Frontend/OpenMP/OMPConstants.h"
 #include <deque>
 #include <memory>
 #include <optional>
@@ -167,12 +163,6 @@ class ObjCMessageExpr;
 class ObjCMethodDecl;
 class ObjCPropertyDecl;
 class ObjCProtocolDecl;
-class OMPThreadPrivateDecl;
-class OMPRequiresDecl;
-class OMPDeclareReductionDecl;
-class OMPDeclareSimdDecl;
-class OMPClause;
-struct OMPVarListLocTy;
 struct OverloadCandidate;
 enum class OverloadCandidateParamOrder : char;
 enum OverloadCandidateRewriteKind : unsigned;
@@ -187,6 +177,7 @@ class QualType;
 class SemaCUDA;
 class SemaHLSL;
 class SemaOpenACC;
+class SemaOpenMP;
 class SemaSYCL;
 class StandardConversionSequence;
 class Stmt;
@@ -480,7 +471,6 @@ class Sema final : public SemaBase {
   // 35. Code Completion (SemaCodeComplete.cpp)
   // 36. FixIt Helpers (SemaFixItUtils.cpp)
   // 37. Name Lookup for RISC-V Vector Intrinsic (SemaRISCVVectorLookup.cpp)
-  // 38. OpenMP Directives and Clauses (SemaOpenMP.cpp)
 
   /// \name Semantic Analysis
   /// Implementations are in Sema.cpp
@@ -997,6 +987,11 @@ class Sema final : public SemaBase {
     return *OpenACCPtr;
   }
 
+  SemaOpenMP &OpenMP() {
+    assert(OpenMPPtr && "SemaOpenMP is dead");
+    return *OpenMPPtr;
+  }
+
   SemaSYCL &SYCL() {
     assert(SYCLPtr);
     return *SYCLPtr;
@@ -1035,6 +1030,7 @@ class Sema final : public SemaBase {
   std::unique_ptr<SemaCUDA> CUDAPtr;
   std::unique_ptr<SemaHLSL> HLSLPtr;
   std::unique_ptr<SemaOpenACC> OpenACCPtr;
+  std::unique_ptr<SemaOpenMP> OpenMPPtr;
   std::unique_ptr<SemaSYCL> SYCLPtr;
 
   ///@}
@@ -3443,14 +3439,6 @@ class Sema final : public SemaBase {
 
   sema::LambdaScopeInfo *RebuildLambdaScopeInfo(CXXMethodDecl *CallOperator);
 
-  /// The declarator \p D defines a function in the scope \p S which is nested
-  /// in an `omp begin/end declare variant` scope. In this method we create a
-  /// declaration for \p D and rename \p D according to the OpenMP context
-  /// selector of the surrounding scope. Return all base functions in \p Bases.
-  void ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
-      Scope *S, Declarator &D, MultiTemplateParamsArg TemplateParameterLists,
-      SmallVectorImpl<FunctionDecl *> &Bases);
-
   // Heuristically tells if the function is `get_return_object` member of a
   // coroutine promise_type by matching the function name.
   static bool CanBeGetReturnObject(const FunctionDecl *FD);
@@ -5533,32 +5521,6 @@ class Sema final : public SemaBase {
                                               Expr *ColumnIdx,
                                               SourceLocation RBLoc);
 
-  ExprResult ActOnOMPArraySectionExpr(Expr *Base, SourceLocation LBLoc,
-                                      Expr *LowerBound,
-                                      SourceLocation ColonLocFirst,
-                                      SourceLocation ColonLocSecond,
-                                      Expr *Length, Expr *Stride,
-                                      SourceLocation RBLoc);
-  ExprResult ActOnOMPArrayShapingExpr(Expr *Base, SourceLocation LParenLoc,
-                                      SourceLocation RParenLoc,
-                                      ArrayRef<Expr *> Dims,
-                                      ArrayRef<SourceRange> Brackets);
-
-  /// Data structure for iterator expression.
-  struct OMPIteratorData {
-    IdentifierInfo *DeclIdent = nullptr;
-    SourceLocation DeclIdentLoc;
-    ParsedType Type;
-    OMPIteratorExpr::IteratorRange Range;
-    SourceLocation AssignLoc;
-    SourceLocation ColonLoc;
-    SourceLocation SecColonLoc;
-  };
-
-  ExprResult ActOnOMPIteratorExpr(Scope *S, SourceLocation IteratorKwLoc,
-                                  SourceLocation LLoc, SourceLocation RLoc,
-                                  ArrayRef<OMPIteratorData> Data);
-
   bool ConvertArgumentsForCall(CallExpr *Call, Expr *Fn, FunctionDecl *FDecl,
                                const FunctionProtoType *Proto,
                                ArrayRef<Expr *> Args, SourceLocation RParenLoc,
@@ -12863,1373 +12825,6 @@ class Sema final : public SemaBase {
   std::unique_ptr<sema::RISCVIntrinsicManager> RVIntrinsicManager;
 
   ///@}
-
-  //
-  //
-  // -------------------------------------------------------------------------
-  //
-  //
-
-  /// \name OpenMP Directives and Clauses
-  /// Implementations are in SemaOpenMP.cpp
-  ///@{
-
-public:
-  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
-  /// context is "used as device code".
-  ///
-  /// - If CurContext is a `declare target` function or it is known that the
-  /// function is emitted for the device, emits the diagnostics immediately.
-  /// - If CurContext is a non-`declare target` function and we are compiling
-  ///   for the device, creates a diagnostic which is emitted if and when we
-  ///   realize that the function will be codegen'ed.
-  ///
-  /// Example usage:
-  ///
-  ///  // Variable-length arrays are not allowed in NVPTX device code.
-  ///  if (diagIfOpenMPDeviceCode(Loc, diag::err_vla_unsupported))
-  ///    return ExprError();
-  ///  // Otherwise, continue parsing as normal.
-  SemaDiagnosticBuilder diagIfOpenMPDeviceCode(SourceLocation Loc,
-                                               unsigned DiagID,
-                                               const FunctionDecl *FD);
-
-  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
-  /// context is "used as host code".
-  ///
-  /// - If CurContext is a `declare target` function or it is known that the
-  /// function is emitted for the host, emits the diagnostics immediately.
-  /// - If CurContext is a non-host function, just ignore it.
-  ///
-  /// Example usage:
-  ///
-  ///  // Variable-length arrays are not allowed in NVPTX device code.
-  ///  if (diagIfOpenMPHostode(Loc, diag::err_vla_unsupported))
-  ///    return ExprError();
-  ///  // Otherwise, continue parsing as normal.
-  SemaDiagnosticBuilder diagIfOpenMPHostCode(SourceLocation Loc,
-                                             unsigned DiagID,
-                                             const FunctionDecl *FD);
-
-  /// Register \p D as specialization of all base functions in \p Bases in the
-  /// current `omp begin/end declare variant` scope.
-  void ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
-      Decl *D, SmallVectorImpl<FunctionDecl *> &Bases);
-
-  /// Act on \p D, a function definition inside of an `omp [begin/end] assumes`.
-  void ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D);
-
-  /// Can we exit an OpenMP declare variant scope at the moment.
-  bool isInOpenMPDeclareVariantScope() const {
-    return !OMPDeclareVariantScopes.empty();
-  }
-
-  ExprResult
-  VerifyPositiveIntegerConstantInClause(Expr *Op, OpenMPClauseKind CKind,
-                                        bool StrictlyPositive = true,
-                                        bool SuppressExprDiags = false);
-
-  /// Given the potential call expression \p Call, determine if there is a
-  /// specialization via the OpenMP declare variant mechanism available. If
-  /// there is, return the specialized call expression, otherwise return the
-  /// original \p Call.
-  ExprResult ActOnOpenMPCall(ExprResult Call, Scope *Scope,
-                             SourceLocation LParenLoc, MultiExprArg ArgExprs,
-                             SourceLocation RParenLoc, Expr *ExecConfig);
-
-  /// Handle a `omp begin declare variant`.
-  void ActOnOpenMPBeginDeclareVariant(SourceLocation Loc, OMPTraitInfo &TI);
-
-  /// Handle a `omp end declare variant`.
-  void ActOnOpenMPEndDeclareVariant();
-
-  /// Function tries to capture lambda's captured variables in the OpenMP region
-  /// before the original lambda is captured.
-  void tryCaptureOpenMPLambdas(ValueDecl *V);
-
-  /// Return true if the provided declaration \a VD should be captured by
-  /// reference.
-  /// \param Level Relative level of nested OpenMP construct for that the check
-  /// is performed.
-  /// \param OpenMPCaptureLevel Capture level within an OpenMP construct.
-  bool isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
-                             unsigned OpenMPCaptureLevel) const;
-
-  /// Check if the specified variable is used in one of the private
-  /// clauses (private, firstprivate, lastprivate, reduction etc.) in OpenMP
-  /// constructs.
-  VarDecl *isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo = false,
-                                unsigned StopAt = 0);
-
-  /// The member expression(this->fd) needs to be rebuilt in the template
-  /// instantiation to generate private copy for OpenMP when default
-  /// clause is used. The function will return true if default
-  /// cluse is used.
-  bool isOpenMPRebuildMemberExpr(ValueDecl *D);
-
-  ExprResult getOpenMPCapturedExpr(VarDecl *Capture, ExprValueKind VK,
-                                   ExprObjectKind OK, SourceLocation Loc);
-
-  /// If the current region is a loop-based region, mark the start of the loop
-  /// construct.
-  void startOpenMPLoop();
-
-  /// If the current region is a range loop-based region, mark the start of the
-  /// loop construct.
-  void startOpenMPCXXRangeFor();
-
-  /// Check if the specified variable is used in 'private' clause.
-  /// \param Level Relative level of nested OpenMP construct for that the check
-  /// is performed.
-  OpenMPClauseKind isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
-                                       unsigned CapLevel) const;
-
-  /// Sets OpenMP capture kind (OMPC_private, OMPC_firstprivate, OMPC_map etc.)
-  /// for \p FD based on DSA for the provided corresponding captured declaration
-  /// \p D.
-  void setOpenMPCaptureKind(FieldDecl *FD, const ValueDecl *D, unsigned Level);
-
-  /// Check if the specified variable is captured  by 'target' directive.
-  /// \param Level Relative level of nested OpenMP construct for that the check
-  /// is performed.
-  bool isOpenMPTargetCapturedDecl(const ValueDecl *D, unsigned Level,
-                                  unsigned CaptureLevel) const;
-
-  /// Check if the specified global variable must be captured  by outer capture
-  /// regions.
-  /// \param Level Relative level of nested OpenMP construct for that
-  /// the check is performed.
-  bool isOpenMPGlobalCapturedDecl(ValueDecl *D, unsigned Level,
-                                  unsigned CaptureLevel) const;
-
-  ExprResult PerformOpenMPImplicitIntegerConversion(SourceLocation OpLoc,
-                                                    Expr *Op);
-  /// Called on start of new data sharing attribute block.
-  void StartOpenMPDSABlock(OpenMPDirectiveKind K,
-                           const DeclarationNameInfo &DirName, Scope *CurScope,
-                           SourceLocation Loc);
-  /// Start analysis of clauses.
-  void StartOpenMPClause(OpenMPClauseKind K);
-  /// End analysis of clauses.
-  void EndOpenMPClause();
-  /// Called on end of data sharing attribute block.
-  void EndOpenMPDSABlock(Stmt *CurDirective);
-
-  /// Check if the current region is an OpenMP loop region and if it is,
-  /// mark loop control variable, used in \p Init for loop initialization, as
-  /// private by default.
-  /// \param Init First part of the for loop.
-  void ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init);
-
-  /// Called on well-formed '\#pragma omp metadirective' after parsing
-  /// of the  associated statement.
-  StmtResult ActOnOpenMPMetaDirective(ArrayRef<OMPClause *> Clauses,
-                                      Stmt *AStmt, SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-
-  // OpenMP directives and clauses.
-  /// Called on correct id-expression from the '#pragma omp
-  /// threadprivate'.
-  ExprResult ActOnOpenMPIdExpression(Scope *CurScope, CXXScopeSpec &ScopeSpec,
-                                     const DeclarationNameInfo &Id,
-                                     OpenMPDirectiveKind Kind);
-  /// Called on well-formed '#pragma omp threadprivate'.
-  DeclGroupPtrTy ActOnOpenMPThreadprivateDirective(SourceLocation Loc,
-                                                   ArrayRef<Expr *> VarList);
-  /// Builds a new OpenMPThreadPrivateDecl and checks its correctness.
-  OMPThreadPrivateDecl *CheckOMPThreadPrivateDecl(SourceLocation Loc,
-                                                  ArrayRef<Expr *> VarList);
-  /// Called on well-formed '#pragma omp allocate'.
-  DeclGroupPtrTy ActOnOpenMPAllocateDirective(SourceLocation Loc,
-                                              ArrayRef<Expr *> VarList,
-                                              ArrayRef<OMPClause *> Clauses,
-                                              DeclContext *Owner = nullptr);
-
-  /// Called on well-formed '#pragma omp [begin] assume[s]'.
-  void ActOnOpenMPAssumesDirective(SourceLocation Loc,
-                                   OpenMPDirectiveKind DKind,
-                                   ArrayRef<std::string> Assumptions,
-                                   bool SkippedClauses);
-
-  /// Check if there is an active global `omp begin assumes` directive.
-  bool isInOpenMPAssumeScope() const { return !OMPAssumeScoped.empty(); }
-
-  /// Check if there is an active global `omp assumes` directive.
-  bool hasGlobalOpenMPAssumes() const { return !OMPAssumeGlobal.empty(); }
-
-  /// Called on well-formed '#pragma omp end assumes'.
-  void ActOnOpenMPEndAssumesDirective();
-
-  /// Called on well-formed '#pragma omp requires'.
-  DeclGroupPtrTy ActOnOpenMPRequiresDirective(SourceLocation Loc,
-                                              ArrayRef<OMPClause *> ClauseList);
-  /// Check restrictions on Requires directive
-  OMPRequiresDecl *CheckOMPRequiresDecl(SourceLocation Loc,
-                                        ArrayRef<OMPClause *> Clauses);
-  /// Check if the specified type is allowed to be used in 'omp declare
-  /// reduction' construct.
-  QualType ActOnOpenMPDeclareReductionType(SourceLocation TyLoc,
-                                           TypeResult ParsedType);
-  /// Called on start of '#pragma omp declare reduction'.
-  DeclGroupPtrTy ActOnOpenMPDeclareReductionDirectiveStart(
-      Scope *S, DeclContext *DC, DeclarationName Name,
-      ArrayRef<std::pair<QualType, SourceLocation>> ReductionTypes,
-      AccessSpecifier AS, Decl *PrevDeclInScope = nullptr);
-  /// Initialize declare reduction construct initializer.
-  void ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D);
-  /// Finish current declare reduction construct initializer.
-  void ActOnOpenMPDeclareReductionCombinerEnd(Decl *D, Expr *Combiner);
-  /// Initialize declare reduction construct initializer.
-  /// \return omp_priv variable.
-  VarDecl *ActOnOpenMPDeclareReductionInitializerStart(Scope *S, Decl *D);
-  /// Finish current declare reduction construct initializer.
-  void ActOnOpenMPDeclareReductionInitializerEnd(Decl *D, Expr *Initializer,
-                                                 VarDecl *OmpPrivParm);
-  /// Called at the end of '#pragma omp declare reduction'.
-  DeclGroupPtrTy ActOnOpenMPDeclareReductionDirectiveEnd(
-      Scope *S, DeclGroupPtrTy DeclReductions, bool IsValid);
-
-  /// Check variable declaration in 'omp declare mapper' construct.
-  TypeResult ActOnOpenMPDeclareMapperVarDecl(Scope *S, Declarator &D);
-  /// Check if the specified type is allowed to be used in 'omp declare
-  /// mapper' construct.
-  QualType ActOnOpenMPDeclareMapperType(SourceLocation TyLoc,
-                                        TypeResult ParsedType);
-  /// Called on start of '#pragma omp declare mapper'.
-  DeclGroupPtrTy ActOnOpenMPDeclareMapperDirective(
-      Scope *S, DeclContext *DC, DeclarationName Name, QualType MapperType,
-      SourceLocation StartLoc, DeclarationName VN, AccessSpecifier AS,
-      Expr *MapperVarRef, ArrayRef<OMPClause *> Clauses,
-      Decl *PrevDeclInScope = nullptr);
-  /// Build the mapper variable of '#pragma omp declare mapper'.
-  ExprResult ActOnOpenMPDeclareMapperDirectiveVarDecl(Scope *S,
-                                                      QualType MapperType,
-                                                      SourceLocation StartLoc,
-                                                      DeclarationName VN);
-  void ActOnOpenMPIteratorVarDecl(VarDecl *VD);
-  bool isOpenMPDeclareMapperVarDeclAllowed(const VarDecl *VD) const;
-  const ValueDecl *getOpenMPDeclareMapperVarName() const;
-
-  struct DeclareTargetContextInfo {
-    struct MapInfo {
-      OMPDeclareTargetDeclAttr::MapTypeTy MT;
-      SourceLocation Loc;
-    };
-    /// Explicitly listed variables and functions in a 'to' or 'link' clause.
-    llvm::DenseMap<NamedDecl *, MapInfo> ExplicitlyMapped;
-
-    /// The 'device_type' as parsed from the clause.
-    OMPDeclareTargetDeclAttr::DevTypeTy DT = OMPDeclareTargetDeclAttr::DT_Any;
-
-    /// The directive kind, `begin declare target` or `declare target`.
-    OpenMPDirectiveKind Kind;
-
-    /// The directive with indirect clause.
-    std::optional<Expr *> Indirect;
-
-    /// The directive location.
-    SourceLocation Loc;
-
-    DeclareTargetContextInfo(OpenMPDirectiveKind Kind, SourceLocation Loc)
-        : Kind(Kind), Loc(Loc) {}
-  };
-
-  /// Called on the start of target region i.e. '#pragma omp declare target'.
-  bool ActOnStartOpenMPDeclareTargetContext(DeclareTargetContextInfo &DTCI);
-
-  /// Called at the end of target region i.e. '#pragma omp end declare target'.
-  const DeclareTargetContextInfo ActOnOpenMPEndDeclareTargetDirective();
-
-  /// Called once a target context is completed, that can be when a
-  /// '#pragma omp end declare target' was encountered or when a
-  /// '#pragma omp declare target' without declaration-definition-seq was
-  /// encountered.
-  void ActOnFinishedOpenMPDeclareTargetContext(DeclareTargetContextInfo &DTCI);
-
-  /// Report unterminated 'omp declare target' or 'omp begin declare target' at
-  /// the end of a compilation unit.
-  void DiagnoseUnterminatedOpenMPDeclareTarget();
-
-  /// Searches for the provided declaration name for OpenMP declare target
-  /// directive.
-  NamedDecl *lookupOpenMPDeclareTargetName(Scope *CurScope,
-                                           CXXScopeSpec &ScopeSpec,
-                                           const DeclarationNameInfo &Id);
-
-  /// Called on correct id-expression from the '#pragma omp declare target'.
-  void ActOnOpenMPDeclareTargetName(NamedDecl *ND, SourceLocation Loc,
-                                    OMPDeclareTargetDeclAttr::MapTypeTy MT,
-                                    DeclareTargetContextInfo &DTCI);
-
-  /// Check declaration inside target region.
-  void
-  checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
-                                   SourceLocation IdLoc = SourceLocation());
-
-  /// Adds OMPDeclareTargetDeclAttr to referenced variables in declare target
-  /// directive.
-  void ActOnOpenMPDeclareTargetInitializer(Decl *D);
-
-  /// Finishes analysis of the deferred functions calls that may be declared as
-  /// host/nohost during device/host compilation.
-  void finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
-                                     const FunctionDecl *Callee,
-                                     SourceLocation Loc);
-
-  /// Return true if currently in OpenMP task with untied clause context.
-  bool isInOpenMPTaskUntiedContext() const;
-
-  /// Return true inside OpenMP declare target region.
-  bool isInOpenMPDeclareTargetContext() const {
-    return !DeclareTargetNesting.empty();
-  }
-  /// Return true inside OpenMP target region.
-  bool isInOpenMPTargetExecutionDirective() const;
-
-  /// Return the number of captured regions created for an OpenMP directive.
-  static int getOpenMPCaptureLevels(OpenMPDirectiveKind Kind);
-
-  /// Initialization of captured region for OpenMP region.
-  void ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope);
-
-  /// Called for syntactical loops (ForStmt or CXXForRangeStmt) associated to
-  /// an OpenMP loop directive.
-  StmtResult ActOnOpenMPCanonicalLoop(Stmt *AStmt);
-
-  /// Process a canonical OpenMP loop nest that can either be a canonical
-  /// literal loop (ForStmt or CXXForRangeStmt), or the generated loop of an
-  /// OpenMP loop transformation construct.
-  StmtResult ActOnOpenMPLoopnest(Stmt *AStmt);
-
-  /// End of OpenMP region.
-  ///
-  /// \param S Statement associated with the current OpenMP region.
-  /// \param Clauses List of clauses for the current OpenMP region.
-  ///
-  /// \returns Statement for finished OpenMP region.
-  StmtResult ActOnOpenMPRegionEnd(StmtResult S, ArrayRef<OMPClause *> Clauses);
-  StmtResult ActOnOpenMPExecutableDirective(
-      OpenMPDirectiveKind Kind, const DeclarationNameInfo &DirName,
-      OpenMPDirectiveKind CancelRegion, ArrayRef<OMPClause *> Clauses,
-      Stmt *AStmt, SourceLocation StartLoc, SourceLocation EndLoc,
-      OpenMPDirectiveKind PrevMappedDirective = llvm::omp::OMPD_unknown);
-  /// Called on well-formed '\#pragma omp parallel' after parsing
-  /// of the  associated statement.
-  StmtResult ActOnOpenMPParallelDirective(ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc);
-  using VarsWithInheritedDSAType =
-      llvm::SmallDenseMap<const ValueDecl *, const Expr *, 4>;
-  /// Called on well-formed '\#pragma omp simd' after parsing
-  /// of the associated statement.
-  StmtResult
-  ActOnOpenMPSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                           SourceLocation StartLoc, SourceLocation EndLoc,
-                           VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '#pragma omp tile' after parsing of its clauses and
-  /// the associated statement.
-  StmtResult ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
-                                      Stmt *AStmt, SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed '#pragma omp unroll' after parsing of its clauses
-  /// and the associated statement.
-  StmtResult ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
-                                        Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp for' after parsing
-  /// of the associated statement.
-  StmtResult
-  ActOnOpenMPForDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                          SourceLocation StartLoc, SourceLocation EndLoc,
-                          VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp for simd' after parsing
-  /// of the associated statement.
-  StmtResult
-  ActOnOpenMPForSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                              SourceLocation StartLoc, SourceLocation EndLoc,
-                              VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp sections' after parsing
-  /// of the associated statement.
-  StmtResult ActOnOpenMPSectionsDirective(ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp section' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPSectionDirective(Stmt *AStmt, SourceLocation StartLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp scope' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPScopeDirective(ArrayRef<OMPClause *> Clauses,
-                                       Stmt *AStmt, SourceLocation StartLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp single' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPSingleDirective(ArrayRef<OMPClause *> Clauses,
-                                        Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp master' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPMasterDirective(Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp critical' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPCriticalDirective(const DeclarationNameInfo &DirName,
-                                          ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp parallel for' after parsing
-  /// of the  associated statement.
-  StmtResult ActOnOpenMPParallelForDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel for simd' after
-  /// parsing of the  associated statement.
-  StmtResult ActOnOpenMPParallelForSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel master' after
-  /// parsing of the  associated statement.
-  StmtResult ActOnOpenMPParallelMasterDirective(ArrayRef<OMPClause *> Clauses,
-                                                Stmt *AStmt,
-                                                SourceLocation StartLoc,
-                                                SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp parallel masked' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPParallelMaskedDirective(ArrayRef<OMPClause *> Clauses,
-                                                Stmt *AStmt,
-                                                SourceLocation StartLoc,
-                                                SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp parallel sections' after
-  /// parsing of the  associated statement.
-  StmtResult ActOnOpenMPParallelSectionsDirective(ArrayRef<OMPClause *> Clauses,
-                                                  Stmt *AStmt,
-                                                  SourceLocation StartLoc,
-                                                  SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp task' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
-                                      Stmt *AStmt, SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp taskyield'.
-  StmtResult ActOnOpenMPTaskyieldDirective(SourceLocation StartLoc,
-                                           SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp error'.
-  /// Error direcitive is allowed in both declared and excutable contexts.
-  /// Adding InExContext to identify which context is called from.
-  StmtResult ActOnOpenMPErrorDirective(ArrayRef<OMPClause *> Clauses,
-                                       SourceLocation StartLoc,
-                                       SourceLocation EndLoc,
-                                       bool InExContext = true);
-  /// Called on well-formed '\#pragma omp barrier'.
-  StmtResult ActOnOpenMPBarrierDirective(SourceLocation StartLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp taskwait'.
-  StmtResult ActOnOpenMPTaskwaitDirective(ArrayRef<OMPClause *> Clauses,
-                                          SourceLocation StartLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp taskgroup'.
-  StmtResult ActOnOpenMPTaskgroupDirective(ArrayRef<OMPClause *> Clauses,
-                                           Stmt *AStmt, SourceLocation StartLoc,
-                                           SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp flush'.
-  StmtResult ActOnOpenMPFlushDirective(ArrayRef<OMPClause *> Clauses,
-                                       SourceLocation StartLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp depobj'.
-  StmtResult ActOnOpenMPDepobjDirective(ArrayRef<OMPClause *> Clauses,
-                                        SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp scan'.
-  StmtResult ActOnOpenMPScanDirective(ArrayRef<OMPClause *> Clauses,
-                                      SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp ordered' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
-                                         Stmt *AStmt, SourceLocation StartLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp atomic' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
-                                        Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp target' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
-                                        Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp target data' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp target enter data' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetEnterDataDirective(ArrayRef<OMPClause *> Clauses,
-                                                 SourceLocation StartLoc,
-                                                 SourceLocation EndLoc,
-                                                 Stmt *AStmt);
-  /// Called on well-formed '\#pragma omp target exit data' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetExitDataDirective(ArrayRef<OMPClause *> Clauses,
-                                                SourceLocation StartLoc,
-                                                SourceLocation EndLoc,
-                                                Stmt *AStmt);
-  /// Called on well-formed '\#pragma omp target parallel' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetParallelDirective(ArrayRef<OMPClause *> Clauses,
-                                                Stmt *AStmt,
-                                                SourceLocation StartLoc,
-                                                SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp target parallel for' after
-  /// parsing of the  associated statement.
-  StmtResult ActOnOpenMPTargetParallelForDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp teams' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPTeamsDirective(ArrayRef<OMPClause *> Clauses,
-                                       Stmt *AStmt, SourceLocation StartLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp teams loop' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPTeamsGenericLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target teams loop' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPTargetTeamsGenericLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel loop' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPParallelGenericLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target parallel loop' after parsing
-  /// of the associated statement.
-  StmtResult ActOnOpenMPTargetParallelGenericLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp cancellation point'.
-  StmtResult
-  ActOnOpenMPCancellationPointDirective(SourceLocation StartLoc,
-                                        SourceLocation EndLoc,
-                                        OpenMPDirectiveKind CancelRegion);
-  /// Called on well-formed '\#pragma omp cancel'.
-  StmtResult ActOnOpenMPCancelDirective(ArrayRef<OMPClause *> Clauses,
-                                        SourceLocation StartLoc,
-                                        SourceLocation EndLoc,
-                                        OpenMPDirectiveKind CancelRegion);
-  /// Called on well-formed '\#pragma omp taskloop' after parsing of the
-  /// associated statement.
-  StmtResult
-  ActOnOpenMPTaskLoopDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                               SourceLocation StartLoc, SourceLocation EndLoc,
-                               VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp taskloop simd' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPTaskLoopSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp master taskloop' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPMasterTaskLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp master taskloop simd' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPMasterTaskLoopSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel master taskloop' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPParallelMasterTaskLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel master taskloop simd' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPParallelMasterTaskLoopSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp masked taskloop' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPMaskedTaskLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp masked taskloop simd' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPMaskedTaskLoopSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel masked taskloop' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPParallelMaskedTaskLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp parallel masked taskloop simd' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp distribute' after parsing
-  /// of the associated statement.
-  StmtResult
-  ActOnOpenMPDistributeDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                                 SourceLocation StartLoc, SourceLocation EndLoc,
-                                 VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target update'.
-  StmtResult ActOnOpenMPTargetUpdateDirective(ArrayRef<OMPClause *> Clauses,
-                                              SourceLocation StartLoc,
-                                              SourceLocation EndLoc,
-                                              Stmt *AStmt);
-  /// Called on well-formed '\#pragma omp distribute parallel for' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPDistributeParallelForDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp distribute parallel for simd'
-  /// after parsing of the associated statement.
-  StmtResult ActOnOpenMPDistributeParallelForSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp distribute simd' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPDistributeSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target parallel for simd' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetParallelForSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target simd' after parsing of
-  /// the associated statement.
-  StmtResult
-  ActOnOpenMPTargetSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                                 SourceLocation StartLoc, SourceLocation EndLoc,
-                                 VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp teams distribute' after parsing of
-  /// the associated statement.
-  StmtResult ActOnOpenMPTeamsDistributeDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp teams distribute simd' after parsing
-  /// of the associated statement.
-  StmtResult ActOnOpenMPTeamsDistributeSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp teams distribute parallel for simd'
-  /// after parsing of the associated statement.
-  StmtResult ActOnOpenMPTeamsDistributeParallelForSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp teams distribute parallel for'
-  /// after parsing of the associated statement.
-  StmtResult ActOnOpenMPTeamsDistributeParallelForDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target teams' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPTargetTeamsDirective(ArrayRef<OMPClause *> Clauses,
-                                             Stmt *AStmt,
-                                             SourceLocation StartLoc,
-                                             SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp target teams distribute' after parsing
-  /// of the associated statement.
-  StmtResult ActOnOpenMPTargetTeamsDistributeDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target teams distribute parallel for'
-  /// after parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetTeamsDistributeParallelForDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target teams distribute parallel for
-  /// simd' after parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetTeamsDistributeParallelForSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp target teams distribute simd' after
-  /// parsing of the associated statement.
-  StmtResult ActOnOpenMPTargetTeamsDistributeSimdDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-  /// Called on well-formed '\#pragma omp interop'.
-  StmtResult ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
-                                         SourceLocation StartLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp dispatch' after parsing of the
-  // /associated statement.
-  StmtResult ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed '\#pragma omp masked' after parsing of the
-  // /associated statement.
-  StmtResult ActOnOpenMPMaskedDirective(ArrayRef<OMPClause *> Clauses,
-                                        Stmt *AStmt, SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-
-  /// Called on well-formed '\#pragma omp loop' after parsing of the
-  /// associated statement.
-  StmtResult ActOnOpenMPGenericLoopDirective(
-      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
-      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
-
-  /// Checks correctness of linear modifiers.
-  bool CheckOpenMPLinearModifier(OpenMPLinearClauseKind LinKind,
-                                 SourceLocation LinLoc);
-  /// Checks that the specified declaration matches requirements for the linear
-  /// decls.
-  bool CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
-                             OpenMPLinearClauseKind LinKind, QualType Type,
-                             bool IsDeclareSimd = false);
-
-  /// Called on well-formed '\#pragma omp declare simd' after parsing of
-  /// the associated method/function.
-  DeclGroupPtrTy ActOnOpenMPDeclareSimdDirective(
-      DeclGroupPtrTy DG, OMPDeclareSimdDeclAttr::BranchStateTy BS,
-      Expr *Simdlen, ArrayRef<Expr *> Uniforms, ArrayRef<Expr *> Aligneds,
-      ArrayRef<Expr *> Alignments, ArrayRef<Expr *> Linears,
-      ArrayRef<unsigned> LinModifiers, ArrayRef<Expr *> Steps, SourceRange SR);
-
-  /// Checks '\#pragma omp declare variant' variant function and original
-  /// functions after parsing of the associated method/function.
-  /// \param DG Function declaration to which declare variant directive is
-  /// applied to.
-  /// \param VariantRef Expression that references the variant function, which
-  /// must be used instead of the original one, specified in \p DG.
-  /// \param TI The trait info object representing the match clause.
-  /// \param NumAppendArgs The number of omp_interop_t arguments to account for
-  /// in checking.
-  /// \returns std::nullopt, if the function/variant function are not compatible
-  /// with the pragma, pair of original function/variant ref expression
-  /// otherwise.
-  std::optional<std::pair<FunctionDecl *, Expr *>>
-  checkOpenMPDeclareVariantFunction(DeclGroupPtrTy DG, Expr *VariantRef,
-                                    OMPTraitInfo &TI, unsigned NumAppendArgs,
-                                    SourceRange SR);
-
-  /// Called on well-formed '\#pragma omp declare variant' after parsing of
-  /// the associated method/function.
-  /// \param FD Function declaration to which declare variant directive is
-  /// applied to.
-  /// \param VariantRef Expression that references the variant function, which
-  /// must be used instead of the original one, specified in \p DG.
-  /// \param TI The context traits associated with the function variant.
-  /// \param AdjustArgsNothing The list of 'nothing' arguments.
-  /// \param AdjustArgsNeedDevicePtr The list of 'need_device_ptr' arguments.
-  /// \param AppendArgs The list of 'append_args' arguments.
-  /// \param AdjustArgsLoc The Location of an 'adjust_args' clause.
-  /// \param AppendArgsLoc The Location of an 'append_args' clause.
-  /// \param SR The SourceRange of the 'declare variant' directive.
-  void ActOnOpenMPDeclareVariantDirective(
-      FunctionDecl *FD, Expr *VariantRef, OMPTraitInfo &TI,
-      ArrayRef<Expr *> AdjustArgsNothing,
-      ArrayRef<Expr *> AdjustArgsNeedDevicePtr,
-      ArrayRef<OMPInteropInfo> AppendArgs, SourceLocation AdjustArgsLoc,
-      SourceLocation AppendArgsLoc, SourceRange SR);
-
-  OMPClause *ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, Expr *Expr,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed 'allocator' clause.
-  OMPClause *ActOnOpenMPAllocatorClause(Expr *Allocator,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'if' clause.
-  OMPClause *ActOnOpenMPIfClause(OpenMPDirectiveKind NameModifier,
-                                 Expr *Condition, SourceLocation StartLoc,
-                                 SourceLocation LParenLoc,
-                                 SourceLocation NameModifierLoc,
-                                 SourceLocation ColonLoc,
-                                 SourceLocation EndLoc);
-  /// Called on well-formed 'final' clause.
-  OMPClause *ActOnOpenMPFinalClause(Expr *Condition, SourceLocation StartLoc,
-                                    SourceLocation LParenLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-formed 'num_threads' clause.
-  OMPClause *ActOnOpenMPNumThreadsClause(Expr *NumThreads,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed 'align' clause.
-  OMPClause *ActOnOpenMPAlignClause(Expr *Alignment, SourceLocation StartLoc,
-                                    SourceLocation LParenLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-formed 'safelen' clause.
-  OMPClause *ActOnOpenMPSafelenClause(Expr *Length, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'simdlen' clause.
-  OMPClause *ActOnOpenMPSimdlenClause(Expr *Length, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-form 'sizes' clause.
-  OMPClause *ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
-                                    SourceLocation StartLoc,
-                                    SourceLocation LParenLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-form 'full' clauses.
-  OMPClause *ActOnOpenMPFullClause(SourceLocation StartLoc,
-                                   SourceLocation EndLoc);
-  /// Called on well-form 'partial' clauses.
-  OMPClause *ActOnOpenMPPartialClause(Expr *FactorExpr, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'collapse' clause.
-  OMPClause *ActOnOpenMPCollapseClause(Expr *NumForLoops,
-                                       SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed 'ordered' clause.
-  OMPClause *
-  ActOnOpenMPOrderedClause(SourceLocation StartLoc, SourceLocation EndLoc,
-                           SourceLocation LParenLoc = SourceLocation(),
-                           Expr *NumForLoops = nullptr);
-  /// Called on well-formed 'grainsize' clause.
-  OMPClause *ActOnOpenMPGrainsizeClause(OpenMPGrainsizeClauseModifier Modifier,
-                                        Expr *Size, SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation ModifierLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'num_tasks' clause.
-  OMPClause *ActOnOpenMPNumTasksClause(OpenMPNumTasksClauseModifier Modifier,
-                                       Expr *NumTasks, SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation ModifierLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed 'hint' clause.
-  OMPClause *ActOnOpenMPHintClause(Expr *Hint, SourceLocation StartLoc,
-                                   SourceLocation LParenLoc,
-                                   SourceLocation EndLoc);
-  /// Called on well-formed 'detach' clause.
-  OMPClause *ActOnOpenMPDetachClause(Expr *Evt, SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-
-  OMPClause *ActOnOpenMPSimpleClause(OpenMPClauseKind Kind, unsigned Argument,
-                                     SourceLocation ArgumentLoc,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'when' clause.
-  OMPClause *ActOnOpenMPWhenClause(OMPTraitInfo &TI, SourceLocation StartLoc,
-                                   SourceLocation LParenLoc,
-                                   SourceLocation EndLoc);
-  /// Called on well-formed 'default' clause.
-  OMPClause *ActOnOpenMPDefaultClause(llvm::omp::DefaultKind Kind,
-                                      SourceLocation KindLoc,
-                                      SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'proc_bind' clause.
-  OMPClause *ActOnOpenMPProcBindClause(llvm::omp::ProcBindKind Kind,
-                                       SourceLocation KindLoc,
-                                       SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed 'order' clause.
-  OMPClause *ActOnOpenMPOrderClause(OpenMPOrderClauseModifier Modifier,
-                                    OpenMPOrderClauseKind Kind,
-                                    SourceLocation StartLoc,
-                                    SourceLocation LParenLoc,
-                                    SourceLocation MLoc, SourceLocation KindLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-formed 'update' clause.
-  OMPClause *ActOnOpenMPUpdateClause(OpenMPDependClauseKind Kind,
-                                     SourceLocation KindLoc,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-
-  OMPClause *ActOnOpenMPSingleExprWithArgClause(
-      OpenMPClauseKind Kind, ArrayRef<unsigned> Arguments, Expr *Expr,
-      SourceLocation StartLoc, SourceLocation LParenLoc,
-      ArrayRef<SourceLocation> ArgumentsLoc, SourceLocation DelimLoc,
-      SourceLocation EndLoc);
-  /// Called on well-formed 'schedule' clause.
-  OMPClause *ActOnOpenMPScheduleClause(
-      OpenMPScheduleClauseModifier M1, OpenMPScheduleClauseModifier M2,
-      OpenMPScheduleClauseKind Kind, Expr *ChunkSize, SourceLocation StartLoc,
-      SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc,
-      SourceLocation KindLoc, SourceLocation CommaLoc, SourceLocation EndLoc);
-
-  OMPClause *ActOnOpenMPClause(OpenMPClauseKind Kind, SourceLocation StartLoc,
-                               SourceLocation EndLoc);
-  /// Called on well-formed 'nowait' clause.
-  OMPClause *ActOnOpenMPNowaitClause(SourceLocation StartLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'untied' clause.
-  OMPClause *ActOnOpenMPUntiedClause(SourceLocation StartLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'mergeable' clause.
-  OMPClause *ActOnOpenMPMergeableClause(SourceLocation StartLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'read' clause.
-  OMPClause *ActOnOpenMPReadClause(SourceLocation StartLoc,
-                                   SourceLocation EndLoc);
-  /// Called on well-formed 'write' clause.
-  OMPClause *ActOnOpenMPWriteClause(SourceLocation StartLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-formed 'update' clause.
-  OMPClause *ActOnOpenMPUpdateClause(SourceLocation StartLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'capture' clause.
-  OMPClause *ActOnOpenMPCaptureClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'compare' clause.
-  OMPClause *ActOnOpenMPCompareClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'fail' clause.
-  OMPClause *ActOnOpenMPFailClause(SourceLocation StartLoc,
-                                   SourceLocation EndLoc);
-  OMPClause *ActOnOpenMPFailClause(OpenMPClauseKind Kind,
-                                   SourceLocation KindLoc,
-                                   SourceLocation StartLoc,
-                                   SourceLocation LParenLoc,
-                                   SourceLocation EndLoc);
-
-  /// Called on well-formed 'seq_cst' clause.
-  OMPClause *ActOnOpenMPSeqCstClause(SourceLocation StartLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'acq_rel' clause.
-  OMPClause *ActOnOpenMPAcqRelClause(SourceLocation StartLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'acquire' clause.
-  OMPClause *ActOnOpenMPAcquireClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'release' clause.
-  OMPClause *ActOnOpenMPReleaseClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'relaxed' clause.
-  OMPClause *ActOnOpenMPRelaxedClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'weak' clause.
-  OMPClause *ActOnOpenMPWeakClause(SourceLocation StartLoc,
-                                   SourceLocation EndLoc);
-
-  /// Called on well-formed 'init' clause.
-  OMPClause *
-  ActOnOpenMPInitClause(Expr *InteropVar, OMPInteropInfo &InteropInfo,
-                        SourceLocation StartLoc, SourceLocation LParenLoc,
-                        SourceLocation VarLoc, SourceLocation EndLoc);
-
-  /// Called on well-formed 'use' clause.
-  OMPClause *ActOnOpenMPUseClause(Expr *InteropVar, SourceLocation StartLoc,
-                                  SourceLocation LParenLoc,
-                                  SourceLocation VarLoc, SourceLocation EndLoc);
-
-  /// Called on well-formed 'destroy' clause.
-  OMPClause *ActOnOpenMPDestroyClause(Expr *InteropVar, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation VarLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'novariants' clause.
-  OMPClause *ActOnOpenMPNovariantsClause(Expr *Condition,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc);
-  /// Called on well-formed 'nocontext' clause.
-  OMPClause *ActOnOpenMPNocontextClause(Expr *Condition,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'filter' clause.
-  OMPClause *ActOnOpenMPFilterClause(Expr *ThreadID, SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'threads' clause.
-  OMPClause *ActOnOpenMPThreadsClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'simd' clause.
-  OMPClause *ActOnOpenMPSIMDClause(SourceLocation StartLoc,
-                                   SourceLocation EndLoc);
-  /// Called on well-formed 'nogroup' clause.
-  OMPClause *ActOnOpenMPNogroupClause(SourceLocation StartLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'unified_address' clause.
-  OMPClause *ActOnOpenMPUnifiedAddressClause(SourceLocation StartLoc,
-                                             SourceLocation EndLoc);
-
-  /// Called on well-formed 'unified_address' clause.
-  OMPClause *ActOnOpenMPUnifiedSharedMemoryClause(SourceLocation StartLoc,
-                                                  SourceLocation EndLoc);
-
-  /// Called on well-formed 'reverse_offload' clause.
-  OMPClause *ActOnOpenMPReverseOffloadClause(SourceLocation StartLoc,
-                                             SourceLocation EndLoc);
-
-  /// Called on well-formed 'dynamic_allocators' clause.
-  OMPClause *ActOnOpenMPDynamicAllocatorsClause(SourceLocation StartLoc,
-                                                SourceLocation EndLoc);
-
-  /// Called on well-formed 'atomic_default_mem_order' clause.
-  OMPClause *ActOnOpenMPAtomicDefaultMemOrderClause(
-      OpenMPAtomicDefaultMemOrderClauseKind Kind, SourceLocation KindLoc,
-      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc);
-
-  /// Called on well-formed 'at' clause.
-  OMPClause *ActOnOpenMPAtClause(OpenMPAtClauseKind Kind,
-                                 SourceLocation KindLoc,
-                                 SourceLocation StartLoc,
-                                 SourceLocation LParenLoc,
-                                 SourceLocation EndLoc);
-
-  /// Called on well-formed 'severity' clause.
-  OMPClause *ActOnOpenMPSeverityClause(OpenMPSeverityClauseKind Kind,
-                                       SourceLocation KindLoc,
-                                       SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc);
-
-  /// Called on well-formed 'message' clause.
-  /// passing string for message.
-  OMPClause *ActOnOpenMPMessageClause(Expr *MS, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-
-  /// Data used for processing a list of variables in OpenMP clauses.
-  struct OpenMPVarListDataTy final {
-    Expr *DepModOrTailExpr = nullptr;
-    Expr *IteratorExpr = nullptr;
-    SourceLocation ColonLoc;
-    SourceLocation RLoc;
-    CXXScopeSpec ReductionOrMapperIdScopeSpec;
-    DeclarationNameInfo ReductionOrMapperId;
-    int ExtraModifier = -1; ///< Additional modifier for linear, map, depend or
-                            ///< lastprivate clause.
-    SmallVector<OpenMPMapModifierKind, NumberOfOMPMapClauseModifiers>
-        MapTypeModifiers;
-    SmallVector<SourceLocation, NumberOfOMPMapClauseModifiers>
-        MapTypeModifiersLoc;
-    SmallVector<OpenMPMotionModifierKind, NumberOfOMPMotionModifiers>
-        MotionModifiers;
-    SmallVector<SourceLocation, NumberOfOMPMotionModifiers> MotionModifiersLoc;
-    bool IsMapTypeImplicit = false;
-    SourceLocation ExtraModifierLoc;
-    SourceLocation OmpAllMemoryLoc;
-    SourceLocation
-        StepModifierLoc; /// 'step' modifier location for linear clause
-  };
-
-  OMPClause *ActOnOpenMPVarListClause(OpenMPClauseKind Kind,
-                                      ArrayRef<Expr *> Vars,
-                                      const OMPVarListLocTy &Locs,
-                                      OpenMPVarListDataTy &Data);
-  /// Called on well-formed 'inclusive' clause.
-  OMPClause *ActOnOpenMPInclusiveClause(ArrayRef<Expr *> VarList,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'exclusive' clause.
-  OMPClause *ActOnOpenMPExclusiveClause(ArrayRef<Expr *> VarList,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc);
-  /// Called on well-formed 'allocate' clause.
-  OMPClause *
-  ActOnOpenMPAllocateClause(Expr *Allocator, ArrayRef<Expr *> VarList,
-                            SourceLocation StartLoc, SourceLocation ColonLoc,
-                            SourceLocation LParenLoc, SourceLocation EndLoc);
-  /// Called on well-formed 'private' clause.
-  OMPClause *ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
-                                      SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'firstprivate' clause.
-  OMPClause *ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc);
-  /// Called on well-formed 'lastprivate' clause.
-  OMPClause *ActOnOpenMPLastprivateClause(
-      ArrayRef<Expr *> VarList, OpenMPLastprivateModifier LPKind,
-      SourceLocation LPKindLoc, SourceLocation ColonLoc,
-      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc);
-  /// Called on well-formed 'shared' clause.
-  OMPClause *ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'reduction' clause.
-  OMPClause *ActOnOpenMPReductionClause(
-      ArrayRef<Expr *> VarList, OpenMPReductionClauseModifier Modifier,
-      SourceLocation StartLoc, SourceLocation LParenLoc,
-      SourceLocation ModifierLoc, SourceLocation ColonLoc,
-      SourceLocation EndLoc, CXXScopeSpec &ReductionIdScopeSpec,
-      const DeclarationNameInfo &ReductionId,
-      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
-  /// Called on well-formed 'task_reduction' clause.
-  OMPClause *ActOnOpenMPTaskReductionClause(
-      ArrayRef<Expr *> VarList, SourceLocation StartLoc,
-      SourceLocation LParenLoc, SourceLocation ColonLoc, SourceLocation EndLoc,
-      CXXScopeSpec &ReductionIdScopeSpec,
-      const DeclarationNameInfo &ReductionId,
-      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
-  /// Called on well-formed 'in_reduction' clause.
-  OMPClause *ActOnOpenMPInReductionClause(
-      ArrayRef<Expr *> VarList, SourceLocation StartLoc,
-      SourceLocation LParenLoc, SourceLocation ColonLoc, SourceLocation EndLoc,
-      CXXScopeSpec &ReductionIdScopeSpec,
-      const DeclarationNameInfo &ReductionId,
-      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
-  /// Called on well-formed 'linear' clause.
-  OMPClause *ActOnOpenMPLinearClause(
-      ArrayRef<Expr *> VarList, Expr *Step, SourceLocation StartLoc,
-      SourceLocation LParenLoc, OpenMPLinearClauseKind LinKind,
-      SourceLocation LinLoc, SourceLocation ColonLoc,
-      SourceLocation StepModifierLoc, SourceLocation EndLoc);
-  /// Called on well-formed 'aligned' clause.
-  OMPClause *ActOnOpenMPAlignedClause(ArrayRef<Expr *> VarList, Expr *Alignment,
-                                      SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation ColonLoc,
-                                      SourceLocation EndLoc);
-  /// Called on well-formed 'copyin' clause.
-  OMPClause *ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'copyprivate' clause.
-  OMPClause *ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed 'flush' pseudo clause.
-  OMPClause *ActOnOpenMPFlushClause(ArrayRef<Expr *> VarList,
-                                    SourceLocation StartLoc,
-                                    SourceLocation LParenLoc,
-                                    SourceLocation EndLoc);
-  /// Called on well-formed 'depobj' pseudo clause.
-  OMPClause *ActOnOpenMPDepobjClause(Expr *Depobj, SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'depend' clause.
-  OMPClause *ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
-                                     Expr *DepModifier,
-                                     ArrayRef<Expr *> VarList,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'device' clause.
-  OMPClause *ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
-                                     Expr *Device, SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation ModifierLoc,
-                                     SourceLocation EndLoc);
-  /// Called on well-formed 'map' clause.
-  OMPClause *ActOnOpenMPMapClause(
-      Expr *IteratorModifier, ArrayRef<OpenMPMapModifierKind> MapTypeModifiers,
-      ArrayRef<SourceLocation> MapTypeModifiersLoc,
-      CXXScopeSpec &MapperIdScopeSpec, DeclarationNameInfo &MapperId,
-      OpenMPMapClauseKind MapType, bool IsMapTypeImplicit,
-      SourceLocation MapLoc, SourceLocation ColonLoc, ArrayRef<Expr *> VarList,
-      const OMPVarListLocTy &Locs, bool NoDiagnose = false,
-      ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
-  /// Called on well-formed 'num_teams' clause.
-  OMPClause *ActOnOpenMPNumTeamsClause(Expr *NumTeams, SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed 'thread_limit' clause.
-  OMPClause *ActOnOpenMPThreadLimitClause(Expr *ThreadLimit,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc);
-  /// Called on well-formed 'priority' clause.
-  OMPClause *ActOnOpenMPPriorityClause(Expr *Priority, SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc);
-  /// Called on well-formed 'dist_schedule' clause.
-  OMPClause *ActOnOpenMPDistScheduleClause(
-      OpenMPDistScheduleClauseKind Kind, Expr *ChunkSize,
-      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation KindLoc,
-      SourceLocation CommaLoc, SourceLocation EndLoc);
-  /// Called on well-formed 'defaultmap' clause.
-  OMPClause *ActOnOpenMPDefaultmapClause(
-      OpenMPDefaultmapClauseModifier M, OpenMPDefaultmapClauseKind Kind,
-      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation MLoc,
-      SourceLocation KindLoc, SourceLocation EndLoc);
-  /// Called on well-formed 'to' clause.
-  OMPClause *
-  ActOnOpenMPToClause(ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
-                      ArrayRef<SourceLocation> MotionModifiersLoc,
-                      CXXScopeSpec &MapperIdScopeSpec,
-                      DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
-                      ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
-                      ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
-  /// Called on well-formed 'from' clause.
-  OMPClause *
-  ActOnOpenMPFromClause(ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
-                        ArrayRef<SourceLocation> MotionModifiersLoc,
-                        CXXScopeSpec &MapperIdScopeSpec,
-                        DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
-                        ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
-                        ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
-  /// Called on well-formed 'use_device_ptr' clause.
-  OMPClause *ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
-                                           const OMPVarListLocTy &Locs);
-  /// Called on well-formed 'use_device_addr' clause.
-  OMPClause *ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
-                                            const OMPVarListLocTy &Locs);
-  /// Called on well-formed 'is_device_ptr' clause.
-  OMPClause *ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
-                                          const OMPVarListLocTy &Locs);
-  /// Called on well-formed 'has_device_addr' clause.
-  OMPClause *ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
-                                            const OMPVarListLocTy &Locs);
-  /// Called on well-formed 'nontemporal' clause.
-  OMPClause *ActOnOpenMPNontemporalClause(ArrayRef<Expr *> VarList,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc);
-
-  /// Data for list of allocators.
-  struct UsesAllocatorsData {
-    /// Allocator.
-    Expr *Allocator = nullptr;
-    /// Allocator traits.
-    Expr *AllocatorTraits = nullptr;
-    /// Locations of '(' and ')' symbols.
-    SourceLocation LParenLoc, RParenLoc;
-  };
-  /// Called on well-formed 'uses_allocators' clause.
-  OMPClause *ActOnOpenMPUsesAllocatorClause(SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc,
-                                            ArrayRef<UsesAllocatorsData> Data);
-  /// Called on well-formed 'affinity' clause.
-  OMPClause *ActOnOpenMPAffinityClause(SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation ColonLoc,
-                                       SourceLocation EndLoc, Expr *Modifier,
-                                       ArrayRef<Expr *> Locators);
-  /// Called on a well-formed 'bind' clause.
-  OMPClause *ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
-                                   SourceLocation KindLoc,
-                                   SourceLocation StartLoc,
-                                   SourceLocation LParenLoc,
-                                   SourceLocation EndLoc);
-
-  /// Called on a well-formed 'ompx_dyn_cgroup_mem' clause.
-  OMPClause *ActOnOpenMPXDynCGroupMemClause(Expr *Size, SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc);
-
-  /// Called on well-formed 'doacross' clause.
-  OMPClause *
-  ActOnOpenMPDoacrossClause(OpenMPDoacrossClauseModifier DepType,
-                            SourceLocation DepLoc, SourceLocation ColonLoc,
-                            ArrayRef<Expr *> VarList, SourceLocation StartLoc,
-                            SourceLocation LParenLoc, SourceLocation EndLoc);
-
-  /// Called on a well-formed 'ompx_attribute' clause.
-  OMPClause *ActOnOpenMPXAttributeClause(ArrayRef<const Attr *> Attrs,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc);
-
-  /// Called on a well-formed 'ompx_bare' clause.
-  OMPClause *ActOnOpenMPXBareClause(SourceLocation StartLoc,
-                                    SourceLocation EndLoc);
-
-private:
-  void *VarDataSharingAttributesStack;
-
-  /// Number of nested '#pragma omp declare target' directives.
-  SmallVector<DeclareTargetContextInfo, 4> DeclareTargetNesting;
-
-  /// Initialization of data-sharing attributes stack.
-  void InitDataSharingAttributesStack();
-  void DestroyDataSharingAttributesStack();
-
-  /// Returns OpenMP nesting level for current directive.
-  unsigned getOpenMPNestingLevel() const;
-
-  /// Adjusts the function scopes index for the target-based regions.
-  void adjustOpenMPTargetScopeIndex(unsigned &FunctionScopesIndex,
-                                    unsigned Level) const;
-
-  /// Returns the number of scopes associated with the construct on the given
-  /// OpenMP level.
-  int getNumberOfConstructScopes(unsigned Level) const;
-
-  /// Push new OpenMP function region for non-capturing function.
-  void pushOpenMPFunctionRegion();
-
-  /// Pop OpenMP function region for non-capturing function.
-  void popOpenMPFunctionRegion(const sema::FunctionScopeInfo *OldFSI);
-
-  /// Analyzes and checks a loop nest for use by a loop transformation.
-  ///
-  /// \param Kind          The loop transformation directive kind.
-  /// \param NumLoops      How many nested loops the directive is expecting.
-  /// \param AStmt         Associated statement of the transformation directive.
-  /// \param LoopHelpers   [out] The loop analysis result.
-  /// \param Body          [out] The body code nested in \p NumLoops loop.
-  /// \param OriginalInits [out] Collection of statements and declarations that
-  ///                      must have been executed/declared before entering the
-  ///                      loop.
-  ///
-  /// \return Whether there was any error.
-  bool checkTransformableLoopNest(
-      OpenMPDirectiveKind Kind, Stmt *AStmt, int NumLoops,
-      SmallVectorImpl<OMPLoopBasedDirective::HelperExprs> &LoopHelpers,
-      Stmt *&Body,
-      SmallVectorImpl<SmallVector<llvm::PointerUnion<Stmt *, Decl *>, 0>>
-          &OriginalInits);
-
-  /// Helper to keep information about the current `omp begin/end declare
-  /// variant` nesting.
-  struct OMPDeclareVariantScope {
-    /// The associated OpenMP context selector.
-    OMPTraitInfo *TI;
-
-    /// The associated OpenMP context selector mangling.
-    std::string NameSuffix;
-
-    OMPDeclareVariantScope(OMPTraitInfo &TI);
-  };
-
-  /// Return the OMPTraitInfo for the surrounding scope, if any.
-  OMPTraitInfo *getOMPTraitInfoForSurroundingScope() {
-    return OMPDeclareVariantScopes.empty() ? nullptr
-                                           : OMPDeclareVariantScopes.back().TI;
-  }
-
-  /// The current `omp begin/end declare variant` scopes.
-  SmallVector<OMPDeclareVariantScope, 4> OMPDeclareVariantScopes;
-
-  /// The current `omp begin/end assumes` scopes.
-  SmallVector<OMPAssumeAttr *, 4> OMPAssumeScoped;
-
-  /// All `omp assumes` we encountered so far.
-  SmallVector<OMPAssumeAttr *, 4> OMPAssumeGlobal;
-
-  /// OMPD_loop is mapped to OMPD_for, OMPD_distribute or OMPD_simd depending
-  /// on the parameter of the bind clause. In the methods for the
-  /// mapped directives, check the parameters of the lastprivate clause.
-  bool checkLastPrivateForMappedDirectives(ArrayRef<OMPClause *> Clauses);
-  /// Depending on the bind clause of OMPD_loop map the directive to new
-  /// directives.
-  ///    1) loop bind(parallel) --> OMPD_for
-  ///    2) loop bind(teams) --> OMPD_distribute
-  ///    3) loop bind(thread) --> OMPD_simd
-  /// This is being handled in Sema instead of Codegen because of the need for
-  /// rigorous semantic checking in the new mapped directives.
-  bool mapLoopConstruct(llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
-                        ArrayRef<OMPClause *> Clauses,
-                        OpenMPBindClauseKind &BindKind,
-                        OpenMPDirectiveKind &Kind,
-                        OpenMPDirectiveKind &PrevMappedDirective,
-                        SourceLocation StartLoc, SourceLocation EndLoc,
-                        const DeclarationNameInfo &DirName,
-                        OpenMPDirectiveKind CancelRegion);
-
-  ///@}
 };
 
 DeductionFailureInfo
diff --git a/clang/include/clang/Sema/SemaOpenMP.h b/clang/include/clang/Sema/SemaOpenMP.h
new file mode 100644
index 0000000000000..9927459bbc594
--- /dev/null
+++ b/clang/include/clang/Sema/SemaOpenMP.h
@@ -0,0 +1,1447 @@
+//===----- SemaOpenMP.h -- Semantic Analysis for OpenMP constructs -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file declares semantic analysis for OpenMP constructs and
+/// clauses.
+///
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_SEMAOPENMP_H
+#define LLVM_CLANG_SEMA_SEMAOPENMP_H
+
+#include "clang/AST/Attr.h"
+#include "clang/AST/Decl.h"
+#include "clang/AST/DeclBase.h"
+#include "clang/AST/DeclOpenMP.h"
+#include "clang/AST/DeclarationName.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/ExprOpenMP.h"
+#include "clang/AST/OpenMPClause.h"
+#include "clang/AST/Stmt.h"
+#include "clang/AST/StmtOpenMP.h"
+#include "clang/AST/Type.h"
+#include "clang/Basic/IdentifierTable.h"
+#include "clang/Basic/LLVM.h"
+#include "clang/Basic/OpenMPKinds.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Basic/Specifiers.h"
+#include "clang/Sema/DeclSpec.h"
+#include "clang/Sema/Ownership.h"
+#include "clang/Sema/Scope.h"
+#include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/SemaBase.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/PointerUnion.h"
+#include <optional>
+#include <string>
+#include <utility>
+
+namespace clang {
+
+class SemaOpenMP : public SemaBase {
+public:
+  SemaOpenMP(Sema &S);
+
+  friend class Parser;
+  friend class Sema;
+
+  using DeclGroupPtrTy = OpaquePtr<DeclGroupRef>;
+  using CapturedParamNameType = std::pair<StringRef, QualType>;
+
+  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
+  /// context is "used as device code".
+  ///
+  /// - If CurContext is a `declare target` function or it is known that the
+  /// function is emitted for the device, emits the diagnostics immediately.
+  /// - If CurContext is a non-`declare target` function and we are compiling
+  ///   for the device, creates a diagnostic which is emitted if and when we
+  ///   realize that the function will be codegen'ed.
+  ///
+  /// Example usage:
+  ///
+  ///  // Variable-length arrays are not allowed in NVPTX device code.
+  ///  if (diagIfOpenMPDeviceCode(Loc, diag::err_vla_unsupported))
+  ///    return ExprError();
+  ///  // Otherwise, continue parsing as normal.
+  SemaDiagnosticBuilder diagIfOpenMPDeviceCode(SourceLocation Loc,
+                                               unsigned DiagID,
+                                               const FunctionDecl *FD);
+
+  /// Creates a SemaDiagnosticBuilder that emits the diagnostic if the current
+  /// context is "used as host code".
+  ///
+  /// - If CurContext is a `declare target` function or it is known that the
+  /// function is emitted for the host, emits the diagnostics immediately.
+  /// - If CurContext is a non-host function, just ignore it.
+  ///
+  /// Example usage:
+  ///
+  ///  // Variable-length arrays are not allowed in NVPTX device code.
+  ///  if (diagIfOpenMPHostode(Loc, diag::err_vla_unsupported))
+  ///    return ExprError();
+  ///  // Otherwise, continue parsing as normal.
+  SemaDiagnosticBuilder diagIfOpenMPHostCode(SourceLocation Loc,
+                                             unsigned DiagID,
+                                             const FunctionDecl *FD);
+
+  /// The declarator \p D defines a function in the scope \p S which is nested
+  /// in an `omp begin/end declare variant` scope. In this method we create a
+  /// declaration for \p D and rename \p D according to the OpenMP context
+  /// selector of the surrounding scope. Return all base functions in \p Bases.
+  void ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
+      Scope *S, Declarator &D, MultiTemplateParamsArg TemplateParameterLists,
+      SmallVectorImpl<FunctionDecl *> &Bases);
+
+  /// Register \p D as specialization of all base functions in \p Bases in the
+  /// current `omp begin/end declare variant` scope.
+  void ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
+      Decl *D, SmallVectorImpl<FunctionDecl *> &Bases);
+
+  /// Act on \p D, a function definition inside of an `omp [begin/end] assumes`.
+  void ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D);
+
+  /// Can we exit an OpenMP declare variant scope at the moment.
+  bool isInOpenMPDeclareVariantScope() const {
+    return !OMPDeclareVariantScopes.empty();
+  }
+
+  ExprResult
+  VerifyPositiveIntegerConstantInClause(Expr *Op, OpenMPClauseKind CKind,
+                                        bool StrictlyPositive = true,
+                                        bool SuppressExprDiags = false);
+
+  /// Given the potential call expression \p Call, determine if there is a
+  /// specialization via the OpenMP declare variant mechanism available. If
+  /// there is, return the specialized call expression, otherwise return the
+  /// original \p Call.
+  ExprResult ActOnOpenMPCall(ExprResult Call, Scope *Scope,
+                             SourceLocation LParenLoc, MultiExprArg ArgExprs,
+                             SourceLocation RParenLoc, Expr *ExecConfig);
+
+  /// Handle a `omp begin declare variant`.
+  void ActOnOpenMPBeginDeclareVariant(SourceLocation Loc, OMPTraitInfo &TI);
+
+  /// Handle a `omp end declare variant`.
+  void ActOnOpenMPEndDeclareVariant();
+
+  /// Function tries to capture lambda's captured variables in the OpenMP region
+  /// before the original lambda is captured.
+  void tryCaptureOpenMPLambdas(ValueDecl *V);
+
+  /// Return true if the provided declaration \a VD should be captured by
+  /// reference.
+  /// \param Level Relative level of nested OpenMP construct for that the check
+  /// is performed.
+  /// \param OpenMPCaptureLevel Capture level within an OpenMP construct.
+  bool isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
+                             unsigned OpenMPCaptureLevel) const;
+
+  /// Check if the specified variable is used in one of the private
+  /// clauses (private, firstprivate, lastprivate, reduction etc.) in OpenMP
+  /// constructs.
+  VarDecl *isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo = false,
+                                unsigned StopAt = 0);
+
+  /// The member expression(this->fd) needs to be rebuilt in the template
+  /// instantiation to generate private copy for OpenMP when default
+  /// clause is used. The function will return true if default
+  /// cluse is used.
+  bool isOpenMPRebuildMemberExpr(ValueDecl *D);
+
+  ExprResult getOpenMPCapturedExpr(VarDecl *Capture, ExprValueKind VK,
+                                   ExprObjectKind OK, SourceLocation Loc);
+
+  /// If the current region is a loop-based region, mark the start of the loop
+  /// construct.
+  void startOpenMPLoop();
+
+  /// If the current region is a range loop-based region, mark the start of the
+  /// loop construct.
+  void startOpenMPCXXRangeFor();
+
+  /// Check if the specified variable is used in 'private' clause.
+  /// \param Level Relative level of nested OpenMP construct for that the check
+  /// is performed.
+  OpenMPClauseKind isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
+                                       unsigned CapLevel) const;
+
+  /// Sets OpenMP capture kind (OMPC_private, OMPC_firstprivate, OMPC_map etc.)
+  /// for \p FD based on DSA for the provided corresponding captured declaration
+  /// \p D.
+  void setOpenMPCaptureKind(FieldDecl *FD, const ValueDecl *D, unsigned Level);
+
+  /// Check if the specified variable is captured  by 'target' directive.
+  /// \param Level Relative level of nested OpenMP construct for that the check
+  /// is performed.
+  bool isOpenMPTargetCapturedDecl(const ValueDecl *D, unsigned Level,
+                                  unsigned CaptureLevel) const;
+
+  /// Check if the specified global variable must be captured  by outer capture
+  /// regions.
+  /// \param Level Relative level of nested OpenMP construct for that
+  /// the check is performed.
+  bool isOpenMPGlobalCapturedDecl(ValueDecl *D, unsigned Level,
+                                  unsigned CaptureLevel) const;
+
+  ExprResult PerformOpenMPImplicitIntegerConversion(SourceLocation OpLoc,
+                                                    Expr *Op);
+  /// Called on start of new data sharing attribute block.
+  void StartOpenMPDSABlock(OpenMPDirectiveKind K,
+                           const DeclarationNameInfo &DirName, Scope *CurScope,
+                           SourceLocation Loc);
+  /// Start analysis of clauses.
+  void StartOpenMPClause(OpenMPClauseKind K);
+  /// End analysis of clauses.
+  void EndOpenMPClause();
+  /// Called on end of data sharing attribute block.
+  void EndOpenMPDSABlock(Stmt *CurDirective);
+
+  /// Check if the current region is an OpenMP loop region and if it is,
+  /// mark loop control variable, used in \p Init for loop initialization, as
+  /// private by default.
+  /// \param Init First part of the for loop.
+  void ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init);
+
+  /// Called on well-formed '\#pragma omp metadirective' after parsing
+  /// of the  associated statement.
+  StmtResult ActOnOpenMPMetaDirective(ArrayRef<OMPClause *> Clauses,
+                                      Stmt *AStmt, SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+
+  // OpenMP directives and clauses.
+  /// Called on correct id-expression from the '#pragma omp
+  /// threadprivate'.
+  ExprResult ActOnOpenMPIdExpression(Scope *CurScope, CXXScopeSpec &ScopeSpec,
+                                     const DeclarationNameInfo &Id,
+                                     OpenMPDirectiveKind Kind);
+  /// Called on well-formed '#pragma omp threadprivate'.
+  DeclGroupPtrTy ActOnOpenMPThreadprivateDirective(SourceLocation Loc,
+                                                   ArrayRef<Expr *> VarList);
+  /// Builds a new OpenMPThreadPrivateDecl and checks its correctness.
+  OMPThreadPrivateDecl *CheckOMPThreadPrivateDecl(SourceLocation Loc,
+                                                  ArrayRef<Expr *> VarList);
+  /// Called on well-formed '#pragma omp allocate'.
+  DeclGroupPtrTy ActOnOpenMPAllocateDirective(SourceLocation Loc,
+                                              ArrayRef<Expr *> VarList,
+                                              ArrayRef<OMPClause *> Clauses,
+                                              DeclContext *Owner = nullptr);
+
+  /// Called on well-formed '#pragma omp [begin] assume[s]'.
+  void ActOnOpenMPAssumesDirective(SourceLocation Loc,
+                                   OpenMPDirectiveKind DKind,
+                                   ArrayRef<std::string> Assumptions,
+                                   bool SkippedClauses);
+
+  /// Check if there is an active global `omp begin assumes` directive.
+  bool isInOpenMPAssumeScope() const { return !OMPAssumeScoped.empty(); }
+
+  /// Check if there is an active global `omp assumes` directive.
+  bool hasGlobalOpenMPAssumes() const { return !OMPAssumeGlobal.empty(); }
+
+  /// Called on well-formed '#pragma omp end assumes'.
+  void ActOnOpenMPEndAssumesDirective();
+
+  /// Called on well-formed '#pragma omp requires'.
+  DeclGroupPtrTy ActOnOpenMPRequiresDirective(SourceLocation Loc,
+                                              ArrayRef<OMPClause *> ClauseList);
+  /// Check restrictions on Requires directive
+  OMPRequiresDecl *CheckOMPRequiresDecl(SourceLocation Loc,
+                                        ArrayRef<OMPClause *> Clauses);
+  /// Check if the specified type is allowed to be used in 'omp declare
+  /// reduction' construct.
+  QualType ActOnOpenMPDeclareReductionType(SourceLocation TyLoc,
+                                           TypeResult ParsedType);
+  /// Called on start of '#pragma omp declare reduction'.
+  DeclGroupPtrTy ActOnOpenMPDeclareReductionDirectiveStart(
+      Scope *S, DeclContext *DC, DeclarationName Name,
+      ArrayRef<std::pair<QualType, SourceLocation>> ReductionTypes,
+      AccessSpecifier AS, Decl *PrevDeclInScope = nullptr);
+  /// Initialize declare reduction construct initializer.
+  void ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D);
+  /// Finish current declare reduction construct initializer.
+  void ActOnOpenMPDeclareReductionCombinerEnd(Decl *D, Expr *Combiner);
+  /// Initialize declare reduction construct initializer.
+  /// \return omp_priv variable.
+  VarDecl *ActOnOpenMPDeclareReductionInitializerStart(Scope *S, Decl *D);
+  /// Finish current declare reduction construct initializer.
+  void ActOnOpenMPDeclareReductionInitializerEnd(Decl *D, Expr *Initializer,
+                                                 VarDecl *OmpPrivParm);
+  /// Called at the end of '#pragma omp declare reduction'.
+  DeclGroupPtrTy ActOnOpenMPDeclareReductionDirectiveEnd(
+      Scope *S, DeclGroupPtrTy DeclReductions, bool IsValid);
+
+  /// Check variable declaration in 'omp declare mapper' construct.
+  TypeResult ActOnOpenMPDeclareMapperVarDecl(Scope *S, Declarator &D);
+  /// Check if the specified type is allowed to be used in 'omp declare
+  /// mapper' construct.
+  QualType ActOnOpenMPDeclareMapperType(SourceLocation TyLoc,
+                                        TypeResult ParsedType);
+  /// Called on start of '#pragma omp declare mapper'.
+  DeclGroupPtrTy ActOnOpenMPDeclareMapperDirective(
+      Scope *S, DeclContext *DC, DeclarationName Name, QualType MapperType,
+      SourceLocation StartLoc, DeclarationName VN, AccessSpecifier AS,
+      Expr *MapperVarRef, ArrayRef<OMPClause *> Clauses,
+      Decl *PrevDeclInScope = nullptr);
+  /// Build the mapper variable of '#pragma omp declare mapper'.
+  ExprResult ActOnOpenMPDeclareMapperDirectiveVarDecl(Scope *S,
+                                                      QualType MapperType,
+                                                      SourceLocation StartLoc,
+                                                      DeclarationName VN);
+  void ActOnOpenMPIteratorVarDecl(VarDecl *VD);
+  bool isOpenMPDeclareMapperVarDeclAllowed(const VarDecl *VD) const;
+  const ValueDecl *getOpenMPDeclareMapperVarName() const;
+
+  struct DeclareTargetContextInfo {
+    struct MapInfo {
+      OMPDeclareTargetDeclAttr::MapTypeTy MT;
+      SourceLocation Loc;
+    };
+    /// Explicitly listed variables and functions in a 'to' or 'link' clause.
+    llvm::DenseMap<NamedDecl *, MapInfo> ExplicitlyMapped;
+
+    /// The 'device_type' as parsed from the clause.
+    OMPDeclareTargetDeclAttr::DevTypeTy DT = OMPDeclareTargetDeclAttr::DT_Any;
+
+    /// The directive kind, `begin declare target` or `declare target`.
+    OpenMPDirectiveKind Kind;
+
+    /// The directive with indirect clause.
+    std::optional<Expr *> Indirect;
+
+    /// The directive location.
+    SourceLocation Loc;
+
+    DeclareTargetContextInfo(OpenMPDirectiveKind Kind, SourceLocation Loc)
+        : Kind(Kind), Loc(Loc) {}
+  };
+
+  /// Called on the start of target region i.e. '#pragma omp declare target'.
+  bool ActOnStartOpenMPDeclareTargetContext(DeclareTargetContextInfo &DTCI);
+
+  /// Called at the end of target region i.e. '#pragma omp end declare target'.
+  const DeclareTargetContextInfo ActOnOpenMPEndDeclareTargetDirective();
+
+  /// Called once a target context is completed, that can be when a
+  /// '#pragma omp end declare target' was encountered or when a
+  /// '#pragma omp declare target' without declaration-definition-seq was
+  /// encountered.
+  void ActOnFinishedOpenMPDeclareTargetContext(DeclareTargetContextInfo &DTCI);
+
+  /// Report unterminated 'omp declare target' or 'omp begin declare target' at
+  /// the end of a compilation unit.
+  void DiagnoseUnterminatedOpenMPDeclareTarget();
+
+  /// Searches for the provided declaration name for OpenMP declare target
+  /// directive.
+  NamedDecl *lookupOpenMPDeclareTargetName(Scope *CurScope,
+                                           CXXScopeSpec &ScopeSpec,
+                                           const DeclarationNameInfo &Id);
+
+  /// Called on correct id-expression from the '#pragma omp declare target'.
+  void ActOnOpenMPDeclareTargetName(NamedDecl *ND, SourceLocation Loc,
+                                    OMPDeclareTargetDeclAttr::MapTypeTy MT,
+                                    DeclareTargetContextInfo &DTCI);
+
+  /// Check declaration inside target region.
+  void
+  checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
+                                   SourceLocation IdLoc = SourceLocation());
+
+  /// Adds OMPDeclareTargetDeclAttr to referenced variables in declare target
+  /// directive.
+  void ActOnOpenMPDeclareTargetInitializer(Decl *D);
+
+  /// Finishes analysis of the deferred functions calls that may be declared as
+  /// host/nohost during device/host compilation.
+  void finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
+                                     const FunctionDecl *Callee,
+                                     SourceLocation Loc);
+
+  /// Return true if currently in OpenMP task with untied clause context.
+  bool isInOpenMPTaskUntiedContext() const;
+
+  /// Return true inside OpenMP declare target region.
+  bool isInOpenMPDeclareTargetContext() const {
+    return !DeclareTargetNesting.empty();
+  }
+  /// Return true inside OpenMP target region.
+  bool isInOpenMPTargetExecutionDirective() const;
+
+  /// Return the number of captured regions created for an OpenMP directive.
+  static int getOpenMPCaptureLevels(OpenMPDirectiveKind Kind);
+
+  /// Initialization of captured region for OpenMP region.
+  void ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope);
+
+  /// Called for syntactical loops (ForStmt or CXXForRangeStmt) associated to
+  /// an OpenMP loop directive.
+  StmtResult ActOnOpenMPCanonicalLoop(Stmt *AStmt);
+
+  /// Process a canonical OpenMP loop nest that can either be a canonical
+  /// literal loop (ForStmt or CXXForRangeStmt), or the generated loop of an
+  /// OpenMP loop transformation construct.
+  StmtResult ActOnOpenMPLoopnest(Stmt *AStmt);
+
+  /// End of OpenMP region.
+  ///
+  /// \param S Statement associated with the current OpenMP region.
+  /// \param Clauses List of clauses for the current OpenMP region.
+  ///
+  /// \returns Statement for finished OpenMP region.
+  StmtResult ActOnOpenMPRegionEnd(StmtResult S, ArrayRef<OMPClause *> Clauses);
+  StmtResult ActOnOpenMPExecutableDirective(
+      OpenMPDirectiveKind Kind, const DeclarationNameInfo &DirName,
+      OpenMPDirectiveKind CancelRegion, ArrayRef<OMPClause *> Clauses,
+      Stmt *AStmt, SourceLocation StartLoc, SourceLocation EndLoc,
+      OpenMPDirectiveKind PrevMappedDirective = llvm::omp::OMPD_unknown);
+  /// Called on well-formed '\#pragma omp parallel' after parsing
+  /// of the  associated statement.
+  StmtResult ActOnOpenMPParallelDirective(ArrayRef<OMPClause *> Clauses,
+                                          Stmt *AStmt, SourceLocation StartLoc,
+                                          SourceLocation EndLoc);
+  using VarsWithInheritedDSAType =
+      llvm::SmallDenseMap<const ValueDecl *, const Expr *, 4>;
+  /// Called on well-formed '\#pragma omp simd' after parsing
+  /// of the associated statement.
+  StmtResult
+  ActOnOpenMPSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                           SourceLocation StartLoc, SourceLocation EndLoc,
+                           VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '#pragma omp tile' after parsing of its clauses and
+  /// the associated statement.
+  StmtResult ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
+                                      Stmt *AStmt, SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed '#pragma omp unroll' after parsing of its clauses
+  /// and the associated statement.
+  StmtResult ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp for' after parsing
+  /// of the associated statement.
+  StmtResult
+  ActOnOpenMPForDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                          SourceLocation StartLoc, SourceLocation EndLoc,
+                          VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp for simd' after parsing
+  /// of the associated statement.
+  StmtResult
+  ActOnOpenMPForSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                              SourceLocation StartLoc, SourceLocation EndLoc,
+                              VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp sections' after parsing
+  /// of the associated statement.
+  StmtResult ActOnOpenMPSectionsDirective(ArrayRef<OMPClause *> Clauses,
+                                          Stmt *AStmt, SourceLocation StartLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp section' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPSectionDirective(Stmt *AStmt, SourceLocation StartLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp scope' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPScopeDirective(ArrayRef<OMPClause *> Clauses,
+                                       Stmt *AStmt, SourceLocation StartLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp single' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPSingleDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp master' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPMasterDirective(Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp critical' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPCriticalDirective(const DeclarationNameInfo &DirName,
+                                          ArrayRef<OMPClause *> Clauses,
+                                          Stmt *AStmt, SourceLocation StartLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp parallel for' after parsing
+  /// of the  associated statement.
+  StmtResult ActOnOpenMPParallelForDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel for simd' after
+  /// parsing of the  associated statement.
+  StmtResult ActOnOpenMPParallelForSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel master' after
+  /// parsing of the  associated statement.
+  StmtResult ActOnOpenMPParallelMasterDirective(ArrayRef<OMPClause *> Clauses,
+                                                Stmt *AStmt,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp parallel masked' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPParallelMaskedDirective(ArrayRef<OMPClause *> Clauses,
+                                                Stmt *AStmt,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp parallel sections' after
+  /// parsing of the  associated statement.
+  StmtResult ActOnOpenMPParallelSectionsDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp task' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
+                                      Stmt *AStmt, SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp taskyield'.
+  StmtResult ActOnOpenMPTaskyieldDirective(SourceLocation StartLoc,
+                                           SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp error'.
+  /// Error direcitive is allowed in both declared and excutable contexts.
+  /// Adding InExContext to identify which context is called from.
+  StmtResult ActOnOpenMPErrorDirective(ArrayRef<OMPClause *> Clauses,
+                                       SourceLocation StartLoc,
+                                       SourceLocation EndLoc,
+                                       bool InExContext = true);
+  /// Called on well-formed '\#pragma omp barrier'.
+  StmtResult ActOnOpenMPBarrierDirective(SourceLocation StartLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp taskwait'.
+  StmtResult ActOnOpenMPTaskwaitDirective(ArrayRef<OMPClause *> Clauses,
+                                          SourceLocation StartLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp taskgroup'.
+  StmtResult ActOnOpenMPTaskgroupDirective(ArrayRef<OMPClause *> Clauses,
+                                           Stmt *AStmt, SourceLocation StartLoc,
+                                           SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp flush'.
+  StmtResult ActOnOpenMPFlushDirective(ArrayRef<OMPClause *> Clauses,
+                                       SourceLocation StartLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp depobj'.
+  StmtResult ActOnOpenMPDepobjDirective(ArrayRef<OMPClause *> Clauses,
+                                        SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp scan'.
+  StmtResult ActOnOpenMPScanDirective(ArrayRef<OMPClause *> Clauses,
+                                      SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp ordered' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
+                                         Stmt *AStmt, SourceLocation StartLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp atomic' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp target' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp target data' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
+                                            Stmt *AStmt,
+                                            SourceLocation StartLoc,
+                                            SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp target enter data' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetEnterDataDirective(ArrayRef<OMPClause *> Clauses,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation EndLoc,
+                                                 Stmt *AStmt);
+  /// Called on well-formed '\#pragma omp target exit data' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetExitDataDirective(ArrayRef<OMPClause *> Clauses,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc,
+                                                Stmt *AStmt);
+  /// Called on well-formed '\#pragma omp target parallel' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetParallelDirective(ArrayRef<OMPClause *> Clauses,
+                                                Stmt *AStmt,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp target parallel for' after
+  /// parsing of the  associated statement.
+  StmtResult ActOnOpenMPTargetParallelForDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp teams' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPTeamsDirective(ArrayRef<OMPClause *> Clauses,
+                                       Stmt *AStmt, SourceLocation StartLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp teams loop' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPTeamsGenericLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target teams loop' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPTargetTeamsGenericLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel loop' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPParallelGenericLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target parallel loop' after parsing
+  /// of the associated statement.
+  StmtResult ActOnOpenMPTargetParallelGenericLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp cancellation point'.
+  StmtResult
+  ActOnOpenMPCancellationPointDirective(SourceLocation StartLoc,
+                                        SourceLocation EndLoc,
+                                        OpenMPDirectiveKind CancelRegion);
+  /// Called on well-formed '\#pragma omp cancel'.
+  StmtResult ActOnOpenMPCancelDirective(ArrayRef<OMPClause *> Clauses,
+                                        SourceLocation StartLoc,
+                                        SourceLocation EndLoc,
+                                        OpenMPDirectiveKind CancelRegion);
+  /// Called on well-formed '\#pragma omp taskloop' after parsing of the
+  /// associated statement.
+  StmtResult
+  ActOnOpenMPTaskLoopDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                               SourceLocation StartLoc, SourceLocation EndLoc,
+                               VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp taskloop simd' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPTaskLoopSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp master taskloop' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPMasterTaskLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp master taskloop simd' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPMasterTaskLoopSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel master taskloop' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPParallelMasterTaskLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel master taskloop simd' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPParallelMasterTaskLoopSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp masked taskloop' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPMaskedTaskLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp masked taskloop simd' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPMaskedTaskLoopSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel masked taskloop' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPParallelMaskedTaskLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp parallel masked taskloop simd' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp distribute' after parsing
+  /// of the associated statement.
+  StmtResult
+  ActOnOpenMPDistributeDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                                 SourceLocation StartLoc, SourceLocation EndLoc,
+                                 VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target update'.
+  StmtResult ActOnOpenMPTargetUpdateDirective(ArrayRef<OMPClause *> Clauses,
+                                              SourceLocation StartLoc,
+                                              SourceLocation EndLoc,
+                                              Stmt *AStmt);
+  /// Called on well-formed '\#pragma omp distribute parallel for' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPDistributeParallelForDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp distribute parallel for simd'
+  /// after parsing of the associated statement.
+  StmtResult ActOnOpenMPDistributeParallelForSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp distribute simd' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPDistributeSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target parallel for simd' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetParallelForSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target simd' after parsing of
+  /// the associated statement.
+  StmtResult
+  ActOnOpenMPTargetSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
+                                 SourceLocation StartLoc, SourceLocation EndLoc,
+                                 VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp teams distribute' after parsing of
+  /// the associated statement.
+  StmtResult ActOnOpenMPTeamsDistributeDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp teams distribute simd' after parsing
+  /// of the associated statement.
+  StmtResult ActOnOpenMPTeamsDistributeSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp teams distribute parallel for simd'
+  /// after parsing of the associated statement.
+  StmtResult ActOnOpenMPTeamsDistributeParallelForSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp teams distribute parallel for'
+  /// after parsing of the associated statement.
+  StmtResult ActOnOpenMPTeamsDistributeParallelForDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target teams' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPTargetTeamsDirective(ArrayRef<OMPClause *> Clauses,
+                                             Stmt *AStmt,
+                                             SourceLocation StartLoc,
+                                             SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp target teams distribute' after parsing
+  /// of the associated statement.
+  StmtResult ActOnOpenMPTargetTeamsDistributeDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target teams distribute parallel for'
+  /// after parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetTeamsDistributeParallelForDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target teams distribute parallel for
+  /// simd' after parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetTeamsDistributeParallelForSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp target teams distribute simd' after
+  /// parsing of the associated statement.
+  StmtResult ActOnOpenMPTargetTeamsDistributeSimdDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+  /// Called on well-formed '\#pragma omp interop'.
+  StmtResult ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
+                                         SourceLocation StartLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp dispatch' after parsing of the
+  // /associated statement.
+  StmtResult ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
+                                          Stmt *AStmt, SourceLocation StartLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed '\#pragma omp masked' after parsing of the
+  // /associated statement.
+  StmtResult ActOnOpenMPMaskedDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+
+  /// Called on well-formed '\#pragma omp loop' after parsing of the
+  /// associated statement.
+  StmtResult ActOnOpenMPGenericLoopDirective(
+      ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+      SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA);
+
+  /// Checks correctness of linear modifiers.
+  bool CheckOpenMPLinearModifier(OpenMPLinearClauseKind LinKind,
+                                 SourceLocation LinLoc);
+  /// Checks that the specified declaration matches requirements for the linear
+  /// decls.
+  bool CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
+                             OpenMPLinearClauseKind LinKind, QualType Type,
+                             bool IsDeclareSimd = false);
+
+  /// Called on well-formed '\#pragma omp declare simd' after parsing of
+  /// the associated method/function.
+  DeclGroupPtrTy ActOnOpenMPDeclareSimdDirective(
+      DeclGroupPtrTy DG, OMPDeclareSimdDeclAttr::BranchStateTy BS,
+      Expr *Simdlen, ArrayRef<Expr *> Uniforms, ArrayRef<Expr *> Aligneds,
+      ArrayRef<Expr *> Alignments, ArrayRef<Expr *> Linears,
+      ArrayRef<unsigned> LinModifiers, ArrayRef<Expr *> Steps, SourceRange SR);
+
+  /// Checks '\#pragma omp declare variant' variant function and original
+  /// functions after parsing of the associated method/function.
+  /// \param DG Function declaration to which declare variant directive is
+  /// applied to.
+  /// \param VariantRef Expression that references the variant function, which
+  /// must be used instead of the original one, specified in \p DG.
+  /// \param TI The trait info object representing the match clause.
+  /// \param NumAppendArgs The number of omp_interop_t arguments to account for
+  /// in checking.
+  /// \returns std::nullopt, if the function/variant function are not compatible
+  /// with the pragma, pair of original function/variant ref expression
+  /// otherwise.
+  std::optional<std::pair<FunctionDecl *, Expr *>>
+  checkOpenMPDeclareVariantFunction(DeclGroupPtrTy DG, Expr *VariantRef,
+                                    OMPTraitInfo &TI, unsigned NumAppendArgs,
+                                    SourceRange SR);
+
+  /// Called on well-formed '\#pragma omp declare variant' after parsing of
+  /// the associated method/function.
+  /// \param FD Function declaration to which declare variant directive is
+  /// applied to.
+  /// \param VariantRef Expression that references the variant function, which
+  /// must be used instead of the original one, specified in \p DG.
+  /// \param TI The context traits associated with the function variant.
+  /// \param AdjustArgsNothing The list of 'nothing' arguments.
+  /// \param AdjustArgsNeedDevicePtr The list of 'need_device_ptr' arguments.
+  /// \param AppendArgs The list of 'append_args' arguments.
+  /// \param AdjustArgsLoc The Location of an 'adjust_args' clause.
+  /// \param AppendArgsLoc The Location of an 'append_args' clause.
+  /// \param SR The SourceRange of the 'declare variant' directive.
+  void ActOnOpenMPDeclareVariantDirective(
+      FunctionDecl *FD, Expr *VariantRef, OMPTraitInfo &TI,
+      ArrayRef<Expr *> AdjustArgsNothing,
+      ArrayRef<Expr *> AdjustArgsNeedDevicePtr,
+      ArrayRef<OMPInteropInfo> AppendArgs, SourceLocation AdjustArgsLoc,
+      SourceLocation AppendArgsLoc, SourceRange SR);
+
+  OMPClause *ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, Expr *Expr,
+                                         SourceLocation StartLoc,
+                                         SourceLocation LParenLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed 'allocator' clause.
+  OMPClause *ActOnOpenMPAllocatorClause(Expr *Allocator,
+                                        SourceLocation StartLoc,
+                                        SourceLocation LParenLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'if' clause.
+  OMPClause *ActOnOpenMPIfClause(OpenMPDirectiveKind NameModifier,
+                                 Expr *Condition, SourceLocation StartLoc,
+                                 SourceLocation LParenLoc,
+                                 SourceLocation NameModifierLoc,
+                                 SourceLocation ColonLoc,
+                                 SourceLocation EndLoc);
+  /// Called on well-formed 'final' clause.
+  OMPClause *ActOnOpenMPFinalClause(Expr *Condition, SourceLocation StartLoc,
+                                    SourceLocation LParenLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-formed 'num_threads' clause.
+  OMPClause *ActOnOpenMPNumThreadsClause(Expr *NumThreads,
+                                         SourceLocation StartLoc,
+                                         SourceLocation LParenLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed 'align' clause.
+  OMPClause *ActOnOpenMPAlignClause(Expr *Alignment, SourceLocation StartLoc,
+                                    SourceLocation LParenLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-formed 'safelen' clause.
+  OMPClause *ActOnOpenMPSafelenClause(Expr *Length, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'simdlen' clause.
+  OMPClause *ActOnOpenMPSimdlenClause(Expr *Length, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-form 'sizes' clause.
+  OMPClause *ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
+                                    SourceLocation StartLoc,
+                                    SourceLocation LParenLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-form 'full' clauses.
+  OMPClause *ActOnOpenMPFullClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-form 'partial' clauses.
+  OMPClause *ActOnOpenMPPartialClause(Expr *FactorExpr, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'collapse' clause.
+  OMPClause *ActOnOpenMPCollapseClause(Expr *NumForLoops,
+                                       SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed 'ordered' clause.
+  OMPClause *
+  ActOnOpenMPOrderedClause(SourceLocation StartLoc, SourceLocation EndLoc,
+                           SourceLocation LParenLoc = SourceLocation(),
+                           Expr *NumForLoops = nullptr);
+  /// Called on well-formed 'grainsize' clause.
+  OMPClause *ActOnOpenMPGrainsizeClause(OpenMPGrainsizeClauseModifier Modifier,
+                                        Expr *Size, SourceLocation StartLoc,
+                                        SourceLocation LParenLoc,
+                                        SourceLocation ModifierLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'num_tasks' clause.
+  OMPClause *ActOnOpenMPNumTasksClause(OpenMPNumTasksClauseModifier Modifier,
+                                       Expr *NumTasks, SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation ModifierLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed 'hint' clause.
+  OMPClause *ActOnOpenMPHintClause(Expr *Hint, SourceLocation StartLoc,
+                                   SourceLocation LParenLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-formed 'detach' clause.
+  OMPClause *ActOnOpenMPDetachClause(Expr *Evt, SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+
+  OMPClause *ActOnOpenMPSimpleClause(OpenMPClauseKind Kind, unsigned Argument,
+                                     SourceLocation ArgumentLoc,
+                                     SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'when' clause.
+  OMPClause *ActOnOpenMPWhenClause(OMPTraitInfo &TI, SourceLocation StartLoc,
+                                   SourceLocation LParenLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-formed 'default' clause.
+  OMPClause *ActOnOpenMPDefaultClause(llvm::omp::DefaultKind Kind,
+                                      SourceLocation KindLoc,
+                                      SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'proc_bind' clause.
+  OMPClause *ActOnOpenMPProcBindClause(llvm::omp::ProcBindKind Kind,
+                                       SourceLocation KindLoc,
+                                       SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed 'order' clause.
+  OMPClause *ActOnOpenMPOrderClause(OpenMPOrderClauseModifier Modifier,
+                                    OpenMPOrderClauseKind Kind,
+                                    SourceLocation StartLoc,
+                                    SourceLocation LParenLoc,
+                                    SourceLocation MLoc, SourceLocation KindLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-formed 'update' clause.
+  OMPClause *ActOnOpenMPUpdateClause(OpenMPDependClauseKind Kind,
+                                     SourceLocation KindLoc,
+                                     SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+
+  OMPClause *ActOnOpenMPSingleExprWithArgClause(
+      OpenMPClauseKind Kind, ArrayRef<unsigned> Arguments, Expr *Expr,
+      SourceLocation StartLoc, SourceLocation LParenLoc,
+      ArrayRef<SourceLocation> ArgumentsLoc, SourceLocation DelimLoc,
+      SourceLocation EndLoc);
+  /// Called on well-formed 'schedule' clause.
+  OMPClause *ActOnOpenMPScheduleClause(
+      OpenMPScheduleClauseModifier M1, OpenMPScheduleClauseModifier M2,
+      OpenMPScheduleClauseKind Kind, Expr *ChunkSize, SourceLocation StartLoc,
+      SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc,
+      SourceLocation KindLoc, SourceLocation CommaLoc, SourceLocation EndLoc);
+
+  OMPClause *ActOnOpenMPClause(OpenMPClauseKind Kind, SourceLocation StartLoc,
+                               SourceLocation EndLoc);
+  /// Called on well-formed 'nowait' clause.
+  OMPClause *ActOnOpenMPNowaitClause(SourceLocation StartLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'untied' clause.
+  OMPClause *ActOnOpenMPUntiedClause(SourceLocation StartLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'mergeable' clause.
+  OMPClause *ActOnOpenMPMergeableClause(SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'read' clause.
+  OMPClause *ActOnOpenMPReadClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-formed 'write' clause.
+  OMPClause *ActOnOpenMPWriteClause(SourceLocation StartLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-formed 'update' clause.
+  OMPClause *ActOnOpenMPUpdateClause(SourceLocation StartLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'capture' clause.
+  OMPClause *ActOnOpenMPCaptureClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'compare' clause.
+  OMPClause *ActOnOpenMPCompareClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'fail' clause.
+  OMPClause *ActOnOpenMPFailClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+  OMPClause *ActOnOpenMPFailClause(OpenMPClauseKind Kind,
+                                   SourceLocation KindLoc,
+                                   SourceLocation StartLoc,
+                                   SourceLocation LParenLoc,
+                                   SourceLocation EndLoc);
+
+  /// Called on well-formed 'seq_cst' clause.
+  OMPClause *ActOnOpenMPSeqCstClause(SourceLocation StartLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'acq_rel' clause.
+  OMPClause *ActOnOpenMPAcqRelClause(SourceLocation StartLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'acquire' clause.
+  OMPClause *ActOnOpenMPAcquireClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'release' clause.
+  OMPClause *ActOnOpenMPReleaseClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'relaxed' clause.
+  OMPClause *ActOnOpenMPRelaxedClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'weak' clause.
+  OMPClause *ActOnOpenMPWeakClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+
+  /// Called on well-formed 'init' clause.
+  OMPClause *
+  ActOnOpenMPInitClause(Expr *InteropVar, OMPInteropInfo &InteropInfo,
+                        SourceLocation StartLoc, SourceLocation LParenLoc,
+                        SourceLocation VarLoc, SourceLocation EndLoc);
+
+  /// Called on well-formed 'use' clause.
+  OMPClause *ActOnOpenMPUseClause(Expr *InteropVar, SourceLocation StartLoc,
+                                  SourceLocation LParenLoc,
+                                  SourceLocation VarLoc, SourceLocation EndLoc);
+
+  /// Called on well-formed 'destroy' clause.
+  OMPClause *ActOnOpenMPDestroyClause(Expr *InteropVar, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation VarLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'novariants' clause.
+  OMPClause *ActOnOpenMPNovariantsClause(Expr *Condition,
+                                         SourceLocation StartLoc,
+                                         SourceLocation LParenLoc,
+                                         SourceLocation EndLoc);
+  /// Called on well-formed 'nocontext' clause.
+  OMPClause *ActOnOpenMPNocontextClause(Expr *Condition,
+                                        SourceLocation StartLoc,
+                                        SourceLocation LParenLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'filter' clause.
+  OMPClause *ActOnOpenMPFilterClause(Expr *ThreadID, SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'threads' clause.
+  OMPClause *ActOnOpenMPThreadsClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'simd' clause.
+  OMPClause *ActOnOpenMPSIMDClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-formed 'nogroup' clause.
+  OMPClause *ActOnOpenMPNogroupClause(SourceLocation StartLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'unified_address' clause.
+  OMPClause *ActOnOpenMPUnifiedAddressClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc);
+
+  /// Called on well-formed 'unified_address' clause.
+  OMPClause *ActOnOpenMPUnifiedSharedMemoryClause(SourceLocation StartLoc,
+                                                  SourceLocation EndLoc);
+
+  /// Called on well-formed 'reverse_offload' clause.
+  OMPClause *ActOnOpenMPReverseOffloadClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc);
+
+  /// Called on well-formed 'dynamic_allocators' clause.
+  OMPClause *ActOnOpenMPDynamicAllocatorsClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc);
+
+  /// Called on well-formed 'atomic_default_mem_order' clause.
+  OMPClause *ActOnOpenMPAtomicDefaultMemOrderClause(
+      OpenMPAtomicDefaultMemOrderClauseKind Kind, SourceLocation KindLoc,
+      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc);
+
+  /// Called on well-formed 'at' clause.
+  OMPClause *ActOnOpenMPAtClause(OpenMPAtClauseKind Kind,
+                                 SourceLocation KindLoc,
+                                 SourceLocation StartLoc,
+                                 SourceLocation LParenLoc,
+                                 SourceLocation EndLoc);
+
+  /// Called on well-formed 'severity' clause.
+  OMPClause *ActOnOpenMPSeverityClause(OpenMPSeverityClauseKind Kind,
+                                       SourceLocation KindLoc,
+                                       SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation EndLoc);
+
+  /// Called on well-formed 'message' clause.
+  /// passing string for message.
+  OMPClause *ActOnOpenMPMessageClause(Expr *MS, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+
+  /// Data used for processing a list of variables in OpenMP clauses.
+  struct OpenMPVarListDataTy final {
+    Expr *DepModOrTailExpr = nullptr;
+    Expr *IteratorExpr = nullptr;
+    SourceLocation ColonLoc;
+    SourceLocation RLoc;
+    CXXScopeSpec ReductionOrMapperIdScopeSpec;
+    DeclarationNameInfo ReductionOrMapperId;
+    int ExtraModifier = -1; ///< Additional modifier for linear, map, depend or
+                            ///< lastprivate clause.
+    SmallVector<OpenMPMapModifierKind, NumberOfOMPMapClauseModifiers>
+        MapTypeModifiers;
+    SmallVector<SourceLocation, NumberOfOMPMapClauseModifiers>
+        MapTypeModifiersLoc;
+    SmallVector<OpenMPMotionModifierKind, NumberOfOMPMotionModifiers>
+        MotionModifiers;
+    SmallVector<SourceLocation, NumberOfOMPMotionModifiers> MotionModifiersLoc;
+    bool IsMapTypeImplicit = false;
+    SourceLocation ExtraModifierLoc;
+    SourceLocation OmpAllMemoryLoc;
+    SourceLocation
+        StepModifierLoc; /// 'step' modifier location for linear clause
+  };
+
+  OMPClause *ActOnOpenMPVarListClause(OpenMPClauseKind Kind,
+                                      ArrayRef<Expr *> Vars,
+                                      const OMPVarListLocTy &Locs,
+                                      OpenMPVarListDataTy &Data);
+  /// Called on well-formed 'inclusive' clause.
+  OMPClause *ActOnOpenMPInclusiveClause(ArrayRef<Expr *> VarList,
+                                        SourceLocation StartLoc,
+                                        SourceLocation LParenLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'exclusive' clause.
+  OMPClause *ActOnOpenMPExclusiveClause(ArrayRef<Expr *> VarList,
+                                        SourceLocation StartLoc,
+                                        SourceLocation LParenLoc,
+                                        SourceLocation EndLoc);
+  /// Called on well-formed 'allocate' clause.
+  OMPClause *
+  ActOnOpenMPAllocateClause(Expr *Allocator, ArrayRef<Expr *> VarList,
+                            SourceLocation StartLoc, SourceLocation ColonLoc,
+                            SourceLocation LParenLoc, SourceLocation EndLoc);
+  /// Called on well-formed 'private' clause.
+  OMPClause *ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
+                                      SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'firstprivate' clause.
+  OMPClause *ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
+                                           SourceLocation StartLoc,
+                                           SourceLocation LParenLoc,
+                                           SourceLocation EndLoc);
+  /// Called on well-formed 'lastprivate' clause.
+  OMPClause *ActOnOpenMPLastprivateClause(
+      ArrayRef<Expr *> VarList, OpenMPLastprivateModifier LPKind,
+      SourceLocation LPKindLoc, SourceLocation ColonLoc,
+      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc);
+  /// Called on well-formed 'shared' clause.
+  OMPClause *ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
+                                     SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'reduction' clause.
+  OMPClause *ActOnOpenMPReductionClause(
+      ArrayRef<Expr *> VarList, OpenMPReductionClauseModifier Modifier,
+      SourceLocation StartLoc, SourceLocation LParenLoc,
+      SourceLocation ModifierLoc, SourceLocation ColonLoc,
+      SourceLocation EndLoc, CXXScopeSpec &ReductionIdScopeSpec,
+      const DeclarationNameInfo &ReductionId,
+      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
+  /// Called on well-formed 'task_reduction' clause.
+  OMPClause *ActOnOpenMPTaskReductionClause(
+      ArrayRef<Expr *> VarList, SourceLocation StartLoc,
+      SourceLocation LParenLoc, SourceLocation ColonLoc, SourceLocation EndLoc,
+      CXXScopeSpec &ReductionIdScopeSpec,
+      const DeclarationNameInfo &ReductionId,
+      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
+  /// Called on well-formed 'in_reduction' clause.
+  OMPClause *ActOnOpenMPInReductionClause(
+      ArrayRef<Expr *> VarList, SourceLocation StartLoc,
+      SourceLocation LParenLoc, SourceLocation ColonLoc, SourceLocation EndLoc,
+      CXXScopeSpec &ReductionIdScopeSpec,
+      const DeclarationNameInfo &ReductionId,
+      ArrayRef<Expr *> UnresolvedReductions = std::nullopt);
+  /// Called on well-formed 'linear' clause.
+  OMPClause *ActOnOpenMPLinearClause(
+      ArrayRef<Expr *> VarList, Expr *Step, SourceLocation StartLoc,
+      SourceLocation LParenLoc, OpenMPLinearClauseKind LinKind,
+      SourceLocation LinLoc, SourceLocation ColonLoc,
+      SourceLocation StepModifierLoc, SourceLocation EndLoc);
+  /// Called on well-formed 'aligned' clause.
+  OMPClause *ActOnOpenMPAlignedClause(ArrayRef<Expr *> VarList, Expr *Alignment,
+                                      SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation ColonLoc,
+                                      SourceLocation EndLoc);
+  /// Called on well-formed 'copyin' clause.
+  OMPClause *ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
+                                     SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'copyprivate' clause.
+  OMPClause *ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
+                                          SourceLocation StartLoc,
+                                          SourceLocation LParenLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed 'flush' pseudo clause.
+  OMPClause *ActOnOpenMPFlushClause(ArrayRef<Expr *> VarList,
+                                    SourceLocation StartLoc,
+                                    SourceLocation LParenLoc,
+                                    SourceLocation EndLoc);
+  /// Called on well-formed 'depobj' pseudo clause.
+  OMPClause *ActOnOpenMPDepobjClause(Expr *Depobj, SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'depend' clause.
+  OMPClause *ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
+                                     Expr *DepModifier,
+                                     ArrayRef<Expr *> VarList,
+                                     SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'device' clause.
+  OMPClause *ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
+                                     Expr *Device, SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation ModifierLoc,
+                                     SourceLocation EndLoc);
+  /// Called on well-formed 'map' clause.
+  OMPClause *ActOnOpenMPMapClause(
+      Expr *IteratorModifier, ArrayRef<OpenMPMapModifierKind> MapTypeModifiers,
+      ArrayRef<SourceLocation> MapTypeModifiersLoc,
+      CXXScopeSpec &MapperIdScopeSpec, DeclarationNameInfo &MapperId,
+      OpenMPMapClauseKind MapType, bool IsMapTypeImplicit,
+      SourceLocation MapLoc, SourceLocation ColonLoc, ArrayRef<Expr *> VarList,
+      const OMPVarListLocTy &Locs, bool NoDiagnose = false,
+      ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
+  /// Called on well-formed 'num_teams' clause.
+  OMPClause *ActOnOpenMPNumTeamsClause(Expr *NumTeams, SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed 'thread_limit' clause.
+  OMPClause *ActOnOpenMPThreadLimitClause(Expr *ThreadLimit,
+                                          SourceLocation StartLoc,
+                                          SourceLocation LParenLoc,
+                                          SourceLocation EndLoc);
+  /// Called on well-formed 'priority' clause.
+  OMPClause *ActOnOpenMPPriorityClause(Expr *Priority, SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation EndLoc);
+  /// Called on well-formed 'dist_schedule' clause.
+  OMPClause *ActOnOpenMPDistScheduleClause(
+      OpenMPDistScheduleClauseKind Kind, Expr *ChunkSize,
+      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation KindLoc,
+      SourceLocation CommaLoc, SourceLocation EndLoc);
+  /// Called on well-formed 'defaultmap' clause.
+  OMPClause *ActOnOpenMPDefaultmapClause(
+      OpenMPDefaultmapClauseModifier M, OpenMPDefaultmapClauseKind Kind,
+      SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation MLoc,
+      SourceLocation KindLoc, SourceLocation EndLoc);
+  /// Called on well-formed 'to' clause.
+  OMPClause *
+  ActOnOpenMPToClause(ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
+                      ArrayRef<SourceLocation> MotionModifiersLoc,
+                      CXXScopeSpec &MapperIdScopeSpec,
+                      DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
+                      ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
+                      ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
+  /// Called on well-formed 'from' clause.
+  OMPClause *
+  ActOnOpenMPFromClause(ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
+                        ArrayRef<SourceLocation> MotionModifiersLoc,
+                        CXXScopeSpec &MapperIdScopeSpec,
+                        DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
+                        ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
+                        ArrayRef<Expr *> UnresolvedMappers = std::nullopt);
+  /// Called on well-formed 'use_device_ptr' clause.
+  OMPClause *ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
+                                           const OMPVarListLocTy &Locs);
+  /// Called on well-formed 'use_device_addr' clause.
+  OMPClause *ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
+                                            const OMPVarListLocTy &Locs);
+  /// Called on well-formed 'is_device_ptr' clause.
+  OMPClause *ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
+                                          const OMPVarListLocTy &Locs);
+  /// Called on well-formed 'has_device_addr' clause.
+  OMPClause *ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
+                                            const OMPVarListLocTy &Locs);
+  /// Called on well-formed 'nontemporal' clause.
+  OMPClause *ActOnOpenMPNontemporalClause(ArrayRef<Expr *> VarList,
+                                          SourceLocation StartLoc,
+                                          SourceLocation LParenLoc,
+                                          SourceLocation EndLoc);
+
+  /// Data for list of allocators.
+  struct UsesAllocatorsData {
+    /// Allocator.
+    Expr *Allocator = nullptr;
+    /// Allocator traits.
+    Expr *AllocatorTraits = nullptr;
+    /// Locations of '(' and ')' symbols.
+    SourceLocation LParenLoc, RParenLoc;
+  };
+  /// Called on well-formed 'uses_allocators' clause.
+  OMPClause *ActOnOpenMPUsesAllocatorClause(SourceLocation StartLoc,
+                                            SourceLocation LParenLoc,
+                                            SourceLocation EndLoc,
+                                            ArrayRef<UsesAllocatorsData> Data);
+  /// Called on well-formed 'affinity' clause.
+  OMPClause *ActOnOpenMPAffinityClause(SourceLocation StartLoc,
+                                       SourceLocation LParenLoc,
+                                       SourceLocation ColonLoc,
+                                       SourceLocation EndLoc, Expr *Modifier,
+                                       ArrayRef<Expr *> Locators);
+  /// Called on a well-formed 'bind' clause.
+  OMPClause *ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
+                                   SourceLocation KindLoc,
+                                   SourceLocation StartLoc,
+                                   SourceLocation LParenLoc,
+                                   SourceLocation EndLoc);
+
+  /// Called on a well-formed 'ompx_dyn_cgroup_mem' clause.
+  OMPClause *ActOnOpenMPXDynCGroupMemClause(Expr *Size, SourceLocation StartLoc,
+                                            SourceLocation LParenLoc,
+                                            SourceLocation EndLoc);
+
+  /// Called on well-formed 'doacross' clause.
+  OMPClause *
+  ActOnOpenMPDoacrossClause(OpenMPDoacrossClauseModifier DepType,
+                            SourceLocation DepLoc, SourceLocation ColonLoc,
+                            ArrayRef<Expr *> VarList, SourceLocation StartLoc,
+                            SourceLocation LParenLoc, SourceLocation EndLoc);
+
+  /// Called on a well-formed 'ompx_attribute' clause.
+  OMPClause *ActOnOpenMPXAttributeClause(ArrayRef<const Attr *> Attrs,
+                                         SourceLocation StartLoc,
+                                         SourceLocation LParenLoc,
+                                         SourceLocation EndLoc);
+
+  /// Called on a well-formed 'ompx_bare' clause.
+  OMPClause *ActOnOpenMPXBareClause(SourceLocation StartLoc,
+                                    SourceLocation EndLoc);
+
+  ExprResult ActOnOMPArraySectionExpr(Expr *Base, SourceLocation LBLoc,
+                                      Expr *LowerBound,
+                                      SourceLocation ColonLocFirst,
+                                      SourceLocation ColonLocSecond,
+                                      Expr *Length, Expr *Stride,
+                                      SourceLocation RBLoc);
+  ExprResult ActOnOMPArrayShapingExpr(Expr *Base, SourceLocation LParenLoc,
+                                      SourceLocation RParenLoc,
+                                      ArrayRef<Expr *> Dims,
+                                      ArrayRef<SourceRange> Brackets);
+
+  /// Data structure for iterator expression.
+  struct OMPIteratorData {
+    IdentifierInfo *DeclIdent = nullptr;
+    SourceLocation DeclIdentLoc;
+    ParsedType Type;
+    OMPIteratorExpr::IteratorRange Range;
+    SourceLocation AssignLoc;
+    SourceLocation ColonLoc;
+    SourceLocation SecColonLoc;
+  };
+
+  ExprResult ActOnOMPIteratorExpr(Scope *S, SourceLocation IteratorKwLoc,
+                                  SourceLocation LLoc, SourceLocation RLoc,
+                                  ArrayRef<OMPIteratorData> Data);
+
+private:
+  void *VarDataSharingAttributesStack;
+
+  /// Number of nested '#pragma omp declare target' directives.
+  SmallVector<DeclareTargetContextInfo, 4> DeclareTargetNesting;
+
+  /// Initialization of data-sharing attributes stack.
+  void InitDataSharingAttributesStack();
+  void DestroyDataSharingAttributesStack();
+
+  /// Returns OpenMP nesting level for current directive.
+  unsigned getOpenMPNestingLevel() const;
+
+  /// Adjusts the function scopes index for the target-based regions.
+  void adjustOpenMPTargetScopeIndex(unsigned &FunctionScopesIndex,
+                                    unsigned Level) const;
+
+  /// Returns the number of scopes associated with the construct on the given
+  /// OpenMP level.
+  int getNumberOfConstructScopes(unsigned Level) const;
+
+  /// Push new OpenMP function region for non-capturing function.
+  void pushOpenMPFunctionRegion();
+
+  /// Pop OpenMP function region for non-capturing function.
+  void popOpenMPFunctionRegion(const sema::FunctionScopeInfo *OldFSI);
+
+  /// Analyzes and checks a loop nest for use by a loop transformation.
+  ///
+  /// \param Kind          The loop transformation directive kind.
+  /// \param NumLoops      How many nested loops the directive is expecting.
+  /// \param AStmt         Associated statement of the transformation directive.
+  /// \param LoopHelpers   [out] The loop analysis result.
+  /// \param Body          [out] The body code nested in \p NumLoops loop.
+  /// \param OriginalInits [out] Collection of statements and declarations that
+  ///                      must have been executed/declared before entering the
+  ///                      loop.
+  ///
+  /// \return Whether there was any error.
+  bool checkTransformableLoopNest(
+      OpenMPDirectiveKind Kind, Stmt *AStmt, int NumLoops,
+      SmallVectorImpl<OMPLoopBasedDirective::HelperExprs> &LoopHelpers,
+      Stmt *&Body,
+      SmallVectorImpl<SmallVector<llvm::PointerUnion<Stmt *, Decl *>, 0>>
+          &OriginalInits);
+
+  /// Helper to keep information about the current `omp begin/end declare
+  /// variant` nesting.
+  struct OMPDeclareVariantScope {
+    /// The associated OpenMP context selector.
+    OMPTraitInfo *TI;
+
+    /// The associated OpenMP context selector mangling.
+    std::string NameSuffix;
+
+    OMPDeclareVariantScope(OMPTraitInfo &TI);
+  };
+
+  /// Return the OMPTraitInfo for the surrounding scope, if any.
+  OMPTraitInfo *getOMPTraitInfoForSurroundingScope() {
+    return OMPDeclareVariantScopes.empty() ? nullptr
+                                           : OMPDeclareVariantScopes.back().TI;
+  }
+
+  /// The current `omp begin/end declare variant` scopes.
+  SmallVector<OMPDeclareVariantScope, 4> OMPDeclareVariantScopes;
+
+  /// The current `omp begin/end assumes` scopes.
+  SmallVector<OMPAssumeAttr *, 4> OMPAssumeScoped;
+
+  /// All `omp assumes` we encountered so far.
+  SmallVector<OMPAssumeAttr *, 4> OMPAssumeGlobal;
+
+  /// OMPD_loop is mapped to OMPD_for, OMPD_distribute or OMPD_simd depending
+  /// on the parameter of the bind clause. In the methods for the
+  /// mapped directives, check the parameters of the lastprivate clause.
+  bool checkLastPrivateForMappedDirectives(ArrayRef<OMPClause *> Clauses);
+  /// Depending on the bind clause of OMPD_loop map the directive to new
+  /// directives.
+  ///    1) loop bind(parallel) --> OMPD_for
+  ///    2) loop bind(teams) --> OMPD_distribute
+  ///    3) loop bind(thread) --> OMPD_simd
+  /// This is being handled in Sema instead of Codegen because of the need for
+  /// rigorous semantic checking in the new mapped directives.
+  bool mapLoopConstruct(llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
+                        ArrayRef<OMPClause *> Clauses,
+                        OpenMPBindClauseKind &BindKind,
+                        OpenMPDirectiveKind &Kind,
+                        OpenMPDirectiveKind &PrevMappedDirective,
+                        SourceLocation StartLoc, SourceLocation EndLoc,
+                        const DeclarationNameInfo &DirName,
+                        OpenMPDirectiveKind CancelRegion);
+};
+
+} // namespace clang
+
+#endif // LLVM_CLANG_SEMA_SEMAOPENMP_H
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index 2b934234b7cf5..c881b37507771 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -28,6 +28,7 @@
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaDiagnostic.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringSwitch.h"
@@ -2383,7 +2384,7 @@ Parser::DeclGroupPtrTy Parser::ParseDeclGroup(ParsingDeclSpec &DS,
       }
 
       if (getLangOpts().OpenMP)
-        Actions.startOpenMPCXXRangeFor();
+        Actions.OpenMP().startOpenMPCXXRangeFor();
       if (Tok.is(tok::l_brace))
         FRI->RangeExpr = ParseBraceInitializer();
       else
diff --git a/clang/lib/Parse/ParseExpr.cpp b/clang/lib/Parse/ParseExpr.cpp
index 473ec9afd6018..32d96f81c4c8d 100644
--- a/clang/lib/Parse/ParseExpr.cpp
+++ b/clang/lib/Parse/ParseExpr.cpp
@@ -31,6 +31,7 @@
 #include "clang/Sema/ParsedTemplate.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/SemaCUDA.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/SemaSYCL.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "llvm/ADT/SmallVector.h"
@@ -2075,7 +2076,7 @@ Parser::ParsePostfixExpressionSuffix(ExprResult LHS) {
           // replace this call to ActOnOpenACCArraySectionExpr in the future.
           // Eventually we'll genericize the OPenMPArraySectionExpr type as
           // well.
-          LHS = Actions.ActOnOMPArraySectionExpr(
+          LHS = Actions.OpenMP().ActOnOMPArraySectionExpr(
               LHS.get(), Loc, ArgExprs.empty() ? nullptr : ArgExprs[0],
               ColonLocFirst, ColonLocSecond, Length.get(), Stride.get(), RLoc);
         } else {
@@ -3277,7 +3278,7 @@ Parser::ParseParenExpression(ParenParseOption &ExprType, bool stopIfCastExpr,
     if (ErrorFound) {
       Result = ExprError();
     } else if (!Result.isInvalid()) {
-      Result = Actions.ActOnOMPArrayShapingExpr(
+      Result = Actions.OpenMP().ActOnOMPArrayShapingExpr(
           Result.get(), OpenLoc, RParenLoc, OMPDimensions, OMPBracketsRanges);
     }
     return Result;
diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp
index 814126e321d3b..480201bc06f61 100644
--- a/clang/lib/Parse/ParseOpenMP.cpp
+++ b/clang/lib/Parse/ParseOpenMP.cpp
@@ -21,6 +21,7 @@
 #include "clang/Parse/RAIIObjectsForParser.h"
 #include "clang/Sema/EnterExpressionEvaluationContext.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "llvm/ADT/PointerIntPair.h"
 #include "llvm/ADT/StringSwitch.h"
 #include "llvm/ADT/UniqueVector.h"
@@ -87,7 +88,7 @@ class DeclDirectiveListParserHelper final {
   DeclDirectiveListParserHelper(Parser *P, OpenMPDirectiveKind Kind)
       : P(P), Kind(Kind) {}
   void operator()(CXXScopeSpec &SS, DeclarationNameInfo NameInfo) {
-    ExprResult Res = P->getActions().ActOnOpenMPIdExpression(
+    ExprResult Res = P->getActions().OpenMP().ActOnOpenMPIdExpression(
         P->getCurScope(), SS, NameInfo, Kind);
     if (Res.isUsable())
       Identifiers.push_back(Res.get());
@@ -322,8 +323,8 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
     SourceRange Range;
     TypeResult TR = ParseTypeName(&Range, DeclaratorContext::Prototype, AS);
     if (TR.isUsable()) {
-      QualType ReductionType =
-          Actions.ActOnOpenMPDeclareReductionType(Range.getBegin(), TR);
+      QualType ReductionType = Actions.OpenMP().ActOnOpenMPDeclareReductionType(
+          Range.getBegin(), TR);
       if (!ReductionType.isNull()) {
         ReductionTypes.push_back(
             std::make_pair(ReductionType, Range.getBegin()));
@@ -363,8 +364,10 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
     return DeclGroupPtrTy();
   }
 
-  DeclGroupPtrTy DRD = Actions.ActOnOpenMPDeclareReductionDirectiveStart(
-      getCurScope(), Actions.getCurLexicalContext(), Name, ReductionTypes, AS);
+  DeclGroupPtrTy DRD =
+      Actions.OpenMP().ActOnOpenMPDeclareReductionDirectiveStart(
+          getCurScope(), Actions.getCurLexicalContext(), Name, ReductionTypes,
+          AS);
 
   // Parse <combiner> expression and then parse initializer if any for each
   // correct type.
@@ -375,10 +378,11 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
                                     Scope::CompoundStmtScope |
                                     Scope::OpenMPDirectiveScope);
     // Parse <combiner> expression.
-    Actions.ActOnOpenMPDeclareReductionCombinerStart(getCurScope(), D);
+    Actions.OpenMP().ActOnOpenMPDeclareReductionCombinerStart(getCurScope(), D);
     ExprResult CombinerResult = Actions.ActOnFinishFullExpr(
         ParseExpression().get(), D->getLocation(), /*DiscardedValue*/ false);
-    Actions.ActOnOpenMPDeclareReductionCombinerEnd(D, CombinerResult.get());
+    Actions.OpenMP().ActOnOpenMPDeclareReductionCombinerEnd(
+        D, CombinerResult.get());
 
     if (CombinerResult.isInvalid() && Tok.isNot(tok::r_paren) &&
         Tok.isNot(tok::annot_pragma_openmp_end)) {
@@ -411,8 +415,8 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
                                         Scope::OpenMPDirectiveScope);
         // Parse expression.
         VarDecl *OmpPrivParm =
-            Actions.ActOnOpenMPDeclareReductionInitializerStart(getCurScope(),
-                                                                D);
+            Actions.OpenMP().ActOnOpenMPDeclareReductionInitializerStart(
+                getCurScope(), D);
         // Check if initializer is omp_priv <init_expr> or something else.
         if (Tok.is(tok::identifier) &&
             Tok.getIdentifierInfo()->isStr("omp_priv")) {
@@ -423,7 +427,7 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
               ParseAssignmentExpression().get(), D->getLocation(),
               /*DiscardedValue*/ false);
         }
-        Actions.ActOnOpenMPDeclareReductionInitializerEnd(
+        Actions.OpenMP().ActOnOpenMPDeclareReductionInitializerEnd(
             D, InitializerResult.get(), OmpPrivParm);
         if (InitializerResult.isInvalid() && Tok.isNot(tok::r_paren) &&
             Tok.isNot(tok::annot_pragma_openmp_end)) {
@@ -444,8 +448,8 @@ Parser::ParseOpenMPDeclareReductionDirective(AccessSpecifier AS) {
     else
       TPA.Commit();
   }
-  return Actions.ActOnOpenMPDeclareReductionDirectiveEnd(getCurScope(), DRD,
-                                                         IsCorrect);
+  return Actions.OpenMP().ActOnOpenMPDeclareReductionDirectiveEnd(
+      getCurScope(), DRD, IsCorrect);
 }
 
 void Parser::ParseOpenMPReductionInitializerForDecl(VarDecl *OmpPrivParm) {
@@ -569,8 +573,8 @@ Parser::ParseOpenMPDeclareMapperDirective(AccessSpecifier AS) {
   SourceRange Range;
   TypeResult ParsedType = parseOpenMPDeclareMapperVarDecl(Range, VName, AS);
   if (ParsedType.isUsable())
-    MapperType =
-        Actions.ActOnOpenMPDeclareMapperType(Range.getBegin(), ParsedType);
+    MapperType = Actions.OpenMP().ActOnOpenMPDeclareMapperType(Range.getBegin(),
+                                                               ParsedType);
   if (MapperType.isNull())
     IsCorrect = false;
   if (!IsCorrect) {
@@ -591,11 +595,13 @@ Parser::ParseOpenMPDeclareMapperDirective(AccessSpecifier AS) {
   unsigned ScopeFlags = Scope::FnScope | Scope::DeclScope |
                         Scope::CompoundStmtScope | Scope::OpenMPDirectiveScope;
   ParseScope OMPDirectiveScope(this, ScopeFlags);
-  Actions.StartOpenMPDSABlock(OMPD_declare_mapper, DirName, getCurScope(), Loc);
+  Actions.OpenMP().StartOpenMPDSABlock(OMPD_declare_mapper, DirName,
+                                       getCurScope(), Loc);
 
   // Add the mapper variable declaration.
-  ExprResult MapperVarRef = Actions.ActOnOpenMPDeclareMapperDirectiveVarDecl(
-      getCurScope(), MapperType, Range.getBegin(), VName);
+  ExprResult MapperVarRef =
+      Actions.OpenMP().ActOnOpenMPDeclareMapperDirectiveVarDecl(
+          getCurScope(), MapperType, Range.getBegin(), VName);
 
   // Parse map clauses.
   SmallVector<OMPClause *, 6> Clauses;
@@ -603,7 +609,7 @@ Parser::ParseOpenMPDeclareMapperDirective(AccessSpecifier AS) {
     OpenMPClauseKind CKind = Tok.isAnnotation()
                                  ? OMPC_unknown
                                  : getOpenMPClauseKind(PP.getSpelling(Tok));
-    Actions.StartOpenMPClause(CKind);
+    Actions.OpenMP().StartOpenMPClause(CKind);
     OMPClause *Clause =
         ParseOpenMPClause(OMPD_declare_mapper, CKind, Clauses.empty());
     if (Clause)
@@ -613,7 +619,7 @@ Parser::ParseOpenMPDeclareMapperDirective(AccessSpecifier AS) {
     // Skip ',' if any.
     if (Tok.is(tok::comma))
       ConsumeToken();
-    Actions.EndOpenMPClause();
+    Actions.OpenMP().EndOpenMPClause();
   }
   if (Clauses.empty()) {
     Diag(Tok, diag::err_omp_expected_clause)
@@ -622,9 +628,9 @@ Parser::ParseOpenMPDeclareMapperDirective(AccessSpecifier AS) {
   }
 
   // Exit scope.
-  Actions.EndOpenMPDSABlock(nullptr);
+  Actions.OpenMP().EndOpenMPDSABlock(nullptr);
   OMPDirectiveScope.Exit();
-  DeclGroupPtrTy DG = Actions.ActOnOpenMPDeclareMapperDirective(
+  DeclGroupPtrTy DG = Actions.OpenMP().ActOnOpenMPDeclareMapperDirective(
       getCurScope(), Actions.getCurLexicalContext(), MapperId, MapperType,
       Range.getBegin(), VName, AS, MapperVarRef.get(), Clauses);
   if (!IsCorrect)
@@ -652,7 +658,8 @@ TypeResult Parser::parseOpenMPDeclareMapperVarDecl(SourceRange &Range,
   }
   Name = Actions.GetNameForDeclarator(DeclaratorInfo).getName();
 
-  return Actions.ActOnOpenMPDeclareMapperVarDecl(getCurScope(), DeclaratorInfo);
+  return Actions.OpenMP().ActOnOpenMPDeclareMapperVarDecl(getCurScope(),
+                                                          DeclaratorInfo);
 }
 
 namespace {
@@ -748,7 +755,7 @@ static bool parseDeclareSimdClauses(
       OpenMPClauseKind CKind = getOpenMPClauseKind(ClauseName);
       if (CKind == OMPC_uniform || CKind == OMPC_aligned ||
           CKind == OMPC_linear) {
-        Sema::OpenMPVarListDataTy Data;
+        SemaOpenMP::OpenMPVarListDataTy Data;
         SmallVectorImpl<Expr *> *Vars = &Uniforms;
         if (CKind == OMPC_aligned) {
           Vars = &Aligneds;
@@ -768,7 +775,7 @@ static bool parseDeclareSimdClauses(
           assert(0 <= Data.ExtraModifier &&
                  Data.ExtraModifier <= OMPC_LINEAR_unknown &&
                  "Unexpected linear modifier.");
-          if (P.getActions().CheckOpenMPLinearModifier(
+          if (P.getActions().OpenMP().CheckOpenMPLinearModifier(
                   static_cast<OpenMPLinearClauseKind>(Data.ExtraModifier),
                   Data.ExtraModifierLoc))
             Data.ExtraModifier = OMPC_LINEAR_val;
@@ -816,7 +823,7 @@ Parser::ParseOMPDeclareSimdClauses(Parser::DeclGroupPtrTy Ptr,
   SourceLocation EndLoc = ConsumeAnnotationToken();
   if (IsError)
     return Ptr;
-  return Actions.ActOnOpenMPDeclareSimdDirective(
+  return Actions.OpenMP().ActOnOpenMPDeclareSimdDirective(
       Ptr, BS, Simdlen.get(), Uniforms, Aligneds, Alignments, Linears,
       LinModifiers, Steps, SourceRange(Loc, EndLoc));
 }
@@ -1412,7 +1419,8 @@ void Parser::ParseOMPDeclareVariantClauses(Parser::DeclGroupPtrTy Ptr,
     return;
   }
 
-  OMPTraitInfo *ParentTI = Actions.getOMPTraitInfoForSurroundingScope();
+  OMPTraitInfo *ParentTI =
+      Actions.OpenMP().getOMPTraitInfoForSurroundingScope();
   ASTContext &ASTCtx = Actions.getASTContext();
   OMPTraitInfo &TI = ASTCtx.getNewOMPTraitInfo();
   SmallVector<Expr *, 6> AdjustNothing;
@@ -1445,7 +1453,7 @@ void Parser::ParseOMPDeclareVariantClauses(Parser::DeclGroupPtrTy Ptr,
       case OMPC_adjust_args: {
         AdjustArgsLoc = Tok.getLocation();
         ConsumeToken();
-        Sema::OpenMPVarListDataTy Data;
+        SemaOpenMP::OpenMPVarListDataTy Data;
         SmallVector<Expr *> Vars;
         IsError = ParseOpenMPVarList(OMPD_declare_variant, OMPC_adjust_args,
                                      Vars, Data);
@@ -1486,12 +1494,12 @@ void Parser::ParseOMPDeclareVariantClauses(Parser::DeclGroupPtrTy Ptr,
   }
 
   std::optional<std::pair<FunctionDecl *, Expr *>> DeclVarData =
-      Actions.checkOpenMPDeclareVariantFunction(
+      Actions.OpenMP().checkOpenMPDeclareVariantFunction(
           Ptr, AssociatedFunction.get(), TI, AppendArgs.size(),
           SourceRange(Loc, Tok.getLocation()));
 
   if (DeclVarData && !TI.Sets.empty())
-    Actions.ActOnOpenMPDeclareVariantDirective(
+    Actions.OpenMP().ActOnOpenMPDeclareVariantDirective(
         DeclVarData->first, DeclVarData->second, TI, AdjustNothing,
         AdjustNeedDevicePtr, AppendArgs, AdjustArgsLoc, AppendArgsLoc,
         SourceRange(Loc, Tok.getLocation()));
@@ -1642,7 +1650,7 @@ void Parser::ParseOpenMPClauses(OpenMPDirectiveKind DKind,
     OpenMPClauseKind CKind = Tok.isAnnotation()
                                  ? OMPC_unknown
                                  : getOpenMPClauseKind(PP.getSpelling(Tok));
-    Actions.StartOpenMPClause(CKind);
+    Actions.OpenMP().StartOpenMPClause(CKind);
     OMPClause *Clause = ParseOpenMPClause(
         DKind, CKind, !FirstClauses[unsigned(CKind)].getInt());
     SkipUntil(tok::comma, tok::identifier, tok::annot_pragma_openmp_end,
@@ -1651,13 +1659,13 @@ void Parser::ParseOpenMPClauses(OpenMPDirectiveKind DKind,
     if (Clause != nullptr)
       Clauses.push_back(Clause);
     if (Tok.is(tok::annot_pragma_openmp_end)) {
-      Actions.EndOpenMPClause();
+      Actions.OpenMP().EndOpenMPClause();
       break;
     }
     // Skip ',' if any.
     if (Tok.is(tok::comma))
       ConsumeToken();
-    Actions.EndOpenMPClause();
+    Actions.OpenMP().EndOpenMPClause();
   }
 }
 
@@ -1750,12 +1758,13 @@ void Parser::ParseOpenMPAssumesDirective(OpenMPDirectiveKind DKind,
     Assumptions.push_back(Assumption);
   }
 
-  Actions.ActOnOpenMPAssumesDirective(Loc, DKind, Assumptions, SkippedClauses);
+  Actions.OpenMP().ActOnOpenMPAssumesDirective(Loc, DKind, Assumptions,
+                                               SkippedClauses);
 }
 
 void Parser::ParseOpenMPEndAssumesDirective(SourceLocation Loc) {
-  if (Actions.isInOpenMPAssumeScope())
-    Actions.ActOnOpenMPEndAssumesDirective();
+  if (Actions.OpenMP().isInOpenMPAssumeScope())
+    Actions.OpenMP().ActOnOpenMPEndAssumesDirective();
   else
     Diag(Loc, diag::err_expected_begin_assumes);
 }
@@ -1811,7 +1820,7 @@ parseOpenMPSimpleClause(Parser &P, OpenMPClauseKind Kind) {
 }
 
 void Parser::ParseOMPDeclareTargetClauses(
-    Sema::DeclareTargetContextInfo &DTCI) {
+    SemaOpenMP::DeclareTargetContextInfo &DTCI) {
   SourceLocation DeviceTypeLoc;
   bool RequiresToOrLinkOrIndirectClause = false;
   bool HasToOrLinkOrIndirectClause = false;
@@ -1910,11 +1919,11 @@ void Parser::ParseOMPDeclareTargetClauses(
     if (DTCI.Kind == OMPD_declare_target || HasIdentifier) {
       auto &&Callback = [this, MT, &DTCI](CXXScopeSpec &SS,
                                           DeclarationNameInfo NameInfo) {
-        NamedDecl *ND =
-            Actions.lookupOpenMPDeclareTargetName(getCurScope(), SS, NameInfo);
+        NamedDecl *ND = Actions.OpenMP().lookupOpenMPDeclareTargetName(
+            getCurScope(), SS, NameInfo);
         if (!ND)
           return;
-        Sema::DeclareTargetContextInfo::MapInfo MI{MT, NameInfo.getLoc()};
+        SemaOpenMP::DeclareTargetContextInfo::MapInfo MI{MT, NameInfo.getLoc()};
         bool FirstMapping = DTCI.ExplicitlyMapped.try_emplace(ND, MI).second;
         if (!FirstMapping)
           Diag(NameInfo.getLoc(), diag::err_omp_declare_target_multiple)
@@ -2090,8 +2099,8 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
       skipUntilPragmaOpenMPEnd(DKind);
       // Skip the last annot_pragma_openmp_end.
       ConsumeAnnotationToken();
-      return Actions.ActOnOpenMPThreadprivateDirective(Loc,
-                                                       Helper.getIdentifiers());
+      return Actions.OpenMP().ActOnOpenMPThreadprivateDirective(
+          Loc, Helper.getIdentifiers());
     }
     break;
   }
@@ -2109,7 +2118,7 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
           OpenMPClauseKind CKind =
               Tok.isAnnotation() ? OMPC_unknown
                                  : getOpenMPClauseKind(PP.getSpelling(Tok));
-          Actions.StartOpenMPClause(CKind);
+          Actions.OpenMP().StartOpenMPClause(CKind);
           OMPClause *Clause = ParseOpenMPClause(
               OMPD_allocate, CKind, !FirstClauses[unsigned(CKind)].getInt());
           SkipUntil(tok::comma, tok::identifier, tok::annot_pragma_openmp_end,
@@ -2118,20 +2127,20 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
           if (Clause != nullptr)
             Clauses.push_back(Clause);
           if (Tok.is(tok::annot_pragma_openmp_end)) {
-            Actions.EndOpenMPClause();
+            Actions.OpenMP().EndOpenMPClause();
             break;
           }
           // Skip ',' if any.
           if (Tok.is(tok::comma))
             ConsumeToken();
-          Actions.EndOpenMPClause();
+          Actions.OpenMP().EndOpenMPClause();
         }
         skipUntilPragmaOpenMPEnd(DKind);
       }
       // Skip the last annot_pragma_openmp_end.
       ConsumeAnnotationToken();
-      return Actions.ActOnOpenMPAllocateDirective(Loc, Helper.getIdentifiers(),
-                                                  Clauses);
+      return Actions.OpenMP().ActOnOpenMPAllocateDirective(
+          Loc, Helper.getIdentifiers(), Clauses);
     }
     break;
   }
@@ -2150,7 +2159,7 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
       OpenMPClauseKind CKind = Tok.isAnnotation()
                                    ? OMPC_unknown
                                    : getOpenMPClauseKind(PP.getSpelling(Tok));
-      Actions.StartOpenMPClause(CKind);
+      Actions.OpenMP().StartOpenMPClause(CKind);
       OMPClause *Clause = ParseOpenMPClause(
           OMPD_requires, CKind, !FirstClauses[unsigned(CKind)].getInt());
       SkipUntil(tok::comma, tok::identifier, tok::annot_pragma_openmp_end,
@@ -2159,13 +2168,13 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
       if (Clause != nullptr)
         Clauses.push_back(Clause);
       if (Tok.is(tok::annot_pragma_openmp_end)) {
-        Actions.EndOpenMPClause();
+        Actions.OpenMP().EndOpenMPClause();
         break;
       }
       // Skip ',' if any.
       if (Tok.is(tok::comma))
         ConsumeToken();
-      Actions.EndOpenMPClause();
+      Actions.OpenMP().EndOpenMPClause();
     }
     // Consume final annot_pragma_openmp_end
     if (Clauses.empty()) {
@@ -2175,14 +2184,15 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
       return nullptr;
     }
     ConsumeAnnotationToken();
-    return Actions.ActOnOpenMPRequiresDirective(StartLoc, Clauses);
+    return Actions.OpenMP().ActOnOpenMPRequiresDirective(StartLoc, Clauses);
   }
   case OMPD_error: {
     SmallVector<OMPClause *, 1> Clauses;
     SourceLocation StartLoc = ConsumeToken();
     ParseOpenMPClauses(DKind, Clauses, StartLoc);
-    Actions.ActOnOpenMPErrorDirective(Clauses, StartLoc, SourceLocation(),
-                                      /*InExContext = */ false);
+    Actions.OpenMP().ActOnOpenMPErrorDirective(Clauses, StartLoc,
+                                               SourceLocation(),
+                                               /*InExContext = */ false);
     break;
   }
   case OMPD_assumes:
@@ -2217,7 +2227,8 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
     // { #pragma omp end declare variant }
     //
     ConsumeToken();
-    OMPTraitInfo *ParentTI = Actions.getOMPTraitInfoForSurroundingScope();
+    OMPTraitInfo *ParentTI =
+        Actions.OpenMP().getOMPTraitInfoForSurroundingScope();
     ASTContext &ASTCtx = Actions.getASTContext();
     OMPTraitInfo &TI = ASTCtx.getNewOMPTraitInfo();
     if (parseOMPDeclareVariantMatchClause(Loc, TI, ParentTI)) {
@@ -2248,7 +2259,7 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
         /* ConstructTraits */ ArrayRef<llvm::omp::TraitProperty>());
 
     if (isVariantApplicableInContext(VMI, OMPCtx, /* DeviceSetOnly */ true)) {
-      Actions.ActOnOpenMPBeginDeclareVariant(Loc, TI);
+      Actions.OpenMP().ActOnOpenMPBeginDeclareVariant(Loc, TI);
       break;
     }
 
@@ -2275,8 +2286,8 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
     break;
   }
   case OMPD_end_declare_variant: {
-    if (Actions.isInOpenMPDeclareVariantScope())
-      Actions.ActOnOpenMPEndDeclareVariant();
+    if (Actions.OpenMP().isInOpenMPDeclareVariantScope())
+      Actions.OpenMP().ActOnOpenMPEndDeclareVariant();
     else
       Diag(Loc, diag::err_expected_begin_declare_variant);
     ConsumeToken();
@@ -2331,7 +2342,7 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
   case OMPD_declare_target: {
     SourceLocation DTLoc = ConsumeAnyToken();
     bool HasClauses = Tok.isNot(tok::annot_pragma_openmp_end);
-    Sema::DeclareTargetContextInfo DTCI(DKind, DTLoc);
+    SemaOpenMP::DeclareTargetContextInfo DTCI(DKind, DTLoc);
     if (HasClauses)
       ParseOMPDeclareTargetClauses(DTCI);
     bool HasImplicitMappings = DKind == OMPD_begin_declare_target ||
@@ -2342,24 +2353,24 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
     ConsumeAnyToken();
 
     if (HasImplicitMappings) {
-      Actions.ActOnStartOpenMPDeclareTargetContext(DTCI);
+      Actions.OpenMP().ActOnStartOpenMPDeclareTargetContext(DTCI);
       return nullptr;
     }
 
-    Actions.ActOnFinishedOpenMPDeclareTargetContext(DTCI);
+    Actions.OpenMP().ActOnFinishedOpenMPDeclareTargetContext(DTCI);
     llvm::SmallVector<Decl *, 4> Decls;
     for (auto &It : DTCI.ExplicitlyMapped)
       Decls.push_back(It.first);
     return Actions.BuildDeclaratorGroup(Decls);
   }
   case OMPD_end_declare_target: {
-    if (!Actions.isInOpenMPDeclareTargetContext()) {
+    if (!Actions.OpenMP().isInOpenMPDeclareTargetContext()) {
       Diag(Tok, diag::err_omp_unexpected_directive)
           << 1 << getOpenMPDirectiveName(DKind);
       break;
     }
-    const Sema::DeclareTargetContextInfo &DTCI =
-        Actions.ActOnOpenMPEndDeclareTargetDirective();
+    const SemaOpenMP::DeclareTargetContextInfo &DTCI =
+        Actions.OpenMP().ActOnOpenMPEndDeclareTargetDirective();
     ParseOMPEndDeclareTargetDirective(DTCI.Kind, DKind, DTCI.Loc);
     return nullptr;
   }
@@ -2683,7 +2694,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
     if (!ParseOpenMPSimpleVarList(DKind, Helper,
                                   /*AllowScopeSpecifier=*/false)) {
       skipUntilPragmaOpenMPEnd(DKind);
-      DeclGroupPtrTy Res = Actions.ActOnOpenMPThreadprivateDirective(
+      DeclGroupPtrTy Res = Actions.OpenMP().ActOnOpenMPThreadprivateDirective(
           Loc, Helper.getIdentifiers());
       Directive = Actions.ActOnDeclStmt(Res, Loc, Tok.getLocation());
     }
@@ -2710,7 +2721,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
           OpenMPClauseKind CKind =
               Tok.isAnnotation() ? OMPC_unknown
                                  : getOpenMPClauseKind(PP.getSpelling(Tok));
-          Actions.StartOpenMPClause(CKind);
+          Actions.OpenMP().StartOpenMPClause(CKind);
           OMPClause *Clause = ParseOpenMPClause(
               OMPD_allocate, CKind, !FirstClauses[unsigned(CKind)].getInt());
           SkipUntil(tok::comma, tok::identifier, tok::annot_pragma_openmp_end,
@@ -2719,17 +2730,17 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
           if (Clause != nullptr)
             Clauses.push_back(Clause);
           if (Tok.is(tok::annot_pragma_openmp_end)) {
-            Actions.EndOpenMPClause();
+            Actions.OpenMP().EndOpenMPClause();
             break;
           }
           // Skip ',' if any.
           if (Tok.is(tok::comma))
             ConsumeToken();
-          Actions.EndOpenMPClause();
+          Actions.OpenMP().EndOpenMPClause();
         }
         skipUntilPragmaOpenMPEnd(DKind);
       }
-      DeclGroupPtrTy Res = Actions.ActOnOpenMPAllocateDirective(
+      DeclGroupPtrTy Res = Actions.OpenMP().ActOnOpenMPAllocateDirective(
           Loc, Helper.getIdentifiers(), Clauses);
       Directive = Actions.ActOnDeclStmt(Res, Loc, Tok.getLocation());
     }
@@ -2875,7 +2886,8 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
     if (isOpenMPSimdDirective(DKind))
       ScopeFlags |= Scope::OpenMPSimdDirectiveScope;
     ParseScope OMPDirectiveScope(this, ScopeFlags);
-    Actions.StartOpenMPDSABlock(DKind, DirName, Actions.getCurScope(), Loc);
+    Actions.OpenMP().StartOpenMPDSABlock(DKind, DirName, Actions.getCurScope(),
+                                         Loc);
 
     while (Tok.isNot(tok::annot_pragma_openmp_end)) {
       // If we are parsing for a directive within a metadirective, the directive
@@ -2909,7 +2921,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
       }
       // No more implicit clauses allowed.
       ImplicitClauseAllowed = false;
-      Actions.StartOpenMPClause(CKind);
+      Actions.OpenMP().StartOpenMPClause(CKind);
       HasImplicitClause = false;
       OMPClause *Clause = ParseOpenMPClause(
           DKind, CKind, !FirstClauses[unsigned(CKind)].getInt());
@@ -2922,7 +2934,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
       // Skip ',' if any.
       if (Tok.is(tok::comma))
         ConsumeToken();
-      Actions.EndOpenMPClause();
+      Actions.OpenMP().EndOpenMPClause();
     }
     // End location of the directive.
     EndLoc = Tok.getLocation();
@@ -2953,7 +2965,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
     StmtResult AssociatedStmt;
     if (HasAssociatedStatement) {
       // The body is a block scope like in Lambdas and Blocks.
-      Actions.ActOnOpenMPRegionStart(DKind, getCurScope());
+      Actions.OpenMP().ActOnOpenMPRegionStart(DKind, getCurScope());
       // FIXME: We create a bogus CompoundStmt scope to hold the contents of
       // the captured region. Code elsewhere assumes that any FunctionScopeInfo
       // should have at least one compound statement scope within it.
@@ -2964,30 +2976,33 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
 
         if (AssociatedStmt.isUsable() && isOpenMPLoopDirective(DKind) &&
             getLangOpts().OpenMPIRBuilder)
-          AssociatedStmt = Actions.ActOnOpenMPLoopnest(AssociatedStmt.get());
+          AssociatedStmt =
+              Actions.OpenMP().ActOnOpenMPLoopnest(AssociatedStmt.get());
       }
-      AssociatedStmt = Actions.ActOnOpenMPRegionEnd(AssociatedStmt, Clauses);
+      AssociatedStmt =
+          Actions.OpenMP().ActOnOpenMPRegionEnd(AssociatedStmt, Clauses);
     } else if (DKind == OMPD_target_update || DKind == OMPD_target_enter_data ||
                DKind == OMPD_target_exit_data) {
-      Actions.ActOnOpenMPRegionStart(DKind, getCurScope());
+      Actions.OpenMP().ActOnOpenMPRegionStart(DKind, getCurScope());
       AssociatedStmt = (Sema::CompoundScopeRAII(Actions),
                         Actions.ActOnCompoundStmt(Loc, Loc, std::nullopt,
                                                   /*isStmtExpr=*/false));
-      AssociatedStmt = Actions.ActOnOpenMPRegionEnd(AssociatedStmt, Clauses);
+      AssociatedStmt =
+          Actions.OpenMP().ActOnOpenMPRegionEnd(AssociatedStmt, Clauses);
     }
-    Directive = Actions.ActOnOpenMPExecutableDirective(
+    Directive = Actions.OpenMP().ActOnOpenMPExecutableDirective(
         DKind, DirName, CancelRegion, Clauses, AssociatedStmt.get(), Loc,
         EndLoc);
 
     // Exit scope.
-    Actions.EndOpenMPDSABlock(Directive.get());
+    Actions.OpenMP().EndOpenMPDSABlock(Directive.get());
     OMPDirectiveScope.Exit();
     break;
   }
   case OMPD_declare_target: {
     SourceLocation DTLoc = ConsumeAnyToken();
     bool HasClauses = Tok.isNot(tok::annot_pragma_openmp_end);
-    Sema::DeclareTargetContextInfo DTCI(DKind, DTLoc);
+    SemaOpenMP::DeclareTargetContextInfo DTCI(DKind, DTLoc);
     if (HasClauses)
       ParseOMPDeclareTargetClauses(DTCI);
     bool HasImplicitMappings =
@@ -3003,7 +3018,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
     // Skip the last annot_pragma_openmp_end.
     ConsumeAnyToken();
 
-    Actions.ActOnFinishedOpenMPDeclareTargetContext(DTCI);
+    Actions.OpenMP().ActOnFinishedOpenMPDeclareTargetContext(DTCI);
     break;
   }
   case OMPD_declare_simd:
@@ -3118,7 +3133,7 @@ OMPClause *Parser::ParseOpenMPSizesClause() {
 
   T.consumeClose();
 
-  return Actions.ActOnOpenMPSizesClause(
+  return Actions.OpenMP().ActOnOpenMPSizesClause(
       ValExprs, ClauseNameLoc, T.getOpenLocation(), T.getCloseLocation());
 }
 
@@ -3130,7 +3145,7 @@ OMPClause *Parser::ParseOpenMPUsesAllocatorClause(OpenMPDirectiveKind DKind) {
   BalancedDelimiterTracker T(*this, tok::l_paren, tok::annot_pragma_openmp_end);
   if (T.expectAndConsume(diag::err_expected_lparen_after, "uses_allocator"))
     return nullptr;
-  SmallVector<Sema::UsesAllocatorsData, 4> Data;
+  SmallVector<SemaOpenMP::UsesAllocatorsData, 4> Data;
   do {
     CXXScopeSpec SS;
     Token Replacement;
@@ -3144,7 +3159,7 @@ OMPClause *Parser::ParseOpenMPUsesAllocatorClause(OpenMPDirectiveKind DKind) {
                 StopBeforeMatch);
       break;
     }
-    Sema::UsesAllocatorsData &D = Data.emplace_back();
+    SemaOpenMP::UsesAllocatorsData &D = Data.emplace_back();
     D.Allocator = Allocator.get();
     if (Tok.is(tok::l_paren)) {
       BalancedDelimiterTracker T(*this, tok::l_paren,
@@ -3169,8 +3184,8 @@ OMPClause *Parser::ParseOpenMPUsesAllocatorClause(OpenMPDirectiveKind DKind) {
       ConsumeAnyToken();
   } while (Tok.isNot(tok::r_paren) && Tok.isNot(tok::annot_pragma_openmp_end));
   T.consumeClose();
-  return Actions.ActOnOpenMPUsesAllocatorClause(Loc, T.getOpenLocation(),
-                                                T.getCloseLocation(), Data);
+  return Actions.OpenMP().ActOnOpenMPUsesAllocatorClause(
+      Loc, T.getOpenLocation(), T.getCloseLocation(), Data);
 }
 
 /// Parsing of OpenMP clauses.
@@ -3538,15 +3553,16 @@ OMPClause *Parser::ParseOpenMPSingleExprClause(OpenMPClauseKind Kind,
 
   if (ParseOnly)
     return nullptr;
-  return Actions.ActOnOpenMPSingleExprClause(Kind, Val.get(), Loc, LLoc, RLoc);
+  return Actions.OpenMP().ActOnOpenMPSingleExprClause(Kind, Val.get(), Loc,
+                                                      LLoc, RLoc);
 }
 
 /// Parse indirect clause for '#pragma omp declare target' directive.
 ///  'indirect' '[' '(' invoked-by-fptr ')' ']'
 /// where invoked-by-fptr is a constant boolean expression that evaluates to
 /// true or false at compile time.
-bool Parser::ParseOpenMPIndirectClause(Sema::DeclareTargetContextInfo &DTCI,
-                                       bool ParseOnly) {
+bool Parser::ParseOpenMPIndirectClause(
+    SemaOpenMP::DeclareTargetContextInfo &DTCI, bool ParseOnly) {
   SourceLocation Loc = ConsumeToken();
   SourceLocation RLoc;
 
@@ -3721,15 +3737,16 @@ OMPClause *Parser::ParseOpenMPInteropClause(OpenMPClauseKind Kind,
     return nullptr;
 
   if (Kind == OMPC_init)
-    return Actions.ActOnOpenMPInitClause(InteropVarExpr.get(), InteropInfo, Loc,
-                                         T.getOpenLocation(), VarLoc, RLoc);
+    return Actions.OpenMP().ActOnOpenMPInitClause(
+        InteropVarExpr.get(), InteropInfo, Loc, T.getOpenLocation(), VarLoc,
+        RLoc);
   if (Kind == OMPC_use)
-    return Actions.ActOnOpenMPUseClause(InteropVarExpr.get(), Loc,
-                                        T.getOpenLocation(), VarLoc, RLoc);
+    return Actions.OpenMP().ActOnOpenMPUseClause(
+        InteropVarExpr.get(), Loc, T.getOpenLocation(), VarLoc, RLoc);
 
   if (Kind == OMPC_destroy)
-    return Actions.ActOnOpenMPDestroyClause(InteropVarExpr.get(), Loc,
-                                            T.getOpenLocation(), VarLoc, RLoc);
+    return Actions.OpenMP().ActOnOpenMPDestroyClause(
+        InteropVarExpr.get(), Loc, T.getOpenLocation(), VarLoc, RLoc);
 
   llvm_unreachable("Unexpected interop variable clause.");
 }
@@ -3787,8 +3804,8 @@ OMPClause *Parser::ParseOpenMPOMPXAttributesClause(bool ParseOnly) {
     };
   }
 
-  return Actions.ActOnOpenMPXAttributeClause(Attrs, Loc, T.getOpenLocation(),
-                                             T.getCloseLocation());
+  return Actions.OpenMP().ActOnOpenMPXAttributeClause(
+      Attrs, Loc, T.getOpenLocation(), T.getCloseLocation());
 }
 
 /// Parsing of simple OpenMP clauses like 'default' or 'proc_bind'.
@@ -3823,9 +3840,8 @@ OMPClause *Parser::ParseOpenMPSimpleClause(OpenMPClauseKind Kind,
         << getOpenMPClauseName(OMPC_default) << "5.1";
     return nullptr;
   }
-  return Actions.ActOnOpenMPSimpleClause(Kind, Val->Type,
-                                         Val->TypeLoc, Val->LOpen,
-                                         Val->Loc, Val->RLoc);
+  return Actions.OpenMP().ActOnOpenMPSimpleClause(
+      Kind, Val->Type, Val->TypeLoc, Val->LOpen, Val->Loc, Val->RLoc);
 }
 
 /// Parsing of OpenMP clauses like 'ordered'.
@@ -3860,7 +3876,7 @@ OMPClause *Parser::ParseOpenMPClause(OpenMPClauseKind Kind, bool ParseOnly) {
 
   if (ParseOnly)
     return nullptr;
-  return Actions.ActOnOpenMPClause(Kind, Loc, Tok.getLocation());
+  return Actions.OpenMP().ActOnOpenMPClause(Kind, Loc, Tok.getLocation());
 }
 
 /// Parsing of OpenMP clauses with single expressions and some additional
@@ -4118,7 +4134,7 @@ OMPClause *Parser::ParseOpenMPSingleExprWithArgClause(OpenMPDirectiveKind DKind,
 
   if (ParseOnly)
     return nullptr;
-  return Actions.ActOnOpenMPSingleExprWithArgClause(
+  return Actions.OpenMP().ActOnOpenMPSingleExprWithArgClause(
       Kind, Arg, Val.get(), Loc, T.getOpenLocation(), KLoc, DelimLoc, RLoc);
 }
 
@@ -4184,7 +4200,7 @@ static OpenMPMapModifierKind isMapModifier(Parser &P) {
 }
 
 /// Parse the mapper modifier in map, to, and from clauses.
-bool Parser::parseMapperModifier(Sema::OpenMPVarListDataTy &Data) {
+bool Parser::parseMapperModifier(SemaOpenMP::OpenMPVarListDataTy &Data) {
   // Parse '('.
   BalancedDelimiterTracker T(*this, tok::l_paren, tok::colon);
   if (T.expectAndConsume(diag::err_expected_lparen_after, "mapper")) {
@@ -4216,7 +4232,7 @@ bool Parser::parseMapperModifier(Sema::OpenMPVarListDataTy &Data) {
 /// map([ [map-type-modifier[,] [map-type-modifier[,] ...] map-type : ] list)
 /// where, map-type-modifier ::= always | close | mapper(mapper-identifier) |
 /// present
-bool Parser::parseMapTypeModifiers(Sema::OpenMPVarListDataTy &Data) {
+bool Parser::parseMapTypeModifiers(SemaOpenMP::OpenMPVarListDataTy &Data) {
   while (getCurToken().isNot(tok::colon)) {
     OpenMPMapModifierKind TypeModifier = isMapModifier(*this);
     if (TypeModifier == OMPC_MAP_MODIFIER_always ||
@@ -4282,7 +4298,7 @@ static OpenMPMapClauseKind isMapType(Parser &P) {
 /// Parse map-type in map clause.
 /// map([ [map-type-modifier[,] [map-type-modifier[,] ...] map-type : ] list)
 /// where, map-type ::= to | from | tofrom | alloc | release | delete
-static void parseMapType(Parser &P, Sema::OpenMPVarListDataTy &Data) {
+static void parseMapType(Parser &P, SemaOpenMP::OpenMPVarListDataTy &Data) {
   Token Tok = P.getCurToken();
   if (Tok.is(tok::colon)) {
     P.Diag(Tok, diag::err_omp_map_type_missing);
@@ -4306,7 +4322,7 @@ ExprResult Parser::ParseOpenMPIteratorsExpr() {
     return ExprError();
 
   SourceLocation LLoc = T.getOpenLocation();
-  SmallVector<Sema::OMPIteratorData, 4> Data;
+  SmallVector<SemaOpenMP::OMPIteratorData, 4> Data;
   while (Tok.isNot(tok::r_paren) && Tok.isNot(tok::annot_pragma_openmp_end)) {
     // Check if the type parsing is required.
     ParsedType IteratorType;
@@ -4380,7 +4396,7 @@ ExprResult Parser::ParseOpenMPIteratorsExpr() {
     if (Tok.is(tok::comma))
       ConsumeToken();
 
-    Sema::OMPIteratorData &D = Data.emplace_back();
+    SemaOpenMP::OMPIteratorData &D = Data.emplace_back();
     D.DeclIdent = II;
     D.DeclIdentLoc = IdLoc;
     D.Type = IteratorType;
@@ -4397,12 +4413,12 @@ ExprResult Parser::ParseOpenMPIteratorsExpr() {
   if (!T.consumeClose())
     RLoc = T.getCloseLocation();
 
-  return Actions.ActOnOMPIteratorExpr(getCurScope(), IteratorKwLoc, LLoc, RLoc,
-                                      Data);
+  return Actions.OpenMP().ActOnOMPIteratorExpr(getCurScope(), IteratorKwLoc,
+                                               LLoc, RLoc, Data);
 }
 
 bool Parser::ParseOpenMPReservedLocator(OpenMPClauseKind Kind,
-                                        Sema::OpenMPVarListDataTy &Data,
+                                        SemaOpenMP::OpenMPVarListDataTy &Data,
                                         const LangOptions &LangOpts) {
   // Currently the only reserved locator is 'omp_all_memory' which is only
   // allowed on a depend clause.
@@ -4430,7 +4446,7 @@ bool Parser::ParseOpenMPReservedLocator(OpenMPClauseKind Kind,
 
 /// Parse step size expression. Returns true if parsing is successfull,
 /// otherwise returns false.
-static bool parseStepSize(Parser &P, Sema::OpenMPVarListDataTy &Data,
+static bool parseStepSize(Parser &P, SemaOpenMP::OpenMPVarListDataTy &Data,
                           OpenMPClauseKind CKind, SourceLocation ELoc) {
   ExprResult Tail = P.ParseAssignmentExpression();
   Sema &Actions = P.getActions();
@@ -4451,7 +4467,7 @@ static bool parseStepSize(Parser &P, Sema::OpenMPVarListDataTy &Data,
 bool Parser::ParseOpenMPVarList(OpenMPDirectiveKind DKind,
                                 OpenMPClauseKind Kind,
                                 SmallVectorImpl<Expr *> &Vars,
-                                Sema::OpenMPVarListDataTy &Data) {
+                                SemaOpenMP::OpenMPVarListDataTy &Data) {
   UnqualifiedId UnqualifiedReductionId;
   bool InvalidReductionId = false;
   bool IsInvalidMapperModifier = false;
@@ -4961,7 +4977,7 @@ OMPClause *Parser::ParseOpenMPVarListClause(OpenMPDirectiveKind DKind,
   SourceLocation Loc = Tok.getLocation();
   SourceLocation LOpen = ConsumeToken();
   SmallVector<Expr *, 4> Vars;
-  Sema::OpenMPVarListDataTy Data;
+  SemaOpenMP::OpenMPVarListDataTy Data;
 
   if (ParseOpenMPVarList(DKind, Kind, Vars, Data))
     return nullptr;
@@ -4969,5 +4985,5 @@ OMPClause *Parser::ParseOpenMPVarListClause(OpenMPDirectiveKind DKind,
   if (ParseOnly)
     return nullptr;
   OMPVarListLocTy Locs(Loc, LOpen, Data.RLoc);
-  return Actions.ActOnOpenMPVarListClause(Kind, Vars, Locs, Data);
+  return Actions.OpenMP().ActOnOpenMPVarListClause(Kind, Vars, Locs, Data);
 }
diff --git a/clang/lib/Parse/ParseStmt.cpp b/clang/lib/Parse/ParseStmt.cpp
index 76a3fa8f2627d..629421c01d17d 100644
--- a/clang/lib/Parse/ParseStmt.cpp
+++ b/clang/lib/Parse/ParseStmt.cpp
@@ -22,6 +22,7 @@
 #include "clang/Sema/DeclSpec.h"
 #include "clang/Sema/EnterExpressionEvaluationContext.h"
 #include "clang/Sema/Scope.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "llvm/ADT/STLExtras.h"
 #include <optional>
@@ -2301,7 +2302,7 @@ StmtResult Parser::ParseForStatement(SourceLocation *TrailingElseLoc) {
     // In OpenMP loop region loop control variable must be captured and be
     // private. Perform analysis of first part (if any).
     if (getLangOpts().OpenMP && FirstPart.isUsable()) {
-      Actions.ActOnOpenMPLoopInitialization(ForLoc, FirstPart.get());
+      Actions.OpenMP().ActOnOpenMPLoopInitialization(ForLoc, FirstPart.get());
     }
   }
 
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index 8de202f4f7a0c..a1e32d391ed0c 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -46,6 +46,7 @@
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/SemaOpenACC.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/SemaSYCL.h"
 #include "clang/Sema/TemplateDeduction.h"
 #include "clang/Sema/TemplateInstCallback.h"
@@ -203,6 +204,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       CUDAPtr(std::make_unique<SemaCUDA>(*this)),
       HLSLPtr(std::make_unique<SemaHLSL>(*this)),
       OpenACCPtr(std::make_unique<SemaOpenACC>(*this)),
+      OpenMPPtr(std::make_unique<SemaOpenMP>(*this)),
       SYCLPtr(std::make_unique<SemaSYCL>(*this)),
       MSPointerToMemberRepresentationMethod(
           LangOpts.getMSPointerToMemberRepresentationMethod()),
@@ -226,8 +228,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       StringWithUTF8StringMethod(nullptr),
       ValueWithBytesObjCTypeMethod(nullptr), NSArrayDecl(nullptr),
       ArrayWithObjectsMethod(nullptr), NSDictionaryDecl(nullptr),
-      DictionaryWithObjectsMethod(nullptr), CodeCompleter(CodeCompleter),
-      VarDataSharingAttributesStack(nullptr) {
+      DictionaryWithObjectsMethod(nullptr), CodeCompleter(CodeCompleter) {
   assert(pp.TUKind == TUKind);
   TUScope = nullptr;
 
@@ -252,7 +253,7 @@ Sema::Sema(Preprocessor &pp, ASTContext &ctxt, ASTConsumer &consumer,
       nullptr, ExpressionEvaluationContextRecord::EK_Other);
 
   // Initialization of data sharing attributes stack for OpenMP
-  InitDataSharingAttributesStack();
+  OpenMP().InitDataSharingAttributesStack();
 
   std::unique_ptr<sema::SemaPPCallbacks> Callbacks =
       std::make_unique<sema::SemaPPCallbacks>();
@@ -501,7 +502,7 @@ Sema::~Sema() {
   threadSafety::threadSafetyCleanup(ThreadSafetyDeclCache);
 
   // Destroys data sharing attributes stack for OpenMP
-  DestroyDataSharingAttributesStack();
+  OpenMP().DestroyDataSharingAttributesStack();
 
   // Detach from the PP callback handler which outlives Sema since it's owned
   // by the preprocessor.
@@ -1159,7 +1160,7 @@ void Sema::ActOnEndOfTranslationUnit() {
 
   DiagnoseUnterminatedPragmaAlignPack();
   DiagnoseUnterminatedPragmaAttribute();
-  DiagnoseUnterminatedOpenMPDeclareTarget();
+  OpenMP().DiagnoseUnterminatedOpenMPDeclareTarget();
 
   // All delayed member exception specs should be checked or we end up accepting
   // incompatible declarations.
@@ -1747,7 +1748,7 @@ class DeferredDiagnosticsEmitter
     // Finalize analysis of OpenMP-specific constructs.
     if (Caller && S.LangOpts.OpenMP && UsePath.size() == 1 &&
         (ShouldEmitRootNode || InOMPDeviceContext))
-      S.finalizeOpenMPDelayedAnalysis(Caller, FD, Loc);
+      S.OpenMP().finalizeOpenMPDelayedAnalysis(Caller, FD, Loc);
     if (Caller)
       S.CUDA().DeviceKnownEmittedFns[FD] = {Caller, Loc};
     // Always emit deferred diagnostics for the direct users. This does not
@@ -1899,8 +1900,8 @@ Sema::targetDiag(SourceLocation Loc, unsigned DiagID, const FunctionDecl *FD) {
   FD = FD ? FD : getCurFunctionDecl();
   if (LangOpts.OpenMP)
     return LangOpts.OpenMPIsTargetDevice
-               ? diagIfOpenMPDeviceCode(Loc, DiagID, FD)
-               : diagIfOpenMPHostCode(Loc, DiagID, FD);
+               ? OpenMP().diagIfOpenMPDeviceCode(Loc, DiagID, FD)
+               : OpenMP().diagIfOpenMPHostCode(Loc, DiagID, FD);
   if (getLangOpts().CUDA)
     return getLangOpts().CUDAIsDevice ? CUDA().DiagIfDeviceCode(Loc, DiagID)
                                       : CUDA().DiagIfHostCode(Loc, DiagID);
@@ -2131,7 +2132,7 @@ void Sema::PushFunctionScope() {
     FunctionScopes.push_back(new FunctionScopeInfo(getDiagnostics()));
   }
   if (LangOpts.OpenMP)
-    pushOpenMPFunctionRegion();
+    OpenMP().pushOpenMPFunctionRegion();
 }
 
 void Sema::PushBlockScope(Scope *BlockScope, BlockDecl *Block) {
@@ -2251,7 +2252,7 @@ Sema::PopFunctionScopeInfo(const AnalysisBasedWarnings::Policy *WP,
                                PoppedFunctionScopeDeleter(this));
 
   if (LangOpts.OpenMP)
-    popOpenMPFunctionRegion(Scope.get());
+    OpenMP().popOpenMPFunctionRegion(Scope.get());
 
   // Issue any analysis-based warnings.
   if (WP && D)
@@ -2687,7 +2688,9 @@ void Sema::PushCapturedRegionScope(Scope *S, CapturedDecl *CD, RecordDecl *RD,
                                    unsigned OpenMPCaptureLevel) {
   auto *CSI = new CapturedRegionScopeInfo(
       getDiagnostics(), S, CD, RD, CD->getContextParam(), K,
-      (getLangOpts().OpenMP && K == CR_OpenMP) ? getOpenMPNestingLevel() : 0,
+      (getLangOpts().OpenMP && K == CR_OpenMP)
+          ? OpenMP().getOpenMPNestingLevel()
+          : 0,
       OpenMPCaptureLevel);
   CSI->ReturnType = Context.VoidTy;
   FunctionScopes.push_back(CSI);
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 8b3b9d020db57..390da508518e1 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -48,6 +48,7 @@
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaHLSL.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/SmallString.h"
@@ -6168,11 +6169,12 @@ Decl *Sema::ActOnDeclarator(Scope *S, Declarator &D) {
   // Check if we are in an `omp begin/end declare variant` scope. Handle this
   // declaration only if the `bind_to_declaration` extension is set.
   SmallVector<FunctionDecl *, 4> Bases;
-  if (LangOpts.OpenMP && isInOpenMPDeclareVariantScope())
-    if (getOMPTraitInfoForSurroundingScope()->isExtensionActive(llvm::omp::TraitProperty::
-              implementation_extension_bind_to_declaration))
-    ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
-        S, D, MultiTemplateParamsArg(), Bases);
+  if (LangOpts.OpenMP && OpenMP().isInOpenMPDeclareVariantScope())
+    if (OpenMP().getOMPTraitInfoForSurroundingScope()->isExtensionActive(
+            llvm::omp::TraitProperty::
+                implementation_extension_bind_to_declaration))
+      OpenMP().ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
+          S, D, MultiTemplateParamsArg(), Bases);
 
   Decl *Dcl = HandleDeclarator(S, D, MultiTemplateParamsArg());
 
@@ -6181,7 +6183,8 @@ Decl *Sema::ActOnDeclarator(Scope *S, Declarator &D) {
     Dcl->setTopLevelDeclInObjCContainer();
 
   if (!Bases.empty())
-    ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(Dcl, Bases);
+    OpenMP().ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(Dcl,
+                                                                        Bases);
 
   return Dcl;
 }
@@ -6568,8 +6571,8 @@ NamedDecl *Sema::HandleDeclarator(Scope *S, Declarator &D,
   if (New->getDeclName() && AddToScope)
     PushOnScopeChains(New, S);
 
-  if (isInOpenMPDeclareTargetContext())
-    checkDeclIsAllowedInOpenMPTarget(nullptr, New);
+  if (OpenMP().isInOpenMPDeclareTargetContext())
+    OpenMP().checkDeclIsAllowedInOpenMPTarget(nullptr, New);
 
   return New;
 }
@@ -12268,7 +12271,7 @@ bool Sema::CheckFunctionDeclaration(Scope *S, FunctionDecl *NewFD,
   }
 
   if (LangOpts.OpenMP)
-    ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(NewFD);
+    OpenMP().ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(NewFD);
 
   // Semantic checking for this function declaration (in isolation).
 
@@ -14956,7 +14959,7 @@ Sema::DeclGroupPtrTy Sema::FinalizeDeclaratorGroup(Scope *S, const DeclSpec &DS,
       if (auto *VD = dyn_cast<VarDecl>(D);
           LangOpts.OpenMP && VD && VD->hasAttr<OMPDeclareTargetDeclAttr>() &&
           VD->hasGlobalStorage())
-        ActOnOpenMPDeclareTargetInitializer(D);
+        OpenMP().ActOnOpenMPDeclareTargetInitializer(D);
       // For declarators, there are some additional syntactic-ish checks we need
       // to perform.
       if (auto *DD = dyn_cast<DeclaratorDecl>(D)) {
@@ -15495,8 +15498,8 @@ Sema::ActOnStartOfFunctionDef(Scope *FnBodyScope, Declarator &D,
   // specialization function under the OpenMP context defined as part of the
   // `omp begin declare variant`.
   SmallVector<FunctionDecl *, 4> Bases;
-  if (LangOpts.OpenMP && isInOpenMPDeclareVariantScope())
-    ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
+  if (LangOpts.OpenMP && OpenMP().isInOpenMPDeclareVariantScope())
+    OpenMP().ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
         ParentScope, D, TemplateParameterLists, Bases);
 
   D.setFunctionDefinitionKind(FunctionDefinitionKind::Definition);
@@ -15504,7 +15507,8 @@ Sema::ActOnStartOfFunctionDef(Scope *FnBodyScope, Declarator &D,
   Decl *Dcl = ActOnStartOfFunctionDef(FnBodyScope, DP, SkipBody, BodyKind);
 
   if (!Bases.empty())
-    ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(Dcl, Bases);
+    OpenMP().ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(Dcl,
+                                                                        Bases);
 
   return Dcl;
 }
@@ -20651,7 +20655,7 @@ Sema::FunctionEmissionStatus Sema::getEmissionStatus(const FunctionDecl *FD,
         return FunctionEmissionStatus::OMPDiscarded;
     // If we have an explicit value for the device type, or we are in a target
     // declare context, we need to emit all extern and used symbols.
-    if (isInOpenMPDeclareTargetContext() || DevTy)
+    if (OpenMP().isInOpenMPDeclareTargetContext() || DevTy)
       if (IsEmittedForExternalSymbol())
         return FunctionEmissionStatus::Emitted;
     // Device mode only emits what it must, if it wasn't tagged yet and needed,
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 7669171fea56f..8c6bae545bfd1 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -44,6 +44,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
@@ -962,8 +963,8 @@ Sema::ActOnDecompositionDeclarator(Scope *S, Declarator &D,
     CurContext->addHiddenDecl(New);
   }
 
-  if (isInOpenMPDeclareTargetContext())
-    checkDeclIsAllowedInOpenMPTarget(nullptr, New);
+  if (OpenMP().isInOpenMPDeclareTargetContext())
+    OpenMP().checkDeclIsAllowedInOpenMPTarget(nullptr, New);
 
   return New;
 }
@@ -18654,8 +18655,8 @@ void Sema::MarkVTableUsed(SourceLocation Loc, CXXRecordDecl *Class,
   // Do not mark as used if compiling for the device outside of the target
   // region.
   if (TUKind != TU_Prefix && LangOpts.OpenMP && LangOpts.OpenMPIsTargetDevice &&
-      !isInOpenMPDeclareTargetContext() &&
-      !isInOpenMPTargetExecutionDirective()) {
+      !OpenMP().isInOpenMPDeclareTargetContext() &&
+      !OpenMP().isInOpenMPTargetExecutionDirective()) {
     if (!DefinitionRequired)
       MarkVirtualMembersReferenced(Loc, Class);
     return;
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 189764cb4b6b0..cabffa47c9318 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -52,6 +52,7 @@
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaFixItUtils.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/STLForwardCompat.h"
@@ -360,9 +361,9 @@ bool Sema::DiagnoseUseOfDecl(NamedDecl *D, ArrayRef<SourceLocation> Locs,
   //  at the same location.
   // [OpenMP 5.2] Also allow iterator declared variables.
   if (LangOpts.OpenMP && isa<VarDecl>(D) &&
-      !isOpenMPDeclareMapperVarDeclAllowed(cast<VarDecl>(D))) {
+      !OpenMP().isOpenMPDeclareMapperVarDeclAllowed(cast<VarDecl>(D))) {
     Diag(Loc, diag::err_omp_declare_mapper_wrong_var)
-        << getOpenMPDeclareMapperVarName();
+        << OpenMP().getOpenMPDeclareMapperVarName();
     Diag(D->getLocation(), diag::note_entity_declared_at) << D;
     return true;
   }
@@ -2267,7 +2268,7 @@ NonOdrUseReason Sema::getNonOdrUseReasonInCurrentContext(ValueDecl *D) {
   //   be loaded from the captured.
   if (VarDecl *VD = dyn_cast<VarDecl>(D)) {
     if (VD->getType()->isReferenceType() &&
-        !(getLangOpts().OpenMP && isOpenMPCapturedDecl(D)) &&
+        !(getLangOpts().OpenMP && OpenMP().isOpenMPCapturedDecl(D)) &&
         !isCapturingReferenceToHostVarInCUDADeviceLambda(*this, VD) &&
         VD->isUsableInConstantExpressions(Context))
       return NOUR_Constant;
@@ -5080,9 +5081,10 @@ ExprResult Sema::ActOnArraySubscriptExpr(Scope *S, Expr *base,
 
   if (base && !base->getType().isNull() &&
       base->hasPlaceholderType(BuiltinType::OMPArraySection))
-    return ActOnOMPArraySectionExpr(base, lbLoc, ArgExprs.front(), SourceLocation(),
-                                    SourceLocation(), /*Length*/ nullptr,
-                                    /*Stride=*/nullptr, rbLoc);
+    return OpenMP().ActOnOMPArraySectionExpr(base, lbLoc, ArgExprs.front(),
+                                             SourceLocation(), SourceLocation(),
+                                             /*Length*/ nullptr,
+                                             /*Stride=*/nullptr, rbLoc);
 
   // Since this might be a postfix expression, get rid of ParenListExprs.
   if (isa<ParenListExpr>(base)) {
@@ -5354,558 +5356,6 @@ void Sema::CheckSubscriptAccessOfNoDeref(const ArraySubscriptExpr *E) {
   }
 }
 
-ExprResult Sema::ActOnOMPArraySectionExpr(Expr *Base, SourceLocation LBLoc,
-                                          Expr *LowerBound,
-                                          SourceLocation ColonLocFirst,
-                                          SourceLocation ColonLocSecond,
-                                          Expr *Length, Expr *Stride,
-                                          SourceLocation RBLoc) {
-  if (Base->hasPlaceholderType() &&
-      !Base->hasPlaceholderType(BuiltinType::OMPArraySection)) {
-    ExprResult Result = CheckPlaceholderExpr(Base);
-    if (Result.isInvalid())
-      return ExprError();
-    Base = Result.get();
-  }
-  if (LowerBound && LowerBound->getType()->isNonOverloadPlaceholderType()) {
-    ExprResult Result = CheckPlaceholderExpr(LowerBound);
-    if (Result.isInvalid())
-      return ExprError();
-    Result = DefaultLvalueConversion(Result.get());
-    if (Result.isInvalid())
-      return ExprError();
-    LowerBound = Result.get();
-  }
-  if (Length && Length->getType()->isNonOverloadPlaceholderType()) {
-    ExprResult Result = CheckPlaceholderExpr(Length);
-    if (Result.isInvalid())
-      return ExprError();
-    Result = DefaultLvalueConversion(Result.get());
-    if (Result.isInvalid())
-      return ExprError();
-    Length = Result.get();
-  }
-  if (Stride && Stride->getType()->isNonOverloadPlaceholderType()) {
-    ExprResult Result = CheckPlaceholderExpr(Stride);
-    if (Result.isInvalid())
-      return ExprError();
-    Result = DefaultLvalueConversion(Result.get());
-    if (Result.isInvalid())
-      return ExprError();
-    Stride = Result.get();
-  }
-
-  // Build an unanalyzed expression if either operand is type-dependent.
-  if (Base->isTypeDependent() ||
-      (LowerBound &&
-       (LowerBound->isTypeDependent() || LowerBound->isValueDependent())) ||
-      (Length && (Length->isTypeDependent() || Length->isValueDependent())) ||
-      (Stride && (Stride->isTypeDependent() || Stride->isValueDependent()))) {
-    return new (Context) OMPArraySectionExpr(
-        Base, LowerBound, Length, Stride, Context.DependentTy, VK_LValue,
-        OK_Ordinary, ColonLocFirst, ColonLocSecond, RBLoc);
-  }
-
-  // Perform default conversions.
-  QualType OriginalTy = OMPArraySectionExpr::getBaseOriginalType(Base);
-  QualType ResultTy;
-  if (OriginalTy->isAnyPointerType()) {
-    ResultTy = OriginalTy->getPointeeType();
-  } else if (OriginalTy->isArrayType()) {
-    ResultTy = OriginalTy->getAsArrayTypeUnsafe()->getElementType();
-  } else {
-    return ExprError(
-        Diag(Base->getExprLoc(), diag::err_omp_typecheck_section_value)
-        << Base->getSourceRange());
-  }
-  // C99 6.5.2.1p1
-  if (LowerBound) {
-    auto Res = PerformOpenMPImplicitIntegerConversion(LowerBound->getExprLoc(),
-                                                      LowerBound);
-    if (Res.isInvalid())
-      return ExprError(Diag(LowerBound->getExprLoc(),
-                            diag::err_omp_typecheck_section_not_integer)
-                       << 0 << LowerBound->getSourceRange());
-    LowerBound = Res.get();
-
-    if (LowerBound->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
-        LowerBound->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
-      Diag(LowerBound->getExprLoc(), diag::warn_omp_section_is_char)
-          << 0 << LowerBound->getSourceRange();
-  }
-  if (Length) {
-    auto Res =
-        PerformOpenMPImplicitIntegerConversion(Length->getExprLoc(), Length);
-    if (Res.isInvalid())
-      return ExprError(Diag(Length->getExprLoc(),
-                            diag::err_omp_typecheck_section_not_integer)
-                       << 1 << Length->getSourceRange());
-    Length = Res.get();
-
-    if (Length->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
-        Length->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
-      Diag(Length->getExprLoc(), diag::warn_omp_section_is_char)
-          << 1 << Length->getSourceRange();
-  }
-  if (Stride) {
-    ExprResult Res =
-        PerformOpenMPImplicitIntegerConversion(Stride->getExprLoc(), Stride);
-    if (Res.isInvalid())
-      return ExprError(Diag(Stride->getExprLoc(),
-                            diag::err_omp_typecheck_section_not_integer)
-                       << 1 << Stride->getSourceRange());
-    Stride = Res.get();
-
-    if (Stride->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
-        Stride->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
-      Diag(Stride->getExprLoc(), diag::warn_omp_section_is_char)
-          << 1 << Stride->getSourceRange();
-  }
-
-  // C99 6.5.2.1p1: "shall have type "pointer to *object* type". Similarly,
-  // C++ [expr.sub]p1: The type "T" shall be a completely-defined object
-  // type. Note that functions are not objects, and that (in C99 parlance)
-  // incomplete types are not object types.
-  if (ResultTy->isFunctionType()) {
-    Diag(Base->getExprLoc(), diag::err_omp_section_function_type)
-        << ResultTy << Base->getSourceRange();
-    return ExprError();
-  }
-
-  if (RequireCompleteType(Base->getExprLoc(), ResultTy,
-                          diag::err_omp_section_incomplete_type, Base))
-    return ExprError();
-
-  if (LowerBound && !OriginalTy->isAnyPointerType()) {
-    Expr::EvalResult Result;
-    if (LowerBound->EvaluateAsInt(Result, Context)) {
-      // OpenMP 5.0, [2.1.5 Array Sections]
-      // The array section must be a subset of the original array.
-      llvm::APSInt LowerBoundValue = Result.Val.getInt();
-      if (LowerBoundValue.isNegative()) {
-        Diag(LowerBound->getExprLoc(), diag::err_omp_section_not_subset_of_array)
-            << LowerBound->getSourceRange();
-        return ExprError();
-      }
-    }
-  }
-
-  if (Length) {
-    Expr::EvalResult Result;
-    if (Length->EvaluateAsInt(Result, Context)) {
-      // OpenMP 5.0, [2.1.5 Array Sections]
-      // The length must evaluate to non-negative integers.
-      llvm::APSInt LengthValue = Result.Val.getInt();
-      if (LengthValue.isNegative()) {
-        Diag(Length->getExprLoc(), diag::err_omp_section_length_negative)
-            << toString(LengthValue, /*Radix=*/10, /*Signed=*/true)
-            << Length->getSourceRange();
-        return ExprError();
-      }
-    }
-  } else if (ColonLocFirst.isValid() &&
-             (OriginalTy.isNull() || (!OriginalTy->isConstantArrayType() &&
-                                      !OriginalTy->isVariableArrayType()))) {
-    // OpenMP 5.0, [2.1.5 Array Sections]
-    // When the size of the array dimension is not known, the length must be
-    // specified explicitly.
-    Diag(ColonLocFirst, diag::err_omp_section_length_undefined)
-        << (!OriginalTy.isNull() && OriginalTy->isArrayType());
-    return ExprError();
-  }
-
-  if (Stride) {
-    Expr::EvalResult Result;
-    if (Stride->EvaluateAsInt(Result, Context)) {
-      // OpenMP 5.0, [2.1.5 Array Sections]
-      // The stride must evaluate to a positive integer.
-      llvm::APSInt StrideValue = Result.Val.getInt();
-      if (!StrideValue.isStrictlyPositive()) {
-        Diag(Stride->getExprLoc(), diag::err_omp_section_stride_non_positive)
-            << toString(StrideValue, /*Radix=*/10, /*Signed=*/true)
-            << Stride->getSourceRange();
-        return ExprError();
-      }
-    }
-  }
-
-  if (!Base->hasPlaceholderType(BuiltinType::OMPArraySection)) {
-    ExprResult Result = DefaultFunctionArrayLvalueConversion(Base);
-    if (Result.isInvalid())
-      return ExprError();
-    Base = Result.get();
-  }
-  return new (Context) OMPArraySectionExpr(
-      Base, LowerBound, Length, Stride, Context.OMPArraySectionTy, VK_LValue,
-      OK_Ordinary, ColonLocFirst, ColonLocSecond, RBLoc);
-}
-
-ExprResult Sema::ActOnOMPArrayShapingExpr(Expr *Base, SourceLocation LParenLoc,
-                                          SourceLocation RParenLoc,
-                                          ArrayRef<Expr *> Dims,
-                                          ArrayRef<SourceRange> Brackets) {
-  if (Base->hasPlaceholderType()) {
-    ExprResult Result = CheckPlaceholderExpr(Base);
-    if (Result.isInvalid())
-      return ExprError();
-    Result = DefaultLvalueConversion(Result.get());
-    if (Result.isInvalid())
-      return ExprError();
-    Base = Result.get();
-  }
-  QualType BaseTy = Base->getType();
-  // Delay analysis of the types/expressions if instantiation/specialization is
-  // required.
-  if (!BaseTy->isPointerType() && Base->isTypeDependent())
-    return OMPArrayShapingExpr::Create(Context, Context.DependentTy, Base,
-                                       LParenLoc, RParenLoc, Dims, Brackets);
-  if (!BaseTy->isPointerType() ||
-      (!Base->isTypeDependent() &&
-       BaseTy->getPointeeType()->isIncompleteType()))
-    return ExprError(Diag(Base->getExprLoc(),
-                          diag::err_omp_non_pointer_type_array_shaping_base)
-                     << Base->getSourceRange());
-
-  SmallVector<Expr *, 4> NewDims;
-  bool ErrorFound = false;
-  for (Expr *Dim : Dims) {
-    if (Dim->hasPlaceholderType()) {
-      ExprResult Result = CheckPlaceholderExpr(Dim);
-      if (Result.isInvalid()) {
-        ErrorFound = true;
-        continue;
-      }
-      Result = DefaultLvalueConversion(Result.get());
-      if (Result.isInvalid()) {
-        ErrorFound = true;
-        continue;
-      }
-      Dim = Result.get();
-    }
-    if (!Dim->isTypeDependent()) {
-      ExprResult Result =
-          PerformOpenMPImplicitIntegerConversion(Dim->getExprLoc(), Dim);
-      if (Result.isInvalid()) {
-        ErrorFound = true;
-        Diag(Dim->getExprLoc(), diag::err_omp_typecheck_shaping_not_integer)
-            << Dim->getSourceRange();
-        continue;
-      }
-      Dim = Result.get();
-      Expr::EvalResult EvResult;
-      if (!Dim->isValueDependent() && Dim->EvaluateAsInt(EvResult, Context)) {
-        // OpenMP 5.0, [2.1.4 Array Shaping]
-        // Each si is an integral type expression that must evaluate to a
-        // positive integer.
-        llvm::APSInt Value = EvResult.Val.getInt();
-        if (!Value.isStrictlyPositive()) {
-          Diag(Dim->getExprLoc(), diag::err_omp_shaping_dimension_not_positive)
-              << toString(Value, /*Radix=*/10, /*Signed=*/true)
-              << Dim->getSourceRange();
-          ErrorFound = true;
-          continue;
-        }
-      }
-    }
-    NewDims.push_back(Dim);
-  }
-  if (ErrorFound)
-    return ExprError();
-  return OMPArrayShapingExpr::Create(Context, Context.OMPArrayShapingTy, Base,
-                                     LParenLoc, RParenLoc, NewDims, Brackets);
-}
-
-ExprResult Sema::ActOnOMPIteratorExpr(Scope *S, SourceLocation IteratorKwLoc,
-                                      SourceLocation LLoc, SourceLocation RLoc,
-                                      ArrayRef<OMPIteratorData> Data) {
-  SmallVector<OMPIteratorExpr::IteratorDefinition, 4> ID;
-  bool IsCorrect = true;
-  for (const OMPIteratorData &D : Data) {
-    TypeSourceInfo *TInfo = nullptr;
-    SourceLocation StartLoc;
-    QualType DeclTy;
-    if (!D.Type.getAsOpaquePtr()) {
-      // OpenMP 5.0, 2.1.6 Iterators
-      // In an iterator-specifier, if the iterator-type is not specified then
-      // the type of that iterator is of int type.
-      DeclTy = Context.IntTy;
-      StartLoc = D.DeclIdentLoc;
-    } else {
-      DeclTy = GetTypeFromParser(D.Type, &TInfo);
-      StartLoc = TInfo->getTypeLoc().getBeginLoc();
-    }
-
-    bool IsDeclTyDependent = DeclTy->isDependentType() ||
-                             DeclTy->containsUnexpandedParameterPack() ||
-                             DeclTy->isInstantiationDependentType();
-    if (!IsDeclTyDependent) {
-      if (!DeclTy->isIntegralType(Context) && !DeclTy->isAnyPointerType()) {
-        // OpenMP 5.0, 2.1.6 Iterators, Restrictions, C/C++
-        // The iterator-type must be an integral or pointer type.
-        Diag(StartLoc, diag::err_omp_iterator_not_integral_or_pointer)
-            << DeclTy;
-        IsCorrect = false;
-        continue;
-      }
-      if (DeclTy.isConstant(Context)) {
-        // OpenMP 5.0, 2.1.6 Iterators, Restrictions, C/C++
-        // The iterator-type must not be const qualified.
-        Diag(StartLoc, diag::err_omp_iterator_not_integral_or_pointer)
-            << DeclTy;
-        IsCorrect = false;
-        continue;
-      }
-    }
-
-    // Iterator declaration.
-    assert(D.DeclIdent && "Identifier expected.");
-    // Always try to create iterator declarator to avoid extra error messages
-    // about unknown declarations use.
-    auto *VD = VarDecl::Create(Context, CurContext, StartLoc, D.DeclIdentLoc,
-                               D.DeclIdent, DeclTy, TInfo, SC_None);
-    VD->setImplicit();
-    if (S) {
-      // Check for conflicting previous declaration.
-      DeclarationNameInfo NameInfo(VD->getDeclName(), D.DeclIdentLoc);
-      LookupResult Previous(*this, NameInfo, LookupOrdinaryName,
-                            ForVisibleRedeclaration);
-      Previous.suppressDiagnostics();
-      LookupName(Previous, S);
-
-      FilterLookupForScope(Previous, CurContext, S, /*ConsiderLinkage=*/false,
-                           /*AllowInlineNamespace=*/false);
-      if (!Previous.empty()) {
-        NamedDecl *Old = Previous.getRepresentativeDecl();
-        Diag(D.DeclIdentLoc, diag::err_redefinition) << VD->getDeclName();
-        Diag(Old->getLocation(), diag::note_previous_definition);
-      } else {
-        PushOnScopeChains(VD, S);
-      }
-    } else {
-      CurContext->addDecl(VD);
-    }
-
-    /// Act on the iterator variable declaration.
-    ActOnOpenMPIteratorVarDecl(VD);
-
-    Expr *Begin = D.Range.Begin;
-    if (!IsDeclTyDependent && Begin && !Begin->isTypeDependent()) {
-      ExprResult BeginRes =
-          PerformImplicitConversion(Begin, DeclTy, AA_Converting);
-      Begin = BeginRes.get();
-    }
-    Expr *End = D.Range.End;
-    if (!IsDeclTyDependent && End && !End->isTypeDependent()) {
-      ExprResult EndRes = PerformImplicitConversion(End, DeclTy, AA_Converting);
-      End = EndRes.get();
-    }
-    Expr *Step = D.Range.Step;
-    if (!IsDeclTyDependent && Step && !Step->isTypeDependent()) {
-      if (!Step->getType()->isIntegralType(Context)) {
-        Diag(Step->getExprLoc(), diag::err_omp_iterator_step_not_integral)
-            << Step << Step->getSourceRange();
-        IsCorrect = false;
-        continue;
-      }
-      std::optional<llvm::APSInt> Result =
-          Step->getIntegerConstantExpr(Context);
-      // OpenMP 5.0, 2.1.6 Iterators, Restrictions
-      // If the step expression of a range-specification equals zero, the
-      // behavior is unspecified.
-      if (Result && Result->isZero()) {
-        Diag(Step->getExprLoc(), diag::err_omp_iterator_step_constant_zero)
-            << Step << Step->getSourceRange();
-        IsCorrect = false;
-        continue;
-      }
-    }
-    if (!Begin || !End || !IsCorrect) {
-      IsCorrect = false;
-      continue;
-    }
-    OMPIteratorExpr::IteratorDefinition &IDElem = ID.emplace_back();
-    IDElem.IteratorDecl = VD;
-    IDElem.AssignmentLoc = D.AssignLoc;
-    IDElem.Range.Begin = Begin;
-    IDElem.Range.End = End;
-    IDElem.Range.Step = Step;
-    IDElem.ColonLoc = D.ColonLoc;
-    IDElem.SecondColonLoc = D.SecColonLoc;
-  }
-  if (!IsCorrect) {
-    // Invalidate all created iterator declarations if error is found.
-    for (const OMPIteratorExpr::IteratorDefinition &D : ID) {
-      if (Decl *ID = D.IteratorDecl)
-        ID->setInvalidDecl();
-    }
-    return ExprError();
-  }
-  SmallVector<OMPIteratorHelperData, 4> Helpers;
-  if (!CurContext->isDependentContext()) {
-    // Build number of ityeration for each iteration range.
-    // Ni = ((Stepi > 0) ? ((Endi + Stepi -1 - Begini)/Stepi) :
-    // ((Begini-Stepi-1-Endi) / -Stepi);
-    for (OMPIteratorExpr::IteratorDefinition &D : ID) {
-      // (Endi - Begini)
-      ExprResult Res = CreateBuiltinBinOp(D.AssignmentLoc, BO_Sub, D.Range.End,
-                                          D.Range.Begin);
-      if(!Res.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      ExprResult St, St1;
-      if (D.Range.Step) {
-        St = D.Range.Step;
-        // (Endi - Begini) + Stepi
-        Res = CreateBuiltinBinOp(D.AssignmentLoc, BO_Add, Res.get(), St.get());
-        if (!Res.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // (Endi - Begini) + Stepi - 1
-        Res =
-            CreateBuiltinBinOp(D.AssignmentLoc, BO_Sub, Res.get(),
-                               ActOnIntegerConstant(D.AssignmentLoc, 1).get());
-        if (!Res.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // ((Endi - Begini) + Stepi - 1) / Stepi
-        Res = CreateBuiltinBinOp(D.AssignmentLoc, BO_Div, Res.get(), St.get());
-        if (!Res.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        St1 = CreateBuiltinUnaryOp(D.AssignmentLoc, UO_Minus, D.Range.Step);
-        // (Begini - Endi)
-        ExprResult Res1 = CreateBuiltinBinOp(D.AssignmentLoc, BO_Sub,
-                                             D.Range.Begin, D.Range.End);
-        if (!Res1.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // (Begini - Endi) - Stepi
-        Res1 =
-            CreateBuiltinBinOp(D.AssignmentLoc, BO_Add, Res1.get(), St1.get());
-        if (!Res1.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // (Begini - Endi) - Stepi - 1
-        Res1 =
-            CreateBuiltinBinOp(D.AssignmentLoc, BO_Sub, Res1.get(),
-                               ActOnIntegerConstant(D.AssignmentLoc, 1).get());
-        if (!Res1.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // ((Begini - Endi) - Stepi - 1) / (-Stepi)
-        Res1 =
-            CreateBuiltinBinOp(D.AssignmentLoc, BO_Div, Res1.get(), St1.get());
-        if (!Res1.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        // Stepi > 0.
-        ExprResult CmpRes =
-            CreateBuiltinBinOp(D.AssignmentLoc, BO_GT, D.Range.Step,
-                               ActOnIntegerConstant(D.AssignmentLoc, 0).get());
-        if (!CmpRes.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-        Res = ActOnConditionalOp(D.AssignmentLoc, D.AssignmentLoc, CmpRes.get(),
-                                 Res.get(), Res1.get());
-        if (!Res.isUsable()) {
-          IsCorrect = false;
-          continue;
-        }
-      }
-      Res = ActOnFinishFullExpr(Res.get(), /*DiscardedValue=*/false);
-      if (!Res.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-
-      // Build counter update.
-      // Build counter.
-      auto *CounterVD =
-          VarDecl::Create(Context, CurContext, D.IteratorDecl->getBeginLoc(),
-                          D.IteratorDecl->getBeginLoc(), nullptr,
-                          Res.get()->getType(), nullptr, SC_None);
-      CounterVD->setImplicit();
-      ExprResult RefRes =
-          BuildDeclRefExpr(CounterVD, CounterVD->getType(), VK_LValue,
-                           D.IteratorDecl->getBeginLoc());
-      // Build counter update.
-      // I = Begini + counter * Stepi;
-      ExprResult UpdateRes;
-      if (D.Range.Step) {
-        UpdateRes = CreateBuiltinBinOp(
-            D.AssignmentLoc, BO_Mul,
-            DefaultLvalueConversion(RefRes.get()).get(), St.get());
-      } else {
-        UpdateRes = DefaultLvalueConversion(RefRes.get());
-      }
-      if (!UpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      UpdateRes = CreateBuiltinBinOp(D.AssignmentLoc, BO_Add, D.Range.Begin,
-                                     UpdateRes.get());
-      if (!UpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      ExprResult VDRes =
-          BuildDeclRefExpr(cast<VarDecl>(D.IteratorDecl),
-                           cast<VarDecl>(D.IteratorDecl)->getType(), VK_LValue,
-                           D.IteratorDecl->getBeginLoc());
-      UpdateRes = CreateBuiltinBinOp(D.AssignmentLoc, BO_Assign, VDRes.get(),
-                                     UpdateRes.get());
-      if (!UpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      UpdateRes =
-          ActOnFinishFullExpr(UpdateRes.get(), /*DiscardedValue=*/true);
-      if (!UpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      ExprResult CounterUpdateRes =
-          CreateBuiltinUnaryOp(D.AssignmentLoc, UO_PreInc, RefRes.get());
-      if (!CounterUpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      CounterUpdateRes =
-          ActOnFinishFullExpr(CounterUpdateRes.get(), /*DiscardedValue=*/true);
-      if (!CounterUpdateRes.isUsable()) {
-        IsCorrect = false;
-        continue;
-      }
-      OMPIteratorHelperData &HD = Helpers.emplace_back();
-      HD.CounterVD = CounterVD;
-      HD.Upper = Res.get();
-      HD.Update = UpdateRes.get();
-      HD.CounterUpdate = CounterUpdateRes.get();
-    }
-  } else {
-    Helpers.assign(ID.size(), {});
-  }
-  if (!IsCorrect) {
-    // Invalidate all created iterator declarations if error is found.
-    for (const OMPIteratorExpr::IteratorDefinition &D : ID) {
-      if (Decl *ID = D.IteratorDecl)
-        ID->setInvalidDecl();
-    }
-    return ExprError();
-  }
-  return OMPIteratorExpr::Create(Context, Context.OMPIteratorTy, IteratorKwLoc,
-                                 LLoc, RLoc, ID, Helpers);
-}
-
 ExprResult
 Sema::CreateBuiltinArraySubscriptExpr(Expr *Base, SourceLocation LLoc,
                                       Expr *Idx, SourceLocation RLoc) {
@@ -7190,8 +6640,8 @@ ExprResult Sema::ActOnCallExpr(Scope *Scope, Expr *Fn, SourceLocation LParenLoc,
   }
 
   if (LangOpts.OpenMP)
-    Call = ActOnOpenMPCall(Call, Scope, LParenLoc, ArgExprs, RParenLoc,
-                           ExecConfig);
+    Call = OpenMP().ActOnOpenMPCall(Call, Scope, LParenLoc, ArgExprs, RParenLoc,
+                                    ExecConfig);
   if (LangOpts.CPlusPlus) {
     if (const auto *CE = dyn_cast<CallExpr>(Call.get()))
       DiagnosedUnqualifiedCallsToStdFunctions(*this, CE);
@@ -19193,7 +18643,7 @@ MarkVarDeclODRUsed(ValueDecl *V, SourceLocation Loc, Sema &SemaRef,
   }
   QualType CaptureType, DeclRefType;
   if (SemaRef.LangOpts.OpenMP)
-    SemaRef.tryCaptureOpenMPLambdas(V);
+    SemaRef.OpenMP().tryCaptureOpenMPLambdas(V);
   SemaRef.tryCaptureVariable(V, Loc, Sema::TryCapture_Implicit,
                              /*EllipsisLoc*/ SourceLocation(),
                              /*BuildAndDiagnose*/ true, CaptureType,
@@ -19474,7 +18924,7 @@ static bool captureInBlock(BlockScopeInfo *BSI, ValueDecl *Var,
 
   const bool HasBlocksAttr = Var->hasAttr<BlocksAttr>();
   if (HasBlocksAttr || CaptureType->isReferenceType() ||
-      (S.getLangOpts().OpenMP && S.isOpenMPCapturedDecl(Var))) {
+      (S.getLangOpts().OpenMP && S.OpenMP().isOpenMPCapturedDecl(Var))) {
     // Block capture by reference does not change the capture or
     // declaration reference types.
     ByRef = true;
@@ -19504,7 +18954,7 @@ static bool captureInCapturedRegion(
     ByRef = (Kind == Sema::TryCapture_ExplicitByRef);
   } else if (S.getLangOpts().OpenMP && RSI->CapRegionKind == CR_OpenMP) {
     // Using an LValue reference type is consistent with Lambdas (see below).
-    if (S.isOpenMPCapturedDecl(Var)) {
+    if (S.OpenMP().isOpenMPCapturedDecl(Var)) {
       bool HasConst = DeclRefType.isConstQualified();
       DeclRefType = DeclRefType.getUnqualifiedType();
       // Don't lose diagnostics about assignments to const.
@@ -19512,11 +18962,11 @@ static bool captureInCapturedRegion(
         DeclRefType.addConst();
     }
     // Do not capture firstprivates in tasks.
-    if (S.isOpenMPPrivateDecl(Var, RSI->OpenMPLevel, RSI->OpenMPCaptureLevel) !=
-        OMPC_unknown)
+    if (S.OpenMP().isOpenMPPrivateDecl(Var, RSI->OpenMPLevel,
+                                       RSI->OpenMPCaptureLevel) != OMPC_unknown)
       return true;
-    ByRef = S.isOpenMPCapturedByRef(Var, RSI->OpenMPLevel,
-                                    RSI->OpenMPCaptureLevel);
+    ByRef = S.OpenMP().isOpenMPCapturedByRef(Var, RSI->OpenMPLevel,
+                                             RSI->OpenMPCaptureLevel);
   }
 
   if (ByRef)
@@ -19777,9 +19227,9 @@ bool Sema::tryCaptureVariable(
   // Capture global variables if it is required to use private copy of this
   // variable.
   bool IsGlobal = !VD->hasLocalStorage();
-  if (IsGlobal &&
-      !(LangOpts.OpenMP && isOpenMPCapturedDecl(Var, /*CheckScopeInfo=*/true,
-                                                MaxFunctionScopesIndex)))
+  if (IsGlobal && !(LangOpts.OpenMP &&
+                    OpenMP().isOpenMPCapturedDecl(Var, /*CheckScopeInfo=*/true,
+                                                  MaxFunctionScopesIndex)))
     return true;
 
   if (isa<VarDecl>(Var))
@@ -19897,7 +19347,7 @@ bool Sema::tryCaptureVariable(
             }
             return true;
           }
-          OpenMPClauseKind IsOpenMPPrivateDecl = isOpenMPPrivateDecl(
+          OpenMPClauseKind IsOpenMPPrivateDecl = OpenMP().isOpenMPPrivateDecl(
               Var, RSI->OpenMPLevel, RSI->OpenMPCaptureLevel);
           // If the variable is private (i.e. not captured) and has variably
           // modified type, we still need to capture the type for correct
@@ -19908,7 +19358,8 @@ bool Sema::tryCaptureVariable(
             QualType QTy = Var->getType();
             if (ParmVarDecl *PVD = dyn_cast_or_null<ParmVarDecl>(Var))
               QTy = PVD->getOriginalType();
-            for (int I = 1, E = getNumberOfConstructScopes(RSI->OpenMPLevel);
+            for (int I = 1,
+                     E = OpenMP().getNumberOfConstructScopes(RSI->OpenMPLevel);
                  I < E; ++I) {
               auto *OuterRSI = cast<CapturedRegionScopeInfo>(
                   FunctionScopes[FunctionScopesIndex - I]);
@@ -19920,18 +19371,19 @@ bool Sema::tryCaptureVariable(
           }
           bool IsTargetCap =
               IsOpenMPPrivateDecl != OMPC_private &&
-              isOpenMPTargetCapturedDecl(Var, RSI->OpenMPLevel,
-                                         RSI->OpenMPCaptureLevel);
+              OpenMP().isOpenMPTargetCapturedDecl(Var, RSI->OpenMPLevel,
+                                                  RSI->OpenMPCaptureLevel);
           // Do not capture global if it is not privatized in outer regions.
           bool IsGlobalCap =
-              IsGlobal && isOpenMPGlobalCapturedDecl(Var, RSI->OpenMPLevel,
-                                                     RSI->OpenMPCaptureLevel);
+              IsGlobal && OpenMP().isOpenMPGlobalCapturedDecl(
+                              Var, RSI->OpenMPLevel, RSI->OpenMPCaptureLevel);
 
           // When we detect target captures we are looking from inside the
           // target region, therefore we need to propagate the capture from the
           // enclosing region. Therefore, the capture is not initially nested.
           if (IsTargetCap)
-            adjustOpenMPTargetScopeIndex(FunctionScopesIndex, RSI->OpenMPLevel);
+            OpenMP().adjustOpenMPTargetScopeIndex(FunctionScopesIndex,
+                                                  RSI->OpenMPLevel);
 
           if (IsTargetCap || IsOpenMPPrivateDecl == OMPC_private ||
               (IsGlobal && !IsGlobalCap)) {
@@ -20753,8 +20205,8 @@ static void
 MarkExprReferenced(Sema &SemaRef, SourceLocation Loc, Decl *D, Expr *E,
                    bool MightBeOdrUse,
                    llvm::DenseMap<const VarDecl *, int> &RefsMinusAssignments) {
-  if (SemaRef.isInOpenMPDeclareTargetContext())
-    SemaRef.checkDeclIsAllowedInOpenMPTarget(E, D);
+  if (SemaRef.OpenMP().isInOpenMPDeclareTargetContext())
+    SemaRef.OpenMP().checkDeclIsAllowedInOpenMPTarget(E, D);
 
   if (VarDecl *Var = dyn_cast<VarDecl>(D)) {
     DoMarkVarDeclReferenced(SemaRef, Loc, Var, E, RefsMinusAssignments);
diff --git a/clang/lib/Sema/SemaExprMember.cpp b/clang/lib/Sema/SemaExprMember.cpp
index 32998ae60eafe..7ea6d733fe5a2 100644
--- a/clang/lib/Sema/SemaExprMember.cpp
+++ b/clang/lib/Sema/SemaExprMember.cpp
@@ -9,7 +9,6 @@
 //  This file implements semantic analysis member access expressions.
 //
 //===----------------------------------------------------------------------===//
-#include "clang/Sema/Overload.h"
 #include "clang/AST/ASTLambda.h"
 #include "clang/AST/DeclCXX.h"
 #include "clang/AST/DeclObjC.h"
@@ -18,9 +17,11 @@
 #include "clang/AST/ExprObjC.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/Lookup.h"
+#include "clang/Sema/Overload.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 
 using namespace clang;
 using namespace sema;
@@ -1900,9 +1901,9 @@ Sema::BuildFieldReferenceExpr(Expr *BaseExpr, bool IsArrow,
   if (getLangOpts().OpenMP && IsArrow &&
       !CurContext->isDependentContext() &&
       isa<CXXThisExpr>(Base.get()->IgnoreParenImpCasts())) {
-    if (auto *PrivateCopy = isOpenMPCapturedDecl(Field)) {
-      return getOpenMPCapturedExpr(PrivateCopy, VK, OK,
-                                   MemberNameInfo.getLoc());
+    if (auto *PrivateCopy = OpenMP().isOpenMPCapturedDecl(Field)) {
+      return OpenMP().getOpenMPCapturedExpr(PrivateCopy, VK, OK,
+                                            MemberNameInfo.getLoc());
     }
   }
 
diff --git a/clang/lib/Sema/SemaLambda.cpp b/clang/lib/Sema/SemaLambda.cpp
index 35a51c6c2328d..1743afaf15287 100644
--- a/clang/lib/Sema/SemaLambda.cpp
+++ b/clang/lib/Sema/SemaLambda.cpp
@@ -21,6 +21,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "llvm/ADT/STLExtras.h"
 #include <optional>
@@ -1398,7 +1399,7 @@ void Sema::ActOnStartOfLambdaDefinition(LambdaIntroducer &Intro,
 
   // OpenMP lambdas might get assumumption attributes.
   if (LangOpts.OpenMP)
-    ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Method);
+    OpenMP().ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Method);
 
   handleLambdaNumbering(Class, Method);
 
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index e9efb4721133f..d229ef650bccb 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -11,6 +11,7 @@
 ///
 //===----------------------------------------------------------------------===//
 
+#include "clang/Sema/SemaOpenMP.h"
 #include "TreeTransform.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/ASTMutationListener.h"
@@ -33,6 +34,7 @@
 #include "clang/Sema/ParsedAttr.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/ScopeInfo.h"
+#include "clang/Sema/Sema.h"
 #include "clang/Sema/SemaInternal.h"
 #include "llvm/ADT/IndexedMap.h"
 #include "llvm/ADT/PointerEmbeddedInt.h"
@@ -1808,9 +1810,9 @@ const DSAStackTy::DSAVarData DSAStackTy::getTopDSA(ValueDecl *D,
           return DVar;
         }
         const_iterator End = end();
-        if (!SemaRef.isOpenMPCapturedByRef(D,
-                                           std::distance(ParentIterTarget, End),
-                                           /*OpenMPCaptureLevel=*/0)) {
+        if (!SemaRef.OpenMP().isOpenMPCapturedByRef(
+                D, std::distance(ParentIterTarget, End),
+                /*OpenMPCaptureLevel=*/0)) {
           DVar.RefExpr =
               buildDeclRefExpr(SemaRef, VD, D->getType().getNonReferenceType(),
                                IterTarget->ConstructLoc);
@@ -2018,22 +2020,22 @@ bool DSAStackTy::hasDirective(
   return false;
 }
 
-void Sema::InitDataSharingAttributesStack() {
-  VarDataSharingAttributesStack = new DSAStackTy(*this);
+void SemaOpenMP::InitDataSharingAttributesStack() {
+  VarDataSharingAttributesStack = new DSAStackTy(SemaRef);
 }
 
 #define DSAStack static_cast<DSAStackTy *>(VarDataSharingAttributesStack)
 
-void Sema::pushOpenMPFunctionRegion() { DSAStack->pushFunction(); }
+void SemaOpenMP::pushOpenMPFunctionRegion() { DSAStack->pushFunction(); }
 
-void Sema::popOpenMPFunctionRegion(const FunctionScopeInfo *OldFSI) {
+void SemaOpenMP::popOpenMPFunctionRegion(const FunctionScopeInfo *OldFSI) {
   DSAStack->popFunction(OldFSI);
 }
 
 static bool isOpenMPDeviceDelayedContext(Sema &S) {
   assert(S.LangOpts.OpenMP && S.LangOpts.OpenMPIsTargetDevice &&
          "Expected OpenMP device compilation.");
-  return !S.isInOpenMPTargetExecutionDirective();
+  return !S.OpenMP().isInOpenMPTargetExecutionDirective();
 }
 
 namespace {
@@ -2045,20 +2047,20 @@ enum class FunctionEmissionStatus {
 };
 } // anonymous namespace
 
-Sema::SemaDiagnosticBuilder
-Sema::diagIfOpenMPDeviceCode(SourceLocation Loc, unsigned DiagID,
-                             const FunctionDecl *FD) {
-  assert(LangOpts.OpenMP && LangOpts.OpenMPIsTargetDevice &&
+SemaBase::SemaDiagnosticBuilder
+SemaOpenMP::diagIfOpenMPDeviceCode(SourceLocation Loc, unsigned DiagID,
+                                   const FunctionDecl *FD) {
+  assert(getLangOpts().OpenMP && getLangOpts().OpenMPIsTargetDevice &&
          "Expected OpenMP device compilation.");
 
   SemaDiagnosticBuilder::Kind Kind = SemaDiagnosticBuilder::K_Nop;
   if (FD) {
-    FunctionEmissionStatus FES = getEmissionStatus(FD);
+    Sema::FunctionEmissionStatus FES = SemaRef.getEmissionStatus(FD);
     switch (FES) {
-    case FunctionEmissionStatus::Emitted:
+    case Sema::FunctionEmissionStatus::Emitted:
       Kind = SemaDiagnosticBuilder::K_Immediate;
       break;
-    case FunctionEmissionStatus::Unknown:
+    case Sema::FunctionEmissionStatus::Unknown:
       // TODO: We should always delay diagnostics here in case a target
       //       region is in a function we do not emit. However, as the
       //       current diagnostics are associated with the function containing
@@ -2066,48 +2068,48 @@ Sema::diagIfOpenMPDeviceCode(SourceLocation Loc, unsigned DiagID,
       //       on diagnostics for the target region itself. We need to anchor
       //       the diagnostics with the new generated function *or* ensure we
       //       emit diagnostics associated with the surrounding function.
-      Kind = isOpenMPDeviceDelayedContext(*this)
+      Kind = isOpenMPDeviceDelayedContext(SemaRef)
                  ? SemaDiagnosticBuilder::K_Deferred
                  : SemaDiagnosticBuilder::K_Immediate;
       break;
-    case FunctionEmissionStatus::TemplateDiscarded:
-    case FunctionEmissionStatus::OMPDiscarded:
+    case Sema::FunctionEmissionStatus::TemplateDiscarded:
+    case Sema::FunctionEmissionStatus::OMPDiscarded:
       Kind = SemaDiagnosticBuilder::K_Nop;
       break;
-    case FunctionEmissionStatus::CUDADiscarded:
+    case Sema::FunctionEmissionStatus::CUDADiscarded:
       llvm_unreachable("CUDADiscarded unexpected in OpenMP device compilation");
       break;
     }
   }
 
-  return SemaDiagnosticBuilder(Kind, Loc, DiagID, FD, *this);
+  return SemaDiagnosticBuilder(Kind, Loc, DiagID, FD, SemaRef);
 }
 
-Sema::SemaDiagnosticBuilder Sema::diagIfOpenMPHostCode(SourceLocation Loc,
-                                                       unsigned DiagID,
-                                                       const FunctionDecl *FD) {
-  assert(LangOpts.OpenMP && !LangOpts.OpenMPIsTargetDevice &&
+SemaBase::SemaDiagnosticBuilder
+SemaOpenMP::diagIfOpenMPHostCode(SourceLocation Loc, unsigned DiagID,
+                                 const FunctionDecl *FD) {
+  assert(getLangOpts().OpenMP && !getLangOpts().OpenMPIsTargetDevice &&
          "Expected OpenMP host compilation.");
 
   SemaDiagnosticBuilder::Kind Kind = SemaDiagnosticBuilder::K_Nop;
   if (FD) {
-    FunctionEmissionStatus FES = getEmissionStatus(FD);
+    Sema::FunctionEmissionStatus FES = SemaRef.getEmissionStatus(FD);
     switch (FES) {
-    case FunctionEmissionStatus::Emitted:
+    case Sema::FunctionEmissionStatus::Emitted:
       Kind = SemaDiagnosticBuilder::K_Immediate;
       break;
-    case FunctionEmissionStatus::Unknown:
+    case Sema::FunctionEmissionStatus::Unknown:
       Kind = SemaDiagnosticBuilder::K_Deferred;
       break;
-    case FunctionEmissionStatus::TemplateDiscarded:
-    case FunctionEmissionStatus::OMPDiscarded:
-    case FunctionEmissionStatus::CUDADiscarded:
+    case Sema::FunctionEmissionStatus::TemplateDiscarded:
+    case Sema::FunctionEmissionStatus::OMPDiscarded:
+    case Sema::FunctionEmissionStatus::CUDADiscarded:
       Kind = SemaDiagnosticBuilder::K_Nop;
       break;
     }
   }
 
-  return SemaDiagnosticBuilder(Kind, Loc, DiagID, FD, *this);
+  return SemaDiagnosticBuilder(Kind, Loc, DiagID, FD, SemaRef);
 }
 
 static OpenMPDefaultmapClauseKind
@@ -2124,9 +2126,9 @@ getVariableCategoryFromDecl(const LangOptions &LO, const ValueDecl *VD) {
   return OMPC_DEFAULTMAP_aggregate;
 }
 
-bool Sema::isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
-                                 unsigned OpenMPCaptureLevel) const {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+bool SemaOpenMP::isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
+                                       unsigned OpenMPCaptureLevel) const {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
 
   ASTContext &Ctx = getASTContext();
   bool IsByRef = true;
@@ -2252,7 +2254,7 @@ bool Sema::isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
                  !Ty->isAnyPointerType()) ||
                 !Ty->isScalarType() ||
                 DSAStack->isDefaultmapCapturedByRef(
-                    Level, getVariableCategoryFromDecl(LangOpts, D)) ||
+                    Level, getVariableCategoryFromDecl(getLangOpts(), D)) ||
                 DSAStack->hasExplicitDSA(
                     D,
                     [](OpenMPClauseKind K, bool AppliedToPointee) {
@@ -2303,17 +2305,17 @@ bool Sema::isOpenMPCapturedByRef(const ValueDecl *D, unsigned Level,
   return IsByRef;
 }
 
-unsigned Sema::getOpenMPNestingLevel() const {
+unsigned SemaOpenMP::getOpenMPNestingLevel() const {
   assert(getLangOpts().OpenMP);
   return DSAStack->getNestingLevel();
 }
 
-bool Sema::isInOpenMPTaskUntiedContext() const {
+bool SemaOpenMP::isInOpenMPTaskUntiedContext() const {
   return isOpenMPTaskingDirective(DSAStack->getCurrentDirective()) &&
          DSAStack->isUntiedRegion();
 }
 
-bool Sema::isInOpenMPTargetExecutionDirective() const {
+bool SemaOpenMP::isInOpenMPTargetExecutionDirective() const {
   return (isOpenMPTargetExecutionDirective(DSAStack->getCurrentDirective()) &&
           !DSAStack->isClauseParsingMode()) ||
          DSAStack->hasDirective(
@@ -2324,7 +2326,7 @@ bool Sema::isInOpenMPTargetExecutionDirective() const {
              false);
 }
 
-bool Sema::isOpenMPRebuildMemberExpr(ValueDecl *D) {
+bool SemaOpenMP::isOpenMPRebuildMemberExpr(ValueDecl *D) {
   // Only rebuild for Field.
   if (!dyn_cast<FieldDecl>(D))
     return false;
@@ -2347,9 +2349,9 @@ static OMPCapturedExprDecl *buildCaptureDecl(Sema &S, IdentifierInfo *Id,
                                              DeclContext *CurContext,
                                              bool AsExpression);
 
-VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
-                                    unsigned StopAt) {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+VarDecl *SemaOpenMP::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
+                                          unsigned StopAt) {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
   D = getCanonicalDecl(D);
 
   auto *VD = dyn_cast<VarDecl>(D);
@@ -2368,7 +2370,8 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
   // 'target' we return true so that this global is also mapped to the device.
   //
   if (VD && !VD->hasLocalStorage() &&
-      (getCurCapturedRegion() || getCurBlock() || getCurLambda())) {
+      (SemaRef.getCurCapturedRegion() || SemaRef.getCurBlock() ||
+       SemaRef.getCurLambda())) {
     if (isInOpenMPTargetExecutionDirective()) {
       DSAStackTy::DSAVarData DVarTop =
           DSAStack->getTopDSA(D, DSAStack->isClauseParsingMode());
@@ -2381,8 +2384,9 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
         return nullptr;
       CapturedRegionScopeInfo *CSI = nullptr;
       for (FunctionScopeInfo *FSI : llvm::drop_begin(
-               llvm::reverse(FunctionScopes),
-               CheckScopeInfo ? (FunctionScopes.size() - (StopAt + 1)) : 0)) {
+               llvm::reverse(SemaRef.FunctionScopes),
+               CheckScopeInfo ? (SemaRef.FunctionScopes.size() - (StopAt + 1))
+                              : 0)) {
         if (!isa<CapturingScopeInfo>(FSI))
           return nullptr;
         if (auto *RSI = dyn_cast<CapturedRegionScopeInfo>(FSI))
@@ -2401,7 +2405,7 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
     if (isInOpenMPDeclareTargetContext()) {
       // Try to mark variable as declare target if it is used in capturing
       // regions.
-      if (LangOpts.OpenMP <= 45 &&
+      if (getLangOpts().OpenMP <= 45 &&
           !OMPDeclareTargetDeclAttr::isDeclareTargetDeclaration(VD))
         checkDeclIsAllowedInOpenMPTarget(nullptr, VD);
       return nullptr;
@@ -2411,7 +2415,7 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
   if (CheckScopeInfo) {
     bool OpenMPFound = false;
     for (unsigned I = StopAt + 1; I > 0; --I) {
-      FunctionScopeInfo *FSI = FunctionScopes[I - 1];
+      FunctionScopeInfo *FSI = SemaRef.FunctionScopes[I - 1];
       if (!isa<CapturingScopeInfo>(FSI))
         return nullptr;
       if (auto *RSI = dyn_cast<CapturedRegionScopeInfo>(FSI))
@@ -2476,22 +2480,23 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
       VarDecl *VD = DSAStack->getImplicitFDCapExprDecl(FD);
       if (VD)
         return VD;
-      if (getCurrentThisType().isNull())
+      if (SemaRef.getCurrentThisType().isNull())
         return nullptr;
-      Expr *ThisExpr = BuildCXXThisExpr(SourceLocation(), getCurrentThisType(),
-                                        /*IsImplicit=*/true);
+      Expr *ThisExpr = SemaRef.BuildCXXThisExpr(SourceLocation(),
+                                                SemaRef.getCurrentThisType(),
+                                                /*IsImplicit=*/true);
       const CXXScopeSpec CS = CXXScopeSpec();
-      Expr *ME = BuildMemberExpr(ThisExpr, /*IsArrow=*/true, SourceLocation(),
-                                 NestedNameSpecifierLoc(), SourceLocation(), FD,
-                                 DeclAccessPair::make(FD, FD->getAccess()),
-                                 /*HadMultipleCandidates=*/false,
-                                 DeclarationNameInfo(), FD->getType(),
-                                 VK_LValue, OK_Ordinary);
+      Expr *ME = SemaRef.BuildMemberExpr(
+          ThisExpr, /*IsArrow=*/true, SourceLocation(),
+          NestedNameSpecifierLoc(), SourceLocation(), FD,
+          DeclAccessPair::make(FD, FD->getAccess()),
+          /*HadMultipleCandidates=*/false, DeclarationNameInfo(), FD->getType(),
+          VK_LValue, OK_Ordinary);
       OMPCapturedExprDecl *CD = buildCaptureDecl(
-          *this, FD->getIdentifier(), ME, DVarPrivate.CKind != OMPC_private,
-          CurContext->getParent(), /*AsExpression=*/false);
+          SemaRef, FD->getIdentifier(), ME, DVarPrivate.CKind != OMPC_private,
+          SemaRef.CurContext->getParent(), /*AsExpression=*/false);
       DeclRefExpr *VDPrivateRefExpr = buildDeclRefExpr(
-          *this, CD, CD->getType().getNonReferenceType(), SourceLocation());
+          SemaRef, CD, CD->getType().getNonReferenceType(), SourceLocation());
       VD = cast<VarDecl>(VDPrivateRefExpr->getDecl());
       DSAStack->addImplicitDefaultFirstprivateFD(FD, VD);
       return VD;
@@ -2505,28 +2510,28 @@ VarDecl *Sema::isOpenMPCapturedDecl(ValueDecl *D, bool CheckScopeInfo,
   return nullptr;
 }
 
-void Sema::adjustOpenMPTargetScopeIndex(unsigned &FunctionScopesIndex,
-                                        unsigned Level) const {
+void SemaOpenMP::adjustOpenMPTargetScopeIndex(unsigned &FunctionScopesIndex,
+                                              unsigned Level) const {
   FunctionScopesIndex -= getOpenMPCaptureLevels(DSAStack->getDirective(Level));
 }
 
-void Sema::startOpenMPLoop() {
-  assert(LangOpts.OpenMP && "OpenMP must be enabled.");
+void SemaOpenMP::startOpenMPLoop() {
+  assert(getLangOpts().OpenMP && "OpenMP must be enabled.");
   if (isOpenMPLoopDirective(DSAStack->getCurrentDirective()))
     DSAStack->loopInit();
 }
 
-void Sema::startOpenMPCXXRangeFor() {
-  assert(LangOpts.OpenMP && "OpenMP must be enabled.");
+void SemaOpenMP::startOpenMPCXXRangeFor() {
+  assert(getLangOpts().OpenMP && "OpenMP must be enabled.");
   if (isOpenMPLoopDirective(DSAStack->getCurrentDirective())) {
     DSAStack->resetPossibleLoopCounter();
     DSAStack->loopStart();
   }
 }
 
-OpenMPClauseKind Sema::isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
-                                           unsigned CapLevel) const {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+OpenMPClauseKind SemaOpenMP::isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
+                                                 unsigned CapLevel) const {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
   if (DSAStack->getCurrentDirective() != OMPD_unknown &&
       (!DSAStack->isClauseParsingMode() ||
        DSAStack->getParentDirective() != OMPD_unknown)) {
@@ -2546,7 +2551,8 @@ OpenMPClauseKind Sema::isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
   }
   if (DSAStack->hasExplicitDirective(isOpenMPTaskingDirective, Level)) {
     bool IsTriviallyCopyable =
-        D->getType().getNonReferenceType().isTriviallyCopyableType(Context) &&
+        D->getType().getNonReferenceType().isTriviallyCopyableType(
+            getASTContext()) &&
         !D->getType()
              .getNonReferenceType()
              .getCanonicalType()
@@ -2620,9 +2626,9 @@ OpenMPClauseKind Sema::isOpenMPPrivateDecl(ValueDecl *D, unsigned Level,
              : OMPC_unknown;
 }
 
-void Sema::setOpenMPCaptureKind(FieldDecl *FD, const ValueDecl *D,
-                                unsigned Level) {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+void SemaOpenMP::setOpenMPCaptureKind(FieldDecl *FD, const ValueDecl *D,
+                                      unsigned Level) {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
   D = getCanonicalDecl(D);
   OpenMPClauseKind OMPC = OMPC_unknown;
   for (unsigned I = DSAStack->getNestingLevel() + 1; I > Level; --I) {
@@ -2649,18 +2655,19 @@ void Sema::setOpenMPCaptureKind(FieldDecl *FD, const ValueDecl *D,
                                        NewLevel)) {
       OMPC = OMPC_map;
       if (DSAStack->mustBeFirstprivateAtLevel(
-              NewLevel, getVariableCategoryFromDecl(LangOpts, D)))
+              NewLevel, getVariableCategoryFromDecl(getLangOpts(), D)))
         OMPC = OMPC_firstprivate;
       break;
     }
   }
   if (OMPC != OMPC_unknown)
-    FD->addAttr(OMPCaptureKindAttr::CreateImplicit(Context, unsigned(OMPC)));
+    FD->addAttr(
+        OMPCaptureKindAttr::CreateImplicit(getASTContext(), unsigned(OMPC)));
 }
 
-bool Sema::isOpenMPTargetCapturedDecl(const ValueDecl *D, unsigned Level,
-                                      unsigned CaptureLevel) const {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+bool SemaOpenMP::isOpenMPTargetCapturedDecl(const ValueDecl *D, unsigned Level,
+                                            unsigned CaptureLevel) const {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
   // Return true if the current level is no longer enclosed in a target region.
 
   SmallVector<OpenMPDirectiveKind, 4> Regions;
@@ -2672,9 +2679,9 @@ bool Sema::isOpenMPTargetCapturedDecl(const ValueDecl *D, unsigned Level,
          Regions[CaptureLevel] != OMPD_task;
 }
 
-bool Sema::isOpenMPGlobalCapturedDecl(ValueDecl *D, unsigned Level,
-                                      unsigned CaptureLevel) const {
-  assert(LangOpts.OpenMP && "OpenMP is not allowed");
+bool SemaOpenMP::isOpenMPGlobalCapturedDecl(ValueDecl *D, unsigned Level,
+                                            unsigned CaptureLevel) const {
+  assert(getLangOpts().OpenMP && "OpenMP is not allowed");
   // Return true if the current level is no longer enclosed in a target region.
 
   if (const auto *VD = dyn_cast<VarDecl>(D)) {
@@ -2702,37 +2709,37 @@ bool Sema::isOpenMPGlobalCapturedDecl(ValueDecl *D, unsigned Level,
   return true;
 }
 
-void Sema::DestroyDataSharingAttributesStack() { delete DSAStack; }
+void SemaOpenMP::DestroyDataSharingAttributesStack() { delete DSAStack; }
 
-void Sema::ActOnOpenMPBeginDeclareVariant(SourceLocation Loc,
-                                          OMPTraitInfo &TI) {
+void SemaOpenMP::ActOnOpenMPBeginDeclareVariant(SourceLocation Loc,
+                                                OMPTraitInfo &TI) {
   OMPDeclareVariantScopes.push_back(OMPDeclareVariantScope(TI));
 }
 
-void Sema::ActOnOpenMPEndDeclareVariant() {
+void SemaOpenMP::ActOnOpenMPEndDeclareVariant() {
   assert(isInOpenMPDeclareVariantScope() &&
          "Not in OpenMP declare variant scope!");
 
   OMPDeclareVariantScopes.pop_back();
 }
 
-void Sema::finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
-                                         const FunctionDecl *Callee,
-                                         SourceLocation Loc) {
-  assert(LangOpts.OpenMP && "Expected OpenMP compilation mode.");
+void SemaOpenMP::finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
+                                               const FunctionDecl *Callee,
+                                               SourceLocation Loc) {
+  assert(getLangOpts().OpenMP && "Expected OpenMP compilation mode.");
   std::optional<OMPDeclareTargetDeclAttr::DevTypeTy> DevTy =
       OMPDeclareTargetDeclAttr::getDeviceType(Caller->getMostRecentDecl());
   // Ignore host functions during device analyzis.
-  if (LangOpts.OpenMPIsTargetDevice &&
+  if (getLangOpts().OpenMPIsTargetDevice &&
       (!DevTy || *DevTy == OMPDeclareTargetDeclAttr::DT_Host))
     return;
   // Ignore nohost functions during host analyzis.
-  if (!LangOpts.OpenMPIsTargetDevice && DevTy &&
+  if (!getLangOpts().OpenMPIsTargetDevice && DevTy &&
       *DevTy == OMPDeclareTargetDeclAttr::DT_NoHost)
     return;
   const FunctionDecl *FD = Callee->getMostRecentDecl();
   DevTy = OMPDeclareTargetDeclAttr::getDeviceType(FD);
-  if (LangOpts.OpenMPIsTargetDevice && DevTy &&
+  if (getLangOpts().OpenMPIsTargetDevice && DevTy &&
       *DevTy == OMPDeclareTargetDeclAttr::DT_Host) {
     // Diagnose host function called during device codegen.
     StringRef HostDevTy =
@@ -2743,8 +2750,9 @@ void Sema::finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
         << HostDevTy;
     return;
   }
-  if (!LangOpts.OpenMPIsTargetDevice && !LangOpts.OpenMPOffloadMandatory &&
-      DevTy && *DevTy == OMPDeclareTargetDeclAttr::DT_NoHost) {
+  if (!getLangOpts().OpenMPIsTargetDevice &&
+      !getLangOpts().OpenMPOffloadMandatory && DevTy &&
+      *DevTy == OMPDeclareTargetDeclAttr::DT_NoHost) {
     // In OpenMP 5.2 or later, if the function has a host variant then allow
     // that to be called instead
     auto &&HasHostAttr = [](const FunctionDecl *Callee) {
@@ -2773,21 +2781,21 @@ void Sema::finalizeOpenMPDelayedAnalysis(const FunctionDecl *Caller,
   }
 }
 
-void Sema::StartOpenMPDSABlock(OpenMPDirectiveKind DKind,
-                               const DeclarationNameInfo &DirName,
-                               Scope *CurScope, SourceLocation Loc) {
+void SemaOpenMP::StartOpenMPDSABlock(OpenMPDirectiveKind DKind,
+                                     const DeclarationNameInfo &DirName,
+                                     Scope *CurScope, SourceLocation Loc) {
   DSAStack->push(DKind, DirName, CurScope, Loc);
-  PushExpressionEvaluationContext(
-      ExpressionEvaluationContext::PotentiallyEvaluated);
+  SemaRef.PushExpressionEvaluationContext(
+      Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 }
 
-void Sema::StartOpenMPClause(OpenMPClauseKind K) {
+void SemaOpenMP::StartOpenMPClause(OpenMPClauseKind K) {
   DSAStack->setClauseParsingMode(K);
 }
 
-void Sema::EndOpenMPClause() {
+void SemaOpenMP::EndOpenMPClause() {
   DSAStack->setClauseParsingMode(/*K=*/OMPC_unknown);
-  CleanupVarDeclMarking();
+  SemaRef.CleanupVarDeclMarking();
 }
 
 static std::pair<ValueDecl *, bool>
@@ -2871,7 +2879,7 @@ static void reportOriginalDsa(Sema &SemaRef, const DSAStackTy *Stack,
                               const DSAStackTy::DSAVarData &DVar,
                               bool IsLoopIterVar = false);
 
-void Sema::EndOpenMPDSABlock(Stmt *CurDirective) {
+void SemaOpenMP::EndOpenMPDSABlock(Stmt *CurDirective) {
   // OpenMP [2.14.3.5, Restrictions, C/C++, p.1]
   //  A variable of class type (or array thereof) that appears in a lastprivate
   //  clause requires an accessible, unambiguous default constructor for the
@@ -2898,15 +2906,15 @@ void Sema::EndOpenMPDSABlock(Stmt *CurDirective) {
             // variable is not added to IdResolver, so the code in the OpenMP
             // region uses original variable for proper diagnostics.
             VarDecl *VDPrivate = buildVarDecl(
-                *this, DE->getExprLoc(), Type.getUnqualifiedType(),
+                SemaRef, DE->getExprLoc(), Type.getUnqualifiedType(),
                 VD->getName(), VD->hasAttrs() ? &VD->getAttrs() : nullptr, DRE);
-            ActOnUninitializedDecl(VDPrivate);
+            SemaRef.ActOnUninitializedDecl(VDPrivate);
             if (VDPrivate->isInvalidDecl()) {
               PrivateCopies.push_back(nullptr);
               continue;
             }
             PrivateCopies.push_back(buildDeclRefExpr(
-                *this, VDPrivate, DE->getType(), DE->getExprLoc()));
+                SemaRef, VDPrivate, DE->getType(), DE->getExprLoc()));
           } else {
             // The variable is also a firstprivate, so initialization sequence
             // for private copy is generated already.
@@ -2924,7 +2932,7 @@ void Sema::EndOpenMPDSABlock(Stmt *CurDirective) {
           SourceLocation ELoc;
           SourceRange ERange;
           Expr *SimpleRefExpr = RefExpr;
-          auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+          auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
           if (Res.second)
             // It will be analyzed later.
             PrivateRefs.push_back(RefExpr);
@@ -2977,7 +2985,7 @@ void Sema::EndOpenMPDSABlock(Stmt *CurDirective) {
                  diag::err_omp_allocator_used_in_clauses)
                 << D.Allocator->getSourceRange();
             if (DVar.RefExpr)
-              reportOriginalDsa(*this, DSAStack, VD, DVar);
+              reportOriginalDsa(SemaRef, DSAStack, VD, DVar);
             else
               Diag(MapExpr->getExprLoc(), diag::note_used_here)
                   << MapExpr->getSourceRange();
@@ -2987,14 +2995,14 @@ void Sema::EndOpenMPDSABlock(Stmt *CurDirective) {
       }
     }
     // Check allocate clauses.
-    if (!CurContext->isDependentContext())
-      checkAllocateClauses(*this, DSAStack, D->clauses());
-    checkReductionClauses(*this, DSAStack, D->clauses());
+    if (!SemaRef.CurContext->isDependentContext())
+      checkAllocateClauses(SemaRef, DSAStack, D->clauses());
+    checkReductionClauses(SemaRef, DSAStack, D->clauses());
   }
 
   DSAStack->pop();
-  DiscardCleanupsInEvaluationContext();
-  PopExpressionEvaluationContext();
+  SemaRef.DiscardCleanupsInEvaluationContext();
+  SemaRef.PopExpressionEvaluationContext();
 }
 
 static bool FinishOpenMPLinearClause(OMPLinearClause &Clause, DeclRefExpr *IV,
@@ -3047,27 +3055,28 @@ class VarOrFuncDeclFilterCCC final : public CorrectionCandidateCallback {
 
 } // namespace
 
-ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
-                                         CXXScopeSpec &ScopeSpec,
-                                         const DeclarationNameInfo &Id,
-                                         OpenMPDirectiveKind Kind) {
-  LookupResult Lookup(*this, Id, LookupOrdinaryName);
-  LookupParsedName(Lookup, CurScope, &ScopeSpec, true);
+ExprResult SemaOpenMP::ActOnOpenMPIdExpression(Scope *CurScope,
+                                               CXXScopeSpec &ScopeSpec,
+                                               const DeclarationNameInfo &Id,
+                                               OpenMPDirectiveKind Kind) {
+  ASTContext &Context = getASTContext();
+  LookupResult Lookup(SemaRef, Id, Sema::LookupOrdinaryName);
+  SemaRef.LookupParsedName(Lookup, CurScope, &ScopeSpec, true);
 
   if (Lookup.isAmbiguous())
     return ExprError();
 
   VarDecl *VD;
   if (!Lookup.isSingleResult()) {
-    VarDeclFilterCCC CCC(*this);
+    VarDeclFilterCCC CCC(SemaRef);
     if (TypoCorrection Corrected =
-            CorrectTypo(Id, LookupOrdinaryName, CurScope, nullptr, CCC,
-                        CTK_ErrorRecovery)) {
-      diagnoseTypo(Corrected,
-                   PDiag(Lookup.empty()
-                             ? diag::err_undeclared_var_use_suggest
-                             : diag::err_omp_expected_var_arg_suggest)
-                       << Id.getName());
+            SemaRef.CorrectTypo(Id, Sema::LookupOrdinaryName, CurScope, nullptr,
+                                CCC, Sema::CTK_ErrorRecovery)) {
+      SemaRef.diagnoseTypo(
+          Corrected,
+          SemaRef.PDiag(Lookup.empty() ? diag::err_undeclared_var_use_suggest
+                                       : diag::err_omp_expected_var_arg_suggest)
+              << Id.getName());
       VD = Corrected.getCorrectionDeclAs<VarDecl>();
     } else {
       Diag(Id.getLoc(), Lookup.empty() ? diag::err_undeclared_var_use
@@ -3101,7 +3110,7 @@ ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
   //   A threadprivate directive for file-scope variables must appear outside
   //   any definition or declaration.
   if (CanonicalVD->getDeclContext()->isTranslationUnit() &&
-      !getCurLexicalContext()->isTranslationUnit()) {
+      !SemaRef.getCurLexicalContext()->isTranslationUnit()) {
     Diag(Id.getLoc(), diag::err_omp_var_scope)
         << getOpenMPDirectiveName(Kind) << VD;
     bool IsDecl =
@@ -3116,7 +3125,7 @@ ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
   //   in the class definition, in the same scope in which the member
   //   variables are declared.
   if (CanonicalVD->isStaticDataMember() &&
-      !CanonicalVD->getDeclContext()->Equals(getCurLexicalContext())) {
+      !CanonicalVD->getDeclContext()->Equals(SemaRef.getCurLexicalContext())) {
     Diag(Id.getLoc(), diag::err_omp_var_scope)
         << getOpenMPDirectiveName(Kind) << VD;
     bool IsDecl =
@@ -3131,8 +3140,9 @@ ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
   //   outside any definition or declaration other than the namespace
   //   definition itself.
   if (CanonicalVD->getDeclContext()->isNamespace() &&
-      (!getCurLexicalContext()->isFileContext() ||
-       !getCurLexicalContext()->Encloses(CanonicalVD->getDeclContext()))) {
+      (!SemaRef.getCurLexicalContext()->isFileContext() ||
+       !SemaRef.getCurLexicalContext()->Encloses(
+           CanonicalVD->getDeclContext()))) {
     Diag(Id.getLoc(), diag::err_omp_var_scope)
         << getOpenMPDirectiveName(Kind) << VD;
     bool IsDecl =
@@ -3146,7 +3156,7 @@ ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
   //   A threadprivate directive for static block-scope variables must appear
   //   in the scope of the variable and not in a nested scope.
   if (CanonicalVD->isLocalVarDecl() && CurScope &&
-      !isDeclInScope(ND, getCurLexicalContext(), CurScope)) {
+      !SemaRef.isDeclInScope(ND, SemaRef.getCurLexicalContext(), CurScope)) {
     Diag(Id.getLoc(), diag::err_omp_var_scope)
         << getOpenMPDirectiveName(Kind) << VD;
     bool IsDecl =
@@ -3174,11 +3184,11 @@ ExprResult Sema::ActOnOpenMPIdExpression(Scope *CurScope,
                              Id.getLoc(), ExprType, VK_LValue);
 }
 
-Sema::DeclGroupPtrTy
-Sema::ActOnOpenMPThreadprivateDirective(SourceLocation Loc,
-                                        ArrayRef<Expr *> VarList) {
+SemaOpenMP::DeclGroupPtrTy
+SemaOpenMP::ActOnOpenMPThreadprivateDirective(SourceLocation Loc,
+                                              ArrayRef<Expr *> VarList) {
   if (OMPThreadPrivateDecl *D = CheckOMPThreadPrivateDecl(Loc, VarList)) {
-    CurContext->addDecl(D);
+    SemaRef.CurContext->addDecl(D);
     return DeclGroupPtrTy::make(DeclGroupRef(D));
   }
   return nullptr;
@@ -3215,7 +3225,9 @@ class LocalVarRefChecker final
 } // namespace
 
 OMPThreadPrivateDecl *
-Sema::CheckOMPThreadPrivateDecl(SourceLocation Loc, ArrayRef<Expr *> VarList) {
+SemaOpenMP::CheckOMPThreadPrivateDecl(SourceLocation Loc,
+                                      ArrayRef<Expr *> VarList) {
+  ASTContext &Context = getASTContext();
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     auto *DE = cast<DeclRefExpr>(RefExpr);
@@ -3235,8 +3247,8 @@ Sema::CheckOMPThreadPrivateDecl(SourceLocation Loc, ArrayRef<Expr *> VarList) {
 
     // OpenMP [2.9.2, Restrictions, C/C++, p.10]
     //   A threadprivate variable must not have an incomplete type.
-    if (RequireCompleteType(ILoc, VD->getType(),
-                            diag::err_omp_threadprivate_incomplete_type)) {
+    if (SemaRef.RequireCompleteType(
+            ILoc, VD->getType(), diag::err_omp_threadprivate_incomplete_type)) {
       continue;
     }
 
@@ -3274,7 +3286,7 @@ Sema::CheckOMPThreadPrivateDecl(SourceLocation Loc, ArrayRef<Expr *> VarList) {
     // Check if initial value of threadprivate variable reference variable with
     // local storage (it is not supported by runtime).
     if (const Expr *Init = VD->getAnyInitializer()) {
-      LocalVarRefChecker Checker(*this);
+      LocalVarRefChecker Checker(SemaRef);
       if (Checker.Visit(Init))
         continue;
     }
@@ -3288,8 +3300,8 @@ Sema::CheckOMPThreadPrivateDecl(SourceLocation Loc, ArrayRef<Expr *> VarList) {
   }
   OMPThreadPrivateDecl *D = nullptr;
   if (!Vars.empty()) {
-    D = OMPThreadPrivateDecl::Create(Context, getCurLexicalContext(), Loc,
-                                     Vars);
+    D = OMPThreadPrivateDecl::Create(Context, SemaRef.getCurLexicalContext(),
+                                     Loc, Vars);
     D->setAccess(AS_public);
   }
   return D;
@@ -3395,10 +3407,9 @@ applyOMPAllocateAttribute(Sema &S, VarDecl *VD,
     ML->DeclarationMarkedOpenMPAllocate(VD, A);
 }
 
-Sema::DeclGroupPtrTy
-Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
-                                   ArrayRef<OMPClause *> Clauses,
-                                   DeclContext *Owner) {
+SemaOpenMP::DeclGroupPtrTy SemaOpenMP::ActOnOpenMPAllocateDirective(
+    SourceLocation Loc, ArrayRef<Expr *> VarList, ArrayRef<OMPClause *> Clauses,
+    DeclContext *Owner) {
   assert(Clauses.size() <= 2 && "Expected at most two clauses.");
   Expr *Alignment = nullptr;
   Expr *Allocator = nullptr;
@@ -3407,9 +3418,9 @@ Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
     // allocate directives that appear in a target region must specify an
     // allocator clause unless a requires directive with the dynamic_allocators
     // clause is present in the same compilation unit.
-    if (LangOpts.OpenMPIsTargetDevice &&
+    if (getLangOpts().OpenMPIsTargetDevice &&
         !DSAStack->hasRequiresDeclWithClause<OMPDynamicAllocatorsClause>())
-      targetDiag(Loc, diag::err_expected_allocator_clause);
+      SemaRef.targetDiag(Loc, diag::err_expected_allocator_clause);
   } else {
     for (const OMPClause *C : Clauses)
       if (const auto *AC = dyn_cast<OMPAllocatorClause>(C))
@@ -3420,7 +3431,7 @@ Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
         llvm_unreachable("Unexpected clause on allocate directive");
   }
   OMPAllocateDeclAttr::AllocatorTypeTy AllocatorKind =
-      getAllocatorKind(*this, DSAStack, Allocator);
+      getAllocatorKind(SemaRef, DSAStack, Allocator);
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     auto *DE = cast<DeclRefExpr>(RefExpr);
@@ -3435,7 +3446,7 @@ Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
 
     // If the used several times in the allocate directive, the same allocator
     // must be used.
-    if (checkPreviousOMPAllocateAttribute(*this, DSAStack, RefExpr, VD,
+    if (checkPreviousOMPAllocateAttribute(SemaRef, DSAStack, RefExpr, VD,
                                           AllocatorKind, Allocator))
       continue;
 
@@ -3448,7 +3459,7 @@ Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
         Diag(Allocator->getExprLoc(),
              diag::err_omp_expected_predefined_allocator)
             << Allocator->getSourceRange();
-        bool IsDecl = VD->isThisDeclarationADefinition(Context) ==
+        bool IsDecl = VD->isThisDeclarationADefinition(getASTContext()) ==
                       VarDecl::DeclarationOnly;
         Diag(VD->getLocation(),
              IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -3458,45 +3469,46 @@ Sema::ActOnOpenMPAllocateDirective(SourceLocation Loc, ArrayRef<Expr *> VarList,
     }
 
     Vars.push_back(RefExpr);
-    applyOMPAllocateAttribute(*this, VD, AllocatorKind, Allocator, Alignment,
+    applyOMPAllocateAttribute(SemaRef, VD, AllocatorKind, Allocator, Alignment,
                               DE->getSourceRange());
   }
   if (Vars.empty())
     return nullptr;
   if (!Owner)
-    Owner = getCurLexicalContext();
-  auto *D = OMPAllocateDecl::Create(Context, Owner, Loc, Vars, Clauses);
+    Owner = SemaRef.getCurLexicalContext();
+  auto *D = OMPAllocateDecl::Create(getASTContext(), Owner, Loc, Vars, Clauses);
   D->setAccess(AS_public);
   Owner->addDecl(D);
   return DeclGroupPtrTy::make(DeclGroupRef(D));
 }
 
-Sema::DeclGroupPtrTy
-Sema::ActOnOpenMPRequiresDirective(SourceLocation Loc,
-                                   ArrayRef<OMPClause *> ClauseList) {
+SemaOpenMP::DeclGroupPtrTy
+SemaOpenMP::ActOnOpenMPRequiresDirective(SourceLocation Loc,
+                                         ArrayRef<OMPClause *> ClauseList) {
   OMPRequiresDecl *D = nullptr;
-  if (!CurContext->isFileContext()) {
+  if (!SemaRef.CurContext->isFileContext()) {
     Diag(Loc, diag::err_omp_invalid_scope) << "requires";
   } else {
     D = CheckOMPRequiresDecl(Loc, ClauseList);
     if (D) {
-      CurContext->addDecl(D);
+      SemaRef.CurContext->addDecl(D);
       DSAStack->addRequiresDecl(D);
     }
   }
   return DeclGroupPtrTy::make(DeclGroupRef(D));
 }
 
-void Sema::ActOnOpenMPAssumesDirective(SourceLocation Loc,
-                                       OpenMPDirectiveKind DKind,
-                                       ArrayRef<std::string> Assumptions,
-                                       bool SkippedClauses) {
+void SemaOpenMP::ActOnOpenMPAssumesDirective(SourceLocation Loc,
+                                             OpenMPDirectiveKind DKind,
+                                             ArrayRef<std::string> Assumptions,
+                                             bool SkippedClauses) {
   if (!SkippedClauses && Assumptions.empty())
     Diag(Loc, diag::err_omp_no_clause_for_directive)
         << llvm::omp::getAllAssumeClauseOptions()
         << llvm::omp::getOpenMPDirectiveName(DKind);
 
-  auto *AA = OMPAssumeAttr::Create(Context, llvm::join(Assumptions, ","), Loc);
+  auto *AA =
+      OMPAssumeAttr::Create(getASTContext(), llvm::join(Assumptions, ","), Loc);
   if (DKind == llvm::omp::Directive::OMPD_begin_assumes) {
     OMPAssumeScoped.push_back(AA);
     return;
@@ -3515,7 +3527,7 @@ void Sema::ActOnOpenMPAssumesDirective(SourceLocation Loc,
   // declarations in included headers. To this end, we traverse all existing
   // declaration contexts and annotate function declarations here.
   SmallVector<DeclContext *, 8> DeclContexts;
-  auto *Ctx = CurContext;
+  auto *Ctx = SemaRef.CurContext;
   while (Ctx->getLexicalParent())
     Ctx = Ctx->getLexicalParent();
   DeclContexts.push_back(Ctx);
@@ -3539,13 +3551,14 @@ void Sema::ActOnOpenMPAssumesDirective(SourceLocation Loc,
   }
 }
 
-void Sema::ActOnOpenMPEndAssumesDirective() {
+void SemaOpenMP::ActOnOpenMPEndAssumesDirective() {
   assert(isInOpenMPAssumeScope() && "Not in OpenMP assumes scope!");
   OMPAssumeScoped.pop_back();
 }
 
-OMPRequiresDecl *Sema::CheckOMPRequiresDecl(SourceLocation Loc,
-                                            ArrayRef<OMPClause *> ClauseList) {
+OMPRequiresDecl *
+SemaOpenMP::CheckOMPRequiresDecl(SourceLocation Loc,
+                                 ArrayRef<OMPClause *> ClauseList) {
   /// For target specific clauses, the requires directive cannot be
   /// specified after the handling of any of the target regions in the
   /// current compilation unit.
@@ -3576,8 +3589,8 @@ OMPRequiresDecl *Sema::CheckOMPRequiresDecl(SourceLocation Loc,
   }
 
   if (!DSAStack->hasDuplicateRequiresClause(ClauseList))
-    return OMPRequiresDecl::Create(Context, getCurLexicalContext(), Loc,
-                                   ClauseList);
+    return OMPRequiresDecl::Create(
+        getASTContext(), SemaRef.getCurLexicalContext(), Loc, ClauseList);
   return nullptr;
 }
 
@@ -3695,7 +3708,7 @@ class DSAAttrChecker final : public StmtVisitor<DSAAttrChecker, void> {
   llvm::SmallVector<Expr *, 4> ImplicitMap[DefaultmapKindNum][OMPC_MAP_delete];
   llvm::SmallVector<OpenMPMapModifierKind, NumberOfOMPMapClauseModifiers>
       ImplicitMapModifier[DefaultmapKindNum];
-  Sema::VarsWithInheritedDSAType VarsWithInheritedDSA;
+  SemaOpenMP::VarsWithInheritedDSAType VarsWithInheritedDSA;
   llvm::SmallDenseSet<const ValueDecl *, 4> ImplicitDeclarations;
 
   void VisitSubCaptures(OMPExecutableDirective *S) {
@@ -4161,7 +4174,7 @@ class DSAAttrChecker final : public StmtVisitor<DSAAttrChecker, void> {
   getImplicitMapModifier(OpenMPDefaultmapClauseKind Kind) const {
     return ImplicitMapModifier[Kind];
   }
-  const Sema::VarsWithInheritedDSAType &getVarsWithInheritedDSA() const {
+  const SemaOpenMP::VarsWithInheritedDSAType &getVarsWithInheritedDSA() const {
     return VarsWithInheritedDSA;
   }
 
@@ -4193,7 +4206,9 @@ static void handleDeclareVariantConstructTrait(DSAStackTy *Stack,
   Stack->handleConstructTrait(Traits, ScopeEntry);
 }
 
-void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
+void SemaOpenMP::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind,
+                                        Scope *CurScope) {
+  ASTContext &Context = getASTContext();
   switch (DKind) {
   case OMPD_parallel:
   case OMPD_parallel_for:
@@ -4208,13 +4223,13 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     break;
   }
   case OMPD_target_teams:
@@ -4232,7 +4247,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4242,31 +4257,33 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params, /*OpenMPCaptureLevel=*/0);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params,
+                                     /*OpenMPCaptureLevel=*/0);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
-    SmallVector<Sema::CapturedParamNameType, 2> ParamsTarget;
+    SmallVector<SemaOpenMP::CapturedParamNameType, 2> ParamsTarget;
     if (getLangOpts().OpenMPIsTargetDevice)
       ParamsTarget.push_back(std::make_pair(StringRef("dyn_ptr"), VoidPtrTy));
     ParamsTarget.push_back(
         std::make_pair(StringRef(), QualType())); // __context with shared vars;
     // Start a captured region for 'target' with no implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTarget,
-                             /*OpenMPCaptureLevel=*/1);
-    Sema::CapturedParamNameType ParamsTeamsOrParallel[] = {
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTarget,
+                                     /*OpenMPCaptureLevel=*/1);
+    SemaOpenMP::CapturedParamNameType ParamsTeamsOrParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'teams' or 'parallel'.  Both regions have
     // the same implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTeamsOrParallel, /*OpenMPCaptureLevel=*/2);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTeamsOrParallel,
+                                     /*OpenMPCaptureLevel=*/2);
     break;
   }
   case OMPD_target:
@@ -4279,7 +4296,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4289,21 +4306,22 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params, /*OpenMPCaptureLevel=*/0);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params,
+                                     /*OpenMPCaptureLevel=*/0);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
-    SmallVector<Sema::CapturedParamNameType, 2> ParamsTarget;
+    SmallVector<SemaOpenMP::CapturedParamNameType, 2> ParamsTarget;
     if (getLangOpts().OpenMPIsTargetDevice)
       ParamsTarget.push_back(std::make_pair(StringRef("dyn_ptr"), VoidPtrTy));
     ParamsTarget.push_back(
         std::make_pair(StringRef(), QualType())); // __context with shared vars;
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTarget,
-                             /*OpenMPCaptureLevel=*/1);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTarget,
+                                     /*OpenMPCaptureLevel=*/1);
     break;
   }
   case OMPD_atomic:
@@ -4329,11 +4347,11 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   case OMPD_scope:
   case OMPD_target_data:
   case OMPD_dispatch: {
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     break;
   }
   case OMPD_task: {
@@ -4345,7 +4363,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4355,11 +4373,11 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
     break;
@@ -4386,7 +4404,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4401,11 +4419,11 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".reductions.", VoidPtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
     break;
@@ -4426,19 +4444,20 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
-    Sema::CapturedParamNameType ParamsParallel[] = {
+    SemaOpenMP::CapturedParamNameType ParamsParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'parallel'.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsParallel, /*OpenMPCaptureLevel=*/0);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsParallel,
+                                     /*OpenMPCaptureLevel=*/0);
     QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4453,11 +4472,12 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".reductions.", VoidPtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params, /*OpenMPCaptureLevel=*/1);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params,
+                                     /*OpenMPCaptureLevel=*/1);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
     break;
@@ -4467,15 +4487,15 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
         std::make_pair(".previous.ub.", Context.getSizeType().withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     break;
   }
   // For 'target teams loop', collect all captured regions so codegen can
@@ -4492,7 +4512,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4502,32 +4522,35 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params, /*OpenMPCaptureLevel=*/0);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params,
+                                     /*OpenMPCaptureLevel=*/0);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
-    SmallVector<Sema::CapturedParamNameType, 2> ParamsTarget;
+    SmallVector<SemaOpenMP::CapturedParamNameType, 2> ParamsTarget;
     if (getLangOpts().OpenMPIsTargetDevice)
       ParamsTarget.push_back(std::make_pair(StringRef("dyn_ptr"), VoidPtrTy));
     ParamsTarget.push_back(
         std::make_pair(StringRef(), QualType())); // __context with shared vars;
     // Start a captured region for 'target' with no implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTarget, /*OpenMPCaptureLevel=*/1);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTarget,
+                                     /*OpenMPCaptureLevel=*/1);
 
-    Sema::CapturedParamNameType ParamsTeams[] = {
+    SemaOpenMP::CapturedParamNameType ParamsTeams[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'target' with no implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTeams, /*OpenMPCaptureLevel=*/2);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTeams,
+                                     /*OpenMPCaptureLevel=*/2);
 
-    Sema::CapturedParamNameType ParamsParallel[] = {
+    SemaOpenMP::CapturedParamNameType ParamsParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
@@ -4536,8 +4559,9 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     };
     // Start a captured region for 'teams' or 'parallel'.  Both regions have
     // the same implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsParallel, /*OpenMPCaptureLevel=*/3);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsParallel,
+                                     /*OpenMPCaptureLevel=*/3);
     break;
   }
 
@@ -4548,16 +4572,17 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
 
-    Sema::CapturedParamNameType ParamsTeams[] = {
+    SemaOpenMP::CapturedParamNameType ParamsTeams[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'target' with no implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsTeams, /*OpenMPCaptureLevel=*/0);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsTeams,
+                                     /*OpenMPCaptureLevel=*/0);
 
-    Sema::CapturedParamNameType ParamsParallel[] = {
+    SemaOpenMP::CapturedParamNameType ParamsParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
         std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
@@ -4566,8 +4591,9 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     };
     // Start a captured region for 'teams' or 'parallel'.  Both regions have
     // the same implicit parameters.
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             ParamsParallel, /*OpenMPCaptureLevel=*/1);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, ParamsParallel,
+                                     /*OpenMPCaptureLevel=*/1);
     break;
   }
   case OMPD_target_update:
@@ -4581,7 +4607,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
-    Sema::CapturedParamNameType Params[] = {
+    SemaOpenMP::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
         std::make_pair(".part_id.", KmpInt32PtrTy),
         std::make_pair(".privates.", VoidPtrTy),
@@ -4591,11 +4617,11 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
-    ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
-                             Params);
+    SemaRef.ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope,
+                                     CR_OpenMP, Params);
     // Mark this captured region as inlined, because we don't use outlined
     // function directly.
-    getCurCapturedRegion()->TheCapturedDecl->addAttr(
+    SemaRef.getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
             Context, {}, AlwaysInlineAttr::Keyword_forceinline));
     break;
@@ -4626,15 +4652,15 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   default:
     llvm_unreachable("Unknown OpenMP directive");
   }
-  DSAStack->setContext(CurContext);
+  DSAStack->setContext(SemaRef.CurContext);
   handleDeclareVariantConstructTrait(DSAStack, DKind, /* ScopeEntry */ true);
 }
 
-int Sema::getNumberOfConstructScopes(unsigned Level) const {
+int SemaOpenMP::getNumberOfConstructScopes(unsigned Level) const {
   return getOpenMPCaptureLevels(DSAStack->getDirective(Level));
 }
 
-int Sema::getOpenMPCaptureLevels(OpenMPDirectiveKind DKind) {
+int SemaOpenMP::getOpenMPCaptureLevels(OpenMPDirectiveKind DKind) {
   SmallVector<OpenMPDirectiveKind, 4> CaptureRegions;
   getOpenMPCaptureRegions(CaptureRegions, DKind);
   return CaptureRegions.size();
@@ -4674,7 +4700,7 @@ static OMPCapturedExprDecl *buildCaptureDecl(Sema &S, IdentifierInfo *Id,
 static DeclRefExpr *buildCapture(Sema &S, ValueDecl *D, Expr *CaptureExpr,
                                  bool WithInit) {
   OMPCapturedExprDecl *CD;
-  if (VarDecl *VD = S.isOpenMPCapturedDecl(D))
+  if (VarDecl *VD = S.OpenMP().isOpenMPCapturedDecl(D))
     CD = cast<OMPCapturedExprDecl>(VD);
   else
     CD = buildCaptureDecl(S, D->getIdentifier(), CaptureExpr, WithInit,
@@ -4726,7 +4752,7 @@ class CaptureRegionUnwinderRAII {
       : S(S), ErrorFound(ErrorFound), DKind(DKind) {}
   ~CaptureRegionUnwinderRAII() {
     if (ErrorFound) {
-      int ThisCaptureLevel = S.getOpenMPCaptureLevels(DKind);
+      int ThisCaptureLevel = S.OpenMP().getOpenMPCaptureLevels(DKind);
       while (--ThisCaptureLevel >= 0)
         S.ActOnCapturedRegionError();
     }
@@ -4734,10 +4760,10 @@ class CaptureRegionUnwinderRAII {
 };
 } // namespace
 
-void Sema::tryCaptureOpenMPLambdas(ValueDecl *V) {
+void SemaOpenMP::tryCaptureOpenMPLambdas(ValueDecl *V) {
   // Capture variables captured by reference in lambdas for target-based
   // directives.
-  if (!CurContext->isDependentContext() &&
+  if (!SemaRef.CurContext->isDependentContext() &&
       (isOpenMPTargetExecutionDirective(DSAStack->getCurrentDirective()) ||
        isOpenMPTargetDataManagementDirective(
            DSAStack->getCurrentDirective()))) {
@@ -4757,14 +4783,14 @@ void Sema::tryCaptureOpenMPLambdas(ValueDecl *V) {
           if (LC.getCaptureKind() == LCK_ByRef) {
             VarDecl *VD = cast<VarDecl>(LC.getCapturedVar());
             DeclContext *VDC = VD->getDeclContext();
-            if (!VDC->Encloses(CurContext))
+            if (!VDC->Encloses(SemaRef.CurContext))
               continue;
-            MarkVariableReferenced(LC.getLocation(), VD);
+            SemaRef.MarkVariableReferenced(LC.getLocation(), VD);
           } else if (LC.getCaptureKind() == LCK_This) {
-            QualType ThisTy = getCurrentThisType();
-            if (!ThisTy.isNull() &&
-                Context.typesAreCompatible(ThisTy, ThisCapture->getType()))
-              CheckCXXThisCapture(LC.getLocation());
+            QualType ThisTy = SemaRef.getCurrentThisType();
+            if (!ThisTy.isNull() && getASTContext().typesAreCompatible(
+                                        ThisTy, ThisCapture->getType()))
+              SemaRef.CheckCXXThisCapture(LC.getLocation());
           }
         }
       }
@@ -4804,8 +4830,8 @@ static bool checkOrderedOrderSpecified(Sema &S,
   return false;
 }
 
-StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
-                                      ArrayRef<OMPClause *> Clauses) {
+StmtResult SemaOpenMP::ActOnOpenMPRegionEnd(StmtResult S,
+                                            ArrayRef<OMPClause *> Clauses) {
   handleDeclareVariantConstructTrait(DSAStack, DSAStack->getCurrentDirective(),
                                      /* ScopeEntry */ false);
   if (DSAStack->getCurrentDirective() == OMPD_atomic ||
@@ -4817,7 +4843,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
 
   bool ErrorFound = false;
   CaptureRegionUnwinderRAII CaptureRegionUnwinder(
-      *this, ErrorFound, DSAStack->getCurrentDirective());
+      SemaRef, ErrorFound, DSAStack->getCurrentDirective());
   if (!S.isUsable()) {
     ErrorFound = true;
     return StmtError();
@@ -4831,7 +4857,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
   SmallVector<const OMPClauseWithPreInit *, 4> PICs;
   // This is required for proper codegen.
   for (OMPClause *Clause : Clauses) {
-    if (!LangOpts.OpenMPSimd &&
+    if (!getLangOpts().OpenMPSimd &&
         (isOpenMPTaskingDirective(DSAStack->getCurrentDirective()) ||
          DSAStack->getCurrentDirective() == OMPD_target) &&
         Clause->getClauseKind() == OMPC_in_reduction) {
@@ -4840,7 +4866,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
       auto *IRC = cast<OMPInReductionClause>(Clause);
       for (Expr *E : IRC->taskgroup_descriptors())
         if (E)
-          MarkDeclarationsReferencedInExpr(E);
+          SemaRef.MarkDeclarationsReferencedInExpr(E);
     }
     if (isOpenMPPrivate(Clause->getClauseKind()) ||
         Clause->getClauseKind() == OMPC_copyprivate ||
@@ -4851,7 +4877,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
       // Mark all variables in private list clauses as used in inner region.
       for (Stmt *VarRef : Clause->children()) {
         if (auto *E = cast_or_null<Expr>(VarRef)) {
-          MarkDeclarationsReferencedInExpr(E);
+          SemaRef.MarkDeclarationsReferencedInExpr(E);
         }
       }
       DSAStack->setForceVarCapturing(/*V=*/false);
@@ -4865,7 +4891,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
         PICs.push_back(C);
       if (auto *C = OMPClauseWithPostUpdate::get(Clause)) {
         if (Expr *E = C->getPostUpdateExpr())
-          MarkDeclarationsReferencedInExpr(E);
+          SemaRef.MarkDeclarationsReferencedInExpr(E);
       }
     }
     if (Clause->getClauseKind() == OMPC_schedule)
@@ -4877,7 +4903,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
   }
   // Capture allocator expressions if used.
   for (Expr *E : DSAStack->getInnerAllocators())
-    MarkDeclarationsReferencedInExpr(E);
+    SemaRef.MarkDeclarationsReferencedInExpr(E);
   // OpenMP, 2.7.1 Loop Construct, Restrictions
   // The nonmonotonic modifier cannot be specified if an ordered clause is
   // specified.
@@ -4899,7 +4925,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
   // OpenMP 5.0, 2.9.2 Worksharing-Loop Construct, Restrictions.
   // If an order(concurrent) clause is present, an ordered clause may not appear
   // on the same directive.
-  if (checkOrderedOrderSpecified(*this, Clauses))
+  if (checkOrderedOrderSpecified(SemaRef, Clauses))
     ErrorFound = true;
   if (!LCs.empty() && OC && OC->getNumForLoops()) {
     for (const OMPLinearClause *C : LCs) {
@@ -4936,7 +4962,8 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
             CaptureRegion == OMPD_unknown) {
           if (auto *DS = cast_or_null<DeclStmt>(C->getPreInitStmt())) {
             for (Decl *D : DS->decls())
-              MarkVariableReferenced(D->getLocation(), cast<VarDecl>(D));
+              SemaRef.MarkVariableReferenced(D->getLocation(),
+                                             cast<VarDecl>(D));
           }
         }
       }
@@ -4950,7 +4977,7 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
                ++I) {
             OMPUsesAllocatorsClause::Data D = UAC->getAllocatorData(I);
             if (Expr *E = D.AllocatorTraits)
-              MarkDeclarationsReferencedInExpr(E);
+              SemaRef.MarkDeclarationsReferencedInExpr(E);
           }
           continue;
         }
@@ -4965,17 +4992,17 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
             continue;
           for (Expr *E : RC->copy_array_temps())
             if (E)
-              MarkDeclarationsReferencedInExpr(E);
+              SemaRef.MarkDeclarationsReferencedInExpr(E);
         }
         if (auto *AC = dyn_cast<OMPAlignedClause>(C)) {
           for (Expr *E : AC->varlists())
-            MarkDeclarationsReferencedInExpr(E);
+            SemaRef.MarkDeclarationsReferencedInExpr(E);
         }
       }
     }
     if (++CompletedRegions == CaptureRegions.size())
       DSAStack->setBodyComplete();
-    SR = ActOnCapturedRegionEnd(SR.get());
+    SR = SemaRef.ActOnCapturedRegionEnd(SR.get());
   }
   return SR;
 }
@@ -5782,9 +5809,9 @@ static CapturedStmt *buildLoopVarFunc(Sema &Actions, QualType LoopVarTy,
   // the OpenMPIRBuilder to know additional C/C++ semantics, such as how to
   // invoke a copy constructor.
   QualType TargetParamTy = Ctx.getLValueReferenceType(LoopVarTy);
-  Sema::CapturedParamNameType Params[] = {{"LoopVar", TargetParamTy},
-                                          {"Logical", LogicalTy},
-                                          {StringRef(), QualType()}};
+  SemaOpenMP::CapturedParamNameType Params[] = {{"LoopVar", TargetParamTy},
+                                                {"Logical", LogicalTy},
+                                                {StringRef(), QualType()}};
   Actions.ActOnCapturedRegionStart({}, nullptr, CR_Default, Params);
 
   // Capture the initial iterator which represents the LoopVar value at the
@@ -5835,7 +5862,7 @@ static CapturedStmt *buildLoopVarFunc(Sema &Actions, QualType LoopVarTy,
       AssertSuccess(Actions.ActOnCapturedRegionEnd(Body)));
 }
 
-StmtResult Sema::ActOnOpenMPCanonicalLoop(Stmt *AStmt) {
+StmtResult SemaOpenMP::ActOnOpenMPCanonicalLoop(Stmt *AStmt) {
   ASTContext &Ctx = getASTContext();
 
   // Extract the common elements of ForStmt and CXXForRangeStmt:
@@ -5946,8 +5973,8 @@ StmtResult Sema::ActOnOpenMPCanonicalLoop(Stmt *AStmt) {
     if (IncBin->getOpcode() == BO_AddAssign) {
       Step = IncBin->getRHS();
     } else if (IncBin->getOpcode() == BO_SubAssign) {
-      Step =
-          AssertSuccess(BuildUnaryOp(nullptr, {}, UO_Minus, IncBin->getRHS()));
+      Step = AssertSuccess(
+          SemaRef.BuildUnaryOp(nullptr, {}, UO_Minus, IncBin->getRHS()));
     } else
       llvm_unreachable("unhandled binary increment operator");
   } else if (auto *CondCXXOp = dyn_cast<CXXOperatorCallExpr>(Inc)) {
@@ -5965,7 +5992,7 @@ StmtResult Sema::ActOnOpenMPCanonicalLoop(Stmt *AStmt) {
       break;
     case OO_MinusEqual:
       Step = AssertSuccess(
-          BuildUnaryOp(nullptr, {}, UO_Minus, CondCXXOp->getArg(1)));
+          SemaRef.BuildUnaryOp(nullptr, {}, UO_Minus, CondCXXOp->getArg(1)));
       break;
     default:
       llvm_unreachable("unhandled overloaded increment operator");
@@ -5974,16 +6001,17 @@ StmtResult Sema::ActOnOpenMPCanonicalLoop(Stmt *AStmt) {
     llvm_unreachable("unknown increment expression");
 
   CapturedStmt *DistanceFunc =
-      buildDistanceFunc(*this, LogicalTy, CondRel, LHS, RHS, Step);
+      buildDistanceFunc(SemaRef, LogicalTy, CondRel, LHS, RHS, Step);
   CapturedStmt *LoopVarFunc = buildLoopVarFunc(
-      *this, LVTy, LogicalTy, CounterRef, Step, isa<CXXForRangeStmt>(AStmt));
-  DeclRefExpr *LVRef = BuildDeclRefExpr(LUVDecl, LUVDecl->getType(), VK_LValue,
-                                        {}, nullptr, nullptr, {}, nullptr);
+      SemaRef, LVTy, LogicalTy, CounterRef, Step, isa<CXXForRangeStmt>(AStmt));
+  DeclRefExpr *LVRef =
+      SemaRef.BuildDeclRefExpr(LUVDecl, LUVDecl->getType(), VK_LValue, {},
+                               nullptr, nullptr, {}, nullptr);
   return OMPCanonicalLoop::create(getASTContext(), AStmt, DistanceFunc,
                                   LoopVarFunc, LVRef);
 }
 
-StmtResult Sema::ActOnOpenMPLoopnest(Stmt *AStmt) {
+StmtResult SemaOpenMP::ActOnOpenMPLoopnest(Stmt *AStmt) {
   // Handle a literal loop.
   if (isa<ForStmt>(AStmt) || isa<CXXForRangeStmt>(AStmt))
     return ActOnOpenMPCanonicalLoop(AStmt);
@@ -6128,7 +6156,7 @@ processImplicitMapsWithDefaultMappers(Sema &S, DSAStackTy *Stack,
       continue;
     CXXScopeSpec MapperIdScopeSpec;
     DeclarationNameInfo MapperId;
-    if (OMPClause *NewClause = S.ActOnOpenMPMapClause(
+    if (OMPClause *NewClause = S.OpenMP().ActOnOpenMPMapClause(
             nullptr, C->getMapTypeModifiers(), C->getMapTypeModifiersLoc(),
             MapperIdScopeSpec, MapperId, C->getMapType(),
             /*IsMapTypeImplicit=*/true, SourceLocation(), SourceLocation(),
@@ -6210,14 +6238,12 @@ static bool teamsLoopCanBeParallelFor(Stmt *AStmt, Sema &SemaRef) {
   return Checker.teamsLoopCanBeParallelFor();
 }
 
-bool Sema::mapLoopConstruct(llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
-                            ArrayRef<OMPClause *> Clauses,
-                            OpenMPBindClauseKind &BindKind,
-                            OpenMPDirectiveKind &Kind,
-                            OpenMPDirectiveKind &PrevMappedDirective,
-                            SourceLocation StartLoc, SourceLocation EndLoc,
-                            const DeclarationNameInfo &DirName,
-                            OpenMPDirectiveKind CancelRegion) {
+bool SemaOpenMP::mapLoopConstruct(
+    llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
+    ArrayRef<OMPClause *> Clauses, OpenMPBindClauseKind &BindKind,
+    OpenMPDirectiveKind &Kind, OpenMPDirectiveKind &PrevMappedDirective,
+    SourceLocation StartLoc, SourceLocation EndLoc,
+    const DeclarationNameInfo &DirName, OpenMPDirectiveKind CancelRegion) {
 
   bool UseClausesWithoutBind = false;
 
@@ -6299,7 +6325,7 @@ bool Sema::mapLoopConstruct(llvm::SmallVector<OMPClause *> &ClausesWithoutBind,
   return UseClausesWithoutBind;
 }
 
-StmtResult Sema::ActOnOpenMPExecutableDirective(
+StmtResult SemaOpenMP::ActOnOpenMPExecutableDirective(
     OpenMPDirectiveKind Kind, const DeclarationNameInfo &DirName,
     OpenMPDirectiveKind CancelRegion, ArrayRef<OMPClause *> Clauses,
     Stmt *AStmt, SourceLocation StartLoc, SourceLocation EndLoc,
@@ -6324,8 +6350,8 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   }
 
   // First check CancelRegion which is then used in checkNestingOfRegions.
-  if (checkCancelRegion(*this, Kind, CancelRegion, StartLoc) ||
-      checkNestingOfRegions(*this, DSAStack, DK, DirName, CancelRegion,
+  if (checkCancelRegion(SemaRef, Kind, CancelRegion, StartLoc) ||
+      checkNestingOfRegions(SemaRef, DSAStack, DK, DirName, CancelRegion,
                             BindKind, StartLoc)) {
     return StmtError();
   }
@@ -6344,13 +6370,14 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   } else {
     ClausesWithImplicit.append(Clauses.begin(), Clauses.end());
   }
-  if (AStmt && !CurContext->isDependentContext() && Kind != OMPD_atomic &&
-      Kind != OMPD_critical && Kind != OMPD_section && Kind != OMPD_master &&
-      Kind != OMPD_masked && !isOpenMPLoopTransformationDirective(Kind)) {
+  if (AStmt && !SemaRef.CurContext->isDependentContext() &&
+      Kind != OMPD_atomic && Kind != OMPD_critical && Kind != OMPD_section &&
+      Kind != OMPD_master && Kind != OMPD_masked &&
+      !isOpenMPLoopTransformationDirective(Kind)) {
     assert(isa<CapturedStmt>(AStmt) && "Captured statement expected");
 
     // Check default data sharing attributes for referenced variables.
-    DSAAttrChecker DSAChecker(DSAStack, *this, cast<CapturedStmt>(AStmt));
+    DSAAttrChecker DSAChecker(DSAStack, SemaRef, cast<CapturedStmt>(AStmt));
     int ThisCaptureLevel = getOpenMPCaptureLevels(Kind);
     Stmt *S = AStmt;
     while (--ThisCaptureLevel >= 0)
@@ -6490,8 +6517,8 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
     }
     // Build expressions for implicit maps of data members with 'default'
     // mappers.
-    if (LangOpts.OpenMP >= 50)
-      processImplicitMapsWithDefaultMappers(*this, DSAStack,
+    if (getLangOpts().OpenMP >= 50)
+      processImplicitMapsWithDefaultMappers(SemaRef, DSAStack,
                                             ClausesWithImplicit);
   }
 
@@ -6505,7 +6532,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   case OMPD_simd:
     Res = ActOnOpenMPSimdDirective(ClausesWithImplicit, AStmt, StartLoc, EndLoc,
                                    VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_tile:
@@ -6523,7 +6550,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   case OMPD_for_simd:
     Res = ActOnOpenMPForSimdDirective(ClausesWithImplicit, AStmt, StartLoc,
                                       EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_sections:
@@ -6561,7 +6588,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
     Res = ActOnOpenMPParallelForSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_scope:
@@ -6698,7 +6725,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
     Res = ActOnOpenMPTaskLoopSimdDirective(ClausesWithImplicit, AStmt, StartLoc,
                                            EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_taskloop);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_master_taskloop:
@@ -6715,13 +6742,13 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
     Res = ActOnOpenMPMasterTaskLoopSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_taskloop);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_masked_taskloop_simd:
     Res = ActOnOpenMPMaskedTaskLoopSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 51) {
+    if (getLangOpts().OpenMP >= 51) {
       AllowedNameModifiers.push_back(OMPD_taskloop);
       AllowedNameModifiers.push_back(OMPD_simd);
     }
@@ -6735,7 +6762,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   case OMPD_parallel_masked_taskloop:
     Res = ActOnOpenMPParallelMaskedTaskLoopDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 51) {
+    if (getLangOpts().OpenMP >= 51) {
       AllowedNameModifiers.push_back(OMPD_taskloop);
       AllowedNameModifiers.push_back(OMPD_parallel);
     }
@@ -6745,13 +6772,13 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_taskloop);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_parallel_masked_taskloop_simd:
     Res = ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 51) {
+    if (getLangOpts().OpenMP >= 51) {
       AllowedNameModifiers.push_back(OMPD_taskloop);
       AllowedNameModifiers.push_back(OMPD_parallel);
       AllowedNameModifiers.push_back(OMPD_simd);
@@ -6775,13 +6802,13 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
     Res = ActOnOpenMPDistributeParallelForSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_distribute_simd:
     Res = ActOnOpenMPDistributeSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_target_parallel_for_simd:
@@ -6789,14 +6816,14 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_target);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_target_simd:
     Res = ActOnOpenMPTargetSimdDirective(ClausesWithImplicit, AStmt, StartLoc,
                                          EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_target);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_teams_distribute:
@@ -6806,14 +6833,14 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   case OMPD_teams_distribute_simd:
     Res = ActOnOpenMPTeamsDistributeSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_teams_distribute_parallel_for_simd:
     Res = ActOnOpenMPTeamsDistributeParallelForSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_teams_distribute_parallel_for:
@@ -6842,14 +6869,14 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_target);
     AllowedNameModifiers.push_back(OMPD_parallel);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_target_teams_distribute_simd:
     Res = ActOnOpenMPTargetTeamsDistributeSimdDirective(
         ClausesWithImplicit, AStmt, StartLoc, EndLoc, VarsWithInheritedDSA);
     AllowedNameModifiers.push_back(OMPD_target);
-    if (LangOpts.OpenMP >= 50)
+    if (getLangOpts().OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_interop:
@@ -6906,7 +6933,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   if (DSAStack->getDefaultDSA() == DSA_none ||
       DSAStack->getDefaultDSA() == DSA_private ||
       DSAStack->getDefaultDSA() == DSA_firstprivate) {
-    DSAAttrChecker DSAChecker(DSAStack, *this, nullptr);
+    DSAAttrChecker DSAChecker(DSAStack, SemaRef, nullptr);
     for (OMPClause *C : Clauses) {
       switch (C->getClauseKind()) {
       case OMPC_num_threads:
@@ -7043,13 +7070,13 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   }
 
   if (!AllowedNameModifiers.empty())
-    ErrorFound = checkIfClauses(*this, Kind, Clauses, AllowedNameModifiers) ||
+    ErrorFound = checkIfClauses(SemaRef, Kind, Clauses, AllowedNameModifiers) ||
                  ErrorFound;
 
   if (ErrorFound)
     return StmtError();
 
-  if (!CurContext->isDependentContext() &&
+  if (!SemaRef.CurContext->isDependentContext() &&
       isOpenMPTargetExecutionDirective(Kind) &&
       !(DSAStack->hasRequiresDeclWithClause<OMPUnifiedSharedMemoryClause>() ||
         DSAStack->hasRequiresDeclWithClause<OMPUnifiedAddressClause>() ||
@@ -7062,7 +7089,7 @@ StmtResult Sema::ActOnOpenMPExecutableDirective(
   return Res;
 }
 
-Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareSimdDirective(
+SemaOpenMP::DeclGroupPtrTy SemaOpenMP::ActOnOpenMPDeclareSimdDirective(
     DeclGroupPtrTy DG, OMPDeclareSimdDeclAttr::BranchStateTy BS, Expr *Simdlen,
     ArrayRef<Expr *> Uniforms, ArrayRef<Expr *> Aligneds,
     ArrayRef<Expr *> Alignments, ArrayRef<Expr *> Linears,
@@ -7297,13 +7324,15 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareSimdDirective(
       NewStep = PerformOpenMPImplicitIntegerConversion(Step->getExprLoc(), Step)
                     .get();
       if (NewStep)
-        NewStep =
-            VerifyIntegerConstantExpression(NewStep, /*FIXME*/ AllowFold).get();
+        NewStep = SemaRef
+                      .VerifyIntegerConstantExpression(
+                          NewStep, /*FIXME*/ Sema::AllowFold)
+                      .get();
     }
     NewSteps.push_back(NewStep);
   }
   auto *NewAttr = OMPDeclareSimdDeclAttr::CreateImplicit(
-      Context, BS, SL.get(), const_cast<Expr **>(Uniforms.data()),
+      getASTContext(), BS, SL.get(), const_cast<Expr **>(Uniforms.data()),
       Uniforms.size(), const_cast<Expr **>(Aligneds.data()), Aligneds.size(),
       const_cast<Expr **>(NewAligns.data()), NewAligns.size(),
       const_cast<Expr **>(Linears.data()), Linears.size(),
@@ -7336,7 +7365,7 @@ static void setPrototype(Sema &S, FunctionDecl *FD, FunctionDecl *FDWithProto,
   FD->setParams(Params);
 }
 
-void Sema::ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D) {
+void SemaOpenMP::ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D) {
   if (D->isInvalidDecl())
     return;
   FunctionDecl *FD = nullptr;
@@ -7349,7 +7378,7 @@ void Sema::ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D) {
   // If we are instantiating templates we do *not* apply scoped assumptions but
   // only global ones. We apply scoped assumption to the template definition
   // though.
-  if (!inTemplateInstantiation()) {
+  if (!SemaRef.inTemplateInstantiation()) {
     for (OMPAssumeAttr *AA : OMPAssumeScoped)
       FD->addAttr(AA);
   }
@@ -7357,10 +7386,10 @@ void Sema::ActOnFinishedFunctionDefinitionInOpenMPAssumeScope(Decl *D) {
     FD->addAttr(AA);
 }
 
-Sema::OMPDeclareVariantScope::OMPDeclareVariantScope(OMPTraitInfo &TI)
+SemaOpenMP::OMPDeclareVariantScope::OMPDeclareVariantScope(OMPTraitInfo &TI)
     : TI(&TI), NameSuffix(TI.getMangledName()) {}
 
-void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
+void SemaOpenMP::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
     Scope *S, Declarator &D, MultiTemplateParamsArg TemplateParamLists,
     SmallVectorImpl<FunctionDecl *> &Bases) {
   if (!D.getIdentifier())
@@ -7376,11 +7405,11 @@ void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
     return;
 
   const IdentifierInfo *BaseII = D.getIdentifier();
-  LookupResult Lookup(*this, DeclarationName(BaseII), D.getIdentifierLoc(),
-                      LookupOrdinaryName);
-  LookupParsedName(Lookup, S, &D.getCXXScopeSpec());
+  LookupResult Lookup(SemaRef, DeclarationName(BaseII), D.getIdentifierLoc(),
+                      Sema::LookupOrdinaryName);
+  SemaRef.LookupParsedName(Lookup, S, &D.getCXXScopeSpec());
 
-  TypeSourceInfo *TInfo = GetTypeForDeclarator(D);
+  TypeSourceInfo *TInfo = SemaRef.GetTypeForDeclarator(D);
   QualType FType = TInfo->getType();
 
   bool IsConstexpr =
@@ -7409,7 +7438,7 @@ void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
 
     QualType UDeclTy = UDecl->getType();
     if (!UDeclTy->isDependentType()) {
-      QualType NewType = Context.mergeFunctionTypes(
+      QualType NewType = getASTContext().mergeFunctionTypes(
           FType, UDeclTy, /* OfBlockPointer */ false,
           /* Unqualified */ false, /* AllowCXX */ true);
       if (NewType.isNull())
@@ -7425,7 +7454,7 @@ void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
   // If no base was found we create a declaration that we use as base.
   if (Bases.empty() && UseImplicitBase) {
     D.setFunctionDefinitionKind(FunctionDefinitionKind::Declaration);
-    Decl *BaseD = HandleDeclarator(S, D, TemplateParamLists);
+    Decl *BaseD = SemaRef.HandleDeclarator(S, D, TemplateParamLists);
     BaseD->setImplicit(true);
     if (auto *BaseTemplD = dyn_cast<FunctionTemplateDecl>(BaseD))
       Bases.push_back(BaseTemplD->getTemplatedDecl());
@@ -7437,18 +7466,18 @@ void Sema::ActOnStartOfFunctionDefinitionInOpenMPDeclareVariantScope(
   MangledName += D.getIdentifier()->getName();
   MangledName += getOpenMPVariantManglingSeparatorStr();
   MangledName += DVScope.NameSuffix;
-  IdentifierInfo &VariantII = Context.Idents.get(MangledName);
+  IdentifierInfo &VariantII = getASTContext().Idents.get(MangledName);
 
   VariantII.setMangledOpenMPVariantName(true);
   D.SetIdentifier(&VariantII, D.getBeginLoc());
 }
 
-void Sema::ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
+void SemaOpenMP::ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
     Decl *D, SmallVectorImpl<FunctionDecl *> &Bases) {
   // Do not mark function as is used to prevent its emission if this is the
   // only place where it is used.
   EnterExpressionEvaluationContext Unevaluated(
-      *this, Sema::ExpressionEvaluationContext::Unevaluated);
+      SemaRef, Sema::ExpressionEvaluationContext::Unevaluated);
 
   FunctionDecl *FD = nullptr;
   if (auto *UTemplDecl = dyn_cast<FunctionTemplateDecl>(D))
@@ -7456,14 +7485,14 @@ void Sema::ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
   else
     FD = cast<FunctionDecl>(D);
   auto *VariantFuncRef = DeclRefExpr::Create(
-      Context, NestedNameSpecifierLoc(), SourceLocation(), FD,
+      getASTContext(), NestedNameSpecifierLoc(), SourceLocation(), FD,
       /* RefersToEnclosingVariableOrCapture */ false,
       /* NameLoc */ FD->getLocation(), FD->getType(),
       ExprValueKind::VK_PRValue);
 
   OMPDeclareVariantScope &DVScope = OMPDeclareVariantScopes.back();
   auto *OMPDeclareVariantA = OMPDeclareVariantAttr::CreateImplicit(
-      Context, VariantFuncRef, DVScope.TI,
+      getASTContext(), VariantFuncRef, DVScope.TI,
       /*NothingArgs=*/nullptr, /*NothingArgsSize=*/0,
       /*NeedDevicePtrArgs=*/nullptr, /*NeedDevicePtrArgsSize=*/0,
       /*AppendArgs=*/nullptr, /*AppendArgsSize=*/0);
@@ -7471,10 +7500,11 @@ void Sema::ActOnFinishedFunctionDefinitionInOpenMPDeclareVariantScope(
     BaseFD->addAttr(OMPDeclareVariantA);
 }
 
-ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
-                                 SourceLocation LParenLoc,
-                                 MultiExprArg ArgExprs,
-                                 SourceLocation RParenLoc, Expr *ExecConfig) {
+ExprResult SemaOpenMP::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
+                                       SourceLocation LParenLoc,
+                                       MultiExprArg ArgExprs,
+                                       SourceLocation RParenLoc,
+                                       Expr *ExecConfig) {
   // The common case is a regular call we do not want to specialize at all. Try
   // to make that case fast by bailing early.
   CallExpr *CE = dyn_cast<CallExpr>(Call.get());
@@ -7485,7 +7515,7 @@ ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
   if (!CalleeFnDecl)
     return Call;
 
-  if (LangOpts.OpenMP >= 51 && CalleeFnDecl->getIdentifier() &&
+  if (getLangOpts().OpenMP >= 51 && CalleeFnDecl->getIdentifier() &&
       CalleeFnDecl->getName().starts_with_insensitive("omp_")) {
     // checking for any calls inside an Order region
     if (Scope && Scope->isOpenMPOrderClauseScope())
@@ -7504,7 +7534,8 @@ ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
         << ISATrait;
   };
   TargetOMPContext OMPCtx(Context, std::move(DiagUnknownTrait),
-                          getCurFunctionDecl(), DSAStack->getConstructTraits());
+                          SemaRef.getCurFunctionDecl(),
+                          DSAStack->getConstructTraits());
 
   QualType CalleeFnType = CalleeFnDecl->getType();
 
@@ -7549,7 +7580,7 @@ ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
       // different type than the base function. This is intended and OK but if
       // we cannot create a call the difference is not in the "implementation
       // defined range" we allow.
-      Sema::TentativeAnalysisScope Trap(*this);
+      Sema::TentativeAnalysisScope Trap(SemaRef);
 
       if (auto *SpecializedMethod = dyn_cast<CXXMethodDecl>(BestDecl)) {
         auto *MemberCall = dyn_cast<CXXMemberCallExpr>(CE);
@@ -7558,12 +7589,12 @@ ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
             /* IsArrow */ false, SpecializedMethod, Context.BoundMemberTy,
             MemberCall->getValueKind(), MemberCall->getObjectKind());
       }
-      NewCall = BuildCallExpr(Scope, BestExpr, LParenLoc, ArgExprs, RParenLoc,
-                              ExecConfig);
+      NewCall = SemaRef.BuildCallExpr(Scope, BestExpr, LParenLoc, ArgExprs,
+                                      RParenLoc, ExecConfig);
       if (NewCall.isUsable()) {
         if (CallExpr *NCE = dyn_cast<CallExpr>(NewCall.get())) {
           FunctionDecl *NewCalleeFnDecl = NCE->getDirectCallee();
-          QualType NewType = Context.mergeFunctionTypes(
+          QualType NewType = getASTContext().mergeFunctionTypes(
               CalleeFnType, NewCalleeFnDecl->getType(),
               /* OfBlockPointer */ false,
               /* Unqualified */ false, /* AllowCXX */ true);
@@ -7581,14 +7612,16 @@ ExprResult Sema::ActOnOpenMPCall(ExprResult Call, Scope *Scope,
 
   if (!NewCall.isUsable())
     return Call;
-  return PseudoObjectExpr::Create(Context, CE, {NewCall.get()}, 0);
+  return PseudoObjectExpr::Create(getASTContext(), CE, {NewCall.get()}, 0);
 }
 
 std::optional<std::pair<FunctionDecl *, Expr *>>
-Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
-                                        Expr *VariantRef, OMPTraitInfo &TI,
-                                        unsigned NumAppendArgs,
-                                        SourceRange SR) {
+SemaOpenMP::checkOpenMPDeclareVariantFunction(SemaOpenMP::DeclGroupPtrTy DG,
+                                              Expr *VariantRef,
+                                              OMPTraitInfo &TI,
+                                              unsigned NumAppendArgs,
+                                              SourceRange SR) {
+  ASTContext &Context = getASTContext();
   if (!DG || DG.get().isNull())
     return std::nullopt;
 
@@ -7631,7 +7664,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   // Check if the function was emitted already.
   const FunctionDecl *Definition;
   if (!FD->isThisDeclarationADefinition() && FD->isDefined(Definition) &&
-      (LangOpts.EmitAllDecls || Context.DeclMustBeEmitted(Definition)))
+      (getLangOpts().EmitAllDecls || Context.DeclMustBeEmitted(Definition)))
     Diag(SR.getBegin(), diag::warn_omp_declare_variant_after_emitted)
         << FD->getLocation();
 
@@ -7654,7 +7687,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   // Deal with non-constant score and user condition expressions.
   auto HandleNonConstantScoresAndConditions = [this](Expr *&E,
                                                      bool IsScore) -> bool {
-    if (!E || E->isIntegerConstantExpr(Context))
+    if (!E || E->isIntegerConstantExpr(getASTContext()))
       return false;
 
     if (IsScore) {
@@ -7686,9 +7719,9 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
     // Adjust the function type to account for an extra omp_interop_t for each
     // specified in the append_args clause.
     const TypeDecl *TD = nullptr;
-    LookupResult Result(*this, &Context.Idents.get("omp_interop_t"),
+    LookupResult Result(SemaRef, &Context.Idents.get("omp_interop_t"),
                         SR.getBegin(), Sema::LookupOrdinaryName);
-    if (LookupName(Result, getCurScope())) {
+    if (SemaRef.LookupName(Result, SemaRef.getCurScope())) {
       NamedDecl *ND = Result.getFoundDecl();
       TD = dyn_cast_or_null<TypeDecl>(ND);
     }
@@ -7711,7 +7744,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   // Convert VariantRef expression to the type of the original function to
   // resolve possible conflicts.
   ExprResult VariantRefCast = VariantRef;
-  if (LangOpts.CPlusPlus) {
+  if (getLangOpts().CPlusPlus) {
     QualType FnPtrType;
     auto *Method = dyn_cast<CXXMethodDecl>(FD);
     if (Method && !Method->isStatic()) {
@@ -7722,9 +7755,9 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
       {
         // Build adrr_of unary op to correctly handle type checks for member
         // functions.
-        Sema::TentativeAnalysisScope Trap(*this);
-        ER = CreateBuiltinUnaryOp(VariantRef->getBeginLoc(), UO_AddrOf,
-                                  VariantRef);
+        Sema::TentativeAnalysisScope Trap(SemaRef);
+        ER = SemaRef.CreateBuiltinUnaryOp(VariantRef->getBeginLoc(), UO_AddrOf,
+                                          VariantRef);
       }
       if (!ER.isUsable()) {
         Diag(VariantRef->getExprLoc(), diag::err_omp_function_expected)
@@ -7737,9 +7770,9 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
     }
     QualType VarianPtrType = Context.getPointerType(VariantRef->getType());
     if (VarianPtrType.getUnqualifiedType() != FnPtrType.getUnqualifiedType()) {
-      ImplicitConversionSequence ICS = TryImplicitConversion(
+      ImplicitConversionSequence ICS = SemaRef.TryImplicitConversion(
           VariantRef, FnPtrType.getUnqualifiedType(),
-          /*SuppressUserConversions=*/false, AllowedExplicit::None,
+          /*SuppressUserConversions=*/false, Sema::AllowedExplicit::None,
           /*InOverloadResolution=*/false,
           /*CStyle=*/false,
           /*AllowObjCWritebackConversion=*/false);
@@ -7751,8 +7784,8 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
             << (NumAppendArgs ? 1 : 0) << VariantRef->getSourceRange();
         return std::nullopt;
       }
-      VariantRefCast = PerformImplicitConversion(
-          VariantRef, FnPtrType.getUnqualifiedType(), AA_Converting);
+      VariantRefCast = SemaRef.PerformImplicitConversion(
+          VariantRef, FnPtrType.getUnqualifiedType(), Sema::AA_Converting);
       if (!VariantRefCast.isUsable())
         return std::nullopt;
     }
@@ -7765,7 +7798,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
     }
   }
 
-  ExprResult ER = CheckPlaceholderExpr(VariantRefCast.get());
+  ExprResult ER = SemaRef.CheckPlaceholderExpr(VariantRefCast.get());
   if (!ER.isUsable() ||
       !ER.get()->IgnoreParenImpCasts()->getType()->isFunctionType()) {
     Diag(VariantRef->getExprLoc(), diag::err_omp_function_expected)
@@ -7795,7 +7828,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   }
 
   // Check if function types are compatible in C.
-  if (!LangOpts.CPlusPlus) {
+  if (!getLangOpts().CPlusPlus) {
     QualType NewType =
         Context.mergeFunctionTypes(AdjustedFnType, NewFD->getType());
     if (NewType.isNull()) {
@@ -7807,9 +7840,9 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
     }
     if (NewType->isFunctionProtoType()) {
       if (FD->getType()->isFunctionNoProtoType())
-        setPrototype(*this, FD, NewFD, NewType);
+        setPrototype(SemaRef, FD, NewFD, NewType);
       else if (NewFD->getType()->isFunctionNoProtoType())
-        setPrototype(*this, NewFD, FD, NewType);
+        setPrototype(SemaRef, NewFD, FD, NewType);
     }
   }
 
@@ -7872,15 +7905,15 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   }
 
   // Check general compatibility.
-  if (areMultiversionVariantFunctionsCompatible(
+  if (SemaRef.areMultiversionVariantFunctionsCompatible(
           FD, NewFD, PartialDiagnostic::NullDiagnostic(),
           PartialDiagnosticAt(SourceLocation(),
                               PartialDiagnostic::NullDiagnostic()),
           PartialDiagnosticAt(
               VariantRef->getExprLoc(),
-              PDiag(diag::err_omp_declare_variant_doesnt_support)),
+              SemaRef.PDiag(diag::err_omp_declare_variant_doesnt_support)),
           PartialDiagnosticAt(VariantRef->getExprLoc(),
-                              PDiag(diag::err_omp_declare_variant_diff)
+                              SemaRef.PDiag(diag::err_omp_declare_variant_diff)
                                   << FD->getLocation()),
           /*TemplatesSupported=*/true, /*ConstexprSupported=*/false,
           /*CLinkageMayDiffer=*/true))
@@ -7888,7 +7921,7 @@ Sema::checkOpenMPDeclareVariantFunction(Sema::DeclGroupPtrTy DG,
   return std::make_pair(FD, cast<Expr>(DRE));
 }
 
-void Sema::ActOnOpenMPDeclareVariantDirective(
+void SemaOpenMP::ActOnOpenMPDeclareVariantDirective(
     FunctionDecl *FD, Expr *VariantRef, OMPTraitInfo &TI,
     ArrayRef<Expr *> AdjustArgsNothing,
     ArrayRef<Expr *> AdjustArgsNeedDevicePtr,
@@ -7906,7 +7939,7 @@ void Sema::ActOnOpenMPDeclareVariantDirective(
 
   if (!AllAdjustArgs.empty() || !AppendArgs.empty()) {
     VariantMatchInfo VMI;
-    TI.getAsVariantMatchInfo(Context, VMI);
+    TI.getAsVariantMatchInfo(getASTContext(), VMI);
     if (!llvm::is_contained(
             VMI.ConstructTraits,
             llvm::omp::TraitProperty::construct_dispatch_dispatch)) {
@@ -7949,18 +7982,18 @@ void Sema::ActOnOpenMPDeclareVariantDirective(
   }
 
   auto *NewAttr = OMPDeclareVariantAttr::CreateImplicit(
-      Context, VariantRef, &TI, const_cast<Expr **>(AdjustArgsNothing.data()),
-      AdjustArgsNothing.size(),
+      getASTContext(), VariantRef, &TI,
+      const_cast<Expr **>(AdjustArgsNothing.data()), AdjustArgsNothing.size(),
       const_cast<Expr **>(AdjustArgsNeedDevicePtr.data()),
       AdjustArgsNeedDevicePtr.size(),
       const_cast<OMPInteropInfo *>(AppendArgs.data()), AppendArgs.size(), SR);
   FD->addAttr(NewAttr);
 }
 
-StmtResult Sema::ActOnOpenMPParallelDirective(ArrayRef<OMPClause *> Clauses,
-                                              Stmt *AStmt,
-                                              SourceLocation StartLoc,
-                                              SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPParallelDirective(ArrayRef<OMPClause *> Clauses,
+                                         Stmt *AStmt, SourceLocation StartLoc,
+                                         SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -7972,11 +8005,11 @@ StmtResult Sema::ActOnOpenMPParallelDirective(ArrayRef<OMPClause *> Clauses,
   // longjmp() and throw() must not violate the entry/exit criteria.
   CS->getCapturedDecl()->setNothrow();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPParallelDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
-                                      DSAStack->getTaskgroupReductionRef(),
-                                      DSAStack->isCancelRegion());
+  return OMPParallelDirective::Create(
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
+      DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
 namespace {
@@ -8226,7 +8259,7 @@ bool OpenMPIterationSpaceChecker::setStep(Expr *NewStep, bool Subtract) {
   if (!NewStep->isValueDependent()) {
     // Check that the step is integer expression.
     SourceLocation StepLoc = NewStep->getBeginLoc();
-    ExprResult Val = SemaRef.PerformOpenMPImplicitIntegerConversion(
+    ExprResult Val = SemaRef.OpenMP().PerformOpenMPImplicitIntegerConversion(
         StepLoc, getExprAsWritten(NewStep));
     if (Val.isInvalid())
       return true;
@@ -9248,7 +9281,7 @@ DeclRefExpr *OpenMPIterationSpaceChecker::buildCounterVar(
     DSAStackTy &DSA) const {
   auto *VD = dyn_cast<VarDecl>(LCDecl);
   if (!VD) {
-    VD = SemaRef.isOpenMPCapturedDecl(LCDecl);
+    VD = SemaRef.OpenMP().isOpenMPCapturedDecl(LCDecl);
     DeclRefExpr *Ref = buildDeclRefExpr(
         SemaRef, VD, VD->getType().getNonReferenceType(), DefaultLoc);
     const DSAStackTy::DSAVarData Data =
@@ -9321,14 +9354,15 @@ Expr *OpenMPIterationSpaceChecker::buildOrderedLoopData(
 }
 } // namespace
 
-void Sema::ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init) {
+void SemaOpenMP::ActOnOpenMPLoopInitialization(SourceLocation ForLoc,
+                                               Stmt *Init) {
   assert(getLangOpts().OpenMP && "OpenMP is not active.");
   assert(Init && "Expected loop in canonical form.");
   unsigned AssociatedLoops = DSAStack->getAssociatedLoops();
   if (AssociatedLoops > 0 &&
       isOpenMPLoopDirective(DSAStack->getCurrentDirective())) {
     DSAStack->loopStart();
-    OpenMPIterationSpaceChecker ISC(*this, /*SupportsNonRectangular=*/true,
+    OpenMPIterationSpaceChecker ISC(SemaRef, /*SupportsNonRectangular=*/true,
                                     *DSAStack, ForLoc);
     if (!ISC.checkAndSetInit(Init, /*EmitDiags=*/false)) {
       if (ValueDecl *D = ISC.getLoopDecl()) {
@@ -9338,7 +9372,7 @@ void Sema::ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init) {
           if (VarDecl *Private = isOpenMPCapturedDecl(D)) {
             VD = Private;
           } else {
-            PrivateRef = buildCapture(*this, D, ISC.getLoopDeclRefExpr(),
+            PrivateRef = buildCapture(SemaRef, D, ISC.getLoopDeclRefExpr(),
                                       /*WithInit=*/false);
             VD = cast<VarDecl>(PrivateRef->getDecl());
           }
@@ -9348,10 +9382,10 @@ void Sema::ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init) {
         if (LD != D->getCanonicalDecl()) {
           DSAStack->resetPossibleLoopCounter();
           if (auto *Var = dyn_cast_or_null<VarDecl>(LD))
-            MarkDeclarationsReferencedInExpr(
-                buildDeclRefExpr(*this, const_cast<VarDecl *>(Var),
-                                 Var->getType().getNonLValueExprType(Context),
-                                 ForLoc, /*RefersToCapture=*/true));
+            SemaRef.MarkDeclarationsReferencedInExpr(buildDeclRefExpr(
+                SemaRef, const_cast<VarDecl *>(Var),
+                Var->getType().getNonLValueExprType(getASTContext()), ForLoc,
+                /*RefersToCapture=*/true));
         }
         OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
         // OpenMP [2.14.1.1, Data-sharing Attribute Rules for Variables
@@ -9372,8 +9406,8 @@ void Sema::ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init) {
                 : OMPC_private;
         if (((isOpenMPSimdDirective(DKind) && DVar.CKind != OMPC_unknown &&
               DVar.CKind != PredeterminedCKind && DVar.RefExpr &&
-              (LangOpts.OpenMP <= 45 || (DVar.CKind != OMPC_lastprivate &&
-                                         DVar.CKind != OMPC_private))) ||
+              (getLangOpts().OpenMP <= 45 || (DVar.CKind != OMPC_lastprivate &&
+                                              DVar.CKind != OMPC_private))) ||
              ((isOpenMPWorksharingDirective(DKind) || DKind == OMPD_taskloop ||
                DKind == OMPD_master_taskloop || DKind == OMPD_masked_taskloop ||
                DKind == OMPD_parallel_master_taskloop ||
@@ -9388,7 +9422,7 @@ void Sema::ActOnOpenMPLoopInitialization(SourceLocation ForLoc, Stmt *Init) {
               << getOpenMPClauseName(PredeterminedCKind);
           if (DVar.RefExpr == nullptr)
             DVar.CKind = PredeterminedCKind;
-          reportOriginalDsa(*this, DSAStack, D, DVar,
+          reportOriginalDsa(SemaRef, DSAStack, D, DVar,
                             /*IsLoopIterVar=*/true);
         } else if (LoopDeclRefExpr) {
           // Make the loop iteration variable private (for worksharing
@@ -9428,7 +9462,7 @@ static bool checkOpenMPIterationSpace(
     unsigned CurrentNestedLoopCount, unsigned NestedLoopCount,
     unsigned TotalNestedLoopCount, Expr *CollapseLoopCountExpr,
     Expr *OrderedLoopCountExpr,
-    Sema::VarsWithInheritedDSAType &VarsWithImplicitDSA,
+    SemaOpenMP::VarsWithInheritedDSAType &VarsWithImplicitDSA,
     llvm::MutableArrayRef<LoopIterationSpace> ResultIterSpaces,
     llvm::MapVector<const Expr *, DeclRefExpr *> &Captures) {
   bool SupportsNonRectangular = !isOpenMPLoopTransformationDirective(DKind);
@@ -9817,7 +9851,7 @@ static unsigned
 checkOpenMPLoop(OpenMPDirectiveKind DKind, Expr *CollapseLoopCountExpr,
                 Expr *OrderedLoopCountExpr, Stmt *AStmt, Sema &SemaRef,
                 DSAStackTy &DSA,
-                Sema::VarsWithInheritedDSAType &VarsWithImplicitDSA,
+                SemaOpenMP::VarsWithInheritedDSAType &VarsWithImplicitDSA,
                 OMPLoopBasedDirective::HelperExprs &Built) {
   unsigned NestedLoopCount = 1;
   bool SupportsNonPerfectlyNested = (SemaRef.LangOpts.OpenMP >= 50) &&
@@ -10566,7 +10600,8 @@ static bool checkGenericLoopLastprivate(Sema &S, ArrayRef<OMPClause *> Clauses,
                                         OpenMPDirectiveKind K,
                                         DSAStackTy *Stack);
 
-bool Sema::checkLastPrivateForMappedDirectives(ArrayRef<OMPClause *> Clauses) {
+bool SemaOpenMP::checkLastPrivateForMappedDirectives(
+    ArrayRef<OMPClause *> Clauses) {
 
   // Check for syntax of lastprivate
   // Param of the lastprivate have different meanings in the mapped directives
@@ -10574,16 +10609,15 @@ bool Sema::checkLastPrivateForMappedDirectives(ArrayRef<OMPClause *> Clauses) {
   //      "omp for"  lastprivate vars must be shared
   if (getLangOpts().OpenMP >= 50 &&
       DSAStack->getMappedDirective() == OMPD_loop &&
-      checkGenericLoopLastprivate(*this, Clauses, OMPD_loop, DSAStack)) {
+      checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_loop, DSAStack)) {
     return false;
   }
   return true;
 }
 
-StmtResult
-Sema::ActOnOpenMPSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                               SourceLocation StartLoc, SourceLocation EndLoc,
-                               VarsWithInheritedDSAType &VarsWithImplicitDSA) {
+StmtResult SemaOpenMP::ActOnOpenMPSimdDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
     return StmtError();
 
@@ -10596,38 +10630,37 @@ Sema::ActOnOpenMPSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_simd, getCollapseNumberExpr(Clauses), getOrderedNumberExpr(Clauses),
-      AStmt, *this, *DSAStack, VarsWithImplicitDSA, B);
+      AStmt, SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   auto *SimdDirective = OMPSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getMappedDirective());
   return SimdDirective;
 }
 
-StmtResult
-Sema::ActOnOpenMPForDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
-                              SourceLocation StartLoc, SourceLocation EndLoc,
-                              VarsWithInheritedDSAType &VarsWithImplicitDSA) {
+StmtResult SemaOpenMP::ActOnOpenMPForDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
     return StmtError();
 
@@ -10640,32 +10673,32 @@ Sema::ActOnOpenMPForDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_for, getCollapseNumberExpr(Clauses), getOrderedNumberExpr(Clauses),
-      AStmt, *this, *DSAStack, VarsWithImplicitDSA, B);
+      AStmt, SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
   auto *ForDirective = OMPForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion(),
       DSAStack->getMappedDirective());
   return ForDirective;
 }
 
-StmtResult Sema::ActOnOpenMPForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -10677,37 +10710,37 @@ StmtResult Sema::ActOnOpenMPForSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_for_simd, getCollapseNumberExpr(Clauses),
-                      getOrderedNumberExpr(Clauses), AStmt, *this, *DSAStack,
+                      getOrderedNumberExpr(Clauses), AStmt, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPForSimdDirective::Create(Context, StartLoc, EndLoc, NestedLoopCount,
-                                     Clauses, AStmt, B);
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPForSimdDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                     NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPSectionsDirective(ArrayRef<OMPClause *> Clauses,
-                                              Stmt *AStmt,
-                                              SourceLocation StartLoc,
-                                              SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPSectionsDirective(ArrayRef<OMPClause *> Clauses,
+                                         Stmt *AStmt, SourceLocation StartLoc,
+                                         SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -10736,23 +10769,23 @@ StmtResult Sema::ActOnOpenMPSectionsDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPSectionsDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
-                                      DSAStack->getTaskgroupReductionRef(),
-                                      DSAStack->isCancelRegion());
+  return OMPSectionsDirective::Create(
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
+      DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPSectionDirective(Stmt *AStmt,
-                                             SourceLocation StartLoc,
-                                             SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPSectionDirective(Stmt *AStmt,
+                                                   SourceLocation StartLoc,
+                                                   SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   DSAStack->setParentCancelRegion(DSAStack->isCancelRegion());
 
-  return OMPSectionDirective::Create(Context, StartLoc, EndLoc, AStmt,
+  return OMPSectionDirective::Create(getASTContext(), StartLoc, EndLoc, AStmt,
                                      DSAStack->isCancelRegion());
 }
 
@@ -10764,10 +10797,10 @@ static Expr *getDirectCallExpr(Expr *E) {
   return nullptr;
 }
 
-StmtResult Sema::ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
-                                              Stmt *AStmt,
-                                              SourceLocation StartLoc,
-                                              SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
+                                         Stmt *AStmt, SourceLocation StartLoc,
+                                         SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -10780,7 +10813,7 @@ StmtResult Sema::ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
 
   SourceLocation TargetCallLoc;
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     Expr *TargetCall = nullptr;
 
     auto *E = dyn_cast<Expr>(S);
@@ -10808,10 +10841,10 @@ StmtResult Sema::ActOnOpenMPDispatchDirective(ArrayRef<OMPClause *> Clauses,
     TargetCallLoc = TargetCall->getExprLoc();
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPDispatchDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
-                                      TargetCallLoc);
+  return OMPDispatchDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                      Clauses, AStmt, TargetCallLoc);
 }
 
 static bool checkGenericLoopLastprivate(Sema &S, ArrayRef<OMPClause *> Clauses,
@@ -10839,7 +10872,7 @@ static bool checkGenericLoopLastprivate(Sema &S, ArrayRef<OMPClause *> Clauses,
   return ErrorFound;
 }
 
-StmtResult Sema::ActOnOpenMPGenericLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPGenericLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -10848,7 +10881,7 @@ StmtResult Sema::ActOnOpenMPGenericLoopDirective(
   // OpenMP 5.1 [2.11.7, loop construct, Restrictions]
   // A list item may not appear in a lastprivate clause unless it is the
   // loop iteration variable of a loop that is associated with the construct.
-  if (checkGenericLoopLastprivate(*this, Clauses, OMPD_loop, DSAStack))
+  if (checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_loop, DSAStack))
     return StmtError();
 
   auto *CS = cast<CapturedStmt>(AStmt);
@@ -10863,19 +10896,19 @@ StmtResult Sema::ActOnOpenMPGenericLoopDirective(
   // In presence of clause 'collapse', it will define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_loop, getCollapseNumberExpr(Clauses), getOrderedNumberExpr(Clauses),
-      AStmt, *this, *DSAStack, VarsWithImplicitDSA, B);
+      AStmt, SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
-  return OMPGenericLoopDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPGenericLoopDirective::Create(getASTContext(), StartLoc, EndLoc,
                                          NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsGenericLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTeamsGenericLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -10884,7 +10917,7 @@ StmtResult Sema::ActOnOpenMPTeamsGenericLoopDirective(
   // OpenMP 5.1 [2.11.7, loop construct, Restrictions]
   // A list item may not appear in a lastprivate clause unless it is the
   // loop iteration variable of a loop that is associated with the construct.
-  if (checkGenericLoopLastprivate(*this, Clauses, OMPD_teams_loop, DSAStack))
+  if (checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_teams_loop, DSAStack))
     return StmtError();
 
   auto *CS = cast<CapturedStmt>(AStmt);
@@ -10909,22 +10942,22 @@ StmtResult Sema::ActOnOpenMPTeamsGenericLoopDirective(
   // In presence of clause 'collapse', it will define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_teams_loop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+                      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
   return OMPTeamsGenericLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsGenericLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsGenericLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -10933,7 +10966,7 @@ StmtResult Sema::ActOnOpenMPTargetTeamsGenericLoopDirective(
   // OpenMP 5.1 [2.11.7, loop construct, Restrictions]
   // A list item may not appear in a lastprivate clause unless it is the
   // loop iteration variable of a loop that is associated with the construct.
-  if (checkGenericLoopLastprivate(*this, Clauses, OMPD_target_teams_loop,
+  if (checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_target_teams_loop,
                                   DSAStack))
     return StmtError();
 
@@ -10959,22 +10992,22 @@ StmtResult Sema::ActOnOpenMPTargetTeamsGenericLoopDirective(
   // In presence of clause 'collapse', it will define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_target_teams_loop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+                      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPTargetTeamsGenericLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
-      teamsLoopCanBeParallelFor(AStmt, *this));
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      teamsLoopCanBeParallelFor(AStmt, SemaRef));
 }
 
-StmtResult Sema::ActOnOpenMPParallelGenericLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelGenericLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -10983,7 +11016,8 @@ StmtResult Sema::ActOnOpenMPParallelGenericLoopDirective(
   // OpenMP 5.1 [2.11.7, loop construct, Restrictions]
   // A list item may not appear in a lastprivate clause unless it is the
   // loop iteration variable of a loop that is associated with the construct.
-  if (checkGenericLoopLastprivate(*this, Clauses, OMPD_parallel_loop, DSAStack))
+  if (checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_parallel_loop,
+                                  DSAStack))
     return StmtError();
 
   auto *CS = cast<CapturedStmt>(AStmt);
@@ -11008,21 +11042,21 @@ StmtResult Sema::ActOnOpenMPParallelGenericLoopDirective(
   // In presence of clause 'collapse', it will define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_parallel_loop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+                      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPParallelGenericLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetParallelGenericLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetParallelGenericLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -11031,7 +11065,7 @@ StmtResult Sema::ActOnOpenMPTargetParallelGenericLoopDirective(
   // OpenMP 5.1 [2.11.7, loop construct, Restrictions]
   // A list item may not appear in a lastprivate clause unless it is the
   // loop iteration variable of a loop that is associated with the construct.
-  if (checkGenericLoopLastprivate(*this, Clauses, OMPD_target_parallel_loop,
+  if (checkGenericLoopLastprivate(SemaRef, Clauses, OMPD_target_parallel_loop,
                                   DSAStack))
     return StmtError();
 
@@ -11057,30 +11091,30 @@ StmtResult Sema::ActOnOpenMPTargetParallelGenericLoopDirective(
   // In presence of clause 'collapse', it will define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_target_parallel_loop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+                      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPTargetParallelGenericLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPSingleDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPSingleDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
   assert(isa<CapturedStmt>(AStmt) && "Captured statement expected");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   // OpenMP [2.7.3, single Construct, Restrictions]
   // The copyprivate clause must not be used with the nowait clause.
@@ -11099,33 +11133,35 @@ StmtResult Sema::ActOnOpenMPSingleDirective(ArrayRef<OMPClause *> Clauses,
     }
   }
 
-  return OMPSingleDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPSingleDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                    AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPMasterDirective(Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPMasterDirective(Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPMasterDirective::Create(Context, StartLoc, EndLoc, AStmt);
+  return OMPMasterDirective::Create(getASTContext(), StartLoc, EndLoc, AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPMaskedDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPMaskedDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPMaskedDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPMaskedDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                    AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPCriticalDirective(
+StmtResult SemaOpenMP::ActOnOpenMPCriticalDirective(
     const DeclarationNameInfo &DirName, ArrayRef<OMPClause *> Clauses,
     Stmt *AStmt, SourceLocation StartLoc, SourceLocation EndLoc) {
   if (!AStmt)
@@ -11146,7 +11182,7 @@ StmtResult Sema::ActOnOpenMPCriticalDirective(
           E->isInstantiationDependent()) {
         DependentHint = true;
       } else {
-        Hint = E->EvaluateKnownConstInt(Context);
+        Hint = E->EvaluateKnownConstInt(getASTContext());
         HintLoc = C->getBeginLoc();
       }
     }
@@ -11165,7 +11201,7 @@ StmtResult Sema::ActOnOpenMPCriticalDirective(
       if (const auto *C = Pair.first->getSingleClause<OMPHintClause>()) {
         Diag(C->getBeginLoc(), diag::note_omp_critical_hint_here)
             << 1
-            << toString(C->getHint()->EvaluateKnownConstInt(Context),
+            << toString(C->getHint()->EvaluateKnownConstInt(getASTContext()),
                         /*Radix=*/10, /*Signed=*/false);
       } else {
         Diag(Pair.first->getBeginLoc(), diag::note_omp_critical_no_hint) << 1;
@@ -11173,16 +11209,16 @@ StmtResult Sema::ActOnOpenMPCriticalDirective(
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  auto *Dir = OMPCriticalDirective::Create(Context, DirName, StartLoc, EndLoc,
-                                           Clauses, AStmt);
+  auto *Dir = OMPCriticalDirective::Create(getASTContext(), DirName, StartLoc,
+                                           EndLoc, Clauses, AStmt);
   if (!Pair.first && DirName.getName() && !DependentHint)
     DSAStack->addCriticalWithHint(Dir, Hint);
   return Dir;
 }
 
-StmtResult Sema::ActOnOpenMPParallelForDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelForDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -11201,32 +11237,32 @@ StmtResult Sema::ActOnOpenMPParallelForDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_parallel_for, getCollapseNumberExpr(Clauses),
-                      getOrderedNumberExpr(Clauses), AStmt, *this, *DSAStack,
+                      getOrderedNumberExpr(Clauses), AStmt, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp parallel for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPParallelForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -11245,34 +11281,33 @@ StmtResult Sema::ActOnOpenMPParallelForSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_parallel_for_simd, getCollapseNumberExpr(Clauses),
-                      getOrderedNumberExpr(Clauses), AStmt, *this, *DSAStack,
+                      getOrderedNumberExpr(Clauses), AStmt, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelForSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult
-Sema::ActOnOpenMPParallelMasterDirective(ArrayRef<OMPClause *> Clauses,
-                                         Stmt *AStmt, SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPParallelMasterDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -11285,17 +11320,16 @@ Sema::ActOnOpenMPParallelMasterDirective(ArrayRef<OMPClause *> Clauses,
   // longjmp() and throw() must not violate the entry/exit criteria.
   CS->getCapturedDecl()->setNothrow();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPParallelMasterDirective::Create(
-      Context, StartLoc, EndLoc, Clauses, AStmt,
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
       DSAStack->getTaskgroupReductionRef());
 }
 
-StmtResult
-Sema::ActOnOpenMPParallelMaskedDirective(ArrayRef<OMPClause *> Clauses,
-                                         Stmt *AStmt, SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPParallelMaskedDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -11308,17 +11342,16 @@ Sema::ActOnOpenMPParallelMaskedDirective(ArrayRef<OMPClause *> Clauses,
   // longjmp() and throw() must not violate the entry/exit criteria.
   CS->getCapturedDecl()->setNothrow();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPParallelMaskedDirective::Create(
-      Context, StartLoc, EndLoc, Clauses, AStmt,
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
       DSAStack->getTaskgroupReductionRef());
 }
 
-StmtResult
-Sema::ActOnOpenMPParallelSectionsDirective(ArrayRef<OMPClause *> Clauses,
-                                           Stmt *AStmt, SourceLocation StartLoc,
-                                           SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPParallelSectionsDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -11348,10 +11381,10 @@ Sema::ActOnOpenMPParallelSectionsDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPParallelSectionsDirective::Create(
-      Context, StartLoc, EndLoc, Clauses, AStmt,
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
@@ -11378,16 +11411,17 @@ static bool checkMutuallyExclusiveClauses(
   return ErrorFound;
 }
 
-StmtResult Sema::ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
+                                                Stmt *AStmt,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
   // OpenMP 5.0, 2.10.1 task Construct
   // If a detach clause appears on the directive, then a mergeable clause cannot
   // appear on the same directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_detach, OMPC_mergeable}))
     return StmtError();
 
@@ -11399,26 +11433,26 @@ StmtResult Sema::ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
   // longjmp() and throw() must not violate the entry/exit criteria.
   CS->getCapturedDecl()->setNothrow();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPTaskDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
-                                  DSAStack->isCancelRegion());
+  return OMPTaskDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                  AStmt, DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPTaskyieldDirective(SourceLocation StartLoc,
-                                               SourceLocation EndLoc) {
-  return OMPTaskyieldDirective::Create(Context, StartLoc, EndLoc);
+StmtResult SemaOpenMP::ActOnOpenMPTaskyieldDirective(SourceLocation StartLoc,
+                                                     SourceLocation EndLoc) {
+  return OMPTaskyieldDirective::Create(getASTContext(), StartLoc, EndLoc);
 }
 
-StmtResult Sema::ActOnOpenMPBarrierDirective(SourceLocation StartLoc,
-                                             SourceLocation EndLoc) {
-  return OMPBarrierDirective::Create(Context, StartLoc, EndLoc);
+StmtResult SemaOpenMP::ActOnOpenMPBarrierDirective(SourceLocation StartLoc,
+                                                   SourceLocation EndLoc) {
+  return OMPBarrierDirective::Create(getASTContext(), StartLoc, EndLoc);
 }
 
-StmtResult Sema::ActOnOpenMPErrorDirective(ArrayRef<OMPClause *> Clauses,
-                                           SourceLocation StartLoc,
-                                           SourceLocation EndLoc,
-                                           bool InExContext) {
+StmtResult SemaOpenMP::ActOnOpenMPErrorDirective(ArrayRef<OMPClause *> Clauses,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation EndLoc,
+                                                 bool InExContext) {
   const OMPAtClause *AtC =
       OMPExecutableDirective::getSingleClause<OMPAtClause>(Clauses);
 
@@ -11443,12 +11477,13 @@ StmtResult Sema::ActOnOpenMPErrorDirective(ArrayRef<OMPClause *> Clauses,
     if (!SeverityC || SeverityC->getSeverityKind() != OMPC_SEVERITY_warning)
       return StmtError();
   }
-  return OMPErrorDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPErrorDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses);
 }
 
-StmtResult Sema::ActOnOpenMPTaskwaitDirective(ArrayRef<OMPClause *> Clauses,
-                                              SourceLocation StartLoc,
-                                              SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPTaskwaitDirective(ArrayRef<OMPClause *> Clauses,
+                                         SourceLocation StartLoc,
+                                         SourceLocation EndLoc) {
   const OMPNowaitClause *NowaitC =
       OMPExecutableDirective::getSingleClause<OMPNowaitClause>(Clauses);
   bool HasDependC =
@@ -11459,28 +11494,29 @@ StmtResult Sema::ActOnOpenMPTaskwaitDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  return OMPTaskwaitDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPTaskwaitDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                      Clauses);
 }
 
-StmtResult Sema::ActOnOpenMPTaskgroupDirective(ArrayRef<OMPClause *> Clauses,
-                                               Stmt *AStmt,
-                                               SourceLocation StartLoc,
-                                               SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPTaskgroupDirective(ArrayRef<OMPClause *> Clauses,
+                                          Stmt *AStmt, SourceLocation StartLoc,
+                                          SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
   assert(isa<CapturedStmt>(AStmt) && "Captured statement expected");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPTaskgroupDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                       AStmt,
+  return OMPTaskgroupDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                       Clauses, AStmt,
                                        DSAStack->getTaskgroupReductionRef());
 }
 
-StmtResult Sema::ActOnOpenMPFlushDirective(ArrayRef<OMPClause *> Clauses,
-                                           SourceLocation StartLoc,
-                                           SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPFlushDirective(ArrayRef<OMPClause *> Clauses,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation EndLoc) {
   OMPFlushClause *FC = nullptr;
   OMPClause *OrderClause = nullptr;
   for (OMPClause *C : Clauses) {
@@ -11514,12 +11550,12 @@ StmtResult Sema::ActOnOpenMPFlushDirective(ArrayRef<OMPClause *> Clauses,
         << getOpenMPClauseName(OrderClause->getClauseKind());
     return StmtError();
   }
-  return OMPFlushDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPFlushDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses);
 }
 
-StmtResult Sema::ActOnOpenMPDepobjDirective(ArrayRef<OMPClause *> Clauses,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPDepobjDirective(ArrayRef<OMPClause *> Clauses,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
   if (Clauses.empty()) {
     Diag(StartLoc, diag::err_omp_depobj_expected);
     return StmtError();
@@ -11536,12 +11572,12 @@ StmtResult Sema::ActOnOpenMPDepobjDirective(ArrayRef<OMPClause *> Clauses,
     Diag(Clauses[0]->getEndLoc(), diag::err_omp_depobj_single_clause_expected);
     return StmtError();
   }
-  return OMPDepobjDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPDepobjDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses);
 }
 
-StmtResult Sema::ActOnOpenMPScanDirective(ArrayRef<OMPClause *> Clauses,
-                                          SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPScanDirective(ArrayRef<OMPClause *> Clauses,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
   // Check that exactly one clause is specified.
   if (Clauses.size() != 1) {
     Diag(Clauses.empty() ? EndLoc : Clauses[1]->getBeginLoc(),
@@ -11566,13 +11602,13 @@ StmtResult Sema::ActOnOpenMPScanDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
   DSAStack->setParentHasScanDirective(StartLoc);
-  return OMPScanDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPScanDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses);
 }
 
-StmtResult Sema::ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
-                                             Stmt *AStmt,
-                                             SourceLocation StartLoc,
-                                             SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc) {
   const OMPClause *DependFound = nullptr;
   const OMPClause *DependSourceClause = nullptr;
   const OMPClause *DependSinkClause = nullptr;
@@ -11631,7 +11667,7 @@ StmtResult Sema::ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
     // An ordered construct with the simd clause is the only OpenMP construct
     // that can appear in the simd region.
     Diag(StartLoc, diag::err_omp_prohibited_region_simd)
-        << (LangOpts.OpenMP >= 50 ? 1 : 0);
+        << (getLangOpts().OpenMP >= 50 ? 1 : 0);
     ErrorFound = true;
   } else if ((DependFound || DoacrossFound) && (TC || SC)) {
     SourceLocation Loc =
@@ -11678,10 +11714,11 @@ StmtResult Sema::ActOnOpenMPOrderedDirective(ArrayRef<OMPClause *> Clauses,
   if (AStmt) {
     assert(isa<CapturedStmt>(AStmt) && "Captured statement expected");
 
-    setFunctionHasBranchProtectedScope();
+    SemaRef.setFunctionHasBranchProtectedScope();
   }
 
-  return OMPOrderedDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPOrderedDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                     AStmt);
 }
 
 namespace {
@@ -12739,10 +12776,11 @@ bool OpenMPAtomicCompareCaptureChecker::checkStmt(Stmt *S,
 }
 } // namespace
 
-StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
+  ASTContext &Context = getASTContext();
   // Register location of the first atomic directive.
   DSAStack->addAtomicDirectiveLoc(StartLoc);
   if (!AStmt)
@@ -12945,7 +12983,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
           << ErrorFound << NoteRange;
       return StmtError();
     }
-    if (CurContext->isDependentContext())
+    if (SemaRef.CurContext->isDependentContext())
       V = X = nullptr;
   } else if (AtomicKind == OMPC_write) {
     enum {
@@ -13007,7 +13045,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
           << ErrorFound << NoteRange;
       return StmtError();
     }
-    if (CurContext->isDependentContext())
+    if (SemaRef.CurContext->isDependentContext())
       E = X = nullptr;
   } else if (AtomicKind == OMPC_update || AtomicKind == OMPC_unknown) {
     // If clause is update:
@@ -13018,7 +13056,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
     //  x binop= expr;
     //  x = x binop expr;
     //  x = expr binop x;
-    OpenMPAtomicUpdateChecker Checker(*this);
+    OpenMPAtomicUpdateChecker Checker(SemaRef);
     if (Checker.checkStatement(
             Body,
             (AtomicKind == OMPC_update)
@@ -13026,7 +13064,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
                 : diag::err_omp_atomic_not_expression_statement,
             diag::note_omp_atomic_update))
       return StmtError();
-    if (!CurContext->isDependentContext()) {
+    if (!SemaRef.CurContext->isDependentContext()) {
       E = Checker.getExpr();
       X = Checker.getX();
       UE = Checker.getUpdateExpr();
@@ -13056,7 +13094,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
       if (AtomicBinOp && AtomicBinOp->getOpcode() == BO_Assign) {
         V = AtomicBinOp->getLHS();
         Body = AtomicBinOp->getRHS()->IgnoreParenImpCasts();
-        OpenMPAtomicUpdateChecker Checker(*this);
+        OpenMPAtomicUpdateChecker Checker(SemaRef);
         if (Checker.checkStatement(
                 Body, diag::err_omp_atomic_capture_not_expression_statement,
                 diag::note_omp_atomic_update))
@@ -13081,7 +13119,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
         Diag(NoteLoc, diag::note_omp_atomic_capture) << ErrorFound << NoteRange;
         return StmtError();
       }
-      if (CurContext->isDependentContext())
+      if (SemaRef.CurContext->isDependentContext())
         UE = V = E = X = nullptr;
     } else {
       // If clause is a capture:
@@ -13110,14 +13148,14 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
           if (auto *EWC = dyn_cast<ExprWithCleanups>(Second))
             Second = EWC->getSubExpr()->IgnoreParenImpCasts();
           // Need to find what subexpression is 'v' and what is 'x'.
-          OpenMPAtomicUpdateChecker Checker(*this);
+          OpenMPAtomicUpdateChecker Checker(SemaRef);
           bool IsUpdateExprFound = !Checker.checkStatement(Second);
           BinaryOperator *BinOp = nullptr;
           if (IsUpdateExprFound) {
             BinOp = dyn_cast<BinaryOperator>(First);
             IsUpdateExprFound = BinOp && BinOp->getOpcode() == BO_Assign;
           }
-          if (IsUpdateExprFound && !CurContext->isDependentContext()) {
+          if (IsUpdateExprFound && !SemaRef.CurContext->isDependentContext()) {
             //  { v = x; x++; }
             //  { v = x; x--; }
             //  { v = x; ++x; }
@@ -13147,7 +13185,8 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
               BinOp = dyn_cast<BinaryOperator>(Second);
               IsUpdateExprFound = BinOp && BinOp->getOpcode() == BO_Assign;
             }
-            if (IsUpdateExprFound && !CurContext->isDependentContext()) {
+            if (IsUpdateExprFound &&
+                !SemaRef.CurContext->isDependentContext()) {
               //  { x++; v = x; }
               //  { x--; v = x; }
               //  { ++x; v = x; }
@@ -13244,12 +13283,12 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
       Diag(NoteLoc, diag::note_omp_atomic_capture) << ErrorFound << NoteRange;
       return StmtError();
     }
-    if (CurContext->isDependentContext())
+    if (SemaRef.CurContext->isDependentContext())
       UE = V = E = X = nullptr;
   } else if (AtomicKind == OMPC_compare) {
     if (IsCompareCapture) {
       OpenMPAtomicCompareCaptureChecker::ErrorInfoTy ErrorInfo;
-      OpenMPAtomicCompareCaptureChecker Checker(*this);
+      OpenMPAtomicCompareCaptureChecker Checker(SemaRef);
       if (!Checker.checkStmt(Body, ErrorInfo)) {
         Diag(ErrorInfo.ErrorLoc, diag::err_omp_atomic_compare_capture)
             << ErrorInfo.ErrorRange;
@@ -13269,7 +13308,7 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
       IsPostfixUpdate = Checker.isPostfixUpdate();
     } else {
       OpenMPAtomicCompareChecker::ErrorInfoTy ErrorInfo;
-      OpenMPAtomicCompareChecker Checker(*this);
+      OpenMPAtomicCompareChecker Checker(SemaRef);
       if (!Checker.checkStmt(Body, ErrorInfo)) {
         Diag(ErrorInfo.ErrorLoc, diag::err_omp_atomic_compare)
             << ErrorInfo.ErrorRange;
@@ -13307,17 +13346,17 @@ StmtResult Sema::ActOnOpenMPAtomicDirective(ArrayRef<OMPClause *> Clauses,
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPAtomicDirective::Create(
       Context, StartLoc, EndLoc, Clauses, AStmt,
       {X, V, R, E, UE, D, CE, IsXLHSInRHSPart, IsPostfixUpdate, IsFailOnly});
 }
 
-StmtResult Sema::ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -13374,15 +13413,15 @@ StmtResult Sema::ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPTargetDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPTargetDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                    AStmt);
 }
 
-StmtResult
-Sema::ActOnOpenMPTargetParallelDirective(ArrayRef<OMPClause *> Clauses,
-                                         Stmt *AStmt, SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetParallelDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -13404,14 +13443,14 @@ Sema::ActOnOpenMPTargetParallelDirective(ArrayRef<OMPClause *> Clauses,
     CS->getCapturedDecl()->setNothrow();
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   return OMPTargetParallelDirective::Create(
-      Context, StartLoc, EndLoc, Clauses, AStmt,
+      getASTContext(), StartLoc, EndLoc, Clauses, AStmt,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPTargetParallelForDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetParallelForDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13440,28 +13479,28 @@ StmtResult Sema::ActOnOpenMPTargetParallelForDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_target_parallel_for, getCollapseNumberExpr(Clauses),
-                      getOrderedNumberExpr(Clauses), CS, *this, *DSAStack,
+                      getOrderedNumberExpr(Clauses), CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target parallel for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetParallelForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
@@ -13498,10 +13537,10 @@ static bool isClauseMappable(ArrayRef<OMPClause *> Clauses) {
   return true;
 }
 
-StmtResult Sema::ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
-                                                Stmt *AStmt,
-                                                SourceLocation StartLoc,
-                                                SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
+                                           Stmt *AStmt, SourceLocation StartLoc,
+                                           SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -13511,9 +13550,10 @@ StmtResult Sema::ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
   // At least one map, use_device_addr or use_device_ptr clause must appear on
   // the directive.
   if (!hasClauses(Clauses, OMPC_map, OMPC_use_device_ptr) &&
-      (LangOpts.OpenMP < 50 || !hasClauses(Clauses, OMPC_use_device_addr))) {
+      (getLangOpts().OpenMP < 50 ||
+       !hasClauses(Clauses, OMPC_use_device_addr))) {
     StringRef Expected;
-    if (LangOpts.OpenMP < 50)
+    if (getLangOpts().OpenMP < 50)
       Expected = "'map' or 'use_device_ptr'";
     else
       Expected = "'map', 'use_device_ptr', or 'use_device_addr'";
@@ -13522,16 +13562,15 @@ StmtResult Sema::ActOnOpenMPTargetDataDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPTargetDataDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                        AStmt);
+  return OMPTargetDataDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                        Clauses, AStmt);
 }
 
-StmtResult
-Sema::ActOnOpenMPTargetEnterDataDirective(ArrayRef<OMPClause *> Clauses,
-                                          SourceLocation StartLoc,
-                                          SourceLocation EndLoc, Stmt *AStmt) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetEnterDataDirective(
+    ArrayRef<OMPClause *> Clauses, SourceLocation StartLoc,
+    SourceLocation EndLoc, Stmt *AStmt) {
   if (!AStmt)
     return StmtError();
 
@@ -13561,14 +13600,13 @@ Sema::ActOnOpenMPTargetEnterDataDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  return OMPTargetEnterDataDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                             AStmt);
+  return OMPTargetEnterDataDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                             Clauses, AStmt);
 }
 
-StmtResult
-Sema::ActOnOpenMPTargetExitDataDirective(ArrayRef<OMPClause *> Clauses,
-                                         SourceLocation StartLoc,
-                                         SourceLocation EndLoc, Stmt *AStmt) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetExitDataDirective(
+    ArrayRef<OMPClause *> Clauses, SourceLocation StartLoc,
+    SourceLocation EndLoc, Stmt *AStmt) {
   if (!AStmt)
     return StmtError();
 
@@ -13598,14 +13636,13 @@ Sema::ActOnOpenMPTargetExitDataDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  return OMPTargetExitDataDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                            AStmt);
+  return OMPTargetExitDataDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                            Clauses, AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPTargetUpdateDirective(ArrayRef<OMPClause *> Clauses,
-                                                  SourceLocation StartLoc,
-                                                  SourceLocation EndLoc,
-                                                  Stmt *AStmt) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetUpdateDirective(
+    ArrayRef<OMPClause *> Clauses, SourceLocation StartLoc,
+    SourceLocation EndLoc, Stmt *AStmt) {
   if (!AStmt)
     return StmtError();
 
@@ -13637,13 +13674,14 @@ StmtResult Sema::ActOnOpenMPTargetUpdateDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  return OMPTargetUpdateDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                          AStmt);
+  return OMPTargetUpdateDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                          Clauses, AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsDirective(ArrayRef<OMPClause *> Clauses,
-                                           Stmt *AStmt, SourceLocation StartLoc,
-                                           SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTeamsDirective(ArrayRef<OMPClause *> Clauses,
+                                                 Stmt *AStmt,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -13659,17 +13697,17 @@ StmtResult Sema::ActOnOpenMPTeamsDirective(ArrayRef<OMPClause *> Clauses,
   // longjmp() and throw() must not violate the entry/exit criteria.
   CS->getCapturedDecl()->setNothrow();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
-  return OMPTeamsDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPTeamsDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                   AStmt);
 }
 
-StmtResult
-Sema::ActOnOpenMPCancellationPointDirective(SourceLocation StartLoc,
-                                            SourceLocation EndLoc,
-                                            OpenMPDirectiveKind CancelRegion) {
+StmtResult SemaOpenMP::ActOnOpenMPCancellationPointDirective(
+    SourceLocation StartLoc, SourceLocation EndLoc,
+    OpenMPDirectiveKind CancelRegion) {
   if (DSAStack->isParentNowaitRegion()) {
     Diag(StartLoc, diag::err_omp_parent_cancel_region_nowait) << 0;
     return StmtError();
@@ -13678,14 +13716,13 @@ Sema::ActOnOpenMPCancellationPointDirective(SourceLocation StartLoc,
     Diag(StartLoc, diag::err_omp_parent_cancel_region_ordered) << 0;
     return StmtError();
   }
-  return OMPCancellationPointDirective::Create(Context, StartLoc, EndLoc,
-                                               CancelRegion);
+  return OMPCancellationPointDirective::Create(getASTContext(), StartLoc,
+                                               EndLoc, CancelRegion);
 }
 
-StmtResult Sema::ActOnOpenMPCancelDirective(ArrayRef<OMPClause *> Clauses,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc,
-                                            OpenMPDirectiveKind CancelRegion) {
+StmtResult SemaOpenMP::ActOnOpenMPCancelDirective(
+    ArrayRef<OMPClause *> Clauses, SourceLocation StartLoc,
+    SourceLocation EndLoc, OpenMPDirectiveKind CancelRegion) {
   if (DSAStack->isParentNowaitRegion()) {
     Diag(StartLoc, diag::err_omp_parent_cancel_region_nowait) << 1;
     return StmtError();
@@ -13695,7 +13732,7 @@ StmtResult Sema::ActOnOpenMPCancelDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
   DSAStack->setParentCancelRegion(/*Cancel=*/true);
-  return OMPCancelDirective::Create(Context, StartLoc, EndLoc, Clauses,
+  return OMPCancelDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
                                     CancelRegion);
 }
 
@@ -13726,7 +13763,7 @@ static bool checkReductionClauseWithNogroup(Sema &S,
   return false;
 }
 
-StmtResult Sema::ActOnOpenMPTaskLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTaskLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13738,33 +13775,33 @@ StmtResult Sema::ActOnOpenMPTaskLoopDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_taskloop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPTaskLoopDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPTaskLoopDirective::Create(getASTContext(), StartLoc, EndLoc,
                                       NestedLoopCount, Clauses, AStmt, B,
                                       DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPTaskLoopSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTaskLoopSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13776,21 +13813,21 @@ StmtResult Sema::ActOnOpenMPTaskLoopSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_taskloop_simd, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
@@ -13798,23 +13835,23 @@ StmtResult Sema::ActOnOpenMPTaskLoopSimdDirective(
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPTaskLoopSimdDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPTaskLoopSimdDirective::Create(getASTContext(), StartLoc, EndLoc,
                                           NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPMasterTaskLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPMasterTaskLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13826,33 +13863,33 @@ StmtResult Sema::ActOnOpenMPMasterTaskLoopDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_master_taskloop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPMasterTaskLoopDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPMasterTaskLoopDirective::Create(getASTContext(), StartLoc, EndLoc,
                                             NestedLoopCount, Clauses, AStmt, B,
                                             DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPMaskedTaskLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPMaskedTaskLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13864,33 +13901,33 @@ StmtResult Sema::ActOnOpenMPMaskedTaskLoopDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_masked_taskloop, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPMaskedTaskLoopDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPMaskedTaskLoopDirective::Create(getASTContext(), StartLoc, EndLoc,
                                             NestedLoopCount, Clauses, AStmt, B,
                                             DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPMasterTaskLoopSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPMasterTaskLoopSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13902,21 +13939,21 @@ StmtResult Sema::ActOnOpenMPMasterTaskLoopSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_master_taskloop_simd, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
@@ -13924,23 +13961,23 @@ StmtResult Sema::ActOnOpenMPMasterTaskLoopSimdDirective(
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPMasterTaskLoopSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPMaskedTaskLoopSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPMaskedTaskLoopSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -13952,21 +13989,21 @@ StmtResult Sema::ActOnOpenMPMaskedTaskLoopSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_masked_taskloop_simd, getCollapseNumberExpr(Clauses),
-                      /*OrderedLoopCountExpr=*/nullptr, AStmt, *this, *DSAStack,
-                      VarsWithImplicitDSA, B);
+                      /*OrderedLoopCountExpr=*/nullptr, AStmt, SemaRef,
+                      *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
@@ -13974,23 +14011,23 @@ StmtResult Sema::ActOnOpenMPMaskedTaskLoopSimdDirective(
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPMaskedTaskLoopSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPParallelMasterTaskLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelMasterTaskLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14021,33 +14058,33 @@ StmtResult Sema::ActOnOpenMPParallelMasterTaskLoopDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_parallel_master_taskloop, getCollapseNumberExpr(Clauses),
-      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelMasterTaskLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPParallelMaskedTaskLoopDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelMaskedTaskLoopDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14078,33 +14115,33 @@ StmtResult Sema::ActOnOpenMPParallelMaskedTaskLoopDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_parallel_masked_taskloop, getCollapseNumberExpr(Clauses),
-      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelMaskedTaskLoopDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPParallelMasterTaskLoopSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelMasterTaskLoopSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14135,21 +14172,21 @@ StmtResult Sema::ActOnOpenMPParallelMasterTaskLoopSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_parallel_master_taskloop_simd, getCollapseNumberExpr(Clauses),
-      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
@@ -14157,23 +14194,23 @@ StmtResult Sema::ActOnOpenMPParallelMasterTaskLoopSimdDirective(
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelMasterTaskLoopSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14204,21 +14241,21 @@ StmtResult Sema::ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_parallel_masked_taskloop_simd, getCollapseNumberExpr(Clauses),
-      /*OrderedLoopCountExpr=*/nullptr, CS, *this, *DSAStack,
+      /*OrderedLoopCountExpr=*/nullptr, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
@@ -14226,23 +14263,23 @@ StmtResult Sema::ActOnOpenMPParallelMaskedTaskLoopSimdDirective(
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkMutuallyExclusiveClauses(*this, Clauses,
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
                                     {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
-  if (checkReductionClauseWithNogroup(*this, Clauses))
+  if (checkReductionClauseWithNogroup(SemaRef, Clauses))
     return StmtError();
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPParallelMaskedTaskLoopSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPDistributeDirective(
+StmtResult SemaOpenMP::ActOnOpenMPDistributeDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14258,21 +14295,21 @@ StmtResult Sema::ActOnOpenMPDistributeDirective(
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_distribute, getCollapseNumberExpr(Clauses),
                       nullptr /*ordered not a clause on distribute*/, AStmt,
-                      *this, *DSAStack, VarsWithImplicitDSA, B);
+                      SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   auto *DistributeDirective = OMPDistributeDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getMappedDirective());
   return DistributeDirective;
 }
 
-StmtResult Sema::ActOnOpenMPDistributeParallelForDirective(
+StmtResult SemaOpenMP::ActOnOpenMPDistributeParallelForDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14302,21 +14339,21 @@ StmtResult Sema::ActOnOpenMPDistributeParallelForDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_distribute_parallel_for, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPDistributeParallelForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPDistributeParallelForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPDistributeParallelForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14346,34 +14383,34 @@ StmtResult Sema::ActOnOpenMPDistributeParallelForSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_distribute_parallel_for_simd, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPDistributeParallelForSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPDistributeSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPDistributeSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14402,34 +14439,34 @@ StmtResult Sema::ActOnOpenMPDistributeSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_distribute_simd, getCollapseNumberExpr(Clauses),
-                      nullptr /*ordered not a clause on distribute*/, CS, *this,
-                      *DSAStack, VarsWithImplicitDSA, B);
+                      nullptr /*ordered not a clause on distribute*/, CS,
+                      SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPDistributeSimdDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPDistributeSimdDirective::Create(getASTContext(), StartLoc, EndLoc,
                                             NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetParallelForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetParallelForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14459,33 +14496,33 @@ StmtResult Sema::ActOnOpenMPTargetParallelForSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_target_parallel_for_simd, getCollapseNumberExpr(Clauses),
-      getOrderedNumberExpr(Clauses), CS, *this, *DSAStack, VarsWithImplicitDSA,
-      B);
+      getOrderedNumberExpr(Clauses), CS, SemaRef, *DSAStack,
+      VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target parallel for simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetParallelForSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14514,34 +14551,34 @@ StmtResult Sema::ActOnOpenMPTargetSimdDirective(
   // nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_target_simd, getCollapseNumberExpr(Clauses),
-                      getOrderedNumberExpr(Clauses), CS, *this, *DSAStack,
+                      getOrderedNumberExpr(Clauses), CS, SemaRef, *DSAStack,
                       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
-  return OMPTargetSimdDirective::Create(Context, StartLoc, EndLoc,
+  SemaRef.setFunctionHasBranchProtectedScope();
+  return OMPTargetSimdDirective::Create(getASTContext(), StartLoc, EndLoc,
                                         NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsDistributeDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTeamsDistributeDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14570,23 +14607,23 @@ StmtResult Sema::ActOnOpenMPTeamsDistributeDirective(
   // define the nested loops number.
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_teams_distribute, getCollapseNumberExpr(Clauses),
-                      nullptr /*ordered not a clause on distribute*/, CS, *this,
-                      *DSAStack, VarsWithImplicitDSA, B);
+                      nullptr /*ordered not a clause on distribute*/, CS,
+                      SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp teams distribute loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
   return OMPTeamsDistributeDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsDistributeSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTeamsDistributeSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14616,38 +14653,38 @@ StmtResult Sema::ActOnOpenMPTeamsDistributeSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_teams_distribute_simd, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
 
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp teams distribute simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
   return OMPTeamsDistributeSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsDistributeParallelForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTeamsDistributeParallelForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14678,38 +14715,38 @@ StmtResult Sema::ActOnOpenMPTeamsDistributeParallelForSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_teams_distribute_parallel_for_simd, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
 
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
   return OMPTeamsDistributeParallelForSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTeamsDistributeParallelForDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTeamsDistributeParallelForDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14740,28 +14777,27 @@ StmtResult Sema::ActOnOpenMPTeamsDistributeParallelForDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_teams_distribute_parallel_for, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
 
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   DSAStack->setParentTeamsRegionLoc(StartLoc);
 
   return OMPTeamsDistributeParallelForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsDirective(ArrayRef<OMPClause *> Clauses,
-                                                 Stmt *AStmt,
-                                                 SourceLocation StartLoc,
-                                                 SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
@@ -14783,7 +14819,7 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDirective(ArrayRef<OMPClause *> Clauses,
     // longjmp() and throw() must not violate the entry/exit criteria.
     CS->getCapturedDecl()->setNothrow();
   }
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   const OMPClause *BareClause = nullptr;
   bool HasThreadLimitAndNumTeamsClause = hasClauses(Clauses, OMPC_num_teams) &&
@@ -14798,11 +14834,11 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
   }
 
-  return OMPTargetTeamsDirective::Create(Context, StartLoc, EndLoc, Clauses,
-                                         AStmt);
+  return OMPTargetTeamsDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                         Clauses, AStmt);
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsDistributeDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsDistributeDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14832,20 +14868,20 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDistributeDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_target_teams_distribute, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target teams distribute loop exprs were not built");
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetTeamsDistributeDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsDistributeParallelForDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsDistributeParallelForDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14875,32 +14911,32 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDistributeParallelForDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_target_teams_distribute_parallel_for, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target teams distribute parallel for loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetTeamsDistributeParallelForDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsDistributeParallelForSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsDistributeParallelForSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14931,35 +14967,35 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDistributeParallelForSimdDirective(
   unsigned NestedLoopCount =
       checkOpenMPLoop(OMPD_target_teams_distribute_parallel_for_simd,
                       getCollapseNumberExpr(Clauses),
-                      nullptr /*ordered not a clause on distribute*/, CS, *this,
-                      *DSAStack, VarsWithImplicitDSA, B);
+                      nullptr /*ordered not a clause on distribute*/, CS,
+                      SemaRef, *DSAStack, VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target teams distribute parallel for simd loop exprs were not "
          "built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetTeamsDistributeParallelForSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-StmtResult Sema::ActOnOpenMPTargetTeamsDistributeSimdDirective(
+StmtResult SemaOpenMP::ActOnOpenMPTargetTeamsDistributeSimdDirective(
     ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
     SourceLocation EndLoc, VarsWithInheritedDSAType &VarsWithImplicitDSA) {
   if (!AStmt)
@@ -14989,34 +15025,34 @@ StmtResult Sema::ActOnOpenMPTargetTeamsDistributeSimdDirective(
   // define the nested loops number.
   unsigned NestedLoopCount = checkOpenMPLoop(
       OMPD_target_teams_distribute_simd, getCollapseNumberExpr(Clauses),
-      nullptr /*ordered not a clause on distribute*/, CS, *this, *DSAStack,
+      nullptr /*ordered not a clause on distribute*/, CS, SemaRef, *DSAStack,
       VarsWithImplicitDSA, B);
   if (NestedLoopCount == 0)
     return StmtError();
 
-  assert((CurContext->isDependentContext() || B.builtAll()) &&
+  assert((SemaRef.CurContext->isDependentContext() || B.builtAll()) &&
          "omp target teams distribute simd loop exprs were not built");
 
-  if (!CurContext->isDependentContext()) {
+  if (!SemaRef.CurContext->isDependentContext()) {
     // Finalize the clauses that need pre-built expressions for CodeGen.
     for (OMPClause *C : Clauses) {
       if (auto *LC = dyn_cast<OMPLinearClause>(C))
         if (FinishOpenMPLinearClause(*LC, cast<DeclRefExpr>(B.IterationVarRef),
-                                     B.NumIterations, *this, CurScope,
-                                     DSAStack))
+                                     B.NumIterations, SemaRef,
+                                     SemaRef.getCurScope(), DSAStack))
           return StmtError();
     }
   }
 
-  if (checkSimdlenSafelenSpecified(*this, Clauses))
+  if (checkSimdlenSafelenSpecified(SemaRef, Clauses))
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
   return OMPTargetTeamsDistributeSimdDirective::Create(
-      Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
+      getASTContext(), StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
-bool Sema::checkTransformableLoopNest(
+bool SemaOpenMP::checkTransformableLoopNest(
     OpenMPDirectiveKind Kind, Stmt *AStmt, int NumLoops,
     SmallVectorImpl<OMPLoopBasedDirective::HelperExprs> &LoopHelpers,
     Stmt *&Body,
@@ -15029,7 +15065,7 @@ bool Sema::checkTransformableLoopNest(
                                                         Stmt *CurStmt) {
         VarsWithInheritedDSAType TmpDSA;
         unsigned SingleNumLoops =
-            checkOpenMPLoop(Kind, nullptr, nullptr, CurStmt, *this, *DSAStack,
+            checkOpenMPLoop(Kind, nullptr, nullptr, CurStmt, SemaRef, *DSAStack,
                             TmpDSA, LoopHelpers[Cnt]);
         if (SingleNumLoops == 0)
           return true;
@@ -15065,9 +15101,11 @@ bool Sema::checkTransformableLoopNest(
   return Result;
 }
 
-StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
-                                          Stmt *AStmt, SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
+                                                Stmt *AStmt,
+                                                SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  ASTContext &Context = getASTContext();
   auto SizesClauses =
       OMPExecutableDirective::getClausesOfKind<OMPSizesClause>(Clauses);
   if (SizesClauses.empty()) {
@@ -15091,7 +15129,7 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
     return StmtError();
 
   // Delay tiling to when template is completely instantiated.
-  if (CurContext->isDependentContext())
+  if (SemaRef.CurContext->isDependentContext())
     return OMPTileDirective::Create(Context, StartLoc, EndLoc, Clauses,
                                     NumLoops, AStmt, nullptr, nullptr);
 
@@ -15117,7 +15155,7 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
       std::string FloorCntName =
           (Twine(".floor_") + llvm::utostr(I) + ".iv." + OrigVarName).str();
       VarDecl *FloorCntDecl =
-          buildVarDecl(*this, {}, CntTy, FloorCntName, nullptr, OrigCntVar);
+          buildVarDecl(SemaRef, {}, CntTy, FloorCntName, nullptr, OrigCntVar);
       FloorIndVars[I] = FloorCntDecl;
     }
 
@@ -15130,7 +15168,8 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
       // used by the expressions to derive the original iteration variable's
       // value from the logical iteration number.
       auto *TileCntDecl = cast<VarDecl>(IterVarRef->getDecl());
-      TileCntDecl->setDeclName(&PP.getIdentifierTable().get(TileCntName));
+      TileCntDecl->setDeclName(
+          &SemaRef.PP.getIdentifierTable().get(TileCntName));
       TileIndVars[I] = TileCntDecl;
     }
     for (auto &P : OriginalInits[I]) {
@@ -15159,17 +15198,18 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
     auto *OrigCntVar = cast<DeclRefExpr>(LoopHelper.Counters[0]);
     QualType CntTy = OrigCntVar->getType();
     Expr *DimTileSize = SizesClause->getSizesRefs()[I];
-    Scope *CurScope = getCurScope();
+    Scope *CurScope = SemaRef.getCurScope();
 
     // Commonly used variables.
-    DeclRefExpr *TileIV = buildDeclRefExpr(*this, TileIndVars[I], CntTy,
+    DeclRefExpr *TileIV = buildDeclRefExpr(SemaRef, TileIndVars[I], CntTy,
                                            OrigCntVar->getExprLoc());
-    DeclRefExpr *FloorIV = buildDeclRefExpr(*this, FloorIndVars[I], CntTy,
+    DeclRefExpr *FloorIV = buildDeclRefExpr(SemaRef, FloorIndVars[I], CntTy,
                                             OrigCntVar->getExprLoc());
 
     // For init-statement: auto .tile.iv = .floor.iv
-    AddInitializerToDecl(TileIndVars[I], DefaultLvalueConversion(FloorIV).get(),
-                         /*DirectInit=*/false);
+    SemaRef.AddInitializerToDecl(TileIndVars[I],
+                                 SemaRef.DefaultLvalueConversion(FloorIV).get(),
+                                 /*DirectInit=*/false);
     Decl *CounterDecl = TileIndVars[I];
     StmtResult InitStmt = new (Context)
         DeclStmt(DeclGroupRef::Create(Context, &CounterDecl, 1),
@@ -15179,28 +15219,29 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
 
     // For cond-expression: .tile.iv < min(.floor.iv + DimTileSize,
     // NumIterations)
-    ExprResult EndOfTile = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
-                                      BO_Add, FloorIV, DimTileSize);
+    ExprResult EndOfTile = SemaRef.BuildBinOp(
+        CurScope, LoopHelper.Cond->getExprLoc(), BO_Add, FloorIV, DimTileSize);
     if (!EndOfTile.isUsable())
       return StmtError();
     ExprResult IsPartialTile =
-        BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
-                   NumIterations, EndOfTile.get());
+        SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
+                           NumIterations, EndOfTile.get());
     if (!IsPartialTile.isUsable())
       return StmtError();
-    ExprResult MinTileAndIterSpace = ActOnConditionalOp(
+    ExprResult MinTileAndIterSpace = SemaRef.ActOnConditionalOp(
         LoopHelper.Cond->getBeginLoc(), LoopHelper.Cond->getEndLoc(),
         IsPartialTile.get(), NumIterations, EndOfTile.get());
     if (!MinTileAndIterSpace.isUsable())
       return StmtError();
-    ExprResult CondExpr = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
-                                     BO_LT, TileIV, MinTileAndIterSpace.get());
+    ExprResult CondExpr =
+        SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
+                           TileIV, MinTileAndIterSpace.get());
     if (!CondExpr.isUsable())
       return StmtError();
 
     // For incr-statement: ++.tile.iv
-    ExprResult IncrStmt =
-        BuildUnaryOp(CurScope, LoopHelper.Inc->getExprLoc(), UO_PreInc, TileIV);
+    ExprResult IncrStmt = SemaRef.BuildUnaryOp(
+        CurScope, LoopHelper.Inc->getExprLoc(), UO_PreInc, TileIV);
     if (!IncrStmt.isUsable())
       return StmtError();
 
@@ -15235,16 +15276,16 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
     DeclRefExpr *OrigCntVar = cast<DeclRefExpr>(LoopHelper.Counters[0]);
     QualType CntTy = OrigCntVar->getType();
     Expr *DimTileSize = SizesClause->getSizesRefs()[I];
-    Scope *CurScope = getCurScope();
+    Scope *CurScope = SemaRef.getCurScope();
 
     // Commonly used variables.
-    DeclRefExpr *FloorIV = buildDeclRefExpr(*this, FloorIndVars[I], CntTy,
+    DeclRefExpr *FloorIV = buildDeclRefExpr(SemaRef, FloorIndVars[I], CntTy,
                                             OrigCntVar->getExprLoc());
 
     // For init-statement: auto .floor.iv = 0
-    AddInitializerToDecl(
+    SemaRef.AddInitializerToDecl(
         FloorIndVars[I],
-        ActOnIntegerConstant(LoopHelper.Init->getExprLoc(), 0).get(),
+        SemaRef.ActOnIntegerConstant(LoopHelper.Init->getExprLoc(), 0).get(),
         /*DirectInit=*/false);
     Decl *CounterDecl = FloorIndVars[I];
     StmtResult InitStmt = new (Context)
@@ -15254,14 +15295,15 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
       return StmtError();
 
     // For cond-expression: .floor.iv < NumIterations
-    ExprResult CondExpr = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
-                                     BO_LT, FloorIV, NumIterations);
+    ExprResult CondExpr = SemaRef.BuildBinOp(
+        CurScope, LoopHelper.Cond->getExprLoc(), BO_LT, FloorIV, NumIterations);
     if (!CondExpr.isUsable())
       return StmtError();
 
     // For incr-statement: .floor.iv += DimTileSize
-    ExprResult IncrStmt = BuildBinOp(CurScope, LoopHelper.Inc->getExprLoc(),
-                                     BO_AddAssign, FloorIV, DimTileSize);
+    ExprResult IncrStmt =
+        SemaRef.BuildBinOp(CurScope, LoopHelper.Inc->getExprLoc(), BO_AddAssign,
+                           FloorIV, DimTileSize);
     if (!IncrStmt.isUsable())
       return StmtError();
 
@@ -15276,15 +15318,18 @@ StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
                                   buildPreInits(Context, PreInits));
 }
 
-StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
-                                            Stmt *AStmt,
-                                            SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
+                                                  Stmt *AStmt,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
+  ASTContext &Context = getASTContext();
+  Scope *CurScope = SemaRef.getCurScope();
   // Empty statement should only be possible if there already was an error.
   if (!AStmt)
     return StmtError();
 
-  if (checkMutuallyExclusiveClauses(*this, Clauses, {OMPC_partial, OMPC_full}))
+  if (checkMutuallyExclusiveClauses(SemaRef, Clauses,
+                                    {OMPC_partial, OMPC_full}))
     return StmtError();
 
   const OMPFullClause *FullClause =
@@ -15307,7 +15352,7 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
   unsigned NumGeneratedLoops = PartialClause ? 1 : 0;
 
   // Delay unrolling to when template is completely instantiated.
-  if (CurContext->isDependentContext())
+  if (SemaRef.CurContext->isDependentContext())
     return OMPUnrollDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
                                       NumGeneratedLoops, nullptr, nullptr);
 
@@ -15412,8 +15457,8 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
   assert(Factor > 0 && "Expected positive unroll factor");
   auto MakeFactorExpr = [this, Factor, IVTy, FactorLoc]() {
     return IntegerLiteral::Create(
-        Context, llvm::APInt(Context.getIntWidth(IVTy), Factor), IVTy,
-        FactorLoc);
+        getASTContext(), llvm::APInt(getASTContext().getIntWidth(IVTy), Factor),
+        IVTy, FactorLoc);
   };
 
   // Iteration variable SourceLocations.
@@ -15430,30 +15475,31 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
 
   // Create the iteration variable for the unrolled loop.
   VarDecl *OuterIVDecl =
-      buildVarDecl(*this, {}, IVTy, OuterIVName, nullptr, OrigVar);
+      buildVarDecl(SemaRef, {}, IVTy, OuterIVName, nullptr, OrigVar);
   auto MakeOuterRef = [this, OuterIVDecl, IVTy, OrigVarLoc]() {
-    return buildDeclRefExpr(*this, OuterIVDecl, IVTy, OrigVarLoc);
+    return buildDeclRefExpr(SemaRef, OuterIVDecl, IVTy, OrigVarLoc);
   };
 
   // Iteration variable for the inner loop: Reuse the iteration variable created
   // by checkOpenMPLoop.
   auto *InnerIVDecl = cast<VarDecl>(IterationVarRef->getDecl());
-  InnerIVDecl->setDeclName(&PP.getIdentifierTable().get(InnerIVName));
+  InnerIVDecl->setDeclName(&SemaRef.PP.getIdentifierTable().get(InnerIVName));
   auto MakeInnerRef = [this, InnerIVDecl, IVTy, OrigVarLoc]() {
-    return buildDeclRefExpr(*this, InnerIVDecl, IVTy, OrigVarLoc);
+    return buildDeclRefExpr(SemaRef, InnerIVDecl, IVTy, OrigVarLoc);
   };
 
   // Make a copy of the NumIterations expression for each use: By the AST
   // constraints, every expression object in a DeclContext must be unique.
-  CaptureVars CopyTransformer(*this);
+  CaptureVars CopyTransformer(SemaRef);
   auto MakeNumIterations = [&CopyTransformer, &LoopHelper]() -> Expr * {
     return AssertSuccess(
         CopyTransformer.TransformExpr(LoopHelper.NumIterations));
   };
 
   // Inner For init-statement: auto .unroll_inner.iv = .unrolled.iv
-  ExprResult LValueConv = DefaultLvalueConversion(MakeOuterRef());
-  AddInitializerToDecl(InnerIVDecl, LValueConv.get(), /*DirectInit=*/false);
+  ExprResult LValueConv = SemaRef.DefaultLvalueConversion(MakeOuterRef());
+  SemaRef.AddInitializerToDecl(InnerIVDecl, LValueConv.get(),
+                               /*DirectInit=*/false);
   StmtResult InnerInit = new (Context)
       DeclStmt(DeclGroupRef(InnerIVDecl), OrigVarLocBegin, OrigVarLocEnd);
   if (!InnerInit.isUsable())
@@ -15466,28 +15512,30 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
   // \endcode
   // This conjunction of two conditions allows ScalarEvolution to derive the
   // maximum trip count of the inner loop.
-  ExprResult EndOfTile = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
-                                    BO_Add, MakeOuterRef(), MakeFactorExpr());
+  ExprResult EndOfTile =
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_Add,
+                         MakeOuterRef(), MakeFactorExpr());
   if (!EndOfTile.isUsable())
     return StmtError();
-  ExprResult InnerCond1 = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
-                                     BO_LT, MakeInnerRef(), EndOfTile.get());
+  ExprResult InnerCond1 =
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
+                         MakeInnerRef(), EndOfTile.get());
   if (!InnerCond1.isUsable())
     return StmtError();
   ExprResult InnerCond2 =
-      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT, MakeInnerRef(),
-                 MakeNumIterations());
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
+                         MakeInnerRef(), MakeNumIterations());
   if (!InnerCond2.isUsable())
     return StmtError();
   ExprResult InnerCond =
-      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LAnd,
-                 InnerCond1.get(), InnerCond2.get());
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LAnd,
+                         InnerCond1.get(), InnerCond2.get());
   if (!InnerCond.isUsable())
     return StmtError();
 
   // Inner For incr-statement: ++.unroll_inner.iv
-  ExprResult InnerIncr = BuildUnaryOp(CurScope, LoopHelper.Inc->getExprLoc(),
-                                      UO_PreInc, MakeInnerRef());
+  ExprResult InnerIncr = SemaRef.BuildUnaryOp(
+      CurScope, LoopHelper.Inc->getExprLoc(), UO_PreInc, MakeInnerRef());
   if (!InnerIncr.isUsable())
     return StmtError();
 
@@ -15496,7 +15544,7 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
   InnerBodyStmts.append(LoopHelper.Updates.begin(), LoopHelper.Updates.end());
   InnerBodyStmts.push_back(Body);
   CompoundStmt *InnerBody =
-      CompoundStmt::Create(Context, InnerBodyStmts, FPOptionsOverride(),
+      CompoundStmt::Create(getASTContext(), InnerBodyStmts, FPOptionsOverride(),
                            Body->getBeginLoc(), Body->getEndLoc());
   ForStmt *InnerFor = new (Context)
       ForStmt(Context, InnerInit.get(), InnerCond.get(), nullptr,
@@ -15518,12 +15566,13 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
   LoopHintAttr *UnrollHintAttr =
       LoopHintAttr::CreateImplicit(Context, LoopHintAttr::UnrollCount,
                                    LoopHintAttr::Numeric, MakeFactorExpr());
-  AttributedStmt *InnerUnrolled =
-      AttributedStmt::Create(Context, StartLoc, {UnrollHintAttr}, InnerFor);
+  AttributedStmt *InnerUnrolled = AttributedStmt::Create(
+      getASTContext(), StartLoc, {UnrollHintAttr}, InnerFor);
 
   // Outer For init-statement: auto .unrolled.iv = 0
-  AddInitializerToDecl(
-      OuterIVDecl, ActOnIntegerConstant(LoopHelper.Init->getExprLoc(), 0).get(),
+  SemaRef.AddInitializerToDecl(
+      OuterIVDecl,
+      SemaRef.ActOnIntegerConstant(LoopHelper.Init->getExprLoc(), 0).get(),
       /*DirectInit=*/false);
   StmtResult OuterInit = new (Context)
       DeclStmt(DeclGroupRef(OuterIVDecl), OrigVarLocBegin, OrigVarLocEnd);
@@ -15532,15 +15581,15 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
 
   // Outer For cond-expression: .unrolled.iv < NumIterations
   ExprResult OuterConde =
-      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT, MakeOuterRef(),
-                 MakeNumIterations());
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT,
+                         MakeOuterRef(), MakeNumIterations());
   if (!OuterConde.isUsable())
     return StmtError();
 
   // Outer For incr-statement: .unrolled.iv += Factor
   ExprResult OuterIncr =
-      BuildBinOp(CurScope, LoopHelper.Inc->getExprLoc(), BO_AddAssign,
-                 MakeOuterRef(), MakeFactorExpr());
+      SemaRef.BuildBinOp(CurScope, LoopHelper.Inc->getExprLoc(), BO_AddAssign,
+                         MakeOuterRef(), MakeFactorExpr());
   if (!OuterIncr.isUsable())
     return StmtError();
 
@@ -15555,10 +15604,11 @@ StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
                                     buildPreInits(Context, PreInits));
 }
 
-OMPClause *Sema::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, Expr *Expr,
-                                             SourceLocation StartLoc,
-                                             SourceLocation LParenLoc,
-                                             SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind,
+                                                   Expr *Expr,
+                                                   SourceLocation StartLoc,
+                                                   SourceLocation LParenLoc,
+                                                   SourceLocation EndLoc) {
   OMPClause *Res = nullptr;
   switch (Kind) {
   case OMPC_final:
@@ -16646,19 +16696,17 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
   return CaptureRegion;
 }
 
-OMPClause *Sema::ActOnOpenMPIfClause(OpenMPDirectiveKind NameModifier,
-                                     Expr *Condition, SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation NameModifierLoc,
-                                     SourceLocation ColonLoc,
-                                     SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPIfClause(
+    OpenMPDirectiveKind NameModifier, Expr *Condition, SourceLocation StartLoc,
+    SourceLocation LParenLoc, SourceLocation NameModifierLoc,
+    SourceLocation ColonLoc, SourceLocation EndLoc) {
   Expr *ValExpr = Condition;
   Stmt *HelperValStmt = nullptr;
   OpenMPDirectiveKind CaptureRegion = OMPD_unknown;
   if (!Condition->isValueDependent() && !Condition->isTypeDependent() &&
       !Condition->isInstantiationDependent() &&
       !Condition->containsUnexpandedParameterPack()) {
-    ExprResult Val = CheckBooleanCondition(StartLoc, Condition);
+    ExprResult Val = SemaRef.CheckBooleanCondition(StartLoc, Condition);
     if (Val.isInvalid())
       return nullptr;
 
@@ -16666,57 +16714,60 @@ OMPClause *Sema::ActOnOpenMPIfClause(OpenMPDirectiveKind NameModifier,
 
     OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
     CaptureRegion = getOpenMPCaptureRegionForClause(
-        DKind, OMPC_if, LangOpts.OpenMP, NameModifier);
-    if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-      ValExpr = MakeFullExpr(ValExpr).get();
+        DKind, OMPC_if, getLangOpts().OpenMP, NameModifier);
+    if (CaptureRegion != OMPD_unknown &&
+        !SemaRef.CurContext->isDependentContext()) {
+      ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
       llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-      ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-      HelperValStmt = buildPreInits(Context, Captures);
+      ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+      HelperValStmt = buildPreInits(getASTContext(), Captures);
     }
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPIfClause(NameModifier, ValExpr, HelperValStmt, CaptureRegion, StartLoc,
                   LParenLoc, NameModifierLoc, ColonLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPFinalClause(Expr *Condition,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPFinalClause(Expr *Condition,
+                                              SourceLocation StartLoc,
+                                              SourceLocation LParenLoc,
+                                              SourceLocation EndLoc) {
   Expr *ValExpr = Condition;
   Stmt *HelperValStmt = nullptr;
   OpenMPDirectiveKind CaptureRegion = OMPD_unknown;
   if (!Condition->isValueDependent() && !Condition->isTypeDependent() &&
       !Condition->isInstantiationDependent() &&
       !Condition->containsUnexpandedParameterPack()) {
-    ExprResult Val = CheckBooleanCondition(StartLoc, Condition);
+    ExprResult Val = SemaRef.CheckBooleanCondition(StartLoc, Condition);
     if (Val.isInvalid())
       return nullptr;
 
-    ValExpr = MakeFullExpr(Val.get()).get();
+    ValExpr = SemaRef.MakeFullExpr(Val.get()).get();
 
     OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
-    CaptureRegion =
-        getOpenMPCaptureRegionForClause(DKind, OMPC_final, LangOpts.OpenMP);
-    if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-      ValExpr = MakeFullExpr(ValExpr).get();
+    CaptureRegion = getOpenMPCaptureRegionForClause(DKind, OMPC_final,
+                                                    getLangOpts().OpenMP);
+    if (CaptureRegion != OMPD_unknown &&
+        !SemaRef.CurContext->isDependentContext()) {
+      ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
       llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-      ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-      HelperValStmt = buildPreInits(Context, Captures);
+      ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+      HelperValStmt = buildPreInits(getASTContext(), Captures);
     }
   }
 
-  return new (Context) OMPFinalClause(ValExpr, HelperValStmt, CaptureRegion,
-                                      StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext()) OMPFinalClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-ExprResult Sema::PerformOpenMPImplicitIntegerConversion(SourceLocation Loc,
-                                                        Expr *Op) {
+ExprResult
+SemaOpenMP::PerformOpenMPImplicitIntegerConversion(SourceLocation Loc,
+                                                   Expr *Op) {
   if (!Op)
     return ExprError();
 
-  class IntConvertDiagnoser : public ICEConvertDiagnoser {
+  class IntConvertDiagnoser : public Sema::ICEConvertDiagnoser {
   public:
     IntConvertDiagnoser()
         : ICEConvertDiagnoser(/*AllowScopedEnumerations*/ false, false, true) {}
@@ -16752,7 +16803,7 @@ ExprResult Sema::PerformOpenMPImplicitIntegerConversion(SourceLocation Loc,
       llvm_unreachable("conversion functions are permitted");
     }
   } ConvertDiagnoser;
-  return PerformContextualImplicitConversion(Loc, Op, ConvertDiagnoser);
+  return SemaRef.PerformContextualImplicitConversion(Loc, Op, ConvertDiagnoser);
 }
 
 static bool
@@ -16765,7 +16816,7 @@ isNonNegativeIntegerValue(Expr *&ValExpr, Sema &SemaRef, OpenMPClauseKind CKind,
       !ValExpr->isInstantiationDependent()) {
     SourceLocation Loc = ValExpr->getExprLoc();
     ExprResult Value =
-        SemaRef.PerformOpenMPImplicitIntegerConversion(Loc, ValExpr);
+        SemaRef.OpenMP().PerformOpenMPImplicitIntegerConversion(Loc, ValExpr);
     if (Value.isInvalid())
       return false;
 
@@ -16797,37 +16848,37 @@ isNonNegativeIntegerValue(Expr *&ValExpr, Sema &SemaRef, OpenMPClauseKind CKind,
   return true;
 }
 
-OMPClause *Sema::ActOnOpenMPNumThreadsClause(Expr *NumThreads,
-                                             SourceLocation StartLoc,
-                                             SourceLocation LParenLoc,
-                                             SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNumThreadsClause(Expr *NumThreads,
+                                                   SourceLocation StartLoc,
+                                                   SourceLocation LParenLoc,
+                                                   SourceLocation EndLoc) {
   Expr *ValExpr = NumThreads;
   Stmt *HelperValStmt = nullptr;
 
   // OpenMP [2.5, Restrictions]
   //  The num_threads expression must evaluate to a positive integer value.
-  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_num_threads,
+  if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_num_threads,
                                  /*StrictlyPositive=*/true))
     return nullptr;
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
-  OpenMPDirectiveKind CaptureRegion =
-      getOpenMPCaptureRegionForClause(DKind, OMPC_num_threads, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+  OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause(
+      DKind, OMPC_num_threads, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context) OMPNumThreadsClause(
+  return new (getASTContext()) OMPNumThreadsClause(
       ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-ExprResult Sema::VerifyPositiveIntegerConstantInClause(Expr *E,
-                                                       OpenMPClauseKind CKind,
-                                                       bool StrictlyPositive,
-                                                       bool SuppressExprDiags) {
+ExprResult SemaOpenMP::VerifyPositiveIntegerConstantInClause(
+    Expr *E, OpenMPClauseKind CKind, bool StrictlyPositive,
+    bool SuppressExprDiags) {
   if (!E)
     return ExprError();
   if (E->isValueDependent() || E->isTypeDependent() ||
@@ -16841,14 +16892,16 @@ ExprResult Sema::VerifyPositiveIntegerConstantInClause(Expr *E,
     // expression.
     struct SuppressedDiagnoser : public Sema::VerifyICEDiagnoser {
       SuppressedDiagnoser() : VerifyICEDiagnoser(/*Suppress=*/true) {}
-      Sema::SemaDiagnosticBuilder diagnoseNotICE(Sema &S,
-                                                 SourceLocation Loc) override {
+      SemaBase::SemaDiagnosticBuilder
+      diagnoseNotICE(Sema &S, SourceLocation Loc) override {
         llvm_unreachable("Diagnostic suppressed");
       }
     } Diagnoser;
-    ICE = VerifyIntegerConstantExpression(E, &Result, Diagnoser, AllowFold);
+    ICE = SemaRef.VerifyIntegerConstantExpression(E, &Result, Diagnoser,
+                                                  Sema::AllowFold);
   } else {
-    ICE = VerifyIntegerConstantExpression(E, &Result, /*FIXME*/ AllowFold);
+    ICE = SemaRef.VerifyIntegerConstantExpression(E, &Result,
+                                                  /*FIXME*/ Sema::AllowFold);
   }
   if (ICE.isInvalid())
     return ExprError();
@@ -16872,29 +16925,31 @@ ExprResult Sema::VerifyPositiveIntegerConstantInClause(Expr *E,
   return ICE;
 }
 
-OMPClause *Sema::ActOnOpenMPSafelenClause(Expr *Len, SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSafelenClause(Expr *Len,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   // OpenMP [2.8.1, simd construct, Description]
   // The parameter of the safelen clause must be a constant
   // positive integer expression.
   ExprResult Safelen = VerifyPositiveIntegerConstantInClause(Len, OMPC_safelen);
   if (Safelen.isInvalid())
     return nullptr;
-  return new (Context)
+  return new (getASTContext())
       OMPSafelenClause(Safelen.get(), StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSimdlenClause(Expr *Len, SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSimdlenClause(Expr *Len,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   // OpenMP [2.8.1, simd construct, Description]
   // The parameter of the simdlen clause must be a constant
   // positive integer expression.
   ExprResult Simdlen = VerifyPositiveIntegerConstantInClause(Len, OMPC_simdlen);
   if (Simdlen.isInvalid())
     return nullptr;
-  return new (Context)
+  return new (getASTContext())
       OMPSimdlenClause(Simdlen.get(), StartLoc, LParenLoc, EndLoc);
 }
 
@@ -16954,31 +17009,32 @@ static bool findOMPAllocatorHandleT(Sema &S, SourceLocation Loc,
   return true;
 }
 
-OMPClause *Sema::ActOnOpenMPAllocatorClause(Expr *A, SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPAllocatorClause(Expr *A,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation LParenLoc,
+                                                  SourceLocation EndLoc) {
   // OpenMP [2.11.3, allocate Directive, Description]
   // allocator is an expression of omp_allocator_handle_t type.
-  if (!findOMPAllocatorHandleT(*this, A->getExprLoc(), DSAStack))
+  if (!findOMPAllocatorHandleT(SemaRef, A->getExprLoc(), DSAStack))
     return nullptr;
 
-  ExprResult Allocator = DefaultLvalueConversion(A);
+  ExprResult Allocator = SemaRef.DefaultLvalueConversion(A);
   if (Allocator.isInvalid())
     return nullptr;
-  Allocator = PerformImplicitConversion(Allocator.get(),
-                                        DSAStack->getOMPAllocatorHandleT(),
-                                        Sema::AA_Initializing,
-                                        /*AllowExplicit=*/true);
+  Allocator = SemaRef.PerformImplicitConversion(
+      Allocator.get(), DSAStack->getOMPAllocatorHandleT(),
+      Sema::AA_Initializing,
+      /*AllowExplicit=*/true);
   if (Allocator.isInvalid())
     return nullptr;
-  return new (Context)
+  return new (getASTContext())
       OMPAllocatorClause(Allocator.get(), StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPCollapseClause(Expr *NumForLoops,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPCollapseClause(Expr *NumForLoops,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation LParenLoc,
+                                                 SourceLocation EndLoc) {
   // OpenMP [2.7.1, loop construct, Description]
   // OpenMP [2.8.1, simd construct, Description]
   // OpenMP [2.9.6, distribute construct, Description]
@@ -16988,14 +17044,14 @@ OMPClause *Sema::ActOnOpenMPCollapseClause(Expr *NumForLoops,
       VerifyPositiveIntegerConstantInClause(NumForLoops, OMPC_collapse);
   if (NumForLoopsResult.isInvalid())
     return nullptr;
-  return new (Context)
+  return new (getASTContext())
       OMPCollapseClause(NumForLoopsResult.get(), StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPOrderedClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc,
-                                          SourceLocation LParenLoc,
-                                          Expr *NumForLoops) {
+OMPClause *SemaOpenMP::ActOnOpenMPOrderedClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc,
+                                                SourceLocation LParenLoc,
+                                                Expr *NumForLoops) {
   // OpenMP [2.7.1, loop construct, Description]
   // OpenMP [2.8.1, simd construct, Description]
   // OpenMP [2.9.6, distribute construct, Description]
@@ -17010,14 +17066,15 @@ OMPClause *Sema::ActOnOpenMPOrderedClause(SourceLocation StartLoc,
   } else {
     NumForLoops = nullptr;
   }
-  auto *Clause = OMPOrderedClause::Create(
-      Context, NumForLoops, NumForLoops ? DSAStack->getAssociatedLoops() : 0,
-      StartLoc, LParenLoc, EndLoc);
+  auto *Clause =
+      OMPOrderedClause::Create(getASTContext(), NumForLoops,
+                               NumForLoops ? DSAStack->getAssociatedLoops() : 0,
+                               StartLoc, LParenLoc, EndLoc);
   DSAStack->setOrderedRegion(/*IsOrdered=*/true, NumForLoops, Clause);
   return Clause;
 }
 
-OMPClause *Sema::ActOnOpenMPSimpleClause(
+OMPClause *SemaOpenMP::ActOnOpenMPSimpleClause(
     OpenMPClauseKind Kind, unsigned Argument, SourceLocation ArgumentLoc,
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc) {
   OMPClause *Res = nullptr;
@@ -17159,11 +17216,11 @@ getListOfPossibleValues(OpenMPClauseKind K, unsigned First, unsigned Last,
   return std::string(Out.str());
 }
 
-OMPClause *Sema::ActOnOpenMPDefaultClause(DefaultKind Kind,
-                                          SourceLocation KindKwLoc,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPDefaultClause(DefaultKind Kind,
+                                                SourceLocation KindKwLoc,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   if (Kind == OMP_DEFAULT_unknown) {
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_default, /*First=*/0,
@@ -17189,39 +17246,39 @@ OMPClause *Sema::ActOnOpenMPDefaultClause(DefaultKind Kind,
     llvm_unreachable("DSA unexpected in OpenMP default clause");
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPDefaultClause(Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPProcBindClause(ProcBindKind Kind,
-                                           SourceLocation KindKwLoc,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPProcBindClause(ProcBindKind Kind,
+                                                 SourceLocation KindKwLoc,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation LParenLoc,
+                                                 SourceLocation EndLoc) {
   if (Kind == OMP_PROC_BIND_unknown) {
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_proc_bind,
                                    /*First=*/unsigned(OMP_PROC_BIND_master),
                                    /*Last=*/
-                                   unsigned(LangOpts.OpenMP > 50
+                                   unsigned(getLangOpts().OpenMP > 50
                                                 ? OMP_PROC_BIND_primary
                                                 : OMP_PROC_BIND_spread) +
                                        1)
         << getOpenMPClauseName(OMPC_proc_bind);
     return nullptr;
   }
-  if (Kind == OMP_PROC_BIND_primary && LangOpts.OpenMP < 51)
+  if (Kind == OMP_PROC_BIND_primary && getLangOpts().OpenMP < 51)
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_proc_bind,
                                    /*First=*/unsigned(OMP_PROC_BIND_master),
                                    /*Last=*/
                                    unsigned(OMP_PROC_BIND_spread) + 1)
         << getOpenMPClauseName(OMPC_proc_bind);
-  return new (Context)
+  return new (getASTContext())
       OMPProcBindClause(Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPAtomicDefaultMemOrderClause(
+OMPClause *SemaOpenMP::ActOnOpenMPAtomicDefaultMemOrderClause(
     OpenMPAtomicDefaultMemOrderClauseKind Kind, SourceLocation KindKwLoc,
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc) {
   if (Kind == OMPC_ATOMIC_DEFAULT_MEM_ORDER_unknown) {
@@ -17232,15 +17289,15 @@ OMPClause *Sema::ActOnOpenMPAtomicDefaultMemOrderClause(
         << getOpenMPClauseName(OMPC_atomic_default_mem_order);
     return nullptr;
   }
-  return new (Context) OMPAtomicDefaultMemOrderClause(Kind, KindKwLoc, StartLoc,
-                                                      LParenLoc, EndLoc);
+  return new (getASTContext()) OMPAtomicDefaultMemOrderClause(
+      Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPAtClause(OpenMPAtClauseKind Kind,
-                                     SourceLocation KindKwLoc,
-                                     SourceLocation StartLoc,
-                                     SourceLocation LParenLoc,
-                                     SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPAtClause(OpenMPAtClauseKind Kind,
+                                           SourceLocation KindKwLoc,
+                                           SourceLocation StartLoc,
+                                           SourceLocation LParenLoc,
+                                           SourceLocation EndLoc) {
   if (Kind == OMPC_AT_unknown) {
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_at, /*First=*/0,
@@ -17248,15 +17305,15 @@ OMPClause *Sema::ActOnOpenMPAtClause(OpenMPAtClauseKind Kind,
         << getOpenMPClauseName(OMPC_at);
     return nullptr;
   }
-  return new (Context)
+  return new (getASTContext())
       OMPAtClause(Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSeverityClause(OpenMPSeverityClauseKind Kind,
-                                           SourceLocation KindKwLoc,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSeverityClause(OpenMPSeverityClauseKind Kind,
+                                                 SourceLocation KindKwLoc,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation LParenLoc,
+                                                 SourceLocation EndLoc) {
   if (Kind == OMPC_SEVERITY_unknown) {
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_severity, /*First=*/0,
@@ -17264,28 +17321,30 @@ OMPClause *Sema::ActOnOpenMPSeverityClause(OpenMPSeverityClauseKind Kind,
         << getOpenMPClauseName(OMPC_severity);
     return nullptr;
   }
-  return new (Context)
+  return new (getASTContext())
       OMPSeverityClause(Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPMessageClause(Expr *ME, SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPMessageClause(Expr *ME,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   assert(ME && "NULL expr in Message clause");
   if (!isa<StringLiteral>(ME)) {
     Diag(ME->getBeginLoc(), diag::warn_clause_expected_string)
         << getOpenMPClauseName(OMPC_message);
     return nullptr;
   }
-  return new (Context) OMPMessageClause(ME, StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext())
+      OMPMessageClause(ME, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPOrderClause(
+OMPClause *SemaOpenMP::ActOnOpenMPOrderClause(
     OpenMPOrderClauseModifier Modifier, OpenMPOrderClauseKind Kind,
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation MLoc,
     SourceLocation KindLoc, SourceLocation EndLoc) {
   if (Kind != OMPC_ORDER_concurrent ||
-      (LangOpts.OpenMP < 51 && MLoc.isValid())) {
+      (getLangOpts().OpenMP < 51 && MLoc.isValid())) {
     // Kind should be concurrent,
     // Modifiers introduced in OpenMP 5.1
     static_assert(OMPC_ORDER_unknown > 0,
@@ -17298,7 +17357,7 @@ OMPClause *Sema::ActOnOpenMPOrderClause(
         << getOpenMPClauseName(OMPC_order);
     return nullptr;
   }
-  if (LangOpts.OpenMP >= 51) {
+  if (getLangOpts().OpenMP >= 51) {
     if (Modifier == OMPC_ORDER_MODIFIER_unknown && MLoc.isValid()) {
       Diag(MLoc, diag::err_omp_unexpected_clause_value)
           << getListOfPossibleValues(OMPC_order,
@@ -17315,21 +17374,21 @@ OMPClause *Sema::ActOnOpenMPOrderClause(
       }
     }
   }
-  return new (Context) OMPOrderClause(Kind, KindLoc, StartLoc, LParenLoc,
-                                      EndLoc, Modifier, MLoc);
+  return new (getASTContext()) OMPOrderClause(
+      Kind, KindLoc, StartLoc, LParenLoc, EndLoc, Modifier, MLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUpdateClause(OpenMPDependClauseKind Kind,
-                                         SourceLocation KindKwLoc,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPUpdateClause(OpenMPDependClauseKind Kind,
+                                               SourceLocation KindKwLoc,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   if (Kind == OMPC_DEPEND_unknown || Kind == OMPC_DEPEND_source ||
       Kind == OMPC_DEPEND_sink || Kind == OMPC_DEPEND_depobj) {
     SmallVector<unsigned> Except = {
         OMPC_DEPEND_source, OMPC_DEPEND_sink, OMPC_DEPEND_depobj,
         OMPC_DEPEND_outallmemory, OMPC_DEPEND_inoutallmemory};
-    if (LangOpts.OpenMP < 51)
+    if (getLangOpts().OpenMP < 51)
       Except.push_back(OMPC_DEPEND_inoutset);
     Diag(KindKwLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_depend, /*First=*/0,
@@ -17337,14 +17396,14 @@ OMPClause *Sema::ActOnOpenMPUpdateClause(OpenMPDependClauseKind Kind,
         << getOpenMPClauseName(OMPC_update);
     return nullptr;
   }
-  return OMPUpdateClause::Create(Context, StartLoc, LParenLoc, KindKwLoc, Kind,
-                                 EndLoc);
+  return OMPUpdateClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                 KindKwLoc, Kind, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
+                                              SourceLocation StartLoc,
+                                              SourceLocation LParenLoc,
+                                              SourceLocation EndLoc) {
   for (Expr *SizeExpr : SizeExprs) {
     ExprResult NumForLoopsResult = VerifyPositiveIntegerConstantInClause(
         SizeExpr, OMPC_sizes, /*StrictlyPositive=*/true);
@@ -17353,19 +17412,19 @@ OMPClause *Sema::ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
   }
 
   DSAStack->setAssociatedLoops(SizeExprs.size());
-  return OMPSizesClause::Create(Context, StartLoc, LParenLoc, EndLoc,
+  return OMPSizesClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
                                 SizeExprs);
 }
 
-OMPClause *Sema::ActOnOpenMPFullClause(SourceLocation StartLoc,
-                                       SourceLocation EndLoc) {
-  return OMPFullClause::Create(Context, StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPFullClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc) {
+  return OMPFullClause::Create(getASTContext(), StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPPartialClause(Expr *FactorExpr,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPPartialClause(Expr *FactorExpr,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   if (FactorExpr) {
     // If an argument is specified, it must be a constant (or an unevaluated
     // template expression).
@@ -17376,22 +17435,22 @@ OMPClause *Sema::ActOnOpenMPPartialClause(Expr *FactorExpr,
     FactorExpr = FactorResult.get();
   }
 
-  return OMPPartialClause::Create(Context, StartLoc, LParenLoc, EndLoc,
+  return OMPPartialClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
                                   FactorExpr);
 }
 
-OMPClause *Sema::ActOnOpenMPAlignClause(Expr *A, SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPAlignClause(Expr *A, SourceLocation StartLoc,
+                                              SourceLocation LParenLoc,
+                                              SourceLocation EndLoc) {
   ExprResult AlignVal;
   AlignVal = VerifyPositiveIntegerConstantInClause(A, OMPC_align);
   if (AlignVal.isInvalid())
     return nullptr;
-  return OMPAlignClause::Create(Context, AlignVal.get(), StartLoc, LParenLoc,
-                                EndLoc);
+  return OMPAlignClause::Create(getASTContext(), AlignVal.get(), StartLoc,
+                                LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSingleExprWithArgClause(
+OMPClause *SemaOpenMP::ActOnOpenMPSingleExprWithArgClause(
     OpenMPClauseKind Kind, ArrayRef<unsigned> Argument, Expr *Expr,
     SourceLocation StartLoc, SourceLocation LParenLoc,
     ArrayRef<SourceLocation> ArgumentLoc, SourceLocation DelimLoc,
@@ -17559,13 +17618,13 @@ static bool checkScheduleModifiers(Sema &S, OpenMPScheduleClauseModifier M1,
   return false;
 }
 
-OMPClause *Sema::ActOnOpenMPScheduleClause(
+OMPClause *SemaOpenMP::ActOnOpenMPScheduleClause(
     OpenMPScheduleClauseModifier M1, OpenMPScheduleClauseModifier M2,
     OpenMPScheduleClauseKind Kind, Expr *ChunkSize, SourceLocation StartLoc,
     SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc,
     SourceLocation KindLoc, SourceLocation CommaLoc, SourceLocation EndLoc) {
-  if (checkScheduleModifiers(*this, M1, M2, M1Loc, M2Loc) ||
-      checkScheduleModifiers(*this, M2, M1, M2Loc, M1Loc))
+  if (checkScheduleModifiers(SemaRef, M1, M2, M1Loc, M2Loc) ||
+      checkScheduleModifiers(SemaRef, M2, M1, M2Loc, M1Loc))
     return nullptr;
   // OpenMP, 2.7.1, Loop Construct, Restrictions
   // Either the monotonic modifier or the nonmonotonic modifier can be specified
@@ -17599,7 +17658,7 @@ OMPClause *Sema::ActOnOpenMPScheduleClause(
   // The nonmonotonic modifier can only be specified with schedule(dynamic) or
   // schedule(guided).
   // OpenMP 5.0 does not have this restriction.
-  if (LangOpts.OpenMP < 50 &&
+  if (getLangOpts().OpenMP < 50 &&
       (M1 == OMPC_SCHEDULE_MODIFIER_nonmonotonic ||
        M2 == OMPC_SCHEDULE_MODIFIER_nonmonotonic) &&
       Kind != OMPC_SCHEDULE_dynamic && Kind != OMPC_SCHEDULE_guided) {
@@ -17625,7 +17684,7 @@ OMPClause *Sema::ActOnOpenMPScheduleClause(
       //  chunk_size must be a loop invariant integer expression with a positive
       //  value.
       if (std::optional<llvm::APSInt> Result =
-              ValExpr->getIntegerConstantExpr(Context)) {
+              ValExpr->getIntegerConstantExpr(getASTContext())) {
         if (Result->isSigned() && !Result->isStrictlyPositive()) {
           Diag(ChunkSizeLoc, diag::err_omp_negative_expression_in_clause)
               << "schedule" << 1 << ChunkSize->getSourceRange();
@@ -17633,24 +17692,24 @@ OMPClause *Sema::ActOnOpenMPScheduleClause(
         }
       } else if (getOpenMPCaptureRegionForClause(
                      DSAStack->getCurrentDirective(), OMPC_schedule,
-                     LangOpts.OpenMP) != OMPD_unknown &&
-                 !CurContext->isDependentContext()) {
-        ValExpr = MakeFullExpr(ValExpr).get();
+                     getLangOpts().OpenMP) != OMPD_unknown &&
+                 !SemaRef.CurContext->isDependentContext()) {
+        ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
         llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-        ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-        HelperValStmt = buildPreInits(Context, Captures);
+        ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+        HelperValStmt = buildPreInits(getASTContext(), Captures);
       }
     }
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPScheduleClause(StartLoc, LParenLoc, KindLoc, CommaLoc, EndLoc, Kind,
                         ValExpr, HelperValStmt, M1, M1Loc, M2, M2Loc);
 }
 
-OMPClause *Sema::ActOnOpenMPClause(OpenMPClauseKind Kind,
-                                   SourceLocation StartLoc,
-                                   SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPClause(OpenMPClauseKind Kind,
+                                         SourceLocation StartLoc,
+                                         SourceLocation EndLoc) {
   OMPClause *Res = nullptr;
   switch (Kind) {
   case OMPC_ordered:
@@ -17804,134 +17863,138 @@ OMPClause *Sema::ActOnOpenMPClause(OpenMPClauseKind Kind,
   return Res;
 }
 
-OMPClause *Sema::ActOnOpenMPNowaitClause(SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNowaitClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
   DSAStack->setNowaitRegion();
-  return new (Context) OMPNowaitClause(StartLoc, EndLoc);
+  return new (getASTContext()) OMPNowaitClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUntiedClause(SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPUntiedClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
   DSAStack->setUntiedRegion();
-  return new (Context) OMPUntiedClause(StartLoc, EndLoc);
+  return new (getASTContext()) OMPUntiedClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPMergeableClause(SourceLocation StartLoc,
-                                            SourceLocation EndLoc) {
-  return new (Context) OMPMergeableClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPMergeableClause(SourceLocation StartLoc,
+                                                  SourceLocation EndLoc) {
+  return new (getASTContext()) OMPMergeableClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPReadClause(SourceLocation StartLoc,
-                                       SourceLocation EndLoc) {
-  return new (Context) OMPReadClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPReadClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc) {
+  return new (getASTContext()) OMPReadClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPWriteClause(SourceLocation StartLoc,
-                                        SourceLocation EndLoc) {
-  return new (Context) OMPWriteClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPWriteClause(SourceLocation StartLoc,
+                                              SourceLocation EndLoc) {
+  return new (getASTContext()) OMPWriteClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUpdateClause(SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
-  return OMPUpdateClause::Create(Context, StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPUpdateClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
+  return OMPUpdateClause::Create(getASTContext(), StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPCaptureClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPCaptureClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPCaptureClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPCaptureClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPCompareClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPCompareClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPCompareClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPCompareClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPFailClause(SourceLocation StartLoc,
-                                       SourceLocation EndLoc) {
-  return new (Context) OMPFailClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPFailClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc) {
+  return new (getASTContext()) OMPFailClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPFailClause(
-      OpenMPClauseKind Parameter, SourceLocation KindLoc,
-      SourceLocation StartLoc, SourceLocation LParenLoc,
-      SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPFailClause(OpenMPClauseKind Parameter,
+                                             SourceLocation KindLoc,
+                                             SourceLocation StartLoc,
+                                             SourceLocation LParenLoc,
+                                             SourceLocation EndLoc) {
 
   if (!checkFailClauseParameter(Parameter)) {
     Diag(KindLoc, diag::err_omp_atomic_fail_wrong_or_no_clauses);
     return nullptr;
   }
-  return new (Context)
+  return new (getASTContext())
       OMPFailClause(Parameter, KindLoc, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSeqCstClause(SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
-  return new (Context) OMPSeqCstClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPSeqCstClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
+  return new (getASTContext()) OMPSeqCstClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPAcqRelClause(SourceLocation StartLoc,
-                                         SourceLocation EndLoc) {
-  return new (Context) OMPAcqRelClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPAcqRelClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
+  return new (getASTContext()) OMPAcqRelClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPAcquireClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPAcquireClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPAcquireClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPAcquireClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPReleaseClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPReleaseClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPReleaseClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPReleaseClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPRelaxedClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPRelaxedClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPRelaxedClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPRelaxedClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPWeakClause(SourceLocation StartLoc,
-                                       SourceLocation EndLoc) {
-  return new (Context) OMPWeakClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPWeakClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc) {
+  return new (getASTContext()) OMPWeakClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPThreadsClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPThreadsClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPThreadsClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPThreadsClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPSIMDClause(SourceLocation StartLoc,
-                                       SourceLocation EndLoc) {
-  return new (Context) OMPSIMDClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPSIMDClause(SourceLocation StartLoc,
+                                             SourceLocation EndLoc) {
+  return new (getASTContext()) OMPSIMDClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPNogroupClause(SourceLocation StartLoc,
-                                          SourceLocation EndLoc) {
-  return new (Context) OMPNogroupClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPNogroupClause(SourceLocation StartLoc,
+                                                SourceLocation EndLoc) {
+  return new (getASTContext()) OMPNogroupClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUnifiedAddressClause(SourceLocation StartLoc,
-                                                 SourceLocation EndLoc) {
-  return new (Context) OMPUnifiedAddressClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPUnifiedAddressClause(SourceLocation StartLoc,
+                                                       SourceLocation EndLoc) {
+  return new (getASTContext()) OMPUnifiedAddressClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUnifiedSharedMemoryClause(SourceLocation StartLoc,
-                                                      SourceLocation EndLoc) {
-  return new (Context) OMPUnifiedSharedMemoryClause(StartLoc, EndLoc);
+OMPClause *
+SemaOpenMP::ActOnOpenMPUnifiedSharedMemoryClause(SourceLocation StartLoc,
+                                                 SourceLocation EndLoc) {
+  return new (getASTContext()) OMPUnifiedSharedMemoryClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPReverseOffloadClause(SourceLocation StartLoc,
-                                                 SourceLocation EndLoc) {
-  return new (Context) OMPReverseOffloadClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPReverseOffloadClause(SourceLocation StartLoc,
+                                                       SourceLocation EndLoc) {
+  return new (getASTContext()) OMPReverseOffloadClause(StartLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPDynamicAllocatorsClause(SourceLocation StartLoc,
-                                                    SourceLocation EndLoc) {
-  return new (Context) OMPDynamicAllocatorsClause(StartLoc, EndLoc);
+OMPClause *
+SemaOpenMP::ActOnOpenMPDynamicAllocatorsClause(SourceLocation StartLoc,
+                                               SourceLocation EndLoc) {
+  return new (getASTContext()) OMPDynamicAllocatorsClause(StartLoc, EndLoc);
 }
 
-StmtResult Sema::ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
-                                             SourceLocation StartLoc,
-                                             SourceLocation EndLoc) {
+StmtResult
+SemaOpenMP::ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
+                                        SourceLocation StartLoc,
+                                        SourceLocation EndLoc) {
 
   // OpenMP 5.1 [2.15.1, interop Construct, Restrictions]
   // At least one action-clause must appear on a directive.
@@ -17981,13 +18044,13 @@ StmtResult Sema::ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
 
     if (ClauseKind == OMPC_init) {
       auto *E = cast<OMPInitClause>(C)->getInteropVar();
-      DeclResult = getPrivateItem(*this, E, ELoc, ERange);
+      DeclResult = getPrivateItem(SemaRef, E, ELoc, ERange);
     } else if (ClauseKind == OMPC_use) {
       auto *E = cast<OMPUseClause>(C)->getInteropVar();
-      DeclResult = getPrivateItem(*this, E, ELoc, ERange);
+      DeclResult = getPrivateItem(SemaRef, E, ELoc, ERange);
     } else if (ClauseKind == OMPC_destroy) {
       auto *E = cast<OMPDestroyClause>(C)->getInteropVar();
-      DeclResult = getPrivateItem(*this, E, ELoc, ERange);
+      DeclResult = getPrivateItem(SemaRef, E, ELoc, ERange);
     }
 
     if (DeclResult.first) {
@@ -17999,7 +18062,8 @@ StmtResult Sema::ActOnOpenMPInteropDirective(ArrayRef<OMPClause *> Clauses,
     }
   }
 
-  return OMPInteropDirective::Create(Context, StartLoc, EndLoc, Clauses);
+  return OMPInteropDirective::Create(getASTContext(), StartLoc, EndLoc,
+                                     Clauses);
 }
 
 static bool isValidInteropVariable(Sema &SemaRef, Expr *InteropVarExpr,
@@ -18059,12 +18123,11 @@ static bool isValidInteropVariable(Sema &SemaRef, Expr *InteropVarExpr,
   return true;
 }
 
-OMPClause *
-Sema::ActOnOpenMPInitClause(Expr *InteropVar, OMPInteropInfo &InteropInfo,
-                            SourceLocation StartLoc, SourceLocation LParenLoc,
-                            SourceLocation VarLoc, SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPInitClause(
+    Expr *InteropVar, OMPInteropInfo &InteropInfo, SourceLocation StartLoc,
+    SourceLocation LParenLoc, SourceLocation VarLoc, SourceLocation EndLoc) {
 
-  if (!isValidInteropVariable(*this, InteropVar, VarLoc, OMPC_init))
+  if (!isValidInteropVariable(SemaRef, InteropVar, VarLoc, OMPC_init))
     return nullptr;
 
   // Check prefer_type values.  These foreign-runtime-id values are either
@@ -18073,7 +18136,7 @@ Sema::ActOnOpenMPInitClause(Expr *InteropVar, OMPInteropInfo &InteropInfo,
     if (E->isValueDependent() || E->isTypeDependent() ||
         E->isInstantiationDependent() || E->containsUnexpandedParameterPack())
       continue;
-    if (E->isIntegerConstantExpr(Context))
+    if (E->isIntegerConstantExpr(getASTContext()))
       continue;
     if (isa<StringLiteral>(E))
       continue;
@@ -18081,28 +18144,29 @@ Sema::ActOnOpenMPInitClause(Expr *InteropVar, OMPInteropInfo &InteropInfo,
     return nullptr;
   }
 
-  return OMPInitClause::Create(Context, InteropVar, InteropInfo, StartLoc,
-                               LParenLoc, VarLoc, EndLoc);
+  return OMPInitClause::Create(getASTContext(), InteropVar, InteropInfo,
+                               StartLoc, LParenLoc, VarLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPUseClause(Expr *InteropVar, SourceLocation StartLoc,
-                                      SourceLocation LParenLoc,
-                                      SourceLocation VarLoc,
-                                      SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPUseClause(Expr *InteropVar,
+                                            SourceLocation StartLoc,
+                                            SourceLocation LParenLoc,
+                                            SourceLocation VarLoc,
+                                            SourceLocation EndLoc) {
 
-  if (!isValidInteropVariable(*this, InteropVar, VarLoc, OMPC_use))
+  if (!isValidInteropVariable(SemaRef, InteropVar, VarLoc, OMPC_use))
     return nullptr;
 
-  return new (Context)
+  return new (getASTContext())
       OMPUseClause(InteropVar, StartLoc, LParenLoc, VarLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPDestroyClause(Expr *InteropVar,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation VarLoc,
-                                          SourceLocation EndLoc) {
-  if (!InteropVar && LangOpts.OpenMP >= 52 &&
+OMPClause *SemaOpenMP::ActOnOpenMPDestroyClause(Expr *InteropVar,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation VarLoc,
+                                                SourceLocation EndLoc) {
+  if (!InteropVar && getLangOpts().OpenMP >= 52 &&
       DSAStack->getCurrentDirective() == OMPD_depobj) {
     Diag(StartLoc, diag::err_omp_expected_clause_argument)
         << getOpenMPClauseName(OMPC_destroy)
@@ -18110,100 +18174,103 @@ OMPClause *Sema::ActOnOpenMPDestroyClause(Expr *InteropVar,
     return nullptr;
   }
   if (InteropVar &&
-      !isValidInteropVariable(*this, InteropVar, VarLoc, OMPC_destroy))
+      !isValidInteropVariable(SemaRef, InteropVar, VarLoc, OMPC_destroy))
     return nullptr;
 
-  return new (Context)
+  return new (getASTContext())
       OMPDestroyClause(InteropVar, StartLoc, LParenLoc, VarLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPNovariantsClause(Expr *Condition,
-                                             SourceLocation StartLoc,
-                                             SourceLocation LParenLoc,
-                                             SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNovariantsClause(Expr *Condition,
+                                                   SourceLocation StartLoc,
+                                                   SourceLocation LParenLoc,
+                                                   SourceLocation EndLoc) {
   Expr *ValExpr = Condition;
   Stmt *HelperValStmt = nullptr;
   OpenMPDirectiveKind CaptureRegion = OMPD_unknown;
   if (!Condition->isValueDependent() && !Condition->isTypeDependent() &&
       !Condition->isInstantiationDependent() &&
       !Condition->containsUnexpandedParameterPack()) {
-    ExprResult Val = CheckBooleanCondition(StartLoc, Condition);
+    ExprResult Val = SemaRef.CheckBooleanCondition(StartLoc, Condition);
     if (Val.isInvalid())
       return nullptr;
 
-    ValExpr = MakeFullExpr(Val.get()).get();
+    ValExpr = SemaRef.MakeFullExpr(Val.get()).get();
 
     OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
     CaptureRegion = getOpenMPCaptureRegionForClause(DKind, OMPC_novariants,
-                                                    LangOpts.OpenMP);
-    if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-      ValExpr = MakeFullExpr(ValExpr).get();
+                                                    getLangOpts().OpenMP);
+    if (CaptureRegion != OMPD_unknown &&
+        !SemaRef.CurContext->isDependentContext()) {
+      ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
       llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-      ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-      HelperValStmt = buildPreInits(Context, Captures);
+      ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+      HelperValStmt = buildPreInits(getASTContext(), Captures);
     }
   }
 
-  return new (Context) OMPNovariantsClause(
+  return new (getASTContext()) OMPNovariantsClause(
       ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPNocontextClause(Expr *Condition,
-                                            SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNocontextClause(Expr *Condition,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation LParenLoc,
+                                                  SourceLocation EndLoc) {
   Expr *ValExpr = Condition;
   Stmt *HelperValStmt = nullptr;
   OpenMPDirectiveKind CaptureRegion = OMPD_unknown;
   if (!Condition->isValueDependent() && !Condition->isTypeDependent() &&
       !Condition->isInstantiationDependent() &&
       !Condition->containsUnexpandedParameterPack()) {
-    ExprResult Val = CheckBooleanCondition(StartLoc, Condition);
+    ExprResult Val = SemaRef.CheckBooleanCondition(StartLoc, Condition);
     if (Val.isInvalid())
       return nullptr;
 
-    ValExpr = MakeFullExpr(Val.get()).get();
+    ValExpr = SemaRef.MakeFullExpr(Val.get()).get();
 
     OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
-    CaptureRegion =
-        getOpenMPCaptureRegionForClause(DKind, OMPC_nocontext, LangOpts.OpenMP);
-    if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-      ValExpr = MakeFullExpr(ValExpr).get();
+    CaptureRegion = getOpenMPCaptureRegionForClause(DKind, OMPC_nocontext,
+                                                    getLangOpts().OpenMP);
+    if (CaptureRegion != OMPD_unknown &&
+        !SemaRef.CurContext->isDependentContext()) {
+      ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
       llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-      ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-      HelperValStmt = buildPreInits(Context, Captures);
+      ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+      HelperValStmt = buildPreInits(getASTContext(), Captures);
     }
   }
 
-  return new (Context) OMPNocontextClause(ValExpr, HelperValStmt, CaptureRegion,
-                                          StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext()) OMPNocontextClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPFilterClause(Expr *ThreadID,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPFilterClause(Expr *ThreadID,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   Expr *ValExpr = ThreadID;
   Stmt *HelperValStmt = nullptr;
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
   OpenMPDirectiveKind CaptureRegion =
-      getOpenMPCaptureRegionForClause(DKind, OMPC_filter, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+      getOpenMPCaptureRegionForClause(DKind, OMPC_filter, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context) OMPFilterClause(ValExpr, HelperValStmt, CaptureRegion,
-                                       StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext()) OMPFilterClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPVarListClause(OpenMPClauseKind Kind,
-                                          ArrayRef<Expr *> VarList,
-                                          const OMPVarListLocTy &Locs,
-                                          OpenMPVarListDataTy &Data) {
+OMPClause *SemaOpenMP::ActOnOpenMPVarListClause(OpenMPClauseKind Kind,
+                                                ArrayRef<Expr *> VarList,
+                                                const OMPVarListLocTy &Locs,
+                                                OpenMPVarListDataTy &Data) {
   SourceLocation StartLoc = Locs.StartLoc;
   SourceLocation LParenLoc = Locs.LParenLoc;
   SourceLocation EndLoc = Locs.EndLoc;
@@ -18395,29 +18462,30 @@ OMPClause *Sema::ActOnOpenMPVarListClause(OpenMPClauseKind Kind,
   return Res;
 }
 
-ExprResult Sema::getOpenMPCapturedExpr(VarDecl *Capture, ExprValueKind VK,
-                                       ExprObjectKind OK, SourceLocation Loc) {
-  ExprResult Res = BuildDeclRefExpr(
+ExprResult SemaOpenMP::getOpenMPCapturedExpr(VarDecl *Capture, ExprValueKind VK,
+                                             ExprObjectKind OK,
+                                             SourceLocation Loc) {
+  ExprResult Res = SemaRef.BuildDeclRefExpr(
       Capture, Capture->getType().getNonReferenceType(), VK_LValue, Loc);
   if (!Res.isUsable())
     return ExprError();
   if (OK == OK_Ordinary && !getLangOpts().CPlusPlus) {
-    Res = CreateBuiltinUnaryOp(Loc, UO_Deref, Res.get());
+    Res = SemaRef.CreateBuiltinUnaryOp(Loc, UO_Deref, Res.get());
     if (!Res.isUsable())
       return ExprError();
   }
   if (VK != VK_LValue && Res.get()->isGLValue()) {
-    Res = DefaultLvalueConversion(Res.get());
+    Res = SemaRef.DefaultLvalueConversion(Res.get());
     if (!Res.isUsable())
       return ExprError();
   }
   return Res;
 }
 
-OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
-                                          SourceLocation StartLoc,
-                                          SourceLocation LParenLoc,
-                                          SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   SmallVector<Expr *, 8> PrivateCopies;
   bool IsImplicitClause =
@@ -18427,7 +18495,7 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -18443,7 +18511,8 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     // OpenMP [2.9.3.3, Restrictions, C/C++, p.3]
     //  A variable that appears in a private clause must not have an incomplete
     //  type or a reference type.
-    if (RequireCompleteType(ELoc, Type, diag::err_omp_private_incomplete_type))
+    if (SemaRef.RequireCompleteType(ELoc, Type,
+                                    diag::err_omp_private_incomplete_type))
       continue;
     Type = Type.getNonReferenceType();
 
@@ -18455,7 +18524,7 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     // OpenMP 3.1 [2.9.3.3, private clause, Restrictions]
     // A variable that appears in a private clause must not have a
     // const-qualified type unless it is of class type with a mutable member.
-    if (rejectConstNotMutableType(*this, D, Type, OMPC_private, ELoc))
+    if (rejectConstNotMutableType(SemaRef, D, Type, OMPC_private, ELoc))
       continue;
 
     // OpenMP [2.9.1.1, Data-sharing Attribute Rules for Variables Referenced
@@ -18469,7 +18538,7 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     if (DVar.CKind != OMPC_unknown && DVar.CKind != OMPC_private) {
       Diag(ELoc, diag::err_omp_wrong_dsa) << getOpenMPClauseName(DVar.CKind)
                                           << getOpenMPClauseName(OMPC_private);
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
@@ -18480,7 +18549,7 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
       Diag(ELoc, diag::err_omp_variably_modified_type_not_supported)
           << getOpenMPClauseName(OMPC_private) << Type
           << getOpenMPDirectiveName(CurrDir);
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -18496,7 +18565,8 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     // A list item cannot appear in both a map clause and a data-sharing
     // attribute clause on the same construct unless the construct is a
     // combined construct.
-    if ((LangOpts.OpenMP <= 45 && isOpenMPTargetExecutionDirective(CurrDir)) ||
+    if ((getLangOpts().OpenMP <= 45 &&
+         isOpenMPTargetExecutionDirective(CurrDir)) ||
         CurrDir == OMPD_target) {
       OpenMPClauseKind ConflictKind;
       if (DSAStack->checkMappableExprComponentListsForDecl(
@@ -18510,7 +18580,7 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
             << getOpenMPClauseName(OMPC_private)
             << getOpenMPClauseName(ConflictKind)
             << getOpenMPDirectiveName(CurrDir);
-        reportOriginalDsa(*this, DSAStack, D, DVar);
+        reportOriginalDsa(SemaRef, DSAStack, D, DVar);
         continue;
       }
     }
@@ -18526,28 +18596,28 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
     // proper diagnostics.
     Type = Type.getUnqualifiedType();
     VarDecl *VDPrivate =
-        buildVarDecl(*this, ELoc, Type, D->getName(),
+        buildVarDecl(SemaRef, ELoc, Type, D->getName(),
                      D->hasAttrs() ? &D->getAttrs() : nullptr,
                      VD ? cast<DeclRefExpr>(SimpleRefExpr) : nullptr);
-    ActOnUninitializedDecl(VDPrivate);
+    SemaRef.ActOnUninitializedDecl(VDPrivate);
     if (VDPrivate->isInvalidDecl())
       continue;
     DeclRefExpr *VDPrivateRefExpr = buildDeclRefExpr(
-        *this, VDPrivate, RefExpr->getType().getUnqualifiedType(), ELoc);
+        SemaRef, VDPrivate, RefExpr->getType().getUnqualifiedType(), ELoc);
 
     DeclRefExpr *Ref = nullptr;
-    if (!VD && !CurContext->isDependentContext()) {
+    if (!VD && !SemaRef.CurContext->isDependentContext()) {
       auto *FD = dyn_cast<FieldDecl>(D);
       VarDecl *VD = FD ? DSAStack->getImplicitFDCapExprDecl(FD) : nullptr;
       if (VD)
-        Ref = buildDeclRefExpr(*this, VD, VD->getType().getNonReferenceType(),
+        Ref = buildDeclRefExpr(SemaRef, VD, VD->getType().getNonReferenceType(),
                                RefExpr->getExprLoc());
       else
-        Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/false);
+        Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/false);
     }
     if (!IsImplicitClause)
       DSAStack->addDSA(D, RefExpr->IgnoreParens(), OMPC_private, Ref);
-    Vars.push_back((VD || CurContext->isDependentContext())
+    Vars.push_back((VD || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
     PrivateCopies.push_back(VDPrivateRefExpr);
@@ -18556,14 +18626,14 @@ OMPClause *Sema::ActOnOpenMPPrivateClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPPrivateClause::Create(Context, StartLoc, LParenLoc, EndLoc, Vars,
-                                  PrivateCopies);
+  return OMPPrivateClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
+                                  Vars, PrivateCopies);
 }
 
-OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
-                                               SourceLocation StartLoc,
-                                               SourceLocation LParenLoc,
-                                               SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
+                                                     SourceLocation StartLoc,
+                                                     SourceLocation LParenLoc,
+                                                     SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   SmallVector<Expr *, 8> PrivateCopies;
   SmallVector<Expr *, 8> Inits;
@@ -18577,7 +18647,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -18595,8 +18665,8 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
     // OpenMP [2.9.3.3, Restrictions, C/C++, p.3]
     //  A variable that appears in a private clause must not have an incomplete
     //  type or a reference type.
-    if (RequireCompleteType(ELoc, Type,
-                            diag::err_omp_firstprivate_incomplete_type))
+    if (SemaRef.RequireCompleteType(ELoc, Type,
+                                    diag::err_omp_firstprivate_incomplete_type))
       continue;
     Type = Type.getNonReferenceType();
 
@@ -18604,7 +18674,8 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
     //  A variable of class type (or array thereof) that appears in a private
     //  clause requires an accessible, unambiguous copy constructor for the
     //  class type.
-    QualType ElemType = Context.getBaseElementType(Type).getNonReferenceType();
+    QualType ElemType =
+        getASTContext().getBaseElementType(Type).getNonReferenceType();
 
     // If an implicit firstprivate variable found it was checked already.
     DSAStackTy::DSAVarData TopDVar;
@@ -18613,7 +18684,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
           DSAStack->getTopDSA(D, /*FromParent=*/false);
       TopDVar = DVar;
       OpenMPDirectiveKind CurrDir = DSAStack->getCurrentDirective();
-      bool IsConstant = ElemType.isConstant(Context);
+      bool IsConstant = ElemType.isConstant(getASTContext());
       // OpenMP [2.4.13, Data-sharing Attribute Clauses]
       //  A list item that specifies a given variable may not appear in more
       // than one clause on the same directive, except that a variable may be
@@ -18628,7 +18699,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
         Diag(ELoc, diag::err_omp_wrong_dsa)
             << getOpenMPClauseName(DVar.CKind)
             << getOpenMPClauseName(OMPC_firstprivate);
-        reportOriginalDsa(*this, DSAStack, D, DVar);
+        reportOriginalDsa(SemaRef, DSAStack, D, DVar);
         continue;
       }
 
@@ -18648,7 +18719,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
         Diag(ELoc, diag::err_omp_wrong_dsa)
             << getOpenMPClauseName(DVar.CKind)
             << getOpenMPClauseName(OMPC_firstprivate);
-        reportOriginalDsa(*this, DSAStack, D, DVar);
+        reportOriginalDsa(SemaRef, DSAStack, D, DVar);
         continue;
       }
 
@@ -18679,7 +18750,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
           Diag(ELoc, diag::err_omp_required_access)
               << getOpenMPClauseName(OMPC_firstprivate)
               << getOpenMPClauseName(OMPC_shared);
-          reportOriginalDsa(*this, DSAStack, D, DVar);
+          reportOriginalDsa(SemaRef, DSAStack, D, DVar);
           continue;
         }
       }
@@ -18712,7 +18783,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
              isOpenMPTeamsDirective(DVar.DKind))) {
           Diag(ELoc, diag::err_omp_parallel_reduction_in_task_firstprivate)
               << getOpenMPDirectiveName(DVar.DKind);
-          reportOriginalDsa(*this, DSAStack, D, DVar);
+          reportOriginalDsa(SemaRef, DSAStack, D, DVar);
           continue;
         }
       }
@@ -18725,7 +18796,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
       // A list item cannot appear in both a map clause and a data-sharing
       // attribute clause on the same construct unless the construct is a
       // combined construct.
-      if ((LangOpts.OpenMP <= 45 &&
+      if ((getLangOpts().OpenMP <= 45 &&
            isOpenMPTargetExecutionDirective(CurrDir)) ||
           CurrDir == OMPD_target) {
         OpenMPClauseKind ConflictKind;
@@ -18741,7 +18812,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
               << getOpenMPClauseName(OMPC_firstprivate)
               << getOpenMPClauseName(ConflictKind)
               << getOpenMPDirectiveName(DSAStack->getCurrentDirective());
-          reportOriginalDsa(*this, DSAStack, D, DVar);
+          reportOriginalDsa(SemaRef, DSAStack, D, DVar);
           continue;
         }
       }
@@ -18753,7 +18824,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
       Diag(ELoc, diag::err_omp_variably_modified_type_not_supported)
           << getOpenMPClauseName(OMPC_firstprivate) << Type
           << getOpenMPDirectiveName(DSAStack->getCurrentDirective());
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -18763,7 +18834,7 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
 
     Type = Type.getUnqualifiedType();
     VarDecl *VDPrivate =
-        buildVarDecl(*this, ELoc, Type, D->getName(),
+        buildVarDecl(SemaRef, ELoc, Type, D->getName(),
                      D->hasAttrs() ? &D->getAttrs() : nullptr,
                      VD ? cast<DeclRefExpr>(SimpleRefExpr) : nullptr);
     // Generate helper private variable and initialize it with the value of the
@@ -18776,32 +18847,32 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
     // original array element in CodeGen.
     if (Type->isArrayType()) {
       VarDecl *VDInit =
-          buildVarDecl(*this, RefExpr->getExprLoc(), ElemType, D->getName());
-      VDInitRefExpr = buildDeclRefExpr(*this, VDInit, ElemType, ELoc);
-      Expr *Init = DefaultLvalueConversion(VDInitRefExpr).get();
+          buildVarDecl(SemaRef, RefExpr->getExprLoc(), ElemType, D->getName());
+      VDInitRefExpr = buildDeclRefExpr(SemaRef, VDInit, ElemType, ELoc);
+      Expr *Init = SemaRef.DefaultLvalueConversion(VDInitRefExpr).get();
       ElemType = ElemType.getUnqualifiedType();
-      VarDecl *VDInitTemp = buildVarDecl(*this, RefExpr->getExprLoc(), ElemType,
-                                         ".firstprivate.temp");
+      VarDecl *VDInitTemp = buildVarDecl(SemaRef, RefExpr->getExprLoc(),
+                                         ElemType, ".firstprivate.temp");
       InitializedEntity Entity =
           InitializedEntity::InitializeVariable(VDInitTemp);
       InitializationKind Kind = InitializationKind::CreateCopy(ELoc, ELoc);
 
-      InitializationSequence InitSeq(*this, Entity, Kind, Init);
-      ExprResult Result = InitSeq.Perform(*this, Entity, Kind, Init);
+      InitializationSequence InitSeq(SemaRef, Entity, Kind, Init);
+      ExprResult Result = InitSeq.Perform(SemaRef, Entity, Kind, Init);
       if (Result.isInvalid())
         VDPrivate->setInvalidDecl();
       else
         VDPrivate->setInit(Result.getAs<Expr>());
       // Remove temp variable declaration.
-      Context.Deallocate(VDInitTemp);
+      getASTContext().Deallocate(VDInitTemp);
     } else {
-      VarDecl *VDInit = buildVarDecl(*this, RefExpr->getExprLoc(), Type,
+      VarDecl *VDInit = buildVarDecl(SemaRef, RefExpr->getExprLoc(), Type,
                                      ".firstprivate.temp");
-      VDInitRefExpr = buildDeclRefExpr(*this, VDInit, RefExpr->getType(),
+      VDInitRefExpr = buildDeclRefExpr(SemaRef, VDInit, RefExpr->getType(),
                                        RefExpr->getExprLoc());
-      AddInitializerToDecl(VDPrivate,
-                           DefaultLvalueConversion(VDInitRefExpr).get(),
-                           /*DirectInit=*/false);
+      SemaRef.AddInitializerToDecl(
+          VDPrivate, SemaRef.DefaultLvalueConversion(VDInitRefExpr).get(),
+          /*DirectInit=*/false);
     }
     if (VDPrivate->isInvalidDecl()) {
       if (IsImplicitClause) {
@@ -18810,29 +18881,30 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
       }
       continue;
     }
-    CurContext->addDecl(VDPrivate);
+    SemaRef.CurContext->addDecl(VDPrivate);
     DeclRefExpr *VDPrivateRefExpr = buildDeclRefExpr(
-        *this, VDPrivate, RefExpr->getType().getUnqualifiedType(),
+        SemaRef, VDPrivate, RefExpr->getType().getUnqualifiedType(),
         RefExpr->getExprLoc());
     DeclRefExpr *Ref = nullptr;
-    if (!VD && !CurContext->isDependentContext()) {
+    if (!VD && !SemaRef.CurContext->isDependentContext()) {
       if (TopDVar.CKind == OMPC_lastprivate) {
         Ref = TopDVar.PrivateCopy;
       } else {
         auto *FD = dyn_cast<FieldDecl>(D);
         VarDecl *VD = FD ? DSAStack->getImplicitFDCapExprDecl(FD) : nullptr;
         if (VD)
-          Ref = buildDeclRefExpr(*this, VD, VD->getType().getNonReferenceType(),
-                                 RefExpr->getExprLoc());
+          Ref =
+              buildDeclRefExpr(SemaRef, VD, VD->getType().getNonReferenceType(),
+                               RefExpr->getExprLoc());
         else
-          Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
+          Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
         if (VD || !isOpenMPCapturedDecl(D))
           ExprCaptures.push_back(Ref->getDecl());
       }
     }
     if (!IsImplicitClause)
       DSAStack->addDSA(D, RefExpr->IgnoreParens(), OMPC_firstprivate, Ref);
-    Vars.push_back((VD || CurContext->isDependentContext())
+    Vars.push_back((VD || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
     PrivateCopies.push_back(VDPrivateRefExpr);
@@ -18842,12 +18914,12 @@ OMPClause *Sema::ActOnOpenMPFirstprivateClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPFirstprivateClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                       Vars, PrivateCopies, Inits,
-                                       buildPreInits(Context, ExprCaptures));
+  return OMPFirstprivateClause::Create(
+      getASTContext(), StartLoc, LParenLoc, EndLoc, Vars, PrivateCopies, Inits,
+      buildPreInits(getASTContext(), ExprCaptures));
 }
 
-OMPClause *Sema::ActOnOpenMPLastprivateClause(
+OMPClause *SemaOpenMP::ActOnOpenMPLastprivateClause(
     ArrayRef<Expr *> VarList, OpenMPLastprivateModifier LPKind,
     SourceLocation LPKindLoc, SourceLocation ColonLoc, SourceLocation StartLoc,
     SourceLocation LParenLoc, SourceLocation EndLoc) {
@@ -18871,7 +18943,7 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -18889,8 +18961,8 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
     // OpenMP [2.14.3.5, Restrictions, C/C++, p.2]
     //  A variable that appears in a lastprivate clause must not have an
     //  incomplete type or a reference type.
-    if (RequireCompleteType(ELoc, Type,
-                            diag::err_omp_lastprivate_incomplete_type))
+    if (SemaRef.RequireCompleteType(ELoc, Type,
+                                    diag::err_omp_lastprivate_incomplete_type))
       continue;
     Type = Type.getNonReferenceType();
 
@@ -18902,7 +18974,7 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
     // OpenMP 3.1 [2.9.3.5, lastprivate clause, Restrictions]
     // A variable that appears in a lastprivate clause must not have a
     // const-qualified type unless it is of class type with a mutable member.
-    if (rejectConstNotMutableType(*this, D, Type, OMPC_lastprivate, ELoc))
+    if (rejectConstNotMutableType(SemaRef, D, Type, OMPC_lastprivate, ELoc))
       continue;
 
     // OpenMP 5.0 [2.19.4.5 lastprivate Clause, Restrictions]
@@ -18910,7 +18982,7 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
     // modifier must be a scalar variable.
     if (LPKind == OMPC_LASTPRIVATE_conditional && !Type->isScalarType()) {
       Diag(ELoc, diag::err_omp_lastprivate_conditional_non_scalar);
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -18935,7 +19007,7 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
       Diag(ELoc, diag::err_omp_wrong_dsa)
           << getOpenMPClauseName(DVar.CKind)
           << getOpenMPClauseName(OMPC_lastprivate);
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
@@ -18954,7 +19026,7 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
         Diag(ELoc, diag::err_omp_required_access)
             << getOpenMPClauseName(OMPC_lastprivate)
             << getOpenMPClauseName(OMPC_shared);
-        reportOriginalDsa(*this, DSAStack, D, DVar);
+        reportOriginalDsa(SemaRef, DSAStack, D, DVar);
         continue;
       }
     }
@@ -18967,53 +19039,53 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
     //  A variable of class type (or array thereof) that appears in a
     //  lastprivate clause requires an accessible, unambiguous copy assignment
     //  operator for the class type.
-    Type = Context.getBaseElementType(Type).getNonReferenceType();
-    VarDecl *SrcVD = buildVarDecl(*this, ERange.getBegin(),
+    Type = getASTContext().getBaseElementType(Type).getNonReferenceType();
+    VarDecl *SrcVD = buildVarDecl(SemaRef, ERange.getBegin(),
                                   Type.getUnqualifiedType(), ".lastprivate.src",
                                   D->hasAttrs() ? &D->getAttrs() : nullptr);
     DeclRefExpr *PseudoSrcExpr =
-        buildDeclRefExpr(*this, SrcVD, Type.getUnqualifiedType(), ELoc);
+        buildDeclRefExpr(SemaRef, SrcVD, Type.getUnqualifiedType(), ELoc);
     VarDecl *DstVD =
-        buildVarDecl(*this, ERange.getBegin(), Type, ".lastprivate.dst",
+        buildVarDecl(SemaRef, ERange.getBegin(), Type, ".lastprivate.dst",
                      D->hasAttrs() ? &D->getAttrs() : nullptr);
-    DeclRefExpr *PseudoDstExpr = buildDeclRefExpr(*this, DstVD, Type, ELoc);
+    DeclRefExpr *PseudoDstExpr = buildDeclRefExpr(SemaRef, DstVD, Type, ELoc);
     // For arrays generate assignment operation for single element and replace
     // it by the original array element in CodeGen.
-    ExprResult AssignmentOp = BuildBinOp(/*S=*/nullptr, ELoc, BO_Assign,
-                                         PseudoDstExpr, PseudoSrcExpr);
+    ExprResult AssignmentOp = SemaRef.BuildBinOp(/*S=*/nullptr, ELoc, BO_Assign,
+                                                 PseudoDstExpr, PseudoSrcExpr);
     if (AssignmentOp.isInvalid())
       continue;
-    AssignmentOp =
-        ActOnFinishFullExpr(AssignmentOp.get(), ELoc, /*DiscardedValue*/ false);
+    AssignmentOp = SemaRef.ActOnFinishFullExpr(AssignmentOp.get(), ELoc,
+                                               /*DiscardedValue*/ false);
     if (AssignmentOp.isInvalid())
       continue;
 
     DeclRefExpr *Ref = nullptr;
-    if (!VD && !CurContext->isDependentContext()) {
+    if (!VD && !SemaRef.CurContext->isDependentContext()) {
       if (TopDVar.CKind == OMPC_firstprivate) {
         Ref = TopDVar.PrivateCopy;
       } else {
-        Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/false);
+        Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/false);
         if (!isOpenMPCapturedDecl(D))
           ExprCaptures.push_back(Ref->getDecl());
       }
       if ((TopDVar.CKind == OMPC_firstprivate && !TopDVar.PrivateCopy) ||
           (!isOpenMPCapturedDecl(D) &&
            Ref->getDecl()->hasAttr<OMPCaptureNoInitAttr>())) {
-        ExprResult RefRes = DefaultLvalueConversion(Ref);
+        ExprResult RefRes = SemaRef.DefaultLvalueConversion(Ref);
         if (!RefRes.isUsable())
           continue;
         ExprResult PostUpdateRes =
-            BuildBinOp(DSAStack->getCurScope(), ELoc, BO_Assign, SimpleRefExpr,
-                       RefRes.get());
+            SemaRef.BuildBinOp(DSAStack->getCurScope(), ELoc, BO_Assign,
+                               SimpleRefExpr, RefRes.get());
         if (!PostUpdateRes.isUsable())
           continue;
         ExprPostUpdates.push_back(
-            IgnoredValueConversions(PostUpdateRes.get()).get());
+            SemaRef.IgnoredValueConversions(PostUpdateRes.get()).get());
       }
     }
     DSAStack->addDSA(D, RefExpr->IgnoreParens(), OMPC_lastprivate, Ref);
-    Vars.push_back((VD || CurContext->isDependentContext())
+    Vars.push_back((VD || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
     SrcExprs.push_back(PseudoSrcExpr);
@@ -19024,24 +19096,24 @@ OMPClause *Sema::ActOnOpenMPLastprivateClause(
   if (Vars.empty())
     return nullptr;
 
-  return OMPLastprivateClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                      Vars, SrcExprs, DstExprs, AssignmentOps,
-                                      LPKind, LPKindLoc, ColonLoc,
-                                      buildPreInits(Context, ExprCaptures),
-                                      buildPostUpdate(*this, ExprPostUpdates));
+  return OMPLastprivateClause::Create(
+      getASTContext(), StartLoc, LParenLoc, EndLoc, Vars, SrcExprs, DstExprs,
+      AssignmentOps, LPKind, LPKindLoc, ColonLoc,
+      buildPreInits(getASTContext(), ExprCaptures),
+      buildPostUpdate(SemaRef, ExprPostUpdates));
 }
 
-OMPClause *Sema::ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP lastprivate clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -19063,15 +19135,16 @@ OMPClause *Sema::ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
         DVar.RefExpr) {
       Diag(ELoc, diag::err_omp_wrong_dsa) << getOpenMPClauseName(DVar.CKind)
                                           << getOpenMPClauseName(OMPC_shared);
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
     DeclRefExpr *Ref = nullptr;
-    if (!VD && isOpenMPCapturedDecl(D) && !CurContext->isDependentContext())
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
+    if (!VD && isOpenMPCapturedDecl(D) &&
+        !SemaRef.CurContext->isDependentContext())
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
     DSAStack->addDSA(D, RefExpr->IgnoreParens(), OMPC_shared, Ref);
-    Vars.push_back((VD || !Ref || CurContext->isDependentContext())
+    Vars.push_back((VD || !Ref || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
   }
@@ -19079,7 +19152,8 @@ OMPClause *Sema::ActOnOpenMPSharedClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPSharedClause::Create(Context, StartLoc, LParenLoc, EndLoc, Vars);
+  return OMPSharedClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
+                                 Vars);
 }
 
 namespace {
@@ -20200,7 +20274,7 @@ static bool actOnOMPReductionKindClause(
       } else {
         VarsExpr = Ref = buildCapture(S, D, SimpleRefExpr, /*WithInit=*/false);
       }
-      if (!S.isOpenMPCapturedDecl(D)) {
+      if (!S.OpenMP().isOpenMPCapturedDecl(D)) {
         RD.ExprCaptures.emplace_back(Ref->getDecl());
         if (Ref->getDecl()->hasAttr<OMPCaptureNoInitAttr>()) {
           ExprResult RefRes = S.DefaultLvalueConversion(Ref);
@@ -20250,7 +20324,7 @@ static bool actOnOMPReductionKindClause(
   return RD.Vars.empty();
 }
 
-OMPClause *Sema::ActOnOpenMPReductionClause(
+OMPClause *SemaOpenMP::ActOnOpenMPReductionClause(
     ArrayRef<Expr *> VarList, OpenMPReductionClauseModifier Modifier,
     SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ModifierLoc, SourceLocation ColonLoc, SourceLocation EndLoc,
@@ -20279,77 +20353,80 @@ OMPClause *Sema::ActOnOpenMPReductionClause(
   }
 
   ReductionData RD(VarList.size(), Modifier);
-  if (actOnOMPReductionKindClause(*this, DSAStack, OMPC_reduction, VarList,
+  if (actOnOMPReductionKindClause(SemaRef, DSAStack, OMPC_reduction, VarList,
                                   StartLoc, LParenLoc, ColonLoc, EndLoc,
                                   ReductionIdScopeSpec, ReductionId,
                                   UnresolvedReductions, RD))
     return nullptr;
 
   return OMPReductionClause::Create(
-      Context, StartLoc, LParenLoc, ModifierLoc, ColonLoc, EndLoc, Modifier,
-      RD.Vars, ReductionIdScopeSpec.getWithLocInContext(Context), ReductionId,
+      getASTContext(), StartLoc, LParenLoc, ModifierLoc, ColonLoc, EndLoc,
+      Modifier, RD.Vars,
+      ReductionIdScopeSpec.getWithLocInContext(getASTContext()), ReductionId,
       RD.Privates, RD.LHSs, RD.RHSs, RD.ReductionOps, RD.InscanCopyOps,
       RD.InscanCopyArrayTemps, RD.InscanCopyArrayElems,
-      buildPreInits(Context, RD.ExprCaptures),
-      buildPostUpdate(*this, RD.ExprPostUpdates));
+      buildPreInits(getASTContext(), RD.ExprCaptures),
+      buildPostUpdate(SemaRef, RD.ExprPostUpdates));
 }
 
-OMPClause *Sema::ActOnOpenMPTaskReductionClause(
+OMPClause *SemaOpenMP::ActOnOpenMPTaskReductionClause(
     ArrayRef<Expr *> VarList, SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ColonLoc, SourceLocation EndLoc,
     CXXScopeSpec &ReductionIdScopeSpec, const DeclarationNameInfo &ReductionId,
     ArrayRef<Expr *> UnresolvedReductions) {
   ReductionData RD(VarList.size());
-  if (actOnOMPReductionKindClause(*this, DSAStack, OMPC_task_reduction, VarList,
-                                  StartLoc, LParenLoc, ColonLoc, EndLoc,
-                                  ReductionIdScopeSpec, ReductionId,
+  if (actOnOMPReductionKindClause(SemaRef, DSAStack, OMPC_task_reduction,
+                                  VarList, StartLoc, LParenLoc, ColonLoc,
+                                  EndLoc, ReductionIdScopeSpec, ReductionId,
                                   UnresolvedReductions, RD))
     return nullptr;
 
   return OMPTaskReductionClause::Create(
-      Context, StartLoc, LParenLoc, ColonLoc, EndLoc, RD.Vars,
-      ReductionIdScopeSpec.getWithLocInContext(Context), ReductionId,
+      getASTContext(), StartLoc, LParenLoc, ColonLoc, EndLoc, RD.Vars,
+      ReductionIdScopeSpec.getWithLocInContext(getASTContext()), ReductionId,
       RD.Privates, RD.LHSs, RD.RHSs, RD.ReductionOps,
-      buildPreInits(Context, RD.ExprCaptures),
-      buildPostUpdate(*this, RD.ExprPostUpdates));
+      buildPreInits(getASTContext(), RD.ExprCaptures),
+      buildPostUpdate(SemaRef, RD.ExprPostUpdates));
 }
 
-OMPClause *Sema::ActOnOpenMPInReductionClause(
+OMPClause *SemaOpenMP::ActOnOpenMPInReductionClause(
     ArrayRef<Expr *> VarList, SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ColonLoc, SourceLocation EndLoc,
     CXXScopeSpec &ReductionIdScopeSpec, const DeclarationNameInfo &ReductionId,
     ArrayRef<Expr *> UnresolvedReductions) {
   ReductionData RD(VarList.size());
-  if (actOnOMPReductionKindClause(*this, DSAStack, OMPC_in_reduction, VarList,
+  if (actOnOMPReductionKindClause(SemaRef, DSAStack, OMPC_in_reduction, VarList,
                                   StartLoc, LParenLoc, ColonLoc, EndLoc,
                                   ReductionIdScopeSpec, ReductionId,
                                   UnresolvedReductions, RD))
     return nullptr;
 
   return OMPInReductionClause::Create(
-      Context, StartLoc, LParenLoc, ColonLoc, EndLoc, RD.Vars,
-      ReductionIdScopeSpec.getWithLocInContext(Context), ReductionId,
+      getASTContext(), StartLoc, LParenLoc, ColonLoc, EndLoc, RD.Vars,
+      ReductionIdScopeSpec.getWithLocInContext(getASTContext()), ReductionId,
       RD.Privates, RD.LHSs, RD.RHSs, RD.ReductionOps, RD.TaskgroupDescriptors,
-      buildPreInits(Context, RD.ExprCaptures),
-      buildPostUpdate(*this, RD.ExprPostUpdates));
+      buildPreInits(getASTContext(), RD.ExprCaptures),
+      buildPostUpdate(SemaRef, RD.ExprPostUpdates));
 }
 
-bool Sema::CheckOpenMPLinearModifier(OpenMPLinearClauseKind LinKind,
-                                     SourceLocation LinLoc) {
-  if ((!LangOpts.CPlusPlus && LinKind != OMPC_LINEAR_val) ||
+bool SemaOpenMP::CheckOpenMPLinearModifier(OpenMPLinearClauseKind LinKind,
+                                           SourceLocation LinLoc) {
+  if ((!getLangOpts().CPlusPlus && LinKind != OMPC_LINEAR_val) ||
       LinKind == OMPC_LINEAR_unknown || LinKind == OMPC_LINEAR_step) {
-    Diag(LinLoc, diag::err_omp_wrong_linear_modifier) << LangOpts.CPlusPlus;
+    Diag(LinLoc, diag::err_omp_wrong_linear_modifier)
+        << getLangOpts().CPlusPlus;
     return true;
   }
   return false;
 }
 
-bool Sema::CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
-                                 OpenMPLinearClauseKind LinKind, QualType Type,
-                                 bool IsDeclareSimd) {
+bool SemaOpenMP::CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
+                                       OpenMPLinearClauseKind LinKind,
+                                       QualType Type, bool IsDeclareSimd) {
   const auto *VD = dyn_cast_or_null<VarDecl>(D);
   // A variable must not have an incomplete type or a reference type.
-  if (RequireCompleteType(ELoc, Type, diag::err_omp_linear_incomplete_type))
+  if (SemaRef.RequireCompleteType(ELoc, Type,
+                                  diag::err_omp_linear_incomplete_type))
     return true;
   if ((LinKind == OMPC_LINEAR_uval || LinKind == OMPC_LINEAR_ref) &&
       !Type->isReferenceType()) {
@@ -20365,17 +20442,17 @@ bool Sema::CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
   // not apply to the firstprivate clause, nor to the linear clause on
   // declarative directives (like declare simd).
   if (!IsDeclareSimd &&
-      rejectConstNotMutableType(*this, D, Type, OMPC_linear, ELoc))
+      rejectConstNotMutableType(SemaRef, D, Type, OMPC_linear, ELoc))
     return true;
 
   // A list item must be of integral or pointer type.
   Type = Type.getUnqualifiedType().getCanonicalType();
   const auto *Ty = Type.getTypePtrOrNull();
   if (!Ty || (LinKind != OMPC_LINEAR_ref && !Ty->isDependentType() &&
-              !Ty->isIntegralType(Context) && !Ty->isPointerType())) {
+              !Ty->isIntegralType(getASTContext()) && !Ty->isPointerType())) {
     Diag(ELoc, diag::err_omp_linear_expected_int_or_ptr) << Type;
     if (D) {
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -20386,7 +20463,7 @@ bool Sema::CheckOpenMPLinearDecl(const ValueDecl *D, SourceLocation ELoc,
   return false;
 }
 
-OMPClause *Sema::ActOnOpenMPLinearClause(
+OMPClause *SemaOpenMP::ActOnOpenMPLinearClause(
     ArrayRef<Expr *> VarList, Expr *Step, SourceLocation StartLoc,
     SourceLocation LParenLoc, OpenMPLinearClauseKind LinKind,
     SourceLocation LinLoc, SourceLocation ColonLoc,
@@ -20409,7 +20486,7 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -20431,7 +20508,7 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
     if (DVar.RefExpr) {
       Diag(ELoc, diag::err_omp_wrong_dsa) << getOpenMPClauseName(DVar.CKind)
                                           << getOpenMPClauseName(OMPC_linear);
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
@@ -20441,29 +20518,29 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
 
     // Build private copy of original var.
     VarDecl *Private =
-        buildVarDecl(*this, ELoc, Type, D->getName(),
+        buildVarDecl(SemaRef, ELoc, Type, D->getName(),
                      D->hasAttrs() ? &D->getAttrs() : nullptr,
                      VD ? cast<DeclRefExpr>(SimpleRefExpr) : nullptr);
-    DeclRefExpr *PrivateRef = buildDeclRefExpr(*this, Private, Type, ELoc);
+    DeclRefExpr *PrivateRef = buildDeclRefExpr(SemaRef, Private, Type, ELoc);
     // Build var to save initial value.
-    VarDecl *Init = buildVarDecl(*this, ELoc, Type, ".linear.start");
+    VarDecl *Init = buildVarDecl(SemaRef, ELoc, Type, ".linear.start");
     Expr *InitExpr;
     DeclRefExpr *Ref = nullptr;
-    if (!VD && !CurContext->isDependentContext()) {
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/false);
+    if (!VD && !SemaRef.CurContext->isDependentContext()) {
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/false);
       if (!isOpenMPCapturedDecl(D)) {
         ExprCaptures.push_back(Ref->getDecl());
         if (Ref->getDecl()->hasAttr<OMPCaptureNoInitAttr>()) {
-          ExprResult RefRes = DefaultLvalueConversion(Ref);
+          ExprResult RefRes = SemaRef.DefaultLvalueConversion(Ref);
           if (!RefRes.isUsable())
             continue;
           ExprResult PostUpdateRes =
-              BuildBinOp(DSAStack->getCurScope(), ELoc, BO_Assign,
-                         SimpleRefExpr, RefRes.get());
+              SemaRef.BuildBinOp(DSAStack->getCurScope(), ELoc, BO_Assign,
+                                 SimpleRefExpr, RefRes.get());
           if (!PostUpdateRes.isUsable())
             continue;
           ExprPostUpdates.push_back(
-              IgnoredValueConversions(PostUpdateRes.get()).get());
+              SemaRef.IgnoredValueConversions(PostUpdateRes.get()).get());
         }
       }
     }
@@ -20471,12 +20548,13 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
       InitExpr = VD ? VD->getInit() : SimpleRefExpr;
     else
       InitExpr = VD ? SimpleRefExpr : Ref;
-    AddInitializerToDecl(Init, DefaultLvalueConversion(InitExpr).get(),
-                         /*DirectInit=*/false);
-    DeclRefExpr *InitRef = buildDeclRefExpr(*this, Init, Type, ELoc);
+    SemaRef.AddInitializerToDecl(
+        Init, SemaRef.DefaultLvalueConversion(InitExpr).get(),
+        /*DirectInit=*/false);
+    DeclRefExpr *InitRef = buildDeclRefExpr(SemaRef, Init, Type, ELoc);
 
     DSAStack->addDSA(D, RefExpr->IgnoreParens(), OMPC_linear, Ref);
-    Vars.push_back((VD || CurContext->isDependentContext())
+    Vars.push_back((VD || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
     Privates.push_back(PrivateRef);
@@ -20499,17 +20577,18 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
 
     // Build var to save the step value.
     VarDecl *SaveVar =
-        buildVarDecl(*this, StepLoc, StepExpr->getType(), ".linear.step");
+        buildVarDecl(SemaRef, StepLoc, StepExpr->getType(), ".linear.step");
     ExprResult SaveRef =
-        buildDeclRefExpr(*this, SaveVar, StepExpr->getType(), StepLoc);
-    ExprResult CalcStep =
-        BuildBinOp(CurScope, StepLoc, BO_Assign, SaveRef.get(), StepExpr);
-    CalcStep = ActOnFinishFullExpr(CalcStep.get(), /*DiscardedValue*/ false);
+        buildDeclRefExpr(SemaRef, SaveVar, StepExpr->getType(), StepLoc);
+    ExprResult CalcStep = SemaRef.BuildBinOp(
+        SemaRef.getCurScope(), StepLoc, BO_Assign, SaveRef.get(), StepExpr);
+    CalcStep =
+        SemaRef.ActOnFinishFullExpr(CalcStep.get(), /*DiscardedValue*/ false);
 
     // Warn about zero linear step (it would be probably better specified as
     // making corresponding variables 'const').
     if (std::optional<llvm::APSInt> Result =
-            StepExpr->getIntegerConstantExpr(Context)) {
+            StepExpr->getIntegerConstantExpr(getASTContext())) {
       if (!Result->isNegative() && !Result->isStrictlyPositive())
         Diag(StepLoc, diag::warn_omp_linear_step_zero)
             << Vars[0] << (Vars.size() > 1);
@@ -20520,11 +20599,11 @@ OMPClause *Sema::ActOnOpenMPLinearClause(
     }
   }
 
-  return OMPLinearClause::Create(Context, StartLoc, LParenLoc, LinKind, LinLoc,
-                                 ColonLoc, StepModifierLoc, EndLoc, Vars,
-                                 Privates, Inits, StepExpr, CalcStepExpr,
-                                 buildPreInits(Context, ExprCaptures),
-                                 buildPostUpdate(*this, ExprPostUpdates));
+  return OMPLinearClause::Create(getASTContext(), StartLoc, LParenLoc, LinKind,
+                                 LinLoc, ColonLoc, StepModifierLoc, EndLoc,
+                                 Vars, Privates, Inits, StepExpr, CalcStepExpr,
+                                 buildPreInits(getASTContext(), ExprCaptures),
+                                 buildPostUpdate(SemaRef, ExprPostUpdates));
 }
 
 static bool FinishOpenMPLinearClause(OMPLinearClause &Clause, DeclRefExpr *IV,
@@ -20630,7 +20709,7 @@ static bool FinishOpenMPLinearClause(OMPLinearClause &Clause, DeclRefExpr *IV,
   return HasErrors;
 }
 
-OMPClause *Sema::ActOnOpenMPAlignedClause(
+OMPClause *SemaOpenMP::ActOnOpenMPAlignedClause(
     ArrayRef<Expr *> VarList, Expr *Alignment, SourceLocation StartLoc,
     SourceLocation LParenLoc, SourceLocation ColonLoc, SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
@@ -20639,7 +20718,7 @@ OMPClause *Sema::ActOnOpenMPAlignedClause(
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -20659,7 +20738,7 @@ OMPClause *Sema::ActOnOpenMPAlignedClause(
     if (!Ty || (!Ty->isArrayType() && !Ty->isPointerType())) {
       Diag(ELoc, diag::err_omp_aligned_expected_array_or_ptr)
           << QType << getLangOpts().CPlusPlus << ERange;
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -20679,9 +20758,10 @@ OMPClause *Sema::ActOnOpenMPAlignedClause(
 
     DeclRefExpr *Ref = nullptr;
     if (!VD && isOpenMPCapturedDecl(D))
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
-    Vars.push_back(DefaultFunctionArrayConversion(
-                       (VD || !Ref) ? RefExpr->IgnoreParens() : Ref)
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
+    Vars.push_back(SemaRef
+                       .DefaultFunctionArrayConversion(
+                           (VD || !Ref) ? RefExpr->IgnoreParens() : Ref)
                        .get());
   }
 
@@ -20700,14 +20780,14 @@ OMPClause *Sema::ActOnOpenMPAlignedClause(
   if (Vars.empty())
     return nullptr;
 
-  return OMPAlignedClause::Create(Context, StartLoc, LParenLoc, ColonLoc,
-                                  EndLoc, Vars, Alignment);
+  return OMPAlignedClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                  ColonLoc, EndLoc, Vars, Alignment);
 }
 
-OMPClause *Sema::ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
-                                         SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   SmallVector<Expr *, 8> SrcExprs;
   SmallVector<Expr *, 8> DstExprs;
@@ -20761,26 +20841,28 @@ OMPClause *Sema::ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
     //  A variable of class type (or array thereof) that appears in a
     //  copyin clause requires an accessible, unambiguous copy assignment
     //  operator for the class type.
-    QualType ElemType = Context.getBaseElementType(Type).getNonReferenceType();
+    QualType ElemType =
+        getASTContext().getBaseElementType(Type).getNonReferenceType();
     VarDecl *SrcVD =
-        buildVarDecl(*this, DE->getBeginLoc(), ElemType.getUnqualifiedType(),
+        buildVarDecl(SemaRef, DE->getBeginLoc(), ElemType.getUnqualifiedType(),
                      ".copyin.src", VD->hasAttrs() ? &VD->getAttrs() : nullptr);
     DeclRefExpr *PseudoSrcExpr = buildDeclRefExpr(
-        *this, SrcVD, ElemType.getUnqualifiedType(), DE->getExprLoc());
+        SemaRef, SrcVD, ElemType.getUnqualifiedType(), DE->getExprLoc());
     VarDecl *DstVD =
-        buildVarDecl(*this, DE->getBeginLoc(), ElemType, ".copyin.dst",
+        buildVarDecl(SemaRef, DE->getBeginLoc(), ElemType, ".copyin.dst",
                      VD->hasAttrs() ? &VD->getAttrs() : nullptr);
     DeclRefExpr *PseudoDstExpr =
-        buildDeclRefExpr(*this, DstVD, ElemType, DE->getExprLoc());
+        buildDeclRefExpr(SemaRef, DstVD, ElemType, DE->getExprLoc());
     // For arrays generate assignment operation for single element and replace
     // it by the original array element in CodeGen.
     ExprResult AssignmentOp =
-        BuildBinOp(/*S=*/nullptr, DE->getExprLoc(), BO_Assign, PseudoDstExpr,
-                   PseudoSrcExpr);
+        SemaRef.BuildBinOp(/*S=*/nullptr, DE->getExprLoc(), BO_Assign,
+                           PseudoDstExpr, PseudoSrcExpr);
     if (AssignmentOp.isInvalid())
       continue;
-    AssignmentOp = ActOnFinishFullExpr(AssignmentOp.get(), DE->getExprLoc(),
-                                       /*DiscardedValue*/ false);
+    AssignmentOp =
+        SemaRef.ActOnFinishFullExpr(AssignmentOp.get(), DE->getExprLoc(),
+                                    /*DiscardedValue*/ false);
     if (AssignmentOp.isInvalid())
       continue;
 
@@ -20794,14 +20876,14 @@ OMPClause *Sema::ActOnOpenMPCopyinClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPCopyinClause::Create(Context, StartLoc, LParenLoc, EndLoc, Vars,
-                                 SrcExprs, DstExprs, AssignmentOps);
+  return OMPCopyinClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
+                                 Vars, SrcExprs, DstExprs, AssignmentOps);
 }
 
-OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
-                                              SourceLocation StartLoc,
-                                              SourceLocation LParenLoc,
-                                              SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
+                                                    SourceLocation StartLoc,
+                                                    SourceLocation LParenLoc,
+                                                    SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   SmallVector<Expr *, 8> SrcExprs;
   SmallVector<Expr *, 8> DstExprs;
@@ -20811,7 +20893,7 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -20837,7 +20919,7 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
         Diag(ELoc, diag::err_omp_wrong_dsa)
             << getOpenMPClauseName(DVar.CKind)
             << getOpenMPClauseName(OMPC_copyprivate);
-        reportOriginalDsa(*this, DSAStack, D, DVar);
+        reportOriginalDsa(SemaRef, DSAStack, D, DVar);
         continue;
       }
 
@@ -20850,7 +20932,7 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
           Diag(ELoc, diag::err_omp_required_access)
               << getOpenMPClauseName(OMPC_copyprivate)
               << "threadprivate or private in the enclosing context";
-          reportOriginalDsa(*this, DSAStack, D, DVar);
+          reportOriginalDsa(SemaRef, DSAStack, D, DVar);
           continue;
         }
       }
@@ -20861,7 +20943,7 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
       Diag(ELoc, diag::err_omp_variably_modified_type_not_supported)
           << getOpenMPClauseName(OMPC_copyprivate) << Type
           << getOpenMPDirectiveName(DSAStack->getCurrentDirective());
-      bool IsDecl = !VD || VD->isThisDeclarationADefinition(Context) ==
+      bool IsDecl = !VD || VD->isThisDeclarationADefinition(getASTContext()) ==
                                VarDecl::DeclarationOnly;
       Diag(D->getLocation(),
            IsDecl ? diag::note_previous_decl : diag::note_defined_here)
@@ -20873,22 +20955,23 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
     //  A variable of class type (or array thereof) that appears in a
     //  copyin clause requires an accessible, unambiguous copy assignment
     //  operator for the class type.
-    Type = Context.getBaseElementType(Type.getNonReferenceType())
+    Type = getASTContext()
+               .getBaseElementType(Type.getNonReferenceType())
                .getUnqualifiedType();
     VarDecl *SrcVD =
-        buildVarDecl(*this, RefExpr->getBeginLoc(), Type, ".copyprivate.src",
+        buildVarDecl(SemaRef, RefExpr->getBeginLoc(), Type, ".copyprivate.src",
                      D->hasAttrs() ? &D->getAttrs() : nullptr);
-    DeclRefExpr *PseudoSrcExpr = buildDeclRefExpr(*this, SrcVD, Type, ELoc);
+    DeclRefExpr *PseudoSrcExpr = buildDeclRefExpr(SemaRef, SrcVD, Type, ELoc);
     VarDecl *DstVD =
-        buildVarDecl(*this, RefExpr->getBeginLoc(), Type, ".copyprivate.dst",
+        buildVarDecl(SemaRef, RefExpr->getBeginLoc(), Type, ".copyprivate.dst",
                      D->hasAttrs() ? &D->getAttrs() : nullptr);
-    DeclRefExpr *PseudoDstExpr = buildDeclRefExpr(*this, DstVD, Type, ELoc);
-    ExprResult AssignmentOp = BuildBinOp(
+    DeclRefExpr *PseudoDstExpr = buildDeclRefExpr(SemaRef, DstVD, Type, ELoc);
+    ExprResult AssignmentOp = SemaRef.BuildBinOp(
         DSAStack->getCurScope(), ELoc, BO_Assign, PseudoDstExpr, PseudoSrcExpr);
     if (AssignmentOp.isInvalid())
       continue;
-    AssignmentOp =
-        ActOnFinishFullExpr(AssignmentOp.get(), ELoc, /*DiscardedValue*/ false);
+    AssignmentOp = SemaRef.ActOnFinishFullExpr(AssignmentOp.get(), ELoc,
+                                               /*DiscardedValue*/ false);
     if (AssignmentOp.isInvalid())
       continue;
 
@@ -20897,7 +20980,7 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
     assert(VD || isOpenMPCapturedDecl(D));
     Vars.push_back(
         VD ? RefExpr->IgnoreParens()
-           : buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/false));
+           : buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/false));
     SrcExprs.push_back(PseudoSrcExpr);
     DstExprs.push_back(PseudoDstExpr);
     AssignmentOps.push_back(AssignmentOp.get());
@@ -20906,18 +20989,20 @@ OMPClause *Sema::ActOnOpenMPCopyprivateClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPCopyprivateClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                      Vars, SrcExprs, DstExprs, AssignmentOps);
+  return OMPCopyprivateClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                      EndLoc, Vars, SrcExprs, DstExprs,
+                                      AssignmentOps);
 }
 
-OMPClause *Sema::ActOnOpenMPFlushClause(ArrayRef<Expr *> VarList,
-                                        SourceLocation StartLoc,
-                                        SourceLocation LParenLoc,
-                                        SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPFlushClause(ArrayRef<Expr *> VarList,
+                                              SourceLocation StartLoc,
+                                              SourceLocation LParenLoc,
+                                              SourceLocation EndLoc) {
   if (VarList.empty())
     return nullptr;
 
-  return OMPFlushClause::Create(Context, StartLoc, LParenLoc, EndLoc, VarList);
+  return OMPFlushClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
+                                VarList);
 }
 
 /// Tries to find omp_depend_t. type.
@@ -20937,22 +21022,23 @@ static bool findOMPDependT(Sema &S, SourceLocation Loc, DSAStackTy *Stack,
   return true;
 }
 
-OMPClause *Sema::ActOnOpenMPDepobjClause(Expr *Depobj, SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPDepobjClause(Expr *Depobj,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   if (!Depobj)
     return nullptr;
 
-  bool OMPDependTFound = findOMPDependT(*this, StartLoc, DSAStack);
+  bool OMPDependTFound = findOMPDependT(SemaRef, StartLoc, DSAStack);
 
   // OpenMP 5.0, 2.17.10.1 depobj Construct
   // depobj is an lvalue expression of type omp_depend_t.
   if (!Depobj->isTypeDependent() && !Depobj->isValueDependent() &&
       !Depobj->isInstantiationDependent() &&
       !Depobj->containsUnexpandedParameterPack() &&
-      (OMPDependTFound &&
-       !Context.typesAreCompatible(DSAStack->getOMPDependT(), Depobj->getType(),
-                                   /*CompareUnqualified=*/true))) {
+      (OMPDependTFound && !getASTContext().typesAreCompatible(
+                              DSAStack->getOMPDependT(), Depobj->getType(),
+                              /*CompareUnqualified=*/true))) {
     Diag(Depobj->getExprLoc(), diag::err_omp_expected_omp_depend_t_lvalue)
         << 0 << Depobj->getType() << Depobj->getSourceRange();
   }
@@ -20962,7 +21048,8 @@ OMPClause *Sema::ActOnOpenMPDepobjClause(Expr *Depobj, SourceLocation StartLoc,
         << 1 << Depobj->getSourceRange();
   }
 
-  return OMPDepobjClause::Create(Context, StartLoc, LParenLoc, EndLoc, Depobj);
+  return OMPDepobjClause::Create(getASTContext(), StartLoc, LParenLoc, EndLoc,
+                                 Depobj);
 }
 
 namespace {
@@ -21062,8 +21149,9 @@ ProcessOpenMPDoacrossClauseCommon(Sema &SemaRef, bool IsSource,
         continue;
       }
       if (RHS) {
-        ExprResult RHSRes = SemaRef.VerifyPositiveIntegerConstantInClause(
-            RHS, OMPC_depend, /*StrictlyPositive=*/false);
+        ExprResult RHSRes =
+            SemaRef.OpenMP().VerifyPositiveIntegerConstantInClause(
+                RHS, OMPC_depend, /*StrictlyPositive=*/false);
         if (RHSRes.isInvalid())
           continue;
       }
@@ -21094,11 +21182,10 @@ ProcessOpenMPDoacrossClauseCommon(Sema &SemaRef, bool IsSource,
   return {Vars, OpsOffs, TotalDepCount};
 }
 
-OMPClause *
-Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
-                              Expr *DepModifier, ArrayRef<Expr *> VarList,
-                              SourceLocation StartLoc, SourceLocation LParenLoc,
-                              SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPDependClause(
+    const OMPDependClause::DependDataTy &Data, Expr *DepModifier,
+    ArrayRef<Expr *> VarList, SourceLocation StartLoc, SourceLocation LParenLoc,
+    SourceLocation EndLoc) {
   OpenMPDependClauseKind DepKind = Data.DepKind;
   SourceLocation DepLoc = Data.DepLoc;
   if (DSAStack->getCurrentDirective() == OMPD_ordered &&
@@ -21116,17 +21203,18 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
        DSAStack->getCurrentDirective() == OMPD_depobj) &&
       (DepKind == OMPC_DEPEND_unknown || DepKind == OMPC_DEPEND_source ||
        DepKind == OMPC_DEPEND_sink ||
-       ((LangOpts.OpenMP < 50 ||
+       ((getLangOpts().OpenMP < 50 ||
          DSAStack->getCurrentDirective() == OMPD_depobj) &&
         DepKind == OMPC_DEPEND_depobj))) {
     SmallVector<unsigned, 6> Except = {OMPC_DEPEND_source, OMPC_DEPEND_sink,
                                        OMPC_DEPEND_outallmemory,
                                        OMPC_DEPEND_inoutallmemory};
-    if (LangOpts.OpenMP < 50 || DSAStack->getCurrentDirective() == OMPD_depobj)
+    if (getLangOpts().OpenMP < 50 ||
+        DSAStack->getCurrentDirective() == OMPD_depobj)
       Except.push_back(OMPC_DEPEND_depobj);
-    if (LangOpts.OpenMP < 51)
+    if (getLangOpts().OpenMP < 51)
       Except.push_back(OMPC_DEPEND_inoutset);
-    std::string Expected = (LangOpts.OpenMP >= 50 && !DepModifier)
+    std::string Expected = (getLangOpts().OpenMP >= 50 && !DepModifier)
                                ? "depend modifier(iterator) or "
                                : "";
     Diag(DepLoc, diag::err_omp_unexpected_clause_value)
@@ -21152,7 +21240,7 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
 
   if (DepKind == OMPC_DEPEND_sink || DepKind == OMPC_DEPEND_source) {
     DoacrossDataInfoTy VarOffset = ProcessOpenMPDoacrossClauseCommon(
-        *this, DepKind == OMPC_DEPEND_source, VarList, DSAStack, EndLoc);
+        SemaRef, DepKind == OMPC_DEPEND_source, VarList, DSAStack, EndLoc);
     Vars = VarOffset.Vars;
     OpsOffs = VarOffset.OpsOffs;
     TotalDepCount = VarOffset.TotalDepCount;
@@ -21168,9 +21256,9 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
       SourceLocation ELoc = RefExpr->getExprLoc();
       Expr *SimpleExpr = RefExpr->IgnoreParenCasts();
       if (DepKind != OMPC_DEPEND_sink && DepKind != OMPC_DEPEND_source) {
-        bool OMPDependTFound = LangOpts.OpenMP >= 50;
+        bool OMPDependTFound = getLangOpts().OpenMP >= 50;
         if (OMPDependTFound)
-          OMPDependTFound = findOMPDependT(*this, StartLoc, DSAStack,
+          OMPDependTFound = findOMPDependT(SemaRef, StartLoc, DSAStack,
                                            DepKind == OMPC_DEPEND_depobj);
         if (DepKind == OMPC_DEPEND_depobj) {
           // OpenMP 5.0, 2.17.11 depend Clause, Restrictions, C/C++
@@ -21180,8 +21268,8 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
               !RefExpr->isInstantiationDependent() &&
               !RefExpr->containsUnexpandedParameterPack() &&
               (OMPDependTFound &&
-               !Context.hasSameUnqualifiedType(DSAStack->getOMPDependT(),
-                                               RefExpr->getType()))) {
+               !getASTContext().hasSameUnqualifiedType(
+                   DSAStack->getOMPDependT(), RefExpr->getType()))) {
             Diag(ELoc, diag::err_omp_expected_omp_depend_t_lvalue)
                 << 0 << RefExpr->getType() << RefExpr->getSourceRange();
             continue;
@@ -21212,7 +21300,7 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
             const Expr *Length = OASE->getLength();
             Expr::EvalResult Result;
             if (Length && !Length->isValueDependent() &&
-                Length->EvaluateAsInt(Result, Context) &&
+                Length->EvaluateAsInt(Result, getASTContext()) &&
                 Result.Val.getInt().isZero()) {
               Diag(ELoc,
                    diag::err_omp_depend_zero_length_array_section_not_allowed)
@@ -21232,8 +21320,9 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
                (OMPDependTFound && DSAStack->getOMPDependT().getTypePtr() ==
                                        ExprTy.getTypePtr()))) {
             Diag(ELoc, diag::err_omp_expected_addressable_lvalue_or_array_item)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0) << RefExpr->getSourceRange();
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << RefExpr->getSourceRange();
             continue;
           }
 
@@ -21245,22 +21334,24 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
                    ->isPointerType() &&
               !ASE->getBase()->getType().getNonReferenceType()->isArrayType()) {
             Diag(ELoc, diag::err_omp_expected_addressable_lvalue_or_array_item)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0) << RefExpr->getSourceRange();
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << RefExpr->getSourceRange();
             continue;
           }
 
           ExprResult Res;
           {
-            Sema::TentativeAnalysisScope Trap(*this);
-            Res = CreateBuiltinUnaryOp(ELoc, UO_AddrOf,
-                                       RefExpr->IgnoreParenImpCasts());
+            Sema::TentativeAnalysisScope Trap(SemaRef);
+            Res = SemaRef.CreateBuiltinUnaryOp(ELoc, UO_AddrOf,
+                                               RefExpr->IgnoreParenImpCasts());
           }
           if (!Res.isUsable() && !isa<OMPArraySectionExpr>(SimpleExpr) &&
               !isa<OMPArrayShapingExpr>(SimpleExpr)) {
             Diag(ELoc, diag::err_omp_expected_addressable_lvalue_or_array_item)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0)
-                << (LangOpts.OpenMP >= 50 ? 1 : 0) << RefExpr->getSourceRange();
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << (getLangOpts().OpenMP >= 50 ? 1 : 0)
+                << RefExpr->getSourceRange();
             continue;
           }
         }
@@ -21275,7 +21366,7 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
     return nullptr;
 
   auto *C = OMPDependClause::Create(
-      Context, StartLoc, LParenLoc, EndLoc,
+      getASTContext(), StartLoc, LParenLoc, EndLoc,
       {DepKind, DepLoc, Data.ColonLoc, Data.OmpAllMemoryLoc}, DepModifier, Vars,
       TotalDepCount.getZExtValue());
   if ((DepKind == OMPC_DEPEND_sink || DepKind == OMPC_DEPEND_source) &&
@@ -21284,12 +21375,11 @@ Sema::ActOnOpenMPDependClause(const OMPDependClause::DependDataTy &Data,
   return C;
 }
 
-OMPClause *Sema::ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
-                                         Expr *Device, SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation ModifierLoc,
-                                         SourceLocation EndLoc) {
-  assert((ModifierLoc.isInvalid() || LangOpts.OpenMP >= 50) &&
+OMPClause *SemaOpenMP::ActOnOpenMPDeviceClause(
+    OpenMPDeviceClauseModifier Modifier, Expr *Device, SourceLocation StartLoc,
+    SourceLocation LParenLoc, SourceLocation ModifierLoc,
+    SourceLocation EndLoc) {
+  assert((ModifierLoc.isInvalid() || getLangOpts().OpenMP >= 50) &&
          "Unexpected device modifier in OpenMP < 50.");
 
   bool ErrorFound = false;
@@ -21306,7 +21396,7 @@ OMPClause *Sema::ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
 
   // OpenMP [2.9.1, Restrictions]
   // The device expression must evaluate to a non-negative integer value.
-  ErrorFound = !isNonNegativeIntegerValue(ValExpr, *this, OMPC_device,
+  ErrorFound = !isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_device,
                                           /*StrictlyPositive=*/false) ||
                ErrorFound;
   if (ErrorFound)
@@ -21317,7 +21407,7 @@ OMPClause *Sema::ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
   // the reverse_offload clause must be specified.
   if (Modifier == OMPC_DEVICE_ancestor) {
     if (!DSAStack->hasRequiresDeclWithClause<OMPReverseOffloadClause>()) {
-      targetDiag(
+      SemaRef.targetDiag(
           StartLoc,
           diag::err_omp_device_ancestor_without_requires_reverse_offload);
       ErrorFound = true;
@@ -21326,15 +21416,16 @@ OMPClause *Sema::ActOnOpenMPDeviceClause(OpenMPDeviceClauseModifier Modifier,
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
   OpenMPDirectiveKind CaptureRegion =
-      getOpenMPCaptureRegionForClause(DKind, OMPC_device, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+      getOpenMPCaptureRegionForClause(DKind, OMPC_device, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPDeviceClause(Modifier, ValExpr, HelperValStmt, CaptureRegion, StartLoc,
                       LParenLoc, ModifierLoc, EndLoc);
 }
@@ -22527,7 +22618,7 @@ static void checkMappableExpressionList(
   }
 }
 
-OMPClause *Sema::ActOnOpenMPMapClause(
+OMPClause *SemaOpenMP::ActOnOpenMPMapClause(
     Expr *IteratorModifier, ArrayRef<OpenMPMapModifierKind> MapTypeModifiers,
     ArrayRef<SourceLocation> MapTypeModifiersLoc,
     CXXScopeSpec &MapperIdScopeSpec, DeclarationNameInfo &MapperId,
@@ -22562,7 +22653,7 @@ OMPClause *Sema::ActOnOpenMPMapClause(
   }
 
   MappableVarListInfo MVLI(VarList);
-  checkMappableExpressionList(*this, DSAStack, OMPC_map, MVLI, Locs.StartLoc,
+  checkMappableExpressionList(SemaRef, DSAStack, OMPC_map, MVLI, Locs.StartLoc,
                               MapperIdScopeSpec, MapperId, UnresolvedMappers,
                               MapType, Modifiers, IsMapTypeImplicit,
                               NoDiagnose);
@@ -22570,17 +22661,17 @@ OMPClause *Sema::ActOnOpenMPMapClause(
   // We need to produce a map clause even if we don't have variables so that
   // other diagnostics related with non-existing map clauses are accurate.
   return OMPMapClause::Create(
-      Context, Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
       MVLI.VarComponents, MVLI.UDMapperList, IteratorModifier, Modifiers,
-      ModifiersLoc, MapperIdScopeSpec.getWithLocInContext(Context), MapperId,
-      MapType, IsMapTypeImplicit, MapLoc);
+      ModifiersLoc, MapperIdScopeSpec.getWithLocInContext(getASTContext()),
+      MapperId, MapType, IsMapTypeImplicit, MapLoc);
 }
 
-QualType Sema::ActOnOpenMPDeclareReductionType(SourceLocation TyLoc,
-                                               TypeResult ParsedType) {
+QualType SemaOpenMP::ActOnOpenMPDeclareReductionType(SourceLocation TyLoc,
+                                                     TypeResult ParsedType) {
   assert(ParsedType.isUsable());
 
-  QualType ReductionType = GetTypeFromParser(ParsedType.get());
+  QualType ReductionType = SemaRef.GetTypeFromParser(ParsedType.get());
   if (ReductionType.isNull())
     return QualType();
 
@@ -22608,15 +22699,17 @@ QualType Sema::ActOnOpenMPDeclareReductionType(SourceLocation TyLoc,
   return ReductionType;
 }
 
-Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveStart(
+SemaOpenMP::DeclGroupPtrTy
+SemaOpenMP::ActOnOpenMPDeclareReductionDirectiveStart(
     Scope *S, DeclContext *DC, DeclarationName Name,
     ArrayRef<std::pair<QualType, SourceLocation>> ReductionTypes,
     AccessSpecifier AS, Decl *PrevDeclInScope) {
   SmallVector<Decl *, 8> Decls;
   Decls.reserve(ReductionTypes.size());
 
-  LookupResult Lookup(*this, Name, SourceLocation(), LookupOMPReductionName,
-                      forRedeclarationInCurContext());
+  LookupResult Lookup(SemaRef, Name, SourceLocation(),
+                      Sema::LookupOMPReductionName,
+                      SemaRef.forRedeclarationInCurContext());
   // [OpenMP 4.0], 2.15 declare reduction Directive, Restrictions
   // A reduction-identifier may not be re-declared in the current scope for the
   // same type or for a type that is compatible according to the base language
@@ -22627,12 +22720,12 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveStart(
   if (S != nullptr) {
     // Find previous declaration with the same name not referenced in other
     // declarations.
-    FunctionScopeInfo *ParentFn = getEnclosingFunction();
+    FunctionScopeInfo *ParentFn = SemaRef.getEnclosingFunction();
     InCompoundScope =
         (ParentFn != nullptr) && !ParentFn->CompoundScopes.empty();
-    LookupName(Lookup, S);
-    FilterLookupForScope(Lookup, DC, S, /*ConsiderLinkage=*/false,
-                         /*AllowInlineNamespace=*/false);
+    SemaRef.LookupName(Lookup, S);
+    SemaRef.FilterLookupForScope(Lookup, DC, S, /*ConsiderLinkage=*/false,
+                                 /*AllowInlineNamespace=*/false);
     llvm::DenseMap<OMPDeclareReductionDecl *, bool> UsedAsPrevious;
     LookupResult::Filter Filter = Lookup.makeFilter();
     while (Filter.hasNext()) {
@@ -22675,8 +22768,8 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveStart(
       Invalid = true;
     }
     PreviousRedeclTypes[TyData.first.getCanonicalType()] = TyData.second;
-    auto *DRD = OMPDeclareReductionDecl::Create(Context, DC, TyData.second,
-                                                Name, TyData.first, PrevDRD);
+    auto *DRD = OMPDeclareReductionDecl::Create(
+        getASTContext(), DC, TyData.second, Name, TyData.first, PrevDRD);
     DC->addDecl(DRD);
     DRD->setAccess(AS);
     Decls.push_back(DRD);
@@ -22687,24 +22780,24 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveStart(
   }
 
   return DeclGroupPtrTy::make(
-      DeclGroupRef::Create(Context, Decls.begin(), Decls.size()));
+      DeclGroupRef::Create(getASTContext(), Decls.begin(), Decls.size()));
 }
 
-void Sema::ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D) {
+void SemaOpenMP::ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D) {
   auto *DRD = cast<OMPDeclareReductionDecl>(D);
 
   // Enter new function scope.
-  PushFunctionScope();
-  setFunctionHasBranchProtectedScope();
-  getCurFunction()->setHasOMPDeclareReductionCombiner();
+  SemaRef.PushFunctionScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
+  SemaRef.getCurFunction()->setHasOMPDeclareReductionCombiner();
 
   if (S != nullptr)
-    PushDeclContext(S, DRD);
+    SemaRef.PushDeclContext(S, DRD);
   else
-    CurContext = DRD;
+    SemaRef.CurContext = DRD;
 
-  PushExpressionEvaluationContext(
-      ExpressionEvaluationContext::PotentiallyEvaluated);
+  SemaRef.PushExpressionEvaluationContext(
+      Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 
   QualType ReductionType = DRD->getType();
   // Create 'T* omp_parm;T omp_in;'. All references to 'omp_in' will
@@ -22714,7 +22807,7 @@ void Sema::ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D) {
   // pointers.
   // Create 'T omp_in;' variable.
   VarDecl *OmpInParm =
-      buildVarDecl(*this, D->getLocation(), ReductionType, "omp_in");
+      buildVarDecl(SemaRef, D->getLocation(), ReductionType, "omp_in");
   // Create 'T* omp_parm;T omp_out;'. All references to 'omp_out' will
   // be replaced by '*omp_parm' during codegen. This required because 'omp_out'
   // uses semantics of argument handles by value, but it should be passed by
@@ -22722,28 +22815,29 @@ void Sema::ActOnOpenMPDeclareReductionCombinerStart(Scope *S, Decl *D) {
   // pointers.
   // Create 'T omp_out;' variable.
   VarDecl *OmpOutParm =
-      buildVarDecl(*this, D->getLocation(), ReductionType, "omp_out");
+      buildVarDecl(SemaRef, D->getLocation(), ReductionType, "omp_out");
   if (S != nullptr) {
-    PushOnScopeChains(OmpInParm, S);
-    PushOnScopeChains(OmpOutParm, S);
+    SemaRef.PushOnScopeChains(OmpInParm, S);
+    SemaRef.PushOnScopeChains(OmpOutParm, S);
   } else {
     DRD->addDecl(OmpInParm);
     DRD->addDecl(OmpOutParm);
   }
   Expr *InE =
-      ::buildDeclRefExpr(*this, OmpInParm, ReductionType, D->getLocation());
+      ::buildDeclRefExpr(SemaRef, OmpInParm, ReductionType, D->getLocation());
   Expr *OutE =
-      ::buildDeclRefExpr(*this, OmpOutParm, ReductionType, D->getLocation());
+      ::buildDeclRefExpr(SemaRef, OmpOutParm, ReductionType, D->getLocation());
   DRD->setCombinerData(InE, OutE);
 }
 
-void Sema::ActOnOpenMPDeclareReductionCombinerEnd(Decl *D, Expr *Combiner) {
+void SemaOpenMP::ActOnOpenMPDeclareReductionCombinerEnd(Decl *D,
+                                                        Expr *Combiner) {
   auto *DRD = cast<OMPDeclareReductionDecl>(D);
-  DiscardCleanupsInEvaluationContext();
-  PopExpressionEvaluationContext();
+  SemaRef.DiscardCleanupsInEvaluationContext();
+  SemaRef.PopExpressionEvaluationContext();
 
-  PopDeclContext();
-  PopFunctionScopeInfo();
+  SemaRef.PopDeclContext();
+  SemaRef.PopFunctionScopeInfo();
 
   if (Combiner != nullptr)
     DRD->setCombiner(Combiner);
@@ -22751,20 +22845,21 @@ void Sema::ActOnOpenMPDeclareReductionCombinerEnd(Decl *D, Expr *Combiner) {
     DRD->setInvalidDecl();
 }
 
-VarDecl *Sema::ActOnOpenMPDeclareReductionInitializerStart(Scope *S, Decl *D) {
+VarDecl *SemaOpenMP::ActOnOpenMPDeclareReductionInitializerStart(Scope *S,
+                                                                 Decl *D) {
   auto *DRD = cast<OMPDeclareReductionDecl>(D);
 
   // Enter new function scope.
-  PushFunctionScope();
-  setFunctionHasBranchProtectedScope();
+  SemaRef.PushFunctionScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
   if (S != nullptr)
-    PushDeclContext(S, DRD);
+    SemaRef.PushDeclContext(S, DRD);
   else
-    CurContext = DRD;
+    SemaRef.CurContext = DRD;
 
-  PushExpressionEvaluationContext(
-      ExpressionEvaluationContext::PotentiallyEvaluated);
+  SemaRef.PushExpressionEvaluationContext(
+      Sema::ExpressionEvaluationContext::PotentiallyEvaluated);
 
   QualType ReductionType = DRD->getType();
   // Create 'T* omp_parm;T omp_priv;'. All references to 'omp_priv' will
@@ -22774,7 +22869,7 @@ VarDecl *Sema::ActOnOpenMPDeclareReductionInitializerStart(Scope *S, Decl *D) {
   // pointers.
   // Create 'T omp_priv;' variable.
   VarDecl *OmpPrivParm =
-      buildVarDecl(*this, D->getLocation(), ReductionType, "omp_priv");
+      buildVarDecl(SemaRef, D->getLocation(), ReductionType, "omp_priv");
   // Create 'T* omp_parm;T omp_orig;'. All references to 'omp_orig' will
   // be replaced by '*omp_parm' during codegen. This required because 'omp_orig'
   // uses semantics of argument handles by value, but it should be passed by
@@ -22782,30 +22877,30 @@ VarDecl *Sema::ActOnOpenMPDeclareReductionInitializerStart(Scope *S, Decl *D) {
   // pointers.
   // Create 'T omp_orig;' variable.
   VarDecl *OmpOrigParm =
-      buildVarDecl(*this, D->getLocation(), ReductionType, "omp_orig");
+      buildVarDecl(SemaRef, D->getLocation(), ReductionType, "omp_orig");
   if (S != nullptr) {
-    PushOnScopeChains(OmpPrivParm, S);
-    PushOnScopeChains(OmpOrigParm, S);
+    SemaRef.PushOnScopeChains(OmpPrivParm, S);
+    SemaRef.PushOnScopeChains(OmpOrigParm, S);
   } else {
     DRD->addDecl(OmpPrivParm);
     DRD->addDecl(OmpOrigParm);
   }
   Expr *OrigE =
-      ::buildDeclRefExpr(*this, OmpOrigParm, ReductionType, D->getLocation());
+      ::buildDeclRefExpr(SemaRef, OmpOrigParm, ReductionType, D->getLocation());
   Expr *PrivE =
-      ::buildDeclRefExpr(*this, OmpPrivParm, ReductionType, D->getLocation());
+      ::buildDeclRefExpr(SemaRef, OmpPrivParm, ReductionType, D->getLocation());
   DRD->setInitializerData(OrigE, PrivE);
   return OmpPrivParm;
 }
 
-void Sema::ActOnOpenMPDeclareReductionInitializerEnd(Decl *D, Expr *Initializer,
-                                                     VarDecl *OmpPrivParm) {
+void SemaOpenMP::ActOnOpenMPDeclareReductionInitializerEnd(
+    Decl *D, Expr *Initializer, VarDecl *OmpPrivParm) {
   auto *DRD = cast<OMPDeclareReductionDecl>(D);
-  DiscardCleanupsInEvaluationContext();
-  PopExpressionEvaluationContext();
+  SemaRef.DiscardCleanupsInEvaluationContext();
+  SemaRef.PopExpressionEvaluationContext();
 
-  PopDeclContext();
-  PopFunctionScopeInfo();
+  SemaRef.PopDeclContext();
+  SemaRef.PopFunctionScopeInfo();
 
   if (Initializer != nullptr) {
     DRD->setInitializer(Initializer, OMPDeclareReductionInitKind::Call);
@@ -22819,13 +22914,13 @@ void Sema::ActOnOpenMPDeclareReductionInitializerEnd(Decl *D, Expr *Initializer,
   }
 }
 
-Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveEnd(
+SemaOpenMP::DeclGroupPtrTy SemaOpenMP::ActOnOpenMPDeclareReductionDirectiveEnd(
     Scope *S, DeclGroupPtrTy DeclReductions, bool IsValid) {
   for (Decl *D : DeclReductions.get()) {
     if (IsValid) {
       if (S)
-        PushOnScopeChains(cast<OMPDeclareReductionDecl>(D), S,
-                          /*AddToContext=*/false);
+        SemaRef.PushOnScopeChains(cast<OMPDeclareReductionDecl>(D), S,
+                                  /*AddToContext=*/false);
     } else {
       D->setInvalidDecl();
     }
@@ -22833,25 +22928,26 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareReductionDirectiveEnd(
   return DeclReductions;
 }
 
-TypeResult Sema::ActOnOpenMPDeclareMapperVarDecl(Scope *S, Declarator &D) {
-  TypeSourceInfo *TInfo = GetTypeForDeclarator(D);
+TypeResult SemaOpenMP::ActOnOpenMPDeclareMapperVarDecl(Scope *S,
+                                                       Declarator &D) {
+  TypeSourceInfo *TInfo = SemaRef.GetTypeForDeclarator(D);
   QualType T = TInfo->getType();
   if (D.isInvalidType())
     return true;
 
   if (getLangOpts().CPlusPlus) {
     // Check that there are no default arguments (C++ only).
-    CheckExtraCXXDefaultArguments(D);
+    SemaRef.CheckExtraCXXDefaultArguments(D);
   }
 
-  return CreateParsedType(T, TInfo);
+  return SemaRef.CreateParsedType(T, TInfo);
 }
 
-QualType Sema::ActOnOpenMPDeclareMapperType(SourceLocation TyLoc,
-                                            TypeResult ParsedType) {
+QualType SemaOpenMP::ActOnOpenMPDeclareMapperType(SourceLocation TyLoc,
+                                                  TypeResult ParsedType) {
   assert(ParsedType.isUsable() && "Expect usable parsed mapper type");
 
-  QualType MapperType = GetTypeFromParser(ParsedType.get());
+  QualType MapperType = SemaRef.GetTypeFromParser(ParsedType.get());
   assert(!MapperType.isNull() && "Expect valid mapper type");
 
   // [OpenMP 5.0], 2.19.7.3 declare mapper Directive, Restrictions
@@ -22863,12 +22959,13 @@ QualType Sema::ActOnOpenMPDeclareMapperType(SourceLocation TyLoc,
   return MapperType;
 }
 
-Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareMapperDirective(
+SemaOpenMP::DeclGroupPtrTy SemaOpenMP::ActOnOpenMPDeclareMapperDirective(
     Scope *S, DeclContext *DC, DeclarationName Name, QualType MapperType,
     SourceLocation StartLoc, DeclarationName VN, AccessSpecifier AS,
     Expr *MapperVarRef, ArrayRef<OMPClause *> Clauses, Decl *PrevDeclInScope) {
-  LookupResult Lookup(*this, Name, SourceLocation(), LookupOMPMapperName,
-                      forRedeclarationInCurContext());
+  LookupResult Lookup(SemaRef, Name, SourceLocation(),
+                      Sema::LookupOMPMapperName,
+                      SemaRef.forRedeclarationInCurContext());
   // [OpenMP 5.0], 2.19.7.3 declare mapper Directive, Restrictions
   //  A mapper-identifier may not be redeclared in the current scope for the
   //  same type or for a type that is compatible according to the base language
@@ -22879,12 +22976,12 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareMapperDirective(
   if (S != nullptr) {
     // Find previous declaration with the same name not referenced in other
     // declarations.
-    FunctionScopeInfo *ParentFn = getEnclosingFunction();
+    FunctionScopeInfo *ParentFn = SemaRef.getEnclosingFunction();
     InCompoundScope =
         (ParentFn != nullptr) && !ParentFn->CompoundScopes.empty();
-    LookupName(Lookup, S);
-    FilterLookupForScope(Lookup, DC, S, /*ConsiderLinkage=*/false,
-                         /*AllowInlineNamespace=*/false);
+    SemaRef.LookupName(Lookup, S);
+    SemaRef.FilterLookupForScope(Lookup, DC, S, /*ConsiderLinkage=*/false,
+                                 /*AllowInlineNamespace=*/false);
     llvm::DenseMap<OMPDeclareMapperDecl *, bool> UsedAsPrevious;
     LookupResult::Filter Filter = Lookup.makeFilter();
     while (Filter.hasNext()) {
@@ -22929,13 +23026,14 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareMapperDirective(
   // mappers.
   SmallVector<OMPClause *, 4> ClausesWithImplicit(Clauses.begin(),
                                                   Clauses.end());
-  if (LangOpts.OpenMP >= 50)
-    processImplicitMapsWithDefaultMappers(*this, DSAStack, ClausesWithImplicit);
-  auto *DMD =
-      OMPDeclareMapperDecl::Create(Context, DC, StartLoc, Name, MapperType, VN,
-                                   ClausesWithImplicit, PrevDMD);
+  if (getLangOpts().OpenMP >= 50)
+    processImplicitMapsWithDefaultMappers(SemaRef, DSAStack,
+                                          ClausesWithImplicit);
+  auto *DMD = OMPDeclareMapperDecl::Create(getASTContext(), DC, StartLoc, Name,
+                                           MapperType, VN, ClausesWithImplicit,
+                                           PrevDMD);
   if (S)
-    PushOnScopeChains(DMD, S);
+    SemaRef.PushOnScopeChains(DMD, S);
   else
     DC->addDecl(DMD);
   DMD->setAccess(AS);
@@ -22951,105 +23049,106 @@ Sema::DeclGroupPtrTy Sema::ActOnOpenMPDeclareMapperDirective(
   return DeclGroupPtrTy::make(DeclGroupRef(DMD));
 }
 
-ExprResult
-Sema::ActOnOpenMPDeclareMapperDirectiveVarDecl(Scope *S, QualType MapperType,
-                                               SourceLocation StartLoc,
-                                               DeclarationName VN) {
+ExprResult SemaOpenMP::ActOnOpenMPDeclareMapperDirectiveVarDecl(
+    Scope *S, QualType MapperType, SourceLocation StartLoc,
+    DeclarationName VN) {
   TypeSourceInfo *TInfo =
-      Context.getTrivialTypeSourceInfo(MapperType, StartLoc);
-  auto *VD = VarDecl::Create(Context, Context.getTranslationUnitDecl(),
-                             StartLoc, StartLoc, VN.getAsIdentifierInfo(),
-                             MapperType, TInfo, SC_None);
+      getASTContext().getTrivialTypeSourceInfo(MapperType, StartLoc);
+  auto *VD = VarDecl::Create(
+      getASTContext(), getASTContext().getTranslationUnitDecl(), StartLoc,
+      StartLoc, VN.getAsIdentifierInfo(), MapperType, TInfo, SC_None);
   if (S)
-    PushOnScopeChains(VD, S, /*AddToContext=*/false);
-  Expr *E = buildDeclRefExpr(*this, VD, MapperType, StartLoc);
+    SemaRef.PushOnScopeChains(VD, S, /*AddToContext=*/false);
+  Expr *E = buildDeclRefExpr(SemaRef, VD, MapperType, StartLoc);
   DSAStack->addDeclareMapperVarRef(E);
   return E;
 }
 
-void Sema::ActOnOpenMPIteratorVarDecl(VarDecl *VD) {
+void SemaOpenMP::ActOnOpenMPIteratorVarDecl(VarDecl *VD) {
   if (DSAStack->getDeclareMapperVarRef())
     DSAStack->addIteratorVarDecl(VD);
 }
 
-bool Sema::isOpenMPDeclareMapperVarDeclAllowed(const VarDecl *VD) const {
-  assert(LangOpts.OpenMP && "Expected OpenMP mode.");
+bool SemaOpenMP::isOpenMPDeclareMapperVarDeclAllowed(const VarDecl *VD) const {
+  assert(getLangOpts().OpenMP && "Expected OpenMP mode.");
   const Expr *Ref = DSAStack->getDeclareMapperVarRef();
   if (const auto *DRE = cast_or_null<DeclRefExpr>(Ref)) {
     if (VD->getCanonicalDecl() == DRE->getDecl()->getCanonicalDecl())
       return true;
-    if (VD->isUsableInConstantExpressions(Context))
+    if (VD->isUsableInConstantExpressions(getASTContext()))
       return true;
-    if (LangOpts.OpenMP >= 52 && DSAStack->isIteratorVarDecl(VD))
+    if (getLangOpts().OpenMP >= 52 && DSAStack->isIteratorVarDecl(VD))
       return true;
     return false;
   }
   return true;
 }
 
-const ValueDecl *Sema::getOpenMPDeclareMapperVarName() const {
-  assert(LangOpts.OpenMP && "Expected OpenMP mode.");
+const ValueDecl *SemaOpenMP::getOpenMPDeclareMapperVarName() const {
+  assert(getLangOpts().OpenMP && "Expected OpenMP mode.");
   return cast<DeclRefExpr>(DSAStack->getDeclareMapperVarRef())->getDecl();
 }
 
-OMPClause *Sema::ActOnOpenMPNumTeamsClause(Expr *NumTeams,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNumTeamsClause(Expr *NumTeams,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation LParenLoc,
+                                                 SourceLocation EndLoc) {
   Expr *ValExpr = NumTeams;
   Stmt *HelperValStmt = nullptr;
 
   // OpenMP [teams Constrcut, Restrictions]
   // The num_teams expression must evaluate to a positive integer value.
-  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_num_teams,
+  if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_num_teams,
                                  /*StrictlyPositive=*/true))
     return nullptr;
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
-  OpenMPDirectiveKind CaptureRegion =
-      getOpenMPCaptureRegionForClause(DKind, OMPC_num_teams, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+  OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause(
+      DKind, OMPC_num_teams, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context) OMPNumTeamsClause(ValExpr, HelperValStmt, CaptureRegion,
-                                         StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext()) OMPNumTeamsClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPThreadLimitClause(Expr *ThreadLimit,
-                                              SourceLocation StartLoc,
-                                              SourceLocation LParenLoc,
-                                              SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPThreadLimitClause(Expr *ThreadLimit,
+                                                    SourceLocation StartLoc,
+                                                    SourceLocation LParenLoc,
+                                                    SourceLocation EndLoc) {
   Expr *ValExpr = ThreadLimit;
   Stmt *HelperValStmt = nullptr;
 
   // OpenMP [teams Constrcut, Restrictions]
   // The thread_limit expression must evaluate to a positive integer value.
-  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_thread_limit,
+  if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_thread_limit,
                                  /*StrictlyPositive=*/true))
     return nullptr;
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
   OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause(
-      DKind, OMPC_thread_limit, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+      DKind, OMPC_thread_limit, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context) OMPThreadLimitClause(
+  return new (getASTContext()) OMPThreadLimitClause(
       ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPPriorityClause(Expr *Priority,
-                                           SourceLocation StartLoc,
-                                           SourceLocation LParenLoc,
-                                           SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPPriorityClause(Expr *Priority,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation LParenLoc,
+                                                 SourceLocation EndLoc) {
   Expr *ValExpr = Priority;
   Stmt *HelperValStmt = nullptr;
   OpenMPDirectiveKind CaptureRegion = OMPD_unknown;
@@ -23057,20 +23156,20 @@ OMPClause *Sema::ActOnOpenMPPriorityClause(Expr *Priority,
   // OpenMP [2.9.1, task Constrcut]
   // The priority-value is a non-negative numerical scalar expression.
   if (!isNonNegativeIntegerValue(
-          ValExpr, *this, OMPC_priority,
+          ValExpr, SemaRef, OMPC_priority,
           /*StrictlyPositive=*/false, /*BuildCapture=*/true,
           DSAStack->getCurrentDirective(), &CaptureRegion, &HelperValStmt))
     return nullptr;
 
-  return new (Context) OMPPriorityClause(ValExpr, HelperValStmt, CaptureRegion,
-                                         StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext()) OMPPriorityClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPGrainsizeClause(
+OMPClause *SemaOpenMP::ActOnOpenMPGrainsizeClause(
     OpenMPGrainsizeClauseModifier Modifier, Expr *Grainsize,
     SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ModifierLoc, SourceLocation EndLoc) {
-  assert((ModifierLoc.isInvalid() || LangOpts.OpenMP >= 51) &&
+  assert((ModifierLoc.isInvalid() || getLangOpts().OpenMP >= 51) &&
          "Unexpected grainsize modifier in OpenMP < 51.");
 
   if (ModifierLoc.isValid() && Modifier == OMPC_GRAINSIZE_unknown) {
@@ -23088,23 +23187,23 @@ OMPClause *Sema::ActOnOpenMPGrainsizeClause(
   // OpenMP [2.9.2, taskloop Constrcut]
   // The parameter of the grainsize clause must be a positive integer
   // expression.
-  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_grainsize,
+  if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_grainsize,
                                  /*StrictlyPositive=*/true,
                                  /*BuildCapture=*/true,
                                  DSAStack->getCurrentDirective(),
                                  &CaptureRegion, &HelperValStmt))
     return nullptr;
 
-  return new (Context)
+  return new (getASTContext())
       OMPGrainsizeClause(Modifier, ValExpr, HelperValStmt, CaptureRegion,
                          StartLoc, LParenLoc, ModifierLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPNumTasksClause(
+OMPClause *SemaOpenMP::ActOnOpenMPNumTasksClause(
     OpenMPNumTasksClauseModifier Modifier, Expr *NumTasks,
     SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ModifierLoc, SourceLocation EndLoc) {
-  assert((ModifierLoc.isInvalid() || LangOpts.OpenMP >= 51) &&
+  assert((ModifierLoc.isInvalid() || getLangOpts().OpenMP >= 51) &&
          "Unexpected num_tasks modifier in OpenMP < 51.");
 
   if (ModifierLoc.isValid() && Modifier == OMPC_NUMTASKS_unknown) {
@@ -23123,19 +23222,20 @@ OMPClause *Sema::ActOnOpenMPNumTasksClause(
   // The parameter of the num_tasks clause must be a positive integer
   // expression.
   if (!isNonNegativeIntegerValue(
-          ValExpr, *this, OMPC_num_tasks,
+          ValExpr, SemaRef, OMPC_num_tasks,
           /*StrictlyPositive=*/true, /*BuildCapture=*/true,
           DSAStack->getCurrentDirective(), &CaptureRegion, &HelperValStmt))
     return nullptr;
 
-  return new (Context)
+  return new (getASTContext())
       OMPNumTasksClause(Modifier, ValExpr, HelperValStmt, CaptureRegion,
                         StartLoc, LParenLoc, ModifierLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPHintClause(Expr *Hint, SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPHintClause(Expr *Hint,
+                                             SourceLocation StartLoc,
+                                             SourceLocation LParenLoc,
+                                             SourceLocation EndLoc) {
   // OpenMP [2.13.2, critical construct, Description]
   // ... where hint-expression is an integer constant expression that evaluates
   // to a valid lock hint.
@@ -23143,7 +23243,7 @@ OMPClause *Sema::ActOnOpenMPHintClause(Expr *Hint, SourceLocation StartLoc,
       VerifyPositiveIntegerConstantInClause(Hint, OMPC_hint, false);
   if (HintExpr.isInvalid())
     return nullptr;
-  return new (Context)
+  return new (getASTContext())
       OMPHintClause(HintExpr.get(), StartLoc, LParenLoc, EndLoc);
 }
 
@@ -23163,13 +23263,14 @@ static bool findOMPEventHandleT(Sema &S, SourceLocation Loc,
   return true;
 }
 
-OMPClause *Sema::ActOnOpenMPDetachClause(Expr *Evt, SourceLocation StartLoc,
-                                         SourceLocation LParenLoc,
-                                         SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPDetachClause(Expr *Evt,
+                                               SourceLocation StartLoc,
+                                               SourceLocation LParenLoc,
+                                               SourceLocation EndLoc) {
   if (!Evt->isValueDependent() && !Evt->isTypeDependent() &&
       !Evt->isInstantiationDependent() &&
       !Evt->containsUnexpandedParameterPack()) {
-    if (!findOMPEventHandleT(*this, Evt->getExprLoc(), DSAStack))
+    if (!findOMPEventHandleT(SemaRef, Evt->getExprLoc(), DSAStack))
       return nullptr;
     // OpenMP 5.0, 2.10.1 task Construct.
     // event-handle is a variable of the omp_event_handle_t type.
@@ -23185,9 +23286,9 @@ OMPClause *Sema::ActOnOpenMPDetachClause(Expr *Evt, SourceLocation StartLoc,
           << "omp_event_handle_t" << 0 << Evt->getSourceRange();
       return nullptr;
     }
-    if (!Context.hasSameUnqualifiedType(DSAStack->getOMPEventHandleT(),
-                                        VD->getType()) ||
-        VD->getType().isConstant(Context)) {
+    if (!getASTContext().hasSameUnqualifiedType(DSAStack->getOMPEventHandleT(),
+                                                VD->getType()) ||
+        VD->getType().isConstant(getASTContext())) {
       Diag(Evt->getExprLoc(), diag::err_omp_var_expected)
           << "omp_event_handle_t" << 1 << VD->getType()
           << Evt->getSourceRange();
@@ -23202,15 +23303,16 @@ OMPClause *Sema::ActOnOpenMPDetachClause(Expr *Evt, SourceLocation StartLoc,
       Diag(Evt->getExprLoc(), diag::err_omp_wrong_dsa)
           << getOpenMPClauseName(DVar.CKind)
           << getOpenMPClauseName(OMPC_firstprivate);
-      reportOriginalDsa(*this, DSAStack, VD, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, VD, DVar);
       return nullptr;
     }
   }
 
-  return new (Context) OMPDetachClause(Evt, StartLoc, LParenLoc, EndLoc);
+  return new (getASTContext())
+      OMPDetachClause(Evt, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPDistScheduleClause(
+OMPClause *SemaOpenMP::ActOnOpenMPDistScheduleClause(
     OpenMPDistScheduleClauseKind Kind, Expr *ChunkSize, SourceLocation StartLoc,
     SourceLocation LParenLoc, SourceLocation KindLoc, SourceLocation CommaLoc,
     SourceLocation EndLoc) {
@@ -23241,7 +23343,7 @@ OMPClause *Sema::ActOnOpenMPDistScheduleClause(
       //  chunk_size must be a loop invariant integer expression with a positive
       //  value.
       if (std::optional<llvm::APSInt> Result =
-              ValExpr->getIntegerConstantExpr(Context)) {
+              ValExpr->getIntegerConstantExpr(getASTContext())) {
         if (Result->isSigned() && !Result->isStrictlyPositive()) {
           Diag(ChunkSizeLoc, diag::err_omp_negative_expression_in_clause)
               << "dist_schedule" << ChunkSize->getSourceRange();
@@ -23249,22 +23351,22 @@ OMPClause *Sema::ActOnOpenMPDistScheduleClause(
         }
       } else if (getOpenMPCaptureRegionForClause(
                      DSAStack->getCurrentDirective(), OMPC_dist_schedule,
-                     LangOpts.OpenMP) != OMPD_unknown &&
-                 !CurContext->isDependentContext()) {
-        ValExpr = MakeFullExpr(ValExpr).get();
+                     getLangOpts().OpenMP) != OMPD_unknown &&
+                 !SemaRef.CurContext->isDependentContext()) {
+        ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
         llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-        ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-        HelperValStmt = buildPreInits(Context, Captures);
+        ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+        HelperValStmt = buildPreInits(getASTContext(), Captures);
       }
     }
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPDistScheduleClause(StartLoc, LParenLoc, KindLoc, CommaLoc, EndLoc,
                             Kind, ValExpr, HelperValStmt);
 }
 
-OMPClause *Sema::ActOnOpenMPDefaultmapClause(
+OMPClause *SemaOpenMP::ActOnOpenMPDefaultmapClause(
     OpenMPDefaultmapClauseModifier M, OpenMPDefaultmapClauseKind Kind,
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation MLoc,
     SourceLocation KindLoc, SourceLocation EndLoc) {
@@ -23291,10 +23393,10 @@ OMPClause *Sema::ActOnOpenMPDefaultmapClause(
   } else {
     bool isDefaultmapModifier = (M != OMPC_DEFAULTMAP_MODIFIER_unknown);
     bool isDefaultmapKind = (Kind != OMPC_DEFAULTMAP_unknown) ||
-                            (LangOpts.OpenMP >= 50 && KindLoc.isInvalid());
+                            (getLangOpts().OpenMP >= 50 && KindLoc.isInvalid());
     if (!isDefaultmapKind || !isDefaultmapModifier) {
       StringRef KindValue = "'scalar', 'aggregate', 'pointer'";
-      if (LangOpts.OpenMP == 50) {
+      if (getLangOpts().OpenMP == 50) {
         StringRef ModifierValue = "'alloc', 'from', 'to', 'tofrom', "
                                   "'firstprivate', 'none', 'default'";
         if (!isDefaultmapKind && isDefaultmapModifier) {
@@ -23346,13 +23448,13 @@ OMPClause *Sema::ActOnOpenMPDefaultmapClause(
     DSAStack->setDefaultDMAAttr(M, Kind, StartLoc);
   }
 
-  return new (Context)
+  return new (getASTContext())
       OMPDefaultmapClause(StartLoc, LParenLoc, MLoc, KindLoc, EndLoc, Kind, M);
 }
 
-bool Sema::ActOnStartOpenMPDeclareTargetContext(
+bool SemaOpenMP::ActOnStartOpenMPDeclareTargetContext(
     DeclareTargetContextInfo &DTCI) {
-  DeclContext *CurLexicalContext = getCurLexicalContext();
+  DeclContext *CurLexicalContext = SemaRef.getCurLexicalContext();
   if (!CurLexicalContext->isFileContext() &&
       !CurLexicalContext->isExternCContext() &&
       !CurLexicalContext->isExternCXXContext() &&
@@ -23372,20 +23474,20 @@ bool Sema::ActOnStartOpenMPDeclareTargetContext(
   return true;
 }
 
-const Sema::DeclareTargetContextInfo
-Sema::ActOnOpenMPEndDeclareTargetDirective() {
+const SemaOpenMP::DeclareTargetContextInfo
+SemaOpenMP::ActOnOpenMPEndDeclareTargetDirective() {
   assert(!DeclareTargetNesting.empty() &&
          "check isInOpenMPDeclareTargetContext() first!");
   return DeclareTargetNesting.pop_back_val();
 }
 
-void Sema::ActOnFinishedOpenMPDeclareTargetContext(
+void SemaOpenMP::ActOnFinishedOpenMPDeclareTargetContext(
     DeclareTargetContextInfo &DTCI) {
   for (auto &It : DTCI.ExplicitlyMapped)
     ActOnOpenMPDeclareTargetName(It.first, It.second.Loc, It.second.MT, DTCI);
 }
 
-void Sema::DiagnoseUnterminatedOpenMPDeclareTarget() {
+void SemaOpenMP::DiagnoseUnterminatedOpenMPDeclareTarget() {
   if (DeclareTargetNesting.empty())
     return;
   DeclareTargetContextInfo &DTCI = DeclareTargetNesting.back();
@@ -23393,23 +23495,23 @@ void Sema::DiagnoseUnterminatedOpenMPDeclareTarget() {
       << getOpenMPDirectiveName(DTCI.Kind);
 }
 
-NamedDecl *Sema::lookupOpenMPDeclareTargetName(Scope *CurScope,
-                                               CXXScopeSpec &ScopeSpec,
-                                               const DeclarationNameInfo &Id) {
-  LookupResult Lookup(*this, Id, LookupOrdinaryName);
-  LookupParsedName(Lookup, CurScope, &ScopeSpec, true);
+NamedDecl *SemaOpenMP::lookupOpenMPDeclareTargetName(
+    Scope *CurScope, CXXScopeSpec &ScopeSpec, const DeclarationNameInfo &Id) {
+  LookupResult Lookup(SemaRef, Id, Sema::LookupOrdinaryName);
+  SemaRef.LookupParsedName(Lookup, CurScope, &ScopeSpec, true);
 
   if (Lookup.isAmbiguous())
     return nullptr;
   Lookup.suppressDiagnostics();
 
   if (!Lookup.isSingleResult()) {
-    VarOrFuncDeclFilterCCC CCC(*this);
+    VarOrFuncDeclFilterCCC CCC(SemaRef);
     if (TypoCorrection Corrected =
-            CorrectTypo(Id, LookupOrdinaryName, CurScope, nullptr, CCC,
-                        CTK_ErrorRecovery)) {
-      diagnoseTypo(Corrected, PDiag(diag::err_undeclared_var_use_suggest)
-                                  << Id.getName());
+            SemaRef.CorrectTypo(Id, Sema::LookupOrdinaryName, CurScope, nullptr,
+                                CCC, Sema::CTK_ErrorRecovery)) {
+      SemaRef.diagnoseTypo(Corrected,
+                           SemaRef.PDiag(diag::err_undeclared_var_use_suggest)
+                               << Id.getName());
       checkDeclIsAllowedInOpenMPTarget(nullptr, Corrected.getCorrectionDecl());
       return nullptr;
     }
@@ -23427,9 +23529,9 @@ NamedDecl *Sema::lookupOpenMPDeclareTargetName(Scope *CurScope,
   return ND;
 }
 
-void Sema::ActOnOpenMPDeclareTargetName(NamedDecl *ND, SourceLocation Loc,
-                                        OMPDeclareTargetDeclAttr::MapTypeTy MT,
-                                        DeclareTargetContextInfo &DTCI) {
+void SemaOpenMP::ActOnOpenMPDeclareTargetName(
+    NamedDecl *ND, SourceLocation Loc, OMPDeclareTargetDeclAttr::MapTypeTy MT,
+    DeclareTargetContextInfo &DTCI) {
   assert((isa<VarDecl>(ND) || isa<FunctionDecl>(ND) ||
           isa<FunctionTemplateDecl>(ND)) &&
          "Expected variable, function or function template.");
@@ -23445,7 +23547,7 @@ void Sema::ActOnOpenMPDeclareTargetName(NamedDecl *ND, SourceLocation Loc,
   }
   // Diagnose marking after use as it may lead to incorrect diagnosis and
   // codegen.
-  if (LangOpts.OpenMP >= 50 &&
+  if (getLangOpts().OpenMP >= 50 &&
       (ND->isUsed(/*CheckUsedAttr=*/false) || ND->isReferenced()))
     Diag(Loc, diag::warn_omp_declare_target_after_first_use);
 
@@ -23484,14 +23586,14 @@ void Sema::ActOnOpenMPDeclareTargetName(NamedDecl *ND, SourceLocation Loc,
       IsIndirect = true;
   }
   auto *A = OMPDeclareTargetDeclAttr::CreateImplicit(
-      Context, MT, DTCI.DT, IndirectE, IsIndirect, Level,
+      getASTContext(), MT, DTCI.DT, IndirectE, IsIndirect, Level,
       SourceRange(Loc, Loc));
   ND->addAttr(A);
-  if (ASTMutationListener *ML = Context.getASTMutationListener())
+  if (ASTMutationListener *ML = getASTContext().getASTMutationListener())
     ML->DeclarationMarkedOpenMPDeclareTarget(ND, A);
   checkDeclIsAllowedInOpenMPTarget(nullptr, ND, Loc);
   if (auto *VD = dyn_cast<VarDecl>(ND);
-      LangOpts.OpenMP && VD && VD->hasAttr<OMPDeclareTargetDeclAttr>() &&
+      getLangOpts().OpenMP && VD && VD->hasAttr<OMPDeclareTargetDeclAttr>() &&
       VD->hasGlobalStorage())
     ActOnOpenMPDeclareTargetInitializer(ND);
 }
@@ -23535,8 +23637,8 @@ static bool checkValueDeclInTarget(SourceLocation SL, SourceRange SR,
                            /*FullCheck=*/false);
 }
 
-void Sema::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
-                                            SourceLocation IdLoc) {
+void SemaOpenMP::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
+                                                  SourceLocation IdLoc) {
   if (!D || D->isInvalidDecl())
     return;
   SourceRange SR = E ? E->getSourceRange() : D->getSourceRange();
@@ -23550,7 +23652,7 @@ void Sema::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
     // directive.
     if (DSAStack->isThreadPrivate(VD)) {
       Diag(SL, diag::err_omp_threadprivate_in_target);
-      reportOriginalDsa(*this, DSAStack, VD, DSAStack->getTopDSA(VD, false));
+      reportOriginalDsa(SemaRef, DSAStack, VD, DSAStack->getTopDSA(VD, false));
       return;
     }
   }
@@ -23569,7 +23671,7 @@ void Sema::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
     // Problem if any with var declared with incomplete type will be reported
     // as normal, so no need to check it here.
     if ((E || !VD->getType()->isIncompleteType()) &&
-        !checkValueDeclInTarget(SL, SR, *this, DSAStack, VD))
+        !checkValueDeclInTarget(SL, SR, SemaRef, DSAStack, VD))
       return;
     if (!E && isInOpenMPDeclareTargetContext()) {
       // Checking declaration inside declare target region.
@@ -23589,13 +23691,13 @@ void Sema::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
             IsIndirect = true;
         }
         auto *A = OMPDeclareTargetDeclAttr::CreateImplicit(
-            Context,
+            getASTContext(),
             getLangOpts().OpenMP >= 52 ? OMPDeclareTargetDeclAttr::MT_Enter
                                        : OMPDeclareTargetDeclAttr::MT_To,
             DTCI.DT, IndirectE, IsIndirect, Level,
             SourceRange(DTCI.Loc, DTCI.Loc));
         D->addAttr(A);
-        if (ASTMutationListener *ML = Context.getASTMutationListener())
+        if (ASTMutationListener *ML = getASTContext().getASTMutationListener())
           ML->DeclarationMarkedOpenMPDeclareTarget(D, A);
       }
       return;
@@ -23603,7 +23705,7 @@ void Sema::checkDeclIsAllowedInOpenMPTarget(Expr *E, Decl *D,
   }
   if (!E)
     return;
-  checkDeclInTargetContext(E->getExprLoc(), E->getSourceRange(), *this, D);
+  checkDeclInTargetContext(E->getExprLoc(), E->getSourceRange(), SemaRef, D);
 }
 
 /// This class visits every VarDecl that the initializer references and adds
@@ -23649,13 +23751,13 @@ class GlobalDeclRefChecker final
 /// Adding OMPDeclareTargetDeclAttr to variables with static storage
 /// duration that are referenced in the initializer expression list of
 /// variables with static storage duration in declare target directive.
-void Sema::ActOnOpenMPDeclareTargetInitializer(Decl *TargetDecl) {
+void SemaOpenMP::ActOnOpenMPDeclareTargetInitializer(Decl *TargetDecl) {
   GlobalDeclRefChecker Checker;
   if (isa<VarDecl>(TargetDecl))
     Checker.declareTargetInitializer(TargetDecl);
 }
 
-OMPClause *Sema::ActOnOpenMPToClause(
+OMPClause *SemaOpenMP::ActOnOpenMPToClause(
     ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
     ArrayRef<SourceLocation> MotionModifiersLoc,
     CXXScopeSpec &MapperIdScopeSpec, DeclarationNameInfo &MapperId,
@@ -23681,18 +23783,18 @@ OMPClause *Sema::ActOnOpenMPToClause(
   }
 
   MappableVarListInfo MVLI(VarList);
-  checkMappableExpressionList(*this, DSAStack, OMPC_to, MVLI, Locs.StartLoc,
+  checkMappableExpressionList(SemaRef, DSAStack, OMPC_to, MVLI, Locs.StartLoc,
                               MapperIdScopeSpec, MapperId, UnresolvedMappers);
   if (MVLI.ProcessedVarList.empty())
     return nullptr;
 
   return OMPToClause::Create(
-      Context, Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
       MVLI.VarComponents, MVLI.UDMapperList, Modifiers, ModifiersLoc,
-      MapperIdScopeSpec.getWithLocInContext(Context), MapperId);
+      MapperIdScopeSpec.getWithLocInContext(getASTContext()), MapperId);
 }
 
-OMPClause *Sema::ActOnOpenMPFromClause(
+OMPClause *SemaOpenMP::ActOnOpenMPFromClause(
     ArrayRef<OpenMPMotionModifierKind> MotionModifiers,
     ArrayRef<SourceLocation> MotionModifiersLoc,
     CXXScopeSpec &MapperIdScopeSpec, DeclarationNameInfo &MapperId,
@@ -23718,19 +23820,20 @@ OMPClause *Sema::ActOnOpenMPFromClause(
   }
 
   MappableVarListInfo MVLI(VarList);
-  checkMappableExpressionList(*this, DSAStack, OMPC_from, MVLI, Locs.StartLoc,
+  checkMappableExpressionList(SemaRef, DSAStack, OMPC_from, MVLI, Locs.StartLoc,
                               MapperIdScopeSpec, MapperId, UnresolvedMappers);
   if (MVLI.ProcessedVarList.empty())
     return nullptr;
 
   return OMPFromClause::Create(
-      Context, Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
       MVLI.VarComponents, MVLI.UDMapperList, Modifiers, ModifiersLoc,
-      MapperIdScopeSpec.getWithLocInContext(Context), MapperId);
+      MapperIdScopeSpec.getWithLocInContext(getASTContext()), MapperId);
 }
 
-OMPClause *Sema::ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
-                                               const OMPVarListLocTy &Locs) {
+OMPClause *
+SemaOpenMP::ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
+                                          const OMPVarListLocTy &Locs) {
   MappableVarListInfo MVLI(VarList);
   SmallVector<Expr *, 8> PrivateCopies;
   SmallVector<Expr *, 8> Inits;
@@ -23740,7 +23843,7 @@ OMPClause *Sema::ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       MVLI.ProcessedVarList.push_back(RefExpr);
@@ -23765,30 +23868,30 @@ OMPClause *Sema::ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
 
     // Build the private variable and the expression that refers to it.
     auto VDPrivate =
-        buildVarDecl(*this, ELoc, Type, D->getName(),
+        buildVarDecl(SemaRef, ELoc, Type, D->getName(),
                      D->hasAttrs() ? &D->getAttrs() : nullptr,
                      VD ? cast<DeclRefExpr>(SimpleRefExpr) : nullptr);
     if (VDPrivate->isInvalidDecl())
       continue;
 
-    CurContext->addDecl(VDPrivate);
+    SemaRef.CurContext->addDecl(VDPrivate);
     DeclRefExpr *VDPrivateRefExpr = buildDeclRefExpr(
-        *this, VDPrivate, RefExpr->getType().getUnqualifiedType(), ELoc);
+        SemaRef, VDPrivate, RefExpr->getType().getUnqualifiedType(), ELoc);
 
     // Add temporary variable to initialize the private copy of the pointer.
     VarDecl *VDInit =
-        buildVarDecl(*this, RefExpr->getExprLoc(), Type, ".devptr.temp");
+        buildVarDecl(SemaRef, RefExpr->getExprLoc(), Type, ".devptr.temp");
     DeclRefExpr *VDInitRefExpr = buildDeclRefExpr(
-        *this, VDInit, RefExpr->getType(), RefExpr->getExprLoc());
-    AddInitializerToDecl(VDPrivate,
-                         DefaultLvalueConversion(VDInitRefExpr).get(),
-                         /*DirectInit=*/false);
+        SemaRef, VDInit, RefExpr->getType(), RefExpr->getExprLoc());
+    SemaRef.AddInitializerToDecl(
+        VDPrivate, SemaRef.DefaultLvalueConversion(VDInitRefExpr).get(),
+        /*DirectInit=*/false);
 
     // If required, build a capture to implement the privatization initialized
     // with the current list item value.
     DeclRefExpr *Ref = nullptr;
     if (!VD)
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
     MVLI.ProcessedVarList.push_back(VD ? RefExpr->IgnoreParens() : Ref);
     PrivateCopies.push_back(VDPrivateRefExpr);
     Inits.push_back(VDInitRefExpr);
@@ -23810,12 +23913,13 @@ OMPClause *Sema::ActOnOpenMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
     return nullptr;
 
   return OMPUseDevicePtrClause::Create(
-      Context, Locs, MVLI.ProcessedVarList, PrivateCopies, Inits,
+      getASTContext(), Locs, MVLI.ProcessedVarList, PrivateCopies, Inits,
       MVLI.VarBaseDeclarations, MVLI.VarComponents);
 }
 
-OMPClause *Sema::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
-                                                const OMPVarListLocTy &Locs) {
+OMPClause *
+SemaOpenMP::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
+                                           const OMPVarListLocTy &Locs) {
   MappableVarListInfo MVLI(VarList);
 
   for (Expr *RefExpr : VarList) {
@@ -23823,7 +23927,7 @@ OMPClause *Sema::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange,
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange,
                               /*AllowArraySection=*/true);
     if (Res.second) {
       // It will be analyzed later.
@@ -23838,7 +23942,7 @@ OMPClause *Sema::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
     // with the current list item value.
     DeclRefExpr *Ref = nullptr;
     if (!VD)
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
     MVLI.ProcessedVarList.push_back(VD ? RefExpr->IgnoreParens() : Ref);
 
     // We need to add a data sharing attribute for this variable to make sure it
@@ -23853,7 +23957,8 @@ OMPClause *Sema::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
     Expr *Component = SimpleRefExpr;
     if (VD && (isa<OMPArraySectionExpr>(RefExpr->IgnoreParenImpCasts()) ||
                isa<ArraySubscriptExpr>(RefExpr->IgnoreParenImpCasts())))
-      Component = DefaultFunctionArrayLvalueConversion(SimpleRefExpr).get();
+      Component =
+          SemaRef.DefaultFunctionArrayLvalueConversion(SimpleRefExpr).get();
     MVLI.VarComponents.back().emplace_back(Component, D,
                                            /*IsNonContiguous=*/false);
   }
@@ -23861,20 +23966,21 @@ OMPClause *Sema::ActOnOpenMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
   if (MVLI.ProcessedVarList.empty())
     return nullptr;
 
-  return OMPUseDeviceAddrClause::Create(Context, Locs, MVLI.ProcessedVarList,
-                                        MVLI.VarBaseDeclarations,
-                                        MVLI.VarComponents);
+  return OMPUseDeviceAddrClause::Create(
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      MVLI.VarComponents);
 }
 
-OMPClause *Sema::ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
-                                              const OMPVarListLocTy &Locs) {
+OMPClause *
+SemaOpenMP::ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
+                                         const OMPVarListLocTy &Locs) {
   MappableVarListInfo MVLI(VarList);
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP is_device_ptr clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       MVLI.ProcessedVarList.push_back(RefExpr);
@@ -23900,7 +24006,7 @@ OMPClause *Sema::ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
           << getOpenMPClauseName(DVar.CKind)
           << getOpenMPClauseName(OMPC_is_device_ptr)
           << getOpenMPDirectiveName(DSAStack->getCurrentDirective());
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
@@ -23944,20 +24050,21 @@ OMPClause *Sema::ActOnOpenMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
   if (MVLI.ProcessedVarList.empty())
     return nullptr;
 
-  return OMPIsDevicePtrClause::Create(Context, Locs, MVLI.ProcessedVarList,
-                                      MVLI.VarBaseDeclarations,
-                                      MVLI.VarComponents);
+  return OMPIsDevicePtrClause::Create(
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      MVLI.VarComponents);
 }
 
-OMPClause *Sema::ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
-                                                const OMPVarListLocTy &Locs) {
+OMPClause *
+SemaOpenMP::ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
+                                           const OMPVarListLocTy &Locs) {
   MappableVarListInfo MVLI(VarList);
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP has_device_addr clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange,
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange,
                               /*AllowArraySection=*/true);
     if (Res.second) {
       // It will be analyzed later.
@@ -23975,7 +24082,7 @@ OMPClause *Sema::ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
           << getOpenMPClauseName(DVar.CKind)
           << getOpenMPClauseName(OMPC_has_device_addr)
           << getOpenMPDirectiveName(DSAStack->getCurrentDirective());
-      reportOriginalDsa(*this, DSAStack, D, DVar);
+      reportOriginalDsa(SemaRef, DSAStack, D, DVar);
       continue;
     }
 
@@ -24000,16 +24107,17 @@ OMPClause *Sema::ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
     auto *VD = dyn_cast<VarDecl>(D);
     if (VD && (isa<OMPArraySectionExpr>(RefExpr->IgnoreParenImpCasts()) ||
                isa<ArraySubscriptExpr>(RefExpr->IgnoreParenImpCasts())))
-      Component = DefaultFunctionArrayLvalueConversion(SimpleRefExpr).get();
+      Component =
+          SemaRef.DefaultFunctionArrayLvalueConversion(SimpleRefExpr).get();
     OMPClauseMappableExprCommon::MappableComponent MC(
         Component, D, /*IsNonContiguous=*/false);
     DSAStack->addMappableExpressionComponents(
         D, MC, /*WhereFoundClauseKind=*/OMPC_has_device_addr);
 
     // Record the expression we've just processed.
-    if (!VD && !CurContext->isDependentContext()) {
+    if (!VD && !SemaRef.CurContext->isDependentContext()) {
       DeclRefExpr *Ref =
-          buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/true);
+          buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/true);
       assert(Ref && "has_device_addr capture failed");
       MVLI.ProcessedVarList.push_back(Ref);
     } else
@@ -24030,27 +24138,27 @@ OMPClause *Sema::ActOnOpenMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
   if (MVLI.ProcessedVarList.empty())
     return nullptr;
 
-  return OMPHasDeviceAddrClause::Create(Context, Locs, MVLI.ProcessedVarList,
-                                        MVLI.VarBaseDeclarations,
-                                        MVLI.VarComponents);
+  return OMPHasDeviceAddrClause::Create(
+      getASTContext(), Locs, MVLI.ProcessedVarList, MVLI.VarBaseDeclarations,
+      MVLI.VarComponents);
 }
 
-OMPClause *Sema::ActOnOpenMPAllocateClause(
+OMPClause *SemaOpenMP::ActOnOpenMPAllocateClause(
     Expr *Allocator, ArrayRef<Expr *> VarList, SourceLocation StartLoc,
     SourceLocation ColonLoc, SourceLocation LParenLoc, SourceLocation EndLoc) {
   if (Allocator) {
     // OpenMP [2.11.4 allocate Clause, Description]
     // allocator is an expression of omp_allocator_handle_t type.
-    if (!findOMPAllocatorHandleT(*this, Allocator->getExprLoc(), DSAStack))
+    if (!findOMPAllocatorHandleT(SemaRef, Allocator->getExprLoc(), DSAStack))
       return nullptr;
 
-    ExprResult AllocatorRes = DefaultLvalueConversion(Allocator);
+    ExprResult AllocatorRes = SemaRef.DefaultLvalueConversion(Allocator);
     if (AllocatorRes.isInvalid())
       return nullptr;
-    AllocatorRes = PerformImplicitConversion(AllocatorRes.get(),
-                                             DSAStack->getOMPAllocatorHandleT(),
-                                             Sema::AA_Initializing,
-                                             /*AllowExplicit=*/true);
+    AllocatorRes = SemaRef.PerformImplicitConversion(
+        AllocatorRes.get(), DSAStack->getOMPAllocatorHandleT(),
+        Sema::AA_Initializing,
+        /*AllowExplicit=*/true);
     if (AllocatorRes.isInvalid())
       return nullptr;
     Allocator = AllocatorRes.get();
@@ -24060,9 +24168,9 @@ OMPClause *Sema::ActOnOpenMPAllocateClause(
     // target region must specify an allocator expression unless a requires
     // directive with the dynamic_allocators clause is present in the same
     // compilation unit.
-    if (LangOpts.OpenMPIsTargetDevice &&
+    if (getLangOpts().OpenMPIsTargetDevice &&
         !DSAStack->hasRequiresDeclWithClause<OMPDynamicAllocatorsClause>())
-      targetDiag(StartLoc, diag::err_expected_allocator_expression);
+      SemaRef.targetDiag(StartLoc, diag::err_expected_allocator_expression);
   }
   // Analyze and build list of variables.
   SmallVector<Expr *, 8> Vars;
@@ -24071,7 +24179,7 @@ OMPClause *Sema::ActOnOpenMPAllocateClause(
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second) {
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -24082,9 +24190,9 @@ OMPClause *Sema::ActOnOpenMPAllocateClause(
 
     auto *VD = dyn_cast<VarDecl>(D);
     DeclRefExpr *Ref = nullptr;
-    if (!VD && !CurContext->isDependentContext())
-      Ref = buildCapture(*this, D, SimpleRefExpr, /*WithInit=*/false);
-    Vars.push_back((VD || CurContext->isDependentContext())
+    if (!VD && !SemaRef.CurContext->isDependentContext())
+      Ref = buildCapture(SemaRef, D, SimpleRefExpr, /*WithInit=*/false);
+    Vars.push_back((VD || SemaRef.CurContext->isDependentContext())
                        ? RefExpr->IgnoreParens()
                        : Ref);
   }
@@ -24094,21 +24202,21 @@ OMPClause *Sema::ActOnOpenMPAllocateClause(
 
   if (Allocator)
     DSAStack->addInnerAllocatorExpr(Allocator);
-  return OMPAllocateClause::Create(Context, StartLoc, LParenLoc, Allocator,
-                                   ColonLoc, EndLoc, Vars);
+  return OMPAllocateClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                   Allocator, ColonLoc, EndLoc, Vars);
 }
 
-OMPClause *Sema::ActOnOpenMPNontemporalClause(ArrayRef<Expr *> VarList,
-                                              SourceLocation StartLoc,
-                                              SourceLocation LParenLoc,
-                                              SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPNontemporalClause(ArrayRef<Expr *> VarList,
+                                                    SourceLocation StartLoc,
+                                                    SourceLocation LParenLoc,
+                                                    SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP nontemporal clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange);
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange);
     if (Res.second)
       // It will be analyzed later.
       Vars.push_back(RefExpr);
@@ -24133,32 +24241,34 @@ OMPClause *Sema::ActOnOpenMPNontemporalClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPNontemporalClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                      Vars);
+  return OMPNontemporalClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                      EndLoc, Vars);
 }
 
-StmtResult Sema::ActOnOpenMPScopeDirective(ArrayRef<OMPClause *> Clauses,
-                                           Stmt *AStmt, SourceLocation StartLoc,
-                                           SourceLocation EndLoc) {
+StmtResult SemaOpenMP::ActOnOpenMPScopeDirective(ArrayRef<OMPClause *> Clauses,
+                                                 Stmt *AStmt,
+                                                 SourceLocation StartLoc,
+                                                 SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
-  setFunctionHasBranchProtectedScope();
+  SemaRef.setFunctionHasBranchProtectedScope();
 
-  return OMPScopeDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt);
+  return OMPScopeDirective::Create(getASTContext(), StartLoc, EndLoc, Clauses,
+                                   AStmt);
 }
 
-OMPClause *Sema::ActOnOpenMPInclusiveClause(ArrayRef<Expr *> VarList,
-                                            SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPInclusiveClause(ArrayRef<Expr *> VarList,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation LParenLoc,
+                                                  SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP nontemporal clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange,
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange,
                               /*AllowArraySection=*/true);
     if (Res.second)
       // It will be analyzed later.
@@ -24185,20 +24295,21 @@ OMPClause *Sema::ActOnOpenMPInclusiveClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPInclusiveClause::Create(Context, StartLoc, LParenLoc, EndLoc, Vars);
+  return OMPInclusiveClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                    EndLoc, Vars);
 }
 
-OMPClause *Sema::ActOnOpenMPExclusiveClause(ArrayRef<Expr *> VarList,
-                                            SourceLocation StartLoc,
-                                            SourceLocation LParenLoc,
-                                            SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPExclusiveClause(ArrayRef<Expr *> VarList,
+                                                  SourceLocation StartLoc,
+                                                  SourceLocation LParenLoc,
+                                                  SourceLocation EndLoc) {
   SmallVector<Expr *, 8> Vars;
   for (Expr *RefExpr : VarList) {
     assert(RefExpr && "NULL expr in OpenMP nontemporal clause.");
     SourceLocation ELoc;
     SourceRange ERange;
     Expr *SimpleRefExpr = RefExpr;
-    auto Res = getPrivateItem(*this, SimpleRefExpr, ELoc, ERange,
+    auto Res = getPrivateItem(SemaRef, SimpleRefExpr, ELoc, ERange,
                               /*AllowArraySection=*/true);
     if (Res.second)
       // It will be analyzed later.
@@ -24228,7 +24339,8 @@ OMPClause *Sema::ActOnOpenMPExclusiveClause(ArrayRef<Expr *> VarList,
   if (Vars.empty())
     return nullptr;
 
-  return OMPExclusiveClause::Create(Context, StartLoc, LParenLoc, EndLoc, Vars);
+  return OMPExclusiveClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                    EndLoc, Vars);
 }
 
 /// Tries to find omp_alloctrait_t type.
@@ -24246,19 +24358,20 @@ static bool findOMPAlloctraitT(Sema &S, SourceLocation Loc, DSAStackTy *Stack) {
   return true;
 }
 
-OMPClause *Sema::ActOnOpenMPUsesAllocatorClause(
+OMPClause *SemaOpenMP::ActOnOpenMPUsesAllocatorClause(
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc,
     ArrayRef<UsesAllocatorsData> Data) {
+  ASTContext &Context = getASTContext();
   // OpenMP [2.12.5, target Construct]
   // allocator is an identifier of omp_allocator_handle_t type.
-  if (!findOMPAllocatorHandleT(*this, StartLoc, DSAStack))
+  if (!findOMPAllocatorHandleT(SemaRef, StartLoc, DSAStack))
     return nullptr;
   // OpenMP [2.12.5, target Construct]
   // allocator-traits-array is an identifier of const omp_alloctrait_t * type.
   if (llvm::any_of(
           Data,
           [](const UsesAllocatorsData &D) { return D.AllocatorTraits; }) &&
-      !findOMPAlloctraitT(*this, StartLoc, DSAStack))
+      !findOMPAlloctraitT(SemaRef, StartLoc, DSAStack))
     return nullptr;
   llvm::SmallPtrSet<CanonicalDeclPtr<Decl>, 4> PredefinedAllocators;
   for (int I = 0; I < OMPAllocateDeclAttr::OMPUserDefinedMemAlloc; ++I) {
@@ -24266,8 +24379,8 @@ OMPClause *Sema::ActOnOpenMPUsesAllocatorClause(
     StringRef Allocator =
         OMPAllocateDeclAttr::ConvertAllocatorTypeTyToStr(AllocatorKind);
     DeclarationName AllocatorName = &Context.Idents.get(Allocator);
-    PredefinedAllocators.insert(LookupSingleName(
-        TUScope, AllocatorName, StartLoc, Sema::LookupAnyName));
+    PredefinedAllocators.insert(SemaRef.LookupSingleName(
+        SemaRef.TUScope, AllocatorName, StartLoc, Sema::LookupAnyName));
   }
 
   SmallVector<OMPUsesAllocatorsClause::Data, 4> NewData;
@@ -24284,7 +24397,7 @@ OMPClause *Sema::ActOnOpenMPUsesAllocatorClause(
       bool IsPredefinedAllocator = false;
       if (DRE) {
         OMPAllocateDeclAttr::AllocatorTypeTy AllocatorTy =
-            getAllocatorKind(*this, DSAStack, AllocatorExpr);
+            getAllocatorKind(SemaRef, DSAStack, AllocatorExpr);
         IsPredefinedAllocator =
             AllocatorTy !=
             OMPAllocateDeclAttr::AllocatorTypeTy::OMPUserDefinedMemAlloc;
@@ -24329,7 +24442,7 @@ OMPClause *Sema::ActOnOpenMPUsesAllocatorClause(
       }
       // No allocator traits - just convert it to rvalue.
       if (!D.AllocatorTraits)
-        AllocatorExpr = DefaultLvalueConversion(AllocatorExpr).get();
+        AllocatorExpr = SemaRef.DefaultLvalueConversion(AllocatorExpr).get();
       DSAStack->addUsesAllocatorsDecl(
           DRE->getDecl(),
           IsPredefinedAllocator
@@ -24376,11 +24489,11 @@ OMPClause *Sema::ActOnOpenMPUsesAllocatorClause(
     NewD.LParenLoc = D.LParenLoc;
     NewD.RParenLoc = D.RParenLoc;
   }
-  return OMPUsesAllocatorsClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                         NewData);
+  return OMPUsesAllocatorsClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                         EndLoc, NewData);
 }
 
-OMPClause *Sema::ActOnOpenMPAffinityClause(
+OMPClause *SemaOpenMP::ActOnOpenMPAffinityClause(
     SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation ColonLoc,
     SourceLocation EndLoc, Expr *Modifier, ArrayRef<Expr *> Locators) {
   SmallVector<Expr *, 8> Vars;
@@ -24403,8 +24516,8 @@ OMPClause *Sema::ActOnOpenMPAffinityClause(
 
     ExprResult Res;
     {
-      Sema::TentativeAnalysisScope Trap(*this);
-      Res = CreateBuiltinUnaryOp(ELoc, UO_AddrOf, SimpleExpr);
+      Sema::TentativeAnalysisScope Trap(SemaRef);
+      Res = SemaRef.CreateBuiltinUnaryOp(ELoc, UO_AddrOf, SimpleExpr);
     }
     if (!Res.isUsable() && !isa<OMPArraySectionExpr>(SimpleExpr) &&
         !isa<OMPArrayShapingExpr>(SimpleExpr)) {
@@ -24415,15 +24528,15 @@ OMPClause *Sema::ActOnOpenMPAffinityClause(
     Vars.push_back(SimpleExpr);
   }
 
-  return OMPAffinityClause::Create(Context, StartLoc, LParenLoc, ColonLoc,
-                                   EndLoc, Modifier, Vars);
+  return OMPAffinityClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                   ColonLoc, EndLoc, Modifier, Vars);
 }
 
-OMPClause *Sema::ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
-                                       SourceLocation KindLoc,
-                                       SourceLocation StartLoc,
-                                       SourceLocation LParenLoc,
-                                       SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
+                                             SourceLocation KindLoc,
+                                             SourceLocation StartLoc,
+                                             SourceLocation LParenLoc,
+                                             SourceLocation EndLoc) {
   if (Kind == OMPC_BIND_unknown) {
     Diag(KindLoc, diag::err_omp_unexpected_clause_value)
         << getListOfPossibleValues(OMPC_bind, /*First=*/0,
@@ -24432,39 +24545,40 @@ OMPClause *Sema::ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
     return nullptr;
   }
 
-  return OMPBindClause::Create(Context, Kind, KindLoc, StartLoc, LParenLoc,
-                               EndLoc);
+  return OMPBindClause::Create(getASTContext(), Kind, KindLoc, StartLoc,
+                               LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPXDynCGroupMemClause(Expr *Size,
-                                                SourceLocation StartLoc,
-                                                SourceLocation LParenLoc,
-                                                SourceLocation EndLoc) {
+OMPClause *SemaOpenMP::ActOnOpenMPXDynCGroupMemClause(Expr *Size,
+                                                      SourceLocation StartLoc,
+                                                      SourceLocation LParenLoc,
+                                                      SourceLocation EndLoc) {
   Expr *ValExpr = Size;
   Stmt *HelperValStmt = nullptr;
 
   // OpenMP [2.5, Restrictions]
   //  The ompx_dyn_cgroup_mem expression must evaluate to a positive integer
   //  value.
-  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_ompx_dyn_cgroup_mem,
+  if (!isNonNegativeIntegerValue(ValExpr, SemaRef, OMPC_ompx_dyn_cgroup_mem,
                                  /*StrictlyPositive=*/false))
     return nullptr;
 
   OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
   OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause(
-      DKind, OMPC_ompx_dyn_cgroup_mem, LangOpts.OpenMP);
-  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
-    ValExpr = MakeFullExpr(ValExpr).get();
+      DKind, OMPC_ompx_dyn_cgroup_mem, getLangOpts().OpenMP);
+  if (CaptureRegion != OMPD_unknown &&
+      !SemaRef.CurContext->isDependentContext()) {
+    ValExpr = SemaRef.MakeFullExpr(ValExpr).get();
     llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
-    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
-    HelperValStmt = buildPreInits(Context, Captures);
+    ValExpr = tryBuildCapture(SemaRef, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(getASTContext(), Captures);
   }
 
-  return new (Context) OMPXDynCGroupMemClause(
+  return new (getASTContext()) OMPXDynCGroupMemClause(
       ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPDoacrossClause(
+OMPClause *SemaOpenMP::ActOnOpenMPDoacrossClause(
     OpenMPDoacrossClauseModifier DepType, SourceLocation DepLoc,
     SourceLocation ColonLoc, ArrayRef<Expr *> VarList, SourceLocation StartLoc,
     SourceLocation LParenLoc, SourceLocation EndLoc) {
@@ -24483,7 +24597,7 @@ OMPClause *Sema::ActOnOpenMPDoacrossClause(
   DSAStackTy::OperatorOffsetTy OpsOffs;
   llvm::APSInt TotalDepCount(/*BitWidth=*/32);
   DoacrossDataInfoTy VarOffset = ProcessOpenMPDoacrossClauseCommon(
-      *this,
+      SemaRef,
       DepType == OMPC_DOACROSS_source ||
           DepType == OMPC_DOACROSS_source_omp_cur_iteration ||
           DepType == OMPC_DOACROSS_sink_omp_cur_iteration,
@@ -24491,22 +24605,587 @@ OMPClause *Sema::ActOnOpenMPDoacrossClause(
   Vars = VarOffset.Vars;
   OpsOffs = VarOffset.OpsOffs;
   TotalDepCount = VarOffset.TotalDepCount;
-  auto *C = OMPDoacrossClause::Create(Context, StartLoc, LParenLoc, EndLoc,
-                                      DepType, DepLoc, ColonLoc, Vars,
+  auto *C = OMPDoacrossClause::Create(getASTContext(), StartLoc, LParenLoc,
+                                      EndLoc, DepType, DepLoc, ColonLoc, Vars,
                                       TotalDepCount.getZExtValue());
   if (DSAStack->isParentOrderedRegion())
     DSAStack->addDoacrossDependClause(C, OpsOffs);
   return C;
 }
 
-OMPClause *Sema::ActOnOpenMPXAttributeClause(ArrayRef<const Attr *> Attrs,
-                                             SourceLocation StartLoc,
-                                             SourceLocation LParenLoc,
-                                             SourceLocation EndLoc) {
-  return new (Context) OMPXAttributeClause(Attrs, StartLoc, LParenLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPXAttributeClause(ArrayRef<const Attr *> Attrs,
+                                                   SourceLocation StartLoc,
+                                                   SourceLocation LParenLoc,
+                                                   SourceLocation EndLoc) {
+  return new (getASTContext())
+      OMPXAttributeClause(Attrs, StartLoc, LParenLoc, EndLoc);
 }
 
-OMPClause *Sema::ActOnOpenMPXBareClause(SourceLocation StartLoc,
-                                        SourceLocation EndLoc) {
-  return new (Context) OMPXBareClause(StartLoc, EndLoc);
+OMPClause *SemaOpenMP::ActOnOpenMPXBareClause(SourceLocation StartLoc,
+                                              SourceLocation EndLoc) {
+  return new (getASTContext()) OMPXBareClause(StartLoc, EndLoc);
+}
+
+ExprResult SemaOpenMP::ActOnOMPArraySectionExpr(
+    Expr *Base, SourceLocation LBLoc, Expr *LowerBound,
+    SourceLocation ColonLocFirst, SourceLocation ColonLocSecond, Expr *Length,
+    Expr *Stride, SourceLocation RBLoc) {
+  ASTContext &Context = getASTContext();
+  if (Base->hasPlaceholderType() &&
+      !Base->hasPlaceholderType(BuiltinType::OMPArraySection)) {
+    ExprResult Result = SemaRef.CheckPlaceholderExpr(Base);
+    if (Result.isInvalid())
+      return ExprError();
+    Base = Result.get();
+  }
+  if (LowerBound && LowerBound->getType()->isNonOverloadPlaceholderType()) {
+    ExprResult Result = SemaRef.CheckPlaceholderExpr(LowerBound);
+    if (Result.isInvalid())
+      return ExprError();
+    Result = SemaRef.DefaultLvalueConversion(Result.get());
+    if (Result.isInvalid())
+      return ExprError();
+    LowerBound = Result.get();
+  }
+  if (Length && Length->getType()->isNonOverloadPlaceholderType()) {
+    ExprResult Result = SemaRef.CheckPlaceholderExpr(Length);
+    if (Result.isInvalid())
+      return ExprError();
+    Result = SemaRef.DefaultLvalueConversion(Result.get());
+    if (Result.isInvalid())
+      return ExprError();
+    Length = Result.get();
+  }
+  if (Stride && Stride->getType()->isNonOverloadPlaceholderType()) {
+    ExprResult Result = SemaRef.CheckPlaceholderExpr(Stride);
+    if (Result.isInvalid())
+      return ExprError();
+    Result = SemaRef.DefaultLvalueConversion(Result.get());
+    if (Result.isInvalid())
+      return ExprError();
+    Stride = Result.get();
+  }
+
+  // Build an unanalyzed expression if either operand is type-dependent.
+  if (Base->isTypeDependent() ||
+      (LowerBound &&
+       (LowerBound->isTypeDependent() || LowerBound->isValueDependent())) ||
+      (Length && (Length->isTypeDependent() || Length->isValueDependent())) ||
+      (Stride && (Stride->isTypeDependent() || Stride->isValueDependent()))) {
+    return new (Context) OMPArraySectionExpr(
+        Base, LowerBound, Length, Stride, Context.DependentTy, VK_LValue,
+        OK_Ordinary, ColonLocFirst, ColonLocSecond, RBLoc);
+  }
+
+  // Perform default conversions.
+  QualType OriginalTy = OMPArraySectionExpr::getBaseOriginalType(Base);
+  QualType ResultTy;
+  if (OriginalTy->isAnyPointerType()) {
+    ResultTy = OriginalTy->getPointeeType();
+  } else if (OriginalTy->isArrayType()) {
+    ResultTy = OriginalTy->getAsArrayTypeUnsafe()->getElementType();
+  } else {
+    return ExprError(
+        Diag(Base->getExprLoc(), diag::err_omp_typecheck_section_value)
+        << Base->getSourceRange());
+  }
+  // C99 6.5.2.1p1
+  if (LowerBound) {
+    auto Res = PerformOpenMPImplicitIntegerConversion(LowerBound->getExprLoc(),
+                                                      LowerBound);
+    if (Res.isInvalid())
+      return ExprError(Diag(LowerBound->getExprLoc(),
+                            diag::err_omp_typecheck_section_not_integer)
+                       << 0 << LowerBound->getSourceRange());
+    LowerBound = Res.get();
+
+    if (LowerBound->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
+        LowerBound->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
+      Diag(LowerBound->getExprLoc(), diag::warn_omp_section_is_char)
+          << 0 << LowerBound->getSourceRange();
+  }
+  if (Length) {
+    auto Res =
+        PerformOpenMPImplicitIntegerConversion(Length->getExprLoc(), Length);
+    if (Res.isInvalid())
+      return ExprError(Diag(Length->getExprLoc(),
+                            diag::err_omp_typecheck_section_not_integer)
+                       << 1 << Length->getSourceRange());
+    Length = Res.get();
+
+    if (Length->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
+        Length->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
+      Diag(Length->getExprLoc(), diag::warn_omp_section_is_char)
+          << 1 << Length->getSourceRange();
+  }
+  if (Stride) {
+    ExprResult Res =
+        PerformOpenMPImplicitIntegerConversion(Stride->getExprLoc(), Stride);
+    if (Res.isInvalid())
+      return ExprError(Diag(Stride->getExprLoc(),
+                            diag::err_omp_typecheck_section_not_integer)
+                       << 1 << Stride->getSourceRange());
+    Stride = Res.get();
+
+    if (Stride->getType()->isSpecificBuiltinType(BuiltinType::Char_S) ||
+        Stride->getType()->isSpecificBuiltinType(BuiltinType::Char_U))
+      Diag(Stride->getExprLoc(), diag::warn_omp_section_is_char)
+          << 1 << Stride->getSourceRange();
+  }
+
+  // C99 6.5.2.1p1: "shall have type "pointer to *object* type". Similarly,
+  // C++ [expr.sub]p1: The type "T" shall be a completely-defined object
+  // type. Note that functions are not objects, and that (in C99 parlance)
+  // incomplete types are not object types.
+  if (ResultTy->isFunctionType()) {
+    Diag(Base->getExprLoc(), diag::err_omp_section_function_type)
+        << ResultTy << Base->getSourceRange();
+    return ExprError();
+  }
+
+  if (SemaRef.RequireCompleteType(Base->getExprLoc(), ResultTy,
+                                  diag::err_omp_section_incomplete_type, Base))
+    return ExprError();
+
+  if (LowerBound && !OriginalTy->isAnyPointerType()) {
+    Expr::EvalResult Result;
+    if (LowerBound->EvaluateAsInt(Result, Context)) {
+      // OpenMP 5.0, [2.1.5 Array Sections]
+      // The array section must be a subset of the original array.
+      llvm::APSInt LowerBoundValue = Result.Val.getInt();
+      if (LowerBoundValue.isNegative()) {
+        Diag(LowerBound->getExprLoc(),
+             diag::err_omp_section_not_subset_of_array)
+            << LowerBound->getSourceRange();
+        return ExprError();
+      }
+    }
+  }
+
+  if (Length) {
+    Expr::EvalResult Result;
+    if (Length->EvaluateAsInt(Result, Context)) {
+      // OpenMP 5.0, [2.1.5 Array Sections]
+      // The length must evaluate to non-negative integers.
+      llvm::APSInt LengthValue = Result.Val.getInt();
+      if (LengthValue.isNegative()) {
+        Diag(Length->getExprLoc(), diag::err_omp_section_length_negative)
+            << toString(LengthValue, /*Radix=*/10, /*Signed=*/true)
+            << Length->getSourceRange();
+        return ExprError();
+      }
+    }
+  } else if (ColonLocFirst.isValid() &&
+             (OriginalTy.isNull() || (!OriginalTy->isConstantArrayType() &&
+                                      !OriginalTy->isVariableArrayType()))) {
+    // OpenMP 5.0, [2.1.5 Array Sections]
+    // When the size of the array dimension is not known, the length must be
+    // specified explicitly.
+    Diag(ColonLocFirst, diag::err_omp_section_length_undefined)
+        << (!OriginalTy.isNull() && OriginalTy->isArrayType());
+    return ExprError();
+  }
+
+  if (Stride) {
+    Expr::EvalResult Result;
+    if (Stride->EvaluateAsInt(Result, Context)) {
+      // OpenMP 5.0, [2.1.5 Array Sections]
+      // The stride must evaluate to a positive integer.
+      llvm::APSInt StrideValue = Result.Val.getInt();
+      if (!StrideValue.isStrictlyPositive()) {
+        Diag(Stride->getExprLoc(), diag::err_omp_section_stride_non_positive)
+            << toString(StrideValue, /*Radix=*/10, /*Signed=*/true)
+            << Stride->getSourceRange();
+        return ExprError();
+      }
+    }
+  }
+
+  if (!Base->hasPlaceholderType(BuiltinType::OMPArraySection)) {
+    ExprResult Result = SemaRef.DefaultFunctionArrayLvalueConversion(Base);
+    if (Result.isInvalid())
+      return ExprError();
+    Base = Result.get();
+  }
+  return new (Context) OMPArraySectionExpr(
+      Base, LowerBound, Length, Stride, Context.OMPArraySectionTy, VK_LValue,
+      OK_Ordinary, ColonLocFirst, ColonLocSecond, RBLoc);
+}
+
+ExprResult SemaOpenMP::ActOnOMPArrayShapingExpr(
+    Expr *Base, SourceLocation LParenLoc, SourceLocation RParenLoc,
+    ArrayRef<Expr *> Dims, ArrayRef<SourceRange> Brackets) {
+  ASTContext &Context = getASTContext();
+  if (Base->hasPlaceholderType()) {
+    ExprResult Result = SemaRef.CheckPlaceholderExpr(Base);
+    if (Result.isInvalid())
+      return ExprError();
+    Result = SemaRef.DefaultLvalueConversion(Result.get());
+    if (Result.isInvalid())
+      return ExprError();
+    Base = Result.get();
+  }
+  QualType BaseTy = Base->getType();
+  // Delay analysis of the types/expressions if instantiation/specialization is
+  // required.
+  if (!BaseTy->isPointerType() && Base->isTypeDependent())
+    return OMPArrayShapingExpr::Create(Context, Context.DependentTy, Base,
+                                       LParenLoc, RParenLoc, Dims, Brackets);
+  if (!BaseTy->isPointerType() ||
+      (!Base->isTypeDependent() &&
+       BaseTy->getPointeeType()->isIncompleteType()))
+    return ExprError(Diag(Base->getExprLoc(),
+                          diag::err_omp_non_pointer_type_array_shaping_base)
+                     << Base->getSourceRange());
+
+  SmallVector<Expr *, 4> NewDims;
+  bool ErrorFound = false;
+  for (Expr *Dim : Dims) {
+    if (Dim->hasPlaceholderType()) {
+      ExprResult Result = SemaRef.CheckPlaceholderExpr(Dim);
+      if (Result.isInvalid()) {
+        ErrorFound = true;
+        continue;
+      }
+      Result = SemaRef.DefaultLvalueConversion(Result.get());
+      if (Result.isInvalid()) {
+        ErrorFound = true;
+        continue;
+      }
+      Dim = Result.get();
+    }
+    if (!Dim->isTypeDependent()) {
+      ExprResult Result =
+          PerformOpenMPImplicitIntegerConversion(Dim->getExprLoc(), Dim);
+      if (Result.isInvalid()) {
+        ErrorFound = true;
+        Diag(Dim->getExprLoc(), diag::err_omp_typecheck_shaping_not_integer)
+            << Dim->getSourceRange();
+        continue;
+      }
+      Dim = Result.get();
+      Expr::EvalResult EvResult;
+      if (!Dim->isValueDependent() && Dim->EvaluateAsInt(EvResult, Context)) {
+        // OpenMP 5.0, [2.1.4 Array Shaping]
+        // Each si is an integral type expression that must evaluate to a
+        // positive integer.
+        llvm::APSInt Value = EvResult.Val.getInt();
+        if (!Value.isStrictlyPositive()) {
+          Diag(Dim->getExprLoc(), diag::err_omp_shaping_dimension_not_positive)
+              << toString(Value, /*Radix=*/10, /*Signed=*/true)
+              << Dim->getSourceRange();
+          ErrorFound = true;
+          continue;
+        }
+      }
+    }
+    NewDims.push_back(Dim);
+  }
+  if (ErrorFound)
+    return ExprError();
+  return OMPArrayShapingExpr::Create(Context, Context.OMPArrayShapingTy, Base,
+                                     LParenLoc, RParenLoc, NewDims, Brackets);
 }
+
+ExprResult SemaOpenMP::ActOnOMPIteratorExpr(Scope *S,
+                                            SourceLocation IteratorKwLoc,
+                                            SourceLocation LLoc,
+                                            SourceLocation RLoc,
+                                            ArrayRef<OMPIteratorData> Data) {
+  ASTContext &Context = getASTContext();
+  SmallVector<OMPIteratorExpr::IteratorDefinition, 4> ID;
+  bool IsCorrect = true;
+  for (const OMPIteratorData &D : Data) {
+    TypeSourceInfo *TInfo = nullptr;
+    SourceLocation StartLoc;
+    QualType DeclTy;
+    if (!D.Type.getAsOpaquePtr()) {
+      // OpenMP 5.0, 2.1.6 Iterators
+      // In an iterator-specifier, if the iterator-type is not specified then
+      // the type of that iterator is of int type.
+      DeclTy = Context.IntTy;
+      StartLoc = D.DeclIdentLoc;
+    } else {
+      DeclTy = Sema::GetTypeFromParser(D.Type, &TInfo);
+      StartLoc = TInfo->getTypeLoc().getBeginLoc();
+    }
+
+    bool IsDeclTyDependent = DeclTy->isDependentType() ||
+                             DeclTy->containsUnexpandedParameterPack() ||
+                             DeclTy->isInstantiationDependentType();
+    if (!IsDeclTyDependent) {
+      if (!DeclTy->isIntegralType(Context) && !DeclTy->isAnyPointerType()) {
+        // OpenMP 5.0, 2.1.6 Iterators, Restrictions, C/C++
+        // The iterator-type must be an integral or pointer type.
+        Diag(StartLoc, diag::err_omp_iterator_not_integral_or_pointer)
+            << DeclTy;
+        IsCorrect = false;
+        continue;
+      }
+      if (DeclTy.isConstant(Context)) {
+        // OpenMP 5.0, 2.1.6 Iterators, Restrictions, C/C++
+        // The iterator-type must not be const qualified.
+        Diag(StartLoc, diag::err_omp_iterator_not_integral_or_pointer)
+            << DeclTy;
+        IsCorrect = false;
+        continue;
+      }
+    }
+
+    // Iterator declaration.
+    assert(D.DeclIdent && "Identifier expected.");
+    // Always try to create iterator declarator to avoid extra error messages
+    // about unknown declarations use.
+    auto *VD =
+        VarDecl::Create(Context, SemaRef.CurContext, StartLoc, D.DeclIdentLoc,
+                        D.DeclIdent, DeclTy, TInfo, SC_None);
+    VD->setImplicit();
+    if (S) {
+      // Check for conflicting previous declaration.
+      DeclarationNameInfo NameInfo(VD->getDeclName(), D.DeclIdentLoc);
+      LookupResult Previous(SemaRef, NameInfo, Sema::LookupOrdinaryName,
+                            Sema::ForVisibleRedeclaration);
+      Previous.suppressDiagnostics();
+      SemaRef.LookupName(Previous, S);
+
+      SemaRef.FilterLookupForScope(Previous, SemaRef.CurContext, S,
+                                   /*ConsiderLinkage=*/false,
+                                   /*AllowInlineNamespace=*/false);
+      if (!Previous.empty()) {
+        NamedDecl *Old = Previous.getRepresentativeDecl();
+        Diag(D.DeclIdentLoc, diag::err_redefinition) << VD->getDeclName();
+        Diag(Old->getLocation(), diag::note_previous_definition);
+      } else {
+        SemaRef.PushOnScopeChains(VD, S);
+      }
+    } else {
+      SemaRef.CurContext->addDecl(VD);
+    }
+
+    /// Act on the iterator variable declaration.
+    ActOnOpenMPIteratorVarDecl(VD);
+
+    Expr *Begin = D.Range.Begin;
+    if (!IsDeclTyDependent && Begin && !Begin->isTypeDependent()) {
+      ExprResult BeginRes =
+          SemaRef.PerformImplicitConversion(Begin, DeclTy, Sema::AA_Converting);
+      Begin = BeginRes.get();
+    }
+    Expr *End = D.Range.End;
+    if (!IsDeclTyDependent && End && !End->isTypeDependent()) {
+      ExprResult EndRes =
+          SemaRef.PerformImplicitConversion(End, DeclTy, Sema::AA_Converting);
+      End = EndRes.get();
+    }
+    Expr *Step = D.Range.Step;
+    if (!IsDeclTyDependent && Step && !Step->isTypeDependent()) {
+      if (!Step->getType()->isIntegralType(Context)) {
+        Diag(Step->getExprLoc(), diag::err_omp_iterator_step_not_integral)
+            << Step << Step->getSourceRange();
+        IsCorrect = false;
+        continue;
+      }
+      std::optional<llvm::APSInt> Result =
+          Step->getIntegerConstantExpr(Context);
+      // OpenMP 5.0, 2.1.6 Iterators, Restrictions
+      // If the step expression of a range-specification equals zero, the
+      // behavior is unspecified.
+      if (Result && Result->isZero()) {
+        Diag(Step->getExprLoc(), diag::err_omp_iterator_step_constant_zero)
+            << Step << Step->getSourceRange();
+        IsCorrect = false;
+        continue;
+      }
+    }
+    if (!Begin || !End || !IsCorrect) {
+      IsCorrect = false;
+      continue;
+    }
+    OMPIteratorExpr::IteratorDefinition &IDElem = ID.emplace_back();
+    IDElem.IteratorDecl = VD;
+    IDElem.AssignmentLoc = D.AssignLoc;
+    IDElem.Range.Begin = Begin;
+    IDElem.Range.End = End;
+    IDElem.Range.Step = Step;
+    IDElem.ColonLoc = D.ColonLoc;
+    IDElem.SecondColonLoc = D.SecColonLoc;
+  }
+  if (!IsCorrect) {
+    // Invalidate all created iterator declarations if error is found.
+    for (const OMPIteratorExpr::IteratorDefinition &D : ID) {
+      if (Decl *ID = D.IteratorDecl)
+        ID->setInvalidDecl();
+    }
+    return ExprError();
+  }
+  SmallVector<OMPIteratorHelperData, 4> Helpers;
+  if (!SemaRef.CurContext->isDependentContext()) {
+    // Build number of ityeration for each iteration range.
+    // Ni = ((Stepi > 0) ? ((Endi + Stepi -1 - Begini)/Stepi) :
+    // ((Begini-Stepi-1-Endi) / -Stepi);
+    for (OMPIteratorExpr::IteratorDefinition &D : ID) {
+      // (Endi - Begini)
+      ExprResult Res = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Sub,
+                                                  D.Range.End, D.Range.Begin);
+      if (!Res.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      ExprResult St, St1;
+      if (D.Range.Step) {
+        St = D.Range.Step;
+        // (Endi - Begini) + Stepi
+        Res = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Add, Res.get(),
+                                         St.get());
+        if (!Res.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // (Endi - Begini) + Stepi - 1
+        Res = SemaRef.CreateBuiltinBinOp(
+            D.AssignmentLoc, BO_Sub, Res.get(),
+            SemaRef.ActOnIntegerConstant(D.AssignmentLoc, 1).get());
+        if (!Res.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // ((Endi - Begini) + Stepi - 1) / Stepi
+        Res = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Div, Res.get(),
+                                         St.get());
+        if (!Res.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        St1 = SemaRef.CreateBuiltinUnaryOp(D.AssignmentLoc, UO_Minus,
+                                           D.Range.Step);
+        // (Begini - Endi)
+        ExprResult Res1 = SemaRef.CreateBuiltinBinOp(
+            D.AssignmentLoc, BO_Sub, D.Range.Begin, D.Range.End);
+        if (!Res1.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // (Begini - Endi) - Stepi
+        Res1 = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Add, Res1.get(),
+                                          St1.get());
+        if (!Res1.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // (Begini - Endi) - Stepi - 1
+        Res1 = SemaRef.CreateBuiltinBinOp(
+            D.AssignmentLoc, BO_Sub, Res1.get(),
+            SemaRef.ActOnIntegerConstant(D.AssignmentLoc, 1).get());
+        if (!Res1.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // ((Begini - Endi) - Stepi - 1) / (-Stepi)
+        Res1 = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Div, Res1.get(),
+                                          St1.get());
+        if (!Res1.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        // Stepi > 0.
+        ExprResult CmpRes = SemaRef.CreateBuiltinBinOp(
+            D.AssignmentLoc, BO_GT, D.Range.Step,
+            SemaRef.ActOnIntegerConstant(D.AssignmentLoc, 0).get());
+        if (!CmpRes.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+        Res = SemaRef.ActOnConditionalOp(D.AssignmentLoc, D.AssignmentLoc,
+                                         CmpRes.get(), Res.get(), Res1.get());
+        if (!Res.isUsable()) {
+          IsCorrect = false;
+          continue;
+        }
+      }
+      Res = SemaRef.ActOnFinishFullExpr(Res.get(), /*DiscardedValue=*/false);
+      if (!Res.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+
+      // Build counter update.
+      // Build counter.
+      auto *CounterVD = VarDecl::Create(Context, SemaRef.CurContext,
+                                        D.IteratorDecl->getBeginLoc(),
+                                        D.IteratorDecl->getBeginLoc(), nullptr,
+                                        Res.get()->getType(), nullptr, SC_None);
+      CounterVD->setImplicit();
+      ExprResult RefRes =
+          SemaRef.BuildDeclRefExpr(CounterVD, CounterVD->getType(), VK_LValue,
+                                   D.IteratorDecl->getBeginLoc());
+      // Build counter update.
+      // I = Begini + counter * Stepi;
+      ExprResult UpdateRes;
+      if (D.Range.Step) {
+        UpdateRes = SemaRef.CreateBuiltinBinOp(
+            D.AssignmentLoc, BO_Mul,
+            SemaRef.DefaultLvalueConversion(RefRes.get()).get(), St.get());
+      } else {
+        UpdateRes = SemaRef.DefaultLvalueConversion(RefRes.get());
+      }
+      if (!UpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      UpdateRes = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Add,
+                                             D.Range.Begin, UpdateRes.get());
+      if (!UpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      ExprResult VDRes =
+          SemaRef.BuildDeclRefExpr(cast<VarDecl>(D.IteratorDecl),
+                                   cast<VarDecl>(D.IteratorDecl)->getType(),
+                                   VK_LValue, D.IteratorDecl->getBeginLoc());
+      UpdateRes = SemaRef.CreateBuiltinBinOp(D.AssignmentLoc, BO_Assign,
+                                             VDRes.get(), UpdateRes.get());
+      if (!UpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      UpdateRes =
+          SemaRef.ActOnFinishFullExpr(UpdateRes.get(), /*DiscardedValue=*/true);
+      if (!UpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      ExprResult CounterUpdateRes = SemaRef.CreateBuiltinUnaryOp(
+          D.AssignmentLoc, UO_PreInc, RefRes.get());
+      if (!CounterUpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      CounterUpdateRes = SemaRef.ActOnFinishFullExpr(CounterUpdateRes.get(),
+                                                     /*DiscardedValue=*/true);
+      if (!CounterUpdateRes.isUsable()) {
+        IsCorrect = false;
+        continue;
+      }
+      OMPIteratorHelperData &HD = Helpers.emplace_back();
+      HD.CounterVD = CounterVD;
+      HD.Upper = Res.get();
+      HD.Update = UpdateRes.get();
+      HD.CounterUpdate = CounterUpdateRes.get();
+    }
+  } else {
+    Helpers.assign(ID.size(), {});
+  }
+  if (!IsCorrect) {
+    // Invalidate all created iterator declarations if error is found.
+    for (const OMPIteratorExpr::IteratorDefinition &D : ID) {
+      if (Decl *ID = D.IteratorDecl)
+        ID->setInvalidDecl();
+    }
+    return ExprError();
+  }
+  return OMPIteratorExpr::Create(Context, Context.OMPIteratorTy, IteratorKwLoc,
+                                 LLoc, RLoc, ID, Helpers);
+}
+
+SemaOpenMP::SemaOpenMP(Sema &S)
+    : SemaBase(S), VarDataSharingAttributesStack(nullptr) {}
diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
index d28c24cfdfd33..a7b33f0db047e 100644
--- a/clang/lib/Sema/SemaStmt.cpp
+++ b/clang/lib/Sema/SemaStmt.cpp
@@ -35,6 +35,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"
@@ -3097,7 +3098,7 @@ StmtResult Sema::BuildCXXForRangeStmt(
   // In OpenMP loop region loop control variable must be private. Perform
   // analysis of first part (if any).
   if (getLangOpts().OpenMP >= 50 && BeginDeclStmt.isUsable())
-    ActOnOpenMPLoopInitialization(ForLoc, BeginDeclStmt.get());
+    OpenMP().ActOnOpenMPLoopInitialization(ForLoc, BeginDeclStmt.get());
 
   return new (Context) CXXForRangeStmt(
       InitStmt, RangeDS, cast_or_null<DeclStmt>(BeginDeclStmt.get()),
@@ -4822,7 +4823,8 @@ buildCapturedStmtCaptureList(Sema &S, CapturedRegionScopeInfo *RSI,
       assert(Cap.isVariableCapture() && "unknown kind of capture");
 
       if (S.getLangOpts().OpenMP && RSI->CapRegionKind == CR_OpenMP)
-        S.setOpenMPCaptureKind(Field, Cap.getVariable(), RSI->OpenMPLevel);
+        S.OpenMP().setOpenMPCaptureKind(Field, Cap.getVariable(),
+                                        RSI->OpenMPLevel);
 
       Captures.push_back(CapturedStmt::Capture(
           Cap.getLocation(),
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index c45a8d1408fff..6d359c5a9a024 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -28,6 +28,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateInstCallback.h"
 #include "llvm/Support/TimeProfiler.h"
@@ -399,7 +400,7 @@ static void instantiateOMPDeclareSimdDeclAttr(
     ++SI;
   }
   LinModifiers.append(Attr.modifiers_begin(), Attr.modifiers_end());
-  (void)S.ActOnOpenMPDeclareSimdDirective(
+  (void)S.OpenMP().ActOnOpenMPDeclareSimdDirective(
       S.ConvertDeclToDeclGroup(New), Attr.getBranchState(), Simdlen.get(),
       Uniforms, Aligneds, Alignments, Linears, LinModifiers, Steps,
       Attr.getRange());
@@ -476,9 +477,9 @@ static void instantiateOMPDeclareVariantAttr(
   // Check function/variant ref for `omp declare variant` but not for `omp
   // begin declare variant` (which use implicit attributes).
   std::optional<std::pair<FunctionDecl *, Expr *>> DeclVarData =
-      S.checkOpenMPDeclareVariantFunction(S.ConvertDeclToDeclGroup(New), E, TI,
-                                          Attr.appendArgs_size(),
-                                          Attr.getRange());
+      S.OpenMP().checkOpenMPDeclareVariantFunction(
+          S.ConvertDeclToDeclGroup(New), E, TI, Attr.appendArgs_size(),
+          Attr.getRange());
 
   if (!DeclVarData)
     return;
@@ -539,7 +540,7 @@ static void instantiateOMPDeclareVariantAttr(
     AppendArgs.emplace_back(II.IsTarget, II.IsTargetSync);
   }
 
-  S.ActOnOpenMPDeclareVariantDirective(
+  S.OpenMP().ActOnOpenMPDeclareVariantDirective(
       FD, E, TI, NothingExprs, NeedDevicePtrExprs, AppendArgs, SourceLocation(),
       SourceLocation(), Attr.getRange());
 }
@@ -3587,7 +3588,7 @@ Decl *TemplateDeclInstantiator::VisitOMPThreadPrivateDecl(
   }
 
   OMPThreadPrivateDecl *TD =
-    SemaRef.CheckOMPThreadPrivateDecl(D->getLocation(), Vars);
+      SemaRef.OpenMP().CheckOMPThreadPrivateDecl(D->getLocation(), Vars);
 
   TD->setAccess(AS_public);
   Owner->addDecl(TD);
@@ -3610,14 +3611,14 @@ Decl *TemplateDeclInstantiator::VisitOMPAllocateDecl(OMPAllocateDecl *D) {
       ExprResult NewE = SemaRef.SubstExpr(AC->getAllocator(), TemplateArgs);
       if (!NewE.isUsable())
         continue;
-      IC = SemaRef.ActOnOpenMPAllocatorClause(
+      IC = SemaRef.OpenMP().ActOnOpenMPAllocatorClause(
           NewE.get(), AC->getBeginLoc(), AC->getLParenLoc(), AC->getEndLoc());
     } else if (auto *AC = dyn_cast<OMPAlignClause>(C)) {
       ExprResult NewE = SemaRef.SubstExpr(AC->getAlignment(), TemplateArgs);
       if (!NewE.isUsable())
         continue;
-      IC = SemaRef.ActOnOpenMPAlignClause(NewE.get(), AC->getBeginLoc(),
-                                          AC->getLParenLoc(), AC->getEndLoc());
+      IC = SemaRef.OpenMP().ActOnOpenMPAlignClause(
+          NewE.get(), AC->getBeginLoc(), AC->getLParenLoc(), AC->getEndLoc());
       // If align clause value ends up being invalid, this can end up null.
       if (!IC)
         continue;
@@ -3625,7 +3626,7 @@ Decl *TemplateDeclInstantiator::VisitOMPAllocateDecl(OMPAllocateDecl *D) {
     Clauses.push_back(IC);
   }
 
-  Sema::DeclGroupPtrTy Res = SemaRef.ActOnOpenMPAllocateDirective(
+  Sema::DeclGroupPtrTy Res = SemaRef.OpenMP().ActOnOpenMPAllocateDirective(
       D->getLocation(), Vars, Clauses, Owner);
   if (Res.get().isNull())
     return nullptr;
@@ -3646,7 +3647,7 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
       D->getType()->containsUnexpandedParameterPack();
   QualType SubstReductionType;
   if (RequiresInstantiation) {
-    SubstReductionType = SemaRef.ActOnOpenMPDeclareReductionType(
+    SubstReductionType = SemaRef.OpenMP().ActOnOpenMPDeclareReductionType(
         D->getLocation(),
         ParsedType::make(SemaRef.SubstType(
             D->getType(), TemplateArgs, D->getLocation(), DeclarationName())));
@@ -3667,7 +3668,7 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
         SemaRef.CurrentInstantiationScope->findInstantiationOf(PrevDeclInScope)
             ->get<Decl *>());
   }
-  auto DRD = SemaRef.ActOnOpenMPDeclareReductionDirectiveStart(
+  auto DRD = SemaRef.OpenMP().ActOnOpenMPDeclareReductionDirectiveStart(
       /*S=*/nullptr, Owner, D->getDeclName(), ReductionTypes, D->getAccess(),
       PrevDeclInScope);
   auto *NewDRD = cast<OMPDeclareReductionDecl>(DRD.get().getSingleDecl());
@@ -3676,7 +3677,7 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
   Expr *SubstInitializer = nullptr;
   // Combiners instantiation sequence.
   if (Combiner) {
-    SemaRef.ActOnOpenMPDeclareReductionCombinerStart(
+    SemaRef.OpenMP().ActOnOpenMPDeclareReductionCombinerStart(
         /*S=*/nullptr, NewDRD);
     SemaRef.CurrentInstantiationScope->InstantiatedLocal(
         cast<DeclRefExpr>(D->getCombinerIn())->getDecl(),
@@ -3688,12 +3689,14 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
     Sema::CXXThisScopeRAII ThisScope(SemaRef, ThisContext, Qualifiers(),
                                      ThisContext);
     SubstCombiner = SemaRef.SubstExpr(Combiner, TemplateArgs).get();
-    SemaRef.ActOnOpenMPDeclareReductionCombinerEnd(NewDRD, SubstCombiner);
+    SemaRef.OpenMP().ActOnOpenMPDeclareReductionCombinerEnd(NewDRD,
+                                                            SubstCombiner);
   }
   // Initializers instantiation sequence.
   if (Init) {
-    VarDecl *OmpPrivParm = SemaRef.ActOnOpenMPDeclareReductionInitializerStart(
-        /*S=*/nullptr, NewDRD);
+    VarDecl *OmpPrivParm =
+        SemaRef.OpenMP().ActOnOpenMPDeclareReductionInitializerStart(
+            /*S=*/nullptr, NewDRD);
     SemaRef.CurrentInstantiationScope->InstantiatedLocal(
         cast<DeclRefExpr>(D->getInitOrig())->getDecl(),
         cast<DeclRefExpr>(NewDRD->getInitOrig())->getDecl());
@@ -3710,8 +3713,8 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
         SemaRef.InstantiateVariableInitializer(OmpPrivParm, OldPrivParm,
                                                TemplateArgs);
     }
-    SemaRef.ActOnOpenMPDeclareReductionInitializerEnd(NewDRD, SubstInitializer,
-                                                      OmpPrivParm);
+    SemaRef.OpenMP().ActOnOpenMPDeclareReductionInitializerEnd(
+        NewDRD, SubstInitializer, OmpPrivParm);
   }
   IsCorrect = IsCorrect && SubstCombiner &&
               (!Init ||
@@ -3720,7 +3723,7 @@ Decl *TemplateDeclInstantiator::VisitOMPDeclareReductionDecl(
                (D->getInitializerKind() != OMPDeclareReductionInitKind::Call &&
                 !SubstInitializer));
 
-  (void)SemaRef.ActOnOpenMPDeclareReductionDirectiveEnd(
+  (void)SemaRef.OpenMP().ActOnOpenMPDeclareReductionDirectiveEnd(
       /*S=*/nullptr, DRD, IsCorrect && !D->isInvalidDecl());
 
   return NewDRD;
@@ -3736,7 +3739,7 @@ TemplateDeclInstantiator::VisitOMPDeclareMapperDecl(OMPDeclareMapperDecl *D) {
   QualType SubstMapperTy;
   DeclarationName VN = D->getVarName();
   if (RequiresInstantiation) {
-    SubstMapperTy = SemaRef.ActOnOpenMPDeclareMapperType(
+    SubstMapperTy = SemaRef.OpenMP().ActOnOpenMPDeclareMapperType(
         D->getLocation(),
         ParsedType::make(SemaRef.SubstType(D->getType(), TemplateArgs,
                                            D->getLocation(), VN)));
@@ -3756,11 +3759,12 @@ TemplateDeclInstantiator::VisitOMPDeclareMapperDecl(OMPDeclareMapperDecl *D) {
   SmallVector<OMPClause *, 6> Clauses;
   // Instantiate the mapper variable.
   DeclarationNameInfo DirName;
-  SemaRef.StartOpenMPDSABlock(llvm::omp::OMPD_declare_mapper, DirName,
-                              /*S=*/nullptr,
-                              (*D->clauselist_begin())->getBeginLoc());
-  ExprResult MapperVarRef = SemaRef.ActOnOpenMPDeclareMapperDirectiveVarDecl(
-      /*S=*/nullptr, SubstMapperTy, D->getLocation(), VN);
+  SemaRef.OpenMP().StartOpenMPDSABlock(llvm::omp::OMPD_declare_mapper, DirName,
+                                       /*S=*/nullptr,
+                                       (*D->clauselist_begin())->getBeginLoc());
+  ExprResult MapperVarRef =
+      SemaRef.OpenMP().ActOnOpenMPDeclareMapperDirectiveVarDecl(
+          /*S=*/nullptr, SubstMapperTy, D->getLocation(), VN);
   SemaRef.CurrentInstantiationScope->InstantiatedLocal(
       cast<DeclRefExpr>(D->getMapperVarRef())->getDecl(),
       cast<DeclRefExpr>(MapperVarRef.get())->getDecl());
@@ -3790,17 +3794,17 @@ TemplateDeclInstantiator::VisitOMPDeclareMapperDecl(OMPDeclareMapperDecl *D) {
         SemaRef.SubstDeclarationNameInfo(OldC->getMapperIdInfo(), TemplateArgs);
     OMPVarListLocTy Locs(OldC->getBeginLoc(), OldC->getLParenLoc(),
                          OldC->getEndLoc());
-    OMPClause *NewC = SemaRef.ActOnOpenMPMapClause(
+    OMPClause *NewC = SemaRef.OpenMP().ActOnOpenMPMapClause(
         OldC->getIteratorModifier(), OldC->getMapTypeModifiers(),
         OldC->getMapTypeModifiersLoc(), SS, NewNameInfo, OldC->getMapType(),
         OldC->isImplicitMapType(), OldC->getMapLoc(), OldC->getColonLoc(),
         NewVars, Locs);
     Clauses.push_back(NewC);
   }
-  SemaRef.EndOpenMPDSABlock(nullptr);
+  SemaRef.OpenMP().EndOpenMPDSABlock(nullptr);
   if (!IsCorrect)
     return nullptr;
-  Sema::DeclGroupPtrTy DG = SemaRef.ActOnOpenMPDeclareMapperDirective(
+  Sema::DeclGroupPtrTy DG = SemaRef.OpenMP().ActOnOpenMPDeclareMapperDirective(
       /*S=*/nullptr, Owner, D->getDeclName(), SubstMapperTy, D->getLocation(),
       VN, D->getAccess(), MapperVarRef.get(), Clauses, PrevDeclInScope);
   Decl *NewDMD = DG.get().getSingleDecl();
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index 404c4e8e31b55..1b31df8d97fba 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -35,6 +35,7 @@
 #include "clang/Sema/ScopeInfo.h"
 #include "clang/Sema/SemaCUDA.h"
 #include "clang/Sema/SemaInternal.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/Template.h"
 #include "clang/Sema/TemplateInstCallback.h"
 #include "llvm/ADT/ArrayRef.h"
@@ -2640,7 +2641,7 @@ QualType Sema::BuildArrayType(QualType T, ArraySizeModifier ASM,
   } else if (isSFINAEContext()) {
     VLADiag = diag::err_vla_in_sfinae;
     VLAIsError = true;
-  } else if (getLangOpts().OpenMP && isInOpenMPTaskUntiedContext()) {
+  } else if (getLangOpts().OpenMP && OpenMP().isInOpenMPTaskUntiedContext()) {
     VLADiag = diag::err_openmp_vla_in_task_untied;
     VLAIsError = true;
   } else if (getLangOpts().CPlusPlus) {
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 9d15f3eacbb0f..acc2d7ff9d427 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -40,6 +40,7 @@
 #include "clang/Sema/SemaDiagnostic.h"
 #include "clang/Sema/SemaInternal.h"
 #include "clang/Sema/SemaOpenACC.h"
+#include "clang/Sema/SemaOpenMP.h"
 #include "clang/Sema/SemaSYCL.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -1656,7 +1657,7 @@ class TreeTransform {
   /// Ensures that the outermost loop in @p LoopStmt is wrapped by a
   /// OMPCanonicalLoop.
   StmtResult RebuildOMPCanonicalLoop(Stmt *LoopStmt) {
-    return getSema().ActOnOpenMPCanonicalLoop(LoopStmt);
+    return getSema().OpenMP().ActOnOpenMPCanonicalLoop(LoopStmt);
   }
 
   /// Build a new OpenMP executable directive.
@@ -1669,7 +1670,7 @@ class TreeTransform {
       Stmt *AStmt, SourceLocation StartLoc, SourceLocation EndLoc,
       OpenMPDirectiveKind PrevMappedDirective = OMPD_unknown) {
 
-    return getSema().ActOnOpenMPExecutableDirective(
+    return getSema().OpenMP().ActOnOpenMPExecutableDirective(
         Kind, DirName, CancelRegion, Clauses, AStmt, StartLoc, EndLoc,
         PrevMappedDirective);
   }
@@ -1684,9 +1685,9 @@ class TreeTransform {
                                 SourceLocation NameModifierLoc,
                                 SourceLocation ColonLoc,
                                 SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPIfClause(NameModifier, Condition, StartLoc,
-                                         LParenLoc, NameModifierLoc, ColonLoc,
-                                         EndLoc);
+    return getSema().OpenMP().ActOnOpenMPIfClause(
+        NameModifier, Condition, StartLoc, LParenLoc, NameModifierLoc, ColonLoc,
+        EndLoc);
   }
 
   /// Build a new OpenMP 'final' clause.
@@ -1696,8 +1697,8 @@ class TreeTransform {
   OMPClause *RebuildOMPFinalClause(Expr *Condition, SourceLocation StartLoc,
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPFinalClause(Condition, StartLoc, LParenLoc,
-                                            EndLoc);
+    return getSema().OpenMP().ActOnOpenMPFinalClause(Condition, StartLoc,
+                                                     LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'num_threads' clause.
@@ -1708,8 +1709,8 @@ class TreeTransform {
                                         SourceLocation StartLoc,
                                         SourceLocation LParenLoc,
                                         SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNumThreadsClause(NumThreads, StartLoc,
-                                                 LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNumThreadsClause(NumThreads, StartLoc,
+                                                          LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'safelen' clause.
@@ -1719,7 +1720,8 @@ class TreeTransform {
   OMPClause *RebuildOMPSafelenClause(Expr *Len, SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPSafelenClause(Len, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPSafelenClause(Len, StartLoc, LParenLoc,
+                                                       EndLoc);
   }
 
   /// Build a new OpenMP 'simdlen' clause.
@@ -1729,28 +1731,30 @@ class TreeTransform {
   OMPClause *RebuildOMPSimdlenClause(Expr *Len, SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPSimdlenClause(Len, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPSimdlenClause(Len, StartLoc, LParenLoc,
+                                                       EndLoc);
   }
 
   OMPClause *RebuildOMPSizesClause(ArrayRef<Expr *> Sizes,
                                    SourceLocation StartLoc,
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPSizesClause(Sizes, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPSizesClause(Sizes, StartLoc, LParenLoc,
+                                                     EndLoc);
   }
 
   /// Build a new OpenMP 'full' clause.
   OMPClause *RebuildOMPFullClause(SourceLocation StartLoc,
                                   SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPFullClause(StartLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPFullClause(StartLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'partial' clause.
   OMPClause *RebuildOMPPartialClause(Expr *Factor, SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPPartialClause(Factor, StartLoc, LParenLoc,
-                                              EndLoc);
+    return getSema().OpenMP().ActOnOpenMPPartialClause(Factor, StartLoc,
+                                                       LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'allocator' clause.
@@ -1760,7 +1764,8 @@ class TreeTransform {
   OMPClause *RebuildOMPAllocatorClause(Expr *A, SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPAllocatorClause(A, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPAllocatorClause(A, StartLoc, LParenLoc,
+                                                         EndLoc);
   }
 
   /// Build a new OpenMP 'collapse' clause.
@@ -1770,8 +1775,8 @@ class TreeTransform {
   OMPClause *RebuildOMPCollapseClause(Expr *Num, SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPCollapseClause(Num, StartLoc, LParenLoc,
-                                               EndLoc);
+    return getSema().OpenMP().ActOnOpenMPCollapseClause(Num, StartLoc,
+                                                        LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'default' clause.
@@ -1782,8 +1787,8 @@ class TreeTransform {
                                      SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDefaultClause(Kind, KindKwLoc,
-                                              StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDefaultClause(
+        Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'proc_bind' clause.
@@ -1795,8 +1800,8 @@ class TreeTransform {
                                       SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPProcBindClause(Kind, KindKwLoc,
-                                               StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPProcBindClause(
+        Kind, KindKwLoc, StartLoc, LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'schedule' clause.
@@ -1808,7 +1813,7 @@ class TreeTransform {
       OpenMPScheduleClauseKind Kind, Expr *ChunkSize, SourceLocation StartLoc,
       SourceLocation LParenLoc, SourceLocation M1Loc, SourceLocation M2Loc,
       SourceLocation KindLoc, SourceLocation CommaLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPScheduleClause(
+    return getSema().OpenMP().ActOnOpenMPScheduleClause(
         M1, M2, Kind, ChunkSize, StartLoc, LParenLoc, M1Loc, M2Loc, KindLoc,
         CommaLoc, EndLoc);
   }
@@ -1820,7 +1825,8 @@ class TreeTransform {
   OMPClause *RebuildOMPOrderedClause(SourceLocation StartLoc,
                                      SourceLocation EndLoc,
                                      SourceLocation LParenLoc, Expr *Num) {
-    return getSema().ActOnOpenMPOrderedClause(StartLoc, EndLoc, LParenLoc, Num);
+    return getSema().OpenMP().ActOnOpenMPOrderedClause(StartLoc, EndLoc,
+                                                       LParenLoc, Num);
   }
 
   /// Build a new OpenMP 'private' clause.
@@ -1831,8 +1837,8 @@ class TreeTransform {
                                      SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPPrivateClause(VarList, StartLoc, LParenLoc,
-                                              EndLoc);
+    return getSema().OpenMP().ActOnOpenMPPrivateClause(VarList, StartLoc,
+                                                       LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'firstprivate' clause.
@@ -1843,8 +1849,8 @@ class TreeTransform {
                                           SourceLocation StartLoc,
                                           SourceLocation LParenLoc,
                                           SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPFirstprivateClause(VarList, StartLoc, LParenLoc,
-                                                   EndLoc);
+    return getSema().OpenMP().ActOnOpenMPFirstprivateClause(VarList, StartLoc,
+                                                            LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'lastprivate' clause.
@@ -1858,7 +1864,7 @@ class TreeTransform {
                                          SourceLocation StartLoc,
                                          SourceLocation LParenLoc,
                                          SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPLastprivateClause(
+    return getSema().OpenMP().ActOnOpenMPLastprivateClause(
         VarList, LPKind, LPKindLoc, ColonLoc, StartLoc, LParenLoc, EndLoc);
   }
 
@@ -1870,8 +1876,8 @@ class TreeTransform {
                                     SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPSharedClause(VarList, StartLoc, LParenLoc,
-                                             EndLoc);
+    return getSema().OpenMP().ActOnOpenMPSharedClause(VarList, StartLoc,
+                                                      LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'reduction' clause.
@@ -1885,7 +1891,7 @@ class TreeTransform {
       SourceLocation EndLoc, CXXScopeSpec &ReductionIdScopeSpec,
       const DeclarationNameInfo &ReductionId,
       ArrayRef<Expr *> UnresolvedReductions) {
-    return getSema().ActOnOpenMPReductionClause(
+    return getSema().OpenMP().ActOnOpenMPReductionClause(
         VarList, Modifier, StartLoc, LParenLoc, ModifierLoc, ColonLoc, EndLoc,
         ReductionIdScopeSpec, ReductionId, UnresolvedReductions);
   }
@@ -1900,7 +1906,7 @@ class TreeTransform {
       CXXScopeSpec &ReductionIdScopeSpec,
       const DeclarationNameInfo &ReductionId,
       ArrayRef<Expr *> UnresolvedReductions) {
-    return getSema().ActOnOpenMPTaskReductionClause(
+    return getSema().OpenMP().ActOnOpenMPTaskReductionClause(
         VarList, StartLoc, LParenLoc, ColonLoc, EndLoc, ReductionIdScopeSpec,
         ReductionId, UnresolvedReductions);
   }
@@ -1916,7 +1922,7 @@ class TreeTransform {
                               CXXScopeSpec &ReductionIdScopeSpec,
                               const DeclarationNameInfo &ReductionId,
                               ArrayRef<Expr *> UnresolvedReductions) {
-    return getSema().ActOnOpenMPInReductionClause(
+    return getSema().OpenMP().ActOnOpenMPInReductionClause(
         VarList, StartLoc, LParenLoc, ColonLoc, EndLoc, ReductionIdScopeSpec,
         ReductionId, UnresolvedReductions);
   }
@@ -1930,9 +1936,9 @@ class TreeTransform {
       SourceLocation LParenLoc, OpenMPLinearClauseKind Modifier,
       SourceLocation ModifierLoc, SourceLocation ColonLoc,
       SourceLocation StepModifierLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPLinearClause(VarList, Step, StartLoc, LParenLoc,
-                                             Modifier, ModifierLoc, ColonLoc,
-                                             StepModifierLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPLinearClause(
+        VarList, Step, StartLoc, LParenLoc, Modifier, ModifierLoc, ColonLoc,
+        StepModifierLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'aligned' clause.
@@ -1944,8 +1950,8 @@ class TreeTransform {
                                      SourceLocation LParenLoc,
                                      SourceLocation ColonLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPAlignedClause(VarList, Alignment, StartLoc,
-                                              LParenLoc, ColonLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPAlignedClause(
+        VarList, Alignment, StartLoc, LParenLoc, ColonLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'copyin' clause.
@@ -1956,8 +1962,8 @@ class TreeTransform {
                                     SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPCopyinClause(VarList, StartLoc, LParenLoc,
-                                             EndLoc);
+    return getSema().OpenMP().ActOnOpenMPCopyinClause(VarList, StartLoc,
+                                                      LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'copyprivate' clause.
@@ -1968,8 +1974,8 @@ class TreeTransform {
                                          SourceLocation StartLoc,
                                          SourceLocation LParenLoc,
                                          SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPCopyprivateClause(VarList, StartLoc, LParenLoc,
-                                                  EndLoc);
+    return getSema().OpenMP().ActOnOpenMPCopyprivateClause(VarList, StartLoc,
+                                                           LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'flush' pseudo clause.
@@ -1980,8 +1986,8 @@ class TreeTransform {
                                    SourceLocation StartLoc,
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPFlushClause(VarList, StartLoc, LParenLoc,
-                                            EndLoc);
+    return getSema().OpenMP().ActOnOpenMPFlushClause(VarList, StartLoc,
+                                                     LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'depobj' pseudo clause.
@@ -1991,8 +1997,8 @@ class TreeTransform {
   OMPClause *RebuildOMPDepobjClause(Expr *Depobj, SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDepobjClause(Depobj, StartLoc, LParenLoc,
-                                             EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDepobjClause(Depobj, StartLoc,
+                                                      LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'depend' pseudo clause.
@@ -2004,8 +2010,8 @@ class TreeTransform {
                                     SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDependClause(Data, DepModifier, VarList,
-                                             StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDependClause(
+        Data, DepModifier, VarList, StartLoc, LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'device' clause.
@@ -2017,8 +2023,8 @@ class TreeTransform {
                                     SourceLocation LParenLoc,
                                     SourceLocation ModifierLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDeviceClause(Modifier, Device, StartLoc,
-                                             LParenLoc, ModifierLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDeviceClause(
+        Modifier, Device, StartLoc, LParenLoc, ModifierLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'map' clause.
@@ -2032,7 +2038,7 @@ class TreeTransform {
       OpenMPMapClauseKind MapType, bool IsMapTypeImplicit,
       SourceLocation MapLoc, SourceLocation ColonLoc, ArrayRef<Expr *> VarList,
       const OMPVarListLocTy &Locs, ArrayRef<Expr *> UnresolvedMappers) {
-    return getSema().ActOnOpenMPMapClause(
+    return getSema().OpenMP().ActOnOpenMPMapClause(
         IteratorModifier, MapTypeModifiers, MapTypeModifiersLoc,
         MapperIdScopeSpec, MapperId, MapType, IsMapTypeImplicit, MapLoc,
         ColonLoc, VarList, Locs,
@@ -2048,8 +2054,8 @@ class TreeTransform {
                                       SourceLocation LParenLoc,
                                       SourceLocation ColonLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPAllocateClause(Allocate, VarList, StartLoc,
-                                               LParenLoc, ColonLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPAllocateClause(
+        Allocate, VarList, StartLoc, LParenLoc, ColonLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'num_teams' clause.
@@ -2059,8 +2065,8 @@ class TreeTransform {
   OMPClause *RebuildOMPNumTeamsClause(Expr *NumTeams, SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNumTeamsClause(NumTeams, StartLoc, LParenLoc,
-                                               EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNumTeamsClause(NumTeams, StartLoc,
+                                                        LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'thread_limit' clause.
@@ -2071,8 +2077,8 @@ class TreeTransform {
                                          SourceLocation StartLoc,
                                          SourceLocation LParenLoc,
                                          SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPThreadLimitClause(ThreadLimit, StartLoc,
-                                                  LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPThreadLimitClause(
+        ThreadLimit, StartLoc, LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'priority' clause.
@@ -2082,8 +2088,8 @@ class TreeTransform {
   OMPClause *RebuildOMPPriorityClause(Expr *Priority, SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPPriorityClause(Priority, StartLoc, LParenLoc,
-                                               EndLoc);
+    return getSema().OpenMP().ActOnOpenMPPriorityClause(Priority, StartLoc,
+                                                        LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'grainsize' clause.
@@ -2095,8 +2101,8 @@ class TreeTransform {
                                        SourceLocation LParenLoc,
                                        SourceLocation ModifierLoc,
                                        SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPGrainsizeClause(Modifier, Device, StartLoc,
-                                                LParenLoc, ModifierLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPGrainsizeClause(
+        Modifier, Device, StartLoc, LParenLoc, ModifierLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'num_tasks' clause.
@@ -2108,8 +2114,8 @@ class TreeTransform {
                                       SourceLocation LParenLoc,
                                       SourceLocation ModifierLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNumTasksClause(Modifier, NumTasks, StartLoc,
-                                               LParenLoc, ModifierLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNumTasksClause(
+        Modifier, NumTasks, StartLoc, LParenLoc, ModifierLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'hint' clause.
@@ -2119,7 +2125,8 @@ class TreeTransform {
   OMPClause *RebuildOMPHintClause(Expr *Hint, SourceLocation StartLoc,
                                   SourceLocation LParenLoc,
                                   SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPHintClause(Hint, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPHintClause(Hint, StartLoc, LParenLoc,
+                                                    EndLoc);
   }
 
   /// Build a new OpenMP 'detach' clause.
@@ -2129,7 +2136,8 @@ class TreeTransform {
   OMPClause *RebuildOMPDetachClause(Expr *Evt, SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDetachClause(Evt, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDetachClause(Evt, StartLoc, LParenLoc,
+                                                      EndLoc);
   }
 
   /// Build a new OpenMP 'dist_schedule' clause.
@@ -2141,7 +2149,7 @@ class TreeTransform {
                                Expr *ChunkSize, SourceLocation StartLoc,
                                SourceLocation LParenLoc, SourceLocation KindLoc,
                                SourceLocation CommaLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDistScheduleClause(
+    return getSema().OpenMP().ActOnOpenMPDistScheduleClause(
         Kind, ChunkSize, StartLoc, LParenLoc, KindLoc, CommaLoc, EndLoc);
   }
 
@@ -2156,9 +2164,9 @@ class TreeTransform {
                      DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
                      ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
                      ArrayRef<Expr *> UnresolvedMappers) {
-    return getSema().ActOnOpenMPToClause(MotionModifiers, MotionModifiersLoc,
-                                         MapperIdScopeSpec, MapperId, ColonLoc,
-                                         VarList, Locs, UnresolvedMappers);
+    return getSema().OpenMP().ActOnOpenMPToClause(
+        MotionModifiers, MotionModifiersLoc, MapperIdScopeSpec, MapperId,
+        ColonLoc, VarList, Locs, UnresolvedMappers);
   }
 
   /// Build a new OpenMP 'from' clause.
@@ -2172,7 +2180,7 @@ class TreeTransform {
                        DeclarationNameInfo &MapperId, SourceLocation ColonLoc,
                        ArrayRef<Expr *> VarList, const OMPVarListLocTy &Locs,
                        ArrayRef<Expr *> UnresolvedMappers) {
-    return getSema().ActOnOpenMPFromClause(
+    return getSema().OpenMP().ActOnOpenMPFromClause(
         MotionModifiers, MotionModifiersLoc, MapperIdScopeSpec, MapperId,
         ColonLoc, VarList, Locs, UnresolvedMappers);
   }
@@ -2183,7 +2191,7 @@ class TreeTransform {
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPUseDevicePtrClause(ArrayRef<Expr *> VarList,
                                           const OMPVarListLocTy &Locs) {
-    return getSema().ActOnOpenMPUseDevicePtrClause(VarList, Locs);
+    return getSema().OpenMP().ActOnOpenMPUseDevicePtrClause(VarList, Locs);
   }
 
   /// Build a new OpenMP 'use_device_addr' clause.
@@ -2192,7 +2200,7 @@ class TreeTransform {
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPUseDeviceAddrClause(ArrayRef<Expr *> VarList,
                                            const OMPVarListLocTy &Locs) {
-    return getSema().ActOnOpenMPUseDeviceAddrClause(VarList, Locs);
+    return getSema().OpenMP().ActOnOpenMPUseDeviceAddrClause(VarList, Locs);
   }
 
   /// Build a new OpenMP 'is_device_ptr' clause.
@@ -2201,7 +2209,7 @@ class TreeTransform {
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPIsDevicePtrClause(ArrayRef<Expr *> VarList,
                                          const OMPVarListLocTy &Locs) {
-    return getSema().ActOnOpenMPIsDevicePtrClause(VarList, Locs);
+    return getSema().OpenMP().ActOnOpenMPIsDevicePtrClause(VarList, Locs);
   }
 
   /// Build a new OpenMP 'has_device_addr' clause.
@@ -2210,7 +2218,7 @@ class TreeTransform {
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPHasDeviceAddrClause(ArrayRef<Expr *> VarList,
                                            const OMPVarListLocTy &Locs) {
-    return getSema().ActOnOpenMPHasDeviceAddrClause(VarList, Locs);
+    return getSema().OpenMP().ActOnOpenMPHasDeviceAddrClause(VarList, Locs);
   }
 
   /// Build a new OpenMP 'defaultmap' clause.
@@ -2224,8 +2232,8 @@ class TreeTransform {
                                         SourceLocation MLoc,
                                         SourceLocation KindLoc,
                                         SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDefaultmapClause(M, Kind, StartLoc, LParenLoc,
-                                                 MLoc, KindLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDefaultmapClause(
+        M, Kind, StartLoc, LParenLoc, MLoc, KindLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'nontemporal' clause.
@@ -2236,8 +2244,8 @@ class TreeTransform {
                                          SourceLocation StartLoc,
                                          SourceLocation LParenLoc,
                                          SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNontemporalClause(VarList, StartLoc, LParenLoc,
-                                                  EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNontemporalClause(VarList, StartLoc,
+                                                           LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'inclusive' clause.
@@ -2248,8 +2256,8 @@ class TreeTransform {
                                        SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPInclusiveClause(VarList, StartLoc, LParenLoc,
-                                                EndLoc);
+    return getSema().OpenMP().ActOnOpenMPInclusiveClause(VarList, StartLoc,
+                                                         LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'exclusive' clause.
@@ -2260,8 +2268,8 @@ class TreeTransform {
                                        SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPExclusiveClause(VarList, StartLoc, LParenLoc,
-                                                EndLoc);
+    return getSema().OpenMP().ActOnOpenMPExclusiveClause(VarList, StartLoc,
+                                                         LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'uses_allocators' clause.
@@ -2269,10 +2277,10 @@ class TreeTransform {
   /// By default, performs semantic analysis to build the new OpenMP clause.
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPUsesAllocatorsClause(
-      ArrayRef<Sema::UsesAllocatorsData> Data, SourceLocation StartLoc,
+      ArrayRef<SemaOpenMP::UsesAllocatorsData> Data, SourceLocation StartLoc,
       SourceLocation LParenLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPUsesAllocatorClause(StartLoc, LParenLoc, EndLoc,
-                                                    Data);
+    return getSema().OpenMP().ActOnOpenMPUsesAllocatorClause(
+        StartLoc, LParenLoc, EndLoc, Data);
   }
 
   /// Build a new OpenMP 'affinity' clause.
@@ -2284,8 +2292,8 @@ class TreeTransform {
                                       SourceLocation ColonLoc,
                                       SourceLocation EndLoc, Expr *Modifier,
                                       ArrayRef<Expr *> Locators) {
-    return getSema().ActOnOpenMPAffinityClause(StartLoc, LParenLoc, ColonLoc,
-                                               EndLoc, Modifier, Locators);
+    return getSema().OpenMP().ActOnOpenMPAffinityClause(
+        StartLoc, LParenLoc, ColonLoc, EndLoc, Modifier, Locators);
   }
 
   /// Build a new OpenMP 'order' clause.
@@ -2296,8 +2304,8 @@ class TreeTransform {
       OpenMPOrderClauseKind Kind, SourceLocation KindKwLoc,
       SourceLocation StartLoc, SourceLocation LParenLoc, SourceLocation EndLoc,
       OpenMPOrderClauseModifier Modifier, SourceLocation ModifierKwLoc) {
-    return getSema().ActOnOpenMPOrderClause(Modifier, Kind, StartLoc, LParenLoc,
-                                            ModifierKwLoc, KindKwLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPOrderClause(
+        Modifier, Kind, StartLoc, LParenLoc, ModifierKwLoc, KindKwLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'init' clause.
@@ -2309,8 +2317,8 @@ class TreeTransform {
                                   SourceLocation LParenLoc,
                                   SourceLocation VarLoc,
                                   SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPInitClause(InteropVar, InteropInfo, StartLoc,
-                                           LParenLoc, VarLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPInitClause(
+        InteropVar, InteropInfo, StartLoc, LParenLoc, VarLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'use' clause.
@@ -2320,8 +2328,8 @@ class TreeTransform {
   OMPClause *RebuildOMPUseClause(Expr *InteropVar, SourceLocation StartLoc,
                                  SourceLocation LParenLoc,
                                  SourceLocation VarLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPUseClause(InteropVar, StartLoc, LParenLoc,
-                                          VarLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPUseClause(InteropVar, StartLoc,
+                                                   LParenLoc, VarLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'destroy' clause.
@@ -2332,8 +2340,8 @@ class TreeTransform {
                                      SourceLocation LParenLoc,
                                      SourceLocation VarLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDestroyClause(InteropVar, StartLoc, LParenLoc,
-                                              VarLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPDestroyClause(
+        InteropVar, StartLoc, LParenLoc, VarLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'novariants' clause.
@@ -2344,8 +2352,8 @@ class TreeTransform {
                                         SourceLocation StartLoc,
                                         SourceLocation LParenLoc,
                                         SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNovariantsClause(Condition, StartLoc, LParenLoc,
-                                                 EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNovariantsClause(Condition, StartLoc,
+                                                          LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'nocontext' clause.
@@ -2355,8 +2363,8 @@ class TreeTransform {
   OMPClause *RebuildOMPNocontextClause(Expr *Condition, SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPNocontextClause(Condition, StartLoc, LParenLoc,
-                                                EndLoc);
+    return getSema().OpenMP().ActOnOpenMPNocontextClause(Condition, StartLoc,
+                                                         LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'filter' clause.
@@ -2366,8 +2374,8 @@ class TreeTransform {
   OMPClause *RebuildOMPFilterClause(Expr *ThreadID, SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPFilterClause(ThreadID, StartLoc, LParenLoc,
-                                             EndLoc);
+    return getSema().OpenMP().ActOnOpenMPFilterClause(ThreadID, StartLoc,
+                                                      LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'bind' clause.
@@ -2379,8 +2387,8 @@ class TreeTransform {
                                   SourceLocation StartLoc,
                                   SourceLocation LParenLoc,
                                   SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPBindClause(Kind, KindLoc, StartLoc, LParenLoc,
-                                           EndLoc);
+    return getSema().OpenMP().ActOnOpenMPBindClause(Kind, KindLoc, StartLoc,
+                                                    LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'ompx_dyn_cgroup_mem' clause.
@@ -2390,8 +2398,8 @@ class TreeTransform {
   OMPClause *RebuildOMPXDynCGroupMemClause(Expr *Size, SourceLocation StartLoc,
                                            SourceLocation LParenLoc,
                                            SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPXDynCGroupMemClause(Size, StartLoc, LParenLoc,
-                                                    EndLoc);
+    return getSema().OpenMP().ActOnOpenMPXDynCGroupMemClause(Size, StartLoc,
+                                                             LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'ompx_attribute' clause.
@@ -2402,8 +2410,8 @@ class TreeTransform {
                                         SourceLocation StartLoc,
                                         SourceLocation LParenLoc,
                                         SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPXAttributeClause(Attrs, StartLoc, LParenLoc,
-                                                 EndLoc);
+    return getSema().OpenMP().ActOnOpenMPXAttributeClause(Attrs, StartLoc,
+                                                          LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'ompx_bare' clause.
@@ -2412,7 +2420,7 @@ class TreeTransform {
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPXBareClause(SourceLocation StartLoc,
                                    SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPXBareClause(StartLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPXBareClause(StartLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'align' clause.
@@ -2422,7 +2430,8 @@ class TreeTransform {
   OMPClause *RebuildOMPAlignClause(Expr *A, SourceLocation StartLoc,
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPAlignClause(A, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPAlignClause(A, StartLoc, LParenLoc,
+                                                     EndLoc);
   }
 
   /// Build a new OpenMP 'at' clause.
@@ -2433,8 +2442,8 @@ class TreeTransform {
                                 SourceLocation StartLoc,
                                 SourceLocation LParenLoc,
                                 SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPAtClause(Kind, KwLoc, StartLoc, LParenLoc,
-                                         EndLoc);
+    return getSema().OpenMP().ActOnOpenMPAtClause(Kind, KwLoc, StartLoc,
+                                                  LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'severity' clause.
@@ -2446,8 +2455,8 @@ class TreeTransform {
                                       SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPSeverityClause(Kind, KwLoc, StartLoc, LParenLoc,
-                                               EndLoc);
+    return getSema().OpenMP().ActOnOpenMPSeverityClause(Kind, KwLoc, StartLoc,
+                                                        LParenLoc, EndLoc);
   }
 
   /// Build a new OpenMP 'message' clause.
@@ -2457,7 +2466,8 @@ class TreeTransform {
   OMPClause *RebuildOMPMessageClause(Expr *MS, SourceLocation StartLoc,
                                      SourceLocation LParenLoc,
                                      SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPMessageClause(MS, StartLoc, LParenLoc, EndLoc);
+    return getSema().OpenMP().ActOnOpenMPMessageClause(MS, StartLoc, LParenLoc,
+                                                       EndLoc);
   }
 
   /// Build a new OpenMP 'doacross' clause.
@@ -2469,7 +2479,7 @@ class TreeTransform {
                            SourceLocation DepLoc, SourceLocation ColonLoc,
                            ArrayRef<Expr *> VarList, SourceLocation StartLoc,
                            SourceLocation LParenLoc, SourceLocation EndLoc) {
-    return getSema().ActOnOpenMPDoacrossClause(
+    return getSema().OpenMP().ActOnOpenMPDoacrossClause(
         DepType, DepLoc, ColonLoc, VarList, StartLoc, LParenLoc, EndLoc);
   }
 
@@ -2777,9 +2787,9 @@ class TreeTransform {
                                         SourceLocation ColonLocSecond,
                                         Expr *Length, Expr *Stride,
                                         SourceLocation RBracketLoc) {
-    return getSema().ActOnOMPArraySectionExpr(Base, LBracketLoc, LowerBound,
-                                              ColonLocFirst, ColonLocSecond,
-                                              Length, Stride, RBracketLoc);
+    return getSema().OpenMP().ActOnOMPArraySectionExpr(
+        Base, LBracketLoc, LowerBound, ColonLocFirst, ColonLocSecond, Length,
+        Stride, RBracketLoc);
   }
 
   /// Build a new array shaping expression.
@@ -2790,19 +2800,20 @@ class TreeTransform {
                                         SourceLocation RParenLoc,
                                         ArrayRef<Expr *> Dims,
                                         ArrayRef<SourceRange> BracketsRanges) {
-    return getSema().ActOnOMPArrayShapingExpr(Base, LParenLoc, RParenLoc, Dims,
-                                              BracketsRanges);
+    return getSema().OpenMP().ActOnOMPArrayShapingExpr(
+        Base, LParenLoc, RParenLoc, Dims, BracketsRanges);
   }
 
   /// Build a new iterator expression.
   ///
   /// By default, performs semantic analysis to build the new expression.
   /// Subclasses may override this routine to provide different behavior.
-  ExprResult RebuildOMPIteratorExpr(
-      SourceLocation IteratorKwLoc, SourceLocation LLoc, SourceLocation RLoc,
-      ArrayRef<Sema::OMPIteratorData> Data) {
-    return getSema().ActOnOMPIteratorExpr(/*Scope=*/nullptr, IteratorKwLoc,
-                                          LLoc, RLoc, Data);
+  ExprResult
+  RebuildOMPIteratorExpr(SourceLocation IteratorKwLoc, SourceLocation LLoc,
+                         SourceLocation RLoc,
+                         ArrayRef<SemaOpenMP::OMPIteratorData> Data) {
+    return getSema().OpenMP().ActOnOMPIteratorExpr(
+        /*Scope=*/nullptr, IteratorKwLoc, LLoc, RLoc, Data);
   }
 
   /// Build a new call expression.
@@ -8060,7 +8071,7 @@ template<typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformForStmt(ForStmt *S) {
   if (getSema().getLangOpts().OpenMP)
-    getSema().startOpenMPLoop();
+    getSema().OpenMP().startOpenMPLoop();
 
   // Transform the initialization statement
   StmtResult Init = getDerived().TransformStmt(S->getInit());
@@ -8070,7 +8081,8 @@ TreeTransform<Derived>::TransformForStmt(ForStmt *S) {
   // In OpenMP loop region loop control variable must be captured and be
   // private. Perform analysis of first part (if any).
   if (getSema().getLangOpts().OpenMP && Init.isUsable())
-    getSema().ActOnOpenMPLoopInitialization(S->getForLoc(), Init.get());
+    getSema().OpenMP().ActOnOpenMPLoopInitialization(S->getForLoc(),
+                                                     Init.get());
 
   // Transform the condition
   Sema::ConditionResult Cond = getDerived().TransformCondition(
@@ -9029,9 +9041,9 @@ StmtResult TreeTransform<Derived>::TransformOMPExecutableDirective(
   for (ArrayRef<OMPClause *>::iterator I = Clauses.begin(), E = Clauses.end();
        I != E; ++I) {
     if (*I) {
-      getDerived().getSema().StartOpenMPClause((*I)->getClauseKind());
+      getDerived().getSema().OpenMP().StartOpenMPClause((*I)->getClauseKind());
       OMPClause *Clause = getDerived().TransformOMPClause(*I);
-      getDerived().getSema().EndOpenMPClause();
+      getDerived().getSema().OpenMP().EndOpenMPClause();
       if (Clause)
         TClauses.push_back(Clause);
     } else {
@@ -9040,8 +9052,9 @@ StmtResult TreeTransform<Derived>::TransformOMPExecutableDirective(
   }
   StmtResult AssociatedStmt;
   if (D->hasAssociatedStmt() && D->getAssociatedStmt()) {
-    getDerived().getSema().ActOnOpenMPRegionStart(D->getDirectiveKind(),
-                                                  /*CurScope=*/nullptr);
+    getDerived().getSema().OpenMP().ActOnOpenMPRegionStart(
+        D->getDirectiveKind(),
+        /*CurScope=*/nullptr);
     StmtResult Body;
     {
       Sema::CompoundScopeRAII CompoundScope(getSema());
@@ -9059,7 +9072,7 @@ StmtResult TreeTransform<Derived>::TransformOMPExecutableDirective(
         Body = getDerived().RebuildOMPCanonicalLoop(Body.get());
     }
     AssociatedStmt =
-        getDerived().getSema().ActOnOpenMPRegionEnd(Body, TClauses);
+        getDerived().getSema().OpenMP().ActOnOpenMPRegionEnd(Body, TClauses);
     if (AssociatedStmt.isInvalid()) {
       return StmtError();
     }
@@ -9100,10 +9113,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPParallelDirective(OMPParallelDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9111,10 +9124,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPSimdDirective(OMPSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_simd, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9122,10 +9135,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTileDirective(OMPTileDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(D->getDirectiveKind(), DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      D->getDirectiveKind(), DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9133,10 +9146,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPUnrollDirective(OMPUnrollDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(D->getDirectiveKind(), DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      D->getDirectiveKind(), DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9144,10 +9157,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPForDirective(OMPForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_for, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_for, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9155,10 +9168,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPForSimdDirective(OMPForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_for_simd, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_for_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9166,10 +9179,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPSectionsDirective(OMPSectionsDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_sections, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_sections, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9177,10 +9190,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPSectionDirective(OMPSectionDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_section, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_section, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9188,10 +9201,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPScopeDirective(OMPScopeDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_scope, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_scope, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9199,10 +9212,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPSingleDirective(OMPSingleDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_single, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_single, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9210,20 +9223,20 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPMasterDirective(OMPMasterDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_master, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_master, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
 template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPCriticalDirective(OMPCriticalDirective *D) {
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_critical, D->getDirectiveName(), nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9231,10 +9244,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelForDirective(
     OMPParallelForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_for, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_for, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9242,10 +9255,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelForSimdDirective(
     OMPParallelForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_for_simd, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_for_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9253,10 +9266,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelMasterDirective(
     OMPParallelMasterDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_master, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_master, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9264,10 +9277,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelMaskedDirective(
     OMPParallelMaskedDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_masked, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_masked, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9275,10 +9288,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelSectionsDirective(
     OMPParallelSectionsDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_sections, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_sections, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9286,10 +9299,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTaskDirective(OMPTaskDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_task, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_task, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9297,10 +9310,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTaskyieldDirective(
     OMPTaskyieldDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_taskyield, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_taskyield, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9308,10 +9321,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPBarrierDirective(OMPBarrierDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_barrier, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_barrier, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9319,10 +9332,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTaskwaitDirective(OMPTaskwaitDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_taskwait, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_taskwait, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9330,10 +9343,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPErrorDirective(OMPErrorDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_error, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_error, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9341,10 +9354,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTaskgroupDirective(
     OMPTaskgroupDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_taskgroup, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_taskgroup, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9352,10 +9365,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPFlushDirective(OMPFlushDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_flush, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_flush, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9363,10 +9376,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPDepobjDirective(OMPDepobjDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_depobj, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_depobj, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9374,10 +9387,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPScanDirective(OMPScanDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_scan, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_scan, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9385,10 +9398,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPOrderedDirective(OMPOrderedDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_ordered, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_ordered, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9396,10 +9409,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPAtomicDirective(OMPAtomicDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_atomic, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_atomic, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9407,10 +9420,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTargetDirective(OMPTargetDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9418,10 +9431,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetDataDirective(
     OMPTargetDataDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_data, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_data, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9429,10 +9442,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetEnterDataDirective(
     OMPTargetEnterDataDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_enter_data, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_enter_data, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9440,10 +9453,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetExitDataDirective(
     OMPTargetExitDataDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_exit_data, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_exit_data, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9451,10 +9464,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetParallelDirective(
     OMPTargetParallelDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_parallel, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_parallel, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9462,10 +9475,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetParallelForDirective(
     OMPTargetParallelForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_parallel_for, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_parallel_for, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9473,10 +9486,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetUpdateDirective(
     OMPTargetUpdateDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_update, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_update, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9484,10 +9497,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTeamsDirective(OMPTeamsDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_teams, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_teams, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9495,10 +9508,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPCancellationPointDirective(
     OMPCancellationPointDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_cancellation_point, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_cancellation_point, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9506,10 +9519,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPCancelDirective(OMPCancelDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_cancel, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_cancel, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9517,10 +9530,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPTaskLoopDirective(OMPTaskLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_taskloop, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_taskloop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9528,10 +9541,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTaskLoopSimdDirective(
     OMPTaskLoopSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_taskloop_simd, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_taskloop_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9539,10 +9552,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPMasterTaskLoopDirective(
     OMPMasterTaskLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_master_taskloop, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_master_taskloop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9550,10 +9563,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPMaskedTaskLoopDirective(
     OMPMaskedTaskLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_masked_taskloop, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_masked_taskloop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9561,10 +9574,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPMasterTaskLoopSimdDirective(
     OMPMasterTaskLoopSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_master_taskloop_simd, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_master_taskloop_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9572,10 +9585,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPMaskedTaskLoopSimdDirective(
     OMPMaskedTaskLoopSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_masked_taskloop_simd, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_masked_taskloop_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9583,10 +9596,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelMasterTaskLoopDirective(
     OMPParallelMasterTaskLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_parallel_master_taskloop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9594,10 +9607,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelMaskedTaskLoopDirective(
     OMPParallelMaskedTaskLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_parallel_masked_taskloop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9606,10 +9619,10 @@ StmtResult
 TreeTransform<Derived>::TransformOMPParallelMasterTaskLoopSimdDirective(
     OMPParallelMasterTaskLoopSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_parallel_master_taskloop_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9618,10 +9631,10 @@ StmtResult
 TreeTransform<Derived>::TransformOMPParallelMaskedTaskLoopSimdDirective(
     OMPParallelMaskedTaskLoopSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_parallel_masked_taskloop_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9629,10 +9642,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPDistributeDirective(
     OMPDistributeDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_distribute, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_distribute, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9640,10 +9653,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPDistributeParallelForDirective(
     OMPDistributeParallelForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_distribute_parallel_for, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9652,10 +9665,10 @@ StmtResult
 TreeTransform<Derived>::TransformOMPDistributeParallelForSimdDirective(
     OMPDistributeParallelForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_distribute_parallel_for_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9663,10 +9676,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPDistributeSimdDirective(
     OMPDistributeSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_distribute_simd, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_distribute_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9674,10 +9687,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetParallelForSimdDirective(
     OMPTargetParallelForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_target_parallel_for_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9685,10 +9698,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetSimdDirective(
     OMPTargetSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_simd, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9696,10 +9709,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTeamsDistributeDirective(
     OMPTeamsDistributeDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_teams_distribute, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_teams_distribute, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9707,10 +9720,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTeamsDistributeSimdDirective(
     OMPTeamsDistributeSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_teams_distribute_simd, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9718,11 +9731,11 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTeamsDistributeParallelForSimdDirective(
     OMPTeamsDistributeParallelForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_teams_distribute_parallel_for_simd, DirName, nullptr,
       D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9730,10 +9743,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTeamsDistributeParallelForDirective(
     OMPTeamsDistributeParallelForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_teams_distribute_parallel_for, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9741,10 +9754,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetTeamsDirective(
     OMPTargetTeamsDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_teams, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_teams, DirName, nullptr, D->getBeginLoc());
   auto Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9752,10 +9765,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetTeamsDistributeDirective(
     OMPTargetTeamsDistributeDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_target_teams_distribute, DirName, nullptr, D->getBeginLoc());
   auto Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9764,11 +9777,11 @@ StmtResult
 TreeTransform<Derived>::TransformOMPTargetTeamsDistributeParallelForDirective(
     OMPTargetTeamsDistributeParallelForDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_target_teams_distribute_parallel_for, DirName, nullptr,
       D->getBeginLoc());
   auto Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9777,11 +9790,11 @@ StmtResult TreeTransform<Derived>::
     TransformOMPTargetTeamsDistributeParallelForSimdDirective(
         OMPTargetTeamsDistributeParallelForSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_target_teams_distribute_parallel_for_simd, DirName, nullptr,
       D->getBeginLoc());
   auto Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9790,10 +9803,10 @@ StmtResult
 TreeTransform<Derived>::TransformOMPTargetTeamsDistributeSimdDirective(
     OMPTargetTeamsDistributeSimdDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
       OMPD_target_teams_distribute_simd, DirName, nullptr, D->getBeginLoc());
   auto Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9801,10 +9814,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPInteropDirective(OMPInteropDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_interop, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_interop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9812,10 +9825,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPDispatchDirective(OMPDispatchDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_dispatch, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_dispatch, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9823,10 +9836,10 @@ template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPMaskedDirective(OMPMaskedDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_masked, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_masked, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9834,10 +9847,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPGenericLoopDirective(
     OMPGenericLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_loop, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_loop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9845,10 +9858,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTeamsGenericLoopDirective(
     OMPTeamsGenericLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_teams_loop, DirName, nullptr,
-                                             D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_teams_loop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9856,10 +9869,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPTargetTeamsGenericLoopDirective(
     OMPTargetTeamsGenericLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_teams_loop, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_teams_loop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9867,10 +9880,10 @@ template <typename Derived>
 StmtResult TreeTransform<Derived>::TransformOMPParallelGenericLoopDirective(
     OMPParallelGenericLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_parallel_loop, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_parallel_loop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -9879,10 +9892,10 @@ StmtResult
 TreeTransform<Derived>::TransformOMPTargetParallelGenericLoopDirective(
     OMPTargetParallelGenericLoopDirective *D) {
   DeclarationNameInfo DirName;
-  getDerived().getSema().StartOpenMPDSABlock(OMPD_target_parallel_loop, DirName,
-                                             nullptr, D->getBeginLoc());
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      OMPD_target_parallel_loop, DirName, nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
-  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
@@ -10972,7 +10985,7 @@ TreeTransform<Derived>::TransformOMPExclusiveClause(OMPExclusiveClause *C) {
 template <typename Derived>
 OMPClause *TreeTransform<Derived>::TransformOMPUsesAllocatorsClause(
     OMPUsesAllocatorsClause *C) {
-  SmallVector<Sema::UsesAllocatorsData, 16> Data;
+  SmallVector<SemaOpenMP::UsesAllocatorsData, 16> Data;
   Data.reserve(C->getNumberOfAllocators());
   for (unsigned I = 0, E = C->getNumberOfAllocators(); I < E; ++I) {
     OMPUsesAllocatorsClause::Data D = C->getAllocatorData(I);
@@ -10985,7 +10998,7 @@ OMPClause *TreeTransform<Derived>::TransformOMPUsesAllocatorsClause(
       if (AllocatorTraits.isInvalid())
         continue;
     }
-    Sema::UsesAllocatorsData &NewD = Data.emplace_back();
+    SemaOpenMP::UsesAllocatorsData &NewD = Data.emplace_back();
     NewD.Allocator = Allocator.get();
     NewD.AllocatorTraits = AllocatorTraits.get();
     NewD.LParenLoc = D.LParenLoc;
@@ -11667,7 +11680,7 @@ template <typename Derived>
 ExprResult
 TreeTransform<Derived>::TransformOMPIteratorExpr(OMPIteratorExpr *E) {
   unsigned NumIterators = E->numOfIterators();
-  SmallVector<Sema::OMPIteratorData, 4> Data(NumIterators);
+  SmallVector<SemaOpenMP::OMPIteratorData, 4> Data(NumIterators);
 
   bool ErrorFound = false;
   bool NeedToRebuild = getDerived().AlwaysRebuild();
@@ -11802,7 +11815,8 @@ TreeTransform<Derived>::TransformMemberExpr(MemberExpr *E) {
     // Skip for member expression of (this->f), rebuilt thisi->f is needed
     // for Openmp where the field need to be privatizized in the case.
     if (!(isa<CXXThisExpr>(E->getBase()) &&
-          getSema().isOpenMPRebuildMemberExpr(cast<ValueDecl>(Member)))) {
+          getSema().OpenMP().isOpenMPRebuildMemberExpr(
+              cast<ValueDecl>(Member)))) {
       // Mark it referenced in the new context regardless.
       // FIXME: this is a bit instantiation-specific.
       SemaRef.MarkMemberReferenced(E);

From 9e7aab951ffba0211193ceb435c6b49e4e19ac24 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Tue, 16 Apr 2024 07:43:13 -0500
Subject: [PATCH 584/886] [CUDA] Rename SM_32 to SM_32_ to work around AIX
 headers (#88779)

Summary:
AIX headers define this, so we need to work around it. In the future
this will be removed but for now we should just rename it to avoid these
issues.
---
 clang/include/clang/Basic/Cuda.h         | 8 +++-----
 clang/lib/Basic/Cuda.cpp                 | 6 +++---
 clang/lib/Basic/Targets/NVPTX.cpp        | 2 +-
 clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp | 2 +-
 4 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/clang/include/clang/Basic/Cuda.h b/clang/include/clang/Basic/Cuda.h
index 38f30543a0f66..ba0e4465a0f5a 100644
--- a/clang/include/clang/Basic/Cuda.h
+++ b/clang/include/clang/Basic/Cuda.h
@@ -50,17 +50,15 @@ const char *CudaVersionToString(CudaVersion V);
 // Input is "Major.Minor"
 CudaVersion CudaStringToVersion(const llvm::Twine &S);
 
-// We have a name conflict with sys/mac.h on AIX
-#ifdef SM_32
-#undef SM_32
-#endif
 enum class CudaArch {
   UNUSED,
   UNKNOWN,
+  // TODO: Deprecate and remove GPU architectures older than sm_52.
   SM_20,
   SM_21,
   SM_30,
-  SM_32,
+  // This has a name conflict with sys/mac.h on AIX, rename it as a workaround.
+  SM_32_,
   SM_35,
   SM_37,
   SM_50,
diff --git a/clang/lib/Basic/Cuda.cpp b/clang/lib/Basic/Cuda.cpp
index 1b1da6a1356f2..113483db5729b 100644
--- a/clang/lib/Basic/Cuda.cpp
+++ b/clang/lib/Basic/Cuda.cpp
@@ -86,7 +86,7 @@ static const CudaArchToStringMap arch_names[] = {
     // clang-format off
     {CudaArch::UNUSED, "", ""},
     SM2(20, "compute_20"), SM2(21, "compute_20"), // Fermi
-    SM(30), SM(32), SM(35), SM(37),  // Kepler
+    SM(30), {CudaArch::SM_32_, "sm_32", "compute_32"}, SM(35), SM(37),  // Kepler
     SM(50), SM(52), SM(53),          // Maxwell
     SM(60), SM(61), SM(62),          // Pascal
     SM(70), SM(72),                  // Volta
@@ -186,7 +186,7 @@ CudaVersion MinVersionForCudaArch(CudaArch A) {
   case CudaArch::SM_20:
   case CudaArch::SM_21:
   case CudaArch::SM_30:
-  case CudaArch::SM_32:
+  case CudaArch::SM_32_:
   case CudaArch::SM_35:
   case CudaArch::SM_37:
   case CudaArch::SM_50:
@@ -231,7 +231,7 @@ CudaVersion MaxVersionForCudaArch(CudaArch A) {
   case CudaArch::SM_21:
     return CudaVersion::CUDA_80;
   case CudaArch::SM_30:
-  case CudaArch::SM_32:
+  case CudaArch::SM_32_:
     return CudaVersion::CUDA_102;
   case CudaArch::SM_35:
   case CudaArch::SM_37:
diff --git a/clang/lib/Basic/Targets/NVPTX.cpp b/clang/lib/Basic/Targets/NVPTX.cpp
index b47c399fef604..8ad9e6e5f5891 100644
--- a/clang/lib/Basic/Targets/NVPTX.cpp
+++ b/clang/lib/Basic/Targets/NVPTX.cpp
@@ -239,7 +239,7 @@ void NVPTXTargetInfo::getTargetDefines(const LangOptions &Opts,
         return "210";
       case CudaArch::SM_30:
         return "300";
-      case CudaArch::SM_32:
+      case CudaArch::SM_32_:
         return "320";
       case CudaArch::SM_35:
         return "350";
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 59ba03c6b8625..eb716520e5ff5 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -3466,7 +3466,7 @@ void CGOpenMPRuntimeGPU::processRequiresDirective(
       case CudaArch::SM_20:
       case CudaArch::SM_21:
       case CudaArch::SM_30:
-      case CudaArch::SM_32:
+      case CudaArch::SM_32_:
       case CudaArch::SM_35:
       case CudaArch::SM_37:
       case CudaArch::SM_50:

From e7fb49c24e4be4780ee4df9829980c5e8ddd511e Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Tue, 16 Apr 2024 09:00:57 -0400
Subject: [PATCH 585/886] Switch release notes links to using markup for github
 issues; NFC

---
 clang/docs/ReleaseNotes.rst | 43 +++++++++++++------------------------
 1 file changed, 15 insertions(+), 28 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 255d2cc044043..db90db6fa4ab0 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -104,8 +104,7 @@ C++20 Feature Support
 
 - Clang now implements [module.import]p7 fully. Clang now will import module
   units transitively for the module units coming from the same module of the
-  current module units.
-  Fixes `#84002 <https://github.com/llvm/llvm-project/issues/84002>`_.
+  current module units. Fixes #GH84002
 
 - Initial support for class template argument deduction (CTAD) for type alias
   templates (`P1814R0 <https://wg21.link/p1814r0>`_).
@@ -135,8 +134,7 @@ C++2c Feature Support
 Resolutions to C++ Defect Reports
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 - Substitute template parameter pack, when it is not explicitly specified
-  in the template parameters, but is deduced from a previous argument.
-  (`#78449: <https://github.com/llvm/llvm-project/issues/78449>`_).
+  in the template parameters, but is deduced from a previous argument. (#GH78449)
 
 - Type qualifications are now ignored when evaluating layout compatibility
   of two types.
@@ -176,8 +174,7 @@ C23 Feature Support
 
 - Clang now generates predefined macros of the form ``__TYPE_FMTB__`` and
   ``__TYPE_FMTb__`` (e.g., ``__UINT_FAST64_FMTB__``) in C23 mode for use with
-  macros typically exposed from ``<inttypes.h>``, such as ``PRIb8``.
-  (`#81896: <https://github.com/llvm/llvm-project/issues/81896>`_).
+  macros typically exposed from ``<inttypes.h>``, such as ``PRIb8``. (#GH81896)
 
 - Clang now supports `N3018 The constexpr specifier for object definitions`
   <https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3018.htm>`_.
@@ -215,7 +212,7 @@ New Compiler Flags
 
 - ``-Wmissing-designated-field-initializers``, grouped under ``-Wmissing-field-initializers``.
   This diagnostic can be disabled to make ``-Wmissing-field-initializers`` behave
-  like it did before Clang 18.x. Fixes (`#56628 <https://github.com/llvm/llvm-project/issues/68933>`_)
+  like it did before Clang 18.x. Fixes #GH56628
 
 Deprecated Compiler Flags
 -------------------------
@@ -254,8 +251,7 @@ Removed Compiler Flags
 
 - The ``-freroll-loops`` flag has been removed. It had no effect since Clang 13.
 - ``-m[no-]unaligned-access`` is removed for RISC-V and LoongArch.
-  ``-m[no-]strict-align``, also supported by GCC, should be used instead.
-  (`#85350 <https://github.com/llvm/llvm-project/pull/85350>`_.)
+  ``-m[no-]strict-align``, also supported by GCC, should be used instead. (#GH85350)
 
 Attribute Changes in Clang
 --------------------------
@@ -325,8 +321,7 @@ Improvements to Clang's diagnostics
   Fixes #GH82512.
 
 - Clang now provides improved warnings for the ``cleanup`` attribute to detect misuse scenarios,
-  such as attempting to call ``free`` on an unallocated object. Fixes
-  `#79443 <https://github.com/llvm/llvm-project/issues/79443>`_.
+  such as attempting to call ``free`` on an unallocated object. Fixes #GH79443.
 
 - Clang no longer warns when the ``bitand`` operator is used with boolean
   operands, distinguishing it from potential typographical errors or unintended
@@ -372,11 +367,10 @@ Improvements to Clang's time-trace
 Bug Fixes in This Version
 -------------------------
 - Clang's ``-Wundefined-func-template`` no longer warns on pure virtual
-  functions.
-  (`#74016 <https://github.com/llvm/llvm-project/issues/74016>`_)
+  functions. (#GH74016)
 
 - Fixed missing warnings when comparing mismatched enumeration constants
-  in C (`#29217 <https://github.com/llvm/llvm-project/issues/29217>`).
+  in C (#GH29217)
 
 - Clang now accepts elaborated-type-specifiers that explicitly specialize
   a member class template for an implicit instantiation of a class template.
@@ -415,7 +409,7 @@ Bug Fixes in This Version
   type only rather than to the complex type (e.g. ``_Complex float / int`` is now evaluated
   as ``_Complex float / float`` rather than ``_Complex float / _Complex float``), as mandated
   by the C standard. This significantly improves codegen of `*` and `/` especially.
-  Fixes (`#31205 <https://github.com/llvm/llvm-project/issues/31205>`_).
+  Fixes #GH31205.
 
 - Fixes an assertion failure on invalid code when trying to define member
   functions in lambdas.
@@ -464,8 +458,7 @@ Bug Fixes to C++ Support
 - Fix a crash when trying to call a varargs function that also has an explicit object parameter. (#GH80971)
 - Fixed a bug where abbreviated function templates would append their invented template parameters to
   an empty template parameter lists.
-- Fix parsing of abominable function types inside type traits.
-  Fixes (`#77585 <https://github.com/llvm/llvm-project/issues/77585>`_)
+- Fix parsing of abominable function types inside type traits. Fixes #GH77585
 - Clang now classifies aggregate initialization in C++17 and newer as constant
   or non-constant more accurately. Previously, only a subset of the initializer
   elements were considered, misclassifying some initializers as constant. Partially fixes
@@ -506,9 +499,7 @@ Bug Fixes to C++ Support
 - Fix a bug where overload resolution falsely reported an ambiguity when it was comparing
   a member-function against a non member function or a member-function with an
   explicit object parameter against a member function with no explicit object parameter
-  when one of the function had more specialized templates.
-  Fixes (`#82509 <https://github.com/llvm/llvm-project/issues/82509>`_)
-  and (`#74494 <https://github.com/llvm/llvm-project/issues/74494>`_)
+  when one of the function had more specialized templates. Fixes #GH82509 and #GH74494
 - Clang now supports direct lambda calls inside of a type alias template declarations.
   This addresses (#GH70601), (#GH76674), (#GH79555), (#GH81145) and (#GH82104).
 - Allow access to a public template alias declaration that refers to friend's
@@ -530,8 +521,7 @@ Bug Fixes to C++ Support
 - Fixed a bug that prevented member function templates of class templates declared with a deduced return type
   from being explicitly specialized for a given implicit instantiation of the class template.
 
-- Fix crash when inheriting from a cv-qualified type. Fixes:
-  (`#35603 <https://github.com/llvm/llvm-project/issues/35603>`_)
+- Fix crash when inheriting from a cv-qualified type. Fixes #GH35603
 - Fix a crash when the using enum declaration uses an anonymous enumeration. Fixes (#GH86790).
 - Handled an edge case in ``getFullyPackExpandedSize`` so that we now avoid a false-positive diagnostic. (#GH84220)
 - Clang now correctly tracks type dependence of by-value captures in lambdas with an explicit
@@ -539,8 +529,7 @@ Bug Fixes to C++ Support
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
 - Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
 - Fix a crash in requires expression with templated base class member function. Fixes (#GH84020).
-- placement new initializes typedef array with correct size
-  (`#GH41441 <https://github.com/llvm/llvm-project/issues/41441>`_)
+- Placement new initializes typedef array with correct size (#GH41441)
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -554,8 +543,7 @@ Miscellaneous Clang Crashes Fixed
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 - Do not attempt to dump the layout of dependent types or invalid declarations
-  when ``-fdump-record-layouts-complete`` is passed.
-  Fixes (`#83684 <https://github.com/llvm/llvm-project/issues/83684>`_).
+  when ``-fdump-record-layouts-complete`` is passed. Fixes #GH83684.
 
 OpenACC Specific Changes
 ------------------------
@@ -605,8 +593,7 @@ Windows Support
   would only be included if AVX was enabled at compile time. This was done to work
   around include times from MSVC STL including ``intrin.h`` under clang-cl.
   Clang-cl now provides ``intrin0.h`` for MSVC STL and therefore all intrinsic
-  features without requiring enablement at compile time.
-  Fixes: (`#53520 <https://github.com/llvm/llvm-project/issues/53520>`_)
+  features without requiring enablement at compile time. Fixes #GH53520
 
 - Improved compile times with MSVC STL. MSVC provides ``intrin0.h`` which is a
   header that only includes intrinsics that are used by MSVC STL to avoid the

From 9d8be2408768912dc113a342050049231e4fc8d1 Mon Sep 17 00:00:00 2001
From: Utkarsh Saxena <usx@google.com>
Date: Tue, 16 Apr 2024 15:30:32 +0200
Subject: [PATCH 586/886] Revert "[codegen] Emit missing cleanups for stmt-expr
 and coro suspensions" and related commits (#88884)

The original change caused widespread breakages in msan/ubsan tests and
causes `use-after-free`. Most likely we are adding more cleanups than
necessary.
---
 clang/lib/CodeGen/CGCall.cpp                  |  13 +-
 clang/lib/CodeGen/CGCleanup.cpp               |  49 +--
 clang/lib/CodeGen/CGCleanup.h                 |  57 +--
 clang/lib/CodeGen/CGDecl.cpp                  |  61 +--
 clang/lib/CodeGen/CGExpr.cpp                  |  12 +-
 clang/lib/CodeGen/CGExprAgg.cpp               |  87 ++--
 clang/lib/CodeGen/CGExprCXX.cpp               |  38 +-
 clang/lib/CodeGen/CodeGenFunction.cpp         |   6 -
 clang/lib/CodeGen/CodeGenFunction.h           |  99 +----
 .../CodeGenCXX/control-flow-in-stmt-expr.cpp  | 409 ------------------
 .../coro-suspend-cleanups.cpp                 |  93 ----
 11 files changed, 128 insertions(+), 796 deletions(-)
 delete mode 100644 clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
 delete mode 100644 clang/test/CodeGenCoroutines/coro-suspend-cleanups.cpp

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 0c860a3ccbd2f..7a0bc6fa77b88 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -4694,11 +4694,11 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
     AggValueSlot Slot = args.isUsingInAlloca()
         ? createPlaceholderSlot(*this, type) : CreateAggTemp(type, "agg.tmp");
 
-    bool DestroyedInCallee = true, NeedsCleanup = true;
+    bool DestroyedInCallee = true, NeedsEHCleanup = true;
     if (const auto *RD = type->getAsCXXRecordDecl())
       DestroyedInCallee = RD->hasNonTrivialDestructor();
     else
-      NeedsCleanup = type.isDestructedType();
+      NeedsEHCleanup = needsEHCleanup(type.isDestructedType());
 
     if (DestroyedInCallee)
       Slot.setExternallyDestructed();
@@ -4707,15 +4707,14 @@ void CodeGenFunction::EmitCallArg(CallArgList &args, const Expr *E,
     RValue RV = Slot.asRValue();
     args.add(RV, type);
 
-    if (DestroyedInCallee && NeedsCleanup) {
+    if (DestroyedInCallee && NeedsEHCleanup) {
       // Create a no-op GEP between the placeholder and the cleanup so we can
       // RAUW it successfully.  It also serves as a marker of the first
       // instruction where the cleanup is active.
-      pushFullExprCleanup<DestroyUnpassedArg>(NormalAndEHCleanup,
-                                              Slot.getAddress(), type);
+      pushFullExprCleanup<DestroyUnpassedArg>(EHCleanup, Slot.getAddress(),
+                                              type);
       // This unreachable is a temporary marker which will be removed later.
-      llvm::Instruction *IsActive =
-          Builder.CreateFlagLoad(llvm::Constant::getNullValue(Int8PtrTy));
+      llvm::Instruction *IsActive = Builder.CreateUnreachable();
       args.addArgCleanupDeactivation(EHStack.stable_begin(), IsActive);
     }
     return;
diff --git a/clang/lib/CodeGen/CGCleanup.cpp b/clang/lib/CodeGen/CGCleanup.cpp
index 8683f19d9da28..e6f8e6873004f 100644
--- a/clang/lib/CodeGen/CGCleanup.cpp
+++ b/clang/lib/CodeGen/CGCleanup.cpp
@@ -634,19 +634,12 @@ static void destroyOptimisticNormalEntry(CodeGenFunction &CGF,
 /// Pops a cleanup block.  If the block includes a normal cleanup, the
 /// current insertion point is threaded through the cleanup, as are
 /// any branch fixups on the cleanup.
-void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
-                                      bool ForDeactivation) {
+void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
   assert(!EHStack.empty() && "cleanup stack is empty!");
   assert(isa<EHCleanupScope>(*EHStack.begin()) && "top not a cleanup!");
   EHCleanupScope &Scope = cast<EHCleanupScope>(*EHStack.begin());
   assert(Scope.getFixupDepth() <= EHStack.getNumBranchFixups());
 
-  // If we are deactivating a normal cleanup, we need to pretend that the
-  // fallthrough is unreachable. We restore this IP before returning.
-  CGBuilderTy::InsertPoint NormalDeactivateOrigIP;
-  if (ForDeactivation && (Scope.isNormalCleanup() || !getLangOpts().EHAsynch)) {
-    NormalDeactivateOrigIP = Builder.saveAndClearIP();
-  }
   // Remember activation information.
   bool IsActive = Scope.isActive();
   Address NormalActiveFlag =
@@ -674,8 +667,7 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
 
   // - whether there's a fallthrough
   llvm::BasicBlock *FallthroughSource = Builder.GetInsertBlock();
-  bool HasFallthrough =
-      FallthroughSource != nullptr && (IsActive || HasExistingBranches);
+  bool HasFallthrough = (FallthroughSource != nullptr && IsActive);
 
   // Branch-through fall-throughs leave the insertion point set to the
   // end of the last cleanup, which points to the current scope.  The
@@ -700,11 +692,7 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
 
   // If we have a prebranched fallthrough into an inactive normal
   // cleanup, rewrite it so that it leads to the appropriate place.
-  if (Scope.isNormalCleanup() && HasPrebranchedFallthrough &&
-      !RequiresNormalCleanup) {
-    // FIXME: Come up with a program which would need forwarding prebranched
-    // fallthrough and add tests. Otherwise delete this and assert against it.
-    assert(!IsActive);
+  if (Scope.isNormalCleanup() && HasPrebranchedFallthrough && !IsActive) {
     llvm::BasicBlock *prebranchDest;
 
     // If the prebranch is semantically branching through the next
@@ -736,8 +724,6 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
     EHStack.popCleanup(); // safe because there are no fixups
     assert(EHStack.getNumBranchFixups() == 0 ||
            EHStack.hasNormalCleanups());
-    if (NormalDeactivateOrigIP.isSet())
-      Builder.restoreIP(NormalDeactivateOrigIP);
     return;
   }
 
@@ -774,19 +760,11 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
   if (!RequiresNormalCleanup) {
     // Mark CPP scope end for passed-by-value Arg temp
     //   per Windows ABI which is "normally" Cleanup in callee
-    if (IsEHa && getInvokeDest()) {
-      // If we are deactivating a normal cleanup then we don't have a
-      // fallthrough. Restore original IP to emit CPP scope ends in the correct
-      // block.
-      if (NormalDeactivateOrigIP.isSet())
-        Builder.restoreIP(NormalDeactivateOrigIP);
-      if (Personality.isMSVCXXPersonality() && Builder.GetInsertBlock())
+    if (IsEHa && getInvokeDest() && Builder.GetInsertBlock()) {
+      if (Personality.isMSVCXXPersonality())
         EmitSehCppScopeEnd();
-      if (NormalDeactivateOrigIP.isSet())
-        NormalDeactivateOrigIP = Builder.saveAndClearIP();
     }
     destroyOptimisticNormalEntry(*this, Scope);
-    Scope.MarkEmitted();
     EHStack.popCleanup();
   } else {
     // If we have a fallthrough and no other need for the cleanup,
@@ -803,7 +781,6 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
       }
 
       destroyOptimisticNormalEntry(*this, Scope);
-      Scope.MarkEmitted();
       EHStack.popCleanup();
 
       EmitCleanup(*this, Fn, cleanupFlags, NormalActiveFlag);
@@ -939,7 +916,6 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
       }
 
       // IV.  Pop the cleanup and emit it.
-      Scope.MarkEmitted();
       EHStack.popCleanup();
       assert(EHStack.hasNormalCleanups() == HasEnclosingCleanups);
 
@@ -1008,8 +984,6 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough,
     }
   }
 
-  if (NormalDeactivateOrigIP.isSet())
-    Builder.restoreIP(NormalDeactivateOrigIP);
   assert(EHStack.hasNormalCleanups() || EHStack.getNumBranchFixups() == 0);
 
   // Emit the EH cleanup if required.
@@ -1299,8 +1273,17 @@ void CodeGenFunction::DeactivateCleanupBlock(EHScopeStack::stable_iterator C,
   // to the current RunCleanupsScope.
   if (C == EHStack.stable_begin() &&
       CurrentCleanupScopeDepth.strictlyEncloses(C)) {
-    PopCleanupBlock(/*FallthroughIsBranchThrough=*/false,
-                    /*ForDeactivation=*/true);
+    // Per comment below, checking EHAsynch is not really necessary
+    // it's there to assure zero-impact w/o EHAsynch option
+    if (!Scope.isNormalCleanup() && getLangOpts().EHAsynch) {
+      PopCleanupBlock();
+    } else {
+      // If it's a normal cleanup, we need to pretend that the
+      // fallthrough is unreachable.
+      CGBuilderTy::InsertPoint SavedIP = Builder.saveAndClearIP();
+      PopCleanupBlock();
+      Builder.restoreIP(SavedIP);
+    }
     return;
   }
 
diff --git a/clang/lib/CodeGen/CGCleanup.h b/clang/lib/CodeGen/CGCleanup.h
index c73c97146abc4..03e4a29d7b3db 100644
--- a/clang/lib/CodeGen/CGCleanup.h
+++ b/clang/lib/CodeGen/CGCleanup.h
@@ -16,11 +16,8 @@
 #include "EHScopeStack.h"
 
 #include "Address.h"
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/IR/Instruction.h"
 
 namespace llvm {
 class BasicBlock;
@@ -269,51 +266,6 @@ class alignas(8) EHCleanupScope : public EHScope {
   };
   mutable struct ExtInfo *ExtInfo;
 
-  /// Erases auxillary allocas and their usages for an unused cleanup.
-  /// Cleanups should mark these allocas as 'used' if the cleanup is
-  /// emitted, otherwise these instructions would be erased.
-  struct AuxillaryAllocas {
-    SmallVector<llvm::Instruction *, 1> AuxAllocas;
-    bool used = false;
-
-    // Records a potentially unused instruction to be erased later.
-    void Add(llvm::AllocaInst *Alloca) { AuxAllocas.push_back(Alloca); }
-
-    // Mark all recorded instructions as used. These will not be erased later.
-    void MarkUsed() {
-      used = true;
-      AuxAllocas.clear();
-    }
-
-    ~AuxillaryAllocas() {
-      if (used)
-        return;
-      llvm::SetVector<llvm::Instruction *> Uses;
-      for (auto *Inst : llvm::reverse(AuxAllocas))
-        CollectUses(Inst, Uses);
-      // Delete uses in the reverse order of insertion.
-      for (auto *I : llvm::reverse(Uses))
-        I->eraseFromParent();
-    }
-
-  private:
-    void CollectUses(llvm::Instruction *I,
-                     llvm::SetVector<llvm::Instruction *> &Uses) {
-      if (!I || !Uses.insert(I))
-        return;
-      for (auto *User : I->users())
-        CollectUses(cast<llvm::Instruction>(User), Uses);
-    }
-  };
-  mutable struct AuxillaryAllocas *AuxAllocas;
-
-  AuxillaryAllocas &getAuxillaryAllocas() {
-    if (!AuxAllocas) {
-      AuxAllocas = new struct AuxillaryAllocas();
-    }
-    return *AuxAllocas;
-  }
-
   /// The number of fixups required by enclosing scopes (not including
   /// this one).  If this is the top cleanup scope, all the fixups
   /// from this index onwards belong to this scope.
@@ -346,7 +298,7 @@ class alignas(8) EHCleanupScope : public EHScope {
                  EHScopeStack::stable_iterator enclosingEH)
       : EHScope(EHScope::Cleanup, enclosingEH),
         EnclosingNormal(enclosingNormal), NormalBlock(nullptr),
-        ActiveFlag(Address::invalid()), ExtInfo(nullptr), AuxAllocas(nullptr),
+        ActiveFlag(Address::invalid()), ExtInfo(nullptr),
         FixupDepth(fixupDepth) {
     CleanupBits.IsNormalCleanup = isNormal;
     CleanupBits.IsEHCleanup = isEH;
@@ -360,15 +312,8 @@ class alignas(8) EHCleanupScope : public EHScope {
   }
 
   void Destroy() {
-    if (AuxAllocas)
-      delete AuxAllocas;
     delete ExtInfo;
   }
-  void AddAuxAllocas(llvm::SmallVector<llvm::AllocaInst *> Allocas) {
-    for (auto *Alloca : Allocas)
-      getAuxillaryAllocas().Add(Alloca);
-  }
-  void MarkEmitted() { getAuxillaryAllocas().MarkUsed(); }
   // Objects of EHCleanupScope are not destructed. Use Destroy().
   ~EHCleanupScope() = delete;
 
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 3f05ebb561da5..ce6d6d8956076 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -19,7 +19,6 @@
 #include "CodeGenFunction.h"
 #include "CodeGenModule.h"
 #include "ConstantEmitter.h"
-#include "EHScopeStack.h"
 #include "PatternInit.h"
 #include "TargetInfo.h"
 #include "clang/AST/ASTContext.h"
@@ -2202,27 +2201,6 @@ void CodeGenFunction::pushDestroy(CleanupKind cleanupKind, Address addr,
                                      destroyer, useEHCleanupForArray);
 }
 
-// Pushes a destroy and defers its deactivation until its
-// CleanupDeactivationScope is exited.
-void CodeGenFunction::pushDestroyAndDeferDeactivation(
-    QualType::DestructionKind dtorKind, Address addr, QualType type) {
-  assert(dtorKind && "cannot push destructor for trivial type");
-
-  CleanupKind cleanupKind = getCleanupKind(dtorKind);
-  pushDestroyAndDeferDeactivation(
-      cleanupKind, addr, type, getDestroyer(dtorKind), cleanupKind & EHCleanup);
-}
-
-void CodeGenFunction::pushDestroyAndDeferDeactivation(
-    CleanupKind cleanupKind, Address addr, QualType type, Destroyer *destroyer,
-    bool useEHCleanupForArray) {
-  llvm::Instruction *DominatingIP =
-      Builder.CreateFlagLoad(llvm::Constant::getNullValue(Int8PtrTy));
-  pushDestroy(cleanupKind, addr, type, destroyer, useEHCleanupForArray);
-  DeferredDeactivationCleanupStack.push_back(
-      {EHStack.stable_begin(), DominatingIP});
-}
-
 void CodeGenFunction::pushStackRestore(CleanupKind Kind, Address SPMem) {
   EHStack.pushCleanup<CallStackRestore>(Kind, SPMem);
 }
@@ -2239,19 +2217,16 @@ void CodeGenFunction::pushLifetimeExtendedDestroy(CleanupKind cleanupKind,
   // If we're not in a conditional branch, we don't need to bother generating a
   // conditional cleanup.
   if (!isInConditionalBranch()) {
+    // Push an EH-only cleanup for the object now.
     // FIXME: When popping normal cleanups, we need to keep this EH cleanup
     // around in case a temporary's destructor throws an exception.
+    if (cleanupKind & EHCleanup)
+      EHStack.pushCleanup<DestroyObject>(
+          static_cast<CleanupKind>(cleanupKind & ~NormalCleanup), addr, type,
+          destroyer, useEHCleanupForArray);
 
-    // Add the cleanup to the EHStack. After the full-expr, this would be
-    // deactivated before being popped from the stack.
-    pushDestroyAndDeferDeactivation(cleanupKind, addr, type, destroyer,
-                                    useEHCleanupForArray);
-
-    // Since this is lifetime-extended, push it once again to the EHStack after
-    // the full expression.
     return pushCleanupAfterFullExprWithActiveFlag<DestroyObject>(
-        cleanupKind, Address::invalid(), addr, type, destroyer,
-        useEHCleanupForArray);
+        cleanupKind, Address::invalid(), addr, type, destroyer, useEHCleanupForArray);
   }
 
   // Otherwise, we should only destroy the object if it's been initialized.
@@ -2266,12 +2241,13 @@ void CodeGenFunction::pushLifetimeExtendedDestroy(CleanupKind cleanupKind,
   Address ActiveFlag = createCleanupActiveFlag();
   SavedType SavedAddr = saveValueInCond(addr);
 
-  pushCleanupAndDeferDeactivation<ConditionalCleanupType>(
-      cleanupKind, SavedAddr, type, destroyer, useEHCleanupForArray);
-  initFullExprCleanupWithFlag(ActiveFlag);
+  if (cleanupKind & EHCleanup) {
+    EHStack.pushCleanup<ConditionalCleanupType>(
+        static_cast<CleanupKind>(cleanupKind & ~NormalCleanup), SavedAddr, type,
+        destroyer, useEHCleanupForArray);
+    initFullExprCleanupWithFlag(ActiveFlag);
+  }
 
-  // Since this is lifetime-extended, push it once again to the EHStack after
-  // the full expression.
   pushCleanupAfterFullExprWithActiveFlag<ConditionalCleanupType>(
       cleanupKind, ActiveFlag, SavedAddr, type, destroyer,
       useEHCleanupForArray);
@@ -2466,9 +2442,9 @@ namespace {
   };
 } // end anonymous namespace
 
-/// pushIrregularPartialArrayCleanup - Push a NormalAndEHCleanup to
-/// destroy already-constructed elements of the given array.  The cleanup may be
-/// popped with DeactivateCleanupBlock or PopCleanupBlock.
+/// pushIrregularPartialArrayCleanup - Push an EH cleanup to destroy
+/// already-constructed elements of the given array.  The cleanup
+/// may be popped with DeactivateCleanupBlock or PopCleanupBlock.
 ///
 /// \param elementType - the immediate element type of the array;
 ///   possibly still an array type
@@ -2477,9 +2453,10 @@ void CodeGenFunction::pushIrregularPartialArrayCleanup(llvm::Value *arrayBegin,
                                                        QualType elementType,
                                                        CharUnits elementAlign,
                                                        Destroyer *destroyer) {
-  pushFullExprCleanup<IrregularPartialArrayDestroy>(
-      NormalAndEHCleanup, arrayBegin, arrayEndPointer, elementType,
-      elementAlign, destroyer);
+  pushFullExprCleanup<IrregularPartialArrayDestroy>(EHCleanup,
+                                                    arrayBegin, arrayEndPointer,
+                                                    elementType, elementAlign,
+                                                    destroyer);
 }
 
 /// pushRegularPartialArrayCleanup - Push an EH cleanup to destroy
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index c85a339f5e3f8..cf696a1c9f560 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -115,16 +115,10 @@ RawAddress CodeGenFunction::CreateTempAlloca(llvm::Type *Ty, CharUnits Align,
 llvm::AllocaInst *CodeGenFunction::CreateTempAlloca(llvm::Type *Ty,
                                                     const Twine &Name,
                                                     llvm::Value *ArraySize) {
-  llvm::AllocaInst *Alloca;
   if (ArraySize)
-    Alloca = Builder.CreateAlloca(Ty, ArraySize, Name);
-  else
-    Alloca = new llvm::AllocaInst(Ty, CGM.getDataLayout().getAllocaAddrSpace(),
-                                  ArraySize, Name, AllocaInsertPt);
-  if (Allocas) {
-    Allocas->Add(Alloca);
-  }
-  return Alloca;
+    return Builder.CreateAlloca(Ty, ArraySize, Name);
+  return new llvm::AllocaInst(Ty, CGM.getDataLayout().getAllocaAddrSpace(),
+                              ArraySize, Name, AllocaInsertPt);
 }
 
 /// CreateDefaultAlignTempAlloca - This creates an alloca with the
diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
index 560a9e2c5ead5..1b9287ea23934 100644
--- a/clang/lib/CodeGen/CGExprAgg.cpp
+++ b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -15,7 +15,6 @@
 #include "CodeGenFunction.h"
 #include "CodeGenModule.h"
 #include "ConstantEmitter.h"
-#include "EHScopeStack.h"
 #include "TargetInfo.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/Attr.h"
@@ -25,7 +24,6 @@
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalVariable.h"
-#include "llvm/IR/Instruction.h"
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Intrinsics.h"
 using namespace clang;
@@ -560,27 +558,24 @@ void AggExprEmitter::EmitArrayInit(Address DestPtr, llvm::ArrayType *AType,
   // For that, we'll need an EH cleanup.
   QualType::DestructionKind dtorKind = elementType.isDestructedType();
   Address endOfInit = Address::invalid();
-  CodeGenFunction::CleanupDeactivationScope deactivation(CGF);
-
-  if (dtorKind) {
-    CodeGenFunction::AllocaTrackerRAII allocaTracker(CGF);
+  EHScopeStack::stable_iterator cleanup;
+  llvm::Instruction *cleanupDominator = nullptr;
+  if (CGF.needsEHCleanup(dtorKind)) {
     // In principle we could tell the cleanup where we are more
     // directly, but the control flow can get so varied here that it
     // would actually be quite complex.  Therefore we go through an
     // alloca.
-    llvm::Instruction *dominatingIP =
-        Builder.CreateFlagLoad(llvm::ConstantInt::getNullValue(CGF.Int8PtrTy));
     endOfInit = CGF.CreateTempAlloca(begin->getType(), CGF.getPointerAlign(),
                                      "arrayinit.endOfInit");
-    Builder.CreateStore(begin, endOfInit);
+    cleanupDominator = Builder.CreateStore(begin, endOfInit);
     CGF.pushIrregularPartialArrayCleanup(begin, endOfInit, elementType,
                                          elementAlign,
                                          CGF.getDestroyer(dtorKind));
-    cast<EHCleanupScope>(*CGF.EHStack.find(CGF.EHStack.stable_begin()))
-        .AddAuxAllocas(allocaTracker.Take());
+    cleanup = CGF.EHStack.stable_begin();
 
-    CGF.DeferredDeactivationCleanupStack.push_back(
-        {CGF.EHStack.stable_begin(), dominatingIP});
+  // Otherwise, remember that we didn't need a cleanup.
+  } else {
+    dtorKind = QualType::DK_none;
   }
 
   llvm::Value *one = llvm::ConstantInt::get(CGF.SizeTy, 1);
@@ -676,6 +671,9 @@ void AggExprEmitter::EmitArrayInit(Address DestPtr, llvm::ArrayType *AType,
 
     CGF.EmitBlock(endBB);
   }
+
+  // Leave the partial-array cleanup if we entered one.
+  if (dtorKind) CGF.DeactivateCleanupBlock(cleanup, cleanupDominator);
 }
 
 //===----------------------------------------------------------------------===//
@@ -1376,8 +1374,9 @@ AggExprEmitter::VisitLambdaExpr(LambdaExpr *E) {
   LValue SlotLV = CGF.MakeAddrLValue(Slot.getAddress(), E->getType());
 
   // We'll need to enter cleanup scopes in case any of the element
-  // initializers throws an exception or contains branch out of the expressions.
-  CodeGenFunction::CleanupDeactivationScope scope(CGF);
+  // initializers throws an exception.
+  SmallVector<EHScopeStack::stable_iterator, 16> Cleanups;
+  llvm::Instruction *CleanupDominator = nullptr;
 
   CXXRecordDecl::field_iterator CurField = E->getLambdaClass()->field_begin();
   for (LambdaExpr::const_capture_init_iterator i = E->capture_init_begin(),
@@ -1396,12 +1395,28 @@ AggExprEmitter::VisitLambdaExpr(LambdaExpr *E) {
     if (QualType::DestructionKind DtorKind =
             CurField->getType().isDestructedType()) {
       assert(LV.isSimple());
-      if (DtorKind)
-        CGF.pushDestroyAndDeferDeactivation(
-            NormalAndEHCleanup, LV.getAddress(CGF), CurField->getType(),
-            CGF.getDestroyer(DtorKind), false);
+      if (CGF.needsEHCleanup(DtorKind)) {
+        if (!CleanupDominator)
+          CleanupDominator = CGF.Builder.CreateAlignedLoad(
+              CGF.Int8Ty,
+              llvm::Constant::getNullValue(CGF.Int8PtrTy),
+              CharUnits::One()); // placeholder
+
+        CGF.pushDestroy(EHCleanup, LV.getAddress(CGF), CurField->getType(),
+                        CGF.getDestroyer(DtorKind), false);
+        Cleanups.push_back(CGF.EHStack.stable_begin());
+      }
     }
   }
+
+  // Deactivate all the partial cleanups in reverse order, which
+  // generally means popping them.
+  for (unsigned i = Cleanups.size(); i != 0; --i)
+    CGF.DeactivateCleanupBlock(Cleanups[i-1], CleanupDominator);
+
+  // Destroy the placeholder if we made one.
+  if (CleanupDominator)
+    CleanupDominator->eraseFromParent();
 }
 
 void AggExprEmitter::VisitExprWithCleanups(ExprWithCleanups *E) {
@@ -1690,7 +1705,14 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
   // We'll need to enter cleanup scopes in case any of the element
   // initializers throws an exception.
   SmallVector<EHScopeStack::stable_iterator, 16> cleanups;
-  CodeGenFunction::CleanupDeactivationScope DeactivateCleanups(CGF);
+  llvm::Instruction *cleanupDominator = nullptr;
+  auto addCleanup = [&](const EHScopeStack::stable_iterator &cleanup) {
+    cleanups.push_back(cleanup);
+    if (!cleanupDominator) // create placeholder once needed
+      cleanupDominator = CGF.Builder.CreateAlignedLoad(
+          CGF.Int8Ty, llvm::Constant::getNullValue(CGF.Int8PtrTy),
+          CharUnits::One());
+  };
 
   unsigned curInitIndex = 0;
 
@@ -1713,8 +1735,10 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
       CGF.EmitAggExpr(InitExprs[curInitIndex++], AggSlot);
 
       if (QualType::DestructionKind dtorKind =
-              Base.getType().isDestructedType())
-        CGF.pushDestroyAndDeferDeactivation(dtorKind, V, Base.getType());
+              Base.getType().isDestructedType()) {
+        CGF.pushDestroy(dtorKind, V, Base.getType());
+        addCleanup(CGF.EHStack.stable_begin());
+      }
     }
   }
 
@@ -1789,10 +1813,10 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
     if (QualType::DestructionKind dtorKind
           = field->getType().isDestructedType()) {
       assert(LV.isSimple());
-      if (dtorKind) {
-        CGF.pushDestroyAndDeferDeactivation(
-            NormalAndEHCleanup, LV.getAddress(CGF), field->getType(),
-            CGF.getDestroyer(dtorKind), false);
+      if (CGF.needsEHCleanup(dtorKind)) {
+        CGF.pushDestroy(EHCleanup, LV.getAddress(CGF), field->getType(),
+                        CGF.getDestroyer(dtorKind), false);
+        addCleanup(CGF.EHStack.stable_begin());
         pushedCleanup = true;
       }
     }
@@ -1805,6 +1829,17 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
         if (GEP->use_empty())
           GEP->eraseFromParent();
   }
+
+  // Deactivate all the partial cleanups in reverse order, which
+  // generally means popping them.
+  assert((cleanupDominator || cleanups.empty()) &&
+         "Missing cleanupDominator before deactivating cleanup blocks");
+  for (unsigned i = cleanups.size(); i != 0; --i)
+    CGF.DeactivateCleanupBlock(cleanups[i-1], cleanupDominator);
+
+  // Destroy the placeholder if we made one.
+  if (cleanupDominator)
+    cleanupDominator->eraseFromParent();
 }
 
 void AggExprEmitter::VisitArrayInitLoopExpr(const ArrayInitLoopExpr *E,
diff --git a/clang/lib/CodeGen/CGExprCXX.cpp b/clang/lib/CodeGen/CGExprCXX.cpp
index a88b29b326bb9..a4fb673284cec 100644
--- a/clang/lib/CodeGen/CGExprCXX.cpp
+++ b/clang/lib/CodeGen/CGExprCXX.cpp
@@ -1008,8 +1008,8 @@ void CodeGenFunction::EmitNewArrayInitializer(
   const Expr *Init = E->getInitializer();
   Address EndOfInit = Address::invalid();
   QualType::DestructionKind DtorKind = ElementType.isDestructedType();
-  CleanupDeactivationScope deactivation(*this);
-  bool pushedCleanup = false;
+  EHScopeStack::stable_iterator Cleanup;
+  llvm::Instruction *CleanupDominator = nullptr;
 
   CharUnits ElementSize = getContext().getTypeSizeInChars(ElementType);
   CharUnits ElementAlign =
@@ -1105,24 +1105,19 @@ void CodeGenFunction::EmitNewArrayInitializer(
     }
 
     // Enter a partial-destruction Cleanup if necessary.
-    if (DtorKind) {
-      AllocaTrackerRAII AllocaTracker(*this);
+    if (needsEHCleanup(DtorKind)) {
       // In principle we could tell the Cleanup where we are more
       // directly, but the control flow can get so varied here that it
       // would actually be quite complex.  Therefore we go through an
       // alloca.
-      llvm::Instruction *DominatingIP =
-          Builder.CreateFlagLoad(llvm::ConstantInt::getNullValue(Int8PtrTy));
       EndOfInit = CreateTempAlloca(BeginPtr.getType(), getPointerAlign(),
                                    "array.init.end");
+      CleanupDominator =
+          Builder.CreateStore(BeginPtr.emitRawPointer(*this), EndOfInit);
       pushIrregularPartialArrayCleanup(BeginPtr.emitRawPointer(*this),
                                        EndOfInit, ElementType, ElementAlign,
                                        getDestroyer(DtorKind));
-      cast<EHCleanupScope>(*EHStack.find(EHStack.stable_begin()))
-          .AddAuxAllocas(AllocaTracker.Take());
-      DeferredDeactivationCleanupStack.push_back(
-          {EHStack.stable_begin(), DominatingIP});
-      pushedCleanup = true;
+      Cleanup = EHStack.stable_begin();
     }
 
     CharUnits StartAlign = CurPtr.getAlignment();
@@ -1169,6 +1164,9 @@ void CodeGenFunction::EmitNewArrayInitializer(
   // initialization.
   llvm::ConstantInt *ConstNum = dyn_cast<llvm::ConstantInt>(NumElements);
   if (ConstNum && ConstNum->getZExtValue() <= InitListElements) {
+    // If there was a Cleanup, deactivate it.
+    if (CleanupDominator)
+      DeactivateCleanupBlock(Cleanup, CleanupDominator);
     return;
   }
 
@@ -1283,14 +1281,13 @@ void CodeGenFunction::EmitNewArrayInitializer(
     Builder.CreateStore(CurPtr.emitRawPointer(*this), EndOfInit);
 
   // Enter a partial-destruction Cleanup if necessary.
-  if (!pushedCleanup && needsEHCleanup(DtorKind)) {
-    llvm::Instruction *DominatingIP =
-        Builder.CreateFlagLoad(llvm::ConstantInt::getNullValue(Int8PtrTy));
-    pushRegularPartialArrayCleanup(BeginPtr.emitRawPointer(*this),
-                                   CurPtr.emitRawPointer(*this), ElementType,
+  if (!CleanupDominator && needsEHCleanup(DtorKind)) {
+    llvm::Value *BeginPtrRaw = BeginPtr.emitRawPointer(*this);
+    llvm::Value *CurPtrRaw = CurPtr.emitRawPointer(*this);
+    pushRegularPartialArrayCleanup(BeginPtrRaw, CurPtrRaw, ElementType,
                                    ElementAlign, getDestroyer(DtorKind));
-    DeferredDeactivationCleanupStack.push_back(
-        {EHStack.stable_begin(), DominatingIP});
+    Cleanup = EHStack.stable_begin();
+    CleanupDominator = Builder.CreateUnreachable();
   }
 
   // Emit the initializer into this element.
@@ -1298,7 +1295,10 @@ void CodeGenFunction::EmitNewArrayInitializer(
                           AggValueSlot::DoesNotOverlap);
 
   // Leave the Cleanup if we entered one.
-  deactivation.ForceDeactivate();
+  if (CleanupDominator) {
+    DeactivateCleanupBlock(Cleanup, CleanupDominator);
+    CleanupDominator->eraseFromParent();
+  }
 
   // Advance to the next element by adjusting the pointer type as necessary.
   llvm::Value *NextPtr = Builder.CreateConstInBoundsGEP1_32(
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 87766a758311d..86a6ddd80cc11 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -91,8 +91,6 @@ CodeGenFunction::CodeGenFunction(CodeGenModule &cgm, bool suppressNewContext)
 
 CodeGenFunction::~CodeGenFunction() {
   assert(LifetimeExtendedCleanupStack.empty() && "failed to emit a cleanup");
-  assert(DeferredDeactivationCleanupStack.empty() &&
-         "missed to deactivate a cleanup");
 
   if (getLangOpts().OpenMP && CurFn)
     CGM.getOpenMPRuntime().functionFinished(*this);
@@ -348,10 +346,6 @@ static void EmitIfUsed(CodeGenFunction &CGF, llvm::BasicBlock *BB) {
 void CodeGenFunction::FinishFunction(SourceLocation EndLoc) {
   assert(BreakContinueStack.empty() &&
          "mismatched push/pop in break/continue stack!");
-  assert(LifetimeExtendedCleanupStack.empty() &&
-         "mismatched push/pop of cleanups in EHStack!");
-  assert(DeferredDeactivationCleanupStack.empty() &&
-         "mismatched activate/deactivate of cleanups!");
 
   bool OnlySimpleReturnStmts = NumSimpleReturnExprs > 0
     && NumSimpleReturnExprs == NumReturnExprs
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index d99188671f1f6..ff1873325d409 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -39,7 +39,6 @@
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Frontend/OpenMP/OMPIRBuilder.h"
-#include "llvm/IR/Instructions.h"
 #include "llvm/IR/ValueHandle.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Transforms/Utils/SanitizerStats.h"
@@ -671,51 +670,6 @@ class CodeGenFunction : public CodeGenTypeCache {
 
   EHScopeStack EHStack;
   llvm::SmallVector<char, 256> LifetimeExtendedCleanupStack;
-
-  // A stack of cleanups which were added to EHStack but have to be deactivated
-  // later before being popped or emitted. These are usually deactivated on
-  // exiting a `CleanupDeactivationScope` scope. For instance, after a
-  // full-expr.
-  //
-  // These are specially useful for correctly emitting cleanups while
-  // encountering branches out of expression (through stmt-expr or coroutine
-  // suspensions).
-  struct DeferredDeactivateCleanup {
-    EHScopeStack::stable_iterator Cleanup;
-    llvm::Instruction *DominatingIP;
-  };
-  llvm::SmallVector<DeferredDeactivateCleanup> DeferredDeactivationCleanupStack;
-
-  // Enters a new scope for capturing cleanups which are deferred to be
-  // deactivated, all of which will be deactivated once the scope is exited.
-  struct CleanupDeactivationScope {
-    CodeGenFunction &CGF;
-    size_t OldDeactivateCleanupStackSize;
-    bool Deactivated;
-    CleanupDeactivationScope(CodeGenFunction &CGF)
-        : CGF(CGF), OldDeactivateCleanupStackSize(
-                        CGF.DeferredDeactivationCleanupStack.size()),
-          Deactivated(false) {}
-
-    void ForceDeactivate() {
-      assert(!Deactivated && "Deactivating already deactivated scope");
-      auto &Stack = CGF.DeferredDeactivationCleanupStack;
-      for (size_t I = Stack.size(); I > OldDeactivateCleanupStackSize; I--) {
-        CGF.DeactivateCleanupBlock(Stack[I - 1].Cleanup,
-                                   Stack[I - 1].DominatingIP);
-        Stack[I - 1].DominatingIP->eraseFromParent();
-      }
-      Stack.resize(OldDeactivateCleanupStackSize);
-      Deactivated = true;
-    }
-
-    ~CleanupDeactivationScope() {
-      if (Deactivated)
-        return;
-      ForceDeactivate();
-    }
-  };
-
   llvm::SmallVector<const JumpDest *, 2> SEHTryEpilogueStack;
 
   llvm::Instruction *CurrentFuncletPad = nullptr;
@@ -921,19 +875,6 @@ class CodeGenFunction : public CodeGenTypeCache {
       new (Buffer + sizeof(Header) + sizeof(T)) RawAddress(ActiveFlag);
   }
 
-  // Push a cleanup onto EHStack and deactivate it later. It is usually
-  // deactivated when exiting a `CleanupDeactivationScope` (for example: after a
-  // full expression).
-  template <class T, class... As>
-  void pushCleanupAndDeferDeactivation(CleanupKind Kind, As... A) {
-    // Placeholder dominating IP for this cleanup.
-    llvm::Instruction *DominatingIP =
-        Builder.CreateFlagLoad(llvm::Constant::getNullValue(Int8PtrTy));
-    EHStack.pushCleanup<T>(Kind, A...);
-    DeferredDeactivationCleanupStack.push_back(
-        {EHStack.stable_begin(), DominatingIP});
-  }
-
   /// Set up the last cleanup that was pushed as a conditional
   /// full-expression cleanup.
   void initFullExprCleanup() {
@@ -957,8 +898,7 @@ class CodeGenFunction : public CodeGenTypeCache {
 
   /// PopCleanupBlock - Will pop the cleanup entry on the stack and
   /// process all branch fixups.
-  void PopCleanupBlock(bool FallThroughIsBranchThrough = false,
-                       bool ForDeactivation = false);
+  void PopCleanupBlock(bool FallThroughIsBranchThrough = false);
 
   /// DeactivateCleanupBlock - Deactivates the given cleanup block.
   /// The block cannot be reactivated.  Pops it if it's the top of the
@@ -986,7 +926,6 @@ class CodeGenFunction : public CodeGenTypeCache {
   class RunCleanupsScope {
     EHScopeStack::stable_iterator CleanupStackDepth, OldCleanupScopeDepth;
     size_t LifetimeExtendedCleanupStackSize;
-    CleanupDeactivationScope DeactivateCleanups;
     bool OldDidCallStackSave;
   protected:
     bool PerformCleanup;
@@ -1001,7 +940,8 @@ class CodeGenFunction : public CodeGenTypeCache {
   public:
     /// Enter a new cleanup scope.
     explicit RunCleanupsScope(CodeGenFunction &CGF)
-        : DeactivateCleanups(CGF), PerformCleanup(true), CGF(CGF) {
+      : PerformCleanup(true), CGF(CGF)
+    {
       CleanupStackDepth = CGF.EHStack.stable_begin();
       LifetimeExtendedCleanupStackSize =
           CGF.LifetimeExtendedCleanupStack.size();
@@ -1031,7 +971,6 @@ class CodeGenFunction : public CodeGenTypeCache {
     void ForceCleanup(std::initializer_list<llvm::Value**> ValuesToReload = {}) {
       assert(PerformCleanup && "Already forced cleanup");
       CGF.DidCallStackSave = OldDidCallStackSave;
-      DeactivateCleanups.ForceDeactivate();
       CGF.PopCleanupBlocks(CleanupStackDepth, LifetimeExtendedCleanupStackSize,
                            ValuesToReload);
       PerformCleanup = false;
@@ -2221,11 +2160,6 @@ class CodeGenFunction : public CodeGenTypeCache {
                      Address addr, QualType type);
   void pushDestroy(CleanupKind kind, Address addr, QualType type,
                    Destroyer *destroyer, bool useEHCleanupForArray);
-  void pushDestroyAndDeferDeactivation(QualType::DestructionKind dtorKind,
-                                       Address addr, QualType type);
-  void pushDestroyAndDeferDeactivation(CleanupKind cleanupKind, Address addr,
-                                       QualType type, Destroyer *destroyer,
-                                       bool useEHCleanupForArray);
   void pushLifetimeExtendedDestroy(CleanupKind kind, Address addr,
                                    QualType type, Destroyer *destroyer,
                                    bool useEHCleanupForArray);
@@ -2764,33 +2698,6 @@ class CodeGenFunction : public CodeGenTypeCache {
                             TBAAAccessInfo *TBAAInfo = nullptr);
   LValue EmitLoadOfPointerLValue(Address Ptr, const PointerType *PtrTy);
 
-private:
-  struct AllocaTracker {
-    void Add(llvm::AllocaInst *I) { Allocas.push_back(I); }
-    llvm::SmallVector<llvm::AllocaInst *> Take() { return std::move(Allocas); }
-
-  private:
-    llvm::SmallVector<llvm::AllocaInst *> Allocas;
-  };
-  AllocaTracker *Allocas = nullptr;
-
-public:
-  // Captures all the allocas created during the scope of its RAII object.
-  struct AllocaTrackerRAII {
-    AllocaTrackerRAII(CodeGenFunction &CGF)
-        : CGF(CGF), OldTracker(CGF.Allocas) {
-      CGF.Allocas = &Tracker;
-    }
-    ~AllocaTrackerRAII() { CGF.Allocas = OldTracker; }
-
-    llvm::SmallVector<llvm::AllocaInst *> Take() { return Tracker.Take(); }
-
-  private:
-    CodeGenFunction &CGF;
-    AllocaTracker *OldTracker;
-    AllocaTracker Tracker;
-  };
-
   /// CreateTempAlloca - This creates an alloca and inserts it into the entry
   /// block if \p ArraySize is nullptr, otherwise inserts it at the current
   /// insertion point of the builder. The caller is responsible for setting an
diff --git a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp b/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
deleted file mode 100644
index 0a51b0e4121c3..0000000000000
--- a/clang/test/CodeGenCXX/control-flow-in-stmt-expr.cpp
+++ /dev/null
@@ -1,409 +0,0 @@
-// RUN: %clang_cc1 --std=c++20 -fexceptions -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefixes=EH %s
-// RUN: %clang_cc1 --std=c++20 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck -check-prefixes=NOEH,CHECK %s
-
-struct Printy {
-  Printy(const char *name) : name(name) {}
-  ~Printy() {}
-  const char *name;
-};
-
-int foo() { return 2; }
-
-struct Printies {
-  Printy a;
-  Printy b;
-  Printy c;
-};
-
-void ParenInit() {
-  // CHECK-LABEL: define dso_local void @_Z9ParenInitv()
-  // CHECK: [[CLEANUP_DEST:%.+]] = alloca i32, align 4
-  Printies ps(Printy("a"), 
-              // CHECK: call void @_ZN6PrintyC1EPKc
-              ({
-                if (foo()) return;
-                // CHECK:     if.then:
-                // CHECK-NEXT:   store i32 1, ptr [[CLEANUP_DEST]], align 4
-                // CHECK-NEXT:   br label %cleanup
-                Printy("b");
-                // CHECK:     if.end:
-                // CHECK-NEXT:  call void @_ZN6PrintyC1EPKc
-              }),
-              ({
-                if (foo()) return;
-                // CHECK:     if.then{{.*}}:
-                // CHECK-NEXT:  store i32 1, ptr [[CLEANUP_DEST]], align 4
-                // CHECK-NEXT:  call void @_ZN6PrintyD1Ev
-                // CHECK-NEXT:  br label %cleanup
-                Printy("c");
-                // CHECK:     if.end{{.*}}:
-                // CHECK-NEXT:  call void @_ZN6PrintyC1EPKc
-                // CHECK-NEXT:  call void @_ZN8PrintiesD1Ev
-                // CHECK-NEXT:  br label %return
-              }));
-  // CHECK:     cleanup:
-  // CHECK-NEXT:  call void @_ZN6PrintyD1Ev
-  // CHECK-NEXT:  br label %return
-}
-
-void break_in_stmt_expr() {
-  // Verify that the "break" in "if.then".calls dtor before jumping to "for.end".
-
-  // CHECK-LABEL: define dso_local void @_Z18break_in_stmt_exprv()
-  Printies p{Printy("a"), 
-            // CHECK: call void @_ZN6PrintyC1EPKc
-            ({
-                for (;;) {
-                    Printies ps{
-                      Printy("b"), 
-                      // CHECK: for.cond:
-                      // CHECK:   call void @_ZN6PrintyC1EPKc
-                      ({
-                        if (foo()) {
-                          break;
-                          // CHECK:       if.then:
-                          // CHECK-NEXT:    call void @_ZN6PrintyD1Ev
-                          // CHECK-NEXT:    br label %for.end
-                        }
-                        Printy("c");
-                        // CHECK:       if.end:
-                        // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-                      }),
-                      Printy("d")};
-                      // CHECK:           call void @_ZN6PrintyC1EPKc
-                      // CHECK-NEXT:      call void @_ZN8PrintiesD1Ev
-                      // CHECK-NEXT:      br label %for.cond
-                }
-                Printy("e");
-  // CHECK:       for.end:
-  // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-              }),
-              Printy("f")};
-  // CHECK:         call void @_ZN6PrintyC1EPKc
-  // CHECK-NEXT:    call void @_ZN8PrintiesD1Ev
-}
-
-void goto_in_stmt_expr() {
-  // Verify that:
-  //  - correct branch fixups for deactivated normal cleanups are generated correctly.
-
-  // CHECK-LABEL: define dso_local void @_Z17goto_in_stmt_exprv()
-  // CHECK: [[CLEANUP_DEST_SLOT:%cleanup.dest.slot.*]] = alloca i32, align 4
-  {
-    Printies p1{Printy("a"), // CHECK: call void @_ZN6PrintyC1EPKc
-                ({
-                  {
-                    Printies p2{Printy("b"),
-                                // CHECK: call void @_ZN6PrintyC1EPKc
-                                ({
-                                  if (foo() == 1) {
-                                    goto in;
-                                    // CHECK:       if.then:
-                                    // CHECK-NEXT:    store i32 2, ptr [[CLEANUP_DEST_SLOT]], align 4
-                                    // CHECK-NEXT:    br label %[[CLEANUP1:.+]]
-                                  }
-                                  if (foo() == 2) {
-                                    goto out;
-                                    // CHECK:       if.then{{.*}}:
-                                    // CHECK-NEXT:    store i32 3, ptr [[CLEANUP_DEST_SLOT]], align 4
-                                    // CHECK-NEXT:    br label %[[CLEANUP1]]
-                                  }
-                                  Printy("c");
-                                  // CHECK:       if.end{{.*}}:
-                                  // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-                                }),
-                                Printy("d")};
-                                // CHECK:           call void @_ZN6PrintyC1EPKc
-                                // CHECK-NEXT:      call void @_ZN8PrintiesD1Ev
-                                // CHECK-NEXT:      br label %in
-
-                  }
-                in:
-                  Printy("e");
-                // CHECK:       in:                                               ; preds = %if.end{{.*}}, %[[CLEANUP1]]
-                // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-                }),
-                Printy("f")};
-                // CHECK:         call void @_ZN6PrintyC1EPKc
-                // CHECK-NEXT:    call void @_ZN8PrintiesD1Ev
-                // CHECK-NEXT:    br label %out
-  }
-out:
-  return;
-  // CHECK:       out:
-  // CHECK-NEXT:    ret void
-
-  // CHECK:       [[CLEANUP1]]:                                          ; preds = %if.then{{.*}}, %if.then
-  // CHECK-NEXT:    call void @_ZN6PrintyD1Ev
-  // CHECK-NEXT:    %cleanup.dest = load i32, ptr [[CLEANUP_DEST_SLOT]], align 4
-  // CHECK-NEXT:    switch i32 %cleanup.dest, label %[[CLEANUP2:.+]] [
-  // CHECK-NEXT:      i32 2, label %in
-  // CHECK-NEXT:    ]
-
-  // CHECK:       [[CLEANUP2]]:                                         ; preds = %[[CLEANUP1]]
-  // CHECK-NEXT:    call void @_ZN6PrintyD1Ev
-  // CHECK-NEXT:    %cleanup.dest{{.*}} = load i32, ptr [[CLEANUP_DEST_SLOT]], align 4
-  // CHECK-NEXT:    switch i32 %cleanup.dest{{.*}}, label %unreachable [
-  // CHECK-NEXT:      i32 3, label %out
-  // CHECK-NEXT:    ]
-}
-
-void ArrayInit() {
-  // Printy arr[4] = {ctorA, ctorB, stmt-exprC, stmt-exprD};
-  // Verify that:
-  //  - We do the necessary stores for array cleanups (endOfInit and last constructed element).
-  //  - We update the array init element correctly for ctorA, ctorB and stmt-exprC.
-  //  - stmt-exprC and stmt-exprD share the array body dtor code (see %cleanup).
-
-  // CHECK-LABEL: define dso_local void @_Z9ArrayInitv()
-  // CHECK: %arrayinit.endOfInit = alloca ptr, align 8
-  // CHECK: %cleanup.dest.slot = alloca i32, align 4
-  // CHECK: %arrayinit.begin = getelementptr inbounds [4 x %struct.Printy], ptr %arr, i64 0, i64 0
-  // CHECK: store ptr %arrayinit.begin, ptr %arrayinit.endOfInit, align 8
-  Printy arr[4] = {
-    Printy("a"),
-    // CHECK: call void @_ZN6PrintyC1EPKc(ptr noundef nonnull align 8 dereferenceable(8) %arrayinit.begin, ptr noundef @.str)
-    // CHECK: [[ARRAYINIT_ELEMENT1:%.+]] = getelementptr inbounds %struct.Printy, ptr %arrayinit.begin, i64 1
-    // CHECK: store ptr [[ARRAYINIT_ELEMENT1]], ptr %arrayinit.endOfInit, align 8
-    Printy("b"),
-    // CHECK: call void @_ZN6PrintyC1EPKc(ptr noundef nonnull align 8 dereferenceable(8) [[ARRAYINIT_ELEMENT1]], ptr noundef @.str.1)
-    // CHECK: [[ARRAYINIT_ELEMENT2:%.+]] = getelementptr inbounds %struct.Printy, ptr [[ARRAYINIT_ELEMENT1]], i64 1
-    // CHECK: store ptr [[ARRAYINIT_ELEMENT2]], ptr %arrayinit.endOfInit, align 8
-    ({
-    // CHECK: br i1 {{.*}}, label %if.then, label %if.end
-      if (foo()) {
-        return;
-      // CHECK:       if.then:
-      // CHECK-NEXT:    store i32 1, ptr %cleanup.dest.slot, align 4
-      // CHECK-NEXT:    br label %cleanup
-      }
-      // CHECK:       if.end:
-      Printy("c");
-      // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-      // CHECK-NEXT:    %arrayinit.element2 = getelementptr inbounds %struct.Printy, ptr %arrayinit.element1, i64 1
-      // CHECK-NEXT:    store ptr %arrayinit.element2, ptr %arrayinit.endOfInit, align 8
-    }),
-    ({
-    // CHECK: br i1 {{%.+}} label %[[IF_THEN2:.+]], label %[[IF_END2:.+]]
-      if (foo()) {
-        return;
-      // CHECK:       [[IF_THEN2]]:
-      // CHECK-NEXT:    store i32 1, ptr %cleanup.dest.slot, align 4
-      // CHECK-NEXT:    br label %cleanup
-      }
-      // CHECK:       [[IF_END2]]:
-      Printy("d");
-      // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-      // CHECK-NEXT:    %array.begin = getelementptr inbounds [4 x %struct.Printy], ptr %arr, i32 0, i32 0
-      // CHECK-NEXT:    %0 = getelementptr inbounds %struct.Printy, ptr %array.begin, i64 4
-      // CHECK-NEXT:    br label %[[ARRAY_DESTROY_BODY1:.+]]
-  }),
-  };
-
-  // CHECK:       [[ARRAY_DESTROY_BODY1]]:
-  // CHECK-NEXT:    %arraydestroy.elementPast{{.*}} = phi ptr [ %0, %[[IF_END2]] ], [ %arraydestroy.element{{.*}}, %[[ARRAY_DESTROY_BODY1]] ]
-  // CHECK-NEXT:    %arraydestroy.element{{.*}} = getelementptr inbounds %struct.Printy, ptr %arraydestroy.elementPast{{.*}}, i64 -1
-  // CHECK-NEXT:    call void @_ZN6PrintyD1Ev
-  // CHECK-NEXT:    %arraydestroy.done{{.*}} = icmp eq ptr %arraydestroy.element{{.*}}, %array.begin
-  // CHECK-NEXT:    br i1 %arraydestroy.done{{.*}}, label %[[ARRAY_DESTROY_DONE1:.+]], label %[[ARRAY_DESTROY_BODY1]]
-
-  // CHECK:       [[ARRAY_DESTROY_DONE1]]:
-  // CHECK-NEXT:    ret void
-
-  // CHECK:       cleanup:
-  // CHECK-NEXT:    %1 = load ptr, ptr %arrayinit.endOfInit, align 8
-  // CHECK-NEXT:    %arraydestroy.isempty = icmp eq ptr %arrayinit.begin, %1
-  // CHECK-NEXT:    br i1 %arraydestroy.isempty, label %[[ARRAY_DESTROY_DONE2:.+]], label %[[ARRAY_DESTROY_BODY2:.+]]
-
-  // CHECK:       [[ARRAY_DESTROY_BODY2]]:
-  // CHECK-NEXT:    %arraydestroy.elementPast = phi ptr [ %1, %cleanup ], [ %arraydestroy.element, %[[ARRAY_DESTROY_BODY2]] ]
-  // CHECK-NEXT:    %arraydestroy.element = getelementptr inbounds %struct.Printy, ptr %arraydestroy.elementPast, i64 -1
-  // CHECK-NEXT:    call void @_ZN6PrintyD1Ev(ptr noundef nonnull align 8 dereferenceable(8) %arraydestroy.element)
-  // CHECK-NEXT:    %arraydestroy.done = icmp eq ptr %arraydestroy.element, %arrayinit.begin
-  // CHECK-NEXT:    br i1 %arraydestroy.done, label %[[ARRAY_DESTROY_DONE2]], label %[[ARRAY_DESTROY_BODY2]]
-
-  // CHECK:       [[ARRAY_DESTROY_DONE2]]:
-  // CHECK-NEXT:    br label %[[ARRAY_DESTROY_DONE1]]
-}
-
-void ArraySubobjects() {
-  struct S {
-    Printy arr1[2];
-    Printy arr2[2];
-    Printy p;
-  };
-  // CHECK-LABEL: define dso_local void @_Z15ArraySubobjectsv()
-  // CHECK: %arrayinit.endOfInit = alloca ptr, align 8
-  S s{{Printy("a"), Printy("b")},
-      // CHECK: call void @_ZN6PrintyC1EPKc
-      // CHECK: call void @_ZN6PrintyC1EPKc
-      {Printy("a"),
-      // CHECK: [[ARRAYINIT_BEGIN:%.+]] = getelementptr inbounds [2 x %struct.Printy]
-      // CHECK: store ptr [[ARRAYINIT_BEGIN]], ptr %arrayinit.endOfInit, align 8
-      // CHECK: call void @_ZN6PrintyC1EPKc
-      // CHECK: [[ARRAYINIT_ELEMENT:%.+]] = getelementptr inbounds %struct.Printy
-      // CHECK: store ptr [[ARRAYINIT_ELEMENT]], ptr %arrayinit.endOfInit, align 8
-      ({
-         if (foo()) {
-           return;
-           // CHECK:      if.then:
-           // CHECK-NEXT:   [[V0:%.+]] = load ptr, ptr %arrayinit.endOfInit, align 8
-           // CHECK-NEXT:   %arraydestroy.isempty = icmp eq ptr [[ARRAYINIT_BEGIN]], [[V0]]
-           // CHECK-NEXT:   br i1 %arraydestroy.isempty, label %[[ARRAY_DESTROY_DONE:.+]], label %[[ARRAY_DESTROY_BODY:.+]]
-         }
-         Printy("b");
-       })
-      },
-      Printy("c")
-      // CHECK:       if.end:
-      // CHECK-NEXT:    call void @_ZN6PrintyC1EPKc
-      // CHECK:         call void @_ZN6PrintyC1EPKc
-      // CHECK-NEXT:    call void @_ZZ15ArraySubobjectsvEN1SD1Ev
-      // CHECK-NEXT:    br label %return
-    };
-    // CHECK:       return:
-    // CHECK-NEXT:    ret void
-
-    // CHECK:       [[ARRAY_DESTROY_BODY]]:
-    // CHECK-NEXT:    %arraydestroy.elementPast = phi ptr [ %0, %if.then ], [ %arraydestroy.element, %[[ARRAY_DESTROY_BODY]] ]
-    // CHECK-NEXT:    %arraydestroy.element = getelementptr inbounds %struct.Printy, ptr %arraydestroy.elementPast, i64 -1
-    // CHECK-NEXT:    call void @_ZN6PrintyD1Ev(ptr noundef nonnull align 8 dereferenceable(8) %arraydestroy.element)
-    // CHECK-NEXT:    %arraydestroy.done = icmp eq ptr %arraydestroy.element, [[ARRAYINIT_BEGIN]]
-    // CHECK-NEXT:    br i1 %arraydestroy.done, label %[[ARRAY_DESTROY_DONE]], label %[[ARRAY_DESTROY_BODY]]
-
-    // CHECK:       [[ARRAY_DESTROY_DONE]]
-    // CHECK-NEXT:    [[ARRAY_BEGIN:%.+]] = getelementptr inbounds [2 x %struct.Printy], ptr %arr1, i32 0, i32 0
-    // CHECK-NEXT:    [[V1:%.+]] = getelementptr inbounds %struct.Printy, ptr [[ARRAY_BEGIN]], i64 2
-    // CHECK-NEXT:    br label %[[ARRAY_DESTROY_BODY2:.+]]
-
-    // CHECK:       [[ARRAY_DESTROY_BODY2]]:
-    // CHECK-NEXT:    %arraydestroy.elementPast5 = phi ptr [ %1, %[[ARRAY_DESTROY_DONE]] ], [ %arraydestroy.element6, %[[ARRAY_DESTROY_BODY2]] ]
-    // CHECK-NEXT:    %arraydestroy.element6 = getelementptr inbounds %struct.Printy, ptr %arraydestroy.elementPast5, i64 -1
-    // CHECK-NEXT:    call void @_ZN6PrintyD1Ev(ptr noundef nonnull align 8 dereferenceable(8) %arraydestroy.element6)
-    // CHECK-NEXT:    %arraydestroy.done7 = icmp eq ptr %arraydestroy.element6, [[ARRAY_BEGIN]]
-    // CHECK-NEXT:    br i1 %arraydestroy.done7, label %[[ARRAY_DESTROY_DONE2:.+]], label %[[ARRAY_DESTROY_BODY2]]
-
-
-    // CHECK:     [[ARRAY_DESTROY_DONE2]]:
-    // CHECK-NEXT:  br label %return
-}
-
-void LambdaInit() {
-  // CHECK-LABEL: define dso_local void @_Z10LambdaInitv()
-  auto S = [a = Printy("a"), b = ({
-                               if (foo()) {
-                                 return;
-                                 // CHECK:       if.then:
-                                 // CHECK-NEXT:    call void @_ZN6PrintyD1Ev
-                                 // CHECK-NEXT:    br label %return
-                               }
-                               Printy("b");
-                             })]() { return a; };
-}
-
-void LifetimeExtended() {
-  // CHECK-LABEL: define dso_local void @_Z16LifetimeExtendedv
-  struct PrintyRefBind {
-    const Printy &a;
-    const Printy &b;
-  };
-  PrintyRefBind ps = {Printy("a"), ({
-                        if (foo()) {
-                          return;
-                          // CHECK: if.then:
-                          // CHECK-NEXT: call void @_ZN6PrintyD1Ev
-                          // CHECK-NEXT: br label %return
-                        }
-                        Printy("b");
-                      })};
-}
-
-void NewArrayInit() {
-  // CHECK-LABEL: define dso_local void @_Z12NewArrayInitv()
-  // CHECK: %array.init.end = alloca ptr, align 8
-  // CHECK: store ptr %0, ptr %array.init.end, align 8
-  Printy *array = new Printy[3]{
-    "a",
-    // CHECK: call void @_ZN6PrintyC1EPKc
-    // CHECK: store ptr %array.exp.next, ptr %array.init.end, align 8
-    "b", 
-    // CHECK: call void @_ZN6PrintyC1EPKc
-    // CHECK: store ptr %array.exp.next1, ptr %array.init.end, align 8
-    ({
-        if (foo()) {
-          return;
-          // CHECK: if.then:
-          // CHECK:   br i1 %arraydestroy.isempty, label %arraydestroy.done{{.*}}, label %arraydestroy.body
-        }
-        "b";
-        // CHECK: if.end:
-        // CHECK:   call void @_ZN6PrintyC1EPKc
-    })};
-  // CHECK:       arraydestroy.body:
-  // CHECK-NEXT:    %arraydestroy.elementPast = phi ptr [ %{{.*}}, %if.then ], [ %arraydestroy.element, %arraydestroy.body ]
-  // CHECK-NEXT:    %arraydestroy.element = getelementptr inbounds %struct.Printy, ptr %arraydestroy.elementPast, i64 -1
-  // CHECK-NEXT:    call void @_ZN6PrintyD1Ev(ptr noundef nonnull align 8 dereferenceable(8) %arraydestroy.element)
-  // CHECK-NEXT:    %arraydestroy.done = icmp eq ptr %arraydestroy.element, %0
-  // CHECK-NEXT:    br i1 %arraydestroy.done, label %arraydestroy.done{{.*}}, label %arraydestroy.body
-
-  // CHECK:       arraydestroy.done{{.*}}:                               ; preds = %arraydestroy.body, %if.then
-  // CHECK-NEXT:    br label %return
-}
-
-void DestroyInConditionalCleanup() {
-  // EH-LABEL: DestroyInConditionalCleanupv()
-  // NOEH-LABEL: DestroyInConditionalCleanupv()
-  struct A {
-    A() {}
-    ~A() {}
-  };
-
-  struct Value {
-    Value(A) {}
-    ~Value() {}
-  };
-
-  struct V2 {
-    Value K;
-    Value V;
-  };
-  // Verify we use conditional cleanups.
-  (void)(foo() ? V2{A(), A()} : V2{A(), A()});
-  // NOEH:   cond.true:
-  // NOEH:      call void @_ZZ27DestroyInConditionalCleanupvEN1AC1Ev
-  // NOEH:      store ptr %{{.*}}, ptr %cond-cleanup.save
-
-  // EH:   cond.true:
-  // EH:        invoke void @_ZZ27DestroyInConditionalCleanupvEN1AC1Ev
-  // EH:        store ptr %{{.*}}, ptr %cond-cleanup.save
-}
-
-void ArrayInitWithContinue() {
-  // CHECK-LABEL: @_Z21ArrayInitWithContinuev
-  // Verify that we start to emit the array destructor.
-  // CHECK: %arrayinit.endOfInit = alloca ptr, align 8
-  for (int i = 0; i < 1; ++i) {
-    Printy arr[2] = {"a", ({
-                       if (foo()) {
-                         continue;
-                       }
-                       "b";
-                     })};
-  }
-}
-
-struct [[clang::trivial_abi]] HasTrivialABI {
-  HasTrivialABI();
-  ~HasTrivialABI();
-};
-void AcceptTrivialABI(HasTrivialABI, int);
-void TrivialABI() {
-  // CHECK-LABEL: define dso_local void @_Z10TrivialABIv()
-  AcceptTrivialABI(HasTrivialABI(), ({
-                     if (foo()) return;
-                     // CHECK:      if.then:
-                     // CHECK-NEXT:   call void @_ZN13HasTrivialABID1Ev
-                     // CHECK-NEXT:   br label %return
-                     0;
-                   }));
-}
diff --git a/clang/test/CodeGenCoroutines/coro-suspend-cleanups.cpp b/clang/test/CodeGenCoroutines/coro-suspend-cleanups.cpp
deleted file mode 100644
index 06cc2069dbe9a..0000000000000
--- a/clang/test/CodeGenCoroutines/coro-suspend-cleanups.cpp
+++ /dev/null
@@ -1,93 +0,0 @@
-// RUN: %clang_cc1 --std=c++20 -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s
-
-#include "Inputs/coroutine.h"
-
-struct Printy {
-  Printy(const char *name) : name(name) {}
-  ~Printy() {}
-  const char *name;
-};
-
-struct coroutine {
-  struct promise_type;
-  std::coroutine_handle<promise_type> handle;
-  ~coroutine() {
-    if (handle) handle.destroy();
-  }
-};
-
-struct coroutine::promise_type {
-  coroutine get_return_object() {
-    return {std::coroutine_handle<promise_type>::from_promise(*this)};
-  }
-  std::suspend_never initial_suspend() noexcept { return {}; }
-  std::suspend_always final_suspend() noexcept { return {}; }
-  void return_void() {}
-  void unhandled_exception() {}
-};
-
-struct Awaiter : std::suspend_always {
-  Printy await_resume() { return {"awaited"}; }
-};
-
-int foo() { return 2; }
-
-coroutine ArrayInitCoro() {
-  // Verify that:
-  //  - We do the necessary stores for array cleanups.
-  //  - Array cleanups are called by await.cleanup.
-  //  - We activate the cleanup after the first element and deactivate it in await.ready (see cleanup.isactive).
-
-  // CHECK-LABEL: define dso_local void @_Z13ArrayInitCorov
-  // CHECK: %arrayinit.endOfInit = alloca ptr, align 8
-  // CHECK: %cleanup.isactive = alloca i1, align 1
-  Printy arr[2] = {
-    Printy("a"),
-    // CHECK:       %arrayinit.begin = getelementptr inbounds [2 x %struct.Printy], ptr %arr.reload.addr, i64 0, i64 0
-    // CHECK-NEXT:  %arrayinit.begin.spill.addr = getelementptr inbounds %_Z13ArrayInitCorov.Frame, ptr %0, i32 0, i32 10
-    // CHECK-NEXT:  store ptr %arrayinit.begin, ptr %arrayinit.begin.spill.addr, align 8
-    // CHECK-NEXT:  store i1 true, ptr %cleanup.isactive.reload.addr, align 1
-    // CHECK-NEXT:  store ptr %arrayinit.begin, ptr %arrayinit.endOfInit.reload.addr, align 8
-    // CHECK-NEXT:  call void @_ZN6PrintyC1EPKc(ptr noundef nonnull align 8 dereferenceable(8) %arrayinit.begin, ptr noundef @.str)
-    // CHECK-NEXT:  %arrayinit.element = getelementptr inbounds %struct.Printy, ptr %arrayinit.begin, i64 1
-    // CHECK-NEXT:  %arrayinit.element.spill.addr = getelementptr inbounds %_Z13ArrayInitCorov.Frame, ptr %0, i32 0, i32 11
-    // CHECK-NEXT:  store ptr %arrayinit.element, ptr %arrayinit.element.spill.addr, align 8
-    // CHECK-NEXT:  store ptr %arrayinit.element, ptr %arrayinit.endOfInit.reload.addr, align 8
-    co_await Awaiter{}
-    // CHECK-NEXT:  @_ZNSt14suspend_always11await_readyEv
-    // CHECK-NEXT:  br i1 %{{.+}}, label %await.ready, label %CoroSave30
-  };
-  // CHECK:       await.cleanup:                                    ; preds = %AfterCoroSuspend{{.*}}
-  // CHECK-NEXT:    br label %cleanup{{.*}}.from.await.cleanup
-
-  // CHECK:       cleanup{{.*}}.from.await.cleanup:                      ; preds = %await.cleanup
-  // CHECK:         br label %cleanup{{.*}}
-
-  // CHECK:       await.ready:
-  // CHECK-NEXT:    %arrayinit.element.reload.addr = getelementptr inbounds %_Z13ArrayInitCorov.Frame, ptr %0, i32 0, i32 11
-  // CHECK-NEXT:    %arrayinit.element.reload = load ptr, ptr %arrayinit.element.reload.addr, align 8
-  // CHECK-NEXT:    call void @_ZN7Awaiter12await_resumeEv
-  // CHECK-NEXT:    store i1 false, ptr %cleanup.isactive.reload.addr, align 1
-  // CHECK-NEXT:    br label %cleanup{{.*}}.from.await.ready
-
-  // CHECK:       cleanup{{.*}}:                                         ; preds = %cleanup{{.*}}.from.await.ready, %cleanup{{.*}}.from.await.cleanup
-  // CHECK:         %cleanup.is_active = load i1, ptr %cleanup.isactive.reload.addr, align 1
-  // CHECK-NEXT:    br i1 %cleanup.is_active, label %cleanup.action, label %cleanup.done
-
-  // CHECK:       cleanup.action:
-  // CHECK:         %arraydestroy.isempty = icmp eq ptr %arrayinit.begin.reload{{.*}}, %{{.*}}
-  // CHECK-NEXT:    br i1 %arraydestroy.isempty, label %arraydestroy.done{{.*}}, label %arraydestroy.body.from.cleanup.action
-  // Ignore rest of the array cleanup.
-}
-
-coroutine ArrayInitWithCoReturn() {
-  // CHECK-LABEL: define dso_local void @_Z21ArrayInitWithCoReturnv
-  // Verify that we start to emit the array destructor.
-  // CHECK: %arrayinit.endOfInit = alloca ptr, align 8
-  Printy arr[2] = {"a", ({
-                      if (foo()) {
-                        co_return;
-                      }
-                      "b";
-                    })};
-}

From 9ce74d6d470437cde82e91631395322ec1c36eeb Mon Sep 17 00:00:00 2001
From: Shilei Tian <i@tianshilei.me>
Date: Tue, 16 Apr 2024 09:34:18 -0400
Subject: [PATCH 587/886] [AMDGPU][CodeGen] Improve handling of memcpy for
 -Os/-Oz compilations (#87632)

We had some instances when LLVM would not inline fixed-count memcpy and
ended up
attempting to lower it a a libcall, which would not work on AMDGPU as
the
address space doesn't meet the requirement, causing compiler crash.

The patch relaxes the threshold used for -Os/-Oz compilation so we're
always allowed
to inline memory copy functions.

This patch basically does the same thing as
https://reviews.llvm.org/D158226 for
AMDGPU.

Fix #88497.
---
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp |    6 +
 llvm/test/CodeGen/AMDGPU/memcpy-libcall.ll    | 2696 +++++++++++++++++
 2 files changed, 2702 insertions(+)
 create mode 100644 llvm/test/CodeGen/AMDGPU/memcpy-libcall.ll

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index f283af6fa07d3..db69d50799e70 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -59,6 +59,12 @@ unsigned AMDGPUTargetLowering::numBitsSigned(SDValue Op, SelectionDAG &DAG) {
 AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
                                            const AMDGPUSubtarget &STI)
     : TargetLowering(TM), Subtarget(&STI) {
+  // Always lower memset, memcpy, and memmove intrinsics to load/store
+  // instructions, rather then generating calls to memset, mempcy or memmove.
+  MaxStoresPerMemset = MaxStoresPerMemsetOptSize = ~0U;
+  MaxStoresPerMemcpy = MaxStoresPerMemcpyOptSize = ~0U;
+  MaxStoresPerMemmove = MaxStoresPerMemmoveOptSize = ~0U;
+
   // Lower floating point store/load to integer store/load to reduce the number
   // of patterns in tablegen.
   setOperationAction(ISD::LOAD, MVT::f32, Promote);
diff --git a/llvm/test/CodeGen/AMDGPU/memcpy-libcall.ll b/llvm/test/CodeGen/AMDGPU/memcpy-libcall.ll
new file mode 100644
index 0000000000000..358f42dfe8dd5
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/memcpy-libcall.ll
@@ -0,0 +1,2696 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 %s -o - | FileCheck %s
+
+%struct.S = type { [32 x i32] }
+
+@shared = addrspace(3) global %struct.S undef, align 4
+
+define amdgpu_kernel void @memcpy_p0_p0_minsize(ptr %dest, ptr readonly %src) #0 {
+; CHECK-LABEL: memcpy_p0_p0_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:1
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:1
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:2
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:2
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:3
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:3
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:4
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:4
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:5
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:5
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:6
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:6
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:7
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:7
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:8
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:8
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:9
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:9
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:10
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:10
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:11
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:11
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:12
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:12
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:13
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:13
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:14
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:14
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:15
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:15
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:16
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:17
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:17
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:18
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:19
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:19
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:20
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:20
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:21
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:21
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:22
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:22
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:23
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:23
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:24
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:24
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:25
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:25
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:26
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:26
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:27
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:27
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:28
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:28
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:29
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:29
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:30
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:30
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:31
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:31
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:32
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:33
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:33
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:34
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:34
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:35
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:35
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:36
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:37
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:37
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:38
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:38
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:39
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:40
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:40
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:41
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:41
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:42
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:42
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:43
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:43
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:44
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:44
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:45
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:45
+; CHECK-NEXT:    flat_load_ubyte v0, v[0:1] offset:46
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v0 offset:46
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p0.i64(ptr %dest, ptr %src, i64 47, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p1_p1_minsize(ptr addrspace(1) %dest, ptr addrspace(1) %src) #0 {
+; CHECK-LABEL: memcpy_p1_p1_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1] offset:32
+; CHECK-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1] offset:39
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) %dest, ptr addrspace(1) %src, i64 47, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p1_p4_minsize(ptr addrspace(1) %global, ptr addrspace(4) %0) #0 {
+; CHECK-LABEL: memcpy_p1_p4_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:32
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:48
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:48
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:64
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:64
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:80
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:80
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:96
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:112
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:112
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p1.p4.i64(ptr addrspace(1) %global, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p5_p4_minsize(ptr addrspace(5) %local, ptr addrspace(4) %0) #0 {
+; CHECK-LABEL: memcpy_p5_p4_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_mov_b64 s[10:11], s[2:3]
+; CHECK-NEXT:    s_mov_b64 s[8:9], s[0:1]
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
+; CHECK-NEXT:    s_load_dword s2, s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    s_add_u32 s8, s8, s7
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1]
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:1
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:2
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:3
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:4
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:5
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:6
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:7
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:8
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:9
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:10
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:11
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:12
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:13
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:14
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:15
+; CHECK-NEXT:    s_addc_u32 s9, s9, 0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s2
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:16
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:17
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(18)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:19
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:1
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:20
+; CHECK-NEXT:    s_waitcnt vmcnt(20)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:2
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:21
+; CHECK-NEXT:    s_waitcnt vmcnt(21)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:3
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:22
+; CHECK-NEXT:    s_waitcnt vmcnt(22)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:4
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:23
+; CHECK-NEXT:    s_waitcnt vmcnt(23)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:5
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:24
+; CHECK-NEXT:    s_waitcnt vmcnt(24)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:6
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:25
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:7
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:26
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:8
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:27
+; CHECK-NEXT:    s_waitcnt vmcnt(27)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:9
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:28
+; CHECK-NEXT:    s_waitcnt vmcnt(28)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:10
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:29
+; CHECK-NEXT:    s_waitcnt vmcnt(29)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:11
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:30
+; CHECK-NEXT:    s_waitcnt vmcnt(30)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:12
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:31
+; CHECK-NEXT:    s_waitcnt vmcnt(31)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:13
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(32)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:14
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:33
+; CHECK-NEXT:    s_waitcnt vmcnt(33)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:15
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:34
+; CHECK-NEXT:    s_waitcnt vmcnt(34)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:16
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:35
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:17
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:18
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:37
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:19
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:38
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:20
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:21
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:40
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:22
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:41
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:23
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:42
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:24
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:43
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:25
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:44
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:26
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:45
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:27
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:46
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:28
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:47
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:29
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:48
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:30
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:49
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:31
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:50
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:32
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:51
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:33
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:52
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:34
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:53
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:35
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:54
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:36
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:55
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:37
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:56
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:38
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:57
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:39
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:58
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:40
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:59
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:41
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:60
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:42
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:61
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:43
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:62
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:44
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:63
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:45
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:64
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:46
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:65
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:47
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:66
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:48
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:67
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:49
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:68
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:50
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:69
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:51
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:70
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:52
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:71
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:53
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:72
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:54
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:73
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:55
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:74
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:56
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:75
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:57
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:76
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:58
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:77
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:59
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:78
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:60
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:79
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:61
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:80
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:62
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:81
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:63
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:82
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:64
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:83
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:65
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:84
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:66
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:85
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:67
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:86
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:68
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:87
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:69
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:88
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:70
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:71
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:89
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:72
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:91
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:73
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:92
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:74
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:93
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:75
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:94
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:76
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:95
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:77
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:78
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:97
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:79
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:98
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:80
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:99
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:81
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:100
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:82
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:101
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:83
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:102
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:84
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:103
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:85
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:104
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:86
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:105
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:87
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:106
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:88
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:107
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:89
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(34)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:91
+; CHECK-NEXT:    s_waitcnt vmcnt(33)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:92
+; CHECK-NEXT:    s_waitcnt vmcnt(32)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:93
+; CHECK-NEXT:    s_waitcnt vmcnt(31)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:94
+; CHECK-NEXT:    s_waitcnt vmcnt(30)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:95
+; CHECK-NEXT:    s_waitcnt vmcnt(29)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(28)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:97
+; CHECK-NEXT:    s_waitcnt vmcnt(27)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:98
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:99
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:100
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:108
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:109
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:110
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:111
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:112
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:113
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:114
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:115
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:116
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:117
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:118
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:119
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:101
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:120
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:102
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:121
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:103
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:122
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:104
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:123
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:105
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:124
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:106
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:125
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:107
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:126
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    global_load_ubyte v21, v0, s[0:1] offset:127
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:108
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:109
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:110
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:111
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:112
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:113
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:114
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:115
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:116
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:117
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:118
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:119
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:120
+; CHECK-NEXT:    s_waitcnt vmcnt(24)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:121
+; CHECK-NEXT:    s_waitcnt vmcnt(23)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:122
+; CHECK-NEXT:    s_waitcnt vmcnt(22)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:123
+; CHECK-NEXT:    s_waitcnt vmcnt(21)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:124
+; CHECK-NEXT:    s_waitcnt vmcnt(20)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:125
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:126
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v21, v1, s[8:11], 0 offen offset:127
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) %local, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p0_p5_minsize(ptr %generic, ptr addrspace(5) %src) #0 {
+; CHECK-LABEL: memcpy_p0_p5_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_mov_b64 s[10:11], s[2:3]
+; CHECK-NEXT:    s_mov_b64 s[8:9], s[0:1]
+; CHECK-NEXT:    s_load_dword s0, s[4:5], 0x8
+; CHECK-NEXT:    s_add_u32 s8, s8, s7
+; CHECK-NEXT:    s_addc_u32 s9, s9, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:1
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:2
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:3
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:4
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:5
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:6
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:7
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:8
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:9
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:10
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:11
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:12
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:13
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:14
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:15
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:16
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:17
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    s_waitcnt vmcnt(17)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:1
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:19
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:2
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:20
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:3
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:21
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:4
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:22
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:5
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:23
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:6
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:24
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:7
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:25
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:8
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:26
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:9
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:27
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:10
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:28
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:11
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:29
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:12
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:30
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:13
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:31
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:14
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:32
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:15
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:33
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:16
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:34
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:17
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:35
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:18
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:19
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:37
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:20
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:38
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:21
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:39
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:22
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:40
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:23
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:41
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:24
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:42
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:25
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:43
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:26
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:44
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:27
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:45
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:28
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:46
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:29
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:47
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:30
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:48
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:31
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:49
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:32
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:50
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:33
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:51
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:34
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:52
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:35
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:53
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:36
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:54
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:37
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:55
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:38
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:56
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:39
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:57
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:40
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:58
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:41
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:59
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:42
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:60
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:43
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:61
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:44
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:62
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:45
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:63
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:46
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:64
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:47
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:65
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:48
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:66
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:49
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:67
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:50
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:68
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:51
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:69
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:52
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:70
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:53
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:71
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:54
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:72
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:55
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:73
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:56
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:74
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:57
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:75
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:58
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:76
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:59
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:77
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:60
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:78
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:61
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:79
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:62
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:80
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:63
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:81
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:64
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:82
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:65
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:83
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:66
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:84
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:67
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:85
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:68
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:86
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:69
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:87
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:70
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:88
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:71
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:89
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:72
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:73
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:74
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:91
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:92
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:75
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:93
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:76
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:94
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:77
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:95
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:78
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:96
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:79
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:97
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:80
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:98
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:81
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:99
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:82
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:100
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:83
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:101
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:84
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:102
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:85
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:103
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:86
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:104
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:87
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:105
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:88
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:106
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:89
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:107
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:90
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:108
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:91
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:92
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:93
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:94
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:95
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:96
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:97
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:98
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:99
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:100
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:101
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:109
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:110
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:111
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:112
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:113
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:114
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:115
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:116
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:117
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:118
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:119
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:102
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:120
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:103
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:121
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:104
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:122
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:105
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:123
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:106
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:124
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:107
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:125
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:108
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:126
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    buffer_load_ubyte v21, v2, s[8:11], 0 offen offset:127
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:109
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:110
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:111
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:112
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:113
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:114
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:115
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:116
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:117
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:118
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:119
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:120
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:121
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:122
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:123
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:124
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:125
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:126
+; CHECK-NEXT:    flat_store_byte v[0:1], v21 offset:127
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p5.i64(ptr %generic, ptr addrspace(5) %src, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p3_p4_minsize(ptr addrspace(4) %0) #0 {
+; CHECK-LABEL: memcpy_p3_p4_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v24, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v24, s[0:1] offset:112
+; CHECK-NEXT:    global_load_dwordx4 v[4:7], v24, s[0:1] offset:96
+; CHECK-NEXT:    global_load_dwordx4 v[8:11], v24, s[0:1] offset:80
+; CHECK-NEXT:    global_load_dwordx4 v[12:15], v24, s[0:1] offset:64
+; CHECK-NEXT:    global_load_dwordx4 v[16:19], v24, s[0:1] offset:48
+; CHECK-NEXT:    global_load_dwordx4 v[20:23], v24, s[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(5)
+; CHECK-NEXT:    ds_write2_b64 v24, v[0:1], v[2:3] offset0:14 offset1:15
+; CHECK-NEXT:    s_waitcnt vmcnt(4)
+; CHECK-NEXT:    ds_write2_b64 v24, v[4:5], v[6:7] offset0:12 offset1:13
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v24, s[0:1] offset:16
+; CHECK-NEXT:    global_load_dwordx4 v[4:7], v24, s[0:1]
+; CHECK-NEXT:    s_waitcnt vmcnt(5)
+; CHECK-NEXT:    ds_write2_b64 v24, v[8:9], v[10:11] offset0:10 offset1:11
+; CHECK-NEXT:    s_waitcnt vmcnt(4)
+; CHECK-NEXT:    ds_write2_b64 v24, v[12:13], v[14:15] offset0:8 offset1:9
+; CHECK-NEXT:    s_waitcnt vmcnt(3)
+; CHECK-NEXT:    ds_write2_b64 v24, v[16:17], v[18:19] offset0:6 offset1:7
+; CHECK-NEXT:    s_waitcnt vmcnt(2)
+; CHECK-NEXT:    ds_write2_b64 v24, v[20:21], v[22:23] offset0:4 offset1:5
+; CHECK-NEXT:    s_waitcnt vmcnt(1)
+; CHECK-NEXT:    ds_write2_b64 v24, v[0:1], v[2:3] offset0:2 offset1:3
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ds_write2_b64 v24, v[4:5], v[6:7] offset1:1
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p3.p4.i64(ptr addrspace(3) @shared, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p0_p3_minsize(ptr %generic) #0 {
+; CHECK-LABEL: memcpy_p0_p3_minsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v2, 0
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:127
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:126
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:125
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:124
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:127
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:126
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:123
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:125
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:124
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:122
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:121
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:123
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:120
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:119
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:122
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:121
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:118
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:120
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:119
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:117
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:116
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:118
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:115
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:114
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:117
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:116
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:113
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:115
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:114
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:112
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:111
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:113
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:110
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:109
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:112
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:111
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:108
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:110
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:109
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:107
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:106
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:108
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:105
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:104
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:107
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:106
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:103
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:105
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:104
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:102
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:101
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:103
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:100
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:99
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:102
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:101
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:98
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:100
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:99
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:97
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:96
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:98
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:95
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:94
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:97
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:96
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:93
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:95
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:94
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:92
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:91
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:93
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:90
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:89
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:92
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:91
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:88
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:90
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:89
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:87
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:86
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:88
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:85
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:84
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:87
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:86
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:83
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:85
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:84
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:82
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:81
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:83
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:80
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:79
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:82
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:81
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:78
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:80
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:79
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:77
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:76
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:78
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:75
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:74
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:77
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:76
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:73
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:75
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:74
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:72
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:71
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:73
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:70
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:69
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:72
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:71
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:68
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:70
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:69
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:67
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:66
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:68
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:65
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:64
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:67
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:66
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:63
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:65
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:64
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:62
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:61
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:63
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:60
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:59
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:62
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:61
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:58
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:60
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:59
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:57
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:56
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:58
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:55
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:54
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:57
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:56
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:53
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:55
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:54
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:52
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:51
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:53
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:50
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:49
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:52
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:51
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:48
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:50
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:49
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:47
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:46
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:48
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:45
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:44
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:47
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:46
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:43
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:45
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:44
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:42
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:41
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:43
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:40
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:39
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:42
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:41
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:38
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:40
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:39
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:37
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:36
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:38
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:35
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:34
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:37
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:36
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:33
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:35
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:34
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:32
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:31
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:33
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:30
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:29
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:32
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:31
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:28
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:30
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:29
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:27
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:26
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:28
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:25
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:24
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:27
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:26
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:23
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:25
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:24
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:22
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:21
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:23
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:20
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:19
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:22
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:21
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:18
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:20
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:19
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:16
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:17
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:18
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:8
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:9
+; CHECK-NEXT:    ds_read_u8 v7, v2 offset:10
+; CHECK-NEXT:    ds_read_u8 v8, v2 offset:11
+; CHECK-NEXT:    ds_read_u8 v9, v2 offset:12
+; CHECK-NEXT:    ds_read_u8 v10, v2 offset:13
+; CHECK-NEXT:    ds_read_u8 v11, v2 offset:14
+; CHECK-NEXT:    ds_read_u8 v12, v2 offset:15
+; CHECK-NEXT:    ds_read_u8 v13, v2
+; CHECK-NEXT:    ds_read_u8 v14, v2 offset:1
+; CHECK-NEXT:    ds_read_u8 v15, v2 offset:2
+; CHECK-NEXT:    ds_read_u8 v16, v2 offset:3
+; CHECK-NEXT:    ds_read_u8 v17, v2 offset:4
+; CHECK-NEXT:    ds_read_u8 v18, v2 offset:5
+; CHECK-NEXT:    ds_read_u8 v19, v2 offset:6
+; CHECK-NEXT:    ds_read_u8 v2, v2 offset:7
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:17
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:16
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:15
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:14
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:13
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:12
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:11
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:10
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:9
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:8
+; CHECK-NEXT:    flat_store_byte v[0:1], v2 offset:7
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:6
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:5
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:4
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:3
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:2
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:1
+; CHECK-NEXT:    flat_store_byte v[0:1], v13
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p3.i64(ptr %generic, ptr addrspace(3) @shared, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p0_p0_optsize(ptr %dest, ptr %src) #1 {
+; CHECK-LABEL: memcpy_p0_p0_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1]
+; CHECK-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:1
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:1
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:2
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:2
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:3
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:3
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:4
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:4
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:5
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:5
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:6
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:6
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:7
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:7
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:8
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:8
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:9
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:9
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:10
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:10
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:11
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:11
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:12
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:12
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:13
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:13
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:14
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:14
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:15
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:15
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:16
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:17
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:17
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:18
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:19
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:19
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:20
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:20
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:21
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:21
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:22
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:22
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:23
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:23
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:24
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:24
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:25
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:25
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:26
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:26
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:27
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:27
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:28
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:28
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:29
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:29
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:30
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:30
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:31
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:31
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:32
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:33
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:33
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:34
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:34
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:35
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:35
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:36
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:37
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:37
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:38
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:38
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:39
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:40
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:40
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:41
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:41
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:42
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:42
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:43
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:43
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:44
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:44
+; CHECK-NEXT:    flat_load_ubyte v4, v[0:1] offset:45
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v4 offset:45
+; CHECK-NEXT:    flat_load_ubyte v0, v[0:1] offset:46
+; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[2:3], v0 offset:46
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p0.i64(ptr %dest, ptr %src, i64 47, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p1_p1_optsize(ptr addrspace(1) %dest, ptr addrspace(1) %src) #1 {
+; CHECK-LABEL: memcpy_p1_p1_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1] offset:32
+; CHECK-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1] offset:39
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) %dest, ptr addrspace(1) %src, i64 47, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p1_p4_optsize(ptr addrspace(1) %global, ptr addrspace(4) %0) #1 {
+; CHECK-LABEL: memcpy_p1_p4_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v4, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:16
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:32
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:48
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:48
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:64
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:64
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:80
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:80
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:96
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3] offset:112
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:112
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p1.p4.i64(ptr addrspace(1) %global, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p5_p4_optsize(ptr addrspace(5) %local, ptr addrspace(4) %0) #1 {
+; CHECK-LABEL: memcpy_p5_p4_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_mov_b64 s[10:11], s[2:3]
+; CHECK-NEXT:    s_mov_b64 s[8:9], s[0:1]
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
+; CHECK-NEXT:    s_load_dword s2, s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    s_add_u32 s8, s8, s7
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1]
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:1
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:2
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:3
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:4
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:5
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:6
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:7
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:8
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:9
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:10
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:11
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:12
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:13
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:14
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:15
+; CHECK-NEXT:    s_addc_u32 s9, s9, 0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s2
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:16
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:17
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(18)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:19
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:1
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:20
+; CHECK-NEXT:    s_waitcnt vmcnt(20)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:2
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:21
+; CHECK-NEXT:    s_waitcnt vmcnt(21)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:3
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:22
+; CHECK-NEXT:    s_waitcnt vmcnt(22)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:4
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:23
+; CHECK-NEXT:    s_waitcnt vmcnt(23)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:5
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:24
+; CHECK-NEXT:    s_waitcnt vmcnt(24)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:6
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:25
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:7
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:26
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:8
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:27
+; CHECK-NEXT:    s_waitcnt vmcnt(27)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:9
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:28
+; CHECK-NEXT:    s_waitcnt vmcnt(28)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:10
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:29
+; CHECK-NEXT:    s_waitcnt vmcnt(29)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:11
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:30
+; CHECK-NEXT:    s_waitcnt vmcnt(30)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:12
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:31
+; CHECK-NEXT:    s_waitcnt vmcnt(31)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:13
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(32)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:14
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:33
+; CHECK-NEXT:    s_waitcnt vmcnt(33)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:15
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:34
+; CHECK-NEXT:    s_waitcnt vmcnt(34)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:16
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:35
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:17
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:18
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:37
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:19
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:38
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:20
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:39
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:21
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:40
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:22
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:41
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:23
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:42
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:24
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:43
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:25
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:44
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:26
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:45
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:27
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:46
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:28
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:47
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:29
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:48
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:30
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:49
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:31
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:50
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:32
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:51
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:33
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:52
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:34
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:53
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:35
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:54
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:36
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:55
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:37
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:56
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:38
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:57
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:39
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:58
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:40
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:59
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:41
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:60
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:42
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:61
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:43
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:62
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:44
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:63
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:45
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:64
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:46
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:65
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:47
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:66
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:48
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:67
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:49
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:68
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:50
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:69
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:51
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:70
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:52
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:71
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:53
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:72
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:54
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:73
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:55
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:74
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:56
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:75
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:57
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:76
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:58
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:77
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:59
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:78
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:60
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:79
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:61
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:80
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:62
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:81
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:63
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:82
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:64
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:83
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:65
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:84
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:66
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:85
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:67
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:86
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:68
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:87
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:69
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:88
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:70
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:71
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:89
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:72
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:91
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:73
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:92
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:74
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:93
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:75
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:94
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:76
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:95
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:77
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:78
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:97
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:79
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:98
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:80
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:99
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:81
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:100
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:82
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:101
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:83
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:102
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:84
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:103
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:85
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:104
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:86
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:105
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:87
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:106
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:88
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:107
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:89
+; CHECK-NEXT:    s_waitcnt vmcnt(35)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(34)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:91
+; CHECK-NEXT:    s_waitcnt vmcnt(33)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:92
+; CHECK-NEXT:    s_waitcnt vmcnt(32)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:93
+; CHECK-NEXT:    s_waitcnt vmcnt(31)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:94
+; CHECK-NEXT:    s_waitcnt vmcnt(30)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:95
+; CHECK-NEXT:    s_waitcnt vmcnt(29)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:96
+; CHECK-NEXT:    s_waitcnt vmcnt(28)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:97
+; CHECK-NEXT:    s_waitcnt vmcnt(27)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:98
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:99
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:100
+; CHECK-NEXT:    global_load_ubyte v2, v0, s[0:1] offset:108
+; CHECK-NEXT:    global_load_ubyte v3, v0, s[0:1] offset:109
+; CHECK-NEXT:    global_load_ubyte v4, v0, s[0:1] offset:110
+; CHECK-NEXT:    global_load_ubyte v5, v0, s[0:1] offset:111
+; CHECK-NEXT:    global_load_ubyte v6, v0, s[0:1] offset:112
+; CHECK-NEXT:    global_load_ubyte v7, v0, s[0:1] offset:113
+; CHECK-NEXT:    global_load_ubyte v15, v0, s[0:1] offset:114
+; CHECK-NEXT:    global_load_ubyte v16, v0, s[0:1] offset:115
+; CHECK-NEXT:    global_load_ubyte v17, v0, s[0:1] offset:116
+; CHECK-NEXT:    global_load_ubyte v18, v0, s[0:1] offset:117
+; CHECK-NEXT:    global_load_ubyte v19, v0, s[0:1] offset:118
+; CHECK-NEXT:    global_load_ubyte v20, v0, s[0:1] offset:119
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:101
+; CHECK-NEXT:    global_load_ubyte v8, v0, s[0:1] offset:120
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:102
+; CHECK-NEXT:    global_load_ubyte v9, v0, s[0:1] offset:121
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:103
+; CHECK-NEXT:    global_load_ubyte v10, v0, s[0:1] offset:122
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:104
+; CHECK-NEXT:    global_load_ubyte v11, v0, s[0:1] offset:123
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:105
+; CHECK-NEXT:    global_load_ubyte v12, v0, s[0:1] offset:124
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:106
+; CHECK-NEXT:    global_load_ubyte v13, v0, s[0:1] offset:125
+; CHECK-NEXT:    s_waitcnt vmcnt(36)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:107
+; CHECK-NEXT:    global_load_ubyte v14, v0, s[0:1] offset:126
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    global_load_ubyte v21, v0, s[0:1] offset:127
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[8:11], 0 offen offset:108
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v3, v1, s[8:11], 0 offen offset:109
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v4, v1, s[8:11], 0 offen offset:110
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v5, v1, s[8:11], 0 offen offset:111
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v6, v1, s[8:11], 0 offen offset:112
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v7, v1, s[8:11], 0 offen offset:113
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v15, v1, s[8:11], 0 offen offset:114
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v16, v1, s[8:11], 0 offen offset:115
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v17, v1, s[8:11], 0 offen offset:116
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v18, v1, s[8:11], 0 offen offset:117
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v19, v1, s[8:11], 0 offen offset:118
+; CHECK-NEXT:    s_waitcnt vmcnt(26)
+; CHECK-NEXT:    buffer_store_byte v20, v1, s[8:11], 0 offen offset:119
+; CHECK-NEXT:    s_waitcnt vmcnt(25)
+; CHECK-NEXT:    buffer_store_byte v8, v1, s[8:11], 0 offen offset:120
+; CHECK-NEXT:    s_waitcnt vmcnt(24)
+; CHECK-NEXT:    buffer_store_byte v9, v1, s[8:11], 0 offen offset:121
+; CHECK-NEXT:    s_waitcnt vmcnt(23)
+; CHECK-NEXT:    buffer_store_byte v10, v1, s[8:11], 0 offen offset:122
+; CHECK-NEXT:    s_waitcnt vmcnt(22)
+; CHECK-NEXT:    buffer_store_byte v11, v1, s[8:11], 0 offen offset:123
+; CHECK-NEXT:    s_waitcnt vmcnt(21)
+; CHECK-NEXT:    buffer_store_byte v12, v1, s[8:11], 0 offen offset:124
+; CHECK-NEXT:    s_waitcnt vmcnt(20)
+; CHECK-NEXT:    buffer_store_byte v13, v1, s[8:11], 0 offen offset:125
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v14, v1, s[8:11], 0 offen offset:126
+; CHECK-NEXT:    s_waitcnt vmcnt(19)
+; CHECK-NEXT:    buffer_store_byte v21, v1, s[8:11], 0 offen offset:127
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) %local, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p0_p5_optsize(ptr %generic, ptr addrspace(5) %src) #1 {
+; CHECK-LABEL: memcpy_p0_p5_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_mov_b64 s[10:11], s[2:3]
+; CHECK-NEXT:    s_mov_b64 s[8:9], s[0:1]
+; CHECK-NEXT:    s_load_dword s0, s[4:5], 0x8
+; CHECK-NEXT:    s_add_u32 s8, s8, s7
+; CHECK-NEXT:    s_addc_u32 s9, s9, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:1
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:2
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:3
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:4
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:5
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:6
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:7
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:8
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:9
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:10
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:11
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:12
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:13
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:14
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:15
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:16
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:17
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    s_waitcnt vmcnt(17)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:18
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:1
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:19
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:2
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:20
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:3
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:21
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:4
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:22
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:5
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:23
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:6
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:24
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:7
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:25
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:8
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:26
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:9
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:27
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:10
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:28
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:11
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:29
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:12
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:30
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:13
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:31
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:14
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:32
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:15
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:33
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:16
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:34
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:17
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:35
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:18
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:36
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:19
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:37
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:20
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:38
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:21
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:39
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:22
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:40
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:23
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:41
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:24
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:42
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:25
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:43
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:26
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:44
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:27
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:45
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:28
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:46
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:29
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:47
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:30
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:48
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:31
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:49
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:32
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:50
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:33
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:51
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:34
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:52
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:35
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:53
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:36
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:54
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:37
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:55
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:38
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:56
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:39
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:57
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:40
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:58
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:41
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:59
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:42
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:60
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:43
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:61
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:44
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:62
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:45
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:63
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:46
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:64
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:47
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:65
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:48
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:66
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:49
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:67
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:50
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:68
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:51
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:69
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:52
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:70
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:53
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:71
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:54
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:72
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:55
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:73
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:56
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:74
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:57
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:75
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:58
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:76
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:59
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:77
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:60
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:78
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:61
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:79
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:62
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:80
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:63
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:81
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:64
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:82
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:65
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:83
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:66
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:84
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:67
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:85
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:68
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:86
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:69
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:87
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:70
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:88
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:71
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:89
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:72
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:90
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:73
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:74
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:91
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:92
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:75
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:93
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:76
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:94
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:77
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:95
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:78
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:96
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:79
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:97
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:80
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:98
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:81
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:99
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:82
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:100
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:83
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:101
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:84
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:102
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:85
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:103
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:86
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:104
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:87
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:105
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:88
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:106
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:89
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:107
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:90
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:108
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:91
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:92
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:93
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:94
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:95
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:96
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:97
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:98
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:99
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:100
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:101
+; CHECK-NEXT:    buffer_load_ubyte v4, v2, s[8:11], 0 offen offset:109
+; CHECK-NEXT:    buffer_load_ubyte v5, v2, s[8:11], 0 offen offset:110
+; CHECK-NEXT:    buffer_load_ubyte v6, v2, s[8:11], 0 offen offset:111
+; CHECK-NEXT:    buffer_load_ubyte v7, v2, s[8:11], 0 offen offset:112
+; CHECK-NEXT:    buffer_load_ubyte v8, v2, s[8:11], 0 offen offset:113
+; CHECK-NEXT:    buffer_load_ubyte v9, v2, s[8:11], 0 offen offset:114
+; CHECK-NEXT:    buffer_load_ubyte v10, v2, s[8:11], 0 offen offset:115
+; CHECK-NEXT:    buffer_load_ubyte v11, v2, s[8:11], 0 offen offset:116
+; CHECK-NEXT:    buffer_load_ubyte v12, v2, s[8:11], 0 offen offset:117
+; CHECK-NEXT:    buffer_load_ubyte v13, v2, s[8:11], 0 offen offset:118
+; CHECK-NEXT:    buffer_load_ubyte v14, v2, s[8:11], 0 offen offset:119
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:102
+; CHECK-NEXT:    buffer_load_ubyte v15, v2, s[8:11], 0 offen offset:120
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:103
+; CHECK-NEXT:    buffer_load_ubyte v16, v2, s[8:11], 0 offen offset:121
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:104
+; CHECK-NEXT:    buffer_load_ubyte v17, v2, s[8:11], 0 offen offset:122
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:105
+; CHECK-NEXT:    buffer_load_ubyte v18, v2, s[8:11], 0 offen offset:123
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:106
+; CHECK-NEXT:    buffer_load_ubyte v19, v2, s[8:11], 0 offen offset:124
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:107
+; CHECK-NEXT:    buffer_load_ubyte v20, v2, s[8:11], 0 offen offset:125
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:108
+; CHECK-NEXT:    buffer_load_ubyte v3, v2, s[8:11], 0 offen offset:126
+; CHECK-NEXT:    s_nop 0
+; CHECK-NEXT:    buffer_load_ubyte v21, v2, s[8:11], 0 offen offset:127
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:109
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:110
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:111
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:112
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:113
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:114
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:115
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:116
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:117
+; CHECK-NEXT:    flat_store_byte v[0:1], v13 offset:118
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:119
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:120
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:121
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:122
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:123
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:124
+; CHECK-NEXT:    flat_store_byte v[0:1], v20 offset:125
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:126
+; CHECK-NEXT:    flat_store_byte v[0:1], v21 offset:127
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p5.i64(ptr %generic, ptr addrspace(5) %src, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p3_p4_optsize(ptr addrspace(4) %0) #1 {
+; CHECK-LABEL: memcpy_p3_p4_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v24, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v24, s[0:1] offset:112
+; CHECK-NEXT:    global_load_dwordx4 v[4:7], v24, s[0:1] offset:96
+; CHECK-NEXT:    global_load_dwordx4 v[8:11], v24, s[0:1] offset:80
+; CHECK-NEXT:    global_load_dwordx4 v[12:15], v24, s[0:1] offset:64
+; CHECK-NEXT:    global_load_dwordx4 v[16:19], v24, s[0:1] offset:48
+; CHECK-NEXT:    global_load_dwordx4 v[20:23], v24, s[0:1] offset:32
+; CHECK-NEXT:    s_waitcnt vmcnt(5)
+; CHECK-NEXT:    ds_write2_b64 v24, v[0:1], v[2:3] offset0:14 offset1:15
+; CHECK-NEXT:    s_waitcnt vmcnt(4)
+; CHECK-NEXT:    ds_write2_b64 v24, v[4:5], v[6:7] offset0:12 offset1:13
+; CHECK-NEXT:    global_load_dwordx4 v[0:3], v24, s[0:1] offset:16
+; CHECK-NEXT:    global_load_dwordx4 v[4:7], v24, s[0:1]
+; CHECK-NEXT:    s_waitcnt vmcnt(5)
+; CHECK-NEXT:    ds_write2_b64 v24, v[8:9], v[10:11] offset0:10 offset1:11
+; CHECK-NEXT:    s_waitcnt vmcnt(4)
+; CHECK-NEXT:    ds_write2_b64 v24, v[12:13], v[14:15] offset0:8 offset1:9
+; CHECK-NEXT:    s_waitcnt vmcnt(3)
+; CHECK-NEXT:    ds_write2_b64 v24, v[16:17], v[18:19] offset0:6 offset1:7
+; CHECK-NEXT:    s_waitcnt vmcnt(2)
+; CHECK-NEXT:    ds_write2_b64 v24, v[20:21], v[22:23] offset0:4 offset1:5
+; CHECK-NEXT:    s_waitcnt vmcnt(1)
+; CHECK-NEXT:    ds_write2_b64 v24, v[0:1], v[2:3] offset0:2 offset1:3
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ds_write2_b64 v24, v[4:5], v[6:7] offset1:1
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p3.p4.i64(ptr addrspace(3) @shared, ptr addrspace(4) %0, i64 128, i1 false)
+  ret void
+}
+
+define amdgpu_kernel void @memcpy_p0_p3_optsize(ptr %generic) #1 {
+; CHECK-LABEL: memcpy_p0_p3_optsize:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v2, 0
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:127
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:126
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:125
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:124
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:127
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:126
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:123
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:125
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:124
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:122
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:121
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:123
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:120
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:119
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:122
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:121
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:118
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:120
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:119
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:117
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:116
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:118
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:115
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:114
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:117
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:116
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:113
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:115
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:114
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:112
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:111
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:113
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:110
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:109
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:112
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:111
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:108
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:110
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:109
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:107
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:106
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:108
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:105
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:104
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:107
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:106
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:103
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:105
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:104
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:102
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:101
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:103
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:100
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:99
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:102
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:101
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:98
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:100
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:99
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:97
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:96
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:98
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:95
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:94
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:97
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:96
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:93
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:95
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:94
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:92
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:91
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:93
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:90
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:89
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:92
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:91
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:88
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:90
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:89
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:87
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:86
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:88
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:85
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:84
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:87
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:86
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:83
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:85
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:84
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:82
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:81
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:83
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:80
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:79
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:82
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:81
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:78
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:80
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:79
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:77
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:76
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:78
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:75
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:74
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:77
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:76
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:73
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:75
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:74
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:72
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:71
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:73
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:70
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:69
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:72
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:71
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:68
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:70
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:69
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:67
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:66
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:68
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:65
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:64
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:67
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:66
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:63
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:65
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:64
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:62
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:61
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:63
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:60
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:59
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:62
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:61
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:58
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:60
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:59
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:57
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:56
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:58
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:55
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:54
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:57
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:56
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:53
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:55
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:54
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:52
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:51
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:53
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:50
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:49
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:52
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:51
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:48
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:50
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:49
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:47
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:46
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:48
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:45
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:44
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:47
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:46
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:43
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:45
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:44
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:42
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:41
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:43
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:40
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:39
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:42
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:41
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:38
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:40
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:39
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:37
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:36
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:38
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:35
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:34
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:37
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:36
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:33
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:35
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:34
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:32
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:31
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:33
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:30
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:29
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:32
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:31
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:28
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:30
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:29
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:27
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:26
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:28
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:25
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:24
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:27
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:26
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:23
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:25
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:24
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:22
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:21
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:23
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:20
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:19
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:22
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:21
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:18
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:20
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:19
+; CHECK-NEXT:    ds_read_u8 v3, v2 offset:16
+; CHECK-NEXT:    ds_read_u8 v5, v2 offset:17
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:18
+; CHECK-NEXT:    ds_read_u8 v4, v2 offset:8
+; CHECK-NEXT:    ds_read_u8 v6, v2 offset:9
+; CHECK-NEXT:    ds_read_u8 v7, v2 offset:10
+; CHECK-NEXT:    ds_read_u8 v8, v2 offset:11
+; CHECK-NEXT:    ds_read_u8 v9, v2 offset:12
+; CHECK-NEXT:    ds_read_u8 v10, v2 offset:13
+; CHECK-NEXT:    ds_read_u8 v11, v2 offset:14
+; CHECK-NEXT:    ds_read_u8 v12, v2 offset:15
+; CHECK-NEXT:    ds_read_u8 v13, v2
+; CHECK-NEXT:    ds_read_u8 v14, v2 offset:1
+; CHECK-NEXT:    ds_read_u8 v15, v2 offset:2
+; CHECK-NEXT:    ds_read_u8 v16, v2 offset:3
+; CHECK-NEXT:    ds_read_u8 v17, v2 offset:4
+; CHECK-NEXT:    ds_read_u8 v18, v2 offset:5
+; CHECK-NEXT:    ds_read_u8 v19, v2 offset:6
+; CHECK-NEXT:    ds_read_u8 v2, v2 offset:7
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    flat_store_byte v[0:1], v5 offset:17
+; CHECK-NEXT:    flat_store_byte v[0:1], v3 offset:16
+; CHECK-NEXT:    flat_store_byte v[0:1], v12 offset:15
+; CHECK-NEXT:    flat_store_byte v[0:1], v11 offset:14
+; CHECK-NEXT:    flat_store_byte v[0:1], v10 offset:13
+; CHECK-NEXT:    flat_store_byte v[0:1], v9 offset:12
+; CHECK-NEXT:    flat_store_byte v[0:1], v8 offset:11
+; CHECK-NEXT:    flat_store_byte v[0:1], v7 offset:10
+; CHECK-NEXT:    flat_store_byte v[0:1], v6 offset:9
+; CHECK-NEXT:    flat_store_byte v[0:1], v4 offset:8
+; CHECK-NEXT:    flat_store_byte v[0:1], v2 offset:7
+; CHECK-NEXT:    flat_store_byte v[0:1], v19 offset:6
+; CHECK-NEXT:    flat_store_byte v[0:1], v18 offset:5
+; CHECK-NEXT:    flat_store_byte v[0:1], v17 offset:4
+; CHECK-NEXT:    flat_store_byte v[0:1], v16 offset:3
+; CHECK-NEXT:    flat_store_byte v[0:1], v15 offset:2
+; CHECK-NEXT:    flat_store_byte v[0:1], v14 offset:1
+; CHECK-NEXT:    flat_store_byte v[0:1], v13
+; CHECK-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.memcpy.p0.p3.i64(ptr %generic, ptr addrspace(3) @shared, i64 128, i1 false)
+  ret void
+}
+
+declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p0.p5.i64(ptr noalias nocapture writeonly, ptr addrspace(5) noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) noalias nocapture writeonly, ptr addrspace(1) noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p1.p4.i64(ptr addrspace(1) noalias nocapture writeonly, ptr addrspace(4) noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) noalias nocapture writeonly, ptr addrspace(4) noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p3.p4.i64(ptr addrspace(3) noalias nocapture writeonly, ptr addrspace(4) noalias nocapture readonly, i64, i1 immarg) #2
+
+declare void @llvm.memcpy.p0.p3.i64(ptr noalias nocapture writeonly, ptr addrspace(3) noalias nocapture readonly, i64, i1 immarg) #2
+
+attributes #0 = { minsize }
+attributes #1 = { optsize }
+attributes #2 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }

From 6ab5927238395798c2e8b657bb59a0304046099d Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 06:13:03 -0700
Subject: [PATCH 588/886] [SLP][NFC]Add a test with the incorrect vectorization
 of smax with unsigned arg.

---
 .../RISCV/smax-unsigned-operand.ll            | 31 +++++++++++++++++++
 1 file changed, 31 insertions(+)
 create mode 100644 llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll

diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll
new file mode 100644
index 0000000000000..577d995f6a765
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll
@@ -0,0 +1,31 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S --passes=slp-vectorizer -mtriple=riscv64-unknown-linux-gnu -mattr=+v < %s | FileCheck %s
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+target triple = "riscv64-unknown-linux-gnu"
+
+@e = global [2 x i8] zeroinitializer
+
+define void @main(ptr noalias %p) {
+; CHECK-LABEL: define void @main(
+; CHECK-SAME: ptr noalias [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[P]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> <i32 0, i32 poison>, i32 [[TMP0]], i32 1
+; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.smax.v2i32(<2 x i32> [[TMP1]], <2 x i32> <i32 0, i32 1>)
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc <2 x i32> [[TMP2]] to <2 x i8>
+; CHECK-NEXT:    store <2 x i8> [[TMP3]], ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 4), align 1
+; CHECK-NEXT:    ret void
+;
+bb:
+  %conv.4 = zext i32 0 to i64
+  %cond.4 = tail call i64 @llvm.smax.i64(i64 %conv.4, i64 0)
+  %conv5.4 = trunc i64 %cond.4 to i8
+  store i8 %conv5.4, ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 4), align 1
+  %0 = load i32, ptr %p, align 4
+  %conv.5 = zext i32 %0 to i64
+  %cond.5 = tail call i64 @llvm.smax.i64(i64 %conv.5, i64 1)
+  %conv5.5 = trunc i64 %cond.5 to i8
+  store i8 %conv5.5, ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 5), align 1
+  ret void
+}
+

From 26ebe16d78b22329d602db0398ce163ad610b0dc Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 06:42:15 -0700
Subject: [PATCH 589/886] [SLP]Fix PR88834: check if unsigned arg can be
 trunced, being used in smax/smin intrinsics.

Need to check that unsigned argument can be safely used in smax/smin
intrinsics by checking if at least single sign bit is cleared, otherwise
its value may be treated as negative instead of positive.
---
 llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp      |  8 +++++++-
 .../SLPVectorizer/RISCV/smax-unsigned-operand.ll     | 12 ++++++++----
 2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index d0bcdceae392b..0cd3ca32933ca 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -14639,10 +14639,16 @@ bool BoUpSLP::collectValuesToDemote(
         assert((ID == Intrinsic::smin || ID == Intrinsic::smax) &&
                "Expected min/max intrinsics only.");
         unsigned SignBits = OrigBitWidth - BitWidth;
+        APInt Mask = APInt::getBitsSetFrom(OrigBitWidth, BitWidth - 1);
         return SignBits <= ComputeNumSignBits(I->getOperand(0), *DL, 0, AC,
                                               nullptr, DT) &&
+               (!isKnownNonNegative(I->getOperand(0), SimplifyQuery(*DL)) ||
+                MaskedValueIsZero(I->getOperand(0), Mask,
+                                  SimplifyQuery(*DL))) &&
                SignBits <= ComputeNumSignBits(I->getOperand(1), *DL, 0, AC,
-                                              nullptr, DT);
+                                              nullptr, DT) &&
+               (!isKnownNonNegative(I->getOperand(1), SimplifyQuery(*DL)) ||
+                MaskedValueIsZero(I->getOperand(1), Mask, SimplifyQuery(*DL)));
       });
     };
     if (ID != Intrinsic::abs) {
diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll
index 577d995f6a765..5db148ac1b485 100644
--- a/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/smax-unsigned-operand.ll
@@ -9,11 +9,15 @@ define void @main(ptr noalias %p) {
 ; CHECK-LABEL: define void @main(
 ; CHECK-SAME: ptr noalias [[P:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[CONV_4:%.*]] = zext i32 0 to i64
+; CHECK-NEXT:    [[COND_4:%.*]] = tail call i64 @llvm.smax.i64(i64 [[CONV_4]], i64 0)
+; CHECK-NEXT:    [[CONV5_4:%.*]] = trunc i64 [[COND_4]] to i8
+; CHECK-NEXT:    store i8 [[CONV5_4]], ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 4), align 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[P]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> <i32 0, i32 poison>, i32 [[TMP0]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.smax.v2i32(<2 x i32> [[TMP1]], <2 x i32> <i32 0, i32 1>)
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc <2 x i32> [[TMP2]] to <2 x i8>
-; CHECK-NEXT:    store <2 x i8> [[TMP3]], ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 4), align 1
+; CHECK-NEXT:    [[CONV_5:%.*]] = zext i32 [[TMP0]] to i64
+; CHECK-NEXT:    [[COND_5:%.*]] = tail call i64 @llvm.smax.i64(i64 [[CONV_5]], i64 1)
+; CHECK-NEXT:    [[CONV5_5:%.*]] = trunc i64 [[COND_5]] to i8
+; CHECK-NEXT:    store i8 [[CONV5_5]], ptr getelementptr inbounds ([11 x i8], ptr @e, i64 0, i64 5), align 1
 ; CHECK-NEXT:    ret void
 ;
 bb:

From e185978321db4500d72fb1186476ee4104c5928b Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 16 Apr 2024 14:23:50 +0100
Subject: [PATCH 590/886] [VectorCombine][X86] Regenerate shuffle.ll +
 shuffle-of-casts.ll

Use v4 of UTC to improve regex matching of argument names to fix a filecheck matching in a future patch
---
 .../VectorCombine/X86/shuffle-of-casts.ll     | 130 +++++++-----
 .../Transforms/VectorCombine/X86/shuffle.ll   | 188 ++++++++++--------
 2 files changed, 185 insertions(+), 133 deletions(-)

diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
index 7d9f7e390b9c0..3a5d2095e2b93 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-casts.ll
@@ -1,12 +1,13 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=sse2 | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx2 | FileCheck %s --check-prefixes=CHECK,AVX
 
 ; standard vector concatenations
 
 define <16 x i32> @concat_zext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @concat_zext_v8i16_v16i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-LABEL: define <16 x i32> @concat_zext_v8i16_v16i32(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[R:%.*]] = zext <16 x i16> [[TMP1]] to <16 x i32>
 ; CHECK-NEXT:    ret <16 x i32> [[R]]
 ;
@@ -17,8 +18,9 @@ define <16 x i32> @concat_zext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
 }
 
 define <16 x i32> @concat_zext_nneg_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @concat_zext_nneg_v8i16_v16i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-LABEL: define <16 x i32> @concat_zext_nneg_v8i16_v16i32(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[R:%.*]] = zext nneg <16 x i16> [[TMP1]] to <16 x i32>
 ; CHECK-NEXT:    ret <16 x i32> [[R]]
 ;
@@ -29,14 +31,16 @@ define <16 x i32> @concat_zext_nneg_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
 }
 
 define <16 x i32> @concat_sext_zext_nneg_v8i16_v8i32(<8 x i16> %a0, <8 x i16> %a1) {
-; SSE-LABEL: @concat_sext_zext_nneg_v8i16_v8i32(
-; SSE-NEXT:    [[X0:%.*]] = sext <8 x i16> [[A0:%.*]] to <8 x i32>
-; SSE-NEXT:    [[X1:%.*]] = zext nneg <8 x i16> [[A1:%.*]] to <8 x i32>
+; SSE-LABEL: define <16 x i32> @concat_sext_zext_nneg_v8i16_v8i32(
+; SSE-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[X0:%.*]] = sext <8 x i16> [[A0]] to <8 x i32>
+; SSE-NEXT:    [[X1:%.*]] = zext nneg <8 x i16> [[A1]] to <8 x i32>
 ; SSE-NEXT:    [[R:%.*]] = shufflevector <8 x i32> [[X0]], <8 x i32> [[X1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; SSE-NEXT:    ret <16 x i32> [[R]]
 ;
-; AVX-LABEL: @concat_sext_zext_nneg_v8i16_v8i32(
-; AVX-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX-LABEL: define <16 x i32> @concat_sext_zext_nneg_v8i16_v8i32(
+; AVX-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; AVX-NEXT:    [[R:%.*]] = sext <16 x i16> [[TMP1]] to <16 x i32>
 ; AVX-NEXT:    ret <16 x i32> [[R]]
 ;
@@ -47,8 +51,9 @@ define <16 x i32> @concat_sext_zext_nneg_v8i16_v8i32(<8 x i16> %a0, <8 x i16> %a
 }
 
 define <16 x i32> @concat_sext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @concat_sext_v8i16_v16i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-LABEL: define <16 x i32> @concat_sext_v8i16_v16i32(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[R:%.*]] = sext <16 x i16> [[TMP1]] to <16 x i32>
 ; CHECK-NEXT:    ret <16 x i32> [[R]]
 ;
@@ -59,8 +64,9 @@ define <16 x i32> @concat_sext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
 }
 
 define <8 x i32> @concat_sext_v4i1_v8i32(<4 x i1> %a0, <4 x i1> %a1) {
-; CHECK-LABEL: @concat_sext_v4i1_v8i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i1> [[A0:%.*]], <4 x i1> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <8 x i32> @concat_sext_v4i1_v8i32(
+; CHECK-SAME: <4 x i1> [[A0:%.*]], <4 x i1> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i1> [[A0]], <4 x i1> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = sext <8 x i1> [[TMP1]] to <8 x i32>
 ; CHECK-NEXT:    ret <8 x i32> [[R]]
 ;
@@ -71,8 +77,9 @@ define <8 x i32> @concat_sext_v4i1_v8i32(<4 x i1> %a0, <4 x i1> %a1) {
 }
 
 define <8 x i16> @concat_trunc_v4i32_v8i16(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @concat_trunc_v4i32_v8i16(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <8 x i16> @concat_trunc_v4i32_v8i16(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0]], <4 x i32> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = trunc <8 x i32> [[TMP1]] to <8 x i16>
 ; CHECK-NEXT:    ret <8 x i16> [[R]]
 ;
@@ -83,8 +90,9 @@ define <8 x i16> @concat_trunc_v4i32_v8i16(<4 x i32> %a0, <4 x i32> %a1) {
 }
 
 define <8 x ptr> @concat_inttoptr_v4i32_v8iptr(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @concat_inttoptr_v4i32_v8iptr(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <8 x ptr> @concat_inttoptr_v4i32_v8iptr(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0]], <4 x i32> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = inttoptr <8 x i32> [[TMP1]] to <8 x ptr>
 ; CHECK-NEXT:    ret <8 x ptr> [[R]]
 ;
@@ -95,8 +103,9 @@ define <8 x ptr> @concat_inttoptr_v4i32_v8iptr(<4 x i32> %a0, <4 x i32> %a1) {
 }
 
 define <16 x i64> @concat_ptrtoint_v8i16_v16i32(<8 x ptr> %a0, <8 x ptr> %a1) {
-; CHECK-LABEL: @concat_ptrtoint_v8i16_v16i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x ptr> [[A0:%.*]], <8 x ptr> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-LABEL: define <16 x i64> @concat_ptrtoint_v8i16_v16i32(
+; CHECK-SAME: <8 x ptr> [[A0:%.*]], <8 x ptr> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x ptr> [[A0]], <8 x ptr> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[R:%.*]] = ptrtoint <16 x ptr> [[TMP1]] to <16 x i64>
 ; CHECK-NEXT:    ret <16 x i64> [[R]]
 ;
@@ -107,14 +116,16 @@ define <16 x i64> @concat_ptrtoint_v8i16_v16i32(<8 x ptr> %a0, <8 x ptr> %a1) {
 }
 
 define <8 x double> @concat_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %a1) {
-; SSE-LABEL: @concat_fpext_v4f32_v8f64(
-; SSE-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE-LABEL: define <8 x double> @concat_fpext_v4f32_v8f64(
+; SSE-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[A0]], <4 x float> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; SSE-NEXT:    [[R:%.*]] = fpext <8 x float> [[TMP1]] to <8 x double>
 ; SSE-NEXT:    ret <8 x double> [[R]]
 ;
-; AVX-LABEL: @concat_fpext_v4f32_v8f64(
-; AVX-NEXT:    [[X0:%.*]] = fpext <4 x float> [[A0:%.*]] to <4 x double>
-; AVX-NEXT:    [[X1:%.*]] = fpext <4 x float> [[A1:%.*]] to <4 x double>
+; AVX-LABEL: define <8 x double> @concat_fpext_v4f32_v8f64(
+; AVX-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[X0:%.*]] = fpext <4 x float> [[A0]] to <4 x double>
+; AVX-NEXT:    [[X1:%.*]] = fpext <4 x float> [[A1]] to <4 x double>
 ; AVX-NEXT:    [[R:%.*]] = shufflevector <4 x double> [[X0]], <4 x double> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; AVX-NEXT:    ret <8 x double> [[R]]
 ;
@@ -125,9 +136,10 @@ define <8 x double> @concat_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %a1)
 }
 
 define <16 x float> @concat_fptrunc_v8f64_v16f32(<8 x double> %a0, <8 x double> %a1) {
-; CHECK-LABEL: @concat_fptrunc_v8f64_v16f32(
-; CHECK-NEXT:    [[X0:%.*]] = fptrunc <8 x double> [[A0:%.*]] to <8 x float>
-; CHECK-NEXT:    [[X1:%.*]] = fptrunc <8 x double> [[A1:%.*]] to <8 x float>
+; CHECK-LABEL: define <16 x float> @concat_fptrunc_v8f64_v16f32(
+; CHECK-SAME: <8 x double> [[A0:%.*]], <8 x double> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = fptrunc <8 x double> [[A0]] to <8 x float>
+; CHECK-NEXT:    [[X1:%.*]] = fptrunc <8 x double> [[A1]] to <8 x float>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x float> [[X0]], <8 x float> [[X1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    ret <16 x float> [[R]]
 ;
@@ -140,8 +152,9 @@ define <16 x float> @concat_fptrunc_v8f64_v16f32(<8 x double> %a0, <8 x double>
 ; commuted vector concatenation
 
 define <16 x i32> @rconcat_sext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @rconcat_sext_v8i16_v16i32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <16 x i32> @rconcat_sext_v8i16_v16i32(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = sext <16 x i16> [[TMP1]] to <16 x i32>
 ; CHECK-NEXT:    ret <16 x i32> [[R]]
 ;
@@ -154,8 +167,9 @@ define <16 x i32> @rconcat_sext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
 ; interleaved shuffle
 
 define <8 x double> @interleave_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %a1) {
-; CHECK-LABEL: @interleave_fpext_v4f32_v8f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-LABEL: define <8 x double> @interleave_fpext_v4f32_v8f64(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[A0]], <4 x float> [[A1]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = fpext <8 x float> [[TMP1]] to <8 x double>
 ; CHECK-NEXT:    ret <8 x double> [[R]]
 ;
@@ -168,8 +182,9 @@ define <8 x double> @interleave_fpext_v4f32_v8f64(<4 x float> %a0, <4 x float> %
 ; bitcasts (same element count)
 
 define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @concat_bitcast_v4i32_v8f32(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <8 x float> @concat_bitcast_v4i32_v8f32(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0]], <4 x i32> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <8 x i32> [[TMP1]] to <8 x float>
 ; CHECK-NEXT:    ret <8 x float> [[R]]
 ;
@@ -182,8 +197,9 @@ define <8 x float> @concat_bitcast_v4i32_v8f32(<4 x i32> %a0, <4 x i32> %a1) {
 ; bitcasts (lower element count)
 
 define <4 x double> @concat_bitcast_v8i16_v4f64(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @concat_bitcast_v8i16_v4f64(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-LABEL: define <4 x double> @concat_bitcast_v8i16_v4f64(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i16> [[A0]], <8 x i16> [[A1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <16 x i16> [[TMP1]] to <4 x double>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
@@ -196,8 +212,9 @@ define <4 x double> @concat_bitcast_v8i16_v4f64(<8 x i16> %a0, <8 x i16> %a1) {
 ; bitcasts (higher element count)
 
 define <16 x i16> @concat_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @concat_bitcast_v4i32_v16i16(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <16 x i16> @concat_bitcast_v4i32_v16i16(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[A0]], <4 x i32> [[A1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <8 x i32> [[TMP1]] to <16 x i16>
 ; CHECK-NEXT:    ret <16 x i16> [[R]]
 ;
@@ -210,11 +227,12 @@ define <16 x i16> @concat_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
 ; negative - multiuse
 
 define <8 x i16> @concat_trunc_v4i32_v8i16_multiuse(<4 x i32> %a0, <4 x i32> %a1, ptr %a2) {
-; CHECK-LABEL: @concat_trunc_v4i32_v8i16_multiuse(
-; CHECK-NEXT:    [[X0:%.*]] = trunc <4 x i32> [[A0:%.*]] to <4 x i16>
-; CHECK-NEXT:    [[X1:%.*]] = trunc <4 x i32> [[A1:%.*]] to <4 x i16>
+; CHECK-LABEL: define <8 x i16> @concat_trunc_v4i32_v8i16_multiuse(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]], ptr [[A2:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = trunc <4 x i32> [[A0]] to <4 x i16>
+; CHECK-NEXT:    [[X1:%.*]] = trunc <4 x i32> [[A1]] to <4 x i16>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i16> [[X0]], <4 x i16> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:    store <4 x i16> [[X0]], ptr [[A2:%.*]], align 8
+; CHECK-NEXT:    store <4 x i16> [[X0]], ptr [[A2]], align 8
 ; CHECK-NEXT:    ret <8 x i16> [[R]]
 ;
   %x0 = trunc <4 x i32> %a0 to <4 x i16>
@@ -227,9 +245,10 @@ define <8 x i16> @concat_trunc_v4i32_v8i16_multiuse(<4 x i32> %a0, <4 x i32> %a1
 ; negative - bitcasts (unscalable higher element count)
 
 define <16 x i16> @revpair_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: @revpair_bitcast_v4i32_v16i16(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0:%.*]] to <8 x i16>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1:%.*]] to <8 x i16>
+; CHECK-LABEL: define <16 x i16> @revpair_bitcast_v4i32_v16i16(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <4 x i32> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <4 x i32> [[A0]] to <8 x i16>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <4 x i32> [[A1]] to <8 x i16>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x i16> [[X0]], <8 x i16> [[X1]], <16 x i32> <i32 1, i32 0, i32 3, i32 3, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
 ; CHECK-NEXT:    ret <16 x i16> [[R]]
 ;
@@ -242,9 +261,10 @@ define <16 x i16> @revpair_bitcast_v4i32_v16i16(<4 x i32> %a0, <4 x i32> %a1) {
 ; negative - bitcasts (unscalable element counts)
 
 define <4 x i32> @shuffle_bitcast_v32i40_v4i32(<32 x i40> %a0, <32 x i40> %a1) {
-; CHECK-LABEL: @shuffle_bitcast_v32i40_v4i32(
-; CHECK-NEXT:    [[X0:%.*]] = bitcast <32 x i40> [[A0:%.*]] to <40 x i32>
-; CHECK-NEXT:    [[X1:%.*]] = bitcast <32 x i40> [[A1:%.*]] to <40 x i32>
+; CHECK-LABEL: define <4 x i32> @shuffle_bitcast_v32i40_v4i32(
+; CHECK-SAME: <32 x i40> [[A0:%.*]], <32 x i40> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = bitcast <32 x i40> [[A0]] to <40 x i32>
+; CHECK-NEXT:    [[X1:%.*]] = bitcast <32 x i40> [[A1]] to <40 x i32>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <40 x i32> [[X0]], <40 x i32> [[X1]], <4 x i32> <i32 0, i32 42, i32 poison, i32 poison>
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
 ;
@@ -257,9 +277,10 @@ define <4 x i32> @shuffle_bitcast_v32i40_v4i32(<32 x i40> %a0, <32 x i40> %a1) {
 ; negative - src type mismatch
 
 define <8 x i32> @concat_sext_v4i8_v4i16_v8i32(<4 x i8> %a0, <4 x i16> %a1) {
-; CHECK-LABEL: @concat_sext_v4i8_v4i16_v8i32(
-; CHECK-NEXT:    [[X0:%.*]] = sext <4 x i8> [[A0:%.*]] to <4 x i32>
-; CHECK-NEXT:    [[X1:%.*]] = sext <4 x i16> [[A1:%.*]] to <4 x i32>
+; CHECK-LABEL: define <8 x i32> @concat_sext_v4i8_v4i16_v8i32(
+; CHECK-SAME: <4 x i8> [[A0:%.*]], <4 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = sext <4 x i8> [[A0]] to <4 x i32>
+; CHECK-NEXT:    [[X1:%.*]] = sext <4 x i16> [[A1]] to <4 x i32>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[X0]], <4 x i32> [[X1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    ret <8 x i32> [[R]]
 ;
@@ -272,9 +293,10 @@ define <8 x i32> @concat_sext_v4i8_v4i16_v8i32(<4 x i8> %a0, <4 x i16> %a1) {
 ; negative - castop mismatch
 
 define <16 x i32> @concat_sext_zext_v8i16_v16i32(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: @concat_sext_zext_v8i16_v16i32(
-; CHECK-NEXT:    [[X0:%.*]] = sext <8 x i16> [[A0:%.*]] to <8 x i32>
-; CHECK-NEXT:    [[X1:%.*]] = zext <8 x i16> [[A1:%.*]] to <8 x i32>
+; CHECK-LABEL: define <16 x i32> @concat_sext_zext_v8i16_v16i32(
+; CHECK-SAME: <8 x i16> [[A0:%.*]], <8 x i16> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[X0:%.*]] = sext <8 x i16> [[A0]] to <8 x i32>
+; CHECK-NEXT:    [[X1:%.*]] = zext <8 x i16> [[A1]] to <8 x i32>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <8 x i32> [[X0]], <8 x i32> [[X1]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
 ; CHECK-NEXT:    ret <16 x i32> [[R]]
 ;
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
index 8337bb37bc549..bb6402347a9b4 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
@@ -1,17 +1,19 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=SSE2 | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=AVX2 | FileCheck %s --check-prefixes=CHECK,AVX
 
 ; x86 does not have a cheap v16i8 shuffle until SSSE3 (pshufb)
 
 define <16 x i8> @bitcast_shuf_narrow_element(<4 x i32> %v) {
-; SSE-LABEL: @bitcast_shuf_narrow_element(
-; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE-LABEL: define <16 x i8> @bitcast_shuf_narrow_element(
+; SSE-SAME: <4 x i32> [[V:%.*]]) #[[ATTR0:[0-9]+]] {
+; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; SSE-NEXT:    [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8>
 ; SSE-NEXT:    ret <16 x i8> [[R]]
 ;
-; AVX-LABEL: @bitcast_shuf_narrow_element(
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8>
+; AVX-LABEL: define <16 x i8> @bitcast_shuf_narrow_element(
+; AVX-SAME: <4 x i32> [[V:%.*]]) #[[ATTR0:[0-9]+]] {
+; AVX-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V]] to <16 x i8>
 ; AVX-NEXT:    [[R:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> poison, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
 ; AVX-NEXT:    ret <16 x i8> [[R]]
 ;
@@ -23,8 +25,9 @@ define <16 x i8> @bitcast_shuf_narrow_element(<4 x i32> %v) {
 ; v4f32 is the same cost as v4i32, so this always works
 
 define <4 x float> @bitcast_shuf_same_size(<4 x i32> %v) {
-; CHECK-LABEL: @bitcast_shuf_same_size(
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <4 x float>
+; CHECK-LABEL: define <4 x float> @bitcast_shuf_same_size(
+; CHECK-SAME: <4 x i32> [[V:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[V]] to <4 x float>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    ret <4 x float> [[R]]
 ;
@@ -36,13 +39,15 @@ define <4 x float> @bitcast_shuf_same_size(<4 x i32> %v) {
 ; Length-changing shuffles
 
 define <16 x i8> @bitcast_shuf_narrow_element_subvector(<2 x i32> %v) {
-; SSE-LABEL: @bitcast_shuf_narrow_element_subvector(
-; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i32> [[V:%.*]], <2 x i32> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0>
+; SSE-LABEL: define <16 x i8> @bitcast_shuf_narrow_element_subvector(
+; SSE-SAME: <2 x i32> [[V:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i32> [[V]], <2 x i32> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0>
 ; SSE-NEXT:    [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8>
 ; SSE-NEXT:    ret <16 x i8> [[R]]
 ;
-; AVX-LABEL: @bitcast_shuf_narrow_element_subvector(
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8>
+; AVX-LABEL: define <16 x i8> @bitcast_shuf_narrow_element_subvector(
+; AVX-SAME: <2 x i32> [[V:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[V]] to <8 x i8>
 ; AVX-NEXT:    [[R:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> poison, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
 ; AVX-NEXT:    ret <16 x i8> [[R]]
 ;
@@ -52,13 +57,15 @@ define <16 x i8> @bitcast_shuf_narrow_element_subvector(<2 x i32> %v) {
 }
 
 define <16 x i16> @bitcast_shuf_narrow_element_concat_subvectors(<2 x i64> %v) {
-; SSE-LABEL: @bitcast_shuf_narrow_element_concat_subvectors(
-; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i64> [[V:%.*]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
+; SSE-LABEL: define <16 x i16> @bitcast_shuf_narrow_element_concat_subvectors(
+; SSE-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i64> [[V]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
 ; SSE-NEXT:    [[R:%.*]] = bitcast <4 x i64> [[SHUF]] to <16 x i16>
 ; SSE-NEXT:    ret <16 x i16> [[R]]
 ;
-; AVX-LABEL: @bitcast_shuf_narrow_element_concat_subvectors(
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <8 x i16>
+; AVX-LABEL: define <16 x i16> @bitcast_shuf_narrow_element_concat_subvectors(
+; AVX-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V]] to <8 x i16>
 ; AVX-NEXT:    [[R:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; AVX-NEXT:    ret <16 x i16> [[R]]
 ;
@@ -68,8 +75,9 @@ define <16 x i16> @bitcast_shuf_narrow_element_concat_subvectors(<2 x i64> %v) {
 }
 
 define <16 x i8> @bitcast_shuf_extract_subvector(<8 x i32> %v) {
-; CHECK-LABEL: @bitcast_shuf_extract_subvector(
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32> [[V:%.*]] to <32 x i8>
+; CHECK-LABEL: define <16 x i8> @bitcast_shuf_extract_subvector(
+; CHECK-SAME: <8 x i32> [[V:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i32> [[V]] to <32 x i8>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <32 x i8> [[TMP1]], <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; CHECK-NEXT:    ret <16 x i8> [[R]]
 ;
@@ -81,8 +89,9 @@ define <16 x i8> @bitcast_shuf_extract_subvector(<8 x i32> %v) {
 ; Negative test - must cast to vector type
 
 define i128 @bitcast_shuf_narrow_element_wrong_type(<4 x i32> %v) {
-; CHECK-LABEL: @bitcast_shuf_narrow_element_wrong_type(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-LABEL: define i128 @bitcast_shuf_narrow_element_wrong_type(
+; CHECK-SAME: <4 x i32> [[V:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to i128
 ; CHECK-NEXT:    ret i128 [[R]]
 ;
@@ -94,8 +103,9 @@ define i128 @bitcast_shuf_narrow_element_wrong_type(<4 x i32> %v) {
 ; Widen shuffle elements
 
 define <4 x i32> @bitcast_shuf_wide_element(<8 x i16> %v) {
-; CHECK-LABEL: @bitcast_shuf_wide_element(
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <4 x i32>
+; CHECK-LABEL: define <4 x i32> @bitcast_shuf_wide_element(
+; CHECK-SAME: <8 x i16> [[V:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[V]] to <4 x i32>
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 1>
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
 ;
@@ -109,8 +119,9 @@ declare void @use(<4 x i32>)
 ; Negative test - don't create an extra shuffle
 
 define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) {
-; CHECK-LABEL: @bitcast_shuf_uses(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-LABEL: define <16 x i8> @bitcast_shuf_uses(
+; CHECK-SAME: <4 x i32> [[V:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i32> [[V]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    call void @use(<4 x i32> [[SHUF]])
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8>
 ; CHECK-NEXT:    ret <16 x i8> [[R]]
@@ -125,8 +136,9 @@ define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) {
 ; TODO - can we remove the empty bitcast(bitcast()) ?
 
 define <4 x i64> @bitcast_shuf_remove_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK-LABEL: @bitcast_shuf_remove_bitcasts(
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-LABEL: define <4 x i64> @bitcast_shuf_remove_bitcasts(
+; CHECK-SAME: <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[A0]], <2 x i64> [[A1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    [[SHUF:%.*]] = bitcast <4 x i64> [[R]] to <8 x i32>
 ; CHECK-NEXT:    [[R1:%.*]] = bitcast <8 x i32> [[SHUF]] to <4 x i64>
 ; CHECK-NEXT:    ret <4 x i64> [[R1]]
@@ -141,9 +153,10 @@ define <4 x i64> @bitcast_shuf_remove_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
 ; shuffle of 2 operands must reduce bitcasts
 
 define <8 x i32> @bitcast_shuf_one_bitcast(<4 x i32> %a0, <2 x i64> %a1) {
-; CHECK-LABEL: @bitcast_shuf_one_bitcast(
-; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[A1:%.*]] to <4 x i32>
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[A0:%.*]], <4 x i32> [[TMP1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-LABEL: define <8 x i32> @bitcast_shuf_one_bitcast(
+; CHECK-SAME: <4 x i32> [[A0:%.*]], <2 x i64> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[A1]] to <4 x i32>
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[A0]], <4 x i32> [[TMP1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    ret <8 x i32> [[R]]
 ;
   %bc0 = bitcast <4 x i32> %a0 to <2 x i64>
@@ -155,8 +168,9 @@ define <8 x i32> @bitcast_shuf_one_bitcast(<4 x i32> %a0, <2 x i64> %a1) {
 ; Negative test - shuffle of 2 operands must not increase bitcasts
 
 define <8 x i32> @bitcast_shuf_too_many_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK-LABEL: @bitcast_shuf_too_many_bitcasts(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-LABEL: define <8 x i32> @bitcast_shuf_too_many_bitcasts(
+; CHECK-SAME: <2 x i64> [[A0:%.*]], <2 x i64> [[A1:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <2 x i64> [[A0]], <2 x i64> [[A1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    [[R:%.*]] = bitcast <4 x i64> [[SHUF]] to <8 x i32>
 ; CHECK-NEXT:    ret <8 x i32> [[R]]
 ;
@@ -166,8 +180,9 @@ define <8 x i32> @bitcast_shuf_too_many_bitcasts(<2 x i64> %a0, <2 x i64> %a1) {
 }
 
 define <2 x i64> @PR35454_1(<2 x i64> %v) {
-; SSE-LABEL: @PR35454_1(
-; SSE-NEXT:    [[BC:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32>
+; SSE-LABEL: define <2 x i64> @PR35454_1(
+; SSE-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[BC:%.*]] = bitcast <2 x i64> [[V]] to <4 x i32>
 ; SSE-NEXT:    [[PERMIL:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; SSE-NEXT:    [[BC1:%.*]] = bitcast <4 x i32> [[PERMIL]] to <16 x i8>
 ; SSE-NEXT:    [[ADD:%.*]] = shl <16 x i8> [[BC1]], <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
@@ -176,8 +191,9 @@ define <2 x i64> @PR35454_1(<2 x i64> %v) {
 ; SSE-NEXT:    [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64>
 ; SSE-NEXT:    ret <2 x i64> [[BC3]]
 ;
-; AVX-LABEL: @PR35454_1(
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <16 x i8>
+; AVX-LABEL: define <2 x i64> @PR35454_1(
+; AVX-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V]] to <16 x i8>
 ; AVX-NEXT:    [[BC1:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> poison, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>
 ; AVX-NEXT:    [[ADD:%.*]] = shl <16 x i8> [[BC1]], <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
 ; AVX-NEXT:    [[BC2:%.*]] = bitcast <16 x i8> [[ADD]] to <4 x i32>
@@ -196,8 +212,9 @@ define <2 x i64> @PR35454_1(<2 x i64> %v) {
 }
 
 define <2 x i64> @PR35454_2(<2 x i64> %v) {
-; SSE-LABEL: @PR35454_2(
-; SSE-NEXT:    [[BC:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32>
+; SSE-LABEL: define <2 x i64> @PR35454_2(
+; SSE-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; SSE-NEXT:    [[BC:%.*]] = bitcast <2 x i64> [[V]] to <4 x i32>
 ; SSE-NEXT:    [[PERMIL:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; SSE-NEXT:    [[BC1:%.*]] = bitcast <4 x i32> [[PERMIL]] to <8 x i16>
 ; SSE-NEXT:    [[ADD:%.*]] = shl <8 x i16> [[BC1]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
@@ -206,8 +223,9 @@ define <2 x i64> @PR35454_2(<2 x i64> %v) {
 ; SSE-NEXT:    [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64>
 ; SSE-NEXT:    ret <2 x i64> [[BC3]]
 ;
-; AVX-LABEL: @PR35454_2(
-; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <8 x i16>
+; AVX-LABEL: define <2 x i64> @PR35454_2(
+; AVX-SAME: <2 x i64> [[V:%.*]]) #[[ATTR0]] {
+; AVX-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[V]] to <8 x i16>
 ; AVX-NEXT:    [[BC1:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> poison, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 2, i32 3, i32 0, i32 1>
 ; AVX-NEXT:    [[ADD:%.*]] = shl <8 x i16> [[BC1]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
 ; AVX-NEXT:    [[BC2:%.*]] = bitcast <8 x i16> [[ADD]] to <4 x i32>
@@ -228,10 +246,11 @@ define <2 x i64> @PR35454_2(<2 x i64> %v) {
 ; Shuffle is much cheaper than fdiv. FMF are intersected.
 
 define <4 x float> @shuf_fdiv_v4f32_yy(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: @shuf_fdiv_v4f32_yy(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[X:%.*]], <4 x float> [[Z:%.*]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[Y:%.*]], <4 x float> poison, <4 x i32> <i32 1, i32 3, i32 1, i32 3>
-; CHECK-NEXT:    [[R:%.*]] = fdiv arcp <4 x float> [[TMP1]], [[TMP2]]
+; CHECK-LABEL: define <4 x float> @shuf_fdiv_v4f32_yy(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> [[Z]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> poison, <4 x i32> <i32 1, i32 3, i32 1, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = fdiv arcp <4 x float> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    ret <4 x float> [[R]]
 ;
   %b0 = fdiv fast <4 x float> %x, %y
@@ -243,11 +262,12 @@ define <4 x float> @shuf_fdiv_v4f32_yy(<4 x float> %x, <4 x float> %y, <4 x floa
 ; Common operand is op0 of the binops.
 
 define <4 x i32> @shuf_add_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: @shuf_add_v4i32_xx(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X:%.*]], <4 x i32> poison, <4 x i32> <i32 poison, i32 poison, i32 2, i32 0>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]], <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
-; CHECK-NEXT:    [[R:%.*]] = add <4 x i32> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    ret <4 x i32> [[R]]
+; CHECK-LABEL: define <4 x i32> @shuf_add_v4i32_xx(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 poison, i32 poison, i32 2, i32 0>
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
+; CHECK-NEXT:    [[R2:%.*]] = add <4 x i32> [[TMP1]], [[R1]]
+; CHECK-NEXT:    ret <4 x i32> [[R2]]
 ;
   %b0 = add <4 x i32> %x, %y
   %b1 = add <4 x i32> %x, %z
@@ -258,9 +278,10 @@ define <4 x i32> @shuf_add_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
 ; For commutative instructions, common operand may be swapped.
 
 define <4 x float> @shuf_fmul_v4f32_xx_swap(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: @shuf_fmul_v4f32_xx_swap(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]], <4 x i32> <i32 0, i32 3, i32 4, i32 7>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X:%.*]], <4 x float> poison, <4 x i32> <i32 0, i32 3, i32 0, i32 3>
+; CHECK-LABEL: define <4 x float> @shuf_fmul_v4f32_xx_swap(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> [[Z]], <4 x i32> <i32 0, i32 3, i32 4, i32 7>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> poison, <4 x i32> <i32 0, i32 3, i32 0, i32 3>
 ; CHECK-NEXT:    [[R:%.*]] = fmul <4 x float> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <4 x float> [[R]]
 ;
@@ -273,9 +294,10 @@ define <4 x float> @shuf_fmul_v4f32_xx_swap(<4 x float> %x, <4 x float> %y, <4 x
 ; For commutative instructions, common operand may be swapped.
 
 define <2 x i64> @shuf_and_v2i64_yy_swap(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
-; CHECK-LABEL: @shuf_and_v2i64_yy_swap(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i64> [[Y:%.*]], <2 x i64> poison, <2 x i32> <i32 1, i32 0>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i64> [[X:%.*]], <2 x i64> [[Z:%.*]], <2 x i32> <i32 3, i32 0>
+; CHECK-LABEL: define <2 x i64> @shuf_and_v2i64_yy_swap(
+; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i64> [[Y]], <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i64> [[X]], <2 x i64> [[Z]], <2 x i32> <i32 3, i32 0>
 ; CHECK-NEXT:    [[R:%.*]] = and <2 x i64> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <2 x i64> [[R]]
 ;
@@ -288,9 +310,10 @@ define <2 x i64> @shuf_and_v2i64_yy_swap(<2 x i64> %x, <2 x i64> %y, <2 x i64> %
 ; non-commutative binop, but common op0
 
 define <4 x i32> @shuf_shl_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: @shuf_shl_v4i32_xx(
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 1, i32 1, i32 2>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]], <4 x i32> <i32 3, i32 1, i32 1, i32 6>
+; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 3, i32 1, i32 1, i32 2>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 3, i32 1, i32 1, i32 6>
 ; CHECK-NEXT:    [[R:%.*]] = shl <4 x i32> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
 ;
@@ -303,11 +326,12 @@ define <4 x i32> @shuf_shl_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
 ; negative test - common operand, but not commutable
 
 define <4 x i32> @shuf_shl_v4i32_xx_swap(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: @shuf_shl_v4i32_xx_swap(
-; CHECK-NEXT:    [[B0:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = shl <4 x i32> [[Z:%.*]], [[X]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 3, i32 2, i32 2, i32 5>
-; CHECK-NEXT:    ret <4 x i32> [[R]]
+; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx_swap(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = shl <4 x i32> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = shl <4 x i32> [[Z]], [[X]]
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 3, i32 2, i32 2, i32 5>
+; CHECK-NEXT:    ret <4 x i32> [[R1]]
 ;
   %b0 = shl <4 x i32> %x, %y
   %b1 = shl <4 x i32> %z, %x
@@ -318,9 +342,10 @@ define <4 x i32> @shuf_shl_v4i32_xx_swap(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
 ; negative test - mismatched opcodes
 
 define <2 x i64> @shuf_sub_add_v2i64_yy(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
-; CHECK-LABEL: @shuf_sub_add_v2i64_yy(
-; CHECK-NEXT:    [[B0:%.*]] = sub <2 x i64> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = add <2 x i64> [[Z:%.*]], [[Y]]
+; CHECK-LABEL: define <2 x i64> @shuf_sub_add_v2i64_yy(
+; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = sub <2 x i64> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = add <2 x i64> [[Z]], [[Y]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[B0]], <2 x i64> [[B1]], <2 x i32> <i32 3, i32 0>
 ; CHECK-NEXT:    ret <2 x i64> [[R]]
 ;
@@ -333,9 +358,10 @@ define <2 x i64> @shuf_sub_add_v2i64_yy(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z
 ; negative test - type change via shuffle
 
 define <8 x float> @shuf_fmul_v4f32_xx_type(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: @shuf_fmul_v4f32_xx_type(
-; CHECK-NEXT:    [[B0:%.*]] = fmul <4 x float> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = fmul <4 x float> [[Z:%.*]], [[X]]
+; CHECK-LABEL: define <8 x float> @shuf_fmul_v4f32_xx_type(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = fmul <4 x float> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = fmul <4 x float> [[Z]], [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <8 x i32> <i32 0, i32 3, i32 4, i32 7, i32 0, i32 1, i32 1, i32 6>
 ; CHECK-NEXT:    ret <8 x float> [[R]]
 ;
@@ -348,10 +374,11 @@ define <8 x float> @shuf_fmul_v4f32_xx_type(<4 x float> %x, <4 x float> %y, <4 x
 ; negative test - uses
 
 define <4 x i32> @shuf_lshr_v4i32_yy_use1(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: @shuf_lshr_v4i32_yy_use1(
-; CHECK-NEXT:    [[B0:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
+; CHECK-LABEL: define <4 x i32> @shuf_lshr_v4i32_yy_use1(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = lshr <4 x i32> [[X]], [[Y]]
 ; CHECK-NEXT:    call void @use(<4 x i32> [[B0]])
-; CHECK-NEXT:    [[B1:%.*]] = lshr <4 x i32> [[Z:%.*]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = lshr <4 x i32> [[Z]], [[Y]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
 ;
@@ -365,9 +392,10 @@ define <4 x i32> @shuf_lshr_v4i32_yy_use1(<4 x i32> %x, <4 x i32> %y, <4 x i32>
 ; negative test - uses
 
 define <4 x i32> @shuf_mul_v4i32_yy_use2(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: @shuf_mul_v4i32_yy_use2(
-; CHECK-NEXT:    [[B0:%.*]] = mul <4 x i32> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = mul <4 x i32> [[Z:%.*]], [[Y]]
+; CHECK-LABEL: define <4 x i32> @shuf_mul_v4i32_yy_use2(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = mul <4 x i32> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = mul <4 x i32> [[Z]], [[Y]]
 ; CHECK-NEXT:    call void @use(<4 x i32> [[B1]])
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
@@ -382,9 +410,10 @@ define <4 x i32> @shuf_mul_v4i32_yy_use2(<4 x i32> %x, <4 x i32> %y, <4 x i32> %
 ; negative test - must have matching operand
 
 define <4 x float> @shuf_fadd_v4f32_no_common_op(<4 x float> %x, <4 x float> %y, <4 x float> %z, <4 x float> %w) {
-; CHECK-LABEL: @shuf_fadd_v4f32_no_common_op(
-; CHECK-NEXT:    [[B0:%.*]] = fadd <4 x float> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = fadd <4 x float> [[Z:%.*]], [[W:%.*]]
+; CHECK-LABEL: define <4 x float> @shuf_fadd_v4f32_no_common_op(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]], <4 x float> [[W:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = fadd <4 x float> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = fadd <4 x float> [[Z]], [[W]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:    ret <4 x float> [[R]]
 ;
@@ -397,9 +426,10 @@ define <4 x float> @shuf_fadd_v4f32_no_common_op(<4 x float> %x, <4 x float> %y,
 ; negative test - binops may be relatively cheap
 
 define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
-; CHECK-LABEL: @shuf_and_v16i16_yy_expensive_shuf(
-; CHECK-NEXT:    [[B0:%.*]] = and <16 x i16> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[B1:%.*]] = and <16 x i16> [[Y]], [[Z:%.*]]
+; CHECK-LABEL: define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(
+; CHECK-SAME: <16 x i16> [[X:%.*]], <16 x i16> [[Y:%.*]], <16 x i16> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = and <16 x i16> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = and <16 x i16> [[Y]], [[Z]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <16 x i16> [[B0]], <16 x i16> [[B1]], <16 x i32> <i32 15, i32 22, i32 25, i32 13, i32 28, i32 0, i32 poison, i32 3, i32 0, i32 30, i32 3, i32 7, i32 9, i32 19, i32 2, i32 22>
 ; CHECK-NEXT:    ret <16 x i16> [[R]]
 ;

From e84b2fb48df882ee1f49bbbca29d44fba22192a8 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 09:48:13 -0400
Subject: [PATCH 591/886] [LV][NFCI]Use integer for cost/trip count
 calculations instead of double, fix possible UB.

Using fp type in the compiler is not the best idea, here it used with
the comparison for equal to 0 and may cause undefined behavior in some
cases.

Reviewers: fhahn

Reviewed By: fhahn

Pull Request: https://github.com/llvm/llvm-project/pull/87241
---
 .../Transforms/Vectorize/LoopVectorize.cpp    | 14 ++---
 .../vectorize-force-tail-with-evl.ll          | 53 +++++++++++++++++++
 2 files changed, 60 insertions(+), 7 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 5535cc55e9321..2057cab46135f 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -9722,7 +9722,7 @@ static bool areRuntimeChecksProfitable(GeneratedRTChecks &Checks,
   }
 
   // The scalar cost should only be 0 when vectorizing with a user specified VF/IC. In those cases, runtime checks should always be generated.
-  double ScalarC = *VF.ScalarCost.getValue();
+  uint64_t ScalarC = *VF.ScalarCost.getValue();
   if (ScalarC == 0)
     return true;
 
@@ -9749,7 +9749,7 @@ static bool areRuntimeChecksProfitable(GeneratedRTChecks &Checks,
   //   RtC + VecC * (TC / VF) + EpiC <  ScalarC * TC
   //
   // Now we can compute the minimum required trip count TC as
-  //   (RtC + EpiC) / (ScalarC - (VecC / VF)) < TC
+  //   VF * (RtC + EpiC) / (ScalarC * VF - VecC) < TC
   //
   // For now we assume the epilogue cost EpiC = 0 for simplicity. Note that
   // the computations are performed on doubles, not integers and the result
@@ -9761,9 +9761,9 @@ static bool areRuntimeChecksProfitable(GeneratedRTChecks &Checks,
       AssumedMinimumVscale = *VScale;
     IntVF *= AssumedMinimumVscale;
   }
-  double VecCOverVF = double(*VF.Cost.getValue()) / IntVF;
-  double RtC = *CheckCost.getValue();
-  double MinTC1 = RtC / (ScalarC - VecCOverVF);
+  uint64_t RtC = *CheckCost.getValue();
+  uint64_t Div = ScalarC * IntVF - *VF.Cost.getValue();
+  uint64_t MinTC1 = Div == 0 ? 0 : divideCeil(RtC * IntVF, Div);
 
   // Second, compute a minimum iteration count so that the cost of the
   // runtime checks is only a fraction of the total scalar loop cost. This
@@ -9772,12 +9772,12 @@ static bool areRuntimeChecksProfitable(GeneratedRTChecks &Checks,
   // * TC. To bound the runtime check to be a fraction 1/X of the scalar
   // cost, compute
   //   RtC < ScalarC * TC * (1 / X)  ==>  RtC * X / ScalarC < TC
-  double MinTC2 = RtC * 10 / ScalarC;
+  uint64_t MinTC2 = divideCeil(RtC * 10, ScalarC);
 
   // Now pick the larger minimum. If it is not a multiple of VF and a scalar
   // epilogue is allowed, choose the next closest multiple of VF. This should
   // partly compensate for ignoring the epilogue cost.
-  uint64_t MinTC = std::ceil(std::max(MinTC1, MinTC2));
+  uint64_t MinTC = std::max(MinTC1, MinTC2);
   if (SEL == CM_ScalarEpilogueAllowed)
     MinTC = alignTo(MinTC, IntVF);
   VF.MinProfitableTripCount = ElementCount::getFixed(MinTC);
diff --git a/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll b/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
index a90b38c6a9605..fe98ca167a089 100644
--- a/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
+++ b/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
@@ -9,6 +9,11 @@
 ; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -force-vector-width=4 \
 ; RUN: -force-target-supports-scalable-vectors -scalable-vectorization=on -S < %s | FileCheck --check-prefix=NO-VP %s
 
+; RUN: opt -passes=loop-vectorize \
+; RUN: -force-tail-folding-style=none \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -force-target-supports-scalable-vectors -scalable-vectorization=on -S < %s | FileCheck --check-prefix=NO-VP-DEF %s
+
 ; The target does not support predicated vectorization.
 define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; IF-EVL-LABEL: @foo(
@@ -80,6 +85,54 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP:       for.cond.cleanup:
 ; NO-VP-NEXT:    ret void
 ;
+; NO-VP-DEF-LABEL: @foo(
+; NO-VP-DEF-NEXT:  entry:
+; NO-VP-DEF-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-DEF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP0]]
+; NO-VP-DEF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-VP-DEF:       vector.ph:
+; NO-VP-DEF-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-DEF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP1]]
+; NO-VP-DEF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-DEF-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-DEF-NEXT:    br label [[VECTOR_BODY:%.*]]
+; NO-VP-DEF:       vector.body:
+; NO-VP-DEF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; NO-VP-DEF-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
+; NO-VP-DEF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP3]]
+; NO-VP-DEF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
+; NO-VP-DEF-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 1 x i32>, ptr [[TMP5]], align 4
+; NO-VP-DEF-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[TMP3]]
+; NO-VP-DEF-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
+; NO-VP-DEF-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 1 x i32>, ptr [[TMP7]], align 4
+; NO-VP-DEF-NEXT:    [[TMP8:%.*]] = add nsw <vscale x 1 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
+; NO-VP-DEF-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP3]]
+; NO-VP-DEF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
+; NO-VP-DEF-NEXT:    store <vscale x 1 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; NO-VP-DEF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
+; NO-VP-DEF-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-DEF-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP-DEF:       middle.block:
+; NO-VP-DEF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-DEF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; NO-VP-DEF:       scalar.ph:
+; NO-VP-DEF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; NO-VP-DEF-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-VP-DEF:       for.body:
+; NO-VP-DEF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-DEF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-DEF-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; NO-VP-DEF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
+; NO-VP-DEF-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
+; NO-VP-DEF-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP12]]
+; NO-VP-DEF-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-DEF-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX4]], align 4
+; NO-VP-DEF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-DEF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-DEF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; NO-VP-DEF:       for.cond.cleanup:
+; NO-VP-DEF-NEXT:    ret void
+;
 entry:
   br label %for.body
 

From 254df2e35c14414564d233902b25305ed17d251c Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 16 Apr 2024 14:52:16 +0100
Subject: [PATCH 592/886] [VectorCombine][X86] shuffle-of-binops.ll - split off
 foldShuffleOfBinops tests from shuffle.ll

---
 .../VectorCombine/X86/shuffle-of-binops.ll    | 204 ++++++++++++++++++
 .../Transforms/VectorCombine/X86/shuffle.ll   | 200 +----------------
 2 files changed, 206 insertions(+), 198 deletions(-)
 create mode 100644 llvm/test/Transforms/VectorCombine/X86/shuffle-of-binops.ll

diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-binops.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-binops.ll
new file mode 100644
index 0000000000000..e2ff343944cf2
--- /dev/null
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-binops.ll
@@ -0,0 +1,204 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=SSE2 | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=AVX2 | FileCheck %s --check-prefixes=CHECK,AVX
+
+declare void @use(<4 x i32>)
+
+; Shuffle is much cheaper than fdiv. FMF are intersected.
+
+define <4 x float> @shuf_fdiv_v4f32_yy(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
+; CHECK-LABEL: define <4 x float> @shuf_fdiv_v4f32_yy(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> [[Z]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> poison, <4 x i32> <i32 1, i32 3, i32 1, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = fdiv arcp <4 x float> [[TMP2]], [[TMP3]]
+; CHECK-NEXT:    ret <4 x float> [[R]]
+;
+  %b0 = fdiv fast <4 x float> %x, %y
+  %b1 = fdiv arcp <4 x float> %z, %y
+  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  ret <4 x float> %r
+}
+
+; Common operand is op0 of the binops.
+
+define <4 x i32> @shuf_add_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
+; CHECK-LABEL: define <4 x i32> @shuf_add_v4i32_xx(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 poison, i32 poison, i32 2, i32 0>
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
+; CHECK-NEXT:    [[R2:%.*]] = add <4 x i32> [[TMP1]], [[R1]]
+; CHECK-NEXT:    ret <4 x i32> [[R2]]
+;
+  %b0 = add <4 x i32> %x, %y
+  %b1 = add <4 x i32> %x, %z
+  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
+  ret <4 x i32> %r
+}
+
+; For commutative instructions, common operand may be swapped.
+
+define <4 x float> @shuf_fmul_v4f32_xx_swap(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
+; CHECK-LABEL: define <4 x float> @shuf_fmul_v4f32_xx_swap(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> [[Z]], <4 x i32> <i32 0, i32 3, i32 4, i32 7>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> poison, <4 x i32> <i32 0, i32 3, i32 0, i32 3>
+; CHECK-NEXT:    [[R:%.*]] = fmul <4 x float> [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    ret <4 x float> [[R]]
+;
+  %b0 = fmul <4 x float> %x, %y
+  %b1 = fmul <4 x float> %z, %x
+  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 0, i32 3, i32 4, i32 7>
+  ret <4 x float> %r
+}
+
+; For commutative instructions, common operand may be swapped.
+
+define <2 x i64> @shuf_and_v2i64_yy_swap(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
+; CHECK-LABEL: define <2 x i64> @shuf_and_v2i64_yy_swap(
+; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i64> [[Y]], <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i64> [[X]], <2 x i64> [[Z]], <2 x i32> <i32 3, i32 0>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i64> [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    ret <2 x i64> [[R]]
+;
+  %b0 = and <2 x i64> %x, %y
+  %b1 = and <2 x i64> %y, %z
+  %r = shufflevector <2 x i64> %b0, <2 x i64> %b1, <2 x i32> <i32 3, i32 0>
+  ret <2 x i64> %r
+}
+
+; non-commutative binop, but common op0
+
+define <4 x i32> @shuf_shl_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
+; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 3, i32 1, i32 1, i32 2>
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 3, i32 1, i32 1, i32 6>
+; CHECK-NEXT:    [[R:%.*]] = shl <4 x i32> [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    ret <4 x i32> [[R]]
+;
+  %b0 = shl <4 x i32> %x, %y
+  %b1 = shl <4 x i32> %x, %z
+  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 3, i32 1, i32 1, i32 6>
+  ret <4 x i32> %r
+}
+
+; negative test - common operand, but not commutable
+
+define <4 x i32> @shuf_shl_v4i32_xx_swap(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
+; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx_swap(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = shl <4 x i32> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = shl <4 x i32> [[Z]], [[X]]
+; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 3, i32 2, i32 2, i32 5>
+; CHECK-NEXT:    ret <4 x i32> [[R1]]
+;
+  %b0 = shl <4 x i32> %x, %y
+  %b1 = shl <4 x i32> %z, %x
+  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 3, i32 2, i32 2, i32 5>
+  ret <4 x i32> %r
+}
+
+; negative test - mismatched opcodes
+
+define <2 x i64> @shuf_sub_add_v2i64_yy(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
+; CHECK-LABEL: define <2 x i64> @shuf_sub_add_v2i64_yy(
+; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = sub <2 x i64> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = add <2 x i64> [[Z]], [[Y]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[B0]], <2 x i64> [[B1]], <2 x i32> <i32 3, i32 0>
+; CHECK-NEXT:    ret <2 x i64> [[R]]
+;
+  %b0 = sub <2 x i64> %x, %y
+  %b1 = add <2 x i64> %z, %y
+  %r = shufflevector <2 x i64> %b0, <2 x i64> %b1, <2 x i32> <i32 3, i32 0>
+  ret <2 x i64> %r
+}
+
+; negative test - type change via shuffle
+
+define <8 x float> @shuf_fmul_v4f32_xx_type(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
+; CHECK-LABEL: define <8 x float> @shuf_fmul_v4f32_xx_type(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = fmul <4 x float> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = fmul <4 x float> [[Z]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <8 x i32> <i32 0, i32 3, i32 4, i32 7, i32 0, i32 1, i32 1, i32 6>
+; CHECK-NEXT:    ret <8 x float> [[R]]
+;
+  %b0 = fmul <4 x float> %x, %y
+  %b1 = fmul <4 x float> %z, %x
+  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <8 x i32> <i32 0, i32 3, i32 4, i32 7, i32 0, i32 1, i32 1, i32 6>
+  ret <8 x float> %r
+}
+
+; negative test - uses
+
+define <4 x i32> @shuf_lshr_v4i32_yy_use1(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
+; CHECK-LABEL: define <4 x i32> @shuf_lshr_v4i32_yy_use1(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = lshr <4 x i32> [[X]], [[Y]]
+; CHECK-NEXT:    call void @use(<4 x i32> [[B0]])
+; CHECK-NEXT:    [[B1:%.*]] = lshr <4 x i32> [[Z]], [[Y]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    ret <4 x i32> [[R]]
+;
+  %b0 = lshr <4 x i32> %x, %y
+  call void @use(<4 x i32> %b0)
+  %b1 = lshr <4 x i32> %z, %y
+  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  ret <4 x i32> %r
+}
+
+; negative test - uses
+
+define <4 x i32> @shuf_mul_v4i32_yy_use2(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
+; CHECK-LABEL: define <4 x i32> @shuf_mul_v4i32_yy_use2(
+; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = mul <4 x i32> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = mul <4 x i32> [[Z]], [[Y]]
+; CHECK-NEXT:    call void @use(<4 x i32> [[B1]])
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    ret <4 x i32> [[R]]
+;
+  %b0 = mul <4 x i32> %x, %y
+  %b1 = mul <4 x i32> %z, %y
+  call void @use(<4 x i32> %b1)
+  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  ret <4 x i32> %r
+}
+
+; negative test - must have matching operand
+
+define <4 x float> @shuf_fadd_v4f32_no_common_op(<4 x float> %x, <4 x float> %y, <4 x float> %z, <4 x float> %w) {
+; CHECK-LABEL: define <4 x float> @shuf_fadd_v4f32_no_common_op(
+; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]], <4 x float> [[W:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = fadd <4 x float> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = fadd <4 x float> [[Z]], [[W]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    ret <4 x float> [[R]]
+;
+  %b0 = fadd <4 x float> %x, %y
+  %b1 = fadd <4 x float> %z, %w
+  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  ret <4 x float> %r
+}
+
+; negative test - binops may be relatively cheap
+
+define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
+; CHECK-LABEL: define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(
+; CHECK-SAME: <16 x i16> [[X:%.*]], <16 x i16> [[Y:%.*]], <16 x i16> [[Z:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[B0:%.*]] = and <16 x i16> [[X]], [[Y]]
+; CHECK-NEXT:    [[B1:%.*]] = and <16 x i16> [[Y]], [[Z]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <16 x i16> [[B0]], <16 x i16> [[B1]], <16 x i32> <i32 15, i32 22, i32 25, i32 13, i32 28, i32 0, i32 poison, i32 3, i32 0, i32 30, i32 3, i32 7, i32 9, i32 19, i32 2, i32 22>
+; CHECK-NEXT:    ret <16 x i16> [[R]]
+;
+  %b0 = and <16 x i16> %x, %y
+  %b1 = and <16 x i16> %y, %z
+  %r = shufflevector <16 x i16> %b0, <16 x i16> %b1, <16 x i32> <i32 15, i32 22, i32 25, i32 13, i32 28, i32 0, i32 poison, i32 3, i32 0, i32 30, i32 3, i32 7, i32 9, i32 19, i32 2, i32 22>
+  ret <16 x i16> %r
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; AVX: {{.*}}
+; SSE: {{.*}}
diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
index bb6402347a9b4..c8c9aa161ae28 100644
--- a/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle.ll
@@ -2,6 +2,8 @@
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=SSE2 | FileCheck %s --check-prefixes=CHECK,SSE
 ; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=AVX2 | FileCheck %s --check-prefixes=CHECK,AVX
 
+declare void @use(<4 x i32>)
+
 ; x86 does not have a cheap v16i8 shuffle until SSSE3 (pshufb)
 
 define <16 x i8> @bitcast_shuf_narrow_element(<4 x i32> %v) {
@@ -114,8 +116,6 @@ define <4 x i32> @bitcast_shuf_wide_element(<8 x i16> %v) {
   ret <4 x i32> %r
 }
 
-declare void @use(<4 x i32>)
-
 ; Negative test - don't create an extra shuffle
 
 define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) {
@@ -242,199 +242,3 @@ define <2 x i64> @PR35454_2(<2 x i64> %v) {
   %bc3 = bitcast <4 x i32> %permil1 to <2 x i64>
   ret <2 x i64> %bc3
 }
-
-; Shuffle is much cheaper than fdiv. FMF are intersected.
-
-define <4 x float> @shuf_fdiv_v4f32_yy(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: define <4 x float> @shuf_fdiv_v4f32_yy(
-; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> [[Z]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> poison, <4 x i32> <i32 1, i32 3, i32 1, i32 3>
-; CHECK-NEXT:    [[R:%.*]] = fdiv arcp <4 x float> [[TMP2]], [[TMP3]]
-; CHECK-NEXT:    ret <4 x float> [[R]]
-;
-  %b0 = fdiv fast <4 x float> %x, %y
-  %b1 = fdiv arcp <4 x float> %z, %y
-  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-  ret <4 x float> %r
-}
-
-; Common operand is op0 of the binops.
-
-define <4 x i32> @shuf_add_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: define <4 x i32> @shuf_add_v4i32_xx(
-; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 poison, i32 poison, i32 2, i32 0>
-; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
-; CHECK-NEXT:    [[R2:%.*]] = add <4 x i32> [[TMP1]], [[R1]]
-; CHECK-NEXT:    ret <4 x i32> [[R2]]
-;
-  %b0 = add <4 x i32> %x, %y
-  %b1 = add <4 x i32> %x, %z
-  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 poison, i32 poison, i32 6, i32 0>
-  ret <4 x i32> %r
-}
-
-; For commutative instructions, common operand may be swapped.
-
-define <4 x float> @shuf_fmul_v4f32_xx_swap(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: define <4 x float> @shuf_fmul_v4f32_xx_swap(
-; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[Y]], <4 x float> [[Z]], <4 x i32> <i32 0, i32 3, i32 4, i32 7>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[X]], <4 x float> poison, <4 x i32> <i32 0, i32 3, i32 0, i32 3>
-; CHECK-NEXT:    [[R:%.*]] = fmul <4 x float> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    ret <4 x float> [[R]]
-;
-  %b0 = fmul <4 x float> %x, %y
-  %b1 = fmul <4 x float> %z, %x
-  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 0, i32 3, i32 4, i32 7>
-  ret <4 x float> %r
-}
-
-; For commutative instructions, common operand may be swapped.
-
-define <2 x i64> @shuf_and_v2i64_yy_swap(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
-; CHECK-LABEL: define <2 x i64> @shuf_and_v2i64_yy_swap(
-; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i64> [[Y]], <2 x i64> poison, <2 x i32> <i32 1, i32 0>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i64> [[X]], <2 x i64> [[Z]], <2 x i32> <i32 3, i32 0>
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i64> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    ret <2 x i64> [[R]]
-;
-  %b0 = and <2 x i64> %x, %y
-  %b1 = and <2 x i64> %y, %z
-  %r = shufflevector <2 x i64> %b0, <2 x i64> %b1, <2 x i32> <i32 3, i32 0>
-  ret <2 x i64> %r
-}
-
-; non-commutative binop, but common op0
-
-define <4 x i32> @shuf_shl_v4i32_xx(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx(
-; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[X]], <4 x i32> poison, <4 x i32> <i32 3, i32 1, i32 1, i32 2>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[Y]], <4 x i32> [[Z]], <4 x i32> <i32 3, i32 1, i32 1, i32 6>
-; CHECK-NEXT:    [[R:%.*]] = shl <4 x i32> [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    ret <4 x i32> [[R]]
-;
-  %b0 = shl <4 x i32> %x, %y
-  %b1 = shl <4 x i32> %x, %z
-  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 3, i32 1, i32 1, i32 6>
-  ret <4 x i32> %r
-}
-
-; negative test - common operand, but not commutable
-
-define <4 x i32> @shuf_shl_v4i32_xx_swap(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: define <4 x i32> @shuf_shl_v4i32_xx_swap(
-; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = shl <4 x i32> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = shl <4 x i32> [[Z]], [[X]]
-; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 3, i32 2, i32 2, i32 5>
-; CHECK-NEXT:    ret <4 x i32> [[R1]]
-;
-  %b0 = shl <4 x i32> %x, %y
-  %b1 = shl <4 x i32> %z, %x
-  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 3, i32 2, i32 2, i32 5>
-  ret <4 x i32> %r
-}
-
-; negative test - mismatched opcodes
-
-define <2 x i64> @shuf_sub_add_v2i64_yy(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
-; CHECK-LABEL: define <2 x i64> @shuf_sub_add_v2i64_yy(
-; CHECK-SAME: <2 x i64> [[X:%.*]], <2 x i64> [[Y:%.*]], <2 x i64> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = sub <2 x i64> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = add <2 x i64> [[Z]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <2 x i64> [[B0]], <2 x i64> [[B1]], <2 x i32> <i32 3, i32 0>
-; CHECK-NEXT:    ret <2 x i64> [[R]]
-;
-  %b0 = sub <2 x i64> %x, %y
-  %b1 = add <2 x i64> %z, %y
-  %r = shufflevector <2 x i64> %b0, <2 x i64> %b1, <2 x i32> <i32 3, i32 0>
-  ret <2 x i64> %r
-}
-
-; negative test - type change via shuffle
-
-define <8 x float> @shuf_fmul_v4f32_xx_type(<4 x float> %x, <4 x float> %y, <4 x float> %z) {
-; CHECK-LABEL: define <8 x float> @shuf_fmul_v4f32_xx_type(
-; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = fmul <4 x float> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = fmul <4 x float> [[Z]], [[X]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <8 x i32> <i32 0, i32 3, i32 4, i32 7, i32 0, i32 1, i32 1, i32 6>
-; CHECK-NEXT:    ret <8 x float> [[R]]
-;
-  %b0 = fmul <4 x float> %x, %y
-  %b1 = fmul <4 x float> %z, %x
-  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <8 x i32> <i32 0, i32 3, i32 4, i32 7, i32 0, i32 1, i32 1, i32 6>
-  ret <8 x float> %r
-}
-
-; negative test - uses
-
-define <4 x i32> @shuf_lshr_v4i32_yy_use1(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: define <4 x i32> @shuf_lshr_v4i32_yy_use1(
-; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = lshr <4 x i32> [[X]], [[Y]]
-; CHECK-NEXT:    call void @use(<4 x i32> [[B0]])
-; CHECK-NEXT:    [[B1:%.*]] = lshr <4 x i32> [[Z]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
-; CHECK-NEXT:    ret <4 x i32> [[R]]
-;
-  %b0 = lshr <4 x i32> %x, %y
-  call void @use(<4 x i32> %b0)
-  %b1 = lshr <4 x i32> %z, %y
-  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
-  ret <4 x i32> %r
-}
-
-; negative test - uses
-
-define <4 x i32> @shuf_mul_v4i32_yy_use2(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
-; CHECK-LABEL: define <4 x i32> @shuf_mul_v4i32_yy_use2(
-; CHECK-SAME: <4 x i32> [[X:%.*]], <4 x i32> [[Y:%.*]], <4 x i32> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = mul <4 x i32> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = mul <4 x i32> [[Z]], [[Y]]
-; CHECK-NEXT:    call void @use(<4 x i32> [[B1]])
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[B0]], <4 x i32> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:    ret <4 x i32> [[R]]
-;
-  %b0 = mul <4 x i32> %x, %y
-  %b1 = mul <4 x i32> %z, %y
-  call void @use(<4 x i32> %b1)
-  %r = shufflevector <4 x i32> %b0, <4 x i32> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-  ret <4 x i32> %r
-}
-
-; negative test - must have matching operand
-
-define <4 x float> @shuf_fadd_v4f32_no_common_op(<4 x float> %x, <4 x float> %y, <4 x float> %z, <4 x float> %w) {
-; CHECK-LABEL: define <4 x float> @shuf_fadd_v4f32_no_common_op(
-; CHECK-SAME: <4 x float> [[X:%.*]], <4 x float> [[Y:%.*]], <4 x float> [[Z:%.*]], <4 x float> [[W:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = fadd <4 x float> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = fadd <4 x float> [[Z]], [[W]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x float> [[B0]], <4 x float> [[B1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:    ret <4 x float> [[R]]
-;
-  %b0 = fadd <4 x float> %x, %y
-  %b1 = fadd <4 x float> %z, %w
-  %r = shufflevector <4 x float> %b0, <4 x float> %b1, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
-  ret <4 x float> %r
-}
-
-; negative test - binops may be relatively cheap
-
-define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
-; CHECK-LABEL: define <16 x i16> @shuf_and_v16i16_yy_expensive_shuf(
-; CHECK-SAME: <16 x i16> [[X:%.*]], <16 x i16> [[Y:%.*]], <16 x i16> [[Z:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:    [[B0:%.*]] = and <16 x i16> [[X]], [[Y]]
-; CHECK-NEXT:    [[B1:%.*]] = and <16 x i16> [[Y]], [[Z]]
-; CHECK-NEXT:    [[R:%.*]] = shufflevector <16 x i16> [[B0]], <16 x i16> [[B1]], <16 x i32> <i32 15, i32 22, i32 25, i32 13, i32 28, i32 0, i32 poison, i32 3, i32 0, i32 30, i32 3, i32 7, i32 9, i32 19, i32 2, i32 22>
-; CHECK-NEXT:    ret <16 x i16> [[R]]
-;
-  %b0 = and <16 x i16> %x, %y
-  %b1 = and <16 x i16> %y, %z
-  %r = shufflevector <16 x i16> %b0, <16 x i16> %b1, <16 x i32> <i32 15, i32 22, i32 25, i32 13, i32 28, i32 0, i32 poison, i32 3, i32 0, i32 30, i32 3, i32 7, i32 9, i32 19, i32 2, i32 22>
-  ret <16 x i16> %r
-}

From 6133878227efc30355c02c2f089e06ce58231a3d Mon Sep 17 00:00:00 2001
From: Erich Keane <ekeane@nvidia.com>
Date: Tue, 16 Apr 2024 06:57:36 -0700
Subject: [PATCH 593/886] [OpenACC] Implement `self` clause for compute
 constructs (#88760)

`self` clauses on compute constructs take an optional condition
expression. We again limit the implementation to ONLY compute constructs
to ensure we get all the rules correct for others. However, this one
will be particularly complicated, as it takes a `var-list` for `update`,
so when we get to that construct/clause combination, we need to do that
as well.

This patch also furthers uses of the `OpenACCClauses.def` as it became
useful while implementing this (as well as some other minor refactors as
I went through).

Finally, `self` and `if` clauses have an interaction with each other, if
an `if` clause evaluates to `true`, the `self` clause has no effect.
While this is intended and can be used 'meaningfully', we are warning on
this with a very granular warning, so that this edge case will be
noticed by newer users, but can be disabled trivially.
---
 clang/include/clang/AST/OpenACCClause.h       |  65 ++++--------
 clang/include/clang/AST/StmtOpenACC.h         |   4 +-
 .../clang/Basic/DiagnosticSemaKinds.td        |   4 +
 clang/include/clang/Basic/OpenACCClauses.def  |   1 +
 clang/include/clang/Basic/OpenACCKinds.h      |   6 ++
 clang/include/clang/Sema/SemaOpenACC.h        |  18 +++-
 clang/lib/AST/OpenACCClause.cpp               |  26 +++++
 clang/lib/AST/StmtProfile.cpp                 |   5 +
 clang/lib/AST/TextNodeDumper.cpp              |   1 +
 clang/lib/Parse/ParseOpenACC.cpp              |  16 ++-
 clang/lib/Sema/SemaOpenACC.cpp                |  68 ++++++++++--
 clang/lib/Sema/TreeTransform.h                | 100 +++++++++++++-----
 clang/lib/Serialization/ASTReader.cpp         |   7 +-
 clang/lib/Serialization/ASTWriter.cpp         |   9 +-
 clang/test/ParserOpenACC/parse-clauses.c      |   5 +-
 .../compute-construct-clause-ast.cpp          |  91 ++++++++++++++++
 .../compute-construct-self-clause.c           |  82 ++++++++++++++
 .../compute-construct-self-clause.cpp         |  99 +++++++++++++++++
 clang/tools/libclang/CIndex.cpp               |   4 +
 19 files changed, 513 insertions(+), 98 deletions(-)
 create mode 100644 clang/test/SemaOpenACC/compute-construct-self-clause.c
 create mode 100644 clang/test/SemaOpenACC/compute-construct-self-clause.cpp

diff --git a/clang/include/clang/AST/OpenACCClause.h b/clang/include/clang/AST/OpenACCClause.h
index 401b8e904a1b7..07587849eb121 100644
--- a/clang/include/clang/AST/OpenACCClause.h
+++ b/clang/include/clang/AST/OpenACCClause.h
@@ -145,6 +145,17 @@ class OpenACCIfClause : public OpenACCClauseWithCondition {
                                  SourceLocation EndLoc);
 };
 
+/// A 'self' clause, which has an optional condition expression.
+class OpenACCSelfClause : public OpenACCClauseWithCondition {
+  OpenACCSelfClause(SourceLocation BeginLoc, SourceLocation LParenLoc,
+                    Expr *ConditionExpr, SourceLocation EndLoc);
+
+public:
+  static OpenACCSelfClause *Create(const ASTContext &C, SourceLocation BeginLoc,
+                                   SourceLocation LParenLoc,
+                                   Expr *ConditionExpr, SourceLocation EndLoc);
+};
+
 template <class Impl> class OpenACCClauseVisitor {
   Impl &getDerived() { return static_cast<Impl &>(*this); }
 
@@ -159,53 +170,13 @@ template <class Impl> class OpenACCClauseVisitor {
       return;
 
     switch (C->getClauseKind()) {
-    case OpenACCClauseKind::Default:
-      VisitDefaultClause(*cast<OpenACCDefaultClause>(C));
-      return;
-    case OpenACCClauseKind::If:
-      VisitIfClause(*cast<OpenACCIfClause>(C));
-      return;
-    case OpenACCClauseKind::Finalize:
-    case OpenACCClauseKind::IfPresent:
-    case OpenACCClauseKind::Seq:
-    case OpenACCClauseKind::Independent:
-    case OpenACCClauseKind::Auto:
-    case OpenACCClauseKind::Worker:
-    case OpenACCClauseKind::Vector:
-    case OpenACCClauseKind::NoHost:
-    case OpenACCClauseKind::Self:
-    case OpenACCClauseKind::Copy:
-    case OpenACCClauseKind::UseDevice:
-    case OpenACCClauseKind::Attach:
-    case OpenACCClauseKind::Delete:
-    case OpenACCClauseKind::Detach:
-    case OpenACCClauseKind::Device:
-    case OpenACCClauseKind::DevicePtr:
-    case OpenACCClauseKind::DeviceResident:
-    case OpenACCClauseKind::FirstPrivate:
-    case OpenACCClauseKind::Host:
-    case OpenACCClauseKind::Link:
-    case OpenACCClauseKind::NoCreate:
-    case OpenACCClauseKind::Present:
-    case OpenACCClauseKind::Private:
-    case OpenACCClauseKind::CopyOut:
-    case OpenACCClauseKind::CopyIn:
-    case OpenACCClauseKind::Create:
-    case OpenACCClauseKind::Reduction:
-    case OpenACCClauseKind::Collapse:
-    case OpenACCClauseKind::Bind:
-    case OpenACCClauseKind::VectorLength:
-    case OpenACCClauseKind::NumGangs:
-    case OpenACCClauseKind::NumWorkers:
-    case OpenACCClauseKind::DeviceNum:
-    case OpenACCClauseKind::DefaultAsync:
-    case OpenACCClauseKind::DeviceType:
-    case OpenACCClauseKind::DType:
-    case OpenACCClauseKind::Async:
-    case OpenACCClauseKind::Tile:
-    case OpenACCClauseKind::Gang:
-    case OpenACCClauseKind::Wait:
-    case OpenACCClauseKind::Invalid:
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  case OpenACCClauseKind::CLAUSE_NAME:                                         \
+    Visit##CLAUSE_NAME##Clause(*cast<OpenACC##CLAUSE_NAME##Clause>(C));        \
+    return;
+#include "clang/Basic/OpenACCClauses.def"
+
+    default:
       llvm_unreachable("Clause visitor not yet implemented");
     }
     llvm_unreachable("Invalid Clause kind");
diff --git a/clang/include/clang/AST/StmtOpenACC.h b/clang/include/clang/AST/StmtOpenACC.h
index 419cb6cada0bc..66f8f844e0b29 100644
--- a/clang/include/clang/AST/StmtOpenACC.h
+++ b/clang/include/clang/AST/StmtOpenACC.h
@@ -142,9 +142,7 @@ class OpenACCComputeConstruct final
                           Stmt *StructuredBlock)
       : OpenACCAssociatedStmtConstruct(OpenACCComputeConstructClass, K, Start,
                                        End, StructuredBlock) {
-    assert((K == OpenACCDirectiveKind::Parallel ||
-            K == OpenACCDirectiveKind::Serial ||
-            K == OpenACCDirectiveKind::Kernels) &&
+    assert(isOpenACCComputeDirectiveKind(K) &&
            "Only parallel, serial, and kernels constructs should be "
            "represented by this type");
 
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 5ec0218aedfe8..44f802c0c28e8 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -12274,4 +12274,8 @@ def note_acc_branch_into_compute_construct
     : Note<"invalid branch into OpenACC Compute Construct">;
 def note_acc_branch_out_of_compute_construct
     : Note<"invalid branch out of OpenACC Compute Construct">;
+def warn_acc_if_self_conflict
+    : Warning<"OpenACC construct 'self' has no effect when an 'if' clause "
+              "evaluates to true">,
+      InGroup<DiagGroup<"openacc-self-if-potential-conflict">>;
 } // end of sema component.
diff --git a/clang/include/clang/Basic/OpenACCClauses.def b/clang/include/clang/Basic/OpenACCClauses.def
index 7fd2720e02ce2..378495d2c0909 100644
--- a/clang/include/clang/Basic/OpenACCClauses.def
+++ b/clang/include/clang/Basic/OpenACCClauses.def
@@ -17,5 +17,6 @@
 
 VISIT_CLAUSE(Default)
 VISIT_CLAUSE(If)
+VISIT_CLAUSE(Self)
 
 #undef VISIT_CLAUSE
diff --git a/clang/include/clang/Basic/OpenACCKinds.h b/clang/include/clang/Basic/OpenACCKinds.h
index 3414df9999170..e3f7417843328 100644
--- a/clang/include/clang/Basic/OpenACCKinds.h
+++ b/clang/include/clang/Basic/OpenACCKinds.h
@@ -146,6 +146,12 @@ inline llvm::raw_ostream &operator<<(llvm::raw_ostream &Out,
   return printOpenACCDirectiveKind(Out, K);
 }
 
+inline bool isOpenACCComputeDirectiveKind(OpenACCDirectiveKind K) {
+  return K == OpenACCDirectiveKind::Parallel ||
+         K == OpenACCDirectiveKind::Serial ||
+         K == OpenACCDirectiveKind::Kernels;
+}
+
 enum class OpenACCAtomicKind {
   Read,
   Write,
diff --git a/clang/include/clang/Sema/SemaOpenACC.h b/clang/include/clang/Sema/SemaOpenACC.h
index c1fe0f5b9c0f6..329dc3945fa2a 100644
--- a/clang/include/clang/Sema/SemaOpenACC.h
+++ b/clang/include/clang/Sema/SemaOpenACC.h
@@ -44,7 +44,8 @@ class SemaOpenACC : public SemaBase {
       Expr *ConditionExpr;
     };
 
-    std::variant<DefaultDetails, ConditionDetails> Details;
+    std::variant<std::monostate, DefaultDetails, ConditionDetails> Details =
+        std::monostate{};
 
   public:
     OpenACCParsedClause(OpenACCDirectiveKind DirKind,
@@ -72,8 +73,17 @@ class SemaOpenACC : public SemaBase {
     }
 
     Expr *getConditionExpr() {
-      assert(ClauseKind == OpenACCClauseKind::If &&
+      assert((ClauseKind == OpenACCClauseKind::If ||
+              (ClauseKind == OpenACCClauseKind::Self &&
+               DirKind != OpenACCDirectiveKind::Update)) &&
              "Parsed clause kind does not have a condition expr");
+
+      // 'self' has an optional ConditionExpr, so be tolerant of that. This will
+      // assert in variant otherwise.
+      if (ClauseKind == OpenACCClauseKind::Self &&
+          std::holds_alternative<std::monostate>(Details))
+        return nullptr;
+
       return std::get<ConditionDetails>(Details).ConditionExpr;
     }
 
@@ -87,7 +97,9 @@ class SemaOpenACC : public SemaBase {
     }
 
     void setConditionDetails(Expr *ConditionExpr) {
-      assert(ClauseKind == OpenACCClauseKind::If &&
+      assert((ClauseKind == OpenACCClauseKind::If ||
+              (ClauseKind == OpenACCClauseKind::Self &&
+               DirKind != OpenACCDirectiveKind::Update)) &&
              "Parsed clause kind does not have a condition expr");
       // In C++ we can count on this being a 'bool', but in C this gets left as
       // some sort of scalar that codegen will have to take care of converting.
diff --git a/clang/lib/AST/OpenACCClause.cpp b/clang/lib/AST/OpenACCClause.cpp
index dcb512cb51417..9c259c8f9bd0a 100644
--- a/clang/lib/AST/OpenACCClause.cpp
+++ b/clang/lib/AST/OpenACCClause.cpp
@@ -48,6 +48,26 @@ OpenACCIfClause::OpenACCIfClause(SourceLocation BeginLoc,
          "Condition expression type not scalar/dependent");
 }
 
+OpenACCSelfClause *OpenACCSelfClause::Create(const ASTContext &C,
+                                             SourceLocation BeginLoc,
+                                             SourceLocation LParenLoc,
+                                             Expr *ConditionExpr,
+                                             SourceLocation EndLoc) {
+  void *Mem = C.Allocate(sizeof(OpenACCIfClause), alignof(OpenACCIfClause));
+  return new (Mem)
+      OpenACCSelfClause(BeginLoc, LParenLoc, ConditionExpr, EndLoc);
+}
+
+OpenACCSelfClause::OpenACCSelfClause(SourceLocation BeginLoc,
+                                     SourceLocation LParenLoc,
+                                     Expr *ConditionExpr, SourceLocation EndLoc)
+    : OpenACCClauseWithCondition(OpenACCClauseKind::Self, BeginLoc, LParenLoc,
+                                 ConditionExpr, EndLoc) {
+  assert((!ConditionExpr || ConditionExpr->isInstantiationDependent() ||
+          ConditionExpr->getType()->isScalarType()) &&
+         "Condition expression type not scalar/dependent");
+}
+
 OpenACCClause::child_range OpenACCClause::children() {
   switch (getClauseKind()) {
   default:
@@ -72,3 +92,9 @@ void OpenACCClausePrinter::VisitDefaultClause(const OpenACCDefaultClause &C) {
 void OpenACCClausePrinter::VisitIfClause(const OpenACCIfClause &C) {
   OS << "if(" << C.getConditionExpr() << ")";
 }
+
+void OpenACCClausePrinter::VisitSelfClause(const OpenACCSelfClause &C) {
+  OS << "self";
+  if (const Expr *CondExpr = C.getConditionExpr())
+    OS << "(" << CondExpr << ")";
+}
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 789e4634bd293..b26d804c6f079 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -2491,6 +2491,11 @@ void OpenACCClauseProfiler::VisitIfClause(const OpenACCIfClause &Clause) {
          "if clause requires a valid condition expr");
   Profiler.VisitStmt(Clause.getConditionExpr());
 }
+
+void OpenACCClauseProfiler::VisitSelfClause(const OpenACCSelfClause &Clause) {
+  if (Clause.hasConditionExpr())
+    Profiler.VisitStmt(Clause.getConditionExpr());
+}
 } // namespace
 
 void StmtProfiler::VisitOpenACCComputeConstruct(
diff --git a/clang/lib/AST/TextNodeDumper.cpp b/clang/lib/AST/TextNodeDumper.cpp
index 688daa64d6197..ff5b3df2d6dfa 100644
--- a/clang/lib/AST/TextNodeDumper.cpp
+++ b/clang/lib/AST/TextNodeDumper.cpp
@@ -398,6 +398,7 @@ void TextNodeDumper::Visit(const OpenACCClause *C) {
       OS << '(' << cast<OpenACCDefaultClause>(C)->getDefaultClauseKind() << ')';
       break;
     case OpenACCClauseKind::If:
+    case OpenACCClauseKind::Self:
       // The condition expression will be printed as a part of the 'children',
       // but print 'clause' here so it is clear what is happening from the dump.
       OS << " clause";
diff --git a/clang/lib/Parse/ParseOpenACC.cpp b/clang/lib/Parse/ParseOpenACC.cpp
index 91f2b8afcf0c2..123be476e928e 100644
--- a/clang/lib/Parse/ParseOpenACC.cpp
+++ b/clang/lib/Parse/ParseOpenACC.cpp
@@ -835,19 +835,23 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
     case OpenACCClauseKind::Default: {
       Token DefKindTok = getCurToken();
 
-      if (expectIdentifierOrKeyword(*this))
-        break;
+      if (expectIdentifierOrKeyword(*this)) {
+        Parens.skipToEnd();
+        return OpenACCCanContinue();
+      }
 
       ConsumeToken();
 
       OpenACCDefaultClauseKind DefKind =
           getOpenACCDefaultClauseKind(DefKindTok);
 
-      if (DefKind == OpenACCDefaultClauseKind::Invalid)
+      if (DefKind == OpenACCDefaultClauseKind::Invalid) {
         Diag(DefKindTok, diag::err_acc_invalid_default_clause_kind);
-      else
-        ParsedClause.setDefaultDetails(DefKind);
+        Parens.skipToEnd();
+        return OpenACCCanContinue();
+      }
 
+      ParsedClause.setDefaultDetails(DefKind);
       break;
     }
     case OpenACCClauseKind::If: {
@@ -977,6 +981,8 @@ Parser::OpenACCClauseParseResult Parser::ParseOpenACCClauseParams(
       case OpenACCClauseKind::Self: {
         assert(DirKind != OpenACCDirectiveKind::Update);
         ExprResult CondExpr = ParseOpenACCConditionExpr();
+        ParsedClause.setConditionDetails(CondExpr.isUsable() ? CondExpr.get()
+                                                             : nullptr);
 
         if (CondExpr.isInvalid()) {
           Parens.skipToEnd();
diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
index 1249136c87650..59f65eaf47a6d 100644
--- a/clang/lib/Sema/SemaOpenACC.cpp
+++ b/clang/lib/Sema/SemaOpenACC.cpp
@@ -15,6 +15,7 @@
 #include "clang/AST/StmtOpenACC.h"
 #include "clang/Basic/DiagnosticSema.h"
 #include "clang/Sema/Sema.h"
+#include "llvm/Support/Casting.h"
 
 using namespace clang;
 
@@ -76,6 +77,19 @@ bool doesClauseApplyToDirective(OpenACCDirectiveKind DirectiveKind,
     default:
       return false;
     }
+  case OpenACCClauseKind::Self:
+    switch (DirectiveKind) {
+    case OpenACCDirectiveKind::Parallel:
+    case OpenACCDirectiveKind::Serial:
+    case OpenACCDirectiveKind::Kernels:
+    case OpenACCDirectiveKind::Update:
+    case OpenACCDirectiveKind::ParallelLoop:
+    case OpenACCDirectiveKind::SerialLoop:
+    case OpenACCDirectiveKind::KernelsLoop:
+      return true;
+    default:
+      return false;
+    }
   default:
     // Do nothing so we can go to the 'unimplemented' diagnostic instead.
     return true;
@@ -121,9 +135,7 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     // Restrictions only properly implemented on 'compute' constructs, and
     // 'compute' constructs are the only construct that can do anything with
     // this yet, so skip/treat as unimplemented in this case.
-    if (Clause.getDirectiveKind() != OpenACCDirectiveKind::Parallel &&
-        Clause.getDirectiveKind() != OpenACCDirectiveKind::Serial &&
-        Clause.getDirectiveKind() != OpenACCDirectiveKind::Kernels)
+    if (!isOpenACCComputeDirectiveKind(Clause.getDirectiveKind()))
       break;
 
     // Don't add an invalid clause to the AST.
@@ -146,9 +158,7 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     // Restrictions only properly implemented on 'compute' constructs, and
     // 'compute' constructs are the only construct that can do anything with
     // this yet, so skip/treat as unimplemented in this case.
-    if (Clause.getDirectiveKind() != OpenACCDirectiveKind::Parallel &&
-        Clause.getDirectiveKind() != OpenACCDirectiveKind::Serial &&
-        Clause.getDirectiveKind() != OpenACCDirectiveKind::Kernels)
+    if (!isOpenACCComputeDirectiveKind(Clause.getDirectiveKind()))
       break;
 
     // There is no prose in the standard that says duplicates aren't allowed,
@@ -160,12 +170,54 @@ SemaOpenACC::ActOnClause(ArrayRef<const OpenACCClause *> ExistingClauses,
     // The parser has ensured that we have a proper condition expr, so there
     // isn't really much to do here.
 
-    // TODO OpenACC: When we implement 'self', this clauses causes us to
-    // 'ignore' the self clause, so we should implement a warning here.
+    // If the 'if' clause is true, it makes the 'self' clause have no effect,
+    // diagnose that here.
+    // TODO OpenACC: When we add these two to other constructs, we might not
+    // want to warn on this (for example, 'update').
+    const auto *Itr =
+        llvm::find_if(ExistingClauses, llvm::IsaPred<OpenACCSelfClause>);
+    if (Itr != ExistingClauses.end()) {
+      Diag(Clause.getBeginLoc(), diag::warn_acc_if_self_conflict);
+      Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+    }
+
     return OpenACCIfClause::Create(
         getASTContext(), Clause.getBeginLoc(), Clause.getLParenLoc(),
         Clause.getConditionExpr(), Clause.getEndLoc());
   }
+
+  case OpenACCClauseKind::Self: {
+    // Restrictions only properly implemented on 'compute' constructs, and
+    // 'compute' constructs are the only construct that can do anything with
+    // this yet, so skip/treat as unimplemented in this case.
+    if (!isOpenACCComputeDirectiveKind(Clause.getDirectiveKind()))
+      break;
+
+    // TODO OpenACC: When we implement this for 'update', this takes a
+    // 'var-list' instead of a condition expression, so semantics/handling has
+    // to happen differently here.
+
+    // There is no prose in the standard that says duplicates aren't allowed,
+    // but this diagnostic is present in other compilers, as well as makes
+    // sense.
+    if (checkAlreadyHasClauseOfKind(*this, ExistingClauses, Clause))
+      return nullptr;
+
+    // If the 'if' clause is true, it makes the 'self' clause have no effect,
+    // diagnose that here.
+    // TODO OpenACC: When we add these two to other constructs, we might not
+    // want to warn on this (for example, 'update').
+    const auto *Itr =
+        llvm::find_if(ExistingClauses, llvm::IsaPred<OpenACCIfClause>);
+    if (Itr != ExistingClauses.end()) {
+      Diag(Clause.getBeginLoc(), diag::warn_acc_if_self_conflict);
+      Diag((*Itr)->getBeginLoc(), diag::note_acc_previous_clause_here);
+    }
+
+    return OpenACCSelfClause::Create(
+        getASTContext(), Clause.getBeginLoc(), Clause.getLParenLoc(),
+        Clause.getConditionExpr(), Clause.getEndLoc());
+  }
   default:
     break;
   }
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index acc2d7ff9d427..0c7fdb357235e 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -11088,6 +11088,77 @@ OMPClause *TreeTransform<Derived>::TransformOMPXBareClause(OMPXBareClause *C) {
 //===----------------------------------------------------------------------===//
 // OpenACC transformation
 //===----------------------------------------------------------------------===//
+namespace {
+template <typename Derived>
+class OpenACCClauseTransform final
+    : public OpenACCClauseVisitor<OpenACCClauseTransform<Derived>> {
+  TreeTransform<Derived> &Self;
+  SemaOpenACC::OpenACCParsedClause &ParsedClause;
+  OpenACCClause *NewClause = nullptr;
+
+public:
+  OpenACCClauseTransform(TreeTransform<Derived> &Self,
+                         SemaOpenACC::OpenACCParsedClause &PC)
+      : Self(Self), ParsedClause(PC) {}
+
+  OpenACCClause *CreatedClause() const { return NewClause; }
+
+#define VISIT_CLAUSE(CLAUSE_NAME)                                              \
+  void Visit##CLAUSE_NAME##Clause(const OpenACC##CLAUSE_NAME##Clause &Clause);
+#include "clang/Basic/OpenACCClauses.def"
+};
+
+template <typename Derived>
+void OpenACCClauseTransform<Derived>::VisitDefaultClause(
+    const OpenACCDefaultClause &C) {
+  ParsedClause.setDefaultDetails(C.getDefaultClauseKind());
+
+  NewClause = OpenACCDefaultClause::Create(
+      Self.getSema().getASTContext(), ParsedClause.getDefaultClauseKind(),
+      ParsedClause.getBeginLoc(), ParsedClause.getLParenLoc(),
+      ParsedClause.getEndLoc());
+}
+
+template <typename Derived>
+void OpenACCClauseTransform<Derived>::VisitIfClause(const OpenACCIfClause &C) {
+  Expr *Cond = const_cast<Expr *>(C.getConditionExpr());
+  assert(Cond && "If constructed with invalid Condition");
+  Sema::ConditionResult Res = Self.TransformCondition(
+      Cond->getExprLoc(), /*Var=*/nullptr, Cond, Sema::ConditionKind::Boolean);
+
+  if (Res.isInvalid() || !Res.get().second)
+    return;
+
+  ParsedClause.setConditionDetails(Res.get().second);
+
+  NewClause = OpenACCIfClause::Create(
+      Self.getSema().getASTContext(), ParsedClause.getBeginLoc(),
+      ParsedClause.getLParenLoc(), ParsedClause.getConditionExpr(),
+      ParsedClause.getEndLoc());
+}
+
+template <typename Derived>
+void OpenACCClauseTransform<Derived>::VisitSelfClause(
+    const OpenACCSelfClause &C) {
+
+  if (C.hasConditionExpr()) {
+    Expr *Cond = const_cast<Expr *>(C.getConditionExpr());
+    Sema::ConditionResult Res =
+        Self.TransformCondition(Cond->getExprLoc(), /*Var=*/nullptr, Cond,
+                                Sema::ConditionKind::Boolean);
+
+    if (Res.isInvalid() || !Res.get().second)
+      return;
+
+    ParsedClause.setConditionDetails(Res.get().second);
+  }
+
+  NewClause = OpenACCSelfClause::Create(
+      Self.getSema().getASTContext(), ParsedClause.getBeginLoc(),
+      ParsedClause.getLParenLoc(), ParsedClause.getConditionExpr(),
+      ParsedClause.getEndLoc());
+}
+} // namespace
 template <typename Derived>
 OpenACCClause *TreeTransform<Derived>::TransformOpenACCClause(
     ArrayRef<const OpenACCClause *> ExistingClauses,
@@ -11100,33 +11171,10 @@ OpenACCClause *TreeTransform<Derived>::TransformOpenACCClause(
   if (const auto *WithParms = dyn_cast<OpenACCClauseWithParams>(OldClause))
     ParsedClause.setLParenLoc(WithParms->getLParenLoc());
 
-  switch (OldClause->getClauseKind()) {
-  case OpenACCClauseKind::Default:
-    // There is nothing to do here as nothing dependent can appear in this
-    // clause. So just set the values so Sema can set the right value.
-    ParsedClause.setDefaultDetails(
-        cast<OpenACCDefaultClause>(OldClause)->getDefaultClauseKind());
-    break;
-  case OpenACCClauseKind::If: {
-    Expr *Cond = const_cast<Expr *>(
-        cast<OpenACCIfClause>(OldClause)->getConditionExpr());
-    assert(Cond && "If constructed with invalid Condition");
-    Sema::ConditionResult Res =
-        TransformCondition(Cond->getExprLoc(), /*Var=*/nullptr, Cond,
-                           Sema::ConditionKind::Boolean);
-
-    if (Res.isInvalid() || !Res.get().second)
-      return nullptr;
-
-    ParsedClause.setConditionDetails(Res.get().second);
-    break;
-  }
-  default:
-    assert(false && "Unhandled OpenACC clause in TreeTransform");
-    return nullptr;
-  }
+  OpenACCClauseTransform<Derived> Transform{*this, ParsedClause};
+  Transform.Visit(OldClause);
 
-  return getSema().OpenACC().ActOnClause(ExistingClauses, ParsedClause);
+  return Transform.CreatedClause();
 }
 
 template <typename Derived>
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index f47d540ea4b86..cf0726460bfca 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -11781,6 +11781,12 @@ OpenACCClause *ASTRecordReader::readOpenACCClause() {
     return OpenACCIfClause::Create(getContext(), BeginLoc, LParenLoc, CondExpr,
                                    EndLoc);
   }
+  case OpenACCClauseKind::Self: {
+    SourceLocation LParenLoc = readSourceLocation();
+    Expr *CondExpr = readBool() ? readSubExpr() : nullptr;
+    return OpenACCSelfClause::Create(getContext(), BeginLoc, LParenLoc,
+                                     CondExpr, EndLoc);
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
@@ -11789,7 +11795,6 @@ OpenACCClause *ASTRecordReader::readOpenACCClause() {
   case OpenACCClauseKind::Worker:
   case OpenACCClauseKind::Vector:
   case OpenACCClauseKind::NoHost:
-  case OpenACCClauseKind::Self:
   case OpenACCClauseKind::Copy:
   case OpenACCClauseKind::UseDevice:
   case OpenACCClauseKind::Attach:
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index ce6fa1feb1eeb..b2a078b6d80f4 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -7524,6 +7524,14 @@ void ASTRecordWriter::writeOpenACCClause(const OpenACCClause *C) {
     AddStmt(const_cast<Expr *>(IC->getConditionExpr()));
     return;
   }
+  case OpenACCClauseKind::Self: {
+    const auto *SC = cast<OpenACCIfClause>(C);
+    writeSourceLocation(SC->getLParenLoc());
+    writeBool(SC->hasConditionExpr());
+    if (SC->hasConditionExpr())
+      AddStmt(const_cast<Expr *>(SC->getConditionExpr()));
+    return;
+  }
   case OpenACCClauseKind::Finalize:
   case OpenACCClauseKind::IfPresent:
   case OpenACCClauseKind::Seq:
@@ -7532,7 +7540,6 @@ void ASTRecordWriter::writeOpenACCClause(const OpenACCClause *C) {
   case OpenACCClauseKind::Worker:
   case OpenACCClauseKind::Vector:
   case OpenACCClauseKind::NoHost:
-  case OpenACCClauseKind::Self:
   case OpenACCClauseKind::Copy:
   case OpenACCClauseKind::UseDevice:
   case OpenACCClauseKind::Attach:
diff --git a/clang/test/ParserOpenACC/parse-clauses.c b/clang/test/ParserOpenACC/parse-clauses.c
index 2369df58308a7..4462f0df540f2 100644
--- a/clang/test/ParserOpenACC/parse-clauses.c
+++ b/clang/test/ParserOpenACC/parse-clauses.c
@@ -376,16 +376,13 @@ void SelfClause() {
 #pragma acc serial self(i > j, seq
   for(;;){}
 
-  // expected-warning@+2{{left operand of comma operator has no effect}}
-  // expected-warning@+1{{OpenACC clause 'self' not yet implemented, clause ignored}}
+  // expected-warning@+1{{left operand of comma operator has no effect}}
 #pragma acc serial self(i, j)
   for(;;){}
 
-  // expected-warning@+1{{OpenACC clause 'self' not yet implemented, clause ignored}}
 #pragma acc serial self(i > j)
   for(;;){}
 
-  // expected-warning@+2{{OpenACC clause 'self' not yet implemented, clause ignored}}
   // expected-warning@+1{{OpenACC clause 'seq' not yet implemented, clause ignored}}
 #pragma acc serial self(1+5>3), seq
   for(;;){}
diff --git a/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
index 018f0b68c7810..6d2efcf81eb6e 100644
--- a/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
+++ b/clang/test/SemaOpenACC/compute-construct-clause-ast.cpp
@@ -110,6 +110,50 @@ void TemplFunc() {
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
 
+#pragma acc serial self
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc kernels self(T::SomeFloat)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc parallel self(T::SomeFloat) if (T::SomeFloat)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+#pragma acc serial if(T::SomeFloat) self(T::SomeFloat)
+  while(true);
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: DependentScopeDeclRefExpr{{.*}} '<dependent type>' lvalue
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'T'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
   // Match the instantiation:
   // CHECK: FunctionDecl{{.*}}TemplFunc{{.*}}implicit_instantiation
   // CHECK-NEXT: TemplateArgument type 'InstTy'
@@ -171,6 +215,53 @@ void TemplFunc() {
   // CHECK-NEXT: WhileStmt
   // CHECK-NEXT: CXXBoolLiteralExpr
   // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}kernels
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}parallel
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
+  // CHECK-NEXT: OpenACCComputeConstruct{{.*}}serial
+  // CHECK-NEXT: if clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: self clause
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'bool' <FloatingToBoolean>
+  // CHECK-NEXT: ImplicitCastExpr{{.*}}'float' <LValueToRValue>
+  // CHECK-NEXT: DeclRefExpr{{.*}} 'const float' lvalue Var{{.*}} 'SomeFloat' 'const float'
+  // CHECK-NEXT: NestedNameSpecifier TypeSpec 'InstTy'
+  // CHECK-NEXT: WhileStmt
+  // CHECK-NEXT: CXXBoolLiteralExpr
+  // CHECK-NEXT: NullStmt
+
 }
 
 struct BoolConversion{ operator bool() const;};
diff --git a/clang/test/SemaOpenACC/compute-construct-self-clause.c b/clang/test/SemaOpenACC/compute-construct-self-clause.c
new file mode 100644
index 0000000000000..fbed2953419a2
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-self-clause.c
@@ -0,0 +1,82 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+void BoolExpr(int *I, float *F) {
+  typedef struct {} SomeStruct;
+  struct C{};
+  // expected-error@+1{{expected expression}}
+#pragma acc parallel self (struct C f())
+  while(0);
+
+  // expected-error@+1{{unexpected type name 'SomeStruct': expected expression}}
+#pragma acc serial self (SomeStruct)
+  while(0);
+
+  // expected-error@+1{{unexpected type name 'SomeStruct': expected expression}}
+#pragma acc serial self (SomeStruct())
+  while(0);
+
+  SomeStruct S;
+  // expected-error@+1{{statement requires expression of scalar type ('SomeStruct' invalid)}}
+#pragma acc serial self (S)
+  while(0);
+
+#pragma acc parallel self (I)
+  while(0);
+
+#pragma acc serial self (F)
+  while(0);
+
+#pragma acc kernels self (*I < *F)
+  while(0);
+}
+
+void WarnMaybeNotUsed(int val1, int val2) {
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self if(val1)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self(val1) if(val1)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(val1) self
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(val1) self(val2)
+  while(0);
+
+  // The below don't warn because one side or the other has an error, thus is
+  // not added to the AST.
+
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel self if(invalid)
+  while(0);
+
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel self(invalid) if(val1)
+  while(0);
+
+  // expected-error@+2{{expected expression}}
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel self() if(invalid)
+  while(0);
+
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel if(invalid) self
+  while(0);
+
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel if(val2) self(invalid)
+  while(0);
+
+  // expected-error@+1{{use of undeclared identifier 'invalid'}}
+#pragma acc parallel if(invalid) self(val1)
+  while(0);
+}
diff --git a/clang/test/SemaOpenACC/compute-construct-self-clause.cpp b/clang/test/SemaOpenACC/compute-construct-self-clause.cpp
new file mode 100644
index 0000000000000..60edbdc2b1191
--- /dev/null
+++ b/clang/test/SemaOpenACC/compute-construct-self-clause.cpp
@@ -0,0 +1,99 @@
+// RUN: %clang_cc1 %s -fopenacc -verify
+
+struct NoBoolConversion{};
+struct BoolConversion{
+  operator bool();
+};
+
+template <typename T, typename U>
+void BoolExpr() {
+  // expected-error@+1{{value of type 'NoBoolConversion' is not contextually convertible to 'bool'}}
+#pragma acc parallel self (NoBoolConversion{})
+  while(0);
+  // expected-error@+2{{no member named 'NotValid' in 'NoBoolConversion'}}
+  // expected-note@#INST{{in instantiation of function template specialization}}
+#pragma acc parallel self (T::NotValid)
+  while(0);
+
+#pragma acc parallel self (BoolConversion{})
+  while(0);
+
+  // expected-error@+1{{value of type 'NoBoolConversion' is not contextually convertible to 'bool'}}
+#pragma acc parallel self (T{})
+  while(0);
+
+#pragma acc parallel self (U{})
+  while(0);
+}
+
+struct HasBool {
+  static constexpr bool B = true;
+};
+
+template<typename T>
+void WarnMaybeNotUsed() {
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self if(T::B)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self(T::B) if(T::B)
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(T::B) self
+  while(0);
+
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(T::B) self(T::B)
+  while(0);
+
+  // We still warn in the cases of dependent failures, since the diagnostic
+  // happens immediately rather than during instantiation.
+
+  // expected-error@+4{{no member named 'Invalid' in 'HasBool'}}
+  // expected-note@#NOT_USED_INST{{in instantiation of function template specialization 'WarnMaybeNotUsed<HasBool>' requested here}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self if(T::Invalid)
+  while(0);
+
+  // expected-error@+3{{no member named 'Invalid' in 'HasBool'}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self(T::Invalid) if(T::B)
+  while(0);
+
+  // expected-error@+3{{no member named 'Invalid' in 'HasBool'}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel self(T::B) if(T::Invalid)
+  while(0);
+
+  // expected-error@+3{{no member named 'Invalid' in 'HasBool'}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(T::Invalid) self
+  while(0);
+
+  // expected-error@+3{{no member named 'Invalid' in 'HasBool'}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(T::Invalid) self(T::B)
+  while(0);
+
+  // expected-error@+3{{no member named 'Invalid' in 'HasBool'}}
+  // expected-warning@+2{{OpenACC construct 'self' has no effect when an 'if' clause evaluates to true}}
+  // expected-note@+1{{previous clause is here}}
+#pragma acc parallel if(T::B) self(T::Invalid)
+  while(0);
+}
+
+void Instantiate() {
+  BoolExpr<NoBoolConversion, BoolConversion>(); // #INST
+  WarnMaybeNotUsed<HasBool>(); // #NOT_USED_INST
+}
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index f304786ff9dff..2ef599d2cd26f 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -2791,6 +2791,10 @@ void OpenACCClauseEnqueue::VisitDefaultClause(const OpenACCDefaultClause &C) {}
 void OpenACCClauseEnqueue::VisitIfClause(const OpenACCIfClause &C) {
   Visitor.AddStmt(C.getConditionExpr());
 }
+void OpenACCClauseEnqueue::VisitSelfClause(const OpenACCSelfClause &C) {
+  if (C.hasConditionExpr())
+    Visitor.AddStmt(C.getConditionExpr());
+}
 } // namespace
 
 void EnqueueVisitor::EnqueueChildren(const OpenACCClause *C) {

From ac791888bbbe58651e597cf7a4b2276424b77a92 Mon Sep 17 00:00:00 2001
From: Benjamin Kramer <benny.kra@googlemail.com>
Date: Tue, 16 Apr 2024 15:58:16 +0200
Subject: [PATCH 594/886] [bazel] Add missing dependency for
 1c076b43c294c7d29d99dd50f6853b33a5b99789

---
 utils/bazel/llvm-project-overlay/mlir/BUILD.bazel | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 653c4bd30600a..03386549a0116 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -11631,6 +11631,7 @@ cc_library(
         ":DialectUtils",
         ":FuncDialect",
         ":IR",
+        ":InferTypeOpInterface",
         ":LinalgDialect",
         ":MathDialect",
         ":Pass",

From a0f8191af3945482f0f7a7c8f030e8c519a795b7 Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 13:12:52 +0100
Subject: [PATCH 595/886] [libclc] Give built bytecode objects a .bc extension.
 NFC

---
 libclc/CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index 7528228b3b7f9..302f559dba7a0 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -326,7 +326,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
         # Generated files are given just as file names, which we must make
         # absolute to the binary directory.
         set( input_file ${CMAKE_CURRENT_BINARY_DIR}/${file} )
-        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${file}.o" )
+        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${file}.bc" )
       else()
         # Other files are originally relative to each SOURCE file, which are
         # then make relative to the libclc root directory. We must normalize
@@ -336,7 +336,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
         get_filename_component( abs_path ${file} ABSOLUTE BASE_DIR ${PROJECT_SOURCE_DIR} )
         file( RELATIVE_PATH root_rel_path ${PROJECT_SOURCE_DIR} ${abs_path} )
         set( input_file ${PROJECT_SOURCE_DIR}/${file} )
-        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${root_rel_path}.o" )
+        set( output_file "${LIBCLC_ARCH_OBJFILE_DIR}/${root_rel_path}.bc" )
       endif()
 
       get_filename_component( file_dir ${file} DIRECTORY )

From 3d118f92081ea0c7048749dc5d08c8e8217be4eb Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 13:59:59 +0100
Subject: [PATCH 596/886] [libclc] Fix dependencies between targets

We need file-level - not target-level - dependencies for these custom
commands to re-trigger when their dependencies change.
---
 libclc/CMakeLists.txt | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index 302f559dba7a0..ed2764847e709 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -364,7 +364,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
       set( spv_suffix ${arch_suffix}.spv )
       add_custom_command( OUTPUT ${spv_suffix}
         COMMAND ${LLVM_SPIRV} ${spvflags} -o ${spv_suffix} ${builtins_link_lib}
-        DEPENDS ${builtins_link_lib_tgt}
+        DEPENDS ${builtins_link_lib}
       )
       add_custom_target( "prepare-${spv_suffix}" ALL DEPENDS "${spv_suffix}" )
       install( FILES ${CMAKE_CURRENT_BINARY_DIR}/${spv_suffix}
@@ -376,7 +376,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
       add_custom_command( OUTPUT ${builtins_opt_lib_tgt}.bc
         COMMAND libclc::opt ${opt_flags} -o ${builtins_opt_lib_tgt}.bc
           ${builtins_link_lib}
-        DEPENDS libclc::opt ${builtins_link_lib_tgt}
+        DEPENDS libclc::opt ${builtins_link_lib}
       )
       add_custom_target( ${builtins_opt_lib_tgt}
         ALL DEPENDS ${builtins_opt_lib_tgt}.bc
@@ -385,12 +385,13 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
         PROPERTIES TARGET_FILE ${builtins_opt_lib_tgt}.bc
       )
 
+      set( builtins_opt_lib $<TARGET_PROPERTY:${builtins_opt_lib_tgt},TARGET_FILE> )
+
       # Add prepare target
       set( obj_suffix ${arch_suffix}.bc )
       add_custom_command( OUTPUT ${obj_suffix}
-        COMMAND prepare_builtins -o ${obj_suffix}
-          $<TARGET_PROPERTY:${builtins_opt_lib_tgt},TARGET_FILE>
-        DEPENDS ${builtins_opt_lib_tgt} prepare_builtins )
+        COMMAND prepare_builtins -o ${obj_suffix} ${builtins_opt_lib}
+        DEPENDS ${builtins_opt_lib} prepare_builtins )
       add_custom_target( prepare-${obj_suffix} ALL DEPENDS ${obj_suffix} )
 
       # nvptx-- targets don't include workitem builtins

From 60de56c743c414240b293a8b8ee10bc2129d7e10 Mon Sep 17 00:00:00 2001
From: Harald van Dijk <harald.vandijk@codeplay.com>
Date: Tue, 16 Apr 2024 15:21:09 +0100
Subject: [PATCH 597/886] [ValueTracking] Restore isKnownNonZero parameter
 order. (#88873)

Prior to #85863, the required parameters of llvm::isKnownNonZero were
Value and DataLayout. After, they are Value, Depth, and SimplifyQuery,
where SimplifyQuery is implicitly constructible from DataLayout. The
change to move Depth before SimplifyQuery needed callers to be updated
unnecessarily, and as commented in #85863, we actually want Depth to be
after SimplifyQuery anyway so that it can be defaulted and the caller
does not need to specify it.
---
 clang/lib/CodeGen/CGCall.cpp                  |   3 +-
 llvm/include/llvm/Analysis/ValueTracking.h    |   2 +-
 llvm/lib/Analysis/BasicAliasAnalysis.cpp      |   3 +-
 llvm/lib/Analysis/InstructionSimplify.cpp     |  25 ++---
 llvm/lib/Analysis/LazyValueInfo.cpp           |   5 +-
 llvm/lib/Analysis/Loads.cpp                   |   4 +-
 llvm/lib/Analysis/ScalarEvolution.cpp         |   2 +-
 llvm/lib/Analysis/ValueTracking.cpp           | 106 +++++++++---------
 llvm/lib/CodeGen/CodeGenPrepare.cpp           |   2 +-
 .../Transforms/IPO/AttributorAttributes.cpp   |   2 +-
 llvm/lib/Transforms/IPO/FunctionAttrs.cpp     |   2 +-
 .../InstCombine/InstCombineAddSub.cpp         |   2 +-
 .../InstCombine/InstCombineAndOrXor.cpp       |   4 +-
 .../InstCombine/InstCombineCalls.cpp          |  11 +-
 .../InstCombine/InstCombineCompares.cpp       |  18 +--
 .../Transforms/InstCombine/InstCombinePHI.cpp |   3 +-
 .../InstCombine/InstructionCombining.cpp      |   2 +-
 .../Instrumentation/MemorySanitizer.cpp       |   4 +-
 .../Utils/PromoteMemoryToRegister.cpp         |   2 +-
 .../lib/Transforms/Utils/SimplifyLibCalls.cpp |  20 ++--
 .../Transforms/Vectorize/VectorCombine.cpp    |   2 +-
 llvm/unittests/Analysis/ValueTrackingTest.cpp |  13 +--
 22 files changed, 113 insertions(+), 124 deletions(-)

diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 7a0bc6fa77b88..3f5463a9a70e9 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -4124,8 +4124,7 @@ static bool isProvablyNull(llvm::Value *addr) {
 }
 
 static bool isProvablyNonNull(Address Addr, CodeGenFunction &CGF) {
-  return llvm::isKnownNonZero(Addr.getBasePointer(), /*Depth=*/0,
-                              CGF.CGM.getDataLayout());
+  return llvm::isKnownNonZero(Addr.getBasePointer(), CGF.CGM.getDataLayout());
 }
 
 /// Emit the actual writing-back of a writeback.
diff --git a/llvm/include/llvm/Analysis/ValueTracking.h b/llvm/include/llvm/Analysis/ValueTracking.h
index 9db0894162afc..e1c41b3b55ccf 100644
--- a/llvm/include/llvm/Analysis/ValueTracking.h
+++ b/llvm/include/llvm/Analysis/ValueTracking.h
@@ -124,7 +124,7 @@ bool isOnlyUsedInZeroEqualityComparison(const Instruction *CxtI);
 /// specified, perform context-sensitive analysis and return true if the
 /// pointer couldn't possibly be null at the specified instruction.
 /// Supports values with integer or pointer type and vectors of integers.
-bool isKnownNonZero(const Value *V, unsigned Depth, const SimplifyQuery &Q);
+bool isKnownNonZero(const Value *V, const SimplifyQuery &Q, unsigned Depth = 0);
 
 /// Return true if the two given values are negation.
 /// Currently can recoginze Value pair:
diff --git a/llvm/lib/Analysis/BasicAliasAnalysis.cpp b/llvm/lib/Analysis/BasicAliasAnalysis.cpp
index b082dfe8fbd21..16ee2ca49d0ec 100644
--- a/llvm/lib/Analysis/BasicAliasAnalysis.cpp
+++ b/llvm/lib/Analysis/BasicAliasAnalysis.cpp
@@ -1283,8 +1283,7 @@ AliasResult BasicAAResult::aliasGEP(
     // VarIndex = Scale*V.
     const VariableGEPIndex &Var = DecompGEP1.VarIndices[0];
     if (Var.Val.TruncBits == 0 &&
-        isKnownNonZero(Var.Val.V, /*Depth=*/0,
-                       SimplifyQuery(DL, DT, &AC, Var.CxtI))) {
+        isKnownNonZero(Var.Val.V, SimplifyQuery(DL, DT, &AC, Var.CxtI))) {
       // Check if abs(V*Scale) >= abs(Scale) holds in the presence of
       // potentially wrapping math.
       auto MultiplyByScaleNoWrap = [](const VariableGEPIndex &Var) {
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 4e6e666922671..8955de6375dec 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -1586,10 +1586,10 @@ static Value *simplifyUnsignedRangeCheck(ICmpInst *ZeroICmp,
     if (match(UnsignedICmp,
               m_c_ICmp(UnsignedPred, m_Specific(Y), m_Specific(A)))) {
       if (UnsignedPred == ICmpInst::ICMP_UGE && IsAnd &&
-          EqPred == ICmpInst::ICMP_NE && isKnownNonZero(B, /*Depth=*/0, Q))
+          EqPred == ICmpInst::ICMP_NE && isKnownNonZero(B, Q))
         return UnsignedICmp;
       if (UnsignedPred == ICmpInst::ICMP_ULT && !IsAnd &&
-          EqPred == ICmpInst::ICMP_EQ && isKnownNonZero(B, /*Depth=*/0, Q))
+          EqPred == ICmpInst::ICMP_EQ && isKnownNonZero(B, Q))
         return UnsignedICmp;
     }
   }
@@ -1607,13 +1607,13 @@ static Value *simplifyUnsignedRangeCheck(ICmpInst *ZeroICmp,
   // X > Y && Y == 0  -->  Y == 0  iff X != 0
   // X > Y || Y == 0  -->  X > Y   iff X != 0
   if (UnsignedPred == ICmpInst::ICMP_UGT && EqPred == ICmpInst::ICMP_EQ &&
-      isKnownNonZero(X, /*Depth=*/0, Q))
+      isKnownNonZero(X, Q))
     return IsAnd ? ZeroICmp : UnsignedICmp;
 
   // X <= Y && Y != 0  -->  X <= Y  iff X != 0
   // X <= Y || Y != 0  -->  Y != 0  iff X != 0
   if (UnsignedPred == ICmpInst::ICMP_ULE && EqPred == ICmpInst::ICMP_NE &&
-      isKnownNonZero(X, /*Depth=*/0, Q))
+      isKnownNonZero(X, Q))
     return IsAnd ? UnsignedICmp : ZeroICmp;
 
   // The transforms below here are expected to be handled more generally with
@@ -2817,10 +2817,9 @@ static Constant *computePointerICmp(CmpInst::Predicate Pred, Value *LHS,
     // the other operand can not be based on the alloc - if it were, then
     // the cmp itself would be a capture.
     Value *MI = nullptr;
-    if (isAllocLikeFn(LHS, TLI) && llvm::isKnownNonZero(RHS, /*Depth=*/0, Q))
+    if (isAllocLikeFn(LHS, TLI) && llvm::isKnownNonZero(RHS, Q))
       MI = LHS;
-    else if (isAllocLikeFn(RHS, TLI) &&
-             llvm::isKnownNonZero(LHS, /*Depth=*/0, Q))
+    else if (isAllocLikeFn(RHS, TLI) && llvm::isKnownNonZero(LHS, Q))
       MI = RHS;
     if (MI) {
       // FIXME: This is incorrect, see PR54002. While we can assume that the
@@ -2976,12 +2975,12 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     return getTrue(ITy);
   case ICmpInst::ICMP_EQ:
   case ICmpInst::ICMP_ULE:
-    if (isKnownNonZero(LHS, /*Depth=*/0, Q))
+    if (isKnownNonZero(LHS, Q))
       return getFalse(ITy);
     break;
   case ICmpInst::ICMP_NE:
   case ICmpInst::ICMP_UGT:
-    if (isKnownNonZero(LHS, /*Depth=*/0, Q))
+    if (isKnownNonZero(LHS, Q))
       return getTrue(ITy);
     break;
   case ICmpInst::ICMP_SLT: {
@@ -2996,7 +2995,7 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     KnownBits LHSKnown = computeKnownBits(LHS, /* Depth */ 0, Q);
     if (LHSKnown.isNegative())
       return getTrue(ITy);
-    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, /*Depth=*/0, Q))
+    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, Q))
       return getFalse(ITy);
     break;
   }
@@ -3012,7 +3011,7 @@ static Value *simplifyICmpWithZero(CmpInst::Predicate Pred, Value *LHS,
     KnownBits LHSKnown = computeKnownBits(LHS, /* Depth */ 0, Q);
     if (LHSKnown.isNegative())
       return getFalse(ITy);
-    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, /*Depth=*/0, Q))
+    if (LHSKnown.isNonNegative() && isKnownNonZero(LHS, Q))
       return getTrue(ITy);
     break;
   }
@@ -3165,7 +3164,7 @@ static Value *simplifyICmpWithBinOpOnLHS(CmpInst::Predicate Pred,
   const APInt *C;
   if ((match(LBO, m_LShr(m_Specific(RHS), m_APInt(C))) && *C != 0) ||
       (match(LBO, m_UDiv(m_Specific(RHS), m_APInt(C))) && *C != 1)) {
-    if (isKnownNonZero(RHS, /*Depth=*/0, Q)) {
+    if (isKnownNonZero(RHS, Q)) {
       switch (Pred) {
       default:
         break;
@@ -3398,7 +3397,7 @@ static Value *simplifyICmpWithBinOp(CmpInst::Predicate Pred, Value *LHS,
       bool NUW = Q.IIQ.hasNoUnsignedWrap(LBO) && Q.IIQ.hasNoUnsignedWrap(RBO);
       bool NSW = Q.IIQ.hasNoSignedWrap(LBO) && Q.IIQ.hasNoSignedWrap(RBO);
       if (!NUW || (ICmpInst::isSigned(Pred) && !NSW) ||
-          !isKnownNonZero(LBO->getOperand(0), /*Depth=*/0, Q))
+          !isKnownNonZero(LBO->getOperand(0), Q))
         break;
       if (Value *V = simplifyICmpInst(Pred, LBO->getOperand(1),
                                       RBO->getOperand(1), Q, MaxRecurse - 1))
diff --git a/llvm/lib/Analysis/LazyValueInfo.cpp b/llvm/lib/Analysis/LazyValueInfo.cpp
index 3223b0564e6c9..6cded828c25f4 100644
--- a/llvm/lib/Analysis/LazyValueInfo.cpp
+++ b/llvm/lib/Analysis/LazyValueInfo.cpp
@@ -645,7 +645,7 @@ LazyValueInfoImpl::solveBlockValueImpl(Value *Val, BasicBlock *BB) {
   // instruction is placed, even if it could legally be hoisted much higher.
   // That is unfortunate.
   PointerType *PT = dyn_cast<PointerType>(BBI->getType());
-  if (PT && isKnownNonZero(BBI, /*Depth=*/0, DL))
+  if (PT && isKnownNonZero(BBI, DL))
     return ValueLatticeElement::getNot(ConstantPointerNull::get(PT));
 
   if (BBI->getType()->isIntegerTy()) {
@@ -1863,8 +1863,7 @@ LazyValueInfo::getPredicateAt(unsigned Pred, Value *V, Constant *C,
   Module *M = CxtI->getModule();
   const DataLayout &DL = M->getDataLayout();
   if (V->getType()->isPointerTy() && C->isNullValue() &&
-      isKnownNonZero(V->stripPointerCastsSameRepresentation(), /*Depth=*/0,
-                     DL)) {
+      isKnownNonZero(V->stripPointerCastsSameRepresentation(), DL)) {
     if (Pred == ICmpInst::ICMP_EQ)
       return LazyValueInfo::False;
     else if (Pred == ICmpInst::ICMP_NE)
diff --git a/llvm/lib/Analysis/Loads.cpp b/llvm/lib/Analysis/Loads.cpp
index b5403408cf2ab..ac508e19c9e01 100644
--- a/llvm/lib/Analysis/Loads.cpp
+++ b/llvm/lib/Analysis/Loads.cpp
@@ -100,7 +100,7 @@ static bool isDereferenceableAndAlignedPointer(
   if (KnownDerefBytes.getBoolValue() && KnownDerefBytes.uge(Size) &&
       !CheckForFreed)
     if (!CheckForNonNull ||
-        isKnownNonZero(V, /*Depth=*/0, SimplifyQuery(DL, DT, AC, CtxI))) {
+        isKnownNonZero(V, SimplifyQuery(DL, DT, AC, CtxI))) {
       // As we recursed through GEPs to get here, we've incrementally checked
       // that each step advanced by a multiple of the alignment. If our base is
       // properly aligned, then the original offset accessed must also be.
@@ -134,7 +134,7 @@ static bool isDereferenceableAndAlignedPointer(
     if (getObjectSize(V, ObjSize, DL, TLI, Opts)) {
       APInt KnownDerefBytes(Size.getBitWidth(), ObjSize);
       if (KnownDerefBytes.getBoolValue() && KnownDerefBytes.uge(Size) &&
-          isKnownNonZero(V, /*Depth=*/0, SimplifyQuery(DL, DT, AC, CtxI)) &&
+          isKnownNonZero(V, SimplifyQuery(DL, DT, AC, CtxI)) &&
           !V->canBeFreed()) {
         // As we recursed through GEPs to get here, we've incrementally
         // checked that each step advanced by a multiple of the alignment. If
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 1c98b0295e525..95440dda3b4c0 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -6900,7 +6900,7 @@ const ConstantRange &ScalarEvolution::getRangeRef(
         uint64_t Rem = MaxVal.urem(Align);
         MaxVal -= APInt(BitWidth, Rem);
         APInt MinVal = APInt::getZero(BitWidth);
-        if (llvm::isKnownNonZero(V, /*Depth=*/0, DL))
+        if (llvm::isKnownNonZero(V, DL))
           MinVal = Align;
         ConservativeResult = ConservativeResult.intersectWith(
             ConstantRange::getNonEmpty(MinVal, MaxVal + 1), RangeType);
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index b3abf016cfb93..e91dc07f31641 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -272,7 +272,7 @@ bool llvm::isKnownToBeAPowerOfTwo(const Value *V, const DataLayout &DL,
 }
 
 static bool isKnownNonZero(const Value *V, const APInt &DemandedElts,
-                           unsigned Depth, const SimplifyQuery &Q);
+                           const SimplifyQuery &Q, unsigned Depth);
 
 bool llvm::isKnownNonNegative(const Value *V, const SimplifyQuery &SQ,
                               unsigned Depth) {
@@ -288,7 +288,7 @@ bool llvm::isKnownPositive(const Value *V, const SimplifyQuery &SQ,
   // this updated.
   KnownBits Known = computeKnownBits(V, Depth, SQ);
   return Known.isNonNegative() &&
-         (Known.isNonZero() || isKnownNonZero(V, Depth, SQ));
+         (Known.isNonZero() || isKnownNonZero(V, SQ, Depth));
 }
 
 bool llvm::isKnownNegative(const Value *V, const SimplifyQuery &SQ,
@@ -868,7 +868,7 @@ static void computeKnownBitsFromShiftOperator(
   bool ShAmtNonZero =
       Known.isNonZero() ||
       (Known.getMaxValue().ult(Known.getBitWidth()) &&
-       isKnownNonZero(I->getOperand(1), DemandedElts, Depth + 1, Q));
+       isKnownNonZero(I->getOperand(1), DemandedElts, Q, Depth + 1));
   Known = KF(Known2, Known, ShAmtNonZero);
 }
 
@@ -2124,7 +2124,7 @@ bool isKnownToBeAPowerOfTwo(const Value *V, bool OrZero, unsigned Depth,
   case Instruction::Mul:
     return isKnownToBeAPowerOfTwo(I->getOperand(1), OrZero, Depth, Q) &&
            isKnownToBeAPowerOfTwo(I->getOperand(0), OrZero, Depth, Q) &&
-           (OrZero || isKnownNonZero(I, Depth, Q));
+           (OrZero || isKnownNonZero(I, Q, Depth));
   case Instruction::And:
     // A power of two and'd with anything is a power of two or zero.
     if (OrZero &&
@@ -2134,7 +2134,7 @@ bool isKnownToBeAPowerOfTwo(const Value *V, bool OrZero, unsigned Depth,
     // X & (-X) is always a power of two or zero.
     if (match(I->getOperand(0), m_Neg(m_Specific(I->getOperand(1)))) ||
         match(I->getOperand(1), m_Neg(m_Specific(I->getOperand(0)))))
-      return OrZero || isKnownNonZero(I->getOperand(0), Depth, Q);
+      return OrZero || isKnownNonZero(I->getOperand(0), Q, Depth);
     return false;
   case Instruction::Add: {
     // Adding a power-of-two or zero to the same power-of-two or zero yields
@@ -2249,7 +2249,7 @@ static bool isGEPKnownNonNull(const GEPOperator *GEP, unsigned Depth,
 
   // If the base pointer is non-null, we cannot walk to a null address with an
   // inbounds GEP in address space zero.
-  if (isKnownNonZero(GEP->getPointerOperand(), Depth, Q))
+  if (isKnownNonZero(GEP->getPointerOperand(), Q, Depth))
     return true;
 
   // Walk the GEP operands and see if any operand introduces a non-zero offset.
@@ -2288,7 +2288,7 @@ static bool isGEPKnownNonNull(const GEPOperator *GEP, unsigned Depth,
     if (Depth++ >= MaxAnalysisRecursionDepth)
       continue;
 
-    if (isKnownNonZero(GTI.getOperand(), Depth, Q))
+    if (isKnownNonZero(GTI.getOperand(), Q, Depth))
       return true;
   }
 
@@ -2441,8 +2441,8 @@ static bool isNonZeroAdd(const APInt &DemandedElts, unsigned Depth,
                          const SimplifyQuery &Q, unsigned BitWidth, Value *X,
                          Value *Y, bool NSW, bool NUW) {
   if (NUW)
-    return isKnownNonZero(Y, DemandedElts, Depth, Q) ||
-           isKnownNonZero(X, DemandedElts, Depth, Q);
+    return isKnownNonZero(Y, DemandedElts, Q, Depth) ||
+           isKnownNonZero(X, DemandedElts, Q, Depth);
 
   KnownBits XKnown = computeKnownBits(X, DemandedElts, Depth, Q);
   KnownBits YKnown = computeKnownBits(Y, DemandedElts, Depth, Q);
@@ -2450,8 +2450,8 @@ static bool isNonZeroAdd(const APInt &DemandedElts, unsigned Depth,
   // If X and Y are both non-negative (as signed values) then their sum is not
   // zero unless both X and Y are zero.
   if (XKnown.isNonNegative() && YKnown.isNonNegative())
-    if (isKnownNonZero(Y, DemandedElts, Depth, Q) ||
-        isKnownNonZero(X, DemandedElts, Depth, Q))
+    if (isKnownNonZero(Y, DemandedElts, Q, Depth) ||
+        isKnownNonZero(X, DemandedElts, Q, Depth))
       return true;
 
   // If X and Y are both negative (as signed values) then their sum is not
@@ -2485,7 +2485,7 @@ static bool isNonZeroSub(const APInt &DemandedElts, unsigned Depth,
                          Value *Y) {
   // TODO: Move this case into isKnownNonEqual().
   if (auto *C = dyn_cast<Constant>(X))
-    if (C->isNullValue() && isKnownNonZero(Y, DemandedElts, Depth, Q))
+    if (C->isNullValue() && isKnownNonZero(Y, DemandedElts, Q, Depth))
       return true;
 
   return ::isKnownNonEqual(X, Y, Depth, Q);
@@ -2497,18 +2497,18 @@ static bool isNonZeroMul(const APInt &DemandedElts, unsigned Depth,
   // If X and Y are non-zero then so is X * Y as long as the multiplication
   // does not overflow.
   if (NSW || NUW)
-    return isKnownNonZero(X, DemandedElts, Depth, Q) &&
-           isKnownNonZero(Y, DemandedElts, Depth, Q);
+    return isKnownNonZero(X, DemandedElts, Q, Depth) &&
+           isKnownNonZero(Y, DemandedElts, Q, Depth);
 
   // If either X or Y is odd, then if the other is non-zero the result can't
   // be zero.
   KnownBits XKnown = computeKnownBits(X, DemandedElts, Depth, Q);
   if (XKnown.One[0])
-    return isKnownNonZero(Y, DemandedElts, Depth, Q);
+    return isKnownNonZero(Y, DemandedElts, Q, Depth);
 
   KnownBits YKnown = computeKnownBits(Y, DemandedElts, Depth, Q);
   if (YKnown.One[0])
-    return XKnown.isNonZero() || isKnownNonZero(X, DemandedElts, Depth, Q);
+    return XKnown.isNonZero() || isKnownNonZero(X, DemandedElts, Q, Depth);
 
   // If there exists any subset of X (sX) and subset of Y (sY) s.t sX * sY is
   // non-zero, then X * Y is non-zero. We can find sX and sY by just taking
@@ -2564,7 +2564,7 @@ static bool isNonZeroShift(const Operator *I, const APInt &DemandedElts,
   // non-zero then at least one non-zero bit must remain.
   if (InvShiftOp(KnownVal.Zero, NumBits - MaxShift)
           .eq(InvShiftOp(APInt::getAllOnes(NumBits), NumBits - MaxShift)) &&
-      isKnownNonZero(I->getOperand(0), DemandedElts, Depth, Q))
+      isKnownNonZero(I->getOperand(0), DemandedElts, Q, Depth))
     return true;
 
   return false;
@@ -2613,7 +2613,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     Type *FromTy = I->getOperand(0)->getType();
     if ((FromTy->isIntOrIntVectorTy() || FromTy->isPtrOrPtrVectorTy()) &&
         (BitWidth % getBitWidth(FromTy->getScalarType(), Q.DL)) == 0)
-      return isKnownNonZero(I->getOperand(0), Depth, Q);
+      return isKnownNonZero(I->getOperand(0), Q, Depth);
   } break;
   case Instruction::IntToPtr:
     // Note that we have to take special care to avoid looking through
@@ -2622,7 +2622,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     if (!isa<ScalableVectorType>(I->getType()) &&
         Q.DL.getTypeSizeInBits(I->getOperand(0)->getType()).getFixedValue() <=
             Q.DL.getTypeSizeInBits(I->getType()).getFixedValue())
-      return isKnownNonZero(I->getOperand(0), Depth, Q);
+      return isKnownNonZero(I->getOperand(0), Q, Depth);
     break;
   case Instruction::PtrToInt:
     // Similar to int2ptr above, we can look through ptr2int here if the cast
@@ -2630,25 +2630,25 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     if (!isa<ScalableVectorType>(I->getType()) &&
         Q.DL.getTypeSizeInBits(I->getOperand(0)->getType()).getFixedValue() <=
             Q.DL.getTypeSizeInBits(I->getType()).getFixedValue())
-      return isKnownNonZero(I->getOperand(0), Depth, Q);
+      return isKnownNonZero(I->getOperand(0), Q, Depth);
     break;
   case Instruction::Sub:
     return isNonZeroSub(DemandedElts, Depth, Q, BitWidth, I->getOperand(0),
                         I->getOperand(1));
   case Instruction::Or:
     // X | Y != 0 if X != 0 or Y != 0.
-    return isKnownNonZero(I->getOperand(1), DemandedElts, Depth, Q) ||
-           isKnownNonZero(I->getOperand(0), DemandedElts, Depth, Q);
+    return isKnownNonZero(I->getOperand(1), DemandedElts, Q, Depth) ||
+           isKnownNonZero(I->getOperand(0), DemandedElts, Q, Depth);
   case Instruction::SExt:
   case Instruction::ZExt:
     // ext X != 0 if X != 0.
-    return isKnownNonZero(I->getOperand(0), Depth, Q);
+    return isKnownNonZero(I->getOperand(0), Q, Depth);
 
   case Instruction::Shl: {
     // shl nsw/nuw can't remove any non-zero bits.
     const OverflowingBinaryOperator *BO = cast<OverflowingBinaryOperator>(I);
     if (Q.IIQ.hasNoUnsignedWrap(BO) || Q.IIQ.hasNoSignedWrap(BO))
-      return isKnownNonZero(I->getOperand(0), Depth, Q);
+      return isKnownNonZero(I->getOperand(0), Q, Depth);
 
     // shl X, Y != 0 if X is odd.  Note that the value of the shift is undefined
     // if the lowest bit is shifted off the end.
@@ -2664,7 +2664,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     // shr exact can only shift out zero bits.
     const PossiblyExactOperator *BO = cast<PossiblyExactOperator>(I);
     if (BO->isExact())
-      return isKnownNonZero(I->getOperand(0), Depth, Q);
+      return isKnownNonZero(I->getOperand(0), Q, Depth);
 
     // shr X, Y != 0 if X is negative.  Note that the value of the shift is not
     // defined if the sign bit is shifted off the end.
@@ -2680,7 +2680,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
     // X / Y
     // div exact can only produce a zero if the dividend is zero.
     if (cast<PossiblyExactOperator>(I)->isExact())
-      return isKnownNonZero(I->getOperand(0), DemandedElts, Depth, Q);
+      return isKnownNonZero(I->getOperand(0), DemandedElts, Q, Depth);
 
     std::optional<bool> XUgeY;
     KnownBits XKnown =
@@ -2730,7 +2730,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       Value *Op;
       Op = IsTrueArm ? I->getOperand(1) : I->getOperand(2);
       // Op is trivially non-zero.
-      if (isKnownNonZero(Op, DemandedElts, Depth, Q))
+      if (isKnownNonZero(Op, DemandedElts, Q, Depth))
         return true;
 
       // The condition of the select dominates the true/false arm. Check if the
@@ -2780,7 +2780,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
         }
       }
       // Finally recurse on the edge and check it directly.
-      return isKnownNonZero(U.get(), DemandedElts, NewDepth, RecQ);
+      return isKnownNonZero(U.get(), DemandedElts, RecQ, NewDepth);
     });
   }
   case Instruction::InsertElement: {
@@ -2802,9 +2802,9 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
 
     // Result is zero if Elt is non-zero and rest of the demanded elts in Vec
     // are non-zero.
-    return (SkipElt || isKnownNonZero(Elt, Depth, Q)) &&
+    return (SkipElt || isKnownNonZero(Elt, Q, Depth)) &&
            (DemandedVecElts.isZero() ||
-            isKnownNonZero(Vec, DemandedVecElts, Depth, Q));
+            isKnownNonZero(Vec, DemandedVecElts, Q, Depth));
   }
   case Instruction::ExtractElement:
     if (const auto *EEI = dyn_cast<ExtractElementInst>(I)) {
@@ -2816,7 +2816,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
         APInt DemandedVecElts = APInt::getAllOnes(NumElts);
         if (CIdx && CIdx->getValue().ult(NumElts))
           DemandedVecElts = APInt::getOneBitSet(NumElts, CIdx->getZExtValue());
-        return isKnownNonZero(Vec, DemandedVecElts, Depth, Q);
+        return isKnownNonZero(Vec, DemandedVecElts, Q, Depth);
       }
     }
     break;
@@ -2831,12 +2831,12 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       break;
     // If demanded elements for both vecs are non-zero, the shuffle is non-zero.
     return (DemandedRHS.isZero() ||
-            isKnownNonZero(Shuf->getOperand(1), DemandedRHS, Depth, Q)) &&
+            isKnownNonZero(Shuf->getOperand(1), DemandedRHS, Q, Depth)) &&
            (DemandedLHS.isZero() ||
-            isKnownNonZero(Shuf->getOperand(0), DemandedLHS, Depth, Q));
+            isKnownNonZero(Shuf->getOperand(0), DemandedLHS, Q, Depth));
   }
   case Instruction::Freeze:
-    return isKnownNonZero(I->getOperand(0), Depth, Q) &&
+    return isKnownNonZero(I->getOperand(0), Q, Depth) &&
            isGuaranteedNotToBePoison(I->getOperand(0), Q.AC, Q.CxtI, Q.DT,
                                      Depth);
   case Instruction::Load: {
@@ -2886,7 +2886,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       if (Call->isReturnNonNull())
         return true;
       if (const auto *RP = getArgumentAliasingToReturnedPointer(Call, true))
-        return isKnownNonZero(RP, Depth, Q);
+        return isKnownNonZero(RP, Q, Depth);
     } else {
       if (MDNode *Ranges = Q.IIQ.getMetadata(Call, LLVMContext::MD_range))
         return rangeMetadataExcludesValue(Ranges, APInt::getZero(BitWidth));
@@ -2896,7 +2896,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
           return true;
       }
       if (const Value *RV = Call->getReturnedArgOperand())
-        if (RV->getType() == I->getType() && isKnownNonZero(RV, Depth, Q))
+        if (RV->getType() == I->getType() && isKnownNonZero(RV, Q, Depth))
           return true;
     }
 
@@ -2908,7 +2908,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       case Intrinsic::bitreverse:
       case Intrinsic::bswap:
       case Intrinsic::ctpop:
-        return isKnownNonZero(II->getArgOperand(0), DemandedElts, Depth, Q);
+        return isKnownNonZero(II->getArgOperand(0), DemandedElts, Q, Depth);
         // NB: We don't do usub_sat here as in any case we can prove its
         // non-zero, we will fold it to `sub nuw` in InstCombine.
       case Intrinsic::ssub_sat:
@@ -2924,11 +2924,11 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       case Intrinsic::vector_reduce_umin:
       case Intrinsic::vector_reduce_smax:
       case Intrinsic::vector_reduce_smin:
-        return isKnownNonZero(II->getArgOperand(0), Depth, Q);
+        return isKnownNonZero(II->getArgOperand(0), Q, Depth);
       case Intrinsic::umax:
       case Intrinsic::uadd_sat:
-        return isKnownNonZero(II->getArgOperand(1), DemandedElts, Depth, Q) ||
-               isKnownNonZero(II->getArgOperand(0), DemandedElts, Depth, Q);
+        return isKnownNonZero(II->getArgOperand(1), DemandedElts, Q, Depth) ||
+               isKnownNonZero(II->getArgOperand(0), DemandedElts, Q, Depth);
       case Intrinsic::smax: {
         // If either arg is strictly positive the result is non-zero. Otherwise
         // the result is non-zero if both ops are non-zero.
@@ -2936,7 +2936,7 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
                              const KnownBits &OpKnown) {
           if (!OpNonZero.has_value())
             OpNonZero = OpKnown.isNonZero() ||
-                        isKnownNonZero(Op, DemandedElts, Depth, Q);
+                        isKnownNonZero(Op, DemandedElts, Q, Depth);
           return *OpNonZero;
         };
         // Avoid re-computing isKnownNonZero.
@@ -2971,8 +2971,8 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       }
         [[fallthrough]];
       case Intrinsic::umin:
-        return isKnownNonZero(II->getArgOperand(0), DemandedElts, Depth, Q) &&
-               isKnownNonZero(II->getArgOperand(1), DemandedElts, Depth, Q);
+        return isKnownNonZero(II->getArgOperand(0), DemandedElts, Q, Depth) &&
+               isKnownNonZero(II->getArgOperand(1), DemandedElts, Q, Depth);
       case Intrinsic::cttz:
         return computeKnownBits(II->getArgOperand(0), DemandedElts, Depth, Q)
             .Zero[0];
@@ -2983,12 +2983,12 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
       case Intrinsic::fshl:
         // If Op0 == Op1, this is a rotate. rotate(x, y) != 0 iff x != 0.
         if (II->getArgOperand(0) == II->getArgOperand(1))
-          return isKnownNonZero(II->getArgOperand(0), DemandedElts, Depth, Q);
+          return isKnownNonZero(II->getArgOperand(0), DemandedElts, Q, Depth);
         break;
       case Intrinsic::vscale:
         return true;
       case Intrinsic::experimental_get_vector_length:
-        return isKnownNonZero(I->getOperand(0), Depth, Q);
+        return isKnownNonZero(I->getOperand(0), Q, Depth);
       default:
         break;
       }
@@ -3010,8 +3010,8 @@ static bool isKnownNonZeroFromOperator(const Operator *I,
 /// specified, perform context-sensitive analysis and return true if the
 /// pointer couldn't possibly be null at the specified instruction.
 /// Supports values with integer or pointer type and vectors of integers.
-bool isKnownNonZero(const Value *V, const APInt &DemandedElts, unsigned Depth,
-                    const SimplifyQuery &Q) {
+bool isKnownNonZero(const Value *V, const APInt &DemandedElts,
+                    const SimplifyQuery &Q, unsigned Depth) {
   Type *Ty = V->getType();
 
 #ifndef NDEBUG
@@ -3101,12 +3101,12 @@ bool isKnownNonZero(const Value *V, const APInt &DemandedElts, unsigned Depth,
   return false;
 }
 
-bool llvm::isKnownNonZero(const Value *V, unsigned Depth,
-                          const SimplifyQuery &Q) {
+bool llvm::isKnownNonZero(const Value *V, const SimplifyQuery &Q,
+                          unsigned Depth) {
   auto *FVTy = dyn_cast<FixedVectorType>(V->getType());
   APInt DemandedElts =
       FVTy ? APInt::getAllOnes(FVTy->getNumElements()) : APInt(1, 1);
-  return ::isKnownNonZero(V, DemandedElts, Depth, Q);
+  return ::isKnownNonZero(V, DemandedElts, Q, Depth);
 }
 
 /// If the pair of operators are the same invertible function, return the
@@ -3253,7 +3253,7 @@ static bool isModifyingBinopOfNonZero(const Value *V1, const Value *V2,
       Op = BO->getOperand(0);
     else
       return false;
-    return isKnownNonZero(Op, Depth + 1, Q);
+    return isKnownNonZero(Op, Q, Depth + 1);
   }
   return false;
 }
@@ -3266,7 +3266,7 @@ static bool isNonEqualMul(const Value *V1, const Value *V2, unsigned Depth,
     const APInt *C;
     return match(OBO, m_Mul(m_Specific(V1), m_APInt(C))) &&
            (OBO->hasNoUnsignedWrap() || OBO->hasNoSignedWrap()) &&
-           !C->isZero() && !C->isOne() && isKnownNonZero(V1, Depth + 1, Q);
+           !C->isZero() && !C->isOne() && isKnownNonZero(V1, Q, Depth + 1);
   }
   return false;
 }
@@ -3279,7 +3279,7 @@ static bool isNonEqualShl(const Value *V1, const Value *V2, unsigned Depth,
     const APInt *C;
     return match(OBO, m_Shl(m_Specific(V1), m_APInt(C))) &&
            (OBO->hasNoUnsignedWrap() || OBO->hasNoSignedWrap()) &&
-           !C->isZero() && isKnownNonZero(V1, Depth + 1, Q);
+           !C->isZero() && isKnownNonZero(V1, Q, Depth + 1);
   }
   return false;
 }
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index 22dbb3198a9f1..e657872c38284 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -2314,7 +2314,7 @@ static bool despeculateCountZeros(IntrinsicInst *CountZeros,
 
   // Bail if the value is never zero.
   Use &Op = CountZeros->getOperandUse(0);
-  if (isKnownNonZero(Op, /*Depth=*/0, *DL))
+  if (isKnownNonZero(Op, *DL))
     return false;
 
   // The intrinsic will be sunk behind a compare against zero and branch.
diff --git a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
index f27d8d64a1040..41b66aafe7d34 100644
--- a/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
+++ b/llvm/lib/Transforms/IPO/AttributorAttributes.cpp
@@ -2453,7 +2453,7 @@ bool AANonNull::isImpliedByIR(Attributor &A, const IRPosition &IRP,
 
   if (llvm::any_of(Worklist, [&](AA::ValueAndContext VAC) {
         return !isKnownNonZero(
-            VAC.getValue(), /*Depth=*/0,
+            VAC.getValue(),
             SimplifyQuery(A.getDataLayout(), DT, AC, VAC.getCtxI()));
       }))
     return false;
diff --git a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
index 14612b251d1a4..7ebf265e17ba1 100644
--- a/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
+++ b/llvm/lib/Transforms/IPO/FunctionAttrs.cpp
@@ -1175,7 +1175,7 @@ static bool isReturnNonNull(Function *F, const SCCNodeSet &SCCNodes,
     Value *RetVal = FlowsToReturn[i];
 
     // If this value is locally known to be non-null, we're good
-    if (isKnownNonZero(RetVal, /*Depth=*/0, DL))
+    if (isKnownNonZero(RetVal, DL))
       continue;
 
     // Otherwise, we need to look upwards since we can't make any local
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
index 07c50d866544b..c59b867b10e7d 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -988,7 +988,7 @@ Instruction *InstCombinerImpl::foldAddWithConstant(BinaryOperator &Add) {
   if (C->isOne()) {
     if (match(Op0, m_ZExt(m_Add(m_Value(X), m_AllOnes())))) {
       const SimplifyQuery Q = SQ.getWithInstruction(&Add);
-      if (llvm::isKnownNonZero(X, /*Depth=*/0, Q))
+      if (llvm::isKnownNonZero(X, Q))
         return new ZExtInst(X, Ty);
     }
   }
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index 2c0c4ee46e809..d311690be64f1 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -1039,9 +1039,9 @@ static Value *foldUnsignedUnderflowCheck(ICmpInst *ZeroICmp,
       match(ZeroCmpOp, m_c_Add(m_Specific(A), m_Value(B))) &&
       (ZeroICmp->hasOneUse() || UnsignedICmp->hasOneUse())) {
     auto GetKnownNonZeroAndOther = [&](Value *&NonZero, Value *&Other) {
-      if (!isKnownNonZero(NonZero, /*Depth=*/0, Q))
+      if (!isKnownNonZero(NonZero, Q))
         std::swap(NonZero, Other);
-      return isKnownNonZero(NonZero, /*Depth=*/0, Q);
+      return isKnownNonZero(NonZero, Q);
     };
 
     // Given  ZeroCmpOp = (A + B)
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index ba5db854647a4..60e4be883f513 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -601,8 +601,7 @@ static Instruction *foldCttzCtlz(IntrinsicInst &II, InstCombinerImpl &IC) {
   // then change the 'ZeroIsPoison' parameter to 'true'
   // because we know the zero behavior can't affect the result.
   if (!Known.One.isZero() ||
-      isKnownNonZero(Op0, /*Depth=*/0,
-                     IC.getSimplifyQuery().getWithInstruction(&II))) {
+      isKnownNonZero(Op0, IC.getSimplifyQuery().getWithInstruction(&II))) {
     if (!match(II.getArgOperand(1), m_One()))
       return IC.replaceOperand(II, 1, IC.Builder.getTrue());
   }
@@ -2067,8 +2066,7 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
     // See if we can deduce non-null.
     if (!CI.hasRetAttr(Attribute::NonNull) &&
         (Known.isNonZero() ||
-         isKnownNonZero(II, /*Depth=*/0,
-                        getSimplifyQuery().getWithInstruction(II)))) {
+         isKnownNonZero(II, getSimplifyQuery().getWithInstruction(II)))) {
       CI.addRetAttr(Attribute::NonNull);
       Changed = true;
     }
@@ -3664,8 +3662,7 @@ Instruction *InstCombinerImpl::visitCallBase(CallBase &Call) {
   for (Value *V : Call.args()) {
     if (V->getType()->isPointerTy() &&
         !Call.paramHasAttr(ArgNo, Attribute::NonNull) &&
-        isKnownNonZero(V, /*Depth=*/0,
-                       getSimplifyQuery().getWithInstruction(&Call)))
+        isKnownNonZero(V, getSimplifyQuery().getWithInstruction(&Call)))
       ArgNos.push_back(ArgNo);
     ArgNo++;
   }
@@ -3845,7 +3842,7 @@ Instruction *InstCombinerImpl::visitCallBase(CallBase &Call) {
 
         // isKnownNonNull -> nonnull attribute
         if (!GCR.hasRetAttr(Attribute::NonNull) &&
-            isKnownNonZero(DerivedPtr, /*Depth=*/0,
+            isKnownNonZero(DerivedPtr,
                            getSimplifyQuery().getWithInstruction(&Call))) {
           GCR.addRetAttr(Attribute::NonNull);
           // We discovered new fact, re-check users.
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index ee783eed190a7..de90907701743 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -1273,12 +1273,12 @@ Instruction *InstCombinerImpl::foldICmpWithZero(ICmpInst &Cmp) {
 
       // if X non-zero and NoOverflow(X * Y)
       //    (icmp eq/ne Y)
-      if (!XKnown.One.isZero() || isKnownNonZero(X, /*Depth=*/0, Q))
+      if (!XKnown.One.isZero() || isKnownNonZero(X, Q))
         return new ICmpInst(Pred, Y, Cmp.getOperand(1));
 
       // if Y non-zero and NoOverflow(X * Y)
       //    (icmp eq/ne X)
-      if (!YKnown.One.isZero() || isKnownNonZero(Y, /*Depth=*/0, Q))
+      if (!YKnown.One.isZero() || isKnownNonZero(Y, Q))
         return new ICmpInst(Pred, X, Cmp.getOperand(1));
     }
     // Note, we are skipping cases:
@@ -3087,7 +3087,7 @@ Instruction *InstCombinerImpl::foldICmpAddConstant(ICmpInst &Cmp,
   // (X + -1) <u C --> X <=u C (if X is never null)
   if (Pred == CmpInst::ICMP_ULT && C2->isAllOnes()) {
     const SimplifyQuery Q = SQ.getWithInstruction(&Cmp);
-    if (llvm::isKnownNonZero(X, /*Depth=*/0, Q))
+    if (llvm::isKnownNonZero(X, Q))
       return new ICmpInst(ICmpInst::ICMP_ULE, X, ConstantInt::get(Ty, C));
   }
 
@@ -4275,7 +4275,7 @@ static Value *foldICmpWithLowBitMaskedVal(ICmpInst::Predicate Pred, Value *Op0,
 
       // Look for: x & ~Mask pred ~Mask
       if (isMaskOrZero(X, /*Not=*/true, Q)) {
-        return !ICmpInst::isSigned(Pred) || isKnownNonZero(X, /*Depth=*/0, Q);
+        return !ICmpInst::isSigned(Pred) || isKnownNonZero(X, Q);
       }
       return false;
     }
@@ -4779,7 +4779,7 @@ static Instruction *foldICmpXorXX(ICmpInst &I, const SimplifyQuery &Q,
   // icmp (X ^ Y_NonZero) s>= X --> icmp (X ^ Y_NonZero) s> X
   // icmp (X ^ Y_NonZero) s<= X --> icmp (X ^ Y_NonZero) s< X
   CmpInst::Predicate PredOut = CmpInst::getStrictPredicate(Pred);
-  if (PredOut != Pred && isKnownNonZero(A, /*Depth=*/0, Q))
+  if (PredOut != Pred && isKnownNonZero(A, Q))
     return new ICmpInst(PredOut, Op0, Op1);
 
   return nullptr;
@@ -5062,11 +5062,11 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
     return new ICmpInst(Pred, C, D);
   // (A - B) u>=/u< A --> B u>/u<= A  iff B != 0
   if (A == Op1 && (Pred == ICmpInst::ICMP_UGE || Pred == ICmpInst::ICMP_ULT) &&
-      isKnownNonZero(B, /*Depth=*/0, Q))
+      isKnownNonZero(B, Q))
     return new ICmpInst(CmpInst::getFlippedStrictnessPredicate(Pred), B, A);
   // C u<=/u> (C - D) --> C u</u>= D  iff B != 0
   if (C == Op0 && (Pred == ICmpInst::ICMP_ULE || Pred == ICmpInst::ICMP_UGT) &&
-      isKnownNonZero(D, /*Depth=*/0, Q))
+      isKnownNonZero(D, Q))
     return new ICmpInst(CmpInst::getFlippedStrictnessPredicate(Pred), C, D);
 
   // icmp (A-B), (C-B) -> icmp A, C for equalities or if there is no overflow.
@@ -5108,13 +5108,13 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
         //    X * Z eq/ne Y * Z -> X eq/ne Y
         if (ZKnown.countMaxTrailingZeros() == 0)
           return new ICmpInst(Pred, X, Y);
-        NonZero = !ZKnown.One.isZero() || isKnownNonZero(Z, /*Depth=*/0, Q);
+        NonZero = !ZKnown.One.isZero() || isKnownNonZero(Z, Q);
         // if Z != 0 and nsw(X * Z) and nsw(Y * Z)
         //    X * Z eq/ne Y * Z -> X eq/ne Y
         if (NonZero && BO0 && BO1 && Op0HasNSW && Op1HasNSW)
           return new ICmpInst(Pred, X, Y);
       } else
-        NonZero = isKnownNonZero(Z, /*Depth=*/0, Q);
+        NonZero = isKnownNonZero(Z, Q);
 
       // If Z != 0 and nuw(X * Z) and nuw(Y * Z)
       //    X * Z u{lt/le/gt/ge}/eq/ne Y * Z -> X u{lt/le/gt/ge}/eq/ne Y
diff --git a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
index 9838e2aa9f3a2..52803e9bea451 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
@@ -1537,8 +1537,7 @@ Instruction *InstCombinerImpl::visitPHINode(PHINode &PN) {
       for (unsigned I = 0, E = PN.getNumIncomingValues(); I != E; ++I) {
         Instruction *CtxI = PN.getIncomingBlock(I)->getTerminator();
         Value *VA = PN.getIncomingValue(I);
-        if (isKnownNonZero(VA, 0,
-                           getSimplifyQuery().getWithInstruction(CtxI))) {
+        if (isKnownNonZero(VA, getSimplifyQuery().getWithInstruction(CtxI))) {
           if (!NonZeroConst)
             NonZeroConst = getAnyNonZeroConstInt(PN);
           if (NonZeroConst != VA) {
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 4c00f2a0ea176..5a144cc737896 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1431,7 +1431,7 @@ Instruction *InstCombinerImpl::foldFBinOpOfIntCastsFromSign(
     if (OpsKnown[OpNo].hasKnownBits() &&
         OpsKnown[OpNo].getKnownBits(SQ).isNonZero())
       return true;
-    return isKnownNonZero(IntOps[OpNo], /*Depth=*/0, SQ);
+    return isKnownNonZero(IntOps[OpNo], SQ);
   };
 
   auto IsNonNeg = [&](unsigned OpNo) -> bool {
diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index a72b0ee9a08e0..ee3531bbd68df 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -1281,7 +1281,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
         // ignored.
         return;
       }
-      if (llvm::isKnownNonZero(ConvertedShadow, /*Depth=*/0, DL)) {
+      if (llvm::isKnownNonZero(ConvertedShadow, DL)) {
         // Copy origin as the value is definitely uninitialized.
         paintOrigin(IRB, updateOrigin(Origin, IRB), OriginPtr, StoreSize,
                     OriginAlignment);
@@ -1427,7 +1427,7 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
           // Skip, value is initialized or const shadow is ignored.
           continue;
         }
-        if (llvm::isKnownNonZero(ConvertedShadow, /*Depth=*/0, DL)) {
+        if (llvm::isKnownNonZero(ConvertedShadow, DL)) {
           // Report as the value is definitely uninitialized.
           insertWarningFn(IRB, ShadowData.Origin);
           if (!MS.Recover)
diff --git a/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp b/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
index f376b5f7d68d4..40d0f6b75d69b 100644
--- a/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
+++ b/llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
@@ -459,7 +459,7 @@ static void convertMetadataToAssumes(LoadInst *LI, Value *Val,
   // we can only do this if the value is known non-poison.
   if (AC && LI->getMetadata(LLVMContext::MD_nonnull) &&
       LI->getMetadata(LLVMContext::MD_noundef) &&
-      !isKnownNonZero(Val, /*Depth=*/0, SimplifyQuery(DL, DT, AC, LI)))
+      !isKnownNonZero(Val, SimplifyQuery(DL, DT, AC, LI)))
     addAssumeNonNull(AC, LI);
 }
 
diff --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
index 7e9e91606fe22..2e68a9c01898c 100644
--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -305,7 +305,7 @@ static void annotateNonNullAndDereferenceable(CallInst *CI, ArrayRef<unsigned> A
   if (ConstantInt *LenC = dyn_cast<ConstantInt>(Size)) {
     annotateNonNullNoUndefBasedOnAccess(CI, ArgNos);
     annotateDereferenceableBytes(CI, ArgNos, LenC->getZExtValue());
-  } else if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
+  } else if (isKnownNonZero(Size, DL)) {
     annotateNonNullNoUndefBasedOnAccess(CI, ArgNos);
     const APInt *X, *Y;
     uint64_t DerefMin = 1;
@@ -394,7 +394,7 @@ Value *LibCallSimplifier::optimizeStrNCat(CallInst *CI, IRBuilderBase &B) {
   Value *Size = CI->getArgOperand(2);
   uint64_t Len;
   annotateNonNullNoUndefBasedOnAccess(CI, 0);
-  if (isKnownNonZero(Size, /*Depth=*/0, DL))
+  if (isKnownNonZero(Size, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 1);
 
   // We don't do anything if length is not constant.
@@ -613,7 +613,7 @@ Value *LibCallSimplifier::optimizeStrNCmp(CallInst *CI, IRBuilderBase &B) {
   if (Str1P == Str2P) // strncmp(x,x,n)  -> 0
     return ConstantInt::get(CI->getType(), 0);
 
-  if (isKnownNonZero(Size, /*Depth=*/0, DL))
+  if (isKnownNonZero(Size, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, {0, 1});
   // Get the length argument if it is constant.
   uint64_t Length;
@@ -749,7 +749,7 @@ Value *LibCallSimplifier::optimizeStpCpy(CallInst *CI, IRBuilderBase &B) {
 
 Value *LibCallSimplifier::optimizeStrLCpy(CallInst *CI, IRBuilderBase &B) {
   Value *Size = CI->getArgOperand(2);
-  if (isKnownNonZero(Size, /*Depth=*/0, DL))
+  if (isKnownNonZero(Size, DL))
     // Like snprintf, the function stores into the destination only when
     // the size argument is nonzero.
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
@@ -833,7 +833,7 @@ Value *LibCallSimplifier::optimizeStringNCpy(CallInst *CI, bool RetEnd,
   Value *Src = CI->getArgOperand(1);
   Value *Size = CI->getArgOperand(2);
 
-  if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
+  if (isKnownNonZero(Size, DL)) {
     // Both st{p,r}ncpy(D, S, N) access the source and destination arrays
     // only when N is nonzero.
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
@@ -926,7 +926,7 @@ Value *LibCallSimplifier::optimizeStringLength(CallInst *CI, IRBuilderBase &B,
   Type *CharTy = B.getIntNTy(CharSize);
 
   if (isOnlyUsedInZeroEqualityComparison(CI) &&
-      (!Bound || isKnownNonZero(Bound, /*Depth=*/0, DL))) {
+      (!Bound || isKnownNonZero(Bound, DL))) {
     // Fold strlen:
     //   strlen(x) != 0 --> *x != 0
     //   strlen(x) == 0 --> *x == 0
@@ -1047,7 +1047,7 @@ Value *LibCallSimplifier::optimizeStrNLen(CallInst *CI, IRBuilderBase &B) {
   if (Value *V = optimizeStringLength(CI, B, 8, Bound))
     return V;
 
-  if (isKnownNonZero(Bound, /*Depth=*/0, DL))
+  if (isKnownNonZero(Bound, DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
   return nullptr;
 }
@@ -1291,7 +1291,7 @@ Value *LibCallSimplifier::optimizeMemChr(CallInst *CI, IRBuilderBase &B) {
   Value *SrcStr = CI->getArgOperand(0);
   Value *Size = CI->getArgOperand(2);
 
-  if (isKnownNonZero(Size, /*Depth=*/0, DL)) {
+  if (isKnownNonZero(Size, DL)) {
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
     if (isOnlyUsedInEqualityComparison(CI, SrcStr))
       return memChrToCharCompare(CI, Size, B, DL);
@@ -2976,7 +2976,7 @@ Value *LibCallSimplifier::optimizeStrToInt(CallInst *CI, IRBuilderBase &B,
     // It would be readonly too, except that it still may write to errno.
     CI->addParamAttr(0, Attribute::NoCapture);
     EndPtr = nullptr;
-  } else if (!isKnownNonZero(EndPtr, /*Depth=*/0, DL))
+  } else if (!isKnownNonZero(EndPtr, DL))
     return nullptr;
 
   StringRef Str;
@@ -3402,7 +3402,7 @@ Value *LibCallSimplifier::optimizeSnPrintF(CallInst *CI, IRBuilderBase &B) {
     return V;
   }
 
-  if (isKnownNonZero(CI->getOperand(1), /*Depth=*/0, DL))
+  if (isKnownNonZero(CI->getOperand(1), DL))
     annotateNonNullNoUndefBasedOnAccess(CI, 0);
   return nullptr;
 }
diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index e0e2f50c89ada..4918cee1fa82a 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -886,7 +886,7 @@ bool VectorCombine::scalarizeVPIntrinsic(Instruction &I) {
     SafeToSpeculate = isSafeToSpeculativelyExecuteWithOpcode(
         *FunctionalOpcode, &VPI, nullptr, &AC, &DT);
   if (!SafeToSpeculate &&
-      !isKnownNonZero(EVL, /*Depth=*/0, SimplifyQuery(*DL, &DT, &AC, &VPI)))
+      !isKnownNonZero(EVL, SimplifyQuery(*DL, &DT, &AC, &VPI)))
     return false;
 
   Value *ScalarVal =
diff --git a/llvm/unittests/Analysis/ValueTrackingTest.cpp b/llvm/unittests/Analysis/ValueTrackingTest.cpp
index 8ebd9b511f39f..8738af91b652b 100644
--- a/llvm/unittests/Analysis/ValueTrackingTest.cpp
+++ b/llvm/unittests/Analysis/ValueTrackingTest.cpp
@@ -2110,8 +2110,7 @@ TEST_F(ValueTrackingTest, isNonZeroRecurrence) {
   )");
   const DataLayout &DL = M->getDataLayout();
   AssumptionCache AC(*F);
-  EXPECT_TRUE(isKnownNonZero(A, /*Depth=*/0,
-                             SimplifyQuery(DL, /*DT=*/nullptr, &AC, CxtI)));
+  EXPECT_TRUE(isKnownNonZero(A, SimplifyQuery(DL, /*DT=*/nullptr, &AC, CxtI)));
 }
 
 TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond) {
@@ -2135,9 +2134,8 @@ TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond) {
   DominatorTree DT(*F);
   const DataLayout &DL = M->getDataLayout();
   const SimplifyQuery SQ(DL, &DT, &AC);
-  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI)), true);
-  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI2)),
-            false);
+  EXPECT_EQ(isKnownNonZero(A, SQ.getWithInstruction(CxtI)), true);
+  EXPECT_EQ(isKnownNonZero(A, SQ.getWithInstruction(CxtI2)), false);
 }
 
 TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond2) {
@@ -2161,9 +2159,8 @@ TEST_F(ValueTrackingTest, KnownNonZeroFromDomCond2) {
   DominatorTree DT(*F);
   const DataLayout &DL = M->getDataLayout();
   const SimplifyQuery SQ(DL, &DT, &AC);
-  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI)), true);
-  EXPECT_EQ(isKnownNonZero(A, /*Depth=*/0, SQ.getWithInstruction(CxtI2)),
-            false);
+  EXPECT_EQ(isKnownNonZero(A, SQ.getWithInstruction(CxtI)), true);
+  EXPECT_EQ(isKnownNonZero(A, SQ.getWithInstruction(CxtI2)), false);
 }
 
 TEST_F(ValueTrackingTest, IsImpliedConditionAnd) {

From 5a34ff12b8f4a73f5dcd4be1b2575dc38cf13bee Mon Sep 17 00:00:00 2001
From: Jeremy Kun <jkun@google.com>
Date: Tue, 16 Apr 2024 07:35:36 -0700
Subject: [PATCH 598/886] fix Polynomial.td doc filename (#88900)

Not sure how best to test this, but I think it fixes the error
https://github.com/llvm/mlir-www/actions/runs/8699908058/job/23859264085#step:7:1111

Co-authored-by: Jeremy Kun <j2kun@users.noreply.github.com>
Co-authored-by: Jacques Pienaar <jpienaar@google.com>
---
 mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
index d8039deb5ee21..dd0384d8b79d6 100644
--- a/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
@@ -1,8 +1,8 @@
 add_mlir_dialect(Polynomial polynomial)
-add_mlir_doc(PolynomialDialect PolynomialDialect Polynomial/ -gen-dialect-doc)
-add_mlir_doc(PolynomialOps PolynomialOps Polynomial/ -gen-op-doc)
-add_mlir_doc(PolynomialAttributes PolynomialAttributes Dialects/ -gen-attrdef-doc)
-add_mlir_doc(PolynomialTypes PolynomialTypes Dialects/ -gen-typedef-doc)
+add_mlir_doc(Polynomial PolynomialDialect Polynomial/ -gen-dialect-doc)
+add_mlir_doc(Polynomial PolynomialOps Polynomial/ -gen-op-doc)
+add_mlir_doc(Polynomial PolynomialAttributes Dialects/ -gen-attrdef-doc)
+add_mlir_doc(Polynomial PolynomialTypes Dialects/ -gen-typedef-doc)
 
 set(LLVM_TARGET_DEFINITIONS Polynomial.td)
 mlir_tablegen(PolynomialAttributes.cpp.inc -gen-attrdef-defs -attrdefs-dialect=polynomial)

From b63247627c9e87e898dec5bf0bea255b3f0eec5c Mon Sep 17 00:00:00 2001
From: Ding Fei <fding@feysh.com>
Date: Tue, 16 Apr 2024 22:38:27 +0800
Subject: [PATCH 599/886] [AST][RecoveryExpr] Fix a crash on c89/c90 invalid
 InitListExpr (#88008) (#88014)

Use refactored `CheckForConstantInitializer()` to skip checking expr
with error.

---------

Co-authored-by: Aaron Ballman <aaron@aaronballman.com>
---
 clang/docs/ReleaseNotes.rst                   |  2 ++
 clang/include/clang/Sema/Sema.h               |  4 ++-
 clang/lib/Sema/SemaDecl.cpp                   | 28 ++++++++-----------
 clang/lib/Sema/SemaExpr.cpp                   |  2 +-
 .../test/Sema/recover-expr-gh88008-nocrash.c  | 11 ++++++++
 5 files changed, 28 insertions(+), 19 deletions(-)
 create mode 100644 clang/test/Sema/recover-expr-gh88008-nocrash.c

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index db90db6fa4ab0..d8ec8bcb8df53 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -417,6 +417,8 @@ Bug Fixes in This Version
 - Fixed a regression in CTAD that a friend declaration that befriends itself may cause
   incorrect constraint substitution. (#GH86769).
 
+- Fixed an assertion failure on invalid InitListExpr in C89 mode (#GH88008).
+
 Bug Fixes to Compiler Builtins
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index a5fe83a539aaf..77150a318ee47 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -55,6 +55,7 @@
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/SemaBase.h"
 #include "clang/Sema/SemaConcept.h"
+#include "clang/Sema/SemaDiagnostic.h"
 #include "clang/Sema/TypoCorrection.h"
 #include "clang/Sema/Weak.h"
 #include "llvm/ADT/ArrayRef.h"
@@ -3427,7 +3428,8 @@ class Sema final : public SemaBase {
       bool ConstexprSupported, bool CLinkageMayDiffer);
 
   /// type checking declaration initializers (C99 6.7.8)
-  bool CheckForConstantInitializer(Expr *e, QualType t);
+  bool CheckForConstantInitializer(
+      Expr *Init, unsigned DiagID = diag::err_init_element_not_constant);
 
   QualType deduceVarTypeFromInitializer(VarDecl *VDecl, DeclarationName Name,
                                         QualType Type, TypeSourceInfo *TSI,
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 390da508518e1..745cf41e204e7 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -12671,7 +12671,7 @@ void Sema::CheckMSVCRTEntryPoint(FunctionDecl *FD) {
   }
 }
 
-bool Sema::CheckForConstantInitializer(Expr *Init, QualType DclT) {
+bool Sema::CheckForConstantInitializer(Expr *Init, unsigned DiagID) {
   // FIXME: Need strict checking.  In C89, we need to check for
   // any assignment, increment, decrement, function-calls, or
   // commas outside of a sizeof.  In C99, it's the same list,
@@ -12689,8 +12689,7 @@ bool Sema::CheckForConstantInitializer(Expr *Init, QualType DclT) {
   const Expr *Culprit;
   if (Init->isConstantInitializer(Context, false, &Culprit))
     return false;
-  Diag(Culprit->getExprLoc(), diag::err_init_element_not_constant)
-    << Culprit->getSourceRange();
+  Diag(Culprit->getExprLoc(), DiagID) << Culprit->getSourceRange();
   return true;
 }
 
@@ -13808,29 +13807,24 @@ void Sema::AddInitializerToDecl(Decl *RealDecl, Expr *Init, bool DirectInit) {
     // OpenCL v1.2 s6.5.3: __constant locals must be constant-initialized.
     // This is true even in C++ for OpenCL.
     } else if (VDecl->getType().getAddressSpace() == LangAS::opencl_constant) {
-      CheckForConstantInitializer(Init, DclT);
+      CheckForConstantInitializer(Init);
 
-    // Otherwise, C++ does not restrict the initializer.
+      // Otherwise, C++ does not restrict the initializer.
     } else if (getLangOpts().CPlusPlus) {
       // do nothing
 
     // C99 6.7.8p4: All the expressions in an initializer for an object that has
     // static storage duration shall be constant expressions or string literals.
     } else if (VDecl->getStorageClass() == SC_Static) {
-      CheckForConstantInitializer(Init, DclT);
+      CheckForConstantInitializer(Init);
 
-    // C89 is stricter than C99 for aggregate initializers.
-    // C89 6.5.7p3: All the expressions [...] in an initializer list
-    // for an object that has aggregate or union type shall be
-    // constant expressions.
+      // C89 is stricter than C99 for aggregate initializers.
+      // C89 6.5.7p3: All the expressions [...] in an initializer list
+      // for an object that has aggregate or union type shall be
+      // constant expressions.
     } else if (!getLangOpts().C99 && VDecl->getType()->isAggregateType() &&
                isa<InitListExpr>(Init)) {
-      const Expr *Culprit;
-      if (!Init->isConstantInitializer(Context, false, &Culprit)) {
-        Diag(Culprit->getExprLoc(),
-             diag::ext_aggregate_init_not_constant)
-          << Culprit->getSourceRange();
-      }
+      CheckForConstantInitializer(Init, diag::ext_aggregate_init_not_constant);
     }
 
     if (auto *E = dyn_cast<ExprWithCleanups>(Init))
@@ -13963,7 +13957,7 @@ void Sema::AddInitializerToDecl(Decl *RealDecl, Expr *Init, bool DirectInit) {
     // Avoid duplicate diagnostics for constexpr variables.
     if (!getLangOpts().CPlusPlus && !VDecl->isInvalidDecl() &&
         !VDecl->isConstexpr())
-      CheckForConstantInitializer(Init, DclT);
+      CheckForConstantInitializer(Init);
   }
 
   QualType InitType = Init->getType();
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index cabffa47c9318..7c3faba0f7881 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -7331,7 +7331,7 @@ Sema::BuildCompoundLiteralExpr(SourceLocation LParenLoc, TypeSourceInfo *TInfo,
     if (!LiteralExpr->isTypeDependent() &&
         !LiteralExpr->isValueDependent() &&
         !literalType->isDependentType()) // C99 6.5.2.5p3
-      if (CheckForConstantInitializer(LiteralExpr, literalType))
+      if (CheckForConstantInitializer(LiteralExpr))
         return ExprError();
   } else if (literalType.getAddressSpace() != LangAS::opencl_private &&
              literalType.getAddressSpace() != LangAS::Default) {
diff --git a/clang/test/Sema/recover-expr-gh88008-nocrash.c b/clang/test/Sema/recover-expr-gh88008-nocrash.c
new file mode 100644
index 0000000000000..5500b33dd0e85
--- /dev/null
+++ b/clang/test/Sema/recover-expr-gh88008-nocrash.c
@@ -0,0 +1,11 @@
+// RUN: %clang_cc1 %s -verify -fsyntax-only -std=c90
+
+struct S {
+  int v;
+};
+
+struct T; // expected-note {{forward declaration of 'struct T'}}
+
+void gh88008_nocrash(struct T *t) {
+  struct S s = { .v = t->y }; // expected-error {{incomplete definition of type 'struct T'}}
+}

From d2d4a1bbdc455a30d600743eb59fb1c69205967a Mon Sep 17 00:00:00 2001
From: XChy <xxs_chy@outlook.com>
Date: Tue, 16 Apr 2024 22:52:19 +0800
Subject: [PATCH 600/886] Revert "[JumpThreading] Thread over BB with only an
 unconditional branch" (#88907)

Reverts llvm/llvm-project#86312
---
 llvm/lib/Transforms/Utils/Local.cpp           |  10 +-
 llvm/test/CodeGen/AArch64/and-sink.ll         |   9 +-
 .../AArch64/combine-comparisons-by-cse.ll     | 122 ++++++++++-------
 llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll   |  18 ++-
 llvm/test/Transforms/JumpThreading/pr79175.ll |   8 +-
 llvm/test/Transforms/JumpThreading/select.ll  |  50 ++++---
 .../Transforms/JumpThreading/thread-prob-7.ll |   8 +-
 .../Transforms/JumpThreading/uncond-no-phi.ll | 123 ------------------
 .../PhaseOrdering/thread-uncond-bb.ll         |  62 ---------
 9 files changed, 126 insertions(+), 284 deletions(-)
 delete mode 100644 llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
 delete mode 100644 llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll

diff --git a/llvm/lib/Transforms/Utils/Local.cpp b/llvm/lib/Transforms/Utils/Local.cpp
index baec51a07fcbf..a42ef0c4e6ae9 100644
--- a/llvm/lib/Transforms/Utils/Local.cpp
+++ b/llvm/lib/Transforms/Utils/Local.cpp
@@ -1019,14 +1019,12 @@ CanRedirectPredsOfEmptyBBToSucc(BasicBlock *BB, BasicBlock *Succ,
                                 const SmallPtrSetImpl<BasicBlock *> &SuccPreds,
                                 BasicBlock *&CommonPred) {
 
-  // When Succ has no phis, BB may be merged into Succ directly. We don't need
-  // to redirect the predecessors of BB in this case.
-  if (Succ->phis().empty())
+  // There must be phis in BB, otherwise BB will be merged into Succ directly
+  if (BB->phis().empty() || Succ->phis().empty())
     return false;
 
-  // BB must have multiple different predecessors, so that at least one of
-  // predecessors can be redirected to Succ, except the common predecessor.
-  if (BB->getUniquePredecessor() || pred_empty(BB))
+  // BB must have predecessors not shared that can be redirected to Succ
+  if (!BB->hasNPredecessorsOrMore(2))
     return false;
 
   // Get single common predecessors of both BB and Succ
diff --git a/llvm/test/CodeGen/AArch64/and-sink.ll b/llvm/test/CodeGen/AArch64/and-sink.ll
index a57e9d54f3078..f298a55dab721 100644
--- a/llvm/test/CodeGen/AArch64/and-sink.ll
+++ b/llvm/test/CodeGen/AArch64/and-sink.ll
@@ -11,14 +11,15 @@
 define dso_local i32 @and_sink1(i32 %a, i1 %c) {
 ; CHECK-LABEL: and_sink1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    tbz w1, #0, .LBB0_2
+; CHECK-NEXT:    tbz w1, #0, .LBB0_3
 ; CHECK-NEXT:  // %bb.1: // %bb0
-; CHECK-NEXT:    tst w0, #0x4
 ; CHECK-NEXT:    adrp x8, A
-; CHECK-NEXT:    cset w0, eq
 ; CHECK-NEXT:    str wzr, [x8, :lo12:A]
+; CHECK-NEXT:    tbnz w0, #2, .LBB0_3
+; CHECK-NEXT:  // %bb.2:
+; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB0_2:
+; CHECK-NEXT:  .LBB0_3: // %bb2
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 
diff --git a/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll b/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
index dde3e81833a63..6449c3e11d667 100644
--- a/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
+++ b/llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
@@ -13,10 +13,10 @@ define i32 @combine_gt_ge_10() #0 {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
-; CHECK-NEXT:    ldr w9, [x8]
+; CHECK-NEXT:    ldr w8, [x8]
+; CHECK-NEXT:    cmp w8, #10
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
-; CHECK-NEXT:    cmp w9, #10
 ; CHECK-NEXT:    b.le .LBB0_3
 ; CHECK-NEXT:  // %bb.1: // %land.lhs.true
 ; CHECK-NEXT:    adrp x9, :got:c
@@ -29,17 +29,18 @@ define i32 @combine_gt_ge_10() #0 {
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB0_3: // %lor.lhs.false
-; CHECK-NEXT:    cmp w9, #10
-; CHECK-NEXT:    b.lt .LBB0_5
+; CHECK-NEXT:    b.lt .LBB0_6
 ; CHECK-NEXT:  .LBB0_4: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    b.ne .LBB0_6
+; CHECK-NEXT:  // %bb.5:
+; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB0_5:
+; CHECK-NEXT:  .LBB0_6: // %if.end
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
@@ -144,10 +145,10 @@ define i32 @combine_lt_ge_5() #0 {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
-; CHECK-NEXT:    ldr w9, [x8]
+; CHECK-NEXT:    ldr w8, [x8]
+; CHECK-NEXT:    cmp w8, #5
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
-; CHECK-NEXT:    cmp w9, #5
 ; CHECK-NEXT:    b.ge .LBB2_3
 ; CHECK-NEXT:  // %bb.1: // %land.lhs.true
 ; CHECK-NEXT:    adrp x9, :got:c
@@ -160,17 +161,18 @@ define i32 @combine_lt_ge_5() #0 {
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:  .LBB2_3: // %lor.lhs.false
-; CHECK-NEXT:    cmp w9, #5
-; CHECK-NEXT:    b.gt .LBB2_5
+; CHECK-NEXT:    b.gt .LBB2_6
 ; CHECK-NEXT:  .LBB2_4: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    b.ne .LBB2_6
+; CHECK-NEXT:  // %bb.5:
+; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB2_5:
+; CHECK-NEXT:  .LBB2_6: // %if.end
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
@@ -497,17 +499,24 @@ define i32 @do_nothing_if_resultant_opcodes_would_differ() #0 {
 ; CHECK-NEXT:  // %bb.3: // %while.cond.while.end_crit_edge
 ; CHECK-NEXT:    ldr w8, [x19]
 ; CHECK-NEXT:  .LBB7_4: // %while.end
-; CHECK-NEXT:    adrp x9, :got:b
-; CHECK-NEXT:    adrp x10, :got:d
-; CHECK-NEXT:    ldr x9, [x9, :got_lo12:b]
-; CHECK-NEXT:    ldr x10, [x10, :got_lo12:d]
-; CHECK-NEXT:    ldp x20, x19, [sp, #16] // 16-byte Folded Reload
+; CHECK-NEXT:    cmp w8, #1
+; CHECK-NEXT:    b.gt .LBB7_7
+; CHECK-NEXT:  // %bb.5: // %land.lhs.true
+; CHECK-NEXT:    adrp x8, :got:b
+; CHECK-NEXT:    adrp x9, :got:d
+; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
+; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
+; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
-; CHECK-NEXT:    ldr w10, [x10]
-; CHECK-NEXT:    cmp w9, w10
-; CHECK-NEXT:    ccmp w8, #2, #0, eq
-; CHECK-NEXT:    mov w8, #123 // =0x7b
-; CHECK-NEXT:    csel w0, w8, wzr, lt
+; CHECK-NEXT:    cmp w8, w9
+; CHECK-NEXT:    b.ne .LBB7_7
+; CHECK-NEXT:  // %bb.6:
+; CHECK-NEXT:    mov w0, #123 // =0x7b
+; CHECK-NEXT:    b .LBB7_8
+; CHECK-NEXT:  .LBB7_7: // %if.end
+; CHECK-NEXT:    mov w0, wzr
+; CHECK-NEXT:  .LBB7_8: // %return
+; CHECK-NEXT:    ldp x20, x19, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp], #32 // 8-byte Folded Reload
 ; CHECK-NEXT:    .cfi_def_cfa_offset 0
 ; CHECK-NEXT:    .cfi_restore w19
@@ -555,42 +564,52 @@ return:                                           ; preds = %if.end, %land.lhs.t
 define i32 @do_nothing_if_compares_can_not_be_adjusted_to_each_other() #0 {
 ; CHECK-LABEL: do_nothing_if_compares_can_not_be_adjusted_to_each_other:
 ; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    stp x30, x19, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w19, -8
+; CHECK-NEXT:    .cfi_offset w30, -16
+; CHECK-NEXT:    .cfi_remember_state
 ; CHECK-NEXT:    adrp x8, :got:a
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:a]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    cmp w8, #0
-; CHECK-NEXT:    b.gt .LBB8_4
+; CHECK-NEXT:    b.gt .LBB8_3
 ; CHECK-NEXT:  // %bb.1: // %while.body.preheader
-; CHECK-NEXT:    stp x30, x19, [sp, #-16]! // 16-byte Folded Spill
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    .cfi_offset w19, -8
-; CHECK-NEXT:    .cfi_offset w30, -16
 ; CHECK-NEXT:    sub w19, w8, #1
 ; CHECK-NEXT:  .LBB8_2: // %while.body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    bl do_something
 ; CHECK-NEXT:    adds w19, w19, #1
 ; CHECK-NEXT:    b.mi .LBB8_2
-; CHECK-NEXT:  // %bb.3:
-; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
-; CHECK-NEXT:    .cfi_def_cfa_offset 0
-; CHECK-NEXT:    .cfi_restore w19
-; CHECK-NEXT:    .cfi_restore w30
-; CHECK-NEXT:  .LBB8_4: // %while.end
+; CHECK-NEXT:  .LBB8_3: // %while.end
+; CHECK-NEXT:    adrp x8, :got:c
+; CHECK-NEXT:    ldr x8, [x8, :got_lo12:c]
+; CHECK-NEXT:    ldr w8, [x8]
+; CHECK-NEXT:    cmn w8, #2
+; CHECK-NEXT:    b.lt .LBB8_6
+; CHECK-NEXT:  // %bb.4: // %land.lhs.true
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:d
-; CHECK-NEXT:    adrp x10, :got:c
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
 ; CHECK-NEXT:    ldr x9, [x9, :got_lo12:d]
-; CHECK-NEXT:    ldr x10, [x10, :got_lo12:c]
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
-; CHECK-NEXT:    ldr w10, [x10]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    mov w8, #-3 // =0xfffffffd
-; CHECK-NEXT:    ccmp w10, w8, #4, eq
-; CHECK-NEXT:    mov w8, #123 // =0x7b
-; CHECK-NEXT:    csel w0, w8, wzr, gt
+; CHECK-NEXT:    b.ne .LBB8_6
+; CHECK-NEXT:  // %bb.5:
+; CHECK-NEXT:    mov w0, #123 // =0x7b
+; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-NEXT:    .cfi_restore w19
+; CHECK-NEXT:    .cfi_restore w30
+; CHECK-NEXT:    ret
+; CHECK-NEXT:  .LBB8_6: // %if.end
+; CHECK-NEXT:    .cfi_restore_state
+; CHECK-NEXT:    mov w0, wzr
+; CHECK-NEXT:    ldp x30, x19, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-NEXT:    .cfi_restore w19
+; CHECK-NEXT:    .cfi_restore w30
 ; CHECK-NEXT:    ret
 entry:
   %0 = load i32, ptr @a, align 4
@@ -763,14 +782,12 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    cmp w8, #0
 ; CHECK-NEXT:    csel x9, x0, xzr, gt
 ; CHECK-NEXT:    str x9, [x1]
-; CHECK-NEXT:    b.le .LBB11_3
+; CHECK-NEXT:    b.le .LBB11_2
 ; CHECK-NEXT:  // %bb.1: // %lor.lhs.false
 ; CHECK-NEXT:    cmp w8, #2
-; CHECK-NEXT:    b.ge .LBB11_5
-; CHECK-NEXT:  // %bb.2:
-; CHECK-NEXT:    mov w0, wzr
-; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB11_3: // %land.lhs.true
+; CHECK-NEXT:    b.ge .LBB11_4
+; CHECK-NEXT:    b .LBB11_6
+; CHECK-NEXT:  .LBB11_2: // %land.lhs.true
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:c
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
@@ -778,11 +795,11 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    b.ne .LBB11_5
-; CHECK-NEXT:  // %bb.4:
+; CHECK-NEXT:    b.ne .LBB11_4
+; CHECK-NEXT:  // %bb.3:
 ; CHECK-NEXT:    mov w0, #1 // =0x1
 ; CHECK-NEXT:    ret
-; CHECK-NEXT:  .LBB11_5: // %land.lhs.true3
+; CHECK-NEXT:  .LBB11_4: // %land.lhs.true3
 ; CHECK-NEXT:    adrp x8, :got:b
 ; CHECK-NEXT:    adrp x9, :got:d
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:b]
@@ -790,7 +807,12 @@ define i32 @combine_gt_ge_sel(i64 %v, ptr %p) #0 {
 ; CHECK-NEXT:    ldr w8, [x8]
 ; CHECK-NEXT:    ldr w9, [x9]
 ; CHECK-NEXT:    cmp w8, w9
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    b.ne .LBB11_6
+; CHECK-NEXT:  // %bb.5:
+; CHECK-NEXT:    mov w0, #1 // =0x1
+; CHECK-NEXT:    ret
+; CHECK-NEXT:  .LBB11_6: // %if.end
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ret
 entry:
   %0 = load i32, ptr @a, align 4
diff --git a/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll b/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
index c33c81841be65..dddc4bd953d7a 100644
--- a/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
+++ b/llvm/test/CodeGen/Hexagon/vect/zext-v4i1.ll
@@ -10,13 +10,12 @@ define i32 @fred(ptr %a0) #0 {
 ; CHECK-LABEL: fred:
 ; CHECK:       // %bb.0: // %b0
 ; CHECK-NEXT:    {
-; CHECK-NEXT:     r1:0 = combine(r0,#0)
-; CHECK-NEXT:     if (p0) jumpr r31
+; CHECK-NEXT:     if (p0) jump:nt .LBB0_2
 ; CHECK-NEXT:    }
-; CHECK-NEXT:  .LBB0_1: // %b2
+; CHECK-NEXT:  // %bb.1: // %b2
 ; CHECK-NEXT:    {
 ; CHECK-NEXT:     r3:2 = combine(#0,#0)
-; CHECK-NEXT:     r1:0 = memd(r1+#0)
+; CHECK-NEXT:     r1:0 = memd(r0+#0)
 ; CHECK-NEXT:    }
 ; CHECK-NEXT:    {
 ; CHECK-NEXT:     p0 = vcmph.eq(r1:0,r3:2)
@@ -28,7 +27,16 @@ define i32 @fred(ptr %a0) #0 {
 ; CHECK-NEXT:     r0 = and(r0,#1)
 ; CHECK-NEXT:    }
 ; CHECK-NEXT:    {
-; CHECK-NEXT:     r0 = !cmp.eq(r0,#11)
+; CHECK-NEXT:     p0 = cmp.eq(r0,#11)
+; CHECK-NEXT:     r0 = #1
+; CHECK-NEXT:    }
+; CHECK-NEXT:    {
+; CHECK-NEXT:     if (p0) r0 = #0
+; CHECK-NEXT:     jumpr r31
+; CHECK-NEXT:    }
+; CHECK-NEXT:  .LBB0_2: // %b14
+; CHECK-NEXT:    {
+; CHECK-NEXT:     r0 = #0
 ; CHECK-NEXT:     jumpr r31
 ; CHECK-NEXT:    }
 b0:
diff --git a/llvm/test/Transforms/JumpThreading/pr79175.ll b/llvm/test/Transforms/JumpThreading/pr79175.ll
index cce30ce079999..2c7ee0770cdc7 100644
--- a/llvm/test/Transforms/JumpThreading/pr79175.ll
+++ b/llvm/test/Transforms/JumpThreading/pr79175.ll
@@ -17,11 +17,11 @@ define i32 @test(i64 %idx, i32 %val) {
 ; CHECK:       cond.end:
 ; CHECK-NEXT:    [[CMP_I:%.*]] = icmp sgt i32 [[VAL]], 0
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[CMP_I]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[TMP0:%.*]], label [[COND_END_THREAD]]
-; CHECK:       0:
-; CHECK-NEXT:    br label [[COND_END_THREAD]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[COND_END_THREAD]], label [[TMP0:%.*]]
 ; CHECK:       cond.end.thread:
-; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ [[VAL]], [[COND_END]] ], [ 0, [[TMP0]] ], [ 0, [[FOR_BODY]] ]
+; CHECK-NEXT:    br label [[TMP0]]
+; CHECK:       0:
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 0, [[COND_END_THREAD]] ], [ [[VAL]], [[COND_END]] ]
 ; CHECK-NEXT:    [[F_IDX:%.*]] = getelementptr inbounds i32, ptr @f, i64 [[IDX]]
 ; CHECK-NEXT:    store i32 [[TMP1]], ptr [[F_IDX]], align 4
 ; CHECK-NEXT:    [[F_RELOAD:%.*]] = load i32, ptr @f, align 4
diff --git a/llvm/test/Transforms/JumpThreading/select.ll b/llvm/test/Transforms/JumpThreading/select.ll
index 27ebf4c25da50..4ec55a66bb8ac 100644
--- a/llvm/test/Transforms/JumpThreading/select.ll
+++ b/llvm/test/Transforms/JumpThreading/select.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals
 ; RUN: opt -S -passes="jump-threading" -debug-only=branch-prob < %s 2>&1 | FileCheck %s
-; RUN: opt -S -passes="require<branch-prob>,jump-threading" -debug-only=branch-prob  -disable-output < %s 2>&1 | FileCheck -check-prefix=CHECK-BPI %s
+; RUN: opt -S -passes="require<branch-prob>,jump-threading" -debug-only=branch-prob < %s 2>&1 | FileCheck -check-prefixes=CHECK,CHECK-BPI %s
 ; REQUIRES: asserts
 
 ; CHECK-BPI-LABEL:  ---- Branch Probability Info : unfold1 ----
@@ -21,7 +21,7 @@ declare void @quux()
 ; booleans where at least one operand is true/false/undef.
 
 ;.
-; CHECK: @anchor = constant [3 x ptr] [ptr blockaddress(@test_indirectbr, %L1), ptr inttoptr (i32 1 to ptr), ptr blockaddress(@test_indirectbr, %L3)]
+; CHECK: @[[ANCHOR:[a-zA-Z0-9_$"\\.-]+]] = constant [3 x ptr] [ptr blockaddress(@test_indirectbr, [[L1:%.*]]), ptr inttoptr (i32 1 to ptr), ptr blockaddress(@test_indirectbr, [[L3:%.*]])]
 ;.
 define void @test_br(i1 %cond, i1 %value) nounwind {
 ; CHECK-LABEL: @test_br(
@@ -66,8 +66,8 @@ define void @test_switch(i1 %cond, i8 %value) nounwind {
 ; CHECK-NEXT:    call void @quux()
 ; CHECK-NEXT:    [[EXPR:%.*]] = select i1 [[COND]], i8 1, i8 [[VALUE:%.*]]
 ; CHECK-NEXT:    switch i8 [[EXPR]], label [[L3:%.*]] [
-; CHECK-NEXT:      i8 1, label [[L1]]
-; CHECK-NEXT:      i8 2, label [[L2:%.*]]
+; CHECK-NEXT:    i8 1, label [[L1]]
+; CHECK-NEXT:    i8 2, label [[L2:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L1:
 ; CHECK-NEXT:    call void @foo()
@@ -192,8 +192,8 @@ define void @test_switch_cmp(i1 %cond, i32 %val, i8 %value) nounwind {
 ; CHECK:       0:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi i8 [ [[VALUE:%.*]], [[L0]] ]
 ; CHECK-NEXT:    switch i8 [[TMP1]], label [[L3:%.*]] [
-; CHECK-NEXT:      i8 1, label [[L1]]
-; CHECK-NEXT:      i8 2, label [[L2:%.*]]
+; CHECK-NEXT:    i8 1, label [[L1]]
+; CHECK-NEXT:    i8 2, label [[L2:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L1:
 ; CHECK-NEXT:    call void @foo()
@@ -237,8 +237,8 @@ define void @test_switch_default(ptr nocapture %status) nounwind {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[STATUS:%.*]], align 4
 ; CHECK-NEXT:    switch i32 [[TMP0]], label [[L2:%.*]] [
-; CHECK-NEXT:      i32 5061, label [[L2_THREAD:%.*]]
-; CHECK-NEXT:      i32 0, label [[L2]]
+; CHECK-NEXT:    i32 5061, label [[L2_THREAD:%.*]]
+; CHECK-NEXT:    i32 0, label [[L2]]
 ; CHECK-NEXT:    ]
 ; CHECK:       L2.thread:
 ; CHECK-NEXT:    store i32 10025, ptr [[STATUS]], align 4
@@ -377,21 +377,21 @@ define i32 @unfold3(i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z, i32 %j) noun
 ; CHECK-NEXT:    br i1 [[CMP_I]], label [[DOTEXIT_THREAD4:%.*]], label [[COND_FALSE_I:%.*]]
 ; CHECK:       cond.false.i:
 ; CHECK-NEXT:    [[CMP4_I:%.*]] = icmp sgt i32 [[U]], [[V]]
-; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD4]], label [[COND_FALSE_6_I:%.*]]
+; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD:%.*]], label [[COND_FALSE_6_I:%.*]]
 ; CHECK:       cond.false.6.i:
 ; CHECK-NEXT:    [[CMP8_I:%.*]] = icmp slt i32 [[W:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    br i1 [[CMP8_I]], label [[DOTEXIT_THREAD4]], label [[COND_FALSE_10_I:%.*]]
 ; CHECK:       cond.false.10.i:
 ; CHECK-NEXT:    [[CMP13_I:%.*]] = icmp sgt i32 [[W]], [[X]]
-; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD4]], label [[DOTEXIT:%.*]]
+; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD]], label [[DOTEXIT:%.*]]
 ; CHECK:       .exit:
 ; CHECK-NEXT:    [[PHITMP:%.*]] = icmp sge i32 [[Y:%.*]], [[Z:%.*]]
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[PHITMP]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD:%.*]], label [[DOTEXIT_THREAD4]]
-; CHECK:       0:
-; CHECK-NEXT:    br label [[DOTEXIT_THREAD4]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD]], label [[DOTEXIT_THREAD4]]
 ; CHECK:       .exit.thread:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[ADD3]], [[DOTEXIT]] ], [ [[J]], [[DOTEXIT_THREAD]] ], [ [[J]], [[COND_FALSE_I]] ], [ [[J]], [[COND_FALSE_10_I]] ], [ [[ADD3]], [[ENTRY:%.*]] ], [ [[ADD3]], [[COND_FALSE_6_I]] ]
+; CHECK-NEXT:    br label [[DOTEXIT_THREAD4]]
+; CHECK:       .exit.thread4:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[J]], [[DOTEXIT_THREAD]] ], [ [[ADD3]], [[DOTEXIT]] ], [ [[ADD3]], [[ENTRY:%.*]] ], [ [[ADD3]], [[COND_FALSE_6_I]] ]
 ; CHECK-NEXT:    ret i32 [[TMP0]]
 ;
 entry:
@@ -430,23 +430,23 @@ define i32 @unfold4(i32 %u, i32 %v, i32 %w, i32 %x, i32 %y, i32 %z, i32 %j) noun
 ; CHECK-NEXT:    br i1 [[CMP_I]], label [[DOTEXIT_THREAD:%.*]], label [[COND_FALSE_I:%.*]]
 ; CHECK:       cond.false.i:
 ; CHECK-NEXT:    [[CMP4_I:%.*]] = icmp sgt i32 [[U]], [[V]]
-; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD]], label [[COND_FALSE_6_I:%.*]]
+; CHECK-NEXT:    br i1 [[CMP4_I]], label [[DOTEXIT_THREAD5:%.*]], label [[COND_FALSE_6_I:%.*]]
 ; CHECK:       cond.false.6.i:
 ; CHECK-NEXT:    [[CMP8_I:%.*]] = icmp slt i32 [[W:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    br i1 [[CMP8_I]], label [[DOTEXIT_THREAD]], label [[COND_FALSE_10_I:%.*]]
 ; CHECK:       cond.false.10.i:
 ; CHECK-NEXT:    [[CMP13_I:%.*]] = icmp sgt i32 [[W]], [[X]]
-; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD]], label [[DOTEXIT:%.*]]
+; CHECK-NEXT:    br i1 [[CMP13_I]], label [[DOTEXIT_THREAD5]], label [[DOTEXIT:%.*]]
 ; CHECK:       .exit:
 ; CHECK-NEXT:    [[CMP19_I:%.*]] = icmp sge i32 [[Y:%.*]], [[Z:%.*]]
 ; CHECK-NEXT:    [[CONV:%.*]] = zext i1 [[CMP19_I]] to i32
 ; CHECK-NEXT:    [[LNOT_I18:%.*]] = icmp eq i32 [[CONV]], 1
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[LNOT_I18]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[TMP1:%.*]], label [[DOTEXIT_THREAD]]
-; CHECK:       0:
-; CHECK-NEXT:    br label [[DOTEXIT_THREAD]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[DOTEXIT_THREAD]], label [[DOTEXIT_THREAD5]]
 ; CHECK:       .exit.thread:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[ADD3]], [[DOTEXIT]] ], [ [[J]], [[TMP1]] ], [ [[J]], [[ENTRY:%.*]] ], [ [[J]], [[COND_FALSE_6_I]] ], [ [[ADD3]], [[COND_FALSE_I]] ], [ [[ADD3]], [[COND_FALSE_10_I]] ]
+; CHECK-NEXT:    br label [[DOTEXIT_THREAD5]]
+; CHECK:       .exit.thread5:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[J]], [[DOTEXIT_THREAD]] ], [ [[ADD3]], [[DOTEXIT]] ], [ [[ADD3]], [[COND_FALSE_I]] ], [ [[ADD3]], [[COND_FALSE_10_I]] ]
 ; CHECK-NEXT:    ret i32 [[TMP0]]
 ;
 entry:
@@ -560,10 +560,10 @@ define void @test_func(ptr nocapture readonly %a, ptr nocapture readonly %b, ptr
 ; CHECK:       if.end:
 ; CHECK-NEXT:    [[LOCAL_VAR_0:%.*]] = phi i32 [ [[TMP1]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    switch i32 [[LOCAL_VAR_0]], label [[SW_DEFAULT]] [
-; CHECK-NEXT:      i32 2, label [[SW_BB]]
-; CHECK-NEXT:      i32 4, label [[SW_BB7]]
-; CHECK-NEXT:      i32 5, label [[SW_BB8:%.*]]
-; CHECK-NEXT:      i32 7, label [[SW_BB9:%.*]]
+; CHECK-NEXT:    i32 2, label [[SW_BB]]
+; CHECK-NEXT:    i32 4, label [[SW_BB7]]
+; CHECK-NEXT:    i32 5, label [[SW_BB8:%.*]]
+; CHECK-NEXT:    i32 7, label [[SW_BB9:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       sw.bb:
 ; CHECK-NEXT:    call void @foo()
@@ -674,5 +674,3 @@ if.end:
 ; CHECK: [[META0:![0-9]+]] = !{!"function_entry_count", i64 1984}
 ; CHECK: [[PROF1]] = !{!"branch_weights", i64 1073741824, i64 3221225472}
 ;.
-;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; CHECK-BPI: {{.*}}
diff --git a/llvm/test/Transforms/JumpThreading/thread-prob-7.ll b/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
index 4623a579be48f..8c9d89871d00b 100644
--- a/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
+++ b/llvm/test/Transforms/JumpThreading/thread-prob-7.ll
@@ -14,15 +14,15 @@ define i32 @func0(i32 %a0, i32 %a1) !prof !0 {
 ; CHECK-NEXT:    br i1 [[CMP1]], label [[BB_JOIN_THREAD:%.*]], label [[TEST2_FALSE:%.*]], !prof [[PROF2:![0-9]+]]
 ; CHECK:       test2_false:
 ; CHECK-NEXT:    call void @foobar()
-; CHECK-NEXT:    br label [[BB_JOIN_THREAD]]
+; CHECK-NEXT:    br label [[TMP0:%.*]]
 ; CHECK:       bb_join:
 ; CHECK-NEXT:    [[C:%.*]] = phi i1 [ [[CX]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[COND_FR:%.*]] = freeze i1 [[C]]
-; CHECK-NEXT:    br i1 [[COND_FR]], label [[BB_JOIN_THREAD1:%.*]], label [[BB_JOIN_THREAD]], !prof [[PROF3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[COND_FR]], label [[BB_JOIN_THREAD]], label [[TMP0]], !prof [[PROF3:![0-9]+]]
 ; CHECK:       bb_join.thread:
-; CHECK-NEXT:    br label [[BB_JOIN_THREAD]]
+; CHECK-NEXT:    br label [[TMP0]]
 ; CHECK:       0:
-; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 7, [[BB_JOIN]] ], [ 7, [[TEST2_FALSE]] ], [ 42, [[TEST2]] ], [ 42, [[BB_JOIN_THREAD1]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ 42, [[BB_JOIN_THREAD]] ], [ 7, [[BB_JOIN]] ], [ 7, [[TEST2_FALSE]] ]
 ; CHECK-NEXT:    ret i32 [[TMP1]]
 ;
 entry:
diff --git a/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll b/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
deleted file mode 100644
index 6104e8f8778bc..0000000000000
--- a/llvm/test/Transforms/JumpThreading/uncond-no-phi.ll
+++ /dev/null
@@ -1,123 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=jump-threading -S < %s | FileCheck %s
-
-define i1 @if_else(i1 %c, i1 %c1) {
-; CHECK-LABEL: define i1 @if_else(
-; CHECK-SAME: i1 [[C:%.*]], i1 [[C1:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br i1 [[C]], label [[THEN:%.*]], label [[RETURN:%.*]]
-; CHECK:       then:
-; CHECK-NEXT:    call void @dummy()
-; CHECK-NEXT:    br i1 [[C1]], label [[ELSE:%.*]], label [[RETURN]]
-; CHECK:       else:
-; CHECK-NEXT:    br label [[RETURN]]
-; CHECK:       return:
-; CHECK-NEXT:    [[RETVAL_0:%.*]] = phi i1 [ false, [[THEN]] ], [ true, [[ENTRY:%.*]] ], [ true, [[ELSE]] ]
-; CHECK-NEXT:    ret i1 [[RETVAL_0]]
-;
-entry:
-  br i1 %c, label %then, label %else
-
-then:
-  call void @dummy()
-  br i1 %c1, label %else, label %return
-
-else:
-  br label %return
-
-return:
-  %retval.0 = phi i1 [ true, %else ], [ false, %then ]
-  ret i1 %retval.0
-}
-
-define i8 @switch_uncond(i8 %arg) {
-; CHECK-LABEL: define i8 @switch_uncond(
-; CHECK-SAME: i8 [[ARG:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    switch i8 [[ARG]], label [[DEFAULT:%.*]] [
-; CHECK-NEXT:      i8 0, label [[BB1:%.*]]
-; CHECK-NEXT:      i8 1, label [[BB3:%.*]]
-; CHECK-NEXT:      i8 2, label [[BB2:%.*]]
-; CHECK-NEXT:      i8 3, label [[END:%.*]]
-; CHECK-NEXT:    ]
-; CHECK:       default:
-; CHECK-NEXT:    unreachable
-; CHECK:       bb:
-; CHECK-NEXT:    call void @dummy()
-; CHECK-NEXT:    br label [[END]]
-; CHECK:       bb1:
-; CHECK-NEXT:    call void @dummy()
-; CHECK-NEXT:    br label [[END]]
-; CHECK:       bb2:
-; CHECK-NEXT:    br label [[END]]
-; CHECK:       end:
-; CHECK-NEXT:    [[PHI:%.*]] = phi i8 [ 1, [[ENTRY:%.*]] ], [ 0, [[BB3]] ], [ 0, [[BB1]] ], [ 0, [[BB2]] ]
-; CHECK-NEXT:    ret i8 [[PHI]]
-;
-entry:
-  switch i8 %arg, label %default [
-  i8 0, label %bb
-  i8 1, label %bb1
-  i8 2, label %bb2
-  i8 3, label %end
-  ]
-
-default:
-  unreachable
-
-bb:
-  call void @dummy()
-  br label %bb2
-
-bb1:
-  call void @dummy()
-  br label %bb2
-
-; Predecessors of %bb2 are %bb and %bb1, they are not identical.
-; So we can thread %bb2.
-bb2:
-  br label %end
-
-end:
-  %phi = phi i8 [ 0, %bb2 ], [ 1, %entry ]
-  ret i8 %phi
-}
-
-define i8 @switch_uncond_fail(i8 %arg) {
-; CHECK-LABEL: define i8 @switch_uncond_fail(
-; CHECK-SAME: i8 [[ARG:%.*]]) {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    switch i8 [[ARG]], label [[DEFAULT:%.*]] [
-; CHECK-NEXT:      i8 0, label [[BB:%.*]]
-; CHECK-NEXT:      i8 1, label [[BB]]
-; CHECK-NEXT:      i8 2, label [[END:%.*]]
-; CHECK-NEXT:    ]
-; CHECK:       default:
-; CHECK-NEXT:    br label [[END]]
-; CHECK:       bb:
-; CHECK-NEXT:    br label [[END]]
-; CHECK:       end:
-; CHECK-NEXT:    [[PHI:%.*]] = phi i8 [ 0, [[BB]] ], [ 1, [[ENTRY:%.*]] ], [ 2, [[DEFAULT]] ]
-; CHECK-NEXT:    ret i8 [[PHI]]
-;
-entry:
-  switch i8 %arg, label %default [
-  i8 0, label %bb
-  i8 1, label %bb
-  i8 2, label %end
-  ]
-
-default:
-  br label %end
-
-; Predecessor of %bb is only %entry (though there are two in predecessor list),
-; thus it's unthreadable.
-bb:
-  br label %end
-
-end:
-  %phi = phi i8 [ 0, %bb ], [ 1, %entry ], [ 2, %default ]
-  ret i8 %phi
-}
-
-declare void @dummy()
diff --git a/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll b/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll
deleted file mode 100644
index 17146d7d5987f..0000000000000
--- a/llvm/test/Transforms/PhaseOrdering/thread-uncond-bb.ll
+++ /dev/null
@@ -1,62 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt < %s -O3 -S | FileCheck %s
-
-define i32 @thread_uncond_bb_cmp(i1 %c, i32 %v) {
-; CHECK-LABEL: define i32 @thread_uncond_bb_cmp(
-; CHECK-SAME: i1 [[C:%.*]], i32 [[V:%.*]]) local_unnamed_addr {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br i1 [[C]], label [[DO_END:%.*]], label [[IF_THEN:%.*]]
-; CHECK:       if.then:
-; CHECK-NEXT:    tail call void @dummy()
-; CHECK-NEXT:    br label [[DO_END]]
-; CHECK:       return:
-; CHECK-NEXT:    [[RETVAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[V]], [[IF_THEN]] ]
-; CHECK-NEXT:    ret i32 [[RETVAL]]
-;
-entry:
-  br i1 %c, label %do.end, label %if.then
-
-if.then:                                          ; preds = %entry
-  call void @dummy()
-  %tobool = icmp eq i32 %v, 0
-  br i1 %tobool, label %do.end, label %return
-
-do.end:                                           ; preds = %entry, %if.then
-  br label %return
-
-return:                                           ; preds = %if.then, %do.end
-  %retval = phi i32 [ 0, %do.end ], [ %v, %if.then ]
-  ret i32 %retval
-}
-
-define i32 @thread_uncond_bb_cmp_zext(i1 %c, i32 %v) {
-; CHECK-LABEL: define i32 @thread_uncond_bb_cmp_zext(
-; CHECK-SAME: i1 [[C:%.*]], i32 [[V:%.*]]) local_unnamed_addr {
-; CHECK-NEXT:  entry:
-; CHECK-NEXT:    br i1 [[C]], label [[DO_END:%.*]], label [[IF_THEN:%.*]]
-; CHECK:       if.then:
-; CHECK-NEXT:    tail call void @dummy()
-; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[V]], 0
-; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = zext i1 [[TOBOOL]] to i32
-; CHECK-NEXT:    br label [[DO_END]]
-; CHECK:       return:
-; CHECK-NEXT:    [[RETVAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
-; CHECK-NEXT:    ret i32 [[RETVAL]]
-;
-entry:
-  br i1 %c, label %do.end, label %if.then
-
-if.then:                                          ; preds = %entry
-  call void @dummy()
-  %tobool = icmp eq i32 %v, 0
-  br i1 %tobool, label %do.end, label %return
-
-do.end:                                           ; preds = %entry, %if.then
-  br label %return
-
-return:                                           ; preds = %if.then, %do.end
-  %retval = phi i32 [ 0, %do.end ], [ 1, %if.then ]
-  ret i32 %retval
-}
-
-declare void @dummy()

From 22629bb22a1bea95eebfc9b3171005de107c38f1 Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Tue, 16 Apr 2024 10:54:28 -0400
Subject: [PATCH 601/886] [libc++] Use availability to rely on key functions
 for bad_expected_access and bad_function_call (#87390)

This patch uses our availability machinery to allow defining a key
function for bad_function_call and bad_expected_access at all times but
only rely on it when we can. This prevents compilers from complaining
about weak vtables and reduces code bloat and the amount of work done by
the dynamic linker.

rdar://111917845
---
 libcxx/include/__availability                 | 16 ++++++++++++
 libcxx/include/__config                       | 26 ++++---------------
 .../include/__expected/bad_expected_access.h  | 13 ++++++----
 libcxx/include/__functional/function.h        |  5 +++-
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...bcxxabi.v1.stable.exceptions.nonew.abilist |  4 +++
 ...xxabi.v1.stable.noexceptions.nonew.abilist |  4 +++
 libcxx/src/CMakeLists.txt                     |  1 +
 libcxx/src/expected.cpp                       | 13 ++++++++++
 libcxx/src/functional.cpp                     |  2 --
 16 files changed, 83 insertions(+), 29 deletions(-)
 create mode 100644 libcxx/src/expected.cpp

diff --git a/libcxx/include/__availability b/libcxx/include/__availability
index bb3ed0a8da521..aa761eb5bfe5e 100644
--- a/libcxx/include/__availability
+++ b/libcxx/include/__availability
@@ -160,6 +160,15 @@
 #  define _LIBCPP_AVAILABILITY_HAS_TZDB 1
 #  define _LIBCPP_AVAILABILITY_TZDB
 
+// These macros determine whether we assume that std::bad_function_call and
+// std::bad_expected_access provide a key function in the dylib. This allows
+// centralizing their vtable and typeinfo instead of having all TUs provide
+// a weak definition that then gets deduplicated.
+#  define _LIBCPP_AVAILABILITY_HAS_BAD_FUNCTION_CALL_KEY_FUNCTION 1
+#  define _LIBCPP_AVAILABILITY_BAD_FUNCTION_CALL_KEY_FUNCTION
+#  define _LIBCPP_AVAILABILITY_HAS_BAD_EXPECTED_ACCESS_KEY_FUNCTION 1
+#  define _LIBCPP_AVAILABILITY_BAD_EXPECTED_ACCESS_KEY_FUNCTION
+
 #elif defined(__APPLE__)
 
 #  define _LIBCPP_AVAILABILITY_HAS_BAD_OPTIONAL_ACCESS                                                                 \
@@ -290,6 +299,13 @@
 #  else
 #    define _LIBCPP_AVAILABILITY_HAS_ADDITIONAL_IOSTREAM_EXPLICIT_INSTANTIATIONS_1 1
 #  endif
+
+#  define _LIBCPP_AVAILABILITY_HAS_BAD_FUNCTION_CALL_KEY_FUNCTION 0
+#  define _LIBCPP_AVAILABILITY_BAD_FUNCTION_CALL_KEY_FUNCTION __attribute__((unavailable))
+
+#  define _LIBCPP_AVAILABILITY_HAS_BAD_EXPECTED_ACCESS_KEY_FUNCTION 0
+#  define _LIBCPP_AVAILABILITY_BAD_EXPECTED_ACCESS_KEY_FUNCTION __attribute__((unavailable))
+
 #else
 
 // ...New vendors can add availability markup here...
diff --git a/libcxx/include/__config b/libcxx/include/__config
index 82782b31c557b..e9fda9cd24ebb 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -120,14 +120,11 @@
 #    define _LIBCPP_ABI_FIX_UNORDERED_NODE_POINTER_UB
 #    define _LIBCPP_ABI_FORWARD_LIST_REMOVE_NODE_POINTER_UB
 #    define _LIBCPP_ABI_FIX_UNORDERED_CONTAINER_SIZE_TYPE
-// Define a key function for `bad_function_call` in the library, to centralize
-// its vtable and typeinfo to libc++ rather than having all other libraries
-// using that class define their own copies.
-#    define _LIBCPP_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
-// Override the default return value of exception::what() for
-// bad_function_call::what() with a string that is specific to
-// bad_function_call (see http://wg21.link/LWG2233). This is an ABI break
-// because it changes the vtable layout of bad_function_call.
+// Override the default return value of exception::what() for bad_function_call::what()
+// with a string that is specific to bad_function_call (see http://wg21.link/LWG2233).
+// This is an ABI break on platforms that sign and authenticate vtable function pointers
+// because it changes the mangling of the virtual function located in the vtable, which
+// changes how it gets signed.
 #    define _LIBCPP_ABI_BAD_FUNCTION_CALL_GOOD_WHAT_MESSAGE
 // Enable optimized version of __do_get_(un)signed which avoids redundant copies.
 #    define _LIBCPP_ABI_OPTIMIZED_LOCALE_NUM_GET
@@ -197,19 +194,6 @@
 #    if defined(__FreeBSD__) && __FreeBSD__ < 14
 #      define _LIBCPP_DEPRECATED_ABI_DISABLE_PAIR_TRIVIAL_COPY_CTOR
 #    endif
-// For XCOFF linkers, we have problems if we see a weak hidden version of a symbol
-// in user code (like you get with -fvisibility-inlines-hidden) and then a strong def
-// in the library, so we need to always rely on the library version.
-#    if defined(_AIX)
-#      define _LIBCPP_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
-#    endif
-#  endif
-
-#  if defined(_LIBCPP_BUILDING_LIBRARY) || _LIBCPP_ABI_VERSION >= 2
-// Define a key function for `bad_function_call` in the library, to centralize
-// its vtable and typeinfo to libc++ rather than having all other libraries
-// using that class define their own copies.
-#    define _LIBCPP_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
 #  endif
 
 // We had some bugs where we use [[no_unique_address]] together with construct_at,
diff --git a/libcxx/include/__expected/bad_expected_access.h b/libcxx/include/__expected/bad_expected_access.h
index 9d490307b6808..ef29fa5088313 100644
--- a/libcxx/include/__expected/bad_expected_access.h
+++ b/libcxx/include/__expected/bad_expected_access.h
@@ -9,6 +9,7 @@
 #ifndef _LIBCPP___EXPECTED_BAD_EXPECTED_ACCESS_H
 #define _LIBCPP___EXPECTED_BAD_EXPECTED_ACCESS_H
 
+#include <__availability>
 #include <__config>
 #include <__exception/exception.h>
 #include <__utility/move.h>
@@ -28,9 +29,11 @@ template <class _Err>
 class bad_expected_access;
 
 _LIBCPP_DIAGNOSTIC_PUSH
+#  if !_LIBCPP_AVAILABILITY_HAS_BAD_EXPECTED_ACCESS_KEY_FUNCTION
 _LIBCPP_CLANG_DIAGNOSTIC_IGNORED("-Wweak-vtables")
+#  endif
 template <>
-class bad_expected_access<void> : public exception {
+class _LIBCPP_EXPORTED_FROM_ABI bad_expected_access<void> : public exception {
 protected:
   _LIBCPP_HIDE_FROM_ABI bad_expected_access() noexcept                                      = default;
   _LIBCPP_HIDE_FROM_ABI bad_expected_access(const bad_expected_access&) noexcept            = default;
@@ -40,11 +43,11 @@ class bad_expected_access<void> : public exception {
   _LIBCPP_HIDE_FROM_ABI_VIRTUAL ~bad_expected_access() override                             = default;
 
 public:
-  // The way this has been designed (by using a class template below) means that we'll already
-  // have a profusion of these vtables in TUs, and the dynamic linker will already have a bunch
-  // of work to do. So it is not worth hiding the <void> specialization in the dylib, given that
-  // it adds deployment target restrictions.
+#  if _LIBCPP_AVAILABILITY_HAS_BAD_EXPECTED_ACCESS_KEY_FUNCTION
+  const char* what() const noexcept override;
+#  else
   _LIBCPP_HIDE_FROM_ABI_VIRTUAL const char* what() const noexcept override { return "bad access to std::expected"; }
+#  endif
 };
 _LIBCPP_DIAGNOSTIC_POP
 
diff --git a/libcxx/include/__functional/function.h b/libcxx/include/__functional/function.h
index 1faa9e92ebd63..36057706933d4 100644
--- a/libcxx/include/__functional/function.h
+++ b/libcxx/include/__functional/function.h
@@ -11,6 +11,7 @@
 #define _LIBCPP___FUNCTIONAL_FUNCTION_H
 
 #include <__assert>
+#include <__availability>
 #include <__config>
 #include <__exception/exception.h>
 #include <__functional/binary_function.h>
@@ -55,7 +56,9 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 // bad_function_call
 
 _LIBCPP_DIAGNOSTIC_PUSH
+#  if !_LIBCPP_AVAILABILITY_HAS_BAD_FUNCTION_CALL_KEY_FUNCTION
 _LIBCPP_CLANG_DIAGNOSTIC_IGNORED("-Wweak-vtables")
+#  endif
 class _LIBCPP_EXPORTED_FROM_ABI bad_function_call : public exception {
 public:
   _LIBCPP_HIDE_FROM_ABI bad_function_call() _NOEXCEPT                                    = default;
@@ -64,7 +67,7 @@ class _LIBCPP_EXPORTED_FROM_ABI bad_function_call : public exception {
 // Note that when a key function is not used, every translation unit that uses
 // bad_function_call will end up containing a weak definition of the vtable and
 // typeinfo.
-#  ifdef _LIBCPP_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
+#  if _LIBCPP_AVAILABILITY_HAS_BAD_FUNCTION_CALL_KEY_FUNCTION
   ~bad_function_call() _NOEXCEPT override;
 #  else
   _LIBCPP_HIDE_FROM_ABI_VIRTUAL ~bad_function_call() _NOEXCEPT override {}
diff --git a/libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
index 46353986f5d7d..64cf368e6e684 100644
--- a/libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -575,6 +575,7 @@
 {'is_defined': True, 'name': '__ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '__ZNKSt3__119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftER11__mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -2073,6 +2074,7 @@
 {'is_defined': True, 'name': '__ZTINSt3__117moneypunct_bynameIwLb1EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119__shared_weak_countE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTINSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'size': 0, 'type': 'OBJECT'}
@@ -2264,6 +2266,7 @@
 {'is_defined': True, 'name': '__ZTSNSt3__117moneypunct_bynameIwLb0EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__117moneypunct_bynameIwLb1EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTSNSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__13pmr15memory_resourceE', 'size': 0, 'type': 'OBJECT'}
@@ -2482,6 +2485,7 @@
 {'is_defined': True, 'name': '__ZTVNSt3__117moneypunct_bynameIwLb1EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119__shared_weak_countE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTVNSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 0, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/i686-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/i686-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
index fec3a4505a0c6..8751dffe23025 100644
--- a/libcxx/lib/abi/i686-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/i686-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -209,6 +209,7 @@
 {'is_defined': True, 'name': '_ZNKSt6__ndk118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNKSt6__ndk119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE10do_unshiftER9mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -1722,6 +1723,7 @@
 {'is_defined': True, 'name': '_ZTINSt6__ndk118__time_get_storageIwEE', 'size': 12, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 12, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119__shared_weak_countE', 'size': 24, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTINSt6__ndk119bad_expected_accessIvEE', 'size': 12, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 12, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 12, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 12, 'type': 'OBJECT'}
@@ -1958,6 +1960,7 @@
 {'is_defined': True, 'name': '_ZTSNSt6__ndk118__time_get_storageIwEE', 'size': 35, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 72, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119__shared_weak_countE', 'size': 33, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTSNSt6__ndk119bad_expected_accessIvEE', 'size': 36, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 73, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 73, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 38, 'type': 'OBJECT'}
@@ -2188,6 +2191,7 @@
 {'is_defined': True, 'name': '_ZTVNSt6__ndk117moneypunct_bynameIwLb1EEE', 'size': 56, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 60, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119__shared_weak_countE', 'size': 28, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTVNSt6__ndk119bad_expected_accessIvEE', 'size': 20, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 48, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/powerpc-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/powerpc-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
index e52cf98dd4c4f..7e223e6652884 100644
--- a/libcxx/lib/abi/powerpc-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/powerpc-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -99,6 +99,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__119bad_expected_accessIvE4whatEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftERPcS2_S2_S3_', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
@@ -910,6 +911,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__117bad_function_callE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__119__shared_weak_countE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
@@ -969,6 +971,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__117bad_function_callE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__119__shared_weak_countE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
@@ -1031,6 +1034,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__117bad_function_callE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__119__shared_weak_countE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/powerpc64-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/powerpc64-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
index 52a04706ddf20..407d0456757af 100644
--- a/libcxx/lib/abi/powerpc64-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/powerpc64-ibm-aix.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -99,6 +99,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__119bad_expected_accessIvE4whatEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftERPcS2_S2_S3_', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'storage_mapping_class': 'DS', 'type': 'FUNC'}
@@ -910,6 +911,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__117bad_function_callE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__119__shared_weak_countE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
@@ -969,6 +971,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__117bad_function_callE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__119__shared_weak_countE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RO', 'type': 'OBJECT'}
@@ -1031,6 +1034,7 @@
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__117__widen_from_utf8ILm32EEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__117bad_function_callE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__119__shared_weak_countE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
+{'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__119bad_expected_accessIvEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDsEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
 {'import_export': 'EXP', 'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IwEE', 'storage_mapping_class': 'RW', 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/x86_64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/x86_64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
index c169b4a992521..d578b41383c0e 100644
--- a/libcxx/lib/abi/x86_64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/x86_64-apple-darwin.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -575,6 +575,7 @@
 {'is_defined': True, 'name': '__ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '__ZNKSt3__119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftER11__mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '__ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -2087,6 +2088,7 @@
 {'is_defined': True, 'name': '__ZTINSt3__118__time_get_storageIwEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119__shared_weak_countE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTINSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'size': 0, 'type': 'OBJECT'}
@@ -2291,6 +2293,7 @@
 {'is_defined': True, 'name': '__ZTSNSt3__117moneypunct_bynameIwLb0EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__117moneypunct_bynameIwLb1EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTSNSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTSNSt3__13pmr15memory_resourceE', 'size': 0, 'type': 'OBJECT'}
@@ -2516,6 +2519,7 @@
 {'is_defined': True, 'name': '__ZTVNSt3__117moneypunct_bynameIwLb1EEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119__shared_weak_countE', 'size': 0, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '__ZTVNSt3__119bad_expected_accessIvEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 0, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '__ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 0, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/x86_64-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/x86_64-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
index efa2189e9c928..fc0f4fcf415e6 100644
--- a/libcxx/lib/abi/x86_64-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/x86_64-linux-android21.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -209,6 +209,7 @@
 {'is_defined': True, 'name': '_ZNKSt6__ndk118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNKSt6__ndk119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE10do_unshiftER9mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt6__ndk120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -1722,6 +1723,7 @@
 {'is_defined': True, 'name': '_ZTINSt6__ndk118__time_get_storageIwEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119__shared_weak_countE', 'size': 40, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTINSt6__ndk119bad_expected_accessIvEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 24, 'type': 'OBJECT'}
@@ -1955,6 +1957,7 @@
 {'is_defined': True, 'name': '_ZTSNSt6__ndk118__time_get_storageIwEE', 'size': 35, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 72, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119__shared_weak_countE', 'size': 33, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTSNSt6__ndk119bad_expected_accessIvEE', 'size': 36, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 73, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 73, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 38, 'type': 'OBJECT'}
@@ -2182,6 +2185,7 @@
 {'is_defined': True, 'name': '_ZTVNSt6__ndk117moneypunct_bynameIwLb1EEE', 'size': 112, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 120, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119__shared_weak_countE', 'size': 56, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTVNSt6__ndk119bad_expected_accessIvEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt6__ndk120__codecvt_utf8_utf16IDiEE', 'size': 96, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/x86_64-unknown-freebsd.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/x86_64-unknown-freebsd.libcxxabi.v1.stable.exceptions.nonew.abilist
index ebda5b0dfba57..4022339562b3a 100644
--- a/libcxx/lib/abi/x86_64-unknown-freebsd.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/x86_64-unknown-freebsd.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -266,6 +266,7 @@
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNKSt3__119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftER11__mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -1695,6 +1696,7 @@
 {'is_defined': True, 'name': '_ZTINSt3__118__time_get_storageIwEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119__shared_weak_countE', 'size': 40, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTINSt3__119bad_expected_accessIvEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'size': 24, 'type': 'OBJECT'}
@@ -1829,6 +1831,7 @@
 {'is_defined': True, 'name': '_ZTSNSt3__118__time_get_storageIwEE', 'size': 32, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 69, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119__shared_weak_countE', 'size': 30, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTSNSt3__119bad_expected_accessIvEE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 35, 'type': 'OBJECT'}
@@ -1962,6 +1965,7 @@
 {'is_defined': True, 'name': '_ZTVNSt3__117moneypunct_bynameIwLb1EEE', 'size': 112, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 120, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119__shared_weak_countE', 'size': 56, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTVNSt3__119bad_expected_accessIvEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 96, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.exceptions.nonew.abilist b/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.exceptions.nonew.abilist
index 6432ad3be3585..574c4504c59b8 100644
--- a/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.exceptions.nonew.abilist
+++ b/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.exceptions.nonew.abilist
@@ -264,6 +264,7 @@
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNKSt3__119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftER11__mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -1696,6 +1697,7 @@
 {'is_defined': True, 'name': '_ZTINSt3__118__time_get_storageIwEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119__shared_weak_countE', 'size': 40, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTINSt3__119bad_expected_accessIvEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'size': 24, 'type': 'OBJECT'}
@@ -1830,6 +1832,7 @@
 {'is_defined': True, 'name': '_ZTSNSt3__118__time_get_storageIwEE', 'size': 32, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 69, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119__shared_weak_countE', 'size': 30, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTSNSt3__119bad_expected_accessIvEE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 35, 'type': 'OBJECT'}
@@ -1963,6 +1966,7 @@
 {'is_defined': True, 'name': '_ZTVNSt3__117moneypunct_bynameIwLb1EEE', 'size': 112, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 120, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119__shared_weak_countE', 'size': 56, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTVNSt3__119bad_expected_accessIvEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 96, 'type': 'OBJECT'}
diff --git a/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.noexceptions.nonew.abilist b/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.noexceptions.nonew.abilist
index 1fe84e17b3f7f..665546699e8de 100644
--- a/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.noexceptions.nonew.abilist
+++ b/libcxx/lib/abi/x86_64-unknown-linux-gnu.libcxxabi.v1.stable.noexceptions.nonew.abilist
@@ -235,6 +235,7 @@
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIcE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__118__time_get_storageIwE15__do_date_orderEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__119__shared_weak_count13__get_deleterERKSt9type_info', 'type': 'FUNC'}
+{'is_defined': True, 'name': '_ZNKSt3__119bad_expected_accessIvE4whatEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE10do_unshiftER11__mbstate_tPcS4_RS4_', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE11do_encodingEv', 'type': 'FUNC'}
 {'is_defined': True, 'name': '_ZNKSt3__120__codecvt_utf8_utf16IDiE13do_max_lengthEv', 'type': 'FUNC'}
@@ -1667,6 +1668,7 @@
 {'is_defined': True, 'name': '_ZTINSt3__118__time_get_storageIwEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119__shared_weak_countE', 'size': 40, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTINSt3__119bad_expected_accessIvEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 24, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTINSt3__120__codecvt_utf8_utf16IDiEE', 'size': 24, 'type': 'OBJECT'}
@@ -1801,6 +1803,7 @@
 {'is_defined': True, 'name': '_ZTSNSt3__118__time_get_storageIwEE', 'size': 32, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 69, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119__shared_weak_countE', 'size': 30, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTSNSt3__119bad_expected_accessIvEE', 'size': 33, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 70, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTSNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 35, 'type': 'OBJECT'}
@@ -1934,6 +1937,7 @@
 {'is_defined': True, 'name': '_ZTVNSt3__117moneypunct_bynameIwLb1EEE', 'size': 112, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__118basic_stringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 120, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119__shared_weak_countE', 'size': 56, 'type': 'OBJECT'}
+{'is_defined': True, 'name': '_ZTVNSt3__119bad_expected_accessIvEE', 'size': 40, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_istringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__119basic_ostringstreamIcNS_11char_traitsIcEENS_9allocatorIcEEEE', 'size': 80, 'type': 'OBJECT'}
 {'is_defined': True, 'name': '_ZTVNSt3__120__codecvt_utf8_utf16IDiEE', 'size': 96, 'type': 'OBJECT'}
diff --git a/libcxx/src/CMakeLists.txt b/libcxx/src/CMakeLists.txt
index 208500ec14fcd..a4a3fee864571 100644
--- a/libcxx/src/CMakeLists.txt
+++ b/libcxx/src/CMakeLists.txt
@@ -10,6 +10,7 @@ set(LIBCXX_SOURCES
   chrono.cpp
   error_category.cpp
   exception.cpp
+  expected.cpp
   filesystem/filesystem_clock.cpp
   filesystem/filesystem_error.cpp
   filesystem/path_parser.h
diff --git a/libcxx/src/expected.cpp b/libcxx/src/expected.cpp
new file mode 100644
index 0000000000000..f30efb5164796
--- /dev/null
+++ b/libcxx/src/expected.cpp
@@ -0,0 +1,13 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <expected>
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+const char* bad_expected_access<void>::what() const noexcept { return "bad access to std::expected"; }
+_LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/src/functional.cpp b/libcxx/src/functional.cpp
index 570bb78e150b7..ef53e3e84da0e 100644
--- a/libcxx/src/functional.cpp
+++ b/libcxx/src/functional.cpp
@@ -10,9 +10,7 @@
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-#ifdef _LIBCPP_ABI_BAD_FUNCTION_CALL_KEY_FUNCTION
 bad_function_call::~bad_function_call() noexcept {}
-#endif
 
 #ifdef _LIBCPP_ABI_BAD_FUNCTION_CALL_GOOD_WHAT_MESSAGE
 const char* bad_function_call::what() const noexcept { return "std::bad_function_call"; }

From 9ddedf07ed80076e0e419940753aeaaf719a09ec Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Tue, 16 Apr 2024 10:57:48 -0400
Subject: [PATCH 602/886] [libc++] Deprecate the C++20 synchronization library
 before C++20 (#86410)

When we initially implemented the C++20 synchronization library, we
reluctantly accepted for the implementation to be backported to C++03
upon request from the person who provided the patch. This was when we
were only starting to have experience with the issues this can create,
so we flinched. Nowadays, we have a much stricter stance about not
backporting features to previous standards.

We have recently started fixing several bugs (and near bugs) in our
implementation of the synchronization library. A recurring theme during
these reviews has been how difficult to understand the current code is,
and upon inspection it becomes clear that being able to use a few recent
C++ features (in particular lambdas) would help a great deal. The code
would still be pretty intricate, but it would be a lot easier to reason
about the flow of callbacks through things like
__thread_poll_with_backoff.

As a result, this patch deprecates support for the synchronization
library before C++20. In the next release, we can remove that support
entirely.
---
 libcxx/.clang-format                          |  1 +
 libcxx/docs/ReleaseNotes/19.rst               |  4 +++
 libcxx/include/__atomic/atomic.h              | 12 ++++---
 libcxx/include/__atomic/atomic_flag.h         | 34 +++++++++++--------
 libcxx/include/__config                       |  8 +++++
 libcxx/include/barrier                        |  2 +-
 libcxx/include/latch                          |  2 +-
 libcxx/include/semaphore                      |  6 ++--
 .../atomic_notify_all.pass.cpp                |  5 ++-
 .../atomic_notify_one.pass.cpp                |  5 ++-
 .../atomic_wait.pass.cpp                      |  5 ++-
 .../atomic_wait_explicit.pass.cpp             |  5 ++-
 .../std/thread/thread.barrier/arrive.pass.cpp |  3 ++
 .../thread.barrier/arrive_and_drop.pass.cpp   |  3 ++
 .../thread.barrier/arrive_and_wait.pass.cpp   |  3 ++
 .../thread/thread.barrier/completion.pass.cpp |  3 ++
 .../thread.barrier/ctor.compile.pass.cpp      |  3 ++
 .../std/thread/thread.barrier/max.pass.cpp    |  3 ++
 .../thread.latch/arrive_and_wait.pass.cpp     |  3 ++
 .../thread/thread.latch/count_down.pass.cpp   |  3 ++
 .../std/thread/thread.latch/ctor.pass.cpp     |  3 ++
 .../test/std/thread/thread.latch/max.pass.cpp |  3 ++
 .../std/thread/thread.latch/try_wait.pass.cpp |  3 ++
 .../thread/thread.semaphore/acquire.pass.cpp  |  3 ++
 .../thread/thread.semaphore/binary.pass.cpp   |  3 ++
 .../thread.semaphore/ctor.compile.pass.cpp    |  3 ++
 .../std/thread/thread.semaphore/max.pass.cpp  |  3 ++
 .../thread/thread.semaphore/release.pass.cpp  |  3 ++
 .../thread/thread.semaphore/timed.pass.cpp    |  3 ++
 .../thread.semaphore/try_acquire.pass.cpp     |  3 ++
 30 files changed, 117 insertions(+), 26 deletions(-)

diff --git a/libcxx/.clang-format b/libcxx/.clang-format
index 39ae1322ffa8a..c37ab817bca90 100644
--- a/libcxx/.clang-format
+++ b/libcxx/.clang-format
@@ -24,6 +24,7 @@ AttributeMacros: [
                   '_LIBCPP_CONSTEXPR_SINCE_CXX23',
                   '_LIBCPP_CONSTEXPR',
                   '_LIBCPP_CONSTINIT',
+                  '_LIBCPP_DEPRECATED_ATOMIC_SYNC',
                   '_LIBCPP_DEPRECATED_IN_CXX11',
                   '_LIBCPP_DEPRECATED_IN_CXX14',
                   '_LIBCPP_DEPRECATED_IN_CXX17',
diff --git a/libcxx/docs/ReleaseNotes/19.rst b/libcxx/docs/ReleaseNotes/19.rst
index e5db17daa4823..45aac88e45502 100644
--- a/libcxx/docs/ReleaseNotes/19.rst
+++ b/libcxx/docs/ReleaseNotes/19.rst
@@ -75,6 +75,10 @@ Improvements and New Features
 Deprecations and Removals
 -------------------------
 
+- The C++20 synchronization library (``<barrier>``, ``<latch>``, ``atomic::wait``, etc.) has been deprecated
+  in language modes prior to C++20. If you are using these features prior to C++20, please update to ``-std=c++20``.
+  In LLVM 20, the C++20 synchronization library will be removed entirely in language modes prior to C++20.
+
 - TODO: The ``LIBCXX_ENABLE_ASSERTIONS`` CMake variable that was used to enable the safe mode has been deprecated and setting
   it triggers an error; use the ``LIBCXX_HARDENING_MODE`` CMake variable with the value ``extensive`` instead. Similarly,
   the ``_LIBCPP_ENABLE_ASSERTIONS`` macro has been deprecated (setting it to ``1`` still enables the extensive mode in
diff --git a/libcxx/include/__atomic/atomic.h b/libcxx/include/__atomic/atomic.h
index 3dfb6937d0325..bd3f659c22df0 100644
--- a/libcxx/include/__atomic/atomic.h
+++ b/libcxx/include/__atomic/atomic.h
@@ -462,22 +462,26 @@ atomic_wait_explicit(const atomic<_Tp>* __o, typename atomic<_Tp>::value_type __
 // atomic_notify_one
 
 template <class _Tp>
-_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void atomic_notify_one(volatile atomic<_Tp>* __o) _NOEXCEPT {
+_LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+atomic_notify_one(volatile atomic<_Tp>* __o) _NOEXCEPT {
   __o->notify_one();
 }
 template <class _Tp>
-_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void atomic_notify_one(atomic<_Tp>* __o) _NOEXCEPT {
+_LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+atomic_notify_one(atomic<_Tp>* __o) _NOEXCEPT {
   __o->notify_one();
 }
 
 // atomic_notify_all
 
 template <class _Tp>
-_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void atomic_notify_all(volatile atomic<_Tp>* __o) _NOEXCEPT {
+_LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+atomic_notify_all(volatile atomic<_Tp>* __o) _NOEXCEPT {
   __o->notify_all();
 }
 template <class _Tp>
-_LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void atomic_notify_all(atomic<_Tp>* __o) _NOEXCEPT {
+_LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+atomic_notify_all(atomic<_Tp>* __o) _NOEXCEPT {
   __o->notify_all();
 }
 
diff --git a/libcxx/include/__atomic/atomic_flag.h b/libcxx/include/__atomic/atomic_flag.h
index 084366237c16e..3ec3366ecaaf9 100644
--- a/libcxx/include/__atomic/atomic_flag.h
+++ b/libcxx/include/__atomic/atomic_flag.h
@@ -49,22 +49,26 @@ struct atomic_flag {
     __cxx_atomic_store(&__a_, _LIBCPP_ATOMIC_FLAG_TYPE(false), __m);
   }
 
-  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void wait(bool __v, memory_order __m = memory_order_seq_cst) const
-      volatile _NOEXCEPT {
+  _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+  wait(bool __v, memory_order __m = memory_order_seq_cst) const volatile _NOEXCEPT {
     std::__atomic_wait(*this, _LIBCPP_ATOMIC_FLAG_TYPE(__v), __m);
   }
-  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
+  _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void
   wait(bool __v, memory_order __m = memory_order_seq_cst) const _NOEXCEPT {
     std::__atomic_wait(*this, _LIBCPP_ATOMIC_FLAG_TYPE(__v), __m);
   }
-  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_one() volatile _NOEXCEPT {
+  _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_one() volatile _NOEXCEPT {
+    std::__atomic_notify_one(*this);
+  }
+  _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_one() _NOEXCEPT {
     std::__atomic_notify_one(*this);
   }
-  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_one() _NOEXCEPT { std::__atomic_notify_one(*this); }
   _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_all() volatile _NOEXCEPT {
     std::__atomic_notify_all(*this);
   }
-  _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_all() _NOEXCEPT { std::__atomic_notify_all(*this); }
+  _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_AVAILABILITY_SYNC _LIBCPP_HIDE_FROM_ABI void notify_all() _NOEXCEPT {
+    std::__atomic_notify_all(*this);
+  }
 
 #if _LIBCPP_STD_VER >= 20
   _LIBCPP_HIDE_FROM_ABI constexpr atomic_flag() _NOEXCEPT : __a_(false) {}
@@ -141,41 +145,43 @@ inline _LIBCPP_HIDE_FROM_ABI void atomic_flag_clear_explicit(atomic_flag* __o, m
   __o->clear(__m);
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_wait(const volatile atomic_flag* __o, bool __v) _NOEXCEPT {
   __o->wait(__v);
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_wait(const atomic_flag* __o, bool __v) _NOEXCEPT {
   __o->wait(__v);
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_wait_explicit(const volatile atomic_flag* __o, bool __v, memory_order __m) _NOEXCEPT {
   __o->wait(__v, __m);
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_wait_explicit(const atomic_flag* __o, bool __v, memory_order __m) _NOEXCEPT {
   __o->wait(__v, __m);
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_notify_one(volatile atomic_flag* __o) _NOEXCEPT {
   __o->notify_one();
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void atomic_flag_notify_one(atomic_flag* __o) _NOEXCEPT {
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+atomic_flag_notify_one(atomic_flag* __o) _NOEXCEPT {
   __o->notify_one();
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
 atomic_flag_notify_all(volatile atomic_flag* __o) _NOEXCEPT {
   __o->notify_all();
 }
 
-inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void atomic_flag_notify_all(atomic_flag* __o) _NOEXCEPT {
+inline _LIBCPP_DEPRECATED_ATOMIC_SYNC _LIBCPP_HIDE_FROM_ABI _LIBCPP_AVAILABILITY_SYNC void
+atomic_flag_notify_all(atomic_flag* __o) _NOEXCEPT {
   __o->notify_all();
 }
 
diff --git a/libcxx/include/__config b/libcxx/include/__config
index e9fda9cd24ebb..9b4155af1e3c6 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -956,6 +956,14 @@ typedef __char32_t char32_t;
 #    define _LIBCPP_DEPRECATED_(m)
 #  endif
 
+#  if _LIBCPP_STD_VER < 20
+#    define _LIBCPP_DEPRECATED_ATOMIC_SYNC                                                                             \
+      _LIBCPP_DEPRECATED_("The C++20 synchronization library has been deprecated prior to C++20. Please update to "    \
+                          "using -std=c++20 if you need to use these facilities.")
+#  else
+#    define _LIBCPP_DEPRECATED_ATOMIC_SYNC /* nothing */
+#  endif
+
 #  if !defined(_LIBCPP_CXX03_LANG)
 #    define _LIBCPP_DEPRECATED_IN_CXX11 _LIBCPP_DEPRECATED
 #  else
diff --git a/libcxx/include/barrier b/libcxx/include/barrier
index c5fd84b91925b..d776078267625 100644
--- a/libcxx/include/barrier
+++ b/libcxx/include/barrier
@@ -257,7 +257,7 @@ public:
 #  endif // !_LIBCPP_HAS_NO_TREE_BARRIER
 
 template <class _CompletionF = __empty_completion>
-class barrier {
+class _LIBCPP_DEPRECATED_ATOMIC_SYNC barrier {
   __barrier_base<_CompletionF> __b_;
 
 public:
diff --git a/libcxx/include/latch b/libcxx/include/latch
index 3cc7258381143..1937617f7dcc6 100644
--- a/libcxx/include/latch
+++ b/libcxx/include/latch
@@ -66,7 +66,7 @@ _LIBCPP_PUSH_MACROS
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-class latch {
+class _LIBCPP_DEPRECATED_ATOMIC_SYNC latch {
   __atomic_base<ptrdiff_t> __a_;
 
 public:
diff --git a/libcxx/include/semaphore b/libcxx/include/semaphore
index 1375ec3f7c04b..cb2f42c106ca8 100644
--- a/libcxx/include/semaphore
+++ b/libcxx/include/semaphore
@@ -127,7 +127,7 @@ private:
 };
 
 template <ptrdiff_t __least_max_value = _LIBCPP_SEMAPHORE_MAX>
-class counting_semaphore {
+class _LIBCPP_DEPRECATED_ATOMIC_SYNC counting_semaphore {
   __atomic_semaphore_base __semaphore_;
 
 public:
@@ -172,7 +172,9 @@ public:
   }
 };
 
-using binary_semaphore = counting_semaphore<1>;
+_LIBCPP_SUPPRESS_DEPRECATED_PUSH
+using binary_semaphore _LIBCPP_DEPRECATED_ATOMIC_SYNC = counting_semaphore<1>;
+_LIBCPP_SUPPRESS_DEPRECATED_POP
 
 _LIBCPP_END_NAMESPACE_STD
 
diff --git a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_all.pass.cpp b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_all.pass.cpp
index 2b9f34b731f87..0ec530c922e70 100644
--- a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_all.pass.cpp
+++ b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_all.pass.cpp
@@ -7,9 +7,12 @@
 //===----------------------------------------------------------------------===//
 //
 // UNSUPPORTED: no-threads
-// XFAIL: c++03
+// UNSUPPORTED: c++03
 // XFAIL: !has-1024-bit-atomics
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <atomic>
diff --git a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_one.pass.cpp b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_one.pass.cpp
index dfa781c566009..c21b67d479ae2 100644
--- a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_one.pass.cpp
+++ b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_notify_one.pass.cpp
@@ -7,9 +7,12 @@
 //===----------------------------------------------------------------------===//
 //
 // UNSUPPORTED: no-threads
-// XFAIL: c++03
+// UNSUPPORTED: c++03
 // XFAIL: !has-1024-bit-atomics
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <atomic>
diff --git a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait.pass.cpp b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait.pass.cpp
index 38142b336e72c..af99113f13499 100644
--- a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait.pass.cpp
+++ b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait.pass.cpp
@@ -7,9 +7,12 @@
 //===----------------------------------------------------------------------===//
 //
 // UNSUPPORTED: no-threads
-// XFAIL: c++03
+// UNSUPPORTED: c++03
 // XFAIL: !has-1024-bit-atomics
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <atomic>
diff --git a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait_explicit.pass.cpp b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait_explicit.pass.cpp
index 2db95a0b67a7f..bb8c64593b54b 100644
--- a/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait_explicit.pass.cpp
+++ b/libcxx/test/std/atomics/atomics.types.operations/atomics.types.operations.wait/atomic_wait_explicit.pass.cpp
@@ -7,9 +7,12 @@
 //===----------------------------------------------------------------------===//
 //
 // UNSUPPORTED: no-threads
-// XFAIL: c++03
+// UNSUPPORTED: c++03
 // XFAIL: !has-1024-bit-atomics
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <atomic>
diff --git a/libcxx/test/std/thread/thread.barrier/arrive.pass.cpp b/libcxx/test/std/thread/thread.barrier/arrive.pass.cpp
index 18cdc6d654ac2..d9d9c1dba6bbb 100644
--- a/libcxx/test/std/thread/thread.barrier/arrive.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/arrive.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <barrier>
diff --git a/libcxx/test/std/thread/thread.barrier/arrive_and_drop.pass.cpp b/libcxx/test/std/thread/thread.barrier/arrive_and_drop.pass.cpp
index 3fc48261de1b1..aff7b26e16f70 100644
--- a/libcxx/test/std/thread/thread.barrier/arrive_and_drop.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/arrive_and_drop.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <barrier>
diff --git a/libcxx/test/std/thread/thread.barrier/arrive_and_wait.pass.cpp b/libcxx/test/std/thread/thread.barrier/arrive_and_wait.pass.cpp
index 2aee8624ae3d5..8c45ba9278f28 100644
--- a/libcxx/test/std/thread/thread.barrier/arrive_and_wait.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/arrive_and_wait.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <barrier>
diff --git a/libcxx/test/std/thread/thread.barrier/completion.pass.cpp b/libcxx/test/std/thread/thread.barrier/completion.pass.cpp
index 7354dbe6ffe8a..633a0c8bf2366 100644
--- a/libcxx/test/std/thread/thread.barrier/completion.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/completion.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <barrier>
diff --git a/libcxx/test/std/thread/thread.barrier/ctor.compile.pass.cpp b/libcxx/test/std/thread/thread.barrier/ctor.compile.pass.cpp
index d47127a18613b..fe7068d2a574c 100644
--- a/libcxx/test/std/thread/thread.barrier/ctor.compile.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/ctor.compile.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <barrier>
 
 // explicit barrier(ptrdiff_t __count, _CompletionF __completion = _CompletionF());
diff --git a/libcxx/test/std/thread/thread.barrier/max.pass.cpp b/libcxx/test/std/thread/thread.barrier/max.pass.cpp
index ec03c5c87a09c..b09a02e1bdef4 100644
--- a/libcxx/test/std/thread/thread.barrier/max.pass.cpp
+++ b/libcxx/test/std/thread/thread.barrier/max.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <barrier>
 
 #include <barrier>
diff --git a/libcxx/test/std/thread/thread.latch/arrive_and_wait.pass.cpp b/libcxx/test/std/thread/thread.latch/arrive_and_wait.pass.cpp
index ddc06d2038cc8..8ca4f37b73b95 100644
--- a/libcxx/test/std/thread/thread.latch/arrive_and_wait.pass.cpp
+++ b/libcxx/test/std/thread/thread.latch/arrive_and_wait.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <latch>
diff --git a/libcxx/test/std/thread/thread.latch/count_down.pass.cpp b/libcxx/test/std/thread/thread.latch/count_down.pass.cpp
index 1503c09509a6c..eb524abd24b98 100644
--- a/libcxx/test/std/thread/thread.latch/count_down.pass.cpp
+++ b/libcxx/test/std/thread/thread.latch/count_down.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <latch>
diff --git a/libcxx/test/std/thread/thread.latch/ctor.pass.cpp b/libcxx/test/std/thread/thread.latch/ctor.pass.cpp
index 1983f6409cb5a..bca4561bd2f74 100644
--- a/libcxx/test/std/thread/thread.latch/ctor.pass.cpp
+++ b/libcxx/test/std/thread/thread.latch/ctor.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <latch>
 
 // inline constexpr explicit latch(ptrdiff_t __expected);
diff --git a/libcxx/test/std/thread/thread.latch/max.pass.cpp b/libcxx/test/std/thread/thread.latch/max.pass.cpp
index 8b9176c8cac57..bcf353ed9712e 100644
--- a/libcxx/test/std/thread/thread.latch/max.pass.cpp
+++ b/libcxx/test/std/thread/thread.latch/max.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <latch>
 
 #include <latch>
diff --git a/libcxx/test/std/thread/thread.latch/try_wait.pass.cpp b/libcxx/test/std/thread/thread.latch/try_wait.pass.cpp
index 70ef2cdf71254..8f354463a8697 100644
--- a/libcxx/test/std/thread/thread.latch/try_wait.pass.cpp
+++ b/libcxx/test/std/thread/thread.latch/try_wait.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <latch>
diff --git a/libcxx/test/std/thread/thread.semaphore/acquire.pass.cpp b/libcxx/test/std/thread/thread.semaphore/acquire.pass.cpp
index 3f6e3107e8bce..22eed736c6b75 100644
--- a/libcxx/test/std/thread/thread.semaphore/acquire.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/acquire.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <semaphore>
diff --git a/libcxx/test/std/thread/thread.semaphore/binary.pass.cpp b/libcxx/test/std/thread/thread.semaphore/binary.pass.cpp
index 111a650b5ea39..c01c78506587c 100644
--- a/libcxx/test/std/thread/thread.semaphore/binary.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/binary.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <semaphore>
diff --git a/libcxx/test/std/thread/thread.semaphore/ctor.compile.pass.cpp b/libcxx/test/std/thread/thread.semaphore/ctor.compile.pass.cpp
index 28ccc0124d489..dcc298ce11ce8 100644
--- a/libcxx/test/std/thread/thread.semaphore/ctor.compile.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/ctor.compile.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <semaphore>
 
 // constexpr explicit counting_semaphore(ptrdiff_t desired);
diff --git a/libcxx/test/std/thread/thread.semaphore/max.pass.cpp b/libcxx/test/std/thread/thread.semaphore/max.pass.cpp
index ca7ad0c92e60e..6f3ed5e345e0b 100644
--- a/libcxx/test/std/thread/thread.semaphore/max.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/max.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // <semaphore>
 
 #include <semaphore>
diff --git a/libcxx/test/std/thread/thread.semaphore/release.pass.cpp b/libcxx/test/std/thread/thread.semaphore/release.pass.cpp
index bf3dd7f7d814f..3c4d179e50433 100644
--- a/libcxx/test/std/thread/thread.semaphore/release.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/release.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <semaphore>
diff --git a/libcxx/test/std/thread/thread.semaphore/timed.pass.cpp b/libcxx/test/std/thread/thread.semaphore/timed.pass.cpp
index 9fa01fc035904..77f15ece221d4 100644
--- a/libcxx/test/std/thread/thread.semaphore/timed.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/timed.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <semaphore>
diff --git a/libcxx/test/std/thread/thread.semaphore/try_acquire.pass.cpp b/libcxx/test/std/thread/thread.semaphore/try_acquire.pass.cpp
index 0d0f7792592fb..ec159daf87a3f 100644
--- a/libcxx/test/std/thread/thread.semaphore/try_acquire.pass.cpp
+++ b/libcxx/test/std/thread/thread.semaphore/try_acquire.pass.cpp
@@ -9,6 +9,9 @@
 // UNSUPPORTED: no-threads
 // UNSUPPORTED: c++03, c++11
 
+// Until we drop support for the synchronization library in C++11/14/17
+// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_DISABLE_DEPRECATION_WARNINGS
+
 // XFAIL: availability-synchronization_library-missing
 
 // <semaphore>

From bd28889732e14ac6baca686c3ec99a82fc9cd89d Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Tue, 16 Apr 2024 07:54:51 -0700
Subject: [PATCH 603/886] [RISCV] Add coverage for strength reduction of mul
 2^N +/- 3/5/9

---
 llvm/test/CodeGen/RISCV/rv64zba.ll | 60 ++++++++++++++++++++++++++++++
 1 file changed, 60 insertions(+)

diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index 0d1d4838c6113..a84b9e5e7962f 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -567,6 +567,66 @@ define i64 @mul96(i64 %a) {
   ret i64 %c
 }
 
+define i64 @mul119(i64 %a) {
+; CHECK-LABEL: mul119:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 119
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 119
+  ret i64 %c
+}
+
+define i64 @mul123(i64 %a) {
+; CHECK-LABEL: mul123:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 123
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 123
+  ret i64 %c
+}
+
+define i64 @mul125(i64 %a) {
+; CHECK-LABEL: mul125:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 125
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 125
+  ret i64 %c
+}
+
+define i64 @mul131(i64 %a) {
+; CHECK-LABEL: mul131:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 131
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 131
+  ret i64 %c
+}
+
+define i64 @mul133(i64 %a) {
+; CHECK-LABEL: mul133:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 133
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 133
+  ret i64 %c
+}
+
+define i64 @mul137(i64 %a) {
+; CHECK-LABEL: mul137:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 137
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 137
+  ret i64 %c
+}
+
 define i64 @mul160(i64 %a) {
 ; RV64I-LABEL: mul160:
 ; RV64I:       # %bb.0:

From 1334c034a73b7bf8a7af08be1c33d24a58127c47 Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek <Krzysztof.Parzyszek@amd.com>
Date: Tue, 16 Apr 2024 10:01:34 -0500
Subject: [PATCH 604/886] [flang] Fix test after
 4078afc6d23e25df6baedad61b224ef86a94d42f

This tests requires the OpenMP runtime to be present, but the way that
the lit config detects it fails when "openmp" is added to RUNTIMES
instead of PROJECTS. This caused the tests to be skipped as unsupported
in local and upstream tests.

The actual bug was a missing word in the message, and putting the check
at the wrong line.
---
 flang/test/Semantics/OpenMP/clause-validity01.f90 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/flang/test/Semantics/OpenMP/clause-validity01.f90 b/flang/test/Semantics/OpenMP/clause-validity01.f90
index 74f154bb0ad67..21b99cb82549a 100644
--- a/flang/test/Semantics/OpenMP/clause-validity01.f90
+++ b/flang/test/Semantics/OpenMP/clause-validity01.f90
@@ -342,8 +342,8 @@
   a = 1.0
   !ERROR: COPYPRIVATE clause is not allowed on the END WORKSHARE directive
   !$omp end workshare nowait copyprivate(a)
+  !ERROR: NOWAIT clause is not allowed on the OMP WORKSHARE directive, use it on OMP END WORKSHARE directive 
   !$omp workshare nowait
-  !ERROR: NOWAIT clause is not allowed on the WORKSHARE directive, use it on OMP END WORKSHARE directive
   !$omp end workshare
   !$omp end parallel
 

From bf1ad1d267b1f911cb9846403d2c3d3250a40870 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 16 Apr 2024 16:01:57 +0100
Subject: [PATCH 605/886] [VectorCombine][X86] Add initial
 shuffle-of-shuffles.ll test cover for #88743

---
 .../VectorCombine/X86/shuffle-of-shuffles.ll  | 51 +++++++++++++++++++
 1 file changed, 51 insertions(+)
 create mode 100644 llvm/test/Transforms/VectorCombine/X86/shuffle-of-shuffles.ll

diff --git a/llvm/test/Transforms/VectorCombine/X86/shuffle-of-shuffles.ll b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-shuffles.ll
new file mode 100644
index 0000000000000..b5b5bb997c6c7
--- /dev/null
+++ b/llvm/test/Transforms/VectorCombine/X86/shuffle-of-shuffles.ll
@@ -0,0 +1,51 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=sse2 | FileCheck %s
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx2 | FileCheck %s
+
+; TODO: fold to identity
+
+define <8 x i32> @concat_extract_subvectors(<8 x i32> %x) {
+; CHECK-LABEL: define <8 x i32> @concat_extract_subvectors(
+; CHECK-SAME: <8 x i32> [[X:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[LO:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; CHECK-NEXT:    [[HI:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    [[CONCAT:%.*]] = shufflevector <4 x i32> [[LO]], <4 x i32> [[HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x i32> [[CONCAT]]
+;
+  %lo = shufflevector <8 x i32> %x, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %hi = shufflevector <8 x i32> %x, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %concat = shufflevector <4 x i32> %lo, <4 x i32> %hi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i32> %concat
+}
+
+; negative test - shuffle contains undef
+
+define <8 x i32> @concat_extract_subvectors_undef(<8 x i32> %x) {
+; CHECK-LABEL: define <8 x i32> @concat_extract_subvectors_undef(
+; CHECK-SAME: <8 x i32> [[X:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[LO:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 8>
+; CHECK-NEXT:    [[HI:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 8>
+; CHECK-NEXT:    [[CONCAT:%.*]] = shufflevector <4 x i32> [[LO]], <4 x i32> [[HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x i32> [[CONCAT]]
+;
+  %lo = shufflevector <8 x i32> %x, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 8>
+  %hi = shufflevector <8 x i32> %x, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 8>
+  %concat = shufflevector <4 x i32> %lo, <4 x i32> %hi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i32> %concat
+}
+
+; negative test - shuffle contains poision
+
+define <8 x i32> @concat_extract_subvectors_poison(<8 x i32> %x) {
+; CHECK-LABEL: define <8 x i32> @concat_extract_subvectors_poison(
+; CHECK-SAME: <8 x i32> [[X:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[LO:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 8>
+; CHECK-NEXT:    [[HI:%.*]] = shufflevector <8 x i32> [[X]], <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 8>
+; CHECK-NEXT:    [[CONCAT:%.*]] = shufflevector <4 x i32> [[LO]], <4 x i32> [[HI]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    ret <8 x i32> [[CONCAT]]
+;
+  %lo = shufflevector <8 x i32> %x, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 8>
+  %hi = shufflevector <8 x i32> %x, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 8>
+  %concat = shufflevector <4 x i32> %lo, <4 x i32> %hi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x i32> %concat
+}

From f8e2ec13a8c6d33cb7b4f37869b4429ddcf43f01 Mon Sep 17 00:00:00 2001
From: Adrian Prantl <aprantl@apple.com>
Date: Tue, 16 Apr 2024 08:29:09 -0700
Subject: [PATCH 606/886] Revert "Add asan tests for libsanitizers. (#88349)"

This reverts commit 82f479ba315a417b6cd01a8c2efdc15c26689f2e due to bot breakage.
---
 lldb/test/API/functionalities/asan/Makefile   |  6 +-
 .../functionalities/asan/TestMemoryHistory.py | 73 +------------------
 .../functionalities/asan/TestReportData.py    | 20 +----
 .../API/functionalities/libsanitizers/util.py |  3 -
 4 files changed, 5 insertions(+), 97 deletions(-)
 delete mode 100644 lldb/test/API/functionalities/libsanitizers/util.py

diff --git a/lldb/test/API/functionalities/asan/Makefile b/lldb/test/API/functionalities/asan/Makefile
index d66696fed7078..4913a18d8cc6f 100644
--- a/lldb/test/API/functionalities/asan/Makefile
+++ b/lldb/test/API/functionalities/asan/Makefile
@@ -1,8 +1,4 @@
 C_SOURCES := main.c
-asan: CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
-asan: all
-
-libsanitizers: CFLAGS_EXTRAS := -fsanitize=address -fsanitize-stable-abi -g -gcolumn-info
-libsanitizers: all
+CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
 
 include Makefile.rules
diff --git a/lldb/test/API/functionalities/asan/TestMemoryHistory.py b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
index ee7939203ead1..00162ae8822c7 100644
--- a/lldb/test/API/functionalities/asan/TestMemoryHistory.py
+++ b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
@@ -9,21 +9,15 @@
 from lldbsuite.test import lldbplatform
 from lldbsuite.test import lldbutil
 
-from functionalities.libsanitizers.util import no_libsanitizers
 
 class AsanTestCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
     @expectedFailureNetBSD
     @skipUnlessAddressSanitizer
     def test(self):
-        self.build(make_targets=["asan"])
+        self.build()
         self.asan_tests()
 
-    @skipIf(oslist=no_match(["macosx"]))
-    def test_libsanitizers_asan(self):
-        self.build(make_targets=["libsanitizers"])
-        self.libsanitizer_tests()
-
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -32,71 +26,6 @@ def setUp(self):
         self.line_free = line_number("main.c", "// free line")
         self.line_breakpoint = line_number("main.c", "// break line")
 
-    # Test line numbers: rdar://126237493
-    def libsanitizer_tests(self):
-        target = self.createTestTarget()
-
-        if no_libsanitizers(self):
-            self.skipTest("libsanitizers not found")
-
-        self.runCmd(
-            "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
-        )
-
-        self.runCmd("run")
-
-        # In libsanitizers, memory history is not supported until a report has been generated
-        self.expect(
-            "thread list",
-            "Process should be stopped due to ASan report",
-            substrs=["stopped", "stop reason = Use of deallocated memory"],
-        )
-
-        # test the 'memory history' command
-        self.expect(
-            "memory history 'pointer'",
-            substrs=[
-                "Memory deallocated by Thread",
-                "a.out`f2",
-                "main.c",
-                "Memory allocated by Thread",
-                "a.out`f1",
-                "main.c",
-            ],
-        )
-
-        # do the same using SB API
-        process = self.dbg.GetSelectedTarget().process
-        val = (
-            process.GetSelectedThread().GetSelectedFrame().EvaluateExpression("pointer")
-        )
-        addr = val.GetValueAsUnsigned()
-        threads = process.GetHistoryThreads(addr)
-        self.assertEqual(threads.GetSize(), 2)
-
-        history_thread = threads.GetThreadAtIndex(0)
-        self.assertTrue(history_thread.num_frames >= 2)
-        self.assertEqual(
-            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
-            "main.c",
-        )
-
-        history_thread = threads.GetThreadAtIndex(1)
-        self.assertTrue(history_thread.num_frames >= 2)
-        self.assertEqual(
-            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
-            "main.c",
-        )
-
-        # let's free the container (SBThreadCollection) and see if the
-        # SBThreads still live
-        threads = None
-        self.assertTrue(history_thread.num_frames >= 2)
-        self.assertEqual(
-            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
-            "main.c",
-        )
-
     def asan_tests(self):
         target = self.createTestTarget()
 
diff --git a/lldb/test/API/functionalities/asan/TestReportData.py b/lldb/test/API/functionalities/asan/TestReportData.py
index de0c1206a57ad..543c5fe66a208 100644
--- a/lldb/test/API/functionalities/asan/TestReportData.py
+++ b/lldb/test/API/functionalities/asan/TestReportData.py
@@ -9,7 +9,6 @@
 from lldbsuite.test.lldbtest import *
 from lldbsuite.test import lldbutil
 
-from functionalities.libsanitizers.util import no_libsanitizers
 
 class AsanTestReportDataCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
@@ -17,14 +16,9 @@ class AsanTestReportDataCase(TestBase):
     @skipUnlessAddressSanitizer
     @skipIf(archs=["i386"], bugnumber="llvm.org/PR36710")
     def test(self):
-        self.build(make_targets=["asan"])
+        self.build()
         self.asan_tests()
 
-    @skipIf(oslist=no_match(["macosx"]))
-    def test_libsanitizers_asan(self):
-        self.build(make_targets=["libsanitizers"])
-        self.asan_tests(libsanitizers=True)
-
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -35,18 +29,10 @@ def setUp(self):
         self.line_crash = line_number("main.c", "// BOOM line")
         self.col_crash = 16
 
-    def asan_tests(self, libsanitizers=False):
+    def asan_tests(self):
         target = self.createTestTarget()
 
-        if libsanitizers and no_libsanitizers(self):
-            self.skipTest("libsanitizers not found")
-
-        if libsanitizers:
-            self.runCmd(
-                "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
-            )
-        else:
-            self.registerSanitizerLibrariesWithTarget(target)
+        self.registerSanitizerLibrariesWithTarget(target)
 
         self.runCmd("run")
 
diff --git a/lldb/test/API/functionalities/libsanitizers/util.py b/lldb/test/API/functionalities/libsanitizers/util.py
deleted file mode 100644
index ad68541aba8d0..0000000000000
--- a/lldb/test/API/functionalities/libsanitizers/util.py
+++ /dev/null
@@ -1,3 +0,0 @@
-def no_libsanitizers(testbase):
-    testbase.runCmd("image list libsystem_sanitizers.dylib", check=False)
-    return not "libsystem_sanitizers.dylib" in testbase.res.GetOutput()

From 8cee94e989b5bf6fb6455087d48eb6c6e0e23c54 Mon Sep 17 00:00:00 2001
From: Harald van Dijk <harald.vandijk@codeplay.com>
Date: Tue, 16 Apr 2024 16:32:57 +0100
Subject: [PATCH 607/886] [RISCV] Fix obvious copy paste error.

CASE_VFMA_OPCODE_VV and CASE_VFMA_CHANGE_OPCODE_VV need to match up if we are
are to avoid "Unexpected opcode" errors, but in CASE_VFMA_CHANGE_OPCODE_VV,
CASE_VFMA_CHANGE_OPCODE_LMULS_MF2 had mistakenly been used instead of
CASE_VFMA_CHANGE_OPCODE_LMULS_MF4.
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 668062c8d33f6..14b5cbea71722 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -3016,7 +3016,7 @@ bool RISCVInstrInfo::findCommutedOpIndices(const MachineInstr &MI,
   CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, TYPE, SEW)
 
 #define CASE_VFMA_CHANGE_OPCODE_VV(OLDOP, NEWOP)                               \
-  CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E16)                     \
+  CASE_VFMA_CHANGE_OPCODE_LMULS_MF4(OLDOP, NEWOP, VV, E16)                     \
   CASE_VFMA_CHANGE_OPCODE_LMULS_MF2(OLDOP, NEWOP, VV, E32)                     \
   CASE_VFMA_CHANGE_OPCODE_LMULS_M1(OLDOP, NEWOP, VV, E64)
 

From 51b42b762112f2e77d032efd16fa0d9d31cde494 Mon Sep 17 00:00:00 2001
From: Mehdi Amini <joker.eph@gmail.com>
Date: Tue, 16 Apr 2024 17:45:02 +0200
Subject: [PATCH 608/886] Revert "[Sema] Mark alias/ifunc targets used and
 consider mangled names" (#88919)

Reverts llvm/llvm-project#87130

Bot is broken with clang crash:
https://lab.llvm.org/buildbot/#/builders/272/builds/14063/steps/6/logs/stdio
---
 clang/lib/Sema/CMakeLists.txt                 |  1 -
 clang/lib/Sema/SemaDeclAttr.cpp               | 44 +++++--------------
 clang/test/AST/ast-dump-attr-json.cpp         |  1 -
 clang/test/Sema/alias-unused-win.cpp          |  2 +-
 clang/test/Sema/alias-unused.cpp              | 16 +++----
 .../llvm-project-overlay/clang/BUILD.bazel    |  1 -
 6 files changed, 19 insertions(+), 46 deletions(-)

diff --git a/clang/lib/Sema/CMakeLists.txt b/clang/lib/Sema/CMakeLists.txt
index a96439df66422..ab3b813a9ccd9 100644
--- a/clang/lib/Sema/CMakeLists.txt
+++ b/clang/lib/Sema/CMakeLists.txt
@@ -1,6 +1,5 @@
 set(LLVM_LINK_COMPONENTS
   Core
-  Demangle
   FrontendHLSL
   FrontendOpenMP
   MC
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index d26f130b5774c..b7b1fbc625a15 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -45,7 +45,6 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
-#include "llvm/Demangle/Demangle.h"
 #include "llvm/IR/Assumptions.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Support/Error.h"
@@ -1984,36 +1983,6 @@ static void handleWeakRefAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   D->addAttr(::new (S.Context) WeakRefAttr(S.Context, AL));
 }
 
-// Mark alias/ifunc target as used. Due to name mangling, we look up the
-// demangled name ignoring parameters (not supported by microsoftDemangle
-// https://github.com/llvm/llvm-project/issues/88825). This should handle the
-// majority of use cases while leaving namespace scope names unmarked.
-static void markUsedForAliasOrIfunc(Sema &S, Decl *D, const ParsedAttr &AL,
-                                    StringRef Str) {
-  std::unique_ptr<char, llvm::FreeDeleter> Demangled;
-  if (S.getASTContext().getCXXABIKind() != TargetCXXABI::Microsoft)
-    Demangled.reset(llvm::itaniumDemangle(Str, /*ParseParams=*/false));
-  std::unique_ptr<MangleContext> MC(S.Context.createMangleContext());
-  SmallString<256> Name;
-
-  const DeclarationNameInfo Target(
-      &S.Context.Idents.get(Demangled ? Demangled.get() : Str), AL.getLoc());
-  LookupResult LR(S, Target, Sema::LookupOrdinaryName);
-  if (S.LookupName(LR, S.TUScope)) {
-    for (NamedDecl *ND : LR) {
-      if (MC->shouldMangleDeclName(ND)) {
-        llvm::raw_svector_ostream Out(Name);
-        Name.clear();
-        MC->mangleName(GlobalDecl(ND), Out);
-      } else {
-        Name = ND->getIdentifier()->getName();
-      }
-      if (Name == Str)
-        ND->markUsed(S.Context);
-    }
-  }
-}
-
 static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   StringRef Str;
   if (!S.checkStringLiteralArgumentAttr(AL, 0, Str))
@@ -2026,7 +1995,6 @@ static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     return;
   }
 
-  markUsedForAliasOrIfunc(S, D, AL, Str);
   D->addAttr(::new (S.Context) IFuncAttr(S.Context, AL, Str));
 }
 
@@ -2061,7 +2029,17 @@ static void handleAliasAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     }
   }
 
-  markUsedForAliasOrIfunc(S, D, AL, Str);
+  // Mark target used to prevent unneeded-internal-declaration warnings.
+  if (!S.LangOpts.CPlusPlus) {
+    // FIXME: demangle Str for C++, as the attribute refers to the mangled
+    // linkage name, not the pre-mangled identifier.
+    const DeclarationNameInfo target(&S.Context.Idents.get(Str), AL.getLoc());
+    LookupResult LR(S, target, Sema::LookupOrdinaryName);
+    if (S.LookupQualifiedName(LR, S.getCurLexicalContext()))
+      for (NamedDecl *ND : LR)
+        ND->markUsed(S.Context);
+  }
+
   D->addAttr(::new (S.Context) AliasAttr(S.Context, AL, Str));
 }
 
diff --git a/clang/test/AST/ast-dump-attr-json.cpp b/clang/test/AST/ast-dump-attr-json.cpp
index 883e584bfedf0..051c2956abfdf 100644
--- a/clang/test/AST/ast-dump-attr-json.cpp
+++ b/clang/test/AST/ast-dump-attr-json.cpp
@@ -46,7 +46,6 @@ __thread __attribute__ ((tls_model ("local-exec"))) int tls_model_var;
 // CHECK-NEXT:    "tokLen": 11
 // CHECK-NEXT:   }
 // CHECK-NEXT:  },
-// CHECK-NEXT:  "isUsed": true,
 // CHECK-NEXT:  "name": "global_decl",
 // CHECK-NEXT:  "mangledName": "global_decl",
 // CHECK-NEXT:  "type": {
diff --git a/clang/test/Sema/alias-unused-win.cpp b/clang/test/Sema/alias-unused-win.cpp
index 97d57a3bbd1e3..47c96d4117517 100644
--- a/clang/test/Sema/alias-unused-win.cpp
+++ b/clang/test/Sema/alias-unused-win.cpp
@@ -7,7 +7,7 @@ extern "C" {
 static int f(void) { return 42; } // cxx-warning{{unused function 'f'}}
 int g(void) __attribute__((alias("f")));
 
-static int foo [] = { 42, 0xDEAD };
+static int foo [] = { 42, 0xDEAD }; // cxx-warning{{variable 'foo' is not needed and will not be emitted}}
 extern typeof(foo) bar __attribute__((unused, alias("foo")));
 
 static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
diff --git a/clang/test/Sema/alias-unused.cpp b/clang/test/Sema/alias-unused.cpp
index c0b541c880e52..dc8e46f072d74 100644
--- a/clang/test/Sema/alias-unused.cpp
+++ b/clang/test/Sema/alias-unused.cpp
@@ -14,26 +14,24 @@ extern typeof(foo) bar __attribute__((unused, alias("foo")));
 /// We report a warning in C++ mode because the internal linkage `resolver` gets
 /// mangled as it does not have a language linkage. GCC does not mangle
 /// `resolver` or report a warning.
-static int (*resolver(void))(void) { return f; } // cxx-warning{{unused function 'resolver'}}
+static int (*resolver(void))(void) { return f; } // expected-warning{{unused function 'resolver'}}
 int ifunc(void) __attribute__((ifunc("resolver")));
 
-static int __attribute__((overloadable)) f0(int x) { return x; }
+static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
 static float __attribute__((overloadable)) f0(float x) { return x; } // expected-warning{{unused function 'f0'}}
 int g0(void) __attribute__((alias("_ZL2f0i")));
 
 #ifdef __cplusplus
-static int f1() { return 42; }
+static int f1() { return 42; } // expected-warning{{unused function 'f1'}}
 int g1(void) __attribute__((alias("_ZL2f1v")));
 }
 
-/// We demangle alias/ifunc target and mark all found functions as used.
-
-static int f2(int) { return 42; } // cxx-warning{{unused function 'f2'}}
-static int f2() { return 42; }
+static int f2(int) { return 42; } // expected-warning{{unused function 'f2'}}
+static int f2() { return 42; } // expected-warning{{unused function 'f2'}}
 int g2() __attribute__((alias("_ZL2f2v")));
 
-static int (*resolver1())() { return f; } // cxx-warning{{unused function 'resolver1'}}
-static int (*resolver1(int))() { return f; }
+static int (*resolver1())() { return f; } // expected-warning{{unused function 'resolver1'}}
+static int (*resolver1(int))() { return f; } // expected-warning{{unused function 'resolver1'}}
 int ifunc1() __attribute__((ifunc("_ZL9resolver1i")));
 
 /// TODO: We should report "unused function" for f3(int).
diff --git a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
index 725ac6bb38120..c2f77e3abca0e 100644
--- a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
@@ -1136,7 +1136,6 @@ cc_library(
         "//llvm:AllTargetsAsmParsers",
         "//llvm:AllTargetsCodeGens",
         "//llvm:Core",
-        "//llvm:Demangle",
         "//llvm:FrontendHLSL",
         "//llvm:FrontendOpenMP",
         "//llvm:MC",

From 9d111286322ec99b32467eef3aeec6b588c49f18 Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 16:48:59 +0100
Subject: [PATCH 609/886] [libclc] Improve clarity of CMake foreach. NFC.

Should be a bit easier to read.
---
 libclc/CMakeLists.txt | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index ed2764847e709..f605c3bbbe9dc 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -50,7 +50,7 @@ if( LIBCLC_STANDALONE_BUILD OR CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DI
   endif()
 
   # Import required tools as targets
-  foreach( tool clang llvm-as llvm-link opt )
+  foreach( tool IN ITEMS clang llvm-as llvm-link opt )
     find_program( LLVM_TOOL_${tool} ${tool} PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
     add_executable( libclc::${tool} IMPORTED GLOBAL )
     set_target_properties( libclc::${tool} PROPERTIES IMPORTED_LOCATION ${LLVM_TOOL_${tool}} )
@@ -68,7 +68,7 @@ else()
     message(FATAL_ERROR "Clang is not enabled, but is required to build libclc in-tree")
   endif()
 
-  foreach( tool clang llvm-as llvm-link opt )
+  foreach( tool IN ITEMS clang llvm-as llvm-link opt )
     add_executable(libclc::${tool} ALIAS ${tool})
   endforeach()
 endif()
@@ -181,7 +181,7 @@ install( FILES ${CMAKE_CURRENT_BINARY_DIR}/libclc.pc DESTINATION "${CMAKE_INSTAL
 install( DIRECTORY generic/include/clc DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}" )
 
 if( ENABLE_RUNTIME_SUBNORMAL )
-  foreach( file subnormal_use_default subnormal_disable )
+  foreach( file IN ITEMS subnormal_use_default subnormal_disable )
     link_bc(
        TARGET ${file}
        INPUTS ${PROJECT_SOURCE_DIR}/generic/lib/${file}.ll

From 2704ebaf2885a16155ab7144f8dd0dd459d77089 Mon Sep 17 00:00:00 2001
From: Slava Zakharin <szakharin@nvidia.com>
Date: Tue, 16 Apr 2024 08:53:12 -0700
Subject: [PATCH 610/886] [flang][runtime] Create CUDA PTX OBJECT library
 target for F18 runtime CUDA build. (#88821)

This is to experiment with distributing F18 runtime CUDA library
in the form of a pure PTX library. The change is under
FLANG_EXPERIMENTAL_CUDA_RUNTIME CMake control.
---
 flang/cmake/modules/AddFlangOffloadRuntime.cmake | 6 +++++-
 flang/lib/Decimal/CMakeLists.txt                 | 2 +-
 flang/runtime/CMakeLists.txt                     | 2 +-
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/flang/cmake/modules/AddFlangOffloadRuntime.cmake b/flang/cmake/modules/AddFlangOffloadRuntime.cmake
index 6fb6213e90fc4..e34d3851187ac 100644
--- a/flang/cmake/modules/AddFlangOffloadRuntime.cmake
+++ b/flang/cmake/modules/AddFlangOffloadRuntime.cmake
@@ -10,7 +10,7 @@ set(FLANG_EXPERIMENTAL_OMP_OFFLOAD_BUILD "off" CACHE STRING
 set(FLANG_OMP_DEVICE_ARCHITECTURES "all" CACHE STRING
   "List of OpenMP device architectures to be used to compile the Fortran runtime (e.g. 'gfx1103;sm_90')")
 
-macro(enable_cuda_compilation files)
+macro(enable_cuda_compilation name files)
   if (FLANG_EXPERIMENTAL_CUDA_RUNTIME)
     if (BUILD_SHARED_LIBS)
       message(FATAL_ERROR
@@ -52,6 +52,10 @@ macro(enable_cuda_compilation files)
       include_directories(AFTER ${FLANG_LIBCUDACXX_PATH}/include)
       add_compile_definitions(RT_USE_LIBCUDACXX=1)
     endif()
+
+    # Add an OBJECT library consisting of CUDA PTX.
+    llvm_add_library(${name}PTX OBJECT PARTIAL_SOURCES_INTENDED ${files})
+    set_property(TARGET obj.${name}PTX PROPERTY CUDA_PTX_COMPILATION ON)
   endif()
 endmacro()
 
diff --git a/flang/lib/Decimal/CMakeLists.txt b/flang/lib/Decimal/CMakeLists.txt
index 3d562b8e3ce1e..880b190f1c581 100644
--- a/flang/lib/Decimal/CMakeLists.txt
+++ b/flang/lib/Decimal/CMakeLists.txt
@@ -55,7 +55,7 @@ set(sources
 )
 
 include(AddFlangOffloadRuntime)
-enable_cuda_compilation("${sources}")
+enable_cuda_compilation(FortranDecimal "${sources}")
 enable_omp_offload_compilation("${sources}")
 
 add_flang_library(FortranDecimal INSTALL_WITH_TOOLCHAIN ${sources})
diff --git a/flang/runtime/CMakeLists.txt b/flang/runtime/CMakeLists.txt
index 2a65a22ab674c..bdd0e07bbfd4d 100644
--- a/flang/runtime/CMakeLists.txt
+++ b/flang/runtime/CMakeLists.txt
@@ -224,7 +224,7 @@ set(supported_files
   utf.cpp
   )
 
-enable_cuda_compilation("${supported_files}")
+enable_cuda_compilation(FortranRuntime "${supported_files}")
 enable_omp_offload_compilation("${supported_files}")
 
 if (NOT TARGET FortranFloat128Math)

From a79783d7ad00c4c10c30f637f4bf13551e47f3dd Mon Sep 17 00:00:00 2001
From: Robin Caloudis <robin.caloudis@gmx.de>
Date: Tue, 16 Apr 2024 17:58:11 +0200
Subject: [PATCH 611/886] [libc][fenv] Use proxy header (#88787)

Include types `fexcept_t` and `fenv_t ` from corresponding proxy
headers, as they are available since
https://github.com/llvm/llvm-project/pull/88467.
---
 libc/src/fenv/fegetexceptflag.h | 2 +-
 libc/src/fenv/fesetexceptflag.h | 2 +-
 libc/src/fenv/feupdateenv.cpp   | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/libc/src/fenv/fegetexceptflag.h b/libc/src/fenv/fegetexceptflag.h
index ad72161e536f8..fcb9598658d43 100644
--- a/libc/src/fenv/fegetexceptflag.h
+++ b/libc/src/fenv/fegetexceptflag.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FEGETEXCEPTFLAG_H
 #define LLVM_LIBC_SRC_FENV_FEGETEXCEPTFLAG_H
 
-#include <fenv.h>
+#include "hdr/types/fexcept_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/fesetexceptflag.h b/libc/src/fenv/fesetexceptflag.h
index 15e62eda1b840..a018358dc9dfc 100644
--- a/libc/src/fenv/fesetexceptflag.h
+++ b/libc/src/fenv/fesetexceptflag.h
@@ -9,7 +9,7 @@
 #ifndef LLVM_LIBC_SRC_FENV_FESETEXCEPTFLAG_H
 #define LLVM_LIBC_SRC_FENV_FESETEXCEPTFLAG_H
 
-#include <fenv.h>
+#include "hdr/types/fexcept_t.h"
 
 namespace LIBC_NAMESPACE {
 
diff --git a/libc/src/fenv/feupdateenv.cpp b/libc/src/fenv/feupdateenv.cpp
index 7e81b9476da91..0664863538155 100644
--- a/libc/src/fenv/feupdateenv.cpp
+++ b/libc/src/fenv/feupdateenv.cpp
@@ -10,7 +10,7 @@
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "src/__support/common.h"
 
-#include <fenv.h>
+#include "hdr/types/fenv_t.h"
 
 namespace LIBC_NAMESPACE {
 

From 38895e6578f7728cfb3d41d0880a0e9b358d9afd Mon Sep 17 00:00:00 2001
From: Robin Caloudis <robin.caloudis@gmx.de>
Date: Tue, 16 Apr 2024 17:58:56 +0200
Subject: [PATCH 612/886] [libc][fenv] Remove unnecessary dependencies (#88788)

Remove the fenv macro dependency from the CMake files as the underlying targets
do not make use of it. Note that we do not have to worry about
[corresponding Bazel targets](https://github.com/llvm/llvm-project/blob/main/utils/bazel/llvm-project-overlay/libc/BUILD.bazel#L1138-L1288),
as they look good.
---
 libc/src/fenv/CMakeLists.txt | 10 ----------
 1 file changed, 10 deletions(-)

diff --git a/libc/src/fenv/CMakeLists.txt b/libc/src/fenv/CMakeLists.txt
index a28a7ca4c2d82..17e9947412062 100644
--- a/libc/src/fenv/CMakeLists.txt
+++ b/libc/src/fenv/CMakeLists.txt
@@ -17,7 +17,6 @@ add_entrypoint_object(
   HDRS
     fesetround.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -30,7 +29,6 @@ add_entrypoint_object(
   HDRS
     feclearexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -43,7 +41,6 @@ add_entrypoint_object(
   HDRS
     feraiseexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -56,7 +53,6 @@ add_entrypoint_object(
   HDRS
     fetestexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -69,7 +65,6 @@ add_entrypoint_object(
   HDRS
     fegetenv.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
@@ -83,7 +78,6 @@ add_entrypoint_object(
   HDRS
     fesetenv.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.hdr.types.fenv_t
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
@@ -111,7 +105,6 @@ add_entrypoint_object(
   HDRS
     fesetexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -166,7 +159,6 @@ add_entrypoint_object(
   HDRS
     feenableexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -179,7 +171,6 @@ add_entrypoint_object(
   HDRS
     fedisableexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2
@@ -192,7 +183,6 @@ add_entrypoint_object(
   HDRS
     fegetexcept.h
   DEPENDS
-    libc.hdr.fenv_macros
     libc.src.__support.FPUtil.fenv_impl
   COMPILE_OPTIONS
     -O2

From 1c6b0f779f66494cb597884c1a52e377bde4bc54 Mon Sep 17 00:00:00 2001
From: Stephen Tozer <stephen.tozer@sony.com>
Date: Tue, 16 Apr 2024 17:07:46 +0100
Subject: [PATCH 613/886] [RemoveDI] Add support for debug records to debugify
 (#87383)

This patch changes debugify to support debug variable records, and
subsequently to no longer convert modules automatically to intrinsics
when entering debugify.
---
 llvm/lib/CodeGen/MachineDebugify.cpp          |  18 ++-
 llvm/lib/Transforms/Utils/Debugify.cpp        | 111 +++++++-----------
 .../check-line-and-variables-x.mir            |   1 +
 .../MIRDebugify/check-line-and-variables.ll   |   1 +
 .../MIRDebugify/check-line-and-variables.mir  |   2 +
 .../MIRDebugify/locations-and-values.mir      |   4 +
 .../MIRDebugify/multifunction-module.mir      |   3 +-
 .../DebugInfo/debugify-bogus-dbg-value.ll     |   1 +
 llvm/test/DebugInfo/debugify-each.ll          |  34 ++++++
 llvm/test/DebugInfo/debugify-export.ll        |   3 +
 llvm/test/DebugInfo/debugify-ignore-phi.ll    |   1 +
 .../debugify-original-no-dbg-info.ll          |   1 +
 .../debugify-report-missing-locs-only.ll      |   1 +
 llvm/test/DebugInfo/debugify.ll               |  27 +++++
 llvm/test/DebugInfo/pr37964.ll                |   1 +
 .../test/DebugInfo/salvage-cast-debug-info.ll |   2 +-
 llvm/test/DebugInfo/verify-di-preserve.ll     |   4 +-
 17 files changed, 142 insertions(+), 73 deletions(-)

diff --git a/llvm/lib/CodeGen/MachineDebugify.cpp b/llvm/lib/CodeGen/MachineDebugify.cpp
index c264e199cf472..bffdd51bfbca7 100644
--- a/llvm/lib/CodeGen/MachineDebugify.cpp
+++ b/llvm/lib/CodeGen/MachineDebugify.cpp
@@ -65,6 +65,7 @@ bool applyDebugifyMetadataToMachineFunction(MachineModuleInfo &MMI,
   // all the others.
   Function *DbgValF = M.getFunction("llvm.dbg.value");
   DbgValueInst *EarliestDVI = nullptr;
+  DbgVariableRecord *EarliestDVR = nullptr;
   DenseMap<unsigned, DILocalVariable *> Line2Var;
   DIExpression *Expr = nullptr;
   if (DbgValF) {
@@ -80,6 +81,20 @@ bool applyDebugifyMetadataToMachineFunction(MachineModuleInfo &MMI,
       Expr = DVI->getExpression();
     }
   }
+  for (BasicBlock &BB : F) {
+    for (Instruction &I : BB) {
+      for (DbgVariableRecord &DVR : filterDbgVars(I.getDbgRecordRange())) {
+        if (!DVR.isDbgValue())
+          continue;
+        unsigned Line = DVR.getDebugLoc().getLine();
+        assert(Line != 0 && "debugify should not insert line 0 locations");
+        Line2Var[Line] = DVR.getVariable();
+        if (!EarliestDVR || Line < EarliestDVR->getDebugLoc().getLine())
+          EarliestDVR = &DVR;
+        Expr = DVR.getExpression();
+      }
+    }
+  }
   if (Line2Var.empty())
     return true;
 
@@ -109,7 +124,8 @@ bool applyDebugifyMetadataToMachineFunction(MachineModuleInfo &MMI,
       // Find a suitable local variable for the DBG_VALUE.
       unsigned Line = MI.getDebugLoc().getLine();
       if (!Line2Var.count(Line))
-        Line = EarliestDVI->getDebugLoc().getLine();
+        Line = EarliestDVI ? EarliestDVI->getDebugLoc().getLine()
+                           : EarliestDVR->getDebugLoc().getLine();
       DILocalVariable *LocalVar = Line2Var[Line];
       assert(LocalVar && "No variable for current line?");
       VarSet.insert(LocalVar);
diff --git a/llvm/lib/Transforms/Utils/Debugify.cpp b/llvm/lib/Transforms/Utils/Debugify.cpp
index 200bad22148f0..fcc82eadac36c 100644
--- a/llvm/lib/Transforms/Utils/Debugify.cpp
+++ b/llvm/lib/Transforms/Utils/Debugify.cpp
@@ -87,10 +87,6 @@ bool llvm::applyDebugifyMetadata(
     return false;
   }
 
-  bool NewDebugMode = M.IsNewDbgInfoFormat;
-  if (NewDebugMode)
-    M.convertFromNewDbgValues();
-
   DIBuilder DIB(M);
   LLVMContext &Ctx = M.getContext();
   auto *Int32Ty = Type::getInt32Ty(Ctx);
@@ -214,9 +210,6 @@ bool llvm::applyDebugifyMetadata(
   if (!M.getModuleFlag(DIVersionKey))
     M.addModuleFlag(Module::Warning, DIVersionKey, DEBUG_METADATA_VERSION);
 
-  if (NewDebugMode)
-    M.convertToNewDbgValues();
-
   return true;
 }
 
@@ -311,10 +304,6 @@ bool llvm::collectDebugInfoMetadata(Module &M,
     return false;
   }
 
-  bool NewDebugMode = M.IsNewDbgInfoFormat;
-  if (NewDebugMode)
-    M.convertFromNewDbgValues();
-
   uint64_t FunctionsCnt = DebugInfoBeforePass.DIFunctions.size();
   // Visit each instruction.
   for (Function &F : Functions) {
@@ -349,20 +338,23 @@ bool llvm::collectDebugInfoMetadata(Module &M,
 
         // Cllect dbg.values and dbg.declare.
         if (DebugifyLevel > Level::Locations) {
-          if (auto *DVI = dyn_cast<DbgVariableIntrinsic>(&I)) {
+          auto HandleDbgVariable = [&](auto *DbgVar) {
             if (!SP)
-              continue;
+              return;
             // Skip inlined variables.
-            if (I.getDebugLoc().getInlinedAt())
-              continue;
+            if (DbgVar->getDebugLoc().getInlinedAt())
+              return;
             // Skip undef values.
-            if (DVI->isKillLocation())
-              continue;
+            if (DbgVar->isKillLocation())
+              return;
 
-            auto *Var = DVI->getVariable();
+            auto *Var = DbgVar->getVariable();
             DebugInfoBeforePass.DIVariables[Var]++;
-            continue;
-          }
+          };
+          for (DbgVariableRecord &DVR : filterDbgVars(I.getDbgRecordRange()))
+            HandleDbgVariable(&DVR);
+          if (auto *DVI = dyn_cast<DbgVariableIntrinsic>(&I))
+            HandleDbgVariable(DVI);
         }
 
         // Skip debug instructions other than dbg.value and dbg.declare.
@@ -379,9 +371,6 @@ bool llvm::collectDebugInfoMetadata(Module &M,
     }
   }
 
-  if (NewDebugMode)
-    M.convertToNewDbgValues();
-
   return true;
 }
 
@@ -561,10 +550,6 @@ bool llvm::checkDebugInfoMetadata(Module &M,
     return false;
   }
 
-  bool NewDebugMode = M.IsNewDbgInfoFormat;
-  if (NewDebugMode)
-    M.convertFromNewDbgValues();
-
   // Map the debug info holding DIs after a pass.
   DebugInfoPerPass DebugInfoAfterPass;
 
@@ -599,20 +584,23 @@ bool llvm::checkDebugInfoMetadata(Module &M,
 
         // Collect dbg.values and dbg.declares.
         if (DebugifyLevel > Level::Locations) {
-          if (auto *DVI = dyn_cast<DbgVariableIntrinsic>(&I)) {
+          auto HandleDbgVariable = [&](auto *DbgVar) {
             if (!SP)
-              continue;
+              return;
             // Skip inlined variables.
-            if (I.getDebugLoc().getInlinedAt())
-              continue;
+            if (DbgVar->getDebugLoc().getInlinedAt())
+              return;
             // Skip undef values.
-            if (DVI->isKillLocation())
-              continue;
+            if (DbgVar->isKillLocation())
+              return;
 
-            auto *Var = DVI->getVariable();
+            auto *Var = DbgVar->getVariable();
             DebugInfoAfterPass.DIVariables[Var]++;
-            continue;
-          }
+          };
+          for (DbgVariableRecord &DVR : filterDbgVars(I.getDbgRecordRange()))
+            HandleDbgVariable(&DVR);
+          if (auto *DVI = dyn_cast<DbgVariableIntrinsic>(&I))
+            HandleDbgVariable(DVI);
         }
 
         // Skip debug instructions other than dbg.value and dbg.declare.
@@ -675,16 +663,14 @@ bool llvm::checkDebugInfoMetadata(Module &M,
   // the debugging information from the previous pass.
   DebugInfoBeforePass = DebugInfoAfterPass;
 
-  if (NewDebugMode)
-    M.convertToNewDbgValues();
-
   LLVM_DEBUG(dbgs() << "\n\n");
   return Result;
 }
 
 namespace {
-/// Return true if a mis-sized diagnostic is issued for \p DVI.
-bool diagnoseMisSizedDbgValue(Module &M, DbgValueInst *DVI) {
+/// Return true if a mis-sized diagnostic is issued for \p DbgVal.
+template <typename DbgValTy>
+bool diagnoseMisSizedDbgValue(Module &M, DbgValTy *DbgVal) {
   // The size of a dbg.value's value operand should match the size of the
   // variable it corresponds to.
   //
@@ -693,22 +679,22 @@ bool diagnoseMisSizedDbgValue(Module &M, DbgValueInst *DVI) {
 
   // For now, don't try to interpret anything more complicated than an empty
   // DIExpression. Eventually we should try to handle OP_deref and fragments.
-  if (DVI->getExpression()->getNumElements())
+  if (DbgVal->getExpression()->getNumElements())
     return false;
 
-  Value *V = DVI->getVariableLocationOp(0);
+  Value *V = DbgVal->getVariableLocationOp(0);
   if (!V)
     return false;
 
   Type *Ty = V->getType();
   uint64_t ValueOperandSize = getAllocSizeInBits(M, Ty);
-  std::optional<uint64_t> DbgVarSize = DVI->getFragmentSizeInBits();
+  std::optional<uint64_t> DbgVarSize = DbgVal->getFragmentSizeInBits();
   if (!ValueOperandSize || !DbgVarSize)
     return false;
 
   bool HasBadSize = false;
   if (Ty->isIntegerTy()) {
-    auto Signedness = DVI->getVariable()->getSignedness();
+    auto Signedness = DbgVal->getVariable()->getSignedness();
     if (Signedness && *Signedness == DIBasicType::Signedness::Signed)
       HasBadSize = ValueOperandSize < *DbgVarSize;
   } else {
@@ -718,7 +704,7 @@ bool diagnoseMisSizedDbgValue(Module &M, DbgValueInst *DVI) {
   if (HasBadSize) {
     dbg() << "ERROR: dbg.value operand has size " << ValueOperandSize
           << ", but its variable has size " << *DbgVarSize << ": ";
-    DVI->print(dbg());
+    DbgVal->print(dbg());
     dbg() << "\n";
   }
   return HasBadSize;
@@ -735,10 +721,6 @@ bool checkDebugifyMetadata(Module &M,
     return false;
   }
 
-  bool NewDebugMode = M.IsNewDbgInfoFormat;
-  if (NewDebugMode)
-    M.convertFromNewDbgValues();
-
   auto getDebugifyOperand = [&](unsigned Idx) -> unsigned {
     return mdconst::extract<ConstantInt>(NMD->getOperand(Idx)->getOperand(0))
         ->getZExtValue();
@@ -780,18 +762,23 @@ bool checkDebugifyMetadata(Module &M,
     }
 
     // Find missing variables and mis-sized debug values.
-    for (Instruction &I : instructions(F)) {
-      auto *DVI = dyn_cast<DbgValueInst>(&I);
-      if (!DVI)
-        continue;
-
+    auto CheckForMisSized = [&](auto *DbgVal) {
       unsigned Var = ~0U;
-      (void)to_integer(DVI->getVariable()->getName(), Var, 10);
+      (void)to_integer(DbgVal->getVariable()->getName(), Var, 10);
       assert(Var <= OriginalNumVars && "Unexpected name for DILocalVariable");
-      bool HasBadSize = diagnoseMisSizedDbgValue(M, DVI);
+      bool HasBadSize = diagnoseMisSizedDbgValue(M, DbgVal);
       if (!HasBadSize)
         MissingVars.reset(Var - 1);
       HasErrors |= HasBadSize;
+    };
+    for (Instruction &I : instructions(F)) {
+      for (DbgVariableRecord &DVR : filterDbgVars(I.getDbgRecordRange()))
+        if (DVR.isDbgValue() || DVR.isDbgAssign())
+          CheckForMisSized(&DVR);
+      auto *DVI = dyn_cast<DbgValueInst>(&I);
+      if (!DVI)
+        continue;
+      CheckForMisSized(DVI);
     }
   }
 
@@ -820,9 +807,6 @@ bool checkDebugifyMetadata(Module &M,
   if (Strip)
     Ret = stripDebugifyMetadata(M);
 
-  if (NewDebugMode)
-    M.convertToNewDbgValues();
-
   return Ret;
 }
 
@@ -1052,10 +1036,6 @@ FunctionPass *createCheckDebugifyFunctionPass(
 
 PreservedAnalyses NewPMCheckDebugifyPass::run(Module &M,
                                               ModuleAnalysisManager &) {
-  bool NewDebugMode = M.IsNewDbgInfoFormat;
-  if (NewDebugMode)
-    M.convertFromNewDbgValues();
-
   if (Mode == DebugifyMode::SyntheticDebugInfo)
     checkDebugifyMetadata(M, M.functions(), NameOfWrappedPass,
                                    "CheckModuleDebugify", Strip, StatsMap);
@@ -1065,9 +1045,6 @@ PreservedAnalyses NewPMCheckDebugifyPass::run(Module &M,
       "CheckModuleDebugify (original debuginfo)", NameOfWrappedPass,
       OrigDIVerifyBugsReportFilePath);
 
-  if (NewDebugMode)
-    M.convertToNewDbgValues();
-
   return PreservedAnalyses::all();
 }
 
diff --git a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables-x.mir b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables-x.mir
index eaa627966347f..40ea01189f2cd 100644
--- a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables-x.mir
+++ b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables-x.mir
@@ -1,5 +1,6 @@
 # REQUIRES: x86-registered-target
 # RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-check-debugify -o - %s 2>&1 | FileCheck %s
+# RUN: llc --experimental-debuginfo-iterators=false -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-check-debugify -o - %s 2>&1 | FileCheck %s
 --- |
   ; ModuleID = 'check-line-and-variables.mir'
   source_filename = "check-line-and-variables.c"
diff --git a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.ll b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.ll
index 9033fd2f147c4..56c7cf45705a7 100644
--- a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.ll
+++ b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.ll
@@ -1,4 +1,5 @@
 ; RUN: llc -debugify-check-and-strip-all-safe -o - %s 2>&1 | FileCheck %s
+; RUN: llc --experimental-debuginfo-iterators=false -debugify-check-and-strip-all-safe -o - %s 2>&1 | FileCheck %s
 
 ; ModuleID = 'main.c'
 source_filename = "main.c"
diff --git a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.mir b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.mir
index 9eb722258b703..0805a7f4cfc6c 100644
--- a/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.mir
+++ b/llvm/test/CodeGen/Generic/MIRDebugify/check-line-and-variables.mir
@@ -1,6 +1,8 @@
 # REQUIRES: x86-registered-target
 # RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-debugify,dead-mi-elimination,mir-check-debugify -o - %s 2>&1 | FileCheck %s
 # RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-debugify,mir-check-debugify -o - %s 2>&1 | FileCheck %s --check-prefix=CHECK-PASS
+# RUN: llc --experimental-debuginfo-iterators=false -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-debugify,dead-mi-elimination,mir-check-debugify -o - %s 2>&1 | FileCheck %s
+# RUN: llc --experimental-debuginfo-iterators=false -mtriple=x86_64-unknown-linux-gnu -run-pass=mir-debugify,mir-check-debugify -o - %s 2>&1 | FileCheck %s --check-prefix=CHECK-PASS
 --- |
   ; ModuleID = 'check-line-and-variables.mir'
   source_filename = "check-line-and-variables.ll"
diff --git a/llvm/test/CodeGen/Generic/MIRDebugify/locations-and-values.mir b/llvm/test/CodeGen/Generic/MIRDebugify/locations-and-values.mir
index 59dcff9efd4d5..3035fb8eab3f8 100644
--- a/llvm/test/CodeGen/Generic/MIRDebugify/locations-and-values.mir
+++ b/llvm/test/CodeGen/Generic/MIRDebugify/locations-and-values.mir
@@ -2,6 +2,10 @@
 # RUN: llc -run-pass=mir-debugify -debugify-level=locations -o - %s | FileCheck --check-prefixes=ALL --implicit-check-not=dbg.value %s
 # RUN: llc -run-pass=mir-debugify,mir-strip-debug,mir-debugify -o - %s | FileCheck --check-prefixes=ALL,VALUE %s
 # RUN: llc -run-pass=mir-debugify,mir-strip-debug -o - %s | FileCheck --check-prefix=STRIP %s
+# RUN: llc --experimental-debuginfo-iterators=false -run-pass=mir-debugify -o - %s | FileCheck --check-prefixes=ALL,VALUE %s
+# RUN: llc --experimental-debuginfo-iterators=false -run-pass=mir-debugify -debugify-level=locations -o - %s | FileCheck --check-prefixes=ALL --implicit-check-not=dbg.value %s
+# RUN: llc --experimental-debuginfo-iterators=false -run-pass=mir-debugify,mir-strip-debug,mir-debugify -o - %s | FileCheck --check-prefixes=ALL,VALUE %s
+# RUN: llc --experimental-debuginfo-iterators=false -run-pass=mir-debugify,mir-strip-debug -o - %s | FileCheck --check-prefix=STRIP %s
 
 --- |
   ; ModuleID = 'loc-only.ll'
diff --git a/llvm/test/CodeGen/Generic/MIRDebugify/multifunction-module.mir b/llvm/test/CodeGen/Generic/MIRDebugify/multifunction-module.mir
index fe4fcc1a15bb8..8079db926e1b0 100644
--- a/llvm/test/CodeGen/Generic/MIRDebugify/multifunction-module.mir
+++ b/llvm/test/CodeGen/Generic/MIRDebugify/multifunction-module.mir
@@ -1,6 +1,5 @@
-# FIXME: Remove rm after a few weeks.
-# RUN: rm -f %S/multifunction-module.s
 # RUN: llc -run-pass=mir-debugify,mir-check-debugify -o - %s 2>&1 | FileCheck %s
+# RUN: llc --experimental-debuginfo-iterators=false -run-pass=mir-debugify,mir-check-debugify -o - %s 2>&1 | FileCheck %s
 
 # CHECK: Machine IR debug info check: PASS
 # CHECK-NOT: Assertion `Var <= NumVars && "Unexpected name for DILocalVariable"'
diff --git a/llvm/test/DebugInfo/debugify-bogus-dbg-value.ll b/llvm/test/DebugInfo/debugify-bogus-dbg-value.ll
index 4990979f10c53..55e436b1a93b2 100644
--- a/llvm/test/DebugInfo/debugify-bogus-dbg-value.ll
+++ b/llvm/test/DebugInfo/debugify-bogus-dbg-value.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -passes=check-debugify < %s 2>&1 | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false -passes=check-debugify < %s 2>&1 | FileCheck %s
 
 define <2 x i64> @test-fun(<2 x i64> %A) !dbg !6 {
   %and = and <2 x i64> %A, <i64 23, i64 42>, !dbg !14
diff --git a/llvm/test/DebugInfo/debugify-each.ll b/llvm/test/DebugInfo/debugify-each.ll
index e9241dedb6960..7685b57b5dd15 100644
--- a/llvm/test/DebugInfo/debugify-each.ll
+++ b/llvm/test/DebugInfo/debugify-each.ll
@@ -40,6 +40,40 @@
 ; RUN: opt -debugify-each -passes=globalopt -S -o /dev/null < %s 2> %t
 ; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS-ONE
 
+; Repeat the same checks with debug intrinsics enabled.
+; RUN: opt --experimental-debuginfo-iterators=false -debugify-each -O3 -S -o /dev/null < %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS
+; RUN: FileCheck %s -input-file=%t -check-prefix=FUNCTION-PASS
+; RUN: opt --experimental-debuginfo-iterators=false -disable-output -debugify-each -passes='default<O3>' %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS
+; RUN: FileCheck %s -input-file=%t -check-prefix=FUNCTION-PASS
+
+; RUN: opt --experimental-debuginfo-iterators=false -enable-debugify -debugify-each -O3 -S -o /dev/null < %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS
+; RUN: FileCheck %s -input-file=%t -check-prefix=FUNCTION-PASS
+
+; RUN: opt --experimental-debuginfo-iterators=false -debugify-each -passes='instrprof,instrprof,sroa,sccp' -S -o /dev/null < %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS
+; RUN: FileCheck %s -input-file=%t -check-prefix=FUNCTION-PASS
+
+; RUN: opt --experimental-debuginfo-iterators=false -debugify-each -O1 < %s | opt -O2 -o /dev/null
+
+; RUN: opt --experimental-debuginfo-iterators=false -disable-output -debugify-quiet -debugify-each -O1 < %s 2>&1 | count 0
+
+; RUN: opt --experimental-debuginfo-iterators=false -O1 < %s -S -o %t.before
+; RUN: opt --experimental-debuginfo-iterators=false -O1 -debugify-each < %s -S -o %t.after
+; RUN: diff %t.before %t.after
+
+; RUN: opt --experimental-debuginfo-iterators=false -O1 < %s | llvm-dis -o %t.before
+; RUN: opt --experimental-debuginfo-iterators=false -O1 -debugify-each < %s | llvm-dis -o %t.after
+; RUN: diff %t.before %t.after
+
+; RUN: opt --experimental-debuginfo-iterators=false -debugify-each -passes=instsimplify -S -o /dev/null < %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=FUNCTION-PASS-ONE
+
+; RUN: opt --experimental-debuginfo-iterators=false -debugify-each -passes=globalopt -S -o /dev/null < %s 2> %t
+; RUN: FileCheck %s -input-file=%t -check-prefix=MODULE-PASS-ONE
+
 define void @foo(i32 %arg) {
   call i32 asm "bswap $0", "=r,r"(i32 %arg)
   ret void
diff --git a/llvm/test/DebugInfo/debugify-export.ll b/llvm/test/DebugInfo/debugify-export.ll
index 6e5952d433da9..30333ca908b0d 100644
--- a/llvm/test/DebugInfo/debugify-export.ll
+++ b/llvm/test/DebugInfo/debugify-export.ll
@@ -1,6 +1,9 @@
 ; RUN: opt %s -disable-output -debugify-each -debugify-quiet -debugify-export - -passes=globalopt | FileCheck %s
 ; RUN: opt %s -disable-output -debugify-each -debugify-quiet -debugify-export - -passes=globalopt | FileCheck %s
 
+; RUN: opt --experimental-debuginfo-iterators=false %s -disable-output -debugify-each -debugify-quiet -debugify-export - -passes=globalopt | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false %s -disable-output -debugify-each -debugify-quiet -debugify-export - -passes=globalopt | FileCheck %s
+
 ; CHECK: Pass Name
 ; CHECK-SAME: # of missing debug values
 ; CHECK-SAME: # of missing locations
diff --git a/llvm/test/DebugInfo/debugify-ignore-phi.ll b/llvm/test/DebugInfo/debugify-ignore-phi.ll
index 322ccafa22ac8..643df1d960485 100644
--- a/llvm/test/DebugInfo/debugify-ignore-phi.ll
+++ b/llvm/test/DebugInfo/debugify-ignore-phi.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -passes=check-debugify < %s -S 2>&1 | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false -passes=check-debugify < %s -S 2>&1 | FileCheck %s
 
 define void @test_phi(i1 %cond) !dbg !6 {
   br i1 %cond, label %1, label %2, !dbg !11
diff --git a/llvm/test/DebugInfo/debugify-original-no-dbg-info.ll b/llvm/test/DebugInfo/debugify-original-no-dbg-info.ll
index 941b294fb8556..4cbbfc5c215e2 100644
--- a/llvm/test/DebugInfo/debugify-original-no-dbg-info.ll
+++ b/llvm/test/DebugInfo/debugify-original-no-dbg-info.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -verify-debuginfo-preserve -passes=instcombine -S -o - < %s 2>&1 | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false -verify-debuginfo-preserve -passes=instcombine -S -o - < %s 2>&1 | FileCheck %s
 
 ; CHECK: ModuleDebugify (original debuginfo): Skipping module without debug info
 ; CHECK-NEXT: CheckModuleDebugify (original debuginfo): Skipping module without debug info
diff --git a/llvm/test/DebugInfo/debugify-report-missing-locs-only.ll b/llvm/test/DebugInfo/debugify-report-missing-locs-only.ll
index 1c5daa19c6484..04b7636f025a0 100644
--- a/llvm/test/DebugInfo/debugify-report-missing-locs-only.ll
+++ b/llvm/test/DebugInfo/debugify-report-missing-locs-only.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -passes=check-debugify < %s -S -o - 2>&1 | FileCheck %s -implicit-check-not "WARNING: Instruction with empty DebugLoc in function bar"
+; RUN: opt --experimental-debuginfo-iterators=false -passes=check-debugify < %s -S -o - 2>&1 | FileCheck %s -implicit-check-not "WARNING: Instruction with empty DebugLoc in function bar"
 
 ; CHECK: WARNING: Instruction with empty DebugLoc in function foo --   ret void
 define void @foo() !dbg !6 {
diff --git a/llvm/test/DebugInfo/debugify.ll b/llvm/test/DebugInfo/debugify.ll
index 5ce6795d41b6b..191015f825933 100644
--- a/llvm/test/DebugInfo/debugify.ll
+++ b/llvm/test/DebugInfo/debugify.ll
@@ -25,6 +25,33 @@
 ; RUN: opt -enable-debugify -O1 < %s | opt -O2 -o /dev/null
 ; RUN: opt -passes=debugify,mem2reg,check-debugify < %s | opt -O2 -o /dev/null
 
+;; Perform the same checks again for intrinsic debug info
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify -S -o - < %s | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify -S -o - < %s | FileCheck %s
+
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,debugify -S -o - < %s 2>&1 | \
+; RUN:   FileCheck %s -check-prefix=CHECK-REPEAT
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,debugify -S -o - < %s 2>&1 | \
+; RUN:   FileCheck %s -check-prefix=CHECK-REPEAT
+
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,check-debugify -S -o - < %s | \
+; RUN:   FileCheck %s -implicit-check-not="CheckModuleDebugify: FAIL"
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,check-debugify -S -o - < %s | \
+; RUN:   FileCheck %s -implicit-check-not="CheckModuleDebugify: FAIL"
+; RUN: opt --experimental-debuginfo-iterators=false -enable-debugify -passes=verify -S -o - < %s | \
+; RUN:   FileCheck %s -implicit-check-not="CheckModuleDebugify: FAIL"
+
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,strip,check-debugify -S -o - < %s 2>&1 | \
+; RUN:   FileCheck %s -check-prefix=CHECK-WARN
+
+; RUN: opt --experimental-debuginfo-iterators=false -enable-debugify -passes=strip -S -o - < %s 2>&1 | \
+; RUN:   FileCheck %s -check-prefix=CHECK-WARN
+
+; RUN: opt --experimental-debuginfo-iterators=false -enable-debugify -S -o - < %s 2>&1 | FileCheck %s -check-prefix=PASS
+
+; RUN: opt --experimental-debuginfo-iterators=false -enable-debugify -O1 < %s | opt -O2 -o /dev/null
+; RUN: opt --experimental-debuginfo-iterators=false -passes=debugify,mem2reg,check-debugify < %s | opt -O2 -o /dev/null
+
 ; CHECK-LABEL: define void @foo
 define void @foo() {
 ; CHECK: ret void, !dbg ![[RET1:.*]]
diff --git a/llvm/test/DebugInfo/pr37964.ll b/llvm/test/DebugInfo/pr37964.ll
index 9581f1a6b35dc..63db67d2bd37f 100644
--- a/llvm/test/DebugInfo/pr37964.ll
+++ b/llvm/test/DebugInfo/pr37964.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -disable-output -debugify-each -passes=gvn < %s 2>&1 | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false -disable-output -debugify-each -passes=gvn < %s 2>&1 | FileCheck %s
 
 ; CHECK-NOT: ERROR: Instruction with empty DebugLoc in function _Z3bazv --  {{%.*}} = phi
 ; CHECK: CheckFunctionDebugify [GVNPass]: PASS
diff --git a/llvm/test/DebugInfo/salvage-cast-debug-info.ll b/llvm/test/DebugInfo/salvage-cast-debug-info.ll
index 4676aee3d4e48..b72f717a4f2de 100644
--- a/llvm/test/DebugInfo/salvage-cast-debug-info.ll
+++ b/llvm/test/DebugInfo/salvage-cast-debug-info.ll
@@ -1,5 +1,5 @@
 ; RUN: opt %s -passes=debugify,early-cse -earlycse-debug-hash -S | FileCheck %s
-; RUN: opt %s -passes=debugify,early-cse -earlycse-debug-hash -S --try-experimental-debuginfo-iterators | FileCheck %s
+; RUN: opt --experimental-debuginfo-iterators=false %s -passes=debugify,early-cse -earlycse-debug-hash -S | FileCheck %s
 define i32 @foo(i64 %nose, i32 %more) {
 ; CHECK-LABEL: @foo(
 ; CHECK: call void @llvm.dbg.value(metadata i64 %nose, metadata [[V1:![0-9]+]], metadata !DIExpression(DW_OP_LLVM_convert, 64, DW_ATE_unsigned, DW_OP_LLVM_convert, 32, DW_ATE_unsigned
diff --git a/llvm/test/DebugInfo/verify-di-preserve.ll b/llvm/test/DebugInfo/verify-di-preserve.ll
index a2f1b1dd78dc5..92fc62a0b34c4 100644
--- a/llvm/test/DebugInfo/verify-di-preserve.ll
+++ b/llvm/test/DebugInfo/verify-di-preserve.ll
@@ -1,10 +1,10 @@
 ; RUN: opt %s -verify-debuginfo-preserve -passes=instcombine -disable-output 2>&1 | FileCheck --check-prefix=VERIFY %s
-; RUN: opt --try-experimental-debuginfo-iterators %s -verify-debuginfo-preserve -passes=instcombine -disable-output 2>&1 | FileCheck --check-prefix=VERIFY %s
+; RUN: opt --experimental-debuginfo-iterators=false %s -verify-debuginfo-preserve -passes=instcombine -disable-output 2>&1 | FileCheck --check-prefix=VERIFY %s
 
 ; VERIFY: CheckModuleDebugify (original debuginfo):
 
 ; RUN: opt %s -verify-each-debuginfo-preserve -O2 -disable-output 2>&1 | FileCheck --check-prefix=VERIFY-EACH %s
-; RUN: opt %s  --try-experimental-debuginfo-iterators -verify-each-debuginfo-preserve -O2 -disable-output 2>&1 | FileCheck --check-prefix=VERIFY-EACH %s
+; RUN: opt %s --experimental-debuginfo-iterators=false -verify-each-debuginfo-preserve -O2 -disable-output 2>&1 | FileCheck --check-prefix=VERIFY-EACH %s
 
 ; VERIFY-EACH: DeadArgumentEliminationPass
 ; VERIFY-EACH: GlobalDCEPass

From 45eabd1362527d1b3a27a90f7479865785b763ee Mon Sep 17 00:00:00 2001
From: YunQiang Su <syq@debian.org>
Date: Wed, 17 Apr 2024 00:52:15 +0800
Subject: [PATCH 614/886] [flang,test] Add -resource-dir option to
 msvc-dependent-lib-flags.f90 (#88894)

For aarch64-windows-msvc, clang_rt.builtins is placed in windows subdir
instead of triple subdir, and the name of clang_rt.builtins is
clang_rt.builtins-aarch64.lib.

So let's use `-resource-dir` option to fix test failure.

Please see talk for PR#87866.
---
 flang/test/Driver/msvc-dependent-lib-flags.f90 | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/flang/test/Driver/msvc-dependent-lib-flags.f90 b/flang/test/Driver/msvc-dependent-lib-flags.f90
index 643dbe9e949cb..6cfc969e92b20 100644
--- a/flang/test/Driver/msvc-dependent-lib-flags.f90
+++ b/flang/test/Driver/msvc-dependent-lib-flags.f90
@@ -1,7 +1,7 @@
-! RUN: %flang -### --target=aarch64-windows-msvc %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC
-! RUN: %flang -### --target=aarch64-windows-msvc -fms-runtime-lib=static_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DEBUG
-! RUN: %flang -### --target=aarch64-windows-msvc -fms-runtime-lib=dll %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL
-! RUN: %flang -### --target=aarch64-windows-msvc -fms-runtime-lib=dll_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL-DEBUG
+! RUN: %flang -### --target=aarch64-windows-msvc -resource-dir=%S/Inputs/resource_dir %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC
+! RUN: %flang -### --target=aarch64-windows-msvc -resource-dir=%S/Inputs/resource_dir -fms-runtime-lib=static_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DEBUG
+! RUN: %flang -### --target=aarch64-windows-msvc -resource-dir=%S/Inputs/resource_dir -fms-runtime-lib=dll %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL
+! RUN: %flang -### --target=aarch64-windows-msvc -resource-dir=%S/Inputs/resource_dir -fms-runtime-lib=dll_dbg %S/Inputs/hello.f90 -v 2>&1 | FileCheck %s --check-prefixes=MSVC-DLL-DEBUG
 
 ! MSVC: -fc1
 ! MSVC-SAME: --dependent-lib=clang_rt.builtins.lib

From 22e6bf77ad8781810fc81fff4c447c03cdf6f419 Mon Sep 17 00:00:00 2001
From: Volodymyr Sapsai <vsapsai@apple.com>
Date: Tue, 16 Apr 2024 10:12:26 -0700
Subject: [PATCH 615/886] [unused-includes][Serialization] Remove unused
 includes. NFC. (#88790)

---
 clang/include/clang/Serialization/ModuleFileExtension.h | 1 -
 clang/lib/Serialization/ASTReader.cpp                   | 1 -
 clang/lib/Serialization/ASTWriterDecl.cpp               | 1 -
 clang/lib/Serialization/ASTWriterStmt.cpp               | 1 -
 clang/lib/Serialization/GeneratePCH.cpp                 | 1 -
 clang/lib/Serialization/GlobalModuleIndex.cpp           | 1 -
 clang/lib/Serialization/ModuleFileExtension.cpp         | 2 +-
 clang/lib/Serialization/PCHContainerOperations.cpp      | 2 --
 8 files changed, 1 insertion(+), 9 deletions(-)

diff --git a/clang/include/clang/Serialization/ModuleFileExtension.h b/clang/include/clang/Serialization/ModuleFileExtension.h
index d7d456c8b5db8..50ce401516275 100644
--- a/clang/include/clang/Serialization/ModuleFileExtension.h
+++ b/clang/include/clang/Serialization/ModuleFileExtension.h
@@ -9,7 +9,6 @@
 #ifndef LLVM_CLANG_SERIALIZATION_MODULEFILEEXTENSION_H
 #define LLVM_CLANG_SERIALIZATION_MODULEFILEEXTENSION_H
 
-#include "llvm/ADT/IntrusiveRefCntPtr.h"
 #include "llvm/Support/ExtensibleRTTI.h"
 #include "llvm/Support/HashBuilder.h"
 #include "llvm/Support/MD5.h"
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index cf0726460bfca..b28df03b4a95e 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -31,7 +31,6 @@
 #include "clang/AST/ExternalASTSource.h"
 #include "clang/AST/NestedNameSpecifier.h"
 #include "clang/AST/ODRDiagsEmitter.h"
-#include "clang/AST/ODRHash.h"
 #include "clang/AST/OpenACCClause.h"
 #include "clang/AST/OpenMPClause.h"
 #include "clang/AST/RawCommentList.h"
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index d0d49bcdf991a..c6db107e0ca42 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -16,7 +16,6 @@
 #include "clang/AST/DeclTemplate.h"
 #include "clang/AST/DeclVisitor.h"
 #include "clang/AST/Expr.h"
-#include "clang/AST/ODRHash.h"
 #include "clang/AST/OpenMPClause.h"
 #include "clang/AST/PrettyDeclStackTrace.h"
 #include "clang/Basic/SourceManager.h"
diff --git a/clang/lib/Serialization/ASTWriterStmt.cpp b/clang/lib/Serialization/ASTWriterStmt.cpp
index e3816181e2b2b..a736a7b0ef726 100644
--- a/clang/lib/Serialization/ASTWriterStmt.cpp
+++ b/clang/lib/Serialization/ASTWriterStmt.cpp
@@ -19,7 +19,6 @@
 #include "clang/AST/ExprOpenMP.h"
 #include "clang/AST/StmtVisitor.h"
 #include "clang/Lex/Token.h"
-#include "clang/Sema/DeclSpec.h"
 #include "clang/Serialization/ASTRecordWriter.h"
 #include "llvm/Bitstream/BitstreamWriter.h"
 using namespace clang;
diff --git a/clang/lib/Serialization/GeneratePCH.cpp b/clang/lib/Serialization/GeneratePCH.cpp
index 2fece29f34487..bed74399098d7 100644
--- a/clang/lib/Serialization/GeneratePCH.cpp
+++ b/clang/lib/Serialization/GeneratePCH.cpp
@@ -17,7 +17,6 @@
 #include "clang/Lex/HeaderSearchOptions.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/SemaConsumer.h"
-#include "clang/Serialization/ASTReader.h"
 #include "clang/Serialization/ASTWriter.h"
 #include "llvm/Bitstream/BitstreamWriter.h"
 
diff --git a/clang/lib/Serialization/GlobalModuleIndex.cpp b/clang/lib/Serialization/GlobalModuleIndex.cpp
index 8ff10f6a8621e..f09ceb8d31620 100644
--- a/clang/lib/Serialization/GlobalModuleIndex.cpp
+++ b/clang/lib/Serialization/GlobalModuleIndex.cpp
@@ -13,7 +13,6 @@
 #include "clang/Serialization/GlobalModuleIndex.h"
 #include "ASTReaderInternals.h"
 #include "clang/Basic/FileManager.h"
-#include "clang/Lex/HeaderSearch.h"
 #include "clang/Serialization/ASTBitCodes.h"
 #include "clang/Serialization/ModuleFile.h"
 #include "clang/Serialization/PCHContainerOperations.h"
diff --git a/clang/lib/Serialization/ModuleFileExtension.cpp b/clang/lib/Serialization/ModuleFileExtension.cpp
index 95fff41e0d7a8..729529b5fca18 100644
--- a/clang/lib/Serialization/ModuleFileExtension.cpp
+++ b/clang/lib/Serialization/ModuleFileExtension.cpp
@@ -6,7 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 #include "clang/Serialization/ModuleFileExtension.h"
-#include "llvm/ADT/Hashing.h"
+
 using namespace clang;
 
 char ModuleFileExtension::ID = 0;
diff --git a/clang/lib/Serialization/PCHContainerOperations.cpp b/clang/lib/Serialization/PCHContainerOperations.cpp
index 56ca3394385b4..4aedb7debcff2 100644
--- a/clang/lib/Serialization/PCHContainerOperations.cpp
+++ b/clang/lib/Serialization/PCHContainerOperations.cpp
@@ -12,8 +12,6 @@
 
 #include "clang/Serialization/PCHContainerOperations.h"
 #include "clang/AST/ASTConsumer.h"
-#include "clang/Lex/ModuleLoader.h"
-#include "llvm/Bitstream/BitstreamReader.h"
 #include "llvm/Support/raw_ostream.h"
 #include <utility>
 

From b566810add5b7c5695bdd2c39710b78af9dc83ba Mon Sep 17 00:00:00 2001
From: Volodymyr Sapsai <vsapsai@apple.com>
Date: Tue, 16 Apr 2024 10:13:15 -0700
Subject: [PATCH 616/886] [unused-includes] PCHContainerOperations uses
 MemoryBufferRef, not MemoryBuffer. NFC. (#88794)

---
 clang/include/clang/Serialization/PCHContainerOperations.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/include/clang/Serialization/PCHContainerOperations.h b/clang/include/clang/Serialization/PCHContainerOperations.h
index ddfddf2dafadf..c9a7e334ce6eb 100644
--- a/clang/include/clang/Serialization/PCHContainerOperations.h
+++ b/clang/include/clang/Serialization/PCHContainerOperations.h
@@ -12,7 +12,7 @@
 #include "clang/Basic/Module.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringMap.h"
-#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/MemoryBufferRef.h"
 #include <memory>
 
 namespace llvm {

From 454d4496970f665200b5b300578894d78405b6ca Mon Sep 17 00:00:00 2001
From: Xing Xue <xingxue@outlook.com>
Date: Tue, 16 Apr 2024 13:13:49 -0400
Subject: [PATCH 617/886] [OpenMP] Use a memory fence before incrementing the
 dispatch buffer index (#87995)

This patch uses a memory fence in function `__kmp_dispatch_next()` to
flush pending memory write invalidates before incrementing the
`volatile` variable `buffer_index` to fix intermittent time-outs of
OpenMP runtime LIT test cases `env/kmp_set_dispatch_buf.c` and
`worksharing/for/kmp_set_dispatch_buf.c`, noting that the same is needed
for incrementing `buffer_index` in function `__kmpc_next_section()`
(line 2600 of `kmp_dispatch.cpp`).
---
 openmp/runtime/src/kmp_dispatch.cpp | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/openmp/runtime/src/kmp_dispatch.cpp b/openmp/runtime/src/kmp_dispatch.cpp
index ac85b2b3f2fcd..fc33376511817 100644
--- a/openmp/runtime/src/kmp_dispatch.cpp
+++ b/openmp/runtime/src/kmp_dispatch.cpp
@@ -2397,6 +2397,8 @@ static int __kmp_dispatch_next(ident_t *loc, int gtid, kmp_int32 *p_last,
           sh->u.s.ordered_iteration = 0;
         }
 
+        KMP_MB(); /* Flush all pending memory write invalidates.  */
+
         sh->buffer_index += __kmp_dispatch_num_buffers;
         KD_TRACE(100, ("__kmp_dispatch_next: T#%d change buffer_index:%d\n",
                        gtid, sh->buffer_index));

From 22bba85d82637d0446928ff6ff41f98583f3d3b2 Mon Sep 17 00:00:00 2001
From: Xing Xue <xingxue@outlook.com>
Date: Tue, 16 Apr 2024 13:14:29 -0400
Subject: [PATCH 618/886] [OpenMP][test][AIX] Make 64 the max number of threads
 for capacity tests in AIX 32-bit (#88739)

This patch makes 64 the max number of threads for 2 capacity tests in
AIX 32-bit mode rather than `XFAIL`ing them.
---
 .../capacity_mix_threads.cpp                  | 20 +++++++++++++------
 .../hidden_helper_task/capacity_nthreads.cpp  | 20 +++++++++++++------
 2 files changed, 28 insertions(+), 12 deletions(-)

diff --git a/openmp/runtime/test/tasking/hidden_helper_task/capacity_mix_threads.cpp b/openmp/runtime/test/tasking/hidden_helper_task/capacity_mix_threads.cpp
index 3f2ceef0c4add..36825dbebafb5 100644
--- a/openmp/runtime/test/tasking/hidden_helper_task/capacity_mix_threads.cpp
+++ b/openmp/runtime/test/tasking/hidden_helper_task/capacity_mix_threads.cpp
@@ -1,7 +1,4 @@
 // RUN: %libomp-cxx-compile-and-run
-//
-// AIX runs out of resource in 32-bit with 4*omp_get_max_threads() threads.
-// XFAIL: aix && ppc
 
 #include <omp.h>
 
@@ -11,6 +8,12 @@
 #include <thread>
 #include <vector>
 
+// AIX runs out of resource in 32-bit if 4*omp_get_max_threads() is more
+// than 64 threads with the default stack size.
+#if defined(_AIX) && !__LP64__
+#define MAX_THREADS 64
+#endif
+
 void dummy_root() {
   // omp_get_max_threads() will do middle initialization
   int nthreads = omp_get_max_threads();
@@ -18,9 +21,14 @@ void dummy_root() {
 }
 
 int main(int argc, char *argv[]) {
-  const int N = std::min(std::max(std::max(32, 4 * omp_get_max_threads()),
-                                  4 * omp_get_num_procs()),
-                         std::numeric_limits<int>::max());
+  int N = std::min(std::max(std::max(32, 4 * omp_get_max_threads()),
+                            4 * omp_get_num_procs()),
+                   std::numeric_limits<int>::max());
+
+#if defined(_AIX) && !__LP64__
+  if (N > MAX_THREADS)
+    N = MAX_THREADS;
+#endif
 
   std::vector<int> data(N);
 
diff --git a/openmp/runtime/test/tasking/hidden_helper_task/capacity_nthreads.cpp b/openmp/runtime/test/tasking/hidden_helper_task/capacity_nthreads.cpp
index f7405d00255cb..1cceee95e704b 100644
--- a/openmp/runtime/test/tasking/hidden_helper_task/capacity_nthreads.cpp
+++ b/openmp/runtime/test/tasking/hidden_helper_task/capacity_nthreads.cpp
@@ -1,7 +1,4 @@
 // RUN: %libomp-cxx-compile-and-run
-//
-// AIX runs out of resource in 32-bit with 4*omp_get_max_threads() threads.
-// XFAIL: aix && ppc
 
 #include <omp.h>
 
@@ -10,10 +7,21 @@
 #include <limits>
 #include <vector>
 
+// AIX runs out of resource in 32-bit if 4*omp_get_max_threads() is more
+// than 64 threads with the default stacksize.
+#if defined(_AIX) && !__LP64__
+#define MAX_THREADS 64
+#endif
+
 int main(int argc, char *argv[]) {
-  const int N = std::min(std::max(std::max(32, 4 * omp_get_max_threads()),
-                                  4 * omp_get_num_procs()),
-                         std::numeric_limits<int>::max());
+  int N = std::min(std::max(std::max(32, 4 * omp_get_max_threads()),
+                            4 * omp_get_num_procs()),
+                   std::numeric_limits<int>::max());
+
+#if defined(_AIX) && !__LP64__
+  if (N > MAX_THREADS)
+    N = MAX_THREADS;
+#endif
 
   std::vector<int> data(N);
 

From 8137bd9e03d636a27701a85b6efe899f9571cac5 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 10:16:48 -0700
Subject: [PATCH 619/886] [memprof] Use CSId to construct MemProfRecord
 (#88362)

We are in the process of referring to call stacks with CallStackId in
IndexedMemProfRecord and IndexedAllocationInfo instead of holding call
stacks inline (both in memory and the serialized format).  Doing so
deduplicates call stacks and reduces the MemProf profile file size.

Before we can eliminate the two fields holding call stacks inline:

- IndexedAllocationInfo::CallStack
- IndexedMemProfRecord::CallSites

we need to eliminate all the read operations on them.

This patch is a step toward that direction.  Specifically, we
eliminate the read operations in the context of MemProfReader and
RawMemProfReader.  A subsequent patch will eliminate the read
operations during the serialization.
---
 llvm/include/llvm/ProfileData/MemProf.h       |  8 ++
 llvm/include/llvm/ProfileData/MemProfReader.h | 20 ++++-
 llvm/lib/ProfileData/MemProf.cpp              | 18 ++++
 llvm/lib/ProfileData/MemProfReader.cpp        | 26 ++++++
 llvm/unittests/ProfileData/MemProfTest.cpp    | 85 +++++++++++++++++++
 5 files changed, 153 insertions(+), 4 deletions(-)

diff --git a/llvm/include/llvm/ProfileData/MemProf.h b/llvm/include/llvm/ProfileData/MemProf.h
index 0431c182276ec..3520034fb1c94 100644
--- a/llvm/include/llvm/ProfileData/MemProf.h
+++ b/llvm/include/llvm/ProfileData/MemProf.h
@@ -16,6 +16,8 @@
 namespace llvm {
 namespace memprof {
 
+struct MemProfRecord;
+
 // The versions of the indexed MemProf format
 enum IndexedVersion : uint64_t {
   // Version 0: This version didn't have a version field.
@@ -392,6 +394,12 @@ struct IndexedMemProfRecord {
                                           const unsigned char *Buffer,
                                           IndexedVersion Version);
 
+  // Convert IndexedMemProfRecord to MemProfRecord.  Callback is used to
+  // translate CallStackId to call stacks with frames inline.
+  MemProfRecord toMemProfRecord(
+      std::function<const llvm::SmallVector<Frame>(const CallStackId)> Callback)
+      const;
+
   // Returns the GUID for the function name after canonicalization. For
   // memprof, we remove any .llvm suffix added by LTO. MemProfRecords are
   // mapped to functions using this GUID.
diff --git a/llvm/include/llvm/ProfileData/MemProfReader.h b/llvm/include/llvm/ProfileData/MemProfReader.h
index 89f49a20a6089..1f84fefad03e3 100644
--- a/llvm/include/llvm/ProfileData/MemProfReader.h
+++ b/llvm/include/llvm/ProfileData/MemProfReader.h
@@ -70,8 +70,20 @@ class MemProfReader {
       Callback =
           std::bind(&MemProfReader::idToFrame, this, std::placeholders::_1);
 
+    auto CallStackCallback = [&](CallStackId CSId) {
+      llvm::SmallVector<Frame> CallStack;
+      auto Iter = CSIdToCallStack.find(CSId);
+      assert(Iter != CSIdToCallStack.end());
+      for (FrameId Id : Iter->second)
+        CallStack.push_back(Callback(Id));
+      return CallStack;
+    };
+
     const IndexedMemProfRecord &IndexedRecord = Iter->second;
-    GuidRecord = {Iter->first, MemProfRecord(IndexedRecord, Callback)};
+    GuidRecord = {
+        Iter->first,
+        IndexedRecord.toMemProfRecord(CallStackCallback),
+    };
     Iter++;
     return Error::success();
   }
@@ -84,9 +96,7 @@ class MemProfReader {
   // Initialize the MemProfReader with the frame mappings and profile contents.
   MemProfReader(
       llvm::DenseMap<FrameId, Frame> FrameIdMap,
-      llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData)
-      : IdToFrame(std::move(FrameIdMap)),
-        FunctionProfileData(std::move(ProfData)) {}
+      llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData);
 
 protected:
   // A helper method to extract the frame from the IdToFrame map.
@@ -97,6 +107,8 @@ class MemProfReader {
   }
   // A mapping from FrameId (a hash of the contents) to the frame.
   llvm::DenseMap<FrameId, Frame> IdToFrame;
+  // A mapping from CallStackId to the call stack.
+  llvm::DenseMap<CallStackId, llvm::SmallVector<FrameId>> CSIdToCallStack;
   // A mapping from function GUID, hash of the canonical function symbol to the
   // memprof profile data for that function, i.e allocation and callsite info.
   llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> FunctionProfileData;
diff --git a/llvm/lib/ProfileData/MemProf.cpp b/llvm/lib/ProfileData/MemProf.cpp
index 97414505f1c13..1ca0a02d3cbde 100644
--- a/llvm/lib/ProfileData/MemProf.cpp
+++ b/llvm/lib/ProfileData/MemProf.cpp
@@ -224,6 +224,24 @@ IndexedMemProfRecord::deserialize(const MemProfSchema &Schema,
   llvm_unreachable("unsupported MemProf version");
 }
 
+MemProfRecord IndexedMemProfRecord::toMemProfRecord(
+    std::function<const llvm::SmallVector<Frame>(const CallStackId)> Callback)
+    const {
+  MemProfRecord Record;
+
+  for (const memprof::IndexedAllocationInfo &IndexedAI : AllocSites) {
+    memprof::AllocationInfo AI;
+    AI.Info = IndexedAI.Info;
+    AI.CallStack = Callback(IndexedAI.CSId);
+    Record.AllocSites.push_back(AI);
+  }
+
+  for (memprof::CallStackId CSId : CallSiteIds)
+    Record.CallSites.push_back(Callback(CSId));
+
+  return Record;
+}
+
 GlobalValue::GUID IndexedMemProfRecord::getGUID(const StringRef FunctionName) {
   // Canonicalize the function name to drop suffixes such as ".llvm.". Note
   // we do not drop any ".__uniq." suffixes, as getCanonicalFnName does not drop
diff --git a/llvm/lib/ProfileData/MemProfReader.cpp b/llvm/lib/ProfileData/MemProfReader.cpp
index 580867a9083fd..91556f036c777 100644
--- a/llvm/lib/ProfileData/MemProfReader.cpp
+++ b/llvm/lib/ProfileData/MemProfReader.cpp
@@ -183,6 +183,28 @@ std::string getBuildIdString(const SegmentEntry &Entry) {
 }
 } // namespace
 
+MemProfReader::MemProfReader(
+    llvm::DenseMap<FrameId, Frame> FrameIdMap,
+    llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData)
+    : IdToFrame(std::move(FrameIdMap)),
+      FunctionProfileData(std::move(ProfData)) {
+  // Populate CSId in each IndexedAllocationInfo and IndexedMemProfRecord
+  // while storing CallStack in CSIdToCallStack.
+  for (auto &KV : FunctionProfileData) {
+    IndexedMemProfRecord &Record = KV.second;
+    for (auto &AS : Record.AllocSites) {
+      CallStackId CSId = hashCallStack(AS.CallStack);
+      AS.CSId = CSId;
+      CSIdToCallStack.insert({CSId, AS.CallStack});
+    }
+    for (auto &CS : Record.CallSites) {
+      CallStackId CSId = hashCallStack(CS);
+      Record.CallSiteIds.push_back(CSId);
+      CSIdToCallStack.insert({CSId, CS});
+    }
+  }
+}
+
 Expected<std::unique_ptr<RawMemProfReader>>
 RawMemProfReader::create(const Twine &Path, const StringRef ProfiledBinary,
                          bool KeepName) {
@@ -445,6 +467,7 @@ Error RawMemProfReader::mapRawProfileToRecords() {
     }
 
     CallStackId CSId = hashCallStack(Callstack);
+    CSIdToCallStack.insert({CSId, Callstack});
 
     // We attach the memprof record to each function bottom-up including the
     // first non-inline frame.
@@ -467,7 +490,10 @@ Error RawMemProfReader::mapRawProfileToRecords() {
     auto Result = FunctionProfileData.insert({Id, IndexedMemProfRecord()});
     IndexedMemProfRecord &Record = Result.first->second;
     for (LocationPtr Loc : Locs) {
+      CallStackId CSId = hashCallStack(*Loc);
+      CSIdToCallStack.insert({CSId, *Loc});
       Record.CallSites.push_back(*Loc);
+      Record.CallSiteIds.push_back(CSId);
     }
   }
 
diff --git a/llvm/unittests/ProfileData/MemProfTest.cpp b/llvm/unittests/ProfileData/MemProfTest.cpp
index 9cf307472d656..ab9227e9df881 100644
--- a/llvm/unittests/ProfileData/MemProfTest.cpp
+++ b/llvm/unittests/ProfileData/MemProfTest.cpp
@@ -21,9 +21,11 @@ using ::llvm::DILineInfo;
 using ::llvm::DILineInfoSpecifier;
 using ::llvm::DILocal;
 using ::llvm::StringRef;
+using ::llvm::memprof::CallStackId;
 using ::llvm::memprof::CallStackMap;
 using ::llvm::memprof::Frame;
 using ::llvm::memprof::FrameId;
+using ::llvm::memprof::IndexedAllocationInfo;
 using ::llvm::memprof::IndexedMemProfRecord;
 using ::llvm::memprof::MemInfoBlock;
 using ::llvm::memprof::MemProfReader;
@@ -36,6 +38,7 @@ using ::llvm::memprof::SegmentEntry;
 using ::llvm::object::SectionedAddress;
 using ::llvm::symbolize::SymbolizableModule;
 using ::testing::Return;
+using ::testing::SizeIs;
 
 class MockSymbolizer : public SymbolizableModule {
 public:
@@ -432,4 +435,86 @@ TEST(MemProf, BaseMemProfReader) {
   EXPECT_THAT(Records[0].AllocSites[0].CallStack[1],
               FrameContains("bar", 10U, 2U, false));
 }
+
+TEST(MemProf, IndexedMemProfRecordToMemProfRecord) {
+  // Verify that MemProfRecord can be constructed from IndexedMemProfRecord with
+  // CallStackIds only.
+
+  llvm::DenseMap<FrameId, Frame> FrameIdMap;
+  Frame F1(1, 0, 0, false);
+  Frame F2(2, 0, 0, false);
+  Frame F3(3, 0, 0, false);
+  Frame F4(4, 0, 0, false);
+  FrameIdMap.insert({F1.hash(), F1});
+  FrameIdMap.insert({F2.hash(), F2});
+  FrameIdMap.insert({F3.hash(), F3});
+  FrameIdMap.insert({F4.hash(), F4});
+
+  llvm::DenseMap<CallStackId, llvm::SmallVector<FrameId>> CallStackIdMap;
+  llvm::SmallVector<FrameId> CS1 = {F1.hash(), F2.hash()};
+  llvm::SmallVector<FrameId> CS2 = {F1.hash(), F3.hash()};
+  llvm::SmallVector<FrameId> CS3 = {F2.hash(), F3.hash()};
+  llvm::SmallVector<FrameId> CS4 = {F2.hash(), F4.hash()};
+  CallStackIdMap.insert({llvm::memprof::hashCallStack(CS1), CS1});
+  CallStackIdMap.insert({llvm::memprof::hashCallStack(CS2), CS2});
+  CallStackIdMap.insert({llvm::memprof::hashCallStack(CS3), CS3});
+  CallStackIdMap.insert({llvm::memprof::hashCallStack(CS4), CS4});
+
+  IndexedMemProfRecord IndexedRecord;
+  IndexedAllocationInfo AI;
+  AI.CSId = llvm::memprof::hashCallStack(CS1);
+  IndexedRecord.AllocSites.push_back(AI);
+  AI.CSId = llvm::memprof::hashCallStack(CS2);
+  IndexedRecord.AllocSites.push_back(AI);
+  IndexedRecord.CallSiteIds.push_back(llvm::memprof::hashCallStack(CS3));
+  IndexedRecord.CallSiteIds.push_back(llvm::memprof::hashCallStack(CS4));
+
+  bool CSIdMissing = false;
+  bool FrameIdMissing = false;
+
+  auto Callback = [&](CallStackId CSId) -> llvm::SmallVector<Frame> {
+    llvm::SmallVector<Frame> CallStack;
+    llvm::SmallVector<FrameId> FrameIds;
+
+    auto Iter = CallStackIdMap.find(CSId);
+    if (Iter == CallStackIdMap.end())
+      CSIdMissing = true;
+    else
+      FrameIds = Iter->second;
+
+    for (FrameId Id : FrameIds) {
+      Frame F(0, 0, 0, false);
+      auto Iter = FrameIdMap.find(Id);
+      if (Iter == FrameIdMap.end())
+        FrameIdMissing = true;
+      else
+        F = Iter->second;
+      CallStack.push_back(F);
+    }
+
+    return CallStack;
+  };
+
+  MemProfRecord Record = IndexedRecord.toMemProfRecord(Callback);
+
+  // Make sure that all lookups are successful.
+  ASSERT_FALSE(CSIdMissing);
+  ASSERT_FALSE(FrameIdMissing);
+
+  // Verify the contents of Record.
+  ASSERT_THAT(Record.AllocSites, SizeIs(2));
+  ASSERT_THAT(Record.AllocSites[0].CallStack, SizeIs(2));
+  EXPECT_EQ(Record.AllocSites[0].CallStack[0].hash(), F1.hash());
+  EXPECT_EQ(Record.AllocSites[0].CallStack[1].hash(), F2.hash());
+  ASSERT_THAT(Record.AllocSites[1].CallStack, SizeIs(2));
+  EXPECT_EQ(Record.AllocSites[1].CallStack[0].hash(), F1.hash());
+  EXPECT_EQ(Record.AllocSites[1].CallStack[1].hash(), F3.hash());
+  ASSERT_THAT(Record.CallSites, SizeIs(2));
+  ASSERT_THAT(Record.CallSites[0], SizeIs(2));
+  EXPECT_EQ(Record.CallSites[0][0].hash(), F2.hash());
+  EXPECT_EQ(Record.CallSites[0][1].hash(), F3.hash());
+  ASSERT_THAT(Record.CallSites[1], SizeIs(2));
+  EXPECT_EQ(Record.CallSites[1][0].hash(), F2.hash());
+  EXPECT_EQ(Record.CallSites[1][1].hash(), F4.hash());
+}
 } // namespace

From 8cd8ebe153391993a3668d2ac8d2994d9491f3ef Mon Sep 17 00:00:00 2001
From: Aiden Grossman <aidengrossman@google.com>
Date: Tue, 16 Apr 2024 10:25:02 -0700
Subject: [PATCH 620/886] [MLGO] Use double comparison facilities for reg alloc
 scoring tests (#88862)

This patch switches from using direct equality (ASSERT_EQ) to the
floating point comparison facilities (ASSERT_DOUBLE_EQ) within google
test to avoid weird floating point problems. There is at least one
downstream that maintains a patch for issues cropping up from the direct
equality.


https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/main/llvm17/allocscore.patch
---
 llvm/unittests/CodeGen/RegAllocScoreTest.cpp | 27 ++++++++++----------
 1 file changed, 14 insertions(+), 13 deletions(-)

diff --git a/llvm/unittests/CodeGen/RegAllocScoreTest.cpp b/llvm/unittests/CodeGen/RegAllocScoreTest.cpp
index ff7146eaf9439..eae517f9d01cf 100644
--- a/llvm/unittests/CodeGen/RegAllocScoreTest.cpp
+++ b/llvm/unittests/CodeGen/RegAllocScoreTest.cpp
@@ -166,19 +166,20 @@ TEST(RegAllocScoreTest, Counts) {
   ASSERT_EQ(MF->size(), 2U);
   const auto TotalScore =
       llvm::calculateRegAllocScore(*MF, MBBFreqMock, IsRemat);
-  ASSERT_EQ(Freq1, TotalScore.copyCounts());
-  ASSERT_EQ(2.0 * Freq1 + Freq2, TotalScore.loadCounts());
-  ASSERT_EQ(Freq1 + Freq2, TotalScore.storeCounts());
-  ASSERT_EQ(Freq2, TotalScore.loadStoreCounts());
-  ASSERT_EQ(Freq1, TotalScore.cheapRematCounts());
-  ASSERT_EQ(Freq2, TotalScore.expensiveRematCounts());
-  ASSERT_EQ(TotalScore.getScore(),
-            TotalScore.copyCounts() * CopyWeight +
-                TotalScore.loadCounts() * LoadWeight +
-                TotalScore.storeCounts() * StoreWeight +
-                TotalScore.loadStoreCounts() * (LoadWeight + StoreWeight) +
-                TotalScore.cheapRematCounts() * CheapRematWeight +
-                TotalScore.expensiveRematCounts() * ExpensiveRematWeight
+  ASSERT_DOUBLE_EQ(Freq1, TotalScore.copyCounts());
+  ASSERT_DOUBLE_EQ(2.0 * Freq1 + Freq2, TotalScore.loadCounts());
+  ASSERT_DOUBLE_EQ(Freq1 + Freq2, TotalScore.storeCounts());
+  ASSERT_DOUBLE_EQ(Freq2, TotalScore.loadStoreCounts());
+  ASSERT_DOUBLE_EQ(Freq1, TotalScore.cheapRematCounts());
+  ASSERT_DOUBLE_EQ(Freq2, TotalScore.expensiveRematCounts());
+  ASSERT_DOUBLE_EQ(
+      TotalScore.getScore(),
+      TotalScore.copyCounts() * CopyWeight +
+          TotalScore.loadCounts() * LoadWeight +
+          TotalScore.storeCounts() * StoreWeight +
+          TotalScore.loadStoreCounts() * (LoadWeight + StoreWeight) +
+          TotalScore.cheapRematCounts() * CheapRematWeight +
+          TotalScore.expensiveRematCounts() * ExpensiveRematWeight
 
   );
 }

From 75054525ae58f26c86e418382164540760871186 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Tue, 16 Apr 2024 18:42:23 +0200
Subject: [PATCH 621/886] [clang][Interp] Implement VectorSplat casts

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 29 ++++++++++++++++++++++++
 clang/lib/AST/Interp/ByteCodeExprGen.h   |  9 +++++++-
 clang/test/AST/Interp/vectors.cpp        | 17 ++++++++++++--
 3 files changed, 52 insertions(+), 3 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 00c4a9f161304..6b4b51aac41e8 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -398,6 +398,35 @@ bool ByteCodeExprGen<Emitter>::VisitCastExpr(const CastExpr *CE) {
     return true;
   }
 
+  case CK_VectorSplat: {
+    assert(!classify(CE->getType()));
+    assert(classify(SubExpr->getType()));
+    assert(CE->getType()->isVectorType());
+
+    if (DiscardResult)
+      return this->discard(SubExpr);
+
+    assert(Initializing); // FIXME: Not always correct.
+    const auto *VT = CE->getType()->getAs<VectorType>();
+    PrimType ElemT = classifyPrim(SubExpr);
+    unsigned ElemOffset = allocateLocalPrimitive(
+        SubExpr, ElemT, /*IsConst=*/true, /*IsExtended=*/false);
+
+    if (!this->visit(SubExpr))
+      return false;
+    if (!this->emitSetLocal(ElemT, ElemOffset, CE))
+      return false;
+
+    for (unsigned I = 0; I != VT->getNumElements(); ++I) {
+      if (!this->emitGetLocal(ElemT, ElemOffset, CE))
+        return false;
+      if (!this->emitInitElem(ElemT, I, CE))
+        return false;
+    }
+
+    return true;
+  }
+
   case CK_ToVoid:
     return discard(SubExpr);
 
diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.h b/clang/lib/AST/Interp/ByteCodeExprGen.h
index db0d73ce23f7c..7e9dc8631fc0d 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.h
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.h
@@ -148,13 +148,20 @@ class ByteCodeExprGen : public ConstStmtVisitor<ByteCodeExprGen<Emitter>, bool>,
     return Ctx.classify(Ty);
   }
 
-  /// Classifies a known primitive type
+  /// Classifies a known primitive type.
   PrimType classifyPrim(QualType Ty) const {
     if (auto T = classify(Ty)) {
       return *T;
     }
     llvm_unreachable("not a primitive type");
   }
+  /// Classifies a known primitive expression.
+  PrimType classifyPrim(const Expr *E) const {
+    if (auto T = classify(E))
+      return *T;
+    llvm_unreachable("not a primitive type");
+  }
+
   /// Evaluates an expression and places the result on the stack. If the
   /// expression is of composite type, a local variable will be created
   /// and a pointer to said variable will be placed on the stack.
diff --git a/clang/test/AST/Interp/vectors.cpp b/clang/test/AST/Interp/vectors.cpp
index 6c5d916f51f56..5c4694f122d81 100644
--- a/clang/test/AST/Interp/vectors.cpp
+++ b/clang/test/AST/Interp/vectors.cpp
@@ -1,10 +1,23 @@
 // RUN: %clang_cc1 -fexperimental-new-constant-interpreter -verify=expected,both %s
 // RUN: %clang_cc1 -verify=ref,both %s
 
-// ref-no-diagnostics
-
 typedef int __attribute__((vector_size(16))) VI4;
 constexpr VI4 A = {1,2,3,4};
+static_assert(A[0] == 1, ""); // ref-error {{not an integral constant expression}}
+static_assert(A[1] == 2, ""); // ref-error {{not an integral constant expression}}
+static_assert(A[2] == 3, ""); // ref-error {{not an integral constant expression}}
+static_assert(A[3] == 4, ""); // ref-error {{not an integral constant expression}}
+
+/// VectorSplat casts
+typedef __attribute__(( ext_vector_type(4) )) float float4;
+constexpr float4 vec4_0 = (float4)0.5f;
+static_assert(vec4_0[0] == 0.5, ""); // ref-error {{not an integral constant expression}}
+static_assert(vec4_0[1] == 0.5, ""); // ref-error {{not an integral constant expression}}
+static_assert(vec4_0[2] == 0.5, ""); // ref-error {{not an integral constant expression}}
+static_assert(vec4_0[3] == 0.5, ""); // ref-error {{not an integral constant expression}}
+constexpr int vec4_0_discarded = ((float4)12.0f, 0);
+
+
 
 /// From constant-expression-cxx11.cpp
 namespace Vector {

From 184ba038ac1d444980b3e554b0057f3f30c516ab Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Tue, 16 Apr 2024 10:46:27 -0700
Subject: [PATCH 622/886] [RISCV] Avoid matching 3/5/9 * 2^N as 2^N + 2/4/8
 (e.g. 24) (#88937)

The former is better as a zero extend can be folded into the sll,
whereas the later currently produces a seperate zext.w due to bad
interactions with other combines.
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp |  6 ++++++
 llvm/test/CodeGen/RISCV/addimm-mulimm.ll    |  9 ++++-----
 llvm/test/CodeGen/RISCV/rv64zba.ll          | 22 +++++++++++++++++++++
 3 files changed, 32 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 765838aafb58d..de2ad639f0d6c 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -13416,6 +13416,12 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
     return SDValue();
   uint64_t MulAmt = CNode->getZExtValue();
 
+  // 3/5/9 * 2^N -> shXadd (sll X, C), (sll X, C)
+  // Matched in tablegen, avoid perturbing patterns.
+  for (uint64_t Divisor : {3, 5, 9})
+    if (MulAmt % Divisor == 0 && isPowerOf2_64(MulAmt / Divisor))
+      return SDValue();
+
   // If this is a power 2 + 2/4/8, we can use a shift followed by a single
   // shXadd. First check if this a sum of two power of 2s because that's
   // easy. Then count how many zeros are up to the first bit.
diff --git a/llvm/test/CodeGen/RISCV/addimm-mulimm.ll b/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
index 10103f071462c..48fa69e104565 100644
--- a/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
+++ b/llvm/test/CodeGen/RISCV/addimm-mulimm.ll
@@ -551,9 +551,8 @@ define i64 @add_mul_combine_infinite_loop(i64 %x) {
 ; RV32IMB-NEXT:    sh3add a1, a1, a2
 ; RV32IMB-NEXT:    sh1add a0, a0, a0
 ; RV32IMB-NEXT:    slli a2, a0, 3
-; RV32IMB-NEXT:    li a3, 1
-; RV32IMB-NEXT:    slli a3, a3, 11
-; RV32IMB-NEXT:    sh3add a0, a0, a3
+; RV32IMB-NEXT:    addi a0, a2, 2047
+; RV32IMB-NEXT:    addi a0, a0, 1
 ; RV32IMB-NEXT:    sltu a2, a0, a2
 ; RV32IMB-NEXT:    add a1, a1, a2
 ; RV32IMB-NEXT:    ret
@@ -562,8 +561,8 @@ define i64 @add_mul_combine_infinite_loop(i64 %x) {
 ; RV64IMB:       # %bb.0:
 ; RV64IMB-NEXT:    addi a0, a0, 86
 ; RV64IMB-NEXT:    sh1add a0, a0, a0
-; RV64IMB-NEXT:    slli a0, a0, 3
-; RV64IMB-NEXT:    addi a0, a0, -16
+; RV64IMB-NEXT:    li a1, -16
+; RV64IMB-NEXT:    sh3add a0, a0, a1
 ; RV64IMB-NEXT:    ret
   %tmp0 = mul i64 %x, 24
   %tmp1 = add i64 %tmp0, 2048
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index a84b9e5e7962f..c3c757656be93 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -2490,3 +2490,25 @@ define ptr @test_gep_gep_dont_crash(ptr %p, i64 %a1, i64 %a2) {
   %gep2 = getelementptr i64, ptr %gep1, i64 %a1
   ret ptr %gep2
 }
+
+define i64 @regression(i32 signext %x, i32 signext %y) {
+; RV64I-LABEL: regression:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    subw a0, a0, a1
+; RV64I-NEXT:    slli a0, a0, 32
+; RV64I-NEXT:    li a1, 3
+; RV64I-NEXT:    slli a1, a1, 35
+; RV64I-NEXT:    mulhu a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: regression:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    subw a0, a0, a1
+; RV64ZBA-NEXT:    slli.uw a0, a0, 3
+; RV64ZBA-NEXT:    sh1add a0, a0, a0
+; RV64ZBA-NEXT:    ret
+  %sub = sub i32 %x, %y
+  %ext = zext i32 %sub to i64
+  %res = mul nuw nsw i64 %ext, 24
+  ret i64 %res
+}

From 4082a7554521572a65a5a0008c4661a534df659d Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Tue, 16 Apr 2024 13:48:13 -0400
Subject: [PATCH 623/886] Improve stack usage to increase recursive
 initialization depth (#88546)

We were crashing due to stack exhaustion on rather reasonable C++
template code. After some investigation, I found that we have a
stack-allocated object that was huge: `InitializationSequence` was 7016
bytes. This caused an overflow with deep call stacks in initialization
code.

With these change, `InitializationSequence` is now 248 bytes.

With the original code, testing RelWithDebInfo on Windows 10, all the
tests in SemaCXX took about 6s 800ms. The max template depth I could
reach on my machine using the code in the issue was 708. After that, I
would get `-Wstack-exhausted` warnings until crashing at 976
instantiations.

With these changes on the same machine, all the tests in SemaCXX took
about 6s 500ms. The max template depth I could reach was 1492. After
that, I would get `-Wstack-exhausted` warnings until crashing at 2898
instantiations.

This improves the behavior of #88330 but there's still an outstanding
question of why we run out of stack space and crash in some
circumstances before we're able to issue a diagnostic about stack space
exhaustion.
---
 clang/docs/ReleaseNotes.rst               |  6 ++
 clang/include/clang/Sema/Initialization.h |  6 +-
 clang/include/clang/Sema/Overload.h       | 70 ++++++-----------------
 clang/lib/Sema/SemaInit.cpp               | 26 +++++----
 clang/lib/Sema/SemaOverload.cpp           | 21 +++----
 5 files changed, 54 insertions(+), 75 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index d8ec8bcb8df53..e6c345a2f5c0f 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -203,6 +203,12 @@ Non-comprehensive list of changes in this release
 - ``__typeof_unqual__`` is available in all C modes as an extension, which behaves
   like ``typeof_unqual`` from C23, similar to ``__typeof__`` and ``typeof``.
 
+- Improved stack usage with C++ initialization code. This allows significantly
+  more levels of recursive initialization before reaching stack exhaustion
+  limits. This will positively impact recursive template instantiation code,
+  but should also reduce memory overhead for initializations in general.
+  Fixes #GH88330
+
 New Compiler Flags
 ------------------
 - ``-fsanitize=implicit-bitfield-conversion`` checks implicit truncation and
diff --git a/clang/include/clang/Sema/Initialization.h b/clang/include/clang/Sema/Initialization.h
index 2072cd8d1c3ef..1ceacf0f49f56 100644
--- a/clang/include/clang/Sema/Initialization.h
+++ b/clang/include/clang/Sema/Initialization.h
@@ -1134,7 +1134,7 @@ class InitializationSequence {
   OverloadingResult FailedOverloadResult;
 
   /// The candidate set created when initialization failed.
-  OverloadCandidateSet FailedCandidateSet;
+  std::unique_ptr<OverloadCandidateSet> FailedCandidateSet;
 
   /// The incomplete type that caused a failure.
   QualType FailedIncompleteType;
@@ -1403,7 +1403,9 @@ class InitializationSequence {
   /// Retrieve a reference to the candidate set when overload
   /// resolution fails.
   OverloadCandidateSet &getFailedCandidateSet() {
-    return FailedCandidateSet;
+    assert(FailedCandidateSet &&
+           "this should have been allocated in the constructor!");
+    return *FailedCandidateSet;
   }
 
   /// Get the overloading result, for when the initialization
diff --git a/clang/include/clang/Sema/Overload.h b/clang/include/clang/Sema/Overload.h
index 76311b00d2fc5..e6f88bbf7c4f4 100644
--- a/clang/include/clang/Sema/Overload.h
+++ b/clang/include/clang/Sema/Overload.h
@@ -37,6 +37,7 @@
 #include <cassert>
 #include <cstddef>
 #include <cstdint>
+#include <memory>
 #include <utility>
 
 namespace clang {
@@ -874,7 +875,8 @@ class Sema;
     ConversionFixItGenerator Fix;
 
     /// Viable - True to indicate that this overload candidate is viable.
-    bool Viable : 1;
+    LLVM_PREFERRED_TYPE(bool)
+    unsigned Viable : 1;
 
     /// Whether this candidate is the best viable function, or tied for being
     /// the best viable function.
@@ -883,12 +885,14 @@ class Sema;
     /// was part of the ambiguity kernel: the minimal non-empty set of viable
     /// candidates such that all elements of the ambiguity kernel are better
     /// than all viable candidates not in the ambiguity kernel.
-    bool Best : 1;
+    LLVM_PREFERRED_TYPE(bool)
+    unsigned Best : 1;
 
     /// IsSurrogate - True to indicate that this candidate is a
     /// surrogate for a conversion to a function pointer or reference
     /// (C++ [over.call.object]).
-    bool IsSurrogate : 1;
+    LLVM_PREFERRED_TYPE(bool)
+    unsigned IsSurrogate : 1;
 
     /// IgnoreObjectArgument - True to indicate that the first
     /// argument's conversion, which for this function represents the
@@ -897,18 +901,20 @@ class Sema;
     /// implicit object argument is just a placeholder) or a
     /// non-static member function when the call doesn't have an
     /// object argument.
-    bool IgnoreObjectArgument : 1;
+    LLVM_PREFERRED_TYPE(bool)
+    unsigned IgnoreObjectArgument : 1;
 
     /// True if the candidate was found using ADL.
-    CallExpr::ADLCallKind IsADLCandidate : 1;
+    LLVM_PREFERRED_TYPE(CallExpr::ADLCallKind)
+    unsigned IsADLCandidate : 1;
 
     /// Whether this is a rewritten candidate, and if so, of what kind?
     LLVM_PREFERRED_TYPE(OverloadCandidateRewriteKind)
     unsigned RewriteKind : 2;
 
     /// FailureKind - The reason why this candidate is not viable.
-    /// Actually an OverloadFailureKind.
-    unsigned char FailureKind;
+    LLVM_PREFERRED_TYPE(OverloadFailureKind)
+    unsigned FailureKind : 5;
 
     /// The number of call arguments that were explicitly provided,
     /// to be used while performing partial ordering of function templates.
@@ -972,7 +978,9 @@ class Sema;
   private:
     friend class OverloadCandidateSet;
     OverloadCandidate()
-        : IsSurrogate(false), IsADLCandidate(CallExpr::NotADL), RewriteKind(CRK_None) {}
+        : IsSurrogate(false),
+          IsADLCandidate(static_cast<unsigned>(CallExpr::NotADL)),
+          RewriteKind(CRK_None) {}
   };
 
   /// OverloadCandidateSet - A set of overload candidates, used in C++
@@ -1070,51 +1078,16 @@ class Sema;
     };
 
   private:
-    SmallVector<OverloadCandidate, 16> Candidates;
-    llvm::SmallPtrSet<uintptr_t, 16> Functions;
-
-    // Allocator for ConversionSequenceLists. We store the first few of these
-    // inline to avoid allocation for small sets.
-    llvm::BumpPtrAllocator SlabAllocator;
+    SmallVector<OverloadCandidate, 4> Candidates;
+    llvm::SmallPtrSet<uintptr_t, 4> Functions;
 
     SourceLocation Loc;
     CandidateSetKind Kind;
     OperatorRewriteInfo RewriteInfo;
 
-    constexpr static unsigned NumInlineBytes =
-        24 * sizeof(ImplicitConversionSequence);
-    unsigned NumInlineBytesUsed = 0;
-    alignas(void *) char InlineSpace[NumInlineBytes];
-
     // Address space of the object being constructed.
     LangAS DestAS = LangAS::Default;
 
-    /// If we have space, allocates from inline storage. Otherwise, allocates
-    /// from the slab allocator.
-    /// FIXME: It would probably be nice to have a SmallBumpPtrAllocator
-    /// instead.
-    /// FIXME: Now that this only allocates ImplicitConversionSequences, do we
-    /// want to un-generalize this?
-    template <typename T>
-    T *slabAllocate(unsigned N) {
-      // It's simpler if this doesn't need to consider alignment.
-      static_assert(alignof(T) == alignof(void *),
-                    "Only works for pointer-aligned types.");
-      static_assert(std::is_trivial<T>::value ||
-                        std::is_same<ImplicitConversionSequence, T>::value,
-                    "Add destruction logic to OverloadCandidateSet::clear().");
-
-      unsigned NBytes = sizeof(T) * N;
-      if (NBytes > NumInlineBytes - NumInlineBytesUsed)
-        return SlabAllocator.Allocate<T>(N);
-      char *FreeSpaceStart = InlineSpace + NumInlineBytesUsed;
-      assert(uintptr_t(FreeSpaceStart) % alignof(void *) == 0 &&
-             "Misaligned storage!");
-
-      NumInlineBytesUsed += NBytes;
-      return reinterpret_cast<T *>(FreeSpaceStart);
-    }
-
     void destroyCandidates();
 
   public:
@@ -1163,12 +1136,7 @@ class Sema;
     ConversionSequenceList
     allocateConversionSequences(unsigned NumConversions) {
       ImplicitConversionSequence *Conversions =
-          slabAllocate<ImplicitConversionSequence>(NumConversions);
-
-      // Construct the new objects.
-      for (unsigned I = 0; I != NumConversions; ++I)
-        new (&Conversions[I]) ImplicitConversionSequence();
-
+          new ImplicitConversionSequence[NumConversions];
       return ConversionSequenceList(Conversions, NumConversions);
     }
 
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index fb7a80ab02846..791c0b6e6df23 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -6114,7 +6114,8 @@ InitializationSequence::InitializationSequence(
     Sema &S, const InitializedEntity &Entity, const InitializationKind &Kind,
     MultiExprArg Args, bool TopLevelOfInitList, bool TreatUnavailableAsInvalid)
     : FailedOverloadResult(OR_Success),
-      FailedCandidateSet(Kind.getLocation(), OverloadCandidateSet::CSK_Normal) {
+      FailedCandidateSet(new OverloadCandidateSet(
+          Kind.getLocation(), OverloadCandidateSet::CSK_Normal)) {
   InitializeFrom(S, Entity, Kind, Args, TopLevelOfInitList,
                  TreatUnavailableAsInvalid);
 }
@@ -9735,7 +9736,7 @@ bool InitializationSequence::Diagnose(Sema &S,
     switch (FailedOverloadResult) {
     case OR_Ambiguous:
 
-      FailedCandidateSet.NoteCandidates(
+      FailedCandidateSet->NoteCandidates(
           PartialDiagnosticAt(
               Kind.getLocation(),
               Failure == FK_UserConversionOverloadFailed
@@ -9749,7 +9750,8 @@ bool InitializationSequence::Diagnose(Sema &S,
       break;
 
     case OR_No_Viable_Function: {
-      auto Cands = FailedCandidateSet.CompleteCandidates(S, OCD_AllCandidates, Args);
+      auto Cands =
+          FailedCandidateSet->CompleteCandidates(S, OCD_AllCandidates, Args);
       if (!S.RequireCompleteType(Kind.getLocation(),
                                  DestType.getNonReferenceType(),
                           diag::err_typecheck_nonviable_condition_incomplete,
@@ -9759,13 +9761,13 @@ bool InitializationSequence::Diagnose(Sema &S,
           << OnlyArg->getType() << Args[0]->getSourceRange()
           << DestType.getNonReferenceType();
 
-      FailedCandidateSet.NoteCandidates(S, Args, Cands);
+      FailedCandidateSet->NoteCandidates(S, Args, Cands);
       break;
     }
     case OR_Deleted: {
       OverloadCandidateSet::iterator Best;
-      OverloadingResult Ovl
-        = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
+      OverloadingResult Ovl =
+          FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
 
       StringLiteral *Msg = Best->Function->getDeletedMessage();
       S.Diag(Kind.getLocation(), diag::err_typecheck_deleted_function)
@@ -9949,7 +9951,7 @@ bool InitializationSequence::Diagnose(Sema &S,
     // bad.
     switch (FailedOverloadResult) {
       case OR_Ambiguous:
-        FailedCandidateSet.NoteCandidates(
+        FailedCandidateSet->NoteCandidates(
             PartialDiagnosticAt(Kind.getLocation(),
                                 S.PDiag(diag::err_ovl_ambiguous_init)
                                     << DestType << ArgsRange),
@@ -10003,7 +10005,7 @@ bool InitializationSequence::Diagnose(Sema &S,
           break;
         }
 
-        FailedCandidateSet.NoteCandidates(
+        FailedCandidateSet->NoteCandidates(
             PartialDiagnosticAt(
                 Kind.getLocation(),
                 S.PDiag(diag::err_ovl_no_viable_function_in_init)
@@ -10013,8 +10015,8 @@ bool InitializationSequence::Diagnose(Sema &S,
 
       case OR_Deleted: {
         OverloadCandidateSet::iterator Best;
-        OverloadingResult Ovl
-          = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
+        OverloadingResult Ovl =
+            FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
         if (Ovl != OR_Deleted) {
           S.Diag(Kind.getLocation(), diag::err_ovl_deleted_init)
               << DestType << ArgsRange;
@@ -10093,8 +10095,8 @@ bool InitializationSequence::Diagnose(Sema &S,
     S.Diag(Kind.getLocation(), diag::err_selected_explicit_constructor)
       << Args[0]->getSourceRange();
     OverloadCandidateSet::iterator Best;
-    OverloadingResult Ovl
-      = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
+    OverloadingResult Ovl =
+        FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
     (void)Ovl;
     assert(Ovl == OR_Success && "Inconsistent overload resolution");
     CXXConstructorDecl *CtorDecl = cast<CXXConstructorDecl>(Best->Function);
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 227ef564ba3e0..bcde0d86cf10f 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -1057,8 +1057,7 @@ bool OverloadCandidateSet::OperatorRewriteInfo::shouldAddReversed(
 
 void OverloadCandidateSet::destroyCandidates() {
   for (iterator i = begin(), e = end(); i != e; ++i) {
-    for (auto &C : i->Conversions)
-      C.~ImplicitConversionSequence();
+    delete[] i->Conversions.data();
     if (!i->Viable && i->FailureKind == ovl_fail_bad_deduction)
       i->DeductionFailure.Destroy();
   }
@@ -1066,8 +1065,6 @@ void OverloadCandidateSet::destroyCandidates() {
 
 void OverloadCandidateSet::clear(CandidateSetKind CSK) {
   destroyCandidates();
-  SlabAllocator.Reset();
-  NumInlineBytesUsed = 0;
   Candidates.clear();
   Functions.clear();
   Kind = CSK;
@@ -6983,7 +6980,7 @@ void Sema::AddOverloadCandidate(
   Candidate.RewriteKind =
       CandidateSet.getRewriteInfo().getRewriteKind(Function, PO);
   Candidate.IsSurrogate = false;
-  Candidate.IsADLCandidate = IsADLCandidate;
+  Candidate.IsADLCandidate = static_cast<unsigned>(IsADLCandidate);
   Candidate.IgnoreObjectArgument = false;
   Candidate.ExplicitCallArguments = Args.size();
 
@@ -7815,7 +7812,7 @@ void Sema::AddTemplateOverloadCandidate(
     Candidate.RewriteKind =
       CandidateSet.getRewriteInfo().getRewriteKind(Candidate.Function, PO);
     Candidate.IsSurrogate = false;
-    Candidate.IsADLCandidate = IsADLCandidate;
+    Candidate.IsADLCandidate = static_cast<unsigned>(IsADLCandidate);
     // Ignore the object argument if there is one, since we don't have an object
     // type.
     Candidate.IgnoreObjectArgument =
@@ -14125,7 +14122,8 @@ static ExprResult FinishOverloadedCallExpr(Sema &SemaRef, Scope *S, Expr *Fn,
       return ExprError();
     return SemaRef.BuildResolvedCallExpr(
         Res.get(), FDecl, LParenLoc, Args, RParenLoc, ExecConfig,
-        /*IsExecConfig=*/false, (*Best)->IsADLCandidate);
+        /*IsExecConfig=*/false,
+        static_cast<CallExpr::ADLCallKind>((*Best)->IsADLCandidate));
   }
 
   case OR_No_Viable_Function: {
@@ -14184,7 +14182,8 @@ static ExprResult FinishOverloadedCallExpr(Sema &SemaRef, Scope *S, Expr *Fn,
       return ExprError();
     return SemaRef.BuildResolvedCallExpr(
         Res.get(), FDecl, LParenLoc, Args, RParenLoc, ExecConfig,
-        /*IsExecConfig=*/false, (*Best)->IsADLCandidate);
+        /*IsExecConfig=*/false,
+        static_cast<CallExpr::ADLCallKind>((*Best)->IsADLCandidate));
   }
   }
 
@@ -14491,7 +14490,8 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
       Args[0] = Input;
       CallExpr *TheCall = CXXOperatorCallExpr::Create(
           Context, Op, FnExpr.get(), ArgsArray, ResultTy, VK, OpLoc,
-          CurFPFeatureOverrides(), Best->IsADLCandidate);
+          CurFPFeatureOverrides(),
+          static_cast<CallExpr::ADLCallKind>(Best->IsADLCandidate));
 
       if (CheckCallReturnType(FnDecl->getReturnType(), OpLoc, TheCall, FnDecl))
         return ExprError();
@@ -14909,7 +14909,8 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
         // members; CodeGen should take care not to emit the this pointer.
         TheCall = CXXOperatorCallExpr::Create(
             Context, ChosenOp, FnExpr.get(), Args, ResultTy, VK, OpLoc,
-            CurFPFeatureOverrides(), Best->IsADLCandidate);
+            CurFPFeatureOverrides(),
+            static_cast<CallExpr::ADLCallKind>(Best->IsADLCandidate));
 
         if (const auto *Method = dyn_cast<CXXMethodDecl>(FnDecl);
             Method && Method->isImplicitObjectMemberFunction()) {

From aefff774a0d6f75565243263555f2513ac3c9fdf Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Tue, 16 Apr 2024 21:50:22 +0400
Subject: [PATCH 624/886] [clang] Migrate DR tests to `static_assert` (#88611)

This patch touches a number of tests that run in C++98 mode that have
been using array size as a context that requires a constant expression,
replacing it with a `static_assert` backported via a macro. This reduces
noise in expected directives that comes from diagnostics around VLAs.

This patch also showcases that DR tests would benefit from folding in
constant expressions in C++98 mode, but I'm not sure it's even on the
table. If it is, I'd be happy to prepare a PR for that, and rebase this
PR on top of it.

CC @AaronBallman
---
 clang/test/CXX/drs/dr0xx.cpp  | 14 +++++---
 clang/test/CXX/drs/dr16xx.cpp |  5 ++-
 clang/test/CXX/drs/dr1xx.cpp  | 61 +++++++++++++++++++----------------
 clang/test/CXX/drs/dr2xx.cpp  | 15 ++++++---
 clang/test/CXX/drs/dr3xx.cpp  | 38 ++++++++++++----------
 clang/test/CXX/drs/dr4xx.cpp  | 60 +++++++++++++++++-----------------
 clang/test/CXX/drs/dr5xx.cpp  | 13 +++++---
 clang/test/CXX/drs/dr6xx.cpp  |  6 ++--
 8 files changed, 117 insertions(+), 95 deletions(-)

diff --git a/clang/test/CXX/drs/dr0xx.cpp b/clang/test/CXX/drs/dr0xx.cpp
index a304862885c64..6c600bbc7c3f6 100644
--- a/clang/test/CXX/drs/dr0xx.cpp
+++ b/clang/test/CXX/drs/dr0xx.cpp
@@ -5,6 +5,11 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors -triple %itanium_abi_triple
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors -triple %itanium_abi_triple
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
 namespace cwg1 { // cwg1: no
   namespace X { extern "C" void cwg1_f(int a = 1); }
   namespace Y { extern "C" void cwg1_f(int a = 1); }
@@ -897,7 +902,7 @@ namespace cwg54 { // cwg54: 2.8
 
 namespace cwg55 { // cwg55: yes
   enum E { e = 5 };
-  int test[(e + 1 == 6) ? 1 : -1];
+  static_assert(e + 1 == 6, "");
 }
 
 namespace cwg56 { // cwg56: yes
@@ -1163,10 +1168,9 @@ namespace cwg75 { // cwg75: yes
 
 namespace cwg76 { // cwg76: yes
   const volatile int n = 1;
-  int arr[n]; // #cwg76-vla
-  // expected-error@#cwg76-vla {{variable length arrays in C++ are a Clang extension}}
-  //   expected-note@#cwg76-vla {{read of volatile-qualified type 'const volatile int' is not allowed in a constant expression}}
-  // expected-error@#cwg76-vla {{variable length array declaration not allowed at file scope}}
+  static_assert(n, "");
+  // expected-error@-1 {{static assertion expression is not an integral constant expression}}
+  //   expected-note@-2 {{read of volatile-qualified type 'const volatile int' is not allowed in a constant expression}}
 }
 
 namespace cwg77 { // cwg77: yes
diff --git a/clang/test/CXX/drs/dr16xx.cpp b/clang/test/CXX/drs/dr16xx.cpp
index 6d7bb7619f8b8..cf6b45ceabf2c 100644
--- a/clang/test/CXX/drs/dr16xx.cpp
+++ b/clang/test/CXX/drs/dr16xx.cpp
@@ -153,10 +153,9 @@ namespace cwg1645 { // cwg1645: 3.9
 
 namespace cwg1652 { // cwg1652: 3.6
   int a, b;
-  int arr[&a + 1 == &b ? 1 : 2];
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
+  static_assert(&a + 1 == &b, "");
+  // expected-error@-1 {{static assertion expression is not an integral constant expression}}
   //   expected-note@-2 {{comparison against pointer '&a + 1' that points past the end of a complete object has unspecified value}}
-  // expected-error@-3 {{variable length array declaration not allowed at file scope}}
 }
 
 namespace cwg1653 { // cwg1653: 4 c++17
diff --git a/clang/test/CXX/drs/dr1xx.cpp b/clang/test/CXX/drs/dr1xx.cpp
index 5b497dda047d6..a8f9b705a9866 100644
--- a/clang/test/CXX/drs/dr1xx.cpp
+++ b/clang/test/CXX/drs/dr1xx.cpp
@@ -5,6 +5,17 @@
 // RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-unknown %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-unknown %s -verify=expected,since-cxx11,since-cxx17 -fexceptions -fcxx-exceptions -pedantic-errors
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
+#if __cplusplus == 199711L
+#define __enable_constant_folding(x) (__builtin_constant_p(x) ? (x) : (x))
+#else
+#define __enable_constant_folding
+#endif
+
 namespace cwg100 { // cwg100: yes
   template<const char (*)[4]> struct A {}; // #cwg100-A
   template<const char (&)[4]> struct B {}; // #cwg100-B
@@ -736,8 +747,8 @@ namespace cwg147 { // cwg147: yes
 
 namespace cwg148 { // cwg148: yes
   struct A { int A::*p; };
-  int check1[__is_pod(int(A::*)) ? 1 : -1];
-  int check2[__is_pod(A) ? 1 : -1];
+  static_assert(__is_pod(int(A::*)), "");
+  static_assert(__is_pod(A), "");
 }
 
 // cwg149: na
@@ -745,13 +756,7 @@ namespace cwg148 { // cwg148: yes
 namespace cwg151 { // cwg151: 3.1
   struct X {};
   typedef int X::*p;
-#if __cplusplus < 201103L
-#define fold(x) (__builtin_constant_p(0) ? (x) : (x))
-#else
-#define fold
-#endif
-  int check[fold(p() == 0) ? 1 : -1];
-#undef fold
+  static_assert(__enable_constant_folding(p() == 0), "");
 }
 
 namespace cwg152 { // cwg152: yes
@@ -956,42 +961,42 @@ namespace cwg171 {
 
 namespace cwg172 { // cwg172: yes
   enum { zero };
-  int check1[-1 < zero ? 1 : -1];
+  static_assert(-1 < zero, "");
 
   enum { x = -1, y = (unsigned int)-1 };
-  int check2[sizeof(x) > sizeof(int) ? 1 : -1];
+  static_assert(sizeof(x) > sizeof(int), "");
 
   enum { a = (unsigned int)-1 / 2 };
-  int check3a[sizeof(a) == sizeof(int) ? 1 : -1];
-  int check3b[-a < 0 ? 1 : -1];
+  static_assert(sizeof(a) == sizeof(int), "");
+  static_assert(-a < 0, "");
 
   enum { b = (unsigned int)-1 / 2 + 1 };
-  int check4a[sizeof(b) == sizeof(unsigned int) ? 1 : -1];
-  int check4b[-b > 0 ? 1 : -1];
+  static_assert(sizeof(b) == sizeof(unsigned int), "");
+  static_assert(-b > 0, "");
 
   enum { c = (unsigned long)-1 / 2 };
-  int check5a[sizeof(c) == sizeof(long) ? 1 : -1];
-  int check5b[-c < 0 ? 1 : -1];
+  static_assert(sizeof(c) == sizeof(long), "");
+  static_assert(-c < 0, "");
 
   enum { d = (unsigned long)-1 / 2 + 1 };
-  int check6a[sizeof(d) == sizeof(unsigned long) ? 1 : -1];
-  int check6b[-d > 0 ? 1 : -1];
+  static_assert(sizeof(d) == sizeof(unsigned long), "");
+  static_assert(-d > 0, "");
 
   enum { e = (unsigned long long)-1 / 2 };
   // cxx98-error@-1 {{'long long' is a C++11 extension}}
-  int check7a[sizeof(e) == sizeof(long) ? 1 : -1];
-  int check7b[-e < 0 ? 1 : -1];
+  static_assert(sizeof(e) == sizeof(long), "");
+  static_assert(-e < 0, "");
 
   enum { f = (unsigned long long)-1 / 2 + 1 };
   // cxx98-error@-1 {{'long long' is a C++11 extension}}
-  int check8a[sizeof(f) == sizeof(unsigned long) ? 1 : -1];
-  int check8b[-f > 0 ? 1 : -1];
+  static_assert(sizeof(f) == sizeof(unsigned long), "");
+  static_assert(-f > 0, "");
 }
 
 namespace cwg173 { // cwg173: yes
-  int check[('0' + 1 == '1' && '0' + 2 == '2' && '0' + 3 == '3' &&
-             '0' + 4 == '4' && '0' + 5 == '5' && '0' + 6 == '6' &&
-             '0' + 7 == '7' && '0' + 8 == '8' && '0' + 9 == '9') ? 1 : -1];
+  static_assert('0' + 1 == '1' && '0' + 2 == '2' && '0' + 3 == '3' &&
+                '0' + 4 == '4' && '0' + 5 == '5' && '0' + 6 == '6' &&
+                '0' + 7 == '7' && '0' + 8 == '8' && '0' + 9 == '9', "");
 }
 
 // cwg174: sup 1012
@@ -1070,7 +1075,7 @@ namespace cwg177 { // cwg177: yes
 }
 
 namespace cwg178 { // cwg178: yes
-  int check[int() == 0 ? 1 : -1];
+  static_assert(int() == 0, "");
 #if __cplusplus >= 201103L
   static_assert(int{} == 0, "");
   struct S { int a, b; };
@@ -1180,7 +1185,7 @@ namespace cwg187 { // cwg187: sup 481
 
 namespace cwg188 { // cwg188: yes
   char c[10];
-  int check[sizeof(0, c) == 10 ? 1 : -1];
+  static_assert(sizeof(0, c) == 10, "");
 }
 
 // cwg190 FIXME: add codegen test for tbaa
diff --git a/clang/test/CXX/drs/dr2xx.cpp b/clang/test/CXX/drs/dr2xx.cpp
index e655e7226d51d..5d3e8ce4bea3b 100644
--- a/clang/test/CXX/drs/dr2xx.cpp
+++ b/clang/test/CXX/drs/dr2xx.cpp
@@ -10,10 +10,15 @@
 typedef __SIZE_TYPE__ size_t;
 // cxx98-error@-1 0-1 {{'long long' is a C++11 extension}}
 
-#if __cplusplus < 201103L
-#define fold(x) (__builtin_constant_p(x) ? (x) : (x))
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
+#if __cplusplus == 199711L
+#define __enable_constant_folding(x) (__builtin_constant_p(x) ? (x) : (x))
 #else
-#define fold
+#define __enable_constant_folding
 #endif
 
 namespace cwg200 { // cwg200: dup 214
@@ -31,7 +36,7 @@ namespace cwg200 { // cwg200: dup 214
 namespace cwg202 { // cwg202: 3.1
   template<typename T> T f();
   template<int (*g)()> struct X {
-    int arr[fold(g == &f<int>) ? 1 : -1];
+    static_assert(__enable_constant_folding(g == &f<int>), "");
   };
   template struct X<f>;
 }
@@ -1024,7 +1029,7 @@ namespace cwg275 { // cwg275: no
 namespace cwg277 { // cwg277: 3.1
   typedef int *intp;
   int *p = intp();
-  int a[fold(intp() ? -1 : 1)];
+  static_assert(__enable_constant_folding(!intp()), "");
 }
 
 namespace cwg280 { // cwg280: 2.9
diff --git a/clang/test/CXX/drs/dr3xx.cpp b/clang/test/CXX/drs/dr3xx.cpp
index 6d1c6958ac8eb..3e9228fe21fb6 100644
--- a/clang/test/CXX/drs/dr3xx.cpp
+++ b/clang/test/CXX/drs/dr3xx.cpp
@@ -5,6 +5,17 @@
 // RUN: %clang_cc1 -std=c++11 -verify=expected,cxx98-14,cxx98-17,cxx98-20,cxx11-14,since-cxx11 -triple %itanium_abi_triple %s -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++98 -verify=expected,cxx98-14,cxx98-17,cxx98-20,cxx98 -triple %itanium_abi_triple %s -fexceptions -fcxx-exceptions -pedantic-errors
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
+#if __cplusplus == 199711L
+#define __enable_constant_folding(x) (__builtin_constant_p(x) ? (x) : (x))
+#else
+#define __enable_constant_folding
+#endif
+
 namespace cwg300 { // cwg300: yes
   template<typename R, typename A> void f(R (&)(A)) {}
   int g(int);
@@ -396,7 +407,7 @@ namespace cwg324 { // cwg324: 3.6
 
 namespace cwg326 { // cwg326: 3.1
   struct S {};
-  int test[__is_trivially_constructible(S, const S&) ? 1 : -1];
+  static_assert(__is_trivially_constructible(S, const S&), "");
 }
 
 namespace cwg327 { // cwg327: dup 538
@@ -653,7 +664,7 @@ namespace cwg339 { // cwg339: 2.8
 
   template<typename T> A<sizeof(f(T()))> make_A();
 
-  int a[conv_int<char>::value ? 1 : -1];
+  static_assert(conv_int<char>::value, "");
   bool b = conv_int2<char>(A<1>());
   A<1> c = make_A<char>();
 }
@@ -1099,21 +1110,14 @@ namespace cwg364 { // cwg364: yes
 #endif
 
 namespace cwg367 { // cwg367: yes
-  // FIXME: These diagnostics are terrible. Don't diagnose an ill-formed global
-  // array as being a VLA!
-  int a[true ? throw 0 : 4];
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
-  // expected-error@-2 {{variable length array declaration not allowed at file scope}}
-  int b[true ? 4 : throw 0];
-  // cxx98-error@-1 {{variable length arrays in C++ are a Clang extension}}
-  // cxx98-error@-2 {{variable length array folded to constant array as an extension}}
-  int c[true ? *new int : 4];
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
+  static_assert(__enable_constant_folding(true ? throw 0 : 4), "");
+  // expected-error@-1 {{expression is not an integral constant expression}}
+  static_assert(__enable_constant_folding(true ? 4 : throw 0), "");
+  static_assert(__enable_constant_folding(true ? *new int : 4), "");
+  // expected-error@-1 {{expression is not an integral constant expression}}
   //   expected-note@-2 {{read of uninitialized object is not allowed in a constant expression}}
-  // expected-error@-3 {{variable length array declaration not allowed at file scope}}
-  int d[true ? 4 : *new int];
-  // cxx98-error@-1 {{variable length arrays in C++ are a Clang extension}}
-  // cxx98-error@-2 {{variable length array folded to constant array as an extension}}
+  static_assert(__enable_constant_folding(true ? 4 : *new int), "");
+
 }
 
 namespace cwg368 { // cwg368: 3.6
@@ -1325,7 +1329,7 @@ namespace cwg383 { // cwg383: yes
   struct B { ~B(); };
   union C { C &operator=(const C&); };
   union D { ~D(); };
-  int check[(__is_pod(A) || __is_pod(B) || __is_pod(C) || __is_pod(D)) ? -1 : 1];
+  static_assert(!__is_pod(A) && !__is_pod(B) && !__is_pod(C) && !__is_pod(D), "");
 }
 
 namespace cwg384 { // cwg384: yes
diff --git a/clang/test/CXX/drs/dr4xx.cpp b/clang/test/CXX/drs/dr4xx.cpp
index 611b791470785..07162cc28f6b6 100644
--- a/clang/test/CXX/drs/dr4xx.cpp
+++ b/clang/test/CXX/drs/dr4xx.cpp
@@ -6,6 +6,11 @@
 // RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++23 %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: env ASAN_OPTIONS=detect_stack_use_after_return=0 %clang_cc1 -std=c++2c %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
 // FIXME: __SIZE_TYPE__ expands to 'long long' on some targets.
 __extension__ typedef __SIZE_TYPE__ size_t;
 
@@ -217,7 +222,7 @@ namespace cwg407 { // cwg407: 3.8
 }
 
 namespace cwg408 { // cwg408: 3.4
-  template<int N> void g() { int arr[N != 1 ? 1 : -1]; }
+  template<int N> void g() { static_assert(N != 1, ""); }
   template<> void g<2>() { }
 
   template<typename T> struct S {
@@ -239,7 +244,7 @@ namespace cwg408 { // cwg408: 3.4
   };
   template<typename T> int R<T>::arr[1];
   template<typename T> void R<T>::f() {
-    int arr[sizeof(arr) != sizeof(int) ? 1 : -1];
+    static_assert(sizeof(arr) != sizeof(int), "");
   }
   template<> int R<int>::arr[2];
   template void R<int>::f();
@@ -842,11 +847,10 @@ namespace cwg451 { // cwg451: yes
   // expected-warning@-1 {{division by zero is undefined}}
   const int b = 1 / 0; // #cwg451-b
   // expected-warning@-1 {{division by zero is undefined}}
-  int arr[b]; // #cwg451-arr
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
+  static_assert(b, "");
+  // expected-error@-1 {{expression is not an integral constant expression}}
   //   expected-note@-2 {{initializer of 'b' is not a constant expression}}
   //   expected-note@#cwg451-b {{declared here}}
-  // expected-error@#cwg451-arr {{variable length array declaration not allowed at file scope}}
 }
 
 namespace cwg452 { // cwg452: yes
@@ -876,11 +880,10 @@ namespace cwg456 { // cwg456: yes
 namespace cwg457 { // cwg457: yes
   const int a = 1;
   const volatile int b = 1;
-  int ax[a];
-  int bx[b];
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
+  static_assert(a, "");
+  static_assert(b, "");
+  // expected-error@-1 {{expression is not an integral constant expression}}
   //   expected-note@-2 {{read of volatile-qualified type 'const volatile int' is not allowed in a constant expression}}
-  // expected-error@-3 {{variable length array declaration not allowed at file scope}}
 
   enum E {
     ea = a,
@@ -1276,20 +1279,18 @@ namespace cwg482 { // cwg482: 3.5
 
 namespace cwg483 { // cwg483: yes
   namespace climits {
-    int check1[__SCHAR_MAX__ >= 127 ? 1 : -1];
-    int check2[__SHRT_MAX__ >= 32767 ? 1 : -1];
-    int check3[__INT_MAX__ >= 32767 ? 1 : -1];
-    int check4[__LONG_MAX__ >= 2147483647 ? 1 : -1];
-    int check5[__LONG_LONG_MAX__ >= 9223372036854775807 ? 1 : -1];
-    // cxx98-error@-1 {{'long long' is a C++11 extension}}
-    // cxx98-error@-2 0-1{{'long long' is a C++11 extension}}
+    static_assert(__SCHAR_MAX__ >= 127, "");
+    static_assert(__SHRT_MAX__ >= 32767, "");
+    static_assert(__INT_MAX__ >= 32767, "");
+    static_assert(__LONG_MAX__ >= 2147483647, "");
+    static_assert(__LONG_LONG_MAX__ >= 9223372036854775807, "");
   }
   namespace cstdint {
-    int check1[__PTRDIFF_WIDTH__ >= 16 ? 1 : -1];
-    int check2[__SIG_ATOMIC_WIDTH__ >= 8 ? 1 : -1];
-    int check3[__SIZE_WIDTH__ >= 16 ? 1 : -1];
-    int check4[__WCHAR_WIDTH__ >= 8 ? 1 : -1];
-    int check5[__WINT_WIDTH__ >= 16 ? 1 : -1];
+    static_assert(__PTRDIFF_WIDTH__ >= 16, "");
+    static_assert(__SIG_ATOMIC_WIDTH__ >= 8, "");
+    static_assert(__SIZE_WIDTH__ >= 16, "");
+    static_assert(__WCHAR_WIDTH__ >= 8, "");
+    static_assert(__WINT_WIDTH__ >= 16, "");
   }
 }
 
@@ -1366,11 +1367,10 @@ namespace cwg486 { // cwg486: yes
 namespace cwg487 { // cwg487: yes
   enum E { e };
   int operator+(int, E); // #cwg487-operator-plus
-  int i[4 + e]; // #cwg487-i
-  // expected-error@-1 {{variable length arrays in C++ are a Clang extension}}
+  static_assert(4 + e, "");
+  // expected-error@-1 {{expression is not an integral constant expression}}
   //   since-cxx11-note@-2 {{non-constexpr function 'operator+' cannot be used in a constant expression}}
   //   since-cxx11-note@#cwg487-operator-plus {{declared here}}
-  // expected-error@#cwg487-i {{variable length array declaration not allowed at file scope}}
 }
 
 namespace cwg488 { // cwg488: yes c++11
@@ -1485,13 +1485,13 @@ namespace cwg495 { // cwg495: 3.5
 namespace cwg496 { // cwg496: sup 2094
   struct A { int n; };
   struct B { volatile int n; };
-  int check1[ __is_trivially_copyable(const int) ? 1 : -1];
+  static_assert(__is_trivially_copyable(const int), "");
   // This checks the cwg2094 behavior, not cwg496
-  int check2[ __is_trivially_copyable(volatile int) ? 1 : -1];
-  int check3[ __is_trivially_constructible(A, const A&) ? 1 : -1];
-  int check4[ __is_trivially_constructible(B, const B&) ? 1 : -1];
-  int check5[ __is_trivially_assignable(A, const A&) ? 1 : -1];
-  int check6[ __is_trivially_assignable(B, const B&) ? 1 : -1];
+  static_assert(__is_trivially_copyable(volatile int), "");
+  static_assert(__is_trivially_constructible(A, const A&), "");
+  static_assert(__is_trivially_constructible(B, const B&), "");
+  static_assert(__is_trivially_assignable(A, const A&), "");
+  static_assert(__is_trivially_assignable(B, const B&), "");
 }
 
 namespace cwg497 { // cwg497: sup 253
diff --git a/clang/test/CXX/drs/dr5xx.cpp b/clang/test/CXX/drs/dr5xx.cpp
index 0fe64102d70b0..9d890f981348a 100644
--- a/clang/test/CXX/drs/dr5xx.cpp
+++ b/clang/test/CXX/drs/dr5xx.cpp
@@ -5,6 +5,11 @@
 // RUN: %clang_cc1 -std=c++20 %s -verify=expected,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 // RUN: %clang_cc1 -std=c++23 %s -verify=expected,since-cxx23,since-cxx20,since-cxx17,since-cxx11 -fexceptions -fcxx-exceptions -pedantic-errors
 
+#if __cplusplus == 199711L
+#define static_assert(...) __extension__ _Static_assert(__VA_ARGS__)
+// cxx98-error@-1 {{variadic macros are a C99 feature}}
+#endif
+
 // FIXME: This is included to avoid a diagnostic with no source location
 // pointing at the implicit operator new. We can't match such a diagnostic
 // with -verify.
@@ -819,7 +824,7 @@ namespace cwg565 { // cwg565: yes
 
 namespace cwg566 { // cwg566: yes
 #if __cplusplus >= 201103L
-  int check[int(-3.99) == -3 ? 1 : -1];
+  static_assert(int(-3.99) == -3, "");
 #endif
 }
 
@@ -834,7 +839,7 @@ namespace cwg568 { // cwg568: 3.0 c++11
   public:
     int n;
   };
-  int check_trivial[__is_trivial(trivial) ? 1 : -1];
+  static_assert(__is_trivial(trivial), "");
 
   struct std_layout {
     std_layout();
@@ -843,7 +848,7 @@ namespace cwg568 { // cwg568: 3.0 c++11
   private:
     int n;
   };
-  int check_std_layout[__is_standard_layout(std_layout) ? 1 : -1];
+  static_assert(__is_standard_layout(std_layout), "");
 
   struct aggregate {
     int x;
@@ -885,7 +890,7 @@ namespace cwg570 { // cwg570: dup 633
 
 namespace cwg572 { // cwg572: yes
   enum E { a = 1, b = 2 };
-  int check[a + b == 3 ? 1 : -1];
+  static_assert(a + b == 3, "");
 }
 
 namespace cwg573 { // cwg573: no
diff --git a/clang/test/CXX/drs/dr6xx.cpp b/clang/test/CXX/drs/dr6xx.cpp
index 9d3613ae8589e..069102d9c5975 100644
--- a/clang/test/CXX/drs/dr6xx.cpp
+++ b/clang/test/CXX/drs/dr6xx.cpp
@@ -144,7 +144,7 @@ namespace cwg608 { // cwg608: yes
   struct D : B, C {};
 }
 
-int cwg610[-0u == 0u ? 1 : -1]; // cwg610: yes
+static_assert(-0u == 0u, ""); // cwg610: yes
 
 namespace cwg611 { // cwg611: yes
   int k;
@@ -190,8 +190,8 @@ namespace cwg613 { // cwg613: yes c++11
   }
 }
 
-int cwg614_a[(-1) / 2 == 0 ? 1 : -1]; // cwg614: yes
-int cwg614_b[(-1) % 2 == -1 ? 1 : -1];
+static_assert((-1) / 2 == 0, ""); // cwg614: yes
+static_assert((-1) % 2 == -1, "");
 
 namespace cwg615 { // cwg615: yes
   int f();

From 6b83fe552990966fdad0e5693a79b02b87d9526e Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Tue, 16 Apr 2024 11:03:53 -0700
Subject: [PATCH 625/886] [RISCV] Strength reduce mul by 2^n + 2/4/8 + 1
 (#88911)

With zba, we can expand this to (add (shl X, C1), (shXadd X, X)).

Note that this is our first expansion to a three instruction sequence. I
believe this to general be a reasonable tradeoff for most architectures,
but we may want to (someday) consider a tuning flag here.

I plan to support 2^n + (2/4/8 + 1) eventually as well, but that comes
behind 2^N - 2^M. Both are also three instruction sequences.

---------

Co-authored-by: Min-Yih Hsu <min@myhsu.dev>
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 37 +++++++++++++++
 llvm/test/CodeGen/RISCV/rv64zba.ll          | 51 +++++++++++++++------
 2 files changed, 73 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index de2ad639f0d6c..dc7c6f83b9857 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -13437,6 +13437,43 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
       return DAG.getNode(ISD::ADD, DL, VT, Shift1, Shift2);
     }
   }
+
+  // 2^(1,2,3) * 3,5,9 + 1 -> (shXadd (shYadd x, x), x)
+  // Matched in tablegen, avoid perturbing patterns.
+  switch (MulAmt) {
+  case 11:
+  case 13:
+  case 19:
+  case 21:
+  case 25:
+  case 27:
+  case 29:
+  case 37:
+  case 41:
+  case 45:
+  case 73:
+  case 91:
+    return SDValue();
+  default:
+    break;
+  }
+
+  // 2^n + 2/4/8 + 1 -> (add (shl X, C1), (shXadd X, X))
+  if (MulAmt > 2 && isPowerOf2_64((MulAmt - 1) & (MulAmt - 2))) {
+    unsigned ScaleShift = llvm::countr_zero(MulAmt - 1);
+    if (ScaleShift >= 1 && ScaleShift < 4) {
+      unsigned ShiftAmt = Log2_64(((MulAmt - 1) & (MulAmt - 2)));
+      SDLoc DL(N);
+      SDValue Shift1 = DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0),
+                                   DAG.getConstant(ShiftAmt, DL, VT));
+      SDValue Shift2 = DAG.getNode(ISD::SHL, DL, VT, N->getOperand(0),
+                                   DAG.getConstant(ScaleShift, DL, VT));
+      return DAG.getNode(
+          ISD::ADD, DL, VT, Shift1,
+          DAG.getNode(ISD::ADD, DL, VT, Shift2, N->getOperand(0)));
+    }
+  }
+
   return SDValue();
 }
 
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index c3c757656be93..b4c80b60e0bad 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -598,31 +598,52 @@ define i64 @mul125(i64 %a) {
 }
 
 define i64 @mul131(i64 %a) {
-; CHECK-LABEL: mul131:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 131
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul131:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 131
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul131:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    sh1add a1, a0, a0
+; RV64ZBA-NEXT:    slli a0, a0, 7
+; RV64ZBA-NEXT:    add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 131
   ret i64 %c
 }
 
 define i64 @mul133(i64 %a) {
-; CHECK-LABEL: mul133:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 133
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul133:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 133
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul133:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    sh2add a1, a0, a0
+; RV64ZBA-NEXT:    slli a0, a0, 7
+; RV64ZBA-NEXT:    add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 133
   ret i64 %c
 }
 
 define i64 @mul137(i64 %a) {
-; CHECK-LABEL: mul137:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    li a1, 137
-; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    ret
+; RV64I-LABEL: mul137:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 137
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul137:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    sh3add a1, a0, a0
+; RV64ZBA-NEXT:    slli a0, a0, 7
+; RV64ZBA-NEXT:    add a0, a0, a1
+; RV64ZBA-NEXT:    ret
   %c = mul i64 %a, 137
   ret i64 %c
 }

From 1c2afbae9af22b58190c10e3517242d01d89d612 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Tue, 16 Apr 2024 11:05:45 -0700
Subject: [PATCH 626/886] [CodeGen,test] Test llvm-libc style alias attribute
 with UsingShadowDecl

The pattern is quite involved and deserves a specific codegen test.
This test would catch the bug in the first attempt of #87130
---
 clang/test/CodeGen/alias.cpp | 25 ++++++++++++++++++++-----
 1 file changed, 20 insertions(+), 5 deletions(-)

diff --git a/clang/test/CodeGen/alias.cpp b/clang/test/CodeGen/alias.cpp
index 17c1e1ae32f03..a468c31d369ed 100644
--- a/clang/test/CodeGen/alias.cpp
+++ b/clang/test/CodeGen/alias.cpp
@@ -1,27 +1,42 @@
-// RUN: %clang_cc1 -triple x86_64-linux -verify -emit-llvm-only %s
-// RUN: not %clang_cc1 -triple x86_64-linux -emit-llvm-only -fdiagnostics-parseable-fixits %s 2>&1 | FileCheck %s
+// RUN: %clang_cc1 -triple x86_64-linux -verify -emit-llvm-only -DERR %s
+// RUN: not %clang_cc1 -triple x86_64-linux -emit-llvm-only -fdiagnostics-parseable-fixits -DERR %s 2>&1 | FileCheck %s --check-prefix=FIXIT
+// RUN: %clang_cc1 -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s
 
+#ifdef ERR
 void *f1_ifunc(void) { return nullptr; }
 void f1(void) __attribute__((alias("f1_ifunc")));
 // expected-error@-1 {{alias must point to a defined variable or function}}
 // expected-note@-2 {{must refer to its mangled name}}
 // expected-note@-3 {{function by that name is mangled as}}
-// CHECK: fix-it:"{{.*}}":{[[@LINE-4]]:30-[[@LINE-4]]:47}:"alias(\"_Z8f1_ifuncv\")"
+// FIXIT: fix-it:"{{.*}}":{[[@LINE-4]]:30-[[@LINE-4]]:47}:"alias(\"_Z8f1_ifuncv\")"
 
 void *f6_resolver_resolver(void) { return 0; }
 void *f6_resolver(void) __attribute__((alias("f6_resolver_resolver")));
 // expected-error@-1 {{alias must point to a defined variable or function}}
 // expected-note@-2 {{must refer to its mangled name}}
 // expected-note@-3 {{function by that name is mangled as}}
-// CHECK: fix-it:"{{.*}}":{[[@LINE-4]]:40-[[@LINE-4]]:69}:"alias(\"_Z20f6_resolver_resolverv\")"
+// FIXIT: fix-it:"{{.*}}":{[[@LINE-4]]:40-[[@LINE-4]]:69}:"alias(\"_Z20f6_resolver_resolverv\")"
 void f6(void) __attribute__((alias("f6_resolver")));
 // expected-error@-1 {{alias must point to a defined variable or function}}
 // expected-note@-2 {{must refer to its mangled name}}
 // expected-note@-3 {{function by that name is mangled as}}
-// CHECK: fix-it:"{{.*}}":{[[@LINE-4]]:30-[[@LINE-4]]:50}:"alias(\"_Z11f6_resolverv\")"
+// FIXIT: fix-it:"{{.*}}":{[[@LINE-4]]:30-[[@LINE-4]]:50}:"alias(\"_Z11f6_resolverv\")"
 
 __attribute__((unused, alias("resolver"), deprecated("hahahaha, isn't C great?")))
 void func();
 // expected-error@-2 {{alias must point to a defined variable or function}}
 // expected-note@-3 {{must refer to its mangled name}}
+#endif
 
+// CHECK: @_ZN4libc4log2Ed ={{.*}} alias double (double), ptr @log2
+// CHECK: define{{.*}} @log2(
+namespace libc { double log2(double x); }
+extern "C" double log2(double);
+namespace std { using ::log2; }
+using std::log2;
+
+namespace libc {
+decltype(libc::log2) __log2_impl__ __asm__("log2");
+decltype(libc::log2) log2 [[gnu::alias("log2")]];
+double __log2_impl__(double x) { return x; }
+}

From 5462b27026dee886fb896980d6ad9487200a6cbe Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:07:03 +0200
Subject: [PATCH 627/886] [NFC][libc++][TZDB] Refactors argument order.
 (#85781)

Putting the output reference argument first looks more sensible.
---
 libcxx/include/__chrono/formatter.h | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)

diff --git a/libcxx/include/__chrono/formatter.h b/libcxx/include/__chrono/formatter.h
index 4ad59382a4148..b64cae529a294 100644
--- a/libcxx/include/__chrono/formatter.h
+++ b/libcxx/include/__chrono/formatter.h
@@ -79,7 +79,7 @@ namespace __formatter {
 // small). Therefore a duration uses its own conversion.
 template <class _CharT, class _Rep, class _Period>
 _LIBCPP_HIDE_FROM_ABI void
-__format_sub_seconds(const chrono::duration<_Rep, _Period>& __value, basic_stringstream<_CharT>& __sstr) {
+__format_sub_seconds(basic_stringstream<_CharT>& __sstr, const chrono::duration<_Rep, _Period>& __value) {
   __sstr << std::use_facet<numpunct<_CharT>>(__sstr.getloc()).decimal_point();
 
   using __duration = chrono::duration<_Rep, _Period>;
@@ -110,13 +110,13 @@ __format_sub_seconds(const chrono::duration<_Rep, _Period>& __value, basic_strin
 }
 
 template <class _CharT, __is_time_point _Tp>
-_LIBCPP_HIDE_FROM_ABI void __format_sub_seconds(const _Tp& __value, basic_stringstream<_CharT>& __sstr) {
-  __formatter::__format_sub_seconds(__value.time_since_epoch(), __sstr);
+_LIBCPP_HIDE_FROM_ABI void __format_sub_seconds(basic_stringstream<_CharT>& __sstr, const _Tp& __value) {
+  __formatter::__format_sub_seconds(__sstr, __value.time_since_epoch());
 }
 
 template <class _CharT, class _Duration>
 _LIBCPP_HIDE_FROM_ABI void
-__format_sub_seconds(const chrono::hh_mm_ss<_Duration>& __value, basic_stringstream<_CharT>& __sstr) {
+__format_sub_seconds(basic_stringstream<_CharT>& __sstr, const chrono::hh_mm_ss<_Duration>& __value) {
   __sstr << std::use_facet<numpunct<_CharT>>(__sstr.getloc()).decimal_point();
   if constexpr (chrono::treat_as_floating_point_v<typename _Duration::rep>)
     std::format_to(std::ostreambuf_iterator<_CharT>{__sstr},
@@ -143,7 +143,7 @@ consteval bool __use_fraction() {
 }
 
 template <class _CharT>
-_LIBCPP_HIDE_FROM_ABI void __format_year(int __year, basic_stringstream<_CharT>& __sstr) {
+_LIBCPP_HIDE_FROM_ABI void __format_year(basic_stringstream<_CharT>& __sstr, int __year) {
   if (__year < 0) {
     __sstr << _CharT('-');
     __year = -__year;
@@ -159,7 +159,7 @@ _LIBCPP_HIDE_FROM_ABI void __format_year(int __year, basic_stringstream<_CharT>&
 }
 
 template <class _CharT>
-_LIBCPP_HIDE_FROM_ABI void __format_century(int __year, basic_stringstream<_CharT>& __sstr) {
+_LIBCPP_HIDE_FROM_ABI void __format_century(basic_stringstream<_CharT>& __sstr, int __year) {
   // TODO FMT Write an issue
   // [tab:time.format.spec]
   //   %C The year divided by 100 using floored division. If the result is a
@@ -172,7 +172,7 @@ _LIBCPP_HIDE_FROM_ABI void __format_century(int __year, basic_stringstream<_Char
 
 template <class _CharT, class _Tp>
 _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
-    const _Tp& __value, basic_stringstream<_CharT>& __sstr, basic_string_view<_CharT> __chrono_specs) {
+    basic_stringstream<_CharT>& __sstr, const _Tp& __value, basic_string_view<_CharT> __chrono_specs) {
   tm __t              = std::__convert_to_tm<tm>(__value);
   const auto& __facet = std::use_facet<time_put<_CharT>>(__sstr.getloc());
   for (auto __it = __chrono_specs.begin(); __it != __chrono_specs.end(); ++__it) {
@@ -196,7 +196,7 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
         // strftime's output is only defined in the range [00, 99].
         int __year = __t.tm_year + 1900;
         if (__year < 1000 || __year > 9999)
-          __formatter::__format_century(__year, __sstr);
+          __formatter::__format_century(__sstr, __year);
         else
           __facet.put(
               {__sstr}, __sstr, _CharT(' '), std::addressof(__t), std::to_address(__s), std::to_address(__it + 1));
@@ -242,7 +242,7 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
         __facet.put(
             {__sstr}, __sstr, _CharT(' '), std::addressof(__t), std::to_address(__s), std::to_address(__it + 1));
         if constexpr (__use_fraction<_Tp>())
-          __formatter::__format_sub_seconds(__value, __sstr);
+          __formatter::__format_sub_seconds(__sstr, __value);
         break;
 
         // Unlike time_put and strftime the formatting library requires %Y
@@ -283,13 +283,13 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
         // Depending on the platform's libc the range of supported years is
         // limited. Intead of of testing all conditions use the internal
         // implementation unconditionally.
-        __formatter::__format_year(__t.tm_year + 1900, __sstr);
+        __formatter::__format_year(__sstr, __t.tm_year + 1900);
         break;
 
       case _CharT('F'): {
         int __year = __t.tm_year + 1900;
         if (__year < 1000) {
-          __formatter::__format_year(__year, __sstr);
+          __formatter::__format_year(__sstr, __year);
           __sstr << std::format(_LIBCPP_STATICALLY_WIDEN(_CharT, "-{:02}-{:02}"), __t.tm_mon + 1, __t.tm_mday);
         } else
           __facet.put(
@@ -310,7 +310,7 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
             ++__it;
             __facet.put(
                 {__sstr}, __sstr, _CharT(' '), std::addressof(__t), std::to_address(__s), std::to_address(__it + 1));
-            __formatter::__format_sub_seconds(__value, __sstr);
+            __formatter::__format_sub_seconds(__sstr, __value);
             break;
           }
         }
@@ -512,7 +512,7 @@ __format_chrono(const _Tp& __value,
     if constexpr (chrono::__is_duration<_Tp>::value) {
       if (__value < __value.zero())
         __sstr << _CharT('-');
-      __formatter::__format_chrono_using_chrono_specs(chrono::abs(__value), __sstr, __chrono_specs);
+      __formatter::__format_chrono_using_chrono_specs(__sstr, chrono::abs(__value), __chrono_specs);
       // TODO FMT When keeping the precision it will truncate the string.
       // Note that the behaviour what the precision does isn't specified.
       __specs.__precision_ = -1;
@@ -556,7 +556,7 @@ __format_chrono(const _Tp& __value,
           __sstr << _CharT('-');
       }
 
-      __formatter::__format_chrono_using_chrono_specs(__value, __sstr, __chrono_specs);
+      __formatter::__format_chrono_using_chrono_specs(__sstr, __value, __chrono_specs);
     }
   }
 

From a75c9d059791f5d175f6c263d114d59e51b46120 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:18:34 +0200
Subject: [PATCH 628/886] [NFC][libc++] Moves ios_base's forward declaration.
 (#88027)

According to our synopsis it belonged to ios_fwd. This is not true in
the C++11 version of the Standard, I did not validate against C++98.

Moving this to ios's forward where it's declared in the standard allows
removing a module quirk. An earlier removal of std::vectors forward
declaration allows to remove all quirks for the iosfwd module part.

Since iosfwd includes __fwd/ios.h this does not change the required
includes.
---
 libcxx/include/__fwd/ios.h          | 2 ++
 libcxx/include/iosfwd               | 3 ---
 libcxx/utils/libcxx/test/modules.py | 2 --
 3 files changed, 2 insertions(+), 5 deletions(-)

diff --git a/libcxx/include/__fwd/ios.h b/libcxx/include/__fwd/ios.h
index 82c865d58cc75..48350709d4ce2 100644
--- a/libcxx/include/__fwd/ios.h
+++ b/libcxx/include/__fwd/ios.h
@@ -18,6 +18,8 @@
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
+class _LIBCPP_EXPORTED_FROM_ABI ios_base;
+
 template <class _CharT, class _Traits = char_traits<_CharT> >
 class _LIBCPP_TEMPLATE_VIS basic_ios;
 
diff --git a/libcxx/include/iosfwd b/libcxx/include/iosfwd
index 9af5e05031850..2481667dd972c 100644
--- a/libcxx/include/iosfwd
+++ b/libcxx/include/iosfwd
@@ -25,7 +25,6 @@ template<>            struct char_traits<wchar_t>;
 
 template<class T>     class allocator;
 
-class ios_base;
 template <class charT, class traits = char_traits<charT> > class basic_ios;
 
 template <class charT, class traits = char_traits<charT> > class basic_streambuf;
@@ -124,8 +123,6 @@ using wosyncstream = basic_osyncstream<wchar_t>;  // C++20
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-class _LIBCPP_EXPORTED_FROM_ABI ios_base;
-
 template <class _CharT, class _Traits = char_traits<_CharT> >
 class _LIBCPP_TEMPLATE_VIS istreambuf_iterator;
 template <class _CharT, class _Traits = char_traits<_CharT> >
diff --git a/libcxx/utils/libcxx/test/modules.py b/libcxx/utils/libcxx/test/modules.py
index 3f3c7999a1a21..44c6292ff1140 100644
--- a/libcxx/utils/libcxx/test/modules.py
+++ b/libcxx/utils/libcxx/test/modules.py
@@ -26,8 +26,6 @@
 # The operators are added for private types like __iom_t10.
 SkipDeclarations["iomanip"] = ["std::operator<<", "std::operator>>"]
 
-SkipDeclarations["iosfwd"] = ["std::ios_base", "std::vector"]
-
 # This header also provides declarations in the namespace that might be
 # an error.
 SkipDeclarations["filesystem"] = [

From 9cd3e92f05fcc2c9168a7abc56d08f0d33bfdfdf Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:19:53 +0200
Subject: [PATCH 629/886] [libc++][modules] Removes some validation quirks.
 (#88031)

Recent unrelated header cleanups caused these quirks to become obsolete.
---
 libcxx/utils/libcxx/test/modules.py | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/libcxx/utils/libcxx/test/modules.py b/libcxx/utils/libcxx/test/modules.py
index 44c6292ff1140..aab7651c7bb03 100644
--- a/libcxx/utils/libcxx/test/modules.py
+++ b/libcxx/utils/libcxx/test/modules.py
@@ -52,8 +52,6 @@
     "std::operator==",
 ]
 
-# Declared in the forward header since std::string uses std::allocator
-SkipDeclarations["string"] = ["std::allocator"]
 # TODO MODULES remove zombie names
 # https://libcxx.llvm.org/Status/Cxx20.html#note-p0619
 SkipDeclarations["memory"] = [
@@ -61,9 +59,6 @@
     "std::get_temporary_buffer",
 ]
 
-# TODO MODULES this should be part of ios instead
-SkipDeclarations["streambuf"] = ["std::basic_ios"]
-
 # include/__type_traits/is_swappable.h
 SkipDeclarations["type_traits"] = [
     "std::swap",

From 41a830500aa5556a65198607ec751d8e3254c949 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:20:37 +0200
Subject: [PATCH 630/886] [libc++] Removes deprecated
 _LIBCPP_ENABLE_<VERSION>_REMOVED_FEATURES macros (#88548)

We marked those macros as deprecated in the last release with the intent
of
removing them in LLVM 19. This commit performs the removal.
---
 libcxx/docs/ReleaseNotes/19.rst               |  2 +-
 libcxx/docs/UsingLibcxx.rst                   | 12 -------
 libcxx/include/__config                       | 26 --------------
 ...le_removed_cpp17_features.compile.pass.cpp | 36 -------------------
 ...moved_cpp17_features.deprecated.verify.cpp | 20 -----------
 ...moved_cpp20_features.deprecated.verify.cpp | 20 -----------
 6 files changed, 1 insertion(+), 115 deletions(-)
 delete mode 100644 libcxx/test/libcxx/depr/enable_removed_cpp17_features.compile.pass.cpp
 delete mode 100644 libcxx/test/libcxx/depr/enable_removed_cpp17_features.deprecated.verify.cpp
 delete mode 100644 libcxx/test/libcxx/depr/enable_removed_cpp20_features.deprecated.verify.cpp

diff --git a/libcxx/docs/ReleaseNotes/19.rst b/libcxx/docs/ReleaseNotes/19.rst
index 45aac88e45502..53cc7a77d1af4 100644
--- a/libcxx/docs/ReleaseNotes/19.rst
+++ b/libcxx/docs/ReleaseNotes/19.rst
@@ -97,7 +97,7 @@ Deprecations and Removals
 - The ``_LIBCPP_ENABLE_CXX20_REMOVED_ALLOCATOR_MEMBERS`` and ``_LIBCPP_ENABLE_CXX20_REMOVED_ALLOCATOR_VOID_SPECIALIZATION``
   macros have been removed in LLVM 19.
 
-- TODO: The ``_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES`` and ``_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES`` macros have
+- The ``_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES`` and ``_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES`` macros have
   been removed in LLVM 19. C++17 and C++20 removed features can still be re-enabled individually.
 
 - The ``_LIBCPP_INLINE_VISIBILITY`` and ``_VSTD`` macros have been removed in LLVM 19.
diff --git a/libcxx/docs/UsingLibcxx.rst b/libcxx/docs/UsingLibcxx.rst
index c0e85ad4d5e24..8f945656de1ca 100644
--- a/libcxx/docs/UsingLibcxx.rst
+++ b/libcxx/docs/UsingLibcxx.rst
@@ -208,12 +208,6 @@ safety annotations.
 
 C++17 Specific Configuration Macros
 -----------------------------------
-**_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES**:
-  This macro is used to re-enable all the features removed in C++17. The effect
-  is equivalent to manually defining each macro listed below.
-  This macro is deprecated and will be removed in LLVM-19. Use the
-  individual macros listed below.
-
 **_LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR**:
   This macro is used to re-enable `auto_ptr`.
 
@@ -238,12 +232,6 @@ C++20 Specific Configuration Macros
   This macro is used to re-enable the function
   ``std::shared_ptr<...>::unique()``.
 
-**_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES**:
-  This macro is used to re-enable all the features removed in C++20. The effect
-  is equivalent to manually defining each macro listed below.
-  This macro is deprecated and will be removed in LLVM-19. Use the
-  individual macros listed below.
-
 **_LIBCPP_ENABLE_CXX20_REMOVED_BINDER_TYPEDEFS**:
   This macro is used to re-enable the `argument_type`, `result_type`,
   `first_argument_type`, and `second_argument_type` members of class
diff --git a/libcxx/include/__config b/libcxx/include/__config
index 9b4155af1e3c6..4ccef2ca0d73b 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -16,17 +16,6 @@
 #  pragma GCC system_header
 #endif
 
-#if defined(_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES) && !defined(_LIBCPP_DISABLE_DEPRECATION_WARNINGS)
-#  pragma clang deprecated(                                                                                            \
-      _LIBCPP_ENABLE_CXX17_REMOVED_FEATURES,                                                                           \
-      "_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES is deprecated in LLVM 18 and will be removed in LLVM 19")
-#endif
-#if defined(_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES) && !defined(_LIBCPP_DISABLE_DEPRECATION_WARNINGS)
-#  pragma clang deprecated(                                                                                            \
-      _LIBCPP_ENABLE_CXX20_REMOVED_FEATURES,                                                                           \
-      "_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES is deprecated in LLVM 18 and will be removed in LLVM 19")
-#endif
-
 #if defined(__apple_build_version__)
 // Given AppleClang XX.Y.Z, _LIBCPP_APPLE_CLANG_VER is XXYZ (e.g. AppleClang 14.0.3 => 1403)
 #  define _LIBCPP_COMPILER_CLANG_BASED
@@ -1230,21 +1219,6 @@ typedef __char32_t char32_t;
 #    define _LIBCPP_IF_WIDE_CHARACTERS(...) __VA_ARGS__
 #  endif
 
-#  if defined(_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES)
-#    define _LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR
-#    define _LIBCPP_ENABLE_CXX17_REMOVED_BINDERS
-#    define _LIBCPP_ENABLE_CXX17_REMOVED_RANDOM_SHUFFLE
-#    define _LIBCPP_ENABLE_CXX17_REMOVED_UNEXPECTED_FUNCTIONS
-#    define _LIBCPP_ENABLE_CXX17_REMOVED_UNARY_BINARY_FUNCTION
-#  endif // _LIBCPP_ENABLE_CXX17_REMOVED_FEATURES
-
-#  if defined(_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES)
-#    define _LIBCPP_ENABLE_CXX20_REMOVED_BINDER_TYPEDEFS
-#    define _LIBCPP_ENABLE_CXX20_REMOVED_NEGATORS
-#    define _LIBCPP_ENABLE_CXX20_REMOVED_RAW_STORAGE_ITERATOR
-#    define _LIBCPP_ENABLE_CXX20_REMOVED_TYPE_TRAITS
-#  endif // _LIBCPP_ENABLE_CXX20_REMOVED_FEATURES
-
 // clang-format off
 #  define _LIBCPP_PUSH_MACROS _Pragma("push_macro(\"min\")") _Pragma("push_macro(\"max\")") _Pragma("push_macro(\"refresh\")") _Pragma("push_macro(\"move\")") _Pragma("push_macro(\"erase\")")
 #  define _LIBCPP_POP_MACROS _Pragma("pop_macro(\"min\")") _Pragma("pop_macro(\"max\")") _Pragma("pop_macro(\"refresh\")") _Pragma("pop_macro(\"move\")") _Pragma("pop_macro(\"erase\")")
diff --git a/libcxx/test/libcxx/depr/enable_removed_cpp17_features.compile.pass.cpp b/libcxx/test/libcxx/depr/enable_removed_cpp17_features.compile.pass.cpp
deleted file mode 100644
index 1b7acad3cfa46..0000000000000
--- a/libcxx/test/libcxx/depr/enable_removed_cpp17_features.compile.pass.cpp
+++ /dev/null
@@ -1,36 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// Test that defining _LIBCPP_ENABLE_CXX17_REMOVED_FEATURES correctly defines
-// _LIBCPP_ENABLE_CXX17_REMOVED_FOO for each individual component macro.
-
-// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES -Wno-deprecated-pragma
-
-#include <__config>
-
-#include "test_macros.h"
-
-#ifndef _LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR
-#  error _LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR must be defined
-#endif
-
-#ifndef _LIBCPP_ENABLE_CXX17_REMOVED_BINDERS
-#  error _LIBCPP_ENABLE_CXX17_REMOVED_BINDERS must be defined
-#endif
-
-#ifndef _LIBCPP_ENABLE_CXX17_REMOVED_RANDOM_SHUFFLE
-#  error _LIBCPP_ENABLE_CXX17_REMOVED_RANDOM_SHUFFLE must be defined
-#endif
-
-#ifndef _LIBCPP_ENABLE_CXX17_REMOVED_UNEXPECTED_FUNCTIONS
-#error _LIBCPP_ENABLE_CXX17_REMOVED_UNEXPECTED_FUNCTIONS must be defined
-#endif
-
-#ifndef _LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR
-#error _LIBCPP_ENABLE_CXX17_REMOVED_AUTO_PTR must be defined
-#endif
diff --git a/libcxx/test/libcxx/depr/enable_removed_cpp17_features.deprecated.verify.cpp b/libcxx/test/libcxx/depr/enable_removed_cpp17_features.deprecated.verify.cpp
deleted file mode 100644
index 059c1b3ead4f1..0000000000000
--- a/libcxx/test/libcxx/depr/enable_removed_cpp17_features.deprecated.verify.cpp
+++ /dev/null
@@ -1,20 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// <__config>
-
-// Ensure that defining _LIBCPP_ENABLE_CXX17_REMOVED_FEATURES yields a
-// deprecation warning. We intend to issue a deprecation warning in LLVM 18
-// and remove the macro entirely in LLVM 19. As such, this test will be quite
-// short lived.
-
-// UNSUPPORTED: clang-modules-build
-
-// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES
-
-#include <__config> // expected-warning@* 1+ {{macro '_LIBCPP_ENABLE_CXX17_REMOVED_FEATURES' has been marked as deprecated}}
diff --git a/libcxx/test/libcxx/depr/enable_removed_cpp20_features.deprecated.verify.cpp b/libcxx/test/libcxx/depr/enable_removed_cpp20_features.deprecated.verify.cpp
deleted file mode 100644
index 163ff7d8fbda0..0000000000000
--- a/libcxx/test/libcxx/depr/enable_removed_cpp20_features.deprecated.verify.cpp
+++ /dev/null
@@ -1,20 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-// <__config>
-
-// Ensure that defining _LIBCPP_ENABLE_CXX20_REMOVED_FEATURES yields a
-// deprecation warning. We intend to issue a deprecation warning in LLVM 18
-// and remove the macro entirely in LLVM 19. As such, this test will be quite
-// short lived.
-
-// UNSUPPORTED: clang-modules-build
-
-// ADDITIONAL_COMPILE_FLAGS: -D_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES
-
-#include <version> // expected-warning@* 1+ {{macro '_LIBCPP_ENABLE_CXX20_REMOVED_FEATURES' has been marked as deprecated}}

From 388da6a31b7ba3062f9306b894656e265b9b33eb Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:21:39 +0200
Subject: [PATCH 631/886] [libc++][test] Removes Clang 16 validation. (#88558)

---
 libcxx/test/libcxx/containers/sequences/vector/asan.pass.cpp | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/libcxx/test/libcxx/containers/sequences/vector/asan.pass.cpp b/libcxx/test/libcxx/containers/sequences/vector/asan.pass.cpp
index 588ce2a3d17ed..614323b1ffd7b 100644
--- a/libcxx/test/libcxx/containers/sequences/vector/asan.pass.cpp
+++ b/libcxx/test/libcxx/containers/sequences/vector/asan.pass.cpp
@@ -29,8 +29,7 @@ void do_exit() {
 
 int main(int, char**)
 {
-#if TEST_STD_VER >= 11 && TEST_CLANG_VER >= 1600
-  // TODO(LLVM-18): Remove the special-casing
+#if TEST_STD_VER >= 11
   {
     typedef int T;
     typedef cpp17_input_iterator<T*> MyInputIter;
@@ -52,7 +51,7 @@ int main(int, char**)
     assert(v[1] == 'b');
     assert(is_contiguous_container_asan_correct(v));
   }
-#endif
+#endif // TEST_STD_VER >= 11
   {
     typedef cpp17_input_iterator<int*> MyInputIter;
     // Sould not trigger ASan.

From 8e0a4a89f940d17b520bbca040981f54195d3ea4 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:22:48 +0200
Subject: [PATCH 632/886] [libc++][doc] Documents -DLIBCXX_INSTALL_MODULES=ON.
 (#88547)

Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
---
 libcxx/docs/BuildingLibcxx.rst | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/libcxx/docs/BuildingLibcxx.rst b/libcxx/docs/BuildingLibcxx.rst
index 28145ed1049e0..a0a0cdb433974 100644
--- a/libcxx/docs/BuildingLibcxx.rst
+++ b/libcxx/docs/BuildingLibcxx.rst
@@ -206,6 +206,12 @@ libc++ specific options
 
   Toggle the installation of the libc++ headers.
 
+.. option:: LIBCXX_INSTALL_MODULES:BOOL
+
+  **Default**: ``OFF``
+
+  Toggle the installation of the experimental libc++ module sources.
+
 .. option:: LIBCXX_ENABLE_SHARED:BOOL
 
   **Default**: ``ON``

From 002297bdaa63f3f9f56b0051110ccf48f31c6825 Mon Sep 17 00:00:00 2001
From: Nico Weber <thakis@chromium.org>
Date: Tue, 16 Apr 2024 14:28:58 -0400
Subject: [PATCH 633/886] [gn] port 22629bb22a1b

---
 llvm/utils/gn/secondary/libcxx/src/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
index 1f6879358f22b..955854c7a134b 100644
--- a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
@@ -125,6 +125,7 @@ cxx_sources = [
   "condition_variable_destructor.cpp",
   "error_category.cpp",
   "exception.cpp",
+  "expected.cpp",
   "fstream.cpp",
   "functional.cpp",
   "future.cpp",

From 8debcf03c535e14ee47b14fddfcaeae3f32d1317 Mon Sep 17 00:00:00 2001
From: Peiming Liu <peiming@google.com>
Date: Tue, 16 Apr 2024 11:31:09 -0700
Subject: [PATCH 634/886] [mlir][sparse] introduce sparse_tensor.iterate
 operation (#88807)

A `sparse_tensor.iterate` iterates over a sparse iteration space
extracted from `sparse_tensor.extract_iteration_space` operation
introduced in https://github.com/llvm/llvm-project/pull/88554.

*DO NOT MERGE* before https://github.com/llvm/llvm-project/pull/88554
---
 .../Dialect/SparseTensor/IR/SparseTensor.h    |  38 ++
 .../SparseTensor/IR/SparseTensorAttrDefs.td   |  15 +
 .../SparseTensor/IR/SparseTensorOps.td        | 152 +++++++-
 .../SparseTensor/IR/SparseTensorTypes.td      |  95 +++++
 .../SparseTensor/IR/SparseTensorDialect.cpp   | 365 ++++++++++++++++++
 mlir/test/Dialect/SparseTensor/invalid.mlir   | 139 +++++++
 mlir/test/Dialect/SparseTensor/roundtrip.mlir |  53 +++
 .../SparseTensor/sparse_itertion_licm.mlir    |  26 ++
 8 files changed, 882 insertions(+), 1 deletion(-)
 create mode 100644 mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
index 5e523ec428aef..081a9b8cad8d6 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
@@ -17,9 +17,13 @@
 #include "mlir/IR/OpDefinition.h"
 #include "mlir/IR/OpImplementation.h"
 #include "mlir/IR/TensorEncoding.h"
+#include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/InferTypeOpInterface.h"
+#include "mlir/Interfaces/LoopLikeInterface.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 
+#include "llvm/ADT/bit.h"
+
 //===----------------------------------------------------------------------===//
 //
 // Type aliases to help code be more self-documenting. Unfortunately
@@ -41,6 +45,40 @@ using Level = uint64_t;
 /// including the value `ShapedType::kDynamic` (for shapes).
 using Size = int64_t;
 
+/// A simple wrapper to encode a bitset of defined  (at most 64) levels.
+class LevelSet {
+  uint64_t bits = 0;
+
+public:
+  LevelSet() = default;
+  explicit LevelSet(uint64_t bits) : bits(bits) {}
+  operator uint64_t() const { return bits; }
+
+  LevelSet &set(unsigned i) {
+    assert(i < 64);
+    bits |= 1 << i;
+    return *this;
+  }
+
+  LevelSet &operator|=(LevelSet lhs) {
+    bits |= static_cast<uint64_t>(lhs);
+    return *this;
+  }
+
+  LevelSet &lshift(unsigned offset) {
+    bits = bits << offset;
+    return *this;
+  }
+
+  bool operator[](unsigned i) const {
+    assert(i < 64);
+    return (bits & (1 << i)) != 0;
+  }
+
+  unsigned count() const { return llvm::popcount(bits); }
+  bool empty() const { return bits == 0; }
+};
+
 } // namespace sparse_tensor
 } // namespace mlir
 
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
index 4a9b9169ae4b8..d5398a98f5b17 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
@@ -19,6 +19,21 @@ class SparseTensor_Attr<string name,
                         list<Trait> traits = []>
     : AttrDef<SparseTensor_Dialect, name, traits>;
 
+//===----------------------------------------------------------------------===//
+// A simple bitset attribute wrapped over a single int64_t to encode a set of
+// sparse tensor levels.
+//===----------------------------------------------------------------------===//
+
+def LevelSetAttr :
+    TypedAttrBase<
+      I64, "IntegerAttr",
+      And<[CPred<"::llvm::isa<::mlir::IntegerAttr>($_self)">,
+           CPred<"::llvm::cast<::mlir::IntegerAttr>($_self).getType().isInteger(64)">]>,
+      "LevelSet attribute"> {
+  let returnType = [{::mlir::sparse_tensor::LevelSet}];
+  let convertFromStorage = [{::mlir::sparse_tensor::LevelSet($_self.getValue().getZExtValue())}];
+}
+
 //===----------------------------------------------------------------------===//
 // These attributes are just like `IndexAttr` except that they clarify whether
 // the index refers to a dimension (an axis of the semantic tensor) or a level
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
index 0cfc64f9988a0..b43d716d5e864 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
@@ -15,6 +15,8 @@ include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td"
 include "mlir/Dialect/SparseTensor/IR/SparseTensorInterfaces.td"
 include "mlir/Interfaces/InferTypeOpInterface.td"
 include "mlir/Interfaces/SideEffectInterfaces.td"
+include "mlir/Interfaces/ControlFlowInterfaces.td"
+include "mlir/Interfaces/LoopLikeInterface.td"
 
 //===----------------------------------------------------------------------===//
 // Base class.
@@ -1277,7 +1279,7 @@ def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResu
 
 def SparseTensor_YieldOp : SparseTensor_Op<"yield", [Pure, Terminator,
     ParentOneOf<["BinaryOp", "UnaryOp", "ReduceOp", "SelectOp",
-                 "ForeachOp"]>]>,
+                 "ForeachOp", "IterateOp"]>]>,
     Arguments<(ins Variadic<AnyType>:$results)> {
   let summary = "Yield from sparse_tensor set-like operations";
   let description = [{
@@ -1430,6 +1432,154 @@ def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
   let hasVerifier = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Operations.
+//===----------------------------------------------------------------------===//
+
+def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
+    [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
+
+  let arguments = (ins AnySparseTensor:$tensor,
+                       Optional<AnySparseIterator>:$parentIter,
+                       LevelAttr:$loLvl, LevelAttr:$hiLvl);
+
+  let results = (outs AnySparseIterSpace:$resultSpace);
+
+  let summary = "Extract an iteration space from a sparse tensor between certain levels";
+  let description = [{
+      Extracts a `!sparse_tensor.iter_space` from a sparse tensor between
+      certian (consecutive) levels.
+
+      `tensor`: the input sparse tensor that defines the iteration space.
+      `parentIter`: the iterator for the previous level, at which the iteration space
+      at the current levels will be extracted.
+      `loLvl`, `hiLvl`: the level range between [loLvl, hiLvl) in the input tensor that
+      the returned iteration space covers. `hiLvl - loLvl` defines the dimension of the
+      iteration space.
+
+      Example:
+      ```mlir
+      // Extracts a 1-D iteration space from a COO tensor at level 1.
+      %space = sparse_tensor.iteration.extract_space %sp at %it1 lvls = 1
+        : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+      ```
+  }];
+
+
+  let extraClassDeclaration = [{
+    std::pair<Level, Level> getLvlRange() {
+      return std::make_pair(getLoLvl(), getHiLvl());
+    }
+    unsigned getSpaceDim() {
+      return getHiLvl() - getLoLvl();
+    }
+    ArrayRef<::mlir::sparse_tensor::LevelType> getSpaceLvlTypes() {
+      return getResultSpace().getType().getLvlTypes();
+    }
+  }];
+
+  let hasVerifier = 1;
+  let assemblyFormat = "$tensor (`at` $parentIter^)? `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) "
+                       " attr-dict `:` type($tensor) (`,` type($parentIter)^)?";
+}
+
+def IterateOp : SparseTensor_Op<"iterate",
+    [RecursiveMemoryEffects, RecursivelySpeculatable,
+     DeclareOpInterfaceMethods<LoopLikeOpInterface,
+      ["getInitsMutable", "getLoopResults", "getRegionIterArgs",
+       "getYieldedValuesMutable"]>,
+     DeclareOpInterfaceMethods<RegionBranchOpInterface,
+      ["getEntrySuccessorOperands"]>,
+     SingleBlockImplicitTerminator<"sparse_tensor::YieldOp">]> {
+
+  let arguments = (ins AnySparseIterSpace:$iterSpace,
+                       Variadic<AnyType>:$initArgs,
+                       LevelSetAttr:$crdUsedLvls);
+  let results = (outs Variadic<AnyType>:$results);
+  let regions = (region SizedRegion<1>:$region);
+
+  let summary = "Iterate over a sparse iteration space";
+  let description = [{
+      The `sparse_tensor.iterate` operations represents a loop over the
+      provided iteration space extracted from a specific sparse tensor.
+      The operation defines an SSA value for a sparse iterator that points
+      to the current stored element in the sparse tensor and SSA values
+      for coordinates of the stored element. The coordinates are always
+      converted to `index` type despite of the underlying sparse tensor
+      storage. When coordinates are not used, the SSA values can be skipped
+      by `_` symbols, which usually leads to simpler generated code after
+      sparsification. For example:
+
+      ```mlir
+      // The coordinate for level 0 is not used when iterating over a 2-D
+      // iteration space.
+      %sparse_tensor.iterate %iterator in %space at(_, %crd_1)
+        : !sparse_tensor.iter_space<#CSR, lvls = 0 to 2>
+      ```
+
+      `sparse_tensor.iterate` can also operate on loop-carried variables
+      and returns the final values after loop termination.
+      The initial values of the variables are passed as additional SSA operands
+      to the iterator SSA value and used coordinate SSA values mentioned
+      above. The operation region has an argument for the iterator, variadic
+      arguments for specified (used) coordiates and followed by one argument
+      for each loop-carried variable, representing the value of the variable
+      at the current iteration.
+      The body region must contain exactly one block that terminates with
+      `sparse_tensor.yield`.
+
+      `sparse_tensor.iterate` results hold the final values after the last
+      iteration. If the `sparse_tensor.iterate` defines any values, a yield
+      must be explicitly present.
+      The number and types of the `sparse_tensor.iterate` results must match
+      the initial values in the iter_args binding and the yield operands.
+
+
+      A nested `sparse_tensor.iterate` example that prints all the coordinates
+      stored in the sparse input:
+
+      ```mlir
+      func.func @nested_iterate(%sp : tensor<4x8xf32, #COO>) {
+        // Iterates over the first level of %sp
+        %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+        %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd0)
+            : !sparse_tensor.iter_space<#COO, lvls = 0 to 1>  {
+          // Iterates over the second level of %sp
+          %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1
+              : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0 to 1>
+          %r2 = sparse_tensor.iterate %it2 in %l2 at (crd1)
+              : !sparse_tensor.iter_space<#COO, lvls = 1 to 2>  {
+             vector.print %crd0 : index
+             vector.print %crd1 : index
+          }
+        }
+      }
+
+      ```
+  }];
+
+  let extraClassDeclaration = [{
+    unsigned getSpaceDim() {
+      return getIterSpace().getType().getSpaceDim();
+    }
+    BlockArgument getIterator() {
+      return getRegion().getArguments().front();
+    }
+    Block::BlockArgListType getCrds() {
+      // The first block argument is iterator, the remaining arguments are
+      // referenced coordinates.
+      return getRegion().getArguments().slice(1, getCrdUsedLvls().count());
+    }
+    unsigned getNumRegionIterArgs() {
+      return getRegion().getArguments().size() - 1 - getCrdUsedLvls().count();
+    }
+  }];
+
+  let hasVerifier = 1;
+  let hasRegionVerifier = 1;
+  let hasCustomAssemblyFormat = 1;
+}
+
 //===----------------------------------------------------------------------===//
 // Sparse Tensor Debugging and Test-Only Operations.
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
index 185cff46ae25d..264a0a5b3bee6 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
@@ -72,4 +72,99 @@ def SparseTensorStorageSpecifier
     : Type<CPred<"::llvm::isa<::mlir::sparse_tensor::StorageSpecifierType>($_self)">, "metadata",
           "::mlir::sparse_tensor::StorageSpecifierType">;
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Types.
+//===----------------------------------------------------------------------===//
+
+def SparseTensor_IterSpace : SparseTensor_Type<"IterSpace"> {
+  let mnemonic = "iter_space";
+
+  let description = [{
+    A sparse iteration space that represents an abstract N-D (sparse) iteration space
+    extracted from a sparse tensor.
+
+    Examples:
+
+    ```mlir
+    // An iteration space extracted from a CSR tensor between levels [0, 2).
+    !iter_space<#CSR, lvls = 0 to 2>
+    ```
+  }];
+
+  let parameters = (ins
+     SparseTensorEncodingAttr : $encoding,
+     "Level" : $loLvl,
+     "Level" : $hiLvl
+  );
+
+  let extraClassDeclaration = [{
+     /// The the dimension of the iteration space.
+     unsigned getSpaceDim() const {
+       return getHiLvl() - getLoLvl();
+     }
+
+     /// Get the level types for the iteration space.
+     ArrayRef<LevelType> getLvlTypes() const {
+       return getEncoding().getLvlTypes().slice(getLoLvl(), getSpaceDim());
+     }
+
+     /// Whether the iteration space is unique (i.e., no duplicated coordinate).
+     bool isUnique() {
+       return !getLvlTypes().back().isa<LevelPropNonDefault::Nonunique>();
+     }
+
+     /// Get the corresponding iterator type.
+     ::mlir::sparse_tensor::IteratorType getIteratorType() const;
+  }];
+
+  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
+}
+
+def SparseTensor_Iterator : SparseTensor_Type<"Iterator"> {
+  let mnemonic = "iterator";
+
+  let description = [{
+    An iterator that points to the current element in the corresponding iteration space.
+
+    Examples:
+
+    ```mlir
+    // An iterator that iterates over a iteration space of type `!iter_space<#CSR, lvls = 0 to 2>`
+    !iterator<#CSR, lvls = 0 to 2>
+    ```
+  }];
+
+  let parameters = (ins
+     SparseTensorEncodingAttr : $encoding,
+     "Level" : $loLvl,
+     "Level" : $hiLvl
+  );
+
+  let extraClassDeclaration = [{
+     /// Get the corresponding iteration space type.
+     ::mlir::sparse_tensor::IterSpaceType getIterSpaceType() const;
+
+     unsigned getSpaceDim() const { return getIterSpaceType().getSpaceDim(); }
+     ArrayRef<LevelType> getLvlTypes() const { return getIterSpaceType().getLvlTypes(); }
+     bool isUnique() { return getIterSpaceType().isUnique(); }
+  }];
+
+  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
+}
+
+def IsSparseSparseIterSpaceTypePred
+    : CPred<"::llvm::isa<::mlir::sparse_tensor::IterSpaceType>($_self)">;
+
+def IsSparseSparseIteratorTypePred
+    : CPred<"::llvm::isa<::mlir::sparse_tensor::IteratorType>($_self)">;
+
+def AnySparseIterSpace
+    : Type<IsSparseSparseIterSpaceTypePred, "sparse iteration space",
+          "::mlir::sparse_tensor::IterSpaceType">;
+
+def AnySparseIterator
+    : Type<IsSparseSparseIteratorTypePred, "sparse iterator",
+          "::mlir::sparse_tensor::IteratorType">;
+
+
 #endif // SPARSETENSOR_TYPES
diff --git a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
index e9058394d33da..36908def09f40 100644
--- a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
+++ b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
@@ -30,6 +30,14 @@
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.cpp.inc"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrEnums.cpp.inc"
 
+// Forward declarations, following custom print/parsing methods are referenced
+// by the generated code for SparseTensorTypes.td.
+static mlir::ParseResult parseLevelRange(mlir::AsmParser &,
+                                         mlir::sparse_tensor::Level &,
+                                         mlir::sparse_tensor::Level &);
+static void printLevelRange(mlir::AsmPrinter &, mlir::sparse_tensor::Level,
+                            mlir::sparse_tensor::Level);
+
 #define GET_TYPEDEF_CLASSES
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.cpp.inc"
 
@@ -1953,6 +1961,363 @@ LogicalResult SortOp::verify() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Operations.
+//===----------------------------------------------------------------------===//
+
+IterSpaceType IteratorType::getIterSpaceType() const {
+  return IterSpaceType::get(getContext(), getEncoding(), getLoLvl(),
+                            getHiLvl());
+}
+
+IteratorType IterSpaceType::getIteratorType() const {
+  return IteratorType::get(getContext(), getEncoding(), getLoLvl(), getHiLvl());
+}
+
+/// Parses a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static ParseResult parseLevelRange(AsmParser &parser, Level &lvlLo,
+                                   Level &lvlHi) {
+  if (parser.parseInteger(lvlLo))
+    return failure();
+
+  if (succeeded(parser.parseOptionalKeyword("to"))) {
+    if (parser.parseInteger(lvlHi))
+      return failure();
+  } else {
+    lvlHi = lvlLo + 1;
+  }
+
+  if (lvlHi <= lvlLo)
+    parser.emitError(parser.getNameLoc(),
+                     "expect larger level upper bound than lower bound");
+
+  return success();
+}
+
+/// Parses a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static ParseResult parseLevelRange(OpAsmParser &parser, IntegerAttr &lvlLoAttr,
+                                   IntegerAttr &lvlHiAttr) {
+  Level lvlLo, lvlHi;
+  if (parseLevelRange(parser, lvlLo, lvlHi))
+    return failure();
+
+  lvlLoAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlLo);
+  lvlHiAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlHi);
+  return success();
+}
+
+/// Prints a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static void printLevelRange(AsmPrinter &p, Level lo, Level hi) {
+
+  if (lo + 1 == hi)
+    p << lo;
+  else
+    p << lo << " to " << hi;
+}
+
+/// Prints a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static void printLevelRange(OpAsmPrinter &p, Operation *, IntegerAttr lvlLo,
+                            IntegerAttr lvlHi) {
+  unsigned lo = lvlLo.getValue().getZExtValue();
+  unsigned hi = lvlHi.getValue().getZExtValue();
+  printLevelRange(p, lo, hi);
+}
+
+static ParseResult
+parseSparseSpaceLoop(OpAsmParser &parser, OperationState &state,
+                     SmallVectorImpl<OpAsmParser::Argument> &iterators,
+                     SmallVectorImpl<OpAsmParser::Argument> &iterArgs) {
+  SmallVector<OpAsmParser::UnresolvedOperand> spaces;
+  SmallVector<OpAsmParser::UnresolvedOperand> initArgs;
+
+  // Parses "%iters, ... in %spaces, ..."
+  if (parser.parseArgumentList(iterators) || parser.parseKeyword("in") ||
+      parser.parseOperandList(spaces))
+    return failure();
+
+  if (iterators.size() != spaces.size())
+    return parser.emitError(
+        parser.getNameLoc(),
+        "mismatch in number of sparse iterators and sparse spaces");
+
+  // Parse "at(%crd0, _, ...)"
+  LevelSet crdUsedLvlSet;
+  bool hasUsedCrds = succeeded(parser.parseOptionalKeyword("at"));
+  unsigned lvlCrdCnt = 0;
+  if (hasUsedCrds) {
+    ParseResult crdList = parser.parseCommaSeparatedList(
+        OpAsmParser::Delimiter::Paren, [&]() -> ParseResult {
+          if (parser.parseOptionalKeyword("_")) {
+            if (parser.parseArgument(iterArgs.emplace_back()))
+              return failure();
+            // Always use IndexType for the coordinate.
+            crdUsedLvlSet.set(lvlCrdCnt);
+            iterArgs.back().type = parser.getBuilder().getIndexType();
+          }
+          lvlCrdCnt += 1;
+          return success();
+        });
+    if (failed(crdList)) {
+      return parser.emitError(
+          parser.getNameLoc(),
+          "expecting SSA value or \"_\" for level coordinates");
+    }
+  }
+  // Set the CrdUsedLvl bitset.
+  state.addAttribute("crdUsedLvls",
+                     parser.getBuilder().getI64IntegerAttr(crdUsedLvlSet));
+
+  // Parse "iter_args(%arg = %init, ...)"
+  bool hasIterArgs = succeeded(parser.parseOptionalKeyword("iter_args"));
+  if (hasIterArgs)
+    if (parser.parseAssignmentList(iterArgs, initArgs))
+      return failure();
+
+  SmallVector<Type> iterSpaceTps;
+  // parse ": sparse_tensor.iter_space -> ret"
+  if (parser.parseColon() || parser.parseTypeList(iterSpaceTps))
+    return failure();
+  if (iterSpaceTps.size() != spaces.size())
+    return parser.emitError(parser.getNameLoc(),
+                            "mismatch in number of iteration space operands "
+                            "and iteration space types");
+
+  for (auto [it, tp] : llvm::zip_equal(iterators, iterSpaceTps)) {
+    IterSpaceType spaceTp = llvm::dyn_cast<IterSpaceType>(tp);
+    if (!spaceTp)
+      return parser.emitError(parser.getNameLoc(),
+                              "expected sparse_tensor.iter_space type for "
+                              "iteration space operands");
+    if (hasUsedCrds && spaceTp.getSpaceDim() != lvlCrdCnt)
+      return parser.emitError(parser.getNameLoc(),
+                              "mismatch in number of iteration space dimension "
+                              "and specified coordinates");
+    it.type = spaceTp.getIteratorType();
+  }
+
+  if (hasIterArgs)
+    if (parser.parseArrowTypeList(state.types))
+      return failure();
+
+  // Resolves input operands.
+  if (parser.resolveOperands(spaces, iterSpaceTps, parser.getNameLoc(),
+                             state.operands))
+    return failure();
+
+  if (hasIterArgs) {
+    unsigned numCrds = crdUsedLvlSet.count();
+    // Strip off leading args that used for coordinates.
+    MutableArrayRef args = MutableArrayRef(iterArgs).drop_front(numCrds);
+    if (args.size() != initArgs.size() || args.size() != state.types.size()) {
+      return parser.emitError(
+          parser.getNameLoc(),
+          "mismatch in number of iteration arguments and return values");
+    }
+
+    for (auto [it, init, tp] : llvm::zip_equal(args, initArgs, state.types)) {
+      it.type = tp;
+      if (parser.resolveOperand(init, tp, state.operands))
+        return failure();
+    }
+  }
+  return success();
+}
+
+LogicalResult ExtractIterSpaceOp::inferReturnTypes(
+    MLIRContext *ctx, std::optional<Location> loc, ValueRange ops,
+    DictionaryAttr attr, OpaqueProperties prop, RegionRange region,
+    SmallVectorImpl<mlir::Type> &ret) {
+
+  ExtractIterSpaceOp::Adaptor adaptor(ops, attr, prop, region);
+  SparseTensorType stt = getSparseTensorType(adaptor.getTensor());
+  ret.push_back(IterSpaceType::get(ctx, stt.getEncoding(), adaptor.getLoLvl(),
+                                   adaptor.getHiLvl()));
+  return success();
+}
+
+LogicalResult ExtractIterSpaceOp::verify() {
+  if (getLoLvl() >= getHiLvl())
+    return emitOpError("expected smaller level low than level high");
+
+  TypedValue<IteratorType> pIter = getParentIter();
+  if ((pIter && getLoLvl() == 0) || (!pIter && getLoLvl() != 0)) {
+    return emitOpError(
+        "parent iterator should be specified iff level lower bound equals 0");
+  }
+
+  if (pIter) {
+    IterSpaceType spaceTp = getResultSpace().getType();
+    if (pIter.getType().getEncoding() != spaceTp.getEncoding())
+      return emitOpError(
+          "mismatch in parent iterator encoding and iteration space encoding.");
+
+    if (spaceTp.getLoLvl() != pIter.getType().getHiLvl())
+      return emitOpError("parent iterator should be used to extract an "
+                         "iteration space from a consecutive level.");
+  }
+
+  return success();
+}
+
+ParseResult IterateOp::parse(OpAsmParser &parser, OperationState &result) {
+  OpAsmParser::Argument iterator;
+  OpAsmParser::UnresolvedOperand iterSpace;
+
+  SmallVector<OpAsmParser::Argument> iters, iterArgs;
+  if (parseSparseSpaceLoop(parser, result, iters, iterArgs))
+    return failure();
+  if (iters.size() != 1)
+    return parser.emitError(parser.getNameLoc(),
+                            "expected only one iterator/iteration space");
+
+  iters.append(iterArgs);
+  Region *body = result.addRegion();
+  if (parser.parseRegion(*body, iters))
+    return failure();
+
+  IterateOp::ensureTerminator(*body, parser.getBuilder(), result.location);
+
+  // Parse the optional attribute list.
+  if (parser.parseOptionalAttrDict(result.attributes))
+    return failure();
+
+  return success();
+}
+
+/// Prints the initialization list in the form of
+///   <prefix>(%inner = %outer, %inner2 = %outer2, <...>)
+/// where 'inner' values are assumed to be region arguments and 'outer' values
+/// are regular SSA values.
+static void printInitializationList(OpAsmPrinter &p,
+                                    Block::BlockArgListType blocksArgs,
+                                    ValueRange initializers,
+                                    StringRef prefix = "") {
+  assert(blocksArgs.size() == initializers.size() &&
+         "expected same length of arguments and initializers");
+  if (initializers.empty())
+    return;
+
+  p << prefix << '(';
+  llvm::interleaveComma(llvm::zip(blocksArgs, initializers), p, [&](auto it) {
+    p << std::get<0>(it) << " = " << std::get<1>(it);
+  });
+  p << ")";
+}
+
+static void printUsedCrdsList(OpAsmPrinter &p, unsigned spaceDim,
+                              Block::BlockArgListType blocksArgs,
+                              LevelSet crdUsedLvls) {
+  if (crdUsedLvls.empty())
+    return;
+
+  p << " at(";
+  for (unsigned i = 0; i < spaceDim; i++) {
+    if (crdUsedLvls[i]) {
+      p << blocksArgs.front();
+      blocksArgs = blocksArgs.drop_front();
+    } else {
+      p << "_";
+    }
+    if (i != spaceDim - 1)
+      p << ", ";
+  }
+  assert(blocksArgs.empty());
+  p << ")";
+}
+
+void IterateOp::print(OpAsmPrinter &p) {
+  p << " " << getIterator() << " in " << getIterSpace();
+  printUsedCrdsList(p, getSpaceDim(), getCrds(), getCrdUsedLvls());
+  printInitializationList(p, getRegionIterArgs(), getInitArgs(), " iter_args");
+
+  p << " : " << getIterSpace().getType() << " ";
+  if (!getInitArgs().empty())
+    p << "-> (" << getInitArgs().getTypes() << ") ";
+
+  p.printRegion(getRegion(), /*printEntryBlockArgs=*/false,
+                /*printBlockTerminators=*/!getInitArgs().empty());
+}
+
+LogicalResult IterateOp::verify() {
+  if (getInitArgs().size() != getNumResults()) {
+    return emitOpError(
+        "mismatch in number of loop-carried values and defined values");
+  }
+  return success();
+}
+
+LogicalResult IterateOp::verifyRegions() {
+  if (getIterator().getType() != getIterSpace().getType().getIteratorType())
+    return emitOpError("mismatch in iterator and iteration space type");
+  if (getNumRegionIterArgs() != getNumResults())
+    return emitOpError(
+        "mismatch in number of basic block args and defined values");
+
+  auto initArgs = getInitArgs();
+  auto iterArgs = getRegionIterArgs();
+  auto yieldVals = getYieldedValues();
+  auto opResults = getResults();
+  if (!llvm::all_equal({initArgs.size(), iterArgs.size(), yieldVals.size(),
+                        opResults.size()})) {
+    return emitOpError() << "number mismatch between iter args and results.";
+  }
+
+  for (auto [i, init, iter, yield, ret] :
+       llvm::enumerate(initArgs, iterArgs, yieldVals, opResults)) {
+    if (init.getType() != ret.getType())
+      return emitOpError() << "types mismatch between " << i
+                           << "th iter operand and defined value";
+    if (iter.getType() != ret.getType())
+      return emitOpError() << "types mismatch between " << i
+                           << "th iter region arg and defined value";
+    if (yield.getType() != ret.getType())
+      return emitOpError() << "types mismatch between " << i
+                           << "th yield value and defined value";
+  }
+
+  return success();
+}
+
+/// IterateOp implemented OpInterfaces' methods.
+SmallVector<Region *> IterateOp::getLoopRegions() { return {&getRegion()}; }
+
+MutableArrayRef<OpOperand> IterateOp::getInitsMutable() {
+  return getInitArgsMutable();
+}
+
+Block::BlockArgListType IterateOp::getRegionIterArgs() {
+  return getRegion().getArguments().take_back(getNumRegionIterArgs());
+}
+
+std::optional<MutableArrayRef<OpOperand>> IterateOp::getYieldedValuesMutable() {
+  return cast<sparse_tensor::YieldOp>(
+             getRegion().getBlocks().front().getTerminator())
+      .getResultsMutable();
+}
+
+std::optional<ResultRange> IterateOp::getLoopResults() { return getResults(); }
+
+OperandRange IterateOp::getEntrySuccessorOperands(RegionBranchPoint point) {
+  return getInitArgs();
+}
+
+void IterateOp::getSuccessorRegions(RegionBranchPoint point,
+                                    SmallVectorImpl<RegionSuccessor> &regions) {
+  // Both the operation itself and the region may be branching into the body or
+  // back into the operation itself.
+  regions.push_back(RegionSuccessor(&getRegion(), getRegionIterArgs()));
+  // It is possible for loop not to enter the body.
+  regions.push_back(RegionSuccessor(getResults()));
+}
+
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Dialect Setups.
+//===----------------------------------------------------------------------===//
+
 /// Materialize a single constant operation from a given attribute value with
 /// the desired resultant type.
 Operation *SparseTensorDialect::materializeConstant(OpBuilder &builder,
diff --git a/mlir/test/Dialect/SparseTensor/invalid.mlir b/mlir/test/Dialect/SparseTensor/invalid.mlir
index 7f5c05190fc9a..b13024cd4ed99 100644
--- a/mlir/test/Dialect/SparseTensor/invalid.mlir
+++ b/mlir/test/Dialect/SparseTensor/invalid.mlir
@@ -1012,3 +1012,142 @@ func.func @sparse_print(%arg0: tensor<10x10xf64>) {
   sparse_tensor.print %arg0 : tensor<10x10xf64>
   return
 }
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 2>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' expect larger level upper bound than lower bound}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 to 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 2>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 1 : tensor<4x8xf32, #COO>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+#CSR = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : dense,
+    j : compressed
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#CSR, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op mismatch in parent iterator encoding and iteration space encoding.}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#CSR, lvls = 0>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be used to extract an iteration space from a consecutive level.}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return
+}
+
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  // expected-error @+1 {{'sparse_tensor.iterate' op different number of region iter_args and yielded values: 2 != 1}}
+  %r1, %r2 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%si = %i, %sj = %j): !sparse_tensor.iter_space<#COO, lvls = 0> -> (index, index) {
+    sparse_tensor.yield %si : index
+  }
+  return %r1 : index
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+// expected-note@+1 {{prior use here}}
+func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index) -> f32 {
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  // expected-error @+1 {{use of value '%i' expects different type than prior uses: 'f32' vs 'index'}}
+  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%outer = %i): !sparse_tensor.iter_space<#COO, lvls = 0> -> f32 {
+    sparse_tensor.yield %outer : f32
+  }
+  return %r1 : f32
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  // expected-error @+1 {{'sparse_tensor.iterate' op 0-th region iter_arg and 0-th yielded value have different type: 'index' != 'f32'}}
+  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%si = %i): !sparse_tensor.iter_space<#COO, lvls = 0> -> index {
+    %y = arith.constant 1.0 :  f32
+    sparse_tensor.yield %y : f32
+  }
+  return %r1 : index
+}
diff --git a/mlir/test/Dialect/SparseTensor/roundtrip.mlir b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
index 12f69c1d37b9c..e9a898f16b41d 100644
--- a/mlir/test/Dialect/SparseTensor/roundtrip.mlir
+++ b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
@@ -738,3 +738,56 @@ func.func @sparse_has_runtime() -> i1 {
   %has_runtime = sparse_tensor.has_runtime_library
   return %has_runtime : i1
 }
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+// CHECK-LABEL:   func.func @sparse_extract_iter_space(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf32, #sparse{{[0-9]*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: !sparse_tensor.iterator<#sparse{{[0-9]*}}, lvls = 0>)
+// CHECK:           %[[VAL_2:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] lvls = 0
+// CHECK:           %[[VAL_3:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] at %[[VAL_1]] lvls = 1
+// CHECK:           return %[[VAL_2]], %[[VAL_3]] : !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 0>, !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 1>
+// CHECK:         }
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>)
+  -> (!sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>) {
+  // Extracting the iteration space for the first level needs no parent iterator.
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  // Extracting the iteration space for the second level needs a parent iterator.
+  %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return %l1, %l2 : !sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>
+}
+
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+// CHECK-LABEL:   func.func @sparse_iterate(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf32, #sparse{{[0-9]*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: index,
+// CHECK-SAME:      %[[VAL_2:.*]]: index) -> index {
+// CHECK:           %[[VAL_3:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] lvls = 0 : tensor<4x8xf32, #sparse{{[0-9]*}}>
+// CHECK:           %[[VAL_4:.*]] = sparse_tensor.iterate %[[VAL_5:.*]] in %[[VAL_3]] at(%[[VAL_6:.*]]) iter_args(%[[VAL_7:.*]] = %[[VAL_1]]) : !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 0> -> (index) {
+// CHECK:             sparse_tensor.yield %[[VAL_7]] : index
+// CHECK:           }
+// CHECK:           return %[[VAL_4]] : index
+// CHECK:         }
+func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%outer = %i): !sparse_tensor.iter_space<#COO, lvls = 0 to 1> -> index {
+    sparse_tensor.yield %outer : index
+  }
+  return %r1 : index
+}
diff --git a/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir b/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir
new file mode 100644
index 0000000000000..e7158d04b37fe
--- /dev/null
+++ b/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir
@@ -0,0 +1,26 @@
+// RUN: mlir-opt %s --loop-invariant-code-motion | FileCheck %s
+
+#CSR = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : dense,
+    j : compressed
+  )
+}>
+
+// Make sure that pure instructions are hoisted outside the loop.
+//
+// CHECK: sparse_tensor.values
+// CHECK: sparse_tensor.positions
+// CHECK: sparse_tensor.coordinate
+// CHECK: sparse_tensor.iterate
+func.func @sparse_iterate(%sp : tensor<?x?xf64, #CSR>) {
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<?x?xf64, #CSR>
+  sparse_tensor.iterate %it1 in %l1 at (%crd) : !sparse_tensor.iter_space<#CSR, lvls = 0> {
+    %0 = sparse_tensor.values %sp : tensor<?x?xf64, #CSR> to memref<?xf64>
+    %1 = sparse_tensor.positions %sp { level = 1 : index } : tensor<?x?xf64, #CSR> to memref<?xindex>
+    %2 = sparse_tensor.coordinates  %sp { level = 1 : index } : tensor<?x?xf64, #CSR> to memref<?xindex>
+    "test.op"(%0, %1, %2) : (memref<?xf64>, memref<?xindex>, memref<?xindex>) -> ()
+  }
+
+  return
+}

From b9556532c7391a2acb77ab0f7d7b36e1dc382b24 Mon Sep 17 00:00:00 2001
From: Peiming Liu <peiming@google.com>
Date: Tue, 16 Apr 2024 11:31:33 -0700
Subject: [PATCH 635/886] Revert "[mlir][sparse] introduce
 sparse_tensor.iterate operation" (#88953)

Reverts llvm/llvm-project#88807 (merged by mistake)
---
 .../Dialect/SparseTensor/IR/SparseTensor.h    |  38 --
 .../SparseTensor/IR/SparseTensorAttrDefs.td   |  15 -
 .../SparseTensor/IR/SparseTensorOps.td        | 152 +-------
 .../SparseTensor/IR/SparseTensorTypes.td      |  95 -----
 .../SparseTensor/IR/SparseTensorDialect.cpp   | 365 ------------------
 mlir/test/Dialect/SparseTensor/invalid.mlir   | 139 -------
 mlir/test/Dialect/SparseTensor/roundtrip.mlir |  53 ---
 .../SparseTensor/sparse_itertion_licm.mlir    |  26 --
 8 files changed, 1 insertion(+), 882 deletions(-)
 delete mode 100644 mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
index 081a9b8cad8d6..5e523ec428aef 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensor.h
@@ -17,13 +17,9 @@
 #include "mlir/IR/OpDefinition.h"
 #include "mlir/IR/OpImplementation.h"
 #include "mlir/IR/TensorEncoding.h"
-#include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/InferTypeOpInterface.h"
-#include "mlir/Interfaces/LoopLikeInterface.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 
-#include "llvm/ADT/bit.h"
-
 //===----------------------------------------------------------------------===//
 //
 // Type aliases to help code be more self-documenting. Unfortunately
@@ -45,40 +41,6 @@ using Level = uint64_t;
 /// including the value `ShapedType::kDynamic` (for shapes).
 using Size = int64_t;
 
-/// A simple wrapper to encode a bitset of defined  (at most 64) levels.
-class LevelSet {
-  uint64_t bits = 0;
-
-public:
-  LevelSet() = default;
-  explicit LevelSet(uint64_t bits) : bits(bits) {}
-  operator uint64_t() const { return bits; }
-
-  LevelSet &set(unsigned i) {
-    assert(i < 64);
-    bits |= 1 << i;
-    return *this;
-  }
-
-  LevelSet &operator|=(LevelSet lhs) {
-    bits |= static_cast<uint64_t>(lhs);
-    return *this;
-  }
-
-  LevelSet &lshift(unsigned offset) {
-    bits = bits << offset;
-    return *this;
-  }
-
-  bool operator[](unsigned i) const {
-    assert(i < 64);
-    return (bits & (1 << i)) != 0;
-  }
-
-  unsigned count() const { return llvm::popcount(bits); }
-  bool empty() const { return bits == 0; }
-};
-
 } // namespace sparse_tensor
 } // namespace mlir
 
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
index d5398a98f5b17..4a9b9169ae4b8 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.td
@@ -19,21 +19,6 @@ class SparseTensor_Attr<string name,
                         list<Trait> traits = []>
     : AttrDef<SparseTensor_Dialect, name, traits>;
 
-//===----------------------------------------------------------------------===//
-// A simple bitset attribute wrapped over a single int64_t to encode a set of
-// sparse tensor levels.
-//===----------------------------------------------------------------------===//
-
-def LevelSetAttr :
-    TypedAttrBase<
-      I64, "IntegerAttr",
-      And<[CPred<"::llvm::isa<::mlir::IntegerAttr>($_self)">,
-           CPred<"::llvm::cast<::mlir::IntegerAttr>($_self).getType().isInteger(64)">]>,
-      "LevelSet attribute"> {
-  let returnType = [{::mlir::sparse_tensor::LevelSet}];
-  let convertFromStorage = [{::mlir::sparse_tensor::LevelSet($_self.getValue().getZExtValue())}];
-}
-
 //===----------------------------------------------------------------------===//
 // These attributes are just like `IndexAttr` except that they clarify whether
 // the index refers to a dimension (an axis of the semantic tensor) or a level
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
index b43d716d5e864..0cfc64f9988a0 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
@@ -15,8 +15,6 @@ include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td"
 include "mlir/Dialect/SparseTensor/IR/SparseTensorInterfaces.td"
 include "mlir/Interfaces/InferTypeOpInterface.td"
 include "mlir/Interfaces/SideEffectInterfaces.td"
-include "mlir/Interfaces/ControlFlowInterfaces.td"
-include "mlir/Interfaces/LoopLikeInterface.td"
 
 //===----------------------------------------------------------------------===//
 // Base class.
@@ -1279,7 +1277,7 @@ def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResu
 
 def SparseTensor_YieldOp : SparseTensor_Op<"yield", [Pure, Terminator,
     ParentOneOf<["BinaryOp", "UnaryOp", "ReduceOp", "SelectOp",
-                 "ForeachOp", "IterateOp"]>]>,
+                 "ForeachOp"]>]>,
     Arguments<(ins Variadic<AnyType>:$results)> {
   let summary = "Yield from sparse_tensor set-like operations";
   let description = [{
@@ -1432,154 +1430,6 @@ def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
   let hasVerifier = 1;
 }
 
-//===----------------------------------------------------------------------===//
-// Sparse Tensor Iteration Operations.
-//===----------------------------------------------------------------------===//
-
-def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
-    [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
-
-  let arguments = (ins AnySparseTensor:$tensor,
-                       Optional<AnySparseIterator>:$parentIter,
-                       LevelAttr:$loLvl, LevelAttr:$hiLvl);
-
-  let results = (outs AnySparseIterSpace:$resultSpace);
-
-  let summary = "Extract an iteration space from a sparse tensor between certain levels";
-  let description = [{
-      Extracts a `!sparse_tensor.iter_space` from a sparse tensor between
-      certian (consecutive) levels.
-
-      `tensor`: the input sparse tensor that defines the iteration space.
-      `parentIter`: the iterator for the previous level, at which the iteration space
-      at the current levels will be extracted.
-      `loLvl`, `hiLvl`: the level range between [loLvl, hiLvl) in the input tensor that
-      the returned iteration space covers. `hiLvl - loLvl` defines the dimension of the
-      iteration space.
-
-      Example:
-      ```mlir
-      // Extracts a 1-D iteration space from a COO tensor at level 1.
-      %space = sparse_tensor.iteration.extract_space %sp at %it1 lvls = 1
-        : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
-      ```
-  }];
-
-
-  let extraClassDeclaration = [{
-    std::pair<Level, Level> getLvlRange() {
-      return std::make_pair(getLoLvl(), getHiLvl());
-    }
-    unsigned getSpaceDim() {
-      return getHiLvl() - getLoLvl();
-    }
-    ArrayRef<::mlir::sparse_tensor::LevelType> getSpaceLvlTypes() {
-      return getResultSpace().getType().getLvlTypes();
-    }
-  }];
-
-  let hasVerifier = 1;
-  let assemblyFormat = "$tensor (`at` $parentIter^)? `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) "
-                       " attr-dict `:` type($tensor) (`,` type($parentIter)^)?";
-}
-
-def IterateOp : SparseTensor_Op<"iterate",
-    [RecursiveMemoryEffects, RecursivelySpeculatable,
-     DeclareOpInterfaceMethods<LoopLikeOpInterface,
-      ["getInitsMutable", "getLoopResults", "getRegionIterArgs",
-       "getYieldedValuesMutable"]>,
-     DeclareOpInterfaceMethods<RegionBranchOpInterface,
-      ["getEntrySuccessorOperands"]>,
-     SingleBlockImplicitTerminator<"sparse_tensor::YieldOp">]> {
-
-  let arguments = (ins AnySparseIterSpace:$iterSpace,
-                       Variadic<AnyType>:$initArgs,
-                       LevelSetAttr:$crdUsedLvls);
-  let results = (outs Variadic<AnyType>:$results);
-  let regions = (region SizedRegion<1>:$region);
-
-  let summary = "Iterate over a sparse iteration space";
-  let description = [{
-      The `sparse_tensor.iterate` operations represents a loop over the
-      provided iteration space extracted from a specific sparse tensor.
-      The operation defines an SSA value for a sparse iterator that points
-      to the current stored element in the sparse tensor and SSA values
-      for coordinates of the stored element. The coordinates are always
-      converted to `index` type despite of the underlying sparse tensor
-      storage. When coordinates are not used, the SSA values can be skipped
-      by `_` symbols, which usually leads to simpler generated code after
-      sparsification. For example:
-
-      ```mlir
-      // The coordinate for level 0 is not used when iterating over a 2-D
-      // iteration space.
-      %sparse_tensor.iterate %iterator in %space at(_, %crd_1)
-        : !sparse_tensor.iter_space<#CSR, lvls = 0 to 2>
-      ```
-
-      `sparse_tensor.iterate` can also operate on loop-carried variables
-      and returns the final values after loop termination.
-      The initial values of the variables are passed as additional SSA operands
-      to the iterator SSA value and used coordinate SSA values mentioned
-      above. The operation region has an argument for the iterator, variadic
-      arguments for specified (used) coordiates and followed by one argument
-      for each loop-carried variable, representing the value of the variable
-      at the current iteration.
-      The body region must contain exactly one block that terminates with
-      `sparse_tensor.yield`.
-
-      `sparse_tensor.iterate` results hold the final values after the last
-      iteration. If the `sparse_tensor.iterate` defines any values, a yield
-      must be explicitly present.
-      The number and types of the `sparse_tensor.iterate` results must match
-      the initial values in the iter_args binding and the yield operands.
-
-
-      A nested `sparse_tensor.iterate` example that prints all the coordinates
-      stored in the sparse input:
-
-      ```mlir
-      func.func @nested_iterate(%sp : tensor<4x8xf32, #COO>) {
-        // Iterates over the first level of %sp
-        %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-        %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd0)
-            : !sparse_tensor.iter_space<#COO, lvls = 0 to 1>  {
-          // Iterates over the second level of %sp
-          %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1
-              : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0 to 1>
-          %r2 = sparse_tensor.iterate %it2 in %l2 at (crd1)
-              : !sparse_tensor.iter_space<#COO, lvls = 1 to 2>  {
-             vector.print %crd0 : index
-             vector.print %crd1 : index
-          }
-        }
-      }
-
-      ```
-  }];
-
-  let extraClassDeclaration = [{
-    unsigned getSpaceDim() {
-      return getIterSpace().getType().getSpaceDim();
-    }
-    BlockArgument getIterator() {
-      return getRegion().getArguments().front();
-    }
-    Block::BlockArgListType getCrds() {
-      // The first block argument is iterator, the remaining arguments are
-      // referenced coordinates.
-      return getRegion().getArguments().slice(1, getCrdUsedLvls().count());
-    }
-    unsigned getNumRegionIterArgs() {
-      return getRegion().getArguments().size() - 1 - getCrdUsedLvls().count();
-    }
-  }];
-
-  let hasVerifier = 1;
-  let hasRegionVerifier = 1;
-  let hasCustomAssemblyFormat = 1;
-}
-
 //===----------------------------------------------------------------------===//
 // Sparse Tensor Debugging and Test-Only Operations.
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
index 264a0a5b3bee6..185cff46ae25d 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
@@ -72,99 +72,4 @@ def SparseTensorStorageSpecifier
     : Type<CPred<"::llvm::isa<::mlir::sparse_tensor::StorageSpecifierType>($_self)">, "metadata",
           "::mlir::sparse_tensor::StorageSpecifierType">;
 
-//===----------------------------------------------------------------------===//
-// Sparse Tensor Iteration Types.
-//===----------------------------------------------------------------------===//
-
-def SparseTensor_IterSpace : SparseTensor_Type<"IterSpace"> {
-  let mnemonic = "iter_space";
-
-  let description = [{
-    A sparse iteration space that represents an abstract N-D (sparse) iteration space
-    extracted from a sparse tensor.
-
-    Examples:
-
-    ```mlir
-    // An iteration space extracted from a CSR tensor between levels [0, 2).
-    !iter_space<#CSR, lvls = 0 to 2>
-    ```
-  }];
-
-  let parameters = (ins
-     SparseTensorEncodingAttr : $encoding,
-     "Level" : $loLvl,
-     "Level" : $hiLvl
-  );
-
-  let extraClassDeclaration = [{
-     /// The the dimension of the iteration space.
-     unsigned getSpaceDim() const {
-       return getHiLvl() - getLoLvl();
-     }
-
-     /// Get the level types for the iteration space.
-     ArrayRef<LevelType> getLvlTypes() const {
-       return getEncoding().getLvlTypes().slice(getLoLvl(), getSpaceDim());
-     }
-
-     /// Whether the iteration space is unique (i.e., no duplicated coordinate).
-     bool isUnique() {
-       return !getLvlTypes().back().isa<LevelPropNonDefault::Nonunique>();
-     }
-
-     /// Get the corresponding iterator type.
-     ::mlir::sparse_tensor::IteratorType getIteratorType() const;
-  }];
-
-  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
-}
-
-def SparseTensor_Iterator : SparseTensor_Type<"Iterator"> {
-  let mnemonic = "iterator";
-
-  let description = [{
-    An iterator that points to the current element in the corresponding iteration space.
-
-    Examples:
-
-    ```mlir
-    // An iterator that iterates over a iteration space of type `!iter_space<#CSR, lvls = 0 to 2>`
-    !iterator<#CSR, lvls = 0 to 2>
-    ```
-  }];
-
-  let parameters = (ins
-     SparseTensorEncodingAttr : $encoding,
-     "Level" : $loLvl,
-     "Level" : $hiLvl
-  );
-
-  let extraClassDeclaration = [{
-     /// Get the corresponding iteration space type.
-     ::mlir::sparse_tensor::IterSpaceType getIterSpaceType() const;
-
-     unsigned getSpaceDim() const { return getIterSpaceType().getSpaceDim(); }
-     ArrayRef<LevelType> getLvlTypes() const { return getIterSpaceType().getLvlTypes(); }
-     bool isUnique() { return getIterSpaceType().isUnique(); }
-  }];
-
-  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
-}
-
-def IsSparseSparseIterSpaceTypePred
-    : CPred<"::llvm::isa<::mlir::sparse_tensor::IterSpaceType>($_self)">;
-
-def IsSparseSparseIteratorTypePred
-    : CPred<"::llvm::isa<::mlir::sparse_tensor::IteratorType>($_self)">;
-
-def AnySparseIterSpace
-    : Type<IsSparseSparseIterSpaceTypePred, "sparse iteration space",
-          "::mlir::sparse_tensor::IterSpaceType">;
-
-def AnySparseIterator
-    : Type<IsSparseSparseIteratorTypePred, "sparse iterator",
-          "::mlir::sparse_tensor::IteratorType">;
-
-
 #endif // SPARSETENSOR_TYPES
diff --git a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
index 36908def09f40..e9058394d33da 100644
--- a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
+++ b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
@@ -30,14 +30,6 @@
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.cpp.inc"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrEnums.cpp.inc"
 
-// Forward declarations, following custom print/parsing methods are referenced
-// by the generated code for SparseTensorTypes.td.
-static mlir::ParseResult parseLevelRange(mlir::AsmParser &,
-                                         mlir::sparse_tensor::Level &,
-                                         mlir::sparse_tensor::Level &);
-static void printLevelRange(mlir::AsmPrinter &, mlir::sparse_tensor::Level,
-                            mlir::sparse_tensor::Level);
-
 #define GET_TYPEDEF_CLASSES
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.cpp.inc"
 
@@ -1961,363 +1953,6 @@ LogicalResult SortOp::verify() {
   return success();
 }
 
-//===----------------------------------------------------------------------===//
-// Sparse Tensor Iteration Operations.
-//===----------------------------------------------------------------------===//
-
-IterSpaceType IteratorType::getIterSpaceType() const {
-  return IterSpaceType::get(getContext(), getEncoding(), getLoLvl(),
-                            getHiLvl());
-}
-
-IteratorType IterSpaceType::getIteratorType() const {
-  return IteratorType::get(getContext(), getEncoding(), getLoLvl(), getHiLvl());
-}
-
-/// Parses a level range in the form "$lo `to` $hi"
-/// or simply "$lo" if $hi - $lo = 1
-static ParseResult parseLevelRange(AsmParser &parser, Level &lvlLo,
-                                   Level &lvlHi) {
-  if (parser.parseInteger(lvlLo))
-    return failure();
-
-  if (succeeded(parser.parseOptionalKeyword("to"))) {
-    if (parser.parseInteger(lvlHi))
-      return failure();
-  } else {
-    lvlHi = lvlLo + 1;
-  }
-
-  if (lvlHi <= lvlLo)
-    parser.emitError(parser.getNameLoc(),
-                     "expect larger level upper bound than lower bound");
-
-  return success();
-}
-
-/// Parses a level range in the form "$lo `to` $hi"
-/// or simply "$lo" if $hi - $lo = 1
-static ParseResult parseLevelRange(OpAsmParser &parser, IntegerAttr &lvlLoAttr,
-                                   IntegerAttr &lvlHiAttr) {
-  Level lvlLo, lvlHi;
-  if (parseLevelRange(parser, lvlLo, lvlHi))
-    return failure();
-
-  lvlLoAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlLo);
-  lvlHiAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlHi);
-  return success();
-}
-
-/// Prints a level range in the form "$lo `to` $hi"
-/// or simply "$lo" if $hi - $lo = 1
-static void printLevelRange(AsmPrinter &p, Level lo, Level hi) {
-
-  if (lo + 1 == hi)
-    p << lo;
-  else
-    p << lo << " to " << hi;
-}
-
-/// Prints a level range in the form "$lo `to` $hi"
-/// or simply "$lo" if $hi - $lo = 1
-static void printLevelRange(OpAsmPrinter &p, Operation *, IntegerAttr lvlLo,
-                            IntegerAttr lvlHi) {
-  unsigned lo = lvlLo.getValue().getZExtValue();
-  unsigned hi = lvlHi.getValue().getZExtValue();
-  printLevelRange(p, lo, hi);
-}
-
-static ParseResult
-parseSparseSpaceLoop(OpAsmParser &parser, OperationState &state,
-                     SmallVectorImpl<OpAsmParser::Argument> &iterators,
-                     SmallVectorImpl<OpAsmParser::Argument> &iterArgs) {
-  SmallVector<OpAsmParser::UnresolvedOperand> spaces;
-  SmallVector<OpAsmParser::UnresolvedOperand> initArgs;
-
-  // Parses "%iters, ... in %spaces, ..."
-  if (parser.parseArgumentList(iterators) || parser.parseKeyword("in") ||
-      parser.parseOperandList(spaces))
-    return failure();
-
-  if (iterators.size() != spaces.size())
-    return parser.emitError(
-        parser.getNameLoc(),
-        "mismatch in number of sparse iterators and sparse spaces");
-
-  // Parse "at(%crd0, _, ...)"
-  LevelSet crdUsedLvlSet;
-  bool hasUsedCrds = succeeded(parser.parseOptionalKeyword("at"));
-  unsigned lvlCrdCnt = 0;
-  if (hasUsedCrds) {
-    ParseResult crdList = parser.parseCommaSeparatedList(
-        OpAsmParser::Delimiter::Paren, [&]() -> ParseResult {
-          if (parser.parseOptionalKeyword("_")) {
-            if (parser.parseArgument(iterArgs.emplace_back()))
-              return failure();
-            // Always use IndexType for the coordinate.
-            crdUsedLvlSet.set(lvlCrdCnt);
-            iterArgs.back().type = parser.getBuilder().getIndexType();
-          }
-          lvlCrdCnt += 1;
-          return success();
-        });
-    if (failed(crdList)) {
-      return parser.emitError(
-          parser.getNameLoc(),
-          "expecting SSA value or \"_\" for level coordinates");
-    }
-  }
-  // Set the CrdUsedLvl bitset.
-  state.addAttribute("crdUsedLvls",
-                     parser.getBuilder().getI64IntegerAttr(crdUsedLvlSet));
-
-  // Parse "iter_args(%arg = %init, ...)"
-  bool hasIterArgs = succeeded(parser.parseOptionalKeyword("iter_args"));
-  if (hasIterArgs)
-    if (parser.parseAssignmentList(iterArgs, initArgs))
-      return failure();
-
-  SmallVector<Type> iterSpaceTps;
-  // parse ": sparse_tensor.iter_space -> ret"
-  if (parser.parseColon() || parser.parseTypeList(iterSpaceTps))
-    return failure();
-  if (iterSpaceTps.size() != spaces.size())
-    return parser.emitError(parser.getNameLoc(),
-                            "mismatch in number of iteration space operands "
-                            "and iteration space types");
-
-  for (auto [it, tp] : llvm::zip_equal(iterators, iterSpaceTps)) {
-    IterSpaceType spaceTp = llvm::dyn_cast<IterSpaceType>(tp);
-    if (!spaceTp)
-      return parser.emitError(parser.getNameLoc(),
-                              "expected sparse_tensor.iter_space type for "
-                              "iteration space operands");
-    if (hasUsedCrds && spaceTp.getSpaceDim() != lvlCrdCnt)
-      return parser.emitError(parser.getNameLoc(),
-                              "mismatch in number of iteration space dimension "
-                              "and specified coordinates");
-    it.type = spaceTp.getIteratorType();
-  }
-
-  if (hasIterArgs)
-    if (parser.parseArrowTypeList(state.types))
-      return failure();
-
-  // Resolves input operands.
-  if (parser.resolveOperands(spaces, iterSpaceTps, parser.getNameLoc(),
-                             state.operands))
-    return failure();
-
-  if (hasIterArgs) {
-    unsigned numCrds = crdUsedLvlSet.count();
-    // Strip off leading args that used for coordinates.
-    MutableArrayRef args = MutableArrayRef(iterArgs).drop_front(numCrds);
-    if (args.size() != initArgs.size() || args.size() != state.types.size()) {
-      return parser.emitError(
-          parser.getNameLoc(),
-          "mismatch in number of iteration arguments and return values");
-    }
-
-    for (auto [it, init, tp] : llvm::zip_equal(args, initArgs, state.types)) {
-      it.type = tp;
-      if (parser.resolveOperand(init, tp, state.operands))
-        return failure();
-    }
-  }
-  return success();
-}
-
-LogicalResult ExtractIterSpaceOp::inferReturnTypes(
-    MLIRContext *ctx, std::optional<Location> loc, ValueRange ops,
-    DictionaryAttr attr, OpaqueProperties prop, RegionRange region,
-    SmallVectorImpl<mlir::Type> &ret) {
-
-  ExtractIterSpaceOp::Adaptor adaptor(ops, attr, prop, region);
-  SparseTensorType stt = getSparseTensorType(adaptor.getTensor());
-  ret.push_back(IterSpaceType::get(ctx, stt.getEncoding(), adaptor.getLoLvl(),
-                                   adaptor.getHiLvl()));
-  return success();
-}
-
-LogicalResult ExtractIterSpaceOp::verify() {
-  if (getLoLvl() >= getHiLvl())
-    return emitOpError("expected smaller level low than level high");
-
-  TypedValue<IteratorType> pIter = getParentIter();
-  if ((pIter && getLoLvl() == 0) || (!pIter && getLoLvl() != 0)) {
-    return emitOpError(
-        "parent iterator should be specified iff level lower bound equals 0");
-  }
-
-  if (pIter) {
-    IterSpaceType spaceTp = getResultSpace().getType();
-    if (pIter.getType().getEncoding() != spaceTp.getEncoding())
-      return emitOpError(
-          "mismatch in parent iterator encoding and iteration space encoding.");
-
-    if (spaceTp.getLoLvl() != pIter.getType().getHiLvl())
-      return emitOpError("parent iterator should be used to extract an "
-                         "iteration space from a consecutive level.");
-  }
-
-  return success();
-}
-
-ParseResult IterateOp::parse(OpAsmParser &parser, OperationState &result) {
-  OpAsmParser::Argument iterator;
-  OpAsmParser::UnresolvedOperand iterSpace;
-
-  SmallVector<OpAsmParser::Argument> iters, iterArgs;
-  if (parseSparseSpaceLoop(parser, result, iters, iterArgs))
-    return failure();
-  if (iters.size() != 1)
-    return parser.emitError(parser.getNameLoc(),
-                            "expected only one iterator/iteration space");
-
-  iters.append(iterArgs);
-  Region *body = result.addRegion();
-  if (parser.parseRegion(*body, iters))
-    return failure();
-
-  IterateOp::ensureTerminator(*body, parser.getBuilder(), result.location);
-
-  // Parse the optional attribute list.
-  if (parser.parseOptionalAttrDict(result.attributes))
-    return failure();
-
-  return success();
-}
-
-/// Prints the initialization list in the form of
-///   <prefix>(%inner = %outer, %inner2 = %outer2, <...>)
-/// where 'inner' values are assumed to be region arguments and 'outer' values
-/// are regular SSA values.
-static void printInitializationList(OpAsmPrinter &p,
-                                    Block::BlockArgListType blocksArgs,
-                                    ValueRange initializers,
-                                    StringRef prefix = "") {
-  assert(blocksArgs.size() == initializers.size() &&
-         "expected same length of arguments and initializers");
-  if (initializers.empty())
-    return;
-
-  p << prefix << '(';
-  llvm::interleaveComma(llvm::zip(blocksArgs, initializers), p, [&](auto it) {
-    p << std::get<0>(it) << " = " << std::get<1>(it);
-  });
-  p << ")";
-}
-
-static void printUsedCrdsList(OpAsmPrinter &p, unsigned spaceDim,
-                              Block::BlockArgListType blocksArgs,
-                              LevelSet crdUsedLvls) {
-  if (crdUsedLvls.empty())
-    return;
-
-  p << " at(";
-  for (unsigned i = 0; i < spaceDim; i++) {
-    if (crdUsedLvls[i]) {
-      p << blocksArgs.front();
-      blocksArgs = blocksArgs.drop_front();
-    } else {
-      p << "_";
-    }
-    if (i != spaceDim - 1)
-      p << ", ";
-  }
-  assert(blocksArgs.empty());
-  p << ")";
-}
-
-void IterateOp::print(OpAsmPrinter &p) {
-  p << " " << getIterator() << " in " << getIterSpace();
-  printUsedCrdsList(p, getSpaceDim(), getCrds(), getCrdUsedLvls());
-  printInitializationList(p, getRegionIterArgs(), getInitArgs(), " iter_args");
-
-  p << " : " << getIterSpace().getType() << " ";
-  if (!getInitArgs().empty())
-    p << "-> (" << getInitArgs().getTypes() << ") ";
-
-  p.printRegion(getRegion(), /*printEntryBlockArgs=*/false,
-                /*printBlockTerminators=*/!getInitArgs().empty());
-}
-
-LogicalResult IterateOp::verify() {
-  if (getInitArgs().size() != getNumResults()) {
-    return emitOpError(
-        "mismatch in number of loop-carried values and defined values");
-  }
-  return success();
-}
-
-LogicalResult IterateOp::verifyRegions() {
-  if (getIterator().getType() != getIterSpace().getType().getIteratorType())
-    return emitOpError("mismatch in iterator and iteration space type");
-  if (getNumRegionIterArgs() != getNumResults())
-    return emitOpError(
-        "mismatch in number of basic block args and defined values");
-
-  auto initArgs = getInitArgs();
-  auto iterArgs = getRegionIterArgs();
-  auto yieldVals = getYieldedValues();
-  auto opResults = getResults();
-  if (!llvm::all_equal({initArgs.size(), iterArgs.size(), yieldVals.size(),
-                        opResults.size()})) {
-    return emitOpError() << "number mismatch between iter args and results.";
-  }
-
-  for (auto [i, init, iter, yield, ret] :
-       llvm::enumerate(initArgs, iterArgs, yieldVals, opResults)) {
-    if (init.getType() != ret.getType())
-      return emitOpError() << "types mismatch between " << i
-                           << "th iter operand and defined value";
-    if (iter.getType() != ret.getType())
-      return emitOpError() << "types mismatch between " << i
-                           << "th iter region arg and defined value";
-    if (yield.getType() != ret.getType())
-      return emitOpError() << "types mismatch between " << i
-                           << "th yield value and defined value";
-  }
-
-  return success();
-}
-
-/// IterateOp implemented OpInterfaces' methods.
-SmallVector<Region *> IterateOp::getLoopRegions() { return {&getRegion()}; }
-
-MutableArrayRef<OpOperand> IterateOp::getInitsMutable() {
-  return getInitArgsMutable();
-}
-
-Block::BlockArgListType IterateOp::getRegionIterArgs() {
-  return getRegion().getArguments().take_back(getNumRegionIterArgs());
-}
-
-std::optional<MutableArrayRef<OpOperand>> IterateOp::getYieldedValuesMutable() {
-  return cast<sparse_tensor::YieldOp>(
-             getRegion().getBlocks().front().getTerminator())
-      .getResultsMutable();
-}
-
-std::optional<ResultRange> IterateOp::getLoopResults() { return getResults(); }
-
-OperandRange IterateOp::getEntrySuccessorOperands(RegionBranchPoint point) {
-  return getInitArgs();
-}
-
-void IterateOp::getSuccessorRegions(RegionBranchPoint point,
-                                    SmallVectorImpl<RegionSuccessor> &regions) {
-  // Both the operation itself and the region may be branching into the body or
-  // back into the operation itself.
-  regions.push_back(RegionSuccessor(&getRegion(), getRegionIterArgs()));
-  // It is possible for loop not to enter the body.
-  regions.push_back(RegionSuccessor(getResults()));
-}
-
-//===----------------------------------------------------------------------===//
-// Sparse Tensor Dialect Setups.
-//===----------------------------------------------------------------------===//
-
 /// Materialize a single constant operation from a given attribute value with
 /// the desired resultant type.
 Operation *SparseTensorDialect::materializeConstant(OpBuilder &builder,
diff --git a/mlir/test/Dialect/SparseTensor/invalid.mlir b/mlir/test/Dialect/SparseTensor/invalid.mlir
index b13024cd4ed99..7f5c05190fc9a 100644
--- a/mlir/test/Dialect/SparseTensor/invalid.mlir
+++ b/mlir/test/Dialect/SparseTensor/invalid.mlir
@@ -1012,142 +1012,3 @@ func.func @sparse_print(%arg0: tensor<10x10xf64>) {
   sparse_tensor.print %arg0 : tensor<10x10xf64>
   return
 }
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 2>) {
-  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' expect larger level upper bound than lower bound}}
-  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 to 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 2>
-  return
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
-  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
-  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
-  return
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>) {
-  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 1 : tensor<4x8xf32, #COO>
-  return
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-#CSR = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : dense,
-    j : compressed
-  )
-}>
-
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#CSR, lvls = 0>) {
-  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op mismatch in parent iterator encoding and iteration space encoding.}}
-  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#CSR, lvls = 0>
-  return
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
-  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be used to extract an iteration space from a consecutive level.}}
-  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
-  return
-}
-
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-  // expected-error @+1 {{'sparse_tensor.iterate' op different number of region iter_args and yielded values: 2 != 1}}
-  %r1, %r2 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%si = %i, %sj = %j): !sparse_tensor.iter_space<#COO, lvls = 0> -> (index, index) {
-    sparse_tensor.yield %si : index
-  }
-  return %r1 : index
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-// expected-note@+1 {{prior use here}}
-func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index) -> f32 {
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-  // expected-error @+1 {{use of value '%i' expects different type than prior uses: 'f32' vs 'index'}}
-  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%outer = %i): !sparse_tensor.iter_space<#COO, lvls = 0> -> f32 {
-    sparse_tensor.yield %outer : f32
-  }
-  return %r1 : f32
-}
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-  // expected-error @+1 {{'sparse_tensor.iterate' op 0-th region iter_arg and 0-th yielded value have different type: 'index' != 'f32'}}
-  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%si = %i): !sparse_tensor.iter_space<#COO, lvls = 0> -> index {
-    %y = arith.constant 1.0 :  f32
-    sparse_tensor.yield %y : f32
-  }
-  return %r1 : index
-}
diff --git a/mlir/test/Dialect/SparseTensor/roundtrip.mlir b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
index e9a898f16b41d..12f69c1d37b9c 100644
--- a/mlir/test/Dialect/SparseTensor/roundtrip.mlir
+++ b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
@@ -738,56 +738,3 @@ func.func @sparse_has_runtime() -> i1 {
   %has_runtime = sparse_tensor.has_runtime_library
   return %has_runtime : i1
 }
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-// CHECK-LABEL:   func.func @sparse_extract_iter_space(
-// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf32, #sparse{{[0-9]*}}>,
-// CHECK-SAME:      %[[VAL_1:.*]]: !sparse_tensor.iterator<#sparse{{[0-9]*}}, lvls = 0>)
-// CHECK:           %[[VAL_2:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] lvls = 0
-// CHECK:           %[[VAL_3:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] at %[[VAL_1]] lvls = 1
-// CHECK:           return %[[VAL_2]], %[[VAL_3]] : !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 0>, !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 1>
-// CHECK:         }
-func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>)
-  -> (!sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>) {
-  // Extracting the iteration space for the first level needs no parent iterator.
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-  // Extracting the iteration space for the second level needs a parent iterator.
-  %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
-  return %l1, %l2 : !sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>
-}
-
-
-// -----
-
-#COO = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : compressed(nonunique),
-    j : singleton(soa)
-  )
-}>
-
-// CHECK-LABEL:   func.func @sparse_iterate(
-// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf32, #sparse{{[0-9]*}}>,
-// CHECK-SAME:      %[[VAL_1:.*]]: index,
-// CHECK-SAME:      %[[VAL_2:.*]]: index) -> index {
-// CHECK:           %[[VAL_3:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] lvls = 0 : tensor<4x8xf32, #sparse{{[0-9]*}}>
-// CHECK:           %[[VAL_4:.*]] = sparse_tensor.iterate %[[VAL_5:.*]] in %[[VAL_3]] at(%[[VAL_6:.*]]) iter_args(%[[VAL_7:.*]] = %[[VAL_1]]) : !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 0> -> (index) {
-// CHECK:             sparse_tensor.yield %[[VAL_7]] : index
-// CHECK:           }
-// CHECK:           return %[[VAL_4]] : index
-// CHECK:         }
-func.func @sparse_iterate(%sp : tensor<4x8xf32, #COO>, %i : index, %j : index) -> index {
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
-  %r1 = sparse_tensor.iterate %it1 in %l1 at (%crd) iter_args(%outer = %i): !sparse_tensor.iter_space<#COO, lvls = 0 to 1> -> index {
-    sparse_tensor.yield %outer : index
-  }
-  return %r1 : index
-}
diff --git a/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir b/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir
deleted file mode 100644
index e7158d04b37fe..0000000000000
--- a/mlir/test/Dialect/SparseTensor/sparse_itertion_licm.mlir
+++ /dev/null
@@ -1,26 +0,0 @@
-// RUN: mlir-opt %s --loop-invariant-code-motion | FileCheck %s
-
-#CSR = #sparse_tensor.encoding<{
-  map = (i, j) -> (
-    i : dense,
-    j : compressed
-  )
-}>
-
-// Make sure that pure instructions are hoisted outside the loop.
-//
-// CHECK: sparse_tensor.values
-// CHECK: sparse_tensor.positions
-// CHECK: sparse_tensor.coordinate
-// CHECK: sparse_tensor.iterate
-func.func @sparse_iterate(%sp : tensor<?x?xf64, #CSR>) {
-  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<?x?xf64, #CSR>
-  sparse_tensor.iterate %it1 in %l1 at (%crd) : !sparse_tensor.iter_space<#CSR, lvls = 0> {
-    %0 = sparse_tensor.values %sp : tensor<?x?xf64, #CSR> to memref<?xf64>
-    %1 = sparse_tensor.positions %sp { level = 1 : index } : tensor<?x?xf64, #CSR> to memref<?xindex>
-    %2 = sparse_tensor.coordinates  %sp { level = 1 : index } : tensor<?x?xf64, #CSR> to memref<?xindex>
-    "test.op"(%0, %1, %2) : (memref<?xf64>, memref<?xindex>, memref<?xindex>) -> ()
-  }
-
-  return
-}

From 481bd5d416df7a1d24e18cc81ae782e8701de965 Mon Sep 17 00:00:00 2001
From: Peiming Liu <peiming@google.com>
Date: Tue, 16 Apr 2024 11:32:30 -0700
Subject: [PATCH 636/886] [mlir][sparse] introduce
 `sparse_tensor.extract_iteration_space` operation. (#88554)

A `sparse_tensor.extract_space %tensor at %iterator` extracts a *sparse*
iteration space defined `%tensor`, the operation to traverse the
iteration space will be introduced in following PRs.
---
 .../SparseTensor/IR/SparseTensorOps.td        |  60 ++++++++++
 .../SparseTensor/IR/SparseTensorTypes.td      |  97 +++++++++++++++
 .../SparseTensor/IR/SparseTensorDialect.cpp   | 110 ++++++++++++++++++
 mlir/test/Dialect/SparseTensor/invalid.mlir   |  82 +++++++++++++
 mlir/test/Dialect/SparseTensor/roundtrip.mlir |  25 ++++
 5 files changed, 374 insertions(+)

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
index 0cfc64f9988a0..d7121e8320a4b 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
@@ -1430,6 +1430,66 @@ def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
   let hasVerifier = 1;
 }
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Operations.
+//===----------------------------------------------------------------------===//
+
+def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
+    [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
+
+  let arguments = (ins AnySparseTensor:$tensor,
+                       Optional<AnySparseIterator>:$parentIter,
+                       LevelAttr:$loLvl, LevelAttr:$hiLvl);
+
+  let results = (outs AnySparseIterSpace:$resultSpace);
+
+  let summary = "Extracts an iteration space from a sparse tensor between certain levels";
+  let description = [{
+      Extracts a `!sparse_tensor.iter_space` from a sparse tensor between
+      certain (consecutive) levels. For sparse levels, it is usually done by
+      loading a postion range from the underlying sparse tensor storage.
+      E.g., for a compressed level, the iteration space is extracted by
+      [pos[i], pos[i+1]) supposing the the parent iterator points at `i`.
+
+      `tensor`: the input sparse tensor that defines the iteration space.
+      `parentIter`: the iterator for the previous level, at which the iteration space
+      at the current levels will be extracted.
+      `loLvl`, `hiLvl`: the level range between [loLvl, hiLvl) in the input tensor that
+      the returned iteration space covers. `hiLvl - loLvl` defines the dimension of the
+      iteration space.
+
+      The type of returned the value is automatically inferred to
+      `!sparse_tensor.iter_space<#INPUT_ENCODING, lvls = $loLvl to $hiLvl>`.
+      The returned iteration space can then be iterated over by
+      `sparse_tensor.iterate` operations to visit every stored element
+      (usually nonzeros) in the input sparse tensor.
+
+      Example:
+      ```mlir
+      // Extracts a 1-D iteration space from a COO tensor at level 1.
+      %space = sparse_tensor.iteration.extract_space %sp at %it1 lvls = 1
+        : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+      ```
+  }];
+
+
+  let extraClassDeclaration = [{
+    std::pair<Level, Level> getLvlRange() {
+      return std::make_pair(getLoLvl(), getHiLvl());
+    }
+    unsigned getSpaceDim() {
+      return getHiLvl() - getLoLvl();
+    }
+    ArrayRef<::mlir::sparse_tensor::LevelType> getSpaceLvlTypes() {
+      return getResultSpace().getType().getLvlTypes();
+    }
+  }];
+
+  let hasVerifier = 1;
+  let assemblyFormat = "$tensor (`at` $parentIter^)? `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) "
+                       " attr-dict `:` type($tensor) (`,` type($parentIter)^)?";
+}
+
 //===----------------------------------------------------------------------===//
 // Sparse Tensor Debugging and Test-Only Operations.
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
index 185cff46ae25d..79113d8778743 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorTypes.td
@@ -72,4 +72,101 @@ def SparseTensorStorageSpecifier
     : Type<CPred<"::llvm::isa<::mlir::sparse_tensor::StorageSpecifierType>($_self)">, "metadata",
           "::mlir::sparse_tensor::StorageSpecifierType">;
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Types.
+//===----------------------------------------------------------------------===//
+
+def SparseTensor_IterSpace : SparseTensor_Type<"IterSpace"> {
+  let mnemonic = "iter_space";
+
+  let description = [{
+    A sparse iteration space that represents an abstract N-D (sparse) iteration space
+    extracted from a sparse tensor, i.e., a set of (crd_0, crd_1, ..., crd_N) for
+    every stored element (usually nonzeros) in a sparse tensor between the specified
+    [$loLvl, $hiLvl) levels.
+
+    Examples:
+
+    ```mlir
+    // An iteration space extracted from a CSR tensor between levels [0, 2).
+    !iter_space<#CSR, lvls = 0 to 2>
+    ```
+  }];
+
+  let parameters = (ins
+     SparseTensorEncodingAttr : $encoding,
+     "Level" : $loLvl,
+     "Level" : $hiLvl
+  );
+
+  let extraClassDeclaration = [{
+     /// The the dimension of the iteration space.
+     unsigned getSpaceDim() const {
+       return getHiLvl() - getLoLvl();
+     }
+
+     /// Get the level types for the iteration space.
+     ArrayRef<LevelType> getLvlTypes() const {
+       return getEncoding().getLvlTypes().slice(getLoLvl(), getSpaceDim());
+     }
+
+     /// Whether the iteration space is unique (i.e., no duplicated coordinate).
+     bool isUnique() {
+       return !getLvlTypes().back().isa<LevelPropNonDefault::Nonunique>();
+     }
+
+     /// Get the corresponding iterator type.
+     ::mlir::sparse_tensor::IteratorType getIteratorType() const;
+  }];
+
+  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
+}
+
+def SparseTensor_Iterator : SparseTensor_Type<"Iterator"> {
+  let mnemonic = "iterator";
+
+  let description = [{
+    An iterator that points to the current element in the corresponding iteration space.
+
+    Examples:
+
+    ```mlir
+    // An iterator that iterates over a iteration space of type `!iter_space<#CSR, lvls = 0 to 2>`
+    !iterator<#CSR, lvls = 0 to 2>
+    ```
+  }];
+
+  let parameters = (ins
+     SparseTensorEncodingAttr : $encoding,
+     "Level" : $loLvl,
+     "Level" : $hiLvl
+  );
+
+  let extraClassDeclaration = [{
+     /// Get the corresponding iteration space type.
+     ::mlir::sparse_tensor::IterSpaceType getIterSpaceType() const;
+
+     unsigned getSpaceDim() const { return getIterSpaceType().getSpaceDim(); }
+     ArrayRef<LevelType> getLvlTypes() const { return getIterSpaceType().getLvlTypes(); }
+     bool isUnique() { return getIterSpaceType().isUnique(); }
+  }];
+
+  let assemblyFormat="`<` $encoding `,` `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) `>`";
+}
+
+def IsSparseSparseIterSpaceTypePred
+    : CPred<"::llvm::isa<::mlir::sparse_tensor::IterSpaceType>($_self)">;
+
+def IsSparseSparseIteratorTypePred
+    : CPred<"::llvm::isa<::mlir::sparse_tensor::IteratorType>($_self)">;
+
+def AnySparseIterSpace
+    : Type<IsSparseSparseIterSpaceTypePred, "sparse iteration space",
+          "::mlir::sparse_tensor::IterSpaceType">;
+
+def AnySparseIterator
+    : Type<IsSparseSparseIteratorTypePred, "sparse iterator",
+          "::mlir::sparse_tensor::IteratorType">;
+
+
 #endif // SPARSETENSOR_TYPES
diff --git a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
index e9058394d33da..516b0943bdcfa 100644
--- a/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
+++ b/mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp
@@ -30,6 +30,14 @@
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrDefs.cpp.inc"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorAttrEnums.cpp.inc"
 
+// Forward declarations, following custom print/parsing methods are referenced
+// by the generated code for SparseTensorTypes.td.
+static mlir::ParseResult parseLevelRange(mlir::AsmParser &,
+                                         mlir::sparse_tensor::Level &,
+                                         mlir::sparse_tensor::Level &);
+static void printLevelRange(mlir::AsmPrinter &, mlir::sparse_tensor::Level,
+                            mlir::sparse_tensor::Level);
+
 #define GET_TYPEDEF_CLASSES
 #include "mlir/Dialect/SparseTensor/IR/SparseTensorTypes.cpp.inc"
 
@@ -1953,6 +1961,108 @@ LogicalResult SortOp::verify() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// Sparse Tensor Iteration Operations.
+//===----------------------------------------------------------------------===//
+
+IterSpaceType IteratorType::getIterSpaceType() const {
+  return IterSpaceType::get(getContext(), getEncoding(), getLoLvl(),
+                            getHiLvl());
+}
+
+IteratorType IterSpaceType::getIteratorType() const {
+  return IteratorType::get(getContext(), getEncoding(), getLoLvl(), getHiLvl());
+}
+
+/// Parses a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static ParseResult parseLevelRange(AsmParser &parser, Level &lvlLo,
+                                   Level &lvlHi) {
+  if (parser.parseInteger(lvlLo))
+    return failure();
+
+  if (succeeded(parser.parseOptionalKeyword("to"))) {
+    if (parser.parseInteger(lvlHi))
+      return failure();
+  } else {
+    lvlHi = lvlLo + 1;
+  }
+
+  if (lvlHi <= lvlLo)
+    parser.emitError(parser.getNameLoc(),
+                     "expect larger level upper bound than lower bound");
+
+  return success();
+}
+
+/// Parses a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static ParseResult parseLevelRange(OpAsmParser &parser, IntegerAttr &lvlLoAttr,
+                                   IntegerAttr &lvlHiAttr) {
+  Level lvlLo, lvlHi;
+  if (parseLevelRange(parser, lvlLo, lvlHi))
+    return failure();
+
+  lvlLoAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlLo);
+  lvlHiAttr = IntegerAttr::get(parser.getBuilder().getIndexType(), lvlHi);
+  return success();
+}
+
+/// Prints a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static void printLevelRange(AsmPrinter &p, Level lo, Level hi) {
+
+  if (lo + 1 == hi)
+    p << lo;
+  else
+    p << lo << " to " << hi;
+}
+
+/// Prints a level range in the form "$lo `to` $hi"
+/// or simply "$lo" if $hi - $lo = 1
+static void printLevelRange(OpAsmPrinter &p, Operation *, IntegerAttr lvlLo,
+                            IntegerAttr lvlHi) {
+  unsigned lo = lvlLo.getValue().getZExtValue();
+  unsigned hi = lvlHi.getValue().getZExtValue();
+  printLevelRange(p, lo, hi);
+}
+
+LogicalResult ExtractIterSpaceOp::inferReturnTypes(
+    MLIRContext *ctx, std::optional<Location> loc, ValueRange ops,
+    DictionaryAttr attr, OpaqueProperties prop, RegionRange region,
+    SmallVectorImpl<mlir::Type> &ret) {
+
+  ExtractIterSpaceOp::Adaptor adaptor(ops, attr, prop, region);
+  SparseTensorType stt = getSparseTensorType(adaptor.getTensor());
+  ret.push_back(IterSpaceType::get(ctx, stt.getEncoding(), adaptor.getLoLvl(),
+                                   adaptor.getHiLvl()));
+  return success();
+}
+
+LogicalResult ExtractIterSpaceOp::verify() {
+  if (getLoLvl() >= getHiLvl())
+    return emitOpError("expected smaller level low than level high");
+
+  TypedValue<IteratorType> pIter = getParentIter();
+  if ((pIter && getLoLvl() == 0) || (!pIter && getLoLvl() != 0)) {
+    return emitOpError(
+        "parent iterator should be specified iff level lower bound equals 0");
+  }
+
+  if (pIter) {
+    IterSpaceType spaceTp = getResultSpace().getType();
+    if (pIter.getType().getEncoding() != spaceTp.getEncoding())
+      return emitOpError(
+          "mismatch in parent iterator encoding and iteration space encoding.");
+
+    if (spaceTp.getLoLvl() != pIter.getType().getHiLvl())
+      return emitOpError("parent iterator should be used to extract an "
+                         "iteration space from a consecutive level.");
+  }
+
+  return success();
+}
+
 /// Materialize a single constant operation from a given attribute value with
 /// the desired resultant type.
 Operation *SparseTensorDialect::materializeConstant(OpBuilder &builder,
diff --git a/mlir/test/Dialect/SparseTensor/invalid.mlir b/mlir/test/Dialect/SparseTensor/invalid.mlir
index 7f5c05190fc9a..3fa696e1600a9 100644
--- a/mlir/test/Dialect/SparseTensor/invalid.mlir
+++ b/mlir/test/Dialect/SparseTensor/invalid.mlir
@@ -1012,3 +1012,85 @@ func.func @sparse_print(%arg0: tensor<10x10xf64>) {
   sparse_tensor.print %arg0 : tensor<10x10xf64>
   return
 }
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 2>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' expect larger level upper bound than lower bound}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 to 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 2>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 0 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be specified iff level lower bound equals 0}}
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 1 : tensor<4x8xf32, #COO>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+#CSR = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : dense,
+    j : compressed
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#CSR, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op mismatch in parent iterator encoding and iteration space encoding.}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#CSR, lvls = 0>
+  return
+}
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>) {
+  // expected-error@+1 {{'sparse_tensor.extract_iteration_space' op parent iterator should be used to extract an iteration space from a consecutive level.}}
+  %l1 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 2 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return
+}
diff --git a/mlir/test/Dialect/SparseTensor/roundtrip.mlir b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
index 12f69c1d37b9c..d34071279e512 100644
--- a/mlir/test/Dialect/SparseTensor/roundtrip.mlir
+++ b/mlir/test/Dialect/SparseTensor/roundtrip.mlir
@@ -738,3 +738,28 @@ func.func @sparse_has_runtime() -> i1 {
   %has_runtime = sparse_tensor.has_runtime_library
   return %has_runtime : i1
 }
+
+// -----
+
+#COO = #sparse_tensor.encoding<{
+  map = (i, j) -> (
+    i : compressed(nonunique),
+    j : singleton(soa)
+  )
+}>
+
+// CHECK-LABEL:   func.func @sparse_extract_iter_space(
+// CHECK-SAME:      %[[VAL_0:.*]]: tensor<4x8xf32, #sparse{{[0-9]*}}>,
+// CHECK-SAME:      %[[VAL_1:.*]]: !sparse_tensor.iterator<#sparse{{[0-9]*}}, lvls = 0>)
+// CHECK:           %[[VAL_2:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] lvls = 0
+// CHECK:           %[[VAL_3:.*]] = sparse_tensor.extract_iteration_space %[[VAL_0]] at %[[VAL_1]] lvls = 1
+// CHECK:           return %[[VAL_2]], %[[VAL_3]] : !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 0>, !sparse_tensor.iter_space<#sparse{{[0-9]*}}, lvls = 1>
+// CHECK:         }
+func.func @sparse_extract_iter_space(%sp : tensor<4x8xf32, #COO>, %it1 : !sparse_tensor.iterator<#COO, lvls = 0>)
+  -> (!sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>) {
+  // Extracting the iteration space for the first level needs no parent iterator.
+  %l1 = sparse_tensor.extract_iteration_space %sp lvls = 0 : tensor<4x8xf32, #COO>
+  // Extracting the iteration space for the second level needs a parent iterator.
+  %l2 = sparse_tensor.extract_iteration_space %sp at %it1 lvls = 1 : tensor<4x8xf32, #COO>, !sparse_tensor.iterator<#COO, lvls = 0>
+  return %l1, %l2 : !sparse_tensor.iter_space<#COO, lvls = 0>, !sparse_tensor.iter_space<#COO, lvls = 1>
+}

From edb711d2f318b17489692b5f85028fab7ed85b83 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Tue, 16 Apr 2024 13:35:23 -0500
Subject: [PATCH 637/886] [InstCombine] Update `vector_reduce_and` tests to
 actually use `llvm.vector.reduce.and`; NFC

---
 llvm/test/Transforms/InstCombine/known-bits.ll | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/llvm/test/Transforms/InstCombine/known-bits.ll b/llvm/test/Transforms/InstCombine/known-bits.ll
index 85a21332b0788..8b4249b2c25a9 100644
--- a/llvm/test/Transforms/InstCombine/known-bits.ll
+++ b/llvm/test/Transforms/InstCombine/known-bits.ll
@@ -1223,7 +1223,7 @@ define i8 @known_reduce_and(<2 x i8> %xx) {
 ; CHECK-NEXT:    ret i8 1
 ;
   %x = or <2 x i8> %xx, <i8 5, i8 3>
-  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %v = call i8 @llvm.vector.reduce.and(<2 x i8> %x)
   %r = and i8 %v, 1
   ret i8 %r
 }
@@ -1231,12 +1231,12 @@ define i8 @known_reduce_and(<2 x i8> %xx) {
 define i8 @known_reduce_and_fail(<2 x i8> %xx) {
 ; CHECK-LABEL: @known_reduce_and_fail(
 ; CHECK-NEXT:    [[X:%.*]] = or <2 x i8> [[XX:%.*]], <i8 5, i8 3>
-; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.or.v2i8(<2 x i8> [[X]])
+; CHECK-NEXT:    [[V:%.*]] = call i8 @llvm.vector.reduce.and.v2i8(<2 x i8> [[X]])
 ; CHECK-NEXT:    [[R:%.*]] = and i8 [[V]], 2
 ; CHECK-NEXT:    ret i8 [[R]]
 ;
   %x = or <2 x i8> %xx, <i8 5, i8 3>
-  %v = call i8 @llvm.vector.reduce.or(<2 x i8> %x)
+  %v = call i8 @llvm.vector.reduce.and(<2 x i8> %x)
   %r = and i8 %v, 2
   ret i8 %r
 }

From 5c6af605b307213453a9a043532b9293db21b5c6 Mon Sep 17 00:00:00 2001
From: mahtohappy <Happy.Kumar@Windriver.com>
Date: Wed, 17 Apr 2024 00:12:14 +0530
Subject: [PATCH 638/886] [Clang][Sema] placement new initializes typedef array
 with correct size (#88902)

Build Failure Fix
Fixes build failures due to #83124
---
 .../{instantiate-new-placement-size.cpp => PR41441.cpp}       | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
 rename clang/test/SemaCXX/{instantiate-new-placement-size.cpp => PR41441.cpp} (75%)

diff --git a/clang/test/SemaCXX/instantiate-new-placement-size.cpp b/clang/test/SemaCXX/PR41441.cpp
similarity index 75%
rename from clang/test/SemaCXX/instantiate-new-placement-size.cpp
rename to clang/test/SemaCXX/PR41441.cpp
index 7a29d3dee8491..0b012b33fce34 100644
--- a/clang/test/SemaCXX/instantiate-new-placement-size.cpp
+++ b/clang/test/SemaCXX/PR41441.cpp
@@ -1,5 +1,5 @@
-// RUN: %clang -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
-// Issue no: 41441
+// RUN: %clang --target=x86_64-pc-linux -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
+
 #include <new>
 
 // CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)

From b01879ec1ffbd249f9bf3c4f32308443be6ac36b Mon Sep 17 00:00:00 2001
From: Chao Chen <116223022+chencha3@users.noreply.github.com>
Date: Tue, 16 Apr 2024 13:44:14 -0500
Subject: [PATCH 639/886] [MLIR][XeGPU] Add XeGPU scattered ops (#86594)

- Extended TensorDescAttr with scattered attribute
- Add scattered ops: CreateDescOp, PrefetchOp, LoadGatherOp,
StoreScatterOp, UpdateOffsetOp
- Add a block op: UpdateNdOffsetOp

---------

Co-authored-by: Mehdi Amini <joker.eph@gmail.com>
Co-authored-by: Adam Siemieniuk <adam.siemieniuk@intel.com>
---
 mlir/include/mlir/Dialect/XeGPU/IR/XeGPU.h    |   1 +
 .../mlir/Dialect/XeGPU/IR/XeGPUAttrs.td       |  42 +-
 .../include/mlir/Dialect/XeGPU/IR/XeGPUOps.td | 464 +++++++++++++++---
 .../mlir/Dialect/XeGPU/IR/XeGPUTypes.td       |  41 +-
 mlir/lib/Dialect/XeGPU/IR/XeGPUDialect.cpp    |  21 +
 mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp        | 254 +++++++++-
 mlir/test/Dialect/XeGPU/XeGPUOps.mlir         |  62 +++
 mlir/test/Dialect/XeGPU/invalid.mlir          | 159 ++++++
 8 files changed, 937 insertions(+), 107 deletions(-)
 create mode 100644 mlir/test/Dialect/XeGPU/invalid.mlir

diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPU.h b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPU.h
index 87aabdc015fea..eca9255ff3974 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPU.h
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPU.h
@@ -12,6 +12,7 @@
 #include "mlir/Bytecode/BytecodeOpInterface.h"
 #include "mlir/IR/BuiltinTypes.h"
 #include "mlir/IR/Dialect.h"
+#include "mlir/IR/TypeUtilities.h"
 #include "mlir/Interfaces/ShapedOpInterfaces.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 #include "mlir/Interfaces/ViewLikeInterface.h"
diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUAttrs.td b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUAttrs.td
index cd38549f1ccf4..6579d07ec2621 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUAttrs.td
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUAttrs.td
@@ -19,17 +19,36 @@ class XeGPUAttr<string name, string attrMnemonic, list<Trait> traits = [],
 }
 
 def XeGPU_TensorDescAttr: XeGPUAttr<"TensorDesc", "tdesc_attr"> {
+  let summary = [{a composite attribute for `TensorDescType`}];
+  let description = [{`TensorDescAttr` (or `tdesc_attr`) is a composite
+    attribute defined for `TensorDescType` for describing following
+    properties of a `TensorDesc`.
+    1. `memory_scope`: It describes where the data block described by the
+        TensorDesc is located, `Global` device memory or `Shared` local memory.
+        It is default to `Global`.
+    2. `array_length`: It describes how many horizontally consecutive blocks
+        will be loaded by a hardware load instruction. If the TensorDesc shape
+        is 8x16, with array_length = 2. The loaded block shape will be acctually
+        8x32. Its default value is 1.
+    3. `boundary_check`: It is used to indicates the hardware whether to do
+        out-of-boundary check. The default value is true.
+    4. `scattered`: It is used to differenciate TensorDescs created from
+       `create_nd_tdesc` vs from `create_tdesc`.
+  }];
+
   let parameters = (ins
     OptionalParameter<"MemoryScopeAttr">: $memory_scope,
     OptionalParameter<"IntegerAttr", "1">: $array_length,
-    OptionalParameter<"BoolAttr", "true">: $boundary_check
+    OptionalParameter<"BoolAttr", "true">: $boundary_check,
+    OptionalParameter<"BoolAttr", "false">: $scattered
   );
 
   let builders = [
     AttrBuilder<(ins
       CArg<"xegpu::MemoryScope", "xegpu::MemoryScope::Global">:$memory_scope,
       CArg<"int", "1">:$array_length,
-      CArg<"bool", "true">: $boundary_check
+      CArg<"bool", "true">: $boundary_check,
+      CArg<"bool", "false">: $scattered
     )>
   ];
 
@@ -41,15 +60,17 @@ def XeGPU_TensorDescAttr: XeGPUAttr<"TensorDesc", "tdesc_attr"> {
 //===----------------------------------------------------------------------===//
 def XeGPU_MemoryScopeGlobal: I32EnumAttrCase<"Global", 0, "global">;
 def XeGPU_MemoryScopeShared: I32EnumAttrCase<"SLM", 1, "slm">;
-def XeGPU_MemoryScope: I32EnumAttr<"MemoryScope", 
-      "The address space of the memory the tensor descritor is created for", 
+def XeGPU_MemoryScope: I32EnumAttr<"MemoryScope",
+      "The address space of the memory the tensor descritor is created for",
       [XeGPU_MemoryScopeGlobal, XeGPU_MemoryScopeShared]> {
   let genSpecializedAttr = 0;
   let cppNamespace = "::mlir::xegpu";
 }
 
-def XeGPU_MemoryScopeAttr: 
+def XeGPU_MemoryScopeAttr:
   EnumAttr<XeGPU_Dialect, XeGPU_MemoryScope, "memory_scope"> {
+    let summary = [{Describe the location of data described by a `TensorDesc`:
+                 Global device memory (`Global`) or Shared local memory (`SLM`).}];
     let assemblyFormat = "$value";
 }
 
@@ -63,19 +84,18 @@ def XeGPU_CachePolicyInvalid:       I32EnumAttrCase<"READ_INVALIDATE", 3, "read_
 def XeGPU_CachePolicyWriteBack:     I32EnumAttrCase<"WRITE_BACK", 4, "write_back">;            // valid for write only
 def XeGPU_CachePolicyWriteThrough:  I32EnumAttrCase<"WRITE_THROUGH", 5, "write_through">;      // valid for write only
 
-def XeGPU_CachePolicyEnums : I32EnumAttr<"CachePolicy", "Cache policy", 
-  [XeGPU_CachePolicyCached, XeGPU_CachePolicyUncached, 
+def XeGPU_CachePolicyEnums : I32EnumAttr<"CachePolicy", "Cache policy",
+  [XeGPU_CachePolicyCached, XeGPU_CachePolicyUncached,
    XeGPU_CachePolicyStreaming, XeGPU_CachePolicyInvalid,
    XeGPU_CachePolicyWriteBack, XeGPU_CachePolicyWriteThrough]> {
   let genSpecializedAttr = 0;
   let cppNamespace = "::mlir::xegpu";
 }
 
-def XeGPU_CacheHintAttr 
+def XeGPU_CacheHintAttr
   : EnumAttr<XeGPU_Dialect, XeGPU_CachePolicyEnums, "cache_hint"> {
+    let summary = [{Describe the cache settings for prefetch/load/store operators}];
     let assemblyFormat = "`<` $value `>`";
 }
 
-
-
-#endif // MLIR_DIALECT_XEGPU_IR_XEGPUATTRS_TD
+#endif // MLIR_DIALECT_XEGPU_IR_XEGPUATTRS_TD
\ No newline at end of file
diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
index b8ebd1a40c607..c6f7f83441b96 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUOps.td
@@ -47,36 +47,35 @@ class XeGPU_Op<string mnemonic, list<Trait> traits = []>:
 }
 
 
-def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface, 
+def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface,
                         AttrSizedOperandSegments, OffsetSizeAndStrideOpInterface]> {
 
   let summary = "Create nd-tensor descriptor operation";
   let description = [{
     The "create_nd_tdesc" operation creates a TensorDescType which represents
     a sub-view of a 2D memory region (It can be extended to support n-D memory
-    region if needed in future). Elements in the subview continuous in each 
-    dimention. It encodes the following important information for supporting 
+    region if needed in future). Elements in the subview continuous in each
+    dimension. It encodes the following important information for supporting
     Intel hardware features:
 
-    * source: an object representing (starting address/pointer of) a 2D memory region. 
+    * source: an object representing (starting address/pointer of) a 2D memory region.
         It can be either a 2D memref object, or simply a pointer represented by uint64_t type.
-        for the later case, the shape and layout information of the 2D memory region should 
-        be explicitly passed via `dynamic_shape` and `dynamic_strides` parameters.
-    * offsets: two index values represents offsets from the "source" at the each dimension 
+        for the later case, the shape and layout information of the 2D memory region should
+        be explicitly passed via `shape` and `strides` parameters.
+    * offsets: two index values represents offsets from the "source" at the each dimension
         at which the subview of the target memory will be created. It is encoded via two
-        variables, including "dynamic_offsets" and "static_offsets", such that it can
-        accept various forms, such as, operands (e.g., [%c0, %c]) and attributes (e.g., [2, 4])).
-    * shape: the shape information of the memory region pointed by the "source".  It is 
-        typically encoded via the MemRefType of the source, e.g., memref<4096x4096xf16>. 
-        But if "source" is simply a pointer represented as uint64_t type, or a memref 
-        type without shape information e.g., memref<?x?xf16>, the shape information has 
-        to be explicitly passed via the "dynamic_shape" argument. Currently "dynamic_shape" 
-        only accepts operands(e.g., [%c4096, %c4096]), not attributes(e.g., [4096, 4096]).
-    * strides: the strides of the memory region pointed by the "source". Similar to shape, 
-        it is typically encoded via the MemRefType of the source too. But if "source" is 
-        simply a pointer represented as uint64_t type, or a memref type without shape 
-        information e.g., memref<?x?xf16>, the strides information has to be explicitly 
-        passed via the "dynamic_strides" argument. And it currently only accepts operands two.
+        variables, including "offsets" and "const_offsets", such that it can
+        accept various forms, such as, operands (e.g., [%c0, %c]) and attributes (e.g., [2, 4]).
+    * shape: the shape information of the memory region pointed by the "source".  It is
+        typically encoded via the MemRefType of the source, e.g., memref<4096x4096xf16>.
+        But if "source" is simply a pointer represented as uint64_t type, or a memref
+        type without shape information e.g., memref<?x?xf16>, the shape information has
+        to be explicitly passed via the "shape" and "const_shape" arguments.
+    * strides: the strides of the memory region pointed by the "source". Similar to shape,
+        it is typically encoded via the MemRefType of the source too. But if "source" is
+        simply a pointer represented as uint64_t type, or a memref type without shape
+        information e.g., memref<?x?xf16>, the strides information has to be explicitly
+        passed via the "strides" and "const_strides" argument.
 
     Example 1 (suppose the tensor shape inferred by the compiler is 8x16):
     %0 = memref.alloc() : memref<1024x1024xf32>
@@ -97,10 +96,10 @@ def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface
     %1 = xegpu.create_nd_tdesc %0[%c0, %c0], [%h, %w], [%w, %c1]: ui64 -> TensorDesc<8x16xf32>
   }];
 
-  let arguments = (ins 
-    XeGPU_BaseAddrType: $source, 
-    Variadic<Index>: $offsets, 
-    Variadic<Index>: $shape, 
+  let arguments = (ins
+    XeGPU_BaseAddrType: $source,
+    Variadic<Index>: $offsets,
+    Variadic<Index>: $shape,
     Variadic<Index>: $strides,
     DenseI64ArrayAttr: $const_offsets,
     OptionalAttr<DenseI64ArrayAttr>: $const_shape,
@@ -119,12 +118,12 @@ def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface
   let hasVerifier = 1;
 
   let builders = [
-    OpBuilder<(ins "Type": $tdesc, "TypedValue<MemRefType>": $source, 
+    OpBuilder<(ins "Type": $tdesc, "TypedValue<MemRefType>": $source,
                    "llvm::ArrayRef<OpFoldResult>": $offsets)>,
 
-    OpBuilder<(ins "Type": $tdesc, "TypedValue<IntegerType> ": $source, 
+    OpBuilder<(ins "Type": $tdesc, "TypedValue<IntegerType> ": $source,
                    "llvm::ArrayRef<OpFoldResult>": $offsets,
-                   "llvm::ArrayRef<OpFoldResult>": $shape, 
+                   "llvm::ArrayRef<OpFoldResult>": $shape,
                    "llvm::ArrayRef<OpFoldResult>": $strides)>
   ];
 
@@ -159,41 +158,41 @@ def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface
     }
 
     /// wrapper for matching with OffsetSizeAndStrideOpInterface
-    /// If source is IntegerType or `const_shape` is filled, 
+    /// If source is IntegerType or `const_shape` is filled,
     /// it will return `const_shape`, such that mixes of `shape`
-    /// and `const_shape` will be used to represent the shape of 
+    /// and `const_shape` will be used to represent the shape of
     /// source operand. They overide static shape from source memref type.
     ArrayRef<int64_t> getStaticSizes() {
       auto attr = getConstShapeAttr();
       if (getSourceType().isa<IntegerType>() || attr)
         return attr;
-      
+
       auto memrefType = getSourceType().dyn_cast<MemRefType>();
       assert(memrefType && "Incorrect use of getStaticSizes");
       return memrefType.getShape();
     }
 
     /// wrapper for matching with OffsetSizeAndStrideOpInterface
-    /// If source is IntegerType or `const_strides` is filled, it 
+    /// If source is IntegerType or `const_strides` is filled, it
     /// will return `const_strides`, such that mixes of `strides`
-    /// and `const_strides` will be used to represent the strides of 
+    /// and `const_strides` will be used to represent the strides of
     /// source operand. They overide static strides from source memref type.
     ArrayRef<int64_t> getStaticStrides() {
       auto attr = getConstStridesAttr();
       if (getSourceType().isa<IntegerType>() || attr)
         return attr;
-      
+
       auto memrefType = getSourceType().dyn_cast<MemRefType>();
       assert(memrefType && "Incorrect use of getStaticStrides");
       auto [strides, offset] = getStridesAndOffset(memrefType);
-      // reuse the storage of ConstStridesAttr since strides from 
+      // reuse the storage of ConstStridesAttr since strides from
       // memref is not persistant
       setConstStrides(strides);
       attr = getConstStridesAttr();
       return attr;
     }
 
-    /// Return the expected rank of each of the`static_offsets`, 
+    /// Return the expected rank of each of the`static_offsets`,
     /// `static_shape` and `static_strides` attributes.
     std::array<unsigned, 3> getArrayAttrMaxRanks() {
       unsigned rank;
@@ -204,8 +203,8 @@ def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface
       }
       return {rank, rank, rank};
     }
-    
-    /// Return the number of leading operands before the `offsets`, 
+
+    /// Return the number of leading operands before the `offsets`,
     /// `shape` and `strides` operands.
     static unsigned getOffsetSizeAndStrideStartOperandIndex() { return 1; }
 
@@ -214,15 +213,15 @@ def XeGPU_CreateNdDescOp: XeGPU_Op<"create_nd_tdesc", [Pure, ViewLikeOpInterface
 }
 
 def XeGPU_PrefetchNdOp : XeGPU_Op<"prefetch_nd", []> {
-  let summary = "prefetches a nD block to cache";
+  let summary = "prefetches a n-D block to cache";
   let description = [{
-    It issues an instruction to prefetch the data from memory to each 
-    level of the cache based on their cache policy.
+    It issues an instruction to prefetch a block of data from continuous
+    memory regions to each level of the cache based on their cache policy.
 
     Example:
     ```
-      xegpu.prefetch_nd %tdesc {l1_hint = #xegpu.cache_hint<cached>, 
-                                l2_hint = #xegpu.cache_hint<cached>, 
+      xegpu.prefetch_nd %tdesc {l1_hint = #xegpu.cache_hint<cached>,
+                                l2_hint = #xegpu.cache_hint<cached>,
                                 l3_hint = #xegpu.cache_hint<cached>}
         : !xegpu.tensor_desc<8x16xf16>
     ```
@@ -233,34 +232,41 @@ def XeGPU_PrefetchNdOp : XeGPU_Op<"prefetch_nd", []> {
                        OptionalAttr<XeGPU_CacheHintAttr>: $l1_hint,
                        OptionalAttr<XeGPU_CacheHintAttr>: $l2_hint,
                        OptionalAttr<XeGPU_CacheHintAttr>: $l3_hint);
-                       
-  let extraClassDeclaration = extraBaseClassDeclaration;
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+  }];
 
   let assemblyFormat = "$TensorDesc prop-dict attr-dict `:` qualified(type($TensorDesc))";
+
+  let hasVerifier = 1;
 }
 
 
-def XeGPU_LoadNdOp : XeGPU_Op<"load_nd"> {
-  let summary = "loads a n-D block from memory (represented by TensorDesc)" 
+def XeGPU_LoadNdOp : XeGPU_Op<"load_nd", [AllElementTypesMatch<["value", "TensorDesc"]>,
+                                         AllElementCountsMatch<["value", "TensorDesc"]>]> {
+  let summary = "loads a n-D block from memory (represented by TensorDesc)"
                 "to registers (represented by vector)";
   let description = [{
-    LoadNdOp essentially mimics the hardware block read instruction to read 
-    a block of data from memory to register. It takes a set of optional cache 
-    hints for each level of cache, L1, L2 and L3. If hardware does not have a 
+    LoadNdOp essentially mimics the hardware block read instruction to read
+    a block of data from memory to register. It takes a set of optional cache
+    hints for each level of cache, L1, L2 and L3. If hardware does not have a
     correspoding cache, Corresponding cache hint attribute will be masked.
-    vnni transform is an hardware feature for Intel GPU, which is used to 
-    do data packing during the load for B operand of matrix operation, if 
-    the bit width of the data type is less then 32 bits, e.g., fp16. And 
+    vnni transform is an hardware feature for Intel GPU, which is used to
+    do data packing during the load for B operand of matrix operation, if
+    the bit width of the data type is less then 32 bits, e.g., fp16. And
     transpose is another Intel hardware feature, which will do transpose
-    operation when loading the data if the bit width of the data type is 
-    fp32 or fp64. It implies that vnni and transpose cannot exit at the 
+    operation when loading the data if the bit width of the data type is
+    fp32 or fp64. It implies that vnni and transpose cannot exit at the
     same time.
 
     Example:
     ```
       xegpu.load_nd %1 {transpose = [1, 0],
-                        l1_hint = #xegpu.cache_hint<cached>, 
-                        l2_hint = #xegpu.cache_hint<uncached>, 
+                        l1_hint = #xegpu.cache_hint<cached>,
+                        l2_hint = #xegpu.cache_hint<uncached>,
                         l3_hint = #xegpu.cache_hint<streaming>}
               : !xegpu.tensor_desc<8x16xf32> -> vector<16x8xf32>
     ```
@@ -291,20 +297,21 @@ def XeGPU_LoadNdOp : XeGPU_Op<"load_nd"> {
   let hasVerifier = 1;
 }
 
-def XeGPU_StoreNdOp : XeGPU_Op<"store_nd", []> {
+def XeGPU_StoreNdOp : XeGPU_Op<"store_nd", [AllShapesMatch<["value", "TensorDesc"]>,
+                                       AllElementTypesMatch<["value", "TensorDesc"]>]> {
   let summary = "stores a n-D block register region back to memory, currently only supports 2D";
 
   let description = [{
     StoreNdOp essentially mimics the hardware block write instruction io
-    write a block of data from register into the memory region as described 
-    by the TensorDesc. It takes a set of optional cache hints for each level 
-    of cache, L1, L2 and L3. If hardware does not have a correspoding cache, 
+    write a block of data from register into the memory region as described
+    by the TensorDesc. It takes a set of optional cache hints for each level
+    of cache, L1, L2 and L3. If hardware does not have a correspoding cache,
     Corresponding cache hint attribute will be masked.
 
     Example:
     ```
       xegpu.store_nd %3, %2 {l1_hint = #xegpu.cache_hint<uncached>,
-                             l2_hint = #xegpu.cache_hint<write_back>, 
+                             l2_hint = #xegpu.cache_hint<write_back>,
                              l3_hint = #xegpu.cache_hint<write_through>}
                              : vector<8x16xf16>, !xegpu.tensor_desc<8x16xf16>
     ```
@@ -318,11 +325,342 @@ def XeGPU_StoreNdOp : XeGPU_Op<"store_nd", []> {
                        OptionalAttr<XeGPU_CacheHintAttr>: $l2_hint,
                        OptionalAttr<XeGPU_CacheHintAttr>: $l3_hint);
 
-  let extraClassDeclaration = extraBaseClassDeclaration;
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    VectorType getValueType() {
+      return llvm::dyn_cast<VectorType>(getValue().getType());
+    }
 
-  let assemblyFormat = [{$value `,` $TensorDesc prop-dict attr-dict 
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+  }];
+
+  let assemblyFormat = [{$value `,` $TensorDesc prop-dict attr-dict
                         `:` type($value) `,` qualified(type($TensorDesc))}];
   let hasVerifier = 1;
 }
 
+def XeGPU_UpdateNdOffsetOp : XeGPU_Op<"update_nd_offset",
+                [AllTypesMatch<["TensorDesc", "result"]>]> {
+  let summary = "It updates the offsets for the TensorDesc.";
+  let description = [{The op updates the offset of the given TensorDesc.
+    The offsets are relative offset to the current position in the number
+    of elements. It will result in a same type TensorDesc as the input.
+
+  example:
+  ```
+    %2 = xegpu.update_nd_offset %1, [0, 16]: !xegpu.tensor_desc<8x16xf32>
+  ```
+  }];
+
+  let arguments = (ins
+    XeGPU_TensorDesc: $TensorDesc,
+    Variadic<Index>: $offsets,
+    DenseI64ArrayAttr: $const_offsets);
+
+  let results = (outs XeGPU_TensorDesc: $result);
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+
+    SmallVector<OpFoldResult> getMixedOffsets() {
+      Builder b(getContext());
+      return getMixedValues(getConstOffsets(), getOffsets(), b);
+    }
+
+    size_t getNumOffsets() {
+      return getMixedOffsets().size();
+    }
+
+    OpFoldResult getOffset(unsigned idx) {
+      assert(idx < getNumOffsets() && "Invalid out of bound access.");
+      return getMixedOffsets()[idx];
+    }
+  }];
+
+  let assemblyFormat = [{
+    $TensorDesc `,`
+    custom<DynamicIndexList>($offsets, $const_offsets)
+    attr-dict `:` qualified(type($result))
+  }];
+
+  let hasVerifier = 1;
+}
+
+def XeGPU_CreateDescOp: XeGPU_Op<"create_tdesc", [Pure, ViewLikeOpInterface]> {
+  let summary = "create scattered tensor descriptors (TensorDesc).";
+  let description = [{
+    "create_tdesc" is similar to "create_nd_tdesc" in terms that it creates
+    a Tensor Descriptor (TensorDescType) for a memory region. While "create_nd_tdesc"
+    is for creating continuous subviews, "create_tdesc" is for creating non-continuous
+    (scattered) subviews, allowing each work-item in a subgroup specifying their own offset.
+    It accepts the following parameters:
+
+    * source: a 1D memref or pointer (uint64_t) represents the flattened memory object.
+    * offsets: a array containing offsets of each access point. Its size
+      is fixed to the hardware supportted subgroup size, e.g., 16 on PVC,
+      implying each element in the array corresponds to a work-item (SIMT lane)
+      in the subgroup.
+    * chunk_size: [optional attribute] indicates number of continious
+      elements accessed for each offset, default is 1.
+
+    Example 1. It assumes subgroup size is 4, and accesses a[0], a[16], a[32], a[64]
+    ```
+    %a = memref.alloc() : memref<1024xf32>
+    %1 = xegpu.create_tdesc %a[0, 16, 32, 64]: memref<1024xf32> -> TensorDesc<4xf32>
+    ```
+
+    Example 2. It assumes subgroup size is 4, and each workitem access 8 elements.
+               It will access totally 32 data elements: a[0:7], a[16:23], a[32:39], a[64:71]
+    ```
+    %0 = memref.alloc() : memref<1024xf32>
+    %1 = xegpu.create_tdesc %0[0, 16, 32, 64] {chunk_size = 8}: memref<1024xf32> -> TensorDesc<4x8xf32>
+    ```
+
+    Example 3. It is similar to Example 2, but there is some overlaps among workitems.
+               It accesses: a[0:7], a[4:11], a[8:15], a[12:19]
+    ```
+    %0 = memref.alloc() : memref<1024xf32>
+    %1 = xegpu.create_tdesc %0[0, 4, 8, 12] {chunk_size = 8}: memref<1024xf32> -> TensorDesc<4x8xf32>
+    ```
+
+
+
+
+  }];
+
+  let arguments = (ins XeGPU_BaseAddrType: $source,
+                       Variadic<Index>: $offsets,
+                       DenseI64ArrayAttr: $const_offsets,
+                       DefaultValuedAttr<I64Attr, "1">: $chunk_size);
+  let results = (outs XeGPU_TensorDesc:$TensorDesc);
+
+  let builders = [
+    OpBuilder<(ins "xegpu::TensorDescType": $TensorDesc, "Value": $source,
+                   "llvm::ArrayRef<OpFoldResult>": $offsets,
+                   CArg<"uint32_t", "1"> : $chunk_size)>,
+  ];
+
+  let assemblyFormat = [{
+    $source
+    custom<DynamicIndexList>($offsets, $const_offsets)
+    attr-dict `:`  type($source) `->` qualified(type($TensorDesc))
+  }];
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+
+    SmallVector<OpFoldResult> getMixedOffsets() {
+      Builder b(getContext());
+      return getMixedValues(getConstOffsets(), getOffsets(), b);
+    }
+
+    size_t getNumOffsets() {
+      return getMixedOffsets().size();
+    }
+
+    mlir::Value getViewSource() { return getSource(); }
+
+    OpFoldResult getOffset(unsigned idx) {
+      assert(idx < getNumOffsets() && "Invalid out of bound access.");
+      return getMixedOffsets()[idx];
+    }
+  }];
+
+  let hasVerifier = 1;
+}
+
+def XeGPU_PrefetchOp : XeGPU_Op<"prefetch", []> {
+  let summary = "prefetches a set of scattered data points to cache";
+
+  let description = [{
+    It issues instructions to prefetch a set of scattered data points
+    from memory to each level of the cache based on their cache policy.
+    As compared to prefetch_nd, which works on non-scattered TensorDesc,
+    it works on scattered TensorDesc instead.
+
+    Example:
+    ```
+      xegpu.prefetch %tdesc {l1_hint = #xegpu.cache_hint<cached>,
+                             l2_hint = #xegpu.cache_hint<cached>,
+                             l3_hint = #xegpu.cache_hint<cached>}
+        : !xegpu.tensor_desc<16xf16>
+    ```
+
+  }];
+
+  let arguments = (ins XeGPU_TensorDesc: $TensorDesc,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l1_hint,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l2_hint,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l3_hint);
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+  }];
+
+  let assemblyFormat = "$TensorDesc prop-dict attr-dict `:` qualified(type($TensorDesc))";
+
+  let hasVerifier = 1;
+}
+
+def XeGPU_LoadGatherOp : XeGPU_Op<"load", [AllRanksMatch<["value", "TensorDesc"]>,
+                                    AllElementTypesMatch<["value", "TensorDesc"]>,
+                                   AllElementCountsMatch<["value", "TensorDesc"]>]> {
+  let summary = "load a set of scattered data points from memory.";
+
+  let description = [{ It (aka. load) load data per each work-item. The output
+    describes the data being loaded at the subgroup level, so its size is
+    consistent with the number of work-items in a subgroup. When `chunk_size_per_lane`
+    attribute is larger than 1 in TensorDesc, the output vector will be 2D vector,
+    with dim-1 correspoding to the chunk size.
+
+    The mask operand masks out memory access so that it is safe to pass out-of-boundary
+    addresses/offsets as long as they are masked. It applies to slots of SIMD lanes.
+
+  Example:
+  ```
+    %2 = xegpu.load %1, %0 {transpose = [1, 0],
+                            l1_hint = #xegpu.cache_hint<cached>,
+                            l2_hint = #xegpu.cache_hint<uncached>,
+                            l3_hint = #xegpu.cache_hint<uncached>}
+          : !xegpu.tensor_desc<16xf32, #xegpu.tdesc_attr<scattered=true>>, vector<16xi1>
+            -> vector<16xf32>
+  ```
+
+  }];
+
+  let arguments = (ins XeGPU_TensorDesc: $TensorDesc,
+                       XeGPU_MaskType: $mask,
+                       OptionalAttr<DenseI64ArrayAttr>: $transpose,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l1_hint,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l2_hint,
+                       OptionalAttr<XeGPU_CacheHintAttr>: $l3_hint);
+  let results = (outs XeGPU_ValueType: $value);
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+
+    mlir::Type getElementType() {
+      auto type = getValue().getType();
+      return getElementTypeOrSelf(type);
+    }
+
+    Type getValueType() {
+      return getValue().getType();
+    }
+
+    Type getMaskType() {
+      return getMask().getType();
+    }
+
+  }];
+
+  let assemblyFormat = [{$TensorDesc `,` $mask prop-dict attr-dict
+      `:` qualified(type($TensorDesc)) `,` type($mask) `->` type($value)}];
+
+  let hasVerifier = 1;
+}
+
+def XeGPU_StoreScatterOp : XeGPU_Op<"store", [AllShapesMatch<["value", "TensorDesc"]>,
+                                        AllElementTypesMatch<["value", "TensorDesc"]>]> {
+  let summary = "store data to scattered memory locations.";
+  let description = [{ It (aka. store) stores data to scattered memory locations.
+  It has similar semantic to `load_gather`.
+
+  Example:
+  ```
+    %3 = xegpu.store %0, %1, %2 {l1_hint = #xegpu.cache_hint<uncached>,
+                                 l2_hint = #xegpu.cache_hint<write_back>,
+                                 l3_hint = #xegpu.cache_hint<write_through>}
+          : vector<16xf32>, !xegpu.tensor_desc<16xf32, #xegpu.tdesc_attr<scattered=true>>, vector<16xi1>
+  ```
+  }];
+
+  let arguments = (ins
+    XeGPU_ValueType: $value,
+    XeGPU_TensorDesc: $TensorDesc,
+    XeGPU_MaskType: $mask,
+    OptionalAttr<XeGPU_CacheHintAttr>: $l1_hint,
+    OptionalAttr<XeGPU_CacheHintAttr>: $l2_hint,
+    OptionalAttr<XeGPU_CacheHintAttr>: $l3_hint);
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+
+    Type getValueType() {
+      return getValue().getType();
+    }
+
+    Type getMaskType() {
+      return getMask().getType();
+    }
+  }];
+
+  let assemblyFormat = [{$value `,` $TensorDesc `,` $mask prop-dict attr-dict
+            `:` type($value) `,` qualified(type($TensorDesc)) `,` type($mask)}];
+
+  let hasVerifier = 1;
+}
+
+def XeGPU_UpdateOffsetOp: XeGPU_Op<"update_offset",
+          [AllTypesMatch<["TensorDesc", "result"]>]> {
+  let summary = "It updates the offsets for the given tensor descriptor";
+
+  let description = [{It behaves similar to `update_nd_offset` in terms that
+    it updates offset of a TensorDesc, and the offsets are relative offset to
+    the current position in the number of elements. However, `update_nd_offset`
+    is to update the start point of a 2D block, so its offset constains two
+    elements representing the shift in each dimension. `update_offset` is to
+    update the offset per work-item, so its offsets contains values representing
+    shifts for each work-item.
+
+    Example:
+    ```
+      %2 = xegpu.update_offset %1, [32, 32, 32, 32]
+            : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+    ```
+  }];
+
+  let arguments = (ins XeGPU_TensorDesc: $TensorDesc,
+                       Variadic<Index>: $offsets,
+                       DenseI64ArrayAttr: $const_offsets);
+  let results = (outs XeGPU_TensorDesc: $result);
+
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    xegpu::TensorDescType getTensorDescType() {
+      return getTensorDesc().getType();
+    }
+
+    SmallVector<OpFoldResult> getMixedOffsets() {
+      Builder b(getContext());
+      return getMixedValues(getConstOffsets(), getOffsets(), b);
+    }
+
+    size_t getNumOffsets() {
+      return getMixedOffsets().size();
+    }
+
+    OpFoldResult getOffset(unsigned idx) {
+      assert(idx < getNumOffsets() && "Invalid out of bound access.");
+      return getMixedOffsets()[idx];
+    }
+  }];
+
+  let assemblyFormat = [{
+    $TensorDesc `,`
+    custom<DynamicIndexList>($offsets, $const_offsets)
+    attr-dict `:` qualified(type($TensorDesc))
+  }];
+}
+
 #endif // MLIR_DIALECT_XEGPU_IR_XEGPUOPS_TD
diff --git a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUTypes.td b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUTypes.td
index 19ac1693712dd..4cd4e5411653c 100644
--- a/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUTypes.td
+++ b/mlir/include/mlir/Dialect/XeGPU/IR/XeGPUTypes.td
@@ -34,10 +34,10 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
         [ShapedTypeInterface], "::mlir::TensorType"> {
   let summary = "TensorDesc describing regions of interested data.";
   let description = [{
-    TensorDesc is a type designed to describe regions of the interested data as well as some 
-    features that are unique to Intel hardware. Different with the builtin tensor type in MLIR, 
-    it essentially only contains the meta data, and doesn't hold the data by itself. It is designed 
-    to mainly support 2D block load/store and DPAS (matrix multiplication instruction) on Intel GPU. 
+    TensorDesc is a type designed to describe regions of the interested data as well as some
+    features that are unique to Intel hardware. Different with the builtin tensor type in MLIR,
+    it essentially only contains the meta data, and doesn't hold the data by itself. It is designed
+    to mainly support 2D block load/store and DPAS (matrix multiplication instruction) on Intel GPU.
     It encodes the following information:
 
     * shape:  the sizes/shape of the intereted data block, e.g., 8x16 means 8 rows
@@ -46,15 +46,15 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
               is set or not.
     * element_type: the data type of the data element, e.g., f16, f32.
 
-    Similar to the builtin tensor, it also provides an optinal attribute to encoding 
+    Similar to the builtin tensor, it also provides an optinal attribute to encoding
     the following information via the TensorDescAttr object:
-    * memory_scope (xegpu::MemoryScope): [optional] where the data is located, 
+    * memory_scope (xegpu::MemoryScope): [optional] where the data is located,
                 global memory or shared memory. It is default to Global.
     * array_length (int): [optional] The number of contiguous blocks with size as `shape`,
                that will be loaded by block load at a time. It is default to 1.
-    * boundary_check (bool): [optional] indicates whether the operation detects the boundary 
+    * boundary_check (bool): [optional] indicates whether the operation detects the boundary
                 and pads with zero for out-of-boundary access. It is default to do boundary check.
-    
+
 
     Syntax:
 
@@ -63,7 +63,7 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
     element-type ::= float-type | integer-type | index-type
     dim-list := (static-dim-list `x`)?
     static-dim-list ::= decimal-literal `x` decimal-literal
-    attr-list = (, memory_scope = value)? (, arr_len = value)? (, boundary_check = value)?
+    attr-list = (, memory_scope = value)? (, arr_len = value)? (, boundary_check = value)? (, scattered = value)?
     ```
 
     Examples:
@@ -84,6 +84,17 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
                         "mlir::Type": $elementType,
                         OptionalParameter<"mlir::Attribute">: $encoding);
 
+  let builders = [
+    TypeBuilderWithInferredContext<(ins
+      "llvm::ArrayRef<int64_t>": $shape,
+      "mlir::Type": $elementType,
+      CArg<"bool", "false">: $scattered,
+      CArg<"int", "1">: $array_length,
+      CArg<"xegpu::MemoryScope", "xegpu::MemoryScope::Global">:$memory_scope,
+      CArg<"bool", "true">: $boundary_check
+    )>
+  ];
+
   let extraClassDeclaration = [{
     using TensorType::clone;
     using mlir::ShapedType::Trait<TensorDescType>::getElementTypeBitWidth;
@@ -116,7 +127,7 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
       if (attr && attr.getArrayLength())
         return attr.getArrayLength().getInt();
       // return default value
-      return 1; 
+      return 1;
     }
 
     bool getBoundaryCheck() {
@@ -126,10 +137,18 @@ def XeGPU_TensorDesc: XeGPUTypeDef<"TensorDesc", "tensor_desc",
       // return default value
       return true;
     }
+
+    bool getScattered() {
+      auto attr = getEncodingAsTensorDescAttr();
+      if (attr && attr.getScattered())
+        return attr.getScattered().getValue();
+      // return default value
+      return false;
+    }
   }];
 
   let hasCustomAssemblyFormat = true;
-  
+
 }
 
 #endif // MLIR_DIALECT_XEGPU_IR_XEGPUTYPES_TD
diff --git a/mlir/lib/Dialect/XeGPU/IR/XeGPUDialect.cpp b/mlir/lib/Dialect/XeGPU/IR/XeGPUDialect.cpp
index 0b3f4b9c9dbea..24719fe748fe4 100644
--- a/mlir/lib/Dialect/XeGPU/IR/XeGPUDialect.cpp
+++ b/mlir/lib/Dialect/XeGPU/IR/XeGPUDialect.cpp
@@ -32,6 +32,17 @@ void XeGPUDialect::initialize() {
 //===----------------------------------------------------------------------===//
 // XeGPU_TensorDescAttr
 //===----------------------------------------------------------------------===//
+TensorDescAttr TensorDescAttr::get(mlir::MLIRContext *context,
+                                   xegpu::MemoryScope memory_scope,
+                                   int array_length, bool boundary_check,
+                                   bool scattered) {
+  auto scopeAttr = MemoryScopeAttr::get(context, memory_scope);
+  auto lengthAttr =
+      IntegerAttr::get(IntegerType::get(context, 64), array_length);
+  auto boundaryAttr = BoolAttr::get(context, boundary_check);
+  auto scatteredAttr = BoolAttr::get(context, scattered);
+  return Base::get(context, scopeAttr, lengthAttr, boundaryAttr, scatteredAttr);
+}
 
 //===----------------------------------------------------------------------===//
 // XeGPU_TensorDescType
@@ -96,6 +107,16 @@ void TensorDescType::print(::mlir::AsmPrinter &printer) const {
   printer << ">";
 }
 
+TensorDescType TensorDescType::get(llvm::ArrayRef<int64_t> shape,
+                                   mlir::Type elementType, bool scattered,
+                                   int array_length, MemoryScope memory_scope,
+                                   bool boundary_check) {
+  auto context = elementType.getContext();
+  auto attr = TensorDescAttr::get(context, memory_scope, array_length,
+                                  boundary_check, scattered);
+  return Base::get(context, shape, elementType, attr);
+}
+
 } // namespace xegpu
 } // namespace mlir
 
diff --git a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
index 02106f221f323..621986c54d492 100644
--- a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
+++ b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
@@ -9,6 +9,9 @@
 #include "mlir/Dialect/Utils/StaticValueUtils.h"
 #include "mlir/Dialect/XeGPU/IR/XeGPU.h"
 #include "mlir/IR/Builders.h"
+#include "mlir/IR/TypeUtilities.h"
+
+#include "llvm/Support/Debug.h"
 
 #define DEBUG_TYPE "xegpu"
 
@@ -16,8 +19,8 @@ namespace mlir {
 namespace xegpu {
 
 static void transpose(llvm::ArrayRef<int64_t> trans,
-                      std::vector<int64_t> &shape) {
-  std::vector<int64_t> old = shape;
+                      SmallVector<int64_t> &shape) {
+  SmallVector<int64_t> old = shape;
   for (size_t i = 0; i < trans.size(); i++)
     shape[i] = old[trans[i]];
 }
@@ -38,6 +41,38 @@ static std::string makeString(T array, bool breakline = false) {
   return buf;
 }
 
+static SmallVector<int64_t> getShapeOf(Type type) {
+  SmallVector<int64_t> shape;
+  if (auto ty = llvm::dyn_cast<ShapedType>(type))
+    shape = SmallVector<int64_t>(ty.getShape());
+  else
+    shape.push_back(1);
+  return shape;
+}
+
+static int64_t getRankOf(Value val) {
+  auto type = val.getType();
+  if (auto ty = llvm::dyn_cast<ShapedType>(type))
+    return ty.getRank();
+  return 0;
+};
+
+static bool isReadHintOrNone(const CachePolicyAttr &attr) {
+  if (!attr)
+    return true;
+  auto kind = attr.getValue();
+  return kind == CachePolicy::CACHED || kind == CachePolicy::UNCACHED ||
+         kind == CachePolicy::STREAMING || kind == CachePolicy::READ_INVALIDATE;
+}
+
+static bool isWriteHintOrNone(const CachePolicyAttr &attr) {
+  if (!attr)
+    return true;
+  auto kind = attr.getValue();
+  return kind == CachePolicy::CACHED || kind == CachePolicy::UNCACHED ||
+         kind == CachePolicy::WRITE_BACK || kind == CachePolicy::WRITE_THROUGH;
+}
+
 //===----------------------------------------------------------------------===//
 // XeGPU_CreateNdDescOp
 //===----------------------------------------------------------------------===//
@@ -114,6 +149,29 @@ LogicalResult CreateNdDescOp::verify() {
     return emitOpError("TensorDesc should have the same element "
                        "type with the source if it is a memref.\n");
 
+  if (getType().getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
+
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_PrefetchNdOp
+//===----------------------------------------------------------------------===//
+LogicalResult PrefetchNdOp::verify() {
+  auto tdescTy = getTensorDescType();
+  if (tdescTy.getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
+
+  if (!isReadHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isReadHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isReadHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
   return success();
 }
 
@@ -125,22 +183,26 @@ LogicalResult LoadNdOp::verify() {
   auto valueTy = getType();
 
   if (tdescTy.getRank() != 2)
-    return emitOpError(
-        "The TensorDesc for LoadNdOp should be a 2D TensorDesc.");
+    return emitOpError("Expecting a 2D TensorDesc.\n");
+
+  if (tdescTy.getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
 
   if (!valueTy)
     return emitOpError("Invalid result, it should be a VectorType.\n");
 
-  auto tdescElemTy = tdescTy.getElementType();
-  auto valueElemTy = valueTy.getElementType();
+  if (!isReadHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
 
-  if (tdescElemTy != valueElemTy)
-    return emitOpError(
-        "Value should have the same element type as TensorDesc.");
+  if (!isReadHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isReadHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
 
   auto array_len = tdescTy.getArrayLength();
-  auto tdescShape = tdescTy.getShape().vec();
-  auto valueShape = valueTy.getShape().vec();
+  auto tdescShape = getShapeOf(tdescTy);
+  auto valueShape = getShapeOf(valueTy);
 
   if (getTranspose()) {
     auto trans = getTranspose().value();
@@ -174,26 +236,174 @@ LogicalResult LoadNdOp::verify() {
 // XeGPU_StoreNdOp
 //===----------------------------------------------------------------------===//
 LogicalResult StoreNdOp::verify() {
-  auto dstTy = getTensorDesc().getType();               // Tile
-  auto valTy = getValue().getType().cast<VectorType>(); // Vector
+  auto dstTy = getTensorDescType(); // Tile
+  auto valTy = getValueType();      // Vector
 
   if (dstTy.getRank() != 2)
-    return emitOpError("Expecting a 2D TensorDesc shape.\n");
+    return emitOpError("Expecting a 2D TensorDesc.\n");
+
+  if (dstTy.getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
 
   if (!valTy)
     return emitOpError("Exepcting a VectorType result.\n");
 
-  auto dstElemTy = dstTy.getElementType();
-  auto valElemTy = valTy.getElementType();
+  if (!isWriteHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isWriteHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isWriteHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
+  return success();
+}
 
-  if (dstElemTy != valElemTy) {
-    return emitOpError() << "The element type of the value should "
-                            "match the elementtype of the TensorDesc.\n";
+//===----------------------------------------------------------------------===//
+// XeGPU_UpdateNDOffsetOp
+//===----------------------------------------------------------------------===//
+LogicalResult UpdateNdOffsetOp::verify() {
+  auto ty = getTensorDescType();
+  if (ty.getScattered())
+    return emitOpError("Expects a non-scattered TensorDesc.\n");
+
+  // number of offsets specified must match the rank of the tensor descriptor
+  if (ty.getRank() != (int64_t)getNumOffsets()) {
+    return emitOpError("Invalid number of offsets.");
   }
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_CreateDescOp
+//===----------------------------------------------------------------------===//
+void CreateDescOp::build(OpBuilder &builder, OperationState &state,
+                         TensorDescType TensorDesc, Value source,
+                         llvm::ArrayRef<OpFoldResult> offsets,
+                         uint32_t chunk_size) {
+  llvm::SmallVector<int64_t> staticOffsets;
+  llvm::SmallVector<Value> dynamicOffsets;
+  dispatchIndexOpFoldResults(offsets, dynamicOffsets, staticOffsets);
+  build(builder, state, TensorDesc, source, dynamicOffsets, staticOffsets,
+        chunk_size);
+}
+
+LogicalResult CreateDescOp::verify() {
+  auto tdescTy = getTensorDescType();
+  auto chunkSize = getChunkSize();
+
+  if (getRankOf(getSource()) > 1)
+    return emitOpError(
+        "Expecting the source is a 1D memref or pointer (uint64_t).");
+
+  if (!tdescTy.getScattered())
+    return emitOpError("Expects a scattered TensorDesc.\n");
+
+  SmallVector<int64_t> shape({(int64_t)getNumOffsets()});
+  if (chunkSize != 1)
+    shape.push_back(chunkSize);
+
+  auto tdescShape = getShapeOf(tdescTy);
+  if (shape != tdescShape)
+    return emitOpError("Incorrect TensorDesc shape. ")
+           << "Expected is " << makeString(shape) << "\n";
+
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_PrefetchOp
+//===----------------------------------------------------------------------===//
+LogicalResult PrefetchOp::verify() {
+  auto tdescTy = getTensorDescType();
+  if (!tdescTy.getScattered())
+    return emitOpError("Expects a scattered TensorDesc.\n");
+
+  if (!isReadHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isReadHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isReadHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_LoadGatherOp
+//===----------------------------------------------------------------------===//
+LogicalResult LoadGatherOp::verify() {
+  auto tdescTy = getTensorDescType();
+  auto maskTy = getMaskType();
+  auto valueTy = getValueType();
+
+  if (!tdescTy.getScattered())
+    return emitOpError("Expects a scattered TensorDesc.\n");
+
+  if (!isReadHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isReadHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isReadHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
+  auto tdescElemTy = tdescTy.getElementType();
+  auto valueElemTy = getElementType();
+  if (tdescElemTy != valueElemTy)
+    return emitOpError(
+        "Value should have the same element type as TensorDesc.");
+
+  auto maskShape = getShapeOf(maskTy);
+  auto valueShape = getShapeOf(valueTy);
+  auto tdescShape = getShapeOf(tdescTy);
+
+  if (tdescShape[0] != maskShape[0])
+    return emitOpError("dim-0 of the Mask and TensorDesc should be the same.");
+
+  if (getTransposeAttr()) {
+    auto trans = getTranspose().value();
+    if (tdescShape.size() < trans.size())
+      emitWarning("Invalid transpose attr. It is ignored.");
+    else
+      transpose(trans, tdescShape);
+  }
+
+  if (valueShape != tdescShape)
+    return emitOpError("Unexpected result shape")
+           << "(Expected shape: " << makeString(tdescShape)
+           << ", Given shape: " << makeString(valueShape) << ").\n";
+
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// XeGPU_StoreScatterOp
+//===----------------------------------------------------------------------===//
+LogicalResult StoreScatterOp::verify() {
+  auto tdescTy = getTensorDescType();
+  if (!tdescTy.getScattered())
+    return emitOpError("Expects a scattered TensorDesc.\n");
+
+  if (!isWriteHintOrNone(getL1HintAttr()))
+    return emitOpError("invlid l1_hint: ") << getL1HintAttr();
+
+  if (!isWriteHintOrNone(getL2HintAttr()))
+    return emitOpError("invlid l2_hint: ") << getL2HintAttr();
+
+  if (!isWriteHintOrNone(getL3HintAttr()))
+    return emitOpError("invlid l3_hint: ") << getL3HintAttr();
+
+  auto maskTy = getMaskType();
+  auto maskShape = getShapeOf(maskTy);
+  auto tdescShape = getShapeOf(tdescTy);
+  if (tdescShape[0] != maskShape[0])
+    return emitOpError("dim-0 of the Mask and TensorDesc should be the same.");
 
-  if (dstTy.getShape() != valTy.getShape())
-    return emitOpError()
-           << "The result shape should match the TensorDesc shape.\n";
   return success();
 }
 
diff --git a/mlir/test/Dialect/XeGPU/XeGPUOps.mlir b/mlir/test/Dialect/XeGPU/XeGPUOps.mlir
index 039346adbb851..f0945c79a94ac 100644
--- a/mlir/test/Dialect/XeGPU/XeGPUOps.mlir
+++ b/mlir/test/Dialect/XeGPU/XeGPUOps.mlir
@@ -59,4 +59,66 @@ gpu.func @test_store_nd_vc(%dst: memref<24x32xf16>) {
   gpu.return
 }
 
+// CHECK: gpu.func @test_create_update_nd_tdesc_vc(%[[arg0:.*]]: memref<24x32xf32>) {
+gpu.func @test_create_update_nd_tdesc_vc(%src: memref<24x32xf32>) {
+  // CHECK: %[[REG:.*]] = xegpu.create_nd_tdesc %arg0[0, 0] : memref<24x32xf32> -> !xegpu.tensor_desc<8x16xf32>
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf32> -> !xegpu.tensor_desc<8x16xf32>
+  // CHECK: %[[R1:.*]] = xegpu.update_nd_offset %[[REG]], [0, 16] : !xegpu.tensor_desc<8x16xf32>
+  %2 = xegpu.update_nd_offset %1, [0, 16]: !xegpu.tensor_desc<8x16xf32>
+  gpu.return
+}
+
+// CHECK: gpu.func @test_create_tdesc_vc(%[[arg0:.*]]: ui64) {
+gpu.func @test_create_tdesc_vc(%src: ui64) {
+  //CHECK: %[[R0:.*]] = xegpu.create_tdesc %arg0 [0, 8, 16, 24] {chunk_size = 2 : i64} : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  gpu.return
+}
+
+// CHECK: gpu.func @test_prefetch_vc(%[[arg0:.*]]: ui64) {
+gpu.func @test_prefetch_vc(%src: ui64) {
+  //CHECK: %[[R0:.*]] = xegpu.create_tdesc %arg0 [0, 8, 16, 24] {chunk_size = 2 : i64} : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  // CHECK: xegpu.prefetch %[[R0]] <{l1_hint = #xegpu.cache_hint<cached>, l2_hint = #xegpu.cache_hint<uncached>}> : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  xegpu.prefetch %1 <{l1_hint = #xegpu.cache_hint<cached>, l2_hint = #xegpu.cache_hint<uncached>}>: !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>> 
+  gpu.return
+}
+
+// CHECK: gpu.func @test_load_gather_vc(%[[arg0:.*]]: ui64) {
+gpu.func @test_load_gather_vc(%src: ui64) {
+  //CHECK: %[[cst:.*]] = arith.constant dense<true> : vector<4xi1>
+  %0 = arith.constant dense<1>: vector<4xi1>
+  //CHECK: %[[R0:.*]] = xegpu.create_tdesc %arg0 [0, 8, 16, 24] {chunk_size = 2 : i64} : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  //CHECK: %[[R1:.*]] = xegpu.load %[[R0]], %[[cst]] <{l1_hint = #xegpu.cache_hint<cached>, l2_hint = #xegpu.cache_hint<uncached>}>
+  //CHECK-SAME: !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1> -> vector<4x2xf32>
+  %2 = xegpu.load %1, %0 <{l1_hint = #xegpu.cache_hint<cached>, l2_hint = #xegpu.cache_hint<uncached>}>
+        : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1> -> vector<4x2xf32>
+  gpu.return
+}
+
+// CHECK: gpu.func @test_store_scatter_vc(%[[arg0:.*]]: ui64) {
+gpu.func @test_store_scatter_vc(%src: ui64) {
+  //CHECK: %[[c0:.*]] = arith.constant dense<true> : vector<4xi1>
+  %0 = arith.constant dense<1>: vector<4xi1>
+  //CHECK: %[[c1:.*]] = arith.constant dense<2.900000e+00> : vector<4x2xf32>
+  %1 = arith.constant dense<2.9>: vector<4x2xf32>
+  //CHECK: %[[R0:.*]] = xegpu.create_tdesc %arg0 [0, 8, 16, 24] {chunk_size = 2 : i64} : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %2 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  //CHECK: xegpu.store %[[c1]], %[[R0]], %[[c0]] <{l1_hint = #xegpu.cache_hint<write_back>, l2_hint = #xegpu.cache_hint<uncached>}>
+  //CHECK-SAME: vector<4x2xf32>, !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1>
+  xegpu.store %1, %2, %0 <{l1_hint = #xegpu.cache_hint<write_back>, l2_hint = #xegpu.cache_hint<uncached>}>
+        : vector<4x2xf32>, !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1>
+  gpu.return
+}
+
+// CHECK: gpu.func @test_create_update_tdesc_vc(%[[arg0:.*]]: ui64) {
+gpu.func @test_create_update_tdesc_vc(%src: ui64) {
+  //CHECK: %[[R0:.*]] = xegpu.create_tdesc %arg0 [0, 8, 16, 24] {chunk_size = 2 : i64} : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  //CHECK: %[[R1:.*]] = xegpu.update_offset %[[R0]], [32, 32, 32, 32] : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  %2 = xegpu.update_offset %1, [32, 32, 32, 32] : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  gpu.return
+}
+
 }
\ No newline at end of file
diff --git a/mlir/test/Dialect/XeGPU/invalid.mlir b/mlir/test/Dialect/XeGPU/invalid.mlir
new file mode 100644
index 0000000000000..5e29361ec6908
--- /dev/null
+++ b/mlir/test/Dialect/XeGPU/invalid.mlir
@@ -0,0 +1,159 @@
+// RUN: mlir-opt %s -split-input-file -verify-diagnostics
+
+// -----
+func.func @test_create_nd_tdesc_vc_1(%src: memref<24xf32>) {
+  // expected-error@+1 {{Expecting the rank of shape, strides, offsets, source memref type (if source is a memref) and TensorDesc should match with each other. They currenlty are 2D.}}
+  %1 = xegpu.create_nd_tdesc %src[0] : memref<24xf32> -> !xegpu.tensor_desc<8x16xf32>
+  return
+}
+
+// -----
+
+func.func @test_create_nd_tdesc_vc_2(%src: memref<24x32xf32>) {
+  // expected-error@+1 {{TensorDesc should have the same element type with the source if it is a memref}}
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf32> -> !xegpu.tensor_desc<8x16xf16>
+  return
+}
+
+// -----
+func.func @test_prefetch_nd_vc_1(%src: memref<24x32xf16>) {
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf16> -> !xegpu.tensor_desc<8x16xf16>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<write_back>}}
+  xegpu.prefetch_nd %1 <{l1_hint = #xegpu.cache_hint<write_back>}>: !xegpu.tensor_desc<8x16xf16>
+  return
+}
+
+// -----
+func.func @test_prefetch_nd_vc_2(%src: memref<24xf16>) {
+  %1 = xegpu.create_tdesc %src[0, 1, 2, 3, 4, 5, 6, 7]
+        : memref<24xf16> -> !xegpu.tensor_desc<8xf16, #xegpu.tdesc_attr<scattered=true>>
+  // expected-error@+1 {{Expects a non-scattered TensorDesc}}
+  xegpu.prefetch_nd %1 <{l1_hint = #xegpu.cache_hint<cached>}>
+        : !xegpu.tensor_desc<8xf16, #xegpu.tdesc_attr<scattered=true>>
+  return
+}
+
+// -----
+func.func @test_load_nd_vc_1(%src: memref<8x16xf16>) {
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<8x16xf16> -> !xegpu.tensor_desc<8x16xf16>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<write_back>}}
+  %2 = xegpu.load_nd %1 <{l1_hint = #xegpu.cache_hint<write_back>}>
+      : !xegpu.tensor_desc<8x16xf16> -> vector<4x16x2xf16>
+  return
+}
+
+// -----
+func.func @test_load_nd_vc_2(%src: memref<16xf16>) {
+  %1 = xegpu.create_tdesc %src[0, 2, 4, 6, 8, 10, 12, 14] {chunk_size = 2}
+        : memref<16xf16> -> !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>>
+  // expected-error@+1 {{Expects a non-scattered TensorDesc.}}
+  %2 = xegpu.load_nd %1 <{l1_hint = #xegpu.cache_hint<cached>}>
+      : !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>> -> vector<8x2xf16>
+  return
+}
+
+// -----
+func.func @test_store_nd_vc_1(%dst: memref<24x32xf16>) {
+  %1 = arith.constant dense<1.0>: vector<24x32xf16>
+  %2 = xegpu.create_nd_tdesc %dst[0, 0] : memref<24x32xf16> -> !xegpu.tensor_desc<24x32xf16>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<streaming>}}
+  xegpu.store_nd %1, %2 <{l1_hint = #xegpu.cache_hint<streaming>}>: vector<24x32xf16>, !xegpu.tensor_desc<24x32xf16>
+  return
+}
+
+// -----
+func.func @test_store_nd_vc_2(%dst: memref<16xf16>) {
+  %1 = arith.constant dense<1.0>: vector<8x2xf16>
+  %2 = xegpu.create_tdesc %dst[0, 2, 4, 6, 8, 10, 12, 14] {chunk_size = 2}
+        : memref<16xf16> -> !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>>
+  // expected-error@+1 {{Expects a non-scattered TensorDesc}}
+  xegpu.store_nd %1, %2 <{l1_hint = #xegpu.cache_hint<streaming>}>
+        : vector<8x2xf16>, !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>>
+  return
+}
+
+// -----
+func.func @test_update_nd_offset_1(%dst: memref<16xf16>) {
+  %1 = xegpu.create_tdesc %dst[0, 2, 4, 6, 8, 10, 12, 14] {chunk_size = 2}
+        : memref<16xf16> -> !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>>
+  // expected-error@+1 {{Expects a non-scattered TensorDesc}}
+  xegpu.update_nd_offset %1, [0, 2] : !xegpu.tensor_desc<8x2xf16, #xegpu.tdesc_attr<scattered=true>>
+  return
+}
+
+// -----
+func.func @test_create_tdesc_vc_1(%src: ui64) {
+  // expected-error@+1 {{Expects a scattered TensorDesc}}
+  %1 = xegpu.create_tdesc %src[0, 2, 4, 6, 8, 10, 12, 14] {chunk_size = 2}
+        : ui64 -> !xegpu.tensor_desc<8x2xf16>
+  return
+}
+
+// -----
+func.func @test_create_tdesc_vc_2(%src: ui64) {
+  // expected-error@+1 {{Incorrect TensorDesc shape}}
+  %1 = xegpu.create_tdesc %src[0, 2, 4, 6, 8, 10, 12, 14] {chunk_size = 2}
+        : ui64 -> !xegpu.tensor_desc<8x4xf16, #xegpu.tdesc_attr<scattered = true>>
+  return
+}
+
+// -----
+func.func @test_prefetch_vc_1(%src: memref<24x32xf16>) {
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf16> -> !xegpu.tensor_desc<24x32xf16>
+  // expected-error@+1 {{Expects a scattered TensorDesc}}
+  xegpu.prefetch %1 <{l1_hint = #xegpu.cache_hint<write_back>}>: !xegpu.tensor_desc<24x32xf16>
+  return
+}
+
+// -----
+func.func @test_prefetch_vc_2(%src: ui64) {
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64  -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<write_back>}}
+  xegpu.prefetch %1 <{l1_hint = #xegpu.cache_hint<write_back>}>: !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  return
+}
+
+// -----
+func.func @test_load_gather_vc_1(%src: memref<24x32xf16>) {
+  %0 = arith.constant dense<1>: vector<4xi1>
+  %1 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf16> -> !xegpu.tensor_desc<4x2xf16>
+  // expected-error@+1 {{Expects a scattered TensorDesc}}
+  %2 = xegpu.load %1, %0 <{l1_hint = #xegpu.cache_hint<cached>}>
+      : !xegpu.tensor_desc<4x2xf16>, vector<4xi1> -> vector<4x2xf16>
+  return
+}
+
+// -----
+func.func @test_load_gather_vc_2(%src: ui64) {
+  %0 = arith.constant dense<1>: vector<4xi1>
+  %1 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2} : ui64
+        -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<write_back>}}
+  %2 = xegpu.load %1, %0 <{l1_hint = #xegpu.cache_hint<write_back>}>
+        : !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1>
+          -> vector<4x2xf32>
+  return
+}
+
+// -----
+func.func @test_store_scatter_vc_1(%src: memref<24x32xf32>) {
+  %0 = arith.constant dense<1>: vector<4xi1>
+  %1 = arith.constant dense<2.9>: vector<4x2xf32>
+  %2 = xegpu.create_nd_tdesc %src[0, 0] : memref<24x32xf32> -> !xegpu.tensor_desc<4x2xf32>
+  // expected-error@+1 {{Expects a scattered TensorDesc}}
+  xegpu.store %1, %2, %0 <{l1_hint = #xegpu.cache_hint<cached>}>
+        : vector<4x2xf32>, !xegpu.tensor_desc<4x2xf32>, vector<4xi1>
+  return
+}
+
+// -----
+func.func @test_store_scatter_vc_2(%src: ui64) {
+  %0 = arith.constant dense<1>: vector<4xi1>
+  %1 = arith.constant dense<2.9>: vector<4x2xf32>
+  %2 = xegpu.create_tdesc %src[0, 8, 16, 24] {chunk_size = 2}
+          : ui64 -> !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>
+  // expected-error@+1 {{invlid l1_hint: #xegpu.cache_hint<streaming>}}
+  xegpu.store %1, %2, %0 <{l1_hint = #xegpu.cache_hint<streaming>}> : vector<4x2xf32>,
+          !xegpu.tensor_desc<4x2xf32, #xegpu.tdesc_attr<scattered = true>>, vector<4xi1>
+  return
+}
\ No newline at end of file

From 9ec8c961664de3b3fcc1cbd5238e40ec8c9bdddb Mon Sep 17 00:00:00 2001
From: Samira Bazuzi <bazuzi@google.com>
Date: Tue, 16 Apr 2024 14:46:05 -0400
Subject: [PATCH 640/886] [clang][dataflow] Expose getReferencedDecls and
 relocate free functions. (#88754)

Moves free functions from DataflowEnvironment.h/cc and
DataflowAnalysisContext.h/cc to RecordOps and a new ASTOps and exposes
them as needed for current use and to expose getReferencedDecls for
out-of-tree use.

Minimal change in functionality, only to modify the return type of
getReferenceDecls to return the collected decls instead of using output
params.

Tested with `ninja check-clang-tooling`.
---
 clang/docs/tools/clang-formatted-files.txt    |   2 +
 .../clang/Analysis/FlowSensitive/ASTOps.h     |  98 +++++++
 .../FlowSensitive/DataflowAnalysisContext.h   |  28 +-
 .../FlowSensitive/DataflowEnvironment.h       |  36 ---
 clang/lib/Analysis/FlowSensitive/ASTOps.cpp   | 249 ++++++++++++++++++
 .../lib/Analysis/FlowSensitive/CMakeLists.txt |   1 +
 .../FlowSensitive/DataflowAnalysisContext.cpp |  53 +---
 .../FlowSensitive/DataflowEnvironment.cpp     | 177 +------------
 clang/lib/Analysis/FlowSensitive/Transfer.cpp |   2 +
 9 files changed, 359 insertions(+), 287 deletions(-)
 create mode 100644 clang/include/clang/Analysis/FlowSensitive/ASTOps.h
 create mode 100644 clang/lib/Analysis/FlowSensitive/ASTOps.cpp

diff --git a/clang/docs/tools/clang-formatted-files.txt b/clang/docs/tools/clang-formatted-files.txt
index 3089438c23d94..2252d0ccde96d 100644
--- a/clang/docs/tools/clang-formatted-files.txt
+++ b/clang/docs/tools/clang-formatted-files.txt
@@ -123,6 +123,7 @@ clang/include/clang/Analysis/Analyses/CalledOnceCheck.h
 clang/include/clang/Analysis/Analyses/CFGReachabilityAnalysis.h
 clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
 clang/include/clang/Analysis/FlowSensitive/AdornedCFG.h
+clang/include/clang/Analysis/FlowSensitive/ASTOps.h
 clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
 clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h
 clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
@@ -307,6 +308,7 @@ clang/lib/Analysis/CalledOnceCheck.cpp
 clang/lib/Analysis/CloneDetection.cpp
 clang/lib/Analysis/CodeInjector.cpp
 clang/lib/Analysis/FlowSensitive/AdornedCFG.cpp
+clang/lib/Analysis/FlowSensitive/ASTOps.cpp
 clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp
 clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
 clang/lib/Analysis/FlowSensitive/DebugSupport.cpp
diff --git a/clang/include/clang/Analysis/FlowSensitive/ASTOps.h b/clang/include/clang/Analysis/FlowSensitive/ASTOps.h
new file mode 100644
index 0000000000000..27ad32c1694f7
--- /dev/null
+++ b/clang/include/clang/Analysis/FlowSensitive/ASTOps.h
@@ -0,0 +1,98 @@
+//===-- ASTOps.h -------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+//  Operations on AST nodes that are used in flow-sensitive analysis.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_ASTOPS_H
+#define LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_ASTOPS_H
+
+#include "clang/AST/Decl.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/Type.h"
+#include "clang/Analysis/FlowSensitive/StorageLocation.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/SetVector.h"
+
+namespace clang {
+namespace dataflow {
+
+/// Skip past nodes that the CFG does not emit. These nodes are invisible to
+/// flow-sensitive analysis, and should be ignored as they will effectively not
+/// exist.
+///
+///   * `ParenExpr` - The CFG takes the operator precedence into account, but
+///   otherwise omits the node afterwards.
+///
+///   * `ExprWithCleanups` - The CFG will generate the appropriate calls to
+///   destructors and then omit the node.
+///
+const Expr &ignoreCFGOmittedNodes(const Expr &E);
+const Stmt &ignoreCFGOmittedNodes(const Stmt &S);
+
+/// A set of `FieldDecl *`. Use `SmallSetVector` to guarantee deterministic
+/// iteration order.
+using FieldSet = llvm::SmallSetVector<const FieldDecl *, 4>;
+
+/// Returns the set of all fields in the type.
+FieldSet getObjectFields(QualType Type);
+
+/// Returns whether `Fields` and `FieldLocs` contain the same fields.
+bool containsSameFields(const FieldSet &Fields,
+                        const RecordStorageLocation::FieldToLoc &FieldLocs);
+
+/// Helper class for initialization of a record with an `InitListExpr`.
+/// `InitListExpr::inits()` contains the initializers for both the base classes
+/// and the fields of the record; this helper class separates these out into two
+/// different lists. In addition, it deals with special cases associated with
+/// unions.
+class RecordInitListHelper {
+public:
+  // `InitList` must have record type.
+  RecordInitListHelper(const InitListExpr *InitList);
+
+  // Base classes with their associated initializer expressions.
+  ArrayRef<std::pair<const CXXBaseSpecifier *, Expr *>> base_inits() const {
+    return BaseInits;
+  }
+
+  // Fields with their associated initializer expressions.
+  ArrayRef<std::pair<const FieldDecl *, Expr *>> field_inits() const {
+    return FieldInits;
+  }
+
+private:
+  SmallVector<std::pair<const CXXBaseSpecifier *, Expr *>> BaseInits;
+  SmallVector<std::pair<const FieldDecl *, Expr *>> FieldInits;
+
+  // We potentially synthesize an `ImplicitValueInitExpr` for unions. It's a
+  // member variable because we store a pointer to it in `FieldInits`.
+  std::optional<ImplicitValueInitExpr> ImplicitValueInitForUnion;
+};
+
+/// A collection of several types of declarations, all referenced from the same
+/// function.
+struct ReferencedDecls {
+  /// Non-static member variables.
+  FieldSet Fields;
+  /// All variables with static storage duration, notably including static
+  /// member variables and static variables declared within a function.
+  llvm::DenseSet<const VarDecl *> Globals;
+  /// Free functions and member functions which are referenced (but not
+  /// necessarily called).
+  llvm::DenseSet<const FunctionDecl *> Functions;
+};
+
+/// Returns declarations that are declared in or referenced from `FD`.
+ReferencedDecls getReferencedDecls(const FunctionDecl &FD);
+
+} // namespace dataflow
+} // namespace clang
+
+#endif // LLVM_CLANG_ANALYSIS_FLOWSENSITIVE_ASTOPS_H
diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h b/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h
index 909a91059438c..aa2c366cb164a 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysisContext.h
@@ -18,6 +18,7 @@
 #include "clang/AST/Decl.h"
 #include "clang/AST/Expr.h"
 #include "clang/AST/TypeOrdering.h"
+#include "clang/Analysis/FlowSensitive/ASTOps.h"
 #include "clang/Analysis/FlowSensitive/AdornedCFG.h"
 #include "clang/Analysis/FlowSensitive/Arena.h"
 #include "clang/Analysis/FlowSensitive/Solver.h"
@@ -30,38 +31,11 @@
 #include <cassert>
 #include <memory>
 #include <optional>
-#include <type_traits>
-#include <utility>
-#include <vector>
 
 namespace clang {
 namespace dataflow {
 class Logger;
 
-/// Skip past nodes that the CFG does not emit. These nodes are invisible to
-/// flow-sensitive analysis, and should be ignored as they will effectively not
-/// exist.
-///
-///   * `ParenExpr` - The CFG takes the operator precedence into account, but
-///   otherwise omits the node afterwards.
-///
-///   * `ExprWithCleanups` - The CFG will generate the appropriate calls to
-///   destructors and then omit the node.
-///
-const Expr &ignoreCFGOmittedNodes(const Expr &E);
-const Stmt &ignoreCFGOmittedNodes(const Stmt &S);
-
-/// A set of `FieldDecl *`. Use `SmallSetVector` to guarantee deterministic
-/// iteration order.
-using FieldSet = llvm::SmallSetVector<const FieldDecl *, 4>;
-
-/// Returns the set of all fields in the type.
-FieldSet getObjectFields(QualType Type);
-
-/// Returns whether `Fields` and `FieldLocs` contain the same fields.
-bool containsSameFields(const FieldSet &Fields,
-                        const RecordStorageLocation::FieldToLoc &FieldLocs);
-
 struct ContextSensitiveOptions {
   /// The maximum depth to analyze. A value of zero is equivalent to disabling
   /// context-sensitive analysis entirely.
diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
index 706664d7db1c2..4277792219c0a 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowEnvironment.h
@@ -775,42 +775,6 @@ RecordStorageLocation *getImplicitObjectLocation(const CXXMemberCallExpr &MCE,
 RecordStorageLocation *getBaseObjectLocation(const MemberExpr &ME,
                                              const Environment &Env);
 
-/// Returns the fields of a `RecordDecl` that are initialized by an
-/// `InitListExpr`, in the order in which they appear in
-/// `InitListExpr::inits()`.
-/// `Init->getType()` must be a record type.
-std::vector<const FieldDecl *>
-getFieldsForInitListExpr(const InitListExpr *InitList);
-
-/// Helper class for initialization of a record with an `InitListExpr`.
-/// `InitListExpr::inits()` contains the initializers for both the base classes
-/// and the fields of the record; this helper class separates these out into two
-/// different lists. In addition, it deals with special cases associated with
-/// unions.
-class RecordInitListHelper {
-public:
-  // `InitList` must have record type.
-  RecordInitListHelper(const InitListExpr *InitList);
-
-  // Base classes with their associated initializer expressions.
-  ArrayRef<std::pair<const CXXBaseSpecifier *, Expr *>> base_inits() const {
-    return BaseInits;
-  }
-
-  // Fields with their associated initializer expressions.
-  ArrayRef<std::pair<const FieldDecl *, Expr *>> field_inits() const {
-    return FieldInits;
-  }
-
-private:
-  SmallVector<std::pair<const CXXBaseSpecifier *, Expr *>> BaseInits;
-  SmallVector<std::pair<const FieldDecl *, Expr *>> FieldInits;
-
-  // We potentially synthesize an `ImplicitValueInitExpr` for unions. It's a
-  // member variable because we store a pointer to it in `FieldInits`.
-  std::optional<ImplicitValueInitExpr> ImplicitValueInitForUnion;
-};
-
 /// Associates a new `RecordValue` with `Loc` and returns the new value.
 RecordValue &refreshRecordValue(RecordStorageLocation &Loc, Environment &Env);
 
diff --git a/clang/lib/Analysis/FlowSensitive/ASTOps.cpp b/clang/lib/Analysis/FlowSensitive/ASTOps.cpp
new file mode 100644
index 0000000000000..75188aef4d1a4
--- /dev/null
+++ b/clang/lib/Analysis/FlowSensitive/ASTOps.cpp
@@ -0,0 +1,249 @@
+//===-- ASTOps.cc -------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+//  Operations on AST nodes that are used in flow-sensitive analysis.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Analysis/FlowSensitive/ASTOps.h"
+#include "clang/AST/ComputeDependence.h"
+#include "clang/AST/Decl.h"
+#include "clang/AST/DeclBase.h"
+#include "clang/AST/DeclCXX.h"
+#include "clang/AST/Expr.h"
+#include "clang/AST/ExprCXX.h"
+#include "clang/AST/Stmt.h"
+#include "clang/AST/Type.h"
+#include "clang/Analysis/FlowSensitive/StorageLocation.h"
+#include "clang/Basic/LLVM.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/STLExtras.h"
+#include <cassert>
+#include <iterator>
+#include <vector>
+
+#define DEBUG_TYPE "dataflow"
+
+namespace clang::dataflow {
+
+const Expr &ignoreCFGOmittedNodes(const Expr &E) {
+  const Expr *Current = &E;
+  if (auto *EWC = dyn_cast<ExprWithCleanups>(Current)) {
+    Current = EWC->getSubExpr();
+    assert(Current != nullptr);
+  }
+  Current = Current->IgnoreParens();
+  assert(Current != nullptr);
+  return *Current;
+}
+
+const Stmt &ignoreCFGOmittedNodes(const Stmt &S) {
+  if (auto *E = dyn_cast<Expr>(&S))
+    return ignoreCFGOmittedNodes(*E);
+  return S;
+}
+
+// FIXME: Does not precisely handle non-virtual diamond inheritance. A single
+// field decl will be modeled for all instances of the inherited field.
+static void getFieldsFromClassHierarchy(QualType Type, FieldSet &Fields) {
+  if (Type->isIncompleteType() || Type->isDependentType() ||
+      !Type->isRecordType())
+    return;
+
+  for (const FieldDecl *Field : Type->getAsRecordDecl()->fields())
+    Fields.insert(Field);
+  if (auto *CXXRecord = Type->getAsCXXRecordDecl())
+    for (const CXXBaseSpecifier &Base : CXXRecord->bases())
+      getFieldsFromClassHierarchy(Base.getType(), Fields);
+}
+
+/// Gets the set of all fields in the type.
+FieldSet getObjectFields(QualType Type) {
+  FieldSet Fields;
+  getFieldsFromClassHierarchy(Type, Fields);
+  return Fields;
+}
+
+bool containsSameFields(const FieldSet &Fields,
+                        const RecordStorageLocation::FieldToLoc &FieldLocs) {
+  if (Fields.size() != FieldLocs.size())
+    return false;
+  for ([[maybe_unused]] auto [Field, Loc] : FieldLocs)
+    if (!Fields.contains(cast_or_null<FieldDecl>(Field)))
+      return false;
+  return true;
+}
+
+/// Returns the fields of a `RecordDecl` that are initialized by an
+/// `InitListExpr`, in the order in which they appear in
+/// `InitListExpr::inits()`.
+/// `Init->getType()` must be a record type.
+static std::vector<const FieldDecl *>
+getFieldsForInitListExpr(const InitListExpr *InitList) {
+  const RecordDecl *RD = InitList->getType()->getAsRecordDecl();
+  assert(RD != nullptr);
+
+  std::vector<const FieldDecl *> Fields;
+
+  if (InitList->getType()->isUnionType()) {
+    Fields.push_back(InitList->getInitializedFieldInUnion());
+    return Fields;
+  }
+
+  // Unnamed bitfields are only used for padding and do not appear in
+  // `InitListExpr`'s inits. However, those fields do appear in `RecordDecl`'s
+  // field list, and we thus need to remove them before mapping inits to
+  // fields to avoid mapping inits to the wrongs fields.
+  llvm::copy_if(
+      RD->fields(), std::back_inserter(Fields),
+      [](const FieldDecl *Field) { return !Field->isUnnamedBitfield(); });
+  return Fields;
+}
+
+RecordInitListHelper::RecordInitListHelper(const InitListExpr *InitList) {
+  auto *RD = InitList->getType()->getAsCXXRecordDecl();
+  assert(RD != nullptr);
+
+  std::vector<const FieldDecl *> Fields = getFieldsForInitListExpr(InitList);
+  ArrayRef<Expr *> Inits = InitList->inits();
+
+  // Unions initialized with an empty initializer list need special treatment.
+  // For structs/classes initialized with an empty initializer list, Clang
+  // puts `ImplicitValueInitExpr`s in `InitListExpr::inits()`, but for unions,
+  // it doesn't do this -- so we create an `ImplicitValueInitExpr` ourselves.
+  SmallVector<Expr *> InitsForUnion;
+  if (InitList->getType()->isUnionType() && Inits.empty()) {
+    assert(Fields.size() == 1);
+    ImplicitValueInitForUnion.emplace(Fields.front()->getType());
+    InitsForUnion.push_back(&*ImplicitValueInitForUnion);
+    Inits = InitsForUnion;
+  }
+
+  size_t InitIdx = 0;
+
+  assert(Fields.size() + RD->getNumBases() == Inits.size());
+  for (const CXXBaseSpecifier &Base : RD->bases()) {
+    assert(InitIdx < Inits.size());
+    Expr *Init = Inits[InitIdx++];
+    BaseInits.emplace_back(&Base, Init);
+  }
+
+  assert(Fields.size() == Inits.size() - InitIdx);
+  for (const FieldDecl *Field : Fields) {
+    assert(InitIdx < Inits.size());
+    Expr *Init = Inits[InitIdx++];
+    FieldInits.emplace_back(Field, Init);
+  }
+}
+
+static void insertIfGlobal(const Decl &D,
+                           llvm::DenseSet<const VarDecl *> &Globals) {
+  if (auto *V = dyn_cast<VarDecl>(&D))
+    if (V->hasGlobalStorage())
+      Globals.insert(V);
+}
+
+static void insertIfFunction(const Decl &D,
+                             llvm::DenseSet<const FunctionDecl *> &Funcs) {
+  if (auto *FD = dyn_cast<FunctionDecl>(&D))
+    Funcs.insert(FD);
+}
+
+static MemberExpr *getMemberForAccessor(const CXXMemberCallExpr &C) {
+  // Use getCalleeDecl instead of getMethodDecl in order to handle
+  // pointer-to-member calls.
+  const auto *MethodDecl = dyn_cast_or_null<CXXMethodDecl>(C.getCalleeDecl());
+  if (!MethodDecl)
+    return nullptr;
+  auto *Body = dyn_cast_or_null<CompoundStmt>(MethodDecl->getBody());
+  if (!Body || Body->size() != 1)
+    return nullptr;
+  if (auto *RS = dyn_cast<ReturnStmt>(*Body->body_begin()))
+    if (auto *Return = RS->getRetValue())
+      return dyn_cast<MemberExpr>(Return->IgnoreParenImpCasts());
+  return nullptr;
+}
+
+static void getReferencedDecls(const Decl &D, ReferencedDecls &Referenced) {
+  insertIfGlobal(D, Referenced.Globals);
+  insertIfFunction(D, Referenced.Functions);
+  if (const auto *Decomp = dyn_cast<DecompositionDecl>(&D))
+    for (const auto *B : Decomp->bindings())
+      if (auto *ME = dyn_cast_or_null<MemberExpr>(B->getBinding()))
+        // FIXME: should we be using `E->getFoundDecl()`?
+        if (const auto *FD = dyn_cast<FieldDecl>(ME->getMemberDecl()))
+          Referenced.Fields.insert(FD);
+}
+
+/// Traverses `S` and inserts into `Referenced` any declarations that are
+/// declared in or referenced from sub-statements.
+static void getReferencedDecls(const Stmt &S, ReferencedDecls &Referenced) {
+  for (auto *Child : S.children())
+    if (Child != nullptr)
+      getReferencedDecls(*Child, Referenced);
+  if (const auto *DefaultArg = dyn_cast<CXXDefaultArgExpr>(&S))
+    getReferencedDecls(*DefaultArg->getExpr(), Referenced);
+  if (const auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(&S))
+    getReferencedDecls(*DefaultInit->getExpr(), Referenced);
+
+  if (auto *DS = dyn_cast<DeclStmt>(&S)) {
+    if (DS->isSingleDecl())
+      getReferencedDecls(*DS->getSingleDecl(), Referenced);
+    else
+      for (auto *D : DS->getDeclGroup())
+        getReferencedDecls(*D, Referenced);
+  } else if (auto *E = dyn_cast<DeclRefExpr>(&S)) {
+    insertIfGlobal(*E->getDecl(), Referenced.Globals);
+    insertIfFunction(*E->getDecl(), Referenced.Functions);
+  } else if (const auto *C = dyn_cast<CXXMemberCallExpr>(&S)) {
+    // If this is a method that returns a member variable but does nothing else,
+    // model the field of the return value.
+    if (MemberExpr *E = getMemberForAccessor(*C))
+      if (const auto *FD = dyn_cast<FieldDecl>(E->getMemberDecl()))
+        Referenced.Fields.insert(FD);
+  } else if (auto *E = dyn_cast<MemberExpr>(&S)) {
+    // FIXME: should we be using `E->getFoundDecl()`?
+    const ValueDecl *VD = E->getMemberDecl();
+    insertIfGlobal(*VD, Referenced.Globals);
+    insertIfFunction(*VD, Referenced.Functions);
+    if (const auto *FD = dyn_cast<FieldDecl>(VD))
+      Referenced.Fields.insert(FD);
+  } else if (auto *InitList = dyn_cast<InitListExpr>(&S)) {
+    if (InitList->getType()->isRecordType())
+      for (const auto *FD : getFieldsForInitListExpr(InitList))
+        Referenced.Fields.insert(FD);
+  }
+}
+
+ReferencedDecls getReferencedDecls(const FunctionDecl &FD) {
+  ReferencedDecls Result;
+  // Look for global variable and field references in the
+  // constructor-initializers.
+  if (const auto *CtorDecl = dyn_cast<CXXConstructorDecl>(&FD)) {
+    for (const auto *Init : CtorDecl->inits()) {
+      if (Init->isMemberInitializer()) {
+        Result.Fields.insert(Init->getMember());
+      } else if (Init->isIndirectMemberInitializer()) {
+        for (const auto *I : Init->getIndirectMember()->chain())
+          Result.Fields.insert(cast<FieldDecl>(I));
+      }
+      const Expr *E = Init->getInit();
+      assert(E != nullptr);
+      getReferencedDecls(*E, Result);
+    }
+    // Add all fields mentioned in default member initializers.
+    for (const FieldDecl *F : CtorDecl->getParent()->fields())
+      if (const auto *I = F->getInClassInitializer())
+        getReferencedDecls(*I, Result);
+  }
+  getReferencedDecls(*FD.getBody(), Result);
+
+  return Result;
+}
+
+} // namespace clang::dataflow
diff --git a/clang/lib/Analysis/FlowSensitive/CMakeLists.txt b/clang/lib/Analysis/FlowSensitive/CMakeLists.txt
index a3b5d9adc24bd..6631fe27f3d90 100644
--- a/clang/lib/Analysis/FlowSensitive/CMakeLists.txt
+++ b/clang/lib/Analysis/FlowSensitive/CMakeLists.txt
@@ -1,6 +1,7 @@
 add_clang_library(clangAnalysisFlowSensitive
   AdornedCFG.cpp
   Arena.cpp
+  ASTOps.cpp
   DataflowAnalysisContext.cpp
   DataflowEnvironment.cpp
   Formula.cpp
diff --git a/clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp b/clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp
index d520539dd2535..e94fd39c45dc1 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp
@@ -14,6 +14,7 @@
 
 #include "clang/Analysis/FlowSensitive/DataflowAnalysisContext.h"
 #include "clang/AST/ExprCXX.h"
+#include "clang/Analysis/FlowSensitive/ASTOps.h"
 #include "clang/Analysis/FlowSensitive/DebugSupport.h"
 #include "clang/Analysis/FlowSensitive/Formula.h"
 #include "clang/Analysis/FlowSensitive/Logger.h"
@@ -359,55 +360,3 @@ DataflowAnalysisContext::~DataflowAnalysisContext() = default;
 
 } // namespace dataflow
 } // namespace clang
-
-using namespace clang;
-
-const Expr &clang::dataflow::ignoreCFGOmittedNodes(const Expr &E) {
-  const Expr *Current = &E;
-  if (auto *EWC = dyn_cast<ExprWithCleanups>(Current)) {
-    Current = EWC->getSubExpr();
-    assert(Current != nullptr);
-  }
-  Current = Current->IgnoreParens();
-  assert(Current != nullptr);
-  return *Current;
-}
-
-const Stmt &clang::dataflow::ignoreCFGOmittedNodes(const Stmt &S) {
-  if (auto *E = dyn_cast<Expr>(&S))
-    return ignoreCFGOmittedNodes(*E);
-  return S;
-}
-
-// FIXME: Does not precisely handle non-virtual diamond inheritance. A single
-// field decl will be modeled for all instances of the inherited field.
-static void getFieldsFromClassHierarchy(QualType Type,
-                                        clang::dataflow::FieldSet &Fields) {
-  if (Type->isIncompleteType() || Type->isDependentType() ||
-      !Type->isRecordType())
-    return;
-
-  for (const FieldDecl *Field : Type->getAsRecordDecl()->fields())
-    Fields.insert(Field);
-  if (auto *CXXRecord = Type->getAsCXXRecordDecl())
-    for (const CXXBaseSpecifier &Base : CXXRecord->bases())
-      getFieldsFromClassHierarchy(Base.getType(), Fields);
-}
-
-/// Gets the set of all fields in the type.
-clang::dataflow::FieldSet clang::dataflow::getObjectFields(QualType Type) {
-  FieldSet Fields;
-  getFieldsFromClassHierarchy(Type, Fields);
-  return Fields;
-}
-
-bool clang::dataflow::containsSameFields(
-    const clang::dataflow::FieldSet &Fields,
-    const clang::dataflow::RecordStorageLocation::FieldToLoc &FieldLocs) {
-  if (Fields.size() != FieldLocs.size())
-    return false;
-  for ([[maybe_unused]] auto [Field, Loc] : FieldLocs)
-    if (!Fields.contains(cast_or_null<FieldDecl>(Field)))
-      return false;
-  return true;
-}
diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index ee2581143e114..3bf3807268bee 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -17,6 +17,7 @@
 #include "clang/AST/DeclCXX.h"
 #include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/AST/Type.h"
+#include "clang/Analysis/FlowSensitive/ASTOps.h"
 #include "clang/Analysis/FlowSensitive/DataflowLattice.h"
 #include "clang/Analysis/FlowSensitive/Value.h"
 #include "llvm/ADT/DenseMap.h"
@@ -304,93 +305,6 @@ widenKeyToValueMap(const llvm::MapVector<Key, Value *> &CurMap,
   return WidenedMap;
 }
 
-/// Initializes a global storage value.
-static void insertIfGlobal(const Decl &D,
-                           llvm::DenseSet<const VarDecl *> &Vars) {
-  if (auto *V = dyn_cast<VarDecl>(&D))
-    if (V->hasGlobalStorage())
-      Vars.insert(V);
-}
-
-static void insertIfFunction(const Decl &D,
-                             llvm::DenseSet<const FunctionDecl *> &Funcs) {
-  if (auto *FD = dyn_cast<FunctionDecl>(&D))
-    Funcs.insert(FD);
-}
-
-static MemberExpr *getMemberForAccessor(const CXXMemberCallExpr &C) {
-  // Use getCalleeDecl instead of getMethodDecl in order to handle
-  // pointer-to-member calls.
-  const auto *MethodDecl = dyn_cast_or_null<CXXMethodDecl>(C.getCalleeDecl());
-  if (!MethodDecl)
-    return nullptr;
-  auto *Body = dyn_cast_or_null<CompoundStmt>(MethodDecl->getBody());
-  if (!Body || Body->size() != 1)
-    return nullptr;
-  if (auto *RS = dyn_cast<ReturnStmt>(*Body->body_begin()))
-    if (auto *Return = RS->getRetValue())
-      return dyn_cast<MemberExpr>(Return->IgnoreParenImpCasts());
-  return nullptr;
-}
-
-static void
-getFieldsGlobalsAndFuncs(const Decl &D, FieldSet &Fields,
-                         llvm::DenseSet<const VarDecl *> &Vars,
-                         llvm::DenseSet<const FunctionDecl *> &Funcs) {
-  insertIfGlobal(D, Vars);
-  insertIfFunction(D, Funcs);
-  if (const auto *Decomp = dyn_cast<DecompositionDecl>(&D))
-    for (const auto *B : Decomp->bindings())
-      if (auto *ME = dyn_cast_or_null<MemberExpr>(B->getBinding()))
-        // FIXME: should we be using `E->getFoundDecl()`?
-        if (const auto *FD = dyn_cast<FieldDecl>(ME->getMemberDecl()))
-          Fields.insert(FD);
-}
-
-/// Traverses `S` and inserts into `Fields`, `Vars` and `Funcs` any fields,
-/// global variables and functions that are declared in or referenced from
-/// sub-statements.
-static void
-getFieldsGlobalsAndFuncs(const Stmt &S, FieldSet &Fields,
-                         llvm::DenseSet<const VarDecl *> &Vars,
-                         llvm::DenseSet<const FunctionDecl *> &Funcs) {
-  for (auto *Child : S.children())
-    if (Child != nullptr)
-      getFieldsGlobalsAndFuncs(*Child, Fields, Vars, Funcs);
-  if (const auto *DefaultArg = dyn_cast<CXXDefaultArgExpr>(&S))
-    getFieldsGlobalsAndFuncs(*DefaultArg->getExpr(), Fields, Vars, Funcs);
-  if (const auto *DefaultInit = dyn_cast<CXXDefaultInitExpr>(&S))
-    getFieldsGlobalsAndFuncs(*DefaultInit->getExpr(), Fields, Vars, Funcs);
-
-  if (auto *DS = dyn_cast<DeclStmt>(&S)) {
-    if (DS->isSingleDecl())
-      getFieldsGlobalsAndFuncs(*DS->getSingleDecl(), Fields, Vars, Funcs);
-    else
-      for (auto *D : DS->getDeclGroup())
-        getFieldsGlobalsAndFuncs(*D, Fields, Vars, Funcs);
-  } else if (auto *E = dyn_cast<DeclRefExpr>(&S)) {
-    insertIfGlobal(*E->getDecl(), Vars);
-    insertIfFunction(*E->getDecl(), Funcs);
-  } else if (const auto *C = dyn_cast<CXXMemberCallExpr>(&S)) {
-    // If this is a method that returns a member variable but does nothing else,
-    // model the field of the return value.
-    if (MemberExpr *E = getMemberForAccessor(*C))
-      if (const auto *FD = dyn_cast<FieldDecl>(E->getMemberDecl()))
-        Fields.insert(FD);
-  } else if (auto *E = dyn_cast<MemberExpr>(&S)) {
-    // FIXME: should we be using `E->getFoundDecl()`?
-    const ValueDecl *VD = E->getMemberDecl();
-    insertIfGlobal(*VD, Vars);
-    insertIfFunction(*VD, Funcs);
-    if (const auto *FD = dyn_cast<FieldDecl>(VD))
-      Fields.insert(FD);
-  } else if (auto *InitList = dyn_cast<InitListExpr>(&S)) {
-    if (InitList->getType()->isRecordType())
-      for (const auto *FD : getFieldsForInitListExpr(InitList))
-        Fields.insert(FD);
-  }
-}
-
 namespace {
 
 // Visitor that builds a map from record prvalues to result objects.
@@ -653,36 +567,13 @@ void Environment::initialize() {
 void Environment::initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl) {
   assert(FuncDecl->doesThisDeclarationHaveABody());
 
-  FieldSet Fields;
-  llvm::DenseSet<const VarDecl *> Vars;
-  llvm::DenseSet<const FunctionDecl *> Funcs;
-
-  // Look for global variable and field references in the
-  // constructor-initializers.
-  if (const auto *CtorDecl = dyn_cast<CXXConstructorDecl>(FuncDecl)) {
-    for (const auto *Init : CtorDecl->inits()) {
-      if (Init->isMemberInitializer()) {
-        Fields.insert(Init->getMember());
-      } else if (Init->isIndirectMemberInitializer()) {
-        for (const auto *I : Init->getIndirectMember()->chain())
-          Fields.insert(cast<FieldDecl>(I));
-      }
-      const Expr *E = Init->getInit();
-      assert(E != nullptr);
-      getFieldsGlobalsAndFuncs(*E, Fields, Vars, Funcs);
-    }
-    // Add all fields mentioned in default member initializers.
-    for (const FieldDecl *F : CtorDecl->getParent()->fields())
-      if (const auto *I = F->getInClassInitializer())
-          getFieldsGlobalsAndFuncs(*I, Fields, Vars, Funcs);
-  }
-  getFieldsGlobalsAndFuncs(*FuncDecl->getBody(), Fields, Vars, Funcs);
+  ReferencedDecls Referenced = getReferencedDecls(*FuncDecl);
 
   // These have to be added before the lines that follow to ensure that
   // `create*` work correctly for structs.
-  DACtx->addModeledFields(Fields);
+  DACtx->addModeledFields(Referenced.Fields);
 
-  for (const VarDecl *D : Vars) {
+  for (const VarDecl *D : Referenced.Globals) {
     if (getStorageLocation(*D) != nullptr)
       continue;
 
@@ -694,7 +585,7 @@ void Environment::initFieldsGlobalsAndFuncs(const FunctionDecl *FuncDecl) {
     setStorageLocation(*D, createObject(*D, nullptr));
   }
 
-  for (const FunctionDecl *FD : Funcs) {
+  for (const FunctionDecl *FD : Referenced.Functions) {
     if (getStorageLocation(*FD) != nullptr)
       continue;
     auto &Loc = createStorageLocation(*FD);
@@ -1354,64 +1245,6 @@ RecordStorageLocation *getBaseObjectLocation(const MemberExpr &ME,
   return Env.get<RecordStorageLocation>(*Base);
 }
 
-std::vector<const FieldDecl *>
-getFieldsForInitListExpr(const InitListExpr *InitList) {
-  const RecordDecl *RD = InitList->getType()->getAsRecordDecl();
-  assert(RD != nullptr);
-
-  std::vector<const FieldDecl *> Fields;
-
-  if (InitList->getType()->isUnionType()) {
-    Fields.push_back(InitList->getInitializedFieldInUnion());
-    return Fields;
-  }
-
-  // Unnamed bitfields are only used for padding and do not appear in
-  // `InitListExpr`'s inits. However, those fields do appear in `RecordDecl`'s
-  // field list, and we thus need to remove them before mapping inits to
-  // fields to avoid mapping inits to the wrongs fields.
-  llvm::copy_if(
-      RD->fields(), std::back_inserter(Fields),
-      [](const FieldDecl *Field) { return !Field->isUnnamedBitfield(); });
-  return Fields;
-}
-
-RecordInitListHelper::RecordInitListHelper(const InitListExpr *InitList) {
-  auto *RD = InitList->getType()->getAsCXXRecordDecl();
-  assert(RD != nullptr);
-
-  std::vector<const FieldDecl *> Fields = getFieldsForInitListExpr(InitList);
-  ArrayRef<Expr *> Inits = InitList->inits();
-
-  // Unions initialized with an empty initializer list need special treatment.
-  // For structs/classes initialized with an empty initializer list, Clang
-  // puts `ImplicitValueInitExpr`s in `InitListExpr::inits()`, but for unions,
-  // it doesn't do this -- so we create an `ImplicitValueInitExpr` ourselves.
-  SmallVector<Expr *> InitsForUnion;
-  if (InitList->getType()->isUnionType() && Inits.empty()) {
-    assert(Fields.size() == 1);
-    ImplicitValueInitForUnion.emplace(Fields.front()->getType());
-    InitsForUnion.push_back(&*ImplicitValueInitForUnion);
-    Inits = InitsForUnion;
-  }
-
-  size_t InitIdx = 0;
-
-  assert(Fields.size() + RD->getNumBases() == Inits.size());
-  for (const CXXBaseSpecifier &Base : RD->bases()) {
-    assert(InitIdx < Inits.size());
-    Expr *Init = Inits[InitIdx++];
-    BaseInits.emplace_back(&Base, Init);
-  }
-
-  assert(Fields.size() == Inits.size() - InitIdx);
-  for (const FieldDecl *Field : Fields) {
-    assert(InitIdx < Inits.size());
-    Expr *Init = Inits[InitIdx++];
-    FieldInits.emplace_back(Field, Init);
-  }
-}
-
 RecordValue &refreshRecordValue(RecordStorageLocation &Loc, Environment &Env) {
   auto &NewVal = Env.create<RecordValue>(Loc);
   Env.setValue(Loc, NewVal);
diff --git a/clang/lib/Analysis/FlowSensitive/Transfer.cpp b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
index 88a9c0eccbebc..1e034771014ea 100644
--- a/clang/lib/Analysis/FlowSensitive/Transfer.cpp
+++ b/clang/lib/Analysis/FlowSensitive/Transfer.cpp
@@ -20,7 +20,9 @@
 #include "clang/AST/OperationKinds.h"
 #include "clang/AST/Stmt.h"
 #include "clang/AST/StmtVisitor.h"
+#include "clang/Analysis/FlowSensitive/ASTOps.h"
 #include "clang/Analysis/FlowSensitive/AdornedCFG.h"
+#include "clang/Analysis/FlowSensitive/DataflowAnalysisContext.h"
 #include "clang/Analysis/FlowSensitive/DataflowEnvironment.h"
 #include "clang/Analysis/FlowSensitive/NoopAnalysis.h"
 #include "clang/Analysis/FlowSensitive/RecordOps.h"

From bbd64c4ddf08be468ab4eb4c161e28bdab6808bb Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Tue, 16 Apr 2024 11:40:23 -0700
Subject: [PATCH 641/886] [RISCV] Add coverage for strength reduction of mul as
 2^N - 2^M

---
 llvm/test/CodeGen/RISCV/mul.ll | 196 ++++++++++++++++++++++++++++++++-
 1 file changed, 192 insertions(+), 4 deletions(-)

diff --git a/llvm/test/CodeGen/RISCV/mul.ll b/llvm/test/CodeGen/RISCV/mul.ll
index af341dbaadeab..364e8c7b38dac 100644
--- a/llvm/test/CodeGen/RISCV/mul.ll
+++ b/llvm/test/CodeGen/RISCV/mul.ll
@@ -465,6 +465,192 @@ define i32 @mulhu_constant(i32 %a) nounwind {
   ret i32 %4
 }
 
+define i32 @muli32_p14(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p14:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 14
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p14:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 14
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p14:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 14
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p14:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 14
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 14
+  ret i32 %1
+}
+
+define i32 @muli32_p28(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p28:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 28
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p28:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 28
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p28:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 28
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p28:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 28
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 28
+  ret i32 %1
+}
+
+define i32 @muli32_p30(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p30:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 30
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p30:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 30
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p30:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 30
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p30:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 30
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 30
+  ret i32 %1
+}
+
+define i32 @muli32_p56(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p56:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 56
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p56:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 56
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p56:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 56
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p56:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 56
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 56
+  ret i32 %1
+}
+
+define i32 @muli32_p60(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p60:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 60
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p60:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 60
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p60:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 60
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p60:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 60
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 60
+  ret i32 %1
+}
+
+define i32 @muli32_p62(i32 %a) nounwind {
+; RV32I-LABEL: muli32_p62:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    li a1, 62
+; RV32I-NEXT:    tail __mulsi3
+;
+; RV32IM-LABEL: muli32_p62:
+; RV32IM:       # %bb.0:
+; RV32IM-NEXT:    li a1, 62
+; RV32IM-NEXT:    mul a0, a0, a1
+; RV32IM-NEXT:    ret
+;
+; RV64I-LABEL: muli32_p62:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    addi sp, sp, -16
+; RV64I-NEXT:    sd ra, 8(sp) # 8-byte Folded Spill
+; RV64I-NEXT:    li a1, 62
+; RV64I-NEXT:    call __muldi3
+; RV64I-NEXT:    ld ra, 8(sp) # 8-byte Folded Reload
+; RV64I-NEXT:    addi sp, sp, 16
+; RV64I-NEXT:    ret
+;
+; RV64IM-LABEL: muli32_p62:
+; RV64IM:       # %bb.0:
+; RV64IM-NEXT:    li a1, 62
+; RV64IM-NEXT:    mulw a0, a0, a1
+; RV64IM-NEXT:    ret
+  %1 = mul i32 %a, 62
+  ret i32 %1
+}
+
 define i32 @muli32_p65(i32 %a) nounwind {
 ; RV32I-LABEL: muli32_p65:
 ; RV32I:       # %bb.0:
@@ -600,6 +786,8 @@ define i64 @muli64_p63(i64 %a) nounwind {
   ret i64 %1
 }
 
+
+
 define i32 @muli32_m63(i32 %a) nounwind {
 ; RV32I-LABEL: muli32_m63:
 ; RV32I:       # %bb.0:
@@ -1145,10 +1333,10 @@ define i128 @muli128_m3840(i128 %a) nounwind {
 ; RV32I-NEXT:    sltu a7, a6, a4
 ; RV32I-NEXT:    sub t0, t1, t0
 ; RV32I-NEXT:    mv t1, a7
-; RV32I-NEXT:    beq a5, a3, .LBB30_2
+; RV32I-NEXT:    beq a5, a3, .LBB36_2
 ; RV32I-NEXT:  # %bb.1:
 ; RV32I-NEXT:    sltu t1, a5, a3
-; RV32I-NEXT:  .LBB30_2:
+; RV32I-NEXT:  .LBB36_2:
 ; RV32I-NEXT:    sub a2, a2, a1
 ; RV32I-NEXT:    sltu a1, a2, t1
 ; RV32I-NEXT:    sub a1, t0, a1
@@ -1261,10 +1449,10 @@ define i128 @muli128_m63(i128 %a) nounwind {
 ; RV32I-NEXT:    slli t0, a1, 6
 ; RV32I-NEXT:    or a7, t0, a7
 ; RV32I-NEXT:    mv t0, a5
-; RV32I-NEXT:    beq a1, a7, .LBB31_2
+; RV32I-NEXT:    beq a1, a7, .LBB37_2
 ; RV32I-NEXT:  # %bb.1:
 ; RV32I-NEXT:    sltu t0, a1, a7
-; RV32I-NEXT:  .LBB31_2:
+; RV32I-NEXT:  .LBB37_2:
 ; RV32I-NEXT:    srli t1, a1, 26
 ; RV32I-NEXT:    slli t2, a6, 6
 ; RV32I-NEXT:    or t1, t2, t1

From 8885813ebb0a61014d99ac776b8118d935848cc9 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Tue, 16 Apr 2024 20:45:12 +0200
Subject: [PATCH 642/886] [libc++][chrono] Disables a test.

This tests seems problematic on different platforms. There is still a
test that ensures coverage, but in an automatic fashion. This test needs
to be investigated.
---
 .../time.zone.members/get_info.sys_time.pass.cpp             | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
index a751a2fb6347b..d27cf0bd89062 100644
--- a/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
+++ b/libcxx/test/std/time/time.zone/time.zone.timezone/time.zone.members/get_info.sys_time.pass.cpp
@@ -6,7 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-// UNSUPPORTED: c++03, c++11, c++14, c++17
+// TODO TZDB review the test based on review comments in
+// https://github.com/llvm/llvm-project/pull/85619
+
+// UNSUPPORTED: c++03, c++11, c++14, c++17, c++20, c++23, c++26
 // UNSUPPORTED: no-filesystem, no-localization, no-tzdb
 
 // XFAIL: libcpp-has-no-incomplete-tzdb

From 0665669876cd7f51f7572cff3bb97485d78f5de5 Mon Sep 17 00:00:00 2001
From: Fangrui Song <i@maskray.me>
Date: Tue, 16 Apr 2024 11:49:25 -0700
Subject: [PATCH 643/886] [Sema] Mark alias/ifunc targets used and consider
 mangled names

https://reviews.llvm.org/D54188 marked "alias" targets as used in C to
fix -Wunused false positives. This patch extends the approach to handle
mangled names to support global scope names in C++ and the
`overloadable` attribute in C.

(Note: we should skip `UsingShadowDecl`, which would trigger an
assertion failure in `ItaniumMangleContextImpl::mangleCXXName`.
See regression test added by commit 1c2afbae9af22b58190c10e3517242d01d89d612.)

In addition, we mark ifunc targets as used to fix #63957 (temporarily
used by xz; ifunc was removed by
https://github.com/tukaani-project/xz/commit/689ae2427342a2ea1206eb5ca08301baf410e7e0)

While our approach has false negatives for namespace scope names, the
majority of alias/ifunc C++ uses (global scope with no overloads) are
handled.

Note: The following function with internal linkage but C language
linkage type is mangled in Clang but not in GCC. This inconsistency
makes alias/ifunc difficult to use in C++ with portability (#88593).
```
extern "C" {
static void f0() {}
// GCC: void g0() __attribute__((alias("_ZL2f0v")));
// Clang: void g0() __attribute__((alias("f0")));
}
```

Pull Request: https://github.com/llvm/llvm-project/pull/87130
---
 clang/lib/Sema/CMakeLists.txt                 |  1 +
 clang/lib/Sema/SemaDeclAttr.cpp               | 46 ++++++++++++++-----
 clang/test/AST/ast-dump-attr-json.cpp         |  1 +
 clang/test/Sema/alias-unused-win.cpp          |  2 +-
 clang/test/Sema/alias-unused.cpp              | 16 ++++---
 .../llvm-project-overlay/clang/BUILD.bazel    |  1 +
 6 files changed, 48 insertions(+), 19 deletions(-)

diff --git a/clang/lib/Sema/CMakeLists.txt b/clang/lib/Sema/CMakeLists.txt
index ab3b813a9ccd9..a96439df66422 100644
--- a/clang/lib/Sema/CMakeLists.txt
+++ b/clang/lib/Sema/CMakeLists.txt
@@ -1,5 +1,6 @@
 set(LLVM_LINK_COMPONENTS
   Core
+  Demangle
   FrontendHLSL
   FrontendOpenMP
   MC
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index b7b1fbc625a15..c3bf18a3f79e2 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -45,6 +45,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/STLForwardCompat.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/Demangle/Demangle.h"
 #include "llvm/IR/Assumptions.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Support/Error.h"
@@ -1983,6 +1984,38 @@ static void handleWeakRefAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   D->addAttr(::new (S.Context) WeakRefAttr(S.Context, AL));
 }
 
+// Mark alias/ifunc target as used. Due to name mangling, we look up the
+// demangled name ignoring parameters (not supported by microsoftDemangle
+// https://github.com/llvm/llvm-project/issues/88825). This should handle the
+// majority of use cases while leaving namespace scope names unmarked.
+static void markUsedForAliasOrIfunc(Sema &S, Decl *D, const ParsedAttr &AL,
+                                    StringRef Str) {
+  std::unique_ptr<char, llvm::FreeDeleter> Demangled;
+  if (S.getASTContext().getCXXABIKind() != TargetCXXABI::Microsoft)
+    Demangled.reset(llvm::itaniumDemangle(Str, /*ParseParams=*/false));
+  std::unique_ptr<MangleContext> MC(S.Context.createMangleContext());
+  SmallString<256> Name;
+
+  const DeclarationNameInfo Target(
+      &S.Context.Idents.get(Demangled ? Demangled.get() : Str), AL.getLoc());
+  LookupResult LR(S, Target, Sema::LookupOrdinaryName);
+  if (S.LookupName(LR, S.TUScope)) {
+    for (NamedDecl *ND : LR) {
+      if (!isa<FunctionDecl>(ND) && !isa<VarDecl>(ND))
+        continue;
+      if (MC->shouldMangleDeclName(ND)) {
+        llvm::raw_svector_ostream Out(Name);
+        Name.clear();
+        MC->mangleName(GlobalDecl(ND), Out);
+      } else {
+        Name = ND->getIdentifier()->getName();
+      }
+      if (Name == Str)
+        ND->markUsed(S.Context);
+    }
+  }
+}
+
 static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
   StringRef Str;
   if (!S.checkStringLiteralArgumentAttr(AL, 0, Str))
@@ -1995,6 +2028,7 @@ static void handleIFuncAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     return;
   }
 
+  markUsedForAliasOrIfunc(S, D, AL, Str);
   D->addAttr(::new (S.Context) IFuncAttr(S.Context, AL, Str));
 }
 
@@ -2029,17 +2063,7 @@ static void handleAliasAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     }
   }
 
-  // Mark target used to prevent unneeded-internal-declaration warnings.
-  if (!S.LangOpts.CPlusPlus) {
-    // FIXME: demangle Str for C++, as the attribute refers to the mangled
-    // linkage name, not the pre-mangled identifier.
-    const DeclarationNameInfo target(&S.Context.Idents.get(Str), AL.getLoc());
-    LookupResult LR(S, target, Sema::LookupOrdinaryName);
-    if (S.LookupQualifiedName(LR, S.getCurLexicalContext()))
-      for (NamedDecl *ND : LR)
-        ND->markUsed(S.Context);
-  }
-
+  markUsedForAliasOrIfunc(S, D, AL, Str);
   D->addAttr(::new (S.Context) AliasAttr(S.Context, AL, Str));
 }
 
diff --git a/clang/test/AST/ast-dump-attr-json.cpp b/clang/test/AST/ast-dump-attr-json.cpp
index 051c2956abfdf..883e584bfedf0 100644
--- a/clang/test/AST/ast-dump-attr-json.cpp
+++ b/clang/test/AST/ast-dump-attr-json.cpp
@@ -46,6 +46,7 @@ __thread __attribute__ ((tls_model ("local-exec"))) int tls_model_var;
 // CHECK-NEXT:    "tokLen": 11
 // CHECK-NEXT:   }
 // CHECK-NEXT:  },
+// CHECK-NEXT:  "isUsed": true,
 // CHECK-NEXT:  "name": "global_decl",
 // CHECK-NEXT:  "mangledName": "global_decl",
 // CHECK-NEXT:  "type": {
diff --git a/clang/test/Sema/alias-unused-win.cpp b/clang/test/Sema/alias-unused-win.cpp
index 47c96d4117517..97d57a3bbd1e3 100644
--- a/clang/test/Sema/alias-unused-win.cpp
+++ b/clang/test/Sema/alias-unused-win.cpp
@@ -7,7 +7,7 @@ extern "C" {
 static int f(void) { return 42; } // cxx-warning{{unused function 'f'}}
 int g(void) __attribute__((alias("f")));
 
-static int foo [] = { 42, 0xDEAD }; // cxx-warning{{variable 'foo' is not needed and will not be emitted}}
+static int foo [] = { 42, 0xDEAD };
 extern typeof(foo) bar __attribute__((unused, alias("foo")));
 
 static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
diff --git a/clang/test/Sema/alias-unused.cpp b/clang/test/Sema/alias-unused.cpp
index dc8e46f072d74..c0b541c880e52 100644
--- a/clang/test/Sema/alias-unused.cpp
+++ b/clang/test/Sema/alias-unused.cpp
@@ -14,24 +14,26 @@ extern typeof(foo) bar __attribute__((unused, alias("foo")));
 /// We report a warning in C++ mode because the internal linkage `resolver` gets
 /// mangled as it does not have a language linkage. GCC does not mangle
 /// `resolver` or report a warning.
-static int (*resolver(void))(void) { return f; } // expected-warning{{unused function 'resolver'}}
+static int (*resolver(void))(void) { return f; } // cxx-warning{{unused function 'resolver'}}
 int ifunc(void) __attribute__((ifunc("resolver")));
 
-static int __attribute__((overloadable)) f0(int x) { return x; } // expected-warning{{unused function 'f0'}}
+static int __attribute__((overloadable)) f0(int x) { return x; }
 static float __attribute__((overloadable)) f0(float x) { return x; } // expected-warning{{unused function 'f0'}}
 int g0(void) __attribute__((alias("_ZL2f0i")));
 
 #ifdef __cplusplus
-static int f1() { return 42; } // expected-warning{{unused function 'f1'}}
+static int f1() { return 42; }
 int g1(void) __attribute__((alias("_ZL2f1v")));
 }
 
-static int f2(int) { return 42; } // expected-warning{{unused function 'f2'}}
-static int f2() { return 42; } // expected-warning{{unused function 'f2'}}
+/// We demangle alias/ifunc target and mark all found functions as used.
+
+static int f2(int) { return 42; } // cxx-warning{{unused function 'f2'}}
+static int f2() { return 42; }
 int g2() __attribute__((alias("_ZL2f2v")));
 
-static int (*resolver1())() { return f; } // expected-warning{{unused function 'resolver1'}}
-static int (*resolver1(int))() { return f; } // expected-warning{{unused function 'resolver1'}}
+static int (*resolver1())() { return f; } // cxx-warning{{unused function 'resolver1'}}
+static int (*resolver1(int))() { return f; }
 int ifunc1() __attribute__((ifunc("_ZL9resolver1i")));
 
 /// TODO: We should report "unused function" for f3(int).
diff --git a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
index c2f77e3abca0e..725ac6bb38120 100644
--- a/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/clang/BUILD.bazel
@@ -1136,6 +1136,7 @@ cc_library(
         "//llvm:AllTargetsAsmParsers",
         "//llvm:AllTargetsCodeGens",
         "//llvm:Core",
+        "//llvm:Demangle",
         "//llvm:FrontendHLSL",
         "//llvm:FrontendOpenMP",
         "//llvm:MC",

From 5422eb0b841521908c2fc60bd9c7fdc11ded12a1 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 11:50:49 -0700
Subject: [PATCH 644/886] [memprof] Add another constructor to MemProfReader
 (#88952)

This patch enables users of MemProfReader to directly supply mappings
from CallStackId to actual call stacks.

Once the users of the current constructor without CSIdMap switch to
the new constructor, we'll have fewer users of:

- IndexedAllocationInfo::CallStack
- IndexedMemProfRecord::CallSites

bringing us one step closer to the removal of these fields in favor
of:

- IndexedAllocationInfo::CSId
- IndexedMemProfRecord::CallSiteIds
---
 llvm/include/llvm/ProfileData/MemProfReader.h |  9 ++++
 llvm/unittests/ProfileData/MemProfTest.cpp    | 41 +++++++++++++++++++
 2 files changed, 50 insertions(+)

diff --git a/llvm/include/llvm/ProfileData/MemProfReader.h b/llvm/include/llvm/ProfileData/MemProfReader.h
index 1f84fefad03e3..7fa8af184dc93 100644
--- a/llvm/include/llvm/ProfileData/MemProfReader.h
+++ b/llvm/include/llvm/ProfileData/MemProfReader.h
@@ -98,6 +98,15 @@ class MemProfReader {
       llvm::DenseMap<FrameId, Frame> FrameIdMap,
       llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData);
 
+  // Initialize the MemProfReader with the frame mappings, call stack mappings,
+  // and profile contents.
+  MemProfReader(
+      llvm::DenseMap<FrameId, Frame> FrameIdMap,
+      llvm::DenseMap<CallStackId, llvm::SmallVector<FrameId>> CSIdMap,
+      llvm::MapVector<GlobalValue::GUID, IndexedMemProfRecord> ProfData)
+      : IdToFrame(std::move(FrameIdMap)), CSIdToCallStack(std::move(CSIdMap)),
+        FunctionProfileData(std::move(ProfData)) {}
+
 protected:
   // A helper method to extract the frame from the IdToFrame map.
   const Frame &idToFrame(const FrameId Id) const {
diff --git a/llvm/unittests/ProfileData/MemProfTest.cpp b/llvm/unittests/ProfileData/MemProfTest.cpp
index ab9227e9df881..f596919ed039a 100644
--- a/llvm/unittests/ProfileData/MemProfTest.cpp
+++ b/llvm/unittests/ProfileData/MemProfTest.cpp
@@ -436,6 +436,47 @@ TEST(MemProf, BaseMemProfReader) {
               FrameContains("bar", 10U, 2U, false));
 }
 
+TEST(MemProf, BaseMemProfReaderWithCSIdMap) {
+  llvm::DenseMap<FrameId, Frame> FrameIdMap;
+  Frame F1(/*Hash=*/IndexedMemProfRecord::getGUID("foo"), /*LineOffset=*/20,
+           /*Column=*/5, /*IsInlineFrame=*/true);
+  Frame F2(/*Hash=*/IndexedMemProfRecord::getGUID("bar"), /*LineOffset=*/10,
+           /*Column=*/2, /*IsInlineFrame=*/false);
+  FrameIdMap.insert({F1.hash(), F1});
+  FrameIdMap.insert({F2.hash(), F2});
+
+  llvm::DenseMap<CallStackId, llvm::SmallVector<FrameId>> CSIdMap;
+  llvm::SmallVector<FrameId> CallStack = {F1.hash(), F2.hash()};
+  CallStackId CSId = llvm::memprof::hashCallStack(CallStack);
+  CSIdMap.insert({CSId, CallStack});
+
+  llvm::MapVector<llvm::GlobalValue::GUID, IndexedMemProfRecord> ProfData;
+  IndexedMemProfRecord FakeRecord;
+  MemInfoBlock Block;
+  Block.AllocCount = 1U, Block.TotalAccessDensity = 4,
+  Block.TotalLifetime = 200001;
+  FakeRecord.AllocSites.emplace_back(
+      /*CS=*/llvm::SmallVector<FrameId>(),
+      /*CSId=*/llvm::memprof::hashCallStack(CallStack),
+      /*MB=*/Block);
+  ProfData.insert({F1.hash(), FakeRecord});
+
+  MemProfReader Reader(FrameIdMap, CSIdMap, ProfData);
+
+  llvm::SmallVector<MemProfRecord, 1> Records;
+  for (const auto &KeyRecordPair : Reader) {
+    Records.push_back(KeyRecordPair.second);
+  }
+
+  ASSERT_THAT(Records, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites[0].CallStack, SizeIs(2));
+  EXPECT_THAT(Records[0].AllocSites[0].CallStack[0],
+              FrameContains("foo", 20U, 5U, true));
+  EXPECT_THAT(Records[0].AllocSites[0].CallStack[1],
+              FrameContains("bar", 10U, 2U, false));
+}
+
 TEST(MemProf, IndexedMemProfRecordToMemProfRecord) {
   // Verify that MemProfRecord can be constructed from IndexedMemProfRecord with
   // CallStackIds only.

From c7657cf7d1ee57f9cb9133164536591a1842b43c Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 14:54:06 -0400
Subject: [PATCH 645/886] [SLP]Keep externally used GEPs as GEPs, if possible
 instead of extractelement.

If the vectorized GEP instruction can be still kept as a scalar GEP,
better to keep it as scalar instead of extractelement. In many cases it
is more profitable.

Metric: size..text

Program                                                                          size..text
                                                                                 results     results0    diff
                        test-suite :: SingleSource/Benchmarks/Misc/oourafft.test    18911.00    19695.00  4.1%
                   test-suite :: SingleSource/Benchmarks/Misc-C++-EH/spirit.test    59987.00    60707.00  1.2%
       test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1392209.00  1392753.00  0.0%
        test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1392209.00  1392753.00  0.0%
           test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test  1087996.00  1088236.00  0.0%
                         test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   309310.00   309342.00  0.0%
             test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   664661.00   664693.00  0.0%
            test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   664661.00   664693.00  0.0%
        test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12354636.00 12354908.00  0.0%
                  test-suite :: External/SPEC/CFP2006/453.povray/453.povray.test  1152748.00  1152716.00 -0.0%
                       test-suite :: MultiSource/Applications/oggenc/oggenc.test   191787.00   191771.00 -0.0%
                     test-suite :: SingleSource/UnitTests/matrix-types-spec.test   480796.00   480476.00 -0.1%

Misc/oourafft - Extra code gets vectorized
Misc-C++-EH/spirit - same
CFP2017speed/638.imagick_s
CFP2017rate/538.imagick_r - same, extra code gets vectorized
CINT2006/400.perlbench - some extra 4 x ptr stores vectorized
Bullet/bullet - extra 4 x ptr store vectorized
CINT2017rate/525.x264_r
CINT2017speed/625.x264_s - same
CFP2017rate/526.blender_r - extra 8 x float stores (several), some extra
4 x ptr stores
CFP2006/453.povray - 2 x double loads/stores replaced by 4 x double
loads/stores
Applications/oggenc - extra code is vectorized
UnitTests/matrix-types-spec - extra code gets vectorized

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/88877
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 50 ++++++++++++++++++-
 .../SLPVectorizer/X86/extract_in_tree_user.ll |  4 +-
 .../SLPVectorizer/X86/geps-non-pow-2.ll       | 17 ++++---
 .../SLPVectorizer/X86/opaque-ptr.ll           | 19 +++----
 .../X86/reorder-reused-masked-gather2.ll      |  2 +-
 .../SLPVectorizer/X86/stacksave-dependence.ll |  4 +-
 6 files changed, 71 insertions(+), 25 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 0cd3ca32933ca..7694627c3b043 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -1134,6 +1134,7 @@ class BoUpSLP {
     MustGather.clear();
     EntryToLastInstruction.clear();
     ExternalUses.clear();
+    ExternalUsesAsGEPs.clear();
     for (auto &Iter : BlocksSchedules) {
       BlockScheduling *BS = Iter.second.get();
       BS->clear();
@@ -3154,6 +3155,10 @@ class BoUpSLP {
   /// after vectorization.
   UserList ExternalUses;
 
+  /// A list of GEPs which can be reaplced by scalar GEPs instead of
+  /// extractelement instructions.
+  SmallPtrSet<Value *, 4> ExternalUsesAsGEPs;
+
   /// Values used only by @llvm.assume calls.
   SmallPtrSet<const Value *, 32> EphValues;
 
@@ -5541,6 +5546,7 @@ void BoUpSLP::buildExternalUses(
                           << FoundLane << " from " << *Scalar << ".\n");
         ScalarToExtUses.try_emplace(Scalar, ExternalUses.size());
         ExternalUses.emplace_back(Scalar, nullptr, FoundLane);
+        continue;
       }
       for (User *U : Scalar->users()) {
         LLVM_DEBUG(dbgs() << "SLP: Checking user:" << *U << ".\n");
@@ -9925,6 +9931,7 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef<Value *> VectorizedVals) {
   SmallVector<APInt> DemandedElts;
   SmallDenseSet<Value *, 4> UsedInserts;
   DenseSet<std::pair<const TreeEntry *, Type *>> VectorCasts;
+  std::optional<DenseMap<Value *, unsigned>> ValueToExtUses;
   for (ExternalUser &EU : ExternalUses) {
     // We only add extract cost once for the same scalar.
     if (!isa_and_nonnull<InsertElementInst>(EU.User) &&
@@ -10033,12 +10040,40 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef<Value *> VectorizedVals) {
         }
       }
     }
+    // Leave the GEPs as is, they are free in most cases and better to keep them
+    // as GEPs.
+    TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;
+    if (auto *GEP = dyn_cast<GetElementPtrInst>(EU.Scalar)) {
+      if (!ValueToExtUses) {
+        ValueToExtUses.emplace();
+        for_each(enumerate(ExternalUses), [&](const auto &P) {
+          ValueToExtUses->try_emplace(P.value().Scalar, P.index());
+        });
+      }
+      // Can use original GEP, if no operands vectorized or they are marked as
+      // externally used already.
+      bool CanBeUsedAsGEP = all_of(GEP->operands(), [&](Value *V) {
+        if (!getTreeEntry(V))
+          return true;
+        auto It = ValueToExtUses->find(V);
+        if (It != ValueToExtUses->end()) {
+          // Replace all uses to avoid compiler crash.
+          ExternalUses[It->second].User = nullptr;
+          return true;
+        }
+        return false;
+      });
+      if (CanBeUsedAsGEP) {
+        ExtractCost += TTI->getInstructionCost(GEP, CostKind);
+        ExternalUsesAsGEPs.insert(EU.Scalar);
+        continue;
+      }
+    }
 
     // If we plan to rewrite the tree in a smaller type, we will need to sign
     // extend the extracted value back to the original type. Here, we account
     // for the extract and the added cost of the sign extend if needed.
     auto *VecTy = FixedVectorType::get(EU.Scalar->getType(), BundleWidth);
-    TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput;
     auto It = MinBWs.find(getTreeEntry(EU.Scalar));
     if (It != MinBWs.end()) {
       auto *MinTy = IntegerType::get(F->getContext(), It->second.first);
@@ -13161,6 +13196,8 @@ Value *BoUpSLP::vectorizeTree(
       if (Scalar->getType() != Vec->getType()) {
         Value *Ex = nullptr;
         Value *ExV = nullptr;
+        auto *GEP = dyn_cast<GetElementPtrInst>(Scalar);
+        bool ReplaceGEP = GEP && ExternalUsesAsGEPs.contains(GEP);
         auto It = ScalarToEEs.find(Scalar);
         if (It != ScalarToEEs.end()) {
           // No need to emit many extracts, just move the only one in the
@@ -13186,6 +13223,15 @@ Value *BoUpSLP::vectorizeTree(
             if (const TreeEntry *ETE = getTreeEntry(V))
               V = ETE->VectorizedValue;
             Ex = Builder.CreateExtractElement(V, ES->getIndexOperand());
+          } else if (ReplaceGEP) {
+            // Leave the GEPs as is, they are free in most cases and better to
+            // keep them as GEPs.
+            auto *CloneGEP = GEP->clone();
+            CloneGEP->insertBefore(*Builder.GetInsertBlock(),
+                                   Builder.GetInsertPoint());
+            if (GEP->hasName())
+              CloneGEP->takeName(GEP);
+            Ex = CloneGEP;
           } else {
             Ex = Builder.CreateExtractElement(Vec, Lane);
           }
@@ -13224,6 +13270,8 @@ Value *BoUpSLP::vectorizeTree(
       assert((ExternallyUsedValues.count(Scalar) ||
               any_of(Scalar->users(),
                      [&](llvm::User *U) {
+                       if (ExternalUsesAsGEPs.contains(U))
+                         return true;
                        TreeEntry *UseEntry = getTreeEntry(U);
                        return UseEntry &&
                               (UseEntry->State == TreeEntry::Vectorize ||
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll b/llvm/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll
index 096f57d100a50..c600d75ed1e8c 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll
@@ -13,7 +13,7 @@ define i32 @fn1() {
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x ptr> [[TMP1]], <2 x ptr> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i64, <2 x ptr> [[TMP2]], <2 x i64> <i64 11, i64 56>
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x ptr> [[TMP3]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i64 11
 ; CHECK-NEXT:    [[TMP5:%.*]] = ptrtoint <2 x ptr> [[TMP3]] to <2 x i64>
 ; CHECK-NEXT:    store <2 x i64> [[TMP5]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    ret i32 undef
@@ -92,7 +92,7 @@ define void @externally_used_ptrs() {
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x ptr> [[TMP1]], <2 x ptr> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i64, <2 x ptr> [[TMP2]], <2 x i64> <i64 56, i64 11>
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x ptr> [[TMP3]], i32 1
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i64 11
 ; CHECK-NEXT:    [[TMP5:%.*]] = ptrtoint <2 x ptr> [[TMP3]] to <2 x i64>
 ; CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <2 x i64> [[TMP5]], [[TMP6]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/geps-non-pow-2.ll b/llvm/test/Transforms/SLPVectorizer/X86/geps-non-pow-2.ll
index aa67974358306..e459cd8c6955b 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/geps-non-pow-2.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/geps-non-pow-2.ll
@@ -13,25 +13,26 @@ define dso_local i32 @g() local_unnamed_addr {
 ; CHECK:       while.body:
 ; CHECK-NEXT:    [[C_022:%.*]] = phi ptr [ [[C_022_BE:%.*]], [[WHILE_BODY_BACKEDGE:%.*]] ], [ undef, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x ptr> [ [[TMP14:%.*]], [[WHILE_BODY_BACKEDGE]] ], [ undef, [[ENTRY]] ]
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[C_022]], i64 1
-; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[C_022]] to i64
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TMP9]] to i64
 ; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP2]] to i32
+; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, ptr [[C_022]], i64 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, <2 x ptr> [[TMP1]], <2 x i64> <i64 1, i64 1>
 ; CHECK-NEXT:    switch i32 [[TMP3]], label [[WHILE_BODY_BACKEDGE]] [
-; CHECK-NEXT:    i32 2, label [[SW_BB:%.*]]
-; CHECK-NEXT:    i32 4, label [[SW_BB6:%.*]]
+; CHECK-NEXT:      i32 2, label [[SW_BB:%.*]]
+; CHECK-NEXT:      i32 4, label [[SW_BB6:%.*]]
 ; CHECK-NEXT:    ]
 ; CHECK:       sw.bb:
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x ptr> [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = ptrtoint ptr [[TMP5]] to i64
 ; CHECK-NEXT:    [[TMP7:%.*]] = trunc i64 [[TMP6]] to i32
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, <2 x ptr> [[TMP1]], <2 x i64> <i64 2, i64 2>
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x ptr> [[TMP4]], i32 1
-; CHECK-NEXT:    store i32 [[TMP7]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, ptr [[C_022]], i64 2
+; CHECK-NEXT:    store i32 [[TMP7]], ptr [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, <2 x ptr> [[TMP1]], <2 x i64> <i64 2, i64 2>
 ; CHECK-NEXT:    br label [[WHILE_BODY_BACKEDGE]]
 ; CHECK:       sw.bb6:
 ; CHECK-NEXT:    [[INCDEC_PTR8:%.*]] = getelementptr inbounds i32, ptr [[C_022]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = ptrtoint ptr [[INCDEC_PTR]] to i64
 ; CHECK-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, <2 x ptr> [[TMP1]], <2 x i64> <i64 2, i64 2>
@@ -39,7 +40,7 @@ define dso_local i32 @g() local_unnamed_addr {
 ; CHECK-NEXT:    store i32 [[TMP11]], ptr [[TMP13]], align 4
 ; CHECK-NEXT:    br label [[WHILE_BODY_BACKEDGE]]
 ; CHECK:       while.body.backedge:
-; CHECK-NEXT:    [[C_022_BE]] = phi ptr [ [[INCDEC_PTR]], [[WHILE_BODY]] ], [ [[INCDEC_PTR8]], [[SW_BB6]] ], [ [[INCDEC_PTR5]], [[SW_BB]] ]
+; CHECK-NEXT:    [[C_022_BE]] = phi ptr [ [[INCDEC_PTR1]], [[WHILE_BODY]] ], [ [[INCDEC_PTR8]], [[SW_BB6]] ], [ [[INCDEC_PTR5]], [[SW_BB]] ]
 ; CHECK-NEXT:    [[TMP14]] = phi <2 x ptr> [ [[TMP4]], [[WHILE_BODY]] ], [ [[TMP12]], [[SW_BB6]] ], [ [[TMP8]], [[SW_BB]] ]
 ; CHECK-NEXT:    br label [[WHILE_BODY]]
 ; CHECK:       while.end:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/opaque-ptr.ll b/llvm/test/Transforms/SLPVectorizer/X86/opaque-ptr.ll
index 3801fa5c787b6..c40be9690cce1 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/opaque-ptr.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/opaque-ptr.ll
@@ -52,17 +52,14 @@ define void @test(ptr %r, ptr %p, ptr %q) #0 {
 
 define void @test2(ptr %a, ptr %b) {
 ; CHECK-LABEL: @test2(
-; CHECK-NEXT:    [[A1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 1
-; CHECK-NEXT:    [[A2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 2
-; CHECK-NEXT:    [[I1:%.*]] = ptrtoint ptr [[A1]] to i64
-; CHECK-NEXT:    [[B3:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 3
-; CHECK-NEXT:    [[I2:%.*]] = ptrtoint ptr [[B3]] to i64
-; CHECK-NEXT:    [[V1:%.*]] = load i64, ptr [[A1]], align 8
-; CHECK-NEXT:    [[V2:%.*]] = load i64, ptr [[A2]], align 8
-; CHECK-NEXT:    [[ADD1:%.*]] = add i64 [[I1]], [[V1]]
-; CHECK-NEXT:    [[ADD2:%.*]] = add i64 [[I2]], [[V2]]
-; CHECK-NEXT:    store i64 [[ADD1]], ptr [[A1]], align 8
-; CHECK-NEXT:    store i64 [[ADD2]], ptr [[A2]], align 8
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[A:%.*]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x ptr> [[TMP1]], ptr [[B:%.*]], i32 1
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i64, <2 x ptr> [[TMP2]], <2 x i64> <i64 1, i64 3>
+; CHECK-NEXT:    [[A1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 1
+; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint <2 x ptr> [[TMP3]] to <2 x i64>
+; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, ptr [[A1]], align 8
+; CHECK-NEXT:    [[TMP6:%.*]] = add <2 x i64> [[TMP4]], [[TMP5]]
+; CHECK-NEXT:    store <2 x i64> [[TMP6]], ptr [[A1]], align 8
 ; CHECK-NEXT:    ret void
 ;
   %a1 = getelementptr inbounds i64, ptr %a, i64 1
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/reorder-reused-masked-gather2.ll b/llvm/test/Transforms/SLPVectorizer/X86/reorder-reused-masked-gather2.ll
index ddc2a1b819041..30f328293cdaa 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/reorder-reused-masked-gather2.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/reorder-reused-masked-gather2.ll
@@ -9,7 +9,7 @@ define void @"foo"(ptr addrspace(1) %0, ptr addrspace(1) %1) #0 {
 ; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x ptr addrspace(1)> poison, ptr addrspace(1) [[TMP0:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x ptr addrspace(1)> [[TMP3]], <4 x ptr addrspace(1)> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, <4 x ptr addrspace(1)> [[TMP4]], <4 x i64> <i64 8, i64 12, i64 28, i64 24>
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x ptr addrspace(1)> [[TMP5]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP0]], i64 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr addrspace(1) [[TMP1:%.*]], i64 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p1(<4 x ptr addrspace(1)> [[TMP5]], i32 4, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, <4 x float> poison)
 ; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <4 x float> [[TMP8]], <4 x float> poison, <8 x i32> <i32 0, i32 3, i32 0, i32 3, i32 2, i32 1, i32 2, i32 1>
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/stacksave-dependence.ll b/llvm/test/Transforms/SLPVectorizer/X86/stacksave-dependence.ll
index 0125e5fab089b..e93c5244dfbe2 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/stacksave-dependence.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/stacksave-dependence.ll
@@ -35,7 +35,7 @@ define void @allocas(ptr %a, ptr %b, ptr %c) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[V1]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x ptr> [[TMP1]], ptr [[V2]], i32 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, <2 x ptr> [[TMP2]], <2 x i32> <i32 1, i32 1>
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x ptr> [[TMP3]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[V1]], i32 1
 ; CHECK-NEXT:    store ptr [[TMP4]], ptr [[A:%.*]], align 8
 ; CHECK-NEXT:    store <2 x ptr> [[TMP3]], ptr [[B:%.*]], align 8
 ; CHECK-NEXT:    ret void
@@ -127,7 +127,7 @@ define void @stacksave2(ptr %a, ptr %b, ptr %c) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[V1]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x ptr> [[TMP1]], ptr [[V2]], i32 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, <2 x ptr> [[TMP2]], <2 x i32> <i32 1, i32 1>
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x ptr> [[TMP3]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[V1]], i32 1
 ; CHECK-NEXT:    store ptr [[TMP4]], ptr [[A:%.*]], align 8
 ; CHECK-NEXT:    call void @use(ptr inalloca(i8) [[V2]]) #[[ATTR5:[0-9]+]]
 ; CHECK-NEXT:    call void @llvm.stackrestore.p0(ptr [[STACK]])

From 7d4e8c1f3bbfe976f4871c9cf953f76d771b0eda Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 14:55:41 -0400
Subject: [PATCH 646/886] [SLP]Attempt to vectorize long stores, if short one
 failed.

We can try to vectorize long store sequences, if short ones were
unsuccessful because of the non-profitable vectorization. It should not
increase compile time significantly (stores are sorted already,
complexity is n x log n), but vectorize extra code.

Metric: size..text

Program                                                                         size..text
                                                                                results     results0    diff
         test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test  1088012.00  1088236.00  0.0%
                  test-suite :: SingleSource/UnitTests/matrix-types-spec.test   480396.00   480476.00  0.0%
          test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   664613.00   664661.00  0.0%
         test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   664613.00   664661.00  0.0%
        test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2041105.00  2040961.00 -0.0%
                 test-suite :: MultiSource/Applications/JM/lencod/lencod.test   836563.00   836387.00 -0.0%
                 test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test  1035100.00  1032140.00 -0.3%

In all benchmarks extra code gets vectorized

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/88563
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 81 ++++++++++++-------
 .../Transforms/SLPVectorizer/X86/pr46983.ll   | 46 +++--------
 2 files changed, 62 insertions(+), 65 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 7694627c3b043..8ae38550d3095 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15237,39 +15237,60 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
         Size *= 2;
       });
       unsigned StartIdx = 0;
-      for (unsigned Size : CandidateVFs) {
-        for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
-          ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
-          assert(
-              all_of(
-                  Slice,
-                  [&](Value *V) {
-                    return cast<StoreInst>(V)->getValueOperand()->getType() ==
-                           cast<StoreInst>(Slice.front())
-                               ->getValueOperand()
-                               ->getType();
-                  }) &&
-              "Expected all operands of same type.");
-          if (!VectorizedStores.count(Slice.front()) &&
-              !VectorizedStores.count(Slice.back()) &&
-              TriedSequences.insert(std::make_pair(Slice.front(), Slice.back()))
-                  .second &&
-              vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
-            // Mark the vectorized stores so that we don't vectorize them again.
-            VectorizedStores.insert(Slice.begin(), Slice.end());
-            Changed = true;
-            // If we vectorized initial block, no need to try to vectorize it
-            // again.
-            if (Cnt == StartIdx)
-              StartIdx += Size;
-            Cnt += Size;
-            continue;
+      unsigned Repeat = 0;
+      constexpr unsigned MaxAttempts = 2;
+      while (true) {
+        ++Repeat;
+        for (unsigned Size : CandidateVFs) {
+          for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
+            ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
+            assert(
+                all_of(
+                    Slice,
+                    [&](Value *V) {
+                      return cast<StoreInst>(V)->getValueOperand()->getType() ==
+                             cast<StoreInst>(Slice.front())
+                                 ->getValueOperand()
+                                 ->getType();
+                    }) &&
+                "Expected all operands of same type.");
+            if (!VectorizedStores.count(Slice.front()) &&
+                !VectorizedStores.count(Slice.back()) &&
+                TriedSequences
+                    .insert(std::make_pair(Slice.front(), Slice.back()))
+                    .second &&
+                vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
+              // Mark the vectorized stores so that we don't vectorize them
+              // again.
+              VectorizedStores.insert(Slice.begin(), Slice.end());
+              Changed = true;
+              // If we vectorized initial block, no need to try to vectorize
+              // it again.
+              if (Cnt == StartIdx)
+                StartIdx += Size;
+              Cnt += Size;
+              continue;
+            }
+            ++Cnt;
+          }
+          // Check if the whole array was vectorized already - exit.
+          if (StartIdx >= Operands.size()) {
+            Repeat = MaxAttempts;
+            break;
           }
-          ++Cnt;
         }
-        // Check if the whole array was vectorized already - exit.
-        if (StartIdx >= Operands.size())
+        // Check if tried all attempts or no need for the last attempts at all.
+        if (Repeat >= MaxAttempts)
           break;
+        const unsigned MaxTotalNum = bit_floor(Operands.size() - StartIdx);
+        if (MaxVF >= MaxTotalNum)
+          break;
+        // Last attempt to vectorize max number of elements, if all previous
+        // attempts were unsuccessful because of the cost issues.
+        CandidateVFs.clear();
+        for (unsigned Size = MaxTotalNum; Size > MaxVF; Size /= 2) {
+          CandidateVFs.push_back(Size);
+        }
       }
     }
   };
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
index 75505f632a43f..3deab0975ce76 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
@@ -100,41 +100,17 @@ define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
 define void @store_i64(ptr nocapture %0, i32 %1, i32 %2) {
 ; SSE-LABEL: @store_i64(
 ; SSE-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP1:%.*]] to i64
-; SSE-NEXT:    [[TMP5:%.*]] = load i64, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
-; SSE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], [[TMP4]]
-; SSE-NEXT:    [[TMP7:%.*]] = lshr i64 [[TMP6]], 15
-; SSE-NEXT:    [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
-; SSE-NEXT:    [[TMP9:%.*]] = icmp ult i32 [[TMP8]], 255
-; SSE-NEXT:    [[TMP10:%.*]] = and i64 [[TMP7]], 4294967295
-; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i64 [[TMP10]], i64 255
-; SSE-NEXT:    store i64 [[TMP11]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 8
-; SSE-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], [[TMP4]]
-; SSE-NEXT:    [[TMP15:%.*]] = lshr i64 [[TMP14]], 15
-; SSE-NEXT:    [[TMP16:%.*]] = trunc i64 [[TMP15]] to i32
-; SSE-NEXT:    [[TMP17:%.*]] = icmp ult i32 [[TMP16]], 255
-; SSE-NEXT:    [[TMP18:%.*]] = and i64 [[TMP15]], 4294967295
-; SSE-NEXT:    [[TMP19:%.*]] = select i1 [[TMP17]], i64 [[TMP18]], i64 255
-; SSE-NEXT:    store i64 [[TMP19]], ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 16
-; SSE-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], [[TMP4]]
-; SSE-NEXT:    [[TMP23:%.*]] = lshr i64 [[TMP22]], 15
-; SSE-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
-; SSE-NEXT:    [[TMP25:%.*]] = icmp ult i32 [[TMP24]], 255
-; SSE-NEXT:    [[TMP26:%.*]] = and i64 [[TMP23]], 4294967295
-; SSE-NEXT:    [[TMP27:%.*]] = select i1 [[TMP25]], i64 [[TMP26]], i64 255
-; SSE-NEXT:    store i64 [[TMP27]], ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 24
-; SSE-NEXT:    [[TMP29:%.*]] = load i64, ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], [[TMP4]]
-; SSE-NEXT:    [[TMP31:%.*]] = lshr i64 [[TMP30]], 15
-; SSE-NEXT:    [[TMP32:%.*]] = trunc i64 [[TMP31]] to i32
-; SSE-NEXT:    [[TMP33:%.*]] = icmp ult i32 [[TMP32]], 255
-; SSE-NEXT:    [[TMP34:%.*]] = and i64 [[TMP31]], 4294967295
-; SSE-NEXT:    [[TMP35:%.*]] = select i1 [[TMP33]], i64 [[TMP34]], i64 255
-; SSE-NEXT:    store i64 [[TMP35]], ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP5:%.*]] = load <4 x i64>, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
+; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x i64> poison, i64 [[TMP4]], i64 0
+; SSE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:    [[TMP8:%.*]] = mul <4 x i64> [[TMP5]], [[TMP7]]
+; SSE-NEXT:    [[TMP9:%.*]] = lshr <4 x i64> [[TMP8]], <i64 15, i64 15, i64 15, i64 15>
+; SSE-NEXT:    [[TMP10:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
+; SSE-NEXT:    [[TMP11:%.*]] = icmp ult <4 x i32> [[TMP10]], <i32 255, i32 255, i32 255, i32 255>
+; SSE-NEXT:    [[TMP12:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
+; SSE-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>
+; SSE-NEXT:    [[TMP14:%.*]] = zext <4 x i32> [[TMP13]] to <4 x i64>
+; SSE-NEXT:    store <4 x i64> [[TMP14]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @store_i64(

From 3eff86f82cb59d7dfc88e0cc3d8df8282f24f028 Mon Sep 17 00:00:00 2001
From: LLVM GN Syncbot <llvmgnsyncbot@gmail.com>
Date: Tue, 16 Apr 2024 18:59:12 +0000
Subject: [PATCH 647/886] [gn build] Port 9ec8c961664d

---
 .../utils/gn/secondary/clang/lib/Analysis/FlowSensitive/BUILD.gn | 1 +
 1 file changed, 1 insertion(+)

diff --git a/llvm/utils/gn/secondary/clang/lib/Analysis/FlowSensitive/BUILD.gn b/llvm/utils/gn/secondary/clang/lib/Analysis/FlowSensitive/BUILD.gn
index 04f20211b3c71..22433459a7878 100644
--- a/llvm/utils/gn/secondary/clang/lib/Analysis/FlowSensitive/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/lib/Analysis/FlowSensitive/BUILD.gn
@@ -23,6 +23,7 @@ static_library("FlowSensitive") {
     target_gen_dir,
   ]
   sources = [
+    "ASTOps.cpp",
     "AdornedCFG.cpp",
     "Arena.cpp",
     "DataflowAnalysisContext.cpp",

From b0ddbfb77d15e00e08fc36f6ccd8a4fecde465d1 Mon Sep 17 00:00:00 2001
From: Justin Bogner <mail@justinbogner.com>
Date: Tue, 16 Apr 2024 12:09:32 -0700
Subject: [PATCH 648/886] [clang][SPIR-V] Set AS for the SPIR-V logical triple
 (#88939)

This was missed in #88455, causing most of the .hlsl to SPIR-V tests to
fail (such as clang\test\Driver\hlsl-lang-targets-spirv.hlsl)
---
 clang/lib/Basic/Targets/SPIR.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/lib/Basic/Targets/SPIR.h b/clang/lib/Basic/Targets/SPIR.h
index 9a4a8b501460b..44265445ff004 100644
--- a/clang/lib/Basic/Targets/SPIR.h
+++ b/clang/lib/Basic/Targets/SPIR.h
@@ -315,7 +315,7 @@ class LLVM_LIBRARY_VISIBILITY SPIRVTargetInfo : public BaseSPIRVTargetInfo {
     // SPIR-V IDs are represented with a single 32-bit word.
     SizeType = TargetInfo::UnsignedInt;
     resetDataLayout("e-i64:64-v16:16-v24:32-v32:32-v48:64-"
-                    "v96:128-v192:256-v256:256-v512:512-v1024:1024");
+                    "v96:128-v192:256-v256:256-v512:512-v1024:1024-G1");
   }
 
   void getTargetDefines(const LangOptions &Opts,

From c9731a3dccd381849bfede5e09290c0574efa248 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 12:10:05 -0700
Subject: [PATCH 649/886] [mlir] Fix a warning about an extraneous semicolon

This patch fixes:

  mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp:58:2: error: extra ';'
  outside of a function is incompatible with C++98
  [-Werror,-Wc++98-compat-extra-semi]
---
 mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
index 621986c54d492..530c50ef74f7a 100644
--- a/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
+++ b/mlir/lib/Dialect/XeGPU/IR/XeGPUOps.cpp
@@ -55,7 +55,7 @@ static int64_t getRankOf(Value val) {
   if (auto ty = llvm::dyn_cast<ShapedType>(type))
     return ty.getRank();
   return 0;
-};
+}
 
 static bool isReadHintOrNone(const CachePolicyAttr &attr) {
   if (!attr)

From 0a789ea8a829da345e46d8224d73b2ddaba6969f Mon Sep 17 00:00:00 2001
From: erichkeane <ekeane@nvidia.com>
Date: Tue, 16 Apr 2024 12:12:25 -0700
Subject: [PATCH 650/886] Fix test from #83124 and #88902

This just replaces an '#include<new>' with a declaration of array
placement new.
---
 clang/test/SemaCXX/PR41441.cpp | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/clang/test/SemaCXX/PR41441.cpp b/clang/test/SemaCXX/PR41441.cpp
index 0b012b33fce34..d0f2917e52f21 100644
--- a/clang/test/SemaCXX/PR41441.cpp
+++ b/clang/test/SemaCXX/PR41441.cpp
@@ -1,6 +1,9 @@
 // RUN: %clang --target=x86_64-pc-linux -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
 
-#include <new>
+namespace std {
+  using size_t = decltype(sizeof(int));
+};
+void* operator new[](std::size_t, void*) noexcept;
 
 // CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)
 // CHECK: call void @llvm.memset.p0.i64(ptr align 16 %x, i8 0, i64 32, i1 false)

From 9a0a28f8384b2cb534953df33bf124f01f0e0d0e Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn@outlook.com>
Date: Tue, 16 Apr 2024 14:19:12 -0500
Subject: [PATCH 651/886] [Libomptarget] Rework Record & Replay to be a plugin
 member (#88928)

Summary:
Previously, the R&R support was global state initialized by a global
constructor. This is bad because it prevents us from adequately
constraining the lifetime of the library. Additionally, we want to
minimize the amount of global state floating around.

This patch moves the R&R support into a plugin member like everything
else. This means there will be multiple copies of the R&R implementation
floating around, but this was already the case given the fact that we
currently handle everything with dynamic libraries.
---
 .../common/include/PluginInterface.h          | 11 ++++++
 .../common/src/PluginInterface.cpp            | 39 ++++++++++++-------
 2 files changed, 35 insertions(+), 15 deletions(-)

diff --git a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
index 79e8464bfda5c..7f05464f36c1f 100644
--- a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
+++ b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
@@ -45,6 +45,8 @@
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TargetParser/Triple.h"
 
+struct RecordReplayTy;
+
 namespace llvm {
 namespace omp {
 namespace target {
@@ -1031,6 +1033,12 @@ struct GenericPluginTy {
     return *RPCServer;
   }
 
+  /// Get a reference to the R&R interface for this plugin.
+  RecordReplayTy &getRecordAndReplay() const {
+    assert(RecordReplay && "R&R not initialized");
+    return *RecordReplay;
+  }
+
   /// Get the OpenMP requires flags set for this plugin.
   int64_t getRequiresFlags() const { return RequiresFlags; }
 
@@ -1220,6 +1228,9 @@ struct GenericPluginTy {
 
   /// The interface between the plugin and the GPU for host services.
   RPCServerTy *RPCServer;
+
+  /// The interface into the record-and-replay functionality.
+  RecordReplayTy *RecordReplay;
 };
 
 namespace Plugin {
diff --git a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
index b5f3c45c835fd..6df9798f12e3d 100644
--- a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
+++ b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
@@ -362,8 +362,6 @@ struct RecordReplayTy {
   }
 };
 
-static RecordReplayTy RecordReplay;
-
 // Extract the mapping of host function pointers to device function pointers
 // from the entry table. Functions marked as 'indirect' in OpenMP will have
 // offloading entries generated for them which map the host's function pointer
@@ -473,7 +471,8 @@ GenericKernelTy::getKernelLaunchEnvironment(
   // Ctor/Dtor have no arguments, replaying uses the original kernel launch
   // environment. Older versions of the compiler do not generate a kernel
   // launch environment.
-  if (isCtorOrDtor() || RecordReplay.isReplaying() ||
+  if (isCtorOrDtor() ||
+      GenericDevice.Plugin.getRecordAndReplay().isReplaying() ||
       Version < OMP_KERNEL_ARG_MIN_VERSION_WITH_DYN_PTR)
     return nullptr;
 
@@ -562,6 +561,7 @@ Error GenericKernelTy::launch(GenericDeviceTy &GenericDevice, void **ArgPtrs,
 
   // Record the kernel description after we modified the argument count and num
   // blocks/threads.
+  RecordReplayTy &RecordReplay = GenericDevice.Plugin.getRecordAndReplay();
   if (RecordReplay.isRecording()) {
     RecordReplay.saveImage(getName(), getImage());
     RecordReplay.saveKernelInput(getName(), getImage());
@@ -839,9 +839,6 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
     delete MemoryManager;
   MemoryManager = nullptr;
 
-  if (RecordReplay.isRecordingOrReplaying())
-    RecordReplay.deinit();
-
   if (RPCServer)
     if (auto Err = RPCServer->deinitDevice(*this))
       return Err;
@@ -858,6 +855,7 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
 
   return deinitImpl();
 }
+
 Expected<DeviceImageTy *>
 GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
                             const __tgt_device_image *InputTgtImage) {
@@ -892,7 +890,8 @@ GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
     return std::move(Err);
 
   // Setup the global device memory pool if needed.
-  if (!RecordReplay.isReplaying() && shouldSetupDeviceMemoryPool()) {
+  if (!Plugin.getRecordAndReplay().isReplaying() &&
+      shouldSetupDeviceMemoryPool()) {
     uint64_t HeapSize;
     auto SizeOrErr = getDeviceHeapSize(HeapSize);
     if (SizeOrErr) {
@@ -1307,8 +1306,8 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
                                             TargetAllocTy Kind) {
   void *Alloc = nullptr;
 
-  if (RecordReplay.isRecordingOrReplaying())
-    return RecordReplay.alloc(Size);
+  if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
+    return Plugin.getRecordAndReplay().alloc(Size);
 
   switch (Kind) {
   case TARGET_ALLOC_DEFAULT:
@@ -1344,7 +1343,7 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
 
 Error GenericDeviceTy::dataDelete(void *TgtPtr, TargetAllocTy Kind) {
   // Free is a noop when recording or replaying.
-  if (RecordReplay.isRecordingOrReplaying())
+  if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
     return Plugin::success();
 
   int Res;
@@ -1396,6 +1395,7 @@ Error GenericDeviceTy::launchKernel(void *EntryPtr, void **ArgPtrs,
                                     ptrdiff_t *ArgOffsets,
                                     KernelArgsTy &KernelArgs,
                                     __tgt_async_info *AsyncInfo) {
+  RecordReplayTy &RecordReplay = Plugin.getRecordAndReplay();
   AsyncInfoWrapperTy AsyncInfoWrapper(
       *this, RecordReplay.isRecordingOrReplaying() ? nullptr : AsyncInfo);
 
@@ -1495,6 +1495,9 @@ Error GenericPluginTy::init() {
   RPCServer = new RPCServerTy(*this);
   assert(RPCServer && "Invalid RPC server");
 
+  RecordReplay = new RecordReplayTy();
+  assert(RecordReplay && "Invalid Record and Replay handler");
+
   return Plugin::success();
 }
 
@@ -1508,6 +1511,9 @@ Error GenericPluginTy::deinit() {
     assert(!Devices[DeviceId] && "Device was not deinitialized");
   }
 
+  if (RecordReplay && RecordReplay->isRecordingOrReplaying())
+    RecordReplay->deinit();
+
   // There is no global handler if no device is available.
   if (GlobalHandler)
     delete GlobalHandler;
@@ -1515,6 +1521,9 @@ Error GenericPluginTy::deinit() {
   if (RPCServer)
     delete RPCServer;
 
+  if (RecordReplay)
+    delete RecordReplay;
+
   // Perform last deinitializations on the plugin.
   return deinitImpl();
 }
@@ -1630,12 +1639,12 @@ int32_t GenericPluginTy::initialize_record_replay(int32_t DeviceId,
       isRecord ? RecordReplayTy::RRStatusTy::RRRecording
                : RecordReplayTy::RRStatusTy::RRReplaying;
 
-  if (auto Err = RecordReplay.init(&Device, MemorySize, VAddr, Status,
-                                   SaveOutput, ReqPtrArgOffset)) {
+  if (auto Err = RecordReplay->init(&Device, MemorySize, VAddr, Status,
+                                    SaveOutput, ReqPtrArgOffset)) {
     REPORT("WARNING RR did not intialize RR-properly with %lu bytes"
            "(Error: %s)\n",
            MemorySize, toString(std::move(Err)).data());
-    RecordReplay.setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
+    RecordReplay->setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
 
     if (!isRecord) {
       return OFFLOAD_FAIL;
@@ -1984,8 +1993,8 @@ int32_t GenericPluginTy::get_global(__tgt_device_binary Binary, uint64_t Size,
   assert(DevicePtr && "Invalid device global's address");
 
   // Save the loaded globals if we are recording.
-  if (RecordReplay.isRecording())
-    RecordReplay.addEntry(Name, Size, *DevicePtr);
+  if (getRecordAndReplay().isRecording())
+    getRecordAndReplay().addEntry(Name, Size, *DevicePtr);
 
   return OFFLOAD_SUCCESS;
 }

From ed7038ef334eaccdd4104053005cab52804fbfad Mon Sep 17 00:00:00 2001
From: Jeremy Kun <jkun@google.com>
Date: Tue, 16 Apr 2024 12:24:19 -0700
Subject: [PATCH 652/886] specify dialect in polynomial docs (#88933)

I figured out how to test this with `make mlir-doc doxygen-mlir`
---
 mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
index dd0384d8b79d6..79e739953d7cf 100644
--- a/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/Polynomial/IR/CMakeLists.txt
@@ -1,5 +1,5 @@
 add_mlir_dialect(Polynomial polynomial)
-add_mlir_doc(Polynomial PolynomialDialect Polynomial/ -gen-dialect-doc)
+add_mlir_doc(Polynomial PolynomialDialect Polynomial/ -gen-dialect-doc -dialect=polynomial)
 add_mlir_doc(Polynomial PolynomialOps Polynomial/ -gen-op-doc)
 add_mlir_doc(Polynomial PolynomialAttributes Dialects/ -gen-attrdef-doc)
 add_mlir_doc(Polynomial PolynomialTypes Dialects/ -gen-typedef-doc)

From 13ea36db166b7007f8b1e84e0827faaf24eb448e Mon Sep 17 00:00:00 2001
From: "juan.vazquez" <juan.vazquez.test@gmail.com>
Date: Tue, 16 Apr 2024 21:42:59 +0200
Subject: [PATCH 653/886] Fix
 UPCAddressofArraySubscriptGadget::getClaimedVarUseSites() (#88406)

UPCAddressofArraySubscriptGadget::getClaimedVarUseSites should skip
parentheses when accessing the DeclRefExpr, otherwise a crash happens
with parenthesized references.
---
 clang/lib/Analysis/UnsafeBufferUsage.cpp             |  2 +-
 .../warn-unsafe-buffer-usage-suggestions-crashes.cpp | 12 ++++++++++++
 2 files changed, 13 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/SemaCXX/warn-unsafe-buffer-usage-suggestions-crashes.cpp

diff --git a/clang/lib/Analysis/UnsafeBufferUsage.cpp b/clang/lib/Analysis/UnsafeBufferUsage.cpp
index e03fe1b683004..c42e70d5b95ac 100644
--- a/clang/lib/Analysis/UnsafeBufferUsage.cpp
+++ b/clang/lib/Analysis/UnsafeBufferUsage.cpp
@@ -1114,7 +1114,7 @@ class UPCAddressofArraySubscriptGadget : public FixableGadget {
   virtual DeclUseList getClaimedVarUseSites() const override {
     const auto *ArraySubst = cast<ArraySubscriptExpr>(Node->getSubExpr());
     const auto *DRE =
-        cast<DeclRefExpr>(ArraySubst->getBase()->IgnoreImpCasts());
+        cast<DeclRefExpr>(ArraySubst->getBase()->IgnoreParenImpCasts());
     return {DRE};
   }
 };
diff --git a/clang/test/SemaCXX/warn-unsafe-buffer-usage-suggestions-crashes.cpp b/clang/test/SemaCXX/warn-unsafe-buffer-usage-suggestions-crashes.cpp
new file mode 100644
index 0000000000000..bf4faec184ee1
--- /dev/null
+++ b/clang/test/SemaCXX/warn-unsafe-buffer-usage-suggestions-crashes.cpp
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -std=c++20 -Wunsafe-buffer-usage \
+// RUN:            -fsafe-buffer-usage-suggestions \
+// RUN:            %s -verify %s
+
+char * unsafe_pointer; // expected-warning{{'unsafe_pointer' is an unsafe pointer used for buffer access}}
+
+void test(char * param) {
+}
+
+void dre_parenthesized() {
+  test(&(unsafe_pointer)[1]); // no-crash // expected-note{{used in buffer access here}}
+}

From f430e374462efd94d891fcf9fa09d606343c780f Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 12:47:02 -0700
Subject: [PATCH 654/886] [llvm] Drop unaligned from calls to readNext (NFC)
 (#88841)

Now readNext defaults to unaligned accesses.  This patch drops
unaligned to improve readability.
---
 .../llvm/ProfileData/InstrProfReader.h        |  4 +-
 llvm/include/llvm/ProfileData/MemProf.h       | 25 ++++-----
 llvm/lib/MC/MCPseudoProbe.cpp                 |  2 +-
 llvm/lib/ProfileData/InstrProf.cpp            |  4 +-
 llvm/lib/ProfileData/InstrProfReader.cpp      | 53 +++++++------------
 llvm/lib/ProfileData/MemProf.cpp              | 25 +++++----
 llvm/lib/ProfileData/MemProfReader.cpp        | 14 ++---
 llvm/lib/ProfileData/SampleProfReader.cpp     |  2 +-
 8 files changed, 55 insertions(+), 74 deletions(-)

diff --git a/llvm/include/llvm/ProfileData/InstrProfReader.h b/llvm/include/llvm/ProfileData/InstrProfReader.h
index e46570af3873f..f662febb9216b 100644
--- a/llvm/include/llvm/ProfileData/InstrProfReader.h
+++ b/llvm/include/llvm/ProfileData/InstrProfReader.h
@@ -508,9 +508,9 @@ class InstrProfLookupTrait {
     using namespace support;
 
     offset_type KeyLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     offset_type DataLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     return std::make_pair(KeyLen, DataLen);
   }
 
diff --git a/llvm/include/llvm/ProfileData/MemProf.h b/llvm/include/llvm/ProfileData/MemProf.h
index 3520034fb1c94..d43fb1c93bb8e 100644
--- a/llvm/include/llvm/ProfileData/MemProf.h
+++ b/llvm/include/llvm/ProfileData/MemProf.h
@@ -68,7 +68,7 @@ struct PortableMemInfoBlock {
       switch (Id) {
 #define MIBEntryDef(NameTag, Name, Type)                                       \
   case Meta::Name: {                                                           \
-    Name = endian::readNext<Type, llvm::endianness::little, unaligned>(Ptr);   \
+    Name = endian::readNext<Type, llvm::endianness::little>(Ptr);              \
   } break;
 #include "llvm/ProfileData/MIBEntryDef.inc"
 #undef MIBEntryDef
@@ -223,13 +223,12 @@ struct Frame {
     using namespace support;
 
     const uint64_t F =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     const uint32_t L =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint32_t, llvm::endianness::little>(Ptr);
     const uint32_t C =
-        endian::readNext<uint32_t, llvm::endianness::little, unaligned>(Ptr);
-    const bool I =
-        endian::readNext<bool, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint32_t, llvm::endianness::little>(Ptr);
+    const bool I = endian::readNext<bool, llvm::endianness::little>(Ptr);
     return Frame(/*Function=*/F, /*LineOffset=*/L, /*Column=*/C,
                  /*IsInlineFrame=*/I);
   }
@@ -482,16 +481,15 @@ class RecordLookupTrait {
     using namespace support;
 
     offset_type KeyLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     offset_type DataLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     return std::make_pair(KeyLen, DataLen);
   }
 
   uint64_t ReadKey(const unsigned char *D, offset_type /*Unused*/) {
     using namespace support;
-    return endian::readNext<external_key_type, llvm::endianness::little,
-                            unaligned>(D);
+    return endian::readNext<external_key_type, llvm::endianness::little>(D);
   }
 
   data_type ReadData(uint64_t K, const unsigned char *D,
@@ -623,16 +621,15 @@ class FrameLookupTrait {
     using namespace support;
 
     offset_type KeyLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     offset_type DataLen =
-        endian::readNext<offset_type, llvm::endianness::little, unaligned>(D);
+        endian::readNext<offset_type, llvm::endianness::little>(D);
     return std::make_pair(KeyLen, DataLen);
   }
 
   uint64_t ReadKey(const unsigned char *D, offset_type /*Unused*/) {
     using namespace support;
-    return endian::readNext<external_key_type, llvm::endianness::little,
-                            unaligned>(D);
+    return endian::readNext<external_key_type, llvm::endianness::little>(D);
   }
 
   data_type ReadData(uint64_t K, const unsigned char *D,
diff --git a/llvm/lib/MC/MCPseudoProbe.cpp b/llvm/lib/MC/MCPseudoProbe.cpp
index eb3894dbb3c25..cec50322bb9f9 100644
--- a/llvm/lib/MC/MCPseudoProbe.cpp
+++ b/llvm/lib/MC/MCPseudoProbe.cpp
@@ -343,7 +343,7 @@ template <typename T> ErrorOr<T> MCPseudoProbeDecoder::readUnencodedNumber() {
   if (Data + sizeof(T) > End) {
     return std::error_code();
   }
-  T Val = endian::readNext<T, llvm::endianness::little, unaligned>(Data);
+  T Val = endian::readNext<T, llvm::endianness::little>(Data);
   return ErrorOr<T>(Val);
 }
 
diff --git a/llvm/lib/ProfileData/InstrProf.cpp b/llvm/lib/ProfileData/InstrProf.cpp
index a5abf63b010f7..f9ba80bd99c85 100644
--- a/llvm/lib/ProfileData/InstrProf.cpp
+++ b/llvm/lib/ProfileData/InstrProf.cpp
@@ -1135,9 +1135,9 @@ static T swapToHostOrder(const unsigned char *&D, llvm::endianness Orig) {
   using namespace support;
 
   if (Orig == llvm::endianness::little)
-    return endian::readNext<T, llvm::endianness::little, unaligned>(D);
+    return endian::readNext<T, llvm::endianness::little>(D);
   else
-    return endian::readNext<T, llvm::endianness::big, unaligned>(D);
+    return endian::readNext<T, llvm::endianness::big>(D);
 }
 
 static std::unique_ptr<ValueProfData> allocValueProfData(uint32_t TotalSize) {
diff --git a/llvm/lib/ProfileData/InstrProfReader.cpp b/llvm/lib/ProfileData/InstrProfReader.cpp
index a35366a106a32..8574a96a1b06f 100644
--- a/llvm/lib/ProfileData/InstrProfReader.cpp
+++ b/llvm/lib/ProfileData/InstrProfReader.cpp
@@ -115,10 +115,9 @@ readBinaryIdsInternal(const MemoryBuffer &DataBuffer,
 
     uint64_t BILen = 0;
     if (Endian == llvm::endianness::little)
-      BILen =
-          endian::readNext<uint64_t, llvm::endianness::little, unaligned>(BI);
+      BILen = endian::readNext<uint64_t, llvm::endianness::little>(BI);
     else
-      BILen = endian::readNext<uint64_t, llvm::endianness::big, unaligned>(BI);
+      BILen = endian::readNext<uint64_t, llvm::endianness::big>(BI);
 
     if (BILen == 0)
       return make_error<InstrProfError>(instrprof_error::malformed,
@@ -923,8 +922,7 @@ data_type InstrProfLookupTrait::ReadData(StringRef K, const unsigned char *D,
     // Read hash.
     if (D + sizeof(uint64_t) >= End)
       return data_type();
-    uint64_t Hash =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(D);
+    uint64_t Hash = endian::readNext<uint64_t, llvm::endianness::little>(D);
 
     // Initialize number of counters for GET_VERSION(FormatVersion) == 1.
     uint64_t CountsSize = N / sizeof(uint64_t) - 1;
@@ -932,8 +930,7 @@ data_type InstrProfLookupTrait::ReadData(StringRef K, const unsigned char *D,
     if (GET_VERSION(FormatVersion) != IndexedInstrProf::ProfVersion::Version1) {
       if (D + sizeof(uint64_t) > End)
         return data_type();
-      CountsSize =
-          endian::readNext<uint64_t, llvm::endianness::little, unaligned>(D);
+      CountsSize = endian::readNext<uint64_t, llvm::endianness::little>(D);
     }
     // Read counter values.
     if (D + CountsSize * sizeof(uint64_t) > End)
@@ -943,15 +940,14 @@ data_type InstrProfLookupTrait::ReadData(StringRef K, const unsigned char *D,
     CounterBuffer.reserve(CountsSize);
     for (uint64_t J = 0; J < CountsSize; ++J)
       CounterBuffer.push_back(
-          endian::readNext<uint64_t, llvm::endianness::little, unaligned>(D));
+          endian::readNext<uint64_t, llvm::endianness::little>(D));
 
     // Read bitmap bytes for GET_VERSION(FormatVersion) > 10.
     if (GET_VERSION(FormatVersion) > IndexedInstrProf::ProfVersion::Version10) {
       uint64_t BitmapBytes = 0;
       if (D + sizeof(uint64_t) > End)
         return data_type();
-      BitmapBytes =
-          endian::readNext<uint64_t, llvm::endianness::little, unaligned>(D);
+      BitmapBytes = endian::readNext<uint64_t, llvm::endianness::little>(D);
       // Read bitmap byte values.
       if (D + BitmapBytes * sizeof(uint8_t) > End)
         return data_type();
@@ -959,8 +955,7 @@ data_type InstrProfLookupTrait::ReadData(StringRef K, const unsigned char *D,
       BitmapByteBuffer.reserve(BitmapBytes);
       for (uint64_t J = 0; J < BitmapBytes; ++J)
         BitmapByteBuffer.push_back(static_cast<uint8_t>(
-            endian::readNext<uint64_t, llvm::endianness::little, unaligned>(
-                D)));
+            endian::readNext<uint64_t, llvm::endianness::little>(D)));
     }
 
     DataBuffer.emplace_back(K, Hash, std::move(CounterBuffer),
@@ -1256,8 +1251,7 @@ Error IndexedInstrProfReader::readHeader() {
     // memprof::MemProfVersion0 or the MemProf version number in
     // memprof::MemProfVersion1.
     const uint64_t FirstWord =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
 
     memprof::IndexedVersion Version = memprof::Version0;
     if (FirstWord == memprof::Version1) {
@@ -1282,17 +1276,15 @@ Error IndexedInstrProfReader::readHeader() {
     const uint64_t RecordTableOffset =
         Version == memprof::Version0
             ? FirstWord
-            : support::endian::readNext<uint64_t, llvm::endianness::little,
-                                        unaligned>(Ptr);
+            : support::endian::readNext<uint64_t, llvm::endianness::little>(
+                  Ptr);
     // The offset in the stream right before invoking
     // FrameTableGenerator.Emit.
     const uint64_t FramePayloadOffset =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     // The value returned from FrameTableGenerator.Emit.
     const uint64_t FrameTableOffset =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
 
     // Read the schema.
     auto SchemaOr = memprof::readMemProfSchema(Ptr);
@@ -1330,8 +1322,7 @@ Error IndexedInstrProfReader::readHeader() {
     const unsigned char *Ptr = Start + BinaryIdOffset;
     // Read binary ids size.
     BinaryIdsSize =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     if (BinaryIdsSize % sizeof(uint64_t))
       return error(instrprof_error::bad_header);
     // Set the binary ids start.
@@ -1348,8 +1339,7 @@ Error IndexedInstrProfReader::readHeader() {
     const unsigned char *Ptr = Start + VTableNamesOffset;
 
     CompressedVTableNamesLen =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
 
     // Writer first writes the length of compressed string, and then the actual
     // content.
@@ -1369,29 +1359,24 @@ Error IndexedInstrProfReader::readHeader() {
     if (Ptr + 2 * sizeof(uint64_t) > PtrEnd)
       return error(instrprof_error::truncated);
     const uint64_t NumTraces =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     TemporalProfTraceStreamSize =
-        support::endian::readNext<uint64_t, llvm::endianness::little,
-                                  unaligned>(Ptr);
+        support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     for (unsigned i = 0; i < NumTraces; i++) {
       // Expect at least two 64 bit fields: Weight and NumFunctions
       if (Ptr + 2 * sizeof(uint64_t) > PtrEnd)
         return error(instrprof_error::truncated);
       TemporalProfTraceTy Trace;
       Trace.Weight =
-          support::endian::readNext<uint64_t, llvm::endianness::little,
-                                    unaligned>(Ptr);
+          support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
       const uint64_t NumFunctions =
-          support::endian::readNext<uint64_t, llvm::endianness::little,
-                                    unaligned>(Ptr);
+          support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
       // Expect at least NumFunctions 64 bit fields
       if (Ptr + NumFunctions * sizeof(uint64_t) > PtrEnd)
         return error(instrprof_error::truncated);
       for (unsigned j = 0; j < NumFunctions; j++) {
         const uint64_t NameRef =
-            support::endian::readNext<uint64_t, llvm::endianness::little,
-                                      unaligned>(Ptr);
+            support::endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
         Trace.FunctionNameRefs.push_back(NameRef);
       }
       TemporalProfTraces.push_back(std::move(Trace));
diff --git a/llvm/lib/ProfileData/MemProf.cpp b/llvm/lib/ProfileData/MemProf.cpp
index 1ca0a02d3cbde..8e0402dd16e68 100644
--- a/llvm/lib/ProfileData/MemProf.cpp
+++ b/llvm/lib/ProfileData/MemProf.cpp
@@ -144,14 +144,14 @@ static IndexedMemProfRecord deserializeV0(const MemProfSchema &Schema,
 
   // Read the meminfo nodes.
   const uint64_t NumNodes =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   for (uint64_t I = 0; I < NumNodes; I++) {
     IndexedAllocationInfo Node;
     const uint64_t NumFrames =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     for (uint64_t J = 0; J < NumFrames; J++) {
       const FrameId Id =
-          endian::readNext<FrameId, llvm::endianness::little, unaligned>(Ptr);
+          endian::readNext<FrameId, llvm::endianness::little>(Ptr);
       Node.CallStack.push_back(Id);
     }
     Node.CSId = hashCallStack(Node.CallStack);
@@ -162,15 +162,15 @@ static IndexedMemProfRecord deserializeV0(const MemProfSchema &Schema,
 
   // Read the callsite information.
   const uint64_t NumCtxs =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   for (uint64_t J = 0; J < NumCtxs; J++) {
     const uint64_t NumFrames =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     llvm::SmallVector<FrameId> Frames;
     Frames.reserve(NumFrames);
     for (uint64_t K = 0; K < NumFrames; K++) {
       const FrameId Id =
-          endian::readNext<FrameId, llvm::endianness::little, unaligned>(Ptr);
+          endian::readNext<FrameId, llvm::endianness::little>(Ptr);
       Frames.push_back(Id);
     }
     Record.CallSites.push_back(Frames);
@@ -188,11 +188,10 @@ static IndexedMemProfRecord deserializeV2(const MemProfSchema &Schema,
 
   // Read the meminfo nodes.
   const uint64_t NumNodes =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   for (uint64_t I = 0; I < NumNodes; I++) {
     IndexedAllocationInfo Node;
-    Node.CSId =
-        endian::readNext<CallStackId, llvm::endianness::little, unaligned>(Ptr);
+    Node.CSId = endian::readNext<CallStackId, llvm::endianness::little>(Ptr);
     Node.Info.deserialize(Schema, Ptr);
     Ptr += PortableMemInfoBlock::serializedSize();
     Record.AllocSites.push_back(Node);
@@ -200,10 +199,10 @@ static IndexedMemProfRecord deserializeV2(const MemProfSchema &Schema,
 
   // Read the callsite information.
   const uint64_t NumCtxs =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   for (uint64_t J = 0; J < NumCtxs; J++) {
     CallStackId CSId =
-        endian::readNext<CallStackId, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<CallStackId, llvm::endianness::little>(Ptr);
     Record.CallSiteIds.push_back(CSId);
   }
 
@@ -263,7 +262,7 @@ Expected<MemProfSchema> readMemProfSchema(const unsigned char *&Buffer) {
 
   const unsigned char *Ptr = Buffer;
   const uint64_t NumSchemaIds =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   if (NumSchemaIds > static_cast<uint64_t>(Meta::Size)) {
     return make_error<InstrProfError>(instrprof_error::malformed,
                                       "memprof schema invalid");
@@ -272,7 +271,7 @@ Expected<MemProfSchema> readMemProfSchema(const unsigned char *&Buffer) {
   MemProfSchema Result;
   for (size_t I = 0; I < NumSchemaIds; I++) {
     const uint64_t Tag =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     if (Tag >= static_cast<uint64_t>(Meta::Size)) {
       return make_error<InstrProfError>(instrprof_error::malformed,
                                         "memprof schema invalid");
diff --git a/llvm/lib/ProfileData/MemProfReader.cpp b/llvm/lib/ProfileData/MemProfReader.cpp
index 91556f036c777..b4d2c6f043f6d 100644
--- a/llvm/lib/ProfileData/MemProfReader.cpp
+++ b/llvm/lib/ProfileData/MemProfReader.cpp
@@ -86,7 +86,7 @@ llvm::SmallVector<SegmentEntry> readSegmentEntries(const char *Ptr) {
   using namespace support;
 
   const uint64_t NumItemsToRead =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   llvm::SmallVector<SegmentEntry> Items;
   for (uint64_t I = 0; I < NumItemsToRead; I++) {
     Items.push_back(*reinterpret_cast<const SegmentEntry *>(
@@ -100,11 +100,11 @@ readMemInfoBlocks(const char *Ptr) {
   using namespace support;
 
   const uint64_t NumItemsToRead =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   llvm::SmallVector<std::pair<uint64_t, MemInfoBlock>> Items;
   for (uint64_t I = 0; I < NumItemsToRead; I++) {
     const uint64_t Id =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     const MemInfoBlock MIB = *reinterpret_cast<const MemInfoBlock *>(Ptr);
     Items.push_back({Id, MIB});
     // Only increment by size of MIB since readNext implicitly increments.
@@ -117,20 +117,20 @@ CallStackMap readStackInfo(const char *Ptr) {
   using namespace support;
 
   const uint64_t NumItemsToRead =
-      endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
   CallStackMap Items;
 
   for (uint64_t I = 0; I < NumItemsToRead; I++) {
     const uint64_t StackId =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
     const uint64_t NumPCs =
-        endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr);
+        endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
 
     SmallVector<uint64_t> CallStack;
     CallStack.reserve(NumPCs);
     for (uint64_t J = 0; J < NumPCs; J++) {
       CallStack.push_back(
-          endian::readNext<uint64_t, llvm::endianness::little, unaligned>(Ptr));
+          endian::readNext<uint64_t, llvm::endianness::little>(Ptr));
     }
 
     Items[StackId] = CallStack;
diff --git a/llvm/lib/ProfileData/SampleProfReader.cpp b/llvm/lib/ProfileData/SampleProfReader.cpp
index 98d0aa794529c..f91a0e6177ea0 100644
--- a/llvm/lib/ProfileData/SampleProfReader.cpp
+++ b/llvm/lib/ProfileData/SampleProfReader.cpp
@@ -503,7 +503,7 @@ ErrorOr<T> SampleProfileReaderBinary::readUnencodedNumber() {
   }
 
   using namespace support;
-  T Val = endian::readNext<T, llvm::endianness::little, unaligned>(Data);
+  T Val = endian::readNext<T, llvm::endianness::little>(Data);
   return Val;
 }
 

From 281d71604f418eb952e967d9dc4b26241b7f96aa Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Tue, 16 Apr 2024 13:48:04 -0600
Subject: [PATCH 655/886] [X86] Change how we treat functions with explicit
 sections as small/large (#88172)

Following #78348, we should treat functions with an explicit section as
small, unless the section name is (or has the prefix) ".ltext".

Clang emits global initializers into a ".text.startup" section on Linux.
If we mix small/medium code model object files with large code model
object files, we'll end up mixing sections with and without the large
section flag.

Reland of #87838 with a check for non-ELF platforms in
TargetMachine::isLargeGlobalValue(), otherwise MCJIT on Windows tests
fail.
---
 llvm/lib/Target/TargetMachine.cpp             | 26 ++++++++++++++-----
 .../X86/code-model-elf-text-sections.ll       | 23 ++++++++++++++++
 2 files changed, 43 insertions(+), 6 deletions(-)

diff --git a/llvm/lib/Target/TargetMachine.cpp b/llvm/lib/Target/TargetMachine.cpp
index a7fe329b064ee..8ddc742004292 100644
--- a/llvm/lib/Target/TargetMachine.cpp
+++ b/llvm/lib/Target/TargetMachine.cpp
@@ -43,6 +43,12 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
   if (getTargetTriple().getArch() != Triple::x86_64)
     return false;
 
+  // Remaining logic below is ELF-specific. For other object file formats where
+  // the large code model is mostly used for JIT compilation, just look at the
+  // code model.
+  if (!getTargetTriple().isOSBinFormatELF())
+    return getCodeModel() == CodeModel::Large;
+
   auto *GO = GVal->getAliaseeObject();
 
   // Be conservative if we can't find an underlying GlobalObject.
@@ -51,9 +57,20 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
 
   auto *GV = dyn_cast<GlobalVariable>(GO);
 
+  auto IsPrefix = [](StringRef Name, StringRef Prefix) {
+    return Name.consume_front(Prefix) && (Name.empty() || Name[0] == '.');
+  };
+
   // Functions/GlobalIFuncs are only large under the large code model.
-  if (!GV)
+  if (!GV) {
+    // Handle explicit sections as we do for GlobalVariables with an explicit
+    // section, see comments below.
+    if (GO->hasSection()) {
+      StringRef Name = GO->getSection();
+      return IsPrefix(Name, ".ltext");
+    }
     return getCodeModel() == CodeModel::Large;
+  }
 
   if (GV->isThreadLocal())
     return false;
@@ -73,11 +90,8 @@ bool TargetMachine::isLargeGlobalValue(const GlobalValue *GVal) const {
   // data sections. The code model attribute overrides this above.
   if (GV->hasSection()) {
     StringRef Name = GV->getSection();
-    auto IsPrefix = [&](StringRef Prefix) {
-      StringRef S = Name;
-      return S.consume_front(Prefix) && (S.empty() || S[0] == '.');
-    };
-    return IsPrefix(".lbss") || IsPrefix(".ldata") || IsPrefix(".lrodata");
+    return IsPrefix(Name, ".lbss") || IsPrefix(Name, ".ldata") ||
+           IsPrefix(Name, ".lrodata");
   }
 
   // Respect large data threshold for medium and large code models.
diff --git a/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll b/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
index 016c9a4d7b839..66a6fd3767542 100644
--- a/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
+++ b/llvm/test/CodeGen/X86/code-model-elf-text-sections.ll
@@ -13,9 +13,20 @@
 ; RUN: llvm-readelf -S %t | FileCheck %s --check-prefix=LARGE-DS
 
 ; SMALL: .text {{.*}} AX {{.*}}
+; SMALL: .ltext {{.*}} AXl {{.*}}
+; SMALL: .ltext.2 {{.*}} AXl {{.*}}
+; SMALL: .foo {{.*}} AX {{.*}}
 ; SMALL-DS: .text.func {{.*}} AX {{.*}}
+; SMALL-DS: .ltext {{.*}} AXl {{.*}}
+; SMALL-DS: .ltext.2 {{.*}} AXl {{.*}}
+; SMALL-DS: .foo {{.*}} AX {{.*}}
 ; LARGE: .ltext {{.*}} AXl {{.*}}
+; LARGE: .ltext.2 {{.*}} AXl {{.*}}
+; LARGE: .foo {{.*}} AX {{.*}}
 ; LARGE-DS: .ltext.func {{.*}} AXl {{.*}}
+; LARGE-DS: .ltext {{.*}} AXl {{.*}}
+; LARGE-DS: .ltext.2 {{.*}} AXl {{.*}}
+; LARGE-DS: .foo {{.*}} AX {{.*}}
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64--linux"
@@ -23,3 +34,15 @@ target triple = "x86_64--linux"
 define void @func() {
   ret void
 }
+
+define void @ltext() section ".ltext" {
+  ret void
+}
+
+define void @ltext2() section ".ltext.2" {
+  ret void
+}
+
+define void @foo() section ".foo" {
+  ret void
+}

From 191be2a8a8531129c779bf23c4eec86f32c69bf9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
Date: Mon, 8 Apr 2024 15:03:04 +0200
Subject: [PATCH 656/886] update_test_checks: pre-commit a new test

The test shows that name preservation doesn't work properly when
--include-generated-funcs is used.
---
 .../Inputs/stable_ir_values_funcs.ll          | 23 ++++++++++++++++++
 .../Inputs/stable_ir_values_funcs.ll.expected | 24 +++++++++++++++++++
 .../stable_ir_values_funcs.test               |  2 ++
 3 files changed, 49 insertions(+)
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/stable_ir_values_funcs.test

diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll
new file mode 100644
index 0000000000000..b4fd23a3d81ce
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll
@@ -0,0 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --include-generated-funcs --version 3
+; RUN: opt < %s -S | FileCheck %s
+
+; The assumption underlying this test is that there are pre-existing check lines
+; but something has changed, and we would like to avoid needless changes of
+; meta variable names so that diffs end up being easier to read, e.g. avoid
+; changing X_I33 into X_I34 or renumbering the various TMP variables.
+
+define i32 @func({i32, i32} %x, i32 %y) {
+  %x.i34 = extractvalue {i32, i32} %x, 0
+  %1 = add i32 %y, 1
+  %2 = add i32 %x.i34, %1
+  %3 = mul i32 %2, 3
+  ret i32 %3
+}
+
+; CHECK-LABEL: define i32 @func(
+; CHECK-SAME: { i32, i32 } [[X:%.*]], i32 [[Y:%.*]]) {
+; CHECK-NEXT:    [[X_I33:%.*]] = extractvalue { i32, i32 } [[X]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[X_I33]], [[Y]]
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i32 [[TMP1]], 3
+; CHECK-NEXT:    ret i32 [[TMP2]]
+;
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
new file mode 100644
index 0000000000000..1559319ac013a
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
@@ -0,0 +1,24 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --include-generated-funcs --version 3
+; RUN: opt < %s -S | FileCheck %s
+
+; The assumption underlying this test is that there are pre-existing check lines
+; but something has changed, and we would like to avoid needless changes of
+; meta variable names so that diffs end up being easier to read, e.g. avoid
+; changing X_I33 into X_I34 or renumbering the various TMP variables.
+
+define i32 @func({i32, i32} %x, i32 %y) {
+  %x.i34 = extractvalue {i32, i32} %x, 0
+  %1 = add i32 %y, 1
+  %2 = add i32 %x.i34, %1
+  %3 = mul i32 %2, 3
+  ret i32 %3
+}
+
+; CHECK-LABEL: define i32 @func(
+; CHECK-SAME: { i32, i32 } [[X:%.*]], i32 [[Y:%.*]]) {
+; CHECK-NEXT:    [[X_I34:%.*]] = extractvalue { i32, i32 } [[X]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[Y]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[X_I34]], [[TMP1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[TMP2]], 3
+; CHECK-NEXT:    ret i32 [[TMP3]]
+;
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/stable_ir_values_funcs.test b/llvm/test/tools/UpdateTestChecks/update_test_checks/stable_ir_values_funcs.test
new file mode 100644
index 0000000000000..5132fb9a26ff4
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/stable_ir_values_funcs.test
@@ -0,0 +1,2 @@
+# RUN: cp -f %S/Inputs/stable_ir_values_funcs.ll %t.ll && %update_test_checks %t.ll
+# RUN: diff -u %t.ll %S/Inputs/stable_ir_values_funcs.ll.expected

From e770249d955e06f205e91017cd394d8670996168 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
Date: Wed, 3 Apr 2024 18:14:12 +0200
Subject: [PATCH 657/886] update_test_checks: add new test

This test is meant to demonstrate an upcoming change that replaces basic
block labels by FileCheck patterns.
---
 .../update_test_checks/Inputs/phi-labels.ll   | 39 ++++++++++
 .../Inputs/phi-labels.ll.expected             | 71 +++++++++++++++++++
 .../update_test_checks/phi-labels.test        |  5 ++
 3 files changed, 115 insertions(+)
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/phi-labels.test

diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll
new file mode 100644
index 0000000000000..4eb05b943f506
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll
@@ -0,0 +1,39 @@
+; RUN: opt < %s -S | FileCheck %s
+
+define i32 @phi_after_label(i1 %cc) {
+entry:
+  br i1 %cc, label %then, label %end
+
+then:
+  br label %end
+
+end:
+  %r = phi i32 [ 0, %entry ], [ 1, %then ]
+  ret i32 %r
+}
+
+define void @phi_before_label(i32 %bound) {
+entry:
+  br label %loop
+
+loop:
+  %ctr = phi i32 [ 0, %entry ], [ %ctr.next, %loop ]
+  %ctr.next = add i32 %ctr, 1
+  %cc = icmp ult i32 %ctr.next, %bound
+  br i1 %cc, label %loop, label %end
+
+end:
+  ret void
+}
+
+define i32 @phi_after_label_unnamed(i1 %cc) {
+0:
+  br i1 %cc, label %1, label %2
+
+1:
+  br label %2
+
+2:
+  %r = phi i32 [ 0, %0 ], [ 1, %1 ]
+  ret i32 %r
+}
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll.expected
new file mode 100644
index 0000000000000..1d21ebe547f68
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/phi-labels.ll.expected
@@ -0,0 +1,71 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -S | FileCheck %s
+
+define i32 @phi_after_label(i1 %cc) {
+; CHECK-LABEL: define i32 @phi_after_label(
+; CHECK-SAME: i1 [[CC:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 [[CC]], label [[THEN:%.*]], label [[END:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    br label [[END]]
+; CHECK:       end:
+; CHECK-NEXT:    [[R:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ 1, [[THEN]] ]
+; CHECK-NEXT:    ret i32 [[R]]
+;
+entry:
+  br i1 %cc, label %then, label %end
+
+then:
+  br label %end
+
+end:
+  %r = phi i32 [ 0, %entry ], [ 1, %then ]
+  ret i32 %r
+}
+
+define void @phi_before_label(i32 %bound) {
+; CHECK-LABEL: define void @phi_before_label(
+; CHECK-SAME: i32 [[BOUND:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[CTR:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[CTR_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[CTR_NEXT]] = add i32 [[CTR]], 1
+; CHECK-NEXT:    [[CC:%.*]] = icmp ult i32 [[CTR_NEXT]], [[BOUND]]
+; CHECK-NEXT:    br i1 [[CC]], label [[LOOP]], label [[END:%.*]]
+; CHECK:       end:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %ctr = phi i32 [ 0, %entry ], [ %ctr.next, %loop ]
+  %ctr.next = add i32 %ctr, 1
+  %cc = icmp ult i32 %ctr.next, %bound
+  br i1 %cc, label %loop, label %end
+
+end:
+  ret void
+}
+
+define i32 @phi_after_label_unnamed(i1 %cc) {
+; CHECK-LABEL: define i32 @phi_after_label_unnamed(
+; CHECK-SAME: i1 [[CC:%.*]]) {
+; CHECK-NEXT:    br i1 [[CC]], label [[TMP1:%.*]], label [[TMP2:%.*]]
+; CHECK:       1:
+; CHECK-NEXT:    br label [[TMP2]]
+; CHECK:       2:
+; CHECK-NEXT:    [[R:%.*]] = phi i32 [ 0, [[TMP0:%.*]] ], [ 1, [[TMP1]] ]
+; CHECK-NEXT:    ret i32 [[R]]
+;
+0:
+  br i1 %cc, label %1, label %2
+
+1:
+  br label %2
+
+2:
+  %r = phi i32 [ 0, %0 ], [ 1, %1 ]
+  ret i32 %r
+}
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/phi-labels.test b/llvm/test/tools/UpdateTestChecks/update_test_checks/phi-labels.test
new file mode 100644
index 0000000000000..411c84de1dcba
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/phi-labels.test
@@ -0,0 +1,5 @@
+# RUN: cp -f %S/Inputs/phi-labels.ll %t.ll && %update_test_checks --version 4 %t.ll
+# RUN: diff -u %t.ll %S/Inputs/phi-labels.ll.expected
+## Check that running the script again does not change the result:
+# RUN: %update_test_checks %t.ll
+# RUN: diff -u %t.ll %S/Inputs/phi-labels.ll.expected

From 377a2767a9951659b5ec7309abb78da719a4f93b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
Date: Wed, 3 Apr 2024 18:19:59 +0200
Subject: [PATCH 658/886] update_test_checks: remove an unused function

---
 llvm/utils/UpdateTestChecks/common.py | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/llvm/utils/UpdateTestChecks/common.py b/llvm/utils/UpdateTestChecks/common.py
index ecb19d233a8d1..eed36a0cdd73f 100644
--- a/llvm/utils/UpdateTestChecks/common.py
+++ b/llvm/utils/UpdateTestChecks/common.py
@@ -980,10 +980,6 @@ def __init__(
     def is_local_def_ir_value(self):
         return self.ir_prefix == "%"
 
-    # Return true if this kind of IR value is "global", basically if it matches '#{{.*}}'.
-    def is_global_scope_ir_value_match(self, match):
-        return self.global_ir_rhs_regexp is not None
-
     # Return the IR prefix and check prefix we use for this kind or IR value,
     # e.g., (%, TMP) for locals. If the IR prefix is a regex, return the prefix
     # used in the IR output

From 9ec6c5d26321f5d32d97218f850ae7cafda32b2e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Tue, 16 Apr 2024 13:15:56 -0700
Subject: [PATCH 659/886] [flang][cuda] Add fir.deallocate operation (#88839)

Add the fir.cuda_deallocate operation that perform device deallocation
of data hold by a descriptor. This will replace the call to
AllocatableDeallocate from the runtime.

This is a companion operation to the one added in #88586
---
 .../include/flang/Optimizer/Dialect/FIROps.td | 25 +++++++++++++
 flang/lib/Optimizer/Dialect/FIROps.cpp        | 13 +++++++
 flang/test/Fir/cuf-invalid.fir                | 37 +++++++++++++++++++
 flang/test/Fir/cuf.mlir                       |  6 +++
 4 files changed, 81 insertions(+)

diff --git a/flang/include/flang/Optimizer/Dialect/FIROps.td b/flang/include/flang/Optimizer/Dialect/FIROps.td
index c181c7ed62dff..580e840587abb 100644
--- a/flang/include/flang/Optimizer/Dialect/FIROps.td
+++ b/flang/include/flang/Optimizer/Dialect/FIROps.td
@@ -3222,4 +3222,29 @@ def fir_CUDAAllocateOp : fir_Op<"cuda_allocate", [AttrSizedOperandSegments,
   let hasVerifier = 1;
 }
 
+def fir_CUDADeallocateOp : fir_Op<"cuda_deallocate",
+    [MemoryEffects<[MemFree<DefaultResource>]>]> {
+  let summary = "Perform the device deallocation of data of an allocatable";
+
+  let description = [{
+    The fir.cuda_deallocate operation performs the deallocation on the device
+    of the data of an allocatable.
+  }];
+
+  let arguments = (ins Arg<fir_ReferenceType, "", [MemRead, MemWrite]>:$box,
+                       Arg<Optional<AnyRefOrBoxType>, "", [MemWrite]>:$errmsg,
+                       fir_CUDADataAttributeAttr:$cuda_attr,
+                       UnitAttr:$hasStat);
+
+  let results = (outs AnyIntegerType:$stat);
+
+  let assemblyFormat = [{
+    $box `:` qualified(type($box))
+    ( `errmsg` `(` $errmsg^ `:` type($errmsg) `)` )?
+    attr-dict `->` type($stat)
+  }];
+
+  let hasVerifier = 1;
+}
+
 #endif
diff --git a/flang/lib/Optimizer/Dialect/FIROps.cpp b/flang/lib/Optimizer/Dialect/FIROps.cpp
index 88710880174d2..be27256d911b3 100644
--- a/flang/lib/Optimizer/Dialect/FIROps.cpp
+++ b/flang/lib/Optimizer/Dialect/FIROps.cpp
@@ -4012,6 +4012,19 @@ mlir::LogicalResult fir::CUDAAllocateOp::verify() {
   return mlir::success();
 }
 
+mlir::LogicalResult fir::CUDADeallocateOp::verify() {
+  if (!fir::unwrapRefType(getBox().getType()).isa<fir::BaseBoxType>())
+    return emitOpError(
+        "expect box to be a reference to class or box type value");
+  if (getErrmsg() &&
+      !fir::unwrapRefType(getErrmsg().getType()).isa<fir::BoxType>())
+    return emitOpError(
+        "expect errmsg to be a reference to/or a box type value");
+  if (getErrmsg() && !getHasStat())
+    return emitOpError("expect stat attribute when errmsg is provided");
+  return mlir::success();
+}
+
 //===----------------------------------------------------------------------===//
 // FIROpsDialect
 //===----------------------------------------------------------------------===//
diff --git a/flang/test/Fir/cuf-invalid.fir b/flang/test/Fir/cuf-invalid.fir
index 9c5ffe7176a3b..5d3aa55cf346a 100644
--- a/flang/test/Fir/cuf-invalid.fir
+++ b/flang/test/Fir/cuf-invalid.fir
@@ -48,3 +48,40 @@ func.func @_QPsub1() {
   %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> errmsg(%1 : !fir.ref<i32>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
   return
 }
+
+// -----
+
+func.func @_QPsub1() {
+  %1 = fir.alloca i32
+  // expected-error@+1{{'fir.cuda_deallocate' op expect box to be a reference to class or box type value}}
+  %2 = fir.cuda_deallocate %1 : !fir.ref<i32> {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %1 = fir.alloca i32
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  // expected-error@+1{{'fir.cuda_deallocate' op expect errmsg to be a reference to/or a box type value}}
+  %13 = fir.cuda_deallocate %11 : !fir.ref<!fir.box<none>> errmsg(%1 : !fir.ref<i32>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+  return
+}
+
+// -----
+
+func.func @_QPsub1() {
+  %0 = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+  %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+  %c100 = arith.constant 100 : index
+  %7 = fir.alloca !fir.char<1,100> {bindc_name = "msg", uniq_name = "_QFsub1Emsg"}
+  %8:2 = hlfir.declare %7 typeparams %c100 {uniq_name = "_QFsub1Emsg"} : (!fir.ref<!fir.char<1,100>>, index) -> (!fir.ref<!fir.char<1,100>>, !fir.ref<!fir.char<1,100>>)
+  %9 = fir.embox %8#1 : (!fir.ref<!fir.char<1,100>>) -> !fir.box<!fir.char<1,100>>
+  %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
+  %16 = fir.convert %9 : (!fir.box<!fir.char<1,100>>) -> !fir.box<none>
+  // expected-error@+1{{'fir.cuda_deallocate' op expect stat attribute when errmsg is provided}}
+  %13 = fir.cuda_deallocate %11 : !fir.ref<!fir.box<none>> errmsg(%16 : !fir.box<none>) {cuda_attr = #fir.cuda<device>} -> i32
+  return
+}
diff --git a/flang/test/Fir/cuf.mlir b/flang/test/Fir/cuf.mlir
index 67eff31b35b2b..71f0652067fac 100644
--- a/flang/test/Fir/cuf.mlir
+++ b/flang/test/Fir/cuf.mlir
@@ -7,10 +7,12 @@ func.func @_QPsub1() {
   %4:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
   %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
   %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
+  %14 = fir.cuda_deallocate %11 : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
   return
 }
 
 // CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
+// CHECK: fir.cuda_deallocate %{{.*}} : !fir.ref<!fir.box<none>> {cuda_attr = #fir.cuda<device>} -> i32
 
 // -----
 
@@ -66,5 +68,9 @@ func.func @_QPsub1() {
   %11 = fir.convert %4#1 : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> !fir.ref<!fir.box<none>>
   %16 = fir.convert %9 : (!fir.box<!fir.char<1,100>>) -> !fir.box<none>
   %13 = fir.cuda_allocate %11 : !fir.ref<!fir.box<none>> errmsg(%16 : !fir.box<none>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+  %14 = fir.cuda_deallocate %11 : !fir.ref<!fir.box<none>> errmsg(%16 : !fir.box<none>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
   return
 }
+
+// CHECK: fir.cuda_allocate %{{.*}} : !fir.ref<!fir.box<none>> errmsg(%{{.*}} : !fir.box<none>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+// CHECK: fir.cuda_deallocate %{{.*}} : !fir.ref<!fir.box<none>> errmsg(%{{.*}} : !fir.box<none>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32

From 34777c238b117b52dd41a9d12e8b54fb83677a12 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Tue, 16 Apr 2024 21:24:24 +0100
Subject: [PATCH 660/886] [VPlan] Don't mark VPBlendRecipe as phi-like.

VPBlendRecipes don't get lowered to phis and usually do not appear at
the beginning of blocks, due to their masks appearing before them.

This effectively relaxes an over-eager verifier message.

Fixes https://github.com/llvm/llvm-project/issues/88297.
Fixes https://github.com/llvm/llvm-project/issues/88804.
---
 llvm/lib/Transforms/Vectorize/VPlanValue.h    |   4 +-
 .../LoopVectorize/blend-in-header.ll          | 233 ++++++++++++++++++
 2 files changed, 235 insertions(+), 2 deletions(-)
 create mode 100644 llvm/test/Transforms/LoopVectorize/blend-in-header.ll

diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index da3a768552fc5..3f8d4f4fe7d64 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -361,8 +361,8 @@ class VPDef {
     VPWidenMemoryInstructionSC,
     VPWidenSC,
     VPWidenSelectSC,
-    // START: Phi-like recipes. Need to be kept together.
     VPBlendSC,
+    // START: Phi-like recipes. Need to be kept together.
     VPWidenPHISC,
     VPPredInstPHISC,
     // START: SubclassID for recipes that inherit VPHeaderPHIRecipe.
@@ -376,7 +376,7 @@ class VPDef {
     VPReductionPHISC,
     // END: SubclassID for recipes that inherit VPHeaderPHIRecipe
     // END: Phi-like recipes
-    VPFirstPHISC = VPBlendSC,
+    VPFirstPHISC = VPWidenPHISC,
     VPFirstHeaderPHISC = VPCanonicalIVPHISC,
     VPLastHeaderPHISC = VPReductionPHISC,
     VPLastPHISC = VPReductionPHISC,
diff --git a/llvm/test/Transforms/LoopVectorize/blend-in-header.ll b/llvm/test/Transforms/LoopVectorize/blend-in-header.ll
new file mode 100644
index 0000000000000..01e223a324379
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/blend-in-header.ll
@@ -0,0 +1,233 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128-ni:1-p2:32:8:8:32-ni:2"
+
+; Test with blend recipe in header VPBB, from
+; https://github.com/llvm/llvm-project/issues/88297.
+define i64 @pr88297() {
+; CHECK-LABEL: define i64 @pr88297() {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    br i1 false, label [[LOOP_LATCH]], label [[THEN:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 1, [[THEN]] ], [ 0, [[LOOP_HEADER]] ]
+; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[R_LCSSA]]
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 false, label %loop.latch, label %then
+
+then:
+  br label %loop.latch
+
+loop.latch:
+  %r = phi i64 [ 1, %then ], [ 0, %loop.header ]
+  %iv.next = add i32 %iv, 1
+  %icmp = icmp sgt i32 %iv, 1000
+  br i1 %icmp, label %exit, label %loop.header
+
+exit:
+  %r.lcssa = phi i64 [ %r, %loop.latch ]
+  ret i64 %r.lcssa
+}
+
+define i64 @pr88297_incoming_ops_reordered() {
+; CHECK-LABEL: define i64 @pr88297_incoming_ops_reordered() {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    br i1 false, label [[LOOP_LATCH]], label [[THEN:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 0, [[LOOP_HEADER]] ], [ 1, [[THEN]] ]
+; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[R_LCSSA]]
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 false, label %loop.latch, label %then
+
+then:
+  br label %loop.latch
+
+loop.latch:
+  %r = phi i64 [ 0, %loop.header ], [ 1, %then ]
+  %iv.next = add i32 %iv, 1
+  %icmp = icmp sgt i32 %iv, 1000
+  br i1 %icmp, label %exit, label %loop.header
+
+exit:
+  %r.lcssa = phi i64 [ %r, %loop.latch ]
+  ret i64 %r.lcssa
+}
+
+define i64 @invar_cond(i1 %c) {
+; CHECK-LABEL: define i64 @invar_cond(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x i64> zeroinitializer, <4 x i64> <i64 1, i64 1, i64 1, i64 1>
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i64> [[PREDPHI]], i32 3
+; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    br i1 [[C]], label [[LOOP_LATCH]], label [[THEN:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 1, [[THEN]] ], [ 0, [[LOOP_HEADER]] ]
+; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[R_LCSSA]]
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 %c, label %loop.latch, label %then
+
+then:
+  br label %loop.latch
+
+loop.latch:
+  %r = phi i64 [ 1, %then ], [ 0, %loop.header ]
+  %iv.next = add i32 %iv, 1
+  %icmp = icmp sgt i32 %iv, 1000
+  br i1 %icmp, label %exit, label %loop.header
+
+exit:
+  %r.lcssa = phi i64 [ %r, %loop.latch ]
+  ret i64 %r.lcssa
+}
+
+define i64 @invar_cond_incoming_ops_reordered(i1 %c) {
+; CHECK-LABEL: define i64 @invar_cond_incoming_ops_reordered(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true, i1 true, i1 true>
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i64> zeroinitializer
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i64> [[PREDPHI]], i32 3
+; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    br i1 [[C]], label [[LOOP_LATCH]], label [[THEN:%.*]]
+; CHECK:       then:
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 0, [[LOOP_HEADER]] ], [ 1, [[THEN]] ]
+; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[R_LCSSA]]
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i32 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 %c, label %loop.latch, label %then
+
+then:
+  br label %loop.latch
+
+loop.latch:
+  %r = phi i64 [ 0, %loop.header ], [ 1, %then ]
+  %iv.next = add i32 %iv, 1
+  %icmp = icmp sgt i32 %iv, 1000
+  br i1 %icmp, label %exit, label %loop.header
+
+exit:
+  %r.lcssa = phi i64 [ %r, %loop.latch ]
+  ret i64 %r.lcssa
+}
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
+; CHECK: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; CHECK: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]}
+;.

From b6bd41db31c798f3fc82368381fad6d42795f512 Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Thu, 21 Mar 2024 11:01:21 -0500
Subject: [PATCH 661/886] [InstCombine] Add canonicalization of `sitofp` ->
 `uitofp nneg`

This is essentially the same as #82404 but has the `nneg` flag which
allows the backend to reliably undo the transform.

Closes #88299
---
 clang/test/Headers/__clang_hip_math.hip       | 24 +++----
 .../InstCombine/InstCombineCasts.cpp          | 18 ++++-
 .../test/Transforms/InstCombine/add-sitofp.ll | 20 +++---
 .../Transforms/InstCombine/binop-itofp.ll     | 66 +++++++++----------
 .../Transforms/InstCombine/clamp-to-minmax.ll | 10 +--
 llvm/test/Transforms/InstCombine/fpcast.ll    | 24 +++----
 .../Transforms/InstCombine/minmax-fold.ll     | 10 +--
 llvm/test/Transforms/InstCombine/minmax-fp.ll |  2 +-
 llvm/test/Transforms/InstCombine/pr27236.ll   |  2 +-
 llvm/test/Transforms/InstCombine/sitofp.ll    |  2 +-
 .../LoopVectorize/X86/float-induction-x86.ll  |  6 +-
 .../LoopVectorize/float-induction.ll          | 56 ++++++++--------
 12 files changed, 127 insertions(+), 113 deletions(-)

diff --git a/clang/test/Headers/__clang_hip_math.hip b/clang/test/Headers/__clang_hip_math.hip
index 2e5f521a5feae..1271868a53b86 100644
--- a/clang/test/Headers/__clang_hip_math.hip
+++ b/clang/test/Headers/__clang_hip_math.hip
@@ -1685,7 +1685,7 @@ extern "C" __device__ double test_j1(double x) {
 // DEFAULT-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// DEFAULT-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // DEFAULT-NEXT:    [[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:    [[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], [[DIV_I]]
 // DEFAULT-NEXT:    [[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1718,7 +1718,7 @@ extern "C" __device__ double test_j1(double x) {
 // FINITEONLY-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// FINITEONLY-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // FINITEONLY-NEXT:    [[DIV_I:%.*]] = fdiv nnan ninf contract float [[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:    [[MUL8_I:%.*]] = fmul nnan ninf contract float [[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:    [[SUB_I]] = fsub nnan ninf contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1751,7 +1751,7 @@ extern "C" __device__ double test_j1(double x) {
 // APPROX-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// APPROX-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // APPROX-NEXT:    [[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // APPROX-NEXT:    [[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], [[DIV_I]]
 // APPROX-NEXT:    [[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -1788,7 +1788,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // DEFAULT-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// DEFAULT-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // DEFAULT-NEXT:    [[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:    [[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], [[DIV_I]]
 // DEFAULT-NEXT:    [[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -1821,7 +1821,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // FINITEONLY-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// FINITEONLY-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // FINITEONLY-NEXT:    [[DIV_I:%.*]] = fdiv nnan ninf contract double [[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:    [[MUL8_I:%.*]] = fmul nnan ninf contract double [[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:    [[SUB_I]] = fsub nnan ninf contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -1854,7 +1854,7 @@ extern "C" __device__ float test_jnf(int x, float y) {
 // APPROX-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// APPROX-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // APPROX-NEXT:    [[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // APPROX-NEXT:    [[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], [[DIV_I]]
 // APPROX-NEXT:    [[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -4222,7 +4222,7 @@ extern "C" __device__ double test_y1(double x) {
 // DEFAULT-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// DEFAULT-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // DEFAULT-NEXT:    [[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:    [[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], [[DIV_I]]
 // DEFAULT-NEXT:    [[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -4255,7 +4255,7 @@ extern "C" __device__ double test_y1(double x) {
 // FINITEONLY-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// FINITEONLY-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // FINITEONLY-NEXT:    [[DIV_I:%.*]] = fdiv nnan ninf contract float [[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:    [[MUL8_I:%.*]] = fmul nnan ninf contract float [[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:    [[SUB_I]] = fsub nnan ninf contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -4288,7 +4288,7 @@ extern "C" __device__ double test_y1(double x) {
 // APPROX-NEXT:    [[__X1_0_I3:%.*]] = phi float [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[__X0_0_I2:%.*]] = phi float [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to float
+// APPROX-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to float
 // APPROX-NEXT:    [[DIV_I:%.*]] = fdiv contract float [[CONV_I]], [[Y]]
 // APPROX-NEXT:    [[MUL8_I:%.*]] = fmul contract float [[__X1_0_I3]], [[DIV_I]]
 // APPROX-NEXT:    [[SUB_I]] = fsub contract float [[MUL8_I]], [[__X0_0_I2]]
@@ -4325,7 +4325,7 @@ extern "C" __device__ float test_ynf(int x, float y) {
 // DEFAULT-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // DEFAULT-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// DEFAULT-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// DEFAULT-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // DEFAULT-NEXT:    [[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // DEFAULT-NEXT:    [[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], [[DIV_I]]
 // DEFAULT-NEXT:    [[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -4358,7 +4358,7 @@ extern "C" __device__ float test_ynf(int x, float y) {
 // FINITEONLY-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // FINITEONLY-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// FINITEONLY-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// FINITEONLY-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // FINITEONLY-NEXT:    [[DIV_I:%.*]] = fdiv nnan ninf contract double [[CONV_I]], [[Y]]
 // FINITEONLY-NEXT:    [[MUL8_I:%.*]] = fmul nnan ninf contract double [[__X1_0_I3]], [[DIV_I]]
 // FINITEONLY-NEXT:    [[SUB_I]] = fsub nnan ninf contract double [[MUL8_I]], [[__X0_0_I2]]
@@ -4391,7 +4391,7 @@ extern "C" __device__ float test_ynf(int x, float y) {
 // APPROX-NEXT:    [[__X1_0_I3:%.*]] = phi double [ [[SUB_I:%.*]], [[FOR_BODY_I]] ], [ [[CALL_I21_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[__X0_0_I2:%.*]] = phi double [ [[__X1_0_I3]], [[FOR_BODY_I]] ], [ [[CALL_I_I]], [[IF_END4_I]] ]
 // APPROX-NEXT:    [[MUL_I:%.*]] = shl nuw nsw i32 [[__I_0_I4]], 1
-// APPROX-NEXT:    [[CONV_I:%.*]] = sitofp i32 [[MUL_I]] to double
+// APPROX-NEXT:    [[CONV_I:%.*]] = uitofp nneg i32 [[MUL_I]] to double
 // APPROX-NEXT:    [[DIV_I:%.*]] = fdiv contract double [[CONV_I]], [[Y]]
 // APPROX-NEXT:    [[MUL8_I:%.*]] = fmul contract double [[__X1_0_I3]], [[DIV_I]]
 // APPROX-NEXT:    [[SUB_I]] = fsub contract double [[MUL8_I]], [[__X0_0_I2]]
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 437e9b92c7032..d242d3f443def 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -1977,11 +1977,25 @@ Instruction *InstCombinerImpl::visitFPToSI(FPToSIInst &FI) {
 }
 
 Instruction *InstCombinerImpl::visitUIToFP(CastInst &CI) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+    return R;
+  if (!CI.hasNonNeg() && isKnownNonNegative(CI.getOperand(0), SQ)) {
+    CI.setNonNeg();
+    return &CI;
+  }
+  return nullptr;
 }
 
 Instruction *InstCombinerImpl::visitSIToFP(CastInst &CI) {
-  return commonCastTransforms(CI);
+  if (Instruction *R = commonCastTransforms(CI))
+    return R;
+  if (isKnownNonNegative(CI.getOperand(0), SQ)) {
+    auto UI =
+        CastInst::Create(Instruction::UIToFP, CI.getOperand(0), CI.getType());
+    UI->setNonNeg(true);
+    return UI;
+  }
+  return nullptr;
 }
 
 Instruction *InstCombinerImpl::visitIntToPtr(IntToPtrInst &CI) {
diff --git a/llvm/test/Transforms/InstCombine/add-sitofp.ll b/llvm/test/Transforms/InstCombine/add-sitofp.ll
index 2bdc808d9771c..f1afcaf5f85d2 100644
--- a/llvm/test/Transforms/InstCombine/add-sitofp.ll
+++ b/llvm/test/Transforms/InstCombine/add-sitofp.ll
@@ -6,7 +6,7 @@ define double @x(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[A:%.*]], 24
 ; CHECK-NEXT:    [[N:%.*]] = and i32 [[M]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[N]], 1
-; CHECK-NEXT:    [[P:%.*]] = uitofp i32 [[TMP1]] to double
+; CHECK-NEXT:    [[P:%.*]] = uitofp nneg i32 [[TMP1]] to double
 ; CHECK-NEXT:    ret double [[P]]
 ;
   %m = lshr i32 %a, 24
@@ -20,7 +20,7 @@ define double @test(i32 %a) {
 ; CHECK-LABEL: @test(
 ; CHECK-NEXT:    [[A_AND:%.*]] = and i32 [[A:%.*]], 1073741823
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[A_AND]], 1
-; CHECK-NEXT:    [[RES:%.*]] = uitofp i32 [[TMP1]] to double
+; CHECK-NEXT:    [[RES:%.*]] = uitofp nneg i32 [[TMP1]] to double
 ; CHECK-NEXT:    ret double [[RES]]
 ;
   ; Drop two highest bits to guarantee that %a + 1 doesn't overflow
@@ -33,7 +33,7 @@ define double @test(i32 %a) {
 define float @test_neg(i32 %a) {
 ; CHECK-LABEL: @test_neg(
 ; CHECK-NEXT:    [[A_AND:%.*]] = and i32 [[A:%.*]], 1073741823
-; CHECK-NEXT:    [[A_AND_FP:%.*]] = sitofp i32 [[A_AND]] to float
+; CHECK-NEXT:    [[A_AND_FP:%.*]] = uitofp nneg i32 [[A_AND]] to float
 ; CHECK-NEXT:    [[RES:%.*]] = fadd float [[A_AND_FP]], 1.000000e+00
 ; CHECK-NEXT:    ret float [[RES]]
 ;
@@ -49,7 +49,7 @@ define double @test_2(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[A_AND:%.*]] = and i32 [[A:%.*]], 1073741823
 ; CHECK-NEXT:    [[B_AND:%.*]] = and i32 [[B:%.*]], 1073741823
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[A_AND]], [[B_AND]]
-; CHECK-NEXT:    [[RES:%.*]] = uitofp i32 [[TMP1]] to double
+; CHECK-NEXT:    [[RES:%.*]] = uitofp nneg i32 [[TMP1]] to double
 ; CHECK-NEXT:    ret double [[RES]]
 ;
   ; Drop two highest bits to guarantee that %a + %b doesn't overflow
@@ -67,8 +67,8 @@ define float @test_2_neg(i32 %a, i32 %b) {
 ; CHECK-LABEL: @test_2_neg(
 ; CHECK-NEXT:    [[A_AND:%.*]] = and i32 [[A:%.*]], 1073741823
 ; CHECK-NEXT:    [[B_AND:%.*]] = and i32 [[B:%.*]], 1073741823
-; CHECK-NEXT:    [[A_AND_FP:%.*]] = sitofp i32 [[A_AND]] to float
-; CHECK-NEXT:    [[B_AND_FP:%.*]] = sitofp i32 [[B_AND]] to float
+; CHECK-NEXT:    [[A_AND_FP:%.*]] = uitofp nneg i32 [[A_AND]] to float
+; CHECK-NEXT:    [[B_AND_FP:%.*]] = uitofp nneg i32 [[B_AND]] to float
 ; CHECK-NEXT:    [[RES:%.*]] = fadd float [[A_AND_FP]], [[B_AND_FP]]
 ; CHECK-NEXT:    ret float [[RES]]
 ;
@@ -89,7 +89,7 @@ define float @test_3(i32 %a, i32 %b) {
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[A:%.*]], 24
 ; CHECK-NEXT:    [[N:%.*]] = and i32 [[M]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[N]], 1
-; CHECK-NEXT:    [[P:%.*]] = uitofp i32 [[TMP1]] to float
+; CHECK-NEXT:    [[P:%.*]] = uitofp nneg i32 [[TMP1]] to float
 ; CHECK-NEXT:    ret float [[P]]
 ;
   %m = lshr i32 %a, 24
@@ -104,7 +104,7 @@ define <4 x double> @test_4(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-NEXT:    [[A_AND:%.*]] = and <4 x i32> [[A:%.*]], <i32 1073741823, i32 1073741823, i32 1073741823, i32 1073741823>
 ; CHECK-NEXT:    [[B_AND:%.*]] = and <4 x i32> [[B:%.*]], <i32 1073741823, i32 1073741823, i32 1073741823, i32 1073741823>
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw <4 x i32> [[A_AND]], [[B_AND]]
-; CHECK-NEXT:    [[RES:%.*]] = uitofp <4 x i32> [[TMP1]] to <4 x double>
+; CHECK-NEXT:    [[RES:%.*]] = uitofp nneg <4 x i32> [[TMP1]] to <4 x double>
 ; CHECK-NEXT:    ret <4 x double> [[RES]]
 ;
   ; Drop two highest bits to guarantee that %a + %b doesn't overflow
@@ -122,8 +122,8 @@ define <4 x float> @test_4_neg(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: @test_4_neg(
 ; CHECK-NEXT:    [[A_AND:%.*]] = and <4 x i32> [[A:%.*]], <i32 1073741823, i32 1073741823, i32 1073741823, i32 1073741823>
 ; CHECK-NEXT:    [[B_AND:%.*]] = and <4 x i32> [[B:%.*]], <i32 1073741823, i32 1073741823, i32 1073741823, i32 1073741823>
-; CHECK-NEXT:    [[A_AND_FP:%.*]] = sitofp <4 x i32> [[A_AND]] to <4 x float>
-; CHECK-NEXT:    [[B_AND_FP:%.*]] = sitofp <4 x i32> [[B_AND]] to <4 x float>
+; CHECK-NEXT:    [[A_AND_FP:%.*]] = uitofp nneg <4 x i32> [[A_AND]] to <4 x float>
+; CHECK-NEXT:    [[B_AND_FP:%.*]] = uitofp nneg <4 x i32> [[B_AND]] to <4 x float>
 ; CHECK-NEXT:    [[RES:%.*]] = fadd <4 x float> [[A_AND_FP]], [[B_AND_FP]]
 ; CHECK-NEXT:    ret <4 x float> [[RES]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/binop-itofp.ll b/llvm/test/Transforms/InstCombine/binop-itofp.ll
index d72a54e8babc9..097a8196af80f 100644
--- a/llvm/test/Transforms/InstCombine/binop-itofp.ll
+++ b/llvm/test/Transforms/InstCombine/binop-itofp.ll
@@ -21,7 +21,7 @@ define half @test_ui_ui_i8_add_fail_overflow(i8 noundef %x_in, i8 noundef %y_in)
 ; CHECK-LABEL: @test_ui_ui_i8_add_fail_overflow(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 127
 ; CHECK-NEXT:    [[Y:%.*]] = and i8 [[Y_IN:%.*]], -127
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = uitofp i8 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -49,7 +49,7 @@ define half @test_ui_ui_i8_add_C(i8 noundef %x_in) {
 define half @test_ui_ui_i8_add_C_fail_no_repr(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i8_add_C_fail_no_repr(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 127
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], 0xH57F8
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -62,7 +62,7 @@ define half @test_ui_ui_i8_add_C_fail_no_repr(i8 noundef %x_in) {
 define half @test_ui_ui_i8_add_C_fail_overflow(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i8_add_C_fail_overflow(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 127
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], 0xH5808
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -110,7 +110,7 @@ define half @test_ui_si_i8_add(i8 noundef %x_in, i8 noundef %y_in) {
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 63
 ; CHECK-NEXT:    [[Y:%.*]] = and i8 [[Y_IN:%.*]], 63
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i8 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i8 %x_in, 63
@@ -140,7 +140,7 @@ define half @test_ui_si_i8_add_overflow(i8 noundef %x_in, i8 noundef %y_in) {
 define half @test_ui_ui_i8_sub_C(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i8_sub_C(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and i8 [[X_IN:%.*]], 127
-; CHECK-NEXT:    [[R:%.*]] = uitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = or i8 %x_in, 128
@@ -166,7 +166,7 @@ define half @test_si_si_i8_sub(i8 noundef %x_in, i8 noundef %y_in) {
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 63
 ; CHECK-NEXT:    [[Y:%.*]] = or i8 [[Y_IN:%.*]], -64
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub nsw i8 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = sitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i8 %x_in, 63
@@ -181,7 +181,7 @@ define half @test_si_si_i8_sub_fail_overflow(i8 noundef %x_in, i8 noundef %y_in)
 ; CHECK-LABEL: @test_si_si_i8_sub_fail_overflow(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 63
 ; CHECK-NEXT:    [[Y:%.*]] = or i8 [[Y_IN:%.*]], -65
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i8 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fsub half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -198,7 +198,7 @@ define half @test_si_si_i8_sub_C(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_si_si_i8_sub_C(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 63
 ; CHECK-NEXT:    [[TMP1:%.*]] = or disjoint i8 [[X]], 64
-; CHECK-NEXT:    [[R:%.*]] = sitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i8 %x_in, 63
@@ -283,7 +283,7 @@ define half @test_ui_ui_i8_mul_C(i8 noundef %x_in) {
 define half @test_ui_ui_i8_mul_C_fail_overlow(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i8_mul_C_fail_overlow(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 14
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], 0xH4CC0
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -315,7 +315,7 @@ define half @test_si_si_i8_mul_fail_maybe_zero(i8 noundef %x_in, i8 noundef %y_i
 ; CHECK-LABEL: @test_si_si_i8_mul_fail_maybe_zero(
 ; CHECK-NEXT:    [[X:%.*]] = and i8 [[X_IN:%.*]], 7
 ; CHECK-NEXT:    [[Y:%.*]] = or i8 [[Y_IN:%.*]], -8
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i8 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -332,7 +332,7 @@ define half @test_si_si_i8_mul_C_fail_no_repr(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_si_si_i8_mul_C_fail_no_repr(
 ; CHECK-NEXT:    [[XX:%.*]] = and i8 [[X_IN:%.*]], 6
 ; CHECK-NEXT:    [[X:%.*]] = or disjoint i8 [[XX]], 1
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], 0xHC780
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -347,7 +347,7 @@ define half @test_si_si_i8_mul_C_fail_overflow(i8 noundef %x_in) {
 ; CHECK-LABEL: @test_si_si_i8_mul_C_fail_overflow(
 ; CHECK-NEXT:    [[XX:%.*]] = and i8 [[X_IN:%.*]], 6
 ; CHECK-NEXT:    [[X:%.*]] = or disjoint i8 [[XX]], 1
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], 0xHCCC0
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -365,7 +365,7 @@ define half @test_ui_si_i8_mul(i8 noundef %x_in, i8 noundef %y_in) {
 ; CHECK-NEXT:    [[YY:%.*]] = and i8 [[Y_IN:%.*]], 7
 ; CHECK-NEXT:    [[Y:%.*]] = add nuw nsw i8 [[YY]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i8 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %xx = and i8 %x_in, 6
@@ -384,7 +384,7 @@ define half @test_ui_si_i8_mul_fail_maybe_zero(i8 noundef %x_in, i8 noundef %y_i
 ; CHECK-NEXT:    [[X:%.*]] = add nuw nsw i8 [[XX]], 1
 ; CHECK-NEXT:    [[Y:%.*]] = and i8 [[Y_IN:%.*]], 7
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i8 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i8 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i8 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %xx = and i8 %x_in, 7
@@ -401,7 +401,7 @@ define half @test_ui_si_i8_mul_fail_signed(i8 noundef %x_in, i8 noundef %y_in) {
 ; CHECK-NEXT:    [[XX:%.*]] = and i8 [[X_IN:%.*]], 7
 ; CHECK-NEXT:    [[X:%.*]] = add nuw nsw i8 [[XX]], 1
 ; CHECK-NEXT:    [[Y:%.*]] = or i8 [[Y_IN:%.*]], -4
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i8 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i8 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = uitofp i8 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -420,7 +420,7 @@ define half @test_ui_ui_i16_add(i16 noundef %x_in, i16 noundef %y_in) {
 ; CHECK-NEXT:    [[X:%.*]] = and i16 [[X_IN:%.*]], 2047
 ; CHECK-NEXT:    [[Y:%.*]] = and i16 [[Y_IN:%.*]], 2047
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i16 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i16 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i16 %x_in, 2047
@@ -435,8 +435,8 @@ define half @test_ui_ui_i16_add_fail_not_promotable(i16 noundef %x_in, i16 nound
 ; CHECK-LABEL: @test_ui_ui_i16_add_fail_not_promotable(
 ; CHECK-NEXT:    [[X:%.*]] = and i16 [[X_IN:%.*]], 2049
 ; CHECK-NEXT:    [[Y:%.*]] = and i16 [[Y_IN:%.*]], 2047
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i16 [[X]] to half
-; CHECK-NEXT:    [[YF:%.*]] = uitofp i16 [[Y]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i16 [[X]] to half
+; CHECK-NEXT:    [[YF:%.*]] = uitofp nneg i16 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -463,7 +463,7 @@ define half @test_ui_ui_i16_add_C(i16 noundef %x_in) {
 define half @test_ui_ui_i16_add_C_fail_overflow(i16 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i16_add_C_fail_overflow(
 ; CHECK-NEXT:    [[X:%.*]] = and i16 [[X_IN:%.*]], 2047
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i16 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i16 [[X]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], 0xH7BD0
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -541,7 +541,7 @@ define half @test_si_si_i16_sub_fail_no_promotion(i16 noundef %x_in, i16 noundef
 ; CHECK-LABEL: @test_si_si_i16_sub_fail_no_promotion(
 ; CHECK-NEXT:    [[X:%.*]] = and i16 [[X_IN:%.*]], 2047
 ; CHECK-NEXT:    [[Y:%.*]] = or i16 [[Y_IN:%.*]], -2049
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i16 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i16 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i16 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fsub half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -575,7 +575,7 @@ define half @test_ui_si_i16_sub_fail_maybe_signed(i16 noundef %x_in, i16 noundef
 ; CHECK-NEXT:    [[X:%.*]] = or i16 [[X_IN:%.*]], -2048
 ; CHECK-NEXT:    [[Y:%.*]] = and i16 [[Y_IN:%.*]], 2047
 ; CHECK-NEXT:    [[XF:%.*]] = uitofp i16 [[X]] to half
-; CHECK-NEXT:    [[YF:%.*]] = sitofp i16 [[Y]] to half
+; CHECK-NEXT:    [[YF:%.*]] = uitofp nneg i16 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fsub half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -607,8 +607,8 @@ define half @test_ui_ui_i16_mul_fail_no_promotion(i16 noundef %x_in, i16 noundef
 ; CHECK-LABEL: @test_ui_ui_i16_mul_fail_no_promotion(
 ; CHECK-NEXT:    [[X:%.*]] = and i16 [[X_IN:%.*]], 4095
 ; CHECK-NEXT:    [[Y:%.*]] = and i16 [[Y_IN:%.*]], 3
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i16 [[X]] to half
-; CHECK-NEXT:    [[YF:%.*]] = uitofp i16 [[Y]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i16 [[X]] to half
+; CHECK-NEXT:    [[YF:%.*]] = uitofp nneg i16 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -643,7 +643,7 @@ define half @test_si_si_i16_mul_fail_overflow(i16 noundef %x_in, i16 noundef %y_
 ; CHECK-NEXT:    [[XX:%.*]] = and i16 [[X_IN:%.*]], 126
 ; CHECK-NEXT:    [[X:%.*]] = or disjoint i16 [[XX]], 1
 ; CHECK-NEXT:    [[Y:%.*]] = or i16 [[Y_IN:%.*]], -257
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i16 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i16 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i16 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -690,7 +690,7 @@ define half @test_ui_si_i16_mul(i16 noundef %x_in, i16 noundef %y_in) {
 ; CHECK-NEXT:    [[YY:%.*]] = and i16 [[Y_IN:%.*]], 126
 ; CHECK-NEXT:    [[Y:%.*]] = or disjoint i16 [[YY]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i16 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i16 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %xx = and i16 %x_in, 126
@@ -723,7 +723,7 @@ define half @test_ui_ui_i12_add_fail_overflow(i12 noundef %x_in, i12 noundef %y_
 ; CHECK-LABEL: @test_ui_ui_i12_add_fail_overflow(
 ; CHECK-NEXT:    [[X:%.*]] = and i12 [[X_IN:%.*]], 2047
 ; CHECK-NEXT:    [[Y:%.*]] = and i12 [[Y_IN:%.*]], -2047
-; CHECK-NEXT:    [[XF:%.*]] = uitofp i12 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i12 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = uitofp i12 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fadd half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -821,7 +821,7 @@ define half @test_si_si_i12_sub(i12 noundef %x_in, i12 noundef %y_in) {
 ; CHECK-NEXT:    [[X:%.*]] = and i12 [[X_IN:%.*]], 1023
 ; CHECK-NEXT:    [[Y:%.*]] = or i12 [[Y_IN:%.*]], -1024
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub nsw i12 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = sitofp i12 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i12 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i12 %x_in, 1023
@@ -850,7 +850,7 @@ define half @test_ui_ui_i12_mul(i12 noundef %x_in, i12 noundef %y_in) {
 ; CHECK-NEXT:    [[X:%.*]] = and i12 [[X_IN:%.*]], 31
 ; CHECK-NEXT:    [[Y:%.*]] = and i12 [[Y_IN:%.*]], 63
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i12 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i12 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i12 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i12 %x_in, 31
@@ -883,7 +883,7 @@ define half @test_ui_ui_i12_mul_C(i12 noundef %x_in) {
 ; CHECK-LABEL: @test_ui_ui_i12_mul_C(
 ; CHECK-NEXT:    [[X:%.*]] = shl i12 [[X_IN:%.*]], 6
 ; CHECK-NEXT:    [[TMP1:%.*]] = and i12 [[X]], 1984
-; CHECK-NEXT:    [[R:%.*]] = uitofp i12 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i12 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = and i12 %x_in, 31
@@ -915,7 +915,7 @@ define half @test_si_si_i12_mul_fail_overflow(i12 noundef %x_in, i12 noundef %y_
 ; CHECK-NEXT:    [[XX:%.*]] = and i12 [[X_IN:%.*]], 30
 ; CHECK-NEXT:    [[X:%.*]] = or disjoint i12 [[XX]], 1
 ; CHECK-NEXT:    [[Y:%.*]] = or i12 [[Y_IN:%.*]], -128
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i12 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i12 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i12 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -933,7 +933,7 @@ define half @test_si_si_i12_mul_fail_maybe_non_zero(i12 noundef %x_in, i12 nound
 ; CHECK-LABEL: @test_si_si_i12_mul_fail_maybe_non_zero(
 ; CHECK-NEXT:    [[X:%.*]] = and i12 [[X_IN:%.*]], 30
 ; CHECK-NEXT:    [[Y:%.*]] = or i12 [[Y_IN:%.*]], -128
-; CHECK-NEXT:    [[XF:%.*]] = sitofp i12 [[X]] to half
+; CHECK-NEXT:    [[XF:%.*]] = uitofp nneg i12 [[X]] to half
 ; CHECK-NEXT:    [[YF:%.*]] = sitofp i12 [[Y]] to half
 ; CHECK-NEXT:    [[R:%.*]] = fmul half [[XF]], [[YF]]
 ; CHECK-NEXT:    ret half [[R]]
@@ -950,7 +950,7 @@ define half @test_si_si_i12_mul_C(i12 noundef %x_in) {
 ; CHECK-LABEL: @test_si_si_i12_mul_C(
 ; CHECK-NEXT:    [[X:%.*]] = or i12 [[X_IN:%.*]], -64
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nsw i12 [[X]], -16
-; CHECK-NEXT:    [[R:%.*]] = sitofp i12 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i12 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %x = or i12 %x_in, -64
@@ -979,7 +979,7 @@ define half @test_ui_si_i12_mul_nsw(i12 noundef %x_in, i12 noundef %y_in) {
 ; CHECK-NEXT:    [[YY:%.*]] = and i12 [[Y_IN:%.*]], 30
 ; CHECK-NEXT:    [[Y:%.*]] = or disjoint i12 [[YY]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw nsw i12 [[X]], [[Y]]
-; CHECK-NEXT:    [[R:%.*]] = uitofp i12 [[TMP1]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i12 [[TMP1]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %xx = and i12 %x_in, 31
diff --git a/llvm/test/Transforms/InstCombine/clamp-to-minmax.ll b/llvm/test/Transforms/InstCombine/clamp-to-minmax.ll
index 9da9eb36d381f..1dd0b17e9f46d 100644
--- a/llvm/test/Transforms/InstCombine/clamp-to-minmax.ll
+++ b/llvm/test/Transforms/InstCombine/clamp-to-minmax.ll
@@ -472,7 +472,7 @@ define float @ui32_clamp_and_cast_to_float(i32 %x) {
 ; CHECK-LABEL: @ui32_clamp_and_cast_to_float(
 ; CHECK-NEXT:    [[LO_CMP:%.*]] = icmp eq i32 [[X:%.*]], 0
 ; CHECK-NEXT:    [[MIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[X]], i32 255)
-; CHECK-NEXT:    [[MIN:%.*]] = uitofp i32 [[MIN1]] to float
+; CHECK-NEXT:    [[MIN:%.*]] = uitofp nneg i32 [[MIN1]] to float
 ; CHECK-NEXT:    [[R:%.*]] = select i1 [[LO_CMP]], float 1.000000e+00, float [[MIN]]
 ; CHECK-NEXT:    ret float [[R]]
 ;
@@ -488,7 +488,7 @@ define float @ui64_clamp_and_cast_to_float(i64 %x) {
 ; CHECK-LABEL: @ui64_clamp_and_cast_to_float(
 ; CHECK-NEXT:    [[LO_CMP:%.*]] = icmp eq i64 [[X:%.*]], 0
 ; CHECK-NEXT:    [[MIN1:%.*]] = call i64 @llvm.umin.i64(i64 [[X]], i64 255)
-; CHECK-NEXT:    [[MIN:%.*]] = uitofp i64 [[MIN1]] to float
+; CHECK-NEXT:    [[MIN:%.*]] = uitofp nneg i64 [[MIN1]] to float
 ; CHECK-NEXT:    [[R:%.*]] = select i1 [[LO_CMP]], float 1.000000e+00, float [[MIN]]
 ; CHECK-NEXT:    ret float [[R]]
 ;
@@ -504,7 +504,7 @@ define float @mixed_clamp_to_float_1(i32 %x) {
 ; CHECK-LABEL: @mixed_clamp_to_float_1(
 ; CHECK-NEXT:    [[SI_MIN:%.*]] = call i32 @llvm.smin.i32(i32 [[X:%.*]], i32 255)
 ; CHECK-NEXT:    [[R1:%.*]] = call i32 @llvm.smax.i32(i32 [[SI_MIN]], i32 1)
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[R1]] to float
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[R1]] to float
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %si_min_cmp = icmp sgt i32 %x, 255
@@ -539,7 +539,7 @@ define float @mixed_clamp_to_float_2(i32 %x) {
 ; CHECK-LABEL: @mixed_clamp_to_float_2(
 ; CHECK-NEXT:    [[SI_MIN:%.*]] = call i32 @llvm.smin.i32(i32 [[X:%.*]], i32 255)
 ; CHECK-NEXT:    [[R1:%.*]] = call i32 @llvm.smax.i32(i32 [[SI_MIN]], i32 1)
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[R1]] to float
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[R1]] to float
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %si_min_cmp = icmp sgt i32 %x, 255
@@ -572,7 +572,7 @@ define <2 x float> @mixed_clamp_to_float_vec(<2 x i32> %x) {
 ; CHECK-LABEL: @mixed_clamp_to_float_vec(
 ; CHECK-NEXT:    [[SI_MIN:%.*]] = call <2 x i32> @llvm.smin.v2i32(<2 x i32> [[X:%.*]], <2 x i32> <i32 255, i32 255>)
 ; CHECK-NEXT:    [[R1:%.*]] = call <2 x i32> @llvm.smax.v2i32(<2 x i32> [[SI_MIN]], <2 x i32> <i32 1, i32 1>)
-; CHECK-NEXT:    [[R:%.*]] = sitofp <2 x i32> [[R1]] to <2 x float>
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg <2 x i32> [[R1]] to <2 x float>
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %si_min_cmp = icmp sgt <2 x i32> %x, <i32 255, i32 255>
diff --git a/llvm/test/Transforms/InstCombine/fpcast.ll b/llvm/test/Transforms/InstCombine/fpcast.ll
index ac4b88fcddd7e..d2c932ba447e4 100644
--- a/llvm/test/Transforms/InstCombine/fpcast.ll
+++ b/llvm/test/Transforms/InstCombine/fpcast.ll
@@ -170,7 +170,7 @@ define half @sint_to_fptrunc(i32 %x) {
 define half @masked_sint_to_fptrunc1(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fptrunc1(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 16777215
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[M]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %m = and i32 %x, 16777215
@@ -182,7 +182,7 @@ define half @masked_sint_to_fptrunc1(i32 %x) {
 define half @masked_sint_to_fptrunc2(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fptrunc2(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 8
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[M]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %m = lshr i32 %x, 8
@@ -194,7 +194,7 @@ define half @masked_sint_to_fptrunc2(i32 %x) {
 define half @masked_sint_to_fptrunc3(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fptrunc3(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 7
-; CHECK-NEXT:    [[F:%.*]] = sitofp i32 [[M]] to float
+; CHECK-NEXT:    [[F:%.*]] = uitofp nneg i32 [[M]] to float
 ; CHECK-NEXT:    [[R:%.*]] = fptrunc float [[F]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -218,7 +218,7 @@ define double @sint_to_fpext(i32 %x) {
 define double @masked_sint_to_fpext1(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fpext1(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 16777215
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[M]] to double
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
   %m = and i32 %x, 16777215
@@ -230,7 +230,7 @@ define double @masked_sint_to_fpext1(i32 %x) {
 define double @masked_sint_to_fpext2(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fpext2(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 8
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[M]] to double
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
   %m = lshr i32 %x, 8
@@ -242,7 +242,7 @@ define double @masked_sint_to_fpext2(i32 %x) {
 define double @masked_sint_to_fpext3(i32 %x) {
 ; CHECK-LABEL: @masked_sint_to_fpext3(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 7
-; CHECK-NEXT:    [[F:%.*]] = sitofp i32 [[M]] to float
+; CHECK-NEXT:    [[F:%.*]] = uitofp nneg i32 [[M]] to float
 ; CHECK-NEXT:    [[R:%.*]] = fpext float [[F]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
@@ -266,7 +266,7 @@ define half @uint_to_fptrunc(i32 %x) {
 define half @masked_uint_to_fptrunc1(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fptrunc1(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 16777215
-; CHECK-NEXT:    [[R:%.*]] = uitofp i32 [[M]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %m = and i32 %x, 16777215
@@ -278,7 +278,7 @@ define half @masked_uint_to_fptrunc1(i32 %x) {
 define half @masked_uint_to_fptrunc2(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fptrunc2(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 8
-; CHECK-NEXT:    [[R:%.*]] = uitofp i32 [[M]] to half
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
   %m = lshr i32 %x, 8
@@ -290,7 +290,7 @@ define half @masked_uint_to_fptrunc2(i32 %x) {
 define half @masked_uint_to_fptrunc3(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fptrunc3(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 7
-; CHECK-NEXT:    [[F:%.*]] = uitofp i32 [[M]] to float
+; CHECK-NEXT:    [[F:%.*]] = uitofp nneg i32 [[M]] to float
 ; CHECK-NEXT:    [[R:%.*]] = fptrunc float [[F]] to half
 ; CHECK-NEXT:    ret half [[R]]
 ;
@@ -314,7 +314,7 @@ define double @uint_to_fpext(i32 %x) {
 define double @masked_uint_to_fpext1(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fpext1(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[X:%.*]], 16777215
-; CHECK-NEXT:    [[R:%.*]] = uitofp i32 [[M]] to double
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
   %m = and i32 %x, 16777215
@@ -326,7 +326,7 @@ define double @masked_uint_to_fpext1(i32 %x) {
 define double @masked_uint_to_fpext2(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fpext2(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 8
-; CHECK-NEXT:    [[R:%.*]] = uitofp i32 [[M]] to double
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[M]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
   %m = lshr i32 %x, 8
@@ -338,7 +338,7 @@ define double @masked_uint_to_fpext2(i32 %x) {
 define double @masked_uint_to_fpext3(i32 %x) {
 ; CHECK-LABEL: @masked_uint_to_fpext3(
 ; CHECK-NEXT:    [[M:%.*]] = lshr i32 [[X:%.*]], 7
-; CHECK-NEXT:    [[F:%.*]] = uitofp i32 [[M]] to float
+; CHECK-NEXT:    [[F:%.*]] = uitofp nneg i32 [[M]] to float
 ; CHECK-NEXT:    [[R:%.*]] = fpext float [[F]] to double
 ; CHECK-NEXT:    ret double [[R]]
 ;
diff --git a/llvm/test/Transforms/InstCombine/minmax-fold.ll b/llvm/test/Transforms/InstCombine/minmax-fold.ll
index 8391fe33eb9b5..bbbbf9eb6eafe 100644
--- a/llvm/test/Transforms/InstCombine/minmax-fold.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-fold.ll
@@ -131,7 +131,7 @@ define i64 @t9(i32 %a) {
 define float @t10(i32 %x) {
 ; CHECK-LABEL: @t10(
 ; CHECK-NEXT:    [[R1:%.*]] = call i32 @llvm.smax.i32(i32 [[X:%.*]], i32 255)
-; CHECK-NEXT:    [[R:%.*]] = sitofp i32 [[R1]] to float
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i32 [[R1]] to float
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %f_x = sitofp i32 %x to float
@@ -143,7 +143,7 @@ define float @t10(i32 %x) {
 define float @t11(i64 %x) {
 ; CHECK-LABEL: @t11(
 ; CHECK-NEXT:    [[R1:%.*]] = call i64 @llvm.smax.i64(i64 [[X:%.*]], i64 255)
-; CHECK-NEXT:    [[R:%.*]] = sitofp i64 [[R1]] to float
+; CHECK-NEXT:    [[R:%.*]] = uitofp nneg i64 [[R1]] to float
 ; CHECK-NEXT:    ret float [[R]]
 ;
   %f_x = sitofp i64 %x to float
@@ -526,7 +526,7 @@ falselabel:
 define double @PR31751_umin1(i32 %x) {
 ; CHECK-LABEL: @PR31751_umin1(
 ; CHECK-NEXT:    [[SEL:%.*]] = call i32 @llvm.umin.i32(i32 [[X:%.*]], i32 2147483647)
-; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    [[CONV:%.*]] = uitofp nneg i32 [[SEL]] to double
 ; CHECK-NEXT:    ret double [[CONV]]
 ;
   %cmp = icmp slt i32 %x, 0
@@ -538,7 +538,7 @@ define double @PR31751_umin1(i32 %x) {
 define double @PR31751_umin2(i32 %x) {
 ; CHECK-LABEL: @PR31751_umin2(
 ; CHECK-NEXT:    [[SEL:%.*]] = call i32 @llvm.umin.i32(i32 [[X:%.*]], i32 2147483647)
-; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    [[CONV:%.*]] = uitofp nneg i32 [[SEL]] to double
 ; CHECK-NEXT:    ret double [[CONV]]
 ;
   %cmp = icmp ult i32 %x, 2147483647
@@ -550,7 +550,7 @@ define double @PR31751_umin2(i32 %x) {
 define double @PR31751_umin3(i32 %x) {
 ; CHECK-LABEL: @PR31751_umin3(
 ; CHECK-NEXT:    [[SEL:%.*]] = call i32 @llvm.umin.i32(i32 [[X:%.*]], i32 2147483647)
-; CHECK-NEXT:    [[CONV:%.*]] = sitofp i32 [[SEL]] to double
+; CHECK-NEXT:    [[CONV:%.*]] = uitofp nneg i32 [[SEL]] to double
 ; CHECK-NEXT:    ret double [[CONV]]
 ;
   %cmp = icmp ugt i32 %x, 2147483647
diff --git a/llvm/test/Transforms/InstCombine/minmax-fp.ll b/llvm/test/Transforms/InstCombine/minmax-fp.ll
index f89e8a18e6344..b9e46caa63753 100644
--- a/llvm/test/Transforms/InstCombine/minmax-fp.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-fp.ll
@@ -257,7 +257,7 @@ define double @t16(i32 %x) {
 define double @t17(i32 %x) {
 ; CHECK-LABEL: @t17(
 ; CHECK-NEXT:    [[SEL1:%.*]] = call i32 @llvm.smax.i32(i32 [[X:%.*]], i32 2)
-; CHECK-NEXT:    [[SEL:%.*]] = sitofp i32 [[SEL1]] to double
+; CHECK-NEXT:    [[SEL:%.*]] = uitofp nneg i32 [[SEL1]] to double
 ; CHECK-NEXT:    ret double [[SEL]]
 ;
   %cmp = icmp sgt i32 %x, 2
diff --git a/llvm/test/Transforms/InstCombine/pr27236.ll b/llvm/test/Transforms/InstCombine/pr27236.ll
index 61ea344b1bdbd..67c320d352466 100644
--- a/llvm/test/Transforms/InstCombine/pr27236.ll
+++ b/llvm/test/Transforms/InstCombine/pr27236.ll
@@ -4,7 +4,7 @@
 define float @test1(i32 %scale) {
 ; CHECK-LABEL: @test1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @llvm.smax.i32(i32 [[SCALE:%.*]], i32 1)
-; CHECK-NEXT:    [[TMP2:%.*]] = sitofp i32 [[TMP1]] to float
+; CHECK-NEXT:    [[TMP2:%.*]] = uitofp nneg i32 [[TMP1]] to float
 ; CHECK-NEXT:    ret float [[TMP2]]
 ;
   %1 = icmp sgt i32 1, %scale
diff --git a/llvm/test/Transforms/InstCombine/sitofp.ll b/llvm/test/Transforms/InstCombine/sitofp.ll
index cc6b6425eb03c..51eff39cd900e 100644
--- a/llvm/test/Transforms/InstCombine/sitofp.ll
+++ b/llvm/test/Transforms/InstCombine/sitofp.ll
@@ -256,7 +256,7 @@ define i25 @consider_lowbits_masked_input(i25 %A) {
 define i32 @overflow_masked_input(i32 %A) {
 ; CHECK-LABEL: @overflow_masked_input(
 ; CHECK-NEXT:    [[M:%.*]] = and i32 [[A:%.*]], 16777217
-; CHECK-NEXT:    [[B:%.*]] = uitofp i32 [[M]] to float
+; CHECK-NEXT:    [[B:%.*]] = uitofp nneg i32 [[M]] to float
 ; CHECK-NEXT:    [[C:%.*]] = fptoui float [[B]] to i32
 ; CHECK-NEXT:    ret i32 [[C]]
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll b/llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll
index c55e732c90147..59b8ce42380d9 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/float-induction-x86.ll
@@ -25,7 +25,7 @@ define void @fp_iv_loop1(ptr noalias nocapture %A, i32 %N) #0 {
 ; AUTO_VEC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; AUTO_VEC:       vector.ph:
 ; AUTO_VEC-NEXT:    [[N_VEC:%.*]] = and i64 [[ZEXT]], 2147483616
-; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; AUTO_VEC-NEXT:    [[TMP0:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; AUTO_VEC-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP0]], 1.000000e+00
 ; AUTO_VEC-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -201,7 +201,7 @@ define double @external_use_with_fast_math(ptr %a, i64 %n) {
 ; AUTO_VEC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; AUTO_VEC:       vector.ph:
 ; AUTO_VEC-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775792
-; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to double
+; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to double
 ; AUTO_VEC-NEXT:    [[TMP0:%.*]] = fmul fast double [[DOTCAST]], 3.000000e+00
 ; AUTO_VEC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; AUTO_VEC:       vector.body:
@@ -366,7 +366,7 @@ define void @fadd_reassoc_FMF(ptr nocapture %p, i32 %N) {
 ; AUTO_VEC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; AUTO_VEC:       vector.ph:
 ; AUTO_VEC-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 4294967264
-; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; AUTO_VEC-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; AUTO_VEC-NEXT:    [[TMP1:%.*]] = fmul reassoc float [[DOTCAST]], 4.200000e+01
 ; AUTO_VEC-NEXT:    [[IND_END:%.*]] = fadd reassoc float [[TMP1]], 1.000000e+00
 ; AUTO_VEC-NEXT:    br label [[VECTOR_BODY:%.*]]
diff --git a/llvm/test/Transforms/LoopVectorize/float-induction.ll b/llvm/test/Transforms/LoopVectorize/float-induction.ll
index caea114e3d448..bd658c31768a8 100644
--- a/llvm/test/Transforms/LoopVectorize/float-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/float-induction.ll
@@ -29,7 +29,7 @@ define void @fp_iv_loop1_fast_FMF(float %init, ptr noalias nocapture %A, i32 %N)
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
 ; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP1]]
 ; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -84,7 +84,7 @@ define void @fp_iv_loop1_fast_FMF(float %init, ptr noalias nocapture %A, i32 %N)
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483640
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
 ; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP1]]
 ; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -142,7 +142,7 @@ define void @fp_iv_loop1_fast_FMF(float %init, ptr noalias nocapture %A, i32 %N)
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP1]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -193,7 +193,7 @@ define void @fp_iv_loop1_fast_FMF(float %init, ptr noalias nocapture %A, i32 %N)
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = fmul fast float [[FPINC]], [[DOTCAST]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fsub fast float [[INIT:%.*]], [[TMP1]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
@@ -276,7 +276,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, ptr noalias nocapture %A, i32
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = fmul reassoc float [[FPINC]], [[DOTCAST]]
 ; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fsub reassoc float [[INIT:%.*]], [[TMP1]]
 ; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -331,7 +331,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, ptr noalias nocapture %A, i32
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483640
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = fmul reassoc float [[FPINC]], [[DOTCAST]]
 ; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fsub reassoc float [[INIT:%.*]], [[TMP1]]
 ; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -389,7 +389,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, ptr noalias nocapture %A, i32
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = fmul reassoc float [[FPINC]], [[DOTCAST]]
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fsub reassoc float [[INIT:%.*]], [[TMP1]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -442,7 +442,7 @@ define void @fp_iv_loop1_reassoc_FMF(float %init, ptr noalias nocapture %A, i32
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = fmul reassoc float [[FPINC]], [[DOTCAST]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fsub reassoc float [[INIT:%.*]], [[TMP1]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
@@ -526,7 +526,7 @@ define void @fp_iv_loop2(float %init, ptr noalias nocapture %A, i32 %N) #0 {
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], [[INIT:%.*]]
 ; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -574,7 +574,7 @@ define void @fp_iv_loop2(float %init, ptr noalias nocapture %A, i32 %N) #0 {
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483640
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], [[INIT:%.*]]
 ; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -625,7 +625,7 @@ define void @fp_iv_loop2(float %init, ptr noalias nocapture %A, i32 %N) #0 {
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], [[INIT:%.*]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -675,7 +675,7 @@ define void @fp_iv_loop2(float %init, ptr noalias nocapture %A, i32 %N) #0 {
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], [[INIT:%.*]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
@@ -758,10 +758,10 @@ define void @fp_iv_loop3(float %init, ptr noalias nocapture %A, ptr noalias noca
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP1]], 2147483644
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP2:%.*]] = fmul fast float [[DOTCAST]], -5.000000e-01
 ; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP2]], 0x3FB99999A0000000
-; VEC4_INTERL1-NEXT:    [[DOTCAST2:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST2:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP3:%.*]] = fmul fast float [[TMP0]], [[DOTCAST2]]
 ; VEC4_INTERL1-NEXT:    [[IND_END3:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
 ; VEC4_INTERL1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -835,10 +835,10 @@ define void @fp_iv_loop3(float %init, ptr noalias nocapture %A, ptr noalias noca
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP1]], 2147483640
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = fmul fast float [[DOTCAST]], -5.000000e-01
 ; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP2]], 0x3FB99999A0000000
-; VEC4_INTERL2-NEXT:    [[DOTCAST2:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST2:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[TMP0]], [[DOTCAST2]]
 ; VEC4_INTERL2-NEXT:    [[IND_END3:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
 ; VEC4_INTERL2-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[INIT]], i64 0
@@ -922,10 +922,10 @@ define void @fp_iv_loop3(float %init, ptr noalias nocapture %A, ptr noalias noca
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP1]], 2147483646
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP2:%.*]] = fmul fast float [[DOTCAST]], -5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP2]], 0x3FB99999A0000000
-; VEC1_INTERL2-NEXT:    [[DOTCAST2:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST2:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = fmul fast float [[TMP0]], [[DOTCAST2]]
 ; VEC1_INTERL2-NEXT:    [[IND_END3:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1000,10 +1000,10 @@ define void @fp_iv_loop3(float %init, ptr noalias nocapture %A, ptr noalias noca
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP1]], 2147483646
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP2:%.*]] = fmul fast float [[DOTCAST]], -5.000000e-01
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP2]], 0x3FB99999A0000000
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST2:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST2:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP3:%.*]] = fmul fast float [[TMP0]], [[DOTCAST2]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END3:%.*]] = fadd fast float [[TMP3]], [[INIT:%.*]]
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[INIT]], i64 0
@@ -1113,7 +1113,7 @@ define void @fp_iv_loop4(ptr noalias nocapture %A, i32 %N) {
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483644
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC4_INTERL1-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
 ; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1158,7 +1158,7 @@ define void @fp_iv_loop4(ptr noalias nocapture %A, i32 %N) {
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483640
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC4_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
 ; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1206,7 +1206,7 @@ define void @fp_iv_loop4(ptr noalias nocapture %A, i32 %N) {
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC1_INTERL2-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1256,7 +1256,7 @@ define void @fp_iv_loop4(ptr noalias nocapture %A, i32 %N) {
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[TMP0]], 2147483646
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[TMP1:%.*]] = fmul fast float [[DOTCAST]], 5.000000e-01
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[IND_END:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1319,7 +1319,7 @@ define void @non_primary_iv_float_scalar(ptr %A, i64 %N) {
 ; VEC4_INTERL1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL1:       vector.ph:
 ; VEC4_INTERL1-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
-; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL1-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL1-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL1:       vector.body:
 ; VEC4_INTERL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE8:%.*]] ]
@@ -1396,7 +1396,7 @@ define void @non_primary_iv_float_scalar(ptr %A, i64 %N) {
 ; VEC4_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC4_INTERL2:       vector.ph:
 ; VEC4_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
-; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC4_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC4_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC4_INTERL2:       vector.body:
 ; VEC4_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE17:%.*]] ]
@@ -1512,7 +1512,7 @@ define void @non_primary_iv_float_scalar(ptr %A, i64 %N) {
 ; VEC1_INTERL2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC1_INTERL2:       vector.ph:
 ; VEC1_INTERL2-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
-; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC1_INTERL2-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC1_INTERL2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC1_INTERL2:       vector.body:
 ; VEC1_INTERL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]
@@ -1570,7 +1570,7 @@ define void @non_primary_iv_float_scalar(ptr %A, i64 %N) {
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[FOR_BODY:%.*]], label [[VECTOR_PH:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.ph:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
-; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = sitofp i64 [[N_VEC]] to float
+; VEC2_INTERL1_PRED_STORE-NEXT:    [[DOTCAST:%.*]] = uitofp nneg i64 [[N_VEC]] to float
 ; VEC2_INTERL1_PRED_STORE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VEC2_INTERL1_PRED_STORE:       vector.body:
 ; VEC2_INTERL1_PRED_STORE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]

From 885b8d9bb5192267cb2449a9ddec28e20ac9300e Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Tue, 16 Apr 2024 13:16:17 -0700
Subject: [PATCH 662/886] [RISCV] Enable mul strength reduction for XTheadBa

This vendor extension has the same shift_add as zba, and most of the same
patterns are duplicated.  Enable it here too so the configurations don't
diverge.
---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp |  2 +-
 llvm/test/CodeGen/RISCV/rv64xtheadba.ll     | 17 +++++++++++++++++
 2 files changed, 18 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index dc7c6f83b9857..7b4bec2f65b74 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -13408,7 +13408,7 @@ static SDValue expandMul(SDNode *N, SelectionDAG &DAG,
   if (VT != Subtarget.getXLenVT())
     return SDValue();
 
-  if (!Subtarget.hasStdExtZba())
+  if (!Subtarget.hasStdExtZba() && !Subtarget.hasVendorXTHeadBa())
     return SDValue();
 
   ConstantSDNode *CNode = dyn_cast<ConstantSDNode>(N->getOperand(1));
diff --git a/llvm/test/CodeGen/RISCV/rv64xtheadba.ll b/llvm/test/CodeGen/RISCV/rv64xtheadba.ll
index 6f56babf28f5e..1450c86c76d05 100644
--- a/llvm/test/CodeGen/RISCV/rv64xtheadba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64xtheadba.ll
@@ -268,6 +268,23 @@ define i64 @mul96(i64 %a) {
   ret i64 %c
 }
 
+define i64 @mul137(i64 %a) {
+; RV64I-LABEL: mul137:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    li a1, 137
+; RV64I-NEXT:    mul a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64XTHEADBA-LABEL: mul137:
+; RV64XTHEADBA:       # %bb.0:
+; RV64XTHEADBA-NEXT:    th.addsl a1, a0, a0, 3
+; RV64XTHEADBA-NEXT:    slli a0, a0, 7
+; RV64XTHEADBA-NEXT:    add a0, a0, a1
+; RV64XTHEADBA-NEXT:    ret
+  %c = mul i64 %a, 137
+  ret i64 %c
+}
+
 define i64 @mul160(i64 %a) {
 ; RV64I-LABEL: mul160:
 ; RV64I:       # %bb.0:

From c6e01627acf8591830ee1d211cff4d5388095f3d Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Tue, 16 Apr 2024 20:35:35 +0000
Subject: [PATCH 663/886] Revert "Reapply "[LV] Improve AnyOf reduction
 codegen. (#78304)""

This reverts commit c6e38b928c56f562aea68a8e90f02dbdf0eada85.

Causes miscompiles, see comments on #78304.
---
 .../include/llvm/Transforms/Utils/LoopUtils.h |   9 +
 llvm/lib/Transforms/Utils/LoopUtils.cpp       |  24 +-
 .../Vectorize/LoopVectorizationPlanner.h      |   4 +-
 .../Transforms/Vectorize/LoopVectorize.cpp    |  76 +-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  13 +-
 .../LoopVectorize/AArch64/sve-select-cmp.ll   |  88 +--
 .../RISCV/select-cmp-reduction.ll             | 742 +++++++++++++-----
 .../epilog-vectorization-any-of-reductions.ll | 149 +---
 .../LoopVectorize/select-cmp-predicated.ll    | 210 +++--
 .../Transforms/LoopVectorize/select-cmp.ll    | 146 ++--
 ...tion-start-value-may-be-undef-or-poison.ll |  43 +-
 11 files changed, 904 insertions(+), 600 deletions(-)

diff --git a/llvm/include/llvm/Transforms/Utils/LoopUtils.h b/llvm/include/llvm/Transforms/Utils/LoopUtils.h
index 345e09dce0b2b..187ace3a0cbed 100644
--- a/llvm/include/llvm/Transforms/Utils/LoopUtils.h
+++ b/llvm/include/llvm/Transforms/Utils/LoopUtils.h
@@ -372,6 +372,15 @@ RecurKind getMinMaxReductionRecurKind(Intrinsic::ID RdxID);
 /// Returns the comparison predicate used when expanding a min/max reduction.
 CmpInst::Predicate getMinMaxReductionPredicate(RecurKind RK);
 
+/// See RecurrenceDescriptor::isAnyOfPattern for a description of the pattern we
+/// are trying to match. In this pattern, we are only ever selecting between two
+/// values: 1) an initial start value \p StartVal of the reduction PHI, and 2) a
+/// loop invariant value. If any of lane value in \p Left, \p Right is not equal
+/// to \p StartVal, select the loop invariant value. This is done by selecting
+/// \p Right iff \p Left is equal to \p StartVal.
+Value *createAnyOfOp(IRBuilderBase &Builder, Value *StartVal, RecurKind RK,
+                     Value *Left, Value *Right);
+
 /// Returns a Min/Max operation corresponding to MinMaxRecurrenceKind.
 /// The Builder's fast-math-flags must be set to propagate the expected values.
 Value *createMinMaxOp(IRBuilderBase &Builder, RecurKind RK, Value *Left,
diff --git a/llvm/lib/Transforms/Utils/LoopUtils.cpp b/llvm/lib/Transforms/Utils/LoopUtils.cpp
index 9d816c5220532..73c5d63678229 100644
--- a/llvm/lib/Transforms/Utils/LoopUtils.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUtils.cpp
@@ -1034,6 +1034,15 @@ CmpInst::Predicate llvm::getMinMaxReductionPredicate(RecurKind RK) {
   }
 }
 
+Value *llvm::createAnyOfOp(IRBuilderBase &Builder, Value *StartVal,
+                           RecurKind RK, Value *Left, Value *Right) {
+  if (auto VTy = dyn_cast<VectorType>(Left->getType()))
+    StartVal = Builder.CreateVectorSplat(VTy->getElementCount(), StartVal);
+  Value *Cmp =
+      Builder.CreateCmp(CmpInst::ICMP_NE, Left, StartVal, "rdx.select.cmp");
+  return Builder.CreateSelect(Cmp, Left, Right, "rdx.select");
+}
+
 Value *llvm::createMinMaxOp(IRBuilderBase &Builder, RecurKind RK, Value *Left,
                             Value *Right) {
   Type *Ty = Left->getType();
@@ -1142,13 +1151,16 @@ Value *llvm::createAnyOfTargetReduction(IRBuilderBase &Builder, Value *Src,
     NewVal = SI->getTrueValue();
   }
 
+  // Create a splat vector with the new value and compare this to the vector
+  // we want to reduce.
+  ElementCount EC = cast<VectorType>(Src->getType())->getElementCount();
+  Value *Right = Builder.CreateVectorSplat(EC, InitVal);
+  Value *Cmp =
+      Builder.CreateCmp(CmpInst::ICMP_NE, Src, Right, "rdx.select.cmp");
+
   // If any predicate is true it means that we want to select the new value.
-  Value *AnyOf =
-      Src->getType()->isVectorTy() ? Builder.CreateOrReduce(Src) : Src;
-  // The compares in the loop may yield poison, which propagates through the
-  // bitwise ORs. Freeze it here before the condition is used.
-  AnyOf = Builder.CreateFreeze(AnyOf);
-  return Builder.CreateSelect(AnyOf, NewVal, InitVal, "rdx.select");
+  Cmp = Builder.CreateOrReduce(Cmp);
+  return Builder.CreateSelect(Cmp, NewVal, InitVal, "rdx.select");
 }
 
 Value *llvm::createSimpleTargetReduction(IRBuilderBase &Builder, Value *Src,
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
index ece2a34f180cb..ebca2d855a467 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
@@ -68,7 +68,9 @@ class VPBuilder {
 public:
   VPBuilder() = default;
   VPBuilder(VPBasicBlock *InsertBB) { setInsertPoint(InsertBB); }
-  VPBuilder(VPRecipeBase *InsertPt) { setInsertPoint(InsertPt); }
+  VPBuilder(VPRecipeBase *InsertPt) {
+    setInsertPoint(InsertPt->getParent(), InsertPt->getIterator());
+  }
 
   /// Clear the insertion point: created instructions will not be inserted into
   /// a block.
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 2057cab46135f..44885a95bd102 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -3051,8 +3051,9 @@ PHINode *InnerLoopVectorizer::createInductionResumeValue(
   }
 
   // Create phi nodes to merge from the  backedge-taken check block.
-  PHINode *BCResumeVal = PHINode::Create(OrigPhi->getType(), 3, "bc.resume.val",
-                                         LoopScalarPreHeader->getFirstNonPHI());
+  PHINode *BCResumeVal =
+      PHINode::Create(OrigPhi->getType(), 3, "bc.resume.val",
+                      LoopScalarPreHeader->getTerminator()->getIterator());
   // Copy original phi DL over to the new one.
   BCResumeVal->setDebugLoc(OrigPhi->getDebugLoc());
 
@@ -7450,6 +7451,7 @@ static void createAndCollectMergePhiForReduction(
   auto *PhiR = cast<VPReductionPHIRecipe>(RedResult->getOperand(0));
   const RecurrenceDescriptor &RdxDesc = PhiR->getRecurrenceDescriptor();
 
+  TrackingVH<Value> ReductionStartValue = RdxDesc.getRecurrenceStartValue();
   Value *FinalValue =
       State.get(RedResult, VPIteration(State.UF - 1, VPLane::getFirstLane()));
   auto *ResumePhi =
@@ -7474,7 +7476,7 @@ static void createAndCollectMergePhiForReduction(
       BCBlockPhi->addIncoming(ResumePhi->getIncomingValueForBlock(Incoming),
                               Incoming);
     else
-      BCBlockPhi->addIncoming(RdxDesc.getRecurrenceStartValue(), Incoming);
+      BCBlockPhi->addIncoming(ReductionStartValue, Incoming);
   }
 
   auto *OrigPhi = cast<PHINode>(PhiR->getUnderlyingValue());
@@ -7767,10 +7769,11 @@ EpilogueVectorizerEpilogueLoop::createEpilogueVectorizedLoopSkeleton(
 
   // Now, compare the remaining count and if there aren't enough iterations to
   // execute the vectorized epilogue skip to the scalar part.
-  LoopVectorPreHeader->setName("vec.epilog.ph");
-  BasicBlock *VecEpilogueIterationCountCheck =
-      SplitBlock(LoopVectorPreHeader, LoopVectorPreHeader->begin(), DT, LI,
-                 nullptr, "vec.epilog.iter.check", true);
+  BasicBlock *VecEpilogueIterationCountCheck = LoopVectorPreHeader;
+  VecEpilogueIterationCountCheck->setName("vec.epilog.iter.check");
+  LoopVectorPreHeader =
+      SplitBlock(LoopVectorPreHeader, LoopVectorPreHeader->getTerminator(), DT,
+                 LI, nullptr, "vec.epilog.ph");
   emitMinimumVectorEpilogueIterCountCheck(LoopScalarPreHeader,
                                           VecEpilogueIterationCountCheck);
 
@@ -8893,10 +8896,6 @@ VPlanPtr LoopVectorizationPlanner::buildVPlan(VFRange &Range) {
 // A ComputeReductionResult recipe is added to the middle block, also for
 // in-loop reductions which compute their result in-loop, because generating
 // the subsequent bc.merge.rdx phi is driven by ComputeReductionResult recipes.
-//
-// Adjust AnyOf reductions; replace the reduction phi for the selected value
-// with a boolean reduction phi node to check if the condition is true in any
-// iteration. The final value is selected by the final ComputeReductionResult.
 void LoopVectorizationPlanner::adjustRecipesForReductions(
     VPBasicBlock *LatchVPBB, VPlanPtr &Plan, VPRecipeBuilder &RecipeBuilder,
     ElementCount MinVF) {
@@ -9071,41 +9070,6 @@ void LoopVectorizationPlanner::adjustRecipesForReductions(
       continue;
 
     const RecurrenceDescriptor &RdxDesc = PhiR->getRecurrenceDescriptor();
-    // Adjust AnyOf reductions; replace the reduction phi for the selected value
-    // with a boolean reduction phi node to check if the condition is true in
-    // any iteration. The final value is selected by the final
-    // ComputeReductionResult.
-    if (RecurrenceDescriptor::isAnyOfRecurrenceKind(
-            RdxDesc.getRecurrenceKind())) {
-      auto *Select = cast<VPRecipeBase>(*find_if(PhiR->users(), [](VPUser *U) {
-        return isa<VPWidenSelectRecipe>(U) ||
-               (isa<VPReplicateRecipe>(U) &&
-                cast<VPReplicateRecipe>(U)->getUnderlyingInstr()->getOpcode() ==
-                    Instruction::Select);
-      }));
-      VPValue *Cmp = Select->getOperand(0);
-      // If the compare is checking the reduction PHI node, adjust it to check
-      // the start value.
-      if (VPRecipeBase *CmpR = Cmp->getDefiningRecipe()) {
-        for (unsigned I = 0; I != CmpR->getNumOperands(); ++I)
-          if (CmpR->getOperand(I) == PhiR)
-            CmpR->setOperand(I, PhiR->getStartValue());
-      }
-      VPBuilder::InsertPointGuard Guard(Builder);
-      Builder.setInsertPoint(Select);
-
-      // If the true value of the select is the reduction phi, the new value is
-      // selected if the negated condition is true in any iteration.
-      if (Select->getOperand(1) == PhiR)
-        Cmp = Builder.createNot(Cmp);
-      VPValue *Or = Builder.createOr(PhiR, Cmp);
-      Select->getVPSingleValue()->replaceAllUsesWith(Or);
-
-      // Convert the reduction phi to operate on bools.
-      PhiR->setOperand(0, Plan->getOrAddLiveIn(ConstantInt::getFalse(
-                              OrigLoop->getHeader()->getContext())));
-    }
-
     // If tail is folded by masking, introduce selects between the phi
     // and the live-out instruction of each reduction, at the beginning of the
     // dedicated latch block.
@@ -9138,9 +9102,7 @@ void LoopVectorizationPlanner::adjustRecipesForReductions(
     // then extend the loop exit value to enable InstCombine to evaluate the
     // entire expression in the smaller type.
     Type *PhiTy = PhiR->getStartValue()->getLiveInIRValue()->getType();
-    if (MinVF.isVector() && PhiTy != RdxDesc.getRecurrenceType() &&
-        !RecurrenceDescriptor::isAnyOfRecurrenceKind(
-            RdxDesc.getRecurrenceKind())) {
+    if (MinVF.isVector() && PhiTy != RdxDesc.getRecurrenceType()) {
       assert(!PhiR->isInLoop() && "Unexpected truncated inloop reduction!");
       Type *RdxTy = RdxDesc.getRecurrenceType();
       auto *Trunc =
@@ -10181,19 +10143,9 @@ bool LoopVectorizePass::processLoop(Loop *L) {
           Value *ResumeV = nullptr;
           // TODO: Move setting of resume values to prepareToExecute.
           if (auto *ReductionPhi = dyn_cast<VPReductionPHIRecipe>(&R)) {
-            const RecurrenceDescriptor &RdxDesc =
-                ReductionPhi->getRecurrenceDescriptor();
-            RecurKind RK = RdxDesc.getRecurrenceKind();
-            ResumeV = ReductionResumeValues.find(&RdxDesc)->second;
-            if (RecurrenceDescriptor::isAnyOfRecurrenceKind(RK)) {
-              // VPReductionPHIRecipes for AnyOf reductions expect a boolean as
-              // start value; compare the final value from the main vector loop
-              // to the start value.
-              IRBuilder<> Builder(
-                  cast<Instruction>(ResumeV)->getParent()->getFirstNonPHI());
-              ResumeV = Builder.CreateICmpNE(ResumeV,
-                                             RdxDesc.getRecurrenceStartValue());
-            }
+            ResumeV = ReductionResumeValues
+                          .find(&ReductionPhi->getRecurrenceDescriptor())
+                          ->second;
           } else {
             // Create induction resume values for both widened pointer and
             // integer/fp inductions and update the start value of the induction
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 625319954e9b7..9f242a1bee8f6 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -501,8 +501,6 @@ Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) {
     // Reduce all of the unrolled parts into a single vector.
     Value *ReducedPartRdx = RdxParts[0];
     unsigned Op = RecurrenceDescriptor::getOpcode(RK);
-    if (RecurrenceDescriptor::isAnyOfRecurrenceKind(RK))
-      Op = Instruction::Or;
 
     if (PhiR->isOrdered()) {
       ReducedPartRdx = RdxParts[State.UF - 1];
@@ -515,16 +513,19 @@ Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) {
         if (Op != Instruction::ICmp && Op != Instruction::FCmp)
           ReducedPartRdx = Builder.CreateBinOp(
               (Instruction::BinaryOps)Op, RdxPart, ReducedPartRdx, "bin.rdx");
-        else
+        else if (RecurrenceDescriptor::isAnyOfRecurrenceKind(RK)) {
+          TrackingVH<Value> ReductionStartValue =
+              RdxDesc.getRecurrenceStartValue();
+          ReducedPartRdx = createAnyOfOp(Builder, ReductionStartValue, RK,
+                                         ReducedPartRdx, RdxPart);
+        } else
           ReducedPartRdx = createMinMaxOp(Builder, RK, ReducedPartRdx, RdxPart);
       }
     }
 
     // Create the reduction after the loop. Note that inloop reductions create
     // the target reduction in the loop using a Reduction recipe.
-    if ((State.VF.isVector() ||
-         RecurrenceDescriptor::isAnyOfRecurrenceKind(RK)) &&
-        !PhiR->isInLoop()) {
+    if (State.VF.isVector() && !PhiR->isInLoop()) {
       ReducedPartRdx =
           createTargetReduction(Builder, RdxDesc, ReducedPartRdx, OrigPhi);
       // If the reduction can be performed in a smaller type, we need to extend
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-select-cmp.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-select-cmp.ll
index 2470bca1e17b9..1c26ee8479e57 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-select-cmp.ll
@@ -8,41 +8,39 @@ target triple = "aarch64-linux-gnu"
 define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; CHECK-VF4IC1-LABEL: @select_const_i32_from_icmp
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <vscale x 4 x i32>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <vscale x 4 x i32> [[VEC_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[NOT:%*]] = xor <vscale x 4 x i1> [[VEC_ICMP]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <vscale x 4 x i1> [[VEC_ICMP]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR]], i32 7, i32 3
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 7, i32 3
 
 ; CHECK-VF4IC4-LABEL: @select_const_i32_from_icmp
 ; CHECK-VF4IC4:      vector.body:
-; CHECK-VF4IC4:        [[VEC_PHI1:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
-; CHECK-VF4IC4:        [[VEC_PHI2:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
-; CHECK-VF4IC4:        [[VEC_PHI3:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
-; CHECK-VF4IC4:        [[VEC_PHI4:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
+; CHECK-VF4IC4:        [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
+; CHECK-VF4IC4:        [[VEC_PHI2:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
+; CHECK-VF4IC4:        [[VEC_PHI3:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
+; CHECK-VF4IC4:        [[VEC_PHI4:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
 ; CHECK-VF4IC4:        [[VEC_ICMP1:%.*]] = icmp eq <vscale x 4 x i32> {{.*}}, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP2:%.*]] = icmp eq <vscale x 4 x i32> {{.*}}, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP3:%.*]] = icmp eq <vscale x 4 x i32> {{.*}}, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP4:%.*]] = icmp eq <vscale x 4 x i32> {{.*}}, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC4-NEXT:    [[NOT1:%.*]] = xor <vscale x 4 x i1> [[VEC_ICMP1]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC4-NEXT:    [[NOT2:%.*]] = xor <vscale x 4 x i1> [[VEC_ICMP2]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC4-NEXT:    [[NOT3:%.*]] = xor <vscale x 4 x i1> [[VEC_ICMP3]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC4-NEXT:    [[NOT4:%.*]] = xor <vscale x 4 x i1> [[VEC_ICMP4]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL1:%.*]] = or <vscale x 4 x i1> [[VEC_PHI1]], [[NOT1]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL2:%.*]] = or <vscale x 4 x i1> [[VEC_PHI2]], [[NOT2]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL3:%.*]] = or <vscale x 4 x i1> [[VEC_PHI3]], [[NOT3]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL4:%.*]] = or <vscale x 4 x i1> [[VEC_PHI4]], [[NOT4]]
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL1]] = select <vscale x 4 x i1> [[VEC_ICMP1]], <vscale x 4 x i32> [[VEC_PHI1]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL2]] = select <vscale x 4 x i1> [[VEC_ICMP2]], <vscale x 4 x i32> [[VEC_PHI2]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL3]] = select <vscale x 4 x i1> [[VEC_ICMP3]], <vscale x 4 x i32> [[VEC_PHI3]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL4]] = select <vscale x 4 x i1> [[VEC_ICMP4]], <vscale x 4 x i32> [[VEC_PHI4]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC4:      middle.block:
-; CHECK-VF4IC4-NEXT:   [[OR1:%.*]] = or <vscale x 4 x i1> [[VEC_SEL2]], [[VEC_SEL1]]
-; CHECK-VF4IC4-NEXT:   [[OR2:%.*]] = or <vscale x 4 x i1> [[VEC_SEL3]], [[OR1]]
-; CHECK-VF4IC4-NEXT:   [[OR3:%.*]] = or <vscale x 4 x i1> [[VEC_SEL4]], [[OR2]]
-; CHECK-VF4IC4-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[OR3]])
-; CHECK-VF4IC4-NEXT:   [[FR:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC4-NEXT:   {{.*}} = select i1 [[FR]], i32 7, i32 3
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP5:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL1]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL5:%.*]] = select <vscale x 4 x i1> [[VEC_ICMP5]], <vscale x 4 x i32> [[VEC_SEL1]], <vscale x 4 x i32> [[VEC_SEL2]]
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP6:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL5]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL6:%.*]] = select <vscale x 4 x i1> [[VEC_ICMP6]], <vscale x 4 x i32> [[VEC_SEL5]], <vscale x 4 x i32> [[VEC_SEL3]]
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP7:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL6]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL7:%.*]] = select <vscale x 4 x i1> [[VEC_ICMP7]], <vscale x 4 x i32> [[VEC_SEL6]], <vscale x 4 x i32> [[VEC_SEL4]]
+; CHECK-VF4IC4-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL7]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC4-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC4-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 7, i32 3
 entry:
   br label %for.body
 
@@ -64,18 +62,21 @@ exit:                                     ; preds = %for.body
 define i32 @select_i32_from_icmp(ptr nocapture readonly %v, i32 %a, i32 %b, i64 %n) #0 {
 ; CHECK-VF4IC1-LABEL: @select_i32_from_icmp
 ; CHECK-VF4IC1:      vector.ph:
-; CHECK-VF4IC1-NOT:    shufflevector <vscale x 4 x i32>
-; CHECK-VF4IC1-NOT:    shufflevector <vscale x 4 x i32>
+; CHECK-VF4IC1:        [[TMP1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 %a, i64 0
+; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_A:%.*]] = shufflevector <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[TMP2:%.*]] = insertelement <vscale x 4 x i32> poison, i32 %b, i64 0
+; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_B:%.*]] = shufflevector <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[SPLAT_OF_A]], %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <vscale x 4 x i32>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <vscale x 4 x i32> [[VEC_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[NOT:%*]] = xor <vscale x 4 x i1> [[VEC_ICMP]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <vscale x 4 x i1> [[VEC_ICMP]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[SPLAT_OF_B]]
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR]], i32 %b, i32 %a
+; CHECK-VF4IC1-NEXT:   [[FIN_INS:%.*]] = insertelement <vscale x 4 x i32> poison, i32 %a, i64 0
+; CHECK-VF4IC1-NEXT:   [[FIN_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[FIN_INS]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[FIN_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL]], [[FIN_SPLAT]]
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[FIN_CMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 %b, i32 %a
 
 ; CHECK-VF4IC4-LABEL: @select_i32_from_icmp
 ; CHECK-VF4IC4:      vector.body:
@@ -100,15 +101,14 @@ exit:                                     ; preds = %for.body
 define i32 @select_const_i32_from_fcmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; CHECK-VF4IC1-LABEL: @select_const_i32_from_fcmp
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <vscale x 4 x float>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = fcmp fast ueq <vscale x 4 x float> [[VEC_LOAD]], shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[NOT:%*]] = xor <vscale x 4 x i1> [[VEC_ICMP]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <vscale x 4 x i1> [[VEC_ICMP]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR]], i32 1, i32 2
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 1, i32 2
 
 ; CHECK-VF4IC4-LABEL: @select_const_i32_from_fcmp
 ; CHECK-VF4IC4:      vector.body:
@@ -156,17 +156,17 @@ exit:                                     ; preds = %for.body
 define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1, ptr noalias nocapture readonly %src2, i64 %n) #0 {
 ; CHECK-VF4IC1-LABEL: @pred_select_const_i32_from_icmp
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <vscale x 4 x i32>
 ; CHECK-VF4IC1:        [[MASK:%.*]] = icmp sgt <vscale x 4 x i32> [[VEC_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 35, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-VF4IC1:        [[MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr {{%.*}}, i32 4, <vscale x 4 x i1> [[MASK]], <vscale x 4 x i32> poison)
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <vscale x 4 x i32> [[MASKED_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL_TMP:%.*]] = or <vscale x 4 x i1> [[VEC_PHI]], [[VEC_ICMP]]
-; CHECK-VF4IC1:        [[VEC_SEL:%.*]] = select <vscale x 4 x i1> [[MASK]], <vscale x 4 x i1> [[VEC_SEL_TMP]], <vscale x 4 x i1> [[VEC_PHI]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL_TMP:%.*]] = select <vscale x 4 x i1> [[VEC_ICMP]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> [[VEC_PHI]]
+; CHECK-VF4IC1:        [[VEC_SEL:%.*]] = select <vscale x 4 x i1> [[MASK]], <vscale x 4 x i32> [[VEC_SEL_TMP]], <vscale x 4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR]], i32 1, i32 0
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <vscale x 4 x i32> [[VEC_SEL]], zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 1, i32 0
 
 ; CHECK-VF4IC4-LABEL: @pred_select_const_i32_from_icmp
 ; CHECK-VF4IC4:      vector.body:
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
index 2b58acbfe9cc9..8a2dc0abb0de8 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S \
 ; RUN:   < %s | FileCheck %s
 ; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 \
@@ -6,59 +7,109 @@
 target triple = "riscv64"
 
 define i32 @select_icmp(i32 %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0 {
-; CHECK-LABEL: @select_icmp
+; CHECK-LABEL: define i32 @select_icmp(
+; CHECK-SAME: i32 [[X:%.*]], i32 [[Y:%.*]], ptr nocapture readonly [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i64 0
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[NOT:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %y, i32 0
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[TMP4]], zeroinitializer
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i32 [[Y]], i32 0
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[A:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[COND:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP7]], [[X]]
+; CHECK-NEXT:    [[COND]] = select i1 [[CMP1]], i32 [[A]], i32 [[Y]]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[COND_LCSSA:%.*]] = phi i32 [ [[COND]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[COND_LCSSA]]
 ;
-; SCALABLE-LABEL: @select_icmp
+; SCALABLE-LABEL: define i32 @select_icmp(
+; SCALABLE-SAME: i32 [[X:%.*]], i32 [[Y:%.*]], ptr nocapture readonly [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
-; SCALABLE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; SCALABLE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
-; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X:%.*]], i64 0
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X]], i64 0
 ; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[Y]], i64 0
+; SCALABLE-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[TMP4]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
-; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; SCALABLE-NEXT:    [[NOT:%.*]] = xor <vscale x 4 x i1> [[TMP8]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
-; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
-; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; SCALABLE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP6]]
+; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; SCALABLE-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
+; SCALABLE-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[BROADCAST_SPLAT2]]
+; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; SCALABLE-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP9]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP13]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %y, i32 0
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[TMP10]], zeroinitializer
+; SCALABLE-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP12]], i32 [[Y]], i32 0
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[A:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[COND:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDVARS_IV]]
+; SCALABLE-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; SCALABLE-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP13]], [[X]]
+; SCALABLE-NEXT:    [[COND]] = select i1 [[CMP1]], i32 [[A]], i32 [[Y]]
+; SCALABLE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; SCALABLE:       for.end:
+; SCALABLE-NEXT:    [[COND_LCSSA:%.*]] = phi i32 [ [[COND]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[COND_LCSSA]]
 ;
 entry:
   br label %for.body
@@ -79,59 +130,109 @@ for.end:
 }
 
 define i32 @select_fcmp(float %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0 {
-; CHECK-LABEL: @select_fcmp
+; CHECK-LABEL: define i32 @select_fcmp(
+; CHECK-SAME: float [[X:%.*]], i32 [[Y:%.*]], ptr nocapture readonly [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[X:%.*]], i64 0
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[X]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[C]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = fcmp fast olt <4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[NOT:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast olt <4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %y, i32 0
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[TMP4]], zeroinitializer
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i32 [[Y]], i32 0
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[A:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[COND:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[C]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[TMP7:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[CMP1:%.*]] = fcmp fast olt float [[TMP7]], [[X]]
+; CHECK-NEXT:    [[COND]] = select i1 [[CMP1]], i32 [[A]], i32 [[Y]]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[COND_LCSSA:%.*]] = phi i32 [ [[COND]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[COND_LCSSA]]
 ;
-; SCALABLE-LABEL: @select_fcmp
+; SCALABLE-LABEL: define i32 @select_fcmp(
+; SCALABLE-SAME: float [[X:%.*]], i32 [[Y:%.*]], ptr nocapture readonly [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
-; SCALABLE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; SCALABLE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
-; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[X:%.*]], i64 0
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x float> poison, float [[X]], i64 0
 ; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x float> [[BROADCAST_SPLATINSERT]], <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[Y]], i64 0
+; SCALABLE-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i64 [[TMP4]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
-; SCALABLE-NEXT:    [[TMP8:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; SCALABLE-NEXT:    [[NOT:%.*]] = xor <vscale x 4 x i1> [[TMP8]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
-; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
-; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; SCALABLE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[C]], i64 [[TMP6]]
+; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; SCALABLE-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
+; SCALABLE-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[BROADCAST_SPLAT2]]
+; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; SCALABLE-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP9]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP13]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %y, i32 0
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[TMP10]], zeroinitializer
+; SCALABLE-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP12]], i32 [[Y]], i32 0
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[A:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[COND:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[C]], i64 [[INDVARS_IV]]
+; SCALABLE-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; SCALABLE-NEXT:    [[CMP1:%.*]] = fcmp fast olt float [[TMP13]], [[X]]
+; SCALABLE-NEXT:    [[COND]] = select i1 [[CMP1]], i32 [[A]], i32 [[Y]]
+; SCALABLE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
+; SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; SCALABLE:       for.end:
+; SCALABLE-NEXT:    [[COND_LCSSA:%.*]] = phi i32 [ [[COND]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[COND_LCSSA]]
 ;
 entry:
   br label %for.body
@@ -152,55 +253,101 @@ for.end:
 }
 
 define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) #0 {
-; CHECK-LABEL: @select_const_i32_from_icmp
+; CHECK-LABEL: define i32 @select_const_i32_from_icmp(
+; CHECK-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-NEXT:    [[NOT:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP4]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 7, i32 3
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[TMP5]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP7]], i32 7, i32 3
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 3, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP15:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP12:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 3
+; CHECK-NEXT:    [[TMP12]] = select i1 [[TMP11]], i32 [[TMP8]], i32 7
+; CHECK-NEXT:    [[TMP13]] = add nuw nsw i64 [[TMP15]], 1
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], [[N]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP12]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
 ;
-; SCALABLE-LABEL: @select_const_i32_from_icmp
+; SCALABLE-LABEL: define i32 @select_const_i32_from_icmp(
+; SCALABLE-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[TMP4]]
+; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP4]]
 ; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
 ; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[NOT:%.*]] = xor <vscale x 4 x i1> [[TMP8]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; SCALABLE-NEXT:    [[TMP9]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 7, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP9]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP13]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 7, i32 3
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[TMP9]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP13]], i32 7, i32 3
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 3, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[TMP21:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP14:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP21]]
+; SCALABLE-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
+; SCALABLE-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 3
+; SCALABLE-NEXT:    [[TMP18]] = select i1 [[TMP17]], i32 [[TMP14]], i32 7
+; SCALABLE-NEXT:    [[TMP19]] = add nuw nsw i64 [[TMP21]], 1
+; SCALABLE-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[TMP19]], [[N]]
+; SCALABLE-NEXT:    br i1 [[TMP20]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; SCALABLE:       exit:
+; SCALABLE-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP18]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[DOTLCSSA]]
 ;
 entry:
   br label %for.body
@@ -221,55 +368,113 @@ exit:                                     ; preds = %for.body
 }
 
 define i32 @select_i32_from_icmp(ptr nocapture readonly %v, i32 %a, i32 %b, i64 %n) #0 {
-; CHECK-LABEL: @select_i32_from_icmp
+; CHECK-LABEL: define i32 @select_i32_from_icmp(
+; CHECK-SAME: ptr nocapture readonly [[V:%.*]], i32 [[A:%.*]], i32 [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[A]], i64 0
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <4 x i32> [[MINMAX_IDENT_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[B]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-NEXT:    [[NOT:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP4]], <4 x i32> [[VEC_PHI]], <4 x i32> [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %b, i32 %a
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[A]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[TMP5]], [[DOTSPLAT]]
+; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP7]], i32 [[B]], i32 [[A]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP15:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP12:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 3
+; CHECK-NEXT:    [[TMP12]] = select i1 [[TMP11]], i32 [[TMP8]], i32 [[B]]
+; CHECK-NEXT:    [[TMP13]] = add nuw nsw i64 [[TMP15]], 1
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], [[N]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP12]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
 ;
-; SCALABLE-LABEL: @select_i32_from_icmp
+; SCALABLE-LABEL: define i32 @select_i32_from_icmp(
+; SCALABLE-SAME: ptr nocapture readonly [[V:%.*]], i32 [[A:%.*]], i32 [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
+; SCALABLE-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[A]], i64 0
+; SCALABLE-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[MINMAX_IDENT_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[B]], i64 0
+; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[TMP4]]
+; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP4]]
 ; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
 ; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 3, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[NOT:%.*]] = xor <vscale x 4 x i1> [[TMP8]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; SCALABLE-NEXT:    [[TMP9]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[BROADCAST_SPLAT]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP9]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP13]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 %b, i32 %a
+; SCALABLE-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[A]], i64 0
+; SCALABLE-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[DOTSPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[TMP9]], [[DOTSPLAT]]
+; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP13]], i32 [[B]], i32 [[A]]
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[TMP21:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP14:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP21]]
+; SCALABLE-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
+; SCALABLE-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 3
+; SCALABLE-NEXT:    [[TMP18]] = select i1 [[TMP17]], i32 [[TMP14]], i32 [[B]]
+; SCALABLE-NEXT:    [[TMP19]] = add nuw nsw i64 [[TMP21]], 1
+; SCALABLE-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[TMP19]], [[N]]
+; SCALABLE-NEXT:    br i1 [[TMP20]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; SCALABLE:       exit:
+; SCALABLE-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP18]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[DOTLCSSA]]
 ;
 entry:
   br label %for.body
@@ -290,55 +495,101 @@ exit:                                     ; preds = %for.body
 }
 
 define i32 @select_const_i32_from_fcmp(ptr nocapture readonly %v, i64 %n) #0 {
-; CHECK-LABEL: @select_const_i32_from_fcmp
+; CHECK-LABEL: define i32 @select_const_i32_from_fcmp(
+; CHECK-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 2, i32 2, i32 2, i32 2>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp fast ueq <4 x float> [[WIDE_LOAD]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>
-; CHECK-NEXT:    [[NOT:%.*]] = xor <4 x i1> [[TMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP4]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 1, i32 2
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[TMP5]], <i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP7]], i32 1, i32 2
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 2, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP15:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP12:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load float, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = fcmp fast ueq float [[TMP10]], 3.000000e+00
+; CHECK-NEXT:    [[TMP12]] = select i1 [[TMP11]], i32 [[TMP8]], i32 1
+; CHECK-NEXT:    [[TMP13]] = add nuw nsw i64 [[TMP15]], 1
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], [[N]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP12]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[DOTLCSSA]]
 ;
-; SCALABLE-LABEL: @select_const_i32_from_fcmp
+; SCALABLE-LABEL: define i32 @select_const_i32_from_fcmp(
+; SCALABLE-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[V:%.*]], i64 [[TMP4]]
+; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[TMP4]]
 ; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
 ; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = fcmp fast ueq <vscale x 4 x float> [[WIDE_LOAD]], shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[NOT:%.*]] = xor <vscale x 4 x i1> [[TMP8]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[NOT]]
+; SCALABLE-NEXT:    [[TMP9]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[TMP9]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP13]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 1, i32 2
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[TMP9]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; SCALABLE-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP13]], i32 1, i32 2
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 2, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[TMP21:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP14:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[TMP21]]
+; SCALABLE-NEXT:    [[TMP16:%.*]] = load float, ptr [[TMP15]], align 4
+; SCALABLE-NEXT:    [[TMP17:%.*]] = fcmp fast ueq float [[TMP16]], 3.000000e+00
+; SCALABLE-NEXT:    [[TMP18]] = select i1 [[TMP17]], i32 [[TMP14]], i32 1
+; SCALABLE-NEXT:    [[TMP19]] = add nuw nsw i64 [[TMP21]], 1
+; SCALABLE-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[TMP19]], [[N]]
+; SCALABLE-NEXT:    br i1 [[TMP20]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; SCALABLE:       exit:
+; SCALABLE-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP18]], [[FOR_BODY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[DOTLCSSA]]
 ;
 entry:
   br label %for.body
@@ -359,11 +610,41 @@ exit:                                     ; preds = %for.body
 }
 
 define float @select_const_f32_from_icmp(ptr nocapture readonly %v, i64 %n) #0 {
-; CHECK-LABEL: @select_const_f32_from_icmp
-; CHECK-NOT: vector.body
+; CHECK-LABEL: define float @select_const_f32_from_icmp(
+; CHECK-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi fast float [ 3.000000e+00, [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 3
+; CHECK-NEXT:    [[TMP5]] = select fast i1 [[TMP4]], float [[TMP1]], float 7.000000e+00
+; CHECK-NEXT:    [[TMP6]] = add nuw nsw i64 [[TMP0]], 1
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[TMP6]], [[N]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[EXIT:%.*]], label [[FOR_BODY]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi float [ [[TMP5]], [[FOR_BODY]] ]
+; CHECK-NEXT:    ret float [[DOTLCSSA]]
 ;
-; SCALABLE-LABEL: @select_const_f32_from_icmp
-; SCALABLE-NOT: vector.body
+; SCALABLE-LABEL: define float @select_const_f32_from_icmp(
+; SCALABLE-SAME: ptr nocapture readonly [[V:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP1:%.*]] = phi fast float [ 3.000000e+00, [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[TMP0]]
+; SCALABLE-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
+; SCALABLE-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 3
+; SCALABLE-NEXT:    [[TMP5]] = select fast i1 [[TMP4]], float [[TMP1]], float 7.000000e+00
+; SCALABLE-NEXT:    [[TMP6]] = add nuw nsw i64 [[TMP0]], 1
+; SCALABLE-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[TMP6]], [[N]]
+; SCALABLE-NEXT:    br i1 [[TMP7]], label [[EXIT:%.*]], label [[FOR_BODY]]
+; SCALABLE:       exit:
+; SCALABLE-NEXT:    [[DOTLCSSA:%.*]] = phi float [ [[TMP5]], [[FOR_BODY]] ]
+; SCALABLE-NEXT:    ret float [[DOTLCSSA]]
 ;
 entry:
   br label %for.body
@@ -384,63 +665,127 @@ exit:                                     ; preds = %for.body
 }
 
 define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1, ptr noalias nocapture readonly %src2, i64 %n) #0 {
-; CHECK-LABEL: @pred_select_const_i32_from_icmp
+; CHECK-LABEL: define i32 @pred_select_const_i32_from_icmp(
+; CHECK-SAME: ptr noalias nocapture readonly [[SRC1:%.*]], ptr noalias nocapture readonly [[SRC2:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1:%.*]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], <i32 35, i32 35, i32 35, i32 35>
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[SRC2:%.*]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP6]], i32 4, <4 x i1> [[TMP4]], <4 x i32> poison)
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[WIDE_MASKED_LOAD]], <i32 2, i32 2, i32 2, i32 2>
-; CHECK-NEXT:    [[TMP9:%.*]] = or <4 x i1> [[VEC_PHI]], [[TMP8]]
-; CHECK-NEXT:    [[PREDPHI]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP9]], <4 x i1> [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], <i32 35, i32 35, i32 35, i32 35>
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[SRC2]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP5]], i32 4, <4 x i1> [[TMP3]], <4 x i32> poison)
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i32> [[WIDE_MASKED_LOAD]], <i32 2, i32 2, i32 2, i32 2>
+; CHECK-NEXT:    [[TMP7:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32> [[VEC_PHI]]
+; CHECK-NEXT:    [[PREDPHI]] = select <4 x i1> [[TMP3]], <4 x i32> [[TMP7]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[PREDPHI]])
-; CHECK-NEXT:    [[FR:%.*]] = freeze i1 [[TMP12]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 1, i32 0
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i32> [[PREDPHI]], zeroinitializer
+; CHECK-NEXT:    [[TMP9:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP9]], i32 1, i32 0
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I_013:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], [[FOR_INC]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[I_013]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP10]], 35
+; CHECK-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[I_013]]
+; CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
+; CHECK-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP11]], 2
+; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = select i1 [[CMP3]], i32 1, i32 [[R_012]]
+; CHECK-NEXT:    br label [[FOR_INC]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[R_1]] = phi i32 [ [[R_012]], [[FOR_BODY]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
+; CHECK-NEXT:    [[INC]] = add nuw nsw i64 [[I_013]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK:       for.end.loopexit:
+; CHECK-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], [[FOR_INC]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i32 [[R_1_LCSSA]]
 ;
-; SCALABLE-LABEL: @pred_select_const_i32_from_icmp
+; SCALABLE-LABEL: define i32 @pred_select_const_i32_from_icmp(
+; SCALABLE-SAME: ptr noalias nocapture readonly [[SRC1:%.*]], ptr noalias nocapture readonly [[SRC2:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; SCALABLE-NEXT:  entry:
+; SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
+; SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SCALABLE:       vector.ph:
 ; SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; SCALABLE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
-; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 %n, [[TMP3]]
-; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 %n, [[N_MOD_VF]]
-; SCALABLE-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; SCALABLE-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 4
+; SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; SCALABLE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; SCALABLE-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
 ; SCALABLE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SCALABLE:       vector.body:
-; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
-; SCALABLE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[SRC1:%.*]], i64 [[TMP4]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
-; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 35, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[SRC2:%.*]], i64 [[TMP4]]
-; SCALABLE-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 0
-; SCALABLE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> poison)
+; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
+; SCALABLE-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP6]]
+; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; SCALABLE-NEXT:    [[TMP9:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 35, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+; SCALABLE-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[SRC2]], i64 [[TMP6]]
+; SCALABLE-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0
+; SCALABLE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> poison)
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 2, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
-; SCALABLE-NEXT:    [[TMP13:%.*]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP12]]
-; SCALABLE-NEXT:    [[PREDPHI]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> [[VEC_PHI]]
-; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP16]]
-; SCALABLE-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; SCALABLE-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; SCALABLE-NEXT:    [[TMP13:%.*]] = select <vscale x 4 x i1> [[TMP12]], <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x i32> [[VEC_PHI]]
+; SCALABLE-NEXT:    [[PREDPHI]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[TMP13]], <vscale x 4 x i32> [[VEC_PHI]]
+; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; SCALABLE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; SCALABLE:       middle.block:
-; SCALABLE-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[PREDPHI]])
-; SCALABLE-NEXT:    [[FR:%.*]] = freeze i1 [[TMP18]]
-; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR]], i32 1, i32 0
+; SCALABLE-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <vscale x 4 x i32> [[PREDPHI]], zeroinitializer
+; SCALABLE-NEXT:    [[TMP15:%.*]] = call i1 @llvm.vector.reduce.or.nxv4i1(<vscale x 4 x i1> [[RDX_SELECT_CMP]])
+; SCALABLE-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP15]], i32 1, i32 0
+; SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; SCALABLE-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; SCALABLE:       scalar.ph:
+; SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SCALABLE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    br label [[FOR_BODY:%.*]]
+; SCALABLE:       for.body:
+; SCALABLE-NEXT:    [[I_013:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; SCALABLE-NEXT:    [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], [[FOR_INC]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; SCALABLE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[I_013]]
+; SCALABLE-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; SCALABLE-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP16]], 35
+; SCALABLE-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; SCALABLE:       if.then:
+; SCALABLE-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[I_013]]
+; SCALABLE-NEXT:    [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
+; SCALABLE-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP17]], 2
+; SCALABLE-NEXT:    [[SPEC_SELECT:%.*]] = select i1 [[CMP3]], i32 1, i32 [[R_012]]
+; SCALABLE-NEXT:    br label [[FOR_INC]]
+; SCALABLE:       for.inc:
+; SCALABLE-NEXT:    [[R_1]] = phi i32 [ [[R_012]], [[FOR_BODY]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
+; SCALABLE-NEXT:    [[INC]] = add nuw nsw i64 [[I_013]], 1
+; SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
+; SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; SCALABLE:       for.end.loopexit:
+; SCALABLE-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], [[FOR_INC]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; SCALABLE-NEXT:    ret i32 [[R_1_LCSSA]]
 ;
 entry:
   br label %for.body
@@ -472,3 +817,34 @@ for.end.loopexit:                                 ; preds = %for.inc
 }
 
 attributes #0 = { "target-features"="+f,+v" }
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
+; CHECK: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; CHECK: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]}
+; CHECK: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]}
+; CHECK: [[LOOP11]] = distinct !{[[LOOP11]], [[META2]], [[META1]]}
+; CHECK: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]}
+; CHECK: [[LOOP13]] = distinct !{[[LOOP13]], [[META2]], [[META1]]}
+;.
+; SCALABLE: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; SCALABLE: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; SCALABLE: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; SCALABLE: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; SCALABLE: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; SCALABLE: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; SCALABLE: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; SCALABLE: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
+; SCALABLE: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; SCALABLE: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]}
+; SCALABLE: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]}
+; SCALABLE: [[LOOP11]] = distinct !{[[LOOP11]], [[META2]], [[META1]]}
+; SCALABLE: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]}
+; SCALABLE: [[LOOP13]] = distinct !{[[LOOP13]], [[META2]], [[META1]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
index c721da7597b1c..0b872709ec6c6 100644
--- a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
@@ -19,20 +19,20 @@ define i32 @any_of_reduction_epilog(ptr %src, i64 %N) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[TMP1]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
+; CHECK-NEXT:    [[TMP8]] = select <4 x i1> [[TMP4]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ne <4 x i32> [[TMP8]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[TMP8:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP8]], i32 1, i32 0
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP7]], i32 1, i32 0
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
 ; CHECK:       vec.epilog.iter.check:
@@ -42,33 +42,32 @@ define i32 @any_of_reduction_epilog(ptr %src, i64 %N) {
 ; CHECK:       vec.epilog.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ], [ [[RDX_SELECT]], [[VEC_EPILOG_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[BC_MERGE_RDX]], 0
 ; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[TMP0]], 4
 ; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF2]]
-; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP9]], i64 0
-; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <4 x i1> [[MINMAX_IDENT_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[BC_MERGE_RDX]], i64 0
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <4 x i32> [[MINMAX_IDENT_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i1> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i32> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP17:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX5]], 0
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[TMP10]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD7]], zeroinitializer
-; CHECK-NEXT:    [[TMP14]] = or <4 x i1> [[VEC_PHI6]], [[TMP13]]
+; CHECK-NEXT:    [[TMP17]] = select <4 x i1> [[TMP13]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32> [[VEC_PHI6]]
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 4
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP15]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp ne <4 x i32> [[TMP17]], zeroinitializer
 ; CHECK-NEXT:    [[TMP16:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP14]])
-; CHECK-NEXT:    [[TMP17:%.*]] = freeze i1 [[TMP16]]
-; CHECK-NEXT:    [[RDX_SELECT9:%.*]] = select i1 [[TMP17]], i32 1, i32 0
+; CHECK-NEXT:    [[RDX_SELECT9:%.*]] = select i1 [[TMP16]], i32 1, i32 0
 ; CHECK-NEXT:    [[CMP_N4:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N4]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       vec.epilog.scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX10:%.*]] = phi i32 [ 0, [[ITER_CHECK]] ], [ 0, [[VEC_EPILOG_ITER_CHECK]] ], [ [[RDX_SELECT9]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX10:%.*]] = phi i32 [ 0, [[ITER_CHECK]] ], [ [[RDX_SELECT]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[RDX_SELECT9]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -102,104 +101,6 @@ exit:
   ret i32 %select
 }
 
-define i32 @any_of_reduction_epilog_arg_as_start_value(ptr %src, i64 %N, i32 %start) {
-; CHECK-LABEL: define i32 @any_of_reduction_epilog_arg_as_start_value(
-; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]], i32 [[START:%.*]]) {
-; CHECK-NEXT:  iter.check:
-; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], 1
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 4
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
-; CHECK:       vector.main.loop.iter.check:
-; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[TMP0]], 4
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 4
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
-; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP5]])
-; CHECK-NEXT:    [[TMP8:%.*]] = freeze i1 [[TMP7]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP8]], i32 1, i32 [[START]]
-; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
-; CHECK:       vec.epilog.iter.check:
-; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[TMP0]], [[N_VEC]]
-; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
-; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
-; CHECK:       vec.epilog.ph:
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[START]], [[VECTOR_MAIN_LOOP_ITER_CHECK]] ], [ [[RDX_SELECT]], [[VEC_EPILOG_ITER_CHECK]] ]
-; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[BC_MERGE_RDX]], [[START]]
-; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[TMP0]], 4
-; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF2]]
-; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP9]], i64 0
-; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <4 x i1> [[MINMAX_IDENT_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
-; CHECK:       vec.epilog.vector.body:
-; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i1> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX5]], 0
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[TMP10]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
-; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD7]], zeroinitializer
-; CHECK-NEXT:    [[TMP14]] = or <4 x i1> [[VEC_PHI6]], [[TMP13]]
-; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 4
-; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP15]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
-; CHECK:       vec.epilog.middle.block:
-; CHECK-NEXT:    [[TMP16:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP14]])
-; CHECK-NEXT:    [[TMP17:%.*]] = freeze i1 [[TMP16]]
-; CHECK-NEXT:    [[RDX_SELECT9:%.*]] = select i1 [[TMP17]], i32 1, i32 [[START]]
-; CHECK-NEXT:    [[CMP_N4:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[CMP_N4]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
-; CHECK:       vec.epilog.scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX10:%.*]] = phi i32 [ [[START]], [[ITER_CHECK]] ], [ [[START]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[RDX_SELECT9]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    br label [[LOOP:%.*]]
-; CHECK:       loop:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
-; CHECK-NEXT:    [[RED:%.*]] = phi i32 [ [[BC_MERGE_RDX10]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[SELECT:%.*]], [[LOOP]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[LOAD:%.*]] = load i8, ptr [[GEP]], align 1
-; CHECK-NEXT:    [[ICMP:%.*]] = icmp eq i8 [[LOAD]], 0
-; CHECK-NEXT:    [[SELECT]] = select i1 [[ICMP]], i32 1, i32 [[RED]]
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; CHECK-NEXT:    [[ICMP3:%.*]] = icmp eq i64 [[IV]], [[N]]
-; CHECK-NEXT:    br i1 [[ICMP3]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
-; CHECK:       exit:
-; CHECK-NEXT:    [[SELECT_LCSSA:%.*]] = phi i32 [ [[SELECT]], [[LOOP]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ [[RDX_SELECT9]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[SELECT_LCSSA]]
-;
-entry:
-  br label %loop
-
-loop:
-  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
-  %red = phi i32 [ %start, %entry ], [ %select, %loop ]
-  %gep = getelementptr inbounds i8, ptr %src, i64 %iv
-  %load = load i8, ptr %gep, align 1
-  %icmp = icmp eq i8 %load, 0
-  %select = select i1 %icmp, i32 1, i32 %red
-  %iv.next = add i64 %iv, 1
-  %icmp3 = icmp eq i64 %iv, %N
-  br i1 %icmp3, label %exit, label %loop
-
-exit:
-  ret i32 %select
-}
 
 define i1 @any_of_reduction_i1_epilog(i64 %N, i32 %a) {
 ; CHECK-LABEL: define i1 @any_of_reduction_i1_epilog(
@@ -223,15 +124,14 @@ define i1 @any_of_reduction_i1_epilog(i64 %N, i32 %a) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i1> [[TMP1]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP3]] = or <4 x i1> [[VEC_PHI]], [[TMP2]]
+; CHECK-NEXT:    [[TMP3]] = select <4 x i1> [[TMP1]], <4 x i1> [[VEC_PHI]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]])
-; CHECK-NEXT:    [[TMP6:%.*]] = freeze i1 [[TMP5]]
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <4 x i1> [[TMP3]], zeroinitializer
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP]])
 ; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i1 false, i1 false
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -244,11 +144,10 @@ define i1 @any_of_reduction_i1_epilog(i64 %N, i32 %a) {
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i1 [ false, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ], [ [[RDX_SELECT]], [[VEC_EPILOG_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
 ; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp ne i1 [[BC_MERGE_RDX]], false
 ; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[TMP0]], 4
 ; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF2]]
 ; CHECK-NEXT:    [[IND_END5:%.*]] = trunc i64 [[N_VEC3]] to i32
-; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP7]], i64 0
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[BC_MERGE_RDX]], i64 0
 ; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <4 x i1> [[MINMAX_IDENT_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[BC_RESUME_VAL]], i64 0
 ; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
@@ -261,22 +160,21 @@ define i1 @any_of_reduction_i1_epilog(i64 %N, i32 %a) {
 ; CHECK-NEXT:    [[VEC_PHI10:%.*]] = phi <4 x i1> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND11:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT12:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[VEC_IND11]], [[BROADCAST_SPLAT14]]
-; CHECK-NEXT:    [[TMP9:%.*]] = xor <4 x i1> [[TMP8]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT:    [[TMP10]] = or <4 x i1> [[VEC_PHI10]], [[TMP9]]
+; CHECK-NEXT:    [[TMP10]] = select <4 x i1> [[TMP8]], <4 x i1> [[VEC_PHI10]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[INDEX_NEXT15]] = add nuw i64 [[INDEX9]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT12]] = add <4 x i32> [[VEC_IND11]], <i32 4, i32 4, i32 4, i32 4>
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT15]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP11]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
-; CHECK-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP10]])
-; CHECK-NEXT:    [[TMP13:%.*]] = freeze i1 [[TMP12]]
+; CHECK-NEXT:    [[RDX_SELECT_CMP16:%.*]] = icmp ne <4 x i1> [[TMP10]], zeroinitializer
+; CHECK-NEXT:    [[TMP13:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[RDX_SELECT_CMP16]])
 ; CHECK-NEXT:    [[RDX_SELECT16:%.*]] = select i1 [[TMP13]], i1 false, i1 false
 ; CHECK-NEXT:    [[CMP_N8:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N8]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       vec.epilog.scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL7:%.*]] = phi i32 [ [[IND_END5]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END6]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX17:%.*]] = phi i1 [ false, [[ITER_CHECK]] ], [ false, [[VEC_EPILOG_ITER_CHECK]] ], [ [[RDX_SELECT16]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX17:%.*]] = phi i1 [ false, [[ITER_CHECK]] ], [ [[RDX_SELECT]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[RDX_SELECT16]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL4]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -287,7 +185,7 @@ define i1 @any_of_reduction_i1_epilog(i64 %N, i32 %a) {
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add i32 [[IV_2]], 1
 ; CHECK-NEXT:    [[CMP_2:%.*]] = icmp eq i64 [[IV]], [[N]]
-; CHECK-NEXT:    br i1 [[CMP_2]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP_2]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i1 [ [[SEL]], [[LOOP]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ [[RDX_SELECT16]], [[VEC_EPILOG_MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    ret i1 [[SEL_LCSSA]]
@@ -321,7 +219,4 @@ exit:
 ; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]], [[META2]]}
 ; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
 ; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
-; CHECK: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
-; CHECK: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]], [[META2]]}
-; CHECK: [[LOOP10]] = distinct !{[[LOOP10]], [[META2]], [[META1]]}
 ;.
diff --git a/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll b/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
index 6a9f83a9e0aa2..1b4bcf6a3739a 100644
--- a/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
@@ -1,114 +1,155 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -S < %s | FileCheck %s --check-prefix=CHECK-VF2IC1
 ; RUN: opt -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=1 -S < %s | FileCheck %s --check-prefix=CHECK-VF1IC2
 
 define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1, ptr noalias nocapture readonly %src2, i64 %n) {
-; CHECK-VF2IC1-LABEL: @pred_select_const_i32_from_icmp(
+; CHECK-VF2IC1-LABEL: define i32 @pred_select_const_i32_from_icmp(
+; CHECK-VF2IC1-SAME: ptr noalias nocapture readonly [[SRC1:%.*]], ptr noalias nocapture readonly [[SRC2:%.*]], i64 [[N:%.*]]) {
+; CHECK-VF2IC1-NEXT:  entry:
+; CHECK-VF2IC1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
+; CHECK-VF2IC1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-VF2IC1:       vector.ph:
+; CHECK-VF2IC1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; CHECK-VF2IC1-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-VF2IC1-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-VF2IC1:       vector.body:
-; CHECK-VF2IC1:         [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, %vector.ph ], [ [[PREDPHI:%.*]], %pred.load.continue2 ]
-; CHECK-VF2IC1:         [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr {{%.*}}, align 4
-; CHECK-VF2IC1-NEXT:    [[TMP4:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], <i32 35, i32 35>
-; CHECK-VF2IC1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
-; CHECK-VF2IC1-NEXT:    br i1 [[TMP5]], label %pred.load.if, label %pred.load.continue
+; CHECK-VF2IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE2:%.*]] ]
+; CHECK-VF2IC1-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[PRED_LOAD_CONTINUE2]] ]
+; CHECK-VF2IC1-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-VF2IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP0]]
+; CHECK-VF2IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
+; CHECK-VF2IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF2IC1-NEXT:    [[TMP3:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], <i32 35, i32 35>
+; CHECK-VF2IC1-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
+; CHECK-VF2IC1-NEXT:    br i1 [[TMP4]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; CHECK-VF2IC1:       pred.load.if:
-; CHECK-VF2IC1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[SRC2:%.*]], i64 {{%.*}}
-; CHECK-VF2IC1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
-; CHECK-VF2IC1-NEXT:    [[TMP8:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0
-; CHECK-VF2IC1-NEXT:    br label %pred.load.continue
+; CHECK-VF2IC1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[TMP0]]
+; CHECK-VF2IC1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
+; CHECK-VF2IC1-NEXT:    [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 0
+; CHECK-VF2IC1-NEXT:    br label [[PRED_LOAD_CONTINUE]]
 ; CHECK-VF2IC1:       pred.load.continue:
-; CHECK-VF2IC1-NEXT:    [[TMP9:%.*]] = phi <2 x i32> [ poison, %vector.body ], [ [[TMP8]], %pred.load.if ]
-; CHECK-VF2IC1-NEXT:    [[TMP10:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
-; CHECK-VF2IC1-NEXT:    br i1 [[TMP10]], label %pred.load.if1, label %pred.load.continue2
+; CHECK-VF2IC1-NEXT:    [[TMP8:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_LOAD_IF]] ]
+; CHECK-VF2IC1-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
+; CHECK-VF2IC1-NEXT:    br i1 [[TMP9]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2]]
 ; CHECK-VF2IC1:       pred.load.if1:
-; CHECK-VF2IC1:         [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 {{%.*}}
-; CHECK-VF2IC1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
-; CHECK-VF2IC1-NEXT:    [[TMP14:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP13]], i32 1
-; CHECK-VF2IC1-NEXT:    br label %pred.load.continue2
+; CHECK-VF2IC1-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 1
+; CHECK-VF2IC1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[TMP10]]
+; CHECK-VF2IC1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+; CHECK-VF2IC1-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP12]], i32 1
+; CHECK-VF2IC1-NEXT:    br label [[PRED_LOAD_CONTINUE2]]
 ; CHECK-VF2IC1:       pred.load.continue2:
-; CHECK-VF2IC1-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP9]], %pred.load.continue ], [ [[TMP14]], %pred.load.if1 ]
-; CHECK-VF2IC1-NEXT:    [[TMP16:%.*]] = icmp eq <2 x i32> [[TMP15]], <i32 2, i32 2>
-; CHECK-VF2IC1-NEXT:    [[TMP17:%.*]] = or <2 x i1> [[VEC_PHI]], [[TMP16]]
-; CHECK-VF2IC1-NEXT:    [[PREDPHI]] = select <2 x i1> [[TMP4]], <2 x i1> [[TMP17]], <2 x i1> [[VEC_PHI]]
-; CHECK-VF2IC1:         br i1 {{%.*}}, label %middle.block, label %vector.body
+; CHECK-VF2IC1-NEXT:    [[TMP14:%.*]] = phi <2 x i32> [ [[TMP8]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP13]], [[PRED_LOAD_IF1]] ]
+; CHECK-VF2IC1-NEXT:    [[TMP15:%.*]] = icmp eq <2 x i32> [[TMP14]], <i32 2, i32 2>
+; CHECK-VF2IC1-NEXT:    [[TMP16:%.*]] = select <2 x i1> [[TMP15]], <2 x i32> <i32 1, i32 1>, <2 x i32> [[VEC_PHI]]
+; CHECK-VF2IC1-NEXT:    [[PREDPHI]] = select <2 x i1> [[TMP3]], <2 x i32> [[TMP16]], <2 x i32> [[VEC_PHI]]
+; CHECK-VF2IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-VF2IC1-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-VF2IC1-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-VF2IC1:       middle.block:
-; CHECK-VF2IC1-NEXT:    [[TMP20:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[PREDPHI]])
-; CHECK-VF2IC1-NEXT:    [[FR_TMP20:%.*]] = freeze i1 [[TMP20]]
-; CHECK-VF2IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR_TMP20]], i32 1, i32 0
+; CHECK-VF2IC1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <2 x i32> [[PREDPHI]], zeroinitializer
+; CHECK-VF2IC1-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[RDX_SELECT_CMP]])
+; CHECK-VF2IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP18]], i32 1, i32 0
+; CHECK-VF2IC1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-VF2IC1-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-VF2IC1:       scalar.ph:
-; CHECK-VF2IC1:         [[BC_RESUME_VAL:%.*]] = phi i64 [ {{%.*}}, %middle.block ], [ 0, %entry ]
-; CHECK-VF2IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %entry ], [ [[RDX_SELECT]], %middle.block ]
-; CHECK-VF2IC1-NEXT:    br label %for.body
+; CHECK-VF2IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-VF2IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-VF2IC1-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-VF2IC1:       for.body:
-; CHECK-VF2IC1:         [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], %for.inc ], [ [[BC_MERGE_RDX]], %scalar.ph ]
-; CHECK-VF2IC1:         [[TMP21:%.*]] = load i32, ptr {{%.*}}, align 4
-; CHECK-VF2IC1-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP21]], 35
-; CHECK-VF2IC1-NEXT:    br i1 [[CMP1]], label %if.then, label %for.inc
+; CHECK-VF2IC1-NEXT:    [[I_013:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-VF2IC1-NEXT:    [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], [[FOR_INC]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; CHECK-VF2IC1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[I_013]]
+; CHECK-VF2IC1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-VF2IC1-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP19]], 35
+; CHECK-VF2IC1-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-VF2IC1:       if.then:
-; CHECK-VF2IC1:         [[TMP22:%.*]] = load i32, ptr {{%.*}}, align 4
-; CHECK-VF2IC1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP22]], 2
+; CHECK-VF2IC1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[I_013]]
+; CHECK-VF2IC1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
+; CHECK-VF2IC1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP20]], 2
 ; CHECK-VF2IC1-NEXT:    [[SPEC_SELECT:%.*]] = select i1 [[CMP3]], i32 1, i32 [[R_012]]
-; CHECK-VF2IC1-NEXT:    br label %for.inc
+; CHECK-VF2IC1-NEXT:    br label [[FOR_INC]]
 ; CHECK-VF2IC1:       for.inc:
-; CHECK-VF2IC1-NEXT:    [[R_1]] = phi i32 [ [[R_012]], %for.body ], [ [[SPEC_SELECT]], %if.then ]
+; CHECK-VF2IC1-NEXT:    [[R_1]] = phi i32 [ [[R_012]], [[FOR_BODY]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
+; CHECK-VF2IC1-NEXT:    [[INC]] = add nuw nsw i64 [[I_013]], 1
+; CHECK-VF2IC1-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
+; CHECK-VF2IC1-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-VF2IC1:       for.end.loopexit:
-; CHECK-VF2IC1-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], %for.inc ], [ [[RDX_SELECT]], %middle.block ]
+; CHECK-VF2IC1-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], [[FOR_INC]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
 ; CHECK-VF2IC1-NEXT:    ret i32 [[R_1_LCSSA]]
 ;
-; CHECK-VF1IC2-LABEL: @pred_select_const_i32_from_icmp(
+; CHECK-VF1IC2-LABEL: define i32 @pred_select_const_i32_from_icmp(
+; CHECK-VF1IC2-SAME: ptr noalias nocapture readonly [[SRC1:%.*]], ptr noalias nocapture readonly [[SRC2:%.*]], i64 [[N:%.*]]) {
+; CHECK-VF1IC2-NEXT:  entry:
+; CHECK-VF1IC2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 2
+; CHECK-VF1IC2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-VF1IC2:       vector.ph:
+; CHECK-VF1IC2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
+; CHECK-VF1IC2-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-VF1IC2-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-VF1IC2:       vector.body:
-; CHECK-VF1IC2:         [[VEC_PHI:%.*]] = phi i1 [ false, %vector.ph ], [ [[PREDPHI:%.*]], %pred.load.continue3 ]
-; CHECK-VF1IC2-NEXT:    [[VEC_PHI2:%.*]] = phi i1 [ false, %vector.ph ], [ [[PREDPHI5:%.*]], %pred.load.continue3 ]
-; CHECK-VF1IC2:         [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC1:%.*]], i64 {{%.*}}
-; CHECK-VF1IC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 {{%.*}}
-; CHECK-VF1IC2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
-; CHECK-VF1IC2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP1]], align 4
-; CHECK-VF1IC2-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], 35
-; CHECK-VF1IC2-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], 35
-; CHECK-VF1IC2-NEXT:    br i1 [[TMP4]], label %pred.load.if, label %pred.load.continue
+; CHECK-VF1IC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE3:%.*]] ]
+; CHECK-VF1IC2-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[PRED_LOAD_CONTINUE3]] ]
+; CHECK-VF1IC2-NEXT:    [[VEC_PHI1:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[PREDPHI4:%.*]], [[PRED_LOAD_CONTINUE3]] ]
+; CHECK-VF1IC2-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
+; CHECK-VF1IC2-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
+; CHECK-VF1IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP0]]
+; CHECK-VF1IC2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP1]]
+; CHECK-VF1IC2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 4
+; CHECK-VF1IC2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4
+; CHECK-VF1IC2-NEXT:    [[TMP6:%.*]] = icmp sgt i32 [[TMP4]], 35
+; CHECK-VF1IC2-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], 35
+; CHECK-VF1IC2-NEXT:    br i1 [[TMP6]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; CHECK-VF1IC2:       pred.load.if:
-; CHECK-VF1IC2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[SRC2:%.*]], i64 {{%.*}}
-; CHECK-VF1IC2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
-; CHECK-VF1IC2-NEXT:    br label %pred.load.continue
+; CHECK-VF1IC2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[TMP0]]
+; CHECK-VF1IC2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
+; CHECK-VF1IC2-NEXT:    br label [[PRED_LOAD_CONTINUE]]
 ; CHECK-VF1IC2:       pred.load.continue:
-; CHECK-VF1IC2-NEXT:    [[TMP8:%.*]] = phi i32 [ poison, %vector.body ], [ [[TMP7]], %pred.load.if ]
-; CHECK-VF1IC2-NEXT:    br i1 [[TMP5]], label %pred.load.if2, label %pred.load.continue3
+; CHECK-VF1IC2-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_LOAD_IF]] ]
+; CHECK-VF1IC2-NEXT:    br i1 [[TMP7]], label [[PRED_LOAD_IF2:%.*]], label [[PRED_LOAD_CONTINUE3]]
 ; CHECK-VF1IC2:       pred.load.if2:
-; CHECK-VF1IC2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 {{%.*}}
-; CHECK-VF1IC2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
-; CHECK-VF1IC2-NEXT:    br label %pred.load.continue3
+; CHECK-VF1IC2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[TMP1]]
+; CHECK-VF1IC2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+; CHECK-VF1IC2-NEXT:    br label [[PRED_LOAD_CONTINUE3]]
 ; CHECK-VF1IC2:       pred.load.continue3:
-; CHECK-VF1IC2-NEXT:    [[TMP11:%.*]] = phi i32 [ poison, %pred.load.continue ], [ [[TMP10]], %pred.load.if2 ]
-; CHECK-VF1IC2-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[TMP8]], 2
-; CHECK-VF1IC2-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[TMP11]], 2
-; CHECK-VF1IC2-NEXT:    [[TMP14:%.*]] = or i1 [[VEC_PHI]], [[TMP12]]
-; CHECK-VF1IC2-NEXT:    [[TMP15:%.*]] = or i1 [[VEC_PHI2]], [[TMP13]]
-; CHECK-VF1IC2-NEXT:    [[PREDPHI]] = select i1 [[TMP4]], i1 [[TMP14]], i1 [[VEC_PHI]]
-; CHECK-VF1IC2-NEXT:    [[PREDPHI5]] = select i1 [[TMP5]], i1 [[TMP15]], i1 [[VEC_PHI2]]
-; CHECK-VF1IC2:         br i1 {{%.*}}, label %middle.block, label %vector.body
+; CHECK-VF1IC2-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[PRED_LOAD_CONTINUE]] ], [ [[TMP12]], [[PRED_LOAD_IF2]] ]
+; CHECK-VF1IC2-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[TMP10]], 2
+; CHECK-VF1IC2-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP13]], 2
+; CHECK-VF1IC2-NEXT:    [[TMP16:%.*]] = select i1 [[TMP14]], i32 1, i32 [[VEC_PHI]]
+; CHECK-VF1IC2-NEXT:    [[TMP17:%.*]] = select i1 [[TMP15]], i32 1, i32 [[VEC_PHI1]]
+; CHECK-VF1IC2-NEXT:    [[PREDPHI]] = select i1 [[TMP6]], i32 [[TMP16]], i32 [[VEC_PHI]]
+; CHECK-VF1IC2-NEXT:    [[PREDPHI4]] = select i1 [[TMP7]], i32 [[TMP17]], i32 [[VEC_PHI1]]
+; CHECK-VF1IC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-VF1IC2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-VF1IC2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-VF1IC2:       middle.block:
-; CHECK-VF1IC2-NEXT:    [[OR:%.*]] = or i1 [[PREDPHI5]], [[PREDPHI]]
-; CHECK-VF1IC2-NEXT:    [[FR_OR:%.*]] = freeze i1 [[OR]]
-; CHECK-VF1IC2-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR_OR]], i32 1, i32 0
-; CHECK-VF1IC2:         br i1 {{%.*}}, label %for.end.loopexit, label %scalar.ph
+; CHECK-VF1IC2-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[PREDPHI]], 0
+; CHECK-VF1IC2-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[PREDPHI]], i32 [[PREDPHI4]]
+; CHECK-VF1IC2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-VF1IC2-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-VF1IC2:       scalar.ph:
-; CHECK-VF1IC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ {{%.*}}, %middle.block ], [ 0, %entry ]
-; CHECK-VF1IC2-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %entry ], [ [[RDX_SELECT]], %middle.block ]
-; CHECK-VF1IC2-NEXT:    br label %for.body
+; CHECK-VF1IC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-VF1IC2-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
+; CHECK-VF1IC2-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-VF1IC2:       for.body:
-; CHECK-VF1IC2-NEXT:    [[I_013:%.*]] = phi i64 [ [[INC:%.*]], %for.inc ], [ [[BC_RESUME_VAL]], %scalar.ph ]
-; CHECK-VF1IC2-NEXT:    [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], %for.inc ], [ [[BC_MERGE_RDX]], %scalar.ph ]
-; CHECK-VF1IC2:         [[TMP19:%.*]] = load i32, ptr {{%.*}}, align 4
+; CHECK-VF1IC2-NEXT:    [[I_013:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-VF1IC2-NEXT:    [[R_012:%.*]] = phi i32 [ [[R_1:%.*]], [[FOR_INC]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; CHECK-VF1IC2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[I_013]]
+; CHECK-VF1IC2-NEXT:    [[TMP19:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 ; CHECK-VF1IC2-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[TMP19]], 35
-; CHECK-VF1IC2-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label %for.inc
+; CHECK-VF1IC2-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-VF1IC2:       if.then:
-; CHECK-VF1IC2:         [[TMP20:%.*]] = load i32, ptr {{%.*}}, align 4
+; CHECK-VF1IC2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[SRC2]], i64 [[I_013]]
+; CHECK-VF1IC2-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
 ; CHECK-VF1IC2-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP20]], 2
 ; CHECK-VF1IC2-NEXT:    [[SPEC_SELECT:%.*]] = select i1 [[CMP3]], i32 1, i32 [[R_012]]
-; CHECK-VF1IC2-NEXT:    br label %for.inc
+; CHECK-VF1IC2-NEXT:    br label [[FOR_INC]]
 ; CHECK-VF1IC2:       for.inc:
-; CHECK-VF1IC2-NEXT:    [[R_1]] = phi i32 [ [[R_012]], %for.body ], [ [[SPEC_SELECT]], %if.then ]
-; CHECK-VF1IC2:         br i1 {{%.*}}, label %for.end.loopexit, label %for.body
+; CHECK-VF1IC2-NEXT:    [[R_1]] = phi i32 [ [[R_012]], [[FOR_BODY]] ], [ [[SPEC_SELECT]], [[IF_THEN]] ]
+; CHECK-VF1IC2-NEXT:    [[INC]] = add nuw nsw i64 [[I_013]], 1
+; CHECK-VF1IC2-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N]]
+; CHECK-VF1IC2-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-VF1IC2:       for.end.loopexit:
-; CHECK-VF1IC2-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], %for.inc ], [ [[RDX_SELECT]], %middle.block ]
+; CHECK-VF1IC2-NEXT:    [[R_1_LCSSA:%.*]] = phi i32 [ [[R_1]], [[FOR_INC]] ], [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ]
 ; CHECK-VF1IC2-NEXT:    ret i32 [[R_1_LCSSA]]
 ;
 entry:
@@ -139,3 +180,14 @@ for.end.loopexit:                                 ; preds = %for.inc
   %r.1.lcssa = phi i32 [ %r.1, %for.inc ]
   ret i32 %r.1.lcssa
 }
+;.
+; CHECK-VF2IC1: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-VF2IC1: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-VF2IC1: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-VF2IC1: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+;.
+; CHECK-VF1IC2: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-VF1IC2: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-VF1IC2: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-VF1IC2: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/select-cmp.ll b/llvm/test/Transforms/LoopVectorize/select-cmp.ll
index 993b56a05207b..c9f2aaef6d5c8 100644
--- a/llvm/test/Transforms/LoopVectorize/select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-cmp.ll
@@ -5,47 +5,45 @@
 define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) {
 ; CHECK-LABEL: @select_const_i32_from_icmp
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <4 x i32>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <4 x i32> [[VEC_LOAD]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-VF4IC1-NEXT:   [[NOT:%.*]] = xor <4 x i1> [[VEC_ICMP]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_ICMP]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 7, i32 3
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 7, i32 3
 
 ; CHECK-VF4IC4:      vector.body:
-; CHECK-VF4IC4:        [[VEC_PHI1:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
-; CHECK-VF4IC4-NEXT:   [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
-; CHECK-VF4IC4-NEXT:   [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
-; CHECK-VF4IC4-NEXT:   [[VEC_PHI4:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
+; CHECK-VF4IC4:        [[VEC_PHI1:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
+; CHECK-VF4IC4-NEXT:   [[VEC_PHI2:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
+; CHECK-VF4IC4-NEXT:   [[VEC_PHI3:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
+; CHECK-VF4IC4-NEXT:   [[VEC_PHI4:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
 ; CHECK-VF4IC4:        [[VEC_ICMP1:%.*]] = icmp eq <4 x i32> {{.*}}, <i32 3, i32 3, i32 3, i32 3>
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP2:%.*]] = icmp eq <4 x i32> {{.*}}, <i32 3, i32 3, i32 3, i32 3>
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP3:%.*]] = icmp eq <4 x i32> {{.*}}, <i32 3, i32 3, i32 3, i32 3>
 ; CHECK-VF4IC4-NEXT:   [[VEC_ICMP4:%.*]] = icmp eq <4 x i32> {{.*}}, <i32 3, i32 3, i32 3, i32 3>
-; CHECK-VF4IC4-NEXT:   [[NOT1:%.*]] = xor <4 x i1> [[VEC_ICMP1]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC4-NEXT:   [[NOT2:%.*]] = xor <4 x i1> [[VEC_ICMP2]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC4-NEXT:   [[NOT3:%.*]] = xor <4 x i1> [[VEC_ICMP3]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC4-NEXT:   [[NOT4:%.*]] = xor <4 x i1> [[VEC_ICMP4]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL1:%.*]] = or <4 x i1> [[VEC_PHI1]], [[NOT1]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL2:%.*]] = or <4 x i1> [[VEC_PHI2]], [[NOT2]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL3:%.*]] = or <4 x i1> [[VEC_PHI3]], [[NOT3]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL4:%.*]] = or <4 x i1> [[VEC_PHI4]], [[NOT4]]
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL1:%.*]] = select <4 x i1> [[VEC_ICMP1]], <4 x i32> [[VEC_PHI1]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL2:%.*]] = select <4 x i1> [[VEC_ICMP2]], <4 x i32> [[VEC_PHI2]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL3:%.*]] = select <4 x i1> [[VEC_ICMP3]], <4 x i32> [[VEC_PHI3]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL4:%.*]] = select <4 x i1> [[VEC_ICMP4]], <4 x i32> [[VEC_PHI4]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>
 ; CHECK-VF4IC4:      middle.block:
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL5:%.*]] = or <4 x i1>  [[VEC_SEL2]], [[VEC_SEL1]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL6:%.*]] = or <4 x i1> [[VEC_SEL3]], [[VEC_SEL5]]
-; CHECK-VF4IC4-NEXT:   [[VEC_SEL7:%.*]] = or <4 x i1> [[VEC_SEL4]], [[VEC_SEL6]]
-; CHECK-VF4IC4-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL7]])
-; CHECK-VF4IC4-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC4-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 7, i32 3
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP5:%.*]] = icmp ne <4 x i32> [[VEC_SEL1]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL5:%.*]] = select <4 x i1> [[VEC_ICMP5]], <4 x i32> [[VEC_SEL1]], <4 x i32> [[VEC_SEL2]]
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP6:%.*]] = icmp ne <4 x i32> [[VEC_SEL5]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL6:%.*]] = select <4 x i1> [[VEC_ICMP6]], <4 x i32> [[VEC_SEL5]], <4 x i32> [[VEC_SEL3]]
+; CHECK-VF4IC4-NEXT:   [[VEC_ICMP7:%.*]] = icmp ne <4 x i32> [[VEC_SEL6]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC4-NEXT:   [[VEC_SEL_FIN:%.*]] = select <4 x i1> [[VEC_ICMP7]], <4 x i32> [[VEC_SEL6]], <4 x i32> [[VEC_SEL4]]
+; CHECK-VF4IC4-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL_FIN]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC4-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC4-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 7, i32 3
 
 
 ; CHECK-VF1IC4:      vector.body:
-; CHECK-VF1IC4:        [[VEC_PHI1:%.*]] = phi i1 [ false, %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
-; CHECK-VF1IC4-NEXT:   [[VEC_PHI2:%.*]] = phi i1 [ false, %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
-; CHECK-VF1IC4-NEXT:   [[VEC_PHI3:%.*]] = phi i1 [ false, %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
-; CHECK-VF1IC4-NEXT:   [[VEC_PHI4:%.*]] = phi i1 [ false, %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
+; CHECK-VF1IC4:        [[VEC_PHI1:%.*]] = phi i32 [ 3, %vector.ph ], [ [[VEC_SEL1:%.*]], %vector.body ]
+; CHECK-VF1IC4-NEXT:   [[VEC_PHI2:%.*]] = phi i32 [ 3, %vector.ph ], [ [[VEC_SEL2:%.*]], %vector.body ]
+; CHECK-VF1IC4-NEXT:   [[VEC_PHI3:%.*]] = phi i32 [ 3, %vector.ph ], [ [[VEC_SEL3:%.*]], %vector.body ]
+; CHECK-VF1IC4-NEXT:   [[VEC_PHI4:%.*]] = phi i32 [ 3, %vector.ph ], [ [[VEC_SEL4:%.*]], %vector.body ]
 ; CHECK-VF1IC4:        [[VEC_LOAD1:%.*]] = load i32
 ; CHECK-VF1IC4-NEXT:   [[VEC_LOAD2:%.*]] = load i32
 ; CHECK-VF1IC4-NEXT:   [[VEC_LOAD3:%.*]] = load i32
@@ -54,20 +52,17 @@ define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) {
 ; CHECK-VF1IC4-NEXT:   [[VEC_ICMP2:%.*]] = icmp eq i32 [[VEC_LOAD2]], 3
 ; CHECK-VF1IC4-NEXT:   [[VEC_ICMP3:%.*]] = icmp eq i32 [[VEC_LOAD3]], 3
 ; CHECK-VF1IC4-NEXT:   [[VEC_ICMP4:%.*]] = icmp eq i32 [[VEC_LOAD4]], 3
-; CHECK-VF1IC4-NEXT:   [[NOT1:%.*]] = xor i1 [[VEC_ICMP1]], true
-; CHECK-VF1IC4-NEXT:   [[NOT2:%.*]] = xor i1 [[VEC_ICMP2]], true
-; CHECK-VF1IC4-NEXT:   [[NOT3:%.*]] = xor i1 [[VEC_ICMP3]], true
-; CHECK-VF1IC4-NEXT:   [[NOT4:%.*]] = xor i1 [[VEC_ICMP4]], true
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL1:%.*]] = or i1 [[VEC_PHI1]], [[NOT1]]
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL2:%.*]] = or i1 [[VEC_PHI2]], [[NOT2]]
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL3:%.*]] = or i1 [[VEC_PHI3]], [[NOT3]]
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL4:%.*]] = or i1 [[VEC_PHI4]], [[NOT4]]
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL1]] = select i1 [[VEC_ICMP1]], i32 [[VEC_PHI1]], i32 7
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL2]] = select i1 [[VEC_ICMP2]], i32 [[VEC_PHI2]], i32 7
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL3]] = select i1 [[VEC_ICMP3]], i32 [[VEC_PHI3]], i32 7
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL4]] = select i1 [[VEC_ICMP4]], i32 [[VEC_PHI4]], i32 7
 ; CHECK-VF1IC4:      middle.block:
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL5:%.*]] = or i1 [[VEC_SEL2]], [[VEC_SEL1]]
-; CHECK-VF1IC4-NEXT:   [[VEC_SEL6:%.*]] = or i1 [[VEC_SEL3]], [[VEC_SEL5]]
-; CHECK-VF1IC4-NEXT:   [[OR_RDX:%.*]] = or i1  [[VEC_SEL4]], [[VEC_SEL6]]
-; CHECK-VF1IC4-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF1IC4-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 7, i32 3
+; CHECK-VF1IC4-NEXT:   [[VEC_ICMP4:%.*]] = icmp ne i32 [[VEC_SEL1]], 3
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL5:%.*]] = select i1 [[VEC_ICMP4]], i32 [[VEC_SEL1]], i32 [[VEC_SEL2]]
+; CHECK-VF1IC4-NEXT:   [[VEC_ICMP5:%.*]] = icmp ne i32 [[VEC_SEL5]], 3
+; CHECK-VF1IC4-NEXT:   [[VEC_SEL6:%.*]] = select i1 [[VEC_ICMP5]], i32 [[VEC_SEL5]], i32 [[VEC_SEL3]]
+; CHECK-VF1IC4-NEXT:   [[VEC_ICMP6:%.*]] = icmp ne i32 [[VEC_SEL6]], 3
+; CHECK-VF1IC4-NEXT:   {{.*}} = select i1 [[VEC_ICMP6]], i32 [[VEC_SEL6]], i32 [[VEC_SEL4]]
 
 entry:
   br label %for.body
@@ -91,14 +86,14 @@ exit:                                     ; preds = %for.body
 define i32 @select_const_i32_from_icmp2(ptr nocapture readonly %v, i64 %n) {
 ; CHECK-LABEL: @select_const_i32_from_icmp2
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 3, i32 3, i32 3, i32 3>, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <4 x i32>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <4 x i32> [[VEC_LOAD]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[VEC_ICMP]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_ICMP]], <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 7, i32 3
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 7, i32 3
 
 entry:
   br label %for.body
@@ -122,18 +117,21 @@ exit:                                     ; preds = %for.body
 define i32 @select_i32_from_icmp(ptr nocapture readonly %v, i32 %a, i32 %b, i64 %n) {
 ; CHECK-LABEL: @select_i32_from_icmp
 ; CHECK-VF4IC1:      vector.ph:
-; CHECK-VF4IC1-NOT:    shufflevector <4 x i32>
-; CHECK-VF4IC1-NOT:    shufflevector <4 x i32>
+; CHECK-VF4IC1:        [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 %a, i64 0
+; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_A:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 %b, i64 0
+; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_B:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ [[SPLAT_OF_A]], %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <4 x i32>
 ; CHECK-VF4IC1-NEXT:   [[VEC_ICMP:%.*]] = icmp eq <4 x i32> [[VEC_LOAD]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-VF4IC1-NEXT:   [[NOT:%.*]] = xor <4 x i1> [[VEC_ICMP]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_ICMP]], <4 x i32> [[VEC_PHI]], <4 x i32> [[SPLAT_OF_B]]
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 %b, i32 %a
+; CHECK-VF4IC1-NEXT:   [[FIN_INS:%.*]] = insertelement <4 x i32> poison, i32 %a, i64 0
+; CHECK-VF4IC1-NEXT:   [[FIN_SPLAT:%.*]] = shufflevector <4 x i32> [[FIN_INS]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[FIN_CMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], [[FIN_SPLAT]]
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_CMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 %b, i32 %a
 entry:
   br label %for.body
 
@@ -156,15 +154,14 @@ exit:                                     ; preds = %for.body
 define i32 @select_const_i32_from_fcmp_fast(ptr nocapture readonly %v, i64 %n) {
 ; CHECK-LABEL: @select_const_i32_from_fcmp_fast
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 2, i32 2, i32 2, i32 2>, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <4 x float>
 ; CHECK-VF4IC1-NEXT:   [[VEC_FCMP:%.*]] = fcmp fast ueq <4 x float> [[VEC_LOAD]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>
-; CHECK-VF4IC1-NEXT:   [[NOT:%.*]] = xor <4 x i1> [[VEC_FCMP]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_FCMP]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 1, i32 2
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], <i32 2, i32 2, i32 2, i32 2>
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 1, i32 2
 entry:
   br label %for.body
 
@@ -187,15 +184,14 @@ exit:                                     ; preds = %for.body
 define i32 @select_const_i32_from_fcmp(ptr nocapture readonly %v, i64 %n) {
 ; CHECK-LABEL: @select_const_i32_from_fcmp
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 2, i32 2, i32 2, i32 2>, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
 ; CHECK-VF4IC1:        [[VEC_LOAD:%.*]] = load <4 x float>
 ; CHECK-VF4IC1-NEXT:   [[VEC_FCMP:%.*]] = fcmp ueq <4 x float> [[VEC_LOAD]], <float 3.000000e+00, float 3.000000e+00, float 3.000000e+00, float 3.000000e+00>
-; CHECK-VF4IC1-NEXT:   [[NOT:%.*]] = xor <4 x i1> [[VEC_FCMP]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_FCMP]], <4 x i32> [[VEC_PHI]], <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 1, i32 2
+; CHECK-VF4IC1-NEXT:   [[FIN_ICMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], <i32 2, i32 2, i32 2, i32 2>
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_ICMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 1, i32 2
 entry:
   br label %for.body
 
@@ -220,16 +216,18 @@ define i32 @select_i32_from_icmp_same_inputs(i32 %a, i32 %b, i64 %n) {
 ; CHECK-VF4IC1:      vector.ph:
 ; CHECK-VF4IC1:        [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 %a, i64 0
 ; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_A:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
-; CHECK-VF4IC1-NOT:   [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 %b, i64 0
+; CHECK-VF4IC1-NEXT:   [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 %b, i64 0
+; CHECK-VF4IC1-NEXT:   [[SPLAT_OF_B:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-VF4IC1:      vector.body:
-; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
-; CHECK-VF4IC1:        [[VEC_ICMP:%.*]] = icmp eq <4 x i32> [[SPLAT_OF_A]], <i32 3, i32 3, i32 3, i32 3>
-; CHECK-VF4IC1-NEXT:   [[NOT:%.*]] = xor <4 x i1> [[VEC_ICMP]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = or <4 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-VF4IC1:        [[VEC_PHI:%.*]] = phi <4 x i32> [ [[SPLAT_OF_A]], %vector.ph ], [ [[VEC_SEL:%.*]], %vector.body ]
+; CHECK-VF4IC1:        [[VEC_ICMP:%.*]] = icmp eq <4 x i32> [[VEC_PHI]], <i32 3, i32 3, i32 3, i32 3>
+; CHECK-VF4IC1-NEXT:   [[VEC_SEL]] = select <4 x i1> [[VEC_ICMP]], <4 x i32> [[VEC_PHI]], <4 x i32> [[SPLAT_OF_B]]
 ; CHECK-VF4IC1:      middle.block:
-; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[VEC_SEL]])
-; CHECK-VF4IC1-NEXT:   [[FR_OR_RDX:%.*]] = freeze i1 [[OR_RDX]]
-; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[FR_OR_RDX]], i32 %b, i32 %a
+; CHECK-VF4IC1-NEXT:   [[FIN_INS:%.*]] = insertelement <4 x i32> poison, i32 %a, i64 0
+; CHECK-VF4IC1-NEXT:   [[FIN_SPLAT:%.*]] = shufflevector <4 x i32> [[FIN_INS]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-VF4IC1-NEXT:   [[FIN_CMP:%.*]] = icmp ne <4 x i32> [[VEC_SEL]], [[FIN_SPLAT]]
+; CHECK-VF4IC1-NEXT:   [[OR_RDX:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[FIN_CMP]])
+; CHECK-VF4IC1-NEXT:   {{.*}} = select i1 [[OR_RDX]], i32 %b, i32 %a
 entry:
   br label %for.body
 
diff --git a/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll b/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
index 55e61158a79c6..16ab45415b5cc 100644
--- a/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
@@ -8,25 +8,26 @@ define i64 @pr62565_incoming_value_known_undef(i64 %a, ptr %src) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[A]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ undef, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[OFFSET_IDX]], 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <2 x i32> [[WIDE_LOAD]], <i32 1, i32 1>
-; CHECK-NEXT:    [[NOT:%*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
-; CHECK-NEXT:    [[TMP4]] = or <2 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP4]] = select <2 x i1> [[TMP3]], <2 x i64> [[VEC_PHI]], <2 x i64> [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP4]])
-; CHECK-NEXT:    [[FR_TMP6:%.*]] = freeze i1 [[TMP6]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR_TMP6]], i64 [[A]], i64 undef
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <2 x i64> [[TMP4]], undef
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 undef
 ; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
@@ -71,25 +72,26 @@ define i64 @pr62565_incoming_value_known_poison(i64 %a, ptr %src) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[A]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ poison, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[OFFSET_IDX]], 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <2 x i32> [[WIDE_LOAD]], <i32 1, i32 1>
-; CHECK-NEXT:    [[NOT:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
-; CHECK-NEXT:    [[TMP4]] = or <2 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP4]] = select <2 x i1> [[TMP3]], <2 x i64> [[VEC_PHI]], <2 x i64> [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP4]])
-; CHECK-NEXT:    [[FR_TMP6:%.*]] = freeze i1 [[TMP6]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR_TMP6]], i64 [[A]], i64 poison
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <2 x i64> [[TMP4]], poison
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 poison
 ; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
@@ -134,25 +136,30 @@ define i64 @pr62565_incoming_value_may_be_poison(i64 %a, ptr %src, i64 %start) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[START]], i64 0
+; CHECK-NEXT:    [[MINMAX_IDENT_SPLAT:%.*]] = shufflevector <2 x i64> [[MINMAX_IDENT_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[A]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[OFFSET_IDX]], 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <2 x i32> [[WIDE_LOAD]], <i32 1, i32 1>
-; CHECK-NEXT:    [[NOT:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
-; CHECK-NEXT:    [[TMP4]] = or <2 x i1> [[VEC_PHI]], [[NOT]]
+; CHECK-NEXT:    [[TMP4]] = select <2 x i1> [[TMP3]], <2 x i64> [[VEC_PHI]], <2 x i64> [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP4]])
-; CHECK-NEXT:    [[FR_TMP6:%.*]] = freeze i1 [[TMP6]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[FR_TMP6]], i64 [[A]], i64 [[START]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[START]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i64> [[DOTSPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne <2 x i64> [[TMP4]], [[DOTSPLAT]]
+; CHECK-NEXT:    [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[RDX_SELECT_CMP]])
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 [[START]]
 ; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]

From 266b2a26408c42ed1ac84ef38dfc41695423d9da Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 15:28:36 -0500
Subject: [PATCH 664/886] [ValueTracking] Add tests for `computeKnownFPClass`
 of `llvm.vector.reduce.{fmin,fmax,fmaximum,fminimum}`; NFC

---
 .../known-fpclass-reduce-signbit.ll           | 134 ++++++++++++++++++
 .../InstSimplify/known-never-infinity.ll      | 112 +++++++++++++++
 2 files changed, 246 insertions(+)
 create mode 100644 llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll

diff --git a/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll b/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll
new file mode 100644
index 0000000000000..2f3db14b2e1ac
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll
@@ -0,0 +1,134 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; RUN: opt < %s -S -passes=instcombine | FileCheck %s
+
+define i1 @vector_reduce_maximum_signbit(<4 x double> nofpclass(nan nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_maximum_signbit
+; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+define i1 @vector_reduce_maximum_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_maximum_signbit_fail_maybe_nan
+; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+
+define i1 @vector_reduce_minimum_signbit(<4 x double> nofpclass(nan nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_minimum_signbit
+; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+define i1 @vector_reduce_minimum_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_minimum_signbit_fail_maybe_nan
+; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+define i1 @vector_reduce_max_signbit(<4 x double> nofpclass(nan nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_max_signbit
+; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+define i1 @vector_reduce_max_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_max_signbit_fail_maybe_nan
+; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+
+define i1 @vector_reduce_min_signbit(<4 x double> nofpclass(nan nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_min_signbit
+; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+define i1 @vector_reduce_min_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_min_signbit_fail_maybe_nan
+; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+
+
+define i1 @vector_reduce_min_signbit_nnan_from_fmf(<4 x double> nofpclass(nzero) %x) {
+; CHECK-LABEL: define i1 @vector_reduce_min_signbit_nnan_from_fmf
+; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
+; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
+; CHECK-NEXT:    [[OP:%.*]] = call nnan double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[X_ABS]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
+  %op = call nnan double @llvm.vector.reduce.fmin.v4f64(<4 x double> %x.abs)
+  %cmp = fcmp oge double %op, 0.0
+  ret i1 %cmp
+}
+
+
diff --git a/llvm/test/Transforms/InstSimplify/known-never-infinity.ll b/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
index 74039d3ffd56c..470b56a8ef2ae 100644
--- a/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
+++ b/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
@@ -1109,6 +1109,118 @@ define float @fcmp_ult_neginf_implies_class_assert(float %arg) {
   ret float %mul_by_zero
 }
 
+define i1 @isKnownNeverInfinity_vector_reduce_maximum(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_maximum
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_maximum_fail(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_maximum_fail
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_minimum(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_minimum
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_minimum_fail(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_minimum_fail
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_fmax(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmax
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_fmax_fail(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmax_fail
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_fmin(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmin
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
+define i1 @isKnownNeverInfinity_vector_reduce_fmin_fail(<4 x double> %x) {
+; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmin_fail
+; CHECK-SAME: (<4 x double> [[X:%.*]]) {
+; CHECK-NEXT:    [[NINF_X:%.*]] = fadd <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[NINF_X]])
+; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %ninf.x = fadd <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
+  %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %ninf.x)
+  %cmp = fcmp une double %op, 0x7ff0000000000000
+  ret i1 %cmp
+}
+
 declare double @llvm.arithmetic.fence.f64(double)
 declare double @llvm.canonicalize.f64(double)
 declare double @llvm.ceil.f64(double)

From 9eeae4421198b99eab3ae9a4ff678fda26bbda2a Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 10 Apr 2024 14:35:09 -0500
Subject: [PATCH 665/886] [ValueTracking] Implement `computeKnownFPClass` for
 `llvm.vector.reduce.{fmin,fmax,fmaximum,fminimum}`

Closes #88408
---
 llvm/lib/Analysis/ValueTracking.cpp           | 13 ++++++++++
 .../known-fpclass-reduce-signbit.ll           | 25 ++++---------------
 .../InstSimplify/known-never-infinity.ll      | 20 +++------------
 3 files changed, 22 insertions(+), 36 deletions(-)

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index e91dc07f31641..ab2f43e1033fa 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -5032,6 +5032,19 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,
 
       break;
     }
+    case Intrinsic::vector_reduce_fmax:
+    case Intrinsic::vector_reduce_fmin:
+    case Intrinsic::vector_reduce_fmaximum:
+    case Intrinsic::vector_reduce_fminimum: {
+      // reduce min/max will choose an element from one of the vector elements,
+      // so we can infer and class information that is common to all elements.
+      Known = computeKnownFPClass(II->getArgOperand(0), II->getFastMathFlags(),
+                                  InterestedClasses, Depth + 1, Q);
+      // Can only propagate sign if output is never NaN.
+      if (!Known.isKnownNeverNaN())
+        Known.SignBit.reset();
+      break;
+    }
     case Intrinsic::trunc:
     case Intrinsic::floor:
     case Intrinsic::ceil:
diff --git a/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll b/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll
index 2f3db14b2e1ac..f46ea9db751ff 100644
--- a/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll
+++ b/llvm/test/Transforms/InstCombine/known-fpclass-reduce-signbit.ll
@@ -4,10 +4,7 @@
 define i1 @vector_reduce_maximum_signbit(<4 x double> nofpclass(nan nzero) %x) {
 ; CHECK-LABEL: define i1 @vector_reduce_maximum_signbit
 ; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
-; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[X_ABS]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
   %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %x.abs)
@@ -33,10 +30,7 @@ define i1 @vector_reduce_maximum_signbit_fail_maybe_nan(<4 x double> nofpclass(n
 define i1 @vector_reduce_minimum_signbit(<4 x double> nofpclass(nan nzero) %x) {
 ; CHECK-LABEL: define i1 @vector_reduce_minimum_signbit
 ; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
-; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[X_ABS]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
   %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %x.abs)
@@ -61,10 +55,7 @@ define i1 @vector_reduce_minimum_signbit_fail_maybe_nan(<4 x double> nofpclass(n
 define i1 @vector_reduce_max_signbit(<4 x double> nofpclass(nan nzero) %x) {
 ; CHECK-LABEL: define i1 @vector_reduce_max_signbit
 ; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
-; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[X_ABS]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
   %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %x.abs)
@@ -90,10 +81,7 @@ define i1 @vector_reduce_max_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero
 define i1 @vector_reduce_min_signbit(<4 x double> nofpclass(nan nzero) %x) {
 ; CHECK-LABEL: define i1 @vector_reduce_min_signbit
 ; CHECK-SAME: (<4 x double> nofpclass(nan nzero) [[X:%.*]]) {
-; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[X_ABS]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
   %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %x.abs)
@@ -120,10 +108,7 @@ define i1 @vector_reduce_min_signbit_fail_maybe_nan(<4 x double> nofpclass(nzero
 define i1 @vector_reduce_min_signbit_nnan_from_fmf(<4 x double> nofpclass(nzero) %x) {
 ; CHECK-LABEL: define i1 @vector_reduce_min_signbit_nnan_from_fmf
 ; CHECK-SAME: (<4 x double> nofpclass(nzero) [[X:%.*]]) {
-; CHECK-NEXT:    [[X_ABS:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[X]])
-; CHECK-NEXT:    [[OP:%.*]] = call nnan double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[X_ABS]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp oge double [[OP]], 0.000000e+00
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %x.abs = call <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
   %op = call nnan double @llvm.vector.reduce.fmin.v4f64(<4 x double> %x.abs)
diff --git a/llvm/test/Transforms/InstSimplify/known-never-infinity.ll b/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
index 470b56a8ef2ae..4d662c08b1a7a 100644
--- a/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
+++ b/llvm/test/Transforms/InstSimplify/known-never-infinity.ll
@@ -1112,10 +1112,7 @@ define float @fcmp_ult_neginf_implies_class_assert(float %arg) {
 define i1 @isKnownNeverInfinity_vector_reduce_maximum(<4 x double> %x) {
 ; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_maximum
 ; CHECK-SAME: (<4 x double> [[X:%.*]]) {
-; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> [[NINF_X]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
   %op = call double @llvm.vector.reduce.fmaximum.v4f64(<4 x double> %ninf.x)
@@ -1140,10 +1137,7 @@ define i1 @isKnownNeverInfinity_vector_reduce_maximum_fail(<4 x double> %x) {
 define i1 @isKnownNeverInfinity_vector_reduce_minimum(<4 x double> %x) {
 ; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_minimum
 ; CHECK-SAME: (<4 x double> [[X:%.*]]) {
-; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> [[NINF_X]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
   %op = call double @llvm.vector.reduce.fminimum.v4f64(<4 x double> %ninf.x)
@@ -1168,10 +1162,7 @@ define i1 @isKnownNeverInfinity_vector_reduce_minimum_fail(<4 x double> %x) {
 define i1 @isKnownNeverInfinity_vector_reduce_fmax(<4 x double> %x) {
 ; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmax
 ; CHECK-SAME: (<4 x double> [[X:%.*]]) {
-; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> [[NINF_X]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
   %op = call double @llvm.vector.reduce.fmax.v4f64(<4 x double> %ninf.x)
@@ -1196,10 +1187,7 @@ define i1 @isKnownNeverInfinity_vector_reduce_fmax_fail(<4 x double> %x) {
 define i1 @isKnownNeverInfinity_vector_reduce_fmin(<4 x double> %x) {
 ; CHECK-LABEL: define i1 @isKnownNeverInfinity_vector_reduce_fmin
 ; CHECK-SAME: (<4 x double> [[X:%.*]]) {
-; CHECK-NEXT:    [[NINF_X:%.*]] = fadd ninf <4 x double> [[X]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
-; CHECK-NEXT:    [[OP:%.*]] = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> [[NINF_X]])
-; CHECK-NEXT:    [[CMP:%.*]] = fcmp une double [[OP]], 0x7FF0000000000000
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %ninf.x = fadd ninf <4 x double> %x, <double 1.0, double 1.0, double 1.0, double 1.0>
   %op = call double @llvm.vector.reduce.fmin.v4f64(<4 x double> %ninf.x)

From d19bd05c79ad3b1a2c3cb439c3fc60825f66bed7 Mon Sep 17 00:00:00 2001
From: Hubert Tong <hubert-reinterpretcast@users.noreply.github.com>
Date: Tue, 16 Apr 2024 17:26:55 -0400
Subject: [PATCH 666/886] Clang Release Notes: Fix reST formatting

Fix a use of inline code markup to have a non-word character after the ending delimiter as required by reST.
---
 clang/docs/ReleaseNotes.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index e6c345a2f5c0f..4aedfafcb26ae 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -68,7 +68,7 @@ AST Dumping Potentially Breaking Changes
 
 Clang Frontend Potentially Breaking Changes
 -------------------------------------------
-- Removed support for constructing on-stack ``TemplateArgumentList``s; interfaces should instead
+- Removed support for constructing on-stack ``TemplateArgumentList``\ s; interfaces should instead
   use ``ArrayRef<TemplateArgument>`` to pass template arguments. Transitioning internal uses to
   ``ArrayRef<TemplateArgument>`` reduces AST memory usage by 0.4% when compiling clang, and is
   expected to show similar improvements on other workloads.

From 3074060d6a1d7d2e74cb767876bd9e5192d12007 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 14:28:45 -0700
Subject: [PATCH 667/886] [memprof] Use SizeIs (NFC) (#88984)

---
 llvm/unittests/ProfileData/MemProfTest.cpp | 30 +++++++++++-----------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/llvm/unittests/ProfileData/MemProfTest.cpp b/llvm/unittests/ProfileData/MemProfTest.cpp
index f596919ed039a..7e00a80cacf93 100644
--- a/llvm/unittests/ProfileData/MemProfTest.cpp
+++ b/llvm/unittests/ProfileData/MemProfTest.cpp
@@ -183,13 +183,13 @@ TEST(MemProf, FillsValue) {
   // We expect 4 records. We attach alloc site data to foo and bar, i.e.
   // all frames bottom up until we find a non-inline frame. We attach call site
   // data to bar, xyz and abc.
-  ASSERT_EQ(Records.size(), 4U);
+  ASSERT_THAT(Records, SizeIs(4));
 
   // Check the memprof record for foo.
   const llvm::GlobalValue::GUID FooId = IndexedMemProfRecord::getGUID("foo");
   ASSERT_EQ(Records.count(FooId), 1U);
   const MemProfRecord &Foo = Records[FooId];
-  ASSERT_EQ(Foo.AllocSites.size(), 1U);
+  ASSERT_THAT(Foo.AllocSites, SizeIs(1));
   EXPECT_EQ(Foo.AllocSites[0].Info.getAllocCount(), 1U);
   EXPECT_THAT(Foo.AllocSites[0].CallStack[0],
               FrameContains("foo", 5U, 30U, true));
@@ -205,7 +205,7 @@ TEST(MemProf, FillsValue) {
   const llvm::GlobalValue::GUID BarId = IndexedMemProfRecord::getGUID("bar");
   ASSERT_EQ(Records.count(BarId), 1U);
   const MemProfRecord &Bar = Records[BarId];
-  ASSERT_EQ(Bar.AllocSites.size(), 1U);
+  ASSERT_THAT(Bar.AllocSites, SizeIs(1));
   EXPECT_EQ(Bar.AllocSites[0].Info.getAllocCount(), 1U);
   EXPECT_THAT(Bar.AllocSites[0].CallStack[0],
               FrameContains("foo", 5U, 30U, true));
@@ -216,8 +216,8 @@ TEST(MemProf, FillsValue) {
   EXPECT_THAT(Bar.AllocSites[0].CallStack[3],
               FrameContains("abc", 5U, 30U, false));
 
-  ASSERT_EQ(Bar.CallSites.size(), 1U);
-  ASSERT_EQ(Bar.CallSites[0].size(), 2U);
+  ASSERT_THAT(Bar.CallSites, SizeIs(1));
+  ASSERT_THAT(Bar.CallSites[0], SizeIs(2));
   EXPECT_THAT(Bar.CallSites[0][0], FrameContains("foo", 5U, 30U, true));
   EXPECT_THAT(Bar.CallSites[0][1], FrameContains("bar", 51U, 20U, false));
 
@@ -225,8 +225,8 @@ TEST(MemProf, FillsValue) {
   const llvm::GlobalValue::GUID XyzId = IndexedMemProfRecord::getGUID("xyz");
   ASSERT_EQ(Records.count(XyzId), 1U);
   const MemProfRecord &Xyz = Records[XyzId];
-  ASSERT_EQ(Xyz.CallSites.size(), 1U);
-  ASSERT_EQ(Xyz.CallSites[0].size(), 2U);
+  ASSERT_THAT(Xyz.CallSites, SizeIs(1));
+  ASSERT_THAT(Xyz.CallSites[0], SizeIs(2));
   // Expect the entire frame even though in practice we only need the first
   // entry here.
   EXPECT_THAT(Xyz.CallSites[0][0], FrameContains("xyz", 5U, 30U, true));
@@ -237,8 +237,8 @@ TEST(MemProf, FillsValue) {
   ASSERT_EQ(Records.count(AbcId), 1U);
   const MemProfRecord &Abc = Records[AbcId];
   EXPECT_TRUE(Abc.AllocSites.empty());
-  ASSERT_EQ(Abc.CallSites.size(), 1U);
-  ASSERT_EQ(Abc.CallSites[0].size(), 2U);
+  ASSERT_THAT(Abc.CallSites, SizeIs(1));
+  ASSERT_THAT(Abc.CallSites[0], SizeIs(2));
   EXPECT_THAT(Abc.CallSites[0][0], FrameContains("xyz", 5U, 30U, true));
   EXPECT_THAT(Abc.CallSites[0][1], FrameContains("abc", 5U, 30U, false));
 }
@@ -393,9 +393,9 @@ TEST(MemProf, SymbolizationFilter) {
     Records.push_back(KeyRecordPair.second);
   }
 
-  ASSERT_EQ(Records.size(), 1U);
-  ASSERT_EQ(Records[0].AllocSites.size(), 1U);
-  ASSERT_EQ(Records[0].AllocSites[0].CallStack.size(), 1U);
+  ASSERT_THAT(Records, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites[0].CallStack, SizeIs(1));
   EXPECT_THAT(Records[0].AllocSites[0].CallStack[0],
               FrameContains("foo", 5U, 30U, false));
 }
@@ -427,9 +427,9 @@ TEST(MemProf, BaseMemProfReader) {
     Records.push_back(KeyRecordPair.second);
   }
 
-  ASSERT_EQ(Records.size(), 1U);
-  ASSERT_EQ(Records[0].AllocSites.size(), 1U);
-  ASSERT_EQ(Records[0].AllocSites[0].CallStack.size(), 2U);
+  ASSERT_THAT(Records, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites, SizeIs(1));
+  ASSERT_THAT(Records[0].AllocSites[0].CallStack, SizeIs(2));
   EXPECT_THAT(Records[0].AllocSites[0].CallStack[0],
               FrameContains("foo", 20U, 5U, true));
   EXPECT_THAT(Records[0].AllocSites[0].CallStack[1],

From b1385dbd98e877a374ce303fd9d1774faf98e31b Mon Sep 17 00:00:00 2001
From: Michael Jones <michaelrj@google.com>
Date: Tue, 16 Apr 2024 14:39:43 -0700
Subject: [PATCH 668/886] [libc][NFC] fix typo in fenv type proxy headers
 (#88982)

libc.incude.fenv ->
libc.include.fenv
---
 libc/hdr/types/CMakeLists.txt | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/libc/hdr/types/CMakeLists.txt b/libc/hdr/types/CMakeLists.txt
index ecb952b60cc06..f53766777e753 100644
--- a/libc/hdr/types/CMakeLists.txt
+++ b/libc/hdr/types/CMakeLists.txt
@@ -28,7 +28,7 @@ add_proxy_header_library(
     fenv_t.h
   FULL_BUILD_DEPENDS
     libc.include.llvm-libc-types.fenv_t
-    libc.incude.fenv
+    libc.include.fenv
 )
 
 add_proxy_header_library(
@@ -37,5 +37,5 @@ add_proxy_header_library(
     fexcept_t.h
   FULL_BUILD_DEPENDS
     libc.include.llvm-libc-types.fexcept_t
-    libc.incude.fenv
+    libc.include.fenv
 )

From 8aa061ffc75adfab4b3084c918e7d4a3ccd5ba43 Mon Sep 17 00:00:00 2001
From: Peiming Liu <peiming@google.com>
Date: Tue, 16 Apr 2024 15:18:47 -0700
Subject: [PATCH 669/886] [mlir][sparse][NFC] switching to using `let
 argments/results` in td files (#88994)

followed the same style used in "TensorOps.td".
---
 .../SparseTensor/IR/SparseTensorOps.td        | 324 ++++++++++--------
 1 file changed, 174 insertions(+), 150 deletions(-)

diff --git a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
index d7121e8320a4b..4e4441c640ed9 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/IR/SparseTensorOps.td
@@ -27,9 +27,7 @@ class SparseTensor_Op<string mnemonic, list<Trait> traits = []>
 // Sparse Tensor Operations.
 //===----------------------------------------------------------------------===//
 
-def SparseTensor_NewOp : SparseTensor_Op<"new", [Pure]>,
-    Arguments<(ins AnyType:$source)>,
-    Results<(outs AnySparseTensor:$result)> {
+def SparseTensor_NewOp : SparseTensor_Op<"new", [Pure]> {
   string summary = "Materializes a new sparse tensor from given source";
   string description = [{
     Materializes a sparse tensor with contents taken from an opaque pointer
@@ -51,15 +49,14 @@ def SparseTensor_NewOp : SparseTensor_Op<"new", [Pure]>,
     sparse_tensor.new %source : !Source to tensor<1024x1024xf64, #CSR>
     ```
   }];
+
+  let arguments = (ins AnyType:$source);
+  let results = (outs AnySparseTensor:$result);
   let assemblyFormat = "$source attr-dict `:` type($source) `to` type($result)";
 }
 
-def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]>,
-    Arguments<(ins Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$levels,
-                   TensorOf<[AnyType]>:$values)>,
-    Results<(outs AnySparseTensor: $result)> {
+def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]> {
   let summary = "Returns a sparse tensor assembled from the given levels and values";
-
   let description = [{
     Assembles the per-level position and coordinate arrays together with
     the values arrays into a sparse tensor. The order and types of the
@@ -93,6 +90,9 @@ def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]>,
     ```
   }];
 
+  let arguments = (ins Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$levels,
+                   TensorOf<[AnyType]>:$values);
+  let results = (outs AnySparseTensor: $result);
   let assemblyFormat =
     "` ` `(` $levels       `)` `,` $values attr-dict `:`"
     "    `(` type($levels) `)` `,` type($values) `to` type($result)";
@@ -100,16 +100,8 @@ def SparseTensor_AssembleOp : SparseTensor_Op<"assemble", [Pure]>,
   let hasVerifier = 1;
 }
 
-def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVariadicResultSize]>,
-    Arguments<(ins AnySparseTensor:$tensor,
-                   Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$out_levels,
-                   TensorOf<[AnyType]>:$out_values)>,
-    Results<(outs Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$ret_levels,
-                  TensorOf<[AnyType]>:$ret_values,
-                  Variadic<AnyIndexingScalarLike>:$lvl_lens,
-                  AnyIndexingScalarLike:$val_len)> {
+def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVariadicResultSize]> {
   let summary = "Copies the levels and values of the given sparse tensor";
-
   let description = [{
     The disassemble operation is the inverse of `sparse_tensor::assemble`.
     It copies the per-level position and coordinate arrays together with
@@ -143,6 +135,13 @@ def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVaria
     ```
   }];
 
+  let arguments = (ins AnySparseTensor:$tensor,
+                   Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$out_levels,
+                   TensorOf<[AnyType]>:$out_values);
+  let results = (outs Variadic<TensorOf<[AnySignlessIntegerOrIndex]>>:$ret_levels,
+                  TensorOf<[AnyType]>:$ret_values,
+                  Variadic<AnyIndexingScalarLike>:$lvl_lens,
+                  AnyIndexingScalarLike:$val_len);
   let assemblyFormat =
     "$tensor attr-dict `:` type($tensor)"
     "`out_lvls` `(` $out_levels `:` type($out_levels) `)` "
@@ -154,9 +153,7 @@ def SparseTensor_DisassembleOp : SparseTensor_Op<"disassemble", [Pure, SameVaria
 }
 
 def SparseTensor_ConvertOp : SparseTensor_Op<"convert",
-  [Pure, StageWithSortSparseOpInterface]>,
-    Arguments<(ins AnyTensor:$source)>,
-    Results<(outs AnyTensor:$dest)> {
+  [Pure, StageWithSortSparseOpInterface]> {
   string summary = "Converts between different tensor types";
   string description = [{
     Converts one sparse or dense tensor type to another tensor type. The rank
@@ -197,20 +194,22 @@ def SparseTensor_ConvertOp : SparseTensor_Op<"convert",
 
   }];
 
+  let arguments = (ins AnyTensor:$source);
+  let results = (outs AnyTensor:$dest);
+  let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
+
   let extraClassDeclaration = [{
      // Whether the convert can be done by a single step or it would require
      // an extra sort. Inherited from StageWithSortSparseOpInterface.
      bool needsExtraSort();
   }];
 
-  let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
   let hasFolder = 1;
   let hasVerifier = 1;
 }
 
-def SparseTensor_ReinterpretMapOp : SparseTensor_Op<"reinterpret_map", [NoMemoryEffect]>,
-    Arguments<(ins AnySparseTensor:$source)>,
-    Results<(outs AnySparseTensor:$dest)> {
+def SparseTensor_ReinterpretMapOp : SparseTensor_Op<"reinterpret_map",
+    [NoMemoryEffect]> {
   let summary = "Reinterprets the dimension/level maps of the source tensor";
   let description = [{
     Reinterprets the dimension-to-level and level-to-dimension map specified in
@@ -248,19 +247,20 @@ def SparseTensor_ReinterpretMapOp : SparseTensor_Op<"reinterpret_map", [NoMemory
     ```
     }];
 
+  let arguments = (ins AnySparseTensor:$source);
+  let results = (outs AnySparseTensor:$dest);
+  let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
+
   let builders = [
     OpBuilder<(ins "SparseTensorEncodingAttr":$dstEnc, "Value":$source)>
   ];
 
-  let assemblyFormat = "$source attr-dict `:` type($source) `to` type($dest)";
   let hasFolder = 1;
   let hasVerifier = 1;
 }
 
 def SparseTensor_ToPositionsOp : SparseTensor_Op<"positions",
-      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]>,
-    Arguments<(ins AnySparseTensor:$tensor, LevelAttr:$level)>,
-    Results<(outs AnyNon0RankedMemRef:$result)> {
+      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
   let summary = "Extracts the `level`-th positions array of the `tensor`";
   let description = [{
     Returns the positions array of the tensor's storage at the given
@@ -280,14 +280,16 @@ def SparseTensor_ToPositionsOp : SparseTensor_Op<"positions",
        : tensor<64x64xf64, #CSR> to memref<?xindex>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor, LevelAttr:$level);
+  let results = (outs AnyNon0RankedMemRef:$result);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor) `to` type($result)";
+
   let hasVerifier = 1;
 }
 
 def SparseTensor_ToCoordinatesOp : SparseTensor_Op<"coordinates",
-      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]>,
-    Arguments<(ins AnySparseTensor:$tensor, LevelAttr:$level)>,
-    Results<(outs AnyNon0RankedMemRef:$result)> {
+      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
   let summary = "Extracts the `level`-th coordinates array of the `tensor`";
   let description = [{
     Returns the coordinates array of the tensor's storage at the given
@@ -307,14 +309,16 @@ def SparseTensor_ToCoordinatesOp : SparseTensor_Op<"coordinates",
        : tensor<64x64xf64, #CSR> to memref<?xindex>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor, LevelAttr:$level);
+  let results = (outs AnyNon0RankedMemRef:$result);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor) `to` type($result)";
+
   let hasVerifier = 1;
 }
 
 def SparseTensor_ToCoordinatesBufferOp : SparseTensor_Op<"coordinates_buffer",
-      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]>,
-    Arguments<(ins AnySparseTensor:$tensor)>,
-    Results<(outs AnyNon0RankedMemRef:$result)> {
+      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
   let summary = "Extracts the linear coordinates array from a tensor";
   let description = [{
     Returns the linear coordinates array for a sparse tensor with
@@ -339,14 +343,16 @@ def SparseTensor_ToCoordinatesBufferOp : SparseTensor_Op<"coordinates_buffer",
        : tensor<64x64xf64, #COO> to memref<?xindex>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor);
+  let results = (outs AnyNon0RankedMemRef:$result);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor) `to` type($result)";
+
   let hasVerifier = 1;
 }
 
 def SparseTensor_ToValuesOp : SparseTensor_Op<"values",
-      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]>,
-    Arguments<(ins AnySparseTensor:$tensor)>,
-    Results<(outs AnyNon0RankedMemRef:$result)> {
+      [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
   let summary = "Extracts numerical values array from a tensor";
   let description = [{
     Returns the values array of the sparse storage format for the given
@@ -365,13 +371,15 @@ def SparseTensor_ToValuesOp : SparseTensor_Op<"values",
     %1 = sparse_tensor.values %0 : tensor<64x64xf64, #CSR> to memref<?xf64>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor);
+  let results = (outs AnyNon0RankedMemRef:$result);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor) `to` type($result)";
+
   let hasVerifier = 1;
 }
 
-def SparseTensor_NumberOfEntriesOp : SparseTensor_Op<"number_of_entries", [Pure]>,
-    Arguments<(ins AnySparseTensor:$tensor)>,
-    Results<(outs Index:$result)> {
+def SparseTensor_NumberOfEntriesOp : SparseTensor_Op<"number_of_entries", [Pure]> {
   let summary = "Returns the number of entries that are stored in the tensor.";
   let description = [{
     Returns the number of entries that are stored in the given sparse tensor.
@@ -385,14 +393,14 @@ def SparseTensor_NumberOfEntriesOp : SparseTensor_Op<"number_of_entries", [Pure]
     %noe = sparse_tensor.number_of_entries %tensor : tensor<64x64xf64, #CSR>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor);
+  let results = (outs Index:$result);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor)";
 }
 
 def SparseTensor_ConcatenateOp : SparseTensor_Op<"concatenate",
-                                 [Pure, StageWithSortSparseOpInterface]>,
-    Arguments<(ins Variadic<AnyRankedTensor>:$inputs, DimensionAttr:$dimension)>,
-    Results<(outs AnyRankedTensor:$result)> {
-
+      [Pure, StageWithSortSparseOpInterface]> {
   let summary = "Concatenates a list of tensors into a single tensor.";
   let description = [{
      Concatenates a list input tensors and the output tensor with the same
@@ -418,13 +426,14 @@ def SparseTensor_ConcatenateOp : SparseTensor_Op<"concatenate",
      bool needsExtraSort();
   }];
 
+  let arguments = (ins Variadic<AnyRankedTensor>:$inputs, DimensionAttr:$dimension);
+  let results = (outs AnyRankedTensor:$result);
   let assemblyFormat = "$inputs attr-dict `:` type($inputs) `to` type($result)";
+
   let hasVerifier = 1;
 }
 
-def SparseTensor_ToSliceOffsetOp : SparseTensor_Op<"slice.offset", [Pure]>,
-    Arguments<(ins AnySparseTensorSlice:$slice, IndexAttr:$dim)>,
-    Results<(outs Index:$offset)> {
+def SparseTensor_ToSliceOffsetOp : SparseTensor_Op<"slice.offset", [Pure]> {
   let summary = "Extracts the offset of the sparse tensor slice at the given dimension";
   let description = [{
     Extracts the offset of the sparse tensor slice at the given dimension.
@@ -445,13 +454,15 @@ def SparseTensor_ToSliceOffsetOp : SparseTensor_Op<"slice.offset", [Pure]>,
     // %2 = %v2
     ```
   }];
+
+  let arguments = (ins AnySparseTensorSlice:$slice, IndexAttr:$dim);
+  let results = (outs Index:$offset);
   let assemblyFormat = "$slice `at` $dim attr-dict `:` type($slice)";
+
   let hasVerifier = 1;
 }
 
-def SparseTensor_ToSliceStrideOp : SparseTensor_Op<"slice.stride", [Pure]>,
-    Arguments<(ins AnySparseTensorSlice:$slice, IndexAttr:$dim)>,
-    Results<(outs Index:$stride)> {
+def SparseTensor_ToSliceStrideOp : SparseTensor_Op<"slice.stride", [Pure]> {
   let summary = "Extracts the stride of the sparse tensor slice at the given dimension";
   let description = [{
     Extracts the stride of the sparse tensor slice at the given dimension.
@@ -473,7 +484,11 @@ def SparseTensor_ToSliceStrideOp : SparseTensor_Op<"slice.stride", [Pure]>,
 
     ```
   }];
+
+  let arguments = (ins AnySparseTensorSlice:$slice, IndexAttr:$dim);
+  let results = (outs Index:$stride);
   let assemblyFormat = "$slice `at` $dim attr-dict `:` type($slice)";
+
   let hasVerifier = 1;
 }
 
@@ -482,9 +497,7 @@ def SparseTensor_ToSliceStrideOp : SparseTensor_Op<"slice.stride", [Pure]>,
 //===----------------------------------------------------------------------===//
 
 def SparseTensor_StorageSpecifierInitOp : SparseTensor_Op<"storage_specifier.init",
-  [Pure]>,
-    Arguments<(ins Optional<SparseTensorStorageSpecifier>:$source)>,
-    Results<(outs SparseTensorStorageSpecifier:$result)> {
+      [Pure]> {
   let summary = "";
   let description = [{
     Returns an initial storage specifier value.  A storage specifier
@@ -515,6 +528,10 @@ def SparseTensor_StorageSpecifierInitOp : SparseTensor_Op<"storage_specifier.ini
     ```
   }];
 
+  let arguments = (ins Optional<SparseTensorStorageSpecifier>:$source);
+  let results = (outs SparseTensorStorageSpecifier:$result);
+  let assemblyFormat = "attr-dict (`with` $source^)? `:` (`from` qualified(type($source))^ `to`)?"
+                                                        " qualified(type($result))";
   let builders = [
     OpBuilder<(ins "Type":$result),
     [{
@@ -522,15 +539,10 @@ def SparseTensor_StorageSpecifierInitOp : SparseTensor_Op<"storage_specifier.ini
     }]>
   ];
 
-  let assemblyFormat = "attr-dict (`with` $source^)? `:` (`from` qualified(type($source))^ `to`)?"
-                                                        " qualified(type($result))";
+
 }
 
-def SparseTensor_GetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.get", [Pure]>,
-    Arguments<(ins SparseTensorStorageSpecifier:$specifier,
-                   SparseTensorStorageSpecifierKindAttr:$specifierKind,
-                   OptionalAttr<LevelAttr>:$level)>,
-    Results<(outs Index:$result)> {
+def SparseTensor_GetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.get", [Pure]> {
   let summary = "";
   let description = [{
     Returns the requested field of the given storage_specifier.
@@ -543,19 +555,19 @@ def SparseTensor_GetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.get"
     ```
   }];
 
+  let arguments = (ins SparseTensorStorageSpecifier:$specifier,
+                   SparseTensorStorageSpecifierKindAttr:$specifierKind,
+                   OptionalAttr<LevelAttr>:$level);
+  let results = (outs Index:$result);
   let assemblyFormat = "$specifier $specifierKind (`at` $level^)? attr-dict"
                        "`:` qualified(type($specifier))";
+
   let hasVerifier = 1;
   let hasFolder = 1;
 }
 
 def SparseTensor_SetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.set",
-    [Pure, AllTypesMatch<["result", "specifier"]>]>,
-    Arguments<(ins SparseTensorStorageSpecifier:$specifier,
-                   SparseTensorStorageSpecifierKindAttr:$specifierKind,
-                   OptionalAttr<LevelAttr>:$level,
-                   Index:$value)>,
-    Results<(outs SparseTensorStorageSpecifier:$result)> {
+    [Pure, AllTypesMatch<["result", "specifier"]>]> {
   let summary = "";
   let description = [{
     Set the field of the storage specifier to the given input value. Returns
@@ -568,8 +580,15 @@ def SparseTensor_SetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.set"
        : !sparse_tensor.storage_specifier<#COO>
     ```
   }];
+
+  let arguments = (ins SparseTensorStorageSpecifier:$specifier,
+                   SparseTensorStorageSpecifierKindAttr:$specifierKind,
+                   OptionalAttr<LevelAttr>:$level,
+                   Index:$value);
+  let results = (outs SparseTensorStorageSpecifier:$result);
   let assemblyFormat = "$specifier $specifierKind (`at` $level^)? `with` $value"
                        " attr-dict `:` qualified(type($result))";
+
   let hasVerifier = 1;
 }
 
@@ -577,9 +596,7 @@ def SparseTensor_SetStorageSpecifierOp : SparseTensor_Op<"storage_specifier.set"
 // Sparse Tensor Coordinate Operations.
 //===----------------------------------------------------------------------===//
 
-def SparseTensor_LvlOp : SparseTensor_Op<"lvl", [ConditionallySpeculatable, NoMemoryEffect]>,
-    Arguments<(ins AnySparseTensor:$source, Index:$index)>,
-    Results<(outs Index:$result)> {
+def SparseTensor_LvlOp : SparseTensor_Op<"lvl", [ConditionallySpeculatable, NoMemoryEffect]> {
   let summary = "level index operation";
   let description = [{
     The `sparse_tensor.lvl` behaves similar to `tensor.dim` operation.
@@ -615,9 +632,9 @@ def SparseTensor_LvlOp : SparseTensor_Op<"lvl", [ConditionallySpeculatable, NoMe
     ```
   }];
 
-  let assemblyFormat = [{
-    attr-dict $source `,` $index `:` type($source)
-  }];
+  let arguments = (ins AnySparseTensor:$source, Index:$index);
+  let results = (outs Index:$result);
+  let assemblyFormat = "attr-dict $source `,` $index `:` type($source) ";
 
   let builders = [
     OpBuilder<(ins "Value":$source, "int64_t":$index)>
@@ -635,11 +652,7 @@ def SparseTensor_LvlOp : SparseTensor_Op<"lvl", [ConditionallySpeculatable, NoMe
   let hasFolder = 1;
 }
 
-def SparseTensor_CrdTranslateOp : SparseTensor_Op<"crd_translate", [Pure]>,
-    Arguments<(ins Variadic<Index>:$in_crds,
-                   SparseTensorCrdTransDirectionAttr:$direction,
-                   SparseTensorEncodingAttr:$encoder)>,
-    Results<(outs Variadic<Index>:$out_crds)> {
+def SparseTensor_CrdTranslateOp : SparseTensor_Op<"crd_translate", [Pure]> {
   string summary = "Performs coordinate translation between level and dimension coordinate space.";
   string description = [{
     Performs coordinate translation between level and dimension coordinate space according
@@ -652,7 +665,13 @@ def SparseTensor_CrdTranslateOp : SparseTensor_Op<"crd_translate", [Pure]>,
                        : index, index, index, index
     ```
   }];
+
+  let arguments = (ins Variadic<Index>:$in_crds,
+                   SparseTensorCrdTransDirectionAttr:$direction,
+                   SparseTensorEncodingAttr:$encoder);
+  let results = (outs Variadic<Index>:$out_crds);
   let assemblyFormat = "$direction `[` $in_crds `]` `as` $encoder attr-dict `:` type($out_crds)";
+
   let hasVerifier = 1;
   let hasFolder = 1;
 }
@@ -669,13 +688,7 @@ def SparseTensor_PushBackOp : SparseTensor_Op<"push_back",
     [TypesMatchWith<"value type matches element type of inBuffer",
                     "inBuffer", "value",
                     "::llvm::cast<ShapedType>($_self).getElementType()">,
-     AllTypesMatch<["inBuffer", "outBuffer"]>]>,
-    Arguments<(ins Index:$curSize,
-               StridedMemRefRankOf<[AnyType], [1]>:$inBuffer,
-               AnyType:$value, Optional<Index>:$n,
-               UnitAttr:$inbounds)>,
-    Results<(outs StridedMemRefRankOf<[AnyType], [1]>:$outBuffer,
-             Index:$newSize)>  {
+     AllTypesMatch<["inBuffer", "outBuffer"]>]> {
   string summary = "Pushes a value to the back of a given buffer";
   string description = [{
     Pushes `value` to the end of the given sparse tensor storage buffer
@@ -719,6 +732,13 @@ def SparseTensor_PushBackOp : SparseTensor_Op<"push_back",
        : xindex, memref<?xf64>, f64
     ```
   }];
+
+  let arguments = (ins Index:$curSize,
+                       StridedMemRefRankOf<[AnyType], [1]>:$inBuffer,
+                       AnyType:$value, Optional<Index>:$n,
+                       UnitAttr:$inbounds);
+  let results = (outs StridedMemRefRankOf<[AnyType], [1]>:$outBuffer,
+                      Index:$newSize);
   let assemblyFormat = "(`inbounds` $inbounds^)? $curSize `,` $inBuffer"
                        " `,` $value (`,` $n^ )?  attr-dict `:`"
                        " type($curSize) `,` type($inBuffer) `,`"
@@ -732,12 +752,7 @@ def SparseTensor_PushBackOp : SparseTensor_Op<"push_back",
   let hasVerifier = 1;
 }
 
-def SparseTensor_ExpandOp : SparseTensor_Op<"expand", []>,
-    Arguments<(ins AnySparseTensor:$tensor)>,
-    Results<(outs AnyStridedMemRefOfRank<1>:$values,
-                  StridedMemRefRankOf<[I1],[1]>:$filled,
-                  StridedMemRefRankOf<[Index],[1]>:$added,
-                  Index:$count)> {
+def SparseTensor_ExpandOp : SparseTensor_Op<"expand", []> {
   string summary = "Expands an access pattern for insertion";
   string description = [{
     Performs an access pattern expansion for the innermost levels of the
@@ -771,19 +786,19 @@ def SparseTensor_ExpandOp : SparseTensor_Op<"expand", []>,
       : tensor<4x4xf64, #CSR> to memref<?xf64>, memref<?xi1>, memref<?xindex>
     ```
   }];
+
+
+  let arguments = (ins AnySparseTensor:$tensor);
+  let results = (outs AnyStridedMemRefOfRank<1>:$values,
+                      StridedMemRefRankOf<[I1],[1]>:$filled,
+                      StridedMemRefRankOf<[Index],[1]>:$added,
+                      Index:$count);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor) `to` type($values)"
                        " `,` type($filled) `,` type($added)";
 }
 
 def SparseTensor_CompressOp : SparseTensor_Op<"compress",
-    [AllTypesMatch<["tensor", "result"]>]>,
-    Arguments<(ins AnyStridedMemRefOfRank<1>:$values,
-                   StridedMemRefRankOf<[I1],[1]>:$filled,
-                   StridedMemRefRankOf<[Index],[1]>:$added,
-                   Index:$count,
-                   AnySparseTensor:$tensor,
-                   Variadic<Index>:$lvlCoords)>,
-    Results<(outs AnySparseTensor:$result)> {
+    [AllTypesMatch<["tensor", "result"]>]> {
   string summary = "Compressed an access pattern for insertion";
   string description = [{
     Finishes a single access pattern expansion by moving inserted elements
@@ -807,6 +822,14 @@ def SparseTensor_CompressOp : SparseTensor_Op<"compress",
       : memref<?xf64>, memref<?xi1>, memref<?xindex>, tensor<4x4xf64, #CSR>
     ```
   }];
+
+  let arguments = (ins AnyStridedMemRefOfRank<1>:$values,
+                   StridedMemRefRankOf<[I1],[1]>:$filled,
+                   StridedMemRefRankOf<[Index],[1]>:$added,
+                   Index:$count,
+                   AnySparseTensor:$tensor,
+                   Variadic<Index>:$lvlCoords);
+  let results = (outs AnySparseTensor:$result);
   let assemblyFormat = "$values `,` $filled `,` $added `,` $count"
                        " `into` $tensor `[` $lvlCoords `]` attr-dict"
                        " `:` type($values) `,` type($filled) `,` type($added)"
@@ -814,9 +837,7 @@ def SparseTensor_CompressOp : SparseTensor_Op<"compress",
   let hasVerifier = 1;
 }
 
-def SparseTensor_LoadOp : SparseTensor_Op<"load", [SameOperandsAndResultType]>,
-    Arguments<(ins AnySparseTensor:$tensor, UnitAttr:$hasInserts)>,
-    Results<(outs AnyTensor:$result)> {
+def SparseTensor_LoadOp : SparseTensor_Op<"load", [SameOperandsAndResultType]> {
   let summary =
     "Rematerializes tensor from underlying sparse storage format";
   let description = [{
@@ -845,11 +866,13 @@ def SparseTensor_LoadOp : SparseTensor_Op<"load", [SameOperandsAndResultType]>,
     %1 = sparse_tensor.load %0 hasInserts : tensor<16x32xf32, #CSR>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor, UnitAttr:$hasInserts);
+  let results = (outs AnyTensor:$result);
   let assemblyFormat = "$tensor (`hasInserts` $hasInserts^)? attr-dict `:` type($tensor)";
 }
 
-def SparseTensor_OutOp : SparseTensor_Op<"out", []>,
-    Arguments<(ins AnySparseTensor:$tensor, AnyType:$dest)> {
+def SparseTensor_OutOp : SparseTensor_Op<"out", []> {
   string summary = "Outputs a sparse tensor to the given destination";
   string description = [{
     Outputs the contents of a sparse tensor to the destination defined by an
@@ -868,6 +891,8 @@ def SparseTensor_OutOp : SparseTensor_Op<"out", []>,
     sparse_tensor.out %t, %dest : tensor<1024x1024xf64, #CSR>, !Dest
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor, AnyType:$dest);
   let assemblyFormat = "$tensor `,` $dest attr-dict `:` type($tensor) `,` type($dest)";
 }
 
@@ -875,11 +900,7 @@ def SparseTensor_OutOp : SparseTensor_Op<"out", []>,
 // Sparse Tensor Sorting/Ordering Operations.
 //===----------------------------------------------------------------------===//
 
-def SparseTensor_SortOp : SparseTensor_Op<"sort">,
-    Arguments<(ins Index:$n, StridedMemRefRankOf<[AnyInteger, Index], [1]>:$xy,
-               Variadic<StridedMemRefRankOf<[AnyType], [1]>>:$ys,
-               AffineMapAttr:$perm_map, OptionalAttr<IndexAttr>:$ny,
-               SparseTensorSortKindAttr:$algorithm)>  {
+def SparseTensor_SortOp : SparseTensor_Op<"sort"> {
   let summary = "Sorts the arrays in xs and ys lexicographically on the "
                 "integral values found in the xs list";
   let description = [{
@@ -904,16 +925,18 @@ def SparseTensor_SortOp : SparseTensor_Op<"sort">,
     ```
   }];
 
+  let arguments = (ins Index:$n,
+                       StridedMemRefRankOf<[AnyInteger, Index], [1]>:$xy,
+                       Variadic<StridedMemRefRankOf<[AnyType], [1]>>:$ys,
+                       AffineMapAttr:$perm_map, OptionalAttr<IndexAttr>:$ny,
+                       SparseTensorSortKindAttr:$algorithm);
   let assemblyFormat = "$algorithm $n"
                        "`,`$xy (`jointly` $ys^)? attr-dict"
                        "`:` type($xy) (`jointly` type($ys)^)?";
   let hasVerifier = 1;
 }
 
-def SparseTensor_ReorderCOOOp : SparseTensor_Op<"reorder_coo", [Pure]>,
-    Arguments<(ins AnySparseTensor: $input_coo,
-                   SparseTensorSortKindAttr:$algorithm)>,
-    Results<(outs AnySparseTensor: $result_coo)> {
+def SparseTensor_ReorderCOOOp : SparseTensor_Op<"reorder_coo", [Pure]> {
   let summary = "Reorder the input COO such that it has the the same order as "
                 "the output COO";
   let description = [{
@@ -933,6 +956,9 @@ def SparseTensor_ReorderCOOOp : SparseTensor_Op<"reorder_coo", [Pure]>,
     ```
   }];
 
+  let arguments = (ins AnySparseTensor: $input_coo,
+                       SparseTensorSortKindAttr:$algorithm);
+  let results = (outs AnySparseTensor: $result_coo);
   let assemblyFormat = "$algorithm $input_coo attr-dict"
                        "`:` type($input_coo) `to` type($result_coo)";
 
@@ -944,9 +970,7 @@ def SparseTensor_ReorderCOOOp : SparseTensor_Op<"reorder_coo", [Pure]>,
 // Sparse Tensor Syntax Operations.
 //===----------------------------------------------------------------------===//
 
-def SparseTensor_BinaryOp : SparseTensor_Op<"binary", [Pure]>,
-    Arguments<(ins AnyType:$x, AnyType:$y, UnitAttr:$left_identity, UnitAttr:$right_identity)>,
-    Results<(outs AnyType:$output)> {
+def SparseTensor_BinaryOp : SparseTensor_Op<"binary", [Pure]> {
   let summary = "Binary set operation utilized within linalg.generic";
   let description = [{
       Defines a computation within a `linalg.generic` operation that takes two
@@ -1054,18 +1078,24 @@ def SparseTensor_BinaryOp : SparseTensor_Op<"binary", [Pure]>,
   }];
 
   let regions = (region AnyRegion:$overlapRegion, AnyRegion:$leftRegion, AnyRegion:$rightRegion);
+  let arguments = (ins AnyType:$x, AnyType:$y, UnitAttr:$left_identity, UnitAttr:$right_identity);
+  let results = (outs AnyType:$output);
   let assemblyFormat = [{
         $x `,` $y `:` attr-dict type($x) `,` type($y) `to` type($output) `\n`
         `overlap` `=` $overlapRegion `\n`
         `left` `=` (`identity` $left_identity^):($leftRegion)? `\n`
         `right` `=` (`identity` $right_identity^):($rightRegion)?
   }];
+
   let hasVerifier = 1;
 }
 
-def SparseTensor_UnaryOp : SparseTensor_Op<"unary", [Pure]>,
-    Arguments<(ins AnyType:$x)>,
-    Results<(outs AnyType:$output)> {
+def SparseTensor_UnaryOp : SparseTensor_Op<"unary", [Pure]> {
+
+  let arguments = (ins AnyType:$x);
+
+  let results = (outs AnyType:$output);
+
   let summary = "Unary set operation utilized within linalg.generic";
   let description = [{
       Defines a computation with a `linalg.generic` operation that takes a single
@@ -1162,9 +1192,7 @@ def SparseTensor_UnaryOp : SparseTensor_Op<"unary", [Pure]>,
   let hasVerifier = 1;
 }
 
-def SparseTensor_ReduceOp : SparseTensor_Op<"reduce", [Pure, SameOperandsAndResultType]>,
-    Arguments<(ins AnyType:$x, AnyType:$y, AnyType:$identity)>,
-    Results<(outs AnyType:$output)> {
+def SparseTensor_ReduceOp : SparseTensor_Op<"reduce", [Pure, SameOperandsAndResultType]> {
   let summary = "Custom reduction operation utilized within linalg.generic";
   let description = [{
       Defines a computation with a `linalg.generic` operation that takes two
@@ -1208,16 +1236,14 @@ def SparseTensor_ReduceOp : SparseTensor_Op<"reduce", [Pure, SameOperandsAndResu
   }];
 
   let regions = (region SizedRegion<1>:$region);
+  let arguments = (ins AnyType:$x, AnyType:$y, AnyType:$identity);
+  let results = (outs AnyType:$output);
+  let assemblyFormat = "$x `,` $y `,` $identity attr-dict `:` type($output) $region";
 
-  let assemblyFormat = [{
-         $x `,` $y `,` $identity attr-dict `:` type($output) $region
-  }];
   let hasVerifier = 1;
 }
 
-def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResultType]>,
-    Arguments<(ins AnyType:$x)>,
-    Results<(outs AnyType:$output)> {
+def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResultType]> {
   let summary = "Select operation utilized within linalg.generic";
   let description = [{
       Defines an evaluation within a `linalg.generic` operation that takes a single
@@ -1269,16 +1295,16 @@ def SparseTensor_SelectOp : SparseTensor_Op<"select", [Pure, SameOperandsAndResu
   }];
 
   let regions = (region SizedRegion<1>:$region);
-  let assemblyFormat = [{
-         $x attr-dict `:` type($x) $region
-  }];
+  let arguments = (ins AnyType:$x);
+  let results = (outs AnyType:$output);
+  let assemblyFormat = "$x attr-dict `:` type($x) $region";
+
   let hasVerifier = 1;
 }
 
 def SparseTensor_YieldOp : SparseTensor_Op<"yield", [Pure, Terminator,
     ParentOneOf<["BinaryOp", "UnaryOp", "ReduceOp", "SelectOp",
-                 "ForeachOp"]>]>,
-    Arguments<(ins Variadic<AnyType>:$results)> {
+                 "ForeachOp"]>]> {
   let summary = "Yield from sparse_tensor set-like operations";
   let description = [{
       Yields a value from within a `binary`, `unary`, `reduce`,
@@ -1319,17 +1345,12 @@ def SparseTensor_YieldOp : SparseTensor_Op<"yield", [Pure, Terminator,
      }
   }];
 
-  let assemblyFormat = [{
-        $results attr-dict `:` type($results)
-  }];
+  let arguments = (ins Variadic<AnyType>:$results);
+  let assemblyFormat = "$results attr-dict `:` type($results)";
 }
 
 def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
-    [SingleBlockImplicitTerminator<"YieldOp">]>,
-     Arguments<(ins AnyTensor:$tensor,
-                    Variadic<AnyType>:$initArgs,
-                    OptionalAttr<AffineMapAttr>:$order)>,
-     Results<(outs Variadic<AnyType>:$results)> {
+    [SingleBlockImplicitTerminator<"YieldOp">]> {
   let summary = "Iterates over elements in a tensor";
   let description = [{
      Iterates over stored elements in a tensor (which are typically, but not always,
@@ -1424,6 +1445,10 @@ def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
   ];
 
   let regions = (region SizedRegion<1>:$region);
+  let arguments = (ins AnyTensor:$tensor,
+                       Variadic<AnyType>:$initArgs,
+                       OptionalAttr<AffineMapAttr>:$order);
+  let results = (outs Variadic<AnyType>:$results);
   let assemblyFormat = "`in` $tensor (`init``(`$initArgs^`)`)? attr-dict"
                        "    `:` type($tensor) (`,` type($initArgs)^)?"
                        "  (`->` type($results)^)?  `do` $region";
@@ -1436,13 +1461,6 @@ def SparseTensor_ForeachOp : SparseTensor_Op<"foreach",
 
 def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
     [Pure, DeclareOpInterfaceMethods<InferTypeOpInterface>]> {
-
-  let arguments = (ins AnySparseTensor:$tensor,
-                       Optional<AnySparseIterator>:$parentIter,
-                       LevelAttr:$loLvl, LevelAttr:$hiLvl);
-
-  let results = (outs AnySparseIterSpace:$resultSpace);
-
   let summary = "Extracts an iteration space from a sparse tensor between certain levels";
   let description = [{
       Extracts a `!sparse_tensor.iter_space` from a sparse tensor between
@@ -1485,17 +1503,21 @@ def ExtractIterSpaceOp : SparseTensor_Op<"extract_iteration_space",
     }
   }];
 
-  let hasVerifier = 1;
+  let arguments = (ins AnySparseTensor:$tensor,
+                       Optional<AnySparseIterator>:$parentIter,
+                       LevelAttr:$loLvl, LevelAttr:$hiLvl);
+  let results = (outs AnySparseIterSpace:$resultSpace);
   let assemblyFormat = "$tensor (`at` $parentIter^)? `lvls` `=` custom<LevelRange>($loLvl, $hiLvl) "
                        " attr-dict `:` type($tensor) (`,` type($parentIter)^)?";
+
+  let hasVerifier = 1;
 }
 
 //===----------------------------------------------------------------------===//
 // Sparse Tensor Debugging and Test-Only Operations.
 //===----------------------------------------------------------------------===//
 
-def SparseTensor_PrintOp : SparseTensor_Op<"print">,
-    Arguments<(ins AnySparseTensor:$tensor)> {
+def SparseTensor_PrintOp : SparseTensor_Op<"print"> {
   string summary = "Prints a sparse tensor (for testing and debugging)";
   string description = [{
     Prints the individual components of a sparse tensors (the positions,
@@ -1509,6 +1531,8 @@ def SparseTensor_PrintOp : SparseTensor_Op<"print">,
     sparse_tensor.print %tensor : tensor<1024x1024xf64, #CSR>
     ```
   }];
+
+  let arguments = (ins AnySparseTensor:$tensor);
   let assemblyFormat = "$tensor attr-dict `:` type($tensor)";
 }
 

From 9067070d91e9d8cdd8509ffa56a076f08a3d7281 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Tue, 16 Apr 2024 15:40:32 -0700
Subject: [PATCH 670/886] [RISCV] Re-separate unaligned scalar and vector
 memory features in the backend. (#88954)

This is largely a revert of commit
e81796671890b59c110f8e41adc7ca26f8484d20.

As #88029 shows, there exists hardware that only supports unaligned
scalar.

I'm leaving how this gets exposed to the clang interface to a future
patch.
---
 clang/lib/Basic/Targets/RISCV.cpp                |  3 ++-
 clang/lib/Driver/ToolChains/Arch/RISCV.cpp       | 16 +++++++++++-----
 clang/test/Driver/riscv-features.c               |  4 ++--
 llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp |  4 ++--
 llvm/lib/Target/RISCV/RISCVFeatures.td           | 13 +++++++++----
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp      | 16 ++++++++--------
 llvm/lib/Target/RISCV/RISCVProcessors.td         |  6 ++++--
 llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h |  4 ++--
 llvm/test/CodeGen/RISCV/memcpy-inline.ll         |  4 ++--
 llvm/test/CodeGen/RISCV/memcpy.ll                |  4 ++--
 llvm/test/CodeGen/RISCV/memset-inline.ll         |  4 ++--
 llvm/test/CodeGen/RISCV/pr56110.ll               |  2 +-
 .../CodeGen/RISCV/riscv-func-target-feature.ll   |  2 +-
 .../RISCV/rvv/concat-vectors-constant-stride.ll  |  4 ++--
 .../rvv/fixed-vectors-strided-load-combine.ll    |  2 +-
 .../CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll |  4 ++--
 llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll     |  4 ++--
 llvm/test/CodeGen/RISCV/rvv/memset-inline.ll     |  4 ++--
 .../CodeGen/RISCV/rvv/unaligned-loads-stores.ll  |  4 ++--
 llvm/test/CodeGen/RISCV/unaligned-load-store.ll  |  4 ++--
 llvm/utils/TableGen/RISCVTargetDefEmitter.cpp    | 12 ++++++++++--
 21 files changed, 71 insertions(+), 49 deletions(-)

diff --git a/clang/lib/Basic/Targets/RISCV.cpp b/clang/lib/Basic/Targets/RISCV.cpp
index f3d705e1551fe..a7ce9dda34bdd 100644
--- a/clang/lib/Basic/Targets/RISCV.cpp
+++ b/clang/lib/Basic/Targets/RISCV.cpp
@@ -353,7 +353,8 @@ bool RISCVTargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
   if (ISAInfo->hasExtension("zfh") || ISAInfo->hasExtension("zhinx"))
     HasLegalHalfType = true;
 
-  FastUnalignedAccess = llvm::is_contained(Features, "+fast-unaligned-access");
+  FastUnalignedAccess = llvm::is_contained(Features, "+unaligned-scalar-mem") &&
+                        llvm::is_contained(Features, "+unaligned-vector-mem");
 
   if (llvm::is_contained(Features, "+experimental"))
     HasExperimental = true;
diff --git a/clang/lib/Driver/ToolChains/Arch/RISCV.cpp b/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
index b1dd7c4372d47..96b3cc3bb8ffb 100644
--- a/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
@@ -68,8 +68,10 @@ static void getRISCFeaturesFromMcpu(const Driver &D, const Arg *A,
           << A->getSpelling() << Mcpu;
   }
 
-  if (llvm::RISCV::hasFastUnalignedAccess(Mcpu))
-    Features.push_back("+fast-unaligned-access");
+  if (llvm::RISCV::hasFastUnalignedAccess(Mcpu)) {
+    Features.push_back("+unaligned-scalar-mem");
+    Features.push_back("+unaligned-vector-mem");
+  }
 }
 
 void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple,
@@ -168,12 +170,16 @@ void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple,
   }
 
   // Android requires fast unaligned access on RISCV64.
-  if (Triple.isAndroid())
-    Features.push_back("+fast-unaligned-access");
+  if (Triple.isAndroid()) {
+    Features.push_back("+unaligned-scalar-mem");
+    Features.push_back("+unaligned-vector-mem");
+  }
 
   // -mstrict-align is default, unless -mno-strict-align is specified.
   AddTargetFeature(Args, Features, options::OPT_mno_strict_align,
-                   options::OPT_mstrict_align, "fast-unaligned-access");
+                   options::OPT_mstrict_align, "unaligned-scalar-mem");
+  AddTargetFeature(Args, Features, options::OPT_mno_strict_align,
+                   options::OPT_mstrict_align, "unaligned-vector-mem");
 
   // Now add any that the user explicitly requested on the command line,
   // which may override the defaults.
diff --git a/clang/test/Driver/riscv-features.c b/clang/test/Driver/riscv-features.c
index ce4947d2bc47b..5e1db5ba1ed3e 100644
--- a/clang/test/Driver/riscv-features.c
+++ b/clang/test/Driver/riscv-features.c
@@ -38,8 +38,8 @@
 // RUN: %clang --target=riscv32-unknown-elf -### %s -mno-strict-align 2>&1 | FileCheck %s -check-prefix=FAST-UNALIGNED-ACCESS
 // RUN: %clang --target=riscv32-unknown-elf -### %s -mstrict-align 2>&1 | FileCheck %s -check-prefix=NO-FAST-UNALIGNED-ACCESS
 
-// FAST-UNALIGNED-ACCESS: "-target-feature" "+fast-unaligned-access"
-// NO-FAST-UNALIGNED-ACCESS: "-target-feature" "-fast-unaligned-access"
+// FAST-UNALIGNED-ACCESS: "-target-feature" "+unaligned-scalar-mem" "-target-feature" "+unaligned-vector-mem"
+// NO-FAST-UNALIGNED-ACCESS: "-target-feature" "-unaligned-scalar-mem" "-target-feature" "-unaligned-vector-mem"
 
 // RUN: %clang --target=riscv32-unknown-elf -### %s 2>&1 | FileCheck %s -check-prefix=NOUWTABLE
 // RUN: %clang --target=riscv32-unknown-elf -fasynchronous-unwind-tables -### %s 2>&1 | FileCheck %s -check-prefix=UWTABLE
diff --git a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
index 173995f05b51c..d93709ac03420 100644
--- a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
+++ b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
@@ -326,8 +326,8 @@ bool RISCVExpandPseudo::expandRV32ZdinxStore(MachineBasicBlock &MBB,
       .setMemRefs(MMOLo);
 
   if (MBBI->getOperand(2).isGlobal() || MBBI->getOperand(2).isCPI()) {
-    // FIXME: Zdinx RV32 can not work on unaligned memory.
-    assert(!STI->hasFastUnalignedAccess());
+    // FIXME: Zdinx RV32 can not work on unaligned scalar memory.
+    assert(!STI->enableUnalignedScalarMem());
 
     assert(MBBI->getOperand(2).getOffset() % 8 == 0);
     MBBI->getOperand(2).setOffset(MBBI->getOperand(2).getOffset() + 4);
diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index 59962216e0c04..561187c39a4a0 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -1183,10 +1183,15 @@ def FeatureTrailingSeqCstFence : SubtargetFeature<"seq-cst-trailing-fence",
                                           "true",
                                           "Enable trailing fence for seq-cst store.">;
 
-def FeatureFastUnalignedAccess
-   : SubtargetFeature<"fast-unaligned-access", "HasFastUnalignedAccess",
-                      "true", "Has reasonably performant unaligned "
-                      "loads and stores (both scalar and vector)">;
+def FeatureUnalignedScalarMem
+   : SubtargetFeature<"unaligned-scalar-mem", "EnableUnalignedScalarMem",
+                      "true", "Has reasonably performant unaligned scalar "
+                      "loads and stores">;
+
+def FeatureUnalignedVectorMem
+   : SubtargetFeature<"unaligned-vector-mem", "EnableUnalignedVectorMem",
+                      "true", "Has reasonably performant unaligned vector "
+                      "loads and stores">;
 
 def FeaturePostRAScheduler : SubtargetFeature<"use-postra-scheduler",
     "UsePostRAScheduler", "true", "Schedule again after register allocation">;
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 7b4bec2f65b74..b0deb1d266995 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1924,7 +1924,7 @@ bool RISCVTargetLowering::shouldConvertConstantLoadToIntImm(const APInt &Imm,
   // replace. If we don't support unaligned scalar mem, prefer the constant
   // pool.
   // TODO: Can the caller pass down the alignment?
-  if (!Subtarget.hasFastUnalignedAccess())
+  if (!Subtarget.enableUnalignedScalarMem())
     return true;
 
   // Prefer to keep the load if it would require many instructions.
@@ -15837,7 +15837,7 @@ static bool matchIndexAsWiderOp(EVT VT, SDValue Index, SDValue Mask,
   if (WiderElementSize > ST.getELen()/8)
     return false;
 
-  if (!ST.hasFastUnalignedAccess() && BaseAlign < WiderElementSize)
+  if (!ST.enableUnalignedVectorMem() && BaseAlign < WiderElementSize)
     return false;
 
   for (unsigned i = 0; i < Index->getNumOperands(); i++) {
@@ -20663,8 +20663,8 @@ bool RISCVTargetLowering::allowsMisalignedMemoryAccesses(
     unsigned *Fast) const {
   if (!VT.isVector()) {
     if (Fast)
-      *Fast = Subtarget.hasFastUnalignedAccess();
-    return Subtarget.hasFastUnalignedAccess();
+      *Fast = Subtarget.enableUnalignedScalarMem();
+    return Subtarget.enableUnalignedScalarMem();
   }
 
   // All vector implementations must support element alignment
@@ -20680,8 +20680,8 @@ bool RISCVTargetLowering::allowsMisalignedMemoryAccesses(
   // misaligned accesses.  TODO: Work through the codegen implications of
   // allowing such accesses to be formed, and considered fast.
   if (Fast)
-    *Fast = Subtarget.hasFastUnalignedAccess();
-  return Subtarget.hasFastUnalignedAccess();
+    *Fast = Subtarget.enableUnalignedVectorMem();
+  return Subtarget.enableUnalignedVectorMem();
 }
 
 
@@ -20716,7 +20716,7 @@ EVT RISCVTargetLowering::getOptimalMemOpType(const MemOp &Op,
 
   // Do we have sufficient alignment for our preferred VT?  If not, revert
   // to largest size allowed by our alignment criteria.
-  if (PreferredVT != MVT::i8 && !Subtarget.hasFastUnalignedAccess()) {
+  if (PreferredVT != MVT::i8 && !Subtarget.enableUnalignedVectorMem()) {
     Align RequiredAlign(PreferredVT.getStoreSize());
     if (Op.isFixedDstAlign())
       RequiredAlign = std::min(RequiredAlign, Op.getDstAlign());
@@ -20908,7 +20908,7 @@ bool RISCVTargetLowering::isLegalStridedLoadStore(EVT DataType,
   if (!isLegalElementTypeForRVV(ScalarType))
     return false;
 
-  if (!Subtarget.hasFastUnalignedAccess() &&
+  if (!Subtarget.enableUnalignedVectorMem() &&
       Alignment < ScalarType.getStoreSize())
     return false;
 
diff --git a/llvm/lib/Target/RISCV/RISCVProcessors.td b/llvm/lib/Target/RISCV/RISCVProcessors.td
index 739b50749e132..f9a557e02bfe1 100644
--- a/llvm/lib/Target/RISCV/RISCVProcessors.td
+++ b/llvm/lib/Target/RISCV/RISCVProcessors.td
@@ -257,7 +257,8 @@ def SIFIVE_P450 : RISCVProcessorModel<"sifive-p450", SiFiveP400Model,
                                        FeatureStdExtZbb,
                                        FeatureStdExtZbs,
                                        FeatureStdExtZfhmin,
-                                       FeatureFastUnalignedAccess],
+                                       FeatureUnalignedScalarMem,
+                                       FeatureUnalignedVectorMem],
                                       [TuneNoDefaultUnroll,
                                        TuneConditionalCompressedMoveFusion,
                                        TuneLUIADDIFusion,
@@ -295,7 +296,8 @@ def SIFIVE_P670 : RISCVProcessorModel<"sifive-p670", SiFiveP600Model,
                                        FeatureStdExtZvkng,
                                        FeatureStdExtZvksc,
                                        FeatureStdExtZvksg,
-                                       FeatureFastUnalignedAccess],
+                                       FeatureUnalignedScalarMem,
+                                       FeatureUnalignedVectorMem],
                                       [TuneNoDefaultUnroll,
                                        TuneConditionalCompressedMoveFusion,
                                        TuneLUIADDIFusion,
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
index e0c0e6517b6f1..2f9281ab89244 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
@@ -228,7 +228,7 @@ class RISCVTTIImpl : public BasicTTIImplBase<RISCVTTIImpl> {
       return false;
 
     EVT ElemType = DataTypeVT.getScalarType();
-    if (!ST->hasFastUnalignedAccess() && Alignment < ElemType.getStoreSize())
+    if (!ST->enableUnalignedVectorMem() && Alignment < ElemType.getStoreSize())
       return false;
 
     return TLI->isLegalElementTypeForRVV(ElemType);
@@ -253,7 +253,7 @@ class RISCVTTIImpl : public BasicTTIImplBase<RISCVTTIImpl> {
       return false;
 
     EVT ElemType = DataTypeVT.getScalarType();
-    if (!ST->hasFastUnalignedAccess() && Alignment < ElemType.getStoreSize())
+    if (!ST->enableUnalignedVectorMem() && Alignment < ElemType.getStoreSize())
       return false;
 
     return TLI->isLegalElementTypeForRVV(ElemType);
diff --git a/llvm/test/CodeGen/RISCV/memcpy-inline.ll b/llvm/test/CodeGen/RISCV/memcpy-inline.ll
index 343695ee37da8..833e07351eec7 100644
--- a/llvm/test/CodeGen/RISCV/memcpy-inline.ll
+++ b/llvm/test/CodeGen/RISCV/memcpy-inline.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32
 ; RUN: llc < %s -mtriple=riscv64 \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
 
 ; ----------------------------------------------------------------------
diff --git a/llvm/test/CodeGen/RISCV/memcpy.ll b/llvm/test/CodeGen/RISCV/memcpy.ll
index 12ec0881b20d9..02f582339d0b7 100644
--- a/llvm/test/CodeGen/RISCV/memcpy.ll
+++ b/llvm/test/CodeGen/RISCV/memcpy.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32
 ; RUN: llc < %s -mtriple=riscv64 \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
 %struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
 
diff --git a/llvm/test/CodeGen/RISCV/memset-inline.ll b/llvm/test/CodeGen/RISCV/memset-inline.ll
index cc22b77c641e2..55fe81a58805e 100644
--- a/llvm/test/CodeGen/RISCV/memset-inline.ll
+++ b/llvm/test/CodeGen/RISCV/memset-inline.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32
 ; RUN: llc < %s -mtriple=riscv64 -mattr=+m \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+unaligned-scalar-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
 %struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
 
diff --git a/llvm/test/CodeGen/RISCV/pr56110.ll b/llvm/test/CodeGen/RISCV/pr56110.ll
index c795b17419f56..fa441f5fc3aef 100644
--- a/llvm/test/CodeGen/RISCV/pr56110.ll
+++ b/llvm/test/CodeGen/RISCV/pr56110.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=riscv32 | FileCheck %s
-; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access | FileCheck %s
+; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem | FileCheck %s
 
 define void @foo_set(ptr nocapture noundef %a, i32 noundef %v) {
 ; CHECK-LABEL: foo_set:
diff --git a/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll b/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
index a03dadbc1d116..d627ae9c90394 100644
--- a/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
+++ b/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
@@ -36,7 +36,7 @@ entry:
 }
 
 ; CHECK-NOT: .option push
-define void @test5() "target-features"="+fast-unaligned-access" {
+define void @test5() "target-features"="+unaligned-scalar-mem" {
 ; CHECK-LABEL: test5
 ; CHECK-NOT: .option pop
 entry:
diff --git a/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll b/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
index f244810e739d9..ff35043dbd7e7 100644
--- a/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=riscv32 -mattr=+v,+fast-unaligned-access -target-abi=ilp32 \
+; RUN: llc -mtriple=riscv32 -mattr=+v,+unaligned-vector-mem -target-abi=ilp32 \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32
-; RUN: llc -mtriple=riscv64 -mattr=+v,+fast-unaligned-access -target-abi=lp64 \
+; RUN: llc -mtriple=riscv64 -mattr=+v,+unaligned-vector-mem -target-abi=lp64 \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64
 
 define void @constant_forward_stride(ptr %s, ptr %d) {
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
index 657d52354aa39..f0fcc482e2207 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
 ; RUN: llc -mtriple=riscv32 -mattr=+v,+zfh,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,RV32
 ; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,RV64
-; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+fast-unaligned-access -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,RV64,RV64-MISALIGN
+; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+unaligned-vector-mem -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,RV64,RV64-MISALIGN
 
 ; RUN: llc -mtriple=riscv64 -mattr=+f,+zfh,+zve64f,+zvl128b,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,ZVE64F
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
index fffc4d6c08335..36c36a13964c9 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=SLOW,RV32-SLOW
 ; RUN: llc -mtriple=riscv64 -mattr=+m,+v -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s --check-prefixes=SLOW,RV64-SLOW
-; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+fast-unaligned-access -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+unaligned-vector-mem -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s --check-prefixes=FAST,RV32-FAST
-; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+fast-unaligned-access -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+unaligned-vector-mem -verify-machineinstrs < %s \
 ; RUN:   | FileCheck %s --check-prefixes=FAST,RV64-FAST
 
 define <4 x i32> @load_v4i32_align1(ptr %ptr) {
diff --git a/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll b/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
index 485f94ee2a102..53598c609107b 100644
--- a/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32
 ; RUN: llc < %s -mtriple=riscv64 -mattr=+v \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+v,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+v,+unaligned-scalar-mem,+unaligned-vector-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+v,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+v,+unaligned-scalar-mem,+unaligned-vector-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
 
 ; ----------------------------------------------------------------------
diff --git a/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll b/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
index 0e7e914cf68e8..accc18519d626 100644
--- a/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32
 ; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+v,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+v,+unaligned-scalar-mem,,+unaligned-vector-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v,+fast-unaligned-access \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v,+unaligned-scalar-mem,+unaligned-vector-mem \
 ; RUN:   | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
 %struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll b/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
index f488baf5a9d9f..1491bb6c337a0 100644
--- a/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
@@ -3,9 +3,9 @@
 ; RUN:    -verify-machineinstrs | FileCheck %s
 ; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v < %s \
 ; RUN:    -verify-machineinstrs | FileCheck %s
-; RUN: llc -mtriple riscv32 -mattr=+d,+zfh,+zvfh,+v,+fast-unaligned-access < %s \
+; RUN: llc -mtriple riscv32 -mattr=+d,+zfh,+zvfh,+v,+unaligned-vector-mem < %s \
 ; RUN:    -verify-machineinstrs | FileCheck --check-prefix=FAST %s
-; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v,+fast-unaligned-access < %s \
+; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v,+unaligned-vector-mem < %s \
 ; RUN:    -verify-machineinstrs | FileCheck --check-prefix=FAST %s
 
 
diff --git a/llvm/test/CodeGen/RISCV/unaligned-load-store.ll b/llvm/test/CodeGen/RISCV/unaligned-load-store.ll
index 599b0d08629ea..ce0d8fedbfb88 100644
--- a/llvm/test/CodeGen/RISCV/unaligned-load-store.ll
+++ b/llvm/test/CodeGen/RISCV/unaligned-load-store.ll
@@ -3,9 +3,9 @@
 ; RUN:   | FileCheck -check-prefixes=ALL,SLOW,RV32I %s
 ; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefixes=ALL,SLOW,RV64I %s
-; RUN: llc -mtriple=riscv32 -mattr=+fast-unaligned-access -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv32 -mattr=+unaligned-scalar-mem -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefixes=ALL,FAST,RV32I-FAST %s
-; RUN: llc -mtriple=riscv64 -mattr=+fast-unaligned-access -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv64 -mattr=+unaligned-scalar-mem -verify-machineinstrs < %s \
 ; RUN:   | FileCheck -check-prefixes=ALL,FAST,RV64I-FAST %s
 
 ; A collection of cases showing codegen for unaligned loads and stores
diff --git a/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp b/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
index 7a6439cb94910..e57bc6fb507e3 100644
--- a/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
+++ b/llvm/utils/TableGen/RISCVTargetDefEmitter.cpp
@@ -60,11 +60,19 @@ static void EmitRISCVTargetDef(RecordKeeper &RK, raw_ostream &OS) {
     if (MArch.empty())
       MArch = getMArch(*Rec);
 
-    const bool FastUnalignedAccess =
+    bool FastScalarUnalignedAccess =
         any_of(Rec->getValueAsListOfDefs("Features"), [&](auto &Feature) {
-          return Feature->getValueAsString("Name") == "fast-unaligned-access";
+          return Feature->getValueAsString("Name") == "unaligned-scalar-mem";
         });
 
+    bool FastVectorUnalignedAccess =
+        any_of(Rec->getValueAsListOfDefs("Features"), [&](auto &Feature) {
+          return Feature->getValueAsString("Name") == "unaligned-vector-mem";
+        });
+
+    bool FastUnalignedAccess =
+        FastScalarUnalignedAccess && FastVectorUnalignedAccess;
+
     OS << "PROC(" << Rec->getName() << ", "
        << "{\"" << Rec->getValueAsString("Name") << "\"}, "
        << "{\"" << MArch << "\"}, " << FastUnalignedAccess << ")\n";

From 988ffd06722e7e056b239efe497345ac97be33db Mon Sep 17 00:00:00 2001
From: Usama Hameed <u_hameed@apple.com>
Date: Tue, 16 Apr 2024 16:00:14 -0700
Subject: [PATCH 671/886] Add asan tests for libsanitizers. (#88349) (#88962)

The previous patch was reverted because the test fails to build when
libsanitizers is not present. This patch catches the BuildError
exception and skips the test appropriately.

This patch tests LLDB integration with libsanitizers for ASan.

rdar://111856681
---
 lldb/test/API/functionalities/asan/Makefile   |  6 +-
 .../functionalities/asan/TestMemoryHistory.py | 74 ++++++++++++++++++-
 .../functionalities/asan/TestReportData.py    | 21 +++++-
 3 files changed, 94 insertions(+), 7 deletions(-)

diff --git a/lldb/test/API/functionalities/asan/Makefile b/lldb/test/API/functionalities/asan/Makefile
index 4913a18d8cc6f..d66696fed7078 100644
--- a/lldb/test/API/functionalities/asan/Makefile
+++ b/lldb/test/API/functionalities/asan/Makefile
@@ -1,4 +1,8 @@
 C_SOURCES := main.c
-CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
+asan: CFLAGS_EXTRAS := -fsanitize=address -g -gcolumn-info
+asan: all
+
+libsanitizers: CFLAGS_EXTRAS := -fsanitize=address -fsanitize-stable-abi -g -gcolumn-info
+libsanitizers: all
 
 include Makefile.rules
diff --git a/lldb/test/API/functionalities/asan/TestMemoryHistory.py b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
index 00162ae8822c7..41ab25823f5cc 100644
--- a/lldb/test/API/functionalities/asan/TestMemoryHistory.py
+++ b/lldb/test/API/functionalities/asan/TestMemoryHistory.py
@@ -8,16 +8,24 @@
 from lldbsuite.test.lldbtest import *
 from lldbsuite.test import lldbplatform
 from lldbsuite.test import lldbutil
-
+from lldbsuite.test_event.build_exception import BuildError
 
 class AsanTestCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
     @expectedFailureNetBSD
     @skipUnlessAddressSanitizer
     def test(self):
-        self.build()
+        self.build(make_targets=["asan"])
         self.asan_tests()
 
+    @skipIf(oslist=no_match(["macosx"]))
+    def test_libsanitizers_asan(self):
+        try:
+            self.build(make_targets=["libsanitizers"])
+        except BuildError as e:
+            self.skipTest("failed to build with libsanitizers")
+        self.libsanitizer_tests()
+
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -26,6 +34,68 @@ def setUp(self):
         self.line_free = line_number("main.c", "// free line")
         self.line_breakpoint = line_number("main.c", "// break line")
 
+    # Test line numbers: rdar://126237493
+    def libsanitizer_tests(self):
+        target = self.createTestTarget()
+
+        self.runCmd(
+            "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
+        )
+
+        self.runCmd("run")
+
+        # In libsanitizers, memory history is not supported until a report has been generated
+        self.expect(
+            "thread list",
+            "Process should be stopped due to ASan report",
+            substrs=["stopped", "stop reason = Use of deallocated memory"],
+        )
+
+        # test the 'memory history' command
+        self.expect(
+            "memory history 'pointer'",
+            substrs=[
+                "Memory deallocated by Thread",
+                "a.out`f2",
+                "main.c",
+                "Memory allocated by Thread",
+                "a.out`f1",
+                "main.c",
+            ],
+        )
+
+        # do the same using SB API
+        process = self.dbg.GetSelectedTarget().process
+        val = (
+            process.GetSelectedThread().GetSelectedFrame().EvaluateExpression("pointer")
+        )
+        addr = val.GetValueAsUnsigned()
+        threads = process.GetHistoryThreads(addr)
+        self.assertEqual(threads.GetSize(), 2)
+
+        history_thread = threads.GetThreadAtIndex(0)
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
+        history_thread = threads.GetThreadAtIndex(1)
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
+        # let's free the container (SBThreadCollection) and see if the
+        # SBThreads still live
+        threads = None
+        self.assertTrue(history_thread.num_frames >= 2)
+        self.assertEqual(
+            history_thread.frames[1].GetLineEntry().GetFileSpec().GetFilename(),
+            "main.c",
+        )
+
     def asan_tests(self):
         target = self.createTestTarget()
 
diff --git a/lldb/test/API/functionalities/asan/TestReportData.py b/lldb/test/API/functionalities/asan/TestReportData.py
index 543c5fe66a208..5e4c179e2a481 100644
--- a/lldb/test/API/functionalities/asan/TestReportData.py
+++ b/lldb/test/API/functionalities/asan/TestReportData.py
@@ -8,7 +8,7 @@
 from lldbsuite.test.decorators import *
 from lldbsuite.test.lldbtest import *
 from lldbsuite.test import lldbutil
-
+from lldbsuite.test_event.build_exception import BuildError
 
 class AsanTestReportDataCase(TestBase):
     @skipIfFreeBSD  # llvm.org/pr21136 runtimes not yet available by default
@@ -16,9 +16,17 @@ class AsanTestReportDataCase(TestBase):
     @skipUnlessAddressSanitizer
     @skipIf(archs=["i386"], bugnumber="llvm.org/PR36710")
     def test(self):
-        self.build()
+        self.build(make_targets=["asan"])
         self.asan_tests()
 
+    @skipIf(oslist=no_match(["macosx"]))
+    def test_libsanitizers_asan(self):
+        try:
+            self.build(make_targets=["libsanitizers"])
+        except BuildError as e:
+            self.skipTest("failed to build with libsanitizers")
+        self.asan_tests(libsanitizers=True)
+
     def setUp(self):
         # Call super's setUp().
         TestBase.setUp(self)
@@ -29,10 +37,15 @@ def setUp(self):
         self.line_crash = line_number("main.c", "// BOOM line")
         self.col_crash = 16
 
-    def asan_tests(self):
+    def asan_tests(self, libsanitizers=False):
         target = self.createTestTarget()
 
-        self.registerSanitizerLibrariesWithTarget(target)
+        if libsanitizers:
+            self.runCmd(
+                "env SanitizersAddress=1 MallocSanitizerZone=1 MallocSecureAllocator=0"
+            )
+        else:
+            self.registerSanitizerLibrariesWithTarget(target)
 
         self.runCmd("run")
 

From 50a371795bcfe0731f8882e42712dff33cbbef9b Mon Sep 17 00:00:00 2001
From: darkbuck <michael.hliao@gmail.com>
Date: Tue, 16 Apr 2024 19:10:11 -0400
Subject: [PATCH 672/886] [X86] Fix instr desc of CFCMOV's 'mr' variants

- With the memory operand as the destination, 'mr' variants of CFCMOV
  works like STORE and their memory operands should be input operands
  instead of output ones.

Reviewers: XinWang10, arsenm

Pull Request: https://github.com/llvm/llvm-project/pull/88970
---
 llvm/lib/Target/X86/X86InstrCMovSetCC.td | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/X86/X86InstrCMovSetCC.td b/llvm/lib/Target/X86/X86InstrCMovSetCC.td
index 27a0c889a4da3..e27aa4115990e 100644
--- a/llvm/lib/Target/X86/X86InstrCMovSetCC.td
+++ b/llvm/lib/Target/X86/X86InstrCMovSetCC.td
@@ -58,8 +58,8 @@ let SchedRW = [WriteCMOV.Folded, WriteCMOV.ReadAfterFold] in {
 }
 let SchedRW = [WriteCMOV, ReadDefault, ReadDefault, ReadDefault, ReadDefault, ReadDefault],
     Predicates = [HasCMOV, HasCF, In64BitMode], mayStore = 1 in
-  def mr : ITy<0x40, MRMDestMemCC, t, (outs t.MemOperand:$dst),
-                (ins t.RegClass:$src1, ccode:$cond),
+  def mr : ITy<0x40, MRMDestMemCC, t, (outs),
+                (ins t.MemOperand:$dst, t.RegClass:$src1, ccode:$cond),
                 "cfcmov${cond}", unaryop_ndd_args, []>, UseEFLAGS, NF;
 }
 

From 1bc092181bf50d6be95b165e91bd906710710ca7 Mon Sep 17 00:00:00 2001
From: Keith Smiley <keithbsmiley@gmail.com>
Date: Tue, 16 Apr 2024 19:17:27 -0400
Subject: [PATCH 673/886] [bazel] Add support for lldb-server (#88989)

---
 .../llvm-project-overlay/lldb/BUILD.bazel     | 73 ++++++++++++++++++-
 .../lldb/source/Plugins/BUILD.bazel           | 19 +++++
 2 files changed, 90 insertions(+), 2 deletions(-)

diff --git a/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel b/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
index 6dfe8085b9285..1f2b5b476bcc1 100644
--- a/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
@@ -702,6 +702,9 @@ cc_library(
             "//lldb/source/Plugins:PluginSymbolLocatorDebugSymbols",
             "//lldb/source/Plugins:PluginSymbolVendorMacOSX",
         ],
+        "@platforms//os:linux": [
+            "//lldb/source/Plugins:PluginProcessLinux",
+        ],
         "//conditions:default": [],
     }),
 )
@@ -752,7 +755,13 @@ cc_binary(
     data = [
         ":lldb-argdumper",
     ] + select({
-        "@platforms//os:macos": [":debugserver"],
+        "@platforms//os:macos": [
+            ":debugserver",
+            ":lldb-server",
+        ],
+        "@platforms//os:linux": [
+            ":lldb-server",
+        ],
         "//conditions:default": [],
     }),
     deps = [
@@ -799,8 +808,8 @@ cc_library(
         ["tools/debugserver/source/**/*.cpp"],
         exclude = ["tools/debugserver/source/debugserver.cpp"],
     ),
-    tags = ["nobuildkite"],
     local_defines = ["LLDB_USE_OS_LOG"],
+    tags = ["nobuildkite"],
     deps = [
         ":DebugServerCommonHeaders",
         ":DebugServerCommonMacOSXHeaders",
@@ -852,3 +861,63 @@ cc_binary(
     srcs = glob(["tools/argdumper/*.cpp"]),
     deps = ["//llvm:Support"],
 )
+
+gentbl_cc_library(
+    name = "lldb_server_opts_gen",
+    strip_include_prefix = ".",
+    tbl_outs = [(
+        ["-gen-opt-parser-defs"],
+        "LLGSOptions.inc",
+    )],
+    tblgen = "//llvm:llvm-tblgen",
+    td_file = "tools/lldb-server/LLGSOptions.td",
+    deps = ["//llvm:OptParserTdFiles"],
+)
+
+cc_binary(
+    name = "lldb-server",
+    srcs = glob([
+        "tools/lldb-server/*.cpp",
+        "tools/lldb-server/*.h",
+    ]),
+    target_compatible_with = select({
+        "@platforms//os:linux": [],
+        "@platforms//os:macos": [],
+        # TODO: This can theoretically support more platforms, but it hasn't been tested yet
+        "//conditions:default": ["@platforms//:incompatible"],
+    }),
+    deps = [
+        ":Host",
+        ":Initialization",
+        ":Utility",
+        ":Version",
+        ":lldb_server_opts_gen",
+        "//lldb:Target",
+        "//lldb:TargetHeaders",
+        "//lldb/source/Plugins:PluginCPlusPlusLanguage",
+        "//lldb/source/Plugins:PluginExpressionParserClang",
+        "//lldb/source/Plugins:PluginInstructionARM",
+        "//lldb/source/Plugins:PluginInstructionARM64",
+        "//lldb/source/Plugins:PluginInstructionLoongArch",
+        "//lldb/source/Plugins:PluginInstructionMIPS",
+        "//lldb/source/Plugins:PluginInstructionMIPS64",
+        "//lldb/source/Plugins:PluginInstructionRISCV",
+        "//lldb/source/Plugins:PluginObjCLanguage",
+        "//lldb/source/Plugins:PluginProcessGDBRemote",
+        "//lldb/source/Plugins:PluginSymbolFileDWARF",
+        "//lldb/source/Plugins:PluginSymbolFileNativePDB",
+        "//lldb/source/Plugins:PluginSymbolFilePDB",
+        "//lldb/source/Plugins:PluginTypeSystemClang",
+        "//llvm:Option",
+        "//llvm:Support",
+    ] + select({
+        "@platforms//os:linux": [
+            "//lldb/source/Plugins:PluginObjectFileELF",
+            "//lldb/source/Plugins:PluginProcessLinux",
+        ],
+        "@platforms//os:macos": [
+            "//lldb/source/Plugins:PluginObjectFileMachO",
+        ],
+        "//conditions:default": [],
+    }),
+)
diff --git a/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel b/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
index bbc523f54a190..b5f5bed1698a6 100644
--- a/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
@@ -2100,6 +2100,25 @@ cc_library(
     ],
 )
 
+cc_library(
+    name = "PluginProcessLinux",
+    srcs = glob(["Process/Linux/*.cpp"]),
+    hdrs = glob(["Process/Linux/*.h"]),
+    include_prefix = "Plugins",
+    deps = [
+        ":PluginProcessPOSIX",
+        ":PluginProcessUtility",
+        "//lldb:Core",
+        "//lldb:Headers",
+        "//lldb:Host",
+        "//lldb:SymbolHeaders",
+        "//lldb:TargetHeaders",
+        "//lldb:Utility",
+        "//llvm:Support",
+        "//llvm:TargetParser",
+    ],
+)
+
 cc_library(
     name = "PluginScriptedProcess",
     srcs = glob(["Process/scripted/*.cpp"]),

From be50a259f1fe77240b000f6b695b9b6394f4936b Mon Sep 17 00:00:00 2001
From: Andy Kaylor <andrew.kaylor@intel.com>
Date: Tue, 16 Apr 2024 16:22:31 -0700
Subject: [PATCH 674/886] Update foldFMulReassoc to respect absent fast-math
 flags (#88589)

This change updates a few of the transformations in foldFMulReassoc to
respect absent fast-math flags in cases where fmul and fdiv, fadd, or fsub
instructions were being folded but the code was only checking for
fast-math flags on the fmul instruction and was transferring flags to
the folded instruction that were not present on the other original
instructions.

This fixes https://github.com/llvm/llvm-project/issues/82857
---
 llvm/include/llvm/IR/InstrTypes.h             |  27 +++++
 .../InstCombine/InstCombineMulDivRem.cpp      |  38 +++++--
 llvm/test/Transforms/InstCombine/fast-math.ll |   4 +-
 llvm/test/Transforms/InstCombine/fmul-pow.ll  |  30 ++---
 llvm/test/Transforms/InstCombine/fmul.ll      | 104 ++++++++++++------
 5 files changed, 141 insertions(+), 62 deletions(-)

diff --git a/llvm/include/llvm/IR/InstrTypes.h b/llvm/include/llvm/IR/InstrTypes.h
index cfe1b11ade5a4..8e6bef69218c2 100644
--- a/llvm/include/llvm/IR/InstrTypes.h
+++ b/llvm/include/llvm/IR/InstrTypes.h
@@ -24,6 +24,7 @@
 #include "llvm/IR/Attributes.h"
 #include "llvm/IR/CallingConv.h"
 #include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/FMF.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/Instruction.h"
 #include "llvm/IR/LLVMContext.h"
@@ -311,6 +312,32 @@ class BinaryOperator : public Instruction {
     return BO;
   }
 
+  static BinaryOperator *CreateWithFMF(BinaryOps Opc, Value *V1, Value *V2,
+                                       FastMathFlags FMF,
+                                       const Twine &Name = "",
+                                       Instruction *InsertBefore = nullptr) {
+    BinaryOperator *BO = Create(Opc, V1, V2, Name, InsertBefore);
+    BO->setFastMathFlags(FMF);
+    return BO;
+  }
+
+  static BinaryOperator *CreateFAddFMF(Value *V1, Value *V2, FastMathFlags FMF,
+                                       const Twine &Name = "") {
+    return CreateWithFMF(Instruction::FAdd, V1, V2, FMF, Name);
+  }
+  static BinaryOperator *CreateFSubFMF(Value *V1, Value *V2, FastMathFlags FMF,
+                                       const Twine &Name = "") {
+    return CreateWithFMF(Instruction::FSub, V1, V2, FMF, Name);
+  }
+  static BinaryOperator *CreateFMulFMF(Value *V1, Value *V2, FastMathFlags FMF,
+                                       const Twine &Name = "") {
+    return CreateWithFMF(Instruction::FMul, V1, V2, FMF, Name);
+  }
+  static BinaryOperator *CreateFDivFMF(Value *V1, Value *V2, FastMathFlags FMF,
+                                       const Twine &Name = "") {
+    return CreateWithFMF(Instruction::FDiv, V1, V2, FMF, Name);
+  }
+
   static BinaryOperator *CreateFAddFMF(Value *V1, Value *V2,
                                        Instruction *FMFSource,
                                        const Twine &Name = "") {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 48372381a0d1c..7b86fcde8937b 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -624,31 +624,38 @@ Instruction *InstCombinerImpl::foldFMulReassoc(BinaryOperator &I) {
   Value *Op1 = I.getOperand(1);
   Value *X, *Y;
   Constant *C;
+  BinaryOperator *Op0BinOp;
 
   // Reassociate constant RHS with another constant to form constant
   // expression.
-  if (match(Op1, m_Constant(C)) && C->isFiniteNonZeroFP()) {
+  if (match(Op1, m_Constant(C)) && C->isFiniteNonZeroFP() &&
+      match(Op0, m_AllowReassoc(m_BinOp(Op0BinOp)))) {
+    // Everything in this scope folds I with Op0, intersecting their FMF.
+    FastMathFlags FMF = I.getFastMathFlags() & Op0BinOp->getFastMathFlags();
+    IRBuilder<>::FastMathFlagGuard FMFGuard(Builder);
+    Builder.setFastMathFlags(FMF);
     Constant *C1;
     if (match(Op0, m_OneUse(m_FDiv(m_Constant(C1), m_Value(X))))) {
       // (C1 / X) * C --> (C * C1) / X
       Constant *CC1 =
           ConstantFoldBinaryOpOperands(Instruction::FMul, C, C1, DL);
       if (CC1 && CC1->isNormalFP())
-        return BinaryOperator::CreateFDivFMF(CC1, X, &I);
+        return BinaryOperator::CreateFDivFMF(CC1, X, FMF);
     }
     if (match(Op0, m_FDiv(m_Value(X), m_Constant(C1)))) {
+      // FIXME: This seems like it should also be checking for arcp
       // (X / C1) * C --> X * (C / C1)
       Constant *CDivC1 =
           ConstantFoldBinaryOpOperands(Instruction::FDiv, C, C1, DL);
       if (CDivC1 && CDivC1->isNormalFP())
-        return BinaryOperator::CreateFMulFMF(X, CDivC1, &I);
+        return BinaryOperator::CreateFMulFMF(X, CDivC1, FMF);
 
       // If the constant was a denormal, try reassociating differently.
       // (X / C1) * C --> X / (C1 / C)
       Constant *C1DivC =
           ConstantFoldBinaryOpOperands(Instruction::FDiv, C1, C, DL);
       if (C1DivC && Op0->hasOneUse() && C1DivC->isNormalFP())
-        return BinaryOperator::CreateFDivFMF(X, C1DivC, &I);
+        return BinaryOperator::CreateFDivFMF(X, C1DivC, FMF);
     }
 
     // We do not need to match 'fadd C, X' and 'fsub X, C' because they are
@@ -658,26 +665,33 @@ Instruction *InstCombinerImpl::foldFMulReassoc(BinaryOperator &I) {
       // (X + C1) * C --> (X * C) + (C * C1)
       if (Constant *CC1 =
               ConstantFoldBinaryOpOperands(Instruction::FMul, C, C1, DL)) {
-        Value *XC = Builder.CreateFMulFMF(X, C, &I);
-        return BinaryOperator::CreateFAddFMF(XC, CC1, &I);
+        Value *XC = Builder.CreateFMul(X, C);
+        return BinaryOperator::CreateFAddFMF(XC, CC1, FMF);
       }
     }
     if (match(Op0, m_OneUse(m_FSub(m_Constant(C1), m_Value(X))))) {
       // (C1 - X) * C --> (C * C1) - (X * C)
       if (Constant *CC1 =
               ConstantFoldBinaryOpOperands(Instruction::FMul, C, C1, DL)) {
-        Value *XC = Builder.CreateFMulFMF(X, C, &I);
-        return BinaryOperator::CreateFSubFMF(CC1, XC, &I);
+        Value *XC = Builder.CreateFMul(X, C);
+        return BinaryOperator::CreateFSubFMF(CC1, XC, FMF);
       }
     }
   }
 
   Value *Z;
   if (match(&I,
-            m_c_FMul(m_OneUse(m_FDiv(m_Value(X), m_Value(Y))), m_Value(Z)))) {
-    // Sink division: (X / Y) * Z --> (X * Z) / Y
-    Value *NewFMul = Builder.CreateFMulFMF(X, Z, &I);
-    return BinaryOperator::CreateFDivFMF(NewFMul, Y, &I);
+            m_c_FMul(m_AllowReassoc(m_OneUse(m_FDiv(m_Value(X), m_Value(Y)))),
+                     m_Value(Z)))) {
+    BinaryOperator *DivOp = cast<BinaryOperator>(((Z == Op0) ? Op1 : Op0));
+    FastMathFlags FMF = I.getFastMathFlags() & DivOp->getFastMathFlags();
+    if (FMF.allowReassoc()) {
+      // Sink division: (X / Y) * Z --> (X * Z) / Y
+      IRBuilder<>::FastMathFlagGuard FMFGuard(Builder);
+      Builder.setFastMathFlags(FMF);
+      auto *NewFMul = Builder.CreateFMul(X, Z);
+      return BinaryOperator::CreateFDivFMF(NewFMul, Y, FMF);
+    }
   }
 
   // sqrt(X) * sqrt(Y) -> sqrt(X * Y)
diff --git a/llvm/test/Transforms/InstCombine/fast-math.ll b/llvm/test/Transforms/InstCombine/fast-math.ll
index 129d7811cfb86..916955e34efac 100644
--- a/llvm/test/Transforms/InstCombine/fast-math.ll
+++ b/llvm/test/Transforms/InstCombine/fast-math.ll
@@ -562,7 +562,7 @@ define float @fdiv1(float %x) {
 ; CHECK-NEXT:    [[DIV1:%.*]] = fmul fast float [[X:%.*]], 0x3FD7303B60000000
 ; CHECK-NEXT:    ret float [[DIV1]]
 ;
-  %div = fdiv float %x, 0x3FF3333340000000
+  %div = fdiv fast float %x, 0x3FF3333340000000
   %div1 = fdiv fast float %div, 0x4002666660000000
   ret float %div1
 ; 0x3FF3333340000000 = 1.2f
@@ -603,7 +603,7 @@ define float @fdiv3(float %x) {
 ; CHECK-NEXT:    [[DIV1:%.*]] = fdiv fast float [[TMP1]], 0x47EFFFFFE0000000
 ; CHECK-NEXT:    ret float [[DIV1]]
 ;
-  %div = fdiv float %x, 0x47EFFFFFE0000000
+  %div = fdiv fast float %x, 0x47EFFFFFE0000000
   %div1 = fdiv fast float %div, 0x4002666660000000
   ret float %div1
 }
diff --git a/llvm/test/Transforms/InstCombine/fmul-pow.ll b/llvm/test/Transforms/InstCombine/fmul-pow.ll
index 63458e136074c..84592d220d62c 100644
--- a/llvm/test/Transforms/InstCombine/fmul-pow.ll
+++ b/llvm/test/Transforms/InstCombine/fmul-pow.ll
@@ -85,8 +85,8 @@ define double @pow_ab_recip_a_reassoc(double %a, double %b)  {
 ; CHECK-NEXT:    [[M:%.*]] = call reassoc double @llvm.pow.f64(double [[A:%.*]], double [[TMP1]])
 ; CHECK-NEXT:    ret double [[M]]
 ;
-  %r = fdiv double 1.0, %a
-  %p = call double @llvm.pow.f64(double %a, double %b)
+  %r = fdiv reassoc double 1.0, %a
+  %p = call reassoc double @llvm.pow.f64(double %a, double %b)
   %m = fmul reassoc double %r, %p
   ret double %m
 }
@@ -99,8 +99,8 @@ define double @pow_ab_recip_a_reassoc_commute(double %a, double %b)  {
 ; CHECK-NEXT:    [[M:%.*]] = call reassoc double @llvm.pow.f64(double [[A:%.*]], double [[TMP1]])
 ; CHECK-NEXT:    ret double [[M]]
 ;
-  %r = fdiv double 1.0, %a
-  %p = call double @llvm.pow.f64(double %a, double %b)
+  %r = fdiv reassoc double 1.0, %a
+  %p = call reassoc double @llvm.pow.f64(double %a, double %b)
   %m = fmul reassoc double %p, %r
   ret double %m
 }
@@ -109,14 +109,14 @@ define double @pow_ab_recip_a_reassoc_commute(double %a, double %b)  {
 
 define double @pow_ab_recip_a_reassoc_use1(double %a, double %b)  {
 ; CHECK-LABEL: @pow_ab_recip_a_reassoc_use1(
-; CHECK-NEXT:    [[R:%.*]] = fdiv double 1.000000e+00, [[A:%.*]]
-; CHECK-NEXT:    [[P:%.*]] = call double @llvm.pow.f64(double [[A]], double [[B:%.*]])
+; CHECK-NEXT:    [[R:%.*]] = fdiv reassoc double 1.000000e+00, [[A:%.*]]
+; CHECK-NEXT:    [[P:%.*]] = call reassoc double @llvm.pow.f64(double [[A]], double [[B:%.*]])
 ; CHECK-NEXT:    [[M:%.*]] = fmul reassoc double [[R]], [[P]]
 ; CHECK-NEXT:    call void @use(double [[R]])
 ; CHECK-NEXT:    ret double [[M]]
 ;
-  %r = fdiv double 1.0, %a
-  %p = call double @llvm.pow.f64(double %a, double %b)
+  %r = fdiv reassoc double 1.0, %a
+  %p = call reassoc double @llvm.pow.f64(double %a, double %b)
   %m = fmul reassoc double %r, %p
   call void @use(double %r)
   ret double %m
@@ -126,13 +126,13 @@ define double @pow_ab_recip_a_reassoc_use1(double %a, double %b)  {
 
 define double @pow_ab_recip_a_reassoc_use2(double %a, double %b)  {
 ; CHECK-LABEL: @pow_ab_recip_a_reassoc_use2(
-; CHECK-NEXT:    [[P:%.*]] = call double @llvm.pow.f64(double [[A:%.*]], double [[B:%.*]])
+; CHECK-NEXT:    [[P:%.*]] = call reassoc double @llvm.pow.f64(double [[A:%.*]], double [[B:%.*]])
 ; CHECK-NEXT:    [[M:%.*]] = fdiv reassoc double [[P]], [[A]]
 ; CHECK-NEXT:    call void @use(double [[P]])
 ; CHECK-NEXT:    ret double [[M]]
 ;
-  %r = fdiv double 1.0, %a
-  %p = call double @llvm.pow.f64(double %a, double %b)
+  %r = fdiv reassoc double 1.0, %a
+  %p = call reassoc double @llvm.pow.f64(double %a, double %b)
   %m = fmul reassoc double %r, %p
   call void @use(double %p)
   ret double %m
@@ -142,15 +142,15 @@ define double @pow_ab_recip_a_reassoc_use2(double %a, double %b)  {
 
 define double @pow_ab_recip_a_reassoc_use3(double %a, double %b)  {
 ; CHECK-LABEL: @pow_ab_recip_a_reassoc_use3(
-; CHECK-NEXT:    [[R:%.*]] = fdiv double 1.000000e+00, [[A:%.*]]
-; CHECK-NEXT:    [[P:%.*]] = call double @llvm.pow.f64(double [[A]], double [[B:%.*]])
+; CHECK-NEXT:    [[R:%.*]] = fdiv reassoc double 1.000000e+00, [[A:%.*]]
+; CHECK-NEXT:    [[P:%.*]] = call reassoc double @llvm.pow.f64(double [[A]], double [[B:%.*]])
 ; CHECK-NEXT:    [[M:%.*]] = fmul reassoc double [[R]], [[P]]
 ; CHECK-NEXT:    call void @use(double [[R]])
 ; CHECK-NEXT:    call void @use(double [[P]])
 ; CHECK-NEXT:    ret double [[M]]
 ;
-  %r = fdiv double 1.0, %a
-  %p = call double @llvm.pow.f64(double %a, double %b)
+  %r = fdiv reassoc double 1.0, %a
+  %p = call reassoc double @llvm.pow.f64(double %a, double %b)
   %m = fmul reassoc double %r, %p
   call void @use(double %r)
   call void @use(double %p)
diff --git a/llvm/test/Transforms/InstCombine/fmul.ll b/llvm/test/Transforms/InstCombine/fmul.ll
index f6435f0032891..39f9e74f899d1 100644
--- a/llvm/test/Transforms/InstCombine/fmul.ll
+++ b/llvm/test/Transforms/InstCombine/fmul.ll
@@ -633,15 +633,15 @@ define float @log2half(float %x, float %y) {
 
 define float @log2half_commute(float %x1, float %y) {
 ; CHECK-LABEL: @log2half_commute(
+; CHECK-NEXT:    [[X1:%.*]] = fmul fast float [[X2:%.*]], 0x3FC24924A0000000
 ; CHECK-NEXT:    [[TMP1:%.*]] = call fast float @llvm.log2.f32(float [[Y:%.*]])
-; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast float [[TMP1]], [[X1:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast float [[TMP1]], [[X1]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = fsub fast float [[TMP2]], [[X1]]
-; CHECK-NEXT:    [[MUL:%.*]] = fmul fast float [[TMP3]], 0x3FC24924A0000000
-; CHECK-NEXT:    ret float [[MUL]]
+; CHECK-NEXT:    ret float [[TMP3]]
 ;
-  %x = fdiv float %x1, 7.0 ; thwart complexity-based canonicalization
-  %halfy = fmul float %y, 0.5
-  %log2 = call float @llvm.log2.f32(float %halfy)
+  %x = fdiv fast float %x1, 7.0 ; thwart complexity-based canonicalization
+  %halfy = fmul fast float %y, 0.5
+  %log2 = call fast float @llvm.log2.f32(float %halfy)
   %mul = fmul fast float %x, %log2
   ret float %mul
 }
@@ -652,12 +652,50 @@ define float @fdiv_constant_numerator_fmul(float %x) {
 ; CHECK-LABEL: @fdiv_constant_numerator_fmul(
 ; CHECK-NEXT:    [[T3:%.*]] = fdiv reassoc float 1.200000e+07, [[X:%.*]]
 ; CHECK-NEXT:    ret float [[T3]]
+;
+  %t1 = fdiv reassoc float 2.0e+3, %x
+  %t3 = fmul reassoc float %t1, 6.0e+3
+  ret float %t3
+}
+
+; C1/X * C2 => (C1*C2) / X with mixed fast-math flags
+
+define float @fdiv_constant_numerator_fmul_mixed(float %x) {
+; CHECK-LABEL: @fdiv_constant_numerator_fmul_mixed(
+; CHECK-NEXT:    [[T3:%.*]] = fdiv reassoc float 1.200000e+07, [[X:%.*]]
+; CHECK-NEXT:    ret float [[T3]]
+;
+  %t1 = fdiv reassoc float 2.0e+3, %x
+  %t3 = fmul fast float %t1, 6.0e+3
+  ret float %t3
+}
+
+; C1/X * C2 => (C1*C2) / X with full fast-math flags
+
+define float @fdiv_constant_numerator_fmul_fast(float %x) {
+; CHECK-LABEL: @fdiv_constant_numerator_fmul_fast(
+; CHECK-NEXT:    [[T3:%.*]] = fdiv fast float 1.200000e+07, [[X:%.*]]
+; CHECK-NEXT:    ret float [[T3]]
+;
+  %t1 = fdiv fast float 2.0e+3, %x
+  %t3 = fmul fast float %t1, 6.0e+3
+  ret float %t3
+}
+
+; C1/X * C2 => (C1*C2) / X with no fast-math flags on the fdiv
+
+define float @fdiv_constant_numerator_fmul_precdiv(float %x) {
+; CHECK-LABEL: @fdiv_constant_numerator_fmul_precdiv(
+; CHECK-NEXT:    [[T1:%.*]] = fdiv float 2.000000e+03, [[X:%.*]]
+; CHECK-NEXT:    [[T4:%.*]] = fmul reassoc float [[T1]], 6.000000e+03
+; CHECK-NEXT:    ret float [[T4]]
 ;
   %t1 = fdiv float 2.0e+3, %x
   %t3 = fmul reassoc float %t1, 6.0e+3
   ret float %t3
 }
 
+
 ; C1/X * C2 => (C1*C2) / X is disabled if C1/X has multiple uses
 
 @fmul2_external = external global float
@@ -682,7 +720,7 @@ define float @fdiv_constant_denominator_fmul(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fmul reassoc float [[X:%.*]], 3.000000e+00
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fdiv float %x, 2.0e+3
+  %t1 = fdiv reassoc float %x, 2.0e+3
   %t3 = fmul reassoc float %t1, 6.0e+3
   ret float %t3
 }
@@ -692,7 +730,7 @@ define <4 x float> @fdiv_constant_denominator_fmul_vec(<4 x float> %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fmul reassoc <4 x float> [[X:%.*]], <float 3.000000e+00, float 2.000000e+00, float 1.000000e+00, float 1.000000e+00>
 ; CHECK-NEXT:    ret <4 x float> [[T3]]
 ;
-  %t1 = fdiv <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3>
+  %t1 = fdiv reassoc <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3>
   %t3 = fmul reassoc <4 x float> %t1, <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3>
   ret <4 x float> %t3
 }
@@ -705,7 +743,7 @@ define <4 x float> @fdiv_constant_denominator_fmul_vec_constexpr(<4 x float> %x)
 ; CHECK-NEXT:    ret <4 x float> [[T3]]
 ;
   %constExprMul = bitcast i128 trunc (i160 bitcast (<5 x float> <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3, float undef> to i160) to i128) to <4 x float>
-  %t1 = fdiv <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3>
+  %t1 = fdiv reassoc <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3>
   %t3 = fmul reassoc <4 x float> %t1, %constExprMul
   ret <4 x float> %t3
 }
@@ -734,7 +772,7 @@ define float @fdiv_constant_denominator_fmul_denorm(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fmul fast float [[X:%.*]], 0x3760620000000000
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fdiv float %x, 2.0e+3
+  %t1 = fdiv fast float %x, 2.0e+3
   %t3 = fmul fast float %t1, 0x3810000000000000
   ret float %t3
 }
@@ -748,7 +786,7 @@ define float @fdiv_constant_denominator_fmul_denorm_try_harder(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fdiv reassoc float [[X:%.*]], 0x47E8000000000000
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fdiv float %x, 3.0
+  %t1 = fdiv reassoc float %x, 3.0
   %t3 = fmul reassoc float %t1, 0x3810000000000000
   ret float %t3
 }
@@ -776,7 +814,7 @@ define float @fmul_fadd_distribute(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc float [[TMP1]], 6.000000e+00
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t2 = fadd float %x, 2.0
+  %t2 = fadd reassoc float %x, 2.0
   %t3 = fmul reassoc float %t2, 3.0
   ret float %t3
 }
@@ -787,7 +825,7 @@ define <2 x float> @fmul_fadd_distribute_vec(<2 x float> %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc <2 x float> [[TMP1]], <float 1.200000e+07, float 1.200000e+07>
 ; CHECK-NEXT:    ret <2 x float> [[T3]]
 ;
-  %t1 = fadd <2 x float> <float 2.0e+3, float 2.0e+3>, %x
+  %t1 = fadd reassoc <2 x float> <float 2.0e+3, float 2.0e+3>, %x
   %t3 = fmul reassoc <2 x float> %t1, <float 6.0e+3, float 6.0e+3>
   ret <2 x float> %t3
 }
@@ -798,7 +836,7 @@ define <vscale x 2 x float> @fmul_fadd_distribute_scalablevec(<vscale x 2 x floa
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc <vscale x 2 x float> [[TMP1]], shufflevector (<vscale x 2 x float> insertelement (<vscale x 2 x float> poison, float 1.200000e+07, i64 0), <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer)
 ; CHECK-NEXT:    ret <vscale x 2 x float> [[T3]]
 ;
-  %t1 = fadd <vscale x 2 x float> splat (float 2.0e+3), %x
+  %t1 = fadd reassoc <vscale x 2 x float> splat (float 2.0e+3), %x
   %t3 = fmul reassoc <vscale x 2 x float> %t1, splat (float 6.0e+3)
 
 
@@ -813,7 +851,7 @@ define float @fmul_fsub_distribute1(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc float [[TMP1]], -6.000000e+00
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t2 = fsub float %x, 2.0
+  %t2 = fsub reassoc float %x, 2.0
   %t3 = fmul reassoc float %t2, 3.0
   ret float %t3
 }
@@ -826,7 +864,7 @@ define float @fmul_fsub_distribute2(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fsub reassoc float 6.000000e+00, [[TMP1]]
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t2 = fsub float 2.0, %x
+  %t2 = fsub reassoc float 2.0, %x
   %t3 = fmul reassoc float %t2, 3.0
   ret float %t3
 }
@@ -840,8 +878,8 @@ define float @fmul_fadd_fmul_distribute(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd fast float [[TMP1]], 1.000000e+01
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fmul float %x, 6.0
-  %t2 = fadd float %t1, 2.0
+  %t1 = fmul fast float %x, 6.0
+  %t2 = fadd fast float %t1, 2.0
   %t3 = fmul fast float %t2, 5.0
   ret float %t3
 }
@@ -872,8 +910,8 @@ define double @fmul_fadd_fdiv_distribute2(double %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc double [[TMP1]], 0x34000000000000
 ; CHECK-NEXT:    ret double [[T3]]
 ;
-  %t1 = fdiv double %x, 3.0
-  %t2 = fadd double %t1, 5.0
+  %t1 = fdiv reassoc double %x, 3.0
+  %t2 = fadd reassoc double %t1, 5.0
   %t3 = fmul reassoc double %t2, 0x10000000000000
   ret double %t3
 }
@@ -887,8 +925,8 @@ define double @fmul_fadd_fdiv_distribute3(double %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd reassoc double [[TMP1]], 0x34000000000000
 ; CHECK-NEXT:    ret double [[T3]]
 ;
-  %t1 = fdiv double %x, 3.0
-  %t2 = fadd double %t1, 5.0
+  %t1 = fdiv reassoc double %x, 3.0
+  %t2 = fadd reassoc double %t1, 5.0
   %t3 = fmul reassoc double %t2, 0x10000000000000
   ret double %t3
 }
@@ -902,8 +940,8 @@ define float @fmul_fsub_fmul_distribute(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fsub fast float 1.000000e+01, [[TMP1]]
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fmul float %x, 6.0
-  %t2 = fsub float 2.0, %t1
+  %t1 = fmul fast float %x, 6.0
+  %t2 = fsub fast float 2.0, %t1
   %t3 = fmul fast float %t2, 5.0
   ret float %t3
 }
@@ -932,8 +970,8 @@ define float @fmul_fsub_fmul_distribute2(float %x) {
 ; CHECK-NEXT:    [[T3:%.*]] = fadd fast float [[TMP1]], -1.000000e+01
 ; CHECK-NEXT:    ret float [[T3]]
 ;
-  %t1 = fmul float %x, 6.0
-  %t2 = fsub float %t1, 2.0
+  %t1 = fmul fast float %x, 6.0
+  %t2 = fsub fast float %t1, 2.0
   %t3 = fmul fast float %t2, 5.0
   ret float %t3
 }
@@ -986,8 +1024,8 @@ define double @fmul_fdivs_factor_common_denominator(double %x, double %y, double
 ; CHECK-NEXT:    [[MUL:%.*]] = fdiv fast double [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret double [[MUL]]
 ;
-  %div1 = fdiv double %x, %z
-  %div2 = fdiv double %y, %z
+  %div1 = fdiv fast double %x, %z
+  %div2 = fdiv fast double %y, %z
   %mul = fmul fast double %div1, %div2
   ret double %mul
 }
@@ -999,8 +1037,8 @@ define double @fmul_fdivs_factor(double %x, double %y, double %z, double %w) {
 ; CHECK-NEXT:    [[MUL:%.*]] = fdiv reassoc double [[TMP2]], [[Y:%.*]]
 ; CHECK-NEXT:    ret double [[MUL]]
 ;
-  %div1 = fdiv double %x, %y
-  %div2 = fdiv double %z, %w
+  %div1 = fdiv reassoc double %x, %y
+  %div2 = fdiv reassoc double %z, %w
   %mul = fmul reassoc double %div1, %div2
   ret double %mul
 }
@@ -1011,7 +1049,7 @@ define double @fmul_fdiv_factor(double %x, double %y, double %z) {
 ; CHECK-NEXT:    [[MUL:%.*]] = fdiv reassoc double [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    ret double [[MUL]]
 ;
-  %div = fdiv double %x, %y
+  %div = fdiv reassoc double %x, %y
   %mul = fmul reassoc double %div, %z
   ret double %mul
 }
@@ -1022,7 +1060,7 @@ define double @fmul_fdiv_factor_constant1(double %x, double %y) {
 ; CHECK-NEXT:    [[MUL:%.*]] = fdiv reassoc double [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    ret double [[MUL]]
 ;
-  %div = fdiv double %x, %y
+  %div = fdiv reassoc double %x, %y
   %mul = fmul reassoc double %div, 42.0
   ret double %mul
 }
@@ -1033,7 +1071,7 @@ define <2 x float> @fmul_fdiv_factor_constant2(<2 x float> %x, <2 x float> %y) {
 ; CHECK-NEXT:    [[MUL:%.*]] = fdiv reassoc <2 x float> [[TMP1]], <float 4.200000e+01, float 1.200000e+01>
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
-  %div = fdiv <2 x float> %x, <float 42.0, float 12.0>
+  %div = fdiv reassoc <2 x float> %x, <float 42.0, float 12.0>
   %mul = fmul reassoc <2 x float> %div, %y
   ret <2 x float> %mul
 }

From ce5381e22a50f354cf3d1763589f1daf155c481b Mon Sep 17 00:00:00 2001
From: Prashant Kumar <pk5561@gmail.com>
Date: Wed, 17 Apr 2024 05:36:40 +0530
Subject: [PATCH 675/886] =?UTF-8?q?[mlir][vector]=20Determine=20vector=20s?=
 =?UTF-8?q?izes=20from=20the=20result=20shape=20in=20the=20ca=E2=80=A6=20(?=
 =?UTF-8?q?#88249)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

…se of tensor pack

When the vector sizes are not passed as inputs to the vector transform
operation, the vector sizes are queried from the static result shape in
the case of tensor.pack op.
---
 .../Linalg/Transforms/Vectorization.cpp       | 54 +++++++++++++-----
 .../Linalg/vectorization-unsupported.mlir     | 17 ++++++
 mlir/test/Dialect/Linalg/vectorization.mlir   | 55 +++++++++++++++++++
 3 files changed, 113 insertions(+), 13 deletions(-)

diff --git a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
index 25785653a7167..df61381432921 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
@@ -1412,10 +1412,11 @@ static SmallVector<int64_t> getTiledPackShape(tensor::PackOp packOp,
 
 /// Create a TransferReadOp from `source` with static shape `readShape`. If the
 /// vector type for the read is not the same as the type of `source`, then a
-/// mask is created on the read.
+/// mask is created on the read.  If `doMasking` parameter is set to false we
+/// update the `inBounds` attribute instead of masking.
 static Value createReadOrMaskedRead(OpBuilder &builder, Location loc,
                                     Value source, ArrayRef<int64_t> readShape,
-                                    Value padValue) {
+                                    Value padValue, bool doMasking = true) {
   assert(llvm::none_of(readShape,
                        [](int64_t s) { return s == ShapedType::kDynamic; }));
   auto sourceShape = dyn_cast<ShapedType>(source.getType()).getShape();
@@ -1424,14 +1425,21 @@ static Value createReadOrMaskedRead(OpBuilder &builder, Location loc,
   auto vectorType = VectorType::get(readShape, padValue.getType());
   int64_t readRank = readShape.size();
   auto zero = builder.create<arith::ConstantIndexOp>(loc, 0);
+  SmallVector<bool> inBoundsVal(readRank, true);
+  if (!doMasking) {
+    // Update the inBounds attribute.
+    for (unsigned i = 0; i < readRank; i++)
+      inBoundsVal[i] = sourceShape[i] == readShape[i];
+  }
   auto transferReadOp = builder.create<vector::TransferReadOp>(
       loc,
       /*vectorType=*/vectorType,
       /*source=*/source,
       /*indices=*/SmallVector<Value>(readRank, zero),
       /*padding=*/padValue,
-      /*inBounds=*/SmallVector<bool>(readRank, true));
-  if (llvm::equal(readShape, sourceShape)) {
+      /*inBounds=*/inBoundsVal);
+
+  if (llvm::equal(readShape, sourceShape) || !doMasking) {
     return transferReadOp;
   }
   SmallVector<OpFoldResult> mixedSourceDims =
@@ -1482,11 +1490,10 @@ static Operation *createWriteOrMaskedWrite(OpBuilder &builder, Location loc,
   return write;
 }
 
-/// Vectorize tensor::PackOp with (1) static innerTiles and (2) constant
-/// padding value into:
+/// Vectorize tensor::PackOp with (1) static innerTiles (2) constant
+/// padding value and (3) input vector sizes into:
 /// masked_transfer_read->shape_cast->transpose->transfer_write_in_bounds
 /// As in the following example:
-///
 /// %pack = tensor.pack %src inner_dims_pos = [2, 1] inner_tiles = [16, 2]
 ///     into %dst : tensor<32x8x16xf32> -> tensor<32x4x1x16x2xf32>
 ///
@@ -1505,6 +1512,10 @@ static Operation *createWriteOrMaskedWrite(OpBuilder &builder, Location loc,
 ///     %empty[%c0_0, %c0_0, %c0_0, %c0_0, %c0_0]
 ///     {in_bounds = [true, true, true, true, true]}
 ///     : vector<32x4x1x16x2xf32>, tensor<32x4x1x16x2xf32>
+///
+/// If the (3) input vector sizes are not provided, the vector sizes are
+/// determined by the result tensor shape. Also, we update the inBounds
+/// attribute instead of masking.
 static LogicalResult
 vectorizeAsTensorPackOp(RewriterBase &rewriter, tensor::PackOp packOp,
                         ArrayRef<int64_t> inputVectorSizes,
@@ -1525,6 +1536,16 @@ vectorizeAsTensorPackOp(RewriterBase &rewriter, tensor::PackOp packOp,
   (void)status; // prevent unused variable warning on non-assert builds.
   assert(succeeded(status) && "failed to reify result shapes");
 
+  // If the input vector sizes are not provided, then the vector sizes are
+  // determined by the result tensor shape. In case the vector sizes aren't
+  // provided, we update the inBounds attribute instead of masking.
+  bool doMasking = true;
+  if (inputVectorSizes.empty()) {
+    ArrayRef<int64_t> resultTensorShape = packOp.getDestType().getShape();
+    inputVectorSizes = resultTensorShape.take_front(packOp.getSourceRank());
+    doMasking = false;
+  }
+
   // Create masked TransferReadOp.
   SmallVector<int64_t> inputShape(inputVectorSizes);
   auto innerTiles = packOp.getStaticInnerTiles();
@@ -1536,7 +1557,7 @@ vectorizeAsTensorPackOp(RewriterBase &rewriter, tensor::PackOp packOp,
   for (auto [idx, size] : enumerate(innerTiles))
     inputShape[innerDimsPos[idx]] *= size;
   auto maskedRead = createReadOrMaskedRead(rewriter, loc, packOp.getSource(),
-                                           inputShape, padValue);
+                                           inputShape, padValue, doMasking);
 
   // Create ShapeCastOp.
   SmallVector<int64_t> destShape(inputVectorSizes);
@@ -1763,7 +1784,7 @@ vectorizeDynamicLinalgOpPrecondition(linalg::LinalgOp op,
 /// Returns success if `inputVectorSizes` is a valid masking configuraion for
 /// given `shape`, i.e., it meets:
 ///   1. The numbers of elements in both array are equal.
-///   2. `inputVectorSizes` does nos have dynamic dimensions.
+///   2. `inputVectorSizes` does not have dynamic dimensions.
 ///   3. All the values in `inputVectorSizes` are greater than or equal to
 ///      static sizes in `shape`.
 static LogicalResult
@@ -1881,18 +1902,25 @@ static LogicalResult vectorizeLinalgOpPrecondition(
   return success();
 }
 
-/// TODO: Use a matcher to check for a constant padding value.
 static LogicalResult
 vectorizePackOpPrecondition(tensor::PackOp packOp,
                             ArrayRef<int64_t> inputVectorSizes) {
   auto padValue = packOp.getPaddingValue();
-  if (padValue && !padValue.getDefiningOp<arith::ConstantOp>()) {
+  Attribute cstAttr;
+  if (padValue && !matchPattern(padValue, m_Constant(&cstAttr))) {
     LDBG("pad value is not constant: " << packOp << "\n");
     return failure();
   }
-
   ArrayRef<int64_t> resultTensorShape = packOp.getDestType().getShape();
-  if (failed(isValidMaskedInputVector(
+  bool satisfyEmptyCond = true;
+  if (inputVectorSizes.empty()) {
+    if (!packOp.getDestType().hasStaticShape() ||
+        !packOp.getSourceType().hasStaticShape())
+      satisfyEmptyCond = false;
+  }
+
+  if (!satisfyEmptyCond &&
+      failed(isValidMaskedInputVector(
           resultTensorShape.take_front(packOp.getSourceRank()),
           inputVectorSizes)))
     return failure();
diff --git a/mlir/test/Dialect/Linalg/vectorization-unsupported.mlir b/mlir/test/Dialect/Linalg/vectorization-unsupported.mlir
index 9127eac5da951..5d3c07c8e23c1 100644
--- a/mlir/test/Dialect/Linalg/vectorization-unsupported.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization-unsupported.mlir
@@ -109,3 +109,20 @@ module attributes {transform.with_named_sequence} {
     transform.yield
   }
 }
+
+  // -----
+
+func.func @test_pack_no_vectorize_dynamic_shape(%arg0: tensor<?xf32>, %arg1: tensor<4x16xf32>) -> tensor<4x16xf32> {
+  %pad = arith.constant 0.000000e+00 : f32
+  // expected-error @+1 {{Attempted to vectorize, but failed}}
+  %pack = tensor.pack %arg0 padding_value(%pad : f32) inner_dims_pos = [0] inner_tiles = [16] into %arg1 : tensor<?xf32> -> tensor<4x16xf32>
+  return %pack : tensor<4x16xf32>
+}
+
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
+    %0 = transform.structured.match ops{["tensor.pack"]} in %arg0 : (!transform.any_op) -> !transform.any_op
+    transform.structured.vectorize %0 : !transform.any_op
+    transform.yield
+  }
+}
diff --git a/mlir/test/Dialect/Linalg/vectorization.mlir b/mlir/test/Dialect/Linalg/vectorization.mlir
index fd7d3b4767eb2..80a5a4c6702ac 100644
--- a/mlir/test/Dialect/Linalg/vectorization.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization.mlir
@@ -930,3 +930,58 @@ func.func @test_vectorize_unpack_no_masks(%source: tensor<8x8x32x16xf32>, %dest:
     transform.yield
   } 
 }
+
+  // -----
+
+// CHECK-LABEL: test_vectorize_pack_no_vector_sizes
+func.func @test_vectorize_pack_no_vector_sizes(%arg0: tensor<64x4xf32>, %arg1: tensor<2x4x16x2xf32>) -> tensor<2x4x16x2xf32> {
+  %pack = tensor.pack %arg0 outer_dims_perm = [1, 0] inner_dims_pos = [0, 1] inner_tiles = [16, 2] into %arg1 : tensor<64x4xf32> -> tensor<2x4x16x2xf32>
+  return %pack : tensor<2x4x16x2xf32>
+}
+//  CHECK-DAG: %[[cst:.*]] = arith.constant 0.000000e+00 : f32
+//  CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
+//      CHECK: %[[read:.*]] = vector.transfer_read %{{.*}}[%[[c0]], %[[c0]]], %[[cst]]
+// CHECK-SAME:    {in_bounds = [true, true]} : tensor<64x4xf32>, vector<64x4xf32>
+//      CHECK: %[[shape_cast:.*]] = vector.shape_cast %[[read]] : vector<64x4xf32> to vector<4x16x2x2xf32>
+//      CHECK: %[[transpose:.*]] = vector.transpose %[[shape_cast]], [2, 0, 1, 3] : vector<4x16x2x2xf32> to vector<2x4x16x2xf32>
+//  CHECK-DAG: %[[c0_1:.*]] = arith.constant 0 : index
+//  CHECK-DAG: %[[empty:.*]] = tensor.empty() : tensor<2x4x16x2xf32>
+//      CHECK: %[[write:.*]] = vector.transfer_write %[[transpose]], %[[empty]][%[[c0_1]], %[[c0_1]], %[[c0_1]], %[[c0_1]]]
+// CHECK-SAME:   {in_bounds = [true, true, true, true]} : vector<2x4x16x2xf32>, tensor<2x4x16x2xf32>
+//      CHECK: return %[[write]] : tensor<2x4x16x2xf32>
+
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
+    %0 = transform.structured.match ops{["tensor.pack"]} in %arg0 : (!transform.any_op) -> !transform.any_op
+    transform.structured.vectorize %0 : !transform.any_op
+    transform.yield
+  }
+}
+
+  // -----
+
+// CHECK-LABEL: test_vectorize_padded_pack_no_vector_sizes
+func.func @test_vectorize_padded_pack_no_vector_sizes(%arg0: tensor<32x7x15xf32>, %arg1: tensor<32x4x1x16x2xf32>) -> tensor<32x4x1x16x2xf32> {
+  %pad = arith.constant 0.000000e+00 : f32
+  %pack = tensor.pack %arg0 padding_value(%pad : f32) inner_dims_pos = [2, 1] inner_tiles = [16, 2] into %arg1 : tensor<32x7x15xf32> -> tensor<32x4x1x16x2xf32>
+  return %pack : tensor<32x4x1x16x2xf32>
+}
+//  CHECK-DAG: %[[cst:.*]] = arith.constant 0.000000e+00 : f32
+//  CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
+//      CHECK: %[[transfer_read:.*]] =  vector.transfer_read %{{.*}}[%[[c0]], %[[c0]], %[[c0]]], %[[cst]]
+// CHECK-SAME:   {in_bounds = [true, false, false]} : tensor<32x7x15xf32>, vector<32x8x16xf32> 
+//      CHECK: %[[shape_cast:.*]] = vector.shape_cast %[[transfer_read]] : vector<32x8x16xf32> to vector<32x4x2x1x16xf32>
+//      CHECK: %[[transpose:.*]] = vector.transpose %[[shape_cast]], [0, 1, 3, 4, 2] : vector<32x4x2x1x16xf32> to vector<32x4x1x16x2xf32>
+//  CHECK-DAG: %[[c0_1:.*]] = arith.constant 0 : index
+//  CHECK-DAG: %[[empty:.*]] = tensor.empty() : tensor<32x4x1x16x2xf32>
+//      CHECK: %[[write:.*]] = vector.transfer_write %[[transpose]], %[[empty]][%[[c0_1]], %[[c0_1]], %[[c0_1]], %[[c0_1]], %[[c0_1]]]
+// CHECK-SAME:   {in_bounds = [true, true, true, true, true]} : vector<32x4x1x16x2xf32>, tensor<32x4x1x16x2xf32>
+//      CHECK: return %[[write]] : tensor<32x4x1x16x2xf32>
+
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
+    %0 = transform.structured.match ops{["tensor.pack"]} in %arg0 : (!transform.any_op) -> !transform.any_op
+    transform.structured.vectorize %0 : !transform.any_op
+    transform.yield
+  }
+}

From 8c9f45e2decbb68dbf83794f98291b53f59390f8 Mon Sep 17 00:00:00 2001
From: Eli Friedman <efriedma@quicinc.com>
Date: Tue, 16 Apr 2024 17:08:02 -0700
Subject: [PATCH 676/886] [ARM64EC] Fix arm_neon.h on ARM64EC. (#88572)

Since 97fe519d, in ARM64EC mode, we don't define `__aarch64__`. Fix
various preprocessor guards to account for this.
---
 clang/include/clang/Basic/arm_fp16.td |  2 +-
 clang/include/clang/Basic/arm_neon.td | 58 +++++++++++++--------------
 clang/utils/TableGen/NeonEmitter.cpp  |  8 ++--
 3 files changed, 34 insertions(+), 34 deletions(-)

diff --git a/clang/include/clang/Basic/arm_fp16.td b/clang/include/clang/Basic/arm_fp16.td
index cb2a09303e8e1..d36b4617bef5d 100644
--- a/clang/include/clang/Basic/arm_fp16.td
+++ b/clang/include/clang/Basic/arm_fp16.td
@@ -14,7 +14,7 @@
 include "arm_neon_incl.td"
 
 // ARMv8.2-A FP16 intrinsics.
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "fullfp16" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "fullfp16" in {
 
   // Negate
   def VNEGSH          : SInst<"vneg", "11", "Sh">;
diff --git a/clang/include/clang/Basic/arm_neon.td b/clang/include/clang/Basic/arm_neon.td
index 7edac5afafaa9..6d655c39360d3 100644
--- a/clang/include/clang/Basic/arm_neon.td
+++ b/clang/include/clang/Basic/arm_neon.td
@@ -605,11 +605,11 @@ def VQDMULL_LANE  : SOpInst<"vqdmull_lane", "(>Q)..I", "si", OP_QDMULL_LN>;
 def VQDMULH_N     : SOpInst<"vqdmulh_n", "..1", "siQsQi", OP_QDMULH_N>;
 def VQRDMULH_N    : SOpInst<"vqrdmulh_n", "..1", "siQsQi", OP_QRDMULH_N>;
 
-let ArchGuard = "!defined(__aarch64__)" in {
+let ArchGuard = "!defined(__aarch64__) && !defined(__arm64ec__)" in {
 def VQDMULH_LANE  : SOpInst<"vqdmulh_lane", "..qI", "siQsQi", OP_QDMULH_LN>;
 def VQRDMULH_LANE : SOpInst<"vqrdmulh_lane", "..qI", "siQsQi", OP_QRDMULH_LN>;
 }
-let ArchGuard = "defined(__aarch64__)" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)" in {
 def A64_VQDMULH_LANE  : SInst<"vqdmulh_lane", "..(!q)I", "siQsQi">;
 def A64_VQRDMULH_LANE : SInst<"vqrdmulh_lane", "..(!q)I", "siQsQi">;
 }
@@ -686,7 +686,7 @@ multiclass REINTERPRET_CROSS_TYPES<string TypesA, string TypesB> {
 
 // E.3.31 Vector reinterpret cast operations
 def VREINTERPRET : REINTERPRET_CROSS_SELF<"csilUcUsUiUlhfPcPsQcQsQiQlQUcQUsQUiQUlQhQfQPcQPs"> {
-  let ArchGuard = "!defined(__aarch64__)";
+  let ArchGuard = "!defined(__aarch64__) && !defined(__arm64ec__)";
   let BigEndianSafe = 1;
 }
 
@@ -714,7 +714,7 @@ def VADDP   : WInst<"vadd", "...", "PcPsPlQPcQPsQPl">;
 ////////////////////////////////////////////////////////////////////////////////
 // AArch64 Intrinsics
 
-let ArchGuard = "defined(__aarch64__)" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)" in {
 
 ////////////////////////////////////////////////////////////////////////////////
 // Load/Store
@@ -1091,14 +1091,14 @@ let isLaneQ = 1 in {
 def VQDMULH_LANEQ  : SInst<"vqdmulh_laneq", "..QI", "siQsQi">;
 def VQRDMULH_LANEQ : SInst<"vqrdmulh_laneq", "..QI", "siQsQi">;
 }
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.1a" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.1a" in {
 def VQRDMLAH_LANEQ : SOpInst<"vqrdmlah_laneq", "...QI", "siQsQi", OP_QRDMLAH_LN> {
   let isLaneQ = 1;
 }
 def VQRDMLSH_LANEQ : SOpInst<"vqrdmlsh_laneq", "...QI", "siQsQi", OP_QRDMLSH_LN> {
   let isLaneQ = 1;
 }
-} // ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.1a"
+} // ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.1a"
 
 // Note: d type implemented by SCALAR_VMULX_LANE
 def VMULX_LANE : IOpInst<"vmulx_lane", "..qI", "fQfQd", OP_MULX_LN>;
@@ -1143,7 +1143,7 @@ def SHA256H2 : SInst<"vsha256h2", "....", "QUi">;
 def SHA256SU1 : SInst<"vsha256su1", "....", "QUi">;
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "sha3" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "sha3" in {
 def BCAX : SInst<"vbcax", "....", "QUcQUsQUiQUlQcQsQiQl">;
 def EOR3 : SInst<"veor3", "....", "QUcQUsQUiQUlQcQsQiQl">;
 def RAX1 : SInst<"vrax1", "...", "QUl">;
@@ -1153,14 +1153,14 @@ def XAR :  SInst<"vxar", "...I", "QUl">;
 }
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "sha3" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "sha3" in {
 def SHA512SU0 : SInst<"vsha512su0", "...", "QUl">;
 def SHA512su1 : SInst<"vsha512su1", "....", "QUl">;
 def SHA512H : SInst<"vsha512h", "....", "QUl">;
 def SHA512H2 : SInst<"vsha512h2", "....", "QUl">;
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "sm4" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "sm4" in {
 def SM3SS1 : SInst<"vsm3ss1", "....", "QUi">;
 def SM3TT1A : SInst<"vsm3tt1a", "....I", "QUi">;
 def SM3TT1B : SInst<"vsm3tt1b", "....I", "QUi">;
@@ -1170,7 +1170,7 @@ def SM3PARTW1 : SInst<"vsm3partw1", "....", "QUi">;
 def SM3PARTW2 : SInst<"vsm3partw2", "....", "QUi">;
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "sm4" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "sm4" in {
 def SM4E : SInst<"vsm4e", "...", "QUi">;
 def SM4EKEY : SInst<"vsm4ekey", "...", "QUi">;
 }
@@ -1193,7 +1193,7 @@ def FCVTAS_S32 : SInst<"vcvta_s32", "S.", "fQf">;
 def FCVTAU_S32 : SInst<"vcvta_u32", "U.", "fQf">;
 }
 
-let ArchGuard = "defined(__aarch64__)" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)" in {
 def FCVTNS_S64 : SInst<"vcvtn_s64", "S.", "dQd">;
 def FCVTNU_S64 : SInst<"vcvtn_u64", "U.", "dQd">;
 def FCVTPS_S64 : SInst<"vcvtp_s64", "S.", "dQd">;
@@ -1217,7 +1217,7 @@ def FRINTZ_S32 : SInst<"vrnd", "..", "fQf">;
 def FRINTI_S32 : SInst<"vrndi", "..", "fQf">;
 }
 
-let ArchGuard = "defined(__aarch64__) && defined(__ARM_FEATURE_DIRECTED_ROUNDING)" in {
+let ArchGuard = "(defined(__aarch64__) || defined(__arm64ec__)) && defined(__ARM_FEATURE_DIRECTED_ROUNDING)" in {
 def FRINTN_S64 : SInst<"vrndn", "..", "dQd">;
 def FRINTA_S64 : SInst<"vrnda", "..", "dQd">;
 def FRINTP_S64 : SInst<"vrndp", "..", "dQd">;
@@ -1227,7 +1227,7 @@ def FRINTZ_S64 : SInst<"vrnd", "..", "dQd">;
 def FRINTI_S64 : SInst<"vrndi", "..", "dQd">;
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.5a" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.5a" in {
 def FRINT32X_S32 : SInst<"vrnd32x", "..", "fQf">;
 def FRINT32Z_S32 : SInst<"vrnd32z", "..", "fQf">;
 def FRINT64X_S32 : SInst<"vrnd64x", "..", "fQf">;
@@ -1247,7 +1247,7 @@ def FMAXNM_S32 : SInst<"vmaxnm", "...", "fQf">;
 def FMINNM_S32 : SInst<"vminnm", "...", "fQf">;
 }
 
-let ArchGuard = "defined(__aarch64__) && defined(__ARM_FEATURE_NUMERIC_MAXMIN)" in {
+let ArchGuard = "(defined(__aarch64__)  || defined(__arm64ec__)) && defined(__ARM_FEATURE_NUMERIC_MAXMIN)" in {
 def FMAXNM_S64 : SInst<"vmaxnm", "...", "dQd">;
 def FMINNM_S64 : SInst<"vminnm", "...", "dQd">;
 }
@@ -1289,7 +1289,7 @@ def VQTBX4_A64 : WInst<"vqtbx4", "..(4Q)U", "UccPcQUcQcQPc">;
 // itself during generation so, unlike all other intrinsics, this one should
 // include *all* types, not just additional ones.
 def VVREINTERPRET : REINTERPRET_CROSS_SELF<"csilUcUsUiUlhfdPcPsPlQcQsQiQlQUcQUsQUiQUlQhQfQdQPcQPsQPlQPk"> {
-  let ArchGuard = "defined(__aarch64__)";
+  let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)";
   let BigEndianSafe = 1;
 }
 
@@ -1401,7 +1401,7 @@ def SCALAR_SQDMULH : SInst<"vqdmulh", "111", "SsSi">;
 // Scalar Integer Saturating Rounding Doubling Multiply Half High
 def SCALAR_SQRDMULH : SInst<"vqrdmulh", "111", "SsSi">;
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.1a" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.1a" in {
 ////////////////////////////////////////////////////////////////////////////////
 // Signed Saturating Rounding Doubling Multiply Accumulate Returning High Half
 def SCALAR_SQRDMLAH : SInst<"vqrdmlah", "1111", "SsSi">;
@@ -1409,7 +1409,7 @@ def SCALAR_SQRDMLAH : SInst<"vqrdmlah", "1111", "SsSi">;
 ////////////////////////////////////////////////////////////////////////////////
 // Signed Saturating Rounding Doubling Multiply Subtract Returning High Half
 def SCALAR_SQRDMLSH : SInst<"vqrdmlsh", "1111", "SsSi">;
-} // ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.1a"
+} // ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.1a"
 
 ////////////////////////////////////////////////////////////////////////////////
 // Scalar Floating-point Multiply Extended
@@ -1651,7 +1651,7 @@ def SCALAR_VDUP_LANEQ : IInst<"vdup_laneq", "1QI", "ScSsSiSlSfSdSUcSUsSUiSUlSPcS
   let isLaneQ = 1;
 }
 
-} // ArchGuard = "defined(__aarch64__)"
+} // ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)"
 
 // ARMv8.2-A FP16 vector intrinsics for A32/A64.
 let TargetGuard = "fullfp16" in {
@@ -1775,7 +1775,7 @@ def VEXTH      : WInst<"vext", "...I", "hQh">;
 def VREV64H    : WOpInst<"vrev64", "..", "hQh", OP_REV64>;
 
 // ARMv8.2-A FP16 vector intrinsics for A64 only.
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "fullfp16" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "fullfp16" in {
 
   // Vector rounding
   def FRINTIH      : SInst<"vrndi", "..", "hQh">;
@@ -1856,7 +1856,7 @@ let ArchGuard = "defined(__aarch64__)", TargetGuard = "fullfp16" in {
   def FMINNMVH : SInst<"vminnmv", "1.", "hQh">;
 }
 
-let ArchGuard = "defined(__aarch64__)" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)" in {
   // Permutation
   def VTRN1H     : SOpInst<"vtrn1", "...", "hQh", OP_TRN1>;
   def VZIP1H     : SOpInst<"vzip1", "...", "hQh", OP_ZIP1>;
@@ -1876,7 +1876,7 @@ let TargetGuard = "dotprod" in {
   def DOT : SInst<"vdot", "..(<<)(<<)", "iQiUiQUi">;
   def DOT_LANE : SOpInst<"vdot_lane", "..(<<)(<<q)I", "iUiQiQUi", OP_DOT_LN>;
 }
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "dotprod" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "dotprod" in {
   // Variants indexing into a 128-bit vector are A64 only.
   def UDOT_LANEQ : SOpInst<"vdot_laneq", "..(<<)(<<Q)I", "iUiQiQUi", OP_DOT_LNQ> {
     let isLaneQ = 1;
@@ -1884,7 +1884,7 @@ let ArchGuard = "defined(__aarch64__)", TargetGuard = "dotprod" in {
 }
 
 // v8.2-A FP16 fused multiply-add long instructions.
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "fp16fml" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "fp16fml" in {
   def VFMLAL_LOW  : SInst<"vfmlal_low",  ">>..", "hQh">;
   def VFMLSL_LOW  : SInst<"vfmlsl_low",  ">>..", "hQh">;
   def VFMLAL_HIGH : SInst<"vfmlal_high", ">>..", "hQh">;
@@ -1918,7 +1918,7 @@ let TargetGuard = "i8mm" in {
   def VUSDOT_LANE  : SOpInst<"vusdot_lane", "..(<<U)(<<q)I", "iQi", OP_USDOT_LN>;
   def VSUDOT_LANE  : SOpInst<"vsudot_lane", "..(<<)(<<qU)I", "iQi", OP_SUDOT_LN>;
 
-  let ArchGuard = "defined(__aarch64__)" in {
+  let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)" in {
     let isLaneQ = 1 in {
       def VUSDOT_LANEQ  : SOpInst<"vusdot_laneq", "..(<<U)(<<Q)I", "iQi", OP_USDOT_LNQ>;
       def VSUDOT_LANEQ  : SOpInst<"vsudot_laneq", "..(<<)(<<QU)I", "iQi", OP_SUDOT_LNQ>;
@@ -1986,7 +1986,7 @@ let TargetGuard = "v8.3a" in {
 
   defm VCMLA_F32        : VCMLA_ROTS<"f", "uint64x1_t", "uint64x2_t">;
 }
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "v8.3a" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "v8.3a" in {
   def VCADDQ_ROT90_FP64  : SInst<"vcaddq_rot90", "QQQ", "d">;
   def VCADDQ_ROT270_FP64 : SInst<"vcaddq_rot270", "QQQ", "d">;
 
@@ -2058,14 +2058,14 @@ let TargetGuard = "bf16" in {
   def SCALAR_CVT_F32_BF16 : SOpInst<"vcvtah_f32", "(1F>)(1!)", "b", OP_CVT_F32_BF16>;
 }
 
-let ArchGuard = "!defined(__aarch64__)", TargetGuard = "bf16" in {
+let ArchGuard = "!defined(__aarch64__) && !defined(__arm64ec__)", TargetGuard = "bf16" in {
   def VCVT_BF16_F32_A32_INTERNAL : WInst<"__a32_vcvt_bf16", "BQ", "f">;
   def VCVT_BF16_F32_A32 : SOpInst<"vcvt_bf16", "BQ", "f", OP_VCVT_BF16_F32_A32>;
   def VCVT_LOW_BF16_F32_A32 : SOpInst<"vcvt_low_bf16",  "BQ", "Qf", OP_VCVT_BF16_F32_LO_A32>;
   def VCVT_HIGH_BF16_F32_A32 : SOpInst<"vcvt_high_bf16", "BBQ", "Qf", OP_VCVT_BF16_F32_HI_A32>;
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "bf16" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "bf16" in {
   def VCVT_LOW_BF16_F32_A64_INTERNAL : WInst<"__a64_vcvtq_low_bf16", "BQ", "Hf">;
   def VCVT_LOW_BF16_F32_A64 : SOpInst<"vcvt_low_bf16", "BQ", "Qf", OP_VCVT_BF16_F32_LO_A64>;
   def VCVT_HIGH_BF16_F32_A64 : SInst<"vcvt_high_bf16", "BBQ", "Qf">;
@@ -2077,14 +2077,14 @@ let ArchGuard = "defined(__aarch64__)", TargetGuard = "bf16" in {
   def COPYQ_LANEQ_BF16 : IOpInst<"vcopy_laneq", "..I.I", "Qb", OP_COPY_LN>;
 }
 
-let ArchGuard = "!defined(__aarch64__)", TargetGuard = "bf16" in {
+let ArchGuard = "!defined(__aarch64__) && !defined(__arm64ec__)", TargetGuard = "bf16" in {
   let BigEndianSafe = 1 in {
     defm VREINTERPRET_BF : REINTERPRET_CROSS_TYPES<
         "csilUcUsUiUlhfPcPsPlQcQsQiQlQUcQUsQUiQUlQhQfQPcQPsQPl", "bQb">;
   }
 }
 
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "bf16" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "bf16" in {
   let BigEndianSafe = 1 in {
     defm VVREINTERPRET_BF : REINTERPRET_CROSS_TYPES<
         "csilUcUsUiUlhfdPcPsPlQcQsQiQlQUcQUsQUiQUlQhQfQdQPcQPsQPlQPk", "bQb">;
@@ -2092,7 +2092,7 @@ let ArchGuard = "defined(__aarch64__)", TargetGuard = "bf16" in {
 }
 
 // v8.9a/v9.4a LRCPC3 intrinsics
-let ArchGuard = "defined(__aarch64__)", TargetGuard = "rcpc3" in {
+let ArchGuard = "defined(__aarch64__) || defined(__arm64ec__)", TargetGuard = "rcpc3" in {
   def VLDAP1_LANE : WInst<"vldap1_lane", ".(c*!).I", "QUlQlUlldQdPlQPl">;
   def VSTL1_LANE  : WInst<"vstl1_lane", "v*(.!)I", "QUlQlUlldQdPlQPl">;
 }
diff --git a/clang/utils/TableGen/NeonEmitter.cpp b/clang/utils/TableGen/NeonEmitter.cpp
index 04e1acc270500..56f1fdf9ef574 100644
--- a/clang/utils/TableGen/NeonEmitter.cpp
+++ b/clang/utils/TableGen/NeonEmitter.cpp
@@ -2266,7 +2266,7 @@ static void emitNeonTypeDefs(const std::string& types, raw_ostream &OS) {
       InIfdef = false;
     }
     if (!InIfdef && IsA64) {
-      OS << "#ifdef __aarch64__\n";
+      OS << "#if defined(__aarch64__) || defined(__arm64ec__)\n";
       InIfdef = true;
     }
 
@@ -2299,7 +2299,7 @@ static void emitNeonTypeDefs(const std::string& types, raw_ostream &OS) {
         InIfdef = false;
       }
       if (!InIfdef && IsA64) {
-        OS << "#ifdef __aarch64__\n";
+        OS << "#if defined(__aarch64__) || defined(__arm64ec__)\n";
         InIfdef = true;
       }
 
@@ -2381,7 +2381,7 @@ void NeonEmitter::run(raw_ostream &OS) {
   OS << "#include <arm_vector_types.h>\n";
 
   // For now, signedness of polynomial types depends on target
-  OS << "#ifdef __aarch64__\n";
+  OS << "#if defined(__aarch64__) || defined(__arm64ec__)\n";
   OS << "typedef uint8_t poly8_t;\n";
   OS << "typedef uint16_t poly16_t;\n";
   OS << "typedef uint64_t poly64_t;\n";
@@ -2582,7 +2582,7 @@ void NeonEmitter::runVectorTypes(raw_ostream &OS) {
   OS << "typedef float float32_t;\n";
   OS << "typedef __fp16 float16_t;\n";
 
-  OS << "#ifdef __aarch64__\n";
+  OS << "#if defined(__aarch64__) || defined(__arm64ec__)\n";
   OS << "typedef double float64_t;\n";
   OS << "#endif\n\n";
 

From 8c9d814b66f7df274de41225575817188fbeed4f Mon Sep 17 00:00:00 2001
From: Kai Sasaki <lewuathe@gmail.com>
Date: Wed, 17 Apr 2024 09:19:15 +0900
Subject: [PATCH 677/886] [mlir][complex] Fastmath flag for complex angle
 (#88658)

See
https://discourse.llvm.org/t/rfc-fastmath-flags-support-in-complex-dialect/71981
---
 .../ComplexToStandard/ComplexToStandard.cpp         |  3 ++-
 .../ComplexToStandard/convert-to-standard.mlir      | 13 +++++++++++++
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
index 03e578136e590..4a15976d40c76 100644
--- a/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
+++ b/mlir/lib/Conversion/ComplexToStandard/ComplexToStandard.cpp
@@ -1289,13 +1289,14 @@ struct AngleOpConversion : public OpConversionPattern<complex::AngleOp> {
                   ConversionPatternRewriter &rewriter) const override {
     auto loc = op.getLoc();
     auto type = op.getType();
+    arith::FastMathFlagsAttr fmf = op.getFastMathFlagsAttr();
 
     Value real =
         rewriter.create<complex::ReOp>(loc, type, adaptor.getComplex());
     Value imag =
         rewriter.create<complex::ImOp>(loc, type, adaptor.getComplex());
 
-    rewriter.replaceOpWithNewOp<math::Atan2Op>(op, imag, real);
+    rewriter.replaceOpWithNewOp<math::Atan2Op>(op, imag, real, fmf);
 
     return success();
   }
diff --git a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
index fa1d564d6ad35..827ae940165c7 100644
--- a/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
+++ b/mlir/test/Conversion/ComplexToStandard/convert-to-standard.mlir
@@ -2187,3 +2187,16 @@ func.func @complex_tanh_nnan_ninf(%arg: complex<f32>) -> complex<f32> {
 
 // CHECK-COUNT-1: arith.select
 // CHECK-NOT: arith.select
+
+// -----
+
+// CHECK-LABEL:   func.func @complex_angle_with_fmf
+// CHECK-SAME: %[[ARG:.*]]: complex<f32>
+func.func @complex_angle_with_fmf(%arg: complex<f32>) -> f32 {
+  %angle = complex.angle %arg fastmath<nnan,contract> : complex<f32>
+  return %angle : f32
+}
+// CHECK: %[[REAL:.*]] = complex.re %[[ARG]] : complex<f32>
+// CHECK: %[[IMAG:.*]] = complex.im %[[ARG]] : complex<f32>
+// CHECK: %[[RESULT:.*]] = math.atan2 %[[IMAG]], %[[REAL]] fastmath<nnan,contract> : f32
+// CHECK: return %[[RESULT]] : f32
\ No newline at end of file

From efd60556f759fbfa0fc0a5984463daeaef20799c Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Wed, 17 Apr 2024 09:22:57 +0900
Subject: [PATCH 678/886] Revert "[SLP]Attempt to vectorize long stores, if
 short one failed."

This reverts commit 7d4e8c1f3bbfe976f4871c9cf953f76d771b0eda.

Contrary to the commit description, this does cause large
compile-time regressions (up to 10% on individual files).
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 81 +++++++------------
 .../Transforms/SLPVectorizer/X86/pr46983.ll   | 46 ++++++++---
 2 files changed, 65 insertions(+), 62 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 8ae38550d3095..7694627c3b043 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15237,60 +15237,39 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
         Size *= 2;
       });
       unsigned StartIdx = 0;
-      unsigned Repeat = 0;
-      constexpr unsigned MaxAttempts = 2;
-      while (true) {
-        ++Repeat;
-        for (unsigned Size : CandidateVFs) {
-          for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
-            ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
-            assert(
-                all_of(
-                    Slice,
-                    [&](Value *V) {
-                      return cast<StoreInst>(V)->getValueOperand()->getType() ==
-                             cast<StoreInst>(Slice.front())
-                                 ->getValueOperand()
-                                 ->getType();
-                    }) &&
-                "Expected all operands of same type.");
-            if (!VectorizedStores.count(Slice.front()) &&
-                !VectorizedStores.count(Slice.back()) &&
-                TriedSequences
-                    .insert(std::make_pair(Slice.front(), Slice.back()))
-                    .second &&
-                vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
-              // Mark the vectorized stores so that we don't vectorize them
-              // again.
-              VectorizedStores.insert(Slice.begin(), Slice.end());
-              Changed = true;
-              // If we vectorized initial block, no need to try to vectorize
-              // it again.
-              if (Cnt == StartIdx)
-                StartIdx += Size;
-              Cnt += Size;
-              continue;
-            }
-            ++Cnt;
-          }
-          // Check if the whole array was vectorized already - exit.
-          if (StartIdx >= Operands.size()) {
-            Repeat = MaxAttempts;
-            break;
+      for (unsigned Size : CandidateVFs) {
+        for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
+          ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
+          assert(
+              all_of(
+                  Slice,
+                  [&](Value *V) {
+                    return cast<StoreInst>(V)->getValueOperand()->getType() ==
+                           cast<StoreInst>(Slice.front())
+                               ->getValueOperand()
+                               ->getType();
+                  }) &&
+              "Expected all operands of same type.");
+          if (!VectorizedStores.count(Slice.front()) &&
+              !VectorizedStores.count(Slice.back()) &&
+              TriedSequences.insert(std::make_pair(Slice.front(), Slice.back()))
+                  .second &&
+              vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
+            // Mark the vectorized stores so that we don't vectorize them again.
+            VectorizedStores.insert(Slice.begin(), Slice.end());
+            Changed = true;
+            // If we vectorized initial block, no need to try to vectorize it
+            // again.
+            if (Cnt == StartIdx)
+              StartIdx += Size;
+            Cnt += Size;
+            continue;
           }
+          ++Cnt;
         }
-        // Check if tried all attempts or no need for the last attempts at all.
-        if (Repeat >= MaxAttempts)
-          break;
-        const unsigned MaxTotalNum = bit_floor(Operands.size() - StartIdx);
-        if (MaxVF >= MaxTotalNum)
+        // Check if the whole array was vectorized already - exit.
+        if (StartIdx >= Operands.size())
           break;
-        // Last attempt to vectorize max number of elements, if all previous
-        // attempts were unsuccessful because of the cost issues.
-        CandidateVFs.clear();
-        for (unsigned Size = MaxTotalNum; Size > MaxVF; Size /= 2) {
-          CandidateVFs.push_back(Size);
-        }
       }
     }
   };
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
index 3deab0975ce76..75505f632a43f 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
@@ -100,17 +100,41 @@ define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
 define void @store_i64(ptr nocapture %0, i32 %1, i32 %2) {
 ; SSE-LABEL: @store_i64(
 ; SSE-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP1:%.*]] to i64
-; SSE-NEXT:    [[TMP5:%.*]] = load <4 x i64>, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
-; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x i64> poison, i64 [[TMP4]], i64 0
-; SSE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP8:%.*]] = mul <4 x i64> [[TMP5]], [[TMP7]]
-; SSE-NEXT:    [[TMP9:%.*]] = lshr <4 x i64> [[TMP8]], <i64 15, i64 15, i64 15, i64 15>
-; SSE-NEXT:    [[TMP10:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
-; SSE-NEXT:    [[TMP11:%.*]] = icmp ult <4 x i32> [[TMP10]], <i32 255, i32 255, i32 255, i32 255>
-; SSE-NEXT:    [[TMP12:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
-; SSE-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>
-; SSE-NEXT:    [[TMP14:%.*]] = zext <4 x i32> [[TMP13]] to <4 x i64>
-; SSE-NEXT:    store <4 x i64> [[TMP14]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP5:%.*]] = load i64, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
+; SSE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], [[TMP4]]
+; SSE-NEXT:    [[TMP7:%.*]] = lshr i64 [[TMP6]], 15
+; SSE-NEXT:    [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
+; SSE-NEXT:    [[TMP9:%.*]] = icmp ult i32 [[TMP8]], 255
+; SSE-NEXT:    [[TMP10:%.*]] = and i64 [[TMP7]], 4294967295
+; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i64 [[TMP10]], i64 255
+; SSE-NEXT:    store i64 [[TMP11]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 8
+; SSE-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], [[TMP4]]
+; SSE-NEXT:    [[TMP15:%.*]] = lshr i64 [[TMP14]], 15
+; SSE-NEXT:    [[TMP16:%.*]] = trunc i64 [[TMP15]] to i32
+; SSE-NEXT:    [[TMP17:%.*]] = icmp ult i32 [[TMP16]], 255
+; SSE-NEXT:    [[TMP18:%.*]] = and i64 [[TMP15]], 4294967295
+; SSE-NEXT:    [[TMP19:%.*]] = select i1 [[TMP17]], i64 [[TMP18]], i64 255
+; SSE-NEXT:    store i64 [[TMP19]], ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 16
+; SSE-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], [[TMP4]]
+; SSE-NEXT:    [[TMP23:%.*]] = lshr i64 [[TMP22]], 15
+; SSE-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
+; SSE-NEXT:    [[TMP25:%.*]] = icmp ult i32 [[TMP24]], 255
+; SSE-NEXT:    [[TMP26:%.*]] = and i64 [[TMP23]], 4294967295
+; SSE-NEXT:    [[TMP27:%.*]] = select i1 [[TMP25]], i64 [[TMP26]], i64 255
+; SSE-NEXT:    store i64 [[TMP27]], ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 24
+; SSE-NEXT:    [[TMP29:%.*]] = load i64, ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], [[TMP4]]
+; SSE-NEXT:    [[TMP31:%.*]] = lshr i64 [[TMP30]], 15
+; SSE-NEXT:    [[TMP32:%.*]] = trunc i64 [[TMP31]] to i32
+; SSE-NEXT:    [[TMP33:%.*]] = icmp ult i32 [[TMP32]], 255
+; SSE-NEXT:    [[TMP34:%.*]] = and i64 [[TMP31]], 4294967295
+; SSE-NEXT:    [[TMP35:%.*]] = select i1 [[TMP33]], i64 [[TMP34]], i64 255
+; SSE-NEXT:    store i64 [[TMP35]], ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @store_i64(

From 7c2688977567ea5ac1203daa3c452b541ef55f67 Mon Sep 17 00:00:00 2001
From: Michael Liao <michael.hliao@gmail.com>
Date: Tue, 16 Apr 2024 15:20:14 -0400
Subject: [PATCH 679/886] [Sparc] Fix instr desc of special register stores

- Those special register stores are STORE and their memory operands are
  input operands instead of output ones.

Reviewers:
JDevlieghere, arsenm, yinying-lisa-li, koachan, PeimingLiu, jyknight, aartbik, matthias-springer

Reviewed By: arsenm

Pull Request: https://github.com/llvm/llvm-project/pull/88971
---
 llvm/lib/Target/Sparc/SparcInstrInfo.td | 30 ++++++++++++-------------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Target/Sparc/SparcInstrInfo.td b/llvm/lib/Target/Sparc/SparcInstrInfo.td
index 5e792427cca28..4d68f93efeac1 100644
--- a/llvm/lib/Target/Sparc/SparcInstrInfo.td
+++ b/llvm/lib/Target/Sparc/SparcInstrInfo.td
@@ -693,38 +693,38 @@ let DecoderNamespace = "SparcV8", Predicates = [HasNoV9] in {
 }
 
 let rd = 0 in {
-  let Defs = [CPSR] in {
-    def STCSRrr : F3_1<3, 0b110101, (outs (MEMrr $rs1, $rs2):$addr), (ins),
+  let mayStore = 1, Uses = [CPSR] in {
+    def STCSRrr : F3_1<3, 0b110101, (outs), (ins (MEMrr $rs1, $rs2):$addr),
                        "st %csr, [$addr]", [], IIC_st>;
-    def STCSRri : F3_2<3, 0b110101, (outs (MEMri $rs1, $simm13):$addr), (ins),
+    def STCSRri : F3_2<3, 0b110101, (outs), (ins (MEMri $rs1, $simm13):$addr),
                        "st %csr, [$addr]", [], IIC_st>;
   }
-  let Defs = [CPQ] in {
-    def STDCQrr : F3_1<3, 0b110110, (outs (MEMrr $rs1, $rs2):$addr), (ins),
+  let mayStore = 1, Uses = [CPQ] in {
+    def STDCQrr : F3_1<3, 0b110110, (outs), (ins (MEMrr $rs1, $rs2):$addr),
                        "std %cq, [$addr]", [], IIC_std>;
-    def STDCQri : F3_2<3, 0b110110, (outs (MEMri $rs1, $simm13):$addr), (ins),
+    def STDCQri : F3_2<3, 0b110110, (outs), (ins (MEMri $rs1, $simm13):$addr),
                        "std %cq, [$addr]", [], IIC_std>;
   }
 }
 
 let rd = 0 in {
-  let Defs = [FSR] in {
-    def STFSRrr : F3_1<3, 0b100101, (outs (MEMrr $rs1, $rs2):$addr), (ins),
+  let mayStore = 1, Uses = [FSR] in {
+    def STFSRrr : F3_1<3, 0b100101, (outs), (ins (MEMrr $rs1, $rs2):$addr),
 		   "st %fsr, [$addr]", [], IIC_st>;
-    def STFSRri : F3_2<3, 0b100101, (outs (MEMri $rs1, $simm13):$addr), (ins),
+    def STFSRri : F3_2<3, 0b100101, (outs), (ins (MEMri $rs1, $simm13):$addr),
 		   "st %fsr, [$addr]", [], IIC_st>;
   }
-  let Defs = [FQ] in {
-    def STDFQrr : F3_1<3, 0b100110, (outs (MEMrr $rs1, $rs2):$addr), (ins),
+  let mayStore = 1, Defs = [FQ] in {
+    def STDFQrr : F3_1<3, 0b100110, (outs), (ins (MEMrr $rs1, $rs2):$addr),
 		   "std %fq, [$addr]", [], IIC_std>;
-    def STDFQri : F3_2<3, 0b100110, (outs (MEMri $rs1, $simm13):$addr), (ins),
+    def STDFQri : F3_2<3, 0b100110, (outs), (ins (MEMri $rs1, $simm13):$addr),
 		   "std %fq, [$addr]", [], IIC_std>;
   }
 }
-let rd = 1, Defs = [FSR] in {
-  def STXFSRrr : F3_1<3, 0b100101, (outs (MEMrr $rs1, $rs2):$addr), (ins),
+let rd = 1, mayStore = 1, Uses = [FSR] in {
+  def STXFSRrr : F3_1<3, 0b100101, (outs), (ins (MEMrr $rs1, $rs2):$addr),
 		 "stx %fsr, [$addr]", []>, Requires<[HasV9]>;
-  def STXFSRri : F3_2<3, 0b100101, (outs (MEMri $rs1, $simm13):$addr), (ins),
+  def STXFSRri : F3_2<3, 0b100101, (outs), (ins (MEMri $rs1, $simm13):$addr),
 		 "stx %fsr, [$addr]", []>, Requires<[HasV9]>;
 }
 

From 62853a246ef131c4de73b63a141c85a0b14c75a5 Mon Sep 17 00:00:00 2001
From: Michael Liao <michael.hliao@gmail.com>
Date: Tue, 16 Apr 2024 15:20:32 -0400
Subject: [PATCH 680/886] [TableGen][InstrInfoEmitter] Count sub-operands on
 def operands

- If a def operand includes multiple sub-operands, count them when
  generating instr info.
- Found issues in x86 and sparc backends, where memory operands of
  store or store-like instructions are wrongly placed in the output
  list.

Reviewers: jayfoad, arsenm, Pierre-vh

Reviewed By: arsenm

Pull Request: https://github.com/llvm/llvm-project/pull/88972
---
 llvm/test/TableGen/def-multiple-operands.td | 37 +++++++++++++++++++++
 llvm/utils/TableGen/InstrInfoEmitter.cpp    |  8 ++++-
 2 files changed, 44 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/TableGen/def-multiple-operands.td

diff --git a/llvm/test/TableGen/def-multiple-operands.td b/llvm/test/TableGen/def-multiple-operands.td
new file mode 100644
index 0000000000000..b747c58907505
--- /dev/null
+++ b/llvm/test/TableGen/def-multiple-operands.td
@@ -0,0 +1,37 @@
+// RUN: llvm-tblgen -gen-instr-info -I %p/../../include %s | FileCheck %s
+
+include "llvm/Target/Target.td"
+
+def archInstrInfo : InstrInfo {}
+
+def arch : Target {
+  let InstructionSet = archInstrInfo;
+}
+
+def R0 : Register<"r0">;
+def P0 : Register<"p0">;
+def R32 : RegisterClass<"MyNS", [i32], 0, (add R0)>;
+def P1 : RegisterClass<"MyNS", [i1], 0, (add P0)>;
+
+def Reg3Opnd : Operand<OtherVT> {
+  let MIOperandInfo = (ops R32, R32, P1);
+}
+
+// The following checks verify that 'MCInstrDesc' entry for 'InstA' has the
+// expected 'NumOperands' and 'NumDefs', i.e. 'InstA' should have 3 defs out of
+// 4 operands.
+
+// CHECK: archInstrTable {{.* = \{}}
+// CHECK: {{\{}}
+// CHECK: {{\{}} [[ID:[0-9]+]], 4, 3, 13, {{.+\}, \/\/}}
+// CHECK-SAME: Inst #[[ID]] = InstA
+def InstA : Instruction {
+  let Namespace = "MyNS";
+  let Size = 13;
+  // InstA should have 3 defs out of 4 operands.
+  let OutOperandList = (outs Reg3Opnd:$dst);
+  let InOperandList = (ins i32imm:$c);
+  field bits<8> Inst;
+  field bits<8> SoftFail = 0;
+  let hasSideEffects = false;
+}
diff --git a/llvm/utils/TableGen/InstrInfoEmitter.cpp b/llvm/utils/TableGen/InstrInfoEmitter.cpp
index 36f8fa1465393..b3a05e081f637 100644
--- a/llvm/utils/TableGen/InstrInfoEmitter.cpp
+++ b/llvm/utils/TableGen/InstrInfoEmitter.cpp
@@ -1181,9 +1181,15 @@ void InstrInfoEmitter::emitRecord(
     // Each logical operand can be multiple MI operands.
     MinOperands =
         Inst.Operands.back().MIOperandNo + Inst.Operands.back().MINumOperands;
+  // Even the logical output operand may be multiple MI operands.
+  int DefOperands = 0;
+  if (Inst.Operands.NumDefs) {
+    auto &Opnd = Inst.Operands[Inst.Operands.NumDefs - 1];
+    DefOperands = Opnd.MIOperandNo + Opnd.MINumOperands;
+  }
 
   OS << "    { ";
-  OS << Num << ",\t" << MinOperands << ",\t" << Inst.Operands.NumDefs << ",\t"
+  OS << Num << ",\t" << MinOperands << ",\t" << DefOperands << ",\t"
      << Inst.TheDef->getValueAsInt("Size") << ",\t"
      << SchedModels.getSchedClassIdx(Inst) << ",\t";
 

From d0f718e06848774a4e9d0b253cf75c1408b5f41a Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Tue, 16 Apr 2024 17:47:17 -0700
Subject: [PATCH 681/886] Revert "Improve stack usage to increase recursive
 initialization depth" (#89006)

Reverts llvm/llvm-project#88546

Leak and performance regression.
Details in #88546
---
 clang/docs/ReleaseNotes.rst               |  6 --
 clang/include/clang/Sema/Initialization.h |  6 +-
 clang/include/clang/Sema/Overload.h       | 70 +++++++++++++++++------
 clang/lib/Sema/SemaInit.cpp               | 26 ++++-----
 clang/lib/Sema/SemaOverload.cpp           | 21 ++++---
 5 files changed, 75 insertions(+), 54 deletions(-)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 4aedfafcb26ae..3752b6ce15760 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -203,12 +203,6 @@ Non-comprehensive list of changes in this release
 - ``__typeof_unqual__`` is available in all C modes as an extension, which behaves
   like ``typeof_unqual`` from C23, similar to ``__typeof__`` and ``typeof``.
 
-- Improved stack usage with C++ initialization code. This allows significantly
-  more levels of recursive initialization before reaching stack exhaustion
-  limits. This will positively impact recursive template instantiation code,
-  but should also reduce memory overhead for initializations in general.
-  Fixes #GH88330
-
 New Compiler Flags
 ------------------
 - ``-fsanitize=implicit-bitfield-conversion`` checks implicit truncation and
diff --git a/clang/include/clang/Sema/Initialization.h b/clang/include/clang/Sema/Initialization.h
index 1ceacf0f49f56..2072cd8d1c3ef 100644
--- a/clang/include/clang/Sema/Initialization.h
+++ b/clang/include/clang/Sema/Initialization.h
@@ -1134,7 +1134,7 @@ class InitializationSequence {
   OverloadingResult FailedOverloadResult;
 
   /// The candidate set created when initialization failed.
-  std::unique_ptr<OverloadCandidateSet> FailedCandidateSet;
+  OverloadCandidateSet FailedCandidateSet;
 
   /// The incomplete type that caused a failure.
   QualType FailedIncompleteType;
@@ -1403,9 +1403,7 @@ class InitializationSequence {
   /// Retrieve a reference to the candidate set when overload
   /// resolution fails.
   OverloadCandidateSet &getFailedCandidateSet() {
-    assert(FailedCandidateSet &&
-           "this should have been allocated in the constructor!");
-    return *FailedCandidateSet;
+    return FailedCandidateSet;
   }
 
   /// Get the overloading result, for when the initialization
diff --git a/clang/include/clang/Sema/Overload.h b/clang/include/clang/Sema/Overload.h
index e6f88bbf7c4f4..76311b00d2fc5 100644
--- a/clang/include/clang/Sema/Overload.h
+++ b/clang/include/clang/Sema/Overload.h
@@ -37,7 +37,6 @@
 #include <cassert>
 #include <cstddef>
 #include <cstdint>
-#include <memory>
 #include <utility>
 
 namespace clang {
@@ -875,8 +874,7 @@ class Sema;
     ConversionFixItGenerator Fix;
 
     /// Viable - True to indicate that this overload candidate is viable.
-    LLVM_PREFERRED_TYPE(bool)
-    unsigned Viable : 1;
+    bool Viable : 1;
 
     /// Whether this candidate is the best viable function, or tied for being
     /// the best viable function.
@@ -885,14 +883,12 @@ class Sema;
     /// was part of the ambiguity kernel: the minimal non-empty set of viable
     /// candidates such that all elements of the ambiguity kernel are better
     /// than all viable candidates not in the ambiguity kernel.
-    LLVM_PREFERRED_TYPE(bool)
-    unsigned Best : 1;
+    bool Best : 1;
 
     /// IsSurrogate - True to indicate that this candidate is a
     /// surrogate for a conversion to a function pointer or reference
     /// (C++ [over.call.object]).
-    LLVM_PREFERRED_TYPE(bool)
-    unsigned IsSurrogate : 1;
+    bool IsSurrogate : 1;
 
     /// IgnoreObjectArgument - True to indicate that the first
     /// argument's conversion, which for this function represents the
@@ -901,20 +897,18 @@ class Sema;
     /// implicit object argument is just a placeholder) or a
     /// non-static member function when the call doesn't have an
     /// object argument.
-    LLVM_PREFERRED_TYPE(bool)
-    unsigned IgnoreObjectArgument : 1;
+    bool IgnoreObjectArgument : 1;
 
     /// True if the candidate was found using ADL.
-    LLVM_PREFERRED_TYPE(CallExpr::ADLCallKind)
-    unsigned IsADLCandidate : 1;
+    CallExpr::ADLCallKind IsADLCandidate : 1;
 
     /// Whether this is a rewritten candidate, and if so, of what kind?
     LLVM_PREFERRED_TYPE(OverloadCandidateRewriteKind)
     unsigned RewriteKind : 2;
 
     /// FailureKind - The reason why this candidate is not viable.
-    LLVM_PREFERRED_TYPE(OverloadFailureKind)
-    unsigned FailureKind : 5;
+    /// Actually an OverloadFailureKind.
+    unsigned char FailureKind;
 
     /// The number of call arguments that were explicitly provided,
     /// to be used while performing partial ordering of function templates.
@@ -978,9 +972,7 @@ class Sema;
   private:
     friend class OverloadCandidateSet;
     OverloadCandidate()
-        : IsSurrogate(false),
-          IsADLCandidate(static_cast<unsigned>(CallExpr::NotADL)),
-          RewriteKind(CRK_None) {}
+        : IsSurrogate(false), IsADLCandidate(CallExpr::NotADL), RewriteKind(CRK_None) {}
   };
 
   /// OverloadCandidateSet - A set of overload candidates, used in C++
@@ -1078,16 +1070,51 @@ class Sema;
     };
 
   private:
-    SmallVector<OverloadCandidate, 4> Candidates;
-    llvm::SmallPtrSet<uintptr_t, 4> Functions;
+    SmallVector<OverloadCandidate, 16> Candidates;
+    llvm::SmallPtrSet<uintptr_t, 16> Functions;
+
+    // Allocator for ConversionSequenceLists. We store the first few of these
+    // inline to avoid allocation for small sets.
+    llvm::BumpPtrAllocator SlabAllocator;
 
     SourceLocation Loc;
     CandidateSetKind Kind;
     OperatorRewriteInfo RewriteInfo;
 
+    constexpr static unsigned NumInlineBytes =
+        24 * sizeof(ImplicitConversionSequence);
+    unsigned NumInlineBytesUsed = 0;
+    alignas(void *) char InlineSpace[NumInlineBytes];
+
     // Address space of the object being constructed.
     LangAS DestAS = LangAS::Default;
 
+    /// If we have space, allocates from inline storage. Otherwise, allocates
+    /// from the slab allocator.
+    /// FIXME: It would probably be nice to have a SmallBumpPtrAllocator
+    /// instead.
+    /// FIXME: Now that this only allocates ImplicitConversionSequences, do we
+    /// want to un-generalize this?
+    template <typename T>
+    T *slabAllocate(unsigned N) {
+      // It's simpler if this doesn't need to consider alignment.
+      static_assert(alignof(T) == alignof(void *),
+                    "Only works for pointer-aligned types.");
+      static_assert(std::is_trivial<T>::value ||
+                        std::is_same<ImplicitConversionSequence, T>::value,
+                    "Add destruction logic to OverloadCandidateSet::clear().");
+
+      unsigned NBytes = sizeof(T) * N;
+      if (NBytes > NumInlineBytes - NumInlineBytesUsed)
+        return SlabAllocator.Allocate<T>(N);
+      char *FreeSpaceStart = InlineSpace + NumInlineBytesUsed;
+      assert(uintptr_t(FreeSpaceStart) % alignof(void *) == 0 &&
+             "Misaligned storage!");
+
+      NumInlineBytesUsed += NBytes;
+      return reinterpret_cast<T *>(FreeSpaceStart);
+    }
+
     void destroyCandidates();
 
   public:
@@ -1136,7 +1163,12 @@ class Sema;
     ConversionSequenceList
     allocateConversionSequences(unsigned NumConversions) {
       ImplicitConversionSequence *Conversions =
-          new ImplicitConversionSequence[NumConversions];
+          slabAllocate<ImplicitConversionSequence>(NumConversions);
+
+      // Construct the new objects.
+      for (unsigned I = 0; I != NumConversions; ++I)
+        new (&Conversions[I]) ImplicitConversionSequence();
+
       return ConversionSequenceList(Conversions, NumConversions);
     }
 
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index 791c0b6e6df23..fb7a80ab02846 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -6114,8 +6114,7 @@ InitializationSequence::InitializationSequence(
     Sema &S, const InitializedEntity &Entity, const InitializationKind &Kind,
     MultiExprArg Args, bool TopLevelOfInitList, bool TreatUnavailableAsInvalid)
     : FailedOverloadResult(OR_Success),
-      FailedCandidateSet(new OverloadCandidateSet(
-          Kind.getLocation(), OverloadCandidateSet::CSK_Normal)) {
+      FailedCandidateSet(Kind.getLocation(), OverloadCandidateSet::CSK_Normal) {
   InitializeFrom(S, Entity, Kind, Args, TopLevelOfInitList,
                  TreatUnavailableAsInvalid);
 }
@@ -9736,7 +9735,7 @@ bool InitializationSequence::Diagnose(Sema &S,
     switch (FailedOverloadResult) {
     case OR_Ambiguous:
 
-      FailedCandidateSet->NoteCandidates(
+      FailedCandidateSet.NoteCandidates(
           PartialDiagnosticAt(
               Kind.getLocation(),
               Failure == FK_UserConversionOverloadFailed
@@ -9750,8 +9749,7 @@ bool InitializationSequence::Diagnose(Sema &S,
       break;
 
     case OR_No_Viable_Function: {
-      auto Cands =
-          FailedCandidateSet->CompleteCandidates(S, OCD_AllCandidates, Args);
+      auto Cands = FailedCandidateSet.CompleteCandidates(S, OCD_AllCandidates, Args);
       if (!S.RequireCompleteType(Kind.getLocation(),
                                  DestType.getNonReferenceType(),
                           diag::err_typecheck_nonviable_condition_incomplete,
@@ -9761,13 +9759,13 @@ bool InitializationSequence::Diagnose(Sema &S,
           << OnlyArg->getType() << Args[0]->getSourceRange()
           << DestType.getNonReferenceType();
 
-      FailedCandidateSet->NoteCandidates(S, Args, Cands);
+      FailedCandidateSet.NoteCandidates(S, Args, Cands);
       break;
     }
     case OR_Deleted: {
       OverloadCandidateSet::iterator Best;
-      OverloadingResult Ovl =
-          FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
+      OverloadingResult Ovl
+        = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
 
       StringLiteral *Msg = Best->Function->getDeletedMessage();
       S.Diag(Kind.getLocation(), diag::err_typecheck_deleted_function)
@@ -9951,7 +9949,7 @@ bool InitializationSequence::Diagnose(Sema &S,
     // bad.
     switch (FailedOverloadResult) {
       case OR_Ambiguous:
-        FailedCandidateSet->NoteCandidates(
+        FailedCandidateSet.NoteCandidates(
             PartialDiagnosticAt(Kind.getLocation(),
                                 S.PDiag(diag::err_ovl_ambiguous_init)
                                     << DestType << ArgsRange),
@@ -10005,7 +10003,7 @@ bool InitializationSequence::Diagnose(Sema &S,
           break;
         }
 
-        FailedCandidateSet->NoteCandidates(
+        FailedCandidateSet.NoteCandidates(
             PartialDiagnosticAt(
                 Kind.getLocation(),
                 S.PDiag(diag::err_ovl_no_viable_function_in_init)
@@ -10015,8 +10013,8 @@ bool InitializationSequence::Diagnose(Sema &S,
 
       case OR_Deleted: {
         OverloadCandidateSet::iterator Best;
-        OverloadingResult Ovl =
-            FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
+        OverloadingResult Ovl
+          = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
         if (Ovl != OR_Deleted) {
           S.Diag(Kind.getLocation(), diag::err_ovl_deleted_init)
               << DestType << ArgsRange;
@@ -10095,8 +10093,8 @@ bool InitializationSequence::Diagnose(Sema &S,
     S.Diag(Kind.getLocation(), diag::err_selected_explicit_constructor)
       << Args[0]->getSourceRange();
     OverloadCandidateSet::iterator Best;
-    OverloadingResult Ovl =
-        FailedCandidateSet->BestViableFunction(S, Kind.getLocation(), Best);
+    OverloadingResult Ovl
+      = FailedCandidateSet.BestViableFunction(S, Kind.getLocation(), Best);
     (void)Ovl;
     assert(Ovl == OR_Success && "Inconsistent overload resolution");
     CXXConstructorDecl *CtorDecl = cast<CXXConstructorDecl>(Best->Function);
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index bcde0d86cf10f..227ef564ba3e0 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -1057,7 +1057,8 @@ bool OverloadCandidateSet::OperatorRewriteInfo::shouldAddReversed(
 
 void OverloadCandidateSet::destroyCandidates() {
   for (iterator i = begin(), e = end(); i != e; ++i) {
-    delete[] i->Conversions.data();
+    for (auto &C : i->Conversions)
+      C.~ImplicitConversionSequence();
     if (!i->Viable && i->FailureKind == ovl_fail_bad_deduction)
       i->DeductionFailure.Destroy();
   }
@@ -1065,6 +1066,8 @@ void OverloadCandidateSet::destroyCandidates() {
 
 void OverloadCandidateSet::clear(CandidateSetKind CSK) {
   destroyCandidates();
+  SlabAllocator.Reset();
+  NumInlineBytesUsed = 0;
   Candidates.clear();
   Functions.clear();
   Kind = CSK;
@@ -6980,7 +6983,7 @@ void Sema::AddOverloadCandidate(
   Candidate.RewriteKind =
       CandidateSet.getRewriteInfo().getRewriteKind(Function, PO);
   Candidate.IsSurrogate = false;
-  Candidate.IsADLCandidate = static_cast<unsigned>(IsADLCandidate);
+  Candidate.IsADLCandidate = IsADLCandidate;
   Candidate.IgnoreObjectArgument = false;
   Candidate.ExplicitCallArguments = Args.size();
 
@@ -7812,7 +7815,7 @@ void Sema::AddTemplateOverloadCandidate(
     Candidate.RewriteKind =
       CandidateSet.getRewriteInfo().getRewriteKind(Candidate.Function, PO);
     Candidate.IsSurrogate = false;
-    Candidate.IsADLCandidate = static_cast<unsigned>(IsADLCandidate);
+    Candidate.IsADLCandidate = IsADLCandidate;
     // Ignore the object argument if there is one, since we don't have an object
     // type.
     Candidate.IgnoreObjectArgument =
@@ -14122,8 +14125,7 @@ static ExprResult FinishOverloadedCallExpr(Sema &SemaRef, Scope *S, Expr *Fn,
       return ExprError();
     return SemaRef.BuildResolvedCallExpr(
         Res.get(), FDecl, LParenLoc, Args, RParenLoc, ExecConfig,
-        /*IsExecConfig=*/false,
-        static_cast<CallExpr::ADLCallKind>((*Best)->IsADLCandidate));
+        /*IsExecConfig=*/false, (*Best)->IsADLCandidate);
   }
 
   case OR_No_Viable_Function: {
@@ -14182,8 +14184,7 @@ static ExprResult FinishOverloadedCallExpr(Sema &SemaRef, Scope *S, Expr *Fn,
       return ExprError();
     return SemaRef.BuildResolvedCallExpr(
         Res.get(), FDecl, LParenLoc, Args, RParenLoc, ExecConfig,
-        /*IsExecConfig=*/false,
-        static_cast<CallExpr::ADLCallKind>((*Best)->IsADLCandidate));
+        /*IsExecConfig=*/false, (*Best)->IsADLCandidate);
   }
   }
 
@@ -14490,8 +14491,7 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
       Args[0] = Input;
       CallExpr *TheCall = CXXOperatorCallExpr::Create(
           Context, Op, FnExpr.get(), ArgsArray, ResultTy, VK, OpLoc,
-          CurFPFeatureOverrides(),
-          static_cast<CallExpr::ADLCallKind>(Best->IsADLCandidate));
+          CurFPFeatureOverrides(), Best->IsADLCandidate);
 
       if (CheckCallReturnType(FnDecl->getReturnType(), OpLoc, TheCall, FnDecl))
         return ExprError();
@@ -14909,8 +14909,7 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
         // members; CodeGen should take care not to emit the this pointer.
         TheCall = CXXOperatorCallExpr::Create(
             Context, ChosenOp, FnExpr.get(), Args, ResultTy, VK, OpLoc,
-            CurFPFeatureOverrides(),
-            static_cast<CallExpr::ADLCallKind>(Best->IsADLCandidate));
+            CurFPFeatureOverrides(), Best->IsADLCandidate);
 
         if (const auto *Method = dyn_cast<CXXMethodDecl>(FnDecl);
             Method && Method->isImplicitObjectMemberFunction()) {

From 1f35e7227178843679d1d364bc5fc0bcfee2eb95 Mon Sep 17 00:00:00 2001
From: Vitaly Buka <vitalybuka@google.com>
Date: Tue, 16 Apr 2024 17:50:16 -0700
Subject: [PATCH 682/886] [clang][builtin] Implement
 __builtin_allow_runtime_check (#87568)

RFC:
https://discourse.llvm.org/t/rfc-introduce-new-clang-builtin-builtin-allow-runtime-check/78281

---------

Co-authored-by: Noah Goldstein <goldstein.w.n@gmail.com>
Co-authored-by: Aaron Ballman <aaron@aaronballman.com>
---
 clang/docs/LanguageExtensions.rst             | 48 +++++++++++++++++++
 clang/include/clang/Basic/Builtins.td         |  6 +++
 clang/lib/CodeGen/CGBuiltin.cpp               |  9 ++++
 clang/lib/Sema/SemaChecking.cpp               | 11 +++++
 .../CodeGen/builtin-allow-runtime-check.cpp   | 29 +++++++++++
 clang/test/Sema/builtin-allow-runtime-check.c | 24 ++++++++++
 6 files changed, 127 insertions(+)
 create mode 100644 clang/test/CodeGen/builtin-allow-runtime-check.cpp
 create mode 100644 clang/test/Sema/builtin-allow-runtime-check.c

diff --git a/clang/docs/LanguageExtensions.rst b/clang/docs/LanguageExtensions.rst
index 05c8f765b5569..3bead159c8f94 100644
--- a/clang/docs/LanguageExtensions.rst
+++ b/clang/docs/LanguageExtensions.rst
@@ -3466,6 +3466,54 @@ Query for this feature with ``__has_builtin(__builtin_trap)``.
 
 ``__builtin_arm_trap`` is lowered to the ``llvm.aarch64.break`` builtin, and then to ``brk #payload``.
 
+``__builtin_allow_runtime_check``
+---------------------------------
+
+``__builtin_allow_runtime_check`` return true if the check at the current
+program location should be executed. It is expected to be used to implement
+``assert`` like checks which can be safely removed by optimizer.
+
+**Syntax**:
+
+.. code-block:: c++
+
+    bool __builtin_allow_runtime_check(const char* kind)
+
+**Example of use**:
+
+.. code-block:: c++
+
+  if (__builtin_allow_runtime_check("mycheck") && !ExpensiveCheck()) {
+     abort();
+  }
+
+**Description**
+
+``__builtin_allow_runtime_check`` is lowered to ` ``llvm.allow.runtime.check``
+<https://llvm.org/docs/LangRef.html#llvm-allow-runtime-check-intrinsic>`_
+builtin.
+
+The ``__builtin_allow_runtime_check()`` is expected to be used with control
+flow conditions such as in ``if`` to guard expensive runtime checks. The
+specific rules for selecting permitted checks can differ and are controlled by
+the compiler options.
+
+Flags to control checks:
+* ``-mllvm -lower-allow-check-percentile-cutoff-hot=N`` where N is PGO hotness
+cutoff in range ``[0, 999999]`` to disallow checks in hot code.
+* ``-mllvm -lower-allow-check-random-rate=P`` where P is number in range
+``[0.0, 1.0]`` representation probability of keeping a check.
+* If both flags are specified, ``-lower-allow-check-random-rate`` takes
+precedence.
+* If none is specified, ``__builtin_allow_runtime_check`` is lowered as
+``true``, allowing all checks.
+
+Parameter ``kind`` is a string literal representing a user selected kind for
+guarded check. It's unused now. It will enable kind-specific lowering in future.
+E.g. a higher hotness cutoff can be used for more expensive kind of check.
+
+Query for this feature with ``__has_builtin(__builtin_allow_runtime_check)``.
+
 ``__builtin_nondeterministic_value``
 ------------------------------------
 
diff --git a/clang/include/clang/Basic/Builtins.td b/clang/include/clang/Basic/Builtins.td
index d6ceb450bd106..de721a87b3341 100644
--- a/clang/include/clang/Basic/Builtins.td
+++ b/clang/include/clang/Basic/Builtins.td
@@ -1164,6 +1164,12 @@ def Unreachable : Builtin {
   let Prototype = "void()";
 }
 
+def AllowRuntimeCheck : Builtin {
+  let Spellings = ["__builtin_allow_runtime_check"];
+  let Attributes = [NoThrow, Pure, Const];
+  let Prototype = "bool(char const*)";
+}
+
 def ShuffleVector : Builtin {
   let Spellings = ["__builtin_shufflevector"];
   let Attributes = [NoThrow, Const, CustomTypeChecking];
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 9f95697f284c4..a05874e63c73c 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -3436,6 +3436,15 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Builder.CreateAssumption(ConstantInt::getTrue(getLLVMContext()), {OBD});
     return RValue::get(nullptr);
   }
+  case Builtin::BI__builtin_allow_runtime_check: {
+    StringRef Kind =
+        cast<StringLiteral>(E->getArg(0)->IgnoreParenCasts())->getString();
+    LLVMContext &Ctx = CGM.getLLVMContext();
+    llvm::Value *Allow = Builder.CreateCall(
+        CGM.getIntrinsic(llvm::Intrinsic::allow_runtime_check),
+        llvm::MetadataAsValue::get(Ctx, llvm::MDString::get(Ctx, Kind)));
+    return RValue::get(Allow);
+  }
   case Builtin::BI__arithmetic_fence: {
     // Create the builtin call if FastMath is selected, and the target
     // supports the builtin, otherwise just return the argument.
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index 8e21811b67d90..99b0a00083535 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -3233,6 +3233,17 @@ Sema::CheckBuiltinFunctionCall(FunctionDecl *FDecl, unsigned BuiltinID,
     if (BuiltinCountZeroBitsGeneric(*this, TheCall))
       return ExprError();
     break;
+
+  case Builtin::BI__builtin_allow_runtime_check: {
+    Expr *Arg = TheCall->getArg(0);
+    // Check if the argument is a string literal.
+    if (!isa<StringLiteral>(Arg->IgnoreParenImpCasts())) {
+      Diag(TheCall->getBeginLoc(), diag::err_expr_not_string_literal)
+          << Arg->getSourceRange();
+      return ExprError();
+    }
+    break;
+  }
   }
 
   if (getLangOpts().HLSL && CheckHLSLBuiltinFunctionCall(BuiltinID, TheCall))
diff --git a/clang/test/CodeGen/builtin-allow-runtime-check.cpp b/clang/test/CodeGen/builtin-allow-runtime-check.cpp
new file mode 100644
index 0000000000000..db3f59a9d48a1
--- /dev/null
+++ b/clang/test/CodeGen/builtin-allow-runtime-check.cpp
@@ -0,0 +1,29 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 4
+// RUN: %clang_cc1 -cc1 -triple x86_64-pc-linux-gnu -emit-llvm -o - %s | FileCheck %s
+
+static_assert(__has_builtin(__builtin_allow_runtime_check), "");
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z4testv(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = call i1 @llvm.allow.runtime.check(metadata !"mycheck")
+// CHECK-NEXT:    ret i1 [[TMP0]]
+//
+bool test() {
+  return __builtin_allow_runtime_check("mycheck");
+}
+
+// CHECK-LABEL: define dso_local noundef zeroext i1 @_Z10test_twicev(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = call i1 @llvm.allow.runtime.check(metadata !"mycheck")
+// CHECK-NEXT:    [[CONV:%.*]] = zext i1 [[TMP0]] to i32
+// CHECK-NEXT:    [[TMP1:%.*]] = call i1 @llvm.allow.runtime.check(metadata !"mycheck")
+// CHECK-NEXT:    [[CONV1:%.*]] = zext i1 [[TMP1]] to i32
+// CHECK-NEXT:    [[OR:%.*]] = or i32 [[CONV]], [[CONV1]]
+// CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[OR]], 0
+// CHECK-NEXT:    ret i1 [[TOBOOL]]
+//
+bool test_twice() {
+  return __builtin_allow_runtime_check("mycheck") | __builtin_allow_runtime_check("mycheck");
+}
diff --git a/clang/test/Sema/builtin-allow-runtime-check.c b/clang/test/Sema/builtin-allow-runtime-check.c
new file mode 100644
index 0000000000000..b656861000075
--- /dev/null
+++ b/clang/test/Sema/builtin-allow-runtime-check.c
@@ -0,0 +1,24 @@
+// RUN: %clang_cc1 -fsyntax-only -triple x86_64-pc-linux-gnu -verify %s
+// RUN: %clang_cc1 -fsyntax-only -triple aarch64-linux-gnu -verify %s
+
+extern const char *str;
+
+int main(void) {
+  int r = 0;
+
+  r |= __builtin_allow_runtime_check(); // expected-error {{too few arguments to function call}}
+
+  r |= __builtin_allow_runtime_check(str); // expected-error {{expression is not a string literal}}
+
+  r |= __builtin_allow_runtime_check(5); // expected-error {{incompatible integer to pointer conversion}} expected-error {{expression is not a string literal}}
+
+  r |= __builtin_allow_runtime_check("a", "b");  // expected-error {{too many arguments to function call}}
+
+  r |= __builtin_allow_runtime_check("");
+
+  r |= __builtin_allow_runtime_check("check");
+
+  str = __builtin_allow_runtime_check("check2");  // expected-error {{incompatible integer to pointer conversion}}
+
+  return r;
+}

From 52a4d8123c2a9157f2e543945f7b6148da3ecfdb Mon Sep 17 00:00:00 2001
From: Maksim Panchenko <maks@fb.com>
Date: Tue, 16 Apr 2024 17:51:32 -0700
Subject: [PATCH 683/886] [BOLT][NFC] Remove unused function (#89009)

getFileOffsetFor() was replaced with getFileOffsetForAddress().
---
 bolt/include/bolt/Rewrite/RewriteInstance.h | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/bolt/include/bolt/Rewrite/RewriteInstance.h b/bolt/include/bolt/Rewrite/RewriteInstance.h
index 826677cd63b22..7a261f611eaf2 100644
--- a/bolt/include/bolt/Rewrite/RewriteInstance.h
+++ b/bolt/include/bolt/Rewrite/RewriteInstance.h
@@ -368,13 +368,6 @@ class RewriteInstance {
   /// rewritten binary.
   void patchBuildID();
 
-  /// Return file offset corresponding to a given virtual address.
-  uint64_t getFileOffsetFor(uint64_t Address) {
-    assert(Address >= NewTextSegmentAddress &&
-           "address in not in the new text segment");
-    return Address - NewTextSegmentAddress + NewTextSegmentOffset;
-  }
-
   /// Return file offset corresponding to a virtual \p Address.
   /// Return 0 if the address has no mapping in the file, including being
   /// part of .bss section.

From 0af8caeb2fa4d68fcabe6297383d1cdf1cae8b87 Mon Sep 17 00:00:00 2001
From: Maksim Panchenko <maks@fb.com>
Date: Tue, 16 Apr 2024 17:58:47 -0700
Subject: [PATCH 684/886] [BOLT][NFC] Remove another unused function (#89011)

RewriteInstance::isKSymtabSection() is deprecated.
---
 bolt/include/bolt/Rewrite/RewriteInstance.h | 3 ---
 bolt/lib/Rewrite/RewriteInstance.cpp        | 7 -------
 2 files changed, 10 deletions(-)

diff --git a/bolt/include/bolt/Rewrite/RewriteInstance.h b/bolt/include/bolt/Rewrite/RewriteInstance.h
index 7a261f611eaf2..af832b4c7c84c 100644
--- a/bolt/include/bolt/Rewrite/RewriteInstance.h
+++ b/bolt/include/bolt/Rewrite/RewriteInstance.h
@@ -391,9 +391,6 @@ class RewriteInstance {
   /// Return true if the section holds debug information.
   static bool isDebugSection(StringRef SectionName);
 
-  /// Return true if the section holds linux kernel symbol information.
-  static bool isKSymtabSection(StringRef SectionName);
-
   /// Adds Debug section to overwrite.
   static void addToDebugSectionsToOverwrite(const char *Section) {
     DebugSectionsToOverwrite.emplace_back(Section);
diff --git a/bolt/lib/Rewrite/RewriteInstance.cpp b/bolt/lib/Rewrite/RewriteInstance.cpp
index fd2477231142e..4e0096cf988ae 100644
--- a/bolt/lib/Rewrite/RewriteInstance.cpp
+++ b/bolt/lib/Rewrite/RewriteInstance.cpp
@@ -5767,10 +5767,3 @@ bool RewriteInstance::isDebugSection(StringRef SectionName) {
 
   return false;
 }
-
-bool RewriteInstance::isKSymtabSection(StringRef SectionName) {
-  if (SectionName.starts_with("__ksymtab"))
-    return true;
-
-  return false;
-}

From f40f4fcee506deacda0594362509ee7dddcf5e37 Mon Sep 17 00:00:00 2001
From: Congcong Cai <congcongcai0907@163.com>
Date: Wed, 17 Apr 2024 09:57:30 +0800
Subject: [PATCH 685/886] [clang analysis] ExprMutationAnalyzer support
 recursive forwarding reference (#88843)

Reapply for #88765.
Partially fixes: #60895.
---
 .../UnnecessaryValueParamCheck.cpp            |  10 +-
 .../performance/UnnecessaryValueParamCheck.h  |   3 +-
 clang-tools-extra/docs/ReleaseNotes.rst       |   4 +
 .../misc/const-correctness-templates.cpp      |  15 ++
 .../Analysis/Analyses/ExprMutationAnalyzer.h  | 136 ++++++++++++------
 clang/lib/Analysis/ExprMutationAnalyzer.cpp   | 125 +++++++++-------
 .../Analysis/ExprMutationAnalyzerTest.cpp     |  30 ++++
 7 files changed, 220 insertions(+), 103 deletions(-)

diff --git a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp
index 2fa7cd0baf98f..c507043c367a8 100644
--- a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp
+++ b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp
@@ -85,10 +85,10 @@ void UnnecessaryValueParamCheck::check(const MatchFinder::MatchResult &Result) {
 
   TraversalKindScope RAII(*Result.Context, TK_AsIs);
 
-  FunctionParmMutationAnalyzer &Analyzer =
-      MutationAnalyzers.try_emplace(Function, *Function, *Result.Context)
-          .first->second;
-  if (Analyzer.isMutated(Param))
+  FunctionParmMutationAnalyzer *Analyzer =
+      FunctionParmMutationAnalyzer::getFunctionParmMutationAnalyzer(
+          *Function, *Result.Context, MutationAnalyzerCache);
+  if (Analyzer->isMutated(Param))
     return;
 
   const bool IsConstQualified =
@@ -169,7 +169,7 @@ void UnnecessaryValueParamCheck::storeOptions(
 }
 
 void UnnecessaryValueParamCheck::onEndOfTranslationUnit() {
-  MutationAnalyzers.clear();
+  MutationAnalyzerCache.clear();
 }
 
 void UnnecessaryValueParamCheck::handleMoveFix(const ParmVarDecl &Var,
diff --git a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h
index 1872e3bc9bf29..7250bffd20b2f 100644
--- a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h
+++ b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h
@@ -37,8 +37,7 @@ class UnnecessaryValueParamCheck : public ClangTidyCheck {
   void handleMoveFix(const ParmVarDecl &Var, const DeclRefExpr &CopyArgument,
                      const ASTContext &Context);
 
-  llvm::DenseMap<const FunctionDecl *, FunctionParmMutationAnalyzer>
-      MutationAnalyzers;
+  ExprMutationAnalyzer::Memoized MutationAnalyzerCache;
   utils::IncludeInserter Inserter;
   const std::vector<StringRef> AllowedTypes;
 };
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index 4dfbd8ca49ab9..7095c564444fe 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -221,6 +221,10 @@ Changes in existing checks
   <clang-tidy/checks/llvm/header-guard>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
 
+- Improved :doc:`misc-const-correctness
+  <clang-tidy/checks/misc/const-correctness>` check by avoiding infinite recursion
+  for recursive forwarding reference.
+
 - Improved :doc:`misc-definitions-in-headers
   <clang-tidy/checks/misc/definitions-in-headers>` check by replacing the local
   option `HeaderFileExtensions` by the global option of the same name.
diff --git a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
index 9da468128743e..248374a71dd40 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-templates.cpp
@@ -58,3 +58,18 @@ void concatenate3(Args... args)
     (..., (stream << args));
 }
 } // namespace gh70323
+
+namespace gh60895 {
+
+template <class T> void f1(T &&a);
+template <class T> void f2(T &&a);
+template <class T> void f1(T &&a) { f2<T>(a); }
+template <class T> void f2(T &&a) { f1<T>(a); }
+void f() {
+  int x = 0;
+  // CHECK-MESSAGES:[[@LINE-1]]:3: warning: variable 'x' of type 'int' can be declared 'const'
+  // CHECK-FIXES: int const x = 0;
+  f1(x);
+}
+
+} // namespace gh60895
diff --git a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
index 1ceef944fbc34..117173ba9a095 100644
--- a/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
+++ b/clang/include/clang/Analysis/Analyses/ExprMutationAnalyzer.h
@@ -8,11 +8,9 @@
 #ifndef LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 #define LLVM_CLANG_ANALYSIS_ANALYSES_EXPRMUTATIONANALYZER_H
 
-#include <type_traits>
-
-#include "clang/AST/AST.h"
 #include "clang/ASTMatchers/ASTMatchers.h"
 #include "llvm/ADT/DenseMap.h"
+#include <memory>
 
 namespace clang {
 
@@ -21,14 +19,74 @@ class FunctionParmMutationAnalyzer;
 /// Analyzes whether any mutative operations are applied to an expression within
 /// a given statement.
 class ExprMutationAnalyzer {
+  friend class FunctionParmMutationAnalyzer;
+
 public:
+  struct Memoized {
+    using ResultMap = llvm::DenseMap<const Expr *, const Stmt *>;
+    using FunctionParaAnalyzerMap =
+        llvm::SmallDenseMap<const FunctionDecl *,
+                            std::unique_ptr<FunctionParmMutationAnalyzer>>;
+
+    ResultMap Results;
+    ResultMap PointeeResults;
+    FunctionParaAnalyzerMap FuncParmAnalyzer;
+
+    void clear() {
+      Results.clear();
+      PointeeResults.clear();
+      FuncParmAnalyzer.clear();
+    }
+  };
+  struct Analyzer {
+    Analyzer(const Stmt &Stm, ASTContext &Context, Memoized &Memorized)
+        : Stm(Stm), Context(Context), Memorized(Memorized) {}
+
+    const Stmt *findMutation(const Expr *Exp);
+    const Stmt *findMutation(const Decl *Dec);
+
+    const Stmt *findPointeeMutation(const Expr *Exp);
+    const Stmt *findPointeeMutation(const Decl *Dec);
+    static bool isUnevaluated(const Stmt *Smt, const Stmt &Stm,
+                              ASTContext &Context);
+
+  private:
+    using MutationFinder = const Stmt *(Analyzer::*)(const Expr *);
+
+    const Stmt *findMutationMemoized(const Expr *Exp,
+                                     llvm::ArrayRef<MutationFinder> Finders,
+                                     Memoized::ResultMap &MemoizedResults);
+    const Stmt *tryEachDeclRef(const Decl *Dec, MutationFinder Finder);
+
+    bool isUnevaluated(const Expr *Exp);
+
+    const Stmt *findExprMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
+    const Stmt *findDeclMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
+    const Stmt *
+    findExprPointeeMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
+    const Stmt *
+    findDeclPointeeMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
+
+    const Stmt *findDirectMutation(const Expr *Exp);
+    const Stmt *findMemberMutation(const Expr *Exp);
+    const Stmt *findArrayElementMutation(const Expr *Exp);
+    const Stmt *findCastMutation(const Expr *Exp);
+    const Stmt *findRangeLoopMutation(const Expr *Exp);
+    const Stmt *findReferenceMutation(const Expr *Exp);
+    const Stmt *findFunctionArgMutation(const Expr *Exp);
+
+    const Stmt &Stm;
+    ASTContext &Context;
+    Memoized &Memorized;
+  };
+
   ExprMutationAnalyzer(const Stmt &Stm, ASTContext &Context)
-      : Stm(Stm), Context(Context) {}
+      : Memorized(), A(Stm, Context, Memorized) {}
 
   bool isMutated(const Expr *Exp) { return findMutation(Exp) != nullptr; }
   bool isMutated(const Decl *Dec) { return findMutation(Dec) != nullptr; }
-  const Stmt *findMutation(const Expr *Exp);
-  const Stmt *findMutation(const Decl *Dec);
+  const Stmt *findMutation(const Expr *Exp) { return A.findMutation(Exp); }
+  const Stmt *findMutation(const Decl *Dec) { return A.findMutation(Dec); }
 
   bool isPointeeMutated(const Expr *Exp) {
     return findPointeeMutation(Exp) != nullptr;
@@ -36,51 +94,40 @@ class ExprMutationAnalyzer {
   bool isPointeeMutated(const Decl *Dec) {
     return findPointeeMutation(Dec) != nullptr;
   }
-  const Stmt *findPointeeMutation(const Expr *Exp);
-  const Stmt *findPointeeMutation(const Decl *Dec);
+  const Stmt *findPointeeMutation(const Expr *Exp) {
+    return A.findPointeeMutation(Exp);
+  }
+  const Stmt *findPointeeMutation(const Decl *Dec) {
+    return A.findPointeeMutation(Dec);
+  }
+
   static bool isUnevaluated(const Stmt *Smt, const Stmt &Stm,
-                            ASTContext &Context);
+                            ASTContext &Context) {
+    return Analyzer::isUnevaluated(Smt, Stm, Context);
+  }
 
 private:
-  using MutationFinder = const Stmt *(ExprMutationAnalyzer::*)(const Expr *);
-  using ResultMap = llvm::DenseMap<const Expr *, const Stmt *>;
-
-  const Stmt *findMutationMemoized(const Expr *Exp,
-                                   llvm::ArrayRef<MutationFinder> Finders,
-                                   ResultMap &MemoizedResults);
-  const Stmt *tryEachDeclRef(const Decl *Dec, MutationFinder Finder);
-
-  bool isUnevaluated(const Expr *Exp);
-
-  const Stmt *findExprMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
-  const Stmt *findDeclMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
-  const Stmt *
-  findExprPointeeMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
-  const Stmt *
-  findDeclPointeeMutation(ArrayRef<ast_matchers::BoundNodes> Matches);
-
-  const Stmt *findDirectMutation(const Expr *Exp);
-  const Stmt *findMemberMutation(const Expr *Exp);
-  const Stmt *findArrayElementMutation(const Expr *Exp);
-  const Stmt *findCastMutation(const Expr *Exp);
-  const Stmt *findRangeLoopMutation(const Expr *Exp);
-  const Stmt *findReferenceMutation(const Expr *Exp);
-  const Stmt *findFunctionArgMutation(const Expr *Exp);
-
-  const Stmt &Stm;
-  ASTContext &Context;
-  llvm::DenseMap<const FunctionDecl *,
-                 std::unique_ptr<FunctionParmMutationAnalyzer>>
-      FuncParmAnalyzer;
-  ResultMap Results;
-  ResultMap PointeeResults;
+  Memoized Memorized;
+  Analyzer A;
 };
 
 // A convenient wrapper around ExprMutationAnalyzer for analyzing function
 // params.
 class FunctionParmMutationAnalyzer {
 public:
-  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context);
+  static FunctionParmMutationAnalyzer *
+  getFunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context,
+                                  ExprMutationAnalyzer::Memoized &Memorized) {
+    auto it = Memorized.FuncParmAnalyzer.find(&Func);
+    if (it == Memorized.FuncParmAnalyzer.end())
+      it =
+          Memorized.FuncParmAnalyzer
+              .try_emplace(&Func, std::unique_ptr<FunctionParmMutationAnalyzer>(
+                                      new FunctionParmMutationAnalyzer(
+                                          Func, Context, Memorized)))
+              .first;
+    return it->getSecond().get();
+  }
 
   bool isMutated(const ParmVarDecl *Parm) {
     return findMutation(Parm) != nullptr;
@@ -88,8 +135,11 @@ class FunctionParmMutationAnalyzer {
   const Stmt *findMutation(const ParmVarDecl *Parm);
 
 private:
-  ExprMutationAnalyzer BodyAnalyzer;
+  ExprMutationAnalyzer::Analyzer BodyAnalyzer;
   llvm::DenseMap<const ParmVarDecl *, const Stmt *> Results;
+
+  FunctionParmMutationAnalyzer(const FunctionDecl &Func, ASTContext &Context,
+                               ExprMutationAnalyzer::Memoized &Memorized);
 };
 
 } // namespace clang
diff --git a/clang/lib/Analysis/ExprMutationAnalyzer.cpp b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
index bb042760d297a..941322be8f870 100644
--- a/clang/lib/Analysis/ExprMutationAnalyzer.cpp
+++ b/clang/lib/Analysis/ExprMutationAnalyzer.cpp
@@ -186,9 +186,10 @@ template <> struct NodeID<Decl> { static constexpr StringRef value = "decl"; };
 constexpr StringRef NodeID<Expr>::value;
 constexpr StringRef NodeID<Decl>::value;
 
-template <class T, class F = const Stmt *(ExprMutationAnalyzer::*)(const T *)>
+template <class T,
+          class F = const Stmt *(ExprMutationAnalyzer::Analyzer::*)(const T *)>
 const Stmt *tryEachMatch(ArrayRef<ast_matchers::BoundNodes> Matches,
-                         ExprMutationAnalyzer *Analyzer, F Finder) {
+                         ExprMutationAnalyzer::Analyzer *Analyzer, F Finder) {
   const StringRef ID = NodeID<T>::value;
   for (const auto &Nodes : Matches) {
     if (const Stmt *S = (Analyzer->*Finder)(Nodes.getNodeAs<T>(ID)))
@@ -199,33 +200,37 @@ const Stmt *tryEachMatch(ArrayRef<ast_matchers::BoundNodes> Matches,
 
 } // namespace
 
-const Stmt *ExprMutationAnalyzer::findMutation(const Expr *Exp) {
-  return findMutationMemoized(Exp,
-                              {&ExprMutationAnalyzer::findDirectMutation,
-                               &ExprMutationAnalyzer::findMemberMutation,
-                               &ExprMutationAnalyzer::findArrayElementMutation,
-                               &ExprMutationAnalyzer::findCastMutation,
-                               &ExprMutationAnalyzer::findRangeLoopMutation,
-                               &ExprMutationAnalyzer::findReferenceMutation,
-                               &ExprMutationAnalyzer::findFunctionArgMutation},
-                              Results);
+const Stmt *ExprMutationAnalyzer::Analyzer::findMutation(const Expr *Exp) {
+  return findMutationMemoized(
+      Exp,
+      {&ExprMutationAnalyzer::Analyzer::findDirectMutation,
+       &ExprMutationAnalyzer::Analyzer::findMemberMutation,
+       &ExprMutationAnalyzer::Analyzer::findArrayElementMutation,
+       &ExprMutationAnalyzer::Analyzer::findCastMutation,
+       &ExprMutationAnalyzer::Analyzer::findRangeLoopMutation,
+       &ExprMutationAnalyzer::Analyzer::findReferenceMutation,
+       &ExprMutationAnalyzer::Analyzer::findFunctionArgMutation},
+      Memorized.Results);
 }
 
-const Stmt *ExprMutationAnalyzer::findMutation(const Decl *Dec) {
-  return tryEachDeclRef(Dec, &ExprMutationAnalyzer::findMutation);
+const Stmt *ExprMutationAnalyzer::Analyzer::findMutation(const Decl *Dec) {
+  return tryEachDeclRef(Dec, &ExprMutationAnalyzer::Analyzer::findMutation);
 }
 
-const Stmt *ExprMutationAnalyzer::findPointeeMutation(const Expr *Exp) {
-  return findMutationMemoized(Exp, {/*TODO*/}, PointeeResults);
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findPointeeMutation(const Expr *Exp) {
+  return findMutationMemoized(Exp, {/*TODO*/}, Memorized.PointeeResults);
 }
 
-const Stmt *ExprMutationAnalyzer::findPointeeMutation(const Decl *Dec) {
-  return tryEachDeclRef(Dec, &ExprMutationAnalyzer::findPointeeMutation);
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findPointeeMutation(const Decl *Dec) {
+  return tryEachDeclRef(Dec,
+                        &ExprMutationAnalyzer::Analyzer::findPointeeMutation);
 }
 
-const Stmt *ExprMutationAnalyzer::findMutationMemoized(
+const Stmt *ExprMutationAnalyzer::Analyzer::findMutationMemoized(
     const Expr *Exp, llvm::ArrayRef<MutationFinder> Finders,
-    ResultMap &MemoizedResults) {
+    Memoized::ResultMap &MemoizedResults) {
   const auto Memoized = MemoizedResults.find(Exp);
   if (Memoized != MemoizedResults.end())
     return Memoized->second;
@@ -241,8 +246,9 @@ const Stmt *ExprMutationAnalyzer::findMutationMemoized(
   return MemoizedResults[Exp] = nullptr;
 }
 
-const Stmt *ExprMutationAnalyzer::tryEachDeclRef(const Decl *Dec,
-                                                 MutationFinder Finder) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::tryEachDeclRef(const Decl *Dec,
+                                               MutationFinder Finder) {
   const auto Refs = match(
       findAll(
           declRefExpr(to(
@@ -261,8 +267,9 @@ const Stmt *ExprMutationAnalyzer::tryEachDeclRef(const Decl *Dec,
   return nullptr;
 }
 
-bool ExprMutationAnalyzer::isUnevaluated(const Stmt *Exp, const Stmt &Stm,
-                                         ASTContext &Context) {
+bool ExprMutationAnalyzer::Analyzer::isUnevaluated(const Stmt *Exp,
+                                                   const Stmt &Stm,
+                                                   ASTContext &Context) {
   return selectFirst<Stmt>(
              NodeID<Expr>::value,
              match(
@@ -293,33 +300,36 @@ bool ExprMutationAnalyzer::isUnevaluated(const Stmt *Exp, const Stmt &Stm,
                  Stm, Context)) != nullptr;
 }
 
-bool ExprMutationAnalyzer::isUnevaluated(const Expr *Exp) {
+bool ExprMutationAnalyzer::Analyzer::isUnevaluated(const Expr *Exp) {
   return isUnevaluated(Exp, Stm, Context);
 }
 
 const Stmt *
-ExprMutationAnalyzer::findExprMutation(ArrayRef<BoundNodes> Matches) {
-  return tryEachMatch<Expr>(Matches, this, &ExprMutationAnalyzer::findMutation);
+ExprMutationAnalyzer::Analyzer::findExprMutation(ArrayRef<BoundNodes> Matches) {
+  return tryEachMatch<Expr>(Matches, this,
+                            &ExprMutationAnalyzer::Analyzer::findMutation);
 }
 
 const Stmt *
-ExprMutationAnalyzer::findDeclMutation(ArrayRef<BoundNodes> Matches) {
-  return tryEachMatch<Decl>(Matches, this, &ExprMutationAnalyzer::findMutation);
+ExprMutationAnalyzer::Analyzer::findDeclMutation(ArrayRef<BoundNodes> Matches) {
+  return tryEachMatch<Decl>(Matches, this,
+                            &ExprMutationAnalyzer::Analyzer::findMutation);
 }
 
-const Stmt *ExprMutationAnalyzer::findExprPointeeMutation(
+const Stmt *ExprMutationAnalyzer::Analyzer::findExprPointeeMutation(
     ArrayRef<ast_matchers::BoundNodes> Matches) {
-  return tryEachMatch<Expr>(Matches, this,
-                            &ExprMutationAnalyzer::findPointeeMutation);
+  return tryEachMatch<Expr>(
+      Matches, this, &ExprMutationAnalyzer::Analyzer::findPointeeMutation);
 }
 
-const Stmt *ExprMutationAnalyzer::findDeclPointeeMutation(
+const Stmt *ExprMutationAnalyzer::Analyzer::findDeclPointeeMutation(
     ArrayRef<ast_matchers::BoundNodes> Matches) {
-  return tryEachMatch<Decl>(Matches, this,
-                            &ExprMutationAnalyzer::findPointeeMutation);
+  return tryEachMatch<Decl>(
+      Matches, this, &ExprMutationAnalyzer::Analyzer::findPointeeMutation);
 }
 
-const Stmt *ExprMutationAnalyzer::findDirectMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findDirectMutation(const Expr *Exp) {
   // LHS of any assignment operators.
   const auto AsAssignmentLhs =
       binaryOperator(isAssignmentOperator(), hasLHS(canResolveToExpr(Exp)));
@@ -426,7 +436,7 @@ const Stmt *ExprMutationAnalyzer::findDirectMutation(const Expr *Exp) {
   const auto AsNonConstRefReturn =
       returnStmt(hasReturnValue(canResolveToExpr(Exp)));
 
-  // It is used as a non-const-reference for initalizing a range-for loop.
+  // It is used as a non-const-reference for initializing a range-for loop.
   const auto AsNonConstRefRangeInit = cxxForRangeStmt(hasRangeInit(declRefExpr(
       allOf(canResolveToExpr(Exp), hasType(nonConstReferenceType())))));
 
@@ -443,7 +453,8 @@ const Stmt *ExprMutationAnalyzer::findDirectMutation(const Expr *Exp) {
   return selectFirst<Stmt>("stmt", Matches);
 }
 
-const Stmt *ExprMutationAnalyzer::findMemberMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findMemberMutation(const Expr *Exp) {
   // Check whether any member of 'Exp' is mutated.
   const auto MemberExprs = match(
       findAll(expr(anyOf(memberExpr(hasObjectExpression(canResolveToExpr(Exp))),
@@ -456,7 +467,8 @@ const Stmt *ExprMutationAnalyzer::findMemberMutation(const Expr *Exp) {
   return findExprMutation(MemberExprs);
 }
 
-const Stmt *ExprMutationAnalyzer::findArrayElementMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findArrayElementMutation(const Expr *Exp) {
   // Check whether any element of an array is mutated.
   const auto SubscriptExprs = match(
       findAll(arraySubscriptExpr(
@@ -469,7 +481,7 @@ const Stmt *ExprMutationAnalyzer::findArrayElementMutation(const Expr *Exp) {
   return findExprMutation(SubscriptExprs);
 }
 
-const Stmt *ExprMutationAnalyzer::findCastMutation(const Expr *Exp) {
+const Stmt *ExprMutationAnalyzer::Analyzer::findCastMutation(const Expr *Exp) {
   // If the 'Exp' is explicitly casted to a non-const reference type the
   // 'Exp' is considered to be modified.
   const auto ExplicitCast =
@@ -504,7 +516,8 @@ const Stmt *ExprMutationAnalyzer::findCastMutation(const Expr *Exp) {
   return findExprMutation(Calls);
 }
 
-const Stmt *ExprMutationAnalyzer::findRangeLoopMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findRangeLoopMutation(const Expr *Exp) {
   // Keep the ordering for the specific initialization matches to happen first,
   // because it is cheaper to match all potential modifications of the loop
   // variable.
@@ -567,7 +580,8 @@ const Stmt *ExprMutationAnalyzer::findRangeLoopMutation(const Expr *Exp) {
   return findDeclMutation(LoopVars);
 }
 
-const Stmt *ExprMutationAnalyzer::findReferenceMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findReferenceMutation(const Expr *Exp) {
   // Follow non-const reference returned by `operator*()` of move-only classes.
   // These are typically smart pointers with unique ownership so we treat
   // mutation of pointee as mutation of the smart pointer itself.
@@ -599,7 +613,8 @@ const Stmt *ExprMutationAnalyzer::findReferenceMutation(const Expr *Exp) {
   return findDeclMutation(Refs);
 }
 
-const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
+const Stmt *
+ExprMutationAnalyzer::Analyzer::findFunctionArgMutation(const Expr *Exp) {
   const auto NonConstRefParam = forEachArgumentWithParam(
       canResolveToExpr(Exp),
       parmVarDecl(hasType(nonConstReferenceType())).bind("parm"));
@@ -637,10 +652,9 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
     if (const auto *RefType = ParmType->getAs<RValueReferenceType>()) {
       if (!RefType->getPointeeType().getQualifiers() &&
           RefType->getPointeeType()->getAs<TemplateTypeParmType>()) {
-        std::unique_ptr<FunctionParmMutationAnalyzer> &Analyzer =
-            FuncParmAnalyzer[Func];
-        if (!Analyzer)
-          Analyzer.reset(new FunctionParmMutationAnalyzer(*Func, Context));
+        FunctionParmMutationAnalyzer *Analyzer =
+            FunctionParmMutationAnalyzer::getFunctionParmMutationAnalyzer(
+                *Func, Context, Memorized);
         if (Analyzer->findMutation(Parm))
           return Exp;
         continue;
@@ -653,13 +667,15 @@ const Stmt *ExprMutationAnalyzer::findFunctionArgMutation(const Expr *Exp) {
 }
 
 FunctionParmMutationAnalyzer::FunctionParmMutationAnalyzer(
-    const FunctionDecl &Func, ASTContext &Context)
-    : BodyAnalyzer(*Func.getBody(), Context) {
+    const FunctionDecl &Func, ASTContext &Context,
+    ExprMutationAnalyzer::Memoized &Memorized)
+    : BodyAnalyzer(*Func.getBody(), Context, Memorized) {
   if (const auto *Ctor = dyn_cast<CXXConstructorDecl>(&Func)) {
     // CXXCtorInitializer might also mutate Param but they're not part of
     // function body, check them eagerly here since they're typically trivial.
     for (const CXXCtorInitializer *Init : Ctor->inits()) {
-      ExprMutationAnalyzer InitAnalyzer(*Init->getInit(), Context);
+      ExprMutationAnalyzer::Analyzer InitAnalyzer(*Init->getInit(), Context,
+                                                  Memorized);
       for (const ParmVarDecl *Parm : Ctor->parameters()) {
         if (Results.contains(Parm))
           continue;
@@ -675,11 +691,14 @@ FunctionParmMutationAnalyzer::findMutation(const ParmVarDecl *Parm) {
   const auto Memoized = Results.find(Parm);
   if (Memoized != Results.end())
     return Memoized->second;
-
+  // To handle call A -> call B -> call A. Assume parameters of A is not mutated
+  // before analyzing parameters of A. Then when analyzing the second "call A",
+  // FunctionParmMutationAnalyzer can use this memoized value to avoid infinite
+  // recursion.
+  Results[Parm] = nullptr;
   if (const Stmt *S = BodyAnalyzer.findMutation(Parm))
     return Results[Parm] = S;
-
-  return Results[Parm] = nullptr;
+  return Results[Parm];
 }
 
 } // namespace clang
diff --git a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
index f58ce4aebcbfc..9c1dc1a76db63 100644
--- a/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
+++ b/clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp
@@ -977,6 +977,36 @@ TEST(ExprMutationAnalyzerTest, FollowFuncArgModified) {
                          "void f() { int x; g(x); }");
   Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
   EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("g(x)"));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_FALSE(isMutated(Results, AST.get()));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); a++; }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
+
+  AST = buildASTFromCode(
+      StdRemoveReference + StdForward +
+      "template <class T> void f1(T &&a);"
+      "template <class T> void f2(T &&a);"
+      "template <class T> void f1(T &&a) { f2<T>(std::forward<T>(a)); a++; }"
+      "template <class T> void f2(T &&a) { f1<T>(std::forward<T>(a)); }"
+      "void f() { int x; f1(x); }");
+  Results = match(withEnclosingCompound(declRefTo("x")), AST->getASTContext());
+  EXPECT_THAT(mutatedBy(Results, AST.get()), ElementsAre("f1(x)"));
 }
 
 TEST(ExprMutationAnalyzerTest, FollowFuncArgNotModified) {

From 3204f3e30b8e15ce6e5d10d5d7bfbaa7cf5cd1f4 Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Wed, 17 Apr 2024 10:07:57 +0800
Subject: [PATCH 686/886] [RISCV] Convert VTYPE operand check to assert in
 RISCVInsertVSETVLI. NFC

The VTYPE operands of a vsetvli pseudo are always immediates
---
 llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
index a14f9a2835473..a54a1148cf28b 100644
--- a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
@@ -1522,9 +1522,7 @@ static bool canMutatePriorConfig(const MachineInstr &PrevMI,
     }
   }
 
-  if (!PrevMI.getOperand(2).isImm() || !MI.getOperand(2).isImm())
-    return false;
-
+  assert(PrevMI.getOperand(2).isImm() && MI.getOperand(2).isImm());
   auto PriorVType = PrevMI.getOperand(2).getImm();
   auto VType = MI.getOperand(2).getImm();
   return areCompatibleVTYPEs(PriorVType, VType, Used);

From c81e5faa6f55d3e390b5e550f78ab08fc6a65ee9 Mon Sep 17 00:00:00 2001
From: Brandon Wu <brandon.wu@sifive.com>
Date: Wed, 17 Apr 2024 10:42:40 +0800
Subject: [PATCH 687/886] [RISCV] Add CFI information for vector callee-saved
 registers (#86811)

Currently the CFI offset for RVV registers are not handled entirely,
this patch add those information for either stack unwinding or
debugger to work correctly on RVV callee-saved stack object.

Depends On D154576

Differential Revision: https://reviews.llvm.org/D156846
---
 llvm/lib/Target/RISCV/RISCVFrameLowering.cpp  | 109 ++++++++++++++---
 llvm/lib/Target/RISCV/RISCVFrameLowering.h    |   3 +
 llvm/test/CodeGen/RISCV/rvv-cfi-info.ll       | 111 ++++++++++++++++++
 .../rvv/fixed-vectors-insert-subvector.ll     |   2 +
 4 files changed, 207 insertions(+), 18 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/rvv-cfi-info.ll

diff --git a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
index 71672ed7b4ae7..cb41577c5d943 100644
--- a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
@@ -435,6 +435,33 @@ void RISCVFrameLowering::adjustStackForRVV(MachineFunction &MF,
                Flag, getStackAlign());
 }
 
+static void appendScalableVectorExpression(const TargetRegisterInfo &TRI,
+                                           SmallVectorImpl<char> &Expr,
+                                           int FixedOffset, int ScalableOffset,
+                                           llvm::raw_string_ostream &Comment) {
+  unsigned DwarfVLenB = TRI.getDwarfRegNum(RISCV::VLENB, true);
+  uint8_t Buffer[16];
+  if (FixedOffset) {
+    Expr.push_back(dwarf::DW_OP_consts);
+    Expr.append(Buffer, Buffer + encodeSLEB128(FixedOffset, Buffer));
+    Expr.push_back((uint8_t)dwarf::DW_OP_plus);
+    Comment << (FixedOffset < 0 ? " - " : " + ") << std::abs(FixedOffset);
+  }
+
+  Expr.push_back((uint8_t)dwarf::DW_OP_consts);
+  Expr.append(Buffer, Buffer + encodeSLEB128(ScalableOffset, Buffer));
+
+  Expr.push_back((uint8_t)dwarf::DW_OP_bregx);
+  Expr.append(Buffer, Buffer + encodeULEB128(DwarfVLenB, Buffer));
+  Expr.push_back(0);
+
+  Expr.push_back((uint8_t)dwarf::DW_OP_mul);
+  Expr.push_back((uint8_t)dwarf::DW_OP_plus);
+
+  Comment << (ScalableOffset < 0 ? " - " : " + ") << std::abs(ScalableOffset)
+          << " * vlenb";
+}
+
 static MCCFIInstruction createDefCFAExpression(const TargetRegisterInfo &TRI,
                                                Register Reg,
                                                uint64_t FixedOffset,
@@ -452,30 +479,38 @@ static MCCFIInstruction createDefCFAExpression(const TargetRegisterInfo &TRI,
   else
     Comment << printReg(Reg, &TRI);
 
-  uint8_t buffer[16];
-  if (FixedOffset) {
-    Expr.push_back(dwarf::DW_OP_consts);
-    Expr.append(buffer, buffer + encodeSLEB128(FixedOffset, buffer));
-    Expr.push_back((uint8_t)dwarf::DW_OP_plus);
-    Comment << " + " << FixedOffset;
-  }
+  appendScalableVectorExpression(TRI, Expr, FixedOffset, ScalableOffset,
+                                 Comment);
 
-  Expr.push_back((uint8_t)dwarf::DW_OP_consts);
-  Expr.append(buffer, buffer + encodeSLEB128(ScalableOffset, buffer));
+  SmallString<64> DefCfaExpr;
+  uint8_t Buffer[16];
+  DefCfaExpr.push_back(dwarf::DW_CFA_def_cfa_expression);
+  DefCfaExpr.append(Buffer, Buffer + encodeULEB128(Expr.size(), Buffer));
+  DefCfaExpr.append(Expr.str());
 
-  unsigned DwarfVlenb = TRI.getDwarfRegNum(RISCV::VLENB, true);
-  Expr.push_back((uint8_t)dwarf::DW_OP_bregx);
-  Expr.append(buffer, buffer + encodeULEB128(DwarfVlenb, buffer));
-  Expr.push_back(0);
+  return MCCFIInstruction::createEscape(nullptr, DefCfaExpr.str(), SMLoc(),
+                                        Comment.str());
+}
 
-  Expr.push_back((uint8_t)dwarf::DW_OP_mul);
-  Expr.push_back((uint8_t)dwarf::DW_OP_plus);
+static MCCFIInstruction createDefCFAOffset(const TargetRegisterInfo &TRI,
+                                           Register Reg, uint64_t FixedOffset,
+                                           uint64_t ScalableOffset) {
+  assert(ScalableOffset != 0 && "Did not need to adjust CFA for RVV");
+  SmallString<64> Expr;
+  std::string CommentBuffer;
+  llvm::raw_string_ostream Comment(CommentBuffer);
+  Comment << printReg(Reg, &TRI) << "  @ cfa";
 
-  Comment << " + " << ScalableOffset << " * vlenb";
+  // Build up the expression (FixedOffset + ScalableOffset * VLENB).
+  appendScalableVectorExpression(TRI, Expr, FixedOffset, ScalableOffset,
+                                 Comment);
 
   SmallString<64> DefCfaExpr;
-  DefCfaExpr.push_back(dwarf::DW_CFA_def_cfa_expression);
-  DefCfaExpr.append(buffer, buffer + encodeULEB128(Expr.size(), buffer));
+  uint8_t Buffer[16];
+  unsigned DwarfReg = TRI.getDwarfRegNum(Reg, true);
+  DefCfaExpr.push_back(dwarf::DW_CFA_expression);
+  DefCfaExpr.append(Buffer, Buffer + encodeULEB128(DwarfReg, Buffer));
+  DefCfaExpr.append(Buffer, Buffer + encodeULEB128(Expr.size(), Buffer));
   DefCfaExpr.append(Expr.str());
 
   return MCCFIInstruction::createEscape(nullptr, DefCfaExpr.str(), SMLoc(),
@@ -671,6 +706,9 @@ void RISCVFrameLowering::emitPrologue(MachineFunction &MF,
           .addCFIIndex(CFIIndex)
           .setMIFlag(MachineInstr::FrameSetup);
     }
+
+    std::advance(MBBI, getRVVCalleeSavedInfo(MF, CSI).size());
+    emitCalleeSavedRVVPrologCFI(MBB, MBBI, hasFP(MF));
   }
 
   if (hasFP(MF)) {
@@ -1492,6 +1530,41 @@ bool RISCVFrameLowering::spillCalleeSavedRegisters(
   return true;
 }
 
+void RISCVFrameLowering::emitCalleeSavedRVVPrologCFI(
+    MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, bool HasFP) const {
+  MachineFunction *MF = MBB.getParent();
+  const MachineFrameInfo &MFI = MF->getFrameInfo();
+  RISCVMachineFunctionInfo *RVFI = MF->getInfo<RISCVMachineFunctionInfo>();
+  const TargetInstrInfo &TII = *STI.getInstrInfo();
+  DebugLoc DL = MBB.findDebugLoc(MI);
+
+  const auto &RVVCSI = getRVVCalleeSavedInfo(*MF, MFI.getCalleeSavedInfo());
+  if (RVVCSI.empty())
+    return;
+
+  uint64_t FixedSize = getStackSizeWithRVVPadding(*MF);
+  if (!HasFP) {
+    uint64_t ScalarLocalVarSize =
+        MFI.getStackSize() - RVFI->getCalleeSavedStackSize() -
+        RVFI->getRVPushStackSize() - RVFI->getVarArgsSaveSize() +
+        RVFI->getRVVPadding();
+    FixedSize -= ScalarLocalVarSize;
+  }
+
+  for (auto &CS : RVVCSI) {
+    // Insert the spill to the stack frame.
+    int FI = CS.getFrameIdx();
+    if (FI >= 0 && MFI.getStackID(FI) == TargetStackID::ScalableVector) {
+      unsigned CFIIndex = MF->addFrameInst(
+          createDefCFAOffset(*STI.getRegisterInfo(), CS.getReg(), -FixedSize,
+                             MFI.getObjectOffset(FI) / 8));
+      BuildMI(MBB, MI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+          .addCFIIndex(CFIIndex)
+          .setMIFlag(MachineInstr::FrameSetup);
+    }
+  }
+}
+
 bool RISCVFrameLowering::restoreCalleeSavedRegisters(
     MachineBasicBlock &MBB, MachineBasicBlock::iterator MI,
     MutableArrayRef<CalleeSavedInfo> CSI, const TargetRegisterInfo *TRI) const {
diff --git a/llvm/lib/Target/RISCV/RISCVFrameLowering.h b/llvm/lib/Target/RISCV/RISCVFrameLowering.h
index 210f8c1064724..28ab4aff3b9d5 100644
--- a/llvm/lib/Target/RISCV/RISCVFrameLowering.h
+++ b/llvm/lib/Target/RISCV/RISCVFrameLowering.h
@@ -88,6 +88,9 @@ class RISCVFrameLowering : public TargetFrameLowering {
   void adjustStackForRVV(MachineFunction &MF, MachineBasicBlock &MBB,
                          MachineBasicBlock::iterator MBBI, const DebugLoc &DL,
                          int64_t Amount, MachineInstr::MIFlag Flag) const;
+  void emitCalleeSavedRVVPrologCFI(MachineBasicBlock &MBB,
+                                   MachineBasicBlock::iterator MI,
+                                   bool HasFP) const;
   std::pair<int64_t, Align>
   assignRVVStackObjectOffsets(MachineFunction &MF) const;
 };
diff --git a/llvm/test/CodeGen/RISCV/rvv-cfi-info.ll b/llvm/test/CodeGen/RISCV/rvv-cfi-info.ll
new file mode 100644
index 0000000000000..c99388cbdaf44
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv-cfi-info.ll
@@ -0,0 +1,111 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+v,+m -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=OMIT-FP %s
+; RUN: llc -mtriple=riscv64 -mattr=+v,+m -verify-machineinstrs -frame-pointer=all < %s \
+; RUN:   | FileCheck -check-prefix=NO-OMIT-FP %s
+
+define riscv_vector_cc <vscale x 1 x i32> @test_vector_callee_cfi(<vscale x 1 x i32> %va) {
+; OMIT-FP-LABEL: test_vector_callee_cfi:
+; OMIT-FP:       # %bb.0: # %entry
+; OMIT-FP-NEXT:    addi sp, sp, -16
+; OMIT-FP-NEXT:    .cfi_def_cfa_offset 16
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    slli a0, a0, 3
+; OMIT-FP-NEXT:    sub sp, sp, a0
+; OMIT-FP-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    li a1, 6
+; OMIT-FP-NEXT:    mul a0, a0, a1
+; OMIT-FP-NEXT:    add a0, sp, a0
+; OMIT-FP-NEXT:    addi a0, a0, 16
+; OMIT-FP-NEXT:    vs1r.v v1, (a0) # Unknown-size Folded Spill
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    slli a0, a0, 2
+; OMIT-FP-NEXT:    add a0, sp, a0
+; OMIT-FP-NEXT:    addi a0, a0, 16
+; OMIT-FP-NEXT:    vs2r.v v2, (a0) # Unknown-size Folded Spill
+; OMIT-FP-NEXT:    addi a0, sp, 16
+; OMIT-FP-NEXT:    vs4r.v v4, (a0) # Unknown-size Folded Spill
+; OMIT-FP-NEXT:    .cfi_escape 0x10, 0x61, 0x08, 0x11, 0x7e, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v1 @ cfa - 2 * vlenb
+; OMIT-FP-NEXT:    .cfi_escape 0x10, 0x62, 0x08, 0x11, 0x7c, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v2m2 @ cfa - 4 * vlenb
+; OMIT-FP-NEXT:    .cfi_escape 0x10, 0x64, 0x08, 0x11, 0x78, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v4m4 @ cfa - 8 * vlenb
+; OMIT-FP-NEXT:    #APP
+; OMIT-FP-NEXT:    #NO_APP
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    li a1, 6
+; OMIT-FP-NEXT:    mul a0, a0, a1
+; OMIT-FP-NEXT:    add a0, sp, a0
+; OMIT-FP-NEXT:    addi a0, a0, 16
+; OMIT-FP-NEXT:    vl1r.v v1, (a0) # Unknown-size Folded Reload
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    slli a0, a0, 2
+; OMIT-FP-NEXT:    add a0, sp, a0
+; OMIT-FP-NEXT:    addi a0, a0, 16
+; OMIT-FP-NEXT:    vl2r.v v2, (a0) # Unknown-size Folded Reload
+; OMIT-FP-NEXT:    addi a0, sp, 16
+; OMIT-FP-NEXT:    vl4r.v v4, (a0) # Unknown-size Folded Reload
+; OMIT-FP-NEXT:    csrr a0, vlenb
+; OMIT-FP-NEXT:    slli a0, a0, 3
+; OMIT-FP-NEXT:    add sp, sp, a0
+; OMIT-FP-NEXT:    addi sp, sp, 16
+; OMIT-FP-NEXT:    ret
+;
+; NO-OMIT-FP-LABEL: test_vector_callee_cfi:
+; NO-OMIT-FP:       # %bb.0: # %entry
+; NO-OMIT-FP-NEXT:    addi sp, sp, -32
+; NO-OMIT-FP-NEXT:    .cfi_def_cfa_offset 32
+; NO-OMIT-FP-NEXT:    sd ra, 24(sp) # 8-byte Folded Spill
+; NO-OMIT-FP-NEXT:    sd s0, 16(sp) # 8-byte Folded Spill
+; NO-OMIT-FP-NEXT:    .cfi_offset ra, -8
+; NO-OMIT-FP-NEXT:    .cfi_offset s0, -16
+; NO-OMIT-FP-NEXT:    addi s0, sp, 32
+; NO-OMIT-FP-NEXT:    .cfi_def_cfa s0, 0
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 3
+; NO-OMIT-FP-NEXT:    sub sp, sp, a0
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 1
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vs1r.v v1, (a0) # Unknown-size Folded Spill
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 2
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vs2r.v v2, (a0) # Unknown-size Folded Spill
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 3
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vs4r.v v4, (a0) # Unknown-size Folded Spill
+; NO-OMIT-FP-NEXT:    .cfi_escape 0x10, 0x61, 0x0b, 0x11, 0x60, 0x22, 0x11, 0x7e, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v1 @ cfa - 32 - 2 * vlenb
+; NO-OMIT-FP-NEXT:    .cfi_escape 0x10, 0x62, 0x0b, 0x11, 0x60, 0x22, 0x11, 0x7c, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v2m2 @ cfa - 32 - 4 * vlenb
+; NO-OMIT-FP-NEXT:    .cfi_escape 0x10, 0x64, 0x0b, 0x11, 0x60, 0x22, 0x11, 0x78, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # $v4m4 @ cfa - 32 - 8 * vlenb
+; NO-OMIT-FP-NEXT:    #APP
+; NO-OMIT-FP-NEXT:    #NO_APP
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 1
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vl1r.v v1, (a0) # Unknown-size Folded Reload
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 2
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vl2r.v v2, (a0) # Unknown-size Folded Reload
+; NO-OMIT-FP-NEXT:    csrr a0, vlenb
+; NO-OMIT-FP-NEXT:    slli a0, a0, 3
+; NO-OMIT-FP-NEXT:    sub a0, s0, a0
+; NO-OMIT-FP-NEXT:    addi a0, a0, -32
+; NO-OMIT-FP-NEXT:    vl4r.v v4, (a0) # Unknown-size Folded Reload
+; NO-OMIT-FP-NEXT:    addi sp, s0, -32
+; NO-OMIT-FP-NEXT:    ld ra, 24(sp) # 8-byte Folded Reload
+; NO-OMIT-FP-NEXT:    ld s0, 16(sp) # 8-byte Folded Reload
+; NO-OMIT-FP-NEXT:    addi sp, sp, 32
+; NO-OMIT-FP-NEXT:    ret
+entry:
+  call void asm sideeffect "",
+  "~{v1},~{v2},~{v3},~{v4},~{v5},~{v6},~{v7}"()
+
+  ret <vscale x 1 x i32> %va
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
index 1d3c22a02efc0..ab6df1d3e883f 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-subvector.ll
@@ -628,6 +628,7 @@ define void @insert_v2i64_nxv16i64_hi(ptr %psv, ptr %out) {
 ; RV32-NEXT:    vs8r.v v8, (a0)
 ; RV32-NEXT:    vs8r.v v16, (a1)
 ; RV32-NEXT:    addi sp, s0, -80
+; RV32-NEXT:    .cfi_def_cfa sp, 80
 ; RV32-NEXT:    lw ra, 76(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    lw s0, 72(sp) # 4-byte Folded Reload
 ; RV32-NEXT:    addi sp, sp, 80
@@ -661,6 +662,7 @@ define void @insert_v2i64_nxv16i64_hi(ptr %psv, ptr %out) {
 ; RV64-NEXT:    vs8r.v v8, (a0)
 ; RV64-NEXT:    vs8r.v v16, (a1)
 ; RV64-NEXT:    addi sp, s0, -80
+; RV64-NEXT:    .cfi_def_cfa sp, 80
 ; RV64-NEXT:    ld ra, 72(sp) # 8-byte Folded Reload
 ; RV64-NEXT:    ld s0, 64(sp) # 8-byte Folded Reload
 ; RV64-NEXT:    addi sp, sp, 80

From e6ecff8d95b9175e70e0d43e14c2975c8f69d718 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Wed, 17 Apr 2024 10:40:09 +0800
Subject: [PATCH 688/886] [C++20] [Modules] Add Release Notes and Documents for
 Reduced BMI

See
https://discourse.llvm.org/t/rfc-c-20-modules-introduce-thin-bmi-and-decls-hash/74755,
https://github.com/llvm/llvm-project/pull/75894 and
https://github.com/llvm/llvm-project/pull/85050 for the background.
---
 clang/docs/ReleaseNotes.rst             |   3 +
 clang/docs/StandardCPlusPlusModules.rst | 106 ++++++++++++++++++++++++
 2 files changed, 109 insertions(+)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 3752b6ce15760..efc32212f300c 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -214,6 +214,9 @@ New Compiler Flags
   This diagnostic can be disabled to make ``-Wmissing-field-initializers`` behave
   like it did before Clang 18.x. Fixes #GH56628
 
+- ``-fexperimental-modules-reduced-bmi`` enables the Reduced BMI for C++20 named modules.
+  See the document of standard C++ modules for details.
+
 Deprecated Compiler Flags
 -------------------------
 
diff --git a/clang/docs/StandardCPlusPlusModules.rst b/clang/docs/StandardCPlusPlusModules.rst
index c5478bba45f38..8d5529d5d37db 100644
--- a/clang/docs/StandardCPlusPlusModules.rst
+++ b/clang/docs/StandardCPlusPlusModules.rst
@@ -520,6 +520,112 @@ is attached to the global module fragments. For example:
 
 Now the linkage name of ``NS::foo()`` will be ``_ZN2NS3fooEv``.
 
+Reduced BMI
+-----------
+
+To support the 2 phase compilation model, Clang chose to put everything needed to
+produce an object into the BMI. But every consumer of the BMI, except itself, doesn't
+need such informations. It makes the BMI to larger and so may introduce unnecessary
+dependencies into the BMI. To mitigate the problem, we decided to reduce the information
+contained in the BMI.
+
+To be clear, we call the default BMI as Full BMI and the new introduced BMI as Reduced
+BMI.
+
+Users can use ``-fexperimental-modules-reduced-bmi`` flag to enable the Reduced BMI.
+
+For one phase compilation model (CMake implements this model), with
+``-fexperimental-modules-reduced-bmi``, the generated BMI will be Reduced BMI automatically.
+(The output path of the BMI is specified by ``-fmodule-output=`` as usual one phase
+compilation model).
+
+It is still possible to support Reduced BMI in two phase compilation model. With
+``-fexperimental-modules-reduced-bmi``, ``--precompile`` and ``-fmodule-output=`` specified,
+the generated BMI specified by ``-o`` will be full BMI and the BMI specified by
+``-fmodule-output=`` will be Reduced BMI. The dependency graph may be:
+
+.. code-block:: none
+
+  module-unit.cppm --> module-unit.full.pcm -> module-unit.o
+                    |
+                    -> module-unit.reduced.pcm -> consumer1.cpp
+                                               -> consumer2.cpp
+                                               -> ...
+                                               -> consumer_n.cpp
+
+We don't emit diagnostics if ``-fexperimental-modules-reduced-bmi`` is used with a non-module
+unit. This design helps the end users of one phase compilation model to perform experiments
+early without asking for the help of build systems. The users of build systems which supports
+two phase compilation model still need helps from build systems.
+
+Within Reduced BMI, we won't write unreachable entities from GMF, definitions of non-inline
+functions and non-inline variables. This may not be a transparent change.
+`[module.global.frag]ex2 <https://eel.is/c++draft/module.global.frag#example-2>`_ may be a good
+example:
+
+.. code-block:: c++
+
+  // foo.h
+  namespace N {
+    struct X {};
+    int d();
+    int e();
+    inline int f(X, int = d()) { return e(); }
+    int g(X);
+    int h(X);
+  }
+
+  // M.cppm
+  module;
+  #include "foo.h"
+  export module M;
+  template<typename T> int use_f() {
+    N::X x;                       // N::X, N, and :: are decl-reachable from use_f
+    return f(x, 123);             // N::f is decl-reachable from use_f,
+                                  // N::e is indirectly decl-reachable from use_f
+                                  //   because it is decl-reachable from N::f, and
+                                  // N::d is decl-reachable from use_f
+                                  //   because it is decl-reachable from N::f
+                                  //   even though it is not used in this call
+  }
+  template<typename T> int use_g() {
+    N::X x;                       // N::X, N, and :: are decl-reachable from use_g
+    return g((T(), x));           // N::g is not decl-reachable from use_g
+  }
+  template<typename T> int use_h() {
+    N::X x;                       // N::X, N, and :: are decl-reachable from use_h
+    return h((T(), x));           // N::h is not decl-reachable from use_h, but
+                                  // N::h is decl-reachable from use_h<int>
+  }
+  int k = use_h<int>();
+    // use_h<int> is decl-reachable from k, so
+    // N::h is decl-reachable from k
+
+  // M-impl.cpp
+  module M;
+  int a = use_f<int>();           // OK
+  int b = use_g<int>();           // error: no viable function for call to g;
+                                  // g is not decl-reachable from purview of
+                                  // module M's interface, so is discarded
+  int c = use_h<int>();           // OK
+
+In the above example, the function definition of ``N::g`` is elided from the Reduced
+BMI of ``M.cppm``. Then the use of ``use_g<int>`` in ``M-impl.cpp`` fails
+to instantiate. For such issues, users can add references to ``N::g`` in the module purview
+of ``M.cppm`` to make sure it is reachable, e.g., ``using N::g;``.
+
+We think the Reduced BMI is the correct direction. But given it is a drastic change,
+we'd like to make it experimental first to avoid breaking existing users. The roadmap
+of Reduced BMI may be:
+
+1. ``-fexperimental-modules-reduced-bmi`` is opt in for 1~2 releases. The period depends
+on testing feedbacks.
+2. We would announce Reduced BMI is not experimental and introduce ``-fmodules-reduced-bmi``.
+and suggest users to enable this mode. This may takes 1~2 releases too.
+3. Finally we will enable this by default. When that time comes, the term BMI will refer to
+the reduced BMI today and the Full BMI will only be meaningful to build systems which
+loves to support two phase compilations.
+
 Performance Tips
 ----------------
 

From eafd515ecaaa100623eebc7fa4d7c36a361bf708 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Tue, 16 Apr 2024 19:47:52 -0700
Subject: [PATCH 689/886] [clang][deps] Support single-file mode for all
 formats (#88764)

The `clang-scan-deps` tool can be used for fast scanning of batches of
compilation commands passed in via the `-compilation-database` option.
This gets awkward in our tests where we have to resort to using
`.in`/`.template` JSON files and running them through `sed` in order to
embed LIT's `%t` variable into them. However, most of our tests only
need to pass single compilation command, so this dance is entirely
unnecessary.

This patch makes sure the existing "per-file" mode (where the
compilation command is passed in-line after the `--` argument) works for
all output formats, not only `P1689`.
---
 clang/test/ClangScanDeps/error.cpp            | 18 +-----
 clang/tools/clang-scan-deps/ClangScanDeps.cpp | 57 +++++++------------
 2 files changed, 23 insertions(+), 52 deletions(-)

diff --git a/clang/test/ClangScanDeps/error.cpp b/clang/test/ClangScanDeps/error.cpp
index 0095a6c900c3b..593dbf35edca5 100644
--- a/clang/test/ClangScanDeps/error.cpp
+++ b/clang/test/ClangScanDeps/error.cpp
@@ -1,23 +1,10 @@
 // RUN: rm -rf %t
 // RUN: split-file %s %t
 
-//--- missing_tu.json.in
-[{
-  "directory": "DIR",
-  "command": "clang -fsyntax-only DIR/missing_tu.c",
-  "file": "DIR/missing_tu.c"
-}]
-//--- missing_header.json.in
-[{
-  "directory": "DIR",
-  "command": "clang -fsyntax-only DIR/missing_header.c",
-  "file": "DIR/missing_header.c"
-}]
 //--- missing_header.c
 #include "missing.h"
 
-// RUN: sed -e "s|DIR|%/t|g" %t/missing_tu.json.in > %t/missing_tu.json
-// RUN: not clang-scan-deps -compilation-database %t/missing_tu.json 2>%t/missing_tu.errs
+// RUN: not clang-scan-deps -- %clang -c %t/missing_tu.c 2>%t/missing_tu.errs
 // RUN: echo EOF >> %t/missing_tu.errs
 // RUN: cat %t/missing_tu.errs | sed 's:\\\\\?:/:g' | FileCheck %s --check-prefix=CHECK-MISSING-TU -DPREFIX=%/t
 // CHECK-MISSING-TU: Error while scanning dependencies for [[PREFIX]]/missing_tu.c
@@ -26,8 +13,7 @@
 // CHECK-MISSING-TU-NEXT: error:
 // CHECK-MISSING-TU-NEXT: EOF
 
-// RUN: sed -e "s|DIR|%/t|g" %t/missing_header.json.in > %t/missing_header.json
-// RUN: not clang-scan-deps -compilation-database %t/missing_header.json 2>%t/missing_header.errs
+// RUN: not clang-scan-deps -- %clang -c %t/missing_header.c 2>%t/missing_header.errs
 // RUN: echo EOF >> %t/missing_header.errs
 // RUN: cat %t/missing_header.errs | sed 's:\\\\\?:/:g' | FileCheck %s --check-prefix=CHECK-MISSING-HEADER -DPREFIX=%/t
 // CHECK-MISSING-HEADER: Error while scanning dependencies for [[PREFIX]]/missing_header.c
diff --git a/clang/tools/clang-scan-deps/ClangScanDeps.cpp b/clang/tools/clang-scan-deps/ClangScanDeps.cpp
index eaa76dd43e41d..94510515cd440 100644
--- a/clang/tools/clang-scan-deps/ClangScanDeps.cpp
+++ b/clang/tools/clang-scan-deps/ClangScanDeps.cpp
@@ -98,8 +98,8 @@ static bool RoundTripArgs = DoRoundTripDefault;
 static void ParseArgs(int argc, char **argv) {
   ScanDepsOptTable Tbl;
   llvm::StringRef ToolName = argv[0];
-  llvm::BumpPtrAllocator A;
-  llvm::StringSaver Saver{A};
+  llvm::BumpPtrAllocator Alloc;
+  llvm::StringSaver Saver{Alloc};
   llvm::opt::InputArgList Args =
       Tbl.parseArgs(argc, argv, OPT_UNKNOWN, Saver, [&](StringRef Msg) {
         llvm::errs() << Msg << '\n';
@@ -186,14 +186,8 @@ static void ParseArgs(int argc, char **argv) {
     }
   }
 
-  if (const llvm::opt::Arg *A = Args.getLastArg(OPT_compilation_database_EQ)) {
+  if (const llvm::opt::Arg *A = Args.getLastArg(OPT_compilation_database_EQ))
     CompilationDB = A->getValue();
-  } else if (Format != ScanningOutputFormat::P1689) {
-    llvm::errs() << ToolName
-                 << ": for the --compiilation-database option: must be "
-                    "specified at least once!";
-    std::exit(1);
-  }
 
   if (const llvm::opt::Arg *A = Args.getLastArg(OPT_module_name_EQ))
     ModuleName = A->getValue();
@@ -225,9 +219,8 @@ static void ParseArgs(int argc, char **argv) {
 
   RoundTripArgs = Args.hasArg(OPT_round_trip_args);
 
-  if (auto *A = Args.getLastArgNoClaim(OPT_DASH_DASH))
-    CommandLine.insert(CommandLine.end(), A->getValues().begin(),
-                       A->getValues().end());
+  if (const llvm::opt::Arg *A = Args.getLastArgNoClaim(OPT_DASH_DASH))
+    CommandLine.assign(A->getValues().begin(), A->getValues().end());
 }
 
 class SharedStream {
@@ -694,38 +687,28 @@ static std::string getModuleCachePath(ArrayRef<std::string> Args) {
   return std::string(Path);
 }
 
-// getCompilationDataBase - If -compilation-database is set, load the
-// compilation database from the specified file. Otherwise if the we're
-// generating P1689 format, trying to generate the compilation database
-// form specified command line after the positional parameter "--".
+/// Attempts to construct the compilation database from '-compilation-database'
+/// or from the arguments following the positional '--'.
 static std::unique_ptr<tooling::CompilationDatabase>
-getCompilationDataBase(int argc, char **argv, std::string &ErrorMessage) {
+getCompilationDatabase(int argc, char **argv, std::string &ErrorMessage) {
   ParseArgs(argc, argv);
 
+  if (!(CommandLine.empty() ^ CompilationDB.empty())) {
+    llvm::errs() << "The compilation command line must be provided either via "
+                    "'-compilation-database' or after '--'.";
+    return nullptr;
+  }
+
   if (!CompilationDB.empty())
     return tooling::JSONCompilationDatabase::loadFromFile(
         CompilationDB, ErrorMessage,
         tooling::JSONCommandLineSyntax::AutoDetect);
 
-  if (Format != ScanningOutputFormat::P1689) {
-    llvm::errs() << "the --compilation-database option: must be specified at "
-                    "least once!";
-    return nullptr;
-  }
-
-  // Trying to get the input file, the output file and the command line options
-  // from the positional parameter "--".
-  char **DoubleDash = std::find(argv, argv + argc, StringRef("--"));
-  if (DoubleDash == argv + argc) {
-    llvm::errs() << "The command line arguments is required after '--' in "
-                    "P1689 per file mode.";
-    return nullptr;
-  }
-
   llvm::IntrusiveRefCntPtr<DiagnosticsEngine> Diags =
       CompilerInstance::createDiagnostics(new DiagnosticOptions);
   driver::Driver TheDriver(CommandLine[0], llvm::sys::getDefaultTargetTriple(),
                            *Diags);
+  TheDriver.setCheckInputsExist(false);
   std::unique_ptr<driver::Compilation> C(
       TheDriver.BuildCompilation(CommandLine));
   if (!C || C->getJobs().empty())
@@ -740,7 +723,8 @@ getCompilationDataBase(int argc, char **argv, std::string &ErrorMessage) {
 
   FrontendOptions &FEOpts = CI->getFrontendOpts();
   if (FEOpts.Inputs.size() != 1) {
-    llvm::errs() << "Only one input file is allowed in P1689 per file mode.";
+    llvm::errs()
+        << "Exactly one input file is required in the per-file mode ('--').\n";
     return nullptr;
   }
 
@@ -749,8 +733,9 @@ getCompilationDataBase(int argc, char **argv, std::string &ErrorMessage) {
   auto LastCmd = C->getJobs().end();
   LastCmd--;
   if (LastCmd->getOutputFilenames().size() != 1) {
-    llvm::errs() << "The command line should provide exactly one output file "
-                    "in P1689 per file mode.\n";
+    llvm::errs()
+        << "Exactly one output file is required in the per-file mode ('--').\n";
+    return nullptr;
   }
   StringRef OutputFile = LastCmd->getOutputFilenames().front();
 
@@ -790,7 +775,7 @@ getCompilationDataBase(int argc, char **argv, std::string &ErrorMessage) {
 int clang_scan_deps_main(int argc, char **argv, const llvm::ToolContext &) {
   std::string ErrorMessage;
   std::unique_ptr<tooling::CompilationDatabase> Compilations =
-      getCompilationDataBase(argc, argv, ErrorMessage);
+      getCompilationDatabase(argc, argv, ErrorMessage);
   if (!Compilations) {
     llvm::errs() << ErrorMessage << "\n";
     return 1;

From 6a4eaf9b33d8091b7d09b2a30a3fc8993a01db31 Mon Sep 17 00:00:00 2001
From: Jan Svoboda <jan_svoboda@apple.com>
Date: Tue, 16 Apr 2024 19:49:07 -0700
Subject: [PATCH 690/886] [clang][deps] Add `-o` flag to specify output path
 (#88767)

This makes it possible to pass "-o /dev/null" to `clang-scan-deps` and
skip some potentially expensive work, making timings less noisy. Also
removes the need for stream redirection.
---
 clang/test/ClangScanDeps/module-format.c      |  2 +-
 clang/tools/clang-scan-deps/ClangScanDeps.cpp | 34 ++++++++++++++++---
 clang/tools/clang-scan-deps/Opts.td           |  4 ++-
 3 files changed, 34 insertions(+), 6 deletions(-)

diff --git a/clang/test/ClangScanDeps/module-format.c b/clang/test/ClangScanDeps/module-format.c
index 001a011ae0b59..0a6abec80dd90 100644
--- a/clang/test/ClangScanDeps/module-format.c
+++ b/clang/test/ClangScanDeps/module-format.c
@@ -16,7 +16,7 @@
 // RUN: rm -f %t/cdb_pch.json
 // RUN: sed "s|DIR|%/t|g" %S/Inputs/modules-pch/cdb_pch.json > %t/cdb_pch.json
 // RUN: clang-scan-deps -compilation-database %t/cdb_pch.json -format experimental-full \
-// RUN:   -module-files-dir %t/build > %t/result_pch.json
+// RUN:   -module-files-dir %t/build -o %t/result_pch.json
 
 // Explicitly build the PCH:
 //
diff --git a/clang/tools/clang-scan-deps/ClangScanDeps.cpp b/clang/tools/clang-scan-deps/ClangScanDeps.cpp
index 94510515cd440..f42af7e330e17 100644
--- a/clang/tools/clang-scan-deps/ClangScanDeps.cpp
+++ b/clang/tools/clang-scan-deps/ClangScanDeps.cpp
@@ -72,6 +72,7 @@ enum ResourceDirRecipeKind {
   RDRK_InvokeCompiler,
 };
 
+static std::string OutputFileName = "-";
 static ScanningMode ScanMode = ScanningMode::DependencyDirectivesScan;
 static ScanningOutputFormat Format = ScanningOutputFormat::Make;
 static ScanningOptimizations OptimizeArgs;
@@ -175,6 +176,9 @@ static void ParseArgs(int argc, char **argv) {
   if (const llvm::opt::Arg *A = Args.getLastArg(OPT_module_files_dir_EQ))
     ModuleFilesDir = A->getValue();
 
+  if (const llvm::opt::Arg *A = Args.getLastArg(OPT_o))
+    OutputFileName = A->getValue();
+
   EagerLoadModules = Args.hasArg(OPT_eager_load_pcm);
 
   if (const llvm::opt::Arg *A = Args.getLastArg(OPT_j)) {
@@ -419,6 +423,11 @@ class FullDeps {
   }
 
   void printFullOutput(raw_ostream &OS) {
+    // Skip sorting modules and constructing the JSON object if the output
+    // cannot be observed anyway. This makes timings less noisy.
+    if (&OS == &llvm::nulls())
+      return;
+
     // Sort the modules by name to get a deterministic order.
     std::vector<IndexedModuleID> ModuleIDs;
     for (auto &&M : Modules)
@@ -849,8 +858,25 @@ int clang_scan_deps_main(int argc, char **argv, const llvm::ToolContext &) {
       });
 
   SharedStream Errs(llvm::errs());
-  // Print out the dependency results to STDOUT by default.
-  SharedStream DependencyOS(llvm::outs());
+
+  std::optional<llvm::raw_fd_ostream> FileOS;
+  llvm::raw_ostream &ThreadUnsafeDependencyOS = [&]() -> llvm::raw_ostream & {
+    if (OutputFileName == "-")
+      return llvm::outs();
+
+    if (OutputFileName == "/dev/null")
+      return llvm::nulls();
+
+    std::error_code EC;
+    FileOS.emplace(OutputFileName, EC);
+    if (EC) {
+      llvm::errs() << "Failed to open output file '" << OutputFileName
+                   << "': " << llvm::errorCodeToError(EC) << '\n';
+      std::exit(1);
+    }
+    return *FileOS;
+  }();
+  SharedStream DependencyOS(ThreadUnsafeDependencyOS);
 
   std::vector<tooling::CompileCommand> Inputs =
       AdjustingCompilations->getAllCompileCommands();
@@ -991,9 +1017,9 @@ int clang_scan_deps_main(int argc, char **argv, const llvm::ToolContext &) {
       HadErrors = true;
 
   if (Format == ScanningOutputFormat::Full)
-    FD->printFullOutput(llvm::outs());
+    FD->printFullOutput(ThreadUnsafeDependencyOS);
   else if (Format == ScanningOutputFormat::P1689)
-    PD.printDependencies(llvm::outs());
+    PD.printDependencies(ThreadUnsafeDependencyOS);
 
   return HadErrors;
 }
diff --git a/clang/tools/clang-scan-deps/Opts.td b/clang/tools/clang-scan-deps/Opts.td
index 5cd5d1a9fb37b..4837ce6f070d7 100644
--- a/clang/tools/clang-scan-deps/Opts.td
+++ b/clang/tools/clang-scan-deps/Opts.td
@@ -11,6 +11,8 @@ multiclass Eq<string name, string help> {
 def help : Flag<["--"], "help">, HelpText<"Display this help">;
 def version : Flag<["--"], "version">, HelpText<"Display the version">;
 
+def o : Arg<"o", "Destination of the primary output">;
+
 defm mode : Eq<"mode", "The preprocessing mode used to compute the dependencies">;
 
 defm format : Eq<"format", "The output format for the dependencies">;
@@ -37,4 +39,4 @@ def verbose : F<"v", "Use verbose output">;
 
 def round_trip_args : F<"round-trip-args", "verify that command-line arguments are canonical by parsing and re-serializing">;
 
-def DASH_DASH : Option<["--"], "", KIND_REMAINING_ARGS>;
\ No newline at end of file
+def DASH_DASH : Option<["--"], "", KIND_REMAINING_ARGS>;

From f71e25bb669d662f98823d6d81b3f918538c9239 Mon Sep 17 00:00:00 2001
From: Kazu Hirata <kazu@google.com>
Date: Tue, 16 Apr 2024 20:35:34 -0700
Subject: [PATCH 691/886] [memprof] Simplify IndexedMemProfRecord::operator==
 (NFC) (#88986)

llvm::SmallVector::operator== exactly meets our needs.
---
 llvm/include/llvm/ProfileData/MemProf.h | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/llvm/include/llvm/ProfileData/MemProf.h b/llvm/include/llvm/ProfileData/MemProf.h
index d43fb1c93bb8e..7f3956bd73939 100644
--- a/llvm/include/llvm/ProfileData/MemProf.h
+++ b/llvm/include/llvm/ProfileData/MemProf.h
@@ -370,14 +370,9 @@ struct IndexedMemProfRecord {
   size_t serializedSize(IndexedVersion Version) const;
 
   bool operator==(const IndexedMemProfRecord &Other) const {
-    if (Other.AllocSites.size() != AllocSites.size())
+    if (Other.AllocSites != AllocSites)
       return false;
 
-    for (size_t I = 0; I < AllocSites.size(); I++) {
-      if (AllocSites[I] != Other.AllocSites[I])
-        return false;
-    }
-
     if (Other.CallSiteIds != CallSiteIds)
       return false;
     return true;

From fca2a493251597967d5d758ea0748c66dd29371a Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Tue, 16 Apr 2024 21:46:57 -0700
Subject: [PATCH 692/886] [RISCV] Simplify FindRegWithEncoding in
 copyPhysRegVector. NFC (#89001)

Instead of searching all encodings, we can convert the encoding back to
a register and use getMatchingSuperReg.
---
 llvm/lib/Target/RISCV/RISCVInstrInfo.cpp | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 14b5cbea71722..8331fc0b8c302 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -361,15 +361,12 @@ void RISCVInstrInfo::copyPhysRegVector(
     return {RISCVII::LMUL_1, RISCV::VRRegClass, RISCV::VMV1R_V,
             RISCV::PseudoVMV_V_V_M1, RISCV::PseudoVMV_V_I_M1};
   };
-  auto FindRegWithEncoding = [&TRI](const TargetRegisterClass &RegClass,
-                                    uint16_t Encoding) {
-    ArrayRef<MCPhysReg> Regs = RegClass.getRegisters();
-    const auto *FoundReg = llvm::find_if(Regs, [&](MCPhysReg Reg) {
-      return TRI->getEncodingValue(Reg) == Encoding;
-    });
-    // We should be always able to find one valid register.
-    assert(FoundReg != Regs.end());
-    return *FoundReg;
+  auto FindRegWithEncoding = [TRI](const TargetRegisterClass &RegClass,
+                                   uint16_t Encoding) {
+    MCRegister Reg = RISCV::V0 + Encoding;
+    if (&RegClass == &RISCV::VRRegClass)
+      return Reg;
+    return TRI->getMatchingSuperReg(Reg, RISCV::sub_vrm1_0, &RegClass);
   };
   while (I != NumRegs) {
     // For non-segment copying, we only do this once as the registers are always

From a6fcbcce8f79adfb2e4338859f3a41fc2538bad1 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Wed, 17 Apr 2024 07:59:43 +0200
Subject: [PATCH 693/886] [libc++][TZDB] Improves time zone format specifiers.
 (#85797)

Per [tab:time.format.spec]
%z  The offset from UTC as specified in ISO 8601-1:2019, subclause
    5.3.4.1. For example -0430 refers to 4 hours 30 minutes behind UTC.
    If the offset is zero, +0000 is used. The modified commands %Ez and
    %Oz insert a : between the hours and minutes: -04:30. If the offset
    information is not available, an exception of type format_error is
    thrown.

Typically the modified versions Oz or Ez would have wording like

  The modified command %OS produces the locale's alternative
  representation.

In this case the modified version does not depend on the locale.

This change is a preparation for formatting sys_info which has time zone
information. The function time_put<_CharT>::put() does not have proper
time zone support, therefore it's a manual implementation.

Fixes https://github.com/llvm/llvm-project/issues/78184
---
 libcxx/include/__chrono/formatter.h           | 50 ++++++++++++++++++-
 .../time.syn/formatter.file_time.pass.cpp     | 39 ++-------------
 .../time/time.syn/formatter.sys_time.pass.cpp | 39 ++-------------
 3 files changed, 56 insertions(+), 72 deletions(-)

diff --git a/libcxx/include/__chrono/formatter.h b/libcxx/include/__chrono/formatter.h
index b64cae529a294..d932a99f4b998 100644
--- a/libcxx/include/__chrono/formatter.h
+++ b/libcxx/include/__chrono/formatter.h
@@ -10,6 +10,7 @@
 #ifndef _LIBCPP___CHRONO_FORMATTER_H
 #define _LIBCPP___CHRONO_FORMATTER_H
 
+#include <__algorithm/ranges_copy.h>
 #include <__chrono/calendar.h>
 #include <__chrono/concepts.h>
 #include <__chrono/convert_to_tm.h>
@@ -170,10 +171,45 @@ _LIBCPP_HIDE_FROM_ABI void __format_century(basic_stringstream<_CharT>& __sstr,
   __sstr << std::format(_LIBCPP_STATICALLY_WIDEN(_CharT, "{:02}"), __century);
 }
 
+// Implements the %z format specifier according to [tab:time.format.spec], where
+// '__modifier' signals %Oz or %Ez were used. (Both modifiers behave the same,
+// so there is no need to distinguish between them.)
+template <class _CharT>
+_LIBCPP_HIDE_FROM_ABI void
+__format_zone_offset(basic_stringstream<_CharT>& __sstr, chrono::seconds __offset, bool __modifier) {
+  if (__offset < 0s) {
+    __sstr << _CharT('-');
+    __offset = -__offset;
+  } else {
+    __sstr << _CharT('+');
+  }
+
+  chrono::hh_mm_ss __hms{__offset};
+  std::ostreambuf_iterator<_CharT> __out_it{__sstr};
+  if (__modifier)
+    std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:%H:%M}"), __hms);
+  else
+    std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:%H%M}"), __hms);
+}
+
+// Helper to store the time zone information needed for formatting.
+struct _LIBCPP_HIDE_FROM_ABI __time_zone {
+  // Typically these abbreviations are short and fit in the string's internal
+  // buffer.
+  string __abbrev;
+  chrono::seconds __offset;
+};
+
+template <class _Tp>
+_LIBCPP_HIDE_FROM_ABI __time_zone __convert_to_time_zone([[maybe_unused]] const _Tp& __value) {
+  return {"UTC", chrono::seconds{0}};
+}
+
 template <class _CharT, class _Tp>
 _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
     basic_stringstream<_CharT>& __sstr, const _Tp& __value, basic_string_view<_CharT> __chrono_specs) {
   tm __t              = std::__convert_to_tm<tm>(__value);
+  __time_zone __z     = __formatter::__convert_to_time_zone(__value);
   const auto& __facet = std::use_facet<time_put<_CharT>>(__sstr.getloc());
   for (auto __it = __chrono_specs.begin(); __it != __chrono_specs.end(); ++__it) {
     if (*__it == _CharT('%')) {
@@ -296,9 +332,13 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
               {__sstr}, __sstr, _CharT(' '), std::addressof(__t), std::to_address(__s), std::to_address(__it + 1));
       } break;
 
+      case _CharT('z'):
+        __formatter::__format_zone_offset(__sstr, __z.__offset, false);
+        break;
+
       case _CharT('Z'):
-        // TODO FMT Add proper timezone support.
-        __sstr << _LIBCPP_STATICALLY_WIDEN(_CharT, "UTC");
+        // __abbrev is always a char so the copy may convert.
+        ranges::copy(__z.__abbrev, std::ostreambuf_iterator<_CharT>{__sstr});
         break;
 
       case _CharT('O'):
@@ -314,9 +354,15 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
             break;
           }
         }
+
+        // Oz produces the same output as Ez below.
         [[fallthrough]];
       case _CharT('E'):
         ++__it;
+        if (*__it == 'z') {
+          __formatter::__format_zone_offset(__sstr, __z.__offset, true);
+          break;
+        }
         [[fallthrough]];
       default:
         __facet.put(
diff --git a/libcxx/test/std/time/time.syn/formatter.file_time.pass.cpp b/libcxx/test/std/time/time.syn/formatter.file_time.pass.cpp
index b07282593d759..f57841cca8629 100644
--- a/libcxx/test/std/time/time.syn/formatter.file_time.pass.cpp
+++ b/libcxx/test/std/time/time.syn/formatter.file_time.pass.cpp
@@ -904,12 +904,6 @@ static void test_valid_values_date_time() {
 
 template <class CharT>
 static void test_valid_values_time_zone() {
-// The Apple CI gives %z='-0700'	%Ez='-0700'	%Oz='-0700'	%Z='UTC'
-// -0700 looks like the local time where the CI happens to reside, therefore
-// omit this test on Apple.
-// The Windows CI gives %z='-0000', but on local machines set to a different
-// timezone, it gives e.g. %z='+0200'.
-#if !defined(__APPLE__) && !defined(_WIN32)
   using namespace std::literals::chrono_literals;
 
   constexpr std::basic_string_view<CharT> fmt  = SV("{:%%z='%z'%t%%Ez='%Ez'%t%%Oz='%Oz'%t%%Z='%Z'%n}");
@@ -918,48 +912,23 @@ static void test_valid_values_time_zone() {
   const std::locale loc(LOCALE_ja_JP_UTF_8);
   std::locale::global(std::locale(LOCALE_fr_FR_UTF_8));
 
-#  if defined(_AIX)
   // Non localized output using C-locale
-  check(SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         fmt,
         file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
   // Use the global locale (fr_FR)
-  check(SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         lfmt,
         file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
-  // Use supplied locale (ja_JP). This locale has a different alternate.a
+  // Use supplied locale (ja_JP).
   check(loc,
-        SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
-        lfmt,
-        file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#  else                    // defined(_AIX)
-  // Non localized output using C-locale
-  check(SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
-        fmt,
-        file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-
-  // Use the global locale (fr_FR)
-  check(SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
+        SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         lfmt,
         file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
-  // Use supplied locale (ja_JP). This locale has a different alternate.a
-#    if defined(__FreeBSD__)
-  check(loc,
-        SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
-        lfmt,
-        file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#    else
-  check(loc,
-        SV("%z='+0000'\t%Ez='+0000'\t%Oz='+〇'\t%Z='UTC'\n"),
-        lfmt,
-        file_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#    endif
-#  endif                   // defined(_AIX)
   std::locale::global(std::locale::classic());
-#endif // !defined(__APPLE__) && !defined(_WIN32)
 }
 
 template <class CharT>
diff --git a/libcxx/test/std/time/time.syn/formatter.sys_time.pass.cpp b/libcxx/test/std/time/time.syn/formatter.sys_time.pass.cpp
index 2fed270cbade7..3a7d6f9a6b01f 100644
--- a/libcxx/test/std/time/time.syn/formatter.sys_time.pass.cpp
+++ b/libcxx/test/std/time/time.syn/formatter.sys_time.pass.cpp
@@ -900,12 +900,6 @@ static void test_valid_values_date_time() {
 
 template <class CharT>
 static void test_valid_values_time_zone() {
-// The Apple CI gives %z='-0700'	%Ez='-0700'	%Oz='-0700'	%Z='UTC'
-// -0700 looks like the local time where the CI happens to reside, therefore
-// omit this test on Apple.
-// The Windows CI gives %z='-0000', but on local machines set to a different
-// timezone, it gives e.g. %z='+0200'.
-#if !defined(__APPLE__) && !defined(_WIN32)
   using namespace std::literals::chrono_literals;
 
   constexpr std::basic_string_view<CharT> fmt  = SV("{:%%z='%z'%t%%Ez='%Ez'%t%%Oz='%Oz'%t%%Z='%Z'%n}");
@@ -914,48 +908,23 @@ static void test_valid_values_time_zone() {
   const std::locale loc(LOCALE_ja_JP_UTF_8);
   std::locale::global(std::locale(LOCALE_fr_FR_UTF_8));
 
-#  if defined(_AIX)
   // Non localized output using C-locale
-  check(SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         fmt,
         std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
   // Use the global locale (fr_FR)
-  check(SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         lfmt,
         std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
-  // Use supplied locale (ja_JP). This locale has a different alternate.a
+  // Use supplied locale (ja_JP).
   check(loc,
-        SV("%z='UTC'\t%Ez='UTC'\t%Oz='UTC'\t%Z='UTC'\n"),
-        lfmt,
-        std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#  else                                // defined(_AIX)
-  // Non localized output using C-locale
-  check(SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
-        fmt,
-        std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-
-  // Use the global locale (fr_FR)
-  check(SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
+        SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='UTC'\n"),
         lfmt,
         std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
 
-  // Use supplied locale (ja_JP). This locale has a different alternate.a
-#    if defined(__FreeBSD__)
-  check(loc,
-        SV("%z='+0000'\t%Ez='+0000'\t%Oz='+0000'\t%Z='UTC'\n"),
-        lfmt,
-        std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#    else
-  check(loc,
-        SV("%z='+0000'\t%Ez='+0000'\t%Oz='+〇'\t%Z='UTC'\n"),
-        lfmt,
-        std::chrono::sys_seconds(0s)); // 00:00:00 UTC Thursday, 1 January 1970
-#    endif
-#  endif                               // defined(_AIX)
   std::locale::global(std::locale::classic());
-#endif // !defined(__APPLE__) && !defined(_WIN32)
 }
 
 template <class CharT>

From e096c144921daba59963f15e89d2ca6fb32d3a78 Mon Sep 17 00:00:00 2001
From: Balazs Benics <benicsbalazs@gmail.com>
Date: Wed, 17 Apr 2024 08:02:49 +0200
Subject: [PATCH 694/886] [analyzer] Fix a security.cert.env.InvalidPtr crash

Fixes #88181
---
 clang/docs/ReleaseNotes.rst                            |  2 ++
 .../StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp |  6 +++++-
 clang/test/Analysis/invalid-ptr-checker.cpp            | 10 ++++++++++
 3 files changed, 17 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/Analysis/invalid-ptr-checker.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index efc32212f300c..6099f8ab02f44 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -679,6 +679,8 @@ Static Analyzer
   but not under any case blocks if ``unroll-loops=true`` analyzer config is
   set. (#GH68819)
 - Support C++23 static operator calls. (#GH84972)
+- Fixed a crash in ``security.cert.env.InvalidPtr`` checker when accidentally
+  matched user-defined ``strerror`` and similar library functions. (GH#88181)
 
 New features
 ^^^^^^^^^^^^
diff --git a/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
index e5dd907c660d8..b2947f590c4ec 100644
--- a/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
@@ -205,8 +205,12 @@ void InvalidPtrChecker::postPreviousReturnInvalidatingCall(
       CE, LCtx, CE->getType(), C.blockCount());
   State = State->BindExpr(CE, LCtx, RetVal);
 
+  const auto *SymRegOfRetVal =
+      dyn_cast_or_null<SymbolicRegion>(RetVal.getAsRegion());
+  if (!SymRegOfRetVal)
+    return;
+
   // Remember to this region.
-  const auto *SymRegOfRetVal = cast<SymbolicRegion>(RetVal.getAsRegion());
   const MemRegion *MR = SymRegOfRetVal->getBaseRegion();
   State = State->set<PreviousCallResultMap>(FD, MR);
 
diff --git a/clang/test/Analysis/invalid-ptr-checker.cpp b/clang/test/Analysis/invalid-ptr-checker.cpp
new file mode 100644
index 0000000000000..58bb45e0fb842
--- /dev/null
+++ b/clang/test/Analysis/invalid-ptr-checker.cpp
@@ -0,0 +1,10 @@
+// RUN: %clang_analyze_cc1 -analyzer-checker=core,security.cert.env.InvalidPtr -verify %s
+
+// expected-no-diagnostics
+
+namespace other {
+int strerror(int errnum); // custom strerror
+void no_crash_on_custom_strerror() {
+  (void)strerror(0); // no-crash
+}
+} // namespace other

From 024281d4d26344f9613b9115ea1fcbdbdba23235 Mon Sep 17 00:00:00 2001
From: Balazs Benics <benicsbalazs@gmail.com>
Date: Wed, 17 Apr 2024 08:02:49 +0200
Subject: [PATCH 695/886] [analyzer] Harden security.cert.env.InvalidPtr
 checker fn matching

Relates to #88181
---
 .../Checkers/cert/InvalidPtrChecker.cpp       | 25 +++++++++++--------
 1 file changed, 15 insertions(+), 10 deletions(-)

diff --git a/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
index b2947f590c4ec..fefe846b6911f 100644
--- a/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/cert/InvalidPtrChecker.cpp
@@ -48,14 +48,19 @@ class InvalidPtrChecker
   bool InvalidatingGetEnv = false;
 
   // GetEnv can be treated invalidating and non-invalidating as well.
-  const CallDescription GetEnvCall{{"getenv"}, 1};
+  const CallDescription GetEnvCall{CDM::CLibrary, {"getenv"}, 1};
 
   const CallDescriptionMap<HandlerFn> EnvpInvalidatingFunctions = {
-      {{{"setenv"}, 3}, &InvalidPtrChecker::EnvpInvalidatingCall},
-      {{{"unsetenv"}, 1}, &InvalidPtrChecker::EnvpInvalidatingCall},
-      {{{"putenv"}, 1}, &InvalidPtrChecker::EnvpInvalidatingCall},
-      {{{"_putenv_s"}, 2}, &InvalidPtrChecker::EnvpInvalidatingCall},
-      {{{"_wputenv_s"}, 2}, &InvalidPtrChecker::EnvpInvalidatingCall},
+      {{CDM::CLibrary, {"setenv"}, 3},
+       &InvalidPtrChecker::EnvpInvalidatingCall},
+      {{CDM::CLibrary, {"unsetenv"}, 1},
+       &InvalidPtrChecker::EnvpInvalidatingCall},
+      {{CDM::CLibrary, {"putenv"}, 1},
+       &InvalidPtrChecker::EnvpInvalidatingCall},
+      {{CDM::CLibrary, {"_putenv_s"}, 2},
+       &InvalidPtrChecker::EnvpInvalidatingCall},
+      {{CDM::CLibrary, {"_wputenv_s"}, 2},
+       &InvalidPtrChecker::EnvpInvalidatingCall},
   };
 
   void postPreviousReturnInvalidatingCall(const CallEvent &Call,
@@ -63,13 +68,13 @@ class InvalidPtrChecker
 
   // SEI CERT ENV34-C
   const CallDescriptionMap<HandlerFn> PreviousCallInvalidatingFunctions = {
-      {{{"setlocale"}, 2},
+      {{CDM::CLibrary, {"setlocale"}, 2},
        &InvalidPtrChecker::postPreviousReturnInvalidatingCall},
-      {{{"strerror"}, 1},
+      {{CDM::CLibrary, {"strerror"}, 1},
        &InvalidPtrChecker::postPreviousReturnInvalidatingCall},
-      {{{"localeconv"}, 0},
+      {{CDM::CLibrary, {"localeconv"}, 0},
        &InvalidPtrChecker::postPreviousReturnInvalidatingCall},
-      {{{"asctime"}, 1},
+      {{CDM::CLibrary, {"asctime"}, 1},
        &InvalidPtrChecker::postPreviousReturnInvalidatingCall},
   };
 

From b851c7f1fc4fd83ea84d565bbdc30fd0d356788c Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Wed, 17 Apr 2024 08:05:43 +0200
Subject: [PATCH 696/886] [clang][dataflow] Support `StmtExpr` in
 `PropagateResultObject()`. (#88872)

This patch adds a test that assert-fails without the fix.
---
 .../FlowSensitive/DataflowEnvironment.cpp     |  5 ++++
 .../Analysis/FlowSensitive/TransferTest.cpp   | 26 +++++++++++++++++++
 2 files changed, 31 insertions(+)

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index 3bf3807268bee..f2b4a67e5bc97 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -470,6 +470,11 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
       return;
     }
 
+    if (auto *SE = dyn_cast<StmtExpr>(E)) {
+      PropagateResultObject(cast<Expr>(SE->getSubStmt()->body_back()), Loc);
+      return;
+    }
+
     // All other expression nodes that propagate a record prvalue should have
     // exactly one child.
     SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index d8bcc3da4b8b1..d7a51b009712f 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -3182,6 +3182,32 @@ TEST(TransferTest, ResultObjectLocationForStdInitializerListExpr) {
       });
 }
 
+TEST(TransferTest, ResultObjectLocationForStmtExpr) {
+  std::string Code = R"(
+    struct S {};
+    void target() {
+      S s = ({ S(); });
+      // [[p]]
+    }
+  )";
+  using ast_matchers::cxxConstructExpr;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  using ast_matchers::traverse;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *Construct = selectFirst<CXXConstructExpr>(
+            "construct", match(cxxConstructExpr().bind("construct"), ASTCtx));
+
+        EXPECT_EQ(&Env.getResultObjectLocation(*Construct),
+                  &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "s"));
+      });
+}
+
 TEST(TransferTest, ResultObjectLocationPropagatesThroughConditionalOperator) {
   std::string Code = R"(
     struct A {

From 47148832d4e3bf4901430732f1af6673147accb2 Mon Sep 17 00:00:00 2001
From: Hideto Ueno <uenoku.tokotoko@gmail.com>
Date: Wed, 17 Apr 2024 15:09:47 +0900
Subject: [PATCH 697/886] [mlir][python] Add `walk` method to PyOperationBase
 (#87962)

This commit adds `walk` method to PyOperationBase that uses a python
object as a callback, e.g. `op.walk(callback)`. Currently callback must
return a walk result explicitly.

We(SiFive) have implemented walk method with python in our internal
python tool for a while. However the overhead of python is expensive and
it didn't scale well for large MLIR files. Just replacing walk with this
version reduced the entire execution time of the tool by 30~40% and
there are a few configs that the tool takes several hours to finish so
this commit significantly improves tool performance.
---
 mlir/include/mlir-c/IR.h                      | 10 ++-
 .../mlir/Bindings/Python/PybindAdaptors.h     |  1 +
 mlir/lib/Bindings/Python/IRCore.cpp           | 32 +++++++-
 mlir/lib/Bindings/Python/IRModule.h           |  4 +
 mlir/lib/CAPI/IR/IR.cpp                       | 21 +++++-
 mlir/test/CAPI/ir.c                           | 58 +++++++++++---
 mlir/test/python/ir/operation.py              | 75 +++++++++++++++++++
 7 files changed, 184 insertions(+), 17 deletions(-)

diff --git a/mlir/include/mlir-c/IR.h b/mlir/include/mlir-c/IR.h
index 82da511f807a3..32abacf353133 100644
--- a/mlir/include/mlir-c/IR.h
+++ b/mlir/include/mlir-c/IR.h
@@ -705,6 +705,13 @@ MLIR_CAPI_EXPORTED void mlirOperationMoveAfter(MlirOperation op,
 MLIR_CAPI_EXPORTED void mlirOperationMoveBefore(MlirOperation op,
                                                 MlirOperation other);
 
+/// Operation walk result.
+typedef enum MlirWalkResult {
+  MlirWalkResultAdvance,
+  MlirWalkResultInterrupt,
+  MlirWalkResultSkip
+} MlirWalkResult;
+
 /// Traversal order for operation walk.
 typedef enum MlirWalkOrder {
   MlirWalkPreOrder,
@@ -713,7 +720,8 @@ typedef enum MlirWalkOrder {
 
 /// Operation walker type. The handler is passed an (opaque) reference to an
 /// operation and a pointer to a `userData`.
-typedef void (*MlirOperationWalkCallback)(MlirOperation, void *userData);
+typedef MlirWalkResult (*MlirOperationWalkCallback)(MlirOperation,
+                                                    void *userData);
 
 /// Walks operation `op` in `walkOrder` and calls `callback` on that operation.
 /// `*userData` is passed to the callback as well and can be used to tunnel some
diff --git a/mlir/include/mlir/Bindings/Python/PybindAdaptors.h b/mlir/include/mlir/Bindings/Python/PybindAdaptors.h
index 52f6321251919..d8f22c7aa1709 100644
--- a/mlir/include/mlir/Bindings/Python/PybindAdaptors.h
+++ b/mlir/include/mlir/Bindings/Python/PybindAdaptors.h
@@ -18,6 +18,7 @@
 #ifndef MLIR_BINDINGS_PYTHON_PYBINDADAPTORS_H
 #define MLIR_BINDINGS_PYTHON_PYBINDADAPTORS_H
 
+#include <pybind11/functional.h>
 #include <pybind11/pybind11.h>
 #include <pybind11/pytypes.h>
 #include <pybind11/stl.h>
diff --git a/mlir/lib/Bindings/Python/IRCore.cpp b/mlir/lib/Bindings/Python/IRCore.cpp
index 734f2f7f3f94c..d875f4eba2b13 100644
--- a/mlir/lib/Bindings/Python/IRCore.cpp
+++ b/mlir/lib/Bindings/Python/IRCore.cpp
@@ -674,6 +674,7 @@ void PyMlirContext::clearOperationsInside(PyOperationBase &op) {
       data->rootOp.getOperation().getContext()->clearOperation(op);
     else
       data->rootSeen = true;
+    return MlirWalkResult::MlirWalkResultAdvance;
   };
   mlirOperationWalk(op.getOperation(), invalidatingCallback,
                     static_cast<void *>(&data), MlirWalkPreOrder);
@@ -1249,6 +1250,21 @@ void PyOperationBase::writeBytecode(const py::object &fileObject,
                               .str());
 }
 
+void PyOperationBase::walk(
+    std::function<MlirWalkResult(MlirOperation)> callback,
+    MlirWalkOrder walkOrder) {
+  PyOperation &operation = getOperation();
+  operation.checkValid();
+  MlirOperationWalkCallback walkCallback = [](MlirOperation op,
+                                              void *userData) {
+    auto *fn =
+        static_cast<std::function<MlirWalkResult(MlirOperation)> *>(userData);
+    return (*fn)(op);
+  };
+
+  mlirOperationWalk(operation, walkCallback, &callback, walkOrder);
+}
+
 py::object PyOperationBase::getAsm(bool binary,
                                    std::optional<int64_t> largeElementsLimit,
                                    bool enableDebugInfo, bool prettyDebugInfo,
@@ -2511,6 +2527,15 @@ void mlir::python::populateIRCore(py::module &m) {
       .value("NOTE", MlirDiagnosticNote)
       .value("REMARK", MlirDiagnosticRemark);
 
+  py::enum_<MlirWalkOrder>(m, "WalkOrder", py::module_local())
+      .value("PRE_ORDER", MlirWalkPreOrder)
+      .value("POST_ORDER", MlirWalkPostOrder);
+
+  py::enum_<MlirWalkResult>(m, "WalkResult", py::module_local())
+      .value("ADVANCE", MlirWalkResultAdvance)
+      .value("INTERRUPT", MlirWalkResultInterrupt)
+      .value("SKIP", MlirWalkResultSkip);
+
   //----------------------------------------------------------------------------
   // Mapping of Diagnostics.
   //----------------------------------------------------------------------------
@@ -2989,8 +3014,7 @@ void mlir::python::populateIRCore(py::module &m) {
            py::arg("binary") = false, kOperationPrintStateDocstring)
       .def("print",
            py::overload_cast<std::optional<int64_t>, bool, bool, bool, bool,
-                             bool, py::object, bool>(
-               &PyOperationBase::print),
+                             bool, py::object, bool>(&PyOperationBase::print),
            // Careful: Lots of arguments must match up with print method.
            py::arg("large_elements_limit") = py::none(),
            py::arg("enable_debug_info") = false,
@@ -3038,7 +3062,9 @@ void mlir::python::populateIRCore(py::module &m) {
             return operation.createOpView();
           },
           "Detaches the operation from its parent block.")
-      .def("erase", [](PyOperationBase &self) { self.getOperation().erase(); });
+      .def("erase", [](PyOperationBase &self) { self.getOperation().erase(); })
+      .def("walk", &PyOperationBase::walk, py::arg("callback"),
+           py::arg("walk_order") = MlirWalkPostOrder);
 
   py::class_<PyOperation, PyOperationBase>(m, "Operation", py::module_local())
       .def_static("create", &PyOperation::create, py::arg("name"),
diff --git a/mlir/lib/Bindings/Python/IRModule.h b/mlir/lib/Bindings/Python/IRModule.h
index 9acfdde25ae04..b038a0c54d29b 100644
--- a/mlir/lib/Bindings/Python/IRModule.h
+++ b/mlir/lib/Bindings/Python/IRModule.h
@@ -579,6 +579,10 @@ class PyOperationBase {
   void writeBytecode(const pybind11::object &fileObject,
                      std::optional<int64_t> bytecodeVersion);
 
+  // Implement the walk method.
+  void walk(std::function<MlirWalkResult(MlirOperation)> callback,
+            MlirWalkOrder walkOrder);
+
   /// Moves the operation before or after the other operation.
   void moveAfter(PyOperationBase &other);
   void moveBefore(PyOperationBase &other);
diff --git a/mlir/lib/CAPI/IR/IR.cpp b/mlir/lib/CAPI/IR/IR.cpp
index cdb64f4ec4a40..a72cd247e73f6 100644
--- a/mlir/lib/CAPI/IR/IR.cpp
+++ b/mlir/lib/CAPI/IR/IR.cpp
@@ -717,17 +717,34 @@ void mlirOperationMoveBefore(MlirOperation op, MlirOperation other) {
   return unwrap(op)->moveBefore(unwrap(other));
 }
 
+static mlir::WalkResult unwrap(MlirWalkResult result) {
+  switch (result) {
+  case MlirWalkResultAdvance:
+    return mlir::WalkResult::advance();
+
+  case MlirWalkResultInterrupt:
+    return mlir::WalkResult::interrupt();
+
+  case MlirWalkResultSkip:
+    return mlir::WalkResult::skip();
+  }
+}
+
 void mlirOperationWalk(MlirOperation op, MlirOperationWalkCallback callback,
                        void *userData, MlirWalkOrder walkOrder) {
   switch (walkOrder) {
 
   case MlirWalkPreOrder:
     unwrap(op)->walk<mlir::WalkOrder::PreOrder>(
-        [callback, userData](Operation *op) { callback(wrap(op), userData); });
+        [callback, userData](Operation *op) {
+          return unwrap(callback(wrap(op), userData));
+        });
     break;
   case MlirWalkPostOrder:
     unwrap(op)->walk<mlir::WalkOrder::PostOrder>(
-        [callback, userData](Operation *op) { callback(wrap(op), userData); });
+        [callback, userData](Operation *op) {
+          return unwrap(callback(wrap(op), userData));
+        });
   }
 }
 
diff --git a/mlir/test/CAPI/ir.c b/mlir/test/CAPI/ir.c
index 8e79338c57a22..3d05b2a12dd8e 100644
--- a/mlir/test/CAPI/ir.c
+++ b/mlir/test/CAPI/ir.c
@@ -2244,9 +2244,22 @@ typedef struct {
   const char *x;
 } callBackData;
 
-void walkCallBack(MlirOperation op, void *rootOpVoid) {
+MlirWalkResult walkCallBack(MlirOperation op, void *rootOpVoid) {
   fprintf(stderr, "%s: %s\n", ((callBackData *)(rootOpVoid))->x,
           mlirIdentifierStr(mlirOperationGetName(op)).data);
+  return MlirWalkResultAdvance;
+}
+
+MlirWalkResult walkCallBackTestWalkResult(MlirOperation op, void *rootOpVoid) {
+  fprintf(stderr, "%s: %s\n", ((callBackData *)(rootOpVoid))->x,
+          mlirIdentifierStr(mlirOperationGetName(op)).data);
+  if (strcmp(mlirIdentifierStr(mlirOperationGetName(op)).data, "func.func") ==
+      0)
+    return MlirWalkResultSkip;
+  if (strcmp(mlirIdentifierStr(mlirOperationGetName(op)).data, "arith.addi") ==
+      0)
+    return MlirWalkResultInterrupt;
+  return MlirWalkResultAdvance;
 }
 
 int testOperationWalk(MlirContext ctx) {
@@ -2259,6 +2272,9 @@ int testOperationWalk(MlirContext ctx) {
                              "    arith.addi %1, %1: i32\n"
                              "    return\n"
                              "  }\n"
+                             "  func.func @bar() {\n"
+                             "    return\n"
+                             "  }\n"
                              "}";
   MlirModule module =
       mlirModuleCreateParse(ctx, mlirStringRefCreateFromCString(moduleString));
@@ -2266,22 +2282,42 @@ int testOperationWalk(MlirContext ctx) {
   callBackData data;
   data.x = "i love you";
 
-  // CHECK: i love you: arith.constant
-  // CHECK: i love you: arith.addi
-  // CHECK: i love you: func.return
-  // CHECK: i love you: func.func
-  // CHECK: i love you: builtin.module
+  // CHECK-NEXT: i love you: arith.constant
+  // CHECK-NEXT: i love you: arith.addi
+  // CHECK-NEXT: i love you: func.return
+  // CHECK-NEXT: i love you: func.func
+  // CHECK-NEXT: i love you: func.return
+  // CHECK-NEXT: i love you: func.func
+  // CHECK-NEXT: i love you: builtin.module
   mlirOperationWalk(mlirModuleGetOperation(module), walkCallBack,
                     (void *)(&data), MlirWalkPostOrder);
 
   data.x = "i don't love you";
-  // CHECK: i don't love you: builtin.module
-  // CHECK: i don't love you: func.func
-  // CHECK: i don't love you: arith.constant
-  // CHECK: i don't love you: arith.addi
-  // CHECK: i don't love you: func.return
+  // CHECK-NEXT: i don't love you: builtin.module
+  // CHECK-NEXT: i don't love you: func.func
+  // CHECK-NEXT: i don't love you: arith.constant
+  // CHECK-NEXT: i don't love you: arith.addi
+  // CHECK-NEXT: i don't love you: func.return
+  // CHECK-NEXT: i don't love you: func.func
+  // CHECK-NEXT: i don't love you: func.return
   mlirOperationWalk(mlirModuleGetOperation(module), walkCallBack,
                     (void *)(&data), MlirWalkPreOrder);
+
+  data.x = "interrupt";
+  // Interrupted at `arith.addi`
+  // CHECK-NEXT: interrupt: arith.constant
+  // CHECK-NEXT: interrupt: arith.addi
+  mlirOperationWalk(mlirModuleGetOperation(module), walkCallBackTestWalkResult,
+                    (void *)(&data), MlirWalkPostOrder);
+
+  data.x = "skip";
+  // Skip at `func.func`
+  // CHECK-NEXT: skip: builtin.module
+  // CHECK-NEXT: skip: func.func
+  // CHECK-NEXT: skip: func.func
+  mlirOperationWalk(mlirModuleGetOperation(module), walkCallBackTestWalkResult,
+                    (void *)(&data), MlirWalkPreOrder);
+
   mlirModuleDestroy(module);
   return 0;
 }
diff --git a/mlir/test/python/ir/operation.py b/mlir/test/python/ir/operation.py
index 04f8a9936e31f..9666e63bda1e0 100644
--- a/mlir/test/python/ir/operation.py
+++ b/mlir/test/python/ir/operation.py
@@ -1015,3 +1015,78 @@ def testOperationParse():
         print(
             f"op_with_source_name: {o.get_asm(enable_debug_info=True, use_local_scope=True)}"
         )
+
+
+# CHECK-LABEL: TEST: testOpWalk
+@run
+def testOpWalk():
+    ctx = Context()
+    ctx.allow_unregistered_dialects = True
+    module = Module.parse(
+        r"""
+    builtin.module {
+      func.func @f() {
+        func.return
+      }
+    }
+  """,
+        ctx,
+    )
+
+    def callback(op):
+        print(op.name)
+        return WalkResult.ADVANCE
+
+    # Test post-order walk (default).
+    # CHECK-NEXT:  Post-order
+    # CHECK-NEXT:  func.return
+    # CHECK-NEXT:  func.func
+    # CHECK-NEXT:  builtin.module
+    print("Post-order")
+    module.operation.walk(callback)
+
+    # Test pre-order walk.
+    # CHECK-NEXT:  Pre-order
+    # CHECK-NEXT:  builtin.module
+    # CHECK-NEXT:  func.fun
+    # CHECK-NEXT:  func.return
+    print("Pre-order")
+    module.operation.walk(callback, WalkOrder.PRE_ORDER)
+
+    # Test interrput.
+    # CHECK-NEXT:  Interrupt post-order
+    # CHECK-NEXT:  func.return
+    print("Interrupt post-order")
+
+    def callback(op):
+        print(op.name)
+        return WalkResult.INTERRUPT
+
+    module.operation.walk(callback)
+
+    # Test skip.
+    # CHECK-NEXT:  Skip pre-order
+    # CHECK-NEXT:  builtin.module
+    print("Skip pre-order")
+
+    def callback(op):
+        print(op.name)
+        return WalkResult.SKIP
+
+    module.operation.walk(callback, WalkOrder.PRE_ORDER)
+
+    # Test exception.
+    # CHECK: Exception
+    # CHECK-NEXT: func.return
+    # CHECK-NEXT: Exception raised
+    print("Exception")
+
+    def callback(op):
+        print(op.name)
+        raise ValueError
+        return WalkResult.ADVANCE
+
+    try:
+        module.operation.walk(callback)
+    except ValueError:
+        print("Exception raised")

From 1bccbe1f49abc39b9f980cf3f1b171da5541d1a4 Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Wed, 17 Apr 2024 08:17:56 +0200
Subject: [PATCH 698/886] [clang][dataflow] Treat `BuiltinBitCastExpr`
 correctly in `PropagateResultObject()`. (#88875)

This patch includes a test that assert-fails without the fix.
---
 .../FlowSensitive/DataflowEnvironment.cpp     |  6 ++++-
 .../Analysis/FlowSensitive/TransferTest.cpp   | 26 +++++++++++++++++++
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index f2b4a67e5bc97..3f1600d9ac5d8 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -419,7 +419,11 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
     // below them can initialize the same object (or part of it).
     if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
         isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
-        isa<CXXStdInitializerListExpr>(E)) {
+        isa<CXXStdInitializerListExpr>(E) ||
+        // We treat `BuiltinBitCastExpr` as an "original initializer" too as
+        // it may not even be casting from a record type -- and even if it is,
+        // the two objects are in general of unrelated type.
+        isa<BuiltinBitCastExpr>(E)) {
       return;
     }
     if (auto *Op = dyn_cast<BinaryOperator>(E);
diff --git a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
index d7a51b009712f..97ec32126c1dc 100644
--- a/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TransferTest.cpp
@@ -3208,6 +3208,32 @@ TEST(TransferTest, ResultObjectLocationForStmtExpr) {
       });
 }
 
+TEST(TransferTest, ResultObjectLocationForBuiltinBitCastExpr) {
+  std::string Code = R"(
+    struct S { int i; };
+    void target(int i) {
+      S s = __builtin_bit_cast(S, i);
+      // [[p]]
+    }
+  )";
+  using ast_matchers::explicitCastExpr;
+  using ast_matchers::match;
+  using ast_matchers::selectFirst;
+  using ast_matchers::traverse;
+  runDataflow(
+      Code,
+      [](const llvm::StringMap<DataflowAnalysisState<NoopLattice>> &Results,
+         ASTContext &ASTCtx) {
+        const Environment &Env = getEnvironmentAtAnnotation(Results, "p");
+
+        auto *BuiltinBitCast = selectFirst<BuiltinBitCastExpr>(
+            "cast", match(explicitCastExpr().bind("cast"), ASTCtx));
+
+        EXPECT_EQ(&Env.getResultObjectLocation(*BuiltinBitCast),
+                  &getLocForDecl<RecordStorageLocation>(ASTCtx, Env, "s"));
+      });
+}
+
 TEST(TransferTest, ResultObjectLocationPropagatesThroughConditionalOperator) {
   std::string Code = R"(
     struct A {

From 64c649585ca23a0c996d8814d2796cd348441d69 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Wed, 17 Apr 2024 09:20:55 +0300
Subject: [PATCH 699/886] [clang][NFC] Move `Sema::SkipBodyInfo` into namespace
 scope

This makes it forward-declarable, and needed from splitting `Sema` up.
---
 clang/include/clang/Sema/Sema.h  | 16 ++++++++--------
 clang/lib/Parse/ParseDecl.cpp    |  2 +-
 clang/lib/Parse/ParseDeclCXX.cpp |  2 +-
 clang/lib/Parse/ParseObjc.cpp    |  4 ++--
 clang/lib/Parse/Parser.cpp       |  2 +-
 clang/lib/Sema/SemaDecl.cpp      |  4 ++--
 6 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 77150a318ee47..091c5c02f75df 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -352,6 +352,14 @@ class PreferredTypeBuilder {
   llvm::function_ref<QualType()> ComputeType;
 };
 
+struct SkipBodyInfo {
+  SkipBodyInfo() = default;
+  bool ShouldSkip = false;
+  bool CheckSameAsPrevious = false;
+  NamedDecl *Previous = nullptr;
+  NamedDecl *New = nullptr;
+};
+
 /// Describes the result of template argument deduction.
 ///
 /// The TemplateDeductionResult enumeration describes the result of
@@ -2627,14 +2635,6 @@ class Sema final : public SemaBase {
     return Entity->getOwningModule();
   }
 
-  struct SkipBodyInfo {
-    SkipBodyInfo() = default;
-    bool ShouldSkip = false;
-    bool CheckSameAsPrevious = false;
-    NamedDecl *Previous = nullptr;
-    NamedDecl *New = nullptr;
-  };
-
   DeclGroupPtrTy ConvertDeclToDeclGroup(Decl *Ptr, Decl *OwnedType = nullptr);
 
   ParsedType getTypeName(const IdentifierInfo &II, SourceLocation NameLoc,
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index c881b37507771..274ee7b10c178 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -5331,7 +5331,7 @@ void Parser::ParseEnumSpecifier(SourceLocation StartLoc, DeclSpec &DS,
 
   stripTypeAttributesOffDeclSpec(attrs, DS, TUK);
 
-  Sema::SkipBodyInfo SkipBody;
+  SkipBodyInfo SkipBody;
   if (!Name && TUK == Sema::TUK_Definition && Tok.is(tok::l_brace) &&
       NextToken().is(tok::identifier))
     SkipBody = Actions.shouldSkipAnonEnumBody(getCurScope(),
diff --git a/clang/lib/Parse/ParseDeclCXX.cpp b/clang/lib/Parse/ParseDeclCXX.cpp
index cd4803d51bc1d..51fd64b2d01aa 100644
--- a/clang/lib/Parse/ParseDeclCXX.cpp
+++ b/clang/lib/Parse/ParseDeclCXX.cpp
@@ -2092,7 +2092,7 @@ void Parser::ParseClassSpecifier(tok::TokenKind TagTokKind,
   TypeResult TypeResult = true;      // invalid
 
   bool Owned = false;
-  Sema::SkipBodyInfo SkipBody;
+  SkipBodyInfo SkipBody;
   if (TemplateId) {
     // Explicit specialization, class template partial specialization,
     // or explicit instantiation.
diff --git a/clang/lib/Parse/ParseObjc.cpp b/clang/lib/Parse/ParseObjc.cpp
index 887d7a36cee7e..671dcb71e51a3 100644
--- a/clang/lib/Parse/ParseObjc.cpp
+++ b/clang/lib/Parse/ParseObjc.cpp
@@ -375,7 +375,7 @@ Decl *Parser::ParseObjCAtInterfaceDeclaration(SourceLocation AtLoc,
     Actions.ActOnTypedefedProtocols(protocols, protocolLocs,
                                     superClassId, superClassLoc);
 
-  Sema::SkipBodyInfo SkipBody;
+  SkipBodyInfo SkipBody;
   ObjCInterfaceDecl *ClsType = Actions.ActOnStartClassInterface(
       getCurScope(), AtLoc, nameId, nameLoc, typeParameterList, superClassId,
       superClassLoc, typeArgs,
@@ -2133,7 +2133,7 @@ Parser::ParseObjCAtProtocolDeclaration(SourceLocation AtLoc,
                                   /*consumeLastToken=*/true))
     return nullptr;
 
-  Sema::SkipBodyInfo SkipBody;
+  SkipBodyInfo SkipBody;
   ObjCProtocolDecl *ProtoType = Actions.ActOnStartProtocolInterface(
       AtLoc, protocolName, nameLoc, ProtocolRefs.data(), ProtocolRefs.size(),
       ProtocolLocs.data(), EndProtoLoc, attrs, &SkipBody);
diff --git a/clang/lib/Parse/Parser.cpp b/clang/lib/Parse/Parser.cpp
index d6f2b9f448cd5..ef46fc74cedc1 100644
--- a/clang/lib/Parse/Parser.cpp
+++ b/clang/lib/Parse/Parser.cpp
@@ -1441,7 +1441,7 @@ Decl *Parser::ParseFunctionDefinition(ParsingDeclarator &D,
 
   // Tell the actions module that we have entered a function definition with the
   // specified Declarator for the function.
-  Sema::SkipBodyInfo SkipBody;
+  SkipBodyInfo SkipBody;
   Decl *Res = Actions.ActOnStartOfFunctionDef(getCurScope(), D,
                                               TemplateInfo.TemplateParams
                                                   ? *TemplateInfo.TemplateParams
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 745cf41e204e7..19abd5327b73a 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -3037,7 +3037,7 @@ static void checkNewAttributesAfterDef(Sema &S, Decl *New, const Decl *Old) {
 
     if (isa<AliasAttr>(NewAttribute) || isa<IFuncAttr>(NewAttribute)) {
       if (FunctionDecl *FD = dyn_cast<FunctionDecl>(New)) {
-        Sema::SkipBodyInfo SkipBody;
+        SkipBodyInfo SkipBody;
         S.CheckForFunctionRedefinition(FD, cast<FunctionDecl>(Def), &SkipBody);
 
         // If we're skipping this definition, drop the "alias" attribute.
@@ -19999,7 +19999,7 @@ EnumConstantDecl *Sema::CheckEnumConstant(EnumDecl *Enum,
                                   Val, EnumVal);
 }
 
-Sema::SkipBodyInfo Sema::shouldSkipAnonEnumBody(Scope *S, IdentifierInfo *II,
+SkipBodyInfo Sema::shouldSkipAnonEnumBody(Scope *S, IdentifierInfo *II,
                                                 SourceLocation IILoc) {
   if (!(getLangOpts().Modules || getLangOpts().ModulesLocalVisibility) ||
       !getLangOpts().CPlusPlus)

From 16f188761da1df6ba5e6627b8742aacfec8e9ec5 Mon Sep 17 00:00:00 2001
From: YunQiang Su <syq@debian.org>
Date: Wed, 17 Apr 2024 14:23:37 +0800
Subject: [PATCH 700/886] CompilerRT: Normalize
 COMPILER_RT_DEFAULT_TARGET_TRIPLE (#88835)

If LLVM is configured with -DLLVM_DEFAULT_TARGET_TRIPLE, or compiler_rt
is configured with -DCOMPILER_RT_DEFAULT_TARGET_TRIPLE, while the
argument is not normalized, such as Debian-style vendor-less triple,
clang will try to find libclang_rt in lib/<normalized_triple>, while
libclang_rt is placed into lib/<triple_arg>.

Let's also place libclang_rt into lib/<normalized_triple>.
---
 compiler-rt/cmake/Modules/CompilerRTUtils.cmake | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
index e8e5f612d5b03..6d413f6753bc0 100644
--- a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
+++ b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
@@ -368,6 +368,12 @@ macro(construct_compiler_rt_default_triple)
           "Default triple for which compiler-rt runtimes will be built.")
   endif()
 
+  if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
+    execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${COMPILER_RT_DEFAULT_TARGET_TRIPLE} -print-effective-triple
+                    OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
+                    OUTPUT_STRIP_TRAILING_WHITESPACE)
+  endif()
+
   string(REPLACE "-" ";" LLVM_TARGET_TRIPLE_LIST ${COMPILER_RT_DEFAULT_TARGET_TRIPLE})
   list(GET LLVM_TARGET_TRIPLE_LIST 0 COMPILER_RT_DEFAULT_TARGET_ARCH)
 

From b090569685699abe4a8031ad442a0f81e373146b Mon Sep 17 00:00:00 2001
From: Jesse Huang <jesse.huang@sifive.com>
Date: Tue, 16 Apr 2024 23:36:27 -0700
Subject: [PATCH 701/886] [RISCV] Support Zama16b1p0 (#88474)

This patch adds the support for Zama16b version 1.0, which has been
added to RVA23U64 optional extensions recently
---
 clang/test/Preprocessor/riscv-target-features.c | 7 +++++++
 llvm/docs/RISCVUsage.rst                        | 3 ++-
 llvm/lib/Support/RISCVISAInfo.cpp               | 1 +
 llvm/lib/Target/RISCV/RISCVFeatures.td          | 7 +++++++
 llvm/test/CodeGen/RISCV/attributes.ll           | 4 ++++
 llvm/test/MC/RISCV/attribute-arch.s             | 3 +++
 llvm/unittests/Support/RISCVISAInfoTest.cpp     | 1 +
 7 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/clang/test/Preprocessor/riscv-target-features.c b/clang/test/Preprocessor/riscv-target-features.c
index ec7764bb53818..646043681fe33 100644
--- a/clang/test/Preprocessor/riscv-target-features.c
+++ b/clang/test/Preprocessor/riscv-target-features.c
@@ -79,6 +79,7 @@
 // CHECK-NOT: __riscv_za128rs {{.*$}}
 // CHECK-NOT: __riscv_za64rs {{.*$}}
 // CHECK-NOT: __riscv_zacas {{.*$}}
+// CHECK-NOT: __riscv_zama16b {{.*$}}
 // CHECK-NOT: __riscv_zawrs {{.*$}}
 // CHECK-NOT: __riscv_zba {{.*$}}
 // CHECK-NOT: __riscv_zbb {{.*$}}
@@ -704,6 +705,12 @@
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZACAS-EXT %s
 // CHECK-ZACAS-EXT: __riscv_zacas 1000000{{$}}
 
+// RUN: %clang --target=riscv32 -march=rv32izama16b -x c -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-ZAMA16B-EXT %s
+// RUN: %clang --target=riscv64 -march=rv64izama16b  -x c -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-ZAMA16B-EXT %s
+// CHECK-ZAMA16B-EXT: __riscv_zama16b  1000000{{$}}
+
 // RUN: %clang --target=riscv32-unknown-linux-gnu \
 // RUN:   -march=rv32izawrs -E -dM %s \
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZAWRS-EXT %s
diff --git a/llvm/docs/RISCVUsage.rst b/llvm/docs/RISCVUsage.rst
index 6f5eba263def4..a4cf17a8398a8 100644
--- a/llvm/docs/RISCVUsage.rst
+++ b/llvm/docs/RISCVUsage.rst
@@ -119,6 +119,7 @@ on support follow.
      ``Za128rs``       Supported (`See note <#riscv-profiles-extensions-note>`__)
      ``Za64rs``        Supported (`See note <#riscv-profiles-extensions-note>`__)
      ``Zacas``         Supported (`See note <#riscv-zacas-note>`__)
+     ``Zama16b``       Supported (`See note <#riscv-profiles-extensions-note>`__)
      ``Zawrs``         Assembly Support
      ``Zba``           Supported
      ``Zbb``           Supported
@@ -237,7 +238,7 @@ Supported
 
 .. _riscv-profiles-extensions-note:
 
-``Za128rs``, ``Za64rs``, ``Zic64b``, ``Ziccamoa``, ``Ziccif``, ``Zicclsm``, ``Ziccrse``, ``Shcounterenvw``, ``Shgatpa``, ``Shtvala``, ``Shvsatpa``, ``Shvstvala``, ``Shvstvecd``, ``Ssccptr``, ``Sscounterenw``, ``Ssstateen``, ``Ssstrict``, ``Sstvala``, ``Sstvecd``, ``Ssu64xl``, ``Svade``, ``Svbare``
+``Za128rs``, ``Za64rs``, ``Zama16b``, ``Zic64b``, ``Ziccamoa``, ``Ziccif``, ``Zicclsm``, ``Ziccrse``, ``Shcounterenvw``, ``Shgatpa``, ``Shtvala``, ``Shvsatpa``, ``Shvstvala``, ``Shvstvecd``, ``Ssccptr``, ``Sscounterenw``, ``Ssstateen``, ``Ssstrict``, ``Sstvala``, ``Sstvecd``, ``Ssu64xl``, ``Svade``, ``Svbare``
   These extensions are defined as part of the `RISC-V Profiles specification <https://github.com/riscv/riscv-profiles/releases/tag/v1.0>`__.  They do not introduce any new features themselves, but instead describe existing hardware features.
 
   .. _riscv-zacas-note:
diff --git a/llvm/lib/Support/RISCVISAInfo.cpp b/llvm/lib/Support/RISCVISAInfo.cpp
index cbdc64bc7a97b..fa967403ea449 100644
--- a/llvm/lib/Support/RISCVISAInfo.cpp
+++ b/llvm/lib/Support/RISCVISAInfo.cpp
@@ -119,6 +119,7 @@ static const RISCVSupportedExtension SupportedExtensions[] = {
     {"za128rs", {1, 0}},
     {"za64rs", {1, 0}},
     {"zacas", {1, 0}},
+    {"zama16b", {1, 0}},
     {"zawrs", {1, 0}},
 
     {"zba", {1, 0}},
diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index 561187c39a4a0..f830ead5dd692 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -208,6 +208,13 @@ def HasStdExtAOrZalrsc
                          "'A' (Atomic Instructions) or "
                          "'Zalrsc' (Load-Reserved/Store-Conditional)">;
 
+def FeatureStdExtZama16b
+    : SubtargetFeature<"zama16b", "HasStdExtZama16b", "true",
+                       "'Zama16b' (Atomic 16-byte misaligned loads, stores and AMOs)">;
+def HasStdExtZama16b : Predicate<"Subtarget->hasStdExtZama16b()">,
+                       AssemblerPredicate<(all_of FeatureStdExtZama16b),
+                           "'Zama16b' (Atomic 16-byte misaligned loads, stores and AMOs)">;
+
 def FeatureStdExtZawrs : SubtargetFeature<"zawrs", "HasStdExtZawrs", "true",
                                           "'Zawrs' (Wait on Reservation Set)">;
 def HasStdExtZawrs : Predicate<"Subtarget->hasStdExtZawrs()">,
diff --git a/llvm/test/CodeGen/RISCV/attributes.ll b/llvm/test/CodeGen/RISCV/attributes.ll
index 2326599bf3513..080783fdeec02 100644
--- a/llvm/test/CodeGen/RISCV/attributes.ll
+++ b/llvm/test/CodeGen/RISCV/attributes.ll
@@ -115,6 +115,7 @@
 ; RUN: llc -mtriple=riscv32 -mattr=+zacas %s -o - | FileCheck --check-prefix=RV32ZACAS %s
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-zalasr %s -o - | FileCheck --check-prefix=RV32ZALASR %s
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-zalrsc %s -o - | FileCheck --check-prefix=RV32ZALRSC %s
+; RUN: llc -mtriple=riscv32 -mattr=+zama16b %s -o - | FileCheck --check-prefixes=CHECK,RV32ZAMA16B %s
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-zicfilp %s -o - | FileCheck --check-prefix=RV32ZICFILP %s
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-zabha %s -o - | FileCheck --check-prefix=RV32ZABHA %s
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-ssnpm  %s -o - | FileCheck --check-prefix=RV32SSNPM %s
@@ -199,6 +200,7 @@
 ; RUN: llc -mtriple=riscv64 -mattr=+xtheadvdot %s -o - | FileCheck --check-prefixes=CHECK,RV64XTHEADVDOT %s
 ; RUN: llc -mtriple=riscv64 -mattr=+za64rs %s -o - | FileCheck --check-prefixes=CHECK,RV64ZA64RS %s
 ; RUN: llc -mtriple=riscv64 -mattr=+za128rs %s -o - | FileCheck --check-prefixes=CHECK,RV64ZA128RS %s
+; RUN: llc -mtriple=riscv64 -mattr=+zama16b %s -o - | FileCheck --check-prefixes=CHECK,RV64ZAMA16B %s
 ; RUN: llc -mtriple=riscv64 -mattr=+zawrs %s -o - | FileCheck --check-prefixes=CHECK,RV64ZAWRS %s
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-ztso %s -o - | FileCheck --check-prefixes=CHECK,RV64ZTSO %s
 ; RUN: llc -mtriple=riscv64 -mattr=+zca %s -o - | FileCheck --check-prefixes=CHECK,RV64ZCA %s
@@ -370,6 +372,7 @@
 ; RV32ZACAS: .attribute 5, "rv32i2p1_a2p1_zacas1p0"
 ; RV32ZALASR: .attribute 5, "rv32i2p1_zalasr0p1"
 ; RV32ZALRSC: .attribute 5, "rv32i2p1_zalrsc0p2"
+; RV32ZAMA16B: .attribute 5, "rv32i2p1_zama16b1p0"
 ; RV32ZICFILP: .attribute 5, "rv32i2p1_zicfilp0p4"
 ; RV32ZABHA: .attribute 5, "rv32i2p1_a2p1_zabha1p0"
 ; RV32SSNPM: .attribute 5, "rv32i2p1_ssnpm0p8"
@@ -418,6 +421,7 @@
 ; RV64ZICBOZ: .attribute 5, "rv64i2p1_zicboz1p0"
 ; RV64ZA64RS: .attribute 5, "rv64i2p1_za64rs1p0"
 ; RV64ZA128RS: .attribute 5, "rv64i2p1_za128rs1p0"
+; RV64ZAMA16B: .attribute 5, "rv64i2p1_zama16b1p0"
 ; RV64ZAWRS: .attribute 5, "rv64i2p1_zawrs1p0"
 ; RV64ZICBOP: .attribute 5, "rv64i2p1_zicbop1p0"
 ; RV64SHCOUNTERENW: .attribute 5, "rv64i2p1_shcounterenw1p0"
diff --git a/llvm/test/MC/RISCV/attribute-arch.s b/llvm/test/MC/RISCV/attribute-arch.s
index a8f493f781ec3..8835ff22446c8 100644
--- a/llvm/test/MC/RISCV/attribute-arch.s
+++ b/llvm/test/MC/RISCV/attribute-arch.s
@@ -270,6 +270,9 @@
 .attribute arch, "rv32iza64rs1p0"
 # CHECK: attribute      5, "rv32i2p1_za64rs1p0"
 
+.attribute arch, "rv32izama16b"
+# CHECK: attribute      5, "rv32i2p1_zama16b1p0"
+
 .attribute arch, "rv32izawrs1p0"
 # CHECK: attribute      5, "rv32i2p1_zawrs1p0"
 
diff --git a/llvm/unittests/Support/RISCVISAInfoTest.cpp b/llvm/unittests/Support/RISCVISAInfoTest.cpp
index 67012d2e6dc72..caf7bf0a31717 100644
--- a/llvm/unittests/Support/RISCVISAInfoTest.cpp
+++ b/llvm/unittests/Support/RISCVISAInfoTest.cpp
@@ -769,6 +769,7 @@ R"(All available -march extensions for RISC-V
     za128rs              1.0
     za64rs               1.0
     zacas                1.0
+    zama16b              1.0
     zawrs                1.0
     zfa                  1.0
     zfh                  1.0

From d35a64363bb851045387717d2ef7d6449b7b547f Mon Sep 17 00:00:00 2001
From: Mikhail Goncharov <goncharov.mikhail@gmail.com>
Date: Wed, 17 Apr 2024 08:33:07 +0200
Subject: [PATCH 702/886] Revert "Fix test from #83124 and #88902"

This reverts commit 0a789ea8a829da345e46d8224d73b2ddaba6969f.

Breaks builds, see discussion in https://github.com/llvm/llvm-project/pull/83124
---
 clang/test/SemaCXX/PR41441.cpp | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/clang/test/SemaCXX/PR41441.cpp b/clang/test/SemaCXX/PR41441.cpp
index d0f2917e52f21..0b012b33fce34 100644
--- a/clang/test/SemaCXX/PR41441.cpp
+++ b/clang/test/SemaCXX/PR41441.cpp
@@ -1,9 +1,6 @@
 // RUN: %clang --target=x86_64-pc-linux -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
 
-namespace std {
-  using size_t = decltype(sizeof(int));
-};
-void* operator new[](std::size_t, void*) noexcept;
+#include <new>
 
 // CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)
 // CHECK: call void @llvm.memset.p0.i64(ptr align 16 %x, i8 0, i64 32, i1 false)

From dbda478693104f78b142375862d66f3369ad8c78 Mon Sep 17 00:00:00 2001
From: Mikhail Goncharov <goncharov.mikhail@gmail.com>
Date: Wed, 17 Apr 2024 08:33:44 +0200
Subject: [PATCH 703/886] Revert "[Clang][Sema] placement new initializes
 typedef array with correct size (#88902)"

This reverts commit 5c6af605b307213453a9a043532b9293db21b5c6.

Breaks builds, see discussion in https://github.com/llvm/llvm-project/pull/83124
---
 .../{PR41441.cpp => instantiate-new-placement-size.cpp}       | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
 rename clang/test/SemaCXX/{PR41441.cpp => instantiate-new-placement-size.cpp} (75%)

diff --git a/clang/test/SemaCXX/PR41441.cpp b/clang/test/SemaCXX/instantiate-new-placement-size.cpp
similarity index 75%
rename from clang/test/SemaCXX/PR41441.cpp
rename to clang/test/SemaCXX/instantiate-new-placement-size.cpp
index 0b012b33fce34..7a29d3dee8491 100644
--- a/clang/test/SemaCXX/PR41441.cpp
+++ b/clang/test/SemaCXX/instantiate-new-placement-size.cpp
@@ -1,5 +1,5 @@
-// RUN: %clang --target=x86_64-pc-linux -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
-
+// RUN: %clang -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
+// Issue no: 41441
 #include <new>
 
 // CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)

From dd84d23adc84cc0c3d2b8fb8f0c353279d99d27a Mon Sep 17 00:00:00 2001
From: Mikhail Goncharov <goncharov.mikhail@gmail.com>
Date: Wed, 17 Apr 2024 08:36:29 +0200
Subject: [PATCH 704/886] Revert "[Clang][Sema] placement new initializes
 typedef array with correct size (#83124)"

This reverts commit c309dc6d0759b23b570c563f611530ff1a49e1bd.

Breaks builds, see discussion in https://github.com/llvm/llvm-project/pull/83124
---
 clang/docs/ReleaseNotes.rst                   |  1 -
 clang/lib/Sema/TreeTransform.h                | 14 +------------
 .../instantiate-new-placement-size.cpp        | 20 -------------------
 3 files changed, 1 insertion(+), 34 deletions(-)
 delete mode 100644 clang/test/SemaCXX/instantiate-new-placement-size.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 6099f8ab02f44..96ad92b540b47 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -534,7 +534,6 @@ Bug Fixes to C++ Support
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
 - Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
 - Fix a crash in requires expression with templated base class member function. Fixes (#GH84020).
-- Placement new initializes typedef array with correct size (#GH41441)
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 0c7fdb357235e..eb05783a6219d 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -12864,19 +12864,6 @@ TreeTransform<Derived>::TransformCXXNewExpr(CXXNewExpr *E) {
     ArraySize = NewArraySize.get();
   }
 
-  // Per C++0x [expr.new]p5, the type being constructed may be a
-  // typedef of an array type.
-  QualType AllocType = AllocTypeInfo->getType();
-  if (ArraySize) {
-    if (const ConstantArrayType *Array =
-            SemaRef.Context.getAsConstantArrayType(AllocType)) {
-      ArraySize = IntegerLiteral::Create(SemaRef.Context, Array->getSize(),
-                                         SemaRef.Context.getSizeType(),
-                                         E->getBeginLoc());
-      AllocType = Array->getElementType();
-    }
-  }
-
   // Transform the placement arguments (if any).
   bool ArgumentChanged = false;
   SmallVector<Expr*, 8> PlacementArgs;
@@ -12938,6 +12925,7 @@ TreeTransform<Derived>::TransformCXXNewExpr(CXXNewExpr *E) {
     return E;
   }
 
+  QualType AllocType = AllocTypeInfo->getType();
   if (!ArraySize) {
     // If no array size was specified, but the new expression was
     // instantiated with an array type (e.g., "new T" where T is
diff --git a/clang/test/SemaCXX/instantiate-new-placement-size.cpp b/clang/test/SemaCXX/instantiate-new-placement-size.cpp
deleted file mode 100644
index 7a29d3dee8491..0000000000000
--- a/clang/test/SemaCXX/instantiate-new-placement-size.cpp
+++ /dev/null
@@ -1,20 +0,0 @@
-// RUN: %clang -S -fno-discard-value-names -emit-llvm -o - %s | FileCheck %s
-// Issue no: 41441
-#include <new>
-
-// CHECK: call void @llvm.memset.p0.i64(ptr align 1 %x, i8 0, i64 8, i1 false)
-// CHECK: call void @llvm.memset.p0.i64(ptr align 16 %x, i8 0, i64 32, i1 false)
-template <typename TYPE>
-void f()
-{
-    typedef TYPE TArray[8];
-
-    TArray x;
-    new(&x) TArray();
-}
-
-int main()
-{
-    f<char>();
-    f<int>();
-}

From bc3620d3a8b1be9534a5635431b0aa09cc50ff3c Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Wed, 17 Apr 2024 08:50:14 +0200
Subject: [PATCH 705/886] AMDGPU: Move libcall simplify into PeepholeEP
 (#88853)

We were running this immediately on the incoming IR, which
is still littered with temporary allocas obscuring trivial values.
This needs to run after initial SROA to handle sincos insertion.
---
 .../lib/Target/AMDGPU/AMDGPUTargetMachine.cpp | 13 +++-
 .../amdgpu-libcall-sincos-pass-ordering.ll    | 77 +++++++++++++++++++
 .../AMDGPU/amdgpu-simplify-libcall-sincos.ll  | 17 ++--
 llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll |  6 +-
 4 files changed, 96 insertions(+), 17 deletions(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/amdgpu-libcall-sincos-pass-ordering.ll

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index f7e552177d6f5..305a6c8c3b926 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -655,9 +655,6 @@ void AMDGPUTargetMachine::registerPassBuilderCallbacks(
   PB.registerPipelineStartEPCallback(
       [](ModulePassManager &PM, OptimizationLevel Level) {
         FunctionPassManager FPM;
-        FPM.addPass(AMDGPUUseNativeCallsPass());
-        if (EnableLibCallSimplify && Level != OptimizationLevel::O0)
-          FPM.addPass(AMDGPUSimplifyLibCallsPass());
         PM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
         if (EnableHipStdPar)
           PM.addPass(HipStdParAcceleratorCodeSelectionPass());
@@ -681,6 +678,16 @@ void AMDGPUTargetMachine::registerPassBuilderCallbacks(
           PM.addPass(AMDGPUAlwaysInlinePass());
       });
 
+  PB.registerPeepholeEPCallback(
+      [](FunctionPassManager &FPM, OptimizationLevel Level) {
+        if (Level == OptimizationLevel::O0)
+          return;
+
+        FPM.addPass(AMDGPUUseNativeCallsPass());
+        if (EnableLibCallSimplify)
+          FPM.addPass(AMDGPUSimplifyLibCallsPass());
+      });
+
   PB.registerCGSCCOptimizerLateEPCallback(
       [this](CGSCCPassManager &PM, OptimizationLevel Level) {
         if (Level == OptimizationLevel::O0)
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-libcall-sincos-pass-ordering.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-libcall-sincos-pass-ordering.ll
new file mode 100644
index 0000000000000..6b835bb4eef66
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-libcall-sincos-pass-ordering.ll
@@ -0,0 +1,77 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -O1 -amdgpu-prelink %s | FileCheck %s
+
+; Make sure that sin+cos -> sincos simplification happens after
+; initial IR simplifications, otherwise we can't identify the common
+; argument value.
+
+@.str = private unnamed_addr addrspace(4) constant [21 x i8] c"x: %f, y: %f, z: %f\0A\00", align 1
+
+; Should have call to sincos declarations, not calls to the asm pseudo-libcalls
+define protected amdgpu_kernel void @swdev456865(ptr addrspace(1) %out0, ptr addrspace(1) %out1, ptr addrspace(1) %out2, float noundef %x) #0 {
+; CHECK-LABEL: define protected amdgpu_kernel void @swdev456865(
+; CHECK-SAME: ptr addrspace(1) nocapture writeonly [[OUT0:%.*]], ptr addrspace(1) nocapture writeonly [[OUT1:%.*]], ptr addrspace(1) nocapture writeonly [[OUT2:%.*]], float noundef [[X:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
+; CHECK-NEXT:    [[I_I:%.*]] = call float @_Z6sincosfPU3AS5f(float [[X]], ptr addrspace(5) [[__SINCOS_]]) #[[ATTR1:[0-9]+]]
+; CHECK-NEXT:    [[I_I2:%.*]] = load float, ptr addrspace(5) [[__SINCOS_]], align 4
+; CHECK-NEXT:    [[ADD:%.*]] = fadd float [[I_I]], [[I_I2]]
+; CHECK-NEXT:    [[CONV:%.*]] = fpext float [[X]] to double
+; CHECK-NEXT:    [[CONV5:%.*]] = fpext float [[ADD]] to double
+; CHECK-NEXT:    store double [[CONV]], ptr addrspace(1) [[OUT0]], align 8
+; CHECK-NEXT:    store double [[CONV5]], ptr addrspace(1) [[OUT1]], align 8
+; CHECK-NEXT:    store double [[CONV5]], ptr addrspace(1) [[OUT2]], align 8
+; CHECK-NEXT:    ret void
+;
+entry:
+  %x.addr = alloca float, align 4, addrspace(5)
+  %y = alloca float, align 4, addrspace(5)
+  %z = alloca float, align 4, addrspace(5)
+  store float %x, ptr addrspace(5) %x.addr, align 4
+  call void @llvm.lifetime.start.p5(i64 4, ptr addrspace(5) %y)
+  %i = load float, ptr addrspace(5) %x.addr, align 4
+  %call = call float @_Z3sinf(float noundef %i) #3
+  %i1 = load float, ptr addrspace(5) %x.addr, align 4
+  %call1 = call float @_Z3cosf(float noundef %i1) #3
+  %add = fadd float %call, %call1
+  store float %add, ptr addrspace(5) %y, align 4
+  call void @llvm.lifetime.start.p5(i64 4, ptr addrspace(5) %z)
+  %i2 = load float, ptr addrspace(5) %x.addr, align 4
+  %call2 = call float @_Z3cosf(float noundef %i2) #3
+  %i3 = load float, ptr addrspace(5) %x.addr, align 4
+  %call3 = call float @_Z3sinf(float noundef %i3) #3
+  %add4 = fadd float %call2, %call3
+  store float %add4, ptr addrspace(5) %z, align 4
+  %i4 = load float, ptr addrspace(5) %x.addr, align 4
+  %conv = fpext float %i4 to double
+  %i5 = load float, ptr addrspace(5) %y, align 4
+  %conv5 = fpext float %i5 to double
+  %i6 = load float, ptr addrspace(5) %z, align 4
+  %conv6 = fpext float %i6 to double
+  store double %conv, ptr addrspace(1) %out0, align 8
+  store double %conv5, ptr addrspace(1) %out1, align 8
+  store double %conv6, ptr addrspace(1) %out2, align 8
+  call void @llvm.lifetime.end.p5(i64 4, ptr addrspace(5) %z)
+  call void @llvm.lifetime.end.p5(i64 4, ptr addrspace(5) %y)
+  ret void
+}
+
+declare void @llvm.lifetime.start.p5(i64 immarg, ptr addrspace(5) nocapture) #1
+declare void @llvm.lifetime.end.p5(i64 immarg, ptr addrspace(5) nocapture) #1
+
+define internal float @_Z3cosf(float noundef %arg) #2 {
+bb:
+  %i = tail call float asm "pseudo-libcall-cos %0, %1", "=v,v"(float noundef %arg) #2
+  ret float %i
+}
+
+define internal float @_Z3sinf(float noundef %arg) #2 {
+bb:
+  %i = tail call float asm "pseudo-libcall-sin %0, %1", "=v,v"(float noundef %arg) #2
+  ret float %i
+}
+
+attributes #0 = { norecurse nounwind }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { mustprogress nofree norecurse nounwind willreturn memory(none) }
+attributes #3 = { nounwind willreturn memory(none) }
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
index 5c56276eeb0f1..9646d196da42f 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
@@ -884,10 +884,9 @@ entry:
 
 define float @sincos_f32_unused_result_cos(float %x) {
 ; CHECK-LABEL: define float @sincos_f32_unused_result_cos
-; CHECK-SAME: (float [[X:%.*]]) local_unnamed_addr #[[ATTR4]] {
+; CHECK-SAME: (float [[X:%.*]]) local_unnamed_addr #[[ATTR5:[0-9]+]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
-; CHECK-NEXT:    [[TMP0:%.*]] = call contract float @_Z6sincosfPU3AS5f(float [[X]], ptr addrspace(5) [[__SINCOS_]])
+; CHECK-NEXT:    [[TMP0:%.*]] = tail call contract float @_Z3sinf(float [[X]])
 ; CHECK-NEXT:    ret float [[TMP0]]
 ;
 entry:
@@ -900,11 +899,9 @@ entry:
 
 define float @sincos_f32_unused_result_sin(float %x) {
 ; CHECK-LABEL: define float @sincos_f32_unused_result_sin
-; CHECK-SAME: (float [[X:%.*]]) local_unnamed_addr #[[ATTR4]] {
+; CHECK-SAME: (float [[X:%.*]]) local_unnamed_addr #[[ATTR5]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
-; CHECK-NEXT:    [[TMP0:%.*]] = call contract float @_Z6sincosfPU3AS5f(float [[X]], ptr addrspace(5) [[__SINCOS_]])
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(5) [[__SINCOS_]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call contract float @_Z3cosf(float [[X]])
 ; CHECK-NEXT:    ret float [[TMP1]]
 ;
 entry:
@@ -917,13 +914,11 @@ entry:
 
 define void @sincos_f32_repeated_uses(float %x, ptr addrspace(1) %sin_out, ptr addrspace(1) %cos_out) {
 ; CHECK-LABEL: define void @sincos_f32_repeated_uses
-; CHECK-SAME: (float [[X:%.*]], ptr addrspace(1) [[SIN_OUT:%.*]], ptr addrspace(1) [[COS_OUT:%.*]]) local_unnamed_addr #[[ATTR5:[0-9]+]] {
+; CHECK-SAME: (float [[X:%.*]], ptr addrspace(1) [[SIN_OUT:%.*]], ptr addrspace(1) [[COS_OUT:%.*]]) local_unnamed_addr #[[ATTR6:[0-9]+]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
 ; CHECK-NEXT:    [[__SINCOS_3:%.*]] = alloca float, align 4, addrspace(5)
 ; CHECK-NEXT:    [[TMP0:%.*]] = call contract float @_Z6sincosfPU3AS5f(float [[X]], ptr addrspace(5) [[__SINCOS_3]])
-; CHECK-NEXT:    [[TMP1:%.*]] = call contract float @_Z6sincosfPU3AS5f(float [[X]], ptr addrspace(5) [[__SINCOS_]])
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr addrspace(5) [[__SINCOS_]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr addrspace(5) [[__SINCOS_3]], align 4
 ; CHECK-NEXT:    store volatile float [[TMP0]], ptr addrspace(1) [[SIN_OUT]], align 4
 ; CHECK-NEXT:    store volatile float [[TMP0]], ptr addrspace(1) [[SIN_OUT]], align 4
 ; CHECK-NEXT:    store volatile float [[TMP2]], ptr addrspace(1) [[COS_OUT]], align 4
diff --git a/llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll b/llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
index 731a88278e512..204c8140d3f17 100644
--- a/llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
+++ b/llvm/test/CodeGen/AMDGPU/simplify-libcalls.ll
@@ -278,7 +278,7 @@ entry:
 
 ; GCN-LABEL: {{^}}define amdgpu_kernel void @test_pow_half
 ; GCN-POSTLINK: call fast float @_Z3powff(float %tmp, float 5.000000e-01)
-; GCN-PRELINK: %__pow2sqrt = tail call fast float @_Z4sqrtf(float %tmp)
+; GCN-PRELINK: %__pow2sqrt = tail call fast float @llvm.sqrt.f32(float %tmp)
 define amdgpu_kernel void @test_pow_half(ptr addrspace(1) nocapture %a) {
 entry:
   %arrayidx = getelementptr inbounds float, ptr addrspace(1) %a, i64 1
@@ -476,7 +476,7 @@ declare float @_Z5rootnfi(float, i32)
 
 ; GCN-LABEL: {{^}}define amdgpu_kernel void @test_rootn_2
 ; GCN-POSTLINK: call fast float @_Z5rootnfi(float %tmp, i32 2)
-; GCN-PRELINK: %__rootn2sqrt = tail call fast float @_Z4sqrtf(float %tmp)
+; GCN-PRELINK: %__rootn2sqrt = tail call fast float @llvm.sqrt.f32(float %tmp)
 define amdgpu_kernel void @test_rootn_2(ptr addrspace(1) nocapture %a) {
 entry:
   %tmp = load float, ptr addrspace(1) %a, align 4
@@ -838,5 +838,5 @@ entry:
 ; GCN-PRELINK: declare float @_Z4cbrtf(float) local_unnamed_addr #[[$NOUNWIND_READONLY:[0-9]+]]
 
 ; GCN-PRELINK-DAG: attributes #[[$NOUNWIND]] = { nounwind }
-; GCN-PRELINK-DAG: attributes #[[$NOUNWIND_READONLY]] = { nofree nounwind memory(read) }
+; GCN-PRELINK-DAG: attributes #[[$NOUNWIND_READONLY]] = { nounwind memory(read) }
 attributes #0 = { nounwind }

From e11b17a4ed90e74147594012207fc35a60515944 Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Wed, 17 Apr 2024 09:51:24 +0300
Subject: [PATCH 706/886] [clang][NFC] Refactor `Sema::CheckedConversionKind`

Convert it to scoped enum, and move it to namespace scope to enable forward declarations.
---
 clang/include/clang/Sema/Sema.h     | 65 +++++++++++++++----------
 clang/lib/Sema/SemaCast.cpp         | 73 ++++++++++++++---------------
 clang/lib/Sema/SemaExpr.cpp         | 10 ++--
 clang/lib/Sema/SemaExprCXX.cpp      |  8 ++--
 clang/lib/Sema/SemaExprObjC.cpp     | 45 +++++++++---------
 clang/lib/Sema/SemaInit.cpp         | 10 ++--
 clang/lib/Sema/SemaOverload.cpp     | 10 ++--
 clang/lib/Sema/SemaPseudoObject.cpp |  2 +-
 8 files changed, 119 insertions(+), 104 deletions(-)

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 091c5c02f75df..281e3b91de1d0 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -437,6 +437,20 @@ enum class CXXSpecialMemberKind {
   Invalid
 };
 
+/// The kind of conversion being performed.
+enum class CheckedConversionKind {
+  /// An implicit conversion.
+  Implicit,
+  /// A C-style cast.
+  CStyleCast,
+  /// A functional-style cast.
+  FunctionalCast,
+  /// A cast other than a C-style cast.
+  OtherCast,
+  /// A conversion for an operand of a builtin overloaded operator.
+  ForBuiltinOverloadedOp
+};
+
 /// Sema - This implements semantic analysis and AST building for C.
 /// \nosubgrouping
 class Sema final : public SemaBase {
@@ -700,28 +714,27 @@ class Sema final : public SemaBase {
   void checkTypeSupport(QualType Ty, SourceLocation Loc,
                         ValueDecl *D = nullptr);
 
-  /// The kind of conversion being performed.
-  enum CheckedConversionKind {
-    /// An implicit conversion.
-    CCK_ImplicitConversion,
-    /// A C-style cast.
-    CCK_CStyleCast,
-    /// A functional-style cast.
-    CCK_FunctionalCast,
-    /// A cast other than a C-style cast.
-    CCK_OtherCast,
-    /// A conversion for an operand of a builtin overloaded operator.
-    CCK_ForBuiltinOverloadedOp
-  };
+  // /// The kind of conversion being performed.
+  // enum CheckedConversionKind {
+  //   /// An implicit conversion.
+  //   CCK_ImplicitConversion,
+  //   /// A C-style cast.
+  //   CCK_CStyleCast,
+  //   /// A functional-style cast.
+  //   CCK_FunctionalCast,
+  //   /// A cast other than a C-style cast.
+  //   CCK_OtherCast,
+  //   /// A conversion for an operand of a builtin overloaded operator.
+  //   CCK_ForBuiltinOverloadedOp
+  // };
 
   /// ImpCastExprToType - If Expr is not of type 'Type', insert an implicit
   /// cast.  If there is already an implicit cast, merge into the existing one.
   /// If isLvalue, the result of the cast is an lvalue.
-  ExprResult
-  ImpCastExprToType(Expr *E, QualType Type, CastKind CK,
-                    ExprValueKind VK = VK_PRValue,
-                    const CXXCastPath *BasePath = nullptr,
-                    CheckedConversionKind CCK = CCK_ImplicitConversion);
+  ExprResult ImpCastExprToType(
+      Expr *E, QualType Type, CastKind CK, ExprValueKind VK = VK_PRValue,
+      const CXXCastPath *BasePath = nullptr,
+      CheckedConversionKind CCK = CheckedConversionKind::Implicit);
 
   /// ScalarTypeToBooleanCastKind - Returns the cast kind corresponding
   /// to the conversion from scalar type ScalarTy to the Boolean type.
@@ -1781,8 +1794,9 @@ class Sema final : public SemaBase {
 
 public:
   static bool isCast(CheckedConversionKind CCK) {
-    return CCK == CCK_CStyleCast || CCK == CCK_FunctionalCast ||
-           CCK == CCK_OtherCast;
+    return CCK == CheckedConversionKind::CStyleCast ||
+           CCK == CheckedConversionKind::FunctionalCast ||
+           CCK == CheckedConversionKind::OtherCast;
   }
 
   /// ActOnCXXNamedCast - Parse
@@ -6739,11 +6753,10 @@ class Sema final : public SemaBase {
 
   bool IsStringLiteralToNonConstPointerConversion(Expr *From, QualType ToType);
 
-  ExprResult
-  PerformImplicitConversion(Expr *From, QualType ToType,
-                            const ImplicitConversionSequence &ICS,
-                            AssignmentAction Action,
-                            CheckedConversionKind CCK = CCK_ImplicitConversion);
+  ExprResult PerformImplicitConversion(
+      Expr *From, QualType ToType, const ImplicitConversionSequence &ICS,
+      AssignmentAction Action,
+      CheckedConversionKind CCK = CheckedConversionKind::Implicit);
   ExprResult PerformImplicitConversion(Expr *From, QualType ToType,
                                        const StandardConversionSequence &SCS,
                                        AssignmentAction Action,
@@ -7064,7 +7077,7 @@ class Sema final : public SemaBase {
 
   ExprResult PerformQualificationConversion(
       Expr *E, QualType Ty, ExprValueKind VK = VK_PRValue,
-      CheckedConversionKind CCK = CCK_ImplicitConversion);
+      CheckedConversionKind CCK = CheckedConversionKind::Implicit);
 
   bool CanPerformCopyInitialization(const InitializedEntity &Entity,
                                     ExprResult Init);
diff --git a/clang/lib/Sema/SemaCast.cpp b/clang/lib/Sema/SemaCast.cpp
index b0c28531fe873..126fd3797417c 100644
--- a/clang/lib/Sema/SemaCast.cpp
+++ b/clang/lib/Sema/SemaCast.cpp
@@ -155,7 +155,7 @@ namespace {
       Self.CheckCastAlign(SrcExpr.get(), DestType, OpRange);
     }
 
-    void checkObjCConversion(Sema::CheckedConversionKind CCK) {
+    void checkObjCConversion(CheckedConversionKind CCK) {
       assert(Self.getLangOpts().allowsNonTrivialObjCLifetimeQualifiers());
 
       Expr *src = SrcExpr.get();
@@ -248,18 +248,14 @@ static TryCastResult TryStaticMemberPointerUpcast(Sema &Self, ExprResult &SrcExp
                                                CastKind &Kind,
                                                CXXCastPath &BasePath);
 
-static TryCastResult TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr,
-                                           QualType DestType,
-                                           Sema::CheckedConversionKind CCK,
-                                           SourceRange OpRange,
-                                           unsigned &msg, CastKind &Kind,
-                                           bool ListInitialization);
+static TryCastResult
+TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr, QualType DestType,
+                      CheckedConversionKind CCK, SourceRange OpRange,
+                      unsigned &msg, CastKind &Kind, bool ListInitialization);
 static TryCastResult TryStaticCast(Sema &Self, ExprResult &SrcExpr,
-                                   QualType DestType,
-                                   Sema::CheckedConversionKind CCK,
-                                   SourceRange OpRange,
-                                   unsigned &msg, CastKind &Kind,
-                                   CXXCastPath &BasePath,
+                                   QualType DestType, CheckedConversionKind CCK,
+                                   SourceRange OpRange, unsigned &msg,
+                                   CastKind &Kind, CXXCastPath &BasePath,
                                    bool ListInitialization);
 static TryCastResult TryConstCast(Sema &Self, ExprResult &SrcExpr,
                                   QualType DestType, bool CStyle,
@@ -1223,7 +1219,7 @@ void CastOperation::CheckReinterpretCast() {
 
   if (isValidCast(tcr)) {
     if (Self.getLangOpts().allowsNonTrivialObjCLifetimeQualifiers())
-      checkObjCConversion(Sema::CCK_OtherCast);
+      checkObjCConversion(CheckedConversionKind::OtherCast);
     DiagnoseReinterpretUpDownCast(Self, SrcExpr.get(), DestType, OpRange);
 
     if (unsigned DiagID = checkCastFunctionType(Self, SrcExpr, DestType))
@@ -1274,9 +1270,9 @@ void CastOperation::CheckStaticCast() {
   }
 
   unsigned msg = diag::err_bad_cxx_cast_generic;
-  TryCastResult tcr
-    = TryStaticCast(Self, SrcExpr, DestType, Sema::CCK_OtherCast, OpRange, msg,
-                    Kind, BasePath, /*ListInitialization=*/false);
+  TryCastResult tcr =
+      TryStaticCast(Self, SrcExpr, DestType, CheckedConversionKind::OtherCast,
+                    OpRange, msg, Kind, BasePath, /*ListInitialization=*/false);
   if (tcr != TC_Success && msg != 0) {
     if (SrcExpr.isInvalid())
       return;
@@ -1296,7 +1292,7 @@ void CastOperation::CheckStaticCast() {
     if (Kind == CK_BitCast)
       checkCastAlign();
     if (Self.getLangOpts().allowsNonTrivialObjCLifetimeQualifiers())
-      checkObjCConversion(Sema::CCK_OtherCast);
+      checkObjCConversion(CheckedConversionKind::OtherCast);
   } else {
     SrcExpr = ExprError();
   }
@@ -1317,14 +1313,13 @@ static bool IsAddressSpaceConversion(QualType SrcType, QualType DestType) {
 /// possible. If @p CStyle, ignore access restrictions on hierarchy casting
 /// and casting away constness.
 static TryCastResult TryStaticCast(Sema &Self, ExprResult &SrcExpr,
-                                   QualType DestType,
-                                   Sema::CheckedConversionKind CCK,
+                                   QualType DestType, CheckedConversionKind CCK,
                                    SourceRange OpRange, unsigned &msg,
                                    CastKind &Kind, CXXCastPath &BasePath,
                                    bool ListInitialization) {
   // Determine whether we have the semantics of a C-style cast.
-  bool CStyle
-    = (CCK == Sema::CCK_CStyleCast || CCK == Sema::CCK_FunctionalCast);
+  bool CStyle = (CCK == CheckedConversionKind::CStyleCast ||
+                 CCK == CheckedConversionKind::FunctionalCast);
 
   // The order the tests is not entirely arbitrary. There is one conversion
   // that can be handled in two different ways. Given:
@@ -1884,11 +1879,11 @@ TryStaticMemberPointerUpcast(Sema &Self, ExprResult &SrcExpr, QualType SrcType,
 ///
 ///   An expression e can be explicitly converted to a type T using a
 ///   @c static_cast if the declaration "T t(e);" is well-formed [...].
-TryCastResult
-TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr, QualType DestType,
-                      Sema::CheckedConversionKind CCK,
-                      SourceRange OpRange, unsigned &msg,
-                      CastKind &Kind, bool ListInitialization) {
+TryCastResult TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr,
+                                    QualType DestType,
+                                    CheckedConversionKind CCK,
+                                    SourceRange OpRange, unsigned &msg,
+                                    CastKind &Kind, bool ListInitialization) {
   if (DestType->isRecordType()) {
     if (Self.RequireCompleteType(OpRange.getBegin(), DestType,
                                  diag::err_bad_cast_incomplete) ||
@@ -1900,13 +1895,14 @@ TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr, QualType DestType,
   }
 
   InitializedEntity Entity = InitializedEntity::InitializeTemporary(DestType);
-  InitializationKind InitKind
-    = (CCK == Sema::CCK_CStyleCast)
-        ? InitializationKind::CreateCStyleCast(OpRange.getBegin(), OpRange,
-                                               ListInitialization)
-    : (CCK == Sema::CCK_FunctionalCast)
-        ? InitializationKind::CreateFunctionalCast(OpRange, ListInitialization)
-    : InitializationKind::CreateCast(OpRange);
+  InitializationKind InitKind =
+      (CCK == CheckedConversionKind::CStyleCast)
+          ? InitializationKind::CreateCStyleCast(OpRange.getBegin(), OpRange,
+                                                 ListInitialization)
+      : (CCK == CheckedConversionKind::FunctionalCast)
+          ? InitializationKind::CreateFunctionalCast(OpRange,
+                                                     ListInitialization)
+          : InitializationKind::CreateCast(OpRange);
   Expr *SrcExprRaw = SrcExpr.get();
   // FIXME: Per DR242, we should check for an implicit conversion sequence
   // or for a constructor that could be invoked by direct-initialization
@@ -1918,8 +1914,8 @@ TryStaticImplicitCast(Sema &Self, ExprResult &SrcExpr, QualType DestType,
   // There is no other way that works.
   // On the other hand, if we're checking a C-style cast, we've still got
   // the reinterpret_cast way.
-  bool CStyle
-    = (CCK == Sema::CCK_CStyleCast || CCK == Sema::CCK_FunctionalCast);
+  bool CStyle = (CCK == CheckedConversionKind::CStyleCast ||
+                 CCK == CheckedConversionKind::FunctionalCast);
   if (InitSeq.Failed() && (CStyle || !DestType->isReferenceType()))
     return TC_NotApplicable;
 
@@ -2814,8 +2810,9 @@ void CastOperation::CheckCXXCStyleCast(bool FunctionalStyle,
   if (isValidCast(tcr))
     Kind = CK_NoOp;
 
-  Sema::CheckedConversionKind CCK =
-      FunctionalStyle ? Sema::CCK_FunctionalCast : Sema::CCK_CStyleCast;
+  CheckedConversionKind CCK = FunctionalStyle
+                                  ? CheckedConversionKind::FunctionalCast
+                                  : CheckedConversionKind::CStyleCast;
   if (tcr == TC_NotApplicable) {
     tcr = TryAddressSpaceCast(Self, SrcExpr, DestType, /*CStyle*/ true, msg,
                               Kind);
@@ -3201,7 +3198,7 @@ void CastOperation::CheckCStyleCast() {
 
   // ARC imposes extra restrictions on casts.
   if (Self.getLangOpts().allowsNonTrivialObjCLifetimeQualifiers()) {
-    checkObjCConversion(Sema::CCK_CStyleCast);
+    checkObjCConversion(CheckedConversionKind::CStyleCast);
     if (SrcExpr.isInvalid())
       return;
 
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 7c3faba0f7881..d2c77ad61644f 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -10177,8 +10177,9 @@ Sema::CheckSingleAssignmentConstraints(QualType LHSType, ExprResult &CallerRHS,
     // diagnostics and just checking for errors, e.g., during overload
     // resolution, return Incompatible to indicate the failure.
     if (getLangOpts().allowsNonTrivialObjCLifetimeQualifiers() &&
-        CheckObjCConversion(SourceRange(), Ty, E, CCK_ImplicitConversion,
-                            Diagnose, DiagnoseCFAudited) != ACR_okay) {
+        CheckObjCConversion(SourceRange(), Ty, E,
+                            CheckedConversionKind::Implicit, Diagnose,
+                            DiagnoseCFAudited) != ACR_okay) {
       if (!Diagnose)
         return Incompatible;
     }
@@ -12899,14 +12900,15 @@ QualType Sema::CheckCompareOperands(ExprResult &LHS, ExprResult &RHS,
         Expr *E = LHS.get();
         if (getLangOpts().ObjCAutoRefCount)
           CheckObjCConversion(SourceRange(), RHSType, E,
-                              CCK_ImplicitConversion);
+                              CheckedConversionKind::Implicit);
         LHS = ImpCastExprToType(E, RHSType,
                                 RPT ? CK_BitCast :CK_CPointerToObjCPointerCast);
       }
       else {
         Expr *E = RHS.get();
         if (getLangOpts().ObjCAutoRefCount)
-          CheckObjCConversion(SourceRange(), LHSType, E, CCK_ImplicitConversion,
+          CheckObjCConversion(SourceRange(), LHSType, E,
+                              CheckedConversionKind::Implicit,
                               /*Diagnose=*/true,
                               /*DiagnoseCFAudited=*/false, Opc);
         RHS = ImpCastExprToType(E, LHSType,
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index 74ed3fe7bd520..f4a91ececfbb5 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -4250,7 +4250,8 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
                                 AssignmentAction Action,
                                 CheckedConversionKind CCK) {
   // C++ [over.match.oper]p7: [...] operands of class type are converted [...]
-  if (CCK == CCK_ForBuiltinOverloadedOp && !From->getType()->isRecordType())
+  if (CCK == CheckedConversionKind::ForBuiltinOverloadedOp &&
+      !From->getType()->isRecordType())
     return From;
 
   switch (ICS.getKind()) {
@@ -4311,7 +4312,7 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
       // C++ [over.match.oper]p7:
       //   [...] the second standard conversion sequence of a user-defined
       //   conversion sequence is not applied.
-      if (CCK == CCK_ForBuiltinOverloadedOp)
+      if (CCK == CheckedConversionKind::ForBuiltinOverloadedOp)
         return From;
 
       return PerformImplicitConversion(From, ToType, ICS.UserDefined.After,
@@ -4352,7 +4353,8 @@ Sema::PerformImplicitConversion(Expr *From, QualType ToType,
                                 const StandardConversionSequence& SCS,
                                 AssignmentAction Action,
                                 CheckedConversionKind CCK) {
-  bool CStyle = (CCK == CCK_CStyleCast || CCK == CCK_FunctionalCast);
+  bool CStyle = (CCK == CheckedConversionKind::CStyleCast ||
+                 CCK == CheckedConversionKind::FunctionalCast);
 
   // Overall FIXME: we are recomputing too many types here and doing far too
   // much extra work. What this means is that we need to keep track of more
diff --git a/clang/lib/Sema/SemaExprObjC.cpp b/clang/lib/Sema/SemaExprObjC.cpp
index 3148f0db6e20c..b13a9d426983b 100644
--- a/clang/lib/Sema/SemaExprObjC.cpp
+++ b/clang/lib/Sema/SemaExprObjC.cpp
@@ -3745,22 +3745,22 @@ bool Sema::isKnownName(StringRef name) {
 
 template <typename DiagBuilderT>
 static void addFixitForObjCARCConversion(
-    Sema &S, DiagBuilderT &DiagB, Sema::CheckedConversionKind CCK,
+    Sema &S, DiagBuilderT &DiagB, CheckedConversionKind CCK,
     SourceLocation afterLParen, QualType castType, Expr *castExpr,
     Expr *realCast, const char *bridgeKeyword, const char *CFBridgeName) {
   // We handle C-style and implicit casts here.
   switch (CCK) {
-  case Sema::CCK_ImplicitConversion:
-  case Sema::CCK_ForBuiltinOverloadedOp:
-  case Sema::CCK_CStyleCast:
-  case Sema::CCK_OtherCast:
+  case CheckedConversionKind::Implicit:
+  case CheckedConversionKind::ForBuiltinOverloadedOp:
+  case CheckedConversionKind::CStyleCast:
+  case CheckedConversionKind::OtherCast:
     break;
-  case Sema::CCK_FunctionalCast:
+  case CheckedConversionKind::FunctionalCast:
     return;
   }
 
   if (CFBridgeName) {
-    if (CCK == Sema::CCK_OtherCast) {
+    if (CCK == CheckedConversionKind::OtherCast) {
       if (const CXXNamedCastExpr *NCE = dyn_cast<CXXNamedCastExpr>(realCast)) {
         SourceRange range(NCE->getOperatorLoc(),
                           NCE->getAngleBrackets().getEnd());
@@ -3805,9 +3805,9 @@ static void addFixitForObjCARCConversion(
     return;
   }
 
-  if (CCK == Sema::CCK_CStyleCast) {
+  if (CCK == CheckedConversionKind::CStyleCast) {
     DiagB.AddFixItHint(FixItHint::CreateInsertion(afterLParen, bridgeKeyword));
-  } else if (CCK == Sema::CCK_OtherCast) {
+  } else if (CCK == CheckedConversionKind::OtherCast) {
     if (const CXXNamedCastExpr *NCE = dyn_cast<CXXNamedCastExpr>(realCast)) {
       std::string castCode = "(";
       castCode += bridgeKeyword;
@@ -3866,12 +3866,12 @@ static ObjCBridgeRelatedAttr *ObjCBridgeRelatedAttrFromType(QualType T,
   return nullptr;
 }
 
-static void
-diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
-                          QualType castType, ARCConversionTypeClass castACTC,
-                          Expr *castExpr, Expr *realCast,
-                          ARCConversionTypeClass exprACTC,
-                          Sema::CheckedConversionKind CCK) {
+static void diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
+                                      QualType castType,
+                                      ARCConversionTypeClass castACTC,
+                                      Expr *castExpr, Expr *realCast,
+                                      ARCConversionTypeClass exprACTC,
+                                      CheckedConversionKind CCK) {
   SourceLocation loc =
     (castRange.isValid() ? castRange.getBegin() : castExpr->getExprLoc());
 
@@ -3927,7 +3927,7 @@ diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
     assert(CreateRule != ACC_bottom && "This cast should already be accepted.");
     if (CreateRule != ACC_plusOne)
     {
-      auto DiagB = (CCK != Sema::CCK_OtherCast)
+      auto DiagB = (CCK != CheckedConversionKind::OtherCast)
                        ? S.Diag(noteLoc, diag::note_arc_bridge)
                        : S.Diag(noteLoc, diag::note_arc_cstyle_bridge);
 
@@ -3937,7 +3937,7 @@ diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
     }
     if (CreateRule != ACC_plusZero)
     {
-      auto DiagB = (CCK == Sema::CCK_OtherCast && !br)
+      auto DiagB = (CCK == CheckedConversionKind::OtherCast && !br)
                        ? S.Diag(noteLoc, diag::note_arc_cstyle_bridge_transfer)
                              << castExprType
                        : S.Diag(br ? castExpr->getExprLoc() : noteLoc,
@@ -3968,7 +3968,7 @@ diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
     assert(CreateRule != ACC_bottom && "This cast should already be accepted.");
     if (CreateRule != ACC_plusOne)
     {
-      auto DiagB = (CCK != Sema::CCK_OtherCast)
+      auto DiagB = (CCK != CheckedConversionKind::OtherCast)
                        ? S.Diag(noteLoc, diag::note_arc_bridge)
                        : S.Diag(noteLoc, diag::note_arc_cstyle_bridge);
       addFixitForObjCARCConversion(S, DiagB, CCK, afterLParen,
@@ -3977,7 +3977,7 @@ diagnoseObjCARCConversion(Sema &S, SourceRange castRange,
     }
     if (CreateRule != ACC_plusZero)
     {
-      auto DiagB = (CCK == Sema::CCK_OtherCast && !br)
+      auto DiagB = (CCK == CheckedConversionKind::OtherCast && !br)
                        ? S.Diag(noteLoc, diag::note_arc_cstyle_bridge_retained)
                              << castType
                        : S.Diag(br ? castExpr->getExprLoc() : noteLoc,
@@ -4403,7 +4403,8 @@ Sema::CheckObjCConversion(SourceRange castRange, QualType castType,
     // Check for viability and report error if casting an rvalue to a
     // life-time qualifier.
     if (castACTC == ACTC_retainable &&
-        (CCK == CCK_CStyleCast || CCK == CCK_OtherCast) &&
+        (CCK == CheckedConversionKind::CStyleCast ||
+         CCK == CheckedConversionKind::OtherCast) &&
         castType != castExprType) {
       const Type *DT = castType.getTypePtr();
       QualType QDT = castType;
@@ -4517,11 +4518,11 @@ void Sema::diagnoseARCUnbridgedCast(Expr *e) {
   if (CStyleCastExpr *cast = dyn_cast<CStyleCastExpr>(realCast)) {
     castRange = SourceRange(cast->getLParenLoc(), cast->getRParenLoc());
     castType = cast->getTypeAsWritten();
-    CCK = CCK_CStyleCast;
+    CCK = CheckedConversionKind::CStyleCast;
   } else if (ExplicitCastExpr *cast = dyn_cast<ExplicitCastExpr>(realCast)) {
     castRange = cast->getTypeInfoAsWritten()->getTypeLoc().getSourceRange();
     castType = cast->getTypeAsWritten();
-    CCK = CCK_OtherCast;
+    CCK = CheckedConversionKind::OtherCast;
   } else {
     llvm_unreachable("Unexpected ImplicitCastExpr");
   }
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index fb7a80ab02846..e86f7578ff0c0 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -9057,11 +9057,11 @@ ExprResult InitializationSequence::Perform(Sema &S,
         }
       }
 
-      Sema::CheckedConversionKind CCK
-        = Kind.isCStyleCast()? Sema::CCK_CStyleCast
-        : Kind.isFunctionalCast()? Sema::CCK_FunctionalCast
-        : Kind.isExplicitCast()? Sema::CCK_OtherCast
-        : Sema::CCK_ImplicitConversion;
+      CheckedConversionKind CCK =
+          Kind.isCStyleCast()       ? CheckedConversionKind::CStyleCast
+          : Kind.isFunctionalCast() ? CheckedConversionKind::FunctionalCast
+          : Kind.isExplicitCast()   ? CheckedConversionKind::OtherCast
+                                    : CheckedConversionKind::Implicit;
       ExprResult CurInitExprRes =
         S.PerformImplicitConversion(CurInit.get(), Step->Type, *Step->ICS,
                                     getAssignmentAction(Entity), CCK);
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 227ef564ba3e0..adc319e97b762 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -14506,7 +14506,7 @@ Sema::CreateOverloadedUnaryOp(SourceLocation OpLoc, UnaryOperatorKind Opc,
       // operator node.
       ExprResult InputRes = PerformImplicitConversion(
           Input, Best->BuiltinParamTypes[0], Best->Conversions[0], AA_Passing,
-          CCK_ForBuiltinOverloadedOp);
+          CheckedConversionKind::ForBuiltinOverloadedOp);
       if (InputRes.isInvalid())
         return ExprError();
       Input = InputRes.get();
@@ -14989,14 +14989,14 @@ ExprResult Sema::CreateOverloadedBinOp(SourceLocation OpLoc,
         // operator node.
         ExprResult ArgsRes0 = PerformImplicitConversion(
             Args[0], Best->BuiltinParamTypes[0], Best->Conversions[0],
-            AA_Passing, CCK_ForBuiltinOverloadedOp);
+            AA_Passing, CheckedConversionKind::ForBuiltinOverloadedOp);
         if (ArgsRes0.isInvalid())
           return ExprError();
         Args[0] = ArgsRes0.get();
 
         ExprResult ArgsRes1 = PerformImplicitConversion(
             Args[1], Best->BuiltinParamTypes[1], Best->Conversions[1],
-            AA_Passing, CCK_ForBuiltinOverloadedOp);
+            AA_Passing, CheckedConversionKind::ForBuiltinOverloadedOp);
         if (ArgsRes1.isInvalid())
           return ExprError();
         Args[1] = ArgsRes1.get();
@@ -15367,14 +15367,14 @@ ExprResult Sema::CreateOverloadedArraySubscriptExpr(SourceLocation LLoc,
         // operator node.
         ExprResult ArgsRes0 = PerformImplicitConversion(
             Args[0], Best->BuiltinParamTypes[0], Best->Conversions[0],
-            AA_Passing, CCK_ForBuiltinOverloadedOp);
+            AA_Passing, CheckedConversionKind::ForBuiltinOverloadedOp);
         if (ArgsRes0.isInvalid())
           return ExprError();
         Args[0] = ArgsRes0.get();
 
         ExprResult ArgsRes1 = PerformImplicitConversion(
             Args[1], Best->BuiltinParamTypes[1], Best->Conversions[1],
-            AA_Passing, CCK_ForBuiltinOverloadedOp);
+            AA_Passing, CheckedConversionKind::ForBuiltinOverloadedOp);
         if (ArgsRes1.isInvalid())
           return ExprError();
         Args[1] = ArgsRes1.get();
diff --git a/clang/lib/Sema/SemaPseudoObject.cpp b/clang/lib/Sema/SemaPseudoObject.cpp
index 82774760b34d4..c6a0a182d3583 100644
--- a/clang/lib/Sema/SemaPseudoObject.cpp
+++ b/clang/lib/Sema/SemaPseudoObject.cpp
@@ -1136,7 +1136,7 @@ static void CheckKeyForObjCARCConversion(Sema &S, QualType ContainerT,
     return;
   QualType T = Getter->parameters()[0]->getType();
   S.CheckObjCConversion(Key->getSourceRange(), T, Key,
-                        Sema::CCK_ImplicitConversion);
+                        CheckedConversionKind::Implicit);
 }
 
 bool ObjCSubscriptOpBuilder::findAtIndexGetter() {

From 49b209d0d1833a339e66735e1288c1805224603e Mon Sep 17 00:00:00 2001
From: Jan Patrick Lehr <JanPatrick.Lehr@amd.com>
Date: Wed, 17 Apr 2024 09:11:43 +0200
Subject: [PATCH 707/886] Revert "[Libomptarget] Rework Record & Replay to be a
 plugin member" (#89028)

Reverts llvm/llvm-project#88928

This broke the AMDGPU buildbots:
https://lab.llvm.org/buildbot/#/builders/193/builds/50201
https://lab.llvm.org/staging/#/builders/185/builds/5565
https://lab.llvm.org/buildbot/#/builders/259/builds/2955
---
 .../common/include/PluginInterface.h          | 11 ------
 .../common/src/PluginInterface.cpp            | 39 +++++++------------
 2 files changed, 15 insertions(+), 35 deletions(-)

diff --git a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
index 7f05464f36c1f..79e8464bfda5c 100644
--- a/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
+++ b/openmp/libomptarget/plugins-nextgen/common/include/PluginInterface.h
@@ -45,8 +45,6 @@
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TargetParser/Triple.h"
 
-struct RecordReplayTy;
-
 namespace llvm {
 namespace omp {
 namespace target {
@@ -1033,12 +1031,6 @@ struct GenericPluginTy {
     return *RPCServer;
   }
 
-  /// Get a reference to the R&R interface for this plugin.
-  RecordReplayTy &getRecordAndReplay() const {
-    assert(RecordReplay && "R&R not initialized");
-    return *RecordReplay;
-  }
-
   /// Get the OpenMP requires flags set for this plugin.
   int64_t getRequiresFlags() const { return RequiresFlags; }
 
@@ -1228,9 +1220,6 @@ struct GenericPluginTy {
 
   /// The interface between the plugin and the GPU for host services.
   RPCServerTy *RPCServer;
-
-  /// The interface into the record-and-replay functionality.
-  RecordReplayTy *RecordReplay;
 };
 
 namespace Plugin {
diff --git a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
index 6df9798f12e3d..b5f3c45c835fd 100644
--- a/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
+++ b/openmp/libomptarget/plugins-nextgen/common/src/PluginInterface.cpp
@@ -362,6 +362,8 @@ struct RecordReplayTy {
   }
 };
 
+static RecordReplayTy RecordReplay;
+
 // Extract the mapping of host function pointers to device function pointers
 // from the entry table. Functions marked as 'indirect' in OpenMP will have
 // offloading entries generated for them which map the host's function pointer
@@ -471,8 +473,7 @@ GenericKernelTy::getKernelLaunchEnvironment(
   // Ctor/Dtor have no arguments, replaying uses the original kernel launch
   // environment. Older versions of the compiler do not generate a kernel
   // launch environment.
-  if (isCtorOrDtor() ||
-      GenericDevice.Plugin.getRecordAndReplay().isReplaying() ||
+  if (isCtorOrDtor() || RecordReplay.isReplaying() ||
       Version < OMP_KERNEL_ARG_MIN_VERSION_WITH_DYN_PTR)
     return nullptr;
 
@@ -561,7 +562,6 @@ Error GenericKernelTy::launch(GenericDeviceTy &GenericDevice, void **ArgPtrs,
 
   // Record the kernel description after we modified the argument count and num
   // blocks/threads.
-  RecordReplayTy &RecordReplay = GenericDevice.Plugin.getRecordAndReplay();
   if (RecordReplay.isRecording()) {
     RecordReplay.saveImage(getName(), getImage());
     RecordReplay.saveKernelInput(getName(), getImage());
@@ -839,6 +839,9 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
     delete MemoryManager;
   MemoryManager = nullptr;
 
+  if (RecordReplay.isRecordingOrReplaying())
+    RecordReplay.deinit();
+
   if (RPCServer)
     if (auto Err = RPCServer->deinitDevice(*this))
       return Err;
@@ -855,7 +858,6 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) {
 
   return deinitImpl();
 }
-
 Expected<DeviceImageTy *>
 GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
                             const __tgt_device_image *InputTgtImage) {
@@ -890,8 +892,7 @@ GenericDeviceTy::loadBinary(GenericPluginTy &Plugin,
     return std::move(Err);
 
   // Setup the global device memory pool if needed.
-  if (!Plugin.getRecordAndReplay().isReplaying() &&
-      shouldSetupDeviceMemoryPool()) {
+  if (!RecordReplay.isReplaying() && shouldSetupDeviceMemoryPool()) {
     uint64_t HeapSize;
     auto SizeOrErr = getDeviceHeapSize(HeapSize);
     if (SizeOrErr) {
@@ -1306,8 +1307,8 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
                                             TargetAllocTy Kind) {
   void *Alloc = nullptr;
 
-  if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
-    return Plugin.getRecordAndReplay().alloc(Size);
+  if (RecordReplay.isRecordingOrReplaying())
+    return RecordReplay.alloc(Size);
 
   switch (Kind) {
   case TARGET_ALLOC_DEFAULT:
@@ -1343,7 +1344,7 @@ Expected<void *> GenericDeviceTy::dataAlloc(int64_t Size, void *HostPtr,
 
 Error GenericDeviceTy::dataDelete(void *TgtPtr, TargetAllocTy Kind) {
   // Free is a noop when recording or replaying.
-  if (Plugin.getRecordAndReplay().isRecordingOrReplaying())
+  if (RecordReplay.isRecordingOrReplaying())
     return Plugin::success();
 
   int Res;
@@ -1395,7 +1396,6 @@ Error GenericDeviceTy::launchKernel(void *EntryPtr, void **ArgPtrs,
                                     ptrdiff_t *ArgOffsets,
                                     KernelArgsTy &KernelArgs,
                                     __tgt_async_info *AsyncInfo) {
-  RecordReplayTy &RecordReplay = Plugin.getRecordAndReplay();
   AsyncInfoWrapperTy AsyncInfoWrapper(
       *this, RecordReplay.isRecordingOrReplaying() ? nullptr : AsyncInfo);
 
@@ -1495,9 +1495,6 @@ Error GenericPluginTy::init() {
   RPCServer = new RPCServerTy(*this);
   assert(RPCServer && "Invalid RPC server");
 
-  RecordReplay = new RecordReplayTy();
-  assert(RecordReplay && "Invalid Record and Replay handler");
-
   return Plugin::success();
 }
 
@@ -1511,9 +1508,6 @@ Error GenericPluginTy::deinit() {
     assert(!Devices[DeviceId] && "Device was not deinitialized");
   }
 
-  if (RecordReplay && RecordReplay->isRecordingOrReplaying())
-    RecordReplay->deinit();
-
   // There is no global handler if no device is available.
   if (GlobalHandler)
     delete GlobalHandler;
@@ -1521,9 +1515,6 @@ Error GenericPluginTy::deinit() {
   if (RPCServer)
     delete RPCServer;
 
-  if (RecordReplay)
-    delete RecordReplay;
-
   // Perform last deinitializations on the plugin.
   return deinitImpl();
 }
@@ -1639,12 +1630,12 @@ int32_t GenericPluginTy::initialize_record_replay(int32_t DeviceId,
       isRecord ? RecordReplayTy::RRStatusTy::RRRecording
                : RecordReplayTy::RRStatusTy::RRReplaying;
 
-  if (auto Err = RecordReplay->init(&Device, MemorySize, VAddr, Status,
-                                    SaveOutput, ReqPtrArgOffset)) {
+  if (auto Err = RecordReplay.init(&Device, MemorySize, VAddr, Status,
+                                   SaveOutput, ReqPtrArgOffset)) {
     REPORT("WARNING RR did not intialize RR-properly with %lu bytes"
            "(Error: %s)\n",
            MemorySize, toString(std::move(Err)).data());
-    RecordReplay->setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
+    RecordReplay.setStatus(RecordReplayTy::RRStatusTy::RRDeactivated);
 
     if (!isRecord) {
       return OFFLOAD_FAIL;
@@ -1993,8 +1984,8 @@ int32_t GenericPluginTy::get_global(__tgt_device_binary Binary, uint64_t Size,
   assert(DevicePtr && "Invalid device global's address");
 
   // Save the loaded globals if we are recording.
-  if (getRecordAndReplay().isRecording())
-    getRecordAndReplay().addEntry(Name, Size, *DevicePtr);
+  if (RecordReplay.isRecording())
+    RecordReplay.addEntry(Name, Size, *DevicePtr);
 
   return OFFLOAD_SUCCESS;
 }

From 9f3334e9932fc9b55cd3590b140913222454c031 Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Wed, 17 Apr 2024 09:20:55 +0200
Subject: [PATCH 708/886] [mlir][SparseTensor] Add missing dependent dialect to
 pass (#88870)

This commit fixes the following error when stopping the sparse compiler
pipeline after bufferization (e.g., with `test-analysis-only`):

```
LLVM ERROR: Building op `vector.print` but it isn't known in this MLIRContext: the dialect may not be loaded or this operation hasn't been added by the dialect. See also https://mlir.llvm.org/getting_started/Faq/#registered-loaded-dependent-whats-up-with-dialects-management
```
---
 mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td   | 1 +
 .../Transforms/SparsificationAndBufferizationPass.cpp         | 1 +
 .../Integration/Dialect/SparseTensor/CPU/sparse_block3d.mlir  | 4 ++++
 3 files changed, 6 insertions(+)

diff --git a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
index 4706d5ba2f218..2f844cee5ff52 100644
--- a/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
+++ b/mlir/include/mlir/Dialect/SparseTensor/Transforms/Passes.td
@@ -460,6 +460,7 @@ def SparsificationAndBufferization : Pass<"sparsification-and-bufferization", "M
     "memref::MemRefDialect",
     "scf::SCFDialect",
     "sparse_tensor::SparseTensorDialect",
+    "vector::VectorDialect"
   ];
 }
 
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparsificationAndBufferizationPass.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparsificationAndBufferizationPass.cpp
index f497be6e48eba..3a8972072ac3b 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparsificationAndBufferizationPass.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparsificationAndBufferizationPass.cpp
@@ -24,6 +24,7 @@
 #include "mlir/Dialect/SCF/IR/SCF.h"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensor.h"
 #include "mlir/Dialect/SparseTensor/Transforms/Passes.h"
+#include "mlir/Dialect/Vector/IR/VectorOps.h"
 #include "mlir/Pass/PassManager.h"
 #include "mlir/Transforms/Passes.h"
 
diff --git a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_block3d.mlir b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_block3d.mlir
index 2ff73923c8327..467b671500e17 100755
--- a/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_block3d.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_block3d.mlir
@@ -30,6 +30,10 @@
 // Do the same run, but now with direct IR generation and VLA vectorization.
 // RUN: %if mlir_arm_sve_tests %{ %{compile_sve} | %{run_sve} | FileCheck %s %}
 
+// Test that test-bufferization-analysis-only works. This option is useful
+// for understanding why buffer copies were inserted.
+// RUN: mlir-opt %s --sparsifier="test-bufferization-analysis-only" -o /dev/null
+
 #Sparse1 = #sparse_tensor.encoding<{
   map = (i, j, k) -> (
     j : compressed,

From 889dfd4ab35892840f2bd2d6d7fed6fac025e18e Mon Sep 17 00:00:00 2001
From: Guillaume Chatelet <gchatelet@google.com>
Date: Wed, 17 Apr 2024 10:04:22 +0200
Subject: [PATCH 709/886] [libc][msan] Fix "non-constexpr function
 '__msan_unpoison' cannot be used in a constant expression" (#88719)

Prior to this patch, calling `cpp::bit_cast<T>` in `constexpr`
expressions under `-fsanitize=memory` would fail with the following
message "non-constexpr function '__msan_unpoison' cannot be used in a
constant expression".

This patch makes sure that the `__msan_unpoison` expression is guarded
by `!__builtin_is_constant_evaluated()`.
---
 libc/src/__support/macros/sanitizer.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/libc/src/__support/macros/sanitizer.h b/libc/src/__support/macros/sanitizer.h
index bd9b62b7121a1..baf44f7996cab 100644
--- a/libc/src/__support/macros/sanitizer.h
+++ b/libc/src/__support/macros/sanitizer.h
@@ -47,14 +47,13 @@
 // Functions to unpoison memory
 //-----------------------------------------------------------------------------
 
-#if defined(LIBC_HAVE_MEMORY_SANITIZER) && __has_builtin(__builtin_constant_p)
+#if defined(LIBC_HAVE_MEMORY_SANITIZER)
 // Only perform MSAN unpoison in non-constexpr context.
 #include <sanitizer/msan_interface.h>
 #define MSAN_UNPOISON(addr, size)                                              \
   do {                                                                         \
-    if (!__builtin_constant_p(*addr)) {                                        \
+    if (!__builtin_is_constant_evaluated())                                    \
       __msan_unpoison(addr, size);                                             \
-    }                                                                          \
   } while (0)
 #else
 #define MSAN_UNPOISON(ptr, size)

From 17b86d5978af8d171fa28763a9e5eba3ce93713a Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang@intel.com>
Date: Wed, 17 Apr 2024 15:59:37 +0800
Subject: [PATCH 710/886] [X86][NFC] Add test cases for pr88958

---
 llvm/test/CodeGen/X86/combine-ptest.ll | 42 ++++++++++++++++++++++++++
 1 file changed, 42 insertions(+)

diff --git a/llvm/test/CodeGen/X86/combine-ptest.ll b/llvm/test/CodeGen/X86/combine-ptest.ll
index 337edef96beee..3a695bfc6234d 100644
--- a/llvm/test/CodeGen/X86/combine-ptest.ll
+++ b/llvm/test/CodeGen/X86/combine-ptest.ll
@@ -397,6 +397,48 @@ define i1 @PR38788(<4 x i32> %0, <4 x i32> %1) {
   ret i1 %7
 }
 
+define i32 @PR88958_1(ptr %0, <2 x i64> %1) {
+; SSE-LABEL: PR88958_1:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    xorl %eax, %eax
+; SSE-NEXT:    ptest %xmm0, %xmm1
+; SSE-NEXT:    sete %al
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: PR88958_1:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vmovdqa (%rdi), %xmm1
+; AVX-NEXT:    xorl %eax, %eax
+; AVX-NEXT:    vptest %xmm0, %xmm1
+; AVX-NEXT:    sete %al
+; AVX-NEXT:    retq
+  %3 = load <2 x i64>, ptr %0
+  %4 = tail call i32 @llvm.x86.sse41.ptestz(<2 x i64> %3, <2 x i64> %1)
+  ret i32 %4
+}
+
+define i32 @PR88958_2(ptr %0, <2 x i64> %1) {
+; SSE-LABEL: PR88958_2:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    xorl %eax, %eax
+; SSE-NEXT:    ptest %xmm0, %xmm1
+; SSE-NEXT:    setb %al
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: PR88958_2:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vmovdqa (%rdi), %xmm1
+; AVX-NEXT:    xorl %eax, %eax
+; AVX-NEXT:    vptest %xmm0, %xmm1
+; AVX-NEXT:    setb %al
+; AVX-NEXT:    retq
+  %3 = load <2 x i64>, ptr %0
+  %4 = tail call i32 @llvm.x86.sse41.ptestc(<2 x i64> %3, <2 x i64> %1)
+  ret i32 %4
+}
+
 declare i32 @llvm.x86.sse41.ptestz(<2 x i64>, <2 x i64>) nounwind readnone
 declare i32 @llvm.x86.sse41.ptestc(<2 x i64>, <2 x i64>) nounwind readnone
 declare i32 @llvm.x86.sse41.ptestnzc(<2 x i64>, <2 x i64>) nounwind readnone

From d1a69e4a6ee0b04778da7728123c47eef2290564 Mon Sep 17 00:00:00 2001
From: shamithoke <152091883+shamithoke@users.noreply.github.com>
Date: Wed, 17 Apr 2024 14:16:10 +0530
Subject: [PATCH 711/886] Move gfni for bitreverse check out of SSSE3. (#88938)

For lowering bitreverse using GFNI, the check is put under SSSE3. This can be pulled out of SSSE3.

Co-authored-by: shami <shami_thoke@yahoo.com>
---
 llvm/lib/Target/X86/X86ISelLowering.cpp | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index f16a751a166d6..27107f554fccf 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1276,6 +1276,13 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::STRICT_FDIV,        MVT::v2f64, Legal);
   }
 
+  if (Subtarget.hasGFNI()) {
+    setOperationAction(ISD::BITREVERSE, MVT::i8, Custom);
+    setOperationAction(ISD::BITREVERSE, MVT::i16, Custom);
+    setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
+    setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
+  }
+
   if (!Subtarget.useSoftFloat() && Subtarget.hasSSSE3()) {
     setOperationAction(ISD::ABS,                MVT::v16i8, Legal);
     setOperationAction(ISD::ABS,                MVT::v8i16, Legal);
@@ -1286,13 +1293,6 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
       setOperationAction(ISD::CTLZ,             VT, Custom);
     }
 
-    if (Subtarget.hasGFNI()) {
-      setOperationAction(ISD::BITREVERSE, MVT::i8, Custom);
-      setOperationAction(ISD::BITREVERSE, MVT::i16, Custom);
-      setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
-      setOperationAction(ISD::BITREVERSE, MVT::i64, Custom);
-    }
-
     // These might be better off as horizontal vector ops.
     setOperationAction(ISD::ADD,                MVT::i16, Custom);
     setOperationAction(ISD::ADD,                MVT::i32, Custom);

From a16bb0701409376dee3a587ae351a6019d6de4e0 Mon Sep 17 00:00:00 2001
From: David Spickett <david.spickett@linaro.org>
Date: Wed, 17 Apr 2024 09:16:46 +0000
Subject: [PATCH 712/886] [lldb][test] Improve invalid compiler error message

I was debugging space separation issues when passing user arguments
and noticed this error is really hard to read in that scenario.

Put "" around the invalid compiler name so you can tell whether
you have spaces around it that's causing the problem.
---
 lldb/packages/Python/lldbsuite/test/dotest.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lldb/packages/Python/lldbsuite/test/dotest.py b/lldb/packages/Python/lldbsuite/test/dotest.py
index 8c29145ecc527..2ec4a840b9167 100644
--- a/lldb/packages/Python/lldbsuite/test/dotest.py
+++ b/lldb/packages/Python/lldbsuite/test/dotest.py
@@ -248,7 +248,7 @@ def parseOptionsAndInitTestdirs():
             configuration.compiler = which(args.compiler)
         if not is_exe(configuration.compiler):
             logging.error(
-                "%s is not a valid compiler executable; aborting...", args.compiler
+                '"%s" is not a valid compiler executable; aborting...', args.compiler
             )
             sys.exit(-1)
     else:

From d9a5aa8e2d755643cf4e7fa86aa831ed226fe54d Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Wed, 17 Apr 2024 18:22:05 +0900
Subject: [PATCH 713/886] [PatternMatch] Do not accept undef elements in
 m_AllOnes() and friends (#88217)

Change all the cstval_pred_ty based PatternMatch helpers (things like
m_AllOnes and m_Zero) to only allow poison elements inside vector
splats, not undef elements.

Historically, we used to represent non-demanded elements in vectors
using undef. Nowadays, we use poison instead. As such, I believe that
support for undef in vector splats is no longer useful.

At the same time, while poison splat elements are pretty much always
safe to ignore, this is not generally the case for undef elements. We
have existing miscompiles in our tests due to this (see the
masked-merge-*.ll tests changed here) and it's easy to miss such cases
in the future, now that we write tests using poison instead of undef
elements.

I think overall, keeping support for undef elements no longer makes
sense, and we should drop it. Once this is done consistently, I think we
may also consider allowing poison in m_APInt by default, as doing that
change is much less risky than doing the same with undef.

This change involves a substantial amount of test changes. For most
tests, I've just replaced undef with poison, as I don't think there is
value in retaining both. For some tests (where the distinction between
undef and poison is important), I've duplicated tests.
---
 llvm/include/llvm/IR/PatternMatch.h           |  35 +----
 llvm/lib/Analysis/InstructionSimplify.cpp     |  23 ++-
 llvm/lib/IR/Constants.cpp                     |   2 +-
 .../InstCombine/InstCombineAndOrXor.cpp       |  12 +-
 .../InstCombine/X86/x86-vector-shifts.ll      |  30 ++--
 llvm/test/Transforms/InstCombine/abs-1.ll     |  16 +-
 .../Transforms/InstCombine/add-mask-neg.ll    |   6 +-
 llvm/test/Transforms/InstCombine/add.ll       |  28 ++--
 .../Transforms/InstCombine/and-or-icmps.ll    |  61 +++++---
 .../test/Transforms/InstCombine/and-xor-or.ll |   4 +-
 llvm/test/Transforms/InstCombine/and.ll       |  86 +++++------
 llvm/test/Transforms/InstCombine/and2.ll      |  19 ++-
 llvm/test/Transforms/InstCombine/ashr-lshr.ll |  28 ++--
 .../Transforms/InstCombine/ashr-or-mul-abs.ll |   8 +-
 .../InstCombine/binop-and-shifts.ll           |  68 ++++----
 .../InstCombine/binop-of-displaced-shifts.ll  |  22 +--
 ...ern-between-zero-and-positive-threshold.ll |  20 +--
 ...nt-low-bit-mask-and-icmp-eq-to-icmp-ule.ll |   6 +-
 ...nt-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll |  12 +-
 ...t-low-bit-mask-and-icmp-sge-to-icmp-sle.ll |  10 +-
 ...t-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll |  16 +-
 ...t-low-bit-mask-and-icmp-sle-to-icmp-sle.ll |  10 +-
 ...t-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll |  16 +-
 ...t-low-bit-mask-and-icmp-uge-to-icmp-ule.ll |   6 +-
 ...t-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll |  12 +-
 ...t-low-bit-mask-and-icmp-ule-to-icmp-ule.ll |   6 +-
 ...t-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll |  12 +-
 ...ze-low-bit-mask-and-icmp-eq-to-icmp-ule.ll |   8 +-
 ...ze-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll |   8 +-
 ...low-bit-mask-v2-and-icmp-eq-to-icmp-ule.ll |  20 +--
 ...low-bit-mask-v2-and-icmp-ne-to-icmp-ugt.ll |  20 +--
 ...low-bit-mask-v4-and-icmp-eq-to-icmp-ule.ll |   8 +-
 ...low-bit-mask-v4-and-icmp-ne-to-icmp-ugt.ll |   8 +-
 .../InstCombine/cast-int-icmp-eq-0.ll         |  22 +--
 .../InstCombine/cast-unsigned-icmp-eqcmp-0.ll |  36 ++---
 llvm/test/Transforms/InstCombine/cast.ll      |  56 +++----
 .../test/Transforms/InstCombine/ctpop-cttz.ll |   6 +-
 llvm/test/Transforms/InstCombine/ctpop.ll     |  17 +-
 .../Transforms/InstCombine/fabs-as-int.ll     |   6 +-
 llvm/test/Transforms/InstCombine/fabs.ll      |   4 +-
 llvm/test/Transforms/InstCombine/fast-math.ll |   6 +-
 .../Transforms/InstCombine/fcmp-special.ll    |  20 +--
 llvm/test/Transforms/InstCombine/fcmp.ll      |  26 ++--
 llvm/test/Transforms/InstCombine/fdiv.ll      |  22 +--
 llvm/test/Transforms/InstCombine/fma.ll       |   8 +-
 llvm/test/Transforms/InstCombine/fmul.ll      |  42 ++---
 .../Transforms/InstCombine/fneg-as-int.ll     |   6 +-
 .../InstCombine/fneg-fabs-as-int.ll           |   6 +-
 llvm/test/Transforms/InstCombine/fneg.ll      |  36 ++---
 ...c-of-add-of-not-x-and-y-to-sub-x-from-y.ll |  20 +--
 .../fold-sub-of-not-to-inc-of-add.ll          |   6 +-
 llvm/test/Transforms/InstCombine/fpcast.ll    |   6 +-
 llvm/test/Transforms/InstCombine/fsub.ll      |  14 +-
 llvm/test/Transforms/InstCombine/funnel.ll    |  54 +++----
 .../get-lowbitmask-upto-and-including-bit.ll  |  20 +--
 .../hoist-negation-out-of-bias-calculation.ll |   6 +-
 .../hoist-not-from-ashr-operand.ll            |   8 +-
 ...al-to-icmp-eq-of-lshr-val-by-bits-and-0.ll |   4 +-
 ...al-to-icmp-eq-of-lshr-val-by-bits-and-0.ll |  20 +--
 ...al-to-icmp-eq-of-lshr-val-by-bits-and-0.ll |   6 +-
 ...al-to-icmp-ne-of-lshr-val-by-bits-and-0.ll |   6 +-
 ...al-to-icmp-ne-of-lshr-val-by-bits-and-0.ll |   4 +-
 ...al-to-icmp-ne-of-lshr-val-by-bits-and-0.ll |  20 +--
 llvm/test/Transforms/InstCombine/icmp.ll      |  62 ++++----
 .../integer-round-up-pow2-alignment.ll        |  70 ++++-----
 ...rt-variable-mask-in-masked-merge-vector.ll |  30 ++--
 .../InstCombine/lshr-and-negC-icmpeq-zero.ll  |  20 +--
 .../lshr-and-signbit-icmpeq-zero.ll           |  20 +--
 .../InstCombine/masked-merge-add.ll           |  17 +-
 .../Transforms/InstCombine/masked-merge-or.ll |  17 +-
 .../InstCombine/masked-merge-xor.ll           |  17 +-
 .../Transforms/InstCombine/min-positive.ll    |   8 +-
 .../Transforms/InstCombine/minmax-fold.ll     |  11 +-
 .../InstCombine/minmax-intrinsics.ll          |  40 ++---
 .../InstCombine/mul-inseltpoison.ll           |  22 +--
 llvm/test/Transforms/InstCombine/mul.ll       |  30 ++--
 llvm/test/Transforms/InstCombine/not-add.ll   |  16 +-
 llvm/test/Transforms/InstCombine/not.ll       |   4 +-
 ...of-two-or-zero-when-comparing-with-zero.ll |  34 ++--
 .../InstCombine/operand-complexity.ll         |  22 +--
 llvm/test/Transforms/InstCombine/or.ll        |  52 +++----
 ...nput-masking-after-truncation-variant-b.ll |  20 +--
 ...nput-masking-after-truncation-variant-c.ll |  20 +--
 ...nput-masking-after-truncation-variant-d.ll |  20 +--
 ...dant-left-shift-input-masking-variant-b.ll |  20 +--
 ...dant-left-shift-input-masking-variant-c.ll |  12 +-
 ...dant-left-shift-input-masking-variant-d.ll |  12 +-
 llvm/test/Transforms/InstCombine/pr53357.ll   |   8 +-
 ...nput-masking-after-truncation-variant-b.ll |  16 +-
 ...nput-masking-after-truncation-variant-c.ll |  20 +--
 ...nput-masking-after-truncation-variant-d.ll |  20 +--
 ...dant-left-shift-input-masking-variant-b.ll |  18 +--
 ...dant-left-shift-input-masking-variant-c.ll |  12 +-
 ...dant-left-shift-input-masking-variant-d.ll |  10 +-
 .../reuse-constant-from-select-in-icmp.ll     |  26 ++--
 llvm/test/Transforms/InstCombine/rotate.ll    |  57 +++----
 .../InstCombine/saturating-add-sub.ll         |  18 +--
 .../InstCombine/select-of-bittest.ll          |  89 ++++++-----
 llvm/test/Transforms/InstCombine/select.ll    |  33 ++--
 .../Transforms/InstCombine/select_meta.ll     |   8 +-
 .../set-lowbits-mask-canonicalize.ll          |  20 +--
 llvm/test/Transforms/InstCombine/sext.ll      |  28 ++--
 .../shift-amount-reassociation-in-bittest.ll  |  64 ++++----
 ...ount-reassociation-with-truncation-ashr.ll |  26 ++--
 ...ount-reassociation-with-truncation-lshr.ll |  32 ++--
 .../InstCombine/shift-amount-reassociation.ll |  26 ++--
 .../Transforms/InstCombine/shift-logic.ll     |  88 +++++------
 .../InstCombine/shl-and-negC-icmpeq-zero.ll   |  20 +--
 .../shl-and-signbit-icmpeq-zero.ll            |  20 +--
 .../signmask-of-sext-vs-of-shl-of-zext.ll     |  42 +++--
 llvm/test/Transforms/InstCombine/sub-not.ll   |   8 +-
 llvm/test/Transforms/InstCombine/sub.ll       |  44 +++---
 .../InstCombine/trunc-inseltpoison.ll         |  98 ++++++------
 .../InstCombine/trunc-shift-trunc.ll          |  16 +-
 llvm/test/Transforms/InstCombine/trunc.ll     | 102 ++++++------
 ...k-of-overflow-check-via-udiv-of-allones.ll |   6 +-
 ...-mul-overflow-check-via-udiv-of-allones.ll |   6 +-
 ...signext-of-variable-high-bit-extraction.ll |  12 +-
 llvm/test/Transforms/InstCombine/vec_sext.ll  |  14 +-
 .../InstCombine/vector-casts-inseltpoison.ll  |  20 +--
 .../Transforms/InstCombine/vector-casts.ll    |  20 +--
 .../Transforms/InstCombine/vector-urem.ll     |  18 +--
 .../test/Transforms/InstCombine/vector-xor.ll |  66 ++++----
 .../InstCombine/zext-bool-add-sub.ll          |  28 ++--
 llvm/test/Transforms/InstSimplify/AndOrXor.ll |  36 ++---
 llvm/test/Transforms/InstSimplify/call.ll     |  10 +-
 llvm/test/Transforms/InstSimplify/compare.ll  |  66 ++++----
 ...constantfold-add-nuw-allones-to-allones.ll |   8 +-
 .../constantfold-shl-nuw-C-to-C.ll            |   8 +-
 llvm/test/Transforms/InstSimplify/div.ll      |  20 +--
 .../InstSimplify/fast-math-strictfp.ll        |  68 ++++----
 .../test/Transforms/InstSimplify/fast-math.ll |  60 ++++----
 llvm/test/Transforms/InstSimplify/fdiv.ll     |   6 +-
 .../floating-point-arithmetic-strictfp.ll     |  54 +++----
 .../InstSimplify/floating-point-arithmetic.ll |  38 ++---
 .../InstSimplify/floating-point-compare.ll    |  30 ++--
 .../Transforms/InstSimplify/fminmax-folds.ll  |  36 ++---
 llvm/test/Transforms/InstSimplify/fp-nan.ll   |  13 +-
 .../InstSimplify/icmp-bool-constant.ll        |  18 +--
 .../InstSimplify/icmp-not-bool-constant.ll    |  36 ++---
 llvm/test/Transforms/InstSimplify/ldexp.ll    |   5 +-
 llvm/test/Transforms/InstSimplify/mul.ll      |   6 +-
 llvm/test/Transforms/InstSimplify/negate.ll   |  12 +-
 llvm/test/Transforms/InstSimplify/or.ll       |  73 +++++++--
 llvm/test/Transforms/InstSimplify/ptrmask.ll  |   3 +-
 llvm/test/Transforms/InstSimplify/rem.ll      |   6 +-
 .../InstSimplify/saturating-add-sub.ll        |   8 +-
 llvm/test/Transforms/InstSimplify/sdiv.ll     |   6 +-
 .../InstSimplify/select-inseltpoison.ll       |  38 ++---
 llvm/test/Transforms/InstSimplify/select.ll   |  38 ++---
 llvm/test/Transforms/InstSimplify/shift.ll    |  35 +++--
 llvm/test/Transforms/InstSimplify/srem.ll     |   6 +-
 llvm/test/Transforms/InstSimplify/sub.ll      |   6 +-
 llvm/test/Transforms/InstSimplify/xor.ll      |  20 ++-
 llvm/test/Transforms/Reassociate/inverses.ll  |   6 +-
 llvm/test/Transforms/Reassociate/negation.ll  |   8 +-
 llvm/unittests/IR/ConstantsTest.cpp           |  39 ++++-
 llvm/unittests/IR/PatternMatch.cpp            | 145 ++++++++++++++----
 158 files changed, 2042 insertions(+), 1839 deletions(-)

diff --git a/llvm/include/llvm/IR/PatternMatch.h b/llvm/include/llvm/IR/PatternMatch.h
index 92cb79d54afc2..98cc0e5037698 100644
--- a/llvm/include/llvm/IR/PatternMatch.h
+++ b/llvm/include/llvm/IR/PatternMatch.h
@@ -345,7 +345,7 @@ template <int64_t Val> inline constantint_match<Val> m_ConstantInt() {
 
 /// This helper class is used to match constant scalars, vector splats,
 /// and fixed width vectors that satisfy a specified predicate.
-/// For fixed width vector constants, undefined elements are ignored.
+/// For fixed width vector constants, poison elements are ignored.
 template <typename Predicate, typename ConstantVal>
 struct cstval_pred_ty : public Predicate {
   template <typename ITy> bool match(ITy *V) {
@@ -364,19 +364,19 @@ struct cstval_pred_ty : public Predicate {
         // Non-splat vector constant: check each element for a match.
         unsigned NumElts = FVTy->getNumElements();
         assert(NumElts != 0 && "Constant vector with no elements?");
-        bool HasNonUndefElements = false;
+        bool HasNonPoisonElements = false;
         for (unsigned i = 0; i != NumElts; ++i) {
           Constant *Elt = C->getAggregateElement(i);
           if (!Elt)
             return false;
-          if (isa<UndefValue>(Elt))
+          if (isa<PoisonValue>(Elt))
             continue;
           auto *CV = dyn_cast<ConstantVal>(Elt);
           if (!CV || !this->isValue(CV->getValue()))
             return false;
-          HasNonUndefElements = true;
+          HasNonPoisonElements = true;
         }
-        return HasNonUndefElements;
+        return HasNonPoisonElements;
       }
     }
     return false;
@@ -2587,31 +2587,6 @@ m_Not(const ValTy &V) {
   return m_c_Xor(m_AllOnes(), V);
 }
 
-template <typename ValTy> struct NotForbidUndef_match {
-  ValTy Val;
-  NotForbidUndef_match(const ValTy &V) : Val(V) {}
-
-  template <typename OpTy> bool match(OpTy *V) {
-    // We do not use m_c_Xor because that could match an arbitrary APInt that is
-    // not -1 as C and then fail to match the other operand if it is -1.
-    // This code should still work even when both operands are constants.
-    Value *X;
-    const APInt *C;
-    if (m_Xor(m_Value(X), m_APIntForbidUndef(C)).match(V) && C->isAllOnes())
-      return Val.match(X);
-    if (m_Xor(m_APIntForbidUndef(C), m_Value(X)).match(V) && C->isAllOnes())
-      return Val.match(X);
-    return false;
-  }
-};
-
-/// Matches a bitwise 'not' as 'xor V, -1' or 'xor -1, V'. For vectors, the
-/// constant value must be composed of only -1 scalar elements.
-template <typename ValTy>
-inline NotForbidUndef_match<ValTy> m_NotForbidUndef(const ValTy &V) {
-  return NotForbidUndef_match<ValTy>(V);
-}
-
 /// Matches an SMin with LHS and RHS in either order.
 template <typename LHS, typename RHS>
 inline MaxMin_match<ICmpInst, LHS, RHS, smin_pred_ty, true>
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 8955de6375dec..06ba5ca4c6b35 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -1513,7 +1513,7 @@ static Value *simplifyAShrInst(Value *Op0, Value *Op1, bool IsExact,
 
   // -1 >>a X --> -1
   // (-1 << X) a>> X --> -1
-  // Do not return Op0 because it may contain undef elements if it's a vector.
+  // We could return the original -1 constant to preserve poison elements.
   if (match(Op0, m_AllOnes()) ||
       match(Op0, m_Shl(m_AllOnes(), m_Specific(Op1))))
     return Constant::getAllOnesValue(Op0->getType());
@@ -2281,7 +2281,7 @@ static Value *simplifyOrLogic(Value *X, Value *Y) {
   // (B ^ ~A) | (A & B) --> B ^ ~A
   // (~A ^ B) | (B & A) --> ~A ^ B
   // (B ^ ~A) | (B & A) --> B ^ ~A
-  if (match(X, m_c_Xor(m_NotForbidUndef(m_Value(A)), m_Value(B))) &&
+  if (match(X, m_c_Xor(m_Not(m_Value(A)), m_Value(B))) &&
       match(Y, m_c_And(m_Specific(A), m_Specific(B))))
     return X;
 
@@ -2298,31 +2298,29 @@ static Value *simplifyOrLogic(Value *X, Value *Y) {
   // (B & ~A) | ~(A | B) --> ~A
   // (B & ~A) | ~(B | A) --> ~A
   Value *NotA;
-  if (match(X,
-            m_c_And(m_CombineAnd(m_Value(NotA), m_NotForbidUndef(m_Value(A))),
-                    m_Value(B))) &&
+  if (match(X, m_c_And(m_CombineAnd(m_Value(NotA), m_Not(m_Value(A))),
+                       m_Value(B))) &&
       match(Y, m_Not(m_c_Or(m_Specific(A), m_Specific(B)))))
     return NotA;
   // The same is true of Logical And
   // TODO: This could share the logic of the version above if there was a
   // version of LogicalAnd that allowed more than just i1 types.
-  if (match(X, m_c_LogicalAnd(
-                   m_CombineAnd(m_Value(NotA), m_NotForbidUndef(m_Value(A))),
-                   m_Value(B))) &&
+  if (match(X, m_c_LogicalAnd(m_CombineAnd(m_Value(NotA), m_Not(m_Value(A))),
+                              m_Value(B))) &&
       match(Y, m_Not(m_c_LogicalOr(m_Specific(A), m_Specific(B)))))
     return NotA;
 
   // ~(A ^ B) | (A & B) --> ~(A ^ B)
   // ~(A ^ B) | (B & A) --> ~(A ^ B)
   Value *NotAB;
-  if (match(X, m_CombineAnd(m_NotForbidUndef(m_Xor(m_Value(A), m_Value(B))),
+  if (match(X, m_CombineAnd(m_Not(m_Xor(m_Value(A), m_Value(B))),
                             m_Value(NotAB))) &&
       match(Y, m_c_And(m_Specific(A), m_Specific(B))))
     return NotAB;
 
   // ~(A & B) | (A ^ B) --> ~(A & B)
   // ~(A & B) | (B ^ A) --> ~(A & B)
-  if (match(X, m_CombineAnd(m_NotForbidUndef(m_And(m_Value(A), m_Value(B))),
+  if (match(X, m_CombineAnd(m_Not(m_And(m_Value(A), m_Value(B))),
                             m_Value(NotAB))) &&
       match(Y, m_c_Xor(m_Specific(A), m_Specific(B))))
     return NotAB;
@@ -2552,9 +2550,8 @@ static Value *simplifyXorInst(Value *Op0, Value *Op1, const SimplifyQuery &Q,
     // The 'not' op must contain a complete -1 operand (no undef elements for
     // vector) for the transform to be safe.
     Value *NotA;
-    if (match(X,
-              m_c_Or(m_CombineAnd(m_NotForbidUndef(m_Value(A)), m_Value(NotA)),
-                     m_Value(B))) &&
+    if (match(X, m_c_Or(m_CombineAnd(m_Not(m_Value(A)), m_Value(NotA)),
+                        m_Value(B))) &&
         match(Y, m_c_And(m_Specific(A), m_Specific(B))))
       return NotA;
 
diff --git a/llvm/lib/IR/Constants.cpp b/llvm/lib/IR/Constants.cpp
index a5fb497f54ed1..45b359a94b3ab 100644
--- a/llvm/lib/IR/Constants.cpp
+++ b/llvm/lib/IR/Constants.cpp
@@ -316,7 +316,7 @@ bool Constant::isElementWiseEqual(Value *Y) const {
   Constant *C0 = ConstantExpr::getBitCast(const_cast<Constant *>(this), IntTy);
   Constant *C1 = ConstantExpr::getBitCast(cast<Constant>(Y), IntTy);
   Constant *CmpEq = ConstantExpr::getICmp(ICmpInst::ICMP_EQ, C0, C1);
-  return isa<UndefValue>(CmpEq) || match(CmpEq, m_One());
+  return isa<PoisonValue>(CmpEq) || match(CmpEq, m_One());
 }
 
 static bool
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index d311690be64f1..0f4fbf5bbfbbd 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -2538,6 +2538,8 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
     }
   }
 
+  // and(shl(zext(X), Y), SignMask) -> and(sext(X), SignMask)
+  // where Y is a valid shift amount.
   if (match(&I, m_And(m_OneUse(m_Shl(m_ZExt(m_Value(X)), m_Value(Y))),
                       m_SignMask())) &&
       match(Y, m_SpecificInt_ICMP(
@@ -2546,15 +2548,7 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
                          Ty->getScalarSizeInBits() -
                              X->getType()->getScalarSizeInBits())))) {
     auto *SExt = Builder.CreateSExt(X, Ty, X->getName() + ".signext");
-    auto *SanitizedSignMask = cast<Constant>(Op1);
-    // We must be careful with the undef elements of the sign bit mask, however:
-    // the mask elt can be undef iff the shift amount for that lane was undef,
-    // otherwise we need to sanitize undef masks to zero.
-    SanitizedSignMask = Constant::replaceUndefsWith(
-        SanitizedSignMask, ConstantInt::getNullValue(Ty->getScalarType()));
-    SanitizedSignMask =
-        Constant::mergeUndefsWith(SanitizedSignMask, cast<Constant>(Y));
-    return BinaryOperator::CreateAnd(SExt, SanitizedSignMask);
+    return BinaryOperator::CreateAnd(SExt, Op1);
   }
 
   if (Instruction *Z = narrowMaskedBinOp(I))
diff --git a/llvm/test/Transforms/InstCombine/X86/x86-vector-shifts.ll b/llvm/test/Transforms/InstCombine/X86/x86-vector-shifts.ll
index 4600a6654a362..b1e5fa4f9e1c9 100644
--- a/llvm/test/Transforms/InstCombine/X86/x86-vector-shifts.ll
+++ b/llvm/test/Transforms/InstCombine/X86/x86-vector-shifts.ll
@@ -2032,23 +2032,23 @@ define <4 x i64> @avx2_psrlv_q_256_allbig(<4 x i64> %v) {
   ret <4 x i64> %1
 }
 
-; The shift amount is 0 (the undef lane could be 0), so we return the unshifted input.
+; The shift amount is 0 (the poison lane could be 0), so we return the unshifted input.
 
-define <2 x i64> @avx2_psrlv_q_128_undef(<2 x i64> %v) {
-; CHECK-LABEL: @avx2_psrlv_q_128_undef(
+define <2 x i64> @avx2_psrlv_q_128_poison(<2 x i64> %v) {
+; CHECK-LABEL: @avx2_psrlv_q_128_poison(
 ; CHECK-NEXT:    ret <2 x i64> [[V:%.*]]
 ;
-  %1 = insertelement <2 x i64> <i64 0, i64 8>, i64 undef, i64 1
+  %1 = insertelement <2 x i64> <i64 0, i64 8>, i64 poison, i64 1
   %2 = tail call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %v, <2 x i64> %1)
   ret <2 x i64> %2
 }
 
-define <4 x i64> @avx2_psrlv_q_256_undef(<4 x i64> %v) {
-; CHECK-LABEL: @avx2_psrlv_q_256_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <4 x i64> [[V:%.*]], <i64 undef, i64 8, i64 16, i64 31>
+define <4 x i64> @avx2_psrlv_q_256_poison(<4 x i64> %v) {
+; CHECK-LABEL: @avx2_psrlv_q_256_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <4 x i64> [[V:%.*]], <i64 poison, i64 8, i64 16, i64 31>
 ; CHECK-NEXT:    ret <4 x i64> [[TMP1]]
 ;
-  %1 = insertelement <4 x i64> <i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %1 = insertelement <4 x i64> <i64 0, i64 8, i64 16, i64 31>, i64 poison, i64 0
   %2 = tail call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %v, <4 x i64> %1)
   ret <4 x i64> %2
 }
@@ -2435,21 +2435,21 @@ define <4 x i64> @avx2_psllv_q_256_allbig(<4 x i64> %v) {
 
 ; The shift amount is 0 (the undef lane could be 0), so we return the unshifted input.
 
-define <2 x i64> @avx2_psllv_q_128_undef(<2 x i64> %v) {
-; CHECK-LABEL: @avx2_psllv_q_128_undef(
+define <2 x i64> @avx2_psllv_q_128_poison(<2 x i64> %v) {
+; CHECK-LABEL: @avx2_psllv_q_128_poison(
 ; CHECK-NEXT:    ret <2 x i64> [[V:%.*]]
 ;
-  %1 = insertelement <2 x i64> <i64 0, i64 8>, i64 undef, i64 1
+  %1 = insertelement <2 x i64> <i64 0, i64 8>, i64 poison, i64 1
   %2 = tail call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %v, <2 x i64> %1)
   ret <2 x i64> %2
 }
 
-define <4 x i64> @avx2_psllv_q_256_undef(<4 x i64> %v) {
-; CHECK-LABEL: @avx2_psllv_q_256_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = shl <4 x i64> [[V:%.*]], <i64 undef, i64 8, i64 16, i64 31>
+define <4 x i64> @avx2_psllv_q_256_poison(<4 x i64> %v) {
+; CHECK-LABEL: @avx2_psllv_q_256_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <4 x i64> [[V:%.*]], <i64 poison, i64 8, i64 16, i64 31>
 ; CHECK-NEXT:    ret <4 x i64> [[TMP1]]
 ;
-  %1 = insertelement <4 x i64> <i64 0, i64 8, i64 16, i64 31>, i64 undef, i64 0
+  %1 = insertelement <4 x i64> <i64 0, i64 8, i64 16, i64 31>, i64 poison, i64 0
   %2 = tail call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %v, <4 x i64> %1)
   ret <4 x i64> %2
 }
diff --git a/llvm/test/Transforms/InstCombine/abs-1.ll b/llvm/test/Transforms/InstCombine/abs-1.ll
index 7355c560c820b..32bd7a37053ed 100644
--- a/llvm/test/Transforms/InstCombine/abs-1.ll
+++ b/llvm/test/Transforms/InstCombine/abs-1.ll
@@ -63,14 +63,14 @@ define <2 x i8> @abs_canonical_2(<2 x i8> %x) {
   ret <2 x i8> %abs
 }
 
-; Even if a constant has undef elements.
+; Even if a constant has poison elements.
 
-define <2 x i8> @abs_canonical_2_vec_undef_elts(<2 x i8> %x) {
-; CHECK-LABEL: @abs_canonical_2_vec_undef_elts(
+define <2 x i8> @abs_canonical_2_vec_poison_elts(<2 x i8> %x) {
+; CHECK-LABEL: @abs_canonical_2_vec_poison_elts(
 ; CHECK-NEXT:    [[ABS:%.*]] = call <2 x i8> @llvm.abs.v2i8(<2 x i8> [[X:%.*]], i1 false)
 ; CHECK-NEXT:    ret <2 x i8> [[ABS]]
 ;
-  %cmp = icmp sgt <2 x i8> %x, <i8 undef, i8 -1>
+  %cmp = icmp sgt <2 x i8> %x, <i8 poison, i8 -1>
   %neg = sub <2 x i8> zeroinitializer, %x
   %abs = select <2 x i1> %cmp, <2 x i8> %x, <2 x i8> %neg
   ret <2 x i8> %abs
@@ -208,15 +208,15 @@ define <2 x i8> @nabs_canonical_2(<2 x i8> %x) {
   ret <2 x i8> %abs
 }
 
-; Even if a constant has undef elements.
+; Even if a constant has poison elements.
 
-define <2 x i8> @nabs_canonical_2_vec_undef_elts(<2 x i8> %x) {
-; CHECK-LABEL: @nabs_canonical_2_vec_undef_elts(
+define <2 x i8> @nabs_canonical_2_vec_poison_elts(<2 x i8> %x) {
+; CHECK-LABEL: @nabs_canonical_2_vec_poison_elts(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i8> @llvm.abs.v2i8(<2 x i8> [[X:%.*]], i1 false)
 ; CHECK-NEXT:    [[ABS:%.*]] = sub <2 x i8> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    ret <2 x i8> [[ABS]]
 ;
-  %cmp = icmp sgt <2 x i8> %x, <i8 -1, i8 undef>
+  %cmp = icmp sgt <2 x i8> %x, <i8 -1, i8 poison>
   %neg = sub <2 x i8> zeroinitializer, %x
   %abs = select <2 x i1> %cmp, <2 x i8> %neg, <2 x i8> %x
   ret <2 x i8> %abs
diff --git a/llvm/test/Transforms/InstCombine/add-mask-neg.ll b/llvm/test/Transforms/InstCombine/add-mask-neg.ll
index 5fad6155d348e..0e579f3097607 100644
--- a/llvm/test/Transforms/InstCombine/add-mask-neg.ll
+++ b/llvm/test/Transforms/InstCombine/add-mask-neg.ll
@@ -89,8 +89,8 @@ define <2 x i32> @dec_mask_neg_v2i32(<2 x i32> %X) {
   ret <2 x i32> %dec
 }
 
-define <2 x i32> @dec_mask_neg_v2i32_undef(<2 x i32> %X) {
-; CHECK-LABEL: @dec_mask_neg_v2i32_undef(
+define <2 x i32> @dec_mask_neg_v2i32_poison(<2 x i32> %X) {
+; CHECK-LABEL: @dec_mask_neg_v2i32_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i32> [[X:%.*]], <i32 -1, i32 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = xor <2 x i32> [[X]], <i32 -1, i32 -1>
 ; CHECK-NEXT:    [[DEC:%.*]] = and <2 x i32> [[TMP1]], [[TMP2]]
@@ -98,7 +98,7 @@ define <2 x i32> @dec_mask_neg_v2i32_undef(<2 x i32> %X) {
 ;
   %neg = sub <2 x i32> zeroinitializer, %X
   %mask = and <2 x i32> %neg, %X
-  %dec = add <2 x i32> %mask, <i32 -1, i32 undef>
+  %dec = add <2 x i32> %mask, <i32 -1, i32 poison>
   ret <2 x i32> %dec
 }
 
diff --git a/llvm/test/Transforms/InstCombine/add.ll b/llvm/test/Transforms/InstCombine/add.ll
index 408b0c6559b00..39b4ad8055088 100644
--- a/llvm/test/Transforms/InstCombine/add.ll
+++ b/llvm/test/Transforms/InstCombine/add.ll
@@ -150,24 +150,24 @@ define i32 @test5_add_nsw(i32 %A, i32 %B) {
   ret i32 %D
 }
 
-define <2 x i8> @neg_op0_vec_undef_elt(<2 x i8> %a, <2 x i8> %b) {
-; CHECK-LABEL: @neg_op0_vec_undef_elt(
+define <2 x i8> @neg_op0_vec_poison_elt(<2 x i8> %a, <2 x i8> %b) {
+; CHECK-LABEL: @neg_op0_vec_poison_elt(
 ; CHECK-NEXT:    [[R:%.*]] = sub <2 x i8> [[B:%.*]], [[A:%.*]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %nega = sub <2 x i8> <i8 0, i8 undef>, %a
+  %nega = sub <2 x i8> <i8 0, i8 poison>, %a
   %r = add <2 x i8> %nega, %b
   ret <2 x i8> %r
 }
 
-define <2 x i8> @neg_neg_vec_undef_elt(<2 x i8> %a, <2 x i8> %b) {
-; CHECK-LABEL: @neg_neg_vec_undef_elt(
+define <2 x i8> @neg_neg_vec_poison_elt(<2 x i8> %a, <2 x i8> %b) {
+; CHECK-LABEL: @neg_neg_vec_poison_elt(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i8> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = sub <2 x i8> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %nega = sub <2 x i8> <i8 undef, i8 0>, %a
-  %negb = sub <2 x i8> <i8 undef, i8 0>, %b
+  %nega = sub <2 x i8> <i8 poison, i8 0>, %a
+  %negb = sub <2 x i8> <i8 poison, i8 0>, %b
   %r = add <2 x i8> %nega, %negb
   ret <2 x i8> %r
 }
@@ -1196,14 +1196,14 @@ define <2 x i32> @test44_vec_non_matching(<2 x i32> %A) {
   ret <2 x i32> %C
 }
 
-define <2 x i32> @test44_vec_undef(<2 x i32> %A) {
-; CHECK-LABEL: @test44_vec_undef(
-; CHECK-NEXT:    [[B:%.*]] = or <2 x i32> [[A:%.*]], <i32 123, i32 undef>
-; CHECK-NEXT:    [[C:%.*]] = add <2 x i32> [[B]], <i32 -123, i32 undef>
+define <2 x i32> @test44_vec_poison(<2 x i32> %A) {
+; CHECK-LABEL: @test44_vec_poison(
+; CHECK-NEXT:    [[B:%.*]] = or <2 x i32> [[A:%.*]], <i32 123, i32 poison>
+; CHECK-NEXT:    [[C:%.*]] = add nsw <2 x i32> [[B]], <i32 -123, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[C]]
 ;
-  %B = or <2 x i32> %A, <i32 123, i32 undef>
-  %C = add <2 x i32> %B, <i32 -123, i32 undef>
+  %B = or <2 x i32> %A, <i32 123, i32 poison>
+  %C = add <2 x i32> %B, <i32 -123, i32 poison>
   ret <2 x i32> %C
 }
 
@@ -2983,7 +2983,7 @@ define i8 @signum_i8_i8_use3(i8 %x) {
   ret i8 %r
 }
 
-; poison/undef is ok to propagate in shift amount
+; poison is ok to propagate in shift amount
 ; complexity canonicalization guarantees that shift is op0 of add
 
 define <2 x i5> @signum_v2i5_v2i5(<2 x i5> %x) {
diff --git a/llvm/test/Transforms/InstCombine/and-or-icmps.ll b/llvm/test/Transforms/InstCombine/and-or-icmps.ll
index 63b11d0c0bc08..c20f48a985b3e 100644
--- a/llvm/test/Transforms/InstCombine/and-or-icmps.ll
+++ b/llvm/test/Transforms/InstCombine/and-or-icmps.ll
@@ -952,8 +952,8 @@ define i1 @substitute_constant_or_ne_uge_commute_logical(i8 %x, i8 %y) {
 
 ; Negative test - not safe to substitute vector constant with undef element
 
-define <2 x i1> @substitute_constant_or_ne_slt_swap_vec(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @substitute_constant_or_ne_slt_swap_vec(
+define <2 x i1> @substitute_constant_or_ne_slt_swap_vec_undef(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @substitute_constant_or_ne_slt_swap_vec_undef(
 ; CHECK-NEXT:    [[C1:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 42, i8 undef>
 ; CHECK-NEXT:    [[C2:%.*]] = icmp slt <2 x i8> [[Y:%.*]], [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = or <2 x i1> [[C1]], [[C2]]
@@ -965,14 +965,29 @@ define <2 x i1> @substitute_constant_or_ne_slt_swap_vec(<2 x i8> %x, <2 x i8> %y
   ret <2 x i1> %r
 }
 
+; TODO: The poison case would be valid to fold.
+
+define <2 x i1> @substitute_constant_or_ne_slt_swap_vec_poison(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @substitute_constant_or_ne_slt_swap_vec_poison(
+; CHECK-NEXT:    [[C1:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 42, i8 poison>
+; CHECK-NEXT:    [[C2:%.*]] = icmp slt <2 x i8> [[Y:%.*]], [[X]]
+; CHECK-NEXT:    [[R:%.*]] = or <2 x i1> [[C1]], [[C2]]
+; CHECK-NEXT:    ret <2 x i1> [[R]]
+;
+  %c1 = icmp ne <2 x i8> %x, <i8 42, i8 poison>
+  %c2 = icmp slt <2 x i8> %y, %x
+  %r = or <2 x i1> %c1, %c2
+  ret <2 x i1> %r
+}
+
 define <2 x i1> @substitute_constant_or_ne_slt_swap_vec_logical(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @substitute_constant_or_ne_slt_swap_vec_logical(
-; CHECK-NEXT:    [[C1:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 42, i8 undef>
+; CHECK-NEXT:    [[C1:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 42, i8 poison>
 ; CHECK-NEXT:    [[C2:%.*]] = icmp slt <2 x i8> [[Y:%.*]], [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[C1]], <2 x i1> <i1 true, i1 true>, <2 x i1> [[C2]]
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %c1 = icmp ne <2 x i8> %x, <i8 42, i8 undef>
+  %c1 = icmp ne <2 x i8> %x, <i8 42, i8 poison>
   %c2 = icmp slt <2 x i8> %y, %x
   %r = select <2 x i1> %c1, <2 x i1> <i1 true, i1 true>, <2 x i1> %c2
   ret <2 x i1> %r
@@ -2497,29 +2512,29 @@ define <2 x i1> @icmp_eq_m1_and_eq_m1(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[TMP1]], <i8 -1, i8 -1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 undef>
-  %ry = icmp eq <2 x i8> %y, <i8 -1, i8 undef>
+  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 poison>
+  %ry = icmp eq <2 x i8> %y, <i8 -1, i8 poison>
   %r = and <2 x i1> %rx, %ry
   ret <2 x i1> %r
 }
 
-define <2 x i1> @icmp_eq_m1_and_eq_undef_m1(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @icmp_eq_m1_and_eq_undef_m1(
+define <2 x i1> @icmp_eq_m1_and_eq_poison_m1(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @icmp_eq_m1_and_eq_poison_m1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[TMP1]], <i8 -1, i8 -1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 undef>
-  %ry = icmp eq <2 x i8> %y, <i8 undef, i8 -1>
+  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 poison>
+  %ry = icmp eq <2 x i8> %y, <i8 poison, i8 -1>
   %r = and <2 x i1> %rx, %ry
   ret <2 x i1> %r
 }
 
-define <2 x i1> @icmp_eq_undef_and_eq_m1_m2(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @icmp_eq_undef_and_eq_m1_m2(
-; CHECK-NEXT:    ret <2 x i1> zeroinitializer
+define <2 x i1> @icmp_eq_poison_and_eq_m1_m2(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @icmp_eq_poison_and_eq_m1_m2(
+; CHECK-NEXT:    ret <2 x i1> poison
 ;
-  %rx = icmp eq <2 x i8> %x, <i8 undef, i8 undef>
+  %rx = icmp eq <2 x i8> %x, <i8 poison, i8 poison>
   %ry = icmp eq <2 x i8> %y, <i8 -1, i8 -2>
   %r = and <2 x i1> %rx, %ry
   ret <2 x i1> %r
@@ -2527,13 +2542,13 @@ define <2 x i1> @icmp_eq_undef_and_eq_m1_m2(<2 x i8> %x, <2 x i8> %y) {
 
 define <2 x i1> @icmp_ne_m1_and_ne_m1_fail(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @icmp_ne_m1_and_ne_m1_fail(
-; CHECK-NEXT:    [[RX:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 -1, i8 undef>
-; CHECK-NEXT:    [[RY:%.*]] = icmp ne <2 x i8> [[Y:%.*]], <i8 -1, i8 undef>
+; CHECK-NEXT:    [[RX:%.*]] = icmp ne <2 x i8> [[X:%.*]], <i8 -1, i8 poison>
+; CHECK-NEXT:    [[RY:%.*]] = icmp ne <2 x i8> [[Y:%.*]], <i8 -1, i8 poison>
 ; CHECK-NEXT:    [[R:%.*]] = and <2 x i1> [[RX]], [[RY]]
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %rx = icmp ne <2 x i8> %x, <i8 -1, i8 undef>
-  %ry = icmp ne <2 x i8> %y, <i8 -1, i8 undef>
+  %rx = icmp ne <2 x i8> %x, <i8 -1, i8 poison>
+  %ry = icmp ne <2 x i8> %y, <i8 -1, i8 poison>
   %r = and <2 x i1> %rx, %ry
   ret <2 x i1> %r
 }
@@ -2541,13 +2556,13 @@ define <2 x i1> @icmp_ne_m1_and_ne_m1_fail(<2 x i8> %x, <2 x i8> %y) {
 
 define <2 x i1> @icmp_eq_m1_or_eq_m1_fail(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @icmp_eq_m1_or_eq_m1_fail(
-; CHECK-NEXT:    [[RX:%.*]] = icmp eq <2 x i8> [[X:%.*]], <i8 -1, i8 undef>
-; CHECK-NEXT:    [[RY:%.*]] = icmp eq <2 x i8> [[Y:%.*]], <i8 -1, i8 undef>
+; CHECK-NEXT:    [[RX:%.*]] = icmp eq <2 x i8> [[X:%.*]], <i8 -1, i8 poison>
+; CHECK-NEXT:    [[RY:%.*]] = icmp eq <2 x i8> [[Y:%.*]], <i8 -1, i8 poison>
 ; CHECK-NEXT:    [[R:%.*]] = or <2 x i1> [[RX]], [[RY]]
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 undef>
-  %ry = icmp eq <2 x i8> %y, <i8 -1, i8 undef>
+  %rx = icmp eq <2 x i8> %x, <i8 -1, i8 poison>
+  %ry = icmp eq <2 x i8> %y, <i8 -1, i8 poison>
   %r = or <2 x i1> %rx, %ry
   ret <2 x i1> %r
 }
@@ -2560,7 +2575,7 @@ define <2 x i1> @icmp_ne_m1_or_ne_m1(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %rx = icmp ne <2 x i8> %x, <i8 -1, i8 -1>
-  %ry = icmp ne <2 x i8> %y, <i8 -1, i8 undef>
+  %ry = icmp ne <2 x i8> %y, <i8 -1, i8 poison>
   %r = or <2 x i1> %rx, %ry
   ret <2 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/and-xor-or.ll b/llvm/test/Transforms/InstCombine/and-xor-or.ll
index d072dc15cbb2c..b26d6e16c2db2 100644
--- a/llvm/test/Transforms/InstCombine/and-xor-or.ll
+++ b/llvm/test/Transforms/InstCombine/and-xor-or.ll
@@ -843,7 +843,7 @@ define <2 x i6> @not_or_or_not_2i6(<2 x i6> %a0, <2 x i6> %b, <2 x i6> %c) {
 ;
   %a = sdiv <2 x i6> <i6 3, i6 3>, %a0 ; thwart complexity-based canonicalization
   %not1 = xor <2 x i6> %b, <i6 -1, i6 -1>
-  %not2 = xor <2 x i6> %c, <i6 -1, i6 undef>
+  %not2 = xor <2 x i6> %c, <i6 -1, i6 poison>
   %or1 = or <2 x i6> %a, %not1
   %or2 = or <2 x i6> %or1, %not2
   ret <2 x i6> %or2
@@ -4018,7 +4018,7 @@ define <2 x i4> @and_orn_xor_commute1(<2 x i4> %a, <2 x i4> %b) {
 ; CHECK-NEXT:    ret <2 x i4> [[R]]
 ;
   %xor = xor <2 x i4> %a, %b
-  %nota = xor <2 x i4> %a, <i4 -1, i4 undef>
+  %nota = xor <2 x i4> %a, <i4 -1, i4 poison>
   %or = or <2 x i4> %nota, %b
   %r = and <2 x i4> %xor, %or
   ret <2 x i4> %r
diff --git a/llvm/test/Transforms/InstCombine/and.ll b/llvm/test/Transforms/InstCombine/and.ll
index ffd8c2a06c86e..b5250fc1a7849 100644
--- a/llvm/test/Transforms/InstCombine/and.ll
+++ b/llvm/test/Transforms/InstCombine/and.ll
@@ -752,16 +752,16 @@ define <2 x i64> @test36_uniform(<2 x i32> %X) {
   ret <2 x i64> %res
 }
 
-define <2 x i64> @test36_undef(<2 x i32> %X) {
-; CHECK-LABEL: @test36_undef(
+define <2 x i64> @test36_poison(<2 x i32> %X) {
+; CHECK-LABEL: @test36_poison(
 ; CHECK-NEXT:    [[ZEXT:%.*]] = zext <2 x i32> [[X:%.*]] to <2 x i64>
-; CHECK-NEXT:    [[ZSUB:%.*]] = add <2 x i64> [[ZEXT]], <i64 7, i64 undef>
-; CHECK-NEXT:    [[RES:%.*]] = and <2 x i64> [[ZSUB]], <i64 240, i64 undef>
+; CHECK-NEXT:    [[ZSUB:%.*]] = add nuw nsw <2 x i64> [[ZEXT]], <i64 7, i64 poison>
+; CHECK-NEXT:    [[RES:%.*]] = and <2 x i64> [[ZSUB]], <i64 240, i64 poison>
 ; CHECK-NEXT:    ret <2 x i64> [[RES]]
 ;
   %zext = zext <2 x i32> %X to <2 x i64>
-  %zsub = add <2 x i64> %zext, <i64 7, i64 undef>
-  %res = and <2 x i64> %zsub, <i64 240, i64 undef>
+  %zsub = add <2 x i64> %zext, <i64 7, i64 poison>
+  %res = and <2 x i64> %zsub, <i64 240, i64 poison>
   ret <2 x i64> %res
 }
 
@@ -1630,16 +1630,16 @@ define <2 x i8> @lowmask_add_splat(<2 x i8> %x, ptr %p) {
   ret <2 x i8> %r
 }
 
-define <2 x i8> @lowmask_add_splat_undef(<2 x i8> %x, ptr %p) {
-; CHECK-LABEL: @lowmask_add_splat_undef(
-; CHECK-NEXT:    [[A:%.*]] = add <2 x i8> [[X:%.*]], <i8 -64, i8 undef>
+define <2 x i8> @lowmask_add_splat_poison(<2 x i8> %x, ptr %p) {
+; CHECK-LABEL: @lowmask_add_splat_poison(
+; CHECK-NEXT:    [[A:%.*]] = add <2 x i8> [[X:%.*]], <i8 -64, i8 poison>
 ; CHECK-NEXT:    store <2 x i8> [[A]], ptr [[P:%.*]], align 2
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i8> [[A]], <i8 undef, i8 32>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i8> [[X]], <i8 poison, i8 32>
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %a = add <2 x i8> %x, <i8 -64, i8 undef> ; 0xc0
+  %a = add <2 x i8> %x, <i8 -64, i8 poison> ; 0xc0
   store <2 x i8> %a, ptr %p
-  %r = and <2 x i8> %a, <i8 undef, i8 32> ; 0x20
+  %r = and <2 x i8> %a, <i8 poison, i8 32> ; 0x20
   ret <2 x i8> %r
 }
 
@@ -1679,14 +1679,14 @@ define <2 x i8> @flip_masked_bit_uniform(<2 x i8> %A) {
   ret <2 x i8> %C
 }
 
-define <2 x i8> @flip_masked_bit_undef(<2 x i8> %A) {
-; CHECK-LABEL: @flip_masked_bit_undef(
+define <2 x i8> @flip_masked_bit_poison(<2 x i8> %A) {
+; CHECK-LABEL: @flip_masked_bit_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <2 x i8> [[A:%.*]], <i8 -1, i8 -1>
-; CHECK-NEXT:    [[C:%.*]] = and <2 x i8> [[TMP1]], <i8 16, i8 undef>
+; CHECK-NEXT:    [[C:%.*]] = and <2 x i8> [[TMP1]], <i8 16, i8 poison>
 ; CHECK-NEXT:    ret <2 x i8> [[C]]
 ;
-  %B = add <2 x i8> %A, <i8 16, i8 undef>
-  %C = and <2 x i8> %B, <i8 16, i8 undef>
+  %B = add <2 x i8> %A, <i8 16, i8 poison>
+  %C = and <2 x i8> %B, <i8 16, i8 poison>
   ret <2 x i8> %C
 }
 
@@ -2004,7 +2004,7 @@ define i16 @invert_signbit_splat_mask_use2(i8 %x, i16 %y) {
   ret i16 %r
 }
 
-; extra use of sext is ok 
+; extra use of sext is ok
 
 define i16 @invert_signbit_splat_mask_use3(i8 %x, i16 %y) {
 ; CHECK-LABEL: @invert_signbit_splat_mask_use3(
@@ -2120,41 +2120,40 @@ define <3 x i16> @shl_lshr_pow2_const_case1_non_uniform_vec_negative(<3 x i16> %
   ret <3 x i16> %r
 }
 
-define <3 x i16> @shl_lshr_pow2_const_case1_undef1_vec(<3 x i16> %x) {
-; CHECK-LABEL: @shl_lshr_pow2_const_case1_undef1_vec(
+define <3 x i16> @shl_lshr_pow2_const_case1_poison1_vec(<3 x i16> %x) {
+; CHECK-LABEL: @shl_lshr_pow2_const_case1_poison1_vec(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <3 x i16> [[X:%.*]], <i16 8, i16 4, i16 4>
 ; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP1]], <3 x i16> <i16 8, i16 8, i16 8>, <3 x i16> zeroinitializer
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
-  %shl = shl <3 x i16> <i16 undef, i16 16, i16 16>, %x
+  %shl = shl <3 x i16> <i16 poison, i16 16, i16 16>, %x
   %lshr = lshr <3 x i16> %shl, <i16 5, i16 5, i16 5>
   %r = and <3 x i16> %lshr, <i16 8, i16 8, i16 8>
   ret <3 x i16> %r
 }
 
-define <3 x i16> @shl_lshr_pow2_const_case1_undef2_vec(<3 x i16> %x) {
-; CHECK-LABEL: @shl_lshr_pow2_const_case1_undef2_vec(
-; CHECK-NEXT:    [[SHL:%.*]] = shl <3 x i16> <i16 16, i16 16, i16 16>, [[X:%.*]]
-; CHECK-NEXT:    [[LSHR:%.*]] = lshr <3 x i16> [[SHL]], <i16 undef, i16 5, i16 5>
-; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[LSHR]], <i16 8, i16 8, i16 8>
+define <3 x i16> @shl_lshr_pow2_const_case1_poison2_vec(<3 x i16> %x) {
+; CHECK-LABEL: @shl_lshr_pow2_const_case1_poison2_vec(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <3 x i16> [[X:%.*]], <i16 poison, i16 4, i16 4>
+; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP1]], <3 x i16> <i16 8, i16 8, i16 8>, <3 x i16> zeroinitializer
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
   %shl = shl <3 x i16> <i16 16, i16 16, i16 16>, %x
-  %lshr = lshr <3 x i16> %shl, <i16 undef, i16 5, i16 5>
+  %lshr = lshr <3 x i16> %shl, <i16 poison, i16 5, i16 5>
   %r = and <3 x i16> %lshr, <i16 8, i16 8, i16 8>
   ret <3 x i16> %r
 }
 
-define <3 x i16> @shl_lshr_pow2_const_case1_undef3_vec(<3 x i16> %x) {
-; CHECK-LABEL: @shl_lshr_pow2_const_case1_undef3_vec(
+define <3 x i16> @shl_lshr_pow2_const_case1_poison3_vec(<3 x i16> %x) {
+; CHECK-LABEL: @shl_lshr_pow2_const_case1_poison3_vec(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <3 x i16> <i16 16, i16 16, i16 16>, [[X:%.*]]
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <3 x i16> [[SHL]], <i16 5, i16 5, i16 5>
-; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[LSHR]], <i16 undef, i16 8, i16 8>
+; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[LSHR]], <i16 poison, i16 8, i16 8>
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
   %shl = shl <3 x i16> <i16 16, i16 16, i16 16>, %x
   %lshr = lshr <3 x i16> %shl, <i16 5, i16 5, i16 5>
-  %r = and <3 x i16> %lshr, <i16 undef, i16 8, i16 8>
+  %r = and <3 x i16> %lshr, <i16 poison, i16 8, i16 8>
   ret <3 x i16> %r
 }
 
@@ -2417,40 +2416,41 @@ define <3 x i16> @lshr_shl_pow2_const_case1_non_uniform_vec_negative(<3 x i16> %
   ret <3 x i16> %r
 }
 
-define <3 x i16> @lshr_shl_pow2_const_case1_undef1_vec(<3 x i16> %x) {
-; CHECK-LABEL: @lshr_shl_pow2_const_case1_undef1_vec(
+define <3 x i16> @lshr_shl_pow2_const_case1_poison1_vec(<3 x i16> %x) {
+; CHECK-LABEL: @lshr_shl_pow2_const_case1_poison1_vec(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <3 x i16> [[X:%.*]], <i16 -1, i16 12, i16 12>
 ; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP1]], <3 x i16> <i16 128, i16 128, i16 128>, <3 x i16> zeroinitializer
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
-  %lshr = lshr <3 x i16> <i16 undef, i16 8192, i16 8192>, %x
+  %lshr = lshr <3 x i16> <i16 poison, i16 8192, i16 8192>, %x
   %shl = shl <3 x i16> %lshr, <i16 6, i16 6, i16 6>
   %r = and <3 x i16> %shl, <i16 128, i16 128, i16 128>
   ret <3 x i16> %r
 }
 
-define <3 x i16> @lshr_shl_pow2_const_case1_undef2_vec(<3 x i16> %x) {
-; CHECK-LABEL: @lshr_shl_pow2_const_case1_undef2_vec(
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <3 x i16> [[X:%.*]], <i16 undef, i16 12, i16 12>
-; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP1]], <3 x i16> <i16 128, i16 128, i16 128>, <3 x i16> zeroinitializer
+define <3 x i16> @lshr_shl_pow2_const_case1_poison2_vec(<3 x i16> %x) {
+; CHECK-LABEL: @lshr_shl_pow2_const_case1_poison2_vec(
+; CHECK-NEXT:    [[LSHR:%.*]] = lshr <3 x i16> <i16 8192, i16 8192, i16 8192>, [[X:%.*]]
+; CHECK-NEXT:    [[SHL:%.*]] = shl <3 x i16> [[LSHR]], <i16 poison, i16 6, i16 6>
+; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[SHL]], <i16 128, i16 128, i16 128>
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
   %lshr = lshr <3 x i16> <i16 8192, i16 8192, i16 8192>, %x
-  %shl = shl <3 x i16> %lshr, <i16 undef, i16 6, i16 6>
+  %shl = shl <3 x i16> %lshr, <i16 poison, i16 6, i16 6>
   %r = and <3 x i16> %shl, <i16 128, i16 128, i16 128>
   ret <3 x i16> %r
 }
 
-define <3 x i16> @lshr_shl_pow2_const_case1_undef3_vec(<3 x i16> %x) {
-; CHECK-LABEL: @lshr_shl_pow2_const_case1_undef3_vec(
+define <3 x i16> @lshr_shl_pow2_const_case1_poison3_vec(<3 x i16> %x) {
+; CHECK-LABEL: @lshr_shl_pow2_const_case1_poison3_vec(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <3 x i16> <i16 8192, i16 8192, i16 8192>, [[X:%.*]]
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <3 x i16> [[LSHR]], <i16 6, i16 6, i16 6>
-; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[SHL]], <i16 undef, i16 128, i16 128>
+; CHECK-NEXT:    [[R:%.*]] = and <3 x i16> [[SHL]], <i16 poison, i16 128, i16 128>
 ; CHECK-NEXT:    ret <3 x i16> [[R]]
 ;
   %lshr = lshr <3 x i16> <i16 8192, i16 8192, i16 8192>, %x
   %shl = shl <3 x i16> %lshr, <i16 6, i16 6, i16 6>
-  %r = and <3 x i16> %shl, <i16 undef, i16 128, i16 128>
+  %r = and <3 x i16> %shl, <i16 poison, i16 128, i16 128>
   ret <3 x i16> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/and2.ll b/llvm/test/Transforms/InstCombine/and2.ll
index 73bdadc86710e..104486e7638f5 100644
--- a/llvm/test/Transforms/InstCombine/and2.ll
+++ b/llvm/test/Transforms/InstCombine/and2.ll
@@ -168,14 +168,14 @@ define <2 x i8> @and1_shl1_is_cmp_eq_0_vec(<2 x i8> %x) {
   ret <2 x i8> %and
 }
 
-define <2 x i8> @and1_shl1_is_cmp_eq_0_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @and1_shl1_is_cmp_eq_0_vec_undef(
+define <2 x i8> @and1_shl1_is_cmp_eq_0_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @and1_shl1_is_cmp_eq_0_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <2 x i8> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    [[AND:%.*]] = zext <2 x i1> [[TMP1]] to <2 x i8>
 ; CHECK-NEXT:    ret <2 x i8> [[AND]]
 ;
-  %sh = shl <2 x i8> <i8 1, i8 undef>, %x
-  %and = and <2 x i8> %sh, <i8 1, i8 undef>
+  %sh = shl <2 x i8> <i8 1, i8 poison>, %x
+  %and = and <2 x i8> %sh, <i8 1, i8 poison>
   ret <2 x i8> %and
 }
 
@@ -215,14 +215,13 @@ define <2 x i8> @and1_lshr1_is_cmp_eq_0_vec(<2 x i8> %x) {
   ret <2 x i8> %and
 }
 
-define <2 x i8> @and1_lshr1_is_cmp_eq_0_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @and1_lshr1_is_cmp_eq_0_vec_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <2 x i8> [[X:%.*]], zeroinitializer
-; CHECK-NEXT:    [[AND:%.*]] = zext <2 x i1> [[TMP1]] to <2 x i8>
+define <2 x i8> @and1_lshr1_is_cmp_eq_0_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @and1_lshr1_is_cmp_eq_0_vec_poison(
+; CHECK-NEXT:    [[AND:%.*]] = lshr <2 x i8> <i8 1, i8 poison>, [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i8> [[AND]]
 ;
-  %sh = lshr <2 x i8> <i8 1, i8 undef>, %x
-  %and = and <2 x i8> %sh, <i8 1, i8 undef>
+  %sh = lshr <2 x i8> <i8 1, i8 poison>, %x
+  %and = and <2 x i8> %sh, <i8 1, i8 poison>
   ret <2 x i8> %and
 }
 
diff --git a/llvm/test/Transforms/InstCombine/ashr-lshr.ll b/llvm/test/Transforms/InstCombine/ashr-lshr.ll
index 60fa5b2597ba9..ac206dc7999dd 100644
--- a/llvm/test/Transforms/InstCombine/ashr-lshr.ll
+++ b/llvm/test/Transforms/InstCombine/ashr-lshr.ll
@@ -229,24 +229,24 @@ define <2 x i32> @ashr_lshr_inv_nonsplat_vec(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %ret
 }
 
-define <2 x i32> @ashr_lshr_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @ashr_lshr_vec_undef(
+define <2 x i32> @ashr_lshr_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @ashr_lshr_vec_poison(
 ; CHECK-NEXT:    [[CMP12:%.*]] = ashr <2 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[CMP12]]
 ;
-  %cmp = icmp sgt <2 x i32> %x, <i32 undef, i32 -1>
+  %cmp = icmp sgt <2 x i32> %x, <i32 poison, i32 -1>
   %l = lshr <2 x i32> %x, %y
   %r = ashr exact <2 x i32> %x, %y
   %ret = select <2 x i1> %cmp, <2 x i32> %l, <2 x i32> %r
   ret <2 x i32> %ret
 }
 
-define <2 x i32> @ashr_lshr_vec_undef2(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @ashr_lshr_vec_undef2(
+define <2 x i32> @ashr_lshr_vec_poison2(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @ashr_lshr_vec_poison2(
 ; CHECK-NEXT:    [[CMP1:%.*]] = ashr exact <2 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[CMP1]]
 ;
-  %cmp = icmp slt <2 x i32> %x, <i32 1, i32 undef>
+  %cmp = icmp slt <2 x i32> %x, <i32 1, i32 poison>
   %l = lshr exact <2 x i32> %x, %y
   %r = ashr exact <2 x i32> %x, %y
   %ret = select <2 x i1> %cmp, <2 x i32> %r, <2 x i32> %l
@@ -498,14 +498,14 @@ define <3 x i42> @lshr_sub_nsw_splat(<3 x i42> %x, <3 x i42> %y) {
   ret <3 x i42> %shr
 }
 
-define <3 x i42> @lshr_sub_nsw_splat_undef(<3 x i42> %x, <3 x i42> %y) {
-; CHECK-LABEL: @lshr_sub_nsw_splat_undef(
+define <3 x i42> @lshr_sub_nsw_splat_poison(<3 x i42> %x, <3 x i42> %y) {
+; CHECK-LABEL: @lshr_sub_nsw_splat_poison(
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <3 x i42> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[SHR:%.*]] = lshr <3 x i42> [[SUB]], <i42 41, i42 undef, i42 41>
+; CHECK-NEXT:    [[SHR:%.*]] = lshr <3 x i42> [[SUB]], <i42 41, i42 poison, i42 41>
 ; CHECK-NEXT:    ret <3 x i42> [[SHR]]
 ;
   %sub = sub nsw <3 x i42> %x, %y
-  %shr = lshr <3 x i42> %sub, <i42 41, i42 undef, i42 41>
+  %shr = lshr <3 x i42> %sub, <i42 41, i42 poison, i42 41>
   ret <3 x i42> %shr
 }
 
@@ -572,14 +572,14 @@ define <3 x i43> @ashr_sub_nsw_splat(<3 x i43> %x, <3 x i43> %y) {
   ret <3 x i43> %shr
 }
 
-define <3 x i43> @ashr_sub_nsw_splat_undef(<3 x i43> %x, <3 x i43> %y) {
-; CHECK-LABEL: @ashr_sub_nsw_splat_undef(
+define <3 x i43> @ashr_sub_nsw_splat_poison(<3 x i43> %x, <3 x i43> %y) {
+; CHECK-LABEL: @ashr_sub_nsw_splat_poison(
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nsw <3 x i43> [[X:%.*]], [[Y:%.*]]
-; CHECK-NEXT:    [[SHR:%.*]] = ashr <3 x i43> [[SUB]], <i43 42, i43 undef, i43 42>
+; CHECK-NEXT:    [[SHR:%.*]] = ashr <3 x i43> [[SUB]], <i43 42, i43 poison, i43 42>
 ; CHECK-NEXT:    ret <3 x i43> [[SHR]]
 ;
   %sub = sub nsw <3 x i43> %x, %y
-  %shr = ashr <3 x i43> %sub, <i43 42, i43 undef, i43 42>
+  %shr = ashr <3 x i43> %sub, <i43 42, i43 poison, i43 42>
   ret <3 x i43> %shr
 }
 
diff --git a/llvm/test/Transforms/InstCombine/ashr-or-mul-abs.ll b/llvm/test/Transforms/InstCombine/ashr-or-mul-abs.ll
index 3cf312e426edf..46a7f2f1189e2 100644
--- a/llvm/test/Transforms/InstCombine/ashr-or-mul-abs.ll
+++ b/llvm/test/Transforms/InstCombine/ashr-or-mul-abs.ll
@@ -62,13 +62,13 @@ define <4 x i32> @ashr_or_mul_to_abs_vec2(<4 x i32> %X) {
   ret <4 x i32> %i2
 }
 
-define <4 x i32> @ashr_or_mul_to_abs_vec3_undef(<4 x i32> %X) {
-; CHECK-LABEL: @ashr_or_mul_to_abs_vec3_undef(
+define <4 x i32> @ashr_or_mul_to_abs_vec3_poison(<4 x i32> %X) {
+; CHECK-LABEL: @ashr_or_mul_to_abs_vec3_poison(
 ; CHECK-NEXT:    [[I2:%.*]] = call <4 x i32> @llvm.abs.v4i32(<4 x i32> [[X:%.*]], i1 false)
 ; CHECK-NEXT:    ret <4 x i32> [[I2]]
 ;
-  %i = ashr <4 x i32> %X, <i32 31, i32 undef, i32 31, i32 31>
-  %i1 = or <4 x i32> %i, <i32 1, i32 1, i32 1, i32 undef>
+  %i = ashr <4 x i32> %X, <i32 31, i32 poison, i32 31, i32 31>
+  %i1 = or <4 x i32> %i, <i32 1, i32 1, i32 1, i32 poison>
   %i2 = mul <4 x i32> %i1, %X
   ret <4 x i32> %i2
 }
diff --git a/llvm/test/Transforms/InstCombine/binop-and-shifts.ll b/llvm/test/Transforms/InstCombine/binop-and-shifts.ll
index 148963894b89f..f776dc13bb4e5 100644
--- a/llvm/test/Transforms/InstCombine/binop-and-shifts.ll
+++ b/llvm/test/Transforms/InstCombine/binop-and-shifts.ll
@@ -178,27 +178,27 @@ define <2 x i8> @shl_xor_and(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @shl_xor_and(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <2 x i8> [[Y:%.*]], <i8 11, i8 poison>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <2 x i8> [[TMP1]], [[X:%.*]]
-; CHECK-NEXT:    [[BW1:%.*]] = shl <2 x i8> [[TMP2]], <i8 2, i8 undef>
+; CHECK-NEXT:    [[BW1:%.*]] = shl <2 x i8> [[TMP2]], <i8 2, i8 poison>
 ; CHECK-NEXT:    ret <2 x i8> [[BW1]]
 ;
-  %shift1 = shl <2 x i8> %x, <i8 2, i8 undef>
-  %shift2 = shl <2 x i8> %y, <i8 2, i8 undef>
-  %bw2 = xor <2 x i8> %shift2, <i8 44, i8 undef>
+  %shift1 = shl <2 x i8> %x, <i8 2, i8 poison>
+  %shift2 = shl <2 x i8> %y, <i8 2, i8 poison>
+  %bw2 = xor <2 x i8> %shift2, <i8 44, i8 poison>
   %bw1 = and <2 x i8> %bw2, %shift1
   ret <2 x i8> %bw1
 }
 
 define <2 x i8> @shl_xor_and_fail(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @shl_xor_and_fail(
-; CHECK-NEXT:    [[SHIFT1:%.*]] = shl <2 x i8> [[X:%.*]], <i8 2, i8 undef>
-; CHECK-NEXT:    [[SHIFT2:%.*]] = shl <2 x i8> [[Y:%.*]], <i8 undef, i8 2>
-; CHECK-NEXT:    [[BW2:%.*]] = xor <2 x i8> [[SHIFT2]], <i8 44, i8 undef>
+; CHECK-NEXT:    [[SHIFT1:%.*]] = shl <2 x i8> [[X:%.*]], <i8 2, i8 poison>
+; CHECK-NEXT:    [[SHIFT2:%.*]] = shl <2 x i8> [[Y:%.*]], <i8 poison, i8 2>
+; CHECK-NEXT:    [[BW2:%.*]] = xor <2 x i8> [[SHIFT2]], <i8 44, i8 poison>
 ; CHECK-NEXT:    [[BW1:%.*]] = and <2 x i8> [[SHIFT1]], [[BW2]]
 ; CHECK-NEXT:    ret <2 x i8> [[BW1]]
 ;
-  %shift1 = shl <2 x i8> %x, <i8 2, i8 undef>
-  %shift2 = shl <2 x i8> %y, <i8 undef, i8 2>
-  %bw2 = xor <2 x i8> %shift2, <i8 44, i8 undef>
+  %shift1 = shl <2 x i8> %x, <i8 2, i8 poison>
+  %shift2 = shl <2 x i8> %y, <i8 poison, i8 2>
+  %bw2 = xor <2 x i8> %shift2, <i8 44, i8 poison>
   %bw1 = and <2 x i8> %shift1, %bw2
   ret <2 x i8> %bw1
 }
@@ -321,13 +321,13 @@ define <2 x i8> @lshr_add_and(<2 x i8> %x, <2 x i8> %y) {
 define <2 x i8> @lshr_add_or_fail_dif_masks(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: @lshr_add_or_fail_dif_masks(
 ; CHECK-NEXT:    [[SHIFT1:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 3, i8 4>
-; CHECK-NEXT:    [[SHIFT2:%.*]] = lshr <2 x i8> [[Y:%.*]], <i8 undef, i8 3>
-; CHECK-NEXT:    [[BW2:%.*]] = add <2 x i8> [[SHIFT2]], <i8 -1, i8 1>
+; CHECK-NEXT:    [[SHIFT2:%.*]] = lshr <2 x i8> [[Y:%.*]], <i8 poison, i8 3>
+; CHECK-NEXT:    [[BW2:%.*]] = add nsw <2 x i8> [[SHIFT2]], <i8 -1, i8 1>
 ; CHECK-NEXT:    [[BW1:%.*]] = and <2 x i8> [[SHIFT1]], [[BW2]]
 ; CHECK-NEXT:    ret <2 x i8> [[BW1]]
 ;
   %shift1 = lshr <2 x i8> %x, <i8 3, i8 4>
-  %shift2 = lshr <2 x i8> %y, <i8 undef, i8 3>
+  %shift2 = lshr <2 x i8> %y, <i8 poison, i8 3>
   %bw2 = add <2 x i8> %shift2, <i8 255, i8 1>
   %bw1 = and <2 x i8> %shift1, %bw2
   ret <2 x i8> %bw1
@@ -659,8 +659,8 @@ define <4 x i8> @and_ashr_not_vec_commuted(<4 x i8> %x, <4 x i8> %y, <4 x i8> %s
   ret <4 x i8> %and
 }
 
-define <4 x i8> @and_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @and_ashr_not_vec_undef_1(
+define <4 x i8> @and_ashr_not_vec_poison_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @and_ashr_not_vec_poison_1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i8> [[Y:%.*]], <i8 -1, i8 -1, i8 -1, i8 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <4 x i8> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    [[AND:%.*]] = ashr <4 x i8> [[TMP2]], [[SHAMT:%.*]]
@@ -668,18 +668,18 @@ define <4 x i8> @and_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %sh
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 poison, i8 poison, i8 poison>
   %and = and <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %and
 }
 
-define <4 x i8> @and_ashr_not_vec_undef_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @and_ashr_not_vec_undef_2(
-; CHECK-NEXT:    ret <4 x i8> zeroinitializer
+define <4 x i8> @and_ashr_not_vec_poison_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @and_ashr_not_vec_poison_2(
+; CHECK-NEXT:    ret <4 x i8> poison
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 undef, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 poison, i8 poison, i8 poison, i8 poison>
   %and = and <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %and
 }
@@ -793,8 +793,8 @@ define <4 x i8> @or_ashr_not_vec_commuted(<4 x i8> %x, <4 x i8> %y, <4 x i8> %sh
   ret <4 x i8> %or
 }
 
-define <4 x i8> @or_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @or_ashr_not_vec_undef_1(
+define <4 x i8> @or_ashr_not_vec_poison_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @or_ashr_not_vec_poison_1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i8> [[Y:%.*]], <i8 -1, i8 -1, i8 -1, i8 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = or <4 x i8> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    [[OR:%.*]] = ashr <4 x i8> [[TMP2]], [[SHAMT:%.*]]
@@ -802,18 +802,18 @@ define <4 x i8> @or_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %sha
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 poison, i8 poison, i8 poison>
   %or = or <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %or
 }
 
-define <4 x i8> @or_ashr_not_vec_undef_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @or_ashr_not_vec_undef_2(
-; CHECK-NEXT:    ret <4 x i8> <i8 -1, i8 -1, i8 -1, i8 -1>
+define <4 x i8> @or_ashr_not_vec_poison_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @or_ashr_not_vec_poison_2(
+; CHECK-NEXT:    ret <4 x i8> poison
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 undef, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 poison, i8 poison, i8 poison, i8 poison>
   %or = or <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %or
 }
@@ -926,8 +926,8 @@ define <4 x i8> @xor_ashr_not_vec_commuted(<4 x i8> %x, <4 x i8> %y, <4 x i8> %s
   ret <4 x i8> %xor
 }
 
-define <4 x i8> @xor_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @xor_ashr_not_vec_undef_1(
+define <4 x i8> @xor_ashr_not_vec_poison_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @xor_ashr_not_vec_poison_1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i8> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    [[DOTNOT:%.*]] = ashr <4 x i8> [[TMP1]], [[SHAMT:%.*]]
 ; CHECK-NEXT:    [[XOR:%.*]] = xor <4 x i8> [[DOTNOT]], <i8 -1, i8 -1, i8 -1, i8 -1>
@@ -935,18 +935,18 @@ define <4 x i8> @xor_ashr_not_vec_undef_1(<4 x i8> %x, <4 x i8> %y, <4 x i8> %sh
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 -1, i8 poison, i8 poison, i8 poison>
   %xor = xor <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %xor
 }
 
-define <4 x i8> @xor_ashr_not_vec_undef_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
-; CHECK-LABEL: @xor_ashr_not_vec_undef_2(
-; CHECK-NEXT:    ret <4 x i8> undef
+define <4 x i8> @xor_ashr_not_vec_poison_2(<4 x i8> %x, <4 x i8> %y, <4 x i8> %shamt) {
+; CHECK-LABEL: @xor_ashr_not_vec_poison_2(
+; CHECK-NEXT:    ret <4 x i8> poison
 ;
   %x.shift = ashr <4 x i8> %x, %shamt
   %y.shift = ashr <4 x i8> %y, %shamt
-  %y.shift.not = xor <4 x i8> %y.shift, <i8 undef, i8 undef, i8 undef, i8 undef>
+  %y.shift.not = xor <4 x i8> %y.shift, <i8 poison, i8 poison, i8 poison, i8 poison>
   %xor = xor <4 x i8> %x.shift, %y.shift.not
   ret <4 x i8> %xor
 }
diff --git a/llvm/test/Transforms/InstCombine/binop-of-displaced-shifts.ll b/llvm/test/Transforms/InstCombine/binop-of-displaced-shifts.ll
index 27a3c8743368a..a16ad4ddb806f 100644
--- a/llvm/test/Transforms/InstCombine/binop-of-displaced-shifts.ll
+++ b/llvm/test/Transforms/InstCombine/binop-of-displaced-shifts.ll
@@ -202,41 +202,41 @@ define <2 x i8> @shl_or_non_splat(<2 x i8> %x) {
   ret <2 x i8> %binop
 }
 
-define <2 x i8> @shl_or_undef_in_add(<2 x i8> %x) {
-; CHECK-LABEL: define <2 x i8> @shl_or_undef_in_add
+define <2 x i8> @shl_or_poison_in_add(<2 x i8> %x) {
+; CHECK-LABEL: define <2 x i8> @shl_or_poison_in_add
 ; CHECK-SAME: (<2 x i8> [[X:%.*]]) {
 ; CHECK-NEXT:    [[BINOP:%.*]] = shl <2 x i8> <i8 22, i8 poison>, [[X]]
 ; CHECK-NEXT:    ret <2 x i8> [[BINOP]]
 ;
   %shift = shl <2 x i8> <i8 16, i8 16>, %x
-  %add = add <2 x i8> %x, <i8 1, i8 undef>
+  %add = add <2 x i8> %x, <i8 1, i8 poison>
   %shift2 = shl <2 x i8> <i8 3, i8 3>, %add
   %binop = or <2 x i8> %shift, %shift2
   ret <2 x i8> %binop
 }
 
-define <2 x i8> @shl_or_undef_in_shift1(<2 x i8> %x) {
-; CHECK-LABEL: define <2 x i8> @shl_or_undef_in_shift1
+define <2 x i8> @shl_or_poison_in_shift1(<2 x i8> %x) {
+; CHECK-LABEL: define <2 x i8> @shl_or_poison_in_shift1
 ; CHECK-SAME: (<2 x i8> [[X:%.*]]) {
-; CHECK-NEXT:    [[BINOP:%.*]] = shl <2 x i8> <i8 22, i8 -1>, [[X]]
+; CHECK-NEXT:    [[BINOP:%.*]] = shl <2 x i8> <i8 22, i8 poison>, [[X]]
 ; CHECK-NEXT:    ret <2 x i8> [[BINOP]]
 ;
-  %shift = shl <2 x i8> <i8 16, i8 undef>, %x
+  %shift = shl <2 x i8> <i8 16, i8 poison>, %x
   %add = add <2 x i8> %x, <i8 1, i8 1>
   %shift2 = shl <2 x i8> <i8 3, i8 3>, %add
   %binop = or <2 x i8> %shift, %shift2
   ret <2 x i8> %binop
 }
 
-define <2 x i8> @shl_or_undef_in_shift2(<2 x i8> %x) {
-; CHECK-LABEL: define <2 x i8> @shl_or_undef_in_shift2
+define <2 x i8> @shl_or_poison_in_shift2(<2 x i8> %x) {
+; CHECK-LABEL: define <2 x i8> @shl_or_poison_in_shift2
 ; CHECK-SAME: (<2 x i8> [[X:%.*]]) {
-; CHECK-NEXT:    [[BINOP:%.*]] = shl <2 x i8> <i8 22, i8 16>, [[X]]
+; CHECK-NEXT:    [[BINOP:%.*]] = shl <2 x i8> <i8 22, i8 poison>, [[X]]
 ; CHECK-NEXT:    ret <2 x i8> [[BINOP]]
 ;
   %shift = shl <2 x i8> <i8 16, i8 16>, %x
   %add = add <2 x i8> %x, <i8 1, i8 1>
-  %shift2 = shl <2 x i8> <i8 3, i8 undef>, %add
+  %shift2 = shl <2 x i8> <i8 3, i8 poison>, %add
   %binop = or <2 x i8> %shift, %shift2
   ret <2 x i8> %binop
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll b/llvm/test/Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll
index 4547008b76093..c555970ea4348 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-clamp-like-pattern-between-zero-and-positive-threshold.ll
@@ -338,22 +338,22 @@ define <2 x i32> @t18_ult_slt_vec_nonsplat(<2 x i32> %x, <2 x i32> %replacement_
   ret <2 x i32> %r
 }
 
-define <3 x i32> @t19_ult_slt_vec_undef0(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
-; CHECK-LABEL: @t19_ult_slt_vec_undef0(
+define <3 x i32> @t19_ult_slt_vec_poison0(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
+; CHECK-LABEL: @t19_ult_slt_vec_poison0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <3 x i32> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <3 x i32> [[X]], <i32 65535, i32 65535, i32 65535>
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <3 x i1> [[TMP1]], <3 x i32> [[REPLACEMENT_LOW:%.*]], <3 x i32> [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP2]], <3 x i32> [[REPLACEMENT_HIGH:%.*]], <3 x i32> [[TMP3]]
 ; CHECK-NEXT:    ret <3 x i32> [[R]]
 ;
-  %t0 = icmp slt <3 x i32> %x, <i32 65536, i32 undef, i32 65536>
+  %t0 = icmp slt <3 x i32> %x, <i32 65536, i32 poison, i32 65536>
   %t1 = select <3 x i1> %t0, <3 x i32> %replacement_low, <3 x i32> %replacement_high
   %t2 = icmp ult <3 x i32> %x, <i32 65536, i32 65536, i32 65536>
   %r = select <3 x i1> %t2, <3 x i32> %x, <3 x i32> %t1
   ret <3 x i32> %r
 }
-define <3 x i32> @t20_ult_slt_vec_undef1(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
-; CHECK-LABEL: @t20_ult_slt_vec_undef1(
+define <3 x i32> @t20_ult_slt_vec_poison1(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
+; CHECK-LABEL: @t20_ult_slt_vec_poison1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <3 x i32> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <3 x i32> [[X]], <i32 65535, i32 65535, i32 65535>
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <3 x i1> [[TMP1]], <3 x i32> [[REPLACEMENT_LOW:%.*]], <3 x i32> [[X]]
@@ -362,21 +362,21 @@ define <3 x i32> @t20_ult_slt_vec_undef1(<3 x i32> %x, <3 x i32> %replacement_lo
 ;
   %t0 = icmp slt <3 x i32> %x, <i32 65536, i32 65537, i32 65536>
   %t1 = select <3 x i1> %t0, <3 x i32> %replacement_low, <3 x i32> %replacement_high
-  %t2 = icmp ult <3 x i32> %x, <i32 65536, i32 undef, i32 65536>
+  %t2 = icmp ult <3 x i32> %x, <i32 65536, i32 poison, i32 65536>
   %r = select <3 x i1> %t2, <3 x i32> %x, <3 x i32> %t1
   ret <3 x i32> %r
 }
-define <3 x i32> @t21_ult_slt_vec_undef2(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
-; CHECK-LABEL: @t21_ult_slt_vec_undef2(
+define <3 x i32> @t21_ult_slt_vec_poison2(<3 x i32> %x, <3 x i32> %replacement_low, <3 x i32> %replacement_high) {
+; CHECK-LABEL: @t21_ult_slt_vec_poison2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <3 x i32> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sgt <3 x i32> [[X]], <i32 65535, i32 65535, i32 65535>
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <3 x i1> [[TMP1]], <3 x i32> [[REPLACEMENT_LOW:%.*]], <3 x i32> [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = select <3 x i1> [[TMP2]], <3 x i32> [[REPLACEMENT_HIGH:%.*]], <3 x i32> [[TMP3]]
 ; CHECK-NEXT:    ret <3 x i32> [[R]]
 ;
-  %t0 = icmp slt <3 x i32> %x, <i32 65536, i32 undef, i32 65536>
+  %t0 = icmp slt <3 x i32> %x, <i32 65536, i32 poison, i32 65536>
   %t1 = select <3 x i1> %t0, <3 x i32> %replacement_low, <3 x i32> %replacement_high
-  %t2 = icmp ult <3 x i32> %x, <i32 65536, i32 undef, i32 65536>
+  %t2 = icmp ult <3 x i32> %x, <i32 65536, i32 poison, i32 65536>
   %r = select <3 x i1> %t2, <3 x i32> %x, <3 x i32> %t1
   ret <3 x i32> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-eq-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
index 5b7a99d53c308..759770688cf20 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
@@ -79,12 +79,12 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ult <3 x i8> [[X:%.*]], <i8 4, i8 4, i8 4>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp eq <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
index 160d968b9ac4c..95e6d5ac6a5f8 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
@@ -79,22 +79,22 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X:%.*]], <i8 3, i8 3, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp ne <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_nonsplat_undef(
+define <3 x i1> @p3_vec_nonsplat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_nonsplat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X:%.*]], <i8 -1, i8 -1, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 -1, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 -1, i8 poison, i8 3>
   %ret = icmp ne <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
index 60921042d5243..ae503bfb1cfe2 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
@@ -58,12 +58,12 @@ define <2 x i1> @p2_vec_nonsplat_edgecase(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp slt <3 x i8> [[X:%.*]], <i8 4, i8 4, i8 4>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp sge <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
@@ -175,11 +175,11 @@ define <2 x i1> @n3_vec(<2 x i8> %x) {
 
 define <3 x i1> @n4_vec(<3 x i8> %x) {
 ; CHECK-LABEL: @n4_vec(
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X:%.*]], <i8 3, i8 undef, i8 -1>
+; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X:%.*]], <i8 3, i8 poison, i8 -1>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sge <3 x i8> [[TMP0]], [[X]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 -1>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 -1>
   %ret = icmp sge <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll
index 6345e70d7220e..f1333fed2c517 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sgt-to-icmp-sgt.ll
@@ -72,26 +72,26 @@ define <2 x i1> @p2_vec_nonsplat_edgecase() {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef() {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison() {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sgt <3 x i8> [[X]], <i8 3, i8 3, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp sgt <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_nonsplat_undef() {
-; CHECK-LABEL: @p3_vec_nonsplat_undef(
+define <3 x i1> @p3_vec_nonsplat_poison() {
+; CHECK-LABEL: @p3_vec_nonsplat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sgt <3 x i8> [[X]], <i8 15, i8 3, i8 15>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 15, i8 3, i8 undef>
+  %tmp0 = and <3 x i8> %x, <i8 15, i8 3, i8 poison>
   %ret = icmp sgt <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
@@ -212,12 +212,12 @@ define <2 x i1> @n3_vec() {
 define <3 x i1> @n4_vec() {
 ; CHECK-LABEL: @n4_vec(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X]], <i8 3, i8 undef, i8 -1>
+; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X]], <i8 3, i8 poison, i8 -1>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sgt <3 x i8> [[X]], [[TMP0]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 -1>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 -1>
   %ret = icmp sgt <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
index b7aec53fed676..4bed21a525f05 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sle-to-icmp-sle.ll
@@ -72,14 +72,14 @@ define <2 x i1> @p2_vec_nonsplat_edgecase() {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef() {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison() {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp slt <3 x i8> [[X]], <i8 4, i8 4, i8 4>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp sle <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
@@ -200,12 +200,12 @@ define <2 x i1> @n3_vec() {
 define <3 x i1> @n4_vec() {
 ; CHECK-LABEL: @n4_vec(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X]], <i8 3, i8 undef, i8 -1>
+; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X]], <i8 3, i8 poison, i8 -1>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sle <3 x i8> [[X]], [[TMP0]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 -1>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 -1>
   %ret = icmp sle <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll
index 56661d335c4f6..be6e3d0306bcd 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-slt-to-icmp-sgt.ll
@@ -58,22 +58,22 @@ define <2 x i1> @p2_vec_nonsplat_edgecase(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sgt <3 x i8> [[X:%.*]], <i8 3, i8 3, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp slt <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_nonsplat_undef(
+define <3 x i1> @p3_vec_nonsplat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_nonsplat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp sgt <3 x i8> [[X:%.*]], <i8 15, i8 15, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 undef, i8 15, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 poison, i8 15, i8 3>
   %ret = icmp slt <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
@@ -185,11 +185,11 @@ define <2 x i1> @n3(<2 x i8> %x) {
 
 define <3 x i1> @n4(<3 x i8> %x) {
 ; CHECK-LABEL: @n4(
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X:%.*]], <i8 3, i8 undef, i8 -1>
+; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X:%.*]], <i8 3, i8 poison, i8 -1>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp slt <3 x i8> [[TMP0]], [[X]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 -1>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 -1>
   %ret = icmp slt <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-uge-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-uge-to-icmp-ule.ll
index a93e8f779435f..cfd48821b2c1d 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-uge-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-uge-to-icmp-ule.ll
@@ -79,12 +79,12 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ult <3 x i8> [[X:%.*]], <i8 4, i8 4, i8 4>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp uge <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll
index 73ea4d456d246..6f6ba95a81c76 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ugt-to-icmp-ugt.ll
@@ -95,26 +95,26 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1() {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef() {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison() {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X]], <i8 3, i8 3, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp ugt <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_nonsplat_undef() {
-; CHECK-LABEL: @p3_vec_nonsplat_undef(
+define <3 x i1> @p3_vec_nonsplat_poison() {
+; CHECK-LABEL: @p3_vec_nonsplat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X]], <i8 3, i8 3, i8 15>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 15>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 15>
   %ret = icmp ugt <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ule-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ule-to-icmp-ule.ll
index 53886b5f2dc9c..54f00321c4cf0 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ule-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ule-to-icmp-ule.ll
@@ -95,14 +95,14 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1() {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef() {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison() {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[X:%.*]] = call <3 x i8> @gen3x8()
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ult <3 x i8> [[X]], <i8 4, i8 4, i8 4>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %x = call <3 x i8> @gen3x8()
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp ule <3 x i8> %x, %tmp0
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll
index d66be571008c2..008fc6d2d6eda 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-ult-to-icmp-ugt.ll
@@ -80,22 +80,22 @@ define <2 x i1> @p2_vec_nonsplat_edgecase1(<2 x i8> %x) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_splat_undef(
+define <3 x i1> @p3_vec_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_splat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X:%.*]], <i8 3, i8 3, i8 3>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
+  %tmp0 = and <3 x i8> %x, <i8 3, i8 poison, i8 3>
   %ret = icmp ult <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_nonsplat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @p3_vec_nonsplat_undef(
+define <3 x i1> @p3_vec_nonsplat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @p3_vec_nonsplat_poison(
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ugt <3 x i8> [[X:%.*]], <i8 7, i8 31, i8 7>
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = and <3 x i8> %x, <i8 7, i8 31, i8 undef>
+  %tmp0 = and <3 x i8> %x, <i8 7, i8 31, i8 poison>
   %ret = icmp ult <3 x i8> %tmp0, %x
   ret <3 x i1> %ret
 }
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-eq-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
index 38611d8b53a98..dc5658d302d99 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-eq-to-icmp-ule.ll
@@ -40,13 +40,13 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef(
-; CHECK-NEXT:    [[TMP0:%.*]] = lshr <3 x i8> <i8 -1, i8 undef, i8 -1>, [[Y:%.*]]
+define <3 x i1> @p2_vec_poison(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison(
+; CHECK-NEXT:    [[TMP0:%.*]] = lshr <3 x i8> <i8 -1, i8 poison, i8 -1>, [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp uge <3 x i8> [[TMP0]], [[X:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = lshr <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %tmp0 = lshr <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   %tmp1 = and <3 x i8> %tmp0, %x
   %ret = icmp eq <3 x i8> %tmp1, %x
   ret <3 x i1> %ret
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
index 37d317b695f60..8fbbd2bb9907d 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-and-icmp-ne-to-icmp-ugt.ll
@@ -40,13 +40,13 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef(
-; CHECK-NEXT:    [[TMP0:%.*]] = lshr <3 x i8> <i8 -1, i8 undef, i8 -1>, [[Y:%.*]]
+define <3 x i1> @p2_vec_poison(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison(
+; CHECK-NEXT:    [[TMP0:%.*]] = lshr <3 x i8> <i8 -1, i8 poison, i8 -1>, [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ult <3 x i8> [[TMP0]], [[X:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %tmp0 = lshr <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %tmp0 = lshr <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   %tmp1 = and <3 x i8> %tmp0, %x
   %ret = icmp ne <3 x i8> %tmp1, %x
   ret <3 x i1> %ret
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-eq-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-eq-to-icmp-ule.ll
index dfd67eae8aafd..88487b38e2c70 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-eq-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-eq-to-icmp-ule.ll
@@ -44,40 +44,40 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef0(
+define <3 x i1> @p2_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison0(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp eq <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   %t1 = xor <3 x i8> %t0, <i8 -1, i8 -1, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp eq <3 x i8> %t2, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p3_vec_undef0(
+define <3 x i1> @p3_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p3_vec_poison0(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp eq <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %t0 = shl <3 x i8> <i8 -1, i8 -1, i8 -1>, %y
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp eq <3 x i8> %t2, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p4_vec_undef2(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p4_vec_undef2(
+define <3 x i1> @p4_vec_poison2(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p4_vec_poison2(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp eq <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp eq <3 x i8> %t2, %x
   ret <3 x i1> %ret
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-ne-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-ne-to-icmp-ugt.ll
index 608e133ec7f73..b717925fd644f 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-ne-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v2-and-icmp-ne-to-icmp-ugt.ll
@@ -44,40 +44,40 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef0(
+define <3 x i1> @p2_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison0(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ne <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   %t1 = xor <3 x i8> %t0, <i8 -1, i8 -1, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp ne <3 x i8> %t2, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p3_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p3_vec_undef0(
+define <3 x i1> @p3_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p3_vec_poison0(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ne <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
   %t0 = shl <3 x i8> <i8 -1, i8 -1, i8 -1>, %y
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp ne <3 x i8> %t2, %x
   ret <3 x i1> %ret
 }
 
-define <3 x i1> @p4_vec_undef2(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p4_vec_undef2(
+define <3 x i1> @p4_vec_poison2(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p4_vec_poison2(
 ; CHECK-NEXT:    [[X_HIGHBITS:%.*]] = lshr <3 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ne <3 x i8> [[X_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %t2 = and <3 x i8> %t1, %x
   %ret = icmp ne <3 x i8> %t2, %x
   ret <3 x i1> %ret
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-eq-to-icmp-ule.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-eq-to-icmp-ule.ll
index d13129c1248a4..f48d284e085bc 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-eq-to-icmp-ule.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-eq-to-icmp-ule.ll
@@ -54,15 +54,15 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef0(
-; CHECK-NEXT:    [[T0:%.*]] = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, [[Y:%.*]]
+define <3 x i1> @p2_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison0(
+; CHECK-NEXT:    [[T0:%.*]] = shl nsw <3 x i8> <i8 -1, i8 poison, i8 -1>, [[Y:%.*]]
 ; CHECK-NEXT:    call void @use3i8(<3 x i8> [[T0]])
 ; CHECK-NEXT:    [[T1:%.*]] = lshr <3 x i8> <i8 -1, i8 -1, i8 -1>, [[Y]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp uge <3 x i8> [[T1]], [[X:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   call void @use3i8(<3 x i8> %t0)
   %t1 = lshr <3 x i8> %t0, %y
   %t2 = and <3 x i8> %t1, %x
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-ne-to-icmp-ugt.ll b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-ne-to-icmp-ugt.ll
index a1517b36d0b9d..f4b3c67164e49 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-ne-to-icmp-ugt.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-low-bit-mask-v4-and-icmp-ne-to-icmp-ugt.ll
@@ -54,15 +54,15 @@ define <2 x i1> @p1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %ret
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @p2_vec_undef0(
-; CHECK-NEXT:    [[T0:%.*]] = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, [[Y:%.*]]
+define <3 x i1> @p2_vec_poison0(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @p2_vec_poison0(
+; CHECK-NEXT:    [[T0:%.*]] = shl nsw <3 x i8> <i8 -1, i8 poison, i8 -1>, [[Y:%.*]]
 ; CHECK-NEXT:    call void @use3i8(<3 x i8> [[T0]])
 ; CHECK-NEXT:    [[T1:%.*]] = lshr <3 x i8> <i8 -1, i8 -1, i8 -1>, [[Y]]
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ult <3 x i8> [[T1]], [[X:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[RET]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %y
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %y
   call void @use3i8(<3 x i8> %t0)
   %t1 = lshr <3 x i8> %t0, %y
   %t2 = and <3 x i8> %t1, %x
diff --git a/llvm/test/Transforms/InstCombine/cast-int-icmp-eq-0.ll b/llvm/test/Transforms/InstCombine/cast-int-icmp-eq-0.ll
index 9b51a7649992f..7b6d07a14a30e 100644
--- a/llvm/test/Transforms/InstCombine/cast-int-icmp-eq-0.ll
+++ b/llvm/test/Transforms/InstCombine/cast-int-icmp-eq-0.ll
@@ -603,7 +603,7 @@ define i1 @i16_cast_cmp_sgt_int_m1_sitofp_half(i16 %i) {
   ret i1 %cmp
 }
 
-; Verify that vector types and vector constants including undef elements are transformed too.
+; Verify that vector types and vector constants including poison elements are transformed too.
 
 define <3 x i1> @i32_cast_cmp_ne_int_0_sitofp_double_vec(<3 x i32> %i) {
 ; CHECK-LABEL: @i32_cast_cmp_ne_int_0_sitofp_double_vec(
@@ -616,38 +616,38 @@ define <3 x i1> @i32_cast_cmp_ne_int_0_sitofp_double_vec(<3 x i32> %i) {
   ret <3 x i1> %cmp
 }
 
-; TODO: Can we propagate the constant vector with undef element?
+; TODO: Can we propagate the constant vector with poison element?
 
-define <3 x i1> @i32_cast_cmp_eq_int_0_sitofp_float_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_eq_int_0_sitofp_float_vec_undef(
+define <3 x i1> @i32_cast_cmp_eq_int_0_sitofp_float_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_eq_int_0_sitofp_float_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = sitofp <3 x i32> %i to  <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp eq <3 x i32> %b, <i32 0, i32 undef, i32 0>
+  %cmp = icmp eq <3 x i32> %b, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %cmp
 }
 
-define <3 x i1> @i64_cast_cmp_slt_int_1_sitofp_half_vec_undef(<3 x i64> %i) {
-; CHECK-LABEL: @i64_cast_cmp_slt_int_1_sitofp_half_vec_undef(
+define <3 x i1> @i64_cast_cmp_slt_int_1_sitofp_half_vec_poison(<3 x i64> %i) {
+; CHECK-LABEL: @i64_cast_cmp_slt_int_1_sitofp_half_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp slt <3 x i64> [[I:%.*]], <i64 1, i64 1, i64 1>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = sitofp <3 x i64> %i to  <3 x half>
   %b = bitcast <3 x half> %f to <3 x i16>
-  %cmp = icmp slt <3 x i16> %b, <i16 1, i16 undef, i16 1>
+  %cmp = icmp slt <3 x i16> %b, <i16 1, i16 poison, i16 1>
   ret <3 x i1> %cmp
 }
 
-define <3 x i1> @i16_cast_cmp_sgt_int_m1_sitofp_float_vec_undef(<3 x i16> %i) {
-; CHECK-LABEL: @i16_cast_cmp_sgt_int_m1_sitofp_float_vec_undef(
+define <3 x i1> @i16_cast_cmp_sgt_int_m1_sitofp_float_vec_poison(<3 x i16> %i) {
+; CHECK-LABEL: @i16_cast_cmp_sgt_int_m1_sitofp_float_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <3 x i16> [[I:%.*]], <i16 -1, i16 -1, i16 -1>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = sitofp <3 x i16> %i to  <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp sgt <3 x i32> %b, <i32 -1, i32 undef, i32 -1>
+  %cmp = icmp sgt <3 x i32> %b, <i32 -1, i32 poison, i32 -1>
   ret <3 x i1> %cmp
 }
 
diff --git a/llvm/test/Transforms/InstCombine/cast-unsigned-icmp-eqcmp-0.ll b/llvm/test/Transforms/InstCombine/cast-unsigned-icmp-eqcmp-0.ll
index 0752576fad45f..1565fb7c0a6a9 100644
--- a/llvm/test/Transforms/InstCombine/cast-unsigned-icmp-eqcmp-0.ll
+++ b/llvm/test/Transforms/InstCombine/cast-unsigned-icmp-eqcmp-0.ll
@@ -27,14 +27,14 @@ define <2 x i1> @i32_cast_cmp_eq_int_0_uitofp_float_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_float_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_float_vec_undef(
+define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_float_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_float_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp eq <3 x i32> %b, <i32 0, i32 undef, i32 0>
+  %cmp = icmp eq <3 x i32> %b, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %cmp
 }
 
@@ -60,14 +60,14 @@ define <2 x i1> @i32_cast_cmp_ne_int_0_uitofp_float_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_float_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_float_vec_undef(
+define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_float_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_float_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ne <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp ne <3 x i32> %b, <i32 0, i32 undef, i32 0>
+  %cmp = icmp ne <3 x i32> %b, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %cmp
 }
 
@@ -93,14 +93,14 @@ define <2 x i1> @i32_cast_cmp_eq_int_0_uitofp_double_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_double_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_double_vec_undef(
+define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_double_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_double_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x double>
   %b = bitcast <3 x double> %f to <3 x i64>
-  %cmp = icmp eq <3 x i64> %b, <i64 0, i64 undef, i64 0>
+  %cmp = icmp eq <3 x i64> %b, <i64 0, i64 poison, i64 0>
   ret <3 x i1> %cmp
 }
 
@@ -126,14 +126,14 @@ define <2 x i1> @i32_cast_cmp_ne_int_0_uitofp_double_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_double_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_double_vec_undef(
+define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_double_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_double_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ne <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x double>
   %b = bitcast <3 x double> %f to <3 x i64>
-  %cmp = icmp ne <3 x i64> %b, <i64 0, i64 undef, i64 0>
+  %cmp = icmp ne <3 x i64> %b, <i64 0, i64 poison, i64 0>
   ret <3 x i1> %cmp
 }
 
@@ -159,14 +159,14 @@ define <2 x i1> @i32_cast_cmp_eq_int_0_uitofp_half_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_half_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_half_vec_undef(
+define <3 x i1> @i32_cast_cmp_eq_int_0_uitofp_half_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_eq_int_0_uitofp_half_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x half>
   %b = bitcast <3 x half> %f to <3 x i16>
-  %cmp = icmp eq <3 x i16> %b, <i16 0, i16 undef, i16 0>
+  %cmp = icmp eq <3 x i16> %b, <i16 0, i16 poison, i16 0>
   ret <3 x i1> %cmp
 }
 
@@ -192,13 +192,13 @@ define <2 x i1> @i32_cast_cmp_ne_int_0_uitofp_half_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_half_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_half_vec_undef(
+define <3 x i1> @i32_cast_cmp_ne_int_0_uitofp_half_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_ne_int_0_uitofp_half_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ne <3 x i32> [[I:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
   %f = uitofp <3 x i32> %i to <3 x half>
   %b = bitcast <3 x half> %f to <3 x i16>
-  %cmp = icmp ne <3 x i16> %b, <i16 0, i16 undef, i16 0>
+  %cmp = icmp ne <3 x i16> %b, <i16 0, i16 poison, i16 0>
   ret <3 x i1> %cmp
 }
diff --git a/llvm/test/Transforms/InstCombine/cast.ll b/llvm/test/Transforms/InstCombine/cast.ll
index d9c93ba277295..04a3e8931e62c 100644
--- a/llvm/test/Transforms/InstCombine/cast.ll
+++ b/llvm/test/Transforms/InstCombine/cast.ll
@@ -508,18 +508,16 @@ define <2 x i16> @test40vec_nonuniform(<2 x i16> %a) {
   ret <2 x i16> %r
 }
 
-define <2 x i16> @test40vec_undef(<2 x i16> %a) {
-; ALL-LABEL: @test40vec_undef(
-; ALL-NEXT:    [[T:%.*]] = zext <2 x i16> [[A:%.*]] to <2 x i32>
-; ALL-NEXT:    [[T21:%.*]] = lshr <2 x i32> [[T]], <i32 9, i32 undef>
-; ALL-NEXT:    [[T5:%.*]] = shl <2 x i32> [[T]], <i32 8, i32 undef>
-; ALL-NEXT:    [[T32:%.*]] = or <2 x i32> [[T21]], [[T5]]
-; ALL-NEXT:    [[R:%.*]] = trunc <2 x i32> [[T32]] to <2 x i16>
+define <2 x i16> @test40vec_poison(<2 x i16> %a) {
+; ALL-LABEL: @test40vec_poison(
+; ALL-NEXT:    [[T21:%.*]] = lshr <2 x i16> [[A:%.*]], <i16 9, i16 poison>
+; ALL-NEXT:    [[T5:%.*]] = shl <2 x i16> [[A]], <i16 8, i16 poison>
+; ALL-NEXT:    [[R:%.*]] = or disjoint <2 x i16> [[T21]], [[T5]]
 ; ALL-NEXT:    ret <2 x i16> [[R]]
 ;
   %t = zext <2 x i16> %a to <2 x i32>
-  %t21 = lshr <2 x i32> %t, <i32 9, i32 undef>
-  %t5 = shl <2 x i32> %t, <i32 8, i32 undef>
+  %t21 = lshr <2 x i32> %t, <i32 9, i32 poison>
+  %t5 = shl <2 x i32> %t, <i32 8, i32 poison>
   %t32 = or <2 x i32> %t21, %t5
   %r = trunc <2 x i32> %t32 to <2 x i16>
   ret <2 x i16> %r
@@ -1452,7 +1450,7 @@ define i32 @test89() {
 ; LE-LABEL: @test89(
 ; LE-NEXT:    ret i32 6
 ;
-  ret i32 bitcast (<2 x i16> <i16 6, i16 undef> to i32)
+  ret i32 bitcast (<2 x i16> <i16 6, i16 poison> to i32)
 }
 
 define <2 x i32> @test90() {
@@ -1462,7 +1460,7 @@ define <2 x i32> @test90() {
 ; LE-LABEL: @test90(
 ; LE-NEXT:    ret <2 x i32> <i32 0, i32 1006632960>
 ;
-  %t6 = bitcast <4 x half> <half undef, half undef, half undef, half 0xH3C00> to <2 x i32>
+  %t6 = bitcast <4 x half> <half poison, half poison, half poison, half 0xH3C00> to <2 x i32>
   ret <2 x i32> %t6
 }
 
@@ -1537,13 +1535,13 @@ define <2 x i8> @trunc_lshr_sext_uniform(<2 x i8> %A) {
   ret <2 x i8> %D
 }
 
-define <2 x i8> @trunc_lshr_sext_uniform_undef(<2 x i8> %A) {
-; ALL-LABEL: @trunc_lshr_sext_uniform_undef(
-; ALL-NEXT:    [[D:%.*]] = ashr <2 x i8> [[A:%.*]], <i8 6, i8 undef>
+define <2 x i8> @trunc_lshr_sext_uniform_poison(<2 x i8> %A) {
+; ALL-LABEL: @trunc_lshr_sext_uniform_poison(
+; ALL-NEXT:    [[D:%.*]] = ashr <2 x i8> [[A:%.*]], <i8 6, i8 poison>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = sext <2 x i8> %A to <2 x i32>
-  %C = lshr <2 x i32> %B, <i32 6, i32 undef>
+  %C = lshr <2 x i32> %B, <i32 6, i32 poison>
   %D = trunc <2 x i32> %C to <2 x i8>
   ret <2 x i8> %D
 }
@@ -1559,13 +1557,13 @@ define <2 x i8> @trunc_lshr_sext_nonuniform(<2 x i8> %A) {
   ret <2 x i8> %D
 }
 
-define <3 x i8> @trunc_lshr_sext_nonuniform_undef(<3 x i8> %A) {
-; ALL-LABEL: @trunc_lshr_sext_nonuniform_undef(
-; ALL-NEXT:    [[D:%.*]] = ashr <3 x i8> [[A:%.*]], <i8 6, i8 2, i8 undef>
+define <3 x i8> @trunc_lshr_sext_nonuniform_poison(<3 x i8> %A) {
+; ALL-LABEL: @trunc_lshr_sext_nonuniform_poison(
+; ALL-NEXT:    [[D:%.*]] = ashr <3 x i8> [[A:%.*]], <i8 6, i8 2, i8 poison>
 ; ALL-NEXT:    ret <3 x i8> [[D]]
 ;
   %B = sext <3 x i8> %A to <3 x i32>
-  %C = lshr <3 x i32> %B, <i32 6, i32 2, i32 undef>
+  %C = lshr <3 x i32> %B, <i32 6, i32 2, i32 poison>
   %D = trunc <3 x i32> %C to <3 x i8>
   ret <3 x i8> %D
 }
@@ -2014,15 +2012,13 @@ define <2 x i8> @trunc_lshr_zext_uniform(<2 x i8> %A) {
   ret <2 x i8> %D
 }
 
-define <2 x i8> @trunc_lshr_zext_uniform_undef(<2 x i8> %A) {
-; ALL-LABEL: @trunc_lshr_zext_uniform_undef(
-; ALL-NEXT:    [[B:%.*]] = zext <2 x i8> [[A:%.*]] to <2 x i32>
-; ALL-NEXT:    [[C:%.*]] = lshr <2 x i32> [[B]], <i32 6, i32 undef>
-; ALL-NEXT:    [[D:%.*]] = trunc nuw <2 x i32> [[C]] to <2 x i8>
+define <2 x i8> @trunc_lshr_zext_uniform_poison(<2 x i8> %A) {
+; ALL-LABEL: @trunc_lshr_zext_uniform_poison(
+; ALL-NEXT:    [[D:%.*]] = lshr <2 x i8> [[A:%.*]], <i8 6, i8 poison>
 ; ALL-NEXT:    ret <2 x i8> [[D]]
 ;
   %B = zext <2 x i8> %A to <2 x i32>
-  %C = lshr <2 x i32> %B, <i32 6, i32 undef>
+  %C = lshr <2 x i32> %B, <i32 6, i32 poison>
   %D = trunc <2 x i32> %C to <2 x i8>
   ret <2 x i8> %D
 }
@@ -2038,15 +2034,13 @@ define <2 x i8> @trunc_lshr_zext_nonuniform(<2 x i8> %A) {
   ret <2 x i8> %D
 }
 
-define <3 x i8> @trunc_lshr_zext_nonuniform_undef(<3 x i8> %A) {
-; ALL-LABEL: @trunc_lshr_zext_nonuniform_undef(
-; ALL-NEXT:    [[B:%.*]] = zext <3 x i8> [[A:%.*]] to <3 x i32>
-; ALL-NEXT:    [[C:%.*]] = lshr <3 x i32> [[B]], <i32 6, i32 2, i32 undef>
-; ALL-NEXT:    [[D:%.*]] = trunc nuw <3 x i32> [[C]] to <3 x i8>
+define <3 x i8> @trunc_lshr_zext_nonuniform_poison(<3 x i8> %A) {
+; ALL-LABEL: @trunc_lshr_zext_nonuniform_poison(
+; ALL-NEXT:    [[D:%.*]] = lshr <3 x i8> [[A:%.*]], <i8 6, i8 2, i8 poison>
 ; ALL-NEXT:    ret <3 x i8> [[D]]
 ;
   %B = zext <3 x i8> %A to <3 x i32>
-  %C = lshr <3 x i32> %B, <i32 6, i32 2, i32 undef>
+  %C = lshr <3 x i32> %B, <i32 6, i32 2, i32 poison>
   %D = trunc <3 x i32> %C to <3 x i8>
   ret <3 x i8> %D
 }
diff --git a/llvm/test/Transforms/InstCombine/ctpop-cttz.ll b/llvm/test/Transforms/InstCombine/ctpop-cttz.ll
index 5d27f374d8921..70868554bdc1b 100644
--- a/llvm/test/Transforms/InstCombine/ctpop-cttz.ll
+++ b/llvm/test/Transforms/InstCombine/ctpop-cttz.ll
@@ -116,14 +116,14 @@ define <2 x i32> @ctpop3v(<2 x i32> %0) {
   ret <2 x i32> %5
 }
 
-define <2 x i32> @ctpop3v_undef(<2 x i32> %0) {
-; CHECK-LABEL: @ctpop3v_undef(
+define <2 x i32> @ctpop3v_poison(<2 x i32> %0) {
+; CHECK-LABEL: @ctpop3v_poison(
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> [[TMP0:%.*]], i1 false), !range [[RNG0]]
 ; CHECK-NEXT:    ret <2 x i32> [[TMP2]]
 ;
   %2 = sub <2 x i32> zeroinitializer, %0
   %3 = and <2 x i32> %2, %0
-  %4 = add <2 x i32> %3, <i32 -1, i32 undef>
+  %4 = add <2 x i32> %3, <i32 -1, i32 poison>
   %5 = tail call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %4)
   ret <2 x i32> %5
 }
diff --git a/llvm/test/Transforms/InstCombine/ctpop.ll b/llvm/test/Transforms/InstCombine/ctpop.ll
index 27194724b7d83..b3653e5071ba2 100644
--- a/llvm/test/Transforms/InstCombine/ctpop.ll
+++ b/llvm/test/Transforms/InstCombine/ctpop.ll
@@ -155,28 +155,27 @@ define <2 x i32> @_parity_of_not_vec(<2 x i32> %x) {
   ret <2 x i32> %r
 }
 
-define <2 x i32> @_parity_of_not_undef(<2 x i32> %x) {
-; CHECK-LABEL: @_parity_of_not_undef(
+define <2 x i32> @_parity_of_not_poison(<2 x i32> %x) {
+; CHECK-LABEL: @_parity_of_not_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> [[X:%.*]]), !range [[RNG1]]
 ; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[TMP1]], <i32 1, i32 1>
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %neg = xor <2 x i32> %x, <i32 undef ,i32 -1>
+  %neg = xor <2 x i32> %x, <i32 poison ,i32 -1>
   %cnt = tail call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %neg)
   %r = and <2 x i32> %cnt, <i32 1 ,i32 1>
   ret <2 x i32> %r
 }
 
-define <2 x i32> @_parity_of_not_undef2(<2 x i32> %x) {
-; CHECK-LABEL: @_parity_of_not_undef2(
-; CHECK-NEXT:    [[NEG:%.*]] = xor <2 x i32> [[X:%.*]], <i32 -1, i32 -1>
-; CHECK-NEXT:    [[CNT:%.*]] = tail call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> [[NEG]]), !range [[RNG1]]
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[CNT]], <i32 1, i32 undef>
+define <2 x i32> @_parity_of_not_poison2(<2 x i32> %x) {
+; CHECK-LABEL: @_parity_of_not_poison2(
+; CHECK-NEXT:    [[CNT:%.*]] = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> [[X:%.*]]), !range [[RNG1]]
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[CNT]], <i32 1, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %neg = xor <2 x i32> %x, <i32 -1 ,i32 -1>
   %cnt = tail call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %neg)
-  %r = and <2 x i32> %cnt, <i32 1 ,i32 undef>
+  %r = and <2 x i32> %cnt, <i32 1 ,i32 poison>
   ret <2 x i32> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fabs-as-int.ll b/llvm/test/Transforms/InstCombine/fabs-as-int.ll
index f32c00e453f22..4e49ff159f875 100644
--- a/llvm/test/Transforms/InstCombine/fabs-as-int.ll
+++ b/llvm/test/Transforms/InstCombine/fabs-as-int.ll
@@ -137,15 +137,15 @@ define <2 x i32> @not_fabs_as_int_v2f32_nonsplat(<2 x float> %x) {
   ret <2 x i32> %and
 }
 
-define <3 x i32> @fabs_as_int_v3f32_undef(<3 x float> %x) {
-; CHECK-LABEL: define <3 x i32> @fabs_as_int_v3f32_undef
+define <3 x i32> @fabs_as_int_v3f32_poison(<3 x float> %x) {
+; CHECK-LABEL: define <3 x i32> @fabs_as_int_v3f32_poison
 ; CHECK-SAME: (<3 x float> [[X:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <3 x float> @llvm.fabs.v3f32(<3 x float> [[X]])
 ; CHECK-NEXT:    [[AND:%.*]] = bitcast <3 x float> [[TMP1]] to <3 x i32>
 ; CHECK-NEXT:    ret <3 x i32> [[AND]]
 ;
   %bc = bitcast <3 x float> %x to <3 x i32>
-  %and = and <3 x i32> %bc, <i32 2147483647, i32 undef, i32 2147483647>
+  %and = and <3 x i32> %bc, <i32 2147483647, i32 poison, i32 2147483647>
   ret <3 x i32> %and
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fabs.ll b/llvm/test/Transforms/InstCombine/fabs.ll
index 7e380c2e4590a..5ec65784e7a34 100644
--- a/llvm/test/Transforms/InstCombine/fabs.ll
+++ b/llvm/test/Transforms/InstCombine/fabs.ll
@@ -321,7 +321,7 @@ define <2 x float> @select_fcmp_nnan_ole_negzero(<2 x float> %x) {
 ; CHECK-NEXT:    ret <2 x float> [[FABS]]
 ;
   %lezero = fcmp ole <2 x float> %x, <float -0.0, float -0.0>
-  %negx = fsub nnan <2 x float> <float 0.0, float undef>, %x
+  %negx = fsub nnan <2 x float> <float 0.0, float poison>, %x
   %fabs = select <2 x i1> %lezero, <2 x float> %negx, <2 x float> %x
   ret <2 x float> %fabs
 }
@@ -332,7 +332,7 @@ define <2 x float> @select_nnan_fcmp_nnan_ole_negzero(<2 x float> %x) {
 ; CHECK-NEXT:    ret <2 x float> [[FABS]]
 ;
   %lezero = fcmp ole <2 x float> %x, <float -0.0, float -0.0>
-  %negx = fsub nnan <2 x float> <float 0.0, float undef>, %x
+  %negx = fsub nnan <2 x float> <float 0.0, float poison>, %x
   %fabs = select nnan <2 x i1> %lezero, <2 x float> %negx, <2 x float> %x
   ret <2 x float> %fabs
 }
diff --git a/llvm/test/Transforms/InstCombine/fast-math.ll b/llvm/test/Transforms/InstCombine/fast-math.ll
index 916955e34efac..83f2091244e52 100644
--- a/llvm/test/Transforms/InstCombine/fast-math.ll
+++ b/llvm/test/Transforms/InstCombine/fast-math.ll
@@ -541,12 +541,12 @@ define float @fneg2(float %x) {
   ret float %sub
 }
 
-define <2 x float> @fneg2_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fneg2_vec_undef(
+define <2 x float> @fneg2_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fneg2_vec_poison(
 ; CHECK-NEXT:    [[SUB:%.*]] = fneg nsz <2 x float> [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[SUB]]
 ;
-  %sub = fsub nsz <2 x float> <float undef, float 0.0>, %x
+  %sub = fsub nsz <2 x float> <float poison, float 0.0>, %x
   ret <2 x float> %sub
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fcmp-special.ll b/llvm/test/Transforms/InstCombine/fcmp-special.ll
index 88bfe930ffdd6..64bc86f4266c7 100644
--- a/llvm/test/Transforms/InstCombine/fcmp-special.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp-special.ll
@@ -144,21 +144,21 @@ define <2 x i1> @uno_vec_with_nan(<2 x double> %x) {
   ret <2 x i1> %f
 }
 
-define <2 x i1> @uno_vec_with_undef(<2 x double> %x) {
-; CHECK-LABEL: @uno_vec_with_undef(
+define <2 x i1> @uno_vec_with_poison(<2 x double> %x) {
+; CHECK-LABEL: @uno_vec_with_poison(
 ; CHECK-NEXT:    [[F:%.*]] = fcmp uno <2 x double> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[F]]
 ;
-  %f = fcmp uno <2 x double> %x, <double 3.0, double undef>
+  %f = fcmp uno <2 x double> %x, <double 3.0, double poison>
   ret <2 x i1> %f
 }
 
-define <2 x i1> @ord_vec_with_undef(<2 x double> %x) {
-; CHECK-LABEL: @ord_vec_with_undef(
-; CHECK-NEXT:    [[F:%.*]] = fcmp ord <2 x double> [[X:%.*]], <double 0.000000e+00, double undef>
+define <2 x i1> @ord_vec_with_poison(<2 x double> %x) {
+; CHECK-LABEL: @ord_vec_with_poison(
+; CHECK-NEXT:    [[F:%.*]] = fcmp ord <2 x double> [[X:%.*]], <double 0.000000e+00, double poison>
 ; CHECK-NEXT:    ret <2 x i1> [[F]]
 ;
-  %f = fcmp ord <2 x double> %x, <double 0.0, double undef>
+  %f = fcmp ord <2 x double> %x, <double 0.0, double poison>
   ret <2 x i1> %f
 }
 
@@ -224,12 +224,12 @@ define <2 x i1> @negative_zero_olt_vec(<2 x float> %x) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @negative_zero_une_vec_undef(<2 x double> %x) {
-; CHECK-LABEL: @negative_zero_une_vec_undef(
+define <2 x i1> @negative_zero_une_vec_poison(<2 x double> %x) {
+; CHECK-LABEL: @negative_zero_une_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = fcmp nnan une <2 x double> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %r = fcmp nnan une <2 x double> %x, <double -0.0, double undef>
+  %r = fcmp nnan une <2 x double> %x, <double -0.0, double poison>
   ret <2 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fcmp.ll b/llvm/test/Transforms/InstCombine/fcmp.ll
index 069512b0f2d8e..389264e2f7075 100644
--- a/llvm/test/Transforms/InstCombine/fcmp.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp.ll
@@ -102,12 +102,12 @@ define <2 x i1> @unary_fneg_constant_swap_pred_vec(<2 x float> %x) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @fneg_constant_swap_pred_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fneg_constant_swap_pred_vec_undef(
+define <2 x i1> @fneg_constant_swap_pred_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fneg_constant_swap_pred_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = fcmp olt <2 x float> [[X:%.*]], <float -1.000000e+00, float -2.000000e+00>
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %neg = fsub <2 x float> <float undef, float -0.0>, %x
+  %neg = fsub <2 x float> <float poison, float -0.0>, %x
   %cmp = fcmp ogt <2 x float> %neg, <float 1.0, float 2.0>
   ret <2 x i1> %cmp
 }
@@ -234,34 +234,34 @@ define <2 x i1> @fneg_unary_fneg_swap_pred_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @fneg_fneg_swap_pred_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @fneg_fneg_swap_pred_vec_undef(
+define <2 x i1> @fneg_fneg_swap_pred_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @fneg_fneg_swap_pred_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = fcmp ogt <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %neg1 = fsub <2 x float> <float -0.0, float undef>, %x
-  %neg2 = fsub <2 x float> <float undef, float -0.0>, %y
+  %neg1 = fsub <2 x float> <float -0.0, float poison>, %x
+  %neg2 = fsub <2 x float> <float poison, float -0.0>, %y
   %cmp = fcmp olt <2 x float> %neg1, %neg2
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @unary_fneg_fneg_swap_pred_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @unary_fneg_fneg_swap_pred_vec_undef(
+define <2 x i1> @unary_fneg_fneg_swap_pred_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @unary_fneg_fneg_swap_pred_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = fcmp ogt <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
   %neg1 = fneg <2 x float> %x
-  %neg2 = fsub <2 x float> <float undef, float -0.0>, %y
+  %neg2 = fsub <2 x float> <float poison, float -0.0>, %y
   %cmp = fcmp olt <2 x float> %neg1, %neg2
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @fneg_unary_fneg_swap_pred_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @fneg_unary_fneg_swap_pred_vec_undef(
+define <2 x i1> @fneg_unary_fneg_swap_pred_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @fneg_unary_fneg_swap_pred_vec_poison(
 ; CHECK-NEXT:    [[CMP:%.*]] = fcmp ogt <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %neg1 = fsub <2 x float> <float -0.0, float undef>, %x
+  %neg1 = fsub <2 x float> <float -0.0, float poison>, %x
   %neg2 = fneg <2 x float> %y
   %cmp = fcmp olt <2 x float> %neg1, %neg2
   ret <2 x i1> %cmp
diff --git a/llvm/test/Transforms/InstCombine/fdiv.ll b/llvm/test/Transforms/InstCombine/fdiv.ll
index a0710c2bb0484..ca11685c98417 100644
--- a/llvm/test/Transforms/InstCombine/fdiv.ll
+++ b/llvm/test/Transforms/InstCombine/fdiv.ll
@@ -141,12 +141,12 @@ define <2 x float> @not_exact_inverse_vec_arcp(<2 x float> %x) {
   ret <2 x float> %div
 }
 
-define <2 x float> @not_exact_inverse_vec_arcp_with_undef_elt(<2 x float> %x) {
-; CHECK-LABEL: @not_exact_inverse_vec_arcp_with_undef_elt(
-; CHECK-NEXT:    [[DIV:%.*]] = fdiv arcp <2 x float> [[X:%.*]], <float undef, float 3.000000e+00>
+define <2 x float> @not_exact_inverse_vec_arcp_with_poison_elt(<2 x float> %x) {
+; CHECK-LABEL: @not_exact_inverse_vec_arcp_with_poison_elt(
+; CHECK-NEXT:    [[DIV:%.*]] = fdiv arcp <2 x float> [[X:%.*]], <float poison, float 3.000000e+00>
 ; CHECK-NEXT:    ret <2 x float> [[DIV]]
 ;
-  %div = fdiv arcp <2 x float> %x, <float undef, float 3.0>
+  %div = fdiv arcp <2 x float> %x, <float poison, float 3.0>
   ret <2 x float> %div
 }
 
@@ -333,13 +333,13 @@ define <2 x float> @unary_fneg_fneg_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x float> %div
 }
 
-define <2 x float> @fneg_fneg_vec_undef_elts(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @fneg_fneg_vec_undef_elts(
+define <2 x float> @fneg_fneg_vec_poison_elts(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @fneg_fneg_vec_poison_elts(
 ; CHECK-NEXT:    [[DIV:%.*]] = fdiv <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[DIV]]
 ;
-  %xneg = fsub <2 x float> <float undef, float -0.0>, %x
-  %yneg = fsub <2 x float> <float -0.0, float undef>, %y
+  %xneg = fsub <2 x float> <float poison, float -0.0>, %x
+  %yneg = fsub <2 x float> <float -0.0, float poison>, %y
   %div = fdiv <2 x float> %xneg, %yneg
   ret <2 x float> %div
 }
@@ -404,12 +404,12 @@ define <2 x float> @unary_fneg_dividend_constant_divisor_vec(<2 x float> %x) {
   ret <2 x float> %div
 }
 
-define <2 x float> @fneg_dividend_constant_divisor_vec_undef_elt(<2 x float> %x) {
-; CHECK-LABEL: @fneg_dividend_constant_divisor_vec_undef_elt(
+define <2 x float> @fneg_dividend_constant_divisor_vec_poison_elt(<2 x float> %x) {
+; CHECK-LABEL: @fneg_dividend_constant_divisor_vec_poison_elt(
 ; CHECK-NEXT:    [[DIV:%.*]] = fdiv ninf <2 x float> [[X:%.*]], <float -3.000000e+00, float 8.000000e+00>
 ; CHECK-NEXT:    ret <2 x float> [[DIV]]
 ;
-  %neg = fsub <2 x float> <float undef, float -0.0>, %x
+  %neg = fsub <2 x float> <float poison, float -0.0>, %x
   %div = fdiv ninf <2 x float> %neg, <float 3.0, float -8.0>
   ret <2 x float> %div
 }
diff --git a/llvm/test/Transforms/InstCombine/fma.ll b/llvm/test/Transforms/InstCombine/fma.ll
index 8b413ae6f664b..cf3d7f3c525a5 100644
--- a/llvm/test/Transforms/InstCombine/fma.ll
+++ b/llvm/test/Transforms/InstCombine/fma.ll
@@ -60,13 +60,13 @@ define <2 x float> @fma_unary_fneg_x_unary_fneg_y_vec(<2 x float> %x, <2 x float
   ret <2 x float> %fma
 }
 
-define <2 x float> @fma_fneg_x_fneg_y_vec_undef(<2 x float> %x, <2 x float> %y, <2 x float> %z) {
-; CHECK-LABEL: @fma_fneg_x_fneg_y_vec_undef(
+define <2 x float> @fma_fneg_x_fneg_y_vec_poison(<2 x float> %x, <2 x float> %y, <2 x float> %z) {
+; CHECK-LABEL: @fma_fneg_x_fneg_y_vec_poison(
 ; CHECK-NEXT:    [[FMA:%.*]] = call <2 x float> @llvm.fma.v2f32(<2 x float> [[X:%.*]], <2 x float> [[Y:%.*]], <2 x float> [[Z:%.*]])
 ; CHECK-NEXT:    ret <2 x float> [[FMA]]
 ;
-  %xn = fsub <2 x float> <float -0.0, float undef>, %x
-  %yn = fsub <2 x float> <float undef, float -0.0>, %y
+  %xn = fsub <2 x float> <float -0.0, float poison>, %x
+  %yn = fsub <2 x float> <float poison, float -0.0>, %y
   %fma = call <2 x float> @llvm.fma.v2f32(<2 x float> %xn, <2 x float> %yn, <2 x float> %z)
   ret <2 x float> %fma
 }
diff --git a/llvm/test/Transforms/InstCombine/fmul.ll b/llvm/test/Transforms/InstCombine/fmul.ll
index 39f9e74f899d1..e9c86a1270493 100644
--- a/llvm/test/Transforms/InstCombine/fmul.ll
+++ b/llvm/test/Transforms/InstCombine/fmul.ll
@@ -42,12 +42,12 @@ define <2 x float> @unary_neg_constant_vec(<2 x float> %x) {
   ret <2 x float> %mul
 }
 
-define <2 x float> @neg_constant_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @neg_constant_vec_undef(
+define <2 x float> @neg_constant_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @neg_constant_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul ninf <2 x float> [[X:%.*]], <float -2.000000e+00, float -3.000000e+00>
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
-  %sub = fsub <2 x float> <float undef, float -0.0>, %x
+  %sub = fsub <2 x float> <float poison, float -0.0>, %x
   %mul = fmul ninf <2 x float> %sub, <float 2.0, float 3.0>
   ret <2 x float> %mul
 }
@@ -162,34 +162,34 @@ define <2 x float> @neg_unary_neg_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x float> %mul
 }
 
-define <2 x float> @neg_neg_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @neg_neg_vec_undef(
+define <2 x float> @neg_neg_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @neg_neg_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
-  %sub1 = fsub <2 x float> <float -0.0, float undef>, %x
-  %sub2 = fsub <2 x float> <float undef, float -0.0>, %y
+  %sub1 = fsub <2 x float> <float -0.0, float poison>, %x
+  %sub2 = fsub <2 x float> <float poison, float -0.0>, %y
   %mul = fmul arcp <2 x float> %sub1, %sub2
   ret <2 x float> %mul
 }
 
-define <2 x float> @unary_neg_neg_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @unary_neg_neg_vec_undef(
+define <2 x float> @unary_neg_neg_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @unary_neg_neg_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
   %neg = fneg <2 x float> %x
-  %sub = fsub <2 x float> <float undef, float -0.0>, %y
+  %sub = fsub <2 x float> <float poison, float -0.0>, %y
   %mul = fmul arcp <2 x float> %neg, %sub
   ret <2 x float> %mul
 }
 
-define <2 x float> @neg_unary_neg_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @neg_unary_neg_vec_undef(
+define <2 x float> @neg_unary_neg_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @neg_unary_neg_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
-  %sub = fsub <2 x float> <float -0.0, float undef>, %x
+  %sub = fsub <2 x float> <float -0.0, float poison>, %x
   %neg = fneg <2 x float> %y
   %mul = fmul arcp <2 x float> %sub, %neg
   ret <2 x float> %mul
@@ -322,13 +322,13 @@ define <2 x float> @unary_neg_mul_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x float> %mul
 }
 
-define <2 x float> @neg_mul_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @neg_mul_vec_undef(
+define <2 x float> @neg_mul_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @neg_mul_vec_poison(
 ; CHECK-NEXT:    [[SUB:%.*]] = fneg <2 x float> [[X:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = fmul <2 x float> [[SUB]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[MUL]]
 ;
-  %sub = fsub <2 x float> <float undef, float -0.0>, %x
+  %sub = fsub <2 x float> <float poison, float -0.0>, %x
   %mul = fmul <2 x float> %sub, %y
   ret <2 x float> %mul
 }
@@ -388,9 +388,9 @@ define void @test8(ptr %inout, i1 %c1) {
 entry:
   %0 = load i32, ptr %inout, align 4
   %conv = uitofp i32 %0 to float
-  %vecinit = insertelement <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>, float %conv, i32 3
+  %vecinit = insertelement <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float poison>, float %conv, i32 3
   %sub = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %vecinit
-  %1 = shufflevector <4 x float> %sub, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
+  %1 = shufflevector <4 x float> %sub, <4 x float> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
   %mul = fmul <4 x float> zeroinitializer, %1
   br label %for.cond
 
@@ -742,7 +742,7 @@ define <4 x float> @fdiv_constant_denominator_fmul_vec_constexpr(<4 x float> %x)
 ; CHECK-NEXT:    [[T3:%.*]] = fmul reassoc <4 x float> [[X:%.*]], <float 3.000000e+00, float 2.000000e+00, float 1.000000e+00, float 1.000000e+00>
 ; CHECK-NEXT:    ret <4 x float> [[T3]]
 ;
-  %constExprMul = bitcast i128 trunc (i160 bitcast (<5 x float> <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3, float undef> to i160) to i128) to <4 x float>
+  %constExprMul = bitcast i128 trunc (i160 bitcast (<5 x float> <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3, float poison> to i160) to i128) to <4 x float>
   %t1 = fdiv reassoc <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3>
   %t3 = fmul reassoc <4 x float> %t1, %constExprMul
   ret <4 x float> %t3
@@ -1270,7 +1270,7 @@ define <vscale x 2 x float> @mul_scalable_splat_zero(<vscale x 2 x float> %z) {
 ; CHECK-LABEL: @mul_scalable_splat_zero(
 ; CHECK-NEXT:    ret <vscale x 2 x float> zeroinitializer
 ;
-  %shuf = shufflevector <vscale x 2 x float> insertelement (<vscale x 2 x float> undef, float 0.0, i32 0), <vscale x 2 x float> undef, <vscale x 2 x i32> zeroinitializer
+  %shuf = shufflevector <vscale x 2 x float> insertelement (<vscale x 2 x float> poison, float 0.0, i32 0), <vscale x 2 x float> poison, <vscale x 2 x i32> zeroinitializer
   %t3 = fmul fast <vscale x 2 x float> %shuf, %z
   ret <vscale x 2 x float> %t3
 }
@@ -1393,7 +1393,7 @@ define <3 x float> @mul_neg_zero_nnan_ninf_vec(<3 x float> nofpclass(inf nan) %a
 ; CHECK-NEXT:    ret <3 x float> [[RET]]
 ;
 entry:
-  %ret = fmul <3 x float> %a, <float -0.0, float undef, float poison>
+  %ret = fmul <3 x float> %a, <float -0.0, float poison, float poison>
   ret <3 x float> %ret
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fneg-as-int.ll b/llvm/test/Transforms/InstCombine/fneg-as-int.ll
index d28e599cacf36..e3067b0d02461 100644
--- a/llvm/test/Transforms/InstCombine/fneg-as-int.ll
+++ b/llvm/test/Transforms/InstCombine/fneg-as-int.ll
@@ -139,15 +139,15 @@ define <2 x i32> @not_fneg_as_int_v2f32_nonsplat(<2 x float> %x) {
   ret <2 x i32> %xor
 }
 
-define <3 x i32> @fneg_as_int_v3f32_undef(<3 x float> %x) {
-; CHECK-LABEL: define <3 x i32> @fneg_as_int_v3f32_undef
+define <3 x i32> @fneg_as_int_v3f32_poison(<3 x float> %x) {
+; CHECK-LABEL: define <3 x i32> @fneg_as_int_v3f32_poison
 ; CHECK-SAME: (<3 x float> [[X:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = fneg <3 x float> [[X]]
 ; CHECK-NEXT:    [[XOR:%.*]] = bitcast <3 x float> [[TMP1]] to <3 x i32>
 ; CHECK-NEXT:    ret <3 x i32> [[XOR]]
 ;
   %bc = bitcast <3 x float> %x to <3 x i32>
-  %xor = xor <3 x i32> %bc, <i32 -2147483648, i32 undef, i32 -2147483648>
+  %xor = xor <3 x i32> %bc, <i32 -2147483648, i32 poison, i32 -2147483648>
   ret <3 x i32> %xor
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fneg-fabs-as-int.ll b/llvm/test/Transforms/InstCombine/fneg-fabs-as-int.ll
index 9aa8d4944e39a..8c3e6958fe083 100644
--- a/llvm/test/Transforms/InstCombine/fneg-fabs-as-int.ll
+++ b/llvm/test/Transforms/InstCombine/fneg-fabs-as-int.ll
@@ -158,8 +158,8 @@ define <2 x i32> @not_fneg_fabs_as_int_v2f32_nonsplat(<2 x float> %x) {
   ret <2 x i32> %or
 }
 
-define <3 x i32> @fneg_fabs_as_int_v3f32_undef(<3 x float> %x) {
-; CHECK-LABEL: define <3 x i32> @fneg_fabs_as_int_v3f32_undef
+define <3 x i32> @fneg_fabs_as_int_v3f32_poison(<3 x float> %x) {
+; CHECK-LABEL: define <3 x i32> @fneg_fabs_as_int_v3f32_poison
 ; CHECK-SAME: (<3 x float> [[X:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <3 x float> @llvm.fabs.v3f32(<3 x float> [[X]])
 ; CHECK-NEXT:    [[TMP2:%.*]] = fneg <3 x float> [[TMP1]]
@@ -167,7 +167,7 @@ define <3 x i32> @fneg_fabs_as_int_v3f32_undef(<3 x float> %x) {
 ; CHECK-NEXT:    ret <3 x i32> [[OR]]
 ;
   %bc = bitcast <3 x float> %x to <3 x i32>
-  %or = or <3 x i32> %bc, <i32 -2147483648, i32 undef, i32 -2147483648>
+  %or = or <3 x i32> %bc, <i32 -2147483648, i32 poison, i32 -2147483648>
   ret <3 x i32> %or
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fneg.ll b/llvm/test/Transforms/InstCombine/fneg.ll
index ed68ba50d36ee..7c9289c447113 100644
--- a/llvm/test/Transforms/InstCombine/fneg.ll
+++ b/llvm/test/Transforms/InstCombine/fneg.ll
@@ -87,24 +87,24 @@ define float @fmul_fneg_extra_use(float %x) {
   ret float %r
 }
 
-; Try a vector. Use special constants (NaN, INF, undef) because they don't change anything.
+; Try a vector. Use special constants (NaN, INF, poison) because they don't change anything.
 
 define <4 x double> @fmul_fsub_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fmul_fsub_vec(
-; CHECK-NEXT:    [[R:%.*]] = fmul <4 x double> [[X:%.*]], <double -4.200000e+01, double 0xFFF8000000000000, double 0xFFF0000000000000, double undef>
+; CHECK-NEXT:    [[R:%.*]] = fmul <4 x double> [[X:%.*]], <double -4.200000e+01, double 0xFFF8000000000000, double 0xFFF0000000000000, double poison>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %m = fmul <4 x double> %x, <double 42.0, double 0x7FF8000000000000, double 0x7FF0000000000000, double undef>
+  %m = fmul <4 x double> %x, <double 42.0, double 0x7FF8000000000000, double 0x7FF0000000000000, double poison>
   %r = fsub <4 x double> <double -0.0, double -0.0, double -0.0, double -0.0>, %m
   ret <4 x double> %r
 }
 
 define <4 x double> @fmul_fneg_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fmul_fneg_vec(
-; CHECK-NEXT:    [[R:%.*]] = fmul <4 x double> [[X:%.*]], <double -4.200000e+01, double 0xFFF8000000000000, double 0xFFF0000000000000, double undef>
+; CHECK-NEXT:    [[R:%.*]] = fmul <4 x double> [[X:%.*]], <double -4.200000e+01, double 0xFFF8000000000000, double 0xFFF0000000000000, double poison>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %m = fmul <4 x double> %x, <double 42.0, double 0x7FF8000000000000, double 0x7FF0000000000000, double undef>
+  %m = fmul <4 x double> %x, <double 42.0, double 0x7FF8000000000000, double 0x7FF0000000000000, double poison>
   %r = fneg <4 x double> %m
   ret <4 x double> %r
 }
@@ -181,24 +181,24 @@ define float @fdiv_op1_constant_fneg_extra_use(float %x) {
   ret float %r
 }
 
-; Try a vector. Use special constants (NaN, INF, undef) because they don't change anything.
+; Try a vector. Use special constants (NaN, INF, poison) because they don't change anything.
 
 define <4 x double> @fdiv_op1_constant_fsub_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fdiv_op1_constant_fsub_vec(
-; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> [[X:%.*]], <double 4.200000e+01, double 0x7FF800000ABCD000, double 0x7FF0000000000000, double undef>
+; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> [[X:%.*]], <double 4.200000e+01, double 0x7FF800000ABCD000, double 0x7FF0000000000000, double poison>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %d = fdiv <4 x double> %x, <double -42.0, double 0xFFF800000ABCD000, double 0xFFF0000000000000, double undef>
+  %d = fdiv <4 x double> %x, <double -42.0, double 0xFFF800000ABCD000, double 0xFFF0000000000000, double poison>
   %r = fsub <4 x double> <double -0.0, double -0.0, double -0.0, double -0.0>, %d
   ret <4 x double> %r
 }
 
 define <4 x double> @fdiv_op1_constant_fneg_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fdiv_op1_constant_fneg_vec(
-; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> [[X:%.*]], <double 4.200000e+01, double 0x7FF800000ABCD000, double 0x7FF0000000000000, double undef>
+; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> [[X:%.*]], <double 4.200000e+01, double 0x7FF800000ABCD000, double 0x7FF0000000000000, double poison>
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %d = fdiv <4 x double> %x, <double -42.0, double 0xFFF800000ABCD000, double 0xFFF0000000000000, double undef>
+  %d = fdiv <4 x double> %x, <double -42.0, double 0xFFF800000ABCD000, double 0xFFF0000000000000, double poison>
   %r = fneg <4 x double> %d
   ret <4 x double> %r
 }
@@ -335,24 +335,24 @@ define float @fdiv_op0_constant_fneg_extra_use(float %x) {
   ret float %r
 }
 
-; Try a vector. Use special constants (NaN, INF, undef) because they don't change anything.
+; Try a vector. Use special constants (NaN, INF, poison) because they don't change anything.
 
 define <4 x double> @fdiv_op0_constant_fsub_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fdiv_op0_constant_fsub_vec(
-; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> <double 4.200000e+01, double 0xFFF8000000000000, double 0x7FF0000000000000, double undef>, [[X:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> <double 4.200000e+01, double 0xFFF8000000000000, double 0x7FF0000000000000, double poison>, [[X:%.*]]
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %d = fdiv <4 x double> <double -42.0, double 0x7FF8000000000000, double 0xFFF0000000000000, double undef>, %x
+  %d = fdiv <4 x double> <double -42.0, double 0x7FF8000000000000, double 0xFFF0000000000000, double poison>, %x
   %r = fsub <4 x double> <double -0.0, double -0.0, double -0.0, double -0.0>, %d
   ret <4 x double> %r
 }
 
 define <4 x double> @fdiv_op0_constant_fneg_vec(<4 x double> %x) {
 ; CHECK-LABEL: @fdiv_op0_constant_fneg_vec(
-; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> <double 4.200000e+01, double 0xFFF8000000000000, double 0x7FF0000000000000, double undef>, [[X:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = fdiv <4 x double> <double 4.200000e+01, double 0xFFF8000000000000, double 0x7FF0000000000000, double poison>, [[X:%.*]]
 ; CHECK-NEXT:    ret <4 x double> [[R]]
 ;
-  %d = fdiv <4 x double> <double -42.0, double 0x7FF8000000000000, double 0xFFF0000000000000, double undef>, %x
+  %d = fdiv <4 x double> <double -42.0, double 0x7FF8000000000000, double 0xFFF0000000000000, double poison>, %x
   %r = fneg <4 x double> %d
   ret <4 x double> %r
 }
@@ -584,11 +584,11 @@ define <2 x float> @fneg_nsz_fadd_constant_vec(<2 x float> %x) {
 
 define <2 x float> @fake_fneg_nsz_fadd_constant_vec(<2 x float> %x) {
 ; CHECK-LABEL: @fake_fneg_nsz_fadd_constant_vec(
-; CHECK-NEXT:    [[R:%.*]] = fsub nsz <2 x float> <float -4.200000e+01, float undef>, [[X:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = fsub nsz <2 x float> <float -4.200000e+01, float poison>, [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
-  %a = fadd <2 x float> %x, <float 42.0, float undef>
-  %r = fsub nsz <2 x float> <float undef, float -0.0>, %a
+  %a = fadd <2 x float> %x, <float 42.0, float poison>
+  %r = fsub nsz <2 x float> <float poison, float -0.0>, %a
   ret <2 x float> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fold-inc-of-add-of-not-x-and-y-to-sub-x-from-y.ll b/llvm/test/Transforms/InstCombine/fold-inc-of-add-of-not-x-and-y-to-sub-x-from-y.ll
index b482cfdfde197..1fd570bf2635b 100644
--- a/llvm/test/Transforms/InstCombine/fold-inc-of-add-of-not-x-and-y-to-sub-x-from-y.ll
+++ b/llvm/test/Transforms/InstCombine/fold-inc-of-add-of-not-x-and-y-to-sub-x-from-y.ll
@@ -36,36 +36,36 @@ define <4 x i32> @t1_vec_splat(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i32> %t2
 }
 
-define <4 x i32> @t2_vec_undef0(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @t2_vec_undef0(
+define <4 x i32> @t2_vec_poison0(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @t2_vec_poison0(
 ; CHECK-NEXT:    [[T2:%.*]] = sub <4 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[T2]]
 ;
-  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 undef, i32 -1>
+  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 poison, i32 -1>
   %t1 = add <4 x i32> %t0, %y
   %t2 = add <4 x i32> %t1, <i32 1, i32 1, i32 1, i32 1>
   ret <4 x i32> %t2
 }
 
-define <4 x i32> @t3_vec_undef1(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @t3_vec_undef1(
+define <4 x i32> @t3_vec_poison1(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @t3_vec_poison1(
 ; CHECK-NEXT:    [[T2:%.*]] = sub <4 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[T2]]
 ;
   %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 -1, i32 -1>
   %t1 = add <4 x i32> %t0, %y
-  %t2 = add <4 x i32> %t1, <i32 1, i32 1, i32 undef, i32 1>
+  %t2 = add <4 x i32> %t1, <i32 1, i32 1, i32 poison, i32 1>
   ret <4 x i32> %t2
 }
 
-define <4 x i32> @t4_vec_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @t4_vec_undef2(
+define <4 x i32> @t4_vec_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @t4_vec_poison2(
 ; CHECK-NEXT:    [[T2:%.*]] = sub <4 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[T2]]
 ;
-  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 undef, i32 -1>
+  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 poison, i32 -1>
   %t1 = add <4 x i32> %t0, %y
-  %t2 = add <4 x i32> %t1, <i32 1, i32 1, i32 undef, i32 1>
+  %t2 = add <4 x i32> %t1, <i32 1, i32 1, i32 poison, i32 1>
   ret <4 x i32> %t2
 }
 
diff --git a/llvm/test/Transforms/InstCombine/fold-sub-of-not-to-inc-of-add.ll b/llvm/test/Transforms/InstCombine/fold-sub-of-not-to-inc-of-add.ll
index 6f311f05fb017..af580ba57513c 100644
--- a/llvm/test/Transforms/InstCombine/fold-sub-of-not-to-inc-of-add.ll
+++ b/llvm/test/Transforms/InstCombine/fold-sub-of-not-to-inc-of-add.ll
@@ -50,13 +50,13 @@ define <4 x i32> @p1_vector_splat(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @p2_vector_undef(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @p2_vector_undef(
+define <4 x i32> @p2_vector_poison(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @p2_vector_poison(
 ; CHECK-NEXT:    [[T0_NEG:%.*]] = add <4 x i32> [[X:%.*]], <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[T1:%.*]] = add <4 x i32> [[T0_NEG]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[T1]]
 ;
-  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 undef, i32 -1>
+  %t0 = xor <4 x i32> %x, <i32 -1, i32 -1, i32 poison, i32 -1>
   %t1 = sub <4 x i32> %y, %t0
   ret <4 x i32> %t1
 }
diff --git a/llvm/test/Transforms/InstCombine/fpcast.ll b/llvm/test/Transforms/InstCombine/fpcast.ll
index d2c932ba447e4..69daac773a645 100644
--- a/llvm/test/Transforms/InstCombine/fpcast.ll
+++ b/llvm/test/Transforms/InstCombine/fpcast.ll
@@ -51,13 +51,13 @@ define half @unary_fneg_fptrunc(float %a) {
   ret half %c
 }
 
-define <2 x half> @fneg_fptrunc_vec_undef(<2 x float> %a) {
-; CHECK-LABEL: @fneg_fptrunc_vec_undef(
+define <2 x half> @fneg_fptrunc_vec_poison(<2 x float> %a) {
+; CHECK-LABEL: @fneg_fptrunc_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = fptrunc <2 x float> [[A:%.*]] to <2 x half>
 ; CHECK-NEXT:    [[C:%.*]] = fneg <2 x half> [[TMP1]]
 ; CHECK-NEXT:    ret <2 x half> [[C]]
 ;
-  %b = fsub <2 x float> <float -0.0, float undef>, %a
+  %b = fsub <2 x float> <float -0.0, float poison>, %a
   %c = fptrunc <2 x float> %b to <2 x half>
   ret <2 x half> %c
 }
diff --git a/llvm/test/Transforms/InstCombine/fsub.ll b/llvm/test/Transforms/InstCombine/fsub.ll
index 6e13c33b126d5..f1e7086e697e8 100644
--- a/llvm/test/Transforms/InstCombine/fsub.ll
+++ b/llvm/test/Transforms/InstCombine/fsub.ll
@@ -153,12 +153,12 @@ define <2 x float> @constant_op1_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x float> %r
 }
 
-define <2 x float> @constant_op1_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @constant_op1_vec_undef(
-; CHECK-NEXT:    [[R:%.*]] = fadd <2 x float> [[X:%.*]], <float undef, float 4.200000e+01>
+define <2 x float> @constant_op1_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @constant_op1_vec_poison(
+; CHECK-NEXT:    [[R:%.*]] = fadd <2 x float> [[X:%.*]], <float poison, float 4.200000e+01>
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
-  %r = fsub <2 x float> %x, <float undef, float -42.0>
+  %r = fsub <2 x float> %x, <float poison, float -42.0>
   ret <2 x float> %r
 }
 
@@ -204,12 +204,12 @@ define <2 x float> @unary_neg_op1_vec(<2 x float> %x, <2 x float> %y) {
   ret <2 x float> %r
 }
 
-define <2 x float> @neg_op1_vec_undef(<2 x float> %x, <2 x float> %y) {
-; CHECK-LABEL: @neg_op1_vec_undef(
+define <2 x float> @neg_op1_vec_poison(<2 x float> %x, <2 x float> %y) {
+; CHECK-LABEL: @neg_op1_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = fadd <2 x float> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
-  %negy = fsub <2 x float> <float -0.0, float undef>, %y
+  %negy = fsub <2 x float> <float -0.0, float poison>, %y
   %r = fsub <2 x float> %x, %negy
   ret <2 x float> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/funnel.ll b/llvm/test/Transforms/InstCombine/funnel.ll
index 162519e648f3e..a54e6e4642b75 100644
--- a/llvm/test/Transforms/InstCombine/funnel.ll
+++ b/llvm/test/Transforms/InstCombine/funnel.ll
@@ -43,24 +43,24 @@ define <2 x i16> @fshl_v2i16_constant_splat(<2 x i16> %x, <2 x i16> %y) {
   ret <2 x i16> %r
 }
 
-define <2 x i16> @fshl_v2i16_constant_splat_undef0(<2 x i16> %x, <2 x i16> %y) {
-; CHECK-LABEL: @fshl_v2i16_constant_splat_undef0(
+define <2 x i16> @fshl_v2i16_constant_splat_poison0(<2 x i16> %x, <2 x i16> %y) {
+; CHECK-LABEL: @fshl_v2i16_constant_splat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> [[X:%.*]], <2 x i16> [[Y:%.*]], <2 x i16> <i16 1, i16 1>)
 ; CHECK-NEXT:    ret <2 x i16> [[R]]
 ;
-  %shl = shl <2 x i16> %x, <i16 undef, i16 1>
+  %shl = shl <2 x i16> %x, <i16 poison, i16 1>
   %shr = lshr <2 x i16> %y, <i16 15, i16 15>
   %r = or <2 x i16> %shl, %shr
   ret <2 x i16> %r
 }
 
-define <2 x i16> @fshl_v2i16_constant_splat_undef1(<2 x i16> %x, <2 x i16> %y) {
-; CHECK-LABEL: @fshl_v2i16_constant_splat_undef1(
+define <2 x i16> @fshl_v2i16_constant_splat_poison1(<2 x i16> %x, <2 x i16> %y) {
+; CHECK-LABEL: @fshl_v2i16_constant_splat_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> [[X:%.*]], <2 x i16> [[Y:%.*]], <2 x i16> <i16 1, i16 1>)
 ; CHECK-NEXT:    ret <2 x i16> [[R]]
 ;
   %shl = shl <2 x i16> %x, <i16 1, i16 1>
-  %shr = lshr <2 x i16> %y, <i16 15, i16 undef>
+  %shr = lshr <2 x i16> %y, <i16 15, i16 poison>
   %r = or <2 x i16> %shl, %shr
   ret <2 x i16> %r
 }
@@ -78,30 +78,30 @@ define <2 x i17> @fshr_v2i17_constant_splat(<2 x i17> %x, <2 x i17> %y) {
   ret <2 x i17> %r
 }
 
-define <2 x i17> @fshr_v2i17_constant_splat_undef0(<2 x i17> %x, <2 x i17> %y) {
-; CHECK-LABEL: @fshr_v2i17_constant_splat_undef0(
+define <2 x i17> @fshr_v2i17_constant_splat_poison0(<2 x i17> %x, <2 x i17> %y) {
+; CHECK-LABEL: @fshr_v2i17_constant_splat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i17> @llvm.fshl.v2i17(<2 x i17> [[Y:%.*]], <2 x i17> [[X:%.*]], <2 x i17> <i17 5, i17 5>)
 ; CHECK-NEXT:    ret <2 x i17> [[R]]
 ;
-  %shr = lshr <2 x i17> %x, <i17 12, i17 undef>
-  %shl = shl <2 x i17> %y, <i17 undef, i17 5>
+  %shr = lshr <2 x i17> %x, <i17 12, i17 poison>
+  %shl = shl <2 x i17> %y, <i17 poison, i17 5>
   %r = or <2 x i17> %shr, %shl
   ret <2 x i17> %r
 }
 
-define <2 x i17> @fshr_v2i17_constant_splat_undef1(<2 x i17> %x, <2 x i17> %y) {
-; CHECK-LABEL: @fshr_v2i17_constant_splat_undef1(
+define <2 x i17> @fshr_v2i17_constant_splat_poison1(<2 x i17> %x, <2 x i17> %y) {
+; CHECK-LABEL: @fshr_v2i17_constant_splat_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i17> @llvm.fshl.v2i17(<2 x i17> [[Y:%.*]], <2 x i17> [[X:%.*]], <2 x i17> <i17 5, i17 5>)
 ; CHECK-NEXT:    ret <2 x i17> [[R]]
 ;
-  %shr = lshr <2 x i17> %x, <i17 12, i17 undef>
-  %shl = shl <2 x i17> %y, <i17 5, i17 undef>
+  %shr = lshr <2 x i17> %x, <i17 12, i17 poison>
+  %shl = shl <2 x i17> %y, <i17 5, i17 poison>
   %r = or <2 x i17> %shr, %shl
   ret <2 x i17> %r
 }
 
 ; Allow arbitrary shift constants.
-; Support undef elements.
+; Support poison elements.
 
 define <2 x i32> @fshr_v2i32_constant_nonsplat(<2 x i32> %x, <2 x i32> %y) {
 ; CHECK-LABEL: @fshr_v2i32_constant_nonsplat(
@@ -114,24 +114,24 @@ define <2 x i32> @fshr_v2i32_constant_nonsplat(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %r
 }
 
-define <2 x i32> @fshr_v2i32_constant_nonsplat_undef0(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @fshr_v2i32_constant_nonsplat_undef0(
+define <2 x i32> @fshr_v2i32_constant_nonsplat_poison0(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @fshr_v2i32_constant_nonsplat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[Y:%.*]], <2 x i32> [[X:%.*]], <2 x i32> <i32 0, i32 13>)
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %shr = lshr <2 x i32> %x, <i32 undef, i32 19>
+  %shr = lshr <2 x i32> %x, <i32 poison, i32 19>
   %shl = shl <2 x i32> %y, <i32 15, i32 13>
   %r = or <2 x i32> %shl, %shr
   ret <2 x i32> %r
 }
 
-define <2 x i32> @fshr_v2i32_constant_nonsplat_undef1(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @fshr_v2i32_constant_nonsplat_undef1(
-; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[Y:%.*]], <2 x i32> [[X:%.*]], <2 x i32> <i32 15, i32 0>)
+define <2 x i32> @fshr_v2i32_constant_nonsplat_poison1(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @fshr_v2i32_constant_nonsplat_poison1(
+; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[Y:%.*]], <2 x i32> [[X:%.*]], <2 x i32> <i32 15, i32 poison>)
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %shr = lshr <2 x i32> %x, <i32 17, i32 19>
-  %shl = shl <2 x i32> %y, <i32 15, i32 undef>
+  %shl = shl <2 x i32> %y, <i32 15, i32 poison>
   %r = or <2 x i32> %shl, %shr
   ret <2 x i32> %r
 }
@@ -147,13 +147,13 @@ define <2 x i36> @fshl_v2i36_constant_nonsplat(<2 x i36> %x, <2 x i36> %y) {
   ret <2 x i36> %r
 }
 
-define <3 x i36> @fshl_v3i36_constant_nonsplat_undef0(<3 x i36> %x, <3 x i36> %y) {
-; CHECK-LABEL: @fshl_v3i36_constant_nonsplat_undef0(
-; CHECK-NEXT:    [[R:%.*]] = call <3 x i36> @llvm.fshl.v3i36(<3 x i36> [[X:%.*]], <3 x i36> [[Y:%.*]], <3 x i36> <i36 21, i36 11, i36 0>)
+define <3 x i36> @fshl_v3i36_constant_nonsplat_poison0(<3 x i36> %x, <3 x i36> %y) {
+; CHECK-LABEL: @fshl_v3i36_constant_nonsplat_poison0(
+; CHECK-NEXT:    [[R:%.*]] = call <3 x i36> @llvm.fshl.v3i36(<3 x i36> [[X:%.*]], <3 x i36> [[Y:%.*]], <3 x i36> <i36 21, i36 11, i36 poison>)
 ; CHECK-NEXT:    ret <3 x i36> [[R]]
 ;
-  %shl = shl <3 x i36> %x, <i36 21, i36 11, i36 undef>
-  %shr = lshr <3 x i36> %y, <i36 15, i36 25, i36 undef>
+  %shl = shl <3 x i36> %x, <i36 21, i36 11, i36 poison>
+  %shr = lshr <3 x i36> %y, <i36 15, i36 25, i36 poison>
   %r = or <3 x i36> %shl, %shr
   ret <3 x i36> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/get-lowbitmask-upto-and-including-bit.ll b/llvm/test/Transforms/InstCombine/get-lowbitmask-upto-and-including-bit.ll
index 12a81f0cd2f0f..40caa57891369 100644
--- a/llvm/test/Transforms/InstCombine/get-lowbitmask-upto-and-including-bit.ll
+++ b/llvm/test/Transforms/InstCombine/get-lowbitmask-upto-and-including-bit.ll
@@ -41,36 +41,36 @@ define <2 x i8> @t2_vec(<2 x i8> %x) {
   %mask = or <2 x i8> %lowbitmask, %bitmask
   ret <2 x i8> %mask
 }
-define <3 x i8> @t3_vec_undef0(<3 x i8> %x) {
-; CHECK-LABEL: @t3_vec_undef0(
+define <3 x i8> @t3_vec_poison0(<3 x i8> %x) {
+; CHECK-LABEL: @t3_vec_poison0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub <3 x i8> <i8 7, i8 7, i8 7>, [[X:%.*]]
 ; CHECK-NEXT:    [[MASK:%.*]] = lshr <3 x i8> <i8 -1, i8 -1, i8 -1>, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[MASK]]
 ;
-  %bitmask = shl <3 x i8> <i8 1, i8 undef, i8 1>, %x
+  %bitmask = shl <3 x i8> <i8 1, i8 poison, i8 1>, %x
   %lowbitmask = add <3 x i8> %bitmask, <i8 -1, i8 -1, i8 -1>
   %mask = or <3 x i8> %lowbitmask, %bitmask
   ret <3 x i8> %mask
 }
-define <3 x i8> @t4_vec_undef1(<3 x i8> %x) {
-; CHECK-LABEL: @t4_vec_undef1(
+define <3 x i8> @t4_vec_poison1(<3 x i8> %x) {
+; CHECK-LABEL: @t4_vec_poison1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub <3 x i8> <i8 7, i8 7, i8 7>, [[X:%.*]]
 ; CHECK-NEXT:    [[MASK:%.*]] = lshr <3 x i8> <i8 -1, i8 -1, i8 -1>, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[MASK]]
 ;
   %bitmask = shl <3 x i8> <i8 1, i8 1, i8 1>, %x
-  %lowbitmask = add <3 x i8> %bitmask, <i8 -1, i8 undef, i8 -1>
+  %lowbitmask = add <3 x i8> %bitmask, <i8 -1, i8 poison, i8 -1>
   %mask = or <3 x i8> %lowbitmask, %bitmask
   ret <3 x i8> %mask
 }
-define <3 x i8> @t5_vec_undef2(<3 x i8> %x) {
-; CHECK-LABEL: @t5_vec_undef2(
+define <3 x i8> @t5_vec_poison2(<3 x i8> %x) {
+; CHECK-LABEL: @t5_vec_poison2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub <3 x i8> <i8 7, i8 7, i8 7>, [[X:%.*]]
 ; CHECK-NEXT:    [[MASK:%.*]] = lshr <3 x i8> <i8 -1, i8 -1, i8 -1>, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[MASK]]
 ;
-  %bitmask = shl <3 x i8> <i8 1, i8 1, i8 undef>, %x
-  %lowbitmask = add <3 x i8> %bitmask, <i8 -1, i8 undef, i8 -1>
+  %bitmask = shl <3 x i8> <i8 1, i8 1, i8 poison>, %x
+  %lowbitmask = add <3 x i8> %bitmask, <i8 -1, i8 poison, i8 -1>
   %mask = or <3 x i8> %lowbitmask, %bitmask
   ret <3 x i8> %mask
 }
diff --git a/llvm/test/Transforms/InstCombine/hoist-negation-out-of-bias-calculation.ll b/llvm/test/Transforms/InstCombine/hoist-negation-out-of-bias-calculation.ll
index c8f14595ea673..e4cae13519783 100644
--- a/llvm/test/Transforms/InstCombine/hoist-negation-out-of-bias-calculation.ll
+++ b/llvm/test/Transforms/InstCombine/hoist-negation-out-of-bias-calculation.ll
@@ -55,14 +55,14 @@ define <2 x i8> @t2_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i8> %negbias
 }
 
-define <2 x i8> @t3_vec_undef(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @t3_vec_undef(
+define <2 x i8> @t3_vec_poison(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @t3_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i8> [[Y:%.*]], <i8 -1, i8 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <2 x i8> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    [[NEGBIAS:%.*]] = sub <2 x i8> zeroinitializer, [[TMP2]]
 ; CHECK-NEXT:    ret <2 x i8> [[NEGBIAS]]
 ;
-  %negy = sub <2 x i8> <i8 0, i8 undef>, %y
+  %negy = sub <2 x i8> <i8 0, i8 poison>, %y
   %unbiasedx = and <2 x i8> %negy, %x
   %negbias = sub <2 x i8> %unbiasedx, %x
   ret <2 x i8> %negbias
diff --git a/llvm/test/Transforms/InstCombine/hoist-not-from-ashr-operand.ll b/llvm/test/Transforms/InstCombine/hoist-not-from-ashr-operand.ll
index e0242855e2683..2217666f0f49a 100644
--- a/llvm/test/Transforms/InstCombine/hoist-not-from-ashr-operand.ll
+++ b/llvm/test/Transforms/InstCombine/hoist-not-from-ashr-operand.ll
@@ -41,14 +41,14 @@ define <2 x i8> @t2_vec(<2 x i8> %x, <2 x i8> %y) {
   %ashr = ashr <2 x i8> %not_x, %y
   ret <2 x i8> %ashr
 }
-; Note that we must sanitize undef elts of -1 constant to -1 or 0.
-define <2 x i8> @t3_vec_undef(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @t3_vec_undef(
+; Note that we must sanitize poison elts of -1 constant to -1 or 0.
+define <2 x i8> @t3_vec_poison(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @t3_vec_poison(
 ; CHECK-NEXT:    [[NOT_X_NOT:%.*]] = ashr <2 x i8> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[ASHR:%.*]] = xor <2 x i8> [[NOT_X_NOT]], <i8 -1, i8 -1>
 ; CHECK-NEXT:    ret <2 x i8> [[ASHR]]
 ;
-  %not_x = xor <2 x i8> %x, <i8 -1, i8 undef>
+  %not_x = xor <2 x i8> %x, <i8 -1, i8 poison>
   %ashr = ashr <2 x i8> %not_x, %y
   ret <2 x i8> %ashr
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp-uge-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-uge-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
index 5adf476f7a79f..32ef6267cdf8b 100644
--- a/llvm/test/Transforms/InstCombine/icmp-uge-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-uge-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
@@ -56,8 +56,8 @@ define <3 x i1> @p2_vec_undef0(<3 x i8> %val, <3 x i8> %bits) {
 ; CHECK-LABEL: @p2_vec_undef0(
 ; CHECK-NEXT:    [[T0:%.*]] = shl <3 x i8> <i8 1, i8 undef, i8 1>, [[BITS:%.*]]
 ; CHECK-NEXT:    call void @use3i8(<3 x i8> [[T0]])
-; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS]]
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
+; CHECK-NEXT:    [[T1:%.*]] = add <3 x i8> [[T0]], <i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[R:%.*]] = icmp uge <3 x i8> [[T1]], [[VAL:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
   %t0 = shl <3 x i8> <i8 1, i8 undef, i8 1>, %bits
diff --git a/llvm/test/Transforms/InstCombine/icmp-uge-of-not-of-shl-allones-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-uge-of-not-of-shl-allones-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
index 7f4603881f23c..27b02c8c6e936 100644
--- a/llvm/test/Transforms/InstCombine/icmp-uge-of-not-of-shl-allones-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-uge-of-not-of-shl-allones-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
@@ -40,38 +40,38 @@ define <2 x i1> @p1_vec(<2 x i8> %val, <2 x i8> %bits) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef0(
+define <3 x i1> @p2_vec_poison0(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison0(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %bits
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %bits
   %t1 = xor <3 x i8> %t0, <i8 -1, i8 -1, i8 -1>
   %r = icmp uge <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef1(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef1(
+define <3 x i1> @p2_vec_poison1(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison1(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
   %t0 = shl <3 x i8> <i8 -1, i8 -1, i8 -1>, %bits
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %r = icmp uge <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef2(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef2(
+define <3 x i1> @p2_vec_poison2(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison2(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %bits
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %bits
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %r = icmp uge <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp-ugt-of-shl-1-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-ugt-of-shl-1-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
index 550e8bb17229f..72cfb5a9f8bd0 100644
--- a/llvm/test/Transforms/InstCombine/icmp-ugt-of-shl-1-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-ugt-of-shl-1-by-bits-and-val-to-icmp-eq-of-lshr-val-by-bits-and-0.ll
@@ -38,13 +38,13 @@ define <2 x i1> @p1_vec(<2 x i8> %val, <2 x i8> %bits) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef(
+define <3 x i1> @p2_vec_poison(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 1, i8 undef, i8 1>, %bits
+  %t0 = shl <3 x i8> <i8 1, i8 poison, i8 1>, %bits
   %r = icmp ugt <3 x i8> %t0, %val
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp-ule-of-shl-1-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-ule-of-shl-1-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
index 26b667d36728a..79e6914f09531 100644
--- a/llvm/test/Transforms/InstCombine/icmp-ule-of-shl-1-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-ule-of-shl-1-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
@@ -38,13 +38,13 @@ define <2 x i1> @p1_vec(<2 x i8> %val, <2 x i8> %bits) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef(
+define <3 x i1> @p2_vec_poison(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 1, i8 undef, i8 1>, %bits
+  %t0 = shl <3 x i8> <i8 1, i8 poison, i8 1>, %bits
   %r = icmp ule <3 x i8> %t0, %val
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp-ult-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-ult-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
index dd353d44218bf..25894a22f0075 100644
--- a/llvm/test/Transforms/InstCombine/icmp-ult-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-ult-of-add-of-shl-one-by-bits-to-allones-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
@@ -56,8 +56,8 @@ define <3 x i1> @p2_vec_undef0(<3 x i8> %val, <3 x i8> %bits) {
 ; CHECK-LABEL: @p2_vec_undef0(
 ; CHECK-NEXT:    [[T0:%.*]] = shl <3 x i8> <i8 1, i8 undef, i8 1>, [[BITS:%.*]]
 ; CHECK-NEXT:    call void @use3i8(<3 x i8> [[T0]])
-; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS]]
-; CHECK-NEXT:    [[R:%.*]] = icmp ne <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
+; CHECK-NEXT:    [[T1:%.*]] = add <3 x i8> [[T0]], <i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[R:%.*]] = icmp ult <3 x i8> [[T1]], [[VAL:%.*]]
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
   %t0 = shl <3 x i8> <i8 1, i8 undef, i8 1>, %bits
diff --git a/llvm/test/Transforms/InstCombine/icmp-ult-of-not-of-shl-allones-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll b/llvm/test/Transforms/InstCombine/icmp-ult-of-not-of-shl-allones-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
index c7a45c5cdc11a..8441033d4857e 100644
--- a/llvm/test/Transforms/InstCombine/icmp-ult-of-not-of-shl-allones-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-ult-of-not-of-shl-allones-by-bits-and-val-to-icmp-ne-of-lshr-val-by-bits-and-0.ll
@@ -40,38 +40,38 @@ define <2 x i1> @p1_vec(<2 x i8> %val, <2 x i8> %bits) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef0(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef0(
+define <3 x i1> @p2_vec_poison0(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison0(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %bits
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %bits
   %t1 = xor <3 x i8> %t0, <i8 -1, i8 -1, i8 -1>
   %r = icmp ult <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef1(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef1(
+define <3 x i1> @p2_vec_poison1(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison1(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
   %t0 = shl <3 x i8> <i8 -1, i8 -1, i8 -1>, %bits
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %r = icmp ult <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
 
-define <3 x i1> @p2_vec_undef2(<3 x i8> %val, <3 x i8> %bits) {
-; CHECK-LABEL: @p2_vec_undef2(
+define <3 x i1> @p2_vec_poison2(<3 x i8> %val, <3 x i8> %bits) {
+; CHECK-LABEL: @p2_vec_poison2(
 ; CHECK-NEXT:    [[VAL_HIGHBITS:%.*]] = lshr <3 x i8> [[VAL:%.*]], [[BITS:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne <3 x i8> [[VAL_HIGHBITS]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[R]]
 ;
-  %t0 = shl <3 x i8> <i8 -1, i8 undef, i8 -1>, %bits
-  %t1 = xor <3 x i8> %t0, <i8 -1, i8 undef, i8 -1>
+  %t0 = shl <3 x i8> <i8 -1, i8 poison, i8 -1>, %bits
+  %t1 = xor <3 x i8> %t0, <i8 -1, i8 poison, i8 -1>
   %r = icmp ult <3 x i8> %t1, %val
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp.ll b/llvm/test/Transforms/InstCombine/icmp.ll
index 10ab1fe118348..31093c7ca1036 100644
--- a/llvm/test/Transforms/InstCombine/icmp.ll
+++ b/llvm/test/Transforms/InstCombine/icmp.ll
@@ -1790,14 +1790,14 @@ define <2 x i1> @icmp_add20_eq_add57_splat(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_add20_eq_add57_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @icmp_add20_eq_add57_undef(
+define <2 x i1> @icmp_add20_eq_add57_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @icmp_add20_eq_add57_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i32> [[Y:%.*]], <i32 37, i32 37>
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <2 x i32> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
   %1 = add <2 x i32> %x, <i32 20, i32 20>
-  %2 = add <2 x i32> %y, <i32 57, i32 undef>
+  %2 = add <2 x i32> %y, <i32 57, i32 poison>
   %cmp = icmp eq <2 x i32> %1, %2
   ret <2 x i1> %cmp
 }
@@ -1838,14 +1838,14 @@ define <2 x i1> @icmp_sub57_ne_sub20_splat(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_sub57_ne_sub20_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @icmp_sub57_ne_sub20_vec_undef(
+define <2 x i1> @icmp_sub57_ne_sub20_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @icmp_sub57_ne_sub20_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i32> [[X:%.*]], <i32 -37, i32 -37>
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ne <2 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %1 = add <2 x i32> %x, <i32 -57, i32 undef>
-  %2 = add <2 x i32> %y, <i32 -20, i32 undef>
+  %1 = add <2 x i32> %x, <i32 -57, i32 poison>
+  %2 = add <2 x i32> %y, <i32 -20, i32 poison>
   %cmp = icmp ne <2 x i32> %1, %2
   ret <2 x i1> %cmp
 }
@@ -1926,14 +1926,14 @@ define <2 x i1> @icmp_add20_sge_add57_splat(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_add20_sge_add57_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @icmp_add20_sge_add57_undef(
+define <2 x i1> @icmp_add20_sge_add57_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @icmp_add20_sge_add57_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nsw <2 x i32> [[Y:%.*]], <i32 37, i32 37>
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sle <2 x i32> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
   %1 = add nsw <2 x i32> %x, <i32 20, i32 20>
-  %2 = add nsw <2 x i32> %y, <i32 57, i32 undef>
+  %2 = add nsw <2 x i32> %y, <i32 57, i32 poison>
   %cmp = icmp sge <2 x i32> %1, %2
   ret <2 x i1> %cmp
 }
@@ -1975,14 +1975,14 @@ define <2 x i1> @icmp_sub57_sge_sub20_splat(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_sub57_sge_sub20_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @icmp_sub57_sge_sub20_vec_undef(
+define <2 x i1> @icmp_sub57_sge_sub20_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @icmp_sub57_sge_sub20_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nsw <2 x i32> [[X:%.*]], <i32 -37, i32 -37>
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sge <2 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %1 = add nsw <2 x i32> %x, <i32 -57, i32 undef>
-  %2 = add nsw <2 x i32> %y, <i32 -20, i32 undef>
+  %1 = add nsw <2 x i32> %x, <i32 -57, i32 poison>
+  %2 = add nsw <2 x i32> %y, <i32 -20, i32 poison>
   %cmp = icmp sge <2 x i32> %1, %2
   ret <2 x i1> %cmp
 }
@@ -2557,13 +2557,13 @@ define <2 x i1> @or_icmp_eq_B_0_icmp_ult_A_B_uniform(<2 x i64> %a, <2 x i64> %b)
   ret <2 x i1> %3
 }
 
-define <2 x i1> @or_icmp_eq_B_0_icmp_ult_A_B_undef(<2 x i64> %a, <2 x i64> %b) {
-; CHECK-LABEL: @or_icmp_eq_B_0_icmp_ult_A_B_undef(
+define <2 x i1> @or_icmp_eq_B_0_icmp_ult_A_B_poison(<2 x i64> %a, <2 x i64> %b) {
+; CHECK-LABEL: @or_icmp_eq_B_0_icmp_ult_A_B_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i64> [[B:%.*]], <i64 -1, i64 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp uge <2 x i64> [[TMP1]], [[A:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[TMP2]]
 ;
-  %1 = icmp eq <2 x i64> %b, <i64 0, i64 undef>
+  %1 = icmp eq <2 x i64> %b, <i64 0, i64 poison>
   %2 = icmp ult <2 x i64> %a, %b
   %3 = or <2 x i1> %1, %2
   ret <2 x i1> %3
@@ -2606,14 +2606,14 @@ define <2 x i1> @or_icmp_ne_A_0_icmp_ne_B_0_uniform(<2 x i64> %a, <2 x i64> %b)
   ret <2 x i1> %3
 }
 
-define <2 x i1> @or_icmp_ne_A_0_icmp_ne_B_0_undef(<2 x i64> %a, <2 x i64> %b) {
-; CHECK-LABEL: @or_icmp_ne_A_0_icmp_ne_B_0_undef(
+define <2 x i1> @or_icmp_ne_A_0_icmp_ne_B_0_poison(<2 x i64> %a, <2 x i64> %b) {
+; CHECK-LABEL: @or_icmp_ne_A_0_icmp_ne_B_0_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <2 x i64> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <2 x i64> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[TMP2]]
 ;
-  %1 = icmp ne <2 x i64> %a, <i64 0, i64 undef>
-  %2 = icmp ne <2 x i64> %b, <i64 0, i64 undef>
+  %1 = icmp ne <2 x i64> %a, <i64 0, i64 poison>
+  %2 = icmp ne <2 x i64> %b, <i64 0, i64 poison>
   %3 = or <2 x i1> %1, %2
   ret <2 x i1> %3
 }
@@ -2803,13 +2803,13 @@ define <2 x i1> @and_icmp_ne_B_0_icmp_uge_A_B_uniform(<2 x i64> %a, <2 x i64> %b
   ret <2 x i1> %3
 }
 
-define <2 x i1> @and_icmp_ne_B_0_icmp_uge_A_B_undef(<2 x i64> %a, <2 x i64> %b) {
-; CHECK-LABEL: @and_icmp_ne_B_0_icmp_uge_A_B_undef(
+define <2 x i1> @and_icmp_ne_B_0_icmp_uge_A_B_poison(<2 x i64> %a, <2 x i64> %b) {
+; CHECK-LABEL: @and_icmp_ne_B_0_icmp_uge_A_B_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i64> [[B:%.*]], <i64 -1, i64 -1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult <2 x i64> [[TMP1]], [[A:%.*]]
 ; CHECK-NEXT:    ret <2 x i1> [[TMP2]]
 ;
-  %1 = icmp ne <2 x i64> %b, <i64 0, i64 undef>
+  %1 = icmp ne <2 x i64> %b, <i64 0, i64 poison>
   %2 = icmp uge <2 x i64> %a, %b
   %3 = and <2 x i1> %1, %2
   ret <2 x i1> %3
@@ -3272,13 +3272,13 @@ define <2 x i1> @icmp_and_or_lshr_cst_vec_nonuniform(<2 x i32> %x) {
   ret <2 x i1> %ret
 }
 
-define <2 x i1> @icmp_and_or_lshr_cst_vec_undef(<2 x i32> %x) {
-; CHECK-LABEL: @icmp_and_or_lshr_cst_vec_undef(
+define <2 x i1> @icmp_and_or_lshr_cst_vec_poison(<2 x i32> %x) {
+; CHECK-LABEL: @icmp_and_or_lshr_cst_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[X:%.*]], <i32 3, i32 poison>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ne <2 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[RET]]
 ;
-  %shf = lshr <2 x i32> %x, <i32 1, i32 undef>
+  %shf = lshr <2 x i32> %x, <i32 1, i32 poison>
   %or = or <2 x i32> %shf, %x
   %and = and <2 x i32> %or, <i32 1, i32 1>
   %ret = icmp ne <2 x i32> %and, zeroinitializer
@@ -3315,15 +3315,15 @@ define <2 x i1> @icmp_and_or_lshr_cst_vec_nonuniform_commute(<2 x i32> %xp) {
   ret <2 x i1> %ret
 }
 
-define <2 x i1> @icmp_and_or_lshr_cst_vec_undef_commute(<2 x i32> %xp) {
-; CHECK-LABEL: @icmp_and_or_lshr_cst_vec_undef_commute(
+define <2 x i1> @icmp_and_or_lshr_cst_vec_poison_commute(<2 x i32> %xp) {
+; CHECK-LABEL: @icmp_and_or_lshr_cst_vec_poison_commute(
 ; CHECK-NEXT:    [[X:%.*]] = srem <2 x i32> [[XP:%.*]], <i32 42, i32 42>
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[X]], <i32 3, i32 poison>
 ; CHECK-NEXT:    [[RET:%.*]] = icmp ne <2 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[RET]]
 ;
   %x = srem <2 x i32> %xp, <i32 42, i32 -42> ; prevent complexity-based canonicalization
-  %shf = lshr <2 x i32> %x, <i32 1, i32 undef>
+  %shf = lshr <2 x i32> %x, <i32 1, i32 poison>
   %or = or <2 x i32> %x, %shf
   %and = and <2 x i32> %or, <i32 1, i32 1>
   %ret = icmp ne <2 x i32> %and, zeroinitializer
@@ -4360,7 +4360,7 @@ define <2 x i1> @signbit_false_logic(<2 x i5> %x) {
 ; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i5> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %dec = add <2 x i5> %x,  <i5 -1, i5 undef>
+  %dec = add <2 x i5> %x,  <i5 -1, i5 poison>
   %not = xor <2 x i5> %x,  <i5 -1, i5 -1>
   %and = and <2 x i5> %dec, %not
   %r = icmp sgt <2 x i5> %and, <i5 -1, i5 -1>
diff --git a/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll b/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
index 7cef922eaf0ce..c7e0553992b90 100644
--- a/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
+++ b/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
@@ -86,9 +86,9 @@ define <2 x i8> @t4_splat(<2 x i8> %x) {
   ret <2 x i8> %x.roundedup
 }
 
-; Splat-with-undef
-define <2 x i8> @t5_splat_undef_0b0001(<2 x i8> %x) {
-; CHECK-LABEL: @t5_splat_undef_0b0001(
+; Splat-with-poison
+define <2 x i8> @t5_splat_poison_0b0001(<2 x i8> %x) {
+; CHECK-LABEL: @t5_splat_poison_0b0001(
 ; CHECK-NEXT:    [[X_BIASED1:%.*]] = add <2 x i8> [[X:%.*]], <i8 15, i8 15>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = and <2 x i8> [[X_BIASED1]], <i8 -16, i8 -16>
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
@@ -96,43 +96,43 @@ define <2 x i8> @t5_splat_undef_0b0001(<2 x i8> %x) {
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 15>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
   %x.biased = add <2 x i8> %x, <i8 16, i8 16>
-  %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 undef>
+  %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 poison>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t5_splat_undef_0b0010(<2 x i8> %x) {
-; CHECK-LABEL: @t5_splat_undef_0b0010(
+define <2 x i8> @t5_splat_poison_0b0010(<2 x i8> %x) {
+; CHECK-LABEL: @t5_splat_poison_0b0010(
 ; CHECK-NEXT:    [[X_BIASED1:%.*]] = add <2 x i8> [[X:%.*]], <i8 15, i8 15>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = and <2 x i8> [[X_BIASED1]], <i8 -16, i8 -16>
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 15>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
-  %x.biased = add <2 x i8> %x, <i8 16, i8 undef>
+  %x.biased = add <2 x i8> %x, <i8 16, i8 poison>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -16>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t5_splat_undef_0b0100(<2 x i8> %x) {
-; CHECK-LABEL: @t5_splat_undef_0b0100(
+define <2 x i8> @t5_splat_poison_0b0100(<2 x i8> %x) {
+; CHECK-LABEL: @t5_splat_poison_0b0100(
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X:%.*]], <i8 15, i8 15>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 -16>
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 15>
-  %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 undef>
+  %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 poison>
   %x.biased = add <2 x i8> %x, <i8 16, i8 16>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -16>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t5_splat_undef_0b1000(<2 x i8> %x) {
-; CHECK-LABEL: @t5_splat_undef_0b1000(
+define <2 x i8> @t5_splat_poison_0b1000(<2 x i8> %x) {
+; CHECK-LABEL: @t5_splat_poison_0b1000(
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X:%.*]], <i8 15, i8 15>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 -16>
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
-  %x.lowbits = and <2 x i8> %x, <i8 15, i8 undef>
+  %x.lowbits = and <2 x i8> %x, <i8 15, i8 poison>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
   %x.biased = add <2 x i8> %x, <i8 16, i8 16>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -16>
@@ -177,64 +177,64 @@ define <2 x i8> @t7_nonsplat_bias(<2 x i8> %x) {
 }
 
 ; Splat-in-disguise vector tests
-define <2 x i8> @t8_nonsplat_masked_by_undef_0b0001(<2 x i8> %x) {
-; CHECK-LABEL: @t8_nonsplat_masked_by_undef_0b0001(
+define <2 x i8> @t8_nonsplat_masked_by_poison_0b0001(<2 x i8> %x) {
+; CHECK-LABEL: @t8_nonsplat_masked_by_poison_0b0001(
 ; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i8> [[X:%.*]], <i8 15, i8 31>
 ; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i8> [[X_LOWBITS]], zeroinitializer
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X]], <i8 16, i8 32>
-; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 undef>
+; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 poison>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i8> [[X]], <2 x i8> [[X_BIASED_HIGHBITS]]
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 31>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
   %x.biased = add <2 x i8> %x, <i8 16, i8 32>
-  %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 undef>
+  %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 poison>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t8_nonsplat_masked_by_undef_0b0010(<2 x i8> %x) {
-; CHECK-LABEL: @t8_nonsplat_masked_by_undef_0b0010(
+define <2 x i8> @t8_nonsplat_masked_by_poison_0b0010(<2 x i8> %x) {
+; CHECK-LABEL: @t8_nonsplat_masked_by_poison_0b0010(
 ; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i8> [[X:%.*]], <i8 15, i8 31>
 ; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i8> [[X_LOWBITS]], zeroinitializer
-; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X]], <i8 16, i8 undef>
+; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X]], <i8 16, i8 poison>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 -32>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i8> [[X]], <2 x i8> [[X_BIASED_HIGHBITS]]
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 31>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
-  %x.biased = add <2 x i8> %x, <i8 16, i8 undef>
+  %x.biased = add <2 x i8> %x, <i8 16, i8 poison>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -32>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t8_nonsplat_masked_by_undef_0b0100(<2 x i8> %x) {
-; CHECK-LABEL: @t8_nonsplat_masked_by_undef_0b0100(
+define <2 x i8> @t8_nonsplat_masked_by_poison_0b0100(<2 x i8> %x) {
+; CHECK-LABEL: @t8_nonsplat_masked_by_poison_0b0100(
 ; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i8> [[X:%.*]], <i8 15, i8 31>
-; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i8> [[X_LOWBITS]], <i8 0, i8 undef>
+; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i8> [[X_LOWBITS]], <i8 0, i8 poison>
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X]], <i8 16, i8 32>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 -32>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i8> [[X]], <2 x i8> [[X_BIASED_HIGHBITS]]
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i8> %x, <i8 15, i8 31>
-  %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 undef>
+  %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 poison>
   %x.biased = add <2 x i8> %x, <i8 16, i8 32>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -32>
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i8> %x, <2 x i8> %x.biased.highbits
   ret <2 x i8> %x.roundedup
 }
-define <2 x i8> @t8_nonsplat_masked_by_undef_0b1000(<2 x i8> %x) {
-; CHECK-LABEL: @t8_nonsplat_masked_by_undef_0b1000(
-; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i8> [[X:%.*]], <i8 15, i8 undef>
+define <2 x i8> @t8_nonsplat_masked_by_poison_0b1000(<2 x i8> %x) {
+; CHECK-LABEL: @t8_nonsplat_masked_by_poison_0b1000(
+; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i8> [[X:%.*]], <i8 15, i8 poison>
 ; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i8> [[X_LOWBITS]], zeroinitializer
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i8> [[X]], <i8 16, i8 32>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i8> [[X_BIASED]], <i8 -16, i8 -32>
 ; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i8> [[X]], <2 x i8> [[X_BIASED_HIGHBITS]]
 ; CHECK-NEXT:    ret <2 x i8> [[X_ROUNDEDUP]]
 ;
-  %x.lowbits = and <2 x i8> %x, <i8 15, i8 undef>
+  %x.lowbits = and <2 x i8> %x, <i8 15, i8 poison>
   %x.lowbits.are.zero = icmp eq <2 x i8> %x.lowbits, <i8 0, i8 0>
   %x.biased = add <2 x i8> %x, <i8 16, i8 32>
   %x.biased.highbits = and <2 x i8> %x.biased, <i8 -16, i8 -32>
@@ -442,28 +442,28 @@ define i8 @t17_oneuse(i8 %x) {
 define <2 x i4> @t18_replacement_0b0001(<2 x i4> %x) {
 ; CHECK-LABEL: @t18_replacement_0b0001(
 ; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X:%.*]], <i4 3, i4 3>
-; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i4> [[X_BIASED]], <i4 -4, i4 undef>
+; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i4> [[X_BIASED]], <i4 -4, i4 poison>
 ; CHECK-NEXT:    call void @use.v2i4(<2 x i4> [[X_BIASED_HIGHBITS]])
 ; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
 ;
   %x.lowbits = and <2 x i4> %x, <i4 3, i4 3>
   %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 0>
   %x.biased = add <2 x i4> %x, <i4 3, i4 3>
-  %x.biased.highbits = and <2 x i4> %x.biased, <i4 -4, i4 undef>
+  %x.biased.highbits = and <2 x i4> %x.biased, <i4 -4, i4 poison>
   call void @use.v2i4(<2 x i4> %x.biased.highbits)
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i4> %x, <2 x i4> %x.biased.highbits
   ret <2 x i4> %x.roundedup
 }
 define <2 x i4> @t18_replacement_0b0010(<2 x i4> %x) {
 ; CHECK-LABEL: @t18_replacement_0b0010(
-; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X:%.*]], <i4 3, i4 undef>
+; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X:%.*]], <i4 3, i4 poison>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i4> [[X_BIASED]], <i4 -4, i4 -4>
 ; CHECK-NEXT:    call void @use.v2i4(<2 x i4> [[X_BIASED_HIGHBITS]])
 ; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
 ;
   %x.lowbits = and <2 x i4> %x, <i4 3, i4 3>
   %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 0>
-  %x.biased = add <2 x i4> %x, <i4 3, i4 undef>
+  %x.biased = add <2 x i4> %x, <i4 3, i4 poison>
   %x.biased.highbits = and <2 x i4> %x.biased, <i4 -4, i4 -4>
   call void @use.v2i4(<2 x i4> %x.biased.highbits)
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i4> %x, <2 x i4> %x.biased.highbits
@@ -477,7 +477,7 @@ define <2 x i4> @t18_replacement_0b0100(<2 x i4> %x) {
 ; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
 ;
   %x.lowbits = and <2 x i4> %x, <i4 3, i4 3>
-  %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 undef>
+  %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 poison>
   %x.biased = add <2 x i4> %x, <i4 3, i4 3>
   %x.biased.highbits = and <2 x i4> %x.biased, <i4 -4, i4 -4>
   call void @use.v2i4(<2 x i4> %x.biased.highbits)
@@ -491,7 +491,7 @@ define <2 x i4> @t18_replacement_0b1000(<2 x i4> %x) {
 ; CHECK-NEXT:    call void @use.v2i4(<2 x i4> [[X_BIASED_HIGHBITS]])
 ; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
 ;
-  %x.lowbits = and <2 x i4> %x, <i4 3, i4 undef>
+  %x.lowbits = and <2 x i4> %x, <i4 3, i4 poison>
   %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 0>
   %x.biased = add <2 x i4> %x, <i4 3, i4 3>
   %x.biased.highbits = and <2 x i4> %x.biased, <i4 -4, i4 -4>
diff --git a/llvm/test/Transforms/InstCombine/invert-variable-mask-in-masked-merge-vector.ll b/llvm/test/Transforms/InstCombine/invert-variable-mask-in-masked-merge-vector.ll
index 486113202ddd7..a76662c4bc439 100644
--- a/llvm/test/Transforms/InstCombine/invert-variable-mask-in-masked-merge-vector.ll
+++ b/llvm/test/Transforms/InstCombine/invert-variable-mask-in-masked-merge-vector.ll
@@ -20,14 +20,14 @@ define <2 x i4> @vector (<2 x i4> %x, <2 x i4> %y, <2 x i4> %m) {
   ret <2 x i4> %r
 }
 
-define <3 x i4> @vector_undef (<3 x i4> %x, <3 x i4> %y, <3 x i4> %m) {
-; CHECK-LABEL: @vector_undef(
+define <3 x i4> @vector_poison (<3 x i4> %x, <3 x i4> %y, <3 x i4> %m) {
+; CHECK-LABEL: @vector_poison(
 ; CHECK-NEXT:    [[N0:%.*]] = xor <3 x i4> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i4> [[N0]], [[M:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = xor <3 x i4> [[TMP1]], [[X]]
 ; CHECK-NEXT:    ret <3 x i4> [[R]]
 ;
-  %im = xor <3 x i4> %m, <i4 -1, i4 undef, i4 -1>
+  %im = xor <3 x i4> %m, <i4 -1, i4 poison, i4 -1>
   %n0 = xor <3 x i4> %x, %y
   %n1 = and <3 x i4> %n0, %im
   %r  = xor <3 x i4> %n1, %y
@@ -78,17 +78,17 @@ define <2 x i4> @in_constant_varx_6_invmask_nonsplat(<2 x i4> %x, <2 x i4> %mask
   ret <2 x i4> %r
 }
 
-define <3 x i4> @in_constant_varx_6_invmask_undef(<3 x i4> %x, <3 x i4> %mask) {
-; CHECK-LABEL: @in_constant_varx_6_invmask_undef(
-; CHECK-NEXT:    [[N0:%.*]] = xor <3 x i4> [[X:%.*]], <i4 6, i4 undef, i4 7>
+define <3 x i4> @in_constant_varx_6_invmask_poison(<3 x i4> %x, <3 x i4> %mask) {
+; CHECK-LABEL: @in_constant_varx_6_invmask_poison(
+; CHECK-NEXT:    [[N0:%.*]] = xor <3 x i4> [[X:%.*]], <i4 6, i4 poison, i4 7>
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i4> [[N0]], [[MASK:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = xor <3 x i4> [[TMP1]], [[X]]
 ; CHECK-NEXT:    ret <3 x i4> [[R]]
 ;
-  %notmask = xor <3 x i4> %mask, <i4 -1, i4 undef, i4 -1>
-  %n0 = xor <3 x i4> %x, <i4 6, i4 undef, i4 7> ; %x
+  %notmask = xor <3 x i4> %mask, <i4 -1, i4 poison, i4 -1>
+  %n0 = xor <3 x i4> %x, <i4 6, i4 poison, i4 7> ; %x
   %n1 = and <3 x i4> %n0, %notmask
-  %r = xor <3 x i4> %n1, <i4 6, i4 undef, i4 7>
+  %r = xor <3 x i4> %n1, <i4 6, i4 poison, i4 7>
   ret <3 x i4> %r
 }
 
@@ -133,15 +133,15 @@ define <2 x i4> @in_constant_6_vary_invmask_nonsplat(<2 x i4> %y, <2 x i4> %mask
   ret <2 x i4> %r
 }
 
-define <3 x i4> @in_constant_6_vary_invmask_undef(<3 x i4> %y, <3 x i4> %mask) {
-; CHECK-LABEL: @in_constant_6_vary_invmask_undef(
-; CHECK-NEXT:    [[N0:%.*]] = xor <3 x i4> [[Y:%.*]], <i4 6, i4 undef, i4 6>
+define <3 x i4> @in_constant_6_vary_invmask_poison(<3 x i4> %y, <3 x i4> %mask) {
+; CHECK-LABEL: @in_constant_6_vary_invmask_poison(
+; CHECK-NEXT:    [[N0:%.*]] = xor <3 x i4> [[Y:%.*]], <i4 6, i4 poison, i4 6>
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i4> [[N0]], [[MASK:%.*]]
-; CHECK-NEXT:    [[R:%.*]] = xor <3 x i4> [[TMP1]], <i4 6, i4 undef, i4 6>
+; CHECK-NEXT:    [[R:%.*]] = xor <3 x i4> [[TMP1]], <i4 6, i4 poison, i4 6>
 ; CHECK-NEXT:    ret <3 x i4> [[R]]
 ;
-  %notmask = xor <3 x i4> %mask, <i4 -1, i4 undef, i4 -1>
-  %n0 = xor <3 x i4> %y, <i4 6, i4 undef, i4 6> ; %x
+  %notmask = xor <3 x i4> %mask, <i4 -1, i4 poison, i4 -1>
+  %n0 = xor <3 x i4> %y, <i4 6, i4 poison, i4 6> ; %x
   %n1 = and <3 x i4> %n0, %notmask
   %r = xor <3 x i4> %n1, %y
   ret <3 x i4> %r
diff --git a/llvm/test/Transforms/InstCombine/lshr-and-negC-icmpeq-zero.ll b/llvm/test/Transforms/InstCombine/lshr-and-negC-icmpeq-zero.ll
index 847a7940bad8c..5d058b20be720 100644
--- a/llvm/test/Transforms/InstCombine/lshr-and-negC-icmpeq-zero.ll
+++ b/llvm/test/Transforms/InstCombine/lshr-and-negC-icmpeq-zero.ll
@@ -81,39 +81,39 @@ define <4 x i1> @vec_4xi32_lshr_and_negC_eq(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_lshr_and_negC_eq_undef1(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_lshr_and_negC_eq_undef1(
+define <4 x i1> @vec_lshr_and_negC_eq_poison1(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_lshr_and_negC_eq_poison1(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[LSHR]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
-  %and = and <4 x i32> %lshr, <i32 4294967288, i32 undef, i32 4294967288, i32 4294967288>  ; ~7
+  %and = and <4 x i32> %lshr, <i32 4294967288, i32 poison, i32 4294967288, i32 4294967288>  ; ~7
   %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_lshr_and_negC_eq_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_lshr_and_negC_eq_undef2(
+define <4 x i1> @vec_lshr_and_negC_eq_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_lshr_and_negC_eq_poison2(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[LSHR]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
   %and = and <4 x i32> %lshr, <i32 4294967288, i32 4294967288, i32 4294967288, i32 4294967288>  ; ~7
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_lshr_and_negC_eq_undef3(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_lshr_and_negC_eq_undef3(
+define <4 x i1> @vec_lshr_and_negC_eq_poison3(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_lshr_and_negC_eq_poison3(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[LSHR]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
-  %and = and <4 x i32> %lshr, <i32 4294967288, i32 4294967288, i32 undef, i32 4294967288>  ; ~7
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %and = and <4 x i32> %lshr, <i32 4294967288, i32 4294967288, i32 poison, i32 4294967288>  ; ~7
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/lshr-and-signbit-icmpeq-zero.ll b/llvm/test/Transforms/InstCombine/lshr-and-signbit-icmpeq-zero.ll
index 39f4e58b25dc8..0166680309ea8 100644
--- a/llvm/test/Transforms/InstCombine/lshr-and-signbit-icmpeq-zero.ll
+++ b/llvm/test/Transforms/InstCombine/lshr-and-signbit-icmpeq-zero.ll
@@ -81,39 +81,39 @@ define <4 x i1> @vec_4xi32_lshr_and_signbit_eq(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_undef1(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_undef1(
+define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_poison1(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_poison1(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[LSHR]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
-  %and = and <4 x i32> %lshr, <i32 2147483648, i32 undef, i32 2147483648, i32 2147483648>
+  %and = and <4 x i32> %lshr, <i32 2147483648, i32 poison, i32 2147483648, i32 2147483648>
   %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_undef2(
+define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_poison2(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[LSHR]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
   %and = and <4 x i32> %lshr, <i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648>
-  %r = icmp eq <4 x i32> %and, <i32 undef, i32 0, i32 0, i32 0>
+  %r = icmp eq <4 x i32> %and, <i32 poison, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_undef3(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_undef3(
+define <4 x i1> @vec_4xi32_lshr_and_signbit_eq_poison3(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_lshr_and_signbit_eq_poison3(
 ; CHECK-NEXT:    [[LSHR:%.*]] = lshr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[LSHR]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %lshr = lshr <4 x i32> %x, %y
-  %and = and <4 x i32> %lshr, <i32 2147483648, i32 undef, i32 2147483648, i32 2147483648>
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %and = and <4 x i32> %lshr, <i32 2147483648, i32 poison, i32 2147483648, i32 2147483648>
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/masked-merge-add.ll b/llvm/test/Transforms/InstCombine/masked-merge-add.ll
index f655153108a43..0484369e99d6a 100644
--- a/llvm/test/Transforms/InstCombine/masked-merge-add.ll
+++ b/llvm/test/Transforms/InstCombine/masked-merge-add.ll
@@ -51,7 +51,7 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
 ; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
 ; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 undef, i32 -1>
 ; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
-; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    [[RET:%.*]] = add <3 x i32> [[AND]], [[AND1]]
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
   %and = and <3 x i32> %x, %m
@@ -61,6 +61,21 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
   ret <3 x i32> %ret
 }
 
+define <3 x i32> @p_vec_poison(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m) {
+; CHECK-LABEL: @p_vec_poison(
+; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
+; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
+; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    ret <3 x i32> [[RET]]
+;
+  %and = and <3 x i32> %x, %m
+  %neg = xor <3 x i32> %m, <i32 -1, i32 poison, i32 -1>
+  %and1 = and <3 x i32> %neg, %y
+  %ret = add <3 x i32> %and, %and1
+  ret <3 x i32> %ret
+}
+
 ; ============================================================================ ;
 ; Constant mask.
 ; ============================================================================ ;
diff --git a/llvm/test/Transforms/InstCombine/masked-merge-or.ll b/llvm/test/Transforms/InstCombine/masked-merge-or.ll
index b49ec07706e28..0531a532fc7e0 100644
--- a/llvm/test/Transforms/InstCombine/masked-merge-or.ll
+++ b/llvm/test/Transforms/InstCombine/masked-merge-or.ll
@@ -51,7 +51,7 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
 ; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
 ; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 undef, i32 -1>
 ; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
-; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    [[RET:%.*]] = or <3 x i32> [[AND]], [[AND1]]
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
   %and = and <3 x i32> %x, %m
@@ -61,6 +61,21 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
   ret <3 x i32> %ret
 }
 
+define <3 x i32> @p_vec_poison(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m) {
+; CHECK-LABEL: @p_vec_poison(
+; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
+; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
+; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    ret <3 x i32> [[RET]]
+;
+  %and = and <3 x i32> %x, %m
+  %neg = xor <3 x i32> %m, <i32 -1, i32 poison, i32 -1>
+  %and1 = and <3 x i32> %neg, %y
+  %ret = or <3 x i32> %and, %and1
+  ret <3 x i32> %ret
+}
+
 ; ============================================================================ ;
 ; Constant mask.
 ; ============================================================================ ;
diff --git a/llvm/test/Transforms/InstCombine/masked-merge-xor.ll b/llvm/test/Transforms/InstCombine/masked-merge-xor.ll
index a6d201be68cee..74cc7625aebff 100644
--- a/llvm/test/Transforms/InstCombine/masked-merge-xor.ll
+++ b/llvm/test/Transforms/InstCombine/masked-merge-xor.ll
@@ -51,7 +51,7 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
 ; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
 ; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 undef, i32 -1>
 ; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
-; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    [[RET:%.*]] = xor <3 x i32> [[AND]], [[AND1]]
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
   %and = and <3 x i32> %x, %m
@@ -61,6 +61,21 @@ define <3 x i32> @p_vec_undef(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m)
   ret <3 x i32> %ret
 }
 
+define <3 x i32> @p_vec_poison(<3 x i32> %x, <3 x i32> %y, <3 x i32> noundef %m) {
+; CHECK-LABEL: @p_vec_poison(
+; CHECK-NEXT:    [[AND:%.*]] = and <3 x i32> [[X:%.*]], [[M:%.*]]
+; CHECK-NEXT:    [[NEG:%.*]] = xor <3 x i32> [[M]], <i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[AND1:%.*]] = and <3 x i32> [[NEG]], [[Y:%.*]]
+; CHECK-NEXT:    [[RET:%.*]] = or disjoint <3 x i32> [[AND]], [[AND1]]
+; CHECK-NEXT:    ret <3 x i32> [[RET]]
+;
+  %and = and <3 x i32> %x, %m
+  %neg = xor <3 x i32> %m, <i32 -1, i32 poison, i32 -1>
+  %and1 = and <3 x i32> %neg, %y
+  %ret = xor <3 x i32> %and, %and1
+  ret <3 x i32> %ret
+}
+
 ; ============================================================================ ;
 ; Constant mask.
 ; ============================================================================ ;
diff --git a/llvm/test/Transforms/InstCombine/min-positive.ll b/llvm/test/Transforms/InstCombine/min-positive.ll
index 1fb212b738725..d2c2e9018792b 100644
--- a/llvm/test/Transforms/InstCombine/min-positive.ll
+++ b/llvm/test/Transforms/InstCombine/min-positive.ll
@@ -67,16 +67,16 @@ define <2 x i1> @smin_commute_vec(<2 x i32> %x, <2 x i32> %other) {
   ret <2 x i1> %test
 }
 
-define <2 x i1> @smin_commute_vec_undef_elts(<2 x i32> %x, <2 x i32> %other) {
-; CHECK-LABEL: @smin_commute_vec_undef_elts(
-; CHECK-NEXT:    [[TEST:%.*]] = icmp sgt <2 x i32> [[OTHER:%.*]], <i32 0, i32 undef>
+define <2 x i1> @smin_commute_vec_poison_elts(<2 x i32> %x, <2 x i32> %other) {
+; CHECK-LABEL: @smin_commute_vec_poison_elts(
+; CHECK-NEXT:    [[TEST:%.*]] = icmp sgt <2 x i32> [[OTHER:%.*]], <i32 0, i32 poison>
 ; CHECK-NEXT:    ret <2 x i1> [[TEST]]
 ;
   %notneg = and <2 x i32> %x, <i32 7, i32 7>
   %positive = or <2 x i32> %notneg, <i32 1, i32 1>
   %cmp = icmp slt <2 x i32> %other, %positive
   %sel = select <2 x i1> %cmp, <2 x i32> %other, <2 x i32> %positive
-  %test = icmp sgt <2 x i32> %sel, <i32 0, i32 undef>
+  %test = icmp sgt <2 x i32> %sel, <i32 0, i32 poison>
   ret <2 x i1> %test
 }
 ; %positive might be zero
diff --git a/llvm/test/Transforms/InstCombine/minmax-fold.ll b/llvm/test/Transforms/InstCombine/minmax-fold.ll
index bbbbf9eb6eafe..8b47dc7a28079 100644
--- a/llvm/test/Transforms/InstCombine/minmax-fold.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-fold.ll
@@ -1360,14 +1360,15 @@ define i8 @PR14613_smax(i8 %x) {
 
 define i8 @PR46271(<2 x i8> %x) {
 ; CHECK-LABEL: @PR46271(
-; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x i8> @llvm.smax.v2i8(<2 x i8> [[X:%.*]], <2 x i8> <i8 -1, i8 -1>)
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i8> [[X:%.*]], <i8 poison, i8 -1>
+; CHECK-NEXT:    [[A_INV:%.*]] = icmp slt <2 x i8> [[X]], zeroinitializer
+; CHECK-NEXT:    [[TMP1:%.*]] = select <2 x i1> [[A_INV]], <2 x i8> <i8 poison, i8 0>, <2 x i8> [[TMP3]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i8> [[TMP1]], i64 1
-; CHECK-NEXT:    [[R:%.*]] = xor i8 [[TMP2]], -1
-; CHECK-NEXT:    ret i8 [[R]]
+; CHECK-NEXT:    ret i8 [[TMP2]]
 ;
   %a = icmp sgt <2 x i8> %x, <i8 -1, i8 -1>
-  %b = select <2 x i1> %a, <2 x i8> %x, <2 x i8> <i8 undef, i8 -1>
-  %not = xor <2 x i8> %b, <i8 undef, i8 -1>
+  %b = select <2 x i1> %a, <2 x i8> %x, <2 x i8> <i8 poison, i8 -1>
+  %not = xor <2 x i8> %b, <i8 poison, i8 -1>
   %r = extractelement <2 x i8> %not, i32 1
   ret i8 %r
 }
diff --git a/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll b/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
index bd1a47bbfcc19..a76f0f84ba340 100644
--- a/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/minmax-intrinsics.ll
@@ -393,7 +393,7 @@ define i8 @smax_of_nots(i8 %x, i8 %y) {
   ret i8 %m
 }
 
-; Vectors are ok (including undef lanes of not ops)
+; Vectors are ok (including poison lanes of not ops)
 
 define <3 x i8> @smin_of_nots(<3 x i8> %x, <3 x i8> %y) {
 ; CHECK-LABEL: @smin_of_nots(
@@ -401,8 +401,8 @@ define <3 x i8> @smin_of_nots(<3 x i8> %x, <3 x i8> %y) {
 ; CHECK-NEXT:    [[M:%.*]] = xor <3 x i8> [[TMP1]], <i8 -1, i8 -1, i8 -1>
 ; CHECK-NEXT:    ret <3 x i8> [[M]]
 ;
-  %notx = xor <3 x i8> %x, <i8 -1, i8 undef, i8 -1>
-  %noty = xor <3 x i8> %y, <i8 -1, i8 -1, i8 undef>
+  %notx = xor <3 x i8> %x, <i8 -1, i8 poison, i8 -1>
+  %noty = xor <3 x i8> %y, <i8 -1, i8 -1, i8 poison>
   %m = call <3 x i8> @llvm.smin.v3i8(<3 x i8> %notx, <3 x i8> %noty)
   ret <3 x i8> %m
 }
@@ -473,16 +473,16 @@ define i8 @smax_of_not_and_const(i8 %x) {
   ret i8 %m
 }
 
-; Vectors are ok (including undef lanes of not ops and min/max constant operand)
+; Vectors are ok (including poison lanes of not ops and min/max constant operand)
 
 define <3 x i8> @smin_of_not_and_const(<3 x i8> %x) {
 ; CHECK-LABEL: @smin_of_not_and_const(
-; CHECK-NEXT:    [[TMP1:%.*]] = call <3 x i8> @llvm.smax.v3i8(<3 x i8> [[X:%.*]], <3 x i8> <i8 -43, i8 undef, i8 -44>)
+; CHECK-NEXT:    [[TMP1:%.*]] = call <3 x i8> @llvm.smax.v3i8(<3 x i8> [[X:%.*]], <3 x i8> <i8 -43, i8 poison, i8 -44>)
 ; CHECK-NEXT:    [[M:%.*]] = xor <3 x i8> [[TMP1]], <i8 -1, i8 -1, i8 -1>
 ; CHECK-NEXT:    ret <3 x i8> [[M]]
 ;
-  %notx = xor <3 x i8> %x, <i8 -1, i8 -1, i8 undef>
-  %m = call <3 x i8> @llvm.smin.v3i8(<3 x i8> <i8 42, i8 undef, i8 43>, <3 x i8> %notx)
+  %notx = xor <3 x i8> %x, <i8 -1, i8 -1, i8 poison>
+  %m = call <3 x i8> @llvm.smin.v3i8(<3 x i8> <i8 42, i8 poison, i8 43>, <3 x i8> %notx)
   ret <3 x i8> %m
 }
 
@@ -706,7 +706,7 @@ define <3 x i8> @smax_negation_vec(<3 x i8> %x) {
 ; CHECK-NEXT:    [[R:%.*]] = call <3 x i8> @llvm.abs.v3i8(<3 x i8> [[X:%.*]], i1 false)
 ; CHECK-NEXT:    ret <3 x i8> [[R]]
 ;
-  %s = sub <3 x i8> <i8 0, i8 undef, i8 0>, %x
+  %s = sub <3 x i8> <i8 0, i8 poison, i8 0>, %x
   %r = call <3 x i8> @llvm.smax.v3i8(<3 x i8> %x, <3 x i8> %s)
   ret <3 x i8> %r
 }
@@ -912,7 +912,7 @@ define <3 x i8> @umin_non_zero_idiom4(<3 x i8> %a) {
 ; CHECK-NEXT:    [[RES:%.*]] = zext <3 x i1> [[TMP1]] to <3 x i8>
 ; CHECK-NEXT:    ret <3 x i8> [[RES]]
 ;
-  %res = call <3 x i8> @llvm.umin.v3i8(<3 x i8> %a, <3 x i8> <i8 1, i8 undef, i8 undef>)
+  %res = call <3 x i8> @llvm.umin.v3i8(<3 x i8> %a, <3 x i8> <i8 1, i8 poison, i8 poison>)
   ret <3 x i8> %res
 }
 
@@ -2118,15 +2118,15 @@ define i8 @umin_offset_uses(i8 %x) {
   ret i8 %m
 }
 
-; TODO: This could transform, but undef element must not propagate to the new add.
+; TODO: This could transform
 
-define <3 x i8> @umax_vector_splat_undef(<3 x i8> %x) {
-; CHECK-LABEL: @umax_vector_splat_undef(
-; CHECK-NEXT:    [[A:%.*]] = add nuw <3 x i8> [[X:%.*]], <i8 undef, i8 64, i8 64>
+define <3 x i8> @umax_vector_splat_poison(<3 x i8> %x) {
+; CHECK-LABEL: @umax_vector_splat_poison(
+; CHECK-NEXT:    [[A:%.*]] = add nuw <3 x i8> [[X:%.*]], <i8 poison, i8 64, i8 64>
 ; CHECK-NEXT:    [[R:%.*]] = call <3 x i8> @llvm.umax.v3i8(<3 x i8> [[A]], <3 x i8> <i8 13, i8 -126, i8 -126>)
 ; CHECK-NEXT:    ret <3 x i8> [[R]]
 ;
-  %a = add nuw <3 x i8> %x, <i8 undef, i8 64, i8 64>
+  %a = add nuw <3 x i8> %x, <i8 poison, i8 64, i8 64>
   %r = call <3 x i8> @llvm.umax.v3i8(<3 x i8> %a, <3 x i8> <i8 13, i8 130, i8 130>)
   ret <3 x i8> %r
 }
@@ -2506,8 +2506,8 @@ entry:
   ret i8 %val
 }
 
-define <3 x i8> @fold_umax_with_knownbits_info_undef_in_splat(<3 x i8> %a, <3 x i8> %b) {
-; CHECK-LABEL: @fold_umax_with_knownbits_info_undef_in_splat(
+define <3 x i8> @fold_umax_with_knownbits_info_poison_in_splat(<3 x i8> %a, <3 x i8> %b) {
+; CHECK-LABEL: @fold_umax_with_knownbits_info_poison_in_splat(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[A1:%.*]] = or <3 x i8> [[A:%.*]], <i8 1, i8 1, i8 1>
 ; CHECK-NEXT:    [[A2:%.*]] = shl <3 x i8> [[B:%.*]], <i8 1, i8 1, i8 1>
@@ -2518,7 +2518,7 @@ entry:
   %a1 = or <3 x i8> %a, <i8 1, i8 1, i8 1>
   %a2 = shl <3 x i8> %b, <i8 1, i8 1, i8 1>
   %sub = sub <3 x i8> %a1, %a2
-  %val = call <3 x i8> @llvm.umax.v3i8(<3 x i8> %sub, <3 x i8> <i8 1, i8 undef, i8 1>)
+  %val = call <3 x i8> @llvm.umax.v3i8(<3 x i8> %sub, <3 x i8> <i8 1, i8 poison, i8 1>)
   ret <3 x i8> %val
 }
 
@@ -2535,8 +2535,8 @@ entry:
   ret i8 %val
 }
 
-define <3 x i8> @fold_umin_with_knownbits_info_undef_in_splat(<3 x i8> %a, <3 x i8> %b) {
-; CHECK-LABEL: @fold_umin_with_knownbits_info_undef_in_splat(
+define <3 x i8> @fold_umin_with_knownbits_info_poison_in_splat(<3 x i8> %a, <3 x i8> %b) {
+; CHECK-LABEL: @fold_umin_with_knownbits_info_poison_in_splat(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    ret <3 x i8> <i8 3, i8 3, i8 3>
 ;
@@ -2544,7 +2544,7 @@ entry:
   %a1 = or <3 x i8> %a, <i8 3, i8 3, i8 3>
   %a2 = shl <3 x i8> %b, <i8 2, i8 2, i8 2>
   %sub = sub <3 x i8> %a1, %a2
-  %val = call <3 x i8> @llvm.umin.v3i8(<3 x i8> %sub, <3 x i8> <i8 3, i8 undef, i8 3>)
+  %val = call <3 x i8> @llvm.umin.v3i8(<3 x i8> %sub, <3 x i8> <i8 3, i8 poison, i8 3>)
   ret <3 x i8> %val
 }
 
diff --git a/llvm/test/Transforms/InstCombine/mul-inseltpoison.ll b/llvm/test/Transforms/InstCombine/mul-inseltpoison.ll
index 8fe4261bbf009..f47c5577075cb 100644
--- a/llvm/test/Transforms/InstCombine/mul-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/mul-inseltpoison.ll
@@ -784,7 +784,7 @@ define <2 x i8> @negate_if_false_commute(<2 x i8> %px, <2 x i1> %cond) {
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %x = sdiv <2 x i8> <i8 42, i8 5>, %px  ; thwart complexity-based canonicalization
-  %sel = select <2 x i1> %cond, <2 x i8> <i8 1, i8 undef>, <2 x i8> <i8 -1, i8 -1>
+  %sel = select <2 x i1> %cond, <2 x i8> <i8 1, i8 poison>, <2 x i8> <i8 -1, i8 -1>
   %r = mul <2 x i8> %x, %sel
   ret <2 x i8> %r
 }
@@ -931,7 +931,7 @@ define <vscale x 2 x i64> @mul_scalable_splat_zero(<vscale x 2 x i64> %z) {
 ; CHECK-LABEL: @mul_scalable_splat_zero(
 ; CHECK-NEXT:    ret <vscale x 2 x i64> zeroinitializer
 ;
-  %shuf = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> undef, i64 0, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+  %shuf = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 0, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
   %t3 = mul <vscale x 2 x i64> %shuf, %z
   ret <vscale x 2 x i64> %t3
 }
@@ -973,14 +973,14 @@ define <2 x i32> @mulsub1_vec_nonuniform(<2 x i32> %a0, <2 x i32> %a1) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @mulsub1_vec_nonuniform_undef(<2 x i32> %a0, <2 x i32> %a1) {
-; CHECK-LABEL: @mulsub1_vec_nonuniform_undef(
+define <2 x i32> @mulsub1_vec_nonuniform_poison(<2 x i32> %a0, <2 x i32> %a1) {
+; CHECK-LABEL: @mulsub1_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[SUB_NEG:%.*]] = sub <2 x i32> [[A0:%.*]], [[A1:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[SUB_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %sub = sub <2 x i32> %a1, %a0
-  %mul = mul <2 x i32> %sub, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %sub, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
@@ -1017,14 +1017,14 @@ define <2 x i32> @mulsub2_vec_nonuniform(<2 x i32> %a0) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @mulsub2_vec_nonuniform_undef(<2 x i32> %a0) {
-; CHECK-LABEL: @mulsub2_vec_nonuniform_undef(
+define <2 x i32> @mulsub2_vec_nonuniform_poison(<2 x i32> %a0) {
+; CHECK-LABEL: @mulsub2_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[SUB_NEG:%.*]] = add <2 x i32> [[A0:%.*]], <i32 -16, i32 -32>
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[SUB_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %sub = sub <2 x i32> <i32 16, i32 32>, %a0
-  %mul = mul <2 x i32> %sub, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %sub, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
@@ -1061,14 +1061,14 @@ define <2 x i32> @muladd2_vec_nonuniform(<2 x i32> %a0) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @muladd2_vec_nonuniform_undef(<2 x i32> %a0) {
-; CHECK-LABEL: @muladd2_vec_nonuniform_undef(
+define <2 x i32> @muladd2_vec_nonuniform_poison(<2 x i32> %a0) {
+; CHECK-LABEL: @muladd2_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[ADD_NEG:%.*]] = sub <2 x i32> <i32 -16, i32 -32>, [[A0:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[ADD_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %add = add <2 x i32> %a0, <i32 16, i32 32>
-  %mul = mul <2 x i32> %add, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %add, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
diff --git a/llvm/test/Transforms/InstCombine/mul.ll b/llvm/test/Transforms/InstCombine/mul.ll
index d4a689c60786e..227ca4a6d5cfa 100644
--- a/llvm/test/Transforms/InstCombine/mul.ll
+++ b/llvm/test/Transforms/InstCombine/mul.ll
@@ -1496,7 +1496,7 @@ define <2 x i8> @negate_if_false_commute(<2 x i8> %px, <2 x i1> %cond) {
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %x = sdiv <2 x i8> <i8 42, i8 5>, %px  ; thwart complexity-based canonicalization
-  %sel = select <2 x i1> %cond, <2 x i8> <i8 1, i8 undef>, <2 x i8> <i8 -1, i8 -1>
+  %sel = select <2 x i1> %cond, <2 x i8> <i8 1, i8 poison>, <2 x i8> <i8 -1, i8 -1>
   %r = mul <2 x i8> %x, %sel
   ret <2 x i8> %r
 }
@@ -1643,7 +1643,7 @@ define <vscale x 2 x i64> @mul_scalable_splat_zero(<vscale x 2 x i64> %z) {
 ; CHECK-LABEL: @mul_scalable_splat_zero(
 ; CHECK-NEXT:    ret <vscale x 2 x i64> zeroinitializer
 ;
-  %shuf = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> undef, i64 0, i32 0), <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %shuf = shufflevector <vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 0, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
   %t3 = mul <vscale x 2 x i64> %shuf, %z
   ret <vscale x 2 x i64> %t3
 }
@@ -1752,14 +1752,14 @@ define <2 x i32> @mulsub1_vec_nonuniform(<2 x i32> %a0, <2 x i32> %a1) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @mulsub1_vec_nonuniform_undef(<2 x i32> %a0, <2 x i32> %a1) {
-; CHECK-LABEL: @mulsub1_vec_nonuniform_undef(
+define <2 x i32> @mulsub1_vec_nonuniform_poison(<2 x i32> %a0, <2 x i32> %a1) {
+; CHECK-LABEL: @mulsub1_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[SUB_NEG:%.*]] = sub <2 x i32> [[A0:%.*]], [[A1:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[SUB_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %sub = sub <2 x i32> %a1, %a0
-  %mul = mul <2 x i32> %sub, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %sub, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
@@ -1796,14 +1796,14 @@ define <2 x i32> @mulsub2_vec_nonuniform(<2 x i32> %a0) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @mulsub2_vec_nonuniform_undef(<2 x i32> %a0) {
-; CHECK-LABEL: @mulsub2_vec_nonuniform_undef(
+define <2 x i32> @mulsub2_vec_nonuniform_poison(<2 x i32> %a0) {
+; CHECK-LABEL: @mulsub2_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[SUB_NEG:%.*]] = add <2 x i32> [[A0:%.*]], <i32 -16, i32 -32>
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[SUB_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %sub = sub <2 x i32> <i32 16, i32 32>, %a0
-  %mul = mul <2 x i32> %sub, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %sub, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
@@ -1819,15 +1819,15 @@ define i8 @mulsub_nsw(i8 %a1, i8 %a2) {
 }
 
 ; It would be safe to keep the nsw on the shl here, but only because the mul
-; to shl transform happens to replace undef with 0.
-define <2 x i8> @mulsub_nsw_undef(<2 x i8> %a1, <2 x i8> %a2) {
-; CHECK-LABEL: @mulsub_nsw_undef(
+; to shl transform happens to replace poison with 0.
+define <2 x i8> @mulsub_nsw_poison(<2 x i8> %a1, <2 x i8> %a2) {
+; CHECK-LABEL: @mulsub_nsw_poison(
 ; CHECK-NEXT:    [[A_NEG:%.*]] = sub nsw <2 x i8> [[A2:%.*]], [[A1:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i8> [[A_NEG]], <i8 1, i8 0>
 ; CHECK-NEXT:    ret <2 x i8> [[MUL]]
 ;
   %a = sub nsw <2 x i8> %a1, %a2
-  %mul = mul nsw <2 x i8> %a, <i8 -2, i8 undef>
+  %mul = mul nsw <2 x i8> %a, <i8 -2, i8 poison>
   ret <2 x i8> %mul
 }
 
@@ -1864,14 +1864,14 @@ define <2 x i32> @muladd2_vec_nonuniform(<2 x i32> %a0) {
   ret <2 x i32> %mul
 }
 
-define <2 x i32> @muladd2_vec_nonuniform_undef(<2 x i32> %a0) {
-; CHECK-LABEL: @muladd2_vec_nonuniform_undef(
+define <2 x i32> @muladd2_vec_nonuniform_poison(<2 x i32> %a0) {
+; CHECK-LABEL: @muladd2_vec_nonuniform_poison(
 ; CHECK-NEXT:    [[ADD_NEG:%.*]] = sub <2 x i32> <i32 -16, i32 -32>, [[A0:%.*]]
 ; CHECK-NEXT:    [[MUL:%.*]] = shl <2 x i32> [[ADD_NEG]], <i32 2, i32 0>
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %add = add <2 x i32> %a0, <i32 16, i32 32>
-  %mul = mul <2 x i32> %add, <i32 -4, i32 undef>
+  %mul = mul <2 x i32> %add, <i32 -4, i32 poison>
   ret <2 x i32> %mul
 }
 
diff --git a/llvm/test/Transforms/InstCombine/not-add.ll b/llvm/test/Transforms/InstCombine/not-add.ll
index 877f558ffd503..9ba37b6bba39e 100644
--- a/llvm/test/Transforms/InstCombine/not-add.ll
+++ b/llvm/test/Transforms/InstCombine/not-add.ll
@@ -115,26 +115,26 @@ define <4 x i32> @vector_test(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i32> %nota
 }
 
-define <4 x i32> @vector_test_undef(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vector_test_undef(
+define <4 x i32> @vector_test_poison(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vector_test_poison(
 ; CHECK-NEXT:    [[NOTA:%.*]] = sub <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[NOTA]]
 ;
-  %notx = xor <4 x i32> %x, <i32 -1, i32 undef, i32 undef, i32 -1>
+  %notx = xor <4 x i32> %x, <i32 -1, i32 poison, i32 poison, i32 -1>
   %a = add <4 x i32> %notx, %y
-  %nota = xor <4 x i32> %a, <i32 -1, i32 -1, i32 undef, i32 undef>
+  %nota = xor <4 x i32> %a, <i32 -1, i32 -1, i32 poison, i32 poison>
   ret <4 x i32> %nota
 }
 
 
-define <4 x i32> @vector_test_undef_nsw_nuw(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vector_test_undef_nsw_nuw(
+define <4 x i32> @vector_test_poison_nsw_nuw(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vector_test_poison_nsw_nuw(
 ; CHECK-NEXT:    [[NOTA:%.*]] = sub nuw nsw <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[NOTA]]
 ;
-  %notx = xor <4 x i32> %x, <i32 -1, i32 undef, i32 undef, i32 -1>
+  %notx = xor <4 x i32> %x, <i32 -1, i32 poison, i32 poison, i32 -1>
   %a = add nsw nuw <4 x i32> %notx, %y
-  %nota = xor <4 x i32> %a, <i32 -1, i32 -1, i32 undef, i32 undef>
+  %nota = xor <4 x i32> %a, <i32 -1, i32 -1, i32 poison, i32 poison>
   ret <4 x i32> %nota
 }
 
diff --git a/llvm/test/Transforms/InstCombine/not.ll b/llvm/test/Transforms/InstCombine/not.ll
index 98b5d98041560..0c2c6195e3240 100644
--- a/llvm/test/Transforms/InstCombine/not.ll
+++ b/llvm/test/Transforms/InstCombine/not.ll
@@ -430,9 +430,9 @@ define <3 x i5> @not_or_neg_commute_vec(<3 x i5> %x, <3 x i5> %p)  {
 ; CHECK-NEXT:    ret <3 x i5> [[NOT]]
 ;
   %y = mul <3 x i5> %p, <i5 1, i5 2, i5 3> ; thwart complexity-based-canonicalization
-  %s = sub <3 x i5> <i5 0, i5 0, i5 undef>, %x
+  %s = sub <3 x i5> <i5 0, i5 0, i5 poison>, %x
   %o = or <3 x i5> %y, %s
-  %not = xor <3 x i5> %o, <i5 -1, i5 undef, i5 -1>
+  %not = xor <3 x i5> %o, <i5 -1, i5 poison, i5 -1>
   ret <3 x i5> %not
 }
 
diff --git a/llvm/test/Transforms/InstCombine/omit-urem-of-power-of-two-or-zero-when-comparing-with-zero.ll b/llvm/test/Transforms/InstCombine/omit-urem-of-power-of-two-or-zero-when-comparing-with-zero.ll
index c16633efe4ce3..3fd4a17d972af 100644
--- a/llvm/test/Transforms/InstCombine/omit-urem-of-power-of-two-or-zero-when-comparing-with-zero.ll
+++ b/llvm/test/Transforms/InstCombine/omit-urem-of-power-of-two-or-zero-when-comparing-with-zero.ll
@@ -95,41 +95,41 @@ define <4 x i1> @p5_vector_urem_by_const__nonsplat(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i1> %t2
 }
 
-define <4 x i1> @p6_vector_urem_by_const__nonsplat_undef0(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @p6_vector_urem_by_const__nonsplat_undef0(
-; CHECK-NEXT:    [[T0:%.*]] = and <4 x i32> [[X:%.*]], <i32 128, i32 128, i32 undef, i32 128>
-; CHECK-NEXT:    [[T1:%.*]] = urem <4 x i32> [[T0]], <i32 6, i32 6, i32 6, i32 6>
-; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T1]], zeroinitializer
+; The poison value in the vector makes the whole function UB.
+
+define <4 x i1> @p6_vector_urem_by_const__nonsplat_poison0(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @p6_vector_urem_by_const__nonsplat_poison0(
+; CHECK-NEXT:    [[T0:%.*]] = and <4 x i32> [[X:%.*]], <i32 128, i32 128, i32 poison, i32 128>
+; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T0]], zeroinitializer
 ; CHECK-NEXT:    ret <4 x i1> [[T2]]
 ;
-  %t0 = and <4 x i32> %x, <i32 128, i32 128, i32 undef, i32 128>
+  %t0 = and <4 x i32> %x, <i32 128, i32 128, i32 poison, i32 128>
   %t1 = urem <4 x i32> %t0, <i32 6, i32 6, i32 6, i32 6> ; '6' is clearly not a power of two
   %t2 = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %t2
 }
 
-define <4 x i1> @p7_vector_urem_by_const__nonsplat_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @p7_vector_urem_by_const__nonsplat_undef2(
+define <4 x i1> @p7_vector_urem_by_const__nonsplat_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @p7_vector_urem_by_const__nonsplat_poison2(
 ; CHECK-NEXT:    [[T0:%.*]] = and <4 x i32> [[X:%.*]], <i32 128, i32 128, i32 128, i32 128>
-; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T0]], <i32 0, i32 0, i32 undef, i32 0>
+; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T0]], <i32 0, i32 0, i32 poison, i32 0>
 ; CHECK-NEXT:    ret <4 x i1> [[T2]]
 ;
   %t0 = and <4 x i32> %x, <i32 128, i32 128, i32 128, i32 128> ; clearly a power-of-two or zero
   %t1 = urem <4 x i32> %t0, <i32 6, i32 6, i32 6, i32 6> ; '6' is clearly not a power of two
-  %t2 = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
+  %t2 = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 poison, i32 0>
   ret <4 x i1> %t2
 }
 
-define <4 x i1> @p8_vector_urem_by_const__nonsplat_undef3(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @p8_vector_urem_by_const__nonsplat_undef3(
-; CHECK-NEXT:    [[T0:%.*]] = and <4 x i32> [[X:%.*]], <i32 128, i32 128, i32 undef, i32 128>
-; CHECK-NEXT:    [[T1:%.*]] = urem <4 x i32> [[T0]], <i32 6, i32 6, i32 6, i32 6>
-; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T1]], <i32 0, i32 0, i32 undef, i32 0>
+define <4 x i1> @p8_vector_urem_by_const__nonsplat_poison3(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @p8_vector_urem_by_const__nonsplat_poison3(
+; CHECK-NEXT:    [[T0:%.*]] = and <4 x i32> [[X:%.*]], <i32 128, i32 128, i32 poison, i32 128>
+; CHECK-NEXT:    [[T2:%.*]] = icmp eq <4 x i32> [[T0]], <i32 0, i32 0, i32 poison, i32 0>
 ; CHECK-NEXT:    ret <4 x i1> [[T2]]
 ;
-  %t0 = and <4 x i32> %x, <i32 128, i32 128, i32 undef, i32 128>
+  %t0 = and <4 x i32> %x, <i32 128, i32 128, i32 poison, i32 128>
   %t1 = urem <4 x i32> %t0, <i32 6, i32 6, i32 6, i32 6> ; '6' is clearly not a power of two
-  %t2 = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
+  %t2 = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 poison, i32 0>
   ret <4 x i1> %t2
 }
 
diff --git a/llvm/test/Transforms/InstCombine/operand-complexity.ll b/llvm/test/Transforms/InstCombine/operand-complexity.ll
index 62cfc76d9d24e..541a15275b617 100644
--- a/llvm/test/Transforms/InstCombine/operand-complexity.ll
+++ b/llvm/test/Transforms/InstCombine/operand-complexity.ll
@@ -29,15 +29,15 @@ define <2 x i8> @neg_vec(<2 x i8> %x) {
   ret <2 x i8> %r
 }
 
-define <2 x i8> @neg_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @neg_vec_undef(
+define <2 x i8> @neg_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @neg_vec_poison(
 ; CHECK-NEXT:    [[BO:%.*]] = udiv <2 x i8> [[X:%.*]], <i8 42, i8 -42>
-; CHECK-NEXT:    [[NEGX:%.*]] = sub <2 x i8> <i8 0, i8 undef>, [[X]]
+; CHECK-NEXT:    [[NEGX:%.*]] = sub <2 x i8> <i8 0, i8 poison>, [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = xor <2 x i8> [[BO]], [[NEGX]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %bo = udiv <2 x i8> %x, <i8 42, i8 -42>
-  %negx = sub <2 x i8> <i8 0, i8 undef>, %x
+  %negx = sub <2 x i8> <i8 0, i8 poison>, %x
   %r = xor <2 x i8> %negx, %bo
   ret <2 x i8> %r
 }
@@ -70,15 +70,15 @@ define <2 x i8> @not_vec(<2 x i8> %x) {
   ret <2 x i8> %r
 }
 
-define <2 x i8> @not_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @not_vec_undef(
+define <2 x i8> @not_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @not_vec_poison(
 ; CHECK-NEXT:    [[BO:%.*]] = udiv <2 x i8> [[X:%.*]], <i8 42, i8 -42>
-; CHECK-NEXT:    [[NOTX:%.*]] = xor <2 x i8> [[X]], <i8 -1, i8 undef>
+; CHECK-NEXT:    [[NOTX:%.*]] = xor <2 x i8> [[X]], <i8 -1, i8 poison>
 ; CHECK-NEXT:    [[R:%.*]] = mul <2 x i8> [[BO]], [[NOTX]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %bo = udiv <2 x i8> %x, <i8 42, i8 -42>
-  %notx = xor <2 x i8> <i8 -1, i8 undef>, %x
+  %notx = xor <2 x i8> <i8 -1, i8 poison>, %x
   %r = mul <2 x i8> %notx, %bo
   ret <2 x i8> %r
 }
@@ -134,8 +134,8 @@ define <2 x float> @fneg_vec(<2 x float> %x) {
   ret <2 x float> %r
 }
 
-define <2 x float> @fneg_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fneg_vec_undef(
+define <2 x float> @fneg_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fneg_vec_poison(
 ; CHECK-NEXT:    [[BO:%.*]] = fdiv <2 x float> [[X:%.*]], <float 4.200000e+01, float -4.200000e+01>
 ; CHECK-NEXT:    [[FNEGX:%.*]] = fneg <2 x float> [[X]]
 ; CHECK-NEXT:    [[R:%.*]] = fmul <2 x float> [[BO]], [[FNEGX]]
@@ -143,7 +143,7 @@ define <2 x float> @fneg_vec_undef(<2 x float> %x) {
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %bo = fdiv <2 x float> %x, <float 42.0, float -42.0>
-  %fnegx = fsub <2 x float> <float -0.0, float undef>, %x
+  %fnegx = fsub <2 x float> <float -0.0, float poison>, %x
   %r = fmul <2 x float> %fnegx, %bo
   call void @use_vec(<2 x float> %fnegx)
   ret <2 x float> %r
diff --git a/llvm/test/Transforms/InstCombine/or.ll b/llvm/test/Transforms/InstCombine/or.ll
index 1b1a6ffbf0f2d..6e2085a8bb6c7 100644
--- a/llvm/test/Transforms/InstCombine/or.ll
+++ b/llvm/test/Transforms/InstCombine/or.ll
@@ -262,26 +262,26 @@ define <2 x i1> @and_icmp_eq_0_vector(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i1> %D
 }
 
-define <2 x i1> @and_icmp_eq_0_vector_undef1(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @and_icmp_eq_0_vector_undef1(
+define <2 x i1> @and_icmp_eq_0_vector_poison1(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @and_icmp_eq_0_vector_poison1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <2 x i32> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[D:%.*]] = icmp eq <2 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[D]]
 ;
-  %C1 = icmp eq <2 x i32> %A, <i32 0, i32 undef>
-  %C2 = icmp eq <2 x i32> %B, <i32 0, i32 undef>
+  %C1 = icmp eq <2 x i32> %A, <i32 0, i32 poison>
+  %C2 = icmp eq <2 x i32> %B, <i32 0, i32 poison>
   %D = and <2 x i1> %C1, %C2
   ret <2 x i1> %D
 }
 
-define <2 x i1> @and_icmp_eq_0_vector_undef2(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @and_icmp_eq_0_vector_undef2(
+define <2 x i1> @and_icmp_eq_0_vector_poison2(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @and_icmp_eq_0_vector_poison2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <2 x i32> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[D:%.*]] = icmp eq <2 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[D]]
 ;
-  %C1 = icmp eq <2 x i32> %A, <i32 0, i32 undef>
-  %C2 = icmp eq <2 x i32> %B, <i32 undef, i32 0>
+  %C1 = icmp eq <2 x i32> %A, <i32 0, i32 poison>
+  %C2 = icmp eq <2 x i32> %B, <i32 poison, i32 0>
   %D = and <2 x i1> %C1, %C2
   ret <2 x i1> %D
 }
@@ -566,17 +566,17 @@ define <2 x i1> @test37_uniform(<2 x i32> %x) {
   ret <2 x i1> %ret1
 }
 
-define <2 x i1> @test37_undef(<2 x i32> %x) {
-; CHECK-LABEL: @test37_undef(
-; CHECK-NEXT:    [[ADD1:%.*]] = add <2 x i32> [[X:%.*]], <i32 7, i32 undef>
-; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult <2 x i32> [[ADD1]], <i32 30, i32 undef>
-; CHECK-NEXT:    [[CMP2:%.*]] = icmp eq <2 x i32> [[X]], <i32 23, i32 undef>
+define <2 x i1> @test37_poison(<2 x i32> %x) {
+; CHECK-LABEL: @test37_poison(
+; CHECK-NEXT:    [[ADD1:%.*]] = add <2 x i32> [[X:%.*]], <i32 7, i32 poison>
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult <2 x i32> [[ADD1]], <i32 30, i32 poison>
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp eq <2 x i32> [[X]], <i32 23, i32 poison>
 ; CHECK-NEXT:    [[RET1:%.*]] = or <2 x i1> [[CMP1]], [[CMP2]]
 ; CHECK-NEXT:    ret <2 x i1> [[RET1]]
 ;
-  %add1 = add <2 x i32> %x, <i32 7, i32 undef>
-  %cmp1 = icmp ult <2 x i32> %add1, <i32 30, i32 undef>
-  %cmp2 = icmp eq <2 x i32> %x, <i32 23, i32 undef>
+  %add1 = add <2 x i32> %x, <i32 7, i32 poison>
+  %cmp1 = icmp ult <2 x i32> %add1, <i32 30, i32 poison>
+  %cmp2 = icmp eq <2 x i32> %x, <i32 23, i32 poison>
   %ret1 = or <2 x i1> %cmp1, %cmp2
   ret <2 x i1> %ret1
 }
@@ -874,19 +874,19 @@ define <2 x i1> @test46_uniform(<2 x i8> %c)  {
   ret <2 x i1> %or
 }
 
-define <2 x i1> @test46_undef(<2 x i8> %c)  {
-; CHECK-LABEL: @test46_undef(
-; CHECK-NEXT:    [[C_OFF:%.*]] = add <2 x i8> [[C:%.*]], <i8 -97, i8 undef>
-; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult <2 x i8> [[C_OFF]], <i8 26, i8 undef>
-; CHECK-NEXT:    [[C_OFF17:%.*]] = add <2 x i8> [[C]], <i8 -65, i8 undef>
-; CHECK-NEXT:    [[CMP2:%.*]] = icmp ult <2 x i8> [[C_OFF17]], <i8 26, i8 undef>
+define <2 x i1> @test46_poison(<2 x i8> %c)  {
+; CHECK-LABEL: @test46_poison(
+; CHECK-NEXT:    [[C_OFF:%.*]] = add <2 x i8> [[C:%.*]], <i8 -97, i8 poison>
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult <2 x i8> [[C_OFF]], <i8 26, i8 poison>
+; CHECK-NEXT:    [[C_OFF17:%.*]] = add <2 x i8> [[C]], <i8 -65, i8 poison>
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp ult <2 x i8> [[C_OFF17]], <i8 26, i8 poison>
 ; CHECK-NEXT:    [[OR:%.*]] = or <2 x i1> [[CMP1]], [[CMP2]]
 ; CHECK-NEXT:    ret <2 x i1> [[OR]]
 ;
-  %c.off = add <2 x i8> %c, <i8 -97, i8 undef>
-  %cmp1 = icmp ult <2 x i8> %c.off, <i8 26, i8 undef>
-  %c.off17 = add <2 x i8> %c, <i8 -65, i8 undef>
-  %cmp2 = icmp ult <2 x i8> %c.off17, <i8 26, i8 undef>
+  %c.off = add <2 x i8> %c, <i8 -97, i8 poison>
+  %cmp1 = icmp ult <2 x i8> %c.off, <i8 26, i8 poison>
+  %c.off17 = add <2 x i8> %c, <i8 -65, i8 poison>
+  %cmp2 = icmp ult <2 x i8> %c.off17, <i8 26, i8 poison>
   %or = or <2 x i1> %cmp1, %cmp2
   ret <2 x i1> %or
 }
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-b.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-b.ll
index f0c2f129e3df3..5ed7d641df65b 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-b.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-b.ll
@@ -89,13 +89,13 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t7
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
-; CHECK-NEXT:    [[T0:%.*]] = add <8 x i32> [[NBITS:%.*]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
+; CHECK-NEXT:    [[T0:%.*]] = add <8 x i32> [[NBITS:%.*]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
 ; CHECK-NEXT:    [[T1:%.*]] = zext <8 x i32> [[T0]] to <8 x i64>
-; CHECK-NEXT:    [[T2:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T1]]
-; CHECK-NEXT:    [[T3:%.*]] = xor <8 x i64> [[T2]], <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>
-; CHECK-NEXT:    [[T4:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, [[NBITS]]
+; CHECK-NEXT:    [[T2:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T1]]
+; CHECK-NEXT:    [[T3:%.*]] = xor <8 x i64> [[T2]], <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>
+; CHECK-NEXT:    [[T4:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, [[NBITS]]
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T2]])
@@ -106,11 +106,11 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    [[T7:%.*]] = and <8 x i32> [[TMP2]], <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 poison, i32 2147483647>
 ; CHECK-NEXT:    ret <8 x i32> [[T7]]
 ;
-  %t0 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+  %t0 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
   %t1 = zext <8 x i32> %t0 to <8 x i64>
-  %t2 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t1 ; shifting by nbits-1
-  %t3 = xor <8 x i64> %t2, <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>
-  %t4 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, %nbits
+  %t2 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t1 ; shifting by nbits-1
+  %t3 = xor <8 x i64> %t2, <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>
+  %t4 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, %nbits
 
   call void @use8xi32(<8 x i32> %t0)
   call void @use8xi64(<8 x i64> %t1)
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-c.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-c.ll
index 46d1de5781b71..1a711e58c333b 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-c.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-c.ll
@@ -73,11 +73,11 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t5
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 undef, i32 -33>
+; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 poison, i32 -33>
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T2]])
@@ -87,8 +87,8 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T5]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
-  %t2 = add <8 x i32> %nbits, <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 undef, i32 -33>
+  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
+  %t2 = add <8 x i32> %nbits, <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 poison, i32 -33>
 
   call void @use8xi64(<8 x i64> %t0)
   call void @use8xi64(<8 x i64> %t1)
@@ -103,8 +103,8 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-LABEL: @t3_vec_nonsplat(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 undef, i32 65>
+; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 poison, i32 65>
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T2]])
@@ -114,8 +114,8 @@ define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T5]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
-  %t2 = add <8 x i32> %nbits, <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 undef, i32 65>
+  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
+  %t2 = add <8 x i32> %nbits, <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 poison, i32 65>
 
   call void @use8xi64(<8 x i64> %t0)
   call void @use8xi64(<8 x i64> %t1)
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-d.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-d.ll
index 48873852cfc7c..cd0098ecdb0a6 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-d.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-after-truncation-variant-d.ll
@@ -81,12 +81,12 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t6
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
 ; CHECK-NEXT:    [[T2:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 undef, i32 -33>
+; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 poison, i32 -33>
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T2]])
@@ -97,9 +97,9 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T6]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
+  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
   %t2 = lshr <8 x i64> %t1, %t0
-  %t3 = add <8 x i32> %nbits, <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 undef, i32 -33>
+  %t3 = add <8 x i32> %nbits, <i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 -33, i32 poison, i32 -33>
 
   call void @use8xi64(<8 x i64> %t0)
   call void @use8xi64(<8 x i64> %t1)
@@ -115,9 +115,9 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-LABEL: @t3_vec_nonsplat(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
 ; CHECK-NEXT:    [[T2:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 undef, i32 65>
+; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 poison, i32 65>
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T2]])
@@ -128,9 +128,9 @@ define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T6]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
+  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
   %t2 = lshr <8 x i64> %t1, %t0
-  %t3 = add <8 x i32> %nbits, <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 undef, i32 65>
+  %t3 = add <8 x i32> %nbits, <i32 -64, i32 -63, i32 -33, i32 -32, i32 63, i32 64, i32 poison, i32 65>
 
   call void @use8xi64(<8 x i64> %t0)
   call void @use8xi64(<8 x i64> %t1)
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-b.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-b.ll
index 8b3f01bcb7691..1debf111b18cd 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-b.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-b.ll
@@ -71,12 +71,12 @@ define <8 x i32> @t1_vec_splat(<8 x i32> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t5
 }
 
-define <8 x i32> @t1_vec_splat_undef(<8 x i32> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t1_vec_splat_undef(
-; CHECK-NEXT:    [[T0:%.*]] = add <8 x i32> [[NBITS:%.*]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = xor <8 x i32> [[T1]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
-; CHECK-NEXT:    [[T4:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, [[NBITS]]
+define <8 x i32> @t1_vec_splat_poison(<8 x i32> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t1_vec_splat_poison(
+; CHECK-NEXT:    [[T0:%.*]] = add <8 x i32> [[NBITS:%.*]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = xor <8 x i32> [[T1]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[T4:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, [[NBITS]]
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T0]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T1]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T2]])
@@ -85,11 +85,11 @@ define <8 x i32> @t1_vec_splat_undef(<8 x i32> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    [[T5:%.*]] = and <8 x i32> [[TMP1]], <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 poison, i32 2147483647>
 ; CHECK-NEXT:    ret <8 x i32> [[T5]]
 ;
-  %t0 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
-  %t1 = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, %t0
-  %t2 = xor <8 x i32> %t1, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+  %t0 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
+  %t1 = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, %t0
+  %t2 = xor <8 x i32> %t1, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
   %t3 = and <8 x i32> %t2, %x
-  %t4 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, %nbits
+  %t4 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, %nbits
   call void @use8xi32(<8 x i32> %t0)
   call void @use8xi32(<8 x i32> %t1)
   call void @use8xi32(<8 x i32> %t2)
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-c.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-c.ll
index 58a905063fac4..55d0b3f80a519 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-c.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-c.ll
@@ -55,19 +55,19 @@ define <8 x i32> @t1_vec_splat(<8 x i32> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t3
 }
 
-define <8 x i32> @t1_vec_splat_undef(<8 x i32> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t1_vec_splat_undef(
-; CHECK-NEXT:    [[T0:%.*]] = lshr <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, [[NBITS:%.*]]
-; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+define <8 x i32> @t1_vec_splat_poison(<8 x i32> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t1_vec_splat_poison(
+; CHECK-NEXT:    [[T0:%.*]] = lshr <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, [[NBITS:%.*]]
+; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T0]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T2]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl <8 x i32> [[X:%.*]], [[T2]]
 ; CHECK-NEXT:    [[T3:%.*]] = and <8 x i32> [[TMP1]], <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 poison, i32 2147483647>
 ; CHECK-NEXT:    ret <8 x i32> [[T3]]
 ;
-  %t0 = lshr <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, %nbits
+  %t0 = lshr <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, %nbits
   %t1 = and <8 x i32> %t0, %x
-  %t2 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+  %t2 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
   call void @use8xi32(<8 x i32> %t0)
   call void @use8xi32(<8 x i32> %t2)
   %t3 = shl <8 x i32> %t1, %t2 ; shift is smaller than mask
diff --git a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-d.ll b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-d.ll
index 9c096d1418a5b..7ad99a6bb0a38 100644
--- a/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-d.ll
+++ b/llvm/test/Transforms/InstCombine/partally-redundant-left-shift-input-masking-variant-d.ll
@@ -63,11 +63,11 @@ define <8 x i32> @t2_vec_splat(<8 x i32> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t4
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i32> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
-; CHECK-NEXT:    [[T0:%.*]] = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, [[NBITS:%.*]]
+define <8 x i32> @t2_vec_splat_poison(<8 x i32> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
+; CHECK-NEXT:    [[T0:%.*]] = shl nsw <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, [[NBITS]]
-; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T0]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T1]])
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[T3]])
@@ -75,10 +75,10 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i32> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    [[T4:%.*]] = and <8 x i32> [[TMP1]], <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647, i32 poison, i32 2147483647>
 ; CHECK-NEXT:    ret <8 x i32> [[T4]]
 ;
-  %t0 = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>, %nbits
+  %t0 = shl <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>, %nbits
   %t1 = lshr <8 x i32> %t0, %nbits
   %t2 = and <8 x i32> %t1, %x
-  %t3 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 undef, i32 -1>
+  %t3 = add <8 x i32> %nbits, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 poison, i32 -1>
   call void @use8xi32(<8 x i32> %t0)
   call void @use8xi32(<8 x i32> %t1)
   call void @use8xi32(<8 x i32> %t3)
diff --git a/llvm/test/Transforms/InstCombine/pr53357.ll b/llvm/test/Transforms/InstCombine/pr53357.ll
index 0a6d2993ce46a..0ae690869c1c4 100644
--- a/llvm/test/Transforms/InstCombine/pr53357.ll
+++ b/llvm/test/Transforms/InstCombine/pr53357.ll
@@ -30,16 +30,16 @@ define <2 x i32> @src_vec(<2 x i32> noundef %0, <2 x i32> noundef %1) {
   ret <2 x i32> %6
 }
 
-; vector version of src with undef values
-define <2 x i32> @src_vec_undef(<2 x i32> noundef %0, <2 x i32> noundef %1) {
-; CHECK-LABEL: @src_vec_undef(
+; vector version of src with poison values
+define <2 x i32> @src_vec_poison(<2 x i32> noundef %0, <2 x i32> noundef %1) {
+; CHECK-LABEL: @src_vec_poison(
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i32> [[TMP1:%.*]], [[TMP0:%.*]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i32> [[TMP3]], <i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <2 x i32> [[TMP4]]
 ;
   %3 = and <2 x i32> %1, %0
   %4 = or  <2 x i32> %1, %0
-  %5 = xor <2 x i32> %4, <i32 -1, i32 undef>
+  %5 = xor <2 x i32> %4, <i32 -1, i32 poison>
   %6 = add <2 x i32> %3, %5
   ret <2 x i32> %6
 }
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-b.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-b.ll
index d49cfe990d82d..cb6775e689b8c 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-b.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-b.ll
@@ -89,12 +89,12 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t6
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = xor <8 x i64> [[T1]], <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>
-; CHECK-NEXT:    [[T3:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, [[NBITS]]
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = xor <8 x i64> [[T1]], <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>
+; CHECK-NEXT:    [[T3:%.*]] = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, [[NBITS]]
 ; CHECK-NEXT:    [[T4:%.*]] = and <8 x i64> [[T2]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use8xi32(<8 x i32> [[NBITS]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
@@ -107,9 +107,9 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T6]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
-  %t2 = xor <8 x i64> %t1, <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>
-  %t3 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 undef, i32 32>, %nbits
+  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
+  %t2 = xor <8 x i64> %t1, <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>
+  %t3 = sub <8 x i32> <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 poison, i32 32>, %nbits
   %t4 = and <8 x i64> %t2, %x
 
   call void @use8xi32(<8 x i32> %nbits)
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-c.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-c.ll
index fbbeffbba630b..a78246781c7f9 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-c.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-c.ll
@@ -77,11 +77,11 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t5
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 undef, i32 -32>
+; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 poison, i32 -32>
 ; CHECK-NEXT:    [[T3:%.*]] = and <8 x i64> [[T1]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
@@ -92,8 +92,8 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T5]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
-  %t2 = add <8 x i32> %nbits, <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 undef, i32 -32>
+  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
+  %t2 = add <8 x i32> %nbits, <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 poison, i32 -32>
   %t3 = and <8 x i64> %t1, %x
 
   call void @use8xi64(<8 x i64> %t0)
@@ -109,8 +109,8 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-LABEL: @t3_vec_nonsplat(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 undef, i32 64>
+; CHECK-NEXT:    [[T1:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T2:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 poison, i32 64>
 ; CHECK-NEXT:    [[T3:%.*]] = and <8 x i64> [[T1]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
@@ -121,8 +121,8 @@ define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T5]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
-  %t2 = add <8 x i32> %nbits, <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 undef, i32 64>
+  %t1 = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
+  %t2 = add <8 x i32> %nbits, <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 poison, i32 64>
   %t3 = and <8 x i64> %t1, %x
 
   call void @use8xi64(<8 x i64> %t0)
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-d.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-d.ll
index 1a977f67a6a5a..b79ab79097527 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-d.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-after-truncation-variant-d.ll
@@ -85,12 +85,12 @@ define <8 x i32> @t1_vec_splat(<8 x i64> %x, <8 x i32> %nbits) {
   ret <8 x i32> %t6
 }
 
-define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
-; CHECK-LABEL: @t2_vec_splat_undef(
+define <8 x i32> @t2_vec_splat_poison(<8 x i64> %x, <8 x i32> %nbits) {
+; CHECK-LABEL: @t2_vec_splat_poison(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
 ; CHECK-NEXT:    [[T2:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 undef, i32 -32>
+; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 poison, i32 -32>
 ; CHECK-NEXT:    [[T4:%.*]] = and <8 x i64> [[T2]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
@@ -102,9 +102,9 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T6]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
+  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
   %t2 = lshr <8 x i64> %t1, %t0
-  %t3 = add <8 x i32> %nbits, <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 undef, i32 -32>
+  %t3 = add <8 x i32> %nbits, <i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 -32, i32 poison, i32 -32>
   %t4 = and <8 x i64> %t2, %x
 
   call void @use8xi64(<8 x i64> %t0)
@@ -121,9 +121,9 @@ define <8 x i32> @t2_vec_splat_undef(<8 x i64> %x, <8 x i32> %nbits) {
 define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-LABEL: @t3_vec_nonsplat(
 ; CHECK-NEXT:    [[T0:%.*]] = zext <8 x i32> [[NBITS:%.*]] to <8 x i64>
-; CHECK-NEXT:    [[T1:%.*]] = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, [[T0]]
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, [[T0]]
 ; CHECK-NEXT:    [[T2:%.*]] = lshr <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, [[T0]]
-; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 undef, i32 64>
+; CHECK-NEXT:    [[T3:%.*]] = add <8 x i32> [[NBITS]], <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 poison, i32 64>
 ; CHECK-NEXT:    [[T4:%.*]] = and <8 x i64> [[T2]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T0]])
 ; CHECK-NEXT:    call void @use8xi64(<8 x i64> [[T1]])
@@ -135,9 +135,9 @@ define <8 x i32> @t3_vec_nonsplat(<8 x i64> %x, <8 x i32> %nbits) {
 ; CHECK-NEXT:    ret <8 x i32> [[T6]]
 ;
   %t0 = zext <8 x i32> %nbits to <8 x i64>
-  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 undef, i64 -1>, %t0
+  %t1 = shl <8 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 poison, i64 -1>, %t0
   %t2 = lshr <8 x i64> %t1, %t0
-  %t3 = add <8 x i32> %nbits, <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 undef, i32 64>
+  %t3 = add <8 x i32> %nbits, <i32 -32, i32 -1, i32 0, i32 1, i32 31, i32 32, i32 poison, i32 64>
   %t4 = and <8 x i64> %t2, %x
 
   call void @use8xi64(<8 x i64> %t0)
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-b.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-b.ll
index ddaef5f4b47c8..4b955a894fcfe 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-b.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-b.ll
@@ -155,12 +155,12 @@ define <3 x i32> @t4_vec_nonsplat(<3 x i32> %x, <3 x i32> %nbits) {
   ret <3 x i32> %t5
 }
 
-define <3 x i32> @t5_vec_undef(<3 x i32> %x, <3 x i32> %nbits) {
-; CHECK-LABEL: @t5_vec_undef(
-; CHECK-NEXT:    [[T1:%.*]] = shl <3 x i32> <i32 -1, i32 undef, i32 -1>, [[NBITS:%.*]]
-; CHECK-NEXT:    [[T2:%.*]] = xor <3 x i32> [[T1]], <i32 -1, i32 undef, i32 -1>
+define <3 x i32> @t5_vec_poison(<3 x i32> %x, <3 x i32> %nbits) {
+; CHECK-LABEL: @t5_vec_poison(
+; CHECK-NEXT:    [[T1:%.*]] = shl nsw <3 x i32> <i32 -1, i32 poison, i32 -1>, [[NBITS:%.*]]
+; CHECK-NEXT:    [[T2:%.*]] = xor <3 x i32> [[T1]], <i32 -1, i32 poison, i32 -1>
 ; CHECK-NEXT:    [[T3:%.*]] = and <3 x i32> [[T2]], [[X:%.*]]
-; CHECK-NEXT:    [[T4:%.*]] = sub <3 x i32> <i32 32, i32 undef, i32 32>, [[NBITS]]
+; CHECK-NEXT:    [[T4:%.*]] = sub <3 x i32> <i32 32, i32 poison, i32 32>, [[NBITS]]
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[NBITS]])
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T1]])
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T2]])
@@ -169,11 +169,11 @@ define <3 x i32> @t5_vec_undef(<3 x i32> %x, <3 x i32> %nbits) {
 ; CHECK-NEXT:    [[T5:%.*]] = shl <3 x i32> [[X]], [[T4]]
 ; CHECK-NEXT:    ret <3 x i32> [[T5]]
 ;
-  %t0 = add <3 x i32> %nbits, <i32 0, i32 undef, i32 0>
-  %t1 = shl <3 x i32> <i32 -1, i32 undef, i32 -1>, %t0
-  %t2 = xor <3 x i32> %t1, <i32 -1, i32 undef, i32 -1>
+  %t0 = add <3 x i32> %nbits, <i32 0, i32 poison, i32 0>
+  %t1 = shl <3 x i32> <i32 -1, i32 poison, i32 -1>, %t0
+  %t2 = xor <3 x i32> %t1, <i32 -1, i32 poison, i32 -1>
   %t3 = and <3 x i32> %t2, %x
-  %t4 = sub <3 x i32> <i32 32, i32 undef, i32 32>, %nbits
+  %t4 = sub <3 x i32> <i32 32, i32 poison, i32 32>, %nbits
   call void @use3xi32(<3 x i32> %t0)
   call void @use3xi32(<3 x i32> %t1)
   call void @use3xi32(<3 x i32> %t2)
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-c.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-c.ll
index c7747cfafcff5..8428ef67d6b86 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-c.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-c.ll
@@ -99,20 +99,20 @@ define <3 x i32> @t3_vec_nonsplat(<3 x i32> %x, <3 x i32> %nbits) {
   ret <3 x i32> %t3
 }
 
-define <3 x i32> @t4_vec_undef(<3 x i32> %x, <3 x i32> %nbits) {
-; CHECK-LABEL: @t4_vec_undef(
-; CHECK-NEXT:    [[T0:%.*]] = lshr <3 x i32> <i32 -1, i32 undef, i32 -1>, [[NBITS:%.*]]
+define <3 x i32> @t4_vec_poison(<3 x i32> %x, <3 x i32> %nbits) {
+; CHECK-LABEL: @t4_vec_poison(
+; CHECK-NEXT:    [[T0:%.*]] = lshr <3 x i32> <i32 -1, i32 poison, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[T1:%.*]] = and <3 x i32> [[T0]], [[X:%.*]]
-; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[NBITS]], <i32 1, i32 undef, i32 1>
+; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[NBITS]], <i32 1, i32 poison, i32 1>
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T0]])
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T1]])
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T2]])
 ; CHECK-NEXT:    [[T3:%.*]] = shl <3 x i32> [[X]], [[T2]]
 ; CHECK-NEXT:    ret <3 x i32> [[T3]]
 ;
-  %t0 = lshr <3 x i32> <i32 -1, i32 undef, i32 -1>, %nbits
+  %t0 = lshr <3 x i32> <i32 -1, i32 poison, i32 -1>, %nbits
   %t1 = and <3 x i32> %t0, %x
-  %t2 = add <3 x i32> %nbits, <i32 1, i32 undef, i32 1>
+  %t2 = add <3 x i32> %nbits, <i32 1, i32 poison, i32 1>
   call void @use3xi32(<3 x i32> %t0)
   call void @use3xi32(<3 x i32> %t1)
   call void @use3xi32(<3 x i32> %t2)
diff --git a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-d.ll b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-d.ll
index 549729fe8b59c..5d8ff9e9fb71b 100644
--- a/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-d.ll
+++ b/llvm/test/Transforms/InstCombine/redundant-left-shift-input-masking-variant-d.ll
@@ -115,9 +115,9 @@ define <3 x i32> @t3_vec_nonsplat(<3 x i32> %x, <3 x i32> %nbits) {
   ret <3 x i32> %t4
 }
 
-define <3 x i32> @t4_vec_undef(<3 x i32> %x, <3 x i32> %nbits) {
-; CHECK-LABEL: @t4_vec_undef(
-; CHECK-NEXT:    [[T0:%.*]] = shl <3 x i32> <i32 -1, i32 undef, i32 -1>, [[NBITS:%.*]]
+define <3 x i32> @t4_vec_poison(<3 x i32> %x, <3 x i32> %nbits) {
+; CHECK-LABEL: @t4_vec_poison(
+; CHECK-NEXT:    [[T0:%.*]] = shl nsw <3 x i32> <i32 -1, i32 poison, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[T1:%.*]] = lshr <3 x i32> <i32 -1, i32 -1, i32 -1>, [[NBITS]]
 ; CHECK-NEXT:    [[T2:%.*]] = and <3 x i32> [[T1]], [[X:%.*]]
 ; CHECK-NEXT:    call void @use3xi32(<3 x i32> [[T0]])
@@ -127,10 +127,10 @@ define <3 x i32> @t4_vec_undef(<3 x i32> %x, <3 x i32> %nbits) {
 ; CHECK-NEXT:    [[T4:%.*]] = shl <3 x i32> [[X]], [[NBITS]]
 ; CHECK-NEXT:    ret <3 x i32> [[T4]]
 ;
-  %t0 = shl <3 x i32> <i32 -1, i32 undef, i32 -1>, %nbits
+  %t0 = shl <3 x i32> <i32 -1, i32 poison, i32 -1>, %nbits
   %t1 = lshr <3 x i32> %t0, %nbits
   %t2 = and <3 x i32> %t1, %x
-  %t3 = add <3 x i32> %nbits, <i32 0, i32 undef, i32 0>
+  %t3 = add <3 x i32> %nbits, <i32 0, i32 poison, i32 0>
   call void @use3xi32(<3 x i32> %t0)
   call void @use3xi32(<3 x i32> %t1)
   call void @use3xi32(<3 x i32> %t2)
diff --git a/llvm/test/Transforms/InstCombine/reuse-constant-from-select-in-icmp.ll b/llvm/test/Transforms/InstCombine/reuse-constant-from-select-in-icmp.ll
index fd0d942ad840b..301ead708a08f 100644
--- a/llvm/test/Transforms/InstCombine/reuse-constant-from-select-in-icmp.ll
+++ b/llvm/test/Transforms/InstCombine/reuse-constant-from-select-in-icmp.ll
@@ -102,36 +102,36 @@ define <2 x i32> @p7_vec_splat_sgt(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %r
 }
 
-; Vectors with undef
+; Vectors with poison
 
-define <2 x i32> @p8_vec_nonsplat_undef0(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @p8_vec_nonsplat_undef0(
+define <2 x i32> @p8_vec_nonsplat_poison0(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @p8_vec_nonsplat_poison0(
 ; CHECK-NEXT:    [[T_INV:%.*]] = icmp ugt <2 x i32> [[X:%.*]], <i32 65535, i32 65535>
 ; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[T_INV]], <2 x i32> <i32 65535, i32 65535>, <2 x i32> [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %t = icmp ult <2 x i32> %x, <i32 65536, i32 undef>
+  %t = icmp ult <2 x i32> %x, <i32 65536, i32 poison>
   %r = select <2 x i1> %t, <2 x i32> %y, <2 x i32> <i32 65535, i32 65535>
   ret <2 x i32> %r
 }
-define <2 x i32> @p9_vec_nonsplat_undef1(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @p9_vec_nonsplat_undef1(
+define <2 x i32> @p9_vec_nonsplat_poison1(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @p9_vec_nonsplat_poison1(
 ; CHECK-NEXT:    [[T_INV:%.*]] = icmp ugt <2 x i32> [[X:%.*]], <i32 65535, i32 65535>
-; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[T_INV]], <2 x i32> <i32 65535, i32 undef>, <2 x i32> [[Y:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[T_INV]], <2 x i32> <i32 65535, i32 poison>, <2 x i32> [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %t = icmp ult <2 x i32> %x, <i32 65536, i32 65536>
-  %r = select <2 x i1> %t, <2 x i32> %y, <2 x i32> <i32 65535, i32 undef>
+  %r = select <2 x i1> %t, <2 x i32> %y, <2 x i32> <i32 65535, i32 poison>
   ret <2 x i32> %r
 }
-define <2 x i32> @p10_vec_nonsplat_undef2(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @p10_vec_nonsplat_undef2(
+define <2 x i32> @p10_vec_nonsplat_poison2(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @p10_vec_nonsplat_poison2(
 ; CHECK-NEXT:    [[T_INV:%.*]] = icmp ugt <2 x i32> [[X:%.*]], <i32 65535, i32 65535>
-; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[T_INV]], <2 x i32> <i32 65535, i32 undef>, <2 x i32> [[Y:%.*]]
+; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[T_INV]], <2 x i32> <i32 65535, i32 poison>, <2 x i32> [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %t = icmp ult <2 x i32> %x, <i32 65536, i32 undef>
-  %r = select <2 x i1> %t, <2 x i32> %y, <2 x i32> <i32 65535, i32 undef>
+  %t = icmp ult <2 x i32> %x, <i32 65536, i32 poison>
+  %r = select <2 x i1> %t, <2 x i32> %y, <2 x i32> <i32 65535, i32 poison>
   ret <2 x i32> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/rotate.ll b/llvm/test/Transforms/InstCombine/rotate.ll
index 6c70c791fd881..eec623e2f193a 100644
--- a/llvm/test/Transforms/InstCombine/rotate.ll
+++ b/llvm/test/Transforms/InstCombine/rotate.ll
@@ -65,24 +65,24 @@ define <2 x i16> @rotl_v2i16_constant_splat(<2 x i16> %x) {
   ret <2 x i16> %r
 }
 
-define <2 x i16> @rotl_v2i16_constant_splat_undef0(<2 x i16> %x) {
-; CHECK-LABEL: @rotl_v2i16_constant_splat_undef0(
+define <2 x i16> @rotl_v2i16_constant_splat_poison0(<2 x i16> %x) {
+; CHECK-LABEL: @rotl_v2i16_constant_splat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> [[X:%.*]], <2 x i16> [[X]], <2 x i16> <i16 1, i16 1>)
 ; CHECK-NEXT:    ret <2 x i16> [[R]]
 ;
-  %shl = shl <2 x i16> %x, <i16 undef, i16 1>
+  %shl = shl <2 x i16> %x, <i16 poison, i16 1>
   %shr = lshr <2 x i16> %x, <i16 15, i16 15>
   %r = or <2 x i16> %shl, %shr
   ret <2 x i16> %r
 }
 
-define <2 x i16> @rotl_v2i16_constant_splat_undef1(<2 x i16> %x) {
-; CHECK-LABEL: @rotl_v2i16_constant_splat_undef1(
+define <2 x i16> @rotl_v2i16_constant_splat_poison1(<2 x i16> %x) {
+; CHECK-LABEL: @rotl_v2i16_constant_splat_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> [[X:%.*]], <2 x i16> [[X]], <2 x i16> <i16 1, i16 1>)
 ; CHECK-NEXT:    ret <2 x i16> [[R]]
 ;
   %shl = shl <2 x i16> %x, <i16 1, i16 1>
-  %shr = lshr <2 x i16> %x, <i16 15, i16 undef>
+  %shr = lshr <2 x i16> %x, <i16 15, i16 poison>
   %r = or <2 x i16> %shl, %shr
   ret <2 x i16> %r
 }
@@ -100,30 +100,30 @@ define <2 x i17> @rotr_v2i17_constant_splat(<2 x i17> %x) {
   ret <2 x i17> %r
 }
 
-define <2 x i17> @rotr_v2i17_constant_splat_undef0(<2 x i17> %x) {
-; CHECK-LABEL: @rotr_v2i17_constant_splat_undef0(
+define <2 x i17> @rotr_v2i17_constant_splat_poison0(<2 x i17> %x) {
+; CHECK-LABEL: @rotr_v2i17_constant_splat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i17> @llvm.fshl.v2i17(<2 x i17> [[X:%.*]], <2 x i17> [[X]], <2 x i17> <i17 12, i17 12>)
 ; CHECK-NEXT:    ret <2 x i17> [[R]]
 ;
-  %shl = shl <2 x i17> %x, <i17 12, i17 undef>
-  %shr = lshr <2 x i17> %x, <i17 undef, i17 5>
+  %shl = shl <2 x i17> %x, <i17 12, i17 poison>
+  %shr = lshr <2 x i17> %x, <i17 poison, i17 5>
   %r = or <2 x i17> %shr, %shl
   ret <2 x i17> %r
 }
 
-define <2 x i17> @rotr_v2i17_constant_splat_undef1(<2 x i17> %x) {
-; CHECK-LABEL: @rotr_v2i17_constant_splat_undef1(
+define <2 x i17> @rotr_v2i17_constant_splat_poison1(<2 x i17> %x) {
+; CHECK-LABEL: @rotr_v2i17_constant_splat_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i17> @llvm.fshl.v2i17(<2 x i17> [[X:%.*]], <2 x i17> [[X]], <2 x i17> <i17 12, i17 12>)
 ; CHECK-NEXT:    ret <2 x i17> [[R]]
 ;
-  %shl = shl <2 x i17> %x, <i17 12, i17 undef>
-  %shr = lshr <2 x i17> %x, <i17 5, i17 undef>
+  %shl = shl <2 x i17> %x, <i17 12, i17 poison>
+  %shr = lshr <2 x i17> %x, <i17 5, i17 poison>
   %r = or <2 x i17> %shr, %shl
   ret <2 x i17> %r
 }
 
 ; Allow arbitrary shift constants.
-; Support undef elements.
+; Support poison elements.
 
 define <2 x i32> @rotr_v2i32_constant_nonsplat(<2 x i32> %x) {
 ; CHECK-LABEL: @rotr_v2i32_constant_nonsplat(
@@ -136,17 +136,6 @@ define <2 x i32> @rotr_v2i32_constant_nonsplat(<2 x i32> %x) {
   ret <2 x i32> %r
 }
 
-define <2 x i32> @rotr_v2i32_constant_nonsplat_undef0(<2 x i32> %x) {
-; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_undef0(
-; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 0, i32 19>)
-; CHECK-NEXT:    ret <2 x i32> [[R]]
-;
-  %shl = shl <2 x i32> %x, <i32 undef, i32 19>
-  %shr = lshr <2 x i32> %x, <i32 15, i32 13>
-  %r = or <2 x i32> %shl, %shr
-  ret <2 x i32> %r
-}
-
 define <2 x i32> @rotr_v2i32_constant_nonsplat_poison0(<2 x i32> %x) {
 ; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_poison0(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 poison, i32 19>)
@@ -158,13 +147,13 @@ define <2 x i32> @rotr_v2i32_constant_nonsplat_poison0(<2 x i32> %x) {
   ret <2 x i32> %r
 }
 
-define <2 x i32> @rotr_v2i32_constant_nonsplat_undef1(<2 x i32> %x) {
-; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_undef1(
+define <2 x i32> @rotr_v2i32_constant_nonsplat_poison1(<2 x i32> %x) {
+; CHECK-LABEL: @rotr_v2i32_constant_nonsplat_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[X:%.*]], <2 x i32> [[X]], <2 x i32> <i32 17, i32 0>)
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %shl = shl <2 x i32> %x, <i32 17, i32 19>
-  %shr = lshr <2 x i32> %x, <i32 15, i32 undef>
+  %shr = lshr <2 x i32> %x, <i32 15, i32 poison>
   %r = or <2 x i32> %shl, %shr
   ret <2 x i32> %r
 }
@@ -180,13 +169,13 @@ define <2 x i36> @rotl_v2i36_constant_nonsplat(<2 x i36> %x) {
   ret <2 x i36> %r
 }
 
-define <3 x i36> @rotl_v3i36_constant_nonsplat_undef0(<3 x i36> %x) {
-; CHECK-LABEL: @rotl_v3i36_constant_nonsplat_undef0(
-; CHECK-NEXT:    [[R:%.*]] = call <3 x i36> @llvm.fshl.v3i36(<3 x i36> [[X:%.*]], <3 x i36> [[X]], <3 x i36> <i36 21, i36 11, i36 0>)
+define <3 x i36> @rotl_v3i36_constant_nonsplat_poison0(<3 x i36> %x) {
+; CHECK-LABEL: @rotl_v3i36_constant_nonsplat_poison0(
+; CHECK-NEXT:    [[R:%.*]] = call <3 x i36> @llvm.fshl.v3i36(<3 x i36> [[X:%.*]], <3 x i36> [[X]], <3 x i36> <i36 21, i36 11, i36 poison>)
 ; CHECK-NEXT:    ret <3 x i36> [[R]]
 ;
-  %shl = shl <3 x i36> %x, <i36 21, i36 11, i36 undef>
-  %shr = lshr <3 x i36> %x, <i36 15, i36 25, i36 undef>
+  %shl = shl <3 x i36> %x, <i36 21, i36 11, i36 poison>
+  %shr = lshr <3 x i36> %x, <i36 15, i36 25, i36 poison>
   %r = or <3 x i36> %shl, %shr
   ret <3 x i36> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/saturating-add-sub.ll b/llvm/test/Transforms/InstCombine/saturating-add-sub.ll
index c1bb6941d4568..57977a72cd08f 100644
--- a/llvm/test/Transforms/InstCombine/saturating-add-sub.ll
+++ b/llvm/test/Transforms/InstCombine/saturating-add-sub.ll
@@ -559,14 +559,14 @@ define <2 x i8> @test_simplify_decrement_vec(<2 x i8> %a) {
   ret <2 x i8> %i2
 }
 
-define <2 x i8> @test_simplify_decrement_vec_undef(<2 x i8> %a) {
-; CHECK-LABEL: @test_simplify_decrement_vec_undef(
+define <2 x i8> @test_simplify_decrement_vec_poison(<2 x i8> %a) {
+; CHECK-LABEL: @test_simplify_decrement_vec_poison(
 ; CHECK-NEXT:    [[I2:%.*]] = call <2 x i8> @llvm.usub.sat.v2i8(<2 x i8> [[A:%.*]], <2 x i8> <i8 1, i8 1>)
 ; CHECK-NEXT:    ret <2 x i8> [[I2]]
 ;
   %i = icmp eq <2 x i8> %a, <i8 0, i8 0>
   %i1 = sub <2 x i8> %a, <i8 1, i8 1>
-  %i2 = select <2 x i1> %i, <2 x i8> <i8 0, i8 undef>, <2 x i8> %i1
+  %i2 = select <2 x i1> %i, <2 x i8> <i8 0, i8 poison>, <2 x i8> %i1
   ret <2 x i8> %i2
 }
 
@@ -1818,14 +1818,14 @@ define <4 x i32> @uadd_sat_constant_vec_commute(<4 x i32> %x) {
 
 define <4 x i32> @uadd_sat_constant_vec_commute_undefs(<4 x i32> %x) {
 ; CHECK-LABEL: @uadd_sat_constant_vec_commute_undefs(
-; CHECK-NEXT:    [[A:%.*]] = add <4 x i32> [[X:%.*]], <i32 42, i32 42, i32 42, i32 undef>
-; CHECK-NEXT:    [[C:%.*]] = icmp ult <4 x i32> [[X]], <i32 -43, i32 -43, i32 undef, i32 -43>
-; CHECK-NEXT:    [[R:%.*]] = select <4 x i1> [[C]], <4 x i32> [[A]], <4 x i32> <i32 -1, i32 undef, i32 -1, i32 -1>
+; CHECK-NEXT:    [[A:%.*]] = add <4 x i32> [[X:%.*]], <i32 42, i32 42, i32 42, i32 poison>
+; CHECK-NEXT:    [[C:%.*]] = icmp ult <4 x i32> [[X]], <i32 -43, i32 -43, i32 poison, i32 -43>
+; CHECK-NEXT:    [[R:%.*]] = select <4 x i1> [[C]], <4 x i32> [[A]], <4 x i32> <i32 -1, i32 poison, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i32> [[R]]
 ;
-  %a = add <4 x i32> %x, <i32 42, i32 42, i32 42, i32 undef>
-  %c = icmp ult <4 x i32> %x, <i32 -43, i32 -43, i32 undef, i32 -43>
-  %r = select <4 x i1> %c, <4 x i32> %a, <4 x i32> <i32 -1, i32 undef, i32 -1, i32 -1>
+  %a = add <4 x i32> %x, <i32 42, i32 42, i32 42, i32 poison>
+  %c = icmp ult <4 x i32> %x, <i32 -43, i32 -43, i32 poison, i32 -43>
+  %r = select <4 x i1> %c, <4 x i32> %a, <4 x i32> <i32 -1, i32 poison, i32 -1, i32 -1>
   ret <4 x i32> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/select-of-bittest.ll b/llvm/test/Transforms/InstCombine/select-of-bittest.ll
index a6f14cbfbfadf..e3eb76de459e2 100644
--- a/llvm/test/Transforms/InstCombine/select-of-bittest.ll
+++ b/llvm/test/Transforms/InstCombine/select-of-bittest.ll
@@ -80,19 +80,18 @@ define <2 x i32> @and_lshr_and_vec_v2(<2 x i32> %arg) {
   ret <2 x i32> %t4
 }
 
-define <3 x i32> @and_lshr_and_vec_undef(<3 x i32> %arg) {
-; CHECK-LABEL: @and_lshr_and_vec_undef(
+define <3 x i32> @and_lshr_and_vec_poison(<3 x i32> %arg) {
+; CHECK-LABEL: @and_lshr_and_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i32> [[ARG:%.*]], <i32 3, i32 poison, i32 3>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <3 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    [[T4:%.*]] = zext <3 x i1> [[TMP2]] to <3 x i32>
 ; CHECK-NEXT:    ret <3 x i32> [[T4]]
 ;
-  %t = and <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t1 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
-  %t2 = lshr <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t3 = and <3 x i32> %t2, <i32 1, i32 undef, i32 1>
-  ; The second element of %t4 is poison because it is (undef ? poison : undef).
-  %t4 = select <3 x i1> %t1, <3 x i32> %t3, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t = and <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t1 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
+  %t2 = lshr <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t3 = and <3 x i32> %t2, <i32 1, i32 poison, i32 1>
+  %t4 = select <3 x i1> %t1, <3 x i32> %t3, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t4
 }
 
@@ -138,17 +137,17 @@ define <2 x i32> @and_and_vec(<2 x i32> %arg) {
   ret <2 x i32> %t3
 }
 
-define <3 x i32> @and_and_vec_undef(<3 x i32> %arg) {
-; CHECK-LABEL: @and_and_vec_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i32> [[ARG:%.*]], <i32 3, i32 -1, i32 3>
+define <3 x i32> @and_and_vec_poison(<3 x i32> %arg) {
+; CHECK-LABEL: @and_and_vec_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i32> [[ARG:%.*]], <i32 3, i32 poison, i32 3>
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <3 x i32> [[TMP1]], zeroinitializer
 ; CHECK-NEXT:    [[T3:%.*]] = zext <3 x i1> [[TMP2]] to <3 x i32>
 ; CHECK-NEXT:    ret <3 x i32> [[T3]]
 ;
-  %t = and <3 x i32> %arg, <i32 2, i32 undef, i32 2>
-  %t1 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
-  %t2 = and <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t3 = select <3 x i1> %t1, <3 x i32> %t2, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t = and <3 x i32> %arg, <i32 2, i32 poison, i32 2>
+  %t1 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
+  %t2 = and <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t3 = select <3 x i1> %t1, <3 x i32> %t2, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t3
 }
 
@@ -221,8 +220,8 @@ define <2 x i32> @f_var0_vec(<2 x i32> %arg, <2 x i32> %arg1) {
   ret <2 x i32> %t5
 }
 
-define <3 x i32> @f_var0_vec_undef(<3 x i32> %arg, <3 x i32> %arg1) {
-; CHECK-LABEL: @f_var0_vec_undef(
+define <3 x i32> @f_var0_vec_poison(<3 x i32> %arg, <3 x i32> %arg1) {
+; CHECK-LABEL: @f_var0_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <3 x i32> [[ARG1:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[ARG:%.*]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
@@ -230,11 +229,11 @@ define <3 x i32> @f_var0_vec_undef(<3 x i32> %arg, <3 x i32> %arg1) {
 ; CHECK-NEXT:    ret <3 x i32> [[T5]]
 ;
   %t = and <3 x i32> %arg, %arg1
-  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
-  %t3 = lshr <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t4 = and <3 x i32> %t3, <i32 1, i32 undef, i32 1>
-  ; The second element of %t5 is poison because it is (undef ? poison : undef).
-  %t5 = select <3 x i1> %t2, <3 x i32> %t4, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
+  %t3 = lshr <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t4 = and <3 x i32> %t3, <i32 1, i32 poison, i32 1>
+  ; The second element of %t5 is poison because it is (poison ? poison : poison).
+  %t5 = select <3 x i1> %t2, <3 x i32> %t4, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t5
 }
 
@@ -284,8 +283,8 @@ define <2 x i32> @f_var1_vec(<2 x i32> %arg, <2 x i32> %arg1) {
   ret <2 x i32> %t4
 }
 
-define <3 x i32> @f_var1_vec_undef(<3 x i32> %arg, <3 x i32> %arg1) {
-; CHECK-LABEL: @f_var1_vec_undef(
+define <3 x i32> @f_var1_vec_poison(<3 x i32> %arg, <3 x i32> %arg1) {
+; CHECK-LABEL: @f_var1_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <3 x i32> [[ARG1:%.*]], <i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[ARG:%.*]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
@@ -293,9 +292,9 @@ define <3 x i32> @f_var1_vec_undef(<3 x i32> %arg, <3 x i32> %arg1) {
 ; CHECK-NEXT:    ret <3 x i32> [[T4]]
 ;
   %t = and <3 x i32> %arg, %arg1
-  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
-  %t3 = and <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t4 = select <3 x i1> %t2, <3 x i32> %t3, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
+  %t3 = and <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t4 = select <3 x i1> %t2, <3 x i32> %t3, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t4
 }
 
@@ -354,20 +353,20 @@ define <2 x i32> @f_var2_vec(<2 x i32> %arg, <2 x i32> %arg1) {
   ret <2 x i32> %t5
 }
 
-define <3 x i32> @f_var2_vec_undef(<3 x i32> %arg, <3 x i32> %arg1) {
-; CHECK-LABEL: @f_var2_vec_undef(
-; CHECK-NEXT:    [[T:%.*]] = and <3 x i32> [[ARG:%.*]], <i32 1, i32 undef, i32 1>
-; CHECK-NEXT:    [[T2:%.*]] = icmp eq <3 x i32> [[T]], <i32 0, i32 undef, i32 0>
+define <3 x i32> @f_var2_vec_poison(<3 x i32> %arg, <3 x i32> %arg1) {
+; CHECK-LABEL: @f_var2_vec_poison(
+; CHECK-NEXT:    [[T:%.*]] = and <3 x i32> [[ARG:%.*]], <i32 1, i32 poison, i32 1>
+; CHECK-NEXT:    [[T2:%.*]] = icmp eq <3 x i32> [[T]], <i32 0, i32 poison, i32 0>
 ; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[ARG]], [[ARG1:%.*]]
-; CHECK-NEXT:    [[T4:%.*]] = and <3 x i32> [[T3]], <i32 1, i32 undef, i32 1>
-; CHECK-NEXT:    [[T5:%.*]] = select <3 x i1> [[T2]], <3 x i32> [[T4]], <3 x i32> <i32 1, i32 undef, i32 1>
+; CHECK-NEXT:    [[T4:%.*]] = and <3 x i32> [[T3]], <i32 1, i32 poison, i32 1>
+; CHECK-NEXT:    [[T5:%.*]] = select <3 x i1> [[T2]], <3 x i32> [[T4]], <3 x i32> <i32 1, i32 poison, i32 1>
 ; CHECK-NEXT:    ret <3 x i32> [[T5]]
 ;
-  %t = and <3 x i32> %arg, <i32 1, i32 undef, i32 1>
-  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
+  %t = and <3 x i32> %arg, <i32 1, i32 poison, i32 1>
+  %t2 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
   %t3 = lshr <3 x i32> %arg, %arg1
-  %t4 = and <3 x i32> %t3, <i32 1, i32 undef, i32 1>
-  %t5 = select <3 x i1> %t2, <3 x i32> %t4, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t4 = and <3 x i32> %t3, <i32 1, i32 poison, i32 1>
+  %t5 = select <3 x i1> %t2, <3 x i32> %t4, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t5
 }
 
@@ -427,20 +426,20 @@ define <2 x i32> @f_var3_splatvec(<2 x i32> %arg, <2 x i32> %arg1, <2 x i32> %ar
   ret <2 x i32> %t6
 }
 
-define <3 x i32> @f_var3_vec_undef(<3 x i32> %arg, <3 x i32> %arg1, <3 x i32> %arg2) {
-; CHECK-LABEL: @f_var3_vec_undef(
+define <3 x i32> @f_var3_vec_poison(<3 x i32> %arg, <3 x i32> %arg1, <3 x i32> %arg2) {
+; CHECK-LABEL: @f_var3_vec_poison(
 ; CHECK-NEXT:    [[T:%.*]] = and <3 x i32> [[ARG:%.*]], [[ARG1:%.*]]
-; CHECK-NEXT:    [[T3:%.*]] = icmp eq <3 x i32> [[T]], <i32 0, i32 undef, i32 0>
+; CHECK-NEXT:    [[T3:%.*]] = icmp eq <3 x i32> [[T]], <i32 0, i32 poison, i32 0>
 ; CHECK-NEXT:    [[T4:%.*]] = lshr <3 x i32> [[ARG]], [[ARG2:%.*]]
-; CHECK-NEXT:    [[T5:%.*]] = and <3 x i32> [[T4]], <i32 1, i32 undef, i32 1>
-; CHECK-NEXT:    [[T6:%.*]] = select <3 x i1> [[T3]], <3 x i32> [[T5]], <3 x i32> <i32 1, i32 undef, i32 1>
+; CHECK-NEXT:    [[T5:%.*]] = and <3 x i32> [[T4]], <i32 1, i32 poison, i32 1>
+; CHECK-NEXT:    [[T6:%.*]] = select <3 x i1> [[T3]], <3 x i32> [[T5]], <3 x i32> <i32 1, i32 poison, i32 1>
 ; CHECK-NEXT:    ret <3 x i32> [[T6]]
 ;
   %t = and <3 x i32> %arg, %arg1
-  %t3 = icmp eq <3 x i32> %t, <i32 0, i32 undef, i32 0>
+  %t3 = icmp eq <3 x i32> %t, <i32 0, i32 poison, i32 0>
   %t4 = lshr <3 x i32> %arg, %arg2
-  %t5 = and <3 x i32> %t4, <i32 1, i32 undef, i32 1>
-  %t6 = select <3 x i1> %t3, <3 x i32> %t5, <3 x i32> <i32 1, i32 undef, i32 1>
+  %t5 = and <3 x i32> %t4, <i32 1, i32 poison, i32 1>
+  %t6 = select <3 x i1> %t3, <3 x i32> %t5, <3 x i32> <i32 1, i32 poison, i32 1>
   ret <3 x i32> %t6
 }
 
diff --git a/llvm/test/Transforms/InstCombine/select.ll b/llvm/test/Transforms/InstCombine/select.ll
index bd8145ab2a35b..8654691c6f875 100644
--- a/llvm/test/Transforms/InstCombine/select.ll
+++ b/llvm/test/Transforms/InstCombine/select.ll
@@ -3109,45 +3109,46 @@ define <4 x i32> @mul_select_eq_zero_vector(<4 x i32> %x, <4 x i32> %y) {
 }
 
 ; Check that a select is folded into multiplication if condition's operand
-; is a vector consisting of zeros and undefs.
-; select (<k x elt> x == {0, undef, ...}), <k x elt> 0, <k x elt> x * y --> freeze(y) * x
-define <2 x i32> @mul_select_eq_undef_vector(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @mul_select_eq_undef_vector(
-; CHECK-NEXT:    [[Y_FR:%.*]] = freeze <2 x i32> [[Y:%.*]]
+; is a vector consisting of zeros and poisons.
+; select (<k x elt> x == {0, poison, ...}), <k x elt> 0, <k x elt> x * y --> freeze(y) * x
+define <2 x i32> @mul_select_eq_poison_vector(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @mul_select_eq_poison_vector(
+; CHECK-NEXT:    [[C:%.*]] = icmp eq <2 x i32> [[Y_FR:%.*]], <i32 0, i32 poison>
 ; CHECK-NEXT:    [[M:%.*]] = mul <2 x i32> [[Y_FR]], [[X:%.*]]
-; CHECK-NEXT:    ret <2 x i32> [[M]]
+; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[C]], <2 x i32> <i32 0, i32 42>, <2 x i32> [[M]]
+; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %c = icmp eq <2 x i32> %x, <i32 0, i32 undef>
+  %c = icmp eq <2 x i32> %x, <i32 0, i32 poison>
   %m = mul <2 x i32> %x, %y
   %r = select <2 x i1> %c, <2 x i32> <i32 0, i32 42>, <2 x i32> %m
   ret <2 x i32> %r
 }
 
 ; Check that a select is folded into multiplication if other select's operand
-; is a vector consisting of zeros and undefs.
-; select (<k x elt> x == 0), <k x elt> {0, undef, ...}, <k x elt> x * y --> freeze(y) * x
-define <2 x i32> @mul_select_eq_zero_sel_undef_vector(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @mul_select_eq_zero_sel_undef_vector(
+; is a vector consisting of zeros and poisons.
+; select (<k x elt> x == 0), <k x elt> {0, poison, ...}, <k x elt> x * y --> freeze(y) * x
+define <2 x i32> @mul_select_eq_zero_sel_poison_vector(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @mul_select_eq_zero_sel_poison_vector(
 ; CHECK-NEXT:    [[Y_FR:%.*]] = freeze <2 x i32> [[Y:%.*]]
 ; CHECK-NEXT:    [[M:%.*]] = mul <2 x i32> [[Y_FR]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[M]]
 ;
   %c = icmp eq <2 x i32> %x, zeroinitializer
   %m = mul <2 x i32> %x, %y
-  %r = select <2 x i1> %c, <2 x i32> <i32 0, i32 undef>, <2 x i32> %m
+  %r = select <2 x i1> %c, <2 x i32> <i32 0, i32 poison>, <2 x i32> %m
   ret <2 x i32> %r
 }
 
 ; Negative test: select should not be folded into mul because
 ; condition's operand and select's operand do not merge into zero vector.
-define <2 x i32> @mul_select_eq_undef_vector_not_merging_to_zero(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @mul_select_eq_undef_vector_not_merging_to_zero(
-; CHECK-NEXT:    [[C:%.*]] = icmp eq <2 x i32> [[X:%.*]], <i32 0, i32 undef>
+define <2 x i32> @mul_select_eq_poison_vector_not_merging_to_zero(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @mul_select_eq_poison_vector_not_merging_to_zero(
+; CHECK-NEXT:    [[C:%.*]] = icmp eq <2 x i32> [[X:%.*]], <i32 0, i32 poison>
 ; CHECK-NEXT:    [[M:%.*]] = mul <2 x i32> [[X]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[C]], <2 x i32> <i32 1, i32 0>, <2 x i32> [[M]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %c = icmp eq <2 x i32> %x, <i32 0, i32 undef>
+  %c = icmp eq <2 x i32> %x, <i32 0, i32 poison>
   %m = mul <2 x i32> %x, %y
   %r = select <2 x i1> %c, <2 x i32> <i32 1, i32 0>, <2 x i32> %m
   ret <2 x i32> %r
diff --git a/llvm/test/Transforms/InstCombine/select_meta.ll b/llvm/test/Transforms/InstCombine/select_meta.ll
index aa794e82e0fdc..3898fd9fa1f57 100644
--- a/llvm/test/Transforms/InstCombine/select_meta.ll
+++ b/llvm/test/Transforms/InstCombine/select_meta.ll
@@ -301,15 +301,15 @@ define <2 x i32> @not_cond_vec(<2 x i1> %c, <2 x i32> %tv, <2 x i32> %fv) {
   ret <2 x i32> %r
 }
 
-; Should match vector 'not' with undef element.
+; Should match vector 'not' with poison element.
 ; The condition is inverted, and the select ops are swapped. The metadata should be swapped.
 
-define <2 x i32> @not_cond_vec_undef(<2 x i1> %c, <2 x i32> %tv, <2 x i32> %fv) {
-; CHECK-LABEL: @not_cond_vec_undef(
+define <2 x i32> @not_cond_vec_poison(<2 x i1> %c, <2 x i32> %tv, <2 x i32> %fv) {
+; CHECK-LABEL: @not_cond_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = select <2 x i1> [[C:%.*]], <2 x i32> [[FV:%.*]], <2 x i32> [[TV:%.*]], !prof [[PROF1]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %notc = xor <2 x i1> %c, <i1 undef, i1 true>
+  %notc = xor <2 x i1> %c, <i1 poison, i1 true>
   %r = select <2 x i1> %notc, <2 x i32> %tv, <2 x i32> %fv, !prof !1
   ret <2 x i32> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/set-lowbits-mask-canonicalize.ll b/llvm/test/Transforms/InstCombine/set-lowbits-mask-canonicalize.ll
index 3ee0224eb1d03..a3c8d3393d04f 100644
--- a/llvm/test/Transforms/InstCombine/set-lowbits-mask-canonicalize.ll
+++ b/llvm/test/Transforms/InstCombine/set-lowbits-mask-canonicalize.ll
@@ -196,36 +196,36 @@ define <2 x i32> @shl_add_vec(<2 x i32> %NBits) {
   ret <2 x i32> %ret
 }
 
-define <3 x i32> @shl_add_vec_undef0(<3 x i32> %NBits) {
-; CHECK-LABEL: @shl_add_vec_undef0(
+define <3 x i32> @shl_add_vec_poison0(<3 x i32> %NBits) {
+; CHECK-LABEL: @shl_add_vec_poison0(
 ; CHECK-NEXT:    [[NOTMASK:%.*]] = shl nsw <3 x i32> <i32 -1, i32 -1, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = xor <3 x i32> [[NOTMASK]], <i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
-  %setbit = shl <3 x i32> <i32 1, i32 undef, i32 1>, %NBits
+  %setbit = shl <3 x i32> <i32 1, i32 poison, i32 1>, %NBits
   %ret = add <3 x i32> %setbit, <i32 -1, i32 -1, i32 -1>
   ret <3 x i32> %ret
 }
 
-define <3 x i32> @shl_add_vec_undef1(<3 x i32> %NBits) {
-; CHECK-LABEL: @shl_add_vec_undef1(
+define <3 x i32> @shl_add_vec_poison1(<3 x i32> %NBits) {
+; CHECK-LABEL: @shl_add_vec_poison1(
 ; CHECK-NEXT:    [[NOTMASK:%.*]] = shl nsw <3 x i32> <i32 -1, i32 -1, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = xor <3 x i32> [[NOTMASK]], <i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
   %setbit = shl <3 x i32> <i32 1, i32 1, i32 1>, %NBits
-  %ret = add <3 x i32> %setbit, <i32 -1, i32 undef, i32 -1>
+  %ret = add <3 x i32> %setbit, <i32 -1, i32 poison, i32 -1>
   ret <3 x i32> %ret
 }
 
-define <3 x i32> @shl_add_vec_undef2(<3 x i32> %NBits) {
-; CHECK-LABEL: @shl_add_vec_undef2(
+define <3 x i32> @shl_add_vec_poison2(<3 x i32> %NBits) {
+; CHECK-LABEL: @shl_add_vec_poison2(
 ; CHECK-NEXT:    [[NOTMASK:%.*]] = shl nsw <3 x i32> <i32 -1, i32 -1, i32 -1>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[RET:%.*]] = xor <3 x i32> [[NOTMASK]], <i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <3 x i32> [[RET]]
 ;
-  %setbit = shl <3 x i32> <i32 1, i32 undef, i32 1>, %NBits
-  %ret = add <3 x i32> %setbit, <i32 -1, i32 undef, i32 -1>
+  %setbit = shl <3 x i32> <i32 1, i32 poison, i32 1>, %NBits
+  %ret = add <3 x i32> %setbit, <i32 -1, i32 poison, i32 -1>
   ret <3 x i32> %ret
 }
 
diff --git a/llvm/test/Transforms/InstCombine/sext.ll b/llvm/test/Transforms/InstCombine/sext.ll
index e3b6058ce7f80..6d263cfcda057 100644
--- a/llvm/test/Transforms/InstCombine/sext.ll
+++ b/llvm/test/Transforms/InstCombine/sext.ll
@@ -167,39 +167,39 @@ define <2 x i32> @test10_vec_nonuniform(<2 x i32> %i) {
   ret <2 x i32> %D
 }
 
-define <2 x i32> @test10_vec_undef0(<2 x i32> %i) {
-; CHECK-LABEL: @test10_vec_undef0(
-; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i32> [[I:%.*]], <i32 30, i32 undef>
-; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i32> [[D1]], <i32 30, i32 undef>
+define <2 x i32> @test10_vec_poison0(<2 x i32> %i) {
+; CHECK-LABEL: @test10_vec_poison0(
+; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i32> [[I:%.*]], <i32 30, i32 poison>
+; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i32> [[D1]], <i32 30, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[D]]
 ;
   %A = trunc <2 x i32> %i to <2 x i8>
   %B = shl <2 x i8> %A, <i8 6, i8 0>
-  %C = ashr <2 x i8> %B, <i8 6, i8 undef>
+  %C = ashr <2 x i8> %B, <i8 6, i8 poison>
   %D = sext <2 x i8> %C to <2 x i32>
   ret <2 x i32> %D
 }
-define <2 x i32> @test10_vec_undef1(<2 x i32> %i) {
-; CHECK-LABEL: @test10_vec_undef1(
+define <2 x i32> @test10_vec_poison1(<2 x i32> %i) {
+; CHECK-LABEL: @test10_vec_poison1(
 ; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i32> [[I:%.*]], <i32 30, i32 undef>
 ; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i32> [[D1]], <i32 30, i32 undef>
 ; CHECK-NEXT:    ret <2 x i32> [[D]]
 ;
   %A = trunc <2 x i32> %i to <2 x i8>
-  %B = shl <2 x i8> %A, <i8 6, i8 undef>
+  %B = shl <2 x i8> %A, <i8 6, i8 poison>
   %C = ashr <2 x i8> %B, <i8 6, i8 0>
   %D = sext <2 x i8> %C to <2 x i32>
   ret <2 x i32> %D
 }
-define <2 x i32> @test10_vec_undef2(<2 x i32> %i) {
-; CHECK-LABEL: @test10_vec_undef2(
-; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i32> [[I:%.*]], <i32 30, i32 undef>
-; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i32> [[D1]], <i32 30, i32 undef>
+define <2 x i32> @test10_vec_poison2(<2 x i32> %i) {
+; CHECK-LABEL: @test10_vec_poison2(
+; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i32> [[I:%.*]], <i32 30, i32 poison>
+; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i32> [[D1]], <i32 30, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[D]]
 ;
   %A = trunc <2 x i32> %i to <2 x i8>
-  %B = shl <2 x i8> %A, <i8 6, i8 undef>
-  %C = ashr <2 x i8> %B, <i8 6, i8 undef>
+  %B = shl <2 x i8> %A, <i8 6, i8 poison>
+  %C = ashr <2 x i8> %B, <i8 6, i8 poison>
   %D = sext <2 x i8> %C to <2 x i32>
   ret <2 x i32> %D
 }
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest.ll
index 0262db1a01e5c..96d429c62a88f 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-in-bittest.ll
@@ -143,34 +143,34 @@ define <2 x i1> @t8_const_lshr_shl_ne_vec_nonsplat(<2 x i32> %x, <2 x i32> %y) {
   %t3 = icmp ne <2 x i32> %t2, <i32 0, i32 0>
   ret <2 x i1> %t3
 }
-define <3 x i1> @t9_const_lshr_shl_ne_vec_undef0(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t9_const_lshr_shl_ne_vec_undef0(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t9_const_lshr_shl_ne_vec_poison0(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t9_const_lshr_shl_ne_vec_poison0(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
-  %t0 = lshr <3 x i32> %x, <i32 1, i32 undef, i32 1>
+  %t0 = lshr <3 x i32> %x, <i32 1, i32 poison, i32 1>
   %t1 = shl <3 x i32> %y, <i32 1, i32 1, i32 1>
   %t2 = and <3 x i32> %t1, %t0
   %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 0, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t10_const_lshr_shl_ne_vec_undef1(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t10_const_lshr_shl_ne_vec_undef1(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t10_const_lshr_shl_ne_vec_poison1(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t10_const_lshr_shl_ne_vec_poison1(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
   %t0 = lshr <3 x i32> %x, <i32 1, i32 1, i32 1>
-  %t1 = shl <3 x i32> %y, <i32 1, i32 undef, i32 1>
+  %t1 = shl <3 x i32> %y, <i32 1, i32 poison, i32 1>
   %t2 = and <3 x i32> %t1, %t0
   %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 0, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t11_const_lshr_shl_ne_vec_undef2(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t11_const_lshr_shl_ne_vec_undef2(
+define <3 x i1> @t11_const_lshr_shl_ne_vec_poison2(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t11_const_lshr_shl_ne_vec_poison2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 2, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
@@ -179,59 +179,59 @@ define <3 x i1> @t11_const_lshr_shl_ne_vec_undef2(<3 x i32> %x, <3 x i32> %y) {
   %t0 = lshr <3 x i32> %x, <i32 1, i32 1, i32 1>
   %t1 = shl <3 x i32> %y, <i32 1, i32 1, i32 1>
   %t2 = and <3 x i32> %t1, %t0
-  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 undef, i32 0>
+  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t12_const_lshr_shl_ne_vec_undef3(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t12_const_lshr_shl_ne_vec_undef3(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t12_const_lshr_shl_ne_vec_poison3(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t12_const_lshr_shl_ne_vec_poison3(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
-  %t0 = lshr <3 x i32> %x, <i32 1, i32 undef, i32 1>
-  %t1 = shl <3 x i32> %y, <i32 1, i32 undef, i32 1>
+  %t0 = lshr <3 x i32> %x, <i32 1, i32 poison, i32 1>
+  %t1 = shl <3 x i32> %y, <i32 1, i32 poison, i32 1>
   %t2 = and <3 x i32> %t1, %t0
   %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 0, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t13_const_lshr_shl_ne_vec_undef4(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t13_const_lshr_shl_ne_vec_undef4(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t13_const_lshr_shl_ne_vec_poison4(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t13_const_lshr_shl_ne_vec_poison4(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
   %t0 = lshr <3 x i32> %x, <i32 1, i32 1, i32 1>
-  %t1 = shl <3 x i32> %y, <i32 1, i32 undef, i32 1>
+  %t1 = shl <3 x i32> %y, <i32 1, i32 poison, i32 1>
   %t2 = and <3 x i32> %t1, %t0
-  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 undef, i32 0>
+  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t14_const_lshr_shl_ne_vec_undef5(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t14_const_lshr_shl_ne_vec_undef5(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t14_const_lshr_shl_ne_vec_poison5(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t14_const_lshr_shl_ne_vec_poison5(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
-  %t0 = lshr <3 x i32> %x, <i32 1, i32 undef, i32 1>
+  %t0 = lshr <3 x i32> %x, <i32 1, i32 poison, i32 1>
   %t1 = shl <3 x i32> %y, <i32 1, i32 1, i32 1>
   %t2 = and <3 x i32> %t1, %t0
-  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 undef, i32 0>
+  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %t3
 }
-define <3 x i1> @t15_const_lshr_shl_ne_vec_undef6(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t15_const_lshr_shl_ne_vec_undef6(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 undef, i32 2>
+define <3 x i1> @t15_const_lshr_shl_ne_vec_poison6(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t15_const_lshr_shl_ne_vec_poison6(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 2, i32 poison, i32 2>
 ; CHECK-NEXT:    [[TMP2:%.*]] = and <3 x i32> [[TMP1]], [[Y:%.*]]
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ne <3 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    ret <3 x i1> [[T3]]
 ;
-  %t0 = lshr <3 x i32> %x, <i32 1, i32 undef, i32 1>
-  %t1 = shl <3 x i32> %y, <i32 1, i32 undef, i32 1>
+  %t0 = lshr <3 x i32> %x, <i32 1, i32 poison, i32 1>
+  %t1 = shl <3 x i32> %y, <i32 1, i32 poison, i32 1>
   %t2 = and <3 x i32> %t1, %t0
-  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 undef, i32 0>
+  %t3 = icmp ne <3 x i32> %t2, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %t3
 }
 
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
index 84dd4c57ebc61..9efc30cc9d916 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-ashr.ll
@@ -42,13 +42,13 @@ define <2 x i16> @t1_vec_splat(<2 x i32> %x, <2 x i16> %y) {
   ret <2 x i16> %t5
 }
 
-define <3 x i16> @t3_vec_nonsplat_undef0(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t3_vec_nonsplat_undef0(
-; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
+define <3 x i16> @t3_vec_nonsplat_poison0(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t3_vec_nonsplat_poison0(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
 ; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
-  %t0 = sub <3 x i16> <i16 32, i16 undef, i16 32>, %y
+  %t0 = sub <3 x i16> <i16 32, i16 poison, i16 32>, %y
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = ashr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
@@ -57,9 +57,9 @@ define <3 x i16> @t3_vec_nonsplat_undef0(<3 x i32> %x, <3 x i16> %y) {
   ret <3 x i16> %t5
 }
 
-define <3 x i16> @t4_vec_nonsplat_undef1(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t4_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
+define <3 x i16> @t4_vec_nonsplat_poison1(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t4_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
 ; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
@@ -67,22 +67,22 @@ define <3 x i16> @t4_vec_nonsplat_undef1(<3 x i32> %x, <3 x i16> %y) {
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = ashr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
-  %t4 = add <3 x i16> %y, <i16 -1, i16 undef, i16 -1>
+  %t4 = add <3 x i16> %y, <i16 -1, i16 poison, i16 -1>
   %t5 = ashr <3 x i16> %t3, %t4
   ret <3 x i16> %t5
 }
 
-define <3 x i16> @t5_vec_nonsplat_undef1(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t5_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
+define <3 x i16> @t5_vec_nonsplat_poison1(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t5_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
 ; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
-  %t0 = sub <3 x i16> <i16 32, i16 undef, i16 32>, %y
+  %t0 = sub <3 x i16> <i16 32, i16 poison, i16 32>, %y
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = ashr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
-  %t4 = add <3 x i16> %y, <i16 -1, i16 undef, i16 -1>
+  %t4 = add <3 x i16> %y, <i16 -1, i16 poison, i16 -1>
   %t5 = ashr <3 x i16> %t3, %t4
   ret <3 x i16> %t5
 }
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
index 214ec88d2e551..c31b6ed3ea2ba 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation-with-truncation-lshr.ll
@@ -42,13 +42,13 @@ define <2 x i16> @t1_vec_splat(<2 x i32> %x, <2 x i16> %y) {
   ret <2 x i16> %t5
 }
 
-define <3 x i16> @t3_vec_nonsplat_undef0(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t3_vec_nonsplat_undef0(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
-; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
+define <3 x i16> @t3_vec_nonsplat_poison0(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t3_vec_nonsplat_poison0(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
-  %t0 = sub <3 x i16> <i16 32, i16 undef, i16 32>, %y
+  %t0 = sub <3 x i16> <i16 32, i16 poison, i16 32>, %y
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = lshr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
@@ -57,32 +57,32 @@ define <3 x i16> @t3_vec_nonsplat_undef0(<3 x i32> %x, <3 x i16> %y) {
   ret <3 x i16> %t5
 }
 
-define <3 x i16> @t4_vec_nonsplat_undef1(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t4_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
-; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
+define <3 x i16> @t4_vec_nonsplat_poison1(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t4_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
   %t0 = sub <3 x i16> <i16 32, i16 32, i16 32>, %y
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = lshr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
-  %t4 = add <3 x i16> %y, <i16 -1, i16 undef, i16 -1>
+  %t4 = add <3 x i16> %y, <i16 -1, i16 poison, i16 -1>
   %t5 = lshr <3 x i16> %t3, %t4
   ret <3 x i16> %t5
 }
 
-define <3 x i16> @t5_vec_nonsplat_undef1(<3 x i32> %x, <3 x i16> %y) {
-; CHECK-LABEL: @t5_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 0, i32 31>
-; CHECK-NEXT:    [[T5:%.*]] = trunc <3 x i32> [[TMP1]] to <3 x i16>
+define <3 x i16> @t5_vec_nonsplat_poison1(<3 x i32> %x, <3 x i16> %y) {
+; CHECK-LABEL: @t5_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 31, i32 poison, i32 31>
+; CHECK-NEXT:    [[T5:%.*]] = trunc nuw nsw <3 x i32> [[TMP1]] to <3 x i16>
 ; CHECK-NEXT:    ret <3 x i16> [[T5]]
 ;
-  %t0 = sub <3 x i16> <i16 32, i16 undef, i16 32>, %y
+  %t0 = sub <3 x i16> <i16 32, i16 poison, i16 32>, %y
   %t1 = zext <3 x i16> %t0 to <3 x i32>
   %t2 = lshr <3 x i32> %x, %t1
   %t3 = trunc <3 x i32> %t2 to <3 x i16>
-  %t4 = add <3 x i16> %y, <i16 -1, i16 undef, i16 -1>
+  %t4 = add <3 x i16> %y, <i16 -1, i16 poison, i16 -1>
   %t5 = lshr <3 x i16> %t3, %t4
   ret <3 x i16> %t5
 }
diff --git a/llvm/test/Transforms/InstCombine/shift-amount-reassociation.ll b/llvm/test/Transforms/InstCombine/shift-amount-reassociation.ll
index b96bcd6bab4f1..6bbe4c5151e45 100644
--- a/llvm/test/Transforms/InstCombine/shift-amount-reassociation.ll
+++ b/llvm/test/Transforms/InstCombine/shift-amount-reassociation.ll
@@ -48,38 +48,38 @@ define <2 x i32> @t2_vec_nonsplat(<2 x i32> %x, <2 x i32> %y) {
 
 ; Basic vector tests
 
-define <3 x i32> @t3_vec_nonsplat_undef0(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t3_vec_nonsplat_undef0(
-; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 undef, i32 30>
+define <3 x i32> @t3_vec_nonsplat_poison0(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t3_vec_nonsplat_poison0(
+; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 poison, i32 30>
 ; CHECK-NEXT:    ret <3 x i32> [[T3]]
 ;
-  %t0 = sub <3 x i32> <i32 32, i32 undef, i32 32>, %y
+  %t0 = sub <3 x i32> <i32 32, i32 poison, i32 32>, %y
   %t1 = lshr <3 x i32> %x, %t0
   %t2 = add <3 x i32> %y, <i32 -2, i32 -2, i32 -2>
   %t3 = lshr <3 x i32> %t1, %t2
   ret <3 x i32> %t3
 }
 
-define <3 x i32> @t4_vec_nonsplat_undef1(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t4_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 undef, i32 30>
+define <3 x i32> @t4_vec_nonsplat_poison1(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t4_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 poison, i32 30>
 ; CHECK-NEXT:    ret <3 x i32> [[T3]]
 ;
   %t0 = sub <3 x i32> <i32 32, i32 32, i32 32>, %y
   %t1 = lshr <3 x i32> %x, %t0
-  %t2 = add <3 x i32> %y, <i32 -2, i32 undef, i32 -2>
+  %t2 = add <3 x i32> %y, <i32 -2, i32 poison, i32 -2>
   %t3 = lshr <3 x i32> %t1, %t2
   ret <3 x i32> %t3
 }
 
-define <3 x i32> @t5_vec_nonsplat_undef1(<3 x i32> %x, <3 x i32> %y) {
-; CHECK-LABEL: @t5_vec_nonsplat_undef1(
-; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 undef, i32 30>
+define <3 x i32> @t5_vec_nonsplat_poison1(<3 x i32> %x, <3 x i32> %y) {
+; CHECK-LABEL: @t5_vec_nonsplat_poison1(
+; CHECK-NEXT:    [[T3:%.*]] = lshr <3 x i32> [[X:%.*]], <i32 30, i32 poison, i32 30>
 ; CHECK-NEXT:    ret <3 x i32> [[T3]]
 ;
-  %t0 = sub <3 x i32> <i32 32, i32 undef, i32 32>, %y
+  %t0 = sub <3 x i32> <i32 32, i32 poison, i32 32>, %y
   %t1 = lshr <3 x i32> %x, %t0
-  %t2 = add <3 x i32> %y, <i32 -2, i32 undef, i32 -2>
+  %t2 = add <3 x i32> %y, <i32 -2, i32 poison, i32 -2>
   %t3 = lshr <3 x i32> %t1, %t2
   ret <3 x i32> %t3
 }
diff --git a/llvm/test/Transforms/InstCombine/shift-logic.ll b/llvm/test/Transforms/InstCombine/shift-logic.ll
index c982b45b504e9..b591400c6a260 100644
--- a/llvm/test/Transforms/InstCombine/shift-logic.ll
+++ b/llvm/test/Transforms/InstCombine/shift-logic.ll
@@ -44,18 +44,18 @@ define i16 @shl_or(i16 %x, i16 %py) {
   ret i16 %sh1
 }
 
-define <2 x i16> @shl_or_undef(<2 x i16> %x, <2 x i16> %py) {
-; CHECK-LABEL: @shl_or_undef(
+define <2 x i16> @shl_or_poison(<2 x i16> %x, <2 x i16> %py) {
+; CHECK-LABEL: @shl_or_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i16> [[PY:%.*]], <i16 42, i16 42>
-; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i16> [[X:%.*]], <i16 12, i16 undef>
-; CHECK-NEXT:    [[TMP2:%.*]] = shl <2 x i16> [[Y]], <i16 7, i16 undef>
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i16> [[X:%.*]], <i16 12, i16 poison>
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw <2 x i16> [[Y]], <i16 7, i16 poison>
 ; CHECK-NEXT:    [[SH1:%.*]] = or <2 x i16> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <2 x i16> [[SH1]]
 ;
   %y = srem <2 x i16> %py, <i16 42, i16 42> ; thwart complexity-based canonicalization
-  %sh0 = shl <2 x i16> %x, <i16 5, i16 undef>
+  %sh0 = shl <2 x i16> %x, <i16 5, i16 poison>
   %r = or <2 x i16> %y, %sh0
-  %sh1 = shl <2 x i16> %r, <i16 7, i16 undef>
+  %sh1 = shl <2 x i16> %r, <i16 7, i16 poison>
   ret <2 x i16> %sh1
 }
 
@@ -100,18 +100,18 @@ define i64 @lshr_and(i64 %x, i64 %py) {
   ret i64 %sh1
 }
 
-define <2 x i64> @lshr_and_undef(<2 x i64> %x, <2 x i64> %py) {
-; CHECK-LABEL: @lshr_and_undef(
+define <2 x i64> @lshr_and_poison(<2 x i64> %x, <2 x i64> %py) {
+; CHECK-LABEL: @lshr_and_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i64> [[PY:%.*]], <i64 42, i64 42>
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 12, i64 undef>
-; CHECK-NEXT:    [[TMP2:%.*]] = lshr <2 x i64> [[Y]], <i64 7, i64 undef>
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 12, i64 poison>
+; CHECK-NEXT:    [[TMP2:%.*]] = lshr <2 x i64> [[Y]], <i64 7, i64 poison>
 ; CHECK-NEXT:    [[SH1:%.*]] = and <2 x i64> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <2 x i64> [[SH1]]
 ;
   %y = srem <2 x i64> %py, <i64 42, i64 42> ; thwart complexity-based canonicalization
-  %sh0 = lshr <2 x i64> %x, <i64 5, i64 undef>
+  %sh0 = lshr <2 x i64> %x, <i64 5, i64 poison>
   %r = and <2 x i64> %y, %sh0
-  %sh1 = lshr <2 x i64> %r, <i64 7, i64 undef>
+  %sh1 = lshr <2 x i64> %r, <i64 7, i64 poison>
   ret <2 x i64> %sh1
 }
 
@@ -212,16 +212,16 @@ define i32 @ashr_overshift_xor(i32 %x, i32 %y) {
   ret i32 %sh1
 }
 
-define <2 x i32> @ashr_undef_undef_xor(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @ashr_undef_undef_xor(
-; CHECK-NEXT:    [[SH0:%.*]] = ashr <2 x i32> [[X:%.*]], <i32 15, i32 undef>
+define <2 x i32> @ashr_poison_poison_xor(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @ashr_poison_poison_xor(
+; CHECK-NEXT:    [[SH0:%.*]] = ashr <2 x i32> [[X:%.*]], <i32 15, i32 poison>
 ; CHECK-NEXT:    [[R:%.*]] = xor <2 x i32> [[SH0]], [[Y:%.*]]
-; CHECK-NEXT:    [[SH1:%.*]] = ashr <2 x i32> [[R]], <i32 undef, i32 17>
+; CHECK-NEXT:    [[SH1:%.*]] = ashr <2 x i32> [[R]], <i32 poison, i32 17>
 ; CHECK-NEXT:    ret <2 x i32> [[SH1]]
 ;
-  %sh0 = ashr <2 x i32> %x, <i32 15, i32 undef>
+  %sh0 = ashr <2 x i32> %x, <i32 15, i32 poison>
   %r = xor <2 x i32> %y, %sh0
-  %sh1 = ashr <2 x i32> %r, <i32 undef, i32 17>
+  %sh1 = ashr <2 x i32> %r, <i32 poison, i32 17>
   ret <2 x i32> %sh1
 }
 
@@ -390,18 +390,18 @@ define <2 x i8> @shl_add_nonuniform(<2 x i8> %x, <2 x i8> %y) {
 }
 
 
-define <2 x i64> @shl_add_undef(<2 x i64> %x, <2 x i64> %py) {
-; CHECK-LABEL: @shl_add_undef(
+define <2 x i64> @shl_add_poison(<2 x i64> %x, <2 x i64> %py) {
+; CHECK-LABEL: @shl_add_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i64> [[PY:%.*]], <i64 42, i64 42>
-; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i64> [[X:%.*]], <i64 12, i64 undef>
-; CHECK-NEXT:    [[TMP2:%.*]] = shl <2 x i64> [[Y]], <i64 7, i64 undef>
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i64> [[X:%.*]], <i64 12, i64 poison>
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw <2 x i64> [[Y]], <i64 7, i64 poison>
 ; CHECK-NEXT:    [[SH1:%.*]] = add <2 x i64> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    ret <2 x i64> [[SH1]]
 ;
   %y = srem <2 x i64> %py, <i64 42, i64 42> ; thwart complexity-based canonicalization
-  %sh0 = shl <2 x i64> %x, <i64 5, i64 undef>
+  %sh0 = shl <2 x i64> %x, <i64 5, i64 poison>
   %r = add <2 x i64> %y, %sh0
-  %sh1 = shl <2 x i64> %r, <i64 7, i64 undef>
+  %sh1 = shl <2 x i64> %r, <i64 7, i64 poison>
   ret <2 x i64> %sh1
 }
 
@@ -432,18 +432,18 @@ define <2 x i8> @lshr_add_nonuniform(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i8> %sh1
 }
 
-define <2 x i64> @lshr_add_undef(<2 x i64> %x, <2 x i64> %py) {
-; CHECK-LABEL: @lshr_add_undef(
+define <2 x i64> @lshr_add_poison(<2 x i64> %x, <2 x i64> %py) {
+; CHECK-LABEL: @lshr_add_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i64> [[PY:%.*]], <i64 42, i64 42>
-; CHECK-NEXT:    [[SH0:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 5, i64 undef>
-; CHECK-NEXT:    [[R:%.*]] = add <2 x i64> [[Y]], [[SH0]]
-; CHECK-NEXT:    [[SH1:%.*]] = lshr <2 x i64> [[R]], <i64 7, i64 undef>
+; CHECK-NEXT:    [[SH0:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 5, i64 poison>
+; CHECK-NEXT:    [[R:%.*]] = add nsw <2 x i64> [[Y]], [[SH0]]
+; CHECK-NEXT:    [[SH1:%.*]] = lshr <2 x i64> [[R]], <i64 7, i64 poison>
 ; CHECK-NEXT:    ret <2 x i64> [[SH1]]
 ;
   %y = srem <2 x i64> %py, <i64 42, i64 42> ; thwart complexity-based canonicalization
-  %sh0 = lshr <2 x i64> %x, <i64 5, i64 undef>
+  %sh0 = lshr <2 x i64> %x, <i64 5, i64 poison>
   %r = add <2 x i64> %y, %sh0
-  %sh1 = lshr <2 x i64> %r, <i64 7, i64 undef>
+  %sh1 = lshr <2 x i64> %r, <i64 7, i64 poison>
   ret <2 x i64> %sh1
 }
 
@@ -488,18 +488,18 @@ define <2 x i8> @shl_sub_nonuniform(<2 x i8> %x, <2 x i8> %y) {
 }
 
 
-define <2 x i64> @shl_sub_undef(<2 x i64> %x, <2 x i64> %py) {
-; CHECK-LABEL: @shl_sub_undef(
+define <2 x i64> @shl_sub_poison(<2 x i64> %x, <2 x i64> %py) {
+; CHECK-LABEL: @shl_sub_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i64> [[PY:%.*]], <i64 42, i64 42>
-; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i64> [[X:%.*]], <i64 12, i64 undef>
-; CHECK-NEXT:    [[TMP2:%.*]] = shl <2 x i64> [[Y]], <i64 7, i64 undef>
+; CHECK-NEXT:    [[TMP1:%.*]] = shl <2 x i64> [[X:%.*]], <i64 12, i64 poison>
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw <2 x i64> [[Y]], <i64 7, i64 poison>
 ; CHECK-NEXT:    [[SH1:%.*]] = sub <2 x i64> [[TMP2]], [[TMP1]]
 ; CHECK-NEXT:    ret <2 x i64> [[SH1]]
 ;
   %y = srem <2 x i64> %py, <i64 42, i64 42> ; thwart complexity-based canonicalization
-  %sh0 = shl <2 x i64> %x, <i64 5, i64 undef>
+  %sh0 = shl <2 x i64> %x, <i64 5, i64 poison>
   %r = sub <2 x i64> %y, %sh0
-  %sh1 = shl <2 x i64> %r, <i64 7, i64 undef>
+  %sh1 = shl <2 x i64> %r, <i64 7, i64 poison>
   ret <2 x i64> %sh1
 }
 
@@ -530,17 +530,17 @@ define <2 x i8> @lshr_sub_nonuniform(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i8> %sh1
 }
 
-define <2 x i64> @lshr_sub_undef(<2 x i64> %x, <2 x i64> %py) {
-; CHECK-LABEL: @lshr_sub_undef(
+define <2 x i64> @lshr_sub_poison(<2 x i64> %x, <2 x i64> %py) {
+; CHECK-LABEL: @lshr_sub_poison(
 ; CHECK-NEXT:    [[Y:%.*]] = srem <2 x i64> [[PY:%.*]], <i64 42, i64 42>
-; CHECK-NEXT:    [[SH0:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 5, i64 undef>
-; CHECK-NEXT:    [[R:%.*]] = sub <2 x i64> [[Y]], [[SH0]]
-; CHECK-NEXT:    [[SH1:%.*]] = lshr <2 x i64> [[R]], <i64 7, i64 undef>
+; CHECK-NEXT:    [[SH0:%.*]] = lshr <2 x i64> [[X:%.*]], <i64 5, i64 poison>
+; CHECK-NEXT:    [[R:%.*]] = sub nsw <2 x i64> [[Y]], [[SH0]]
+; CHECK-NEXT:    [[SH1:%.*]] = lshr <2 x i64> [[R]], <i64 7, i64 poison>
 ; CHECK-NEXT:    ret <2 x i64> [[SH1]]
 ;
   %y = srem <2 x i64> %py, <i64 42, i64 42> ; thwart complexity-based canonicalization
-  %sh0 = lshr <2 x i64> %x, <i64 5, i64 undef>
+  %sh0 = lshr <2 x i64> %x, <i64 5, i64 poison>
   %r = sub <2 x i64> %y, %sh0
-  %sh1 = lshr <2 x i64> %r, <i64 7, i64 undef>
+  %sh1 = lshr <2 x i64> %r, <i64 7, i64 poison>
   ret <2 x i64> %sh1
 }
diff --git a/llvm/test/Transforms/InstCombine/shl-and-negC-icmpeq-zero.ll b/llvm/test/Transforms/InstCombine/shl-and-negC-icmpeq-zero.ll
index 406dc72f2646e..daa4955796594 100644
--- a/llvm/test/Transforms/InstCombine/shl-and-negC-icmpeq-zero.ll
+++ b/llvm/test/Transforms/InstCombine/shl-and-negC-icmpeq-zero.ll
@@ -81,39 +81,39 @@ define <4 x i1> @vec_4xi32_shl_and_negC_eq(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_shl_and_negC_eq_undef1(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_shl_and_negC_eq_undef1(
+define <4 x i1> @vec_shl_and_negC_eq_poison1(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_shl_and_negC_eq_poison1(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[SHL]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
-  %and = and <4 x i32> %shl, <i32 4294967288, i32 undef, i32 4294967288, i32 4294967288>  ; ~7
+  %and = and <4 x i32> %shl, <i32 4294967288, i32 poison, i32 4294967288, i32 4294967288>  ; ~7
   %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_shl_and_negC_eq_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_shl_and_negC_eq_undef2(
+define <4 x i1> @vec_shl_and_negC_eq_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_shl_and_negC_eq_poison2(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[SHL]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
   %and = and <4 x i32> %shl, <i32 4294967288, i32 4294967288, i32 4294967288, i32 4294967288>  ; ~7
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_shl_and_negC_eq_undef3(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_shl_and_negC_eq_undef3(
+define <4 x i1> @vec_shl_and_negC_eq_poison3(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_shl_and_negC_eq_poison3(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp ult <4 x i32> [[SHL]], <i32 8, i32 8, i32 8, i32 8>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
-  %and = and <4 x i32> %shl, <i32 4294967288, i32 4294967288, i32 undef, i32 4294967288>  ; ~7
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %and = and <4 x i32> %shl, <i32 4294967288, i32 4294967288, i32 poison, i32 4294967288>  ; ~7
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/shl-and-signbit-icmpeq-zero.ll b/llvm/test/Transforms/InstCombine/shl-and-signbit-icmpeq-zero.ll
index 4c2c876e3925b..dcc181945357d 100644
--- a/llvm/test/Transforms/InstCombine/shl-and-signbit-icmpeq-zero.ll
+++ b/llvm/test/Transforms/InstCombine/shl-and-signbit-icmpeq-zero.ll
@@ -81,39 +81,39 @@ define <4 x i1> @vec_4xi32_shl_and_signbit_eq(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_shl_and_signbit_eq_undef1(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_undef1(
+define <4 x i1> @vec_4xi32_shl_and_signbit_eq_poison1(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_poison1(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[SHL]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
-  %and = and <4 x i32> %shl, <i32 2147483648, i32 undef, i32 2147483648, i32 2147483648>
+  %and = and <4 x i32> %shl, <i32 2147483648, i32 poison, i32 2147483648, i32 2147483648>
   %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_shl_and_signbit_eq_undef2(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_undef2(
+define <4 x i1> @vec_4xi32_shl_and_signbit_eq_poison2(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_poison2(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[SHL]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
   %and = and <4 x i32> %shl, <i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648>
-  %r = icmp eq <4 x i32> %and, <i32 undef, i32 0, i32 0, i32 0>
+  %r = icmp eq <4 x i32> %and, <i32 poison, i32 0, i32 0, i32 0>
   ret <4 x i1> %r
 }
 
-define <4 x i1> @vec_4xi32_shl_and_signbit_eq_undef3(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_undef3(
+define <4 x i1> @vec_4xi32_shl_and_signbit_eq_poison3(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @vec_4xi32_shl_and_signbit_eq_poison3(
 ; CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = icmp sgt <4 x i32> [[SHL]], <i32 -1, i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    ret <4 x i1> [[R]]
 ;
   %shl = shl <4 x i32> %x, %y
-  %and = and <4 x i32> %shl, <i32 2147483648, i32 undef, i32 2147483648, i32 2147483648>
-  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 undef>
+  %and = and <4 x i32> %shl, <i32 2147483648, i32 poison, i32 2147483648, i32 2147483648>
+  %r = icmp eq <4 x i32> %and, <i32 0, i32 0, i32 0, i32 poison>
   ret <4 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll b/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll
index aeb4c8bb62cba..e7505721cad60 100644
--- a/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll
+++ b/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll
@@ -129,40 +129,56 @@ define <2 x i32> @t8(<2 x i16> %x) {
   %r = and <2 x i32> %i1, <i32 -2147483648, i32 -2147483648>
   ret <2 x i32> %r
 }
+
 define <2 x i32> @t9(<2 x i16> %x) {
 ; CHECK-LABEL: @t9(
-; CHECK-NEXT:    [[X_SIGNEXT:%.*]] = sext <2 x i16> [[X:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[X_SIGNEXT]], <i32 -2147483648, i32 undef>
+; CHECK-NEXT:    [[I1:%.*]] = sext <2 x i16> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[I1]], <i32 -2147483648, i32 -2147483648>
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %i0 = zext <2 x i16> %x to <2 x i32>
-  %i1 = shl <2 x i32> %i0, <i32 16, i32 undef>
+  %i1 = shl <2 x i32> %i0, <i32 16, i32 poison>
   %r = and <2 x i32> %i1, <i32 -2147483648, i32 -2147483648>
-  ; Here undef can be propagated into the mask.
   ret <2 x i32> %r
 }
-define <2 x i32> @t10(<2 x i16> %x) {
-; CHECK-LABEL: @t10(
-; CHECK-NEXT:    [[X_SIGNEXT:%.*]] = sext <2 x i16> [[X:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[X_SIGNEXT]], <i32 -2147483648, i32 0>
+
+; If we folded this, we wouldn't be able to keep the undef mask.
+define <2 x i32> @t10_undef(<2 x i16> %x) {
+; CHECK-LABEL: @t10_undef(
+; CHECK-NEXT:    [[I0:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT:    [[I1:%.*]] = shl nuw <2 x i32> [[I0]], <i32 16, i32 16>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[I1]], <i32 -2147483648, i32 undef>
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %i0 = zext <2 x i16> %x to <2 x i32>
   %i1 = shl <2 x i32> %i0, <i32 16, i32 16>
   %r = and <2 x i32> %i1, <i32 -2147483648, i32 undef>
-  ; CAREFUL! We can't keep undef mask here, since high bits are no longer zero,
+  ret <2 x i32> %r
+}
+
+define <2 x i32> @t10_poison(<2 x i16> %x) {
+; CHECK-LABEL: @t10_poison(
+; CHECK-NEXT:    [[I1:%.*]] = sext <2 x i16> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[I1]], <i32 -2147483648, i32 poison>
+; CHECK-NEXT:    ret <2 x i32> [[R]]
+;
+  %i0 = zext <2 x i16> %x to <2 x i32>
+  %i1 = shl <2 x i32> %i0, <i32 16, i32 16>
+  %r = and <2 x i32> %i1, <i32 -2147483648, i32 poison>
+  ; CAREFUL! We can't keep poison mask here, since high bits are no longer zero,
   ; we must sanitize it to 0.
   ret <2 x i32> %r
 }
+
 define <2 x i32> @t11(<2 x i16> %x) {
 ; CHECK-LABEL: @t11(
 ; CHECK-NEXT:    [[X_SIGNEXT:%.*]] = sext <2 x i16> [[X:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[X_SIGNEXT]], <i32 -2147483648, i32 undef>
+; CHECK-NEXT:    [[R:%.*]] = and <2 x i32> [[X_SIGNEXT]], <i32 -2147483648, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
   %i0 = zext <2 x i16> %x to <2 x i32>
-  %i1 = shl <2 x i32> %i0, <i32 16, i32 undef>
-  %r = and <2 x i32> %i1, <i32 -2147483648, i32 undef>
-  ; Here undef mask is fine.
+  %i1 = shl <2 x i32> %i0, <i32 16, i32 poison>
+  %r = and <2 x i32> %i1, <i32 -2147483648, i32 poison>
+  ; Here poison mask is fine.
   ret <2 x i32> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/sub-not.ll b/llvm/test/Transforms/InstCombine/sub-not.ll
index ec36754d3e9b1..89ccf5aa3c8f4 100644
--- a/llvm/test/Transforms/InstCombine/sub-not.ll
+++ b/llvm/test/Transforms/InstCombine/sub-not.ll
@@ -34,7 +34,7 @@ define <2 x i8> @sub_not_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %s = sub <2 x i8> %x, %y
-  %r = xor <2 x i8> %s, <i8 -1, i8 undef>
+  %r = xor <2 x i8> %s, <i8 -1, i8 poison>
   ret <2 x i8> %r
 }
 
@@ -69,7 +69,7 @@ define <2 x i8> @dec_sub_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %s = sub <2 x i8> %x, %y
-  %r = add <2 x i8> %s, <i8 -1, i8 undef>
+  %r = add <2 x i8> %s, <i8 -1, i8 poison>
   ret <2 x i8> %r
 }
 
@@ -103,7 +103,7 @@ define <2 x i8> @sub_inc_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    [[R:%.*]] = add <2 x i8> [[S_NEG]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %s = add <2 x i8> %x, <i8 undef, i8 1>
+  %s = add <2 x i8> %x, <i8 poison, i8 1>
   %r = sub <2 x i8> %y, %s
   ret <2 x i8> %r
 }
@@ -138,7 +138,7 @@ define <2 x i8> @sub_dec_vec(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-NEXT:    [[R:%.*]] = add <2 x i8> [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %s = add <2 x i8> %x, <i8 undef, i8 -1>
+  %s = add <2 x i8> %x, <i8 poison, i8 -1>
   %r = sub <2 x i8> %s, %y
   ret <2 x i8> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/sub.ll b/llvm/test/Transforms/InstCombine/sub.ll
index 249b5673c8acf..a84e389f13c3b 100644
--- a/llvm/test/Transforms/InstCombine/sub.ll
+++ b/llvm/test/Transforms/InstCombine/sub.ll
@@ -130,44 +130,44 @@ define <2 x i32> @neg_nsw_sub_nsw_vec(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %r
 }
 
-define <2 x i32> @neg_sub_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @neg_sub_vec_undef(
+define <2 x i32> @neg_sub_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @neg_sub_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %neg = sub <2 x i32> <i32 0, i32 undef>, %x
+  %neg = sub <2 x i32> <i32 0, i32 poison>, %x
   %r = sub <2 x i32> %y, %neg
   ret <2 x i32> %r
 }
 
-define <2 x i32> @neg_nsw_sub_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @neg_nsw_sub_vec_undef(
+define <2 x i32> @neg_nsw_sub_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @neg_nsw_sub_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %neg = sub nsw <2 x i32> <i32 undef, i32 0>, %x
+  %neg = sub nsw <2 x i32> <i32 poison, i32 0>, %x
   %r = sub <2 x i32> %y, %neg
   ret <2 x i32> %r
 }
 
-define <2 x i32> @neg_sub_nsw_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @neg_sub_nsw_vec_undef(
+define <2 x i32> @neg_sub_nsw_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @neg_sub_nsw_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = add <2 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %neg = sub <2 x i32> <i32 undef, i32 0>, %x
+  %neg = sub <2 x i32> <i32 poison, i32 0>, %x
   %r = sub nsw <2 x i32> %y, %neg
   ret <2 x i32> %r
 }
 
 ; This should not drop 'nsw'.
 
-define <2 x i32> @neg_nsw_sub_nsw_vec_undef(<2 x i32> %x, <2 x i32> %y) {
-; CHECK-LABEL: @neg_nsw_sub_nsw_vec_undef(
+define <2 x i32> @neg_nsw_sub_nsw_vec_poison(<2 x i32> %x, <2 x i32> %y) {
+; CHECK-LABEL: @neg_nsw_sub_nsw_vec_poison(
 ; CHECK-NEXT:    [[R:%.*]] = add nsw <2 x i32> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[R]]
 ;
-  %neg = sub nsw <2 x i32> <i32 0, i32 undef>, %x
+  %neg = sub nsw <2 x i32> <i32 0, i32 poison>, %x
   %r = sub nsw <2 x i32> %y, %neg
   ret <2 x i32> %r
 }
@@ -205,13 +205,13 @@ define <2 x i8> @notnotsub_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i8> %sub
 }
 
-define <2 x i8> @notnotsub_vec_undef_elts(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: @notnotsub_vec_undef_elts(
+define <2 x i8> @notnotsub_vec_poison_elts(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @notnotsub_vec_poison_elts(
 ; CHECK-NEXT:    [[SUB:%.*]] = sub <2 x i8> [[Y:%.*]], [[X:%.*]]
 ; CHECK-NEXT:    ret <2 x i8> [[SUB]]
 ;
-  %nx = xor <2 x i8> %x, <i8 undef, i8 -1>
-  %ny = xor <2 x i8> %y, <i8 -1, i8 undef>
+  %nx = xor <2 x i8> %x, <i8 poison, i8 -1>
+  %ny = xor <2 x i8> %y, <i8 -1, i8 poison>
   %sub = sub <2 x i8> %nx, %ny
   ret <2 x i8> %sub
 }
@@ -2351,12 +2351,12 @@ define <2 x i8> @sub_to_and_vector1(<2 x i8> %x) {
 
 define <2 x i8> @sub_to_and_vector2(<2 x i8> %x) {
 ; CHECK-LABEL: @sub_to_and_vector2(
-; CHECK-NEXT:    [[SUB:%.*]] = sub nuw <2 x i8> <i8 71, i8 undef>, [[X:%.*]]
+; CHECK-NEXT:    [[SUB:%.*]] = sub nuw <2 x i8> <i8 71, i8 poison>, [[X:%.*]]
 ; CHECK-NEXT:    [[AND:%.*]] = and <2 x i8> [[SUB]], <i8 120, i8 120>
 ; CHECK-NEXT:    [[R:%.*]] = sub nsw <2 x i8> <i8 77, i8 77>, [[AND]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
-  %sub = sub nuw <2 x i8> <i8 71, i8 undef>, %x
+  %sub = sub nuw <2 x i8> <i8 71, i8 poison>, %x
   %and = and <2 x i8> %sub, <i8 120, i8 120>
   %r = sub <2 x i8>  <i8 77, i8 77>, %and
   ret <2 x i8> %r
@@ -2366,12 +2366,12 @@ define <2 x i8> @sub_to_and_vector2(<2 x i8> %x) {
 define <2 x i8> @sub_to_and_vector3(<2 x i8> %x) {
 ; CHECK-LABEL: @sub_to_and_vector3(
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nuw <2 x i8> <i8 71, i8 71>, [[X:%.*]]
-; CHECK-NEXT:    [[AND:%.*]] = and <2 x i8> [[SUB]], <i8 120, i8 undef>
+; CHECK-NEXT:    [[AND:%.*]] = and <2 x i8> [[SUB]], <i8 120, i8 poison>
 ; CHECK-NEXT:    [[R:%.*]] = sub nsw <2 x i8> <i8 44, i8 44>, [[AND]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %sub = sub nuw <2 x i8> <i8 71, i8 71>, %x
-  %and = and <2 x i8> %sub, <i8 120, i8 undef>
+  %and = and <2 x i8> %sub, <i8 120, i8 poison>
   %r = sub <2 x i8>  <i8 44, i8 44>, %and
   ret <2 x i8> %r
 }
@@ -2381,12 +2381,12 @@ define <2 x i8> @sub_to_and_vector4(<2 x i8> %x) {
 ; CHECK-LABEL: @sub_to_and_vector4(
 ; CHECK-NEXT:    [[SUB:%.*]] = sub nuw <2 x i8> <i8 71, i8 71>, [[X:%.*]]
 ; CHECK-NEXT:    [[AND:%.*]] = and <2 x i8> [[SUB]], <i8 120, i8 120>
-; CHECK-NEXT:    [[R:%.*]] = sub <2 x i8> <i8 88, i8 undef>, [[AND]]
+; CHECK-NEXT:    [[R:%.*]] = sub nsw <2 x i8> <i8 88, i8 poison>, [[AND]]
 ; CHECK-NEXT:    ret <2 x i8> [[R]]
 ;
   %sub = sub nuw <2 x i8> <i8 71, i8 71>, %x
   %and = and <2 x i8> %sub, <i8 120, i8 120>
-  %r = sub <2 x i8>  <i8 88, i8 undef>, %and
+  %r = sub <2 x i8>  <i8 88, i8 poison>, %and
   ret <2 x i8> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
index 4c857125365a9..063006ba5eea8 100644
--- a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
@@ -49,15 +49,15 @@ define <2 x i64> @test1_vec_nonuniform(<2 x i64> %a) {
   ret <2 x i64> %d
 }
 
-define <2 x i64> @test1_vec_undef(<2 x i64> %a) {
-; CHECK-LABEL: @test1_vec_undef(
+define <2 x i64> @test1_vec_poison(<2 x i64> %a) {
+; CHECK-LABEL: @test1_vec_poison(
 ; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[D:%.*]] = and <2 x i64> [[A]], <i64 15, i64 0>
+; CHECK-NEXT:    [[D:%.*]] = and <2 x i64> [[A]], <i64 15, i64 poison>
 ; CHECK-NEXT:    call void @use_vec(<2 x i32> [[B]])
 ; CHECK-NEXT:    ret <2 x i64> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = and <2 x i32> %b, <i32 15, i32 undef>
+  %c = and <2 x i32> %b, <i32 15, i32 poison>
   %d = zext <2 x i32> %c to <2 x i64>
   call void @use_vec(<2 x i32> %b)
   ret <2 x i64> %d
@@ -111,17 +111,17 @@ define <2 x i64> @test2_vec_nonuniform(<2 x i64> %a) {
   ret <2 x i64> %d
 }
 
-define <2 x i64> @test2_vec_undef(<2 x i64> %a) {
-; CHECK-LABEL: @test2_vec_undef(
+define <2 x i64> @test2_vec_poison(<2 x i64> %a) {
+; CHECK-LABEL: @test2_vec_poison(
 ; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i64> [[A]], <i64 36, i64 undef>
-; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i64> [[D1]], <i64 36, i64 undef>
+; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i64> [[A]], <i64 36, i64 poison>
+; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i64> [[D1]], <i64 36, i64 poison>
 ; CHECK-NEXT:    call void @use_vec(<2 x i32> [[B]])
 ; CHECK-NEXT:    ret <2 x i64> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = shl <2 x i32> %b, <i32 4, i32 undef>
-  %q = ashr <2 x i32> %c, <i32 4, i32 undef>
+  %c = shl <2 x i32> %b, <i32 4, i32 poison>
+  %q = ashr <2 x i32> %c, <i32 4, i32 poison>
   %d = sext <2 x i32> %q to <2 x i64>
   call void @use_vec(<2 x i32> %b)
   ret <2 x i64> %d
@@ -300,18 +300,17 @@ define <2 x i64> @test8_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test8_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test8_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = shl <2 x i128> [[D]], <i128 32, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = or <2 x i128> [[E]], [[C]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test8_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test8_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[E:%.*]] = shl nuw <2 x i64> [[D]], <i64 32, i64 poison>
+; CHECK-NEXT:    [[G:%.*]] = or disjoint <2 x i64> [[E]], [[C]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = shl <2 x i128> %D, <i128 32, i128 undef>
+  %E = shl <2 x i128> %D, <i128 32, i128 poison>
   %F = or <2 x i128> %E, %C
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -388,18 +387,17 @@ define <2 x i64> @test11_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test11_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test11_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = shl <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test11_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test11_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = shl nuw nsw <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = shl <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -453,18 +451,17 @@ define <2 x i64> @test12_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test12_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test12_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = lshr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc nuw nsw <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test12_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test12_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = lshr <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = lshr <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -518,18 +515,17 @@ define <2 x i64> @test13_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test13_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test13_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = sext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = ashr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc nsw <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test13_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test13_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = sext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = ashr <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = sext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = ashr <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -766,13 +762,13 @@ define <2 x i32> @trunc_shl_v2i32_v2i64_uniform(<2 x i64> %val) {
   ret <2 x i32> %trunc
 }
 
-define <2 x i32> @trunc_shl_v2i32_v2i64_undef(<2 x i64> %val) {
-; CHECK-LABEL: @trunc_shl_v2i32_v2i64_undef(
+define <2 x i32> @trunc_shl_v2i32_v2i64_poison(<2 x i64> %val) {
+; CHECK-LABEL: @trunc_shl_v2i32_v2i64_poison(
 ; CHECK-NEXT:    [[VAL_TR:%.*]] = trunc <2 x i64> [[VAL:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[TRUNC:%.*]] = shl <2 x i32> [[VAL_TR]], <i32 31, i32 undef>
+; CHECK-NEXT:    [[TRUNC:%.*]] = shl <2 x i32> [[VAL_TR]], <i32 31, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[TRUNC]]
 ;
-  %shl = shl <2 x i64> %val, <i64 31, i64 undef>
+  %shl = shl <2 x i64> %val, <i64 31, i64 poison>
   %trunc = trunc <2 x i64> %shl to <2 x i32>
   ret <2 x i32> %trunc
 }
@@ -917,7 +913,7 @@ define <4 x i8> @wide_shuf(<4 x i32> %x) {
   ret <4 x i8> %trunc
 }
 
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
 
 define <4 x i8> @wide_splat1(<4 x i32> %x) {
 ; CHECK-LABEL: @wide_splat1(
@@ -931,7 +927,7 @@ define <4 x i8> @wide_splat1(<4 x i32> %x) {
 }
 
 ; Test weird types.
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
 
 define <3 x i31> @wide_splat2(<3 x i33> %x) {
 ; CHECK-LABEL: @wide_splat2(
@@ -945,8 +941,8 @@ define <3 x i31> @wide_splat2(<3 x i33> %x) {
 }
 
 ; FIXME:
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
-; A mask with undef elements should still be considered a splat mask.
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
+; A mask with poison elements should still be considered a splat mask.
 
 define <3 x i31> @wide_splat3(<3 x i33> %x) {
 ; CHECK-LABEL: @wide_splat3(
@@ -954,7 +950,7 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) {
 ; CHECK-NEXT:    [[TRUNC:%.*]] = trunc <3 x i33> [[SHUF]] to <3 x i31>
 ; CHECK-NEXT:    ret <3 x i31> [[TRUNC]]
 ;
-  %shuf = shufflevector <3 x i33> %x, <3 x i33> poison, <3 x i32> <i32 undef, i32 1, i32 1>
+  %shuf = shufflevector <3 x i33> %x, <3 x i33> poison, <3 x i32> <i32 poison, i32 1, i32 1>
   %trunc = trunc <3 x i33> %shuf to <3 x i31>
   ret <3 x i31> %trunc
 }
diff --git a/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll b/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
index 2c5f428cf98de..c50a3d06d24b9 100644
--- a/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-shift-trunc.ll
@@ -56,14 +56,14 @@ define <2 x i8> @trunc_lshr_trunc_nonuniform(<2 x i64> %a) {
   ret <2 x i8> %d
 }
 
-define <2 x i8> @trunc_lshr_trunc_uniform_undef(<2 x i64> %a) {
-; CHECK-LABEL: @trunc_lshr_trunc_uniform_undef(
-; CHECK-NEXT:    [[C1:%.*]] = lshr <2 x i64> [[A:%.*]], <i64 24, i64 undef>
+define <2 x i8> @trunc_lshr_trunc_uniform_poison(<2 x i64> %a) {
+; CHECK-LABEL: @trunc_lshr_trunc_uniform_poison(
+; CHECK-NEXT:    [[C1:%.*]] = lshr <2 x i64> [[A:%.*]], <i64 24, i64 poison>
 ; CHECK-NEXT:    [[D:%.*]] = trunc <2 x i64> [[C1]] to <2 x i8>
 ; CHECK-NEXT:    ret <2 x i8> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = lshr <2 x i32> %b, <i32 24, i32 undef>
+  %c = lshr <2 x i32> %b, <i32 24, i32 poison>
   %d = trunc <2 x i32> %c to <2 x i8>
   ret <2 x i8> %d
 }
@@ -142,14 +142,14 @@ define <2 x i8> @trunc_ashr_trunc_nonuniform(<2 x i64> %a) {
   ret <2 x i8> %d
 }
 
-define <2 x i8> @trunc_ashr_trunc_uniform_undef(<2 x i64> %a) {
-; CHECK-LABEL: @trunc_ashr_trunc_uniform_undef(
-; CHECK-NEXT:    [[C1:%.*]] = ashr <2 x i64> [[A:%.*]], <i64 8, i64 undef>
+define <2 x i8> @trunc_ashr_trunc_uniform_poison(<2 x i64> %a) {
+; CHECK-LABEL: @trunc_ashr_trunc_uniform_poison(
+; CHECK-NEXT:    [[C1:%.*]] = ashr <2 x i64> [[A:%.*]], <i64 8, i64 poison>
 ; CHECK-NEXT:    [[D:%.*]] = trunc <2 x i64> [[C1]] to <2 x i8>
 ; CHECK-NEXT:    ret <2 x i8> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = ashr <2 x i32> %b, <i32 8, i32 undef>
+  %c = ashr <2 x i32> %b, <i32 8, i32 poison>
   %d = trunc <2 x i32> %c to <2 x i8>
   ret <2 x i8> %d
 }
diff --git a/llvm/test/Transforms/InstCombine/trunc.ll b/llvm/test/Transforms/InstCombine/trunc.ll
index c77d7269f2cf7..e59b2bea6684c 100644
--- a/llvm/test/Transforms/InstCombine/trunc.ll
+++ b/llvm/test/Transforms/InstCombine/trunc.ll
@@ -49,15 +49,15 @@ define <2 x i64> @test1_vec_nonuniform(<2 x i64> %a) {
   ret <2 x i64> %d
 }
 
-define <2 x i64> @test1_vec_undef(<2 x i64> %a) {
-; CHECK-LABEL: @test1_vec_undef(
+define <2 x i64> @test1_vec_poison(<2 x i64> %a) {
+; CHECK-LABEL: @test1_vec_poison(
 ; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[D:%.*]] = and <2 x i64> [[A]], <i64 15, i64 0>
+; CHECK-NEXT:    [[D:%.*]] = and <2 x i64> [[A]], <i64 15, i64 poison>
 ; CHECK-NEXT:    call void @use_vec(<2 x i32> [[B]])
 ; CHECK-NEXT:    ret <2 x i64> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = and <2 x i32> %b, <i32 15, i32 undef>
+  %c = and <2 x i32> %b, <i32 15, i32 poison>
   %d = zext <2 x i32> %c to <2 x i64>
   call void @use_vec(<2 x i32> %b)
   ret <2 x i64> %d
@@ -111,17 +111,17 @@ define <2 x i64> @test2_vec_nonuniform(<2 x i64> %a) {
   ret <2 x i64> %d
 }
 
-define <2 x i64> @test2_vec_undef(<2 x i64> %a) {
-; CHECK-LABEL: @test2_vec_undef(
+define <2 x i64> @test2_vec_poison(<2 x i64> %a) {
+; CHECK-LABEL: @test2_vec_poison(
 ; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i64> [[A]], <i64 36, i64 undef>
-; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i64> [[D1]], <i64 36, i64 undef>
+; CHECK-NEXT:    [[D1:%.*]] = shl <2 x i64> [[A]], <i64 36, i64 poison>
+; CHECK-NEXT:    [[D:%.*]] = ashr exact <2 x i64> [[D1]], <i64 36, i64 poison>
 ; CHECK-NEXT:    call void @use_vec(<2 x i32> [[B]])
 ; CHECK-NEXT:    ret <2 x i64> [[D]]
 ;
   %b = trunc <2 x i64> %a to <2 x i32>
-  %c = shl <2 x i32> %b, <i32 4, i32 undef>
-  %q = ashr <2 x i32> %c, <i32 4, i32 undef>
+  %c = shl <2 x i32> %b, <i32 4, i32 poison>
+  %q = ashr <2 x i32> %c, <i32 4, i32 poison>
   %d = sext <2 x i32> %q to <2 x i64>
   call void @use_vec(<2 x i32> %b)
   ret <2 x i64> %d
@@ -300,18 +300,17 @@ define <2 x i64> @test8_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test8_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test8_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = shl <2 x i128> [[D]], <i128 32, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = or <2 x i128> [[E]], [[C]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test8_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test8_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[E:%.*]] = shl nuw <2 x i64> [[D]], <i64 32, i64 poison>
+; CHECK-NEXT:    [[G:%.*]] = or disjoint <2 x i64> [[E]], [[C]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = shl <2 x i128> %D, <i128 32, i128 undef>
+  %E = shl <2 x i128> %D, <i128 32, i128 poison>
   %F = or <2 x i128> %E, %C
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -388,18 +387,17 @@ define <2 x i64> @test11_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test11_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test11_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = shl <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test11_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test11_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = shl nuw nsw <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = shl <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -453,18 +451,17 @@ define <2 x i64> @test12_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test12_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test12_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = lshr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc nuw nsw <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test12_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test12_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = zext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = lshr <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = zext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = lshr <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -518,18 +515,17 @@ define <2 x i64> @test13_vec_nonuniform(<2 x i32> %A, <2 x i32> %B) {
   ret <2 x i64> %G
 }
 
-define <2 x i64> @test13_vec_undef(<2 x i32> %A, <2 x i32> %B) {
-; CHECK-LABEL: @test13_vec_undef(
-; CHECK-NEXT:    [[C:%.*]] = sext <2 x i32> [[A:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[D:%.*]] = zext <2 x i32> [[B:%.*]] to <2 x i128>
-; CHECK-NEXT:    [[E:%.*]] = and <2 x i128> [[D]], <i128 31, i128 undef>
-; CHECK-NEXT:    [[F:%.*]] = ashr <2 x i128> [[C]], [[E]]
-; CHECK-NEXT:    [[G:%.*]] = trunc nsw <2 x i128> [[F]] to <2 x i64>
+define <2 x i64> @test13_vec_poison(<2 x i32> %A, <2 x i32> %B) {
+; CHECK-LABEL: @test13_vec_poison(
+; CHECK-NEXT:    [[C:%.*]] = sext <2 x i32> [[A:%.*]] to <2 x i64>
+; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i32> [[B:%.*]], <i32 31, i32 poison>
+; CHECK-NEXT:    [[E:%.*]] = zext nneg <2 x i32> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[G:%.*]] = ashr <2 x i64> [[C]], [[E]]
 ; CHECK-NEXT:    ret <2 x i64> [[G]]
 ;
   %C = sext <2 x i32> %A to <2 x i128>
   %D = zext <2 x i32> %B to <2 x i128>
-  %E = and <2 x i128> %D, <i128 31, i128 undef>
+  %E = and <2 x i128> %D, <i128 31, i128 poison>
   %F = ashr <2 x i128> %C, %E
   %G = trunc <2 x i128> %F to <2 x i64>
   ret <2 x i64> %G
@@ -766,13 +762,13 @@ define <2 x i32> @trunc_shl_v2i32_v2i64_uniform(<2 x i64> %val) {
   ret <2 x i32> %trunc
 }
 
-define <2 x i32> @trunc_shl_v2i32_v2i64_undef(<2 x i64> %val) {
-; CHECK-LABEL: @trunc_shl_v2i32_v2i64_undef(
+define <2 x i32> @trunc_shl_v2i32_v2i64_poison(<2 x i64> %val) {
+; CHECK-LABEL: @trunc_shl_v2i32_v2i64_poison(
 ; CHECK-NEXT:    [[VAL_TR:%.*]] = trunc <2 x i64> [[VAL:%.*]] to <2 x i32>
-; CHECK-NEXT:    [[TRUNC:%.*]] = shl <2 x i32> [[VAL_TR]], <i32 31, i32 undef>
+; CHECK-NEXT:    [[TRUNC:%.*]] = shl <2 x i32> [[VAL_TR]], <i32 31, i32 poison>
 ; CHECK-NEXT:    ret <2 x i32> [[TRUNC]]
 ;
-  %shl = shl <2 x i64> %val, <i64 31, i64 undef>
+  %shl = shl <2 x i64> %val, <i64 31, i64 poison>
   %trunc = trunc <2 x i64> %shl to <2 x i32>
   ret <2 x i32> %trunc
 }
@@ -917,7 +913,7 @@ define <4 x i8> @wide_shuf(<4 x i32> %x) {
   ret <4 x i8> %trunc
 }
 
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
 
 define <4 x i8> @wide_splat1(<4 x i32> %x) {
 ; CHECK-LABEL: @wide_splat1(
@@ -925,13 +921,13 @@ define <4 x i8> @wide_splat1(<4 x i32> %x) {
 ; CHECK-NEXT:    [[TRUNC:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
 ; CHECK-NEXT:    ret <4 x i8> [[TRUNC]]
 ;
-  %shuf = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+  %shuf = shufflevector <4 x i32> %x, <4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
   %trunc = trunc <4 x i32> %shuf to <4 x i8>
   ret <4 x i8> %trunc
 }
 
 ; Test weird types.
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
 
 define <3 x i31> @wide_splat2(<3 x i33> %x) {
 ; CHECK-LABEL: @wide_splat2(
@@ -939,14 +935,14 @@ define <3 x i31> @wide_splat2(<3 x i33> %x) {
 ; CHECK-NEXT:    [[TRUNC:%.*]] = shufflevector <3 x i31> [[TMP1]], <3 x i31> poison, <3 x i32> <i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    ret <3 x i31> [[TRUNC]]
 ;
-  %shuf = shufflevector <3 x i33> %x, <3 x i33> undef, <3 x i32> <i32 1, i32 1, i32 1>
+  %shuf = shufflevector <3 x i33> %x, <3 x i33> poison, <3 x i32> <i32 1, i32 1, i32 1>
   %trunc = trunc <3 x i33> %shuf to <3 x i31>
   ret <3 x i31> %trunc
 }
 
 ; FIXME:
-; trunc (shuffle X, undef, SplatMask) --> shuffle (trunc X), undef, SplatMask
-; A mask with undef elements should still be considered a splat mask.
+; trunc (shuffle X, poison, SplatMask) --> shuffle (trunc X), poison, SplatMask
+; A mask with poison elements should still be considered a splat mask.
 
 define <3 x i31> @wide_splat3(<3 x i33> %x) {
 ; CHECK-LABEL: @wide_splat3(
@@ -954,7 +950,7 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) {
 ; CHECK-NEXT:    [[TRUNC:%.*]] = trunc <3 x i33> [[SHUF]] to <3 x i31>
 ; CHECK-NEXT:    ret <3 x i31> [[TRUNC]]
 ;
-  %shuf = shufflevector <3 x i33> %x, <3 x i33> undef, <3 x i32> <i32 undef, i32 1, i32 1>
+  %shuf = shufflevector <3 x i33> %x, <3 x i33> poison, <3 x i32> <i32 poison, i32 1, i32 1>
   %trunc = trunc <3 x i33> %shuf to <3 x i31>
   ret <3 x i31> %trunc
 }
diff --git a/llvm/test/Transforms/InstCombine/unsigned-mul-lack-of-overflow-check-via-udiv-of-allones.ll b/llvm/test/Transforms/InstCombine/unsigned-mul-lack-of-overflow-check-via-udiv-of-allones.ll
index 1ffcfb4424e31..241d9cbcde338 100644
--- a/llvm/test/Transforms/InstCombine/unsigned-mul-lack-of-overflow-check-via-udiv-of-allones.ll
+++ b/llvm/test/Transforms/InstCombine/unsigned-mul-lack-of-overflow-check-via-udiv-of-allones.ll
@@ -30,14 +30,14 @@ define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @t2_vec_undef(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @t2_vec_undef(
+define <3 x i1> @t2_vec_poison(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @t2_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = call { <3 x i8>, <3 x i1> } @llvm.umul.with.overflow.v3i8(<3 x i8> [[X:%.*]], <3 x i8> [[Y:%.*]])
 ; CHECK-NEXT:    [[MUL_OV:%.*]] = extractvalue { <3 x i8>, <3 x i1> } [[MUL]], 1
 ; CHECK-NEXT:    [[MUL_NOT_OV:%.*]] = xor <3 x i1> [[MUL_OV]], <i1 true, i1 true, i1 true>
 ; CHECK-NEXT:    ret <3 x i1> [[MUL_NOT_OV]]
 ;
-  %t0 = udiv <3 x i8> <i8 -1, i8 undef, i8 -1>, %x
+  %t0 = udiv <3 x i8> <i8 -1, i8 poison, i8 -1>, %x
   %r = icmp uge <3 x i8> %t0, %y
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/unsigned-mul-overflow-check-via-udiv-of-allones.ll b/llvm/test/Transforms/InstCombine/unsigned-mul-overflow-check-via-udiv-of-allones.ll
index 710a09f6e16a1..7eb08bdd6016c 100644
--- a/llvm/test/Transforms/InstCombine/unsigned-mul-overflow-check-via-udiv-of-allones.ll
+++ b/llvm/test/Transforms/InstCombine/unsigned-mul-overflow-check-via-udiv-of-allones.ll
@@ -28,13 +28,13 @@ define <2 x i1> @t1_vec(<2 x i8> %x, <2 x i8> %y) {
   ret <2 x i1> %r
 }
 
-define <3 x i1> @t2_vec_undef(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @t2_vec_undef(
+define <3 x i1> @t2_vec_poison(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @t2_vec_poison(
 ; CHECK-NEXT:    [[MUL:%.*]] = call { <3 x i8>, <3 x i1> } @llvm.umul.with.overflow.v3i8(<3 x i8> [[X:%.*]], <3 x i8> [[Y:%.*]])
 ; CHECK-NEXT:    [[MUL_OV:%.*]] = extractvalue { <3 x i8>, <3 x i1> } [[MUL]], 1
 ; CHECK-NEXT:    ret <3 x i1> [[MUL_OV]]
 ;
-  %t0 = udiv <3 x i8> <i8 -1, i8 undef, i8 -1>, %x
+  %t0 = udiv <3 x i8> <i8 -1, i8 poison, i8 -1>, %x
   %r = icmp ult <3 x i8> %t0, %y
   ret <3 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstCombine/variable-signext-of-variable-high-bit-extraction.ll b/llvm/test/Transforms/InstCombine/variable-signext-of-variable-high-bit-extraction.ll
index adacf3ce99b2f..262942aa1219b 100644
--- a/llvm/test/Transforms/InstCombine/variable-signext-of-variable-high-bit-extraction.ll
+++ b/llvm/test/Transforms/InstCombine/variable-signext-of-variable-high-bit-extraction.ll
@@ -203,20 +203,20 @@ define <2 x i32> @t4_vec(<2 x i64> %data, <2 x i32> %nbits) {
   ret <2 x i32> %signextended
 }
 
-define <3 x i32> @t5_vec_undef(<3 x i64> %data, <3 x i32> %nbits) {
-; CHECK-LABEL: @t5_vec_undef(
-; CHECK-NEXT:    [[SKIP_HIGH:%.*]] = sub <3 x i32> <i32 64, i32 64, i32 undef>, [[NBITS:%.*]]
+define <3 x i32> @t5_vec_poison(<3 x i64> %data, <3 x i32> %nbits) {
+; CHECK-LABEL: @t5_vec_poison(
+; CHECK-NEXT:    [[SKIP_HIGH:%.*]] = sub <3 x i32> <i32 64, i32 64, i32 poison>, [[NBITS:%.*]]
 ; CHECK-NEXT:    [[SKIP_HIGH_WIDE:%.*]] = zext nneg <3 x i32> [[SKIP_HIGH]] to <3 x i64>
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <3 x i64> [[DATA:%.*]], [[SKIP_HIGH_WIDE]]
 ; CHECK-NEXT:    [[SIGNEXTENDED:%.*]] = trunc <3 x i64> [[TMP1]] to <3 x i32>
 ; CHECK-NEXT:    ret <3 x i32> [[SIGNEXTENDED]]
 ;
-  %skip_high = sub <3 x i32> <i32 64, i32 64, i32 undef>, %nbits
+  %skip_high = sub <3 x i32> <i32 64, i32 64, i32 poison>, %nbits
   %skip_high_wide = zext <3 x i32> %skip_high to <3 x i64>
   %extracted = lshr <3 x i64> %data, %skip_high_wide
   %extracted_narrow = trunc <3 x i64> %extracted to <3 x i32>
-  %num_high_bits_to_smear_narrow0 = sub <3 x i32> <i32 32, i32 32, i32 undef>, %nbits
-  %num_high_bits_to_smear_narrow1 = sub <3 x i32> <i32 undef, i32 32, i32 32>, %nbits
+  %num_high_bits_to_smear_narrow0 = sub <3 x i32> <i32 32, i32 32, i32 poison>, %nbits
+  %num_high_bits_to_smear_narrow1 = sub <3 x i32> <i32 poison, i32 32, i32 32>, %nbits
   %signbit_positioned = shl <3 x i32> %extracted_narrow, %num_high_bits_to_smear_narrow0
   %signextended = ashr <3 x i32> %signbit_positioned, %num_high_bits_to_smear_narrow1
   ret <3 x i32> %signextended
diff --git a/llvm/test/Transforms/InstCombine/vec_sext.ll b/llvm/test/Transforms/InstCombine/vec_sext.ll
index a880d5e562725..9f5f957f49445 100644
--- a/llvm/test/Transforms/InstCombine/vec_sext.ll
+++ b/llvm/test/Transforms/InstCombine/vec_sext.ll
@@ -42,24 +42,24 @@ define <4 x i32> @vec_select_alternate_sign_bit_test(<4 x i32> %a, <4 x i32> %b)
   ret <4 x i32> %cond
 }
 
-define <2 x i32> @is_negative_undef_elt(<2 x i32> %a) {
-; CHECK-LABEL: @is_negative_undef_elt(
+define <2 x i32> @is_negative_poison_elt(<2 x i32> %a) {
+; CHECK-LABEL: @is_negative_poison_elt(
 ; CHECK-NEXT:    [[A_LOBIT:%.*]] = ashr <2 x i32> [[A:%.*]], <i32 31, i32 31>
 ; CHECK-NEXT:    ret <2 x i32> [[A_LOBIT]]
 ;
-  %cmp = icmp slt <2 x i32> %a, <i32 0, i32 undef>
+  %cmp = icmp slt <2 x i32> %a, <i32 0, i32 poison>
   %sext = sext <2 x i1> %cmp to <2 x i32>
   ret <2 x i32> %sext
 
 }
 
-define <2 x i32> @is_positive_undef_elt(<2 x i32> %a) {
-; CHECK-LABEL: @is_positive_undef_elt(
-; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <2 x i32> [[A:%.*]], <i32 undef, i32 -1>
+define <2 x i32> @is_positive_poison_elt(<2 x i32> %a) {
+; CHECK-LABEL: @is_positive_poison_elt(
+; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <2 x i32> [[A:%.*]], <i32 poison, i32 -1>
 ; CHECK-NEXT:    [[SEXT:%.*]] = sext <2 x i1> [[CMP]] to <2 x i32>
 ; CHECK-NEXT:    ret <2 x i32> [[SEXT]]
 ;
-  %cmp = icmp sgt <2 x i32> %a, <i32 undef, i32 -1>
+  %cmp = icmp sgt <2 x i32> %a, <i32 poison, i32 -1>
   %sext = sext <2 x i1> %cmp to <2 x i32>
   ret <2 x i32> %sext
 }
diff --git a/llvm/test/Transforms/InstCombine/vector-casts-inseltpoison.ll b/llvm/test/Transforms/InstCombine/vector-casts-inseltpoison.ll
index cf1b72fbcf3e1..a87364600ba30 100644
--- a/llvm/test/Transforms/InstCombine/vector-casts-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/vector-casts-inseltpoison.ll
@@ -26,26 +26,26 @@ define <2 x i1> @and_cmp_is_trunc(<2 x i64> %a) {
 
 ; This is trunc.
 
-define <2 x i1> @and_cmp_is_trunc_even_with_undef_elt(<2 x i64> %a) {
-; CHECK-LABEL: @and_cmp_is_trunc_even_with_undef_elt(
+define <2 x i1> @and_cmp_is_trunc_even_with_poison_elt(<2 x i64> %a) {
+; CHECK-LABEL: @and_cmp_is_trunc_even_with_poison_elt(
 ; CHECK-NEXT:    [[R:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %t = and <2 x i64> %a, <i64 undef, i64 1>
+  %t = and <2 x i64> %a, <i64 poison, i64 1>
   %r = icmp ne <2 x i64> %t, zeroinitializer
   ret <2 x i1> %r
 }
 
-; TODO: This could be just 1 instruction (trunc), but our undef matching is incomplete.
+; TODO: This could be just 1 instruction (trunc), but our poison matching is incomplete.
 
-define <2 x i1> @and_cmp_is_trunc_even_with_undef_elts(<2 x i64> %a) {
-; CHECK-LABEL: @and_cmp_is_trunc_even_with_undef_elts(
-; CHECK-NEXT:    [[T:%.*]] = and <2 x i64> [[A:%.*]], <i64 undef, i64 1>
-; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i64> [[T]], <i64 undef, i64 0>
+define <2 x i1> @and_cmp_is_trunc_even_with_poison_elts(<2 x i64> %a) {
+; CHECK-LABEL: @and_cmp_is_trunc_even_with_poison_elts(
+; CHECK-NEXT:    [[T:%.*]] = and <2 x i64> [[A:%.*]], <i64 poison, i64 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i64> [[T]], <i64 poison, i64 0>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %t = and <2 x i64> %a, <i64 undef, i64 1>
-  %r = icmp ne <2 x i64> %t, <i64 undef, i64 0>
+  %t = and <2 x i64> %a, <i64 poison, i64 1>
+  %r = icmp ne <2 x i64> %t, <i64 poison, i64 0>
   ret <2 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/vector-casts.ll b/llvm/test/Transforms/InstCombine/vector-casts.ll
index 281fc5f6011ea..fd2a4ffdfb709 100644
--- a/llvm/test/Transforms/InstCombine/vector-casts.ll
+++ b/llvm/test/Transforms/InstCombine/vector-casts.ll
@@ -26,26 +26,26 @@ define <2 x i1> @and_cmp_is_trunc(<2 x i64> %a) {
 
 ; This is trunc.
 
-define <2 x i1> @and_cmp_is_trunc_even_with_undef_elt(<2 x i64> %a) {
-; CHECK-LABEL: @and_cmp_is_trunc_even_with_undef_elt(
+define <2 x i1> @and_cmp_is_trunc_even_with_poison_elt(<2 x i64> %a) {
+; CHECK-LABEL: @and_cmp_is_trunc_even_with_poison_elt(
 ; CHECK-NEXT:    [[R:%.*]] = trunc <2 x i64> [[A:%.*]] to <2 x i1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %t = and <2 x i64> %a, <i64 undef, i64 1>
+  %t = and <2 x i64> %a, <i64 poison, i64 1>
   %r = icmp ne <2 x i64> %t, zeroinitializer
   ret <2 x i1> %r
 }
 
-; TODO: This could be just 1 instruction (trunc), but our undef matching is incomplete.
+; TODO: This could be just 1 instruction (trunc), but our poison matching is incomplete.
 
-define <2 x i1> @and_cmp_is_trunc_even_with_undef_elts(<2 x i64> %a) {
-; CHECK-LABEL: @and_cmp_is_trunc_even_with_undef_elts(
-; CHECK-NEXT:    [[T:%.*]] = and <2 x i64> [[A:%.*]], <i64 undef, i64 1>
-; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i64> [[T]], <i64 undef, i64 0>
+define <2 x i1> @and_cmp_is_trunc_even_with_poison_elts(<2 x i64> %a) {
+; CHECK-LABEL: @and_cmp_is_trunc_even_with_poison_elts(
+; CHECK-NEXT:    [[T:%.*]] = and <2 x i64> [[A:%.*]], <i64 poison, i64 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i64> [[T]], <i64 poison, i64 0>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %t = and <2 x i64> %a, <i64 undef, i64 1>
-  %r = icmp ne <2 x i64> %t, <i64 undef, i64 0>
+  %t = and <2 x i64> %a, <i64 poison, i64 1>
+  %r = icmp ne <2 x i64> %t, <i64 poison, i64 0>
   ret <2 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/vector-urem.ll b/llvm/test/Transforms/InstCombine/vector-urem.ll
index d5c77470a20f8..627789a03ef6c 100644
--- a/llvm/test/Transforms/InstCombine/vector-urem.ll
+++ b/llvm/test/Transforms/InstCombine/vector-urem.ll
@@ -19,11 +19,11 @@ define <4 x i32> @test_v4i32_const_pow2(<4 x i32> %a0) {
   ret <4 x i32> %1
 }
 
-define <4 x i32> @test_v4i32_const_pow2_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_const_pow2_undef(
+define <4 x i32> @test_v4i32_const_pow2_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_const_pow2_poison(
 ; CHECK-NEXT:    ret <4 x i32> poison
 ;
-  %1 = urem <4 x i32> %a0, <i32 1, i32 2, i32 4, i32 undef>
+  %1 = urem <4 x i32> %a0, <i32 1, i32 2, i32 4, i32 poison>
   ret <4 x i32> %1
 }
 
@@ -37,13 +37,13 @@ define <4 x i32> @test_v4i32_one(<4 x i32> %a0) {
   ret <4 x i32> %1
 }
 
-define <4 x i32> @test_v4i32_one_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_one_undef(
+define <4 x i32> @test_v4i32_one_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_one_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne <4 x i32> [[A0:%.*]], <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    [[TMP2:%.*]] = zext <4 x i1> [[TMP1]] to <4 x i32>
 ; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
-  %1 = urem <4 x i32> <i32 1, i32 1, i32 1, i32 undef>, %a0
+  %1 = urem <4 x i32> <i32 1, i32 1, i32 1, i32 poison>, %a0
   ret <4 x i32> %1
 }
 
@@ -71,10 +71,10 @@ define <4 x i32> @test_v4i32_negconst(<4 x i32> %a0) {
   ret <4 x i32> %1
 }
 
-define <4 x i32> @test_v4i32_negconst_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_negconst_undef(
+define <4 x i32> @test_v4i32_negconst_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_negconst_poison(
 ; CHECK-NEXT:    ret <4 x i32> poison
 ;
-  %1 = urem <4 x i32> %a0, <i32 -3, i32 -5, i32 -7, i32 undef>
+  %1 = urem <4 x i32> %a0, <i32 -3, i32 -5, i32 -7, i32 poison>
   ret <4 x i32> %1
 }
diff --git a/llvm/test/Transforms/InstCombine/vector-xor.ll b/llvm/test/Transforms/InstCombine/vector-xor.ll
index 171dd6e35b4e1..ee593b5d15e8e 100644
--- a/llvm/test/Transforms/InstCombine/vector-xor.ll
+++ b/llvm/test/Transforms/InstCombine/vector-xor.ll
@@ -53,14 +53,14 @@ define <4 x i32> @test_v4i32_xor_bswap_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_xor_bswap_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_xor_bswap_const_undef(
+define <4 x i32> @test_v4i32_xor_bswap_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_xor_bswap_const_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> [[A0:%.*]])
-; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 undef, i32 0, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 poison, i32 0, i32 2, i32 3>
 ; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
   %1 = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> %a0)
-  %2 = xor  <4 x i32> %1, <i32 undef, i32 0, i32 2, i32 3>
+  %2 = xor  <4 x i32> %1, <i32 poison, i32 0, i32 2, i32 3>
   ret <4 x i32> %2
 }
 
@@ -105,14 +105,14 @@ define <4 x i32> @test_v4i32_not_ashr_not(<4 x i32> %x, <4 x i32> %y) {
   ret <4 x i32> %3
 }
 
-define <4 x i32> @test_v4i32_not_ashr_not_undef(<4 x i32> %x, <4 x i32> %y) {
-; CHECK-LABEL: @test_v4i32_not_ashr_not_undef(
+define <4 x i32> @test_v4i32_not_ashr_not_poison(<4 x i32> %x, <4 x i32> %y) {
+; CHECK-LABEL: @test_v4i32_not_ashr_not_poison(
 ; CHECK-NEXT:    [[DOTNOT:%.*]] = ashr <4 x i32> [[X:%.*]], [[Y:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[DOTNOT]]
 ;
-  %1 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 undef>, %x
+  %1 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 poison>, %x
   %2 = ashr <4 x i32> %1, %y
-  %3 = xor  <4 x i32> <i32 -1, i32 -1, i32 undef, i32 -1>, %2
+  %3 = xor  <4 x i32> <i32 -1, i32 -1, i32 poison, i32 -1>, %2
   ret <4 x i32> %3
 }
 
@@ -138,13 +138,13 @@ define <4 x i32> @test_v4i32_not_ashr_negative_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_not_ashr_negative_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_not_ashr_negative_const_undef(
+define <4 x i32> @test_v4i32_not_ashr_negative_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_not_ashr_negative_const_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <4 x i32> <i32 2, i32 4, i32 0, i32 8>, [[A0:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 ;
-  %1 = ashr <4 x i32> <i32 -3, i32 -5, i32 undef, i32 -9>, %a0
-  %2 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 undef>, %1
+  %1 = ashr <4 x i32> <i32 -3, i32 -5, i32 poison, i32 -9>, %a0
+  %2 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 poison>, %1
   ret <4 x i32> %2
 }
 
@@ -170,13 +170,13 @@ define <4 x i32> @test_v4i32_not_lshr_nonnegative_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_not_lshr_nonnegative_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_not_lshr_nonnegative_const_undef(
+define <4 x i32> @test_v4i32_not_lshr_nonnegative_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_not_lshr_nonnegative_const_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <4 x i32> <i32 -4, i32 -6, i32 -1, i32 -10>, [[A0:%.*]]
 ; CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 ;
-  %1 = lshr <4 x i32> <i32  3, i32  5, i32 undef, i32  9>, %a0
-  %2 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 undef>, %1
+  %1 = lshr <4 x i32> <i32  3, i32  5, i32 poison, i32  9>, %a0
+  %2 = xor  <4 x i32> <i32 -1, i32 -1, i32 -1, i32 poison>, %1
   ret <4 x i32> %2
 }
 
@@ -202,13 +202,13 @@ define <4 x i32> @test_v4i32_not_sub_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_not_sub_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_not_sub_const_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[A0:%.*]], <i32 -4, i32 undef, i32 0, i32 -16>
+define <4 x i32> @test_v4i32_not_sub_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_not_sub_const_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[A0:%.*]], <i32 -4, i32 poison, i32 0, i32 -16>
 ; CHECK-NEXT:    ret <4 x i32> [[TMP1]]
 ;
-  %1 = sub <4 x i32> <i32  3, i32 undef, i32 -1, i32 15>, %a0
-  %2 = xor <4 x i32> <i32 -1, i32 -1, i32 -1, i32 undef>, %1
+  %1 = sub <4 x i32> <i32  3, i32 poison, i32 -1, i32 15>, %a0
+  %2 = xor <4 x i32> <i32 -1, i32 -1, i32 -1, i32 poison>, %1
   ret <4 x i32> %2
 }
 
@@ -235,14 +235,14 @@ define <4 x i32> @test_v4i32_xor_signmask_sub_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_xor_signmask_sub_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_xor_signmask_sub_const_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = sub <4 x i32> <i32 3, i32 undef, i32 -1, i32 15>, [[A0:%.*]]
-; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 undef>
+define <4 x i32> @test_v4i32_xor_signmask_sub_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_xor_signmask_sub_const_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = sub <4 x i32> <i32 3, i32 poison, i32 -1, i32 15>, [[A0:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 poison>
 ; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
-  %1 = sub <4 x i32> <i32  3, i32 undef, i32 -1, i32 15>, %a0
-  %2 = xor <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 undef>, %1
+  %1 = sub <4 x i32> <i32  3, i32 poison, i32 -1, i32 15>, %a0
+  %2 = xor <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 poison>, %1
   ret <4 x i32> %2
 }
 
@@ -269,13 +269,13 @@ define <4 x i32> @test_v4i32_xor_signmask_add_const(<4 x i32> %a0) {
   ret <4 x i32> %2
 }
 
-define <4 x i32> @test_v4i32_xor_signmask_add_const_undef(<4 x i32> %a0) {
-; CHECK-LABEL: @test_v4i32_xor_signmask_add_const_undef(
-; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[A0:%.*]], <i32 3, i32 undef, i32 -1, i32 15>
-; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 undef>
+define <4 x i32> @test_v4i32_xor_signmask_add_const_poison(<4 x i32> %a0) {
+; CHECK-LABEL: @test_v4i32_xor_signmask_add_const_poison(
+; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[A0:%.*]], <i32 3, i32 poison, i32 -1, i32 15>
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 poison>
 ; CHECK-NEXT:    ret <4 x i32> [[TMP2]]
 ;
-  %1 = add <4 x i32> <i32  3, i32 undef, i32 -1, i32 15>, %a0
-  %2 = xor <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 undef>, %1
+  %1 = add <4 x i32> <i32  3, i32 poison, i32 -1, i32 15>, %a0
+  %2 = xor <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 poison>, %1
   ret <4 x i32> %2
 }
diff --git a/llvm/test/Transforms/InstCombine/zext-bool-add-sub.ll b/llvm/test/Transforms/InstCombine/zext-bool-add-sub.ll
index 7fed952a7ff7e..12739b5686a0a 100644
--- a/llvm/test/Transforms/InstCombine/zext-bool-add-sub.ll
+++ b/llvm/test/Transforms/InstCombine/zext-bool-add-sub.ll
@@ -126,13 +126,13 @@ define <2 x i64> @zext_negate_vec(<2 x i1> %A) {
   ret <2 x i64> %sub
 }
 
-define <2 x i64> @zext_negate_vec_undef_elt(<2 x i1> %A) {
-; CHECK-LABEL: @zext_negate_vec_undef_elt(
+define <2 x i64> @zext_negate_vec_poison_elt(<2 x i1> %A) {
+; CHECK-LABEL: @zext_negate_vec_poison_elt(
 ; CHECK-NEXT:    [[EXT_NEG:%.*]] = sext <2 x i1> [[A:%.*]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[EXT_NEG]]
 ;
   %ext = zext <2 x i1> %A to <2 x i64>
-  %sub = sub <2 x i64> <i64 0, i64 undef>, %ext
+  %sub = sub <2 x i64> <i64 0, i64 poison>, %ext
   ret <2 x i64> %sub
 }
 
@@ -169,13 +169,13 @@ define <2 x i64> @zext_sub_const_vec(<2 x i1> %A) {
   ret <2 x i64> %sub
 }
 
-define <2 x i64> @zext_sub_const_vec_undef_elt(<2 x i1> %A) {
-; CHECK-LABEL: @zext_sub_const_vec_undef_elt(
-; CHECK-NEXT:    [[SUB:%.*]] = select <2 x i1> [[A:%.*]], <2 x i64> <i64 41, i64 undef>, <2 x i64> <i64 42, i64 undef>
+define <2 x i64> @zext_sub_const_vec_poison_elt(<2 x i1> %A) {
+; CHECK-LABEL: @zext_sub_const_vec_poison_elt(
+; CHECK-NEXT:    [[SUB:%.*]] = select <2 x i1> [[A:%.*]], <2 x i64> <i64 41, i64 poison>, <2 x i64> <i64 42, i64 poison>
 ; CHECK-NEXT:    ret <2 x i64> [[SUB]]
 ;
   %ext = zext <2 x i1> %A to <2 x i64>
-  %sub = sub <2 x i64> <i64 42, i64 undef>, %ext
+  %sub = sub <2 x i64> <i64 42, i64 poison>, %ext
   ret <2 x i64> %sub
 }
 
@@ -212,13 +212,13 @@ define <2 x i64> @sext_negate_vec(<2 x i1> %A) {
   ret <2 x i64> %sub
 }
 
-define <2 x i64> @sext_negate_vec_undef_elt(<2 x i1> %A) {
-; CHECK-LABEL: @sext_negate_vec_undef_elt(
+define <2 x i64> @sext_negate_vec_poison_elt(<2 x i1> %A) {
+; CHECK-LABEL: @sext_negate_vec_poison_elt(
 ; CHECK-NEXT:    [[EXT_NEG:%.*]] = zext <2 x i1> [[A:%.*]] to <2 x i64>
 ; CHECK-NEXT:    ret <2 x i64> [[EXT_NEG]]
 ;
   %ext = sext <2 x i1> %A to <2 x i64>
-  %sub = sub <2 x i64> <i64 0, i64 undef>, %ext
+  %sub = sub <2 x i64> <i64 0, i64 poison>, %ext
   ret <2 x i64> %sub
 }
 
@@ -255,13 +255,13 @@ define <2 x i64> @sext_sub_const_vec(<2 x i1> %A) {
   ret <2 x i64> %sub
 }
 
-define <2 x i64> @sext_sub_const_vec_undef_elt(<2 x i1> %A) {
-; CHECK-LABEL: @sext_sub_const_vec_undef_elt(
-; CHECK-NEXT:    [[SUB:%.*]] = select <2 x i1> [[A:%.*]], <2 x i64> <i64 undef, i64 43>, <2 x i64> <i64 undef, i64 42>
+define <2 x i64> @sext_sub_const_vec_poison_elt(<2 x i1> %A) {
+; CHECK-LABEL: @sext_sub_const_vec_poison_elt(
+; CHECK-NEXT:    [[SUB:%.*]] = select <2 x i1> [[A:%.*]], <2 x i64> <i64 poison, i64 43>, <2 x i64> <i64 poison, i64 42>
 ; CHECK-NEXT:    ret <2 x i64> [[SUB]]
 ;
   %ext = sext <2 x i1> %A to <2 x i64>
-  %sub = sub <2 x i64> <i64 undef, i64 42>, %ext
+  %sub = sub <2 x i64> <i64 poison, i64 42>, %ext
   ret <2 x i64> %sub
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/AndOrXor.ll b/llvm/test/Transforms/InstSimplify/AndOrXor.ll
index 494b6bcd2b66d..2e3a605224203 100644
--- a/llvm/test/Transforms/InstSimplify/AndOrXor.ll
+++ b/llvm/test/Transforms/InstSimplify/AndOrXor.ll
@@ -12,11 +12,11 @@ define i8 @and0(i8 %x) {
   ret i8 %r
 }
 
-define <2 x i8> @and0_vec_undef_elt(<2 x i8> %x) {
-; CHECK-LABEL: @and0_vec_undef_elt(
+define <2 x i8> @and0_vec_poison_elt(<2 x i8> %x) {
+; CHECK-LABEL: @and0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i8> zeroinitializer
 ;
-  %r = and <2 x i8> %x, <i8 undef, i8 0>
+  %r = and <2 x i8> %x, <i8 poison, i8 0>
   ret <2 x i8> %r
 }
 
@@ -31,14 +31,14 @@ define <2 x i32> @add_nsw_signbit(<2 x i32> %x) {
   ret <2 x i32> %z
 }
 
-; Undef elements in either constant vector are ok.
+; Poison elements in either constant vector are ok.
 
-define <2 x i32> @add_nsw_signbit_undef(<2 x i32> %x) {
-; CHECK-LABEL: @add_nsw_signbit_undef(
+define <2 x i32> @add_nsw_signbit_poison(<2 x i32> %x) {
+; CHECK-LABEL: @add_nsw_signbit_poison(
 ; CHECK-NEXT:    ret <2 x i32> [[X:%.*]]
 ;
-  %y = xor <2 x i32> %x, <i32 undef, i32 -2147483648>
-  %z = add nsw <2 x i32> %y, <i32 -2147483648, i32 undef>
+  %y = xor <2 x i32> %x, <i32 poison, i32 -2147483648>
+  %z = add nsw <2 x i32> %y, <i32 -2147483648, i32 poison>
   ret <2 x i32> %z
 }
 
@@ -53,14 +53,14 @@ define <2 x i5> @add_nuw_signbit(<2 x i5> %x) {
   ret <2 x i5> %z
 }
 
-; Undef elements in either constant vector are ok.
+; Poison elements in either constant vector are ok.
 
-define <2 x i5> @add_nuw_signbit_undef(<2 x i5> %x) {
-; CHECK-LABEL: @add_nuw_signbit_undef(
+define <2 x i5> @add_nuw_signbit_poison(<2 x i5> %x) {
+; CHECK-LABEL: @add_nuw_signbit_poison(
 ; CHECK-NEXT:    ret <2 x i5> [[X:%.*]]
 ;
-  %y = xor <2 x i5> %x, <i5 -16, i5 undef>
-  %z = add nuw <2 x i5> %y, <i5 undef, i5 -16>
+  %y = xor <2 x i5> %x, <i5 -16, i5 poison>
+  %z = add nuw <2 x i5> %y, <i5 poison, i5 -16>
   ret <2 x i5> %z
 }
 
@@ -584,7 +584,7 @@ define <2 x i32> @or_xor_andn_commute2(<2 x i32> %a, <2 x i32> %b) {
 ; CHECK-NEXT:    ret <2 x i32> [[XOR]]
 ;
   %xor = xor <2 x i32> %a, %b
-  %neg = xor <2 x i32> %b, <i32 -1, i32 undef>
+  %neg = xor <2 x i32> %b, <i32 -1, i32 poison>
   %and = and <2 x i32> %a, %neg
   %or = or <2 x i32> %xor, %and
   ret <2 x i32> %or
@@ -708,15 +708,13 @@ define <2 x i32> @or_xorn_and_commute2_undef(<2 x i32> %a, <2 x i32> %b) {
   ret <2 x i32> %or
 }
 
-; TODO: Unlike the above test, this is safe to fold.
+; Unlike the above test, this is safe to fold.
 
 define <2 x i32> @or_xorn_and_commute2_poison(<2 x i32> %a, <2 x i32> %b) {
 ; CHECK-LABEL: @or_xorn_and_commute2_poison(
 ; CHECK-NEXT:    [[NEGA:%.*]] = xor <2 x i32> [[A:%.*]], <i32 poison, i32 -1>
-; CHECK-NEXT:    [[AND:%.*]] = and <2 x i32> [[B:%.*]], [[A]]
-; CHECK-NEXT:    [[XOR:%.*]] = xor <2 x i32> [[B]], [[NEGA]]
-; CHECK-NEXT:    [[OR:%.*]] = or <2 x i32> [[XOR]], [[AND]]
-; CHECK-NEXT:    ret <2 x i32> [[OR]]
+; CHECK-NEXT:    [[XOR:%.*]] = xor <2 x i32> [[B:%.*]], [[NEGA]]
+; CHECK-NEXT:    ret <2 x i32> [[XOR]]
 ;
   %nega = xor <2 x i32> %a, <i32 poison, i32 -1>
   %and = and <2 x i32> %b, %a
diff --git a/llvm/test/Transforms/InstSimplify/call.ll b/llvm/test/Transforms/InstSimplify/call.ll
index 52c207a276046..c6f6b65f89dc2 100644
--- a/llvm/test/Transforms/InstSimplify/call.ll
+++ b/llvm/test/Transforms/InstSimplify/call.ll
@@ -976,7 +976,7 @@ define <2 x i8> @fshr_zero_vec(<2 x i8> %shamt) {
 ; CHECK-LABEL: @fshr_zero_vec(
 ; CHECK-NEXT:    ret <2 x i8> zeroinitializer
 ;
-  %r = call <2 x i8> @llvm.fshr.v2i8(<2 x i8> zeroinitializer, <2 x i8> <i8 0, i8 undef>, <2 x i8> %shamt)
+  %r = call <2 x i8> @llvm.fshr.v2i8(<2 x i8> zeroinitializer, <2 x i8> <i8 0, i8 poison>, <2 x i8> %shamt)
   ret <2 x i8> %r
 }
 
@@ -984,7 +984,7 @@ define <2 x i7> @fshl_ones_vec(<2 x i7> %shamt) {
 ; CHECK-LABEL: @fshl_ones_vec(
 ; CHECK-NEXT:    ret <2 x i7> <i7 -1, i7 -1>
 ;
-  %r = call <2 x i7> @llvm.fshl.v2i7(<2 x i7> <i7 undef, i7 -1>, <2 x i7> <i7 -1, i7 undef>, <2 x i7> %shamt)
+  %r = call <2 x i7> @llvm.fshl.v2i7(<2 x i7> <i7 poison, i7 -1>, <2 x i7> <i7 -1, i7 poison>, <2 x i7> %shamt)
   ret <2 x i7> %r
 }
 
@@ -1466,7 +1466,7 @@ define <3 x i33> @cttz_shl1_vec(<3 x i33> %x) {
 ; CHECK-LABEL: @cttz_shl1_vec(
 ; CHECK-NEXT:    ret <3 x i33> [[X:%.*]]
 ;
-  %s = shl <3 x i33> <i33 1, i33 1, i33 undef>, %x
+  %s = shl <3 x i33> <i33 1, i33 1, i33 poison>, %x
   %r = call <3 x i33> @llvm.cttz.v3i33(<3 x i33> %s, i1 false)
   ret <3 x i33> %r
 }
@@ -1509,7 +1509,7 @@ define <3 x i33> @ctlz_lshr_sign_bit_vec(<3 x i33> %x) {
 ; CHECK-LABEL: @ctlz_lshr_sign_bit_vec(
 ; CHECK-NEXT:    ret <3 x i33> [[X:%.*]]
 ;
-  %s = lshr <3 x i33> <i33 undef, i33 4294967296, i33 4294967296>, %x
+  %s = lshr <3 x i33> <i33 poison, i33 4294967296, i33 4294967296>, %x
   %r = call <3 x i33> @llvm.ctlz.v3i33(<3 x i33> %s, i1 false)
   ret <3 x i33> %r
 }
@@ -1549,7 +1549,7 @@ define <3 x i33> @ctlz_ashr_sign_bit_vec(<3 x i33> %x) {
 ; CHECK-LABEL: @ctlz_ashr_sign_bit_vec(
 ; CHECK-NEXT:    ret <3 x i33> zeroinitializer
 ;
-  %s = ashr <3 x i33> <i33 4294967296, i33 undef, i33 4294967296>, %x
+  %s = ashr <3 x i33> <i33 4294967296, i33 poison, i33 4294967296>, %x
   %r = call <3 x i33> @llvm.ctlz.v3i33(<3 x i33> %s, i1 true)
   ret <3 x i33> %r
 }
diff --git a/llvm/test/Transforms/InstSimplify/compare.ll b/llvm/test/Transforms/InstSimplify/compare.ll
index 1e90f0edbd800..724912d90bd86 100644
--- a/llvm/test/Transforms/InstSimplify/compare.ll
+++ b/llvm/test/Transforms/InstSimplify/compare.ll
@@ -1659,21 +1659,21 @@ define <2 x i1> @icmp_shl_1_ugt_signmask(<2 x i8> %V) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_shl_1_ugt_signmask_undef(<2 x i8> %V) {
-; CHECK-LABEL: @icmp_shl_1_ugt_signmask_undef(
+define <2 x i1> @icmp_shl_1_ugt_signmask_poison(<2 x i8> %V) {
+; CHECK-LABEL: @icmp_shl_1_ugt_signmask_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %shl = shl <2 x i8> <i8 1, i8 1>, %V
-  %cmp = icmp ugt <2 x i8> %shl, <i8 128, i8 undef>
+  %cmp = icmp ugt <2 x i8> %shl, <i8 128, i8 poison>
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_shl_1_ugt_signmask_undef2(<2 x i8> %V) {
-; CHECK-LABEL: @icmp_shl_1_ugt_signmask_undef2(
+define <2 x i1> @icmp_shl_1_ugt_signmask_poison2(<2 x i8> %V) {
+; CHECK-LABEL: @icmp_shl_1_ugt_signmask_poison2(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
-  %shl = shl <2 x i8> <i8 1, i8 undef>, %V
-  %cmp = icmp ugt <2 x i8> %shl, <i8 undef, i8 128>
+  %shl = shl <2 x i8> <i8 1, i8 poison>, %V
+  %cmp = icmp ugt <2 x i8> %shl, <i8 poison, i8 128>
   ret <2 x i1> %cmp
 }
 
@@ -1695,21 +1695,21 @@ define <2 x i1> @icmp_shl_1_ule_signmask(<2 x i8> %V) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_shl_1_ule_signmask_undef(<2 x i8> %V) {
-; CHECK-LABEL: @icmp_shl_1_ule_signmask_undef(
+define <2 x i1> @icmp_shl_1_ule_signmask_poison(<2 x i8> %V) {
+; CHECK-LABEL: @icmp_shl_1_ule_signmask_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %shl = shl <2 x i8> <i8 1, i8 1>, %V
-  %cmp = icmp ule <2 x i8> %shl, <i8 128, i8 undef>
+  %cmp = icmp ule <2 x i8> %shl, <i8 128, i8 poison>
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @icmp_shl_1_ule_signmask_undef2(<2 x i8> %V) {
-; CHECK-LABEL: @icmp_shl_1_ule_signmask_undef2(
+define <2 x i1> @icmp_shl_1_ule_signmask_poison2(<2 x i8> %V) {
+; CHECK-LABEL: @icmp_shl_1_ule_signmask_poison2(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %shl = shl <2 x i8> <i8 1, i8 undef>, %V
-  %cmp = icmp ule <2 x i8> %shl, <i8 undef, i8 128>
+  %shl = shl <2 x i8> <i8 1, i8 poison>, %V
+  %cmp = icmp ule <2 x i8> %shl, <i8 poison, i8 128>
   ret <2 x i1> %cmp
 }
 
@@ -1731,12 +1731,12 @@ define <2 x i1> @shl_1_cmp_eq_nonpow2_splat(<2 x i32> %x) {
   ret <2 x i1> %c
 }
 
-define <2 x i1> @shl_1_cmp_eq_nonpow2_splat_undef(<2 x i32> %x) {
-; CHECK-LABEL: @shl_1_cmp_eq_nonpow2_splat_undef(
+define <2 x i1> @shl_1_cmp_eq_nonpow2_splat_poison(<2 x i32> %x) {
+; CHECK-LABEL: @shl_1_cmp_eq_nonpow2_splat_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %s = shl <2 x i32> <i32 1, i32 1>, %x
-  %c = icmp eq <2 x i32> %s, <i32 31, i32 undef>
+  %c = icmp eq <2 x i32> %s, <i32 31, i32 poison>
   ret <2 x i1> %c
 }
 
@@ -1758,12 +1758,12 @@ define <2 x i1> @shl_1_cmp_ne_nonpow2_splat(<2 x i32> %x) {
   ret <2 x i1> %c
 }
 
-define <2 x i1> @shl_1_cmp_ne_nonpow2_splat_undef(<2 x i32> %x) {
-; CHECK-LABEL: @shl_1_cmp_ne_nonpow2_splat_undef(
+define <2 x i1> @shl_1_cmp_ne_nonpow2_splat_poison(<2 x i32> %x) {
+; CHECK-LABEL: @shl_1_cmp_ne_nonpow2_splat_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %s = shl <2 x i32> <i32 undef, i32 1>, %x
-  %c = icmp ne <2 x i32> %s, <i32 42, i32 undef>
+  %s = shl <2 x i32> <i32 poison, i32 1>, %x
+  %c = icmp ne <2 x i32> %s, <i32 42, i32 poison>
   ret <2 x i1> %c
 }
 
@@ -1776,12 +1776,12 @@ define i1 @shl_pow2_cmp_eq_nonpow2(i32 %x) {
   ret i1 %c
 }
 
-define <2 x i1> @shl_pow21_cmp_ne_nonpow2_splat_undef(<2 x i32> %x) {
-; CHECK-LABEL: @shl_pow21_cmp_ne_nonpow2_splat_undef(
+define <2 x i1> @shl_pow21_cmp_ne_nonpow2_splat_poison(<2 x i32> %x) {
+; CHECK-LABEL: @shl_pow21_cmp_ne_nonpow2_splat_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %s = shl <2 x i32> <i32 undef, i32 4>, %x
-  %c = icmp ne <2 x i32> %s, <i32 31, i32 undef>
+  %s = shl <2 x i32> <i32 poison, i32 4>, %x
+  %c = icmp ne <2 x i32> %s, <i32 31, i32 poison>
   ret <2 x i1> %c
 }
 
@@ -1820,12 +1820,12 @@ define i1 @shl_pow2_cmp_eq_zero_nuw(i32 %x) {
   ret i1 %c
 }
 
-define <2 x i1> @shl_pow2_cmp_ne_zero_nuw_splat_undef(<2 x i32> %x) {
-; CHECK-LABEL: @shl_pow2_cmp_ne_zero_nuw_splat_undef(
+define <2 x i1> @shl_pow2_cmp_ne_zero_nuw_splat_poison(<2 x i32> %x) {
+; CHECK-LABEL: @shl_pow2_cmp_ne_zero_nuw_splat_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %s = shl nuw <2 x i32> <i32 16, i32 undef>, %x
-  %c = icmp ne <2 x i32> %s, <i32 undef, i32 0>
+  %s = shl nuw <2 x i32> <i32 16, i32 poison>, %x
+  %c = icmp ne <2 x i32> %s, <i32 poison, i32 0>
   ret <2 x i1> %c
 }
 
@@ -1838,12 +1838,12 @@ define i1 @shl_pow2_cmp_ne_zero_nsw(i32 %x) {
   ret i1 %c
 }
 
-define <2 x i1> @shl_pow2_cmp_eq_zero_nsw_splat_undef(<2 x i32> %x) {
-; CHECK-LABEL: @shl_pow2_cmp_eq_zero_nsw_splat_undef(
+define <2 x i1> @shl_pow2_cmp_eq_zero_nsw_splat_poison(<2 x i32> %x) {
+; CHECK-LABEL: @shl_pow2_cmp_eq_zero_nsw_splat_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
-  %s = shl nsw <2 x i32> <i32 undef, i32 16>, %x
-  %c = icmp eq <2 x i32> %s, <i32 0, i32 undef>
+  %s = shl nsw <2 x i32> <i32 poison, i32 16>, %x
+  %c = icmp eq <2 x i32> %s, <i32 0, i32 poison>
   ret <2 x i1> %c
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/constantfold-add-nuw-allones-to-allones.ll b/llvm/test/Transforms/InstSimplify/constantfold-add-nuw-allones-to-allones.ll
index 7c9d9a9e2c7ce..92d6cc30d6248 100644
--- a/llvm/test/Transforms/InstSimplify/constantfold-add-nuw-allones-to-allones.ll
+++ b/llvm/test/Transforms/InstSimplify/constantfold-add-nuw-allones-to-allones.ll
@@ -63,11 +63,11 @@ define <2 x i8> @add_vec(<2 x i8> %x) {
   ret <2 x i8> %ret
 }
 
-define <3 x i8> @add_vec_undef(<3 x i8> %x) {
-; CHECK-LABEL: @add_vec_undef(
-; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 undef, i8 -1>
+define <3 x i8> @add_vec_poison(<3 x i8> %x) {
+; CHECK-LABEL: @add_vec_poison(
+; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 poison, i8 -1>
 ;
-  %ret = add nuw <3 x i8> %x, <i8 -1, i8 undef, i8 -1>
+  %ret = add nuw <3 x i8> %x, <i8 -1, i8 poison, i8 -1>
   ret <3 x i8> %ret
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/constantfold-shl-nuw-C-to-C.ll b/llvm/test/Transforms/InstSimplify/constantfold-shl-nuw-C-to-C.ll
index b5b5773fee538..3f4a08807a4b4 100644
--- a/llvm/test/Transforms/InstSimplify/constantfold-shl-nuw-C-to-C.ll
+++ b/llvm/test/Transforms/InstSimplify/constantfold-shl-nuw-C-to-C.ll
@@ -78,11 +78,11 @@ define <2 x i8> @shl_vec(<2 x i8> %x) {
   ret <2 x i8> %ret
 }
 
-define <3 x i8> @shl_vec_undef(<3 x i8> %x) {
-; CHECK-LABEL: @shl_vec_undef(
-; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 undef, i8 -1>
+define <3 x i8> @shl_vec_poison(<3 x i8> %x) {
+; CHECK-LABEL: @shl_vec_poison(
+; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 poison, i8 -1>
 ;
-  %ret = shl nuw <3 x i8> <i8 -1, i8 undef, i8 -1>, %x
+  %ret = shl nuw <3 x i8> <i8 -1, i8 poison, i8 -1>, %x
   ret <3 x i8> %ret
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/div.ll b/llvm/test/Transforms/InstSimplify/div.ll
index e13b6f139bcf5..5ca2e8837b924 100644
--- a/llvm/test/Transforms/InstSimplify/div.ll
+++ b/llvm/test/Transforms/InstSimplify/div.ll
@@ -17,11 +17,11 @@ define <2 x i32> @zero_dividend_vector(<2 x i32> %A) {
   ret <2 x i32> %B
 }
 
-define <2 x i32> @zero_dividend_vector_undef_elt(<2 x i32> %A) {
-; CHECK-LABEL: @zero_dividend_vector_undef_elt(
+define <2 x i32> @zero_dividend_vector_poison_elt(<2 x i32> %A) {
+; CHECK-LABEL: @zero_dividend_vector_poison_elt(
 ; CHECK-NEXT:    ret <2 x i32> zeroinitializer
 ;
-  %B = sdiv <2 x i32> <i32 0, i32 undef>, %A
+  %B = sdiv <2 x i32> <i32 0, i32 poison>, %A
   ret <2 x i32> %B
 }
 
@@ -59,23 +59,23 @@ define <2 x i8> @udiv_zero_elt_vec(<2 x i8> %x) {
   ret <2 x i8> %div
 }
 
-define <2 x i8> @sdiv_undef_elt_vec(<2 x i8> %x) {
-; CHECK-LABEL: @sdiv_undef_elt_vec(
+define <2 x i8> @sdiv_poison_elt_vec(<2 x i8> %x) {
+; CHECK-LABEL: @sdiv_poison_elt_vec(
 ; CHECK-NEXT:    ret <2 x i8> poison
 ;
-  %div = sdiv <2 x i8> %x, <i8 -42, i8 undef>
+  %div = sdiv <2 x i8> %x, <i8 -42, i8 poison>
   ret <2 x i8> %div
 }
 
-define <2 x i8> @udiv_undef_elt_vec(<2 x i8> %x) {
-; CHECK-LABEL: @udiv_undef_elt_vec(
+define <2 x i8> @udiv_poison_elt_vec(<2 x i8> %x) {
+; CHECK-LABEL: @udiv_poison_elt_vec(
 ; CHECK-NEXT:    ret <2 x i8> poison
 ;
-  %div = udiv <2 x i8> %x, <i8 undef, i8 42>
+  %div = udiv <2 x i8> %x, <i8 poison, i8 42>
   ret <2 x i8> %div
 }
 
-; Division-by-zero is undef. UB in any vector lane means the whole op is undef.
+; Division-by-zero is poison. UB in any vector lane means the whole op is poison.
 ; Thus, we can simplify this: if any element of 'y' is 0, we can do anything.
 ; Therefore, assume that all elements of 'y' must be 1.
 
diff --git a/llvm/test/Transforms/InstSimplify/fast-math-strictfp.ll b/llvm/test/Transforms/InstSimplify/fast-math-strictfp.ll
index 4938987baccc2..b1d772890aff8 100644
--- a/llvm/test/Transforms/InstSimplify/fast-math-strictfp.ll
+++ b/llvm/test/Transforms/InstSimplify/fast-math-strictfp.ll
@@ -18,11 +18,11 @@ define float @mul_zero_2(float %a) #0 {
   ret float %b
 }
 
-define <2 x float> @mul_zero_nsz_nnan_vec_undef(<2 x float> %a) #0 {
-; CHECK-LABEL: @mul_zero_nsz_nnan_vec_undef(
+define <2 x float> @mul_zero_nsz_nnan_vec_poison(<2 x float> %a) #0 {
+; CHECK-LABEL: @mul_zero_nsz_nnan_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> zeroinitializer
 ;
-  %b = call nsz nnan <2 x float> @llvm.experimental.constrained.fmul.v2f32(<2 x float> %a, <2 x float><float 0.0, float undef>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %b = call nsz nnan <2 x float> @llvm.experimental.constrained.fmul.v2f32(<2 x float> %a, <2 x float><float 0.0, float poison>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %b
 }
 
@@ -98,13 +98,13 @@ define <2 x float> @fadd_unary_fnegx_commute_vec(<2 x float> %x) #0 {
   ret <2 x float> %r
 }
 
-define <2 x float> @fadd_fnegx_commute_vec_undef(<2 x float> %x) #0 {
-; CHECK-LABEL: @fadd_fnegx_commute_vec_undef(
-; CHECK-NEXT:    [[NEGX:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.000000e+00>, <2 x float> [[X:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fadd_fnegx_commute_vec_poison(<2 x float> %x) #0 {
+; CHECK-LABEL: @fadd_fnegx_commute_vec_poison(
+; CHECK-NEXT:    [[NEGX:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.000000e+00>, <2 x float> [[X:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    [[R:%.*]] = call nnan <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> [[X]], <2 x float> [[NEGX]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
-  %negx = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.0>, <2 x float> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %negx = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.0>, <2 x float> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   %r = call nnan <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> %negx, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %r
 }
@@ -240,34 +240,34 @@ define float @fneg_x(float %a) #0 {
   ret float %ret
 }
 
-define <2 x float> @fsub_0_0_x_vec_undef1(<2 x float> %a) #0 {
-; CHECK-LABEL: @fsub_0_0_x_vec_undef1(
-; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float undef>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fsub_0_0_x_vec_poison1(<2 x float> %a) #0 {
+; CHECK-LABEL: @fsub_0_0_x_vec_poison1(
+; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float poison>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    [[RET:%.*]] = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> zeroinitializer, <2 x float> [[T1]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[RET]]
 ;
-  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.0, float undef>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.0, float poison>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   %ret = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> zeroinitializer, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
-define <2 x float> @fneg_x_vec_undef1(<2 x float> %a) #0 {
-; CHECK-LABEL: @fneg_x_vec_undef1(
+define <2 x float> @fneg_x_vec_poison1(<2 x float> %a) #0 {
+; CHECK-LABEL: @fneg_x_vec_poison1(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fneg <2 x float> %a
-  %ret = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.0, float undef>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.0, float poison>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_0_0_x_vec_undef2(<2 x float> %a) #0 {
-; CHECK-LABEL: @fsub_0_0_x_vec_undef2(
+define <2 x float> @fsub_0_0_x_vec_poison2(<2 x float> %a) #0 {
+; CHECK-LABEL: @fsub_0_0_x_vec_poison2(
 ; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> zeroinitializer, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
-; CHECK-NEXT:    [[RET:%.*]] = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.000000e+00>, <2 x float> [[T1]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+; CHECK-NEXT:    [[RET:%.*]] = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.000000e+00>, <2 x float> [[T1]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[RET]]
 ;
   %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> zeroinitializer, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
-  %ret = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.0>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.0>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
@@ -281,11 +281,11 @@ define <2 x float> @fadd_zero_nsz_vec(<2 x float> %x) #0 {
   ret <2 x float> %r
 }
 
-define <2 x float> @fadd_zero_nsz_vec_undef(<2 x float> %x) #0 {
-; CHECK-LABEL: @fadd_zero_nsz_vec_undef(
+define <2 x float> @fadd_zero_nsz_vec_poison(<2 x float> %x) #0 {
+; CHECK-LABEL: @fadd_zero_nsz_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> [[X:%.*]]
 ;
-  %r = call nsz <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> <float 0.0, float undef>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call nsz <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> <float 0.0, float poison>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %r
 }
 
@@ -375,11 +375,11 @@ define double @fdiv_zero_by_x(double %x) #0 {
   ret double %r
 }
 
-define <2 x double> @fdiv_zero_by_x_vec_undef(<2 x double> %x) #0 {
-; CHECK-LABEL: @fdiv_zero_by_x_vec_undef(
+define <2 x double> @fdiv_zero_by_x_vec_poison(<2 x double> %x) #0 {
+; CHECK-LABEL: @fdiv_zero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> zeroinitializer
 ;
-  %r = call nnan nsz <2 x double> @llvm.experimental.constrained.fdiv.v2f64(<2 x double> <double 0.0, double undef>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call nnan nsz <2 x double> @llvm.experimental.constrained.fdiv.v2f64(<2 x double> <double 0.0, double poison>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x double> %r
 }
 
@@ -394,11 +394,11 @@ define double @frem_zero_by_x(double %x) #0 {
   ret double %r
 }
 
-define <2 x double> @frem_poszero_by_x_vec_undef(<2 x double> %x) #0 {
-; CHECK-LABEL: @frem_poszero_by_x_vec_undef(
+define <2 x double> @frem_poszero_by_x_vec_poison(<2 x double> %x) #0 {
+; CHECK-LABEL: @frem_poszero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> zeroinitializer
 ;
-  %r = call nnan <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double> <double 0.0, double undef>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call nnan <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double> <double 0.0, double poison>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x double> %r
 }
 
@@ -413,11 +413,11 @@ define double @frem_negzero_by_x(double %x) #0 {
   ret double %r
 }
 
-define <2 x double> @frem_negzero_by_x_vec_undef(<2 x double> %x) #0 {
-; CHECK-LABEL: @frem_negzero_by_x_vec_undef(
+define <2 x double> @frem_negzero_by_x_vec_poison(<2 x double> %x) #0 {
+; CHECK-LABEL: @frem_negzero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> <double -0.000000e+00, double -0.000000e+00>
 ;
-  %r = call nnan <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double> <double undef, double -0.0>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call nnan <2 x double> @llvm.experimental.constrained.frem.v2f64(<2 x double> <double poison, double -0.0>, <2 x double> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x double> %r
 }
 
@@ -493,13 +493,13 @@ define float @fdiv_neg_swapped2(float %f) #0 {
   ret float %div
 }
 
-define <2 x float> @fdiv_neg_vec_undef_elt(<2 x float> %f) #0 {
-; CHECK-LABEL: @fdiv_neg_vec_undef_elt(
-; CHECK-NEXT:    [[NEG:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float undef>, <2 x float> [[F:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fdiv_neg_vec_poison_elt(<2 x float> %f) #0 {
+; CHECK-LABEL: @fdiv_neg_vec_poison_elt(
+; CHECK-NEXT:    [[NEG:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float poison>, <2 x float> [[F:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    [[DIV:%.*]] = call nnan <2 x float> @llvm.experimental.constrained.fdiv.v2f32(<2 x float> [[F]], <2 x float> [[NEG]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[DIV]]
 ;
-  %neg = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float undef>, <2 x float> %f, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %neg = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float 0.000000e+00, float poison>, <2 x float> %f, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   %div = call nnan <2 x float> @llvm.experimental.constrained.fdiv.v2f32(<2 x float> %f, <2 x float> %neg, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %div
 }
diff --git a/llvm/test/Transforms/InstSimplify/fast-math.ll b/llvm/test/Transforms/InstSimplify/fast-math.ll
index d1818e6346d7a..287f30b162f80 100644
--- a/llvm/test/Transforms/InstSimplify/fast-math.ll
+++ b/llvm/test/Transforms/InstSimplify/fast-math.ll
@@ -18,11 +18,11 @@ define float @mul_zero_2(float %a) {
   ret float %b
 }
 
-define <2 x float> @mul_zero_nsz_nnan_vec_undef(<2 x float> %a) {
-; CHECK-LABEL: @mul_zero_nsz_nnan_vec_undef(
+define <2 x float> @mul_zero_nsz_nnan_vec_poison(<2 x float> %a) {
+; CHECK-LABEL: @mul_zero_nsz_nnan_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> zeroinitializer
 ;
-  %b = fmul nsz nnan <2 x float> %a, <float 0.0, float undef>
+  %b = fmul nsz nnan <2 x float> %a, <float 0.0, float poison>
   ret <2 x float> %b
 }
 
@@ -94,11 +94,11 @@ define <2 x float> @fadd_unary_fnegx_commute_vec(<2 x float> %x) {
   ret <2 x float> %r
 }
 
-define <2 x float> @fadd_fnegx_commute_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fadd_fnegx_commute_vec_undef(
+define <2 x float> @fadd_fnegx_commute_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fadd_fnegx_commute_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> zeroinitializer
 ;
-  %negx = fsub <2 x float> <float undef, float -0.0>, %x
+  %negx = fsub <2 x float> <float poison, float -0.0>, %x
   %r = fadd nnan <2 x float> %x, %negx
   ret <2 x float> %r
 }
@@ -226,30 +226,30 @@ define float @fneg_x(float %a) {
   ret float %ret
 }
 
-define <2 x float> @fsub_0_0_x_vec_undef1(<2 x float> %a) {
-; CHECK-LABEL: @fsub_0_0_x_vec_undef1(
+define <2 x float> @fsub_0_0_x_vec_poison1(<2 x float> %a) {
+; CHECK-LABEL: @fsub_0_0_x_vec_poison1(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
-  %t1 = fsub <2 x float> <float 0.0, float undef>, %a
+  %t1 = fsub <2 x float> <float 0.0, float poison>, %a
   %ret = fsub nsz <2 x float> zeroinitializer, %t1
   ret <2 x float> %ret
 }
 
-define <2 x float> @fneg_x_vec_undef1(<2 x float> %a) {
-; CHECK-LABEL: @fneg_x_vec_undef1(
+define <2 x float> @fneg_x_vec_poison1(<2 x float> %a) {
+; CHECK-LABEL: @fneg_x_vec_poison1(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fneg <2 x float> %a
-  %ret = fsub nsz <2 x float> <float 0.0, float undef>, %t1
+  %ret = fsub nsz <2 x float> <float 0.0, float poison>, %t1
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_0_0_x_vec_undef2(<2 x float> %a) {
-; CHECK-LABEL: @fsub_0_0_x_vec_undef2(
+define <2 x float> @fsub_0_0_x_vec_poison2(<2 x float> %a) {
+; CHECK-LABEL: @fsub_0_0_x_vec_poison2(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fsub <2 x float> zeroinitializer, %a
-  %ret = fsub nsz <2 x float> <float undef, float -0.0>, %t1
+  %ret = fsub nsz <2 x float> <float poison, float -0.0>, %t1
   ret <2 x float> %ret
 }
 
@@ -263,11 +263,11 @@ define <2 x float> @fadd_zero_nsz_vec(<2 x float> %x) {
   ret <2 x float> %r
 }
 
-define <2 x float> @fadd_zero_nsz_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fadd_zero_nsz_vec_undef(
+define <2 x float> @fadd_zero_nsz_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fadd_zero_nsz_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> [[X:%.*]]
 ;
-  %r = fadd nsz <2 x float> %x, <float 0.0, float undef>
+  %r = fadd nsz <2 x float> %x, <float 0.0, float poison>
   ret <2 x float> %r
 }
 
@@ -357,11 +357,11 @@ define double @fdiv_zero_by_x(double %x) {
   ret double %r
 }
 
-define <2 x double> @fdiv_zero_by_x_vec_undef(<2 x double> %x) {
-; CHECK-LABEL: @fdiv_zero_by_x_vec_undef(
+define <2 x double> @fdiv_zero_by_x_vec_poison(<2 x double> %x) {
+; CHECK-LABEL: @fdiv_zero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> zeroinitializer
 ;
-  %r = fdiv nnan nsz <2 x double> <double 0.0, double undef>, %x
+  %r = fdiv nnan nsz <2 x double> <double 0.0, double poison>, %x
   ret <2 x double> %r
 }
 
@@ -376,11 +376,11 @@ define double @frem_zero_by_x(double %x) {
   ret double %r
 }
 
-define <2 x double> @frem_poszero_by_x_vec_undef(<2 x double> %x) {
-; CHECK-LABEL: @frem_poszero_by_x_vec_undef(
+define <2 x double> @frem_poszero_by_x_vec_poison(<2 x double> %x) {
+; CHECK-LABEL: @frem_poszero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> zeroinitializer
 ;
-  %r = frem nnan <2 x double> <double 0.0, double undef>, %x
+  %r = frem nnan <2 x double> <double 0.0, double poison>, %x
   ret <2 x double> %r
 }
 
@@ -395,11 +395,11 @@ define double @frem_negzero_by_x(double %x) {
   ret double %r
 }
 
-define <2 x double> @frem_negzero_by_x_vec_undef(<2 x double> %x) {
-; CHECK-LABEL: @frem_negzero_by_x_vec_undef(
+define <2 x double> @frem_negzero_by_x_vec_poison(<2 x double> %x) {
+; CHECK-LABEL: @frem_negzero_by_x_vec_poison(
 ; CHECK-NEXT:    ret <2 x double> <double -0.000000e+00, double -0.000000e+00>
 ;
-  %r = frem nnan <2 x double> <double undef, double -0.0>, %x
+  %r = frem nnan <2 x double> <double poison, double -0.0>, %x
   ret <2 x double> %r
 }
 
@@ -467,11 +467,11 @@ define float @fdiv_neg_swapped2(float %f) {
   ret float %div
 }
 
-define <2 x float> @fdiv_neg_vec_undef_elt(<2 x float> %f) {
-; CHECK-LABEL: @fdiv_neg_vec_undef_elt(
+define <2 x float> @fdiv_neg_vec_poison_elt(<2 x float> %f) {
+; CHECK-LABEL: @fdiv_neg_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x float> <float -1.000000e+00, float -1.000000e+00>
 ;
-  %neg = fsub <2 x float> <float 0.0, float undef>, %f
+  %neg = fsub <2 x float> <float 0.0, float poison>, %f
   %div = fdiv nnan <2 x float> %f, %neg
   ret <2 x float> %div
 }
diff --git a/llvm/test/Transforms/InstSimplify/fdiv.ll b/llvm/test/Transforms/InstSimplify/fdiv.ll
index 38e31257e185a..fb59011b91d5b 100644
--- a/llvm/test/Transforms/InstSimplify/fdiv.ll
+++ b/llvm/test/Transforms/InstSimplify/fdiv.ll
@@ -110,11 +110,11 @@ define <2 x float> @fdiv_nnan_ninf_by_undef_v2f32(<2 x float> %x) {
   ret <2 x float> %fdiv
 }
 
-define <2 x float> @fdiv_nnan_ninf_by_zero_undef_v2f32(<2 x float> %x) {
-; CHECK-LABEL: @fdiv_nnan_ninf_by_zero_undef_v2f32(
+define <2 x float> @fdiv_nnan_ninf_by_zero_poison_v2f32(<2 x float> %x) {
+; CHECK-LABEL: @fdiv_nnan_ninf_by_zero_poison_v2f32(
 ; CHECK-NEXT:    ret <2 x float> poison
 ;
-  %fdiv = fdiv nnan ninf <2 x float> %x, <float 0.0, float undef>
+  %fdiv = fdiv nnan ninf <2 x float> %x, <float 0.0, float poison>
   ret <2 x float> %fdiv
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/floating-point-arithmetic-strictfp.ll b/llvm/test/Transforms/InstSimplify/floating-point-arithmetic-strictfp.ll
index e4748a2402923..32ea4cb7cd198 100644
--- a/llvm/test/Transforms/InstSimplify/floating-point-arithmetic-strictfp.ll
+++ b/llvm/test/Transforms/InstSimplify/floating-point-arithmetic-strictfp.ll
@@ -24,23 +24,23 @@ define <2 x float> @fsub_-0_x_vec(<2 x float> %a) #0 {
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_-0_x_vec_undef_elts(<2 x float> %a) #0 {
-; CHECK-LABEL: @fsub_-0_x_vec_undef_elts(
-; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float -0.000000e+00, float undef>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fsub_-0_x_vec_poison_elts(<2 x float> %a) #0 {
+; CHECK-LABEL: @fsub_-0_x_vec_poison_elts(
+; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float -0.000000e+00, float poison>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    [[RET:%.*]] = fneg <2 x float> [[T1]]
 ; CHECK-NEXT:    ret <2 x float> [[RET]]
 ;
-  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float undef>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float poison>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   %ret = fneg <2 x float> %t1
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_negzero_vec_undef_elts(<2 x float> %x) #0 {
-; CHECK-LABEL: @fsub_negzero_vec_undef_elts(
-; CHECK-NEXT:    [[R:%.*]] = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.000000e+00>, <2 x float> [[X:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fsub_negzero_vec_poison_elts(<2 x float> %x) #0 {
+; CHECK-LABEL: @fsub_negzero_vec_poison_elts(
+; CHECK-NEXT:    [[R:%.*]] = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.000000e+00>, <2 x float> [[X:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
-  %r = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float undef, float -0.0>, <2 x float> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call nsz <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float poison, float -0.0>, <2 x float> %x, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %r
 }
 
@@ -86,23 +86,23 @@ define <2 x float> @fneg_x_vec(<2 x float> %a) #0 {
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_-0_-0_x_vec_undef_elts(<2 x float> %a) #0 {
-; CHECK-LABEL: @fsub_-0_-0_x_vec_undef_elts(
-; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float undef, float -0.000000e+00>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
-; CHECK-NEXT:    [[RET:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float -0.000000e+00, float undef>, <2 x float> [[T1]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fsub_-0_-0_x_vec_poison_elts(<2 x float> %a) #0 {
+; CHECK-LABEL: @fsub_-0_-0_x_vec_poison_elts(
+; CHECK-NEXT:    [[T1:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float poison, float -0.000000e+00>, <2 x float> [[A:%.*]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
+; CHECK-NEXT:    [[RET:%.*]] = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> <float -0.000000e+00, float poison>, <2 x float> [[T1]], metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[RET]]
 ;
-  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float undef, float -0.0>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
-  %ret = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float undef>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %t1 = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float poison, float -0.0>, <2 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float poison>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
-define <2 x float> @fneg_x_vec_undef_elts(<2 x float> %a) #0 {
-; CHECK-LABEL: @fneg_x_vec_undef_elts(
+define <2 x float> @fneg_x_vec_poison_elts(<2 x float> %a) #0 {
+; CHECK-LABEL: @fneg_x_vec_poison_elts(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fneg <2 x float> %a
-  %ret = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float undef>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float><float -0.0, float poison>, <2 x float> %t1, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
@@ -139,11 +139,11 @@ define float @fsub_x_0(float %x) #0 {
   ret float %r
 }
 
-define <2 x float> @fsub_x_0_vec_undef(<2 x float> %x) #0 {
-; CHECK-LABEL: @fsub_x_0_vec_undef(
+define <2 x float> @fsub_x_0_vec_poison(<2 x float> %x) #0 {
+; CHECK-LABEL: @fsub_x_0_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> [[X:%.*]]
 ;
-  %r = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> %x, <2 x float><float undef, float 0.0>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %r = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> %x, <2 x float><float poison, float 0.0>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %r
 }
 
@@ -156,11 +156,11 @@ define float @fadd_x_n0(float %a) #0 {
   ret float %ret
 }
 
-define <2 x float> @fadd_x_n0_vec_undef_elt(<2 x float> %a) #0 {
-; CHECK-LABEL: @fadd_x_n0_vec_undef_elt(
+define <2 x float> @fadd_x_n0_vec_poison_elt(<2 x float> %a) #0 {
+; CHECK-LABEL: @fadd_x_n0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
-  %ret = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %a, <2 x float> <float -0.0, float undef>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %a, <2 x float> <float -0.0, float poison>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
@@ -174,12 +174,12 @@ define float @fadd_x_p0(float %a) #0 {
   ret float %ret
 }
 
-define <2 x float> @fadd_x_p0_vec_undef_elt(<2 x float> %a) #0 {
-; CHECK-LABEL: @fadd_x_p0_vec_undef_elt(
-; CHECK-NEXT:    [[RET:%.*]] = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> [[A:%.*]], <2 x float> <float 0.000000e+00, float undef>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+define <2 x float> @fadd_x_p0_vec_poison_elt(<2 x float> %a) #0 {
+; CHECK-LABEL: @fadd_x_p0_vec_poison_elt(
+; CHECK-NEXT:    [[RET:%.*]] = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> [[A:%.*]], <2 x float> <float 0.000000e+00, float poison>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
 ; CHECK-NEXT:    ret <2 x float> [[RET]]
 ;
-  %ret = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %a, <2 x float> <float 0.0, float undef>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
+  %ret = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %a, <2 x float> <float 0.0, float poison>, metadata !"round.tonearest", metadata !"fpexcept.ignore")
   ret <2 x float> %ret
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/floating-point-arithmetic.ll b/llvm/test/Transforms/InstSimplify/floating-point-arithmetic.ll
index 5d17504c09df6..7a35f09f03b99 100644
--- a/llvm/test/Transforms/InstSimplify/floating-point-arithmetic.ll
+++ b/llvm/test/Transforms/InstSimplify/floating-point-arithmetic.ll
@@ -30,20 +30,20 @@ define <2 x float> @fsub_-0_x_vec(<2 x float> %a) {
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_-0_x_vec_undef_elts(<2 x float> %a) {
-; CHECK-LABEL: @fsub_-0_x_vec_undef_elts(
+define <2 x float> @fsub_-0_x_vec_poison_elts(<2 x float> %a) {
+; CHECK-LABEL: @fsub_-0_x_vec_poison_elts(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
-  %t1 = fsub <2 x float> <float -0.0, float undef>, %a
+  %t1 = fsub <2 x float> <float -0.0, float poison>, %a
   %ret = fneg <2 x float> %t1
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_negzero_vec_undef_elts(<2 x float> %x) {
-; CHECK-LABEL: @fsub_negzero_vec_undef_elts(
+define <2 x float> @fsub_negzero_vec_poison_elts(<2 x float> %x) {
+; CHECK-LABEL: @fsub_negzero_vec_poison_elts(
 ; CHECK-NEXT:    ret <2 x float> [[X:%.*]]
 ;
-  %r = fsub nsz <2 x float> %x, <float undef, float -0.0>
+  %r = fsub nsz <2 x float> %x, <float poison, float -0.0>
   ret <2 x float> %r
 }
 
@@ -85,21 +85,21 @@ define <2 x float> @fneg_x_vec(<2 x float> %a) {
   ret <2 x float> %ret
 }
 
-define <2 x float> @fsub_-0_-0_x_vec_undef_elts(<2 x float> %a) {
-; CHECK-LABEL: @fsub_-0_-0_x_vec_undef_elts(
+define <2 x float> @fsub_-0_-0_x_vec_poison_elts(<2 x float> %a) {
+; CHECK-LABEL: @fsub_-0_-0_x_vec_poison_elts(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
-  %t1 = fsub <2 x float> <float undef, float -0.0>, %a
-  %ret = fsub <2 x float> <float -0.0, float undef>, %t1
+  %t1 = fsub <2 x float> <float poison, float -0.0>, %a
+  %ret = fsub <2 x float> <float -0.0, float poison>, %t1
   ret <2 x float> %ret
 }
 
-define <2 x float> @fneg_x_vec_undef_elts(<2 x float> %a) {
-; CHECK-LABEL: @fneg_x_vec_undef_elts(
+define <2 x float> @fneg_x_vec_poison_elts(<2 x float> %a) {
+; CHECK-LABEL: @fneg_x_vec_poison_elts(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
   %t1 = fneg <2 x float> %a
-  %ret = fsub <2 x float> <float -0.0, float undef>, %t1
+  %ret = fsub <2 x float> <float -0.0, float poison>, %t1
   ret <2 x float> %ret
 }
 
@@ -136,11 +136,11 @@ define float @fsub_x_0(float %x) {
   ret float %r
 }
 
-define <2 x float> @fsub_x_0_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @fsub_x_0_vec_undef(
+define <2 x float> @fsub_x_0_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @fsub_x_0_vec_poison(
 ; CHECK-NEXT:    ret <2 x float> [[X:%.*]]
 ;
-  %r = fsub <2 x float> %x, <float undef, float 0.0>
+  %r = fsub <2 x float> %x, <float poison, float 0.0>
   ret <2 x float> %r
 }
 
@@ -153,11 +153,11 @@ define float @fadd_x_n0(float %a) {
   ret float %ret
 }
 
-define <2 x float> @fadd_x_n0_vec_undef_elt(<2 x float> %a) {
-; CHECK-LABEL: @fadd_x_n0_vec_undef_elt(
+define <2 x float> @fadd_x_n0_vec_poison_elt(<2 x float> %a) {
+; CHECK-LABEL: @fadd_x_n0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x float> [[A:%.*]]
 ;
-  %ret = fadd <2 x float> %a, <float -0.0, float undef>
+  %ret = fadd <2 x float> %a, <float -0.0, float poison>
   ret <2 x float> %ret
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/floating-point-compare.ll b/llvm/test/Transforms/InstSimplify/floating-point-compare.ll
index 3c1794c81284d..70f0321039ea9 100644
--- a/llvm/test/Transforms/InstSimplify/floating-point-compare.ll
+++ b/llvm/test/Transforms/InstSimplify/floating-point-compare.ll
@@ -547,30 +547,30 @@ define <2 x i1> @fabs_is_not_negative_anyzero(<2 x float> %V) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @fabs_is_not_negative_negzero_undef(<3 x float> %V) {
-; CHECK-LABEL: @fabs_is_not_negative_negzero_undef(
+define <3 x i1> @fabs_is_not_negative_negzero_poison(<3 x float> %V) {
+; CHECK-LABEL: @fabs_is_not_negative_negzero_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %abs = call <3 x float> @llvm.fabs.v3f32(<3 x float> %V)
-  %cmp = fcmp olt <3 x float> %abs, <float -0.0, float -0.0, float undef>
+  %cmp = fcmp olt <3 x float> %abs, <float -0.0, float -0.0, float poison>
   ret <3 x i1> %cmp
 }
 
-define <3 x i1> @fabs_is_not_negative_poszero_undef(<3 x float> %V) {
-; CHECK-LABEL: @fabs_is_not_negative_poszero_undef(
+define <3 x i1> @fabs_is_not_negative_poszero_poison(<3 x float> %V) {
+; CHECK-LABEL: @fabs_is_not_negative_poszero_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %abs = call <3 x float> @llvm.fabs.v3f32(<3 x float> %V)
-  %cmp = fcmp olt <3 x float> %abs, <float 0.0, float 0.0, float undef>
+  %cmp = fcmp olt <3 x float> %abs, <float 0.0, float 0.0, float poison>
   ret <3 x i1> %cmp
 }
 
-define <3 x i1> @fabs_is_not_negative_anyzero_undef(<3 x float> %V) {
-; CHECK-LABEL: @fabs_is_not_negative_anyzero_undef(
+define <3 x i1> @fabs_is_not_negative_anyzero_poison(<3 x float> %V) {
+; CHECK-LABEL: @fabs_is_not_negative_anyzero_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %abs = call <3 x float> @llvm.fabs.v3f32(<3 x float> %V)
-  %cmp = fcmp olt <3 x float> %abs, <float 0.0, float -0.0, float undef>
+  %cmp = fcmp olt <3 x float> %abs, <float 0.0, float -0.0, float poison>
   ret <3 x i1> %cmp
 }
 
@@ -1335,19 +1335,19 @@ define <2 x i1> @orderedCompareWithNaNVector(<2 x double> %A) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @orderedCompareWithNaNVector_undef_elt(<2 x double> %A) {
-; CHECK-LABEL: @orderedCompareWithNaNVector_undef_elt(
+define <2 x i1> @orderedCompareWithNaNVector_poison_elt(<2 x double> %A) {
+; CHECK-LABEL: @orderedCompareWithNaNVector_poison_elt(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
-  %cmp = fcmp olt <2 x double> %A, <double 0xFFFFFFFFFFFFFFFF, double undef>
+  %cmp = fcmp olt <2 x double> %A, <double 0xFFFFFFFFFFFFFFFF, double poison>
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @unorderedCompareWithNaNVector_undef_elt(<2 x double> %A) {
-; CHECK-LABEL: @unorderedCompareWithNaNVector_undef_elt(
+define <2 x i1> @unorderedCompareWithNaNVector_poison_elt(<2 x double> %A) {
+; CHECK-LABEL: @unorderedCompareWithNaNVector_poison_elt(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %cmp = fcmp ult <2 x double> %A, <double undef, double 0xFFFFFFFFFFFFFFFF>
+  %cmp = fcmp ult <2 x double> %A, <double poison, double 0xFFFFFFFFFFFFFFFF>
   ret <2 x i1> %cmp
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/fminmax-folds.ll b/llvm/test/Transforms/InstSimplify/fminmax-folds.ll
index a8a9e96a652fa..668a93ddf5a42 100644
--- a/llvm/test/Transforms/InstSimplify/fminmax-folds.ll
+++ b/llvm/test/Transforms/InstSimplify/fminmax-folds.ll
@@ -493,7 +493,7 @@ define <2 x double> @maxnum_nan_op0_vec(<2 x double> %x) {
 ; CHECK-LABEL: @maxnum_nan_op0_vec(
 ; CHECK-NEXT:    ret <2 x double> [[X:%.*]]
 ;
-  %r = call <2 x double> @llvm.maxnum.v2f64(<2 x double> <double 0x7ff8000000000000, double undef>, <2 x double> %x)
+  %r = call <2 x double> @llvm.maxnum.v2f64(<2 x double> <double 0x7ff8000000000000, double poison>, <2 x double> %x)
   ret <2 x double> %r
 }
 
@@ -509,7 +509,7 @@ define <2 x double> @minnum_nan_op0_vec(<2 x double> %x) {
 ; CHECK-LABEL: @minnum_nan_op0_vec(
 ; CHECK-NEXT:    ret <2 x double> [[X:%.*]]
 ;
-  %r = call <2 x double> @llvm.minnum.v2f64(<2 x double> <double undef, double 0x7ff8000dead00000>, <2 x double> %x)
+  %r = call <2 x double> @llvm.minnum.v2f64(<2 x double> <double poison, double 0x7ff8000dead00000>, <2 x double> %x)
   ret <2 x double> %r
 }
 
@@ -873,19 +873,19 @@ define double @minimum_nan_op1(double %x) {
   ret double %r
 }
 
-define <2 x double> @maximum_nan_op0_vec_partial_undef(<2 x double> %x) {
-; CHECK-LABEL: @maximum_nan_op0_vec_partial_undef(
-; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>
+define <2 x double> @maximum_nan_op0_vec_partial_poison(<2 x double> %x) {
+; CHECK-LABEL: @maximum_nan_op0_vec_partial_poison(
+; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double poison>
 ;
-  %r = call <2 x double> @llvm.maximum.v2f64(<2 x double> <double 0x7ff8000000000000, double undef>, <2 x double> %x)
+  %r = call <2 x double> @llvm.maximum.v2f64(<2 x double> <double 0x7ff8000000000000, double poison>, <2 x double> %x)
   ret <2 x double> %r
 }
 
-define <2 x double> @maximum_nan_op1_vec_partial_undef(<2 x double> %x) {
-; CHECK-LABEL: @maximum_nan_op1_vec_partial_undef(
-; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double 0x7FF8000000000000>
+define <2 x double> @maximum_nan_op1_vec_partial_poison(<2 x double> %x) {
+; CHECK-LABEL: @maximum_nan_op1_vec_partial_poison(
+; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double poison>
 ;
-  %r = call <2 x double> @llvm.maximum.v2f64(<2 x double> %x, <2 x double> <double 0x7ff8000000000000, double undef>)
+  %r = call <2 x double> @llvm.maximum.v2f64(<2 x double> %x, <2 x double> <double 0x7ff8000000000000, double poison>)
   ret <2 x double> %r
 }
 
@@ -897,19 +897,19 @@ define <2 x double> @maximum_nan_op1_vec(<2 x double> %x) {
   ret <2 x double> %r
 }
 
-define <2 x double> @minimum_nan_op0_vec_partial_undef(<2 x double> %x) {
-; CHECK-LABEL: @minimum_nan_op0_vec_partial_undef(
-; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double 0x7FF8000DEAD00000>
+define <2 x double> @minimum_nan_op0_vec_partial_poison(<2 x double> %x) {
+; CHECK-LABEL: @minimum_nan_op0_vec_partial_poison(
+; CHECK-NEXT:    ret <2 x double> <double poison, double 0x7FF8000DEAD00000>
 ;
-  %r = call <2 x double> @llvm.minimum.v2f64(<2 x double> <double undef, double 0x7ff8000dead00000>, <2 x double> %x)
+  %r = call <2 x double> @llvm.minimum.v2f64(<2 x double> <double poison, double 0x7ff8000dead00000>, <2 x double> %x)
   ret <2 x double> %r
 }
 
-define <2 x double> @minimum_nan_op1_vec_partial_undef(<2 x double> %x) {
-; CHECK-LABEL: @minimum_nan_op1_vec_partial_undef(
-; CHECK-NEXT:    ret <2 x double> <double 0x7FF8000000000000, double 0x7FF8000DEAD00000>
+define <2 x double> @minimum_nan_op1_vec_partial_poison(<2 x double> %x) {
+; CHECK-LABEL: @minimum_nan_op1_vec_partial_poison(
+; CHECK-NEXT:    ret <2 x double> <double poison, double 0x7FF8000DEAD00000>
 ;
-  %r = call <2 x double> @llvm.minimum.v2f64(<2 x double> %x, <2 x double> <double undef, double 0x7ff8000dead00000>)
+  %r = call <2 x double> @llvm.minimum.v2f64(<2 x double> %x, <2 x double> <double poison, double 0x7ff8000dead00000>)
   ret <2 x double> %r
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/fp-nan.ll b/llvm/test/Transforms/InstSimplify/fp-nan.ll
index cb0bed3790782..bb557500822c1 100644
--- a/llvm/test/Transforms/InstSimplify/fp-nan.ll
+++ b/llvm/test/Transforms/InstSimplify/fp-nan.ll
@@ -163,13 +163,13 @@ define <2 x double> @fsub_nan_poison_op1(<2 x double> %x) {
   ret <2 x double> %r
 }
 
-; Vector with undef element
+; Vector with poison element
 
-define <2 x double> @frem_nan_undef_op0(<2 x double> %x) {
-; CHECK-LABEL: @frem_nan_undef_op0(
-; CHECK-NEXT:    ret <2 x double> <double 0xFFFF00000000DEAD, double 0x7FF8000000000000>
+define <2 x double> @frem_nan_poison_op0(<2 x double> %x) {
+; CHECK-LABEL: @frem_nan_poison_op0(
+; CHECK-NEXT:    ret <2 x double> <double 0xFFFF00000000DEAD, double poison>
 ;
-  %r = frem <2 x double> <double 0xFFFF00000000DEAD, double undef>, %x
+  %r = frem <2 x double> <double 0xFFFF00000000DEAD, double poison>, %x
   ret <2 x double> %r
 }
 
@@ -177,7 +177,8 @@ define <2 x double> @frem_nan_undef_op0(<2 x double> %x) {
 
 define <3 x double> @fadd_nan_poison_undef_op1(<3 x double> %x) {
 ; CHECK-LABEL: @fadd_nan_poison_undef_op1(
-; CHECK-NEXT:    ret <3 x double> <double 0xFFFF00000000DEAD, double poison, double 0x7FF8000000000000>
+; CHECK-NEXT:    [[R:%.*]] = fadd <3 x double> [[X:%.*]], <double 0xFFFF00000000DEAD, double poison, double undef>
+; CHECK-NEXT:    ret <3 x double> [[R]]
 ;
   %r = fadd <3 x double> %x, <double 0xFFFF00000000DEAD, double poison, double undef>
   ret <3 x double> %r
diff --git a/llvm/test/Transforms/InstSimplify/icmp-bool-constant.ll b/llvm/test/Transforms/InstSimplify/icmp-bool-constant.ll
index 6205225098a7a..a501f995b6c97 100644
--- a/llvm/test/Transforms/InstSimplify/icmp-bool-constant.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp-bool-constant.ll
@@ -12,11 +12,11 @@ define <2 x i1> @eq_t(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @eq_t_undef_elt(<2 x i1> %a) {
-; CHECK-LABEL: @eq_t_undef_elt(
+define <2 x i1> @eq_t_poison_elt(<2 x i1> %a) {
+; CHECK-LABEL: @eq_t_poison_elt(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %r = icmp eq <2 x i1> %a, <i1 undef, i1 true>
+  %r = icmp eq <2 x i1> %a, <i1 poison, i1 true>
   ret <2 x i1> %r
 }
 
@@ -54,11 +54,11 @@ define <2 x i1> @ugt_t(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @ugt_t_undef_elt(<2 x i1> %a) {
-; CHECK-LABEL: @ugt_t_undef_elt(
+define <2 x i1> @ugt_t_poison_elt(<2 x i1> %a) {
+; CHECK-LABEL: @ugt_t_poison_elt(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
-  %r = icmp ugt <2 x i1> %a, <i1 true, i1 undef>
+  %r = icmp ugt <2 x i1> %a, <i1 true, i1 poison>
   ret <2 x i1> %r
 }
 
@@ -161,11 +161,11 @@ define <2 x i1> @sge_t(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @sge_t_undef_elt(<2 x i1> %a) {
-; CHECK-LABEL: @sge_t_undef_elt(
+define <2 x i1> @sge_t_poison_elt(<2 x i1> %a) {
+; CHECK-LABEL: @sge_t_poison_elt(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %r = icmp sge <2 x i1> %a, <i1 true, i1 undef>
+  %r = icmp sge <2 x i1> %a, <i1 true, i1 poison>
   ret <2 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/icmp-not-bool-constant.ll b/llvm/test/Transforms/InstSimplify/icmp-not-bool-constant.ll
index f4a0b6ddf6621..045d773bf3284 100644
--- a/llvm/test/Transforms/InstSimplify/icmp-not-bool-constant.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp-not-bool-constant.ll
@@ -33,11 +33,11 @@ define <2 x i1> @eq_f_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @eq_f_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @eq_f_not_undef(
+define <2 x i1> @eq_f_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @eq_f_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp eq <2 x i1> %not, <i1 false, i1 false>
   ret <2 x i1> %r
 }
@@ -60,11 +60,11 @@ define <2 x i1> @ne_t_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @ne_t_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @ne_t_not_undef(
+define <2 x i1> @ne_t_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @ne_t_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp ne <2 x i1> %not, <i1 true, i1 true>
   ret <2 x i1> %r
 }
@@ -116,11 +116,11 @@ define <2 x i1> @ult_t_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @ult_t_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @ult_t_not_undef(
+define <2 x i1> @ult_t_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @ult_t_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp ult <2 x i1> %not, <i1 true, i1 true>
   ret <2 x i1> %r
 }
@@ -152,11 +152,11 @@ define <2 x i1> @sgt_t_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @sgt_t_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @sgt_t_not_undef(
+define <2 x i1> @sgt_t_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @sgt_t_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp sgt <2 x i1> %not, <i1 true, i1 true>
   ret <2 x i1> %r
 }
@@ -235,11 +235,11 @@ define <2 x i1> @ule_f_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @ule_f_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @ule_f_not_undef(
+define <2 x i1> @ule_f_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @ule_f_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp ule <2 x i1> %not, <i1 false, i1 false>
   ret <2 x i1> %r
 }
@@ -271,11 +271,11 @@ define <2 x i1> @sge_f_not_swap(<2 x i1> %a) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @sge_f_not_undef(<2 x i1> %a) {
-; CHECK-LABEL: @sge_f_not_undef(
+define <2 x i1> @sge_f_not_poison(<2 x i1> %a) {
+; CHECK-LABEL: @sge_f_not_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[A:%.*]]
 ;
-  %not = xor <2 x i1> %a, <i1 undef, i1 true>
+  %not = xor <2 x i1> %a, <i1 poison, i1 true>
   %r = icmp sge <2 x i1> %not, <i1 false, i1 false>
   ret <2 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstSimplify/ldexp.ll b/llvm/test/Transforms/InstSimplify/ldexp.ll
index c6bb0141199f2..d39f6a1e49673 100644
--- a/llvm/test/Transforms/InstSimplify/ldexp.ll
+++ b/llvm/test/Transforms/InstSimplify/ldexp.ll
@@ -57,11 +57,12 @@ define void @ldexp_f32_exp0(float %x) {
 define void @ldexp_v2f32_exp0(<2 x float> %x) {
 ; CHECK-LABEL: @ldexp_v2f32_exp0(
 ; CHECK-NEXT:    store volatile <2 x float> [[X:%.*]], ptr addrspace(1) undef, align 8
-; CHECK-NEXT:    store volatile <2 x float> [[X]], ptr addrspace(1) undef, align 8
+; CHECK-NEXT:    [[PART_UNDEF1:%.*]] = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> [[X]], <2 x i32> <i32 undef, i32 0>)
+; CHECK-NEXT:    store volatile <2 x float> [[PART_UNDEF1]], ptr addrspace(1) undef, align 8
 ; CHECK-NEXT:    store volatile <2 x float> [[X]], ptr addrspace(1) undef, align 8
 ; CHECK-NEXT:    ret void
 ;
-  %part.undef0 = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> %x, <2 x i32> <i32 0, i32 undef>)
+  %part.undef0 = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> %x, <2 x i32> <i32 0, i32 poison>)
   store volatile <2 x float> %part.undef0, ptr addrspace(1) undef
 
   %part.undef1 = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> %x, <2 x i32> <i32 undef, i32 0>)
diff --git a/llvm/test/Transforms/InstSimplify/mul.ll b/llvm/test/Transforms/InstSimplify/mul.ll
index 8ae7f1eaac92b..a1b03a30fe4f4 100644
--- a/llvm/test/Transforms/InstSimplify/mul.ll
+++ b/llvm/test/Transforms/InstSimplify/mul.ll
@@ -34,11 +34,11 @@ define <16 x i8> @mul_by_0_vec(<16 x i8> %a) {
   ret <16 x i8> %b
 }
 
-define <2 x i8> @mul_by_0_vec_undef_elt(<2 x i8> %a) {
-; CHECK-LABEL: @mul_by_0_vec_undef_elt(
+define <2 x i8> @mul_by_0_vec_poison_elt(<2 x i8> %a) {
+; CHECK-LABEL: @mul_by_0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i8> zeroinitializer
 ;
-  %b = mul <2 x i8> %a, <i8 undef, i8 0>
+  %b = mul <2 x i8> %a, <i8 poison, i8 0>
   ret <2 x i8> %b
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/negate.ll b/llvm/test/Transforms/InstSimplify/negate.ll
index d72a0db6d445c..d07029becd1fe 100644
--- a/llvm/test/Transforms/InstSimplify/negate.ll
+++ b/llvm/test/Transforms/InstSimplify/negate.ll
@@ -17,11 +17,11 @@ define <2 x i32> @negate_nuw_vec(<2 x i32> %x) {
   ret <2 x i32> %neg
 }
 
-define <2 x i32> @negate_nuw_vec_undef_elt(<2 x i32> %x) {
-; CHECK-LABEL: @negate_nuw_vec_undef_elt(
+define <2 x i32> @negate_nuw_vec_poison_elt(<2 x i32> %x) {
+; CHECK-LABEL: @negate_nuw_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i32> zeroinitializer
 ;
-  %neg = sub nuw <2 x i32> <i32 0, i32 undef>, %x
+  %neg = sub nuw <2 x i32> <i32 0, i32 poison>, %x
   ret <2 x i32> %neg
 }
 
@@ -43,12 +43,12 @@ define <2 x i8> @negate_zero_or_minsigned_nsw_vec(<2 x i8> %x) {
   ret <2 x i8> %neg
 }
 
-define <2 x i8> @negate_zero_or_minsigned_nsw_vec_undef_elt(<2 x i8> %x) {
-; CHECK-LABEL: @negate_zero_or_minsigned_nsw_vec_undef_elt(
+define <2 x i8> @negate_zero_or_minsigned_nsw_vec_poison_elt(<2 x i8> %x) {
+; CHECK-LABEL: @negate_zero_or_minsigned_nsw_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i8> zeroinitializer
 ;
   %signbit = shl <2 x i8> %x, <i8 7, i8 7>
-  %neg = sub nsw <2 x i8> <i8 undef, i8 0>, %signbit
+  %neg = sub nsw <2 x i8> <i8 poison, i8 0>, %signbit
   ret <2 x i8> %neg
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/or.ll b/llvm/test/Transforms/InstSimplify/or.ll
index 913b760dd331c..f241c6987b9e7 100644
--- a/llvm/test/Transforms/InstSimplify/or.ll
+++ b/llvm/test/Transforms/InstSimplify/or.ll
@@ -17,11 +17,11 @@ define i32 @all_ones(i32 %A) {
   ret i32 %B
 }
 
-define <3 x i8> @all_ones_vec_with_undef_elt(<3 x i8> %A) {
-; CHECK-LABEL: @all_ones_vec_with_undef_elt(
+define <3 x i8> @all_ones_vec_with_poison_elt(<3 x i8> %A) {
+; CHECK-LABEL: @all_ones_vec_with_poison_elt(
 ; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 -1, i8 -1>
 ;
-  %B = or <3 x i8> %A, <i8 -1, i8 undef, i8 -1>
+  %B = or <3 x i8> %A, <i8 -1, i8 poison, i8 -1>
   ret <3 x i8> %B
 }
 
@@ -68,11 +68,11 @@ define i32 @or_not(i32 %A) {
   ret i32 %B
 }
 
-define <2 x i4> @or_not_commute_vec_undef(<2 x i4> %A) {
-; CHECK-LABEL: @or_not_commute_vec_undef(
+define <2 x i4> @or_not_commute_vec_poison(<2 x i4> %A) {
+; CHECK-LABEL: @or_not_commute_vec_poison(
 ; CHECK-NEXT:    ret <2 x i4> <i4 -1, i4 -1>
 ;
-  %NotA = xor <2 x i4> %A, <i4 -1, i4 undef>
+  %NotA = xor <2 x i4> %A, <i4 -1, i4 poison>
   %B = or <2 x i4> %NotA, %A
   ret <2 x i4> %B
 }
@@ -335,7 +335,7 @@ define <2 x i1> @or_with_not_op_commute4(<2 x i1> %a, <2 x i1> %b) {
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %ab = and <2 x i1> %b, %a
-  %not = xor <2 x i1> %ab, <i1 -1, i1 undef>
+  %not = xor <2 x i1> %ab, <i1 -1, i1 poison>
   %r = or <2 x i1> %not, %a
   ret <2 x i1> %r
 }
@@ -515,6 +515,21 @@ define <2 x i4> @and_or_not_or_commute7_undef_elt(<2 x i4> %A, <2 x i4> %B) {
   ret <2 x i4> %r
 }
 
+; doing the same with poison is safe.
+
+define <2 x i4> @and_or_not_or_commute7_poison_elt(<2 x i4> %A, <2 x i4> %B) {
+; CHECK-LABEL: @and_or_not_or_commute7_poison_elt(
+; CHECK-NEXT:    [[NOTA:%.*]] = xor <2 x i4> [[A:%.*]], <i4 poison, i4 -1>
+; CHECK-NEXT:    ret <2 x i4> [[NOTA]]
+;
+  %nota = xor <2 x i4> %A, <i4 poison, i4 -1>
+  %and = and <2 x i4> %B, %nota
+  %or = or <2 x i4> %B, %A
+  %notab = xor <2 x i4> %or, <i4 -1, i4 -1>
+  %r = or <2 x i4> %notab, %and
+  ret <2 x i4> %r
+}
+
 ; (A | B) | (A ^ B) --> A | B
 
 define i69 @or_or_xor(i69 %A, i69 %B) {
@@ -769,6 +784,21 @@ define <2 x i4> @or_nxor_and_undef_elt(<2 x i4> %a, <2 x i4> %b) {
   ret <2 x i4> %r
 }
 
+; Same with poison is safe.
+
+define <2 x i4> @or_nxor_and_poison_elt(<2 x i4> %a, <2 x i4> %b) {
+; CHECK-LABEL: @or_nxor_and_poison_elt(
+; CHECK-NEXT:    [[XOR:%.*]] = xor <2 x i4> [[A:%.*]], [[B:%.*]]
+; CHECK-NEXT:    [[NOT:%.*]] = xor <2 x i4> [[XOR]], <i4 -1, i4 poison>
+; CHECK-NEXT:    ret <2 x i4> [[NOT]]
+;
+  %and = and <2 x i4> %b, %a
+  %xor = xor <2 x i4> %a, %b
+  %not = xor <2 x i4> %xor, <i4 -1, i4 poison>
+  %r = or <2 x i4> %not, %and
+  ret <2 x i4> %r
+}
+
 ; ~(A ^ B) | (A | B) --> -1
 
 define i4 @or_nxor_or_commute0(i4 %a, i4 %b) {
@@ -849,15 +879,15 @@ define i4 @or_nxor_or_wrong_val2(i4 %a, i4 %b, i4 %c) {
   ret i4 %r
 }
 
-; negative test - undef in 'not' is allowed
+; negative test - poison in 'not' is allowed
 
-define <2 x i4> @or_nxor_or_undef_elt(<2 x i4> %a, <2 x i4> %b) {
-; CHECK-LABEL: @or_nxor_or_undef_elt(
+define <2 x i4> @or_nxor_or_poison_elt(<2 x i4> %a, <2 x i4> %b) {
+; CHECK-LABEL: @or_nxor_or_poison_elt(
 ; CHECK-NEXT:    ret <2 x i4> <i4 -1, i4 -1>
 ;
   %or = or <2 x i4> %b, %a
   %xor = xor <2 x i4> %a, %b
-  %not = xor <2 x i4> %xor, <i4 -1, i4 undef>
+  %not = xor <2 x i4> %xor, <i4 -1, i4 poison>
   %r = or <2 x i4> %or, %not
   ret <2 x i4> %r
 }
@@ -966,12 +996,12 @@ define i32  @or_xor_not_op_or_commute7(i32 %a, i32 %b){
   ret i32  %r
 }
 
-define <2 x i4> @or_xor_not_op_or_undef_elt(<2 x i4> %a, <2 x i4> %b) {
-; CHECK-LABEL: @or_xor_not_op_or_undef_elt(
+define <2 x i4> @or_xor_not_op_or_poison_elt(<2 x i4> %a, <2 x i4> %b) {
+; CHECK-LABEL: @or_xor_not_op_or_poison_elt(
 ; CHECK-NEXT:    ret <2 x i4> <i4 -1, i4 -1>
 ;
   %xor = xor <2 x i4> %a, %b
-  %nota = xor <2 x i4> %a, <i4 -1, i4 undef>
+  %nota = xor <2 x i4> %a, <i4 -1, i4 poison>
   %or = or <2 x i4>  %nota, %b
   %r = or <2 x i4> %xor, %or
   ret <2 x i4> %r
@@ -1082,6 +1112,21 @@ define <2 x i4> @or_nand_xor_undef_elt(<2 x i4> %x, <2 x i4> %y) {
   ret <2 x i4> %or
 }
 
+; Same with poison is safe.
+
+define <2 x i4> @or_nand_xor_poison_elt(<2 x i4> %x, <2 x i4> %y) {
+; CHECK-LABEL: @or_nand_xor_poison_elt(
+; CHECK-NEXT:    [[AND:%.*]] = and <2 x i4> [[Y:%.*]], [[X:%.*]]
+; CHECK-NEXT:    [[NAND:%.*]] = xor <2 x i4> [[AND]], <i4 poison, i4 -1>
+; CHECK-NEXT:    ret <2 x i4> [[NAND]]
+;
+  %and = and <2 x i4> %y, %x
+  %xor = xor <2 x i4> %x, %y
+  %nand = xor <2 x i4> %and, <i4 poison, i4 -1>
+  %or = or <2 x i4> %xor, %nand
+  ret <2 x i4> %or
+}
+
 declare i32 @llvm.fshl.i32 (i32, i32, i32)
 declare i32 @llvm.fshr.i32 (i32, i32, i32)
 
diff --git a/llvm/test/Transforms/InstSimplify/ptrmask.ll b/llvm/test/Transforms/InstSimplify/ptrmask.ll
index dd83abfdeee46..d2c4a5dd7f035 100644
--- a/llvm/test/Transforms/InstSimplify/ptrmask.ll
+++ b/llvm/test/Transforms/InstSimplify/ptrmask.ll
@@ -40,7 +40,8 @@ define <2 x ptr addrspace(1) > @ptrmask_simplify_poison_and_zero_i32_vec_fail(<2
 define <2 x ptr> @ptrmask_simplify_undef_and_ones_vec(<2 x ptr> %p) {
 ; CHECK-LABEL: define <2 x ptr> @ptrmask_simplify_undef_and_ones_vec
 ; CHECK-SAME: (<2 x ptr> [[P:%.*]]) {
-; CHECK-NEXT:    ret <2 x ptr> [[P]]
+; CHECK-NEXT:    [[R:%.*]] = call <2 x ptr> @llvm.ptrmask.v2p0.v2i64(<2 x ptr> [[P]], <2 x i64> <i64 undef, i64 -1>)
+; CHECK-NEXT:    ret <2 x ptr> [[R]]
 ;
   %r = call <2 x ptr> @llvm.ptrmask.v2p1.v2i64(<2 x ptr> %p, <2 x i64> <i64 undef, i64 -1>)
   ret <2 x ptr> %r
diff --git a/llvm/test/Transforms/InstSimplify/rem.ll b/llvm/test/Transforms/InstSimplify/rem.ll
index 5af3b5f7c5e0b..a46db0342042f 100644
--- a/llvm/test/Transforms/InstSimplify/rem.ll
+++ b/llvm/test/Transforms/InstSimplify/rem.ll
@@ -17,11 +17,11 @@ define <2 x i32> @zero_dividend_vector(<2 x i32> %A) {
   ret <2 x i32> %B
 }
 
-define <2 x i32> @zero_dividend_vector_undef_elt(<2 x i32> %A) {
-; CHECK-LABEL: @zero_dividend_vector_undef_elt(
+define <2 x i32> @zero_dividend_vector_poison_elt(<2 x i32> %A) {
+; CHECK-LABEL: @zero_dividend_vector_poison_elt(
 ; CHECK-NEXT:    ret <2 x i32> zeroinitializer
 ;
-  %B = urem <2 x i32> <i32 undef, i32 0>, %A
+  %B = urem <2 x i32> <i32 poison, i32 0>, %A
   ret <2 x i32> %B
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/saturating-add-sub.ll b/llvm/test/Transforms/InstSimplify/saturating-add-sub.ll
index 6fb12612f2f72..40b22c619f768 100644
--- a/llvm/test/Transforms/InstSimplify/saturating-add-sub.ll
+++ b/llvm/test/Transforms/InstSimplify/saturating-add-sub.ll
@@ -44,7 +44,7 @@ define <2 x i8> @uadd_vector_0_commute(<2 x i8> %a) {
 ; CHECK-LABEL: @uadd_vector_0_commute(
 ; CHECK-NEXT:    ret <2 x i8> [[A:%.*]]
 ;
-  %x2v = call <2 x i8> @llvm.uadd.sat.v2i8(<2 x i8> <i8 0, i8 undef>, <2 x i8> %a)
+  %x2v = call <2 x i8> @llvm.uadd.sat.v2i8(<2 x i8> <i8 0, i8 poison>, <2 x i8> %a)
   ret <2 x i8> %x2v
 }
 
@@ -156,7 +156,7 @@ define <2 x i8> @sadd_vector_0(<2 x i8> %a) {
 ; CHECK-LABEL: @sadd_vector_0(
 ; CHECK-NEXT:    ret <2 x i8> [[A:%.*]]
 ;
-  %y1v = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> %a, <2 x i8> <i8 undef, i8 0>)
+  %y1v = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> %a, <2 x i8> <i8 poison, i8 0>)
   ret <2 x i8> %y1v
 }
 
@@ -205,10 +205,10 @@ define i8 @sadd_scalar_maxval_commute(i8 %a) {
 
 define <2 x i8> @sadd_vector_maxval_commute(<2 x i8> %a) {
 ; CHECK-LABEL: @sadd_vector_maxval_commute(
-; CHECK-NEXT:    [[Y4V:%.*]] = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> <i8 undef, i8 127>, <2 x i8> [[A:%.*]])
+; CHECK-NEXT:    [[Y4V:%.*]] = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> <i8 poison, i8 127>, <2 x i8> [[A:%.*]])
 ; CHECK-NEXT:    ret <2 x i8> [[Y4V]]
 ;
-  %y4v = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> <i8 undef, i8 127>, <2 x i8> %a)
+  %y4v = call <2 x i8> @llvm.sadd.sat.v2i8(<2 x i8> <i8 poison, i8 127>, <2 x i8> %a)
   ret <2 x i8> %y4v
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/sdiv.ll b/llvm/test/Transforms/InstSimplify/sdiv.ll
index 2514d90b01235..99092802cab02 100644
--- a/llvm/test/Transforms/InstSimplify/sdiv.ll
+++ b/llvm/test/Transforms/InstSimplify/sdiv.ll
@@ -158,11 +158,11 @@ define <2 x i32> @knownnegation_commute_vec_bad3(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %div
 }
 
-define <3 x i32> @negated_operand_vec_undef(<3 x i32> %x) {
-; CHECK-LABEL: @negated_operand_vec_undef(
+define <3 x i32> @negated_operand_vec_poison(<3 x i32> %x) {
+; CHECK-LABEL: @negated_operand_vec_poison(
 ; CHECK-NEXT:    ret <3 x i32> <i32 -1, i32 -1, i32 -1>
 ;
-  %negx = sub nsw <3 x i32> <i32 0, i32 undef, i32 0>, %x
+  %negx = sub nsw <3 x i32> <i32 0, i32 poison, i32 0>, %x
   %div = sdiv <3 x i32> %negx, %x
   ret <3 x i32> %div
 }
diff --git a/llvm/test/Transforms/InstSimplify/select-inseltpoison.ll b/llvm/test/Transforms/InstSimplify/select-inseltpoison.ll
index 2a4ce85ed11f8..fcf8c31b25eed 100644
--- a/llvm/test/Transforms/InstSimplify/select-inseltpoison.ll
+++ b/llvm/test/Transforms/InstSimplify/select-inseltpoison.ll
@@ -17,11 +17,11 @@ define <2 x i1> @bool_true_or_false_vec(<2 x i1> %cond) {
   ret <2 x i1> %s
 }
 
-define <2 x i1> @bool_true_or_false_vec_undef(<2 x i1> %cond) {
-; CHECK-LABEL: @bool_true_or_false_vec_undef(
+define <2 x i1> @bool_true_or_false_vec_poison(<2 x i1> %cond) {
+; CHECK-LABEL: @bool_true_or_false_vec_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[COND:%.*]]
 ;
-  %s = select <2 x i1> %cond, <2 x i1> <i1 undef, i1 true>, <2 x i1> <i1 false, i1 undef>
+  %s = select <2 x i1> %cond, <2 x i1> <i1 poison, i1 true>, <2 x i1> <i1 false, i1 poison>
   ret <2 x i1> %s
 }
 
@@ -57,27 +57,27 @@ define <2 x i32> @equal_arms_vec(<2 x i1> %cond, <2 x i32> %x) {
   ret <2 x i32> %V
 }
 
-define <2 x i32> @equal_arms_vec_undef(<2 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_undef(
+define <2 x i32> @equal_arms_vec_poison(<2 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_poison(
 ; CHECK-NEXT:    ret <2 x i32> <i32 42, i32 42>
 ;
-  %V = select <2 x i1> %cond, <2 x i32> <i32 42, i32 undef>, <2 x i32> <i32 undef, i32 42>
+  %V = select <2 x i1> %cond, <2 x i32> <i32 42, i32 poison>, <2 x i32> <i32 poison, i32 42>
   ret <2 x i32> %V
 }
 
-define <3 x float> @equal_arms_vec_less_undef(<3 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_less_undef(
+define <3 x float> @equal_arms_vec_less_poison(<3 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_less_poison(
 ; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float 4.200000e+01, float 4.300000e+01>
 ;
-  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float undef, float 43.0>, <3 x float> <float 42.0, float 42.0, float 43.0>
+  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float poison, float 43.0>, <3 x float> <float 42.0, float 42.0, float 43.0>
   ret <3 x float> %V
 }
 
-define <3 x float> @equal_arms_vec_more_undef(<3 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_more_undef(
-; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float undef, float 4.300000e+01>
+define <3 x float> @equal_arms_vec_more_poison(<3 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_more_poison(
+; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float poison, float 4.300000e+01>
 ;
-  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float undef, float undef>, <3 x float> <float undef, float undef, float 43.0>
+  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float poison, float poison>, <3 x float> <float poison, float poison, float 43.0>
   ret <3 x float> %V
 }
 
@@ -105,19 +105,19 @@ define <2 x i8> @vsel_mixedvec() {
   ret <2 x i8> %s
 }
 
-define <3 x i8> @vsel_undef_true_op(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @vsel_undef_true_op(
+define <3 x i8> @vsel_poison_true_op(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @vsel_poison_true_op(
 ; CHECK-NEXT:    ret <3 x i8> [[X:%.*]]
 ;
-  %s = select <3 x i1><i1 1, i1 undef, i1 1>, <3 x i8> %x, <3 x i8> %y
+  %s = select <3 x i1><i1 1, i1 poison, i1 1>, <3 x i8> %x, <3 x i8> %y
   ret <3 x i8> %s
 }
 
-define <3 x i4> @vsel_undef_false_op(<3 x i4> %x, <3 x i4> %y) {
-; CHECK-LABEL: @vsel_undef_false_op(
+define <3 x i4> @vsel_poison_false_op(<3 x i4> %x, <3 x i4> %y) {
+; CHECK-LABEL: @vsel_poison_false_op(
 ; CHECK-NEXT:    ret <3 x i4> [[Y:%.*]]
 ;
-  %s = select <3 x i1><i1 0, i1 undef, i1 undef>, <3 x i4> %x, <3 x i4> %y
+  %s = select <3 x i1><i1 0, i1 poison, i1 poison>, <3 x i4> %x, <3 x i4> %y
   ret <3 x i4> %s
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/select.ll b/llvm/test/Transforms/InstSimplify/select.ll
index fe93a0c3f2125..40c1460e3ebc3 100644
--- a/llvm/test/Transforms/InstSimplify/select.ll
+++ b/llvm/test/Transforms/InstSimplify/select.ll
@@ -25,11 +25,11 @@ define <2 x i1> @bool_true_or_false_vec(<2 x i1> %cond) {
   ret <2 x i1> %s
 }
 
-define <2 x i1> @bool_true_or_false_vec_undef(<2 x i1> %cond) {
-; CHECK-LABEL: @bool_true_or_false_vec_undef(
+define <2 x i1> @bool_true_or_false_vec_poison(<2 x i1> %cond) {
+; CHECK-LABEL: @bool_true_or_false_vec_poison(
 ; CHECK-NEXT:    ret <2 x i1> [[COND:%.*]]
 ;
-  %s = select <2 x i1> %cond, <2 x i1> <i1 undef, i1 true>, <2 x i1> <i1 false, i1 undef>
+  %s = select <2 x i1> %cond, <2 x i1> <i1 poison, i1 true>, <2 x i1> <i1 false, i1 poison>
   ret <2 x i1> %s
 }
 
@@ -65,27 +65,27 @@ define <2 x i32> @equal_arms_vec(<2 x i1> %cond, <2 x i32> %x) {
   ret <2 x i32> %V
 }
 
-define <2 x i32> @equal_arms_vec_undef(<2 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_undef(
+define <2 x i32> @equal_arms_vec_poison(<2 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_poison(
 ; CHECK-NEXT:    ret <2 x i32> <i32 42, i32 42>
 ;
-  %V = select <2 x i1> %cond, <2 x i32> <i32 42, i32 undef>, <2 x i32> <i32 undef, i32 42>
+  %V = select <2 x i1> %cond, <2 x i32> <i32 42, i32 poison>, <2 x i32> <i32 poison, i32 42>
   ret <2 x i32> %V
 }
 
-define <3 x float> @equal_arms_vec_less_undef(<3 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_less_undef(
+define <3 x float> @equal_arms_vec_less_poison(<3 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_less_poison(
 ; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float 4.200000e+01, float 4.300000e+01>
 ;
-  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float undef, float 43.0>, <3 x float> <float 42.0, float 42.0, float 43.0>
+  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float poison, float 43.0>, <3 x float> <float 42.0, float 42.0, float 43.0>
   ret <3 x float> %V
 }
 
-define <3 x float> @equal_arms_vec_more_undef(<3 x i1> %cond) {
-; CHECK-LABEL: @equal_arms_vec_more_undef(
-; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float undef, float 4.300000e+01>
+define <3 x float> @equal_arms_vec_more_poison(<3 x i1> %cond) {
+; CHECK-LABEL: @equal_arms_vec_more_poison(
+; CHECK-NEXT:    ret <3 x float> <float 4.200000e+01, float poison, float 4.300000e+01>
 ;
-  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float undef, float undef>, <3 x float> <float undef, float undef, float 43.0>
+  %V = select <3 x i1> %cond, <3 x float> <float 42.0, float poison, float poison>, <3 x float> <float poison, float poison, float 43.0>
   ret <3 x float> %V
 }
 
@@ -113,19 +113,19 @@ define <2 x i8> @vsel_mixedvec() {
   ret <2 x i8> %s
 }
 
-define <3 x i8> @vsel_undef_true_op(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @vsel_undef_true_op(
+define <3 x i8> @vsel_poison_true_op(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @vsel_poison_true_op(
 ; CHECK-NEXT:    ret <3 x i8> [[X:%.*]]
 ;
-  %s = select <3 x i1><i1 1, i1 undef, i1 1>, <3 x i8> %x, <3 x i8> %y
+  %s = select <3 x i1><i1 1, i1 poison, i1 1>, <3 x i8> %x, <3 x i8> %y
   ret <3 x i8> %s
 }
 
-define <3 x i4> @vsel_undef_false_op(<3 x i4> %x, <3 x i4> %y) {
-; CHECK-LABEL: @vsel_undef_false_op(
+define <3 x i4> @vsel_poison_false_op(<3 x i4> %x, <3 x i4> %y) {
+; CHECK-LABEL: @vsel_poison_false_op(
 ; CHECK-NEXT:    ret <3 x i4> [[Y:%.*]]
 ;
-  %s = select <3 x i1><i1 0, i1 undef, i1 undef>, <3 x i4> %x, <3 x i4> %y
+  %s = select <3 x i1><i1 0, i1 poison, i1 poison>, <3 x i4> %x, <3 x i4> %y
   ret <3 x i4> %s
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/shift.ll b/llvm/test/Transforms/InstSimplify/shift.ll
index b562c3c164d52..a816fcbdeeee0 100644
--- a/llvm/test/Transforms/InstSimplify/shift.ll
+++ b/llvm/test/Transforms/InstSimplify/shift.ll
@@ -17,11 +17,11 @@ define i41 @shl_0(i41 %X) {
   ret i41 %B
 }
 
-define <2 x i41> @shl_0_vec_undef_elt(<2 x i41> %X) {
-; CHECK-LABEL: @shl_0_vec_undef_elt(
+define <2 x i41> @shl_0_vec_poison_elt(<2 x i41> %X) {
+; CHECK-LABEL: @shl_0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i41> zeroinitializer
 ;
-  %B = shl <2 x i41> <i41 0, i41 undef>, %X
+  %B = shl <2 x i41> <i41 0, i41 poison>, %X
   ret <2 x i41> %B
 }
 
@@ -41,11 +41,11 @@ define i39 @ashr_0(i39 %X) {
   ret i39 %B
 }
 
-define <2 x i141> @ashr_0_vec_undef_elt(<2 x i141> %X) {
-; CHECK-LABEL: @ashr_0_vec_undef_elt(
+define <2 x i141> @ashr_0_vec_poison_elt(<2 x i141> %X) {
+; CHECK-LABEL: @ashr_0_vec_poison_elt(
 ; CHECK-NEXT:    ret <2 x i141> zeroinitializer
 ;
-  %B = shl <2 x i141> <i141 undef, i141 0>, %X
+  %B = shl <2 x i141> <i141 poison, i141 0>, %X
   ret <2 x i141> %B
 }
 
@@ -113,11 +113,11 @@ define i32 @ashr_all_ones(i32 %A) {
   ret i32 %B
 }
 
-define <3 x i8> @ashr_all_ones_vec_with_undef_elts(<3 x i8> %x, <3 x i8> %y) {
-; CHECK-LABEL: @ashr_all_ones_vec_with_undef_elts(
+define <3 x i8> @ashr_all_ones_vec_with_poison_elts(<3 x i8> %x, <3 x i8> %y) {
+; CHECK-LABEL: @ashr_all_ones_vec_with_poison_elts(
 ; CHECK-NEXT:    ret <3 x i8> <i8 -1, i8 -1, i8 -1>
 ;
-  %sh = ashr <3 x i8> <i8 undef, i8 -1, i8 undef>, %y
+  %sh = ashr <3 x i8> <i8 poison, i8 -1, i8 poison>, %y
   ret <3 x i8> %sh
 }
 
@@ -306,11 +306,22 @@ define <2 x i7> @all_ones_left_right_splat(<2 x i7> %x) {
 
 ; Poison could propagate, but undef must not.
 
-define <3 x i7> @all_ones_left_right_splat_poison_undef_elt(<3 x i7> %x) {
-; CHECK-LABEL: @all_ones_left_right_splat_poison_undef_elt(
+define <3 x i7> @all_ones_left_right_splat_undef_elt(<3 x i7> %x) {
+; CHECK-LABEL: @all_ones_left_right_splat_undef_elt(
+; CHECK-NEXT:    [[LEFT:%.*]] = shl <3 x i7> <i7 undef, i7 -1, i7 undef>, [[X:%.*]]
+; CHECK-NEXT:    [[RIGHT:%.*]] = ashr <3 x i7> [[LEFT]], [[X]]
+; CHECK-NEXT:    ret <3 x i7> [[RIGHT]]
+;
+  %left = shl <3 x i7> <i7 undef, i7 -1, i7 undef>, %x
+  %right = ashr <3 x i7> %left, %x
+  ret <3 x i7> %right
+}
+
+define <3 x i7> @all_ones_left_right_splat_poison__elt(<3 x i7> %x) {
+; CHECK-LABEL: @all_ones_left_right_splat_poison__elt(
 ; CHECK-NEXT:    ret <3 x i7> <i7 -1, i7 -1, i7 -1>
 ;
-  %left = shl <3 x i7> <i7 poison, i7 -1, i7 undef>, %x
+  %left = shl <3 x i7> <i7 poison, i7 -1, i7 poison>, %x
   %right = ashr <3 x i7> %left, %x
   ret <3 x i7> %right
 }
diff --git a/llvm/test/Transforms/InstSimplify/srem.ll b/llvm/test/Transforms/InstSimplify/srem.ll
index b1cbdf35b3c7c..ab726832e517b 100644
--- a/llvm/test/Transforms/InstSimplify/srem.ll
+++ b/llvm/test/Transforms/InstSimplify/srem.ll
@@ -39,11 +39,11 @@ define <2 x i32> @knownnegation_commute_vec(<2 x i32> %x, <2 x i32> %y) {
   ret <2 x i32> %rem
 }
 
-define <3 x i32> @negated_operand_vec_undef(<3 x i32> %x) {
-; CHECK-LABEL: @negated_operand_vec_undef(
+define <3 x i32> @negated_operand_vec_poison(<3 x i32> %x) {
+; CHECK-LABEL: @negated_operand_vec_poison(
 ; CHECK-NEXT:    ret <3 x i32> zeroinitializer
 ;
-  %negx = sub <3 x i32> <i32 0, i32 undef, i32 0>, %x
+  %negx = sub <3 x i32> <i32 0, i32 poison, i32 0>, %x
   %rem = srem <3 x i32> %negx, %x
   ret <3 x i32> %rem
 }
diff --git a/llvm/test/Transforms/InstSimplify/sub.ll b/llvm/test/Transforms/InstSimplify/sub.ll
index deb0ee33cd920..fd88fc15716c8 100644
--- a/llvm/test/Transforms/InstSimplify/sub.ll
+++ b/llvm/test/Transforms/InstSimplify/sub.ll
@@ -29,7 +29,7 @@ define <2 x i32> @sub_zero_vec(<2 x i32> %A) {
 ; CHECK-LABEL: @sub_zero_vec(
 ; CHECK-NEXT:    ret <2 x i32> [[A:%.*]]
 ;
-  %B = sub <2 x i32> %A, <i32 0, i32 undef>
+  %B = sub <2 x i32> %A, <i32 0, i32 poison>
   ret <2 x i32> %B
 }
 
@@ -46,8 +46,8 @@ define <2 x i32> @neg_neg_vec(<2 x i32> %A) {
 ; CHECK-LABEL: @neg_neg_vec(
 ; CHECK-NEXT:    ret <2 x i32> [[A:%.*]]
 ;
-  %B = sub <2 x i32> <i32 0, i32 undef>, %A
-  %C = sub <2 x i32> <i32 0, i32 undef>, %B
+  %B = sub <2 x i32> <i32 0, i32 poison>, %A
+  %C = sub <2 x i32> <i32 0, i32 poison>, %B
   ret <2 x i32> %C
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/xor.ll b/llvm/test/Transforms/InstSimplify/xor.ll
index 0e23cc66c1652..229e943a3836f 100644
--- a/llvm/test/Transforms/InstSimplify/xor.ll
+++ b/llvm/test/Transforms/InstSimplify/xor.ll
@@ -156,6 +156,20 @@ define <2 x i4> @xor_and_or_not_undef_elt(<2 x i4> %a, <2 x i4> %b) {
   ret <2 x i4> %r
 }
 
+; but correct to propagate poison element
+
+define <2 x i4> @xor_and_or_not_poison_elt(<2 x i4> %a, <2 x i4> %b) {
+; CHECK-LABEL: @xor_and_or_not_poison_elt(
+; CHECK-NEXT:    [[NOT:%.*]] = xor <2 x i4> [[A:%.*]], <i4 -1, i4 poison>
+; CHECK-NEXT:    ret <2 x i4> [[NOT]]
+;
+  %and = and <2 x i4> %b, %a
+  %not = xor <2 x i4> %a, <i4 -1, i4 poison>
+  %or = or <2 x i4> %not, %b
+  %r = xor <2 x i4> %or, %and
+  ret <2 x i4> %r
+}
+
 define i4 @xor_or_and_not_commute0(i4 %a, i4 %b) {
 ; CHECK-LABEL: @xor_or_and_not_commute0(
 ; CHECK-NEXT:    ret i4 [[A:%.*]]
@@ -277,11 +291,11 @@ define i4 @xor_or_and_not_wrong_val2(i4 %a, i4 %b, i4 %c) {
   ret i4 %r
 }
 
-define <2 x i4> @xor_or_and_not_undef_elt(<2 x i4> %a, <2 x i4> %b) {
-; CHECK-LABEL: @xor_or_and_not_undef_elt(
+define <2 x i4> @xor_or_and_not_poison_elt(<2 x i4> %a, <2 x i4> %b) {
+; CHECK-LABEL: @xor_or_and_not_poison_elt(
 ; CHECK-NEXT:    ret <2 x i4> [[A:%.*]]
 ;
-  %not = xor <2 x i4> %a, <i4 -1, i4 undef>
+  %not = xor <2 x i4> %a, <i4 -1, i4 poison>
   %and = and <2 x i4> %b, %not
   %or = or <2 x i4> %a, %b
   %r = xor <2 x i4> %or, %and
diff --git a/llvm/test/Transforms/Reassociate/inverses.ll b/llvm/test/Transforms/Reassociate/inverses.ll
index b6962c6872a9a..a9d0c4fb03222 100644
--- a/llvm/test/Transforms/Reassociate/inverses.ll
+++ b/llvm/test/Transforms/Reassociate/inverses.ll
@@ -12,12 +12,12 @@ define i32 @test1(i32 %a, i32 %b) {
   ret i32 %t5
 }
 
-define <2 x i32> @not_op_vec_undef(<2 x i32> %a, <2 x i32> %b) {
-; CHECK-LABEL: @not_op_vec_undef(
+define <2 x i32> @not_op_vec_poison(<2 x i32> %a, <2 x i32> %b) {
+; CHECK-LABEL: @not_op_vec_poison(
 ; CHECK-NEXT:    ret <2 x i32> zeroinitializer
 ;
   %t2 = and <2 x i32> %b, %a
-  %t4 = xor <2 x i32> %a, <i32 -1, i32 undef>
+  %t4 = xor <2 x i32> %a, <i32 -1, i32 poison>
   %t5 = and <2 x i32> %t2, %t4
   ret <2 x i32> %t5
 }
diff --git a/llvm/test/Transforms/Reassociate/negation.ll b/llvm/test/Transforms/Reassociate/negation.ll
index 4718d9d87ae1b..14ae86fb94aab 100644
--- a/llvm/test/Transforms/Reassociate/negation.ll
+++ b/llvm/test/Transforms/Reassociate/negation.ll
@@ -31,16 +31,16 @@ define i32 @test2(i32 %a, i32 %b, i32 %z) {
   ret i32 %f
 }
 
-define <2 x i32> @negate_vec_undefs(<2 x i32> %a, <2 x i32> %b, <2 x i32> %z) {
-; CHECK-LABEL: @negate_vec_undefs(
+define <2 x i32> @negate_vec_poisons(<2 x i32> %a, <2 x i32> %b, <2 x i32> %z) {
+; CHECK-LABEL: @negate_vec_poisons(
 ; CHECK-NEXT:    [[E:%.*]] = mul <2 x i32> [[A:%.*]], <i32 40, i32 40>
 ; CHECK-NEXT:    [[F:%.*]] = mul <2 x i32> [[E]], [[Z:%.*]]
 ; CHECK-NEXT:    ret <2 x i32> [[F]]
 ;
   %d = mul <2 x i32> %z, <i32 40, i32 40>
-  %c = sub <2 x i32> <i32 0, i32 undef>, %d
+  %c = sub <2 x i32> <i32 0, i32 poison>, %d
   %e = mul <2 x i32> %a, %c
-  %f = sub <2 x i32> <i32 0, i32 undef>, %e
+  %f = sub <2 x i32> <i32 0, i32 poison>, %e
   ret <2 x i32> %f
 }
 
diff --git a/llvm/unittests/IR/ConstantsTest.cpp b/llvm/unittests/IR/ConstantsTest.cpp
index 1d6a92c498b06..8f0a507c0fd18 100644
--- a/llvm/unittests/IR/ConstantsTest.cpp
+++ b/llvm/unittests/IR/ConstantsTest.cpp
@@ -581,7 +581,7 @@ TEST(ConstantsTest, containsUndefElemTest) {
   }
 }
 
-// Check that undefined elements in vector constants are matched
+// Check that poison elements in vector constants are matched
 // correctly for both integer and floating-point types. Just don't
 // crash on vectors of pointers (could be handled?).
 
@@ -590,6 +590,7 @@ TEST(ConstantsTest, isElementWiseEqual) {
 
   Type *Int32Ty = Type::getInt32Ty(Context);
   Constant *CU = UndefValue::get(Int32Ty);
+  Constant *CP = PoisonValue::get(Int32Ty);
   Constant *C1 = ConstantInt::get(Int32Ty, 1);
   Constant *C2 = ConstantInt::get(Int32Ty, 2);
 
@@ -597,15 +598,25 @@ TEST(ConstantsTest, isElementWiseEqual) {
   Constant *C12U1 = ConstantVector::get({C1, C2, CU, C1});
   Constant *C12U2 = ConstantVector::get({C1, C2, CU, C2});
   Constant *C12U21 = ConstantVector::get({C1, C2, CU, C2, C1});
+  Constant *C12P1 = ConstantVector::get({C1, C2, CP, C1});
+  Constant *C12P2 = ConstantVector::get({C1, C2, CP, C2});
+  Constant *C12P21 = ConstantVector::get({C1, C2, CP, C2, C1});
 
-  EXPECT_TRUE(C1211->isElementWiseEqual(C12U1));
-  EXPECT_TRUE(C12U1->isElementWiseEqual(C1211));
+  EXPECT_FALSE(C1211->isElementWiseEqual(C12U1));
+  EXPECT_FALSE(C12U1->isElementWiseEqual(C1211));
   EXPECT_FALSE(C12U2->isElementWiseEqual(C12U1));
   EXPECT_FALSE(C12U1->isElementWiseEqual(C12U2));
   EXPECT_FALSE(C12U21->isElementWiseEqual(C12U2));
 
+  EXPECT_TRUE(C1211->isElementWiseEqual(C12P1));
+  EXPECT_TRUE(C12P1->isElementWiseEqual(C1211));
+  EXPECT_FALSE(C12P2->isElementWiseEqual(C12P1));
+  EXPECT_FALSE(C12P1->isElementWiseEqual(C12P2));
+  EXPECT_FALSE(C12P21->isElementWiseEqual(C12P2));
+
   Type *FltTy = Type::getFloatTy(Context);
   Constant *CFU = UndefValue::get(FltTy);
+  Constant *CFP = PoisonValue::get(FltTy);
   Constant *CF1 = ConstantFP::get(FltTy, 1.0);
   Constant *CF2 = ConstantFP::get(FltTy, 2.0);
 
@@ -613,25 +624,41 @@ TEST(ConstantsTest, isElementWiseEqual) {
   Constant *CF12U1 = ConstantVector::get({CF1, CF2, CFU, CF1});
   Constant *CF12U2 = ConstantVector::get({CF1, CF2, CFU, CF2});
   Constant *CFUU1U = ConstantVector::get({CFU, CFU, CF1, CFU});
+  Constant *CF12P1 = ConstantVector::get({CF1, CF2, CFP, CF1});
+  Constant *CF12P2 = ConstantVector::get({CF1, CF2, CFP, CF2});
+  Constant *CFPP1P = ConstantVector::get({CFP, CFP, CF1, CFP});
 
-  EXPECT_TRUE(CF1211->isElementWiseEqual(CF12U1));
-  EXPECT_TRUE(CF12U1->isElementWiseEqual(CF1211));
-  EXPECT_TRUE(CFUU1U->isElementWiseEqual(CF12U1));
+  EXPECT_FALSE(CF1211->isElementWiseEqual(CF12U1));
+  EXPECT_FALSE(CF12U1->isElementWiseEqual(CF1211));
+  EXPECT_FALSE(CFUU1U->isElementWiseEqual(CF12U1));
   EXPECT_FALSE(CF12U2->isElementWiseEqual(CF12U1));
   EXPECT_FALSE(CF12U1->isElementWiseEqual(CF12U2));
 
+  EXPECT_TRUE(CF1211->isElementWiseEqual(CF12P1));
+  EXPECT_TRUE(CF12P1->isElementWiseEqual(CF1211));
+  EXPECT_TRUE(CFPP1P->isElementWiseEqual(CF12P1));
+  EXPECT_FALSE(CF12P2->isElementWiseEqual(CF12P1));
+  EXPECT_FALSE(CF12P1->isElementWiseEqual(CF12P2));
+
   PointerType *PtrTy = PointerType::get(Context, 0);
   Constant *CPU = UndefValue::get(PtrTy);
+  Constant *CPP = PoisonValue::get(PtrTy);
   Constant *CP0 = ConstantPointerNull::get(PtrTy);
 
   Constant *CP0000 = ConstantVector::get({CP0, CP0, CP0, CP0});
   Constant *CP00U0 = ConstantVector::get({CP0, CP0, CPU, CP0});
   Constant *CP00U = ConstantVector::get({CP0, CP0, CPU});
+  Constant *CP00P0 = ConstantVector::get({CP0, CP0, CPP, CP0});
+  Constant *CP00P = ConstantVector::get({CP0, CP0, CPP});
 
   EXPECT_FALSE(CP0000->isElementWiseEqual(CP00U0));
   EXPECT_FALSE(CP00U0->isElementWiseEqual(CP0000));
   EXPECT_FALSE(CP0000->isElementWiseEqual(CP00U));
   EXPECT_FALSE(CP00U->isElementWiseEqual(CP00U0));
+  EXPECT_FALSE(CP0000->isElementWiseEqual(CP00P0));
+  EXPECT_FALSE(CP00P0->isElementWiseEqual(CP0000));
+  EXPECT_FALSE(CP0000->isElementWiseEqual(CP00P));
+  EXPECT_FALSE(CP00P->isElementWiseEqual(CP00P0));
 }
 
 // Check that vector/aggregate constants correctly store undef and poison
diff --git a/llvm/unittests/IR/PatternMatch.cpp b/llvm/unittests/IR/PatternMatch.cpp
index 4d0c2e4220fec..133012684d16d 100644
--- a/llvm/unittests/IR/PatternMatch.cpp
+++ b/llvm/unittests/IR/PatternMatch.cpp
@@ -1184,6 +1184,8 @@ TEST_F(PatternMatchTest, VectorUndefInt) {
   Type *VectorTy = FixedVectorType::get(ScalarTy, 4);
   Constant *ScalarUndef = UndefValue::get(ScalarTy);
   Constant *VectorUndef = UndefValue::get(VectorTy);
+  Constant *ScalarPoison = PoisonValue::get(ScalarTy);
+  Constant *VectorPoison = PoisonValue::get(VectorTy);
   Constant *ScalarZero = Constant::getNullValue(ScalarTy);
   Constant *VectorZero = Constant::getNullValue(VectorTy);
 
@@ -1194,17 +1196,30 @@ TEST_F(PatternMatchTest, VectorUndefInt) {
   Elems.push_back(ScalarZero);
   Constant *VectorZeroUndef = ConstantVector::get(Elems);
 
+  SmallVector<Constant *, 4> Elems2;
+  Elems2.push_back(ScalarPoison);
+  Elems2.push_back(ScalarZero);
+  Elems2.push_back(ScalarPoison);
+  Elems2.push_back(ScalarZero);
+  Constant *VectorZeroPoison = ConstantVector::get(Elems2);
+
   EXPECT_TRUE(match(ScalarUndef, m_Undef()));
+  EXPECT_TRUE(match(ScalarPoison, m_Undef()));
   EXPECT_TRUE(match(VectorUndef, m_Undef()));
+  EXPECT_TRUE(match(VectorPoison, m_Undef()));
   EXPECT_FALSE(match(ScalarZero, m_Undef()));
   EXPECT_FALSE(match(VectorZero, m_Undef()));
   EXPECT_FALSE(match(VectorZeroUndef, m_Undef()));
+  EXPECT_FALSE(match(VectorZeroPoison, m_Undef()));
 
   EXPECT_FALSE(match(ScalarUndef, m_Zero()));
+  EXPECT_FALSE(match(ScalarPoison, m_Zero()));
   EXPECT_FALSE(match(VectorUndef, m_Zero()));
+  EXPECT_FALSE(match(VectorPoison, m_Zero()));
+  EXPECT_FALSE(match(VectorZeroUndef, m_Zero()));
   EXPECT_TRUE(match(ScalarZero, m_Zero()));
   EXPECT_TRUE(match(VectorZero, m_Zero()));
-  EXPECT_TRUE(match(VectorZeroUndef, m_Zero()));
+  EXPECT_TRUE(match(VectorZeroPoison, m_Zero()));
 
   const APInt *C;
   // Regardless of whether undefs are allowed,
@@ -1249,6 +1264,8 @@ TEST_F(PatternMatchTest, VectorUndefFloat) {
   Type *VectorTy = FixedVectorType::get(ScalarTy, 4);
   Constant *ScalarUndef = UndefValue::get(ScalarTy);
   Constant *VectorUndef = UndefValue::get(VectorTy);
+  Constant *ScalarPoison = PoisonValue::get(ScalarTy);
+  Constant *VectorPoison = PoisonValue::get(VectorTy);
   Constant *ScalarZero = Constant::getNullValue(ScalarTy);
   Constant *VectorZero = Constant::getNullValue(VectorTy);
   Constant *ScalarPosInf = ConstantFP::getInfinity(ScalarTy, false);
@@ -1258,72 +1275,116 @@ TEST_F(PatternMatchTest, VectorUndefFloat) {
   Constant *VectorZeroUndef =
       ConstantVector::get({ScalarUndef, ScalarZero, ScalarUndef, ScalarZero});
 
+  Constant *VectorZeroPoison =
+      ConstantVector::get({ScalarPoison, ScalarZero, ScalarPoison, ScalarZero});
+
   Constant *VectorInfUndef = ConstantVector::get(
       {ScalarPosInf, ScalarNegInf, ScalarUndef, ScalarPosInf});
 
+  Constant *VectorInfPoison = ConstantVector::get(
+      {ScalarPosInf, ScalarNegInf, ScalarPoison, ScalarPosInf});
+
   Constant *VectorNaNUndef =
       ConstantVector::get({ScalarUndef, ScalarNaN, ScalarNaN, ScalarNaN});
 
+  Constant *VectorNaNPoison =
+      ConstantVector::get({ScalarPoison, ScalarNaN, ScalarNaN, ScalarNaN});
+
   EXPECT_TRUE(match(ScalarUndef, m_Undef()));
   EXPECT_TRUE(match(VectorUndef, m_Undef()));
+  EXPECT_TRUE(match(ScalarPoison, m_Undef()));
+  EXPECT_TRUE(match(VectorPoison, m_Undef()));
   EXPECT_FALSE(match(ScalarZero, m_Undef()));
   EXPECT_FALSE(match(VectorZero, m_Undef()));
   EXPECT_FALSE(match(VectorZeroUndef, m_Undef()));
   EXPECT_FALSE(match(VectorInfUndef, m_Undef()));
   EXPECT_FALSE(match(VectorNaNUndef, m_Undef()));
+  EXPECT_FALSE(match(VectorZeroPoison, m_Undef()));
+  EXPECT_FALSE(match(VectorInfPoison, m_Undef()));
+  EXPECT_FALSE(match(VectorNaNPoison, m_Undef()));
 
   EXPECT_FALSE(match(ScalarUndef, m_AnyZeroFP()));
   EXPECT_FALSE(match(VectorUndef, m_AnyZeroFP()));
+  EXPECT_FALSE(match(ScalarPoison, m_AnyZeroFP()));
+  EXPECT_FALSE(match(VectorPoison, m_AnyZeroFP()));
   EXPECT_TRUE(match(ScalarZero, m_AnyZeroFP()));
   EXPECT_TRUE(match(VectorZero, m_AnyZeroFP()));
-  EXPECT_TRUE(match(VectorZeroUndef, m_AnyZeroFP()));
+  EXPECT_FALSE(match(VectorZeroUndef, m_AnyZeroFP()));
   EXPECT_FALSE(match(VectorInfUndef, m_AnyZeroFP()));
   EXPECT_FALSE(match(VectorNaNUndef, m_AnyZeroFP()));
+  EXPECT_TRUE(match(VectorZeroPoison, m_AnyZeroFP()));
+  EXPECT_FALSE(match(VectorInfPoison, m_AnyZeroFP()));
+  EXPECT_FALSE(match(VectorNaNPoison, m_AnyZeroFP()));
 
   EXPECT_FALSE(match(ScalarUndef, m_NaN()));
   EXPECT_FALSE(match(VectorUndef, m_NaN()));
   EXPECT_FALSE(match(VectorZeroUndef, m_NaN()));
+  EXPECT_FALSE(match(ScalarPoison, m_NaN()));
+  EXPECT_FALSE(match(VectorPoison, m_NaN()));
+  EXPECT_FALSE(match(VectorZeroPoison, m_NaN()));
   EXPECT_FALSE(match(ScalarPosInf, m_NaN()));
   EXPECT_FALSE(match(ScalarNegInf, m_NaN()));
   EXPECT_TRUE(match(ScalarNaN, m_NaN()));
   EXPECT_FALSE(match(VectorInfUndef, m_NaN()));
-  EXPECT_TRUE(match(VectorNaNUndef, m_NaN()));
+  EXPECT_FALSE(match(VectorNaNUndef, m_NaN()));
+  EXPECT_FALSE(match(VectorInfPoison, m_NaN()));
+  EXPECT_TRUE(match(VectorNaNPoison, m_NaN()));
 
   EXPECT_FALSE(match(ScalarUndef, m_NonNaN()));
   EXPECT_FALSE(match(VectorUndef, m_NonNaN()));
-  EXPECT_TRUE(match(VectorZeroUndef, m_NonNaN()));
+  EXPECT_FALSE(match(VectorZeroUndef, m_NonNaN()));
+  EXPECT_FALSE(match(ScalarPoison, m_NonNaN()));
+  EXPECT_FALSE(match(VectorPoison, m_NonNaN()));
+  EXPECT_TRUE(match(VectorZeroPoison, m_NonNaN()));
   EXPECT_TRUE(match(ScalarPosInf, m_NonNaN()));
   EXPECT_TRUE(match(ScalarNegInf, m_NonNaN()));
   EXPECT_FALSE(match(ScalarNaN, m_NonNaN()));
-  EXPECT_TRUE(match(VectorInfUndef, m_NonNaN()));
+  EXPECT_FALSE(match(VectorInfUndef, m_NonNaN()));
   EXPECT_FALSE(match(VectorNaNUndef, m_NonNaN()));
+  EXPECT_TRUE(match(VectorInfPoison, m_NonNaN()));
+  EXPECT_FALSE(match(VectorNaNPoison, m_NonNaN()));
 
   EXPECT_FALSE(match(ScalarUndef, m_Inf()));
   EXPECT_FALSE(match(VectorUndef, m_Inf()));
   EXPECT_FALSE(match(VectorZeroUndef, m_Inf()));
+  EXPECT_FALSE(match(ScalarPoison, m_Inf()));
+  EXPECT_FALSE(match(VectorPoison, m_Inf()));
+  EXPECT_FALSE(match(VectorZeroPoison, m_Inf()));
   EXPECT_TRUE(match(ScalarPosInf, m_Inf()));
   EXPECT_TRUE(match(ScalarNegInf, m_Inf()));
   EXPECT_FALSE(match(ScalarNaN, m_Inf()));
-  EXPECT_TRUE(match(VectorInfUndef, m_Inf()));
+  EXPECT_FALSE(match(VectorInfUndef, m_Inf()));
   EXPECT_FALSE(match(VectorNaNUndef, m_Inf()));
+  EXPECT_TRUE(match(VectorInfPoison, m_Inf()));
+  EXPECT_FALSE(match(VectorNaNPoison, m_Inf()));
 
   EXPECT_FALSE(match(ScalarUndef, m_NonInf()));
   EXPECT_FALSE(match(VectorUndef, m_NonInf()));
-  EXPECT_TRUE(match(VectorZeroUndef, m_NonInf()));
+  EXPECT_FALSE(match(VectorZeroUndef, m_NonInf()));
+  EXPECT_FALSE(match(ScalarPoison, m_NonInf()));
+  EXPECT_FALSE(match(VectorPoison, m_NonInf()));
+  EXPECT_TRUE(match(VectorZeroPoison, m_NonInf()));
   EXPECT_FALSE(match(ScalarPosInf, m_NonInf()));
   EXPECT_FALSE(match(ScalarNegInf, m_NonInf()));
   EXPECT_TRUE(match(ScalarNaN, m_NonInf()));
   EXPECT_FALSE(match(VectorInfUndef, m_NonInf()));
-  EXPECT_TRUE(match(VectorNaNUndef, m_NonInf()));
+  EXPECT_FALSE(match(VectorNaNUndef, m_NonInf()));
+  EXPECT_FALSE(match(VectorInfPoison, m_NonInf()));
+  EXPECT_TRUE(match(VectorNaNPoison, m_NonInf()));
 
   EXPECT_FALSE(match(ScalarUndef, m_Finite()));
   EXPECT_FALSE(match(VectorUndef, m_Finite()));
-  EXPECT_TRUE(match(VectorZeroUndef, m_Finite()));
+  EXPECT_FALSE(match(VectorZeroUndef, m_Finite()));
+  EXPECT_FALSE(match(ScalarPoison, m_Finite()));
+  EXPECT_FALSE(match(VectorPoison, m_Finite()));
+  EXPECT_TRUE(match(VectorZeroPoison, m_Finite()));
   EXPECT_FALSE(match(ScalarPosInf, m_Finite()));
   EXPECT_FALSE(match(ScalarNegInf, m_Finite()));
   EXPECT_FALSE(match(ScalarNaN, m_Finite()));
   EXPECT_FALSE(match(VectorInfUndef, m_Finite()));
   EXPECT_FALSE(match(VectorNaNUndef, m_Finite()));
+  EXPECT_FALSE(match(VectorInfPoison, m_Finite()));
+  EXPECT_FALSE(match(VectorNaNPoison, m_Finite()));
 
   const APFloat *C;
   // Regardless of whether undefs are allowed,
@@ -1707,38 +1768,57 @@ TEST_F(PatternMatchTest, ConstantPredicateType) {
 
   Constant *CMixedU32 = ConstantVector::get({CU32Max, CU32Zero, CU32DeadBeef});
   Constant *CU32Undef = UndefValue::get(U32Ty);
+  Constant *CU32Poison = PoisonValue::get(U32Ty);
   Constant *CU32MaxWithUndef =
       ConstantVector::get({CU32Undef, CU32Max, CU32Undef});
+  Constant *CU32MaxWithPoison =
+      ConstantVector::get({CU32Poison, CU32Max, CU32Poison});
 
   EXPECT_FALSE(match(CMixedU32, cst_pred_ty<is_unsigned_max_pred>()));
   EXPECT_FALSE(match(CMixedU32, cst_pred_ty<is_unsigned_zero_pred>()));
   EXPECT_TRUE(match(CMixedU32, cst_pred_ty<always_true_pred<APInt>>()));
   EXPECT_FALSE(match(CMixedU32, cst_pred_ty<always_false_pred<APInt>>()));
 
-  EXPECT_TRUE(match(CU32MaxWithUndef, cst_pred_ty<is_unsigned_max_pred>()));
+  EXPECT_FALSE(match(CU32MaxWithUndef, cst_pred_ty<is_unsigned_max_pred>()));
   EXPECT_FALSE(match(CU32MaxWithUndef, cst_pred_ty<is_unsigned_zero_pred>()));
-  EXPECT_TRUE(match(CU32MaxWithUndef, cst_pred_ty<always_true_pred<APInt>>()));
+  EXPECT_FALSE(match(CU32MaxWithUndef, cst_pred_ty<always_true_pred<APInt>>()));
   EXPECT_FALSE(
       match(CU32MaxWithUndef, cst_pred_ty<always_false_pred<APInt>>()));
 
+  EXPECT_TRUE(match(CU32MaxWithPoison, cst_pred_ty<is_unsigned_max_pred>()));
+  EXPECT_FALSE(match(CU32MaxWithPoison, cst_pred_ty<is_unsigned_zero_pred>()));
+  EXPECT_TRUE(match(CU32MaxWithPoison, cst_pred_ty<always_true_pred<APInt>>()));
+  EXPECT_FALSE(
+      match(CU32MaxWithPoison, cst_pred_ty<always_false_pred<APInt>>()));
+
   // Float arbitrary vector
 
   Constant *CMixedF32 = ConstantVector::get({CF32NaN, CF32Zero, CF32Pi});
   Constant *CF32Undef = UndefValue::get(F32Ty);
+  Constant *CF32Poison = PoisonValue::get(F32Ty);
   Constant *CF32NaNWithUndef =
       ConstantVector::get({CF32Undef, CF32NaN, CF32Undef});
+  Constant *CF32NaNWithPoison =
+      ConstantVector::get({CF32Poison, CF32NaN, CF32Poison});
 
   EXPECT_FALSE(match(CMixedF32, cstfp_pred_ty<is_float_nan_pred>()));
   EXPECT_FALSE(match(CMixedF32, cstfp_pred_ty<is_float_zero_pred>()));
   EXPECT_TRUE(match(CMixedF32, cstfp_pred_ty<always_true_pred<APFloat>>()));
   EXPECT_FALSE(match(CMixedF32, cstfp_pred_ty<always_false_pred<APFloat>>()));
 
-  EXPECT_TRUE(match(CF32NaNWithUndef, cstfp_pred_ty<is_float_nan_pred>()));
+  EXPECT_FALSE(match(CF32NaNWithUndef, cstfp_pred_ty<is_float_nan_pred>()));
   EXPECT_FALSE(match(CF32NaNWithUndef, cstfp_pred_ty<is_float_zero_pred>()));
-  EXPECT_TRUE(
+  EXPECT_FALSE(
       match(CF32NaNWithUndef, cstfp_pred_ty<always_true_pred<APFloat>>()));
   EXPECT_FALSE(
       match(CF32NaNWithUndef, cstfp_pred_ty<always_false_pred<APFloat>>()));
+
+  EXPECT_TRUE(match(CF32NaNWithPoison, cstfp_pred_ty<is_float_nan_pred>()));
+  EXPECT_FALSE(match(CF32NaNWithPoison, cstfp_pred_ty<is_float_zero_pred>()));
+  EXPECT_TRUE(
+      match(CF32NaNWithPoison, cstfp_pred_ty<always_true_pred<APFloat>>()));
+  EXPECT_FALSE(
+      match(CF32NaNWithPoison, cstfp_pred_ty<always_false_pred<APFloat>>()));
 }
 
 TEST_F(PatternMatchTest, InsertValue) {
@@ -1888,35 +1968,44 @@ TEST_F(PatternMatchTest, NotForbidUndef) {
   Type *ScalarTy = IRB.getInt8Ty();
   Type *VectorTy = FixedVectorType::get(ScalarTy, 3);
   Constant *ScalarUndef = UndefValue::get(ScalarTy);
+  Constant *ScalarPoison = PoisonValue::get(ScalarTy);
   Constant *ScalarOnes = Constant::getAllOnesValue(ScalarTy);
   Constant *VectorZero = Constant::getNullValue(VectorTy);
   Constant *VectorOnes = Constant::getAllOnesValue(VectorTy);
 
-  SmallVector<Constant *, 3> MixedElems;
-  MixedElems.push_back(ScalarOnes);
-  MixedElems.push_back(ScalarOnes);
-  MixedElems.push_back(ScalarUndef);
-  Constant *VectorMixed = ConstantVector::get(MixedElems);
+  SmallVector<Constant *, 3> MixedElemsUndef;
+  MixedElemsUndef.push_back(ScalarOnes);
+  MixedElemsUndef.push_back(ScalarOnes);
+  MixedElemsUndef.push_back(ScalarUndef);
+  Constant *VectorMixedUndef = ConstantVector::get(MixedElemsUndef);
+
+  SmallVector<Constant *, 3> MixedElemsPoison;
+  MixedElemsPoison.push_back(ScalarOnes);
+  MixedElemsPoison.push_back(ScalarOnes);
+  MixedElemsPoison.push_back(ScalarPoison);
+  Constant *VectorMixedPoison = ConstantVector::get(MixedElemsPoison);
 
   Value *Not = IRB.CreateXor(VectorZero, VectorOnes);
   Value *X;
-  EXPECT_TRUE(match(Not, m_Not(m_Value())));
-  EXPECT_TRUE(match(Not, m_NotForbidUndef(m_Value(X))));
+  EXPECT_TRUE(match(Not, m_Not(m_Value(X))));
   EXPECT_TRUE(match(X, m_Zero()));
 
   Value *NotCommute = IRB.CreateXor(VectorOnes, VectorZero);
   Value *Y;
-  EXPECT_TRUE(match(NotCommute, m_Not(m_Value())));
-  EXPECT_TRUE(match(NotCommute, m_NotForbidUndef(m_Value(Y))));
+  EXPECT_TRUE(match(NotCommute, m_Not(m_Value(Y))));
   EXPECT_TRUE(match(Y, m_Zero()));
 
-  Value *NotWithUndefs = IRB.CreateXor(VectorZero, VectorMixed);
-  EXPECT_TRUE(match(NotWithUndefs, m_Not(m_Value())));
-  EXPECT_FALSE(match(NotWithUndefs, m_NotForbidUndef(m_Value())));
+  Value *NotWithUndefs = IRB.CreateXor(VectorZero, VectorMixedUndef);
+  EXPECT_FALSE(match(NotWithUndefs, m_Not(m_Value())));
+
+  Value *NotWithPoisons = IRB.CreateXor(VectorZero, VectorMixedPoison);
+  EXPECT_TRUE(match(NotWithPoisons, m_Not(m_Value())));
+
+  Value *NotWithUndefsCommute = IRB.CreateXor(VectorMixedUndef, VectorZero);
+  EXPECT_FALSE(match(NotWithUndefsCommute, m_Not(m_Value())));
 
-  Value *NotWithUndefsCommute = IRB.CreateXor(VectorMixed, VectorZero);
-  EXPECT_TRUE(match(NotWithUndefsCommute, m_Not(m_Value())));
-  EXPECT_FALSE(match(NotWithUndefsCommute, m_NotForbidUndef(m_Value(X))));
+  Value *NotWithPoisonsCommute = IRB.CreateXor(VectorMixedPoison, VectorZero);
+  EXPECT_TRUE(match(NotWithPoisonsCommute, m_Not(m_Value())));
 }
 
 template <typename T> struct MutableConstTest : PatternMatchTest { };

From 971237dab259bdaa619403fc6472bd1758d4dc18 Mon Sep 17 00:00:00 2001
From: jeanPerier <jperier@nvidia.com>
Date: Wed, 17 Apr 2024 11:31:29 +0200
Subject: [PATCH 714/886] [flang] Retain internal and BIND(C) host procedure
 link in FIR (#87796)

Currently, it is not possible to find back which fun.func is the host
procedure of some internal procedure because the mangling of the
internal procedure does not contain info about the BIND(C) name of the
host.
This info may be useful to ensure dwarf DW_TAG_subprogram of internal
procedures are nested under DW_TAG_subprogram of host procedures for
instance.
---
 flang/include/flang/Lower/CallInterface.h     |  4 --
 .../flang/Optimizer/Dialect/FIROpsSupport.h   | 10 ++---
 flang/lib/Lower/CallInterface.cpp             | 40 +++++++++++++------
 .../Optimizer/Transforms/ArrayValueCopy.cpp   |  2 +-
 .../HLFIR/internal-procedures-bindc-host.f90  | 39 ++++++++++++++++++
 .../test/Lower/HLFIR/internal-procedures.f90  |  6 +--
 flang/test/Lower/OpenACC/acc-routine04.f90    |  2 +-
 .../threadprivate-host-association-2.f90      |  2 +-
 .../OpenMP/threadprivate-host-association.f90 |  2 +-
 flang/test/Lower/character-elemental.f90      | 15 ++++---
 .../Lower/equivalence-with-host-assoc.f90     | 16 ++++----
 .../Lower/explicit-interface-results-2.f90    |  4 +-
 .../test/Lower/host-associated-functions.f90  |  6 +--
 flang/test/Lower/host-associated-globals.f90  |  6 +--
 flang/test/Lower/host-associated.f90          | 32 +++++++--------
 flang/test/Lower/polymorphic.f90              |  2 +-
 16 files changed, 119 insertions(+), 69 deletions(-)
 create mode 100644 flang/test/Lower/HLFIR/internal-procedures-bindc-host.f90

diff --git a/flang/include/flang/Lower/CallInterface.h b/flang/include/flang/Lower/CallInterface.h
index 80b0576425377..a11e81b6593de 100644
--- a/flang/include/flang/Lower/CallInterface.h
+++ b/flang/include/flang/Lower/CallInterface.h
@@ -391,9 +391,6 @@ class CallerInterface : public CallInterface<CallerInterface> {
     llvm_unreachable("getting host associated type in CallerInterface");
   }
 
-  /// Set attributes on MLIR function.
-  void setFuncAttrs(mlir::func::FuncOp) const {}
-
 private:
   /// Check that the input vector is complete.
   bool verifyActualInputs() const;
@@ -444,7 +441,6 @@ class CalleeInterface : public CallInterface<CalleeInterface> {
   bool hasHostAssociated() const;
   mlir::Type getHostAssociatedTy() const;
   mlir::Value getHostAssociatedTuple() const;
-  void setFuncAttrs(mlir::func::FuncOp) const;
 
 private:
   Fortran::lower::pft::FunctionLikeUnit &funit;
diff --git a/flang/include/flang/Optimizer/Dialect/FIROpsSupport.h b/flang/include/flang/Optimizer/Dialect/FIROpsSupport.h
index 3266ea3aa7fdc..46b62d8de8d37 100644
--- a/flang/include/flang/Optimizer/Dialect/FIROpsSupport.h
+++ b/flang/include/flang/Optimizer/Dialect/FIROpsSupport.h
@@ -104,9 +104,9 @@ static constexpr llvm::StringRef getHostAssocAttrName() {
   return "fir.host_assoc";
 }
 
-/// Attribute to mark an internal procedure.
-static constexpr llvm::StringRef getInternalProcedureAttrName() {
-  return "fir.internal_proc";
+/// Attribute to link an internal procedure to its host procedure symbol.
+static constexpr llvm::StringRef getHostSymbolAttrName() {
+  return "fir.host_symbol";
 }
 
 /// Attribute containing the original name of a function from before the
@@ -122,8 +122,8 @@ bool hasHostAssociationArgument(mlir::func::FuncOp func);
 /// Is the function, \p func an internal procedure ?
 /// Some internal procedures may have access to saved host procedure
 /// variables even when they do not have a tuple argument.
-inline bool isInternalPorcedure(mlir::func::FuncOp func) {
-  return func->hasAttr(fir::getInternalProcedureAttrName());
+inline bool isInternalProcedure(mlir::func::FuncOp func) {
+  return func->hasAttr(fir::getHostSymbolAttrName());
 }
 
 /// Tell if \p value is:
diff --git a/flang/lib/Lower/CallInterface.cpp b/flang/lib/Lower/CallInterface.cpp
index 05a0c10c70974..2d4d17a2ef12e 100644
--- a/flang/lib/Lower/CallInterface.cpp
+++ b/flang/lib/Lower/CallInterface.cpp
@@ -575,13 +575,6 @@ mlir::Value Fortran::lower::CalleeInterface::getHostAssociatedTuple() const {
   return converter.hostAssocTupleValue();
 }
 
-void Fortran::lower::CalleeInterface::setFuncAttrs(
-    mlir::func::FuncOp func) const {
-  if (funit.parentHasHostAssoc())
-    func->setAttr(fir::getInternalProcedureAttrName(),
-                  mlir::UnitAttr::get(func->getContext()));
-}
-
 //===----------------------------------------------------------------------===//
 // CallInterface implementation: this part is common to both caller and callee.
 //===----------------------------------------------------------------------===//
@@ -589,6 +582,34 @@ void Fortran::lower::CalleeInterface::setFuncAttrs(
 static void addSymbolAttribute(mlir::func::FuncOp func,
                                const Fortran::semantics::Symbol &sym,
                                mlir::MLIRContext &mlirContext) {
+  const Fortran::semantics::Symbol &ultimate = sym.GetUltimate();
+  // The link between an internal procedure and its host procedure is lost
+  // in FIR if the host is BIND(C) since the internal mangling will not
+  // allow retrieving the host bind(C) name, and therefore func.func symbol.
+  // Preserve it as an attribute so that this can be later retrieved.
+  if (Fortran::semantics::ClassifyProcedure(ultimate) ==
+      Fortran::semantics::ProcedureDefinitionClass::Internal) {
+    if (ultimate.owner().kind() ==
+        Fortran::semantics::Scope::Kind::Subprogram) {
+      if (const Fortran::semantics::Symbol *hostProcedure =
+              ultimate.owner().symbol()) {
+        std::string hostName = Fortran::lower::mangle::mangleName(
+            *hostProcedure, /*keepExternalInScope=*/true);
+        func->setAttr(
+            fir::getHostSymbolAttrName(),
+            mlir::SymbolRefAttr::get(
+                &mlirContext, mlir::StringAttr::get(&mlirContext, hostName)));
+      }
+    } else if (ultimate.owner().kind() ==
+               Fortran::semantics::Scope::Kind::MainProgram) {
+      func->setAttr(fir::getHostSymbolAttrName(),
+                    mlir::SymbolRefAttr::get(
+                        &mlirContext,
+                        mlir::StringAttr::get(
+                            &mlirContext, fir::NameUniquer::doProgramEntry())));
+    }
+  }
+
   // Only add this on bind(C) functions for which the symbol is not reflected in
   // the current context.
   if (!Fortran::semantics::IsBindCProcedure(sym))
@@ -686,7 +707,6 @@ void Fortran::lower::CallInterface<T>::declare() {
       for (const auto &placeHolder : llvm::enumerate(inputs))
         if (!placeHolder.value().attributes.empty())
           func.setArgAttrs(placeHolder.index(), placeHolder.value().attributes);
-      side().setFuncAttrs(func);
 
       setCUDAAttributes(func, side().getProcedureSymbol(), characteristic);
     }
@@ -1599,10 +1619,6 @@ class SignatureBuilder
     return proc;
   }
 
-  /// Set internal procedure attribute on MLIR function. Internal procedure
-  /// are defined in the current file and will not go through SignatureBuilder.
-  void setFuncAttrs(mlir::func::FuncOp) const {}
-
   /// This is not the description of an indirect call.
   static constexpr bool isIndirectCall() { return false; }
 
diff --git a/flang/lib/Optimizer/Transforms/ArrayValueCopy.cpp b/flang/lib/Optimizer/Transforms/ArrayValueCopy.cpp
index 675314ed9da03..18ca5711bfea8 100644
--- a/flang/lib/Optimizer/Transforms/ArrayValueCopy.cpp
+++ b/flang/lib/Optimizer/Transforms/ArrayValueCopy.cpp
@@ -728,7 +728,7 @@ conservativeCallConflict(llvm::ArrayRef<mlir::Operation *> reaches) {
       if (auto callee =
               call.getCallableForCallee().dyn_cast<mlir::SymbolRefAttr>()) {
         auto module = op->getParentOfType<mlir::ModuleOp>();
-        return isInternalPorcedure(
+        return isInternalProcedure(
             module.lookupSymbol<mlir::func::FuncOp>(callee));
       }
     return false;
diff --git a/flang/test/Lower/HLFIR/internal-procedures-bindc-host.f90 b/flang/test/Lower/HLFIR/internal-procedures-bindc-host.f90
new file mode 100644
index 0000000000000..07f60b98b0941
--- /dev/null
+++ b/flang/test/Lower/HLFIR/internal-procedures-bindc-host.f90
@@ -0,0 +1,39 @@
+! Test fir.host_sym attribute to retain link between internal
+! and host procedure in FIR even when BIND(C) is involved.
+
+! RUN: bbc -emit-hlfir -o - %s | FileCheck %s
+! RUN: bbc -emit-hlfir -o - %s | fir-opt -external-name-interop -o - |FileCheck %s --check-prefix=AFTER_RENAME_PASS
+
+subroutine foo() bind(c, name="some_c_name")
+  call bar()
+contains
+ subroutine bar()
+ end subroutine
+end subroutine
+! CHECK: func.func @some_c_name()
+! CHECK: func.func private @_QFfooPbar() attributes {fir.host_symbol = @some_c_name, llvm.linkage = #llvm.linkage<internal>}
+! AFTER_RENAME_PASS: func.func @some_c_name()
+! AFTER_RENAME_PASS: func.func private @_QFfooPbar() attributes {fir.host_symbol = @some_c_name, llvm.linkage = #llvm.linkage<internal>}
+
+subroutine notbindc()
+  call bar()
+contains
+ subroutine bar()
+ end subroutine
+end subroutine
+! CHECK: func.func @_QPnotbindc()
+! CHECK: func.func private @_QFnotbindcPbar() attributes {fir.host_symbol = @_QPnotbindc, llvm.linkage = #llvm.linkage<internal>}
+! AFTER_RENAME_PASS: func.func @notbindc_() attributes {fir.internal_name = "_QPnotbindc"}
+! AFTER_RENAME_PASS: func.func private @_QFnotbindcPbar() attributes {fir.host_symbol = @notbindc_, llvm.linkage = #llvm.linkage<internal>}
+
+
+! Main program
+call bar()
+contains
+ subroutine bar()
+ end subroutine
+end
+! CHECK: func.func @_QQmain()
+! CHECK: func.func private @_QFPbar() attributes {fir.host_symbol = @_QQmain, llvm.linkage = #llvm.linkage<internal>}
+! AFTER_RENAME_PASS: func.func @_QQmain()
+! AFTER_RENAME_PASS: func.func private @_QFPbar() attributes {fir.host_symbol = @_QQmain, llvm.linkage = #llvm.linkage<internal>}
diff --git a/flang/test/Lower/HLFIR/internal-procedures.f90 b/flang/test/Lower/HLFIR/internal-procedures.f90
index c898903b6fbe1..fff7125897ddf 100644
--- a/flang/test/Lower/HLFIR/internal-procedures.f90
+++ b/flang/test/Lower/HLFIR/internal-procedures.f90
@@ -10,7 +10,7 @@ subroutine internal
 end subroutine
 end subroutine
 ! CHECK-LABEL: func.func private @_QFtest_explicit_shape_arrayPinternal(
-! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<!fir.box<!fir.array<?xf32>>>>, i32) -> !fir.ref<!fir.box<!fir.array<?xf32>>>
 ! CHECK:  %[[VAL_3:.*]] = fir.load %[[VAL_2]] : !fir.ref<!fir.box<!fir.array<?xf32>>>
@@ -28,7 +28,7 @@ subroutine internal
 end subroutine
 end subroutine
 ! CHECK-LABEL: func.func private @_QFtest_assumed_shapePinternal(
-! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<!fir.box<!fir.array<?xf32>>>>, i32) -> !fir.ref<!fir.box<!fir.array<?xf32>>>
 ! CHECK:  %[[VAL_3:.*]] = fir.load %[[VAL_2]] : !fir.ref<!fir.box<!fir.array<?xf32>>>
@@ -45,7 +45,7 @@ subroutine internal()
 end subroutine
 end subroutine
 ! CHECK-LABEL:   func.func private @_QFtest_scalar_charPinternal(
-! CHECK-SAME:                               %[[VAL_0:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:                               %[[VAL_0:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<!fir.boxchar<1>>>, i32) -> !fir.ref<!fir.boxchar<1>>
 ! CHECK:  %[[VAL_3:.*]] = fir.load %[[VAL_2]] : !fir.ref<!fir.boxchar<1>>
diff --git a/flang/test/Lower/OpenACC/acc-routine04.f90 b/flang/test/Lower/OpenACC/acc-routine04.f90
index 2339c23eaaf85..f603376163901 100644
--- a/flang/test/Lower/OpenACC/acc-routine04.f90
+++ b/flang/test/Lower/OpenACC/acc-routine04.f90
@@ -31,4 +31,4 @@ subroutine sub2()
 ! CHECK: acc.routine @acc_routine_0 func(@_QMdummy_modPsub1) seq
 ! CHECK: func.func @_QMdummy_modPsub1(%arg0: !fir.ref<i32> {fir.bindc_name = "i"}) attributes {acc.routine_info = #acc.routine_info<[@acc_routine_0]>}
 ! CHECK: func.func @_QQmain() attributes {fir.bindc_name = "test_acc_routine"}
-! CHECK: func.func private @_QFPsub2() attributes {acc.routine_info = #acc.routine_info<[@acc_routine_1]>, llvm.linkage = #llvm.linkage<internal>}
+! CHECK: func.func private @_QFPsub2() attributes {acc.routine_info = #acc.routine_info<[@acc_routine_1]>, fir.host_symbol = @_QQmain, llvm.linkage = #llvm.linkage<internal>}
diff --git a/flang/test/Lower/OpenMP/threadprivate-host-association-2.f90 b/flang/test/Lower/OpenMP/threadprivate-host-association-2.f90
index b47bff5bebb0b..a8d29baf74f22 100644
--- a/flang/test/Lower/OpenMP/threadprivate-host-association-2.f90
+++ b/flang/test/Lower/OpenMP/threadprivate-host-association-2.f90
@@ -12,7 +12,7 @@
 !CHECK:   fir.call @_QFPsub() fastmath<contract> : () -> ()
 !CHECK:   return
 !CHECK: }
-!CHECK: func.func private @_QFPsub() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+!CHECK: func.func private @_QFPsub() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 !CHECK:   %[[A:.*]] = fir.alloca i32 {bindc_name = "a", uniq_name = "_QFEa"}
 !CHECK:   %[[A_DECL:.*]]:2 = hlfir.declare %[[A]] {uniq_name = "_QFEa"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK:   %[[A_ADDR:.*]] = fir.address_of(@_QFEa) : !fir.ref<i32>
diff --git a/flang/test/Lower/OpenMP/threadprivate-host-association.f90 b/flang/test/Lower/OpenMP/threadprivate-host-association.f90
index 98f7b51bb9711..096e510c19c69 100644
--- a/flang/test/Lower/OpenMP/threadprivate-host-association.f90
+++ b/flang/test/Lower/OpenMP/threadprivate-host-association.f90
@@ -11,7 +11,7 @@
 !CHECK:   fir.call @_QFPsub() fastmath<contract> : () -> ()
 !CHECK:   return
 !CHECK: }
-!CHECK: func.func private @_QFPsub() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+!CHECK: func.func private @_QFPsub() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 !CHECK:   %[[A:.*]] = fir.address_of(@_QFEa) : !fir.ref<i32>
 !CHECK:   %[[A_DECL:.*]]:2 = hlfir.declare %[[A]] {uniq_name = "_QFEa"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK:   %[[TP_A:.*]] = omp.threadprivate %[[A_DECL]]#1 : !fir.ref<i32> -> !fir.ref<i32>
diff --git a/flang/test/Lower/character-elemental.f90 b/flang/test/Lower/character-elemental.f90
index 6c46454176f53..9a9cf8bf2d9c6 100644
--- a/flang/test/Lower/character-elemental.f90
+++ b/flang/test/Lower/character-elemental.f90
@@ -5,6 +5,12 @@ subroutine substring_main
   character*7 :: string(2) = ['12     ', '12     ']
   integer :: result(2)
   integer :: ival
+interface
+  elemental function inner(arg)
+    character(len=*), intent(in) :: arg
+    integer :: inner
+  end function inner
+end interface
 
   ival = 1
   ! CHECK: %[[a0:.*]] = fir.alloca i32 {bindc_name = "ival", uniq_name = "_QFsubstring_mainEival"}
@@ -26,14 +32,7 @@ subroutine substring_main
   ! CHECK: %[[a14:.*]] = fir.coordinate_of %[[a13]], %[[a12]] : (!fir.ref<!fir.array<7x!fir.char<1>>>, index) -> !fir.ref<!fir.char<1>>
   ! CHECK: %[[a15:.*]] = fir.convert %[[a14]] : (!fir.ref<!fir.char<1>>) -> !fir.ref<!fir.char<1,?>>
   ! CHECK: %[[a16:.*]] = fir.emboxchar %[[a15]], {{.*}} : (!fir.ref<!fir.char<1,?>>, index) -> !fir.boxchar<1>
-  ! CHECK: %[[a17:.*]] = fir.call @_QFsubstring_mainPinner(%[[a16]]) {{.*}}: (!fir.boxchar<1>) -> i32
+  ! CHECK: %[[a17:.*]] = fir.call @_QPinner(%[[a16]]) {{.*}}: (!fir.boxchar<1>) -> i32
   result = inner(string(1:2)(ival:ival))
   print *, result
-contains
-  elemental function inner(arg)
-    character(len=*), intent(in) :: arg
-    integer :: inner
-
-    inner = len(arg)
-  end function inner
 end subroutine substring_main
diff --git a/flang/test/Lower/equivalence-with-host-assoc.f90 b/flang/test/Lower/equivalence-with-host-assoc.f90
index 0ffb1bc5bf9ee..b8ce72f3787c0 100644
--- a/flang/test/Lower/equivalence-with-host-assoc.f90
+++ b/flang/test/Lower/equivalence-with-host-assoc.f90
@@ -10,7 +10,7 @@ subroutine inner
     i1 = j1
   end subroutine inner
 end subroutine test1
-! FIR-LABEL:   func.func private @_QFtest1Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! FIR-LABEL:   func.func private @_QFtest1Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! FIR:           %[[VAL_0:.*]] = fir.address_of(@_QFtest1Ei1) : !fir.ref<!fir.array<1xi32>>
 ! FIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<!fir.array<1xi32>>) -> !fir.ref<!fir.array<4xi8>>
 ! FIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -24,7 +24,7 @@ end subroutine test1
 ! FIR:           return
 ! FIR:         }
 
-! HLFIR-LABEL:   func.func private @_QFtest1Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! HLFIR-LABEL:   func.func private @_QFtest1Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! HLFIR:           %[[VAL_0:.*]] = fir.address_of(@_QFtest1Ei1) : !fir.ref<!fir.array<1xi32>>
 ! HLFIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<!fir.array<1xi32>>) -> !fir.ref<!fir.array<4xi8>>
 ! HLFIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -54,7 +54,7 @@ subroutine inner
     end subroutine inner
   end subroutine host
 end module test2
-! FIR-LABEL:   func.func private @_QMtest2FhostPinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! FIR-LABEL:   func.func private @_QMtest2FhostPinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! FIR:           %[[VAL_0:.*]] = fir.address_of(@_QMtest2FhostEf1) : !fir.ref<!fir.array<1xi32>>
 ! FIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<!fir.array<1xi32>>) -> !fir.ref<!fir.array<4xi8>>
 ! FIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -68,7 +68,7 @@ end module test2
 ! FIR:           return
 ! FIR:         }
 
-! HLFIR-LABEL:   func.func private @_QMtest2FhostPinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! HLFIR-LABEL:   func.func private @_QMtest2FhostPinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! HLFIR:           %[[VAL_0:.*]] = fir.address_of(@_QMtest2FhostEf1) : !fir.ref<!fir.array<1xi32>>
 ! HLFIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<!fir.array<1xi32>>) -> !fir.ref<!fir.array<4xi8>>
 ! HLFIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -94,7 +94,7 @@ subroutine inner
     i1 = j1 + k1
   end subroutine inner
 end subroutine test3
-! FIR-LABEL:   func.func private @_QFtest3Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! FIR-LABEL:   func.func private @_QFtest3Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! FIR:           %[[VAL_0:.*]] = fir.address_of(@blk_) : !fir.ref<tuple<i32>>
 ! FIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<tuple<i32>>) -> !fir.ref<!fir.array<?xi8>>
 ! FIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -115,7 +115,7 @@ end subroutine test3
 ! FIR:           return
 ! FIR:         }
 
-! HLFIR-LABEL:   func.func private @_QFtest3Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! HLFIR-LABEL:   func.func private @_QFtest3Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! HLFIR:           %[[VAL_0:.*]] = fir.address_of(@blk_) : !fir.ref<tuple<i32>>
 ! HLFIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<tuple<i32>>) -> !fir.ref<!fir.array<?xi8>>
 ! HLFIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -149,7 +149,7 @@ subroutine inner
     i1 = j1 + k1
   end subroutine inner
 end subroutine test4
-! FIR-LABEL:   func.func private @_QFtest4Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! FIR-LABEL:   func.func private @_QFtest4Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! FIR:           %[[VAL_0:.*]] = fir.address_of(@blk_) : !fir.ref<tuple<i32>>
 ! FIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<tuple<i32>>) -> !fir.ref<!fir.array<?xi8>>
 ! FIR:           %[[VAL_2:.*]] = arith.constant 0 : index
@@ -170,7 +170,7 @@ end subroutine test4
 ! FIR:           return
 ! FIR:         }
 
-! HLFIR-LABEL:   func.func private @_QFtest4Pinner() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! HLFIR-LABEL:   func.func private @_QFtest4Pinner() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! HLFIR:           %[[VAL_0:.*]] = fir.address_of(@blk_) : !fir.ref<tuple<i32>>
 ! HLFIR:           %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<tuple<i32>>) -> !fir.ref<!fir.array<?xi8>>
 ! HLFIR:           %[[VAL_2:.*]] = arith.constant 0 : index
diff --git a/flang/test/Lower/explicit-interface-results-2.f90 b/flang/test/Lower/explicit-interface-results-2.f90
index 86aae720e7fcf..a63ee5fc91794 100644
--- a/flang/test/Lower/explicit-interface-results-2.f90
+++ b/flang/test/Lower/explicit-interface-results-2.f90
@@ -70,7 +70,7 @@ subroutine host4()
   call internal_proc_a()
 contains
 ! CHECK-LABEL: func private @_QFhost4Pinternal_proc_a
-! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine internal_proc_a()
     call takes_array(return_array())
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
@@ -94,7 +94,7 @@ subroutine host5()
   implicit none
   call internal_proc_a()
 contains
-! CHECK-LABEL: func private @_QFhost5Pinternal_proc_a() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-LABEL: func private @_QFhost5Pinternal_proc_a() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine internal_proc_a()
     call takes_array(return_array())
 ! CHECK:  %[[VAL_0:.*]] = fir.address_of(@_QMsome_moduleEn_module) : !fir.ref<i32>
diff --git a/flang/test/Lower/host-associated-functions.f90 b/flang/test/Lower/host-associated-functions.f90
index 78d081748c2f4..d67a74fa39980 100644
--- a/flang/test/Lower/host-associated-functions.f90
+++ b/flang/test/Lower/host-associated-functions.f90
@@ -20,7 +20,7 @@ subroutine capture_char_func_dummy(char_func_dummy, n)
   call internal()
 contains
   ! CHECK-LABEL: func private @_QFcapture_char_func_dummyPinternal(
-  ! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>, !fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>, !fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine internal()
   ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
   ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>, !fir.ref<i32>>>, i32) -> !fir.ref<tuple<!fir.boxproc<() -> ()>, i64>>
@@ -56,7 +56,7 @@ subroutine capture_char_func_assumed_dummy(char_func_dummy)
   call internal()
 contains
 ! CHECK-LABEL: func private @_QFcapture_char_func_assumed_dummyPinternal(
-! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine internal()
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<tuple<!fir.boxproc<() -> ()>, i64>>>, i32) -> !fir.ref<tuple<!fir.boxproc<() -> ()>, i64>>
@@ -110,7 +110,7 @@ function array_func()
 contains
   subroutine internal()
 ! CHECK-LABEL: func private @_QFcapture_array_funcPinternal(
-! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:  %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 0 : i32
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<tuple<!fir.ref<i32>>>, i32) -> !fir.llvm_ptr<!fir.ref<i32>>
 ! CHECK:  %[[VAL_3:.*]] = fir.load %[[VAL_2]] : !fir.llvm_ptr<!fir.ref<i32>>
diff --git a/flang/test/Lower/host-associated-globals.f90 b/flang/test/Lower/host-associated-globals.f90
index fe612e777aeaa..c91a5a46af0d5 100644
--- a/flang/test/Lower/host-associated-globals.f90
+++ b/flang/test/Lower/host-associated-globals.f90
@@ -37,7 +37,7 @@ subroutine bar()
     print *, j_in_equiv, not_in_equiv
  end subroutine
 end subroutine
-! CHECK-LABEL: func.func private @_QFtest_commonPbar() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-LABEL: func.func private @_QFtest_commonPbar() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_0:.*]] = fir.address_of(@x_) : !fir.ref<!fir.array<12xi8>>
 ! CHECK:  %[[VAL_1:.*]] = fir.convert %[[VAL_0]] : (!fir.ref<!fir.array<12xi8>>) -> !fir.ref<!fir.array<?xi8>>
 ! CHECK:  %[[VAL_2:.*]] = arith.constant 4 : index
@@ -59,7 +59,7 @@ subroutine bar()
     print *, j_in_equiv, not_in_equiv
  end subroutine
 end subroutine
-! CHECK-LABEL: func.func private @_QFsaved_equivPbar() attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-LABEL: func.func private @_QFsaved_equivPbar() attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_0:.*]] = fir.address_of(@_QFsaved_equivEi) : !fir.ref<!fir.array<8xi8>>
 ! CHECK:  %[[VAL_1:.*]] = arith.constant 4 : index
 ! CHECK:  %[[VAL_2:.*]] = fir.coordinate_of %[[VAL_0]], %[[VAL_1]] : (!fir.ref<!fir.array<8xi8>>, index) -> !fir.ref<i8>
@@ -80,7 +80,7 @@ subroutine bar()
  end subroutine
 end subroutine
 ! CHECK-LABEL: func.func private @_QFmixed_capturePbar(
-! CHECK-SAME:    %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:    %[[VAL_0:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:  %[[VAL_1:.*]] = fir.address_of(@_QFmixed_captureEsaved_i) : !fir.ref<!fir.array<4xi8>>
 ! CHECK:  %[[VAL_2:.*]] = arith.constant 0 : index
 ! CHECK:  %[[VAL_3:.*]] = fir.coordinate_of %[[VAL_1]], %[[VAL_2]] : (!fir.ref<!fir.array<4xi8>>, index) -> !fir.ref<i8>
diff --git a/flang/test/Lower/host-associated.f90 b/flang/test/Lower/host-associated.f90
index f88903c8af80f..cdc7e6a05288a 100644
--- a/flang/test/Lower/host-associated.f90
+++ b/flang/test/Lower/host-associated.f90
@@ -20,7 +20,7 @@ subroutine test1
   print *, i
 contains
   ! CHECK-LABEL: func private @_QFtest1Ptest1_internal(
-  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   ! CHECK: %[[iaddr:.*]] = fir.coordinate_of %[[arg]], %c0
   ! CHECK: %[[i:.*]] = fir.load %[[iaddr]] : !fir.llvm_ptr<!fir.ref<i32>>
   ! CHECK: %[[val:.*]] = fir.call @_QPifoo() {{.*}}: () -> i32
@@ -47,7 +47,7 @@ subroutine test2
   print *, a, b
 contains
   ! CHECK-LABEL: func private @_QFtest2Ptest2_internal(
-  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<f32>, !fir.ref<f32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<f32>, !fir.ref<f32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test2_internal
     ! CHECK: %[[a:.*]] = fir.coordinate_of %[[arg]], %c0
     ! CHECK: %[[aa:.*]] = fir.load %[[a]] : !fir.llvm_ptr<!fir.ref<f32>>
@@ -62,7 +62,7 @@ subroutine test2_internal
   end subroutine test2_internal
 
   ! CHECK-LABEL: func private @_QFtest2Ptest2_inner(
-  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<f32>, !fir.ref<f32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[arg:[^:]*]]: !fir.ref<tuple<!fir.ref<f32>, !fir.ref<f32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test2_inner
     ! CHECK: %[[a:.*]] = fir.coordinate_of %[[arg]], %c0
     ! CHECK: %[[aa:.*]] = fir.load %[[a]] : !fir.llvm_ptr<!fir.ref<f32>>
@@ -96,7 +96,7 @@ subroutine test6(c)
 
 contains
   ! CHECK-LABEL: func private @_QFtest6Ptest6_inner(
-  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test6_inner
     ! CHECK: %[[coor:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.boxchar<1>>>, i32) -> !fir.ref<!fir.boxchar<1>>
     ! CHECK: %[[load:.*]] = fir.load %[[coor]] : !fir.ref<!fir.boxchar<1>>
@@ -138,7 +138,7 @@ subroutine test3(p,q,i)
   
 contains
   ! CHECK-LABEL: func private @_QFtest3Ptest3_inner(
-  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test3_inner
     ! CHECK: %[[pcoor:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>>>, i32) -> !fir.ref<!fir.box<!fir.array<?xf32>>>
     ! CHECK: %[[p:.*]] = fir.load %[[pcoor]] : !fir.ref<!fir.box<!fir.array<?xf32>>>
@@ -185,7 +185,7 @@ subroutine test3a(p)
   
 contains
   ! CHECK: func private @_QFtest3aPtest3a_inner(
-  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.box<!fir.array<10xf32>>, !fir.box<!fir.array<10xf32>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.box<!fir.array<10xf32>>, !fir.box<!fir.array<10xf32>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test3a_inner
     ! CHECK: %[[pcoor:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.box<!fir.array<10xf32>>, !fir.box<!fir.array<10xf32>>>>, i32) -> !fir.ref<!fir.box<!fir.array<10xf32>>>
     ! CHECK: %[[p:.*]] = fir.load %[[pcoor]] : !fir.ref<!fir.box<!fir.array<10xf32>>>
@@ -229,7 +229,7 @@ subroutine test4
   
 contains
   ! CHECK-LABEL: func private @_QFtest4Ptest4_inner(
-  ! CHECK-SAME:%[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME:%[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test4_inner
     ! CHECK: %[[ptup:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>>>, i32) -> !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<f32>>>>
     ! CHECK: %[[p:.*]] = fir.load %[[ptup]] : !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<f32>>>>
@@ -271,7 +271,7 @@ subroutine test5
   
 contains
   ! CHECK-LABEL: func private @_QFtest5Ptest5_inner(
-  ! CHECK-SAME:%[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+  ! CHECK-SAME:%[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
   subroutine test5_inner
     ! CHECK: %[[ptup:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>>>, i32) -> !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>>
     ! CHECK: %[[p:.*]] = fir.load %[[ptup]] : !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>>
@@ -309,7 +309,7 @@ subroutine test7(j, k)
 contains
 
 ! CHECK-LABEL: func private @_QFtest7Ptest7_inner(
-! CHECK-SAME: %[[i:.*]]: !fir.ref<i32>{{.*}}, %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) -> i32 attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[i:.*]]: !fir.ref<i32>{{.*}}, %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) -> i32 attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 elemental integer function test7_inner(i)
   implicit none
   integer, intent(in) :: i
@@ -330,7 +330,7 @@ subroutine issue990()
   call bar()
 contains
 ! CHECK-LABEL: func private @_QFissue990Pbar(
-! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 subroutine bar()
   integer :: stmt_func, i
   stmt_func(i) = i + captured
@@ -352,7 +352,7 @@ subroutine issue990b()
   call bar()
 contains
 ! CHECK-LABEL: func private @_QFissue990bPbar(
-! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 subroutine bar()
   ! CHECK: %[[tupAddr:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.ref<i32>>>, i32) -> !fir.llvm_ptr<!fir.ref<i32>>
   ! CHECK: %[[addr:.*]] = fir.load %[[tupAddr]] : !fir.llvm_ptr<!fir.ref<i32>>
@@ -373,7 +373,7 @@ real function dummy_proc(x)
  call bar()
 contains
 ! CHECK-LABEL: func private @_QFtest8Pbar(
-! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxproc<() -> ()>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxproc<() -> ()>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 subroutine bar()
   ! CHECK: %[[tupAddr:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.boxproc<() -> ()>>>, i32) -> !fir.ref<!fir.boxproc<() -> ()>>
   ! CHECK: %[[dummyProc:.*]] = fir.load %[[tupAddr]] : !fir.ref<!fir.boxproc<() -> ()>>
@@ -393,7 +393,7 @@ subroutine dummy_proc()
  call bar()
 contains
 ! CHECK-LABEL: func private @_QFtest9Pbar(
-! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxproc<() -> ()>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.boxproc<() -> ()>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 subroutine bar()
   ! CHECK: %[[tupAddr:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.boxproc<() -> ()>>>, i32) -> !fir.ref<!fir.boxproc<() -> ()>>
   ! CHECK: %[[dummyProc:.*]] = fir.load %[[tupAddr]] : !fir.ref<!fir.boxproc<() -> ()>>
@@ -416,7 +416,7 @@ subroutine test10(i)
  call bar()
 contains
 ! CHECK-LABEL: func private @_QFtest10Pbar(
-! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[tup:.*]]: !fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 subroutine bar()
   ! CHECK: %[[tupAddr:.*]] = fir.coordinate_of %[[tup]], %c0{{.*}} : (!fir.ref<tuple<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>>>, i32) -> !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>>
   ! CHECK: fir.load %[[tupAddr]] : !fir.llvm_ptr<!fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>>
@@ -435,7 +435,7 @@ subroutine bar()
 
 ! CHECK-LABEL: func private @_QFtest_proc_dummyPtest_proc_dummy_a(
 ! CHECK-SAME:          %[[VAL_0:.*]]: !fir.ref<i32> {fir.bindc_name = "j"},
-! CHECK-SAME:          %[[VAL_1:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:          %[[VAL_1:.*]]: !fir.ref<tuple<!fir.ref<i32>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK:         %[[VAL_2:.*]] = arith.constant 0 : i32
 ! CHECK:         %[[VAL_3:.*]] = fir.coordinate_of %[[VAL_1]], %[[VAL_2]] : (!fir.ref<tuple<!fir.ref<i32>>>, i32) -> !fir.llvm_ptr<!fir.ref<i32>>
 ! CHECK:         %[[VAL_4:.*]] = fir.load %[[VAL_3]] : !fir.llvm_ptr<!fir.ref<i32>>
@@ -528,7 +528,7 @@ end subroutine test_proc_dummy_other
 ! CHECK-LABEL: func private @_QFtest_proc_dummy_charPgen_message(
 ! CHECK-SAME:                                            %[[VAL_0:.*]]: !fir.ref<!fir.char<1,10>>,
 ! CHECK-SAME:                                            %[[VAL_1:.*]]: index,
-! CHECK-SAME:                                            %[[VAL_2:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) -> !fir.boxchar<1> attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME:                                            %[[VAL_2:.*]]: !fir.ref<tuple<!fir.boxchar<1>>> {fir.host_assoc}) -> !fir.boxchar<1> attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK-DAG:         %[[VAL_3:.*]] = arith.constant 0 : i32
 ! CHECK-DAG:         %[[VAL_4:.*]] = arith.constant 10 : index
 ! CHECK-DAG:         %[[VAL_5:.*]] = arith.constant false
diff --git a/flang/test/Lower/polymorphic.f90 b/flang/test/Lower/polymorphic.f90
index e031b4805dc5b..70c1f768e389a 100644
--- a/flang/test/Lower/polymorphic.f90
+++ b/flang/test/Lower/polymorphic.f90
@@ -520,7 +520,7 @@ subroutine internal
   end subroutine
 
 ! CHECK-LABEL: func.func private @_QMpolymorphic_testFhost_assocPinternal(
-! CHECK-SAME: %[[TUPLE:.*]]: !fir.ref<tuple<!fir.class<!fir.type<_QMpolymorphic_testTp1{a:i32,b:i32}>>>> {fir.host_assoc}) attributes {fir.internal_proc, llvm.linkage = #llvm.linkage<internal>} {
+! CHECK-SAME: %[[TUPLE:.*]]: !fir.ref<tuple<!fir.class<!fir.type<_QMpolymorphic_testTp1{a:i32,b:i32}>>>> {fir.host_assoc}) attributes {fir.host_symbol = {{.*}}, llvm.linkage = #llvm.linkage<internal>} {
 ! CHECK: %[[POS_IN_TUPLE:.*]] = arith.constant 0 : i32
 ! CHECK: %[[COORD_OF_CLASS:.*]] = fir.coordinate_of %[[TUPLE]], %[[POS_IN_TUPLE]] : (!fir.ref<tuple<!fir.class<!fir.type<_QMpolymorphic_testTp1{a:i32,b:i32}>>>>, i32) -> !fir.ref<!fir.class<!fir.type<_QMpolymorphic_testTp1{a:i32,b:i32}>>>
 ! CHECK: %[[CLASS:.*]] = fir.load %[[COORD_OF_CLASS]] : !fir.ref<!fir.class<!fir.type<_QMpolymorphic_testTp1{a:i32,b:i32}>>>

From b512df660ef136f8bbd0895bf862a827923a6714 Mon Sep 17 00:00:00 2001
From: Vyacheslav Levytskyy <vyacheslav.levytskyy@intel.com>
Date: Wed, 17 Apr 2024 11:50:37 +0200
Subject: [PATCH 715/886] [SPIR-V] Improve Tablegen instruction selection and
 account for a pointer size of the target (#88725)

This PR resolves the issue that SPIR-V Backend uses the notion of a
pointer size of the target, most notably, in legalizer code, but
Tablegen instruction selection in SPIR-V Backend doesn't account for a
pointer size of the target. See
https://github.com/llvm/llvm-project/issues/88723 for a detailed
description. There are 3 test cases attached to the PR that reproduced
the issue, when dealing with spirv32-spirv64 differences, and are
working correctly now with this PR.
---
 llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp |  3 +-
 llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp      |  6 +-
 llvm/lib/Target/SPIRV/SPIRVInstrInfo.td       | 18 ++++--
 .../Target/SPIRV/SPIRVInstructionSelector.cpp |  8 ++-
 llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp  |  5 +-
 llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp   | 62 ++++++++++++++-----
 .../Target/SPIRV/SPIRVRegisterBankInfo.cpp    | 20 +-----
 llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td   |  7 +--
 llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td    | 49 ++++++++-------
 .../CodeGen/SPIRV/instructions/select-phi.ll  |  4 ++
 .../SPIRV/instructions/select-ptr-load.ll     | 25 ++++++++
 .../test/CodeGen/SPIRV/instructions/select.ll |  3 +
 .../SPIRV/{select.ll => select-builtin.ll}    |  2 +
 13 files changed, 138 insertions(+), 74 deletions(-)
 create mode 100644 llvm/test/CodeGen/SPIRV/instructions/select-ptr-load.ll
 rename llvm/test/CodeGen/SPIRV/{select.ll => select-builtin.ll} (67%)

diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
index 05e41e06248e3..cebe230d3e8ce 100644
--- a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
@@ -653,7 +653,8 @@ Register SPIRVGlobalRegistry::buildGlobalVariable(
   auto MRI = MIRBuilder.getMRI();
   assert(MRI->getType(ResVReg).isPointer() && "Pointer type is expected");
   if (Reg != ResVReg) {
-    LLT RegLLTy = LLT::pointer(MRI->getType(ResVReg).getAddressSpace(), 32);
+    LLT RegLLTy =
+        LLT::pointer(MRI->getType(ResVReg).getAddressSpace(), getPointerSize());
     MRI->setType(Reg, RegLLTy);
     assignSPIRVTypeToVReg(BaseType, Reg, MIRBuilder.getMF());
   } else {
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
index aacfecc1e313f..af98f2f880459 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.cpp
@@ -247,8 +247,10 @@ void SPIRVInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
 bool SPIRVInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
   if (MI.getOpcode() == SPIRV::GET_ID || MI.getOpcode() == SPIRV::GET_fID ||
-      MI.getOpcode() == SPIRV::GET_pID || MI.getOpcode() == SPIRV::GET_vfID ||
-      MI.getOpcode() == SPIRV::GET_vID || MI.getOpcode() == SPIRV::GET_vpID) {
+      MI.getOpcode() == SPIRV::GET_pID32 ||
+      MI.getOpcode() == SPIRV::GET_pID64 || MI.getOpcode() == SPIRV::GET_vfID ||
+      MI.getOpcode() == SPIRV::GET_vID || MI.getOpcode() == SPIRV::GET_vpID32 ||
+      MI.getOpcode() == SPIRV::GET_vpID64) {
     auto &MRI = MI.getMF()->getRegInfo();
     MRI.replaceRegWith(MI.getOperand(0).getReg(), MI.getOperand(1).getReg());
     MI.eraseFromParent();
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
index a3f981457c8da..151d0ec1fe569 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
+++ b/llvm/lib/Target/SPIRV/SPIRVInstrInfo.td
@@ -19,10 +19,12 @@ let isCodeGenOnly=1 in {
   def DECL_TYPE: Pseudo<(outs ANYID:$dst_id), (ins ANYID:$src_id, TYPE:$src_ty)>;
   def GET_ID: Pseudo<(outs ID:$dst_id), (ins ANYID:$src)>;
   def GET_fID: Pseudo<(outs fID:$dst_id), (ins ANYID:$src)>;
-  def GET_pID: Pseudo<(outs pID:$dst_id), (ins ANYID:$src)>;
+  def GET_pID32: Pseudo<(outs pID32:$dst_id), (ins ANYID:$src)>;
+  def GET_pID64: Pseudo<(outs pID64:$dst_id), (ins ANYID:$src)>;
   def GET_vID: Pseudo<(outs vID:$dst_id), (ins ANYID:$src)>;
   def GET_vfID: Pseudo<(outs vfID:$dst_id), (ins ANYID:$src)>;
-  def GET_vpID: Pseudo<(outs vpID:$dst_id), (ins ANYID:$src)>;
+  def GET_vpID32: Pseudo<(outs vpID32:$dst_id), (ins ANYID:$src)>;
+  def GET_vpID64: Pseudo<(outs vpID64:$dst_id), (ins ANYID:$src)>;
 }
 
 def SPVTypeBin : SDTypeProfile<1, 2, []>;
@@ -66,8 +68,10 @@ multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genP =
     def SIVCond: TernOpTyped<name, opCode, vID, ID, node>;
   }
   if genP then {
-    def SPSCond: TernOpTyped<name, opCode, ID, pID, node>;
-    def SPVCond: TernOpTyped<name, opCode, vID, pID, node>;
+    def SPSCond32: TernOpTyped<name, opCode, ID, pID32, node>;
+    def SPVCond32: TernOpTyped<name, opCode, vID, pID32, node>;
+    def SPSCond64: TernOpTyped<name, opCode, ID, pID64, node>;
+    def SPVCond64: TernOpTyped<name, opCode, vID, pID64, node>;
   }
   if genV then {
     if genF then {
@@ -79,8 +83,10 @@ multiclass TernOpTypedGen<string name, bits<16> opCode, SDNode node, bit genP =
       def VIVCond: TernOpTyped<name, opCode, vID, vID, node>;
     }
     if genP then {
-      def VPSCond: TernOpTyped<name, opCode, ID, vpID, node>;
-      def VPVCond: TernOpTyped<name, opCode, vID, vpID, node>;
+      def VPSCond32: TernOpTyped<name, opCode, ID, vpID32, node>;
+      def VPVCond32: TernOpTyped<name, opCode, vID, vpID32, node>;
+      def VPSCond64: TernOpTyped<name, opCode, ID, vpID64, node>;
+      def VPVCond64: TernOpTyped<name, opCode, vID, vpID64, node>;
     }
   }
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index 200fe38298c02..72e5a7bcac983 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -290,14 +290,18 @@ bool SPIRVInstructionSelector::select(MachineInstr &I) {
   // If it's not a GMIR instruction, we've selected it already.
   if (!isPreISelGenericOpcode(Opcode)) {
     if (Opcode == SPIRV::ASSIGN_TYPE) { // These pseudos aren't needed any more.
-      auto *Def = MRI->getVRegDef(I.getOperand(1).getReg());
+      Register DstReg = I.getOperand(0).getReg();
+      Register SrcReg = I.getOperand(1).getReg();
+      auto *Def = MRI->getVRegDef(SrcReg);
       if (isTypeFoldingSupported(Def->getOpcode())) {
+        if (MRI->getType(DstReg).isPointer())
+          MRI->setType(DstReg, LLT::scalar(32));
         bool Res = selectImpl(I, *CoverageInfo);
         assert(Res || Def->getOpcode() == TargetOpcode::G_CONSTANT);
         if (Res)
           return Res;
       }
-      MRI->replaceRegWith(I.getOperand(1).getReg(), I.getOperand(0).getReg());
+      MRI->replaceRegWith(SrcReg, DstReg);
       I.removeFromParent();
       return true;
     } else if (I.getNumDefs() == 1) {
diff --git a/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
index f069a92ac6868..d652b5de60808 100644
--- a/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVPostLegalizer.cpp
@@ -55,8 +55,9 @@ extern void processInstr(MachineInstr &MI, MachineIRBuilder &MIB,
 
 static bool isMetaInstrGET(unsigned Opcode) {
   return Opcode == SPIRV::GET_ID || Opcode == SPIRV::GET_fID ||
-         Opcode == SPIRV::GET_pID || Opcode == SPIRV::GET_vID ||
-         Opcode == SPIRV::GET_vfID || Opcode == SPIRV::GET_vpID;
+         Opcode == SPIRV::GET_pID32 || Opcode == SPIRV::GET_pID64 ||
+         Opcode == SPIRV::GET_vID || Opcode == SPIRV::GET_vfID ||
+         Opcode == SPIRV::GET_vpID32 || Opcode == SPIRV::GET_vpID64;
 }
 
 static bool mayBeInserted(unsigned Opcode) {
diff --git a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
index 2c964595fc39e..ed03154d4c8dd 100644
--- a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
@@ -223,11 +223,12 @@ static SPIRVType *propagateSPIRVType(MachineInstr *MI, SPIRVGlobalRegistry *GR,
 }
 
 static std::pair<Register, unsigned>
-createNewIdReg(Register ValReg, unsigned Opcode, MachineRegisterInfo &MRI,
+createNewIdReg(SPIRVType *SpvType, Register SrcReg, MachineRegisterInfo &MRI,
                const SPIRVGlobalRegistry &GR) {
-  LLT NewT = LLT::scalar(32);
-  SPIRVType *SpvType = GR.getSPIRVTypeForVReg(ValReg);
+  if (!SpvType)
+    SpvType = GR.getSPIRVTypeForVReg(SrcReg);
   assert(SpvType && "VReg is expected to have SPIRV type");
+  LLT NewT = LLT::scalar(32);
   bool IsFloat = SpvType->getOpcode() == SPIRV::OpTypeFloat;
   bool IsVectorFloat =
       SpvType->getOpcode() == SPIRV::OpTypeVector &&
@@ -236,14 +237,38 @@ createNewIdReg(Register ValReg, unsigned Opcode, MachineRegisterInfo &MRI,
   IsFloat |= IsVectorFloat;
   auto GetIdOp = IsFloat ? SPIRV::GET_fID : SPIRV::GET_ID;
   auto DstClass = IsFloat ? &SPIRV::fIDRegClass : &SPIRV::IDRegClass;
-  if (MRI.getType(ValReg).isPointer()) {
-    NewT = LLT::pointer(0, 32);
-    GetIdOp = SPIRV::GET_pID;
-    DstClass = &SPIRV::pIDRegClass;
-  } else if (MRI.getType(ValReg).isVector()) {
+  if (MRI.getType(SrcReg).isPointer()) {
+    unsigned PtrSz = GR.getPointerSize();
+    NewT = LLT::pointer(0, PtrSz);
+    bool IsVec = MRI.getType(SrcReg).isVector();
+    if (IsVec)
+      NewT = LLT::fixed_vector(2, NewT);
+    if (PtrSz == 64) {
+      if (IsVec) {
+        GetIdOp = SPIRV::GET_vpID64;
+        DstClass = &SPIRV::vpID64RegClass;
+      } else {
+        GetIdOp = SPIRV::GET_pID64;
+        DstClass = &SPIRV::pID64RegClass;
+      }
+    } else {
+      if (IsVec) {
+        GetIdOp = SPIRV::GET_vpID32;
+        DstClass = &SPIRV::vpID32RegClass;
+      } else {
+        GetIdOp = SPIRV::GET_pID32;
+        DstClass = &SPIRV::pID32RegClass;
+      }
+    }
+  } else if (MRI.getType(SrcReg).isVector()) {
     NewT = LLT::fixed_vector(2, NewT);
-    GetIdOp = IsFloat ? SPIRV::GET_vfID : SPIRV::GET_vID;
-    DstClass = IsFloat ? &SPIRV::vfIDRegClass : &SPIRV::vIDRegClass;
+    if (IsFloat) {
+      GetIdOp = SPIRV::GET_vfID;
+      DstClass = &SPIRV::vfIDRegClass;
+    } else {
+      GetIdOp = SPIRV::GET_vID;
+      DstClass = &SPIRV::vIDRegClass;
+    }
   }
   Register IdReg = MRI.createGenericVirtualRegister(NewT);
   MRI.setRegClass(IdReg, DstClass);
@@ -264,6 +289,7 @@ Register insertAssignInstr(Register Reg, Type *Ty, SPIRVType *SpirvTy,
   MIB.setInsertPt(*Def->getParent(),
                   (Def->getNextNode() ? Def->getNextNode()->getIterator()
                                       : Def->getParent()->end()));
+  SpirvTy = SpirvTy ? SpirvTy : GR->getOrCreateSPIRVType(Ty, MIB);
   Register NewReg = MRI.createGenericVirtualRegister(MRI.getType(Reg));
   if (auto *RC = MRI.getRegClassOrNull(Reg)) {
     MRI.setRegClass(NewReg, RC);
@@ -271,7 +297,6 @@ Register insertAssignInstr(Register Reg, Type *Ty, SPIRVType *SpirvTy,
     MRI.setRegClass(NewReg, &SPIRV::IDRegClass);
     MRI.setRegClass(Reg, &SPIRV::IDRegClass);
   }
-  SpirvTy = SpirvTy ? SpirvTy : GR->getOrCreateSPIRVType(Ty, MIB);
   GR->assignSPIRVTypeToVReg(SpirvTy, Reg, MIB.getMF());
   // This is to make it convenient for Legalizer to get the SPIRVType
   // when processing the actual MI (i.e. not pseudo one).
@@ -290,11 +315,11 @@ Register insertAssignInstr(Register Reg, Type *Ty, SPIRVType *SpirvTy,
 
 void processInstr(MachineInstr &MI, MachineIRBuilder &MIB,
                   MachineRegisterInfo &MRI, SPIRVGlobalRegistry *GR) {
-  unsigned Opc = MI.getOpcode();
   assert(MI.getNumDefs() > 0 && MRI.hasOneUse(MI.getOperand(0).getReg()));
   MachineInstr &AssignTypeInst =
       *(MRI.use_instr_begin(MI.getOperand(0).getReg()));
-  auto NewReg = createNewIdReg(MI.getOperand(0).getReg(), Opc, MRI, *GR).first;
+  auto NewReg =
+      createNewIdReg(nullptr, MI.getOperand(0).getReg(), MRI, *GR).first;
   AssignTypeInst.getOperand(1).setReg(NewReg);
   MI.getOperand(0).setReg(NewReg);
   MIB.setInsertPt(*MI.getParent(),
@@ -303,7 +328,7 @@ void processInstr(MachineInstr &MI, MachineIRBuilder &MIB,
   for (auto &Op : MI.operands()) {
     if (!Op.isReg() || Op.isDef())
       continue;
-    auto IdOpInfo = createNewIdReg(Op.getReg(), Opc, MRI, *GR);
+    auto IdOpInfo = createNewIdReg(nullptr, Op.getReg(), MRI, *GR);
     MIB.buildInstr(IdOpInfo.second).addDef(IdOpInfo.first).addUse(Op.getReg());
     Op.setReg(IdOpInfo.first);
   }
@@ -419,6 +444,7 @@ static void processInstrsWithTypeFolding(MachineFunction &MF,
         processInstr(MI, MIB, MRI, GR);
     }
   }
+
   for (MachineBasicBlock &MBB : MF) {
     for (MachineInstr &MI : MBB) {
       // We need to rewrite dst types for ASSIGN_TYPE instrs to be able
@@ -431,16 +457,18 @@ static void processInstrsWithTypeFolding(MachineFunction &MF,
       if (!isTypeFoldingSupported(Opcode))
         continue;
       Register DstReg = MI.getOperand(0).getReg();
-      if (MRI.getType(DstReg).isVector())
+      bool IsDstPtr = MRI.getType(DstReg).isPointer();
+      if (IsDstPtr || MRI.getType(DstReg).isVector())
         MRI.setRegClass(DstReg, &SPIRV::IDRegClass);
       // Don't need to reset type of register holding constant and used in
-      // G_ADDRSPACE_CAST, since it braaks legalizer.
+      // G_ADDRSPACE_CAST, since it breaks legalizer.
       if (Opcode == TargetOpcode::G_CONSTANT && MRI.hasOneUse(DstReg)) {
         MachineInstr &UseMI = *MRI.use_instr_begin(DstReg);
         if (UseMI.getOpcode() == TargetOpcode::G_ADDRSPACE_CAST)
           continue;
       }
-      MRI.setType(DstReg, LLT::scalar(32));
+      MRI.setType(DstReg, IsDstPtr ? LLT::pointer(0, GR->getPointerSize())
+                                   : LLT::scalar(32));
     }
   }
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
index 5983c9229cb3c..ecd99f1840d7e 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterBankInfo.cpp
@@ -27,23 +27,7 @@ using namespace llvm;
 const RegisterBank &
 SPIRVRegisterBankInfo::getRegBankFromRegClass(const TargetRegisterClass &RC,
                                               LLT Ty) const {
-  switch (RC.getID()) {
-  case SPIRV::TYPERegClassID:
+  if (RC.getID() == SPIRV::TYPERegClassID)
     return SPIRV::TYPERegBank;
-  case SPIRV::pIDRegClassID:
-  case SPIRV::IDRegClassID:
-    return SPIRV::IDRegBank;
-  case SPIRV::fIDRegClassID:
-    return SPIRV::fIDRegBank;
-  case SPIRV::vIDRegClassID:
-    return SPIRV::vIDRegBank;
-  case SPIRV::vfIDRegClassID:
-    return SPIRV::vfIDRegBank;
-  case SPIRV::vpIDRegClassID:
-    return SPIRV::vpIDRegBank;
-  case SPIRV::ANYIDRegClassID:
-  case SPIRV::ANYRegClassID:
-    return SPIRV::IDRegBank;
-  }
-  llvm_unreachable("Unknown register class");
+  return SPIRV::IDRegBank;
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td b/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
index c7f1e172f3d4f..dea2ef402d3d9 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterBanks.td
@@ -8,9 +8,6 @@
 
 // Although RegisterBankSelection is disabled we need to distinct the banks
 // as InstructionSelector RegClass checking code relies on them
-def IDRegBank : RegisterBank<"IDBank", [ID]>;
-def fIDRegBank : RegisterBank<"fIDBank", [fID]>;
-def vIDRegBank : RegisterBank<"vIDBank", [vID]>;
-def vfIDRegBank : RegisterBank<"vfIDBank", [vfID]>;
-def vpIDRegBank : RegisterBank<"vpIDBank", [vpID]>;
+
 def TYPERegBank : RegisterBank<"TYPEBank", [TYPE]>;
+def IDRegBank : RegisterBank<"IDBank", [ID, fID, pID32, pID64, vID, vfID, vpID32, vpID64]>;
diff --git a/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td b/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
index 6d2bfb91a97f1..9231d22e8d836 100644
--- a/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
+++ b/llvm/lib/Target/SPIRV/SPIRVRegisterInfo.td
@@ -11,39 +11,46 @@
 //===----------------------------------------------------------------------===//
 
 let Namespace = "SPIRV" in {
-  def p0 : PtrValueType <i32, 0>;
-
-  class P0Vec<ValueType scalar>
-      : PtrValueType <scalar, 0> {
-    let nElem = 2;
-    let ElementType = p0;
-    let isInteger = false;
-    let isFP = false;
-    let isVector = true;
+  // Pointer types for patterns with the GlobalISelEmitter
+  def p32 : PtrValueType <i32, 0>;
+  def p64 : PtrValueType <i64, 0>;
+
+  class VTPtrVec<int nelem, PtrValueType ptr>
+      : VTVec<nelem, ValueType<ptr.Size, ptr.Value>, ptr.Value> {
+    int isPointer = true;
   }
 
-  def v2p0 : P0Vec<i32>;
-  // All registers are for 32-bit identifiers, so have a single dummy register
+  def v2p32 : VTPtrVec<2, p32>;
+  def v2p64 : VTPtrVec<2, p64>;
 
-  // Class for registers that are the result of OpTypeXXX instructions
+  // Class for type registers
   def TYPE0 : Register<"TYPE0">;
   def TYPE : RegisterClass<"SPIRV", [i32], 32, (add TYPE0)>;
 
-  // Class for every other non-type ID
+  // Class for non-type registers
   def ID0 : Register<"ID0">;
-  def ID : RegisterClass<"SPIRV", [i32], 32, (add ID0)>;
   def fID0 : Register<"fID0">;
-  def fID : RegisterClass<"SPIRV", [f32], 32, (add fID0)>;
-  def pID0 : Register<"pID0">;
-  def pID : RegisterClass<"SPIRV", [p0], 32, (add pID0)>;
+  def pID320 : Register<"pID320">;
+  def pID640 : Register<"pID640">;
   def vID0 : Register<"vID0">;
-  def vID : RegisterClass<"SPIRV", [v2i32], 32, (add vID0)>;
   def vfID0 : Register<"vfID0">;
+  def vpID320 : Register<"vpID320">;
+  def vpID640 : Register<"vpID640">;
+
+  def ID : RegisterClass<"SPIRV", [i32], 32, (add ID0)>;
+  def fID : RegisterClass<"SPIRV", [f32], 32, (add fID0)>;
+  def pID32 : RegisterClass<"SPIRV", [p32], 32, (add pID320)>;
+  def pID64 : RegisterClass<"SPIRV", [p64], 32, (add pID640)>;
+  def vID : RegisterClass<"SPIRV", [v2i32], 32, (add vID0)>;
   def vfID : RegisterClass<"SPIRV", [v2f32], 32, (add vfID0)>;
-  def vpID0 : Register<"vpID0">;
-  def vpID : RegisterClass<"SPIRV", [v2p0], 32, (add vpID0)>;
+  def vpID32 : RegisterClass<"SPIRV", [v2p32], 32, (add vpID320)>;
+  def vpID64 : RegisterClass<"SPIRV", [v2p64], 32, (add vpID640)>;
 
-  def ANYID : RegisterClass<"SPIRV", [i32, f32, p0, v2i32, v2f32], 32, (add ID, fID, pID, vID, vfID)>;
+  def ANYID : RegisterClass<
+      "SPIRV",
+      [i32, f32, p32, p64, v2i32, v2f32, v2p32, v2p64],
+      32,
+      (add ID0, fID0, pID320, pID640, vID0, vfID0, vpID320, vpID640)>;
 
   // A few instructions like OpName can take ids from both type and non-type
   // instructions, so we need a super-class to allow for both to count as valid
diff --git a/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll b/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll
index afc75c616f023..3828fe89e60ae 100644
--- a/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll
+++ b/llvm/test/CodeGen/SPIRV/instructions/select-phi.ll
@@ -1,6 +1,10 @@
 ; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s
+
 ; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown --translator-compatibility-mode %s -o - -filetype=obj | spirv-val %}
 ; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown --translator-compatibility-mode %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
 
 ; CHECK-DAG: %[[Char:.*]] = OpTypeInt 8 0
 ; CHECK-DAG: %[[Long:.*]] = OpTypeInt 32 0
diff --git a/llvm/test/CodeGen/SPIRV/instructions/select-ptr-load.ll b/llvm/test/CodeGen/SPIRV/instructions/select-ptr-load.ll
new file mode 100644
index 0000000000000..0ff28952f8081
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/instructions/select-ptr-load.ll
@@ -0,0 +1,25 @@
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-SPIRV-DAG: %[[Float:.*]] = OpTypeFloat 32
+; CHECK-SPIRV-DAG: %[[FloatPtr:.*]] = OpTypePointer Function %[[Float]]
+; CHECK-SPIRV: OpInBoundsPtrAccessChain %[[FloatPtr]]
+; CHECK-SPIRV: OpInBoundsPtrAccessChain %[[FloatPtr]]
+; CHECK-SPIRV: OpSelect %[[FloatPtr]]
+; CHECK-SPIRV: OpLoad %[[Float]]
+
+%struct = type { [3 x float] }
+
+define spir_kernel void @bar(i1 %sw) {
+entry:
+  %var1 = alloca %struct
+  %var2 = alloca %struct
+  %elem1 = getelementptr inbounds [3 x float], ptr %var1, i64 0, i64 0
+  %elem2 = getelementptr inbounds [3 x float], ptr %var2, i64 0, i64 1
+  %elem = select i1 %sw, ptr %elem1, ptr %elem2
+  %res = load float, ptr %elem
+  ret void
+}
diff --git a/llvm/test/CodeGen/SPIRV/instructions/select.ll b/llvm/test/CodeGen/SPIRV/instructions/select.ll
index c4176b17abb44..9234b97157d9d 100644
--- a/llvm/test/CodeGen/SPIRV/instructions/select.ll
+++ b/llvm/test/CodeGen/SPIRV/instructions/select.ll
@@ -1,6 +1,9 @@
 ; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s
 ; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
 
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
 ; CHECK-DAG:  OpName [[SCALARi32:%.+]] "select_i32"
 ; CHECK-DAG:  OpName [[SCALARPTR:%.+]] "select_ptr"
 ; CHECK-DAG:  OpName [[VEC2i32:%.+]] "select_i32v2"
diff --git a/llvm/test/CodeGen/SPIRV/select.ll b/llvm/test/CodeGen/SPIRV/select-builtin.ll
similarity index 67%
rename from llvm/test/CodeGen/SPIRV/select.ll
rename to llvm/test/CodeGen/SPIRV/select-builtin.ll
index b34e91be1dbcd..6717970d160fc 100644
--- a/llvm/test/CodeGen/SPIRV/select.ll
+++ b/llvm/test/CodeGen/SPIRV/select-builtin.ll
@@ -1,4 +1,6 @@
 ; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
 
 ; CHECK-SPIRV: OpSelect
 

From 42d801d4e42ff8c47c3a24d562774851e3a424f5 Mon Sep 17 00:00:00 2001
From: Vyacheslav Levytskyy <vyacheslav.levytskyy@intel.com>
Date: Wed, 17 Apr 2024 11:50:55 +0200
Subject: [PATCH 716/886] [SPIR-V] Account for zext in a llvm intrinsic call
 (#88903)

This PR addresses an issue that may arise when an integer argument size
differs from a machine word size for the target in a call to llvm
intrinsic. The following example demonstrates the issue:

```
@__const.test.arr = private unnamed_addr addrspace(2) constant [3 x i32] [i32 1, i32 2, i32 3]

define spir_func void @test() {
entry:
  %arr = alloca [3 x i32], align 4
  %dest = bitcast ptr %arr to ptr
  call void @llvm.memcpy.p0.p2.i32(ptr align 4 %dest, ptr addrspace(2) align 4 @__const.test.arr, i32 1024, i1 false)
  ret void
}

declare void @llvm.memcpy.p0.p2.i32(ptr nocapture writeonly, ptr addrspace(2) nocapture readonly, i32, i1)
```

Depending on the target this code may work or may fail without this PR
due to the fact that IR Translation step introduces additional `zext`
when type of the 3rd argument of `@llvm.memcpy.p0.p2.i32` differs from
machine word.

This PR addresses the issue by adding type deduction for a newly
inserted G_ZEXT generic opcode.
---
 llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp   | 25 +++++++++
 .../CodeGen/SPIRV/transcoding/memcpy-zext.ll  | 43 ++++++++++++++++
 .../spirv-private-array-initialization.ll     | 51 +++++++++++--------
 3 files changed, 99 insertions(+), 20 deletions(-)
 create mode 100644 llvm/test/CodeGen/SPIRV/transcoding/memcpy-zext.ll

diff --git a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
index ed03154d4c8dd..d16f6d5bf67ef 100644
--- a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
@@ -171,6 +171,12 @@ static void insertBitcasts(MachineFunction &MF, SPIRVGlobalRegistry *GR,
 //   %1 = G_GLOBAL_VALUE
 //   %2 = COPY %1
 //   %3 = G_ADDRSPACE_CAST %2
+//
+// or
+//
+//  %1 = G_ZEXT %2
+//  G_MEMCPY ... %2 ...
+//
 // New registers have no SPIRVType and no register class info.
 //
 // Set SPIRVType for GV, propagate it from GV to other instructions,
@@ -200,6 +206,24 @@ static SPIRVType *propagateSPIRVType(MachineInstr *MI, SPIRVGlobalRegistry *GR,
         SpirvTy = GR->getOrCreateSPIRVType(Ty, MIB);
         break;
       }
+      case TargetOpcode::G_ZEXT: {
+        if (MI->getOperand(1).isReg()) {
+          if (MachineInstr *DefInstr =
+                  MRI.getVRegDef(MI->getOperand(1).getReg())) {
+            if (SPIRVType *Def = propagateSPIRVType(DefInstr, GR, MRI, MIB)) {
+              unsigned CurrentBW = GR->getScalarOrVectorBitWidth(Def);
+              unsigned ExpectedBW =
+                  std::max(MRI.getType(Reg).getScalarSizeInBits(), CurrentBW);
+              unsigned NumElements = GR->getScalarOrVectorComponentCount(Def);
+              SpirvTy = GR->getOrCreateSPIRVIntegerType(ExpectedBW, MIB);
+              if (NumElements > 1)
+                SpirvTy =
+                    GR->getOrCreateSPIRVVectorType(SpirvTy, NumElements, MIB);
+            }
+          }
+        }
+        break;
+      }
       case TargetOpcode::G_TRUNC:
       case TargetOpcode::G_ADDRSPACE_CAST:
       case TargetOpcode::G_PTR_ADD:
@@ -415,6 +439,7 @@ static void generateAssignInstrs(MachineFunction &MF, SPIRVGlobalRegistry *GR,
         }
         insertAssignInstr(Reg, Ty, nullptr, GR, MIB, MRI);
       } else if (MI.getOpcode() == TargetOpcode::G_TRUNC ||
+                 MI.getOpcode() == TargetOpcode::G_ZEXT ||
                  MI.getOpcode() == TargetOpcode::G_GLOBAL_VALUE ||
                  MI.getOpcode() == TargetOpcode::COPY ||
                  MI.getOpcode() == TargetOpcode::G_ADDRSPACE_CAST) {
diff --git a/llvm/test/CodeGen/SPIRV/transcoding/memcpy-zext.ll b/llvm/test/CodeGen/SPIRV/transcoding/memcpy-zext.ll
new file mode 100644
index 0000000000000..ea0197548a815
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/transcoding/memcpy-zext.ll
@@ -0,0 +1,43 @@
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-32
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-64
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-64-DAG: %[[#i64:]] = OpTypeInt 64 0
+
+; CHECK-DAG:    %[[#i8:]] = OpTypeInt 8 0
+; CHECK-DAG:    %[[#i32:]] = OpTypeInt 32 0
+; CHECK-DAG:    %[[#one:]] = OpConstant %[[#i32]] 1
+; CHECK-DAG:    %[[#two:]] = OpConstant %[[#i32]] 2
+; CHECK-DAG:    %[[#three:]] = OpConstant %[[#i32]] 3
+; CHECK-DAG:    %[[#i32x3:]] = OpTypeArray %[[#i32]] %[[#three]]
+; CHECK-DAG:    %[[#test_arr_init:]] = OpConstantComposite %[[#i32x3]] %[[#one]] %[[#two]] %[[#three]]
+; CHECK-DAG:    %[[#szconst1024:]] = OpConstant %[[#i32]] 1024
+; CHECK-DAG:    %[[#szconst42:]] = OpConstant %[[#i8]] 42
+; CHECK-DAG:    %[[#const_i32x3_ptr:]] = OpTypePointer UniformConstant %[[#i32x3]]
+; CHECK-DAG:    %[[#test_arr:]] = OpVariable %[[#const_i32x3_ptr]] UniformConstant %[[#test_arr_init]]
+; CHECK-DAG:    %[[#i32x3_ptr:]] = OpTypePointer Function %[[#i32x3]]
+; CHECK:        %[[#arr:]] = OpVariable %[[#i32x3_ptr]] Function
+
+; CHECK-32:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#szconst1024]]
+; CHECK-64:     %[[#szconstext1024:]] = OpUConvert %[[#i64:]] %[[#szconst1024:]]
+; CHECK-64:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#szconstext1024]]
+
+; CHECK-32:     %[[#szconstext42:]] = OpUConvert %[[#i32:]] %[[#szconst42:]]
+; CHECK-32:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#szconstext42]]
+; CHECK-64:     %[[#szconstext42:]] = OpUConvert %[[#i64:]] %[[#szconst42:]]
+; CHECK-64:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#szconstext42]]
+
+@__const.test.arr = private unnamed_addr addrspace(2) constant [3 x i32] [i32 1, i32 2, i32 3]
+
+define spir_func void @test() {
+entry:
+  %arr = alloca [3 x i32], align 4
+  %dest = bitcast ptr %arr to ptr
+  call void @llvm.memcpy.p0.p2.i32(ptr align 4 %dest, ptr addrspace(2) align 4 @__const.test.arr, i32 1024, i1 false)
+  call void @llvm.memcpy.p0.p2.i8(ptr align 4 %dest, ptr addrspace(2) align 4 @__const.test.arr, i8 42, i1 false)
+  ret void
+}
+
+declare void @llvm.memcpy.p0.p2.i32(ptr nocapture writeonly, ptr addrspace(2) nocapture readonly, i32, i1)
+declare void @llvm.memcpy.p0.p2.i8(ptr nocapture writeonly, ptr addrspace(2) nocapture readonly, i8, i1)
diff --git a/llvm/test/CodeGen/SPIRV/transcoding/spirv-private-array-initialization.ll b/llvm/test/CodeGen/SPIRV/transcoding/spirv-private-array-initialization.ll
index e0172ec3c1bdb..04fb39118034c 100644
--- a/llvm/test/CodeGen/SPIRV/transcoding/spirv-private-array-initialization.ll
+++ b/llvm/test/CodeGen/SPIRV/transcoding/spirv-private-array-initialization.ll
@@ -1,23 +1,34 @@
-; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV
-;
-; CHECK-SPIRV-DAG: %[[#i32:]] = OpTypeInt 32 0
-; CHECK-SPIRV-DAG: %[[#one:]] = OpConstant %[[#i32]] 1
-; CHECK-SPIRV-DAG: %[[#two:]] = OpConstant %[[#i32]] 2
-; CHECK-SPIRV-DAG: %[[#three:]] = OpConstant %[[#i32]] 3
-; CHECK-SPIRV-DAG: %[[#i32x3:]] = OpTypeArray %[[#i32]] %[[#three]]
-; CHECK-SPIRV-DAG: %[[#test_arr_init:]] = OpConstantComposite %[[#i32x3]] %[[#one]] %[[#two]] %[[#three]]
-; CHECK-SPIRV-DAG: %[[#twelve:]] = OpConstant %[[#i32]] 12
-; CHECK-SPIRV-DAG: %[[#const_i32x3_ptr:]] = OpTypePointer UniformConstant %[[#i32x3]]
-
-; CHECK-SPIRV:     %[[#test_arr2:]] = OpVariable %[[#const_i32x3_ptr]] UniformConstant %[[#test_arr_init]]
-; CHECK-SPIRV:     %[[#test_arr:]] = OpVariable %[[#const_i32x3_ptr]] UniformConstant %[[#test_arr_init]]
-
-; CHECK-SPIRV-DAG: %[[#i32x3_ptr:]] = OpTypePointer Function %[[#i32x3]]
-
-; CHECK-SPIRV:     %[[#arr:]] = OpVariable %[[#i32x3_ptr]] Function
-; CHECK-SPIRV:     %[[#arr2:]] = OpVariable %[[#i32x3_ptr]] Function
-; CHECK-SPIRV:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#twelve]] Aligned 4
-; CHECK-SPIRV:     OpCopyMemorySized %[[#arr2]] %[[#test_arr2]] %[[#twelve]] Aligned 4
+; RUN: llc -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK-SPIRV,CHECK-SPIRV-32
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+; RUN: llc -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK-SPIRV,CHECK-SPIRV-64
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv64-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-SPIRV-64-DAG: %[[#i64:]] = OpTypeInt 64 0
+
+; CHECK-SPIRV-DAG:    %[[#i32:]] = OpTypeInt 32 0
+; CHECK-SPIRV-DAG:    %[[#one:]] = OpConstant %[[#i32]] 1
+; CHECK-SPIRV-DAG:    %[[#two:]] = OpConstant %[[#i32]] 2
+; CHECK-SPIRV-DAG:    %[[#three:]] = OpConstant %[[#i32]] 3
+; CHECK-SPIRV-DAG:    %[[#i32x3:]] = OpTypeArray %[[#i32]] %[[#three]]
+; CHECK-SPIRV-DAG:    %[[#test_arr_init:]] = OpConstantComposite %[[#i32x3]] %[[#one]] %[[#two]] %[[#three]]
+; CHECK-SPIRV-DAG:    %[[#twelve:]] = OpConstant %[[#i32]] 12
+; CHECK-SPIRV-DAG:    %[[#const_i32x3_ptr:]] = OpTypePointer UniformConstant %[[#i32x3]]
+
+; CHECK-SPIRV:        %[[#test_arr2:]] = OpVariable %[[#const_i32x3_ptr]] UniformConstant %[[#test_arr_init]]
+; CHECK-SPIRV:        %[[#test_arr:]] = OpVariable %[[#const_i32x3_ptr]] UniformConstant %[[#test_arr_init]]
+
+; CHECK-SPIRV-DAG:    %[[#i32x3_ptr:]] = OpTypePointer Function %[[#i32x3]]
+
+; CHECK-SPIRV:        %[[#arr:]] = OpVariable %[[#i32x3_ptr]] Function
+; CHECK-SPIRV:        %[[#arr2:]] = OpVariable %[[#i32x3_ptr]] Function
+
+; CHECK-SPIRV-32:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#twelve]] Aligned 4
+; CHECK-SPIRV-32:     OpCopyMemorySized %[[#arr2]] %[[#test_arr2]] %[[#twelve]] Aligned 4
+
+; CHECK-SPIRV-64:     %[[#twelvezext1:]] = OpUConvert %[[#i64:]] %[[#twelve:]]
+; CHECK-SPIRV-64:     OpCopyMemorySized %[[#arr]] %[[#test_arr]] %[[#twelvezext1]] Aligned 4
+; CHECK-SPIRV-64:     %[[#twelvezext2:]] = OpUConvert %[[#i64:]] %[[#twelve:]]
+; CHECK-SPIRV-64:     OpCopyMemorySized %[[#arr2]] %[[#test_arr2]] %[[#twelvezext2]] Aligned 4
 
 
 @__const.test.arr = private unnamed_addr addrspace(2) constant [3 x i32] [i32 1, i32 2, i32 3], align 4

From fa61f062a515be92a98cac64a9193498918c1225 Mon Sep 17 00:00:00 2001
From: harishch4 <harishcse44@gmail.com>
Date: Wed, 17 Apr 2024 15:21:45 +0530
Subject: [PATCH 717/886] Fix threadprivate variable scope inside BLOCK
 construct. (#88921)

When a local variable inside a BLOCK construct is used as threadprivate
variable, llvm-flang throws below error:

> error: The THREADPRIVATE directive and the common block or variable in
it must appear in the same declaration section of a scoping unit
---
 flang/lib/Semantics/check-omp-structure.cpp     |  2 +-
 flang/test/Lower/OpenMP/threadprivate-hlfir.f90 |  1 +
 flang/test/Semantics/OpenMP/threadprivate07.f90 | 15 +++++++++++++++
 3 files changed, 17 insertions(+), 1 deletion(-)
 create mode 100644 flang/test/Semantics/OpenMP/threadprivate07.f90

diff --git a/flang/lib/Semantics/check-omp-structure.cpp b/flang/lib/Semantics/check-omp-structure.cpp
index e85d8d1f7ab53..bafa242a79302 100644
--- a/flang/lib/Semantics/check-omp-structure.cpp
+++ b/flang/lib/Semantics/check-omp-structure.cpp
@@ -1048,7 +1048,7 @@ void OmpStructureChecker::CheckThreadprivateOrDeclareTargetVar(
                       name->symbol->GetUltimate().owner();
                   if (!curScope.IsTopLevel()) {
                     const semantics::Scope &declScope =
-                        GetProgramUnitContaining(curScope);
+                        GetProgramUnitOrBlockConstructContaining(curScope);
                     const semantics::Symbol *sym{
                         declScope.parent().FindSymbol(name->symbol->name())};
                     if (sym &&
diff --git a/flang/test/Lower/OpenMP/threadprivate-hlfir.f90 b/flang/test/Lower/OpenMP/threadprivate-hlfir.f90
index d39ae1e701183..7d02987c5eade 100644
--- a/flang/test/Lower/OpenMP/threadprivate-hlfir.f90
+++ b/flang/test/Lower/OpenMP/threadprivate-hlfir.f90
@@ -24,3 +24,4 @@ subroutine sub()
     print *, a
   !$omp end parallel
 end subroutine
+
diff --git a/flang/test/Semantics/OpenMP/threadprivate07.f90 b/flang/test/Semantics/OpenMP/threadprivate07.f90
new file mode 100644
index 0000000000000..c9a006ca0e083
--- /dev/null
+++ b/flang/test/Semantics/OpenMP/threadprivate07.f90
@@ -0,0 +1,15 @@
+! RUN: %python %S/../test_errors.py %s %flang_fc1 -fopenmp
+
+! Check Threadprivate Directive with local variable of a BLOCK construct.
+
+program main
+  call sub1()
+  print *, 'pass'
+end program main
+
+subroutine sub1()
+  BLOCK
+    integer, save :: a
+    !$omp threadprivate(a)
+  END BLOCK
+end subroutine

From cbe148b730a04fc95eda9a43903f0af487884a96 Mon Sep 17 00:00:00 2001
From: Mel Chen <mel.chen@sifive.com>
Date: Wed, 17 Apr 2024 17:59:52 +0800
Subject: [PATCH 718/886] [LV][NFC] Remove the declaration of function
 `fixReduction`. (#88491)

---
 llvm/lib/Transforms/Vectorize/LoopVectorize.cpp | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 44885a95bd102..20059f9d62d55 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -616,9 +616,6 @@ class InnerLoopVectorizer {
   void fixFixedOrderRecurrence(VPFirstOrderRecurrencePHIRecipe *PhiR,
                                VPTransformState &State);
 
-  /// Create code for the loop exit value of the reduction.
-  void fixReduction(VPReductionPHIRecipe *Phi, VPTransformState &State);
-
   /// Iteratively sink the scalarized operands of a predicated instruction into
   /// the block that was created for it.
   void sinkScalarOperands(Instruction *PredInst);

From a9bafe91dd088c5fa6f074c14dd3a1af25f00457 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Wed, 17 Apr 2024 11:00:58 +0100
Subject: [PATCH 719/886] [VPlan] Split VPWidenMemoryInstructionRecipe (NFCI).
 (#87411)

This patch introduces a new VPWidenMemoryRecipe base class and distinct
sub-classes to model loads and stores.

This is a first step in an effort to simplify and modularize code
generation for widened loads and stores and enable adding further more
specialized memory recipes.

PR: https://github.com/llvm/llvm-project/pull/87411
---
 .../Transforms/Vectorize/LoopVectorize.cpp    | 191 +++++++++---------
 .../Transforms/Vectorize/VPRecipeBuilder.h    |   6 +-
 llvm/lib/Transforms/Vectorize/VPlan.h         | 161 +++++++++------
 .../Transforms/Vectorize/VPlanAnalysis.cpp    |   9 +-
 llvm/lib/Transforms/Vectorize/VPlanAnalysis.h |   4 +-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  44 ++--
 .../Transforms/Vectorize/VPlanTransforms.cpp  |  23 +--
 llvm/lib/Transforms/Vectorize/VPlanValue.h    |   5 +-
 .../Transforms/Vectorize/VPlanVerifier.cpp    |   2 +-
 .../AArch64/vector-reverse-mask4.ll           |   2 +-
 .../LoopVectorize/X86/masked_load_store.ll    |  12 +-
 .../Transforms/Vectorize/VPlanHCFGTest.cpp    |   4 +-
 .../Transforms/Vectorize/VPlanTest.cpp        |   9 +-
 13 files changed, 249 insertions(+), 223 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 20059f9d62d55..a8272f4502535 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -545,11 +545,6 @@ class InnerLoopVectorizer {
   // Return true if any runtime check is added.
   bool areSafetyChecksAdded() { return AddedSafetyChecks; }
 
-  /// A type for vectorized values in the new loop. Each value from the
-  /// original loop, when vectorized, is represented by UF vector values in the
-  /// new unrolled loop, where UF is the unroll factor.
-  using VectorParts = SmallVector<Value *, 2>;
-
   /// A helper function to scalarize a single Instruction in the innermost loop.
   /// Generates a sequence of scalar instances for each lane between \p MinLane
   /// and \p MaxLane, times each part between \p MinPart and \p MaxPart,
@@ -8086,7 +8081,7 @@ void VPRecipeBuilder::createBlockInMask(BasicBlock *BB) {
   BlockMaskCache[BB] = BlockMask;
 }
 
-VPWidenMemoryInstructionRecipe *
+VPWidenMemoryRecipe *
 VPRecipeBuilder::tryToWidenMemory(Instruction *I, ArrayRef<VPValue *> Operands,
                                   VFRange &Range) {
   assert((isa<LoadInst>(I) || isa<StoreInst>(I)) &&
@@ -8131,12 +8126,12 @@ VPRecipeBuilder::tryToWidenMemory(Instruction *I, ArrayRef<VPValue *> Operands,
     Ptr = VectorPtr;
   }
   if (LoadInst *Load = dyn_cast<LoadInst>(I))
-    return new VPWidenMemoryInstructionRecipe(*Load, Ptr, Mask, Consecutive,
-                                              Reverse, I->getDebugLoc());
+    return new VPWidenLoadRecipe(*Load, Ptr, Mask, Consecutive, Reverse,
+                                 I->getDebugLoc());
 
   StoreInst *Store = cast<StoreInst>(I);
-  return new VPWidenMemoryInstructionRecipe(
-      *Store, Ptr, Operands[0], Mask, Consecutive, Reverse, I->getDebugLoc());
+  return new VPWidenStoreRecipe(*Store, Ptr, Operands[0], Mask, Consecutive,
+                                Reverse, I->getDebugLoc());
 }
 
 /// Creates a VPWidenIntOrFpInductionRecpipe for \p Phi. If needed, it will also
@@ -8775,13 +8770,12 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
   // for this VPlan, replace the Recipes widening its memory instructions with a
   // single VPInterleaveRecipe at its insertion point.
   for (const auto *IG : InterleaveGroups) {
-    auto *Recipe = cast<VPWidenMemoryInstructionRecipe>(
-        RecipeBuilder.getRecipe(IG->getInsertPos()));
+    auto *Recipe =
+        cast<VPWidenMemoryRecipe>(RecipeBuilder.getRecipe(IG->getInsertPos()));
     SmallVector<VPValue *, 4> StoredValues;
     for (unsigned i = 0; i < IG->getFactor(); ++i)
       if (auto *SI = dyn_cast_or_null<StoreInst>(IG->getMember(i))) {
-        auto *StoreR =
-            cast<VPWidenMemoryInstructionRecipe>(RecipeBuilder.getRecipe(SI));
+        auto *StoreR = cast<VPWidenStoreRecipe>(RecipeBuilder.getRecipe(SI));
         StoredValues.push_back(StoreR->getStoredValue());
       }
 
@@ -9368,92 +9362,27 @@ static Instruction *lowerLoadUsingVectorIntrinsics(IRBuilderBase &Builder,
   return Call;
 }
 
-void VPWidenMemoryInstructionRecipe::execute(VPTransformState &State) {
-  VPValue *StoredValue = isStore() ? getStoredValue() : nullptr;
-
-  // Attempt to issue a wide load.
-  LoadInst *LI = dyn_cast<LoadInst>(&Ingredient);
-  StoreInst *SI = dyn_cast<StoreInst>(&Ingredient);
-
-  assert((LI || SI) && "Invalid Load/Store instruction");
-  assert((!SI || StoredValue) && "No stored value provided for widened store");
-  assert((!LI || !StoredValue) && "Stored value provided for widened load");
+void VPWidenLoadRecipe::execute(VPTransformState &State) {
+  auto *LI = cast<LoadInst>(&Ingredient);
 
   Type *ScalarDataTy = getLoadStoreType(&Ingredient);
-
   auto *DataTy = VectorType::get(ScalarDataTy, State.VF);
   const Align Alignment = getLoadStoreAlignment(&Ingredient);
-  bool CreateGatherScatter = !isConsecutive();
+  bool CreateGather = !isConsecutive();
 
   auto &Builder = State.Builder;
-  InnerLoopVectorizer::VectorParts BlockInMaskParts(State.UF);
-  bool isMaskRequired = getMask();
-  if (isMaskRequired) {
-    // Mask reversal is only needed for non-all-one (null) masks, as reverse of
-    // a null all-one mask is a null mask.
-    for (unsigned Part = 0; Part < State.UF; ++Part) {
-      Value *Mask = State.get(getMask(), Part);
+  State.setDebugLocFrom(getDebugLoc());
+  for (unsigned Part = 0; Part < State.UF; ++Part) {
+    Value *NewLI;
+    Value *Mask = nullptr;
+    if (auto *VPMask = getMask()) {
+      // Mask reversal is only needed for non-all-one (null) masks, as reverse
+      // of a null all-one mask is a null mask.
+      Mask = State.get(VPMask, Part);
       if (isReverse())
         Mask = Builder.CreateVectorReverse(Mask, "reverse");
-      BlockInMaskParts[Part] = Mask;
-    }
-  }
-
-  // Handle Stores:
-  if (SI) {
-    State.setDebugLocFrom(getDebugLoc());
-
-    for (unsigned Part = 0; Part < State.UF; ++Part) {
-      Instruction *NewSI = nullptr;
-      Value *StoredVal = State.get(StoredValue, Part);
-      // TODO: split this into several classes for better design.
-      if (State.EVL) {
-        assert(State.UF == 1 && "Expected only UF == 1 when vectorizing with "
-                                "explicit vector length.");
-        assert(cast<VPInstruction>(State.EVL)->getOpcode() ==
-                   VPInstruction::ExplicitVectorLength &&
-               "EVL must be VPInstruction::ExplicitVectorLength.");
-        Value *EVL = State.get(State.EVL, VPIteration(0, 0));
-        // If EVL is not nullptr, then EVL must be a valid value set during plan
-        // creation, possibly default value = whole vector register length. EVL
-        // is created only if TTI prefers predicated vectorization, thus if EVL
-        // is not nullptr it also implies preference for predicated
-        // vectorization.
-        // FIXME: Support reverse store after vp_reverse is added.
-        Value *MaskPart = isMaskRequired ? BlockInMaskParts[Part] : nullptr;
-        NewSI = lowerStoreUsingVectorIntrinsics(
-            Builder, State.get(getAddr(), Part, !CreateGatherScatter),
-            StoredVal, CreateGatherScatter, MaskPart, EVL, Alignment);
-      } else if (CreateGatherScatter) {
-        Value *MaskPart = isMaskRequired ? BlockInMaskParts[Part] : nullptr;
-        Value *VectorGep = State.get(getAddr(), Part);
-        NewSI = Builder.CreateMaskedScatter(StoredVal, VectorGep, Alignment,
-                                            MaskPart);
-      } else {
-        if (isReverse()) {
-          // If we store to reverse consecutive memory locations, then we need
-          // to reverse the order of elements in the stored value.
-          StoredVal = Builder.CreateVectorReverse(StoredVal, "reverse");
-          // We don't want to update the value in the map as it might be used in
-          // another expression. So don't call resetVectorValue(StoredVal).
-        }
-        auto *VecPtr = State.get(getAddr(), Part, /*IsScalar*/ true);
-        if (isMaskRequired)
-          NewSI = Builder.CreateMaskedStore(StoredVal, VecPtr, Alignment,
-                                            BlockInMaskParts[Part]);
-        else
-          NewSI = Builder.CreateAlignedStore(StoredVal, VecPtr, Alignment);
-      }
-      State.addMetadata(NewSI, SI);
     }
-    return;
-  }
 
-  // Handle loads.
-  assert(LI && "Must have a load instruction");
-  State.setDebugLocFrom(getDebugLoc());
-  for (unsigned Part = 0; Part < State.UF; ++Part) {
-    Value *NewLI;
     // TODO: split this into several classes for better design.
     if (State.EVL) {
       assert(State.UF == 1 && "Expected only UF == 1 when vectorizing with "
@@ -9468,22 +9397,20 @@ void VPWidenMemoryInstructionRecipe::execute(VPTransformState &State) {
       // is not nullptr it also implies preference for predicated
       // vectorization.
       // FIXME: Support reverse loading after vp_reverse is added.
-      Value *MaskPart = isMaskRequired ? BlockInMaskParts[Part] : nullptr;
       NewLI = lowerLoadUsingVectorIntrinsics(
-          Builder, DataTy, State.get(getAddr(), Part, !CreateGatherScatter),
-          CreateGatherScatter, MaskPart, EVL, Alignment);
-    } else if (CreateGatherScatter) {
-      Value *MaskPart = isMaskRequired ? BlockInMaskParts[Part] : nullptr;
+          Builder, DataTy, State.get(getAddr(), Part, !CreateGather),
+          CreateGather, Mask, EVL, Alignment);
+    } else if (CreateGather) {
       Value *VectorGep = State.get(getAddr(), Part);
-      NewLI = Builder.CreateMaskedGather(DataTy, VectorGep, Alignment, MaskPart,
+      NewLI = Builder.CreateMaskedGather(DataTy, VectorGep, Alignment, Mask,
                                          nullptr, "wide.masked.gather");
       State.addMetadata(NewLI, LI);
     } else {
       auto *VecPtr = State.get(getAddr(), Part, /*IsScalar*/ true);
-      if (isMaskRequired)
-        NewLI = Builder.CreateMaskedLoad(
-            DataTy, VecPtr, Alignment, BlockInMaskParts[Part],
-            PoisonValue::get(DataTy), "wide.masked.load");
+      if (Mask)
+        NewLI = Builder.CreateMaskedLoad(DataTy, VecPtr, Alignment, Mask,
+                                         PoisonValue::get(DataTy),
+                                         "wide.masked.load");
       else
         NewLI =
             Builder.CreateAlignedLoad(DataTy, VecPtr, Alignment, "wide.load");
@@ -9494,7 +9421,69 @@ void VPWidenMemoryInstructionRecipe::execute(VPTransformState &State) {
         NewLI = Builder.CreateVectorReverse(NewLI, "reverse");
     }
 
-    State.set(getVPSingleValue(), NewLI, Part);
+    State.set(this, NewLI, Part);
+  }
+}
+
+void VPWidenStoreRecipe::execute(VPTransformState &State) {
+  auto *SI = cast<StoreInst>(&Ingredient);
+
+  VPValue *StoredVPValue = getStoredValue();
+  bool CreateScatter = !isConsecutive();
+  const Align Alignment = getLoadStoreAlignment(&Ingredient);
+
+  auto &Builder = State.Builder;
+  State.setDebugLocFrom(getDebugLoc());
+
+  for (unsigned Part = 0; Part < State.UF; ++Part) {
+    Instruction *NewSI = nullptr;
+    Value *Mask = nullptr;
+    if (auto *VPMask = getMask()) {
+      // Mask reversal is only needed for non-all-one (null) masks, as reverse
+      // of a null all-one mask is a null mask.
+      Mask = State.get(VPMask, Part);
+      if (isReverse())
+        Mask = Builder.CreateVectorReverse(Mask, "reverse");
+    }
+
+    Value *StoredVal = State.get(StoredVPValue, Part);
+    if (isReverse()) {
+      assert(!State.EVL && "reversing not yet implemented with EVL");
+      // If we store to reverse consecutive memory locations, then we need
+      // to reverse the order of elements in the stored value.
+      StoredVal = Builder.CreateVectorReverse(StoredVal, "reverse");
+      // We don't want to update the value in the map as it might be used in
+      // another expression. So don't call resetVectorValue(StoredVal).
+    }
+    // TODO: split this into several classes for better design.
+    if (State.EVL) {
+      assert(State.UF == 1 && "Expected only UF == 1 when vectorizing with "
+                              "explicit vector length.");
+      assert(cast<VPInstruction>(State.EVL)->getOpcode() ==
+                 VPInstruction::ExplicitVectorLength &&
+             "EVL must be VPInstruction::ExplicitVectorLength.");
+      Value *EVL = State.get(State.EVL, VPIteration(0, 0));
+      // If EVL is not nullptr, then EVL must be a valid value set during plan
+      // creation, possibly default value = whole vector register length. EVL
+      // is created only if TTI prefers predicated vectorization, thus if EVL
+      // is not nullptr it also implies preference for predicated
+      // vectorization.
+      // FIXME: Support reverse store after vp_reverse is added.
+      NewSI = lowerStoreUsingVectorIntrinsics(
+          Builder, State.get(getAddr(), Part, !CreateScatter), StoredVal,
+          CreateScatter, Mask, EVL, Alignment);
+    } else if (CreateScatter) {
+      Value *VectorGep = State.get(getAddr(), Part);
+      NewSI =
+          Builder.CreateMaskedScatter(StoredVal, VectorGep, Alignment, Mask);
+    } else {
+      auto *VecPtr = State.get(getAddr(), Part, /*IsScalar*/ true);
+      if (Mask)
+        NewSI = Builder.CreateMaskedStore(StoredVal, VecPtr, Alignment, Mask);
+      else
+        NewSI = Builder.CreateAlignedStore(StoredVal, VecPtr, Alignment);
+    }
+    State.addMetadata(NewSI, SI);
   }
 }
 
diff --git a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
index 605b47fa0a46b..b4c7ab02f928f 100644
--- a/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
+++ b/llvm/lib/Transforms/Vectorize/VPRecipeBuilder.h
@@ -69,9 +69,9 @@ class VPRecipeBuilder {
   /// Check if the load or store instruction \p I should widened for \p
   /// Range.Start and potentially masked. Such instructions are handled by a
   /// recipe that takes an additional VPInstruction for the mask.
-  VPWidenMemoryInstructionRecipe *tryToWidenMemory(Instruction *I,
-                                                   ArrayRef<VPValue *> Operands,
-                                                   VFRange &Range);
+  VPWidenMemoryRecipe *tryToWidenMemory(Instruction *I,
+                                        ArrayRef<VPValue *> Operands,
+                                        VFRange &Range);
 
   /// Check if an induction recipe should be constructed for \p Phi. If so build
   /// and return it. If not, return null.
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index d86a81d4fb4c7..148227f1f1a57 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -875,7 +875,8 @@ class VPSingleDefRecipe : public VPRecipeBase, public VPValue {
       return true;
     case VPRecipeBase::VPInterleaveSC:
     case VPRecipeBase::VPBranchOnMaskSC:
-    case VPRecipeBase::VPWidenMemoryInstructionSC:
+    case VPRecipeBase::VPWidenLoadSC:
+    case VPRecipeBase::VPWidenStoreSC:
       // TODO: Widened stores don't define a value, but widened loads do. Split
       // the recipes to be able to make widened loads VPSingleDefRecipes.
       return false;
@@ -2280,68 +2281,62 @@ class VPPredInstPHIRecipe : public VPSingleDefRecipe {
   }
 };
 
-/// A Recipe for widening load/store operations.
-/// The recipe uses the following VPValues:
-/// - For load: Address, optional mask
-/// - For store: Address, stored value, optional mask
-/// TODO: We currently execute only per-part unless a specific instance is
-/// provided.
-class VPWidenMemoryInstructionRecipe : public VPRecipeBase {
+/// A common base class for widening memory operations. An optional mask can be
+/// provided as the last operand.
+class VPWidenMemoryRecipe : public VPRecipeBase {
+protected:
   Instruction &Ingredient;
 
-  // Whether the loaded-from / stored-to addresses are consecutive.
+  /// Whether the accessed addresses are consecutive.
   bool Consecutive;
 
-  // Whether the consecutive loaded/stored addresses are in reverse order.
+  /// Whether the consecutive accessed addresses are in reverse order.
   bool Reverse;
 
+  /// Whether the memory access is masked.
+  bool IsMasked = false;
+
   void setMask(VPValue *Mask) {
+    assert(!IsMasked && "cannot re-set mask");
     if (!Mask)
       return;
     addOperand(Mask);
+    IsMasked = true;
   }
 
-  bool isMasked() const {
-    return isStore() ? getNumOperands() == 3 : getNumOperands() == 2;
+  VPWidenMemoryRecipe(const char unsigned SC, Instruction &I,
+                      std::initializer_list<VPValue *> Operands,
+                      bool Consecutive, bool Reverse, DebugLoc DL)
+      : VPRecipeBase(SC, Operands, DL), Ingredient(I), Consecutive(Consecutive),
+        Reverse(Reverse) {
+    assert((Consecutive || !Reverse) && "Reverse implies consecutive");
   }
 
 public:
-  VPWidenMemoryInstructionRecipe(LoadInst &Load, VPValue *Addr, VPValue *Mask,
-                                 bool Consecutive, bool Reverse, DebugLoc DL)
-      : VPRecipeBase(VPDef::VPWidenMemoryInstructionSC, {Addr}, DL),
-        Ingredient(Load), Consecutive(Consecutive), Reverse(Reverse) {
-    assert((Consecutive || !Reverse) && "Reverse implies consecutive");
-    new VPValue(this, &Load);
-    setMask(Mask);
-  }
+  VPWidenMemoryRecipe *clone() override = 0;
 
-  VPWidenMemoryInstructionRecipe(StoreInst &Store, VPValue *Addr,
-                                 VPValue *StoredValue, VPValue *Mask,
-                                 bool Consecutive, bool Reverse, DebugLoc DL)
-      : VPRecipeBase(VPDef::VPWidenMemoryInstructionSC, {Addr, StoredValue},
-                     DL),
-        Ingredient(Store), Consecutive(Consecutive), Reverse(Reverse) {
-    assert((Consecutive || !Reverse) && "Reverse implies consecutive");
-    setMask(Mask);
+  static inline bool classof(const VPRecipeBase *R) {
+    return R->getVPDefID() == VPDef::VPWidenLoadSC ||
+           R->getVPDefID() == VPDef::VPWidenStoreSC;
   }
 
-  VPWidenMemoryInstructionRecipe *clone() override {
-    if (isStore())
-      return new VPWidenMemoryInstructionRecipe(
-          cast<StoreInst>(Ingredient), getAddr(), getStoredValue(), getMask(),
-          Consecutive, Reverse, getDebugLoc());
-
-    return new VPWidenMemoryInstructionRecipe(cast<LoadInst>(Ingredient),
-                                              getAddr(), getMask(), Consecutive,
-                                              Reverse, getDebugLoc());
+  static inline bool classof(const VPUser *U) {
+    auto *R = dyn_cast<VPRecipeBase>(U);
+    return R && classof(R);
   }
 
-  VP_CLASSOF_IMPL(VPDef::VPWidenMemoryInstructionSC)
+  /// Return whether the loaded-from / stored-to addresses are consecutive.
+  bool isConsecutive() const { return Consecutive; }
+
+  /// Return whether the consecutive loaded/stored addresses are in reverse
+  /// order.
+  bool isReverse() const { return Reverse; }
 
   /// Return the address accessed by this recipe.
-  VPValue *getAddr() const {
-    return getOperand(0); // Address is the 1st, mandatory operand.
-  }
+  VPValue *getAddr() const { return getOperand(0); }
+
+  /// Returns true if the recipe is masked.
+  bool isMasked() const { return IsMasked; }
 
   /// Return the mask used by this recipe. Note that a full mask is represented
   /// by a nullptr.
@@ -2350,23 +2345,34 @@ class VPWidenMemoryInstructionRecipe : public VPRecipeBase {
     return isMasked() ? getOperand(getNumOperands() - 1) : nullptr;
   }
 
-  /// Returns true if this recipe is a store.
-  bool isStore() const { return isa<StoreInst>(Ingredient); }
+  /// Generate the wide load/store.
+  void execute(VPTransformState &State) override {
+    llvm_unreachable("VPWidenMemoryRecipe should not be instantiated.");
+  }
 
-  /// Return the address accessed by this recipe.
-  VPValue *getStoredValue() const {
-    assert(isStore() && "Stored value only available for store instructions");
-    return getOperand(1); // Stored value is the 2nd, mandatory operand.
+  Instruction &getIngredient() const { return Ingredient; }
+};
+
+/// A recipe for widening load operations, using the address to load from and an
+/// optional mask.
+struct VPWidenLoadRecipe final : public VPWidenMemoryRecipe, public VPValue {
+  VPWidenLoadRecipe(LoadInst &Load, VPValue *Addr, VPValue *Mask,
+                    bool Consecutive, bool Reverse, DebugLoc DL)
+      : VPWidenMemoryRecipe(VPDef::VPWidenLoadSC, Load, {Addr}, Consecutive,
+                            Reverse, DL),
+        VPValue(this, &Load) {
+    setMask(Mask);
   }
 
-  // Return whether the loaded-from / stored-to addresses are consecutive.
-  bool isConsecutive() const { return Consecutive; }
+  VPWidenLoadRecipe *clone() override {
+    return new VPWidenLoadRecipe(cast<LoadInst>(Ingredient), getAddr(),
+                                 getMask(), Consecutive, Reverse,
+                                 getDebugLoc());
+  }
 
-  // Return whether the consecutive loaded/stored addresses are in reverse
-  // order.
-  bool isReverse() const { return Reverse; }
+  VP_CLASSOF_IMPL(VPDef::VPWidenLoadSC);
 
-  /// Generate the wide load/store.
+  /// Generate a wide load or gather.
   void execute(VPTransformState &State) override;
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
@@ -2380,16 +2386,51 @@ class VPWidenMemoryInstructionRecipe : public VPRecipeBase {
     assert(is_contained(operands(), Op) &&
            "Op must be an operand of the recipe");
 
-    // Widened, consecutive memory operations only demand the first lane of
-    // their address, unless the same operand is also stored. That latter can
-    // happen with opaque pointers.
-    return Op == getAddr() && isConsecutive() &&
-           (!isStore() || Op != getStoredValue());
+    // Widened, consecutive loads operations only demand the first lane of
+    // their address.
+    return Op == getAddr() && isConsecutive();
   }
-
-  Instruction &getIngredient() const { return Ingredient; }
 };
 
+/// A recipe for widening store operations, using the stored value, the address
+/// to store to and an optional mask.
+struct VPWidenStoreRecipe final : public VPWidenMemoryRecipe {
+  VPWidenStoreRecipe(StoreInst &Store, VPValue *Addr, VPValue *StoredVal,
+                     VPValue *Mask, bool Consecutive, bool Reverse, DebugLoc DL)
+      : VPWidenMemoryRecipe(VPDef::VPWidenStoreSC, Store, {Addr, StoredVal},
+                            Consecutive, Reverse, DL) {
+    setMask(Mask);
+  }
+
+  VPWidenStoreRecipe *clone() override {
+    return new VPWidenStoreRecipe(cast<StoreInst>(Ingredient), getAddr(),
+                                  getStoredValue(), getMask(), Consecutive,
+                                  Reverse, getDebugLoc());
+  }
+
+  VP_CLASSOF_IMPL(VPDef::VPWidenStoreSC);
+
+  /// Return the value stored by this recipe.
+  VPValue *getStoredValue() const { return getOperand(1); }
+
+  /// Generate a wide store or scatter.
+  void execute(VPTransformState &State) override;
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+  /// Print the recipe.
+  void print(raw_ostream &O, const Twine &Indent,
+             VPSlotTracker &SlotTracker) const override;
+#endif
+
+  /// Returns true if the recipe only uses the first lane of operand \p Op.
+  bool onlyFirstLaneUsed(const VPValue *Op) const override {
+    assert(is_contained(operands(), Op) &&
+           "Op must be an operand of the recipe");
+    // Widened, consecutive stores only demand the first lane of their address,
+    // unless the same operand is also stored.
+    return Op == getAddr() && isConsecutive() && Op != getStoredValue();
+  }
+};
 /// Recipe to expand a SCEV expression.
 class VPExpandSCEVRecipe : public VPSingleDefRecipe {
   const SCEV *Expr;
diff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
index c8ae2ee5a30fe..130fb04f586e7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
@@ -108,9 +108,9 @@ Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPWidenCallRecipe *R) {
   return CI.getType();
 }
 
-Type *VPTypeAnalysis::inferScalarTypeForRecipe(
-    const VPWidenMemoryInstructionRecipe *R) {
-  assert(!R->isStore() && "Store recipes should not define any values");
+Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPWidenMemoryRecipe *R) {
+  assert(isa<VPWidenLoadRecipe>(R) &&
+         "Store recipes should not define any values");
   return cast<LoadInst>(&R->getIngredient())->getType();
 }
 
@@ -231,8 +231,7 @@ Type *VPTypeAnalysis::inferScalarType(const VPValue *V) {
             return inferScalarType(R->getOperand(0));
           })
           .Case<VPBlendRecipe, VPInstruction, VPWidenRecipe, VPReplicateRecipe,
-                VPWidenCallRecipe, VPWidenMemoryInstructionRecipe,
-                VPWidenSelectRecipe>(
+                VPWidenCallRecipe, VPWidenMemoryRecipe, VPWidenSelectRecipe>(
               [this](const auto *R) { return inferScalarTypeForRecipe(R); })
           .Case<VPInterleaveRecipe>([V](const VPInterleaveRecipe *R) {
             // TODO: Use info from interleave group.
diff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.h b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.h
index 4e69de7fd6812..7d310b1b31b6f 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.h
@@ -20,7 +20,7 @@ class VPInstruction;
 class VPWidenRecipe;
 class VPWidenCallRecipe;
 class VPWidenIntOrFpInductionRecipe;
-class VPWidenMemoryInstructionRecipe;
+class VPWidenMemoryRecipe;
 struct VPWidenSelectRecipe;
 class VPReplicateRecipe;
 class Type;
@@ -46,7 +46,7 @@ class VPTypeAnalysis {
   Type *inferScalarTypeForRecipe(const VPWidenCallRecipe *R);
   Type *inferScalarTypeForRecipe(const VPWidenRecipe *R);
   Type *inferScalarTypeForRecipe(const VPWidenIntOrFpInductionRecipe *R);
-  Type *inferScalarTypeForRecipe(const VPWidenMemoryInstructionRecipe *R);
+  Type *inferScalarTypeForRecipe(const VPWidenMemoryRecipe *R);
   Type *inferScalarTypeForRecipe(const VPWidenSelectRecipe *R);
   Type *inferScalarTypeForRecipe(const VPReplicateRecipe *R);
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 9f242a1bee8f6..78932643c81fa 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -47,9 +47,8 @@ bool VPRecipeBase::mayWriteToMemory() const {
   switch (getVPDefID()) {
   case VPInterleaveSC:
     return cast<VPInterleaveRecipe>(this)->getNumStoreOperands() > 0;
-  case VPWidenMemoryInstructionSC: {
-    return cast<VPWidenMemoryInstructionRecipe>(this)->isStore();
-  }
+  case VPWidenStoreSC:
+    return true;
   case VPReplicateSC:
   case VPWidenCallSC:
     return cast<Instruction>(getVPSingleValue()->getUnderlyingValue())
@@ -64,6 +63,7 @@ bool VPRecipeBase::mayWriteToMemory() const {
   case VPWidenCastSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
+  case VPWidenLoadSC:
   case VPWidenPHISC:
   case VPWidenSC:
   case VPWidenSelectSC: {
@@ -81,16 +81,16 @@ bool VPRecipeBase::mayWriteToMemory() const {
 
 bool VPRecipeBase::mayReadFromMemory() const {
   switch (getVPDefID()) {
-  case VPWidenMemoryInstructionSC: {
-    return !cast<VPWidenMemoryInstructionRecipe>(this)->isStore();
-  }
+  case VPWidenLoadSC:
+    return true;
   case VPReplicateSC:
   case VPWidenCallSC:
     return cast<Instruction>(getVPSingleValue()->getUnderlyingValue())
         ->mayReadFromMemory();
   case VPBranchOnMaskSC:
-  case VPScalarIVStepsSC:
   case VPPredInstPHISC:
+  case VPScalarIVStepsSC:
+  case VPWidenStoreSC:
     return false;
   case VPBlendSC:
   case VPReductionSC:
@@ -155,12 +155,13 @@ bool VPRecipeBase::mayHaveSideEffects() const {
   }
   case VPInterleaveSC:
     return mayWriteToMemory();
-  case VPWidenMemoryInstructionSC:
-    assert(cast<VPWidenMemoryInstructionRecipe>(this)
-                   ->getIngredient()
-                   .mayHaveSideEffects() == mayWriteToMemory() &&
-           "mayHaveSideffects result for ingredient differs from this "
-           "implementation");
+  case VPWidenLoadSC:
+  case VPWidenStoreSC:
+    assert(
+        cast<VPWidenMemoryRecipe>(this)->getIngredient().mayHaveSideEffects() ==
+            mayWriteToMemory() &&
+        "mayHaveSideffects result for ingredient differs from this "
+        "implementation");
     return mayWriteToMemory();
   case VPReplicateSC: {
     auto *R = cast<VPReplicateRecipe>(this);
@@ -1769,16 +1770,17 @@ void VPPredInstPHIRecipe::print(raw_ostream &O, const Twine &Indent,
   printOperands(O, SlotTracker);
 }
 
-void VPWidenMemoryInstructionRecipe::print(raw_ostream &O, const Twine &Indent,
-                                           VPSlotTracker &SlotTracker) const {
+void VPWidenLoadRecipe::print(raw_ostream &O, const Twine &Indent,
+                              VPSlotTracker &SlotTracker) const {
   O << Indent << "WIDEN ";
+  printAsOperand(O, SlotTracker);
+  O << " = load ";
+  printOperands(O, SlotTracker);
+}
 
-  if (!isStore()) {
-    getVPSingleValue()->printAsOperand(O, SlotTracker);
-    O << " = ";
-  }
-  O << Instruction::getOpcodeName(Ingredient.getOpcode()) << " ";
-
+void VPWidenStoreRecipe::print(raw_ostream &O, const Twine &Indent,
+                               VPSlotTracker &SlotTracker) const {
+  O << Indent << "WIDEN store ";
   printOperands(O, SlotTracker);
 }
 #endif
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 1256e4d8fda50..382bf5ac11405 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -60,14 +60,14 @@ void VPlanTransforms::VPInstructionsToVPRecipes(
         assert(isa<VPInstruction>(&Ingredient) &&
                "only VPInstructions expected here");
         assert(!isa<PHINode>(Inst) && "phis should be handled above");
-        // Create VPWidenMemoryInstructionRecipe for loads and stores.
+        // Create VPWidenMemoryRecipe for loads and stores.
         if (LoadInst *Load = dyn_cast<LoadInst>(Inst)) {
-          NewRecipe = new VPWidenMemoryInstructionRecipe(
+          NewRecipe = new VPWidenLoadRecipe(
               *Load, Ingredient.getOperand(0), nullptr /*Mask*/,
               false /*Consecutive*/, false /*Reverse*/,
               Ingredient.getDebugLoc());
         } else if (StoreInst *Store = dyn_cast<StoreInst>(Inst)) {
-          NewRecipe = new VPWidenMemoryInstructionRecipe(
+          NewRecipe = new VPWidenStoreRecipe(
               *Store, Ingredient.getOperand(1), Ingredient.getOperand(0),
               nullptr /*Mask*/, false /*Consecutive*/, false /*Reverse*/,
               Ingredient.getDebugLoc());
@@ -977,10 +977,9 @@ void VPlanTransforms::truncateToMinimalBitwidths(
            vp_depth_first_deep(Plan.getVectorLoopRegion()))) {
     for (VPRecipeBase &R : make_early_inc_range(*VPBB)) {
       if (!isa<VPWidenRecipe, VPWidenCastRecipe, VPReplicateRecipe,
-               VPWidenSelectRecipe, VPWidenMemoryInstructionRecipe>(&R))
+               VPWidenSelectRecipe, VPWidenMemoryRecipe>(&R))
         continue;
-      if (isa<VPWidenMemoryInstructionRecipe>(&R) &&
-          cast<VPWidenMemoryInstructionRecipe>(&R)->isStore())
+      if (isa<VPWidenStoreRecipe>(&R))
         continue;
 
       VPValue *ResultVPV = R.getVPSingleValue();
@@ -1048,10 +1047,9 @@ void VPlanTransforms::truncateToMinimalBitwidths(
         assert(cast<VPWidenRecipe>(&R)->getOpcode() == Instruction::ICmp &&
                "Only ICmps should not need extending the result.");
 
-      if (isa<VPWidenMemoryInstructionRecipe>(&R)) {
-        assert(!cast<VPWidenMemoryInstructionRecipe>(&R)->isStore() && "stores cannot be narrowed");
+      assert(!isa<VPWidenStoreRecipe>(&R) && "stores cannot be narrowed");
+      if (isa<VPWidenLoadRecipe>(&R))
         continue;
-      }
 
       // Shrink operands by introducing truncates as needed.
       unsigned StartIdx = isa<VPWidenSelectRecipe>(&R) ? 1 : 0;
@@ -1315,7 +1313,7 @@ void VPlanTransforms::addExplicitVectorLength(VPlan &Plan) {
       ConstantInt::getTrue(CanonicalIVPHI->getScalarType()->getContext());
   VPValue *VPTrueMask = Plan.getOrAddLiveIn(TrueMask);
   replaceHeaderPredicateWith(Plan, *VPTrueMask, [](VPUser &U, unsigned) {
-    return isa<VPWidenMemoryInstructionRecipe>(U);
+    return isa<VPWidenMemoryRecipe>(U);
   });
   // Now create the ExplicitVectorLengthPhi recipe in the main loop.
   auto *EVLPhi = new VPEVLBasedIVPHIRecipe(StartV, DebugLoc());
@@ -1371,8 +1369,7 @@ void VPlanTransforms::dropPoisonGeneratingRecipes(
       // instruction. Widen memory instructions involved in address computation
       // will lead to gather/scatter instructions, which don't need to be
       // handled.
-      if (isa<VPWidenMemoryInstructionRecipe>(CurRec) ||
-          isa<VPInterleaveRecipe>(CurRec) ||
+      if (isa<VPWidenMemoryRecipe>(CurRec) || isa<VPInterleaveRecipe>(CurRec) ||
           isa<VPScalarIVStepsRecipe>(CurRec) || isa<VPHeaderPHIRecipe>(CurRec))
         continue;
 
@@ -1420,7 +1417,7 @@ void VPlanTransforms::dropPoisonGeneratingRecipes(
   auto Iter = vp_depth_first_deep(Plan.getEntry());
   for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(Iter)) {
     for (VPRecipeBase &Recipe : *VPBB) {
-      if (auto *WidenRec = dyn_cast<VPWidenMemoryInstructionRecipe>(&Recipe)) {
+      if (auto *WidenRec = dyn_cast<VPWidenMemoryRecipe>(&Recipe)) {
         Instruction &UnderlyingInstr = WidenRec->getIngredient();
         VPRecipeBase *AddrDef = WidenRec->getAddr()->getDefiningRecipe();
         if (AddrDef && WidenRec->isConsecutive() &&
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 3f8d4f4fe7d64..0bbc7ffb4a2fe 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -36,7 +36,6 @@ class VPDef;
 class VPSlotTracker;
 class VPUser;
 class VPRecipeBase;
-class VPWidenMemoryInstructionRecipe;
 
 // This is the base class of the VPlan Def/Use graph, used for modeling the data
 // flow into, within and out of the VPlan. VPValues can stand for live-ins
@@ -51,7 +50,6 @@ class VPValue {
   friend class VPInterleavedAccessInfo;
   friend class VPSlotTracker;
   friend class VPRecipeBase;
-  friend class VPWidenMemoryInstructionRecipe;
 
   const unsigned char SubclassID; ///< Subclass identifier (for isa/dyn_cast).
 
@@ -358,7 +356,8 @@ class VPDef {
     VPWidenCanonicalIVSC,
     VPWidenCastSC,
     VPWidenGEPSC,
-    VPWidenMemoryInstructionSC,
+    VPWidenLoadSC,
+    VPWidenStoreSC,
     VPWidenSC,
     VPWidenSelectSC,
     VPBlendSC,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp b/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
index 12d37fa711db9..5587302207acd 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanVerifier.cpp
@@ -128,7 +128,7 @@ static bool verifyVPBasicBlock(const VPBasicBlock *VPBB,
       }
       return true;
     }
-    if (isa<VPWidenMemoryInstructionRecipe>(R))
+    if (isa<VPWidenMemoryRecipe>(R))
       VPWidenMemRecipe = R;
     return true;
   };
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll b/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll
index d5ace655fdcc1..c22613509be4f 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll
@@ -46,8 +46,8 @@ define void @vector_reverse_mask_v4i1(ptr noalias %a, ptr noalias %cond, i64 %N)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i64 -24
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP7]], i64 -56
 ; CHECK-NEXT:    [[REVERSE3:%.*]] = shufflevector <4 x i1> [[TMP5]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    [[REVERSE4:%.*]] = shufflevector <4 x i1> [[TMP6]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP8]], i32 8, <4 x i1> [[REVERSE3]], <4 x double> poison)
+; CHECK-NEXT:    [[REVERSE4:%.*]] = shufflevector <4 x i1> [[TMP6]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP9]], i32 8, <4 x i1> [[REVERSE4]], <4 x double> poison)
 ; CHECK-NEXT:    [[TMP10:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
 ; CHECK-NEXT:    [[TMP11:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD6]], <double 1.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>
diff --git a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
index eea2894f82794..aea72b7de5f42 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
@@ -1400,15 +1400,15 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX2-NEXT:    [[TMP30:%.*]] = getelementptr double, ptr [[TMP20]], i32 -12
 ; AVX2-NEXT:    [[TMP31:%.*]] = getelementptr double, ptr [[TMP30]], i32 -3
 ; AVX2-NEXT:    [[REVERSE12:%.*]] = shufflevector <4 x i1> [[TMP16]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:    [[REVERSE14:%.*]] = shufflevector <4 x i1> [[TMP17]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:    [[REVERSE17:%.*]] = shufflevector <4 x i1> [[TMP18]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:    [[REVERSE20:%.*]] = shufflevector <4 x i1> [[TMP19]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP25]], i32 8, <4 x i1> [[REVERSE12]], <4 x double> poison), !alias.scope !21
 ; AVX2-NEXT:    [[REVERSE13:%.*]] = shufflevector <4 x double> [[WIDE_MASKED_LOAD]], <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:    [[REVERSE14:%.*]] = shufflevector <4 x i1> [[TMP17]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP27]], i32 8, <4 x i1> [[REVERSE14]], <4 x double> poison), !alias.scope !21
 ; AVX2-NEXT:    [[REVERSE16:%.*]] = shufflevector <4 x double> [[WIDE_MASKED_LOAD15]], <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:    [[REVERSE17:%.*]] = shufflevector <4 x i1> [[TMP18]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD18:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP29]], i32 8, <4 x i1> [[REVERSE17]], <4 x double> poison), !alias.scope !21
 ; AVX2-NEXT:    [[REVERSE19:%.*]] = shufflevector <4 x double> [[WIDE_MASKED_LOAD18]], <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:    [[REVERSE20:%.*]] = shufflevector <4 x i1> [[TMP19]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD21:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP31]], i32 8, <4 x i1> [[REVERSE20]], <4 x double> poison), !alias.scope !21
 ; AVX2-NEXT:    [[REVERSE22:%.*]] = shufflevector <4 x double> [[WIDE_MASKED_LOAD21]], <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:    [[TMP32:%.*]] = fadd <4 x double> [[REVERSE13]], <double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01>
@@ -1524,15 +1524,15 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX512-NEXT:    [[TMP30:%.*]] = getelementptr double, ptr [[TMP20]], i32 -24
 ; AVX512-NEXT:    [[TMP31:%.*]] = getelementptr double, ptr [[TMP30]], i32 -7
 ; AVX512-NEXT:    [[REVERSE12:%.*]] = shufflevector <8 x i1> [[TMP16]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[REVERSE14:%.*]] = shufflevector <8 x i1> [[TMP17]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[REVERSE17:%.*]] = shufflevector <8 x i1> [[TMP18]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[REVERSE20:%.*]] = shufflevector <8 x i1> [[TMP19]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP25]], i32 8, <8 x i1> [[REVERSE12]], <8 x double> poison), !alias.scope !34
 ; AVX512-NEXT:    [[REVERSE13:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:    [[REVERSE14:%.*]] = shufflevector <8 x i1> [[TMP17]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP27]], i32 8, <8 x i1> [[REVERSE14]], <8 x double> poison), !alias.scope !34
 ; AVX512-NEXT:    [[REVERSE16:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD15]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:    [[REVERSE17:%.*]] = shufflevector <8 x i1> [[TMP18]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD18:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP29]], i32 8, <8 x i1> [[REVERSE17]], <8 x double> poison), !alias.scope !34
 ; AVX512-NEXT:    [[REVERSE19:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD18]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:    [[REVERSE20:%.*]] = shufflevector <8 x i1> [[TMP19]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD21:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP31]], i32 8, <8 x i1> [[REVERSE20]], <8 x double> poison), !alias.scope !34
 ; AVX512-NEXT:    [[REVERSE22:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD21]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[TMP32:%.*]] = fadd <8 x double> [[REVERSE13]], <double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01, double 5.000000e-01>
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp
index 777675b623f32..2b25c62ac2f65 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanHCFGTest.cpp
@@ -192,9 +192,9 @@ TEST_F(VPlanHCFGTest, testVPInstructionToVPRecipesInner) {
   auto Iter = VecBB->begin();
   EXPECT_NE(nullptr, dyn_cast<VPWidenPHIRecipe>(&*Iter++));
   EXPECT_NE(nullptr, dyn_cast<VPWidenGEPRecipe>(&*Iter++));
-  EXPECT_NE(nullptr, dyn_cast<VPWidenMemoryInstructionRecipe>(&*Iter++));
+  EXPECT_NE(nullptr, dyn_cast<VPWidenMemoryRecipe>(&*Iter++));
   EXPECT_NE(nullptr, dyn_cast<VPWidenRecipe>(&*Iter++));
-  EXPECT_NE(nullptr, dyn_cast<VPWidenMemoryInstructionRecipe>(&*Iter++));
+  EXPECT_NE(nullptr, dyn_cast<VPWidenMemoryRecipe>(&*Iter++));
   EXPECT_NE(nullptr, dyn_cast<VPWidenRecipe>(&*Iter++));
   EXPECT_NE(nullptr, dyn_cast<VPWidenRecipe>(&*Iter++));
   EXPECT_NE(nullptr, dyn_cast<VPInstruction>(&*Iter++));
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
index cb8737a9e64d2..64e9c06db3fe8 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
@@ -1029,7 +1029,7 @@ TEST(VPRecipeTest, CastVPBranchOnMaskRecipeToVPUser) {
   EXPECT_EQ(&Recipe, BaseR);
 }
 
-TEST(VPRecipeTest, CastVPWidenMemoryInstructionRecipeToVPUserAndVPDef) {
+TEST(VPRecipeTest, CastVPWidenMemoryRecipeToVPUserAndVPDef) {
   LLVMContext C;
 
   IntegerType *Int32 = IntegerType::get(C, 32);
@@ -1038,7 +1038,7 @@ TEST(VPRecipeTest, CastVPWidenMemoryInstructionRecipeToVPUserAndVPDef) {
       new LoadInst(Int32, UndefValue::get(Int32Ptr), "", false, Align(1));
   VPValue Addr;
   VPValue Mask;
-  VPWidenMemoryInstructionRecipe Recipe(*Load, &Addr, &Mask, true, false, {});
+  VPWidenLoadRecipe Recipe(*Load, &Addr, &Mask, true, false, {});
   EXPECT_TRUE(isa<VPUser>(&Recipe));
   VPRecipeBase *BaseR = &Recipe;
   EXPECT_TRUE(isa<VPUser>(BaseR));
@@ -1133,7 +1133,7 @@ TEST(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
         new LoadInst(Int32, UndefValue::get(Int32Ptr), "", false, Align(1));
     VPValue Addr;
     VPValue Mask;
-    VPWidenMemoryInstructionRecipe Recipe(*Load, &Addr, &Mask, true, false, {});
+    VPWidenLoadRecipe Recipe(*Load, &Addr, &Mask, true, false, {});
     EXPECT_FALSE(Recipe.mayHaveSideEffects());
     EXPECT_TRUE(Recipe.mayReadFromMemory());
     EXPECT_FALSE(Recipe.mayWriteToMemory());
@@ -1147,8 +1147,7 @@ TEST(VPRecipeTest, MayHaveSideEffectsAndMayReadWriteMemory) {
     VPValue Addr;
     VPValue Mask;
     VPValue StoredV;
-    VPWidenMemoryInstructionRecipe Recipe(*Store, &Addr, &StoredV, &Mask, false,
-                                          false, {});
+    VPWidenStoreRecipe Recipe(*Store, &Addr, &StoredV, &Mask, false, false, {});
     EXPECT_TRUE(Recipe.mayHaveSideEffects());
     EXPECT_FALSE(Recipe.mayReadFromMemory());
     EXPECT_TRUE(Recipe.mayWriteToMemory());

From f4737a2edd900df661750116821806bb45e4086a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
Date: Wed, 17 Apr 2024 12:24:32 +0200
Subject: [PATCH 720/886] update_test_checks: keep names stable with generated
 functions (#87988)

Collect the original check lines in a manner that is independent of
where the check lines appear in the file. This is so that we keep
FileCheck variable names stable even when --include-generated-funcs is
used.

Reported-by: Ruiling Song <ruiling.song@amd.com>
---
 .../Inputs/stable_ir_values_funcs.ll.expected | 10 ++--
 llvm/utils/UpdateTestChecks/common.py         | 59 +++++++++++--------
 2 files changed, 40 insertions(+), 29 deletions(-)

diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
index 1559319ac013a..86f929ffe36af 100644
--- a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/stable_ir_values_funcs.ll.expected
@@ -16,9 +16,9 @@ define i32 @func({i32, i32} %x, i32 %y) {
 
 ; CHECK-LABEL: define i32 @func(
 ; CHECK-SAME: { i32, i32 } [[X:%.*]], i32 [[Y:%.*]]) {
-; CHECK-NEXT:    [[X_I34:%.*]] = extractvalue { i32, i32 } [[X]], 0
-; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[Y]], 1
-; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[X_I34]], [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[TMP2]], 3
-; CHECK-NEXT:    ret i32 [[TMP3]]
+; CHECK-NEXT:    [[X_I33:%.*]] = extractvalue { i32, i32 } [[X]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[Y]], 1
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[X_I33]], [[TMP3]]
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i32 [[TMP1]], 3
+; CHECK-NEXT:    ret i32 [[TMP2]]
 ;
diff --git a/llvm/utils/UpdateTestChecks/common.py b/llvm/utils/UpdateTestChecks/common.py
index eed36a0cdd73f..15d3d5e527d61 100644
--- a/llvm/utils/UpdateTestChecks/common.py
+++ b/llvm/utils/UpdateTestChecks/common.py
@@ -430,36 +430,47 @@ def collect_original_check_lines(ti: TestInfo, prefix_set: set):
     result[func_name][prefix] is filled with a list of right-hand-sides of check
     lines.
     """
-    result = {}
+    result = collections.defaultdict(lambda: {})
 
+    current_prefix = None
     current_function = None
     for input_line_info in ti.ro_iterlines():
         input_line = input_line_info.line
-        if current_function is not None:
-            if input_line == "":
-                continue
-            if input_line.lstrip().startswith(";"):
-                m = CHECK_RE.match(input_line)
-                if (
-                    m is not None
-                    and m.group(1) in prefix_set
-                    and m.group(2) not in ["LABEL", "SAME"]
-                ):
-                    if m.group(1) not in current_function:
-                        current_function[m.group(1)] = []
-                    current_function[m.group(1)].append(input_line[m.end() :].strip())
-                continue
-            current_function = None
+        if input_line.lstrip().startswith(";"):
+            m = CHECK_RE.match(input_line)
+            if m is not None:
+                prefix = m.group(1)
+                check_kind = m.group(2)
+                line = input_line[m.end() :].strip()
+
+                if prefix != current_prefix:
+                    current_function = None
+                    current_prefix = None
+
+                if check_kind not in ["LABEL", "SAME"]:
+                    if current_function is not None:
+                        current_function.append(line)
+                    continue
 
-        m = IR_FUNCTION_RE.match(input_line)
-        if m is not None:
-            func_name = m.group(1)
-            if ti.args.function is not None and func_name != ti.args.function:
-                # When filtering on a specific function, skip all others.
-                continue
+                if check_kind == "SAME":
+                    continue
+
+                if check_kind == "LABEL":
+                    m = IR_FUNCTION_RE.match(line)
+                    if m is not None:
+                        func_name = m.group(1)
+                        if (
+                            ti.args.function is not None
+                            and func_name != ti.args.function
+                        ):
+                            # When filtering on a specific function, skip all others.
+                            continue
+
+                        current_prefix = prefix
+                        current_function = result[func_name][prefix] = []
+                        continue
 
-            assert func_name not in result
-            current_function = result[func_name] = {}
+        current_function = None
 
     return result
 

From 3eb0ba34b0a2a29c2f34ead2b84fdf9b62cb29c1 Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Wed, 17 Apr 2024 11:28:30 +0100
Subject: [PATCH 721/886] [MLIR][Flang][OpenMP] Make omp.simdloop into a loop
 wrapper (#87365)

This patch updates the definition of `omp.simdloop` to enforce the
restrictions of a wrapper operation. It has been renamed to `omp.simd`,
to better reflect the naming used in the spec. All uses of "simdloop" in
function names have been updated accordingly.

Some changes to Flang lowering and OpenMP to LLVM IR translation are
introduced to prevent the introduction of compilation/test failures. The
eventual long term solution might be different.
---
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 167 +++++++++-----
 .../Fir/convert-to-llvm-openmp-and-fir.fir    | 101 +++++----
 flang/test/Lower/OpenMP/FIR/if-clause.f90     |  23 +-
 flang/test/Lower/OpenMP/FIR/loop-combined.f90 |   2 +-
 .../OpenMP/FIR/parallel-private-clause.f90    |   3 +-
 flang/test/Lower/OpenMP/FIR/simd.f90          | 109 +++++----
 flang/test/Lower/OpenMP/if-clause.f90         |  23 +-
 flang/test/Lower/OpenMP/loop-combined.f90     |   2 +-
 .../Lower/OpenMP/parallel-private-clause.f90  |   3 +-
 flang/test/Lower/OpenMP/simd.f90              | 123 +++++-----
 .../Frontend/OpenMPIRBuilderTest.cpp          |   2 +-
 .../Dialect/OpenMP/OpenMPClauseOperands.h     |   9 +-
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |  58 ++---
 .../Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp  |  17 +-
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  |  45 ++--
 .../OpenMP/OpenMPToLLVMIRTranslation.cpp      |  41 ++--
 .../OpenMPToLLVM/convert-to-llvmir.mlir       |  31 +--
 mlir/test/Dialect/OpenMP/invalid.mlir         | 163 +++++++------
 mlir/test/Dialect/OpenMP/ops.mlir             | 214 +++++++++---------
 mlir/test/Target/LLVMIR/openmp-llvm.mlir      | 157 +++++++------
 20 files changed, 701 insertions(+), 592 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 9b99752236662..c31d63625dbb1 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -502,8 +502,10 @@ struct OpWithBodyGenInfo {
 
   OpWithBodyGenInfo(Fortran::lower::AbstractConverter &converter,
                     Fortran::semantics::SemanticsContext &semaCtx,
-                    mlir::Location loc, Fortran::lower::pft::Evaluation &eval)
-      : converter(converter), semaCtx(semaCtx), loc(loc), eval(eval) {}
+                    mlir::Location loc, Fortran::lower::pft::Evaluation &eval,
+                    llvm::omp::Directive dir)
+      : converter(converter), semaCtx(semaCtx), loc(loc), eval(eval), dir(dir) {
+  }
 
   OpWithBodyGenInfo &setGenNested(bool value) {
     genNested = value;
@@ -546,6 +548,8 @@ struct OpWithBodyGenInfo {
   mlir::Location loc;
   /// [in] current PFT node/evaluation.
   Fortran::lower::pft::Evaluation &eval;
+  /// [in] leaf directive for which to generate the op body.
+  llvm::omp::Directive dir;
   /// [in] whether to generate FIR for nested evaluations
   bool genNested = true;
   /// [in] is this an outer operation - prevents privatization.
@@ -568,8 +572,7 @@ struct OpWithBodyGenInfo {
 ///
 /// \param [in]   op - the operation the body belongs to.
 /// \param [in] info - options controlling code-gen for the construction.
-template <typename Op>
-static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
+static void createBodyOfOp(mlir::Operation &op, OpWithBodyGenInfo &info) {
   fir::FirOpBuilder &firOpBuilder = info.converter.getFirOpBuilder();
 
   auto insertMarker = [](fir::FirOpBuilder &builder) {
@@ -585,10 +588,10 @@ static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
   auto regionArgs =
       [&]() -> llvm::SmallVector<const Fortran::semantics::Symbol *> {
     if (info.genRegionEntryCB != nullptr) {
-      return info.genRegionEntryCB(op);
+      return info.genRegionEntryCB(&op);
     }
 
-    firOpBuilder.createBlock(&op.getRegion());
+    firOpBuilder.createBlock(&op.getRegion(0));
     return {};
   }();
   // Mark the earliest insertion point.
@@ -603,8 +606,8 @@ static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
 
   // Start with privatization, so that the lowering of the nested
   // code will use the right symbols.
-  constexpr bool isLoop = std::is_same_v<Op, mlir::omp::WsloopOp> ||
-                          std::is_same_v<Op, mlir::omp::SimdLoopOp>;
+  bool isLoop = llvm::omp::getDirectiveAssociation(info.dir) ==
+                llvm::omp::Association::Loop;
   bool privatize = info.clauses && !info.outerCombined;
 
   firOpBuilder.setInsertionPoint(marker);
@@ -616,7 +619,7 @@ static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
     }
   }
 
-  if constexpr (std::is_same_v<Op, mlir::omp::ParallelOp>) {
+  if (info.dir == llvm::omp::Directive::OMPD_parallel) {
     threadPrivatizeVars(info.converter, info.eval);
     if (info.clauses) {
       firOpBuilder.setInsertionPoint(marker);
@@ -630,9 +633,9 @@ static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
     // a lot of complications for our approach if the terminator generation
     // is delayed past this point. Insert a temporary terminator here, then
     // delete it.
-    firOpBuilder.setInsertionPointToEnd(&op.getRegion().back());
-    auto *temp = Fortran::lower::genOpenMPTerminator(
-        firOpBuilder, op.getOperation(), info.loc);
+    firOpBuilder.setInsertionPointToEnd(&op.getRegion(0).back());
+    auto *temp =
+        Fortran::lower::genOpenMPTerminator(firOpBuilder, &op, info.loc);
     firOpBuilder.setInsertionPointAfter(marker);
     genNestedEvaluations(info.converter, info.eval);
     temp->erase();
@@ -674,23 +677,36 @@ static void createBodyOfOp(Op &op, OpWithBodyGenInfo &info) {
     return exit;
   };
 
-  if (auto *exitBlock = getUniqueExit(op.getRegion())) {
+  if (auto *exitBlock = getUniqueExit(op.getRegion(0))) {
     firOpBuilder.setInsertionPointToEnd(exitBlock);
-    auto *term = Fortran::lower::genOpenMPTerminator(
-        firOpBuilder, op.getOperation(), info.loc);
+    auto *term =
+        Fortran::lower::genOpenMPTerminator(firOpBuilder, &op, info.loc);
     // Only insert lastprivate code when there actually is an exit block.
     // Such a block may not exist if the nested code produced an infinite
     // loop (this may not make sense in production code, but a user could
     // write that and we should handle it).
     firOpBuilder.setInsertionPoint(term);
     if (privatize) {
+      // DataSharingProcessor::processStep2() may create operations before/after
+      // the one passed as argument. We need to treat loop wrappers and their
+      // nested loop as a unit, so we need to pass the top level wrapper (if
+      // present). Otherwise, these operations will be inserted within a
+      // wrapper region.
+      mlir::Operation *privatizationTopLevelOp = &op;
+      if (auto loopNest = llvm::dyn_cast<mlir::omp::LoopNestOp>(op)) {
+        llvm::SmallVector<mlir::omp::LoopWrapperInterface> wrappers;
+        loopNest.gatherWrappers(wrappers);
+        if (!wrappers.empty())
+          privatizationTopLevelOp = &*wrappers.back();
+      }
+
       if (!info.dsp) {
         assert(tempDsp.has_value());
-        tempDsp->processStep2(op, isLoop);
+        tempDsp->processStep2(privatizationTopLevelOp, isLoop);
       } else {
         if (isLoop && regionArgs.size() > 0)
           info.dsp->setLoopIV(info.converter.getSymbolAddress(*regionArgs[0]));
-        info.dsp->processStep2(op, isLoop);
+        info.dsp->processStep2(privatizationTopLevelOp, isLoop);
       }
     }
   }
@@ -921,7 +937,7 @@ template <typename OpTy, typename... Args>
 static OpTy genOpWithBody(OpWithBodyGenInfo &info, Args &&...args) {
   auto op = info.converter.getFirOpBuilder().create<OpTy>(
       info.loc, std::forward<Args>(args)...);
-  createBodyOfOp<OpTy>(op, info);
+  createBodyOfOp(*op, info);
   return op;
 }
 
@@ -954,6 +970,18 @@ static void genFlushClauses(
     TODO(converter.getCurrentLocation(), "Handle OmpMemoryOrderClause");
 }
 
+static void genLoopNestClauses(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
+    mlir::omp::LoopNestClauseOps &clauseOps,
+    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processCollapse(loc, eval, clauseOps, iv);
+  clauseOps.loopInclusiveAttr = converter.getFirOpBuilder().getUnitAttr();
+}
+
 static void
 genOrderedRegionClauses(Fortran::lower::AbstractConverter &converter,
                         Fortran::semantics::SemanticsContext &semaCtx,
@@ -1002,21 +1030,16 @@ static void genSectionsClauses(Fortran::lower::AbstractConverter &converter,
   }
 }
 
-static void genSimdLoopClauses(
-    Fortran::lower::AbstractConverter &converter,
-    Fortran::semantics::SemanticsContext &semaCtx,
-    Fortran::lower::StatementContext &stmtCtx,
-    Fortran::lower::pft::Evaluation &eval,
-    const Fortran::parser::OmpClauseList &clauses, mlir::Location loc,
-    mlir::omp::SimdLoopClauseOps &clauseOps,
-    llvm::SmallVectorImpl<const Fortran::semantics::Symbol *> &iv) {
+static void genSimdClauses(Fortran::lower::AbstractConverter &converter,
+                           Fortran::semantics::SemanticsContext &semaCtx,
+                           const Fortran::parser::OmpClauseList &clauses,
+                           mlir::Location loc,
+                           mlir::omp::SimdClauseOps &clauseOps) {
   ClauseProcessor cp(converter, semaCtx, clauses);
-  cp.processCollapse(loc, eval, clauseOps, iv);
   cp.processIf(llvm::omp::Directive::OMPD_simd, clauseOps);
   cp.processReduction(loc, clauseOps);
   cp.processSafelen(clauseOps);
   cp.processSimdlen(clauseOps);
-  clauseOps.loopInclusiveAttr = converter.getFirOpBuilder().getUnitAttr();
   // TODO Support delayed privatization.
 
   cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
@@ -1260,7 +1283,9 @@ genCriticalOp(Fortran::lower::AbstractConverter &converter,
   }
 
   return genOpWithBody<mlir::omp::CriticalOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested),
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_critical)
+          .setGenNested(genNested),
       nameAttr);
 }
 
@@ -1295,7 +1320,9 @@ genMasterOp(Fortran::lower::AbstractConverter &converter,
             Fortran::lower::pft::Evaluation &eval, bool genNested,
             mlir::Location loc) {
   return genOpWithBody<mlir::omp::MasterOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested));
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_master)
+          .setGenNested(genNested));
 }
 
 static mlir::omp::OrderedOp
@@ -1317,7 +1344,9 @@ genOrderedRegionOp(Fortran::lower::AbstractConverter &converter,
   genOrderedRegionClauses(converter, semaCtx, clauseList, loc, clauseOps);
 
   return genOpWithBody<mlir::omp::OrderedRegionOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(genNested),
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_ordered)
+          .setGenNested(genNested),
       clauseOps);
 }
 
@@ -1345,7 +1374,8 @@ genParallelOp(Fortran::lower::AbstractConverter &converter,
   };
 
   OpWithBodyGenInfo genInfo =
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_parallel)
           .setGenNested(genNested)
           .setOuterCombined(outerCombined)
           .setClauses(&clauseList)
@@ -1408,7 +1438,8 @@ genSectionOp(Fortran::lower::AbstractConverter &converter,
   // Currently only private/firstprivate clause is handled, and
   // all privatization is done within `omp.section` operations.
   return genOpWithBody<mlir::omp::SectionOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_section)
           .setGenNested(genNested)
           .setClauses(&clauseList));
 }
@@ -1419,23 +1450,39 @@ genSectionsOp(Fortran::lower::AbstractConverter &converter,
               Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
               const mlir::omp::SectionsClauseOps &clauseOps) {
   return genOpWithBody<mlir::omp::SectionsOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval).setGenNested(false),
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_sections)
+          .setGenNested(false),
       clauseOps);
 }
 
-static mlir::omp::SimdLoopOp
-genSimdLoopOp(Fortran::lower::AbstractConverter &converter,
-              Fortran::semantics::SemanticsContext &semaCtx,
-              Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
-              const Fortran::parser::OmpClauseList &clauseList) {
+static mlir::omp::SimdOp
+genSimdOp(Fortran::lower::AbstractConverter &converter,
+          Fortran::semantics::SemanticsContext &semaCtx,
+          Fortran::lower::pft::Evaluation &eval, mlir::Location loc,
+          const Fortran::parser::OmpClauseList &clauseList) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
   DataSharingProcessor dsp(converter, semaCtx, clauseList, eval);
   dsp.processStep1();
 
   Fortran::lower::StatementContext stmtCtx;
-  mlir::omp::SimdLoopClauseOps clauseOps;
+  mlir::omp::LoopNestClauseOps loopClauseOps;
+  mlir::omp::SimdClauseOps simdClauseOps;
   llvm::SmallVector<const Fortran::semantics::Symbol *> iv;
-  genSimdLoopClauses(converter, semaCtx, stmtCtx, eval, clauseList, loc,
-                     clauseOps, iv);
+  genLoopNestClauses(converter, semaCtx, eval, clauseList, loc, loopClauseOps,
+                     iv);
+  genSimdClauses(converter, semaCtx, clauseList, loc, simdClauseOps);
+
+  // Create omp.simd wrapper.
+  auto simdOp = firOpBuilder.create<mlir::omp::SimdOp>(loc, simdClauseOps);
+
+  // TODO: Add reduction-related arguments to the wrapper's entry block.
+  firOpBuilder.createBlock(&simdOp.getRegion());
+  firOpBuilder.setInsertionPoint(
+      Fortran::lower::genOpenMPTerminator(firOpBuilder, simdOp, loc));
+
+  // Create nested omp.loop_nest and fill body with loop contents.
+  auto loopOp = firOpBuilder.create<mlir::omp::LoopNestOp>(loc, loopClauseOps);
 
   auto *nestedEval =
       getCollapsedLoopEval(eval, Fortran::lower::getCollapseValue(clauseList));
@@ -1444,12 +1491,14 @@ genSimdLoopOp(Fortran::lower::AbstractConverter &converter,
     return genLoopVars(op, converter, loc, iv);
   };
 
-  return genOpWithBody<mlir::omp::SimdLoopOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
-          .setClauses(&clauseList)
-          .setDataSharingProcessor(&dsp)
-          .setGenRegionEntryCb(ivCallback),
-      clauseOps);
+  createBodyOfOp(*loopOp,
+                 OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval,
+                                   llvm::omp::Directive::OMPD_simd)
+                     .setClauses(&clauseList)
+                     .setDataSharingProcessor(&dsp)
+                     .setGenRegionEntryCb(ivCallback));
+
+  return simdOp;
 }
 
 static mlir::omp::SingleOp
@@ -1464,7 +1513,8 @@ genSingleOp(Fortran::lower::AbstractConverter &converter,
                    clauseOps);
 
   return genOpWithBody<mlir::omp::SingleOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_single)
           .setGenNested(genNested)
           .setClauses(&beginClauseList),
       clauseOps);
@@ -1645,7 +1695,8 @@ genTaskOp(Fortran::lower::AbstractConverter &converter,
   genTaskClauses(converter, semaCtx, stmtCtx, clauseList, loc, clauseOps);
 
   return genOpWithBody<mlir::omp::TaskOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_task)
           .setGenNested(genNested)
           .setClauses(&clauseList),
       clauseOps);
@@ -1661,7 +1712,8 @@ genTaskgroupOp(Fortran::lower::AbstractConverter &converter,
   genTaskgroupClauses(converter, semaCtx, clauseList, loc, clauseOps);
 
   return genOpWithBody<mlir::omp::TaskgroupOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_taskgroup)
           .setGenNested(genNested)
           .setClauses(&clauseList),
       clauseOps);
@@ -1704,7 +1756,8 @@ genTeamsOp(Fortran::lower::AbstractConverter &converter,
   genTeamsClauses(converter, semaCtx, stmtCtx, clauseList, loc, clauseOps);
 
   return genOpWithBody<mlir::omp::TeamsOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, eval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, eval,
+                        llvm::omp::Directive::OMPD_teams)
           .setGenNested(genNested)
           .setOuterCombined(outerCombined)
           .setClauses(&clauseList),
@@ -1738,7 +1791,8 @@ genWsloopOp(Fortran::lower::AbstractConverter &converter,
   };
 
   return genOpWithBody<mlir::omp::WsloopOp>(
-      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval)
+      OpWithBodyGenInfo(converter, semaCtx, loc, *nestedEval,
+                        llvm::omp::Directive::OMPD_do)
           .setClauses(&beginClauseList)
           .setDataSharingProcessor(&dsp)
           .setReductions(&reductionSyms, &reductionTypes)
@@ -2253,7 +2307,7 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
                        endClauseList, currentLocation);
   } else if (llvm::omp::allSimdSet.test(ompDirective)) {
     // 2.9.3.1 SIMD construct
-    genSimdLoopOp(converter, semaCtx, eval, currentLocation, beginClauseList);
+    genSimdOp(converter, semaCtx, eval, currentLocation, beginClauseList);
   } else {
     genWsloopOp(converter, semaCtx, eval, currentLocation, beginClauseList,
                 endClauseList);
@@ -2341,10 +2395,9 @@ mlir::Operation *Fortran::lower::genOpenMPTerminator(fir::FirOpBuilder &builder,
                                                      mlir::Operation *op,
                                                      mlir::Location loc) {
   if (mlir::isa<mlir::omp::WsloopOp, mlir::omp::DeclareReductionOp,
-                mlir::omp::AtomicUpdateOp, mlir::omp::SimdLoopOp>(op))
+                mlir::omp::AtomicUpdateOp, mlir::omp::LoopNestOp>(op))
     return builder.create<mlir::omp::YieldOp>(loc);
-  else
-    return builder.create<mlir::omp::TerminatorOp>(loc);
+  return builder.create<mlir::omp::TerminatorOp>(loc);
 }
 
 void Fortran::lower::genOpenMPConstruct(
diff --git a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
index 92628af37085a..fa7979e8875af 100644
--- a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
+++ b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
@@ -180,14 +180,16 @@ func.func @_QPsimd1(%arg0: !fir.ref<i32> {fir.bindc_name = "n"}, %arg1: !fir.ref
   omp.parallel  {
     %1 = fir.alloca i32 {adapt.valuebyref, pinned}
     %2 = fir.load %arg0 : !fir.ref<i32>
-    omp.simdloop for (%arg2) : i32 = (%c1_i32) to (%2) step (%c1_i32)  {
-      fir.store %arg2 to %1 : !fir.ref<i32>
-      %3 = fir.load %1 : !fir.ref<i32>
-      %4 = fir.convert %3 : (i32) -> i64
-      %5 = arith.subi %4, %c1_i64 : i64
-      %6 = fir.coordinate_of %arg1, %5 : (!fir.ref<!fir.array<?xi32>>, i64) -> !fir.ref<i32>
-      fir.store %3 to %6 : !fir.ref<i32>
-      omp.yield
+    omp.simd {
+      omp.loop_nest (%arg2) : i32 = (%c1_i32) to (%2) step (%c1_i32) {
+        fir.store %arg2 to %1 : !fir.ref<i32>
+        %3 = fir.load %1 : !fir.ref<i32>
+        %4 = fir.convert %3 : (i32) -> i64
+        %5 = arith.subi %4, %c1_i64 : i64
+        %6 = fir.coordinate_of %arg1, %5 : (!fir.ref<!fir.array<?xi32>>, i64) -> !fir.ref<i32>
+        fir.store %3 to %6 : !fir.ref<i32>
+        omp.yield
+      }
     }
     omp.terminator
   }
@@ -202,8 +204,8 @@ func.func @_QPsimd1(%arg0: !fir.ref<i32> {fir.bindc_name = "n"}, %arg1: !fir.ref
 // CHECK:      %[[ONE_3:.*]] = llvm.mlir.constant(1 : i64) : i64
 // CHECK:      %[[I_VAR:.*]] = llvm.alloca %[[ONE_3]] x i32 {pinned} : (i64) -> !llvm.ptr
 // CHECK:      %[[N:.*]] = llvm.load %[[N_REF]] : !llvm.ptr -> i32
-// CHECK: omp.simdloop
-// CHECK-SAME: (%[[I:.*]]) : i32 = (%[[ONE_2]]) to (%[[N]]) step (%[[ONE_2]]) {
+// CHECK: omp.simd {
+// CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[ONE_2]]) to (%[[N]]) step (%[[ONE_2]]) {
 // CHECK:   llvm.store %[[I]], %[[I_VAR]] : i32, !llvm.ptr
 // CHECK:   %[[I1:.*]] = llvm.load %[[I_VAR]] : !llvm.ptr -> i32
 // CHECK:   %[[I1_EXT:.*]] = llvm.sext %[[I1]] : i32 to i64
@@ -212,6 +214,7 @@ func.func @_QPsimd1(%arg0: !fir.ref<i32> {fir.bindc_name = "n"}, %arg1: !fir.ref
 // CHECK:   llvm.store %[[I1]], %[[ARR_I_REF]] : i32, !llvm.ptr
 // CHECK: omp.yield
 // CHECK: }
+// CHECK: }
 // CHECK: omp.terminator
 // CHECK: }
 // CHECK: llvm.return
@@ -471,55 +474,59 @@ func.func @_QPomp_target() {
 
 // -----
 
-func.func @_QPsimdloop_with_nested_loop() {
+func.func @_QPsimd_with_nested_loop() {
   %0 = fir.alloca i32 {adapt.valuebyref}
-  %1 = fir.alloca !fir.array<10xi32> {bindc_name = "a", uniq_name = "_QFsimdloop_with_nested_loopEa"}
-  %2 = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFsimdloop_with_nested_loopEi"}
-  %3 = fir.alloca i32 {bindc_name = "j", uniq_name = "_QFsimdloop_with_nested_loopEj"}
+  %1 = fir.alloca !fir.array<10xi32> {bindc_name = "a", uniq_name = "_QFsimd_with_nested_loopEa"}
+  %2 = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFsimd_with_nested_loopEi"}
+  %3 = fir.alloca i32 {bindc_name = "j", uniq_name = "_QFsimd_with_nested_loopEj"}
   %c1_i32 = arith.constant 1 : i32
   %c10_i32 = arith.constant 10 : i32
   %c1_i32_0 = arith.constant 1 : i32
-  omp.simdloop   for  (%arg0) : i32 = (%c1_i32) to (%c10_i32) inclusive step (%c1_i32_0) {
-    fir.store %arg0 to %0 : !fir.ref<i32>
-    %c1_i32_1 = arith.constant 1 : i32
-    %4 = fir.convert %c1_i32_1 : (i32) -> index
-    %c10_i32_2 = arith.constant 10 : i32
-    %5 = fir.convert %c10_i32_2 : (i32) -> index
-    %c1 = arith.constant 1 : index
-    %6 = fir.do_loop %arg1 = %4 to %5 step %c1 -> index {
-      %8 = fir.convert %arg1 : (index) -> i32
-      fir.store %8 to %3 : !fir.ref<i32>
-      %9 = fir.load %0 : !fir.ref<i32>
-      %10 = fir.load %0 : !fir.ref<i32>
-      %11 = fir.convert %10 : (i32) -> i64
-      %c1_i64 = arith.constant 1 : i64
-      %12 = arith.subi %11, %c1_i64 : i64
-      %13 = fir.coordinate_of %1, %12 : (!fir.ref<!fir.array<10xi32>>, i64) -> !fir.ref<i32>
-      fir.store %9 to %13 : !fir.ref<i32>
-      %14 = arith.addi %arg1, %c1 : index
-      fir.result %14 : index
+  omp.simd {
+    omp.loop_nest (%arg0) : i32 = (%c1_i32) to (%c10_i32) inclusive step (%c1_i32_0) {
+      fir.store %arg0 to %0 : !fir.ref<i32>
+      %c1_i32_1 = arith.constant 1 : i32
+      %4 = fir.convert %c1_i32_1 : (i32) -> index
+      %c10_i32_2 = arith.constant 10 : i32
+      %5 = fir.convert %c10_i32_2 : (i32) -> index
+      %c1 = arith.constant 1 : index
+      %6 = fir.do_loop %arg1 = %4 to %5 step %c1 -> index {
+        %8 = fir.convert %arg1 : (index) -> i32
+        fir.store %8 to %3 : !fir.ref<i32>
+        %9 = fir.load %0 : !fir.ref<i32>
+        %10 = fir.load %0 : !fir.ref<i32>
+        %11 = fir.convert %10 : (i32) -> i64
+        %c1_i64 = arith.constant 1 : i64
+        %12 = arith.subi %11, %c1_i64 : i64
+        %13 = fir.coordinate_of %1, %12 : (!fir.ref<!fir.array<10xi32>>, i64) -> !fir.ref<i32>
+        fir.store %9 to %13 : !fir.ref<i32>
+        %14 = arith.addi %arg1, %c1 : index
+        fir.result %14 : index
+      }
+      %7 = fir.convert %6 : (index) -> i32
+      fir.store %7 to %3 : !fir.ref<i32>
+      omp.yield
     }
-    %7 = fir.convert %6 : (index) -> i32
-    fir.store %7 to %3 : !fir.ref<i32>
-    omp.yield
   }
   return
 }
 
-// CHECK-LABEL:   llvm.func @_QPsimdloop_with_nested_loop() {
+// CHECK-LABEL:   llvm.func @_QPsimd_with_nested_loop() {
 // CHECK:           %[[LOWER:.*]] = llvm.mlir.constant(1 : i32) : i32
 // CHECK:           %[[UPPER:.*]] = llvm.mlir.constant(10 : i32) : i32
 // CHECK:           %[[STEP:.*]] = llvm.mlir.constant(1 : i32) : i32
-// CHECK:           omp.simdloop   for  (%[[CNT:.*]]) : i32 = (%[[LOWER]]) to (%[[UPPER]]) inclusive step (%[[STEP]]) {
-// CHECK:             llvm.br ^bb1(%[[VAL_1:.*]], %[[VAL_2:.*]] : i64, i64)
-// CHECK:           ^bb1(%[[VAL_3:.*]]: i64, %[[VAL_4:.*]]: i64):
-// CHECK:             %[[VAL_5:.*]] = llvm.mlir.constant(0 : index) : i64
-// CHECK:             %[[VAL_6:.*]] = llvm.icmp "sgt" %[[VAL_4]], %[[VAL_5]] : i64
-// CHECK:             llvm.cond_br %[[VAL_6]], ^bb2, ^bb3
-// CHECK:           ^bb2:
-// CHECK:             llvm.br ^bb1(%[[VAL_7:.*]], %[[VAL_8:.*]] : i64, i64)
-// CHECK:           ^bb3:
-// CHECK:             omp.yield
+// CHECK:           omp.simd {
+// CHECK-NEXT:        omp.loop_nest (%[[CNT:.*]]) : i32 = (%[[LOWER]]) to (%[[UPPER]]) inclusive step (%[[STEP]]) {
+// CHECK:               llvm.br ^bb1(%[[VAL_1:.*]], %[[VAL_2:.*]] : i64, i64)
+// CHECK:             ^bb1(%[[VAL_3:.*]]: i64, %[[VAL_4:.*]]: i64):
+// CHECK:               %[[VAL_5:.*]] = llvm.mlir.constant(0 : index) : i64
+// CHECK:               %[[VAL_6:.*]] = llvm.icmp "sgt" %[[VAL_4]], %[[VAL_5]] : i64
+// CHECK:               llvm.cond_br %[[VAL_6]], ^bb2, ^bb3
+// CHECK:             ^bb2:
+// CHECK:               llvm.br ^bb1(%[[VAL_7:.*]], %[[VAL_8:.*]] : i64, i64)
+// CHECK:             ^bb3:
+// CHECK:               omp.yield
+// CHECK:             }
 // CHECK:           }
 // CHECK:           llvm.return
 // CHECK:         }
diff --git a/flang/test/Lower/OpenMP/FIR/if-clause.f90 b/flang/test/Lower/OpenMP/FIR/if-clause.f90
index a1235be8e61ea..f686b9708fc54 100644
--- a/flang/test/Lower/OpenMP/FIR/if-clause.f90
+++ b/flang/test/Lower/OpenMP/FIR/if-clause.f90
@@ -116,7 +116,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-SAME: if({{.*}})
   ! CHECK:      omp.wsloop
@@ -124,7 +124,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-SAME: if({{.*}})
   ! CHECK:      omp.wsloop
@@ -134,7 +134,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
@@ -147,7 +147,7 @@ program main
   ! ----------------------------------------------------------------------------
   ! SIMD
   ! ----------------------------------------------------------------------------
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp simd
@@ -155,14 +155,14 @@ program main
   end do
   !$omp end simd
 
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp simd if(.true.)
   do i = 1, 10
   end do
   !$omp end simd
 
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp simd if(simd: .true.)
   do i = 1, 10
@@ -281,7 +281,6 @@ program main
   end do
   !$omp end target parallel do
 
-  
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
@@ -360,7 +359,7 @@ program main
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp target simd
@@ -370,7 +369,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(.true.)
   do i = 1, 10
@@ -379,7 +378,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(target: .true.) if(simd: .false.)
   do i = 1, 10
@@ -388,7 +387,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp target simd if(target: .true.)
@@ -399,7 +398,7 @@ program main
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(simd: .true.)
   do i = 1, 10
diff --git a/flang/test/Lower/OpenMP/FIR/loop-combined.f90 b/flang/test/Lower/OpenMP/FIR/loop-combined.f90
index a6cec1beb49c8..6c6618dc9fb57 100644
--- a/flang/test/Lower/OpenMP/FIR/loop-combined.f90
+++ b/flang/test/Lower/OpenMP/FIR/loop-combined.f90
@@ -75,7 +75,7 @@ program main
   ! TARGET SIMD
   ! ----------------------------------------------------------------------------
   ! CHECK: omp.target
-  ! CHECK: omp.simdloop
+  ! CHECK: omp.simd
   !$omp target simd
   do i = 1, 10
   end do
diff --git a/flang/test/Lower/OpenMP/FIR/parallel-private-clause.f90 b/flang/test/Lower/OpenMP/FIR/parallel-private-clause.f90
index 8f5d280943cc2..8b75ecbaae8c7 100644
--- a/flang/test/Lower/OpenMP/FIR/parallel-private-clause.f90
+++ b/flang/test/Lower/OpenMP/FIR/parallel-private-clause.f90
@@ -361,7 +361,8 @@ subroutine simd_loop_1
   ! FIRDialect:     %[[UB:.*]] = arith.constant 9 : i32
   ! FIRDialect:     %[[STEP:.*]] = arith.constant 1 : i32
 
-  ! FIRDialect: omp.simdloop for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
+  ! FIRDialect: omp.simd {
+  ! FIRDialect-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   !$OMP SIMD PRIVATE(r)
   do i=1, 9
   ! FIRDialect:     fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
diff --git a/flang/test/Lower/OpenMP/FIR/simd.f90 b/flang/test/Lower/OpenMP/FIR/simd.f90
index c8c2022d693d4..db7d30295c45d 100644
--- a/flang/test/Lower/OpenMP/FIR/simd.f90
+++ b/flang/test/Lower/OpenMP/FIR/simd.f90
@@ -2,32 +2,34 @@
 
 ! RUN: bbc -fopenmp -emit-fir -hlfir=false %s -o - | FileCheck %s
 
-!CHECK-LABEL: func @_QPsimdloop()
-subroutine simdloop
-integer :: i
+!CHECK-LABEL: func @_QPsimd()
+subroutine simd
+  integer :: i
   !$OMP SIMD
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK-NEXT: %[[UB:.*]] = arith.constant 9 : i32
   ! CHECK-NEXT: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK-NEXT: omp.simdloop for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
+  ! CHECK-NEXT: omp.simd {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i=1, 9
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
     ! CHECK: fir.call @_FortranAioOutputInteger32({{.*}}, %[[LD]]) {{.*}}: (!fir.ref<i8>, i32) -> i1
     print*, i
   end do
-  !$OMP END SIMD 
+  !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_if_clause
-subroutine simdloop_with_if_clause(n, threshold)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_if_clause
+subroutine simd_with_if_clause(n, threshold)
+  integer :: i, n, threshold
   !$OMP SIMD IF( n .GE. threshold )
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
   ! CHECK: %[[COND:.*]] = arith.cmpi sge
-  ! CHECK: omp.simdloop if(%[[COND:.*]]) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd if(%[[COND:.*]]) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -37,14 +39,15 @@ subroutine simdloop_with_if_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause
-subroutine simdloop_with_simdlen_clause(n, threshold)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause
+subroutine simd_with_simdlen_clause(n, threshold)
+  integer :: i, n, threshold
   !$OMP SIMD SIMDLEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -54,15 +57,16 @@ subroutine simdloop_with_simdlen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause_from_param
-subroutine simdloop_with_simdlen_clause_from_param(n, threshold)
-integer :: i, n, threshold
-integer, parameter :: simdlen = 2;
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause_from_param
+subroutine simd_with_simdlen_clause_from_param(n, threshold)
+  integer :: i, n, threshold
+  integer, parameter :: simdlen = 2;
   !$OMP SIMD SIMDLEN(simdlen)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -72,15 +76,16 @@ subroutine simdloop_with_simdlen_clause_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause_from_expr_from_param
-subroutine simdloop_with_simdlen_clause_from_expr_from_param(n, threshold)
-integer :: i, n, threshold
-integer, parameter :: simdlen = 2;
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause_from_expr_from_param
+subroutine simd_with_simdlen_clause_from_expr_from_param(n, threshold)
+  integer :: i, n, threshold
+  integer, parameter :: simdlen = 2;
   !$OMP SIMD SIMDLEN(simdlen*2 + 2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(6) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(6) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -90,14 +95,15 @@ subroutine simdloop_with_simdlen_clause_from_expr_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_safelen_clause
-subroutine simdloop_with_safelen_clause(n, threshold)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_safelen_clause
+subroutine simd_with_safelen_clause(n, threshold)
+  integer :: i, n, threshold
   !$OMP SIMD SAFELEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop safelen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd safelen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -107,15 +113,16 @@ subroutine simdloop_with_safelen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_safelen_clause_from_expr_from_param
-subroutine simdloop_with_safelen_clause_from_expr_from_param(n, threshold)
-integer :: i, n, threshold
-integer, parameter :: safelen = 2;
+!CHECK-LABEL: func @_QPsimd_with_safelen_clause_from_expr_from_param
+subroutine simd_with_safelen_clause_from_expr_from_param(n, threshold)
+  integer :: i, n, threshold
+  integer, parameter :: safelen = 2;
   !$OMP SIMD SAFELEN(safelen*2 + 2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop safelen(6) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd safelen(6) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -125,14 +132,15 @@ subroutine simdloop_with_safelen_clause_from_expr_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_safelen_clause
-subroutine simdloop_with_simdlen_safelen_clause(n, threshold)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_simdlen_safelen_clause
+subroutine simd_with_simdlen_safelen_clause(n, threshold)
+  integer :: i, n, threshold
   !$OMP SIMD SIMDLEN(1) SAFELEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %arg0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(1) safelen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(1) safelen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]] : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]] : !fir.ref<i32>
@@ -142,20 +150,21 @@ subroutine simdloop_with_simdlen_safelen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_collapse_clause
-subroutine simdloop_with_collapse_clause(n)
-integer :: i, j, n
-integer :: A(n,n)
-! CHECK: %[[LOWER_I:.*]] = arith.constant 1 : i32
-! CHECK: %[[UPPER_I:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
-! CHECK: %[[STEP_I:.*]] = arith.constant 1 : i32
-! CHECK: %[[LOWER_J:.*]] = arith.constant 1 : i32
-! CHECK: %[[UPPER_J:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
-! CHECK: %[[STEP_J:.*]] = arith.constant 1 : i32
-! CHECK: omp.simdloop  for (%[[ARG_0:.*]], %[[ARG_1:.*]]) : i32 = (
-! CHECK-SAME:               %[[LOWER_I]], %[[LOWER_J]]) to (
-! CHECK-SAME:               %[[UPPER_I]], %[[UPPER_J]]) inclusive step (
-! CHECK-SAME:               %[[STEP_I]], %[[STEP_J]]) {
+!CHECK-LABEL: func @_QPsimd_with_collapse_clause
+subroutine simd_with_collapse_clause(n)
+  integer :: i, j, n
+  integer :: A(n,n)
+  ! CHECK: %[[LOWER_I:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[UPPER_I:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
+  ! CHECK: %[[STEP_I:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[LOWER_J:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[UPPER_J:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
+  ! CHECK: %[[STEP_J:.*]] = arith.constant 1 : i32
+  ! CHECK: omp.simd {
+  ! CHECK-NEXT: omp.loop_nest (%[[ARG_0:.*]], %[[ARG_1:.*]]) : i32 = (
+  ! CHECK-SAME:               %[[LOWER_I]], %[[LOWER_J]]) to (
+  ! CHECK-SAME:               %[[UPPER_I]], %[[UPPER_J]]) inclusive step (
+  ! CHECK-SAME:               %[[STEP_I]], %[[STEP_J]]) {
   !$OMP SIMD COLLAPSE(2)
   do i = 1, n
     do j = 1, n
diff --git a/flang/test/Lower/OpenMP/if-clause.f90 b/flang/test/Lower/OpenMP/if-clause.f90
index f982bf67b0722..ce4427a0c2cab 100644
--- a/flang/test/Lower/OpenMP/if-clause.f90
+++ b/flang/test/Lower/OpenMP/if-clause.f90
@@ -116,7 +116,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-SAME: if({{.*}})
   ! CHECK:      omp.wsloop
@@ -124,7 +124,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-SAME: if({{.*}})
   ! CHECK:      omp.wsloop
@@ -134,7 +134,7 @@ program main
   do i = 1, 10
   end do
   !$omp end parallel do simd
-  
+
   ! CHECK:      omp.parallel
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
@@ -147,7 +147,7 @@ program main
   ! ----------------------------------------------------------------------------
   ! SIMD
   ! ----------------------------------------------------------------------------
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp simd
@@ -155,14 +155,14 @@ program main
   end do
   !$omp end simd
 
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp simd if(.true.)
   do i = 1, 10
   end do
   !$omp end simd
 
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp simd if(simd: .true.)
   do i = 1, 10
@@ -281,7 +281,6 @@ program main
   end do
   !$omp end target parallel do
 
-  
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
@@ -360,7 +359,7 @@ program main
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp target simd
@@ -370,7 +369,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(.true.)
   do i = 1, 10
@@ -379,7 +378,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(target: .true.) if(simd: .false.)
   do i = 1, 10
@@ -388,7 +387,7 @@ program main
 
   ! CHECK:      omp.target
   ! CHECK-SAME: if({{.*}})
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
   !$omp target simd if(target: .true.)
@@ -399,7 +398,7 @@ program main
   ! CHECK:      omp.target
   ! CHECK-NOT:  if({{.*}})
   ! CHECK-SAME: {
-  ! CHECK:      omp.simdloop
+  ! CHECK:      omp.simd
   ! CHECK-SAME: if({{.*}})
   !$omp target simd if(simd: .true.)
   do i = 1, 10
diff --git a/flang/test/Lower/OpenMP/loop-combined.f90 b/flang/test/Lower/OpenMP/loop-combined.f90
index 70488b6a769ce..298634b3f6f82 100644
--- a/flang/test/Lower/OpenMP/loop-combined.f90
+++ b/flang/test/Lower/OpenMP/loop-combined.f90
@@ -75,7 +75,7 @@ program main
   ! TARGET SIMD
   ! ----------------------------------------------------------------------------
   ! CHECK: omp.target
-  ! CHECK: omp.simdloop
+  ! CHECK: omp.simd
   !$omp target simd
   do i = 1, 10
   end do
diff --git a/flang/test/Lower/OpenMP/parallel-private-clause.f90 b/flang/test/Lower/OpenMP/parallel-private-clause.f90
index 5578b6710da7c..775f7b4f2cb10 100644
--- a/flang/test/Lower/OpenMP/parallel-private-clause.f90
+++ b/flang/test/Lower/OpenMP/parallel-private-clause.f90
@@ -411,7 +411,8 @@ subroutine simd_loop_1
   ! FIRDialect:     %[[UB:.*]] = arith.constant 9 : i32
   ! FIRDialect:     %[[STEP:.*]] = arith.constant 1 : i32
 
-  ! FIRDialect: omp.simdloop for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
+  ! FIRDialect: omp.simd {
+  ! FIRDialect-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   !$OMP SIMD PRIVATE(r)
   do i=1, 9
   ! FIRDialect:     fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
diff --git a/flang/test/Lower/OpenMP/simd.f90 b/flang/test/Lower/OpenMP/simd.f90
index 135b38c792623..190aa61521217 100644
--- a/flang/test/Lower/OpenMP/simd.f90
+++ b/flang/test/Lower/OpenMP/simd.f90
@@ -3,33 +3,35 @@
 !RUN: %flang_fc1 -flang-experimental-hlfir -emit-hlfir -fopenmp %s -o - | FileCheck %s
 !RUN: bbc -hlfir -emit-hlfir -fopenmp %s -o - | FileCheck %s
 
-!CHECK-LABEL: func @_QPsimdloop()
-subroutine simdloop
-integer :: i
+!CHECK-LABEL: func @_QPsimd()
+subroutine simd
+  integer :: i
   !$OMP SIMD
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK-NEXT: %[[UB:.*]] = arith.constant 9 : i32
   ! CHECK-NEXT: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK-NEXT: omp.simdloop for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
+  ! CHECK-NEXT: omp.simd {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i=1, 9
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
     ! CHECK: fir.call @_FortranAioOutputInteger32({{.*}}, %[[LD]]) {{.*}}: (!fir.ref<i8>, i32) -> i1
     print*, i
   end do
-  !$OMP END SIMD 
+  !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_if_clause
-subroutine simdloop_with_if_clause(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_if_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_if_clause
+subroutine simd_with_if_clause(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_if_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
   !$OMP SIMD IF( n .GE. threshold )
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
   ! CHECK: %[[COND:.*]] = arith.cmpi sge
-  ! CHECK: omp.simdloop if(%[[COND:.*]]) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd if(%[[COND:.*]]) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -39,15 +41,16 @@ subroutine simdloop_with_if_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause
-subroutine simdloop_with_simdlen_clause(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_simdlen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause
+subroutine simd_with_simdlen_clause(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_simdlen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
   !$OMP SIMD SIMDLEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -57,16 +60,17 @@ subroutine simdloop_with_simdlen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause_from_param
-subroutine simdloop_with_simdlen_clause_from_param(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_simdlen_clause_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
-integer, parameter :: simdlen = 2;
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause_from_param
+subroutine simd_with_simdlen_clause_from_param(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_simdlen_clause_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
+  integer, parameter :: simdlen = 2;
   !$OMP SIMD SIMDLEN(simdlen)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -76,16 +80,17 @@ subroutine simdloop_with_simdlen_clause_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_clause_from_expr_from_param
-subroutine simdloop_with_simdlen_clause_from_expr_from_param(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_simdlen_clause_from_expr_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
-integer, parameter :: simdlen = 2;
+!CHECK-LABEL: func @_QPsimd_with_simdlen_clause_from_expr_from_param
+subroutine simd_with_simdlen_clause_from_expr_from_param(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_simdlen_clause_from_expr_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
+  integer, parameter :: simdlen = 2;
   !$OMP SIMD SIMDLEN(simdlen*2 + 2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(6) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(6) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -95,15 +100,16 @@ subroutine simdloop_with_simdlen_clause_from_expr_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_safelen_clause
-subroutine simdloop_with_safelen_clause(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_safelen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_safelen_clause
+subroutine simd_with_safelen_clause(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_safelen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
   !$OMP SIMD SAFELEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop safelen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd safelen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -113,16 +119,17 @@ subroutine simdloop_with_safelen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_safelen_clause_from_expr_from_param
-subroutine simdloop_with_safelen_clause_from_expr_from_param(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_safelen_clause_from_expr_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
-integer, parameter :: safelen = 2;
+!CHECK-LABEL: func @_QPsimd_with_safelen_clause_from_expr_from_param
+subroutine simd_with_safelen_clause_from_expr_from_param(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_safelen_clause_from_expr_from_paramEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
+  integer, parameter :: safelen = 2;
   !$OMP SIMD SAFELEN(safelen*2 + 2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop safelen(6) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd safelen(6) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -132,15 +139,16 @@ subroutine simdloop_with_safelen_clause_from_expr_from_param(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_simdlen_safelen_clause
-subroutine simdloop_with_simdlen_safelen_clause(n, threshold)
-  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimdloop_with_simdlen_safelen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-integer :: i, n, threshold
+!CHECK-LABEL: func @_QPsimd_with_simdlen_safelen_clause
+subroutine simd_with_simdlen_safelen_clause(n, threshold)
+  ! CHECK: %[[ARG_N:.*]]:2 = hlfir.declare %{{.*}} {uniq_name = "_QFsimd_with_simdlen_safelen_clauseEn"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+  integer :: i, n, threshold
   !$OMP SIMD SIMDLEN(1) SAFELEN(2)
   ! CHECK: %[[LB:.*]] = arith.constant 1 : i32
   ! CHECK: %[[UB:.*]] = fir.load %[[ARG_N]]#0
   ! CHECK: %[[STEP:.*]] = arith.constant 1 : i32
-  ! CHECK: omp.simdloop simdlen(1) safelen(2) for (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive  step (%[[STEP]]) {
+  ! CHECK: omp.simd simdlen(1) safelen(2) {
+  ! CHECK-NEXT: omp.loop_nest (%[[I:.*]]) : i32 = (%[[LB]]) to (%[[UB]]) inclusive step (%[[STEP]]) {
   do i = 1, n
     ! CHECK: fir.store %[[I]] to %[[LOCAL:.*]]#1 : !fir.ref<i32>
     ! CHECK: %[[LD:.*]] = fir.load %[[LOCAL]]#0 : !fir.ref<i32>
@@ -150,20 +158,21 @@ subroutine simdloop_with_simdlen_safelen_clause(n, threshold)
   !$OMP END SIMD
 end subroutine
 
-!CHECK-LABEL: func @_QPsimdloop_with_collapse_clause
-subroutine simdloop_with_collapse_clause(n)
-integer :: i, j, n
-integer :: A(n,n)
-! CHECK: %[[LOWER_I:.*]] = arith.constant 1 : i32
-! CHECK: %[[UPPER_I:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
-! CHECK: %[[STEP_I:.*]] = arith.constant 1 : i32
-! CHECK: %[[LOWER_J:.*]] = arith.constant 1 : i32
-! CHECK: %[[UPPER_J:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
-! CHECK: %[[STEP_J:.*]] = arith.constant 1 : i32
-! CHECK: omp.simdloop  for (%[[ARG_0:.*]], %[[ARG_1:.*]]) : i32 = (
-! CHECK-SAME:               %[[LOWER_I]], %[[LOWER_J]]) to (
-! CHECK-SAME:               %[[UPPER_I]], %[[UPPER_J]]) inclusive step (
-! CHECK-SAME:               %[[STEP_I]], %[[STEP_J]]) {
+!CHECK-LABEL: func @_QPsimd_with_collapse_clause
+subroutine simd_with_collapse_clause(n)
+  integer :: i, j, n
+  integer :: A(n,n)
+  ! CHECK: %[[LOWER_I:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[UPPER_I:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
+  ! CHECK: %[[STEP_I:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[LOWER_J:.*]] = arith.constant 1 : i32
+  ! CHECK: %[[UPPER_J:.*]] = fir.load %[[PARAM_ARG:.*]] : !fir.ref<i32>
+  ! CHECK: %[[STEP_J:.*]] = arith.constant 1 : i32
+  ! CHECK: omp.simd {
+  ! CHECK-NEXT: omp.loop_nest (%[[ARG_0:.*]], %[[ARG_1:.*]]) : i32 = (
+  ! CHECK-SAME:                %[[LOWER_I]], %[[LOWER_J]]) to (
+  ! CHECK-SAME:                %[[UPPER_I]], %[[UPPER_J]]) inclusive step (
+  ! CHECK-SAME:                %[[STEP_I]], %[[STEP_J]]) {
   !$OMP SIMD COLLAPSE(2)
   do i = 1, n
     do j = 1, n
diff --git a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
index db1c4a8951ad2..8344bca08404e 100644
--- a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
+++ b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
@@ -2097,7 +2097,7 @@ TEST_F(OpenMPIRBuilderTest, ApplySimdlenSafelen) {
   }));
 }
 
-TEST_F(OpenMPIRBuilderTest, ApplySimdLoopIf) {
+TEST_F(OpenMPIRBuilderTest, ApplySimdIf) {
   OpenMPIRBuilder OMPBuilder(*M);
   IRBuilder<> Builder(BB);
   MapVector<Value *, Value *> AlignedVars;
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
index 27a766aceb316..3c5fa23bd4a7f 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauseOperands.h
@@ -251,11 +251,10 @@ using SectionsClauseOps = detail::Clauses<AllocateClauseOps, NowaitClauseOps,
                                           PrivateClauseOps, ReductionClauseOps>;
 
 // TODO `linear` clause.
-using SimdLoopClauseOps =
-    detail::Clauses<AlignedClauseOps, CollapseClauseOps, IfClauseOps,
-                    LoopRelatedOps, NontemporalClauseOps, OrderClauseOps,
-                    PrivateClauseOps, ReductionClauseOps, SafelenClauseOps,
-                    SimdlenClauseOps>;
+using SimdClauseOps =
+    detail::Clauses<AlignedClauseOps, IfClauseOps, NontemporalClauseOps,
+                    OrderClauseOps, PrivateClauseOps, ReductionClauseOps,
+                    SafelenClauseOps, SimdlenClauseOps>;
 
 using SingleClauseOps = detail::Clauses<AllocateClauseOps, CopyprivateClauseOps,
                                         NowaitClauseOps, PrivateClauseOps>;
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 82be7ad31a158..10771f6e854dd 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -562,7 +562,7 @@ def LoopNestOp : OpenMP_Op<"loop_nest", [SameVariadicOperandSize,
     loop operations intended to serve as a stopgap solution until the long-term
     representation of canonical loops is defined. Specifically, this operation
     is intended to serve as a unique source for loop information during the
-    transition to making `omp.distribute`, `omp.simdloop`, `omp.taskloop` and
+    transition to making `omp.distribute`, `omp.simd`, `omp.taskloop` and
     `omp.wsloop` wrapper operations. It is not intended to help with the
     addition of support for loop transformations, non-rectangular loops and
     non-perfectly nested loops.
@@ -722,24 +722,19 @@ def WsloopOp : OpenMP_Op<"wsloop", [AttrSizedOperandSegments,
 // Simd construct [2.9.3.1]
 //===----------------------------------------------------------------------===//
 
-def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
-                         AllTypesMatch<["lowerBound", "upperBound", "step"]>,
-                         DeclareOpInterfaceMethods<LoopWrapperInterface>,
-                         RecursiveMemoryEffects]> {
- let summary = "simd loop construct";
+def SimdOp : OpenMP_Op<"simd", [AttrSizedOperandSegments,
+                       DeclareOpInterfaceMethods<LoopWrapperInterface>,
+                       RecursiveMemoryEffects,
+                       SingleBlockImplicitTerminator<"TerminatorOp">]> {
+ let summary = "simd construct";
   let description = [{
     The simd construct can be applied to a loop to indicate that the loop can be
     transformed into a SIMD loop (that is, multiple iterations of the loop can
-    be executed concurrently using SIMD instructions).. The lower and upper
-    bounds specify a half-open range: the range includes the lower bound but
-    does not include the upper bound. If the `inclusive` attribute is specified
-    then the upper bound is also included.
+    be executed concurrently using SIMD instructions).
 
-    The body region can contain any number of blocks. The region is terminated
-    by "omp.yield" instruction without operands.
-
-    Collapsed loops are represented by the simd-loop having a list of indices,
-    bounds and steps where the size of the list is equal to the collapse value.
+    The body region can contain a single block which must contain a single
+    operation and a terminator. The operation must be another compatible loop
+    wrapper or an `omp.loop_nest`.
 
     The `alignment_values` attribute additionally specifies alignment of each
     corresponding aligned operand. Note that `$aligned_vars` and
@@ -763,32 +758,32 @@ def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
     SIMD chunk can have a distance in the logical iteration space that is
     greater than or equal to the value given in the clause.
     ```
-    omp.simdloop <clauses>
-    for (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
-      // block operations
-      omp.yield
+    omp.simd <clauses> {
+      omp.loop_nest (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
+        %a = load %arrA[%i1, %i2] : memref<?x?xf32>
+        %b = load %arrB[%i1, %i2] : memref<?x?xf32>
+        %sum = arith.addf %a, %b : f32
+        store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
+        omp.yield
+      }
     }
     ```
   }];
 
   // TODO: Add other clauses
-  let arguments = (ins Variadic<IntLikeType>:$lowerBound,
-             Variadic<IntLikeType>:$upperBound,
-             Variadic<IntLikeType>:$step,
-             Variadic<OpenMP_PointerLikeType>:$aligned_vars,
+  let arguments = (ins Variadic<OpenMP_PointerLikeType>:$aligned_vars,
              OptionalAttr<I64ArrayAttr>:$alignment_values,
              Optional<I1>:$if_expr,
              Variadic<OpenMP_PointerLikeType>:$nontemporal_vars,
              OptionalAttr<OrderKindAttr>:$order_val,
              ConfinedAttr<OptionalAttr<I64Attr>, [IntPositive]>:$simdlen,
-             ConfinedAttr<OptionalAttr<I64Attr>, [IntPositive]>:$safelen,
-             UnitAttr:$inclusive
+             ConfinedAttr<OptionalAttr<I64Attr>, [IntPositive]>:$safelen
      );
 
   let regions = (region AnyRegion:$region);
 
   let builders = [
-    OpBuilder<(ins CArg<"const SimdLoopClauseOps &">:$clauses)>
+    OpBuilder<(ins CArg<"const SimdClauseOps &">:$clauses)>
   ];
 
   let assemblyFormat = [{
@@ -800,14 +795,7 @@ def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
           |`order` `(` custom<ClauseAttr>($order_val) `)`
           |`simdlen` `(` $simdlen  `)`
           |`safelen` `(` $safelen  `)`
-    ) `for` custom<LoopControl>($region, $lowerBound, $upperBound, $step,
-                                  type($step), $inclusive) attr-dict
-  }];
-
-  let extraClassDeclaration = [{
-    /// Returns the number of loops in the simd loop nest.
-    unsigned getNumLoops() { return getLowerBound().size(); }
-
+    ) $region attr-dict
   }];
 
   let hasCustomAssemblyFormat = 1;
@@ -818,7 +806,7 @@ def SimdLoopOp : OpenMP_Op<"simdloop", [AttrSizedOperandSegments,
 def YieldOp : OpenMP_Op<"yield",
     [Pure, ReturnLike, Terminator,
      ParentOneOf<["LoopNestOp", "WsloopOp", "DeclareReductionOp",
-     "AtomicUpdateOp", "SimdLoopOp", "PrivateClauseOp"]>]> {
+     "AtomicUpdateOp", "PrivateClauseOp"]>]> {
   let summary = "loop yield and termination operation";
   let description = [{
     "omp.yield" yields SSA values from the OpenMP dialect op region and
diff --git a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
index b9ada0fa0f979..a206c7b228d21 100644
--- a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
+++ b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
@@ -251,11 +251,11 @@ void mlir::configureOpenMPToLLVMConversionLegality(
   });
   target.addDynamicallyLegalOp<
       mlir::omp::AtomicUpdateOp, mlir::omp::CriticalOp, mlir::omp::TargetOp,
-      mlir::omp::TargetDataOp, mlir::omp::OrderedRegionOp,
-      mlir::omp::ParallelOp, mlir::omp::WsloopOp, mlir::omp::SimdLoopOp,
-      mlir::omp::MasterOp, mlir::omp::SectionOp, mlir::omp::SectionsOp,
-      mlir::omp::SingleOp, mlir::omp::TaskgroupOp, mlir::omp::TaskOp,
-      mlir::omp::DeclareReductionOp,
+      mlir::omp::TargetDataOp, mlir::omp::LoopNestOp,
+      mlir::omp::OrderedRegionOp, mlir::omp::ParallelOp, mlir::omp::WsloopOp,
+      mlir::omp::SimdOp, mlir::omp::MasterOp, mlir::omp::SectionOp,
+      mlir::omp::SectionsOp, mlir::omp::SingleOp, mlir::omp::TaskgroupOp,
+      mlir::omp::TaskOp, mlir::omp::DeclareReductionOp,
       mlir::omp::PrivateClauseOp>([&](Operation *op) {
     return std::all_of(op->getRegions().begin(), op->getRegions().end(),
                        [&](Region &region) {
@@ -278,11 +278,12 @@ void mlir::populateOpenMPToLLVMConversionPatterns(LLVMTypeConverter &converter,
       AtomicReadOpConversion, MapInfoOpConversion, ReductionOpConversion,
       MultiRegionOpConversion<omp::DeclareReductionOp>,
       MultiRegionOpConversion<omp::PrivateClauseOp>,
-      RegionOpConversion<omp::CriticalOp>, RegionOpConversion<omp::MasterOp>,
-      ReductionOpConversion, RegionOpConversion<omp::OrderedRegionOp>,
+      RegionOpConversion<omp::CriticalOp>, RegionOpConversion<omp::LoopNestOp>,
+      RegionOpConversion<omp::MasterOp>, ReductionOpConversion,
+      RegionOpConversion<omp::OrderedRegionOp>,
       RegionOpConversion<omp::ParallelOp>, RegionOpConversion<omp::WsloopOp>,
       RegionOpConversion<omp::SectionsOp>, RegionOpConversion<omp::SectionOp>,
-      RegionOpConversion<omp::SimdLoopOp>, RegionOpConversion<omp::SingleOp>,
+      RegionOpConversion<omp::SimdOp>, RegionOpConversion<omp::SingleOp>,
       RegionOpConversion<omp::TaskgroupOp>, RegionOpConversion<omp::TaskOp>,
       RegionOpConversion<omp::TargetDataOp>, RegionOpConversion<omp::TargetOp>,
       RegionLessOpWithVarOperandsConversion<omp::AtomicWriteOp>,
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index e500d0fca741f..caf0ac3f86017 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1604,34 +1604,37 @@ void printLoopControl(OpAsmPrinter &p, Operation *op, Region &region,
 // Simd construct [2.9.3.1]
 //===----------------------------------------------------------------------===//
 
-void SimdLoopOp::build(OpBuilder &builder, OperationState &state,
-                       const SimdLoopClauseOps &clauses) {
+void SimdOp::build(OpBuilder &builder, OperationState &state,
+                   const SimdClauseOps &clauses) {
   MLIRContext *ctx = builder.getContext();
   // TODO Store clauses in op: privateVars, reductionByRefAttr, reductionVars,
   // privatizers, reductionDeclSymbols.
-  SimdLoopOp::build(
-      builder, state, clauses.loopLBVar, clauses.loopUBVar, clauses.loopStepVar,
-      clauses.alignedVars, makeArrayAttr(ctx, clauses.alignmentAttrs),
-      clauses.ifVar, clauses.nontemporalVars, clauses.orderAttr,
-      clauses.simdlenAttr, clauses.safelenAttr, clauses.loopInclusiveAttr);
+  SimdOp::build(builder, state, clauses.alignedVars,
+                makeArrayAttr(ctx, clauses.alignmentAttrs), clauses.ifVar,
+                clauses.nontemporalVars, clauses.orderAttr, clauses.simdlenAttr,
+                clauses.safelenAttr);
 }
 
-LogicalResult SimdLoopOp::verify() {
-  if (this->getLowerBound().empty()) {
-    return emitOpError() << "empty lowerbound for simd loop operation";
-  }
-  if (this->getSimdlen().has_value() && this->getSafelen().has_value() &&
-      this->getSimdlen().value() > this->getSafelen().value()) {
+LogicalResult SimdOp::verify() {
+  if (getSimdlen().has_value() && getSafelen().has_value() &&
+      getSimdlen().value() > getSafelen().value())
     return emitOpError()
            << "simdlen clause and safelen clause are both present, but the "
               "simdlen value is not less than or equal to safelen value";
-  }
-  if (verifyAlignedClause(*this, this->getAlignmentValues(),
-                          this->getAlignedVars())
+
+  if (verifyAlignedClause(*this, getAlignmentValues(), getAlignedVars())
           .failed())
     return failure();
-  if (verifyNontemporalClause(*this, this->getNontemporalVars()).failed())
+
+  if (verifyNontemporalClause(*this, getNontemporalVars()).failed())
     return failure();
+
+  if (!isWrapper())
+    return emitOpError() << "must be a loop wrapper";
+
+  if (getNestedWrapper())
+    return emitOpError() << "must wrap an 'omp.loop_nest' directly";
+
   return success();
 }
 
@@ -1662,9 +1665,9 @@ LogicalResult DistributeOp::verify() {
   if (LoopWrapperInterface nested = getNestedWrapper()) {
     // Check for the allowed leaf constructs that may appear in a composite
     // construct directly after DISTRIBUTE.
-    if (!isa<ParallelOp, SimdLoopOp>(nested))
+    if (!isa<ParallelOp, SimdOp>(nested))
       return emitError() << "only supported nested wrappers are 'omp.parallel' "
-                            "and 'omp.simdloop'";
+                            "and 'omp.simd'";
   }
 
   return success();
@@ -1876,8 +1879,8 @@ LogicalResult TaskloopOp::verify() {
   if (LoopWrapperInterface nested = getNestedWrapper()) {
     // Check for the allowed leaf constructs that may appear in a composite
     // construct directly after TASKLOOP.
-    if (!isa<SimdLoopOp>(nested))
-      return emitError() << "only supported nested wrapper is 'omp.simdloop'";
+    if (!isa<SimdOp>(nested))
+      return emitError() << "only supported nested wrapper is 'omp.simd'";
   }
   return success();
 }
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 300fc8ba56fc5..e89ff9209b034 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -1406,9 +1406,10 @@ convertOmpParallel(omp::ParallelOp opInst, llvm::IRBuilderBase &builder,
 
 /// Converts an OpenMP simd loop into LLVM IR using OpenMPIRBuilder.
 static LogicalResult
-convertOmpSimdLoop(Operation &opInst, llvm::IRBuilderBase &builder,
-                   LLVM::ModuleTranslation &moduleTranslation) {
-  auto loop = cast<omp::SimdLoopOp>(opInst);
+convertOmpSimd(Operation &opInst, llvm::IRBuilderBase &builder,
+               LLVM::ModuleTranslation &moduleTranslation) {
+  auto simdOp = cast<omp::SimdOp>(opInst);
+  auto loopOp = cast<omp::LoopNestOp>(simdOp.getWrappedLoop());
 
   llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
 
@@ -1421,33 +1422,34 @@ convertOmpSimdLoop(Operation &opInst, llvm::IRBuilderBase &builder,
   auto bodyGen = [&](llvm::OpenMPIRBuilder::InsertPointTy ip, llvm::Value *iv) {
     // Make sure further conversions know about the induction variable.
     moduleTranslation.mapValue(
-        loop.getRegion().front().getArgument(loopInfos.size()), iv);
+        loopOp.getRegion().front().getArgument(loopInfos.size()), iv);
 
     // Capture the body insertion point for use in nested loops. BodyIP of the
     // CanonicalLoopInfo always points to the beginning of the entry block of
     // the body.
     bodyInsertPoints.push_back(ip);
 
-    if (loopInfos.size() != loop.getNumLoops() - 1)
+    if (loopInfos.size() != loopOp.getNumLoops() - 1)
       return;
 
     // Convert the body of the loop.
     builder.restoreIP(ip);
-    convertOmpOpRegions(loop.getRegion(), "omp.simdloop.region", builder,
+    convertOmpOpRegions(loopOp.getRegion(), "omp.simd.region", builder,
                         moduleTranslation, bodyGenStatus);
   };
 
   // Delegate actual loop construction to the OpenMP IRBuilder.
-  // TODO: this currently assumes SimdLoop is semantically similar to SCF loop,
-  // i.e. it has a positive step, uses signed integer semantics. Reconsider
-  // this code when SimdLoop clearly supports more cases.
+  // TODO: this currently assumes omp.loop_nest is semantically similar to SCF
+  // loop, i.e. it has a positive step, uses signed integer semantics.
+  // Reconsider this code when the nested loop operation clearly supports more
+  // cases.
   llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
-  for (unsigned i = 0, e = loop.getNumLoops(); i < e; ++i) {
+  for (unsigned i = 0, e = loopOp.getNumLoops(); i < e; ++i) {
     llvm::Value *lowerBound =
-        moduleTranslation.lookupValue(loop.getLowerBound()[i]);
+        moduleTranslation.lookupValue(loopOp.getLowerBound()[i]);
     llvm::Value *upperBound =
-        moduleTranslation.lookupValue(loop.getUpperBound()[i]);
-    llvm::Value *step = moduleTranslation.lookupValue(loop.getStep()[i]);
+        moduleTranslation.lookupValue(loopOp.getUpperBound()[i]);
+    llvm::Value *step = moduleTranslation.lookupValue(loopOp.getStep()[i]);
 
     // Make sure loop trip count are emitted in the preheader of the outermost
     // loop at the latest so that they are all available for the new collapsed
@@ -1473,18 +1475,18 @@ convertOmpSimdLoop(Operation &opInst, llvm::IRBuilderBase &builder,
       ompBuilder->collapseLoops(ompLoc.DL, loopInfos, {});
 
   llvm::ConstantInt *simdlen = nullptr;
-  if (std::optional<uint64_t> simdlenVar = loop.getSimdlen())
+  if (std::optional<uint64_t> simdlenVar = simdOp.getSimdlen())
     simdlen = builder.getInt64(simdlenVar.value());
 
   llvm::ConstantInt *safelen = nullptr;
-  if (std::optional<uint64_t> safelenVar = loop.getSafelen())
+  if (std::optional<uint64_t> safelenVar = simdOp.getSafelen())
     safelen = builder.getInt64(safelenVar.value());
 
   llvm::MapVector<llvm::Value *, llvm::Value *> alignedVars;
   ompBuilder->applySimd(
       loopInfo, alignedVars,
-      loop.getIfExpr() ? moduleTranslation.lookupValue(loop.getIfExpr())
-                       : nullptr,
+      simdOp.getIfExpr() ? moduleTranslation.lookupValue(simdOp.getIfExpr())
+                         : nullptr,
       llvm::omp::OrderKind::OMP_ORDER_unknown, simdlen, safelen);
 
   builder.restoreIP(afterIP);
@@ -3198,8 +3200,8 @@ convertHostOrTargetOperation(Operation *op, llvm::IRBuilderBase &builder,
       .Case([&](omp::WsloopOp) {
         return convertOmpWsloop(*op, builder, moduleTranslation);
       })
-      .Case([&](omp::SimdLoopOp) {
-        return convertOmpSimdLoop(*op, builder, moduleTranslation);
+      .Case([&](omp::SimdOp) {
+        return convertOmpSimd(*op, builder, moduleTranslation);
       })
       .Case([&](omp::AtomicReadOp) {
         return convertOmpAtomicRead(*op, builder, moduleTranslation);
@@ -3421,7 +3423,6 @@ LogicalResult OpenMPDialectLLVMIRTranslationInterface::convertOperation(
       return convertTargetOpsInNest(op, builder, moduleTranslation);
     }
   }
-
   return convertHostOrTargetOperation(op, builder, moduleTranslation);
 }
 
diff --git a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
index dc5d6969ca789..9f45d139b81f2 100644
--- a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
+++ b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
@@ -145,9 +145,10 @@ func.func @threadprivate(%a: !llvm.ptr) -> () {
 
 // -----
 
-// CHECK:      llvm.func @simdloop_block_arg(%[[LOWER:.*]]: i32, %[[UPPER:.*]]: i32, %[[ITER:.*]]: i64) {
-// CHECK:      omp.simdloop   for  (%[[ARG_0:.*]]) : i32 =
-// CHECK-SAME:     (%[[LOWER]]) to (%[[UPPER]]) inclusive step (%[[LOWER]]) {
+// CHECK:      llvm.func @loop_nest_block_arg(%[[LOWER:.*]]: i32, %[[UPPER:.*]]: i32, %[[ITER:.*]]: i64) {
+// CHECK:      omp.simd {
+// CHECK-NEXT: omp.loop_nest (%[[ARG_0:.*]]) : i32 = (%[[LOWER]])
+// CHECK-SAME: to (%[[UPPER]]) inclusive step (%[[LOWER]]) {
 // CHECK:      llvm.br ^[[BB1:.*]](%[[ITER]] : i64)
 // CHECK:        ^[[BB1]](%[[VAL_0:.*]]: i64):
 // CHECK:          %[[VAL_1:.*]] = llvm.icmp "slt" %[[VAL_0]], %[[ITER]] : i64
@@ -157,17 +158,19 @@ func.func @threadprivate(%a: !llvm.ptr) -> () {
 // CHECK:          llvm.br ^[[BB1]](%[[VAL_2]] : i64)
 // CHECK:        ^[[BB3]]:
 // CHECK:          omp.yield
-func.func @simdloop_block_arg(%val : i32, %ub : i32, %i : index) {
-  omp.simdloop   for  (%arg0) : i32 = (%val) to (%ub) inclusive step (%val) {
-    cf.br ^bb1(%i : index)
-  ^bb1(%0: index):
-    %1 = arith.cmpi slt, %0, %i : index
-    cf.cond_br %1, ^bb2, ^bb3
-  ^bb2:
-    %2 = arith.addi %0, %i : index
-    cf.br ^bb1(%2 : index)
-  ^bb3:
-    omp.yield
+func.func @loop_nest_block_arg(%val : i32, %ub : i32, %i : index) {
+  omp.simd {
+    omp.loop_nest (%arg0) : i32 = (%val) to (%ub) inclusive step (%val) {
+      cf.br ^bb1(%i : index)
+    ^bb1(%0: index):
+      %1 = arith.cmpi slt, %0, %i : index
+      cf.cond_br %1, ^bb2, ^bb3
+    ^bb2:
+      %2 = arith.addi %0, %i : index
+      cf.br ^bb1(%2 : index)
+    ^bb3:
+      omp.yield
+    }
   }
   return
 }
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index 7f86a7f5b3182..9323beadf4549 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -243,145 +243,168 @@ llvm.func @test_omp_wsloop_dynamic_wrong_modifier3(%lb : i64, %ub : i64, %step :
 
 // -----
 
-func.func @omp_simdloop(%lb : index, %ub : index, %step : i32) -> () {
-  // expected-error @below {{op failed to verify that all of {lowerBound, upperBound, step} have same type}}
-  "omp.simdloop" (%lb, %ub, %step) ({
-    ^bb0(%iv: index):
-      omp.yield
-  }) {operandSegmentSizes = array<i32: 1,1,1,0,0,0>} :
-    (index, index, i32) -> ()
+func.func @omp_simd() -> () {
+  // expected-error @below {{op must be a loop wrapper}}
+  omp.simd {
+    omp.terminator
+  }
+  return
+}
 
+// -----
+
+func.func @omp_simd_nested_wrapper() -> () {
+  // expected-error @below {{op must wrap an 'omp.loop_nest' directly}}
+  omp.simd {
+    omp.distribute {
+      omp.terminator
+    }
+  }
   return
 }
 
 // -----
 
-func.func @omp_simdloop_pretty_aligned(%lb : index, %ub : index, %step : index,
-                                       %data_var : memref<i32>) -> () {
+func.func @omp_simd_pretty_aligned(%lb : index, %ub : index, %step : index,
+                                   %data_var : memref<i32>) -> () {
   //  expected-error @below {{expected '->'}}
-  omp.simdloop aligned(%data_var : memref<i32>)
-  for (%iv) : index = (%lb) to (%ub) step (%step) {
-    omp.yield
+  omp.simd aligned(%data_var : memref<i32>) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
 // -----
 
-func.func @omp_simdloop_aligned_mismatch(%arg0 : index, %arg1 : index,
-                                         %arg2 : index, %arg3 : memref<i32>,
-                                         %arg4 : memref<i32>) -> () {
+func.func @omp_simd_aligned_mismatch(%arg0 : index, %arg1 : index,
+                                     %arg2 : index, %arg3 : memref<i32>,
+                                     %arg4 : memref<i32>) -> () {
   //  expected-error @below {{op expected as many alignment values as aligned variables}}
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg4) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
+  "omp.simd"(%arg3, %arg4) ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
   }) {alignment_values = [128],
-      operandSegmentSizes = array<i32: 1, 1, 1, 2, 0, 0>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+      operandSegmentSizes = array<i32: 2, 0, 0>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_aligned_negative(%arg0 : index, %arg1 : index,
-                                         %arg2 : index, %arg3 : memref<i32>,
-                                         %arg4 : memref<i32>) -> () {
+func.func @omp_simd_aligned_negative(%arg0 : index, %arg1 : index,
+                                     %arg2 : index, %arg3 : memref<i32>,
+                                     %arg4 : memref<i32>) -> () {
   //  expected-error @below {{op alignment should be greater than 0}}
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg4) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
-  }) {alignment_values = [-1, 128], operandSegmentSizes = array<i32: 1, 1, 1,2, 0, 0>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+  "omp.simd"(%arg3, %arg4) ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
+  }) {alignment_values = [-1, 128], operandSegmentSizes = array<i32: 2, 0, 0>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_unexpected_alignment(%arg0 : index, %arg1 : index,
-                                             %arg2 : index, %arg3 : memref<i32>,
-                                             %arg4 : memref<i32>) -> () {
+func.func @omp_simd_unexpected_alignment(%arg0 : index, %arg1 : index,
+                                         %arg2 : index, %arg3 : memref<i32>,
+                                         %arg4 : memref<i32>) -> () {
   //  expected-error @below {{unexpected alignment values attribute}}
-  "omp.simdloop"(%arg0, %arg1, %arg2) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
-  }) {alignment_values = [1, 128], operandSegmentSizes = array<i32: 1, 1, 1, 0, 0, 0>} : (index, index, index) -> ()
+  "omp.simd"() ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
+  }) {alignment_values = [1, 128]} : () -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_aligned_float(%arg0 : index, %arg1 : index,
-                                      %arg2 : index, %arg3 : memref<i32>,
-                                      %arg4 : memref<i32>) -> () {
+func.func @omp_simd_aligned_float(%arg0 : index, %arg1 : index,
+                                  %arg2 : index, %arg3 : memref<i32>,
+                                  %arg4 : memref<i32>) -> () {
   //  expected-error @below {{failed to satisfy constraint: 64-bit integer array attribute}}
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg4) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
-  }) {alignment_values = [1.5, 128], operandSegmentSizes = array<i32: 1, 1, 1,2, 0, 0>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+  "omp.simd"(%arg3, %arg4) ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
+  }) {alignment_values = [1.5, 128], operandSegmentSizes = array<i32: 2, 0, 0>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_aligned_the_same_var(%arg0 : index, %arg1 : index,
-                                             %arg2 : index, %arg3 : memref<i32>,
-                                             %arg4 : memref<i32>) -> () {
+func.func @omp_simd_aligned_the_same_var(%arg0 : index, %arg1 : index,
+                                         %arg2 : index, %arg3 : memref<i32>,
+                                         %arg4 : memref<i32>) -> () {
   //  expected-error @below {{aligned variable used more than once}}
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg3) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
-  }) {alignment_values = [1, 128], operandSegmentSizes = array<i32: 1, 1, 1,2, 0, 0>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+  "omp.simd"(%arg3, %arg3) ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
+  }) {alignment_values = [1, 128], operandSegmentSizes = array<i32: 2, 0, 0>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_nontemporal_the_same_var(%arg0 : index,
-                                                 %arg1 : index,
-                                                 %arg2 : index,
-                                                 %arg3 : memref<i32>) -> () {
+func.func @omp_simd_nontemporal_the_same_var(%arg0 : index,  %arg1 : index,
+                                             %arg2 : index,
+                                             %arg3 : memref<i32>) -> () {
   //  expected-error @below {{nontemporal variable used more than once}}
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg3) ({
-    ^bb0(%arg5: index):
-      "omp.yield"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 1, 1, 1, 0, 0, 2>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+  "omp.simd"(%arg3, %arg3) ({
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
+  }) {operandSegmentSizes = array<i32: 0, 0, 2>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
 // -----
 
-func.func @omp_simdloop_order_value(%lb : index, %ub : index, %step : index) {
+func.func @omp_simd_order_value(%lb : index, %ub : index, %step : index) {
   // expected-error @below {{invalid clause value: 'default'}}
-  omp.simdloop order(default) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+  omp.simd order(default) {
+    omp.loop_nest (%iv) : index = (%arg0) to (%arg1) step (%arg2) {
+      omp.yield
+    }
   }
   return
 }
 
 // -----
 
-func.func @omp_simdloop_pretty_simdlen(%lb : index, %ub : index, %step : index) -> () {
+func.func @omp_simd_pretty_simdlen(%lb : index, %ub : index, %step : index) -> () {
   // expected-error @below {{op attribute 'simdlen' failed to satisfy constraint: 64-bit signless integer attribute whose value is positive}}
-  omp.simdloop simdlen(0) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+  omp.simd simdlen(0) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
 // -----
 
-func.func @omp_simdloop_pretty_safelen(%lb : index, %ub : index, %step : index) -> () {
+func.func @omp_simd_pretty_safelen(%lb : index, %ub : index, %step : index) -> () {
   // expected-error @below {{op attribute 'safelen' failed to satisfy constraint: 64-bit signless integer attribute whose value is positive}}
-  omp.simdloop safelen(0) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+  omp.simd safelen(0) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
 // -----
 
-func.func @omp_simdloop_pretty_simdlen_safelen(%lb : index, %ub : index, %step : index) -> () {
-  // expected-error @below {{'omp.simdloop' op simdlen clause and safelen clause are both present, but the simdlen value is not less than or equal to safelen value}}
-  omp.simdloop simdlen(2) safelen(1) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+func.func @omp_simd_pretty_simdlen_safelen(%lb : index, %ub : index, %step : index) -> () {
+  // expected-error @below {{op simdlen clause and safelen clause are both present, but the simdlen value is not less than or equal to safelen value}}
+  omp.simd simdlen(2) safelen(1) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
@@ -1720,7 +1743,7 @@ func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
 // -----
 
 func.func @taskloop(%lb: i32, %ub: i32, %step: i32) {
-  // expected-error @below {{only supported nested wrapper is 'omp.simdloop'}}
+  // expected-error @below {{only supported nested wrapper is 'omp.simd'}}
   omp.taskloop {
     omp.distribute {
       omp.loop_nest (%iv) : i32 = (%lb) to (%ub) step (%step) {
@@ -1927,7 +1950,7 @@ func.func @omp_distribute_wrapper() -> () {
 // -----
 
 func.func @omp_distribute_nested_wrapper(%data_var : memref<i32>) -> () {
-  // expected-error @below {{only supported nested wrappers are 'omp.parallel' and 'omp.simdloop'}}
+  // expected-error @below {{only supported nested wrappers are 'omp.parallel' and 'omp.simd'}}
   "omp.distribute"() ({
       "omp.wsloop"() ({
         %0 = arith.constant 0 : i32
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index 802e1795b3fff..e2ca12afc14bd 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -439,154 +439,161 @@ func.func @omp_wsloop_pretty_multiple(%lb1 : i32, %ub1 : i32, %step1 : i32, %lb2
   return
 }
 
-// CHECK-LABEL: omp_simdloop
-func.func @omp_simdloop(%lb : index, %ub : index, %step : index) -> () {
-  // CHECK: omp.simdloop for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  "omp.simdloop" (%lb, %ub, %step) ({
-    ^bb0(%iv: index):
-      omp.yield
-  }) {operandSegmentSizes = array<i32: 1,1,1,0,0,0>} :
-    (index, index, index) -> ()
+// CHECK-LABEL: omp_simd
+func.func @omp_simd(%lb : index, %ub : index, %step : index) -> () {
+  // CHECK: omp.simd
+  "omp.simd" () ({
+    "omp.loop_nest" (%lb, %ub, %step) ({
+    ^bb1(%iv2: index):
+      "omp.yield"() : () -> ()
+    }) : (index, index, index) -> ()
+    "omp.terminator"() : () -> ()
+  }) : () -> ()
 
   return
 }
 
-// CHECK-LABEL: omp_simdloop_aligned_list
-func.func @omp_simdloop_aligned_list(%arg0 : index, %arg1 : index, %arg2 : index,
-                                     %arg3 : memref<i32>, %arg4 : memref<i32>) -> () {
-  // CHECK:      omp.simdloop   aligned(%{{.*}} : memref<i32> -> 32 : i64,
+// CHECK-LABEL: omp_simd_aligned_list
+func.func @omp_simd_aligned_list(%arg0 : index, %arg1 : index, %arg2 : index,
+                                 %arg3 : memref<i32>, %arg4 : memref<i32>) -> () {
+  // CHECK:      omp.simd aligned(
+  // CHECK-SAME: %{{.*}} : memref<i32> -> 32 : i64,
   // CHECK-SAME: %{{.*}} : memref<i32> -> 128 : i64)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg4) ({
-    ^bb0(%arg5: index):
+  "omp.simd"(%arg3, %arg4) ({
+    "omp.loop_nest" (%arg0, %arg1, %arg2) ({
+    ^bb1(%iv2: index):
       "omp.yield"() : () -> ()
+    }) : (index, index, index) -> ()
+    "omp.terminator"() : () -> ()
   }) {alignment_values = [32, 128],
-      operandSegmentSizes = array<i32: 1, 1, 1, 2, 0, 0>} : (index, index, index, memref<i32>, memref<i32>) -> ()
+      operandSegmentSizes = array<i32: 2, 0, 0>} : (memref<i32>, memref<i32>) -> ()
   return
 }
 
-// CHECK-LABEL: omp_simdloop_aligned_single
-func.func @omp_simdloop_aligned_single(%arg0 : index, %arg1 : index, %arg2 : index,
-                                       %arg3 : memref<i32>, %arg4 : memref<i32>) -> () {
-  // CHECK:      omp.simdloop   aligned(%{{.*}} : memref<i32> -> 32 : i64)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3) ({
-    ^bb0(%arg5: index):
+// CHECK-LABEL: omp_simd_aligned_single
+func.func @omp_simd_aligned_single(%arg0 : index, %arg1 : index, %arg2 : index,
+                                   %arg3 : memref<i32>, %arg4 : memref<i32>) -> () {
+  // CHECK: omp.simd aligned(%{{.*}} : memref<i32> -> 32 : i64)
+  "omp.simd"(%arg3) ({
+    "omp.loop_nest" (%arg0, %arg1, %arg2) ({
+    ^bb1(%iv2: index):
       "omp.yield"() : () -> ()
+    }) : (index, index, index) -> ()
+    "omp.terminator"() : () -> ()
   }) {alignment_values = [32],
-      operandSegmentSizes = array<i32: 1, 1, 1, 1, 0, 0>} : (index, index, index, memref<i32>) -> ()
+      operandSegmentSizes = array<i32: 1, 0, 0>} : (memref<i32>) -> ()
   return
 }
 
-// CHECK-LABEL: omp_simdloop_nontemporal_list
-func.func @omp_simdloop_nontemporal_list(%arg0 : index,
-                                         %arg1 : index,
-                                         %arg2 : index,
-                                         %arg3 : memref<i32>,
-                                         %arg4 : memref<i64>) -> () {
-  // CHECK:      omp.simdloop   nontemporal(%{{.*}}, %{{.*}} : memref<i32>, memref<i64>)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3, %arg4) ({
-    ^bb0(%arg5: index):
+// CHECK-LABEL: omp_simd_nontemporal_list
+func.func @omp_simd_nontemporal_list(%arg0 : index, %arg1 : index,
+                                     %arg2 : index, %arg3 : memref<i32>,
+                                     %arg4 : memref<i64>) -> () {
+  // CHECK: omp.simd nontemporal(%{{.*}}, %{{.*}} : memref<i32>, memref<i64>)
+  "omp.simd"(%arg3, %arg4) ({
+    "omp.loop_nest" (%arg0, %arg1, %arg2) ({
+    ^bb1(%iv2: index):
       "omp.yield"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 1, 1, 1, 0, 0, 2>} : (index, index, index, memref<i32>, memref<i64>) -> ()
+    }) : (index, index, index) -> ()
+    "omp.terminator"() : () -> ()
+  }) {operandSegmentSizes = array<i32: 0, 0, 2>} : (memref<i32>, memref<i64>) -> ()
   return
 }
 
-// CHECK-LABEL: omp_simdloop_nontemporal_single
-func.func @omp_simdloop_nontemporal_single(%arg0 : index,
-                                           %arg1 : index,
-                                           %arg2 : index,
-                                           %arg3 : memref<i32>,
-                                           %arg4 : memref<i64>) -> () {
-  // CHECK:      omp.simdloop   nontemporal(%{{.*}} : memref<i32>)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  "omp.simdloop"(%arg0, %arg1, %arg2, %arg3) ({
-    ^bb0(%arg5: index):
+// CHECK-LABEL: omp_simd_nontemporal_single
+func.func @omp_simd_nontemporal_single(%arg0 : index, %arg1 : index,
+                                       %arg2 : index, %arg3 : memref<i32>,
+                                       %arg4 : memref<i64>) -> () {
+  // CHECK: omp.simd nontemporal(%{{.*}} : memref<i32>)
+  "omp.simd"(%arg3) ({
+    "omp.loop_nest" (%arg0, %arg1, %arg2) ({
+    ^bb1(%iv2: index):
       "omp.yield"() : () -> ()
-  }) {operandSegmentSizes = array<i32: 1, 1, 1, 0, 0, 1>} : (index, index, index, memref<i32>) -> ()
+    }) : (index, index, index) -> ()
+    "omp.terminator"() : () -> ()
+  }) {operandSegmentSizes = array<i32: 0, 0, 1>} : (memref<i32>) -> ()
   return
 }
 
-// CHECK-LABEL: omp_simdloop_pretty
-func.func @omp_simdloop_pretty(%lb : index, %ub : index, %step : index) -> () {
-  // CHECK: omp.simdloop for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  omp.simdloop for (%iv) : index = (%lb) to (%ub) step (%step) {
-    omp.yield
+// CHECK-LABEL: omp_simd_pretty
+func.func @omp_simd_pretty(%lb : index, %ub : index, %step : index) -> () {
+  // CHECK: omp.simd {
+  omp.simd {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL:   func.func @omp_simdloop_pretty_aligned(
-func.func @omp_simdloop_pretty_aligned(%lb : index, %ub : index, %step : index,
-                                       %data_var : memref<i32>,
-                                       %data_var1 : memref<i32>) -> () {
-  // CHECK:      omp.simdloop   aligned(%{{.*}} : memref<i32> -> 32 : i64,
+// CHECK-LABEL:   func.func @omp_simd_pretty_aligned(
+func.func @omp_simd_pretty_aligned(%lb : index, %ub : index, %step : index,
+                                   %data_var : memref<i32>,
+                                   %data_var1 : memref<i32>) -> () {
+  // CHECK:      omp.simd aligned(
+  // CHECK-SAME: %{{.*}} : memref<i32> -> 32 : i64,
   // CHECK-SAME: %{{.*}} : memref<i32> -> 128 : i64)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  omp.simdloop aligned(%data_var :  memref<i32> -> 32, %data_var1 : memref<i32> -> 128)
-    for (%iv) : index = (%lb) to (%ub) step (%step) {
+  omp.simd aligned(%data_var :  memref<i32> -> 32, %data_var1 : memref<i32> -> 128) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
       omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL: omp_simdloop_pretty_if
-func.func @omp_simdloop_pretty_if(%lb : index, %ub : index, %step : index, %if_cond : i1) -> () {
-  // CHECK: omp.simdloop if(%{{.*}}) for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  omp.simdloop if(%if_cond) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+// CHECK-LABEL: omp_simd_pretty_if
+func.func @omp_simd_pretty_if(%lb : index, %ub : index, %step : index, %if_cond : i1) -> () {
+  // CHECK: omp.simd if(%{{.*}})
+  omp.simd if(%if_cond) {
+    omp.loop_nest (%iv): index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL:   func.func @omp_simdloop_pretty_nontemporal
-func.func @omp_simdloop_pretty_nontemporal(%lb : index,
-                                           %ub : index,
-                                           %step : index,
-                                           %data_var : memref<i32>,
-                                           %data_var1 : memref<i32>) -> () {
-  // CHECK:      omp.simdloop   nontemporal(%{{.*}}, %{{.*}} : memref<i32>, memref<i32>)
-  // CHECK-SAME: for  (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
-  omp.simdloop nontemporal(%data_var, %data_var1 : memref<i32>, memref<i32>)
-    for (%iv) : index = (%lb) to (%ub) step (%step) {
+// CHECK-LABEL: func.func @omp_simd_pretty_nontemporal
+func.func @omp_simd_pretty_nontemporal(%lb : index, %ub : index, %step : index,
+                                       %data_var : memref<i32>,
+                                       %data_var1 : memref<i32>) -> () {
+  // CHECK: omp.simd nontemporal(%{{.*}}, %{{.*}} : memref<i32>, memref<i32>)
+  omp.simd nontemporal(%data_var, %data_var1 : memref<i32>, memref<i32>) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
       omp.yield
-  }
-  return
-}
-// CHECK-LABEL: omp_simdloop_pretty_order
-func.func @omp_simdloop_pretty_order(%lb : index, %ub : index, %step : index) -> () {
-  // CHECK: omp.simdloop order(concurrent)
-  // CHECK-SAME: for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  omp.simdloop order(concurrent) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL: omp_simdloop_pretty_simdlen
-func.func @omp_simdloop_pretty_simdlen(%lb : index, %ub : index, %step : index) -> () {
-  // CHECK: omp.simdloop simdlen(2) for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  omp.simdloop simdlen(2) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+// CHECK-LABEL: omp_simd_pretty_order
+func.func @omp_simd_pretty_order(%lb : index, %ub : index, %step : index) -> () {
+  // CHECK: omp.simd order(concurrent)
+  omp.simd order(concurrent) {
+    omp.loop_nest (%iv): index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL: omp_simdloop_pretty_safelen
-func.func @omp_simdloop_pretty_safelen(%lb : index, %ub : index, %step : index) -> () {
-  // CHECK: omp.simdloop safelen(2) for (%{{.*}}) : index = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
-  omp.simdloop safelen(2) for (%iv): index = (%lb) to (%ub) step (%step) {
-    omp.yield
+// CHECK-LABEL: omp_simd_pretty_simdlen
+func.func @omp_simd_pretty_simdlen(%lb : index, %ub : index, %step : index) -> () {
+  // CHECK: omp.simd simdlen(2)
+  omp.simd simdlen(2) {
+    omp.loop_nest (%iv): index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
 
-// CHECK-LABEL: omp_simdloop_pretty_multiple
-func.func @omp_simdloop_pretty_multiple(%lb1 : index, %ub1 : index, %step1 : index, %lb2 : index, %ub2 : index, %step2 : index) -> () {
-  // CHECK: omp.simdloop for (%{{.*}}, %{{.*}}) : index = (%{{.*}}, %{{.*}}) to (%{{.*}}, %{{.*}}) step (%{{.*}}, %{{.*}})
-  omp.simdloop for (%iv1, %iv2) : index = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
-    omp.yield
+// CHECK-LABEL: omp_simd_pretty_safelen
+func.func @omp_simd_pretty_safelen(%lb : index, %ub : index, %step : index) -> () {
+  // CHECK: omp.simd safelen(2)
+  omp.simd safelen(2) {
+    omp.loop_nest (%iv): index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
   }
   return
 }
@@ -633,15 +640,13 @@ func.func @omp_distribute(%chunk_size : i32, %data_var : memref<i32>, %arg0 : i3
   }
   // CHECK: omp.distribute
   omp.distribute {
-    // TODO Remove induction variables from omp.simdloop.
-    omp.simdloop for (%iv) : i32 = (%arg0) to (%arg0) step (%arg0) {
+    omp.simd {
       omp.loop_nest (%iv2) : i32 = (%arg0) to (%arg0) step (%arg0) {
         omp.yield
       }
-      omp.yield
     }
   }
-return
+  return
 }
 
 
@@ -2170,14 +2175,11 @@ func.func @omp_taskloop(%lb: i32, %ub: i32, %step: i32) -> () {
 
   // CHECK: omp.taskloop {
   omp.taskloop {
-    // TODO Remove induction variables from omp.simdloop.
-    omp.simdloop for (%iv) : i32 = (%lb) to (%ub) step (%step) {
+    omp.simd {
       omp.loop_nest (%i, %j) : i32 = (%lb, %ub) to (%ub, %lb) step (%step, %step) {
         // CHECK: omp.yield
         omp.yield
       }
-      // CHECK: omp.yield
-      omp.yield
     }
   }
 
diff --git a/mlir/test/Target/LLVMIR/openmp-llvm.mlir b/mlir/test/Target/LLVMIR/openmp-llvm.mlir
index 4cb99c1f1a285..d1390022c1dc4 100644
--- a/mlir/test/Target/LLVMIR/openmp-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-llvm.mlir
@@ -638,10 +638,10 @@ llvm.func @test_omp_wsloop_guided_simd(%lb : i64, %ub : i64, %step : i64) -> ()
 
 // -----
 
-// CHECK-LABEL: @simdloop_simple
-llvm.func @simdloop_simple(%lb : i64, %ub : i64, %step : i64, %arg0: !llvm.ptr) {
-  "omp.simdloop" (%lb, %ub, %step) ({
-    ^bb0(%iv: i64):
+// CHECK-LABEL: @simd_simple
+llvm.func @simd_simple(%lb : i64, %ub : i64, %step : i64, %arg0: !llvm.ptr) {
+  "omp.simd" () ({
+    omp.loop_nest (%iv) : i64 = (%lb) to (%ub) step (%step) {
       %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
       // The form of the emitted IR is controlled by OpenMPIRBuilder and
       // tested there. Just check that the right metadata is added.
@@ -649,8 +649,9 @@ llvm.func @simdloop_simple(%lb : i64, %ub : i64, %step : i64, %arg0: !llvm.ptr)
       %4 = llvm.getelementptr %arg0[%iv] : (!llvm.ptr, i64) -> !llvm.ptr, f32
       llvm.store %3, %4 : f32, !llvm.ptr
       omp.yield
-  }) {operandSegmentSizes = array<i32: 1,1,1,0,0,0>} :
-    (i64, i64, i64) -> ()
+    }
+    "omp.terminator"() : () -> ()
+  }) : () -> ()
 
   llvm.return
 }
@@ -659,34 +660,36 @@ llvm.func @simdloop_simple(%lb : i64, %ub : i64, %step : i64, %arg0: !llvm.ptr)
 
 // -----
 
-// CHECK-LABEL: @simdloop_simple_multiple
-llvm.func @simdloop_simple_multiple(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
-  omp.simdloop for (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
-    %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
-    // The form of the emitted IR is controlled by OpenMPIRBuilder and
-    // tested there. Just check that the right metadata is added and collapsed
-    // loop bound is generated (Collapse clause is represented as a loop with
-    // list of indices, bounds and steps where the size of the list is equal
-    // to the collapse value.)
-    // CHECK: icmp slt i64
-    // CHECK-COUNT-3: select
-    // CHECK: %[[TRIPCOUNT0:.*]] = select
-    // CHECK: br label %[[PREHEADER:.*]]
-    // CHECK: [[PREHEADER]]:
-    // CHECK: icmp slt i64
-    // CHECK-COUNT-3: select
-    // CHECK: %[[TRIPCOUNT1:.*]] = select
-    // CHECK: mul nuw i64 %[[TRIPCOUNT0]], %[[TRIPCOUNT1]]
-    // CHECK: br label %[[COLLAPSED_PREHEADER:.*]]
-    // CHECK: [[COLLAPSED_PREHEADER]]:
-    // CHECK: br label %[[COLLAPSED_HEADER:.*]]
-    // CHECK: llvm.access.group
-    // CHECK-NEXT: llvm.access.group
-    %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    llvm.store %3, %4 : f32, !llvm.ptr
-    llvm.store %3, %5 : f32, !llvm.ptr
-    omp.yield
+// CHECK-LABEL: @simd_simple_multiple
+llvm.func @simd_simple_multiple(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
+  omp.simd {
+    omp.loop_nest (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
+      %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
+      // The form of the emitted IR is controlled by OpenMPIRBuilder and
+      // tested there. Just check that the right metadata is added and collapsed
+      // loop bound is generated (Collapse clause is represented as a loop with
+      // list of indices, bounds and steps where the size of the list is equal
+      // to the collapse value.)
+      // CHECK: icmp slt i64
+      // CHECK-COUNT-3: select
+      // CHECK: %[[TRIPCOUNT0:.*]] = select
+      // CHECK: br label %[[PREHEADER:.*]]
+      // CHECK: [[PREHEADER]]:
+      // CHECK: icmp slt i64
+      // CHECK-COUNT-3: select
+      // CHECK: %[[TRIPCOUNT1:.*]] = select
+      // CHECK: mul nuw i64 %[[TRIPCOUNT0]], %[[TRIPCOUNT1]]
+      // CHECK: br label %[[COLLAPSED_PREHEADER:.*]]
+      // CHECK: [[COLLAPSED_PREHEADER]]:
+      // CHECK: br label %[[COLLAPSED_HEADER:.*]]
+      // CHECK: llvm.access.group
+      // CHECK-NEXT: llvm.access.group
+      %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      llvm.store %3, %4 : f32, !llvm.ptr
+      llvm.store %3, %5 : f32, !llvm.ptr
+      omp.yield
+    }
   }
   llvm.return
 }
@@ -695,19 +698,21 @@ llvm.func @simdloop_simple_multiple(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 :
 
 // -----
 
-// CHECK-LABEL: @simdloop_simple_multiple_simdlen
-llvm.func @simdloop_simple_multiple_simdlen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
-  omp.simdloop simdlen(2) for (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
-    %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
-    // The form of the emitted IR is controlled by OpenMPIRBuilder and
-    // tested there. Just check that the right metadata is added.
-    // CHECK: llvm.access.group
-    // CHECK-NEXT: llvm.access.group
-    %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    llvm.store %3, %4 : f32, !llvm.ptr
-    llvm.store %3, %5 : f32, !llvm.ptr
-    omp.yield
+// CHECK-LABEL: @simd_simple_multiple_simdlen
+llvm.func @simd_simple_multiple_simdlen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
+  omp.simd simdlen(2) {
+    omp.loop_nest (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
+      %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
+      // The form of the emitted IR is controlled by OpenMPIRBuilder and
+      // tested there. Just check that the right metadata is added.
+      // CHECK: llvm.access.group
+      // CHECK-NEXT: llvm.access.group
+      %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      llvm.store %3, %4 : f32, !llvm.ptr
+      llvm.store %3, %5 : f32, !llvm.ptr
+      omp.yield
+    }
   }
   llvm.return
 }
@@ -717,15 +722,17 @@ llvm.func @simdloop_simple_multiple_simdlen(%lb1 : i64, %ub1 : i64, %step1 : i64
 
 // -----
 
-// CHECK-LABEL: @simdloop_simple_multiple_safelen
-llvm.func @simdloop_simple_multiple_safelen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
-  omp.simdloop safelen(2) for (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
-    %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
-    %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    llvm.store %3, %4 : f32, !llvm.ptr
-    llvm.store %3, %5 : f32, !llvm.ptr
-    omp.yield
+// CHECK-LABEL: @simd_simple_multiple_safelen
+llvm.func @simd_simple_multiple_safelen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
+  omp.simd safelen(2) {
+    omp.loop_nest (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
+      %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
+      %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      llvm.store %3, %4 : f32, !llvm.ptr
+      llvm.store %3, %5 : f32, !llvm.ptr
+      omp.yield
+    }
   }
   llvm.return
 }
@@ -734,15 +741,17 @@ llvm.func @simdloop_simple_multiple_safelen(%lb1 : i64, %ub1 : i64, %step1 : i64
 
 // -----
 
-// CHECK-LABEL: @simdloop_simple_multiple_simdlen_safelen
-llvm.func @simdloop_simple_multiple_simdlen_safelen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
-  omp.simdloop simdlen(1) safelen(2) for (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
-    %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
-    %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
-    llvm.store %3, %4 : f32, !llvm.ptr
-    llvm.store %3, %5 : f32, !llvm.ptr
-    omp.yield
+// CHECK-LABEL: @simd_simple_multiple_simdlen_safelen
+llvm.func @simd_simple_multiple_simdlen_safelen(%lb1 : i64, %ub1 : i64, %step1 : i64, %lb2 : i64, %ub2 : i64, %step2 : i64, %arg0: !llvm.ptr, %arg1: !llvm.ptr) {
+  omp.simd simdlen(1) safelen(2) {
+    omp.loop_nest (%iv1, %iv2) : i64 = (%lb1, %lb2) to (%ub1, %ub2) step (%step1, %step2) {
+      %3 = llvm.mlir.constant(2.000000e+00 : f32) : f32
+      %4 = llvm.getelementptr %arg0[%iv1] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      %5 = llvm.getelementptr %arg1[%iv2] : (!llvm.ptr, i64) -> !llvm.ptr, f32
+      llvm.store %3, %4 : f32, !llvm.ptr
+      llvm.store %3, %5 : f32, !llvm.ptr
+      omp.yield
+    }
   }
   llvm.return
 }
@@ -751,8 +760,8 @@ llvm.func @simdloop_simple_multiple_simdlen_safelen(%lb1 : i64, %ub1 : i64, %ste
 
 // -----
 
-// CHECK-LABEL: @simdloop_if
-llvm.func @simdloop_if(%arg0: !llvm.ptr {fir.bindc_name = "n"}, %arg1: !llvm.ptr {fir.bindc_name = "threshold"}) {
+// CHECK-LABEL: @simd_if
+llvm.func @simd_if(%arg0: !llvm.ptr {fir.bindc_name = "n"}, %arg1: !llvm.ptr {fir.bindc_name = "threshold"}) {
   %0 = llvm.mlir.constant(1 : i64) : i64
   %1 = llvm.alloca %0 x i32 {adapt.valuebyref, in_type = i32, operandSegmentSizes = array<i32: 0, 0>} : (i64) -> !llvm.ptr
   %2 = llvm.mlir.constant(1 : i64) : i64
@@ -763,12 +772,14 @@ llvm.func @simdloop_if(%arg0: !llvm.ptr {fir.bindc_name = "n"}, %arg1: !llvm.ptr
   %7 = llvm.load %arg0 : !llvm.ptr -> i32
   %8 = llvm.load %arg1 : !llvm.ptr -> i32
   %9 = llvm.icmp "sge" %7, %8 : i32
-  omp.simdloop   if(%9) for  (%arg2) : i32 = (%4) to (%5) inclusive step (%6) {
-    // The form of the emitted IR is controlled by OpenMPIRBuilder and
-    // tested there. Just check that the right metadata is added.
-    // CHECK: llvm.access.group
-    llvm.store %arg2, %1 : i32, !llvm.ptr
-    omp.yield
+  omp.simd if(%9) {
+    omp.loop_nest (%arg2) : i32 = (%4) to (%5) inclusive step (%6) {
+      // The form of the emitted IR is controlled by OpenMPIRBuilder and
+      // tested there. Just check that the right metadata is added.
+      // CHECK: llvm.access.group
+      llvm.store %arg2, %1 : i32, !llvm.ptr
+      omp.yield
+    }
   }
   llvm.return
 }

From 16b0be613205a37519e2bfec1a904ceaa89636e7 Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Wed, 17 Apr 2024 11:30:11 +0100
Subject: [PATCH 722/886] [MLIR][OpenMP] NFC: Remove LoopControl
 parsing/printing code (#88909)

This patch removes the LoopControl parsing/printing functions that are
no longer used after transitioning `omp.simdloop` and `omp.taskloop`
into loop wrapper operations.
---
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp | 52 --------------------
 1 file changed, 52 deletions(-)

diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index caf0ac3f86017..5d2281ce6094f 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1548,58 +1548,6 @@ void printWsloop(OpAsmPrinter &p, Operation *op, Region &region,
   p.printRegion(region, /*printEntryBlockArgs=*/false);
 }
 
-/// loop-control ::= `(` ssa-id-list `)` `:` type `=`  loop-bounds
-/// loop-bounds := `(` ssa-id-list `)` to `(` ssa-id-list `)` inclusive? steps
-/// steps := `step` `(`ssa-id-list`)`
-ParseResult
-parseLoopControl(OpAsmParser &parser, Region &region,
-                 SmallVectorImpl<OpAsmParser::UnresolvedOperand> &lowerBound,
-                 SmallVectorImpl<OpAsmParser::UnresolvedOperand> &upperBound,
-                 SmallVectorImpl<OpAsmParser::UnresolvedOperand> &steps,
-                 SmallVectorImpl<Type> &loopVarTypes, UnitAttr &inclusive) {
-  // Parse an opening `(` followed by induction variables followed by `)`
-  SmallVector<OpAsmParser::Argument> ivs;
-  Type loopVarType;
-  if (parser.parseArgumentList(ivs, OpAsmParser::Delimiter::Paren) ||
-      parser.parseColonType(loopVarType) ||
-      // Parse loop bounds.
-      parser.parseEqual() ||
-      parser.parseOperandList(lowerBound, ivs.size(),
-                              OpAsmParser::Delimiter::Paren) ||
-      parser.parseKeyword("to") ||
-      parser.parseOperandList(upperBound, ivs.size(),
-                              OpAsmParser::Delimiter::Paren))
-    return failure();
-
-  if (succeeded(parser.parseOptionalKeyword("inclusive")))
-    inclusive = UnitAttr::get(parser.getBuilder().getContext());
-
-  // Parse step values.
-  if (parser.parseKeyword("step") ||
-      parser.parseOperandList(steps, ivs.size(), OpAsmParser::Delimiter::Paren))
-    return failure();
-
-  // Now parse the body.
-  loopVarTypes = SmallVector<Type>(ivs.size(), loopVarType);
-  for (auto &iv : ivs)
-    iv.type = loopVarType;
-
-  return parser.parseRegion(region, ivs);
-}
-
-void printLoopControl(OpAsmPrinter &p, Operation *op, Region &region,
-                      ValueRange lowerBound, ValueRange upperBound,
-                      ValueRange steps, TypeRange loopVarTypes,
-                      UnitAttr inclusive) {
-  auto args = region.front().getArguments();
-  p << " (" << args << ") : " << args[0].getType() << " = (" << lowerBound
-    << ") to (" << upperBound << ") ";
-  if (inclusive)
-    p << "inclusive ";
-  p << "step (" << steps << ") ";
-  p.printRegion(region, /*printEntryBlockArgs=*/false);
-}
-
 //===----------------------------------------------------------------------===//
 // Simd construct [2.9.3.1]
 //===----------------------------------------------------------------------===//

From a02019960b1a693320cd43b0ed6653d95877b94f Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Wed, 17 Apr 2024 18:15:26 +0800
Subject: [PATCH 723/886] [RISCV] Assert only valid AVLs in doLocalPostpass are
 X0 or virtual regs. NFC

In vxrm.mir we were running RISCVInsertVSETVLI on pseudos that already had
vsetvlis inserted and their AVLs set to $noreg. (This happened to work
since doLocalPostpass got rid of the extra vsetvli)

This removes the vsetvlis from the test and enforces that the only valid
AVLs we work with are either X0 or virtual registers (or $noreg before
emitVSETVLIs), since we don't handle physical registers properly in
doLocalPostpass.
---
 llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp | 16 +++++++---------
 llvm/test/CodeGen/RISCV/rvv/vxrm.mir         |  8 ++++----
 2 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
index a54a1148cf28b..6e45f0c703ceb 100644
--- a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
@@ -468,6 +468,7 @@ class VSETVLIInfo {
   bool isUnknown() const { return State == Unknown; }
 
   void setAVLReg(Register Reg) {
+    assert(Reg.isVirtual() || Reg == RISCV::X0 || Reg == RISCV::NoRegister);
     AVLReg = Reg;
     State = AVLIsReg;
   }
@@ -1514,12 +1515,9 @@ static bool canMutatePriorConfig(const MachineInstr &PrevMI,
 
     // If the AVL is a register, we need to make sure MI's AVL dominates PrevMI.
     // For now just check that PrevMI uses the same virtual register.
-    if (AVL.isReg() && AVL.getReg() != RISCV::X0) {
-      if (AVL.getReg().isPhysical())
-        return false;
-      if (!PrevAVL.isReg() || PrevAVL.getReg() != AVL.getReg())
-        return false;
-    }
+    if (AVL.isReg() && AVL.getReg() != RISCV::X0 &&
+        (!PrevAVL.isReg() || PrevAVL.getReg() != AVL.getReg()))
+      return false;
   }
 
   assert(PrevMI.getOperand(2).isImm() && MI.getOperand(2).isImm());
@@ -1543,9 +1541,9 @@ void RISCVInsertVSETVLI::doLocalPostpass(MachineBasicBlock &MBB) {
       continue;
     }
 
-    Register VRegDef = MI.getOperand(0).getReg();
-    if (VRegDef != RISCV::X0 &&
-        !(VRegDef.isVirtual() && MRI->use_nodbg_empty(VRegDef)))
+    Register RegDef = MI.getOperand(0).getReg();
+    assert(RegDef == RISCV::X0 || RegDef.isVirtual());
+    if (RegDef != RISCV::X0 && !MRI->use_nodbg_empty(RegDef))
       Used.demandVL();
 
     if (NextMI) {
diff --git a/llvm/test/CodeGen/RISCV/rvv/vxrm.mir b/llvm/test/CodeGen/RISCV/rvv/vxrm.mir
index 64e191887e092..a588677bec8e2 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vxrm.mir
+++ b/llvm/test/CodeGen/RISCV/rvv/vxrm.mir
@@ -11,9 +11,9 @@ body:     |
     ; MIR-LABEL: name: verify_vxrm
     ; MIR: liveins: $v8, $v9, $x10
     ; MIR-NEXT: {{  $}}
-    ; MIR-NEXT: dead $x0 = PseudoVSETVLI renamable $x10, 197 /* e8, mf8, ta, ma */, implicit-def $vl, implicit-def $vtype
+    ; MIR-NEXT: dead $x0 = PseudoVSETVLI killed renamable $x10, 197 /* e8, mf8, ta, ma */, implicit-def $vl, implicit-def $vtype
     ; MIR-NEXT: WriteVXRMImm 0, implicit-def $vxrm
-    ; MIR-NEXT: renamable $v8 = PseudoVAADD_VV_MF8 undef $v8, renamable $v8, renamable $v9, 0, $noreg, 3 /* e8 */, 0  /* tu, mu */, implicit $vl, implicit $vtype, implicit $vxrm
+    ; MIR-NEXT: renamable $v8 = PseudoVAADD_VV_MF8 undef $v8, killed renamable $v8, killed renamable $v9, 0, $noreg, 3 /* e8 */, 0  /* tu, mu */, implicit $vl, implicit $vtype, implicit $vxrm
     ; MIR-NEXT: PseudoRET implicit $v8
     ; ASM-LABEL: verify_vxrm:
     ; ASM:        # %bb.0:
@@ -23,8 +23,8 @@ body:     |
     ; ASM-NEXT:    ret
     %0:vr = COPY $v8
     %1:vr = COPY $v9
-    dead $x0 = PseudoVSETVLI killed renamable $x10, 197 /* e8, mf8, ta, ma */, implicit-def $vl, implicit-def $vtype
+    %2:gprnox0 = COPY $x10
     %pt:vr = IMPLICIT_DEF
-    renamable $v8 = PseudoVAADD_VV_MF8 %pt, killed renamable $v8, killed renamable $v9, 0, $noreg, 3 /* e8 */, 0
+    renamable $v8 = PseudoVAADD_VV_MF8 %pt, %0, %1, 0, %2, 3 /* e8 */, 0
     PseudoRET implicit $v8
 ...

From a634f3ef39c0c547b87f1ee4ebe02ee3a256587f Mon Sep 17 00:00:00 2001
From: Stephen Tozer <stephen.tozer@sony.com>
Date: Wed, 17 Apr 2024 11:39:18 +0100
Subject: [PATCH 724/886] [RemoveDIs] Update update_test_checks script to
 recognize dbg_records (#87388)

As we've added new IR elements for the RemoveDIs project,
we need the update_test_checks script to understand them. For the
records themselves this is already done automatically, but their
metadata arguments are not recognized as such due to lacking the
`metadata` prefix, which means they won't be checked by the script. This
patch fixes this by adding a check for all `![0-9]+` patterns as long as
they are not at the start of a line (which avoids matching global
values).
---
 .../Inputs/various_ir_values_dbgrecords.ll    | 168 ++++++++++
 .../various_ir_values_dbgrecords.ll.expected  | 238 ++++++++++++++
 ...s_ir_values_dbgrecords.ll.funcsig.expected | 240 ++++++++++++++
 ...ues_dbgrecords.ll.funcsig.globals.expected | 309 ++++++++++++++++++
 ...s_dbgrecords.ll.funcsig.noglobals.expected | 238 ++++++++++++++
 ...ords.ll.funcsig.transitiveglobals.expected | 299 +++++++++++++++++
 .../various_ir_values_dbgrecords.test         |  24 ++
 llvm/utils/UpdateTestChecks/common.py         |   2 +-
 8 files changed, 1517 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.noglobals.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.transitiveglobals.expected
 create mode 100644 llvm/test/tools/UpdateTestChecks/update_test_checks/various_ir_values_dbgrecords.test

diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll
new file mode 100644
index 0000000000000..9a9cc0a06936f
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll
@@ -0,0 +1,168 @@
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+    #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+    #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+    #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+    #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+    #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.expected
new file mode 100644
index 0000000000000..1f9c37ccfbd88
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.expected
@@ -0,0 +1,238 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, !DIAssignID [[DIASSIGNID16:![0-9]+]]
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:      #dbg_assign(i1 undef, [[META13:![0-9]+]], !DIExpression(), [[DIASSIGNID16]], ptr [[A_ADDR]], !DIExpression(), [[META17:![0-9]+]])
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META13]], !DIExpression(), [[META17]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2:[0-9]+]], !dbg [[DBG22:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META14:![0-9]+]], !DIExpression(), [[META23:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META23]], !tbaa [[TBAA24:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG22]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG26:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG28:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG29:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG30:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG31:![0-9]+]], !prof [[PROF32:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG33:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG34:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG35:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG34]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG34]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG36:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG34]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG37:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG37]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG37]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG33]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+  #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+  #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META43:![0-9]+]], !DIExpression(), [[META46:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG47:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META44:![0-9]+]], !DIExpression(), [[META48:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META48]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG47]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG49:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG51:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG52:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG53:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG54:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG55:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG56:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG57:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG56]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG56]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG58:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG56]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG59:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG59]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG59]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG55]], !llvm.loop [[LOOP60:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG62:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+  #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected
new file mode 100644
index 0000000000000..5905e443deff2
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected
@@ -0,0 +1,240 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+; CHECK-LABEL: define {{[^@]+}}@foo
+; CHECK-SAME: (ptr [[A:%.*]]) #[[ATTR0:[0-9]+]] !dbg [[DBG7:![0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, !DIAssignID [[DIASSIGNID16:![0-9]+]]
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:      #dbg_assign(i1 undef, [[META13:![0-9]+]], !DIExpression(), [[DIASSIGNID16]], ptr [[A_ADDR]], !DIExpression(), [[META17:![0-9]+]])
+; CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META13]], !DIExpression(), [[META17]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2:[0-9]+]], !dbg [[DBG22:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META14:![0-9]+]], !DIExpression(), [[META23:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META23]], !tbaa [[TBAA24:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG22]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG26:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG28:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG29:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG30:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG31:![0-9]+]], !prof [[PROF32:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG33:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG34:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG35:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG34]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG34]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG36:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG34]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG37:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG37]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG37]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG33]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+  #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+  #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+; CHECK-LABEL: define {{[^@]+}}@bar
+; CHECK-SAME: (ptr [[A:%.*]]) #[[ATTR0]] !dbg [[DBG41:![0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META43:![0-9]+]], !DIExpression(), [[META46:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG47:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META44:![0-9]+]], !DIExpression(), [[META48:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META48]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG47]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG49:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG51:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG52:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG53:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG54:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG55:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG56:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG57:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG56]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG56]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG58:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG56]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG59:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG59]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG59]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG55]], !llvm.loop [[LOOP60:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG62:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+  #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected
new file mode 100644
index 0000000000000..579d6a437d0e5
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected
@@ -0,0 +1,309 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+; CHECK-LABEL: define {{[^@]+}}@foo
+; CHECK-SAME: (ptr [[A:%.*]]) #[[ATTR0:[0-9]+]] !dbg [[DBG7:![0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, !DIAssignID [[DIASSIGNID16:![0-9]+]]
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:      #dbg_assign(i1 undef, [[META13:![0-9]+]], !DIExpression(), [[DIASSIGNID16]], ptr [[A_ADDR]], !DIExpression(), [[META17:![0-9]+]])
+; CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META13]], !DIExpression(), [[META17]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2:[0-9]+]], !dbg [[DBG22:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META14:![0-9]+]], !DIExpression(), [[META23:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META23]], !tbaa [[TBAA24:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG22]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG26:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG28:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG29:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG30:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG31:![0-9]+]], !prof [[PROF32:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG33:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG34:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG35:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG34]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG34]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG36:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG34]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG37:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG37]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG37]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG33]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+  #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+  #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+; CHECK-LABEL: define {{[^@]+}}@bar
+; CHECK-SAME: (ptr [[A:%.*]]) #[[ATTR0]] !dbg [[DBG41:![0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META43:![0-9]+]], !DIExpression(), [[META46:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG47:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META44:![0-9]+]], !DIExpression(), [[META48:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META48]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG47]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG49:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG51:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG52:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG53:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG54:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG55:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG56:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG57:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG56]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG56]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG58:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG56]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG59:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG59]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG59]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG55]], !llvm.loop [[LOOP60:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG62:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+  #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
+;.
+; CHECK: attributes #[[ATTR0]] = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+; CHECK: attributes #[[ATTR1:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+; CHECK: attributes #[[ATTR2]] = { nounwind }
+;.
+; CHECK: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C99, file: [[META1:![0-9]+]], producer: "{{.*}}clang version {{.*}}", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: [[META2:![0-9]+]], splitDebugInlining: false, nameTableKind: None)
+; CHECK: [[META1]] = !DIFile(filename: "various_ir_values.c", directory: {{.*}})
+; CHECK: [[META2]] = !{}
+; CHECK: [[META3:![0-9]+]] = !{i32 7, !"Dwarf Version", i32 4}
+; CHECK: [[META4:![0-9]+]] = !{i32 2, !"Debug Info Version", i32 3}
+; CHECK: [[META5:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}
+; CHECK: [[META6:![0-9]+]] = !{!"{{.*}}clang version {{.*}}"}
+; CHECK: [[DBG7]] = distinct !DISubprogram(name: "foo", scope: [[META1]], file: [[META1]], line: 1, type: [[META8:![0-9]+]], scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META12:![0-9]+]])
+; CHECK: [[META8]] = !DISubroutineType(types: [[META9:![0-9]+]])
+; CHECK: [[META9]] = !{null, [[META10:![0-9]+]]}
+; CHECK: [[META10]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META11:![0-9]+]], size: 64)
+; CHECK: [[META11]] = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+; CHECK: [[META12]] = !{[[META13]], [[META14]]}
+; CHECK: [[META13]] = !DILocalVariable(name: "A", arg: 1, scope: [[DBG7]], file: [[META1]], line: 1, type: [[META10]])
+; CHECK: [[META14]] = !DILocalVariable(name: "i", scope: [[META15:![0-9]+]], file: [[META1]], line: 3, type: [[META11]])
+; CHECK: [[META15]] = distinct !DILexicalBlock(scope: [[DBG7]], file: [[META1]], line: 3, column: 3)
+; CHECK: [[DIASSIGNID16]] = distinct !DIAssignID()
+; CHECK: [[META17]] = !DILocation(line: 1, column: 15, scope: [[DBG7]])
+; CHECK: [[TBAA18]] = !{[[META19:![0-9]+]], [[META19]], i64 0}
+; CHECK: [[META19]] = !{!"any pointer", [[META20:![0-9]+]], i64 0}
+; CHECK: [[META20]] = !{!"omnipotent char", [[META21:![0-9]+]], i64 0}
+; CHECK: [[META21]] = !{!"Simple C/C++ TBAA"}
+; CHECK: [[DBG22]] = !DILocation(line: 3, column: 8, scope: [[META15]])
+; CHECK: [[META23]] = !DILocation(line: 3, column: 12, scope: [[META15]])
+; CHECK: [[TBAA24]] = !{[[META25:![0-9]+]], [[META25]], i64 0}
+; CHECK: [[META25]] = !{!"int", [[META20]], i64 0}
+; CHECK: [[DBG26]] = !DILocation(line: 3, column: 19, scope: [[META27:![0-9]+]])
+; CHECK: [[META27]] = distinct !DILexicalBlock(scope: [[META15]], file: [[META1]], line: 3, column: 3)
+; CHECK: [[DBG28]] = !DILocation(line: 3, column: 24, scope: [[META27]])
+; CHECK: [[DBG29]] = !DILocation(line: 3, column: 23, scope: [[META27]])
+; CHECK: [[DBG30]] = !DILocation(line: 3, column: 21, scope: [[META27]])
+; CHECK: [[DBG31]] = !DILocation(line: 3, column: 3, scope: [[META15]])
+; CHECK: [[PROF32]] = !{!"branch_weights", i32 1, i32 1048575}
+; CHECK: [[DBG33]] = !DILocation(line: 3, column: 3, scope: [[META27]])
+; CHECK: [[DBG34]] = !DILocation(line: 4, column: 5, scope: [[META27]])
+; CHECK: [[DBG35]] = !DILocation(line: 4, column: 7, scope: [[META27]])
+; CHECK: [[DBG36]] = !DILocation(line: 4, column: 10, scope: [[META27]])
+; CHECK: [[DBG37]] = !DILocation(line: 3, column: 27, scope: [[META27]])
+; CHECK: [[LOOP38]] = distinct !{[[LOOP38]], [[DBG31]], [[META39:![0-9]+]]}
+; CHECK: [[META39]] = !DILocation(line: 4, column: 12, scope: [[META15]])
+; CHECK: [[DBG40]] = !DILocation(line: 5, column: 1, scope: [[DBG7]])
+; CHECK: [[DBG41]] = distinct !DISubprogram(name: "bar", scope: [[META1]], file: [[META1]], line: 7, type: [[META8]], scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META42:![0-9]+]])
+; CHECK: [[META42]] = !{[[META43]], [[META44]]}
+; CHECK: [[META43]] = !DILocalVariable(name: "A", arg: 1, scope: [[DBG41]], file: [[META1]], line: 7, type: [[META10]])
+; CHECK: [[META44]] = !DILocalVariable(name: "i", scope: [[META45:![0-9]+]], file: [[META1]], line: 9, type: [[META11]])
+; CHECK: [[META45]] = distinct !DILexicalBlock(scope: [[DBG41]], file: [[META1]], line: 9, column: 3)
+; CHECK: [[META46]] = !DILocation(line: 7, column: 15, scope: [[DBG41]])
+; CHECK: [[DBG47]] = !DILocation(line: 9, column: 8, scope: [[META45]])
+; CHECK: [[META48]] = !DILocation(line: 9, column: 12, scope: [[META45]])
+; CHECK: [[DBG49]] = !DILocation(line: 9, column: 19, scope: [[META50:![0-9]+]])
+; CHECK: [[META50]] = distinct !DILexicalBlock(scope: [[META45]], file: [[META1]], line: 9, column: 3)
+; CHECK: [[DBG51]] = !DILocation(line: 9, column: 24, scope: [[META50]])
+; CHECK: [[DBG52]] = !DILocation(line: 9, column: 23, scope: [[META50]])
+; CHECK: [[DBG53]] = !DILocation(line: 9, column: 21, scope: [[META50]])
+; CHECK: [[DBG54]] = !DILocation(line: 9, column: 3, scope: [[META45]])
+; CHECK: [[DBG55]] = !DILocation(line: 9, column: 3, scope: [[META50]])
+; CHECK: [[DBG56]] = !DILocation(line: 10, column: 5, scope: [[META50]])
+; CHECK: [[DBG57]] = !DILocation(line: 10, column: 7, scope: [[META50]])
+; CHECK: [[DBG58]] = !DILocation(line: 10, column: 10, scope: [[META50]])
+; CHECK: [[DBG59]] = !DILocation(line: 9, column: 27, scope: [[META50]])
+; CHECK: [[LOOP60]] = distinct !{[[LOOP60]], [[DBG54]], [[META61:![0-9]+]]}
+; CHECK: [[META61]] = !DILocation(line: 10, column: 12, scope: [[META45]])
+; CHECK: [[DBG62]] = !DILocation(line: 11, column: 1, scope: [[DBG41]])
+;.
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.noglobals.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.noglobals.expected
new file mode 100644
index 0000000000000..1f9c37ccfbd88
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.noglobals.expected
@@ -0,0 +1,238 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, !DIAssignID [[DIASSIGNID16:![0-9]+]]
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:      #dbg_assign(i1 undef, [[META13:![0-9]+]], !DIExpression(), [[DIASSIGNID16]], ptr [[A_ADDR]], !DIExpression(), [[META17:![0-9]+]])
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META13]], !DIExpression(), [[META17]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2:[0-9]+]], !dbg [[DBG22:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META14:![0-9]+]], !DIExpression(), [[META23:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META23]], !tbaa [[TBAA24:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG22]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG26:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG28:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG29:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG30:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG31:![0-9]+]], !prof [[PROF32:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG33:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG34:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG35:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG34]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG34]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG36:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG34]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG37:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG37]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG37]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG33]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+  #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+  #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META43:![0-9]+]], !DIExpression(), [[META46:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG47:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META44:![0-9]+]], !DIExpression(), [[META48:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META48]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG47]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG49:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG51:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG52:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG53:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG54:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG55:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG56:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG57:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG56]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG56]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG58:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG56]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG59:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG59]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG59]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG55]], !llvm.loop [[LOOP60:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG62:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+  #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.transitiveglobals.expected b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.transitiveglobals.expected
new file mode 100644
index 0000000000000..e2c426029a6b0
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/Inputs/various_ir_values_dbgrecords.ll.funcsig.transitiveglobals.expected
@@ -0,0 +1,299 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals smart
+; Just run it through opt, no passes needed.
+; RUN: opt < %s -S --write-experimental-debuginfo=true | FileCheck %s
+
+; ModuleID = 'various_ir_values.c'
+source_filename = "various_ir_values.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Function Attrs: nounwind uwtable
+define dso_local void @foo(ptr %A) #0 !dbg !7 {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8, !DIAssignID [[DIASSIGNID16:![0-9]+]]
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:      #dbg_assign(i1 undef, [[META13:![0-9]+]], !DIExpression(), [[DIASSIGNID16]], ptr [[A_ADDR]], !DIExpression(), [[META17:![0-9]+]])
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META13]], !DIExpression(), [[META17]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2:[0-9]+]], !dbg [[DBG22:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META14:![0-9]+]], !DIExpression(), [[META23:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META23]], !tbaa [[TBAA24:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG22]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG26:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG28:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG29:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG30:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG31:![0-9]+]], !prof [[PROF32:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG33:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG34:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG35:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG34]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG34]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG36:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG34]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG37:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG37]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG37]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG33]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8, !DIAssignID !16
+  %i = alloca i32, align 4
+  #dbg_assign(i1 undef, !13, !DIExpression(), !16, ptr %A.addr, !DIExpression(), !17)
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !13, !DIExpression(), !17)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !22
+  #dbg_declare(ptr %i, !14, !DIExpression(), !23)
+  store i32 0, ptr %i, align 4, !dbg !23, !tbaa !24
+  br label %for.cond, !dbg !22
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !26, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !28, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !29, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !30
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !31, !prof !32
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !33
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !34, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !35, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !34
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !34
+  store i32 0, ptr %arrayidx, align 4, !dbg !36, !tbaa !24
+  br label %for.inc, !dbg !34
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !37, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !37
+  store i32 %inc, ptr %i, align 4, !dbg !37, !tbaa !24
+  br label %for.cond, !dbg !33, !llvm.loop !38
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !40
+}
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
+
+; Function Attrs: nounwind uwtable
+define dso_local void @bar(ptr %A) #0 !dbg !41 {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
+; CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT:    store ptr [[A:%.*]], ptr [[A_ADDR]], align 8, !tbaa [[TBAA18]]
+; CHECK-NEXT:      #dbg_declare(ptr [[A_ADDR]], [[META43:![0-9]+]], !DIExpression(), [[META46:![0-9]+]])
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG47:![0-9]+]]
+; CHECK-NEXT:      #dbg_declare(ptr [[I]], [[META44:![0-9]+]], !DIExpression(), [[META48:![0-9]+]])
+; CHECK-NEXT:    store i32 0, ptr [[I]], align 4, !dbg [[META48]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND:%.*]], !dbg [[DBG47]]
+; CHECK:       for.cond:
+; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG49:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG51:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4, !dbg [[DBG52:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], [[TMP2]], !dbg [[DBG53:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_COND_CLEANUP:%.*]], !dbg [[DBG54:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4, ptr [[I]]) #[[ATTR2]], !dbg [[DBG55:![0-9]+]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 8, !dbg [[DBG56:![0-9]+]], !tbaa [[TBAA18]]
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG57:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64, !dbg [[DBG56]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 [[IDXPROM]], !dbg [[DBG56]]
+; CHECK-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !dbg [[DBG58:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_INC:%.*]], !dbg [[DBG56]]
+; CHECK:       for.inc:
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !dbg [[DBG59:![0-9]+]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP5]], 1, !dbg [[DBG59]]
+; CHECK-NEXT:    store i32 [[INC]], ptr [[I]], align 4, !dbg [[DBG59]], !tbaa [[TBAA24]]
+; CHECK-NEXT:    br label [[FOR_COND]], !dbg [[DBG55]], !llvm.loop [[LOOP60:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void, !dbg [[DBG62:![0-9]+]]
+;
+entry:
+  %A.addr = alloca ptr, align 8
+  %i = alloca i32, align 4
+  store ptr %A, ptr %A.addr, align 8, !tbaa !18
+  #dbg_declare(ptr %A.addr, !43, !DIExpression(), !46)
+  call void @llvm.lifetime.start.p0(i64 4, ptr %i) #2, !dbg !47
+  #dbg_declare(ptr %i, !44, !DIExpression(), !48)
+  store i32 0, ptr %i, align 4, !dbg !48, !tbaa !24
+  br label %for.cond, !dbg !47
+
+for.cond:                                         ; preds = %for.inc, %entry
+  %0 = load i32, ptr %i, align 4, !dbg !49, !tbaa !24
+  %1 = load ptr, ptr %A.addr, align 8, !dbg !51, !tbaa !18
+  %2 = load i32, ptr %1, align 4, !dbg !52, !tbaa !24
+  %cmp = icmp slt i32 %0, %2, !dbg !53
+  br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !54
+
+for.cond.cleanup:                                 ; preds = %for.cond
+  call void @llvm.lifetime.end.p0(i64 4, ptr %i) #2, !dbg !55
+  br label %for.end
+
+for.body:                                         ; preds = %for.cond
+  %3 = load ptr, ptr %A.addr, align 8, !dbg !56, !tbaa !18
+  %4 = load i32, ptr %i, align 4, !dbg !57, !tbaa !24
+  %idxprom = sext i32 %4 to i64, !dbg !56
+  %arrayidx = getelementptr inbounds i32, ptr %3, i64 %idxprom, !dbg !56
+  store i32 0, ptr %arrayidx, align 4, !dbg !58, !tbaa !24
+  br label %for.inc, !dbg !56
+
+for.inc:                                          ; preds = %for.body
+  %5 = load i32, ptr %i, align 4, !dbg !59, !tbaa !24
+  %inc = add nsw i32 %5, 1, !dbg !59
+  store i32 %inc, ptr %i, align 4, !dbg !59, !tbaa !24
+  br label %for.cond, !dbg !55, !llvm.loop !60
+
+for.end:                                          ; preds = %for.cond.cleanup
+  ret void, !dbg !62
+}
+
+attributes #0 = { nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "denormal-fp-math"="ieee,ieee" "denormal-fp-math-f32"="ieee,ieee" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #2 = { nounwind }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "various_ir_values.c", directory: "/data/build/llvm-project")
+!2 = !{}
+!3 = !{i32 7, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!6 = !{!"clang version 11.0.0 (git@github.com:llvm/llvm-project.git 1d5da8cd30fce1c0a2c2fa6ba656dbfaa36192c8)"}
+!7 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 1, type: !8, scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!8 = !DISubroutineType(types: !9)
+!9 = !{null, !10}
+!10 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13, !14}
+!13 = !DILocalVariable(name: "A", arg: 1, scope: !7, file: !1, line: 1, type: !10)
+!14 = !DILocalVariable(name: "i", scope: !15, file: !1, line: 3, type: !11)
+!15 = distinct !DILexicalBlock(scope: !7, file: !1, line: 3, column: 3)
+!16 = distinct !DIAssignID()
+!17 = !DILocation(line: 1, column: 15, scope: !7)
+!18 = !{!19, !19, i64 0}
+!19 = !{!"any pointer", !20, i64 0}
+!20 = !{!"omnipotent char", !21, i64 0}
+!21 = !{!"Simple C/C++ TBAA"}
+!22 = !DILocation(line: 3, column: 8, scope: !15)
+!23 = !DILocation(line: 3, column: 12, scope: !15)
+!24 = !{!25, !25, i64 0}
+!25 = !{!"int", !20, i64 0}
+!26 = !DILocation(line: 3, column: 19, scope: !27)
+!27 = distinct !DILexicalBlock(scope: !15, file: !1, line: 3, column: 3)
+!28 = !DILocation(line: 3, column: 24, scope: !27)
+!29 = !DILocation(line: 3, column: 23, scope: !27)
+!30 = !DILocation(line: 3, column: 21, scope: !27)
+!31 = !DILocation(line: 3, column: 3, scope: !15)
+!32 = !{!"branch_weights", i32 1, i32 1048575}
+!33 = !DILocation(line: 3, column: 3, scope: !27)
+!34 = !DILocation(line: 4, column: 5, scope: !27)
+!35 = !DILocation(line: 4, column: 7, scope: !27)
+!36 = !DILocation(line: 4, column: 10, scope: !27)
+!37 = !DILocation(line: 3, column: 27, scope: !27)
+!38 = distinct !{!38, !31, !39}
+!39 = !DILocation(line: 4, column: 12, scope: !15)
+!40 = !DILocation(line: 5, column: 1, scope: !7)
+!41 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 7, type: !8, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !42)
+!42 = !{!43, !44}
+!43 = !DILocalVariable(name: "A", arg: 1, scope: !41, file: !1, line: 7, type: !10)
+!44 = !DILocalVariable(name: "i", scope: !45, file: !1, line: 9, type: !11)
+!45 = distinct !DILexicalBlock(scope: !41, file: !1, line: 9, column: 3)
+!46 = !DILocation(line: 7, column: 15, scope: !41)
+!47 = !DILocation(line: 9, column: 8, scope: !45)
+!48 = !DILocation(line: 9, column: 12, scope: !45)
+!49 = !DILocation(line: 9, column: 19, scope: !50)
+!50 = distinct !DILexicalBlock(scope: !45, file: !1, line: 9, column: 3)
+!51 = !DILocation(line: 9, column: 24, scope: !50)
+!52 = !DILocation(line: 9, column: 23, scope: !50)
+!53 = !DILocation(line: 9, column: 21, scope: !50)
+!54 = !DILocation(line: 9, column: 3, scope: !45)
+!55 = !DILocation(line: 9, column: 3, scope: !50)
+!56 = !DILocation(line: 10, column: 5, scope: !50)
+!57 = !DILocation(line: 10, column: 7, scope: !50)
+!58 = !DILocation(line: 10, column: 10, scope: !50)
+!59 = !DILocation(line: 9, column: 27, scope: !50)
+!60 = distinct !{!60, !54, !61}
+!61 = !DILocation(line: 10, column: 12, scope: !45)
+!62 = !DILocation(line: 11, column: 1, scope: !41)
+;.
+; CHECK: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C99, file: [[META1:![0-9]+]], producer: "{{.*}}clang version {{.*}}", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: [[META2:![0-9]+]], splitDebugInlining: false, nameTableKind: None)
+; CHECK: [[META1]] = !DIFile(filename: "various_ir_values.c", directory: {{.*}})
+; CHECK: [[META2]] = !{}
+; CHECK: [[META7:![0-9]+]] = distinct !DISubprogram(name: "foo", scope: [[META1]], file: [[META1]], line: 1, type: [[META8:![0-9]+]], scopeLine: 1, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META12:![0-9]+]])
+; CHECK: [[META8]] = !DISubroutineType(types: [[META9:![0-9]+]])
+; CHECK: [[META9]] = !{null, [[META10:![0-9]+]]}
+; CHECK: [[META10]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META11:![0-9]+]], size: 64)
+; CHECK: [[META11]] = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+; CHECK: [[META12]] = !{[[META13]], [[META14]]}
+; CHECK: [[META13]] = !DILocalVariable(name: "A", arg: 1, scope: [[META7]], file: [[META1]], line: 1, type: [[META10]])
+; CHECK: [[META14]] = !DILocalVariable(name: "i", scope: [[META15:![0-9]+]], file: [[META1]], line: 3, type: [[META11]])
+; CHECK: [[META15]] = distinct !DILexicalBlock(scope: [[META7]], file: [[META1]], line: 3, column: 3)
+; CHECK: [[DIASSIGNID16]] = distinct !DIAssignID()
+; CHECK: [[META17]] = !DILocation(line: 1, column: 15, scope: [[META7]])
+; CHECK: [[TBAA18]] = !{[[META19:![0-9]+]], [[META19]], i64 0}
+; CHECK: [[META19]] = !{!"any pointer", [[META20:![0-9]+]], i64 0}
+; CHECK: [[META20]] = !{!"omnipotent char", [[META21:![0-9]+]], i64 0}
+; CHECK: [[META21]] = !{!"Simple C/C++ TBAA"}
+; CHECK: [[DBG22]] = !DILocation(line: 3, column: 8, scope: [[META15]])
+; CHECK: [[META23]] = !DILocation(line: 3, column: 12, scope: [[META15]])
+; CHECK: [[TBAA24]] = !{[[META25:![0-9]+]], [[META25]], i64 0}
+; CHECK: [[META25]] = !{!"int", [[META20]], i64 0}
+; CHECK: [[DBG26]] = !DILocation(line: 3, column: 19, scope: [[META27:![0-9]+]])
+; CHECK: [[META27]] = distinct !DILexicalBlock(scope: [[META15]], file: [[META1]], line: 3, column: 3)
+; CHECK: [[DBG28]] = !DILocation(line: 3, column: 24, scope: [[META27]])
+; CHECK: [[DBG29]] = !DILocation(line: 3, column: 23, scope: [[META27]])
+; CHECK: [[DBG30]] = !DILocation(line: 3, column: 21, scope: [[META27]])
+; CHECK: [[DBG31]] = !DILocation(line: 3, column: 3, scope: [[META15]])
+; CHECK: [[PROF32]] = !{!"branch_weights", i32 1, i32 1048575}
+; CHECK: [[DBG33]] = !DILocation(line: 3, column: 3, scope: [[META27]])
+; CHECK: [[DBG34]] = !DILocation(line: 4, column: 5, scope: [[META27]])
+; CHECK: [[DBG35]] = !DILocation(line: 4, column: 7, scope: [[META27]])
+; CHECK: [[DBG36]] = !DILocation(line: 4, column: 10, scope: [[META27]])
+; CHECK: [[DBG37]] = !DILocation(line: 3, column: 27, scope: [[META27]])
+; CHECK: [[LOOP38]] = distinct !{[[LOOP38]], [[DBG31]], [[META39:![0-9]+]]}
+; CHECK: [[META39]] = !DILocation(line: 4, column: 12, scope: [[META15]])
+; CHECK: [[DBG40]] = !DILocation(line: 5, column: 1, scope: [[META7]])
+; CHECK: [[META41:![0-9]+]] = distinct !DISubprogram(name: "bar", scope: [[META1]], file: [[META1]], line: 7, type: [[META8]], scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META42:![0-9]+]])
+; CHECK: [[META42]] = !{[[META43]], [[META44]]}
+; CHECK: [[META43]] = !DILocalVariable(name: "A", arg: 1, scope: [[META41]], file: [[META1]], line: 7, type: [[META10]])
+; CHECK: [[META44]] = !DILocalVariable(name: "i", scope: [[META45:![0-9]+]], file: [[META1]], line: 9, type: [[META11]])
+; CHECK: [[META45]] = distinct !DILexicalBlock(scope: [[META41]], file: [[META1]], line: 9, column: 3)
+; CHECK: [[META46]] = !DILocation(line: 7, column: 15, scope: [[META41]])
+; CHECK: [[DBG47]] = !DILocation(line: 9, column: 8, scope: [[META45]])
+; CHECK: [[META48]] = !DILocation(line: 9, column: 12, scope: [[META45]])
+; CHECK: [[DBG49]] = !DILocation(line: 9, column: 19, scope: [[META50:![0-9]+]])
+; CHECK: [[META50]] = distinct !DILexicalBlock(scope: [[META45]], file: [[META1]], line: 9, column: 3)
+; CHECK: [[DBG51]] = !DILocation(line: 9, column: 24, scope: [[META50]])
+; CHECK: [[DBG52]] = !DILocation(line: 9, column: 23, scope: [[META50]])
+; CHECK: [[DBG53]] = !DILocation(line: 9, column: 21, scope: [[META50]])
+; CHECK: [[DBG54]] = !DILocation(line: 9, column: 3, scope: [[META45]])
+; CHECK: [[DBG55]] = !DILocation(line: 9, column: 3, scope: [[META50]])
+; CHECK: [[DBG56]] = !DILocation(line: 10, column: 5, scope: [[META50]])
+; CHECK: [[DBG57]] = !DILocation(line: 10, column: 7, scope: [[META50]])
+; CHECK: [[DBG58]] = !DILocation(line: 10, column: 10, scope: [[META50]])
+; CHECK: [[DBG59]] = !DILocation(line: 9, column: 27, scope: [[META50]])
+; CHECK: [[LOOP60]] = distinct !{[[LOOP60]], [[DBG54]], [[META61:![0-9]+]]}
+; CHECK: [[META61]] = !DILocation(line: 10, column: 12, scope: [[META45]])
+; CHECK: [[DBG62]] = !DILocation(line: 11, column: 1, scope: [[META41]])
+;.
diff --git a/llvm/test/tools/UpdateTestChecks/update_test_checks/various_ir_values_dbgrecords.test b/llvm/test/tools/UpdateTestChecks/update_test_checks/various_ir_values_dbgrecords.test
new file mode 100644
index 0000000000000..9cc77d894d62c
--- /dev/null
+++ b/llvm/test/tools/UpdateTestChecks/update_test_checks/various_ir_values_dbgrecords.test
@@ -0,0 +1,24 @@
+## Basic test checking that update_test_checks.py works correctly on various "IR value" kinds
+# RUN: cp -f %S/Inputs/various_ir_values_dbgrecords.ll %t.ll && %update_test_checks %t.ll
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.expected
+## Check that running the script again does not change the result:
+# RUN: %update_test_checks %t.ll
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.expected
+## Also try the --function-signature flag
+# RUN: %update_test_checks %t.ll --function-signature
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected
+## Verify that running without the --function-signature flag does not removes
+## the -SAME: lines since the generated file will have --function-signature in
+## an UTC_ARGS: comment in the first line (from the invocation above) which is
+## added to the update invocation below.
+# RUN: %update_test_checks %t.ll
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.expected
+## Also try the --check-globals flag
+# RUN: %update_test_checks %t.ll --check-globals
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected
+# RUN: cp -f %S/Inputs/various_ir_values_dbgrecords.ll %t.ll && %update_test_checks %t.ll --function-signature --check-globals all
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.globals.expected
+# RUN: cp -f %S/Inputs/various_ir_values_dbgrecords.ll %t.ll && %update_test_checks %t.ll --check-globals none
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.noglobals.expected
+# RUN: cp -f %S/Inputs/various_ir_values_dbgrecords.ll %t.ll && %update_test_checks %t.ll --check-globals smart
+# RUN: diff -u %t.ll %S/Inputs/various_ir_values_dbgrecords.ll.funcsig.transitiveglobals.expected
diff --git a/llvm/utils/UpdateTestChecks/common.py b/llvm/utils/UpdateTestChecks/common.py
index 15d3d5e527d61..5595e6f417555 100644
--- a/llvm/utils/UpdateTestChecks/common.py
+++ b/llvm/utils/UpdateTestChecks/common.py
@@ -1082,10 +1082,10 @@ def get_value_use(self, var, match, var_prefix=None):
     NamelessValue(r"TBAA_STRUCT", "!", r"!tbaa.struct ", r"![0-9]+", None),
     NamelessValue(r"RNG", "!", r"!range ", r"![0-9]+", None),
     NamelessValue(r"LOOP", "!", r"!llvm.loop ", r"![0-9]+", None),
-    NamelessValue(r"META", "!", r"metadata ", r"![0-9]+", None),
     NamelessValue(r"META", "!", r"", r"![0-9]+", r"(?:distinct |)!.*"),
     NamelessValue(r"ACC_GRP", "!", r"!llvm.access.group ", r"![0-9]+", None),
     NamelessValue(r"META", "!", r"![a-z.]+ ", r"![0-9]+", None),
+    NamelessValue(r"META", "!", r"[, (]", r"![0-9]+", None),
 ]
 
 global_nameless_values = [

From 792d437b56adfb3416daf8105942d4899fb82763 Mon Sep 17 00:00:00 2001
From: NagyDonat <donat.nagy@ericsson.com>
Date: Wed, 17 Apr 2024 12:52:23 +0200
Subject: [PATCH 725/886] [clang-tidy NFC] Fix a typo in docs for
 sizeof-expression (#88912)

"Till heaven and earth pass, one jot, or one tittle shall not pass of
the law"
---
 .../docs/clang-tidy/checks/bugprone/sizeof-expression.rst       | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang-tools-extra/docs/clang-tidy/checks/bugprone/sizeof-expression.rst b/clang-tools-extra/docs/clang-tidy/checks/bugprone/sizeof-expression.rst
index a3e88b837d375..c37df1706eb4e 100644
--- a/clang-tools-extra/docs/clang-tidy/checks/bugprone/sizeof-expression.rst
+++ b/clang-tools-extra/docs/clang-tidy/checks/bugprone/sizeof-expression.rst
@@ -190,6 +190,6 @@ Options
 
 .. option:: WarnOnSizeOfPointerToAggregate
 
-   When `true, the check will warn on an expression like
+   When `true`, the check will warn on an expression like
    ``sizeof(expr)`` where the expression is a pointer
    to aggregate. Default is `true`.

From 5f3e106de3cd5ce6d7ba37fb11f6ad740cb430c5 Mon Sep 17 00:00:00 2001
From: Pavel Labath <pavel@labath.sk>
Date: Wed, 17 Apr 2024 13:05:15 +0200
Subject: [PATCH 726/886] [lldb/linux] Make sure the process continues running
 after a detach (#88494)

Fixes #85084

Whenever an inferior thread stops, lldb-server sends a SIGSTOP to all
other threads in the process to force them to stop as well. If those
threads stop on their own before they get a signal, this SIGSTOP will
remain pending and be delivered the next time the process resumes.

Normally, this is not a problem, because lldb-server will detect this
stale SIGSTOP and resume the process. However, if we detach from the
process while it has these SIGSTOPs pending, they will get immediately
delivered, and the process will remain stopped (most likely forever).

This patch fixes that by sending a SIGCONT just before detaching from
the process. This signal cancels out any pending SIGSTOPs, and ensures
it is able to run after we detach. It does have one somewhat unfortunate
side-effect that in that the process's SIGCONT handler (if it has one)
will get executed spuriously (from the process's POV).

This could be _sometimes_ avoided by tracking which threads got send a
SIGSTOP, and whether those threads stopped due to it. From what I could
tell by observing its behavior, this is what gdb does. I have not tried
to replicate that behavior here because it adds a nontrivial amount of
complexity and the result is still uncertain -- we still need to send a
SIGCONT (and execute the handler) when any thread stops for some other
reason (and leaves our SIGSTOP hanging). Furthermore, since SIGSTOPs
don't stack, it's also possible that our SIGSTOP/SIGCONT combination
will cancel a genuine SIGSTOP being sent to the debugger application (by
someone else), and there is nothing we can do about that. For this
reason I think it's simplest and most predictible to just always send a
SIGCONT when detaching, but if it turns out this is breaking something,
we can consider implementing something more elaborate.

One alternative I did try is to use PTRACE_INTERRUPT to suspend the
threads instead of a SIGSTOP. PTRACE_INTERUPT requires using
PTRACE_SEIZE to attach to the process, which also made this solution
somewhat complicated, but the main problem with that approach is that
PTRACE_INTERRUPT is not considered to be a signal-delivery-stop, which
means it's not possible to resume it while injecting another signal to
the inferior (which some of our tests expect to be able to do). This
limitation could be worked around by forcing the thread into a signal
delivery stop whenever we need to do this, but this additional
complication is what made me think this approach is also not worthwhile.

This patch should fix (at least some of) the problems with
TestConcurrentVFork, but I've also added a dedicated test for checking
that a process keeps running after we detach. Although the problem I'm
fixing here is linux-specific, the core functinoality of not stopping
after a detach should function the same way everywhere.
---
 .../Process/Linux/NativeProcessLinux.cpp      |  4 ++
 .../commands/process/detach-resumes/Makefile  |  4 ++
 .../detach-resumes/TestDetachResumes.py       | 59 +++++++++++++++++++
 .../commands/process/detach-resumes/main.cpp  | 48 +++++++++++++++
 .../concurrent_vfork/TestConcurrentVFork.py   | 16 -----
 5 files changed, 115 insertions(+), 16 deletions(-)
 create mode 100644 lldb/test/API/commands/process/detach-resumes/Makefile
 create mode 100644 lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
 create mode 100644 lldb/test/API/commands/process/detach-resumes/main.cpp

diff --git a/lldb/source/Plugins/Process/Linux/NativeProcessLinux.cpp b/lldb/source/Plugins/Process/Linux/NativeProcessLinux.cpp
index 5d2b4b03fe60c..59fc8726b7673 100644
--- a/lldb/source/Plugins/Process/Linux/NativeProcessLinux.cpp
+++ b/lldb/source/Plugins/Process/Linux/NativeProcessLinux.cpp
@@ -1089,6 +1089,10 @@ Status NativeProcessLinux::Detach() {
   if (GetID() == LLDB_INVALID_PROCESS_ID)
     return error;
 
+  // Cancel out any SIGSTOPs we may have sent while stopping the process.
+  // Otherwise, the process may stop as soon as we detach from it.
+  kill(GetID(), SIGCONT);
+
   for (const auto &thread : m_threads) {
     Status e = Detach(thread->GetID());
     if (e.Fail())
diff --git a/lldb/test/API/commands/process/detach-resumes/Makefile b/lldb/test/API/commands/process/detach-resumes/Makefile
new file mode 100644
index 0000000000000..c46619c662348
--- /dev/null
+++ b/lldb/test/API/commands/process/detach-resumes/Makefile
@@ -0,0 +1,4 @@
+CXX_SOURCES := main.cpp
+ENABLE_THREADS := YES
+
+include Makefile.rules
diff --git a/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py b/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
new file mode 100644
index 0000000000000..57727294ddc3d
--- /dev/null
+++ b/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
@@ -0,0 +1,59 @@
+"""
+Test that the process continues running after we detach from it.
+"""
+
+import lldb
+import time
+from lldbsuite.test.decorators import *
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test import lldbutil
+
+
+class DetachResumesTestCase(TestBase):
+    NO_DEBUG_INFO_TESTCASE = True
+
+    def test_detach_resumes(self):
+        self.build()
+        exe = self.getBuildArtifact()
+
+        # The inferior will use this file to let us know it is ready to be
+        # attached.
+        sync_file_path = lldbutil.append_to_process_working_directory(
+            self, "sync_file_%d" % (int(time.time()))
+        )
+
+        # And this one to let us know it is running after we've detached from
+        # it.
+        exit_file_path = lldbutil.append_to_process_working_directory(
+            self, "exit_file_%d" % (int(time.time()))
+        )
+
+        popen = self.spawnSubprocess(
+            self.getBuildArtifact(exe), [sync_file_path, exit_file_path]
+        )
+        lldbutil.wait_for_file_on_target(self, sync_file_path)
+
+        self.runCmd("process attach -p " + str(popen.pid))
+
+        # Set a breakpoint at a place that will be called by multiple threads
+        # simultaneously. On systems (e.g. linux) where the debugger needs to
+        # send signals to suspend threads, these signals will race with threads
+        # hitting the breakpoint (and stopping on their own).
+        bpno = lldbutil.run_break_set_by_symbol(self, "break_here")
+
+        # And let the inferior know it can call the function.
+        self.runCmd("expr -- wait_for_debugger_flag = false")
+
+        self.runCmd("continue")
+
+        self.expect(
+            "thread list",
+            STOPPED_DUE_TO_BREAKPOINT,
+            substrs=["stopped", "stop reason = breakpoint"],
+        )
+
+        # Detach, the process should keep running after this, and not be stopped
+        # by the signals that the debugger may have used to suspend the threads.
+        self.runCmd("detach")
+
+        lldbutil.wait_for_file_on_target(self, exit_file_path)
diff --git a/lldb/test/API/commands/process/detach-resumes/main.cpp b/lldb/test/API/commands/process/detach-resumes/main.cpp
new file mode 100644
index 0000000000000..e8050fef2c385
--- /dev/null
+++ b/lldb/test/API/commands/process/detach-resumes/main.cpp
@@ -0,0 +1,48 @@
+#include "pseudo_barrier.h"
+#include <chrono>
+#include <fcntl.h>
+#include <fstream>
+#include <stdio.h>
+#include <thread>
+#include <vector>
+
+pseudo_barrier_t barrier;
+
+constexpr size_t nthreads = 5;
+volatile bool wait_for_debugger_flag = true;
+
+void break_here() {}
+
+void tfunc() {
+  pseudo_barrier_wait(barrier);
+
+  break_here();
+}
+
+int main(int argc, char const *argv[]) {
+  lldb_enable_attach();
+
+  if (argc < 3)
+    return 1;
+
+  // Create a file to signal that this process has started up.
+  std::ofstream(argv[1]).close();
+
+  // And wait for it to attach.
+  for (int i = 0; i < 100 && wait_for_debugger_flag; ++i)
+    std::this_thread::sleep_for(std::chrono::seconds(1));
+
+  // Fire up the threads and have them call break_here() simultaneously.
+  pseudo_barrier_init(barrier, nthreads);
+  std::vector<std::thread> threads;
+  for (size_t i = 0; i < nthreads; ++i)
+    threads.emplace_back(tfunc);
+
+  for (std::thread &t : threads)
+    t.join();
+
+  // Create the file to let the debugger know we're running.
+  std::ofstream(argv[2]).close();
+
+  return 0;
+}
diff --git a/lldb/test/API/functionalities/fork/concurrent_vfork/TestConcurrentVFork.py b/lldb/test/API/functionalities/fork/concurrent_vfork/TestConcurrentVFork.py
index 1790bd497f4e6..2dcbb728549fb 100644
--- a/lldb/test/API/functionalities/fork/concurrent_vfork/TestConcurrentVFork.py
+++ b/lldb/test/API/functionalities/fork/concurrent_vfork/TestConcurrentVFork.py
@@ -48,8 +48,6 @@ def follow_child_helper(self, use_fork, call_exec):
         self.expect("continue", patterns=[r"exited with status = 1[0-4]"])
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_parent_vfork_no_exec(self):
         """
         Make sure that debugging concurrent vfork() from multiple threads won't crash lldb during follow-parent.
@@ -58,8 +56,6 @@ def test_follow_parent_vfork_no_exec(self):
         self.follow_parent_helper(use_fork=False, call_exec=False)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_parent_fork_no_exec(self):
         """
         Make sure that debugging concurrent fork() from multiple threads won't crash lldb during follow-parent.
@@ -68,8 +64,6 @@ def test_follow_parent_fork_no_exec(self):
         self.follow_parent_helper(use_fork=True, call_exec=False)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_parent_vfork_call_exec(self):
         """
         Make sure that debugging concurrent vfork() from multiple threads won't crash lldb during follow-parent.
@@ -78,8 +72,6 @@ def test_follow_parent_vfork_call_exec(self):
         self.follow_parent_helper(use_fork=False, call_exec=True)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_parent_fork_call_exec(self):
         """
         Make sure that debugging concurrent vfork() from multiple threads won't crash lldb during follow-parent.
@@ -88,8 +80,6 @@ def test_follow_parent_fork_call_exec(self):
         self.follow_parent_helper(use_fork=True, call_exec=True)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_child_vfork_no_exec(self):
         """
         Make sure that debugging concurrent vfork() from multiple threads won't crash lldb during follow-child.
@@ -98,8 +88,6 @@ def test_follow_child_vfork_no_exec(self):
         self.follow_child_helper(use_fork=False, call_exec=False)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_child_fork_no_exec(self):
         """
         Make sure that debugging concurrent fork() from multiple threads won't crash lldb during follow-child.
@@ -108,8 +96,6 @@ def test_follow_child_fork_no_exec(self):
         self.follow_child_helper(use_fork=True, call_exec=False)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_child_vfork_call_exec(self):
         """
         Make sure that debugging concurrent vfork() from multiple threads won't crash lldb during follow-child.
@@ -118,8 +104,6 @@ def test_follow_child_vfork_call_exec(self):
         self.follow_child_helper(use_fork=False, call_exec=True)
 
     @skipUnlessPlatform(["linux"])
-    # See https://github.com/llvm/llvm-project/issues/85084.
-    @skipIf(oslist=["linux"], archs=["aarch64", "arm"])
     def test_follow_child_fork_call_exec(self):
         """
         Make sure that debugging concurrent fork() from multiple threads won't crash lldb during follow-child.

From c8dca5bc0733e2fba81008fc33fcad1f45ba666a Mon Sep 17 00:00:00 2001
From: Sergio Afonso <safonsof@amd.com>
Date: Wed, 17 Apr 2024 12:17:50 +0100
Subject: [PATCH 727/886] [Flang][OpenMP][Lower] Refactor lowering of compound
 constructs (#87070)

This patch simplifies the lowering from PFT to MLIR of OpenMP compound
constructs (i.e. combined and composite).

The new approach consists of iteratively processing the outermost leaf
construct of the given combined construct until it cannot be split
further. Both leaf constructs and composite ones have `gen...()`
functions that are called when appropriate.

This approach enables treating a leaf construct the same way regardless
of if it appeared as part of a combined construct, and it also enables
the lowering of composite constructs as a single unit.

Previous corner cases are now handled in a more straightforward way and
comments pointing to the relevant spec section are added. Directive sets
are also completed with missing LOOP related constructs.
---
 .../flang/Semantics/openmp-directive-sets.h   |  57 ++-
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 430 ++++++++++++------
 2 files changed, 333 insertions(+), 154 deletions(-)

diff --git a/flang/include/flang/Semantics/openmp-directive-sets.h b/flang/include/flang/Semantics/openmp-directive-sets.h
index 91773ae3ea9a3..842d251b682aa 100644
--- a/flang/include/flang/Semantics/openmp-directive-sets.h
+++ b/flang/include/flang/Semantics/openmp-directive-sets.h
@@ -32,14 +32,14 @@ static const OmpDirectiveSet topDistributeSet{
 
 static const OmpDirectiveSet allDistributeSet{
     OmpDirectiveSet{
-        llvm::omp::OMPD_target_teams_distribute,
-        llvm::omp::OMPD_target_teams_distribute_parallel_do,
-        llvm::omp::OMPD_target_teams_distribute_parallel_do_simd,
-        llvm::omp::OMPD_target_teams_distribute_simd,
-        llvm::omp::OMPD_teams_distribute,
-        llvm::omp::OMPD_teams_distribute_parallel_do,
-        llvm::omp::OMPD_teams_distribute_parallel_do_simd,
-        llvm::omp::OMPD_teams_distribute_simd,
+        Directive::OMPD_target_teams_distribute,
+        Directive::OMPD_target_teams_distribute_parallel_do,
+        Directive::OMPD_target_teams_distribute_parallel_do_simd,
+        Directive::OMPD_target_teams_distribute_simd,
+        Directive::OMPD_teams_distribute,
+        Directive::OMPD_teams_distribute_parallel_do,
+        Directive::OMPD_teams_distribute_parallel_do_simd,
+        Directive::OMPD_teams_distribute_simd,
     } | topDistributeSet,
 };
 
@@ -63,10 +63,24 @@ static const OmpDirectiveSet allDoSet{
     } | topDoSet,
 };
 
+static const OmpDirectiveSet topLoopSet{
+    Directive::OMPD_loop,
+};
+
+static const OmpDirectiveSet allLoopSet{
+    OmpDirectiveSet{
+        Directive::OMPD_parallel_loop,
+        Directive::OMPD_target_parallel_loop,
+        Directive::OMPD_target_teams_loop,
+        Directive::OMPD_teams_loop,
+    } | topLoopSet,
+};
+
 static const OmpDirectiveSet topParallelSet{
     Directive::OMPD_parallel,
     Directive::OMPD_parallel_do,
     Directive::OMPD_parallel_do_simd,
+    Directive::OMPD_parallel_loop,
     Directive::OMPD_parallel_masked_taskloop,
     Directive::OMPD_parallel_masked_taskloop_simd,
     Directive::OMPD_parallel_master_taskloop,
@@ -82,6 +96,7 @@ static const OmpDirectiveSet allParallelSet{
         Directive::OMPD_target_parallel,
         Directive::OMPD_target_parallel_do,
         Directive::OMPD_target_parallel_do_simd,
+        Directive::OMPD_target_parallel_loop,
         Directive::OMPD_target_teams_distribute_parallel_do,
         Directive::OMPD_target_teams_distribute_parallel_do_simd,
         Directive::OMPD_teams_distribute_parallel_do,
@@ -118,12 +133,14 @@ static const OmpDirectiveSet topTargetSet{
     Directive::OMPD_target_parallel,
     Directive::OMPD_target_parallel_do,
     Directive::OMPD_target_parallel_do_simd,
+    Directive::OMPD_target_parallel_loop,
     Directive::OMPD_target_simd,
     Directive::OMPD_target_teams,
     Directive::OMPD_target_teams_distribute,
     Directive::OMPD_target_teams_distribute_parallel_do,
     Directive::OMPD_target_teams_distribute_parallel_do_simd,
     Directive::OMPD_target_teams_distribute_simd,
+    Directive::OMPD_target_teams_loop,
 };
 
 static const OmpDirectiveSet allTargetSet{topTargetSet};
@@ -156,11 +173,12 @@ static const OmpDirectiveSet topTeamsSet{
 
 static const OmpDirectiveSet allTeamsSet{
     OmpDirectiveSet{
-        llvm::omp::OMPD_target_teams,
-        llvm::omp::OMPD_target_teams_distribute,
-        llvm::omp::OMPD_target_teams_distribute_parallel_do,
-        llvm::omp::OMPD_target_teams_distribute_parallel_do_simd,
-        llvm::omp::OMPD_target_teams_distribute_simd,
+        Directive::OMPD_target_teams,
+        Directive::OMPD_target_teams_distribute,
+        Directive::OMPD_target_teams_distribute_parallel_do,
+        Directive::OMPD_target_teams_distribute_parallel_do_simd,
+        Directive::OMPD_target_teams_distribute_simd,
+        Directive::OMPD_target_teams_loop,
     } | topTeamsSet,
 };
 
@@ -178,6 +196,14 @@ static const OmpDirectiveSet allDistributeSimdSet{
 static const OmpDirectiveSet allDoSimdSet{allDoSet & allSimdSet};
 static const OmpDirectiveSet allTaskloopSimdSet{allTaskloopSet & allSimdSet};
 
+static const OmpDirectiveSet compositeConstructSet{
+    Directive::OMPD_distribute_parallel_do,
+    Directive::OMPD_distribute_parallel_do_simd,
+    Directive::OMPD_distribute_simd,
+    Directive::OMPD_do_simd,
+    Directive::OMPD_taskloop_simd,
+};
+
 static const OmpDirectiveSet blockConstructSet{
     Directive::OMPD_master,
     Directive::OMPD_ordered,
@@ -201,12 +227,14 @@ static const OmpDirectiveSet loopConstructSet{
     Directive::OMPD_distribute_simd,
     Directive::OMPD_do,
     Directive::OMPD_do_simd,
+    Directive::OMPD_loop,
     Directive::OMPD_masked_taskloop,
     Directive::OMPD_masked_taskloop_simd,
     Directive::OMPD_master_taskloop,
     Directive::OMPD_master_taskloop_simd,
     Directive::OMPD_parallel_do,
     Directive::OMPD_parallel_do_simd,
+    Directive::OMPD_parallel_loop,
     Directive::OMPD_parallel_masked_taskloop,
     Directive::OMPD_parallel_masked_taskloop_simd,
     Directive::OMPD_parallel_master_taskloop,
@@ -214,17 +242,20 @@ static const OmpDirectiveSet loopConstructSet{
     Directive::OMPD_simd,
     Directive::OMPD_target_parallel_do,
     Directive::OMPD_target_parallel_do_simd,
+    Directive::OMPD_target_parallel_loop,
     Directive::OMPD_target_simd,
     Directive::OMPD_target_teams_distribute,
     Directive::OMPD_target_teams_distribute_parallel_do,
     Directive::OMPD_target_teams_distribute_parallel_do_simd,
     Directive::OMPD_target_teams_distribute_simd,
+    Directive::OMPD_target_teams_loop,
     Directive::OMPD_taskloop,
     Directive::OMPD_taskloop_simd,
     Directive::OMPD_teams_distribute,
     Directive::OMPD_teams_distribute_parallel_do,
     Directive::OMPD_teams_distribute_parallel_do_simd,
     Directive::OMPD_teams_distribute_simd,
+    Directive::OMPD_teams_loop,
     Directive::OMPD_tile,
     Directive::OMPD_unroll,
 };
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index c31d63625dbb1..bb38082b245ef 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -488,6 +488,81 @@ markDeclareTarget(mlir::Operation *op,
   declareTargetOp.setDeclareTarget(deviceType, captureClause);
 }
 
+/// Split a combined directive into an outer leaf directive and the (possibly
+/// combined) rest of the combined directive. Composite directives and
+/// non-compound directives are not split, in which case it will return the
+/// input directive as its first output and an empty value as its second output.
+static std::pair<llvm::omp::Directive, std::optional<llvm::omp::Directive>>
+splitCombinedDirective(llvm::omp::Directive dir) {
+  using D = llvm::omp::Directive;
+  switch (dir) {
+  case D::OMPD_masked_taskloop:
+    return {D::OMPD_masked, D::OMPD_taskloop};
+  case D::OMPD_masked_taskloop_simd:
+    return {D::OMPD_masked, D::OMPD_taskloop_simd};
+  case D::OMPD_master_taskloop:
+    return {D::OMPD_master, D::OMPD_taskloop};
+  case D::OMPD_master_taskloop_simd:
+    return {D::OMPD_master, D::OMPD_taskloop_simd};
+  case D::OMPD_parallel_do:
+    return {D::OMPD_parallel, D::OMPD_do};
+  case D::OMPD_parallel_do_simd:
+    return {D::OMPD_parallel, D::OMPD_do_simd};
+  case D::OMPD_parallel_masked:
+    return {D::OMPD_parallel, D::OMPD_masked};
+  case D::OMPD_parallel_masked_taskloop:
+    return {D::OMPD_parallel, D::OMPD_masked_taskloop};
+  case D::OMPD_parallel_masked_taskloop_simd:
+    return {D::OMPD_parallel, D::OMPD_masked_taskloop_simd};
+  case D::OMPD_parallel_master:
+    return {D::OMPD_parallel, D::OMPD_master};
+  case D::OMPD_parallel_master_taskloop:
+    return {D::OMPD_parallel, D::OMPD_master_taskloop};
+  case D::OMPD_parallel_master_taskloop_simd:
+    return {D::OMPD_parallel, D::OMPD_master_taskloop_simd};
+  case D::OMPD_parallel_sections:
+    return {D::OMPD_parallel, D::OMPD_sections};
+  case D::OMPD_parallel_workshare:
+    return {D::OMPD_parallel, D::OMPD_workshare};
+  case D::OMPD_target_parallel:
+    return {D::OMPD_target, D::OMPD_parallel};
+  case D::OMPD_target_parallel_do:
+    return {D::OMPD_target, D::OMPD_parallel_do};
+  case D::OMPD_target_parallel_do_simd:
+    return {D::OMPD_target, D::OMPD_parallel_do_simd};
+  case D::OMPD_target_simd:
+    return {D::OMPD_target, D::OMPD_simd};
+  case D::OMPD_target_teams:
+    return {D::OMPD_target, D::OMPD_teams};
+  case D::OMPD_target_teams_distribute:
+    return {D::OMPD_target, D::OMPD_teams_distribute};
+  case D::OMPD_target_teams_distribute_parallel_do:
+    return {D::OMPD_target, D::OMPD_teams_distribute_parallel_do};
+  case D::OMPD_target_teams_distribute_parallel_do_simd:
+    return {D::OMPD_target, D::OMPD_teams_distribute_parallel_do_simd};
+  case D::OMPD_target_teams_distribute_simd:
+    return {D::OMPD_target, D::OMPD_teams_distribute_simd};
+  case D::OMPD_teams_distribute:
+    return {D::OMPD_teams, D::OMPD_distribute};
+  case D::OMPD_teams_distribute_parallel_do:
+    return {D::OMPD_teams, D::OMPD_distribute_parallel_do};
+  case D::OMPD_teams_distribute_parallel_do_simd:
+    return {D::OMPD_teams, D::OMPD_distribute_parallel_do_simd};
+  case D::OMPD_teams_distribute_simd:
+    return {D::OMPD_teams, D::OMPD_distribute_simd};
+  case D::OMPD_parallel_loop:
+    return {D::OMPD_parallel, D::OMPD_loop};
+  case D::OMPD_target_parallel_loop:
+    return {D::OMPD_target, D::OMPD_parallel_loop};
+  case D::OMPD_target_teams_loop:
+    return {D::OMPD_target, D::OMPD_teams_loop};
+  case D::OMPD_teams_loop:
+    return {D::OMPD_teams, D::OMPD_loop};
+  default:
+    return {dir, std::nullopt};
+  }
+}
+
 //===----------------------------------------------------------------------===//
 // Op body generation helper structures and functions
 //===----------------------------------------------------------------------===//
@@ -1676,7 +1751,6 @@ static OpTy genTargetEnterExitUpdateDataOp(
   } else {
     llvm_unreachable("Unexpected TARGET DATA construct");
   }
-
   mlir::omp::TargetEnterExitUpdateDataClauseOps clauseOps;
   genTargetEnterExitUpdateDataClauses(converter, semaCtx, stmtCtx, clauseList,
                                       loc, directive, clauseOps);
@@ -1804,16 +1878,44 @@ genWsloopOp(Fortran::lower::AbstractConverter &converter,
 // Code generation functions for composite constructs
 //===----------------------------------------------------------------------===//
 
-static void genCompositeDoSimd(
+static void genCompositeDistributeParallelDo(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OmpClauseList &beginClauseList,
+    const Fortran::parser::OmpClauseList *endClauseList, mlir::Location loc) {
+  TODO(loc, "Composite DISTRIBUTE PARALLEL DO");
+}
+
+static void genCompositeDistributeParallelDoSimd(
+    Fortran::lower::AbstractConverter &converter,
+    Fortran::semantics::SemanticsContext &semaCtx,
+    Fortran::lower::pft::Evaluation &eval,
+    const Fortran::parser::OmpClauseList &beginClauseList,
+    const Fortran::parser::OmpClauseList *endClauseList, mlir::Location loc) {
+  TODO(loc, "Composite DISTRIBUTE PARALLEL DO SIMD");
+}
+
+static void genCompositeDistributeSimd(
     Fortran::lower::AbstractConverter &converter,
     Fortran::semantics::SemanticsContext &semaCtx,
-    Fortran::lower::pft::Evaluation &eval, llvm::omp::Directive ompDirective,
+    Fortran::lower::pft::Evaluation &eval,
     const Fortran::parser::OmpClauseList &beginClauseList,
     const Fortran::parser::OmpClauseList *endClauseList, mlir::Location loc) {
+  TODO(loc, "Composite DISTRIBUTE SIMD");
+}
+
+static void
+genCompositeDoSimd(Fortran::lower::AbstractConverter &converter,
+                   Fortran::semantics::SemanticsContext &semaCtx,
+                   Fortran::lower::pft::Evaluation &eval,
+                   const Fortran::parser::OmpClauseList &beginClauseList,
+                   const Fortran::parser::OmpClauseList *endClauseList,
+                   mlir::Location loc) {
   ClauseProcessor cp(converter, semaCtx, beginClauseList);
   cp.processTODO<clause::Aligned, clause::Allocate, clause::Linear,
-                 clause::Order, clause::Safelen, clause::Simdlen>(loc,
-                                                                  ompDirective);
+                 clause::Order, clause::Safelen, clause::Simdlen>(
+      loc, llvm::omp::OMPD_do_simd);
   // TODO: Add support for vectorization - add vectorization hints inside loop
   // body.
   // OpenMP standard does not specify the length of vector instructions.
@@ -1825,6 +1927,16 @@ static void genCompositeDoSimd(
   genWsloopOp(converter, semaCtx, eval, loc, beginClauseList, endClauseList);
 }
 
+static void
+genCompositeTaskloopSimd(Fortran::lower::AbstractConverter &converter,
+                         Fortran::semantics::SemanticsContext &semaCtx,
+                         Fortran::lower::pft::Evaluation &eval,
+                         const Fortran::parser::OmpClauseList &beginClauseList,
+                         const Fortran::parser::OmpClauseList *endClauseList,
+                         mlir::Location loc) {
+  TODO(loc, "Composite TASKLOOP SIMD");
+}
+
 //===----------------------------------------------------------------------===//
 // OpenMPDeclarativeConstruct visitors
 //===----------------------------------------------------------------------===//
@@ -2082,13 +2194,18 @@ genOMP(Fortran::lower::AbstractConverter &converter,
       std::get<Fortran::parser::OmpBeginBlockDirective>(blockConstruct.t);
   const auto &endBlockDirective =
       std::get<Fortran::parser::OmpEndBlockDirective>(blockConstruct.t);
-  const auto &directive =
-      std::get<Fortran::parser::OmpBlockDirective>(beginBlockDirective.t);
+  mlir::Location currentLocation =
+      converter.genLocation(beginBlockDirective.source);
+  const auto origDirective =
+      std::get<Fortran::parser::OmpBlockDirective>(beginBlockDirective.t).v;
   const auto &beginClauseList =
       std::get<Fortran::parser::OmpClauseList>(beginBlockDirective.t);
   const auto &endClauseList =
       std::get<Fortran::parser::OmpClauseList>(endBlockDirective.t);
 
+  assert(llvm::omp::blockConstructSet.test(origDirective) &&
+         "Expected block construct");
+
   for (const Fortran::parser::OmpClause &clause : beginClauseList.v) {
     mlir::Location clauseLocation = converter.genLocation(clause.source);
     if (!std::get_if<Fortran::parser::OmpClause::If>(&clause.u) &&
@@ -2124,93 +2241,74 @@ genOMP(Fortran::lower::AbstractConverter &converter,
       TODO(clauseLocation, "OpenMP Block construct clause");
   }
 
-  bool singleDirective = true;
-  mlir::Location currentLocation = converter.genLocation(directive.source);
-  switch (directive.v) {
-  case llvm::omp::Directive::OMPD_master:
-    genMasterOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation);
-    break;
-  case llvm::omp::Directive::OMPD_ordered:
-    genOrderedRegionOp(converter, semaCtx, eval, /*genNested=*/true,
-                       currentLocation, beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_parallel:
-    genParallelOp(converter, symTable, semaCtx, eval, /*genNested=*/true,
-                  currentLocation, beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_single:
-    genSingleOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-                beginClauseList, endClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_target:
-    genTargetOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
+  std::optional<llvm::omp::Directive> nextDir = origDirective;
+  bool outermostLeafConstruct = true;
+  while (nextDir) {
+    llvm::omp::Directive leafDir;
+    std::tie(leafDir, nextDir) = splitCombinedDirective(*nextDir);
+    const bool genNested = !nextDir;
+    const bool outerCombined = outermostLeafConstruct && nextDir.has_value();
+    switch (leafDir) {
+    case llvm::omp::Directive::OMPD_master:
+      // 2.16 MASTER construct.
+      genMasterOp(converter, semaCtx, eval, genNested, currentLocation);
+      break;
+    case llvm::omp::Directive::OMPD_ordered:
+      // 2.17.9 ORDERED construct.
+      genOrderedRegionOp(converter, semaCtx, eval, genNested, currentLocation,
+                         beginClauseList);
+      break;
+    case llvm::omp::Directive::OMPD_parallel:
+      // 2.6 PARALLEL construct.
+      genParallelOp(converter, symTable, semaCtx, eval, genNested,
+                    currentLocation, beginClauseList, outerCombined);
+      break;
+    case llvm::omp::Directive::OMPD_single:
+      // 2.8.2 SINGLE construct.
+      genSingleOp(converter, semaCtx, eval, genNested, currentLocation,
+                  beginClauseList, endClauseList);
+      break;
+    case llvm::omp::Directive::OMPD_target:
+      // 2.12.5 TARGET construct.
+      genTargetOp(converter, semaCtx, eval, genNested, currentLocation,
+                  beginClauseList, outerCombined);
+      break;
+    case llvm::omp::Directive::OMPD_target_data:
+      // 2.12.2 TARGET DATA construct.
+      genTargetDataOp(converter, semaCtx, eval, genNested, currentLocation,
+                      beginClauseList);
+      break;
+    case llvm::omp::Directive::OMPD_task:
+      // 2.10.1 TASK construct.
+      genTaskOp(converter, semaCtx, eval, genNested, currentLocation,
                 beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_target_data:
-    genTargetDataOp(converter, semaCtx, eval, /*genNested=*/true,
-                    currentLocation, beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_task:
-    genTaskOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-              beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_taskgroup:
-    genTaskgroupOp(converter, semaCtx, eval, /*genNested=*/true,
-                   currentLocation, beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_teams:
-    genTeamsOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-               beginClauseList);
-    break;
-  case llvm::omp::Directive::OMPD_workshare:
-    // FIXME: Workshare is not a commonly used OpenMP construct, an
-    // implementation for this feature will come later. For the codes
-    // that use this construct, add a single construct for now.
-    genSingleOp(converter, semaCtx, eval, /*genNested=*/true, currentLocation,
-                beginClauseList, endClauseList);
-    break;
-  default:
-    singleDirective = false;
-    break;
-  }
-
-  if (singleDirective)
-    return;
-
-  // Codegen for combined directives
-  bool combinedDirective = false;
-  if ((llvm::omp::allTargetSet & llvm::omp::blockConstructSet)
-          .test(directive.v)) {
-    genTargetOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-                beginClauseList, /*outerCombined=*/true);
-    combinedDirective = true;
-  }
-  if ((llvm::omp::allTeamsSet & llvm::omp::blockConstructSet)
-          .test(directive.v)) {
-    genTeamsOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-               beginClauseList);
-    combinedDirective = true;
-  }
-  if ((llvm::omp::allParallelSet & llvm::omp::blockConstructSet)
-          .test(directive.v)) {
-    bool outerCombined =
-        directive.v != llvm::omp::Directive::OMPD_target_parallel;
-    genParallelOp(converter, symTable, semaCtx, eval, /*genNested=*/false,
-                  currentLocation, beginClauseList, outerCombined);
-    combinedDirective = true;
-  }
-  if ((llvm::omp::workShareSet & llvm::omp::blockConstructSet)
-          .test(directive.v)) {
-    genSingleOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-                beginClauseList, endClauseList);
-    combinedDirective = true;
+      break;
+    case llvm::omp::Directive::OMPD_taskgroup:
+      // 2.17.6 TASKGROUP construct.
+      genTaskgroupOp(converter, semaCtx, eval, genNested, currentLocation,
+                     beginClauseList);
+      break;
+    case llvm::omp::Directive::OMPD_teams:
+      // 2.7 TEAMS construct.
+      // FIXME Pass the outerCombined argument or rename it to better describe
+      // what it represents if it must always be `false` in this context.
+      genTeamsOp(converter, semaCtx, eval, genNested, currentLocation,
+                 beginClauseList);
+      break;
+    case llvm::omp::Directive::OMPD_workshare:
+      // 2.8.3 WORKSHARE construct.
+      // FIXME: Workshare is not a commonly used OpenMP construct, an
+      // implementation for this feature will come later. For the codes
+      // that use this construct, add a single construct for now.
+      genSingleOp(converter, semaCtx, eval, genNested, currentLocation,
+                  beginClauseList, endClauseList);
+      break;
+    default:
+      llvm_unreachable("Unexpected block construct");
+      break;
+    }
+    outermostLeafConstruct = false;
   }
-  if (!combinedDirective)
-    TODO(currentLocation, "Unhandled block directive (" +
-                              llvm::omp::getOpenMPDirectiveName(directive.v) +
-                              ")");
-
-  genNestedEvaluations(converter, eval);
 }
 
 static void
@@ -2248,9 +2346,12 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
       std::get<Fortran::parser::OmpClauseList>(beginLoopDirective.t);
   mlir::Location currentLocation =
       converter.genLocation(beginLoopDirective.source);
-  const auto ompDirective =
+  const auto origDirective =
       std::get<Fortran::parser::OmpLoopDirective>(beginLoopDirective.t).v;
 
+  assert(llvm::omp::loopConstructSet.test(origDirective) &&
+         "Expected loop construct");
+
   const auto *endClauseList = [&]() {
     using RetTy = const Fortran::parser::OmpClauseList *;
     if (auto &endLoopDirective =
@@ -2262,56 +2363,103 @@ static void genOMP(Fortran::lower::AbstractConverter &converter,
     return RetTy();
   }();
 
-  bool validDirective = false;
-  if (llvm::omp::topTaskloopSet.test(ompDirective)) {
-    validDirective = true;
-    genTaskloopOp(converter, semaCtx, eval, currentLocation, beginClauseList);
-  } else {
-    // Create omp.{target, teams, distribute, parallel} nested operations
-    if ((llvm::omp::allTargetSet & llvm::omp::loopConstructSet)
-            .test(ompDirective)) {
-      validDirective = true;
-      genTargetOp(converter, semaCtx, eval, /*genNested=*/false,
-                  currentLocation, beginClauseList, /*outerCombined=*/true);
-    }
-    if ((llvm::omp::allTeamsSet & llvm::omp::loopConstructSet)
-            .test(ompDirective)) {
-      validDirective = true;
-      genTeamsOp(converter, semaCtx, eval, /*genNested=*/false, currentLocation,
-                 beginClauseList, /*outerCombined=*/true);
-    }
-    if (llvm::omp::allDistributeSet.test(ompDirective)) {
-      validDirective = true;
-      genDistributeOp(converter, semaCtx, eval, /*genNested=*/false,
-                      currentLocation, beginClauseList);
-    }
-    if ((llvm::omp::allParallelSet & llvm::omp::loopConstructSet)
-            .test(ompDirective)) {
-      validDirective = true;
-      genParallelOp(converter, symTable, semaCtx, eval, /*genNested=*/false,
-                    currentLocation, beginClauseList, /*outerCombined=*/true);
+  std::optional<llvm::omp::Directive> nextDir = origDirective;
+  while (nextDir) {
+    llvm::omp::Directive leafDir;
+    std::tie(leafDir, nextDir) = splitCombinedDirective(*nextDir);
+    if (llvm::omp::compositeConstructSet.test(leafDir)) {
+      assert(!nextDir && "Composite construct cannot be split");
+      switch (leafDir) {
+      case llvm::omp::Directive::OMPD_distribute_parallel_do:
+        // 2.9.4.3 DISTRIBUTE PARALLEL Worksharing-Loop construct.
+        genCompositeDistributeParallelDo(converter, semaCtx, eval,
+                                         beginClauseList, endClauseList,
+                                         currentLocation);
+        break;
+      case llvm::omp::Directive::OMPD_distribute_parallel_do_simd:
+        // 2.9.4.4 DISTRIBUTE PARALLEL Worksharing-Loop SIMD construct.
+        genCompositeDistributeParallelDoSimd(converter, semaCtx, eval,
+                                             beginClauseList, endClauseList,
+                                             currentLocation);
+        break;
+      case llvm::omp::Directive::OMPD_distribute_simd:
+        // 2.9.4.2 DISTRIBUTE SIMD construct.
+        genCompositeDistributeSimd(converter, semaCtx, eval, beginClauseList,
+                                   endClauseList, currentLocation);
+        break;
+      case llvm::omp::Directive::OMPD_do_simd:
+        // 2.9.3.2 Worksharing-Loop SIMD construct.
+        genCompositeDoSimd(converter, semaCtx, eval, beginClauseList,
+                           endClauseList, currentLocation);
+        break;
+      case llvm::omp::Directive::OMPD_taskloop_simd:
+        // 2.10.3 TASKLOOP SIMD construct.
+        genCompositeTaskloopSimd(converter, semaCtx, eval, beginClauseList,
+                                 endClauseList, currentLocation);
+        break;
+      default:
+        llvm_unreachable("Unexpected composite construct");
+      }
+    } else {
+      const bool genNested = !nextDir;
+      switch (leafDir) {
+      case llvm::omp::Directive::OMPD_distribute:
+        // 2.9.4.1 DISTRIBUTE construct.
+        genDistributeOp(converter, semaCtx, eval, genNested, currentLocation,
+                        beginClauseList);
+        break;
+      case llvm::omp::Directive::OMPD_do:
+        // 2.9.2 Worksharing-Loop construct.
+        genWsloopOp(converter, semaCtx, eval, currentLocation, beginClauseList,
+                    endClauseList);
+        break;
+      case llvm::omp::Directive::OMPD_parallel:
+        // 2.6 PARALLEL construct.
+        // FIXME This is not necessarily always the outer leaf construct of a
+        // combined construct in this context (e.g. DISTRIBUTE PARALLEL DO).
+        // Maybe rename the argument if it represents something else or
+        // initialize it properly.
+        genParallelOp(converter, symTable, semaCtx, eval, genNested,
+                      currentLocation, beginClauseList,
+                      /*outerCombined=*/true);
+        break;
+      case llvm::omp::Directive::OMPD_simd:
+        // 2.9.3.1 SIMD construct.
+        genSimdOp(converter, semaCtx, eval, currentLocation, beginClauseList);
+        break;
+      case llvm::omp::Directive::OMPD_target:
+        // 2.12.5 TARGET construct.
+        genTargetOp(converter, semaCtx, eval, genNested, currentLocation,
+                    beginClauseList, /*outerCombined=*/true);
+        break;
+      case llvm::omp::Directive::OMPD_taskloop:
+        // 2.10.2 TASKLOOP construct.
+        genTaskloopOp(converter, semaCtx, eval, currentLocation,
+                      beginClauseList);
+        break;
+      case llvm::omp::Directive::OMPD_teams:
+        // 2.7 TEAMS construct.
+        // FIXME This is not necessarily always the outer leaf construct of a
+        // combined construct in this constext (e.g. TARGET TEAMS DISTRIBUTE).
+        // Maybe rename the argument if it represents something else or
+        // initialize it properly.
+        genTeamsOp(converter, semaCtx, eval, genNested, currentLocation,
+                   beginClauseList, /*outerCombined=*/true);
+        break;
+      case llvm::omp::Directive::OMPD_loop:
+      case llvm::omp::Directive::OMPD_masked:
+      case llvm::omp::Directive::OMPD_master:
+      case llvm::omp::Directive::OMPD_tile:
+      case llvm::omp::Directive::OMPD_unroll:
+        TODO(currentLocation, "Unhandled loop directive (" +
+                                  llvm::omp::getOpenMPDirectiveName(leafDir) +
+                                  ")");
+        break;
+      default:
+        llvm_unreachable("Unexpected loop construct");
+      }
     }
   }
-  if ((llvm::omp::allDoSet | llvm::omp::allSimdSet).test(ompDirective))
-    validDirective = true;
-
-  if (!validDirective) {
-    TODO(currentLocation, "Unhandled loop directive (" +
-                              llvm::omp::getOpenMPDirectiveName(ompDirective) +
-                              ")");
-  }
-
-  if (llvm::omp::allDoSimdSet.test(ompDirective)) {
-    // 2.9.3.2 Workshare SIMD construct
-    genCompositeDoSimd(converter, semaCtx, eval, ompDirective, beginClauseList,
-                       endClauseList, currentLocation);
-  } else if (llvm::omp::allSimdSet.test(ompDirective)) {
-    // 2.9.3.1 SIMD construct
-    genSimdOp(converter, semaCtx, eval, currentLocation, beginClauseList);
-  } else {
-    genWsloopOp(converter, semaCtx, eval, currentLocation, beginClauseList,
-                endClauseList);
-  }
 }
 
 static void

From 06eedffe0d2782922e63cc25cb927f4acdaf7b30 Mon Sep 17 00:00:00 2001
From: NagyDonat <donat.nagy@ericsson.com>
Date: Wed, 17 Apr 2024 13:26:51 +0200
Subject: [PATCH 728/886] [analyzer] Use explicit call description mode in
 iterator checkers (#88913)

This commit explicitly specifies the matching mode (C library function,
any non-method function, or C++ method) for the `CallDescription`s
constructed in the iterator/container checkers.

This change won't cause major functional changes, but isn't NFC because
it ensures that e.g. call descriptions for a non-method function won't
accidentally match a method that has the same name.

Separate commits will perform (or have already performed) this change in
other checkers. My goal is to ensure that the call description mode is
always explicitly specified and eliminate (or strongly restrict) the
vague "may be either a method or a simple function" mode that's the
current default.

I'm handling the iterator checkers in this separate commit because
they're infamously complex; but I don't expect any trouble because this
transition doesn't interact with the "central" logic of iterator
handling.
---
 .../Checkers/ContainerModeling.cpp            | 33 +++++----
 .../Checkers/DebugContainerModeling.cpp       |  4 +-
 .../Checkers/DebugIteratorModeling.cpp        |  6 +-
 .../Checkers/IteratorModeling.cpp             |  7 +-
 .../Checkers/IteratorRangeChecker.cpp         | 11 ++-
 .../Checkers/STLAlgorithmModeling.cpp         | 67 ++++++++++++-------
 6 files changed, 80 insertions(+), 48 deletions(-)

diff --git a/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
index 009c0d3fb9368..55ed809bfed6c 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ContainerModeling.cpp
@@ -72,26 +72,31 @@ class ContainerModeling
                                                    SVal) const;
 
   CallDescriptionMap<NoItParamFn> NoIterParamFunctions = {
-      {{{"clear"}, 0}, &ContainerModeling::handleClear},
-      {{{"assign"}, 2}, &ContainerModeling::handleAssign},
-      {{{"push_back"}, 1}, &ContainerModeling::handlePushBack},
-      {{{"emplace_back"}, 1}, &ContainerModeling::handlePushBack},
-      {{{"pop_back"}, 0}, &ContainerModeling::handlePopBack},
-      {{{"push_front"}, 1}, &ContainerModeling::handlePushFront},
-      {{{"emplace_front"}, 1}, &ContainerModeling::handlePushFront},
-      {{{"pop_front"}, 0}, &ContainerModeling::handlePopFront},
+      {{CDM::CXXMethod, {"clear"}, 0}, &ContainerModeling::handleClear},
+      {{CDM::CXXMethod, {"assign"}, 2}, &ContainerModeling::handleAssign},
+      {{CDM::CXXMethod, {"push_back"}, 1}, &ContainerModeling::handlePushBack},
+      {{CDM::CXXMethod, {"emplace_back"}, 1},
+       &ContainerModeling::handlePushBack},
+      {{CDM::CXXMethod, {"pop_back"}, 0}, &ContainerModeling::handlePopBack},
+      {{CDM::CXXMethod, {"push_front"}, 1},
+       &ContainerModeling::handlePushFront},
+      {{CDM::CXXMethod, {"emplace_front"}, 1},
+       &ContainerModeling::handlePushFront},
+      {{CDM::CXXMethod, {"pop_front"}, 0}, &ContainerModeling::handlePopFront},
   };
 
   CallDescriptionMap<OneItParamFn> OneIterParamFunctions = {
-      {{{"insert"}, 2}, &ContainerModeling::handleInsert},
-      {{{"emplace"}, 2}, &ContainerModeling::handleInsert},
-      {{{"erase"}, 1}, &ContainerModeling::handleErase},
-      {{{"erase_after"}, 1}, &ContainerModeling::handleEraseAfter},
+      {{CDM::CXXMethod, {"insert"}, 2}, &ContainerModeling::handleInsert},
+      {{CDM::CXXMethod, {"emplace"}, 2}, &ContainerModeling::handleInsert},
+      {{CDM::CXXMethod, {"erase"}, 1}, &ContainerModeling::handleErase},
+      {{CDM::CXXMethod, {"erase_after"}, 1},
+       &ContainerModeling::handleEraseAfter},
   };
 
   CallDescriptionMap<TwoItParamFn> TwoIterParamFunctions = {
-      {{{"erase"}, 2}, &ContainerModeling::handleErase},
-      {{{"erase_after"}, 2}, &ContainerModeling::handleEraseAfter},
+      {{CDM::CXXMethod, {"erase"}, 2}, &ContainerModeling::handleErase},
+      {{CDM::CXXMethod, {"erase_after"}, 2},
+       &ContainerModeling::handleEraseAfter},
   };
 };
 
diff --git a/clang/lib/StaticAnalyzer/Checkers/DebugContainerModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/DebugContainerModeling.cpp
index 72186a99d9435..d3830a01dd0cb 100644
--- a/clang/lib/StaticAnalyzer/Checkers/DebugContainerModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/DebugContainerModeling.cpp
@@ -42,9 +42,9 @@ class DebugContainerModeling
                                                  CheckerContext &) const;
 
   CallDescriptionMap<FnCheck> Callbacks = {
-      {{{"clang_analyzer_container_begin"}, 1},
+      {{CDM::SimpleFunc, {"clang_analyzer_container_begin"}, 1},
        &DebugContainerModeling::analyzerContainerBegin},
-      {{{"clang_analyzer_container_end"}, 1},
+      {{CDM::SimpleFunc, {"clang_analyzer_container_end"}, 1},
        &DebugContainerModeling::analyzerContainerEnd},
   };
 
diff --git a/clang/lib/StaticAnalyzer/Checkers/DebugIteratorModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/DebugIteratorModeling.cpp
index 79ab71d7829db..203743dacda63 100644
--- a/clang/lib/StaticAnalyzer/Checkers/DebugIteratorModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/DebugIteratorModeling.cpp
@@ -43,11 +43,11 @@ class DebugIteratorModeling
                                                  CheckerContext &) const;
 
   CallDescriptionMap<FnCheck> Callbacks = {
-      {{{"clang_analyzer_iterator_position"}, 1},
+      {{CDM::SimpleFunc, {"clang_analyzer_iterator_position"}, 1},
        &DebugIteratorModeling::analyzerIteratorPosition},
-      {{{"clang_analyzer_iterator_container"}, 1},
+      {{CDM::SimpleFunc, {"clang_analyzer_iterator_container"}, 1},
        &DebugIteratorModeling::analyzerIteratorContainer},
-      {{{"clang_analyzer_iterator_validity"}, 1},
+      {{CDM::SimpleFunc, {"clang_analyzer_iterator_validity"}, 1},
        &DebugIteratorModeling::analyzerIteratorValidity},
   };
 
diff --git a/clang/lib/StaticAnalyzer/Checkers/IteratorModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/IteratorModeling.cpp
index a95e811c2a418..5649454b4cd47 100644
--- a/clang/lib/StaticAnalyzer/Checkers/IteratorModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/IteratorModeling.cpp
@@ -129,19 +129,20 @@ class IteratorModeling
   CallDescriptionMap<AdvanceFn> AdvanceLikeFunctions = {
       // template<class InputIt, class Distance>
       // void advance(InputIt& it, Distance n);
-      {{{"std", "advance"}, 2}, &IteratorModeling::handleAdvance},
+      {{CDM::SimpleFunc, {"std", "advance"}, 2},
+       &IteratorModeling::handleAdvance},
 
       // template<class BidirIt>
       // BidirIt prev(
       //   BidirIt it,
       //   typename std::iterator_traits<BidirIt>::difference_type n = 1);
-      {{{"std", "prev"}, 2}, &IteratorModeling::handlePrev},
+      {{CDM::SimpleFunc, {"std", "prev"}, 2}, &IteratorModeling::handlePrev},
 
       // template<class ForwardIt>
       // ForwardIt next(
       //   ForwardIt it,
       //   typename std::iterator_traits<ForwardIt>::difference_type n = 1);
-      {{{"std", "next"}, 2}, &IteratorModeling::handleNext},
+      {{CDM::SimpleFunc, {"std", "next"}, 2}, &IteratorModeling::handleNext},
   };
 
 public:
diff --git a/clang/lib/StaticAnalyzer/Checkers/IteratorRangeChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/IteratorRangeChecker.cpp
index d2b61fb92483c..4dd2f700a2a0e 100644
--- a/clang/lib/StaticAnalyzer/Checkers/IteratorRangeChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/IteratorRangeChecker.cpp
@@ -56,10 +56,15 @@ class IteratorRangeChecker
   using AdvanceFn = void (IteratorRangeChecker::*)(CheckerContext &, SVal,
                                                    SVal) const;
 
+  // FIXME: these three functions are also listed in IteratorModeling.cpp,
+  // perhaps unify their handling?
   CallDescriptionMap<AdvanceFn> AdvanceFunctions = {
-      {{{"std", "advance"}, 2}, &IteratorRangeChecker::verifyAdvance},
-      {{{"std", "prev"}, 2}, &IteratorRangeChecker::verifyPrev},
-      {{{"std", "next"}, 2}, &IteratorRangeChecker::verifyNext},
+      {{CDM::SimpleFunc, {"std", "advance"}, 2},
+       &IteratorRangeChecker::verifyAdvance},
+      {{CDM::SimpleFunc, {"std", "prev"}, 2},
+       &IteratorRangeChecker::verifyPrev},
+      {{CDM::SimpleFunc, {"std", "next"}, 2},
+       &IteratorRangeChecker::verifyNext},
   };
 };
 
diff --git a/clang/lib/StaticAnalyzer/Checkers/STLAlgorithmModeling.cpp b/clang/lib/StaticAnalyzer/Checkers/STLAlgorithmModeling.cpp
index a5173a05636a0..e037719b90298 100644
--- a/clang/lib/StaticAnalyzer/Checkers/STLAlgorithmModeling.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/STLAlgorithmModeling.cpp
@@ -33,29 +33,50 @@ class STLAlgorithmModeling : public Checker<eval::Call> {
                                                 const CallExpr *) const;
 
   const CallDescriptionMap<FnCheck> Callbacks = {
-    {{{"std", "find"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_if"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_if"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_if_not"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_if_not"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_first_of"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_first_of"}, 5}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_first_of"}, 6}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_end"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_end"}, 5}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "find_end"}, 6}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "lower_bound"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "lower_bound"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "upper_bound"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "upper_bound"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search"}, 3}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search"}, 5}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search"}, 6}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search_n"}, 4}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search_n"}, 5}, &STLAlgorithmModeling::evalFind},
-    {{{"std", "search_n"}, 6}, &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find"}, 3}, &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find"}, 4}, &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_if"}, 3},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_if"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_if_not"}, 3},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_if_not"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_first_of"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_first_of"}, 5},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_first_of"}, 6},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_end"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_end"}, 5},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "find_end"}, 6},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "lower_bound"}, 3},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "lower_bound"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "upper_bound"}, 3},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "upper_bound"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search"}, 3},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search"}, 5},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search"}, 6},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search_n"}, 4},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search_n"}, 5},
+       &STLAlgorithmModeling::evalFind},
+      {{CDM::SimpleFunc, {"std", "search_n"}, 6},
+       &STLAlgorithmModeling::evalFind},
   };
 
 public:

From 631c5e818ef8bb0f61fd3bb44cc4449be2142e2b Mon Sep 17 00:00:00 2001
From: MbjYjbpivj <mbjyjbpivj@gmail.com>
Date: Wed, 17 Apr 2024 20:15:30 +0800
Subject: [PATCH 729/886] [mlir] fix intNEQValue summary (#89029)

Fix the summary of intNEQValue.
---
 mlir/include/mlir/IR/CommonAttrConstraints.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/include/mlir/IR/CommonAttrConstraints.td b/mlir/include/mlir/IR/CommonAttrConstraints.td
index 0312ac7ec1d8d..0d69bb0717a59 100644
--- a/mlir/include/mlir/IR/CommonAttrConstraints.td
+++ b/mlir/include/mlir/IR/CommonAttrConstraints.td
@@ -755,7 +755,7 @@ class AllAttrOf<list<AttrConstraint> constraints> : AttrConstraint<
 
 class IntNEQValue<int n> : AttrConstraint<
     CPred<"::llvm::cast<::mlir::IntegerAttr>($_self).getInt() != " # n>,
-    "whose minimum value is " # n>;
+    "whose value is not " # n>;
 
 class IntMinValue<int n> : AttrConstraint<
     CPred<"::llvm::cast<::mlir::IntegerAttr>($_self).getInt() >= " # n>,

From d57907d0b4f292f148310695ed011fe5a0585d6b Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Wed, 17 Apr 2024 08:21:48 -0400
Subject: [PATCH 730/886] [libc++] Add missing iterator requirement checks in
 the PSTL (#88127)

Also add tests for those, and add a few missing requirements to testing
iterators in the test suite.
---
 .../__algorithm/pstl_any_all_none_of.h        |   6 +-
 libcxx/include/__algorithm/pstl_copy.h        |  13 ++
 libcxx/include/__algorithm/pstl_count.h       |   5 +
 libcxx/include/__algorithm/pstl_equal.h       |   9 +
 libcxx/include/__algorithm/pstl_fill.h        |   6 +-
 libcxx/include/__algorithm/pstl_find.h        |   6 +-
 libcxx/include/__algorithm/pstl_for_each.h    |   4 +-
 libcxx/include/__algorithm/pstl_generate.h    |   5 +-
 .../include/__algorithm/pstl_is_partitioned.h |   2 +
 libcxx/include/__algorithm/pstl_merge.h       |   5 +
 libcxx/include/__algorithm/pstl_move.h        |   5 +
 libcxx/include/__algorithm/pstl_replace.h     |  13 ++
 libcxx/include/__algorithm/pstl_rotate_copy.h |   5 +
 libcxx/include/__algorithm/pstl_sort.h        |   3 +
 libcxx/include/__algorithm/pstl_stable_sort.h |   2 +
 libcxx/include/__algorithm/pstl_transform.h   |  16 +-
 .../__iterator/cpp17_iterator_concepts.h      |  38 ++--
 libcxx/include/__numeric/pstl_reduce.h        |   3 +
 .../include/__numeric/pstl_transform_reduce.h |   6 +
 .../cpp17_iterator_concepts.verify.cpp        | 125 ++++++------
 .../pstl.iterator-requirements.verify.cpp     | 192 ++++++++++++++++++
 libcxx/test/support/test_iterators.h          |  16 +-
 22 files changed, 383 insertions(+), 102 deletions(-)
 create mode 100644 libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp

diff --git a/libcxx/include/__algorithm/pstl_any_all_none_of.h b/libcxx/include/__algorithm/pstl_any_all_none_of.h
index 4b1e0e61b5421..911a7e42b3fa3 100644
--- a/libcxx/include/__algorithm/pstl_any_all_none_of.h
+++ b/libcxx/include/__algorithm/pstl_any_all_none_of.h
@@ -60,7 +60,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI bool
 any_of(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "any_of requires a ForwardIterator");
   auto __res = std::__any_of(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -99,7 +99,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI bool
 all_of(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Pred __pred) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "all_of requires a ForwardIterator");
   auto __res = std::__all_of(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -136,7 +136,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI bool
 none_of(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Pred __pred) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "none_of requires a ForwardIterator");
   auto __res = std::__none_of(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_copy.h b/libcxx/include/__algorithm/pstl_copy.h
index 1069dcec0e117..f35bb9713ef14 100644
--- a/libcxx/include/__algorithm/pstl_copy.h
+++ b/libcxx/include/__algorithm/pstl_copy.h
@@ -16,6 +16,7 @@
 #include <__config>
 #include <__functional/identity.h>
 #include <__iterator/concepts.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_constant_evaluated.h>
 #include <__type_traits/is_execution_policy.h>
@@ -67,6 +68,12 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator
 copy(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _ForwardOutIterator __result) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardIterator, "copy(first, last, result) requires [first, last) to be ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardOutIterator, "copy(first, last, result) requires result to be a ForwardIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(*__first), "copy(first, last, result) requires result to be an OutputIterator");
   auto __res = std::__copy(__policy, std::move(__first), std::move(__last), std::move(__result));
   if (!__res)
     std::__throw_bad_alloc();
@@ -106,6 +113,12 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator
 copy_n(_ExecutionPolicy&& __policy, _ForwardIterator __first, _Size __n, _ForwardOutIterator __result) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardIterator, "copy_n(first, n, result) requires first to be a ForwardIterator");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardOutIterator, "copy_n(first, n, result) requires result to be a ForwardIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(*__first), "copy_n(first, n, result) requires result to be an OutputIterator");
   auto __res = std::__copy_n(__policy, std::move(__first), std::move(__n), std::move(__result));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_count.h b/libcxx/include/__algorithm/pstl_count.h
index 2781f6bfd3c9e..6ff57cac334eb 100644
--- a/libcxx/include/__algorithm/pstl_count.h
+++ b/libcxx/include/__algorithm/pstl_count.h
@@ -17,6 +17,7 @@
 #include <__atomic/atomic.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/pstl_transform_reduce.h>
 #include <__type_traits/enable_if.h>
@@ -70,6 +71,8 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI __iter_diff_t<_ForwardIterator>
 count_if(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardIterator, "count_if(first, last, pred) requires [first, last) to be ForwardIterators");
   auto __res = std::__count_if(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -106,6 +109,8 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI __iter_diff_t<_ForwardIterator>
 count(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(
+      _ForwardIterator, "count(first, last, val) requires [first, last) to be ForwardIterators");
   auto __res = std::__count(__policy, std::move(__first), std::move(__last), __value);
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_equal.h b/libcxx/include/__algorithm/pstl_equal.h
index d235c0f4f4197..0b38197d7f63d 100644
--- a/libcxx/include/__algorithm/pstl_equal.h
+++ b/libcxx/include/__algorithm/pstl_equal.h
@@ -13,6 +13,7 @@
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/pstl_transform_reduce.h>
 #include <__utility/move.h>
@@ -74,6 +75,8 @@ equal(_ExecutionPolicy&& __policy,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _Pred __pred) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "equal requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "equal requires ForwardIterators");
   auto __res = std::__equal(__policy, std::move(__first1), std::move(__last1), std::move(__first2), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -86,6 +89,8 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<__remove_cvref_t<_ExecutionPolicy>>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI bool
 equal(_ExecutionPolicy&& __policy, _ForwardIterator1 __first1, _ForwardIterator1 __last1, _ForwardIterator2 __first2) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "equal requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "equal requires ForwardIterators");
   return std::equal(__policy, std::move(__first1), std::move(__last1), std::move(__first2), std::equal_to{});
 }
 
@@ -145,6 +150,8 @@ equal(_ExecutionPolicy&& __policy,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2,
       _Pred __pred) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "equal requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "equal requires ForwardIterators");
   auto __res = std::__equal(
       __policy, std::move(__first1), std::move(__last1), std::move(__first2), std::move(__last2), std::move(__pred));
   if (!__res)
@@ -162,6 +169,8 @@ equal(_ExecutionPolicy&& __policy,
       _ForwardIterator1 __last1,
       _ForwardIterator2 __first2,
       _ForwardIterator2 __last2) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "equal requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "equal requires ForwardIterators");
   return std::equal(
       __policy, std::move(__first1), std::move(__last1), std::move(__first2), std::move(__last2), std::equal_to{});
 }
diff --git a/libcxx/include/__algorithm/pstl_fill.h b/libcxx/include/__algorithm/pstl_fill.h
index 488b49a0feec9..fd248506bc4b9 100644
--- a/libcxx/include/__algorithm/pstl_fill.h
+++ b/libcxx/include/__algorithm/pstl_fill.h
@@ -43,7 +43,6 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __fill(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) noexcept {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
   return std::__pstl_frontend_dispatch(
       _LIBCPP_PSTL_CUSTOMIZATION_POINT(__pstl_fill, _RawPolicy),
       [&](_ForwardIterator __g_first, _ForwardIterator __g_last, const _Tp& __g_value) {
@@ -63,7 +62,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 fill(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "fill requires ForwardIterators");
   if (!std::__fill(__policy, std::move(__first), std::move(__last), __value))
     std::__throw_bad_alloc();
 }
@@ -79,7 +78,6 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __fill_n(_ExecutionPolicy&& __policy, _ForwardIterator&& __first, _SizeT&& __n, const _Tp& __value) noexcept {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
   return std::__pstl_frontend_dispatch(
       _LIBCPP_PSTL_CUSTOMIZATION_POINT(__pstl_fill_n, _RawPolicy),
       [&](_ForwardIterator __g_first, _SizeT __g_n, const _Tp& __g_value) {
@@ -102,7 +100,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 fill_n(_ExecutionPolicy&& __policy, _ForwardIterator __first, _SizeT __n, const _Tp& __value) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "fill_n requires ForwardIterators");
   if (!std::__fill_n(__policy, std::move(__first), std::move(__n), __value))
     std::__throw_bad_alloc();
 }
diff --git a/libcxx/include/__algorithm/pstl_find.h b/libcxx/include/__algorithm/pstl_find.h
index 5b694db68aead..3b30a7bc9b456 100644
--- a/libcxx/include/__algorithm/pstl_find.h
+++ b/libcxx/include/__algorithm/pstl_find.h
@@ -50,7 +50,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardIterator
 find_if(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "find_if requires ForwardIterators");
   auto __res = std::__find_if(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -88,7 +88,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardIterator
 find_if_not(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "find_if_not requires ForwardIterators");
   auto __res = std::__find_if_not(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
@@ -125,7 +125,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardIterator
 find(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, const _Tp& __value) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "find requires ForwardIterators");
   auto __res = std::__find(__policy, std::move(__first), std::move(__last), __value);
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_for_each.h b/libcxx/include/__algorithm/pstl_for_each.h
index bb7b5a61a6dc0..a9ebed74a62fd 100644
--- a/libcxx/include/__algorithm/pstl_for_each.h
+++ b/libcxx/include/__algorithm/pstl_for_each.h
@@ -53,7 +53,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 for_each(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Function __func) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "for_each requires ForwardIterators");
   if (!std::__for_each(__policy, std::move(__first), std::move(__last), std::move(__func)))
     std::__throw_bad_alloc();
 }
@@ -93,7 +93,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 for_each_n(_ExecutionPolicy&& __policy, _ForwardIterator __first, _Size __size, _Function __func) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "for_each_n requires a ForwardIterator");
   auto __res = std::__for_each_n(__policy, std::move(__first), std::move(__size), std::move(__func));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_generate.h b/libcxx/include/__algorithm/pstl_generate.h
index 7133c6f4f4c62..886af290d7f25 100644
--- a/libcxx/include/__algorithm/pstl_generate.h
+++ b/libcxx/include/__algorithm/pstl_generate.h
@@ -42,7 +42,6 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI optional<__empty>
 __generate(_ExecutionPolicy&& __policy, _ForwardIterator&& __first, _ForwardIterator&& __last, _Generator&& __gen) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
   return std::__pstl_frontend_dispatch(
       _LIBCPP_PSTL_CUSTOMIZATION_POINT(__pstl_generate, _RawPolicy),
       [&__policy](_ForwardIterator __g_first, _ForwardIterator __g_last, _Generator __g_gen) {
@@ -63,7 +62,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 generate(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Generator __gen) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "generate requires ForwardIterators");
   if (!std::__generate(__policy, std::move(__first), std::move(__last), std::move(__gen)))
     std::__throw_bad_alloc();
 }
@@ -100,7 +99,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 generate_n(_ExecutionPolicy&& __policy, _ForwardIterator __first, _Size __n, _Generator __gen) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "generate_n requires a ForwardIterator");
   if (!std::__generate_n(__policy, std::move(__first), std::move(__n), std::move(__gen)))
     std::__throw_bad_alloc();
 }
diff --git a/libcxx/include/__algorithm/pstl_is_partitioned.h b/libcxx/include/__algorithm/pstl_is_partitioned.h
index b654302122072..108bb1e432526 100644
--- a/libcxx/include/__algorithm/pstl_is_partitioned.h
+++ b/libcxx/include/__algorithm/pstl_is_partitioned.h
@@ -14,6 +14,7 @@
 #include <__algorithm/pstl_find.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
@@ -62,6 +63,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_NODISCARD_EXT _LIBCPP_HIDE_FROM_ABI bool
 is_partitioned(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _Predicate __pred) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "is_partitioned requires ForwardIterators");
   auto __res = std::__is_partitioned(__policy, std::move(__first), std::move(__last), std::move(__pred));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_merge.h b/libcxx/include/__algorithm/pstl_merge.h
index 3d262db6bc0c1..d03cd8c7fbd58 100644
--- a/libcxx/include/__algorithm/pstl_merge.h
+++ b/libcxx/include/__algorithm/pstl_merge.h
@@ -12,6 +12,7 @@
 #include <__algorithm/pstl_backend.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
@@ -70,6 +71,10 @@ merge(_ExecutionPolicy&& __policy,
       _ForwardIterator2 __last2,
       _ForwardOutIterator __result,
       _Comp __comp = {}) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "merge requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "merge requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, decltype(*__first1), "merge requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, decltype(*__first2), "merge requires an OutputIterator");
   auto __res = std::__merge(
       __policy,
       std::move(__first1),
diff --git a/libcxx/include/__algorithm/pstl_move.h b/libcxx/include/__algorithm/pstl_move.h
index d8441f1a6c2e1..f4c8c1fbb2e87 100644
--- a/libcxx/include/__algorithm/pstl_move.h
+++ b/libcxx/include/__algorithm/pstl_move.h
@@ -15,6 +15,7 @@
 #include <__algorithm/pstl_transform.h>
 #include <__config>
 #include <__functional/identity.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_constant_evaluated.h>
@@ -69,6 +70,10 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator
 move(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last, _ForwardOutIterator __result) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "move requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "move requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(std::move(*__first)), "move requires an OutputIterator");
   auto __res = std::__move(__policy, std::move(__first), std::move(__last), std::move(__result));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_replace.h b/libcxx/include/__algorithm/pstl_replace.h
index b1caf3fd4ac0a..73ac11cda26a9 100644
--- a/libcxx/include/__algorithm/pstl_replace.h
+++ b/libcxx/include/__algorithm/pstl_replace.h
@@ -14,6 +14,7 @@
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__algorithm/pstl_transform.h>
 #include <__config>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/remove_cvref.h>
@@ -74,6 +75,7 @@ replace_if(_ExecutionPolicy&& __policy,
            _ForwardIterator __last,
            _Pred __pred,
            const _Tp& __new_value) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "replace_if requires ForwardIterators");
   auto __res = std::__replace_if(__policy, std::move(__first), std::move(__last), std::move(__pred), __new_value);
   if (!__res)
     std::__throw_bad_alloc();
@@ -121,6 +123,7 @@ replace(_ExecutionPolicy&& __policy,
         _ForwardIterator __last,
         const _Tp& __old_value,
         const _Tp& __new_value) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "replace requires ForwardIterators");
   if (!std::__replace(__policy, std::move(__first), std::move(__last), __old_value, __new_value))
     std::__throw_bad_alloc();
 }
@@ -177,6 +180,11 @@ _LIBCPP_HIDE_FROM_ABI void replace_copy_if(
     _ForwardOutIterator __result,
     _Pred __pred,
     const _Tp& __new_value) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "replace_copy_if requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "replace_copy_if requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(*__first), "replace_copy_if requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, const _Tp&, "replace_copy requires an OutputIterator");
   if (!std::__replace_copy_if(
           __policy, std::move(__first), std::move(__last), std::move(__result), std::move(__pred), __new_value))
     std::__throw_bad_alloc();
@@ -233,6 +241,11 @@ _LIBCPP_HIDE_FROM_ABI void replace_copy(
     _ForwardOutIterator __result,
     const _Tp& __old_value,
     const _Tp& __new_value) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "replace_copy requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "replace_copy requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(*__first), "replace_copy requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, const _Tp&, "replace_copy requires an OutputIterator");
   if (!std::__replace_copy(
           __policy, std::move(__first), std::move(__last), std::move(__result), __old_value, __new_value))
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__algorithm/pstl_rotate_copy.h b/libcxx/include/__algorithm/pstl_rotate_copy.h
index 346aab1d4a55c..adab3958fe311 100644
--- a/libcxx/include/__algorithm/pstl_rotate_copy.h
+++ b/libcxx/include/__algorithm/pstl_rotate_copy.h
@@ -12,6 +12,7 @@
 #include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_copy.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/is_execution_policy.h>
 #include <optional>
 
@@ -69,6 +70,10 @@ _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator rotate_copy(
     _ForwardIterator __middle,
     _ForwardIterator __last,
     _ForwardOutIterator __result) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "rotate_copy requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "rotate_copy requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(*__first), "rotate_copy requires an OutputIterator");
   auto __res =
       std::__rotate_copy(__policy, std::move(__first), std::move(__middle), std::move(__last), std::move(__result));
   if (!__res)
diff --git a/libcxx/include/__algorithm/pstl_sort.h b/libcxx/include/__algorithm/pstl_sort.h
index a931f768111a2..65bc794ca6f4c 100644
--- a/libcxx/include/__algorithm/pstl_sort.h
+++ b/libcxx/include/__algorithm/pstl_sort.h
@@ -14,6 +14,7 @@
 #include <__algorithm/pstl_stable_sort.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
 #include <__utility/empty.h>
@@ -60,6 +61,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 sort(_ExecutionPolicy&& __policy, _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp) {
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(_RandomAccessIterator, "sort requires RandomAccessIterators");
   if (!std::__sort(__policy, std::move(__first), std::move(__last), std::move(__comp)))
     std::__throw_bad_alloc();
 }
@@ -70,6 +72,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void
 sort(_ExecutionPolicy&& __policy, _RandomAccessIterator __first, _RandomAccessIterator __last) {
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(_RandomAccessIterator, "sort requires RandomAccessIterators");
   std::sort(std::forward<_ExecutionPolicy>(__policy), std::move(__first), std::move(__last), less{});
 }
 
diff --git a/libcxx/include/__algorithm/pstl_stable_sort.h b/libcxx/include/__algorithm/pstl_stable_sort.h
index 8ea0bb3f9a8d5..79b94557e3dc3 100644
--- a/libcxx/include/__algorithm/pstl_stable_sort.h
+++ b/libcxx/include/__algorithm/pstl_stable_sort.h
@@ -12,6 +12,7 @@
 #include <__algorithm/pstl_backend.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
@@ -48,6 +49,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI void stable_sort(
     _ExecutionPolicy&& __policy, _RandomAccessIterator __first, _RandomAccessIterator __last, _Comp __comp = {}) {
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(_RandomAccessIterator, "stable_sort requires RandomAccessIterators");
   if (!std::__stable_sort(__policy, std::move(__first), std::move(__last), std::move(__comp)))
     std::__throw_bad_alloc();
 }
diff --git a/libcxx/include/__algorithm/pstl_transform.h b/libcxx/include/__algorithm/pstl_transform.h
index f95938782fc3b..a01a64a43cf1a 100644
--- a/libcxx/include/__algorithm/pstl_transform.h
+++ b/libcxx/include/__algorithm/pstl_transform.h
@@ -58,9 +58,10 @@ _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator transform(
     _ForwardIterator __last,
     _ForwardOutIterator __result,
     _UnaryOperation __op) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator);
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator);
-  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, decltype(__op(*__first)));
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "transform requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "transform requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(__op(*__first)), "transform requires an OutputIterator");
   auto __res = std::__transform(__policy, std::move(__first), std::move(__last), std::move(__result), std::move(__op));
   if (!__res)
     std::__throw_bad_alloc();
@@ -100,10 +101,11 @@ _LIBCPP_HIDE_FROM_ABI _ForwardOutIterator transform(
     _ForwardIterator2 __first2,
     _ForwardOutIterator __result,
     _BinaryOperation __op) {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1);
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2);
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator);
-  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(_ForwardOutIterator, decltype(__op(*__first1, *__first2)));
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "transform requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "transform requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardOutIterator, "transform requires an OutputIterator");
+  _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(
+      _ForwardOutIterator, decltype(__op(*__first1, *__first2)), "transform requires an OutputIterator");
   auto __res = std::__transform(
       __policy, std::move(__first1), std::move(__last1), std::move(__first2), std::move(__result), std::move(__op));
   if (!__res)
diff --git a/libcxx/include/__iterator/cpp17_iterator_concepts.h b/libcxx/include/__iterator/cpp17_iterator_concepts.h
index cdb561e68452a..9d5a392582da4 100644
--- a/libcxx/include/__iterator/cpp17_iterator_concepts.h
+++ b/libcxx/include/__iterator/cpp17_iterator_concepts.h
@@ -157,29 +157,31 @@ concept __cpp17_random_access_iterator =
 _LIBCPP_END_NAMESPACE_STD
 
 #  ifndef _LIBCPP_DISABLE_ITERATOR_CHECKS
-#    define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t) static_assert(::std::__cpp17_input_iterator<iter_t>);
-#    define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t)                                                     \
-      static_assert(::std::__cpp17_output_iterator<iter_t, write_t>);
-#    define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t) static_assert(::std::__cpp17_forward_iterator<iter_t>);
-#    define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t)                                                       \
-      static_assert(::std::__cpp17_bidirectional_iterator<iter_t>);
-#    define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t)                                                       \
-      static_assert(::std::__cpp17_random_access_iterator<iter_t>);
+#    define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t, message)                                                      \
+      static_assert(::std::__cpp17_input_iterator<iter_t>, message)
+#    define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t, message)                                            \
+      static_assert(::std::__cpp17_output_iterator<iter_t, write_t>, message)
+#    define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t, message)                                                    \
+      static_assert(::std::__cpp17_forward_iterator<iter_t>, message)
+#    define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t, message)                                              \
+      static_assert(::std::__cpp17_bidirectional_iterator<iter_t>, message)
+#    define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t, message)                                              \
+      static_assert(::std::__cpp17_random_access_iterator<iter_t>, message)
 #  else
-#    define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t)
-#    define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t)
-#    define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t)
-#    define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t)
-#    define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t)
+#    define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t, message) static_assert(true)
+#    define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t, message) static_assert(true)
+#    define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t, message) static_assert(true)
+#    define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t, message) static_assert(true)
+#    define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t, message) static_assert(true)
 #  endif
 
 #else // _LIBCPP_STD_VER >= 20
 
-#  define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t)
-#  define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t)
-#  define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t)
-#  define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t)
-#  define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t)
+#  define _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(iter_t, message) static_assert(true)
+#  define _LIBCPP_REQUIRE_CPP17_OUTPUT_ITERATOR(iter_t, write_t, message) static_assert(true)
+#  define _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(iter_t, message) static_assert(true)
+#  define _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(iter_t, message) static_assert(true)
+#  define _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(iter_t, message) static_assert(true)
 
 #endif // _LIBCPP_STD_VER >= 20
 
diff --git a/libcxx/include/__numeric/pstl_reduce.h b/libcxx/include/__numeric/pstl_reduce.h
index f9f666c2bb38b..d678b9480070b 100644
--- a/libcxx/include/__numeric/pstl_reduce.h
+++ b/libcxx/include/__numeric/pstl_reduce.h
@@ -12,6 +12,7 @@
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__functional/identity.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/pstl_transform_reduce.h>
 #include <__type_traits/is_execution_policy.h>
@@ -66,6 +67,7 @@ reduce(_ExecutionPolicy&& __policy,
        _ForwardIterator __last,
        _Tp __init,
        _BinaryOperation __op = {}) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "reduce requires ForwardIterators");
   auto __res = std::__reduce(__policy, std::move(__first), std::move(__last), std::move(__init), std::move(__op));
   if (!__res)
     std::__throw_bad_alloc();
@@ -94,6 +96,7 @@ template <class _ExecutionPolicy,
           enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI __iter_value_type<_ForwardIterator>
 reduce(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __last) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "reduce requires ForwardIterators");
   auto __res = std::__reduce(__policy, std::move(__first), std::move(__last));
   if (!__res)
     std::__throw_bad_alloc();
diff --git a/libcxx/include/__numeric/pstl_transform_reduce.h b/libcxx/include/__numeric/pstl_transform_reduce.h
index 07ecf0d9956bb..2d2621dc8dadb 100644
--- a/libcxx/include/__numeric/pstl_transform_reduce.h
+++ b/libcxx/include/__numeric/pstl_transform_reduce.h
@@ -13,6 +13,7 @@
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__functional/operations.h>
+#include <__iterator/cpp17_iterator_concepts.h>
 #include <__numeric/transform_reduce.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
@@ -72,6 +73,8 @@ _LIBCPP_HIDE_FROM_ABI _Tp transform_reduce(
     _Tp __init,
     _BinaryOperation1 __reduce,
     _BinaryOperation2 __transform) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "transform_reduce requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "transform_reduce requires ForwardIterators");
   auto __res = std::__transform_reduce(
       __policy,
       std::move(__first1),
@@ -99,6 +102,8 @@ _LIBCPP_HIDE_FROM_ABI _Tp transform_reduce(
     _ForwardIterator1 __last1,
     _ForwardIterator2 __first2,
     _Tp __init) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator1, "transform_reduce requires ForwardIterators");
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator2, "transform_reduce requires ForwardIterators");
   return std::transform_reduce(__policy, __first1, __last1, __first2, __init, plus{}, multiplies{});
 }
 
@@ -140,6 +145,7 @@ _LIBCPP_HIDE_FROM_ABI _Tp transform_reduce(
     _Tp __init,
     _BinaryOperation __reduce,
     _UnaryOperation __transform) {
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "transform_reduce requires ForwardIterators");
   auto __res = std::__transform_reduce(
       __policy, std::move(__first), std::move(__last), std::move(__init), std::move(__reduce), std::move(__transform));
   if (!__res)
diff --git a/libcxx/test/libcxx/algorithms/cpp17_iterator_concepts.verify.cpp b/libcxx/test/libcxx/algorithms/cpp17_iterator_concepts.verify.cpp
index 344543d5f19ff..544a9744b7909 100644
--- a/libcxx/test/libcxx/algorithms/cpp17_iterator_concepts.verify.cpp
+++ b/libcxx/test/libcxx/algorithms/cpp17_iterator_concepts.verify.cpp
@@ -16,29 +16,29 @@
 #include <cstddef>
 
 struct missing_deref {
-  using difference_type = std::ptrdiff_t;
+  using difference_type   = std::ptrdiff_t;
   using iterator_category = std::input_iterator_tag;
-  using value_type = int;
-  using reference = int&;
+  using value_type        = int;
+  using reference         = int&;
 
   missing_deref& operator++();
 };
 
 struct missing_preincrement {
-  using difference_type = std::ptrdiff_t;
+  using difference_type   = std::ptrdiff_t;
   using iterator_category = std::input_iterator_tag;
-  using value_type = int;
-  using reference = int&;
+  using value_type        = int;
+  using reference         = int&;
 
   int& operator*();
 };
 
 template <class Derived>
 struct valid_iterator {
-  using difference_type = std::ptrdiff_t;
+  using difference_type   = std::ptrdiff_t;
   using iterator_category = std::input_iterator_tag;
-  using value_type = int;
-  using reference = int&;
+  using value_type        = int;
+  using reference         = int&;
 
   int& operator*() const;
   Derived& operator++();
@@ -51,30 +51,30 @@ struct valid_iterator {
 };
 
 struct not_move_constructible : valid_iterator<not_move_constructible> {
-  not_move_constructible(const not_move_constructible&) = default;
-  not_move_constructible(not_move_constructible&&) = delete;
-  not_move_constructible& operator=(not_move_constructible&&) = default;
+  not_move_constructible(const not_move_constructible&)            = default;
+  not_move_constructible(not_move_constructible&&)                 = delete;
+  not_move_constructible& operator=(not_move_constructible&&)      = default;
   not_move_constructible& operator=(const not_move_constructible&) = default;
 };
 
 struct not_copy_constructible : valid_iterator<not_copy_constructible> {
-  not_copy_constructible(const not_copy_constructible&) = delete;
-  not_copy_constructible(not_copy_constructible&&) = default;
-  not_copy_constructible& operator=(not_copy_constructible&&) = default;
+  not_copy_constructible(const not_copy_constructible&)            = delete;
+  not_copy_constructible(not_copy_constructible&&)                 = default;
+  not_copy_constructible& operator=(not_copy_constructible&&)      = default;
   not_copy_constructible& operator=(const not_copy_constructible&) = default;
 };
 
 struct not_move_assignable : valid_iterator<not_move_assignable> {
-  not_move_assignable(const not_move_assignable&) = default;
-  not_move_assignable(not_move_assignable&&) = default;
-  not_move_assignable& operator=(not_move_assignable&&) = delete;
+  not_move_assignable(const not_move_assignable&)            = default;
+  not_move_assignable(not_move_assignable&&)                 = default;
+  not_move_assignable& operator=(not_move_assignable&&)      = delete;
   not_move_assignable& operator=(const not_move_assignable&) = default;
 };
 
 struct not_copy_assignable : valid_iterator<not_copy_assignable> {
-  not_copy_assignable(const not_copy_assignable&) = default;
-  not_copy_assignable(not_copy_assignable&&) = default;
-  not_copy_assignable& operator=(not_copy_assignable&&) = default;
+  not_copy_assignable(const not_copy_assignable&)            = default;
+  not_copy_assignable(not_copy_assignable&&)                 = default;
+  not_copy_assignable& operator=(not_copy_assignable&&)      = default;
   not_copy_assignable& operator=(const not_copy_assignable&) = delete;
 };
 
@@ -89,7 +89,6 @@ void check_iterator_requirements() {
   static_assert(std::__cpp17_iterator<missing_preincrement>); // expected-error {{static assertion failed}}
   // expected-note@*:* {{cannot increment value of type 'missing_preincrement'}}
 
-
   static_assert(std::__cpp17_iterator<not_move_constructible>); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'not_move_constructible' does not satisfy '__cpp17_move_constructible'}}
 
@@ -115,11 +114,13 @@ bool operator==(not_unequality_comparable, not_unequality_comparable);
 bool operator!=(not_unequality_comparable, not_unequality_comparable) = delete;
 
 void check_input_iterator_requirements() {
-  _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(not_equality_comparable); // expected-error {{static assertion failed}}
+  // clang-format off
+  _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(not_equality_comparable, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{'__lhs == __rhs' would be invalid: overload resolution selected deleted operator '=='}}
 
-  _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(not_unequality_comparable); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_INPUT_ITERATOR(not_unequality_comparable, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{'__lhs != __rhs' would be invalid: overload resolution selected deleted operator '!='}}
+  // clang-format on
 }
 
 template <class Derived>
@@ -138,9 +139,9 @@ struct postincrement_not_ref : valid_iterator<postincrement_not_ref> {};
 bool operator==(postincrement_not_ref, postincrement_not_ref);
 
 void check_forward_iterator_requirements() {
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(not_default_constructible); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(not_default_constructible, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'not_default_constructible' does not satisfy '__cpp17_default_constructible'}}
-  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(postincrement_not_ref); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(postincrement_not_ref, ""); // expected-error {{static assertion failed}}
 #ifndef _AIX
   // expected-note@*:* {{because type constraint 'convertible_to<valid_iterator<postincrement_not_ref>::Proxy, const postincrement_not_ref &>' was not satisfied}}
 #endif
@@ -155,7 +156,6 @@ struct missing_postdecrement : valid_forward_iterator<missing_postdecrement> {
 };
 
 struct not_returning_iter_reference : valid_forward_iterator<not_returning_iter_reference> {
-
   struct Proxy {
     operator const not_returning_iter_reference&();
 
@@ -167,12 +167,14 @@ struct not_returning_iter_reference : valid_forward_iterator<not_returning_iter_
 };
 
 void check_bidirectional_iterator_requirements() {
-  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(missing_predecrement); // expected-error {{static assertion failed}}
+  // clang-format off
+  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(missing_predecrement, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{cannot decrement value of type 'missing_predecrement'}}
-  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(missing_postdecrement); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(missing_postdecrement, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{cannot decrement value of type 'missing_postdecrement'}}
-  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(not_returning_iter_reference); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_BIDIRECTIONAL_ITERATOR(not_returning_iter_reference, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because type constraint 'same_as<int, __iter_reference<not_returning_iter_reference> >' was not satisfied}}
+  // clang-format on
 }
 
 template <class Derived>
@@ -246,7 +248,8 @@ struct missing_minus_const_iter_const_iter : valid_random_access_iterator<missin
   friend difference_type operator-(missing_minus_const_iter_const_iter&, missing_minus_const_iter_const_iter&);
   friend difference_type operator-(const missing_minus_const_iter_const_iter&, missing_minus_const_iter_const_iter&);
   friend difference_type operator-(missing_minus_const_iter_const_iter&, const missing_minus_const_iter_const_iter&);
-  friend difference_type operator-(const missing_minus_const_iter_const_iter&, const missing_minus_const_iter_const_iter&) = delete;
+  friend difference_type
+  operator-(const missing_minus_const_iter_const_iter&, const missing_minus_const_iter_const_iter&) = delete;
 };
 
 struct missing_subscript_operator : valid_random_access_iterator<missing_subscript_operator> {
@@ -359,62 +362,64 @@ struct missing_const_const_greater_eq : valid_random_access_iterator<missing_con
 };
 
 void check_random_access_iterator() {
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_equals); // expected-error {{static assertion failed}}
+  // clang-format off
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_equals, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter += __n' would be invalid: overload resolution selected deleted operator '+='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_equals); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_equals, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter -= __n' would be invalid: overload resolution selected deleted operator '-='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter + __n' would be invalid: overload resolution selected deleted operator '+'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_diff_iter); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_diff_iter, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__n + __iter' would be invalid: overload resolution selected deleted operator '+'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff_const_mut); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff_const_mut, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) + __n' would be invalid: overload resolution selected deleted operator '+'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff_mut_const); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_plus_iter_diff_mut_const, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__n + std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '+'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_diff_const); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_diff_const, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) - __n' would be invalid: overload resolution selected deleted operator '-'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_iter); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_iter, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter - __iter' would be invalid: overload resolution selected deleted operator '-'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_const_iter_iter); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_const_iter_iter, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) - __iter' would be invalid: overload resolution selected deleted operator '-'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_const_iter); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_iter_const_iter, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter - std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '-'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_const_iter_const_iter); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_minus_const_iter_const_iter, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) - std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '-'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_subscript_operator); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_subscript_operator, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter[__n]' would be invalid: overload resolution selected deleted operator '[]'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_subscript_operator); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_subscript_operator, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter)[__n]' would be invalid: overload resolution selected deleted operator '[]'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_less); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_less, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter < __iter' would be invalid: overload resolution selected deleted operator '<'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_less); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_less, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) < __iter' would be invalid: overload resolution selected deleted operator '<'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_less); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_less, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter < std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '<'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_less); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_less, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) < std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '<'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_greater); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_greater, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter > __iter' would be invalid: overload resolution selected deleted operator '>'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_greater); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_greater, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) > __iter' would be invalid: overload resolution selected deleted operator '>'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_greater); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_greater, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter > std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '>'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_greater); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_greater, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) > std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '>'}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_less_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_less_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter <= __iter' would be invalid: overload resolution selected deleted operator '<='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_less_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_less_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) <= __iter' would be invalid: overload resolution selected deleted operator '<='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_less_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_less_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter <= std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '<='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_less_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_less_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) <= std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '<='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_greater_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_greater_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter >= __iter' would be invalid: overload resolution selected deleted operator '>='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_greater_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_mut_greater_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) >= __iter' would be invalid: overload resolution selected deleted operator '>='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_greater_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_mut_const_greater_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because '__iter >= std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '>='}}
-  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_greater_eq); // expected-error {{static assertion failed}}
+  _LIBCPP_REQUIRE_CPP17_RANDOM_ACCESS_ITERATOR(missing_const_const_greater_eq, ""); // expected-error {{static assertion failed}}
   // expected-note@*:* {{because 'std::as_const(__iter) >= std::as_const(__iter)' would be invalid: overload resolution selected deleted operator '>='}}
+  // clang-format on
 }
diff --git a/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp b/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
new file mode 100644
index 0000000000000..98e3509752e16
--- /dev/null
+++ b/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
@@ -0,0 +1,192 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, c++11, c++14
+// REQUIRES: stdlib=libc++
+// UNSUPPORTED: libcpp-has-no-incomplete-pstl
+
+// <algorithm>
+// <numeric>
+
+// Make sure that all PSTL algorithms contain checks for iterator requirements.
+// This is not a requirement from the Standard, but we strive to catch misuse in
+// the PSTL both because we can, and because iterator category mistakes in the
+// PSTL can lead to subtle bugs.
+
+// Ignore spurious errors after the initial static_assert failure.
+// ADDITIONAL_COMPILE_FLAGS: -Xclang -verify-ignore-unexpected=error
+
+// We only diagnose this in C++20 and above because we implement the checks with concepts.
+// UNSUPPORTED: c++17
+
+#include <algorithm>
+#include <cstddef>
+#include <numeric>
+
+#include "test_iterators.h"
+
+using non_forward_iterator = cpp17_input_iterator<int*>;
+struct non_output_iterator : forward_iterator<int*> {
+  constexpr int const& operator*() const; // prevent it from being an output iterator
+};
+
+void f(non_forward_iterator non_fwd, non_output_iterator non_output, std::execution::sequenced_policy pol) {
+  auto pred     = [](auto&&...) -> bool { return true; };
+  auto func     = [](auto&&...) -> int { return 1; };
+  int* it       = nullptr;
+  int* out      = nullptr;
+  std::size_t n = 0;
+  int val       = 0;
+
+  {
+    (void)std::any_of(pol, non_fwd, non_fwd, pred);  // expected-error@*:* {{static assertion failed: any_of}}
+    (void)std::all_of(pol, non_fwd, non_fwd, pred);  // expected-error@*:* {{static assertion failed: all_of}}
+    (void)std::none_of(pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: none_of}}
+  }
+
+  {
+    (void)std::copy(pol, non_fwd, non_fwd, it); // expected-error@*:* {{static assertion failed: copy}}
+    (void)std::copy(pol, it, it, non_fwd);      // expected-error@*:* {{static assertion failed: copy}}
+    (void)std::copy(pol, it, it, non_output);   // expected-error@*:* {{static assertion failed: copy}}
+  }
+  {
+    (void)std::copy_n(pol, non_fwd, n, it);    // expected-error@*:* {{static assertion failed: copy_n}}
+    (void)std::copy_n(pol, it, n, non_fwd);    // expected-error@*:* {{static assertion failed: copy_n}}
+    (void)std::copy_n(pol, it, n, non_output); // expected-error@*:* {{static assertion failed: copy_n}}
+  }
+
+  {
+    (void)std::count(pol, non_fwd, non_fwd, val);     // expected-error@*:* {{static assertion failed: count}}
+    (void)std::count_if(pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: count_if}}
+  }
+
+  {
+    (void)std::equal(pol, non_fwd, non_fwd, it);       // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, it, it, non_fwd);            // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, non_fwd, non_fwd, it, pred); // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, it, it, non_fwd, pred);      // expected-error@*:* {{static assertion failed: equal}}
+
+    (void)std::equal(pol, non_fwd, non_fwd, it, it);       // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, it, it, non_fwd, non_fwd);       // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, non_fwd, non_fwd, it, it, pred); // expected-error@*:* {{static assertion failed: equal}}
+    (void)std::equal(pol, it, it, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: equal}}
+  }
+
+  {
+    (void)std::fill(pol, non_fwd, non_fwd, val); // expected-error@*:* {{static assertion failed: fill}}
+    (void)std::fill_n(pol, non_fwd, n, val);     // expected-error@*:* {{static assertion failed: fill_n}}
+  }
+
+  {
+    (void)std::find(pol, non_fwd, non_fwd, val);         // expected-error@*:* {{static assertion failed: find}}
+    (void)std::find_if(pol, non_fwd, non_fwd, pred);     // expected-error@*:* {{static assertion failed: find_if}}
+    (void)std::find_if_not(pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: find_if_not}}
+  }
+
+  {
+    (void)std::for_each(pol, non_fwd, non_fwd, func); // expected-error@*:* {{static assertion failed: for_each}}
+    (void)std::for_each_n(pol, non_fwd, n, func);     // expected-error@*:* {{static assertion failed: for_each_n}}
+  }
+
+  {
+    (void)std::generate(pol, non_fwd, non_fwd, func); // expected-error@*:* {{static assertion failed: generate}}
+    (void)std::generate_n(pol, non_fwd, n, func);     // expected-error@*:* {{static assertion failed: generate_n}}
+  }
+
+  {
+    (void)std::is_partitioned(
+        pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: is_partitioned}}
+  }
+
+  {
+    (void)std::merge(pol, non_fwd, non_fwd, it, it, out); // expected-error@*:* {{static assertion failed: merge}}
+    (void)std::merge(pol, it, it, non_fwd, non_fwd, out); // expected-error@*:* {{static assertion failed: merge}}
+    (void)std::merge(pol, it, it, it, it, non_output);    // expected-error@*:* {{static assertion failed: merge}}
+
+    (void)std::merge(pol, non_fwd, non_fwd, it, it, out, pred); // expected-error@*:* {{static assertion failed: merge}}
+    (void)std::merge(pol, it, it, non_fwd, non_fwd, out, pred); // expected-error@*:* {{static assertion failed: merge}}
+    (void)std::merge(pol, it, it, it, it, non_output, pred);    // expected-error@*:* {{static assertion failed: merge}}
+  }
+
+  {
+    (void)std::move(pol, non_fwd, non_fwd, out); // expected-error@*:* {{static assertion failed: move}}
+    (void)std::move(pol, it, it, non_fwd);       // expected-error@*:* {{static assertion failed: move}}
+    (void)std::move(pol, it, it, non_output);    // expected-error@*:* {{static assertion failed: move}}
+  }
+
+  {
+    (void)std::replace_if(
+        pol, non_fwd, non_fwd, pred, val);               // expected-error@*:* {{static assertion failed: replace_if}}
+    (void)std::replace(pol, non_fwd, non_fwd, val, val); // expected-error@*:* {{static assertion failed: replace}}
+
+    (void)std::replace_copy_if(
+        pol, non_fwd, non_fwd, out, pred, val); // expected-error@*:* {{static assertion failed: replace_copy_if}}
+    (void)std::replace_copy_if(
+        pol, it, it, non_fwd, pred, val); // expected-error@*:* {{static assertion failed: replace_copy_if}}
+    (void)std::replace_copy_if(
+        pol, it, it, non_output, pred, val); // expected-error@*:* {{static assertion failed: replace_copy_if}}
+
+    (void)std::replace_copy(
+        pol, non_fwd, non_fwd, out, val, val); // expected-error@*:* {{static assertion failed: replace_copy}}
+    (void)std::replace_copy(
+        pol, it, it, non_fwd, val, val); // expected-error@*:* {{static assertion failed: replace_copy}}
+    (void)std::replace_copy(
+        pol, it, it, non_output, val, val); // expected-error@*:* {{static assertion failed: replace_copy}}
+  }
+
+  {
+    (void)std::rotate_copy(
+        pol, non_fwd, non_fwd, non_fwd, out);            // expected-error@*:* {{static assertion failed: rotate_copy}}
+    (void)std::rotate_copy(pol, it, it, it, non_fwd);    // expected-error@*:* {{static assertion failed: rotate_copy}}
+    (void)std::rotate_copy(pol, it, it, it, non_output); // expected-error@*:* {{static assertion failed: rotate_copy}}
+  }
+
+  {
+    (void)std::sort(pol, non_fwd, non_fwd);       // expected-error@*:* {{static assertion failed: sort}}
+    (void)std::sort(pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: sort}}
+  }
+
+  {
+    (void)std::stable_sort(pol, non_fwd, non_fwd);       // expected-error@*:* {{static assertion failed: stable_sort}}
+    (void)std::stable_sort(pol, non_fwd, non_fwd, pred); // expected-error@*:* {{static assertion failed: stable_sort}}
+  }
+
+  {
+    (void)std::transform(pol, non_fwd, non_fwd, out, func); // expected-error@*:* {{static assertion failed: transform}}
+    (void)std::transform(pol, it, it, non_fwd, func);       // expected-error@*:* {{static assertion failed: transform}}
+    (void)std::transform(pol, it, it, non_output, func);    // expected-error@*:* {{static assertion failed: transform}}
+
+    (void)std::transform(
+        pol, non_fwd, non_fwd, it, out, func);             // expected-error@*:* {{static assertion failed: transform}}
+    (void)std::transform(pol, it, it, non_fwd, out, func); // expected-error@*:* {{static assertion failed: transform}}
+    (void)std::transform(pol, it, it, it, non_fwd, func);  // expected-error@*:* {{static assertion failed: transform}}
+    (void)std::transform(
+        pol, it, it, it, non_output, func); // expected-error@*:* {{static assertion failed: transform}}
+  }
+
+  {
+    (void)std::reduce(pol, non_fwd, non_fwd);            // expected-error@*:* {{static assertion failed: reduce}}
+    (void)std::reduce(pol, non_fwd, non_fwd, val);       // expected-error@*:* {{static assertion failed: reduce}}
+    (void)std::reduce(pol, non_fwd, non_fwd, val, func); // expected-error@*:* {{static assertion failed: reduce}}
+  }
+
+  {
+    (void)std::transform_reduce(
+        pol, non_fwd, non_fwd, it, val); // expected-error@*:* {{static assertion failed: transform_reduce}}
+    (void)std::transform_reduce(
+        pol, it, it, non_fwd, val); // expected-error@*:* {{static assertion failed: transform_reduce}}
+
+    (void)std::transform_reduce(
+        pol, non_fwd, non_fwd, it, val, func, func); // expected-error@*:* {{static assertion failed: transform_reduce}}
+    (void)std::transform_reduce(
+        pol, it, it, non_fwd, val, func, func); // expected-error@*:* {{static assertion failed: transform_reduce}}
+
+    (void)std::transform_reduce(
+        pol, non_fwd, non_fwd, val, func, func); // expected-error@*:* {{static assertion failed: transform_reduce}}
+  }
+}
diff --git a/libcxx/test/support/test_iterators.h b/libcxx/test/support/test_iterators.h
index 7ffb74990fa4d..aa819ecd4733b 100644
--- a/libcxx/test/support/test_iterators.h
+++ b/libcxx/test/support/test_iterators.h
@@ -1484,9 +1484,14 @@ class iterator_wrapper {
     return tmp;
   }
 
-  iterator_wrapper& operator+=(difference_type i) {
+  Derived& operator+=(difference_type i) {
     iter_ += i;
-    return *this;
+    return static_cast<Derived&>(*this);
+  }
+
+  Derived& operator-=(difference_type i) {
+    iter_ -= i;
+    return static_cast<Derived&>(*this);
   }
 
   friend decltype(iter_ - iter_) operator-(const iterator_wrapper& lhs, const iterator_wrapper& rhs) {
@@ -1503,8 +1508,15 @@ class iterator_wrapper {
     return iter;
   }
 
+  friend Derived operator+(difference_type i, Derived iter) { return iter + i; }
+
   friend bool operator==(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ == rhs.iter_; }
   friend bool operator!=(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ != rhs.iter_; }
+
+  friend bool operator>(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ > rhs.iter_; }
+  friend bool operator<(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ < rhs.iter_; }
+  friend bool operator<=(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ <= rhs.iter_; }
+  friend bool operator>=(const iterator_wrapper& lhs, const iterator_wrapper& rhs) { return lhs.iter_ >= rhs.iter_; }
 };
 
 class iterator_error : std::runtime_error {

From 6c7853080451a7f9ba612f57b41a076061b2a1a9 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 17 Apr 2024 13:18:55 +0100
Subject: [PATCH 731/886] [X86] vector-shuffle-combining-sse41.ll - add missing
 AVX1/2/512 check prefixes

---
 .../X86/vector-shuffle-combining-sse41.ll     | 133 +++++++++++++++++-
 1 file changed, 129 insertions(+), 4 deletions(-)

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
index 33851f56fe8de..8d213d2577433 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-sse41.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=AVX,AVX512
 
 ; Combine tests involving SSE41 target shuffles (BLEND,INSERTPS,MOVZX)
 
@@ -29,6 +29,30 @@ define <4 x i32> @combine_blend_of_permutes_v4i32(<2 x i64> %a0, <2 x i64> %a1)
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3],xmm2[4,5],xmm0[6,7]
 ; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_blend_of_permutes_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_blend_of_permutes_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: combine_blend_of_permutes_v4i32:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    # kill: def $xmm1 killed $xmm1 def $zmm1
+; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [2,19,0,17]
+; AVX512-NEXT:    vpermt2d %zmm1, %zmm2, %zmm0
+; AVX512-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   %s0 = shufflevector <2 x i64> %a0, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
   %s1 = shufflevector <2 x i64> %a1, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
   %x0 = bitcast <2 x i64> %s0 to <4 x i32>
@@ -71,6 +95,107 @@ define <16 x i8> @PR50049(ptr %p1, ptr %p2) {
 ; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    retq
+;
+; AVX1-LABEL: PR50049:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
+; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [1,4,7,10,13,128,128,128,128,128,128,u,u,u,u,u]
+; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,0,3,6,9,12,15,u,u,u,u,u]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpor %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa (%rsi), %xmm2
+; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm5
+; AVX1-NEXT:    vmovdqa 32(%rsi), %xmm6
+; AVX1-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX1-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpor %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [5,6,7,8,9,10,128,128,128,128,128,0,1,2,3,4]
+; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpor %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX1-NEXT:    vpmullw %xmm5, %xmm1, %xmm1
+; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX1-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: PR50049:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
+; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [1,4,7,10,13,128,128,128,128,128,128,u,u,u,u,u]
+; AVX2-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,0,3,6,9,12,15,u,u,u,u,u]
+; AVX2-NEXT:    vpshufb %xmm4, %xmm0, %xmm0
+; AVX2-NEXT:    vpor %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa (%rsi), %xmm2
+; AVX2-NEXT:    vmovdqa 16(%rsi), %xmm5
+; AVX2-NEXT:    vmovdqa 32(%rsi), %xmm6
+; AVX2-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX2-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX2-NEXT:    vpor %xmm3, %xmm2, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [5,6,7,8,9,10,128,128,128,128,128,0,1,2,3,4]
+; AVX2-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128]
+; AVX2-NEXT:    vpshufb %xmm4, %xmm5, %xmm5
+; AVX2-NEXT:    vpor %xmm5, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX2-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: PR50049:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
+; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
+; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [1,4,7,10,13,128,128,128,128,128,128,u,u,u,u,u]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,0,3,6,9,12,15,u,u,u,u,u]
+; AVX512-NEXT:    vpshufb %xmm4, %xmm0, %xmm0
+; AVX512-NEXT:    vpor %xmm2, %xmm0, %xmm0
+; AVX512-NEXT:    vmovdqa (%rsi), %xmm2
+; AVX512-NEXT:    vmovdqa 16(%rsi), %xmm5
+; AVX512-NEXT:    vmovdqa 32(%rsi), %xmm6
+; AVX512-NEXT:    vpshufb %xmm3, %xmm6, %xmm3
+; AVX512-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
+; AVX512-NEXT:    vpor %xmm3, %xmm2, %xmm2
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [5,6,7,8,9,10,128,128,128,128,128,0,1,2,3,4]
+; AVX512-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX512-NEXT:    vmovdqa {{.*#+}} xmm4 = [128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128]
+; AVX512-NEXT:    vpshufb %xmm4, %xmm5, %xmm5
+; AVX512-NEXT:    vpor %xmm5, %xmm2, %xmm2
+; AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX512-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX512-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
+; AVX512-NEXT:    vpor %xmm1, %xmm0, %xmm0
+; AVX512-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
+; AVX512-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   %x1 = load <48 x i8>, ptr %p1, align 16
   %x2 = load <48 x i8>, ptr %p2, align 16
   %s1 = shufflevector <48 x i8> %x1, <48 x i8> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>

From 37b26bf48b9894ed0c13fd1aede23472660fb75e Mon Sep 17 00:00:00 2001
From: "Oleksandr \"Alex\" Zinenko" <zinenko@google.com>
Date: Wed, 17 Apr 2024 14:24:51 +0200
Subject: [PATCH 732/886] [mlir] transform.apply_patterns support more config
 options (#88484)

Greedy rewrite driver has options to control the number of rewrites
applies. Expose those via the corresponding transform op.
---
 .../mlir/Dialect/Transform/IR/TransformOps.td |  5 +++-
 .../lib/Dialect/Transform/IR/TransformOps.cpp |  7 +++++
 .../Transform/test-pattern-application.mlir   | 30 +++++++++++++++++++
 3 files changed, 41 insertions(+), 1 deletion(-)

diff --git a/mlir/include/mlir/Dialect/Transform/IR/TransformOps.td b/mlir/include/mlir/Dialect/Transform/IR/TransformOps.td
index 21c9595860d4c..fbac1ffb621fd 100644
--- a/mlir/include/mlir/Dialect/Transform/IR/TransformOps.td
+++ b/mlir/include/mlir/Dialect/Transform/IR/TransformOps.td
@@ -331,7 +331,10 @@ def ApplyPatternsOp : TransformDialectOp<"apply_patterns",
   }];
 
   let arguments = (ins
-    TransformHandleTypeInterface:$target, UnitAttr:$apply_cse);
+    TransformHandleTypeInterface:$target,
+    UnitAttr:$apply_cse,
+    DefaultValuedAttr<I64Attr, "static_cast<uint64_t>(-1)">:$max_iterations,
+    DefaultValuedAttr<I64Attr, "static_cast<uint64_t>(-1)">:$max_num_rewrites);
   let results = (outs);
   let regions = (region MaxSizedRegion<1>:$patterns);
 
diff --git a/mlir/lib/Dialect/Transform/IR/TransformOps.cpp b/mlir/lib/Dialect/Transform/IR/TransformOps.cpp
index dc19022219e5b..53f958caa0bdb 100644
--- a/mlir/lib/Dialect/Transform/IR/TransformOps.cpp
+++ b/mlir/lib/Dialect/Transform/IR/TransformOps.cpp
@@ -396,6 +396,13 @@ DiagnosedSilenceableFailure transform::ApplyPatternsOp::applyToOne(
       static_cast<RewriterBase::Listener *>(rewriter.getListener());
   FrozenRewritePatternSet frozenPatterns(std::move(patterns));
 
+  config.maxIterations = getMaxIterations() == static_cast<uint64_t>(-1)
+                             ? GreedyRewriteConfig::kNoLimit
+                             : getMaxIterations();
+  config.maxNumRewrites = getMaxNumRewrites() == static_cast<uint64_t>(-1)
+                              ? GreedyRewriteConfig::kNoLimit
+                              : getMaxNumRewrites();
+
   // Apply patterns and CSE repetitively until a fixpoint is reached. If no CSE
   // was requested, apply the greedy pattern rewrite only once. (The greedy
   // pattern rewrite driver already iterates to a fixpoint internally.)
diff --git a/mlir/test/Dialect/Transform/test-pattern-application.mlir b/mlir/test/Dialect/Transform/test-pattern-application.mlir
index fa8a555af9218..f78b4b6f6798c 100644
--- a/mlir/test/Dialect/Transform/test-pattern-application.mlir
+++ b/mlir/test/Dialect/Transform/test-pattern-application.mlir
@@ -26,6 +26,36 @@ module attributes {transform.with_named_sequence} {
 
 // -----
 
+// CHECK-LABEL: @limited_updates
+func.func @limited_updates() {
+  "test.container"() ({
+    // Only one is replaced.
+    // CHECK: "test.foo"() {replace_with_new_op = "test.foo"}
+    // CHECK: "test.foo"() : ()
+    %0 = "test.foo"() {replace_with_new_op = "test.foo"} : () -> (i32)
+    %1 = "test.foo"() {replace_with_new_op = "test.foo"} : () -> (i32)
+  }) : () -> ()
+  return
+}
+
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @__transform_main(%arg0: !transform.any_op) {
+    // Pattern application will fail because of the upper limit, wrap in
+    // sequence to suppress the error message.
+    transform.sequence %arg0 : !transform.any_op failures(suppress) {
+    ^bb0(%arg1: !transform.any_op):
+      %0 = transform.structured.match ops{["test.container"]} in %arg1 : (!transform.any_op) -> !transform.any_op
+      %1 = transform.structured.match ops{["test.foo"]} in %arg1 : (!transform.any_op) -> !transform.any_op
+      transform.apply_patterns to %0 {
+        transform.apply_patterns.transform.test_patterns
+      }  {max_num_rewrites = 1} : !transform.any_op
+    }
+    transform.yield
+  }
+}
+
+// -----
+
 func.func @replacement_op_not_found() {
   "test.container"() ({
     // expected-note @below {{[0] replaced op}}

From 4536ad47579d8d61f372ab85128bcfaed58a1256 Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Wed, 17 Apr 2024 20:24:59 +0800
Subject: [PATCH 733/886] [RISCV] Fix clang-tidy warning about else after
 return. NFC

---
 llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
index 6e45f0c703ceb..aab91adbb64be 100644
--- a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
@@ -1551,7 +1551,9 @@ void RISCVInsertVSETVLI::doLocalPostpass(MachineBasicBlock &MBB) {
         ToDelete.push_back(&MI);
         // Leave NextMI unchanged
         continue;
-      } else if (canMutatePriorConfig(MI, *NextMI, Used, *MRI)) {
+      }
+
+      if (canMutatePriorConfig(MI, *NextMI, Used, *MRI)) {
         if (!isVLPreservingConfig(*NextMI)) {
           MI.getOperand(0).setReg(NextMI->getOperand(0).getReg());
           MI.getOperand(0).setIsDead(false);

From 86a78284e7ce2ecc7a9283c7d141566a32371492 Mon Sep 17 00:00:00 2001
From: Quentin Dian <dianqk@dianqk.net>
Date: Wed, 17 Apr 2024 20:27:09 +0800
Subject: [PATCH 734/886] [TailDuplicator] Add maximum predecessors and
 successors to consider tail duplicating blocks (#78582)

Fixes #78578.

Duplicating a BB which has both multiple predecessors and successors
will result in a complex CFG and also may cause huge amount of PHI
nodes. See
https://github.com/llvm/llvm-project/issues/78578#issuecomment-1962363580
for a detailed description of the limit.
---
 llvm/lib/CodeGen/TailDuplicator.cpp           |  20 ++
 .../CodeGen/X86/tail-dup-pred-succ-size.mir   | 260 ++++++++++++++++++
 2 files changed, 280 insertions(+)
 create mode 100644 llvm/test/CodeGen/X86/tail-dup-pred-succ-size.mir

diff --git a/llvm/lib/CodeGen/TailDuplicator.cpp b/llvm/lib/CodeGen/TailDuplicator.cpp
index 5ed67bd0a121e..f5dd21cb92701 100644
--- a/llvm/lib/CodeGen/TailDuplicator.cpp
+++ b/llvm/lib/CodeGen/TailDuplicator.cpp
@@ -68,6 +68,18 @@ static cl::opt<unsigned> TailDupIndirectBranchSize(
              "end with indirect branches."), cl::init(20),
     cl::Hidden);
 
+static cl::opt<unsigned>
+    TailDupPredSize("tail-dup-pred-size",
+                    cl::desc("Maximum predecessors (maximum successors at the "
+                             "same time) to consider tail duplicating blocks."),
+                    cl::init(16), cl::Hidden);
+
+static cl::opt<unsigned>
+    TailDupSuccSize("tail-dup-succ-size",
+                    cl::desc("Maximum successors (maximum predecessors at the "
+                             "same time) to consider tail duplicating blocks."),
+                    cl::init(16), cl::Hidden);
+
 static cl::opt<bool>
     TailDupVerify("tail-dup-verify",
                   cl::desc("Verify sanity of PHI instructions during taildup"),
@@ -565,6 +577,14 @@ bool TailDuplicator::shouldTailDuplicate(bool IsSimple,
   if (TailBB.isSuccessor(&TailBB))
     return false;
 
+  // Duplicating a BB which has both multiple predecessors and successors will
+  // result in a complex CFG and also may cause huge amount of PHI nodes. If we
+  // want to remove this limitation, we have to address
+  // https://github.com/llvm/llvm-project/issues/78578.
+  if (TailBB.pred_size() > TailDupPredSize &&
+      TailBB.succ_size() > TailDupSuccSize)
+    return false;
+
   // Set the limit on the cost to duplicate. When optimizing for size,
   // duplicate only one, because one branch instruction can be eliminated to
   // compensate for the duplication.
diff --git a/llvm/test/CodeGen/X86/tail-dup-pred-succ-size.mir b/llvm/test/CodeGen/X86/tail-dup-pred-succ-size.mir
new file mode 100644
index 0000000000000..67f8cc72e0d72
--- /dev/null
+++ b/llvm/test/CodeGen/X86/tail-dup-pred-succ-size.mir
@@ -0,0 +1,260 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=early-tailduplication -tail-dup-pred-size=3 -tail-dup-succ-size=3 %s -o - | FileCheck %s -check-prefix=LIMIT
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=early-tailduplication -tail-dup-pred-size=4 -tail-dup-succ-size=4 %s -o - | FileCheck %s -check-prefix=NOLIMIT
+
+---
+name:            foo
+tracksRegLiveness: true
+jumpTable:
+  kind:            block-address
+  entries:
+    - id:              0
+      blocks:          [ '%bb.2', '%bb.3', '%bb.4', '%bb.5' ]
+    - id:              1
+      blocks:          [ '%bb.9', '%bb.10', '%bb.11', '%bb.12' ]
+body:             |
+  ; LIMIT-LABEL: name: foo
+  ; LIMIT: bb.0:
+  ; LIMIT-NEXT:   successors: %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000), %bb.5(0x20000000)
+  ; LIMIT-NEXT:   liveins: $rdi, $esi
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[COPY:%[0-9]+]]:gr32 = COPY $esi
+  ; LIMIT-NEXT:   [[COPY1:%[0-9]+]]:gr64 = COPY $rdi
+  ; LIMIT-NEXT:   [[SHR32ri:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 1, implicit-def dead $eflags
+  ; LIMIT-NEXT:   [[AND32ri:%[0-9]+]]:gr32 = AND32ri [[SHR32ri]], 7, implicit-def dead $eflags
+  ; LIMIT-NEXT:   [[SUBREG_TO_REG:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, killed [[AND32ri]], %subreg.sub_32bit
+  ; LIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG]], %jump-table.0, $noreg
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.2:
+  ; LIMIT-NEXT:   successors: %bb.7(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   JMP_1 %bb.7
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.3:
+  ; LIMIT-NEXT:   successors: %bb.7(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm1:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri1:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm1]], 1, implicit-def dead $eflags
+  ; LIMIT-NEXT:   JMP_1 %bb.7
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.4:
+  ; LIMIT-NEXT:   successors: %bb.7(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm2:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri2:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm2]], 2, implicit-def dead $eflags
+  ; LIMIT-NEXT:   JMP_1 %bb.7
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.5:
+  ; LIMIT-NEXT:   successors: %bb.7(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm3:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri3:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm3]], 3, implicit-def dead $eflags
+  ; LIMIT-NEXT:   JMP_1 %bb.7
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.6:
+  ; LIMIT-NEXT:   successors:
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.7:
+  ; LIMIT-NEXT:   successors: %bb.9(0x20000000), %bb.10(0x20000000), %bb.11(0x20000000), %bb.12(0x20000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[PHI:%[0-9]+]]:gr32 = PHI [[SHR32ri3]], %bb.5, [[SHR32ri2]], %bb.4, [[SHR32ri1]], %bb.3, [[MOV32rm]], %bb.2
+  ; LIMIT-NEXT:   [[SHR32ri4:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 2, implicit-def dead $eflags
+  ; LIMIT-NEXT:   [[AND32ri1:%[0-9]+]]:gr32 = AND32ri [[SHR32ri4]], 7, implicit-def dead $eflags
+  ; LIMIT-NEXT:   [[SUBREG_TO_REG1:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, killed [[AND32ri1]], %subreg.sub_32bit
+  ; LIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG1]], %jump-table.1, $noreg
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.9:
+  ; LIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm4:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   JMP_1 %bb.13
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.10:
+  ; LIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm5:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri5:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm5]], 1, implicit-def dead $eflags
+  ; LIMIT-NEXT:   JMP_1 %bb.13
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.11:
+  ; LIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm6:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri6:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm6]], 2, implicit-def dead $eflags
+  ; LIMIT-NEXT:   JMP_1 %bb.13
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.12:
+  ; LIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT:   [[MOV32rm7:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; LIMIT-NEXT:   [[SHR32ri7:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm7]], 6, implicit-def dead $eflags
+  ; LIMIT-NEXT: {{  $}}
+  ; LIMIT-NEXT: bb.13:
+  ; LIMIT-NEXT:   [[PHI1:%[0-9]+]]:gr32 = PHI [[SHR32ri7]], %bb.12, [[SHR32ri6]], %bb.11, [[SHR32ri5]], %bb.10, [[MOV32rm4]], %bb.9
+  ; LIMIT-NEXT:   [[OR32rr:%[0-9]+]]:gr32 = OR32rr [[PHI1]], [[PHI]], implicit-def dead $eflags
+  ; LIMIT-NEXT:   $eax = COPY [[OR32rr]]
+  ; LIMIT-NEXT:   RET 0, $eax
+  ;
+  ; NOLIMIT-LABEL: name: foo
+  ; NOLIMIT: bb.0:
+  ; NOLIMIT-NEXT:   successors: %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000), %bb.5(0x20000000)
+  ; NOLIMIT-NEXT:   liveins: $rdi, $esi
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[COPY:%[0-9]+]]:gr32 = COPY $esi
+  ; NOLIMIT-NEXT:   [[COPY1:%[0-9]+]]:gr64 = COPY $rdi
+  ; NOLIMIT-NEXT:   [[SHR32ri:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 1, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[AND32ri:%[0-9]+]]:gr32 = AND32ri [[SHR32ri]], 7, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SUBREG_TO_REG:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, killed [[AND32ri]], %subreg.sub_32bit
+  ; NOLIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG]], %jump-table.0, $noreg
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.2:
+  ; NOLIMIT-NEXT:   successors: %bb.9(0x20000000), %bb.10(0x20000000), %bb.11(0x20000000), %bb.12(0x20000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[MOV32rm:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri1:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[AND32ri1:%[0-9]+]]:gr32 = AND32ri [[SHR32ri1]], 7, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SUBREG_TO_REG1:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, [[AND32ri1]], %subreg.sub_32bit
+  ; NOLIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG1]], %jump-table.1, $noreg
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.3:
+  ; NOLIMIT-NEXT:   successors: %bb.9(0x20000000), %bb.10(0x20000000), %bb.11(0x20000000), %bb.12(0x20000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[MOV32rm1:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri2:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm1]], 1, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SHR32ri3:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[AND32ri2:%[0-9]+]]:gr32 = AND32ri [[SHR32ri3]], 7, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SUBREG_TO_REG2:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, [[AND32ri2]], %subreg.sub_32bit
+  ; NOLIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG2]], %jump-table.1, $noreg
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.4:
+  ; NOLIMIT-NEXT:   successors: %bb.9(0x20000000), %bb.10(0x20000000), %bb.11(0x20000000), %bb.12(0x20000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[MOV32rm2:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri4:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm2]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SHR32ri5:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[AND32ri3:%[0-9]+]]:gr32 = AND32ri [[SHR32ri5]], 7, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SUBREG_TO_REG3:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, [[AND32ri3]], %subreg.sub_32bit
+  ; NOLIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG3]], %jump-table.1, $noreg
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.5:
+  ; NOLIMIT-NEXT:   successors: %bb.9(0x20000000), %bb.10(0x20000000), %bb.11(0x20000000), %bb.12(0x20000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[MOV32rm3:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri6:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm3]], 3, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SHR32ri7:%[0-9]+]]:gr32 = SHR32ri [[COPY]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[AND32ri4:%[0-9]+]]:gr32 = AND32ri [[SHR32ri7]], 7, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   [[SUBREG_TO_REG4:%[0-9]+]]:gr64_nosp = SUBREG_TO_REG 0, [[AND32ri4]], %subreg.sub_32bit
+  ; NOLIMIT-NEXT:   JMP64m $noreg, 8, [[SUBREG_TO_REG4]], %jump-table.1, $noreg
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.6:
+  ; NOLIMIT-NEXT:   successors:
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.9:
+  ; NOLIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[PHI:%[0-9]+]]:gr32 = PHI [[MOV32rm]], %bb.2, [[SHR32ri2]], %bb.3, [[SHR32ri4]], %bb.4, [[SHR32ri6]], %bb.5
+  ; NOLIMIT-NEXT:   [[MOV32rm4:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   JMP_1 %bb.13
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.10:
+  ; NOLIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[PHI1:%[0-9]+]]:gr32 = PHI [[MOV32rm]], %bb.2, [[SHR32ri2]], %bb.3, [[SHR32ri4]], %bb.4, [[SHR32ri6]], %bb.5
+  ; NOLIMIT-NEXT:   [[MOV32rm5:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri8:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm5]], 1, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   JMP_1 %bb.13
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.11:
+  ; NOLIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[PHI2:%[0-9]+]]:gr32 = PHI [[MOV32rm]], %bb.2, [[SHR32ri2]], %bb.3, [[SHR32ri4]], %bb.4, [[SHR32ri6]], %bb.5
+  ; NOLIMIT-NEXT:   [[MOV32rm6:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri9:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm6]], 2, implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   JMP_1 %bb.13
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.12:
+  ; NOLIMIT-NEXT:   successors: %bb.13(0x80000000)
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT:   [[PHI3:%[0-9]+]]:gr32 = PHI [[MOV32rm]], %bb.2, [[SHR32ri2]], %bb.3, [[SHR32ri4]], %bb.4, [[SHR32ri6]], %bb.5
+  ; NOLIMIT-NEXT:   [[MOV32rm7:%[0-9]+]]:gr32 = MOV32rm [[COPY1]], 1, $noreg, 0, $noreg
+  ; NOLIMIT-NEXT:   [[SHR32ri10:%[0-9]+]]:gr32 = SHR32ri [[MOV32rm7]], 6, implicit-def dead $eflags
+  ; NOLIMIT-NEXT: {{  $}}
+  ; NOLIMIT-NEXT: bb.13:
+  ; NOLIMIT-NEXT:   [[PHI4:%[0-9]+]]:gr32 = PHI [[PHI]], %bb.9, [[PHI1]], %bb.10, [[PHI2]], %bb.11, [[PHI3]], %bb.12
+  ; NOLIMIT-NEXT:   [[PHI5:%[0-9]+]]:gr32 = PHI [[SHR32ri10]], %bb.12, [[SHR32ri9]], %bb.11, [[SHR32ri8]], %bb.10, [[MOV32rm4]], %bb.9
+  ; NOLIMIT-NEXT:   [[OR32rr:%[0-9]+]]:gr32 = OR32rr [[PHI5]], [[PHI4]], implicit-def dead $eflags
+  ; NOLIMIT-NEXT:   $eax = COPY [[OR32rr]]
+  ; NOLIMIT-NEXT:   RET 0, $eax
+  bb.0:
+    liveins: $rdi, $esi
+
+    %11:gr32 = COPY $esi
+    %10:gr64 = COPY $rdi
+    %13:gr32 = SHR32ri %11, 1, implicit-def dead $eflags
+    %14:gr32 = AND32ri %13, 7, implicit-def dead $eflags
+    %12:gr64_nosp = SUBREG_TO_REG 0, killed %14, %subreg.sub_32bit
+
+  bb.1:
+    successors: %bb.2, %bb.3, %bb.4, %bb.5
+
+    JMP64m $noreg, 8, %12, %jump-table.0, $noreg
+
+  bb.2:
+    %0:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    JMP_1 %bb.7
+
+  bb.3:
+    %17:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %1:gr32 = SHR32ri %17, 1, implicit-def dead $eflags
+    JMP_1 %bb.7
+
+  bb.4:
+    %16:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %2:gr32 = SHR32ri %16, 2, implicit-def dead $eflags
+    JMP_1 %bb.7
+
+  bb.5:
+    %15:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %3:gr32 = SHR32ri %15, 3, implicit-def dead $eflags
+    JMP_1 %bb.7
+
+  bb.6:
+    successors:
+
+  bb.7:
+    %4:gr32 = PHI %3, %bb.5, %2, %bb.4, %1, %bb.3, %0, %bb.2
+    %19:gr32 = SHR32ri %11, 2, implicit-def dead $eflags
+    %20:gr32 = AND32ri %19, 7, implicit-def dead $eflags
+    %18:gr64_nosp = SUBREG_TO_REG 0, killed %20, %subreg.sub_32bit
+
+  bb.8:
+    successors: %bb.9, %bb.10, %bb.11, %bb.12
+
+    JMP64m $noreg, 8, %18, %jump-table.1, $noreg
+
+  bb.9:
+    %5:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    JMP_1 %bb.13
+
+  bb.10:
+    %23:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %6:gr32 = SHR32ri %23, 1, implicit-def dead $eflags
+    JMP_1 %bb.13
+
+  bb.11:
+    %22:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %7:gr32 = SHR32ri %22, 2, implicit-def dead $eflags
+    JMP_1 %bb.13
+
+  bb.12:
+    %21:gr32 = MOV32rm %10, 1, $noreg, 0, $noreg
+    %8:gr32 = SHR32ri %21, 6, implicit-def dead $eflags
+
+  bb.13:
+    %9:gr32 = PHI %8, %bb.12, %7, %bb.11, %6, %bb.10, %5, %bb.9
+    %24:gr32 = OR32rr %9, %4, implicit-def dead $eflags
+    $eax = COPY %24
+    RET 0, $eax
+
+...

From 915c84b1480bb3c6d2e44ca83822d2c2304b763a Mon Sep 17 00:00:00 2001
From: Pavel Labath <pavel@labath.sk>
Date: Wed, 17 Apr 2024 14:33:18 +0200
Subject: [PATCH 735/886] [lldb] Fix evaluation of expressions with static
 initializers (#89063)

After 281d71604f418eb952e967d9dc4b26241b7f96a, llvm generates 32-bit
relocations, which overflow when we load these objects into high memory.
Interestingly, setting the code model to "large" does not help here
(perhaps it is the default?). I'm not completely sure that this is the
right thing to do, but it doesn't seem to cause any ill effects. I'll
follow up with the author of that patch about the expected behavior
here.
---
 lldb/source/Expression/IRExecutionUnit.cpp | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/lldb/source/Expression/IRExecutionUnit.cpp b/lldb/source/Expression/IRExecutionUnit.cpp
index cb9bee8733e15..7ad0e5ff22b2f 100644
--- a/lldb/source/Expression/IRExecutionUnit.cpp
+++ b/lldb/source/Expression/IRExecutionUnit.cpp
@@ -13,6 +13,7 @@
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
+#include "llvm/Support/CodeGen.h"
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/raw_ostream.h"
 
@@ -279,10 +280,13 @@ void IRExecutionUnit::GetRunnableInfo(Status &error, lldb::addr_t &func_addr,
   llvm::EngineBuilder builder(std::move(m_module_up));
   llvm::Triple triple(m_module->getTargetTriple());
 
+  // PIC needed for ELF to avoid generating 32-bit relocations (which overflow
+  // if the object is loaded into high memory).
+  bool want_pic = triple.isOSBinFormatMachO() || triple.isOSBinFormatELF();
+
   builder.setEngineKind(llvm::EngineKind::JIT)
       .setErrorStr(&error_string)
-      .setRelocationModel(triple.isOSBinFormatMachO() ? llvm::Reloc::PIC_
-                                                      : llvm::Reloc::Static)
+      .setRelocationModel(want_pic ? llvm::Reloc::PIC_ : llvm::Reloc::Static)
       .setMCJITMemoryManager(std::make_unique<MemoryManager>(*this))
       .setOptLevel(llvm::CodeGenOptLevel::Less);
 

From 79726ef5d20f3bf6510b67d7dd3378b8a41db768 Mon Sep 17 00:00:00 2001
From: "Kevin P. Neal" <kevin.neal@sas.com>
Date: Wed, 17 Apr 2024 08:34:25 -0400
Subject: [PATCH 736/886] [VP] Correct lowering of predicated fma and faddmul
 to avoid strictfp. (#85272)

Correct missing cases in a switch that result in @llvm.vp.fma.v4f32
getting lowered to a constrained fma intrinsic. Vector predicated
lowering to contrained intrinsics is not supported currently, and
there's no consensus on the path forward. We certainly shouldn't be
introducing constrained intrinsics into a function that isn't strictfp.

Problem found with D146845.
---
 llvm/include/llvm/IR/Intrinsics.h             |   4 +
 llvm/lib/CodeGen/ExpandVectorPredication.cpp  |  12 +-
 llvm/lib/IR/Function.cpp                      |  22 ++-
 .../Generic/expand-vp-fp-intrinsics.ll        | 176 ++++++++++++++++++
 4 files changed, 203 insertions(+), 11 deletions(-)
 create mode 100644 llvm/test/CodeGen/Generic/expand-vp-fp-intrinsics.ll

diff --git a/llvm/include/llvm/IR/Intrinsics.h b/llvm/include/llvm/IR/Intrinsics.h
index 0dfe9f029f9b1..92eae344ce729 100644
--- a/llvm/include/llvm/IR/Intrinsics.h
+++ b/llvm/include/llvm/IR/Intrinsics.h
@@ -105,6 +105,10 @@ namespace Intrinsic {
   /// Map a MS builtin name to an intrinsic ID.
   ID getIntrinsicForMSBuiltin(const char *Prefix, StringRef BuiltinName);
 
+  /// Returns true if the intrinsic ID is for one of the "Constrained
+  /// Floating-Point Intrinsics".
+  bool isConstrainedFPIntrinsic(ID QID);
+
   /// This is a type descriptor which explains the type requirements of an
   /// intrinsic. This is returned by getIntrinsicInfoTableEntries.
   struct IITDescriptor {
diff --git a/llvm/lib/CodeGen/ExpandVectorPredication.cpp b/llvm/lib/CodeGen/ExpandVectorPredication.cpp
index 0fe4cfefdb160..8e623c85b737b 100644
--- a/llvm/lib/CodeGen/ExpandVectorPredication.cpp
+++ b/llvm/lib/CodeGen/ExpandVectorPredication.cpp
@@ -340,6 +340,8 @@ Value *CachingVPExpander::expandPredicationToFPCall(
     replaceOperation(*NewOp, VPI);
     return NewOp;
   }
+  case Intrinsic::fma:
+  case Intrinsic::fmuladd:
   case Intrinsic::experimental_constrained_fma:
   case Intrinsic::experimental_constrained_fmuladd: {
     Value *Op0 = VPI.getOperand(0);
@@ -347,8 +349,12 @@ Value *CachingVPExpander::expandPredicationToFPCall(
     Value *Op2 = VPI.getOperand(2);
     Function *Fn = Intrinsic::getDeclaration(
         VPI.getModule(), UnpredicatedIntrinsicID, {VPI.getType()});
-    Value *NewOp =
-        Builder.CreateConstrainedFPCall(Fn, {Op0, Op1, Op2}, VPI.getName());
+    Value *NewOp;
+    if (Intrinsic::isConstrainedFPIntrinsic(UnpredicatedIntrinsicID))
+      NewOp =
+          Builder.CreateConstrainedFPCall(Fn, {Op0, Op1, Op2}, VPI.getName());
+    else
+      NewOp = Builder.CreateCall(Fn, {Op0, Op1, Op2}, VPI.getName());
     replaceOperation(*NewOp, VPI);
     return NewOp;
   }
@@ -731,6 +737,8 @@ Value *CachingVPExpander::expandPredication(VPIntrinsic &VPI) {
   case Intrinsic::vp_minnum:
   case Intrinsic::vp_maximum:
   case Intrinsic::vp_minimum:
+  case Intrinsic::vp_fma:
+  case Intrinsic::vp_fmuladd:
     return expandPredicationToFPCall(Builder, VPI,
                                      VPI.getFunctionalIntrinsicID().value());
   case Intrinsic::vp_load:
diff --git a/llvm/lib/IR/Function.cpp b/llvm/lib/IR/Function.cpp
index 96953ac49c19b..818a167560de6 100644
--- a/llvm/lib/IR/Function.cpp
+++ b/llvm/lib/IR/Function.cpp
@@ -499,15 +499,7 @@ static MutableArrayRef<Argument> makeArgArray(Argument *Args, size_t Count) {
 }
 
 bool Function::isConstrainedFPIntrinsic() const {
-  switch (getIntrinsicID()) {
-#define INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC)                         \
-  case Intrinsic::INTRINSIC:
-#include "llvm/IR/ConstrainedOps.def"
-    return true;
-#undef INSTRUCTION
-  default:
-    return false;
-  }
+  return Intrinsic::isConstrainedFPIntrinsic(getIntrinsicID());
 }
 
 void Function::clearArguments() {
@@ -1486,6 +1478,18 @@ Function *Intrinsic::getDeclaration(Module *M, ID id, ArrayRef<Type*> Tys) {
 #include "llvm/IR/IntrinsicImpl.inc"
 #undef GET_LLVM_INTRINSIC_FOR_MS_BUILTIN
 
+bool Intrinsic::isConstrainedFPIntrinsic(ID QID) {
+  switch (QID) {
+#define INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC)                         \
+  case Intrinsic::INTRINSIC:
+#include "llvm/IR/ConstrainedOps.def"
+    return true;
+#undef INSTRUCTION
+  default:
+    return false;
+  }
+}
+
 using DeferredIntrinsicMatchPair =
     std::pair<Type *, ArrayRef<Intrinsic::IITDescriptor>>;
 
diff --git a/llvm/test/CodeGen/Generic/expand-vp-fp-intrinsics.ll b/llvm/test/CodeGen/Generic/expand-vp-fp-intrinsics.ll
new file mode 100644
index 0000000000000..bc89ddea6b85a
--- /dev/null
+++ b/llvm/test/CodeGen/Generic/expand-vp-fp-intrinsics.ll
@@ -0,0 +1,176 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -expandvp -S < %s | FileCheck %s
+
+define void @vp_fadd_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fadd_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:    [[RES1:%.*]] = fadd <4 x float> [[A0]], [[A1]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fadd.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fadd.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_fsub_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fsub_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = fsub <4 x float> [[A0]], [[A1]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fsub.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fsub.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_fmul_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fmul_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = fmul <4 x float> [[A0]], [[A1]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fmul.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fmul.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_fdiv_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fdiv_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = fdiv <4 x float> [[A0]], [[A1]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fdiv.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fdiv.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_frem_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_frem_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = frem <4 x float> [[A0]], [[A1]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.frem.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.frem.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_fabs_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fabs_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[A0]])
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fabs.v4f32(<4 x float> %a0, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fabs.v4f32(<4 x float>, <4 x i1>, i32)
+
+define void @vp_sqrt_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_sqrt_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> [[A0]])
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.sqrt.v4f32(<4 x float> %a0, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.sqrt.v4f32(<4 x float>, <4 x i1>, i32)
+
+define void @vp_fneg_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i32 %vp) nounwind {
+; CHECK-LABEL: define void @vp_fneg_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i32 [[VP:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = fneg <4 x float> [[A0]]
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> %a0, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 %vp)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fneg.v4f32(<4 x float>, <4 x i1>, i32)
+
+define void @vp_fma_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i4 %a5) nounwind {
+; CHECK-LABEL: define void @vp_fma_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i4 [[A5:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = call <4 x float> @llvm.fma.v4f32(<4 x float> [[A0]], <4 x float> [[A1]], <4 x float> [[A1]])
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 4)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, <4 x i1>, i32)
+
+define void @vp_fmuladd_v4f32(<4 x float> %a0, <4 x float> %a1, ptr %out, i4 %a5) nounwind {
+; CHECK-LABEL: define void @vp_fmuladd_v4f32(
+; CHECK-SAME: <4 x float> [[A0:%.*]], <4 x float> [[A1:%.*]], ptr [[OUT:%.*]], i4 [[A5:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:    [[RES1:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[A0]], <4 x float> [[A1]], <4 x float> [[A1]])
+; CHECK-NEXT:    store <4 x float> [[RES1]], ptr [[OUT]], align 16
+; CHECK-NEXT:    ret void
+;
+  %res = call <4 x float> @llvm.vp.fmuladd.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a1, <4 x i1> <i1 -1, i1 -1, i1 -1, i1 -1>, i32 4)
+  store <4 x float> %res, ptr %out
+  ret void
+}
+declare <4 x float> @llvm.vp.fmuladd.v4f32(<4 x float>, <4 x float>, <4 x float>, <4 x i1>, i32)
+
+declare <4 x float> @llvm.vp.maxnum.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+define <4 x float> @vfmax_vv_v4f32(<4 x float> %va, <4 x float> %vb, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: define <4 x float> @vfmax_vv_v4f32(
+; CHECK-SAME: <4 x float> [[VA:%.*]], <4 x float> [[VB:%.*]], <4 x i1> [[M:%.*]], i32 zeroext [[EVL:%.*]]) {
+; CHECK-NEXT:    [[V1:%.*]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VA]], <4 x float> [[VB]])
+; CHECK-NEXT:    ret <4 x float> [[V1]]
+;
+  %v = call <4 x float> @llvm.vp.maxnum.v4f32(<4 x float> %va, <4 x float> %vb, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <8 x float> @llvm.vp.maxnum.v8f32(<8 x float>, <8 x float>, <8 x i1>, i32)
+define <8 x float> @vfmax_vv_v8f32(<8 x float> %va, <8 x float> %vb, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: define <8 x float> @vfmax_vv_v8f32(
+; CHECK-SAME: <8 x float> [[VA:%.*]], <8 x float> [[VB:%.*]], <8 x i1> [[M:%.*]], i32 zeroext [[EVL:%.*]]) {
+; CHECK-NEXT:    [[V1:%.*]] = call <8 x float> @llvm.maxnum.v8f32(<8 x float> [[VA]], <8 x float> [[VB]])
+; CHECK-NEXT:    ret <8 x float> [[V1]]
+;
+  %v = call <8 x float> @llvm.vp.maxnum.v8f32(<8 x float> %va, <8 x float> %vb, <8 x i1> %m, i32 %evl)
+  ret <8 x float> %v
+}
+
+declare <4 x float> @llvm.vp.minnum.v4f32(<4 x float>, <4 x float>, <4 x i1>, i32)
+define <4 x float> @vfmin_vv_v4f32(<4 x float> %va, <4 x float> %vb, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: define <4 x float> @vfmin_vv_v4f32(
+; CHECK-SAME: <4 x float> [[VA:%.*]], <4 x float> [[VB:%.*]], <4 x i1> [[M:%.*]], i32 zeroext [[EVL:%.*]]) {
+; CHECK-NEXT:    [[V1:%.*]] = call <4 x float> @llvm.minnum.v4f32(<4 x float> [[VA]], <4 x float> [[VB]])
+; CHECK-NEXT:    ret <4 x float> [[V1]]
+;
+  %v = call <4 x float> @llvm.vp.minnum.v4f32(<4 x float> %va, <4 x float> %vb, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <8 x float> @llvm.vp.minnum.v8f32(<8 x float>, <8 x float>, <8 x i1>, i32)
+define <8 x float> @vfmin_vv_v8f32(<8 x float> %va, <8 x float> %vb, <8 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: define <8 x float> @vfmin_vv_v8f32(
+; CHECK-SAME: <8 x float> [[VA:%.*]], <8 x float> [[VB:%.*]], <8 x i1> [[M:%.*]], i32 zeroext [[EVL:%.*]]) {
+; CHECK-NEXT:    [[V1:%.*]] = call <8 x float> @llvm.minnum.v8f32(<8 x float> [[VA]], <8 x float> [[VB]])
+; CHECK-NEXT:    ret <8 x float> [[V1]]
+;
+  %v = call <8 x float> @llvm.vp.minnum.v8f32(<8 x float> %va, <8 x float> %vb, <8 x i1> %m, i32 %evl)
+  ret <8 x float> %v
+}

From 7b8625ec16efcb6d11e14a80e08c65dbfe68db9f Mon Sep 17 00:00:00 2001
From: Fabian Ritter <fabian.ritter@amd.com>
Date: Wed, 17 Apr 2024 14:54:14 +0200
Subject: [PATCH 737/886] [AMDGPU][Docs] Fix broken link to HRF memory model
 reference (#88696)

The link to the Heterogeneous-race-free Memory Models ASPLOS'14 paper by
Hower et al. pointed to a bogus website, probably because the domain
ownership has changed.
This patch updates it to a version hosted on research.cs.wisc.edu.
---
 llvm/docs/AMDGPUUsage.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 22c1d1f186ea5..7da5d8e41f6f8 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -16024,7 +16024,7 @@ Additional Documentation
 .. [CLANG-ATTR] `Attributes in Clang <https://clang.llvm.org/docs/AttributeReference.html>`__
 .. [DWARF] `DWARF Debugging Information Format <http://dwarfstd.org/>`__
 .. [ELF] `Executable and Linkable Format (ELF) <http://www.sco.com/developers/gabi/>`__
-.. [HRF] `Heterogeneous-race-free Memory Models <http://benedictgaster.org/wp-content/uploads/2014/01/asplos269-FINAL.pdf>`__
+.. [HRF] `Heterogeneous-race-free Memory Models <https://research.cs.wisc.edu/multifacet/papers/asplos14_hrf.pdf>`__
 .. [HSA] `Heterogeneous System Architecture (HSA) Foundation <http://www.hsafoundation.com/>`__
 .. [MsgPack] `Message Pack <http://www.msgpack.org/>`__
 .. [OpenCL] `The OpenCL Specification Version 2.0 <http://www.khronos.org/registry/cl/specs/opencl-2.0.pdf>`__

From 971ec1f0eea324d4a1eec6709e2c97e1798a6002 Mon Sep 17 00:00:00 2001
From: DianQK <dianqk@dianqk.net>
Date: Wed, 17 Apr 2024 20:54:17 +0800
Subject: [PATCH 738/886] [Inline] Regenerate inline-switch-default-2.ll (NFC)

---
 .../Inline/inline-switch-default-2.ll         | 176 ------------------
 1 file changed, 176 deletions(-)

diff --git a/llvm/test/Transforms/Inline/inline-switch-default-2.ll b/llvm/test/Transforms/Inline/inline-switch-default-2.ll
index 8d3e24c798df8..82dae1c27648f 100644
--- a/llvm/test/Transforms/Inline/inline-switch-default-2.ll
+++ b/llvm/test/Transforms/Inline/inline-switch-default-2.ll
@@ -4,50 +4,6 @@
 ; Check for scenarios without TTI.
 
 define i64 @foo1(i64 %a) {
-; LOOKUPTABLE-LABEL: define i64 @foo1(
-; LOOKUPTABLE-SAME: i64 [[TMP0:%.*]]) {
-; LOOKUPTABLE-NEXT:    switch i64 [[TMP0]], label [[DEFAULT_BRANCH_I:%.*]] [
-; LOOKUPTABLE-NEXT:      i64 0, label [[BRANCH_0_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 2, label [[BRANCH_2_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 4, label [[BRANCH_4_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 6, label [[BRANCH_6_I:%.*]]
-; LOOKUPTABLE-NEXT:    ]
-; LOOKUPTABLE:       branch_0.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR1_EXIT:%.*]]
-; LOOKUPTABLE:       branch_2.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR1_EXIT]]
-; LOOKUPTABLE:       branch_4.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR1_EXIT]]
-; LOOKUPTABLE:       branch_6.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR1_EXIT]]
-; LOOKUPTABLE:       default_branch.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR1_EXIT]]
-; LOOKUPTABLE:       bar1.exit:
-; LOOKUPTABLE-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0_I]] ], [ 9, [[BRANCH_2_I]] ], [ 2, [[BRANCH_4_I]] ], [ 7, [[BRANCH_6_I]] ], [ 3, [[DEFAULT_BRANCH_I]] ]
-; LOOKUPTABLE-NEXT:    ret i64 [[TMP2]]
-;
-; SWITCH-LABEL: define i64 @foo1(
-; SWITCH-SAME: i64 [[TMP0:%.*]]) {
-; SWITCH-NEXT:    switch i64 [[TMP0]], label [[DEFAULT_BRANCH_I:%.*]] [
-; SWITCH-NEXT:      i64 0, label [[BRANCH_0_I:%.*]]
-; SWITCH-NEXT:      i64 2, label [[BRANCH_2_I:%.*]]
-; SWITCH-NEXT:      i64 4, label [[BRANCH_4_I:%.*]]
-; SWITCH-NEXT:      i64 6, label [[BRANCH_6_I:%.*]]
-; SWITCH-NEXT:    ]
-; SWITCH:       branch_0.i:
-; SWITCH-NEXT:    br label [[BAR1_EXIT:%.*]]
-; SWITCH:       branch_2.i:
-; SWITCH-NEXT:    br label [[BAR1_EXIT]]
-; SWITCH:       branch_4.i:
-; SWITCH-NEXT:    br label [[BAR1_EXIT]]
-; SWITCH:       branch_6.i:
-; SWITCH-NEXT:    br label [[BAR1_EXIT]]
-; SWITCH:       default_branch.i:
-; SWITCH-NEXT:    br label [[BAR1_EXIT]]
-; SWITCH:       bar1.exit:
-; SWITCH-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0_I]] ], [ 9, [[BRANCH_2_I]] ], [ 2, [[BRANCH_4_I]] ], [ 7, [[BRANCH_6_I]] ], [ 3, [[DEFAULT_BRANCH_I]] ]
-; SWITCH-NEXT:    ret i64 [[TMP2]]
-;
 ; CHECK-LABEL: define i64 @foo1(
 ; CHECK-SAME: i64 [[A:%.*]]) {
 ; CHECK-NEXT:    [[B:%.*]] = call i64 @bar1(i64 [[A]])
@@ -58,50 +14,6 @@ define i64 @foo1(i64 %a) {
 }
 
 define i64 @foo2(i64 %a) {
-; LOOKUPTABLE-LABEL: define i64 @foo2(
-; LOOKUPTABLE-SAME: i64 [[TMP0:%.*]]) {
-; LOOKUPTABLE-NEXT:    switch i64 [[TMP0]], label [[UNREACHABLEDEFAULT_I:%.*]] [
-; LOOKUPTABLE-NEXT:      i64 0, label [[BRANCH_0_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 2, label [[BRANCH_2_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 4, label [[BRANCH_4_I:%.*]]
-; LOOKUPTABLE-NEXT:      i64 6, label [[BRANCH_6_I:%.*]]
-; LOOKUPTABLE-NEXT:    ]
-; LOOKUPTABLE:       branch_0.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR2_EXIT:%.*]]
-; LOOKUPTABLE:       branch_2.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR2_EXIT]]
-; LOOKUPTABLE:       branch_4.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR2_EXIT]]
-; LOOKUPTABLE:       branch_6.i:
-; LOOKUPTABLE-NEXT:    br label [[BAR2_EXIT]]
-; LOOKUPTABLE:       unreachabledefault.i:
-; LOOKUPTABLE-NEXT:    unreachable
-; LOOKUPTABLE:       bar2.exit:
-; LOOKUPTABLE-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0_I]] ], [ 9, [[BRANCH_2_I]] ], [ 2, [[BRANCH_4_I]] ], [ 7, [[BRANCH_6_I]] ]
-; LOOKUPTABLE-NEXT:    ret i64 [[TMP2]]
-;
-; SWITCH-LABEL: define i64 @foo2(
-; SWITCH-SAME: i64 [[TMP0:%.*]]) {
-; SWITCH-NEXT:    switch i64 [[TMP0]], label [[UNREACHABLEDEFAULT_I:%.*]] [
-; SWITCH-NEXT:      i64 0, label [[BRANCH_0_I:%.*]]
-; SWITCH-NEXT:      i64 2, label [[BRANCH_2_I:%.*]]
-; SWITCH-NEXT:      i64 4, label [[BRANCH_4_I:%.*]]
-; SWITCH-NEXT:      i64 6, label [[BRANCH_6_I:%.*]]
-; SWITCH-NEXT:    ]
-; SWITCH:       branch_0.i:
-; SWITCH-NEXT:    br label [[BAR2_EXIT:%.*]]
-; SWITCH:       branch_2.i:
-; SWITCH-NEXT:    br label [[BAR2_EXIT]]
-; SWITCH:       branch_4.i:
-; SWITCH-NEXT:    br label [[BAR2_EXIT]]
-; SWITCH:       branch_6.i:
-; SWITCH-NEXT:    br label [[BAR2_EXIT]]
-; SWITCH:       unreachabledefault.i:
-; SWITCH-NEXT:    unreachable
-; SWITCH:       bar2.exit:
-; SWITCH-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0_I]] ], [ 9, [[BRANCH_2_I]] ], [ 2, [[BRANCH_4_I]] ], [ 7, [[BRANCH_6_I]] ]
-; SWITCH-NEXT:    ret i64 [[TMP2]]
-;
 ; CHECK-LABEL: define i64 @foo2(
 ; CHECK-SAME: i64 [[A:%.*]]) {
 ; CHECK-NEXT:    switch i64 [[A]], label [[UNREACHABLEDEFAULT_I:%.*]] [
@@ -129,50 +41,6 @@ define i64 @foo2(i64 %a) {
 }
 
 define i64 @bar1(i64 %a) {
-; LOOKUPTABLE-LABEL: define i64 @bar1(
-; LOOKUPTABLE-SAME: i64 [[TMP0:%.*]]) {
-; LOOKUPTABLE-NEXT:    switch i64 [[TMP0]], label [[DEFAULT_BRANCH:%.*]] [
-; LOOKUPTABLE-NEXT:      i64 0, label [[BRANCH_0:%.*]]
-; LOOKUPTABLE-NEXT:      i64 2, label [[BRANCH_2:%.*]]
-; LOOKUPTABLE-NEXT:      i64 4, label [[BRANCH_4:%.*]]
-; LOOKUPTABLE-NEXT:      i64 6, label [[BRANCH_6:%.*]]
-; LOOKUPTABLE-NEXT:    ]
-; LOOKUPTABLE:       branch_0:
-; LOOKUPTABLE-NEXT:    br label [[EXIT:%.*]]
-; LOOKUPTABLE:       branch_2:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       branch_4:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       branch_6:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       default_branch:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       exit:
-; LOOKUPTABLE-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0]] ], [ 9, [[BRANCH_2]] ], [ 2, [[BRANCH_4]] ], [ 7, [[BRANCH_6]] ], [ 3, [[DEFAULT_BRANCH]] ]
-; LOOKUPTABLE-NEXT:    ret i64 [[TMP2]]
-;
-; SWITCH-LABEL: define i64 @bar1(
-; SWITCH-SAME: i64 [[TMP0:%.*]]) {
-; SWITCH-NEXT:    switch i64 [[TMP0]], label [[DEFAULT_BRANCH:%.*]] [
-; SWITCH-NEXT:      i64 0, label [[BRANCH_0:%.*]]
-; SWITCH-NEXT:      i64 2, label [[BRANCH_2:%.*]]
-; SWITCH-NEXT:      i64 4, label [[BRANCH_4:%.*]]
-; SWITCH-NEXT:      i64 6, label [[BRANCH_6:%.*]]
-; SWITCH-NEXT:    ]
-; SWITCH:       branch_0:
-; SWITCH-NEXT:    br label [[EXIT:%.*]]
-; SWITCH:       branch_2:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       branch_4:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       branch_6:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       default_branch:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       exit:
-; SWITCH-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0]] ], [ 9, [[BRANCH_2]] ], [ 2, [[BRANCH_4]] ], [ 7, [[BRANCH_6]] ], [ 3, [[DEFAULT_BRANCH]] ]
-; SWITCH-NEXT:    ret i64 [[TMP2]]
-;
 ; CHECK-LABEL: define i64 @bar1(
 ; CHECK-SAME: i64 [[A:%.*]]) {
 ; CHECK-NEXT:    switch i64 [[A]], label [[DEFAULT_BRANCH:%.*]] [
@@ -223,50 +91,6 @@ exit:
 }
 
 define i64 @bar2(i64 %a) {
-; LOOKUPTABLE-LABEL: define i64 @bar2(
-; LOOKUPTABLE-SAME: i64 [[TMP0:%.*]]) {
-; LOOKUPTABLE-NEXT:    switch i64 [[TMP0]], label [[UNREACHABLEDEFAULT:%.*]] [
-; LOOKUPTABLE-NEXT:      i64 0, label [[BRANCH_0:%.*]]
-; LOOKUPTABLE-NEXT:      i64 2, label [[BRANCH_2:%.*]]
-; LOOKUPTABLE-NEXT:      i64 4, label [[BRANCH_4:%.*]]
-; LOOKUPTABLE-NEXT:      i64 6, label [[BRANCH_6:%.*]]
-; LOOKUPTABLE-NEXT:    ]
-; LOOKUPTABLE:       branch_0:
-; LOOKUPTABLE-NEXT:    br label [[EXIT:%.*]]
-; LOOKUPTABLE:       branch_2:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       branch_4:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       branch_6:
-; LOOKUPTABLE-NEXT:    br label [[EXIT]]
-; LOOKUPTABLE:       unreachabledefault:
-; LOOKUPTABLE-NEXT:    unreachable
-; LOOKUPTABLE:       exit:
-; LOOKUPTABLE-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0]] ], [ 9, [[BRANCH_2]] ], [ 2, [[BRANCH_4]] ], [ 7, [[BRANCH_6]] ]
-; LOOKUPTABLE-NEXT:    ret i64 [[TMP2]]
-;
-; SWITCH-LABEL: define i64 @bar2(
-; SWITCH-SAME: i64 [[TMP0:%.*]]) {
-; SWITCH-NEXT:    switch i64 [[TMP0]], label [[UNREACHABLEDEFAULT:%.*]] [
-; SWITCH-NEXT:      i64 0, label [[BRANCH_0:%.*]]
-; SWITCH-NEXT:      i64 2, label [[BRANCH_2:%.*]]
-; SWITCH-NEXT:      i64 4, label [[BRANCH_4:%.*]]
-; SWITCH-NEXT:      i64 6, label [[BRANCH_6:%.*]]
-; SWITCH-NEXT:    ]
-; SWITCH:       branch_0:
-; SWITCH-NEXT:    br label [[EXIT:%.*]]
-; SWITCH:       branch_2:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       branch_4:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       branch_6:
-; SWITCH-NEXT:    br label [[EXIT]]
-; SWITCH:       unreachabledefault:
-; SWITCH-NEXT:    unreachable
-; SWITCH:       exit:
-; SWITCH-NEXT:    [[TMP2:%.*]] = phi i64 [ 5, [[BRANCH_0]] ], [ 9, [[BRANCH_2]] ], [ 2, [[BRANCH_4]] ], [ 7, [[BRANCH_6]] ]
-; SWITCH-NEXT:    ret i64 [[TMP2]]
-;
 ; CHECK-LABEL: define i64 @bar2(
 ; CHECK-SAME: i64 [[A:%.*]]) {
 ; CHECK-NEXT:    switch i64 [[A]], label [[UNREACHABLEDEFAULT:%.*]] [

From 73140daebbf522dbb14dc4b2f3c67dc0aa1a62dd Mon Sep 17 00:00:00 2001
From: "Oleksandr \"Alex\" Zinenko" <zinenko@google.com>
Date: Wed, 17 Apr 2024 15:01:59 +0200
Subject: [PATCH 739/886] [mlir] expose transform dialect symbol merge to
 python (#87690)

This functionality is available in C++, make it available in Python
directly to operate on transform modules.
---
 .../mlir-c/Dialect/Transform/Interpreter.h    | 12 ++-
 .../Bindings/Python/TransformInterpreter.cpp  | 15 ++++
 .../lib/CAPI/Dialect/TransformInterpreter.cpp |  9 +++
 .../transform/interpreter/__init__.py         | 10 ++-
 .../python/dialects/transform_interpreter.py  | 76 +++++++++++++++++++
 5 files changed, 120 insertions(+), 2 deletions(-)

diff --git a/mlir/include/mlir-c/Dialect/Transform/Interpreter.h b/mlir/include/mlir-c/Dialect/Transform/Interpreter.h
index 00095d5040a0e..fa320324234e8 100644
--- a/mlir/include/mlir-c/Dialect/Transform/Interpreter.h
+++ b/mlir/include/mlir-c/Dialect/Transform/Interpreter.h
@@ -60,7 +60,7 @@ MLIR_CAPI_EXPORTED void
 mlirTransformOptionsDestroy(MlirTransformOptions transformOptions);
 
 //----------------------------------------------------------------------------//
-// Transform interpreter.
+// Transform interpreter and utilities.
 //----------------------------------------------------------------------------//
 
 /// Applies the transformation script starting at the given transform root
@@ -72,6 +72,16 @@ MLIR_CAPI_EXPORTED MlirLogicalResult mlirTransformApplyNamedSequence(
     MlirOperation payload, MlirOperation transformRoot,
     MlirOperation transformModule, MlirTransformOptions transformOptions);
 
+/// Merge the symbols from `other` into `target`, potentially renaming them to
+/// avoid conflicts. Private symbols may be renamed during the merge, public
+/// symbols must have at most one declaration. A name conflict in public symbols
+/// is reported as an error before returning a failure.
+///
+/// Note that this clones the `other` operation unlike the C++ counterpart that
+/// takes ownership.
+MLIR_CAPI_EXPORTED MlirLogicalResult
+mlirMergeSymbolsIntoFromClone(MlirOperation target, MlirOperation other);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/mlir/lib/Bindings/Python/TransformInterpreter.cpp b/mlir/lib/Bindings/Python/TransformInterpreter.cpp
index 6517f8c39dfad..f6b4532b1b6be 100644
--- a/mlir/lib/Bindings/Python/TransformInterpreter.cpp
+++ b/mlir/lib/Bindings/Python/TransformInterpreter.cpp
@@ -82,6 +82,21 @@ static void populateTransformInterpreterSubmodule(py::module &m) {
       py::arg("payload_root"), py::arg("transform_root"),
       py::arg("transform_module"),
       py::arg("transform_options") = PyMlirTransformOptions());
+
+  m.def(
+      "copy_symbols_and_merge_into",
+      [](MlirOperation target, MlirOperation other) {
+        mlir::python::CollectDiagnosticsToStringScope scope(
+            mlirOperationGetContext(target));
+
+        MlirLogicalResult result = mlirMergeSymbolsIntoFromClone(target, other);
+        if (mlirLogicalResultIsFailure(result)) {
+          throw py::value_error(
+              "Failed to merge symbols.\nDiagnostic message " +
+              scope.takeMessage());
+        }
+      },
+      py::arg("target"), py::arg("other"));
 }
 
 PYBIND11_MODULE(_mlirTransformInterpreter, m) {
diff --git a/mlir/lib/CAPI/Dialect/TransformInterpreter.cpp b/mlir/lib/CAPI/Dialect/TransformInterpreter.cpp
index eb6951dc5584d..145455e1c1b3d 100644
--- a/mlir/lib/CAPI/Dialect/TransformInterpreter.cpp
+++ b/mlir/lib/CAPI/Dialect/TransformInterpreter.cpp
@@ -15,6 +15,7 @@
 #include "mlir/CAPI/IR.h"
 #include "mlir/CAPI/Support.h"
 #include "mlir/CAPI/Wrap.h"
+#include "mlir/Dialect/Transform/IR/Utils.h"
 #include "mlir/Dialect/Transform/Interfaces/TransformInterfaces.h"
 #include "mlir/Dialect/Transform/Transforms/TransformInterpreterUtils.h"
 
@@ -71,4 +72,12 @@ MlirLogicalResult mlirTransformApplyNamedSequence(
       unwrap(payload), unwrap(transformRoot),
       cast<ModuleOp>(unwrap(transformModule)), *unwrap(transformOptions)));
 }
+
+MlirLogicalResult mlirMergeSymbolsIntoFromClone(MlirOperation target,
+                                                MlirOperation other) {
+  OwningOpRef<Operation *> otherOwning(unwrap(other)->clone());
+  LogicalResult result = transform::detail::mergeSymbolsInto(
+      unwrap(target), std::move(otherOwning));
+  return wrap(result);
+}
 }
diff --git a/mlir/python/mlir/dialects/transform/interpreter/__init__.py b/mlir/python/mlir/dialects/transform/interpreter/__init__.py
index 6145b99224eb5..34cdc43cb617f 100644
--- a/mlir/python/mlir/dialects/transform/interpreter/__init__.py
+++ b/mlir/python/mlir/dialects/transform/interpreter/__init__.py
@@ -5,7 +5,6 @@
 from ....ir import Operation
 from ...._mlir_libs import _mlirTransformInterpreter as _cextTransformInterpreter
 
-
 TransformOptions = _cextTransformInterpreter.TransformOptions
 
 
@@ -31,3 +30,12 @@ def apply_named_sequence(
         _cextTransformInterpreter.apply_named_sequence(*args)
     else:
         _cextTransformInterpreter(*args, transform_options)
+
+
+def copy_symbols_and_merge_into(target, other):
+    """Copies symbols from other into target, renaming private symbols to avoid
+    duplicates. Raises an error if copying would lead to duplicate public
+    symbols."""
+    _cextTransformInterpreter.copy_symbols_and_merge_into(
+        _unpack_operation(target), _unpack_operation(other)
+    )
diff --git a/mlir/test/python/dialects/transform_interpreter.py b/mlir/test/python/dialects/transform_interpreter.py
index 740c49f76a26c..807a98c493279 100644
--- a/mlir/test/python/dialects/transform_interpreter.py
+++ b/mlir/test/python/dialects/transform_interpreter.py
@@ -54,3 +54,79 @@ def failed():
         assert (
             "must implement TransformOpInterface to be used as transform root" in str(e)
         )
+
+
+print_root_via_include_module = """
+module @print_root_via_include_module attributes {transform.with_named_sequence} {
+  transform.named_sequence private @callee1(%root: !transform.any_op {transform.readonly})
+  transform.named_sequence private @callee2(%root: !transform.any_op {transform.readonly})
+  transform.named_sequence @__transform_main(%root: !transform.any_op) {
+    transform.include @callee2 failures(propagate)
+        (%root) : (!transform.any_op) -> ()
+    transform.yield
+  }
+}"""
+
+callee2_definition = """
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence private @callee1(%root: !transform.any_op {transform.readonly})
+  transform.named_sequence @callee2(%root: !transform.any_op {transform.readonly}) {
+    transform.include @callee1 failures(propagate)
+        (%root) : (!transform.any_op) -> ()
+    transform.yield
+  }
+}
+"""
+
+callee1_definition = """
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @callee1(%root: !transform.any_op {transform.readonly}) {
+    transform.print %root { name = \"from interpreter\" }: !transform.any_op
+    transform.yield
+  }
+}
+"""
+
+
+@test_in_context
+def include():
+    main = ir.Module.parse(print_root_via_include_module)
+    callee1 = ir.Module.parse(callee1_definition)
+    callee2 = ir.Module.parse(callee2_definition)
+    interp.copy_symbols_and_merge_into(main, callee1)
+    interp.copy_symbols_and_merge_into(main, callee2)
+
+    # CHECK: @print_root_via_include_module
+    # CHECK: transform.named_sequence @__transform_main
+    # CHECK: transform.include @callee2
+    #
+    # CHECK: transform.named_sequence @callee1
+    # CHECK: transform.print
+    #
+    # CHECK: transform.named_sequence @callee2
+    # CHECK: transform.include @callee1
+    interp.apply_named_sequence(main, main.body.operations[0], main)
+
+
+@test_in_context
+def partial_include():
+    main = ir.Module.parse(print_root_via_include_module)
+    callee2 = ir.Module.parse(callee2_definition)
+    interp.copy_symbols_and_merge_into(main, callee2)
+
+    try:
+        interp.apply_named_sequence(main, main.body.operations[0], main)
+    except ValueError as e:
+        assert "Failed to apply" in str(e)
+
+
+@test_in_context
+def repeated_include():
+    main = ir.Module.parse(print_root_via_include_module)
+    callee2 = ir.Module.parse(callee2_definition)
+    interp.copy_symbols_and_merge_into(main, callee2)
+
+    try:
+        interp.copy_symbols_and_merge_into(main, callee2)
+    except ValueError as e:
+        assert "doubly defined symbol @callee2" in str(e)

From 20d653fdb2d4d6eafa4575cd954beaf7ecad113a Mon Sep 17 00:00:00 2001
From: Krzysztof Parzyszek <Krzysztof.Parzyszek@amd.com>
Date: Wed, 17 Apr 2024 08:04:19 -0500
Subject: [PATCH 740/886] [LLVM][CodeGen] Fix register lane liveness tracking
 in RegisterPressure (#88892)

Re-enable an old assertion in `decreaseSetPressure`.
---
 llvm/lib/CodeGen/RegisterPressure.cpp | 39 ++++++++++++++-------------
 1 file changed, 21 insertions(+), 18 deletions(-)

diff --git a/llvm/lib/CodeGen/RegisterPressure.cpp b/llvm/lib/CodeGen/RegisterPressure.cpp
index f86aa3a167202..3fa22447f416b 100644
--- a/llvm/lib/CodeGen/RegisterPressure.cpp
+++ b/llvm/lib/CodeGen/RegisterPressure.cpp
@@ -64,7 +64,7 @@ static void increaseSetPressure(std::vector<unsigned> &CurrSetPressure,
 static void decreaseSetPressure(std::vector<unsigned> &CurrSetPressure,
                                 const MachineRegisterInfo &MRI, Register Reg,
                                 LaneBitmask PrevMask, LaneBitmask NewMask) {
-  //assert((NewMask & !PrevMask) == 0 && "Must not add bits");
+  assert((NewMask & ~PrevMask).none() && "Must not add bits");
   if (NewMask.any() || PrevMask.none())
     return;
 
@@ -617,17 +617,11 @@ void RegisterOperands::adjustLaneLiveness(const LiveIntervals &LIS,
       ++I;
     }
   }
-  for (auto *I = Uses.begin(); I != Uses.end();) {
-    LaneBitmask LiveBefore = getLiveLanesAt(LIS, MRI, true, I->RegUnit,
-                                            Pos.getBaseIndex());
-    LaneBitmask LaneMask = I->LaneMask & LiveBefore;
-    if (LaneMask.none()) {
-      I = Uses.erase(I);
-    } else {
-      I->LaneMask = LaneMask;
-      ++I;
-    }
-  }
+
+  // For uses just copy the information from LIS.
+  for (auto &[RegUnit, LaneMask] : Uses)
+    LaneMask = getLiveLanesAt(LIS, MRI, true, RegUnit, Pos.getBaseIndex());
+
   if (AddFlagsMI != nullptr) {
     for (const RegisterMaskPair &P : DeadDefs) {
       Register RegUnit = P.RegUnit;
@@ -1060,18 +1054,27 @@ void RegPressureTracker::bumpUpwardPressure(const MachineInstr *MI) {
   // Kill liveness at live defs.
   for (const RegisterMaskPair &P : RegOpers.Defs) {
     Register Reg = P.RegUnit;
-    LaneBitmask LiveLanes = LiveRegs.contains(Reg);
+    LaneBitmask LiveAfter = LiveRegs.contains(Reg);
     LaneBitmask UseLanes = getRegLanes(RegOpers.Uses, Reg);
     LaneBitmask DefLanes = P.LaneMask;
-    LaneBitmask LiveAfter = (LiveLanes & ~DefLanes) | UseLanes;
-    decreaseRegPressure(Reg, LiveLanes, LiveAfter);
+    LaneBitmask LiveBefore = (LiveAfter & ~DefLanes) | UseLanes;
+
+    // There may be parts of the register that were dead before the
+    // instruction, but became live afterwards. Similarly, some parts
+    // may have been killed in this instruction.
+    decreaseRegPressure(Reg, LiveAfter, LiveAfter & LiveBefore);
+    increaseRegPressure(Reg, LiveAfter, ~LiveAfter & LiveBefore);
   }
   // Generate liveness for uses.
   for (const RegisterMaskPair &P : RegOpers.Uses) {
     Register Reg = P.RegUnit;
-    LaneBitmask LiveLanes = LiveRegs.contains(Reg);
-    LaneBitmask LiveAfter = LiveLanes | P.LaneMask;
-    increaseRegPressure(Reg, LiveLanes, LiveAfter);
+    // If this register was also in a def operand, we've handled it
+    // with defs.
+    if (getRegLanes(RegOpers.Defs, Reg).any())
+      continue;
+    LaneBitmask LiveAfter = LiveRegs.contains(Reg);
+    LaneBitmask LiveBefore = LiveAfter | P.LaneMask;
+    increaseRegPressure(Reg, LiveAfter, LiveBefore);
   }
 }
 

From 1fc72dbc807fb138cafd05501e2e31beaa574693 Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Wed, 17 Apr 2024 21:10:56 +0800
Subject: [PATCH 741/886] [RISCV] Add test for doLocalPostpass issue not
 checking if VL was modified. NFC

---
 .../test/CodeGen/RISCV/rvv/vsetvli-insert.mir | 25 +++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir b/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
index 1850abe6363bc..d9a87c2cb12a8 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
+++ b/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
@@ -76,6 +76,10 @@
     ret void
   }
 
+  define void @postpass_modify_vl() {
+    ret void
+  }
+
   declare <vscale x 1 x i64> @llvm.riscv.vadd.nxv1i64.nxv1i64.i64(<vscale x 1 x i64>, <vscale x 1 x i64>, <vscale x 1 x i64>, i64) #1
 
   declare <vscale x 1 x i64> @llvm.riscv.vle.nxv1i64.i64(<vscale x 1 x i64>, ptr nocapture, i64) #4
@@ -503,3 +507,24 @@ body:             |
     %5:vr = PseudoVMV_V_I_MF2 $noreg, 1, 2, 5, 0
     PseudoRET
 ...
+---
+name:            postpass_modify_vl
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $x1
+    ; CHECK-LABEL: name: postpass_modify_vl
+    ; CHECK: liveins: $x1
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: $x0 = PseudoVSETIVLI 3, 216 /* e64, m1, ta, ma */, implicit-def $vl, implicit-def $vtype
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $vtype
+    ; CHECK-NEXT: $vl = COPY $x1
+    ; CHECK-NEXT: [[PseudoVADD_VV_M1_:%[0-9]+]]:vr = PseudoVADD_VV_M1 $noreg, $noreg, $noreg, 3, 6 /* e64 */, 0 /* tu, mu */, implicit $vl, implicit $vtype
+    ; CHECK-NEXT: PseudoRET
+    dead $x0 = PseudoVSETIVLI 3, 216, implicit-def $vl, implicit-def $vtype
+    %1:gpr = COPY $vtype
+    $vl = COPY $x1
+    dead $x0 = PseudoVSETIVLI 3, 216, implicit-def $vl, implicit-def $vtype
+    %4:vr = PseudoVADD_VV_M1 $noreg, $noreg, $noreg, 3, 6, 0
+    PseudoRET
+...

From 76ad2897480a85532eee93daf041246881772693 Mon Sep 17 00:00:00 2001
From: Zaara Syeda <syzaara@ca.ibm.com>
Date: Wed, 17 Apr 2024 09:24:53 -0400
Subject: [PATCH 742/886] [PowerPC] 32-bit large code-model support for
 toc-data (#85129)

This patch adds the pseudo op ADDItocL for 32-bit large code-model
support for toc-data.
---
 llvm/lib/Target/PowerPC/P10InstrResources.td |  2 +-
 llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp    | 61 ++++++++++++------
 llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp  | 35 +++++++---
 llvm/lib/Target/PowerPC/PPCInstrInfo.cpp     |  1 +
 llvm/lib/Target/PowerPC/PPCInstrInfo.td      |  4 +-
 llvm/lib/Target/PowerPC/PPCMacroFusion.def   | 10 +--
 llvm/test/CodeGen/PowerPC/toc-data.ll        | 67 ++++++++++++++++++++
 7 files changed, 144 insertions(+), 36 deletions(-)

diff --git a/llvm/lib/Target/PowerPC/P10InstrResources.td b/llvm/lib/Target/PowerPC/P10InstrResources.td
index 5015ba887d0b6..32cebb65cb569 100644
--- a/llvm/lib/Target/PowerPC/P10InstrResources.td
+++ b/llvm/lib/Target/PowerPC/P10InstrResources.td
@@ -881,7 +881,7 @@ def : InstRW<[P10W_FX_3C, P10W_DISP_ANY],
 // 3 Cycles ALU operations, 1 input operands
 def : InstRW<[P10W_FX_3C, P10W_DISP_ANY, P10FX_Read],
       (instrs
-    ADDI, ADDI8, ADDIdtprelL32, ADDItlsldLADDR32, ADDItocL8, LI, LI8,
+    ADDI, ADDI8, ADDIdtprelL32, ADDItlsldLADDR32, ADDItocL, ADDItocL8, LI, LI8,
     ADDIC, ADDIC8,
     ADDIS, ADDIS8, ADDISdtprelHA32, ADDIStocHA, ADDIStocHA8, LIS, LIS8,
     ADDME, ADDME8,
diff --git a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
index 1c57b92057fff..6e1002c45d81c 100644
--- a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
+++ b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
@@ -1148,15 +1148,27 @@ void PPCAsmPrinter::emitInstruction(const MachineInstr *MI) {
 
     MCSymbolRefExpr::VariantKind VK = GetVKForMO(MO);
 
-    // Always use TOC on AIX. Map the global address operand to be a reference
-    // to the TOC entry we will synthesize later. 'TOCEntry' is a label used to
-    // reference the storage allocated in the TOC which contains the address of
-    // 'MOSymbol'.
-    MCSymbol *TOCEntry =
-        lookUpOrCreateTOCEntry(MOSymbol, getTOCEntryTypeForMO(MO), VK);
-    const MCExpr *Exp = MCSymbolRefExpr::create(TOCEntry,
-                                                MCSymbolRefExpr::VK_PPC_U,
-                                                OutContext);
+    // If the symbol isn't toc-data then use the TOC on AIX.
+    // Map the global address operand to be a reference to the TOC entry we
+    // will synthesize later. 'TOCEntry' is a label used to reference the
+    // storage allocated in the TOC which contains the address of 'MOSymbol'.
+    // If the toc-data attribute is used, the TOC entry contains the data
+    // rather than the address of the MOSymbol.
+    if (![](const MachineOperand &MO) {
+          if (!MO.isGlobal())
+            return false;
+
+          const GlobalVariable *GV = dyn_cast<GlobalVariable>(MO.getGlobal());
+          if (!GV)
+            return false;
+
+          return GV->hasAttribute("toc-data");
+        }(MO)) {
+      MOSymbol = lookUpOrCreateTOCEntry(MOSymbol, getTOCEntryTypeForMO(MO), VK);
+    }
+
+    const MCExpr *Exp = MCSymbolRefExpr::create(
+        MOSymbol, MCSymbolRefExpr::VK_PPC_U, OutContext);
     TmpInst.getOperand(2) = MCOperand::createExpr(Exp);
     EmitToStreamer(*OutStreamer, TmpInst);
     return;
@@ -1273,25 +1285,32 @@ void PPCAsmPrinter::emitInstruction(const MachineInstr *MI) {
     EmitToStreamer(*OutStreamer, TmpInst);
     return;
   }
+  case PPC::ADDItocL:
   case PPC::ADDItocL8: {
-    // Transform %xd = ADDItocL8 %xs, @sym
+    // Transform %xd = ADDItocL %xs, @sym
     LowerPPCMachineInstrToMCInst(MI, TmpInst, *this);
 
-    // Change the opcode to ADDI8. If the global address is external, then
-    // generate a TOC entry and reference that. Otherwise, reference the
-    // symbol directly.
-    TmpInst.setOpcode(PPC::ADDI8);
+    unsigned Op = MI->getOpcode();
+
+    // Change the opcode to load address for tocdata
+    TmpInst.setOpcode(Op == PPC::ADDItocL8 ? PPC::ADDI8 : PPC::LA);
 
     const MachineOperand &MO = MI->getOperand(2);
-    assert((MO.isGlobal() || MO.isCPI()) && "Invalid operand for ADDItocL8.");
+    assert((Op == PPC::ADDItocL8)
+               ? (MO.isGlobal() || MO.isCPI())
+               : MO.isGlobal() && "Invalid operand for ADDItocL8.");
+    assert(!(MO.isGlobal() && Subtarget->isGVIndirectSymbol(MO.getGlobal())) &&
+           "Interposable definitions must use indirect accesses.");
 
-    LLVM_DEBUG(assert(
-        !(MO.isGlobal() && Subtarget->isGVIndirectSymbol(MO.getGlobal())) &&
-        "Interposable definitions must use indirect access."));
+    // Map the operand to its corresponding MCSymbol.
+    const MCSymbol *const MOSymbol = getMCSymbolForTOCPseudoMO(MO, *this);
+
+    const MCExpr *Exp = MCSymbolRefExpr::create(
+        MOSymbol,
+        Op == PPC::ADDItocL8 ? MCSymbolRefExpr::VK_PPC_TOC_LO
+                             : MCSymbolRefExpr::VK_PPC_L,
+        OutContext);
 
-    const MCExpr *Exp =
-        MCSymbolRefExpr::create(getMCSymbolForTOCPseudoMO(MO, *this),
-                                MCSymbolRefExpr::VK_PPC_TOC_LO, OutContext);
     TmpInst.getOperand(2) = MCOperand::createExpr(Exp);
     EmitToStreamer(*OutStreamer, TmpInst);
     return;
diff --git a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index af82b6cdb1809..2f647daa4bcb5 100644
--- a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -510,7 +510,7 @@ SDNode *PPCDAGToDAGISel::getGlobalBaseReg() {
 }
 
 // Check if a SDValue has the toc-data attribute.
-static bool hasTocDataAttr(SDValue Val, unsigned PointerSize) {
+static bool hasTocDataAttr(SDValue Val) {
   GlobalAddressSDNode *GA = dyn_cast<GlobalAddressSDNode>(Val);
   if (!GA)
     return false;
@@ -6115,8 +6115,7 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
 
       assert(isAIXABI && "ELF ABI already handled");
 
-      if (hasTocDataAttr(N->getOperand(0),
-                         CurDAG->getDataLayout().getPointerSize())) {
+      if (hasTocDataAttr(N->getOperand(0))) {
         replaceWith(PPC::ADDItoc, N, MVT::i32);
         return;
       }
@@ -6128,8 +6127,7 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
     if (isPPC64 && CModel == CodeModel::Small) {
       assert(isAIXABI && "ELF ABI handled in common SelectCode");
 
-      if (hasTocDataAttr(N->getOperand(0),
-                         CurDAG->getDataLayout().getPointerSize())) {
+      if (hasTocDataAttr(N->getOperand(0))) {
         replaceWith(PPC::ADDItoc8, N, MVT::i64);
         return;
       }
@@ -6144,9 +6142,10 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
            " ELF/AIX or 32-bit AIX in the following.");
 
     // Transforms the ISD::TOC_ENTRY node for 32-bit AIX large code model mode
-    // or 64-bit medium (ELF-only) or large (ELF and AIX) code model code. We
-    // generate two instructions as described below. The first source operand
-    // is a symbol reference. If it must be toc-referenced according to
+    // or 64-bit medium (ELF-only) or large (ELF and AIX) code model code non
+    // toc-data symbols.
+    // We generate two instructions as described below. The first source
+    // operand is a symbol reference. If it must be toc-referenced according to
     // Subtarget, we generate:
     // [32-bit AIX]
     //   LWZtocL(@sym, ADDIStocHA(%r2, @sym))
@@ -6154,6 +6153,13 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
     //   LDtocL(@sym, ADDIStocHA8(%x2, @sym))
     // Otherwise we generate:
     //   ADDItocL8(ADDIStocHA8(%x2, @sym), @sym)
+
+    // For large code model toc-data symbols we generate:
+    // [32-bit AIX]
+    //   ADDItocL(ADDIStocHA(%x2, @sym), @sym)
+    // [64-bit AIX]
+    //   Currently not supported.
+
     SDValue GA = N->getOperand(0);
     SDValue TOCbase = N->getOperand(1);
 
@@ -6161,6 +6167,19 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
     SDNode *Tmp = CurDAG->getMachineNode(
         isPPC64 ? PPC::ADDIStocHA8 : PPC::ADDIStocHA, dl, VT, TOCbase, GA);
 
+    // On AIX if the symbol has the toc-data attribute it will be defined
+    // in the TOC entry, so we use an ADDItocL similar to the medium code
+    // model ELF abi.
+    if (isAIXABI && hasTocDataAttr(GA)) {
+      if (isPPC64)
+        report_fatal_error(
+            "64-bit large code model toc-data not yet supported");
+
+      ReplaceNode(N, CurDAG->getMachineNode(PPC::ADDItocL, dl, VT,
+                                            SDValue(Tmp, 0), GA));
+      return;
+    }
+
     if (PPCLowering->isAccessedAsGotIndirect(GA)) {
       // If it is accessed as got-indirect, we need an extra LWZ/LD to load
       // the address.
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
index 93874d65531ae..b32f178ca38e6 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -1090,6 +1090,7 @@ bool PPCInstrInfo::isReallyTriviallyReMaterializable(
   case PPC::LIS8:
   case PPC::ADDIStocHA:
   case PPC::ADDIStocHA8:
+  case PPC::ADDItocL:
   case PPC::ADDItocL8:
   case PPC::LOAD_STACK_GUARD:
   case PPC::PPCLdFixedAddr:
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.td b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
index 6423e692d88c3..43e3902cf4074 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
@@ -3346,11 +3346,13 @@ def ADDIStocHA : PPCEmitTimePseudo<(outs gprc:$rD), (ins gprc_nor0:$reg, tocentr
                        "#ADDIStocHA",
                        [(set i32:$rD,
                          (PPCtoc_entry i32:$reg, tglobaladdr:$disp))]>;
-// Local Data Transform
+// TOC Data Transform AIX
 def ADDItoc : PPCEmitTimePseudo<(outs gprc:$rD), (ins tocentry32:$disp, gprc:$reg),
                    "#ADDItoc",
                    [(set i32:$rD,
                      (PPCtoc_entry tglobaladdr:$disp, i32:$reg))]>;
+def ADDItocL : PPCEmitTimePseudo<(outs gprc:$rD), (ins gprc_nor0:$reg, tocentry32:$disp),
+                   "#ADDItocL", []>;
 
 // Get Global (GOT) Base Register offset, from the word immediately preceding
 // the function label.
diff --git a/llvm/lib/Target/PowerPC/PPCMacroFusion.def b/llvm/lib/Target/PowerPC/PPCMacroFusion.def
index fb6e656edb8b9..1a61ae23e5d72 100644
--- a/llvm/lib/Target/PowerPC/PPCMacroFusion.def
+++ b/llvm/lib/Target/PowerPC/PPCMacroFusion.def
@@ -32,7 +32,7 @@
 // {addi} followed by one of these {lxvd2x, lxvw4x, lxvdsx, lvebx, lvehx,
 // lvewx, lvx, lxsdx}
 FUSION_FEATURE(AddiLoad, hasAddiLoadFusion, 2, \
-               FUSION_OP_SET(ADDI, ADDI8, ADDItocL8), \
+               FUSION_OP_SET(ADDI, ADDI8, ADDItocL, ADDItocL8), \
                FUSION_OP_SET(LXVD2X, LXVW4X, LXVDSX, LVEBX, LVEHX, LVEWX, \
                              LVX, LXSDX))
 
@@ -134,13 +134,13 @@ FUSION_FEATURE(XorisXori, hasWideImmFusion, 1, FUSION_OP_SET(XORIS, XORIS8),
 
 // addis rx,ra,si - addi rt,rx,SI, SI >= 0
 FUSION_FEATURE(AddisAddi, hasWideImmFusion, 1,
-               FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8),
-               FUSION_OP_SET(ADDI, ADDI8, ADDItocL8))
+               FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8, ADDIStocHA),
+               FUSION_OP_SET(ADDI, ADDI8, ADDItocL8, ADDItocL))
 
 // addi rx,ra,si - addis rt,rx,SI, ra > 0, SI >= 2
 FUSION_FEATURE(AddiAddis, hasWideImmFusion, 1,
-               FUSION_OP_SET(ADDI, ADDI8, ADDItocL8),
-               FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8))
+               FUSION_OP_SET(ADDI, ADDI8, ADDItocL8, ADDItocL),
+               FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8, ADDIStocHA))
 
 // mtctr - { bcctr,bcctrl }
 FUSION_FEATURE(ZeroMoveCTR, hasZeroMoveFusion, -1,
diff --git a/llvm/test/CodeGen/PowerPC/toc-data.ll b/llvm/test/CodeGen/PowerPC/toc-data.ll
index cbf3be9fcaad0..7f7afe76cfcde 100644
--- a/llvm/test/CodeGen/PowerPC/toc-data.ll
+++ b/llvm/test/CodeGen/PowerPC/toc-data.ll
@@ -12,6 +12,14 @@
 ; RUN: llc -mtriple powerpc-ibm-aix-xcoff -verify-machineinstrs -O0 < %s | FileCheck %s --check-prefix TEST32
 ; RUN: llc -mtriple powerpc64-ibm-aix-xcoff -verify-machineinstrs -O0 < %s | FileCheck %s --check-prefix TEST64
 
+; RUN: llc -mtriple powerpc-ibm-aix-xcoff -code-model=large -verify-machineinstrs < %s \
+; RUN:     -stop-before=ppc-vsx-copy | FileCheck %s --check-prefix CHECK32LARGE
+; RUN: llc -mtriple powerpc-ibm-aix-xcoff -code-model=large -verify-machineinstrs < %s | FileCheck %s --check-prefix TEST32LARGE
+
+; Global variables i and f have the toc-data attribute.
+; In the following functions, those writing to or reading from
+; variables i and f should use the toc-data access pattern.
+; All remaining variables should use the regular toc access sequence.
 @i = dso_local global i32 0, align 4 #0
 @d = dso_local local_unnamed_addr global double 3.141590e+00, align 8
 @f = dso_local local_unnamed_addr global float 0x4005BE76C0000000, align 4 #0
@@ -44,6 +52,16 @@ define dso_local void @write_int(i32 signext %in) {
 ; TEST64:           la 4, i[TD](2)
 ; TEST64-NEXT:      stw 3, 0(4)
 
+; CHECK32LARGE: name:            write_int
+; CHECK32LARGE:      %[[SCRATCH1:[0-9]+]]:gprc_and_gprc_nor0 = ADDIStocHA $r2, @i
+; CHECK32LARGE-NEXT: %[[SCRATCH2:[0-9]+]]:gprc_and_gprc_nor0 = ADDItocL killed %[[SCRATCH1]], @i
+; CHECK32LARGE-NEXT: STW %{{[0-9]+}}, 0, killed %[[SCRATCH2]] :: (store (s32) into @i)
+
+; FIXME: peephole optimization opportunity for lower part relocation @l to the consuming stw
+; TEST32LARGE:         .write_int:
+; TEST32LARGE:          addis 4, i[TD]@u(2)
+; TEST32LARGE-NEXT:	la 4, i[TD]@l(4)
+; TEST32LARGE-NEXT:	stw 3, 0(4)
 
 define dso_local i64 @read_ll() {
   entry:
@@ -70,6 +88,15 @@ define dso_local i64 @read_ll() {
 ; TEST64:         ld 3, L..C0(2)
 ; TEST64-NEXT:    ld 3, 0(3)
 
+; CHECK32LARGE: name:            read_ll
+; CHECK32LARGE: %[[SCRATCH1:[0-9]+]]:gprc_and_gprc_nor0 = ADDIStocHA $r2, @ll
+; CHECK32LARGE: LWZtocL @ll, killed %[[SCRATCH1]] :: (load (s32) from got)
+
+; TEST32LARGE:         .read_ll:
+; TEST32LARGE:          addis 3, L..C0@u(2)
+; TEST32LARGE-NEXT:	lwz 4, L..C0@l(3)
+; TEST32LARGE-NEXT:	lwz 3, 0(4)
+; TEST32LARGE-NEXT:	lwz 4, 4(4)
 
 define dso_local float @read_float() {
   entry:
@@ -96,6 +123,16 @@ define dso_local float @read_float() {
 ; TEST64:         la 3, f[TD](2)
 ; TEST64-NEXT:    lfs 1, 0(3)
 
+; CHECK32LARGE: name:            read_float
+; CHECK32LARGE:      %[[SCRATCH1:[0-9]+]]:gprc_and_gprc_nor0 = ADDIStocHA $r2, @f
+; CHECK32LARGE-NEXT: %[[SCRATCH2:[0-9]+]]:gprc_and_gprc_nor0 = ADDItocL killed %[[SCRATCH1]], @f
+; CHECK32LARGE-NEXT: LFS 0, killed %[[SCRATCH2]] :: (dereferenceable load (s32) from @f)
+
+; FIXME: peephole optimization opportunity for lower part relocation @l to the consuming lfs
+; TEST32LARGE:         .read_float:
+; TEST32LARGE:          addis 3, f[TD]@u(2)
+; TEST32LARGE-NEXT:	la 3, f[TD]@l(3)
+; TEST32LARGE-NEXT:	lfs 1, 0(3)
 
 define dso_local void @write_double(double %in) {
   entry:
@@ -121,6 +158,14 @@ define dso_local void @write_double(double %in) {
 ; TEST64:         ld 3, L..C1(2)
 ; TEST64-NEXT:    stfd 1, 0(3)
 
+; CHECK32LARGE: name:            write_double
+; CHECK32LARGE: %[[SCRATCH1:[0-9]+]]:gprc_and_gprc_nor0 = ADDIStocHA $r2, @d
+; CHECK32LARGE: LWZtocL @d, killed %[[SCRATCH1]] :: (load (s32) from got)
+
+; TEST32LARGE:         .write_double:
+; TEST32LARGE:          addis 3, L..C1@u(2)
+; TEST32LARGE-NEXT:	lwz 3, L..C1@l(3)
+; TEST32LARGE-NEXT:	stfd 1, 0(3)
 
 define dso_local nonnull ptr @addr() {
   entry:
@@ -144,6 +189,15 @@ define dso_local nonnull ptr @addr() {
 ; TEST64:       .addr
 ; TEST64:         la 3, i[TD](2)
 
+; CHECK32LARGE: name:            addr
+; CHECK32LARGE:      %[[SCRATCH1:[0-9]+]]:gprc_and_gprc_nor0 = ADDIStocHA $r2, @i
+; CHECK32LARGE-NEXT: %[[SCRATCH2:[0-9]+]]:gprc = ADDItocL killed %[[SCRATCH1]], @i
+; CHECK32LARGE-NEXT: $r3 = COPY %[[SCRATCH2]]
+
+; TEST32LARGE:         .addr:
+; TEST32LARGE:          addis 3, i[TD]@u(2)
+; TEST32LARGE-NEXT:	la 3, i[TD]@l(3)
+
 ; TEST32:         .toc
 ; TEST32:           .tc ll[TC],ll[RW]
 ; TEST32-NOT:       .csect ll[TD]
@@ -170,4 +224,17 @@ define dso_local nonnull ptr @addr() {
 ; TEST64-NEXT:      .globl f[TD]
 ; TEST64-NOT:       .tc f[TD],f[RW]
 
+; TEST32LARGE:         .toc
+; TEST32LARGE:           .tc ll[TE],ll[RW]
+; TEST32LARGE-NOT:       .csect ll[TD]
+; TEST32LARGE:           .tc d[TE],d[RW]
+; TEST32LARGE-NOT:       .csect d[TD],2
+; TEST32LARGE:           .csect i[TD],2
+; TEST32LARGE-NEXT:      .globl  i[TD]
+; TEST32LARGE-NEXT:      .align  2
+; TEST32LARGE-NOT:       .tc i[TE],i[RW]
+; TEST32LARGE:           .csect f[TD],2
+; TEST32LARGE-NEXT:      .globl f[TD]
+; TEST32LARGE-NOT:       .tc f[TE],f[RW]
+
 attributes #0 = { "toc-data" }

From edbeae373489a2e710f328ceba50b4740c738217 Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Wed, 17 Apr 2024 21:12:08 +0800
Subject: [PATCH 743/886] [RISCV] Explicitly bail if something modifies
 VL/VTYPE in doLocalPostpass

If an instruction between MI and NextMI uses VL or VTYPE we demand the
respective fields so as to not clobber them at their uses. But we don't
consider if something might modify VL or VTYPE, and will happily coalesce
two vsetvlis when we need to preserve them.

This fixes this by skipping to the next vsetvli. Demanding the fields isn't
enough, as we need to preserve the VL and VTYPE values even if no fields
are demanded.

In practice this doesn't happen, presumably due to there not being any
instructions that write to VL or VTYPE without reading them. But I noticed
this whilst working on a separate patch and split it out.
---
 llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp   | 3 +++
 llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir | 3 ++-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
index aab91adbb64be..fa37d1ccccd73 100644
--- a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
@@ -1538,6 +1538,9 @@ void RISCVInsertVSETVLI::doLocalPostpass(MachineBasicBlock &MBB) {
 
     if (!isVectorConfigInstr(MI)) {
       doUnion(Used, getDemanded(MI, MRI, ST));
+      if (MI.isCall() || MI.isInlineAsm() || MI.modifiesRegister(RISCV::VL) ||
+          MI.modifiesRegister(RISCV::VTYPE))
+        NextMI = nullptr;
       continue;
     }
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir b/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
index d9a87c2cb12a8..e8620c848f8d3 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
+++ b/llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.mir
@@ -516,9 +516,10 @@ body:             |
     ; CHECK-LABEL: name: postpass_modify_vl
     ; CHECK: liveins: $x1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: $x0 = PseudoVSETIVLI 3, 216 /* e64, m1, ta, ma */, implicit-def $vl, implicit-def $vtype
+    ; CHECK-NEXT: dead $x0 = PseudoVSETIVLI 3, 216 /* e64, m1, ta, ma */, implicit-def $vl, implicit-def $vtype
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:gpr = COPY $vtype
     ; CHECK-NEXT: $vl = COPY $x1
+    ; CHECK-NEXT: dead $x0 = PseudoVSETIVLI 3, 216 /* e64, m1, ta, ma */, implicit-def $vl, implicit-def $vtype
     ; CHECK-NEXT: [[PseudoVADD_VV_M1_:%[0-9]+]]:vr = PseudoVADD_VV_M1 $noreg, $noreg, $noreg, 3, 6 /* e64 */, 0 /* tu, mu */, implicit $vl, implicit $vtype
     ; CHECK-NEXT: PseudoRET
     dead $x0 = PseudoVSETIVLI 3, 216, implicit-def $vl, implicit-def $vtype

From 38205717501237f2b7a57eaabe65a8367e5f91c3 Mon Sep 17 00:00:00 2001
From: Aaron Ballman <aaron@aaronballman.com>
Date: Tue, 16 Apr 2024 14:57:27 -0400
Subject: [PATCH 744/886] [C99] Remove WG14 N522 from the C status page

This paper is about type compatibility rules that changed in C99, but
this is only applicable across translation units and so there's nothing
for us to test. The specific change was that C89 allowed different tag
types (e.g., struct and union) to be compatible and C99 tightened that
restriction. This is a case where the user gets whatever they get if
they link two TUs with incompatible tag types.
---
 clang/www/c_status.html | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/clang/www/c_status.html b/clang/www/c_status.html
index 9893170ae8473..dfc1afefda245 100644
--- a/clang/www/c_status.html
+++ b/clang/www/c_status.html
@@ -295,11 +295,6 @@ <h2 id="c99">C99 implementation status</h2>
       <td>N570</td>
       <td class="full" align="center">Yes</td>
     </tr>
-    <tr>
-      <td>new structure type compatibility (tag compatibility)</td>
-      <td>N522</td>
-      <td class="unknown" align="center">Unknown</td>
-    </tr>
     <tr>
       <td>additional predefined macro names</td>
       <td>Unknown</td>

From 41b7341d6b27adf81262a5a0bd4e430675b73bbb Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Wed, 17 Apr 2024 14:56:46 +0100
Subject: [PATCH 745/886] [VPlan] Factor out helper to recursively collect all
 users (NFCI).

Factor out logic to collect all users recursively to be re-used
in https://github.com/llvm/llvm-project/pull/87816.
---
 llvm/lib/Transforms/Vectorize/VPlan.h         |  5 ++++
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 30 +++++++++----------
 2 files changed, 19 insertions(+), 16 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 148227f1f1a57..334b10e2e5d09 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1049,6 +1049,11 @@ class VPRecipeWithIRFlags : public VPSingleDefRecipe {
            R->getVPDefID() == VPRecipeBase::VPVectorPointerSC;
   }
 
+  static inline bool classof(const VPUser *U) {
+    auto *R = dyn_cast<VPRecipeBase>(U);
+    return R && classof(R);
+  }
+
   /// Drop all poison-generating flags.
   void dropPoisonGeneratingFlags() {
     // NOTE: This needs to be kept in-sync with
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 382bf5ac11405..78d0b5b95c5ec 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -852,6 +852,18 @@ bool VPlanTransforms::adjustFixedOrderRecurrences(VPlan &Plan,
   return true;
 }
 
+static SmallVector<VPUser *> collectUsersRecursively(VPValue *V) {
+  SetVector<VPUser *> Users(V->user_begin(), V->user_end());
+  for (unsigned I = 0; I != Users.size(); ++I) {
+    VPRecipeBase *Cur = dyn_cast<VPRecipeBase>(Users[I]);
+    if (!Cur || isa<VPHeaderPHIRecipe>(Cur))
+      continue;
+    for (VPValue *V : Cur->definedValues())
+      Users.insert(V->user_begin(), V->user_end());
+  }
+  return Users.takeVector();
+}
+
 void VPlanTransforms::clearReductionWrapFlags(VPlan &Plan) {
   for (VPRecipeBase &R :
        Plan.getVectorLoopRegion()->getEntryBasicBlock()->phis()) {
@@ -863,24 +875,10 @@ void VPlanTransforms::clearReductionWrapFlags(VPlan &Plan) {
     if (RK != RecurKind::Add && RK != RecurKind::Mul)
       continue;
 
-    SmallSetVector<VPValue *, 8> Worklist;
-    Worklist.insert(PhiR);
-
-    for (unsigned I = 0; I != Worklist.size(); ++I) {
-      VPValue *Cur = Worklist[I];
-      if (auto *RecWithFlags =
-              dyn_cast<VPRecipeWithIRFlags>(Cur->getDefiningRecipe())) {
+    for (VPUser *U : collectUsersRecursively(PhiR))
+      if (auto *RecWithFlags = dyn_cast<VPRecipeWithIRFlags>(U)) {
         RecWithFlags->dropPoisonGeneratingFlags();
       }
-
-      for (VPUser *U : Cur->users()) {
-        auto *UserRecipe = dyn_cast<VPRecipeBase>(U);
-        if (!UserRecipe)
-          continue;
-        for (VPValue *V : UserRecipe->definedValues())
-          Worklist.insert(V);
-      }
-    }
   }
 }
 

From 856d1c44103f09f2ed0448001de9dcda63055733 Mon Sep 17 00:00:00 2001
From: Jay Foad <jay.foad@amd.com>
Date: Wed, 17 Apr 2024 14:58:13 +0100
Subject: [PATCH 746/886] [AMDGPU] Fix predicates for BUFFER_ATOMIC_FMIN/FMAX
 patterns (#89066)

Use OtherPredicates to avoid interfering with other uses of
SubtargetPredicate for GFX12.
---
 llvm/lib/Target/AMDGPU/BUFInstructions.td     |  2 +-
 .../AMDGPU/fp-min-max-buffer-atomics.ll       | 72 +++++++++++++++++++
 2 files changed, 73 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AMDGPU/BUFInstructions.td b/llvm/lib/Target/AMDGPU/BUFInstructions.td
index 273f92abf3546..8053d89aeb0a8 100644
--- a/llvm/lib/Target/AMDGPU/BUFInstructions.td
+++ b/llvm/lib/Target/AMDGPU/BUFInstructions.td
@@ -1726,7 +1726,7 @@ let SubtargetPredicate = isGFX12Plus in {
   defm : SIBufferAtomicPat_Common<"SIbuffer_atomic_cond_sub_u32", i32, "BUFFER_ATOMIC_COND_SUB_U32_VBUFFER", ["noret"]>;
 }
 
-let SubtargetPredicate = isGFX6GFX7GFX10Plus in {
+let OtherPredicates = [isGFX6GFX7GFX10Plus] in {
   defm : SIBufferAtomicPat<"SIbuffer_atomic_fmin", f32, "BUFFER_ATOMIC_FMIN">;
   defm : SIBufferAtomicPat<"SIbuffer_atomic_fmax", f32, "BUFFER_ATOMIC_FMAX">;
 }
diff --git a/llvm/test/CodeGen/AMDGPU/fp-min-max-buffer-atomics.ll b/llvm/test/CodeGen/AMDGPU/fp-min-max-buffer-atomics.ll
index 0c62b52eb92af..587340c7aa342 100644
--- a/llvm/test/CodeGen/AMDGPU/fp-min-max-buffer-atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp-min-max-buffer-atomics.ll
@@ -4,12 +4,14 @@
 ; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck %s -check-prefix=GFX10
 ; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1030 -verify-machineinstrs | FileCheck %s -check-prefix=GFX1030
 ; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs | FileCheck %s -check-prefix=GFX1100
+; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs | FileCheck %s -check-prefix=GFX12
 
 ; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s -check-prefix=G_SI
 ; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=hawaii -verify-machineinstrs | FileCheck %s  -check-prefix=G_GFX7
 ; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck %s -check-prefix=G_GFX10
 ; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=gfx1030 -verify-machineinstrs | FileCheck %s -check-prefix=G_GFX1030
 ; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs | FileCheck %s -check-prefix=G_GFX1100
+; RUN: llc < %s -global-isel -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs | FileCheck %s -check-prefix=GFX12
 
 declare float @llvm.amdgcn.raw.buffer.atomic.fmin.f32(float, <4 x i32>, i32, i32, i32 immarg)
 declare float @llvm.amdgcn.raw.buffer.atomic.fmax.f32(float, <4 x i32>, i32, i32, i32 immarg)
@@ -70,6 +72,18 @@ define amdgpu_kernel void @raw_buffer_atomic_min_noret_f32(<4 x i32> inreg %rsrc
 ; GFX1100-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_min_noret_f32:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[4:5], s[0:1], 0x34
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX12-NEXT:    buffer_atomic_min_num_f32 v0, v1, s[0:3], null offen
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_min_noret_f32:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xd
@@ -170,6 +184,15 @@ define amdgpu_ps void @raw_buffer_atomic_min_rtn_f32(<4 x i32> inreg %rsrc, floa
 ; GFX1100-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_min_rtn_f32:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    buffer_atomic_min_num_f32 v0, v1, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    global_store_b32 v[0:1], v0, off
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_min_rtn_f32:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    buffer_atomic_fmin v0, v1, s[0:3], 0 offen glc
@@ -292,6 +315,20 @@ define amdgpu_kernel void @raw_buffer_atomic_min_rtn_f32_off4_slc(<4 x i32> inre
 ; GFX1100-NEXT:    ds_store_b32 v1, v0
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b96 s[4:6], s[0:1], 0x34
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX12-NEXT:    s_mov_b32 s4, 4
+; GFX12-NEXT:    buffer_atomic_min_num_f32 v0, v1, s[0:3], s4 offen th:TH_ATOMIC_NT_RETURN
+; GFX12-NEXT:    v_mov_b32_e32 v1, s6
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    ds_store_b32 v1, v0
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0xd
@@ -427,6 +464,18 @@ define amdgpu_kernel void @raw_buffer_atomic_max_noret_f32(<4 x i32> inreg %rsrc
 ; GFX1100-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_max_noret_f32:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[4:5], s[0:1], 0x34
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX12-NEXT:    buffer_atomic_max_num_f32 v0, v1, s[0:3], null offen
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_max_noret_f32:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0xd
@@ -527,6 +576,15 @@ define amdgpu_ps void @raw_buffer_atomic_max_rtn_f32(<4 x i32> inreg %rsrc, floa
 ; GFX1100-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_max_rtn_f32:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    buffer_atomic_max_num_f32 v0, v1, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    global_store_b32 v[0:1], v0, off
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_max_rtn_f32:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    buffer_atomic_fmax v0, v1, s[0:3], 0 offen glc
@@ -641,6 +699,20 @@ define amdgpu_kernel void @raw_buffer_atomic_max_rtn_f32_off4_slc(<4 x i32> inre
 ; GFX1100-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1100-NEXT:    s_endpgm
 ;
+; GFX12-LABEL: raw_buffer_atomic_max_rtn_f32_off4_slc:
+; GFX12:       ; %bb.0: ; %main_body
+; GFX12-NEXT:    s_load_b256 s[0:7], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s5
+; GFX12-NEXT:    s_mov_b32 s4, 4
+; GFX12-NEXT:    buffer_atomic_max_num_f32 v0, v1, s[0:3], s4 offen th:TH_ATOMIC_NT_RETURN
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    global_store_b32 v1, v0, s[6:7]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+;
 ; G_SI-LABEL: raw_buffer_atomic_max_rtn_f32_off4_slc:
 ; G_SI:       ; %bb.0: ; %main_body
 ; G_SI-NEXT:    s_load_dwordx8 s[0:7], s[0:1], 0x9

From 4f88c2311130791cf69da34b743b1b3ba7584a7b Mon Sep 17 00:00:00 2001
From: Guray Ozen <guray.ozen@gmail.com>
Date: Wed, 17 Apr 2024 15:59:18 +0200
Subject: [PATCH 747/886] [mlir][py] Add NVGPU's `TensorMapDescriptorType` in
 py bindings (#88855)

This PR adds NVGPU dialects' TensorMapDescriptorType in the py bindings.

This is a follow-up issue from [this
PR](https://github.com/llvm/llvm-project/pull/87153#discussion_r1546193095)
---
 mlir/include/mlir-c/Dialect/NVGPU.h       | 11 ++++++
 mlir/lib/Bindings/Python/DialectNVGPU.cpp | 41 +++++++++++++++++++++++
 mlir/lib/CAPI/Dialect/NVGPU.cpp           | 18 ++++++++++
 mlir/python/CMakeLists.txt                | 13 +++++++
 mlir/python/mlir/dialects/nvgpu.py        |  1 +
 mlir/test/python/dialects/nvgpu.py        | 17 ++++++++++
 6 files changed, 101 insertions(+)
 create mode 100644 mlir/lib/Bindings/Python/DialectNVGPU.cpp

diff --git a/mlir/include/mlir-c/Dialect/NVGPU.h b/mlir/include/mlir-c/Dialect/NVGPU.h
index 580d566794c09..e58015a4a3421 100644
--- a/mlir/include/mlir-c/Dialect/NVGPU.h
+++ b/mlir/include/mlir-c/Dialect/NVGPU.h
@@ -11,6 +11,7 @@
 #define MLIR_C_DIALECT_NVGPU_H
 
 #include "mlir-c/IR.h"
+#include "mlir-c/Support.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -18,6 +19,16 @@ extern "C" {
 
 MLIR_DECLARE_CAPI_DIALECT_REGISTRATION(NVGPU, nvgpu);
 
+//===---------------------------------------------------------------------===//
+// TensorMapDescriptorType
+//===---------------------------------------------------------------------===//
+
+MLIR_CAPI_EXPORTED bool mlirTypeIsANVGPUTensorMapDescriptorType(MlirType type);
+
+MLIR_CAPI_EXPORTED MlirType mlirNVGPUTensorMapDescriptorTypeGet(
+    MlirContext ctx, MlirType tensorMemrefType, int swizzle, int l2promo,
+    int oobFill, int interleave);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/mlir/lib/Bindings/Python/DialectNVGPU.cpp b/mlir/lib/Bindings/Python/DialectNVGPU.cpp
new file mode 100644
index 0000000000000..341e4d55bcf21
--- /dev/null
+++ b/mlir/lib/Bindings/Python/DialectNVGPU.cpp
@@ -0,0 +1,41 @@
+//===--- DialectNvgpu.cpp - Pybind module for Nvgpu dialect API support ---===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir-c/Dialect/NVGPU.h"
+#include "mlir-c/IR.h"
+#include "mlir/Bindings/Python/PybindAdaptors.h"
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+using namespace llvm;
+using namespace mlir;
+using namespace mlir::python;
+using namespace mlir::python::adaptors;
+
+static void populateDialectNvgpuSubmodule(const pybind11::module &m) {
+  auto nvgpuTensorMapDescriptorType = mlir_type_subclass(
+      m, "TensorMapDescriptorType", mlirTypeIsANVGPUTensorMapDescriptorType);
+
+  nvgpuTensorMapDescriptorType.def_classmethod(
+      "get",
+      [](py::object cls, MlirType tensorMemrefType, int swizzle, int l2promo,
+         int oobFill, int interleave, MlirContext ctx) {
+        return cls(mlirNVGPUTensorMapDescriptorTypeGet(
+            ctx, tensorMemrefType, swizzle, l2promo, oobFill, interleave));
+      },
+      "Gets an instance of TensorMapDescriptorType in the same context",
+      py::arg("cls"), py::arg("tensor_type"), py::arg("swizzle"),
+      py::arg("l2promo"), py::arg("oob_fill"), py::arg("interleave"),
+      py::arg("ctx") = py::none());
+}
+
+PYBIND11_MODULE(_mlirDialectsNvgpu, m) {
+  m.doc() = "MLIR NVGPU dialect.";
+
+  populateDialectNvgpuSubmodule(m);
+}
diff --git a/mlir/lib/CAPI/Dialect/NVGPU.cpp b/mlir/lib/CAPI/Dialect/NVGPU.cpp
index 02d10954a0377..e6da529e1b6b5 100644
--- a/mlir/lib/CAPI/Dialect/NVGPU.cpp
+++ b/mlir/lib/CAPI/Dialect/NVGPU.cpp
@@ -9,5 +9,23 @@
 #include "mlir-c/Dialect/NVGPU.h"
 #include "mlir/CAPI/Registration.h"
 #include "mlir/Dialect/NVGPU/IR/NVGPUDialect.h"
+#include "mlir/IR/BuiltinTypes.h"
+
+using namespace mlir;
+using namespace mlir::nvgpu;
 
 MLIR_DEFINE_CAPI_DIALECT_REGISTRATION(NVGPU, nvgpu, mlir::nvgpu::NVGPUDialect)
+
+bool mlirTypeIsANVGPUTensorMapDescriptorType(MlirType type) {
+  return isa<nvgpu::TensorMapDescriptorType>(unwrap(type));
+}
+
+MlirType mlirNVGPUTensorMapDescriptorTypeGet(MlirContext ctx,
+                                             MlirType tensorMemrefType,
+                                             int swizzle, int l2promo,
+                                             int oobFill, int interleave) {
+  return wrap(nvgpu::TensorMapDescriptorType::get(
+      unwrap(ctx), cast<MemRefType>(unwrap(tensorMemrefType)),
+      TensorMapSwizzleKind(swizzle), TensorMapL2PromoKind(l2promo),
+      TensorMapOOBKind(oobFill), TensorMapInterleaveKind(interleave)));
+}
diff --git a/mlir/python/CMakeLists.txt b/mlir/python/CMakeLists.txt
index c27ee688a0408..0a2dc0754c09d 100644
--- a/mlir/python/CMakeLists.txt
+++ b/mlir/python/CMakeLists.txt
@@ -524,6 +524,19 @@ declare_mlir_python_extension(MLIRPythonExtension.Dialects.Quant.Pybind
     MLIRCAPIQuant
 )
 
+declare_mlir_python_extension(MLIRPythonExtension.Dialects.NVGPU.Pybind
+  MODULE_NAME _mlirDialectsNvgpu
+  ADD_TO_PARENT MLIRPythonSources.Dialects.nvgpu
+  ROOT_DIR "${PYTHON_SOURCE_DIR}"
+  SOURCES
+    DialectNVGPU.cpp
+  PRIVATE_LINK_LIBS
+    LLVMSupport
+  EMBED_CAPI_LINK_LIBS
+    MLIRCAPIIR
+    MLIRCAPINVGPU
+)
+
 declare_mlir_python_extension(MLIRPythonExtension.Dialects.PDL.Pybind
   MODULE_NAME _mlirDialectsPDL
   ADD_TO_PARENT MLIRPythonSources.Dialects.pdl
diff --git a/mlir/python/mlir/dialects/nvgpu.py b/mlir/python/mlir/dialects/nvgpu.py
index 2f6993b768ca5..e19bf610ea33a 100644
--- a/mlir/python/mlir/dialects/nvgpu.py
+++ b/mlir/python/mlir/dialects/nvgpu.py
@@ -4,3 +4,4 @@
 
 from ._nvgpu_ops_gen import *
 from ._nvgpu_enum_gen import *
+from .._mlir_libs._mlirDialectsNvgpu import *
diff --git a/mlir/test/python/dialects/nvgpu.py b/mlir/test/python/dialects/nvgpu.py
index 3158388f0e686..6df32bdd3c273 100644
--- a/mlir/test/python/dialects/nvgpu.py
+++ b/mlir/test/python/dialects/nvgpu.py
@@ -15,6 +15,23 @@ def constructAndPrintInModule(f):
     return f
 
 
+# CHECK-LABEL: testTypes
+@constructAndPrintInModule
+def testTypes():
+    tensorMemrefType = MemRefType.get(
+        (128, 64), F16Type.get(), memory_space=Attribute.parse("3")
+    )
+    # CHECK: !nvgpu.tensormap.descriptor<tensor = memref<128x64xf16, 3>, swizzle = swizzle_128b, l2promo = l2promo_256b, oob = nan, interleave = none>
+    tma_desc = nvgpu.TensorMapDescriptorType.get(
+        tensorMemrefType,
+        nvgpu.TensorMapSwizzleKind.SWIZZLE_128B,
+        nvgpu.TensorMapL2PromoKind.L2PROMO_256B,
+        nvgpu.TensorMapOOBKind.OOB_NAN,
+        nvgpu.TensorMapInterleaveKind.INTERLEAVE_NONE,
+    )
+    print(tma_desc)
+
+
 # CHECK-LABEL: testSmoke
 @constructAndPrintInModule
 def testSmoke():

From fda04b1caaf1a61b208f23e717a2db6d9b861f5a Mon Sep 17 00:00:00 2001
From: Rajveer Singh Bharadwaj <rajveer.developer@icloud.com>
Date: Wed, 17 Apr 2024 19:37:18 +0530
Subject: [PATCH 748/886] [libc] Replace mentions of `LIBC_FULLBUILD` with
 `LLVM_LIBC_FULL_BUILD` in 'examples/' (#88657)

Resolves #88328
---
 libc/examples/README.md      | 2 +-
 libc/examples/examples.cmake | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/libc/examples/README.md b/libc/examples/README.md
index 36b886090c6c1..1bc4a67294f2a 100644
--- a/libc/examples/README.md
+++ b/libc/examples/README.md
@@ -59,7 +59,7 @@ have installed them, you have to inform CMake that we are linking against the
 full libc as follows:
 
 ```bash
-cmake ../ -G <GEN> -DLIBC_FULLBUILD=ON    \
+cmake ../ -G <GEN> -DLLVM_LIBC_FULL_BUILD=ON    \
   -DCMAKE_SYSROOT=<SYSROOT>               \
   -DCMAKE_C_COMPILER=<SYSROOT>/bin/clang  \
   -DCMAKE_TRY_COMPILE_TARGET_TYPE=STATIC_LIBRARY
diff --git a/libc/examples/examples.cmake b/libc/examples/examples.cmake
index 81e99e3cbede9..6bb6b41c252f5 100644
--- a/libc/examples/examples.cmake
+++ b/libc/examples/examples.cmake
@@ -4,13 +4,13 @@ function(add_example name)
     ${ARGN}
   )
 
-  if(LIBC_FULLBUILD)
+  if(LLVM_LIBC_FULL_BUILD)
     target_link_options(${name} PRIVATE -static -rtlib=compiler-rt -fuse-ld=lld)
   elseif(LIBC_OVERLAY_ARCHIVE_DIR)
     target_link_directories(${name} PRIVATE ${LIBC_OVERLAY_ARCHIVE_DIR})
     target_link_options(${name} PRIVATE -l:libllvmlibc.a)
   else()
-    message(FATAL_ERROR "Either LIBC_FULLBUILD should be on or "
+    message(FATAL_ERROR "Either LLVM_LIBC_FULL_BUILD should be on or "
                         "LIBC_OVERLAY_ARCHIVE_DIR should be set.")
   endif()
 endfunction()

From d558c090fc78beb6737098f058a084635b893567 Mon Sep 17 00:00:00 2001
From: yronglin <yronglin777@gmail.com>
Date: Wed, 17 Apr 2024 22:11:04 +0800
Subject: [PATCH 749/886] [NFC] Clean dead code in ParsedAttr.h (#89064)

Signed-off-by: yronglin <yronglin777@gmail.com>
---
 clang/include/clang/Sema/ParsedAttr.h | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/clang/include/clang/Sema/ParsedAttr.h b/clang/include/clang/Sema/ParsedAttr.h
index e3857b2f07d9e..25a5fa05b21c7 100644
--- a/clang/include/clang/Sema/ParsedAttr.h
+++ b/clang/include/clang/Sema/ParsedAttr.h
@@ -94,7 +94,7 @@ struct PropertyData {
       : GetterId(getterId), SetterId(setterId) {}
 };
 
-} // namespace
+} // namespace detail
 
 /// Wraps an identifier and optional source location for the identifier.
 struct IdentifierLoc {
@@ -743,11 +743,6 @@ class AttributePool {
                      IdentifierInfo *scopeName, SourceLocation scopeLoc,
                      ArgsUnion *args, unsigned numArgs, ParsedAttr::Form form,
                      SourceLocation ellipsisLoc = SourceLocation()) {
-    size_t temp =
-        ParsedAttr::totalSizeToAlloc<ArgsUnion, detail::AvailabilityData,
-                                     detail::TypeTagForDatatypeData, ParsedType,
-                                     detail::PropertyData>(numArgs, 0, 0, 0, 0);
-    (void)temp;
     void *memory = allocate(
         ParsedAttr::totalSizeToAlloc<ArgsUnion, detail::AvailabilityData,
                                      detail::TypeTagForDatatypeData, ParsedType,

From dbafcd7e6d5b6c3aff74307b2ed5ee1834f275b0 Mon Sep 17 00:00:00 2001
From: Pavel Labath <pavel@labath.sk>
Date: Wed, 17 Apr 2024 14:29:09 +0000
Subject: [PATCH 750/886] [lldb] XFAIL TestDetachResumes on windows

---
 .../API/commands/process/detach-resumes/TestDetachResumes.py     | 1 +
 1 file changed, 1 insertion(+)

diff --git a/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py b/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
index 57727294ddc3d..797db55a45b9b 100644
--- a/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
+++ b/lldb/test/API/commands/process/detach-resumes/TestDetachResumes.py
@@ -12,6 +12,7 @@
 class DetachResumesTestCase(TestBase):
     NO_DEBUG_INFO_TESTCASE = True
 
+    @expectedFailureAll(oslist=["windows"], bugnumber="llvm.org/pr89077")
     def test_detach_resumes(self):
         self.build()
         exe = self.getBuildArtifact()

From e49043512dbdc68319093da46e95a1e331ef837e Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 17 Apr 2024 14:55:53 +0100
Subject: [PATCH 751/886] [CostModel][X86] Update BITREVERSE costs for GFNI
 targets

Inspired by the recent patches by @shamithoke - we have real scheduler model numbers for GFNI instructions now, allowing us to calculate an upper bounds costs table instead of performing it analytically.
---
 .../lib/Target/X86/X86TargetTransformInfo.cpp |  41 +++--
 .../CostModel/X86/bitreverse-codesize.ll      |  64 +++----
 .../CostModel/X86/bitreverse-latency.ll       | 160 +++++++++---------
 .../CostModel/X86/bitreverse-sizelatency.ll   | 160 +++++++++---------
 .../test/Analysis/CostModel/X86/bitreverse.ll | 136 +++++++--------
 5 files changed, 284 insertions(+), 277 deletions(-)

diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index b466624e13348..38064f9792699 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -3820,6 +3820,24 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     { ISD::FSQRT,      MVT::v2f64,   { 27, 27,  1,  1 } }, // vsqrtpd
     { ISD::FSQRT,      MVT::v4f64,   { 54, 54,  1,  3 } }, // vsqrtpd
   };
+  static const CostKindTblEntry GFNICostTbl[] = {
+    { ISD::BITREVERSE, MVT::i8,      {  3,  3,  3,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::i16,     {  3,  3,  4,  6 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::i32,     {  3,  3,  4,  5 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::i64,     {  3,  3,  4,  6 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v16i8,   {  1,  6,  1,  2 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v32i8,   {  1,  6,  1,  2 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v64i8,   {  1,  6,  1,  2 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v8i16,   {  1,  8,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v16i16,  {  1,  9,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v32i16,  {  1,  9,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v4i32,   {  1,  8,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v8i32,   {  1,  9,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v16i32,  {  1,  9,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v2i64,   {  1,  8,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v4i64,   {  1,  9,  2,  4 } }, // gf2p8affineqb
+    { ISD::BITREVERSE, MVT::v8i64,   {  1,  9,  2,  4 } }, // gf2p8affineqb
+  };
   static const CostKindTblEntry GLMCostTbl[] = {
     { ISD::FSQRT,      MVT::f32,     { 19, 20, 1, 1 } }, // sqrtss
     { ISD::FSQRT,      MVT::v4f32,   { 37, 41, 1, 5 } }, // sqrtps
@@ -4156,23 +4174,6 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(OpTy);
     MVT MTy = LT.second;
 
-    // Attempt to lookup cost.
-    if (ISD == ISD::BITREVERSE && ST->hasGFNI() && ST->hasSSSE3() &&
-        MTy.isVector()) {
-      // With PSHUFB the code is very similar for all types. If we have integer
-      // byte operations, we just need a GF2P8AFFINEQB for vXi8. For other types
-      // we also need a PSHUFB.
-      unsigned Cost = MTy.getVectorElementType() == MVT::i8 ? 1 : 2;
-
-      // Without byte operations, we need twice as many GF2P8AFFINEQB and PSHUFB
-      // instructions. We also need an extract and an insert.
-      if (!(MTy.is128BitVector() || (ST->hasAVX2() && MTy.is256BitVector()) ||
-            (ST->hasBWI() && MTy.is512BitVector())))
-        Cost = Cost * 2 + 2;
-
-      return LT.first * Cost;
-    }
-
     // Without BMI/LZCNT see if we're only looking for a *_ZERO_UNDEF cost.
     if (((ISD == ISD::CTTZ && !ST->hasBMI()) ||
          (ISD == ISD::CTLZ && !ST->hasLZCNT())) &&
@@ -4230,6 +4231,12 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
           return adjustTableCost(Entry->ISD, *KindCost, LT.first,
                                  ICA.getFlags());
 
+    if (ST->hasGFNI())
+      if (const auto *Entry = CostTableLookup(GFNICostTbl, ISD, MTy))
+        if (auto KindCost = Entry->Cost[CostKind])
+          return adjustTableCost(Entry->ISD, *KindCost, LT.first,
+                                 ICA.getFlags());
+
     if (ST->hasCDI())
       if (const auto *Entry = CostTableLookup(AVX512CDCostTbl, ISD, MTy))
         if (auto KindCost = Entry->Cost[CostKind])
diff --git a/llvm/test/Analysis/CostModel/X86/bitreverse-codesize.ll b/llvm/test/Analysis/CostModel/X86/bitreverse-codesize.ll
index 90e49e6ccb81c..e02ba761aa8d2 100644
--- a/llvm/test/Analysis/CostModel/X86/bitreverse-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/bitreverse-codesize.ll
@@ -40,23 +40,23 @@ define i64 @var_bitreverse_i64(i64 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
   %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
@@ -77,23 +77,23 @@ define i32 @var_bitreverse_i32(i32 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
   %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
@@ -114,23 +114,23 @@ define i16 @var_bitreverse_i16(i16 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
   %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
@@ -151,23 +151,23 @@ define i8 @var_bitreverse_i8(i8 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
   %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
@@ -270,7 +270,7 @@ define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i64'
@@ -323,7 +323,7 @@ define <8 x i64> @var_bitreverse_v8i64(<8 x i64> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i64'
@@ -331,7 +331,7 @@ define <8 x i64> @var_bitreverse_v8i64(<8 x i64> %a) {
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i64'
@@ -421,7 +421,7 @@ define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i32'
@@ -474,7 +474,7 @@ define <16 x i32> @var_bitreverse_v16i32(<16 x i32> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i32'
@@ -482,7 +482,7 @@ define <16 x i32> @var_bitreverse_v16i32(<16 x i32> %a) {
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i32'
@@ -572,7 +572,7 @@ define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i16'
@@ -625,7 +625,7 @@ define <32 x i16> @var_bitreverse_v32i16(<32 x i16> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i16'
@@ -633,7 +633,7 @@ define <32 x i16> @var_bitreverse_v32i16(<32 x i16> %a) {
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i16'
@@ -723,7 +723,7 @@ define <32 x i8> @var_bitreverse_v32i8(<32 x i8> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i8'
@@ -776,7 +776,7 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v64i8'
@@ -784,7 +784,7 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v64i8'
diff --git a/llvm/test/Analysis/CostModel/X86/bitreverse-latency.ll b/llvm/test/Analysis/CostModel/X86/bitreverse-latency.ll
index e7d0d05f82429..ba231b985c26b 100644
--- a/llvm/test/Analysis/CostModel/X86/bitreverse-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/bitreverse-latency.ll
@@ -40,23 +40,23 @@ define i64 @var_bitreverse_i64(i64 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
   %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
@@ -77,23 +77,23 @@ define i32 @var_bitreverse_i32(i32 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
   %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
@@ -114,23 +114,23 @@ define i16 @var_bitreverse_i16(i16 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
   %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
@@ -151,23 +151,23 @@ define i8 @var_bitreverse_i8(i8 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
   %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
@@ -221,23 +221,23 @@ define <2 x i64> @var_bitreverse_v2i64(<2 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v2i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
   %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
@@ -270,23 +270,23 @@ define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
   %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
@@ -323,23 +323,23 @@ define <8 x i64> @var_bitreverse_v8i64(<8 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
   %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
@@ -376,23 +376,23 @@ define <4 x i32> @var_bitreverse_v4i32(<4 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
   %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
@@ -425,23 +425,23 @@ define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
   %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
@@ -478,23 +478,23 @@ define <16 x i32> @var_bitreverse_v16i32(<16 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
   %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
@@ -531,23 +531,23 @@ define <8 x i16> @var_bitreverse_v8i16(<8 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
   %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
@@ -580,23 +580,23 @@ define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
   %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
@@ -633,23 +633,23 @@ define <32 x i16> @var_bitreverse_v32i16(<32 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v32i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
   %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
@@ -686,23 +686,23 @@ define <16 x i8> @var_bitreverse_v16i8(<16 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
   %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
@@ -735,23 +735,23 @@ define <32 x i8> @var_bitreverse_v32i8(<32 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v32i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
   %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
@@ -788,23 +788,23 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v64i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
   %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
diff --git a/llvm/test/Analysis/CostModel/X86/bitreverse-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/bitreverse-sizelatency.ll
index 1ba93af46ebdc..d60fac228fc06 100644
--- a/llvm/test/Analysis/CostModel/X86/bitreverse-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/bitreverse-sizelatency.ll
@@ -40,23 +40,23 @@ define i64 @var_bitreverse_i64(i64 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i64 %bitreverse
 ;
   %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
@@ -77,23 +77,23 @@ define i32 @var_bitreverse_i32(i32 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 %bitreverse
 ;
   %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
@@ -114,23 +114,23 @@ define i16 @var_bitreverse_i16(i16 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i16 %bitreverse
 ;
   %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
@@ -151,23 +151,23 @@ define i8 @var_bitreverse_i8(i8 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i8 %bitreverse
 ;
   %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
@@ -217,23 +217,23 @@ define <2 x i64> @var_bitreverse_v2i64(<2 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v2i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <2 x i64> %bitreverse
 ;
   %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
@@ -270,23 +270,23 @@ define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i64> %bitreverse
 ;
   %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
@@ -323,23 +323,23 @@ define <8 x i64> @var_bitreverse_v8i64(<8 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i64> %bitreverse
 ;
   %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
@@ -372,23 +372,23 @@ define <4 x i32> @var_bitreverse_v4i32(<4 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <4 x i32> %bitreverse
 ;
   %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
@@ -425,23 +425,23 @@ define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i32> %bitreverse
 ;
   %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
@@ -478,23 +478,23 @@ define <16 x i32> @var_bitreverse_v16i32(<16 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i32> %bitreverse
 ;
   %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
@@ -527,23 +527,23 @@ define <8 x i16> @var_bitreverse_v8i16(<8 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %bitreverse
 ;
   %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
@@ -580,23 +580,23 @@ define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %bitreverse
 ;
   %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
@@ -633,23 +633,23 @@ define <32 x i16> @var_bitreverse_v32i16(<32 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v32i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %bitreverse
 ;
   %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
@@ -682,23 +682,23 @@ define <16 x i8> @var_bitreverse_v16i8(<16 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %bitreverse
 ;
   %bitreverse = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> %a)
@@ -735,23 +735,23 @@ define <32 x i8> @var_bitreverse_v32i8(<32 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v32i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %bitreverse
 ;
   %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
@@ -788,23 +788,23 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v64i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %bitreverse
 ;
   %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
diff --git a/llvm/test/Analysis/CostModel/X86/bitreverse.ll b/llvm/test/Analysis/CostModel/X86/bitreverse.ll
index 0b76b0d527ba1..a890147fee465 100644
--- a/llvm/test/Analysis/CostModel/X86/bitreverse.ll
+++ b/llvm/test/Analysis/CostModel/X86/bitreverse.ll
@@ -40,23 +40,23 @@ define i64 @var_bitreverse_i64(i64 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i64 %bitreverse
 ;
   %bitreverse = call i64 @llvm.bitreverse.i64(i64 %a)
@@ -77,23 +77,23 @@ define i32 @var_bitreverse_i32(i32 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 %bitreverse
 ;
   %bitreverse = call i32 @llvm.bitreverse.i32(i32 %a)
@@ -114,23 +114,23 @@ define i16 @var_bitreverse_i16(i16 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i16 %bitreverse
 ;
   %bitreverse = call i16 @llvm.bitreverse.i16(i16 %a)
@@ -151,23 +151,23 @@ define i8 @var_bitreverse_i8(i8 %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_i8'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_i8'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_i8'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i8 %bitreverse
 ;
   %bitreverse = call i8 @llvm.bitreverse.i8(i8 %a)
@@ -217,23 +217,23 @@ define <2 x i64> @var_bitreverse_v2i64(<2 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v2i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v2i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bitreverse
 ;
   %bitreverse = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
@@ -270,23 +270,23 @@ define <4 x i64> @var_bitreverse_v4i64(<4 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bitreverse
 ;
   %bitreverse = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
@@ -323,23 +323,23 @@ define <8 x i64> @var_bitreverse_v8i64(<8 x i64> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i64'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i64'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bitreverse
 ;
   %bitreverse = call <8 x i64> @llvm.bitreverse.v8i64(<8 x i64> %a)
@@ -372,23 +372,23 @@ define <4 x i32> @var_bitreverse_v4i32(<4 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v4i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v4i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bitreverse
 ;
   %bitreverse = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> %a)
@@ -425,23 +425,23 @@ define <8 x i32> @var_bitreverse_v8i32(<8 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bitreverse
 ;
   %bitreverse = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
@@ -478,23 +478,23 @@ define <16 x i32> @var_bitreverse_v16i32(<16 x i32> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i32'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i32'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bitreverse
 ;
   %bitreverse = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
@@ -527,23 +527,23 @@ define <8 x i16> @var_bitreverse_v8i16(<8 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v8i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v8i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bitreverse
 ;
   %bitreverse = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> %a)
@@ -580,23 +580,23 @@ define <16 x i16> @var_bitreverse_v16i16(<16 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v16i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v16i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bitreverse
 ;
   %bitreverse = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
@@ -633,23 +633,23 @@ define <32 x i16> @var_bitreverse_v32i16(<32 x i16> %a) {
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNISSE-LABEL: 'var_bitreverse_v32i16'
-; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v32i16'
-; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
+; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
 ; GFNIAVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bitreverse
 ;
   %bitreverse = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
@@ -743,7 +743,7 @@ define <32 x i8> @var_bitreverse_v32i8(<32 x i8> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v32i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v32i8'
@@ -796,7 +796,7 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; GFNISSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX2-LABEL: 'var_bitreverse_v64i8'
@@ -804,7 +804,7 @@ define <64 x i8> @var_bitreverse_v64i8(<64 x i8> %a) {
 ; GFNIAVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512F-LABEL: 'var_bitreverse_v64i8'
-; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
+; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bitreverse = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
 ; GFNIAVX512F-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bitreverse
 ;
 ; GFNIAVX512BW-LABEL: 'var_bitreverse_v64i8'

From 4a5ab13bf5a94ec7f0eabaf24dfe1a5ee720b860 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 17 Apr 2024 15:46:24 +0100
Subject: [PATCH 752/886] [VectorCombine] Remove single quotes from
 "-passes=vector-combine"

These confuse the update_test_checks.py script when run by DOS cmd.exe
---
 llvm/test/Transforms/VectorCombine/AArch64/select-shuffle.ll    | 2 +-
 llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll | 2 +-
 llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/llvm/test/Transforms/VectorCombine/AArch64/select-shuffle.ll b/llvm/test/Transforms/VectorCombine/AArch64/select-shuffle.ll
index eae0879004839..b27c026dbccf0 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/select-shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/select-shuffle.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes='vector-combine' -S %s | FileCheck %s
+; RUN: opt -passes=vector-combine -S %s | FileCheck %s
 
 target triple = "aarch64"
 
diff --git a/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
index d96dfec849167..a1d4ca1a740e7 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes='vector-combine' -S %s | FileCheck %s
+; RUN: opt -passes=vector-combine -S %s | FileCheck %s
 
 target triple = "aarch64"
 
diff --git a/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll b/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
index d69cb75664a8c..a22575ccb1ca2 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -passes='vector-combine' -S %s | FileCheck %s
+; RUN: opt -passes=vector-combine -S %s | FileCheck %s
 
 target triple = "aarch64"
 

From 5d314353fbec1a15cd8900f466dcdcf2af40e8c9 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Wed, 17 Apr 2024 15:53:15 +0100
Subject: [PATCH 753/886] [VPlan] Check for VPWidenLoadRecipe directly in
 truncateToMinBW. (NFCI).

Since ne
After a separate recipe has been introduced for wide loads in
a9bafe91dd0, we can directly check for load recipes in the early
bail-out and remove the redundant bail out for stores.
---
 llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 78d0b5b95c5ec..901ecd10c69d8 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -975,9 +975,7 @@ void VPlanTransforms::truncateToMinimalBitwidths(
            vp_depth_first_deep(Plan.getVectorLoopRegion()))) {
     for (VPRecipeBase &R : make_early_inc_range(*VPBB)) {
       if (!isa<VPWidenRecipe, VPWidenCastRecipe, VPReplicateRecipe,
-               VPWidenSelectRecipe, VPWidenMemoryRecipe>(&R))
-        continue;
-      if (isa<VPWidenStoreRecipe>(&R))
+               VPWidenSelectRecipe, VPWidenLoadRecipe>(&R))
         continue;
 
       VPValue *ResultVPV = R.getVPSingleValue();

From 812963f6aa2adb5e990f273b8ce1a0eabcdefd7f Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Wed, 17 Apr 2024 17:02:07 +0200
Subject: [PATCH 754/886] [libc++][chrono] Improves date formatting. (#86127)

The formatting of years has been done manually since the results of %Y
outside the "typical" range may produce unexpected values. The same
applies to %F which is identical to %Y-%m-%d. None of these conversion
specifiers is affected by the locale used. So it's trivial to manually
handle this case.

This removes several platform specific ifdefs from the tests.
---
 libcxx/include/__chrono/formatter.h           | 16 ++++-----
 .../time.cal.ymd.nonmembers/ostream.pass.cpp  | 34 -------------------
 .../sys_date.ostream.pass.cpp                 | 34 -------------------
 .../formatter.year_month_day.pass.cpp         | 24 -------------
 4 files changed, 7 insertions(+), 101 deletions(-)

diff --git a/libcxx/include/__chrono/formatter.h b/libcxx/include/__chrono/formatter.h
index d932a99f4b998..f76e7b2ea0e86 100644
--- a/libcxx/include/__chrono/formatter.h
+++ b/libcxx/include/__chrono/formatter.h
@@ -322,15 +322,13 @@ _LIBCPP_HIDE_FROM_ABI void __format_chrono_using_chrono_specs(
         __formatter::__format_year(__sstr, __t.tm_year + 1900);
         break;
 
-      case _CharT('F'): {
-        int __year = __t.tm_year + 1900;
-        if (__year < 1000) {
-          __formatter::__format_year(__sstr, __year);
-          __sstr << std::format(_LIBCPP_STATICALLY_WIDEN(_CharT, "-{:02}-{:02}"), __t.tm_mon + 1, __t.tm_mday);
-        } else
-          __facet.put(
-              {__sstr}, __sstr, _CharT(' '), std::addressof(__t), std::to_address(__s), std::to_address(__it + 1));
-      } break;
+      case _CharT('F'):
+        // Depending on the platform's libc the range of supported years is
+        // limited. Instead of testing all conditions use the internal
+        // implementation unconditionally.
+        __formatter::__format_year(__sstr, __t.tm_year + 1900);
+        __sstr << std::format(_LIBCPP_STATICALLY_WIDEN(_CharT, "-{:02}-{:02}"), __t.tm_mon + 1, __t.tm_mday);
+        break;
 
       case _CharT('z'):
         __formatter::__format_zone_offset(__sstr, __z.__offset, false);
diff --git a/libcxx/test/std/time/time.cal/time.cal.ymd/time.cal.ymd.nonmembers/ostream.pass.cpp b/libcxx/test/std/time/time.cal/time.cal.ymd/time.cal.ymd.nonmembers/ostream.pass.cpp
index ffc737fcad5dd..20ffb165558e3 100644
--- a/libcxx/test/std/time/time.cal/time.cal.ymd/time.cal.ymd.nonmembers/ostream.pass.cpp
+++ b/libcxx/test/std/time/time.cal/time.cal.ymd/time.cal.ymd.nonmembers/ostream.pass.cpp
@@ -13,9 +13,6 @@
 // TODO FMT This test should not require std::to_chars(floating-point)
 // XFAIL: availability-fp_to_chars-missing
 
-// TODO FMT Investigate Windows issues.
-// XFAIL: msvc
-
 // REQUIRES: locale.fr_FR.UTF-8
 // REQUIRES: locale.ja_JP.UTF-8
 
@@ -89,20 +86,9 @@ static void test() {
   TEST_EQUAL(stream_c_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}),
              SV("2000-02-29"));
-
-#if defined(_AIX)
-  TEST_EQUAL(stream_c_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_c_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV(""));
-#else                 //  defined(_AIX)
   TEST_EQUAL(stream_c_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 
   TEST_EQUAL(stream_fr_FR_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{-32'768}, std::chrono::month{1}, std::chrono::day{1}}),
@@ -122,19 +108,9 @@ static void test() {
   TEST_EQUAL(stream_fr_FR_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}),
              SV("2000-02-29"));
-#if defined(_AIX)
-  TEST_EQUAL(stream_fr_FR_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_fr_FR_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV(""));
-#else                 // defined(_AIX)
   TEST_EQUAL(stream_fr_FR_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 
   TEST_EQUAL(stream_ja_JP_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{-32'768}, std::chrono::month{1}, std::chrono::day{1}}),
@@ -154,19 +130,9 @@ static void test() {
   TEST_EQUAL(stream_ja_JP_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}),
              SV("2000-02-29"));
-#if defined(_AIX)
-  TEST_EQUAL(stream_ja_JP_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_ja_JP_locale<CharT>(
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
-             SV(""));
-#else                 // defined(_AIX)
   TEST_EQUAL(stream_ja_JP_locale<CharT>(
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 }
 
 int main(int, char**) {
diff --git a/libcxx/test/std/time/time.clock/time.clock.system/sys_date.ostream.pass.cpp b/libcxx/test/std/time/time.clock/time.clock.system/sys_date.ostream.pass.cpp
index 7af3ebf776807..c5645884cfed0 100644
--- a/libcxx/test/std/time/time.clock/time.clock.system/sys_date.ostream.pass.cpp
+++ b/libcxx/test/std/time/time.clock/time.clock.system/sys_date.ostream.pass.cpp
@@ -13,9 +13,6 @@
 // TODO FMT This test should not require std::to_chars(floating-point)
 // XFAIL: availability-fp_to_chars-missing
 
-// TODO FMT Investigate Windows issues.
-// XFAIL: msvc
-
 // REQUIRES: locale.fr_FR.UTF-8
 // REQUIRES: locale.ja_JP.UTF-8
 
@@ -81,20 +78,9 @@ static void test() {
   TEST_EQUAL(stream_c_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}}),
              SV("2000-02-29"));
-
-#if defined(_AIX)
-  TEST_EQUAL(stream_c_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_c_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV(""));
-#else                 //  defined(_AIX)
   TEST_EQUAL(stream_c_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 
   // multiples of days are considered days.
   TEST_EQUAL(stream_c_locale<CharT>(std::chrono::sys_time<std::chrono::weeks>{std::chrono::weeks{3}}),
@@ -112,19 +98,9 @@ static void test() {
   TEST_EQUAL(stream_fr_FR_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}}),
              SV("2000-02-29"));
-#if defined(_AIX)
-  TEST_EQUAL(stream_fr_FR_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_fr_FR_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV(""));
-#else                 // defined(_AIX)
   TEST_EQUAL(stream_fr_FR_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 
   // multiples of days are considered days.
   TEST_EQUAL(stream_fr_FR_locale<CharT>(std::chrono::sys_time<std::chrono::weeks>{std::chrono::weeks{3}}),
@@ -142,19 +118,9 @@ static void test() {
   TEST_EQUAL(stream_ja_JP_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{2000}, std::chrono::month{2}, std::chrono::day{29}}}),
              SV("2000-02-29"));
-#if defined(_AIX)
-  TEST_EQUAL(stream_ja_JP_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV("+32767-12-31"));
-#elif defined(_WIN32) // defined(_AIX)
-  TEST_EQUAL(stream_ja_JP_locale<CharT>(std::chrono::sys_days{
-                 std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
-             SV(""));
-#else                 // defined(_AIX)
   TEST_EQUAL(stream_ja_JP_locale<CharT>(std::chrono::sys_days{
                  std::chrono::year_month_day{std::chrono::year{32'767}, std::chrono::month{12}, std::chrono::day{31}}}),
              SV("32767-12-31"));
-#endif                // defined(_AIX)
 
   // multiples of days are considered days.
   TEST_EQUAL(stream_ja_JP_locale<CharT>(std::chrono::sys_time<std::chrono::weeks>{std::chrono::weeks{3}}),
diff --git a/libcxx/test/std/time/time.syn/formatter.year_month_day.pass.cpp b/libcxx/test/std/time/time.syn/formatter.year_month_day.pass.cpp
index 5a2b7afa37a86..1f2af1cb0530d 100644
--- a/libcxx/test/std/time/time.syn/formatter.year_month_day.pass.cpp
+++ b/libcxx/test/std/time/time.syn/formatter.year_month_day.pass.cpp
@@ -62,17 +62,6 @@ static void test_no_chrono_specs() {
         std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{2}, std::chrono::day{31}});
 
   // Valid year, invalid month, valid day
-#ifdef _WIN32
-  check(SV(" is not a valid date"),
-        SV("{}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{31}});
-  check(SV("****** is not a valid date******"),
-        SV("{:*^32}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{31}});
-  check(SV("*********** is not a valid date"),
-        SV("{:*>31}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{31}});
-#else  // _WIN32
   check(SV("1970-00-31 is not a valid date"),
         SV("{}"),
         std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{31}});
@@ -82,20 +71,8 @@ static void test_no_chrono_specs() {
   check(SV("*1970-00-31 is not a valid date"),
         SV("{:*>31}"),
         std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{31}});
-#endif // _WIN32
 
   // Valid year, invalid month, invalid day
-#ifdef _WIN32
-  check(SV(" is not a valid date"),
-        SV("{}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{32}});
-  check(SV("****** is not a valid date******"),
-        SV("{:*^32}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{32}});
-  check(SV("*********** is not a valid date"),
-        SV("{:*>31}"),
-        std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{32}});
-#else  // _WIN32
   check(SV("1970-00-32 is not a valid date"),
         SV("{}"),
         std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{32}});
@@ -105,7 +82,6 @@ static void test_no_chrono_specs() {
   check(SV("*1970-00-32 is not a valid date"),
         SV("{:*>31}"),
         std::chrono::year_month_day{std::chrono::year{1970}, std::chrono::month{0}, std::chrono::day{32}});
-#endif // _WIN32
 
   // Invalid year, valid month, valid day
   check(SV("-32768-01-31 is not a valid date"),

From 458328ae23d318a5055d5bac66426b8551bce01f Mon Sep 17 00:00:00 2001
From: Vlad Serebrennikov <serebrennikov.vladislav@gmail.com>
Date: Wed, 17 Apr 2024 18:01:55 +0300
Subject: [PATCH 755/886] [clang][NFC] Refactor `Sema::RedeclarationKind`

This patch converts the enum into scoped enum, and moves it into its own header for the time being. It's definition is needed in `Sema.h`, and is going to be needed in upcoming `SemaObjC.h`. `Lookup.h` can't hold it, because it includes `Sema.h`.
---
 clang/include/clang/Sema/Lookup.h             | 43 ++++++++++---------
 clang/include/clang/Sema/Redeclaration.h      | 31 +++++++++++++
 clang/include/clang/Sema/Sema.h               | 40 ++++-------------
 clang/lib/Interpreter/Interpreter.cpp         |  3 +-
 clang/lib/Interpreter/InterpreterUtils.cpp    |  2 +-
 clang/lib/Sema/SemaDecl.cpp                   | 24 ++++++-----
 clang/lib/Sema/SemaDeclCXX.cpp                | 28 ++++++------
 clang/lib/Sema/SemaExprCXX.cpp                |  2 +-
 clang/lib/Sema/SemaExprMember.cpp             |  2 +-
 clang/lib/Sema/SemaLookup.cpp                 | 16 ++++++-
 clang/lib/Sema/SemaOpenMP.cpp                 |  2 +-
 clang/lib/Sema/SemaTemplate.cpp               |  5 ++-
 .../lib/Sema/SemaTemplateInstantiateDecl.cpp  |  8 ++--
 13 files changed, 118 insertions(+), 88 deletions(-)
 create mode 100644 clang/include/clang/Sema/Redeclaration.h

diff --git a/clang/include/clang/Sema/Lookup.h b/clang/include/clang/Sema/Lookup.h
index 2f2f2607a937f..0db5b847038ff 100644
--- a/clang/include/clang/Sema/Lookup.h
+++ b/clang/include/clang/Sema/Lookup.h
@@ -153,28 +153,30 @@ class LookupResult {
 
   using iterator = UnresolvedSetImpl::iterator;
 
-  LookupResult(Sema &SemaRef, const DeclarationNameInfo &NameInfo,
-               Sema::LookupNameKind LookupKind,
-               Sema::RedeclarationKind Redecl = Sema::NotForRedeclaration)
+  LookupResult(
+      Sema &SemaRef, const DeclarationNameInfo &NameInfo,
+      Sema::LookupNameKind LookupKind,
+      RedeclarationKind Redecl = RedeclarationKind::NotForRedeclaration)
       : SemaPtr(&SemaRef), NameInfo(NameInfo), LookupKind(LookupKind),
-        Redecl(Redecl != Sema::NotForRedeclaration),
-        ExternalRedecl(Redecl == Sema::ForExternalRedeclaration),
-        DiagnoseAccess(Redecl == Sema::NotForRedeclaration),
-        DiagnoseAmbiguous(Redecl == Sema::NotForRedeclaration) {
+        Redecl(Redecl != RedeclarationKind::NotForRedeclaration),
+        ExternalRedecl(Redecl == RedeclarationKind::ForExternalRedeclaration),
+        DiagnoseAccess(Redecl == RedeclarationKind::NotForRedeclaration),
+        DiagnoseAmbiguous(Redecl == RedeclarationKind::NotForRedeclaration) {
     configure();
   }
 
   // TODO: consider whether this constructor should be restricted to take
   // as input a const IdentifierInfo* (instead of Name),
   // forcing other cases towards the constructor taking a DNInfo.
-  LookupResult(Sema &SemaRef, DeclarationName Name, SourceLocation NameLoc,
-               Sema::LookupNameKind LookupKind,
-               Sema::RedeclarationKind Redecl = Sema::NotForRedeclaration)
+  LookupResult(
+      Sema &SemaRef, DeclarationName Name, SourceLocation NameLoc,
+      Sema::LookupNameKind LookupKind,
+      RedeclarationKind Redecl = RedeclarationKind::NotForRedeclaration)
       : SemaPtr(&SemaRef), NameInfo(Name, NameLoc), LookupKind(LookupKind),
-        Redecl(Redecl != Sema::NotForRedeclaration),
-        ExternalRedecl(Redecl == Sema::ForExternalRedeclaration),
-        DiagnoseAccess(Redecl == Sema::NotForRedeclaration),
-        DiagnoseAmbiguous(Redecl == Sema::NotForRedeclaration) {
+        Redecl(Redecl != RedeclarationKind::NotForRedeclaration),
+        ExternalRedecl(Redecl == RedeclarationKind::ForExternalRedeclaration),
+        DiagnoseAccess(Redecl == RedeclarationKind::NotForRedeclaration),
+        DiagnoseAmbiguous(Redecl == RedeclarationKind::NotForRedeclaration) {
     configure();
   }
 
@@ -285,9 +287,10 @@ class LookupResult {
     return ExternalRedecl;
   }
 
-  Sema::RedeclarationKind redeclarationKind() const {
-    return ExternalRedecl ? Sema::ForExternalRedeclaration :
-           Redecl ? Sema::ForVisibleRedeclaration : Sema::NotForRedeclaration;
+  RedeclarationKind redeclarationKind() const {
+    return ExternalRedecl ? RedeclarationKind::ForExternalRedeclaration
+           : Redecl       ? RedeclarationKind::ForVisibleRedeclaration
+                          : RedeclarationKind::NotForRedeclaration;
   }
 
   /// Specify whether hidden declarations are visible, e.g.,
@@ -615,9 +618,9 @@ class LookupResult {
   }
 
   /// Change this lookup's redeclaration kind.
-  void setRedeclarationKind(Sema::RedeclarationKind RK) {
-    Redecl = (RK != Sema::NotForRedeclaration);
-    ExternalRedecl = (RK == Sema::ForExternalRedeclaration);
+  void setRedeclarationKind(RedeclarationKind RK) {
+    Redecl = (RK != RedeclarationKind::NotForRedeclaration);
+    ExternalRedecl = (RK == RedeclarationKind::ForExternalRedeclaration);
     configure();
   }
 
diff --git a/clang/include/clang/Sema/Redeclaration.h b/clang/include/clang/Sema/Redeclaration.h
new file mode 100644
index 0000000000000..ae18b922f5cd9
--- /dev/null
+++ b/clang/include/clang/Sema/Redeclaration.h
@@ -0,0 +1,31 @@
+//===- Redeclaration.h - Redeclarations--------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+//  This file defines RedeclarationKind enum.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CLANG_SEMA_REDECLARATION_H
+#define LLVM_CLANG_SEMA_REDECLARATION_H
+
+/// Specifies whether (or how) name lookup is being performed for a
+/// redeclaration (vs. a reference).
+enum class RedeclarationKind {
+  /// The lookup is a reference to this name that is not for the
+  /// purpose of redeclaring the name.
+  NotForRedeclaration = 0,
+  /// The lookup results will be used for redeclaration of a name,
+  /// if an entity by that name already exists and is visible.
+  ForVisibleRedeclaration,
+  /// The lookup results will be used for redeclaration of a name
+  /// with external linkage; non-visible lookup results with external linkage
+  /// may also be found.
+  ForExternalRedeclaration
+};
+
+#endif // LLVM_CLANG_SEMA_REDECLARATION_H
\ No newline at end of file
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 281e3b91de1d0..1e89dfc58d92b 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -52,6 +52,7 @@
 #include "clang/Sema/IdentifierResolver.h"
 #include "clang/Sema/ObjCMethodList.h"
 #include "clang/Sema/Ownership.h"
+#include "clang/Sema/Redeclaration.h"
 #include "clang/Sema/Scope.h"
 #include "clang/Sema/SemaBase.h"
 #include "clang/Sema/SemaConcept.h"
@@ -7443,40 +7444,17 @@ class Sema final : public SemaBase {
   typedef std::function<ExprResult(Sema &, TypoExpr *, TypoCorrection)>
       TypoRecoveryCallback;
 
-  /// Specifies whether (or how) name lookup is being performed for a
-  /// redeclaration (vs. a reference).
-  enum RedeclarationKind {
-    /// The lookup is a reference to this name that is not for the
-    /// purpose of redeclaring the name.
-    NotForRedeclaration = 0,
-    /// The lookup results will be used for redeclaration of a name,
-    /// if an entity by that name already exists and is visible.
-    ForVisibleRedeclaration,
-    /// The lookup results will be used for redeclaration of a name
-    /// with external linkage; non-visible lookup results with external linkage
-    /// may also be found.
-    ForExternalRedeclaration
-  };
-
-  RedeclarationKind forRedeclarationInCurContext() const {
-    // A declaration with an owning module for linkage can never link against
-    // anything that is not visible. We don't need to check linkage here; if
-    // the context has internal linkage, redeclaration lookup won't find things
-    // from other TUs, and we can't safely compute linkage yet in general.
-    if (cast<Decl>(CurContext)
-            ->getOwningModuleForLinkage(/*IgnoreLinkage*/ true))
-      return ForVisibleRedeclaration;
-    return ForExternalRedeclaration;
-  }
+  RedeclarationKind forRedeclarationInCurContext() const;
 
   /// Look up a name, looking for a single declaration.  Return
   /// null if the results were absent, ambiguous, or overloaded.
   ///
   /// It is preferable to use the elaborated form and explicitly handle
   /// ambiguity and overloaded.
-  NamedDecl *LookupSingleName(Scope *S, DeclarationName Name,
-                              SourceLocation Loc, LookupNameKind NameKind,
-                              RedeclarationKind Redecl = NotForRedeclaration);
+  NamedDecl *LookupSingleName(
+      Scope *S, DeclarationName Name, SourceLocation Loc,
+      LookupNameKind NameKind,
+      RedeclarationKind Redecl = RedeclarationKind::NotForRedeclaration);
   bool LookupBuiltin(LookupResult &R);
   void LookupNecessaryTypesForBuiltin(Scope *S, unsigned ID);
   bool LookupName(LookupResult &R, Scope *S, bool AllowBuiltinCreation = false,
@@ -7488,9 +7466,9 @@ class Sema final : public SemaBase {
   bool LookupParsedName(LookupResult &R, Scope *S, CXXScopeSpec *SS,
                         bool AllowBuiltinCreation = false,
                         bool EnteringContext = false);
-  ObjCProtocolDecl *
-  LookupProtocol(IdentifierInfo *II, SourceLocation IdLoc,
-                 RedeclarationKind Redecl = NotForRedeclaration);
+  ObjCProtocolDecl *LookupProtocol(
+      IdentifierInfo *II, SourceLocation IdLoc,
+      RedeclarationKind Redecl = RedeclarationKind::NotForRedeclaration);
   bool LookupInSuper(LookupResult &R, CXXRecordDecl *Class);
 
   void LookupOverloadedOperatorName(OverloadedOperatorKind Op, Scope *S,
diff --git a/clang/lib/Interpreter/Interpreter.cpp b/clang/lib/Interpreter/Interpreter.cpp
index cf31456b6950a..b20e6efcebfd1 100644
--- a/clang/lib/Interpreter/Interpreter.cpp
+++ b/clang/lib/Interpreter/Interpreter.cpp
@@ -550,7 +550,8 @@ std::unique_ptr<RuntimeInterfaceBuilder> Interpreter::FindRuntimeInterface() {
 
   auto LookupInterface = [&](Expr *&Interface, llvm::StringRef Name) {
     LookupResult R(S, &Ctx.Idents.get(Name), SourceLocation(),
-                   Sema::LookupOrdinaryName, Sema::ForVisibleRedeclaration);
+                   Sema::LookupOrdinaryName,
+                   RedeclarationKind::ForVisibleRedeclaration);
     S.LookupQualifiedName(R, Ctx.getTranslationUnitDecl());
     if (R.empty())
       return false;
diff --git a/clang/lib/Interpreter/InterpreterUtils.cpp b/clang/lib/Interpreter/InterpreterUtils.cpp
index c19cf6aa3156c..45f6322b8461e 100644
--- a/clang/lib/Interpreter/InterpreterUtils.cpp
+++ b/clang/lib/Interpreter/InterpreterUtils.cpp
@@ -72,7 +72,7 @@ NamedDecl *LookupNamed(Sema &S, llvm::StringRef Name,
                        const DeclContext *Within) {
   DeclarationName DName = &S.Context.Idents.get(Name);
   LookupResult R(S, DName, SourceLocation(), Sema::LookupOrdinaryName,
-                 Sema::ForVisibleRedeclaration);
+                 RedeclarationKind::ForVisibleRedeclaration);
 
   R.suppressDiagnostics();
 
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 19abd5327b73a..455ccb45b4068 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -5374,7 +5374,7 @@ static bool CheckAnonMemberRedeclaration(Sema &SemaRef, Scope *S,
   LookupResult R(SemaRef, Name, NameLoc,
                  Owner->isRecord() ? Sema::LookupMemberName
                                    : Sema::LookupOrdinaryName,
-                 Sema::ForVisibleRedeclaration);
+                 RedeclarationKind::ForVisibleRedeclaration);
   if (!SemaRef.LookupName(R, S)) return false;
 
   // Pick a representative declaration.
@@ -6470,7 +6470,8 @@ NamedDecl *Sema::HandleDeclarator(Scope *S, Declarator &D,
 
     if (IsLinkageLookup) {
       Previous.clear(LookupRedeclarationWithLinkage);
-      Previous.setRedeclarationKind(ForExternalRedeclaration);
+      Previous.setRedeclarationKind(
+          RedeclarationKind::ForExternalRedeclaration);
     }
 
     LookupName(Previous, S, CreateBuiltins);
@@ -8521,7 +8522,8 @@ void Sema::CheckShadow(Scope *S, VarDecl *D) {
     return;
 
   LookupResult R(*this, D->getDeclName(), D->getLocation(),
-                 Sema::LookupOrdinaryName, Sema::ForVisibleRedeclaration);
+                 Sema::LookupOrdinaryName,
+                 RedeclarationKind::ForVisibleRedeclaration);
   LookupName(R, S);
   if (NamedDecl *ShadowedDecl = getShadowedDeclaration(D, R))
     CheckShadow(D, ShadowedDecl, R);
@@ -9161,7 +9163,7 @@ static NamedDecl *DiagnoseInvalidRedeclaration(
   LookupResult Prev(SemaRef, Name, NewFD->getLocation(),
                     IsLocalFriend ? Sema::LookupLocalFriendName
                                   : Sema::LookupOrdinaryName,
-                    Sema::ForVisibleRedeclaration);
+                    RedeclarationKind::ForVisibleRedeclaration);
 
   NewFD->setInvalidDecl();
   if (IsLocalFriend)
@@ -15196,7 +15198,7 @@ Decl *Sema::ActOnParamDeclarator(Scope *S, Declarator &D,
   const IdentifierInfo *II = D.getIdentifier();
   if (II) {
     LookupResult R(*this, II, D.getIdentifierLoc(), LookupOrdinaryName,
-                   ForVisibleRedeclaration);
+                   RedeclarationKind::ForVisibleRedeclaration);
     LookupName(R, S);
     if (!R.empty()) {
       NamedDecl *PrevDecl = *R.begin();
@@ -17428,7 +17430,7 @@ Sema::ActOnTag(Scope *S, unsigned TagSpec, TagUseKind TUK, SourceLocation KWLoc,
 
   RedeclarationKind Redecl = forRedeclarationInCurContext();
   if (TUK == TUK_Friend || TUK == TUK_Reference)
-    Redecl = NotForRedeclaration;
+    Redecl = RedeclarationKind::NotForRedeclaration;
 
   /// Create a new tag decl in C/ObjC. Since the ODR-like semantics for ObjC/C
   /// implemented asks for structural equivalence checking, the returned decl
@@ -18589,7 +18591,7 @@ FieldDecl *Sema::HandleField(Scope *S, RecordDecl *Record,
   // Check to see if this name was declared as a member previously
   NamedDecl *PrevDecl = nullptr;
   LookupResult Previous(*this, II, Loc, LookupMemberName,
-                        ForVisibleRedeclaration);
+                        RedeclarationKind::ForVisibleRedeclaration);
   LookupName(Previous, S);
   switch (Previous.getResultKind()) {
     case LookupResult::Found:
@@ -18993,8 +18995,9 @@ Decl *Sema::ActOnIvar(Scope *S, SourceLocation DeclStart, Declarator &D,
     NewID->setInvalidDecl();
 
   if (II) {
-    NamedDecl *PrevDecl = LookupSingleName(S, II, Loc, LookupMemberName,
-                                           ForVisibleRedeclaration);
+    NamedDecl *PrevDecl =
+        LookupSingleName(S, II, Loc, LookupMemberName,
+                         RedeclarationKind::ForVisibleRedeclaration);
     if (PrevDecl && isDeclInScope(PrevDecl, EnclosingContext, S)
         && !isa<TagDecl>(PrevDecl)) {
       Diag(Loc, diag::err_duplicate_member) << II;
@@ -20039,7 +20042,8 @@ Decl *Sema::ActOnEnumConstant(Scope *S, Decl *theEnumDecl, Decl *lastEnumConst,
 
   // Verify that there isn't already something declared with this name in this
   // scope.
-  LookupResult R(*this, Id, IdLoc, LookupOrdinaryName, ForVisibleRedeclaration);
+  LookupResult R(*this, Id, IdLoc, LookupOrdinaryName,
+                 RedeclarationKind::ForVisibleRedeclaration);
   LookupName(R, S);
   NamedDecl *PrevDecl = R.getAsSingle<NamedDecl>();
 
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 8c6bae545bfd1..591016243b0ac 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -896,7 +896,7 @@ Sema::ActOnDecompositionDeclarator(Scope *S, Declarator &D,
     assert(VarName && "Cannot have an unnamed binding declaration");
 
     LookupResult Previous(*this, NameInfo, LookupOrdinaryName,
-                          ForVisibleRedeclaration);
+                          RedeclarationKind::ForVisibleRedeclaration);
     LookupName(Previous, S,
                /*CreateBuiltins*/DC->getRedeclContext()->isTranslationUnit());
 
@@ -951,7 +951,7 @@ Sema::ActOnDecompositionDeclarator(Scope *S, Declarator &D,
   DeclarationNameInfo NameInfo((IdentifierInfo *)nullptr,
                                Decomp.getLSquareLoc());
   LookupResult Previous(*this, NameInfo, LookupOrdinaryName,
-                        ForVisibleRedeclaration);
+                        RedeclarationKind::ForVisibleRedeclaration);
 
   // Build the variable that holds the non-decomposed object.
   bool AddToScope = true;
@@ -11715,7 +11715,7 @@ Decl *Sema::ActOnStartNamespaceDef(Scope *NamespcScope,
     // look through using directives, just look for any ordinary names
     // as if by qualified name lookup.
     LookupResult R(*this, II, IdentLoc, LookupOrdinaryName,
-                   ForExternalRedeclaration);
+                   RedeclarationKind::ForExternalRedeclaration);
     LookupQualifiedName(R, CurContext->getRedeclContext());
     NamedDecl *PrevDecl =
         R.isSingleResult() ? R.getRepresentativeDecl() : nullptr;
@@ -12916,7 +12916,7 @@ NamedDecl *Sema::BuildUsingDeclaration(
 
   // Do the redeclaration lookup in the current scope.
   LookupResult Previous(*this, UsingName, LookupUsingDeclName,
-                        ForVisibleRedeclaration);
+                        RedeclarationKind::ForVisibleRedeclaration);
   Previous.setHideTags(false);
   if (S) {
     LookupName(Previous, S);
@@ -13159,7 +13159,7 @@ NamedDecl *Sema::BuildUsingEnumDeclaration(Scope *S, AccessSpecifier AS,
     /// In class scope, check if this is a duplicate, for better a diagnostic.
     DeclarationNameInfo UsingEnumName(ED->getDeclName(), NameLoc);
     LookupResult Previous(*this, UsingEnumName, LookupUsingDeclName,
-                          ForVisibleRedeclaration);
+                          RedeclarationKind::ForVisibleRedeclaration);
 
     LookupName(Previous, S);
 
@@ -13192,7 +13192,7 @@ NamedDecl *Sema::BuildUsingEnumDeclaration(Scope *S, AccessSpecifier AS,
     UsingShadowDecl *PrevDecl = nullptr;
     DeclarationNameInfo DNI(EC->getDeclName(), EC->getLocation());
     LookupResult Previous(*this, DNI, LookupOrdinaryName,
-                          ForVisibleRedeclaration);
+                          RedeclarationKind::ForVisibleRedeclaration);
     LookupName(Previous, S);
     FilterUsingLookup(S, Previous);
 
@@ -13587,7 +13587,7 @@ Decl *Sema::ActOnAliasDeclaration(Scope *S, AccessSpecifier AS,
   LookupResult Previous(*this, NameInfo, LookupOrdinaryName,
                         TemplateParamLists.size()
                             ? forRedeclarationInCurContext()
-                            : ForVisibleRedeclaration);
+                            : RedeclarationKind::ForVisibleRedeclaration);
   LookupName(Previous, S);
 
   // Warn about shadowing the name of a template parameter.
@@ -13737,7 +13737,7 @@ Decl *Sema::ActOnNamespaceAliasDef(Scope *S, SourceLocation NamespaceLoc,
 
   // Check if we have a previous declaration with the same name.
   LookupResult PrevR(*this, Alias, AliasLoc, LookupOrdinaryName,
-                     ForVisibleRedeclaration);
+                     RedeclarationKind::ForVisibleRedeclaration);
   LookupName(PrevR, S);
 
   // Check we're not shadowing a template parameter.
@@ -13983,7 +13983,7 @@ void Sema::CheckImplicitSpecialMemberDeclaration(Scope *S, FunctionDecl *FD) {
   // implicit special members with this name.
   DeclarationName Name = FD->getDeclName();
   LookupResult R(*this, Name, SourceLocation(), LookupOrdinaryName,
-                 ForExternalRedeclaration);
+                 RedeclarationKind::ForExternalRedeclaration);
   for (auto *D : FD->getParent()->lookup(Name))
     if (auto *Acceptable = R.getAcceptableDecl(D))
       R.addDecl(Acceptable);
@@ -17113,9 +17113,9 @@ Decl *Sema::ActOnExceptionDeclarator(Scope *S, Declarator &D) {
   }
 
   const IdentifierInfo *II = D.getIdentifier();
-  if (NamedDecl *PrevDecl = LookupSingleName(S, II, D.getIdentifierLoc(),
-                                             LookupOrdinaryName,
-                                             ForVisibleRedeclaration)) {
+  if (NamedDecl *PrevDecl =
+          LookupSingleName(S, II, D.getIdentifierLoc(), LookupOrdinaryName,
+                           RedeclarationKind::ForVisibleRedeclaration)) {
     // The scope should be freshly made just for us. There is just no way
     // it contains any previous declaration, except for function parameters in
     // a function-try-block's catch statement.
@@ -17906,7 +17906,7 @@ NamedDecl *Sema::ActOnFriendFunctionDecl(Scope *S, Declarator &D,
   DeclContext *DC;
   Scope *DCScope = S;
   LookupResult Previous(*this, NameInfo, LookupOrdinaryName,
-                        ForExternalRedeclaration);
+                        RedeclarationKind::ForExternalRedeclaration);
 
   bool isTemplateId = D.getName().getKind() == UnqualifiedIdKind::IK_TemplateId;
 
@@ -19242,7 +19242,7 @@ MSPropertyDecl *Sema::HandleMSProperty(Scope *S, RecordDecl *Record,
   // Check to see if this name was declared as a member previously
   NamedDecl *PrevDecl = nullptr;
   LookupResult Previous(*this, II, Loc, LookupMemberName,
-                        ForVisibleRedeclaration);
+                        RedeclarationKind::ForVisibleRedeclaration);
   LookupName(Previous, S);
   switch (Previous.getResultKind()) {
   case LookupResult::Found:
diff --git a/clang/lib/Sema/SemaExprCXX.cpp b/clang/lib/Sema/SemaExprCXX.cpp
index f4a91ececfbb5..7582cbd75fec0 100644
--- a/clang/lib/Sema/SemaExprCXX.cpp
+++ b/clang/lib/Sema/SemaExprCXX.cpp
@@ -9153,7 +9153,7 @@ Sema::CheckMicrosoftIfExistsSymbol(Scope *S,
 
   // Do the redeclaration lookup in the current scope.
   LookupResult R(*this, TargetNameInfo, Sema::LookupAnyName,
-                 Sema::NotForRedeclaration);
+                 RedeclarationKind::NotForRedeclaration);
   LookupParsedName(R, S, &SS);
   R.suppressDiagnostics();
 
diff --git a/clang/lib/Sema/SemaExprMember.cpp b/clang/lib/Sema/SemaExprMember.cpp
index 7ea6d733fe5a2..c79128bc8f39e 100644
--- a/clang/lib/Sema/SemaExprMember.cpp
+++ b/clang/lib/Sema/SemaExprMember.cpp
@@ -728,7 +728,7 @@ static bool LookupMemberExprInRecord(Sema &SemaRef, LookupResult &R,
     Sema &SemaRef;
     DeclarationNameInfo NameInfo;
     Sema::LookupNameKind LookupKind;
-    Sema::RedeclarationKind Redecl;
+    RedeclarationKind Redecl;
   };
   QueryState Q = {R.getSema(), R.getLookupNameInfo(), R.getLookupKind(),
                   R.redeclarationKind()};
diff --git a/clang/lib/Sema/SemaLookup.cpp b/clang/lib/Sema/SemaLookup.cpp
index d65f52b8efe81..55af414df39f5 100644
--- a/clang/lib/Sema/SemaLookup.cpp
+++ b/clang/lib/Sema/SemaLookup.cpp
@@ -4449,7 +4449,8 @@ LabelDecl *Sema::LookupOrCreateLabel(IdentifierInfo *II, SourceLocation Loc,
   }
 
   // Not a GNU local label.
-  Res = LookupSingleName(CurScope, II, Loc, LookupLabel, NotForRedeclaration);
+  Res = LookupSingleName(CurScope, II, Loc, LookupLabel,
+                         RedeclarationKind::NotForRedeclaration);
   // If we found a label, check to see if it is in the same context as us.
   // When in a Block, we don't want to reuse a label in an enclosing function.
   if (Res && Res->getDeclContext() != CurContext)
@@ -5889,7 +5890,8 @@ void Sema::clearDelayedTypo(TypoExpr *TE) {
 
 void Sema::ActOnPragmaDump(Scope *S, SourceLocation IILoc, IdentifierInfo *II) {
   DeclarationNameInfo Name(II, IILoc);
-  LookupResult R(*this, Name, LookupAnyName, Sema::NotForRedeclaration);
+  LookupResult R(*this, Name, LookupAnyName,
+                 RedeclarationKind::NotForRedeclaration);
   R.suppressDiagnostics();
   R.setHideTags(false);
   LookupName(R, S);
@@ -5899,3 +5901,13 @@ void Sema::ActOnPragmaDump(Scope *S, SourceLocation IILoc, IdentifierInfo *II) {
 void Sema::ActOnPragmaDump(Expr *E) {
   E->dump();
 }
+
+RedeclarationKind Sema::forRedeclarationInCurContext() const {
+  // A declaration with an owning module for linkage can never link against
+  // anything that is not visible. We don't need to check linkage here; if
+  // the context has internal linkage, redeclaration lookup won't find things
+  // from other TUs, and we can't safely compute linkage yet in general.
+  if (cast<Decl>(CurContext)->getOwningModuleForLinkage(/*IgnoreLinkage*/ true))
+    return RedeclarationKind::ForVisibleRedeclaration;
+  return RedeclarationKind::ForExternalRedeclaration;
+}
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index d229ef650bccb..3e9f6cba25076 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -24944,7 +24944,7 @@ ExprResult SemaOpenMP::ActOnOMPIteratorExpr(Scope *S,
       // Check for conflicting previous declaration.
       DeclarationNameInfo NameInfo(VD->getDeclName(), D.DeclIdentLoc);
       LookupResult Previous(SemaRef, NameInfo, Sema::LookupOrdinaryName,
-                            Sema::ForVisibleRedeclaration);
+                            RedeclarationKind::ForVisibleRedeclaration);
       Previous.suppressDiagnostics();
       SemaRef.LookupName(Previous, S);
 
diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
index 95171359f0ab1..f4b6e1ceb6f02 100644
--- a/clang/lib/Sema/SemaTemplate.cpp
+++ b/clang/lib/Sema/SemaTemplate.cpp
@@ -972,8 +972,9 @@ void Sema::translateTemplateArguments(const ASTTemplateArgsPtr &TemplateArgsIn,
 static void maybeDiagnoseTemplateParameterShadow(Sema &SemaRef, Scope *S,
                                                  SourceLocation Loc,
                                                  const IdentifierInfo *Name) {
-  NamedDecl *PrevDecl = SemaRef.LookupSingleName(
-      S, Name, Loc, Sema::LookupOrdinaryName, Sema::ForVisibleRedeclaration);
+  NamedDecl *PrevDecl =
+      SemaRef.LookupSingleName(S, Name, Loc, Sema::LookupOrdinaryName,
+                               RedeclarationKind::ForVisibleRedeclaration);
   if (PrevDecl && PrevDecl->isTemplateParameter())
     SemaRef.DiagnoseTemplateParameterShadow(Loc, PrevDecl);
 }
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index 6d359c5a9a024..caa07abb61fe3 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -2296,7 +2296,7 @@ Decl *TemplateDeclInstantiator::VisitFunctionDecl(
       SemaRef, Function->getDeclName(), SourceLocation(),
       D->isLocalExternDecl() ? Sema::LookupRedeclarationWithLinkage
                              : Sema::LookupOrdinaryName,
-      D->isLocalExternDecl() ? Sema::ForExternalRedeclaration
+      D->isLocalExternDecl() ? RedeclarationKind::ForExternalRedeclaration
                              : SemaRef.forRedeclarationInCurContext());
 
   if (DependentFunctionTemplateSpecializationInfo *DFTSI =
@@ -2697,7 +2697,7 @@ Decl *TemplateDeclInstantiator::VisitCXXMethodDecl(
     Method->setInvalidDecl();
 
   LookupResult Previous(SemaRef, NameInfo, Sema::LookupOrdinaryName,
-                        Sema::ForExternalRedeclaration);
+                        RedeclarationKind::ForExternalRedeclaration);
 
   bool IsExplicitSpecialization = false;
 
@@ -3365,7 +3365,7 @@ Decl *TemplateDeclInstantiator::VisitUsingDecl(UsingDecl *D) {
   // fact, it's not really even possible in non-class scopes).
   bool CheckRedeclaration = Owner->isRecord();
   LookupResult Prev(SemaRef, NameInfo, Sema::LookupUsingDeclName,
-                    Sema::ForVisibleRedeclaration);
+                    RedeclarationKind::ForVisibleRedeclaration);
 
   UsingDecl *NewUD = UsingDecl::Create(SemaRef.Context, Owner,
                                        D->getUsingLoc(),
@@ -5388,7 +5388,7 @@ void Sema::BuildVariableInstantiation(
       *this, NewVar->getDeclName(), NewVar->getLocation(),
       NewVar->isLocalExternDecl() ? Sema::LookupRedeclarationWithLinkage
                                   : Sema::LookupOrdinaryName,
-      NewVar->isLocalExternDecl() ? Sema::ForExternalRedeclaration
+      NewVar->isLocalExternDecl() ? RedeclarationKind::ForExternalRedeclaration
                                   : forRedeclarationInCurContext());
 
   if (NewVar->isLocalExternDecl() && OldVar->getPreviousDecl() &&

From 950bb097e11d6ee26533c00519c62df994322228 Mon Sep 17 00:00:00 2001
From: Dinar Temirbulatov <Dinar.Temirbulatov@arm.com>
Date: Wed, 17 Apr 2024 15:30:40 +0000
Subject: [PATCH 756/886] Revert "[Clang][AArch64] Warn when calling
 non/streaming about vector size difference (#79842)"

This reverts commit 4e85e1ffcaf161736e27a24c291c1177be865976
---
 clang/include/clang/Basic/DiagnosticGroups.td |   3 -
 .../clang/Basic/DiagnosticSemaKinds.td        |  10 --
 clang/lib/Sema/SemaChecking.cpp               |  22 +--
 clang/lib/Sema/SemaDecl.cpp                   |  12 +-
 .../Sema/aarch64-incompat-sm-builtin-calls.c  |   6 +-
 clang/test/Sema/aarch64-sme-func-attrs.c      | 136 +-----------------
 6 files changed, 5 insertions(+), 184 deletions(-)

diff --git a/clang/include/clang/Basic/DiagnosticGroups.td b/clang/include/clang/Basic/DiagnosticGroups.td
index 47747d8704b6c..5251774ff4efd 100644
--- a/clang/include/clang/Basic/DiagnosticGroups.td
+++ b/clang/include/clang/Basic/DiagnosticGroups.td
@@ -1412,9 +1412,6 @@ def MultiGPU: DiagGroup<"multi-gpu">;
 // libc and the CRT to be skipped.
 def AVRRtlibLinkingQuirks : DiagGroup<"avr-rtlib-linking-quirks">;
 
-// A warning group related to AArch64 SME function attribues.
-def AArch64SMEAttributes : DiagGroup<"aarch64-sme-attributes">;
-
 // A warning group for things that will change semantics in the future.
 def FutureCompat : DiagGroup<"future-compat">;
 
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 44f802c0c28e8..30a8543489f48 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -3751,16 +3751,6 @@ def err_sme_definition_using_za_in_non_sme_target : Error<
   "function using ZA state requires 'sme'">;
 def err_sme_definition_using_zt0_in_non_sme2_target : Error<
   "function using ZT0 state requires 'sme2'">;
-def warn_sme_streaming_pass_return_vl_to_non_streaming : Warning<
-  "passing a VL-dependent argument to/from a function that has a different"
-  " streaming-mode. The streaming and non-streaming vector lengths may be"
-  " different">,
-  InGroup<AArch64SMEAttributes>, DefaultIgnore;
-def warn_sme_locally_streaming_has_vl_args_returns : Warning<
-  "passing/returning a VL-dependent argument to/from a __arm_locally_streaming"
-  " function. The streaming and non-streaming vector"
-  " lengths may be different">,
-  InGroup<AArch64SMEAttributes>, DefaultIgnore;
 def err_conflicting_attributes_arm_state : Error<
   "conflicting attributes for state '%0'">;
 def err_sme_streaming_cannot_be_multiversioned : Error<
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index 99b0a00083535..d814327d23470 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -7949,7 +7949,6 @@ void Sema::checkCall(NamedDecl *FDecl, const FunctionProtoType *Proto,
     // For variadic functions, we may have more args than parameters.
     // For some K&R functions, we may have less args than parameters.
     const auto N = std::min<unsigned>(Proto->getNumParams(), Args.size());
-    bool AnyScalableArgsOrRet = Proto->getReturnType()->isSizelessVectorType();
     for (unsigned ArgIdx = 0; ArgIdx < N; ++ArgIdx) {
       // Args[ArgIdx] can be null in malformed code.
       if (const Expr *Arg = Args[ArgIdx]) {
@@ -7963,8 +7962,6 @@ void Sema::checkCall(NamedDecl *FDecl, const FunctionProtoType *Proto,
           checkAIXMemberAlignment((Arg->getExprLoc()), Arg);
 
         QualType ParamTy = Proto->getParamType(ArgIdx);
-        if (ParamTy->isSizelessVectorType())
-          AnyScalableArgsOrRet = true;
         QualType ArgTy = Arg->getType();
         CheckArgAlignment(Arg->getExprLoc(), FDecl, std::to_string(ArgIdx + 1),
                           ArgTy, ParamTy);
@@ -7985,23 +7982,6 @@ void Sema::checkCall(NamedDecl *FDecl, const FunctionProtoType *Proto,
       }
     }
 
-    // If the call requires a streaming-mode change and has scalable vector
-    // arguments or return values, then warn the user that the streaming and
-    // non-streaming vector lengths may be different.
-    const auto *CallerFD = dyn_cast<FunctionDecl>(CurContext);
-    if (CallerFD && (!FD || !FD->getBuiltinID()) && AnyScalableArgsOrRet) {
-      bool IsCalleeStreaming =
-          ExtInfo.AArch64SMEAttributes & FunctionType::SME_PStateSMEnabledMask;
-      bool IsCalleeStreamingCompatible =
-          ExtInfo.AArch64SMEAttributes &
-          FunctionType::SME_PStateSMCompatibleMask;
-      ArmStreamingType CallerFnType = getArmStreamingFnType(CallerFD);
-      if (!IsCalleeStreamingCompatible &&
-          (CallerFnType == ArmStreamingCompatible ||
-           ((CallerFnType == ArmStreaming) ^ IsCalleeStreaming)))
-        Diag(Loc, diag::warn_sme_streaming_pass_return_vl_to_non_streaming);
-    }
-
     FunctionType::ArmStateValue CalleeArmZAState =
         FunctionType::getArmZAState(ExtInfo.AArch64SMEAttributes);
     FunctionType::ArmStateValue CalleeArmZT0State =
@@ -8010,7 +7990,7 @@ void Sema::checkCall(NamedDecl *FDecl, const FunctionProtoType *Proto,
         CalleeArmZT0State != FunctionType::ARM_None) {
       bool CallerHasZAState = false;
       bool CallerHasZT0State = false;
-      if (CallerFD) {
+      if (const auto *CallerFD = dyn_cast<FunctionDecl>(CurContext)) {
         auto *Attr = CallerFD->getAttr<ArmNewAttr>();
         if (Attr && Attr->isNewZA())
           CallerHasZAState = true;
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 455ccb45b4068..1bde99d6fce74 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -12408,22 +12408,12 @@ bool Sema::CheckFunctionDeclaration(Scope *S, FunctionDecl *NewFD,
   }
 
   // Check if the function definition uses any AArch64 SME features without
-  // having the '+sme' feature enabled and warn user if sme locally streaming
-  // function returns or uses arguments with VL-based types.
+  // having the '+sme' feature enabled.
   if (DeclIsDefn) {
     const auto *Attr = NewFD->getAttr<ArmNewAttr>();
     bool UsesSM = NewFD->hasAttr<ArmLocallyStreamingAttr>();
     bool UsesZA = Attr && Attr->isNewZA();
     bool UsesZT0 = Attr && Attr->isNewZT0();
-
-    if (NewFD->hasAttr<ArmLocallyStreamingAttr>()) {
-      if (NewFD->getReturnType()->isSizelessVectorType() ||
-          llvm::any_of(NewFD->parameters(), [](ParmVarDecl *P) {
-            return P->getOriginalType()->isSizelessVectorType();
-          }))
-        Diag(NewFD->getLocation(),
-             diag::warn_sme_locally_streaming_has_vl_args_returns);
-    }
     if (const auto *FPT = NewFD->getType()->getAs<FunctionProtoType>()) {
       FunctionProtoType::ExtProtoInfo EPI = FPT->getExtProtoInfo();
       UsesSM |=
diff --git a/clang/test/Sema/aarch64-incompat-sm-builtin-calls.c b/clang/test/Sema/aarch64-incompat-sm-builtin-calls.c
index 6a1feeb9bf539..55c97c73e8b69 100644
--- a/clang/test/Sema/aarch64-incompat-sm-builtin-calls.c
+++ b/clang/test/Sema/aarch64-incompat-sm-builtin-calls.c
@@ -1,6 +1,6 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1  -triple aarch64-none-linux-gnu -target-feature +sve \
-// RUN:   -target-feature +sme2 -target-feature +sve2 -target-feature +neon -Waarch64-sme-attributes -fsyntax-only -verify %s
+// RUN:   -target-feature +sme2 -target-feature +sve2 -target-feature +neon -fsyntax-only -verify %s
 
 // REQUIRES: aarch64-registered-target
 
@@ -33,7 +33,6 @@ svuint32_t incompat_sve_sm(svbool_t pg, svuint32_t a, int16_t b) __arm_streaming
   return __builtin_sve_svld1_gather_u32base_index_u32(pg, a, b);
 }
 
-// expected-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
 __arm_locally_streaming svuint32_t incompat_sve_ls(svbool_t pg, svuint32_t a, int64_t b) {
   // expected-warning@+1 {{builtin call has undefined behaviour when called from a streaming function}}
   return __builtin_sve_svld1_gather_u32base_index_u32(pg, a, b);
@@ -49,7 +48,6 @@ svuint32_t incompat_sve2_sm(svbool_t pg, svuint32_t a, int64_t b) __arm_streamin
   return __builtin_sve_svldnt1_gather_u32base_index_u32(pg, a, b);
 }
 
-// expected-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
 __arm_locally_streaming svuint32_t incompat_sve2_ls(svbool_t pg, svuint32_t a, int64_t b) {
   // expected-warning@+1 {{builtin call has undefined behaviour when called from a streaming function}}
   return __builtin_sve_svldnt1_gather_u32base_index_u32(pg, a, b);
@@ -70,7 +68,6 @@ svfloat64_t streaming_caller_sve(svbool_t pg, svfloat64_t a, float64_t b) __arm_
   return svadd_n_f64_m(pg, a, b);
 }
 
-// expected-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
 __arm_locally_streaming svfloat64_t locally_streaming_caller_sve(svbool_t pg, svfloat64_t a, float64_t b) {
   // expected-no-warning
   return svadd_n_f64_m(pg, a, b);
@@ -86,7 +83,6 @@ svint16_t streaming_caller_sve2(svint16_t op1, svint16_t op2) __arm_streaming {
   return svmul_lane_s16(op1, op2, 0);
 }
 
-// expected-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
 __arm_locally_streaming svint16_t locally_streaming_caller_sve2(svint16_t op1, svint16_t op2) {
   // expected-no-warning
   return svmul_lane_s16(op1, op2, 0);
diff --git a/clang/test/Sema/aarch64-sme-func-attrs.c b/clang/test/Sema/aarch64-sme-func-attrs.c
index 12de16509ccb8..bfc8768c3f36e 100644
--- a/clang/test/Sema/aarch64-sme-func-attrs.c
+++ b/clang/test/Sema/aarch64-sme-func-attrs.c
@@ -1,5 +1,5 @@
-// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme2 -target-feature +sve -Waarch64-sme-attributes -fsyntax-only -verify %s
-// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme2 -target-feature +sve -Waarch64-sme-attributes -fsyntax-only -verify=expected-cpp -x c++ %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme2 -fsyntax-only -verify %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme2 -fsyntax-only -verify=expected-cpp -x c++ %s
 
 // Valid attributes
 
@@ -496,135 +496,3 @@ void fmv_caller() {
     just_fine();
     incompatible_locally_streaming();
 }
-
-void sme_streaming_with_vl_arg(__SVInt8_t a) __arm_streaming { }
-
-__SVInt8_t sme_streaming_returns_vl(void) __arm_streaming { __SVInt8_t r; return r; }
-
-void sme_streaming_compatible_with_vl_arg(__SVInt8_t a) __arm_streaming_compatible { }
-
-__SVInt8_t sme_streaming_compatible_returns_vl(void) __arm_streaming_compatible { __SVInt8_t r; return r; }
-
-void sme_no_streaming_with_vl_arg(__SVInt8_t a) { }
-
-__SVInt8_t sme_no_streaming_returns_vl(void) { __SVInt8_t r; return r; }
-
-// expected-warning@+2 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
-// expected-cpp-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
-__arm_locally_streaming void sme_locally_streaming_with_vl_arg(__SVInt8_t a) { }
-
-// expected-warning@+2 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
-// expected-cpp-warning@+1 {{passing/returning a VL-dependent argument to/from a __arm_locally_streaming function. The streaming and non-streaming vector lengths may be different}}
-__arm_locally_streaming __SVInt8_t sme_locally_streaming_returns_vl(void) { __SVInt8_t r; return r; }
-
-void sme_no_streaming_calling_streaming_with_vl_args() {
-  __SVInt8_t a;
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  sme_streaming_with_vl_arg(a);
-}
-
-void sme_no_streaming_calling_streaming_with_return_vl() {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  __SVInt8_t r = sme_streaming_returns_vl();
-}
-
-void sme_streaming_calling_non_streaming_with_vl_args(void) __arm_streaming {
-  __SVInt8_t a;
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  sme_no_streaming_with_vl_arg(a);
-}
-
-void sme_streaming_calling_non_streaming_with_return_vl(void) __arm_streaming {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  __SVInt8_t r = sme_no_streaming_returns_vl();
-}
-
-void sme_no_streaming_calling_streaming_with_vl_args_param(__SVInt8_t arg, void (*sc)( __SVInt8_t arg) __arm_streaming) {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  sc(arg);
-}
-
-__SVInt8_t sme_no_streaming_calling_streaming_return_vl_param(__SVInt8_t (*s)(void) __arm_streaming) {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  return s();
-}
-
-void sme_streaming_compatible_calling_streaming_with_vl_args(__SVInt8_t arg) __arm_streaming_compatible {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  sme_streaming_with_vl_arg(arg);
-}
-
-void sme_streaming_compatible_calling_sme_streaming_return_vl(void) __arm_streaming_compatible {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  __SVInt8_t r = sme_streaming_returns_vl();
-}
-
-void sme_streaming_compatible_calling_no_streaming_with_vl_args(__SVInt8_t arg) __arm_streaming_compatible {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  sme_no_streaming_with_vl_arg(arg);
-}
-
-void sme_streaming_compatible_calling_no_sme_streaming_return_vl(void) __arm_streaming_compatible {
-  // expected-warning@+2 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  // expected-cpp-warning@+1 {{passing a VL-dependent argument to/from a function that has a different streaming-mode. The streaming and non-streaming vector lengths may be different}}
-  __SVInt8_t r = sme_no_streaming_returns_vl();
-}
-
-void sme_streaming_calling_streaming(__SVInt8_t arg, void (*s)( __SVInt8_t arg) __arm_streaming) __arm_streaming {
-  s(arg);
-}
-
-__SVInt8_t sme_streaming_calling_streaming_return_vl(__SVInt8_t (*s)(void) __arm_streaming) __arm_streaming {
-  return s();
-}
-
-void sme_streaming_calling_streaming_with_vl_args(__SVInt8_t a) __arm_streaming {
-  sme_streaming_with_vl_arg(a);
-}
-
-void sme_streaming_calling_streaming_with_return_vl(void) __arm_streaming {
-  __SVInt8_t r = sme_streaming_returns_vl();
-}
-
-void sme_streaming_calling_streaming_compatible_with_vl_args(__SVInt8_t a) __arm_streaming {
-  sme_streaming_compatible_with_vl_arg(a);
-}
-
-void sme_streaming_calling_streaming_compatible_with_return_vl(void) __arm_streaming {
-  __SVInt8_t r = sme_streaming_compatible_returns_vl();
-}
-
-void sme_no_streaming_calling_streaming_compatible_with_vl_args() {
-  __SVInt8_t a;
-  sme_streaming_compatible_with_vl_arg(a);
-}
-
-void sme_no_streaming_calling_streaming_compatible_with_return_vl() {
-  __SVInt8_t r = sme_streaming_compatible_returns_vl();
-}
-
-void sme_no_streaming_calling_non_streaming_compatible_with_vl_args() {
-  __SVInt8_t a;
-  sme_no_streaming_with_vl_arg(a);
-}
-
-void sme_no_streaming_calling_non_streaming_compatible_with_return_vl() {
-  __SVInt8_t r = sme_no_streaming_returns_vl();
-}
-
-void sme_streaming_compatible_calling_streaming_compatible_with_vl_args(__SVInt8_t arg) __arm_streaming_compatible {
-  sme_streaming_compatible_with_vl_arg(arg);
-}
-
-void sme_streaming_compatible_calling_streaming_compatible_with_return_vl(void) __arm_streaming_compatible {
-  __SVInt8_t r = sme_streaming_compatible_returns_vl();
-}

From b854a2323337be2633b1135f590678a17e9d1ade Mon Sep 17 00:00:00 2001
From: Robin Caloudis <robin.caloudis@gmx.de>
Date: Wed, 17 Apr 2024 17:38:47 +0200
Subject: [PATCH 757/886] [libc][c23][fenv] Implement fetestexceptflag (#87828)

Provide C23 `fetestexceptflag` function according to 7.6.4.6 in the
latest [revision of the C
standard](https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3096.pdf)
from 2023-04-02.

Closes https://github.com/llvm/llvm-project/issues/87565.
---
 libc/config/baremetal/arm/entrypoints.txt     |  1 +
 libc/config/baremetal/riscv/entrypoints.txt   |  1 +
 libc/config/darwin/arm/entrypoints.txt        |  1 +
 libc/config/darwin/x86_64/entrypoints.txt     |  1 +
 libc/config/linux/aarch64/entrypoints.txt     |  1 +
 libc/config/linux/arm/entrypoints.txt         |  1 +
 libc/config/linux/riscv/entrypoints.txt       |  1 +
 libc/config/linux/x86_64/entrypoints.txt      |  1 +
 libc/config/windows/entrypoints.txt           |  1 +
 libc/docs/c23.rst                             |  2 +-
 libc/docs/fenv.rst                            |  4 +--
 libc/spec/stdc.td                             |  5 +++
 libc/src/fenv/CMakeLists.txt                  | 13 ++++++++
 libc/src/fenv/fetestexceptflag.cpp            | 23 +++++++++++++
 libc/src/fenv/fetestexceptflag.h              | 20 ++++++++++++
 libc/test/src/fenv/CMakeLists.txt             |  1 +
 libc/test/src/fenv/exception_flags_test.cpp   | 32 +++++++++++++++----
 .../llvm-project-overlay/libc/BUILD.bazel     | 10 ++++++
 .../libc/test/src/fenv/BUILD.bazel            |  1 +
 19 files changed, 110 insertions(+), 10 deletions(-)
 create mode 100644 libc/src/fenv/fetestexceptflag.cpp
 create mode 100644 libc/src/fenv/fetestexceptflag.h

diff --git a/libc/config/baremetal/arm/entrypoints.txt b/libc/config/baremetal/arm/entrypoints.txt
index f33f9430c7920..4e3d1cb9f5337 100644
--- a/libc/config/baremetal/arm/entrypoints.txt
+++ b/libc/config/baremetal/arm/entrypoints.txt
@@ -201,6 +201,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/baremetal/riscv/entrypoints.txt b/libc/config/baremetal/riscv/entrypoints.txt
index dad187fa0496d..7efd9bcd5b3cb 100644
--- a/libc/config/baremetal/riscv/entrypoints.txt
+++ b/libc/config/baremetal/riscv/entrypoints.txt
@@ -201,6 +201,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/darwin/arm/entrypoints.txt b/libc/config/darwin/arm/entrypoints.txt
index aea2f6d5771e8..e1303265b9ac4 100644
--- a/libc/config/darwin/arm/entrypoints.txt
+++ b/libc/config/darwin/arm/entrypoints.txt
@@ -112,6 +112,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/darwin/x86_64/entrypoints.txt b/libc/config/darwin/x86_64/entrypoints.txt
index 09fe3d7b47687..02912decadcf7 100644
--- a/libc/config/darwin/x86_64/entrypoints.txt
+++ b/libc/config/darwin/x86_64/entrypoints.txt
@@ -106,6 +106,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     # libc.src.fenv.fesetround
     # libc.src.fenv.feraiseexcept
     # libc.src.fenv.fetestexcept
+    # libc.src.fenv.fetestexceptflag
     # libc.src.fenv.feupdateenv
 
     ## Currently disabled for failing tests.
diff --git a/libc/config/linux/aarch64/entrypoints.txt b/libc/config/linux/aarch64/entrypoints.txt
index 2952baacdd67f..1ac6bd9300008 100644
--- a/libc/config/linux/aarch64/entrypoints.txt
+++ b/libc/config/linux/aarch64/entrypoints.txt
@@ -324,6 +324,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/linux/arm/entrypoints.txt b/libc/config/linux/arm/entrypoints.txt
index 35fd588a9a6c4..335981ff7dc7c 100644
--- a/libc/config/linux/arm/entrypoints.txt
+++ b/libc/config/linux/arm/entrypoints.txt
@@ -192,6 +192,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/linux/riscv/entrypoints.txt b/libc/config/linux/riscv/entrypoints.txt
index 47c03a61c45a9..87e82e5eb9a06 100644
--- a/libc/config/linux/riscv/entrypoints.txt
+++ b/libc/config/linux/riscv/entrypoints.txt
@@ -332,6 +332,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index 8fdd4575e27e2..70f130a4399a3 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -346,6 +346,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/config/windows/entrypoints.txt b/libc/config/windows/entrypoints.txt
index c46c947bf3135..71216530c4041 100644
--- a/libc/config/windows/entrypoints.txt
+++ b/libc/config/windows/entrypoints.txt
@@ -110,6 +110,7 @@ set(TARGET_LIBM_ENTRYPOINTS
     libc.src.fenv.fesetround
     libc.src.fenv.feraiseexcept
     libc.src.fenv.fetestexcept
+    libc.src.fenv.fetestexceptflag
     libc.src.fenv.feupdateenv
 
     # math.h entrypoints
diff --git a/libc/docs/c23.rst b/libc/docs/c23.rst
index 4138c9d7104f3..44724fe1660cb 100644
--- a/libc/docs/c23.rst
+++ b/libc/docs/c23.rst
@@ -21,7 +21,7 @@ Additions:
 * fenv.h
 
   * fesetexcept |check|
-  * fetestexceptflag
+  * fetestexceptflag |check|
   * fegetmode
   * fesetmode
 * math.h
diff --git a/libc/docs/fenv.rst b/libc/docs/fenv.rst
index 6574fb7246ddd..1dee5515e1174 100644
--- a/libc/docs/fenv.rst
+++ b/libc/docs/fenv.rst
@@ -42,7 +42,7 @@ fenv.h Functions
     - |check|
     - 7.6.6.3
   * - fesetexcept
-    -
+    - |check|
     - 7.6.4.4
   * - fesetexceptflag
     - |check|
@@ -57,7 +57,7 @@ fenv.h Functions
     - |check|
     - 7.6.4.7
   * - fetestexceptflag
-    -
+    - |check|
     - 7.6.4.6
   * - feupdateenv
     - |check|
diff --git a/libc/spec/stdc.td b/libc/spec/stdc.td
index 63d0449867114..01aa7c70b3b9d 100644
--- a/libc/spec/stdc.td
+++ b/libc/spec/stdc.td
@@ -149,6 +149,11 @@ def StdC : StandardSpec<"stdc"> {
               RetValSpec<IntType>,
               [ArgSpec<IntType>]
           >,
+          FunctionSpec<
+              "fetestexceptflag",
+              RetValSpec<IntType>,
+              [ArgSpec<ConstFExceptTPtr>, ArgSpec<IntType>]
+          >,
           FunctionSpec<
               "feraiseexcept",
               RetValSpec<IntType>,
diff --git a/libc/src/fenv/CMakeLists.txt b/libc/src/fenv/CMakeLists.txt
index 17e9947412062..c5431b1b9d55e 100644
--- a/libc/src/fenv/CMakeLists.txt
+++ b/libc/src/fenv/CMakeLists.txt
@@ -58,6 +58,19 @@ add_entrypoint_object(
     -O2
 )
 
+add_entrypoint_object(
+  fetestexceptflag
+  SRCS
+    fetestexceptflag.cpp
+  HDRS
+    fetestexceptflag.h
+  DEPENDS
+    libc.hdr.types.fexcept_t
+    libc.src.__support.FPUtil.fenv_impl
+  COMPILE_OPTIONS
+    -O2
+)
+
 add_entrypoint_object(
   fegetenv
   SRCS
diff --git a/libc/src/fenv/fetestexceptflag.cpp b/libc/src/fenv/fetestexceptflag.cpp
new file mode 100644
index 0000000000000..63453350a199f
--- /dev/null
+++ b/libc/src/fenv/fetestexceptflag.cpp
@@ -0,0 +1,23 @@
+//===-- Implementation of fetestexceptflag function -----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/fenv/fetestexceptflag.h"
+#include "hdr/types/fexcept_t.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/common.h"
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, fetestexceptflag,
+                   (const fexcept_t *flagp, int excepts)) {
+  static_assert(sizeof(int) >= sizeof(fexcept_t),
+                "fexcept_t value cannot fit in an int value.");
+  return *flagp | fputil::test_except(excepts);
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/fenv/fetestexceptflag.h b/libc/src/fenv/fetestexceptflag.h
new file mode 100644
index 0000000000000..1c8b0b843f547
--- /dev/null
+++ b/libc/src/fenv/fetestexceptflag.h
@@ -0,0 +1,20 @@
+//===-- Implementation header for fetestexceptflag --------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_FENV_FETESTEXCEPTFLAG_H
+#define LLVM_LIBC_SRC_FENV_FETESTEXCEPTFLAG_H
+
+#include "hdr/types/fexcept_t.h"
+
+namespace LIBC_NAMESPACE {
+
+int fetestexceptflag(const fexcept_t *, int excepts);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_FENV_FETESTEXCEPTFLAG_H
diff --git a/libc/test/src/fenv/CMakeLists.txt b/libc/test/src/fenv/CMakeLists.txt
index 577735599dc01..f277b65e2d42b 100644
--- a/libc/test/src/fenv/CMakeLists.txt
+++ b/libc/test/src/fenv/CMakeLists.txt
@@ -48,6 +48,7 @@ add_libc_unittest(
   DEPENDS
     libc.src.fenv.fegetexceptflag
     libc.src.fenv.fesetexceptflag
+    libc.src.fenv.fetestexceptflag
     libc.src.__support.FPUtil.fenv_impl
 )
 
diff --git a/libc/test/src/fenv/exception_flags_test.cpp b/libc/test/src/fenv/exception_flags_test.cpp
index d1d8bfcc53db5..9d2be6426a6d0 100644
--- a/libc/test/src/fenv/exception_flags_test.cpp
+++ b/libc/test/src/fenv/exception_flags_test.cpp
@@ -1,4 +1,4 @@
-//===-- Unittests for fegetexceptflag and fesetexceptflag -----------------===//
+//===-- Unittests for fe{get|set|test}exceptflag --------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -9,11 +9,12 @@
 #include "hdr/types/fexcept_t.h"
 #include "src/fenv/fegetexceptflag.h"
 #include "src/fenv/fesetexceptflag.h"
+#include "src/fenv/fetestexceptflag.h"
 
 #include "src/__support/FPUtil/FEnvImpl.h"
 #include "test/UnitTest/Test.h"
 
-TEST(LlvmLibcFenvTest, GetExceptFlagAndSetExceptFlag) {
+TEST(LlvmLibcFenvTest, GetSetTestExceptFlag) {
   // We will disable all exceptions to prevent invocation of the exception
   // handler.
   LIBC_NAMESPACE::fputil::disable_except(FE_ALL_EXCEPT);
@@ -39,19 +40,36 @@ TEST(LlvmLibcFenvTest, GetExceptFlagAndSetExceptFlag) {
     ASSERT_EQ(LIBC_NAMESPACE::fesetexceptflag(&eflags, FE_ALL_EXCEPT), 0);
     ASSERT_NE(LIBC_NAMESPACE::fputil::test_except(FE_ALL_EXCEPT) & e, 0);
 
+    // Exception flags are exactly the flags corresponding to the previously
+    // raised exception.
+    ASSERT_EQ(LIBC_NAMESPACE::fetestexceptflag(&eflags, FE_ALL_EXCEPT),
+              LIBC_NAMESPACE::fputil::test_except(FE_ALL_EXCEPT));
+
     // Cleanup. We clear all excepts as raising excepts like FE_OVERFLOW
     // can also raise FE_INEXACT.
     LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
   }
 
-  // Next, we will raise one exception and save the flags.
+  // Next, we will raise one exception, save the flag and clear all exceptions.
   LIBC_NAMESPACE::fputil::raise_except(FE_INVALID);
-  fexcept_t eflags;
-  LIBC_NAMESPACE::fegetexceptflag(&eflags, FE_ALL_EXCEPT);
-  // Clear all exceptions and raise two other exceptions.
+  fexcept_t invalid_flag;
+  LIBC_NAMESPACE::fegetexceptflag(&invalid_flag, FE_ALL_EXCEPT);
+  ASSERT_EQ(LIBC_NAMESPACE::fetestexceptflag(&invalid_flag, FE_ALL_EXCEPT),
+            FE_INVALID);
   LIBC_NAMESPACE::fputil::clear_except(FE_ALL_EXCEPT);
+
+  // Raise two other exceptions and verify that they are set.
   LIBC_NAMESPACE::fputil::raise_except(FE_OVERFLOW | FE_INEXACT);
+  fexcept_t overflow_and_inexact_flag;
+  LIBC_NAMESPACE::fegetexceptflag(&overflow_and_inexact_flag, FE_ALL_EXCEPT);
+  ASSERT_EQ(LIBC_NAMESPACE::fetestexceptflag(&overflow_and_inexact_flag,
+                                             FE_ALL_EXCEPT),
+            FE_OVERFLOW | FE_INEXACT);
+  ASSERT_EQ(LIBC_NAMESPACE::fetestexceptflag(&overflow_and_inexact_flag,
+                                             FE_OVERFLOW | FE_INEXACT),
+            FE_OVERFLOW | FE_INEXACT);
+
   // When we set the flags and test, we should only see FE_INVALID.
-  LIBC_NAMESPACE::fesetexceptflag(&eflags, FE_ALL_EXCEPT);
+  LIBC_NAMESPACE::fesetexceptflag(&invalid_flag, FE_ALL_EXCEPT);
   EXPECT_EQ(LIBC_NAMESPACE::fputil::test_except(FE_ALL_EXCEPT), FE_INVALID);
 }
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index fb37f113b310a..6029cc3fee610 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -1145,6 +1145,16 @@ libc_function(
     ],
 )
 
+libc_function(
+    name = "fetestexceptflag",
+    srcs = ["src/fenv/fetestexceptflag.cpp"],
+    hdrs = ["src/fenv/fetestexceptflag.h"],
+    deps = [
+        ":__support_common",
+        ":__support_fputil_fenv_impl",
+    ],
+)
+
 libc_function(
     name = "feclearexcept",
     srcs = ["src/fenv/feclearexcept.cpp"],
diff --git a/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
index 359db0723dfd3..fc3ab3da3587c 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/src/fenv/BUILD.bazel
@@ -75,6 +75,7 @@ libc_test(
     libc_function_deps = [
         "//libc:fegetexceptflag",
         "//libc:fesetexceptflag",
+        "//libc:fetestexceptflag",
     ],
     deps = [
         "//libc:__support_fputil_fenv_impl",

From 8656d4c6a7a742c6fa6ee02c2ace7415163e65e4 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Wed, 17 Apr 2024 11:41:03 -0400
Subject: [PATCH 758/886] [Clang][Parse] Diagnose requires expressions with
 explicit object parameters (#88974)

Clang currently allows the following:
```
auto x = requires (this int) { true; };
```
This patch addresses that.
---
 clang/docs/ReleaseNotes.rst                       |  2 ++
 clang/include/clang/Basic/DiagnosticParseKinds.td |  2 ++
 clang/lib/Parse/ParseDecl.cpp                     | 15 ++++++++++++++-
 .../CXX/dcl.decl/dcl.meaning/dcl.fct/p6-cxx23.cpp |  7 +++++++
 4 files changed, 25 insertions(+), 1 deletion(-)
 create mode 100644 clang/test/CXX/dcl.decl/dcl.meaning/dcl.fct/p6-cxx23.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 96ad92b540b47..c19ad9fba58f3 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -364,6 +364,8 @@ Improvements to Clang's diagnostics
 - Clang now uses the correct type-parameter-key (``class`` or ``typename``) when printing
   template template parameter declarations.
 
+- Clang now diagnoses requires expressions with explicit object parameters.
+
 Improvements to Clang's time-trace
 ----------------------------------
 
diff --git a/clang/include/clang/Basic/DiagnosticParseKinds.td b/clang/include/clang/Basic/DiagnosticParseKinds.td
index bb9ca2a50cc06..66405095d51de 100644
--- a/clang/include/clang/Basic/DiagnosticParseKinds.td
+++ b/clang/include/clang/Basic/DiagnosticParseKinds.td
@@ -863,6 +863,8 @@ def err_empty_requires_expr : Error<
   "a requires expression must contain at least one requirement">;
 def err_requires_expr_parameter_list_ellipsis : Error<
   "varargs not allowed in requires expression">;
+def err_requires_expr_explicit_object_parameter: Error<
+  "a requires expression cannot have an explicit object parameter">;
 def err_expected_semi_requirement : Error<
   "expected ';' at end of requirement">;
 def err_requires_expr_missing_arrow : Error<
diff --git a/clang/lib/Parse/ParseDecl.cpp b/clang/lib/Parse/ParseDecl.cpp
index 274ee7b10c178..5f26b5a9e46be 100644
--- a/clang/lib/Parse/ParseDecl.cpp
+++ b/clang/lib/Parse/ParseDecl.cpp
@@ -7660,8 +7660,21 @@ void Parser::ParseParameterDeclarationClause(
     // Parse a C++23 Explicit Object Parameter
     // We do that in all language modes to produce a better diagnostic.
     SourceLocation ThisLoc;
-    if (getLangOpts().CPlusPlus && Tok.is(tok::kw_this))
+    if (getLangOpts().CPlusPlus && Tok.is(tok::kw_this)) {
       ThisLoc = ConsumeToken();
+      // C++23 [dcl.fct]p6:
+      //   An explicit-object-parameter-declaration is a parameter-declaration
+      //   with a this specifier. An explicit-object-parameter-declaration
+      //   shall appear only as the first parameter-declaration of a
+      //   parameter-declaration-list of either:
+      //   - a member-declarator that declares a member function, or
+      //   - a lambda-declarator.
+      //
+      // The parameter-declaration-list of a requires-expression is not such
+      // a context.
+      if (DeclaratorCtx == DeclaratorContext::RequiresExpr)
+        Diag(ThisLoc, diag::err_requires_expr_explicit_object_parameter);
+    }
 
     ParseDeclarationSpecifiers(DS, /*TemplateInfo=*/ParsedTemplateInfo(),
                                AS_none, DeclSpecContext::DSC_normal,
diff --git a/clang/test/CXX/dcl.decl/dcl.meaning/dcl.fct/p6-cxx23.cpp b/clang/test/CXX/dcl.decl/dcl.meaning/dcl.fct/p6-cxx23.cpp
new file mode 100644
index 0000000000000..9c1f30f81a011
--- /dev/null
+++ b/clang/test/CXX/dcl.decl/dcl.meaning/dcl.fct/p6-cxx23.cpp
@@ -0,0 +1,7 @@
+// RUN: %clang_cc1 -std=c++23 -fsyntax-only -verify %s
+
+auto x0 = requires (this int) { true; }; // expected-error {{a requires expression cannot have an explicit object parameter}}
+auto x1 = requires (int, this int) { true; }; // expected-error {{a requires expression cannot have an explicit object parameter}}
+
+template<this auto> // expected-error {{expected template parameter}}
+void f(); // expected-error {{no function template matches function template specialization 'f'}}

From abd5e45a96954d80f6ffe6d8676c0059fae8573b Mon Sep 17 00:00:00 2001
From: Alexander Richardson <alexrichardson@google.com>
Date: Wed, 17 Apr 2024 08:42:41 -0700
Subject: [PATCH 759/886] [compiler-rt] Use __atomic builtins whenever possible

The code in this file dates back to 2012 when Clang's support for atomic
builtins was still quite limited. The bugs referenced in the comment
at the top of the file have long been fixed and using the compiler
builtins directly should now generate slightly better code.
Additionally, this allows using the atomic builtin header for platforms
where the __sync_builtins are lacking (e.g. Arm Morello).

This change does not introduce any code generation changes for
__tsan_read*/__tsan_write* or __tsan_func_{entry,exit} on x86, which
indicates the previously noted compiler issues have been fixed.

We also have to touch the non-clang codepaths here since the only way we
can make this work easily is by making the memory_order enum match the
compiler-provided macros, so we have to update the debug checks that
assumed the enum was always a bitflag.

The one downside of this change is that 32-bit MIPS now definitely
requires libatomic (but that may already have been needed for RMW ops).

Reviewed By: dvyukov

Pull Request: https://github.com/llvm/llvm-project/pull/84439
---
 .../lib/sanitizer_common/CMakeLists.txt       |   3 -
 .../lib/sanitizer_common/sanitizer_atomic.h   |  12 ++
 .../sanitizer_common/sanitizer_atomic_clang.h |  85 ++++++-------
 .../sanitizer_atomic_clang_mips.h             | 117 ------------------
 .../sanitizer_atomic_clang_other.h            |  85 -------------
 .../sanitizer_atomic_clang_x86.h              | 113 -----------------
 .../sanitizer_common/sanitizer_atomic_msvc.h  |   8 +-
 .../compiler-rt/lib/sanitizer_common/BUILD.gn |   3 -
 8 files changed, 56 insertions(+), 370 deletions(-)
 delete mode 100644 compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_mips.h
 delete mode 100644 compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_other.h
 delete mode 100644 compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_x86.h

diff --git a/compiler-rt/lib/sanitizer_common/CMakeLists.txt b/compiler-rt/lib/sanitizer_common/CMakeLists.txt
index f2b4ac72ae157..66f2d259aa5fd 100644
--- a/compiler-rt/lib/sanitizer_common/CMakeLists.txt
+++ b/compiler-rt/lib/sanitizer_common/CMakeLists.txt
@@ -122,9 +122,6 @@ set(SANITIZER_IMPL_HEADERS
   sanitizer_asm.h
   sanitizer_atomic.h
   sanitizer_atomic_clang.h
-  sanitizer_atomic_clang_mips.h
-  sanitizer_atomic_clang_other.h
-  sanitizer_atomic_clang_x86.h
   sanitizer_atomic_msvc.h
   sanitizer_bitvector.h
   sanitizer_bvgraph.h
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic.h
index 46f06957228c9..0609a11ffdebb 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_atomic.h
@@ -18,12 +18,24 @@
 namespace __sanitizer {
 
 enum memory_order {
+// If the __atomic atomic builtins are supported (Clang/GCC), use the
+// compiler provided macro values so that we can map the atomic operations
+// to __atomic_* directly.
+#ifdef __ATOMIC_SEQ_CST
+  memory_order_relaxed = __ATOMIC_RELAXED,
+  memory_order_consume = __ATOMIC_CONSUME,
+  memory_order_acquire = __ATOMIC_ACQUIRE,
+  memory_order_release = __ATOMIC_RELEASE,
+  memory_order_acq_rel = __ATOMIC_ACQ_REL,
+  memory_order_seq_cst = __ATOMIC_SEQ_CST
+#else
   memory_order_relaxed = 1 << 0,
   memory_order_consume = 1 << 1,
   memory_order_acquire = 1 << 2,
   memory_order_release = 1 << 3,
   memory_order_acq_rel = 1 << 4,
   memory_order_seq_cst = 1 << 5
+#endif
 };
 
 struct atomic_uint8_t {
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang.h
index 4318d64d16cfa..1414092e38d7e 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang.h
@@ -14,60 +14,63 @@
 #ifndef SANITIZER_ATOMIC_CLANG_H
 #define SANITIZER_ATOMIC_CLANG_H
 
-#if defined(__i386__) || defined(__x86_64__)
-# include "sanitizer_atomic_clang_x86.h"
-#else
-# include "sanitizer_atomic_clang_other.h"
-#endif
-
 namespace __sanitizer {
 
-// We would like to just use compiler builtin atomic operations
-// for loads and stores, but they are mostly broken in clang:
-// - they lead to vastly inefficient code generation
-// (http://llvm.org/bugs/show_bug.cgi?id=17281)
-// - 64-bit atomic operations are not implemented on x86_32
-// (http://llvm.org/bugs/show_bug.cgi?id=15034)
-// - they are not implemented on ARM
-// error: undefined reference to '__atomic_load_4'
+// We use the compiler builtin atomic operations for loads and stores, which
+// generates correct code for all architectures, but may require libatomic
+// on platforms where e.g. 64-bit atomics are not supported natively.
 
 // See http://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html
 // for mappings of the memory model to different processors.
 
-inline void atomic_signal_fence(memory_order) {
+inline void atomic_signal_fence(memory_order mo) { __atomic_signal_fence(mo); }
+
+inline void atomic_thread_fence(memory_order mo) { __atomic_thread_fence(mo); }
+
+inline void proc_yield(int cnt) {
+  __asm__ __volatile__("" ::: "memory");
+#if defined(__i386__) || defined(__x86_64__)
+  for (int i = 0; i < cnt; i++) __asm__ __volatile__("pause");
   __asm__ __volatile__("" ::: "memory");
+#endif
 }
 
-inline void atomic_thread_fence(memory_order) {
-  __sync_synchronize();
+template <typename T>
+inline typename T::Type atomic_load(const volatile T *a, memory_order mo) {
+  DCHECK(mo == memory_order_relaxed || mo == memory_order_consume ||
+         mo == memory_order_acquire || mo == memory_order_seq_cst);
+  DCHECK(!((uptr)a % sizeof(*a)));
+  return __atomic_load_n(&a->val_dont_use, mo);
 }
 
-template<typename T>
-inline typename T::Type atomic_fetch_add(volatile T *a,
-    typename T::Type v, memory_order mo) {
-  (void)mo;
+template <typename T>
+inline void atomic_store(volatile T *a, typename T::Type v, memory_order mo) {
+  DCHECK(mo == memory_order_relaxed || mo == memory_order_release ||
+         mo == memory_order_seq_cst);
   DCHECK(!((uptr)a % sizeof(*a)));
-  return __sync_fetch_and_add(&a->val_dont_use, v);
+  __atomic_store_n(&a->val_dont_use, v, mo);
 }
 
-template<typename T>
-inline typename T::Type atomic_fetch_sub(volatile T *a,
-    typename T::Type v, memory_order mo) {
+template <typename T>
+inline typename T::Type atomic_fetch_add(volatile T *a, typename T::Type v,
+                                         memory_order mo) {
+  DCHECK(!((uptr)a % sizeof(*a)));
+  return __atomic_fetch_add(&a->val_dont_use, v, mo);
+}
+
+template <typename T>
+inline typename T::Type atomic_fetch_sub(volatile T *a, typename T::Type v,
+                                         memory_order mo) {
   (void)mo;
   DCHECK(!((uptr)a % sizeof(*a)));
-  return __sync_fetch_and_add(&a->val_dont_use, -v);
+  return __atomic_fetch_sub(&a->val_dont_use, v, mo);
 }
 
-template<typename T>
-inline typename T::Type atomic_exchange(volatile T *a,
-    typename T::Type v, memory_order mo) {
+template <typename T>
+inline typename T::Type atomic_exchange(volatile T *a, typename T::Type v,
+                                        memory_order mo) {
   DCHECK(!((uptr)a % sizeof(*a)));
-  if (mo & (memory_order_release | memory_order_acq_rel | memory_order_seq_cst))
-    __sync_synchronize();
-  v = __sync_lock_test_and_set(&a->val_dont_use, v);
-  if (mo == memory_order_seq_cst)
-    __sync_synchronize();
-  return v;
+  return __atomic_exchange_n(&a->val_dont_use, v, mo);
 }
 
 template <typename T>
@@ -82,9 +85,8 @@ inline bool atomic_compare_exchange_strong(volatile T *a, typename T::Type *cmp,
                                    __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST);
 }
 
-template<typename T>
-inline bool atomic_compare_exchange_weak(volatile T *a,
-                                         typename T::Type *cmp,
+template <typename T>
+inline bool atomic_compare_exchange_weak(volatile T *a, typename T::Type *cmp,
                                          typename T::Type xchg,
                                          memory_order mo) {
   return atomic_compare_exchange_strong(a, cmp, xchg, mo);
@@ -92,13 +94,6 @@ inline bool atomic_compare_exchange_weak(volatile T *a,
 
 }  // namespace __sanitizer
 
-// This include provides explicit template instantiations for atomic_uint64_t
-// on MIPS32, which does not directly support 8 byte atomics. It has to
-// proceed the template definitions above.
-#if defined(_MIPS_SIM) && defined(_ABIO32) && _MIPS_SIM == _ABIO32
-#  include "sanitizer_atomic_clang_mips.h"
-#endif
-
 #undef ATOMIC_ORDER
 
 #endif  // SANITIZER_ATOMIC_CLANG_H
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_mips.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_mips.h
deleted file mode 100644
index f3d3052e5b7c5..0000000000000
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_mips.h
+++ /dev/null
@@ -1,117 +0,0 @@
-//===-- sanitizer_atomic_clang_mips.h ---------------------------*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file is a part of ThreadSanitizer/AddressSanitizer runtime.
-// Not intended for direct inclusion. Include sanitizer_atomic.h.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef SANITIZER_ATOMIC_CLANG_MIPS_H
-#define SANITIZER_ATOMIC_CLANG_MIPS_H
-
-namespace __sanitizer {
-
-// MIPS32 does not support atomics > 4 bytes. To address this lack of
-// functionality, the sanitizer library provides helper methods which use an
-// internal spin lock mechanism to emulate atomic operations when the size is
-// 8 bytes.
-static void __spin_lock(volatile int *lock) {
-  while (__sync_lock_test_and_set(lock, 1))
-    while (*lock) {
-    }
-}
-
-static void __spin_unlock(volatile int *lock) { __sync_lock_release(lock); }
-
-// Make sure the lock is on its own cache line to prevent false sharing.
-// Put it inside a struct that is aligned and padded to the typical MIPS
-// cacheline which is 32 bytes.
-static struct {
-  int lock;
-  char pad[32 - sizeof(int)];
-} __attribute__((aligned(32))) lock = {0, {0}};
-
-template <>
-inline atomic_uint64_t::Type atomic_fetch_add(volatile atomic_uint64_t *ptr,
-                                              atomic_uint64_t::Type val,
-                                              memory_order mo) {
-  DCHECK(mo &
-         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
-  DCHECK(!((uptr)ptr % sizeof(*ptr)));
-
-  atomic_uint64_t::Type ret;
-
-  __spin_lock(&lock.lock);
-  ret = *(const_cast<atomic_uint64_t::Type volatile *>(&ptr->val_dont_use));
-  ptr->val_dont_use = ret + val;
-  __spin_unlock(&lock.lock);
-
-  return ret;
-}
-
-template <>
-inline atomic_uint64_t::Type atomic_fetch_sub(volatile atomic_uint64_t *ptr,
-                                              atomic_uint64_t::Type val,
-                                              memory_order mo) {
-  return atomic_fetch_add(ptr, -val, mo);
-}
-
-template <>
-inline bool atomic_compare_exchange_strong(volatile atomic_uint64_t *ptr,
-                                           atomic_uint64_t::Type *cmp,
-                                           atomic_uint64_t::Type xchg,
-                                           memory_order mo) {
-  DCHECK(mo &
-         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
-  DCHECK(!((uptr)ptr % sizeof(*ptr)));
-
-  typedef atomic_uint64_t::Type Type;
-  Type cmpv = *cmp;
-  Type prev;
-  bool ret = false;
-
-  __spin_lock(&lock.lock);
-  prev = *(const_cast<Type volatile *>(&ptr->val_dont_use));
-  if (prev == cmpv) {
-    ret = true;
-    ptr->val_dont_use = xchg;
-  }
-  __spin_unlock(&lock.lock);
-
-  return ret;
-}
-
-template <>
-inline atomic_uint64_t::Type atomic_load(const volatile atomic_uint64_t *ptr,
-                                         memory_order mo) {
-  DCHECK(mo &
-         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
-  DCHECK(!((uptr)ptr % sizeof(*ptr)));
-
-  atomic_uint64_t::Type zero = 0;
-  volatile atomic_uint64_t *Newptr =
-      const_cast<volatile atomic_uint64_t *>(ptr);
-  return atomic_fetch_add(Newptr, zero, mo);
-}
-
-template <>
-inline void atomic_store(volatile atomic_uint64_t *ptr, atomic_uint64_t::Type v,
-                         memory_order mo) {
-  DCHECK(mo &
-         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
-  DCHECK(!((uptr)ptr % sizeof(*ptr)));
-
-  __spin_lock(&lock.lock);
-  ptr->val_dont_use = v;
-  __spin_unlock(&lock.lock);
-}
-
-}  // namespace __sanitizer
-
-#endif  // SANITIZER_ATOMIC_CLANG_MIPS_H
-
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_other.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_other.h
deleted file mode 100644
index 557082a636b87..0000000000000
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_other.h
+++ /dev/null
@@ -1,85 +0,0 @@
-//===-- sanitizer_atomic_clang_other.h --------------------------*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file is a part of ThreadSanitizer/AddressSanitizer runtime.
-// Not intended for direct inclusion. Include sanitizer_atomic.h.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef SANITIZER_ATOMIC_CLANG_OTHER_H
-#define SANITIZER_ATOMIC_CLANG_OTHER_H
-
-namespace __sanitizer {
-
-
-inline void proc_yield(int cnt) {
-  __asm__ __volatile__("" ::: "memory");
-}
-
-template<typename T>
-inline typename T::Type atomic_load(
-    const volatile T *a, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_consume
-      | memory_order_acquire | memory_order_seq_cst));
-  DCHECK(!((uptr)a % sizeof(*a)));
-  typename T::Type v;
-
-  if (sizeof(*a) < 8 || sizeof(void*) == 8) {
-    // Assume that aligned loads are atomic.
-    if (mo == memory_order_relaxed) {
-      v = a->val_dont_use;
-    } else if (mo == memory_order_consume) {
-      // Assume that processor respects data dependencies
-      // (and that compiler won't break them).
-      __asm__ __volatile__("" ::: "memory");
-      v = a->val_dont_use;
-      __asm__ __volatile__("" ::: "memory");
-    } else if (mo == memory_order_acquire) {
-      __asm__ __volatile__("" ::: "memory");
-      v = a->val_dont_use;
-      __sync_synchronize();
-    } else {  // seq_cst
-      // E.g. on POWER we need a hw fence even before the store.
-      __sync_synchronize();
-      v = a->val_dont_use;
-      __sync_synchronize();
-    }
-  } else {
-    __atomic_load(const_cast<typename T::Type volatile *>(&a->val_dont_use), &v,
-                  __ATOMIC_SEQ_CST);
-  }
-  return v;
-}
-
-template<typename T>
-inline void atomic_store(volatile T *a, typename T::Type v, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_release
-      | memory_order_seq_cst));
-  DCHECK(!((uptr)a % sizeof(*a)));
-
-  if (sizeof(*a) < 8 || sizeof(void*) == 8) {
-    // Assume that aligned stores are atomic.
-    if (mo == memory_order_relaxed) {
-      a->val_dont_use = v;
-    } else if (mo == memory_order_release) {
-      __sync_synchronize();
-      a->val_dont_use = v;
-      __asm__ __volatile__("" ::: "memory");
-    } else {  // seq_cst
-      __sync_synchronize();
-      a->val_dont_use = v;
-      __sync_synchronize();
-    }
-  } else {
-    __atomic_store(&a->val_dont_use, &v, __ATOMIC_SEQ_CST);
-  }
-}
-
-}  // namespace __sanitizer
-
-#endif  // #ifndef SANITIZER_ATOMIC_CLANG_OTHER_H
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_x86.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_x86.h
deleted file mode 100644
index b81a354d20987..0000000000000
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_clang_x86.h
+++ /dev/null
@@ -1,113 +0,0 @@
-//===-- sanitizer_atomic_clang_x86.h ----------------------------*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file is a part of ThreadSanitizer/AddressSanitizer runtime.
-// Not intended for direct inclusion. Include sanitizer_atomic.h.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef SANITIZER_ATOMIC_CLANG_X86_H
-#define SANITIZER_ATOMIC_CLANG_X86_H
-
-namespace __sanitizer {
-
-inline void proc_yield(int cnt) {
-  __asm__ __volatile__("" ::: "memory");
-  for (int i = 0; i < cnt; i++)
-    __asm__ __volatile__("pause");
-  __asm__ __volatile__("" ::: "memory");
-}
-
-template<typename T>
-inline typename T::Type atomic_load(
-    const volatile T *a, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_consume
-      | memory_order_acquire | memory_order_seq_cst));
-  DCHECK(!((uptr)a % sizeof(*a)));
-  typename T::Type v;
-
-  if (sizeof(*a) < 8 || sizeof(void*) == 8) {
-    // Assume that aligned loads are atomic.
-    if (mo == memory_order_relaxed) {
-      v = a->val_dont_use;
-    } else if (mo == memory_order_consume) {
-      // Assume that processor respects data dependencies
-      // (and that compiler won't break them).
-      __asm__ __volatile__("" ::: "memory");
-      v = a->val_dont_use;
-      __asm__ __volatile__("" ::: "memory");
-    } else if (mo == memory_order_acquire) {
-      __asm__ __volatile__("" ::: "memory");
-      v = a->val_dont_use;
-      // On x86 loads are implicitly acquire.
-      __asm__ __volatile__("" ::: "memory");
-    } else {  // seq_cst
-      // On x86 plain MOV is enough for seq_cst store.
-      __asm__ __volatile__("" ::: "memory");
-      v = a->val_dont_use;
-      __asm__ __volatile__("" ::: "memory");
-    }
-  } else {
-    // 64-bit load on 32-bit platform.
-    __asm__ __volatile__(
-        "movq %1, %%mm0;"  // Use mmx reg for 64-bit atomic moves
-        "movq %%mm0, %0;"  // (ptr could be read-only)
-        "emms;"            // Empty mmx state/Reset FP regs
-        : "=m" (v)
-        : "m" (a->val_dont_use)
-        : // mark the mmx registers as clobbered
-#ifdef __MMX__
-          "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7",
-#endif  // #ifdef __MMX__
-          "memory");
-  }
-  return v;
-}
-
-template<typename T>
-inline void atomic_store(volatile T *a, typename T::Type v, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_release
-      | memory_order_seq_cst));
-  DCHECK(!((uptr)a % sizeof(*a)));
-
-  if (sizeof(*a) < 8 || sizeof(void*) == 8) {
-    // Assume that aligned stores are atomic.
-    if (mo == memory_order_relaxed) {
-      a->val_dont_use = v;
-    } else if (mo == memory_order_release) {
-      // On x86 stores are implicitly release.
-      __asm__ __volatile__("" ::: "memory");
-      a->val_dont_use = v;
-      __asm__ __volatile__("" ::: "memory");
-    } else {  // seq_cst
-      // On x86 stores are implicitly release.
-      __asm__ __volatile__("" ::: "memory");
-      a->val_dont_use = v;
-      __sync_synchronize();
-    }
-  } else {
-    // 64-bit store on 32-bit platform.
-    __asm__ __volatile__(
-        "movq %1, %%mm0;"  // Use mmx reg for 64-bit atomic moves
-        "movq %%mm0, %0;"
-        "emms;"            // Empty mmx state/Reset FP regs
-        : "=m" (a->val_dont_use)
-        : "m" (v)
-        : // mark the mmx registers as clobbered
-#ifdef __MMX__
-          "mm0", "mm1", "mm2", "mm3", "mm4", "mm5", "mm6", "mm7",
-#endif  // #ifdef __MMX__
-          "memory");
-    if (mo == memory_order_seq_cst)
-      __sync_synchronize();
-  }
-}
-
-}  // namespace __sanitizer
-
-#endif  // #ifndef SANITIZER_ATOMIC_CLANG_X86_H
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_msvc.h b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_msvc.h
index 31317adcdfc99..d80bfdbf6a081 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_atomic_msvc.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_atomic_msvc.h
@@ -70,8 +70,8 @@ inline void proc_yield(int cnt) {
 template<typename T>
 inline typename T::Type atomic_load(
     const volatile T *a, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_consume
-      | memory_order_acquire | memory_order_seq_cst));
+  DCHECK(mo == memory_order_relaxed || mo == memory_order_consume ||
+         mo == memory_order_acquire || mo == memory_order_seq_cst);
   DCHECK(!((uptr)a % sizeof(*a)));
   typename T::Type v;
   // FIXME(dvyukov): 64-bit load is not atomic on 32-bits.
@@ -87,8 +87,8 @@ inline typename T::Type atomic_load(
 
 template<typename T>
 inline void atomic_store(volatile T *a, typename T::Type v, memory_order mo) {
-  DCHECK(mo & (memory_order_relaxed | memory_order_release
-      | memory_order_seq_cst));
+  DCHECK(mo == memory_order_relaxed || mo == memory_order_release ||
+         mo == memory_order_seq_cst);
   DCHECK(!((uptr)a % sizeof(*a)));
   // FIXME(dvyukov): 64-bit store is not atomic on 32-bits.
   if (mo == memory_order_relaxed) {
diff --git a/llvm/utils/gn/secondary/compiler-rt/lib/sanitizer_common/BUILD.gn b/llvm/utils/gn/secondary/compiler-rt/lib/sanitizer_common/BUILD.gn
index 0519073239430..f7f1fce10bf5f 100644
--- a/llvm/utils/gn/secondary/compiler-rt/lib/sanitizer_common/BUILD.gn
+++ b/llvm/utils/gn/secondary/compiler-rt/lib/sanitizer_common/BUILD.gn
@@ -33,9 +33,6 @@ source_set("sources") {
     "sanitizer_asm.h",
     "sanitizer_atomic.h",
     "sanitizer_atomic_clang.h",
-    "sanitizer_atomic_clang_mips.h",
-    "sanitizer_atomic_clang_other.h",
-    "sanitizer_atomic_clang_x86.h",
     "sanitizer_atomic_msvc.h",
     "sanitizer_bitvector.h",
     "sanitizer_bvgraph.h",

From a88ea8fbb3953c2fe2887438baf342e381a79d8b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Wed, 17 Apr 2024 08:43:11 -0700
Subject: [PATCH 760/886] [flang][cuda] Update memory effect on
 fir.cuda_allocate op (#88930)

Add MemRead effect on the box operand as the descriptor might be read
when performing the allocation of the data.

Also update the expected type of the box operand to be a reference.
Check in the verifier that this is a reference to a box or class type.

This addresses the comment made post commit on #88586
---
 flang/include/flang/Optimizer/Dialect/FIROps.td | 2 +-
 flang/lib/Optimizer/Dialect/FIROps.cpp          | 2 +-
 flang/test/Fir/cuf-invalid.fir                  | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/flang/include/flang/Optimizer/Dialect/FIROps.td b/flang/include/flang/Optimizer/Dialect/FIROps.td
index 580e840587abb..92790a691e473 100644
--- a/flang/include/flang/Optimizer/Dialect/FIROps.td
+++ b/flang/include/flang/Optimizer/Dialect/FIROps.td
@@ -3200,7 +3200,7 @@ def fir_CUDAAllocateOp : fir_Op<"cuda_allocate", [AttrSizedOperandSegments,
     is initialized before with the standard flang runtime calls.
   }];
 
-  let arguments = (ins Arg<AnyRefOrBoxType, "", [MemWrite]>:$box,
+  let arguments = (ins Arg<fir_ReferenceType, "", [MemRead, MemWrite]>:$box,
                        Arg<Optional<AnyRefOrBoxType>, "", [MemWrite]>:$errmsg,
                        Optional<AnyIntegerType>:$stream,
                        Arg<Optional<AnyRefOrBoxType>, "", [MemWrite]>:$pinned,
diff --git a/flang/lib/Optimizer/Dialect/FIROps.cpp b/flang/lib/Optimizer/Dialect/FIROps.cpp
index be27256d911b3..5c24c95db427a 100644
--- a/flang/lib/Optimizer/Dialect/FIROps.cpp
+++ b/flang/lib/Optimizer/Dialect/FIROps.cpp
@@ -3998,7 +3998,7 @@ mlir::LogicalResult fir::CUDAAllocateOp::verify() {
     return emitOpError("pinned and stream cannot appears at the same time");
   if (!fir::unwrapRefType(getBox().getType()).isa<fir::BaseBoxType>())
     return emitOpError(
-        "expect box to be a reference to/or a class or box type value");
+        "expect box to be a reference to a class or box type value");
   if (getSource() &&
       !fir::unwrapRefType(getSource().getType()).isa<fir::BaseBoxType>())
     return emitOpError(
diff --git a/flang/test/Fir/cuf-invalid.fir b/flang/test/Fir/cuf-invalid.fir
index 5d3aa55cf346a..6c533a32ccf9b 100644
--- a/flang/test/Fir/cuf-invalid.fir
+++ b/flang/test/Fir/cuf-invalid.fir
@@ -16,7 +16,7 @@ func.func @_QPsub1() {
 
 func.func @_QPsub1() {
   %1 = fir.alloca i32
-  // expected-error@+1{{'fir.cuda_allocate' op expect box to be a reference to/or a class or box type value}}
+  // expected-error@+1{{'fir.cuda_allocate' op expect box to be a reference to a class or box type value}}
   %2 = fir.cuda_allocate %1 : !fir.ref<i32> {cuda_attr = #fir.cuda<device>} -> i32
   return
 }

From da70f2cdcde8cb96e75ce0236db1fb5353407a69 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Wed, 17 Apr 2024 08:43:25 -0700
Subject: [PATCH 761/886] [flang][cuda] Lower ALLOCATE for device variable
 (#88980)

Replace the runtime call to `AllocatableAllocate` for CUDA device
variable to the newly added `fir.cuda_allocate` operation.
---
 flang/lib/Lower/Allocatable.cpp            |  57 +++++++++--
 flang/test/Lower/CUDA/cuda-allocatable.cuf | 107 +++++++++++++++++++++
 2 files changed, 154 insertions(+), 10 deletions(-)
 create mode 100644 flang/test/Lower/CUDA/cuda-allocatable.cuf

diff --git a/flang/lib/Lower/Allocatable.cpp b/flang/lib/Lower/Allocatable.cpp
index 42e78fc96e444..1d434d512d0c5 100644
--- a/flang/lib/Lower/Allocatable.cpp
+++ b/flang/lib/Lower/Allocatable.cpp
@@ -14,6 +14,7 @@
 #include "flang/Evaluate/tools.h"
 #include "flang/Lower/AbstractConverter.h"
 #include "flang/Lower/ConvertType.h"
+#include "flang/Lower/ConvertVariable.h"
 #include "flang/Lower/IterationSpace.h"
 #include "flang/Lower/Mangler.h"
 #include "flang/Lower/OpenACC.h"
@@ -368,20 +369,17 @@ class AllocateStmtHelper {
               [&](const Fortran::parser::AllocOpt::Mold &mold) {
                 moldExpr = Fortran::semantics::GetExpr(mold.v.value());
               },
-              [&](const Fortran::parser::AllocOpt::Stream &) {
-                TODO(loc, "CUDA ALLOCATE(STREAM=)");
+              [&](const Fortran::parser::AllocOpt::Stream &stream) {
+                streamExpr = Fortran::semantics::GetExpr(stream.v.value());
               },
-              [&](const Fortran::parser::AllocOpt::Pinned &) {
-                TODO(loc, "CUDA ALLOCATE(PINNED=)");
+              [&](const Fortran::parser::AllocOpt::Pinned &pinned) {
+                pinnedExpr = Fortran::semantics::GetExpr(pinned.v.value());
               },
           },
           allocOption.u);
   }
 
   void lowerAllocation(const Allocation &alloc) {
-    if (Fortran::semantics::HasCUDAAttr(alloc.getSymbol()))
-      TODO(loc, "Allocation of variable with CUDA attributes");
-
     fir::MutableBoxValue boxAddr =
         genMutableBoxValue(converter, loc, alloc.getAllocObj());
 
@@ -456,7 +454,8 @@ class AllocateStmtHelper {
                            const fir::MutableBoxValue &box) {
     if (!box.isDerived() && !errorManager.hasStatSpec() &&
         !alloc.type.IsPolymorphic() && !alloc.hasCoarraySpec() &&
-        !useAllocateRuntime && !box.isPointer()) {
+        !useAllocateRuntime && !box.isPointer() &&
+        !Fortran::semantics::HasCUDAAttr(alloc.getSymbol())) {
       // Pointers must use PointerAllocate so that their deallocations
       // can be validated.
       genInlinedAllocation(alloc, box);
@@ -472,7 +471,12 @@ class AllocateStmtHelper {
       genSetType(alloc, box, loc);
     genSetDeferredLengthParameters(alloc, box);
     genAllocateObjectBounds(alloc, box);
-    mlir::Value stat = genRuntimeAllocate(builder, loc, box, errorManager);
+    mlir::Value stat;
+    if (!Fortran::semantics::HasCUDAAttr(alloc.getSymbol()))
+      stat = genRuntimeAllocate(builder, loc, box, errorManager);
+    else
+      stat =
+          genCudaAllocate(builder, loc, box, errorManager, alloc.getSymbol());
     fir::factory::syncMutableBoxFromIRBox(builder, loc, box);
     postAllocationAction(alloc);
     errorManager.assignStat(builder, loc, stat);
@@ -602,7 +606,10 @@ class AllocateStmtHelper {
       genSetDeferredLengthParameters(alloc, box);
     genAllocateObjectBounds(alloc, box);
     mlir::Value stat;
-    if (isSource)
+    if (Fortran::semantics::HasCUDAAttr(alloc.getSymbol()))
+      stat =
+          genCudaAllocate(builder, loc, box, errorManager, alloc.getSymbol());
+    else if (isSource)
       stat = genRuntimeAllocateSource(builder, loc, box, exv, errorManager);
     else
       stat = genRuntimeAllocate(builder, loc, box, errorManager);
@@ -717,6 +724,34 @@ class AllocateStmtHelper {
     return nullptr;
   }
 
+  mlir::Value genCudaAllocate(fir::FirOpBuilder &builder, mlir::Location loc,
+                              const fir::MutableBoxValue &box,
+                              ErrorManager &errorManager,
+                              const Fortran::semantics::Symbol &sym) {
+    Fortran::lower::StatementContext stmtCtx;
+    fir::CUDADataAttributeAttr cudaAttr =
+        Fortran::lower::translateSymbolCUDADataAttribute(builder.getContext(),
+                                                         sym);
+    mlir::Value errmsg = errMsgExpr ? errorManager.errMsgAddr : nullptr;
+    mlir::Value stream =
+        streamExpr
+            ? fir::getBase(converter.genExprValue(loc, *streamExpr, stmtCtx))
+            : nullptr;
+    mlir::Value pinned =
+        pinnedExpr
+            ? fir::getBase(converter.genExprAddr(loc, *pinnedExpr, stmtCtx))
+            : nullptr;
+    mlir::Value source = sourceExpr ? fir::getBase(sourceExv) : nullptr;
+
+    // Keep return type the same as a standard AllocatableAllocate call.
+    mlir::Type retTy = fir::runtime::getModel<int>()(builder.getContext());
+    return builder
+        .create<fir::CUDAAllocateOp>(
+            loc, retTy, box.getAddr(), errmsg, stream, pinned, source, cudaAttr,
+            errorManager.hasStatSpec() ? builder.getUnitAttr() : nullptr)
+        .getResult();
+  }
+
   Fortran::lower::AbstractConverter &converter;
   fir::FirOpBuilder &builder;
   const Fortran::parser::AllocateStmt &stmt;
@@ -724,6 +759,8 @@ class AllocateStmtHelper {
   const Fortran::lower::SomeExpr *moldExpr{nullptr};
   const Fortran::lower::SomeExpr *statExpr{nullptr};
   const Fortran::lower::SomeExpr *errMsgExpr{nullptr};
+  const Fortran::lower::SomeExpr *pinnedExpr{nullptr};
+  const Fortran::lower::SomeExpr *streamExpr{nullptr};
   // If the allocate has a type spec, lenParams contains the
   // value of the length parameters that were specified inside.
   llvm::SmallVector<mlir::Value> lenParams;
diff --git a/flang/test/Lower/CUDA/cuda-allocatable.cuf b/flang/test/Lower/CUDA/cuda-allocatable.cuf
new file mode 100644
index 0000000000000..55223011e8d9e
--- /dev/null
+++ b/flang/test/Lower/CUDA/cuda-allocatable.cuf
@@ -0,0 +1,107 @@
+! RUN: bbc -emit-hlfir -fcuda %s -o - | FileCheck %s
+
+! Test lowering of CUDA allocatable allocate/deallocate statements.
+
+subroutine sub1()
+  real, allocatable, device :: a(:)
+  allocate(a(10))
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub1()
+! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub1Ea"}
+! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub1Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<device>} -> i32
+
+subroutine sub2()
+  real, allocatable, managed :: a(:)
+  integer :: istat
+  allocate(a(10), stat=istat)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub2()
+! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub2Ea"}
+! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub2Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[ISTAT:.*]] = fir.alloca i32 {bindc_name = "istat", uniq_name = "_QFsub2Eistat"}
+! CHECK: %[[ISTAT_DECL:.*]]:2 = hlfir.declare %[[ISTAT]] {uniq_name = "_QFsub2Eistat"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %[[STAT:.*]] = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<managed>, hasStat} -> i32
+! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>
+
+subroutine sub3()
+  integer, allocatable, pinned :: a(:,:)
+  logical :: plog
+  allocate(a(20,30), pinned = plog)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub3()
+! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x?xi32>>> {bindc_name = "a", uniq_name = "_QFsub3Ea"}
+! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub3Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>>)
+! CHECK: %[[PLOG:.*]] = fir.alloca !fir.logical<4> {bindc_name = "plog", uniq_name = "_QFsub3Eplog"}
+! CHECK: %[[PLOG_DECL:.*]]:2 = hlfir.declare %5 {uniq_name = "_QFsub3Eplog"} : (!fir.ref<!fir.logical<4>>) -> (!fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>)
+! CHECK-2: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xi32>>>> pinned(%[[PLOG_DECL]]#1 : !fir.ref<!fir.logical<4>>) {cuda_attr = #fir.cuda<pinned>} -> i32
+
+subroutine sub4()
+  real, allocatable, unified :: a(:)
+  integer :: istream
+  allocate(a(10), stream=istream)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub4()
+! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub4Ea"}
+! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %0 {cuda_attr = #fir.cuda<unified>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub4Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[ISTREAM:.*]] = fir.alloca i32 {bindc_name = "istream", uniq_name = "_QFsub4Eistream"}
+! CHECK: %[[ISTREAM_DECL:.*]]:2 = hlfir.declare %[[ISTREAM]] {uniq_name = "_QFsub4Eistream"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %[[STREAM:.*]] = fir.load %[[ISTREAM_DECL]]#0 : !fir.ref<i32>
+! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> stream(%[[STREAM]] : i32) {cuda_attr = #fir.cuda<unified>} -> i32
+
+subroutine sub5()
+  real, allocatable, device :: a(:)
+  real, allocatable :: b(:)
+  allocate(a, source=b)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub5()
+! CHECK: %[[BOX_A:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub5Ea"}
+! CHECK: %[[BOX_A_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub5Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[BOX_B:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "b", uniq_name = "_QFsub5Eb"}
+! CHECK: %[[BOX_B_DECL:.*]]:2 = hlfir.declare %[[BOX_B]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub5Eb"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[LOAD_B:.*]] = fir.load %[[BOX_B_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
+! CHECK: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> source(%[[LOAD_B]] : !fir.box<!fir.heap<!fir.array<?xf32>>>) {cuda_attr = #fir.cuda<device>} -> i32
+
+subroutine sub6()
+  real, allocatable, device :: a(:)
+  real, allocatable :: b(:)
+  allocate(a, mold=b)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub6()
+! CHECK: %[[BOX_A:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub6Ea"}
+! CHECK: %[[BOX_A_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub6Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[BOX_B:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "b", uniq_name = "_QFsub6Eb"}
+! CHECK: %[[BOX_B_DECL:.*]]:2 = hlfir.declare %[[BOX_B]] {fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub6Eb"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[LOAD_B:.*]] = fir.load %[[BOX_B_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>
+! CHECK: fir.call @_FortranAAllocatableApplyMold
+! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_A_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<device>} -> i32
+
+subroutine sub7()
+  real, allocatable, device :: a(:)
+  integer :: istat
+  character(50) :: err
+  allocate(a(100), stat=istat, errmsg=err)
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub7()
+! CHECK: %[[BOX:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?xf32>>> {bindc_name = "a", uniq_name = "_QFsub7Ea"}
+! CHECK: %[[BOX_DECL:.*]]:2 = hlfir.declare %[[BOX]] {cuda_attr = #fir.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFsub7Ea"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>>)
+! CHECK: %[[ERR:.*]] = fir.alloca !fir.char<1,50> {bindc_name = "err", uniq_name = "_QFsub7Eerr"}
+! CHECK: %[[ERR_DECL:.*]]:2 = hlfir.declare %[[ERR]] typeparams %{{.*}} {uniq_name = "_QFsub7Eerr"} : (!fir.ref<!fir.char<1,50>>, index) -> (!fir.ref<!fir.char<1,50>>, !fir.ref<!fir.char<1,50>>)
+! CHECK: %[[ISTAT:.*]] = fir.alloca i32 {bindc_name = "istat", uniq_name = "_QFsub7Eistat"}
+! CHECK: %[[ISTAT_DECL:.*]]:2 = hlfir.declare %[[ISTAT]] {uniq_name = "_QFsub7Eistat"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[ERR_BOX:.*]] = fir.embox %[[ERR_DECL]]#1 : (!fir.ref<!fir.char<1,50>>) -> !fir.box<!fir.char<1,50>>
+! CHECK: fir.call @_FortranAAllocatableSetBounds
+! CHECK: %[[STAT:.*]] = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> errmsg(%[[ERR_BOX]] : !fir.box<!fir.char<1,50>>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>

From 19c6a7feca6e1558ef7cbe18efd2477c1126899d Mon Sep 17 00:00:00 2001
From: Alexandros Lamprineas <alexandros.lamprineas@arm.com>
Date: Wed, 17 Apr 2024 17:16:58 +0100
Subject: [PATCH 762/886] [FMV] Remove useless features according the latest
 ACLE spec. (#88965)

As explained in https://github.com/ARM-software/acle/pull/315 we
are deprecating features which aren't adding any value. These are:

sha1, pmull, dit, dgh, ebf16, sve-bf16, sve-ebf16, sve-i8mm,
sve2-pmull128, memtag2, memtag3, ssbs2, bti, ls64_v, ls64_accdata
---
 clang/test/CodeGen/aarch64-cpu-supports.c     |  24 +-
 .../CodeGen/aarch64-mixed-target-attributes.c |  12 +-
 .../test/CodeGen/attr-target-clones-aarch64.c |  77 ++---
 clang/test/CodeGen/attr-target-version.c      | 326 +++++++++---------
 .../CodeGenCXX/attr-target-clones-aarch64.cpp |  41 +--
 clang/test/CodeGenCXX/attr-target-version.cpp |  22 +-
 clang/test/Sema/aarch64-cpu-supports.c        |   2 +-
 clang/test/Sema/attr-target-clones-aarch64.c  |  21 +-
 clang/test/Sema/attr-target-version.c         |  18 +-
 clang/test/SemaCXX/attr-target-version.cpp    |  16 +-
 compiler-rt/lib/builtins/cpu_model/aarch64.c  |  17 +-
 .../builtins/cpu_model/aarch64/fmv/apple.inc  |   6 +-
 .../cpu_model/aarch64/fmv/fuchsia.inc         |   4 -
 .../builtins/cpu_model/aarch64/fmv/mrs.inc    |  42 +--
 .../llvm/TargetParser/AArch64TargetParser.h   |  36 +-
 15 files changed, 287 insertions(+), 377 deletions(-)

diff --git a/clang/test/CodeGen/aarch64-cpu-supports.c b/clang/test/CodeGen/aarch64-cpu-supports.c
index c54b7475a3fd5..7fad9724dfb6c 100644
--- a/clang/test/CodeGen/aarch64-cpu-supports.c
+++ b/clang/test/CodeGen/aarch64-cpu-supports.c
@@ -1,15 +1,17 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals --version 2
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -emit-llvm -o - %s | FileCheck %s
 
+//.
 // CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
+//.
 // CHECK-LABEL: define dso_local i32 @main
 // CHECK-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70368744177664
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70368744177664
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34359738368
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34359738368
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 // CHECK:       if.then:
@@ -17,8 +19,8 @@
 // CHECK-NEXT:    br label [[RETURN:%.*]]
 // CHECK:       if.end:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 9070970929152
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 9070970929152
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 17716740096
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 17716740096
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[IF_THEN1:%.*]], label [[IF_END2:%.*]]
 // CHECK:       if.then1:
@@ -26,8 +28,8 @@
 // CHECK-NEXT:    br label [[RETURN]]
 // CHECK:       if.end2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 166633186212708352
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 166633186212708352
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 5222680231936
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 5222680231936
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[IF_THEN3:%.*]], label [[IF_END4:%.*]]
 // CHECK:       if.then3:
@@ -49,10 +51,10 @@ int main(void) {
   if (__builtin_cpu_supports("sb"))
     return 1;
 
-  if (__builtin_cpu_supports("sve2-pmull128+memtag"))
+  if (__builtin_cpu_supports("sve2-aes+memtag"))
     return 2;
 
-  if (__builtin_cpu_supports("sme2+ls64_v+wfxt"))
+  if (__builtin_cpu_supports("sme2+ls64+wfxt"))
     return 3;
 
   if (__builtin_cpu_supports("avx2"))
@@ -60,3 +62,9 @@ int main(void) {
 
   return 0;
 }
+//.
+// CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+//.
+// CHECK: [[META0:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}
+// CHECK: [[META1:![0-9]+]] = !{!"{{.*}}clang version {{.*}}"}
+//.
diff --git a/clang/test/CodeGen/aarch64-mixed-target-attributes.c b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
index aef6ce36ab1c0..be290ff9ecee6 100644
--- a/clang/test/CodeGen/aarch64-mixed-target-attributes.c
+++ b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
@@ -69,8 +69,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -143,8 +143,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -210,8 +210,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
diff --git a/clang/test/CodeGen/attr-target-clones-aarch64.c b/clang/test/CodeGen/attr-target-clones-aarch64.c
index 8c8b951e9118d..c715001f6a722 100644
--- a/clang/test/CodeGen/attr-target-clones-aarch64.c
+++ b/clang/test/CodeGen/attr-target-clones-aarch64.c
@@ -3,7 +3,7 @@
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature -fmv -S -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK-NOFMV
 
 int __attribute__((target_clones("lse+aes", "sve2"))) ftc(void) { return 0; }
-int __attribute__((target_clones("sha2", "sha2+memtag2", " default "))) ftc_def(void) { return 1; }
+int __attribute__((target_clones("sha2", "sha2+memtag", " default "))) ftc_def(void) { return 1; }
 int __attribute__((target_clones("sha2", "default"))) ftc_dup1(void) { return 2; }
 int __attribute__((target_clones("fp", "crc+dotprod"))) ftc_dup2(void) { return 3; }
 int foo() {
@@ -12,7 +12,7 @@ int foo() {
 
 inline int __attribute__((target_clones("rng+simd", "rcpc+predres", "sve2-aes+wfxt"))) ftc_inline1(void) { return 1; }
 inline int __attribute__((target_clones("fp16", "fcma+sve2-bitperm", "default"))) ftc_inline2(void);
-inline int __attribute__((target_clones("bti", "sve+sb"))) ftc_inline3(void) { return 3; }
+inline int __attribute__((target_clones("mops", "sve+sb"))) ftc_inline3(void) { return 3; }
 
 int __attribute__((target_clones("default"))) ftc_direct(void) { return 4; }
 
@@ -56,16 +56,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16512
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16512
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8320
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8320
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc._MaesMlse
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 68719476736
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 68719476736
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 268435456
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 268435456
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -81,7 +81,7 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: @ftc_def._Mmemtag2Msha2(
+// CHECK-LABEL: @ftc_def._MmemtagMsha2(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
@@ -90,16 +90,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 17592186048512
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 17592186048512
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 17179871232
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 17179871232
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @ftc_def._Mmemtag2Msha2
+// CHECK-NEXT:    ret ptr @ftc_def._MmemtagMsha2
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 4096
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 4096
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2048
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2048
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -118,8 +118,8 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4096
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4096
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2048
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2048
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -198,16 +198,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 18014535948435456
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 18014535948435456
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 550292684800
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 550292684800
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline1._Msve2-aesMwfxt
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 140737492549632
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 140737492549632
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 68720001024
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 68720001024
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -228,16 +228,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 549757911040
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 549757911040
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1074003968
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1074003968
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline2._MfcmaMsve2-bitperm
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 65536
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 65536
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16384
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16384
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -250,20 +250,20 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70369817919488
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70369817919488
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34393292800
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34393292800
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline3._MsbMsve
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 1125899906842624
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 1125899906842624
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 17592186044416
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 17592186044416
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @ftc_inline3._Mbti
+// CHECK-NEXT:    ret ptr @ftc_inline3._Mmops
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    ret ptr @ftc_inline3.default
 //
@@ -329,7 +329,7 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: @ftc_inline3._Mbti(
+// CHECK-LABEL: @ftc_inline3._Mmops(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -407,17 +407,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK: attributes #[[ATTR0:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+lse,+neon" }
 // CHECK: attributes #[[ATTR1:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2" }
 // CHECK: attributes #[[ATTR2:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+sha2" }
-// CHECK: attributes #[[ATTR3:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+mte,+neon,+sha2" }
-// CHECK: attributes #[[ATTR4:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
-// CHECK: attributes #[[ATTR5:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+dotprod,+fp-armv8,+neon" }
-// CHECK: attributes #[[ATTR6:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
-// CHECK: attributes #[[ATTR7:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
-// CHECK: attributes #[[ATTR8:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+complxnum,+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-bitperm" }
-// CHECK: attributes #[[ATTR9:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rand" }
-// CHECK: attributes #[[ATTR10:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+predres,+rcpc" }
-// CHECK: attributes #[[ATTR11:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+wfxt" }
-// CHECK: attributes #[[ATTR12:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bti" }
-// CHECK: attributes #[[ATTR13:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sb,+sve" }
+// CHECK: attributes #[[ATTR3:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
+// CHECK: attributes #[[ATTR4:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+dotprod,+fp-armv8,+neon" }
+// CHECK: attributes #[[ATTR5:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+// CHECK: attributes #[[ATTR6:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
+// CHECK: attributes #[[ATTR7:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+complxnum,+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-bitperm" }
+// CHECK: attributes #[[ATTR8:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rand" }
+// CHECK: attributes #[[ATTR9:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+predres,+rcpc" }
+// CHECK: attributes #[[ATTR10:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+wfxt" }
+// CHECK: attributes #[[ATTR11:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops" }
+// CHECK: attributes #[[ATTR12:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sb,+sve" }
 //.
 // CHECK-NOFMV: attributes #[[ATTR0:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
 // CHECK-NOFMV: attributes #[[ATTR1:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
diff --git a/clang/test/CodeGen/attr-target-version.c b/clang/test/CodeGen/attr-target-version.c
index dd4cbbf5a8986..e71370e6d91df 100644
--- a/clang/test/CodeGen/attr-target-version.c
+++ b/clang/test/CodeGen/attr-target-version.c
@@ -5,11 +5,11 @@
 int __attribute__((target_version("rng+flagm+fp16fml"))) fmv(void) { return 1; }
 int __attribute__((target_version("flagm2+sme-i16i64"))) fmv(void) { return 2; }
 int __attribute__((target_version("lse+sha2"))) fmv(void) { return 3; }
-int __attribute__((target_version("dotprod+ls64_accdata"))) fmv(void) { return 4; }
+int __attribute__((target_version("dotprod+ls64"))) fmv(void) { return 4; }
 int __attribute__((target_version("fp16fml+memtag"))) fmv(void) { return 5; }
 int __attribute__((target_version("fp+aes"))) fmv(void) { return 6; }
-int __attribute__((target_version("crc+ls64_v"))) fmv(void) { return 7; }
-int __attribute__((target_version("bti"))) fmv(void) { return 8; }
+int __attribute__((target_version("crc+ls64"))) fmv(void) { return 7; }
+int __attribute__((target_version("mops"))) fmv(void) { return 8; }
 int __attribute__((target_version("sme2"))) fmv(void) { return 9; }
 int __attribute__((target_version("default"))) fmv(void);
 int __attribute__((target_version("ls64+simd"))) fmv_one(void) { return 1; }
@@ -17,25 +17,25 @@ int __attribute__((target_version("dpb"))) fmv_one(void) { return 2; }
 int __attribute__((target_version("default"))) fmv_one(void);
 int __attribute__((target_version("fp"))) fmv_two(void) { return 1; }
 int __attribute__((target_version("simd"))) fmv_two(void) { return 2; }
-int __attribute__((target_version("dgh"))) fmv_two(void) { return 3; }
+int __attribute__((target_version("frintts"))) fmv_two(void) { return 3; }
 int __attribute__((target_version("fp16+simd"))) fmv_two(void) { return 4; }
 int __attribute__((target_version("default"))) fmv_two(void);
 int foo() {
   return fmv()+fmv_one()+fmv_two();
 }
 
-inline int __attribute__((target_version("sha1+pmull+f64mm"))) fmv_inline(void) { return 1; }
+inline int __attribute__((target_version("crypto+f64mm"))) fmv_inline(void) { return 1; }
 inline int __attribute__((target_version("fp16+fcma+rdma+sme+ fp16 "))) fmv_inline(void) { return 2; }
 inline int __attribute__((target_version("sha3+i8mm+f32mm"))) fmv_inline(void) { return 12; }
-inline int __attribute__((target_version("dit+sve-ebf16"))) fmv_inline(void) { return 8; }
+inline int __attribute__((target_version("sme2+ssbs"))) fmv_inline(void) { return 8; }
 inline int __attribute__((target_version("dpb+rcpc2 "))) fmv_inline(void) { return 6; }
 inline int __attribute__((target_version(" dpb2 + jscvt"))) fmv_inline(void) { return 7; }
 inline int __attribute__((target_version("rcpc+frintts"))) fmv_inline(void) { return 3; }
-inline int __attribute__((target_version("sve+sve-bf16"))) fmv_inline(void) { return 4; }
+inline int __attribute__((target_version("sve+sme"))) fmv_inline(void) { return 4; }
 inline int __attribute__((target_version("sve2-aes+sve2-sha3"))) fmv_inline(void) { return 5; }
-inline int __attribute__((target_version("sve2+sve2-pmull128+sve2-bitperm"))) fmv_inline(void) { return 9; }
-inline int __attribute__((target_version("sve2-sm4+memtag2"))) fmv_inline(void) { return 10; }
-inline int __attribute__((target_version("memtag3+rcpc3+mops"))) fmv_inline(void) { return 11; }
+inline int __attribute__((target_version("sve2+sve2-aes+sve2-bitperm"))) fmv_inline(void) { return 9; }
+inline int __attribute__((target_version("sve2-sm4+memtag"))) fmv_inline(void) { return 10; }
+inline int __attribute__((target_version("memtag+rcpc3+mops"))) fmv_inline(void) { return 11; }
 inline int __attribute__((target_version("aes+dotprod"))) fmv_inline(void) { return 13; }
 inline int __attribute__((target_version("simd+fp16fml"))) fmv_inline(void) { return 14; }
 inline int __attribute__((target_version("fp+sm4"))) fmv_inline(void) { return 15; }
@@ -186,7 +186,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._MdotprodMls64_accdata
+// CHECK-LABEL: define {{[^@]+}}@fmv._MdotprodMls64
 // CHECK-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 4
@@ -207,14 +207,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._McrcMls64_v
+// CHECK-LABEL: define {{[^@]+}}@fmv._McrcMls64
 // CHECK-SAME: () #[[ATTR6:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 7
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._Mbti
+// CHECK-LABEL: define {{[^@]+}}@fmv._Mmops
 // CHECK-SAME: () #[[ATTR7:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 8
@@ -256,7 +256,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_two._Mdgh
+// CHECK-LABEL: define {{[^@]+}}@fmv_two._Mfrintts
 // CHECK-SAME: () #[[ATTR11:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
@@ -271,7 +271,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@foo
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @fmv()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @fmv_one()
@@ -293,68 +293,68 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:    ret ptr @fmv._MflagmMfp16fmlMrng
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 72057594037927940
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 72057594037927940
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2199023255556
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2199023255556
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
 // CHECK-NEXT:    ret ptr @fmv._Mflagm2Msme-i16i64
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 9007199254741008
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 9007199254741008
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 274877906960
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 274877906960
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[RESOLVER_RETURN3:%.*]], label [[RESOLVER_ELSE4:%.*]]
 // CHECK:       resolver_return3:
-// CHECK-NEXT:    ret ptr @fmv._MdotprodMls64_accdata
+// CHECK-NEXT:    ret ptr @fmv._MdotprodMls64
 // CHECK:       resolver_else4:
 // CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 4503599627371520
-// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 4503599627371520
+// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 274877907968
+// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 274877907968
 // CHECK-NEXT:    [[TMP15:%.*]] = and i1 true, [[TMP14]]
 // CHECK-NEXT:    br i1 [[TMP15]], label [[RESOLVER_RETURN5:%.*]], label [[RESOLVER_ELSE6:%.*]]
 // CHECK:       resolver_return5:
-// CHECK-NEXT:    ret ptr @fmv._McrcMls64_v
+// CHECK-NEXT:    ret ptr @fmv._McrcMls64
 // CHECK:       resolver_else6:
 // CHECK-NEXT:    [[TMP16:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 8796093022216
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 8796093022216
+// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 17179869192
+// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 17179869192
 // CHECK-NEXT:    [[TMP19:%.*]] = and i1 true, [[TMP18]]
 // CHECK-NEXT:    br i1 [[TMP19]], label [[RESOLVER_RETURN7:%.*]], label [[RESOLVER_ELSE8:%.*]]
 // CHECK:       resolver_return7:
 // CHECK-NEXT:    ret ptr @fmv._Mfp16fmlMmemtag
 // CHECK:       resolver_else8:
 // CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 16640
-// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 16640
+// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 8448
+// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 8448
 // CHECK-NEXT:    [[TMP23:%.*]] = and i1 true, [[TMP22]]
 // CHECK-NEXT:    br i1 [[TMP23]], label [[RESOLVER_RETURN9:%.*]], label [[RESOLVER_ELSE10:%.*]]
 // CHECK:       resolver_return9:
 // CHECK-NEXT:    ret ptr @fmv._MaesMfp
 // CHECK:       resolver_else10:
 // CHECK-NEXT:    [[TMP24:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 4224
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 4224
+// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 2176
+// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 2176
 // CHECK-NEXT:    [[TMP27:%.*]] = and i1 true, [[TMP26]]
 // CHECK-NEXT:    br i1 [[TMP27]], label [[RESOLVER_RETURN11:%.*]], label [[RESOLVER_ELSE12:%.*]]
 // CHECK:       resolver_return11:
 // CHECK-NEXT:    ret ptr @fmv._MlseMsha2
 // CHECK:       resolver_else12:
 // CHECK-NEXT:    [[TMP28:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 144115188075855872
-// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 144115188075855872
+// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 17592186044416
+// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 17592186044416
 // CHECK-NEXT:    [[TMP31:%.*]] = and i1 true, [[TMP30]]
 // CHECK-NEXT:    br i1 [[TMP31]], label [[RESOLVER_RETURN13:%.*]], label [[RESOLVER_ELSE14:%.*]]
 // CHECK:       resolver_return13:
-// CHECK-NEXT:    ret ptr @fmv._Msme2
+// CHECK-NEXT:    ret ptr @fmv._Mmops
 // CHECK:       resolver_else14:
 // CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 1125899906842624
-// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 1125899906842624
+// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 4398046511104
+// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 4398046511104
 // CHECK-NEXT:    [[TMP35:%.*]] = and i1 true, [[TMP34]]
 // CHECK-NEXT:    br i1 [[TMP35]], label [[RESOLVER_RETURN15:%.*]], label [[RESOLVER_ELSE16:%.*]]
 // CHECK:       resolver_return15:
-// CHECK-NEXT:    ret ptr @fmv._Mbti
+// CHECK-NEXT:    ret ptr @fmv._Msme2
 // CHECK:       resolver_else16:
 // CHECK-NEXT:    ret ptr @fmv.default
 //
@@ -363,16 +363,16 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2251799813685760
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2251799813685760
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877907456
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877907456
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_one._Mls64Msimd
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 262144
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 262144
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 32768
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 32768
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -385,20 +385,20 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 66048
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 66048
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16896
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16896
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_two._Mfp16Msimd
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 33554432
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 33554432
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @fmv_two._Mdgh
+// CHECK-NEXT:    ret ptr @fmv_two._Mfrintts
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
 // CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 512
@@ -421,49 +421,49 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_e.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 20
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_d._Msb
-// CHECK-SAME: () #[[ATTR13:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR14:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_d.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 111
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_c._Mssbs
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_c.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@goo
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @fmv_inline()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @fmv_e()
@@ -477,96 +477,96 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4398048673856
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4398048673856
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8590213184
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8590213184
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_inline._MfcmaMfp16MrdmMsme
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 864726312827224064
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 864726312827224064
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 26405458935808
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 26405458935808
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @fmv_inline._Mmemtag3MmopsMrcpc3
+// CHECK-NEXT:    ret ptr @fmv_inline._MmemtagMmopsMrcpc3
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 893353197568
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 893353197568
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 1879048192
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 1879048192
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[RESOLVER_RETURN3:%.*]], label [[RESOLVER_ELSE4:%.*]]
 // CHECK:       resolver_return3:
-// CHECK-NEXT:    ret ptr @fmv_inline._Msve2Msve2-bitpermMsve2-pmull128
+// CHECK-NEXT:    ret ptr @fmv_inline._Msve2Msve2-aesMsve2-bitperm
 // CHECK:       resolver_else4:
 // CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 34359773184
-// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 34359773184
+// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 71307264
+// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 71307264
 // CHECK-NEXT:    [[TMP15:%.*]] = and i1 true, [[TMP14]]
 // CHECK-NEXT:    br i1 [[TMP15]], label [[RESOLVER_RETURN5:%.*]], label [[RESOLVER_ELSE6:%.*]]
 // CHECK:       resolver_return5:
-// CHECK-NEXT:    ret ptr @fmv_inline._Mf64mmMpmullMsha1
+// CHECK-NEXT:    ret ptr @fmv_inline._Mf32mmMi8mmMsha3
 // CHECK:       resolver_else6:
 // CHECK-NEXT:    [[TMP16:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 17246986240
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 17246986240
+// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 4535485464576
+// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 4535485464576
 // CHECK-NEXT:    [[TMP19:%.*]] = and i1 true, [[TMP18]]
 // CHECK-NEXT:    br i1 [[TMP19]], label [[RESOLVER_RETURN7:%.*]], label [[RESOLVER_ELSE8:%.*]]
 // CHECK:       resolver_return7:
-// CHECK-NEXT:    ret ptr @fmv_inline._Mf32mmMi8mmMsha3
+// CHECK-NEXT:    ret ptr @fmv_inline._Msme2Mssbs
 // CHECK:       resolver_else8:
 // CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 19791209299968
-// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 19791209299968
+// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 21474836480
+// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 21474836480
 // CHECK-NEXT:    [[TMP23:%.*]] = and i1 true, [[TMP22]]
 // CHECK-NEXT:    br i1 [[TMP23]], label [[RESOLVER_RETURN9:%.*]], label [[RESOLVER_ELSE10:%.*]]
 // CHECK:       resolver_return9:
-// CHECK-NEXT:    ret ptr @fmv_inline._Mmemtag2Msve2-sm4
+// CHECK-NEXT:    ret ptr @fmv_inline._MmemtagMsve2-sm4
 // CHECK:       resolver_else10:
 // CHECK-NEXT:    [[TMP24:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 1236950581248
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 1236950581248
+// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 8623489024
+// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 8623489024
 // CHECK-NEXT:    [[TMP27:%.*]] = and i1 true, [[TMP26]]
 // CHECK-NEXT:    br i1 [[TMP27]], label [[RESOLVER_RETURN11:%.*]], label [[RESOLVER_ELSE12:%.*]]
 // CHECK:       resolver_return11:
-// CHECK-NEXT:    ret ptr @fmv_inline._Msve2-aesMsve2-sha3
+// CHECK-NEXT:    ret ptr @fmv_inline._MsmeMsve
 // CHECK:       resolver_else12:
 // CHECK-NEXT:    [[TMP28:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 4295098368
-// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 4295098368
+// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 2684354560
+// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 2684354560
 // CHECK-NEXT:    [[TMP31:%.*]] = and i1 true, [[TMP30]]
 // CHECK-NEXT:    br i1 [[TMP31]], label [[RESOLVER_RETURN13:%.*]], label [[RESOLVER_ELSE14:%.*]]
 // CHECK:       resolver_return13:
-// CHECK-NEXT:    ret ptr @fmv_inline._MditMsve-ebf16
+// CHECK-NEXT:    ret ptr @fmv_inline._Msve2-aesMsve2-sha3
 // CHECK:       resolver_else14:
 // CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 3221225472
-// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 3221225472
+// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 281475110928384
+// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 281475110928384
 // CHECK-NEXT:    [[TMP35:%.*]] = and i1 true, [[TMP34]]
 // CHECK-NEXT:    br i1 [[TMP35]], label [[RESOLVER_RETURN15:%.*]], label [[RESOLVER_ELSE16:%.*]]
 // CHECK:       resolver_return15:
-// CHECK-NEXT:    ret ptr @fmv_inline._MsveMsve-bf16
+// CHECK-NEXT:    ret ptr @fmv_inline._McryptoMf64mm
 // CHECK:       resolver_else16:
 // CHECK-NEXT:    [[TMP36:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP37:%.*]] = and i64 [[TMP36]], 20971520
-// CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i64 [[TMP37]], 20971520
+// CHECK-NEXT:    [[TMP37:%.*]] = and i64 [[TMP36]], 2621440
+// CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i64 [[TMP37]], 2621440
 // CHECK-NEXT:    [[TMP39:%.*]] = and i1 true, [[TMP38]]
 // CHECK-NEXT:    br i1 [[TMP39]], label [[RESOLVER_RETURN17:%.*]], label [[RESOLVER_ELSE18:%.*]]
 // CHECK:       resolver_return17:
 // CHECK-NEXT:    ret ptr @fmv_inline._MfrinttsMrcpc
 // CHECK:       resolver_else18:
 // CHECK-NEXT:    [[TMP40:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP41:%.*]] = and i64 [[TMP40]], 8650752
-// CHECK-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[TMP41]], 8650752
+// CHECK-NEXT:    [[TMP41:%.*]] = and i64 [[TMP40]], 1081344
+// CHECK-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[TMP41]], 1081344
 // CHECK-NEXT:    [[TMP43:%.*]] = and i1 true, [[TMP42]]
 // CHECK-NEXT:    br i1 [[TMP43]], label [[RESOLVER_RETURN19:%.*]], label [[RESOLVER_ELSE20:%.*]]
 // CHECK:       resolver_return19:
 // CHECK-NEXT:    ret ptr @fmv_inline._MdpbMrcpc2
 // CHECK:       resolver_else20:
 // CHECK-NEXT:    [[TMP44:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP45:%.*]] = and i64 [[TMP44]], 1572864
-// CHECK-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[TMP45]], 1572864
+// CHECK-NEXT:    [[TMP45:%.*]] = and i64 [[TMP44]], 196608
+// CHECK-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[TMP45]], 196608
 // CHECK-NEXT:    [[TMP47:%.*]] = and i1 true, [[TMP46]]
 // CHECK-NEXT:    br i1 [[TMP47]], label [[RESOLVER_RETURN21:%.*]], label [[RESOLVER_ELSE22:%.*]]
 // CHECK:       resolver_return21:
@@ -581,8 +581,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:    ret ptr @fmv_inline._Mfp16fmlMsimd
 // CHECK:       resolver_else24:
 // CHECK-NEXT:    [[TMP52:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP53:%.*]] = and i64 [[TMP52]], 16400
-// CHECK-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[TMP53]], 16400
+// CHECK-NEXT:    [[TMP53:%.*]] = and i64 [[TMP52]], 8208
+// CHECK-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[TMP53]], 8208
 // CHECK-NEXT:    [[TMP55:%.*]] = and i1 true, [[TMP54]]
 // CHECK-NEXT:    br i1 [[TMP55]], label [[RESOLVER_RETURN25:%.*]], label [[RESOLVER_ELSE26:%.*]]
 // CHECK:       resolver_return25:
@@ -611,8 +611,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2251799813685248
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2251799813685248
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877906944
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877906944
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -625,8 +625,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70368744177664
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70368744177664
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34359738368
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34359738368
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -639,8 +639,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 281474976710656
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 281474976710656
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 137438953472
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 137438953472
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -651,7 +651,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@recur
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    call void @reca()
 // CHECK-NEXT:    ret void
@@ -659,7 +659,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@main
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 0, ptr [[RETVAL]], align 4
@@ -670,7 +670,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@hoo
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FP1:%.*]] = alloca ptr, align 8
 // CHECK-NEXT:    [[FP2:%.*]] = alloca ptr, align 8
@@ -687,14 +687,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_forward_default_decl._Mmops
-// CHECK-SAME: () #[[ATTR14:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR7]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_extern_forward_default_decl._Mdotprod
-// CHECK-SAME: () #[[ATTR15:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR3]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
@@ -708,14 +708,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_default_def._Msve
-// CHECK-SAME: () #[[ATTR16:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR15:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_default_def.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
@@ -729,56 +729,56 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_default_def.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_forward_default_def.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_forward_default_def._Mlse
-// CHECK-SAME: () #[[ATTR17:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR16:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_without_default._Mrdm
-// CHECK-SAME: () #[[ATTR18:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR17:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@default_def_with_version_decls.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@used_def_without_default_decl._Mjscvt
-// CHECK-SAME: () #[[ATTR21:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR20:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@used_def_without_default_decl._Mrdm
-// CHECK-SAME: () #[[ATTR18]] {
+// CHECK-SAME: () #[[ATTR17]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@caller
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @used_def_without_default_decl()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @used_decl_without_default_decl()
@@ -790,8 +790,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -812,8 +812,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -831,92 +831,92 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mf64mmMpmullMsha1
-// CHECK-SAME: () #[[ATTR22:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._McryptoMf64mm
+// CHECK-SAME: () #[[ATTR21:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfcmaMfp16MrdmMsme
-// CHECK-SAME: () #[[ATTR23:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR22:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mf32mmMi8mmMsha3
-// CHECK-SAME: () #[[ATTR24:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR23:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 12
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MditMsve-ebf16
-// CHECK-SAME: () #[[ATTR25:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msme2Mssbs
+// CHECK-SAME: () #[[ATTR8]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 8
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MdpbMrcpc2
-// CHECK-SAME: () #[[ATTR26:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR24:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 6
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mdpb2Mjscvt
-// CHECK-SAME: () #[[ATTR27:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR25:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 7
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfrinttsMrcpc
-// CHECK-SAME: () #[[ATTR28:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR26:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MsveMsve-bf16
-// CHECK-SAME: () #[[ATTR29:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MsmeMsve
+// CHECK-SAME: () #[[ATTR27:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 4
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2-aesMsve2-sha3
-// CHECK-SAME: () #[[ATTR30:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR28:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 5
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2Msve2-bitpermMsve2-pmull128
-// CHECK-SAME: () #[[ATTR31:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2Msve2-aesMsve2-bitperm
+// CHECK-SAME: () #[[ATTR29:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 9
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mmemtag2Msve2-sm4
-// CHECK-SAME: () #[[ATTR32:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MmemtagMsve2-sm4
+// CHECK-SAME: () #[[ATTR30:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 10
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mmemtag3MmopsMrcpc3
-// CHECK-SAME: () #[[ATTR33:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MmemtagMmopsMrcpc3
+// CHECK-SAME: () #[[ATTR31:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 11
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MaesMdotprod
-// CHECK-SAME: () #[[ATTR15]] {
+// CHECK-SAME: () #[[ATTR3]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 13
 //
@@ -930,21 +930,21 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfpMsm4
-// CHECK-SAME: () #[[ATTR34:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR32:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 15
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MlseMrdm
-// CHECK-SAME: () #[[ATTR35:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR33:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 16
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline.default
-// CHECK-SAME: () #[[ATTR11]] {
+// CHECK-SAME: () #[[ATTR13]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -953,8 +953,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1073741824
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1073741824
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 33554432
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 33554432
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -967,8 +967,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 65536
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 65536
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16384
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16384
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -995,8 +995,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -1132,39 +1132,37 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+flagm,+fp16fml,+fullfp16,+neon,+rand,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR1]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+altnzcv,+bf16,+flagm,+sme,+sme-i16i64,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR2]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+sha2,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+ls64,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp16fml,+fullfp16,+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR5]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR6]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bti,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR8]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme,+sme2,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR9:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR10]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR11]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR11]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR12]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR13]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+sb,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR14]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR17]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR18]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR20:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fullfp16,+neon,+rdm,+sme,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+dit,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR26]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR28]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+mte,+neon,+sve,+sve2,+sve2-sm4,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR33]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,+mte,+rcpc,+rcpc3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR34]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+sm4,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR35]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR13]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR14]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+sb,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR17]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR18:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR20]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fullfp16,+neon,+sha2,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fullfp16,+neon,+rdm,+sme,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,+rcpc,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR26]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,+rcpc,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fullfp16,+neon,+sme,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR28]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-sm4,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,+rcpc,+rcpc3,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+sm4,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR33]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+rdm,-fp-armv8,-v9.5a" }
 //.
 // CHECK-NOFMV: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
 // CHECK-NOFMV: attributes #[[ATTR1:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
diff --git a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
index 7953f902bf09b..d439f96982eb4 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
@@ -1,8 +1,8 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals --include-generated-funcs
 // RUN: %clang_cc1 -std=c++11 -triple aarch64-linux-gnu -emit-llvm %s -o - | FileCheck %s
 
-int __attribute__((target_clones("ls64_v+fp16", "default"))) foo_ovl(int) { return 1; }
-int __attribute__((target_clones("ls64_accdata+ls64"))) foo_ovl(void) { return 2; }
+int __attribute__((target_clones("ls64+fp16", "default"))) foo_ovl(int) { return 1; }
+int __attribute__((target_clones("sme+ls64"))) foo_ovl(void) { return 2; }
 
 int bar() {
   return foo_ovl(1) + foo_ovl();
@@ -35,9 +35,6 @@ void run_foo_tml() {
   Mc4.foo_tml();
 }
 
-
-
-
 //.
 // CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
 // CHECK: @_Z7foo_ovli.ifunc = weak_odr alias i32 (i32), ptr @_Z7foo_ovli
@@ -49,7 +46,7 @@ void run_foo_tml() {
 // CHECK: @_ZN7MyClassIssE7foo_tmlEv = weak_odr ifunc i32 (ptr), ptr @_ZN7MyClassIssE7foo_tmlEv.resolver
 // CHECK: @_ZN7MyClassIisE7foo_tmlEv = weak_odr ifunc i32 (ptr), ptr @_ZN7MyClassIisE7foo_tmlEv.resolver
 //.
-// CHECK-LABEL: @_Z7foo_ovli._Mfp16Mls64_v(
+// CHECK-LABEL: @_Z7foo_ovli._Mfp16Mls64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 [[TMP0:%.*]], ptr [[DOTADDR]], align 4
@@ -60,17 +57,17 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4503599627436032
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4503599627436032
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877923328
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877923328
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z7foo_ovli._Mfp16Mls64_v
+// CHECK-NEXT:    ret ptr @_Z7foo_ovli._Mfp16Mls64
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z7foo_ovli.default
 //
 //
-// CHECK-LABEL: @_Z7foo_ovlv._Mls64Mls64_accdata(
+// CHECK-LABEL: @_Z7foo_ovlv._Mls64Msme(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
@@ -79,12 +76,12 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 11258999068426240
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 11258999068426240
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 283467841536
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 283467841536
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z7foo_ovlv._Mls64Mls64_accdata
+// CHECK-NEXT:    ret ptr @_Z7foo_ovlv._Mls64Msme
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z7foo_ovlv.default
 //
@@ -114,16 +111,16 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36310271995674624
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36310271995674624
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1236950581248
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1236950581248
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @_ZN7MyClassIssE7foo_tmlEv._Msme-f64f64Mssbs
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16777216
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16777216
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -136,16 +133,16 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36310271995674624
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36310271995674624
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1236950581248
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1236950581248
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @_ZN7MyClassIisE7foo_tmlEv._Msme-f64f64Mssbs
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16777216
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16777216
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -231,7 +228,7 @@ void run_foo_tml() {
 //
 //.
 // CHECK: attributes #[[ATTR0:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
-// CHECK: attributes #[[ATTR1:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ls64" }
+// CHECK: attributes #[[ATTR1:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme" }
 // CHECK: attributes #[[ATTR2:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // CHECK: attributes #[[ATTR3:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint" }
 // CHECK: attributes #[[ATTR4:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme,+sme-f64f64" }
diff --git a/clang/test/CodeGenCXX/attr-target-version.cpp b/clang/test/CodeGenCXX/attr-target-version.cpp
index 8b7273fe3bb51..24433b8f20b7d 100644
--- a/clang/test/CodeGenCXX/attr-target-version.cpp
+++ b/clang/test/CodeGenCXX/attr-target-version.cpp
@@ -3,7 +3,7 @@
 
 int __attribute__((target_version("sme-f64f64+bf16"))) foo(int) { return 1; }
 int __attribute__((target_version("default"))) foo(int) { return 2; }
-int __attribute__((target_version("sm4+ebf16"))) foo(void) { return 3; }
+int __attribute__((target_version("sm4+bf16"))) foo(void) { return 3; }
 int __attribute__((target_version("default"))) foo(void) { return 4; }
 
 struct MyClass {
@@ -89,7 +89,7 @@ int bar() {
 // CHECK-NEXT:    ret i32 2
 //
 //
-// CHECK-LABEL: @_Z3foov._Mebf16Msm4(
+// CHECK-LABEL: @_Z3foov._Mbf16Msm4(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -246,8 +246,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36028797153181696
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36028797153181696
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1099520016384
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1099520016384
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -260,12 +260,12 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 268435488
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 268435488
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8388640
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8388640
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z3foov._Mebf16Msm4
+// CHECK-NEXT:    ret ptr @_Z3foov._Mbf16Msm4
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z3foov.default
 //
@@ -274,8 +274,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1073741824
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1073741824
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 33554432
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 33554432
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -288,8 +288,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 65536
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 65536
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16384
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16384
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
diff --git a/clang/test/Sema/aarch64-cpu-supports.c b/clang/test/Sema/aarch64-cpu-supports.c
index ddeed7c5bc9e9..cf8e043dfb1e3 100644
--- a/clang/test/Sema/aarch64-cpu-supports.c
+++ b/clang/test/Sema/aarch64-cpu-supports.c
@@ -20,7 +20,7 @@ int test_aarch64_features(void) {
   // expected-warning@+1 {{invalid cpu feature string}}
   if (__builtin_cpu_supports("default"))
     return 6;
-  if (__builtin_cpu_supports(" ssbs + bti "))
+  if (__builtin_cpu_supports(" ssbs + crc "))
     return 7;
   return 0;
 }
diff --git a/clang/test/Sema/attr-target-clones-aarch64.c b/clang/test/Sema/attr-target-clones-aarch64.c
index bc3fceab82825..511e8a9411804 100644
--- a/clang/test/Sema/attr-target-clones-aarch64.c
+++ b/clang/test/Sema/attr-target-clones-aarch64.c
@@ -9,7 +9,7 @@ void __attribute__((target_clones("ssbs+ls64"))) warn2(void);
 
 // expected-error@+2 {{'target_clones' and 'target_version' attributes are not compatible}}
 // expected-note@+1 {{conflicting attribute is here}}
-void __attribute__((target_version("sve-bf16"), target_clones("sme+memtag"))) not_compat(void);
+void __attribute__((target_version("bf16"), target_clones("sme+memtag"))) not_compat(void);
 
 int redecl(void);
 int __attribute__((target_clones("frintts", "simd+fp", "default"))) redecl(void) { return 1; }
@@ -21,26 +21,25 @@ int __attribute__((target_clones("sve+dotprod"))) redecl3(void);
 int redecl3(void);
 
 int __attribute__((target_clones("rng", "fp16fml+fp", "default"))) redecl4(void);
-// expected-error@+3 {{'target_clones' attribute does not match previous declaration}}
+// expected-error@+2 {{'target_clones' attribute does not match previous declaration}}
 // expected-note@-2 {{previous declaration is here}}
-// expected-warning@+1 {{version list contains entries that don't impact code generation}}
-int __attribute__((target_clones("dgh+memtag+rpres+ls64_v", "ebf16+dpb+sha1", "default"))) redecl4(void) { return 1; }
+int __attribute__((target_clones("dpb2+memtag+rpres+ls64", "bf16+dpb+sha2", "default"))) redecl4(void) { return 1; }
 
 int __attribute__((target_version("flagm2"))) redef2(void) { return 1; }
 // expected-error@+2 {{multiversioned function redeclarations require identical target attributes}}
 // expected-note@-2 {{previous declaration is here}}
 int __attribute__((target_clones("flagm2", "default"))) redef2(void) { return 1; }
 
-int __attribute__((target_clones("f32mm", "f64mm", "sha1+fp"))) redef3(void) { return 1; }
+int __attribute__((target_clones("f32mm", "f64mm", "sha3+fp"))) redef3(void) { return 1; }
 // expected-error@+2 {{'target_clones' attribute does not match previous declaration}}
 // expected-note@-2 {{previous declaration is here}}
-int __attribute__((target_clones("f32mm", "sha1+fp", "f64mm"))) redef3(void) { return 1; }
+int __attribute__((target_clones("f32mm", "sha3+fp", "f64mm"))) redef3(void) { return 1; }
 
 int __attribute__((target_clones("rdm+lse+rdm", "lse+rdm"))) dup1(void) { return 1; }
 // expected-warning@+1 {{version list contains duplicate entries}}
 int __attribute__((target_clones("rdm+lse+rdm", "rdm+lse+rdm"))) dup2(void) { return 2; }
 // expected-warning@+1 {{version list contains duplicate entries}}
-int __attribute__((target_clones("rcpc2+sve2-pmull128", "rcpc2+sve2-pmull128"))) dup3(void) { return 3; }
+int __attribute__((target_clones("rcpc2+sve2-aes", "rcpc2+sve2-aes"))) dup3(void) { return 3; }
 // expected-warning@+1 {{version list contains duplicate entries}}
 void __attribute__((target_clones("sha3", "default", "default"))) dup4(void);
 // expected-warning@+2 {{version list contains duplicate entries}}
@@ -49,7 +48,7 @@ int __attribute__((target_clones("fp", "fp", "crc+dotprod", "dotprod+crc"))) dup
 
 // expected-warning@+1 {{version list contains duplicate entries}}
 int __attribute__((target_clones("fp16+memtag", "memtag+fp16"))) dup6(void) { return 6; }
-int __attribute__((target_clones("simd+ssbs2", "simd+dpb2"))) dup7(void) { return 7; }
+int __attribute__((target_clones("simd+ssbs", "simd+dpb2"))) dup7(void) { return 7; }
 
 // expected-warning@+1 {{unsupported '' in the 'target_clones' attribute string;}}
 void __attribute__((target_clones(""))) empty_target_1(void);
@@ -72,13 +71,13 @@ empty_target_5(void);
 void __attribute__((target_clones("sve2-bitperm", "sve2-bitperm")))
 dupe_normal(void);
 
-void __attribute__((target_clones("default"), target_clones("memtag3+bti"))) dupe_normal2(void);
+void __attribute__((target_clones("default"), target_clones("memtag+mops"))) dupe_normal2(void);
 
 int mv_after_use(void);
 int useage(void) {
   return mv_after_use();
 }
 // expected-error@+1 {{function declaration cannot become a multiversioned function after first usage}}
-int __attribute__((target_clones("sve2-sha3+ssbs2", "sm4"))) mv_after_use(void) { return 1; }
+int __attribute__((target_clones("sve2-sha3+ssbs", "sm4"))) mv_after_use(void) { return 1; }
 // expected-error@+1 {{'main' cannot be a multiversioned function}}
-int __attribute__((target_clones("sve-i8mm"))) main() { return 1; }
+int __attribute__((target_clones("sve2-aes"))) main() { return 1; }
diff --git a/clang/test/Sema/attr-target-version.c b/clang/test/Sema/attr-target-version.c
index cd5be459456eb..4ff8e1eee619e 100644
--- a/clang/test/Sema/attr-target-version.c
+++ b/clang/test/Sema/attr-target-version.c
@@ -16,7 +16,7 @@ int __attribute__((target_version("aes"))) foo(void) { return 1; }
 int __attribute__((target_version("default"))) foo(void) { return 2; }
 
 //expected-note@+1 {{previous definition is here}}
-int __attribute__((target_version("sha3 + pmull "))) foo(void) { return 1; }
+int __attribute__((target_version("sha3 + aes "))) foo(void) { return 1; }
 //expected-note@-1 {{previous definition is here}}
 
 //expected-error@+1 {{redefinition of 'foo'}}
@@ -32,11 +32,11 @@ __attribute__ ((target("bf16,sve,sve2,dotprod"))) int func(void) { return 1; }
 __attribute__ ((target("default"))) int func(void) { return 0; }
 
 //expected-note@+1 {{previous declaration is here}}
-void __attribute__((target_version("bti+flagm2"))) one(void) {}
+void __attribute__((target_version("mops+flagm2"))) one(void) {}
 //expected-error@+1 {{multiversioned function redeclarations require identical target attributes}}
-void __attribute__((target_version("flagm2+bti"))) one(void) {}
+void __attribute__((target_version("flagm2+mops"))) one(void) {}
 
-void __attribute__((target_version("ssbs+sha1"))) two(void) {}
+void __attribute__((target_version("ssbs+sha2"))) two(void) {}
 void __attribute__((target_version("ssbs+fp16fml"))) two(void) {}
 
 //expected-error@+1 {{'main' cannot be a multiversioned function}}
@@ -44,7 +44,7 @@ int __attribute__((target_version("lse"))) main(void) { return 1; }
 
 // It is ok for the default version to appear first.
 int default_first(void) { return 1; }
-int __attribute__((target_version("dit"))) default_first(void) { return 2; }
+int __attribute__((target_version("crc"))) default_first(void) { return 2; }
 int __attribute__((target_version("mops"))) default_first(void) { return 3; }
 
 // It is ok if the default version is between other versions.
@@ -77,7 +77,7 @@ void __attribute__((target_version("rdm+rng+crc"))) redef(void) {}
 void __attribute__((target_version("rdm+rng+crc"))) redef(void) {}
 
 int def(void);
-void __attribute__((target_version("dit"))) nodef(void);
+void __attribute__((target_version("crc"))) nodef(void);
 void __attribute__((target_version("ls64"))) nodef(void);
 void __attribute__((target_version("aes"))) ovl(void);
 void __attribute__((target_version("default"))) ovl(void);
@@ -89,12 +89,12 @@ int bar() {
   return def();
 }
 // expected-error@+1 {{function declaration cannot become a multiversioned function after first usage}}
-int __attribute__((target_version("sha1"))) def(void) { return 1; }
+int __attribute__((target_version("sha3"))) def(void) { return 1; }
 
 int __attribute__((target_version("sve"))) prot();
 // expected-error@-1 {{multiversioned function must have a prototype}}
 
-int __attribute__((target_version("pmull"))) rtype(int);
+int __attribute__((target_version("aes"))) rtype(int);
 // expected-error@+1 {{multiversioned function declaration has a different return type}}
 float __attribute__((target_version("rdm"))) rtype(int);
 
@@ -102,7 +102,7 @@ int __attribute__((target_version("sha2"))) combine(void) { return 1; }
 // expected-error@+1 {{multiversioned function declaration has a different calling convention}}
 int __attribute__((aarch64_vector_pcs, target_version("sha3"))) combine(void) { return 2; }
 
-int __attribute__((target_version("fp+aes+pmull+rcpc"))) unspec_args() { return -1; }
+int __attribute__((target_version("fp+aes+sha3+rcpc"))) unspec_args() { return -1; }
 // expected-error@-1 {{multiversioned function must have a prototype}}
 // expected-error@+1 {{multiversioned function must have a prototype}}
 int __attribute__((target_version("default"))) unspec_args() { return 0; }
diff --git a/clang/test/SemaCXX/attr-target-version.cpp b/clang/test/SemaCXX/attr-target-version.cpp
index b3385f043590f..ae2923cc303c2 100644
--- a/clang/test/SemaCXX/attr-target-version.cpp
+++ b/clang/test/SemaCXX/attr-target-version.cpp
@@ -25,13 +25,13 @@ int __attribute__((target_version("dpb"))) diff_link(void);
 
 int __attribute__((target_version("memtag"))) diff_link1(void) { return 1; }
 //expected-error@+1 {{multiversioned function declaration has a different linkage}}
-static int __attribute__((target_version("bti"))) diff_link1(void);
+static int __attribute__((target_version("mops"))) diff_link1(void);
 
 int __attribute__((target_version("flagm2"))) diff_link2(void) { return 1; }
 extern int __attribute__((target_version("flagm"))) diff_link2(void);
 
 namespace {
-static int __attribute__((target_version("memtag3"))) diff_link2(void) { return 2; }
+static int __attribute__((target_version("memtag"))) diff_link2(void) { return 2; }
 int __attribute__((target_version("sve2-bitperm"))) diff_link2(void) { return 1; }
 } // namespace
 
@@ -49,7 +49,7 @@ double __attribute__((target_version("rcpc"))) diff_type1(void);
 
 auto __attribute__((target_version("rcpc2"))) diff_type2(void) -> int { return 1; }
 //expected-error@+1 {{multiversioned function declaration has a different return type}}
-auto __attribute__((target_version("sve-bf16"))) diff_type2(void) -> long { return (long)1; }
+auto __attribute__((target_version("sve2-aes"))) diff_type2(void) -> long { return (long)1; }
 
 int __attribute__((target_version("fp16fml"))) diff_type3(void) noexcept(false) { return 1; }
 //expected-error@+2 {{exception specification in declaration does not match previous declaration}}
@@ -75,7 +75,7 @@ auto __attribute__((target_version("dpb2"))) ret3(void) -> int { return 1; }
 class Cls {
   __attribute__((target_version("rng"))) Cls();
   // expected-error@-1 {{attribute 'target_version' multiversioned functions do not yet support constructors}}
-  __attribute__((target_version("sve-i8mm"))) ~Cls();
+  __attribute__((target_version("sve2-aes"))) ~Cls();
   // expected-error@-1 {{attribute 'target_version' multiversioned functions do not yet support destructors}}
 
   Cls &__attribute__((target_version("f32mm"))) operator=(const Cls &) = default;
@@ -98,11 +98,11 @@ __attribute__((target_version("jscvt"))) void Decl();
 } // namespace Nms
 
 class Out {
-  int __attribute__((target_version("bti"))) func(void);
-  int __attribute__((target_version("ssbs2"))) func(void);
+  int __attribute__((target_version("mops"))) func(void);
+  int __attribute__((target_version("ssbs"))) func(void);
 };
-int __attribute__((target_version("bti"))) Out::func(void) { return 1; }
-int __attribute__((target_version("ssbs2"))) Out::func(void) { return 2; }
+int __attribute__((target_version("mops"))) Out::func(void) { return 1; }
+int __attribute__((target_version("ssbs"))) Out::func(void) { return 2; }
 // expected-error@+3 {{out-of-line definition of 'func' does not match any declaration in 'Out'}}
 // expected-note@-3 {{member declaration nearly matches}}
 // expected-note@-3 {{member declaration nearly matches}}
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64.c b/compiler-rt/lib/builtins/cpu_model/aarch64.c
index 17bddfca46f09..1ac4d85a0c139 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64.c
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64.c
@@ -67,13 +67,10 @@ enum CPUFeatures {
   FEAT_FP,
   FEAT_SIMD,
   FEAT_CRC,
-  FEAT_SHA1,
   FEAT_SHA2,
   FEAT_SHA3,
   FEAT_AES,
-  FEAT_PMULL,
   FEAT_FP16,
-  FEAT_DIT,
   FEAT_DPB,
   FEAT_DPB2,
   FEAT_JSCVT,
@@ -81,35 +78,23 @@ enum CPUFeatures {
   FEAT_RCPC,
   FEAT_RCPC2,
   FEAT_FRINTTS,
-  FEAT_DGH,
   FEAT_I8MM,
   FEAT_BF16,
-  FEAT_EBF16,
   FEAT_RPRES,
   FEAT_SVE,
-  FEAT_SVE_BF16,
-  FEAT_SVE_EBF16,
-  FEAT_SVE_I8MM,
   FEAT_SVE_F32MM,
   FEAT_SVE_F64MM,
   FEAT_SVE2,
   FEAT_SVE_AES,
-  FEAT_SVE_PMULL128,
   FEAT_SVE_BITPERM,
   FEAT_SVE_SHA3,
   FEAT_SVE_SM4,
   FEAT_SME,
   FEAT_MEMTAG,
-  FEAT_MEMTAG2,
-  FEAT_MEMTAG3,
   FEAT_SB,
   FEAT_PREDRES,
   FEAT_SSBS,
-  FEAT_SSBS2,
-  FEAT_BTI,
   FEAT_LS64,
-  FEAT_LS64_V,
-  FEAT_LS64_ACCDATA,
   FEAT_WFXT,
   FEAT_SME_F64,
   FEAT_SME_I64,
@@ -117,7 +102,7 @@ enum CPUFeatures {
   FEAT_RCPC3,
   FEAT_MOPS,
   FEAT_MAX,
-  FEAT_EXT = 62, // Reserved to indicate presence of additional features field
+  FEAT_EXT = 47, // Reserved to indicate presence of additional features field
                  // in __aarch64_cpu_features
   FEAT_INIT      // Used as flag of features initialization completion
 };
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
index 6fef109567b61..19114c4abdfca 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
@@ -35,13 +35,10 @@ void __init_cpu_features_resolver(void) {
       {"hw.optional.floatingpoint", FEAT_FP},
       {"hw.optional.AdvSIMD", FEAT_SIMD},
       {"hw.optional.armv8_crc32", FEAT_CRC},
-      {"hw.optional.arm.FEAT_SHA1", FEAT_SHA1},
       {"hw.optional.arm.FEAT_SHA256", FEAT_SHA2},
       {"hw.optional.arm.FEAT_SHA3", FEAT_SHA3},
       {"hw.optional.arm.FEAT_AES", FEAT_AES},
-      {"hw.optional.arm.FEAT_PMULL", FEAT_PMULL},
       {"hw.optional.arm.FEAT_FP16", FEAT_FP16},
-      {"hw.optional.arm.FEAT_DIT", FEAT_DIT},
       {"hw.optional.arm.FEAT_DPB", FEAT_DPB},
       {"hw.optional.arm.FEAT_DPB2", FEAT_DPB2},
       {"hw.optional.arm.FEAT_JSCVT", FEAT_JSCVT},
@@ -53,8 +50,7 @@ void __init_cpu_features_resolver(void) {
       {"hw.optional.arm.FEAT_BF16", FEAT_BF16},
       {"hw.optional.arm.FEAT_SB", FEAT_SB},
       {"hw.optional.arm.FEAT_SPECRES", FEAT_PREDRES},
-      {"hw.optional.arm.FEAT_SSBS", FEAT_SSBS2},
-      {"hw.optional.arm.FEAT_BTI", FEAT_BTI},
+      {"hw.optional.arm.FEAT_SSBS", FEAT_SSBS},
   };
 
   for (size_t I = 0, E = sizeof(feature_checks) / sizeof(feature_checks[0]);
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
index d8e0280f40416..579aa00dc5c72 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
@@ -22,10 +22,6 @@ void __init_cpu_features_resolver() {
     setCPUFeature(FEAT_SIMD);
   if (features & ZX_ARM64_FEATURE_ISA_AES)
     setCPUFeature(FEAT_AES);
-  if (features & ZX_ARM64_FEATURE_ISA_PMULL)
-    setCPUFeature(FEAT_PMULL);
-  if (features & ZX_ARM64_FEATURE_ISA_SHA1)
-    setCPUFeature(FEAT_SHA1);
   if (features & ZX_ARM64_FEATURE_ISA_SHA256)
     setCPUFeature(FEAT_SHA2);
   if (features & ZX_ARM64_FEATURE_ISA_CRC32)
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
index 32a21a2fba9a3..54797852e4aaa 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
@@ -16,8 +16,6 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     hwcap2 = arg->_hwcap2;
   if (hwcap & HWCAP_CRC32)
     setCPUFeature(FEAT_CRC);
-  if (hwcap & HWCAP_PMULL)
-    setCPUFeature(FEAT_PMULL);
   if (hwcap & HWCAP_FLAGM)
     setCPUFeature(FEAT_FLAGM);
   if (hwcap2 & HWCAP2_FLAGM2) {
@@ -34,16 +32,12 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     setCPUFeature(FEAT_FP16);
     setCPUFeature(FEAT_FP);
   }
-  if (hwcap & HWCAP_DIT)
-    setCPUFeature(FEAT_DIT);
   if (hwcap & HWCAP_ASIMDRDM)
     setCPUFeature(FEAT_RDM);
   if (hwcap & HWCAP_ILRCPC)
     setCPUFeature(FEAT_RCPC2);
   if (hwcap & HWCAP_AES)
     setCPUFeature(FEAT_AES);
-  if (hwcap & HWCAP_SHA1)
-    setCPUFeature(FEAT_SHA1);
   if (hwcap & HWCAP_SHA2)
     setCPUFeature(FEAT_SHA2);
   if (hwcap & HWCAP_JSCVT)
@@ -53,22 +47,11 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
   if (hwcap & HWCAP_SB)
     setCPUFeature(FEAT_SB);
   if (hwcap & HWCAP_SSBS)
-    setCPUFeature(FEAT_SSBS2);
-  if (hwcap2 & HWCAP2_MTE) {
+    setCPUFeature(FEAT_SSBS);
+  if (hwcap2 & HWCAP2_MTE)
     setCPUFeature(FEAT_MEMTAG);
-    setCPUFeature(FEAT_MEMTAG2);
-  }
-  if (hwcap2 & HWCAP2_MTE3) {
-    setCPUFeature(FEAT_MEMTAG);
-    setCPUFeature(FEAT_MEMTAG2);
-    setCPUFeature(FEAT_MEMTAG3);
-  }
   if (hwcap2 & HWCAP2_SVEAES)
     setCPUFeature(FEAT_SVE_AES);
-  if (hwcap2 & HWCAP2_SVEPMULL) {
-    setCPUFeature(FEAT_SVE_AES);
-    setCPUFeature(FEAT_SVE_PMULL128);
-  }
   if (hwcap2 & HWCAP2_SVEBITPERM)
     setCPUFeature(FEAT_SVE_BITPERM);
   if (hwcap2 & HWCAP2_SVESHA3)
@@ -83,22 +66,12 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     setCPUFeature(FEAT_RNG);
   if (hwcap2 & HWCAP2_I8MM)
     setCPUFeature(FEAT_I8MM);
-  if (hwcap2 & HWCAP2_EBF16)
-    setCPUFeature(FEAT_EBF16);
-  if (hwcap2 & HWCAP2_SVE_EBF16)
-    setCPUFeature(FEAT_SVE_EBF16);
-  if (hwcap2 & HWCAP2_DGH)
-    setCPUFeature(FEAT_DGH);
   if (hwcap2 & HWCAP2_FRINT)
     setCPUFeature(FEAT_FRINTTS);
-  if (hwcap2 & HWCAP2_SVEI8MM)
-    setCPUFeature(FEAT_SVE_I8MM);
   if (hwcap2 & HWCAP2_SVEF32MM)
     setCPUFeature(FEAT_SVE_F32MM);
   if (hwcap2 & HWCAP2_SVEF64MM)
     setCPUFeature(FEAT_SVE_F64MM);
-  if (hwcap2 & HWCAP2_BTI)
-    setCPUFeature(FEAT_BTI);
   if (hwcap2 & HWCAP2_RPRES)
     setCPUFeature(FEAT_RPRES);
   if (hwcap2 & HWCAP2_WFXT)
@@ -141,9 +114,6 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
       // ID_AA64ZFR0_EL1.SVEver == 0b0001
       if (extractBits(ftr, 0, 4) == 0x1)
         setCPUFeature(FEAT_SVE2);
-      // ID_AA64ZFR0_EL1.BF16 != 0b0000
-      if (extractBits(ftr, 20, 4) != 0x0)
-        setCPUFeature(FEAT_SVE_BF16);
     }
     getCPUFeature(ID_AA64ISAR0_EL1, ftr);
     // ID_AA64ISAR0_EL1.SHA3 != 0b0000
@@ -168,12 +138,6 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     // ID_AA64ISAR1_EL1.LS64 >= 0b0001
     if (extractBits(ftr, 60, 4) >= 0x1)
       setCPUFeature(FEAT_LS64);
-    // ID_AA64ISAR1_EL1.LS64 >= 0b0010
-    if (extractBits(ftr, 60, 4) >= 0x2)
-      setCPUFeature(FEAT_LS64_V);
-    // ID_AA64ISAR1_EL1.LS64 >= 0b0011
-    if (extractBits(ftr, 60, 4) >= 0x3)
-      setCPUFeature(FEAT_LS64_ACCDATA);
   } else {
     // Set some features in case of no CPUID support
     if (hwcap & (HWCAP_FP | HWCAP_FPHP)) {
@@ -187,8 +151,6 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
       setCPUFeature(FEAT_RCPC);
     if (hwcap2 & HWCAP2_BF16 || hwcap2 & HWCAP2_EBF16)
       setCPUFeature(FEAT_BF16);
-    if (hwcap2 & HWCAP2_SVEBF16)
-      setCPUFeature(FEAT_SVE_BF16);
     if (hwcap2 & HWCAP2_SVE2 && hwcap & HWCAP_SVE)
       setCPUFeature(FEAT_SVE2);
     if (hwcap & HWCAP_SHA3)
diff --git a/llvm/include/llvm/TargetParser/AArch64TargetParser.h b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
index 805b963a7a13c..bcb36beb3cc0b 100644
--- a/llvm/include/llvm/TargetParser/AArch64TargetParser.h
+++ b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
@@ -46,13 +46,10 @@ enum CPUFeatures {
   FEAT_FP,
   FEAT_SIMD,
   FEAT_CRC,
-  FEAT_SHA1,
   FEAT_SHA2,
   FEAT_SHA3,
   FEAT_AES,
-  FEAT_PMULL,
   FEAT_FP16,
-  FEAT_DIT,
   FEAT_DPB,
   FEAT_DPB2,
   FEAT_JSCVT,
@@ -60,35 +57,23 @@ enum CPUFeatures {
   FEAT_RCPC,
   FEAT_RCPC2,
   FEAT_FRINTTS,
-  FEAT_DGH,
   FEAT_I8MM,
   FEAT_BF16,
-  FEAT_EBF16,
   FEAT_RPRES,
   FEAT_SVE,
-  FEAT_SVE_BF16,
-  FEAT_SVE_EBF16,
-  FEAT_SVE_I8MM,
   FEAT_SVE_F32MM,
   FEAT_SVE_F64MM,
   FEAT_SVE2,
   FEAT_SVE_AES,
-  FEAT_SVE_PMULL128,
   FEAT_SVE_BITPERM,
   FEAT_SVE_SHA3,
   FEAT_SVE_SM4,
   FEAT_SME,
   FEAT_MEMTAG,
-  FEAT_MEMTAG2,
-  FEAT_MEMTAG3,
   FEAT_SB,
   FEAT_PREDRES,
   FEAT_SSBS,
-  FEAT_SSBS2,
-  FEAT_BTI,
   FEAT_LS64,
-  FEAT_LS64_V,
-  FEAT_LS64_ACCDATA,
   FEAT_WFXT,
   FEAT_SME_F64,
   FEAT_SME_I64,
@@ -96,12 +81,12 @@ enum CPUFeatures {
   FEAT_RCPC3,
   FEAT_MOPS,
   FEAT_MAX,
-  FEAT_EXT = 62,
+  FEAT_EXT = 47,
   FEAT_INIT
 };
 
-static_assert(FEAT_MAX < 62,
-              "Number of features in CPUFeatures are limited to 62 entries");
+static_assert(FEAT_MAX < 47,
+              "Number of features in CPUFeatures are limited to 47 entries");
 
 // Arch extension modifiers for CPUs. These are labelled with their Arm ARM
 // feature name (though the canonical reference for those is AArch64.td)
@@ -215,17 +200,13 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"b16b16", AArch64::AEK_B16B16, "+b16b16", "-b16b16", FEAT_INIT, "", 0},
     {"bf16", AArch64::AEK_BF16, "+bf16", "-bf16", FEAT_BF16, "+bf16", 280},
     {"brbe", AArch64::AEK_BRBE, "+brbe", "-brbe", FEAT_INIT, "", 0},
-    {"bti", AArch64::AEK_NONE, {}, {}, FEAT_BTI, "+bti", 510},
     {"crc", AArch64::AEK_CRC, "+crc", "-crc", FEAT_CRC, "+crc", 110},
     {"crypto", AArch64::AEK_CRYPTO, "+crypto", "-crypto", FEAT_INIT, "+aes,+sha2", 0},
     {"cssc", AArch64::AEK_CSSC, "+cssc", "-cssc", FEAT_INIT, "", 0},
     {"d128", AArch64::AEK_D128, "+d128", "-d128", FEAT_INIT, "", 0},
-    {"dgh", AArch64::AEK_NONE, {}, {}, FEAT_DGH, "", 260},
-    {"dit", AArch64::AEK_NONE, {}, {}, FEAT_DIT, "+dit", 180},
     {"dotprod", AArch64::AEK_DOTPROD, "+dotprod", "-dotprod", FEAT_DOTPROD, "+dotprod,+fp-armv8,+neon", 104},
     {"dpb", AArch64::AEK_NONE, {}, {}, FEAT_DPB, "+ccpp", 190},
     {"dpb2", AArch64::AEK_NONE, {}, {}, FEAT_DPB2, "+ccpp,+ccdp", 200},
-    {"ebf16", AArch64::AEK_NONE, {}, {}, FEAT_EBF16, "+bf16", 290},
     {"f32mm", AArch64::AEK_F32MM, "+f32mm", "-f32mm", FEAT_SVE_F32MM, "+sve,+f32mm,+fullfp16,+fp-armv8,+neon", 350},
     {"f64mm", AArch64::AEK_F64MM, "+f64mm", "-f64mm", FEAT_SVE_F64MM, "+sve,+f64mm,+fullfp16,+fp-armv8,+neon", 360},
     {"fcma", AArch64::AEK_FCMA, "+complxnum", "-complxnum", FEAT_FCMA, "+fp-armv8,+neon,+complxnum", 220},
@@ -239,17 +220,12 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"i8mm", AArch64::AEK_I8MM, "+i8mm", "-i8mm", FEAT_I8MM, "+i8mm", 270},
     {"ite", AArch64::AEK_ITE, "+ite", "-ite", FEAT_INIT, "", 0},
     {"jscvt", AArch64::AEK_JSCVT, "+jsconv", "-jsconv", FEAT_JSCVT, "+fp-armv8,+neon,+jsconv", 210},
-    {"ls64_accdata", AArch64::AEK_NONE, {}, {}, FEAT_LS64_ACCDATA, "+ls64", 540},
-    {"ls64_v", AArch64::AEK_NONE, {}, {}, FEAT_LS64_V, "", 530},
     {"ls64", AArch64::AEK_LS64, "+ls64", "-ls64", FEAT_LS64, "", 520},
     {"lse", AArch64::AEK_LSE, "+lse", "-lse", FEAT_LSE, "+lse", 80},
     {"lse128", AArch64::AEK_LSE128, "+lse128", "-lse128", FEAT_INIT, "", 0},
     {"memtag", AArch64::AEK_MTE, "+mte", "-mte", FEAT_MEMTAG, "", 440},
-    {"memtag2", AArch64::AEK_NONE, {}, {}, FEAT_MEMTAG2, "+mte", 450},
-    {"memtag3", AArch64::AEK_NONE, {}, {}, FEAT_MEMTAG3, "+mte", 460},
     {"mops", AArch64::AEK_MOPS, "+mops", "-mops", FEAT_MOPS, "+mops", 650},
     {"pauth", AArch64::AEK_PAUTH, "+pauth", "-pauth", FEAT_INIT, "", 0},
-    {"pmull", AArch64::AEK_NONE, {}, {}, FEAT_PMULL, "+aes,+fp-armv8,+neon", 160},
     {"pmuv3", AArch64::AEK_PERFMON, "+perfmon", "-perfmon", FEAT_INIT, "", 0},
     {"predres", AArch64::AEK_PREDRES, "+predres", "-predres", FEAT_PREDRES, "+predres", 480},
     {"predres2", AArch64::AEK_SPECRES2, "+specres2", "-specres2", FEAT_INIT, "", 0},
@@ -263,7 +239,6 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"rng", AArch64::AEK_RAND, "+rand", "-rand", FEAT_RNG, "+rand", 10},
     {"rpres", AArch64::AEK_NONE, {}, {}, FEAT_RPRES, "", 300},
     {"sb", AArch64::AEK_SB, "+sb", "-sb", FEAT_SB, "+sb", 470},
-    {"sha1", AArch64::AEK_NONE, {}, {}, FEAT_SHA1, "+fp-armv8,+neon", 120},
     {"sha2", AArch64::AEK_SHA2, "+sha2", "-sha2", FEAT_SHA2, "+sha2,+fp-armv8,+neon", 130},
     {"sha3", AArch64::AEK_SHA3, "+sha3", "-sha3", FEAT_SHA3, "+sha3,+sha2,+fp-armv8,+neon", 140},
     {"simd", AArch64::AEK_SIMD, "+neon", "-neon", FEAT_SIMD, "+fp-armv8,+neon", 100},
@@ -275,14 +250,9 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"sme2", AArch64::AEK_SME2, "+sme2", "-sme2", FEAT_SME2, "+sme2,+sme,+bf16", 580},
     {"sme2p1", AArch64::AEK_SME2p1, "+sme2p1", "-sme2p1", FEAT_INIT, "+sme2p1,+sme2,+sme,+bf16", 0},
     {"ssbs", AArch64::AEK_SSBS, "+ssbs", "-ssbs", FEAT_SSBS, "", 490},
-    {"ssbs2", AArch64::AEK_NONE, {}, {}, FEAT_SSBS2, "+ssbs", 500},
-    {"sve-bf16", AArch64::AEK_NONE, {}, {}, FEAT_SVE_BF16, "+sve,+bf16,+fullfp16,+fp-armv8,+neon", 320},
-    {"sve-ebf16", AArch64::AEK_NONE, {}, {}, FEAT_SVE_EBF16, "+sve,+bf16,+fullfp16,+fp-armv8,+neon", 330},
-    {"sve-i8mm", AArch64::AEK_NONE, {}, {}, FEAT_SVE_I8MM, "+sve,+i8mm,+fullfp16,+fp-armv8,+neon", 340},
     {"sve", AArch64::AEK_SVE, "+sve", "-sve", FEAT_SVE, "+sve,+fullfp16,+fp-armv8,+neon", 310},
     {"sve2-aes", AArch64::AEK_SVE2AES, "+sve2-aes", "-sve2-aes", FEAT_SVE_AES, "+sve2,+sve,+sve2-aes,+fullfp16,+fp-armv8,+neon", 380},
     {"sve2-bitperm", AArch64::AEK_SVE2BITPERM, "+sve2-bitperm", "-sve2-bitperm", FEAT_SVE_BITPERM, "+sve2,+sve,+sve2-bitperm,+fullfp16,+fp-armv8,+neon", 400},
-    {"sve2-pmull128", AArch64::AEK_NONE, {}, {}, FEAT_SVE_PMULL128, "+sve2,+sve,+sve2-aes,+fullfp16,+fp-armv8,+neon", 390},
     {"sve2-sha3", AArch64::AEK_SVE2SHA3, "+sve2-sha3", "-sve2-sha3", FEAT_SVE_SHA3, "+sve2,+sve,+sve2-sha3,+fullfp16,+fp-armv8,+neon", 410},
     {"sve2-sm4", AArch64::AEK_SVE2SM4, "+sve2-sm4", "-sve2-sm4", FEAT_SVE_SM4, "+sve2,+sve,+sve2-sm4,+fullfp16,+fp-armv8,+neon", 420},
     {"sve2", AArch64::AEK_SVE2, "+sve2", "-sve2", FEAT_SVE2, "+sve2,+sve,+fullfp16,+fp-armv8,+neon", 370},

From 06947b9f8d258fe66fc69f1e7c0197cb621da3a5 Mon Sep 17 00:00:00 2001
From: Nick Desaulniers <nickdesaulniers@users.noreply.github.com>
Date: Wed, 17 Apr 2024 09:31:29 -0700
Subject: [PATCH 763/886] [libc][POSIX][pthreads] implement pthread_condattr_t
 functions (#88987)

Implement:
- pthread_condattr_destroy
- pthread_condattr_getclock
- pthread_condattr_getpshared
- pthread_condattr_init
- pthread_condattr_setclock
- pthread_condattr_setpshared

Fixes: #88581
---
 libc/config/linux/api.td                      | 29 ++++++--
 libc/config/linux/x86_64/entrypoints.txt      |  6 ++
 libc/include/CMakeLists.txt                   |  1 +
 libc/include/llvm-libc-types/CMakeLists.txt   |  3 +-
 .../llvm-libc-types/pthread_condattr_t.h      | 18 +++++
 libc/include/pthread.h.def                    |  6 ++
 libc/spec/posix.td                            | 64 ++++++++++++++++-
 libc/src/pthread/CMakeLists.txt               | 65 +++++++++++++++++
 libc/src/pthread/pthread_condattr_destroy.cpp | 23 ++++++
 libc/src/pthread/pthread_condattr_destroy.h   | 20 ++++++
 .../src/pthread/pthread_condattr_getclock.cpp | 25 +++++++
 libc/src/pthread/pthread_condattr_getclock.h  | 22 ++++++
 .../pthread/pthread_condattr_getpshared.cpp   | 24 +++++++
 .../src/pthread/pthread_condattr_getpshared.h | 21 ++++++
 libc/src/pthread/pthread_condattr_init.cpp    | 24 +++++++
 libc/src/pthread/pthread_condattr_init.h      | 20 ++++++
 .../src/pthread/pthread_condattr_setclock.cpp | 30 ++++++++
 libc/src/pthread/pthread_condattr_setclock.h  | 21 ++++++
 .../pthread/pthread_condattr_setpshared.cpp   | 28 ++++++++
 .../src/pthread/pthread_condattr_setpshared.h | 20 ++++++
 libc/test/src/pthread/CMakeLists.txt          | 12 ++++
 .../src/pthread/pthread_condattr_test.cpp     | 71 +++++++++++++++++++
 22 files changed, 545 insertions(+), 8 deletions(-)
 create mode 100644 libc/include/llvm-libc-types/pthread_condattr_t.h
 create mode 100644 libc/src/pthread/pthread_condattr_destroy.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_destroy.h
 create mode 100644 libc/src/pthread/pthread_condattr_getclock.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_getclock.h
 create mode 100644 libc/src/pthread/pthread_condattr_getpshared.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_getpshared.h
 create mode 100644 libc/src/pthread/pthread_condattr_init.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_init.h
 create mode 100644 libc/src/pthread/pthread_condattr_setclock.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_setclock.h
 create mode 100644 libc/src/pthread/pthread_condattr_setpshared.cpp
 create mode 100644 libc/src/pthread/pthread_condattr_setpshared.h
 create mode 100644 libc/test/src/pthread/pthread_condattr_test.cpp

diff --git a/libc/config/linux/api.td b/libc/config/linux/api.td
index 9964971f191b7..5fb92a9c299cc 100644
--- a/libc/config/linux/api.td
+++ b/libc/config/linux/api.td
@@ -175,6 +175,7 @@ def PThreadAPI : PublicAPI<"pthread.h"> {
       "__pthread_start_t",
       "__pthread_tss_dtor_t",
       "pthread_attr_t",
+      "pthread_condattr_t",
       "pthread_mutex_t",
       "pthread_mutexattr_t",
       "pthread_t",
@@ -241,10 +242,30 @@ def SysSendfileAPI : PublicAPI<"sys/sendfile.h"> {
 }
 
 def SysTypesAPI : PublicAPI<"sys/types.h"> {
-  let Types = ["blkcnt_t", "blksize_t", "clockid_t", "dev_t", "gid_t", "ino_t",
-               "mode_t", "nlink_t", "off_t", "pid_t", "pthread_attr_t", "pthread_key_t",
-               "pthread_mutex_t", "pthread_mutexattr_t", "pthread_once_t", "pthread_t",
-               "size_t", "ssize_t", "suseconds_t", "time_t", "uid_t"];
+  let Types = [
+    "blkcnt_t",
+    "blksize_t",
+    "clockid_t",
+    "dev_t",
+    "gid_t",
+    "ino_t",
+    "mode_t",
+    "nlink_t",
+    "off_t",
+    "pid_t",
+    "pthread_attr_t",
+    "pthread_condattr_t",
+    "pthread_key_t",
+    "pthread_mutex_t",
+    "pthread_mutexattr_t",
+    "pthread_once_t",
+    "pthread_t",
+    "size_t",
+    "ssize_t",
+    "suseconds_t",
+    "time_t",
+    "uid_t"
+  ];
 }
 
 def SysUtsNameAPI : PublicAPI<"sys/utsname.h"> {
diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index 70f130a4399a3..2d8136536b218 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -639,6 +639,12 @@ if(LLVM_LIBC_FULL_BUILD)
     libc.src.pthread.pthread_attr_setguardsize
     libc.src.pthread.pthread_attr_setstack
     libc.src.pthread.pthread_attr_setstacksize
+    libc.src.pthread.pthread_condattr_destroy
+    libc.src.pthread.pthread_condattr_getclock
+    libc.src.pthread.pthread_condattr_getpshared
+    libc.src.pthread.pthread_condattr_init
+    libc.src.pthread.pthread_condattr_setclock
+    libc.src.pthread.pthread_condattr_setpshared
     libc.src.pthread.pthread_create
     libc.src.pthread.pthread_detach
     libc.src.pthread.pthread_equal
diff --git a/libc/include/CMakeLists.txt b/libc/include/CMakeLists.txt
index b85366c8deafe..f5ba2791af3fb 100644
--- a/libc/include/CMakeLists.txt
+++ b/libc/include/CMakeLists.txt
@@ -321,6 +321,7 @@ add_gen_header(
     .llvm-libc-types.__pthread_start_t
     .llvm-libc-types.__pthread_tss_dtor_t
     .llvm-libc-types.pthread_attr_t
+    .llvm-libc-types.pthread_condattr_t
     .llvm-libc-types.pthread_mutex_t
     .llvm-libc-types.pthread_mutexattr_t
     .llvm-libc-types.pthread_t
diff --git a/libc/include/llvm-libc-types/CMakeLists.txt b/libc/include/llvm-libc-types/CMakeLists.txt
index 93a79e1477b33..f26fc0729dc94 100644
--- a/libc/include/llvm-libc-types/CMakeLists.txt
+++ b/libc/include/llvm-libc-types/CMakeLists.txt
@@ -49,11 +49,12 @@ add_header(pid_t HDR pid_t.h)
 add_header(posix_spawn_file_actions_t HDR posix_spawn_file_actions_t.h)
 add_header(posix_spawnattr_t HDR posix_spawnattr_t.h)
 add_header(pthread_attr_t HDR pthread_attr_t.h DEPENDS .size_t)
+add_header(pthread_condattr_t HDR pthread_condattr_t.h DEPENDS .clockid_t)
 add_header(pthread_key_t HDR pthread_key_t.h)
 add_header(pthread_mutex_t HDR pthread_mutex_t.h DEPENDS .__futex_word .__mutex_type)
-add_header(pthread_t HDR pthread_t.h DEPENDS .__thread_type)
 add_header(pthread_mutexattr_t HDR pthread_mutexattr_t.h)
 add_header(pthread_once_t HDR pthread_once_t.h DEPENDS .__futex_word)
+add_header(pthread_t HDR pthread_t.h DEPENDS .__thread_type)
 add_header(rlim_t HDR rlim_t.h)
 add_header(time_t HDR time_t.h)
 add_header(stack_t HDR stack_t.h)
diff --git a/libc/include/llvm-libc-types/pthread_condattr_t.h b/libc/include/llvm-libc-types/pthread_condattr_t.h
new file mode 100644
index 0000000000000..b91fc2950aa3f
--- /dev/null
+++ b/libc/include/llvm-libc-types/pthread_condattr_t.h
@@ -0,0 +1,18 @@
+//===-- Definition of pthread_condattr_t type -----------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+#ifndef LLVM_LIBC_TYPES_PTHREAD_CONDATTR_T_H
+#define LLVM_LIBC_TYPES_PTHREAD_CONDATTR_T_H
+
+#include "clockid_t.h"
+
+typedef struct {
+  clockid_t clock;
+  int pshared;
+} pthread_condattr_t;
+
+#endif // LLVM_LIBC_TYPES_PTHREAD_CONDATTR_T_H
diff --git a/libc/include/pthread.h.def b/libc/include/pthread.h.def
index abeb839ee83d1..a94d770657e10 100644
--- a/libc/include/pthread.h.def
+++ b/libc/include/pthread.h.def
@@ -11,6 +11,9 @@
 
 #include "__llvm-libc-common.h"
 
+// TODO: move to a pthreads-macros.h file:
+// https://github.com/llvm/llvm-project/issues/88997
+
 #define PTHREAD_STACK_MIN (1 << 14) // 16KB
 
 #define PTHREAD_MUTEX_INITIALIZER {0}
@@ -32,6 +35,9 @@ enum {
   PTHREAD_MUTEX_ROBUST = 0x1,
 };
 
+#define PTHREAD_PROCESS_PRIVATE 0
+#define PTHREAD_PROCESS_SHARED 1
+
 %%public_api()
 
 #endif // LLVM_LIBC_PTHREAD_H
diff --git a/libc/spec/posix.td b/libc/spec/posix.td
index 7095a3964ee3f..0c88dbd848a3f 100644
--- a/libc/spec/posix.td
+++ b/libc/spec/posix.td
@@ -26,6 +26,7 @@ def UidT : NamedType<"uid_t">;
 def GidT : NamedType<"gid_t">;
 def DevT : NamedType<"dev_t">;
 def ClockIdT : NamedType<"clockid_t">;
+def RestrictedClockIdTPtr : RestrictedPtrType<ClockIdT>;
 def BlkSizeT : NamedType<"blksize_t">;
 def BlkCntT : NamedType<"blkcnt_t">;
 def NLinkT : NamedType<"nlink_t">;
@@ -105,6 +106,10 @@ def POSIX : StandardSpec<"POSIX"> {
   ConstType ConstPThreadAttrTPtr = ConstType<PThreadAttrTPtr>;
   ConstType ConstRestrictedPThreadAttrTPtr = ConstType<RestrictedPThreadAttrTPtr>;
 
+  NamedType PThreadCondAttrTType = NamedType<"pthread_condattr_t">;
+  PtrType PThreadCondAttrTPtr = PtrType<PThreadCondAttrTType>;
+  ConstType ConstRestrictedPThreadCondAttrTPtr = ConstType<RestrictedPtrType<PThreadCondAttrTType>>;
+
   NamedType PThreadMutexAttrTType = NamedType<"pthread_mutexattr_t">;
   PtrType PThreadMutexAttrTPtr = PtrType<PThreadMutexAttrTType>;
   RestrictedPtrType RestrictedPThreadMutexAttrTPtr = RestrictedPtrType<PThreadMutexAttrTType>;
@@ -980,7 +985,9 @@ def POSIX : StandardSpec<"POSIX"> {
     [], // Macros
     [
         AtForkCallbackT,
+        ClockIdT,
         PThreadAttrTType,
+        PThreadCondAttrTType,
         PThreadKeyT,
         PThreadMutexAttrTType,
         PThreadMutexTType,
@@ -1047,6 +1054,36 @@ def POSIX : StandardSpec<"POSIX"> {
           RetValSpec<IntType>,
           [ArgSpec<PThreadAttrTPtr>, ArgSpec<VoidPtr>, ArgSpec<SizeTType>]
       >,
+      FunctionSpec<
+          "pthread_condattr_destroy",
+          RetValSpec<IntType>,
+          [ArgSpec<PThreadCondAttrTPtr>]
+      >,
+      FunctionSpec<
+          "pthread_condattr_getclock",
+          RetValSpec<IntType>,
+          [ArgSpec<ConstRestrictedPThreadCondAttrTPtr>, ArgSpec<RestrictedClockIdTPtr>]
+      >,
+      FunctionSpec<
+          "pthread_condattr_getpshared",
+          RetValSpec<IntType>,
+          [ArgSpec<ConstRestrictedPThreadCondAttrTPtr>, ArgSpec<RestrictedIntPtr>]
+      >,
+      FunctionSpec<
+          "pthread_condattr_init",
+          RetValSpec<IntType>,
+          [ArgSpec<PThreadCondAttrTPtr>]
+      >,
+      FunctionSpec<
+          "pthread_condattr_setclock",
+          RetValSpec<IntType>,
+          [ArgSpec<PThreadCondAttrTPtr>, ArgSpec<ClockIdT>]
+      >,
+      FunctionSpec<
+          "pthread_condattr_setpshared",
+          RetValSpec<IntType>,
+          [ArgSpec<PThreadCondAttrTPtr>, ArgSpec<IntType>]
+      >,
       FunctionSpec<
           "pthread_create",
           RetValSpec<IntType>,
@@ -1522,9 +1559,30 @@ def POSIX : StandardSpec<"POSIX"> {
   HeaderSpec SysTypes = HeaderSpec<
     "sys/types.h",
     [], // Macros
-    [BlkCntT, BlkSizeT, ClockIdT, DevT, GidT, InoT, ModeTType, NLinkT, OffTType, PidT,
-     PThreadAttrTType, PThreadKeyT, PThreadMutexTType, PThreadMutexAttrTType, PThreadOnceT, PThreadTType,
-     SizeTType, SSizeTType, SuSecondsT, TimeTType, UidT],
+    [
+      BlkCntT,
+      BlkSizeT,
+      ClockIdT,
+      DevT,
+      GidT,
+      InoT,
+      ModeTType,
+      NLinkT,
+      OffTType,
+      PThreadAttrTType,
+      PThreadCondAttrTType,
+      PThreadKeyT,
+      PThreadMutexAttrTType,
+      PThreadMutexTType,
+      PThreadOnceT,
+      PThreadTType,
+      PidT,
+      SSizeTType,
+      SizeTType,
+      SuSecondsT,
+      TimeTType,
+      UidT
+    ], // Types
     [], // Enumerations
     []  // Functions
   >;
diff --git a/libc/src/pthread/CMakeLists.txt b/libc/src/pthread/CMakeLists.txt
index d5e6c802a8452..3d6cf6dde010b 100644
--- a/libc/src/pthread/CMakeLists.txt
+++ b/libc/src/pthread/CMakeLists.txt
@@ -100,6 +100,71 @@ add_entrypoint_object(
     libc.src.pthread.pthread_attr_setstacksize
 )
 
+add_entrypoint_object(
+  pthread_condattr_destroy
+  SRCS
+    pthread_condattr_destroy.cpp
+  HDRS
+    pthread_condattr_destroy.h
+  DEPENDS
+    libc.include.pthread
+)
+
+add_entrypoint_object(
+  pthread_condattr_getclock
+  SRCS
+    pthread_condattr_getclock.cpp
+  HDRS
+    pthread_condattr_getclock.h
+  DEPENDS
+    libc.include.pthread
+    libc.include.sys_types
+)
+
+add_entrypoint_object(
+  pthread_condattr_getpshared
+  SRCS
+    pthread_condattr_getpshared.cpp
+  HDRS
+    pthread_condattr_getpshared.h
+  DEPENDS
+    libc.include.pthread
+)
+
+add_entrypoint_object(
+  pthread_condattr_init
+  SRCS
+    pthread_condattr_init.cpp
+  HDRS
+    pthread_condattr_init.h
+  DEPENDS
+    libc.include.pthread
+    libc.include.time
+)
+
+add_entrypoint_object(
+  pthread_condattr_setclock
+  SRCS
+    pthread_condattr_setclock.cpp
+  HDRS
+    pthread_condattr_setclock.h
+  DEPENDS
+    libc.include.errno
+    libc.include.pthread
+    libc.include.sys_types
+    libc.include.time
+)
+
+add_entrypoint_object(
+  pthread_condattr_setpshared
+  SRCS
+    pthread_condattr_setpshared.cpp
+  HDRS
+    pthread_condattr_setpshared.h
+  DEPENDS
+    libc.include.pthread
+)
+
 add_header_library(
   pthread_mutexattr
   HDRS
diff --git a/libc/src/pthread/pthread_condattr_destroy.cpp b/libc/src/pthread/pthread_condattr_destroy.cpp
new file mode 100644
index 0000000000000..45cc011a4a92d
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_destroy.cpp
@@ -0,0 +1,23 @@
+//===-- Implementation of the pthread_condattr_destroy --------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_destroy.h"
+
+#include "src/__support/common.h"
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_destroy, (pthread_condattr_t * attr)) {
+  // Initializing a pthread_condattr_t acquires no resources, so this is a
+  // no-op.
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_destroy.h b/libc/src/pthread/pthread_condattr_destroy.h
new file mode 100644
index 0000000000000..2910fa9f96168
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_destroy.h
@@ -0,0 +1,20 @@
+//===-- Implementation header for pthread_condattr_destroy ------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_DESTROY_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_DESTROY_H
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_destroy(pthread_condattr_t *attr);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_DESTROY_H
diff --git a/libc/src/pthread/pthread_condattr_getclock.cpp b/libc/src/pthread/pthread_condattr_getclock.cpp
new file mode 100644
index 0000000000000..a3a3963f4f429
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_getclock.cpp
@@ -0,0 +1,25 @@
+//===-- Implementation of the pthread_condattr_getclock -------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_getclock.h"
+
+#include "src/__support/common.h"
+
+#include <pthread.h>   // pthread_condattr_t
+#include <sys/types.h> // clockid_t
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_getclock,
+                   (const pthread_condattr_t *__restrict attr,
+                    clockid_t *__restrict clock_id)) {
+  *clock_id = attr->clock;
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_getclock.h b/libc/src/pthread/pthread_condattr_getclock.h
new file mode 100644
index 0000000000000..d5878c4f45b53
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_getclock.h
@@ -0,0 +1,22 @@
+//===-- Implementation header for pthread_condattr_getclock -----*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_GETCLOCK_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_GETCLOCK_H
+
+#include <pthread.h>   // pthread_condattr_t
+#include <sys/types.h> // clockid_t
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_getclock(const pthread_condattr_t *__restrict attr,
+                              clockid_t *__restrict clock_id);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_GETCLOCK_H
diff --git a/libc/src/pthread/pthread_condattr_getpshared.cpp b/libc/src/pthread/pthread_condattr_getpshared.cpp
new file mode 100644
index 0000000000000..0c5fdc115c25d
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_getpshared.cpp
@@ -0,0 +1,24 @@
+//===-- Implementation of the pthread_condattr_getpshared -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_getpshared.h"
+
+#include "src/__support/common.h"
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_getpshared,
+                   (const pthread_condattr_t *__restrict attr,
+                    int *__restrict pshared)) {
+  *pshared = attr->pshared;
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_getpshared.h b/libc/src/pthread/pthread_condattr_getpshared.h
new file mode 100644
index 0000000000000..3d7a0c1d357c6
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_getpshared.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for pthread_condattr_getpshared ---*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_PSHARED_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_PSHARED_H
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_getpshared(const pthread_condattr_t *__restrict attr,
+                                int *__restrict pshared);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_PSHARED_H
diff --git a/libc/src/pthread/pthread_condattr_init.cpp b/libc/src/pthread/pthread_condattr_init.cpp
new file mode 100644
index 0000000000000..54633b2e3a5ea
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_init.cpp
@@ -0,0 +1,24 @@
+//===-- Implementation of the pthread_condattr_init -----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_init.h"
+
+#include "src/__support/common.h"
+
+#include <pthread.h> // pthread_condattr_t, PTHREAD_PROCESS_PRIVATE
+#include <time.h>    // CLOCK_REALTIME
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_init, (pthread_condattr_t * attr)) {
+  attr->clock = CLOCK_REALTIME;
+  attr->pshared = PTHREAD_PROCESS_PRIVATE;
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_init.h b/libc/src/pthread/pthread_condattr_init.h
new file mode 100644
index 0000000000000..9f3c06bb6f4ae
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_init.h
@@ -0,0 +1,20 @@
+//===-- Implementation header for pthread_condattr_init ---------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_INIT_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_INIT_H
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_init(pthread_condattr_t *attr);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_INIT_H
diff --git a/libc/src/pthread/pthread_condattr_setclock.cpp b/libc/src/pthread/pthread_condattr_setclock.cpp
new file mode 100644
index 0000000000000..6eca8b30ef7f8
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_setclock.cpp
@@ -0,0 +1,30 @@
+//===-- Implementation of the pthread_condattr_setclock -------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_setclock.h"
+
+#include "src/__support/common.h"
+
+#include <errno.h>     // EINVAL
+#include <pthread.h>   // pthread_condattr_t
+#include <sys/types.h> // clockid_t
+#include <time.h>      // CLOCK_MONOTONIC, CLOCK_REALTIME
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_setclock,
+                   (pthread_condattr_t * attr, clockid_t clock)) {
+
+  if (clock != CLOCK_MONOTONIC && clock != CLOCK_REALTIME)
+    return EINVAL;
+
+  attr->clock = clock;
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_setclock.h b/libc/src/pthread/pthread_condattr_setclock.h
new file mode 100644
index 0000000000000..328766fe78833
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_setclock.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for pthread_condattr_setclock -----*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETCLOCK_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETCLOCK_H
+
+#include <pthread.h>
+#include <sys/types.h> // clockid_t
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_setclock(pthread_condattr_t *attr, clockid_t clock);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETCLOCK_H
diff --git a/libc/src/pthread/pthread_condattr_setpshared.cpp b/libc/src/pthread/pthread_condattr_setpshared.cpp
new file mode 100644
index 0000000000000..7f1560acad843
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_setpshared.cpp
@@ -0,0 +1,28 @@
+//===-- Implementation of the pthread_condattr_setpshared -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "pthread_condattr_setpshared.h"
+
+#include "src/__support/common.h"
+
+#include <errno.h> // EINVAL
+#include <pthread.h> // pthread_condattr_t, PTHREAD_PROCESS_SHARED, PTHREAD_PROCESS_PRIVATE
+
+namespace LIBC_NAMESPACE {
+
+LLVM_LIBC_FUNCTION(int, pthread_condattr_setpshared,
+                   (pthread_condattr_t * attr, int pshared)) {
+
+  if (pshared != PTHREAD_PROCESS_SHARED && pshared != PTHREAD_PROCESS_PRIVATE)
+    return EINVAL;
+
+  attr->pshared = pshared;
+  return 0;
+}
+
+} // namespace LIBC_NAMESPACE
diff --git a/libc/src/pthread/pthread_condattr_setpshared.h b/libc/src/pthread/pthread_condattr_setpshared.h
new file mode 100644
index 0000000000000..8083bdec78cc4
--- /dev/null
+++ b/libc/src/pthread/pthread_condattr_setpshared.h
@@ -0,0 +1,20 @@
+//===-- Implementation header for pthread_condattr_setpshared ---*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETPSHARED_H
+#define LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETPSHARED_H
+
+#include <pthread.h>
+
+namespace LIBC_NAMESPACE {
+
+int pthread_condattr_setpshared(pthread_condattr_t *attr, int pshared);
+
+} // namespace LIBC_NAMESPACE
+
+#endif // LLVM_LIBC_SRC_PTHREAD_PTHREAD_CONDATTR_SETPSHARED_H
diff --git a/libc/test/src/pthread/CMakeLists.txt b/libc/test/src/pthread/CMakeLists.txt
index fb0d22ab9d2a5..46f38422a5374 100644
--- a/libc/test/src/pthread/CMakeLists.txt
+++ b/libc/test/src/pthread/CMakeLists.txt
@@ -39,3 +39,15 @@ add_libc_unittest(
     libc.src.pthread.pthread_mutexattr_setrobust
     libc.src.pthread.pthread_mutexattr_settype
 )
+
+add_libc_unittest(
+  pthread_condattr_test
+  SUITE
+    libc_pthread_unittests
+  SRCS
+    pthread_condattr_test.cpp
+  DEPENDS
+    libc.include.errno
+    libc.include.pthread
+    libc.include.time
+  )
diff --git a/libc/test/src/pthread/pthread_condattr_test.cpp b/libc/test/src/pthread/pthread_condattr_test.cpp
new file mode 100644
index 0000000000000..accb62de92e45
--- /dev/null
+++ b/libc/test/src/pthread/pthread_condattr_test.cpp
@@ -0,0 +1,71 @@
+//===-- Unittests for pthread_condattr_t ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "test/UnitTest/Test.h"
+
+#include <errno.h>
+#include <pthread.h>
+#include <time.h>
+
+TEST(LlvmLibcPThreadCondAttrTest, InitAndDestroy) {
+  pthread_condattr_t cond;
+  ASSERT_EQ(pthread_condattr_init(&cond), 0);
+  ASSERT_EQ(pthread_condattr_destroy(&cond), 0);
+}
+
+TEST(LlvmLibcPThreadCondAttrTest, GetDefaultValues) {
+  pthread_condattr_t cond;
+
+  // Invalid clock id.
+  clockid_t clock = 7;
+  // Invalid value.
+  int pshared = 42;
+
+  ASSERT_EQ(pthread_condattr_init(&cond), 0);
+  ASSERT_EQ(pthread_condattr_getclock(&cond, &clock), 0);
+  ASSERT_EQ(clock, CLOCK_REALTIME);
+  ASSERT_EQ(pthread_condattr_getpshared(&cond, &pshared), 0);
+  ASSERT_EQ(pshared, PTHREAD_PROCESS_PRIVATE);
+  ASSERT_EQ(pthread_condattr_destroy(&cond), 0);
+}
+
+TEST(LlvmLibcPThreadCondAttrTest, SetGoodValues) {
+  pthread_condattr_t cond;
+
+  // Invalid clock id.
+  clockid_t clock = 7;
+  // Invalid value.
+  int pshared = 42;
+
+  ASSERT_EQ(pthread_condattr_init(&cond), 0);
+  ASSERT_EQ(pthread_condattr_setclock(&cond, CLOCK_MONOTONIC), 0);
+  ASSERT_EQ(pthread_condattr_getclock(&cond, &clock), 0);
+  ASSERT_EQ(clock, CLOCK_MONOTONIC);
+  ASSERT_EQ(pthread_condattr_setpshared(&cond, PTHREAD_PROCESS_SHARED), 0);
+  ASSERT_EQ(pthread_condattr_getpshared(&cond, &pshared), 0);
+  ASSERT_EQ(pshared, PTHREAD_PROCESS_SHARED);
+  ASSERT_EQ(pthread_condattr_destroy(&cond), 0);
+}
+
+TEST(LlvmLibcPThreadCondAttrTest, SetBadValues) {
+  pthread_condattr_t cond;
+
+  // Invalid clock id.
+  clockid_t clock = 7;
+  // Invalid value.
+  int pshared = 42;
+
+  ASSERT_EQ(pthread_condattr_init(&cond), 0);
+  ASSERT_EQ(pthread_condattr_setclock(&cond, clock), EINVAL);
+  ASSERT_EQ(pthread_condattr_getclock(&cond, &clock), 0);
+  ASSERT_EQ(clock, CLOCK_REALTIME);
+  ASSERT_EQ(pthread_condattr_setpshared(&cond, pshared), EINVAL);
+  ASSERT_EQ(pthread_condattr_getpshared(&cond, &pshared), 0);
+  ASSERT_EQ(pshared, PTHREAD_PROCESS_PRIVATE);
+  ASSERT_EQ(pthread_condattr_destroy(&cond), 0);
+}

From 4edeaffbf255137861f5153eb1a6183d956efede Mon Sep 17 00:00:00 2001
From: fabrizio-indirli <fabrizio.indirli@arm.com>
Date: Wed, 17 Apr 2024 17:43:22 +0100
Subject: [PATCH 764/886] [mlir][tosa] Fix tosa.Resize-to-linalg lowering
 (#88514)

---
 .../Conversion/TosaToLinalg/TosaToLinalg.cpp  |  21 ++--
 .../TosaToLinalg/tosa-to-linalg-resize.mlir   | 112 ++++++++----------
 2 files changed, 60 insertions(+), 73 deletions(-)

diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
index d8dd1c93722b0..af19ebaea937d 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -1582,17 +1582,16 @@ class GenericResizeConverter : public OpRewritePattern<tosa::ResizeOp> {
         }
         // x = x * scale_d + offset;
         // ix = floor(x / scale_n)
-        // dx = x / scale_n - ix
-        Value val = b.create<arith::UIToFPOp>(floatTy, in);
-        scaleN = b.create<arith::UIToFPOp>(floatTy, scaleN);
-        scaleD = b.create<arith::UIToFPOp>(floatTy, scaleD);
-        offset = b.create<arith::SIToFPOp>(floatTy, offset);
-        val = b.create<arith::MulFOp>(val, scaleD);
-        val = b.create<arith::AddFOp>(val, offset);
-        val = b.create<arith::DivFOp>(val, scaleN);
-        index = b.create<math::FloorOp>(val);
-        delta = b.create<arith::SubFOp>(val, index);
-        index = b.create<arith::FPToSIOp>(b.getI32Type(), index);
+        Value val = b.create<arith::MulIOp>(in, scaleD);
+        val = b.create<arith::AddIOp>(val, offset);
+        index = b.create<arith::FloorDivSIOp>(val, scaleN);
+
+        // rx = x % scale_n
+        // dx = rx / scale_n
+        Value r = b.create<arith::RemSIOp>(val, scaleN);
+        Value rFp = b.create<arith::SIToFPOp>(floatTy, r);
+        Value scaleNfp = b.create<arith::UIToFPOp>(floatTy, scaleN);
+        delta = b.create<arith::DivFOp>(rFp, scaleNfp);
       };
 
       // Compute the ix and dx values for the X and Y dimensions - int case.
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-resize.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-resize.mlir
index 468e92e2a2661..d42d0a46692d4 100644
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-resize.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg-resize.mlir
@@ -304,42 +304,36 @@ func.func @resize_nearest_fp32(%input: tensor<1x50x48x1xf32>) -> () {
   // CHECK-DAG: %[[XMAX:.*]] = arith.constant 47
   // CHECK: %[[Y:.+]] = arith.index_cast %[[IDX1]]
   // CHECK: %[[X:.+]] = arith.index_cast %[[IDX2]]
-  // CHECK-DAG: %[[ISCALE_Y_N:.*]] = arith.constant 64
-  // CHECK-DAG: %[[ISCALE_Y_D:.*]] = arith.constant 2
-  // CHECK-DAG: %[[ISCALE_X_N:.*]] = arith.constant 64
-  // CHECK-DAG: %[[ISCALE_X_D:.*]] = arith.constant 2
-  // CHECK-DAG: %[[IOFFSET_Y:.*]] = arith.constant -31
-  // CHECK-DAG: %[[IOFFSET_X:.*]] = arith.constant -31
-  // CHECK-DAG: %[[IBORDER_Y:.*]] = arith.constant 31
-  // CHECK-DAG: %[[IBORDER_X:.*]] = arith.constant 31
-
-  // CHECK: %[[Y0:.+]] = arith.uitofp %[[Y]]
-  // CHECK: %[[SCALE_Y_N:.*]] = arith.uitofp %[[ISCALE_Y_N]]
-  // CHECK: %[[SCALE_Y_D:.*]] = arith.uitofp %[[ISCALE_Y_D]]
-  // CHECK: %[[OFFSET_Y:.*]] = arith.sitofp %[[IOFFSET_Y]]
-  // CHECK: %[[VAL_29:.*]] = arith.mulf %[[Y0]], %[[SCALE_Y_D]]
-  // CHECK: %[[VAL_31:.*]] = arith.addf %[[VAL_29]], %[[OFFSET_Y]]
-  // CHECK: %[[VAL_33:.*]] = arith.divf %[[VAL_31]], %[[SCALE_Y_N]]
-  // CHECK: %[[VAL_35:.*]] = math.floor %[[VAL_33]]
-  // CHECK: %[[D_Y:.*]] = arith.subf %[[VAL_33]], %[[VAL_35]]
-  // CHECK: %[[VAL_39:.*]] = arith.fptosi %[[VAL_35]]
-
-  // CHECK: %[[X0:.+]] = arith.uitofp %[[X]]
-  // CHECK: %[[SCALE_X_N:.*]] = arith.uitofp %[[ISCALE_X_N]]
-  // CHECK: %[[SCALE_X_D:.*]] = arith.uitofp %[[ISCALE_X_D]]
-  // CHECK: %[[OFFSET_X:.*]] = arith.sitofp %[[IOFFSET_X]]
-  // CHECK: %[[VAL_30:.*]] = arith.mulf %[[X0]], %[[SCALE_X_D]]
-  // CHECK: %[[VAL_32:.*]] = arith.addf %[[VAL_30]], %[[OFFSET_X]]
-  // CHECK: %[[VAL_34:.*]] = arith.divf %[[VAL_32]], %[[SCALE_X_N]]
-  // CHECK: %[[VAL_36:.*]] = math.floor %[[VAL_34]]
-  // CHECK: %[[D_X:.*]] = arith.subf %[[VAL_34]], %[[VAL_36]]
-  // CHECK: %[[VAL_40:.*]] = arith.fptosi %[[VAL_36]]
+  // CHECK-DAG: %[[SCALE_Y_N:.*]] = arith.constant 64
+  // CHECK-DAG: %[[SCALE_Y_D:.*]] = arith.constant 2
+  // CHECK-DAG: %[[SCALE_X_N:.*]] = arith.constant 64
+  // CHECK-DAG: %[[SCALE_X_D:.*]] = arith.constant 2
+  // CHECK-DAG: %[[OFFSET_Y:.*]] = arith.constant -31
+  // CHECK-DAG: %[[OFFSET_X:.*]] = arith.constant -31
+  // CHECK-DAG: %[[BORDER_Y:.*]] = arith.constant 31
+  // CHECK-DAG: %[[BORDER_X:.*]] = arith.constant 31
+
+  // CHECK: %[[VAL_29:.*]] = arith.muli %[[Y]], %[[SCALE_Y_D]]
+  // CHECK: %[[Y_TEMP:.*]] = arith.addi %[[VAL_29]], %[[OFFSET_Y]]
+  // CHECK: %[[IY_TEMP:.*]] = arith.floordivsi %[[Y_TEMP]], %[[SCALE_Y_N]]
+  // CHECK: %[[RY:.*]] = arith.remsi %[[Y_TEMP]], %[[SCALE_Y_N]]
+  // CHECK: %[[RY_FP:.*]] = arith.sitofp %[[RY]]
+  // CHECK: %[[SCALE_Y_N_FP:.*]] = arith.uitofp %[[SCALE_Y_N]]
+  // CHECK: %[[D_Y:.*]] = arith.divf %[[RY_FP]], %[[SCALE_Y_N_FP]]
+
+  // CHECK: %[[VAL_30:.*]] = arith.muli %[[X]], %[[SCALE_X_D]]
+  // CHECK: %[[X_TEMP:.*]] = arith.addi %[[VAL_30]], %[[OFFSET_X]]
+  // CHECK: %[[IX_TEMP:.*]] = arith.floordivsi %[[X_TEMP]], %[[SCALE_X_N]]
+  // CHECK: %[[RX:.*]] = arith.remsi %[[X_TEMP]], %[[SCALE_X_N]]
+  // CHECK: %[[RX_FP:.*]] = arith.sitofp %[[RX]]
+  // CHECK: %[[SCALE_X_N_FP:.*]] = arith.uitofp %[[SCALE_X_N]]
+  // CHECK: %[[D_X:.*]] = arith.divf %[[RX_FP]], %[[SCALE_X_N_FP]]
 
   // CHECK-DAG: %[[ONE:.*]] = arith.constant 1
   // CHECK-DAG: %[[HALF:.*]] = arith.constant 5.000000e-01
   // CHECK: %[[PRED_Y:.*]] = arith.cmpf oge, %[[D_Y]], %[[HALF]]
   // CHECK: %[[ROUND_Y:.*]] = arith.select %[[PRED_Y]], %[[ONE]], %[[ZERO]]
-  // CHECK: %[[VAL_48:.*]] = arith.addi %[[VAL_39]], %[[ROUND_Y]]
+  // CHECK: %[[VAL_48:.*]] = arith.addi %[[IY_TEMP]], %[[ROUND_Y]]
   // CHECK: %[[LOWER:.*]] = arith.maxsi %[[ZERO]], %[[VAL_48]]
   // CHECK: %[[CLAMPED:.*]] = arith.minsi %[[YMAX]], %[[LOWER]]
   // CHECK: %[[IDY:.*]] = arith.index_cast %[[CLAMPED]]
@@ -347,7 +341,7 @@ func.func @resize_nearest_fp32(%input: tensor<1x50x48x1xf32>) -> () {
   // CHECK-DAG: %[[HALF:.*]] = arith.constant 5.000000e-01
   // CHECK: %[[PRED_X:.*]] = arith.cmpf oge, %[[D_X]], %[[HALF]]
   // CHECK: %[[ROUND_X:.*]] = arith.select %[[PRED_X]], %[[ONE]], %[[ZERO]]
-  // CHECK: %[[VAL_49:.*]] = arith.addi %[[VAL_40]], %[[ROUND_X]]
+  // CHECK: %[[VAL_49:.*]] = arith.addi %[[IX_TEMP]], %[[ROUND_X]]
   // CHECK: %[[LOWER:.*]] = arith.maxsi %[[ZERO]], %[[VAL_49]]
   // CHECK: %[[CLAMPED:.*]] = arith.minsi %[[XMAX]], %[[LOWER]]
   // CHECK: %[[IDX:.*]] = arith.index_cast %[[CLAMPED]]
@@ -374,36 +368,30 @@ func.func @resize_bilinear_fp(%input: tensor<1x23x24x1xf32>) -> () {
   // CHECK-DAG: %[[X_MAX:.*]] = arith.constant 23
   // CHECK: %[[Y:.+]] = arith.index_cast %[[IDX_1]]
   // CHECK: %[[X:.+]] = arith.index_cast %[[IDX_2]]
-  // CHECK-DAG: %[[ISCALE_Y_N:.*]] = arith.constant 4
-  // CHECK-DAG: %[[ISCALE_Y_D:.*]] = arith.constant 1
-  // CHECK-DAG: %[[ISCALE_X_N:.*]] = arith.constant 4
-  // CHECK-DAG: %[[ISCALE_X_D:.*]] = arith.constant 1
-  // CHECK-DAG: %[[IOFFSET_Y:.*]] = arith.constant 0
-  // CHECK-DAG: %[[IOFFSET_X:.*]] = arith.constant 0
-  // CHECK-DAG: %[[IBORDER_Y:.*]] = arith.constant 0
-  // CHECK-DAG: %[[IBORDER_X:.*]] = arith.constant 0
-
-  // CHECK: %[[Y0:.+]] = arith.uitofp %[[Y]]
-  // CHECK: %[[SCALE_Y_N:.*]] = arith.uitofp %[[ISCALE_Y_N]]
-  // CHECK: %[[SCALE_Y_D:.*]] = arith.uitofp %[[ISCALE_Y_D]]
-  // CHECK: %[[OFFSET_Y:.*]] = arith.sitofp %[[IOFFSET_Y]]
-  // CHECK: %[[VAL_29:.*]] = arith.mulf %[[Y0]], %[[SCALE_Y_D]]
-  // CHECK: %[[VAL_31:.*]] = arith.addf %[[VAL_29]], %[[OFFSET_Y]]
-  // CHECK: %[[VAL_33:.*]] = arith.divf %[[VAL_31]], %[[SCALE_Y_N]]
-  // CHECK: %[[VAL_35:.*]] = math.floor %[[VAL_33]]
-  // CHECK: %[[D_Y:.*]] = arith.subf %[[VAL_33]], %[[VAL_35]]
-  // CHECK: %[[I_Y:.*]] = arith.fptosi %[[VAL_35]]
-
-  // CHECK: %[[X0:.+]] = arith.uitofp %[[X]]
-  // CHECK: %[[SCALE_X_N:.*]] = arith.uitofp %[[ISCALE_X_N]]
-  // CHECK: %[[SCALE_X_D:.*]] = arith.uitofp %[[ISCALE_X_D]]
-  // CHECK: %[[OFFSET_X:.*]] = arith.sitofp %[[IOFFSET_X]]
-  // CHECK: %[[VAL_30:.*]] = arith.mulf %[[X0]], %[[SCALE_X_D]]
-  // CHECK: %[[VAL_32:.*]] = arith.addf %[[VAL_30]], %[[OFFSET_X]]
-  // CHECK: %[[VAL_34:.*]] = arith.divf %[[VAL_32]], %[[SCALE_X_N]]
-  // CHECK: %[[VAL_36:.*]] = math.floor %[[VAL_34]]
-  // CHECK: %[[D_X:.*]] = arith.subf %[[VAL_34]], %[[VAL_36]]
-  // CHECK: %[[I_X:.*]] = arith.fptosi %[[VAL_36]]
+  // CHECK-DAG: %[[SCALE_Y_N:.*]] = arith.constant 4
+  // CHECK-DAG: %[[SCALE_Y_D:.*]] = arith.constant 1
+  // CHECK-DAG: %[[SCALE_X_N:.*]] = arith.constant 4
+  // CHECK-DAG: %[[SCALE_X_D:.*]] = arith.constant 1
+  // CHECK-DAG: %[[OFFSET_Y:.*]] = arith.constant 0
+  // CHECK-DAG: %[[OFFSET_X:.*]] = arith.constant 0
+  // CHECK-DAG: %[[BORDER_Y:.*]] = arith.constant 0
+  // CHECK-DAG: %[[BORDER_X:.*]] = arith.constant 0
+
+  // CHECK: %[[VAL_29:.*]] = arith.muli %[[Y]], %[[SCALE_Y_D]]
+  // CHECK: %[[Y_TEMP:.*]] = arith.addi %[[VAL_29]], %[[OFFSET_Y]]
+  // CHECK: %[[I_Y:.*]] = arith.floordivsi %[[Y_TEMP]], %[[SCALE_Y_N]]
+  // CHECK: %[[RY:.*]] = arith.remsi %[[Y_TEMP]], %[[SCALE_Y_N]]
+  // CHECK: %[[RY_FP:.*]] = arith.sitofp %[[RY]]
+  // CHECK: %[[SCALE_Y_N_FP:.*]] = arith.uitofp %[[SCALE_Y_N]]
+  // CHECK: %[[D_Y:.*]] = arith.divf %[[RY_FP]], %[[SCALE_Y_N_FP]]
+
+  // CHECK: %[[VAL_30:.*]] = arith.muli %[[X]], %[[SCALE_X_D]]
+  // CHECK: %[[X_TEMP:.*]] = arith.addi %[[VAL_30]], %[[OFFSET_X]]
+  // CHECK: %[[I_X:.*]] = arith.floordivsi %[[X_TEMP]], %[[SCALE_X_N]]
+  // CHECK: %[[RX:.*]] = arith.remsi %[[X_TEMP]], %[[SCALE_X_N]]
+  // CHECK: %[[RX_FP:.*]] = arith.sitofp %[[RX]]
+  // CHECK: %[[SCALE_X_N_FP:.*]] = arith.uitofp %[[SCALE_X_N]]
+  // CHECK: %[[D_X:.*]] = arith.divf %[[RX_FP]], %[[SCALE_X_N_FP]]
 
   // Compute the left, right, and top indices for the bilinear interpolation.
 

From 564f9abfcc3b99a01843f88b5a2c7309bfab5a33 Mon Sep 17 00:00:00 2001
From: Jorge Gorbe Moya <jgorbe@google.com>
Date: Wed, 17 Apr 2024 10:11:03 -0700
Subject: [PATCH 765/886] [bazel][mlir] Add missing dep after
 4f88c2311130791cf69da34b743b1b3ba7584a7b

---
 utils/bazel/llvm-project-overlay/mlir/BUILD.bazel | 1 +
 1 file changed, 1 insertion(+)

diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 03386549a0116..58538b66c5e0c 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -577,6 +577,7 @@ mlir_c_api_cc_library(
     ],
     includes = ["include"],
     deps = [
+        ":IR",
         ":NVGPUDialect",
     ],
 )

From e59632bdfd4c70caed437216af17d335858686fd Mon Sep 17 00:00:00 2001
From: Michael Maitland <michaeltmaitland@gmail.com>
Date: Wed, 17 Apr 2024 06:19:11 -0700
Subject: [PATCH 766/886] [RISCV] Fix typo in RISCVScheduleV.td that was
 introduced in 60a1158

---
 llvm/lib/Target/RISCV/RISCVScheduleV.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVScheduleV.td b/llvm/lib/Target/RISCV/RISCVScheduleV.td
index 5993884bc2c1c..5be06d4c3f7e7 100644
--- a/llvm/lib/Target/RISCV/RISCVScheduleV.td
+++ b/llvm/lib/Target/RISCV/RISCVScheduleV.td
@@ -1079,7 +1079,7 @@ defm "" : LMULReadAdvance<"ReadVFCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceW<"ReadVFWCvtIToFV", 0>;
 defm "" : LMULReadAdvanceFW<"ReadVFWCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFWCvtFToFV", 0>;
-defm "" : LMULSEWReadAdvanceFW<"SEWReadVFNCvtIToFV", 0>;
+defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtIToFV", 0>;
 defm "" : LMULReadAdvanceW<"ReadVFNCvtFToIV", 0>;
 defm "" : LMULSEWReadAdvanceFW<"ReadVFNCvtFToFV", 0>;
 

From 676d3bafc09d0c331a04b813804407334de12917 Mon Sep 17 00:00:00 2001
From: Jorge Gorbe Moya <jgorbe@google.com>
Date: Wed, 17 Apr 2024 10:20:07 -0700
Subject: [PATCH 767/886] [bazel][libc] Add missing dep after
 b854a2323337be2633b1135f590678a17e9d1ade

---
 utils/bazel/llvm-project-overlay/libc/BUILD.bazel | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index 6029cc3fee610..be02c22704321 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -786,7 +786,7 @@ libc_support_library(
         ":errno",
         ":hdr_fenv_macros",
         ":hdr_math_macros",
-	":types_fenv_t"
+        ":types_fenv_t",
     ],
 )
 
@@ -1152,6 +1152,7 @@ libc_function(
     deps = [
         ":__support_common",
         ":__support_fputil_fenv_impl",
+        ":types_fexcept_t",
     ],
 )
 
@@ -1272,7 +1273,7 @@ libc_function(
     deps = [
         ":__support_common",
         ":__support_fputil_fenv_impl",
-	":types_fexcept_t"
+        ":types_fexcept_t",
     ],
 )
 
@@ -1283,7 +1284,7 @@ libc_function(
     deps = [
         ":__support_common",
         ":__support_fputil_fenv_impl",
-	":types_fexcept_t",
+        ":types_fexcept_t",
     ],
 )
 

From 693a458287d019c5c6a66fe3019d099df2978cdb Mon Sep 17 00:00:00 2001
From: Abdul Raheem <55028856+abdulraheembeigh@users.noreply.github.com>
Date: Wed, 17 Apr 2024 18:24:04 +0100
Subject: [PATCH 768/886] [MLIR] Update doc comment in ViewLikeInterface.td
 (NFC) (#89074)

Signed-off: Abdul Raheem Beigh <abdulraheembeigh@gmail.com>
---
 mlir/include/mlir/Interfaces/ViewLikeInterface.td | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/include/mlir/Interfaces/ViewLikeInterface.td b/mlir/include/mlir/Interfaces/ViewLikeInterface.td
index ea5bb1b5ac485..9397f271e1bc6 100644
--- a/mlir/include/mlir/Interfaces/ViewLikeInterface.td
+++ b/mlir/include/mlir/Interfaces/ViewLikeInterface.td
@@ -158,7 +158,7 @@ def OffsetSizeAndStrideOpInterface : OpInterface<"OffsetSizeAndStrideOpInterface
     >,
     InterfaceMethod<
       /*desc=*/[{
-        Return a vector of all the static or dynamic sizes of the op.
+        Return a vector of all the static or dynamic offsets of the op.
       }],
       /*retTy=*/"::llvm::SmallVector<::mlir::OpFoldResult, 4>",
       /*methodName=*/"getMixedOffsets",

From 6f7160eedb2db02f37d4ffd52fff7b0cf88b3fdc Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Tue, 16 Apr 2024 14:55:41 -0400
Subject: [PATCH 769/886] [SLP]Attempt to vectorize long stores, if short one
 failed.

We can try to vectorize long store sequences, if short ones were
unsuccessful because of the non-profitable vectorization. It should not
increase compile time significantly (stores are sorted already,
complexity is n x log n), but vectorize extra code.

Metric: size..text

Program                                                                         size..text
                                                                                results     results0    diff
         test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test  1088012.00  1088236.00  0.0%
                  test-suite :: SingleSource/UnitTests/matrix-types-spec.test   480396.00   480476.00  0.0%
          test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   664613.00   664661.00  0.0%
         test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   664613.00   664661.00  0.0%
        test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2041105.00  2040961.00 -0.0%
                 test-suite :: MultiSource/Applications/JM/lencod/lencod.test   836563.00   836387.00 -0.0%
                 test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test  1035100.00  1032140.00 -0.3%

In all benchmarks extra code gets vectorized

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/88563
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 106 ++++++++++++------
 .../Transforms/SLPVectorizer/X86/pr46983.ll   |  46 ++------
 2 files changed, 80 insertions(+), 72 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 7694627c3b043..806e8085038b3 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15164,10 +15164,6 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
   BoUpSLP::ValueSet VectorizedStores;
   bool Changed = false;
 
-  // Stores the pair of stores (first_store, last_store) in a range, that were
-  // already tried to be vectorized. Allows to skip the store ranges that were
-  // already tried to be vectorized but the attempts were unsuccessful.
-  DenseSet<std::pair<Value *, Value *>> TriedSequences;
   struct StoreDistCompare {
     bool operator()(const std::pair<unsigned, int> &Op1,
                     const std::pair<unsigned, int> &Op2) const {
@@ -15209,8 +15205,10 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
       Type *ValueTy = StoreTy;
       if (auto *Trunc = dyn_cast<TruncInst>(Store->getValueOperand()))
         ValueTy = Trunc->getSrcTy();
-      unsigned MinVF = PowerOf2Ceil(TTI->getStoreMinimumVF(
-          R.getMinVF(DL->getTypeStoreSizeInBits(StoreTy)), StoreTy, ValueTy));
+      unsigned MinVF = std::max<unsigned>(
+          2, PowerOf2Ceil(TTI->getStoreMinimumVF(
+                 R.getMinVF(DL->getTypeStoreSizeInBits(StoreTy)), StoreTy,
+                 ValueTy)));
 
       if (MaxVF < MinVF) {
         LLVM_DEBUG(dbgs() << "SLP: Vectorization infeasible as MaxVF (" << MaxVF
@@ -15236,40 +15234,74 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
         VF = Size > MaxVF ? NonPowerOf2VF : Size;
         Size *= 2;
       });
-      unsigned StartIdx = 0;
-      for (unsigned Size : CandidateVFs) {
-        for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
-          ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
-          assert(
-              all_of(
-                  Slice,
-                  [&](Value *V) {
-                    return cast<StoreInst>(V)->getValueOperand()->getType() ==
-                           cast<StoreInst>(Slice.front())
-                               ->getValueOperand()
-                               ->getType();
-                  }) &&
-              "Expected all operands of same type.");
-          if (!VectorizedStores.count(Slice.front()) &&
-              !VectorizedStores.count(Slice.back()) &&
-              TriedSequences.insert(std::make_pair(Slice.front(), Slice.back()))
-                  .second &&
-              vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
-            // Mark the vectorized stores so that we don't vectorize them again.
-            VectorizedStores.insert(Slice.begin(), Slice.end());
-            Changed = true;
-            // If we vectorized initial block, no need to try to vectorize it
-            // again.
-            if (Cnt == StartIdx)
-              StartIdx += Size;
-            Cnt += Size;
-            continue;
+      unsigned End = Operands.size();
+      unsigned Repeat = 0;
+      constexpr unsigned MaxAttempts = 2;
+      SmallBitVector Range(Operands.size());
+      while (true) {
+        ++Repeat;
+        for (unsigned Size : CandidateVFs) {
+          int StartIdx = Range.find_first_unset();
+          while (StartIdx != -1) {
+            int EndIdx = Range.find_next(StartIdx);
+            unsigned Sz = EndIdx == -1 ? End : EndIdx;
+            for (unsigned Cnt = StartIdx; Cnt + Size <= Sz;) {
+              ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
+              assert(all_of(Slice,
+                            [&](Value *V) {
+                              return cast<StoreInst>(V)
+                                         ->getValueOperand()
+                                         ->getType() ==
+                                     cast<StoreInst>(Slice.front())
+                                         ->getValueOperand()
+                                         ->getType();
+                            }) &&
+                     "Expected all operands of same type.");
+              if (vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
+                // Mark the vectorized stores so that we don't vectorize them
+                // again.
+                VectorizedStores.insert(Slice.begin(), Slice.end());
+                // Mark the vectorized stores so that we don't vectorize them
+                // again.
+                Changed = true;
+                // If we vectorized initial block, no need to try to vectorize
+                // it again.
+                Range.set(Cnt, Cnt + Size);
+                if (Cnt < StartIdx + MinVF)
+                  Range.set(StartIdx, Cnt);
+                if (Cnt > EndIdx - Size - MinVF) {
+                  Range.set(Cnt + Size, EndIdx);
+                  End = Cnt;
+                }
+                Cnt += Size;
+                continue;
+              }
+              ++Cnt;
+            }
+            if (Sz >= End)
+              break;
+            StartIdx = Range.find_next_unset(EndIdx);
           }
-          ++Cnt;
         }
-        // Check if the whole array was vectorized already - exit.
-        if (StartIdx >= Operands.size())
+        // All values vectorize - exit.
+        if (Range.all())
+          break;
+        // Check if tried all attempts or no need for the last attempts at all.
+        if (Repeat >= MaxAttempts)
+          break;
+        constexpr unsigned MaxVFScale = 4;
+        constexpr unsigned StoresLimit = 16;
+        const unsigned MaxTotalNum = std::min(
+            std::max<unsigned>(StoresLimit, MaxVFScale * MaxVF),
+            bit_floor(static_cast<unsigned>(Range.find_last_unset() -
+                                            Range.find_first_unset() + 1)));
+        if (MaxVF >= MaxTotalNum)
           break;
+        // Last attempt to vectorize max number of elements, if all previous
+        // attempts were unsuccessful because of the cost issues.
+        CandidateVFs.clear();
+        for (unsigned Size = MaxTotalNum; Size > MaxVF; Size /= 2)
+          CandidateVFs.push_back(Size);
       }
     }
   };
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
index 75505f632a43f..3deab0975ce76 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
@@ -100,41 +100,17 @@ define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
 define void @store_i64(ptr nocapture %0, i32 %1, i32 %2) {
 ; SSE-LABEL: @store_i64(
 ; SSE-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP1:%.*]] to i64
-; SSE-NEXT:    [[TMP5:%.*]] = load i64, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
-; SSE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], [[TMP4]]
-; SSE-NEXT:    [[TMP7:%.*]] = lshr i64 [[TMP6]], 15
-; SSE-NEXT:    [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
-; SSE-NEXT:    [[TMP9:%.*]] = icmp ult i32 [[TMP8]], 255
-; SSE-NEXT:    [[TMP10:%.*]] = and i64 [[TMP7]], 4294967295
-; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i64 [[TMP10]], i64 255
-; SSE-NEXT:    store i64 [[TMP11]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 8
-; SSE-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], [[TMP4]]
-; SSE-NEXT:    [[TMP15:%.*]] = lshr i64 [[TMP14]], 15
-; SSE-NEXT:    [[TMP16:%.*]] = trunc i64 [[TMP15]] to i32
-; SSE-NEXT:    [[TMP17:%.*]] = icmp ult i32 [[TMP16]], 255
-; SSE-NEXT:    [[TMP18:%.*]] = and i64 [[TMP15]], 4294967295
-; SSE-NEXT:    [[TMP19:%.*]] = select i1 [[TMP17]], i64 [[TMP18]], i64 255
-; SSE-NEXT:    store i64 [[TMP19]], ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 16
-; SSE-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], [[TMP4]]
-; SSE-NEXT:    [[TMP23:%.*]] = lshr i64 [[TMP22]], 15
-; SSE-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
-; SSE-NEXT:    [[TMP25:%.*]] = icmp ult i32 [[TMP24]], 255
-; SSE-NEXT:    [[TMP26:%.*]] = and i64 [[TMP23]], 4294967295
-; SSE-NEXT:    [[TMP27:%.*]] = select i1 [[TMP25]], i64 [[TMP26]], i64 255
-; SSE-NEXT:    store i64 [[TMP27]], ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 24
-; SSE-NEXT:    [[TMP29:%.*]] = load i64, ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
-; SSE-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], [[TMP4]]
-; SSE-NEXT:    [[TMP31:%.*]] = lshr i64 [[TMP30]], 15
-; SSE-NEXT:    [[TMP32:%.*]] = trunc i64 [[TMP31]] to i32
-; SSE-NEXT:    [[TMP33:%.*]] = icmp ult i32 [[TMP32]], 255
-; SSE-NEXT:    [[TMP34:%.*]] = and i64 [[TMP31]], 4294967295
-; SSE-NEXT:    [[TMP35:%.*]] = select i1 [[TMP33]], i64 [[TMP34]], i64 255
-; SSE-NEXT:    store i64 [[TMP35]], ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP5:%.*]] = load <4 x i64>, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
+; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x i64> poison, i64 [[TMP4]], i64 0
+; SSE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:    [[TMP8:%.*]] = mul <4 x i64> [[TMP5]], [[TMP7]]
+; SSE-NEXT:    [[TMP9:%.*]] = lshr <4 x i64> [[TMP8]], <i64 15, i64 15, i64 15, i64 15>
+; SSE-NEXT:    [[TMP10:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
+; SSE-NEXT:    [[TMP11:%.*]] = icmp ult <4 x i32> [[TMP10]], <i32 255, i32 255, i32 255, i32 255>
+; SSE-NEXT:    [[TMP12:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
+; SSE-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>
+; SSE-NEXT:    [[TMP14:%.*]] = zext <4 x i32> [[TMP13]] to <4 x i64>
+; SSE-NEXT:    store <4 x i64> [[TMP14]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @store_i64(

From eefee382186005d3662958e076c8e61e286ea1ab Mon Sep 17 00:00:00 2001
From: Nick Desaulniers <nickdesaulniers@users.noreply.github.com>
Date: Wed, 17 Apr 2024 10:34:23 -0700
Subject: [PATCH 770/886] [libc] set cmake dependencies for condattr test
 (#89103)

The entrypoints are not yet exposed on non-x86. Express this dependency
to
unbreak post submit.

Fixes #88987
---
 libc/src/pthread/pthread_condattr_destroy.cpp | 3 ++-
 libc/test/src/pthread/CMakeLists.txt          | 6 ++++++
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/libc/src/pthread/pthread_condattr_destroy.cpp b/libc/src/pthread/pthread_condattr_destroy.cpp
index 45cc011a4a92d..41994c6941ffe 100644
--- a/libc/src/pthread/pthread_condattr_destroy.cpp
+++ b/libc/src/pthread/pthread_condattr_destroy.cpp
@@ -14,7 +14,8 @@
 
 namespace LIBC_NAMESPACE {
 
-LLVM_LIBC_FUNCTION(int, pthread_condattr_destroy, (pthread_condattr_t * attr)) {
+LLVM_LIBC_FUNCTION(int, pthread_condattr_destroy,
+                   (pthread_condattr_t * attr [[gnu::unused]])) {
   // Initializing a pthread_condattr_t acquires no resources, so this is a
   // no-op.
   return 0;
diff --git a/libc/test/src/pthread/CMakeLists.txt b/libc/test/src/pthread/CMakeLists.txt
index 46f38422a5374..51954a5babd2c 100644
--- a/libc/test/src/pthread/CMakeLists.txt
+++ b/libc/test/src/pthread/CMakeLists.txt
@@ -50,4 +50,10 @@ add_libc_unittest(
     libc.include.errno
     libc.include.pthread
     libc.include.time
+    libc.src.pthread.pthread_condattr_destroy
+    libc.src.pthread.pthread_condattr_getclock
+    libc.src.pthread.pthread_condattr_getpshared
+    libc.src.pthread.pthread_condattr_init
+    libc.src.pthread.pthread_condattr_setclock
+    libc.src.pthread.pthread_condattr_setpshared
   )

From 825536039d667eeb933c590fe40c358fdea03a8d Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 17 Apr 2024 16:38:31 +0100
Subject: [PATCH 771/886] [CostModel][X86] Add basic GFNI target test coverage
 for shift/rotate costs

---
 .../Analysis/CostModel/X86/fshl-codesize.ll   | 181 +++++++++++++
 .../Analysis/CostModel/X86/fshl-latency.ll    | 181 +++++++++++++
 .../CostModel/X86/fshl-sizelatency.ll         | 241 ++++++++++++++++++
 llvm/test/Analysis/CostModel/X86/fshl.ll      | 181 +++++++++++++
 .../Analysis/CostModel/X86/fshr-codesize.ll   | 181 +++++++++++++
 .../Analysis/CostModel/X86/fshr-latency.ll    | 181 +++++++++++++
 .../CostModel/X86/fshr-sizelatency.ll         | 241 ++++++++++++++++++
 llvm/test/Analysis/CostModel/X86/fshr.ll      | 181 +++++++++++++
 .../CostModel/X86/vshift-ashr-codesize.ll     |  85 ++++++
 .../X86/vshift-ashr-cost-inseltpoison.ll      |  87 +++++++
 .../CostModel/X86/vshift-ashr-cost.ll         |  87 +++++++
 .../CostModel/X86/vshift-ashr-latency.ll      |  93 +++++++
 .../CostModel/X86/vshift-ashr-sizelatency.ll  |  93 +++++++
 .../CostModel/X86/vshift-lshr-codesize.ll     |  65 +++++
 .../X86/vshift-lshr-cost-inseltpoison.ll      |  77 ++++++
 .../CostModel/X86/vshift-lshr-cost.ll         |  77 ++++++
 .../CostModel/X86/vshift-lshr-latency.ll      |  75 ++++++
 .../CostModel/X86/vshift-lshr-sizelatency.ll  |  77 ++++++
 .../CostModel/X86/vshift-shl-codesize.ll      |  57 +++++
 .../X86/vshift-shl-cost-inseltpoison.ll       |  81 ++++++
 .../Analysis/CostModel/X86/vshift-shl-cost.ll |  81 ++++++
 .../CostModel/X86/vshift-shl-latency.ll       |  61 +++++
 .../CostModel/X86/vshift-shl-sizelatency.ll   |  73 ++++++
 23 files changed, 2737 insertions(+)

diff --git a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
index c212ecfc3eb90..c46e32ffb4ad3 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1690,6 +1811,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1775,6 +1903,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -2023,6 +2158,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2123,6 +2267,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2335,6 +2488,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2420,6 +2580,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2616,6 +2783,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2694,6 +2868,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
index 487adfe79d944..fa32497c63ec7 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -939,6 +1011,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1024,6 +1103,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1109,6 +1195,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1198,6 +1291,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1276,6 +1376,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1354,6 +1461,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1432,6 +1546,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1642,6 +1763,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1727,6 +1855,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1975,6 +2110,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2075,6 +2219,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2287,6 +2440,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2372,6 +2532,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2568,6 +2735,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2646,6 +2820,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
index 52ad7e13c84c6..832a574a9b332 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1562,6 +1683,13 @@ define void @var_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
@@ -1647,6 +1775,13 @@ define void @var_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
@@ -1732,6 +1867,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1817,6 +1959,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1919,6 +2068,15 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -2019,6 +2177,15 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -2119,6 +2286,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2219,6 +2395,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2303,6 +2488,13 @@ define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
@@ -2388,6 +2580,13 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -2473,6 +2672,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2558,6 +2764,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2640,6 +2853,13 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
@@ -2718,6 +2938,13 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %a32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -2796,6 +3023,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2874,6 +3108,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index 4e688c29c7ea3..311d8d5ed7d2a 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -932,6 +1004,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1017,6 +1096,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1102,6 +1188,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1191,6 +1284,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshl.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1269,6 +1369,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshl.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1347,6 +1454,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1425,6 +1539,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1635,6 +1756,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1720,6 +1848,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1968,6 +2103,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2068,6 +2212,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2273,6 +2426,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2358,6 +2518,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2554,6 +2721,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2632,6 +2806,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
index 0e76246fad85f..f9d30e4ced3ec 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1690,6 +1811,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1775,6 +1903,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -2023,6 +2158,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2123,6 +2267,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2335,6 +2488,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2420,6 +2580,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2616,6 +2783,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2694,6 +2868,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
index 73e32dc92a69b..ed2227591847a 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 51 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -939,6 +1011,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1024,6 +1103,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1109,6 +1195,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 57 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1198,6 +1291,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1276,6 +1376,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1354,6 +1461,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1432,6 +1546,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1642,6 +1763,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1727,6 +1855,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1975,6 +2110,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2075,6 +2219,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2287,6 +2440,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2372,6 +2532,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2568,6 +2735,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2646,6 +2820,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
index 2d106c6dd3069..8931781f70bdc 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -572,6 +610,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -681,6 +728,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 66 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -790,6 +846,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -881,6 +946,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -966,6 +1038,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1051,6 +1130,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1136,6 +1222,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1225,6 +1318,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1310,6 +1410,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 26 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 52 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1395,6 +1502,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 54 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1480,6 +1594,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 58 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1562,6 +1683,13 @@ define void @var_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %c128)
@@ -1647,6 +1775,13 @@ define void @var_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %c128)
@@ -1732,6 +1867,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1817,6 +1959,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1919,6 +2068,15 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -2019,6 +2177,15 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -2119,6 +2286,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2219,6 +2395,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2303,6 +2488,13 @@ define void @constant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 1, i64 7>)
@@ -2388,6 +2580,13 @@ define void @constant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -2473,6 +2672,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2558,6 +2764,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2640,6 +2853,13 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
@@ -2718,6 +2938,13 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -2796,6 +3023,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2874,6 +3108,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index 9565630677d52..ca9ddcc52938d 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -12,6 +12,7 @@
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=bdver2 | FileCheck %s --check-prefixes=XOP
 ; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
@@ -97,6 +98,13 @@ define void @var_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8 x i64
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 %c64)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %c128)
@@ -182,6 +190,13 @@ define void @var_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16 x i3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 %c32)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %c128)
@@ -267,6 +282,13 @@ define void @var_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %c128)
@@ -352,6 +374,13 @@ define void @var_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %c128)
@@ -463,6 +492,15 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
   %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
@@ -563,6 +601,15 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
   %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
@@ -663,6 +710,15 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -763,6 +819,15 @@ define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -854,6 +919,13 @@ define void @constant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 1, i64 7, i64 15, i64 31, i64 1, i64 7, i64 15, i64 31>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 1, i64 7>)
@@ -932,6 +1004,13 @@ define void @constant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 7)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 4, i32 5, i32 6, i32 7>)
@@ -1017,6 +1096,13 @@ define void @constant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -1102,6 +1188,13 @@ define void @constant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -1191,6 +1284,13 @@ define void @splatconstant_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i64'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %b64, i64 7)
   %V2I64  = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> <i64 7, i64 7>)
@@ -1269,6 +1369,13 @@ define void @splatconstant_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i32'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %b32, i32 5)
   %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
@@ -1347,6 +1454,13 @@ define void @splatconstant_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %b16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -1425,6 +1539,13 @@ define void @splatconstant_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_funnel_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %b8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
@@ -1635,6 +1756,13 @@ define void @var_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 %c16)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %c128)
@@ -1720,6 +1848,13 @@ define void @var_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'var_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %c256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %c512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 %c8)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %c128)
@@ -1968,6 +2103,15 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
   %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
@@ -2068,6 +2212,15 @@ define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
   %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -2273,6 +2426,13 @@ define void @constant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 7)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>)
@@ -2358,6 +2518,13 @@ define void @constant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'constant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 7)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
@@ -2554,6 +2721,13 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I16    = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
   %V8I16  = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
@@ -2632,6 +2806,13 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
   %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
index 8e3f38c796e6e..a3c24bdd1a882 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -327,6 +340,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -389,6 +406,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -778,6 +799,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -827,6 +854,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -882,6 +915,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -937,6 +976,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1158,6 +1203,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1191,6 +1240,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1224,6 +1277,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1253,6 +1310,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1315,6 +1376,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1531,6 +1596,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1568,6 +1637,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1601,6 +1674,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1634,6 +1711,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1667,6 +1748,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
index d4ece3b2a1134..67679f2cb8566 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector arithmetic shift right instructions.
 
@@ -228,6 +229,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -269,6 +274,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -310,6 +319,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -392,6 +409,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -433,6 +454,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -950,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
@@ -1017,6 +1054,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1226,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1267,6 +1314,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1308,6 +1359,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1349,6 +1404,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1390,6 +1449,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1431,6 +1494,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1631,6 +1698,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1672,6 +1743,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1713,6 +1788,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1754,6 +1833,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1795,6 +1878,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
index f8fb2d76a778f..efd378ee0b5a6 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector arithmetic shift right instructions.
 
@@ -228,6 +229,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -269,6 +274,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -310,6 +319,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -392,6 +409,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -433,6 +454,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -950,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -1017,6 +1054,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1226,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1267,6 +1314,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1308,6 +1359,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1349,6 +1404,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1390,6 +1449,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1431,6 +1494,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1631,6 +1698,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1672,6 +1743,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1713,6 +1788,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1754,6 +1833,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1795,6 +1878,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
index af5278e355cfb..cd4189d4a7f84 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -331,6 +344,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -364,6 +381,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -397,6 +418,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +859,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -889,6 +920,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -944,6 +981,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -999,6 +1042,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1220,6 +1269,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1253,6 +1306,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1286,6 +1343,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1319,6 +1380,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1352,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1385,6 +1454,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1649,6 +1722,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1686,6 +1763,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1723,6 +1804,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1760,6 +1845,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1797,6 +1886,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
index 8286b3967af60..84ccad0294155 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -265,6 +270,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -298,6 +307,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -331,6 +344,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -364,6 +381,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -397,6 +418,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -786,6 +811,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = ashr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -835,6 +866,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %shift = ashr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -890,6 +927,12 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = ashr <32 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %insert = insertelement <32 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
@@ -945,6 +988,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = ashr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1166,6 +1215,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = ashr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1199,6 +1252,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1232,6 +1289,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1265,6 +1326,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1298,6 +1363,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1331,6 +1400,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1547,6 +1620,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = ashr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1584,6 +1661,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = ashr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1617,6 +1698,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = ashr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1654,6 +1739,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1691,6 +1780,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = ashr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
index 7c506af62e817..a0e15bb8ff738 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -397,6 +414,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -762,6 +783,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -909,6 +936,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1138,6 +1171,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1171,6 +1208,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1204,6 +1245,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1233,6 +1278,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1295,6 +1344,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1495,6 +1548,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1532,6 +1589,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1631,6 +1692,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
index 855d2b8e88e3a..34ea1a644f6cc 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -355,6 +368,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -396,6 +413,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -437,6 +458,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -826,6 +851,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -991,6 +1022,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1200,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1282,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1282,6 +1327,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1323,6 +1372,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1364,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1405,6 +1462,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1593,6 +1654,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1634,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1700,6 +1769,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1741,6 +1814,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
index 55894dc932957..93b844eddf182 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -232,6 +233,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -355,6 +368,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -396,6 +413,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -437,6 +458,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BWVL-LABEL: 'var_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -826,6 +851,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -991,6 +1022,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1200,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v8i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1282,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1282,6 +1327,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1323,6 +1372,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1364,6 +1417,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1405,6 +1462,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'constant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1593,6 +1654,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1634,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1700,6 +1769,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1741,6 +1814,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
index 92fe253d38e73..61620e2cc97ed 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -277,6 +282,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -421,6 +438,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -834,6 +855,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -889,6 +916,12 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <16 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %insert = insertelement <16 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
@@ -993,6 +1026,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1222,6 +1261,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1259,6 +1302,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1296,6 +1343,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1333,6 +1384,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1403,6 +1458,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1643,6 +1702,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1676,6 +1739,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1709,6 +1776,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1771,6 +1842,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
index fe8a7d91cc375..e6b6ac75b65d5 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -368,6 +385,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -401,6 +422,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -766,6 +791,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -913,6 +944,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %shift = lshr <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1142,6 +1179,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = lshr <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1175,6 +1216,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1208,6 +1253,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1237,6 +1286,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = lshr <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1270,6 +1323,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1303,6 +1360,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1503,6 +1564,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = lshr <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1540,6 +1605,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = lshr <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1606,6 +1675,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = lshr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1643,6 +1716,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = lshr <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
index 07229e22873c4..265658b1e3a2d 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-codesize.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=code-size -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -273,6 +278,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -306,6 +315,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -335,6 +348,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -397,6 +414,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -762,6 +783,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -915,6 +942,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1186,6 +1219,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1215,6 +1252,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1277,6 +1318,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1477,6 +1522,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1514,6 +1563,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1613,6 +1666,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
index d8737d2365649..f093d1c8ca358 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector shift left instructions.
 
@@ -236,6 +237,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -281,6 +286,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -326,6 +335,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -371,6 +384,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -416,6 +433,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -461,6 +482,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -868,6 +893,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> poison, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
@@ -1051,6 +1082,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> poison, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
@@ -1305,6 +1342,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1354,6 +1395,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1399,6 +1444,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1444,6 +1493,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1489,6 +1542,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1677,6 +1734,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1718,6 +1779,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1784,6 +1849,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1825,6 +1894,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
@@ -2044,6 +2117,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
 ; SLM-LABEL: 'test6'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test6'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
 ;
   %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <16 x i16> %shl
@@ -2162,6 +2239,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
 ; SLM-LABEL: 'test9'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test9'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
 ;
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
index b254ff27daccb..09521cfa2d23d 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector shift left instructions.
 
@@ -236,6 +237,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -281,6 +286,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -326,6 +335,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-LABEL: 'var_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -371,6 +384,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -416,6 +433,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, %b
   ret <32 x i8> %shift
@@ -461,6 +482,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-LABEL: 'var_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, %b
   ret <64 x i8> %shift
@@ -868,6 +893,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -1051,6 +1082,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1305,6 +1342,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1354,6 +1395,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; SLM-LABEL: 'constant_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1399,6 +1444,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1444,6 +1493,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <32 x i8> %shift
@@ -1489,6 +1542,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-LABEL: 'constant_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <64 x i8> %shift
@@ -1677,6 +1734,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1718,6 +1779,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1784,6 +1849,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1825,6 +1894,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BWVL-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BWVL-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
@@ -2044,6 +2117,10 @@ define <16 x i16> @test6(<16 x i16> %a) {
 ; SLM-LABEL: 'test6'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test6'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %shl
 ;
   %shl = shl <16 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <16 x i16> %shl
@@ -2162,6 +2239,10 @@ define <32 x i16> @test9(<32 x i16> %a) {
 ; SLM-LABEL: 'test9'
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
+;
+; AVX512GFNI-LABEL: 'test9'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %shl
 ;
   %shl = shl <32 x i16> %a, <i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11, i16 1, i16 1, i16 2, i16 3, i16 7, i16 0, i16 9, i16 11>
   ret <32 x i16> %shl
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
index c53dbc5f586aa..42c91144ff6f9 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-latency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=latency -mcpu=tigerlake | FileCheck %s --check-prefixes=AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -240,6 +241,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, %b
   ret <8 x i16> %shift
@@ -277,6 +282,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, %b
   ret <16 x i16> %shift
@@ -314,6 +323,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; AVX512BW-LABEL: 'var_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, %b
   ret <32 x i16> %shift
@@ -351,6 +364,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; AVX512BW-LABEL: 'var_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, %b
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, %b
   ret <16 x i8> %shift
@@ -830,6 +847,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -983,6 +1006,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1208,6 +1237,10 @@ define <8 x i16> @constant_shift_v8i16(<8 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v8i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
   %shift = shl <8 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <8 x i16> %shift
@@ -1241,6 +1274,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1274,6 +1311,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1311,6 +1352,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
   ret <16 x i8> %shift
@@ -1617,6 +1662,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1650,6 +1699,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1683,6 +1736,10 @@ define <16 x i8> @splatconstant_shift_v16i8(<16 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
   %shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <16 x i8> %shift
@@ -1745,6 +1802,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
index cbf0f24b90fb8..47b24df063ef7 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-sizelatency.ll
@@ -15,6 +15,7 @@
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=slm | FileCheck %s --check-prefixes=CHECK,SSE,SLM
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=goldmont | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
 ; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=btver2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=size-latency -mcpu=tigerlake | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512GFNI
 
 ; Verify the cost of vector logical shift right instructions.
 
@@ -263,6 +264,10 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v8i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <8 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v8i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %shift = shl <8 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <8 x i16> %shift
@@ -303,6 +308,10 @@ define <16 x i16> @var_shift_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v16i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %shift = shl <16 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
@@ -343,6 +352,10 @@ define <32 x i16> @var_shift_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i16> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v32i16'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %shift = shl <32 x i16> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
@@ -383,6 +396,10 @@ define <16 x i8> @var_shift_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <16 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v16i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %shift = shl <16 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
@@ -423,6 +440,10 @@ define <32 x i8> @var_shift_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <32 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v32i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <32 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
@@ -463,6 +484,10 @@ define <64 x i8> @var_shift_v64i8(<64 x i8> %a, <64 x i8> %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %shift = shl <64 x i8> %a, %b
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'var_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <64 x i8> %a, %b
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
 ; BTVER2-LABEL: 'var_shift_v64i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %shift = shl <64 x i8> %a, %b
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
@@ -824,6 +849,12 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i16> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %insert = insertelement <32 x i16> undef, i16 %b, i32 0
   %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
@@ -995,6 +1026,12 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %shift = shl <64 x i8> %a, %splat
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %insert = insertelement <64 x i8> undef, i8 %b, i32 0
   %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
@@ -1238,6 +1275,10 @@ define <16 x i16> @constant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <16 x i16> %shift
@@ -1271,6 +1312,10 @@ define <32 x i16> @constant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'constant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'constant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 0, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>
   ret <32 x i16> %shift
@@ -1309,6 +1354,10 @@ define <16 x i8> @constant_shift_v16i8(<16 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v16i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v16i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %shift = shl <16 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i8> %shift
@@ -1349,6 +1398,10 @@ define <32 x i8> @constant_shift_v32i8(<32 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v32i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %shift = shl <32 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
@@ -1389,6 +1442,10 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) {
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
+; AVX512GFNI-LABEL: 'constant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
 ; BTVER2-LABEL: 'constant_shift_v64i8'
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %shift = shl <64 x i8> %a, <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>
 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
@@ -1571,6 +1628,10 @@ define <16 x i16> @splatconstant_shift_v16i16(<16 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v16i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v16i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <16 x i16> %shift
 ;
   %shift = shl <16 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <16 x i16> %shift
@@ -1604,6 +1665,10 @@ define <32 x i16> @splatconstant_shift_v32i16(<32 x i16> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i16'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i16> %shift
 ;
   %shift = shl <32 x i16> %a, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
   ret <32 x i16> %shift
@@ -1662,6 +1727,10 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v32i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <32 x i8> %shift
 ;
   %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
@@ -1695,6 +1764,10 @@ define <64 x i8> @splatconstant_shift_v64i8(<64 x i8> %a) {
 ; AVX512BW-LABEL: 'splatconstant_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
+;
+; AVX512GFNI-LABEL: 'splatconstant_shift_v64i8'
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret <64 x i8> %shift
 ;
   %shift = shl <64 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <64 x i8> %shift

From da04e4afd3cae13581cac85688fbf10a5848655f Mon Sep 17 00:00:00 2001
From: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed, 17 Apr 2024 12:05:26 -0500
Subject: [PATCH 772/886] [InstCombine] Use `auto *` instead of `auto` in
 `visitSIToFP`; NFC

---
 llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index d242d3f443def..4537a47da2ced 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -1990,7 +1990,7 @@ Instruction *InstCombinerImpl::visitSIToFP(CastInst &CI) {
   if (Instruction *R = commonCastTransforms(CI))
     return R;
   if (isKnownNonNegative(CI.getOperand(0), SQ)) {
-    auto UI =
+    auto *UI =
         CastInst::Create(Instruction::UIToFP, CI.getOperand(0), CI.getType());
     UI->setNonNeg(true);
     return UI;

From d423d80e560d8bf7ca493596d9f34a9e1f0eede7 Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Wed, 17 Apr 2024 13:36:53 -0400
Subject: [PATCH 773/886] [libc++][pstl] Promote CPU backends to top-level
 backends (#88968)

This patch removes the two-level backend dispatching mechanism we had in
the PSTL. Instead of selecting both a PSTL backend and a PSTL CPU
backend, we now only select a top-level PSTL backend. This greatly
simplifies the PSTL configuration layer.

While this patch technically removes some flexibility from the PSTL
configuration mechanism because CPU backends are not considered
separately, it opens the door to a much more powerful configuration
mechanism based on chained backends in a follow-up patch.

This is a step towards overhauling the PSTL dispatching mechanism.
---
 libcxx/CMakeLists.txt                         | 18 +++----
 libcxx/cmake/caches/Apple.cmake               |  2 +-
 libcxx/include/CMakeLists.txt                 | 11 ++---
 .../__algorithm/pstl_backends/cpu_backend.h   | 23 ---------
 .../pstl_backends/cpu_backends/any_of.h       |  2 +-
 .../pstl_backends/cpu_backends/backend.h      | 45 ------------------
 .../pstl_backends/cpu_backends/fill.h         |  2 +-
 .../pstl_backends/cpu_backends/find_if.h      |  2 +-
 .../pstl_backends/cpu_backends/for_each.h     |  2 +-
 .../pstl_backends/cpu_backends/merge.h        |  2 +-
 .../pstl_backends/cpu_backends/stable_sort.h  |  2 +-
 .../pstl_backends/cpu_backends/transform.h    |  2 +-
 .../cpu_backends/transform_reduce.h           |  2 +-
 libcxx/include/__algorithm/pstl_copy.h        |  2 +-
 libcxx/include/__algorithm/pstl_count.h       |  2 +-
 libcxx/include/__algorithm/pstl_find.h        |  2 +-
 libcxx/include/__algorithm/pstl_for_each.h    |  2 +-
 libcxx/include/__algorithm/pstl_generate.h    |  2 +-
 .../include/__algorithm/pstl_is_partitioned.h |  2 +-
 libcxx/include/__algorithm/pstl_merge.h       |  2 +-
 libcxx/include/__algorithm/pstl_move.h        |  2 +-
 libcxx/include/__algorithm/pstl_replace.h     |  2 +-
 libcxx/include/__algorithm/pstl_rotate_copy.h |  2 +-
 libcxx/include/__algorithm/pstl_sort.h        |  2 +-
 libcxx/include/__algorithm/pstl_stable_sort.h |  2 +-
 libcxx/include/__algorithm/pstl_transform.h   |  2 +-
 libcxx/include/__config_site.in               |  6 +--
 .../include/__numeric/pstl_transform_reduce.h |  2 +-
 .../backends}/libdispatch.h                   | 19 ++++++--
 .../cpu_backends => __pstl/backends}/serial.h | 19 ++++++--
 .../thread.h => __pstl/backends/std_thread.h} | 19 ++++++--
 libcxx/include/__pstl/configuration.h         | 27 +++++++++++
 .../configuration_fwd.h}                      | 25 +++++++---
 libcxx/include/libcxx.imp                     |  6 ---
 libcxx/include/module.modulemap               | 47 ++++++++-----------
 libcxx/src/CMakeLists.txt                     |  2 +-
 libcxx/src/pstl/libdispatch.cpp               |  2 +-
 ...pstl.libdispatch.chunk_partitions.pass.cpp |  4 +-
 .../apple/system-install-properties.sh.cpp    |  2 +-
 libcxx/utils/libcxx/test/features.py          |  2 +-
 40 files changed, 154 insertions(+), 169 deletions(-)
 delete mode 100644 libcxx/include/__algorithm/pstl_backends/cpu_backend.h
 delete mode 100644 libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/backends}/libdispatch.h (94%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/backends}/serial.h (78%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends/thread.h => __pstl/backends/std_thread.h} (79%)
 create mode 100644 libcxx/include/__pstl/configuration.h
 rename libcxx/include/{__algorithm/pstl_backend.h => __pstl/configuration_fwd.h} (93%)

diff --git a/libcxx/CMakeLists.txt b/libcxx/CMakeLists.txt
index 043d5a8295c1a..2977c26646cb2 100644
--- a/libcxx/CMakeLists.txt
+++ b/libcxx/CMakeLists.txt
@@ -300,9 +300,9 @@ option(LIBCXX_HAS_EXTERNAL_THREAD_API
    This option may only be set to ON when LIBCXX_ENABLE_THREADS=ON." OFF)
 
 if (LIBCXX_ENABLE_THREADS)
-  set(LIBCXX_PSTL_CPU_BACKEND "std_thread" CACHE STRING "Which PSTL CPU backend to use")
+  set(LIBCXX_PSTL_BACKEND "std_thread" CACHE STRING "Which PSTL backend to use")
 else()
-  set(LIBCXX_PSTL_CPU_BACKEND "serial" CACHE STRING "Which PSTL CPU backend to use")
+  set(LIBCXX_PSTL_BACKEND "serial" CACHE STRING "Which PSTL backend to use")
 endif()
 
 # Misc options ----------------------------------------------------------------
@@ -792,14 +792,14 @@ elseif (LIBCXX_HARDENING_MODE STREQUAL "debug")
   config_define(8 _LIBCPP_HARDENING_MODE_DEFAULT)
 endif()
 
-if (LIBCXX_PSTL_CPU_BACKEND STREQUAL "serial")
-  config_define(1 _LIBCPP_PSTL_CPU_BACKEND_SERIAL)
-elseif(LIBCXX_PSTL_CPU_BACKEND STREQUAL "std_thread")
-  config_define(1 _LIBCPP_PSTL_CPU_BACKEND_THREAD)
-elseif(LIBCXX_PSTL_CPU_BACKEND STREQUAL "libdispatch")
-  config_define(1 _LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH)
+if (LIBCXX_PSTL_BACKEND STREQUAL "serial")
+  config_define(1 _LIBCPP_PSTL_BACKEND_SERIAL)
+elseif(LIBCXX_PSTL_BACKEND STREQUAL "std_thread")
+  config_define(1 _LIBCPP_PSTL_BACKEND_STD_THREAD)
+elseif(LIBCXX_PSTL_BACKEND STREQUAL "libdispatch")
+  config_define(1 _LIBCPP_PSTL_BACKEND_LIBDISPATCH)
 else()
-  message(FATAL_ERROR "LIBCXX_PSTL_CPU_BACKEND is set to ${LIBCXX_PSTL_CPU_BACKEND}, which is not a valid backend.
+  message(FATAL_ERROR "LIBCXX_PSTL_BACKEND is set to ${LIBCXX_PSTL_BACKEND}, which is not a valid backend.
                        Valid backends are: serial, std_thread and libdispatch")
 endif()
 
diff --git a/libcxx/cmake/caches/Apple.cmake b/libcxx/cmake/caches/Apple.cmake
index cec13c08acf10..8768653e620ad 100644
--- a/libcxx/cmake/caches/Apple.cmake
+++ b/libcxx/cmake/caches/Apple.cmake
@@ -7,7 +7,7 @@ set(LIBCXX_ENABLE_STATIC ON CACHE BOOL "")
 set(LIBCXX_ENABLE_SHARED ON CACHE BOOL "")
 set(LIBCXX_CXX_ABI libcxxabi CACHE STRING "")
 set(LIBCXX_ENABLE_VENDOR_AVAILABILITY_ANNOTATIONS ON CACHE BOOL "")
-set(LIBCXX_PSTL_CPU_BACKEND libdispatch CACHE STRING "")
+set(LIBCXX_PSTL_BACKEND libdispatch CACHE STRING "")
 
 set(LIBCXX_HERMETIC_STATIC_LIBRARY ON CACHE BOOL "")
 set(LIBCXXABI_HERMETIC_STATIC_LIBRARY ON CACHE BOOL "")
diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index a2af1d9915be4..ee4979bfc6f89 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -73,18 +73,12 @@ set(files
   __algorithm/pop_heap.h
   __algorithm/prev_permutation.h
   __algorithm/pstl_any_all_none_of.h
-  __algorithm/pstl_backend.h
-  __algorithm/pstl_backends/cpu_backend.h
   __algorithm/pstl_backends/cpu_backends/any_of.h
-  __algorithm/pstl_backends/cpu_backends/backend.h
   __algorithm/pstl_backends/cpu_backends/fill.h
   __algorithm/pstl_backends/cpu_backends/find_if.h
   __algorithm/pstl_backends/cpu_backends/for_each.h
-  __algorithm/pstl_backends/cpu_backends/libdispatch.h
   __algorithm/pstl_backends/cpu_backends/merge.h
-  __algorithm/pstl_backends/cpu_backends/serial.h
   __algorithm/pstl_backends/cpu_backends/stable_sort.h
-  __algorithm/pstl_backends/cpu_backends/thread.h
   __algorithm/pstl_backends/cpu_backends/transform.h
   __algorithm/pstl_backends/cpu_backends/transform_reduce.h
   __algorithm/pstl_copy.h
@@ -594,6 +588,11 @@ set(files
   __numeric/transform_exclusive_scan.h
   __numeric/transform_inclusive_scan.h
   __numeric/transform_reduce.h
+  __pstl/backends/libdispatch.h
+  __pstl/backends/serial.h
+  __pstl/backends/std_thread.h
+  __pstl/configuration.h
+  __pstl/configuration_fwd.h
   __pstl/cpu_algos/cpu_traits.h
   __random/bernoulli_distribution.h
   __random/binomial_distribution.h
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backend.h b/libcxx/include/__algorithm/pstl_backends/cpu_backend.h
deleted file mode 100644
index 53eae58f96095..0000000000000
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backend.h
+++ /dev/null
@@ -1,23 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
-
-#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
-#include <__algorithm/pstl_backends/cpu_backends/fill.h>
-#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
-#include <__algorithm/pstl_backends/cpu_backends/merge.h>
-#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
-#include <__config>
-
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
index 3755d288047e0..3db4765da64b2 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
@@ -11,12 +11,12 @@
 
 #include <__algorithm/any_of.h>
 #include <__algorithm/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__atomic/atomic.h>
 #include <__atomic/memory_order.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/concepts.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
deleted file mode 100644
index cb9425862a2b0..0000000000000
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/backend.h
+++ /dev/null
@@ -1,45 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_BACKEND_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_BACKEND_H
-
-#include <__config>
-#include <cstddef>
-
-#if defined(_LIBCPP_PSTL_CPU_BACKEND_SERIAL)
-#  include <__algorithm/pstl_backends/cpu_backends/serial.h>
-#elif defined(_LIBCPP_PSTL_CPU_BACKEND_THREAD)
-#  include <__algorithm/pstl_backends/cpu_backends/thread.h>
-#elif defined(_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH)
-#  include <__algorithm/pstl_backends/cpu_backends/libdispatch.h>
-#else
-#  error "Invalid CPU backend choice"
-#endif
-
-#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
-#  pragma GCC system_header
-#endif
-
-#if !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
-
-_LIBCPP_BEGIN_NAMESPACE_STD
-
-#  if defined(_LIBCPP_PSTL_CPU_BACKEND_SERIAL)
-using __cpu_backend_tag = __pstl::__serial_backend_tag;
-#  elif defined(_LIBCPP_PSTL_CPU_BACKEND_THREAD)
-using __cpu_backend_tag = __pstl::__std_thread_backend_tag;
-#  elif defined(_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH)
-using __cpu_backend_tag = __pstl::__libdispatch_backend_tag;
-#  endif
-
-_LIBCPP_END_NAMESPACE_STD
-
-#endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && && _LIBCPP_STD_VER >= 17
-
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_BACKEND_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
index 0c20bdff62675..b5a49f8417d32 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
@@ -10,9 +10,9 @@
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FILL_H
 
 #include <__algorithm/fill.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
index 626293faef692..2b1754ea3a755 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
@@ -10,12 +10,12 @@
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FIND_IF_H
 
 #include <__algorithm/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__atomic/atomic.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
index d637084e151d8..6db212ead8ae6 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
@@ -10,9 +10,9 @@
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKNEDS_FOR_EACH_H
 
 #include <__algorithm/for_each.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
index c93f4051c9d09..f3e59e8c02854 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
@@ -10,9 +10,9 @@
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_MERGE_H
 
 #include <__algorithm/merge.h>
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
index 8c60cf897ff86..9ad8cc8fb0f2d 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
@@ -9,9 +9,9 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_STABLE_SORT_H
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_STABLE_SORT_H
 
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__algorithm/stable_sort.h>
 #include <__config>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/empty.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
index 4b9b296866832..65e166d847e12 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
@@ -9,11 +9,11 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_H
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_H
 
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__algorithm/transform.h>
 #include <__config>
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
index c074eea9861c1..af481d505bb91 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
+++ b/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
@@ -9,11 +9,11 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_REDUCE_H
 #define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_REDUCE_H
 
-#include <__algorithm/pstl_backends/cpu_backends/backend.h>
 #include <__config>
 #include <__iterator/concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/transform_reduce.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__type_traits/desugars_to.h>
 #include <__type_traits/is_arithmetic.h>
diff --git a/libcxx/include/__algorithm/pstl_copy.h b/libcxx/include/__algorithm/pstl_copy.h
index f35bb9713ef14..0fcea33c3919f 100644
--- a/libcxx/include/__algorithm/pstl_copy.h
+++ b/libcxx/include/__algorithm/pstl_copy.h
@@ -10,13 +10,13 @@
 #define _LIBCPP___ALGORITHM_PSTL_COPY_H
 
 #include <__algorithm/copy_n.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__algorithm/pstl_transform.h>
 #include <__config>
 #include <__functional/identity.h>
 #include <__iterator/concepts.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_constant_evaluated.h>
 #include <__type_traits/is_execution_policy.h>
diff --git a/libcxx/include/__algorithm/pstl_count.h b/libcxx/include/__algorithm/pstl_count.h
index 6ff57cac334eb..64c84d855e4f6 100644
--- a/libcxx/include/__algorithm/pstl_count.h
+++ b/libcxx/include/__algorithm/pstl_count.h
@@ -11,7 +11,6 @@
 
 #include <__algorithm/count.h>
 #include <__algorithm/for_each.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_for_each.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__atomic/atomic.h>
@@ -20,6 +19,7 @@
 #include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
 #include <__numeric/pstl_transform_reduce.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_find.h b/libcxx/include/__algorithm/pstl_find.h
index 3b30a7bc9b456..b4c4dfb2ffb6f 100644
--- a/libcxx/include/__algorithm/pstl_find.h
+++ b/libcxx/include/__algorithm/pstl_find.h
@@ -11,10 +11,10 @@
 
 #include <__algorithm/comp.h>
 #include <__algorithm/find.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_for_each.h b/libcxx/include/__algorithm/pstl_for_each.h
index a9ebed74a62fd..a99eb6d97fd27 100644
--- a/libcxx/include/__algorithm/pstl_for_each.h
+++ b/libcxx/include/__algorithm/pstl_for_each.h
@@ -11,11 +11,11 @@
 
 #include <__algorithm/for_each.h>
 #include <__algorithm/for_each_n.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__iterator/concepts.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_generate.h b/libcxx/include/__algorithm/pstl_generate.h
index 886af290d7f25..350c0e4798be6 100644
--- a/libcxx/include/__algorithm/pstl_generate.h
+++ b/libcxx/include/__algorithm/pstl_generate.h
@@ -9,12 +9,12 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_GENERATE_H
 #define _LIBCPP___ALGORITHM_PSTL_GENERATE_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_for_each.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_is_partitioned.h b/libcxx/include/__algorithm/pstl_is_partitioned.h
index 108bb1e432526..c016b388e3784 100644
--- a/libcxx/include/__algorithm/pstl_is_partitioned.h
+++ b/libcxx/include/__algorithm/pstl_is_partitioned.h
@@ -10,11 +10,11 @@
 #define _LIBCPP___ALGORITHM_PSTL_IS_PARITTIONED
 
 #include <__algorithm/pstl_any_all_none_of.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_find.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_merge.h b/libcxx/include/__algorithm/pstl_merge.h
index d03cd8c7fbd58..87f634a67f588 100644
--- a/libcxx/include/__algorithm/pstl_merge.h
+++ b/libcxx/include/__algorithm/pstl_merge.h
@@ -9,10 +9,10 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_MERGE_H
 #define _LIBCPP___ALGORITHM_PSTL_MERGE_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_move.h b/libcxx/include/__algorithm/pstl_move.h
index f4c8c1fbb2e87..3155ddedf91bb 100644
--- a/libcxx/include/__algorithm/pstl_move.h
+++ b/libcxx/include/__algorithm/pstl_move.h
@@ -10,13 +10,13 @@
 #define _LIBCPP___ALGORITHM_PSTL_MOVE_H
 
 #include <__algorithm/copy_n.h>
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__algorithm/pstl_transform.h>
 #include <__config>
 #include <__functional/identity.h>
 #include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_constant_evaluated.h>
 #include <__type_traits/is_execution_policy.h>
diff --git a/libcxx/include/__algorithm/pstl_replace.h b/libcxx/include/__algorithm/pstl_replace.h
index 73ac11cda26a9..b2ded54dfe25f 100644
--- a/libcxx/include/__algorithm/pstl_replace.h
+++ b/libcxx/include/__algorithm/pstl_replace.h
@@ -9,13 +9,13 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_REPLACE_H
 #define _LIBCPP___ALGORITHM_PSTL_REPLACE_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_for_each.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__algorithm/pstl_transform.h>
 #include <__config>
 #include <__iterator/cpp17_iterator_concepts.h>
 #include <__iterator/iterator_traits.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/remove_cvref.h>
 #include <__utility/move.h>
diff --git a/libcxx/include/__algorithm/pstl_rotate_copy.h b/libcxx/include/__algorithm/pstl_rotate_copy.h
index adab3958fe311..1a32b710877c1 100644
--- a/libcxx/include/__algorithm/pstl_rotate_copy.h
+++ b/libcxx/include/__algorithm/pstl_rotate_copy.h
@@ -9,10 +9,10 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_ROTATE_COPY_H
 #define _LIBCPP___ALGORITHM_PSTL_ROTATE_COPY_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_copy.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/is_execution_policy.h>
 #include <optional>
 
diff --git a/libcxx/include/__algorithm/pstl_sort.h b/libcxx/include/__algorithm/pstl_sort.h
index 65bc794ca6f4c..769dd81af77e0 100644
--- a/libcxx/include/__algorithm/pstl_sort.h
+++ b/libcxx/include/__algorithm/pstl_sort.h
@@ -9,12 +9,12 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_SORT_H
 #define _LIBCPP___ALGORITHM_PSTL_SORT_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__algorithm/pstl_stable_sort.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
 #include <__utility/empty.h>
diff --git a/libcxx/include/__algorithm/pstl_stable_sort.h b/libcxx/include/__algorithm/pstl_stable_sort.h
index 79b94557e3dc3..f5e0dd40f72b4 100644
--- a/libcxx/include/__algorithm/pstl_stable_sort.h
+++ b/libcxx/include/__algorithm/pstl_stable_sort.h
@@ -9,10 +9,10 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_STABLE_SORT_H
 #define _LIBCPP___ALGORITHM_PSTL_STABLE_SORT_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__algorithm/pstl_transform.h b/libcxx/include/__algorithm/pstl_transform.h
index a01a64a43cf1a..80e1d6b496f2e 100644
--- a/libcxx/include/__algorithm/pstl_transform.h
+++ b/libcxx/include/__algorithm/pstl_transform.h
@@ -9,9 +9,9 @@
 #ifndef _LIBCPP___ALGORITHM_PSTL_TRANSFORM_H
 #define _LIBCPP___ALGORITHM_PSTL_TRANSFORM_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__config>
 #include <__iterator/cpp17_iterator_concepts.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/enable_if.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__type_traits/remove_cvref.h>
diff --git a/libcxx/include/__config_site.in b/libcxx/include/__config_site.in
index 7c002c5bfcf8e..89a14609ee3f9 100644
--- a/libcxx/include/__config_site.in
+++ b/libcxx/include/__config_site.in
@@ -32,9 +32,9 @@
 #cmakedefine _LIBCPP_INSTRUMENTED_WITH_ASAN
 
 // PSTL backends
-#cmakedefine _LIBCPP_PSTL_CPU_BACKEND_SERIAL
-#cmakedefine _LIBCPP_PSTL_CPU_BACKEND_THREAD
-#cmakedefine _LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH
+#cmakedefine _LIBCPP_PSTL_BACKEND_SERIAL
+#cmakedefine _LIBCPP_PSTL_BACKEND_STD_THREAD
+#cmakedefine _LIBCPP_PSTL_BACKEND_LIBDISPATCH
 
 // Hardening.
 #cmakedefine _LIBCPP_HARDENING_MODE_DEFAULT @_LIBCPP_HARDENING_MODE_DEFAULT@
diff --git a/libcxx/include/__numeric/pstl_transform_reduce.h b/libcxx/include/__numeric/pstl_transform_reduce.h
index 2d2621dc8dadb..fe41b1c86f3b1 100644
--- a/libcxx/include/__numeric/pstl_transform_reduce.h
+++ b/libcxx/include/__numeric/pstl_transform_reduce.h
@@ -9,12 +9,12 @@
 #ifndef _LIBCPP___NUMERIC_PSTL_TRANSFORM_REDUCE_H
 #define _LIBCPP___NUMERIC_PSTL_TRANSFORM_REDUCE_H
 
-#include <__algorithm/pstl_backend.h>
 #include <__algorithm/pstl_frontend_dispatch.h>
 #include <__config>
 #include <__functional/operations.h>
 #include <__iterator/cpp17_iterator_concepts.h>
 #include <__numeric/transform_reduce.h>
+#include <__pstl/configuration.h>
 #include <__type_traits/is_execution_policy.h>
 #include <__utility/move.h>
 #include <optional>
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h b/libcxx/include/__pstl/backends/libdispatch.h
similarity index 94%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
rename to libcxx/include/__pstl/backends/libdispatch.h
index 17faadf55dd4f..977b06b9a489c 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/libdispatch.h
+++ b/libcxx/include/__pstl/backends/libdispatch.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_LIBDISPATCH_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_LIBDISPATCH_H
+#ifndef _LIBCPP___PSTL_BACKENDS_LIBDISPATCH_H
+#define _LIBCPP___PSTL_BACKENDS_LIBDISPATCH_H
 
 #include <__algorithm/inplace_merge.h>
 #include <__algorithm/lower_bound.h>
@@ -23,6 +23,7 @@
 #include <__memory/construct_at.h>
 #include <__memory/unique_ptr.h>
 #include <__numeric/reduce.h>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/exception_guard.h>
@@ -40,8 +41,6 @@ _LIBCPP_PUSH_MACROS
 _LIBCPP_BEGIN_NAMESPACE_STD
 namespace __pstl {
 
-struct __libdispatch_backend_tag {};
-
 namespace __libdispatch {
 // ::dispatch_apply is marked as __attribute__((nothrow)) because it doesn't let exceptions propagate, and neither do
 // we.
@@ -349,4 +348,14 @@ _LIBCPP_END_NAMESPACE_STD
 
 _LIBCPP_POP_MACROS
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_LIBDISPATCH_H
+// Implement PSTL algorithms based on the __cpu_traits specialized above
+#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
+#include <__algorithm/pstl_backends/cpu_backends/fill.h>
+#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
+#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
+#include <__algorithm/pstl_backends/cpu_backends/merge.h>
+#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+
+#endif // _LIBCPP___PSTL_BACKENDS_LIBDISPATCH_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h b/libcxx/include/__pstl/backends/serial.h
similarity index 78%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
rename to libcxx/include/__pstl/backends/serial.h
index 7544619a8eefd..8bb8945093096 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/serial.h
+++ b/libcxx/include/__pstl/backends/serial.h
@@ -7,10 +7,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_SERIAL_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_SERIAL_H
+#ifndef _LIBCPP___PSTL_BACKENDS_SERIAL_H
+#define _LIBCPP___PSTL_BACKENDS_SERIAL_H
 
 #include <__config>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/move.h>
@@ -29,8 +30,6 @@ _LIBCPP_PUSH_MACROS
 _LIBCPP_BEGIN_NAMESPACE_STD
 namespace __pstl {
 
-struct __serial_backend_tag {};
-
 template <>
 struct __cpu_traits<__serial_backend_tag> {
   template <class _RandomAccessIterator, class _Fp>
@@ -82,4 +81,14 @@ _LIBCPP_POP_MACROS
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_SERIAL_H
+// Implement PSTL algorithms based on the __cpu_traits specialized above
+#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
+#include <__algorithm/pstl_backends/cpu_backends/fill.h>
+#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
+#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
+#include <__algorithm/pstl_backends/cpu_backends/merge.h>
+#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+
+#endif // _LIBCPP___PSTL_BACKENDS_SERIAL_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h b/libcxx/include/__pstl/backends/std_thread.h
similarity index 79%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
rename to libcxx/include/__pstl/backends/std_thread.h
index 2acf912264a00..ab09f42cfdd8d 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/thread.h
+++ b/libcxx/include/__pstl/backends/std_thread.h
@@ -6,11 +6,12 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_THREAD_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_THREAD_H
+#ifndef _LIBCPP___PSTL_BACKENDS_STD_THREAD_H
+#define _LIBCPP___PSTL_BACKENDS_STD_THREAD_H
 
 #include <__assert>
 #include <__config>
+#include <__pstl/configuration_fwd.h>
 #include <__pstl/cpu_algos/cpu_traits.h>
 #include <__utility/empty.h>
 #include <__utility/move.h>
@@ -32,8 +33,6 @@ _LIBCPP_PUSH_MACROS
 _LIBCPP_BEGIN_NAMESPACE_STD
 namespace __pstl {
 
-struct __std_thread_backend_tag {};
-
 template <>
 struct __cpu_traits<__std_thread_backend_tag> {
   template <class _RandomAccessIterator, class _Fp>
@@ -85,4 +84,14 @@ _LIBCPP_END_NAMESPACE_STD
 
 _LIBCPP_POP_MACROS
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_THREAD_H
+// Implement PSTL algorithms based on the __cpu_traits specialized above
+#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
+#include <__algorithm/pstl_backends/cpu_backends/fill.h>
+#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
+#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
+#include <__algorithm/pstl_backends/cpu_backends/merge.h>
+#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform.h>
+#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+
+#endif // _LIBCPP___PSTL_BACKENDS_STD_THREAD_H
diff --git a/libcxx/include/__pstl/configuration.h b/libcxx/include/__pstl/configuration.h
new file mode 100644
index 0000000000000..d32bd21df1f9e
--- /dev/null
+++ b/libcxx/include/__pstl/configuration.h
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___PSTL_CONFIGURATION_H
+#define _LIBCPP___PSTL_CONFIGURATION_H
+
+#include <__config>
+#include <__pstl/configuration_fwd.h>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+#if defined(_LIBCPP_PSTL_BACKEND_SERIAL)
+#  include <__pstl/backends/serial.h>
+#elif defined(_LIBCPP_PSTL_BACKEND_STD_THREAD)
+#  include <__pstl/backends/std_thread.h>
+#elif defined(_LIBCPP_PSTL_BACKEND_LIBDISPATCH)
+#  include <__pstl/backends/libdispatch.h>
+#endif
+
+#endif // _LIBCPP___PSTL_CONFIGURATION_H
diff --git a/libcxx/include/__algorithm/pstl_backend.h b/libcxx/include/__pstl/configuration_fwd.h
similarity index 93%
rename from libcxx/include/__algorithm/pstl_backend.h
rename to libcxx/include/__pstl/configuration_fwd.h
index 3af03ce2fbc8e..995fcfce847cb 100644
--- a/libcxx/include/__algorithm/pstl_backend.h
+++ b/libcxx/include/__pstl/configuration_fwd.h
@@ -6,10 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKEND_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKEND_H
+#ifndef _LIBCPP___PSTL_CONFIGURATION_FWD_H
+#define _LIBCPP___PSTL_CONFIGURATION_FWD_H
 
-#include <__algorithm/pstl_backends/cpu_backend.h>
 #include <__config>
 #include <execution>
 
@@ -191,6 +190,20 @@ into a program termination at the front-end level. When a backend returns a dise
 frontend will turn that into a call to `std::__throw_bad_alloc();` to report the internal failure to the user.
 */
 
+namespace __pstl {
+struct __libdispatch_backend_tag {};
+struct __serial_backend_tag {};
+struct __std_thread_backend_tag {};
+} // namespace __pstl
+
+#  if defined(_LIBCPP_PSTL_BACKEND_SERIAL)
+using __cpu_backend_tag = __pstl::__serial_backend_tag;
+#  elif defined(_LIBCPP_PSTL_BACKEND_STD_THREAD)
+using __cpu_backend_tag = __pstl::__std_thread_backend_tag;
+#  elif defined(_LIBCPP_PSTL_BACKEND_LIBDISPATCH)
+using __cpu_backend_tag = __pstl::__libdispatch_backend_tag;
+#  endif
+
 template <class _ExecutionPolicy>
 struct __select_backend;
 
@@ -206,8 +219,8 @@ struct __select_backend<std::execution::unsequenced_policy> {
 };
 #  endif
 
-#  if defined(_LIBCPP_PSTL_CPU_BACKEND_SERIAL) || defined(_LIBCPP_PSTL_CPU_BACKEND_THREAD) ||                          \
-      defined(_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH)
+#  if defined(_LIBCPP_PSTL_BACKEND_SERIAL) || defined(_LIBCPP_PSTL_BACKEND_STD_THREAD) ||                              \
+      defined(_LIBCPP_PSTL_BACKEND_LIBDISPATCH)
 template <>
 struct __select_backend<std::execution::parallel_policy> {
   using type = __cpu_backend_tag;
@@ -229,4 +242,4 @@ _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKEND_H
+#endif // _LIBCPP___PSTL_CONFIGURATION_FWD_H
diff --git a/libcxx/include/libcxx.imp b/libcxx/include/libcxx.imp
index 8820fb8c0936f..a4e2690fc55c9 100644
--- a/libcxx/include/libcxx.imp
+++ b/libcxx/include/libcxx.imp
@@ -73,18 +73,12 @@
   { include: [ "<__algorithm/pop_heap.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/prev_permutation.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_any_all_none_of.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backend.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backend.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/any_of.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/backend.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/fill.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/find_if.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/for_each.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/libdispatch.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/merge.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/serial.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/stable_sort.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/thread.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/transform.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_backends/cpu_backends/transform_reduce.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_copy.h>", "private", "<algorithm>", "public" ] },
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index ce133e471deb7..f996c2cc05459 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -714,32 +714,14 @@ module std_private_algorithm_partition_point                             [system
 module std_private_algorithm_pop_heap                                    [system] { header "__algorithm/pop_heap.h" }
 module std_private_algorithm_prev_permutation                            [system] { header "__algorithm/prev_permutation.h" }
 module std_private_algorithm_pstl_any_all_none_of                        [system] { header "__algorithm/pstl_any_all_none_of.h" }
-module std_private_algorithm_pstl_backend                                [system] {
-  header "__algorithm/pstl_backend.h"
-  export *
-}
-module std_private_algorithm_pstl_backends_cpu_backend                   [system] {
-  header "__algorithm/pstl_backends/cpu_backend.h"
-  export *
-}
-module std_private_algorithm_pstl_backends_cpu_backends_any_of           [system] { header "__algorithm/pstl_backends/cpu_backends/any_of.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_backend          [system] {
-  header "__algorithm/pstl_backends/cpu_backends/backend.h"
-  export *
-}
-module std_private_algorithm_pstl_backends_cpu_backends_fill             [system] { header "__algorithm/pstl_backends/cpu_backends/fill.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_find_if          [system] { header "__algorithm/pstl_backends/cpu_backends/find_if.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_for_each         [system] { header "__algorithm/pstl_backends/cpu_backends/for_each.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_libdispatch      [system] { header "__algorithm/pstl_backends/cpu_backends/libdispatch.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_merge            [system] { header "__algorithm/pstl_backends/cpu_backends/merge.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_serial           [system] { textual header "__algorithm/pstl_backends/cpu_backends/serial.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_stable_sort      [system] { header "__algorithm/pstl_backends/cpu_backends/stable_sort.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_thread           [system] { textual header "__algorithm/pstl_backends/cpu_backends/thread.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_transform        [system] {
-  header "__algorithm/pstl_backends/cpu_backends/transform.h"
-  export std_private_algorithm_transform
-}
-module std_private_algorithm_pstl_backends_cpu_backends_transform_reduce [system] { header "__algorithm/pstl_backends/cpu_backends/transform_reduce.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_any_of           [system] { textual header "__algorithm/pstl_backends/cpu_backends/any_of.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_fill             [system] { textual header "__algorithm/pstl_backends/cpu_backends/fill.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_find_if          [system] { textual header "__algorithm/pstl_backends/cpu_backends/find_if.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_for_each         [system] { textual header "__algorithm/pstl_backends/cpu_backends/for_each.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_merge            [system] { textual header "__algorithm/pstl_backends/cpu_backends/merge.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_stable_sort      [system] { textual header "__algorithm/pstl_backends/cpu_backends/stable_sort.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_transform        [system] { textual header "__algorithm/pstl_backends/cpu_backends/transform.h" }
+module std_private_algorithm_pstl_backends_cpu_backends_transform_reduce [system] { textual header "__algorithm/pstl_backends/cpu_backends/transform_reduce.h" }
 module std_private_algorithm_pstl_copy                                   [system] { header "__algorithm/pstl_copy.h" }
 module std_private_algorithm_pstl_count                                  [system] { header "__algorithm/pstl_count.h" }
 module std_private_algorithm_pstl_equal                                  [system] { header "__algorithm/pstl_equal.h" }
@@ -1613,7 +1595,18 @@ module std_private_numeric_transform_exclusive_scan [system] { header "__numeric
 module std_private_numeric_transform_inclusive_scan [system] { header "__numeric/transform_inclusive_scan.h" }
 module std_private_numeric_transform_reduce         [system] { header "__numeric/transform_reduce.h" }
 
-module std_private_pstl_cpu_algos_cpu_traits [system] { header "__pstl/cpu_algos/cpu_traits.h" }
+module std_private_pstl_backends_libdispatch      [system] { header "__pstl/backends/libdispatch.h" }
+module std_private_pstl_backends_serial           [system] { header "__pstl/backends/serial.h" }
+module std_private_pstl_backends_std_thread       [system] { header "__pstl/backends/std_thread.h" }
+module std_private_pstl_cpu_algos_cpu_traits      [system] { header "__pstl/cpu_algos/cpu_traits.h" }
+module std_private_pstl_configuration_fwd         [system] {
+  header "__pstl/configuration_fwd.h"
+  export *
+}
+module std_private_pstl_configuration             [system] {
+  header "__pstl/configuration.h"
+  export *
+}
 
 module std_private_queue_fwd [system] { header "__fwd/queue.h" }
 
diff --git a/libcxx/src/CMakeLists.txt b/libcxx/src/CMakeLists.txt
index a4a3fee864571..8b28d1b891895 100644
--- a/libcxx/src/CMakeLists.txt
+++ b/libcxx/src/CMakeLists.txt
@@ -327,7 +327,7 @@ set(LIBCXX_EXPERIMENTAL_SOURCES
   experimental/keep.cpp
   )
 
-if (LIBCXX_PSTL_CPU_BACKEND STREQUAL "libdispatch")
+if (LIBCXX_PSTL_BACKEND STREQUAL "libdispatch")
   list(APPEND LIBCXX_EXPERIMENTAL_SOURCES
     pstl/libdispatch.cpp
     )
diff --git a/libcxx/src/pstl/libdispatch.cpp b/libcxx/src/pstl/libdispatch.cpp
index d997a9c73463d..3dca702341c85 100644
--- a/libcxx/src/pstl/libdispatch.cpp
+++ b/libcxx/src/pstl/libdispatch.cpp
@@ -7,8 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 #include <__algorithm/min.h>
-#include <__algorithm/pstl_backends/cpu_backends/libdispatch.h>
 #include <__config>
+#include <__pstl/backends/libdispatch.h>
 #include <dispatch/dispatch.h>
 
 _LIBCPP_BEGIN_NAMESPACE_STD
diff --git a/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp b/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
index 8c7016a80b811..b48ac02dd79c5 100644
--- a/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
+++ b/libcxx/test/libcxx/algorithms/pstl.libdispatch.chunk_partitions.pass.cpp
@@ -8,11 +8,11 @@
 
 // <algorithm>
 
-// REQUIRES: libcpp-pstl-cpu-backend-libdispatch
+// REQUIRES: libcpp-pstl-backend-libdispatch
 
 // __chunk_partitions __partition_chunks(ptrdiff_t);
 
-#include <__algorithm/pstl_backends/cpu_backends/libdispatch.h>
+#include <__pstl/backends/libdispatch.h>
 #include <cassert>
 #include <cstddef>
 
diff --git a/libcxx/test/libcxx/vendor/apple/system-install-properties.sh.cpp b/libcxx/test/libcxx/vendor/apple/system-install-properties.sh.cpp
index 3e2e080368f4c..4ea27401e35d4 100644
--- a/libcxx/test/libcxx/vendor/apple/system-install-properties.sh.cpp
+++ b/libcxx/test/libcxx/vendor/apple/system-install-properties.sh.cpp
@@ -45,4 +45,4 @@
 
 // Make sure we use the libdispatch backend for the PSTL.
 //
-// RUN: grep "%{include-dir}/__config_site" -e '#define _LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH'
+// RUN: grep "%{include-dir}/__config_site" -e '#define _LIBCPP_PSTL_BACKEND_LIBDISPATCH'
diff --git a/libcxx/utils/libcxx/test/features.py b/libcxx/utils/libcxx/test/features.py
index 6ff16309546ba..c81b56b1af547 100644
--- a/libcxx/utils/libcxx/test/features.py
+++ b/libcxx/utils/libcxx/test/features.py
@@ -318,7 +318,7 @@ def _getAndroidDeviceApi(cfg):
     "_LIBCPP_HAS_NO_WIDE_CHARACTERS": "no-wide-characters",
     "_LIBCPP_HAS_NO_TIME_ZONE_DATABASE": "no-tzdb",
     "_LIBCPP_HAS_NO_UNICODE": "libcpp-has-no-unicode",
-    "_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH": "libcpp-pstl-cpu-backend-libdispatch",
+    "_LIBCPP_PSTL_BACKEND_LIBDISPATCH": "libcpp-pstl-backend-libdispatch",
 }
 for macro, feature in macros.items():
     DEFAULT_FEATURES.append(

From 8d49ce176414cd4d0d5d276fd721d9226e17e810 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Wed, 17 Apr 2024 18:38:24 +0100
Subject: [PATCH 774/886] [GlobalISel][AArch64] Add LLRINT support (#88702)

This hooks up G_INTRINSIC_LLRINT instructions, very similar to the lrint
nodes that already exist. On AArch64 they are treated the same as lrint
with the default return types.
---
 llvm/include/llvm/Support/TargetOpcodes.def   |  3 +
 llvm/include/llvm/Target/GenericOpcodes.td    |  6 ++
 .../Target/GlobalISel/SelectionDAGCompat.td   |  1 +
 llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp  |  2 +
 .../CodeGen/GlobalISel/LegalizerHelper.cpp    |  6 +-
 .../AArch64/GISel/AArch64LegalizerInfo.cpp    |  2 +-
 .../AArch64/GISel/AArch64RegisterBankInfo.cpp |  1 +
 .../AArch64/GlobalISel/arm64-irtranslator.ll  | 10 ++
 .../AArch64/GlobalISel/legalize-llrint.mir    | 98 +++++++++++++++++++
 .../GlobalISel/legalizer-info-validation.mir  |  4 +
 llvm/test/CodeGen/AArch64/llrint-conv-fp16.ll |  2 +
 llvm/test/CodeGen/AArch64/llrint-conv.ll      | 58 +++++++----
 .../builtins/match-table-replacerreg.td       | 22 ++---
 .../match-table-imms.td                       | 30 +++---
 .../match-table-intrinsics.td                 | 16 +--
 .../match-table-patfrag-root.td               | 30 +++---
 .../GlobalISelCombinerEmitter/match-table.td  | 62 ++++++------
 llvm/test/TableGen/GlobalISelEmitter.td       |  2 +-
 18 files changed, 251 insertions(+), 104 deletions(-)
 create mode 100644 llvm/test/CodeGen/AArch64/GlobalISel/legalize-llrint.mir

diff --git a/llvm/include/llvm/Support/TargetOpcodes.def b/llvm/include/llvm/Support/TargetOpcodes.def
index 5765926d6d93d..cb98f96af522f 100644
--- a/llvm/include/llvm/Support/TargetOpcodes.def
+++ b/llvm/include/llvm/Support/TargetOpcodes.def
@@ -351,6 +351,9 @@ HANDLE_TARGET_OPCODE(G_INTRINSIC_ROUND)
 /// INTRINSIC round to integer intrinsic.
 HANDLE_TARGET_OPCODE(G_INTRINSIC_LRINT)
 
+/// INTRINSIC long round to integer intrinsic.
+HANDLE_TARGET_OPCODE(G_INTRINSIC_LLRINT)
+
 /// INTRINSIC roundeven intrinsic.
 HANDLE_TARGET_OPCODE(G_INTRINSIC_ROUNDEVEN)
 
diff --git a/llvm/include/llvm/Target/GenericOpcodes.td b/llvm/include/llvm/Target/GenericOpcodes.td
index d0f471eb29b6f..e8cf8fcb647f4 100644
--- a/llvm/include/llvm/Target/GenericOpcodes.td
+++ b/llvm/include/llvm/Target/GenericOpcodes.td
@@ -1089,6 +1089,12 @@ def G_INTRINSIC_LRINT : GenericInstruction {
   let hasSideEffects = false;
 }
 
+def G_INTRINSIC_LLRINT : GenericInstruction {
+  let OutOperandList = (outs type0:$dst);
+  let InOperandList = (ins type1:$src);
+  let hasSideEffects = false;
+}
+
 def G_INTRINSIC_ROUNDEVEN : GenericInstruction {
   let OutOperandList = (outs type0:$dst);
   let InOperandList = (ins type0:$src1);
diff --git a/llvm/include/llvm/Target/GlobalISel/SelectionDAGCompat.td b/llvm/include/llvm/Target/GlobalISel/SelectionDAGCompat.td
index dd4e7d790bc6b..8fa0e4b86d6dc 100644
--- a/llvm/include/llvm/Target/GlobalISel/SelectionDAGCompat.td
+++ b/llvm/include/llvm/Target/GlobalISel/SelectionDAGCompat.td
@@ -157,6 +157,7 @@ def : GINodeEquiv<G_INTRINSIC_TRUNC, ftrunc>;
 def : GINodeEquiv<G_INTRINSIC_ROUND, fround>;
 def : GINodeEquiv<G_INTRINSIC_ROUNDEVEN, froundeven>;
 def : GINodeEquiv<G_INTRINSIC_LRINT, lrint>;
+def : GINodeEquiv<G_INTRINSIC_LLRINT, llrint>;
 def : GINodeEquiv<G_FCOPYSIGN, fcopysign>;
 def : GINodeEquiv<G_SMIN, smin>;
 def : GINodeEquiv<G_SMAX, smax>;
diff --git a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
index 9b4575f7f34d4..0b6aae3759756 100644
--- a/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/IRTranslator.cpp
@@ -1955,6 +1955,8 @@ unsigned IRTranslator::getSimpleIntrinsicOpcode(Intrinsic::ID ID) {
       return TargetOpcode::G_PTRMASK;
     case Intrinsic::lrint:
       return TargetOpcode::G_INTRINSIC_LRINT;
+    case Intrinsic::llrint:
+      return TargetOpcode::G_INTRINSIC_LLRINT;
     // FADD/FMUL require checking the FMF, so are handled elsewhere.
     case Intrinsic::vector_reduce_fmin:
       return TargetOpcode::G_VECREDUCE_FMIN;
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 156353296cfc1..d55091e2e7173 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -474,6 +474,8 @@ static RTLIB::Libcall getRTLibDesc(unsigned Opcode, unsigned Size) {
     RTLIBCASE(ROUNDEVEN_F);
   case TargetOpcode::G_INTRINSIC_LRINT:
     RTLIBCASE(LRINT_F);
+  case TargetOpcode::G_INTRINSIC_LLRINT:
+    RTLIBCASE(LLRINT_F);
   }
   llvm_unreachable("Unknown libcall function");
 }
@@ -1061,7 +1063,8 @@ LegalizerHelper::libcall(MachineInstr &MI, LostDebugLocObserver &LocObserver) {
       return Status;
     break;
   }
-  case TargetOpcode::G_INTRINSIC_LRINT: {
+  case TargetOpcode::G_INTRINSIC_LRINT:
+  case TargetOpcode::G_INTRINSIC_LLRINT: {
     LLT LLTy = MRI.getType(MI.getOperand(1).getReg());
     unsigned Size = LLTy.getSizeInBits();
     Type *HLTy = getFloatTypeForLLT(Ctx, LLTy);
@@ -2661,6 +2664,7 @@ LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
   case TargetOpcode::G_FPTOSI:
   case TargetOpcode::G_FPTOUI:
   case TargetOpcode::G_INTRINSIC_LRINT:
+  case TargetOpcode::G_INTRINSIC_LLRINT:
   case TargetOpcode::G_IS_FPCLASS:
     Observer.changingInstr(MI);
 
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
index 661ea151d1a0c..85dd0f2eb192d 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
@@ -262,7 +262,7 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
       .minScalar(0, s32)
       .scalarize(0);
 
-  getActionDefinitionsBuilder(G_INTRINSIC_LRINT)
+  getActionDefinitionsBuilder({G_INTRINSIC_LRINT, G_INTRINSIC_LLRINT})
       .legalFor({{s64, MinFPScalar}, {s64, s32}, {s64, s64}})
       .libcallFor({{s64, s128}})
       .minScalarOrElt(1, MinFPScalar);
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
index d5c4ce1888e78..44ba9f0429e67 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64RegisterBankInfo.cpp
@@ -793,6 +793,7 @@ AArch64RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const {
   case TargetOpcode::G_FPTOSI:
   case TargetOpcode::G_FPTOUI:
   case TargetOpcode::G_INTRINSIC_LRINT:
+  case TargetOpcode::G_INTRINSIC_LLRINT:
     if (MRI.getType(MI.getOperand(0).getReg()).isVector())
       break;
     OpRegBankIdx = {PMI_FirstGPR, PMI_FirstFPR};
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll b/llvm/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
index a131f35e66d03..a61931b898aea 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/arm64-irtranslator.ll
@@ -1385,6 +1385,16 @@ define i32 @test_intrinsic_lrint(float %a) {
   ret i32 %res
 }
 
+declare i32 @llvm.llrint.i32.f32(float)
+define i32 @test_intrinsic_llrint(float %a) {
+; CHECK-LABEL: name: test_intrinsic_llrint
+; CHECK: [[A:%[0-9]+]]:_(s32) = COPY $s0
+; CHECK: [[RES:%[0-9]+]]:_(s32) = G_INTRINSIC_LLRINT [[A]]
+; CHECK: $w0 = COPY [[RES]]
+  %res = call i32 @llvm.llrint.i32.f32(float %a)
+  ret i32 %res
+}
+
 declare i32 @llvm.ctlz.i32(i32, i1)
 define i32 @test_ctlz_intrinsic_zero_not_undef(i32 %a) {
 ; CHECK-LABEL: name: test_ctlz_intrinsic_zero_not_undef
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/legalize-llrint.mir b/llvm/test/CodeGen/AArch64/GlobalISel/legalize-llrint.mir
new file mode 100644
index 0000000000000..f77649b793951
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/legalize-llrint.mir
@@ -0,0 +1,98 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -verify-machineinstrs -mtriple aarch64-unknown-unknown -run-pass=legalizer %s -o - | FileCheck %s
+---
+name:            testmsws
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$s0' }
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: testmsws
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: [[INTRINSIC_LLRINT:%[0-9]+]]:_(s64) = G_INTRINSIC_LLRINT [[COPY]](s32)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[INTRINSIC_LLRINT]](s64)
+    ; CHECK-NEXT: $w0 = COPY [[TRUNC]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %0:_(s32) = COPY $s0
+    %1:_(s64) = G_INTRINSIC_LLRINT %0(s32)
+    %2:_(s32) = G_TRUNC %1(s64)
+    $w0 = COPY %2(s32)
+    RET_ReallyLR implicit $w0
+
+...
+---
+name:            testmsxs
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$s0' }
+body:             |
+  bb.1:
+    liveins: $s0
+
+    ; CHECK-LABEL: name: testmsxs
+    ; CHECK: liveins: $s0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $s0
+    ; CHECK-NEXT: [[INTRINSIC_LLRINT:%[0-9]+]]:_(s64) = G_INTRINSIC_LLRINT [[COPY]](s32)
+    ; CHECK-NEXT: $x0 = COPY [[INTRINSIC_LLRINT]](s64)
+    ; CHECK-NEXT: RET_ReallyLR implicit $x0
+    %0:_(s32) = COPY $s0
+    %1:_(s64) = G_INTRINSIC_LLRINT %0(s32)
+    $x0 = COPY %1(s64)
+    RET_ReallyLR implicit $x0
+
+...
+---
+name:            testmswd
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$d0' }
+body:             |
+  bb.1:
+    liveins: $d0
+
+    ; CHECK-LABEL: name: testmswd
+    ; CHECK: liveins: $d0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $d0
+    ; CHECK-NEXT: [[INTRINSIC_LLRINT:%[0-9]+]]:_(s64) = G_INTRINSIC_LLRINT [[COPY]](s64)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[INTRINSIC_LLRINT]](s64)
+    ; CHECK-NEXT: $w0 = COPY [[TRUNC]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %0:_(s64) = COPY $d0
+    %1:_(s64) = G_INTRINSIC_LLRINT %0(s64)
+    %2:_(s32) = G_TRUNC %1(s64)
+    $w0 = COPY %2(s32)
+    RET_ReallyLR implicit $w0
+
+...
+---
+name:            testmsxd
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$d0' }
+body:             |
+  bb.1:
+    liveins: $d0
+
+    ; CHECK-LABEL: name: testmsxd
+    ; CHECK: liveins: $d0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $d0
+    ; CHECK-NEXT: [[INTRINSIC_LLRINT:%[0-9]+]]:_(s64) = G_INTRINSIC_LLRINT [[COPY]](s64)
+    ; CHECK-NEXT: $x0 = COPY [[INTRINSIC_LLRINT]](s64)
+    ; CHECK-NEXT: RET_ReallyLR implicit $x0
+    %0:_(s64) = COPY $d0
+    %1:_(s64) = G_INTRINSIC_LLRINT %0(s64)
+    $x0 = COPY %1(s64)
+    RET_ReallyLR implicit $x0
+
+...
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir b/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
index 0793f3983c8e5..098726b0a980d 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir
@@ -156,6 +156,10 @@
 # DEBUG-NEXT: G_INTRINSIC_LRINT (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
 # DEBUG-NEXT: .. the first uncovered type index: 2, OK
 # DEBUG-NEXT: .. the first uncovered imm index: 0, OK
+# DEBUG-NEXT: G_INTRINSIC_LLRINT (opcode {{[0-9]+}}): 2 type indices, 0 imm indices
+# DEBUG-NEXT: .. opcode {{[0-9]+}} is aliased to {{[0-9]+}}
+# DEBUG-NEXT: .. the first uncovered type index: 2, OK
+# DEBUG-NEXT: .. the first uncovered imm index: 0, OK
 # DEBUG-NEXT: G_INTRINSIC_ROUNDEVEN (opcode {{[0-9]+}}): 1 type index, 0 imm indices
 # DEBUG-NEXT: .. opcode {{[0-9]+}} is aliased to {{[0-9]+}}
 # DEBUG-NEXT: .. type index coverage check SKIPPED: user-defined predicate detected
diff --git a/llvm/test/CodeGen/AArch64/llrint-conv-fp16.ll b/llvm/test/CodeGen/AArch64/llrint-conv-fp16.ll
index 1adbbab76abf5..7e28c863b07a9 100644
--- a/llvm/test/CodeGen/AArch64/llrint-conv-fp16.ll
+++ b/llvm/test/CodeGen/AArch64/llrint-conv-fp16.ll
@@ -1,6 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
 ; RUN: llc < %s -mtriple=aarch64 | FileCheck %s --check-prefixes=CHECK-NOFP16
 ; RUN: llc < %s -mtriple=aarch64 -mattr=+fullfp16 | FileCheck %s --check-prefixes=CHECK-FP16
+; RUN: llc < %s -mtriple=aarch64 -global-isel | FileCheck %s --check-prefixes=CHECK-NOFP16
+; RUN: llc < %s -mtriple=aarch64 -mattr=+fullfp16 -global-isel | FileCheck %s --check-prefixes=CHECK-FP16
 
 define i16 @testmhhs(half %x) {
 ; CHECK-NOFP16-LABEL: testmhhs:
diff --git a/llvm/test/CodeGen/AArch64/llrint-conv.ll b/llvm/test/CodeGen/AArch64/llrint-conv.ll
index fa11b007eeb3d..3a6396d120f79 100644
--- a/llvm/test/CodeGen/AArch64/llrint-conv.ll
+++ b/llvm/test/CodeGen/AArch64/llrint-conv.ll
@@ -1,59 +1,75 @@
-; RUN: llc < %s -mtriple=aarch64 -mattr=+neon | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc < %s -mtriple=aarch64 | FileCheck %s
+; RUN: llc < %s -mtriple=aarch64 -global-isel | FileCheck %s
 
-; CHECK-LABEL: testmsws:
-; CHECK:       frintx  [[REG:s[0-9]]], s0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK:       ret
 define i32 @testmsws(float %x) {
+; CHECK-LABEL: testmsws:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx s0, s0
+; CHECK-NEXT:    fcvtzs x0, s0
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.llrint.f32(float %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsxs:
-; CHECK:       frintx  [[REG:s[0-9]]], s0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK-NEXT:  ret
 define i64 @testmsxs(float %x) {
+; CHECK-LABEL: testmsxs:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx s0, s0
+; CHECK-NEXT:    fcvtzs x0, s0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.llrint.f32(float %x)
   ret i64 %0
 }
 
-; CHECK-LABEL: testmswd:
-; CHECK:       frintx  [[REG:d[0-9]]], d0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK:       ret
 define i32 @testmswd(double %x) {
+; CHECK-LABEL: testmswd:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx d0, d0
+; CHECK-NEXT:    fcvtzs x0, d0
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.llrint.f64(double %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsxd:
-; CHECK:       frintx  [[REG:d[0-9]]], d0
-; CHECK-NEXT:  fcvtzs  x0, [[REG]]
-; CHECK-nEXT:  ret
 define i64 @testmsxd(double %x) {
+; CHECK-LABEL: testmsxd:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    frintx d0, d0
+; CHECK-NEXT:    fcvtzs x0, d0
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.llrint.f64(double %x)
   ret i64 %0
 }
 
-; CHECK-LABEL: testmswl:
-; CHECK:       bl      llrintl
 define i32 @testmswl(fp128 %x) {
+; CHECK-LABEL: testmswl:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -16
+; CHECK-NEXT:    bl llrintl
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
+; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
 entry:
   %0 = tail call i64 @llvm.llrint.f128(fp128 %x)
   %conv = trunc i64 %0 to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: testmsll:
-; CHECK:       b       llrintl
 define i64 @testmsll(fp128 %x) {
+; CHECK-LABEL: testmsll:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    b llrintl
 entry:
   %0 = tail call i64 @llvm.llrint.f128(fp128 %x)
   ret i64 %0
diff --git a/llvm/test/TableGen/GlobalISelCombinerEmitter/builtins/match-table-replacerreg.td b/llvm/test/TableGen/GlobalISelCombinerEmitter/builtins/match-table-replacerreg.td
index 40a831d7e9e8f..ebb95ccb21040 100644
--- a/llvm/test/TableGen/GlobalISelCombinerEmitter/builtins/match-table-replacerreg.td
+++ b/llvm/test/TableGen/GlobalISelCombinerEmitter/builtins/match-table-replacerreg.td
@@ -28,11 +28,11 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 
 //      CHECK: const uint8_t *GenMyCombiner::getMatchTable() const {
 // CHECK-NEXT:   constexpr static uint8_t MatchTable0[] = {
-// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(69), GIMT_Encode2(186), /*)*//*default:*//*Label 2*/ GIMT_Encode4(562),
-// CHECK-NEXT:     /*TargetOpcode::G_UNMERGE_VALUES*//*Label 0*/ GIMT_Encode4(478), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_FNEG*//*Label 1*/ GIMT_Encode4(530),
-// CHECK-NEXT:     // Label 0: @478
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 3*/ GIMT_Encode4(529), // Rule ID 1 //
+// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2({{[0-9]+}}), GIMT_Encode2({{[0-9]+}}), /*)*//*default:*//*Label 2*/ GIMT_Encode4([[L562:[0-9]+]]),
+// CHECK-NEXT:     /*TargetOpcode::G_UNMERGE_VALUES*//*Label 0*/ GIMT_Encode4([[L478:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_FNEG*//*Label 1*/ GIMT_Encode4([[L530:[0-9]+]]),
+// CHECK-NEXT:     // Label 0: @[[L478]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 3*/ GIMT_Encode4([[L529:[0-9]+]]), // Rule ID 1 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule1Enabled),
 // CHECK-NEXT:       GIM_CheckNumOperands, /*MI*/0, /*Expected*/3,
 // CHECK-NEXT:       // MIs[0] a
@@ -57,10 +57,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_ReplaceRegWithTempReg, /*OldInsnID*/0, /*OldOpIdx*/1, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 3: @529
+// CHECK-NEXT:     // Label 3: @[[L529]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 1: @530
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4(561), // Rule ID 0 //
+// CHECK-NEXT:     // Label 1: @[[L530]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4([[L561:[0-9]+]]), // Rule ID 0 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       // MIs[0] dst
 // CHECK-NEXT:       // No operand predicates
@@ -75,10 +75,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_ReplaceReg, /*OldInsnID*/0, /*OldOpIdx*/0, /*NewInsnId*/1, /*NewOpIdx*/1,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 4: @561
+// CHECK-NEXT:     // Label 4: @[[L561]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 2: @562
+// CHECK-NEXT:     // Label 2: @[[L562]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     }; // Size: 563 bytes
+// CHECK-NEXT:     }; // Size: {{[0-9]+}} bytes
 // CHECK-NEXT:   return MatchTable0;
 // CHECK-NEXT: }
diff --git a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-imms.td b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-imms.td
index 751b1318ecc01..6004a17d351be 100644
--- a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-imms.td
+++ b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-imms.td
@@ -34,12 +34,12 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 
 //      CHECK: const uint8_t *GenMyCombiner::getMatchTable() const {
 // CHECK-NEXT:   constexpr static uint8_t MatchTable0[] = {
-// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(19), GIMT_Encode2(132), /*)*//*default:*//*Label 3*/ GIMT_Encode4(579),
-// CHECK-NEXT:     /*TargetOpcode::COPY*//*Label 0*/ GIMT_Encode4(462), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_CONSTANT*//*Label 1*/ GIMT_Encode4(493), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 2*/ GIMT_Encode4(539),
-// CHECK-NEXT:     // Label 0: @462
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4(492), // Rule ID 0 //
+// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(19), GIMT_Encode2({{[0-9]+}}), /*)*//*default:*//*Label 3*/ GIMT_Encode4([[L579:[0-9]+]]),
+// CHECK-NEXT:     /*TargetOpcode::COPY*//*Label 0*/ GIMT_Encode4([[L462:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_CONSTANT*//*Label 1*/ GIMT_Encode4([[L493:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 2*/ GIMT_Encode4({{[0-9]+}}),
+// CHECK-NEXT:     // Label 0: @[[L462]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4([[L492:[0-9]+]]), // Rule ID 0 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       GIM_CheckType, /*MI*/0, /*Op*/1, /*Type*/GILLT_s32,
 // CHECK-NEXT:       // MIs[0] a
@@ -51,10 +51,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddImm8, /*InsnID*/0, /*Imm*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 4: @492
+// CHECK-NEXT:     // Label 4: @[[L492]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 1: @493
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 5*/ GIMT_Encode4(538), // Rule ID 2 //
+// CHECK-NEXT:     // Label 1: @[[L493]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 5*/ GIMT_Encode4([[L538:[0-9]+]]), // Rule ID 2 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule2Enabled),
 // CHECK-NEXT:       GIM_CheckType, /*MI*/0, /*Op*/1, /*Type*/GILLT_s32,
 // CHECK-NEXT:       // MIs[0] a
@@ -66,10 +66,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddCImm, /*InsnID*/0, /*Type*/GILLT_s32, /*Imm*/GIMT_Encode8(42),
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 5: @538
+// CHECK-NEXT:     // Label 5: @[[L538]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 2: @539
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 6*/ GIMT_Encode4(578), // Rule ID 1 //
+// CHECK-NEXT:     // Label 2: @{{[0-9]+}}
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 6*/ GIMT_Encode4([[L578:[0-9]+]]), // Rule ID 1 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule1Enabled),
 // CHECK-NEXT:       // MIs[0] a
 // CHECK-NEXT:       // No operand predicates
@@ -83,10 +83,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/0, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 6: @578
+// CHECK-NEXT:     // Label 6: @[[L578]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 3: @579
+// CHECK-NEXT:     // Label 3: @[[L579]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     }; // Size: 580 bytes
+// CHECK-NEXT:     }; // Size: {{[0-9]+}} bytes
 // CHECK-NEXT:   return MatchTable0;
 // CHECK-NEXT: }
diff --git a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-intrinsics.td b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-intrinsics.td
index e8e6d3e74f402..b2dd8b6684b1d 100644
--- a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-intrinsics.td
+++ b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-intrinsics.td
@@ -29,11 +29,11 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 
 //      CHECK: const uint8_t *GenMyCombiner::getMatchTable() const {
 // CHECK-NEXT:   constexpr static uint8_t MatchTable0[] = {
-// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(119), GIMT_Encode2(121), /*)*//*default:*//*Label 2*/ GIMT_Encode4(132),
+// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2({{[0-9]+}}), GIMT_Encode2({{[0-9]+}}), /*)*//*default:*//*Label 2*/ GIMT_Encode4([[L132:[0-9]+]]),
 // CHECK-NEXT:     /*TargetOpcode::G_INTRINSIC*//*Label 0*/ GIMT_Encode4(18),
-// CHECK-NEXT:     /*TargetOpcode::G_INTRINSIC_W_SIDE_EFFECTS*//*Label 1*/ GIMT_Encode4(73),
+// CHECK-NEXT:     /*TargetOpcode::G_INTRINSIC_W_SIDE_EFFECTS*//*Label 1*/ GIMT_Encode4([[L73:[0-9]+]]),
 // CHECK-NEXT:     // Label 0: @18
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 3*/ GIMT_Encode4(72), // Rule ID 0 //
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 3*/ GIMT_Encode4([[L72:[0-9]+]]), // Rule ID 0 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       GIM_CheckNumOperands, /*MI*/0, /*Expected*/3,
 // CHECK-NEXT:       GIM_CheckIntrinsicID, /*MI*/0, /*Op*/1, GIMT_Encode2(Intrinsic::1in_1out),
@@ -52,10 +52,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/1, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 3: @72
+// CHECK-NEXT:     // Label 3: @[[L72]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 1: @73
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4(131), // Rule ID 1 //
+// CHECK-NEXT:     // Label 1: @[[L73]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4([[L131:[0-9]+]]), // Rule ID 1 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule1Enabled),
 // CHECK-NEXT:       GIM_CheckNumOperands, /*MI*/0, /*Expected*/3,
 // CHECK-NEXT:       GIM_CheckIntrinsicID, /*MI*/0, /*Op*/1, GIMT_Encode2(Intrinsic::sideeffects_1in_1out),
@@ -76,9 +76,9 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_MergeMemOperands, /*InsnID*/1, /*NumInsns*/1, /*MergeInsnID's*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 4: @131
+// CHECK-NEXT:     // Label 4: @[[L131]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 2: @132
+// CHECK-NEXT:     // Label 2: @[[L132]]
 // CHECK-NEXT:     GIM_Reject,
 // CHECK-NEXT:     }; // Size: 133 bytes
 // CHECK-NEXT:   return MatchTable0;
diff --git a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-patfrag-root.td b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-patfrag-root.td
index 26a0ec6235e30..016ab05ca01e4 100644
--- a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-patfrag-root.td
+++ b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table-patfrag-root.td
@@ -28,12 +28,12 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 
 //      CHECK: const uint8_t *GenMyCombiner::getMatchTable() const {
 // CHECK-NEXT:   constexpr static uint8_t MatchTable0[] = {
-// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(124), GIMT_Encode2(187), /*)*//*default:*//*Label 3*/ GIMT_Encode4(380),
-// CHECK-NEXT:     /*TargetOpcode::G_TRUNC*//*Label 0*/ GIMT_Encode4(262), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 1*/ GIMT_Encode4(298), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_FPEXT*//*Label 2*/ GIMT_Encode4(344),
-// CHECK-NEXT:     // Label 0: @262
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4(297), // Rule ID 1 //
+// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2({{[0-9]+}}), GIMT_Encode2({{[0-9]+}}), /*)*//*default:*//*Label 3*/ GIMT_Encode4([[L380:[0-9]+]]),
+// CHECK-NEXT:     /*TargetOpcode::G_TRUNC*//*Label 0*/ GIMT_Encode4([[L262:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 1*/ GIMT_Encode4([[L298:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_FPEXT*//*Label 2*/ GIMT_Encode4([[L344:[0-9]+]]),
+// CHECK-NEXT:     // Label 0: @[[L262]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 4*/ GIMT_Encode4([[L297:[0-9]+]]), // Rule ID 1 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       // MIs[0] root
 // CHECK-NEXT:       // No operand predicates
@@ -47,10 +47,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/0, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 4: @297
+// CHECK-NEXT:     // Label 4: @[[L297]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 1: @298
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 5*/ GIMT_Encode4(343), // Rule ID 0 //
+// CHECK-NEXT:     // Label 1: @[[L298]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 5*/ GIMT_Encode4([[L343:[0-9]+]]), // Rule ID 0 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       // MIs[0] root
 // CHECK-NEXT:       // No operand predicates
@@ -68,10 +68,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/0, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 5: @343
+// CHECK-NEXT:     // Label 5: @[[L343]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 2: @344
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 6*/ GIMT_Encode4(379), // Rule ID 2 //
+// CHECK-NEXT:     // Label 2: @[[L344]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 6*/ GIMT_Encode4([[L379:[0-9]+]]), // Rule ID 2 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       // MIs[0] root
 // CHECK-NEXT:       // No operand predicates
@@ -85,10 +85,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/0, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 6: @379
+// CHECK-NEXT:     // Label 6: @[[L379]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 3: @380
+// CHECK-NEXT:     // Label 3: @[[L380]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     }; // Size: 381 bytes
+// CHECK-NEXT:     }; // Size: {{[0-9]+}} bytes
 // CHECK-NEXT:   return MatchTable0;
 // CHECK-NEXT: }
diff --git a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table.td b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table.td
index 0189d3d056fc0..02085c1fd2666 100644
--- a/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table.td
+++ b/llvm/test/TableGen/GlobalISelCombinerEmitter/match-table.td
@@ -135,15 +135,15 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // Verify match table.
 // CHECK:      const uint8_t *GenMyCombiner::getMatchTable() const {
 // CHECK-NEXT:   constexpr static uint8_t MatchTable0[] = {
-// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(19), GIMT_Encode2(132), /*)*//*default:*//*Label 6*/ GIMT_Encode4(677),
-// CHECK-NEXT:     /*TargetOpcode::COPY*//*Label 0*/ GIMT_Encode4(462), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_AND*//*Label 1*/ GIMT_Encode4(504), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_STORE*//*Label 2*/ GIMT_Encode4(557), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_TRUNC*//*Label 3*/ GIMT_Encode4(599), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_SEXT*//*Label 4*/ GIMT_Encode4(624), GIMT_Encode4(0),
-// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 5*/ GIMT_Encode4(637),
-// CHECK-NEXT:     // Label 0: @462
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 7*/ GIMT_Encode4(491), // Rule ID 4 //
+// CHECK-NEXT:     GIM_SwitchOpcode, /*MI*/0, /*[*/GIMT_Encode2(19), GIMT_Encode2({{[0-9]+}}), /*)*//*default:*//*Label 6*/ GIMT_Encode4([[L677:[0-9]+]]),
+// CHECK-NEXT:     /*TargetOpcode::COPY*//*Label 0*/ GIMT_Encode4([[L462:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_AND*//*Label 1*/ GIMT_Encode4([[L504:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_STORE*//*Label 2*/ GIMT_Encode4([[L557:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_TRUNC*//*Label 3*/ GIMT_Encode4([[L599:[0-9]+]]), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_SEXT*//*Label 4*/ GIMT_Encode4([[L624:[0-9]+]]), GIMT_Encode4(0),
+// CHECK-NEXT:     /*TargetOpcode::G_ZEXT*//*Label 5*/ GIMT_Encode4([[L637:[0-9]+]]),
+// CHECK-NEXT:     // Label 0: @[[L462]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 7*/ GIMT_Encode4([[L491:[0-9]+]]), // Rule ID 4 //
 // CHECK-NEXT:       GIM_CheckFeatures, GIMT_Encode2(GIFBS_HasAnswerToEverything),
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule3Enabled),
 // CHECK-NEXT:       // MIs[0] a
@@ -158,8 +158,8 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       // Combiner Rule #3: InstTest1
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner0),
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 7: @491
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 8*/ GIMT_Encode4(503), // Rule ID 3 //
+// CHECK-NEXT:     // Label 7: @[[L491]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 8*/ GIMT_Encode4([[L503:[0-9]+]]), // Rule ID 3 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule2Enabled),
 // CHECK-NEXT:       // MIs[0] a
 // CHECK-NEXT:       // No operand predicates
@@ -168,10 +168,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       // Combiner Rule #2: InstTest0
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner1),
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 8: @503
+// CHECK-NEXT:     // Label 8: @[[L503]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 1: @504
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 9*/ GIMT_Encode4(556), // Rule ID 6 //
+// CHECK-NEXT:     // Label 1: @[[L504]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 9*/ GIMT_Encode4([[L556:[0-9]+]]), // Rule ID 6 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule5Enabled),
 // CHECK-NEXT:       GIM_CheckType, /*MI*/0, /*Op*/2, /*Type*/GILLT_s32,
 // CHECK-NEXT:       // MIs[0] dst
@@ -189,10 +189,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_Copy, /*NewInsnID*/0, /*OldInsnID*/1, /*OpIdx*/1, // z
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 9: @556
+// CHECK-NEXT:     // Label 9: @[[L556]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 2: @557
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 10*/ GIMT_Encode4(598), // Rule ID 5 //
+// CHECK-NEXT:     // Label 2: @[[L557]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 10*/ GIMT_Encode4([[L598:[0-9]+]]), // Rule ID 5 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule4Enabled),
 // CHECK-NEXT:       // MIs[0] tmp
 // CHECK-NEXT:       GIM_RecordInsnIgnoreCopies, /*DefineMI*/1, /*MI*/0, /*OpIdx*/0, // MIs[1]
@@ -210,32 +210,32 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner2),
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 10: @598
+// CHECK-NEXT:     // Label 10: @[[L598]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 3: @599
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 11*/ GIMT_Encode4(611), // Rule ID 0 //
+// CHECK-NEXT:     // Label 3: @[[L599]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 11*/ GIMT_Encode4([[L611:[0-9]+]]), // Rule ID 0 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule0Enabled),
 // CHECK-NEXT:       // Combiner Rule #0: WipOpcodeTest0; wip_match_opcode 'G_TRUNC'
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner0),
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 11: @611
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 12*/ GIMT_Encode4(623), // Rule ID 1 //
+// CHECK-NEXT:     // Label 11: @[[L611]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 12*/ GIMT_Encode4([[L623:[0-9]+]]), // Rule ID 1 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule1Enabled),
 // CHECK-NEXT:       // Combiner Rule #1: WipOpcodeTest1; wip_match_opcode 'G_TRUNC'
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner0),
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 12: @623
+// CHECK-NEXT:     // Label 12: @[[L623]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 4: @624
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 13*/ GIMT_Encode4(636), // Rule ID 2 //
+// CHECK-NEXT:     // Label 4: @[[L624]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 13*/ GIMT_Encode4([[L636:[0-9]+]]), // Rule ID 2 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule1Enabled),
 // CHECK-NEXT:       // Combiner Rule #1: WipOpcodeTest1; wip_match_opcode 'G_SEXT'
 // CHECK-NEXT:       GIR_CustomAction, GIMT_Encode2(GICXXCustomAction_CombineApplyGICombiner0),
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 13: @636
+// CHECK-NEXT:     // Label 13: @[[L636]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 5: @637
-// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 14*/ GIMT_Encode4(676), // Rule ID 7 //
+// CHECK-NEXT:     // Label 5: @[[L637]]
+// CHECK-NEXT:     GIM_Try, /*On fail goto*//*Label 14*/ GIMT_Encode4([[L676:[0-9]+]]), // Rule ID 7 //
 // CHECK-NEXT:       GIM_CheckSimplePredicate, GIMT_Encode2(GICXXPred_Simple_IsRule6Enabled),
 // CHECK-NEXT:       // MIs[0] dst
 // CHECK-NEXT:       // No operand predicates
@@ -250,10 +250,10 @@ def MyCombiner: GICombiner<"GenMyCombiner", [
 // CHECK-NEXT:       GIR_AddSimpleTempRegister, /*InsnID*/0, /*TempRegID*/0,
 // CHECK-NEXT:       GIR_EraseFromParent, /*InsnID*/0,
 // CHECK-NEXT:       GIR_Done,
-// CHECK-NEXT:     // Label 14: @676
+// CHECK-NEXT:     // Label 14: @[[L676]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     // Label 6: @677
+// CHECK-NEXT:     // Label 6: @[[L677]]
 // CHECK-NEXT:     GIM_Reject,
-// CHECK-NEXT:     }; // Size: 678 bytes
+// CHECK-NEXT:     }; // Size: {{[0-9]+}} bytes
 // CHECK-NEXT:   return MatchTable0;
 // CHECK-NEXT: }
diff --git a/llvm/test/TableGen/GlobalISelEmitter.td b/llvm/test/TableGen/GlobalISelEmitter.td
index f79b792b37a36..82ecc4495e80a 100644
--- a/llvm/test/TableGen/GlobalISelEmitter.td
+++ b/llvm/test/TableGen/GlobalISelEmitter.td
@@ -518,7 +518,7 @@ def : Pat<(frag GPR32:$src1, complex:$src2, complex:$src3),
 // R00O-NEXT:  GIM_Reject,
 // R00O:       // Label [[DEFAULT_NUM]]: @[[DEFAULT]]
 // R00O-NEXT:  GIM_Reject,
-// R00O-NEXT:  }; // Size: 2023 bytes
+// R00O-NEXT:  }; // Size: 2027 bytes
 
 def INSNBOB : I<(outs GPR32:$dst), (ins GPR32:$src1, GPR32:$src2, GPR32:$src3, GPR32:$src4),
                  [(set GPR32:$dst,

From 2c22a0c16d1cb844eac142156ba67098627a336c Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Fri, 12 Apr 2024 12:35:40 -0700
Subject: [PATCH 775/886] [InstCombine] Add test case for turning sub into xor
 using dominating condition. NFC

I plan to disable using dominating conditions for turning sub into
xor, but first we need that demonstrates it currently happens.
---
 llvm/test/Transforms/InstCombine/sub-xor.ll | 23 +++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/llvm/test/Transforms/InstCombine/sub-xor.ll b/llvm/test/Transforms/InstCombine/sub-xor.ll
index 71da73d51ae37..b4e87d0405fc4 100644
--- a/llvm/test/Transforms/InstCombine/sub-xor.ll
+++ b/llvm/test/Transforms/InstCombine/sub-xor.ll
@@ -157,3 +157,26 @@ define <2 x i8> @xor_add_splat_undef(<2 x i8> %x) {
   %add = add <2 x i8> %xor, <i8 42, i8 42>
   ret <2 x i8> %add
 }
+
+define i32 @xor_dominating_cond(i32 %x) {
+; CHECK-LABEL: @xor_dominating_cond(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[COND:%.*]] = icmp ult i32 [[X:%.*]], 256
+; CHECK-NEXT:    br i1 [[COND]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[A:%.*]] = xor i32 [[X]], 255
+; CHECK-NEXT:    ret i32 [[A]]
+; CHECK:       if.end:
+; CHECK-NEXT:    ret i32 [[X]]
+;
+entry:
+  %cond = icmp ult i32 %x, 256
+  br i1 %cond, label %if.then, label %if.end
+
+if.then:
+  %a = sub i32 255, %x
+  ret i32 %a
+
+if.end:
+  ret i32 %x
+}

From 421a8c5892b7e59f27b2c21452f81fa789a758fd Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Tue, 16 Apr 2024 21:25:28 -0700
Subject: [PATCH 776/886] [InstCombine] Add phase ordering test for #88239. NFC

---
 .../Transforms/PhaseOrdering/X86/pr88239.ll   | 55 +++++++++++++++++++
 1 file changed, 55 insertions(+)
 create mode 100644 llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll

diff --git a/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll b/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll
new file mode 100644
index 0000000000000..3afa1904fb249
--- /dev/null
+++ b/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll
@@ -0,0 +1,55 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes="default<O3>" -mcpu=skx -S | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @foo(ptr noalias noundef %0, ptr noalias noundef %1) optsize {
+; CHECK-LABEL: define void @foo(
+; CHECK-SAME: ptr noalias nocapture noundef readonly [[TMP0:%.*]], ptr noalias nocapture noundef writeonly [[TMP1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  vector.ph:
+; CHECK-NEXT:    br label [[TMP4:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[TMP2:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[TMP4]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>, [[TMP2]] ], [ [[VEC_IND_NEXT:%.*]], [[TMP4]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = and <8 x i64> [[VEC_IND]], <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
+; CHECK-NEXT:    [[TMP3:%.*]] = xor <8 x i64> [[TMP6]], <i64 255, i64 255, i64 255, i64 255, i64 255, i64 255, i64 255, i64 255>
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], <8 x i64> [[TMP3]]
+; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = tail call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> [[TMP7]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i32> poison)
+; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_GATHER]], <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw i64 [[INDVARS_IV]], 8
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 256
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[MIDDLE_BLOCK:%.*]], label [[TMP4]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    ret void
+;
+  br label %3
+
+3:                                                ; preds = %7, %2
+  %4 = phi i32 [ 0, %2 ], [ %15, %7 ]
+  %5 = icmp slt i32 %4, 256
+  br i1 %5, label %7, label %6
+
+6:                                                ; preds = %3
+  ret void
+
+7:                                                ; preds = %3
+  %8 = sub nsw i32 255, %4
+  %9 = zext nneg i32 %8 to i64
+  %10 = getelementptr inbounds i32, ptr %0, i64 %9
+  %11 = load i32, ptr %10, align 4
+  %12 = add nsw i32 %11, 5
+  %13 = sext i32 %4 to i64
+  %14 = getelementptr inbounds i32, ptr %1, i64 %13
+  store i32 %12, ptr %14, align 4
+  %15 = add nsw i32 %4, 1
+  br label %3
+}
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+;.

From ed741ffe893698cd14c6785ac2ee7031d9d344a6 Mon Sep 17 00:00:00 2001
From: Nathan Lanza <nathanlanza@gmail.com>
Date: Wed, 17 Apr 2024 14:06:01 -0400
Subject: [PATCH 777/886] [github] Add ClangIR to new-prs-labeler.yml (#86088)

---
 .github/new-prs-labeler.yml | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/.github/new-prs-labeler.yml b/.github/new-prs-labeler.yml
index a0428336d300f..1502d64a7d3e3 100644
--- a/.github/new-prs-labeler.yml
+++ b/.github/new-prs-labeler.yml
@@ -1,3 +1,9 @@
+ClangIR:
+  - clang/include/clang/CIR/**/*
+  - clang/lib/CIR/**/*
+  - clang/tools/cir-*/**/*
+  - clang/test/CIR/**/*
+
 clang:dataflow:
   - clang/include/clang/Analysis/FlowSensitive/**/*
   - clang/lib/Analysis/FlowSensitive/**/*

From c02ed29ec151d1d555c3735efef2ab215126ddbf Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 17 Apr 2024 19:01:44 +0100
Subject: [PATCH 778/886] [CostModel][X86] Recognise vector rotation by uniform
 constant patterns

Adds suitable costs for AVX512 targets (we still rely on default expansion for AVX2 and earlier)
---
 .../lib/Target/X86/X86TargetTransformInfo.cpp | 41 +++++++++++++---
 .../Analysis/CostModel/X86/fshl-codesize.ll   | 24 +++++-----
 .../Analysis/CostModel/X86/fshl-latency.ll    | 24 +++++-----
 .../CostModel/X86/fshl-sizelatency.ll         | 24 +++++-----
 llvm/test/Analysis/CostModel/X86/fshl.ll      | 22 ++++-----
 .../Analysis/CostModel/X86/fshr-codesize.ll   | 48 +++++++++----------
 .../Analysis/CostModel/X86/fshr-latency.ll    | 24 +++++-----
 .../CostModel/X86/fshr-sizelatency.ll         | 48 +++++++++----------
 llvm/test/Analysis/CostModel/X86/fshr.ll      | 22 ++++-----
 .../SLPVectorizer/X86/arith-fshl-rot.ll       | 40 ++++++++++------
 .../SLPVectorizer/X86/arith-fshr-rot.ll       | 40 ++++++++++------
 11 files changed, 204 insertions(+), 153 deletions(-)

diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 38064f9792699..d111c4d4ecc1a 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -3402,6 +3402,9 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     { ISD::ROTR,       MVT::v32i16,  {  1,  1,  1,  1 } },
     { ISD::ROTR,       MVT::v16i16,  {  1,  1,  1,  1 } },
     { ISD::ROTR,       MVT::v8i16,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v32i16,  {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v16i16,  {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v8i16,   {  1,  1,  1,  1 } },
   };
   static const CostKindTblEntry AVX512BITALGCostTbl[] = {
     { ISD::CTPOP,      MVT::v32i16,  {  1,  1,  1,  1 } },
@@ -3498,6 +3501,12 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     { ISD::ROTR,       MVT::v64i8,   {  5,  6, 12, 14 } },
     { ISD::ROTR,       MVT::v32i8,   {  5, 14,  6,  9 } },
     { ISD::ROTR,       MVT::v16i8,   {  5, 14,  6,  9 } },
+    { X86ISD::VROTLI,  MVT::v32i16,  {  2,  5,  3,  3 } },
+    { X86ISD::VROTLI,  MVT::v16i16,  {  1,  5,  3,  3 } },
+    { X86ISD::VROTLI,  MVT::v8i16,   {  1,  5,  3,  3 } },
+    { X86ISD::VROTLI,  MVT::v64i8,   {  2,  9,  3,  4 } },
+    { X86ISD::VROTLI,  MVT::v32i8,   {  1,  9,  3,  4 } },
+    { X86ISD::VROTLI,  MVT::v16i8,   {  1,  8,  3,  4 } },
     { ISD::SADDSAT,    MVT::v32i16,  {  1 } },
     { ISD::SADDSAT,    MVT::v64i8,   {  1 } },
     { ISD::SMAX,       MVT::v32i16,  {  1,  1,  1,  1 } },
@@ -3556,6 +3565,12 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     { ISD::ROTR,       MVT::v16i32,  {  1,  1,  1,  1 } },
     { ISD::ROTR,       MVT::v8i32,   {  1,  1,  1,  1 } },
     { ISD::ROTR,       MVT::v4i32,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v8i64,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v4i64,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v2i64,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v16i32,  {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v8i32,   {  1,  1,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v4i32,   {  1,  1,  1,  1 } },
     { ISD::SMAX,       MVT::v8i64,   {  1,  3,  1,  1 } },
     { ISD::SMAX,       MVT::v16i32,  {  1,  1,  1,  1 } },
     { ISD::SMAX,       MVT::v32i16,  {  3,  7,  5,  5 } },
@@ -3642,7 +3657,15 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     { ISD::ROTR,       MVT::v2i64,   {  1,  3,  3,  3 } },
     { ISD::ROTR,       MVT::v4i32,   {  1,  3,  3,  3 } },
     { ISD::ROTR,       MVT::v8i16,   {  1,  3,  3,  3 } },
-    { ISD::ROTR,       MVT::v16i8,   {  1,  3,  3,  3 } }
+    { ISD::ROTR,       MVT::v16i8,   {  1,  3,  3,  3 } },
+    { X86ISD::VROTLI,  MVT::v4i64,   {  4,  7,  5,  6 } },
+    { X86ISD::VROTLI,  MVT::v8i32,   {  4,  7,  5,  6 } },
+    { X86ISD::VROTLI,  MVT::v16i16,  {  4,  7,  5,  6 } },
+    { X86ISD::VROTLI,  MVT::v32i8,   {  4,  7,  5,  6 } },
+    { X86ISD::VROTLI,  MVT::v2i64,   {  1,  3,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v4i32,   {  1,  3,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v8i16,   {  1,  3,  1,  1 } },
+    { X86ISD::VROTLI,  MVT::v16i8,   {  1,  3,  1,  1 } },
   };
   static const CostKindTblEntry AVX2CostTbl[] = {
     { ISD::ABS,        MVT::v2i64,   {  2,  4,  3,  5 } }, // VBLENDVPD(X,VPSUBQ(0,X),X)
@@ -4096,9 +4119,11 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
       const SmallVectorImpl<const Value *> &Args = ICA.getArgs();
       if (Args[0] == Args[1]) {
         ISD = ISD::ROTL;
-        // Handle scalar constant rotation amounts.
-        // TODO: Handle vector + funnel-shift cases.
-        if (isa_and_nonnull<ConstantInt>(Args[2]))
+        // Handle uniform constant rotation amounts.
+        // TODO: Handle funnel-shift cases.
+        const APInt *Amt;
+        if (Args[2] &&
+            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowUndef(Amt)))
           ISD = X86ISD::VROTLI;
       }
     }
@@ -4109,10 +4134,12 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     if (!ICA.isTypeBasedOnly()) {
       const SmallVectorImpl<const Value *> &Args = ICA.getArgs();
       if (Args[0] == Args[1]) {
-        // Handle scalar constant rotation amount.
-        // TODO: Handle vector + funnel-shift cases.
         ISD = ISD::ROTR;
-        if (isa_and_nonnull<ConstantInt>(Args[2]))
+        // Handle uniform constant rotation amount.
+        // TODO: Handle funnel-shift cases.
+        const APInt *Amt;
+        if (Args[2] &&
+            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowUndef(Amt)))
           ISD = X86ISD::VROTLI;
       }
     }
diff --git a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
index c46e32ffb4ad3..a7585a4d9f39e 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-codesize.ll
@@ -2744,9 +2744,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -2829,9 +2829,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2843,9 +2843,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2871,9 +2871,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
index fa32497c63ec7..7105f713fdc34 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-latency.ll
@@ -2696,9 +2696,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -2781,9 +2781,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2795,9 +2795,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2823,9 +2823,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
index 832a574a9b332..5d7361e293176 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl-sizelatency.ll
@@ -2984,9 +2984,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -3069,9 +3069,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -3083,9 +3083,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -3111,9 +3111,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index 311d8d5ed7d2a..1cbdab09acd90 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -2682,8 +2682,8 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshl.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
@@ -2767,9 +2767,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2781,9 +2781,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2809,9 +2809,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshl.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
index f9d30e4ced3ec..ecc861dd7f8ee 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-codesize.ll
@@ -2644,9 +2644,9 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i64'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I64    = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
@@ -2701,9 +2701,9 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i32'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I32   = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
@@ -2744,9 +2744,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -2779,9 +2779,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i16'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
@@ -2829,9 +2829,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2843,9 +2843,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2864,16 +2864,16 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i8'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
index ed2227591847a..0142ad77849ca 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-latency.ll
@@ -2696,9 +2696,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -2781,9 +2781,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2795,9 +2795,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2823,9 +2823,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
index 8931781f70bdc..6dafb20a0aeed 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr-sizelatency.ll
@@ -2849,9 +2849,9 @@ define void @splatconstant_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i64'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I64 = call i64 @llvm.fshr.i64(i64 %a64, i64 %a64, i64 7)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> <i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> <i64 7, i64 7, i64 7, i64 7>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i64'
@@ -2934,9 +2934,9 @@ define void @splatconstant_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i32'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I32 = call i32 @llvm.fshr.i32(i32 %a32, i32 %a32, i32 5)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> <i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i32'
@@ -2984,9 +2984,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i16'
@@ -3019,9 +3019,9 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i16'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i16'
@@ -3069,9 +3069,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -3083,9 +3083,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -3104,16 +3104,16 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; XOP-LABEL: 'splatconstant_rotate_i8'
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; XOP-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; XOP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; XOP-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index ca9ddcc52938d..ada1b9c5bdc4b 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -2682,8 +2682,8 @@ define void @splatconstant_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a25
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i16'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I16 = call i16 @llvm.fshr.i16(i16 %a16, i16 %a16, i16 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
@@ -2767,9 +2767,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512BW-LABEL: 'splatconstant_rotate_i8'
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512BW-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; AVX512DQ-LABEL: 'splatconstant_rotate_i8'
@@ -2781,9 +2781,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512VBMI2-LABEL: 'splatconstant_rotate_i8'
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512VBMI2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; SLM-LABEL: 'splatconstant_rotate_i8'
@@ -2809,9 +2809,9 @@ define void @splatconstant_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <
 ;
 ; AVX512GFNI-LABEL: 'splatconstant_rotate_i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %I8 = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
-; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
+; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>)
 ; AVX512GFNI-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %I8    = call i8 @llvm.fshr.i8(i8 %a8, i8 %a8, i8 3)
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/arith-fshl-rot.ll b/llvm/test/Transforms/SLPVectorizer/X86/arith-fshl-rot.ll
index 639aa0a1c6a2c..9b8480cd0088a 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/arith-fshl-rot.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/arith-fshl-rot.ll
@@ -834,22 +834,34 @@ define void @fshl_v2i32_uniformconst() {
 ; SSE-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
 ; SSE-NEXT:    ret void
 ;
-; AVX-LABEL: @fshl_v2i32_uniformconst(
-; AVX-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
-; AVX-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
-; AVX-NEXT:    [[R0:%.*]] = call i32 @llvm.fshl.i32(i32 [[A0]], i32 [[A0]], i32 1)
-; AVX-NEXT:    [[R1:%.*]] = call i32 @llvm.fshl.i32(i32 [[A1]], i32 [[A1]], i32 1)
-; AVX-NEXT:    store i32 [[R0]], ptr @d32, align 4
-; AVX-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
-; AVX-NEXT:    ret void
+; AVX1-LABEL: @fshl_v2i32_uniformconst(
+; AVX1-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
+; AVX1-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
+; AVX1-NEXT:    [[R0:%.*]] = call i32 @llvm.fshl.i32(i32 [[A0]], i32 [[A0]], i32 1)
+; AVX1-NEXT:    [[R1:%.*]] = call i32 @llvm.fshl.i32(i32 [[A1]], i32 [[A1]], i32 1)
+; AVX1-NEXT:    store i32 [[R0]], ptr @d32, align 4
+; AVX1-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX1-NEXT:    ret void
+;
+; AVX2-LABEL: @fshl_v2i32_uniformconst(
+; AVX2-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
+; AVX2-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
+; AVX2-NEXT:    [[R0:%.*]] = call i32 @llvm.fshl.i32(i32 [[A0]], i32 [[A0]], i32 1)
+; AVX2-NEXT:    [[R1:%.*]] = call i32 @llvm.fshl.i32(i32 [[A1]], i32 [[A1]], i32 1)
+; AVX2-NEXT:    store i32 [[R0]], ptr @d32, align 4
+; AVX2-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX2-NEXT:    ret void
+;
+; AVX256-LABEL: @fshl_v2i32_uniformconst(
+; AVX256-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr @a32, align 4
+; AVX256-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>)
+; AVX256-NEXT:    store <2 x i32> [[TMP2]], ptr @d32, align 4
+; AVX256-NEXT:    ret void
 ;
 ; AVX512-LABEL: @fshl_v2i32_uniformconst(
-; AVX512-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
-; AVX512-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
-; AVX512-NEXT:    [[R0:%.*]] = call i32 @llvm.fshl.i32(i32 [[A0]], i32 [[A0]], i32 1)
-; AVX512-NEXT:    [[R1:%.*]] = call i32 @llvm.fshl.i32(i32 [[A1]], i32 [[A1]], i32 1)
-; AVX512-NEXT:    store i32 [[R0]], ptr @d32, align 4
-; AVX512-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX512-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr @a32, align 4
+; AVX512-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>)
+; AVX512-NEXT:    store <2 x i32> [[TMP2]], ptr @d32, align 4
 ; AVX512-NEXT:    ret void
 ;
   %a0  = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 0 ), align 4
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/arith-fshr-rot.ll b/llvm/test/Transforms/SLPVectorizer/X86/arith-fshr-rot.ll
index c557c9647551a..f3e73d0e6840e 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/arith-fshr-rot.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/arith-fshr-rot.ll
@@ -834,22 +834,34 @@ define void @fshr_v2i32_uniformconst() {
 ; SSE-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
 ; SSE-NEXT:    ret void
 ;
-; AVX-LABEL: @fshr_v2i32_uniformconst(
-; AVX-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
-; AVX-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
-; AVX-NEXT:    [[R0:%.*]] = call i32 @llvm.fshr.i32(i32 [[A0]], i32 [[A0]], i32 1)
-; AVX-NEXT:    [[R1:%.*]] = call i32 @llvm.fshr.i32(i32 [[A1]], i32 [[A1]], i32 1)
-; AVX-NEXT:    store i32 [[R0]], ptr @d32, align 4
-; AVX-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
-; AVX-NEXT:    ret void
+; AVX1-LABEL: @fshr_v2i32_uniformconst(
+; AVX1-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
+; AVX1-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
+; AVX1-NEXT:    [[R0:%.*]] = call i32 @llvm.fshr.i32(i32 [[A0]], i32 [[A0]], i32 1)
+; AVX1-NEXT:    [[R1:%.*]] = call i32 @llvm.fshr.i32(i32 [[A1]], i32 [[A1]], i32 1)
+; AVX1-NEXT:    store i32 [[R0]], ptr @d32, align 4
+; AVX1-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX1-NEXT:    ret void
+;
+; AVX2-LABEL: @fshr_v2i32_uniformconst(
+; AVX2-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
+; AVX2-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
+; AVX2-NEXT:    [[R0:%.*]] = call i32 @llvm.fshr.i32(i32 [[A0]], i32 [[A0]], i32 1)
+; AVX2-NEXT:    [[R1:%.*]] = call i32 @llvm.fshr.i32(i32 [[A1]], i32 [[A1]], i32 1)
+; AVX2-NEXT:    store i32 [[R0]], ptr @d32, align 4
+; AVX2-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX2-NEXT:    ret void
+;
+; AVX256-LABEL: @fshr_v2i32_uniformconst(
+; AVX256-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr @a32, align 4
+; AVX256-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>)
+; AVX256-NEXT:    store <2 x i32> [[TMP2]], ptr @d32, align 4
+; AVX256-NEXT:    ret void
 ;
 ; AVX512-LABEL: @fshr_v2i32_uniformconst(
-; AVX512-NEXT:    [[A0:%.*]] = load i32, ptr @a32, align 4
-; AVX512-NEXT:    [[A1:%.*]] = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 1), align 4
-; AVX512-NEXT:    [[R0:%.*]] = call i32 @llvm.fshr.i32(i32 [[A0]], i32 [[A0]], i32 1)
-; AVX512-NEXT:    [[R1:%.*]] = call i32 @llvm.fshr.i32(i32 [[A1]], i32 [[A1]], i32 1)
-; AVX512-NEXT:    store i32 [[R0]], ptr @d32, align 4
-; AVX512-NEXT:    store i32 [[R1]], ptr getelementptr inbounds ([16 x i32], ptr @d32, i32 0, i64 1), align 4
+; AVX512-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr @a32, align 4
+; AVX512-NEXT:    [[TMP2:%.*]] = call <2 x i32> @llvm.fshr.v2i32(<2 x i32> [[TMP1]], <2 x i32> [[TMP1]], <2 x i32> <i32 1, i32 1>)
+; AVX512-NEXT:    store <2 x i32> [[TMP2]], ptr @d32, align 4
 ; AVX512-NEXT:    ret void
 ;
   %a0  = load i32, ptr getelementptr inbounds ([16 x i32], ptr @a32, i32 0, i64 0 ), align 4

From 58a08e154c804051aaca9151a8053aea3ec15646 Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Wed, 17 Apr 2024 11:15:25 -0700
Subject: [PATCH 779/886] [RISCV] Add coverage for strength reduction of mul by
 small negative immediates

---
 llvm/test/CodeGen/RISCV/rv32zba.ll | 93 ++++++++++++++++++++++++++++++
 llvm/test/CodeGen/RISCV/rv64zba.ll | 93 ++++++++++++++++++++++++++++++
 2 files changed, 186 insertions(+)

diff --git a/llvm/test/CodeGen/RISCV/rv32zba.ll b/llvm/test/CodeGen/RISCV/rv32zba.ll
index cc632a09c8054..a78f823d31841 100644
--- a/llvm/test/CodeGen/RISCV/rv32zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv32zba.ll
@@ -645,3 +645,96 @@ define i32 @addshl_5_8(i32 %a, i32 %b) {
   %e = add i32 %c, %d
   ret i32 %e
 }
+
+define i32 @mul_neg1(i32 %a) {
+; CHECK-LABEL: mul_neg1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -1
+  ret i32 %c
+}
+
+define i32 @mul_neg2(i32 %a) {
+; CHECK-LABEL: mul_neg2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 1
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -2
+  ret i32 %c
+}
+
+define i32 @mul_neg3(i32 %a) {
+; RV32I-LABEL: mul_neg3:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a1, a0, 1
+; RV32I-NEXT:    neg a0, a0
+; RV32I-NEXT:    sub a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32ZBA-LABEL: mul_neg3:
+; RV32ZBA:       # %bb.0:
+; RV32ZBA-NEXT:    sh1add a0, a0, a0
+; RV32ZBA-NEXT:    neg a0, a0
+; RV32ZBA-NEXT:    ret
+  %c = mul i32 %a, -3
+  ret i32 %c
+}
+
+define i32 @mul_neg4(i32 %a) {
+; CHECK-LABEL: mul_neg4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 2
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -4
+  ret i32 %c
+}
+
+define i32 @mul_neg5(i32 %a) {
+; RV32I-LABEL: mul_neg5:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    slli a1, a0, 2
+; RV32I-NEXT:    neg a0, a0
+; RV32I-NEXT:    sub a0, a0, a1
+; RV32I-NEXT:    ret
+;
+; RV32ZBA-LABEL: mul_neg5:
+; RV32ZBA:       # %bb.0:
+; RV32ZBA-NEXT:    sh2add a0, a0, a0
+; RV32ZBA-NEXT:    neg a0, a0
+; RV32ZBA-NEXT:    ret
+  %c = mul i32 %a, -5
+  ret i32 %c
+}
+
+define i32 @mul_neg6(i32 %a) {
+; CHECK-LABEL: mul_neg6:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, -6
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -6
+  ret i32 %c
+}
+
+define i32 @mul_neg7(i32 %a) {
+; CHECK-LABEL: mul_neg7:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a1, a0, 3
+; CHECK-NEXT:    sub a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -7
+  ret i32 %c
+}
+
+define i32 @mul_neg8(i32 %a) {
+; CHECK-LABEL: mul_neg8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 3
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i32 %a, -8
+  ret i32 %c
+}
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index b4c80b60e0bad..6939185947f44 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -2533,3 +2533,96 @@ define i64 @regression(i32 signext %x, i32 signext %y) {
   %res = mul nuw nsw i64 %ext, 24
   ret i64 %res
 }
+
+define i64 @mul_neg1(i64 %a) {
+; CHECK-LABEL: mul_neg1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -1
+  ret i64 %c
+}
+
+define i64 @mul_neg2(i64 %a) {
+; CHECK-LABEL: mul_neg2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 1
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -2
+  ret i64 %c
+}
+
+define i64 @mul_neg3(i64 %a) {
+; RV64I-LABEL: mul_neg3:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a1, a0, 1
+; RV64I-NEXT:    neg a0, a0
+; RV64I-NEXT:    sub a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul_neg3:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    sh1add a0, a0, a0
+; RV64ZBA-NEXT:    neg a0, a0
+; RV64ZBA-NEXT:    ret
+  %c = mul i64 %a, -3
+  ret i64 %c
+}
+
+define i64 @mul_neg4(i64 %a) {
+; CHECK-LABEL: mul_neg4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 2
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -4
+  ret i64 %c
+}
+
+define i64 @mul_neg5(i64 %a) {
+; RV64I-LABEL: mul_neg5:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    slli a1, a0, 2
+; RV64I-NEXT:    neg a0, a0
+; RV64I-NEXT:    sub a0, a0, a1
+; RV64I-NEXT:    ret
+;
+; RV64ZBA-LABEL: mul_neg5:
+; RV64ZBA:       # %bb.0:
+; RV64ZBA-NEXT:    sh2add a0, a0, a0
+; RV64ZBA-NEXT:    neg a0, a0
+; RV64ZBA-NEXT:    ret
+  %c = mul i64 %a, -5
+  ret i64 %c
+}
+
+define i64 @mul_neg6(i64 %a) {
+; CHECK-LABEL: mul_neg6:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, -6
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -6
+  ret i64 %c
+}
+
+define i64 @mul_neg7(i64 %a) {
+; CHECK-LABEL: mul_neg7:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a1, a0, 3
+; CHECK-NEXT:    sub a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -7
+  ret i64 %c
+}
+
+define i64 @mul_neg8(i64 %a) {
+; CHECK-LABEL: mul_neg8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 3
+; CHECK-NEXT:    neg a0, a0
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, -8
+  ret i64 %c
+}

From cc82f1290a1e2157a6c0530d78d8cc84d2b8553d Mon Sep 17 00:00:00 2001
From: Usman Nadeem <mnadeem@quicinc.com>
Date: Wed, 17 Apr 2024 11:42:52 -0700
Subject: [PATCH 780/886] [AArch64] Update latencies for Cortex-A510 scheduling
 model (#87293)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Updated according to the Software Optimization Guide for Arm®
Cortex®‑A510 Core Revision: r1p3 Issue 6.0.
---
 llvm/lib/Target/AArch64/AArch64SchedA510.td   |  145 +-
 .../AArch64/GlobalISel/combine-udiv.ll        |    8 +-
 llvm/test/CodeGen/AArch64/aarch64-addv.ll     |   10 +-
 .../AArch64/aarch64-dup-ext-scalable.ll       |   40 +-
 llvm/test/CodeGen/AArch64/aarch64-smull.ll    |    2 +-
 llvm/test/CodeGen/AArch64/active_lane_mask.ll |  169 +-
 .../CodeGen/AArch64/arm64-convert-v4f64.ll    |   14 +-
 llvm/test/CodeGen/AArch64/arm64-vabs.ll       |  106 +-
 llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll     |   28 +-
 llvm/test/CodeGen/AArch64/arm64-vhadd.ll      |    8 +-
 ...interleaving-add-mull-scalable-contract.ll |   64 +-
 ...x-deinterleaving-add-mull-scalable-fast.ll |   63 +-
 ...complex-deinterleaving-f16-mul-scalable.ll |   23 +-
 ...complex-deinterleaving-f32-mul-scalable.ll |   18 +-
 ...complex-deinterleaving-f64-mul-scalable.ll |   18 +-
 ...complex-deinterleaving-i16-mul-scalable.ll |    3 +-
 ...rleaving-reductions-predicated-scalable.ll |   64 +-
 ...plex-deinterleaving-reductions-scalable.ll |   26 +-
 .../complex-deinterleaving-splat-scalable.ll  |   16 +-
 .../AArch64/concat_vector-truncate-combine.ll |    2 +-
 .../AArch64/dag-combine-concat-vectors.ll     |   12 +-
 .../div-rem-pair-recomposition-signed.ll      |    6 +-
 .../div-rem-pair-recomposition-unsigned.ll    |    6 +-
 llvm/test/CodeGen/AArch64/extbinopload.ll     |  121 +-
 llvm/test/CodeGen/AArch64/fcmp.ll             |  108 +-
 llvm/test/CodeGen/AArch64/fdiv-combine.ll     |    2 +-
 .../fold-int-pow2-with-fmul-or-fdiv.ll        |    2 +-
 .../CodeGen/AArch64/fp-veclib-expansion.ll    |    8 +-
 llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll |    8 +-
 .../test/CodeGen/AArch64/fptosi-sat-vector.ll |   20 +-
 llvm/test/CodeGen/AArch64/funnel-shift-rot.ll |    4 +-
 ...st-and-by-const-from-lshr-in-eqcmp-zero.ll |    2 +-
 llvm/test/CodeGen/AArch64/icmp.ll             |    4 +-
 llvm/test/CodeGen/AArch64/insert-extend.ll    |   54 +-
 .../insert-subvector-res-legalization.ll      |   70 +-
 .../AArch64/intrinsic-cttz-elts-sve.ll        |   91 +-
 llvm/test/CodeGen/AArch64/itofp.ll            | 1132 ++++++------
 llvm/test/CodeGen/AArch64/ldexp.ll            |    6 +-
 .../CodeGen/AArch64/llvm-ir-to-intrinsic.ll   |   43 +-
 llvm/test/CodeGen/AArch64/load-insert-zero.ll |   28 +-
 llvm/test/CodeGen/AArch64/logic-shift.ll      |   48 +-
 .../AArch64/named-vector-shuffles-sve.ll      |   76 +-
 llvm/test/CodeGen/AArch64/neon-dotreduce.ll   |  758 ++++----
 llvm/test/CodeGen/AArch64/neon-extadd.ll      |   12 +-
 llvm/test/CodeGen/AArch64/neon-shift-neg.ll   |   24 +-
 .../CodeGen/AArch64/predicated-add-sub.ll     |   22 +-
 .../AArch64/ragreedy-local-interval-cost.ll   |  262 ++-
 llvm/test/CodeGen/AArch64/rcpc3-sve.ll        |    4 +-
 llvm/test/CodeGen/AArch64/reassocmls.ll       |    6 +-
 llvm/test/CodeGen/AArch64/reduce-shuffle.ll   |  356 ++--
 llvm/test/CodeGen/AArch64/sat-add.ll          |    8 +-
 llvm/test/CodeGen/AArch64/sext.ll             |   36 +-
 .../CodeGen/AArch64/sink-addsub-of-const.ll   |   24 +-
 ...ate-sm-changing-call-disable-coalescing.ll |  130 +-
 .../sme-streaming-compatible-interface.ll     |    2 +-
 .../AArch64/sme2-intrinsics-fp-dots.ll        |   12 +-
 .../AArch64/sme2-intrinsics-int-dots.ll       |   42 +-
 .../CodeGen/AArch64/sme2-intrinsics-max.ll    |  305 ++-
 .../CodeGen/AArch64/sme2-intrinsics-min.ll    |  163 +-
 .../CodeGen/AArch64/sme2-intrinsics-mlall.ll  |   54 +-
 .../CodeGen/AArch64/sme2-intrinsics-rshl.ll   |  161 +-
 .../AArch64/sme2-intrinsics-sqdmulh.ll        |   89 +-
 .../CodeGen/AArch64/split-vector-insert.ll    |  265 ++-
 .../CodeGen/AArch64/srem-seteq-vec-splat.ll   |   16 +-
 llvm/test/CodeGen/AArch64/srem-vector-lkk.ll  |    4 +-
 llvm/test/CodeGen/AArch64/sve-abd.ll          |   18 +-
 llvm/test/CodeGen/AArch64/sve-bitcast.ll      |   20 +-
 .../AArch64/sve-calling-convention-mixed.ll   |  116 +-
 llvm/test/CodeGen/AArch64/sve-cmp-folds.ll    |    4 +-
 llvm/test/CodeGen/AArch64/sve-doublereduct.ll |    8 +-
 llvm/test/CodeGen/AArch64/sve-expand-div.ll   |   16 +-
 .../CodeGen/AArch64/sve-extract-element.ll    |   10 +-
 .../sve-extract-fixed-from-scalable-vector.ll |    8 +-
 .../AArch64/sve-extract-fixed-vector.ll       |   16 +-
 .../AArch64/sve-extract-scalable-vector.ll    |   18 +-
 llvm/test/CodeGen/AArch64/sve-fcmp.ll         |    2 +-
 llvm/test/CodeGen/AArch64/sve-fcopysign.ll    |   18 +-
 llvm/test/CodeGen/AArch64/sve-fcvt.ll         |   64 +-
 .../sve-fixed-length-addressing-modes.ll      |    4 +-
 .../AArch64/sve-fixed-length-build-vector.ll  |   13 +-
 .../AArch64/sve-fixed-length-concat.ll        |  112 +-
 .../sve-fixed-length-extract-vector-elt.ll    |   24 +-
 .../AArch64/sve-fixed-length-fcopysign.ll     |   16 +-
 .../sve-fixed-length-fp-extend-trunc.ll       |    6 +-
 .../AArch64/sve-fixed-length-fp-select.ll     |  234 +--
 .../AArch64/sve-fixed-length-fp-to-int.ll     |   24 +-
 .../CodeGen/AArch64/sve-fixed-length-fp128.ll |    8 +-
 .../sve-fixed-length-frame-offests-crash.ll   |   32 +-
 .../sve-fixed-length-insert-vector-elt.ll     |  222 +--
 .../AArch64/sve-fixed-length-int-arith.ll     |    4 +-
 .../AArch64/sve-fixed-length-int-div.ll       |   56 +-
 .../AArch64/sve-fixed-length-int-extends.ll   |   12 +-
 .../AArch64/sve-fixed-length-int-rem.ll       |   88 +-
 .../AArch64/sve-fixed-length-int-select.ll    |  312 ++--
 .../AArch64/sve-fixed-length-int-to-fp.ll     |   24 +-
 .../AArch64/sve-fixed-length-mask-opt.ll      |    6 +-
 .../sve-fixed-length-masked-128bit-loads.ll   |    2 +-
 .../sve-fixed-length-masked-128bit-stores.ll  |   18 +-
 .../AArch64/sve-fixed-length-masked-gather.ll |  148 +-
 .../AArch64/sve-fixed-length-masked-loads.ll  |   32 +-
 .../sve-fixed-length-masked-scatter.ll        |  156 +-
 .../AArch64/sve-fixed-length-masked-stores.ll |   42 +-
 .../AArch64/sve-fixed-length-shuffles.ll      |   26 +-
 .../AArch64/sve-fixed-length-splat-vector.ll  |   84 +-
 .../AArch64/sve-fixed-length-trunc-stores.ll  |   16 +-
 .../sve-fixed-length-vector-shuffle-tbl.ll    |    6 +-
 .../CodeGen/AArch64/sve-fp-int-min-max.ll     |    2 +-
 .../test/CodeGen/AArch64/sve-fp-reciprocal.ll |    6 +-
 .../CodeGen/AArch64/sve-fp-reduce-fadda.ll    |    4 +-
 llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll   |  282 +--
 llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll   |   32 +-
 .../AArch64/sve-gather-scatter-addr-opts.ll   |   28 +-
 llvm/test/CodeGen/AArch64/sve-hadd.ll         |   20 +-
 .../AArch64/sve-implicit-zero-filling.ll      |    8 +-
 .../CodeGen/AArch64/sve-insert-element.ll     |   88 +-
 .../test/CodeGen/AArch64/sve-insert-vector.ll |   22 +-
 .../test/CodeGen/AArch64/sve-int-arith-imm.ll |   34 +-
 llvm/test/CodeGen/AArch64/sve-int-arith.ll    |   20 +-
 llvm/test/CodeGen/AArch64/sve-int-reduce.ll   |   62 +-
 .../CodeGen/AArch64/sve-intrinsics-index.ll   |    2 +-
 .../AArch64/sve-intrinsics-int-arith-imm.ll   |   38 +-
 .../AArch64/sve-intrinsics-logical-imm.ll     |    2 +-
 llvm/test/CodeGen/AArch64/sve-ld-post-inc.ll  |   30 +-
 llvm/test/CodeGen/AArch64/sve-ld1r.ll         |   16 +-
 .../sve-lsr-scaled-index-addressing-mode.ll   |    4 +-
 .../AArch64/sve-masked-gather-legalize.ll     |    8 +-
 .../CodeGen/AArch64/sve-masked-ldst-sext.ll   |    2 +-
 .../CodeGen/AArch64/sve-masked-ldst-zext.ll   |    6 +-
 .../AArch64/sve-masked-scatter-legalize.ll    |   20 +-
 .../CodeGen/AArch64/sve-masked-scatter.ll     |    2 +-
 llvm/test/CodeGen/AArch64/sve-pr62151.ll      |    2 +-
 llvm/test/CodeGen/AArch64/sve-pred-arith.ll   |   40 +-
 .../test/CodeGen/AArch64/sve-pred-selectop.ll |  108 +-
 .../CodeGen/AArch64/sve-pred-selectop2.ll     |   76 +-
 .../CodeGen/AArch64/sve-pred-selectop3.ll     |   28 +-
 .../AArch64/sve-ptest-removal-cmple.ll        |    8 +-
 .../CodeGen/AArch64/sve-redundant-store.ll    |    2 +-
 .../CodeGen/AArch64/sve-split-extract-elt.ll  |   42 +-
 llvm/test/CodeGen/AArch64/sve-split-fcvt.ll   |   26 +-
 .../CodeGen/AArch64/sve-split-fp-reduce.ll    |    2 +-
 .../CodeGen/AArch64/sve-split-insert-elt.ll   |   26 +-
 .../CodeGen/AArch64/sve-split-int-reduce.ll   |    8 +-
 llvm/test/CodeGen/AArch64/sve-split-load.ll   |    4 +-
 llvm/test/CodeGen/AArch64/sve-split-store.ll  |    8 +-
 .../CodeGen/AArch64/sve-srem-combine-loop.ll  |    2 +-
 .../sve-st1-addressing-mode-reg-imm.ll        |    8 +-
 llvm/test/CodeGen/AArch64/sve-stepvector.ll   |    2 +-
 ...treaming-mode-fixed-length-bit-counting.ll |   36 +-
 ...e-streaming-mode-fixed-length-ext-loads.ll |   10 +-
 ...e-streaming-mode-fixed-length-fcopysign.ll |   26 +-
 ...ve-streaming-mode-fixed-length-fp-arith.ll |   48 +-
 ...streaming-mode-fixed-length-fp-compares.ll |   44 +-
 ...-streaming-mode-fixed-length-fp-convert.ll |    4 +-
 ...aming-mode-fixed-length-fp-extend-trunc.ll |   10 +-
 .../sve-streaming-mode-fixed-length-fp-fma.ll |    6 +-
 ...e-streaming-mode-fixed-length-fp-minmax.ll |   24 +-
 ...e-streaming-mode-fixed-length-fp-reduce.ll |   30 +-
 ...streaming-mode-fixed-length-fp-rounding.ll |   42 +-
 ...e-streaming-mode-fixed-length-fp-select.ll |   18 +-
 ...e-streaming-mode-fixed-length-fp-to-int.ll |  118 +-
 ...-streaming-mode-fixed-length-fp-vselect.ll |    6 +-
 ...ing-mode-fixed-length-insert-vector-elt.ll |   42 +-
 ...e-streaming-mode-fixed-length-int-arith.ll |   16 +-
 ...treaming-mode-fixed-length-int-compares.ll |   14 +-
 ...sve-streaming-mode-fixed-length-int-div.ll |   34 +-
 ...eaming-mode-fixed-length-int-immediates.ll |    8 +-
 ...-streaming-mode-fixed-length-int-minmax.ll |   32 +-
 ...ve-streaming-mode-fixed-length-int-mulh.ll |   20 +-
 ...-streaming-mode-fixed-length-int-reduce.ll |   40 +-
 ...sve-streaming-mode-fixed-length-int-rem.ll |   64 +-
 ...-streaming-mode-fixed-length-int-select.ll |   28 +-
 ...-streaming-mode-fixed-length-int-shifts.ll |   32 +-
 ...e-streaming-mode-fixed-length-int-to-fp.ll |   56 +-
 ...streaming-mode-fixed-length-int-vselect.ll |   10 +-
 ...-streaming-mode-fixed-length-log-reduce.ll |   24 +-
 ...treaming-mode-fixed-length-masked-store.ll |   16 +-
 ...eaming-mode-fixed-length-optimize-ptrue.ll |   10 +-
 ...streaming-mode-fixed-length-permute-rev.ll |   24 +-
 ...g-mode-fixed-length-permute-zip-uzp-trn.ll |   30 +-
 .../sve-streaming-mode-fixed-length-ptest.ll  |   78 +-
 .../sve-streaming-mode-fixed-length-rev.ll    |   14 +-
 ...e-streaming-mode-fixed-length-sdiv-pow2.ll |    8 +-
 ...sve-streaming-mode-fixed-length-shuffle.ll |    2 +-
 .../sve-streaming-mode-fixed-length-stores.ll |    4 +-
 .../sve-streaming-mode-fixed-length-trunc.ll  |  354 ++--
 llvm/test/CodeGen/AArch64/sve-trunc.ll        |   36 +-
 llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll |   20 +-
 .../sve-uunpklo-load-uzp1-store-combine.ll    |    8 +-
 .../test/CodeGen/AArch64/sve-vecreduce-dot.ll |    2 +-
 .../CodeGen/AArch64/sve-vecreduce-fold.ll     |    2 +-
 llvm/test/CodeGen/AArch64/sve2-fcopysign.ll   |   16 +-
 .../AArch64/sve2-fixed-length-fcopysign.ll    |   16 +-
 .../AArch64/sve2-intrinsics-combine-rshrnb.ll |   22 +-
 .../AArch64/sve2-intrinsics-int-arith-imm.ll  |    2 +-
 llvm/test/CodeGen/AArch64/sve2-rsh.ll         |    2 +-
 llvm/test/CodeGen/AArch64/sve2-xar.ll         |    2 +-
 .../AArch64/sve2p1-intrinsics-selx2.ll        |   32 +-
 .../AArch64/sve2p1-intrinsics-selx4.ll        |   64 +-
 .../AArch64/sve2p1-intrinsics-stores.ll       |   64 +-
 llvm/test/CodeGen/AArch64/uadd_sat_vec.ll     |   10 +-
 .../CodeGen/AArch64/urem-seteq-vec-nonzero.ll |   10 +-
 llvm/test/CodeGen/AArch64/vec_uaddo.ll        |    8 +-
 llvm/test/CodeGen/AArch64/vecreduce-add.ll    |  546 +++---
 llvm/test/CodeGen/AArch64/vector-fcopysign.ll |   46 +-
 llvm/test/CodeGen/AArch64/vector-gep.ll       |    4 +-
 .../test/CodeGen/AArch64/vselect-constants.ll |    5 +-
 llvm/test/CodeGen/AArch64/zext-to-tbl.ll      |   50 +-
 .../AArch64/Cortex/A510-neon-instructions.s   |  376 ++--
 .../AArch64/Cortex/A510-sve-instructions.s    | 1636 ++++++++---------
 209 files changed, 6488 insertions(+), 6481 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64SchedA510.td b/llvm/lib/Target/AArch64/AArch64SchedA510.td
index 68343674bc819..9456878946151 100644
--- a/llvm/lib/Target/AArch64/AArch64SchedA510.td
+++ b/llvm/lib/Target/AArch64/AArch64SchedA510.td
@@ -254,7 +254,7 @@ def : InstRW<[WriteIS], (instrs RBITWr, RBITXr)>;
 // Compute pointer authentication code for data address
 // Compute pointer authentication code, using generic key
 // Compute pointer authentication code for instruction address
-def : InstRW<[CortexA510Write<3, CortexA510UnitPAC>], (instregex "^AUT", "^PAC")>;
+def : InstRW<[CortexA510Write<5, CortexA510UnitPAC>], (instregex "^AUT", "^PAC")>;
 
 // Branch and link, register, with pointer authentication
 // Branch, register, with pointer authentication
@@ -401,30 +401,30 @@ def : InstRW<[CortexA510WriteFPALU_F3], (instrs FCSELHrrr, FCSELSrrr, FCSELDrrr)
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU]ABDv(2i32|4i16|8i8)")>;
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU]ABDv(16i8|4i32|8i16)")>;
 // ASIMD absolute diff accum
-def : InstRW<[CortexA510Write<8, CortexA510UnitVALU>], (instregex "[SU]ABAL?v")>;
+def : InstRW<[CortexA510Write<6, CortexA510UnitVALU>], (instregex "[SU]ABAL?v")>;
 // ASIMD absolute diff long
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU]ABDLv")>;
 // ASIMD arith #1
-def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "(ADD|SUB|NEG)v(1i64|2i32|4i16|8i8)",
-  "[SU]R?HADDv(2i32|4i16|8i8)", "[SU]HSUBv(2i32|4i16|8i8)")>;
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "(ADD|SUB|NEG)v(2i64|4i32|8i16|16i8)",
-  "[SU]R?HADDv(8i16|4i32|16i8)", "[SU]HSUBv(8i16|4i32|16i8)")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "(ADD|SUB|NEG)v",
+  "[SU]R?HADDv", "[SU]HSUBv")>;
 // ASIMD arith #2
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "ABSv(1i64|2i32|4i16|8i8)$",
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "ABSv(1i64|2i32|4i16|8i8)$",
   "[SU]ADDLPv(2i32_v1i64|4i16_v2i32|8i8_v4i16)$",
-  "([SU]QADD|[SU]QSUB|SQNEG|SUQADD|USQADD)v(1i16|1i32|1i64|1i8|2i32|4i16|8i8)$",
   "ADDPv(2i32|4i16|8i8)$")>;
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "ABSv(2i64|4i32|8i16|16i8)$",
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "([SU]QADD|[SU]QSUB|SQNEG|SUQADD|USQADD)v(1i16|1i32|1i64|1i8|2i32|4i16|8i8)$")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "ABSv(2i64|4i32|8i16|16i8)$",
   "[SU]ADDLPv(16i8_v8i16|4i32_v2i64|8i16_v4i32)$",
-  "([SU]QADD|[SU]QSUB|SQNEG|SUQADD|USQADD)v(16i8|2i64|4i32|8i16)$",
   "ADDPv(16i8|2i64|4i32|8i16)$")>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "([SU]QADD|[SU]QSUB|SQNEG|SUQADD|USQADD)v(16i8|2i64|4i32|8i16)$")>;
 // ASIMD arith #3
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex  "SADDLv", "UADDLv", "SADDWv",
-  "UADDWv", "SSUBLv", "USUBLv", "SSUBWv", "USUBWv", "ADDHNv", "SUBHNv")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex  "SADDLv", "UADDLv", "SADDWv",
+  "UADDWv", "SSUBLv", "USUBLv", "SSUBWv", "USUBWv")>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex  "ADDHNv", "SUBHNv")>;
 // ASIMD arith #5
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "RADDHNv", "RSUBHNv")>;
+def : InstRW<[CortexA510Write<8, CortexA510UnitVALU>], (instregex "RADDHNv", "RSUBHNv")>;
 // ASIMD arith, reduce
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex  "ADDVv", "SADDLVv", "UADDLVv")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex  "ADDVv")>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex  "SADDLVv", "UADDLVv")>;
 // ASIMD compare #1
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "CM(EQ|GE|GT|HI|HS|LE|LT)v(1i64|2i32|4i16|8i8)")>;
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "CM(EQ|GE|GT|HI|HS|LE|LT)v(2i64|4i32|8i16|16i8)")>;
@@ -437,8 +437,8 @@ def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "(AND|EOR|NOT|
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "(AND|EOR|NOT|ORN)v16i8",
   "(ORR|BIC)v(16i8|4i32|8i16)$", "MVNIv(4i32|4s|8i16)")>;
 // ASIMD max/min, basic
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "[SU](MIN|MAX)P?v(2i32|4i16|8i8)")>;
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "[SU](MIN|MAX)P?v(16i8|4i132|8i16)")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU](MIN|MAX)P?v(2i32|4i16|8i8)")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU](MIN|MAX)P?v(16i8|4i132|8i16)")>;
 // SIMD max/min, reduce
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "[SU](MAX|MIN)Vv")>;
 // ASIMD multiply, by element
@@ -467,12 +467,12 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "[SU]MULLv", "
 // ASIMD polynomial (8x8) multiply long
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instrs PMULLv8i8, PMULLv16i8)>;
 // ASIMD pairwise add and accumulate
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "[SU]ADALPv")>;
+def : InstRW<[CortexA510MCWrite<7, 2, CortexA510UnitVALU>], (instregex "[SU]ADALPv")>;
 // ASIMD shift accumulate
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "[SU]SRA(d|v2i32|v4i16|v8i8)")>;
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "[SU]SRAv(16i8|2i64|4i32|8i16)")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU]SRA(d|v2i32|v4i16|v8i8)")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "[SU]SRAv(16i8|2i64|4i32|8i16)")>;
 // ASIMD shift accumulate #2
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "[SU]RSRA[vd]")>;
+def : InstRW<[CortexA510MCWrite<7, 2, CortexA510UnitVALU>], (instregex "[SU]RSRA[vd]")>;
 // ASIMD shift by immed
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "SHLd$", "SHLv",
   "SLId$", "SRId$", "[SU]SHR[vd]", "SHRNv")>;
@@ -504,7 +504,7 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "[SU]QRSHLv(2i
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^AES[DE]rr$", "^AESI?MCrr")>;
 
 // Crypto polynomial (64x64) multiply long
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instrs PMULLv1i64, PMULLv2i64)>;
+def : InstRW<[CortexA510MCWrite<4, 0, CortexA510UnitVMC>], (instrs PMULLv1i64, PMULLv2i64)>;
 
 // Crypto SHA1 hash acceleration op
 // Crypto SHA1 schedule acceleration ops
@@ -512,25 +512,26 @@ def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^SHA1(H|SU0|S
 
 // Crypto SHA1 hash acceleration ops
 // Crypto SHA256 hash acceleration ops
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instregex "^SHA1[CMP]", "^SHA256H2?")>;
+def : InstRW<[CortexA510MCWrite<4, 0, CortexA510UnitVMC>], (instregex "^SHA1[CMP]", "^SHA256H2?")>;
 
 // Crypto SHA256 schedule acceleration ops
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instregex "^SHA256SU[01]")>;
+def : InstRW<[CortexA510MCWrite<4, 0, CortexA510UnitVMC>], (instregex "^SHA256SU[01]")>;
 
 // Crypto SHA512 hash acceleration ops
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instregex "^SHA512(H|H2|SU0|SU1)")>;
+def : InstRW<[CortexA510MCWrite<9, 0, CortexA510UnitVMC>], (instregex "^SHA512(H|H2|SU0|SU1)")>;
 
 // Crypto SHA3 ops
-def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instrs BCAX, EOR3, XAR)>;
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instrs RAX1)>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instrs BCAX, EOR3)>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instrs XAR)>;
+def : InstRW<[CortexA510MCWrite<9, 0, CortexA510UnitVMC>], (instrs RAX1)>;
 
 
 // Crypto SM3 ops
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instregex "^SM3PARTW[12]$", "^SM3SS1$",
+def : InstRW<[CortexA510MCWrite<9, 0, CortexA510UnitVMC>], (instregex "^SM3PARTW[12]$", "^SM3SS1$",
                                                             "^SM3TT[12][AB]$")>;
 
 // Crypto SM4 ops
-def : InstRW<[CortexA510MCWrite<8, 0, CortexA510UnitVMC>], (instrs SM4E, SM4ENCKEY)>;
+def : InstRW<[CortexA510MCWrite<9, 0, CortexA510UnitVMC>], (instrs SM4E, SM4ENCKEY)>;
 
 // CRC
 // -----------------------------------------------------------------------------
@@ -540,25 +541,25 @@ def : InstRW<[CortexA510MCWrite<2, 0, CortexA510UnitMAC>], (instregex "^CRC32")>
 // SVE Predicate instructions
 
 // Loop control, based on predicate
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs BRKA_PPmP, BRKA_PPzP,
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs BRKA_PPmP, BRKA_PPzP,
                                                   BRKB_PPmP, BRKB_PPzP)>;
 
 // Loop control, based on predicate and flag setting
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs BRKAS_PPzP, BRKBS_PPzP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs BRKAS_PPzP, BRKBS_PPzP)>;
 
 // Loop control, propagating
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs BRKN_PPzP, BRKPA_PPzPP, BRKPB_PPzPP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs BRKN_PPzP, BRKPA_PPzPP, BRKPB_PPzPP)>;
 
 // Loop control, propagating and flag setting
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs BRKNS_PPzP)>;
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs BRKPAS_PPzPP, BRKPBS_PPzPP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs BRKNS_PPzP)>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU0>], (instrs BRKPAS_PPzPP, BRKPBS_PPzPP)>;
 
 
 // Loop control, based on GPR
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>],
              (instregex "^WHILE(GE|GT|HI|HS|LE|LO|LS|LT)_P(WW|XX)_[BHSD]")>;
 
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^WHILE(RW|WR)_PXX_[BHSD]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^WHILE(RW|WR)_PXX_[BHSD]")>;
 
 // Loop terminate
 def : InstRW<[CortexA510Write<1, CortexA510UnitALU>], (instregex "^CTERM(EQ|NE)_(WW|XX)")>;
@@ -569,20 +570,20 @@ def : InstRW<[CortexA510Write<1, CortexA510UnitALU>], (instrs ADDPL_XXI, ADDVL_X
 def : InstRW<[CortexA510Write<1, CortexA510UnitALU>],
              (instregex "^CNT[BHWD]_XPiI")>;
 
-def : InstRW<[CortexA510Write<1, CortexA510UnitALU>],
+def : InstRW<[CortexA510Write<3, CortexA510UnitALU>],
              (instregex "^(INC|DEC)[BHWD]_XPiI")>;
 
-def : InstRW<[CortexA510Write<1, CortexA510UnitALU>],
+def : InstRW<[CortexA510Write<4, CortexA510UnitALU>],
              (instregex "^(SQINC|SQDEC|UQINC|UQDEC)[BHWD]_[XW]Pi(Wd)?I")>;
 
 // Predicate counting scalar, active predicate
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU0>],
              (instregex "^CNTP_XPP_[BHSD]")>;
 
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU0>],
              (instregex "^(DEC|INC)P_XP_[BHSD]")>;
 
-def : InstRW<[CortexA510Write<8, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<9, CortexA510UnitVALU0>],
              (instregex "^(SQDEC|SQINC|UQDEC|UQINC)P_XP_[BHSD]",
                         "^(UQDEC|UQINC)P_WP_[BHSD]",
                         "^(SQDEC|SQINC|UQDEC|UQINC)P_XPWd_[BHSD]")>;
@@ -593,39 +594,39 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>],
              (instregex "^(DEC|INC|SQDEC|SQINC|UQDEC|UQINC)P_ZP_[HSD]")>;
 
 // Predicate logical
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>],
              (instregex "^(AND|BIC|EOR|NAND|NOR|ORN|ORR)_PPzPP")>;
 
 // Predicate logical, flag setting
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>],
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>],
              (instregex "^(ANDS|BICS|EORS|NANDS|NORS|ORNS|ORRS)_PPzPP")>;
 
 // Predicate reverse
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^REV_PP_[BHSD]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^REV_PP_[BHSD]")>;
 
 // Predicate select
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs SEL_PPPP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs SEL_PPPP)>;
 
 // Predicate set
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^PFALSE", "^PTRUE_[BHSD]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^PFALSE", "^PTRUE_[BHSD]")>;
 
 // Predicate set/initialize, set flags
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^PTRUES_[BHSD]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^PTRUES_[BHSD]")>;
 
 // Predicate find first/next
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^PFIRST_B", "^PNEXT_[BHSD]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^PFIRST_B", "^PNEXT_[BHSD]")>;
 
 // Predicate test
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs PTEST_PP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs PTEST_PP)>;
 
 // Predicate transpose
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^TRN[12]_PPP_[BHSDQ]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^TRN[12]_PPP_[BHSDQ]")>;
 
 // Predicate unpack and widen
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instrs PUNPKHI_PP, PUNPKLO_PP)>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instrs PUNPKHI_PP, PUNPKLO_PP)>;
 
 // Predicate zip/unzip
-def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^(ZIP|UZP)[12]_PPP_[BHSDQ]")>;
+def : InstRW<[CortexA510Write<2, CortexA510UnitVALU0>], (instregex "^(ZIP|UZP)[12]_PPP_[BHSDQ]")>;
 
 
 // SVE integer instructions
@@ -634,10 +635,10 @@ def : InstRW<[CortexA510Write<6, CortexA510UnitVALU0>], (instregex "^(ZIP|UZP)[1
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^[SU]ABD_(ZPmZ|ZPZZ)_[BHSD]")>;
 
 // Arithmetic, absolute diff accum
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "^[SU]ABA_ZZZ_[BHSD]")>;
+def : InstRW<[CortexA510MCWrite<6, 2, CortexA510UnitVALU>], (instregex "^[SU]ABA_ZZZ_[BHSD]")>;
 
 // Arithmetic, absolute diff accum long
-def : InstRW<[CortexA510MCWrite<8, 2, CortexA510UnitVALU>], (instregex "^[SU]ABAL[TB]_ZZZ_[HSD]")>;
+def : InstRW<[CortexA510MCWrite<6, 2, CortexA510UnitVALU>], (instregex "^[SU]ABAL[TB]_ZZZ_[HSD]")>;
 
 // Arithmetic, absolute diff long
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^[SU]ABDL[TB]_ZZZ_[HSD]")>;
@@ -651,20 +652,22 @@ def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>],
                         "^(ADD|SUB|SUBR)_ZI_[BHSD]",
                         "^ADR_[SU]XTW_ZZZ_D_[0123]",
                         "^ADR_LSL_ZZZ_[SD]_[0123]",
-                        "^[SU](ADD|SUB)[LW][BT]_ZZZ_[HSD]",
+                        "^[SU]H(ADD|SUB|SUBR)_ZPmZ_[BHSD]")>;
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>],
+             (instregex "^[SU](ADD|SUB)[LW][BT]_ZZZ_[HSD]",
                         "^SADDLBT_ZZZ_[HSD]",
-                        "^[SU]H(ADD|SUB|SUBR)_ZPmZ_[BHSD]",
                         "^SSUBL(BT|TB)_ZZZ_[HSD]")>;
 
 // Arithmetic, complex
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>],
-             (instregex "^R?(ADD|SUB)HN[BT]_ZZZ_[BHS]",
-                        "^SQ(ABS|NEG)_ZPmZ_[BHSD]",
+             (instregex "^SQ(ABS|NEG)_ZPmZ_[BHSD]",
                         "^SQ(ADD|SUB|SUBR)_ZPmZ_?[BHSD]",
                         "^[SU]Q(ADD|SUB)_ZZZ_[BHSD]",
                         "^[SU]Q(ADD|SUB)_ZI_[BHSD]",
                         "^(SRH|SUQ|UQ|USQ|URH)ADD_ZPmZ_[BHSD]",
                         "^(UQSUB|UQSUBR)_ZPmZ_[BHSD]")>;
+def : InstRW<[CortexA510Write<8, CortexA510UnitVALU>],
+             (instregex "^R?(ADD|SUB)HN[BT]_ZZZ_[BHS]")>;
 
 // Arithmetic, large integer
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^(AD|SB)CL[BT]_ZZZ_[SD]")>;
@@ -735,14 +738,14 @@ def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^(BSL|BSL1N|B
 
 // Count/reverse bits
 def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^(CLS|CLZ|RBIT)_ZPmZ_[BHSD]")>;
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^CNT_ZPmZ_[BH]")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^CNT_ZPmZ_[BH]")>;
 def : InstRW<[CortexA510Write<8, CortexA510UnitVALU>], (instregex "^CNT_ZPmZ_S")>;
 def : InstRW<[CortexA510Write<12, CortexA510UnitVALU>], (instregex "^CNT_ZPmZ_D")>;
 // Broadcast logical bitmask immediate to vector
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instrs DUPM_ZI)>;
 
 // Compare and set flags
-def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>],
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>],
              (instregex "^CMP(EQ|GE|GT|HI|HS|LE|LO|LS|LT|NE)_PPzZ[IZ]_[BHSD]",
                         "^CMP(EQ|GE|GT|HI|HS|LE|LO|LS|LT|NE)_WIDE_PPzZZ_[BHS]")>;
 
@@ -939,12 +942,14 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVMAC>], (instregex "^SQRDMULH_ZZZ
 // Multiply/multiply long, (8x8) polynomial
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^PMUL_ZZZ_B")>;
 
-def : InstRW<[CortexA510Write<6, CortexA510UnitVMC>], (instregex "^PMULL[BT]_ZZZ_[HDQ]")>;
+def : InstRW<[CortexA510Write<9, CortexA510UnitVMC>], (instregex "^PMULL[BT]_ZZZ_[HDQ]")>;
 
 
 // Predicate counting vector
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>],
+             (instregex "^(DEC|INC)[HWD]_ZPiI")>;
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>],
-             (instregex "^(DEC|INC|SQDEC|SQINC|UQDEC|UQINC)[HWD]_ZPiI")>;
+             (instregex "^(SQDEC|SQINC|UQDEC|UQINC)[HWD]_ZPiI")>;
 
 // Reciprocal estimate
 def : InstRW<[CortexA510Write<4, CortexA510UnitVMAC>], (instregex "^URECPE_ZPmZ_S", "^URSQRTE_ZPmZ_S")>;
@@ -965,7 +970,7 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVALU0>], (instregex "^[SU](ADD|MA
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU0>], (instregex "^(ANDV|EORV|ORV)_VPZ_[BHSD]")>;
 
 // Reverse, vector
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^REV_ZZ_[BHSD]",
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^REV_ZZ_[BHSD]",
                                            "^REVB_ZPmZ_[HSD]",
                                            "^REVH_ZPmZ_[SD]",
                                            "^REVW_ZPmZ_D")>;
@@ -980,13 +985,13 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^TBL_ZZZZ?_[B
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^TBX_ZZZ_[BHSD]")>;
 
 // Transpose, vector form
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^TRN[12]_ZZZ_[BHSDQ]")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^TRN[12]_ZZZ_[BHSDQ]")>;
 
 // Unpack and extend
 def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^[SU]UNPK(HI|LO)_ZZ_[HSD]")>;
 
 // Zip/unzip
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^(UZP|ZIP)[12]_ZZZ_[BHSDQ]")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^(UZP|ZIP)[12]_ZZZ_[BHSDQ]")>;
 
 // SVE floating-point instructions
 // -----------------------------------------------------------------------------
@@ -1142,7 +1147,7 @@ def : InstRW<[CortexA510Write<4, CortexA510UnitVMAC>], (instregex "^FTMAD_ZZI_[H
 
 // Floating point trigonometric, miscellaneous
 def : InstRW<[CortexA510Write<4, CortexA510UnitVMAC>], (instregex "^FTSMUL_ZZZ_[HSD]")>;
-def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^FTSSEL_ZZZ_[HSD]")>;
+def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^FTSSEL_ZZZ_[HSD]")>;
 
 
 // SVE BFloat16 (BF16) instructions
@@ -1251,12 +1256,12 @@ def : InstRW<[CortexA510MCWrite<7, 7, CortexA510UnitLdSt>],
                         "^GLD(FF)?1D(_SCALED)?$")>;
 
 // Gather load, 32-bit scaled offset
-def : InstRW<[CortexA510MCWrite<9, 9, CortexA510UnitLd>],
+def : InstRW<[CortexA510MCWrite<7, 7, CortexA510UnitLd>],
              (instregex "^GLD(FF)?1S?[HW]_S_[SU]XTW_SCALED$",
                         "^GLD(FF)?1W_[SU]XTW_SCALED")>;
 
 // Gather load, 32-bit unpacked unscaled offset
-def : InstRW<[CortexA510MCWrite<9, 9, CortexA510UnitLd>], (instregex "^GLD(FF)?1S?[BH]_S_[SU]XTW$",
+def : InstRW<[CortexA510MCWrite<7, 7, CortexA510UnitLd>], (instregex "^GLD(FF)?1S?[BH]_S_[SU]XTW$",
                                               "^GLD(FF)?1W_[SU]XTW$")>;
 
 def : InstRW<[CortexA510Write<0, CortexA510UnitVALU>], (instregex "^PRF(B|H|W|D).*")>;
@@ -1377,12 +1382,12 @@ def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^AES[DE]_ZZZ_
                                            "^AESI?MC_ZZ_B$")>;
 
 // Crypto SHA3 ops
-def : InstRW<[CortexA510Write<3, CortexA510UnitVALU>], (instregex "^(BCAX|EOR3)_ZZZZ$",
+def : InstRW<[CortexA510Write<4, CortexA510UnitVALU>], (instregex "^(BCAX|EOR3)_ZZZZ$",
                                             "^XAR_ZZZI_[BHSD]$")>;
 
-def : InstRW<[CortexA510MC_RC0Write<8, CortexA510UnitVMC>], (instregex "^RAX1_ZZZ_D$")>;
+def : InstRW<[CortexA510MC_RC0Write<9, CortexA510UnitVMC>], (instregex "^RAX1_ZZZ_D$")>;
 
 // Crypto SM4 ops
-def : InstRW<[CortexA510MC_RC0Write<8, CortexA510UnitVMC>], (instregex "^SM4E(KEY)?_ZZZ_S$")>;
+def : InstRW<[CortexA510MC_RC0Write<9, CortexA510UnitVMC>], (instregex "^SM4E(KEY)?_ZZZ_S$")>;
 
 }
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/combine-udiv.ll b/llvm/test/CodeGen/AArch64/GlobalISel/combine-udiv.ll
index ceef0c49a45ec..9a525151ca328 100644
--- a/llvm/test/CodeGen/AArch64/GlobalISel/combine-udiv.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/combine-udiv.ll
@@ -140,10 +140,10 @@ define <8 x i16> @combine_vec_udiv_nonuniform3(<8 x i16> %x) {
 ; GISEL-NEXT:    umull2 v2.4s, v0.8h, v1.8h
 ; GISEL-NEXT:    umull v1.4s, v0.4h, v1.4h
 ; GISEL-NEXT:    uzp2 v1.8h, v1.8h, v2.8h
+; GISEL-NEXT:    ldr q2, [x8, :lo12:.LCPI3_0]
 ; GISEL-NEXT:    sub v0.8h, v0.8h, v1.8h
 ; GISEL-NEXT:    usra v1.8h, v0.8h, #1
-; GISEL-NEXT:    ldr q0, [x8, :lo12:.LCPI3_0]
-; GISEL-NEXT:    neg v0.8h, v0.8h
+; GISEL-NEXT:    neg v0.8h, v2.8h
 ; GISEL-NEXT:    ushl v0.8h, v1.8h, v0.8h
 ; GISEL-NEXT:    ret
   %1 = udiv <8 x i16> %x, <i16 7, i16 23, i16 25, i16 27, i16 31, i16 47, i16 63, i16 127>
@@ -170,13 +170,13 @@ define <16 x i8> @combine_vec_udiv_nonuniform4(<16 x i8> %x) {
 ; GISEL-LABEL: combine_vec_udiv_nonuniform4:
 ; GISEL:       // %bb.0:
 ; GISEL-NEXT:    adrp x8, .LCPI4_2
+; GISEL-NEXT:    adrp x9, .LCPI4_0
 ; GISEL-NEXT:    ldr q1, [x8, :lo12:.LCPI4_2]
 ; GISEL-NEXT:    adrp x8, .LCPI4_1
+; GISEL-NEXT:    ldr q4, [x9, :lo12:.LCPI4_0]
 ; GISEL-NEXT:    ldr q3, [x8, :lo12:.LCPI4_1]
-; GISEL-NEXT:    adrp x8, .LCPI4_0
 ; GISEL-NEXT:    umull2 v2.8h, v0.16b, v1.16b
 ; GISEL-NEXT:    umull v1.8h, v0.8b, v1.8b
-; GISEL-NEXT:    ldr q4, [x8, :lo12:.LCPI4_0]
 ; GISEL-NEXT:    uzp2 v1.16b, v1.16b, v2.16b
 ; GISEL-NEXT:    neg v2.16b, v3.16b
 ; GISEL-NEXT:    shl v3.16b, v4.16b, #7
diff --git a/llvm/test/CodeGen/AArch64/aarch64-addv.ll b/llvm/test/CodeGen/AArch64/aarch64-addv.ll
index b77d591347659..ee035ec1941d5 100644
--- a/llvm/test/CodeGen/AArch64/aarch64-addv.ll
+++ b/llvm/test/CodeGen/AArch64/aarch64-addv.ll
@@ -101,12 +101,12 @@ define i32 @oversized_ADDV_256(ptr noalias nocapture readonly %arg1, ptr noalias
 ; GISEL-NEXT:    ushll v2.8h, v2.8b, #0
 ; GISEL-NEXT:    usubl v3.4s, v1.4h, v2.4h
 ; GISEL-NEXT:    usubl2 v1.4s, v1.8h, v2.8h
-; GISEL-NEXT:    neg v2.4s, v3.4s
-; GISEL-NEXT:    neg v4.4s, v1.4s
-; GISEL-NEXT:    cmgt v5.4s, v0.4s, v3.4s
+; GISEL-NEXT:    cmgt v2.4s, v0.4s, v3.4s
 ; GISEL-NEXT:    cmgt v0.4s, v0.4s, v1.4s
-; GISEL-NEXT:    bif v2.16b, v3.16b, v5.16b
-; GISEL-NEXT:    bsl v0.16b, v4.16b, v1.16b
+; GISEL-NEXT:    neg v4.4s, v3.4s
+; GISEL-NEXT:    neg v5.4s, v1.4s
+; GISEL-NEXT:    bsl v2.16b, v4.16b, v3.16b
+; GISEL-NEXT:    bsl v0.16b, v5.16b, v1.16b
 ; GISEL-NEXT:    add v0.4s, v2.4s, v0.4s
 ; GISEL-NEXT:    addv s0, v0.4s
 ; GISEL-NEXT:    fmov w0, s0
diff --git a/llvm/test/CodeGen/AArch64/aarch64-dup-ext-scalable.ll b/llvm/test/CodeGen/AArch64/aarch64-dup-ext-scalable.ll
index fdeae9f326ad8..36b81d8e495ce 100644
--- a/llvm/test/CodeGen/AArch64/aarch64-dup-ext-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/aarch64-dup-ext-scalable.ll
@@ -4,8 +4,8 @@
 define <vscale x 2 x i16> @dupsext_v2i8_v2i16(i8 %src, <vscale x 2 x i16> %b) {
 ; CHECK-LABEL: dupsext_v2i8_v2i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sxtb w8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -20,8 +20,8 @@ entry:
 define <vscale x 4 x i16> @dupsext_v4i8_v4i16(i8 %src, <vscale x 4 x i16> %b) {
 ; CHECK-LABEL: dupsext_v4i8_v4i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sxtb w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -36,8 +36,8 @@ entry:
 define <vscale x 8 x i16> @dupsext_v8i8_v8i16(i8 %src, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: dupsext_v8i8_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    sxtb w8, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -52,8 +52,8 @@ entry:
 define <vscale x 2 x i32> @dupsext_v2i8_v2i32(i8 %src, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: dupsext_v2i8_v2i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sxtb w8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -68,8 +68,8 @@ entry:
 define <vscale x 4 x i32> @dupsext_v4i8_v4i32(i8 %src, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: dupsext_v4i8_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sxtb w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -84,9 +84,9 @@ entry:
 define <vscale x 2 x i64> @dupsext_v2i8_v2i64(i8 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupsext_v2i8_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    sxtb x8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -101,8 +101,8 @@ entry:
 define <vscale x 2 x i32> @dupsext_v2i16_v2i32(i16 %src, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: dupsext_v2i16_v2i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sxth w8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -117,8 +117,8 @@ entry:
 define <vscale x 4 x i32> @dupsext_v4i16_v4i32(i16 %src, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: dupsext_v4i16_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sxth w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -133,9 +133,9 @@ entry:
 define <vscale x 2 x i64> @dupsext_v2i16_v2i64(i16 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupsext_v2i16_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    sxth x8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -150,9 +150,9 @@ entry:
 define <vscale x 2 x i64> @dupsext_v2i32_v2i64(i32 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupsext_v2i32_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    sxtw x8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -167,8 +167,8 @@ entry:
 define <vscale x 2 x i16> @dupzext_v2i8_v2i16(i8 %src, <vscale x 2 x i16> %b) {
 ; CHECK-LABEL: dupzext_v2i8_v2i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and w8, w0, #0xff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -183,8 +183,8 @@ entry:
 define <vscale x 4 x i16> @dupzext_v4i8_v4i16(i8 %src, <vscale x 4 x i16> %b) {
 ; CHECK-LABEL: dupzext_v4i8_v4i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0xff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -199,8 +199,8 @@ entry:
 define <vscale x 8 x i16> @dupzext_v8i8_v8i16(i8 %src, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: dupzext_v8i8_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and w8, w0, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -215,8 +215,8 @@ entry:
 define <vscale x 2 x i32> @dupzext_v2i8_v2i32(i8 %src, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: dupzext_v2i8_v2i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and w8, w0, #0xff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -231,8 +231,8 @@ entry:
 define <vscale x 4 x i32> @dupzext_v4i8_v4i32(i8 %src, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: dupzext_v4i8_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0xff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -247,9 +247,9 @@ entry:
 define <vscale x 2 x i64> @dupzext_v2i8_v2i64(i8 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupzext_v2i8_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0xff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -264,8 +264,8 @@ entry:
 define <vscale x 2 x i32> @dupzext_v2i16_v2i32(i16 %src, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: dupzext_v2i16_v2i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and w8, w0, #0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -280,8 +280,8 @@ entry:
 define <vscale x 4 x i32> @dupzext_v4i16_v4i32(i16 %src, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: dupzext_v4i16_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0xffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -296,9 +296,9 @@ entry:
 define <vscale x 2 x i64> @dupzext_v2i16_v2i64(i16 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupzext_v2i16_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -313,8 +313,8 @@ entry:
 define <vscale x 2 x i64> @dupzext_v2i32_v2i64(i32 %src, <vscale x 2 x i64> %b) {
 ; CHECK-LABEL: dupzext_v2i32_v2i64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/aarch64-smull.ll b/llvm/test/CodeGen/AArch64/aarch64-smull.ll
index 61a4f64ac2bfc..540471a05901a 100644
--- a/llvm/test/CodeGen/AArch64/aarch64-smull.ll
+++ b/llvm/test/CodeGen/AArch64/aarch64-smull.ll
@@ -257,8 +257,8 @@ define <2 x i64> @smull_zext_v2i32_v2i64(ptr %A, ptr %B) nounwind {
 ; CHECK-SVE-LABEL: smull_zext_v2i32_v2i64:
 ; CHECK-SVE:       // %bb.0:
 ; CHECK-SVE-NEXT:    ldrh w8, [x0]
-; CHECK-SVE-NEXT:    ptrue p0.d, vl2
 ; CHECK-SVE-NEXT:    ldrh w9, [x0, #2]
+; CHECK-SVE-NEXT:    ptrue p0.d, vl2
 ; CHECK-SVE-NEXT:    ldr d0, [x1]
 ; CHECK-SVE-NEXT:    fmov d1, x8
 ; CHECK-SVE-NEXT:    sshll v0.2d, v0.2s, #0
diff --git a/llvm/test/CodeGen/AArch64/active_lane_mask.ll b/llvm/test/CodeGen/AArch64/active_lane_mask.ll
index a65c5d6667794..43122c8c953fc 100644
--- a/llvm/test/CodeGen/AArch64/active_lane_mask.ll
+++ b/llvm/test/CodeGen/AArch64/active_lane_mask.ll
@@ -131,9 +131,9 @@ define <vscale x 2 x i1> @lane_mask_nxv2i1_i8(i8 %index, i8 %TC) {
 ; CHECK-NEXT:    index z0.d, #0, #1
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0xff
+; CHECK-NEXT:    // kill: def $w1 killed $w1 def $x1
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
-; CHECK-NEXT:    // kill: def $w1 killed $w1 def $x1
 ; CHECK-NEXT:    and x8, x1, #0xff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xff
 ; CHECK-NEXT:    add z0.d, z0.d, z1.d
@@ -153,6 +153,7 @@ define <vscale x 32 x i1> @lane_mask_nxv32i1_i32(i32 %index, i32 %TC) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
@@ -160,16 +161,16 @@ define <vscale x 32 x i1> @lane_mask_nxv32i1_i32(i32 %index, i32 %TC) {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    index z0.s, #0, #1
 ; CHECK-NEXT:    mov z1.s, w0
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z25.s, w1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.d, z0.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    uqadd z6.s, z0.s, z1.s
 ; CHECK-NEXT:    incw z0.s, all, mul #4
 ; CHECK-NEXT:    incw z2.s
 ; CHECK-NEXT:    incw z3.s, all, mul #2
-; CHECK-NEXT:    cmphi p2.s, p0/z, z25.s, z6.s
 ; CHECK-NEXT:    uqadd z0.s, z0.s, z1.s
+; CHECK-NEXT:    cmphi p2.s, p0/z, z25.s, z6.s
 ; CHECK-NEXT:    mov z4.d, z2.d
 ; CHECK-NEXT:    uqadd z5.s, z2.s, z1.s
 ; CHECK-NEXT:    uqadd z7.s, z3.s, z1.s
@@ -177,25 +178,26 @@ define <vscale x 32 x i1> @lane_mask_nxv32i1_i32(i32 %index, i32 %TC) {
 ; CHECK-NEXT:    incw z3.s, all, mul #4
 ; CHECK-NEXT:    cmphi p5.s, p0/z, z25.s, z0.s
 ; CHECK-NEXT:    incw z4.s, all, mul #2
-; CHECK-NEXT:    cmphi p1.s, p0/z, z25.s, z5.s
-; CHECK-NEXT:    cmphi p3.s, p0/z, z25.s, z7.s
 ; CHECK-NEXT:    uqadd z2.s, z2.s, z1.s
 ; CHECK-NEXT:    uqadd z3.s, z3.s, z1.s
+; CHECK-NEXT:    cmphi p1.s, p0/z, z25.s, z5.s
+; CHECK-NEXT:    cmphi p3.s, p0/z, z25.s, z7.s
 ; CHECK-NEXT:    uqadd z24.s, z4.s, z1.s
 ; CHECK-NEXT:    incw z4.s, all, mul #4
-; CHECK-NEXT:    uzp1 p1.h, p2.h, p1.h
 ; CHECK-NEXT:    cmphi p6.s, p0/z, z25.s, z2.s
-; CHECK-NEXT:    cmphi p2.s, p0/z, z25.s, z3.s
+; CHECK-NEXT:    cmphi p7.s, p0/z, z25.s, z3.s
+; CHECK-NEXT:    uzp1 p1.h, p2.h, p1.h
 ; CHECK-NEXT:    uqadd z1.s, z4.s, z1.s
 ; CHECK-NEXT:    cmphi p4.s, p0/z, z25.s, z24.s
-; CHECK-NEXT:    uzp1 p3.h, p3.h, p4.h
 ; CHECK-NEXT:    cmphi p0.s, p0/z, z25.s, z1.s
-; CHECK-NEXT:    uzp1 p4.h, p5.h, p6.h
+; CHECK-NEXT:    uzp1 p2.h, p3.h, p4.h
+; CHECK-NEXT:    uzp1 p3.h, p5.h, p6.h
 ; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p2.h, p2.h, p0.h
-; CHECK-NEXT:    uzp1 p0.b, p1.b, p3.b
-; CHECK-NEXT:    uzp1 p1.b, p4.b, p2.b
+; CHECK-NEXT:    uzp1 p4.h, p7.h, p0.h
+; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p0.b, p1.b, p2.b
+; CHECK-NEXT:    uzp1 p1.b, p3.b, p4.b
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -208,96 +210,97 @@ define <vscale x 32 x i1> @lane_mask_nxv32i1_i64(i64 %index, i64 %TC) {
 ; CHECK-LABEL: lane_mask_nxv32i1_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
-; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p10, [sp, #1, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    str p9, [sp, #2, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p8, [sp, #3, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    str z8, [sp, #1, mul vl] // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    index z1.d, #0, #1
+; CHECK-NEXT:    .cfi_escape 0x10, 0x48, 0x0a, 0x11, 0x70, 0x22, 0x11, 0x78, 0x92, 0x2e, 0x00, 0x1e, 0x22 // $d8 @ cfa - 16 - 8 * VG
+; CHECK-NEXT:    index z5.d, #0, #1
 ; CHECK-NEXT:    mov z0.d, x0
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z3.d, x1
-; CHECK-NEXT:    mov z2.d, z1.d
-; CHECK-NEXT:    mov z4.d, z1.d
-; CHECK-NEXT:    mov z6.d, z1.d
-; CHECK-NEXT:    uqadd z25.d, z1.d, z0.d
-; CHECK-NEXT:    incd z1.d, all, mul #8
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z2.d, z5.d
+; CHECK-NEXT:    mov z1.d, z5.d
+; CHECK-NEXT:    mov z4.d, z5.d
+; CHECK-NEXT:    uqadd z25.d, z5.d, z0.d
+; CHECK-NEXT:    incd z5.d, all, mul #8
 ; CHECK-NEXT:    incd z2.d
-; CHECK-NEXT:    incd z4.d, all, mul #2
-; CHECK-NEXT:    incd z6.d, all, mul #4
-; CHECK-NEXT:    cmphi p1.d, p0/z, z3.d, z25.d
-; CHECK-NEXT:    uqadd z1.d, z1.d, z0.d
-; CHECK-NEXT:    mov z5.d, z2.d
-; CHECK-NEXT:    uqadd z26.d, z2.d, z0.d
+; CHECK-NEXT:    incd z1.d, all, mul #2
+; CHECK-NEXT:    incd z4.d, all, mul #4
+; CHECK-NEXT:    uqadd z5.d, z5.d, z0.d
+; CHECK-NEXT:    cmphi p3.d, p0/z, z3.d, z25.d
+; CHECK-NEXT:    mov z6.d, z2.d
 ; CHECK-NEXT:    mov z7.d, z2.d
-; CHECK-NEXT:    mov z24.d, z4.d
-; CHECK-NEXT:    uqadd z27.d, z4.d, z0.d
-; CHECK-NEXT:    uqadd z28.d, z6.d, z0.d
+; CHECK-NEXT:    mov z24.d, z1.d
+; CHECK-NEXT:    uqadd z26.d, z2.d, z0.d
+; CHECK-NEXT:    uqadd z27.d, z1.d, z0.d
+; CHECK-NEXT:    uqadd z28.d, z4.d, z0.d
 ; CHECK-NEXT:    incd z2.d, all, mul #8
+; CHECK-NEXT:    incd z1.d, all, mul #8
 ; CHECK-NEXT:    incd z4.d, all, mul #8
-; CHECK-NEXT:    incd z6.d, all, mul #8
-; CHECK-NEXT:    incd z5.d, all, mul #2
+; CHECK-NEXT:    incd z6.d, all, mul #2
 ; CHECK-NEXT:    incd z7.d, all, mul #4
-; CHECK-NEXT:    cmphi p2.d, p0/z, z3.d, z26.d
 ; CHECK-NEXT:    incd z24.d, all, mul #4
-; CHECK-NEXT:    cmphi p3.d, p0/z, z3.d, z27.d
-; CHECK-NEXT:    cmphi p5.d, p0/z, z3.d, z28.d
+; CHECK-NEXT:    cmphi p4.d, p0/z, z3.d, z26.d
+; CHECK-NEXT:    cmphi p2.d, p0/z, z3.d, z27.d
+; CHECK-NEXT:    cmphi p1.d, p0/z, z3.d, z28.d
+; CHECK-NEXT:    mov z31.d, z6.d
+; CHECK-NEXT:    uqadd z29.d, z6.d, z0.d
+; CHECK-NEXT:    uqadd z30.d, z7.d, z0.d
+; CHECK-NEXT:    uqadd z8.d, z24.d, z0.d
+; CHECK-NEXT:    incd z6.d, all, mul #8
+; CHECK-NEXT:    incd z7.d, all, mul #8
+; CHECK-NEXT:    incd z24.d, all, mul #8
 ; CHECK-NEXT:    uqadd z2.d, z2.d, z0.d
+; CHECK-NEXT:    uqadd z1.d, z1.d, z0.d
+; CHECK-NEXT:    incd z31.d, all, mul #4
 ; CHECK-NEXT:    uqadd z4.d, z4.d, z0.d
+; CHECK-NEXT:    uzp1 p3.s, p3.s, p4.s
+; CHECK-NEXT:    cmphi p5.d, p0/z, z3.d, z29.d
+; CHECK-NEXT:    cmphi p7.d, p0/z, z3.d, z30.d
 ; CHECK-NEXT:    uqadd z6.d, z6.d, z0.d
-; CHECK-NEXT:    mov z26.d, z5.d
-; CHECK-NEXT:    uqadd z25.d, z5.d, z0.d
-; CHECK-NEXT:    uqadd z27.d, z7.d, z0.d
-; CHECK-NEXT:    incd z5.d, all, mul #8
-; CHECK-NEXT:    incd z7.d, all, mul #8
-; CHECK-NEXT:    uzp1 p1.s, p1.s, p2.s
-; CHECK-NEXT:    incd z26.d, all, mul #4
-; CHECK-NEXT:    cmphi p8.d, p0/z, z3.d, z2.d
-; CHECK-NEXT:    cmphi p4.d, p0/z, z3.d, z25.d
-; CHECK-NEXT:    uqadd z25.d, z24.d, z0.d
-; CHECK-NEXT:    incd z24.d, all, mul #8
-; CHECK-NEXT:    uqadd z5.d, z5.d, z0.d
+; CHECK-NEXT:    cmphi p6.d, p0/z, z3.d, z8.d
+; CHECK-NEXT:    ldr z8, [sp, #1, mul vl] // 16-byte Folded Reload
 ; CHECK-NEXT:    uqadd z7.d, z7.d, z0.d
-; CHECK-NEXT:    cmphi p6.d, p0/z, z3.d, z27.d
-; CHECK-NEXT:    uqadd z28.d, z26.d, z0.d
-; CHECK-NEXT:    incd z26.d, all, mul #8
-; CHECK-NEXT:    uzp1 p3.s, p3.s, p4.s
+; CHECK-NEXT:    uqadd z25.d, z31.d, z0.d
+; CHECK-NEXT:    incd z31.d, all, mul #8
 ; CHECK-NEXT:    uqadd z24.d, z24.d, z0.d
-; CHECK-NEXT:    cmphi p7.d, p0/z, z3.d, z25.d
-; CHECK-NEXT:    cmphi p4.d, p0/z, z3.d, z1.d
-; CHECK-NEXT:    uzp1 p5.s, p5.s, p6.s
-; CHECK-NEXT:    cmphi p6.d, p0/z, z3.d, z4.d
-; CHECK-NEXT:    cmphi p9.d, p0/z, z3.d, z5.d
-; CHECK-NEXT:    cmphi p10.d, p0/z, z3.d, z7.d
-; CHECK-NEXT:    uqadd z0.d, z26.d, z0.d
-; CHECK-NEXT:    cmphi p2.d, p0/z, z3.d, z28.d
-; CHECK-NEXT:    uzp1 p4.s, p4.s, p8.s
-; CHECK-NEXT:    cmphi p8.d, p0/z, z3.d, z24.d
-; CHECK-NEXT:    uzp1 p6.s, p6.s, p9.s
+; CHECK-NEXT:    cmphi p4.d, p0/z, z3.d, z5.d
+; CHECK-NEXT:    uzp1 p2.s, p2.s, p5.s
+; CHECK-NEXT:    cmphi p5.d, p0/z, z3.d, z2.d
+; CHECK-NEXT:    cmphi p9.d, p0/z, z3.d, z6.d
+; CHECK-NEXT:    uqadd z0.d, z31.d, z0.d
+; CHECK-NEXT:    uzp1 p1.s, p1.s, p7.s
+; CHECK-NEXT:    cmphi p7.d, p0/z, z3.d, z1.d
+; CHECK-NEXT:    cmphi p8.d, p0/z, z3.d, z25.d
+; CHECK-NEXT:    uzp1 p2.h, p3.h, p2.h
+; CHECK-NEXT:    cmphi p3.d, p0/z, z3.d, z7.d
+; CHECK-NEXT:    uzp1 p4.s, p4.s, p5.s
+; CHECK-NEXT:    uzp1 p5.s, p7.s, p9.s
 ; CHECK-NEXT:    ldr p9, [sp, #2, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p1.h, p1.h, p3.h
-; CHECK-NEXT:    uzp1 p2.s, p7.s, p2.s
-; CHECK-NEXT:    cmphi p7.d, p0/z, z3.d, z6.d
+; CHECK-NEXT:    uzp1 p6.s, p6.s, p8.s
+; CHECK-NEXT:    cmphi p8.d, p0/z, z3.d, z4.d
+; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p4.h, p4.h, p5.h
+; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p1.h, p1.h, p6.h
+; CHECK-NEXT:    cmphi p6.d, p0/z, z3.d, z24.d
 ; CHECK-NEXT:    cmphi p0.d, p0/z, z3.d, z0.d
-; CHECK-NEXT:    uzp1 p7.s, p7.s, p10.s
-; CHECK-NEXT:    ldr p10, [sp, #1, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p0.s, p8.s, p0.s
+; CHECK-NEXT:    uzp1 p3.s, p8.s, p3.s
 ; CHECK-NEXT:    ldr p8, [sp, #3, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p3.h, p4.h, p6.h
+; CHECK-NEXT:    uzp1 p0.s, p6.s, p0.s
 ; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p2.h, p5.h, p2.h
-; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p4.h, p7.h, p0.h
-; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p0.b, p1.b, p2.b
-; CHECK-NEXT:    uzp1 p1.b, p3.b, p4.b
+; CHECK-NEXT:    uzp1 p3.h, p3.h, p0.h
+; CHECK-NEXT:    uzp1 p0.b, p2.b, p1.b
+; CHECK-NEXT:    uzp1 p1.b, p4.b, p3.b
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %active.lane.mask = call <vscale x 32 x i1> @llvm.get.active.lane.mask.nxv32i1.i64(i64 %index, i64 %TC)
@@ -459,12 +462,12 @@ define <4 x i1> @lane_mask_v4i1_i8(i8 %index, i8 %TC) {
 ; CHECK-NEXT:    adrp x8, .LCPI26_0
 ; CHECK-NEXT:    movi d2, #0xff00ff00ff00ff
 ; CHECK-NEXT:    ldr d1, [x8, :lo12:.LCPI26_0]
+; CHECK-NEXT:    dup v3.4h, w1
 ; CHECK-NEXT:    bic v0.4h, #255, lsl #8
+; CHECK-NEXT:    bic v3.4h, #255, lsl #8
 ; CHECK-NEXT:    add v0.4h, v0.4h, v1.4h
-; CHECK-NEXT:    dup v1.4h, w1
 ; CHECK-NEXT:    umin v0.4h, v0.4h, v2.4h
-; CHECK-NEXT:    bic v1.4h, #255, lsl #8
-; CHECK-NEXT:    cmhi v0.4h, v1.4h, v0.4h
+; CHECK-NEXT:    cmhi v0.4h, v3.4h, v0.4h
 ; CHECK-NEXT:    ret
   %active.lane.mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i8(i8 %index, i8 %TC)
   ret <4 x i1> %active.lane.mask
@@ -480,9 +483,9 @@ define <2 x i1> @lane_mask_v2i1_i8(i8 %index, i8 %TC) {
 ; CHECK-NEXT:    dup v3.2s, w1
 ; CHECK-NEXT:    and v1.8b, v1.8b, v0.8b
 ; CHECK-NEXT:    add v1.2s, v1.2s, v2.2s
-; CHECK-NEXT:    umin v1.2s, v1.2s, v0.2s
-; CHECK-NEXT:    and v0.8b, v3.8b, v0.8b
-; CHECK-NEXT:    cmhi v0.2s, v0.2s, v1.2s
+; CHECK-NEXT:    and v2.8b, v3.8b, v0.8b
+; CHECK-NEXT:    umin v0.2s, v1.2s, v0.2s
+; CHECK-NEXT:    cmhi v0.2s, v2.2s, v0.2s
 ; CHECK-NEXT:    ret
   %active.lane.mask = call <2 x i1> @llvm.get.active.lane.mask.v2i1.i8(i8 %index, i8 %TC)
   ret <2 x i1> %active.lane.mask
diff --git a/llvm/test/CodeGen/AArch64/arm64-convert-v4f64.ll b/llvm/test/CodeGen/AArch64/arm64-convert-v4f64.ll
index 3007e7ce771e6..508f68d6f14d4 100644
--- a/llvm/test/CodeGen/AArch64/arm64-convert-v4f64.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-convert-v4f64.ll
@@ -54,19 +54,19 @@ define <4 x bfloat> @uitofp_v4i64_to_v4bf16(ptr %ptr) {
 ; CHECK-LABEL: uitofp_v4i64_to_v4bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ldp q0, q1, [x0]
-; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    ucvtf v0.2d, v0.2d
 ; CHECK-NEXT:    ucvtf v1.2d, v1.2d
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    fcmeq v2.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    bit v0.16b, v1.16b, v2.16b
+; CHECK-NEXT:    add v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
 ; CHECK-NEXT:    shrn v0.4h, v0.4s, #16
 ; CHECK-NEXT:    ret
   %tmp1 = load <4 x i64>, ptr %ptr
diff --git a/llvm/test/CodeGen/AArch64/arm64-vabs.ll b/llvm/test/CodeGen/AArch64/arm64-vabs.ll
index fe4da2e7cf36b..5b45ba2552cef 100644
--- a/llvm/test/CodeGen/AArch64/arm64-vabs.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-vabs.ll
@@ -257,12 +257,12 @@ define i16 @uabd16b_rdx(ptr %a, ptr %b) {
 ; CHECK-GI-NEXT:    movi.2d v0, #0000000000000000
 ; CHECK-GI-NEXT:    usubl.8h v3, v1, v2
 ; CHECK-GI-NEXT:    usubl2.8h v1, v1, v2
-; CHECK-GI-NEXT:    neg.8h v2, v3
-; CHECK-GI-NEXT:    neg.8h v4, v1
-; CHECK-GI-NEXT:    cmgt.8h v5, v0, v3
+; CHECK-GI-NEXT:    cmgt.8h v2, v0, v3
 ; CHECK-GI-NEXT:    cmgt.8h v0, v0, v1
-; CHECK-GI-NEXT:    bif.16b v2, v3, v5
-; CHECK-GI-NEXT:    bsl.16b v0, v4, v1
+; CHECK-GI-NEXT:    neg.8h v4, v3
+; CHECK-GI-NEXT:    neg.8h v5, v1
+; CHECK-GI-NEXT:    bsl.16b v2, v4, v3
+; CHECK-GI-NEXT:    bsl.16b v0, v5, v1
 ; CHECK-GI-NEXT:    add.8h v0, v2, v0
 ; CHECK-GI-NEXT:    addv.8h h0, v0
 ; CHECK-GI-NEXT:    fmov w0, s0
@@ -299,18 +299,18 @@ define i32 @uabd16b_rdx_i32(<16 x i8> %a, <16 x i8> %b) {
 ; CHECK-GI-NEXT:    usubl2.4s v3, v3, v4
 ; CHECK-GI-NEXT:    usubl.4s v4, v0, v1
 ; CHECK-GI-NEXT:    usubl2.4s v0, v0, v1
-; CHECK-GI-NEXT:    neg.4s v6, v5
-; CHECK-GI-NEXT:    neg.4s v7, v3
 ; CHECK-GI-NEXT:    cmgt.4s v1, v2, v5
-; CHECK-GI-NEXT:    neg.4s v16, v4
-; CHECK-GI-NEXT:    neg.4s v17, v0
-; CHECK-GI-NEXT:    cmgt.4s v18, v2, v3
-; CHECK-GI-NEXT:    cmgt.4s v19, v2, v4
+; CHECK-GI-NEXT:    cmgt.4s v6, v2, v3
+; CHECK-GI-NEXT:    neg.4s v16, v5
+; CHECK-GI-NEXT:    cmgt.4s v7, v2, v4
 ; CHECK-GI-NEXT:    cmgt.4s v2, v2, v0
-; CHECK-GI-NEXT:    bsl.16b v1, v6, v5
-; CHECK-GI-NEXT:    bit.16b v3, v7, v18
-; CHECK-GI-NEXT:    bit.16b v4, v16, v19
-; CHECK-GI-NEXT:    bit.16b v0, v17, v2
+; CHECK-GI-NEXT:    neg.4s v17, v3
+; CHECK-GI-NEXT:    neg.4s v18, v4
+; CHECK-GI-NEXT:    neg.4s v19, v0
+; CHECK-GI-NEXT:    bsl.16b v1, v16, v5
+; CHECK-GI-NEXT:    bit.16b v3, v17, v6
+; CHECK-GI-NEXT:    bit.16b v4, v18, v7
+; CHECK-GI-NEXT:    bit.16b v0, v19, v2
 ; CHECK-GI-NEXT:    add.4s v1, v1, v3
 ; CHECK-GI-NEXT:    add.4s v0, v4, v0
 ; CHECK-GI-NEXT:    add.4s v0, v1, v0
@@ -347,18 +347,18 @@ define i32 @sabd16b_rdx_i32(<16 x i8> %a, <16 x i8> %b) {
 ; CHECK-GI-NEXT:    ssubl2.4s v3, v3, v4
 ; CHECK-GI-NEXT:    ssubl.4s v4, v0, v1
 ; CHECK-GI-NEXT:    ssubl2.4s v0, v0, v1
-; CHECK-GI-NEXT:    neg.4s v6, v5
-; CHECK-GI-NEXT:    neg.4s v7, v3
 ; CHECK-GI-NEXT:    cmgt.4s v1, v2, v5
-; CHECK-GI-NEXT:    neg.4s v16, v4
-; CHECK-GI-NEXT:    neg.4s v17, v0
-; CHECK-GI-NEXT:    cmgt.4s v18, v2, v3
-; CHECK-GI-NEXT:    cmgt.4s v19, v2, v4
+; CHECK-GI-NEXT:    cmgt.4s v6, v2, v3
+; CHECK-GI-NEXT:    neg.4s v16, v5
+; CHECK-GI-NEXT:    cmgt.4s v7, v2, v4
 ; CHECK-GI-NEXT:    cmgt.4s v2, v2, v0
-; CHECK-GI-NEXT:    bsl.16b v1, v6, v5
-; CHECK-GI-NEXT:    bit.16b v3, v7, v18
-; CHECK-GI-NEXT:    bit.16b v4, v16, v19
-; CHECK-GI-NEXT:    bit.16b v0, v17, v2
+; CHECK-GI-NEXT:    neg.4s v17, v3
+; CHECK-GI-NEXT:    neg.4s v18, v4
+; CHECK-GI-NEXT:    neg.4s v19, v0
+; CHECK-GI-NEXT:    bsl.16b v1, v16, v5
+; CHECK-GI-NEXT:    bit.16b v3, v17, v6
+; CHECK-GI-NEXT:    bit.16b v4, v18, v7
+; CHECK-GI-NEXT:    bit.16b v0, v19, v2
 ; CHECK-GI-NEXT:    add.4s v1, v1, v3
 ; CHECK-GI-NEXT:    add.4s v0, v4, v0
 ; CHECK-GI-NEXT:    add.4s v0, v1, v0
@@ -396,12 +396,12 @@ define i32 @uabd8h_rdx(ptr %a, ptr %b) {
 ; CHECK-GI-NEXT:    movi.2d v0, #0000000000000000
 ; CHECK-GI-NEXT:    usubl.4s v3, v1, v2
 ; CHECK-GI-NEXT:    usubl2.4s v1, v1, v2
-; CHECK-GI-NEXT:    neg.4s v2, v3
-; CHECK-GI-NEXT:    neg.4s v4, v1
-; CHECK-GI-NEXT:    cmgt.4s v5, v0, v3
+; CHECK-GI-NEXT:    cmgt.4s v2, v0, v3
 ; CHECK-GI-NEXT:    cmgt.4s v0, v0, v1
-; CHECK-GI-NEXT:    bif.16b v2, v3, v5
-; CHECK-GI-NEXT:    bsl.16b v0, v4, v1
+; CHECK-GI-NEXT:    neg.4s v4, v3
+; CHECK-GI-NEXT:    neg.4s v5, v1
+; CHECK-GI-NEXT:    bsl.16b v2, v4, v3
+; CHECK-GI-NEXT:    bsl.16b v0, v5, v1
 ; CHECK-GI-NEXT:    add.4s v0, v2, v0
 ; CHECK-GI-NEXT:    addv.4s s0, v0
 ; CHECK-GI-NEXT:    fmov w0, s0
@@ -428,15 +428,15 @@ define i32 @sabd8h_rdx(<8 x i16> %a, <8 x i16> %b) {
 ;
 ; CHECK-GI-LABEL: sabd8h_rdx:
 ; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
 ; CHECK-GI-NEXT:    ssubl.4s v3, v0, v1
 ; CHECK-GI-NEXT:    ssubl2.4s v0, v0, v1
-; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
-; CHECK-GI-NEXT:    neg.4s v1, v3
-; CHECK-GI-NEXT:    neg.4s v4, v0
-; CHECK-GI-NEXT:    cmgt.4s v5, v2, v3
+; CHECK-GI-NEXT:    neg.4s v4, v3
+; CHECK-GI-NEXT:    neg.4s v5, v0
+; CHECK-GI-NEXT:    cmgt.4s v1, v2, v3
 ; CHECK-GI-NEXT:    cmgt.4s v2, v2, v0
-; CHECK-GI-NEXT:    bif.16b v1, v3, v5
-; CHECK-GI-NEXT:    bit.16b v0, v4, v2
+; CHECK-GI-NEXT:    bsl.16b v1, v4, v3
+; CHECK-GI-NEXT:    bit.16b v0, v5, v2
 ; CHECK-GI-NEXT:    add.4s v0, v1, v0
 ; CHECK-GI-NEXT:    addv.4s s0, v0
 ; CHECK-GI-NEXT:    fmov w0, s0
@@ -461,10 +461,10 @@ define i32 @uabdl4s_rdx_i32(<4 x i16> %a, <4 x i16> %b) {
 ;
 ; CHECK-GI-LABEL: uabdl4s_rdx_i32:
 ; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
 ; CHECK-GI-NEXT:    usubl.4s v0, v0, v1
-; CHECK-GI-NEXT:    movi.2d v1, #0000000000000000
+; CHECK-GI-NEXT:    cmgt.4s v1, v2, v0
 ; CHECK-GI-NEXT:    neg.4s v2, v0
-; CHECK-GI-NEXT:    cmgt.4s v1, v1, v0
 ; CHECK-GI-NEXT:    bit.16b v0, v2, v1
 ; CHECK-GI-NEXT:    addv.4s s0, v0
 ; CHECK-GI-NEXT:    fmov w0, s0
@@ -499,12 +499,12 @@ define i64 @uabd4s_rdx(ptr %a, ptr %b, i32 %h) {
 ; CHECK-GI-NEXT:    movi.2d v0, #0000000000000000
 ; CHECK-GI-NEXT:    usubl.2d v3, v1, v2
 ; CHECK-GI-NEXT:    usubl2.2d v1, v1, v2
-; CHECK-GI-NEXT:    neg.2d v2, v3
-; CHECK-GI-NEXT:    neg.2d v4, v1
-; CHECK-GI-NEXT:    cmgt.2d v5, v0, v3
+; CHECK-GI-NEXT:    cmgt.2d v2, v0, v3
 ; CHECK-GI-NEXT:    cmgt.2d v0, v0, v1
-; CHECK-GI-NEXT:    bif.16b v2, v3, v5
-; CHECK-GI-NEXT:    bsl.16b v0, v4, v1
+; CHECK-GI-NEXT:    neg.2d v4, v3
+; CHECK-GI-NEXT:    neg.2d v5, v1
+; CHECK-GI-NEXT:    bsl.16b v2, v4, v3
+; CHECK-GI-NEXT:    bsl.16b v0, v5, v1
 ; CHECK-GI-NEXT:    add.2d v0, v2, v0
 ; CHECK-GI-NEXT:    addp.2d d0, v0
 ; CHECK-GI-NEXT:    fmov x0, d0
@@ -531,15 +531,15 @@ define i64 @sabd4s_rdx(<4 x i32> %a, <4 x i32> %b) {
 ;
 ; CHECK-GI-LABEL: sabd4s_rdx:
 ; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
 ; CHECK-GI-NEXT:    ssubl.2d v3, v0, v1
 ; CHECK-GI-NEXT:    ssubl2.2d v0, v0, v1
-; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
-; CHECK-GI-NEXT:    neg.2d v1, v3
-; CHECK-GI-NEXT:    neg.2d v4, v0
-; CHECK-GI-NEXT:    cmgt.2d v5, v2, v3
+; CHECK-GI-NEXT:    neg.2d v4, v3
+; CHECK-GI-NEXT:    neg.2d v5, v0
+; CHECK-GI-NEXT:    cmgt.2d v1, v2, v3
 ; CHECK-GI-NEXT:    cmgt.2d v2, v2, v0
-; CHECK-GI-NEXT:    bif.16b v1, v3, v5
-; CHECK-GI-NEXT:    bit.16b v0, v4, v2
+; CHECK-GI-NEXT:    bsl.16b v1, v4, v3
+; CHECK-GI-NEXT:    bit.16b v0, v5, v2
 ; CHECK-GI-NEXT:    add.2d v0, v1, v0
 ; CHECK-GI-NEXT:    addp.2d d0, v0
 ; CHECK-GI-NEXT:    fmov x0, d0
@@ -564,10 +564,10 @@ define i64 @uabdl2d_rdx_i64(<2 x i32> %a, <2 x i32> %b) {
 ;
 ; CHECK-GI-LABEL: uabdl2d_rdx_i64:
 ; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
 ; CHECK-GI-NEXT:    usubl.2d v0, v0, v1
-; CHECK-GI-NEXT:    movi.2d v1, #0000000000000000
+; CHECK-GI-NEXT:    cmgt.2d v1, v2, v0
 ; CHECK-GI-NEXT:    neg.2d v2, v0
-; CHECK-GI-NEXT:    cmgt.2d v1, v1, v0
 ; CHECK-GI-NEXT:    bit.16b v0, v2, v1
 ; CHECK-GI-NEXT:    addp.2d d0, v0
 ; CHECK-GI-NEXT:    fmov x0, d0
@@ -1796,10 +1796,10 @@ define <2 x i64> @uabd_i32(<2 x i32> %a, <2 x i32> %b) {
 ;
 ; CHECK-GI-LABEL: uabd_i32:
 ; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    movi.2d v2, #0000000000000000
 ; CHECK-GI-NEXT:    ssubl.2d v0, v0, v1
-; CHECK-GI-NEXT:    movi.2d v1, #0000000000000000
+; CHECK-GI-NEXT:    cmgt.2d v1, v2, v0
 ; CHECK-GI-NEXT:    neg.2d v2, v0
-; CHECK-GI-NEXT:    cmgt.2d v1, v1, v0
 ; CHECK-GI-NEXT:    bit.16b v0, v2, v1
 ; CHECK-GI-NEXT:    ret
   %aext = sext <2 x i32> %a to <2 x i64>
diff --git a/llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll b/llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll
index cafee32ada686..d4cc154ac6afc 100644
--- a/llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll
@@ -205,15 +205,15 @@ define <2 x bfloat> @test_vcvt_bf16_f64(<2 x double> %v) nounwind readnone ssp {
 ; GENERIC-LABEL: test_vcvt_bf16_f64:
 ; GENERIC:       // %bb.0:
 ; GENERIC-NEXT:    fcvtxn v0.2s, v0.2d
-; GENERIC-NEXT:    movi.4s v1, #127, msl #8
-; GENERIC-NEXT:    movi.4s v2, #1
+; GENERIC-NEXT:    movi.4s v1, #1
+; GENERIC-NEXT:    movi.4s v2, #127, msl #8
 ; GENERIC-NEXT:    ushr.4s v3, v0, #16
-; GENERIC-NEXT:    add.4s v1, v0, v1
-; GENERIC-NEXT:    and.16b v2, v3, v2
-; GENERIC-NEXT:    add.4s v1, v2, v1
-; GENERIC-NEXT:    fcmeq.4s v2, v0, v0
+; GENERIC-NEXT:    add.4s v2, v0, v2
+; GENERIC-NEXT:    and.16b v1, v3, v1
+; GENERIC-NEXT:    fcmeq.4s v3, v0, v0
 ; GENERIC-NEXT:    orr.4s v0, #64, lsl #16
-; GENERIC-NEXT:    bit.16b v0, v1, v2
+; GENERIC-NEXT:    add.4s v1, v1, v2
+; GENERIC-NEXT:    bit.16b v0, v1, v3
 ; GENERIC-NEXT:    shrn.4h v0, v0, #16
 ; GENERIC-NEXT:    ret
 ;
@@ -238,15 +238,15 @@ define <2 x bfloat> @test_vcvt_bf16_f64(<2 x double> %v) nounwind readnone ssp {
 ; GISEL-LABEL: test_vcvt_bf16_f64:
 ; GISEL:       // %bb.0:
 ; GISEL-NEXT:    fcvtxn v0.2s, v0.2d
-; GISEL-NEXT:    movi.4s v1, #127, msl #8
-; GISEL-NEXT:    movi.4s v2, #1
+; GISEL-NEXT:    movi.4s v1, #1
+; GISEL-NEXT:    movi.4s v2, #127, msl #8
 ; GISEL-NEXT:    ushr.4s v3, v0, #16
-; GISEL-NEXT:    add.4s v1, v0, v1
-; GISEL-NEXT:    and.16b v2, v3, v2
-; GISEL-NEXT:    add.4s v1, v2, v1
-; GISEL-NEXT:    fcmeq.4s v2, v0, v0
+; GISEL-NEXT:    add.4s v2, v0, v2
+; GISEL-NEXT:    and.16b v1, v3, v1
+; GISEL-NEXT:    fcmeq.4s v3, v0, v0
 ; GISEL-NEXT:    orr.4s v0, #64, lsl #16
-; GISEL-NEXT:    bit.16b v0, v1, v2
+; GISEL-NEXT:    add.4s v1, v1, v2
+; GISEL-NEXT:    bit.16b v0, v1, v3
 ; GISEL-NEXT:    shrn.4h v0, v0, #16
 ; GISEL-NEXT:    ret
   %vcvt1.i = fptrunc <2 x double> %v to <2 x bfloat>
diff --git a/llvm/test/CodeGen/AArch64/arm64-vhadd.ll b/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
index dda610e5dd3cb..e754f01daa2a9 100644
--- a/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
@@ -903,10 +903,10 @@ define <2 x i16> @hadd8x2_sext_lsr(<2 x i8> %src1, <2 x i8> %src2) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    shl.2s v0, v0, #24
 ; CHECK-NEXT:    shl.2s v1, v1, #24
+; CHECK-NEXT:    movi d2, #0x00ffff0000ffff
 ; CHECK-NEXT:    sshr.2s v0, v0, #24
 ; CHECK-NEXT:    ssra.2s v0, v1, #24
-; CHECK-NEXT:    movi d1, #0x00ffff0000ffff
-; CHECK-NEXT:    and.8b v0, v0, v1
+; CHECK-NEXT:    and.8b v0, v0, v2
 ; CHECK-NEXT:    ushr.2s v0, v0, #1
 ; CHECK-NEXT:    ret
   %zextsrc1 = sext <2 x i8> %src1 to <2 x i16>
@@ -968,10 +968,10 @@ define <4 x i16> @rhadd8_sext_lsr(<4 x i8> %src1, <4 x i8> %src2) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    shl.4h v0, v0, #8
 ; CHECK-NEXT:    shl.4h v1, v1, #8
+; CHECK-NEXT:    movi.4h v2, #1
 ; CHECK-NEXT:    sshr.4h v0, v0, #8
 ; CHECK-NEXT:    ssra.4h v0, v1, #8
-; CHECK-NEXT:    movi.4h v1, #1
-; CHECK-NEXT:    add.4h v0, v0, v1
+; CHECK-NEXT:    add.4h v0, v0, v2
 ; CHECK-NEXT:    ushr.4h v0, v0, #1
 ; CHECK-NEXT:    ret
   %zextsrc1 = sext <4 x i8> %src1 to <4 x i16>
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-contract.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-contract.ll
index ebf5ce20d4ecc..86b1d5d195ffd 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-contract.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-contract.ll
@@ -7,21 +7,22 @@ target triple = "aarch64-unknown-linux-gnu"
 define <vscale x 4 x double> @mull_add(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c) {
 ; CHECK-LABEL: mull_add:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    uzp1 z6.d, z0.d, z1.d
-; CHECK-NEXT:    uzp2 z7.d, z2.d, z3.d
-; CHECK-NEXT:    uzp2 z0.d, z0.d, z1.d
+; CHECK-NEXT:    uzp2 z6.d, z0.d, z1.d
+; CHECK-NEXT:    uzp1 z0.d, z0.d, z1.d
+; CHECK-NEXT:    uzp2 z1.d, z2.d, z3.d
+; CHECK-NEXT:    uzp1 z2.d, z2.d, z3.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    uzp1 z1.d, z2.d, z3.d
-; CHECK-NEXT:    fmul z2.d, z6.d, z7.d
-; CHECK-NEXT:    fmul z3.d, z0.d, z7.d
-; CHECK-NEXT:    fmad z0.d, p0/m, z1.d, z2.d
-; CHECK-NEXT:    fnmsb z1.d, p0/m, z6.d, z3.d
-; CHECK-NEXT:    uzp2 z2.d, z4.d, z5.d
-; CHECK-NEXT:    uzp1 z3.d, z4.d, z5.d
-; CHECK-NEXT:    fadd z2.d, z0.d, z2.d
+; CHECK-NEXT:    fmul z7.d, z0.d, z1.d
+; CHECK-NEXT:    fmul z1.d, z6.d, z1.d
+; CHECK-NEXT:    movprfx z3, z7
+; CHECK-NEXT:    fmla z3.d, p0/m, z6.d, z2.d
+; CHECK-NEXT:    fnmsb z0.d, p0/m, z2.d, z1.d
+; CHECK-NEXT:    uzp2 z1.d, z4.d, z5.d
+; CHECK-NEXT:    uzp1 z2.d, z4.d, z5.d
+; CHECK-NEXT:    fadd z2.d, z2.d, z0.d
 ; CHECK-NEXT:    fadd z1.d, z3.d, z1.d
-; CHECK-NEXT:    zip1 z0.d, z1.d, z2.d
-; CHECK-NEXT:    zip2 z1.d, z1.d, z2.d
+; CHECK-NEXT:    zip1 z0.d, z2.d, z1.d
+; CHECK-NEXT:    zip2 z1.d, z2.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -49,21 +50,21 @@ entry:
 define <vscale x 4 x double> @mul_add_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_add_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #0
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #0
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z4.d, #0
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #0
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #90
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #90
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #90
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z4.d, #90
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #90
-; CHECK-NEXT:    fadd z0.d, z25.d, z27.d
 ; CHECK-NEXT:    fadd z1.d, z26.d, z24.d
+; CHECK-NEXT:    fadd z0.d, z25.d, z27.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -100,21 +101,21 @@ entry:
 define <vscale x 4 x double> @mul_sub_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_sub_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #0
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #0
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z4.d, #0
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #0
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #90
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #90
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #90
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z4.d, #90
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #90
-; CHECK-NEXT:    fsub z0.d, z25.d, z27.d
 ; CHECK-NEXT:    fsub z1.d, z26.d, z24.d
+; CHECK-NEXT:    fsub z0.d, z25.d, z27.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -151,21 +152,21 @@ entry:
 define <vscale x 4 x double> @mul_conj_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_conj_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #0
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #0
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z4.d, z6.d, #0
-; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #0
+; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #270
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z2.d, z0.d, #90
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z3.d, z1.d, #90
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z4.d, z6.d, #270
-; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #270
-; CHECK-NEXT:    fadd z0.d, z25.d, z27.d
 ; CHECK-NEXT:    fadd z1.d, z26.d, z24.d
+; CHECK-NEXT:    fadd z0.d, z25.d, z27.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -206,8 +207,8 @@ define <vscale x 4 x double> @mul_add_rot_mull(<vscale x 4 x double> %a, <vscale
 ; CHECK-NEXT:    mov z25.d, #0 // =0x0
 ; CHECK-NEXT:    uzp1 z4.d, z4.d, z5.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    and z25.d, z25.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    mov z26.d, z24.d
+; CHECK-NEXT:    and z25.d, z25.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    and z26.d, z26.d, #0x8000000000000000
 ; CHECK-NEXT:    orr z5.d, z25.d, z26.d
 ; CHECK-NEXT:    fadd z5.d, z4.d, z5.d
@@ -220,18 +221,19 @@ define <vscale x 4 x double> @mul_add_rot_mull(<vscale x 4 x double> %a, <vscale
 ; CHECK-NEXT:    fsub z4.d, z4.d, z24.d
 ; CHECK-NEXT:    uzp2 z24.d, z6.d, z7.d
 ; CHECK-NEXT:    uzp1 z6.d, z6.d, z7.d
-; CHECK-NEXT:    fmul z3.d, z0.d, z1.d
+; CHECK-NEXT:    fmul z26.d, z0.d, z1.d
 ; CHECK-NEXT:    fmul z1.d, z25.d, z1.d
-; CHECK-NEXT:    fmul z7.d, z4.d, z24.d
+; CHECK-NEXT:    fmul z3.d, z4.d, z24.d
 ; CHECK-NEXT:    fmul z24.d, z5.d, z24.d
-; CHECK-NEXT:    fmla z3.d, p0/m, z25.d, z2.d
+; CHECK-NEXT:    movprfx z7, z26
+; CHECK-NEXT:    fmla z7.d, p0/m, z25.d, z2.d
 ; CHECK-NEXT:    fnmsb z0.d, p0/m, z2.d, z1.d
-; CHECK-NEXT:    movprfx z1, z7
+; CHECK-NEXT:    movprfx z1, z3
 ; CHECK-NEXT:    fmla z1.d, p0/m, z6.d, z5.d
 ; CHECK-NEXT:    movprfx z2, z24
 ; CHECK-NEXT:    fnmls z2.d, p0/m, z4.d, z6.d
 ; CHECK-NEXT:    fadd z2.d, z0.d, z2.d
-; CHECK-NEXT:    fadd z1.d, z3.d, z1.d
+; CHECK-NEXT:    fadd z1.d, z7.d, z1.d
 ; CHECK-NEXT:    zip1 z0.d, z2.d, z1.d
 ; CHECK-NEXT:    zip2 z1.d, z2.d, z1.d
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
index 1751f99b3e92c..edf580e334e80 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-add-mull-scalable-fast.ll
@@ -41,19 +41,19 @@ entry:
 define <vscale x 4 x double> @mul_add_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_add_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
-; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #0
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
+; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #0
-; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #90
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #90
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
+; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #90
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #90
-; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
 ; CHECK-NEXT:    mov z1.d, z24.d
+; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -90,19 +90,19 @@ entry:
 define <vscale x 4 x double> @mul_sub_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_sub_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
-; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #270
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #270
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
+; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #270
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #0
-; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #180
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z5.d, #180
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
+; CHECK-NEXT:    fcmla z25.d, p0/m, z6.d, z4.d, #180
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #90
-; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
 ; CHECK-NEXT:    mov z1.d, z24.d
+; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -139,19 +139,19 @@ entry:
 define <vscale x 4 x double> @mul_conj_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_conj_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #0
-; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #90
-; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z6.d, #0
+; CHECK-NEXT:    fcmla z25.d, p0/m, z0.d, z2.d, #90
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #0
-; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z6.d, #270
+; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z6.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z5.d, z7.d, #270
-; CHECK-NEXT:    mov z0.d, z25.d
+; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z6.d, #270
 ; CHECK-NEXT:    mov z1.d, z24.d
+; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    ret
 entry:
   %strided.vec = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -188,24 +188,25 @@ entry:
 define <vscale x 4 x double> @mul_add_rot_mull(<vscale x 4 x double> %a, <vscale x 4 x double> %b, <vscale x 4 x double> %c, <vscale x 4 x double> %d) {
 ; CHECK-LABEL: mul_add_rot_mull:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    uzp1 z24.d, z2.d, z3.d
+; CHECK-NEXT:    uzp2 z24.d, z2.d, z3.d
 ; CHECK-NEXT:    uzp2 z25.d, z0.d, z1.d
-; CHECK-NEXT:    uzp2 z2.d, z2.d, z3.d
-; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    uzp1 z2.d, z2.d, z3.d
 ; CHECK-NEXT:    uzp1 z0.d, z0.d, z1.d
+; CHECK-NEXT:    uzp2 z1.d, z4.d, z5.d
 ; CHECK-NEXT:    uzp1 z26.d, z6.d, z7.d
-; CHECK-NEXT:    fmul z1.d, z24.d, z25.d
-; CHECK-NEXT:    fmul z3.d, z2.d, z25.d
-; CHECK-NEXT:    uzp2 z25.d, z4.d, z5.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uzp1 z4.d, z4.d, z5.d
 ; CHECK-NEXT:    uzp2 z5.d, z6.d, z7.d
-; CHECK-NEXT:    fmla z3.d, p0/m, z26.d, z25.d
-; CHECK-NEXT:    fmla z1.d, p0/m, z2.d, z0.d
-; CHECK-NEXT:    movprfx z2, z3
-; CHECK-NEXT:    fmla z2.d, p0/m, z5.d, z4.d
-; CHECK-NEXT:    fmla z1.d, p0/m, z26.d, z4.d
-; CHECK-NEXT:    fnmls z2.d, p0/m, z24.d, z0.d
-; CHECK-NEXT:    fmls z1.d, p0/m, z5.d, z25.d
+; CHECK-NEXT:    fmul z3.d, z2.d, z25.d
+; CHECK-NEXT:    fmul z25.d, z24.d, z25.d
+; CHECK-NEXT:    fmla z3.d, p0/m, z24.d, z0.d
+; CHECK-NEXT:    movprfx z24, z25
+; CHECK-NEXT:    fmla z24.d, p0/m, z26.d, z1.d
+; CHECK-NEXT:    movprfx z6, z24
+; CHECK-NEXT:    fmla z6.d, p0/m, z5.d, z4.d
+; CHECK-NEXT:    fmla z3.d, p0/m, z26.d, z4.d
+; CHECK-NEXT:    fnmsb z2.d, p0/m, z0.d, z6.d
+; CHECK-NEXT:    fmsb z1.d, p0/m, z5.d, z3.d
 ; CHECK-NEXT:    zip1 z0.d, z2.d, z1.d
 ; CHECK-NEXT:    zip2 z1.d, z2.d, z1.d
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f16-mul-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f16-mul-scalable.ll
index 611cf44ea7ee8..cb285c05b2e86 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f16-mul-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f16-mul-scalable.ll
@@ -16,9 +16,10 @@ define <vscale x 4 x half> @complex_mul_v4f16(<vscale x 4 x half> %a, <vscale x
 ; CHECK-NEXT:    uzp1 z0.d, z0.d, z2.d
 ; CHECK-NEXT:    uzp2 z2.d, z1.d, z3.d
 ; CHECK-NEXT:    uzp1 z1.d, z1.d, z3.d
-; CHECK-NEXT:    movprfx z3, z2
-; CHECK-NEXT:    fmul z3.h, p0/m, z3.h, z0.h
+; CHECK-NEXT:    movprfx z5, z2
+; CHECK-NEXT:    fmul z5.h, p0/m, z5.h, z0.h
 ; CHECK-NEXT:    fmul z2.h, p0/m, z2.h, z4.h
+; CHECK-NEXT:    movprfx z3, z5
 ; CHECK-NEXT:    fmla z3.h, p0/m, z1.h, z4.h
 ; CHECK-NEXT:    fnmsb z0.h, p0/m, z1.h, z2.h
 ; CHECK-NEXT:    zip2 z1.d, z0.d, z3.d
@@ -46,8 +47,8 @@ entry:
 define <vscale x 8 x half> @complex_mul_v8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b) {
 ; CHECK-LABEL: complex_mul_v8f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fcmla z2.h, p0/m, z1.h, z0.h, #0
 ; CHECK-NEXT:    fcmla z2.h, p0/m, z1.h, z0.h, #90
 ; CHECK-NEXT:    mov z0.d, z2.d
@@ -72,15 +73,15 @@ entry:
 define <vscale x 16 x half> @complex_mul_v16f16(<vscale x 16 x half> %a, <vscale x 16 x half> %b) {
 ; CHECK-LABEL: complex_mul_v16f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z4.h, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z5.d, z4.d
-; CHECK-NEXT:    fcmla z5.h, p0/m, z2.h, z0.h, #0
 ; CHECK-NEXT:    fcmla z4.h, p0/m, z3.h, z1.h, #0
-; CHECK-NEXT:    fcmla z5.h, p0/m, z2.h, z0.h, #90
+; CHECK-NEXT:    fcmla z5.h, p0/m, z2.h, z0.h, #0
 ; CHECK-NEXT:    fcmla z4.h, p0/m, z3.h, z1.h, #90
-; CHECK-NEXT:    mov z0.d, z5.d
+; CHECK-NEXT:    fcmla z5.h, p0/m, z2.h, z0.h, #90
 ; CHECK-NEXT:    mov z1.d, z4.d
+; CHECK-NEXT:    mov z0.d, z5.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.experimental.vector.deinterleave2.nxv16f16(<vscale x 16 x half> %a)
@@ -103,23 +104,23 @@ entry:
 define <vscale x 32 x half> @complex_mul_v32f16(<vscale x 32 x half> %a, <vscale x 32 x half> %b) {
 ; CHECK-LABEL: complex_mul_v32f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z24.h, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.h, p0/m, z7.h, z3.h, #0
 ; CHECK-NEXT:    fcmla z25.h, p0/m, z4.h, z0.h, #0
 ; CHECK-NEXT:    fcmla z26.h, p0/m, z5.h, z1.h, #0
 ; CHECK-NEXT:    fcmla z27.h, p0/m, z6.h, z2.h, #0
-; CHECK-NEXT:    fcmla z24.h, p0/m, z7.h, z3.h, #0
+; CHECK-NEXT:    fcmla z24.h, p0/m, z7.h, z3.h, #90
 ; CHECK-NEXT:    fcmla z25.h, p0/m, z4.h, z0.h, #90
 ; CHECK-NEXT:    fcmla z26.h, p0/m, z5.h, z1.h, #90
 ; CHECK-NEXT:    fcmla z27.h, p0/m, z6.h, z2.h, #90
-; CHECK-NEXT:    fcmla z24.h, p0/m, z7.h, z3.h, #90
+; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    mov z1.d, z26.d
 ; CHECK-NEXT:    mov z2.d, z27.d
-; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 16 x half>, <vscale x 16 x half> } @llvm.experimental.vector.deinterleave2.nxv32f16(<vscale x 32 x half> %a)
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f32-mul-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f32-mul-scalable.ll
index 0f5e9a2202ddd..1e2afb78de1b0 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f32-mul-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f32-mul-scalable.ll
@@ -7,8 +7,8 @@ target triple = "aarch64"
 define <vscale x 4 x float> @complex_mul_v4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
 ; CHECK-LABEL: complex_mul_v4f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmla z2.s, p0/m, z1.s, z0.s, #0
 ; CHECK-NEXT:    fcmla z2.s, p0/m, z1.s, z0.s, #90
 ; CHECK-NEXT:    mov z0.d, z2.d
@@ -34,15 +34,15 @@ entry:
 define <vscale x 8 x float> @complex_mul_v8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b) {
 ; CHECK-LABEL: complex_mul_v8f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z4.s, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z5.d, z4.d
-; CHECK-NEXT:    fcmla z5.s, p0/m, z2.s, z0.s, #0
 ; CHECK-NEXT:    fcmla z4.s, p0/m, z3.s, z1.s, #0
-; CHECK-NEXT:    fcmla z5.s, p0/m, z2.s, z0.s, #90
+; CHECK-NEXT:    fcmla z5.s, p0/m, z2.s, z0.s, #0
 ; CHECK-NEXT:    fcmla z4.s, p0/m, z3.s, z1.s, #90
-; CHECK-NEXT:    mov z0.d, z5.d
+; CHECK-NEXT:    fcmla z5.s, p0/m, z2.s, z0.s, #90
 ; CHECK-NEXT:    mov z1.d, z4.d
+; CHECK-NEXT:    mov z0.d, z5.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.experimental.vector.deinterleave2.nxv8f32(<vscale x 8 x float> %a)
@@ -65,23 +65,23 @@ entry:
 define <vscale x 16 x float> @complex_mul_v16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b) {
 ; CHECK-LABEL: complex_mul_v16f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z24.s, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.s, p0/m, z7.s, z3.s, #0
 ; CHECK-NEXT:    fcmla z25.s, p0/m, z4.s, z0.s, #0
 ; CHECK-NEXT:    fcmla z26.s, p0/m, z5.s, z1.s, #0
 ; CHECK-NEXT:    fcmla z27.s, p0/m, z6.s, z2.s, #0
-; CHECK-NEXT:    fcmla z24.s, p0/m, z7.s, z3.s, #0
+; CHECK-NEXT:    fcmla z24.s, p0/m, z7.s, z3.s, #90
 ; CHECK-NEXT:    fcmla z25.s, p0/m, z4.s, z0.s, #90
 ; CHECK-NEXT:    fcmla z26.s, p0/m, z5.s, z1.s, #90
 ; CHECK-NEXT:    fcmla z27.s, p0/m, z6.s, z2.s, #90
-; CHECK-NEXT:    fcmla z24.s, p0/m, z7.s, z3.s, #90
+; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    mov z1.d, z26.d
 ; CHECK-NEXT:    mov z2.d, z27.d
-; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 8 x float>, <vscale x 8 x float> } @llvm.experimental.vector.deinterleave2.nxv16f32(<vscale x 16 x float> %a)
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f64-mul-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f64-mul-scalable.ll
index 1fe554bdc616e..17a239a09a033 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-f64-mul-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-f64-mul-scalable.ll
@@ -7,8 +7,8 @@ target triple = "aarch64"
 define <vscale x 2 x double> @complex_mul_v2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b) {
 ; CHECK-LABEL: complex_mul_v2f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fcmla z2.d, p0/m, z1.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z2.d, p0/m, z1.d, z0.d, #90
 ; CHECK-NEXT:    mov z0.d, z2.d
@@ -34,15 +34,15 @@ entry:
 define <vscale x 4 x double> @complex_mul_v4f64(<vscale x 4 x double> %a, <vscale x 4 x double> %b) {
 ; CHECK-LABEL: complex_mul_v4f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z4.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z5.d, z4.d
-; CHECK-NEXT:    fcmla z5.d, p0/m, z2.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z4.d, p0/m, z3.d, z1.d, #0
-; CHECK-NEXT:    fcmla z5.d, p0/m, z2.d, z0.d, #90
+; CHECK-NEXT:    fcmla z5.d, p0/m, z2.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z4.d, p0/m, z3.d, z1.d, #90
-; CHECK-NEXT:    mov z0.d, z5.d
+; CHECK-NEXT:    fcmla z5.d, p0/m, z2.d, z0.d, #90
 ; CHECK-NEXT:    mov z1.d, z4.d
+; CHECK-NEXT:    mov z0.d, z5.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.experimental.vector.deinterleave2.nxv4f64(<vscale x 4 x double> %a)
@@ -65,23 +65,23 @@ entry:
 define <vscale x 8 x double> @complex_mul_v8f64(<vscale x 8 x double> %a, <vscale x 8 x double> %b) {
 ; CHECK-LABEL: complex_mul_v8f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z24.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, z24.d
 ; CHECK-NEXT:    mov z26.d, z24.d
 ; CHECK-NEXT:    mov z27.d, z24.d
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z3.d, #0
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z0.d, #0
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z5.d, z1.d, #0
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z2.d, #0
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z3.d, #0
+; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z3.d, #90
 ; CHECK-NEXT:    fcmla z25.d, p0/m, z4.d, z0.d, #90
 ; CHECK-NEXT:    fcmla z26.d, p0/m, z5.d, z1.d, #90
 ; CHECK-NEXT:    fcmla z27.d, p0/m, z6.d, z2.d, #90
-; CHECK-NEXT:    fcmla z24.d, p0/m, z7.d, z3.d, #90
+; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    mov z0.d, z25.d
 ; CHECK-NEXT:    mov z1.d, z26.d
 ; CHECK-NEXT:    mov z2.d, z27.d
-; CHECK-NEXT:    mov z3.d, z24.d
 ; CHECK-NEXT:    ret
 entry:
   %a.deinterleaved = tail call { <vscale x 4 x double>, <vscale x 4 x double> } @llvm.experimental.vector.deinterleave2.nxv8f64(<vscale x 8 x double> %a)
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-i16-mul-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-i16-mul-scalable.ll
index 1b8a21b66ade9..07488b623b98d 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-i16-mul-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-i16-mul-scalable.ll
@@ -16,8 +16,9 @@ define <vscale x 4 x i16> @complex_mul_v4i16(<vscale x 4 x i16> %a, <vscale x 4
 ; CHECK-NEXT:    uzp2 z0.d, z0.d, z2.d
 ; CHECK-NEXT:    uzp1 z2.d, z1.d, z3.d
 ; CHECK-NEXT:    uzp2 z1.d, z1.d, z3.d
-; CHECK-NEXT:    mul z3.d, z2.d, z0.d
+; CHECK-NEXT:    mul z5.d, z2.d, z0.d
 ; CHECK-NEXT:    mul z2.d, z2.d, z4.d
+; CHECK-NEXT:    movprfx z3, z5
 ; CHECK-NEXT:    mla z3.d, p0/m, z1.d, z4.d
 ; CHECK-NEXT:    msb z0.d, p0/m, z1.d, z2.d
 ; CHECK-NEXT:    zip2 z1.d, z0.d, z3.d
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
index 467c3c254fc2d..19318fdeeca70 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
@@ -14,37 +14,37 @@ target triple = "aarch64"
 define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: complex_mul_v2f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w9, #100 // =0x64
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
-; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    whilelo p1.d, xzr, x9
+; CHECK-NEXT:    mov w9, #100 // =0x64
 ; CHECK-NEXT:    cntd x10
+; CHECK-NEXT:    whilelo p1.d, xzr, x9
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    rdvl x11, #2
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x12, x10
 ; CHECK-NEXT:    zip2 z0.d, z1.d, z1.d
 ; CHECK-NEXT:    zip1 z1.d, z1.d, z1.d
 ; CHECK-NEXT:  .LBB0_1: // %vector.body
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    zip2 p3.d, p1.d, p1.d
+; CHECK-NEXT:    zip2 p2.d, p1.d, p1.d
 ; CHECK-NEXT:    add x13, x0, x8
 ; CHECK-NEXT:    add x14, x1, x8
-; CHECK-NEXT:    zip1 p2.d, p1.d, p1.d
+; CHECK-NEXT:    zip1 p1.d, p1.d, p1.d
 ; CHECK-NEXT:    mov z6.d, z1.d
 ; CHECK-NEXT:    mov z7.d, z0.d
-; CHECK-NEXT:    whilelo p1.d, x12, x9
+; CHECK-NEXT:    ld1d { z2.d }, p2/z, [x13, #1, mul vl]
+; CHECK-NEXT:    ld1d { z4.d }, p2/z, [x14, #1, mul vl]
 ; CHECK-NEXT:    add x8, x8, x11
-; CHECK-NEXT:    add x12, x12, x10
-; CHECK-NEXT:    ld1d { z2.d }, p3/z, [x13, #1, mul vl]
-; CHECK-NEXT:    ld1d { z4.d }, p3/z, [x14, #1, mul vl]
-; CHECK-NEXT:    ld1d { z3.d }, p2/z, [x13]
-; CHECK-NEXT:    ld1d { z5.d }, p2/z, [x14]
+; CHECK-NEXT:    ld1d { z3.d }, p1/z, [x13]
+; CHECK-NEXT:    ld1d { z5.d }, p1/z, [x14]
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #0
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #90
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT:    mov z0.d, p3/m, z7.d
-; CHECK-NEXT:    mov z1.d, p2/m, z6.d
+; CHECK-NEXT:    mov z0.d, p2/m, z7.d
+; CHECK-NEXT:    mov z1.d, p1/m, z6.d
+; CHECK-NEXT:    whilelo p1.d, x12, x9
+; CHECK-NEXT:    add x12, x12, x10
 ; CHECK-NEXT:    b.mi .LBB0_1
 ; CHECK-NEXT:  // %bb.2: // %exit.block
 ; CHECK-NEXT:    uzp1 z2.d, z1.d, z0.d
@@ -114,10 +114,10 @@ define %"class.std::complex" @complex_mul_predicated_v2f64(ptr %a, ptr %b, ptr %
 ; CHECK-LABEL: complex_mul_predicated_v2f64:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntd x10
-; CHECK-NEXT:    neg x11, x10
 ; CHECK-NEXT:    mov w12, #100 // =0x64
+; CHECK-NEXT:    neg x11, x10
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    mov x9, xzr
 ; CHECK-NEXT:    and x11, x11, x12
@@ -133,20 +133,20 @@ define %"class.std::complex" @complex_mul_predicated_v2f64(ptr %a, ptr %b, ptr %
 ; CHECK-NEXT:    mov z7.d, z0.d
 ; CHECK-NEXT:    add x9, x9, x10
 ; CHECK-NEXT:    add x8, x8, x12
-; CHECK-NEXT:    cmpne p1.d, p0/z, z2.d, #0
+; CHECK-NEXT:    cmpne p2.d, p0/z, z2.d, #0
 ; CHECK-NEXT:    cmp x11, x9
-; CHECK-NEXT:    zip2 p2.d, p1.d, p1.d
-; CHECK-NEXT:    zip1 p1.d, p1.d, p1.d
-; CHECK-NEXT:    ld1d { z2.d }, p2/z, [x13, #1, mul vl]
-; CHECK-NEXT:    ld1d { z4.d }, p2/z, [x14, #1, mul vl]
-; CHECK-NEXT:    ld1d { z3.d }, p1/z, [x13]
-; CHECK-NEXT:    ld1d { z5.d }, p1/z, [x14]
+; CHECK-NEXT:    zip2 p1.d, p2.d, p2.d
+; CHECK-NEXT:    zip1 p2.d, p2.d, p2.d
+; CHECK-NEXT:    ld1d { z2.d }, p1/z, [x13, #1, mul vl]
+; CHECK-NEXT:    ld1d { z4.d }, p1/z, [x14, #1, mul vl]
+; CHECK-NEXT:    ld1d { z3.d }, p2/z, [x13]
+; CHECK-NEXT:    ld1d { z5.d }, p2/z, [x14]
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #0
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #90
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT:    mov z0.d, p2/m, z7.d
-; CHECK-NEXT:    mov z1.d, p1/m, z6.d
+; CHECK-NEXT:    mov z0.d, p1/m, z7.d
+; CHECK-NEXT:    mov z1.d, p2/m, z6.d
 ; CHECK-NEXT:    b.ne .LBB1_1
 ; CHECK-NEXT:  // %bb.2: // %exit.block
 ; CHECK-NEXT:    uzp1 z2.d, z1.d, z0.d
@@ -217,8 +217,8 @@ exit.block:                                     ; preds = %vector.body
 define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, ptr %cond) {
 ; CHECK-LABEL: complex_mul_predicated_x2_v2f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w10, #100 // =0x64
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
+; CHECK-NEXT:    mov w10, #100 // =0x64
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    whilelo p1.d, xzr, x10
 ; CHECK-NEXT:    mov x8, xzr
@@ -236,19 +236,19 @@ define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, pt
 ; CHECK-NEXT:    mov z7.d, z0.d
 ; CHECK-NEXT:    add x9, x9, x11
 ; CHECK-NEXT:    add x8, x8, x12
-; CHECK-NEXT:    cmpne p1.d, p1/z, z2.d, #0
-; CHECK-NEXT:    zip2 p3.d, p1.d, p1.d
-; CHECK-NEXT:    zip1 p2.d, p1.d, p1.d
-; CHECK-NEXT:    whilelo p1.d, x9, x10
-; CHECK-NEXT:    ld1d { z2.d }, p3/z, [x13, #1, mul vl]
-; CHECK-NEXT:    ld1d { z4.d }, p3/z, [x14, #1, mul vl]
+; CHECK-NEXT:    cmpne p2.d, p1/z, z2.d, #0
+; CHECK-NEXT:    zip2 p1.d, p2.d, p2.d
+; CHECK-NEXT:    zip1 p2.d, p2.d, p2.d
+; CHECK-NEXT:    ld1d { z2.d }, p1/z, [x13, #1, mul vl]
+; CHECK-NEXT:    ld1d { z4.d }, p1/z, [x14, #1, mul vl]
 ; CHECK-NEXT:    ld1d { z3.d }, p2/z, [x13]
 ; CHECK-NEXT:    ld1d { z5.d }, p2/z, [x14]
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #0
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z4.d, z2.d, #90
 ; CHECK-NEXT:    fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT:    mov z0.d, p3/m, z7.d
+; CHECK-NEXT:    mov z0.d, p1/m, z7.d
+; CHECK-NEXT:    whilelo p1.d, x9, x10
 ; CHECK-NEXT:    mov z1.d, p2/m, z6.d
 ; CHECK-NEXT:    b.mi .LBB2_1
 ; CHECK-NEXT:  // %bb.2: // %exit.block
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
index 1696ac8709d40..664d99a3627b5 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
@@ -15,11 +15,11 @@ define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: complex_mul_v2f64:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cntd x9
-; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    neg x9, x9
 ; CHECK-NEXT:    mov w10, #100 // =0x64
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    and x10, x9, x10
 ; CHECK-NEXT:    rdvl x11, #2
@@ -101,18 +101,18 @@ exit.block:                                     ; preds = %vector.body
 define %"class.std::complex" @complex_mul_nonzero_init_v2f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: complex_mul_nonzero_init_v2f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    fmov d0, #1.00000000
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
-; CHECK-NEXT:    fmov d2, #2.00000000
 ; CHECK-NEXT:    cntd x9
-; CHECK-NEXT:    mov w10, #100 // =0x64
-; CHECK-NEXT:    ptrue p1.b
+; CHECK-NEXT:    fmov d2, #2.00000000
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    neg x9, x9
+; CHECK-NEXT:    ptrue p1.b
+; CHECK-NEXT:    mov w10, #100 // =0x64
 ; CHECK-NEXT:    mov x8, xzr
+; CHECK-NEXT:    sel z3.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    and x10, x9, x10
 ; CHECK-NEXT:    rdvl x11, #2
-; CHECK-NEXT:    sel z3.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    mov z1.d, p0/m, z2.d
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    zip2 z0.d, z1.d, z3.d
@@ -190,12 +190,12 @@ define %"class.std::complex" @complex_mul_v2f64_unrolled(ptr %a, ptr %b) {
 ; CHECK-LABEL: complex_mul_v2f64_unrolled:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cntw x9
-; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    neg x9, x9
 ; CHECK-NEXT:    mov w10, #1000 // =0x3e8
+; CHECK-NEXT:    neg x9, x9
 ; CHECK-NEXT:    rdvl x12, #2
+; CHECK-NEXT:    ptrue p1.b
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    and x10, x9, x10
 ; CHECK-NEXT:    zip2 z0.d, z1.d, z1.d
@@ -324,10 +324,10 @@ define dso_local %"class.std::complex" @reduction_mix(ptr %a, ptr %b, ptr noalia
 ; CHECK-LABEL: reduction_mix:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    mov z2.d, #0 // =0x0
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntd x9
-; CHECK-NEXT:    neg x10, x9
 ; CHECK-NEXT:    mov w11, #100 // =0x64
+; CHECK-NEXT:    neg x10, x9
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    and x10, x10, x11
 ; CHECK-NEXT:    rdvl x11, #2
@@ -349,8 +349,8 @@ define dso_local %"class.std::complex" @reduction_mix(ptr %a, ptr %b, ptr noalia
 ; CHECK-NEXT:    uaddv d2, p0, z2.d
 ; CHECK-NEXT:    uzp2 z3.d, z1.d, z0.d
 ; CHECK-NEXT:    uzp1 z1.d, z1.d, z0.d
-; CHECK-NEXT:    fmov x8, d2
 ; CHECK-NEXT:    faddv d0, p0, z3.d
+; CHECK-NEXT:    fmov x8, d2
 ; CHECK-NEXT:    faddv d1, p0, z1.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 killed $z1
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-splat-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-splat-scalable.ll
index 742a7099559f7..17bf5ba6eb48b 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-splat-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-splat-scalable.ll
@@ -8,8 +8,8 @@ target triple = "aarch64"
 define <vscale x 4 x double> @complex_mul_const(<vscale x 4 x double> %a, <vscale x 4 x double> %b) {
 ; CHECK-LABEL: complex_mul_const:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z4.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fmov z7.d, #3.00000000
 ; CHECK-NEXT:    fmov z24.d, #11.00000000
 ; CHECK-NEXT:    mov z6.d, z4.d
@@ -55,25 +55,25 @@ entry:
 define <vscale x 4 x double> @complex_mul_non_const(<vscale x 4 x double> %a, <vscale x 4 x double> %b, [2 x double] %c) {
 ; CHECK-LABEL: complex_mul_non_const:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z6.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $d5 killed $d5 def $z5
 ; CHECK-NEXT:    // kill: def $d4 killed $d4 def $z4
 ; CHECK-NEXT:    mov z5.d, d5
 ; CHECK-NEXT:    mov z4.d, d4
 ; CHECK-NEXT:    mov z24.d, z6.d
 ; CHECK-NEXT:    mov z7.d, z6.d
-; CHECK-NEXT:    zip2 z25.d, z4.d, z5.d
-; CHECK-NEXT:    zip1 z4.d, z4.d, z5.d
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #0
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z0.d, z2.d, #0
 ; CHECK-NEXT:    fcmla z24.d, p0/m, z1.d, z3.d, #90
+; CHECK-NEXT:    zip2 z1.d, z4.d, z5.d
 ; CHECK-NEXT:    fcmla z7.d, p0/m, z0.d, z2.d, #90
+; CHECK-NEXT:    zip1 z2.d, z4.d, z5.d
 ; CHECK-NEXT:    mov z0.d, z6.d
-; CHECK-NEXT:    fcmla z6.d, p0/m, z24.d, z25.d, #0
-; CHECK-NEXT:    fcmla z0.d, p0/m, z7.d, z4.d, #0
-; CHECK-NEXT:    fcmla z6.d, p0/m, z24.d, z25.d, #90
-; CHECK-NEXT:    fcmla z0.d, p0/m, z7.d, z4.d, #90
+; CHECK-NEXT:    fcmla z6.d, p0/m, z24.d, z1.d, #0
+; CHECK-NEXT:    fcmla z0.d, p0/m, z7.d, z2.d, #0
+; CHECK-NEXT:    fcmla z6.d, p0/m, z24.d, z1.d, #90
+; CHECK-NEXT:    fcmla z0.d, p0/m, z7.d, z2.d, #90
 ; CHECK-NEXT:    mov z1.d, z6.d
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/concat_vector-truncate-combine.ll b/llvm/test/CodeGen/AArch64/concat_vector-truncate-combine.ll
index 1c254f9ed935d..e6d5a2ac0fd79 100644
--- a/llvm/test/CodeGen/AArch64/concat_vector-truncate-combine.ll
+++ b/llvm/test/CodeGen/AArch64/concat_vector-truncate-combine.ll
@@ -96,8 +96,8 @@ entry:
 define void @test_concat_fptrunc_v4f64_to_v4f32(ptr %ptr) #1 {
 ; CHECK-LABEL: test_concat_fptrunc_v4f64_to_v4f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fmov z0.s, #1.00000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/dag-combine-concat-vectors.ll b/llvm/test/CodeGen/AArch64/dag-combine-concat-vectors.ll
index 4a2e85c715f7a..83c7f73800af1 100644
--- a/llvm/test/CodeGen/AArch64/dag-combine-concat-vectors.ll
+++ b/llvm/test/CodeGen/AArch64/dag-combine-concat-vectors.ll
@@ -9,8 +9,8 @@ define fastcc i8 @allocno_reload_assign() {
 ; CHECK-LABEL: allocno_reload_assign:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.b, #0 // =0x0
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z16.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    uunpklo z1.h, z0.b
 ; CHECK-NEXT:    uunpkhi z0.h, z0.b
@@ -48,12 +48,12 @@ define fastcc i8 @allocno_reload_assign() {
 ; CHECK-NEXT:    punpklo p4.h, p3.b
 ; CHECK-NEXT:    punpkhi p3.h, p3.b
 ; CHECK-NEXT:    st1b { z2.d }, p4, [z16.d]
-; CHECK-NEXT:    punpklo p4.h, p2.b
-; CHECK-NEXT:    punpkhi p2.h, p2.b
 ; CHECK-NEXT:    st1b { z3.d }, p3, [z16.d]
-; CHECK-NEXT:    punpklo p3.h, p4.b
-; CHECK-NEXT:    st1b { z4.d }, p3, [z16.d]
-; CHECK-NEXT:    punpkhi p3.h, p4.b
+; CHECK-NEXT:    punpklo p3.h, p2.b
+; CHECK-NEXT:    punpkhi p2.h, p2.b
+; CHECK-NEXT:    punpklo p4.h, p3.b
+; CHECK-NEXT:    punpkhi p3.h, p3.b
+; CHECK-NEXT:    st1b { z4.d }, p4, [z16.d]
 ; CHECK-NEXT:    st1b { z5.d }, p3, [z16.d]
 ; CHECK-NEXT:    punpklo p3.h, p2.b
 ; CHECK-NEXT:    punpkhi p2.h, p2.b
diff --git a/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-signed.ll b/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-signed.ll
index 49ad3ae7d6290..6e13ae6feb66b 100644
--- a/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-signed.ll
+++ b/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-signed.ll
@@ -228,13 +228,13 @@ define <2 x i64> @vector_i128_i64(<2 x i64> %x, <2 x i64> %y, ptr %divdst) nounw
 ; ALL-NEXT:    sdiv x9, x9, x8
 ; ALL-NEXT:    sdiv x11, x11, x10
 ; ALL-NEXT:    mul x8, x9, x8
+; ALL-NEXT:    fmov d2, x9
 ; ALL-NEXT:    fmov d1, x8
 ; ALL-NEXT:    mul x10, x11, x10
+; ALL-NEXT:    mov v2.d[1], x11
+; ALL-NEXT:    str q2, [x0]
 ; ALL-NEXT:    mov v1.d[1], x10
 ; ALL-NEXT:    sub v0.2d, v0.2d, v1.2d
-; ALL-NEXT:    fmov d1, x9
-; ALL-NEXT:    mov v1.d[1], x11
-; ALL-NEXT:    str q1, [x0]
 ; ALL-NEXT:    ret
   %div = sdiv <2 x i64> %x, %y
   store <2 x i64> %div, ptr %divdst, align 16
diff --git a/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-unsigned.ll b/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-unsigned.ll
index 3bc50b2f03d83..650219e03b8a7 100644
--- a/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-unsigned.ll
+++ b/llvm/test/CodeGen/AArch64/div-rem-pair-recomposition-unsigned.ll
@@ -228,13 +228,13 @@ define <2 x i64> @vector_i128_i64(<2 x i64> %x, <2 x i64> %y, ptr %divdst) nounw
 ; ALL-NEXT:    udiv x9, x9, x8
 ; ALL-NEXT:    udiv x11, x11, x10
 ; ALL-NEXT:    mul x8, x9, x8
+; ALL-NEXT:    fmov d2, x9
 ; ALL-NEXT:    fmov d1, x8
 ; ALL-NEXT:    mul x10, x11, x10
+; ALL-NEXT:    mov v2.d[1], x11
+; ALL-NEXT:    str q2, [x0]
 ; ALL-NEXT:    mov v1.d[1], x10
 ; ALL-NEXT:    sub v0.2d, v0.2d, v1.2d
-; ALL-NEXT:    fmov d1, x9
-; ALL-NEXT:    mov v1.d[1], x11
-; ALL-NEXT:    str q1, [x0]
 ; ALL-NEXT:    ret
   %div = udiv <2 x i64> %x, %y
   store <2 x i64> %div, ptr %divdst, align 16
diff --git a/llvm/test/CodeGen/AArch64/extbinopload.ll b/llvm/test/CodeGen/AArch64/extbinopload.ll
index dff4831330deb..bd9d9b99622e3 100644
--- a/llvm/test/CodeGen/AArch64/extbinopload.ll
+++ b/llvm/test/CodeGen/AArch64/extbinopload.ll
@@ -502,9 +502,9 @@ define <16 x i32> @double_bv_4xv4i8_i32(ptr %p, ptr %q, ptr %r, ptr %s, ptr %t,
 ; CHECK-NEXT:    usubl v4.8h, v4.8b, v5.8b
 ; CHECK-NEXT:    ld1 { v6.s }[1], [x7], #4
 ; CHECK-NEXT:    ld1 { v7.s }[1], [x7]
-; CHECK-NEXT:    usubl v5.8h, v6.8b, v7.8b
 ; CHECK-NEXT:    shll v0.4s, v4.4h, #16
 ; CHECK-NEXT:    shll2 v4.4s, v4.8h, #16
+; CHECK-NEXT:    usubl v5.8h, v6.8b, v7.8b
 ; CHECK-NEXT:    saddw v0.4s, v0.4s, v1.4h
 ; CHECK-NEXT:    saddw2 v1.4s, v4.4s, v1.8h
 ; CHECK-NEXT:    shll v6.4s, v5.4h, #16
@@ -647,10 +647,10 @@ define <16 x i32> @extrause_load(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ldr s1, [x0]
 ; CHECK-NEXT:    add x8, x3, #8
-; CHECK-NEXT:    add x11, x3, #12
+; CHECK-NEXT:    add x11, x1, #12
 ; CHECK-NEXT:    str s1, [x4]
 ; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    ldp s0, s4, [x2]
+; CHECK-NEXT:    ldr s0, [x2]
 ; CHECK-NEXT:    ushll v2.8h, v0.8b, #0
 ; CHECK-NEXT:    umov w9, v2.h[0]
 ; CHECK-NEXT:    umov w10, v2.h[1]
@@ -664,31 +664,32 @@ define <16 x i32> @extrause_load(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-NEXT:    add x9, x1, #4
 ; CHECK-NEXT:    mov v1.d[1], v2.d[0]
 ; CHECK-NEXT:    mov v0.b[11], w10
-; CHECK-NEXT:    add x10, x1, #12
+; CHECK-NEXT:    add x10, x3, #12
 ; CHECK-NEXT:    bic v1.8h, #255, lsl #8
 ; CHECK-NEXT:    ld1 { v0.s }[3], [x3], #4
-; CHECK-NEXT:    ldr s3, [x0, #12]
-; CHECK-NEXT:    ldp s2, s7, [x0, #4]
-; CHECK-NEXT:    ld1 { v4.s }[1], [x3]
-; CHECK-NEXT:    ldp s5, s6, [x2, #8]
-; CHECK-NEXT:    ld1 { v3.s }[1], [x10]
-; CHECK-NEXT:    ld1 { v2.s }[1], [x9]
-; CHECK-NEXT:    ld1 { v5.s }[1], [x8]
-; CHECK-NEXT:    ld1 { v6.s }[1], [x11]
+; CHECK-NEXT:    ldr s5, [x0, #4]
+; CHECK-NEXT:    ldp s2, s3, [x2, #4]
+; CHECK-NEXT:    ldr s7, [x2, #12]
+; CHECK-NEXT:    ldp s6, s4, [x0, #8]
+; CHECK-NEXT:    ld1 { v5.s }[1], [x9]
+; CHECK-NEXT:    ld1 { v7.s }[1], [x10]
+; CHECK-NEXT:    ld1 { v3.s }[1], [x8]
+; CHECK-NEXT:    ld1 { v2.s }[1], [x3]
 ; CHECK-NEXT:    add x8, x1, #8
-; CHECK-NEXT:    ld1 { v7.s }[1], [x8]
-; CHECK-NEXT:    uaddl v2.8h, v2.8b, v3.8b
-; CHECK-NEXT:    ushll v3.8h, v5.8b, #0
-; CHECK-NEXT:    uaddl v4.8h, v4.8b, v6.8b
-; CHECK-NEXT:    uaddw v1.8h, v1.8h, v7.8b
+; CHECK-NEXT:    ld1 { v4.s }[1], [x11]
+; CHECK-NEXT:    ld1 { v6.s }[1], [x8]
+; CHECK-NEXT:    ushll v3.8h, v3.8b, #0
+; CHECK-NEXT:    uaddl v2.8h, v2.8b, v7.8b
+; CHECK-NEXT:    uaddl v4.8h, v5.8b, v4.8b
+; CHECK-NEXT:    uaddw v1.8h, v1.8h, v6.8b
 ; CHECK-NEXT:    uaddw2 v5.8h, v3.8h, v0.16b
-; CHECK-NEXT:    ushll v0.4s, v2.4h, #3
+; CHECK-NEXT:    ushll v6.4s, v2.4h, #3
 ; CHECK-NEXT:    ushll2 v2.4s, v2.8h, #3
-; CHECK-NEXT:    ushll v6.4s, v4.4h, #3
+; CHECK-NEXT:    ushll v0.4s, v4.4h, #3
 ; CHECK-NEXT:    ushll2 v3.4s, v4.8h, #3
 ; CHECK-NEXT:    uaddw v0.4s, v0.4s, v1.4h
-; CHECK-NEXT:    uaddw2 v1.4s, v2.4s, v1.8h
-; CHECK-NEXT:    uaddw2 v3.4s, v3.4s, v5.8h
+; CHECK-NEXT:    uaddw2 v1.4s, v3.4s, v1.8h
+; CHECK-NEXT:    uaddw2 v3.4s, v2.4s, v5.8h
 ; CHECK-NEXT:    uaddw v2.4s, v6.4s, v5.4h
 ; CHECK-NEXT:    ret
   %lp1 = load <4 x i8>, ptr %p
@@ -762,35 +763,35 @@ define <16 x i32> @extrause_shuffle(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-NEXT:    add x8, x1, #8
 ; CHECK-NEXT:    ldr s6, [x1, #12]
 ; CHECK-NEXT:    ldp s17, s18, [x2, #8]
-; CHECK-NEXT:    ldp s2, s3, [x2]
+; CHECK-NEXT:    ldp s3, s5, [x2]
 ; CHECK-NEXT:    add x9, x3, #8
 ; CHECK-NEXT:    mov v4.16b, v1.16b
 ; CHECK-NEXT:    ldp s7, s16, [x0]
-; CHECK-NEXT:    ldr s5, [x3, #12]
+; CHECK-NEXT:    ldr s2, [x3, #12]
 ; CHECK-NEXT:    mov v1.s[1], v6.s[0]
-; CHECK-NEXT:    ld1 { v2.s }[1], [x3], #4
+; CHECK-NEXT:    ld1 { v3.s }[1], [x3], #4
 ; CHECK-NEXT:    mov v4.s[1], v6.s[0]
 ; CHECK-NEXT:    ld1 { v7.s }[1], [x1], #4
 ; CHECK-NEXT:    ld1 { v16.s }[1], [x1]
-; CHECK-NEXT:    ld1 { v3.s }[1], [x3]
-; CHECK-NEXT:    ld1 { v0.s }[1], [x8]
+; CHECK-NEXT:    ld1 { v5.s }[1], [x3]
 ; CHECK-NEXT:    ld1 { v17.s }[1], [x9]
+; CHECK-NEXT:    ld1 { v0.s }[1], [x8]
 ; CHECK-NEXT:    mov v4.s[2], v18.s[0]
-; CHECK-NEXT:    mov v18.s[1], v5.s[0]
+; CHECK-NEXT:    mov v18.s[1], v2.s[0]
 ; CHECK-NEXT:    uaddl v1.8h, v16.8b, v1.8b
 ; CHECK-NEXT:    uaddl v6.8h, v7.8b, v0.8b
-; CHECK-NEXT:    uaddl v2.8h, v2.8b, v17.8b
-; CHECK-NEXT:    uaddl v3.8h, v3.8b, v18.8b
+; CHECK-NEXT:    uaddl v7.8h, v3.8b, v17.8b
 ; CHECK-NEXT:    ushll v0.4s, v1.4h, #3
 ; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #3
-; CHECK-NEXT:    mov v4.s[3], v5.s[0]
+; CHECK-NEXT:    uaddl v5.8h, v5.8b, v18.8b
+; CHECK-NEXT:    mov v4.s[3], v2.s[0]
 ; CHECK-NEXT:    uaddw v0.4s, v0.4s, v6.4h
 ; CHECK-NEXT:    uaddw2 v1.4s, v1.4s, v6.8h
-; CHECK-NEXT:    ushll v7.4s, v3.4h, #3
-; CHECK-NEXT:    ushll2 v3.4s, v3.8h, #3
+; CHECK-NEXT:    ushll v16.4s, v5.4h, #3
+; CHECK-NEXT:    ushll2 v3.4s, v5.8h, #3
 ; CHECK-NEXT:    str q4, [x4]
-; CHECK-NEXT:    uaddw2 v3.4s, v3.4s, v2.8h
-; CHECK-NEXT:    uaddw v2.4s, v7.4s, v2.4h
+; CHECK-NEXT:    uaddw2 v3.4s, v3.4s, v7.8h
+; CHECK-NEXT:    uaddw v2.4s, v16.4s, v7.4h
 ; CHECK-NEXT:    ret
   %lp1 = load <4 x i8>, ptr %p
   %p2 = getelementptr i8, ptr %p, i32 4
@@ -873,8 +874,8 @@ define <16 x i32> @extrause_ext(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-NEXT:    ld1 { v4.s }[1], [x11]
 ; CHECK-NEXT:    ld1 { v2.s }[1], [x3]
 ; CHECK-NEXT:    ld1 { v0.s }[1], [x10]
-; CHECK-NEXT:    ld1 { v7.s }[1], [x9]
 ; CHECK-NEXT:    ld1 { v6.s }[1], [x8]
+; CHECK-NEXT:    ld1 { v7.s }[1], [x9]
 ; CHECK-NEXT:    uaddl v5.8h, v5.8b, v4.8b
 ; CHECK-NEXT:    uaddl v2.8h, v2.8b, v0.8b
 ; CHECK-NEXT:    ushll v16.8h, v0.8b, #0
@@ -972,8 +973,8 @@ define <16 x i32> @extrause_add(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-NEXT:    ld1 { v5.s }[1], [x11]
 ; CHECK-NEXT:    ld1 { v3.s }[1], [x3]
 ; CHECK-NEXT:    ld1 { v7.s }[1], [x10]
-; CHECK-NEXT:    ld1 { v4.s }[1], [x9]
 ; CHECK-NEXT:    ld1 { v6.s }[1], [x8]
+; CHECK-NEXT:    ld1 { v4.s }[1], [x9]
 ; CHECK-NEXT:    uaddl v5.8h, v1.8b, v5.8b
 ; CHECK-NEXT:    uaddl v7.8h, v3.8b, v7.8b
 ; CHECK-NEXT:    uaddl v1.8h, v0.8b, v4.8b
@@ -1072,23 +1073,23 @@ define <16 x i32> @extrause_ext2(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-NEXT:    ld1 { v6.s }[1], [x9]
 ; CHECK-NEXT:    ld1 { v4.s }[1], [x8]
 ; CHECK-NEXT:    uaddl v7.8h, v3.8b, v7.8b
-; CHECK-NEXT:    uaddl v3.8h, v1.8b, v5.8b
+; CHECK-NEXT:    uaddl v1.8h, v1.8b, v5.8b
 ; CHECK-NEXT:    uaddl v2.8h, v2.8b, v6.8b
 ; CHECK-NEXT:    uaddl v4.8h, v0.8b, v4.8b
 ; CHECK-NEXT:    ushll v0.4s, v7.4h, #3
-; CHECK-NEXT:    ushll2 v1.4s, v7.8h, #3
-; CHECK-NEXT:    ushll v5.4s, v3.4h, #3
-; CHECK-NEXT:    ushll2 v6.4s, v3.8h, #3
-; CHECK-NEXT:    ushll2 v16.4s, v3.8h, #0
-; CHECK-NEXT:    ushll v17.4s, v3.4h, #0
-; CHECK-NEXT:    uaddw2 v1.4s, v1.4s, v2.8h
+; CHECK-NEXT:    ushll2 v3.4s, v7.8h, #3
+; CHECK-NEXT:    ushll v6.4s, v1.4h, #3
+; CHECK-NEXT:    ushll2 v16.4s, v1.8h, #3
+; CHECK-NEXT:    ushll2 v5.4s, v1.8h, #0
+; CHECK-NEXT:    ushll v17.4s, v1.4h, #0
+; CHECK-NEXT:    ushll2 v18.4s, v7.8h, #0
+; CHECK-NEXT:    uaddw2 v1.4s, v3.4s, v2.8h
 ; CHECK-NEXT:    uaddw v0.4s, v0.4s, v2.4h
-; CHECK-NEXT:    uaddw v2.4s, v5.4s, v4.4h
-; CHECK-NEXT:    uaddw2 v3.4s, v6.4s, v4.8h
-; CHECK-NEXT:    ushll2 v4.4s, v7.8h, #0
-; CHECK-NEXT:    ushll v5.4s, v7.4h, #0
-; CHECK-NEXT:    stp q17, q16, [x4, #32]
-; CHECK-NEXT:    stp q5, q4, [x4]
+; CHECK-NEXT:    uaddw v2.4s, v6.4s, v4.4h
+; CHECK-NEXT:    uaddw2 v3.4s, v16.4s, v4.8h
+; CHECK-NEXT:    ushll v4.4s, v7.4h, #0
+; CHECK-NEXT:    stp q17, q5, [x4, #32]
+; CHECK-NEXT:    stp q4, q18, [x4]
 ; CHECK-NEXT:    ret
   %lp1 = load <4 x i8>, ptr %p
   %p2 = getelementptr i8, ptr %p, i32 4
@@ -1157,32 +1158,32 @@ define <16 x i32> @extrause_ext2(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 define <16 x i32> @extrause_shl(ptr %p, ptr %q, ptr %r, ptr %s, ptr %z) {
 ; CHECK-LABEL: extrause_shl:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp s0, s1, [x0]
+; CHECK-NEXT:    ldp s1, s2, [x0]
 ; CHECK-NEXT:    add x10, x3, #12
-; CHECK-NEXT:    ldp s2, s3, [x2]
+; CHECK-NEXT:    ldp s0, s3, [x2]
 ; CHECK-NEXT:    add x11, x1, #12
 ; CHECK-NEXT:    ldp s4, s5, [x0, #8]
 ; CHECK-NEXT:    add x8, x3, #8
 ; CHECK-NEXT:    ldp s6, s7, [x2, #8]
 ; CHECK-NEXT:    add x9, x1, #8
-; CHECK-NEXT:    ld1 { v2.s }[1], [x3], #4
-; CHECK-NEXT:    ld1 { v0.s }[1], [x1], #4
-; CHECK-NEXT:    ld1 { v1.s }[1], [x1]
+; CHECK-NEXT:    ld1 { v0.s }[1], [x3], #4
+; CHECK-NEXT:    ld1 { v1.s }[1], [x1], #4
+; CHECK-NEXT:    ld1 { v2.s }[1], [x1]
 ; CHECK-NEXT:    ld1 { v5.s }[1], [x11]
 ; CHECK-NEXT:    ld1 { v3.s }[1], [x3]
 ; CHECK-NEXT:    ld1 { v7.s }[1], [x10]
 ; CHECK-NEXT:    ld1 { v4.s }[1], [x9]
 ; CHECK-NEXT:    ld1 { v6.s }[1], [x8]
-; CHECK-NEXT:    uaddl v1.8h, v1.8b, v5.8b
+; CHECK-NEXT:    uaddl v2.8h, v2.8b, v5.8b
 ; CHECK-NEXT:    uaddl v3.8h, v3.8b, v7.8b
-; CHECK-NEXT:    uaddl v4.8h, v0.8b, v4.8b
-; CHECK-NEXT:    uaddl v2.8h, v2.8b, v6.8b
-; CHECK-NEXT:    ushll v5.4s, v1.4h, #3
+; CHECK-NEXT:    uaddl v4.8h, v1.8b, v4.8b
+; CHECK-NEXT:    ushll v5.4s, v2.4h, #3
+; CHECK-NEXT:    ushll2 v7.4s, v2.8h, #3
+; CHECK-NEXT:    uaddl v2.8h, v0.8b, v6.8b
 ; CHECK-NEXT:    ushll v6.4s, v3.4h, #3
-; CHECK-NEXT:    ushll2 v7.4s, v1.8h, #3
 ; CHECK-NEXT:    ushll2 v16.4s, v3.8h, #3
-; CHECK-NEXT:    uaddw v0.4s, v5.4s, v4.4h
 ; CHECK-NEXT:    uaddw2 v1.4s, v7.4s, v4.8h
+; CHECK-NEXT:    uaddw v0.4s, v5.4s, v4.4h
 ; CHECK-NEXT:    stp q5, q7, [x4]
 ; CHECK-NEXT:    uaddw2 v3.4s, v16.4s, v2.8h
 ; CHECK-NEXT:    uaddw v2.4s, v6.4s, v2.4h
diff --git a/llvm/test/CodeGen/AArch64/fcmp.ll b/llvm/test/CodeGen/AArch64/fcmp.ll
index 9916aeeab1cad..b1ca88975a621 100644
--- a/llvm/test/CodeGen/AArch64/fcmp.ll
+++ b/llvm/test/CodeGen/AArch64/fcmp.ll
@@ -280,10 +280,10 @@ define <3 x i32> @v3f64_i32(<3 x double> %a, <3 x double> %b, <3 x i32> %d, <3 x
 ; CHECK-GI-NEXT:    uzp1 v0.4s, v0.4s, v2.4s
 ; CHECK-GI-NEXT:    fmov s2, w8
 ; CHECK-GI-NEXT:    mov v2.s[1], w8
-; CHECK-GI-NEXT:    neg v3.4s, v1.4s
 ; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v1.4s
+; CHECK-GI-NEXT:    neg v1.4s, v1.4s
 ; CHECK-GI-NEXT:    mov v2.s[2], w8
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v3.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-GI-NEXT:    eor v1.16b, v0.16b, v2.16b
 ; CHECK-GI-NEXT:    and v0.16b, v6.16b, v0.16b
 ; CHECK-GI-NEXT:    and v1.16b, v7.16b, v1.16b
@@ -348,10 +348,10 @@ define <3 x float> @v3f32_float(<3 x float> %a, <3 x float> %b, <3 x float> %d,
 ; CHECK-GI-NEXT:    mov w8, #-1 // =0xffffffff
 ; CHECK-GI-NEXT:    fmov s1, w8
 ; CHECK-GI-NEXT:    mov v1.s[1], w8
-; CHECK-GI-NEXT:    neg v5.4s, v4.4s
 ; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v4.4s
+; CHECK-GI-NEXT:    neg v4.4s, v4.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v4.4s
 ; CHECK-GI-NEXT:    mov v1.s[2], w8
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v5.4s
 ; CHECK-GI-NEXT:    eor v1.16b, v0.16b, v1.16b
 ; CHECK-GI-NEXT:    and v0.16b, v2.16b, v0.16b
 ; CHECK-GI-NEXT:    and v1.16b, v3.16b, v1.16b
@@ -426,10 +426,10 @@ define <3 x i32> @v3f32_i32(<3 x float> %a, <3 x float> %b, <3 x i32> %d, <3 x i
 ; CHECK-GI-NEXT:    mov w8, #-1 // =0xffffffff
 ; CHECK-GI-NEXT:    fmov s1, w8
 ; CHECK-GI-NEXT:    mov v1.s[1], w8
-; CHECK-GI-NEXT:    neg v5.4s, v4.4s
 ; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v4.4s
+; CHECK-GI-NEXT:    neg v4.4s, v4.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v4.4s
 ; CHECK-GI-NEXT:    mov v1.s[2], w8
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v5.4s
 ; CHECK-GI-NEXT:    eor v1.16b, v0.16b, v1.16b
 ; CHECK-GI-NEXT:    and v0.16b, v2.16b, v0.16b
 ; CHECK-GI-NEXT:    and v1.16b, v3.16b, v1.16b
@@ -545,8 +545,8 @@ define <7 x half> @v7f16_half(<7 x half> %a, <7 x half> %b, <7 x half> %d, <7 x
 ; CHECK-GI-NOFP16-LABEL: v7f16_half:
 ; CHECK-GI-NOFP16:       // %bb.0: // %entry
 ; CHECK-GI-NOFP16-NEXT:    mov w8, #15 // =0xf
-; CHECK-GI-NOFP16-NEXT:    mov h6, v0.h[4]
-; CHECK-GI-NOFP16-NEXT:    mov h7, v0.h[5]
+; CHECK-GI-NOFP16-NEXT:    mov h5, v0.h[4]
+; CHECK-GI-NOFP16-NEXT:    mov h6, v0.h[5]
 ; CHECK-GI-NOFP16-NEXT:    fmov s4, w8
 ; CHECK-GI-NOFP16-NEXT:    mov h16, v1.h[4]
 ; CHECK-GI-NOFP16-NEXT:    mov w8, #65535 // =0xffff
@@ -555,32 +555,32 @@ define <7 x half> @v7f16_half(<7 x half> %a, <7 x half> %b, <7 x half> %d, <7 x
 ; CHECK-GI-NOFP16-NEXT:    mov h19, v1.h[6]
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
-; CHECK-GI-NOFP16-NEXT:    mov v5.16b, v4.16b
-; CHECK-GI-NOFP16-NEXT:    mov v6.h[1], v7.h[0]
-; CHECK-GI-NOFP16-NEXT:    fmov s7, w8
+; CHECK-GI-NOFP16-NEXT:    mov v7.16b, v4.16b
+; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v6.h[0]
+; CHECK-GI-NOFP16-NEXT:    fmov s6, w8
 ; CHECK-GI-NOFP16-NEXT:    mov v16.h[1], v17.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.16b, v7.16b
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[1], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.16b, v6.16b
 ; CHECK-GI-NOFP16-NEXT:    fcmgt v0.4s, v1.4s, v0.4s
-; CHECK-GI-NOFP16-NEXT:    mov v6.h[2], v18.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[1], v7.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v18.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[1], v6.h[0]
 ; CHECK-GI-NOFP16-NEXT:    mov v16.h[2], v19.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v6.4h
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[2], v7.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[2], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    fcvtl v5.4s, v5.4h
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[2], v6.h[0]
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v16.4s, v16.4h
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[3], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[3], v7.h[0]
-; CHECK-GI-NOFP16-NEXT:    fcmgt v1.4s, v16.4s, v6.4s
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[4], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[4], v7.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[3], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[3], v6.h[0]
+; CHECK-GI-NOFP16-NEXT:    fcmgt v1.4s, v16.4s, v5.4s
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[4], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[4], v6.h[0]
 ; CHECK-GI-NOFP16-NEXT:    uzp1 v0.8h, v0.8h, v1.8h
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[5], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[5], v7.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v5.h[6], v4.h[0]
-; CHECK-GI-NOFP16-NEXT:    mov v17.h[6], v7.h[0]
-; CHECK-GI-NOFP16-NEXT:    neg v1.8h, v5.8h
-; CHECK-GI-NOFP16-NEXT:    ushl v0.8h, v0.8h, v5.8h
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[5], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[5], v6.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v7.h[6], v4.h[0]
+; CHECK-GI-NOFP16-NEXT:    mov v17.h[6], v6.h[0]
+; CHECK-GI-NOFP16-NEXT:    ushl v0.8h, v0.8h, v7.8h
+; CHECK-GI-NOFP16-NEXT:    neg v1.8h, v7.8h
 ; CHECK-GI-NOFP16-NEXT:    sshl v0.8h, v0.8h, v1.8h
 ; CHECK-GI-NOFP16-NEXT:    eor v1.16b, v0.16b, v17.16b
 ; CHECK-GI-NOFP16-NEXT:    and v0.16b, v2.16b, v0.16b
@@ -609,8 +609,8 @@ define <7 x half> @v7f16_half(<7 x half> %a, <7 x half> %b, <7 x half> %d, <7 x
 ; CHECK-GI-FP16-NEXT:    mov v7.h[5], v6.h[0]
 ; CHECK-GI-FP16-NEXT:    mov v5.h[6], v4.h[0]
 ; CHECK-GI-FP16-NEXT:    mov v7.h[6], v6.h[0]
-; CHECK-GI-FP16-NEXT:    neg v1.8h, v5.8h
 ; CHECK-GI-FP16-NEXT:    ushl v0.8h, v0.8h, v5.8h
+; CHECK-GI-FP16-NEXT:    neg v1.8h, v5.8h
 ; CHECK-GI-FP16-NEXT:    sshl v0.8h, v0.8h, v1.8h
 ; CHECK-GI-FP16-NEXT:    eor v1.16b, v0.16b, v7.16b
 ; CHECK-GI-FP16-NEXT:    and v0.16b, v2.16b, v0.16b
@@ -1047,6 +1047,7 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
 ; CHECK-GI-NOFP16-NEXT:    mov h6, v0.h[6]
 ; CHECK-GI-NOFP16-NEXT:    mov h7, v1.h[6]
 ; CHECK-GI-NOFP16-NEXT:    fmov s16, w0
+; CHECK-GI-NOFP16-NEXT:    fmov s18, w4
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
 ; CHECK-GI-NOFP16-NEXT:    mov v2.h[1], v3.h[0]
@@ -1054,6 +1055,7 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
 ; CHECK-GI-NOFP16-NEXT:    mov v4.h[1], v5.h[0]
 ; CHECK-GI-NOFP16-NEXT:    ldr s5, [sp]
 ; CHECK-GI-NOFP16-NEXT:    mov v16.s[1], w1
+; CHECK-GI-NOFP16-NEXT:    mov v18.s[1], w5
 ; CHECK-GI-NOFP16-NEXT:    mov v3.s[1], w8
 ; CHECK-GI-NOFP16-NEXT:    fmov w9, s5
 ; CHECK-GI-NOFP16-NEXT:    fmov s5, w7
@@ -1069,27 +1071,25 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
 ; CHECK-GI-NOFP16-NEXT:    mov v3.s[2], w8
 ; CHECK-GI-NOFP16-NEXT:    mov w8, #-1 // =0xffffffff
 ; CHECK-GI-NOFP16-NEXT:    mov v7.s[1], v17.s[0]
-; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v2.4h
 ; CHECK-GI-NOFP16-NEXT:    ldr s17, [sp, #40]
+; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v2.4h
+; CHECK-GI-NOFP16-NEXT:    mov v18.s[2], w6
 ; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v4.4h
 ; CHECK-GI-NOFP16-NEXT:    mov v16.s[3], w3
 ; CHECK-GI-NOFP16-NEXT:    mov v5.s[2], w9
-; CHECK-GI-NOFP16-NEXT:    neg v18.4s, v3.4s
 ; CHECK-GI-NOFP16-NEXT:    mov v7.s[2], v17.s[0]
 ; CHECK-GI-NOFP16-NEXT:    fcmgt v2.4s, v4.4s, v2.4s
 ; CHECK-GI-NOFP16-NEXT:    fmov s4, w8
 ; CHECK-GI-NOFP16-NEXT:    mov v4.s[1], w8
 ; CHECK-GI-NOFP16-NEXT:    ushl v2.4s, v2.4s, v3.4s
-; CHECK-GI-NOFP16-NEXT:    fmov s3, w4
-; CHECK-GI-NOFP16-NEXT:    mov v3.s[1], w5
+; CHECK-GI-NOFP16-NEXT:    neg v3.4s, v3.4s
 ; CHECK-GI-NOFP16-NEXT:    mov v4.s[2], w8
-; CHECK-GI-NOFP16-NEXT:    sshl v2.4s, v2.4s, v18.4s
+; CHECK-GI-NOFP16-NEXT:    sshl v2.4s, v2.4s, v3.4s
 ; CHECK-GI-NOFP16-NEXT:    fmov w8, s6
-; CHECK-GI-NOFP16-NEXT:    mov v3.s[2], w6
-; CHECK-GI-NOFP16-NEXT:    eor v1.16b, v2.16b, v4.16b
 ; CHECK-GI-NOFP16-NEXT:    mov v5.s[3], w8
+; CHECK-GI-NOFP16-NEXT:    eor v1.16b, v2.16b, v4.16b
+; CHECK-GI-NOFP16-NEXT:    and v2.16b, v18.16b, v2.16b
 ; CHECK-GI-NOFP16-NEXT:    and v1.16b, v7.16b, v1.16b
-; CHECK-GI-NOFP16-NEXT:    and v2.16b, v3.16b, v2.16b
 ; CHECK-GI-NOFP16-NEXT:    bsl v0.16b, v16.16b, v5.16b
 ; CHECK-GI-NOFP16-NEXT:    orr v1.16b, v2.16b, v1.16b
 ; CHECK-GI-NOFP16-NEXT:    mov s2, v0.s[1]
@@ -1111,30 +1111,32 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
 ; CHECK-GI-FP16-NEXT:    fcmgt v0.8h, v1.8h, v0.8h
 ; CHECK-GI-FP16-NEXT:    mov w10, #31 // =0x1f
 ; CHECK-GI-FP16-NEXT:    ldr s3, [sp]
-; CHECK-GI-FP16-NEXT:    fmov s1, w10
+; CHECK-GI-FP16-NEXT:    fmov s2, w10
 ; CHECK-GI-FP16-NEXT:    fmov s6, w0
 ; CHECK-GI-FP16-NEXT:    ldr s4, [sp, #8]
+; CHECK-GI-FP16-NEXT:    fmov s17, w4
 ; CHECK-GI-FP16-NEXT:    ldr s7, [sp, #24]
 ; CHECK-GI-FP16-NEXT:    ldr s16, [sp, #32]
 ; CHECK-GI-FP16-NEXT:    umov w8, v0.h[4]
 ; CHECK-GI-FP16-NEXT:    umov w9, v0.h[5]
-; CHECK-GI-FP16-NEXT:    mov v1.s[1], w10
+; CHECK-GI-FP16-NEXT:    mov v2.s[1], w10
 ; CHECK-GI-FP16-NEXT:    mov v6.s[1], w1
+; CHECK-GI-FP16-NEXT:    mov v17.s[1], w5
 ; CHECK-GI-FP16-NEXT:    mov v7.s[1], v16.s[0]
 ; CHECK-GI-FP16-NEXT:    ldr s16, [sp, #40]
-; CHECK-GI-FP16-NEXT:    fmov s2, w8
+; CHECK-GI-FP16-NEXT:    fmov s1, w8
 ; CHECK-GI-FP16-NEXT:    umov w8, v0.h[6]
-; CHECK-GI-FP16-NEXT:    mov v1.s[2], w10
+; CHECK-GI-FP16-NEXT:    mov v2.s[2], w10
 ; CHECK-GI-FP16-NEXT:    ushll v0.4s, v0.4h, #0
 ; CHECK-GI-FP16-NEXT:    mov v6.s[2], w2
+; CHECK-GI-FP16-NEXT:    mov v17.s[2], w6
 ; CHECK-GI-FP16-NEXT:    mov v7.s[2], v16.s[0]
-; CHECK-GI-FP16-NEXT:    mov v2.s[1], w9
+; CHECK-GI-FP16-NEXT:    mov v1.s[1], w9
 ; CHECK-GI-FP16-NEXT:    mov w9, #-1 // =0xffffffff
 ; CHECK-GI-FP16-NEXT:    fmov s5, w9
-; CHECK-GI-FP16-NEXT:    neg v17.4s, v1.4s
 ; CHECK-GI-FP16-NEXT:    shl v0.4s, v0.4s, #31
 ; CHECK-GI-FP16-NEXT:    mov v6.s[3], w3
-; CHECK-GI-FP16-NEXT:    mov v2.s[2], w8
+; CHECK-GI-FP16-NEXT:    mov v1.s[2], w8
 ; CHECK-GI-FP16-NEXT:    fmov w8, s3
 ; CHECK-GI-FP16-NEXT:    fmov s3, w7
 ; CHECK-GI-FP16-NEXT:    mov v5.s[1], w9
@@ -1142,26 +1144,24 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
 ; CHECK-GI-FP16-NEXT:    mov v3.s[1], w8
 ; CHECK-GI-FP16-NEXT:    fmov w8, s4
 ; CHECK-GI-FP16-NEXT:    ldr s4, [sp, #16]
-; CHECK-GI-FP16-NEXT:    ushl v1.4s, v2.4s, v1.4s
-; CHECK-GI-FP16-NEXT:    fmov s2, w4
+; CHECK-GI-FP16-NEXT:    ushl v1.4s, v1.4s, v2.4s
+; CHECK-GI-FP16-NEXT:    neg v2.4s, v2.4s
 ; CHECK-GI-FP16-NEXT:    mov v5.s[2], w9
-; CHECK-GI-FP16-NEXT:    mov v2.s[1], w5
 ; CHECK-GI-FP16-NEXT:    mov v3.s[2], w8
-; CHECK-GI-FP16-NEXT:    sshl v1.4s, v1.4s, v17.4s
+; CHECK-GI-FP16-NEXT:    sshl v1.4s, v1.4s, v2.4s
 ; CHECK-GI-FP16-NEXT:    fmov w8, s4
-; CHECK-GI-FP16-NEXT:    eor v4.16b, v1.16b, v5.16b
-; CHECK-GI-FP16-NEXT:    mov v2.s[2], w6
+; CHECK-GI-FP16-NEXT:    eor v2.16b, v1.16b, v5.16b
+; CHECK-GI-FP16-NEXT:    and v1.16b, v17.16b, v1.16b
 ; CHECK-GI-FP16-NEXT:    mov v3.s[3], w8
-; CHECK-GI-FP16-NEXT:    and v1.16b, v2.16b, v1.16b
-; CHECK-GI-FP16-NEXT:    and v2.16b, v7.16b, v4.16b
+; CHECK-GI-FP16-NEXT:    and v2.16b, v7.16b, v2.16b
 ; CHECK-GI-FP16-NEXT:    bsl v0.16b, v6.16b, v3.16b
 ; CHECK-GI-FP16-NEXT:    orr v1.16b, v1.16b, v2.16b
 ; CHECK-GI-FP16-NEXT:    mov s2, v0.s[1]
 ; CHECK-GI-FP16-NEXT:    mov s3, v0.s[2]
 ; CHECK-GI-FP16-NEXT:    mov s4, v0.s[3]
-; CHECK-GI-FP16-NEXT:    fmov w0, s0
 ; CHECK-GI-FP16-NEXT:    mov s5, v1.s[1]
 ; CHECK-GI-FP16-NEXT:    mov s6, v1.s[2]
+; CHECK-GI-FP16-NEXT:    fmov w0, s0
 ; CHECK-GI-FP16-NEXT:    fmov w4, s1
 ; CHECK-GI-FP16-NEXT:    fmov w1, s2
 ; CHECK-GI-FP16-NEXT:    fmov w2, s3
diff --git a/llvm/test/CodeGen/AArch64/fdiv-combine.ll b/llvm/test/CodeGen/AArch64/fdiv-combine.ll
index 1ed63f3ef2507..0627250d07791 100644
--- a/llvm/test/CodeGen/AArch64/fdiv-combine.ll
+++ b/llvm/test/CodeGen/AArch64/fdiv-combine.ll
@@ -171,8 +171,8 @@ entry:
 define <vscale x 2 x double> @splat_fdiv_nxv2f64(double %D, <vscale x 2 x double> %a) #1 {
 ; CHECK-LABEL: splat_fdiv_nxv2f64:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, d0
 ; CHECK-NEXT:    fdivr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/fold-int-pow2-with-fmul-or-fdiv.ll b/llvm/test/CodeGen/AArch64/fold-int-pow2-with-fmul-or-fdiv.ll
index 03e64f8b785b0..a78addc490086 100644
--- a/llvm/test/CodeGen/AArch64/fold-int-pow2-with-fmul-or-fdiv.ll
+++ b/llvm/test/CodeGen/AArch64/fold-int-pow2-with-fmul-or-fdiv.ll
@@ -604,8 +604,8 @@ define fastcc i1 @quantum_hadamard(i32 %0) {
 define <vscale x 4 x float> @fdiv_pow2_nx4xfloat(<vscale x 4 x i32> %i) "target-features"="+sve" {
 ; CHECK-LABEL: fdiv_pow2_nx4xfloat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    lslr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fmov z1.s, #9.00000000
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
diff --git a/llvm/test/CodeGen/AArch64/fp-veclib-expansion.ll b/llvm/test/CodeGen/AArch64/fp-veclib-expansion.ll
index 67c056c780cc8..2c8e2190f8209 100644
--- a/llvm/test/CodeGen/AArch64/fp-veclib-expansion.ll
+++ b/llvm/test/CodeGen/AArch64/fp-veclib-expansion.ll
@@ -62,9 +62,9 @@ define <vscale x 4 x float> @frem_nxv4f32(<vscale x 4 x float> %unused, <vscale
 ; ARMPL-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
 ; ARMPL-NEXT:    .cfi_def_cfa_offset 16
 ; ARMPL-NEXT:    .cfi_offset w30, -16
-; ARMPL-NEXT:    ptrue p0.s
 ; ARMPL-NEXT:    mov z0.d, z1.d
 ; ARMPL-NEXT:    mov z1.d, z2.d
+; ARMPL-NEXT:    ptrue p0.s
 ; ARMPL-NEXT:    bl armpl_svfmod_f32_x
 ; ARMPL-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
 ; ARMPL-NEXT:    ret
@@ -74,9 +74,9 @@ define <vscale x 4 x float> @frem_nxv4f32(<vscale x 4 x float> %unused, <vscale
 ; SLEEF-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
 ; SLEEF-NEXT:    .cfi_def_cfa_offset 16
 ; SLEEF-NEXT:    .cfi_offset w30, -16
-; SLEEF-NEXT:    ptrue p0.s
 ; SLEEF-NEXT:    mov z0.d, z1.d
 ; SLEEF-NEXT:    mov z1.d, z2.d
+; SLEEF-NEXT:    ptrue p0.s
 ; SLEEF-NEXT:    bl _ZGVsMxvv_fmodf
 ; SLEEF-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
 ; SLEEF-NEXT:    ret
@@ -90,9 +90,9 @@ define <vscale x 2 x double> @frem_strict_nxv2f64(<vscale x 2 x double> %unused,
 ; ARMPL-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
 ; ARMPL-NEXT:    .cfi_def_cfa_offset 16
 ; ARMPL-NEXT:    .cfi_offset w30, -16
-; ARMPL-NEXT:    ptrue p0.d
 ; ARMPL-NEXT:    mov z0.d, z1.d
 ; ARMPL-NEXT:    mov z1.d, z2.d
+; ARMPL-NEXT:    ptrue p0.d
 ; ARMPL-NEXT:    bl armpl_svfmod_f64_x
 ; ARMPL-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
 ; ARMPL-NEXT:    ret
@@ -102,9 +102,9 @@ define <vscale x 2 x double> @frem_strict_nxv2f64(<vscale x 2 x double> %unused,
 ; SLEEF-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
 ; SLEEF-NEXT:    .cfi_def_cfa_offset 16
 ; SLEEF-NEXT:    .cfi_offset w30, -16
-; SLEEF-NEXT:    ptrue p0.d
 ; SLEEF-NEXT:    mov z0.d, z1.d
 ; SLEEF-NEXT:    mov z1.d, z2.d
+; SLEEF-NEXT:    ptrue p0.d
 ; SLEEF-NEXT:    bl _ZGVsMxvv_fmod
 ; SLEEF-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
 ; SLEEF-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll b/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
index 301d28fd7be56..2ea581359af6f 100644
--- a/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
+++ b/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
@@ -194,10 +194,10 @@ define <2 x i16> @ustest_f64i16(<2 x double> %x) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    fcvtzs v0.2d, v0.2d
 ; CHECK-NEXT:    movi d1, #0x00ffff0000ffff
+; CHECK-NEXT:    movi v2.2d, #0000000000000000
 ; CHECK-NEXT:    xtn v0.2s, v0.2d
 ; CHECK-NEXT:    smin v0.2s, v0.2s, v1.2s
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    smax v0.2s, v0.2s, v1.2s
+; CHECK-NEXT:    smax v0.2s, v0.2s, v2.2s
 ; CHECK-NEXT:    ret
 entry:
   %conv = fptosi <2 x double> %x to <2 x i32>
@@ -833,10 +833,10 @@ define <2 x i16> @ustest_f64i16_mm(<2 x double> %x) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    fcvtzs v0.2d, v0.2d
 ; CHECK-NEXT:    movi d1, #0x00ffff0000ffff
+; CHECK-NEXT:    movi v2.2d, #0000000000000000
 ; CHECK-NEXT:    xtn v0.2s, v0.2d
 ; CHECK-NEXT:    smin v0.2s, v0.2s, v1.2s
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    smax v0.2s, v0.2s, v1.2s
+; CHECK-NEXT:    smax v0.2s, v0.2s, v2.2s
 ; CHECK-NEXT:    ret
 entry:
   %conv = fptosi <2 x double> %x to <2 x i32>
diff --git a/llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll b/llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll
index 92fd3183393ea..c45885a38f159 100644
--- a/llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll
+++ b/llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll
@@ -697,9 +697,9 @@ define <2 x i1> @test_signed_v2f32_v2i1(<2 x float> %f) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-NEXT:    fcvtzs v0.2s, v0.2s
+; CHECK-NEXT:    movi v2.2d, #0xffffffffffffffff
 ; CHECK-NEXT:    smin v0.2s, v0.2s, v1.2s
-; CHECK-NEXT:    movi v1.2d, #0xffffffffffffffff
-; CHECK-NEXT:    smax v0.2s, v0.2s, v1.2s
+; CHECK-NEXT:    smax v0.2s, v0.2s, v2.2s
 ; CHECK-NEXT:    ret
     %x = call <2 x i1> @llvm.fptosi.sat.v2f32.v2i1(<2 x float> %f)
     ret <2 x i1> %x
@@ -1620,9 +1620,9 @@ define <4 x i1> @test_signed_v4f16_v4i1(<4 x half> %f) {
 ; CHECK-FP16:       // %bb.0:
 ; CHECK-FP16-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-FP16-NEXT:    fcvtzs v0.4h, v0.4h
+; CHECK-FP16-NEXT:    movi v2.2d, #0xffffffffffffffff
 ; CHECK-FP16-NEXT:    smin v0.4h, v0.4h, v1.4h
-; CHECK-FP16-NEXT:    movi v1.2d, #0xffffffffffffffff
-; CHECK-FP16-NEXT:    smax v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT:    smax v0.4h, v0.4h, v2.4h
 ; CHECK-FP16-NEXT:    ret
     %x = call <4 x i1> @llvm.fptosi.sat.v4f16.v4i1(<4 x half> %f)
     ret <4 x i1> %x
@@ -1668,9 +1668,9 @@ define <4 x i13> @test_signed_v4f16_v4i13(<4 x half> %f) {
 ; CHECK-FP16:       // %bb.0:
 ; CHECK-FP16-NEXT:    fcvtzs v0.4h, v0.4h
 ; CHECK-FP16-NEXT:    mvni v1.4h, #240, lsl #8
+; CHECK-FP16-NEXT:    movi v2.4h, #240, lsl #8
 ; CHECK-FP16-NEXT:    smin v0.4h, v0.4h, v1.4h
-; CHECK-FP16-NEXT:    movi v1.4h, #240, lsl #8
-; CHECK-FP16-NEXT:    smax v0.4h, v0.4h, v1.4h
+; CHECK-FP16-NEXT:    smax v0.4h, v0.4h, v2.4h
 ; CHECK-FP16-NEXT:    ret
     %x = call <4 x i13> @llvm.fptosi.sat.v4f16.v4i13(<4 x half> %f)
     ret <4 x i13> %x
@@ -2103,9 +2103,9 @@ define <8 x i1> @test_signed_v8f16_v8i1(<8 x half> %f) {
 ; CHECK-FP16:       // %bb.0:
 ; CHECK-FP16-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-FP16-NEXT:    fcvtzs v0.8h, v0.8h
+; CHECK-FP16-NEXT:    movi v2.2d, #0xffffffffffffffff
 ; CHECK-FP16-NEXT:    smin v0.8h, v0.8h, v1.8h
-; CHECK-FP16-NEXT:    movi v1.2d, #0xffffffffffffffff
-; CHECK-FP16-NEXT:    smax v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:    smax v0.8h, v0.8h, v2.8h
 ; CHECK-FP16-NEXT:    xtn v0.8b, v0.8h
 ; CHECK-FP16-NEXT:    ret
     %x = call <8 x i1> @llvm.fptosi.sat.v8f16.v8i1(<8 x half> %f)
@@ -2254,9 +2254,9 @@ define <8 x i13> @test_signed_v8f16_v8i13(<8 x half> %f) {
 ; CHECK-FP16:       // %bb.0:
 ; CHECK-FP16-NEXT:    fcvtzs v0.8h, v0.8h
 ; CHECK-FP16-NEXT:    mvni v1.8h, #240, lsl #8
+; CHECK-FP16-NEXT:    movi v2.8h, #240, lsl #8
 ; CHECK-FP16-NEXT:    smin v0.8h, v0.8h, v1.8h
-; CHECK-FP16-NEXT:    movi v1.8h, #240, lsl #8
-; CHECK-FP16-NEXT:    smax v0.8h, v0.8h, v1.8h
+; CHECK-FP16-NEXT:    smax v0.8h, v0.8h, v2.8h
 ; CHECK-FP16-NEXT:    ret
     %x = call <8 x i13> @llvm.fptosi.sat.v8f16.v8i13(<8 x half> %f)
     ret <8 x i13> %x
diff --git a/llvm/test/CodeGen/AArch64/funnel-shift-rot.ll b/llvm/test/CodeGen/AArch64/funnel-shift-rot.ll
index 181f2185893e4..d39c09524e1ad 100644
--- a/llvm/test/CodeGen/AArch64/funnel-shift-rot.ll
+++ b/llvm/test/CodeGen/AArch64/funnel-shift-rot.ll
@@ -78,9 +78,9 @@ define <4 x i32> @rotl_v4i32(<4 x i32> %x, <4 x i32> %z) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v2.4s, #31
 ; CHECK-NEXT:    neg v3.4s, v1.4s
+; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
 ; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    neg v2.4s, v2.4s
+; CHECK-NEXT:    neg v2.4s, v3.4s
 ; CHECK-NEXT:    ushl v1.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ushl v0.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    orr v0.16b, v1.16b, v0.16b
diff --git a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
index 97511639ec8cf..cb9f04a7fac48 100644
--- a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
+++ b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
@@ -202,8 +202,8 @@ define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; CHECK-LABEL: vec_4xi32_nonsplat_eq:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v1.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI13_0
+; CHECK-NEXT:    neg v1.4s, v1.4s
 ; CHECK-NEXT:    ldr q2, [x8, :lo12:.LCPI13_0]
 ; CHECK-NEXT:    ushl v1.4s, v2.4s, v1.4s
 ; CHECK-NEXT:    and v0.16b, v1.16b, v0.16b
diff --git a/llvm/test/CodeGen/AArch64/icmp.ll b/llvm/test/CodeGen/AArch64/icmp.ll
index e7352fe03d01a..8e10847e7aae3 100644
--- a/llvm/test/CodeGen/AArch64/icmp.ll
+++ b/llvm/test/CodeGen/AArch64/icmp.ll
@@ -179,10 +179,10 @@ define <3 x i32> @v3i32_i32(<3 x i32> %a, <3 x i32> %b, <3 x i32> %d, <3 x i32>
 ; CHECK-GI-NEXT:    mov w8, #-1 // =0xffffffff
 ; CHECK-GI-NEXT:    fmov s1, w8
 ; CHECK-GI-NEXT:    mov v1.s[1], w8
-; CHECK-GI-NEXT:    neg v5.4s, v4.4s
 ; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v4.4s
+; CHECK-GI-NEXT:    neg v4.4s, v4.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v4.4s
 ; CHECK-GI-NEXT:    mov v1.s[2], w8
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v5.4s
 ; CHECK-GI-NEXT:    eor v1.16b, v0.16b, v1.16b
 ; CHECK-GI-NEXT:    and v0.16b, v2.16b, v0.16b
 ; CHECK-GI-NEXT:    and v1.16b, v3.16b, v1.16b
diff --git a/llvm/test/CodeGen/AArch64/insert-extend.ll b/llvm/test/CodeGen/AArch64/insert-extend.ll
index e4d2b516b8fbf..0b730f6e77156 100644
--- a/llvm/test/CodeGen/AArch64/insert-extend.ll
+++ b/llvm/test/CodeGen/AArch64/insert-extend.ll
@@ -64,104 +64,104 @@ define i32 @large(ptr nocapture noundef readonly %p1, i32 noundef %st1, ptr noca
 ; CHECK-NEXT:    ldr d3, [x11]
 ; CHECK-NEXT:    ldr d4, [x10, x8]
 ; CHECK-NEXT:    ldr d5, [x11, x9]
+; CHECK-NEXT:    shll2 v6.4s, v0.8h, #16
 ; CHECK-NEXT:    usubl v2.8h, v2.8b, v3.8b
 ; CHECK-NEXT:    usubl v3.8h, v4.8b, v5.8b
-; CHECK-NEXT:    shll2 v4.4s, v0.8h, #16
-; CHECK-NEXT:    shll2 v5.4s, v1.8h, #16
-; CHECK-NEXT:    saddw v0.4s, v4.4s, v0.4h
+; CHECK-NEXT:    shll2 v4.4s, v1.8h, #16
+; CHECK-NEXT:    saddw v0.4s, v6.4s, v0.4h
 ; CHECK-NEXT:    shll2 v6.4s, v2.8h, #16
-; CHECK-NEXT:    saddw v1.4s, v5.4s, v1.4h
-; CHECK-NEXT:    shll2 v4.4s, v3.8h, #16
-; CHECK-NEXT:    saddw v2.4s, v6.4s, v2.4h
-; CHECK-NEXT:    saddw v3.4s, v4.4s, v3.4h
+; CHECK-NEXT:    shll2 v5.4s, v3.8h, #16
+; CHECK-NEXT:    saddw v1.4s, v4.4s, v1.4h
 ; CHECK-NEXT:    rev64 v4.4s, v0.4s
+; CHECK-NEXT:    saddw v2.4s, v6.4s, v2.4h
+; CHECK-NEXT:    saddw v3.4s, v5.4s, v3.4h
 ; CHECK-NEXT:    rev64 v5.4s, v1.4s
 ; CHECK-NEXT:    rev64 v6.4s, v2.4s
-; CHECK-NEXT:    rev64 v7.4s, v3.4s
 ; CHECK-NEXT:    sub v4.4s, v0.4s, v4.4s
 ; CHECK-NEXT:    addp v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    rev64 v7.4s, v3.4s
 ; CHECK-NEXT:    sub v5.4s, v1.4s, v5.4s
 ; CHECK-NEXT:    sub v6.4s, v2.4s, v6.4s
 ; CHECK-NEXT:    addp v2.4s, v3.4s, v2.4s
-; CHECK-NEXT:    sub v7.4s, v3.4s, v7.4s
 ; CHECK-NEXT:    zip1 v16.4s, v5.4s, v4.4s
-; CHECK-NEXT:    ext v1.16b, v2.16b, v2.16b, #8
+; CHECK-NEXT:    sub v7.4s, v3.4s, v7.4s
 ; CHECK-NEXT:    zip2 v3.4s, v6.4s, v7.4s
 ; CHECK-NEXT:    mov v6.s[1], v7.s[0]
 ; CHECK-NEXT:    ext v7.16b, v5.16b, v16.16b, #8
 ; CHECK-NEXT:    mov v5.s[3], v4.s[2]
-; CHECK-NEXT:    uzp1 v4.4s, v1.4s, v0.4s
-; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v0.4s
+; CHECK-NEXT:    ext v4.16b, v2.16b, v2.16b, #8
 ; CHECK-NEXT:    mov v6.d[1], v7.d[1]
 ; CHECK-NEXT:    mov v3.d[1], v5.d[1]
+; CHECK-NEXT:    uzp1 v1.4s, v4.4s, v0.4s
+; CHECK-NEXT:    uzp2 v4.4s, v4.4s, v0.4s
 ; CHECK-NEXT:    uzp2 v5.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    uzp1 v0.4s, v2.4s, v0.4s
-; CHECK-NEXT:    sub v1.4s, v4.4s, v1.4s
 ; CHECK-NEXT:    add v2.4s, v3.4s, v6.4s
 ; CHECK-NEXT:    sub v3.4s, v6.4s, v3.4s
+; CHECK-NEXT:    sub v1.4s, v1.4s, v4.4s
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
-; CHECK-NEXT:    rev64 v6.4s, v1.4s
 ; CHECK-NEXT:    rev64 v4.4s, v2.4s
 ; CHECK-NEXT:    rev64 v5.4s, v3.4s
-; CHECK-NEXT:    addp v16.4s, v1.4s, v3.4s
+; CHECK-NEXT:    rev64 v6.4s, v1.4s
 ; CHECK-NEXT:    rev64 v7.4s, v0.4s
+; CHECK-NEXT:    addp v16.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    addp v17.4s, v0.4s, v2.4s
-; CHECK-NEXT:    sub v1.4s, v1.4s, v6.4s
 ; CHECK-NEXT:    sub v3.4s, v3.4s, v5.4s
 ; CHECK-NEXT:    sub v2.4s, v2.4s, v4.4s
+; CHECK-NEXT:    sub v1.4s, v1.4s, v6.4s
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v7.4s
 ; CHECK-NEXT:    zip1 v18.4s, v17.4s, v17.4s
-; CHECK-NEXT:    ext v6.16b, v1.16b, v16.16b, #8
 ; CHECK-NEXT:    ext v4.16b, v17.16b, v2.16b, #4
 ; CHECK-NEXT:    ext v5.16b, v16.16b, v3.16b, #4
 ; CHECK-NEXT:    mov v20.16b, v3.16b
+; CHECK-NEXT:    ext v6.16b, v1.16b, v16.16b, #8
 ; CHECK-NEXT:    ext v7.16b, v0.16b, v17.16b, #4
 ; CHECK-NEXT:    mov v21.16b, v2.16b
 ; CHECK-NEXT:    trn2 v0.4s, v18.4s, v0.4s
-; CHECK-NEXT:    ext v19.16b, v6.16b, v1.16b, #4
-; CHECK-NEXT:    mov v1.s[2], v16.s[1]
 ; CHECK-NEXT:    mov v20.s[2], v16.s[3]
 ; CHECK-NEXT:    zip2 v4.4s, v4.4s, v17.4s
 ; CHECK-NEXT:    zip2 v5.4s, v5.4s, v16.4s
 ; CHECK-NEXT:    mov v21.s[2], v17.s[3]
+; CHECK-NEXT:    ext v19.16b, v6.16b, v1.16b, #4
 ; CHECK-NEXT:    ext v7.16b, v7.16b, v7.16b, #4
-; CHECK-NEXT:    mov v18.16b, v1.16b
+; CHECK-NEXT:    mov v1.s[2], v16.s[1]
 ; CHECK-NEXT:    ext v2.16b, v2.16b, v4.16b, #12
 ; CHECK-NEXT:    ext v3.16b, v3.16b, v5.16b, #12
 ; CHECK-NEXT:    uzp2 v4.4s, v6.4s, v19.4s
 ; CHECK-NEXT:    mov v5.16b, v7.16b
 ; CHECK-NEXT:    mov v6.16b, v20.16b
+; CHECK-NEXT:    mov v18.16b, v1.16b
 ; CHECK-NEXT:    mov v19.16b, v21.16b
-; CHECK-NEXT:    mov v18.s[1], v16.s[0]
 ; CHECK-NEXT:    sub v7.4s, v0.4s, v7.4s
 ; CHECK-NEXT:    mov v6.s[1], v16.s[2]
 ; CHECK-NEXT:    mov v5.s[0], v17.s[1]
+; CHECK-NEXT:    mov v18.s[1], v16.s[0]
 ; CHECK-NEXT:    mov v19.s[1], v17.s[2]
 ; CHECK-NEXT:    sub v1.4s, v1.4s, v4.4s
 ; CHECK-NEXT:    sub v16.4s, v20.4s, v3.4s
 ; CHECK-NEXT:    sub v17.4s, v21.4s, v2.4s
-; CHECK-NEXT:    add v4.4s, v18.4s, v4.4s
 ; CHECK-NEXT:    add v3.4s, v6.4s, v3.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v5.4s
+; CHECK-NEXT:    add v4.4s, v18.4s, v4.4s
 ; CHECK-NEXT:    add v2.4s, v19.4s, v2.4s
-; CHECK-NEXT:    mov v4.d[1], v1.d[1]
 ; CHECK-NEXT:    mov v3.d[1], v16.d[1]
 ; CHECK-NEXT:    mov v0.d[1], v7.d[1]
+; CHECK-NEXT:    mov v4.d[1], v1.d[1]
 ; CHECK-NEXT:    mov v2.d[1], v17.d[1]
-; CHECK-NEXT:    cmlt v6.8h, v4.8h, #0
 ; CHECK-NEXT:    cmlt v1.8h, v3.8h, #0
 ; CHECK-NEXT:    cmlt v5.8h, v0.8h, #0
+; CHECK-NEXT:    cmlt v6.8h, v4.8h, #0
 ; CHECK-NEXT:    cmlt v7.8h, v2.8h, #0
-; CHECK-NEXT:    add v4.4s, v6.4s, v4.4s
 ; CHECK-NEXT:    add v3.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
+; CHECK-NEXT:    add v4.4s, v6.4s, v4.4s
 ; CHECK-NEXT:    add v2.4s, v7.4s, v2.4s
 ; CHECK-NEXT:    eor v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    eor v0.16b, v0.16b, v5.16b
 ; CHECK-NEXT:    eor v2.16b, v2.16b, v7.16b
 ; CHECK-NEXT:    eor v3.16b, v4.16b, v6.16b
-; CHECK-NEXT:    eor v0.16b, v0.16b, v5.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v3.4s
+; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
diff --git a/llvm/test/CodeGen/AArch64/insert-subvector-res-legalization.ll b/llvm/test/CodeGen/AArch64/insert-subvector-res-legalization.ll
index 29f9c0336bbcc..542b2e90ffc15 100644
--- a/llvm/test/CodeGen/AArch64/insert-subvector-res-legalization.ll
+++ b/llvm/test/CodeGen/AArch64/insert-subvector-res-legalization.ll
@@ -9,9 +9,9 @@ define <vscale x 8 x i8> @vec_scalable_subvec_scalable_idx_zero_i8(ptr %a, ptr %
 ; CHECK-LABEL: vec_scalable_subvec_scalable_idx_zero_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ld1b { z0.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1b { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ld1b { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
 ; CHECK-NEXT:    uzp1 z0.h, z1.h, z0.h
 ; CHECK-NEXT:    ret
@@ -25,9 +25,9 @@ define <vscale x 8 x i8> @vec_scalable_subvec_scalable_idx_nonzero_i8(ptr %a, pt
 ; CHECK-LABEL: vec_scalable_subvec_scalable_idx_nonzero_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ld1b { z0.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1b { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ld1b { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -41,9 +41,9 @@ define <vscale x 4 x i16> @vec_scalable_subvec_scalable_idx_zero_i16(ptr %a, ptr
 ; CHECK-LABEL: vec_scalable_subvec_scalable_idx_zero_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ld1h { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
 ; CHECK-NEXT:    uzp1 z0.s, z1.s, z0.s
 ; CHECK-NEXT:    ret
@@ -57,9 +57,9 @@ define <vscale x 4 x i16> @vec_scalable_subvec_scalable_idx_nonzero_i16(ptr %a,
 ; CHECK-LABEL: vec_scalable_subvec_scalable_idx_nonzero_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ld1h { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -76,10 +76,10 @@ define <vscale x 8 x i8> @vec_scalable_subvec_fixed_idx_zero_i8(ptr %a, ptr %b)
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr d0, [x1]
-; CHECK-NEXT:    ptrue p1.h, vl8
-; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-NEXT:    ld1b { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    sel z0.h, p1, z0.h, z1.h
+; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
+; CHECK-NEXT:    ptrue p0.h, vl8
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 8 x i8>, ptr %a
   %subvec = load <8 x i8>, ptr %b
@@ -92,19 +92,19 @@ define <vscale x 8 x i8> @vec_scalable_subvec_fixed_idx_nonzero_i8(ptr %a, ptr %
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cnth x8
-; CHECK-NEXT:    ldr d0, [x1]
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ldr d1, [x1]
 ; CHECK-NEXT:    sub x8, x8, #8
 ; CHECK-NEXT:    mov w9, #8 // =0x8
 ; CHECK-NEXT:    cmp x8, #8
-; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
+; CHECK-NEXT:    ld1b { z0.h }, p0/z, [x0]
+; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #1
-; CHECK-NEXT:    ld1b { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    st1h { z1.h }, p0, [sp]
-; CHECK-NEXT:    str q0, [x9, x8]
+; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
+; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -120,10 +120,10 @@ define <vscale x 4 x i16> @vec_scalable_subvec_fixed_idx_zero_i16(ptr %a, ptr %b
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr d0, [x1]
-; CHECK-NEXT:    ptrue p1.s, vl4
-; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
 ; CHECK-NEXT:    ld1h { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
+; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 4 x i16>, ptr %a
   %subvec = load <4 x i16>, ptr %b
@@ -136,19 +136,19 @@ define <vscale x 4 x i16> @vec_scalable_subvec_fixed_idx_nonzero_i16(ptr %a, ptr
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cntw x8
-; CHECK-NEXT:    ldr d0, [x1]
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ldr d1, [x1]
 ; CHECK-NEXT:    sub x8, x8, #4
 ; CHECK-NEXT:    mov w9, #4 // =0x4
 ; CHECK-NEXT:    cmp x8, #4
-; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
+; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0]
+; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #2
-; CHECK-NEXT:    ld1h { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    st1w { z1.s }, p0, [sp]
-; CHECK-NEXT:    str q0, [x9, x8]
+; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
+; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -164,10 +164,10 @@ define <vscale x 2 x i32> @vec_scalable_subvec_fixed_idx_zero_i32(ptr %a, ptr %b
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [x1]
-; CHECK-NEXT:    ptrue p1.d, vl2
-; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
 ; CHECK-NEXT:    ld1w { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 2 x i32>, ptr %a
   %subvec = load <2 x i32>, ptr %b
@@ -180,19 +180,19 @@ define <vscale x 2 x i32> @vec_scalable_subvec_fixed_idx_nonzero_i32(ptr %a, ptr
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntd x8
-; CHECK-NEXT:    ldr d0, [x1]
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ldr d1, [x1]
 ; CHECK-NEXT:    sub x8, x8, #2
 ; CHECK-NEXT:    mov w9, #2 // =0x2
 ; CHECK-NEXT:    cmp x8, #2
-; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
+; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x0]
+; CHECK-NEXT:    ushll v1.2d, v1.2s, #0
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #3
-; CHECK-NEXT:    ld1w { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    st1d { z1.d }, p0, [sp]
-; CHECK-NEXT:    str q0, [x9, x8]
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/intrinsic-cttz-elts-sve.ll b/llvm/test/CodeGen/AArch64/intrinsic-cttz-elts-sve.ll
index 1a4ab6ab334a6..9bd2ed240810d 100644
--- a/llvm/test/CodeGen/AArch64/intrinsic-cttz-elts-sve.ll
+++ b/llvm/test/CodeGen/AArch64/intrinsic-cttz-elts-sve.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme < %s | FileCheck %s
 
@@ -31,28 +32,28 @@ define i32 @ctz_nxv32i1(<vscale x 32 x i1> %a) #0 {
 ; CHECK-NEXT:    neg x8, x8
 ; CHECK-NEXT:    punpklo p3.h, p1.b
 ; CHECK-NEXT:    rdvl x9, #2
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    rdvl x8, #-1
-; CHECK-NEXT:    punpkhi p1.h, p1.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    mov z2.h, w8
-; CHECK-NEXT:    inch z0.h, all, mul #4
+; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    mov z3.h, p2/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    ptrue p2.h
+; CHECK-NEXT:    inch z0.h, all, mul #4
 ; CHECK-NEXT:    mov z5.h, p3/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    add z1.h, z0.h, z1.h
-; CHECK-NEXT:    add z4.h, z0.h, z2.h
 ; CHECK-NEXT:    mov z6.h, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z7.h, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    add z1.h, z0.h, z1.h
+; CHECK-NEXT:    add z4.h, z0.h, z2.h
 ; CHECK-NEXT:    and z0.d, z0.d, z3.d
 ; CHECK-NEXT:    add z2.h, z1.h, z2.h
 ; CHECK-NEXT:    and z3.d, z4.d, z5.d
 ; CHECK-NEXT:    and z1.d, z1.d, z6.d
 ; CHECK-NEXT:    and z2.d, z2.d, z7.d
-; CHECK-NEXT:    umax z0.h, p2/m, z0.h, z3.h
-; CHECK-NEXT:    umax z1.h, p2/m, z1.h, z2.h
-; CHECK-NEXT:    umax z0.h, p2/m, z0.h, z1.h
-; CHECK-NEXT:    umaxv h0, p2, z0.h
+; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z3.h
+; CHECK-NEXT:    umax z1.h, p0/m, z1.h, z2.h
+; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    umaxv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    sub w8, w9, w8
 ; CHECK-NEXT:    and w0, w8, #0xffff
@@ -65,12 +66,12 @@ define i32 @ctz_nxv4i32(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: ctz_nxv4i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    index z1.s, #0, #-1
 ; CHECK-NEXT:    cntw x9
-; CHECK-NEXT:    incw z1.s
 ; CHECK-NEXT:    cmpne p1.s, p0/z, z0.s, #0
-; CHECK-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    and z0.d, z1.d, z0.d
+; CHECK-NEXT:    index z0.s, #0, #-1
+; CHECK-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    incw z0.s
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
 ; CHECK-NEXT:    and z0.s, z0.s, #0xff
 ; CHECK-NEXT:    umaxv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w8, s0
@@ -87,38 +88,38 @@ define i64 @vscale_4096(<vscale x 16 x i8> %a) #1 {
 ; CHECK-LABEL: vscale_4096:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    index z1.s, #0, #-1
 ; CHECK-NEXT:    cntw x8
-; CHECK-NEXT:    cnth x9
 ; CHECK-NEXT:    neg x8, x8
-; CHECK-NEXT:    mov z1.s, w8
-; CHECK-NEXT:    neg x8, x9
 ; CHECK-NEXT:    rdvl x9, #1
-; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
-; CHECK-NEXT:    index z0.s, #0, #-1
+; CHECK-NEXT:    mov z0.s, w8
+; CHECK-NEXT:    cnth x8
+; CHECK-NEXT:    neg x8, x8
+; CHECK-NEXT:    incw z1.s, all, mul #4
+; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
-; CHECK-NEXT:    incw z0.s, all, mul #4
-; CHECK-NEXT:    add z1.s, z0.s, z1.s
-; CHECK-NEXT:    add z5.s, z0.s, z2.s
+; CHECK-NEXT:    add z0.s, z1.s, z0.s
+; CHECK-NEXT:    add z4.s, z1.s, z2.s
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpkhi p3.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    add z2.s, z1.s, z2.s
-; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    mov z3.s, p2/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    ptrue p2.s
-; CHECK-NEXT:    mov z4.s, p3/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    add z2.s, z0.s, z2.s
+; CHECK-NEXT:    punpklo p1.h, p1.b
+; CHECK-NEXT:    mov z5.s, p3/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z6.s, p0/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z7.s, p1/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    and z1.d, z1.d, z3.d
-; CHECK-NEXT:    and z2.d, z2.d, z4.d
-; CHECK-NEXT:    and z3.d, z5.d, z6.d
-; CHECK-NEXT:    and z0.d, z0.d, z7.d
-; CHECK-NEXT:    umax z1.s, p2/m, z1.s, z2.s
-; CHECK-NEXT:    umax z0.s, p2/m, z0.s, z3.s
-; CHECK-NEXT:    umax z0.s, p2/m, z0.s, z1.s
-; CHECK-NEXT:    umaxv s0, p2, z0.s
+; CHECK-NEXT:    and z0.d, z0.d, z3.d
+; CHECK-NEXT:    and z2.d, z2.d, z5.d
+; CHECK-NEXT:    and z3.d, z4.d, z6.d
+; CHECK-NEXT:    and z1.d, z1.d, z7.d
+; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z2.s
+; CHECK-NEXT:    umax z1.s, p0/m, z1.s, z3.s
+; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    umaxv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    sub w0, w9, w8
 ; CHECK-NEXT:    ret
@@ -130,21 +131,21 @@ define i64 @vscale_4096_poison(<vscale x 16 x i8> %a) #1 {
 ; CHECK-LABEL: vscale_4096_poison:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    index z1.h, #0, #-1
 ; CHECK-NEXT:    cnth x8
-; CHECK-NEXT:    rdvl x9, #1
 ; CHECK-NEXT:    neg x8, x8
-; CHECK-NEXT:    mov z1.h, w8
+; CHECK-NEXT:    rdvl x9, #1
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
-; CHECK-NEXT:    index z0.h, #0, #-1
+; CHECK-NEXT:    mov z0.h, w8
+; CHECK-NEXT:    inch z1.h, all, mul #2
 ; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    inch z0.h, all, mul #2
-; CHECK-NEXT:    add z1.h, z0.h, z1.h
+; CHECK-NEXT:    add z0.h, z1.h, z0.h
 ; CHECK-NEXT:    mov z2.h, p1/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z3.h, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    and z1.d, z1.d, z2.d
-; CHECK-NEXT:    and z0.d, z0.d, z3.d
+; CHECK-NEXT:    and z0.d, z0.d, z2.d
+; CHECK-NEXT:    and z1.d, z1.d, z3.d
 ; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    umaxv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w8, s0
@@ -161,16 +162,16 @@ define i32 @ctz_nxv8i1_no_range(<vscale x 8 x i1> %a) {
 ; CHECK-LABEL: ctz_nxv8i1_no_range:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    index z0.s, #0, #-1
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    cntw x8
-; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    neg x8, x8
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    cnth x9
 ; CHECK-NEXT:    mov z1.s, w8
-; CHECK-NEXT:    incw z0.s, all, mul #2
 ; CHECK-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z3.s, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    incw z0.s, all, mul #2
 ; CHECK-NEXT:    add z1.s, z0.s, z1.s
 ; CHECK-NEXT:    and z0.d, z0.d, z2.d
 ; CHECK-NEXT:    and z1.d, z1.d, z3.d
@@ -212,8 +213,8 @@ define i32 @ctz_nxv16i1_poison(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %a) {
 define i32 @ctz_and_nxv16i1(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
 ; CHECK-LABEL: ctz_and_nxv16i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, z1.b
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    brkb p0.b, p1/z, p0.b
 ; CHECK-NEXT:    cntp x0, p0, p0.b
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
diff --git a/llvm/test/CodeGen/AArch64/itofp.ll b/llvm/test/CodeGen/AArch64/itofp.ll
index f5a7b5dc9f492..ae4ced258bb8e 100644
--- a/llvm/test/CodeGen/AArch64/itofp.ll
+++ b/llvm/test/CodeGen/AArch64/itofp.ll
@@ -6555,18 +6555,18 @@ define <3 x bfloat> @stofp_v3i64_v3bf16(<3 x i64> %a) {
 ; CHECK-NEXT:    // kill: def $d2 killed $d2 def $q2
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    scvtf v1.2d, v2.2d
-; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    scvtf v0.2d, v0.2d
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    fcmeq v2.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    bit v0.16b, v1.16b, v2.16b
+; CHECK-NEXT:    add v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
 ; CHECK-NEXT:    shrn v0.4h, v0.4s, #16
 ; CHECK-NEXT:    ret
 entry:
@@ -6582,18 +6582,18 @@ define <3 x bfloat> @utofp_v3i64_v3bf16(<3 x i64> %a) {
 ; CHECK-NEXT:    // kill: def $d2 killed $d2 def $q2
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    ucvtf v1.2d, v2.2d
-; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    ucvtf v0.2d, v0.2d
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    fcmeq v2.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    bit v0.16b, v1.16b, v2.16b
+; CHECK-NEXT:    add v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
 ; CHECK-NEXT:    shrn v0.4h, v0.4s, #16
 ; CHECK-NEXT:    ret
 entry:
@@ -6606,17 +6606,17 @@ define <4 x bfloat> @stofp_v4i64_v4bf16(<4 x i64> %a) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    scvtf v0.2d, v0.2d
 ; CHECK-NEXT:    scvtf v1.2d, v1.2d
-; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    fcmeq v2.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    bit v0.16b, v1.16b, v2.16b
+; CHECK-NEXT:    add v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
 ; CHECK-NEXT:    shrn v0.4h, v0.4s, #16
 ; CHECK-NEXT:    ret
 entry:
@@ -6629,17 +6629,17 @@ define <4 x bfloat> @utofp_v4i64_v4bf16(<4 x i64> %a) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ucvtf v0.2d, v0.2d
 ; CHECK-NEXT:    ucvtf v1.2d, v1.2d
-; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v2.16b, v3.16b, v2.16b
-; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    fcmeq v2.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    bit v0.16b, v1.16b, v2.16b
+; CHECK-NEXT:    add v1.4s, v1.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
 ; CHECK-NEXT:    shrn v0.4h, v0.4s, #16
 ; CHECK-NEXT:    ret
 entry:
@@ -6658,22 +6658,22 @@ define <8 x bfloat> @stofp_v8i64_v8bf16(<8 x i64> %a) {
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    movi v3.4s, #1
+; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v4.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v5.4s, v0.4s, #16
-; CHECK-NEXT:    add v6.4s, v2.4s, v1.4s
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v4.16b, v4.16b, v3.16b
-; CHECK-NEXT:    and v3.16b, v5.16b, v3.16b
+; CHECK-NEXT:    add v6.4s, v2.4s, v3.4s
+; CHECK-NEXT:    add v3.4s, v0.4s, v3.4s
+; CHECK-NEXT:    and v4.16b, v4.16b, v1.16b
+; CHECK-NEXT:    and v1.16b, v5.16b, v1.16b
 ; CHECK-NEXT:    fcmeq v5.4s, v2.4s, v2.4s
 ; CHECK-NEXT:    orr v2.4s, #64, lsl #16
 ; CHECK-NEXT:    add v4.4s, v4.4s, v6.4s
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
-; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
+; CHECK-NEXT:    fcmeq v6.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
 ; CHECK-NEXT:    bit v2.16b, v4.16b, v5.16b
-; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
+; CHECK-NEXT:    bit v0.16b, v1.16b, v6.16b
 ; CHECK-NEXT:    uzp2 v0.8h, v0.8h, v2.8h
 ; CHECK-NEXT:    ret
 entry:
@@ -6692,22 +6692,22 @@ define <8 x bfloat> @utofp_v8i64_v8bf16(<8 x i64> %a) {
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
 ; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    movi v3.4s, #1
+; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v4.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v5.4s, v0.4s, #16
-; CHECK-NEXT:    add v6.4s, v2.4s, v1.4s
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
-; CHECK-NEXT:    and v4.16b, v4.16b, v3.16b
-; CHECK-NEXT:    and v3.16b, v5.16b, v3.16b
+; CHECK-NEXT:    add v6.4s, v2.4s, v3.4s
+; CHECK-NEXT:    add v3.4s, v0.4s, v3.4s
+; CHECK-NEXT:    and v4.16b, v4.16b, v1.16b
+; CHECK-NEXT:    and v1.16b, v5.16b, v1.16b
 ; CHECK-NEXT:    fcmeq v5.4s, v2.4s, v2.4s
 ; CHECK-NEXT:    orr v2.4s, #64, lsl #16
 ; CHECK-NEXT:    add v4.4s, v4.4s, v6.4s
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
-; CHECK-NEXT:    fcmeq v3.4s, v0.4s, v0.4s
+; CHECK-NEXT:    fcmeq v6.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
 ; CHECK-NEXT:    bit v2.16b, v4.16b, v5.16b
-; CHECK-NEXT:    bit v0.16b, v1.16b, v3.16b
+; CHECK-NEXT:    bit v0.16b, v1.16b, v6.16b
 ; CHECK-NEXT:    uzp2 v0.8h, v0.8h, v2.8h
 ; CHECK-NEXT:    ret
 entry:
@@ -6718,51 +6718,51 @@ entry:
 define <16 x bfloat> @stofp_v16i64_v16bf16(<16 x i64> %a) {
 ; CHECK-LABEL: stofp_v16i64_v16bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    scvtf v2.2d, v2.2d
 ; CHECK-NEXT:    scvtf v0.2d, v0.2d
+; CHECK-NEXT:    scvtf v2.2d, v2.2d
 ; CHECK-NEXT:    scvtf v6.2d, v6.2d
 ; CHECK-NEXT:    scvtf v4.2d, v4.2d
-; CHECK-NEXT:    scvtf v3.2d, v3.2d
 ; CHECK-NEXT:    scvtf v1.2d, v1.2d
+; CHECK-NEXT:    scvtf v3.2d, v3.2d
 ; CHECK-NEXT:    scvtf v7.2d, v7.2d
 ; CHECK-NEXT:    scvtf v5.2d, v5.2d
-; CHECK-NEXT:    fcvtn v2.2s, v2.2d
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
+; CHECK-NEXT:    fcvtn v2.2s, v2.2d
 ; CHECK-NEXT:    fcvtn v6.2s, v6.2d
 ; CHECK-NEXT:    fcvtn v4.2s, v4.2d
-; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v6.4s, v7.2d
 ; CHECK-NEXT:    fcvtn2 v4.4s, v5.2d
-; CHECK-NEXT:    movi v3.4s, #1
-; CHECK-NEXT:    ushr v5.4s, v2.4s, #16
+; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v7.4s, v0.4s, #16
-; CHECK-NEXT:    add v17.4s, v2.4s, v1.4s
-; CHECK-NEXT:    add v19.4s, v0.4s, v1.4s
+; CHECK-NEXT:    ushr v5.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v16.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v18.4s, v4.4s, #16
-; CHECK-NEXT:    add v20.4s, v6.4s, v1.4s
-; CHECK-NEXT:    add v1.4s, v4.4s, v1.4s
-; CHECK-NEXT:    and v5.16b, v5.16b, v3.16b
-; CHECK-NEXT:    and v7.16b, v7.16b, v3.16b
-; CHECK-NEXT:    and v16.16b, v16.16b, v3.16b
-; CHECK-NEXT:    and v3.16b, v18.16b, v3.16b
-; CHECK-NEXT:    fcmeq v18.4s, v0.4s, v0.4s
-; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    add v5.4s, v5.4s, v17.4s
+; CHECK-NEXT:    ushr v17.4s, v4.4s, #16
+; CHECK-NEXT:    add v19.4s, v0.4s, v3.4s
+; CHECK-NEXT:    add v18.4s, v2.4s, v3.4s
+; CHECK-NEXT:    add v20.4s, v6.4s, v3.4s
+; CHECK-NEXT:    add v3.4s, v4.4s, v3.4s
+; CHECK-NEXT:    and v7.16b, v7.16b, v1.16b
+; CHECK-NEXT:    and v5.16b, v5.16b, v1.16b
+; CHECK-NEXT:    and v16.16b, v16.16b, v1.16b
+; CHECK-NEXT:    and v1.16b, v17.16b, v1.16b
 ; CHECK-NEXT:    fcmeq v17.4s, v2.4s, v2.4s
+; CHECK-NEXT:    orr v2.4s, #64, lsl #16
 ; CHECK-NEXT:    add v7.4s, v7.4s, v19.4s
 ; CHECK-NEXT:    fcmeq v19.4s, v6.4s, v6.4s
-; CHECK-NEXT:    orr v2.4s, #64, lsl #16
-; CHECK-NEXT:    add v16.4s, v16.4s, v20.4s
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
+; CHECK-NEXT:    add v5.4s, v5.4s, v18.4s
+; CHECK-NEXT:    fcmeq v18.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    fcmeq v3.4s, v4.4s, v4.4s
+; CHECK-NEXT:    add v16.4s, v16.4s, v20.4s
+; CHECK-NEXT:    orr v0.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v6.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v4.4s, #64, lsl #16
 ; CHECK-NEXT:    bit v2.16b, v5.16b, v17.16b
-; CHECK-NEXT:    bit v0.16b, v7.16b, v18.16b
 ; CHECK-NEXT:    mov v5.16b, v19.16b
+; CHECK-NEXT:    bit v0.16b, v7.16b, v18.16b
 ; CHECK-NEXT:    bif v1.16b, v4.16b, v3.16b
 ; CHECK-NEXT:    bsl v5.16b, v16.16b, v6.16b
 ; CHECK-NEXT:    uzp2 v0.8h, v0.8h, v2.8h
@@ -6776,51 +6776,51 @@ entry:
 define <16 x bfloat> @utofp_v16i64_v16bf16(<16 x i64> %a) {
 ; CHECK-LABEL: utofp_v16i64_v16bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ucvtf v2.2d, v2.2d
 ; CHECK-NEXT:    ucvtf v0.2d, v0.2d
+; CHECK-NEXT:    ucvtf v2.2d, v2.2d
 ; CHECK-NEXT:    ucvtf v6.2d, v6.2d
 ; CHECK-NEXT:    ucvtf v4.2d, v4.2d
-; CHECK-NEXT:    ucvtf v3.2d, v3.2d
 ; CHECK-NEXT:    ucvtf v1.2d, v1.2d
+; CHECK-NEXT:    ucvtf v3.2d, v3.2d
 ; CHECK-NEXT:    ucvtf v7.2d, v7.2d
 ; CHECK-NEXT:    ucvtf v5.2d, v5.2d
-; CHECK-NEXT:    fcvtn v2.2s, v2.2d
 ; CHECK-NEXT:    fcvtn v0.2s, v0.2d
+; CHECK-NEXT:    fcvtn v2.2s, v2.2d
 ; CHECK-NEXT:    fcvtn v6.2s, v6.2d
 ; CHECK-NEXT:    fcvtn v4.2s, v4.2d
-; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v0.4s, v1.2d
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
+; CHECK-NEXT:    fcvtn2 v2.4s, v3.2d
 ; CHECK-NEXT:    fcvtn2 v6.4s, v7.2d
 ; CHECK-NEXT:    fcvtn2 v4.4s, v5.2d
-; CHECK-NEXT:    movi v3.4s, #1
-; CHECK-NEXT:    ushr v5.4s, v2.4s, #16
+; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v7.4s, v0.4s, #16
-; CHECK-NEXT:    add v17.4s, v2.4s, v1.4s
-; CHECK-NEXT:    add v19.4s, v0.4s, v1.4s
+; CHECK-NEXT:    ushr v5.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v16.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v18.4s, v4.4s, #16
-; CHECK-NEXT:    add v20.4s, v6.4s, v1.4s
-; CHECK-NEXT:    add v1.4s, v4.4s, v1.4s
-; CHECK-NEXT:    and v5.16b, v5.16b, v3.16b
-; CHECK-NEXT:    and v7.16b, v7.16b, v3.16b
-; CHECK-NEXT:    and v16.16b, v16.16b, v3.16b
-; CHECK-NEXT:    and v3.16b, v18.16b, v3.16b
-; CHECK-NEXT:    fcmeq v18.4s, v0.4s, v0.4s
-; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    add v5.4s, v5.4s, v17.4s
+; CHECK-NEXT:    ushr v17.4s, v4.4s, #16
+; CHECK-NEXT:    add v19.4s, v0.4s, v3.4s
+; CHECK-NEXT:    add v18.4s, v2.4s, v3.4s
+; CHECK-NEXT:    add v20.4s, v6.4s, v3.4s
+; CHECK-NEXT:    add v3.4s, v4.4s, v3.4s
+; CHECK-NEXT:    and v7.16b, v7.16b, v1.16b
+; CHECK-NEXT:    and v5.16b, v5.16b, v1.16b
+; CHECK-NEXT:    and v16.16b, v16.16b, v1.16b
+; CHECK-NEXT:    and v1.16b, v17.16b, v1.16b
 ; CHECK-NEXT:    fcmeq v17.4s, v2.4s, v2.4s
+; CHECK-NEXT:    orr v2.4s, #64, lsl #16
 ; CHECK-NEXT:    add v7.4s, v7.4s, v19.4s
 ; CHECK-NEXT:    fcmeq v19.4s, v6.4s, v6.4s
-; CHECK-NEXT:    orr v2.4s, #64, lsl #16
-; CHECK-NEXT:    add v16.4s, v16.4s, v20.4s
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
+; CHECK-NEXT:    add v5.4s, v5.4s, v18.4s
+; CHECK-NEXT:    fcmeq v18.4s, v0.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    fcmeq v3.4s, v4.4s, v4.4s
+; CHECK-NEXT:    add v16.4s, v16.4s, v20.4s
+; CHECK-NEXT:    orr v0.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v6.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v4.4s, #64, lsl #16
 ; CHECK-NEXT:    bit v2.16b, v5.16b, v17.16b
-; CHECK-NEXT:    bit v0.16b, v7.16b, v18.16b
 ; CHECK-NEXT:    mov v5.16b, v19.16b
+; CHECK-NEXT:    bit v0.16b, v7.16b, v18.16b
 ; CHECK-NEXT:    bif v1.16b, v4.16b, v3.16b
 ; CHECK-NEXT:    bsl v5.16b, v16.16b, v6.16b
 ; CHECK-NEXT:    uzp2 v0.8h, v0.8h, v2.8h
@@ -6834,107 +6834,107 @@ entry:
 define <32 x bfloat> @stofp_v32i64_v32bf16(<32 x i64> %a) {
 ; CHECK-LABEL: stofp_v32i64_v32bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    scvtf v16.2d, v2.2d
-; CHECK-NEXT:    scvtf v17.2d, v0.2d
-; CHECK-NEXT:    scvtf v18.2d, v3.2d
-; CHECK-NEXT:    scvtf v19.2d, v6.2d
-; CHECK-NEXT:    ldp q24, q23, [sp, #96]
-; CHECK-NEXT:    scvtf v21.2d, v1.2d
-; CHECK-NEXT:    scvtf v22.2d, v4.2d
+; CHECK-NEXT:    scvtf v17.2d, v2.2d
+; CHECK-NEXT:    scvtf v18.2d, v0.2d
+; CHECK-NEXT:    scvtf v19.2d, v3.2d
+; CHECK-NEXT:    scvtf v3.2d, v6.2d
+; CHECK-NEXT:    ldp q21, q20, [sp, #32]
+; CHECK-NEXT:    scvtf v4.2d, v4.2d
 ; CHECK-NEXT:    scvtf v6.2d, v7.2d
-; CHECK-NEXT:    scvtf v7.2d, v5.2d
-; CHECK-NEXT:    movi v3.4s, #127, msl #8
-; CHECK-NEXT:    movi v2.4s, #1
-; CHECK-NEXT:    fcvtn v0.2s, v16.2d
-; CHECK-NEXT:    ldp q20, q16, [sp, #32]
-; CHECK-NEXT:    fcvtn v1.2s, v17.2d
-; CHECK-NEXT:    ldp q5, q17, [sp]
-; CHECK-NEXT:    fcvtn v4.2s, v19.2d
-; CHECK-NEXT:    scvtf v23.2d, v23.2d
+; CHECK-NEXT:    scvtf v5.2d, v5.2d
+; CHECK-NEXT:    ldp q24, q23, [sp, #64]
+; CHECK-NEXT:    movi v16.4s, #1
+; CHECK-NEXT:    fcvtn v0.2s, v17.2d
+; CHECK-NEXT:    scvtf v17.2d, v1.2d
+; CHECK-NEXT:    fcvtn v1.2s, v18.2d
+; CHECK-NEXT:    fcvtn v3.2s, v3.2d
+; CHECK-NEXT:    ldp q18, q7, [sp]
+; CHECK-NEXT:    scvtf v21.2d, v21.2d
+; CHECK-NEXT:    fcvtn v4.2s, v4.2d
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    scvtf v20.2d, v20.2d
-; CHECK-NEXT:    scvtf v16.2d, v16.2d
-; CHECK-NEXT:    fcvtn2 v0.4s, v18.2d
-; CHECK-NEXT:    ldp q19, q18, [sp, #64]
-; CHECK-NEXT:    scvtf v25.2d, v5.2d
-; CHECK-NEXT:    fcvtn v5.2s, v22.2d
-; CHECK-NEXT:    fcvtn2 v1.4s, v21.2d
-; CHECK-NEXT:    scvtf v21.2d, v24.2d
-; CHECK-NEXT:    scvtf v17.2d, v17.2d
-; CHECK-NEXT:    fcvtn2 v4.4s, v6.2d
+; CHECK-NEXT:    fcvtn2 v0.4s, v19.2d
+; CHECK-NEXT:    ldp q22, q19, [sp, #96]
+; CHECK-NEXT:    fcvtn2 v1.4s, v17.2d
+; CHECK-NEXT:    fcvtn2 v3.4s, v6.2d
+; CHECK-NEXT:    scvtf v18.2d, v18.2d
+; CHECK-NEXT:    scvtf v17.2d, v24.2d
+; CHECK-NEXT:    fcvtn v6.2s, v21.2d
+; CHECK-NEXT:    fcvtn2 v4.4s, v5.2d
+; CHECK-NEXT:    scvtf v22.2d, v22.2d
+; CHECK-NEXT:    scvtf v21.2d, v23.2d
+; CHECK-NEXT:    scvtf v7.2d, v7.2d
+; CHECK-NEXT:    ushr v24.4s, v0.4s, #16
+; CHECK-NEXT:    add v5.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    scvtf v19.2d, v19.2d
-; CHECK-NEXT:    scvtf v6.2d, v18.2d
-; CHECK-NEXT:    fcvtn v18.2s, v20.2d
-; CHECK-NEXT:    ushr v22.4s, v0.4s, #16
-; CHECK-NEXT:    add v20.4s, v0.4s, v3.4s
-; CHECK-NEXT:    fcvtn2 v5.4s, v7.2d
-; CHECK-NEXT:    fcvtn v24.2s, v25.2d
-; CHECK-NEXT:    ushr v7.4s, v1.4s, #16
-; CHECK-NEXT:    fcvtn v21.2s, v21.2d
-; CHECK-NEXT:    add v26.4s, v1.4s, v3.4s
-; CHECK-NEXT:    ushr v27.4s, v4.4s, #16
-; CHECK-NEXT:    fcvtn v19.2s, v19.2d
-; CHECK-NEXT:    fcvtn2 v18.4s, v16.2d
-; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
-; CHECK-NEXT:    and v7.16b, v7.16b, v2.16b
-; CHECK-NEXT:    fcmeq v25.4s, v0.4s, v0.4s
+; CHECK-NEXT:    ushr v23.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v3.4s, #16
+; CHECK-NEXT:    fcvtn v18.2s, v18.2d
+; CHECK-NEXT:    fcvtn2 v6.4s, v20.2d
+; CHECK-NEXT:    add v26.4s, v1.4s, v2.4s
+; CHECK-NEXT:    fcvtn v17.2s, v17.2d
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
+; CHECK-NEXT:    fcvtn v22.2s, v22.2d
+; CHECK-NEXT:    fcmeq v20.4s, v0.4s, v0.4s
+; CHECK-NEXT:    and v23.16b, v23.16b, v16.16b
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    ushr v28.4s, v5.4s, #16
-; CHECK-NEXT:    fcvtn2 v24.4s, v17.2d
-; CHECK-NEXT:    add v17.4s, v5.4s, v3.4s
-; CHECK-NEXT:    fcvtn2 v21.4s, v23.2d
-; CHECK-NEXT:    and v16.16b, v27.16b, v2.16b
-; CHECK-NEXT:    add v20.4s, v22.4s, v20.4s
-; CHECK-NEXT:    fcvtn2 v19.4s, v6.2d
-; CHECK-NEXT:    add v7.4s, v7.4s, v26.4s
-; CHECK-NEXT:    ushr v26.4s, v18.4s, #16
-; CHECK-NEXT:    and v23.16b, v28.16b, v2.16b
-; CHECK-NEXT:    add v22.4s, v4.4s, v3.4s
-; CHECK-NEXT:    fcmeq v6.4s, v1.4s, v1.4s
-; CHECK-NEXT:    ushr v27.4s, v24.4s, #16
-; CHECK-NEXT:    add v30.4s, v24.4s, v3.4s
+; CHECK-NEXT:    fcmeq v27.4s, v3.4s, v3.4s
+; CHECK-NEXT:    fcvtn2 v18.4s, v7.2d
+; CHECK-NEXT:    add v7.4s, v3.4s, v2.4s
+; CHECK-NEXT:    orr v3.4s, #64, lsl #16
+; CHECK-NEXT:    add v5.4s, v24.4s, v5.4s
+; CHECK-NEXT:    and v24.16b, v25.16b, v16.16b
+; CHECK-NEXT:    ushr v25.4s, v4.4s, #16
+; CHECK-NEXT:    fcvtn2 v22.4s, v19.2d
+; CHECK-NEXT:    add v19.4s, v23.4s, v26.4s
+; CHECK-NEXT:    ushr v26.4s, v6.4s, #16
+; CHECK-NEXT:    fcvtn2 v17.4s, v21.2d
+; CHECK-NEXT:    fcmeq v21.4s, v1.4s, v1.4s
 ; CHECK-NEXT:    orr v1.4s, #64, lsl #16
-; CHECK-NEXT:    ushr v28.4s, v21.4s, #16
-; CHECK-NEXT:    add v31.4s, v21.4s, v3.4s
-; CHECK-NEXT:    and v26.16b, v26.16b, v2.16b
-; CHECK-NEXT:    add v17.4s, v23.4s, v17.4s
-; CHECK-NEXT:    add v23.4s, v18.4s, v3.4s
-; CHECK-NEXT:    ushr v29.4s, v19.4s, #16
-; CHECK-NEXT:    and v27.16b, v27.16b, v2.16b
-; CHECK-NEXT:    add v3.4s, v19.4s, v3.4s
-; CHECK-NEXT:    add v16.4s, v16.4s, v22.4s
-; CHECK-NEXT:    and v28.16b, v28.16b, v2.16b
-; CHECK-NEXT:    fcmeq v22.4s, v4.4s, v4.4s
-; CHECK-NEXT:    orr v4.4s, #64, lsl #16
-; CHECK-NEXT:    and v2.16b, v29.16b, v2.16b
-; CHECK-NEXT:    fcmeq v29.4s, v5.4s, v5.4s
-; CHECK-NEXT:    orr v5.4s, #64, lsl #16
-; CHECK-NEXT:    add v23.4s, v26.4s, v23.4s
-; CHECK-NEXT:    fcmeq v26.4s, v18.4s, v18.4s
-; CHECK-NEXT:    add v27.4s, v27.4s, v30.4s
-; CHECK-NEXT:    fcmeq v30.4s, v24.4s, v24.4s
+; CHECK-NEXT:    and v23.16b, v25.16b, v16.16b
+; CHECK-NEXT:    add v25.4s, v4.4s, v2.4s
+; CHECK-NEXT:    add v7.4s, v24.4s, v7.4s
+; CHECK-NEXT:    ushr v24.4s, v18.4s, #16
+; CHECK-NEXT:    add v30.4s, v18.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v5.16b, v20.16b
+; CHECK-NEXT:    ushr v28.4s, v22.4s, #16
+; CHECK-NEXT:    add v31.4s, v22.4s, v2.4s
+; CHECK-NEXT:    add v23.4s, v23.4s, v25.4s
+; CHECK-NEXT:    and v25.16b, v26.16b, v16.16b
+; CHECK-NEXT:    add v26.4s, v6.4s, v2.4s
+; CHECK-NEXT:    ushr v29.4s, v17.4s, #16
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
+; CHECK-NEXT:    add v2.4s, v17.4s, v2.4s
+; CHECK-NEXT:    and v28.16b, v28.16b, v16.16b
+; CHECK-NEXT:    bit v3.16b, v7.16b, v27.16b
+; CHECK-NEXT:    bit v1.16b, v19.16b, v21.16b
+; CHECK-NEXT:    add v25.4s, v25.4s, v26.4s
+; CHECK-NEXT:    fcmeq v26.4s, v6.4s, v6.4s
+; CHECK-NEXT:    orr v6.4s, #64, lsl #16
+; CHECK-NEXT:    and v16.16b, v29.16b, v16.16b
+; CHECK-NEXT:    add v24.4s, v24.4s, v30.4s
+; CHECK-NEXT:    fcmeq v30.4s, v18.4s, v18.4s
 ; CHECK-NEXT:    add v28.4s, v28.4s, v31.4s
-; CHECK-NEXT:    fcmeq v31.4s, v21.4s, v21.4s
-; CHECK-NEXT:    add v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    fcmeq v3.4s, v19.4s, v19.4s
+; CHECK-NEXT:    fcmeq v31.4s, v22.4s, v22.4s
+; CHECK-NEXT:    fcmeq v29.4s, v4.4s, v4.4s
+; CHECK-NEXT:    orr v4.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v18.4s, #64, lsl #16
-; CHECK-NEXT:    orr v24.4s, #64, lsl #16
-; CHECK-NEXT:    orr v21.4s, #64, lsl #16
-; CHECK-NEXT:    orr v19.4s, #64, lsl #16
-; CHECK-NEXT:    bit v1.16b, v7.16b, v6.16b
-; CHECK-NEXT:    bit v4.16b, v16.16b, v22.16b
-; CHECK-NEXT:    mov v6.16b, v26.16b
-; CHECK-NEXT:    mov v7.16b, v30.16b
-; CHECK-NEXT:    mov v16.16b, v31.16b
-; CHECK-NEXT:    bit v0.16b, v20.16b, v25.16b
-; CHECK-NEXT:    bit v5.16b, v17.16b, v29.16b
-; CHECK-NEXT:    bsl v3.16b, v2.16b, v19.16b
-; CHECK-NEXT:    bsl v6.16b, v23.16b, v18.16b
-; CHECK-NEXT:    bsl v7.16b, v27.16b, v24.16b
-; CHECK-NEXT:    bsl v16.16b, v28.16b, v21.16b
+; CHECK-NEXT:    orr v22.4s, #64, lsl #16
+; CHECK-NEXT:    mov v5.16b, v26.16b
+; CHECK-NEXT:    add v2.4s, v16.4s, v2.4s
+; CHECK-NEXT:    fcmeq v16.4s, v17.4s, v17.4s
+; CHECK-NEXT:    orr v17.4s, #64, lsl #16
 ; CHECK-NEXT:    uzp2 v0.8h, v1.8h, v0.8h
-; CHECK-NEXT:    uzp2 v1.8h, v5.8h, v4.8h
-; CHECK-NEXT:    uzp2 v2.8h, v7.8h, v6.8h
-; CHECK-NEXT:    uzp2 v3.8h, v3.8h, v16.8h
+; CHECK-NEXT:    mov v7.16b, v31.16b
+; CHECK-NEXT:    bit v4.16b, v23.16b, v29.16b
+; CHECK-NEXT:    bsl v5.16b, v25.16b, v6.16b
+; CHECK-NEXT:    mov v6.16b, v30.16b
+; CHECK-NEXT:    bsl v16.16b, v2.16b, v17.16b
+; CHECK-NEXT:    bsl v7.16b, v28.16b, v22.16b
+; CHECK-NEXT:    bsl v6.16b, v24.16b, v18.16b
+; CHECK-NEXT:    uzp2 v1.8h, v4.8h, v3.8h
+; CHECK-NEXT:    uzp2 v3.8h, v16.8h, v7.8h
+; CHECK-NEXT:    uzp2 v2.8h, v6.8h, v5.8h
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <32 x i64> %a to <32 x bfloat>
@@ -6944,107 +6944,107 @@ entry:
 define <32 x bfloat> @utofp_v32i64_v32bf16(<32 x i64> %a) {
 ; CHECK-LABEL: utofp_v32i64_v32bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ucvtf v16.2d, v2.2d
-; CHECK-NEXT:    ucvtf v17.2d, v0.2d
-; CHECK-NEXT:    ucvtf v18.2d, v3.2d
-; CHECK-NEXT:    ucvtf v19.2d, v6.2d
-; CHECK-NEXT:    ldp q24, q23, [sp, #96]
-; CHECK-NEXT:    ucvtf v21.2d, v1.2d
-; CHECK-NEXT:    ucvtf v22.2d, v4.2d
+; CHECK-NEXT:    ucvtf v17.2d, v2.2d
+; CHECK-NEXT:    ucvtf v18.2d, v0.2d
+; CHECK-NEXT:    ucvtf v19.2d, v3.2d
+; CHECK-NEXT:    ucvtf v3.2d, v6.2d
+; CHECK-NEXT:    ldp q21, q20, [sp, #32]
+; CHECK-NEXT:    ucvtf v4.2d, v4.2d
 ; CHECK-NEXT:    ucvtf v6.2d, v7.2d
-; CHECK-NEXT:    ucvtf v7.2d, v5.2d
-; CHECK-NEXT:    movi v3.4s, #127, msl #8
-; CHECK-NEXT:    movi v2.4s, #1
-; CHECK-NEXT:    fcvtn v0.2s, v16.2d
-; CHECK-NEXT:    ldp q20, q16, [sp, #32]
-; CHECK-NEXT:    fcvtn v1.2s, v17.2d
-; CHECK-NEXT:    ldp q5, q17, [sp]
-; CHECK-NEXT:    fcvtn v4.2s, v19.2d
-; CHECK-NEXT:    ucvtf v23.2d, v23.2d
+; CHECK-NEXT:    ucvtf v5.2d, v5.2d
+; CHECK-NEXT:    ldp q24, q23, [sp, #64]
+; CHECK-NEXT:    movi v16.4s, #1
+; CHECK-NEXT:    fcvtn v0.2s, v17.2d
+; CHECK-NEXT:    ucvtf v17.2d, v1.2d
+; CHECK-NEXT:    fcvtn v1.2s, v18.2d
+; CHECK-NEXT:    fcvtn v3.2s, v3.2d
+; CHECK-NEXT:    ldp q18, q7, [sp]
+; CHECK-NEXT:    ucvtf v21.2d, v21.2d
+; CHECK-NEXT:    fcvtn v4.2s, v4.2d
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    ucvtf v20.2d, v20.2d
-; CHECK-NEXT:    ucvtf v16.2d, v16.2d
-; CHECK-NEXT:    fcvtn2 v0.4s, v18.2d
-; CHECK-NEXT:    ldp q19, q18, [sp, #64]
-; CHECK-NEXT:    ucvtf v25.2d, v5.2d
-; CHECK-NEXT:    fcvtn v5.2s, v22.2d
-; CHECK-NEXT:    fcvtn2 v1.4s, v21.2d
-; CHECK-NEXT:    ucvtf v21.2d, v24.2d
-; CHECK-NEXT:    ucvtf v17.2d, v17.2d
-; CHECK-NEXT:    fcvtn2 v4.4s, v6.2d
+; CHECK-NEXT:    fcvtn2 v0.4s, v19.2d
+; CHECK-NEXT:    ldp q22, q19, [sp, #96]
+; CHECK-NEXT:    fcvtn2 v1.4s, v17.2d
+; CHECK-NEXT:    fcvtn2 v3.4s, v6.2d
+; CHECK-NEXT:    ucvtf v18.2d, v18.2d
+; CHECK-NEXT:    ucvtf v17.2d, v24.2d
+; CHECK-NEXT:    fcvtn v6.2s, v21.2d
+; CHECK-NEXT:    fcvtn2 v4.4s, v5.2d
+; CHECK-NEXT:    ucvtf v22.2d, v22.2d
+; CHECK-NEXT:    ucvtf v21.2d, v23.2d
+; CHECK-NEXT:    ucvtf v7.2d, v7.2d
+; CHECK-NEXT:    ushr v24.4s, v0.4s, #16
+; CHECK-NEXT:    add v5.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    ucvtf v19.2d, v19.2d
-; CHECK-NEXT:    ucvtf v6.2d, v18.2d
-; CHECK-NEXT:    fcvtn v18.2s, v20.2d
-; CHECK-NEXT:    ushr v22.4s, v0.4s, #16
-; CHECK-NEXT:    add v20.4s, v0.4s, v3.4s
-; CHECK-NEXT:    fcvtn2 v5.4s, v7.2d
-; CHECK-NEXT:    fcvtn v24.2s, v25.2d
-; CHECK-NEXT:    ushr v7.4s, v1.4s, #16
-; CHECK-NEXT:    fcvtn v21.2s, v21.2d
-; CHECK-NEXT:    add v26.4s, v1.4s, v3.4s
-; CHECK-NEXT:    ushr v27.4s, v4.4s, #16
-; CHECK-NEXT:    fcvtn v19.2s, v19.2d
-; CHECK-NEXT:    fcvtn2 v18.4s, v16.2d
-; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
-; CHECK-NEXT:    and v7.16b, v7.16b, v2.16b
-; CHECK-NEXT:    fcmeq v25.4s, v0.4s, v0.4s
+; CHECK-NEXT:    ushr v23.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v3.4s, #16
+; CHECK-NEXT:    fcvtn v18.2s, v18.2d
+; CHECK-NEXT:    fcvtn2 v6.4s, v20.2d
+; CHECK-NEXT:    add v26.4s, v1.4s, v2.4s
+; CHECK-NEXT:    fcvtn v17.2s, v17.2d
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
+; CHECK-NEXT:    fcvtn v22.2s, v22.2d
+; CHECK-NEXT:    fcmeq v20.4s, v0.4s, v0.4s
+; CHECK-NEXT:    and v23.16b, v23.16b, v16.16b
 ; CHECK-NEXT:    orr v0.4s, #64, lsl #16
-; CHECK-NEXT:    ushr v28.4s, v5.4s, #16
-; CHECK-NEXT:    fcvtn2 v24.4s, v17.2d
-; CHECK-NEXT:    add v17.4s, v5.4s, v3.4s
-; CHECK-NEXT:    fcvtn2 v21.4s, v23.2d
-; CHECK-NEXT:    and v16.16b, v27.16b, v2.16b
-; CHECK-NEXT:    add v20.4s, v22.4s, v20.4s
-; CHECK-NEXT:    fcvtn2 v19.4s, v6.2d
-; CHECK-NEXT:    add v7.4s, v7.4s, v26.4s
-; CHECK-NEXT:    ushr v26.4s, v18.4s, #16
-; CHECK-NEXT:    and v23.16b, v28.16b, v2.16b
-; CHECK-NEXT:    add v22.4s, v4.4s, v3.4s
-; CHECK-NEXT:    fcmeq v6.4s, v1.4s, v1.4s
-; CHECK-NEXT:    ushr v27.4s, v24.4s, #16
-; CHECK-NEXT:    add v30.4s, v24.4s, v3.4s
+; CHECK-NEXT:    fcmeq v27.4s, v3.4s, v3.4s
+; CHECK-NEXT:    fcvtn2 v18.4s, v7.2d
+; CHECK-NEXT:    add v7.4s, v3.4s, v2.4s
+; CHECK-NEXT:    orr v3.4s, #64, lsl #16
+; CHECK-NEXT:    add v5.4s, v24.4s, v5.4s
+; CHECK-NEXT:    and v24.16b, v25.16b, v16.16b
+; CHECK-NEXT:    ushr v25.4s, v4.4s, #16
+; CHECK-NEXT:    fcvtn2 v22.4s, v19.2d
+; CHECK-NEXT:    add v19.4s, v23.4s, v26.4s
+; CHECK-NEXT:    ushr v26.4s, v6.4s, #16
+; CHECK-NEXT:    fcvtn2 v17.4s, v21.2d
+; CHECK-NEXT:    fcmeq v21.4s, v1.4s, v1.4s
 ; CHECK-NEXT:    orr v1.4s, #64, lsl #16
-; CHECK-NEXT:    ushr v28.4s, v21.4s, #16
-; CHECK-NEXT:    add v31.4s, v21.4s, v3.4s
-; CHECK-NEXT:    and v26.16b, v26.16b, v2.16b
-; CHECK-NEXT:    add v17.4s, v23.4s, v17.4s
-; CHECK-NEXT:    add v23.4s, v18.4s, v3.4s
-; CHECK-NEXT:    ushr v29.4s, v19.4s, #16
-; CHECK-NEXT:    and v27.16b, v27.16b, v2.16b
-; CHECK-NEXT:    add v3.4s, v19.4s, v3.4s
-; CHECK-NEXT:    add v16.4s, v16.4s, v22.4s
-; CHECK-NEXT:    and v28.16b, v28.16b, v2.16b
-; CHECK-NEXT:    fcmeq v22.4s, v4.4s, v4.4s
-; CHECK-NEXT:    orr v4.4s, #64, lsl #16
-; CHECK-NEXT:    and v2.16b, v29.16b, v2.16b
-; CHECK-NEXT:    fcmeq v29.4s, v5.4s, v5.4s
-; CHECK-NEXT:    orr v5.4s, #64, lsl #16
-; CHECK-NEXT:    add v23.4s, v26.4s, v23.4s
-; CHECK-NEXT:    fcmeq v26.4s, v18.4s, v18.4s
-; CHECK-NEXT:    add v27.4s, v27.4s, v30.4s
-; CHECK-NEXT:    fcmeq v30.4s, v24.4s, v24.4s
+; CHECK-NEXT:    and v23.16b, v25.16b, v16.16b
+; CHECK-NEXT:    add v25.4s, v4.4s, v2.4s
+; CHECK-NEXT:    add v7.4s, v24.4s, v7.4s
+; CHECK-NEXT:    ushr v24.4s, v18.4s, #16
+; CHECK-NEXT:    add v30.4s, v18.4s, v2.4s
+; CHECK-NEXT:    bit v0.16b, v5.16b, v20.16b
+; CHECK-NEXT:    ushr v28.4s, v22.4s, #16
+; CHECK-NEXT:    add v31.4s, v22.4s, v2.4s
+; CHECK-NEXT:    add v23.4s, v23.4s, v25.4s
+; CHECK-NEXT:    and v25.16b, v26.16b, v16.16b
+; CHECK-NEXT:    add v26.4s, v6.4s, v2.4s
+; CHECK-NEXT:    ushr v29.4s, v17.4s, #16
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
+; CHECK-NEXT:    add v2.4s, v17.4s, v2.4s
+; CHECK-NEXT:    and v28.16b, v28.16b, v16.16b
+; CHECK-NEXT:    bit v3.16b, v7.16b, v27.16b
+; CHECK-NEXT:    bit v1.16b, v19.16b, v21.16b
+; CHECK-NEXT:    add v25.4s, v25.4s, v26.4s
+; CHECK-NEXT:    fcmeq v26.4s, v6.4s, v6.4s
+; CHECK-NEXT:    orr v6.4s, #64, lsl #16
+; CHECK-NEXT:    and v16.16b, v29.16b, v16.16b
+; CHECK-NEXT:    add v24.4s, v24.4s, v30.4s
+; CHECK-NEXT:    fcmeq v30.4s, v18.4s, v18.4s
 ; CHECK-NEXT:    add v28.4s, v28.4s, v31.4s
-; CHECK-NEXT:    fcmeq v31.4s, v21.4s, v21.4s
-; CHECK-NEXT:    add v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    fcmeq v3.4s, v19.4s, v19.4s
+; CHECK-NEXT:    fcmeq v31.4s, v22.4s, v22.4s
+; CHECK-NEXT:    fcmeq v29.4s, v4.4s, v4.4s
+; CHECK-NEXT:    orr v4.4s, #64, lsl #16
 ; CHECK-NEXT:    orr v18.4s, #64, lsl #16
-; CHECK-NEXT:    orr v24.4s, #64, lsl #16
-; CHECK-NEXT:    orr v21.4s, #64, lsl #16
-; CHECK-NEXT:    orr v19.4s, #64, lsl #16
-; CHECK-NEXT:    bit v1.16b, v7.16b, v6.16b
-; CHECK-NEXT:    bit v4.16b, v16.16b, v22.16b
-; CHECK-NEXT:    mov v6.16b, v26.16b
-; CHECK-NEXT:    mov v7.16b, v30.16b
-; CHECK-NEXT:    mov v16.16b, v31.16b
-; CHECK-NEXT:    bit v0.16b, v20.16b, v25.16b
-; CHECK-NEXT:    bit v5.16b, v17.16b, v29.16b
-; CHECK-NEXT:    bsl v3.16b, v2.16b, v19.16b
-; CHECK-NEXT:    bsl v6.16b, v23.16b, v18.16b
-; CHECK-NEXT:    bsl v7.16b, v27.16b, v24.16b
-; CHECK-NEXT:    bsl v16.16b, v28.16b, v21.16b
+; CHECK-NEXT:    orr v22.4s, #64, lsl #16
+; CHECK-NEXT:    mov v5.16b, v26.16b
+; CHECK-NEXT:    add v2.4s, v16.4s, v2.4s
+; CHECK-NEXT:    fcmeq v16.4s, v17.4s, v17.4s
+; CHECK-NEXT:    orr v17.4s, #64, lsl #16
 ; CHECK-NEXT:    uzp2 v0.8h, v1.8h, v0.8h
-; CHECK-NEXT:    uzp2 v1.8h, v5.8h, v4.8h
-; CHECK-NEXT:    uzp2 v2.8h, v7.8h, v6.8h
-; CHECK-NEXT:    uzp2 v3.8h, v3.8h, v16.8h
+; CHECK-NEXT:    mov v7.16b, v31.16b
+; CHECK-NEXT:    bit v4.16b, v23.16b, v29.16b
+; CHECK-NEXT:    bsl v5.16b, v25.16b, v6.16b
+; CHECK-NEXT:    mov v6.16b, v30.16b
+; CHECK-NEXT:    bsl v16.16b, v2.16b, v17.16b
+; CHECK-NEXT:    bsl v7.16b, v28.16b, v22.16b
+; CHECK-NEXT:    bsl v6.16b, v24.16b, v18.16b
+; CHECK-NEXT:    uzp2 v1.8h, v4.8h, v3.8h
+; CHECK-NEXT:    uzp2 v3.8h, v16.8h, v7.8h
+; CHECK-NEXT:    uzp2 v2.8h, v6.8h, v5.8h
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <32 x i64> %a to <32 x bfloat>
@@ -7059,9 +7059,9 @@ define <2 x bfloat> @stofp_v2i32_v2bf16(<2 x i32> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <2 x i32> %a to <2 x bfloat>
@@ -7076,9 +7076,9 @@ define <2 x bfloat> @utofp_v2i32_v2bf16(<2 x i32> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <2 x i32> %a to <2 x bfloat>
@@ -7092,9 +7092,9 @@ define <3 x bfloat> @stofp_v3i32_v3bf16(<3 x i32> %a) {
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <3 x i32> %a to <3 x bfloat>
@@ -7108,9 +7108,9 @@ define <3 x bfloat> @utofp_v3i32_v3bf16(<3 x i32> %a) {
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <3 x i32> %a to <3 x bfloat>
@@ -7124,9 +7124,9 @@ define <4 x bfloat> @stofp_v4i32_v4bf16(<4 x i32> %a) {
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <4 x i32> %a to <4 x bfloat>
@@ -7140,9 +7140,9 @@ define <4 x bfloat> @utofp_v4i32_v4bf16(<4 x i32> %a) {
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <4 x i32> %a to <4 x bfloat>
@@ -7155,15 +7155,15 @@ define <8 x bfloat> @stofp_v8i32_v8bf16(<8 x i32> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    movi v2.4s, #1
 ; CHECK-NEXT:    scvtf v1.4s, v1.4s
+; CHECK-NEXT:    movi v5.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v4.4s, v1.4s, #16
 ; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
 ; CHECK-NEXT:    and v2.16b, v4.16b, v2.16b
 ; CHECK-NEXT:    add v0.4s, v3.4s, v0.4s
-; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v3.4s
-; CHECK-NEXT:    addhn2 v0.8h, v1.4s, v3.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v5.4s
+; CHECK-NEXT:    addhn2 v0.8h, v1.4s, v5.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <8 x i32> %a to <8 x bfloat>
@@ -7176,15 +7176,15 @@ define <8 x bfloat> @utofp_v8i32_v8bf16(<8 x i32> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    movi v2.4s, #1
 ; CHECK-NEXT:    ucvtf v1.4s, v1.4s
+; CHECK-NEXT:    movi v5.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v4.4s, v1.4s, #16
 ; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
 ; CHECK-NEXT:    and v2.16b, v4.16b, v2.16b
 ; CHECK-NEXT:    add v0.4s, v3.4s, v0.4s
-; CHECK-NEXT:    movi v3.4s, #127, msl #8
 ; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v3.4s
-; CHECK-NEXT:    addhn2 v0.8h, v1.4s, v3.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v5.4s
+; CHECK-NEXT:    addhn2 v0.8h, v1.4s, v5.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <8 x i32> %a to <8 x bfloat>
@@ -7194,28 +7194,28 @@ entry:
 define <16 x bfloat> @stofp_v16i32_v16bf16(<16 x i32> %a) {
 ; CHECK-LABEL: stofp_v16i32_v16bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    scvtf v2.4s, v2.4s
-; CHECK-NEXT:    movi v4.4s, #1
-; CHECK-NEXT:    scvtf v1.4s, v1.4s
+; CHECK-NEXT:    scvtf v0.4s, v0.4s
+; CHECK-NEXT:    scvtf v4.4s, v1.4s
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    scvtf v3.4s, v3.4s
+; CHECK-NEXT:    movi v17.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v5.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v6.4s, v2.4s, #16
-; CHECK-NEXT:    ushr v7.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v7.4s, v4.4s, #16
 ; CHECK-NEXT:    ushr v16.4s, v3.4s, #16
-; CHECK-NEXT:    and v5.16b, v5.16b, v4.16b
-; CHECK-NEXT:    and v6.16b, v6.16b, v4.16b
+; CHECK-NEXT:    and v5.16b, v5.16b, v1.16b
+; CHECK-NEXT:    and v6.16b, v6.16b, v1.16b
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
 ; CHECK-NEXT:    add v2.4s, v6.4s, v2.4s
-; CHECK-NEXT:    movi v6.4s, #127, msl #8
-; CHECK-NEXT:    and v5.16b, v7.16b, v4.16b
-; CHECK-NEXT:    and v4.16b, v16.16b, v4.16b
-; CHECK-NEXT:    add v5.4s, v5.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v6.4s
-; CHECK-NEXT:    add v3.4s, v4.4s, v3.4s
-; CHECK-NEXT:    addhn v1.4h, v2.4s, v6.4s
-; CHECK-NEXT:    addhn2 v0.8h, v5.4s, v6.4s
-; CHECK-NEXT:    addhn2 v1.8h, v3.4s, v6.4s
+; CHECK-NEXT:    and v5.16b, v7.16b, v1.16b
+; CHECK-NEXT:    and v6.16b, v16.16b, v1.16b
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v17.4s
+; CHECK-NEXT:    addhn v1.4h, v2.4s, v17.4s
+; CHECK-NEXT:    add v2.4s, v5.4s, v4.4s
+; CHECK-NEXT:    add v3.4s, v6.4s, v3.4s
+; CHECK-NEXT:    addhn2 v0.8h, v2.4s, v17.4s
+; CHECK-NEXT:    addhn2 v1.8h, v3.4s, v17.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <16 x i32> %a to <16 x bfloat>
@@ -7225,28 +7225,28 @@ entry:
 define <16 x bfloat> @utofp_v16i32_v16bf16(<16 x i32> %a) {
 ; CHECK-LABEL: utofp_v16i32_v16bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ucvtf v2.4s, v2.4s
-; CHECK-NEXT:    movi v4.4s, #1
-; CHECK-NEXT:    ucvtf v1.4s, v1.4s
+; CHECK-NEXT:    ucvtf v0.4s, v0.4s
+; CHECK-NEXT:    ucvtf v4.4s, v1.4s
+; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    ucvtf v3.4s, v3.4s
+; CHECK-NEXT:    movi v17.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v5.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v6.4s, v2.4s, #16
-; CHECK-NEXT:    ushr v7.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v7.4s, v4.4s, #16
 ; CHECK-NEXT:    ushr v16.4s, v3.4s, #16
-; CHECK-NEXT:    and v5.16b, v5.16b, v4.16b
-; CHECK-NEXT:    and v6.16b, v6.16b, v4.16b
+; CHECK-NEXT:    and v5.16b, v5.16b, v1.16b
+; CHECK-NEXT:    and v6.16b, v6.16b, v1.16b
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
 ; CHECK-NEXT:    add v2.4s, v6.4s, v2.4s
-; CHECK-NEXT:    movi v6.4s, #127, msl #8
-; CHECK-NEXT:    and v5.16b, v7.16b, v4.16b
-; CHECK-NEXT:    and v4.16b, v16.16b, v4.16b
-; CHECK-NEXT:    add v5.4s, v5.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v6.4s
-; CHECK-NEXT:    add v3.4s, v4.4s, v3.4s
-; CHECK-NEXT:    addhn v1.4h, v2.4s, v6.4s
-; CHECK-NEXT:    addhn2 v0.8h, v5.4s, v6.4s
-; CHECK-NEXT:    addhn2 v1.8h, v3.4s, v6.4s
+; CHECK-NEXT:    and v5.16b, v7.16b, v1.16b
+; CHECK-NEXT:    and v6.16b, v16.16b, v1.16b
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v17.4s
+; CHECK-NEXT:    addhn v1.4h, v2.4s, v17.4s
+; CHECK-NEXT:    add v2.4s, v5.4s, v4.4s
+; CHECK-NEXT:    add v3.4s, v6.4s, v3.4s
+; CHECK-NEXT:    addhn2 v0.8h, v2.4s, v17.4s
+; CHECK-NEXT:    addhn2 v1.8h, v3.4s, v17.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <16 x i32> %a to <16 x bfloat>
@@ -7262,42 +7262,42 @@ define <32 x bfloat> @stofp_v32i32_v32bf16(<32 x i32> %a) {
 ; CHECK-NEXT:    scvtf v6.4s, v6.4s
 ; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    scvtf v1.4s, v1.4s
-; CHECK-NEXT:    scvtf v3.4s, v3.4s
+; CHECK-NEXT:    scvtf v17.4s, v3.4s
 ; CHECK-NEXT:    scvtf v5.4s, v5.4s
 ; CHECK-NEXT:    scvtf v7.4s, v7.4s
-; CHECK-NEXT:    ushr v17.4s, v0.4s, #16
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
+; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v18.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v19.4s, v4.4s, #16
 ; CHECK-NEXT:    ushr v20.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v21.4s, v1.4s, #16
-; CHECK-NEXT:    ushr v22.4s, v3.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v5.4s, #16
-; CHECK-NEXT:    and v17.16b, v17.16b, v16.16b
+; CHECK-NEXT:    ushr v22.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v7.4s, #16
+; CHECK-NEXT:    and v3.16b, v3.16b, v16.16b
 ; CHECK-NEXT:    and v18.16b, v18.16b, v16.16b
 ; CHECK-NEXT:    and v19.16b, v19.16b, v16.16b
 ; CHECK-NEXT:    and v20.16b, v20.16b, v16.16b
-; CHECK-NEXT:    and v21.16b, v21.16b, v16.16b
-; CHECK-NEXT:    and v22.16b, v22.16b, v16.16b
-; CHECK-NEXT:    add v0.4s, v17.4s, v0.4s
-; CHECK-NEXT:    ushr v17.4s, v7.4s, #16
+; CHECK-NEXT:    add v0.4s, v3.4s, v0.4s
+; CHECK-NEXT:    and v3.16b, v22.16b, v16.16b
 ; CHECK-NEXT:    add v2.4s, v18.4s, v2.4s
-; CHECK-NEXT:    movi v18.4s, #127, msl #8
 ; CHECK-NEXT:    add v4.4s, v19.4s, v4.4s
 ; CHECK-NEXT:    add v6.4s, v20.4s, v6.4s
-; CHECK-NEXT:    and v19.16b, v23.16b, v16.16b
-; CHECK-NEXT:    add v20.4s, v22.4s, v3.4s
-; CHECK-NEXT:    and v16.16b, v17.16b, v16.16b
-; CHECK-NEXT:    add v17.4s, v21.4s, v1.4s
+; CHECK-NEXT:    and v18.16b, v23.16b, v16.16b
+; CHECK-NEXT:    and v19.16b, v24.16b, v16.16b
+; CHECK-NEXT:    and v16.16b, v25.16b, v16.16b
+; CHECK-NEXT:    add v20.4s, v3.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v21.4s
+; CHECK-NEXT:    addhn v1.4h, v2.4s, v21.4s
+; CHECK-NEXT:    addhn v2.4h, v4.4s, v21.4s
+; CHECK-NEXT:    addhn v3.4h, v6.4s, v21.4s
+; CHECK-NEXT:    add v4.4s, v18.4s, v17.4s
 ; CHECK-NEXT:    add v5.4s, v19.4s, v5.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v18.4s
-; CHECK-NEXT:    addhn v1.4h, v2.4s, v18.4s
-; CHECK-NEXT:    addhn v2.4h, v4.4s, v18.4s
-; CHECK-NEXT:    add v4.4s, v16.4s, v7.4s
-; CHECK-NEXT:    addhn v3.4h, v6.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v17.4s, v18.4s
-; CHECK-NEXT:    addhn2 v1.8h, v20.4s, v18.4s
-; CHECK-NEXT:    addhn2 v2.8h, v5.4s, v18.4s
-; CHECK-NEXT:    addhn2 v3.8h, v4.4s, v18.4s
+; CHECK-NEXT:    add v6.4s, v16.4s, v7.4s
+; CHECK-NEXT:    addhn2 v0.8h, v20.4s, v21.4s
+; CHECK-NEXT:    addhn2 v1.8h, v4.4s, v21.4s
+; CHECK-NEXT:    addhn2 v2.8h, v5.4s, v21.4s
+; CHECK-NEXT:    addhn2 v3.8h, v6.4s, v21.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <32 x i32> %a to <32 x bfloat>
@@ -7313,42 +7313,42 @@ define <32 x bfloat> @utofp_v32i32_v32bf16(<32 x i32> %a) {
 ; CHECK-NEXT:    ucvtf v6.4s, v6.4s
 ; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    ucvtf v1.4s, v1.4s
-; CHECK-NEXT:    ucvtf v3.4s, v3.4s
+; CHECK-NEXT:    ucvtf v17.4s, v3.4s
 ; CHECK-NEXT:    ucvtf v5.4s, v5.4s
 ; CHECK-NEXT:    ucvtf v7.4s, v7.4s
-; CHECK-NEXT:    ushr v17.4s, v0.4s, #16
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
+; CHECK-NEXT:    ushr v3.4s, v0.4s, #16
 ; CHECK-NEXT:    ushr v18.4s, v2.4s, #16
 ; CHECK-NEXT:    ushr v19.4s, v4.4s, #16
 ; CHECK-NEXT:    ushr v20.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v21.4s, v1.4s, #16
-; CHECK-NEXT:    ushr v22.4s, v3.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v5.4s, #16
-; CHECK-NEXT:    and v17.16b, v17.16b, v16.16b
+; CHECK-NEXT:    ushr v22.4s, v1.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v7.4s, #16
+; CHECK-NEXT:    and v3.16b, v3.16b, v16.16b
 ; CHECK-NEXT:    and v18.16b, v18.16b, v16.16b
 ; CHECK-NEXT:    and v19.16b, v19.16b, v16.16b
 ; CHECK-NEXT:    and v20.16b, v20.16b, v16.16b
-; CHECK-NEXT:    and v21.16b, v21.16b, v16.16b
-; CHECK-NEXT:    and v22.16b, v22.16b, v16.16b
-; CHECK-NEXT:    add v0.4s, v17.4s, v0.4s
-; CHECK-NEXT:    ushr v17.4s, v7.4s, #16
+; CHECK-NEXT:    add v0.4s, v3.4s, v0.4s
+; CHECK-NEXT:    and v3.16b, v22.16b, v16.16b
 ; CHECK-NEXT:    add v2.4s, v18.4s, v2.4s
-; CHECK-NEXT:    movi v18.4s, #127, msl #8
 ; CHECK-NEXT:    add v4.4s, v19.4s, v4.4s
 ; CHECK-NEXT:    add v6.4s, v20.4s, v6.4s
-; CHECK-NEXT:    and v19.16b, v23.16b, v16.16b
-; CHECK-NEXT:    add v20.4s, v22.4s, v3.4s
-; CHECK-NEXT:    and v16.16b, v17.16b, v16.16b
-; CHECK-NEXT:    add v17.4s, v21.4s, v1.4s
+; CHECK-NEXT:    and v18.16b, v23.16b, v16.16b
+; CHECK-NEXT:    and v19.16b, v24.16b, v16.16b
+; CHECK-NEXT:    and v16.16b, v25.16b, v16.16b
+; CHECK-NEXT:    add v20.4s, v3.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v21.4s
+; CHECK-NEXT:    addhn v1.4h, v2.4s, v21.4s
+; CHECK-NEXT:    addhn v2.4h, v4.4s, v21.4s
+; CHECK-NEXT:    addhn v3.4h, v6.4s, v21.4s
+; CHECK-NEXT:    add v4.4s, v18.4s, v17.4s
 ; CHECK-NEXT:    add v5.4s, v19.4s, v5.4s
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v18.4s
-; CHECK-NEXT:    addhn v1.4h, v2.4s, v18.4s
-; CHECK-NEXT:    addhn v2.4h, v4.4s, v18.4s
-; CHECK-NEXT:    add v4.4s, v16.4s, v7.4s
-; CHECK-NEXT:    addhn v3.4h, v6.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v17.4s, v18.4s
-; CHECK-NEXT:    addhn2 v1.8h, v20.4s, v18.4s
-; CHECK-NEXT:    addhn2 v2.8h, v5.4s, v18.4s
-; CHECK-NEXT:    addhn2 v3.8h, v4.4s, v18.4s
+; CHECK-NEXT:    add v6.4s, v16.4s, v7.4s
+; CHECK-NEXT:    addhn2 v0.8h, v20.4s, v21.4s
+; CHECK-NEXT:    addhn2 v1.8h, v4.4s, v21.4s
+; CHECK-NEXT:    addhn2 v2.8h, v5.4s, v21.4s
+; CHECK-NEXT:    addhn2 v3.8h, v6.4s, v21.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <32 x i32> %a to <32 x bfloat>
@@ -7364,9 +7364,9 @@ define <2 x bfloat> @stofp_v2i16_v2bf16(<2 x i16> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <2 x i16> %a to <2 x bfloat>
@@ -7382,9 +7382,9 @@ define <2 x bfloat> @utofp_v2i16_v2bf16(<2 x i16> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <2 x i16> %a to <2 x bfloat>
@@ -7399,9 +7399,9 @@ define <3 x bfloat> @stofp_v3i16_v3bf16(<3 x i16> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <3 x i16> %a to <3 x bfloat>
@@ -7416,9 +7416,9 @@ define <3 x bfloat> @utofp_v3i16_v3bf16(<3 x i16> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <3 x i16> %a to <3 x bfloat>
@@ -7433,9 +7433,9 @@ define <4 x bfloat> @stofp_v4i16_v4bf16(<4 x i16> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <4 x i16> %a to <4 x bfloat>
@@ -7450,9 +7450,9 @@ define <4 x bfloat> @utofp_v4i16_v4bf16(<4 x i16> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <4 x i16> %a to <4 x bfloat>
@@ -7513,27 +7513,27 @@ define <16 x bfloat> @stofp_v16i16_v16bf16(<16 x i16> %a) {
 ; CHECK-NEXT:    sshll2 v0.4s, v0.8h, #0
 ; CHECK-NEXT:    sshll2 v1.4s, v1.8h, #0
 ; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v7.4s, #127, msl #8
 ; CHECK-NEXT:    scvtf v3.4s, v3.4s
 ; CHECK-NEXT:    scvtf v4.4s, v4.4s
-; CHECK-NEXT:    scvtf v6.4s, v0.4s
-; CHECK-NEXT:    scvtf v7.4s, v1.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    ushr v5.4s, v3.4s, #16
-; CHECK-NEXT:    ushr v0.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v16.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v17.4s, v7.4s, #16
-; CHECK-NEXT:    and v5.16b, v5.16b, v2.16b
+; CHECK-NEXT:    scvtf v5.4s, v0.4s
+; CHECK-NEXT:    scvtf v6.4s, v1.4s
+; CHECK-NEXT:    ushr v0.4s, v3.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v16.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v17.4s, v6.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
 ; CHECK-NEXT:    and v16.16b, v16.16b, v2.16b
 ; CHECK-NEXT:    and v2.16b, v17.16b, v2.16b
-; CHECK-NEXT:    add v5.4s, v5.4s, v1.4s
-; CHECK-NEXT:    add v18.4s, v0.4s, v1.4s
-; CHECK-NEXT:    add v2.4s, v2.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v3.4s, v5.4s
-; CHECK-NEXT:    add v3.4s, v16.4s, v1.4s
-; CHECK-NEXT:    addhn v1.4h, v4.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v3.4s
-; CHECK-NEXT:    addhn2 v1.8h, v7.4s, v2.4s
+; CHECK-NEXT:    add v0.4s, v0.4s, v7.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v7.4s
+; CHECK-NEXT:    add v2.4s, v2.4s, v7.4s
+; CHECK-NEXT:    addhn v0.4h, v3.4s, v0.4s
+; CHECK-NEXT:    addhn v1.4h, v4.4s, v1.4s
+; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
+; CHECK-NEXT:    addhn2 v0.8h, v5.4s, v3.4s
+; CHECK-NEXT:    addhn2 v1.8h, v6.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <16 x i16> %a to <16 x bfloat>
@@ -7548,27 +7548,27 @@ define <16 x bfloat> @utofp_v16i16_v16bf16(<16 x i16> %a) {
 ; CHECK-NEXT:    ushll2 v0.4s, v0.8h, #0
 ; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
 ; CHECK-NEXT:    movi v2.4s, #1
+; CHECK-NEXT:    movi v7.4s, #127, msl #8
 ; CHECK-NEXT:    ucvtf v3.4s, v3.4s
 ; CHECK-NEXT:    ucvtf v4.4s, v4.4s
-; CHECK-NEXT:    ucvtf v6.4s, v0.4s
-; CHECK-NEXT:    ucvtf v7.4s, v1.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    ushr v5.4s, v3.4s, #16
-; CHECK-NEXT:    ushr v0.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v16.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v17.4s, v7.4s, #16
-; CHECK-NEXT:    and v5.16b, v5.16b, v2.16b
+; CHECK-NEXT:    ucvtf v5.4s, v0.4s
+; CHECK-NEXT:    ucvtf v6.4s, v1.4s
+; CHECK-NEXT:    ushr v0.4s, v3.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v16.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v17.4s, v6.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
 ; CHECK-NEXT:    and v16.16b, v16.16b, v2.16b
 ; CHECK-NEXT:    and v2.16b, v17.16b, v2.16b
-; CHECK-NEXT:    add v5.4s, v5.4s, v1.4s
-; CHECK-NEXT:    add v18.4s, v0.4s, v1.4s
-; CHECK-NEXT:    add v2.4s, v2.4s, v1.4s
-; CHECK-NEXT:    addhn v0.4h, v3.4s, v5.4s
-; CHECK-NEXT:    add v3.4s, v16.4s, v1.4s
-; CHECK-NEXT:    addhn v1.4h, v4.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v3.4s
-; CHECK-NEXT:    addhn2 v1.8h, v7.4s, v2.4s
+; CHECK-NEXT:    add v0.4s, v0.4s, v7.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v7.4s
+; CHECK-NEXT:    add v2.4s, v2.4s, v7.4s
+; CHECK-NEXT:    addhn v0.4h, v3.4s, v0.4s
+; CHECK-NEXT:    addhn v1.4h, v4.4s, v1.4s
+; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
+; CHECK-NEXT:    addhn2 v0.8h, v5.4s, v3.4s
+; CHECK-NEXT:    addhn2 v1.8h, v6.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <16 x i16> %a to <16 x bfloat>
@@ -7578,56 +7578,56 @@ entry:
 define <32 x bfloat> @stofp_v32i16_v32bf16(<32 x i16> %a) {
 ; CHECK-LABEL: stofp_v32i16_v32bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    sshll v5.4s, v1.4h, #0
-; CHECK-NEXT:    sshll v4.4s, v0.4h, #0
+; CHECK-NEXT:    sshll v4.4s, v1.4h, #0
+; CHECK-NEXT:    sshll v5.4s, v0.4h, #0
 ; CHECK-NEXT:    sshll v6.4s, v2.4h, #0
 ; CHECK-NEXT:    sshll v7.4s, v3.4h, #0
 ; CHECK-NEXT:    sshll2 v0.4s, v0.8h, #0
-; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    sshll2 v1.4s, v1.8h, #0
 ; CHECK-NEXT:    sshll2 v2.4s, v2.8h, #0
 ; CHECK-NEXT:    sshll2 v3.4s, v3.8h, #0
+; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    scvtf v5.4s, v5.4s
 ; CHECK-NEXT:    scvtf v4.4s, v4.4s
 ; CHECK-NEXT:    scvtf v6.4s, v6.4s
 ; CHECK-NEXT:    scvtf v7.4s, v7.4s
-; CHECK-NEXT:    scvtf v19.4s, v0.4s
-; CHECK-NEXT:    movi v18.4s, #127, msl #8
-; CHECK-NEXT:    scvtf v20.4s, v1.4s
-; CHECK-NEXT:    scvtf v21.4s, v2.4s
-; CHECK-NEXT:    scvtf v22.4s, v3.4s
+; CHECK-NEXT:    scvtf v17.4s, v0.4s
+; CHECK-NEXT:    scvtf v18.4s, v1.4s
+; CHECK-NEXT:    scvtf v19.4s, v2.4s
+; CHECK-NEXT:    scvtf v20.4s, v3.4s
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v0.4s, v5.4s, #16
-; CHECK-NEXT:    ushr v17.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v1.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v2.4s, v7.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v20.4s, #16
-; CHECK-NEXT:    ushr v25.4s, v22.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v2.4s, v6.4s, #16
+; CHECK-NEXT:    ushr v3.4s, v7.4s, #16
+; CHECK-NEXT:    ushr v22.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v18.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v19.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v20.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v16.16b
-; CHECK-NEXT:    and v3.16b, v17.16b, v16.16b
 ; CHECK-NEXT:    and v1.16b, v1.16b, v16.16b
 ; CHECK-NEXT:    and v2.16b, v2.16b, v16.16b
-; CHECK-NEXT:    ushr v17.4s, v19.4s, #16
+; CHECK-NEXT:    and v3.16b, v3.16b, v16.16b
+; CHECK-NEXT:    and v22.16b, v22.16b, v16.16b
 ; CHECK-NEXT:    and v23.16b, v23.16b, v16.16b
-; CHECK-NEXT:    add v24.4s, v0.4s, v18.4s
-; CHECK-NEXT:    ushr v0.4s, v21.4s, #16
-; CHECK-NEXT:    add v3.4s, v3.4s, v18.4s
-; CHECK-NEXT:    add v26.4s, v1.4s, v18.4s
-; CHECK-NEXT:    add v27.4s, v2.4s, v18.4s
-; CHECK-NEXT:    and v17.16b, v17.16b, v16.16b
-; CHECK-NEXT:    and v28.16b, v0.16b, v16.16b
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
 ; CHECK-NEXT:    and v16.16b, v25.16b, v16.16b
-; CHECK-NEXT:    addhn v0.4h, v4.4s, v3.4s
-; CHECK-NEXT:    addhn v1.4h, v5.4s, v24.4s
-; CHECK-NEXT:    add v4.4s, v17.4s, v18.4s
-; CHECK-NEXT:    addhn v2.4h, v6.4s, v26.4s
-; CHECK-NEXT:    add v5.4s, v23.4s, v18.4s
-; CHECK-NEXT:    addhn v3.4h, v7.4s, v27.4s
-; CHECK-NEXT:    add v6.4s, v28.4s, v18.4s
-; CHECK-NEXT:    add v16.4s, v16.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v19.4s, v4.4s
-; CHECK-NEXT:    addhn2 v1.8h, v20.4s, v5.4s
-; CHECK-NEXT:    addhn2 v2.8h, v21.4s, v6.4s
-; CHECK-NEXT:    addhn2 v3.8h, v22.4s, v16.4s
+; CHECK-NEXT:    add v0.4s, v0.4s, v21.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v21.4s
+; CHECK-NEXT:    add v2.4s, v2.4s, v21.4s
+; CHECK-NEXT:    add v3.4s, v3.4s, v21.4s
+; CHECK-NEXT:    addhn v0.4h, v5.4s, v0.4s
+; CHECK-NEXT:    addhn v1.4h, v4.4s, v1.4s
+; CHECK-NEXT:    addhn v2.4h, v6.4s, v2.4s
+; CHECK-NEXT:    addhn v3.4h, v7.4s, v3.4s
+; CHECK-NEXT:    add v4.4s, v22.4s, v21.4s
+; CHECK-NEXT:    add v5.4s, v23.4s, v21.4s
+; CHECK-NEXT:    add v6.4s, v24.4s, v21.4s
+; CHECK-NEXT:    add v7.4s, v16.4s, v21.4s
+; CHECK-NEXT:    addhn2 v0.8h, v17.4s, v4.4s
+; CHECK-NEXT:    addhn2 v1.8h, v18.4s, v5.4s
+; CHECK-NEXT:    addhn2 v2.8h, v19.4s, v6.4s
+; CHECK-NEXT:    addhn2 v3.8h, v20.4s, v7.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <32 x i16> %a to <32 x bfloat>
@@ -7637,56 +7637,56 @@ entry:
 define <32 x bfloat> @utofp_v32i16_v32bf16(<32 x i16> %a) {
 ; CHECK-LABEL: utofp_v32i16_v32bf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ushll v5.4s, v1.4h, #0
-; CHECK-NEXT:    ushll v4.4s, v0.4h, #0
+; CHECK-NEXT:    ushll v4.4s, v1.4h, #0
+; CHECK-NEXT:    ushll v5.4s, v0.4h, #0
 ; CHECK-NEXT:    ushll v6.4s, v2.4h, #0
 ; CHECK-NEXT:    ushll v7.4s, v3.4h, #0
 ; CHECK-NEXT:    ushll2 v0.4s, v0.8h, #0
-; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    ushll2 v1.4s, v1.8h, #0
 ; CHECK-NEXT:    ushll2 v2.4s, v2.8h, #0
 ; CHECK-NEXT:    ushll2 v3.4s, v3.8h, #0
+; CHECK-NEXT:    movi v16.4s, #1
 ; CHECK-NEXT:    ucvtf v5.4s, v5.4s
 ; CHECK-NEXT:    ucvtf v4.4s, v4.4s
 ; CHECK-NEXT:    ucvtf v6.4s, v6.4s
 ; CHECK-NEXT:    ucvtf v7.4s, v7.4s
-; CHECK-NEXT:    ucvtf v19.4s, v0.4s
-; CHECK-NEXT:    movi v18.4s, #127, msl #8
-; CHECK-NEXT:    ucvtf v20.4s, v1.4s
-; CHECK-NEXT:    ucvtf v21.4s, v2.4s
-; CHECK-NEXT:    ucvtf v22.4s, v3.4s
+; CHECK-NEXT:    ucvtf v17.4s, v0.4s
+; CHECK-NEXT:    ucvtf v18.4s, v1.4s
+; CHECK-NEXT:    ucvtf v19.4s, v2.4s
+; CHECK-NEXT:    ucvtf v20.4s, v3.4s
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
 ; CHECK-NEXT:    ushr v0.4s, v5.4s, #16
-; CHECK-NEXT:    ushr v17.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v1.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v2.4s, v7.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v20.4s, #16
-; CHECK-NEXT:    ushr v25.4s, v22.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v2.4s, v6.4s, #16
+; CHECK-NEXT:    ushr v3.4s, v7.4s, #16
+; CHECK-NEXT:    ushr v22.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v18.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v19.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v20.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v16.16b
-; CHECK-NEXT:    and v3.16b, v17.16b, v16.16b
 ; CHECK-NEXT:    and v1.16b, v1.16b, v16.16b
 ; CHECK-NEXT:    and v2.16b, v2.16b, v16.16b
-; CHECK-NEXT:    ushr v17.4s, v19.4s, #16
+; CHECK-NEXT:    and v3.16b, v3.16b, v16.16b
+; CHECK-NEXT:    and v22.16b, v22.16b, v16.16b
 ; CHECK-NEXT:    and v23.16b, v23.16b, v16.16b
-; CHECK-NEXT:    add v24.4s, v0.4s, v18.4s
-; CHECK-NEXT:    ushr v0.4s, v21.4s, #16
-; CHECK-NEXT:    add v3.4s, v3.4s, v18.4s
-; CHECK-NEXT:    add v26.4s, v1.4s, v18.4s
-; CHECK-NEXT:    add v27.4s, v2.4s, v18.4s
-; CHECK-NEXT:    and v17.16b, v17.16b, v16.16b
-; CHECK-NEXT:    and v28.16b, v0.16b, v16.16b
+; CHECK-NEXT:    and v24.16b, v24.16b, v16.16b
 ; CHECK-NEXT:    and v16.16b, v25.16b, v16.16b
-; CHECK-NEXT:    addhn v0.4h, v4.4s, v3.4s
-; CHECK-NEXT:    addhn v1.4h, v5.4s, v24.4s
-; CHECK-NEXT:    add v4.4s, v17.4s, v18.4s
-; CHECK-NEXT:    addhn v2.4h, v6.4s, v26.4s
-; CHECK-NEXT:    add v5.4s, v23.4s, v18.4s
-; CHECK-NEXT:    addhn v3.4h, v7.4s, v27.4s
-; CHECK-NEXT:    add v6.4s, v28.4s, v18.4s
-; CHECK-NEXT:    add v16.4s, v16.4s, v18.4s
-; CHECK-NEXT:    addhn2 v0.8h, v19.4s, v4.4s
-; CHECK-NEXT:    addhn2 v1.8h, v20.4s, v5.4s
-; CHECK-NEXT:    addhn2 v2.8h, v21.4s, v6.4s
-; CHECK-NEXT:    addhn2 v3.8h, v22.4s, v16.4s
+; CHECK-NEXT:    add v0.4s, v0.4s, v21.4s
+; CHECK-NEXT:    add v1.4s, v1.4s, v21.4s
+; CHECK-NEXT:    add v2.4s, v2.4s, v21.4s
+; CHECK-NEXT:    add v3.4s, v3.4s, v21.4s
+; CHECK-NEXT:    addhn v0.4h, v5.4s, v0.4s
+; CHECK-NEXT:    addhn v1.4h, v4.4s, v1.4s
+; CHECK-NEXT:    addhn v2.4h, v6.4s, v2.4s
+; CHECK-NEXT:    addhn v3.4h, v7.4s, v3.4s
+; CHECK-NEXT:    add v4.4s, v22.4s, v21.4s
+; CHECK-NEXT:    add v5.4s, v23.4s, v21.4s
+; CHECK-NEXT:    add v6.4s, v24.4s, v21.4s
+; CHECK-NEXT:    add v7.4s, v16.4s, v21.4s
+; CHECK-NEXT:    addhn2 v0.8h, v17.4s, v4.4s
+; CHECK-NEXT:    addhn2 v1.8h, v18.4s, v5.4s
+; CHECK-NEXT:    addhn2 v2.8h, v19.4s, v6.4s
+; CHECK-NEXT:    addhn2 v3.8h, v20.4s, v7.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <32 x i16> %a to <32 x bfloat>
@@ -7768,9 +7768,9 @@ define <3 x bfloat> @stofp_v3i8_v3bf16(<3 x i8> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <3 x i8> %a to <3 x bfloat>
@@ -7789,9 +7789,9 @@ define <3 x bfloat> @utofp_v3i8_v3bf16(<3 x i8> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <3 x i8> %a to <3 x bfloat>
@@ -7808,9 +7808,9 @@ define <4 x bfloat> @stofp_v4i8_v4bf16(<4 x i8> %a) {
 ; CHECK-NEXT:    scvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <4 x i8> %a to <4 x bfloat>
@@ -7826,9 +7826,9 @@ define <4 x bfloat> @utofp_v4i8_v4bf16(<4 x i8> %a) {
 ; CHECK-NEXT:    ucvtf v0.4s, v0.4s
 ; CHECK-NEXT:    ushr v2.4s, v0.4s, #16
 ; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    movi v2.4s, #127, msl #8
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    movi v1.4s, #127, msl #8
-; CHECK-NEXT:    addhn v0.4h, v0.4s, v1.4s
+; CHECK-NEXT:    addhn v0.4h, v0.4s, v2.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <4 x i8> %a to <4 x bfloat>
@@ -7909,11 +7909,11 @@ define <16 x bfloat> @stofp_v16i8_v16bf16(<16 x i8> %a) {
 ; CHECK-NEXT:    add v5.4s, v5.4s, v7.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v7.4s
 ; CHECK-NEXT:    addhn v1.4h, v3.4s, v5.4s
-; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
-; CHECK-NEXT:    add v5.4s, v17.4s, v7.4s
 ; CHECK-NEXT:    addhn v0.4h, v4.4s, v0.4s
+; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
+; CHECK-NEXT:    add v4.4s, v17.4s, v7.4s
 ; CHECK-NEXT:    addhn2 v1.8h, v2.4s, v3.4s
-; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v5.4s
+; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v4.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <16 x i8> %a to <16 x bfloat>
@@ -7946,11 +7946,11 @@ define <16 x bfloat> @utofp_v16i8_v16bf16(<16 x i8> %a) {
 ; CHECK-NEXT:    add v5.4s, v5.4s, v7.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v7.4s
 ; CHECK-NEXT:    addhn v1.4h, v3.4s, v5.4s
-; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
-; CHECK-NEXT:    add v5.4s, v17.4s, v7.4s
 ; CHECK-NEXT:    addhn v0.4h, v4.4s, v0.4s
+; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
+; CHECK-NEXT:    add v4.4s, v17.4s, v7.4s
 ; CHECK-NEXT:    addhn2 v1.8h, v2.4s, v3.4s
-; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v5.4s
+; CHECK-NEXT:    addhn2 v0.8h, v6.4s, v4.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <16 x i8> %a to <16 x bfloat>
@@ -7961,14 +7961,14 @@ define <32 x bfloat> @stofp_v32i8_v32bf16(<32 x i8> %a) {
 ; CHECK-LABEL: stofp_v32i8_v32bf16:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    sshll2 v3.8h, v0.16b, #0
-; CHECK-NEXT:    sshll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    sshll v0.8h, v0.8b, #0
+; CHECK-NEXT:    sshll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
 ; CHECK-NEXT:    movi v2.4s, #1
-; CHECK-NEXT:    movi v20.4s, #127, msl #8
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
 ; CHECK-NEXT:    sshll v5.4s, v3.4h, #0
-; CHECK-NEXT:    sshll v6.4s, v4.4h, #0
-; CHECK-NEXT:    sshll v7.4s, v0.4h, #0
+; CHECK-NEXT:    sshll v6.4s, v0.4h, #0
+; CHECK-NEXT:    sshll v7.4s, v4.4h, #0
 ; CHECK-NEXT:    sshll v16.4s, v1.4h, #0
 ; CHECK-NEXT:    sshll2 v3.4s, v3.8h, #0
 ; CHECK-NEXT:    sshll2 v4.4s, v4.8h, #0
@@ -7980,40 +7980,40 @@ define <32 x bfloat> @stofp_v32i8_v32bf16(<32 x i8> %a) {
 ; CHECK-NEXT:    scvtf v16.4s, v16.4s
 ; CHECK-NEXT:    scvtf v17.4s, v3.4s
 ; CHECK-NEXT:    scvtf v4.4s, v4.4s
-; CHECK-NEXT:    scvtf v19.4s, v0.4s
-; CHECK-NEXT:    scvtf v21.4s, v1.4s
-; CHECK-NEXT:    ushr v3.4s, v5.4s, #16
-; CHECK-NEXT:    ushr v18.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v0.4s, v7.4s, #16
-; CHECK-NEXT:    ushr v1.4s, v16.4s, #16
-; CHECK-NEXT:    ushr v22.4s, v17.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v24.4s, v19.4s, #16
-; CHECK-NEXT:    ushr v25.4s, v21.4s, #16
-; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
-; CHECK-NEXT:    and v18.16b, v18.16b, v2.16b
+; CHECK-NEXT:    scvtf v18.4s, v0.4s
+; CHECK-NEXT:    scvtf v19.4s, v1.4s
+; CHECK-NEXT:    ushr v0.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v3.4s, v6.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v7.4s, #16
+; CHECK-NEXT:    ushr v20.4s, v16.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v22.4s, v18.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v19.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
 ; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
+; CHECK-NEXT:    and v20.16b, v20.16b, v2.16b
 ; CHECK-NEXT:    and v23.16b, v23.16b, v2.16b
 ; CHECK-NEXT:    and v24.16b, v24.16b, v2.16b
-; CHECK-NEXT:    and v2.16b, v25.16b, v2.16b
-; CHECK-NEXT:    add v3.4s, v3.4s, v20.4s
-; CHECK-NEXT:    add v18.4s, v18.4s, v20.4s
-; CHECK-NEXT:    add v0.4s, v0.4s, v20.4s
-; CHECK-NEXT:    add v26.4s, v1.4s, v20.4s
-; CHECK-NEXT:    addhn v1.4h, v5.4s, v3.4s
-; CHECK-NEXT:    addhn v3.4h, v6.4s, v18.4s
-; CHECK-NEXT:    addhn v0.4h, v7.4s, v0.4s
-; CHECK-NEXT:    add v5.4s, v22.4s, v20.4s
-; CHECK-NEXT:    add v6.4s, v24.4s, v20.4s
-; CHECK-NEXT:    add v7.4s, v23.4s, v20.4s
-; CHECK-NEXT:    add v18.4s, v2.4s, v20.4s
-; CHECK-NEXT:    addhn v2.4h, v16.4s, v26.4s
-; CHECK-NEXT:    addhn2 v0.8h, v19.4s, v6.4s
-; CHECK-NEXT:    addhn2 v1.8h, v17.4s, v5.4s
+; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
+; CHECK-NEXT:    and v25.16b, v25.16b, v2.16b
+; CHECK-NEXT:    add v0.4s, v0.4s, v21.4s
+; CHECK-NEXT:    add v3.4s, v3.4s, v21.4s
+; CHECK-NEXT:    add v26.4s, v1.4s, v21.4s
+; CHECK-NEXT:    add v20.4s, v20.4s, v21.4s
+; CHECK-NEXT:    addhn v1.4h, v5.4s, v0.4s
+; CHECK-NEXT:    addhn v0.4h, v6.4s, v3.4s
+; CHECK-NEXT:    addhn v3.4h, v7.4s, v26.4s
+; CHECK-NEXT:    addhn v2.4h, v16.4s, v20.4s
+; CHECK-NEXT:    add v5.4s, v22.4s, v21.4s
+; CHECK-NEXT:    add v6.4s, v23.4s, v21.4s
+; CHECK-NEXT:    add v7.4s, v24.4s, v21.4s
+; CHECK-NEXT:    add v16.4s, v25.4s, v21.4s
+; CHECK-NEXT:    addhn2 v0.8h, v18.4s, v5.4s
+; CHECK-NEXT:    addhn2 v1.8h, v17.4s, v6.4s
 ; CHECK-NEXT:    addhn2 v3.8h, v4.4s, v7.4s
-; CHECK-NEXT:    addhn2 v2.8h, v21.4s, v18.4s
+; CHECK-NEXT:    addhn2 v2.8h, v19.4s, v16.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = sitofp <32 x i8> %a to <32 x bfloat>
@@ -8024,14 +8024,14 @@ define <32 x bfloat> @utofp_v32i8_v32bf16(<32 x i8> %a) {
 ; CHECK-LABEL: utofp_v32i8_v32bf16:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ushll2 v3.8h, v0.16b, #0
-; CHECK-NEXT:    ushll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
+; CHECK-NEXT:    ushll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
 ; CHECK-NEXT:    movi v2.4s, #1
-; CHECK-NEXT:    movi v20.4s, #127, msl #8
+; CHECK-NEXT:    movi v21.4s, #127, msl #8
 ; CHECK-NEXT:    ushll v5.4s, v3.4h, #0
-; CHECK-NEXT:    ushll v6.4s, v4.4h, #0
-; CHECK-NEXT:    ushll v7.4s, v0.4h, #0
+; CHECK-NEXT:    ushll v6.4s, v0.4h, #0
+; CHECK-NEXT:    ushll v7.4s, v4.4h, #0
 ; CHECK-NEXT:    ushll v16.4s, v1.4h, #0
 ; CHECK-NEXT:    ushll2 v3.4s, v3.8h, #0
 ; CHECK-NEXT:    ushll2 v4.4s, v4.8h, #0
@@ -8043,40 +8043,40 @@ define <32 x bfloat> @utofp_v32i8_v32bf16(<32 x i8> %a) {
 ; CHECK-NEXT:    ucvtf v16.4s, v16.4s
 ; CHECK-NEXT:    ucvtf v17.4s, v3.4s
 ; CHECK-NEXT:    ucvtf v4.4s, v4.4s
-; CHECK-NEXT:    ucvtf v19.4s, v0.4s
-; CHECK-NEXT:    ucvtf v21.4s, v1.4s
-; CHECK-NEXT:    ushr v3.4s, v5.4s, #16
-; CHECK-NEXT:    ushr v18.4s, v6.4s, #16
-; CHECK-NEXT:    ushr v0.4s, v7.4s, #16
-; CHECK-NEXT:    ushr v1.4s, v16.4s, #16
-; CHECK-NEXT:    ushr v22.4s, v17.4s, #16
-; CHECK-NEXT:    ushr v23.4s, v4.4s, #16
-; CHECK-NEXT:    ushr v24.4s, v19.4s, #16
-; CHECK-NEXT:    ushr v25.4s, v21.4s, #16
-; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
-; CHECK-NEXT:    and v18.16b, v18.16b, v2.16b
+; CHECK-NEXT:    ucvtf v18.4s, v0.4s
+; CHECK-NEXT:    ucvtf v19.4s, v1.4s
+; CHECK-NEXT:    ushr v0.4s, v5.4s, #16
+; CHECK-NEXT:    ushr v3.4s, v6.4s, #16
+; CHECK-NEXT:    ushr v1.4s, v7.4s, #16
+; CHECK-NEXT:    ushr v20.4s, v16.4s, #16
+; CHECK-NEXT:    ushr v23.4s, v17.4s, #16
+; CHECK-NEXT:    ushr v24.4s, v4.4s, #16
+; CHECK-NEXT:    ushr v22.4s, v18.4s, #16
+; CHECK-NEXT:    ushr v25.4s, v19.4s, #16
 ; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    and v3.16b, v3.16b, v2.16b
 ; CHECK-NEXT:    and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
+; CHECK-NEXT:    and v20.16b, v20.16b, v2.16b
 ; CHECK-NEXT:    and v23.16b, v23.16b, v2.16b
 ; CHECK-NEXT:    and v24.16b, v24.16b, v2.16b
-; CHECK-NEXT:    and v2.16b, v25.16b, v2.16b
-; CHECK-NEXT:    add v3.4s, v3.4s, v20.4s
-; CHECK-NEXT:    add v18.4s, v18.4s, v20.4s
-; CHECK-NEXT:    add v0.4s, v0.4s, v20.4s
-; CHECK-NEXT:    add v26.4s, v1.4s, v20.4s
-; CHECK-NEXT:    addhn v1.4h, v5.4s, v3.4s
-; CHECK-NEXT:    addhn v3.4h, v6.4s, v18.4s
-; CHECK-NEXT:    addhn v0.4h, v7.4s, v0.4s
-; CHECK-NEXT:    add v5.4s, v22.4s, v20.4s
-; CHECK-NEXT:    add v6.4s, v24.4s, v20.4s
-; CHECK-NEXT:    add v7.4s, v23.4s, v20.4s
-; CHECK-NEXT:    add v18.4s, v2.4s, v20.4s
-; CHECK-NEXT:    addhn v2.4h, v16.4s, v26.4s
-; CHECK-NEXT:    addhn2 v0.8h, v19.4s, v6.4s
-; CHECK-NEXT:    addhn2 v1.8h, v17.4s, v5.4s
+; CHECK-NEXT:    and v22.16b, v22.16b, v2.16b
+; CHECK-NEXT:    and v25.16b, v25.16b, v2.16b
+; CHECK-NEXT:    add v0.4s, v0.4s, v21.4s
+; CHECK-NEXT:    add v3.4s, v3.4s, v21.4s
+; CHECK-NEXT:    add v26.4s, v1.4s, v21.4s
+; CHECK-NEXT:    add v20.4s, v20.4s, v21.4s
+; CHECK-NEXT:    addhn v1.4h, v5.4s, v0.4s
+; CHECK-NEXT:    addhn v0.4h, v6.4s, v3.4s
+; CHECK-NEXT:    addhn v3.4h, v7.4s, v26.4s
+; CHECK-NEXT:    addhn v2.4h, v16.4s, v20.4s
+; CHECK-NEXT:    add v5.4s, v22.4s, v21.4s
+; CHECK-NEXT:    add v6.4s, v23.4s, v21.4s
+; CHECK-NEXT:    add v7.4s, v24.4s, v21.4s
+; CHECK-NEXT:    add v16.4s, v25.4s, v21.4s
+; CHECK-NEXT:    addhn2 v0.8h, v18.4s, v5.4s
+; CHECK-NEXT:    addhn2 v1.8h, v17.4s, v6.4s
 ; CHECK-NEXT:    addhn2 v3.8h, v4.4s, v7.4s
-; CHECK-NEXT:    addhn2 v2.8h, v21.4s, v18.4s
+; CHECK-NEXT:    addhn2 v2.8h, v19.4s, v16.4s
 ; CHECK-NEXT:    ret
 entry:
   %c = uitofp <32 x i8> %a to <32 x bfloat>
diff --git a/llvm/test/CodeGen/AArch64/ldexp.ll b/llvm/test/CodeGen/AArch64/ldexp.ll
index 4b491051a88aa..ba04ba1d7bb6a 100644
--- a/llvm/test/CodeGen/AArch64/ldexp.ll
+++ b/llvm/test/CodeGen/AArch64/ldexp.ll
@@ -4,9 +4,9 @@
 define double @testExp(double %val, i32 %a) {
 ; CHECK-LABEL: testExp:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    sxtw x8, w0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    fmov d1, x8
 ; CHECK-NEXT:    fscale z0.d, p0/m, z0.d, z1.d
@@ -22,8 +22,8 @@ declare double @ldexp(double, i32) memory(none)
 define float @testExpf(float %val, i32 %a) {
 ; CHECK-LABEL: testExpf:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fmov s1, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $z0
 ; CHECK-NEXT:    fscale z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $z0
@@ -49,9 +49,9 @@ declare fp128 @ldexpl(fp128, i32) memory(none)
 define half @testExpf16(half %val, i32 %a) {
 ; CHECK-LABEL: testExpf16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcvt s0, h0
 ; CHECK-NEXT:    fmov s1, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fscale z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fcvt h0, s0
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll b/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
index ab15bf564ec42..59a460923e8b7 100644
--- a/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
+++ b/llvm/test/CodeGen/AArch64/llvm-ir-to-intrinsic.ll
@@ -10,9 +10,9 @@ define <vscale x 16 x i8> @sdiv_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sunpkhi z2.h, z1.b
 ; CHECK-NEXT:    sunpkhi z3.h, z0.b
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z1.h, z1.b
 ; CHECK-NEXT:    sunpklo z0.h, z0.b
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z4.s, z2.h
 ; CHECK-NEXT:    sunpkhi z5.s, z3.h
 ; CHECK-NEXT:    sunpklo z2.s, z2.h
@@ -36,11 +36,11 @@ define <vscale x 16 x i8> @sdiv_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 8 x i16> @sdiv_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: sdiv_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z2.s, z1.h
 ; CHECK-NEXT:    sunpkhi z3.s, z0.h
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z2.h
@@ -140,9 +140,9 @@ define <vscale x 16 x i8> @srem_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 8 x i16> @srem_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: srem_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z2.s, z1.h
 ; CHECK-NEXT:    sunpkhi z3.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z4.s, z0.h
 ; CHECK-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    sunpklo z3.s, z1.h
@@ -188,9 +188,9 @@ define <vscale x 16 x i8> @udiv_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    uunpkhi z2.h, z1.b
 ; CHECK-NEXT:    uunpkhi z3.h, z0.b
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z1.h, z1.b
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z4.s, z2.h
 ; CHECK-NEXT:    uunpkhi z5.s, z3.h
 ; CHECK-NEXT:    uunpklo z2.s, z2.h
@@ -214,11 +214,11 @@ define <vscale x 16 x i8> @udiv_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 8 x i16> @udiv_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: udiv_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z2.s, z1.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
 ; CHECK-NEXT:    uunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z2.h
@@ -261,9 +261,9 @@ define <vscale x 8 x i32> @udiv_split_i32(<vscale x 8 x i32> %a, <vscale x 8 x i
 define <vscale x 2 x i32> @udiv_widen_i32(<vscale x 2 x i32> %a, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: udiv_widen_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    udiv z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %div = udiv <vscale x 2 x i32> %a, %b
@@ -319,9 +319,9 @@ define <vscale x 16 x i8> @urem_i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 8 x i16> @urem_i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
 ; CHECK-LABEL: urem_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z2.s, z1.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z4.s, z0.h
 ; CHECK-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    uunpklo z3.s, z1.h
@@ -558,9 +558,9 @@ define <vscale x 4 x i64> @umin_split_i64(<vscale x 4 x i64> %a, <vscale x 4 x i
 define <vscale x 8 x i8> @umin_promote_i8(<vscale x 8 x i8> %a, <vscale x 8 x i8> %b) {
 ; CHECK-LABEL: umin_promote_i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %cmp = icmp ult <vscale x 8 x i8> %a, %b
@@ -704,9 +704,9 @@ define <vscale x 16 x i16> @umax_split_i16(<vscale x 16 x i16> %a, <vscale x 16
 define <vscale x 2 x i32> @umax_promote_i32(<vscale x 2 x i32> %a, <vscale x 2 x i32> %b) {
 ; CHECK-LABEL: umax_promote_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %cmp = icmp ugt <vscale x 2 x i32> %a, %b
@@ -883,8 +883,8 @@ define <vscale x 4 x i64> @lsl_split_i64(<vscale x 4 x i64> %a, <vscale x 4 x i6
 define <vscale x 4 x i16> @lsl_promote_i16(<vscale x 4 x i16> %a, <vscale x 4 x i16> %b){
 ; CHECK-LABEL: lsl_promote_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    lsl z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %shl = shl <vscale x 4 x i16> %a, %b
@@ -982,9 +982,9 @@ define <vscale x 2 x i64> @lsr_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
 define <vscale x 8 x i8> @lsr_promote_i8(<vscale x 8 x i8> %a, <vscale x 8 x i8> %b){
 ; CHECK-LABEL: lsr_promote_i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    lsr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %shr = lshr <vscale x 8 x i8> %a, %b
@@ -1081,10 +1081,10 @@ declare <vscale x 2 x i64> @llvm.fshr.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x
 define <vscale x 2 x i64> @fshl_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c){
 ; CHECK-LABEL: fshl_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z3.d, #63 // =0x3f
 ; CHECK-NEXT:    mov z4.d, z2.d
 ; CHECK-NEXT:    lsr z1.d, z1.d, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    bic z2.d, z3.d, z2.d
 ; CHECK-NEXT:    and z4.d, z4.d, #0x3f
 ; CHECK-NEXT:    lsl z0.d, p0/m, z0.d, z4.d
@@ -1098,17 +1098,16 @@ define <vscale x 2 x i64> @fshl_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b
 define <vscale x 4 x i64> @fshl_illegal_i64(<vscale x 4 x i64> %a, <vscale x 4 x i64> %b, <vscale x 4 x i64> %c){
 ; CHECK-LABEL: fshl_illegal_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z6.d, #63 // =0x3f
-; CHECK-NEXT:    mov z7.d, z4.d
 ; CHECK-NEXT:    lsr z2.d, z2.d, #1
 ; CHECK-NEXT:    lsr z3.d, z3.d, #1
-; CHECK-NEXT:    bic z4.d, z6.d, z4.d
-; CHECK-NEXT:    and z7.d, z7.d, #0x3f
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    bic z7.d, z6.d, z4.d
+; CHECK-NEXT:    and z4.d, z4.d, #0x3f
 ; CHECK-NEXT:    bic z6.d, z6.d, z5.d
 ; CHECK-NEXT:    and z5.d, z5.d, #0x3f
-; CHECK-NEXT:    lsl z0.d, p0/m, z0.d, z7.d
-; CHECK-NEXT:    lsr z2.d, p0/m, z2.d, z4.d
+; CHECK-NEXT:    lsl z0.d, p0/m, z0.d, z4.d
+; CHECK-NEXT:    lsr z2.d, p0/m, z2.d, z7.d
 ; CHECK-NEXT:    lsr z3.d, p0/m, z3.d, z6.d
 ; CHECK-NEXT:    lsl z1.d, p0/m, z1.d, z5.d
 ; CHECK-NEXT:    orr z0.d, z0.d, z2.d
@@ -1121,9 +1120,9 @@ define <vscale x 4 x i64> @fshl_illegal_i64(<vscale x 4 x i64> %a, <vscale x 4 x
 define <vscale x 2 x i64> @fshl_rot_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b){
 ; CHECK-LABEL: fshl_rot_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    subr z1.d, z1.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    and z1.d, z1.d, #0x3f
 ; CHECK-NEXT:    lslr z2.d, p0/m, z2.d, z0.d
@@ -1138,11 +1137,11 @@ define <vscale x 2 x i64> @fshl_rot_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64
 define <vscale x 4 x i64> @fshl_rot_illegal_i64(<vscale x 4 x i64> %a, <vscale x 4 x i64> %b){
 ; CHECK-LABEL: fshl_rot_illegal_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z4.d, z2.d
 ; CHECK-NEXT:    subr z2.d, z2.d, #0 // =0x0
 ; CHECK-NEXT:    mov z5.d, z3.d
 ; CHECK-NEXT:    subr z3.d, z3.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z4.d, z4.d, #0x3f
 ; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    and z5.d, z5.d, #0x3f
@@ -1175,10 +1174,10 @@ define <vscale x 2 x i64> @fshl_rot_const_i64(<vscale x 2 x i64> %a){
 define <vscale x 2 x i64> @fshr_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c){
 ; CHECK-LABEL: fshr_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z3.d, #63 // =0x3f
 ; CHECK-NEXT:    mov z4.d, z2.d
 ; CHECK-NEXT:    lsl z0.d, z0.d, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    bic z2.d, z3.d, z2.d
 ; CHECK-NEXT:    and z4.d, z4.d, #0x3f
 ; CHECK-NEXT:    lsr z1.d, p0/m, z1.d, z4.d
@@ -1192,9 +1191,9 @@ define <vscale x 2 x i64> @fshr_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b
 define <vscale x 2 x i64> @fshr_rot_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b){
 ; CHECK-LABEL: fshr_rot_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    subr z1.d, z1.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    and z1.d, z1.d, #0x3f
 ; CHECK-NEXT:    lsrr z2.d, p0/m, z2.d, z0.d
diff --git a/llvm/test/CodeGen/AArch64/load-insert-zero.ll b/llvm/test/CodeGen/AArch64/load-insert-zero.ll
index 993af08a66ddd..23d545459295f 100644
--- a/llvm/test/CodeGen/AArch64/load-insert-zero.ll
+++ b/llvm/test/CodeGen/AArch64/load-insert-zero.ll
@@ -469,18 +469,18 @@ define void @predictor_4x4_neon(ptr nocapture noundef writeonly %0, i64 noundef
 ; CHECK-NEXT:    lsr w8, w8, #24
 ; CHECK-NEXT:    uaddl v0.8h, v0.8b, v1.8b
 ; CHECK-NEXT:    urhadd v1.8b, v1.8b, v2.8b
-; CHECK-NEXT:    str s1, [x0]
 ; CHECK-NEXT:    add v0.8h, v0.8h, v3.8h
 ; CHECK-NEXT:    dup v3.8b, w8
+; CHECK-NEXT:    str s1, [x0]
 ; CHECK-NEXT:    lsl x8, x1, #1
 ; CHECK-NEXT:    rshrn v0.8b, v0.8h, #2
 ; CHECK-NEXT:    zip1 v2.2s, v1.2s, v3.2s
 ; CHECK-NEXT:    str s0, [x0, x1]
 ; CHECK-NEXT:    zip1 v3.2s, v0.2s, v3.2s
 ; CHECK-NEXT:    ext v2.8b, v2.8b, v0.8b, #1
+; CHECK-NEXT:    ext v1.8b, v3.8b, v0.8b, #1
 ; CHECK-NEXT:    str s2, [x0, x8]
 ; CHECK-NEXT:    add x8, x8, x1
-; CHECK-NEXT:    ext v1.8b, v3.8b, v0.8b, #1
 ; CHECK-NEXT:    str s1, [x0, x8]
 ; CHECK-NEXT:    ret
   %5 = load i32, ptr %2, align 4
@@ -608,9 +608,9 @@ define void @predictor_4x4_neon_new(ptr nocapture noundef writeonly %0, i64 noun
 define <vscale x 8 x i8> @loadnxv8i8(ptr %p) {
 ; CHECK-LABEL: loadnxv8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
 ; CHECK-NEXT:    ldrb w8, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    mov z0.h, p0/m, w8
 ; CHECK-NEXT:    ret
   %l = load i8, ptr %p
@@ -631,9 +631,9 @@ define <vscale x 16 x i8> @loadnxv16i8(ptr %p) {
 define <vscale x 4 x i16> @loadnxv4i16(ptr %p) {
 ; CHECK-LABEL: loadnxv4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    mov z0.s, #0 // =0x0
 ; CHECK-NEXT:    ldrh w8, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    mov z0.s, p0/m, w8
 ; CHECK-NEXT:    ret
   %l = load i16, ptr %p
@@ -654,9 +654,9 @@ define <vscale x 8 x i16> @loadnxv8i16(ptr %p) {
 define <vscale x 2 x i32> @loadnxv2i32(ptr %p) {
 ; CHECK-LABEL: loadnxv2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, #0 // =0x0
 ; CHECK-NEXT:    ldr w8, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, p0/m, x8
 ; CHECK-NEXT:    ret
   %l = load i32, ptr %p
@@ -688,9 +688,9 @@ define <vscale x 2 x i64> @loadnxv2i64(ptr %p) {
 define <vscale x 4 x half> @loadnxv4f16(ptr %p) {
 ; CHECK-LABEL: loadnxv4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
@@ -715,9 +715,9 @@ define <vscale x 8 x half> @loadnxv8f16(ptr %p) {
 define <vscale x 4 x bfloat> @loadnxv4bf16(ptr %p) {
 ; CHECK-LABEL: loadnxv4bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
@@ -742,9 +742,9 @@ define <vscale x 8 x bfloat> @loadnxv8bf16(ptr %p) {
 define <vscale x 2 x float> @loadnxv2f32(ptr %p) {
 ; CHECK-LABEL: loadnxv2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    mov z0.s, #0 // =0x0
@@ -782,9 +782,9 @@ define <vscale x 2 x double> @loadnxv2f64(ptr %p) {
 define <vscale x 8 x i8> @loadnxv8i8_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv8i8_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
 ; CHECK-NEXT:    ldrb w8, [x0, #1]
+; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    mov z0.h, p0/m, w8
 ; CHECK-NEXT:    ret
   %g = getelementptr inbounds i8, ptr %p, i64 1
@@ -807,9 +807,9 @@ define <vscale x 16 x i8> @loadnxv16i8_offset(ptr %p) {
 define <vscale x 4 x i16> @loadnxv4i16_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv4i16_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    mov z0.s, #0 // =0x0
 ; CHECK-NEXT:    ldurh w8, [x0, #1]
+; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    mov z0.s, p0/m, w8
 ; CHECK-NEXT:    ret
   %g = getelementptr inbounds i8, ptr %p, i64 1
@@ -832,9 +832,9 @@ define <vscale x 8 x i16> @loadnxv8i16_offset(ptr %p) {
 define <vscale x 2 x i32> @loadnxv2i32_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv2i32_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, #0 // =0x0
 ; CHECK-NEXT:    ldur w8, [x0, #1]
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, p0/m, x8
 ; CHECK-NEXT:    ret
   %g = getelementptr inbounds i8, ptr %p, i64 1
@@ -869,9 +869,9 @@ define <vscale x 2 x i64> @loadnxv2i64_offset(ptr %p) {
 define <vscale x 4 x half> @loadnxv4f16_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv4f16_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
@@ -898,9 +898,9 @@ define <vscale x 8 x half> @loadnxv8f16_offset(ptr %p) {
 define <vscale x 4 x bfloat> @loadnxv4bf16_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv4bf16_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
@@ -927,9 +927,9 @@ define <vscale x 8 x bfloat> @loadnxv8bf16_offset(ptr %p) {
 define <vscale x 2 x float> @loadnxv2f32_offset(ptr %p) {
 ; CHECK-LABEL: loadnxv2f32_offset:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    mov z0.s, #0 // =0x0
diff --git a/llvm/test/CodeGen/AArch64/logic-shift.ll b/llvm/test/CodeGen/AArch64/logic-shift.ll
index 39f82dd4593fb..31047954401cf 100644
--- a/llvm/test/CodeGen/AArch64/logic-shift.ll
+++ b/llvm/test/CodeGen/AArch64/logic-shift.ll
@@ -34,9 +34,9 @@ define i32 @or_lshr_commute1(i32 %x0, i32 %x1, i32 %y, i32 %z) {
 define <8 x i16> @or_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y, <8 x i16> %z) {
 ; CHECK-LABEL: or_lshr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.8h, v2.8h
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.8h, v0.8h, v2.8h
+; CHECK-NEXT:    neg v1.8h, v2.8h
+; CHECK-NEXT:    ushl v0.8h, v0.8h, v1.8h
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <8 x i16> %x0, %y
@@ -49,9 +49,9 @@ define <8 x i16> @or_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y, <
 define <2 x i64> @or_lshr_commute3(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %y, <2 x i64> %z) {
 ; CHECK-LABEL: or_lshr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.2d, v2.2d
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.2d, v0.2d, v2.2d
+; CHECK-NEXT:    neg v1.2d, v2.2d
+; CHECK-NEXT:    ushl v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <2 x i64> %x0, %y
@@ -94,9 +94,9 @@ define i64 @or_ashr_commute1(i64 %x0, i64 %x1, i64 %y, i64 %z) {
 define <4 x i32> @or_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y, <4 x i32> %z) {
 ; CHECK-LABEL: or_ashr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.4s, v2.4s
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    neg v1.4s, v2.4s
+; CHECK-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <4 x i32> %x0, %y
@@ -109,9 +109,9 @@ define <4 x i32> @or_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y, <
 define <16 x i8> @or_ashr_commute3(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %y, <16 x i8> %z) {
 ; CHECK-LABEL: or_ashr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.16b, v2.16b
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    neg v1.16b, v2.16b
+; CHECK-NEXT:    sshl v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <16 x i8> %x0, %y
@@ -262,9 +262,9 @@ define i32 @xor_lshr_commute1(i32 %x0, i32 %x1, i32 %y, i32 %z) {
 define <8 x i16> @xor_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y, <8 x i16> %z) {
 ; CHECK-LABEL: xor_lshr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.8h, v2.8h
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.8h, v0.8h, v2.8h
+; CHECK-NEXT:    neg v1.8h, v2.8h
+; CHECK-NEXT:    ushl v0.8h, v0.8h, v1.8h
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <8 x i16> %x0, %y
@@ -277,9 +277,9 @@ define <8 x i16> @xor_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y,
 define <2 x i64> @xor_lshr_commute3(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %y, <2 x i64> %z) {
 ; CHECK-LABEL: xor_lshr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.2d, v2.2d
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.2d, v0.2d, v2.2d
+; CHECK-NEXT:    neg v1.2d, v2.2d
+; CHECK-NEXT:    ushl v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <2 x i64> %x0, %y
@@ -322,9 +322,9 @@ define i64 @xor_ashr_commute1(i64 %x0, i64 %x1, i64 %y, i64 %z) {
 define <4 x i32> @xor_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y, <4 x i32> %z) {
 ; CHECK-LABEL: xor_ashr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.4s, v2.4s
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    neg v1.4s, v2.4s
+; CHECK-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <4 x i32> %x0, %y
@@ -337,9 +337,9 @@ define <4 x i32> @xor_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y,
 define <16 x i8> @xor_ashr_commute3(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %y, <16 x i8> %z) {
 ; CHECK-LABEL: xor_ashr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.16b, v2.16b
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    neg v1.16b, v2.16b
+; CHECK-NEXT:    sshl v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <16 x i8> %x0, %y
@@ -490,9 +490,9 @@ define i32 @and_lshr_commute1(i32 %x0, i32 %x1, i32 %y, i32 %z) {
 define <8 x i16> @and_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y, <8 x i16> %z) {
 ; CHECK-LABEL: and_lshr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.8h, v2.8h
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.8h, v0.8h, v2.8h
+; CHECK-NEXT:    neg v1.8h, v2.8h
+; CHECK-NEXT:    ushl v0.8h, v0.8h, v1.8h
 ; CHECK-NEXT:    and v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <8 x i16> %x0, %y
@@ -505,9 +505,9 @@ define <8 x i16> @and_lshr_commute2(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %y,
 define <2 x i64> @and_lshr_commute3(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %y, <2 x i64> %z) {
 ; CHECK-LABEL: and_lshr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.2d, v2.2d
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    ushl v0.2d, v0.2d, v2.2d
+; CHECK-NEXT:    neg v1.2d, v2.2d
+; CHECK-NEXT:    ushl v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    and v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = lshr <2 x i64> %x0, %y
@@ -550,9 +550,9 @@ define i64 @and_ashr_commute1(i64 %x0, i64 %x1, i64 %y, i64 %z) {
 define <4 x i32> @and_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y, <4 x i32> %z) {
 ; CHECK-LABEL: and_ashr_commute2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.4s, v2.4s
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    neg v1.4s, v2.4s
+; CHECK-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    and v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <4 x i32> %x0, %y
@@ -565,9 +565,9 @@ define <4 x i32> @and_ashr_commute2(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %y,
 define <16 x i8> @and_ashr_commute3(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %y, <16 x i8> %z) {
 ; CHECK-LABEL: and_ashr_commute3:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    neg v2.16b, v2.16b
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    sshl v0.16b, v0.16b, v2.16b
+; CHECK-NEXT:    neg v1.16b, v2.16b
+; CHECK-NEXT:    sshl v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    and v0.16b, v0.16b, v3.16b
 ; CHECK-NEXT:    ret
   %sh1 = ashr <16 x i8> %x0, %y
diff --git a/llvm/test/CodeGen/AArch64/named-vector-shuffles-sve.ll b/llvm/test/CodeGen/AArch64/named-vector-shuffles-sve.ll
index 06570b4539cc1..fac96e07de541 100644
--- a/llvm/test/CodeGen/AArch64/named-vector-shuffles-sve.ll
+++ b/llvm/test/CodeGen/AArch64/named-vector-shuffles-sve.ll
@@ -258,10 +258,10 @@ define <vscale x 2 x i1> @splice_nxv2i1_idx(<vscale x 2 x i1> %a, <vscale x 2 x
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.d, p0/z, #1 // =0x1
-; CHECK-NEXT:    ptrue p2.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ext z1.b, z1.b, z0.b, #8
 ; CHECK-NEXT:    and z1.d, z1.d, #0x1
-; CHECK-NEXT:    cmpne p0.d, p2/z, z1.d, #0
+; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    ret
   %res = call <vscale x 2 x i1> @llvm.experimental.vector.splice.nxv2i1(<vscale x 2 x i1> %a, <vscale x 2 x i1> %b, i32 1)
   ret <vscale x 2 x i1> %res
@@ -273,10 +273,10 @@ define <vscale x 4 x i1> @splice_nxv4i1_idx(<vscale x 4 x i1> %a, <vscale x 4 x
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.s, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    ptrue p2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ext z1.b, z1.b, z0.b, #8
 ; CHECK-NEXT:    and z1.s, z1.s, #0x1
-; CHECK-NEXT:    cmpne p0.s, p2/z, z1.s, #0
+; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
 ; CHECK-NEXT:    ret
   %res = call <vscale x 4 x i1> @llvm.experimental.vector.splice.nxv4i1(<vscale x 4 x i1> %a, <vscale x 4 x i1> %b, i32 2)
   ret <vscale x 4 x i1> %res
@@ -288,10 +288,10 @@ define <vscale x 8 x i1> @splice_nxv8i1_idx(<vscale x 8 x i1> %a, <vscale x 8 x
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.h, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.h, p0/z, #1 // =0x1
-; CHECK-NEXT:    ptrue p2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ext z1.b, z1.b, z0.b, #8
 ; CHECK-NEXT:    and z1.h, z1.h, #0x1
-; CHECK-NEXT:    cmpne p0.h, p2/z, z1.h, #0
+; CHECK-NEXT:    cmpne p0.h, p0/z, z1.h, #0
 ; CHECK-NEXT:    ret
   %res = call <vscale x 8 x i1> @llvm.experimental.vector.splice.nxv8i1(<vscale x 8 x i1> %a, <vscale x 8 x i1> %b, i32 4)
   ret <vscale x 8 x i1> %res
@@ -303,10 +303,10 @@ define <vscale x 16 x i1> @splice_nxv16i1_idx(<vscale x 16 x i1> %a, <vscale x 1
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.b, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.b, p0/z, #1 // =0x1
-; CHECK-NEXT:    ptrue p2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ext z1.b, z1.b, z0.b, #8
 ; CHECK-NEXT:    and z1.b, z1.b, #0x1
-; CHECK-NEXT:    cmpne p0.b, p2/z, z1.b, #0
+; CHECK-NEXT:    cmpne p0.b, p0/z, z1.b, #0
 ; CHECK-NEXT:    ret
   %res = call <vscale x 16 x i1> @llvm.experimental.vector.splice.nxv16i1(<vscale x 16 x i1> %a, <vscale x 16 x i1> %b, i32 8)
   ret <vscale x 16 x i1> %res
@@ -350,16 +350,16 @@ define <vscale x 16 x float> @splice_nxv16f32_16(<vscale x 16 x float> %a, <vsca
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-8
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    rdvl x8, #1
 ; CHECK-NEXT:    mov w9, #16 // =0x10
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, #16
+; CHECK-NEXT:    st1w { z3.s }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    add x10, x9, x8, lsl #2
-; CHECK-NEXT:    st1w { z3.s }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1w { z2.s }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    add x10, x9, x8, lsl #2
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    st1w { z7.s }, p0, [sp, #7, mul vl]
@@ -452,16 +452,16 @@ define <vscale x 16 x i8> @splice_nxv16i8_neg17(<vscale x 16 x i8> %a, <vscale x
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-2
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w9, #17 // =0x11
-; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    cmp x8, #17
-; CHECK-NEXT:    add x10, x10, x8
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    sub x8, x10, x8
+; CHECK-NEXT:    mov x10, sp
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    add x8, x10, x8
 ; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK-NEXT:    st1b { z1.b }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    sub x8, x8, x9
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x8]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -497,16 +497,16 @@ define <vscale x 8 x i16> @splice_nxv8i16_neg9(<vscale x 8 x i16> %a, <vscale x
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-2
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w9, #18 // =0x12
-; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    cmp x8, #18
-; CHECK-NEXT:    add x10, x10, x8
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    sub x8, x10, x8
+; CHECK-NEXT:    mov x10, sp
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    add x8, x10, x8
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    st1h { z1.h }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    sub x8, x8, x9
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x8]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -608,16 +608,16 @@ define <vscale x 8 x half> @splice_nxv8f16_neg9(<vscale x 8 x half> %a, <vscale
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-2
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w9, #18 // =0x12
-; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    cmp x8, #18
-; CHECK-NEXT:    add x10, x10, x8
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    sub x8, x10, x8
+; CHECK-NEXT:    mov x10, sp
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    add x8, x10, x8
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    st1h { z1.h }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    sub x8, x8, x9
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x8]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -699,9 +699,9 @@ define <vscale x 2 x i1> @splice_nxv2i1(<vscale x 2 x i1> %a, <vscale x 2 x i1>
 ; CHECK-NEXT:    ptrue p2.d, vl1
 ; CHECK-NEXT:    mov z0.d, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.d, p0/z, #1 // =0x1
+; CHECK-NEXT:    rev p0.d, p2.d
+; CHECK-NEXT:    splice z1.d, p0, z1.d, z0.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    rev p2.d, p2.d
-; CHECK-NEXT:    splice z1.d, p2, z1.d, z0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0x1
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    ret
@@ -716,9 +716,9 @@ define <vscale x 4 x i1> @splice_nxv4i1(<vscale x 4 x i1> %a, <vscale x 4 x i1>
 ; CHECK-NEXT:    ptrue p2.s, vl1
 ; CHECK-NEXT:    mov z0.s, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.s, p0/z, #1 // =0x1
+; CHECK-NEXT:    rev p0.s, p2.s
+; CHECK-NEXT:    splice z1.s, p0, z1.s, z0.s
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    rev p2.s, p2.s
-; CHECK-NEXT:    splice z1.s, p2, z1.s, z0.s
 ; CHECK-NEXT:    and z1.s, z1.s, #0x1
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
 ; CHECK-NEXT:    ret
@@ -733,9 +733,9 @@ define <vscale x 8 x i1> @splice_nxv8i1(<vscale x 8 x i1> %a, <vscale x 8 x i1>
 ; CHECK-NEXT:    ptrue p2.h, vl1
 ; CHECK-NEXT:    mov z0.h, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.h, p0/z, #1 // =0x1
+; CHECK-NEXT:    rev p0.h, p2.h
+; CHECK-NEXT:    splice z1.h, p0, z1.h, z0.h
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    rev p2.h, p2.h
-; CHECK-NEXT:    splice z1.h, p2, z1.h, z0.h
 ; CHECK-NEXT:    and z1.h, z1.h, #0x1
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z1.h, #0
 ; CHECK-NEXT:    ret
@@ -750,9 +750,9 @@ define <vscale x 16 x i1> @splice_nxv16i1(<vscale x 16 x i1> %a, <vscale x 16 x
 ; CHECK-NEXT:    ptrue p2.b, vl1
 ; CHECK-NEXT:    mov z0.b, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.b, p0/z, #1 // =0x1
+; CHECK-NEXT:    rev p0.b, p2.b
+; CHECK-NEXT:    splice z1.b, p0, z1.b, z0.b
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    rev p2.b, p2.b
-; CHECK-NEXT:    splice z1.b, p2, z1.b, z0.b
 ; CHECK-NEXT:    and z1.b, z1.b, #0x1
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z1.b, #0
 ; CHECK-NEXT:    ret
@@ -783,8 +783,8 @@ define <vscale x 8 x i32> @splice_nxv8i32(<vscale x 8 x i32> %a, <vscale x 8 x i
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    add x8, x9, x8
 ; CHECK-NEXT:    mov x9, #-8 // =0xfffffffffffffff8
-; CHECK-NEXT:    sub x10, x8, #32
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    sub x10, x8, #32
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    st1w { z3.s }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1w { z2.s }, p0, [sp, #2, mul vl]
@@ -803,16 +803,16 @@ define <vscale x 16 x float> @splice_nxv16f32_neg17(<vscale x 16 x float> %a, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-8
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    rdvl x8, #4
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w9, #68 // =0x44
-; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    cmp x8, #68
+; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    csel x9, x8, x9, lo
 ; CHECK-NEXT:    add x8, x10, x8
-; CHECK-NEXT:    sub x8, x8, x9
 ; CHECK-NEXT:    st1w { z3.s }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1w { z2.s }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    sub x8, x8, x9
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    st1w { z7.s }, p0, [sp, #7, mul vl]
diff --git a/llvm/test/CodeGen/AArch64/neon-dotreduce.ll b/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
index 706aa4ad1b466..736f66c935e74 100644
--- a/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
+++ b/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
@@ -845,9 +845,9 @@ define i32 @test_sdot_v24i8_double(<24 x i8> %a, <24 x i8> %b, <24 x i8> %c, <24
 ; CHECK-NEXT:    ld1 { v5.b }[15], [x8]
 ; CHECK-NEXT:    ld1 { v7.b }[7], [x10]
 ; CHECK-NEXT:    addp v1.2s, v19.2s, v19.2s
+; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    sdot v16.4s, v5.16b, v4.16b
 ; CHECK-NEXT:    sdot v18.2s, v7.8b, v6.8b
-; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    fmov w9, s1
 ; CHECK-NEXT:    addv s2, v16.4s
 ; CHECK-NEXT:    addp v3.2s, v18.2s, v18.2s
@@ -975,8 +975,8 @@ define i32 @test_sdot_v24i8_double_nomla(<24 x i8> %a, <24 x i8> %b, <24 x i8> %
 ; CHECK-NEXT:    addv s3, v5.4s
 ; CHECK-NEXT:    addp v1.2s, v17.2s, v17.2s
 ; CHECK-NEXT:    addp v2.2s, v7.2s, v7.2s
-; CHECK-NEXT:    addv s0, v6.4s
 ; CHECK-NEXT:    fmov w10, s3
+; CHECK-NEXT:    addv s0, v6.4s
 ; CHECK-NEXT:    fmov w9, s1
 ; CHECK-NEXT:    fmov w11, s2
 ; CHECK-NEXT:    fmov w8, s0
@@ -998,26 +998,26 @@ entry:
 define i32 @test_udot_v25i8(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %sum) {
 ; CHECK-LABEL: test_udot_v25i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp q4, q0, [x0]
+; CHECK-NEXT:    ldp q2, q0, [x0]
 ; CHECK-NEXT:    ldp q5, q1, [x1]
-; CHECK-NEXT:    ushll2 v2.8h, v0.16b, #0
-; CHECK-NEXT:    ushll v6.8h, v4.8b, #0
+; CHECK-NEXT:    ushll2 v3.8h, v0.16b, #0
+; CHECK-NEXT:    ushll v6.8h, v2.8b, #0
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
-; CHECK-NEXT:    ushll2 v3.8h, v1.16b, #0
+; CHECK-NEXT:    ushll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    ushll v7.8h, v5.8b, #0
 ; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    umull v2.4s, v3.4h, v2.4h
-; CHECK-NEXT:    movi v3.2d, #0000000000000000
+; CHECK-NEXT:    ushll2 v2.8h, v2.16b, #0
+; CHECK-NEXT:    umull v3.4s, v4.4h, v3.4h
+; CHECK-NEXT:    movi v4.2d, #0000000000000000
 ; CHECK-NEXT:    umull2 v16.4s, v7.8h, v6.8h
 ; CHECK-NEXT:    umull v6.4s, v7.4h, v6.4h
-; CHECK-NEXT:    mov v3.s[0], v2.s[0]
-; CHECK-NEXT:    ushll2 v2.8h, v4.16b, #0
-; CHECK-NEXT:    ushll2 v4.8h, v5.16b, #0
-; CHECK-NEXT:    umlal v6.4s, v1.4h, v0.4h
+; CHECK-NEXT:    mov v4.s[0], v3.s[0]
+; CHECK-NEXT:    ushll2 v3.8h, v5.16b, #0
 ; CHECK-NEXT:    umlal2 v16.4s, v1.8h, v0.8h
-; CHECK-NEXT:    umlal v3.4s, v4.4h, v2.4h
-; CHECK-NEXT:    umlal2 v16.4s, v4.8h, v2.8h
-; CHECK-NEXT:    add v0.4s, v6.4s, v3.4s
+; CHECK-NEXT:    umlal v6.4s, v1.4h, v0.4h
+; CHECK-NEXT:    umlal v4.4s, v3.4h, v2.4h
+; CHECK-NEXT:    umlal2 v16.4s, v3.8h, v2.8h
+; CHECK-NEXT:    add v0.4s, v6.4s, v4.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v16.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
@@ -1039,17 +1039,17 @@ define i32 @test_udot_v25i8_nomla(ptr nocapture readonly %a1) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ldp q2, q1, [x0]
 ; CHECK-NEXT:    movi v0.2d, #0000000000000000
-; CHECK-NEXT:    ushll v3.8h, v1.8b, #0
+; CHECK-NEXT:    ushll2 v3.8h, v1.16b, #0
+; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
 ; CHECK-NEXT:    ushll v4.8h, v2.8b, #0
-; CHECK-NEXT:    ushll2 v1.8h, v1.16b, #0
 ; CHECK-NEXT:    ushll2 v2.8h, v2.16b, #0
-; CHECK-NEXT:    uaddl2 v5.4s, v4.8h, v3.8h
-; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    uaddl v3.4s, v4.4h, v3.4h
-; CHECK-NEXT:    mov v0.s[0], v1.s[0]
-; CHECK-NEXT:    uaddw2 v1.4s, v5.4s, v2.8h
+; CHECK-NEXT:    ushll v3.4s, v3.4h, #0
+; CHECK-NEXT:    uaddl2 v5.4s, v4.8h, v1.8h
+; CHECK-NEXT:    uaddl v1.4s, v4.4h, v1.4h
+; CHECK-NEXT:    mov v0.s[0], v3.s[0]
+; CHECK-NEXT:    uaddw2 v3.4s, v5.4s, v2.8h
+; CHECK-NEXT:    add v1.4s, v1.4s, v3.4s
 ; CHECK-NEXT:    uaddw v0.4s, v0.4s, v2.4h
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w0, s0
@@ -1063,26 +1063,26 @@ entry:
 define i32 @test_sdot_v25i8(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %sum) {
 ; CHECK-LABEL: test_sdot_v25i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp q4, q0, [x0]
+; CHECK-NEXT:    ldp q2, q0, [x0]
 ; CHECK-NEXT:    ldp q5, q1, [x1]
-; CHECK-NEXT:    sshll2 v2.8h, v0.16b, #0
-; CHECK-NEXT:    sshll v6.8h, v4.8b, #0
+; CHECK-NEXT:    sshll2 v3.8h, v0.16b, #0
+; CHECK-NEXT:    sshll v6.8h, v2.8b, #0
 ; CHECK-NEXT:    sshll v0.8h, v0.8b, #0
-; CHECK-NEXT:    sshll2 v3.8h, v1.16b, #0
+; CHECK-NEXT:    sshll2 v4.8h, v1.16b, #0
 ; CHECK-NEXT:    sshll v7.8h, v5.8b, #0
 ; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
-; CHECK-NEXT:    smull v2.4s, v3.4h, v2.4h
-; CHECK-NEXT:    movi v3.2d, #0000000000000000
+; CHECK-NEXT:    sshll2 v2.8h, v2.16b, #0
+; CHECK-NEXT:    smull v3.4s, v4.4h, v3.4h
+; CHECK-NEXT:    movi v4.2d, #0000000000000000
 ; CHECK-NEXT:    smull2 v16.4s, v7.8h, v6.8h
 ; CHECK-NEXT:    smull v6.4s, v7.4h, v6.4h
-; CHECK-NEXT:    mov v3.s[0], v2.s[0]
-; CHECK-NEXT:    sshll2 v2.8h, v4.16b, #0
-; CHECK-NEXT:    sshll2 v4.8h, v5.16b, #0
-; CHECK-NEXT:    smlal v6.4s, v1.4h, v0.4h
+; CHECK-NEXT:    mov v4.s[0], v3.s[0]
+; CHECK-NEXT:    sshll2 v3.8h, v5.16b, #0
 ; CHECK-NEXT:    smlal2 v16.4s, v1.8h, v0.8h
-; CHECK-NEXT:    smlal v3.4s, v4.4h, v2.4h
-; CHECK-NEXT:    smlal2 v16.4s, v4.8h, v2.8h
-; CHECK-NEXT:    add v0.4s, v6.4s, v3.4s
+; CHECK-NEXT:    smlal v6.4s, v1.4h, v0.4h
+; CHECK-NEXT:    smlal v4.4s, v3.4h, v2.4h
+; CHECK-NEXT:    smlal2 v16.4s, v3.8h, v2.8h
+; CHECK-NEXT:    add v0.4s, v6.4s, v4.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v16.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
@@ -1105,222 +1105,222 @@ define i32 @test_sdot_v25i8_double(<25 x i8> %a, <25 x i8> %b, <25 x i8> %c, <25
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    fmov s4, w0
-; CHECK-NEXT:    ldr b0, [sp, #80]
-; CHECK-NEXT:    add x8, sp, #88
 ; CHECK-NEXT:    ldr b1, [sp, #16]
-; CHECK-NEXT:    add x10, sp, #24
-; CHECK-NEXT:    ldr b2, [sp, #280]
-; CHECK-NEXT:    ld1 { v0.b }[1], [x8]
+; CHECK-NEXT:    ldr b0, [sp, #80]
+; CHECK-NEXT:    add x11, sp, #24
 ; CHECK-NEXT:    ldr b3, [sp, #216]
+; CHECK-NEXT:    add x10, sp, #88
+; CHECK-NEXT:    ldr b2, [sp, #280]
+; CHECK-NEXT:    ld1 { v1.b }[1], [x11]
 ; CHECK-NEXT:    add x11, sp, #224
-; CHECK-NEXT:    mov v4.b[1], w1
-; CHECK-NEXT:    ld1 { v1.b }[1], [x10]
+; CHECK-NEXT:    ldr b4, [sp, #152]
+; CHECK-NEXT:    ldr b6, [sp, #480]
+; CHECK-NEXT:    ld1 { v0.b }[1], [x10]
 ; CHECK-NEXT:    add x10, sp, #288
-; CHECK-NEXT:    ldr b5, [sp, #152]
-; CHECK-NEXT:    add x9, sp, #96
-; CHECK-NEXT:    ld1 { v2.b }[1], [x10]
+; CHECK-NEXT:    add x12, sp, #160
 ; CHECK-NEXT:    ld1 { v3.b }[1], [x11]
-; CHECK-NEXT:    add x10, sp, #160
-; CHECK-NEXT:    ld1 { v0.b }[2], [x9]
-; CHECK-NEXT:    ld1 { v5.b }[1], [x10]
-; CHECK-NEXT:    add x10, sp, #32
-; CHECK-NEXT:    add x11, sp, #296
-; CHECK-NEXT:    mov v4.b[2], w2
-; CHECK-NEXT:    ld1 { v1.b }[2], [x10]
-; CHECK-NEXT:    add x10, sp, #232
+; CHECK-NEXT:    add x11, sp, #488
+; CHECK-NEXT:    ld1 { v2.b }[1], [x10]
+; CHECK-NEXT:    ld1 { v4.b }[1], [x12]
+; CHECK-NEXT:    ld1 { v6.b }[1], [x11]
+; CHECK-NEXT:    add x11, sp, #32
+; CHECK-NEXT:    add x9, sp, #96
 ; CHECK-NEXT:    add x8, sp, #104
-; CHECK-NEXT:    ld1 { v2.b }[2], [x11]
-; CHECK-NEXT:    ld1 { v3.b }[2], [x10]
+; CHECK-NEXT:    ld1 { v1.b }[2], [x11]
+; CHECK-NEXT:    add x11, sp, #232
+; CHECK-NEXT:    ld1 { v0.b }[2], [x9]
+; CHECK-NEXT:    add x9, sp, #296
+; CHECK-NEXT:    ld1 { v3.b }[2], [x11]
 ; CHECK-NEXT:    add x11, sp, #168
+; CHECK-NEXT:    ld1 { v2.b }[2], [x9]
+; CHECK-NEXT:    ld1 { v4.b }[2], [x11]
+; CHECK-NEXT:    add x11, sp, #40
+; CHECK-NEXT:    ld1 { v1.b }[3], [x11]
 ; CHECK-NEXT:    ld1 { v0.b }[3], [x8]
-; CHECK-NEXT:    add x8, sp, #40
-; CHECK-NEXT:    ld1 { v5.b }[2], [x11]
-; CHECK-NEXT:    ld1 { v1.b }[3], [x8]
-; CHECK-NEXT:    add x8, sp, #240
-; CHECK-NEXT:    mov v4.b[3], w3
-; CHECK-NEXT:    ld1 { v3.b }[3], [x8]
-; CHECK-NEXT:    add x8, sp, #176
-; CHECK-NEXT:    add x12, sp, #112
-; CHECK-NEXT:    add x13, sp, #48
-; CHECK-NEXT:    add x9, sp, #120
-; CHECK-NEXT:    ld1 { v5.b }[3], [x8]
-; CHECK-NEXT:    ld1 { v0.b }[4], [x12]
-; CHECK-NEXT:    add x12, sp, #184
-; CHECK-NEXT:    ld1 { v1.b }[4], [x13]
-; CHECK-NEXT:    add x15, sp, #56
-; CHECK-NEXT:    add x14, sp, #128
-; CHECK-NEXT:    mov v4.b[4], w4
-; CHECK-NEXT:    add x11, sp, #304
-; CHECK-NEXT:    add x13, sp, #256
-; CHECK-NEXT:    ld1 { v5.b }[4], [x12]
-; CHECK-NEXT:    ld1 { v0.b }[5], [x9]
-; CHECK-NEXT:    add x9, sp, #192
-; CHECK-NEXT:    add x12, sp, #248
-; CHECK-NEXT:    ld1 { v1.b }[5], [x15]
-; CHECK-NEXT:    add x15, sp, #200
-; CHECK-NEXT:    ld1 { v3.b }[4], [x12]
-; CHECK-NEXT:    ld1 { v2.b }[3], [x11]
+; CHECK-NEXT:    add x8, sp, #304
+; CHECK-NEXT:    add x10, sp, #112
+; CHECK-NEXT:    add x11, sp, #240
+; CHECK-NEXT:    add x13, sp, #56
+; CHECK-NEXT:    ld1 { v2.b }[3], [x8]
+; CHECK-NEXT:    add x8, sp, #48
+; CHECK-NEXT:    ld1 { v3.b }[3], [x11]
+; CHECK-NEXT:    ld1 { v1.b }[4], [x8]
+; CHECK-NEXT:    ld1 { v0.b }[4], [x10]
+; CHECK-NEXT:    add x15, sp, #312
+; CHECK-NEXT:    add x12, sp, #120
+; CHECK-NEXT:    add x8, sp, #248
 ; CHECK-NEXT:    add x11, sp, #64
-; CHECK-NEXT:    mov v4.b[5], w5
-; CHECK-NEXT:    ld1 { v5.b }[5], [x9]
-; CHECK-NEXT:    ld1 { v0.b }[6], [x14]
-; CHECK-NEXT:    ldr b6, [sp, #352]
-; CHECK-NEXT:    add x10, sp, #136
-; CHECK-NEXT:    ld1 { v1.b }[6], [x11]
-; CHECK-NEXT:    add x11, sp, #360
-; CHECK-NEXT:    ld1 { v3.b }[5], [x13]
+; CHECK-NEXT:    ld1 { v2.b }[4], [x15]
+; CHECK-NEXT:    ld1 { v3.b }[4], [x8]
+; CHECK-NEXT:    add x15, sp, #320
+; CHECK-NEXT:    ld1 { v1.b }[5], [x13]
+; CHECK-NEXT:    ld1 { v0.b }[5], [x12]
 ; CHECK-NEXT:    ldr b18, [sp, #552]
-; CHECK-NEXT:    ld1 { v5.b }[6], [x15]
-; CHECK-NEXT:    add x14, sp, #208
-; CHECK-NEXT:    ld1 { v6.b }[1], [x11]
-; CHECK-NEXT:    mov v4.b[6], w6
-; CHECK-NEXT:    ld1 { v0.b }[7], [x10]
-; CHECK-NEXT:    add x10, sp, #560
-; CHECK-NEXT:    add x9, sp, #264
-; CHECK-NEXT:    ld1 { v18.b }[1], [x10]
+; CHECK-NEXT:    add x14, sp, #128
+; CHECK-NEXT:    add x16, sp, #256
+; CHECK-NEXT:    ldr b16, [sp, #352]
+; CHECK-NEXT:    ld1 { v2.b }[5], [x15]
+; CHECK-NEXT:    add x15, sp, #176
+; CHECK-NEXT:    ld1 { v3.b }[5], [x16]
+; CHECK-NEXT:    ld1 { v1.b }[6], [x11]
+; CHECK-NEXT:    add x11, sp, #560
+; CHECK-NEXT:    ld1 { v0.b }[6], [x14]
+; CHECK-NEXT:    add x16, sp, #360
+; CHECK-NEXT:    ld1 { v4.b }[3], [x15]
+; CHECK-NEXT:    ld1 { v18.b }[1], [x11]
+; CHECK-NEXT:    add x10, sp, #72
+; CHECK-NEXT:    ld1 { v16.b }[1], [x16]
+; CHECK-NEXT:    add x9, sp, #136
+; CHECK-NEXT:    add x14, sp, #184
+; CHECK-NEXT:    ld1 { v1.b }[7], [x10]
 ; CHECK-NEXT:    add x10, sp, #568
-; CHECK-NEXT:    ld1 { v5.b }[7], [x14]
-; CHECK-NEXT:    ld1 { v3.b }[6], [x9]
+; CHECK-NEXT:    ld1 { v0.b }[7], [x9]
+; CHECK-NEXT:    ld1 { v4.b }[4], [x14]
 ; CHECK-NEXT:    add x9, sp, #368
-; CHECK-NEXT:    ld1 { v6.b }[2], [x9]
-; CHECK-NEXT:    add x11, sp, #488
-; CHECK-NEXT:    ldr b7, [sp, #144]
-; CHECK-NEXT:    mov v4.b[7], w7
 ; CHECK-NEXT:    ld1 { v18.b }[2], [x10]
-; CHECK-NEXT:    add x10, sp, #376
-; CHECK-NEXT:    sshll v17.8h, v5.8b, #0
-; CHECK-NEXT:    ldr b5, [sp, #480]
-; CHECK-NEXT:    sshll v7.8h, v7.8b, #0
-; CHECK-NEXT:    ld1 { v6.b }[3], [x10]
+; CHECK-NEXT:    add x11, sp, #496
+; CHECK-NEXT:    ld1 { v16.b }[2], [x9]
+; CHECK-NEXT:    fmov s5, w0
+; CHECK-NEXT:    add x9, sp, #192
+; CHECK-NEXT:    ld1 { v6.b }[2], [x11]
 ; CHECK-NEXT:    add x10, sp, #576
-; CHECK-NEXT:    add x8, sp, #312
-; CHECK-NEXT:    ld1 { v5.b }[1], [x11]
+; CHECK-NEXT:    ld1 { v4.b }[5], [x9]
+; CHECK-NEXT:    add x9, sp, #376
 ; CHECK-NEXT:    ld1 { v18.b }[3], [x10]
-; CHECK-NEXT:    add x11, sp, #496
-; CHECK-NEXT:    sshll v16.8h, v4.8b, #0
-; CHECK-NEXT:    ldr b4, [sp, #344]
-; CHECK-NEXT:    add x10, sp, #384
-; CHECK-NEXT:    ld1 { v6.b }[4], [x10]
+; CHECK-NEXT:    add x11, sp, #504
+; CHECK-NEXT:    ld1 { v16.b }[3], [x9]
+; CHECK-NEXT:    mov v5.b[1], w1
+; CHECK-NEXT:    ldr b7, [sp, #144]
+; CHECK-NEXT:    ldr b17, [sp, #344]
+; CHECK-NEXT:    add x9, sp, #200
+; CHECK-NEXT:    ld1 { v6.b }[3], [x11]
 ; CHECK-NEXT:    add x10, sp, #584
-; CHECK-NEXT:    ld1 { v2.b }[4], [x8]
-; CHECK-NEXT:    sshll v19.8h, v4.8b, #0
-; CHECK-NEXT:    ld1 { v5.b }[2], [x11]
+; CHECK-NEXT:    ld1 { v4.b }[6], [x9]
+; CHECK-NEXT:    add x9, sp, #384
 ; CHECK-NEXT:    ld1 { v18.b }[4], [x10]
-; CHECK-NEXT:    smull2 v4.4s, v16.8h, v17.8h
-; CHECK-NEXT:    smull v16.4s, v16.4h, v17.4h
-; CHECK-NEXT:    ldr b17, [sp, #416]
-; CHECK-NEXT:    add x11, sp, #504
-; CHECK-NEXT:    add x10, sp, #424
-; CHECK-NEXT:    add x16, sp, #320
-; CHECK-NEXT:    smull v19.4s, v7.4h, v19.4h
-; CHECK-NEXT:    movi v7.2d, #0000000000000000
-; CHECK-NEXT:    ld1 { v5.b }[3], [x11]
-; CHECK-NEXT:    add x11, sp, #392
-; CHECK-NEXT:    ld1 { v17.b }[1], [x10]
-; CHECK-NEXT:    add x10, sp, #592
-; CHECK-NEXT:    ld1 { v2.b }[5], [x16]
-; CHECK-NEXT:    ld1 { v6.b }[5], [x11]
-; CHECK-NEXT:    ld1 { v18.b }[5], [x10]
+; CHECK-NEXT:    sshll v7.8h, v7.8b, #0
+; CHECK-NEXT:    sshll v17.8h, v17.8b, #0
 ; CHECK-NEXT:    add x11, sp, #512
-; CHECK-NEXT:    add x10, sp, #432
+; CHECK-NEXT:    ld1 { v16.b }[4], [x9]
+; CHECK-NEXT:    ld1 { v6.b }[4], [x11]
+; CHECK-NEXT:    add x11, sp, #592
+; CHECK-NEXT:    mov v5.b[2], w2
+; CHECK-NEXT:    add x10, sp, #392
+; CHECK-NEXT:    ldr b19, [sp, #680]
+; CHECK-NEXT:    ld1 { v18.b }[5], [x11]
+; CHECK-NEXT:    smull v7.4s, v7.4h, v17.4h
+; CHECK-NEXT:    ldr b17, [sp, #416]
+; CHECK-NEXT:    ld1 { v16.b }[5], [x10]
+; CHECK-NEXT:    add x10, sp, #688
 ; CHECK-NEXT:    add x12, sp, #328
-; CHECK-NEXT:    mov v7.s[0], v19.s[0]
-; CHECK-NEXT:    ld1 { v5.b }[4], [x11]
-; CHECK-NEXT:    add x11, sp, #400
-; CHECK-NEXT:    ld1 { v17.b }[2], [x10]
+; CHECK-NEXT:    add x9, sp, #424
+; CHECK-NEXT:    ld1 { v19.b }[1], [x10]
 ; CHECK-NEXT:    add x10, sp, #600
-; CHECK-NEXT:    ldr b19, [sp, #680]
 ; CHECK-NEXT:    ldr b20, [sp, #616]
 ; CHECK-NEXT:    ld1 { v2.b }[6], [x12]
-; CHECK-NEXT:    ld1 { v6.b }[6], [x11]
+; CHECK-NEXT:    ld1 { v17.b }[1], [x9]
+; CHECK-NEXT:    add x11, sp, #400
 ; CHECK-NEXT:    ld1 { v18.b }[6], [x10]
-; CHECK-NEXT:    add x11, sp, #688
 ; CHECK-NEXT:    add x12, sp, #624
-; CHECK-NEXT:    ld1 { v19.b }[1], [x11]
+; CHECK-NEXT:    mov v5.b[3], w3
+; CHECK-NEXT:    ld1 { v16.b }[6], [x11]
+; CHECK-NEXT:    add x11, sp, #696
 ; CHECK-NEXT:    ld1 { v20.b }[1], [x12]
-; CHECK-NEXT:    add x10, sp, #408
+; CHECK-NEXT:    add x9, sp, #432
+; CHECK-NEXT:    ld1 { v19.b }[2], [x11]
 ; CHECK-NEXT:    add x11, sp, #608
-; CHECK-NEXT:    add x12, sp, #440
-; CHECK-NEXT:    ld1 { v6.b }[7], [x10]
+; CHECK-NEXT:    ld1 { v17.b }[2], [x9]
+; CHECK-NEXT:    add x10, sp, #408
 ; CHECK-NEXT:    ld1 { v18.b }[7], [x11]
-; CHECK-NEXT:    ld1 { v17.b }[3], [x12]
-; CHECK-NEXT:    add x10, sp, #696
 ; CHECK-NEXT:    add x11, sp, #632
-; CHECK-NEXT:    ld1 { v19.b }[2], [x10]
-; CHECK-NEXT:    add x10, sp, #448
+; CHECK-NEXT:    ld1 { v16.b }[7], [x10]
 ; CHECK-NEXT:    ld1 { v20.b }[2], [x11]
-; CHECK-NEXT:    add x11, sp, #640
-; CHECK-NEXT:    sshll v6.8h, v6.8b, #0
-; CHECK-NEXT:    ld1 { v17.b }[4], [x10]
+; CHECK-NEXT:    mov v5.b[4], w4
 ; CHECK-NEXT:    add x10, sp, #704
-; CHECK-NEXT:    sshll v18.8h, v18.8b, #0
+; CHECK-NEXT:    add x12, sp, #440
 ; CHECK-NEXT:    ld1 { v19.b }[3], [x10]
-; CHECK-NEXT:    add x10, sp, #712
-; CHECK-NEXT:    add x12, sp, #520
-; CHECK-NEXT:    ld1 { v20.b }[3], [x11]
-; CHECK-NEXT:    add x11, sp, #648
-; CHECK-NEXT:    ldr b21, [sp, #544]
-; CHECK-NEXT:    smull2 v22.4s, v6.8h, v18.8h
-; CHECK-NEXT:    smull v6.4s, v6.4h, v18.4h
-; CHECK-NEXT:    ldr b18, [sp, #744]
-; CHECK-NEXT:    ld1 { v19.b }[4], [x10]
-; CHECK-NEXT:    ld1 { v5.b }[5], [x12]
-; CHECK-NEXT:    add x12, sp, #656
-; CHECK-NEXT:    ld1 { v20.b }[4], [x11]
-; CHECK-NEXT:    add x11, sp, #456
-; CHECK-NEXT:    sshll v21.8h, v21.8b, #0
-; CHECK-NEXT:    ld1 { v17.b }[5], [x11]
-; CHECK-NEXT:    add x11, sp, #720
+; CHECK-NEXT:    add x10, sp, #448
+; CHECK-NEXT:    ld1 { v17.b }[3], [x12]
+; CHECK-NEXT:    add x12, sp, #640
+; CHECK-NEXT:    sshll v21.8h, v16.8b, #0
+; CHECK-NEXT:    ld1 { v20.b }[3], [x12]
 ; CHECK-NEXT:    sshll v18.8h, v18.8b, #0
+; CHECK-NEXT:    add x11, sp, #712
+; CHECK-NEXT:    mov v5.b[5], w5
+; CHECK-NEXT:    ld1 { v19.b }[4], [x11]
+; CHECK-NEXT:    add x9, sp, #520
+; CHECK-NEXT:    ld1 { v17.b }[4], [x10]
+; CHECK-NEXT:    add x10, sp, #648
+; CHECK-NEXT:    ldr b22, [sp, #544]
+; CHECK-NEXT:    ld1 { v20.b }[4], [x10]
+; CHECK-NEXT:    smull2 v16.4s, v21.8h, v18.8h
+; CHECK-NEXT:    smull v18.4s, v21.4h, v18.4h
+; CHECK-NEXT:    ldr b21, [sp, #744]
+; CHECK-NEXT:    add x11, sp, #720
+; CHECK-NEXT:    ld1 { v6.b }[5], [x9]
+; CHECK-NEXT:    add x9, sp, #456
 ; CHECK-NEXT:    ld1 { v19.b }[5], [x11]
+; CHECK-NEXT:    mov v5.b[6], w6
+; CHECK-NEXT:    ld1 { v17.b }[5], [x9]
+; CHECK-NEXT:    add x9, sp, #656
+; CHECK-NEXT:    sshll v22.8h, v22.8b, #0
+; CHECK-NEXT:    sshll v21.8h, v21.8b, #0
+; CHECK-NEXT:    ld1 { v20.b }[5], [x9]
 ; CHECK-NEXT:    add x10, sp, #528
-; CHECK-NEXT:    add x11, sp, #464
-; CHECK-NEXT:    ld1 { v20.b }[5], [x12]
-; CHECK-NEXT:    ld1 { v5.b }[6], [x10]
-; CHECK-NEXT:    add x12, sp, #728
-; CHECK-NEXT:    add x13, sp, #664
-; CHECK-NEXT:    add x8, sp, #72
-; CHECK-NEXT:    ld1 { v17.b }[6], [x11]
-; CHECK-NEXT:    ld1 { v19.b }[6], [x12]
-; CHECK-NEXT:    ld1 { v1.b }[7], [x8]
+; CHECK-NEXT:    add x11, sp, #728
+; CHECK-NEXT:    ld1 { v6.b }[6], [x10]
+; CHECK-NEXT:    add x10, sp, #464
+; CHECK-NEXT:    ld1 { v19.b }[6], [x11]
+; CHECK-NEXT:    add x11, sp, #664
+; CHECK-NEXT:    ld1 { v17.b }[6], [x10]
+; CHECK-NEXT:    smull v21.4s, v22.4h, v21.4h
+; CHECK-NEXT:    movi v22.2d, #0000000000000000
+; CHECK-NEXT:    ld1 { v20.b }[6], [x11]
+; CHECK-NEXT:    mov v5.b[7], w7
+; CHECK-NEXT:    add x9, sp, #536
+; CHECK-NEXT:    add x10, sp, #736
+; CHECK-NEXT:    add x11, sp, #208
+; CHECK-NEXT:    add x13, sp, #264
+; CHECK-NEXT:    ld1 { v6.b }[7], [x9]
+; CHECK-NEXT:    ld1 { v19.b }[7], [x10]
+; CHECK-NEXT:    ld1 { v4.b }[7], [x11]
+; CHECK-NEXT:    add x9, sp, #472
+; CHECK-NEXT:    add x10, sp, #672
+; CHECK-NEXT:    ld1 { v3.b }[6], [x13]
+; CHECK-NEXT:    ld1 { v17.b }[7], [x9]
+; CHECK-NEXT:    ld1 { v20.b }[7], [x10]
 ; CHECK-NEXT:    add x8, sp, #336
-; CHECK-NEXT:    ld1 { v20.b }[6], [x13]
-; CHECK-NEXT:    add x9, sp, #272
-; CHECK-NEXT:    smull v18.4s, v21.4h, v18.4h
+; CHECK-NEXT:    mov v22.s[0], v21.s[0]
 ; CHECK-NEXT:    movi v21.2d, #0000000000000000
-; CHECK-NEXT:    add x10, sp, #536
+; CHECK-NEXT:    sshll v5.8h, v5.8b, #0
+; CHECK-NEXT:    sshll v6.8h, v6.8b, #0
+; CHECK-NEXT:    sshll v19.8h, v19.8b, #0
 ; CHECK-NEXT:    ld1 { v2.b }[7], [x8]
-; CHECK-NEXT:    ld1 { v3.b }[7], [x9]
-; CHECK-NEXT:    ld1 { v5.b }[7], [x10]
-; CHECK-NEXT:    add x8, sp, #472
-; CHECK-NEXT:    add x9, sp, #736
-; CHECK-NEXT:    add x10, sp, #672
-; CHECK-NEXT:    ld1 { v17.b }[7], [x8]
-; CHECK-NEXT:    ld1 { v19.b }[7], [x9]
-; CHECK-NEXT:    ld1 { v20.b }[7], [x10]
+; CHECK-NEXT:    add x8, sp, #272
+; CHECK-NEXT:    sshll v4.8h, v4.8b, #0
+; CHECK-NEXT:    ld1 { v3.b }[7], [x8]
+; CHECK-NEXT:    sshll v17.8h, v17.8b, #0
+; CHECK-NEXT:    sshll v20.8h, v20.8b, #0
 ; CHECK-NEXT:    sshll v0.8h, v0.8b, #0
-; CHECK-NEXT:    mov v21.s[0], v18.s[0]
 ; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
+; CHECK-NEXT:    smlal v18.4s, v6.4h, v19.4h
+; CHECK-NEXT:    smlal2 v16.4s, v6.8h, v19.8h
+; CHECK-NEXT:    mov v21.s[0], v7.s[0]
+; CHECK-NEXT:    smull v6.4s, v5.4h, v4.4h
 ; CHECK-NEXT:    sshll v2.8h, v2.8b, #0
 ; CHECK-NEXT:    sshll v3.8h, v3.8b, #0
-; CHECK-NEXT:    sshll v5.8h, v5.8b, #0
-; CHECK-NEXT:    sshll v17.8h, v17.8b, #0
-; CHECK-NEXT:    sshll v18.8h, v19.8b, #0
-; CHECK-NEXT:    sshll v19.8h, v20.8b, #0
-; CHECK-NEXT:    smlal v16.4s, v0.4h, v2.4h
+; CHECK-NEXT:    smlal v22.4s, v17.4h, v20.4h
+; CHECK-NEXT:    smull2 v4.4s, v5.8h, v4.8h
+; CHECK-NEXT:    smlal v21.4s, v1.4h, v3.4h
+; CHECK-NEXT:    smlal2 v16.4s, v17.8h, v20.8h
+; CHECK-NEXT:    smlal v6.4s, v0.4h, v2.4h
+; CHECK-NEXT:    add v5.4s, v18.4s, v22.4s
 ; CHECK-NEXT:    smlal2 v4.4s, v0.8h, v2.8h
-; CHECK-NEXT:    smlal v7.4s, v1.4h, v3.4h
-; CHECK-NEXT:    smlal v6.4s, v5.4h, v18.4h
-; CHECK-NEXT:    smlal2 v22.4s, v5.8h, v18.8h
-; CHECK-NEXT:    smlal v21.4s, v17.4h, v19.4h
+; CHECK-NEXT:    add v0.4s, v6.4s, v21.4s
+; CHECK-NEXT:    add v2.4s, v5.4s, v16.4s
 ; CHECK-NEXT:    smlal2 v4.4s, v1.8h, v3.8h
-; CHECK-NEXT:    add v0.4s, v16.4s, v7.4s
-; CHECK-NEXT:    add v1.4s, v6.4s, v21.4s
-; CHECK-NEXT:    smlal2 v22.4s, v17.8h, v19.8h
+; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v4.4s
-; CHECK-NEXT:    add v1.4s, v1.4s, v22.4s
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -1584,34 +1584,34 @@ entry:
 define i32 @test_udot_v33i8(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %sum) {
 ; CHECK-LABEL: test_udot_v33i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp q2, q3, [x0]
-; CHECK-NEXT:    movi v18.2d, #0000000000000000
-; CHECK-NEXT:    ldp q4, q5, [x1]
 ; CHECK-NEXT:    ldr b0, [x0, #32]
 ; CHECK-NEXT:    ldr b1, [x1, #32]
+; CHECK-NEXT:    ldp q2, q4, [x0]
+; CHECK-NEXT:    ldp q3, q6, [x1]
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
-; CHECK-NEXT:    ushll v6.8h, v2.8b, #0
-; CHECK-NEXT:    ushll2 v2.8h, v2.16b, #0
 ; CHECK-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-NEXT:    ushll v7.8h, v4.8b, #0
-; CHECK-NEXT:    ushll2 v4.8h, v4.16b, #0
-; CHECK-NEXT:    ushll2 v16.8h, v3.16b, #0
-; CHECK-NEXT:    ushll v3.8h, v3.8b, #0
-; CHECK-NEXT:    ushll2 v19.8h, v5.16b, #0
-; CHECK-NEXT:    ushll v5.8h, v5.8b, #0
+; CHECK-NEXT:    ushll v5.8h, v2.8b, #0
+; CHECK-NEXT:    ushll2 v2.8h, v2.16b, #0
+; CHECK-NEXT:    ushll2 v16.8h, v4.16b, #0
+; CHECK-NEXT:    ushll v7.8h, v3.8b, #0
+; CHECK-NEXT:    ushll2 v3.8h, v3.16b, #0
+; CHECK-NEXT:    ushll v4.8h, v4.8b, #0
 ; CHECK-NEXT:    umull v0.4s, v1.4h, v0.4h
-; CHECK-NEXT:    umull2 v1.4s, v7.8h, v6.8h
-; CHECK-NEXT:    umull2 v17.4s, v4.8h, v2.8h
-; CHECK-NEXT:    umull v2.4s, v4.4h, v2.4h
-; CHECK-NEXT:    umlal2 v17.4s, v19.8h, v16.8h
-; CHECK-NEXT:    umlal2 v1.4s, v5.8h, v3.8h
-; CHECK-NEXT:    mov v18.s[0], v0.s[0]
-; CHECK-NEXT:    umlal v2.4s, v19.4h, v16.4h
-; CHECK-NEXT:    add v0.4s, v1.4s, v17.4s
-; CHECK-NEXT:    umlal v18.4s, v7.4h, v6.4h
-; CHECK-NEXT:    umlal v18.4s, v5.4h, v3.4h
-; CHECK-NEXT:    add v0.4s, v2.4s, v0.4s
-; CHECK-NEXT:    add v0.4s, v18.4s, v0.4s
+; CHECK-NEXT:    movi v1.2d, #0000000000000000
+; CHECK-NEXT:    ushll2 v19.8h, v6.16b, #0
+; CHECK-NEXT:    ushll v6.8h, v6.8b, #0
+; CHECK-NEXT:    umull2 v17.4s, v7.8h, v5.8h
+; CHECK-NEXT:    umull2 v18.4s, v3.8h, v2.8h
+; CHECK-NEXT:    mov v1.s[0], v0.s[0]
+; CHECK-NEXT:    umull v0.4s, v3.4h, v2.4h
+; CHECK-NEXT:    umlal2 v18.4s, v19.8h, v16.8h
+; CHECK-NEXT:    umlal2 v17.4s, v6.8h, v4.8h
+; CHECK-NEXT:    umlal v1.4s, v7.4h, v5.4h
+; CHECK-NEXT:    umlal v0.4s, v19.4h, v16.4h
+; CHECK-NEXT:    add v2.4s, v17.4s, v18.4s
+; CHECK-NEXT:    umlal v1.4s, v6.4h, v4.4h
+; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    add w0, w8, w2
@@ -1639,14 +1639,14 @@ define i32 @test_udot_v33i8_nomla(ptr nocapture readonly %a1) {
 ; CHECK-NEXT:    ushll2 v2.8h, v2.16b, #0
 ; CHECK-NEXT:    ushll2 v3.8h, v3.16b, #0
 ; CHECK-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-NEXT:    uaddl2 v6.4s, v3.8h, v2.8h
-; CHECK-NEXT:    uaddl v2.4s, v3.4h, v2.4h
+; CHECK-NEXT:    uaddl2 v6.4s, v5.8h, v4.8h
 ; CHECK-NEXT:    mov v0.s[0], v1.s[0]
-; CHECK-NEXT:    uaddl2 v1.4s, v5.8h, v4.8h
-; CHECK-NEXT:    add v1.4s, v1.4s, v6.4s
+; CHECK-NEXT:    uaddl2 v1.4s, v3.8h, v2.8h
+; CHECK-NEXT:    uaddl v2.4s, v3.4h, v2.4h
+; CHECK-NEXT:    add v1.4s, v6.4s, v1.4s
 ; CHECK-NEXT:    uaddw v0.4s, v0.4s, v5.4h
-; CHECK-NEXT:    uaddw v0.4s, v0.4s, v4.4h
 ; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
+; CHECK-NEXT:    uaddw v0.4s, v0.4s, v4.4h
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w0, s0
@@ -1660,34 +1660,34 @@ entry:
 define i32 @test_sdot_v33i8(ptr nocapture readonly %a, ptr nocapture readonly %b, i32 %sum) {
 ; CHECK-LABEL: test_sdot_v33i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldp q2, q3, [x0]
-; CHECK-NEXT:    movi v18.2d, #0000000000000000
-; CHECK-NEXT:    ldp q4, q5, [x1]
 ; CHECK-NEXT:    ldr b0, [x0, #32]
 ; CHECK-NEXT:    ldr b1, [x1, #32]
+; CHECK-NEXT:    ldp q2, q4, [x0]
+; CHECK-NEXT:    ldp q3, q6, [x1]
 ; CHECK-NEXT:    sshll v0.8h, v0.8b, #0
-; CHECK-NEXT:    sshll v6.8h, v2.8b, #0
-; CHECK-NEXT:    sshll2 v2.8h, v2.16b, #0
 ; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
-; CHECK-NEXT:    sshll v7.8h, v4.8b, #0
-; CHECK-NEXT:    sshll2 v4.8h, v4.16b, #0
-; CHECK-NEXT:    sshll2 v16.8h, v3.16b, #0
-; CHECK-NEXT:    sshll v3.8h, v3.8b, #0
-; CHECK-NEXT:    sshll2 v19.8h, v5.16b, #0
-; CHECK-NEXT:    sshll v5.8h, v5.8b, #0
+; CHECK-NEXT:    sshll v5.8h, v2.8b, #0
+; CHECK-NEXT:    sshll2 v2.8h, v2.16b, #0
+; CHECK-NEXT:    sshll2 v16.8h, v4.16b, #0
+; CHECK-NEXT:    sshll v7.8h, v3.8b, #0
+; CHECK-NEXT:    sshll2 v3.8h, v3.16b, #0
+; CHECK-NEXT:    sshll v4.8h, v4.8b, #0
 ; CHECK-NEXT:    smull v0.4s, v1.4h, v0.4h
-; CHECK-NEXT:    smull2 v1.4s, v7.8h, v6.8h
-; CHECK-NEXT:    smull2 v17.4s, v4.8h, v2.8h
-; CHECK-NEXT:    smull v2.4s, v4.4h, v2.4h
-; CHECK-NEXT:    smlal2 v17.4s, v19.8h, v16.8h
-; CHECK-NEXT:    smlal2 v1.4s, v5.8h, v3.8h
-; CHECK-NEXT:    mov v18.s[0], v0.s[0]
-; CHECK-NEXT:    smlal v2.4s, v19.4h, v16.4h
-; CHECK-NEXT:    add v0.4s, v1.4s, v17.4s
-; CHECK-NEXT:    smlal v18.4s, v7.4h, v6.4h
-; CHECK-NEXT:    smlal v18.4s, v5.4h, v3.4h
-; CHECK-NEXT:    add v0.4s, v2.4s, v0.4s
-; CHECK-NEXT:    add v0.4s, v18.4s, v0.4s
+; CHECK-NEXT:    movi v1.2d, #0000000000000000
+; CHECK-NEXT:    sshll2 v19.8h, v6.16b, #0
+; CHECK-NEXT:    sshll v6.8h, v6.8b, #0
+; CHECK-NEXT:    smull2 v17.4s, v7.8h, v5.8h
+; CHECK-NEXT:    smull2 v18.4s, v3.8h, v2.8h
+; CHECK-NEXT:    mov v1.s[0], v0.s[0]
+; CHECK-NEXT:    smull v0.4s, v3.4h, v2.4h
+; CHECK-NEXT:    smlal2 v18.4s, v19.8h, v16.8h
+; CHECK-NEXT:    smlal2 v17.4s, v6.8h, v4.8h
+; CHECK-NEXT:    smlal v1.4s, v7.4h, v5.4h
+; CHECK-NEXT:    smlal v0.4s, v19.4h, v16.4h
+; CHECK-NEXT:    add v2.4s, v17.4s, v18.4s
+; CHECK-NEXT:    smlal v1.4s, v6.4h, v4.4h
+; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
+; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    add w0, w8, w2
@@ -2018,151 +2018,151 @@ define i32 @test_sdot_v33i8_double_nomla(<33 x i8> %a, <33 x i8> %b, <33 x i8> %
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ldr b1, [sp, #80]
+; CHECK-NEXT:    ldr b0, [sp, #80]
 ; CHECK-NEXT:    add x8, sp, #88
 ; CHECK-NEXT:    ldr b2, [sp, #144]
-; CHECK-NEXT:    add x9, sp, #152
-; CHECK-NEXT:    ldr b3, [sp, #16]
-; CHECK-NEXT:    add x12, sp, #32
-; CHECK-NEXT:    ld1 { v1.b }[1], [x8]
-; CHECK-NEXT:    ld1 { v2.b }[1], [x9]
 ; CHECK-NEXT:    add x9, sp, #96
-; CHECK-NEXT:    add x8, sp, #24
-; CHECK-NEXT:    add x11, sp, #112
-; CHECK-NEXT:    fmov s0, w0
-; CHECK-NEXT:    ld1 { v3.b }[1], [x8]
-; CHECK-NEXT:    add x8, sp, #160
-; CHECK-NEXT:    ldr b4, [sp, #480]
-; CHECK-NEXT:    ld1 { v1.b }[2], [x9]
-; CHECK-NEXT:    add x9, sp, #104
-; CHECK-NEXT:    ld1 { v2.b }[2], [x8]
+; CHECK-NEXT:    add x10, sp, #152
+; CHECK-NEXT:    add x11, sp, #160
+; CHECK-NEXT:    ld1 { v0.b }[1], [x8]
+; CHECK-NEXT:    ld1 { v2.b }[1], [x10]
+; CHECK-NEXT:    add x8, sp, #104
+; CHECK-NEXT:    ldr b3, [sp, #16]
+; CHECK-NEXT:    ldr b5, [sp, #480]
+; CHECK-NEXT:    fmov s1, w0
+; CHECK-NEXT:    add x10, sp, #24
+; CHECK-NEXT:    add x13, sp, #488
+; CHECK-NEXT:    ldr b4, [sp, #608]
+; CHECK-NEXT:    ld1 { v0.b }[2], [x9]
+; CHECK-NEXT:    ld1 { v2.b }[2], [x11]
+; CHECK-NEXT:    add x12, sp, #112
+; CHECK-NEXT:    ld1 { v3.b }[1], [x10]
+; CHECK-NEXT:    ld1 { v5.b }[1], [x13]
+; CHECK-NEXT:    add x10, sp, #616
+; CHECK-NEXT:    mov v1.b[1], w1
+; CHECK-NEXT:    ld1 { v4.b }[1], [x10]
+; CHECK-NEXT:    add x10, sp, #32
+; CHECK-NEXT:    ld1 { v0.b }[3], [x8]
 ; CHECK-NEXT:    add x8, sp, #168
-; CHECK-NEXT:    add x10, sp, #120
-; CHECK-NEXT:    add x13, sp, #48
-; CHECK-NEXT:    ld1 { v3.b }[2], [x12]
-; CHECK-NEXT:    add x12, sp, #40
-; CHECK-NEXT:    ldr b5, [sp, #608]
-; CHECK-NEXT:    ld1 { v1.b }[3], [x9]
+; CHECK-NEXT:    add x9, sp, #120
 ; CHECK-NEXT:    ld1 { v2.b }[3], [x8]
-; CHECK-NEXT:    mov v0.b[1], w1
-; CHECK-NEXT:    add x9, sp, #128
-; CHECK-NEXT:    add x14, sp, #184
-; CHECK-NEXT:    ldr b16, [sp, #544]
-; CHECK-NEXT:    ld1 { v3.b }[3], [x12]
-; CHECK-NEXT:    add x12, sp, #176
-; CHECK-NEXT:    ldr b17, [sp, #672]
-; CHECK-NEXT:    ld1 { v1.b }[4], [x11]
-; CHECK-NEXT:    add x11, sp, #488
-; CHECK-NEXT:    ld1 { v2.b }[4], [x12]
-; CHECK-NEXT:    ld1 { v4.b }[1], [x11]
-; CHECK-NEXT:    mov v0.b[2], w2
-; CHECK-NEXT:    add x11, sp, #192
-; CHECK-NEXT:    ld1 { v3.b }[4], [x13]
-; CHECK-NEXT:    add x13, sp, #616
-; CHECK-NEXT:    add x12, sp, #56
-; CHECK-NEXT:    ld1 { v1.b }[5], [x10]
-; CHECK-NEXT:    ld1 { v5.b }[1], [x13]
-; CHECK-NEXT:    add x13, sp, #496
+; CHECK-NEXT:    add x8, sp, #496
+; CHECK-NEXT:    ld1 { v3.b }[2], [x10]
+; CHECK-NEXT:    ld1 { v5.b }[2], [x8]
+; CHECK-NEXT:    add x8, sp, #176
+; CHECK-NEXT:    ldr b6, [sp, #544]
+; CHECK-NEXT:    ld1 { v0.b }[4], [x12]
+; CHECK-NEXT:    add x14, sp, #552
+; CHECK-NEXT:    ldr b7, [sp, #672]
+; CHECK-NEXT:    ld1 { v2.b }[4], [x8]
+; CHECK-NEXT:    add x13, sp, #40
+; CHECK-NEXT:    ld1 { v6.b }[1], [x14]
+; CHECK-NEXT:    mov v1.b[2], w2
+; CHECK-NEXT:    add x11, sp, #128
+; CHECK-NEXT:    ld1 { v3.b }[3], [x13]
+; CHECK-NEXT:    ld1 { v0.b }[5], [x9]
+; CHECK-NEXT:    add x9, sp, #680
+; CHECK-NEXT:    add x13, sp, #184
+; CHECK-NEXT:    ld1 { v7.b }[1], [x9]
+; CHECK-NEXT:    ld1 { v2.b }[5], [x13]
+; CHECK-NEXT:    add x13, sp, #624
+; CHECK-NEXT:    add x15, sp, #504
 ; CHECK-NEXT:    ld1 { v4.b }[2], [x13]
-; CHECK-NEXT:    ld1 { v2.b }[5], [x14]
-; CHECK-NEXT:    add x14, sp, #680
-; CHECK-NEXT:    ld1 { v17.b }[1], [x14]
-; CHECK-NEXT:    add x13, sp, #504
-; CHECK-NEXT:    ld1 { v3.b }[5], [x12]
-; CHECK-NEXT:    ld1 { v1.b }[6], [x9]
-; CHECK-NEXT:    add x9, sp, #552
-; CHECK-NEXT:    add x12, sp, #688
-; CHECK-NEXT:    ld1 { v16.b }[1], [x9]
-; CHECK-NEXT:    add x9, sp, #624
-; CHECK-NEXT:    ld1 { v4.b }[3], [x13]
-; CHECK-NEXT:    ld1 { v2.b }[6], [x11]
+; CHECK-NEXT:    add x10, sp, #136
+; CHECK-NEXT:    ld1 { v0.b }[6], [x11]
 ; CHECK-NEXT:    add x11, sp, #560
-; CHECK-NEXT:    add x8, sp, #136
-; CHECK-NEXT:    ld1 { v17.b }[2], [x12]
-; CHECK-NEXT:    ld1 { v5.b }[2], [x9]
-; CHECK-NEXT:    ld1 { v1.b }[7], [x8]
-; CHECK-NEXT:    ld1 { v16.b }[2], [x11]
-; CHECK-NEXT:    add x8, sp, #512
-; CHECK-NEXT:    mov v0.b[3], w3
-; CHECK-NEXT:    ld1 { v4.b }[4], [x8]
-; CHECK-NEXT:    add x8, sp, #568
-; CHECK-NEXT:    add x9, sp, #696
-; CHECK-NEXT:    add x11, sp, #632
-; CHECK-NEXT:    ld1 { v17.b }[3], [x9]
-; CHECK-NEXT:    add x9, sp, #520
-; CHECK-NEXT:    ld1 { v16.b }[3], [x8]
-; CHECK-NEXT:    ld1 { v5.b }[3], [x11]
-; CHECK-NEXT:    add x8, sp, #640
-; CHECK-NEXT:    ld1 { v4.b }[5], [x9]
+; CHECK-NEXT:    ld1 { v5.b }[3], [x15]
+; CHECK-NEXT:    ld1 { v6.b }[2], [x11]
+; CHECK-NEXT:    add x11, sp, #688
+; CHECK-NEXT:    mov v1.b[3], w3
+; CHECK-NEXT:    ld1 { v7.b }[2], [x11]
+; CHECK-NEXT:    add x9, sp, #632
+; CHECK-NEXT:    add x11, sp, #512
+; CHECK-NEXT:    ld1 { v0.b }[7], [x10]
+; CHECK-NEXT:    ld1 { v4.b }[3], [x9]
+; CHECK-NEXT:    add x9, sp, #568
+; CHECK-NEXT:    add x10, sp, #696
+; CHECK-NEXT:    ld1 { v6.b }[3], [x9]
+; CHECK-NEXT:    ld1 { v5.b }[4], [x11]
+; CHECK-NEXT:    ld1 { v7.b }[3], [x10]
+; CHECK-NEXT:    add x9, sp, #640
+; CHECK-NEXT:    mov v1.b[4], w4
+; CHECK-NEXT:    ld1 { v4.b }[4], [x9]
 ; CHECK-NEXT:    add x9, sp, #576
-; CHECK-NEXT:    add x11, sp, #704
+; CHECK-NEXT:    add x10, sp, #704
+; CHECK-NEXT:    add x11, sp, #520
+; CHECK-NEXT:    ld1 { v6.b }[4], [x9]
 ; CHECK-NEXT:    ldr b18, [sp, #736]
-; CHECK-NEXT:    mov v0.b[4], w4
-; CHECK-NEXT:    ld1 { v17.b }[4], [x11]
-; CHECK-NEXT:    ld1 { v16.b }[4], [x9]
-; CHECK-NEXT:    ld1 { v5.b }[4], [x8]
-; CHECK-NEXT:    add x9, sp, #528
-; CHECK-NEXT:    sshll v18.8h, v18.8b, #0
-; CHECK-NEXT:    add x8, sp, #648
+; CHECK-NEXT:    ld1 { v7.b }[4], [x10]
+; CHECK-NEXT:    ld1 { v5.b }[5], [x11]
+; CHECK-NEXT:    add x12, sp, #192
+; CHECK-NEXT:    add x8, sp, #48
+; CHECK-NEXT:    ld1 { v2.b }[6], [x12]
+; CHECK-NEXT:    add x9, sp, #648
+; CHECK-NEXT:    ld1 { v3.b }[4], [x8]
+; CHECK-NEXT:    add x10, sp, #528
 ; CHECK-NEXT:    add x11, sp, #584
 ; CHECK-NEXT:    add x12, sp, #712
-; CHECK-NEXT:    ld1 { v4.b }[6], [x9]
-; CHECK-NEXT:    movi v7.2d, #0000000000000000
-; CHECK-NEXT:    ld1 { v16.b }[5], [x11]
-; CHECK-NEXT:    ld1 { v17.b }[5], [x12]
-; CHECK-NEXT:    ld1 { v5.b }[5], [x8]
-; CHECK-NEXT:    mov v0.b[5], w5
-; CHECK-NEXT:    add x9, sp, #536
-; CHECK-NEXT:    sshll v18.4s, v18.4h, #0
-; CHECK-NEXT:    add x8, sp, #656
+; CHECK-NEXT:    sshll v18.8h, v18.8b, #0
+; CHECK-NEXT:    mov v1.b[5], w5
+; CHECK-NEXT:    ld1 { v6.b }[5], [x11]
+; CHECK-NEXT:    ld1 { v7.b }[5], [x12]
+; CHECK-NEXT:    ld1 { v4.b }[5], [x9]
+; CHECK-NEXT:    ld1 { v5.b }[6], [x10]
+; CHECK-NEXT:    add x14, sp, #56
+; CHECK-NEXT:    movi v17.2d, #0000000000000000
+; CHECK-NEXT:    ld1 { v3.b }[5], [x14]
+; CHECK-NEXT:    add x9, sp, #656
+; CHECK-NEXT:    add x10, sp, #536
 ; CHECK-NEXT:    add x11, sp, #592
 ; CHECK-NEXT:    add x12, sp, #720
-; CHECK-NEXT:    ld1 { v4.b }[7], [x9]
-; CHECK-NEXT:    ld1 { v16.b }[6], [x11]
-; CHECK-NEXT:    ld1 { v17.b }[6], [x12]
-; CHECK-NEXT:    ld1 { v5.b }[6], [x8]
-; CHECK-NEXT:    ldr b6, [sp, #208]
-; CHECK-NEXT:    add x10, sp, #64
-; CHECK-NEXT:    mov v7.s[0], v18.s[0]
-; CHECK-NEXT:    mov v0.b[6], w6
-; CHECK-NEXT:    ld1 { v3.b }[6], [x10]
+; CHECK-NEXT:    sshll v18.4s, v18.4h, #0
+; CHECK-NEXT:    ldr b16, [sp, #208]
+; CHECK-NEXT:    ld1 { v6.b }[6], [x11]
+; CHECK-NEXT:    ld1 { v7.b }[6], [x12]
+; CHECK-NEXT:    ld1 { v4.b }[6], [x9]
+; CHECK-NEXT:    ld1 { v5.b }[7], [x10]
+; CHECK-NEXT:    add x8, sp, #64
+; CHECK-NEXT:    mov v1.b[6], w6
+; CHECK-NEXT:    sshll v16.8h, v16.8b, #0
+; CHECK-NEXT:    ld1 { v3.b }[6], [x8]
 ; CHECK-NEXT:    add x8, sp, #664
 ; CHECK-NEXT:    add x9, sp, #600
 ; CHECK-NEXT:    add x10, sp, #728
-; CHECK-NEXT:    sshll v4.8h, v4.8b, #0
-; CHECK-NEXT:    sshll v6.8h, v6.8b, #0
-; CHECK-NEXT:    ld1 { v16.b }[7], [x9]
-; CHECK-NEXT:    ld1 { v17.b }[7], [x10]
-; CHECK-NEXT:    ld1 { v5.b }[7], [x8]
+; CHECK-NEXT:    mov v17.s[0], v18.s[0]
+; CHECK-NEXT:    ld1 { v6.b }[7], [x9]
+; CHECK-NEXT:    ld1 { v7.b }[7], [x10]
+; CHECK-NEXT:    ld1 { v4.b }[7], [x8]
+; CHECK-NEXT:    sshll v5.8h, v5.8b, #0
 ; CHECK-NEXT:    movi v18.2d, #0000000000000000
-; CHECK-NEXT:    mov v0.b[7], w7
+; CHECK-NEXT:    sshll v16.4s, v16.4h, #0
+; CHECK-NEXT:    mov v1.b[7], w7
 ; CHECK-NEXT:    add x9, sp, #200
-; CHECK-NEXT:    add x10, sp, #72
-; CHECK-NEXT:    saddw v7.4s, v7.4s, v4.4h
-; CHECK-NEXT:    sshll v6.4s, v6.4h, #0
-; CHECK-NEXT:    sshll v16.8h, v16.8b, #0
-; CHECK-NEXT:    sshll v17.8h, v17.8b, #0
-; CHECK-NEXT:    sshll v5.8h, v5.8b, #0
+; CHECK-NEXT:    add x8, sp, #72
 ; CHECK-NEXT:    ld1 { v2.b }[7], [x9]
-; CHECK-NEXT:    ld1 { v3.b }[7], [x10]
-; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
-; CHECK-NEXT:    mov v18.s[0], v6.s[0]
+; CHECK-NEXT:    sshll v6.8h, v6.8b, #0
+; CHECK-NEXT:    ld1 { v3.b }[7], [x8]
+; CHECK-NEXT:    sshll v7.8h, v7.8b, #0
+; CHECK-NEXT:    sshll v4.8h, v4.8b, #0
+; CHECK-NEXT:    saddw v17.4s, v17.4s, v5.4h
+; CHECK-NEXT:    mov v18.s[0], v16.s[0]
 ; CHECK-NEXT:    sshll v0.8h, v0.8b, #0
-; CHECK-NEXT:    saddl2 v6.4s, v17.8h, v16.8h
-; CHECK-NEXT:    saddl2 v4.4s, v5.8h, v4.8h
-; CHECK-NEXT:    saddl v16.4s, v17.4h, v16.4h
-; CHECK-NEXT:    saddw v5.4s, v7.4s, v5.4h
+; CHECK-NEXT:    sshll v1.8h, v1.8b, #0
 ; CHECK-NEXT:    sshll v2.8h, v2.8b, #0
 ; CHECK-NEXT:    sshll v3.8h, v3.8b, #0
-; CHECK-NEXT:    saddl2 v17.4s, v0.8h, v1.8h
-; CHECK-NEXT:    saddw v0.4s, v18.4s, v0.4h
+; CHECK-NEXT:    saddl2 v16.4s, v7.8h, v6.8h
+; CHECK-NEXT:    saddl2 v5.4s, v4.8h, v5.8h
+; CHECK-NEXT:    saddl v6.4s, v7.4h, v6.4h
+; CHECK-NEXT:    saddw v4.4s, v17.4s, v4.4h
+; CHECK-NEXT:    saddl2 v17.4s, v1.8h, v0.8h
+; CHECK-NEXT:    saddw v1.4s, v18.4s, v1.4h
 ; CHECK-NEXT:    saddl2 v7.4s, v3.8h, v2.8h
-; CHECK-NEXT:    add v4.4s, v4.4s, v6.4s
-; CHECK-NEXT:    saddl v2.4s, v3.4h, v2.4h
 ; CHECK-NEXT:    add v5.4s, v5.4s, v16.4s
-; CHECK-NEXT:    saddw v0.4s, v0.4s, v1.4h
+; CHECK-NEXT:    saddl v2.4s, v3.4h, v2.4h
+; CHECK-NEXT:    add v4.4s, v4.4s, v6.4s
+; CHECK-NEXT:    saddw v0.4s, v1.4s, v0.4h
 ; CHECK-NEXT:    add v6.4s, v17.4s, v7.4s
-; CHECK-NEXT:    add v1.4s, v5.4s, v4.4s
+; CHECK-NEXT:    add v1.4s, v4.4s, v5.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    add v1.4s, v6.4s, v1.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
diff --git a/llvm/test/CodeGen/AArch64/neon-extadd.ll b/llvm/test/CodeGen/AArch64/neon-extadd.ll
index 913205f327536..16200435c5c31 100644
--- a/llvm/test/CodeGen/AArch64/neon-extadd.ll
+++ b/llvm/test/CodeGen/AArch64/neon-extadd.ll
@@ -515,15 +515,15 @@ define <20 x i32> @v20(<20 x i8> %s0, <20 x i8> %s1) {
 ; CHECK-NEXT:    mov v0.b[6], w6
 ; CHECK-NEXT:    ld1 { v1.b }[7], [x9]
 ; CHECK-NEXT:    uaddl v2.8h, v3.8b, v2.8b
+; CHECK-NEXT:    ushll v3.4s, v4.4h, #0
 ; CHECK-NEXT:    mov v0.b[7], w7
-; CHECK-NEXT:    ushll2 v3.4s, v2.8h, #0
-; CHECK-NEXT:    ushll v2.4s, v2.4h, #0
 ; CHECK-NEXT:    uaddl v0.8h, v0.8b, v1.8b
-; CHECK-NEXT:    ushll v1.4s, v4.4h, #0
-; CHECK-NEXT:    stp q3, q1, [x8, #48]
-; CHECK-NEXT:    ushll2 v1.4s, v0.8h, #0
+; CHECK-NEXT:    ushll2 v1.4s, v2.8h, #0
+; CHECK-NEXT:    ushll v2.4s, v2.4h, #0
+; CHECK-NEXT:    stp q1, q3, [x8, #48]
+; CHECK-NEXT:    ushll2 v3.4s, v0.8h, #0
 ; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
-; CHECK-NEXT:    stp q1, q2, [x8, #16]
+; CHECK-NEXT:    stp q3, q2, [x8, #16]
 ; CHECK-NEXT:    str q0, [x8]
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/neon-shift-neg.ll b/llvm/test/CodeGen/AArch64/neon-shift-neg.ll
index 881bbf315e8e9..45272143e8592 100644
--- a/llvm/test/CodeGen/AArch64/neon-shift-neg.ll
+++ b/llvm/test/CodeGen/AArch64/neon-shift-neg.ll
@@ -375,8 +375,8 @@ entry:
 define <vscale x 2 x i64> @shrn64x2(<vscale x 2 x i64> %a, i64 %b) {
 ; CHECK-LABEL: shrn64x2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    neg x8, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    asr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -391,8 +391,8 @@ entry:
 define <vscale x 4 x i32> @shrn32x4(<vscale x 4 x i32> %a, i32 %b) {
 ; CHECK-LABEL: shrn32x4:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    asr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -407,8 +407,8 @@ entry:
 define <vscale x 8 x i16> @shrn16x8(<vscale x 8 x i16> %a, i16 %b) {
 ; CHECK-LABEL: shrn16x8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    asr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -423,8 +423,8 @@ entry:
 define <vscale x 16 x i8> @shrn8x16(<vscale x 16 x i8> %a, i8 %b) {
 ; CHECK-LABEL: shrn8x16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z1.b, w8
 ; CHECK-NEXT:    asr z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
@@ -439,8 +439,8 @@ entry:
 define <vscale x 2 x i64> @lshrn64x2(<vscale x 2 x i64> %a, i64 %b) {
 ; CHECK-LABEL: lshrn64x2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    neg x8, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    lsr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -455,8 +455,8 @@ entry:
 define <vscale x 4 x i32> @lshrn32x4(<vscale x 4 x i32> %a, i32 %b) {
 ; CHECK-LABEL: lshrn32x4:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    lsr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -471,8 +471,8 @@ entry:
 define <vscale x 8 x i16> @lshrn16x8(<vscale x 8 x i16> %a, i16 %b) {
 ; CHECK-LABEL: lshrn16x8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    lsr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -487,8 +487,8 @@ entry:
 define <vscale x 16 x i8> @lshrn8x16(<vscale x 16 x i8> %a, i8 %b) {
 ; CHECK-LABEL: lshrn8x16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z1.b, w8
 ; CHECK-NEXT:    lsr z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
@@ -503,8 +503,8 @@ entry:
 define <vscale x 2 x i64> @shln64x2(<vscale x 2 x i64> %a, i64 %b) {
 ; CHECK-LABEL: shln64x2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    neg x8, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    lsl z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
@@ -519,8 +519,8 @@ entry:
 define <vscale x 4 x i32> @shln32x4(<vscale x 4 x i32> %a, i32 %b) {
 ; CHECK-LABEL: shln32x4:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    lsl z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -535,8 +535,8 @@ entry:
 define <vscale x 8 x i16> @shln16x8(<vscale x 8 x i16> %a, i16 %b) {
 ; CHECK-LABEL: shln16x8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    lsl z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -551,8 +551,8 @@ entry:
 define <vscale x 16 x i8> @shln8x16(<vscale x 16 x i8> %a, i8 %b) {
 ; CHECK-LABEL: shln8x16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    neg w8, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z1.b, w8
 ; CHECK-NEXT:    lsl z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/predicated-add-sub.ll b/llvm/test/CodeGen/AArch64/predicated-add-sub.ll
index 6b3cfc040cb3d..20088354bdb75 100644
--- a/llvm/test/CodeGen/AArch64/predicated-add-sub.ll
+++ b/llvm/test/CodeGen/AArch64/predicated-add-sub.ll
@@ -83,11 +83,11 @@ define <vscale x 2 x i64> @zext.add.2xi64(<vscale x 2 x i64> %a, <vscale x 2 x i
 define <vscale x 8 x i32> @zext.add.8xi32(<vscale x 8 x i32> %a, <vscale x 8 x i1> %v) #0 {
 ; CHECK-LABEL: zext.add.8xi32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    mov z2.s, #1 // =0x1
+; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    add z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    add z0.s, p0/m, z0.s, z2.s
+; CHECK-NEXT:    add z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    ret
   %extend = zext <vscale x 8 x i1> %v to <vscale x 8 x i32>
   %result = add <vscale x 8 x i32> %a, %extend
@@ -103,8 +103,8 @@ define <vscale x 16 x i32> @zext.add.16xi32(<vscale x 16 x i32> %a, <vscale x 16
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpklo p3.h, p0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z3.s, p2/m, z3.s, z4.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z2.s, p1/m, z2.s, z4.s
 ; CHECK-NEXT:    add z0.s, p3/m, z0.s, z4.s
 ; CHECK-NEXT:    add z1.s, p0/m, z1.s, z4.s
@@ -194,8 +194,8 @@ define <vscale x 2 x i64> @zext.sub.2xi64(<vscale x 2 x i64> %a, <vscale x 2 x i
 define <vscale x 8 x i32> @zext.sub.8xi32(<vscale x 8 x i32> %a, <vscale x 8 x i1> %v) #0 {
 ; CHECK-LABEL: zext.sub.8xi32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z0.s, p1/m, z0.s, z2.s
 ; CHECK-NEXT:    add z1.s, p0/m, z1.s, z2.s
@@ -214,8 +214,8 @@ define <vscale x 16 x i32> @zext.sub.16xi32(<vscale x 16 x i32> %a, <vscale x 16
 ; CHECK-NEXT:    punpklo p2.h, p1.b
 ; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    punpklo p3.h, p0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z0.s, p2/m, z0.s, z4.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z1.s, p1/m, z1.s, z4.s
 ; CHECK-NEXT:    add z2.s, p3/m, z2.s, z4.s
 ; CHECK-NEXT:    add z3.s, p0/m, z3.s, z4.s
@@ -305,11 +305,11 @@ define <vscale x 2 x i64> @sext.add.2xi64(<vscale x 2 x i64> %a, <vscale x 2 x i
 define <vscale x 8 x i32> @sext.add.8xi32(<vscale x 8 x i32> %a, <vscale x 8 x i1> %v) #0 {
 ; CHECK-LABEL: sext.add.8xi32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    add z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    add z0.s, p0/m, z0.s, z2.s
+; CHECK-NEXT:    add z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    ret
   %extend = sext <vscale x 8 x i1> %v to <vscale x 8 x i32>
   %result = add <vscale x 8 x i32> %a, %extend
@@ -325,8 +325,8 @@ define <vscale x 16 x i32> @sext.add.16xi32(<vscale x 16 x i32> %a, <vscale x 16
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpklo p3.h, p0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z3.s, p2/m, z3.s, z4.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    add z2.s, p1/m, z2.s, z4.s
 ; CHECK-NEXT:    add z0.s, p3/m, z0.s, z4.s
 ; CHECK-NEXT:    add z1.s, p0/m, z1.s, z4.s
@@ -416,11 +416,11 @@ define <vscale x 2 x i64> @sext.sub.2xi64(<vscale x 2 x i64> %a, <vscale x 2 x i
 define <vscale x 8 x i32> @sext.sub.8xi32(<vscale x 8 x i32> %a, <vscale x 8 x i1> %v) #0 {
 ; CHECK-LABEL: sext.sub.8xi32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    sub z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    sub z0.s, p0/m, z0.s, z2.s
+; CHECK-NEXT:    sub z1.s, p1/m, z1.s, z2.s
 ; CHECK-NEXT:    ret
   %extend = sext <vscale x 8 x i1> %v to <vscale x 8 x i32>
   %result = sub <vscale x 8 x i32> %a, %extend
@@ -436,8 +436,8 @@ define <vscale x 16 x i32> @sext.sub.16xi32(<vscale x 16 x i32> %a, <vscale x 16
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpklo p3.h, p0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    sub z3.s, p2/m, z3.s, z4.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    sub z2.s, p1/m, z2.s, z4.s
 ; CHECK-NEXT:    sub z0.s, p3/m, z0.s, z4.s
 ; CHECK-NEXT:    sub z1.s, p0/m, z1.s, z4.s
diff --git a/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll b/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
index 934ff44900c04..866b27b81d885 100644
--- a/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
+++ b/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
@@ -28,197 +28,193 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
 ; CHECK-NEXT:    .cfi_offset b13, -80
 ; CHECK-NEXT:    .cfi_offset b14, -88
 ; CHECK-NEXT:    .cfi_offset b15, -96
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
-; CHECK-NEXT:    // implicit-def: $q6
+; CHECK-NEXT:    movi v2.2d, #0000000000000000
+; CHECK-NEXT:    // implicit-def: $q1
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    mov x9, xzr
 ; CHECK-NEXT:    adrp x10, B+48
 ; CHECK-NEXT:    add x10, x10, :lo12:B+48
 ; CHECK-NEXT:    adrp x11, A
 ; CHECK-NEXT:    add x11, x11, :lo12:A
-; CHECK-NEXT:    // kill: killed $q6
-; CHECK-NEXT:    // implicit-def: $q6
+; CHECK-NEXT:    // kill: killed $q1
+; CHECK-NEXT:    // implicit-def: $q1
 ; CHECK-NEXT:    // implicit-def: $q0
-; CHECK-NEXT:    // implicit-def: $q2
 ; CHECK-NEXT:    // implicit-def: $q3
 ; CHECK-NEXT:    // implicit-def: $q4
 ; CHECK-NEXT:    // implicit-def: $q5
 ; CHECK-NEXT:    // implicit-def: $q7
-; CHECK-NEXT:    // implicit-def: $q16
-; CHECK-NEXT:    // implicit-def: $q17
 ; CHECK-NEXT:    // implicit-def: $q10
+; CHECK-NEXT:    // implicit-def: $q17
+; CHECK-NEXT:    // implicit-def: $q6
+; CHECK-NEXT:    // implicit-def: $q18
 ; CHECK-NEXT:    // implicit-def: $q19
 ; CHECK-NEXT:    // implicit-def: $q20
 ; CHECK-NEXT:    // implicit-def: $q21
 ; CHECK-NEXT:    // implicit-def: $q22
 ; CHECK-NEXT:    // implicit-def: $q23
 ; CHECK-NEXT:    // implicit-def: $q24
-; CHECK-NEXT:    // implicit-def: $q25
+; CHECK-NEXT:    // implicit-def: $q9
 ; CHECK-NEXT:    // implicit-def: $q27
-; CHECK-NEXT:    // implicit-def: $q26
+; CHECK-NEXT:    // implicit-def: $q12
 ; CHECK-NEXT:    // implicit-def: $q28
-; CHECK-NEXT:    // implicit-def: $q30
-; CHECK-NEXT:    // implicit-def: $q18
+; CHECK-NEXT:    // implicit-def: $q14
+; CHECK-NEXT:    // implicit-def: $q15
 ; CHECK-NEXT:    // implicit-def: $q29
+; CHECK-NEXT:    // implicit-def: $q30
+; CHECK-NEXT:    // implicit-def: $q11
 ; CHECK-NEXT:    // implicit-def: $q31
-; CHECK-NEXT:    // implicit-def: $q12
 ; CHECK-NEXT:    // implicit-def: $q13
-; CHECK-NEXT:    // implicit-def: $q11
-; CHECK-NEXT:    // kill: killed $q6
-; CHECK-NEXT:    // implicit-def: $q6
-; CHECK-NEXT:    // kill: killed $q6
+; CHECK-NEXT:    // kill: killed $q1
+; CHECK-NEXT:    // implicit-def: $q1
+; CHECK-NEXT:    // kill: killed $q1
 ; CHECK-NEXT:  .LBB0_1: // %for.cond1.preheader
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-NEXT:    str q14, [sp, #32] // 16-byte Folded Spill
 ; CHECK-NEXT:    ldr q14, [x8]
 ; CHECK-NEXT:    mov x12, xzr
-; CHECK-NEXT:    str q18, [sp, #32] // 16-byte Folded Spill
 ; CHECK-NEXT:    ldr x14, [x12]
-; CHECK-NEXT:    ldr q15, [x12]
-; CHECK-NEXT:    add x7, x11, x8
+; CHECK-NEXT:    stp q29, q15, [sp] // 32-byte Folded Spill
+; CHECK-NEXT:    add x19, x11, x8
 ; CHECK-NEXT:    fmov x15, d14
 ; CHECK-NEXT:    mov x16, v14.d[1]
-; CHECK-NEXT:    ldr q18, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    fmov x18, d15
-; CHECK-NEXT:    mov x13, v15.d[1]
-; CHECK-NEXT:    ldr x5, [x8]
+; CHECK-NEXT:    ldr q15, [x12]
 ; CHECK-NEXT:    ldr q14, [x10], #64
-; CHECK-NEXT:    ldr x7, [x7, #128]
+; CHECK-NEXT:    mov v8.16b, v28.16b
+; CHECK-NEXT:    fmov x13, d15
+; CHECK-NEXT:    mov x18, v15.d[1]
+; CHECK-NEXT:    mov v28.16b, v24.16b
 ; CHECK-NEXT:    mul x17, x15, x14
-; CHECK-NEXT:    mov v6.16b, v0.16b
-; CHECK-NEXT:    mov v9.16b, v27.16b
 ; CHECK-NEXT:    mov x12, v14.d[1]
 ; CHECK-NEXT:    fmov x4, d14
-; CHECK-NEXT:    mov v27.16b, v23.16b
+; CHECK-NEXT:    mov v24.16b, v20.16b
+; CHECK-NEXT:    mov v20.16b, v17.16b
+; CHECK-NEXT:    mov v17.16b, v5.16b
 ; CHECK-NEXT:    mul x1, x16, x14
-; CHECK-NEXT:    mov v23.16b, v19.16b
-; CHECK-NEXT:    mov v19.16b, v7.16b
-; CHECK-NEXT:    mov v7.16b, v2.16b
-; CHECK-NEXT:    stp q26, q31, [sp] // 32-byte Folded Spill
-; CHECK-NEXT:    mov v31.16b, v22.16b
-; CHECK-NEXT:    mul x0, x18, x14
-; CHECK-NEXT:    mov v26.16b, v10.16b
-; CHECK-NEXT:    mov v22.16b, v5.16b
+; CHECK-NEXT:    ldr q5, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    ldr x5, [x8]
+; CHECK-NEXT:    ldr x19, [x19, #128]
+; CHECK-NEXT:    mov v29.16b, v21.16b
+; CHECK-NEXT:    mov v21.16b, v0.16b
+; CHECK-NEXT:    mul x0, x13, x14
+; CHECK-NEXT:    mov v25.16b, v6.16b
+; CHECK-NEXT:    mov v6.16b, v2.16b
 ; CHECK-NEXT:    fmov d15, x17
-; CHECK-NEXT:    mov v5.16b, v1.16b
-; CHECK-NEXT:    mov v8.16b, v20.16b
-; CHECK-NEXT:    mul x2, x13, x14
-; CHECK-NEXT:    mov v20.16b, v16.16b
-; CHECK-NEXT:    mov v16.16b, v3.16b
-; CHECK-NEXT:    mov v10.16b, v21.16b
-; CHECK-NEXT:    mov v21.16b, v17.16b
-; CHECK-NEXT:    mov v17.16b, v4.16b
+; CHECK-NEXT:    mov v26.16b, v22.16b
+; CHECK-NEXT:    mov v22.16b, v18.16b
+; CHECK-NEXT:    mul x2, x18, x14
+; CHECK-NEXT:    mov v18.16b, v7.16b
+; CHECK-NEXT:    mov v7.16b, v3.16b
+; CHECK-NEXT:    mov v16.16b, v4.16b
+; CHECK-NEXT:    add x8, x8, #8
+; CHECK-NEXT:    add x9, x9, #1
 ; CHECK-NEXT:    mov v15.d[1], x1
 ; CHECK-NEXT:    mul x3, x12, x14
-; CHECK-NEXT:    add x8, x8, #8
-; CHECK-NEXT:    fmov d14, x0
 ; CHECK-NEXT:    cmp x8, #64
-; CHECK-NEXT:    add x9, x9, #1
+; CHECK-NEXT:    fmov d14, x0
 ; CHECK-NEXT:    mul x14, x4, x14
-; CHECK-NEXT:    add v18.2d, v18.2d, v15.2d
-; CHECK-NEXT:    mul x19, x15, x5
+; CHECK-NEXT:    add v5.2d, v5.2d, v15.2d
+; CHECK-NEXT:    mul x20, x15, x5
 ; CHECK-NEXT:    mov v14.d[1], x2
-; CHECK-NEXT:    mul x15, x15, x7
+; CHECK-NEXT:    mul x15, x15, x19
 ; CHECK-NEXT:    fmov d0, x14
-; CHECK-NEXT:    str q18, [sp, #64] // 16-byte Folded Spill
-; CHECK-NEXT:    ldp q18, q15, [sp, #32] // 32-byte Folded Reload
-; CHECK-NEXT:    mul x6, x16, x5
-; CHECK-NEXT:    fmov d1, x19
+; CHECK-NEXT:    str q5, [sp, #64] // 16-byte Folded Spill
+; CHECK-NEXT:    ldr q5, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT:    mul x21, x13, x19
+; CHECK-NEXT:    add v5.2d, v5.2d, v14.2d
+; CHECK-NEXT:    fmov d3, x20
+; CHECK-NEXT:    mul x7, x16, x5
 ; CHECK-NEXT:    mov v0.d[1], x3
-; CHECK-NEXT:    mul x16, x16, x7
-; CHECK-NEXT:    fmov d2, x15
-; CHECK-NEXT:    add v15.2d, v15.2d, v14.2d
-; CHECK-NEXT:    mul x21, x18, x7
-; CHECK-NEXT:    mov v1.d[1], x6
-; CHECK-NEXT:    mul x0, x4, x7
-; CHECK-NEXT:    str q15, [sp, #48] // 16-byte Folded Spill
-; CHECK-NEXT:    add v15.2d, v11.2d, v14.2d
-; CHECK-NEXT:    mov v2.d[1], x16
-; CHECK-NEXT:    ldr q11, [sp, #80] // 16-byte Folded Reload
-; CHECK-NEXT:    mul x20, x13, x7
-; CHECK-NEXT:    fmov d3, x21
-; CHECK-NEXT:    add v11.2d, v11.2d, v0.2d
-; CHECK-NEXT:    add v12.2d, v12.2d, v1.2d
-; CHECK-NEXT:    mul x22, x12, x7
-; CHECK-NEXT:    fmov d4, x0
-; CHECK-NEXT:    add v18.2d, v18.2d, v2.2d
-; CHECK-NEXT:    mov v2.16b, v7.16b
-; CHECK-NEXT:    mul x14, x18, x5
-; CHECK-NEXT:    mov v7.16b, v19.16b
-; CHECK-NEXT:    mov v19.16b, v23.16b
-; CHECK-NEXT:    mov v3.d[1], x20
-; CHECK-NEXT:    mov v23.16b, v27.16b
-; CHECK-NEXT:    add v27.2d, v9.2d, v1.2d
-; CHECK-NEXT:    mul x15, x4, x5
-; CHECK-NEXT:    str q11, [sp, #80] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v11.16b, v15.16b
-; CHECK-NEXT:    mov v4.d[1], x22
-; CHECK-NEXT:    add v19.2d, v19.2d, v1.2d
-; CHECK-NEXT:    add v7.2d, v7.2d, v1.2d
+; CHECK-NEXT:    fmov d1, x15
+; CHECK-NEXT:    mul x16, x16, x19
+; CHECK-NEXT:    str q5, [sp, #48] // 16-byte Folded Spill
+; CHECK-NEXT:    add v5.2d, v13.2d, v14.2d
+; CHECK-NEXT:    fmov d2, x21
+; CHECK-NEXT:    ldr q13, [sp, #80] // 16-byte Folded Reload
+; CHECK-NEXT:    mul x6, x18, x5
+; CHECK-NEXT:    ldp q15, q14, [sp, #16] // 32-byte Folded Reload
+; CHECK-NEXT:    mov v3.d[1], x7
+; CHECK-NEXT:    add v13.2d, v13.2d, v0.2d
+; CHECK-NEXT:    mul x18, x18, x19
+; CHECK-NEXT:    mov v1.d[1], x16
+; CHECK-NEXT:    mul x22, x4, x19
+; CHECK-NEXT:    str q13, [sp, #80] // 16-byte Folded Spill
+; CHECK-NEXT:    mov v13.16b, v5.16b
+; CHECK-NEXT:    mov v5.16b, v17.16b
+; CHECK-NEXT:    mov v17.16b, v20.16b
+; CHECK-NEXT:    mov v20.16b, v24.16b
 ; CHECK-NEXT:    mul x13, x13, x5
-; CHECK-NEXT:    add v23.2d, v23.2d, v1.2d
-; CHECK-NEXT:    add v1.2d, v5.2d, v1.2d
-; CHECK-NEXT:    fmov d14, x14
-; CHECK-NEXT:    add v30.2d, v30.2d, v3.2d
-; CHECK-NEXT:    mov v3.16b, v16.16b
+; CHECK-NEXT:    mov v24.16b, v28.16b
+; CHECK-NEXT:    add v11.2d, v11.2d, v3.2d
+; CHECK-NEXT:    mov v2.d[1], x18
+; CHECK-NEXT:    add v15.2d, v15.2d, v1.2d
+; CHECK-NEXT:    add v27.2d, v27.2d, v3.2d
+; CHECK-NEXT:    mul x17, x12, x19
+; CHECK-NEXT:    add v23.2d, v23.2d, v3.2d
+; CHECK-NEXT:    add v19.2d, v19.2d, v3.2d
+; CHECK-NEXT:    fmov d4, x22
+; CHECK-NEXT:    add v10.2d, v10.2d, v3.2d
+; CHECK-NEXT:    mul x14, x4, x5
+; CHECK-NEXT:    fmov d0, x13
+; CHECK-NEXT:    add v14.2d, v14.2d, v2.2d
+; CHECK-NEXT:    add v2.2d, v6.2d, v3.2d
 ; CHECK-NEXT:    mul x12, x12, x5
-; CHECK-NEXT:    mov v16.16b, v20.16b
-; CHECK-NEXT:    mov v5.16b, v22.16b
-; CHECK-NEXT:    fmov d0, x15
-; CHECK-NEXT:    add v28.2d, v28.2d, v4.2d
-; CHECK-NEXT:    mov v4.16b, v17.16b
-; CHECK-NEXT:    mov v17.16b, v21.16b
-; CHECK-NEXT:    mov v21.16b, v10.16b
-; CHECK-NEXT:    mov v10.16b, v26.16b
-; CHECK-NEXT:    mov v14.d[1], x13
-; CHECK-NEXT:    mov v22.16b, v31.16b
-; CHECK-NEXT:    ldp q26, q31, [sp] // 32-byte Folded Reload
-; CHECK-NEXT:    mov v0.d[1], x12
-; CHECK-NEXT:    add v13.2d, v13.2d, v14.2d
-; CHECK-NEXT:    add v31.2d, v31.2d, v14.2d
-; CHECK-NEXT:    add v26.2d, v26.2d, v14.2d
-; CHECK-NEXT:    add v24.2d, v24.2d, v14.2d
-; CHECK-NEXT:    add v22.2d, v22.2d, v14.2d
-; CHECK-NEXT:    add v20.2d, v8.2d, v14.2d
-; CHECK-NEXT:    add v10.2d, v10.2d, v14.2d
-; CHECK-NEXT:    add v16.2d, v16.2d, v14.2d
-; CHECK-NEXT:    add v5.2d, v5.2d, v14.2d
-; CHECK-NEXT:    add v3.2d, v3.2d, v14.2d
-; CHECK-NEXT:    add v2.2d, v2.2d, v14.2d
-; CHECK-NEXT:    add v29.2d, v29.2d, v0.2d
-; CHECK-NEXT:    add v25.2d, v25.2d, v0.2d
-; CHECK-NEXT:    add v21.2d, v21.2d, v0.2d
+; CHECK-NEXT:    mov v3.16b, v7.16b
+; CHECK-NEXT:    mov v7.16b, v18.16b
+; CHECK-NEXT:    mov v4.d[1], x17
+; CHECK-NEXT:    mov v18.16b, v22.16b
+; CHECK-NEXT:    mov v0.d[1], x6
+; CHECK-NEXT:    fmov d1, x14
+; CHECK-NEXT:    add v28.2d, v8.2d, v4.2d
+; CHECK-NEXT:    mov v1.d[1], x12
+; CHECK-NEXT:    add v31.2d, v31.2d, v0.2d
+; CHECK-NEXT:    add v30.2d, v30.2d, v0.2d
+; CHECK-NEXT:    add v12.2d, v12.2d, v0.2d
+; CHECK-NEXT:    add v24.2d, v24.2d, v0.2d
+; CHECK-NEXT:    add v22.2d, v26.2d, v0.2d
+; CHECK-NEXT:    add v20.2d, v20.2d, v0.2d
+; CHECK-NEXT:    add v18.2d, v18.2d, v0.2d
 ; CHECK-NEXT:    add v17.2d, v17.2d, v0.2d
-; CHECK-NEXT:    add v4.2d, v4.2d, v0.2d
-; CHECK-NEXT:    add v0.2d, v6.2d, v0.2d
+; CHECK-NEXT:    add v7.2d, v7.2d, v0.2d
+; CHECK-NEXT:    add v4.2d, v16.2d, v0.2d
+; CHECK-NEXT:    add v3.2d, v3.2d, v0.2d
+; CHECK-NEXT:    mov v0.16b, v21.16b
+; CHECK-NEXT:    mov v21.16b, v29.16b
+; CHECK-NEXT:    ldr q29, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    add v9.2d, v9.2d, v1.2d
+; CHECK-NEXT:    add v6.2d, v25.2d, v1.2d
+; CHECK-NEXT:    add v5.2d, v5.2d, v1.2d
+; CHECK-NEXT:    add v29.2d, v29.2d, v1.2d
+; CHECK-NEXT:    add v21.2d, v21.2d, v1.2d
+; CHECK-NEXT:    add v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    b.ne .LBB0_1
 ; CHECK-NEXT:  // %bb.2: // %for.cond.cleanup
-; CHECK-NEXT:    ldr q6, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT:    ldr q1, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    adrp x8, C
 ; CHECK-NEXT:    add x8, x8, :lo12:C
-; CHECK-NEXT:    stp q12, q31, [x8, #80]
+; CHECK-NEXT:    stp q11, q30, [x8, #80]
 ; CHECK-NEXT:    ldp x20, x19, [sp, #176] // 16-byte Folded Reload
-; CHECK-NEXT:    str q6, [x8]
-; CHECK-NEXT:    ldr q6, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    str q29, [x8, #112]
+; CHECK-NEXT:    str q1, [x8]
+; CHECK-NEXT:    ldr q1, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    stp q15, q14, [x8, #144]
 ; CHECK-NEXT:    ldp x22, x21, [sp, #160] // 16-byte Folded Reload
-; CHECK-NEXT:    stp q6, q11, [x8, #16]
-; CHECK-NEXT:    ldr q6, [sp, #80] // 16-byte Folded Reload
-; CHECK-NEXT:    stp q18, q30, [x8, #144]
-; CHECK-NEXT:    ldp d9, d8, [sp, #144] // 16-byte Folded Reload
-; CHECK-NEXT:    stp q6, q13, [x8, #48]
+; CHECK-NEXT:    stp q1, q13, [x8, #16]
+; CHECK-NEXT:    ldr q1, [sp, #80] // 16-byte Folded Reload
+; CHECK-NEXT:    stp q28, q12, [x8, #176]
 ; CHECK-NEXT:    ldp d13, d12, [sp, #112] // 16-byte Folded Reload
-; CHECK-NEXT:    stp q28, q26, [x8, #176]
+; CHECK-NEXT:    stp q1, q31, [x8, #48]
 ; CHECK-NEXT:    ldp d15, d14, [sp, #96] // 16-byte Folded Reload
-; CHECK-NEXT:    stp q19, q10, [x8, #336]
+; CHECK-NEXT:    stp q9, q24, [x8, #240]
+; CHECK-NEXT:    ldp d9, d8, [sp, #144] // 16-byte Folded Reload
+; CHECK-NEXT:    stp q19, q18, [x8, #336]
+; CHECK-NEXT:    stp q10, q7, [x8, #400]
 ; CHECK-NEXT:    ldp d11, d10, [sp, #128] // 16-byte Folded Reload
+; CHECK-NEXT:    str q29, [x8, #112]
 ; CHECK-NEXT:    str q27, [x8, #208]
-; CHECK-NEXT:    stp q25, q24, [x8, #240]
 ; CHECK-NEXT:    stp q23, q22, [x8, #272]
 ; CHECK-NEXT:    stp q21, q20, [x8, #304]
-; CHECK-NEXT:    stp q17, q16, [x8, #368]
-; CHECK-NEXT:    stp q7, q5, [x8, #400]
-; CHECK-NEXT:    stp q4, q3, [x8, #432]
-; CHECK-NEXT:    stp q1, q2, [x8, #464]
+; CHECK-NEXT:    stp q6, q17, [x8, #368]
+; CHECK-NEXT:    stp q5, q4, [x8, #432]
+; CHECK-NEXT:    stp q2, q3, [x8, #464]
 ; CHECK-NEXT:    str q0, [x8, #496]
 ; CHECK-NEXT:    add sp, sp, #192
 ; CHECK-NEXT:    .cfi_def_cfa_offset 0
diff --git a/llvm/test/CodeGen/AArch64/rcpc3-sve.ll b/llvm/test/CodeGen/AArch64/rcpc3-sve.ll
index 47e3381517499..6b03e5d12bfd3 100644
--- a/llvm/test/CodeGen/AArch64/rcpc3-sve.ll
+++ b/llvm/test/CodeGen/AArch64/rcpc3-sve.ll
@@ -8,8 +8,8 @@
 define hidden <vscale x 2 x i64> @test_load_sve_lane0(ptr nocapture noundef readonly %a, <vscale x 2 x i64> noundef %b) local_unnamed_addr {
 ; CHECK-LABEL: test_load_sve_lane0:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    ldapr x8, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, p0/m, x8
 ; CHECK-NEXT:    ret
   %1 = load atomic i64, ptr %a acquire, align 8
@@ -20,9 +20,9 @@ define hidden <vscale x 2 x i64> @test_load_sve_lane0(ptr nocapture noundef read
 define hidden <vscale x 2 x i64> @test_load_sve_lane1(ptr nocapture noundef readonly %a, <vscale x 2 x i64> noundef %b) local_unnamed_addr {
 ; CHECK-LABEL: test_load_sve_lane1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    ldapr x8, [x0]
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
diff --git a/llvm/test/CodeGen/AArch64/reassocmls.ll b/llvm/test/CodeGen/AArch64/reassocmls.ll
index 381caffba92eb..acbf9fc584a2e 100644
--- a/llvm/test/CodeGen/AArch64/reassocmls.ll
+++ b/llvm/test/CodeGen/AArch64/reassocmls.ll
@@ -79,7 +79,7 @@ define i64 @mls_i64_C(i64 %a, i64 %b, i64 %c, i64 %d, i64 %e) {
 ; CHECK-LABEL: mls_i64_C:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mul x8, x2, x1
-; CHECK-NEXT:    mov w9, #10
+; CHECK-NEXT:    mov w9, #10 // =0xa
 ; CHECK-NEXT:    madd x8, x4, x3, x8
 ; CHECK-NEXT:    sub x0, x9, x8
 ; CHECK-NEXT:    ret
@@ -290,9 +290,9 @@ define <vscale x 8 x i16> @smlsl_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i8
 define <vscale x 8 x i16> @umlsl_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i8> %b, <vscale x 8 x i8> %c, <vscale x 8 x i8> %d, <vscale x 8 x i8> %e) {
 ; CHECK-LABEL: umlsl_nxv8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z3.h, z3.h, #0xff
 ; CHECK-NEXT:    and z4.h, z4.h, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    and z2.h, z2.h, #0xff
 ; CHECK-NEXT:    mls z0.h, p0/m, z4.h, z3.h
@@ -326,8 +326,8 @@ define <vscale x 8 x i16> @mls_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16>
 define <vscale x 8 x i16> @mla_nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c, <vscale x 8 x i16> %d, <vscale x 8 x i16> %e) {
 ; CHECK-LABEL: mla_nxv8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mul z1.h, z2.h, z1.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mla z1.h, p0/m, z4.h, z3.h
 ; CHECK-NEXT:    add z0.h, z1.h, z0.h
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/reduce-shuffle.ll b/llvm/test/CodeGen/AArch64/reduce-shuffle.ll
index a080a7403811f..325ab444205bf 100644
--- a/llvm/test/CodeGen/AArch64/reduce-shuffle.ll
+++ b/llvm/test/CodeGen/AArch64/reduce-shuffle.ll
@@ -20,111 +20,111 @@ define i32 @v1(ptr nocapture noundef readonly %p1, i32 noundef %i1, ptr nocaptur
 ; CHECK-NEXT:    ldr d6, [x10, x8]
 ; CHECK-NEXT:    ldr d5, [x11]
 ; CHECK-NEXT:    ldr d7, [x11, x9]
-; CHECK-NEXT:    usubl v2.8h, v2.8b, v3.8b
 ; CHECK-NEXT:    usubl v0.8h, v0.8b, v1.8b
-; CHECK-NEXT:    usubl v1.8h, v4.8b, v5.8b
+; CHECK-NEXT:    usubl v1.8h, v2.8b, v3.8b
+; CHECK-NEXT:    usubl v2.8h, v4.8b, v5.8b
 ; CHECK-NEXT:    usubl v3.8h, v6.8b, v7.8b
-; CHECK-NEXT:    shll2 v4.4s, v2.8h, #16
-; CHECK-NEXT:    shll2 v5.4s, v0.8h, #16
+; CHECK-NEXT:    shll2 v4.4s, v0.8h, #16
+; CHECK-NEXT:    shll2 v5.4s, v1.8h, #16
 ; CHECK-NEXT:    shll2 v6.4s, v3.8h, #16
-; CHECK-NEXT:    shll2 v7.4s, v1.8h, #16
-; CHECK-NEXT:    saddw v2.4s, v4.4s, v2.4h
-; CHECK-NEXT:    saddw v0.4s, v5.4s, v0.4h
+; CHECK-NEXT:    shll2 v7.4s, v2.8h, #16
+; CHECK-NEXT:    saddw v0.4s, v4.4s, v0.4h
+; CHECK-NEXT:    saddw v1.4s, v5.4s, v1.4h
 ; CHECK-NEXT:    saddw v3.4s, v6.4s, v3.4h
-; CHECK-NEXT:    saddw v1.4s, v7.4s, v1.4h
+; CHECK-NEXT:    saddw v2.4s, v7.4s, v2.4h
+; CHECK-NEXT:    zip1 v4.4s, v1.4s, v0.4s
+; CHECK-NEXT:    zip2 v6.4s, v1.4s, v0.4s
+; CHECK-NEXT:    uzp2 v5.4s, v3.4s, v2.4s
 ; CHECK-NEXT:    mov v7.16b, v2.16b
-; CHECK-NEXT:    zip1 v4.4s, v2.4s, v0.4s
-; CHECK-NEXT:    zip2 v6.4s, v2.4s, v0.4s
-; CHECK-NEXT:    uzp2 v5.4s, v3.4s, v1.4s
-; CHECK-NEXT:    mov v17.16b, v1.16b
-; CHECK-NEXT:    zip2 v16.4s, v1.4s, v3.4s
-; CHECK-NEXT:    mov v7.s[3], v0.s[2]
-; CHECK-NEXT:    ext v18.16b, v3.16b, v3.16b, #12
-; CHECK-NEXT:    ext v2.16b, v2.16b, v4.16b, #8
-; CHECK-NEXT:    mov v17.s[1], v3.s[0]
+; CHECK-NEXT:    ext v17.16b, v3.16b, v3.16b, #12
+; CHECK-NEXT:    zip2 v18.4s, v3.4s, v2.4s
+; CHECK-NEXT:    ext v16.16b, v1.16b, v4.16b, #8
+; CHECK-NEXT:    mov v1.s[3], v0.s[2]
+; CHECK-NEXT:    mov v7.s[1], v3.s[0]
 ; CHECK-NEXT:    uzp2 v0.4s, v5.4s, v3.4s
-; CHECK-NEXT:    zip2 v5.4s, v3.4s, v1.4s
-; CHECK-NEXT:    mov v3.s[0], v1.s[1]
-; CHECK-NEXT:    ext v1.16b, v1.16b, v18.16b, #12
-; CHECK-NEXT:    mov v16.d[1], v7.d[1]
-; CHECK-NEXT:    mov v17.d[1], v2.d[1]
+; CHECK-NEXT:    zip2 v5.4s, v2.4s, v3.4s
+; CHECK-NEXT:    mov v3.s[0], v2.s[1]
+; CHECK-NEXT:    ext v2.16b, v2.16b, v17.16b, #12
+; CHECK-NEXT:    mov v18.d[1], v1.d[1]
+; CHECK-NEXT:    mov v7.d[1], v16.d[1]
 ; CHECK-NEXT:    mov v0.d[1], v6.d[1]
-; CHECK-NEXT:    mov v5.d[1], v7.d[1]
 ; CHECK-NEXT:    mov v3.d[1], v4.d[1]
-; CHECK-NEXT:    mov v1.d[1], v6.d[1]
-; CHECK-NEXT:    add v0.4s, v0.4s, v5.4s
-; CHECK-NEXT:    add v2.4s, v3.4s, v17.4s
-; CHECK-NEXT:    sub v3.4s, v17.4s, v3.4s
-; CHECK-NEXT:    sub v1.4s, v16.4s, v1.4s
+; CHECK-NEXT:    mov v5.d[1], v1.d[1]
+; CHECK-NEXT:    mov v2.d[1], v6.d[1]
+; CHECK-NEXT:    add v0.4s, v0.4s, v18.4s
+; CHECK-NEXT:    add v1.4s, v3.4s, v7.4s
+; CHECK-NEXT:    sub v3.4s, v7.4s, v3.4s
+; CHECK-NEXT:    sub v2.4s, v5.4s, v2.4s
 ; CHECK-NEXT:    rev64 v4.4s, v0.4s
-; CHECK-NEXT:    rev64 v5.4s, v2.4s
-; CHECK-NEXT:    add v6.4s, v1.4s, v3.4s
-; CHECK-NEXT:    sub v1.4s, v3.4s, v1.4s
+; CHECK-NEXT:    rev64 v6.4s, v1.4s
+; CHECK-NEXT:    sub v5.4s, v3.4s, v2.4s
+; CHECK-NEXT:    add v2.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    mov v4.d[1], v0.d[1]
-; CHECK-NEXT:    mov v5.d[1], v2.d[1]
-; CHECK-NEXT:    rev64 v3.4s, v1.4s
-; CHECK-NEXT:    sub v2.4s, v2.4s, v4.4s
-; CHECK-NEXT:    add v0.4s, v0.4s, v5.4s
-; CHECK-NEXT:    rev64 v4.4s, v6.4s
-; CHECK-NEXT:    rev64 v5.4s, v2.4s
-; CHECK-NEXT:    rev64 v7.4s, v0.4s
-; CHECK-NEXT:    addp v16.4s, v0.4s, v6.4s
-; CHECK-NEXT:    addp v17.4s, v2.4s, v1.4s
-; CHECK-NEXT:    sub v4.4s, v6.4s, v4.4s
-; CHECK-NEXT:    sub v1.4s, v1.4s, v3.4s
-; CHECK-NEXT:    sub v2.4s, v2.4s, v5.4s
-; CHECK-NEXT:    sub v0.4s, v0.4s, v7.4s
-; CHECK-NEXT:    zip1 v21.4s, v16.4s, v16.4s
-; CHECK-NEXT:    ext v5.16b, v17.16b, v1.16b, #4
-; CHECK-NEXT:    ext v6.16b, v16.16b, v4.16b, #4
-; CHECK-NEXT:    mov v18.16b, v1.16b
-; CHECK-NEXT:    mov v19.16b, v4.16b
-; CHECK-NEXT:    ext v3.16b, v2.16b, v17.16b, #8
-; CHECK-NEXT:    ext v7.16b, v0.16b, v16.16b, #4
-; CHECK-NEXT:    mov v18.s[2], v17.s[3]
-; CHECK-NEXT:    zip2 v5.4s, v5.4s, v17.4s
-; CHECK-NEXT:    zip2 v6.4s, v6.4s, v16.4s
-; CHECK-NEXT:    mov v19.s[2], v16.s[3]
-; CHECK-NEXT:    trn2 v0.4s, v21.4s, v0.4s
-; CHECK-NEXT:    ext v20.16b, v3.16b, v2.16b, #4
-; CHECK-NEXT:    ext v7.16b, v7.16b, v7.16b, #4
-; CHECK-NEXT:    mov v2.s[2], v17.s[1]
-; CHECK-NEXT:    ext v1.16b, v1.16b, v5.16b, #12
-; CHECK-NEXT:    ext v4.16b, v4.16b, v6.16b, #12
-; CHECK-NEXT:    mov v5.16b, v18.16b
-; CHECK-NEXT:    uzp2 v3.4s, v3.4s, v20.4s
-; CHECK-NEXT:    mov v6.16b, v7.16b
-; CHECK-NEXT:    mov v20.16b, v19.16b
-; CHECK-NEXT:    mov v21.16b, v2.16b
-; CHECK-NEXT:    mov v5.s[1], v17.s[2]
-; CHECK-NEXT:    sub v7.4s, v0.4s, v7.4s
-; CHECK-NEXT:    mov v6.s[0], v16.s[1]
-; CHECK-NEXT:    mov v20.s[1], v16.s[2]
-; CHECK-NEXT:    sub v16.4s, v19.4s, v4.4s
-; CHECK-NEXT:    mov v21.s[1], v17.s[0]
-; CHECK-NEXT:    sub v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    sub v17.4s, v18.4s, v1.4s
-; CHECK-NEXT:    add v1.4s, v5.4s, v1.4s
+; CHECK-NEXT:    mov v6.d[1], v1.d[1]
+; CHECK-NEXT:    rev64 v3.4s, v5.4s
+; CHECK-NEXT:    rev64 v7.4s, v2.4s
+; CHECK-NEXT:    sub v1.4s, v1.4s, v4.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v6.4s
-; CHECK-NEXT:    add v4.4s, v20.4s, v4.4s
-; CHECK-NEXT:    add v3.4s, v21.4s, v3.4s
-; CHECK-NEXT:    mov v1.d[1], v17.d[1]
-; CHECK-NEXT:    mov v0.d[1], v7.d[1]
-; CHECK-NEXT:    mov v4.d[1], v16.d[1]
-; CHECK-NEXT:    mov v3.d[1], v2.d[1]
+; CHECK-NEXT:    sub v3.4s, v5.4s, v3.4s
+; CHECK-NEXT:    addp v4.4s, v1.4s, v5.4s
+; CHECK-NEXT:    sub v5.4s, v2.4s, v7.4s
+; CHECK-NEXT:    addp v2.4s, v0.4s, v2.4s
+; CHECK-NEXT:    rev64 v6.4s, v0.4s
+; CHECK-NEXT:    rev64 v7.4s, v1.4s
+; CHECK-NEXT:    ext v16.16b, v4.16b, v3.16b, #4
+; CHECK-NEXT:    ext v17.16b, v2.16b, v5.16b, #4
+; CHECK-NEXT:    sub v0.4s, v0.4s, v6.4s
+; CHECK-NEXT:    sub v1.4s, v1.4s, v7.4s
+; CHECK-NEXT:    mov v7.16b, v3.16b
+; CHECK-NEXT:    zip2 v6.4s, v16.4s, v4.4s
+; CHECK-NEXT:    mov v16.16b, v5.16b
+; CHECK-NEXT:    zip2 v17.4s, v17.4s, v2.4s
+; CHECK-NEXT:    ext v18.16b, v0.16b, v2.16b, #4
+; CHECK-NEXT:    mov v7.s[2], v4.s[3]
+; CHECK-NEXT:    mov v21.16b, v1.16b
+; CHECK-NEXT:    mov v16.s[2], v2.s[3]
+; CHECK-NEXT:    ext v5.16b, v5.16b, v17.16b, #12
+; CHECK-NEXT:    zip1 v17.4s, v2.4s, v2.4s
+; CHECK-NEXT:    ext v3.16b, v3.16b, v6.16b, #12
+; CHECK-NEXT:    ext v18.16b, v18.16b, v18.16b, #4
+; CHECK-NEXT:    mov v19.16b, v7.16b
+; CHECK-NEXT:    ext v6.16b, v1.16b, v4.16b, #8
+; CHECK-NEXT:    mov v21.s[2], v4.s[1]
+; CHECK-NEXT:    mov v20.16b, v16.16b
+; CHECK-NEXT:    mov v19.s[1], v4.s[2]
+; CHECK-NEXT:    trn2 v0.4s, v17.4s, v0.4s
+; CHECK-NEXT:    sub v16.4s, v16.4s, v5.4s
+; CHECK-NEXT:    mov v17.16b, v18.16b
+; CHECK-NEXT:    ext v1.16b, v6.16b, v1.16b, #4
+; CHECK-NEXT:    sub v7.4s, v7.4s, v3.4s
+; CHECK-NEXT:    mov v20.s[1], v2.s[2]
+; CHECK-NEXT:    mov v17.s[0], v2.s[1]
+; CHECK-NEXT:    mov v2.16b, v21.16b
+; CHECK-NEXT:    add v3.4s, v19.4s, v3.4s
+; CHECK-NEXT:    uzp2 v1.4s, v6.4s, v1.4s
+; CHECK-NEXT:    add v5.4s, v20.4s, v5.4s
+; CHECK-NEXT:    mov v2.s[1], v4.s[0]
+; CHECK-NEXT:    sub v4.4s, v0.4s, v18.4s
+; CHECK-NEXT:    mov v3.d[1], v7.d[1]
+; CHECK-NEXT:    add v0.4s, v0.4s, v17.4s
+; CHECK-NEXT:    mov v5.d[1], v16.d[1]
+; CHECK-NEXT:    sub v6.4s, v21.4s, v1.4s
+; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
+; CHECK-NEXT:    mov v0.d[1], v4.d[1]
+; CHECK-NEXT:    cmlt v4.8h, v3.8h, #0
+; CHECK-NEXT:    cmlt v2.8h, v5.8h, #0
+; CHECK-NEXT:    mov v1.d[1], v6.d[1]
+; CHECK-NEXT:    add v3.4s, v4.4s, v3.4s
+; CHECK-NEXT:    cmlt v6.8h, v0.8h, #0
+; CHECK-NEXT:    add v5.4s, v2.4s, v5.4s
+; CHECK-NEXT:    eor v3.16b, v3.16b, v4.16b
 ; CHECK-NEXT:    cmlt v7.8h, v1.8h, #0
-; CHECK-NEXT:    cmlt v2.8h, v0.8h, #0
-; CHECK-NEXT:    cmlt v6.8h, v4.8h, #0
-; CHECK-NEXT:    cmlt v5.8h, v3.8h, #0
+; CHECK-NEXT:    add v0.4s, v6.4s, v0.4s
+; CHECK-NEXT:    eor v2.16b, v5.16b, v2.16b
 ; CHECK-NEXT:    add v1.4s, v7.4s, v1.4s
-; CHECK-NEXT:    add v0.4s, v2.4s, v0.4s
-; CHECK-NEXT:    add v4.4s, v6.4s, v4.4s
-; CHECK-NEXT:    add v3.4s, v5.4s, v3.4s
+; CHECK-NEXT:    eor v0.16b, v0.16b, v6.16b
+; CHECK-NEXT:    add v2.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    eor v1.16b, v1.16b, v7.16b
-; CHECK-NEXT:    eor v0.16b, v0.16b, v2.16b
-; CHECK-NEXT:    eor v2.16b, v3.16b, v5.16b
-; CHECK-NEXT:    eor v3.16b, v4.16b, v6.16b
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    addv s0, v0.4s
@@ -278,13 +278,13 @@ define i32 @v2(ptr nocapture noundef readonly %p1, i32 noundef %i1, ptr nocaptur
 ; CHECK-NEXT:    mov v1.d[1], v6.d[1]
 ; CHECK-NEXT:    add v2.4s, v2.4s, v16.4s
 ; CHECK-NEXT:    add v3.4s, v4.4s, v17.4s
+; CHECK-NEXT:    rev64 v5.4s, v2.4s
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    sub v1.4s, v17.4s, v4.4s
-; CHECK-NEXT:    rev64 v5.4s, v2.4s
 ; CHECK-NEXT:    rev64 v6.4s, v3.4s
+; CHECK-NEXT:    mov v5.d[1], v2.d[1]
 ; CHECK-NEXT:    sub v4.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    mov v5.d[1], v2.d[1]
 ; CHECK-NEXT:    mov v6.d[1], v3.d[1]
 ; CHECK-NEXT:    sub v3.4s, v3.4s, v5.4s
 ; CHECK-NEXT:    add v1.4s, v2.4s, v6.4s
@@ -304,40 +304,40 @@ define i32 @v2(ptr nocapture noundef readonly %p1, i32 noundef %i1, ptr nocaptur
 ; CHECK-NEXT:    mov v4.d[1], v16.d[1]
 ; CHECK-NEXT:    mov v1.d[1], v7.d[1]
 ; CHECK-NEXT:    add v0.4s, v17.4s, v1.4s
-; CHECK-NEXT:    sub v1.4s, v1.4s, v17.4s
 ; CHECK-NEXT:    add v2.4s, v18.4s, v4.4s
+; CHECK-NEXT:    sub v1.4s, v1.4s, v17.4s
 ; CHECK-NEXT:    sub v3.4s, v4.4s, v18.4s
-; CHECK-NEXT:    zip2 v4.4s, v0.4s, v1.4s
-; CHECK-NEXT:    ext v5.16b, v0.16b, v0.16b, #4
-; CHECK-NEXT:    ext v6.16b, v2.16b, v2.16b, #4
+; CHECK-NEXT:    ext v4.16b, v0.16b, v0.16b, #4
+; CHECK-NEXT:    ext v5.16b, v2.16b, v2.16b, #4
+; CHECK-NEXT:    zip2 v6.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    zip2 v7.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    zip2 v16.4s, v3.4s, v2.4s
 ; CHECK-NEXT:    zip2 v17.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    zip1 v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    zip1 v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    ext v1.16b, v5.16b, v1.16b, #8
-; CHECK-NEXT:    ext v18.16b, v6.16b, v3.16b, #8
-; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
-; CHECK-NEXT:    sub v4.4s, v4.4s, v17.4s
-; CHECK-NEXT:    sub v0.4s, v0.4s, v2.4s
-; CHECK-NEXT:    ext v1.16b, v1.16b, v5.16b, #4
-; CHECK-NEXT:    ext v5.16b, v18.16b, v6.16b, #4
-; CHECK-NEXT:    cmlt v2.8h, v4.8h, #0
-; CHECK-NEXT:    cmlt v6.8h, v3.8h, #0
-; CHECK-NEXT:    add v3.4s, v6.4s, v3.4s
-; CHECK-NEXT:    add v4.4s, v2.4s, v4.4s
-; CHECK-NEXT:    add v1.4s, v5.4s, v1.4s
+; CHECK-NEXT:    ext v18.16b, v4.16b, v1.16b, #8
+; CHECK-NEXT:    ext v19.16b, v5.16b, v3.16b, #8
+; CHECK-NEXT:    zip1 v1.4s, v2.4s, v3.4s
+; CHECK-NEXT:    add v2.4s, v16.4s, v7.4s
+; CHECK-NEXT:    sub v3.4s, v6.4s, v17.4s
+; CHECK-NEXT:    ext v4.16b, v18.16b, v4.16b, #4
+; CHECK-NEXT:    ext v5.16b, v19.16b, v5.16b, #4
+; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    cmlt v1.8h, v3.8h, #0
+; CHECK-NEXT:    cmlt v6.8h, v2.8h, #0
+; CHECK-NEXT:    add v4.4s, v5.4s, v4.4s
 ; CHECK-NEXT:    cmlt v5.8h, v0.8h, #0
+; CHECK-NEXT:    add v2.4s, v6.4s, v2.4s
+; CHECK-NEXT:    add v3.4s, v1.4s, v3.4s
+; CHECK-NEXT:    cmlt v7.8h, v4.8h, #0
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
-; CHECK-NEXT:    eor v2.16b, v4.16b, v2.16b
-; CHECK-NEXT:    eor v3.16b, v3.16b, v6.16b
-; CHECK-NEXT:    cmlt v4.8h, v1.8h, #0
-; CHECK-NEXT:    add v2.4s, v3.4s, v2.4s
-; CHECK-NEXT:    add v1.4s, v4.4s, v1.4s
+; CHECK-NEXT:    eor v2.16b, v2.16b, v6.16b
+; CHECK-NEXT:    eor v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    add v3.4s, v7.4s, v4.4s
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v5.16b
-; CHECK-NEXT:    add v0.4s, v0.4s, v2.4s
-; CHECK-NEXT:    eor v1.16b, v1.16b, v4.16b
-; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
+; CHECK-NEXT:    eor v2.16b, v3.16b, v7.16b
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    add v0.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    lsr w9, w8, #16
@@ -461,93 +461,93 @@ define i32 @v3(ptr nocapture noundef readonly %p1, i32 noundef %i1, ptr nocaptur
 ; CHECK-NEXT:    ldr d3, [x11, x9]
 ; CHECK-NEXT:    ldr d4, [x10]
 ; CHECK-NEXT:    ldr d5, [x11]
+; CHECK-NEXT:    shll2 v6.4s, v0.8h, #16
 ; CHECK-NEXT:    usubl v2.8h, v2.8b, v3.8b
-; CHECK-NEXT:    usubl v3.8h, v4.8b, v5.8b
-; CHECK-NEXT:    shll2 v4.4s, v0.8h, #16
-; CHECK-NEXT:    shll2 v5.4s, v1.8h, #16
-; CHECK-NEXT:    saddw v0.4s, v4.4s, v0.4h
-; CHECK-NEXT:    shll2 v4.4s, v2.8h, #16
-; CHECK-NEXT:    saddw v1.4s, v5.4s, v1.4h
-; CHECK-NEXT:    shll2 v5.4s, v3.8h, #16
-; CHECK-NEXT:    saddw v2.4s, v4.4s, v2.4h
-; CHECK-NEXT:    saddw v3.4s, v5.4s, v3.4h
-; CHECK-NEXT:    rev64 v4.4s, v0.4s
+; CHECK-NEXT:    shll2 v3.4s, v1.8h, #16
+; CHECK-NEXT:    usubl v4.8h, v4.8b, v5.8b
+; CHECK-NEXT:    saddw v0.4s, v6.4s, v0.4h
+; CHECK-NEXT:    shll2 v5.4s, v2.8h, #16
+; CHECK-NEXT:    saddw v1.4s, v3.4s, v1.4h
+; CHECK-NEXT:    shll2 v3.4s, v4.8h, #16
+; CHECK-NEXT:    rev64 v6.4s, v0.4s
+; CHECK-NEXT:    saddw v2.4s, v5.4s, v2.4h
 ; CHECK-NEXT:    rev64 v5.4s, v1.4s
-; CHECK-NEXT:    rev64 v6.4s, v2.4s
-; CHECK-NEXT:    rev64 v7.4s, v3.4s
-; CHECK-NEXT:    sub v4.4s, v0.4s, v4.4s
+; CHECK-NEXT:    saddw v3.4s, v3.4s, v4.4h
+; CHECK-NEXT:    rev64 v4.4s, v2.4s
+; CHECK-NEXT:    sub v6.4s, v0.4s, v6.4s
 ; CHECK-NEXT:    addp v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    rev64 v7.4s, v3.4s
 ; CHECK-NEXT:    sub v5.4s, v1.4s, v5.4s
-; CHECK-NEXT:    sub v6.4s, v2.4s, v6.4s
+; CHECK-NEXT:    sub v4.4s, v2.4s, v4.4s
 ; CHECK-NEXT:    addp v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    sub v1.4s, v3.4s, v7.4s
+; CHECK-NEXT:    ext v1.16b, v5.16b, v6.16b, #4
+; CHECK-NEXT:    sub v7.4s, v3.4s, v7.4s
 ; CHECK-NEXT:    ext v3.16b, v0.16b, v0.16b, #8
-; CHECK-NEXT:    ext v7.16b, v5.16b, v4.16b, #4
-; CHECK-NEXT:    mov v4.s[3], v5.s[2]
-; CHECK-NEXT:    zip2 v16.4s, v6.4s, v1.4s
-; CHECK-NEXT:    zip1 v1.4s, v6.4s, v1.4s
-; CHECK-NEXT:    uzp2 v6.4s, v2.4s, v0.4s
-; CHECK-NEXT:    ext v5.16b, v7.16b, v5.16b, #4
+; CHECK-NEXT:    mov v6.s[3], v5.s[2]
+; CHECK-NEXT:    zip2 v16.4s, v4.4s, v7.4s
+; CHECK-NEXT:    zip1 v4.4s, v4.4s, v7.4s
+; CHECK-NEXT:    ext v1.16b, v1.16b, v5.16b, #4
+; CHECK-NEXT:    uzp2 v5.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    uzp1 v0.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    uzp1 v7.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    uzp2 v2.4s, v2.4s, v3.4s
-; CHECK-NEXT:    mov v16.d[1], v4.d[1]
-; CHECK-NEXT:    rev64 v3.4s, v6.4s
-; CHECK-NEXT:    mov v1.d[1], v5.d[1]
+; CHECK-NEXT:    mov v16.d[1], v6.d[1]
+; CHECK-NEXT:    mov v4.d[1], v1.d[1]
+; CHECK-NEXT:    rev64 v1.4s, v5.4s
 ; CHECK-NEXT:    rev64 v0.4s, v0.4s
 ; CHECK-NEXT:    sub v2.4s, v7.4s, v2.4s
-; CHECK-NEXT:    sub v4.4s, v1.4s, v16.4s
-; CHECK-NEXT:    add v0.4s, v3.4s, v0.4s
-; CHECK-NEXT:    add v1.4s, v16.4s, v1.4s
-; CHECK-NEXT:    zip1 v3.4s, v2.4s, v4.4s
+; CHECK-NEXT:    sub v3.4s, v4.4s, v16.4s
+; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    add v1.4s, v16.4s, v4.4s
+; CHECK-NEXT:    zip1 v4.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    zip1 v5.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    uzp2 v6.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    zip2 v7.4s, v0.4s, v1.4s
-; CHECK-NEXT:    zip2 v17.4s, v2.4s, v4.4s
-; CHECK-NEXT:    ext v16.16b, v2.16b, v3.16b, #8
+; CHECK-NEXT:    zip2 v17.4s, v2.4s, v3.4s
+; CHECK-NEXT:    ext v16.16b, v2.16b, v4.16b, #8
 ; CHECK-NEXT:    trn2 v5.4s, v0.4s, v5.4s
 ; CHECK-NEXT:    uzp2 v6.4s, v6.4s, v0.4s
-; CHECK-NEXT:    mov v2.s[3], v4.s[2]
+; CHECK-NEXT:    mov v2.s[3], v3.s[2]
 ; CHECK-NEXT:    mov v0.s[1], v1.s[1]
 ; CHECK-NEXT:    mov v5.d[1], v16.d[1]
 ; CHECK-NEXT:    mov v6.d[1], v17.d[1]
 ; CHECK-NEXT:    mov v7.d[1], v2.d[1]
-; CHECK-NEXT:    mov v0.d[1], v3.d[1]
+; CHECK-NEXT:    mov v0.d[1], v4.d[1]
 ; CHECK-NEXT:    add v1.4s, v6.4s, v7.4s
-; CHECK-NEXT:    sub v2.4s, v7.4s, v6.4s
-; CHECK-NEXT:    add v3.4s, v5.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v5.4s, v0.4s
+; CHECK-NEXT:    sub v3.4s, v7.4s, v6.4s
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v5.4s
-; CHECK-NEXT:    zip2 v4.4s, v1.4s, v2.4s
-; CHECK-NEXT:    ext v5.16b, v1.16b, v1.16b, #4
-; CHECK-NEXT:    ext v6.16b, v3.16b, v3.16b, #4
-; CHECK-NEXT:    zip2 v7.4s, v2.4s, v1.4s
-; CHECK-NEXT:    zip2 v16.4s, v0.4s, v3.4s
-; CHECK-NEXT:    zip2 v17.4s, v3.4s, v0.4s
-; CHECK-NEXT:    zip1 v1.4s, v1.4s, v2.4s
-; CHECK-NEXT:    ext v2.16b, v5.16b, v2.16b, #8
-; CHECK-NEXT:    ext v18.16b, v6.16b, v0.16b, #8
-; CHECK-NEXT:    zip1 v0.4s, v3.4s, v0.4s
-; CHECK-NEXT:    add v3.4s, v16.4s, v7.4s
-; CHECK-NEXT:    sub v4.4s, v4.4s, v17.4s
-; CHECK-NEXT:    ext v2.16b, v2.16b, v5.16b, #4
-; CHECK-NEXT:    ext v5.16b, v18.16b, v6.16b, #4
+; CHECK-NEXT:    ext v4.16b, v1.16b, v1.16b, #4
+; CHECK-NEXT:    ext v5.16b, v2.16b, v2.16b, #4
+; CHECK-NEXT:    zip2 v6.4s, v1.4s, v3.4s
+; CHECK-NEXT:    zip2 v7.4s, v3.4s, v1.4s
+; CHECK-NEXT:    zip2 v16.4s, v0.4s, v2.4s
+; CHECK-NEXT:    zip2 v17.4s, v2.4s, v0.4s
+; CHECK-NEXT:    zip1 v1.4s, v1.4s, v3.4s
+; CHECK-NEXT:    ext v18.16b, v4.16b, v3.16b, #8
+; CHECK-NEXT:    ext v19.16b, v5.16b, v0.16b, #8
+; CHECK-NEXT:    zip1 v0.4s, v2.4s, v0.4s
+; CHECK-NEXT:    add v2.4s, v16.4s, v7.4s
+; CHECK-NEXT:    sub v3.4s, v6.4s, v17.4s
+; CHECK-NEXT:    ext v4.16b, v18.16b, v4.16b, #4
+; CHECK-NEXT:    ext v5.16b, v19.16b, v5.16b, #4
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    cmlt v1.8h, v4.8h, #0
-; CHECK-NEXT:    cmlt v6.8h, v3.8h, #0
-; CHECK-NEXT:    add v3.4s, v6.4s, v3.4s
-; CHECK-NEXT:    add v4.4s, v1.4s, v4.4s
-; CHECK-NEXT:    add v2.4s, v5.4s, v2.4s
+; CHECK-NEXT:    cmlt v1.8h, v3.8h, #0
+; CHECK-NEXT:    cmlt v6.8h, v2.8h, #0
+; CHECK-NEXT:    add v4.4s, v5.4s, v4.4s
 ; CHECK-NEXT:    cmlt v5.8h, v0.8h, #0
+; CHECK-NEXT:    add v2.4s, v6.4s, v2.4s
+; CHECK-NEXT:    add v3.4s, v1.4s, v3.4s
+; CHECK-NEXT:    cmlt v7.8h, v4.8h, #0
 ; CHECK-NEXT:    add v0.4s, v5.4s, v0.4s
-; CHECK-NEXT:    eor v1.16b, v4.16b, v1.16b
-; CHECK-NEXT:    eor v3.16b, v3.16b, v6.16b
-; CHECK-NEXT:    cmlt v4.8h, v2.8h, #0
-; CHECK-NEXT:    add v1.4s, v3.4s, v1.4s
-; CHECK-NEXT:    add v2.4s, v4.4s, v2.4s
+; CHECK-NEXT:    eor v2.16b, v2.16b, v6.16b
+; CHECK-NEXT:    eor v1.16b, v3.16b, v1.16b
+; CHECK-NEXT:    add v3.4s, v7.4s, v4.4s
 ; CHECK-NEXT:    eor v0.16b, v0.16b, v5.16b
+; CHECK-NEXT:    add v1.4s, v2.4s, v1.4s
+; CHECK-NEXT:    eor v2.16b, v3.16b, v7.16b
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    eor v1.16b, v2.16b, v4.16b
-; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
+; CHECK-NEXT:    add v0.4s, v2.4s, v0.4s
 ; CHECK-NEXT:    addv s0, v0.4s
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    lsr w9, w8, #16
diff --git a/llvm/test/CodeGen/AArch64/sat-add.ll b/llvm/test/CodeGen/AArch64/sat-add.ll
index 86c224bee990a..2deb19be24821 100644
--- a/llvm/test/CodeGen/AArch64/sat-add.ll
+++ b/llvm/test/CodeGen/AArch64/sat-add.ll
@@ -346,9 +346,9 @@ define <16 x i8> @unsigned_sat_constant_v16i8_using_min(<16 x i8> %x) {
 ; CHECK-LABEL: unsigned_sat_constant_v16i8_using_min:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.16b, #213
+; CHECK-NEXT:    movi v2.16b, #42
 ; CHECK-NEXT:    umin v0.16b, v0.16b, v1.16b
-; CHECK-NEXT:    movi v1.16b, #42
-; CHECK-NEXT:    add v0.16b, v0.16b, v1.16b
+; CHECK-NEXT:    add v0.16b, v0.16b, v2.16b
 ; CHECK-NEXT:    ret
   %c = icmp ult <16 x i8> %x, <i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43>
   %s = select <16 x i1> %c, <16 x i8> %x, <16 x i8> <i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43, i8 -43>
@@ -384,9 +384,9 @@ define <8 x i16> @unsigned_sat_constant_v8i16_using_min(<8 x i16> %x) {
 ; CHECK-LABEL: unsigned_sat_constant_v8i16_using_min:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mvni v1.8h, #42
+; CHECK-NEXT:    movi v2.8h, #42
 ; CHECK-NEXT:    umin v0.8h, v0.8h, v1.8h
-; CHECK-NEXT:    movi v1.8h, #42
-; CHECK-NEXT:    add v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    add v0.8h, v0.8h, v2.8h
 ; CHECK-NEXT:    ret
   %c = icmp ult <8 x i16> %x, <i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43>
   %s = select <8 x i1> %c, <8 x i16> %x, <8 x i16> <i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43, i16 -43>
diff --git a/llvm/test/CodeGen/AArch64/sext.ll b/llvm/test/CodeGen/AArch64/sext.ll
index 3e0d5dd875097..5237a3491de9b 100644
--- a/llvm/test/CodeGen/AArch64/sext.ll
+++ b/llvm/test/CodeGen/AArch64/sext.ll
@@ -245,15 +245,15 @@ define <3 x i32> @sext_v3i8_v3i32(<3 x i8> %a) {
 ; CHECK-GI-LABEL: sext_v3i8_v3i32:
 ; CHECK-GI:       // %bb.0: // %entry
 ; CHECK-GI-NEXT:    mov w8, #24 // =0x18
-; CHECK-GI-NEXT:    fmov s1, w0
-; CHECK-GI-NEXT:    fmov s0, w8
-; CHECK-GI-NEXT:    mov v1.s[1], w1
-; CHECK-GI-NEXT:    mov v0.s[1], w8
-; CHECK-GI-NEXT:    mov v1.s[2], w2
-; CHECK-GI-NEXT:    mov v0.s[2], w8
-; CHECK-GI-NEXT:    neg v2.4s, v0.4s
-; CHECK-GI-NEXT:    ushl v0.4s, v1.4s, v0.4s
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v2.4s
+; CHECK-GI-NEXT:    fmov s0, w0
+; CHECK-GI-NEXT:    fmov s1, w8
+; CHECK-GI-NEXT:    mov v0.s[1], w1
+; CHECK-GI-NEXT:    mov v1.s[1], w8
+; CHECK-GI-NEXT:    mov v0.s[2], w2
+; CHECK-GI-NEXT:    mov v1.s[2], w8
+; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v1.4s
+; CHECK-GI-NEXT:    neg v1.4s, v1.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-GI-NEXT:    ret
 entry:
   %c = sext <3 x i8> %a to <3 x i32>
@@ -408,15 +408,15 @@ define <3 x i32> @sext_v3i10_v3i32(<3 x i10> %a) {
 ; CHECK-GI-LABEL: sext_v3i10_v3i32:
 ; CHECK-GI:       // %bb.0: // %entry
 ; CHECK-GI-NEXT:    mov w8, #22 // =0x16
-; CHECK-GI-NEXT:    fmov s1, w0
-; CHECK-GI-NEXT:    fmov s0, w8
-; CHECK-GI-NEXT:    mov v1.s[1], w1
-; CHECK-GI-NEXT:    mov v0.s[1], w8
-; CHECK-GI-NEXT:    mov v1.s[2], w2
-; CHECK-GI-NEXT:    mov v0.s[2], w8
-; CHECK-GI-NEXT:    neg v2.4s, v0.4s
-; CHECK-GI-NEXT:    ushl v0.4s, v1.4s, v0.4s
-; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v2.4s
+; CHECK-GI-NEXT:    fmov s0, w0
+; CHECK-GI-NEXT:    fmov s1, w8
+; CHECK-GI-NEXT:    mov v0.s[1], w1
+; CHECK-GI-NEXT:    mov v1.s[1], w8
+; CHECK-GI-NEXT:    mov v0.s[2], w2
+; CHECK-GI-NEXT:    mov v1.s[2], w8
+; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v1.4s
+; CHECK-GI-NEXT:    neg v1.4s, v1.4s
+; CHECK-GI-NEXT:    sshl v0.4s, v0.4s, v1.4s
 ; CHECK-GI-NEXT:    ret
 entry:
   %c = sext <3 x i10> %a to <3 x i32>
diff --git a/llvm/test/CodeGen/AArch64/sink-addsub-of-const.ll b/llvm/test/CodeGen/AArch64/sink-addsub-of-const.ll
index f4f75bb9c7825..88e062d2c999c 100644
--- a/llvm/test/CodeGen/AArch64/sink-addsub-of-const.ll
+++ b/llvm/test/CodeGen/AArch64/sink-addsub-of-const.ll
@@ -158,8 +158,8 @@ define i32 @sink_sub_from_const_to_sub2(i32 %a, i32 %b) {
 define <4 x i32> @vec_sink_add_of_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_add_of_const_to_add0:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI12_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI12_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -170,8 +170,8 @@ define <4 x i32> @vec_sink_add_of_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_add_of_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_add_of_const_to_add1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI13_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI13_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -186,8 +186,8 @@ define <4 x i32> @vec_sink_add_of_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_of_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_of_const_to_add0:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI14_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI14_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -198,8 +198,8 @@ define <4 x i32> @vec_sink_sub_of_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_of_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_of_const_to_add1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI15_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI15_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -214,8 +214,8 @@ define <4 x i32> @vec_sink_sub_of_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_from_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_from_const_to_add0:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    adrp x8, .LCPI16_0
+; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI16_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -226,8 +226,8 @@ define <4 x i32> @vec_sink_sub_from_const_to_add0(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_from_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_from_const_to_add1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    adrp x8, .LCPI17_0
+; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI17_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -242,8 +242,8 @@ define <4 x i32> @vec_sink_sub_from_const_to_add1(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_add_of_const_to_sub:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI18_0
+; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI18_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -254,8 +254,8 @@ define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_add_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_add_of_const_to_sub2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    adrp x8, .LCPI19_0
+; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI19_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -270,8 +270,8 @@ define <4 x i32> @vec_sink_add_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_of_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_of_const_to_sub:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI20_0
+; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI20_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -282,8 +282,8 @@ define <4 x i32> @vec_sink_sub_of_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_of_const_to_sub2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    adrp x8, .LCPI21_0
+; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI21_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -298,8 +298,8 @@ define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_from_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_from_const_to_sub:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI22_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI22_0]
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
@@ -310,8 +310,8 @@ define <4 x i32> @vec_sink_sub_from_const_to_sub(<4 x i32> %a, <4 x i32> %b) {
 define <4 x i32> @vec_sink_sub_from_const_to_sub2(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: vec_sink_sub_from_const_to_sub2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    adrp x8, .LCPI23_0
+; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI23_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll b/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
index 0c674c5685e00..1d1bae42c9e30 100644
--- a/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
+++ b/llvm/test/CodeGen/AArch64/sme-pstate-sm-changing-call-disable-coalescing.ll
@@ -28,8 +28,8 @@ define void @dont_coalesce_arg_i8(i8 %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl use_i8
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr z0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
@@ -61,8 +61,8 @@ define void @dont_coalesce_arg_i16(i16 %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl use_i16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
@@ -94,8 +94,8 @@ define void @dont_coalesce_arg_i32(i32 %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl use_i32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr z0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
@@ -127,8 +127,8 @@ define void @dont_coalesce_arg_i64(i64 %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl use_i64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
@@ -165,8 +165,8 @@ define void @dont_coalesce_arg_f16(half %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr h0, [sp, #14] // 2-byte Folded Reload
 ; CHECK-NEXT:    bl use_f16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -205,8 +205,8 @@ define void @dont_coalesce_arg_f32(float %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
 ; CHECK-NEXT:    bl use_f32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -245,8 +245,8 @@ define void @dont_coalesce_arg_f64(double %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_f64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -290,8 +290,8 @@ define void @dont_coalesce_arg_v1i8(<1 x i8> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v16i8
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -331,8 +331,8 @@ define void @dont_coalesce_arg_v1i16(<1 x i16> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v8i16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -372,8 +372,8 @@ define void @dont_coalesce_arg_v1i32(<1 x i32> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v4i32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -413,8 +413,8 @@ define void @dont_coalesce_arg_v1i64(<1 x i64> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v2i64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -454,8 +454,8 @@ define void @dont_coalesce_arg_v1f16(<1 x half> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr h0, [sp, #14] // 2-byte Folded Reload
 ; CHECK-NEXT:    bl use_v8f16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -495,8 +495,8 @@ define void @dont_coalesce_arg_v1f32(<1 x float> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v4f32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -536,8 +536,8 @@ define void @dont_coalesce_arg_v1f64(<1 x double> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    bl use_v2f64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -581,8 +581,8 @@ define void @dont_coalesce_arg_v16i8(<16 x i8> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v16i8
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -621,8 +621,8 @@ define void @dont_coalesce_arg_v8i16(<8 x i16> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v8i16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -661,8 +661,8 @@ define void @dont_coalesce_arg_v4i32(<4 x i32> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v4i32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -701,8 +701,8 @@ define void @dont_coalesce_arg_v2i64(<2 x i64> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v2i64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -741,8 +741,8 @@ define void @dont_coalesce_arg_v8f16(<8 x half> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v8f16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -781,8 +781,8 @@ define void @dont_coalesce_arg_v8bf16(<8 x bfloat> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v8bf16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -821,8 +821,8 @@ define void @dont_coalesce_arg_v4f32(<4 x float> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v4f32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -861,8 +861,8 @@ define void @dont_coalesce_arg_v2f64(<2 x double> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl use_v2f64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr z0, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -894,9 +894,9 @@ define void @dont_coalesce_arg_v8i1(<8 x i1> %arg, ptr %ptr) #0 {
 ; CHECK-NEXT:    stp x30, x19, [sp, #80] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov z1.d, z0.d
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    add x8, sp, #16
 ; CHECK-NEXT:    mov x19, x0
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -942,13 +942,13 @@ define void @dont_coalesce_res_i8(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl get_i8
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    fmov s0, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldp d9, d8, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
 ; CHECK-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp], #80 // 16-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call i8 @get_i8()
@@ -969,13 +969,13 @@ define void @dont_coalesce_res_i16(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl get_i16
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fmov s0, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldp d9, d8, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
 ; CHECK-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp], #80 // 16-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call i16 @get_i16()
@@ -996,13 +996,13 @@ define void @dont_coalesce_res_i32(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl get_i32
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fmov s0, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp d9, d8, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
 ; CHECK-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp], #80 // 16-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call i32 @get_i32()
@@ -1023,13 +1023,13 @@ define void @dont_coalesce_res_i64(ptr %ptr) #0 {
 ; CHECK-NEXT:    smstop sm
 ; CHECK-NEXT:    bl get_i64
 ; CHECK-NEXT:    smstart sm
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fmov d0, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp d9, d8, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
 ; CHECK-NEXT:    ldp x30, x19, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    ldp d13, d12, [sp, #16] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp], #80 // 16-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call i64 @get_i64()
@@ -1056,11 +1056,11 @@ define void @dont_coalesce_res_f16(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr h0, [sp, #14] // 2-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call half @get_f16()
@@ -1086,11 +1086,11 @@ define void @dont_coalesce_res_f32(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr s0, [sp, #12] // 4-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call float @get_f32()
@@ -1116,11 +1116,11 @@ define void @dont_coalesce_res_f64(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call double @get_f64()
@@ -1150,11 +1150,11 @@ define void @dont_coalesce_res_v1i8(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x i8> @get_v1i8()
@@ -1181,11 +1181,11 @@ define void @dont_coalesce_res_v1i16(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x i16> @get_v1i16()
@@ -1212,11 +1212,11 @@ define void @dont_coalesce_res_v1i32(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x i32> @get_v1i32()
@@ -1243,11 +1243,11 @@ define void @dont_coalesce_res_v1i64(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x i64> @get_v1i64()
@@ -1275,11 +1275,11 @@ define void @dont_coalesce_res_v1f16(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr h0, [sp, #14] // 2-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x half> @get_v1f16()
@@ -1306,11 +1306,11 @@ define void @dont_coalesce_res_v1f32(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x float> @get_v1f32()
@@ -1337,11 +1337,11 @@ define void @dont_coalesce_res_v1f64(ptr %ptr) #0 {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr d0, [sp, #8] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <1 x double> @get_v1f64()
@@ -1373,11 +1373,11 @@ define void @dont_coalesce_res_v16i8(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1b { z0.b }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <16 x i8> @get_v16i8()
@@ -1404,11 +1404,11 @@ define void @dont_coalesce_res_v8i16(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <8 x i16> @get_v8i16()
@@ -1435,11 +1435,11 @@ define void @dont_coalesce_res_v4i32(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <4 x i32> @get_v4i32()
@@ -1466,11 +1466,11 @@ define void @dont_coalesce_res_v2i64(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <2 x i64> @get_v2i64()
@@ -1497,11 +1497,11 @@ define void @dont_coalesce_res_v8f16(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1h { z0.h }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <8 x half> @get_v8f16()
@@ -1528,11 +1528,11 @@ define void @dont_coalesce_res_v4f32(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1w { z0.s }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <4 x float> @get_v4f32()
@@ -1559,11 +1559,11 @@ define void @dont_coalesce_res_v2f64(ptr %ptr) #0 {
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    ldp d9, d8, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
+; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d11, d10, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d13, d12, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d15, d14, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    st1d { z0.d }, p0, [x19]
-; CHECK-NEXT:    ldp x30, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #96
 ; CHECK-NEXT:    ret
   %res = call <2 x double> @get_v2f64()
diff --git a/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll b/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
index 47b24290d3c85..1e16f140676ba 100644
--- a/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
+++ b/llvm/test/CodeGen/AArch64/sme-streaming-compatible-interface.ll
@@ -151,8 +151,8 @@ define <2 x double> @streaming_compatible_with_neon_vectors(<2 x double> %arg) "
 ; CHECK-NEXT:  // %bb.3:
 ; CHECK-NEXT:    smstart sm
 ; CHECK-NEXT:  .LBB4_4:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr z1, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-fp-dots.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-fp-dots.ll
index 0097968b1171d..b4fd5a2272e7e 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-fp-dots.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-fp-dots.ll
@@ -26,16 +26,16 @@ define void @fdot_multi_za32_f16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @fdot_multi_za32_f16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1, <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3,
 ; CHECK-LABEL: fdot_multi_za32_f16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    fdot za.s[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    fdot za.s[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -71,16 +71,16 @@ define void @bfdot_multi_za32_bf16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused,
 define void @fdot_multi_za32_bf16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1, <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3,
 ; CHECK-LABEL: fdot_multi_za32_bf16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    bfdot za.s[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    bfdot za.s[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-int-dots.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-int-dots.ll
index 6d98604837115..e154a4df86efe 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-int-dots.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-int-dots.ll
@@ -26,16 +26,16 @@ define void @udot_multi_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @udot_multi_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
 ; CHECK-LABEL: udot_multi_za32_u16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -68,16 +68,16 @@ define void @udot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <v
 define void @udot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
 ; CHECK-LABEL: udot_multi_za32_u8_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -110,16 +110,16 @@ define void @udot_multi_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @udot_multi_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
 ; CHECK-LABEL: udot_multi_za64_u16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    udot za.d[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    udot za.d[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -152,16 +152,16 @@ define void @usdot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @usdot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
 ; CHECK-LABEL: usdot_multi_za32_u8_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    usdot za.s[w8, 7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -197,16 +197,16 @@ define void @sdot_multi_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @sdot_multi_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
 ; CHECK-LABEL: sdot_multi_za32_u16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -239,16 +239,16 @@ define void @sdot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <v
 define void @sdot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
 ; CHECK-LABEL: sdot_multi_za32_u8_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -281,16 +281,16 @@ define void @sdot_multi_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <
 define void @sdot_multi_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
 ; CHECK-LABEL: sdot_multi_za64_u16_vg1x4:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    sdot za.d[w8, 7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-max.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-max.ll
index e95d29f65e55e..92e8877927ea5 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-max.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-max.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 -verify-machineinstrs < %s | FileCheck %s
 
 ; SMAX (Single, x2)
@@ -151,8 +152,7 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_single_x2
 
 ; SMAX (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_max_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_max_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_s8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -170,8 +170,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_max_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_max_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_s16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -189,8 +188,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_max_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_max_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_s32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -208,8 +206,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_max_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_max_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_s64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -229,8 +226,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; UMAX (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_max_single_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_max_single_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_u8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -248,8 +244,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_max_single_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_max_single_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_u16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -267,8 +262,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_max_single_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_max_single_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_u32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -286,8 +280,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_max_single_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_max_single_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_u64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -307,8 +300,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; FMAX (SINGLE, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_max_single_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_max_single_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -326,8 +318,7 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_max_single_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_single_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -345,8 +336,7 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_max_single_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_single_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
 ; CHECK-LABEL: multi_vec_max_single_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -537,104 +527,100 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x2_
 
 ; SMAX (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_max_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
-                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_max_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_s8:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smax { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
               @llvm.aarch64.sve.smax.x4.nxv16i8(<vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
                                                 <vscale x 16 x i8> %zm1,  <vscale x 16 x i8> %zm2,  <vscale x 16 x i8> %zm3,  <vscale x 16 x i8> %zm4)
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_max_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
-                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_max_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_s16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smax { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
               @llvm.aarch64.sve.smax.x4.nxv8i16(<vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
                                                 <vscale x 8 x i16> %zm1,  <vscale x 8 x i16> %zm2,  <vscale x 8 x i16> %zm3,  <vscale x 8 x i16> %zm4)
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_max_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
-                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_max_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_s32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smax { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
               @llvm.aarch64.sve.smax.x4.nxv4i32(<vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
                                                 <vscale x 4 x i32> %zm1,  <vscale x 4 x i32> %zm2,  <vscale x 4 x i32> %zm3,  <vscale x 4 x i32> %zm4)
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_max_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
-                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_max_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_s64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smax { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
               @llvm.aarch64.sve.smax.x4.nxv2i64(<vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
                                                 <vscale x 2 x i64> %zm1,  <vscale x 2 x i64> %zm2,  <vscale x 2 x i64> %zm3,  <vscale x 2 x i64> %zm4)
@@ -643,104 +629,100 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; UMAX (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_max_multi_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
-                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_max_multi_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_u8:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umax { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
               @llvm.aarch64.sve.umax.x4.nxv16i8(<vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
                                                 <vscale x 16 x i8> %zm1,  <vscale x 16 x i8> %zm2,  <vscale x 16 x i8> %zm3,  <vscale x 16 x i8> %zm4)
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_max_multi_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
-                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_max_multi_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_u16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umax { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
               @llvm.aarch64.sve.umax.x4.nxv8i16(<vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
                                                 <vscale x 8 x i16> %zm1,  <vscale x 8 x i16> %zm2,  <vscale x 8 x i16> %zm3,  <vscale x 8 x i16> %zm4)
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_max_multi_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
-                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_max_multi_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_u32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umax { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
               @llvm.aarch64.sve.umax.x4.nxv4i32(<vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
                                                 <vscale x 4 x i32> %zm1,  <vscale x 4 x i32> %zm2,  <vscale x 4 x i32> %zm3,  <vscale x 4 x i32> %zm4)
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_max_multi_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
-                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_max_multi_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_u64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umax { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
               @llvm.aarch64.sve.umax.x4.nxv2i64(<vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
                                                 <vscale x 2 x i64> %zm1,  <vscale x 2 x i64> %zm2,  <vscale x 2 x i64> %zm3,  <vscale x 2 x i64> %zm4)
@@ -749,78 +731,75 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; FMAX (Multi, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_max_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
-                            <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_max_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_f16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmax { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
   %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
               @llvm.aarch64.sve.fmax.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
                                                 <vscale x 8 x half> %zm1,  <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_max_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
-                            <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_f32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmax { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
   %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
               @llvm.aarch64.sve.fmax.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
                                                 <vscale x 4 x float> %zm1,  <vscale x 4 x float> %zm2,  <vscale x 4 x float> %zm3,  <vscale x 4 x float> %zm4)
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_max_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
-                            <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
 ; CHECK-LABEL: multi_vec_max_multi_x4_f64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmax { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
   %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
               @llvm.aarch64.sve.fmax.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
                                                 <vscale x 2 x double> %zm1,  <vscale x 2 x double> %zm2,  <vscale x 2 x double> %zm3,  <vscale x 2 x double> %zm4)
@@ -870,8 +849,7 @@ define { <vscale x 2 x double>, <vscale x 2 x double> }  @multi_vec_maxnm_single
 
 ; FMAXNM (Single, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_maxnm_single_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_maxnm_single_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
 ; CHECK-LABEL: multi_vec_maxnm_single_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -889,8 +867,7 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_maxnm_single_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_maxnm_single_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
 ; CHECK-LABEL: multi_vec_maxnm_single_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -908,8 +885,7 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_maxnm_single_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_maxnm_single_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
 ; CHECK-LABEL: multi_vec_maxnm_single_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -976,19 +952,18 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_maxnm_x2_f64(
 
 ; FMAXNM (Multi, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_maxnm_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_maxnm_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
 ; CHECK-LABEL: multi_vec_maxnm_x4_f16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmaxnm { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -996,24 +971,23 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-              @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, 
+              @llvm.aarch64.sve.fmaxnm.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
                                                   <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_maxnm_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_maxnm_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
 ; CHECK-LABEL: multi_vec_maxnm_x4_f32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmaxnm { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -1026,19 +1000,18 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_maxnm_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_maxnm_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
 ; CHECK-LABEL: multi_vec_maxnm_x4_f64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmaxnm { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-min.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-min.ll
index 21a55c6436acd..363f9ba5d3530 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-min.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-min.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 -verify-machineinstrs < %s | FileCheck %s
 
 ; SMIN (Single, x2)
@@ -151,8 +152,7 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_single_x2
 
 ; SMIN (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_min_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_min_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_s8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -170,8 +170,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_min_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_min_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_s16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -189,8 +188,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_min_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_min_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_s32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -208,8 +206,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_min_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_min_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_s64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -229,8 +226,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; UMIN (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_min_single_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_min_single_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_u8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -248,8 +244,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_min_single_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_min_single_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_u16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -267,8 +262,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_min_single_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_min_single_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_u32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -286,8 +280,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_min_single_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_min_single_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_u64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -307,8 +300,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; FMIN (SINGLE, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_min_single_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_single_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -326,8 +318,7 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_min_single_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_single_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -345,8 +336,7 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_min_single_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_single_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
 ; CHECK-LABEL: multi_vec_min_single_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -537,104 +527,100 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_multi_x2_
 
 ; SMIN (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_min_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
-                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_min_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_s8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smin { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
               @llvm.aarch64.sve.smin.x4.nxv16i8(<vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
                                                 <vscale x 16 x i8> %zm1,  <vscale x 16 x i8> %zm2,  <vscale x 16 x i8> %zm3,  <vscale x 16 x i8> %zm4)
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_min_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
-                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_min_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_s16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smin { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
               @llvm.aarch64.sve.smin.x4.nxv8i16(<vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
                                                 <vscale x 8 x i16> %zm1,  <vscale x 8 x i16> %zm2,  <vscale x 8 x i16> %zm3,  <vscale x 8 x i16> %zm4)
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_min_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
-                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_min_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_s32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smin { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
               @llvm.aarch64.sve.smin.x4.nxv4i32(<vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
                                                 <vscale x 4 x i32> %zm1,  <vscale x 4 x i32> %zm2,  <vscale x 4 x i32> %zm3,  <vscale x 4 x i32> %zm4)
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_min_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
-                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_min_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_s64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    smin { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
               @llvm.aarch64.sve.smin.x4.nxv2i64(<vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
                                                 <vscale x 2 x i64> %zm1,  <vscale x 2 x i64> %zm2,  <vscale x 2 x i64> %zm3,  <vscale x 2 x i64> %zm4)
@@ -643,104 +629,100 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; UMIN (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_min_multi_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
-                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_min_multi_x4_u8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_u8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umin { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                           <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
               @llvm.aarch64.sve.umin.x4.nxv16i8(<vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
                                                 <vscale x 16 x i8> %zm1,  <vscale x 16 x i8> %zm2,  <vscale x 16 x i8> %zm3,  <vscale x 16 x i8> %zm4)
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_min_multi_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
-                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_min_multi_x4_u16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_u16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umin { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
               @llvm.aarch64.sve.umin.x4.nxv8i16(<vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
                                                 <vscale x 8 x i16> %zm1,  <vscale x 8 x i16> %zm2,  <vscale x 8 x i16> %zm3,  <vscale x 8 x i16> %zm4)
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_min_multi_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
-                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_min_multi_x4_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_u32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umin { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
               @llvm.aarch64.sve.umin.x4.nxv4i32(<vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
                                                 <vscale x 4 x i32> %zm1,  <vscale x 4 x i32> %zm2,  <vscale x 4 x i32> %zm3,  <vscale x 4 x i32> %zm4)
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_min_multi_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
-                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_min_multi_x4_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_u64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    umin { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
               @llvm.aarch64.sve.umin.x4.nxv2i64(<vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
                                                 <vscale x 2 x i64> %zm1,  <vscale x 2 x i64> %zm2,  <vscale x 2 x i64> %zm3,  <vscale x 2 x i64> %zm4)
@@ -749,78 +731,75 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 
 ; FMIN (Multi, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_min_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
-                            <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmin { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
   %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
               @llvm.aarch64.sve.fmin.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
                                                 <vscale x 8 x half> %zm1,  <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_min_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
-                            <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmin { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
   %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
               @llvm.aarch64.sve.fmin.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
                                                 <vscale x 4 x float> %zm1,  <vscale x 4 x float> %zm2,  <vscale x 4 x float> %zm3,  <vscale x 4 x float> %zm4)
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_min_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
-                            <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
 ; CHECK-LABEL: multi_vec_min_multi_x4_f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fmin { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                            <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
   %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
               @llvm.aarch64.sve.fmin.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
                                                 <vscale x 2 x double> %zm1,  <vscale x 2 x double> %zm2,  <vscale x 2 x double> %zm3,  <vscale x 2 x double> %zm4)
@@ -870,8 +849,7 @@ define { <vscale x 2 x double>, <vscale x 2 x double> }  @multi_vec_minnm_single
 
 ; FMINNM (Single, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_minnm_single_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_minnm_single_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm) {
 ; CHECK-LABEL: multi_vec_minnm_single_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -889,8 +867,7 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_minnm_single_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_minnm_single_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm) {
 ; CHECK-LABEL: multi_vec_minnm_single_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -908,8 +885,7 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_minnm_single_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_minnm_single_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm) {
 ; CHECK-LABEL: multi_vec_minnm_single_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -976,19 +952,18 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_minnm_x2_f64(
 
 ; FMINNM (Multi, x4)
 
-define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
-@multi_vec_minnm_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_minnm_x4_f16(<vscale x 8 x half> %dummy, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
 ; CHECK-LABEL: multi_vec_minnm_x4_f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fminnm { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -1001,19 +976,18 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
-define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
-@multi_vec_minnm_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_minnm_x4_f32(<vscale x 8 x half> %dummy, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
 ; CHECK-LABEL: multi_vec_minnm_x4_f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fminnm { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -1026,19 +1000,18 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
-define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
-@multi_vec_minnm_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_minnm_x4_f64(<vscale x 8 x half> %dummy, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
 ; CHECK-LABEL: multi_vec_minnm_x4_f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z30.d, z7.d
 ; CHECK-NEXT:    mov z27.d, z4.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z29.d, z6.d
 ; CHECK-NEXT:    mov z26.d, z3.d
 ; CHECK-NEXT:    mov z28.d, z5.d
 ; CHECK-NEXT:    mov z25.d, z2.d
-; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    fminnm { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-mlall.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-mlall.ll
index f766bfcff4d1d..346afc611eb75 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-mlall.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-mlall.ll
@@ -142,16 +142,16 @@ define void @multi_vector_mul_add_multi_long_vg4x2_s16(i32 %slice, <vscale x 8 x
 define void @multi_vector_mul_add_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_s8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -164,16 +164,16 @@ define void @multi_vector_mul_add_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x
 define void @multi_vector_mul_add_multi_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_s16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -418,16 +418,16 @@ define void @multi_vector_mul_add_multi_long_vg4x2_u16(i32 %slice, <vscale x 8 x
 define void @multi_vector_mul_add_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_u8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -440,16 +440,16 @@ define void @multi_vector_mul_add_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x
 define void @multi_vector_mul_add_multi_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_u16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -694,16 +694,16 @@ define void @multi_vector_mul_sub_multi_long_vg4x2_s16(i32 %slice, <vscale x 8 x
 define void @multi_vector_mul_sub_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_s8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -716,16 +716,16 @@ define void @multi_vector_mul_sub_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x
 define void @multi_vector_mul_sub_multi_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_s16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -970,16 +970,16 @@ define void @multi_vector_mul_sub_multi_long_vg4x2_u16(i32 %slice, <vscale x 8 x
 define void @multi_vector_mul_sub_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_u8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
@@ -992,16 +992,16 @@ define void @multi_vector_mul_sub_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x
 define void @multi_vector_mul_sub_multi_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_u16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ret
@@ -1275,16 +1275,16 @@ define void @multi_vector_mul_add_multi_unsigned_long_vg4x2_u8(i32 %slice, <vsca
 define void @multi_vector_mul_add_multi_unsigned_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
 ; CHECK-LABEL: multi_vector_mul_add_multi_unsigned_long_vg4x4_u8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-rshl.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-rshl.ll
index d138a3af43852..12a940ff03e29 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-rshl.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-rshl.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 -verify-machineinstrs < %s | FileCheck %s
 
 ; SRSHL (Single, x2)
@@ -56,8 +57,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_single
 
 ; SRSHL (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_rounding_shl_single_x4_s8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_rounding_shl_single_x4_s8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_s8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -75,8 +75,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_rounding_shl_single_x4_s16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_rounding_shl_single_x4_s16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_s16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -94,8 +93,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_rounding_shl_single_x4_s32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_rounding_shl_single_x4_s32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_s32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -113,8 +111,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_rounding_shl_single_x4_s64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_single_x4_s64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_s64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -188,8 +185,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_single
 
 ; URSHL (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_rounding_shl_single_x4_u8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_rounding_shl_single_x4_u8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_u8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -207,8 +203,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_rounding_shl_single_x4_u16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_rounding_shl_single_x4_u16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_u16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -226,8 +221,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_rounding_shl_single_x4_u32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_rounding_shl_single_x4_u32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_u32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -245,8 +239,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_rounding_shl_single_x4_u64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_single_x4_u64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_rounding_shl_single_x4_u64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -328,19 +321,18 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_x2_s64
 
 ; SRSHL (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_rounding_shl_x4_s8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_rounding_shl_x4_s8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_s8:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    srshl { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -353,19 +345,18 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_rounding_shl_x4_s16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_rounding_shl_x4_s16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_s16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    srshl { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -378,19 +369,18 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_rounding_shl_x4_s32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_rounding_shl_x4_s32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_s32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    srshl { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -403,19 +393,18 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_rounding_shl_x4_s64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_x4_s64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_s64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    srshl { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -492,19 +481,18 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_uhl_x2_u64
 
 ; URSHL (Multi, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_rounding_shl_x4_u8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_rounding_shl_x4_u8(<vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_u8:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    urshl { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -517,19 +505,18 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_rounding_shl_x4_u16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_rounding_shl_x4_u16(<vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_u16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    urshl { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -542,19 +529,18 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_rounding_shl_x4_u32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_rounding_shl_x4_u32(<vscale x 4 x i32> %dummy, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_u32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    urshl { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
@@ -567,19 +553,18 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_rounding_shl_x4_u64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_rounding_shl_x4_u64(<vscale x 2 x i64> %dummy, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
 ; CHECK-LABEL: multi_vec_rounding_shl_x4_u64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    urshl { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-sqdmulh.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-sqdmulh.ll
index 9c5dff6c3bf6f..e71afe213d8a5 100644
--- a/llvm/test/CodeGen/AArch64/sme2-intrinsics-sqdmulh.ll
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-sqdmulh.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 -verify-machineinstrs < %s | FileCheck %s
 
 ; SQDMULH (Single, x2)
@@ -56,8 +57,7 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_sat_double_mulh_sin
 
 ; SQDMULH (Single, x4)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_sat_double_mulh_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_sat_double_mulh_single_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4, <vscale x 16 x i8> %zm) {
 ; CHECK-LABEL: multi_vec_sat_double_mulh_single_x4_s8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -75,8 +75,7 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_sat_double_mulh_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_sat_double_mulh_single_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4, <vscale x 8 x i16> %zm) {
 ; CHECK-LABEL: multi_vec_sat_double_mulh_single_x4_s16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -94,8 +93,7 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_sat_double_mulh_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_sat_double_mulh_single_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4, <vscale x 4 x i32> %zm) {
 ; CHECK-LABEL: multi_vec_sat_double_mulh_single_x4_s32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -113,8 +111,7 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_sat_double_mulh_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_sat_double_mulh_single_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4, <vscale x 2 x i64> %zm) {
 ; CHECK-LABEL: multi_vec_sat_double_mulh_single_x4_s64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z27.d, z4.d
@@ -196,104 +193,100 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_sat_double_mulh_mul
 
 ; SQDMULH (x4, Multi)
 
-define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
-@multi_vec_sat_double_mulh_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
-                                       <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @multi_vec_sat_double_mulh_multi_x4_s8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
 ; CHECK-LABEL: multi_vec_sat_double_mulh_multi_x4_s8:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1b { z31.b }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    sqdmulh { z24.b - z27.b }, { z24.b - z27.b }, { z28.b - z31.b }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                                       <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4) {
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> }
               @llvm.aarch64.sve.sqdmulh.vgx4.nxv16i8(<vscale x 16 x i8> %zdn1, <vscale x 16 x i8> %zdn2, <vscale x 16 x i8> %zdn3, <vscale x 16 x i8> %zdn4,
                                                      <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3, <vscale x 16 x i8> %zm4)
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
-define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
-@multi_vec_sat_double_mulh_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
-                                        <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @multi_vec_sat_double_mulh_multi_x4_s16(<vscale x 8 x i16> %unused, <vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
 ; CHECK-LABEL: multi_vec_sat_double_mulh_multi_x4_s16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1h { z31.h }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    sqdmulh { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                                        <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4) {
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> }
               @llvm.aarch64.sve.sqdmulh.vgx4.nxv8i16(<vscale x 8 x i16> %zdn1, <vscale x 8 x i16> %zdn2, <vscale x 8 x i16> %zdn3, <vscale x 8 x i16> %zdn4,
                                                      <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3, <vscale x 8 x i16> %zm4)
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
-define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
-@multi_vec_sat_double_mulh_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
-                                        <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @multi_vec_sat_double_mulh_multi_x4_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
 ; CHECK-LABEL: multi_vec_sat_double_mulh_multi_x4_s32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1w { z31.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    sqdmulh { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                                        <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4) {
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> }
               @llvm.aarch64.sve.sqdmulh.vgx4.nxv4i32(<vscale x 4 x i32> %zdn1, <vscale x 4 x i32> %zdn2, <vscale x 4 x i32> %zdn3, <vscale x 4 x i32> %zdn4,
                                                      <vscale x 4 x i32> %zm1, <vscale x 4 x i32> %zm2, <vscale x 4 x i32> %zm3, <vscale x 4 x i32> %zm4)
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
-define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
-@multi_vec_sat_double_mulh_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
-                                        <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @multi_vec_sat_double_mulh_multi_x4_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
 ; CHECK-LABEL: multi_vec_sat_double_mulh_multi_x4_s64:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov z30.d, z7.d
+; CHECK-NEXT:    mov z27.d, z4.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov	z30.d, z7.d
-; CHECK-NEXT:    mov	z27.d, z4.d
-; CHECK-NEXT:    mov	z29.d, z6.d
-; CHECK-NEXT:    mov	z26.d, z3.d
-; CHECK-NEXT:    mov	z28.d, z5.d
-; CHECK-NEXT:    mov	z25.d, z2.d
-; CHECK-NEXT:    mov	z24.d, z1.d
+; CHECK-NEXT:    mov z29.d, z6.d
+; CHECK-NEXT:    mov z26.d, z3.d
+; CHECK-NEXT:    mov z28.d, z5.d
+; CHECK-NEXT:    mov z25.d, z2.d
 ; CHECK-NEXT:    ld1d { z31.d }, p0/z, [x0]
+; CHECK-NEXT:    mov z24.d, z1.d
 ; CHECK-NEXT:    sqdmulh { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
 ; CHECK-NEXT:    mov z0.d, z24.d
 ; CHECK-NEXT:    mov z1.d, z25.d
 ; CHECK-NEXT:    mov z2.d, z26.d
 ; CHECK-NEXT:    mov z3.d, z27.d
 ; CHECK-NEXT:    ret
+                                        <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4) {
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> }
               @llvm.aarch64.sve.sqdmulh.vgx4.nxv2i64(<vscale x 2 x i64> %zdn1, <vscale x 2 x i64> %zdn2, <vscale x 2 x i64> %zdn3, <vscale x 2 x i64> %zdn4,
                                                      <vscale x 2 x i64> %zm1, <vscale x 2 x i64> %zm2, <vscale x 2 x i64> %zm3, <vscale x 2 x i64> %zm4)
diff --git a/llvm/test/CodeGen/AArch64/split-vector-insert.ll b/llvm/test/CodeGen/AArch64/split-vector-insert.ll
index a507296338f93..b8ab9a00c6981 100644
--- a/llvm/test/CodeGen/AArch64/split-vector-insert.ll
+++ b/llvm/test/CodeGen/AArch64/split-vector-insert.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc < %s -debug-only=legalize-types 2>&1 | FileCheck %s --check-prefix=CHECK-LEGALIZATION
 ; RUN: llc < %s | FileCheck %s
 ; REQUIRES: asserts
@@ -9,54 +10,94 @@ declare <vscale x 2 x i64> @llvm.vector.insert.nxv2i64.v8i64(<vscale x 2 x i64>,
 declare <vscale x 2 x double> @llvm.vector.insert.nxv2f64.v8f64(<vscale x 2 x double>, <8 x double>, i64)
 
 define <vscale x 2 x i64> @test_nxv2i64_v8i64(<vscale x 2 x i64> %a, <8 x i64> %b) #0 {
-; CHECK-LEGALIZATION: Legally typed node: [[T1:t[0-9]+]]: nxv2i64 = insert_subvector {{t[0-9]+}}, {{t[0-9]+}}, Constant:i64<0>
-; CHECK-LEGALIZATION: Legally typed node: [[T2:t[0-9]+]]: nxv2i64 = insert_subvector [[T1]], {{t[0-9]+}}, Constant:i64<2>
-; CHECK-LEGALIZATION: Legally typed node: [[T3:t[0-9]+]]: nxv2i64 = insert_subvector [[T2]], {{t[0-9]+}}, Constant:i64<4>
-; CHECK-LEGALIZATION: Legally typed node: [[T4:t[0-9]+]]: nxv2i64 = insert_subvector [[T3]], {{t[0-9]+}}, Constant:i64<6>
-
+; CHECK-LEGALIZATION-LABEL: test_nxv2i64_v8i64:
+; CHECK-LEGALIZATION:       // %bb.0:
+; CHECK-LEGALIZATION-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-LEGALIZATION-NEXT:    .cfi_offset w29, -16
+; CHECK-LEGALIZATION-NEXT:    addvl sp, sp, #-3
+; CHECK-LEGALIZATION-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
+; CHECK-LEGALIZATION-NEXT:    cntd x8
+; CHECK-LEGALIZATION-NEXT:    ptrue p0.d, vl2
+; CHECK-LEGALIZATION-NEXT:    mov w9, #2 // =0x2
+; CHECK-LEGALIZATION-NEXT:    sub x8, x8, #2
+; CHECK-LEGALIZATION-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-LEGALIZATION-NEXT:    mov x10, sp
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #2
+; CHECK-LEGALIZATION-NEXT:    mov z0.d, p0/m, z1.d
+; CHECK-LEGALIZATION-NEXT:    ptrue p0.d
+; CHECK-LEGALIZATION-NEXT:    csel x9, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #4
+; CHECK-LEGALIZATION-NEXT:    lsl x9, x9, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-LEGALIZATION-NEXT:    str q2, [x10, x9]
+; CHECK-LEGALIZATION-NEXT:    mov w9, #4 // =0x4
+; CHECK-LEGALIZATION-NEXT:    addvl x10, sp, #1
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp]
+; CHECK-LEGALIZATION-NEXT:    csel x9, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #6
+; CHECK-LEGALIZATION-NEXT:    lsl x9, x9, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
+; CHECK-LEGALIZATION-NEXT:    str q3, [x10, x9]
+; CHECK-LEGALIZATION-NEXT:    mov w9, #6 // =0x6
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
+; CHECK-LEGALIZATION-NEXT:    csel x8, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    addvl x9, sp, #2
+; CHECK-LEGALIZATION-NEXT:    lsl x8, x8, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
+; CHECK-LEGALIZATION-NEXT:    str q4, [x9, x8]
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
+; CHECK-LEGALIZATION-NEXT:    addvl sp, sp, #3
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa wsp, 16
+; CHECK-LEGALIZATION-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-LEGALIZATION-NEXT:    .cfi_restore w29
+; CHECK-LEGALIZATION-NEXT:    ret
+;
 ; CHECK-LABEL: test_nxv2i64_v8i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT: str	x29, [sp, #-16]!                // 8-byte Folded Spill
-; CHECK-NEXT: .cfi_def_cfa_offset 16
-; CHECK-NEXT: .cfi_offset w29, -16
-; CHECK-NEXT: addvl	sp, sp, #-3
-; CHECK-NEXT: .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
-; CHECK-NEXT: ptrue	p1.d, vl2
-; CHECK-NEXT: cntd	x8
-; CHECK-NEXT: mov	w9, #2                          // =0x2
-; CHECK-NEXT: ptrue	p0.d
-; CHECK-NEXT: sub	x8, x8, #2
-; CHECK-NEXT:                                       // kill: def $q1 killed $q1 def $z1
-; CHECK-NEXT: mov	x10, sp
-; CHECK-NEXT: cmp	x8, #2
-; CHECK-NEXT: csel	x9, x8, x9, lo
-; CHECK-NEXT: cmp	x8, #4
-; CHECK-NEXT: lsl	x9, x9, #3
-; CHECK-NEXT: mov	z0.d, p1/m, z1.d
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp]
-; CHECK-NEXT: str	q2, [x10, x9]
-; CHECK-NEXT: mov	w9, #4                          // =0x4
-; CHECK-NEXT: addvl	x10, sp, #1
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp]
-; CHECK-NEXT: csel	x9, x8, x9, lo
-; CHECK-NEXT: cmp	x8, #6
-; CHECK-NEXT: lsl	x9, x9, #3
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp, #1, mul vl]
-; CHECK-NEXT: str	q3, [x10, x9]
-; CHECK-NEXT: mov	w9, #6                          // =0x6
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp, #1, mul vl]
-; CHECK-NEXT: csel	x8, x8, x9, lo
-; CHECK-NEXT: addvl	x9, sp, #2
-; CHECK-NEXT: lsl	x8, x8, #3
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp, #2, mul vl]
-; CHECK-NEXT: str	q4, [x9, x8]
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp, #2, mul vl]
-; CHECK-NEXT: addvl	sp, sp, #3
-; CHECK-NEXT: .cfi_def_cfa wsp, 16
-; CHECK-NEXT: ldr	x29, [sp], #16                  // 8-byte Folded Reload
-; CHECK-NEXT: .cfi_def_cfa_offset 0
-; CHECK-NEXT: .cfi_restore w29
-; CHECK-NEXT: ret
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    addvl sp, sp, #-3
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
+; CHECK-NEXT:    cntd x8
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    mov w9, #2 // =0x2
+; CHECK-NEXT:    sub x8, x8, #2
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    mov x10, sp
+; CHECK-NEXT:    cmp x8, #2
+; CHECK-NEXT:    mov z0.d, p0/m, z1.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    cmp x8, #4
+; CHECK-NEXT:    lsl x9, x9, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-NEXT:    str q2, [x10, x9]
+; CHECK-NEXT:    mov w9, #4 // =0x4
+; CHECK-NEXT:    addvl x10, sp, #1
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    cmp x8, #6
+; CHECK-NEXT:    lsl x9, x9, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    str q3, [x10, x9]
+; CHECK-NEXT:    mov w9, #6 // =0x6
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    addvl x9, sp, #2
+; CHECK-NEXT:    lsl x8, x8, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    str q4, [x9, x8]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
+; CHECK-NEXT:    addvl sp, sp, #3
+; CHECK-NEXT:    .cfi_def_cfa wsp, 16
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-NEXT:    .cfi_restore w29
+; CHECK-NEXT:    ret
+
 
 
 
@@ -66,54 +107,94 @@ define <vscale x 2 x i64> @test_nxv2i64_v8i64(<vscale x 2 x i64> %a, <8 x i64> %
 }
 
 define <vscale x 2 x double> @test_nxv2f64_v8f64(<vscale x 2 x double> %a, <8 x double> %b) #0 {
-; CHECK-LEGALIZATION: Legally typed node: [[T1:t[0-9]+]]: nxv2f64 = insert_subvector {{t[0-9]+}}, {{t[0-9]+}}, Constant:i64<0>
-; CHECK-LEGALIZATION: Legally typed node: [[T2:t[0-9]+]]: nxv2f64 = insert_subvector [[T1]], {{t[0-9]+}}, Constant:i64<2>
-; CHECK-LEGALIZATION: Legally typed node: [[T3:t[0-9]+]]: nxv2f64 = insert_subvector [[T2]], {{t[0-9]+}}, Constant:i64<4>
-; CHECK-LEGALIZATION: Legally typed node: [[T4:t[0-9]+]]: nxv2f64 = insert_subvector [[T3]], {{t[0-9]+}}, Constant:i64<6>
-
+; CHECK-LEGALIZATION-LABEL: test_nxv2f64_v8f64:
+; CHECK-LEGALIZATION:       // %bb.0:
+; CHECK-LEGALIZATION-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-LEGALIZATION-NEXT:    .cfi_offset w29, -16
+; CHECK-LEGALIZATION-NEXT:    addvl sp, sp, #-3
+; CHECK-LEGALIZATION-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
+; CHECK-LEGALIZATION-NEXT:    cntd x8
+; CHECK-LEGALIZATION-NEXT:    ptrue p0.d, vl2
+; CHECK-LEGALIZATION-NEXT:    mov w9, #2 // =0x2
+; CHECK-LEGALIZATION-NEXT:    sub x8, x8, #2
+; CHECK-LEGALIZATION-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-LEGALIZATION-NEXT:    mov x10, sp
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #2
+; CHECK-LEGALIZATION-NEXT:    mov z0.d, p0/m, z1.d
+; CHECK-LEGALIZATION-NEXT:    ptrue p0.d
+; CHECK-LEGALIZATION-NEXT:    csel x9, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #4
+; CHECK-LEGALIZATION-NEXT:    lsl x9, x9, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-LEGALIZATION-NEXT:    str q2, [x10, x9]
+; CHECK-LEGALIZATION-NEXT:    mov w9, #4 // =0x4
+; CHECK-LEGALIZATION-NEXT:    addvl x10, sp, #1
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp]
+; CHECK-LEGALIZATION-NEXT:    csel x9, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    cmp x8, #6
+; CHECK-LEGALIZATION-NEXT:    lsl x9, x9, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
+; CHECK-LEGALIZATION-NEXT:    str q3, [x10, x9]
+; CHECK-LEGALIZATION-NEXT:    mov w9, #6 // =0x6
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
+; CHECK-LEGALIZATION-NEXT:    csel x8, x8, x9, lo
+; CHECK-LEGALIZATION-NEXT:    addvl x9, sp, #2
+; CHECK-LEGALIZATION-NEXT:    lsl x8, x8, #3
+; CHECK-LEGALIZATION-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
+; CHECK-LEGALIZATION-NEXT:    str q4, [x9, x8]
+; CHECK-LEGALIZATION-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
+; CHECK-LEGALIZATION-NEXT:    addvl sp, sp, #3
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa wsp, 16
+; CHECK-LEGALIZATION-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-LEGALIZATION-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-LEGALIZATION-NEXT:    .cfi_restore w29
+; CHECK-LEGALIZATION-NEXT:    ret
+;
 ; CHECK-LABEL: test_nxv2f64_v8f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT: str	x29, [sp, #-16]!                // 8-byte Folded Spill
-; CHECK-NEXT: .cfi_def_cfa_offset 16
-; CHECK-NEXT: .cfi_offset w29, -16
-; CHECK-NEXT: addvl	sp, sp, #-3
-; CHECK-NEXT: .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
-; CHECK-NEXT: ptrue	p1.d, vl2
-; CHECK-NEXT: cntd	x8
-; CHECK-NEXT: mov	w9, #2                          // =0x2
-; CHECK-NEXT: ptrue	p0.d
-; CHECK-NEXT: sub	x8, x8, #2
-; CHECK-NEXT:                                       // kill: def $q1 killed $q1 def $z1
-; CHECK-NEXT: mov	x10, sp
-; CHECK-NEXT: cmp	x8, #2
-; CHECK-NEXT: csel	x9, x8, x9, lo
-; CHECK-NEXT: cmp	x8, #4
-; CHECK-NEXT: lsl	x9, x9, #3
-; CHECK-NEXT: mov	z0.d, p1/m, z1.d
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp]
-; CHECK-NEXT: str	q2, [x10, x9]
-; CHECK-NEXT: mov	w9, #4                          // =0x4
-; CHECK-NEXT: addvl	x10, sp, #1
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp]
-; CHECK-NEXT: csel	x9, x8, x9, lo
-; CHECK-NEXT: cmp	x8, #6
-; CHECK-NEXT: lsl	x9, x9, #3
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp, #1, mul vl]
-; CHECK-NEXT: str	q3, [x10, x9]
-; CHECK-NEXT: mov	w9, #6                          // =0x6
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp, #1, mul vl]
-; CHECK-NEXT: csel	x8, x8, x9, lo
-; CHECK-NEXT: addvl	x9, sp, #2
-; CHECK-NEXT: lsl	x8, x8, #3
-; CHECK-NEXT: st1d	{ z0.d }, p0, [sp, #2, mul vl]
-; CHECK-NEXT: str	q4, [x9, x8]
-; CHECK-NEXT: ld1d	{ z0.d }, p0/z, [sp, #2, mul vl]
-; CHECK-NEXT: addvl	sp, sp, #3
-; CHECK-NEXT: .cfi_def_cfa wsp, 16
-; CHECK-NEXT: ldr	x29, [sp], #16                  // 8-byte Folded Reload
-; CHECK-NEXT: .cfi_def_cfa_offset 0
-; CHECK-NEXT: .cfi_restore w29
-; CHECK-NEXT: ret
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    addvl sp, sp, #-3
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
+; CHECK-NEXT:    cntd x8
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    mov w9, #2 // =0x2
+; CHECK-NEXT:    sub x8, x8, #2
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    mov x10, sp
+; CHECK-NEXT:    cmp x8, #2
+; CHECK-NEXT:    mov z0.d, p0/m, z1.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    cmp x8, #4
+; CHECK-NEXT:    lsl x9, x9, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-NEXT:    str q2, [x10, x9]
+; CHECK-NEXT:    mov w9, #4 // =0x4
+; CHECK-NEXT:    addvl x10, sp, #1
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
+; CHECK-NEXT:    csel x9, x8, x9, lo
+; CHECK-NEXT:    cmp x8, #6
+; CHECK-NEXT:    lsl x9, x9, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    str q3, [x10, x9]
+; CHECK-NEXT:    mov w9, #6 // =0x6
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    addvl x9, sp, #2
+; CHECK-NEXT:    lsl x8, x8, #3
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    str q4, [x9, x8]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
+; CHECK-NEXT:    addvl sp, sp, #3
+; CHECK-NEXT:    .cfi_def_cfa wsp, 16
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    .cfi_def_cfa_offset 0
+; CHECK-NEXT:    .cfi_restore w29
+; CHECK-NEXT:    ret
+
 
 
 
diff --git a/llvm/test/CodeGen/AArch64/srem-seteq-vec-splat.ll b/llvm/test/CodeGen/AArch64/srem-seteq-vec-splat.ll
index 1d9cb88260b60..c0c0ae5c9d1fe 100644
--- a/llvm/test/CodeGen/AArch64/srem-seteq-vec-splat.ll
+++ b/llvm/test/CodeGen/AArch64/srem-seteq-vec-splat.ll
@@ -111,6 +111,7 @@ define <4 x i32> @test_srem_odd_undef1(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_srem_odd_undef1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #34079 // =0x851f
+; CHECK-NEXT:    movi v3.4s, #25
 ; CHECK-NEXT:    movk w8, #20971, lsl #16
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    smull2 v2.2d, v0.4s, v1.4s
@@ -118,9 +119,8 @@ define <4 x i32> @test_srem_odd_undef1(<4 x i32> %X) nounwind {
 ; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v2.4s
 ; CHECK-NEXT:    sshr v2.4s, v1.4s, #3
 ; CHECK-NEXT:    usra v2.4s, v1.4s, #31
-; CHECK-NEXT:    movi v1.4s, #25
-; CHECK-NEXT:    mls v0.4s, v2.4s, v1.4s
 ; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    mls v0.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -134,6 +134,7 @@ define <4 x i32> @test_srem_even_undef1(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_srem_even_undef1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #34079 // =0x851f
+; CHECK-NEXT:    movi v3.4s, #100
 ; CHECK-NEXT:    movk w8, #20971, lsl #16
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    smull2 v2.2d, v0.4s, v1.4s
@@ -141,9 +142,8 @@ define <4 x i32> @test_srem_even_undef1(<4 x i32> %X) nounwind {
 ; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v2.4s
 ; CHECK-NEXT:    sshr v2.4s, v1.4s, #5
 ; CHECK-NEXT:    usra v2.4s, v1.4s, #31
-; CHECK-NEXT:    movi v1.4s, #100
-; CHECK-NEXT:    mls v0.4s, v2.4s, v1.4s
 ; CHECK-NEXT:    movi v1.4s, #1
+; CHECK-NEXT:    mls v0.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret
@@ -201,11 +201,11 @@ define <4 x i32> @test_srem_pow2(<4 x i32> %X) nounwind {
 define <4 x i32> @test_srem_int_min(<4 x i32> %X) nounwind {
 ; CHECK-LABEL: test_srem_int_min:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    cmlt v1.4s, v0.4s, #0
-; CHECK-NEXT:    mov v2.16b, v0.16b
-; CHECK-NEXT:    usra v2.4s, v1.4s, #1
+; CHECK-NEXT:    cmlt v2.4s, v0.4s, #0
+; CHECK-NEXT:    mov v3.16b, v0.16b
 ; CHECK-NEXT:    movi v1.4s, #128, lsl #24
-; CHECK-NEXT:    and v1.16b, v2.16b, v1.16b
+; CHECK-NEXT:    usra v3.4s, v2.4s, #1
+; CHECK-NEXT:    and v1.16b, v3.16b, v1.16b
 ; CHECK-NEXT:    add v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    movi v1.4s, #1
 ; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
diff --git a/llvm/test/CodeGen/AArch64/srem-vector-lkk.ll b/llvm/test/CodeGen/AArch64/srem-vector-lkk.ll
index 0598af7c98063..a74f0c86fe185 100644
--- a/llvm/test/CodeGen/AArch64/srem-vector-lkk.ll
+++ b/llvm/test/CodeGen/AArch64/srem-vector-lkk.ll
@@ -245,6 +245,7 @@ define <4 x i32> @fold_srem_v4i32(<4 x i32> %x) {
 ; CHECK-LABEL: fold_srem_v4i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #26215 // =0x6667
+; CHECK-NEXT:    movi v3.4s, #10
 ; CHECK-NEXT:    movk w8, #26214, lsl #16
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    smull2 v2.2d, v0.4s, v1.4s
@@ -252,8 +253,7 @@ define <4 x i32> @fold_srem_v4i32(<4 x i32> %x) {
 ; CHECK-NEXT:    uzp2 v1.4s, v1.4s, v2.4s
 ; CHECK-NEXT:    sshr v2.4s, v1.4s, #2
 ; CHECK-NEXT:    usra v2.4s, v1.4s, #31
-; CHECK-NEXT:    movi v1.4s, #10
-; CHECK-NEXT:    mls v0.4s, v2.4s, v1.4s
+; CHECK-NEXT:    mls v0.4s, v2.4s, v3.4s
 ; CHECK-NEXT:    ret
   %1 = srem <4 x i32> %x, <i32 10, i32 10, i32 10, i32 10>
   ret <4 x i32> %1
diff --git a/llvm/test/CodeGen/AArch64/sve-abd.ll b/llvm/test/CodeGen/AArch64/sve-abd.ll
index 95aec0a492619..7b492229e3d23 100644
--- a/llvm/test/CodeGen/AArch64/sve-abd.ll
+++ b/llvm/test/CodeGen/AArch64/sve-abd.ll
@@ -24,10 +24,10 @@ define <vscale x 16 x i8> @sabd_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 16 x i8> @sabd_b_promoted_ops(<vscale x 16 x i1> %a, <vscale x 16 x i1> %b) #0 {
 ; CHECK-LABEL: sabd_b_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    mov z0.b, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z1.b, p1/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    sabd z0.b, p2/m, z0.b, z1.b
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    sabd z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
   %a.sext = sext <vscale x 16 x i1> %a to <vscale x 16 x i8>
   %b.sext = sext <vscale x 16 x i1> %b to <vscale x 16 x i8>
@@ -144,10 +144,10 @@ define <vscale x 16 x i8> @uabd_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 define <vscale x 16 x i8> @uabd_b_promoted_ops(<vscale x 16 x i1> %a, <vscale x 16 x i1> %b) #0 {
 ; CHECK-LABEL: uabd_b_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    mov z0.b, p0/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.b, p1/z, #1 // =0x1
-; CHECK-NEXT:    uabd z0.b, p2/m, z0.b, z1.b
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    uabd z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 16 x i1> %a to <vscale x 16 x i8>
   %b.zext = zext <vscale x 16 x i1> %b to <vscale x 16 x i8>
@@ -173,9 +173,9 @@ define <vscale x 8 x i16> @uabd_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
 define <vscale x 8 x i16> @uabd_h_promoted_ops(<vscale x 8 x i8> %a, <vscale x 8 x i8> %b) #0 {
 ; CHECK-LABEL: uabd_h_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    uabd z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 8 x i8> %a to <vscale x 8 x i16>
@@ -202,9 +202,9 @@ define <vscale x 4 x i32> @uabd_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b)
 define <vscale x 4 x i32> @uabd_s_promoted_ops(<vscale x 4 x i16> %a, <vscale x 4 x i16> %b) #0 {
 ; CHECK-LABEL: uabd_s_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uabd z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 4 x i16> %a to <vscale x 4 x i32>
@@ -231,9 +231,9 @@ define <vscale x 2 x i64> @uabd_d(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
 define <vscale x 2 x i64> @uabd_d_promoted_ops(<vscale x 2 x i32> %a, <vscale x 2 x i32> %b) #0 {
 ; CHECK-LABEL: uabd_d_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
 ; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uabd z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 2 x i32> %a to <vscale x 2 x i64>
@@ -248,8 +248,8 @@ define <vscale x 2 x i64> @uabd_d_promoted_ops(<vscale x 2 x i32> %a, <vscale x
 define <vscale x 4 x i32> @uabd_non_matching_extension(<vscale x 4 x i32> %a, <vscale x 4 x i8> %b) #0 {
 ; CHECK-LABEL: uabd_non_matching_extension:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z1.s, z1.s, #0xff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uabd z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 4 x i32> %a to <vscale x 4 x i64>
@@ -265,9 +265,9 @@ define <vscale x 4 x i32> @uabd_non_matching_extension(<vscale x 4 x i32> %a, <v
 define <vscale x 4 x i32> @uabd_non_matching_promoted_ops(<vscale x 4 x i8> %a, <vscale x 4 x i16> %b) #0 {
 ; CHECK-LABEL: uabd_non_matching_promoted_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.s, z0.s, #0xff
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uabd z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %a.zext = zext <vscale x 4 x i8> %a to <vscale x 4 x i32>
diff --git a/llvm/test/CodeGen/AArch64/sve-bitcast.ll b/llvm/test/CodeGen/AArch64/sve-bitcast.ll
index 7dd568fc837a3..95f43ba512632 100644
--- a/llvm/test/CodeGen/AArch64/sve-bitcast.ll
+++ b/llvm/test/CodeGen/AArch64/sve-bitcast.ll
@@ -1698,8 +1698,8 @@ define <vscale x 1 x i64> @bitcast_nxv8i8_to_nxv1i64(<vscale x 8 x i8> %v) #0 {
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.b
 ; CHECK_BE-NEXT:    uzp1 z0.b, z0.b, z0.b
+; CHECK_BE-NEXT:    ptrue p0.b
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -1720,8 +1720,8 @@ define <vscale x 1 x i64> @bitcast_nxv4i16_to_nxv1i64(<vscale x 4 x i16> %v) #0
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.h
 ; CHECK_BE-NEXT:    uzp1 z0.h, z0.h, z0.h
+; CHECK_BE-NEXT:    ptrue p0.h
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -1742,8 +1742,8 @@ define <vscale x 1 x i64> @bitcast_nxv2i32_to_nxv1i64(<vscale x 2 x i32> %v) #0
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -2218,8 +2218,8 @@ define <vscale x 1 x double> @bitcast_nxv8i8_to_nxv1f64(<vscale x 8 x i8> %v) #0
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.b
 ; CHECK_BE-NEXT:    uzp1 z0.b, z0.b, z0.b
+; CHECK_BE-NEXT:    ptrue p0.b
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -2240,8 +2240,8 @@ define <vscale x 1 x double> @bitcast_nxv4i16_to_nxv1f64(<vscale x 4 x i16> %v)
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.h
 ; CHECK_BE-NEXT:    uzp1 z0.h, z0.h, z0.h
+; CHECK_BE-NEXT:    ptrue p0.h
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -2262,8 +2262,8 @@ define <vscale x 1 x double> @bitcast_nxv2i32_to_nxv1f64(<vscale x 2 x i32> %v)
 ; CHECK_BE:       // %bb.0:
 ; CHECK_BE-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK_BE-NEXT:    addvl sp, sp, #-1
-; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    ptrue p1.d
 ; CHECK_BE-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK_BE-NEXT:    ld1d { z0.d }, p1/z, [sp]
@@ -2827,11 +2827,11 @@ define <vscale x 1 x i32> @bitcast_nxv2f16_to_nxv1i32(<vscale x 2 x half> %v) #0
 ; CHECK_BE-NEXT:    addvl sp, sp, #-2
 ; CHECK_BE-NEXT:    ptrue p0.d
 ; CHECK_BE-NEXT:    ptrue p1.h
-; CHECK_BE-NEXT:    ptrue p2.s
 ; CHECK_BE-NEXT:    st1h { z0.d }, p0, [sp]
+; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    ld1h { z0.h }, p1/z, [sp]
 ; CHECK_BE-NEXT:    st1h { z0.h }, p1, [sp, #1, mul vl]
-; CHECK_BE-NEXT:    ld1w { z0.s }, p2/z, [sp, #1, mul vl]
+; CHECK_BE-NEXT:    ld1w { z0.s }, p0/z, [sp, #1, mul vl]
 ; CHECK_BE-NEXT:    addvl sp, sp, #2
 ; CHECK_BE-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK_BE-NEXT:    ret
@@ -2860,11 +2860,11 @@ define <vscale x 1 x i32> @bitcast_nxv2bf16_to_nxv1i32(<vscale x 2 x bfloat> %v)
 ; CHECK_BE-NEXT:    addvl sp, sp, #-2
 ; CHECK_BE-NEXT:    ptrue p0.d
 ; CHECK_BE-NEXT:    ptrue p1.h
-; CHECK_BE-NEXT:    ptrue p2.s
 ; CHECK_BE-NEXT:    st1h { z0.d }, p0, [sp]
+; CHECK_BE-NEXT:    ptrue p0.s
 ; CHECK_BE-NEXT:    ld1h { z0.h }, p1/z, [sp]
 ; CHECK_BE-NEXT:    st1h { z0.h }, p1, [sp, #1, mul vl]
-; CHECK_BE-NEXT:    ld1w { z0.s }, p2/z, [sp, #1, mul vl]
+; CHECK_BE-NEXT:    ld1w { z0.s }, p0/z, [sp, #1, mul vl]
 ; CHECK_BE-NEXT:    addvl sp, sp, #2
 ; CHECK_BE-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK_BE-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll b/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
index 56b023086ea24..3b7b03e6ef61f 100644
--- a/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
+++ b/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
@@ -64,12 +64,12 @@ define float @foo2(ptr %x0, ptr %x1) nounwind {
 ; CHECK-NEXT:    add x9, sp, #16
 ; CHECK-NEXT:    mov w2, #2 // =0x2
 ; CHECK-NEXT:    mov w3, #3 // =0x3
+; CHECK-NEXT:    ld4d { z1.d - z4.d }, p0/z, [x0]
+; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    mov w4, #4 // =0x4
 ; CHECK-NEXT:    mov w5, #5 // =0x5
 ; CHECK-NEXT:    mov w6, #6 // =0x6
 ; CHECK-NEXT:    mov w7, #7 // =0x7
-; CHECK-NEXT:    ld4d { z1.d - z4.d }, p0/z, [x0]
-; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    ld4d { z16.d - z19.d }, p0/z, [x1]
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w1, #1 // =0x1
@@ -182,8 +182,8 @@ entry:
 define double @foo5(i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, ptr %ptr1, ptr %ptr2, double %x0, <vscale x 8 x double> %x1, <vscale x 8 x double> %x2) nounwind {
 ; CHECK-LABEL: foo5:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr x8, [sp]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ld1d { z5.d }, p0/z, [x8, #1, mul vl]
 ; CHECK-NEXT:    ld1d { z6.d }, p0/z, [x8]
 ; CHECK-NEXT:    ld1d { z7.d }, p0/z, [x8, #3, mul vl]
@@ -229,10 +229,10 @@ entry:
 define void @aavpcs1(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, i32 %s6, <vscale x 4 x i32> %s7, <vscale x 4 x i32> %s8, <vscale x 4 x i32> %s9, <vscale x 4 x i32> %s10, <vscale x 4 x i32> %s11, <vscale x 4 x i32> %s12, <vscale x 4 x i32> %s13, <vscale x 4 x i32> %s14, <vscale x 4 x i32> %s15, <vscale x 4 x i32> %s16, ptr %ptr) nounwind {
 ; CHECK-LABEL: aavpcs1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x8]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ld1w { z24.s }, p0/z, [x7]
+; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x8]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x9]
 ; CHECK-NEXT:    st1w { z1.s }, p0, [x9]
 ; CHECK-NEXT:    st1w { z2.s }, p0, [x9]
@@ -261,12 +261,12 @@ entry:
 define void @aavpcs2(float %s0, float %s1, float %s2, float %s3, float %s4, float %s5, float %s6, <vscale x 4 x float> %s7, <vscale x 4 x float> %s8, <vscale x 4 x float> %s9, <vscale x 4 x float> %s10, <vscale x 4 x float> %s11, <vscale x 4 x float> %s12,<vscale x 4 x float> %s13,<vscale x 4 x float> %s14,<vscale x 4 x float> %s15,<vscale x 4 x float> %s16,ptr %ptr) nounwind {
 ; CHECK-LABEL: aavpcs2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp x8, x9, [sp]
-; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x7]
 ; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x6]
+; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8]
 ; CHECK-NEXT:    ld1w { z4.s }, p0/z, [x5]
 ; CHECK-NEXT:    ld1w { z5.s }, p0/z, [x1]
 ; CHECK-NEXT:    ld1w { z6.s }, p0/z, [x4]
@@ -299,11 +299,10 @@ entry:
 define void @aavpcs3(float %s0, float %s1, float %s2, float %s3, float %s4, float %s5, float %s6, float %s7, <vscale x 4 x float> %s8, <vscale x 4 x float> %s9, <vscale x 4 x float> %s10, <vscale x 4 x float> %s11, <vscale x 4 x float> %s12, <vscale x 4 x float> %s13, <vscale x 4 x float> %s14, <vscale x 4 x float> %s15, <vscale x 4 x float> %s16, <vscale x 4 x float> %s17, <vscale x 16 x i1> %p0, ptr %ptr) nounwind {
 ; CHECK-LABEL: aavpcs3:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr x8, [sp]
-; CHECK-NEXT:    ldr x9, [sp, #16]
-; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x8]
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8]
+; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x7]
 ; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x1]
 ; CHECK-NEXT:    ld1w { z4.s }, p0/z, [x6]
@@ -311,15 +310,16 @@ define void @aavpcs3(float %s0, float %s1, float %s2, float %s3, float %s4, floa
 ; CHECK-NEXT:    ld1w { z6.s }, p0/z, [x2]
 ; CHECK-NEXT:    ld1w { z7.s }, p0/z, [x4]
 ; CHECK-NEXT:    ld1w { z24.s }, p0/z, [x3]
-; CHECK-NEXT:    st1w { z0.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z3.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z6.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z24.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z7.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z5.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z4.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z2.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z1.s }, p0, [x9]
+; CHECK-NEXT:    ldr x8, [sp, #16]
+; CHECK-NEXT:    st1w { z1.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z3.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z6.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z24.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z7.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z5.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z4.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z2.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z0.s }, p0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   store volatile <vscale x 4 x float> %s8, ptr %ptr
@@ -339,8 +339,8 @@ entry:
 define void @aavpcs4(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, i32 %s6, i32 %s7, <vscale x 4 x i32> %s8, <vscale x 4 x i32> %s9, <vscale x 4 x i32> %s10, <vscale x 4 x i32> %s11, <vscale x 4 x i32> %s12, <vscale x 4 x i32> %s13, <vscale x 4 x i32> %s14, <vscale x 4 x i32> %s15, <vscale x 4 x i32> %s16, <vscale x 4 x i32> %s17, ptr %ptr) nounwind {
 ; CHECK-LABEL: aavpcs4:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr x8, [sp]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr x9, [sp, #16]
 ; CHECK-NEXT:    ld1w { z24.s }, p0/z, [x8]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x9]
@@ -371,11 +371,10 @@ entry:
 define <vscale x 4 x float> @aavpcs5(float %s0, float %s1, float %s2, float %s3, float %s4, float %s5, float %s6, float %s7, <vscale x 4 x float> %s8, <vscale x 4 x float> %s9, <vscale x 4 x float> %s10, <vscale x 4 x float> %s11, <vscale x 4 x float> %s12, <vscale x 4 x float> %s13, <vscale x 4 x float> %s14, <vscale x 4 x float> %s15, <vscale x 4 x float> %s16, <vscale x 4 x float> %s17, ptr %ptr) nounwind {
 ; CHECK-LABEL: aavpcs5:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr x8, [sp]
-; CHECK-NEXT:    ldr x9, [sp, #16]
-; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x8]
+; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x7]
 ; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x1]
 ; CHECK-NEXT:    ld1w { z4.s }, p0/z, [x6]
@@ -383,15 +382,16 @@ define <vscale x 4 x float> @aavpcs5(float %s0, float %s1, float %s2, float %s3,
 ; CHECK-NEXT:    ld1w { z6.s }, p0/z, [x2]
 ; CHECK-NEXT:    ld1w { z7.s }, p0/z, [x4]
 ; CHECK-NEXT:    ld1w { z24.s }, p0/z, [x3]
-; CHECK-NEXT:    st1w { z0.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z3.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z6.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z24.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z7.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z5.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z4.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z2.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z1.s }, p0, [x9]
+; CHECK-NEXT:    ldr x8, [sp, #16]
+; CHECK-NEXT:    st1w { z0.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z3.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z6.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z24.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z7.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z5.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z4.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z2.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z1.s }, p0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   store volatile <vscale x 4 x float> %s8, ptr %ptr
@@ -409,11 +409,10 @@ entry:
 define void @aapcs1(float %s0, float %s1, float %s2, float %s3, float %s4, float %s5, float %s6, float %s7, <vscale x 4 x float> %s8, <vscale x 4 x float> %s9, <vscale x 4 x float> %s10, <vscale x 4 x float> %s11, <vscale x 4 x float> %s12, <vscale x 4 x float> %s13, <vscale x 4 x float> %s14, <vscale x 4 x float> %s15, <vscale x 4 x float> %s16, <vscale x 4 x float> %s17, ptr %ptr) nounwind {
 ; CHECK-LABEL: aapcs1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldr x8, [sp]
-; CHECK-NEXT:    ldr x9, [sp, #16]
-; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x8]
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8]
+; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x7]
 ; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x1]
 ; CHECK-NEXT:    ld1w { z4.s }, p0/z, [x6]
@@ -421,15 +420,16 @@ define void @aapcs1(float %s0, float %s1, float %s2, float %s3, float %s4, float
 ; CHECK-NEXT:    ld1w { z6.s }, p0/z, [x2]
 ; CHECK-NEXT:    ld1w { z7.s }, p0/z, [x4]
 ; CHECK-NEXT:    ld1w { z16.s }, p0/z, [x3]
-; CHECK-NEXT:    st1w { z0.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z3.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z6.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z16.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z7.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z5.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z4.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z2.s }, p0, [x9]
-; CHECK-NEXT:    st1w { z1.s }, p0, [x9]
+; CHECK-NEXT:    ldr x8, [sp, #16]
+; CHECK-NEXT:    st1w { z1.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z3.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z6.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z16.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z7.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z5.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z4.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z2.s }, p0, [x8]
+; CHECK-NEXT:    st1w { z0.s }, p0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   store volatile <vscale x 4 x float> %s8, ptr %ptr
@@ -486,13 +486,13 @@ define void @non_sve_caller_high_range_non_sve_callee_high_range(float %f0, floa
 ; CHECK-NEXT:    fmov s2, #2.00000000
 ; CHECK-NEXT:    fmov s3, #3.00000000
 ; CHECK-NEXT:    fmov s4, #4.00000000
-; CHECK-NEXT:    fmov s5, #5.00000000
-; CHECK-NEXT:    fmov s6, #6.00000000
-; CHECK-NEXT:    fmov s7, #7.00000000
 ; CHECK-NEXT:    ld1w { z16.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z17.s }, p0/z, [x1]
 ; CHECK-NEXT:    addvl x0, sp, #1
+; CHECK-NEXT:    fmov s5, #5.00000000
+; CHECK-NEXT:    fmov s6, #6.00000000
 ; CHECK-NEXT:    mov x1, sp
+; CHECK-NEXT:    fmov s7, #7.00000000
 ; CHECK-NEXT:    st1w { z17.s }, p0, [sp]
 ; CHECK-NEXT:    st1w { z16.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    bl non_sve_callee_high_range
@@ -548,20 +548,20 @@ define <vscale x 4 x float> @sve_caller_non_sve_callee_high_range(<vscale x 4 x
 ; CHECK-NEXT:    .cfi_escape 0x10, 0x4d, 0x0a, 0x11, 0x70, 0x22, 0x11, 0x50, 0x92, 0x2e, 0x00, 0x1e, 0x22 // $d13 @ cfa - 16 - 48 * VG
 ; CHECK-NEXT:    .cfi_escape 0x10, 0x4e, 0x0a, 0x11, 0x70, 0x22, 0x11, 0x48, 0x92, 0x2e, 0x00, 0x1e, 0x22 // $d14 @ cfa - 16 - 56 * VG
 ; CHECK-NEXT:    .cfi_escape 0x10, 0x4f, 0x0a, 0x11, 0x70, 0x22, 0x11, 0x40, 0x92, 0x2e, 0x00, 0x1e, 0x22 // $d15 @ cfa - 16 - 64 * VG
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z25.d, z0.d
 ; CHECK-NEXT:    str z0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    movi d0, #0000000000000000
 ; CHECK-NEXT:    mov z24.d, z1.d
-; CHECK-NEXT:    addvl x0, sp, #2
 ; CHECK-NEXT:    fmov s1, #1.00000000
+; CHECK-NEXT:    addvl x0, sp, #2
 ; CHECK-NEXT:    fmov s2, #2.00000000
-; CHECK-NEXT:    addvl x1, sp, #1
 ; CHECK-NEXT:    fmov s3, #3.00000000
+; CHECK-NEXT:    addvl x1, sp, #1
 ; CHECK-NEXT:    fmov s4, #4.00000000
 ; CHECK-NEXT:    fmov s5, #5.00000000
 ; CHECK-NEXT:    fmov s6, #6.00000000
 ; CHECK-NEXT:    fmov s7, #7.00000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1w { z24.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z25.s }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    bl non_sve_callee_high_range
@@ -704,20 +704,20 @@ define void @verify_all_operands_are_initialised() {
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x20, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 32 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w30, -8
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movi d0, #0000000000000000
-; CHECK-NEXT:    fmov s1, #1.00000000
 ; CHECK-NEXT:    fmov z16.s, #9.00000000
-; CHECK-NEXT:    mov w8, #1090519040 // =0x41000000
-; CHECK-NEXT:    add x0, sp, #16
+; CHECK-NEXT:    add x8, sp, #16
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    fmov s1, #1.00000000
 ; CHECK-NEXT:    fmov s2, #2.00000000
 ; CHECK-NEXT:    fmov s3, #3.00000000
-; CHECK-NEXT:    add x9, sp, #16
+; CHECK-NEXT:    add x0, sp, #16
 ; CHECK-NEXT:    fmov s4, #4.00000000
 ; CHECK-NEXT:    fmov s5, #5.00000000
+; CHECK-NEXT:    st1w { z16.s }, p0, [x8]
+; CHECK-NEXT:    mov w8, #1090519040 // =0x41000000
 ; CHECK-NEXT:    fmov s6, #6.00000000
 ; CHECK-NEXT:    fmov s7, #7.00000000
-; CHECK-NEXT:    st1w { z16.s }, p0, [x9]
 ; CHECK-NEXT:    str w8, [sp]
 ; CHECK-NEXT:    bl func_f8_and_v0_passed_via_memory
 ; CHECK-NEXT:    addvl sp, sp, #1
diff --git a/llvm/test/CodeGen/AArch64/sve-cmp-folds.ll b/llvm/test/CodeGen/AArch64/sve-cmp-folds.ll
index c8a36e47efca6..1b78c253c58dd 100644
--- a/llvm/test/CodeGen/AArch64/sve-cmp-folds.ll
+++ b/llvm/test/CodeGen/AArch64/sve-cmp-folds.ll
@@ -115,10 +115,10 @@ define i1 @foo_last(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov x8, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    whilels p1.s, xzr, x8
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    lastb w8, p1, z0.s
+; CHECK-NEXT:    whilels p0.s, xzr, x8
+; CHECK-NEXT:    lastb w8, p0, z0.s
 ; CHECK-NEXT:    and w0, w8, #0x1
 ; CHECK-NEXT:    ret
   %vcond = fcmp oeq <vscale x 4 x float> %a, %b
diff --git a/llvm/test/CodeGen/AArch64/sve-doublereduct.ll b/llvm/test/CodeGen/AArch64/sve-doublereduct.ll
index f5721cd0fd793..7bc31d44bb654 100644
--- a/llvm/test/CodeGen/AArch64/sve-doublereduct.ll
+++ b/llvm/test/CodeGen/AArch64/sve-doublereduct.ll
@@ -87,8 +87,8 @@ define float @fmaximum_f32(<vscale x 8 x float> %a, <vscale x 4 x float> %b) {
 define i32 @add_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: add_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add z0.s, z0.s, z2.s
 ; CHECK-NEXT:    uaddv d0, p0, z0.s
 ; CHECK-NEXT:    fmov x0, d0
@@ -160,8 +160,8 @@ define i16 @add_ext_v32i16(<vscale x 32 x i8> %a, <vscale x 16 x i8> %b) {
 define i32 @and_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: and_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.d, z0.d, z2.d
 ; CHECK-NEXT:    andv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -175,8 +175,8 @@ define i32 @and_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 define i32 @or_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: or_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    orr z0.d, z0.d, z2.d
 ; CHECK-NEXT:    orv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -190,8 +190,8 @@ define i32 @or_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 define i32 @xor_i32(<vscale x 8 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: xor_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    eor3 z0.d, z0.d, z1.d, z2.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    eorv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-expand-div.ll b/llvm/test/CodeGen/AArch64/sve-expand-div.ll
index fe5cdc9387728..180c64e0a7de1 100644
--- a/llvm/test/CodeGen/AArch64/sve-expand-div.ll
+++ b/llvm/test/CodeGen/AArch64/sve-expand-div.ll
@@ -10,8 +10,8 @@
 define <vscale x 16 x i8> @sdiv_i8(<vscale x 16 x i8> %a) #0 {
 ; CHECK-LABEL: sdiv_i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z1.b, #86 // =0x56
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    smulh z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    lsr z1.b, z0.b, #7
 ; CHECK-NEXT:    add z0.b, z0.b, z1.b
@@ -23,8 +23,8 @@ define <vscale x 16 x i8> @sdiv_i8(<vscale x 16 x i8> %a) #0 {
 define <vscale x 8 x i16> @sdiv_i16(<vscale x 8 x i16> %a) #0 {
 ; CHECK-LABEL: sdiv_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #21846 // =0x5556
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    smulh z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    lsr z1.h, z0.h, #15
@@ -37,8 +37,8 @@ define <vscale x 8 x i16> @sdiv_i16(<vscale x 8 x i16> %a) #0 {
 define <vscale x 4 x i32> @sdiv_i32(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: sdiv_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #21846 // =0x5556
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movk w8, #21845, lsl #16
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    smulh z0.s, p0/m, z0.s, z1.s
@@ -52,8 +52,8 @@ define <vscale x 4 x i32> @sdiv_i32(<vscale x 4 x i32> %a) #0 {
 define <vscale x 2 x i64> @sdiv_i64(<vscale x 2 x i64> %a) #0 {
 ; CHECK-LABEL: sdiv_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #6148914691236517205 // =0x5555555555555555
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movk x8, #21846
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    smulh z0.d, p0/m, z0.d, z1.d
@@ -71,8 +71,8 @@ define <vscale x 2 x i64> @sdiv_i64(<vscale x 2 x i64> %a) #0 {
 define <vscale x 16 x i8> @udiv_i8(<vscale x 16 x i8> %a) #0 {
 ; CHECK-LABEL: udiv_i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z1.b, #-85 // =0xffffffffffffffab
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    umulh z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    lsr z0.b, z0.b, #1
 ; CHECK-NEXT:    ret
@@ -83,8 +83,8 @@ define <vscale x 16 x i8> @udiv_i8(<vscale x 16 x i8> %a) #0 {
 define <vscale x 8 x i16> @udiv_i16(<vscale x 8 x i16> %a) #0 {
 ; CHECK-LABEL: udiv_i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #-21845 // =0xffffaaab
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    umulh z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    lsr z0.h, z0.h, #1
@@ -96,8 +96,8 @@ define <vscale x 8 x i16> @udiv_i16(<vscale x 8 x i16> %a) #0 {
 define <vscale x 4 x i32> @udiv_i32(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: udiv_i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #43691 // =0xaaab
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movk w8, #43690, lsl #16
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    umulh z0.s, p0/m, z0.s, z1.s
@@ -110,8 +110,8 @@ define <vscale x 4 x i32> @udiv_i32(<vscale x 4 x i32> %a) #0 {
 define <vscale x 2 x i64> @udiv_i64(<vscale x 2 x i64> %a) #0 {
 ; CHECK-LABEL: udiv_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-6148914691236517206 // =0xaaaaaaaaaaaaaaaa
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movk x8, #43691
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    umulh z0.d, p0/m, z0.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-extract-element.ll b/llvm/test/CodeGen/AArch64/sve-extract-element.ll
index a3c34b53baa07..6d4f5963881e5 100644
--- a/llvm/test/CodeGen/AArch64/sve-extract-element.ll
+++ b/llvm/test/CodeGen/AArch64/sve-extract-element.ll
@@ -616,8 +616,8 @@ define i1 @test_last_8xi1(<vscale x 8 x i1> %a) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov x8, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z0.h, p0/z, #1 // =0x1
-; CHECK-NEXT:    whilels p1.h, xzr, x8
-; CHECK-NEXT:    lastb w8, p1, z0.h
+; CHECK-NEXT:    whilels p0.h, xzr, x8
+; CHECK-NEXT:    lastb w8, p0, z0.h
 ; CHECK-NEXT:    and w0, w8, #0x1
 ; CHECK-NEXT:    ret
   %vscale = call i64 @llvm.vscale.i64()
@@ -630,10 +630,10 @@ define i1 @test_last_8xi1(<vscale x 8 x i1> %a) #0 {
 define i1 @test_lanex_4xi1(<vscale x 4 x i1> %a, i32 %x) #0 {
 ; CHECK-LABEL: test_lanex_4xi1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, w0
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    whilels p1.s, xzr, x8
-; CHECK-NEXT:    lastb w8, p1, z0.s
+; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    whilels p0.s, xzr, x8
+; CHECK-NEXT:    lastb w8, p0, z0.s
 ; CHECK-NEXT:    and w0, w8, #0x1
 ; CHECK-NEXT:    ret
   %b = extractelement <vscale x 4 x i1> %a, i32 %x
diff --git a/llvm/test/CodeGen/AArch64/sve-extract-fixed-from-scalable-vector.ll b/llvm/test/CodeGen/AArch64/sve-extract-fixed-from-scalable-vector.ll
index bc1c563810f35..b9c531fe33526 100644
--- a/llvm/test/CodeGen/AArch64/sve-extract-fixed-from-scalable-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-extract-fixed-from-scalable-vector.ll
@@ -100,16 +100,16 @@ define <2 x i64> @extract_v2i64_nxv8i64_8(<vscale x 8 x i64> %arg) {
 ; CHECK-NEXT:    addvl sp, sp, #-4
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 32 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cnth x8
 ; CHECK-NEXT:    mov w9, #8 // =0x8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sub x8, x8, #2
 ; CHECK-NEXT:    cmp x8, #8
+; CHECK-NEXT:    st1d { z3.d }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #3
-; CHECK-NEXT:    st1d { z3.d }, p0, [sp, #3, mul vl]
-; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    st1d { z1.d }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    ldr q0, [x9, x8]
@@ -183,10 +183,10 @@ define <4 x i1> @extract_v4i1_nxv32i1_16(<vscale x 32 x i1> %arg) {
 ; CHECK-NEXT:    addvl sp, sp, #-8
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0d, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0xc0, 0x00, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 64 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    mov z0.b, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.b, p0/z, #1 // =0x1
 ; CHECK-NEXT:    mov x8, sp
+; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    add x8, x8, #16
 ; CHECK-NEXT:    st1b { z0.b }, p2, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1b { z1.b }, p2, [sp]
diff --git a/llvm/test/CodeGen/AArch64/sve-extract-fixed-vector.ll b/llvm/test/CodeGen/AArch64/sve-extract-fixed-vector.ll
index e2f8dad03ef6f..88268104889fd 100644
--- a/llvm/test/CodeGen/AArch64/sve-extract-fixed-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-extract-fixed-vector.ll
@@ -17,15 +17,15 @@ define <2 x i64> @extract_v2i64_nxv2i64_idx2(<vscale x 2 x i64> %vec) nounwind {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntd x8
 ; CHECK-NEXT:    mov w9, #2 // =0x2
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sub x8, x8, #2
 ; CHECK-NEXT:    cmp x8, #2
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #3
-; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    ldr q0, [x9, x8]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -50,15 +50,15 @@ define <4 x i32> @extract_v4i32_nxv4i32_idx4(<vscale x 4 x i32> %vec) nounwind {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cntw x8
 ; CHECK-NEXT:    mov w9, #4 // =0x4
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sub x8, x8, #4
 ; CHECK-NEXT:    cmp x8, #4
+; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #2
-; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    ldr q0, [x9, x8]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -115,15 +115,15 @@ define <8 x i16> @extract_v8i16_nxv8i16_idx8(<vscale x 8 x i16> %vec) nounwind {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cnth x8
 ; CHECK-NEXT:    mov w9, #8 // =0x8
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    sub x8, x8, #8
 ; CHECK-NEXT:    cmp x8, #8
+; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #1
-; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    ldr q0, [x9, x8]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -214,14 +214,14 @@ define <16 x i8> @extract_v16i8_nxv16i8_idx16(<vscale x 16 x i8> %vec) nounwind
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w9, #16 // =0x10
 ; CHECK-NEXT:    sub x8, x8, #16
 ; CHECK-NEXT:    cmp x8, #16
+; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK-NEXT:    ldr q0, [x9, x8]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sve-extract-scalable-vector.ll b/llvm/test/CodeGen/AArch64/sve-extract-scalable-vector.ll
index e60a2f142922f..3c0bd501f45d8 100644
--- a/llvm/test/CodeGen/AArch64/sve-extract-scalable-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-extract-scalable-vector.ll
@@ -65,27 +65,25 @@ define <vscale x 14 x i1> @extract_nxv14i1_nxv28i1_14(<vscale x 28 x i1> %in) uw
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
-; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    punpkhi p0.h, p0.b
-; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    punpklo p2.h, p2.b
 ; CHECK-NEXT:    punpkhi p3.h, p1.b
-; CHECK-NEXT:    punpklo p1.h, p1.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    punpkhi p4.h, p2.b
 ; CHECK-NEXT:    punpklo p2.h, p2.b
+; CHECK-NEXT:    uzp1 p4.s, p4.s, p0.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    punpkhi p5.h, p3.b
-; CHECK-NEXT:    punpklo p3.h, p3.b
-; CHECK-NEXT:    punpkhi p6.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    uzp1 p2.s, p5.s, p2.s
+; CHECK-NEXT:    punpklo p3.h, p3.b
+; CHECK-NEXT:    punpkhi p5.h, p1.b
+; CHECK-NEXT:    punpklo p1.h, p1.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    uzp1 p3.s, p5.s, p3.s
 ; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p3.s, p6.s, p3.s
-; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p4.s, p4.s, p0.s
 ; CHECK-NEXT:    uzp1 p0.s, p0.s, p1.s
 ; CHECK-NEXT:    uzp1 p1.h, p2.h, p4.h
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sve-fcmp.ll b/llvm/test/CodeGen/AArch64/sve-fcmp.ll
index f7e3b6d0171ac..35cbe65c6a8b8 100644
--- a/llvm/test/CodeGen/AArch64/sve-fcmp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fcmp.ll
@@ -374,8 +374,8 @@ define <vscale x 4 x i1> @one_zero(<vscale x 4 x float> %x) {
 define <vscale x 4 x i1> @ueq_zero(<vscale x 4 x float> %x) {
 ; CHECK-LABEL: ueq_zero:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmuo p1.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    sel p0.b, p0, p0.b, p1.b
diff --git a/llvm/test/CodeGen/AArch64/sve-fcopysign.ll b/llvm/test/CodeGen/AArch64/sve-fcopysign.ll
index f15807597ac21..78843e392e536 100644
--- a/llvm/test/CodeGen/AArch64/sve-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fcopysign.ll
@@ -63,10 +63,10 @@ define <vscale x 4 x float> @test_copysign_v4f32_v4f64(<vscale x 4 x float> %a,
 ; CHECK-EXTEND-ROUND-NEXT:    ptrue p0.d
 ; CHECK-EXTEND-ROUND-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-EXTEND-ROUND-NEXT:    uunpklo z0.d, z0.s
-; CHECK-EXTEND-ROUND-NEXT:    and z3.s, z3.s, #0x7fffffff
-; CHECK-EXTEND-ROUND-NEXT:    and z0.s, z0.s, #0x7fffffff
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z2.s, p0/m, z2.d
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z1.s, p0/m, z1.d
+; CHECK-EXTEND-ROUND-NEXT:    and z3.s, z3.s, #0x7fffffff
+; CHECK-EXTEND-ROUND-NEXT:    and z0.s, z0.s, #0x7fffffff
 ; CHECK-EXTEND-ROUND-NEXT:    and z2.s, z2.s, #0x80000000
 ; CHECK-EXTEND-ROUND-NEXT:    and z1.s, z1.s, #0x80000000
 ; CHECK-EXTEND-ROUND-NEXT:    orr z2.d, z3.d, z2.d
@@ -115,9 +115,9 @@ declare <vscale x 2 x double> @llvm.copysign.v2f64(<vscale x 2 x double> %a, <vs
 define <vscale x 4 x double> @test_copysign_v4f64_v4f32(<vscale x 4 x double> %a, <vscale x 4 x float> %b) #0 {
 ; CHECK-LABEL: test_copysign_v4f64_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z3.d, z2.s
 ; CHECK-NEXT:    uunpkhi z2.d, z2.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z0.d, z0.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    and z1.d, z1.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    fcvt z3.d, p0/m, z3.s
@@ -193,10 +193,10 @@ define <vscale x 4 x half> @test_copysign_v4f16_v4f64(<vscale x 4 x half> %a, <v
 ; CHECK-EXTEND-ROUND-NEXT:    ptrue p0.d
 ; CHECK-EXTEND-ROUND-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-EXTEND-ROUND-NEXT:    uunpklo z0.d, z0.s
-; CHECK-EXTEND-ROUND-NEXT:    and z3.h, z3.h, #0x7fff
-; CHECK-EXTEND-ROUND-NEXT:    and z0.h, z0.h, #0x7fff
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z2.h, p0/m, z2.d
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z1.h, p0/m, z1.d
+; CHECK-EXTEND-ROUND-NEXT:    and z3.h, z3.h, #0x7fff
+; CHECK-EXTEND-ROUND-NEXT:    and z0.h, z0.h, #0x7fff
 ; CHECK-EXTEND-ROUND-NEXT:    and z2.h, z2.h, #0x8000
 ; CHECK-EXTEND-ROUND-NEXT:    and z1.h, z1.h, #0x8000
 ; CHECK-EXTEND-ROUND-NEXT:    orr z2.d, z3.d, z2.d
@@ -240,10 +240,10 @@ define <vscale x 8 x half> @test_copysign_v8f16_v8f32(<vscale x 8 x half> %a, <v
 ; CHECK-EXTEND-ROUND-NEXT:    ptrue p0.s
 ; CHECK-EXTEND-ROUND-NEXT:    uunpkhi z3.s, z0.h
 ; CHECK-EXTEND-ROUND-NEXT:    uunpklo z0.s, z0.h
-; CHECK-EXTEND-ROUND-NEXT:    and z3.h, z3.h, #0x7fff
-; CHECK-EXTEND-ROUND-NEXT:    and z0.h, z0.h, #0x7fff
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z2.h, p0/m, z2.s
 ; CHECK-EXTEND-ROUND-NEXT:    fcvt z1.h, p0/m, z1.s
+; CHECK-EXTEND-ROUND-NEXT:    and z3.h, z3.h, #0x7fff
+; CHECK-EXTEND-ROUND-NEXT:    and z0.h, z0.h, #0x7fff
 ; CHECK-EXTEND-ROUND-NEXT:    and z2.h, z2.h, #0x8000
 ; CHECK-EXTEND-ROUND-NEXT:    and z1.h, z1.h, #0x8000
 ; CHECK-EXTEND-ROUND-NEXT:    orr z2.d, z3.d, z2.d
@@ -261,9 +261,9 @@ define <vscale x 8 x half> @test_copysign_v8f16_v8f32(<vscale x 8 x half> %a, <v
 define <vscale x 4 x half> @test_copysign_nxv4f32_nxv4f16(<vscale x 4 x float> %a, <vscale x 4 x float> %b) #0 {
 ; CHECK-NO-EXTEND-ROUND-LABEL: test_copysign_nxv4f32_nxv4f16:
 ; CHECK-NO-EXTEND-ROUND:       // %bb.0:
-; CHECK-NO-EXTEND-ROUND-NEXT:    ptrue p0.s
 ; CHECK-NO-EXTEND-ROUND-NEXT:    and z1.s, z1.s, #0x80000000
 ; CHECK-NO-EXTEND-ROUND-NEXT:    and z0.s, z0.s, #0x7fffffff
+; CHECK-NO-EXTEND-ROUND-NEXT:    ptrue p0.s
 ; CHECK-NO-EXTEND-ROUND-NEXT:    orr z0.d, z0.d, z1.d
 ; CHECK-NO-EXTEND-ROUND-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NO-EXTEND-ROUND-NEXT:    ret
@@ -285,9 +285,9 @@ define <vscale x 4 x half> @test_copysign_nxv4f32_nxv4f16(<vscale x 4 x float> %
 define <vscale x 2 x float> @test_copysign_nxv2f64_nxv2f32(<vscale x 2 x double> %a, <vscale x 2 x double> %b) #0 {
 ; CHECK-NO-EXTEND-ROUND-LABEL: test_copysign_nxv2f64_nxv2f32:
 ; CHECK-NO-EXTEND-ROUND:       // %bb.0:
-; CHECK-NO-EXTEND-ROUND-NEXT:    ptrue p0.d
 ; CHECK-NO-EXTEND-ROUND-NEXT:    and z1.d, z1.d, #0x8000000000000000
 ; CHECK-NO-EXTEND-ROUND-NEXT:    and z0.d, z0.d, #0x7fffffffffffffff
+; CHECK-NO-EXTEND-ROUND-NEXT:    ptrue p0.d
 ; CHECK-NO-EXTEND-ROUND-NEXT:    orr z0.d, z0.d, z1.d
 ; CHECK-NO-EXTEND-ROUND-NEXT:    fcvt z0.s, p0/m, z0.d
 ; CHECK-NO-EXTEND-ROUND-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-fcvt.ll b/llvm/test/CodeGen/AArch64/sve-fcvt.ll
index 0fe38bf9ae718..fc5128fffad36 100644
--- a/llvm/test/CodeGen/AArch64/sve-fcvt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fcvt.ll
@@ -454,9 +454,9 @@ define <vscale x 2 x i64> @fcvtzu_d_nxv2f64(<vscale x 2 x double> %a) {
 define <vscale x 2 x half> @scvtf_h_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: scvtf_h_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.h, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    scvtf z0.h, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 2 x i1> %a to <vscale x 2 x half>
   ret <vscale x 2 x half> %res
@@ -495,9 +495,9 @@ define <vscale x 2 x half> @scvtf_h_nxv2i64(<vscale x 2 x i64> %a) {
 define <vscale x 3 x half> @scvtf_h_nxv3i1(<vscale x 3 x i1> %a) {
 ; CHECK-LABEL: scvtf_h_nxv3i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    scvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 3 x i1> %a to <vscale x 3 x half>
   ret <vscale x 3 x half> %res
@@ -516,9 +516,9 @@ define <vscale x 3 x half> @scvtf_h_nxv3i16(<vscale x 3 x i16> %a) {
 define <vscale x 4 x half> @scvtf_h_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK-LABEL: scvtf_h_nxv4i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    scvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 4 x i1> %a to <vscale x 4 x half>
   ret <vscale x 4 x half> %res
@@ -547,9 +547,9 @@ define <vscale x 4 x half> @scvtf_h_nxv4i32(<vscale x 4 x i32> %a) {
 define <vscale x 7 x half> @scvtf_h_nxv7i1(<vscale x 7 x i1> %a) {
 ; CHECK-LABEL: scvtf_h_nxv7i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z0.h, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.h, p1/m, z0.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    scvtf z0.h, p0/m, z0.h
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 7 x i1> %a to <vscale x 7 x half>
   ret <vscale x 7 x half> %res
@@ -568,9 +568,9 @@ define <vscale x 7 x half> @scvtf_h_nxv7i16(<vscale x 7 x i16> %a) {
 define <vscale x 8 x half> @scvtf_h_nxv8i1(<vscale x 8 x i1> %a) {
 ; CHECK-LABEL: scvtf_h_nxv8i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z0.h, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.h, p1/m, z0.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    scvtf z0.h, p0/m, z0.h
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 8 x i1> %a to <vscale x 8 x half>
   ret <vscale x 8 x half> %res
@@ -589,9 +589,9 @@ define <vscale x 8 x half> @scvtf_h_nxv8i16(<vscale x 8 x i16> %a) {
 define <vscale x 2 x float> @scvtf_s_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: scvtf_s_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.s, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    scvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 2 x i1> %a to <vscale x 2 x float>
   ret <vscale x 2 x float> %res
@@ -620,9 +620,9 @@ define <vscale x 2 x float> @scvtf_s_nxv2i64(<vscale x 2 x i64> %a) {
 define <vscale x 3 x float> @scvtf_s_nxv3i1(<vscale x 3 x i1> %a) {
 ; CHECK-LABEL: scvtf_s_nxv3i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.s, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    scvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 3 x i1> %a to <vscale x 3 x float>
   ret <vscale x 3 x float> %res
@@ -641,9 +641,9 @@ define <vscale x 3 x float> @scvtf_s_nxv3i32(<vscale x 3 x i32> %a) {
 define <vscale x 4 x float> @scvtf_s_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK-LABEL: scvtf_s_nxv4i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.s, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    scvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 4 x i1> %a to <vscale x 4 x float>
   ret <vscale x 4 x float> %res
@@ -662,9 +662,9 @@ define <vscale x 4 x float> @scvtf_s_nxv4i32(<vscale x 4 x i32> %a) {
 define <vscale x 2 x double> @scvtf_d_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: scvtf_d_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    scvtf z0.d, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = sitofp <vscale x 2 x i1> %a to <vscale x 2 x double>
   ret <vscale x 2 x double> %res
@@ -695,9 +695,9 @@ define <vscale x 2 x double> @scvtf_d_nxv2i64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x half> @ucvtf_h_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: ucvtf_h_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.h, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 2 x i1> %a to <vscale x 2 x half>
   ret <vscale x 2 x half> %res
@@ -736,9 +736,9 @@ define <vscale x 2 x half> @ucvtf_h_nxv2i64(<vscale x 2 x i64> %a) {
 define <vscale x 3 x half> @ucvtf_h_nxv3i1(<vscale x 3 x i1> %a) {
 ; CHECK-LABEL: ucvtf_h_nxv3i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 3 x i1> %a to <vscale x 3 x half>
   ret <vscale x 3 x half> %res
@@ -767,9 +767,9 @@ define <vscale x 3 x half> @ucvtf_h_nxv3i32(<vscale x 3 x i32> %a) {
 define <vscale x 4 x half> @ucvtf_h_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK-LABEL: ucvtf_h_nxv4i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 4 x i1> %a to <vscale x 4 x half>
   ret <vscale x 4 x half> %res
@@ -798,9 +798,9 @@ define <vscale x 4 x half> @ucvtf_h_nxv4i32(<vscale x 4 x i32> %a) {
 define <vscale x 8 x half> @ucvtf_h_nxv8i1(<vscale x 8 x i1> %a) {
 ; CHECK-LABEL: ucvtf_h_nxv8i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z0.h, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.h, p1/m, z0.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.h
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 8 x i1> %a to <vscale x 8 x half>
   ret <vscale x 8 x half> %res
@@ -819,9 +819,9 @@ define <vscale x 8 x half> @ucvtf_h_nxv8i16(<vscale x 8 x i16> %a) {
 define <vscale x 2 x float> @ucvtf_s_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: ucvtf_s_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.s, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 2 x i1> %a to <vscale x 2 x float>
   ret <vscale x 2 x float> %res
@@ -850,9 +850,9 @@ define <vscale x 2 x float> @ucvtf_s_nxv2i64(<vscale x 2 x i64> %a) {
 define <vscale x 4 x float> @ucvtf_s_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK-LABEL: ucvtf_s_nxv4i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.s, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 4 x i1> %a to <vscale x 4 x float>
   ret <vscale x 4 x float> %res
@@ -871,9 +871,9 @@ define <vscale x 4 x float> @ucvtf_s_nxv4i32(<vscale x 4 x i32> %a) {
 define <vscale x 2 x double> @ucvtf_d_nxv2i1(<vscale x 2 x i1> %a) {
 ; CHECK-LABEL: ucvtf_d_nxv2i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p0/z, #1 // =0x1
-; CHECK-NEXT:    ucvtf z0.d, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %res = uitofp <vscale x 2 x i1> %a to <vscale x 2 x double>
   ret <vscale x 2 x double> %res
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-addressing-modes.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-addressing-modes.ll
index ed7ea657874a4..28e1412c524a0 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-addressing-modes.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-addressing-modes.ll
@@ -7,8 +7,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @masked_gather_base_plus_stride_v8f32(ptr %dst, ptr %src) #0 {
 ; CHECK-LABEL: masked_gather_base_plus_stride_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    index z0.s, #0, #7
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x1, z0.s, sxtw #2]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -21,8 +21,8 @@ define void @masked_gather_base_plus_stride_v8f32(ptr %dst, ptr %src) #0 {
 define void @masked_gather_base_plus_stride_v4f64(ptr %dst, ptr %src) #0 {
 ; CHECK-LABEL: masked_gather_base_plus_stride_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mov x8, #-32 // =0xffffffffffffffe0
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    index z0.d, #-2, x8
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1, z0.d, lsl #3]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
index ad482118ec0bb..47fda39d84001 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-build-vector.ll
@@ -7,8 +7,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @build_vector_7_inc1_v32i8(ptr %a) #0 {
 ; VBITS_GE_256-LABEL: build_vector_7_inc1_v32i8:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
 ; VBITS_GE_256-NEXT:    index z0.b, #7, #1
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
   store <32 x i8> <i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31, i8 32, i8 33, i8 34, i8 35, i8 36, i8 37, i8 38>, ptr %a, align 1
@@ -18,8 +18,8 @@ define void @build_vector_7_inc1_v32i8(ptr %a) #0 {
 define void @build_vector_0_inc2_v16i16(ptr %a) #0 {
 ; VBITS_GE_256-LABEL: build_vector_0_inc2_v16i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    index z0.h, #0, #2
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
   store <16 x i16> <i16 0, i16 2, i16 4, i16 6, i16 8, i16 10, i16 12, i16 14, i16 16, i16 18, i16 20, i16 22, i16 24, i16 26, i16 28, i16 30>, ptr %a, align 2
@@ -30,8 +30,8 @@ define void @build_vector_0_inc2_v16i16(ptr %a) #0 {
 define void @build_vector_0_dec3_v8i32(ptr %a) #0 {
 ; VBITS_GE_256-LABEL: build_vector_0_dec3_v8i32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    index z0.s, #0, #-3
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
   store <8 x i32> <i32 0, i32 -3, i32 -6, i32 -9, i32 -12, i32 -15, i32 -18, i32 -21>, ptr %a, align 4
@@ -42,8 +42,8 @@ define void @build_vector_0_dec3_v8i32(ptr %a) #0 {
 define void @build_vector_minus2_dec32_v4i64(ptr %a) #0 {
 ; VBITS_GE_256-LABEL: build_vector_minus2_dec32_v4i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #-32 // =0xffffffffffffffe0
+; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    index z0.d, #-2, x8
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
@@ -53,11 +53,6 @@ define void @build_vector_minus2_dec32_v4i64(ptr %a) #0 {
 
 ; Constant but not a sequence.
 define void @build_vector_no_stride_v4i64(ptr %a) #0 {
-; VBITS_GE_256-LABEL:  .LCPI4_0:
-; VBITS_GE_256:         .xword  0
-; VBITS_GE_256-NEXT:    .xword  4
-; VBITS_GE_256-NEXT:    .xword  1
-; VBITS_GE_256-NEXT:    .xword  8
 ; VBITS_GE_256-LABEL: build_vector_no_stride_v4i64:
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-concat.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-concat.ll
index 65cb448cac117..f7751131005e3 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-concat.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-concat.ll
@@ -38,9 +38,9 @@ define void @concat_v32i8(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.b, vl32
 ; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
-; CHECK-NEXT:    st1b { z0.b }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.b, vl32
+; CHECK-NEXT:    st1b { z0.b }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i8>, ptr %a
   %op2 = load <16 x i8>, ptr %b
@@ -66,11 +66,11 @@ define void @concat_v64i8(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v64i8:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.b, vl32
-; VBITS_GE_512-NEXT:    ptrue p1.b, vl64
 ; VBITS_GE_512-NEXT:    ld1b { z0.b }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1b { z1.b }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.b, p0, z0.b, z1.b
-; VBITS_GE_512-NEXT:    st1b { z0.b }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.b, vl64
+; VBITS_GE_512-NEXT:    st1b { z0.b }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <32 x i8>, ptr %a
   %op2 = load <32 x i8>, ptr %b
@@ -90,11 +90,11 @@ define void @concat_v128i8(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v128i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b, vl64
-; CHECK-NEXT:    ptrue p1.b, vl128
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
 ; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
-; CHECK-NEXT:    st1b { z0.b }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.b, vl128
+; CHECK-NEXT:    st1b { z0.b }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <64 x i8>, ptr %a
   %op2 = load <64 x i8>, ptr %b
@@ -122,11 +122,11 @@ define void @concat_v256i8(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v256i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b, vl128
-; CHECK-NEXT:    ptrue p1.b, vl256
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
 ; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
-; CHECK-NEXT:    st1b { z0.b }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.b, vl256
+; CHECK-NEXT:    st1b { z0.b }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <128 x i8>, ptr %a
   %op2 = load <128 x i8>, ptr %b
@@ -198,9 +198,9 @@ define void @concat_v16i16(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.h, vl16
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl16
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <8 x i16>, ptr %a
   %op2 = load <8 x i16>, ptr %b
@@ -224,11 +224,11 @@ define void @concat_v32i16(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v32i16:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.h, vl16
-; VBITS_GE_512-NEXT:    ptrue p1.h, vl32
 ; VBITS_GE_512-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.h, p0, z0.h, z1.h
-; VBITS_GE_512-NEXT:    st1h { z0.h }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
+; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
   %op2 = load <16 x i16>, ptr %b
@@ -244,11 +244,11 @@ define void @concat_v64i16(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v64i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl32
-; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl64
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i16>, ptr %a
   %op2 = load <32 x i16>, ptr %b
@@ -268,11 +268,11 @@ define void @concat_v128i16(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v128i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl64
-; CHECK-NEXT:    ptrue p1.h, vl128
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl128
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <64 x i16>, ptr %a
   %op2 = load <64 x i16>, ptr %b
@@ -328,9 +328,9 @@ define void @concat_v8i32(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.s, vl8
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl8
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <4 x i32>, ptr %a
   %op2 = load <4 x i32>, ptr %b
@@ -353,11 +353,11 @@ define void @concat_v16i32(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v16i32:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
-; VBITS_GE_512-NEXT:    ptrue p1.s, vl16
 ; VBITS_GE_512-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.s, p0, z0.s, z1.s
-; VBITS_GE_512-NEXT:    st1w { z0.s }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
+; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <8 x i32>, ptr %a
   %op2 = load <8 x i32>, ptr %b
@@ -371,11 +371,11 @@ define void @concat_v32i32(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v32i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl16
-; CHECK-NEXT:    ptrue p1.s, vl32
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl32
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i32>, ptr %a
   %op2 = load <16 x i32>, ptr %b
@@ -391,11 +391,11 @@ define void @concat_v64i32(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v64i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl32
-; CHECK-NEXT:    ptrue p1.s, vl64
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl64
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i32>, ptr %a
   %op2 = load <32 x i32>, ptr %b
@@ -433,9 +433,9 @@ define void @concat_v4i64(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.d, vl4
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl4
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <2 x i64>, ptr %a
   %op2 = load <2 x i64>, ptr %b
@@ -458,11 +458,11 @@ define void @concat_v8i64(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v8i64:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl4
-; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
 ; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.d, p0, z0.d, z1.d
-; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <4 x i64>, ptr %a
   %op2 = load <4 x i64>, ptr %b
@@ -475,11 +475,11 @@ define void @concat_v16i64(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v16i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl8
-; CHECK-NEXT:    ptrue p1.d, vl16
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl16
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <8 x i64>, ptr %a
   %op2 = load <8 x i64>, ptr %b
@@ -493,11 +493,11 @@ define void @concat_v32i64(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v32i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl16
-; CHECK-NEXT:    ptrue p1.d, vl32
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl32
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i64>, ptr %a
   %op2 = load <16 x i64>, ptr %b
@@ -541,9 +541,9 @@ define void @concat_v16f16(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.h, vl16
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl16
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <8 x half>, ptr %a
   %op2 = load <8 x half>, ptr %b
@@ -567,11 +567,11 @@ define void @concat_v32f16(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v32f16:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.h, vl16
-; VBITS_GE_512-NEXT:    ptrue p1.h, vl32
 ; VBITS_GE_512-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.h, p0, z0.h, z1.h
-; VBITS_GE_512-NEXT:    st1h { z0.h }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
+; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <16 x half>, ptr %a
   %op2 = load <16 x half>, ptr %b
@@ -587,11 +587,11 @@ define void @concat_v64f16(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v64f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl32
-; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl64
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x half>, ptr %a
   %op2 = load <32 x half>, ptr %b
@@ -611,11 +611,11 @@ define void @concat_v128f16(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v128f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl64
-; CHECK-NEXT:    ptrue p1.h, vl128
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
-; CHECK-NEXT:    st1h { z0.h }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.h, vl128
+; CHECK-NEXT:    st1h { z0.h }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <64 x half>, ptr %a
   %op2 = load <64 x half>, ptr %b
@@ -671,9 +671,9 @@ define void @concat_v8f32(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.s, vl8
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl8
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <4 x float>, ptr %a
   %op2 = load <4 x float>, ptr %b
@@ -696,11 +696,11 @@ define void @concat_v16f32(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v16f32:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
-; VBITS_GE_512-NEXT:    ptrue p1.s, vl16
 ; VBITS_GE_512-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.s, p0, z0.s, z1.s
-; VBITS_GE_512-NEXT:    st1w { z0.s }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
+; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <8 x float>, ptr %a
   %op2 = load <8 x float>, ptr %b
@@ -714,11 +714,11 @@ define void @concat_v32f32(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v32f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl16
-; CHECK-NEXT:    ptrue p1.s, vl32
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl32
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x float>, ptr %a
   %op2 = load <16 x float>, ptr %b
@@ -734,11 +734,11 @@ define void @concat_v64f32(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v64f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl32
-; CHECK-NEXT:    ptrue p1.s, vl64
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    st1w { z0.s }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.s, vl64
+; CHECK-NEXT:    st1w { z0.s }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x float>, ptr %a
   %op2 = load <32 x float>, ptr %b
@@ -776,9 +776,9 @@ define void @concat_v4f64(ptr %a, ptr %b, ptr %c) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
-; CHECK-NEXT:    ptrue p1.d, vl4
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl4
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <2 x double>, ptr %a
   %op2 = load <2 x double>, ptr %b
@@ -801,11 +801,11 @@ define void @concat_v8f64(ptr %a, ptr %b, ptr %c) #0 {
 ; VBITS_GE_512-LABEL: concat_v8f64:
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl4
-; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
 ; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    splice z0.d, p0, z0.d, z1.d
-; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [x2]
+; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x2]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <4 x double>, ptr %a
   %op2 = load <4 x double>, ptr %b
@@ -818,11 +818,11 @@ define void @concat_v16f64(ptr %a, ptr %b, ptr %c) vscale_range(8,0) #0 {
 ; CHECK-LABEL: concat_v16f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl8
-; CHECK-NEXT:    ptrue p1.d, vl16
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl16
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <8 x double>, ptr %a
   %op2 = load <8 x double>, ptr %b
@@ -836,11 +836,11 @@ define void @concat_v32f64(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #0 {
 ; CHECK-LABEL: concat_v32f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl16
-; CHECK-NEXT:    ptrue p1.d, vl32
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
-; CHECK-NEXT:    st1d { z0.d }, p1, [x2]
+; CHECK-NEXT:    ptrue p0.d, vl32
+; CHECK-NEXT:    st1d { z0.d }, p0, [x2]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x double>, ptr %a
   %op2 = load <16 x double>, ptr %b
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-extract-vector-elt.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-extract-vector-elt.ll
index 485124c1d59ed..ad4efeaf39247 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-extract-vector-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-extract-vector-elt.ll
@@ -70,9 +70,9 @@ define half @extractelement_v64f16(ptr %a) vscale_range(8,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov w8, #63 // =0x3f
-; CHECK-NEXT:    whilels p1.h, xzr, x8
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
-; CHECK-NEXT:    lastb h0, p1, z0.h
+; CHECK-NEXT:    whilels p0.h, xzr, x8
+; CHECK-NEXT:    lastb h0, p0, z0.h
 ; CHECK-NEXT:    ret
     %op1 = load <64 x half>, ptr %a
     %r = extractelement <64 x half> %op1, i64 63
@@ -84,9 +84,9 @@ define half @extractelement_v128f16(ptr %a) vscale_range(16,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov w8, #127 // =0x7f
-; CHECK-NEXT:    whilels p1.h, xzr, x8
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0]
-; CHECK-NEXT:    lastb h0, p1, z0.h
+; CHECK-NEXT:    whilels p0.h, xzr, x8
+; CHECK-NEXT:    lastb h0, p0, z0.h
 ; CHECK-NEXT:    ret
     %op1 = load <128 x half>, ptr %a
     %r = extractelement <128 x half> %op1, i64 127
@@ -154,9 +154,9 @@ define float @extractelement_v32f32(ptr %a) vscale_range(8,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    mov w8, #31 // =0x1f
-; CHECK-NEXT:    whilels p1.s, xzr, x8
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    lastb s0, p1, z0.s
+; CHECK-NEXT:    whilels p0.s, xzr, x8
+; CHECK-NEXT:    lastb s0, p0, z0.s
 ; CHECK-NEXT:    ret
     %op1 = load <32 x float>, ptr %a
     %r = extractelement <32 x float> %op1, i64 31
@@ -168,9 +168,9 @@ define float @extractelement_v64f32(ptr %a) vscale_range(16,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    mov w8, #63 // =0x3f
-; CHECK-NEXT:    whilels p1.s, xzr, x8
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; CHECK-NEXT:    lastb s0, p1, z0.s
+; CHECK-NEXT:    whilels p0.s, xzr, x8
+; CHECK-NEXT:    lastb s0, p0, z0.s
 ; CHECK-NEXT:    ret
     %op1 = load <64 x float>, ptr %a
     %r = extractelement <64 x float> %op1, i64 63
@@ -236,9 +236,9 @@ define double @extractelement_v16f64(ptr %a) vscale_range(8,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    mov w8, #15 // =0xf
-; CHECK-NEXT:    whilels p1.d, xzr, x8
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    lastb d0, p1, z0.d
+; CHECK-NEXT:    whilels p0.d, xzr, x8
+; CHECK-NEXT:    lastb d0, p0, z0.d
 ; CHECK-NEXT:    ret
     %op1 = load <16 x double>, ptr %a
     %r = extractelement <16 x double> %op1, i64 15
@@ -250,9 +250,9 @@ define double @extractelement_v32f64(ptr %a) vscale_range(16,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    mov w8, #31 // =0x1f
-; CHECK-NEXT:    whilels p1.d, xzr, x8
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    lastb d0, p1, z0.d
+; CHECK-NEXT:    whilels p0.d, xzr, x8
+; CHECK-NEXT:    lastb d0, p0, z0.d
 ; CHECK-NEXT:    ret
     %op1 = load <32 x double>, ptr %a
     %r = extractelement <32 x double> %op1, i64 31
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fcopysign.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fcopysign.ll
index bca3dfe5717ef..e77cd9ef55eaf 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fcopysign.ll
@@ -396,9 +396,9 @@ define void @test_copysign_v4f32_v4f64(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mvni v1.4s, #128, lsl #24
 ; CHECK-NEXT:    ldr q2, [x0]
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.s, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    fcvt z0.s, p0/m, z0.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    bit v0.16b, v2.16b, v1.16b
 ; CHECK-NEXT:    str q0, [x0]
@@ -450,12 +450,12 @@ define void @test_copysign_v4f64_v4f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v4f64_v4f32:
 ; CHECK_EXTEND_ROUND:       // %bb.0:
-; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d, vl4
 ; CHECK_EXTEND_ROUND-NEXT:    ldr q0, [x1]
+; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d, vl4
 ; CHECK_EXTEND_ROUND-NEXT:    uunpklo z0.d, z0.s
-; CHECK_EXTEND_ROUND-NEXT:    fcvt z0.d, p0/m, z0.s
 ; CHECK_EXTEND_ROUND-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; CHECK_EXTEND_ROUND-NEXT:    and z1.d, z1.d, #0x7fffffffffffffff
+; CHECK_EXTEND_ROUND-NEXT:    fcvt z0.d, p0/m, z0.s
 ; CHECK_EXTEND_ROUND-NEXT:    and z0.d, z0.d, #0x8000000000000000
 ; CHECK_EXTEND_ROUND-NEXT:    orr z0.d, z1.d, z0.d
 ; CHECK_EXTEND_ROUND-NEXT:    st1d { z0.d }, p0, [x0]
@@ -494,9 +494,9 @@ define void @test_copysign_v4f16_v4f64(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mvni v1.4h, #128, lsl #8
 ; CHECK-NEXT:    ldr d2, [x0]
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    bit v0.8b, v2.8b, v1.8b
@@ -521,9 +521,9 @@ define void @test_copysign_v8f16_v8f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    mvni v1.8h, #128, lsl #8
 ; CHECK-NEXT:    ldr q2, [x0]
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    bit v0.16b, v2.16b, v1.16b
 ; CHECK-NEXT:    str q0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-extend-trunc.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-extend-trunc.ll
index 6da07b855a5c5..b60988be1e76c 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-extend-trunc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-extend-trunc.ll
@@ -462,11 +462,11 @@ define void @fcvt_v8f64_v8f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.d
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    fcvt z0.h, p1/m, z0.d
-; VBITS_GE_256-NEXT:    fcvt z1.h, p1/m, z1.d
+; VBITS_GE_256-NEXT:    ptrue p0.d
+; VBITS_GE_256-NEXT:    fcvt z0.h, p0/m, z0.d
+; VBITS_GE_256-NEXT:    fcvt z1.h, p0/m, z1.d
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-select.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-select.ll
index 13ebda1df7f9d..d1e9dc13f50e8 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-select.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-select.ll
@@ -34,15 +34,15 @@ define <8 x half> @select_v8f16(<8 x half> %op1, <8 x half> %op2, i1 %mask) vsca
 define void @select_v16f16(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl16
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <16 x half>, ptr %a
   %op2 = load volatile <16 x half>, ptr %b
@@ -54,33 +54,33 @@ define void @select_v16f16(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v32f16(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v32f16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov z0.h, w2
+; VBITS_GE_256-NEXT:    ptrue p0.h
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
-; VBITS_GE_256-NEXT:    ptrue p1.h
+; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    and z0.h, z0.h, #0x1
-; VBITS_GE_256-NEXT:    ld1h { z1.h }, p0/z, [x0, x8, lsl #1]
-; VBITS_GE_256-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1h { z3.h }, p0/z, [x1, x8, lsl #1]
-; VBITS_GE_256-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; VBITS_GE_256-NEXT:    ld1h { z0.h }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.h, p1, z1.h, z3.h
-; VBITS_GE_256-NEXT:    mov z0.h, p1/m, z2.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0, x8, lsl #1]
-; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; VBITS_GE_256-NEXT:    ld1h { z0.h }, p1/z, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    ld1h { z1.h }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1h { z2.h }, p1/z, [x1, x8, lsl #1]
+; VBITS_GE_256-NEXT:    ld1h { z3.h }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.h, p0, z0.h, z2.h
+; VBITS_GE_256-NEXT:    sel z1.h, p0, z1.h, z3.h
+; VBITS_GE_256-NEXT:    st1h { z0.h }, p1, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v32f16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    mov z0.h, w2
-; VBITS_GE_512-NEXT:    ptrue p1.h
+; VBITS_GE_512-NEXT:    ptrue p0.h
+; VBITS_GE_512-NEXT:    ptrue p1.h, vl32
 ; VBITS_GE_512-NEXT:    and z0.h, z0.h, #0x1
-; VBITS_GE_512-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; VBITS_GE_512-NEXT:    sel z0.h, p1, z1.h, z2.h
-; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x0]
+; VBITS_GE_512-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; VBITS_GE_512-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.h, p0, z0.h, z1.h
+; VBITS_GE_512-NEXT:    st1h { z0.h }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <32 x half>, ptr %a
   %op2 = load volatile <32 x half>, ptr %b
@@ -92,15 +92,15 @@ define void @select_v32f16(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v64f16(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v64f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <64 x half>, ptr %a
   %op2 = load volatile <64 x half>, ptr %b
@@ -112,15 +112,15 @@ define void @select_v64f16(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v128f16(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v128f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl128
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <128 x half>, ptr %a
   %op2 = load volatile <128 x half>, ptr %b
@@ -158,15 +158,15 @@ define <4 x float> @select_v4f32(<4 x float> %op1, <4 x float> %op2, i1 %mask) v
 define void @select_v8f32(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl8
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <8 x float>, ptr %a
   %op2 = load volatile <8 x float>, ptr %b
@@ -178,33 +178,33 @@ define void @select_v8f32(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v16f32(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v16f32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    and w8, w2, #0x1
-; VBITS_GE_256-NEXT:    ptrue p1.s
+; VBITS_GE_256-NEXT:    ptrue p0.s
 ; VBITS_GE_256-NEXT:    mov z0.s, w8
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0, x8, lsl #2]
-; VBITS_GE_256-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1w { z3.s }, p0/z, [x1, x8, lsl #2]
-; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.s, p1, z1.s, z3.s
-; VBITS_GE_256-NEXT:    mov z0.s, p1/m, z2.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0, x8, lsl #2]
-; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; VBITS_GE_256-NEXT:    ld1w { z0.s }, p1/z, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    ld1w { z1.s }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1w { z2.s }, p1/z, [x1, x8, lsl #2]
+; VBITS_GE_256-NEXT:    ld1w { z3.s }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.s, p0, z0.s, z2.s
+; VBITS_GE_256-NEXT:    sel z1.s, p0, z1.s, z3.s
+; VBITS_GE_256-NEXT:    st1w { z0.s }, p1, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v16f32:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    and w8, w2, #0x1
-; VBITS_GE_512-NEXT:    ptrue p1.s
+; VBITS_GE_512-NEXT:    ptrue p0.s
 ; VBITS_GE_512-NEXT:    mov z0.s, w8
-; VBITS_GE_512-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; VBITS_GE_512-NEXT:    sel z0.s, p1, z1.s, z2.s
-; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x0]
+; VBITS_GE_512-NEXT:    ptrue p1.s, vl16
+; VBITS_GE_512-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; VBITS_GE_512-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.s, p0, z0.s, z1.s
+; VBITS_GE_512-NEXT:    st1w { z0.s }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <16 x float>, ptr %a
   %op2 = load volatile <16 x float>, ptr %b
@@ -216,15 +216,15 @@ define void @select_v16f32(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v32f32(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v32f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl32
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <32 x float>, ptr %a
   %op2 = load volatile <32 x float>, ptr %b
@@ -236,15 +236,15 @@ define void @select_v32f32(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v64f32(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v64f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl64
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <64 x float>, ptr %a
   %op2 = load volatile <64 x float>, ptr %b
@@ -282,16 +282,16 @@ define <2 x double> @select_v2f64(<2 x double> %op1, <2 x double> %op2, i1 %mask
 define void @select_v4f64(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl4
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <4 x double>, ptr %a
   %op2 = load volatile <4 x double>, ptr %b
@@ -303,35 +303,35 @@ define void @select_v4f64(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v8f64(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v8f64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; VBITS_GE_256-NEXT:    and x8, x2, #0x1
-; VBITS_GE_256-NEXT:    ptrue p1.d
+; VBITS_GE_256-NEXT:    ptrue p0.d
 ; VBITS_GE_256-NEXT:    mov z0.d, x8
+; VBITS_GE_256-NEXT:    ptrue p1.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.d, p1, z1.d, z3.d
-; VBITS_GE_256-NEXT:    mov z0.d, p1/m, z2.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1d { z2.d }, p1/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z3.d }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.d, p0, z0.d, z2.d
+; VBITS_GE_256-NEXT:    sel z1.d, p0, z1.d, z3.d
+; VBITS_GE_256-NEXT:    st1d { z0.d }, p1, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v8f64:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; VBITS_GE_512-NEXT:    and x8, x2, #0x1
-; VBITS_GE_512-NEXT:    ptrue p1.d
+; VBITS_GE_512-NEXT:    ptrue p0.d
 ; VBITS_GE_512-NEXT:    mov z0.d, x8
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; VBITS_GE_512-NEXT:    sel z0.d, p1, z1.d, z2.d
-; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
+; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
+; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.d, p0, z0.d, z1.d
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <8 x double>, ptr %a
   %op2 = load volatile <8 x double>, ptr %b
@@ -343,16 +343,16 @@ define void @select_v8f64(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v16f64(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v16f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl16
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <16 x double>, ptr %a
   %op2 = load volatile <16 x double>, ptr %b
@@ -364,16 +364,16 @@ define void @select_v16f64(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v32f64(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v32f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl32
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <32 x double>, ptr %a
   %op2 = load volatile <32 x double>, ptr %b
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-to-int.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-to-int.ll
index da0cf927d74d2..af54b146c5b66 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-to-int.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp-to-int.ll
@@ -131,8 +131,8 @@ define <4 x i32> @fcvtzu_v4f16_v4i32(<4 x half> %op1) vscale_range(2,0) #0 {
 define void @fcvtzu_v8f16_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: fcvtzu_v8f16_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.h
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -357,7 +357,6 @@ define void @fcvtzu_v16f32_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    fcvtzu z0.s, p0/m, z0.s
@@ -366,7 +365,8 @@ define void @fcvtzu_v16f32_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fcvtzu_v16f32_v16i16:
@@ -532,8 +532,8 @@ define <2 x i64> @fcvtzu_v2f32_v2i64(<2 x float> %op1) vscale_range(2,0) #0 {
 define void @fcvtzu_v4f32_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: fcvtzu_v4f32_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.s
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
@@ -752,7 +752,6 @@ define void @fcvtzu_v8f64_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    fcvtzu z0.d, p0/m, z0.d
@@ -761,7 +760,8 @@ define void @fcvtzu_v8f64_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fcvtzu_v8f64_v8i32:
@@ -1024,8 +1024,8 @@ define <4 x i32> @fcvtzs_v4f16_v4i32(<4 x half> %op1) vscale_range(2,0) #0 {
 define void @fcvtzs_v8f16_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: fcvtzs_v8f16_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.h
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -1250,7 +1250,6 @@ define void @fcvtzs_v16f32_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    fcvtzs z0.s, p0/m, z0.s
@@ -1259,7 +1258,8 @@ define void @fcvtzs_v16f32_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fcvtzs_v16f32_v16i16:
@@ -1425,8 +1425,8 @@ define <2 x i64> @fcvtzs_v2f32_v2i64(<2 x float> %op1) vscale_range(2,0) #0 {
 define void @fcvtzs_v4f32_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: fcvtzs_v4f32_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.s
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
@@ -1645,7 +1645,6 @@ define void @fcvtzs_v8f64_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    fcvtzs z0.d, p0/m, z0.d
@@ -1654,7 +1653,8 @@ define void @fcvtzs_v8f64_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: fcvtzs_v8f64_v8i32:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp128.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp128.ll
index 115f722986b5c..61e04682fa0bf 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-fp128.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-fp128.ll
@@ -85,10 +85,10 @@ define void @fcvt_v4f128_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    str q0, [sp, #32] // 16-byte Folded Spill
 ; CHECK-NEXT:    ldr q0, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl __trunctfdf2
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q1, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    add x8, sp, #128
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    ldr z1, [x8, #1, mul vl] // 16-byte Folded Reload
 ; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
@@ -111,14 +111,14 @@ define void @fcvt_v4f128_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ldr q0, [sp, #112] // 16-byte Folded Reload
 ; CHECK-NEXT:    bl __trunctfdf2
 ; CHECK-NEXT:    ldr q1, [sp, #96] // 16-byte Folded Reload
-; CHECK-NEXT:    ptrue p1.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    add x8, sp, #128
-; CHECK-NEXT:    ptrue p0.d, vl4
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    ldr z1, [x8] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov x8, #4 // =0x4
-; CHECK-NEXT:    splice z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    splice z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x19, x8, lsl #3]
 ; CHECK-NEXT:    add x8, sp, #128
 ; CHECK-NEXT:    ldr z0, [x8, #1, mul vl] // 16-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-frame-offests-crash.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-frame-offests-crash.ll
index a4b5ccd69fdb7..1bd688d23050b 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-frame-offests-crash.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-frame-offests-crash.ll
@@ -21,24 +21,24 @@ define dso_local void @func1(ptr %v1, ptr %v2, ptr %v3, ptr %v4, ptr %v5, ptr %v
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w29, -48
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    add x11, sp, #176
-; CHECK-NEXT:    add x10, sp, #144
-; CHECK-NEXT:    add x9, sp, #112
+; CHECK-NEXT:    add x10, sp, #176
 ; CHECK-NEXT:    add x8, sp, #48
+; CHECK-NEXT:    add x9, sp, #144
 ; CHECK-NEXT:    add x20, sp, #176
-; CHECK-NEXT:    ldp x13, x12, [sp, #328]
 ; CHECK-NEXT:    ldr x15, [sp, #104]
+; CHECK-NEXT:    ld1d { z3.d }, p0/z, [x10]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x8]
+; CHECK-NEXT:    add x8, sp, #112
+; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x9]
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x8]
 ; CHECK-NEXT:    ldur q4, [sp, #88]
-; CHECK-NEXT:    ldp x16, x17, [sp, #208]
+; CHECK-NEXT:    ldp x9, x8, [sp, #328]
 ; CHECK-NEXT:    ldr x19, [sp, #272]
+; CHECK-NEXT:    ldp x11, x10, [sp, #312]
+; CHECK-NEXT:    ldp x13, x12, [sp, #296]
+; CHECK-NEXT:    ldp x18, x14, [sp, #280]
+; CHECK-NEXT:    ldp x16, x17, [sp, #208]
 ; CHECK-NEXT:    ldp x21, x22, [sp, #352]
-; CHECK-NEXT:    ld1d { z3.d }, p0/z, [x11]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x10]
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x9]
-; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x8]
-; CHECK-NEXT:    ldp x8, x14, [sp, #312]
-; CHECK-NEXT:    ldp x10, x9, [sp, #296]
-; CHECK-NEXT:    ldp x18, x11, [sp, #280]
 ; CHECK-NEXT:    st1d { z3.d }, p0, [x20]
 ; CHECK-NEXT:    add x20, sp, #144
 ; CHECK-NEXT:    st1d { z2.d }, p0, [x20]
@@ -53,10 +53,10 @@ define dso_local void @func1(ptr %v1, ptr %v2, ptr %v3, ptr %v4, ptr %v5, ptr %v
 ; CHECK-NEXT:    stp x16, x17, [sp, #208]
 ; CHECK-NEXT:    stur q4, [sp, #88]
 ; CHECK-NEXT:    str x15, [sp, #104]
-; CHECK-NEXT:    stp x11, x10, [sp, #288]
-; CHECK-NEXT:    stp x9, x8, [sp, #304]
-; CHECK-NEXT:    stp x14, x13, [sp, #320]
-; CHECK-NEXT:    str x12, [sp, #336]
+; CHECK-NEXT:    stp x14, x13, [sp, #288]
+; CHECK-NEXT:    stp x12, x11, [sp, #304]
+; CHECK-NEXT:    stp x10, x9, [sp, #320]
+; CHECK-NEXT:    str x8, [sp, #336]
 ; CHECK-NEXT:    ldr x29, [sp], #48 // 8-byte Folded Reload
 ; CHECK-NEXT:    b func2
                              ptr %v9, ptr %v10, ptr %v11, ptr %v12, ptr %v13, ptr %v14,  ptr %v15, ptr %v16,
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
index 977c528e2583a..6f4d257039bca 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
@@ -36,16 +36,16 @@ define <8 x half> @insertelement_v8f16(<8 x half> %op1) vscale_range(2,0) #0 {
 define void @insertelement_v16f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: insertelement_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    mov w8, #15 // =0xf
 ; CHECK-NEXT:    index z0.h, #0, #1
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.h, p1/z, z0.h, z1.h
-; CHECK-NEXT:    fmov h0, #5.00000000
-; CHECK-NEXT:    mov z2.h, p1/m, h0
-; CHECK-NEXT:    st1h { z2.h }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.h, vl16
+; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    fmov h1, #5.00000000
+; CHECK-NEXT:    mov z0.h, p0/m, h1
+; CHECK-NEXT:    st1h { z0.h }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x half>, ptr %a
   %r = insertelement <16 x half> %op1, half 5.0, i64 15
@@ -56,33 +56,33 @@ define void @insertelement_v16f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @insertelement_v32f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: insertelement_v32f16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov w8, #15 // =0xf
 ; VBITS_GE_256-NEXT:    index z0.h, #0, #1
-; VBITS_GE_256-NEXT:    ptrue p1.h
+; VBITS_GE_256-NEXT:    ptrue p0.h
 ; VBITS_GE_256-NEXT:    mov z1.h, w8
+; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
-; VBITS_GE_256-NEXT:    fmov h2, #5.00000000
-; VBITS_GE_256-NEXT:    ld1h { z3.h }, p0/z, [x0, x8, lsl #1]
-; VBITS_GE_256-NEXT:    cmpeq p1.h, p1/z, z0.h, z1.h
-; VBITS_GE_256-NEXT:    ld1h { z0.h }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    mov z3.h, p1/m, h2
-; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x1]
-; VBITS_GE_256-NEXT:    st1h { z3.h }, p0, [x1, x8, lsl #1]
+; VBITS_GE_256-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
+; VBITS_GE_256-NEXT:    fmov h0, #5.00000000
+; VBITS_GE_256-NEXT:    ld1h { z1.h }, p1/z, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    mov z1.h, p0/m, h0
+; VBITS_GE_256-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z0.h }, p1, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: insertelement_v32f16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    mov w8, #31 // =0x1f
 ; VBITS_GE_512-NEXT:    index z0.h, #0, #1
-; VBITS_GE_512-NEXT:    ptrue p1.h
+; VBITS_GE_512-NEXT:    ptrue p0.h
 ; VBITS_GE_512-NEXT:    mov z1.h, w8
-; VBITS_GE_512-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    cmpeq p1.h, p1/z, z0.h, z1.h
-; VBITS_GE_512-NEXT:    fmov h0, #5.00000000
-; VBITS_GE_512-NEXT:    mov z2.h, p1/m, h0
-; VBITS_GE_512-NEXT:    st1h { z2.h }, p0, [x1]
+; VBITS_GE_512-NEXT:    ptrue p1.h, vl32
+; VBITS_GE_512-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
+; VBITS_GE_512-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    fmov h1, #5.00000000
+; VBITS_GE_512-NEXT:    mov z0.h, p0/m, h1
+; VBITS_GE_512-NEXT:    st1h { z0.h }, p1, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <32 x half>, ptr %a
   %r = insertelement <32 x half> %op1, half 5.0, i64 31
@@ -93,16 +93,16 @@ define void @insertelement_v32f16(ptr %a, ptr %b) #0 {
 define void @insertelement_v64f16(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: insertelement_v64f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov w8, #63 // =0x3f
 ; CHECK-NEXT:    index z0.h, #0, #1
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.h, p1/z, z0.h, z1.h
-; CHECK-NEXT:    fmov h0, #5.00000000
-; CHECK-NEXT:    mov z2.h, p1/m, h0
-; CHECK-NEXT:    st1h { z2.h }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.h, vl64
+; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    fmov h1, #5.00000000
+; CHECK-NEXT:    mov z0.h, p0/m, h1
+; CHECK-NEXT:    st1h { z0.h }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <64 x half>, ptr %a
   %r = insertelement <64 x half> %op1, half 5.0, i64 63
@@ -113,16 +113,16 @@ define void @insertelement_v64f16(ptr %a, ptr %b) vscale_range(8,0) #0 {
 define void @insertelement_v128f16(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: insertelement_v128f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov w8, #127 // =0x7f
 ; CHECK-NEXT:    index z0.h, #0, #1
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.h, p1/z, z0.h, z1.h
-; CHECK-NEXT:    fmov h0, #5.00000000
-; CHECK-NEXT:    mov z2.h, p1/m, h0
-; CHECK-NEXT:    st1h { z2.h }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.h, vl128
+; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    fmov h1, #5.00000000
+; CHECK-NEXT:    mov z0.h, p0/m, h1
+; CHECK-NEXT:    st1h { z0.h }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <128 x half>, ptr %a
   %r = insertelement <128 x half> %op1, half 5.0, i64 127
@@ -157,16 +157,16 @@ define <4 x float> @insertelement_v4f32(<4 x float> %op1) vscale_range(2,0) #0 {
 define void @insertelement_v8f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: insertelement_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z0.s, #0, #1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.s, p1/z, z0.s, z1.s
-; CHECK-NEXT:    fmov s0, #5.00000000
-; CHECK-NEXT:    mov z2.s, p1/m, s0
-; CHECK-NEXT:    st1w { z2.s }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.s, vl8
+; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    fmov s1, #5.00000000
+; CHECK-NEXT:    mov z0.s, p0/m, s1
+; CHECK-NEXT:    st1w { z0.s }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <8 x float>, ptr %a
   %r = insertelement <8 x float> %op1, float 5.0, i64 7
@@ -177,33 +177,33 @@ define void @insertelement_v8f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @insertelement_v16f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: insertelement_v16f32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov w8, #7 // =0x7
 ; VBITS_GE_256-NEXT:    index z0.s, #0, #1
-; VBITS_GE_256-NEXT:    ptrue p1.s
+; VBITS_GE_256-NEXT:    ptrue p0.s
 ; VBITS_GE_256-NEXT:    mov z1.s, w8
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    fmov s2, #5.00000000
-; VBITS_GE_256-NEXT:    ld1w { z3.s }, p0/z, [x0, x8, lsl #2]
-; VBITS_GE_256-NEXT:    cmpeq p1.s, p1/z, z0.s, z1.s
-; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    mov z3.s, p1/m, s2
-; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x1]
-; VBITS_GE_256-NEXT:    st1w { z3.s }, p0, [x1, x8, lsl #2]
+; VBITS_GE_256-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
+; VBITS_GE_256-NEXT:    fmov s0, #5.00000000
+; VBITS_GE_256-NEXT:    ld1w { z1.s }, p1/z, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    mov z1.s, p0/m, s0
+; VBITS_GE_256-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z0.s }, p1, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: insertelement_v16f32:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    mov w8, #15 // =0xf
 ; VBITS_GE_512-NEXT:    index z0.s, #0, #1
-; VBITS_GE_512-NEXT:    ptrue p1.s
+; VBITS_GE_512-NEXT:    ptrue p0.s
 ; VBITS_GE_512-NEXT:    mov z1.s, w8
-; VBITS_GE_512-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    cmpeq p1.s, p1/z, z0.s, z1.s
-; VBITS_GE_512-NEXT:    fmov s0, #5.00000000
-; VBITS_GE_512-NEXT:    mov z2.s, p1/m, s0
-; VBITS_GE_512-NEXT:    st1w { z2.s }, p0, [x1]
+; VBITS_GE_512-NEXT:    ptrue p1.s, vl16
+; VBITS_GE_512-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
+; VBITS_GE_512-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    fmov s1, #5.00000000
+; VBITS_GE_512-NEXT:    mov z0.s, p0/m, s1
+; VBITS_GE_512-NEXT:    st1w { z0.s }, p1, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <16 x float>, ptr %a
   %r = insertelement <16 x float> %op1, float 5.0, i64 15
@@ -214,16 +214,16 @@ define void @insertelement_v16f32(ptr %a, ptr %b) #0 {
 define void @insertelement_v32f32(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: insertelement_v32f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    mov w8, #31 // =0x1f
 ; CHECK-NEXT:    index z0.s, #0, #1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.s, p1/z, z0.s, z1.s
-; CHECK-NEXT:    fmov s0, #5.00000000
-; CHECK-NEXT:    mov z2.s, p1/m, s0
-; CHECK-NEXT:    st1w { z2.s }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.s, vl32
+; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    fmov s1, #5.00000000
+; CHECK-NEXT:    mov z0.s, p0/m, s1
+; CHECK-NEXT:    st1w { z0.s }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x float>, ptr %a
   %r = insertelement <32 x float> %op1, float 5.0, i64 31
@@ -234,16 +234,16 @@ define void @insertelement_v32f32(ptr %a, ptr %b) vscale_range(8,0) #0 {
 define void @insertelement_v64f32(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: insertelement_v64f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    mov w8, #63 // =0x3f
 ; CHECK-NEXT:    index z0.s, #0, #1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.s, p1/z, z0.s, z1.s
-; CHECK-NEXT:    fmov s0, #5.00000000
-; CHECK-NEXT:    mov z2.s, p1/m, s0
-; CHECK-NEXT:    st1w { z2.s }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.s, vl64
+; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    fmov s1, #5.00000000
+; CHECK-NEXT:    mov z0.s, p0/m, s1
+; CHECK-NEXT:    st1w { z0.s }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <64 x float>, ptr %a
   %r = insertelement <64 x float> %op1, float 5.0, i64 63
@@ -276,16 +276,16 @@ define <2 x double> @insertelement_v2f64(<2 x double> %op1) vscale_range(2,0) #0
 define void @insertelement_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: insertelement_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z0.d, #0, #1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.d, p1/z, z0.d, z1.d
-; CHECK-NEXT:    fmov d0, #5.00000000
-; CHECK-NEXT:    mov z2.d, p1/m, d0
-; CHECK-NEXT:    st1d { z2.d }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.d, vl4
+; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    fmov d1, #5.00000000
+; CHECK-NEXT:    mov z0.d, p0/m, d1
+; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <4 x double>, ptr %a
   %r = insertelement <4 x double> %op1, double 5.0, i64 3
@@ -296,33 +296,33 @@ define void @insertelement_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @insertelement_v8f64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: insertelement_v8f64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov w8, #3 // =0x3
 ; VBITS_GE_256-NEXT:    index z0.d, #0, #1
-; VBITS_GE_256-NEXT:    ptrue p1.d
+; VBITS_GE_256-NEXT:    ptrue p0.d
 ; VBITS_GE_256-NEXT:    mov z1.d, x8
+; VBITS_GE_256-NEXT:    ptrue p1.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    fmov d2, #5.00000000
-; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    cmpeq p1.d, p1/z, z0.d, z1.d
-; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    mov z3.d, p1/m, d2
-; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x1]
-; VBITS_GE_256-NEXT:    st1d { z3.d }, p0, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
+; VBITS_GE_256-NEXT:    fmov d0, #5.00000000
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    mov z1.d, p0/m, d0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p1, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z0.d }, p1, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: insertelement_v8f64:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    mov w8, #7 // =0x7
 ; VBITS_GE_512-NEXT:    index z0.d, #0, #1
-; VBITS_GE_512-NEXT:    ptrue p1.d
+; VBITS_GE_512-NEXT:    ptrue p0.d
 ; VBITS_GE_512-NEXT:    mov z1.d, x8
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    cmpeq p1.d, p1/z, z0.d, z1.d
-; VBITS_GE_512-NEXT:    fmov d0, #5.00000000
-; VBITS_GE_512-NEXT:    mov z2.d, p1/m, d0
-; VBITS_GE_512-NEXT:    st1d { z2.d }, p0, [x1]
+; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
+; VBITS_GE_512-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    fmov d1, #5.00000000
+; VBITS_GE_512-NEXT:    mov z0.d, p0/m, d1
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load <8 x double>, ptr %a
   %r = insertelement <8 x double> %op1, double 5.0, i64 7
@@ -333,16 +333,16 @@ define void @insertelement_v8f64(ptr %a, ptr %b) #0 {
 define void @insertelement_v16f64(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: insertelement_v16f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    mov w8, #15 // =0xf
 ; CHECK-NEXT:    index z0.d, #0, #1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.d, p1/z, z0.d, z1.d
-; CHECK-NEXT:    fmov d0, #5.00000000
-; CHECK-NEXT:    mov z2.d, p1/m, d0
-; CHECK-NEXT:    st1d { z2.d }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.d, vl16
+; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    fmov d1, #5.00000000
+; CHECK-NEXT:    mov z0.d, p0/m, d1
+; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x double>, ptr %a
   %r = insertelement <16 x double> %op1, double 5.0, i64 15
@@ -353,16 +353,16 @@ define void @insertelement_v16f64(ptr %a, ptr %b) vscale_range(8,0) #0 {
 define void @insertelement_v32f64(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: insertelement_v32f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    mov w8, #31 // =0x1f
 ; CHECK-NEXT:    index z0.d, #0, #1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; CHECK-NEXT:    cmpeq p1.d, p1/z, z0.d, z1.d
-; CHECK-NEXT:    fmov d0, #5.00000000
-; CHECK-NEXT:    mov z2.d, p1/m, d0
-; CHECK-NEXT:    st1d { z2.d }, p0, [x1]
+; CHECK-NEXT:    ptrue p1.d, vl32
+; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    fmov d1, #5.00000000
+; CHECK-NEXT:    mov z0.d, p0/m, d1
+; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x double>, ptr %a
   %r = insertelement <32 x double> %op1, double 5.0, i64 31
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
index b45a62f4e7581..58fca3a2cf8b6 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-arith.ll
@@ -1388,11 +1388,11 @@ define void @abs_v128i16(ptr %a) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    mov x10, #64 // =0x40
 ; CHECK-NEXT:    mov x11, #80 // =0x50
 ; CHECK-NEXT:    mov x12, #32 // =0x20
-; CHECK-NEXT:    mov x13, #48 // =0x30
-; CHECK-NEXT:    mov x14, #16 // =0x10
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0, x9, lsl #1]
 ; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x0, x10, lsl #1]
+; CHECK-NEXT:    mov x13, #48 // =0x30
+; CHECK-NEXT:    mov x14, #16 // =0x10
 ; CHECK-NEXT:    ld1h { z3.h }, p0/z, [x0, x11, lsl #1]
 ; CHECK-NEXT:    ld1h { z4.h }, p0/z, [x0, x12, lsl #1]
 ; CHECK-NEXT:    ld1h { z5.h }, p0/z, [x0, x13, lsl #1]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-div.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-div.ll
index 11ed69513917c..0ddf434eff930 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-div.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-div.ll
@@ -15,9 +15,9 @@ target triple = "aarch64-unknown-linux-gnu"
 define <8 x i8> @sdiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v8i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll v1.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    sshll v0.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.4s, v0.8h, #0
 ; VBITS_GE_128-NEXT:    sshll v1.4s, v1.4h, #0
@@ -94,11 +94,11 @@ define <8 x i8> @sdiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 define <16 x i8> @sdiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v16i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v2.8h, v1.16b, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.8h, v0.16b, #0
 ; VBITS_GE_128-NEXT:    sshll v1.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    sshll v0.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    sshll v2.4s, v2.4h, #0
@@ -203,7 +203,6 @@ define void @sdiv_v128i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: sdiv_v128i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b, vl128
-; CHECK-NEXT:    ptrue p1.h, vl128
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
 ; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x1]
 ; CHECK-NEXT:    ptrue p0.s, vl64
@@ -221,7 +220,8 @@ define void @sdiv_v128i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    uzp1 z1.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    splice z1.h, p0, z1.h, z0.h
-; CHECK-NEXT:    st1b { z1.h }, p1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl128
+; CHECK-NEXT:    st1b { z1.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <128 x i8>, ptr %a
   %op2 = load <128 x i8>, ptr %b
@@ -260,14 +260,14 @@ define void @sdiv_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    sdivr z3.s, p1/m, z3.s, z5.s
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    sdiv z0.s, p1/m, z0.s, z1.s
-; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    uzp1 z1.h, z4.h, z4.h
+; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z1.h, p1, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    splice z3.h, p1, z3.h, z0.h
-; CHECK-NEXT:    ptrue p1.b, vl128
 ; CHECK-NEXT:    uzp1 z0.b, z1.b, z1.b
+; CHECK-NEXT:    ptrue p1.b, vl128
 ; CHECK-NEXT:    uzp1 z1.b, z3.b, z3.b
 ; CHECK-NEXT:    splice z0.b, p1, z0.b, z1.b
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
@@ -284,18 +284,18 @@ define void @sdiv_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <4 x i16> @sdiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v4i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll v1.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    sshll v0.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_128-NEXT:    xtn v0.4h, v0.4s
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: sdiv_v4i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    sshll v1.4s, v1.4h, #0
 ; VBITS_GE_256-NEXT:    sshll v0.4s, v0.4h, #0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    sdivr z1.s, p0/m, z1.s, z0.s
 ; VBITS_GE_256-NEXT:    mov w8, v1.s[1]
 ; VBITS_GE_256-NEXT:    mov v0.16b, v1.16b
@@ -309,9 +309,9 @@ define <4 x i16> @sdiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: sdiv_v4i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    sshll v1.4s, v1.4h, #0
 ; VBITS_GE_512-NEXT:    sshll v0.4s, v0.4h, #0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    sdivr z1.s, p0/m, z1.s, z0.s
 ; VBITS_GE_512-NEXT:    mov w8, v1.s[1]
 ; VBITS_GE_512-NEXT:    mov v0.16b, v1.16b
@@ -329,11 +329,11 @@ define <4 x i16> @sdiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 define <8 x i16> @sdiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v8i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.4s, v0.8h, #0
 ; VBITS_GE_128-NEXT:    sshll v1.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    sshll v0.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_128-NEXT:    uzp1 v0.8h, v0.8h, v2.8h
@@ -341,9 +341,9 @@ define <8 x i16> @sdiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: sdiv_v8i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_256-NEXT:    // kill: def $q0 killed $q0 def $z0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
@@ -353,9 +353,9 @@ define <8 x i16> @sdiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: sdiv_v8i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_512-NEXT:    // kill: def $q0 killed $q0 def $z0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
 ; VBITS_GE_512-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
@@ -369,8 +369,8 @@ define <8 x i16> @sdiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 define void @sdiv_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v16i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q4, q1, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldr q0, [x0, #16]
 ; VBITS_GE_128-NEXT:    sshll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.4s, v0.8h, #0
@@ -542,8 +542,8 @@ define void @sdiv_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @sdiv_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q5, q4, [x1, #32]
 ; VBITS_GE_128-NEXT:    sdivr z0.s, p0/m, z0.s, z1.s
@@ -664,8 +664,8 @@ define void @sdiv_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @sdiv_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: sdiv_v8i64:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q5, q4, [x1, #32]
 ; VBITS_GE_128-NEXT:    sdivr z0.d, p0/m, z0.d, z1.d
@@ -748,9 +748,9 @@ define void @sdiv_v32i64(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <8 x i8> @udiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: udiv_v8i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll v1.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    ushll v0.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.4s, v0.8h, #0
 ; VBITS_GE_128-NEXT:    ushll v1.4s, v1.4h, #0
@@ -827,11 +827,11 @@ define <8 x i8> @udiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 define <16 x i8> @udiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: udiv_v16i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v2.8h, v1.16b, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.8h, v0.16b, #0
 ; VBITS_GE_128-NEXT:    ushll v1.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    ushll v0.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    ushll v2.4s, v2.4h, #0
@@ -980,14 +980,14 @@ define void @udiv_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    udivr z3.s, p1/m, z3.s, z5.s
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    udiv z0.s, p1/m, z0.s, z1.s
-; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    uzp1 z1.h, z4.h, z4.h
+; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z1.h, p1, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    splice z3.h, p1, z3.h, z0.h
-; CHECK-NEXT:    ptrue p1.b, vl128
 ; CHECK-NEXT:    uzp1 z0.b, z1.b, z1.b
+; CHECK-NEXT:    ptrue p1.b, vl128
 ; CHECK-NEXT:    uzp1 z1.b, z3.b, z3.b
 ; CHECK-NEXT:    splice z0.b, p1, z0.b, z1.b
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
@@ -1004,18 +1004,18 @@ define void @udiv_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <4 x i16> @udiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: udiv_v4i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll v1.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    ushll v0.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_128-NEXT:    xtn v0.4h, v0.4s
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: udiv_v4i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    ushll v1.4s, v1.4h, #0
 ; VBITS_GE_256-NEXT:    ushll v0.4s, v0.4h, #0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    udivr z1.s, p0/m, z1.s, z0.s
 ; VBITS_GE_256-NEXT:    mov w8, v1.s[1]
 ; VBITS_GE_256-NEXT:    mov v0.16b, v1.16b
@@ -1029,9 +1029,9 @@ define <4 x i16> @udiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: udiv_v4i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    ushll v1.4s, v1.4h, #0
 ; VBITS_GE_512-NEXT:    ushll v0.4s, v0.4h, #0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    udivr z1.s, p0/m, z1.s, z0.s
 ; VBITS_GE_512-NEXT:    mov w8, v1.s[1]
 ; VBITS_GE_512-NEXT:    mov v0.16b, v1.16b
@@ -1049,11 +1049,11 @@ define <4 x i16> @udiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 define <8 x i16> @udiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: udiv_v8i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.4s, v0.8h, #0
 ; VBITS_GE_128-NEXT:    ushll v1.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    ushll v0.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
 ; VBITS_GE_128-NEXT:    uzp1 v0.8h, v0.8h, v2.8h
@@ -1061,9 +1061,9 @@ define <8 x i16> @udiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: udiv_v8i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_256-NEXT:    // kill: def $q0 killed $q0 def $z0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    uunpklo z1.s, z1.h
 ; VBITS_GE_256-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
@@ -1073,9 +1073,9 @@ define <8 x i16> @udiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: udiv_v8i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_512-NEXT:    // kill: def $q0 killed $q0 def $z0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    uunpklo z1.s, z1.h
 ; VBITS_GE_512-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_512-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
@@ -1089,8 +1089,8 @@ define <8 x i16> @udiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 define void @udiv_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: udiv_v16i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q4, q1, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldr q0, [x0, #16]
 ; VBITS_GE_128-NEXT:    ushll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.4s, v0.8h, #0
@@ -1253,8 +1253,8 @@ define void @udiv_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @udiv_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: udiv_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q5, q4, [x1, #32]
 ; VBITS_GE_128-NEXT:    udivr z0.s, p0/m, z0.s, z1.s
@@ -1375,8 +1375,8 @@ define void @udiv_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @udiv_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: udiv_v8i64:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q5, q4, [x1, #32]
 ; VBITS_GE_128-NEXT:    udivr z0.d, p0/m, z0.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-extends.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-extends.ll
index 756e5f4cddf80..4feb86305f8f6 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-extends.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-extends.ll
@@ -58,8 +58,8 @@ define void @sext_v4i3_v4i64(<4 x i3> %a, ptr %out) vscale_range(2,0) #0 {
 define void @sext_v16i8_v16i16(<16 x i8> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: sext_v16i8_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    sunpklo z0.h, z0.b
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -308,8 +308,8 @@ define void @sext_v32i8_v32i64(ptr %in, ptr %out) vscale_range(16,0) #0 {
 define void @sext_v8i16_v8i32(<8 x i16> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: sext_v8i16_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -472,8 +472,8 @@ define void @sext_v32i16_v32i64(ptr %in, ptr %out) vscale_range(16,0) #0 {
 define void @sext_v4i32_v4i64(<4 x i32> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: sext_v4i32_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -554,8 +554,8 @@ define void @sext_v32i32_v32i64(ptr %in, ptr %out) vscale_range(16,0) #0 {
 define void @zext_v16i8_v16i16(<16 x i8> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: zext_v16i8_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -804,8 +804,8 @@ define void @zext_v32i8_v32i64(ptr %in, ptr %out) vscale_range(16,0) #0 {
 define void @zext_v8i16_v8i32(<8 x i16> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: zext_v8i16_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -968,8 +968,8 @@ define void @zext_v32i16_v32i64(ptr %in, ptr %out) vscale_range(16,0) #0 {
 define void @zext_v4i32_v4i64(<4 x i32> %a, ptr %out) vscale_range(2,0) #0 {
 ; CHECK-LABEL: zext_v4i32_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
index 38444d83c1d7e..2d78945399176 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-rem.ll
@@ -15,9 +15,9 @@ target triple = "aarch64-unknown-linux-gnu"
 define <8 x i8> @srem_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: srem_v8i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll v2.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    sshll v3.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    sshll v2.4s, v2.4h, #0
@@ -97,9 +97,9 @@ define <8 x i8> @srem_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 define <16 x i8> @srem_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: srem_v16i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v2.8h, v1.16b, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.8h, v0.16b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    sshll v2.4s, v2.4h, #0
@@ -277,8 +277,8 @@ define void @srem_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    sdivr z3.s, p1/m, z3.s, z5.s
 ; CHECK-NEXT:    ptrue p1.h, vl64
-; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z4.h, p1, z4.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z2.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z5.h, p1, z5.h, z3.h
@@ -300,9 +300,9 @@ define void @srem_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <4 x i16> @srem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: srem_v4i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll v2.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    sshll v3.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    xtn v2.4h, v2.4s
 ; VBITS_GE_128-NEXT:    mls v0.4h, v2.4h, v1.4h
@@ -310,9 +310,9 @@ define <4 x i16> @srem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: srem_v4i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    sshll v2.4s, v1.4h, #0
 ; VBITS_GE_256-NEXT:    sshll v3.4s, v0.4h, #0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_256-NEXT:    mov w8, v2.s[1]
 ; VBITS_GE_256-NEXT:    mov v3.16b, v2.16b
@@ -326,9 +326,9 @@ define <4 x i16> @srem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: srem_v4i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    sshll v2.4s, v1.4h, #0
 ; VBITS_GE_512-NEXT:    sshll v3.4s, v0.4h, #0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_512-NEXT:    mov w8, v2.s[1]
 ; VBITS_GE_512-NEXT:    mov v3.16b, v2.16b
@@ -346,9 +346,9 @@ define <4 x i16> @srem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: srem_v8i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.4s, v0.8h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    sshll v4.4s, v0.4h, #0
 ; VBITS_GE_128-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    sshll v3.4s, v1.4h, #0
@@ -359,11 +359,11 @@ define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: srem_v8i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_256-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; VBITS_GE_256-NEXT:    sunpklo z2.s, z1.h
 ; VBITS_GE_256-NEXT:    sunpklo z3.s, z0.h
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_256-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; VBITS_GE_256-NEXT:    mls v0.8h, v2.8h, v1.8h
@@ -372,11 +372,11 @@ define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: srem_v8i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_512-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; VBITS_GE_512-NEXT:    sunpklo z2.s, z1.h
 ; VBITS_GE_512-NEXT:    sunpklo z3.s, z0.h
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_512-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; VBITS_GE_512-NEXT:    mls v0.8h, v2.8h, v1.8h
@@ -389,8 +389,8 @@ define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 define void @srem_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: srem_v16i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q4, q1, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldr q0, [x0, #16]
 ; VBITS_GE_128-NEXT:    sshll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    sshll2 v3.4s, v0.8h, #0
@@ -582,25 +582,25 @@ define void @srem_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @srem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: srem_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
 ; VBITS_GE_128-NEXT:    movprfx z4, z1
 ; VBITS_GE_128-NEXT:    sdiv z4.s, p0/m, z4.s, z0.s
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    sdiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    sdiv z7.s, p0/m, z7.s, z6.s
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    sdiv z18.s, p0/m, z18.s, z17.s
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    sdiv z19.s, p0/m, z19.s, z3.s
-; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
-; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
 ; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
 ; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
-; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
+; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
 ; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
+; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: srem_v16i32:
@@ -730,26 +730,26 @@ define void @srem_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @srem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: srem_v8i64:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
 ; VBITS_GE_128-NEXT:    movprfx z4, z1
 ; VBITS_GE_128-NEXT:    sdiv z4.d, p0/m, z4.d, z0.d
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    sdiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    sdiv z7.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    sdiv z18.d, p0/m, z18.d, z17.d
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    sdiv z19.d, p0/m, z19.d, z3.d
-; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
-; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    msb z0.d, p0/m, z4.d, z1.d
 ; VBITS_GE_128-NEXT:    movprfx z1, z2
 ; VBITS_GE_128-NEXT:    mls z1.d, p0/m, z19.d, z3.d
-; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
+; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    stp q0, q1, [x0]
+; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: srem_v8i64:
@@ -833,9 +833,9 @@ define void @srem_v32i64(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <8 x i8> @urem_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: urem_v8i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll v2.8h, v1.8b, #0
 ; VBITS_GE_128-NEXT:    ushll v3.8h, v0.8b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    ushll v2.4s, v2.4h, #0
@@ -915,9 +915,9 @@ define <8 x i8> @urem_v8i8(<8 x i8> %op1, <8 x i8> %op2) #0 {
 define <16 x i8> @urem_v16i8(<16 x i8> %op1, <16 x i8> %op2) #0 {
 ; VBITS_GE_128-LABEL: urem_v16i8:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v2.8h, v1.16b, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.8h, v0.16b, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v4.4s, v2.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v5.4s, v3.8h, #0
 ; VBITS_GE_128-NEXT:    ushll v2.4s, v2.4h, #0
@@ -1095,8 +1095,8 @@ define void @urem_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    udivr z3.s, p1/m, z3.s, z5.s
 ; CHECK-NEXT:    ptrue p1.h, vl64
-; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z4.h, p1, z4.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z2.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z5.h, p1, z5.h, z3.h
@@ -1118,9 +1118,9 @@ define void @urem_v256i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 define <4 x i16> @urem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: urem_v4i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll v2.4s, v1.4h, #0
 ; VBITS_GE_128-NEXT:    ushll v3.4s, v0.4h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    xtn v2.4h, v2.4s
 ; VBITS_GE_128-NEXT:    mls v0.4h, v2.4h, v1.4h
@@ -1128,9 +1128,9 @@ define <4 x i16> @urem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: urem_v4i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    ushll v2.4s, v1.4h, #0
 ; VBITS_GE_256-NEXT:    ushll v3.4s, v0.4h, #0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_256-NEXT:    mov w8, v2.s[1]
 ; VBITS_GE_256-NEXT:    mov v3.16b, v2.16b
@@ -1144,9 +1144,9 @@ define <4 x i16> @urem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: urem_v4i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    ushll v2.4s, v1.4h, #0
 ; VBITS_GE_512-NEXT:    ushll v3.4s, v0.4h, #0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_512-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_512-NEXT:    mov w8, v2.s[1]
 ; VBITS_GE_512-NEXT:    mov v3.16b, v2.16b
@@ -1164,9 +1164,9 @@ define <4 x i16> @urem_v4i16(<4 x i16> %op1, <4 x i16> %op2) #0 {
 define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ; VBITS_GE_128-LABEL: urem_v8i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.4s, v0.8h, #0
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ushll v4.4s, v0.4h, #0
 ; VBITS_GE_128-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_128-NEXT:    ushll v3.4s, v1.4h, #0
@@ -1177,11 +1177,11 @@ define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_256-LABEL: urem_v8i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_256-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; VBITS_GE_256-NEXT:    uunpklo z2.s, z1.h
 ; VBITS_GE_256-NEXT:    uunpklo z3.s, z0.h
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_256-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; VBITS_GE_256-NEXT:    mls v0.8h, v2.8h, v1.8h
@@ -1190,11 +1190,11 @@ define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 ;
 ; VBITS_GE_512-LABEL: urem_v8i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; VBITS_GE_512-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; VBITS_GE_512-NEXT:    uunpklo z2.s, z1.h
 ; VBITS_GE_512-NEXT:    uunpklo z3.s, z0.h
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_512-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; VBITS_GE_512-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; VBITS_GE_512-NEXT:    mls v0.8h, v2.8h, v1.8h
@@ -1207,8 +1207,8 @@ define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) #0 {
 define void @urem_v16i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: urem_v16i16:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q4, q1, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldr q0, [x0, #16]
 ; VBITS_GE_128-NEXT:    ushll2 v2.4s, v1.8h, #0
 ; VBITS_GE_128-NEXT:    ushll2 v3.4s, v0.8h, #0
@@ -1400,25 +1400,25 @@ define void @urem_v8i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @urem_v16i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: urem_v16i32:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
 ; VBITS_GE_128-NEXT:    movprfx z4, z1
 ; VBITS_GE_128-NEXT:    udiv z4.s, p0/m, z4.s, z0.s
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    udiv z19.s, p0/m, z19.s, z3.s
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    udiv z7.s, p0/m, z7.s, z6.s
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    udiv z18.s, p0/m, z18.s, z17.s
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    udiv z19.s, p0/m, z19.s, z3.s
-; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
-; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
 ; VBITS_GE_128-NEXT:    mls v1.4s, v4.4s, v0.4s
 ; VBITS_GE_128-NEXT:    mls v2.4s, v19.4s, v3.4s
-; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    mls v16.4s, v18.4s, v17.4s
+; VBITS_GE_128-NEXT:    mls v5.4s, v7.4s, v6.4s
 ; VBITS_GE_128-NEXT:    stp q1, q2, [x0]
+; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: urem_v16i32:
@@ -1548,26 +1548,26 @@ define void @urem_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @urem_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_128-LABEL: urem_v8i64:
 ; VBITS_GE_128:       // %bb.0:
-; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q0, q3, [x1]
+; VBITS_GE_128-NEXT:    ptrue p0.d, vl2
 ; VBITS_GE_128-NEXT:    ldp q1, q2, [x0]
 ; VBITS_GE_128-NEXT:    ldp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ldp q17, q6, [x1, #32]
 ; VBITS_GE_128-NEXT:    movprfx z4, z1
 ; VBITS_GE_128-NEXT:    udiv z4.d, p0/m, z4.d, z0.d
+; VBITS_GE_128-NEXT:    movprfx z19, z2
+; VBITS_GE_128-NEXT:    udiv z19.d, p0/m, z19.d, z3.d
 ; VBITS_GE_128-NEXT:    movprfx z7, z5
 ; VBITS_GE_128-NEXT:    udiv z7.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    movprfx z18, z16
 ; VBITS_GE_128-NEXT:    udiv z18.d, p0/m, z18.d, z17.d
-; VBITS_GE_128-NEXT:    movprfx z19, z2
-; VBITS_GE_128-NEXT:    udiv z19.d, p0/m, z19.d, z3.d
-; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
-; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    msb z0.d, p0/m, z4.d, z1.d
 ; VBITS_GE_128-NEXT:    movprfx z1, z2
 ; VBITS_GE_128-NEXT:    mls z1.d, p0/m, z19.d, z3.d
-; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
+; VBITS_GE_128-NEXT:    mls z16.d, p0/m, z18.d, z17.d
+; VBITS_GE_128-NEXT:    mls z5.d, p0/m, z7.d, z6.d
 ; VBITS_GE_128-NEXT:    stp q0, q1, [x0]
+; VBITS_GE_128-NEXT:    stp q16, q5, [x0, #32]
 ; VBITS_GE_128-NEXT:    ret
 ;
 ; VBITS_GE_256-LABEL: urem_v8i64:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-select.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-select.ll
index 710dce4de6dda..37396ba7011be 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-select.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-select.ll
@@ -34,14 +34,14 @@ define <16 x i8> @select_v16i8(<16 x i8> %op1, <16 x i8> %op2, i1 %mask) vscale_
 define void @select_v32i8(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl32
 ; CHECK-NEXT:    mov z0.b, w2
-; CHECK-NEXT:    ptrue p1.b
-; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x0]
-; CHECK-NEXT:    ld1b { z2.b }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.b, p1/z, z0.b, #0
-; CHECK-NEXT:    sel z0.b, p1, z1.b, z2.b
-; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    ptrue p1.b, vl32
+; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    ld1b { z0.b }, p1/z, [x0]
+; CHECK-NEXT:    ld1b { z1.b }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.b, p0, z0.b, z1.b
+; CHECK-NEXT:    st1b { z0.b }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <32 x i8>, ptr %a
   %op2 = load volatile <32 x i8>, ptr %b
@@ -53,31 +53,31 @@ define void @select_v32i8(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v64i8(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v64i8:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
 ; VBITS_GE_256-NEXT:    mov z0.b, w2
+; VBITS_GE_256-NEXT:    ptrue p0.b
 ; VBITS_GE_256-NEXT:    mov w8, #32 // =0x20
-; VBITS_GE_256-NEXT:    ptrue p1.b
-; VBITS_GE_256-NEXT:    ld1b { z1.b }, p0/z, [x0, x8]
-; VBITS_GE_256-NEXT:    ld1b { z2.b }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1b { z3.b }, p0/z, [x1, x8]
-; VBITS_GE_256-NEXT:    cmpne p1.b, p1/z, z0.b, #0
-; VBITS_GE_256-NEXT:    ld1b { z0.b }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.b, p1, z1.b, z3.b
-; VBITS_GE_256-NEXT:    mov z0.b, p1/m, z2.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x0, x8]
-; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x0]
+; VBITS_GE_256-NEXT:    ptrue p1.b, vl32
+; VBITS_GE_256-NEXT:    cmpne p0.b, p0/z, z0.b, #0
+; VBITS_GE_256-NEXT:    ld1b { z0.b }, p1/z, [x0, x8]
+; VBITS_GE_256-NEXT:    ld1b { z1.b }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1b { z2.b }, p1/z, [x1, x8]
+; VBITS_GE_256-NEXT:    ld1b { z3.b }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.b, p0, z0.b, z2.b
+; VBITS_GE_256-NEXT:    sel z1.b, p0, z1.b, z3.b
+; VBITS_GE_256-NEXT:    st1b { z0.b }, p1, [x0, x8]
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v64i8:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.b, vl64
 ; VBITS_GE_512-NEXT:    mov z0.b, w2
-; VBITS_GE_512-NEXT:    ptrue p1.b
-; VBITS_GE_512-NEXT:    ld1b { z1.b }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1b { z2.b }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.b, p1/z, z0.b, #0
-; VBITS_GE_512-NEXT:    sel z0.b, p1, z1.b, z2.b
-; VBITS_GE_512-NEXT:    st1b { z0.b }, p0, [x0]
+; VBITS_GE_512-NEXT:    ptrue p0.b
+; VBITS_GE_512-NEXT:    ptrue p1.b, vl64
+; VBITS_GE_512-NEXT:    cmpne p0.b, p0/z, z0.b, #0
+; VBITS_GE_512-NEXT:    ld1b { z0.b }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1b { z1.b }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.b, p0, z0.b, z1.b
+; VBITS_GE_512-NEXT:    st1b { z0.b }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <64 x i8>, ptr %a
   %op2 = load volatile <64 x i8>, ptr %b
@@ -89,14 +89,14 @@ define void @select_v64i8(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v128i8(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v128i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl128
 ; CHECK-NEXT:    mov z0.b, w2
-; CHECK-NEXT:    ptrue p1.b
-; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x0]
-; CHECK-NEXT:    ld1b { z2.b }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.b, p1/z, z0.b, #0
-; CHECK-NEXT:    sel z0.b, p1, z1.b, z2.b
-; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    ptrue p1.b, vl128
+; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    ld1b { z0.b }, p1/z, [x0]
+; CHECK-NEXT:    ld1b { z1.b }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.b, p0, z0.b, z1.b
+; CHECK-NEXT:    st1b { z0.b }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <128 x i8>, ptr %a
   %op2 = load volatile <128 x i8>, ptr %b
@@ -108,14 +108,14 @@ define void @select_v128i8(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v256i8(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v256i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
 ; CHECK-NEXT:    mov z0.b, w2
-; CHECK-NEXT:    ptrue p1.b
-; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x0]
-; CHECK-NEXT:    ld1b { z2.b }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.b, p1/z, z0.b, #0
-; CHECK-NEXT:    sel z0.b, p1, z1.b, z2.b
-; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    ptrue p1.b, vl256
+; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    ld1b { z0.b }, p1/z, [x0]
+; CHECK-NEXT:    ld1b { z1.b }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.b, p0, z0.b, z1.b
+; CHECK-NEXT:    st1b { z0.b }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <256 x i8>, ptr %a
   %op2 = load volatile <256 x i8>, ptr %b
@@ -153,15 +153,15 @@ define <8 x i16> @select_v8i16(<8 x i16> %op1, <8 x i16> %op2, i1 %mask) vscale_
 define void @select_v16i16(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl16
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <16 x i16>, ptr %a
   %op2 = load volatile <16 x i16>, ptr %b
@@ -173,33 +173,33 @@ define void @select_v16i16(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v32i16(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v32i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov z0.h, w2
+; VBITS_GE_256-NEXT:    ptrue p0.h
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
-; VBITS_GE_256-NEXT:    ptrue p1.h
+; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    and z0.h, z0.h, #0x1
-; VBITS_GE_256-NEXT:    ld1h { z1.h }, p0/z, [x0, x8, lsl #1]
-; VBITS_GE_256-NEXT:    ld1h { z2.h }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1h { z3.h }, p0/z, [x1, x8, lsl #1]
-; VBITS_GE_256-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; VBITS_GE_256-NEXT:    ld1h { z0.h }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.h, p1, z1.h, z3.h
-; VBITS_GE_256-NEXT:    mov z0.h, p1/m, z2.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x0, x8, lsl #1]
-; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; VBITS_GE_256-NEXT:    ld1h { z0.h }, p1/z, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    ld1h { z1.h }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1h { z2.h }, p1/z, [x1, x8, lsl #1]
+; VBITS_GE_256-NEXT:    ld1h { z3.h }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.h, p0, z0.h, z2.h
+; VBITS_GE_256-NEXT:    sel z1.h, p0, z1.h, z3.h
+; VBITS_GE_256-NEXT:    st1h { z0.h }, p1, [x0, x8, lsl #1]
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v32i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    mov z0.h, w2
-; VBITS_GE_512-NEXT:    ptrue p1.h
+; VBITS_GE_512-NEXT:    ptrue p0.h
+; VBITS_GE_512-NEXT:    ptrue p1.h, vl32
 ; VBITS_GE_512-NEXT:    and z0.h, z0.h, #0x1
-; VBITS_GE_512-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; VBITS_GE_512-NEXT:    sel z0.h, p1, z1.h, z2.h
-; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x0]
+; VBITS_GE_512-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; VBITS_GE_512-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.h, p0, z0.h, z1.h
+; VBITS_GE_512-NEXT:    st1h { z0.h }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <32 x i16>, ptr %a
   %op2 = load volatile <32 x i16>, ptr %b
@@ -211,15 +211,15 @@ define void @select_v32i16(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v64i16(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v64i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl64
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <64 x i16>, ptr %a
   %op2 = load volatile <64 x i16>, ptr %b
@@ -231,15 +231,15 @@ define void @select_v64i16(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v128i16(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v128i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov z0.h, w2
-; CHECK-NEXT:    ptrue p1.h
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ptrue p1.h, vl128
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.h, p1/z, z0.h, #0
-; CHECK-NEXT:    sel z0.h, p1, z1.h, z2.h
-; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
+; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    ld1h { z0.h }, p1/z, [x0]
+; CHECK-NEXT:    ld1h { z1.h }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <128 x i16>, ptr %a
   %op2 = load volatile <128 x i16>, ptr %b
@@ -277,15 +277,15 @@ define <4 x i32> @select_v4i32(<4 x i32> %op1, <4 x i32> %op2, i1 %mask) vscale_
 define void @select_v8i32(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl8
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <8 x i32>, ptr %a
   %op2 = load volatile <8 x i32>, ptr %b
@@ -297,33 +297,33 @@ define void @select_v8i32(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v16i32(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v16i32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    and w8, w2, #0x1
-; VBITS_GE_256-NEXT:    ptrue p1.s
+; VBITS_GE_256-NEXT:    ptrue p0.s
 ; VBITS_GE_256-NEXT:    mov z0.s, w8
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0, x8, lsl #2]
-; VBITS_GE_256-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1w { z3.s }, p0/z, [x1, x8, lsl #2]
-; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.s, p1, z1.s, z3.s
-; VBITS_GE_256-NEXT:    mov z0.s, p1/m, z2.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0, x8, lsl #2]
-; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; VBITS_GE_256-NEXT:    ld1w { z0.s }, p1/z, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    ld1w { z1.s }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1w { z2.s }, p1/z, [x1, x8, lsl #2]
+; VBITS_GE_256-NEXT:    ld1w { z3.s }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.s, p0, z0.s, z2.s
+; VBITS_GE_256-NEXT:    sel z1.s, p0, z1.s, z3.s
+; VBITS_GE_256-NEXT:    st1w { z0.s }, p1, [x0, x8, lsl #2]
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v16i32:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    and w8, w2, #0x1
-; VBITS_GE_512-NEXT:    ptrue p1.s
+; VBITS_GE_512-NEXT:    ptrue p0.s
 ; VBITS_GE_512-NEXT:    mov z0.s, w8
-; VBITS_GE_512-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; VBITS_GE_512-NEXT:    sel z0.s, p1, z1.s, z2.s
-; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x0]
+; VBITS_GE_512-NEXT:    ptrue p1.s, vl16
+; VBITS_GE_512-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; VBITS_GE_512-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.s, p0, z0.s, z1.s
+; VBITS_GE_512-NEXT:    st1w { z0.s }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <16 x i32>, ptr %a
   %op2 = load volatile <16 x i32>, ptr %b
@@ -335,15 +335,15 @@ define void @select_v16i32(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v32i32(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v32i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl32
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <32 x i32>, ptr %a
   %op2 = load volatile <32 x i32>, ptr %b
@@ -355,15 +355,15 @@ define void @select_v32i32(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v64i32(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v64i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    and w8, w2, #0x1
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.s, p1/z, z0.s, #0
-; CHECK-NEXT:    sel z0.s, p1, z1.s, z2.s
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.s, vl64
+; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    ld1w { z0.s }, p1/z, [x0]
+; CHECK-NEXT:    ld1w { z1.s }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <64 x i32>, ptr %a
   %op2 = load volatile <64 x i32>, ptr %b
@@ -401,16 +401,16 @@ define <2 x i64> @select_v2i64(<2 x i64> %op1, <2 x i64> %op2, i1 %mask) vscale_
 define void @select_v4i64(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 ; CHECK-LABEL: select_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl4
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <4 x i64>, ptr %a
   %op2 = load volatile <4 x i64>, ptr %b
@@ -422,35 +422,35 @@ define void @select_v4i64(ptr %a, ptr %b, i1 %mask) vscale_range(2,0) #0 {
 define void @select_v8i64(ptr %a, ptr %b, i1 %mask) #0 {
 ; VBITS_GE_256-LABEL: select_v8i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; VBITS_GE_256-NEXT:    and x8, x2, #0x1
-; VBITS_GE_256-NEXT:    ptrue p1.d
+; VBITS_GE_256-NEXT:    ptrue p0.d
 ; VBITS_GE_256-NEXT:    mov z0.d, x8
+; VBITS_GE_256-NEXT:    ptrue p1.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sel z1.d, p1, z1.d, z3.d
-; VBITS_GE_256-NEXT:    mov z0.d, p1/m, z2.d
-; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0]
+; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [x0]
+; VBITS_GE_256-NEXT:    ld1d { z2.d }, p1/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z3.d }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    sel z0.d, p0, z0.d, z2.d
+; VBITS_GE_256-NEXT:    sel z1.d, p0, z1.d, z3.d
+; VBITS_GE_256-NEXT:    st1d { z0.d }, p1, [x0, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1d { z1.d }, p1, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: select_v8i64:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; VBITS_GE_512-NEXT:    and x8, x2, #0x1
-; VBITS_GE_512-NEXT:    ptrue p1.d
+; VBITS_GE_512-NEXT:    ptrue p0.d
 ; VBITS_GE_512-NEXT:    mov z0.d, x8
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; VBITS_GE_512-NEXT:    sel z0.d, p1, z1.d, z2.d
-; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
+; VBITS_GE_512-NEXT:    ptrue p1.d, vl8
+; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; VBITS_GE_512-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; VBITS_GE_512-NEXT:    sel z0.d, p0, z0.d, z1.d
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %op1 = load volatile <8 x i64>, ptr %a
   %op2 = load volatile <8 x i64>, ptr %b
@@ -462,16 +462,16 @@ define void @select_v8i64(ptr %a, ptr %b, i1 %mask) #0 {
 define void @select_v16i64(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 ; CHECK-LABEL: select_v16i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl16
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <16 x i64>, ptr %a
   %op2 = load volatile <16 x i64>, ptr %b
@@ -483,16 +483,16 @@ define void @select_v16i64(ptr %a, ptr %b, i1 %mask) vscale_range(8,0) #0 {
 define void @select_v32i64(ptr %a, ptr %b, i1 %mask) vscale_range(16,0) #0 {
 ; CHECK-LABEL: select_v32i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p1.d, p1/z, z0.d, #0
-; CHECK-NEXT:    sel z0.d, p1, z1.d, z2.d
-; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
+; CHECK-NEXT:    ptrue p1.d, vl32
+; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [x0]
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
+; CHECK-NEXT:    st1d { z0.d }, p1, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load volatile <32 x i64>, ptr %a
   %op2 = load volatile <32 x i64>, ptr %b
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-to-fp.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-to-fp.ll
index 50040eaa61e6c..5bb012ae57503 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-int-to-fp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-int-to-fp.ll
@@ -131,8 +131,8 @@ define <4 x float> @ucvtf_v4i16_v4f32(<4 x i16> %op1) vscale_range(2,0) #0 {
 define void @ucvtf_v8i16_v8f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: ucvtf_v8i16_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -354,7 +354,6 @@ define void @ucvtf_v16i32_v16f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ucvtf z0.h, p0/m, z0.s
@@ -363,7 +362,8 @@ define void @ucvtf_v16i32_v16f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ucvtf_v16i32_v16f16:
@@ -535,8 +535,8 @@ define <2 x double> @ucvtf_v2i32_v2f64(<2 x i32> %op1) vscale_range(2,0) #0 {
 define void @ucvtf_v4i32_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: ucvtf_v4i32_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
@@ -759,7 +759,6 @@ define void @ucvtf_v8i64_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ucvtf z0.s, p0/m, z0.d
@@ -768,7 +767,8 @@ define void @ucvtf_v8i64_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: ucvtf_v8i64_v8f32:
@@ -1038,8 +1038,8 @@ define <4 x float> @scvtf_v4i16_v4f32(<4 x i16> %op1) vscale_range(2,0) #0 {
 define void @scvtf_v8i16_v8f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: scvtf_v8i16_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -1273,7 +1273,6 @@ define void @scvtf_v16i32_v16f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    scvtf z0.h, p0/m, z0.s
@@ -1282,7 +1281,8 @@ define void @scvtf_v16i32_v16f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: scvtf_v16i32_v16f16:
@@ -1454,8 +1454,8 @@ define <2 x double> @scvtf_v2i32_v2f64(<2 x i32> %op1) vscale_range(2,0) #0 {
 define void @scvtf_v4i32_v4f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: scvtf_v4i32_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
@@ -1684,7 +1684,6 @@ define void @scvtf_v8i64_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    scvtf z0.s, p0/m, z0.d
@@ -1693,7 +1692,8 @@ define void @scvtf_v8i64_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: scvtf_v8i64_v8f32:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-mask-opt.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-mask-opt.ll
index e23151475014d..f2ad98f8caec9 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-mask-opt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-mask-opt.ll
@@ -246,16 +246,16 @@ define void @masked_gather_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    ld1w { z0.d }, p0/z, [z0.d]
 ; VBITS_GE_256-NEXT:    ld1w { z1.d }, p0/z, [z1.d]
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x0]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_gather_v8i32:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-loads.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-loads.ll
index 67a53d4e15f3b..55d37d1bda5e4 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-loads.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-loads.ll
@@ -11,8 +11,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define <16 x i8> @masked_load_v16i8(ptr %src, <16 x i1> %mask) {
 ; CHECK-LABEL: masked_load_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    shl v0.16b, v0.16b, #7
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    cmlt v0.16b, v0.16b, #0
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-stores.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-stores.ll
index bdd6ce0647016..1a19b77f53c67 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-128bit-stores.ll
@@ -11,12 +11,12 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @masked_store_v16i8(ptr %dst, <16 x i1> %mask) {
 ; CHECK-LABEL: masked_store_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    shl v0.16b, v0.16b, #7
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    cmlt v0.16b, v0.16b, #0
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
-; CHECK-NEXT:    st1b { z1.b }, p0, [x0]
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.v16i8(<16 x i8> zeroinitializer, ptr %dst, i32 8, <16 x i1> %mask)
   ret void
@@ -27,11 +27,11 @@ define void @masked_store_v8f16(ptr %dst, <8 x i1> %mask) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-NEXT:    ptrue p0.h, vl8
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-NEXT:    shl v0.8h, v0.8h, #15
 ; CHECK-NEXT:    cmlt v0.8h, v0.8h, #0
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
-; CHECK-NEXT:    st1h { z1.h }, p0, [x0]
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.v8f16(<8 x half> zeroinitializer, ptr %dst, i32 8, <8 x i1> %mask)
   ret void
@@ -42,11 +42,11 @@ define void @masked_store_v4f32(ptr %dst, <4 x i1> %mask) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-NEXT:    shl v0.4s, v0.4s, #31
 ; CHECK-NEXT:    cmlt v0.4s, v0.4s, #0
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
-; CHECK-NEXT:    st1w { z1.s }, p0, [x0]
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.v4f32(<4 x float> zeroinitializer, ptr %dst, i32 8, <4 x i1> %mask)
   ret void
@@ -57,11 +57,11 @@ define void @masked_store_v2f64(ptr %dst, <2 x i1> %mask) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ushll v0.2d, v0.2s, #0
 ; CHECK-NEXT:    ptrue p0.d, vl2
-; CHECK-NEXT:    movi v1.2d, #0000000000000000
 ; CHECK-NEXT:    shl v0.2d, v0.2d, #63
 ; CHECK-NEXT:    cmlt v0.2d, v0.2d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    st1d { z1.d }, p0, [x0]
+; CHECK-NEXT:    movi v0.2d, #0000000000000000
+; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.v2f64(<2 x double> zeroinitializer, ptr %dst, i32 8, <2 x i1> %mask)
   ret void
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll
index 92fce4584f6a9..27e95489f8ad7 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-gather.ll
@@ -41,11 +41,11 @@ define void @masked_gather_v4i8(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-NEXT:    cmeq v0.4h, v0.4h, #0
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1b { z0.d }, p1/z, [z1.d]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1b { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    st1b { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %cval = load <4 x i8>, ptr %a
@@ -65,7 +65,6 @@ define void @masked_gather_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    cmeq v0.8b, v0.8b, #0
 ; VBITS_GE_256-NEXT:    zip2 v1.8b, v0.8b, v0.8b
 ; VBITS_GE_256-NEXT:    zip1 v0.8b, v0.8b, v0.8b
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    shl v1.4h, v1.4h, #8
 ; VBITS_GE_256-NEXT:    shl v0.4h, v0.4h, #8
 ; VBITS_GE_256-NEXT:    sshr v1.4h, v1.4h, #8
@@ -75,10 +74,11 @@ define void @masked_gather_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
 ; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    ld1b { z1.d }, p1/z, [z2.d]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1b { z0.d }, p0/z, [z2.d]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1b { z1.d }, p1/z, [z1.d]
+; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    ld1b { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
@@ -93,11 +93,11 @@ define void @masked_gather_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    cmeq v0.8b, v0.8b, #0
 ; VBITS_GE_512-NEXT:    sunpklo z0.h, z0.b
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
 ; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_512-NEXT:    ld1b { z0.d }, p0/z, [z1.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    ld1b { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_512-NEXT:    uzp1 z0.b, z0.b, z0.b
@@ -118,11 +118,11 @@ define void @masked_gather_v16i8(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    cmeq v0.16b, v0.16b, #0
 ; CHECK-NEXT:    sunpklo z0.h, z0.b
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1b { z0.d }, p0/z, [z1.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1b { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
@@ -194,10 +194,10 @@ define void @masked_gather_v4i16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    cmeq v0.4h, v0.4h, #0
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    str d0, [x0]
@@ -219,15 +219,15 @@ define void @masked_gather_v8i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    cmeq v0.8h, v0.8h, #0
 ; VBITS_GE_256-NEXT:    sunpklo z1.s, z0.h
 ; VBITS_GE_256-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
+; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    ld1h { z1.d }, p1/z, [z2.d]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1h { z0.d }, p0/z, [z2.d]
+; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    ld1h { z1.d }, p1/z, [z1.d]
+; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
@@ -242,10 +242,10 @@ define void @masked_gather_v8i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    cmeq v0.8h, v0.8h, #0
 ; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_512-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_512-NEXT:    str q0, [x0]
@@ -332,9 +332,9 @@ define void @masked_gather_v4i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    cmeq v0.4s, v0.4s, #0
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z1.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1w { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    str q0, [x0]
 ; CHECK-NEXT:    ret
@@ -354,18 +354,18 @@ define void @masked_gather_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    ptrue p2.d, vl4
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p2/z, [x1]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p2/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    cmpeq p1.s, p0/z, z0.s, #0
-; VBITS_GE_256-NEXT:    punpklo p3.h, p1.b
 ; VBITS_GE_256-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    punpklo p1.h, p1.b
+; VBITS_GE_256-NEXT:    and p1.b, p1/z, p1.b, p2.b
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
-; VBITS_GE_256-NEXT:    and p1.b, p3/z, p3.b, p2.b
-; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    ld1w { z1.d }, p1/z, [z1.d]
+; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p2/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z2.d]
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p2/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
+; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z0.d]
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p1, z1.s, z0.s
 ; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
@@ -460,8 +460,8 @@ define void @masked_gather_v1i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @masked_gather_v2i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_gather_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    cmeq v0.2d, v0.2d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ldr q0, [x1]
@@ -481,9 +481,9 @@ define void @masked_gather_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %cval = load <4 x i64>, ptr %a
@@ -500,13 +500,13 @@ define void @masked_gather_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
-; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
+; VBITS_GE_256-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_256-NEXT:    cmpeq p1.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [z2.d]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p1/z, [z1.d]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    st1d { z1.d }, p0, [x0]
 ; VBITS_GE_256-NEXT:    ret
@@ -515,9 +515,9 @@ define void @masked_gather_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512:       // %bb.0:
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
-; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
 ; VBITS_GE_512-NEXT:    ret
   %cval = load <8 x i64>, ptr %a
@@ -533,9 +533,9 @@ define void @masked_gather_v16i64(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %cval = load <16 x i64>, ptr %a
@@ -551,9 +551,9 @@ define void @masked_gather_v32i64(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z1.d]
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1d { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %cval = load <32 x i64>, ptr %a
@@ -603,10 +603,10 @@ define void @masked_gather_v4f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    fcmeq v0.4h, v0.4h, #0.0
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    str d0, [x0]
@@ -626,17 +626,17 @@ define void @masked_gather_v8f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    fcmeq v0.8h, v0.8h, #0.0
-; VBITS_GE_256-NEXT:    sunpklo z2.s, z0.h
+; VBITS_GE_256-NEXT:    sunpklo z1.s, z0.h
 ; VBITS_GE_256-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
-; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
 ; VBITS_GE_256-NEXT:    sunpklo z0.s, z0.h
-; VBITS_GE_256-NEXT:    sunpklo z2.d, z2.s
+; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z1.d, #0
 ; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
-; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z2.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    ld1h { z1.d }, p1/z, [z1.d]
-; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1h { z0.d }, p0/z, [z2.d]
+; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
@@ -651,10 +651,10 @@ define void @masked_gather_v8f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    fcmeq v0.8h, v0.8h, #0.0
 ; VBITS_GE_512-NEXT:    sunpklo z0.s, z0.h
-; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z0.d, z0.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; VBITS_GE_512-NEXT:    ld1h { z0.d }, p0/z, [z1.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    ld1h { z0.d }, p1/z, [z0.d]
 ; VBITS_GE_512-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_512-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_512-NEXT:    str q0, [x0]
@@ -741,9 +741,9 @@ define void @masked_gather_v4f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    fcmeq v0.4s, v0.4s, #0.0
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z1.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
+; CHECK-NEXT:    ld1w { z0.d }, p1/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    str q0, [x0]
 ; CHECK-NEXT:    ret
@@ -763,18 +763,18 @@ define void @masked_gather_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    ptrue p2.d, vl4
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p2/z, [x1]
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p2/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
-; VBITS_GE_256-NEXT:    punpklo p3.h, p1.b
 ; VBITS_GE_256-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    punpklo p1.h, p1.b
+; VBITS_GE_256-NEXT:    and p1.b, p1/z, p1.b, p2.b
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
-; VBITS_GE_256-NEXT:    and p1.b, p3/z, p3.b, p2.b
-; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    ld1w { z1.d }, p1/z, [z1.d]
+; VBITS_GE_256-NEXT:    sunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p2/z, z0.d, #0
-; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z2.d]
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
+; VBITS_GE_256-NEXT:    ld1d { z0.d }, p2/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
+; VBITS_GE_256-NEXT:    ld1w { z0.d }, p1/z, [z0.d]
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p1, z1.s, z0.s
 ; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x0]
@@ -869,8 +869,8 @@ define void @masked_gather_v1f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @masked_gather_v2f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_gather_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcmeq v0.2d, v0.2d, #0.0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ldr q0, [x1]
@@ -1202,8 +1202,8 @@ define void @masked_gather_passthru(ptr %a, ptr %b, ptr %c) vscale_range(16,0) #
 ; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x2]
 ; CHECK-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    ld1d { z0.d }, p2/z, [x1]
-; CHECK-NEXT:    punpklo p3.h, p1.b
-; CHECK-NEXT:    ld1w { z0.d }, p3/z, [z0.d]
+; CHECK-NEXT:    punpklo p2.h, p1.b
+; CHECK-NEXT:    ld1w { z0.d }, p2/z, [z0.d]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-loads.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-loads.ll
index 467378e7da59b..c22d9e71c51a9 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-loads.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-loads.ll
@@ -35,9 +35,9 @@ define <2 x half> @masked_load_v2f16(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 define <2 x float> @masked_load_v2f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_load_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    ldr d0, [x0]
 ; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    fcmeq v0.2s, v0.2s, v1.2s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
@@ -53,9 +53,9 @@ define <2 x float> @masked_load_v2f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 define <4 x float> @masked_load_v4f32(ptr %ap, ptr %bp) vscale_range(1,0) #0 {
 ; CHECK-LABEL: masked_load_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcmeq v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0]
@@ -401,8 +401,8 @@ define void @masked_load_sext_v32i8i16(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_sext_v16i8i32(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_sext_v16i8i32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    ldr q0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    cmeq v0.16b, v0.16b, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.b, p0/z, z0.b, #0
@@ -436,8 +436,8 @@ define void @masked_load_sext_v16i8i32(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_sext_v8i8i64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_sext_v8i8i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl8
 ; VBITS_GE_256-NEXT:    ldr d0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    cmeq v0.8b, v0.8b, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.b, p0/z, z0.b, #0
@@ -504,8 +504,8 @@ define void @masked_load_sext_v16i16i32(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_sext_v8i16i64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_sext_v8i16i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    ldr q0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    cmeq v0.8h, v0.8h, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.h, p0/z, z0.h, #0
@@ -603,8 +603,8 @@ define void @masked_load_zext_v32i8i16(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_zext_v16i8i32(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_zext_v16i8i32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    ldr q0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    cmeq v0.16b, v0.16b, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.b, p0/z, z0.b, #0
@@ -638,8 +638,8 @@ define void @masked_load_zext_v16i8i32(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_zext_v8i8i64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_zext_v8i8i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl8
 ; VBITS_GE_256-NEXT:    ldr d0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    cmeq v0.8b, v0.8b, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.b, p0/z, z0.b, #0
@@ -706,8 +706,8 @@ define void @masked_load_zext_v16i16i32(ptr %ap, ptr %bp, ptr %c) #0 {
 define void @masked_load_zext_v8i16i64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-LABEL: masked_load_zext_v8i16i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    ldr q0, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    cmeq v0.8h, v0.8h, #0
 ; VBITS_GE_256-NEXT:    cmpne p0.h, p0/z, z0.h, #0
@@ -782,11 +782,11 @@ define void @masked_load_sext_v32i8i16_m16(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    mov z0.h, p1/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    mov z1.h, p2/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p1.b, vl16
-; VBITS_GE_256-NEXT:    ptrue p2.b, vl32
 ; VBITS_GE_256-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; VBITS_GE_256-NEXT:    uzp1 z1.b, z1.b, z1.b
 ; VBITS_GE_256-NEXT:    splice z1.b, p1, z1.b, z0.b
-; VBITS_GE_256-NEXT:    cmpne p1.b, p2/z, z1.b, #0
+; VBITS_GE_256-NEXT:    ptrue p1.b, vl32
+; VBITS_GE_256-NEXT:    cmpne p1.b, p1/z, z1.b, #0
 ; VBITS_GE_256-NEXT:    ld1b { z0.b }, p1/z, [x0]
 ; VBITS_GE_256-NEXT:    sunpklo z1.h, z0.b
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
@@ -1000,11 +1000,11 @@ define void @masked_load_sext_v8i32i64_m64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    mov z0.d, p1/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    mov z1.d, p2/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
-; VBITS_GE_256-NEXT:    ptrue p2.s, vl8
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p1, z1.s, z0.s
-; VBITS_GE_256-NEXT:    cmpne p1.s, p2/z, z1.s, #0
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
+; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z1.s, #0
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p1/z, [x0]
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z0.s
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
@@ -1041,11 +1041,11 @@ define void @masked_load_zext_v32i8i16_m16(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    mov z0.h, p1/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    mov z1.h, p2/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p1.b, vl16
-; VBITS_GE_256-NEXT:    ptrue p2.b, vl32
 ; VBITS_GE_256-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; VBITS_GE_256-NEXT:    uzp1 z1.b, z1.b, z1.b
 ; VBITS_GE_256-NEXT:    splice z1.b, p1, z1.b, z0.b
-; VBITS_GE_256-NEXT:    cmpne p1.b, p2/z, z1.b, #0
+; VBITS_GE_256-NEXT:    ptrue p1.b, vl32
+; VBITS_GE_256-NEXT:    cmpne p1.b, p1/z, z1.b, #0
 ; VBITS_GE_256-NEXT:    ld1b { z0.b }, p1/z, [x0]
 ; VBITS_GE_256-NEXT:    uunpklo z1.h, z0.b
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
@@ -1259,11 +1259,11 @@ define void @masked_load_zext_v8i32i64_m64(ptr %ap, ptr %bp, ptr %c) #0 {
 ; VBITS_GE_256-NEXT:    mov z0.d, p1/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    mov z1.d, p2/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p1.s, vl4
-; VBITS_GE_256-NEXT:    ptrue p2.s, vl8
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p1, z1.s, z0.s
-; VBITS_GE_256-NEXT:    cmpne p1.s, p2/z, z1.s, #0
+; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
+; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z1.s, #0
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p1/z, [x0]
 ; VBITS_GE_256-NEXT:    uunpklo z1.d, z0.s
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-scatter.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-scatter.ll
index e2d341c22efc2..e3e06dcdf17f3 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-scatter.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-scatter.ll
@@ -39,12 +39,12 @@ define void @masked_scatter_v4i8(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-NEXT:    cmeq v1.4h, v0.4h, #0
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1b { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1b { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x i8>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -61,11 +61,11 @@ define void @masked_scatter_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    cmeq v1.8b, v0.8b, #0
 ; VBITS_GE_256-NEXT:    zip1 v3.8b, v0.8b, v0.8b
+; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    zip1 v2.8b, v1.8b, v0.8b
 ; VBITS_GE_256-NEXT:    zip2 v1.8b, v1.8b, v0.8b
 ; VBITS_GE_256-NEXT:    zip2 v0.8b, v0.8b, v0.8b
 ; VBITS_GE_256-NEXT:    uunpklo z3.s, z3.h
-; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    shl v2.4h, v2.4h, #8
 ; VBITS_GE_256-NEXT:    shl v1.4h, v1.4h, #8
 ; VBITS_GE_256-NEXT:    uunpklo z0.s, z0.h
@@ -78,10 +78,10 @@ define void @masked_scatter_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    sunpklo z2.d, z2.s
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z2.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    st1b { z3.d }, p1, [z4.d]
-; VBITS_GE_256-NEXT:    st1b { z0.d }, p0, [z2.d]
+; VBITS_GE_256-NEXT:    st1b { z3.d }, p1, [z2.d]
+; VBITS_GE_256-NEXT:    st1b { z0.d }, p0, [z4.d]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_scatter_v8i8:
@@ -92,12 +92,12 @@ define void @masked_scatter_v8i8(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    uunpklo z0.h, z0.b
 ; VBITS_GE_512-NEXT:    sunpklo z1.h, z1.b
 ; VBITS_GE_512-NEXT:    uunpklo z0.s, z0.h
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_512-NEXT:    uunpklo z0.d, z0.s
 ; VBITS_GE_512-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_512-NEXT:    st1b { z0.d }, p0, [z2.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    st1b { z0.d }, p1, [z1.d]
 ; VBITS_GE_512-NEXT:    ret
   %vals = load <8 x i8>, ptr %a
   %ptrs = load <8 x ptr>, ptr %b
@@ -115,12 +115,12 @@ define void @masked_scatter_v16i8(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    sunpklo z1.h, z1.b
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1b { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1b { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <16 x i8>, ptr %a
   %ptrs = load <16 x ptr>, ptr %b
@@ -135,13 +135,13 @@ define void @masked_scatter_v32i8(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    ptrue p0.b, vl32
 ; CHECK-NEXT:    ptrue p1.d, vl32
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [x0]
-; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
+; CHECK-NEXT:    uunpklo z1.h, z0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
-; CHECK-NEXT:    uunpklo z0.h, z0.b
+; CHECK-NEXT:    uunpklo z0.s, z1.h
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    uunpklo z0.d, z0.s
+; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
 ; CHECK-NEXT:    punpklo p0.h, p0.b
+; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1b { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
@@ -187,10 +187,10 @@ define void @masked_scatter_v4i16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1h { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1h { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x i16>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -208,20 +208,20 @@ define void @masked_scatter_v8i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    cmeq v1.8h, v0.8h, #0
 ; VBITS_GE_256-NEXT:    uunpklo z3.s, z0.h
 ; VBITS_GE_256-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    sunpklo z2.s, z1.h
 ; VBITS_GE_256-NEXT:    ext v1.16b, v1.16b, v1.16b, #8
 ; VBITS_GE_256-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    uunpklo z3.d, z3.s
-; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_256-NEXT:    sunpklo z2.d, z2.s
 ; VBITS_GE_256-NEXT:    uunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z2.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    st1h { z3.d }, p1, [z4.d]
+; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    st1h { z0.d }, p0, [z2.d]
+; VBITS_GE_256-NEXT:    st1h { z3.d }, p1, [z2.d]
+; VBITS_GE_256-NEXT:    st1h { z0.d }, p0, [z4.d]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_scatter_v8i16:
@@ -232,10 +232,10 @@ define void @masked_scatter_v8i16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_512-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_512-NEXT:    uunpklo z0.d, z0.s
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_512-NEXT:    st1h { z0.d }, p0, [z2.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    st1h { z0.d }, p1, [z1.d]
 ; VBITS_GE_512-NEXT:    ret
   %vals = load <8 x i16>, ptr %a
   %ptrs = load <8 x ptr>, ptr %b
@@ -253,9 +253,9 @@ define void @masked_scatter_v16i16(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    punpklo p0.h, p0.b
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <16 x i16>, ptr %a
@@ -274,9 +274,9 @@ define void @masked_scatter_v32i16(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    punpklo p0.h, p0.b
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <32 x i16>, ptr %a
@@ -317,9 +317,9 @@ define void @masked_scatter_v4i32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    cmeq v1.4s, v0.4s, #0
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1w { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1w { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x i32>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -333,21 +333,21 @@ define void @masked_scatter_v8i32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
+; VBITS_GE_256-NEXT:    ptrue p1.d, vl4
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    cmpeq p1.s, p0/z, z0.s, #0
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
+; VBITS_GE_256-NEXT:    ld1d { z3.d }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    ld1d { z4.d }, p1/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
 ; VBITS_GE_256-NEXT:    uunpklo z2.d, z0.s
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
-; VBITS_GE_256-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
-; VBITS_GE_256-NEXT:    punpklo p2.h, p1.b
 ; VBITS_GE_256-NEXT:    uunpklo z0.d, z0.s
+; VBITS_GE_256-NEXT:    mov z1.s, p0/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    punpklo p0.h, p0.b
+; VBITS_GE_256-NEXT:    and p0.b, p0/z, p0.b, p1.b
 ; VBITS_GE_256-NEXT:    ext z1.b, z1.b, z1.b, #16
-; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    st1w { z2.d }, p0, [z3.d]
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_256-NEXT:    and p1.b, p2/z, p2.b, p0.b
-; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    st1w { z2.d }, p1, [z3.d]
+; VBITS_GE_256-NEXT:    cmpne p0.d, p1/z, z1.d, #0
 ; VBITS_GE_256-NEXT:    st1w { z0.d }, p0, [z4.d]
 ; VBITS_GE_256-NEXT:    ret
 ;
@@ -434,8 +434,8 @@ define void @masked_scatter_v1i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 define void @masked_scatter_v2i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_scatter_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    cmeq v1.2d, v0.2d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    ldr q1, [x1]
@@ -454,8 +454,8 @@ define void @masked_scatter_v4i64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    st1d { z0.d }, p0, [z1.d]
+; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    st1d { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x i64>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -484,8 +484,8 @@ define void @masked_scatter_v8i64(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; VBITS_GE_512-NEXT:    cmpeq p0.d, p0/z, z0.d, #0
-; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [z1.d]
+; VBITS_GE_512-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
+; VBITS_GE_512-NEXT:    st1d { z0.d }, p1, [z1.d]
 ; VBITS_GE_512-NEXT:    ret
   %vals = load <8 x i64>, ptr %a
   %ptrs = load <8 x ptr>, ptr %b
@@ -500,8 +500,8 @@ define void @masked_scatter_v16i64(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    st1d { z0.d }, p0, [z1.d]
+; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    st1d { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <16 x i64>, ptr %a
   %ptrs = load <16 x ptr>, ptr %b
@@ -516,8 +516,8 @@ define void @masked_scatter_v32i64(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    st1d { z0.d }, p0, [z1.d]
+; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, #0
+; CHECK-NEXT:    st1d { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <32 x i64>, ptr %a
   %ptrs = load <32 x ptr>, ptr %b
@@ -539,15 +539,15 @@ define void @masked_scatter_v2f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    fcmeq v2.4h, v1.4h, #0.0
 ; CHECK-NEXT:    uunpklo z1.s, z1.h
 ; CHECK-NEXT:    sshll v2.4s, v2.4h, #0
-; CHECK-NEXT:    uunpklo z1.d, z1.s
 ; CHECK-NEXT:    mov v0.h[0], v2.h[0]
 ; CHECK-NEXT:    mov w8, v2.s[1]
 ; CHECK-NEXT:    mov v0.h[1], w8
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    ldr q0, [x1]
-; CHECK-NEXT:    st1h { z1.d }, p0, [z0.d]
+; CHECK-NEXT:    uunpklo z0.d, z1.s
+; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    st1h { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <2 x half>, ptr %a
   %ptrs = load <2 x ptr>, ptr %b
@@ -565,10 +565,10 @@ define void @masked_scatter_v4f16(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1h { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1h { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x half>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -586,20 +586,20 @@ define void @masked_scatter_v8f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256-NEXT:    fcmeq v1.8h, v0.8h, #0.0
 ; VBITS_GE_256-NEXT:    uunpklo z3.s, z0.h
 ; VBITS_GE_256-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    sunpklo z2.s, z1.h
 ; VBITS_GE_256-NEXT:    ext v1.16b, v1.16b, v1.16b, #8
 ; VBITS_GE_256-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_256-NEXT:    uunpklo z3.d, z3.s
-; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_256-NEXT:    sunpklo z2.d, z2.s
 ; VBITS_GE_256-NEXT:    uunpklo z0.d, z0.s
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
 ; VBITS_GE_256-NEXT:    cmpne p1.d, p0/z, z2.d, #0
-; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1, x8, lsl #3]
-; VBITS_GE_256-NEXT:    st1h { z3.d }, p1, [z4.d]
+; VBITS_GE_256-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    st1h { z0.d }, p0, [z2.d]
+; VBITS_GE_256-NEXT:    st1h { z3.d }, p1, [z2.d]
+; VBITS_GE_256-NEXT:    st1h { z0.d }, p0, [z4.d]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_scatter_v8f16:
@@ -610,10 +610,10 @@ define void @masked_scatter_v8f16(ptr %a, ptr %b) #0 {
 ; VBITS_GE_512-NEXT:    uunpklo z0.s, z0.h
 ; VBITS_GE_512-NEXT:    sunpklo z1.s, z1.h
 ; VBITS_GE_512-NEXT:    uunpklo z0.d, z0.s
-; VBITS_GE_512-NEXT:    ld1d { z2.d }, p0/z, [x1]
 ; VBITS_GE_512-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_512-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_512-NEXT:    st1h { z0.d }, p0, [z2.d]
+; VBITS_GE_512-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; VBITS_GE_512-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; VBITS_GE_512-NEXT:    st1h { z0.d }, p1, [z1.d]
 ; VBITS_GE_512-NEXT:    ret
   %vals = load <8 x half>, ptr %a
   %ptrs = load <8 x ptr>, ptr %b
@@ -631,9 +631,9 @@ define void @masked_scatter_v16f16(ptr %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    punpklo p0.h, p0.b
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <16 x half>, ptr %a
@@ -652,9 +652,9 @@ define void @masked_scatter_v32f16(ptr %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x1]
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    punpklo p0.h, p0.b
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z0.d }, p0, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <32 x half>, ptr %a
@@ -695,9 +695,9 @@ define void @masked_scatter_v4f32(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    fcmeq v1.4s, v0.4s, #0.0
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x1]
-; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; CHECK-NEXT:    st1w { z0.d }, p0, [z2.d]
+; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
+; CHECK-NEXT:    ld1d { z1.d }, p0/z, [x1]
+; CHECK-NEXT:    st1w { z0.d }, p1, [z1.d]
 ; CHECK-NEXT:    ret
   %vals = load <4 x float>, ptr %a
   %ptrs = load <4 x ptr>, ptr %b
@@ -711,21 +711,21 @@ define void @masked_scatter_v8f32(ptr %a, ptr %b) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
+; VBITS_GE_256-NEXT:    ptrue p1.d, vl4
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0]
-; VBITS_GE_256-NEXT:    fcmeq p1.s, p0/z, z0.s, #0.0
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
+; VBITS_GE_256-NEXT:    ld1d { z3.d }, p1/z, [x1]
+; VBITS_GE_256-NEXT:    ld1d { z4.d }, p1/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
 ; VBITS_GE_256-NEXT:    uunpklo z2.d, z0.s
 ; VBITS_GE_256-NEXT:    ext z0.b, z0.b, z0.b, #16
-; VBITS_GE_256-NEXT:    mov z1.s, p1/z, #-1 // =0xffffffffffffffff
-; VBITS_GE_256-NEXT:    punpklo p2.h, p1.b
 ; VBITS_GE_256-NEXT:    uunpklo z0.d, z0.s
-; VBITS_GE_256-NEXT:    ld1d { z3.d }, p0/z, [x1]
-; VBITS_GE_256-NEXT:    ld1d { z4.d }, p0/z, [x1, x8, lsl #3]
+; VBITS_GE_256-NEXT:    mov z1.s, p0/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    punpklo p0.h, p0.b
+; VBITS_GE_256-NEXT:    and p0.b, p0/z, p0.b, p1.b
 ; VBITS_GE_256-NEXT:    ext z1.b, z1.b, z1.b, #16
+; VBITS_GE_256-NEXT:    st1w { z2.d }, p0, [z3.d]
 ; VBITS_GE_256-NEXT:    sunpklo z1.d, z1.s
-; VBITS_GE_256-NEXT:    and p1.b, p2/z, p2.b, p0.b
-; VBITS_GE_256-NEXT:    cmpne p0.d, p0/z, z1.d, #0
-; VBITS_GE_256-NEXT:    st1w { z2.d }, p1, [z3.d]
+; VBITS_GE_256-NEXT:    cmpne p0.d, p1/z, z1.d, #0
 ; VBITS_GE_256-NEXT:    st1w { z0.d }, p0, [z4.d]
 ; VBITS_GE_256-NEXT:    ret
 ;
@@ -812,8 +812,8 @@ define void @masked_scatter_v1f64(ptr %a, ptr %b) vscale_range(8,0) #0 {
 define void @masked_scatter_v2f64(ptr %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_scatter_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcmeq v1.2d, v0.2d, #0.0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    ldr q1, [x1]
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-stores.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-stores.ll
index 68fb4cc6afb09..b0d4f79aea110 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-masked-stores.ll
@@ -34,9 +34,9 @@ define void @masked_store_v2f16(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 define void @masked_store_v2f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-LABEL: masked_store_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    ldr d0, [x0]
 ; CHECK-NEXT:    ldr d1, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    fcmeq v1.2s, v0.2s, v1.2s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -51,9 +51,9 @@ define void @masked_store_v2f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 define void @masked_store_v4f32(ptr %ap, ptr %bp) vscale_range(1,0) #0 {
 ; CHECK-LABEL: masked_store_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcmeq v1.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
@@ -161,11 +161,11 @@ define void @masked_store_trunc_v8i64i8(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    mov z3.d, p0/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z2.s, z2.s, z2.s
-; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
+; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    splice z3.s, p0, z3.s, z2.s
-; VBITS_GE_256-NEXT:    cmpne p0.s, p1/z, z3.s, #0
-; VBITS_GE_256-NEXT:    st1b { z1.s }, p0, [x2]
+; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z3.s, #0
+; VBITS_GE_256-NEXT:    st1b { z1.s }, p1, [x2]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_store_trunc_v8i64i8:
@@ -197,14 +197,14 @@ define void @masked_store_trunc_v8i64i16(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    cmpeq p0.d, p0/z, z1.d, z3.d
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
+; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    mov z2.d, p1/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    mov z3.d, p0/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
-; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
-; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    uzp1 z2.s, z2.s, z2.s
-; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    mov v1.d[1], v0.d[0]
+; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    splice z3.s, p0, z3.s, z2.s
 ; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    uzp1 z2.h, z3.h, z3.h
@@ -246,11 +246,11 @@ define void @masked_store_trunc_v8i64i32(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    mov z3.d, p0/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z2.s, z2.s, z2.s
-; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
+; VBITS_GE_256-NEXT:    uzp1 z3.s, z3.s, z3.s
 ; VBITS_GE_256-NEXT:    splice z3.s, p0, z3.s, z2.s
-; VBITS_GE_256-NEXT:    cmpne p0.s, p1/z, z3.s, #0
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x2]
+; VBITS_GE_256-NEXT:    cmpne p1.s, p1/z, z3.s, #0
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x2]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_store_trunc_v8i64i32:
@@ -282,14 +282,14 @@ define void @masked_store_trunc_v16i32i8(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    cmpeq p0.s, p0/z, z1.s, z3.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
+; VBITS_GE_256-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; VBITS_GE_256-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
+; VBITS_GE_256-NEXT:    uzp1 z1.b, z1.b, z1.b
 ; VBITS_GE_256-NEXT:    mov z3.s, p0/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
-; VBITS_GE_256-NEXT:    uzp1 z0.b, z0.b, z0.b
-; VBITS_GE_256-NEXT:    uzp1 z1.b, z1.b, z1.b
 ; VBITS_GE_256-NEXT:    uzp1 z2.h, z2.h, z2.h
-; VBITS_GE_256-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; VBITS_GE_256-NEXT:    mov v1.d[1], v0.d[0]
+; VBITS_GE_256-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; VBITS_GE_256-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; VBITS_GE_256-NEXT:    uzp1 z3.b, z3.b, z3.b
 ; VBITS_GE_256-NEXT:    mov v3.d[1], v2.d[0]
@@ -327,17 +327,17 @@ define void @masked_store_trunc_v16i32i16(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    cmpeq p0.s, p0/z, z1.s, z3.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    mov z2.s, p1/z, #-1 // =0xffffffffffffffff
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
+; VBITS_GE_256-NEXT:    ptrue p1.h, vl8
 ; VBITS_GE_256-NEXT:    mov z3.s, p0/z, #-1 // =0xffffffffffffffff
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    splice z1.h, p1, z1.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; VBITS_GE_256-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; VBITS_GE_256-NEXT:    uzp1 z2.b, z2.b, z2.b
-; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z3.b, z3.b, z3.b
 ; VBITS_GE_256-NEXT:    mov v3.d[1], v2.d[0]
 ; VBITS_GE_256-NEXT:    sunpklo z2.h, z3.b
-; VBITS_GE_256-NEXT:    cmpne p0.h, p1/z, z2.h, #0
+; VBITS_GE_256-NEXT:    cmpne p0.h, p0/z, z2.h, #0
 ; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x2]
 ; VBITS_GE_256-NEXT:    ret
 ;
@@ -375,11 +375,11 @@ define void @masked_store_trunc_v32i16i8(ptr %ap, ptr %bp, ptr %dest) #0 {
 ; VBITS_GE_256-NEXT:    mov z3.h, p0/z, #-1 // =0xffffffffffffffff
 ; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    uzp1 z2.b, z2.b, z2.b
-; VBITS_GE_256-NEXT:    uzp1 z3.b, z3.b, z3.b
 ; VBITS_GE_256-NEXT:    splice z1.b, p0, z1.b, z0.b
+; VBITS_GE_256-NEXT:    uzp1 z3.b, z3.b, z3.b
 ; VBITS_GE_256-NEXT:    splice z3.b, p0, z3.b, z2.b
-; VBITS_GE_256-NEXT:    cmpne p0.b, p1/z, z3.b, #0
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x2]
+; VBITS_GE_256-NEXT:    cmpne p1.b, p1/z, z3.b, #0
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p1, [x2]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: masked_store_trunc_v32i16i8:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-shuffles.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-shuffles.ll
index a5303c901b80f..fb169491b0c90 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-shuffles.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-shuffles.ll
@@ -64,7 +64,6 @@ define void @crash_when_lowering_extract_shuffle(ptr %dst, i1 %cond) vscale_rang
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    mov x8, #16 // =0x10
 ; CHECK-NEXT:    mov x10, #8 // =0x8
-; CHECK-NEXT:    ld1w { z4.s }, p0/z, [x0, x8, lsl #2]
 ; CHECK-NEXT:    mov v2.b[7], w11
 ; CHECK-NEXT:    mov v1.b[7], w9
 ; CHECK-NEXT:    uunpklo z3.h, z3.b
@@ -86,22 +85,23 @@ define void @crash_when_lowering_extract_shuffle(ptr %dst, i1 %cond) vscale_rang
 ; CHECK-NEXT:    asr z2.s, z2.s, #31
 ; CHECK-NEXT:    asr z1.s, z1.s, #31
 ; CHECK-NEXT:    cmpne p1.s, p0/z, z0.s, #0
-; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0, x9, lsl #2]
+; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; CHECK-NEXT:    cmpne p2.s, p0/z, z3.s, #0
-; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x0, x10, lsl #2]
+; CHECK-NEXT:    ld1w { z3.s }, p0/z, [x0, x9, lsl #2]
 ; CHECK-NEXT:    and z2.s, z2.s, #0x1
 ; CHECK-NEXT:    and z1.s, z1.s, #0x1
-; CHECK-NEXT:    mov z4.s, p1/m, #0 // =0x0
-; CHECK-NEXT:    mov z0.s, p2/m, #0 // =0x0
+; CHECK-NEXT:    mov z0.s, p1/m, #0 // =0x0
 ; CHECK-NEXT:    cmpne p3.s, p0/z, z2.s, #0
-; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0]
-; CHECK-NEXT:    cmpne p1.s, p0/z, z1.s, #0
-; CHECK-NEXT:    st1w { z4.s }, p0, [x0, x8, lsl #2]
-; CHECK-NEXT:    st1w { z0.s }, p0, [x0, x9, lsl #2]
-; CHECK-NEXT:    mov z3.s, p3/m, #0 // =0x0
-; CHECK-NEXT:    mov z2.s, p1/m, #0 // =0x0
-; CHECK-NEXT:    st1w { z3.s }, p0, [x0, x10, lsl #2]
-; CHECK-NEXT:    st1w { z2.s }, p0, [x0]
+; CHECK-NEXT:    cmpne p4.s, p0/z, z1.s, #0
+; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0, x10, lsl #2]
+; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0]
+; CHECK-NEXT:    mov z3.s, p2/m, #0 // =0x0
+; CHECK-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
+; CHECK-NEXT:    mov z2.s, p3/m, #0 // =0x0
+; CHECK-NEXT:    mov z1.s, p4/m, #0 // =0x0
+; CHECK-NEXT:    st1w { z3.s }, p0, [x0, x9, lsl #2]
+; CHECK-NEXT:    st1w { z2.s }, p0, [x0, x10, lsl #2]
+; CHECK-NEXT:    st1w { z1.s }, p0, [x0]
 ; CHECK-NEXT:  .LBB1_2: // %exit
 ; CHECK-NEXT:    ret
   %broadcast.splat = shufflevector <32 x i1> zeroinitializer, <32 x i1> zeroinitializer, <32 x i32> zeroinitializer
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-splat-vector.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-splat-vector.ll
index f97ca05f3bdd4..b633057be139c 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-splat-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-splat-vector.ll
@@ -34,8 +34,8 @@ define <16 x i8> @splat_v16i8(i8 %a) vscale_range(2,0) #0 {
 define void @splat_v32i8(i8 %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl32
 ; CHECK-NEXT:    mov z0.b, w0
+; CHECK-NEXT:    ptrue p0.b, vl32
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <32 x i8> undef, i8 %a, i64 0
@@ -47,8 +47,8 @@ define void @splat_v32i8(i8 %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v64i8(i8 %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v64i8:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
 ; VBITS_GE_256-NEXT:    mov z0.b, w0
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
 ; VBITS_GE_256-NEXT:    mov w8, #32 // =0x20
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x1, x8]
 ; VBITS_GE_256-NEXT:    st1b { z0.b }, p0, [x1]
@@ -56,8 +56,8 @@ define void @splat_v64i8(i8 %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v64i8:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.b, vl64
 ; VBITS_GE_512-NEXT:    mov z0.b, w0
+; VBITS_GE_512-NEXT:    ptrue p0.b, vl64
 ; VBITS_GE_512-NEXT:    st1b { z0.b }, p0, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %insert = insertelement <64 x i8> undef, i8 %a, i64 0
@@ -69,8 +69,8 @@ define void @splat_v64i8(i8 %a, ptr %b) #0 {
 define void @splat_v128i8(i8 %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v128i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl128
 ; CHECK-NEXT:    mov z0.b, w0
+; CHECK-NEXT:    ptrue p0.b, vl128
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <128 x i8> undef, i8 %a, i64 0
@@ -82,8 +82,8 @@ define void @splat_v128i8(i8 %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v256i8(i8 %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v256i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl256
 ; CHECK-NEXT:    mov z0.b, w0
+; CHECK-NEXT:    ptrue p0.b, vl256
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <256 x i8> undef, i8 %a, i64 0
@@ -117,8 +117,8 @@ define <8 x i16> @splat_v8i16(i16 %a) vscale_range(2,0) #0 {
 define void @splat_v16i16(i16 %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    mov z0.h, w0
+; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <16 x i16> undef, i16 %a, i64 0
@@ -130,8 +130,8 @@ define void @splat_v16i16(i16 %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v32i16(i16 %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v32i16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov z0.h, w0
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x1, x8, lsl #1]
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x1]
@@ -139,8 +139,8 @@ define void @splat_v32i16(i16 %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v32i16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    mov z0.h, w0
+; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %insert = insertelement <32 x i16> undef, i16 %a, i64 0
@@ -152,8 +152,8 @@ define void @splat_v32i16(i16 %a, ptr %b) #0 {
 define void @splat_v64i16(i16 %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v64i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov z0.h, w0
+; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <64 x i16> undef, i16 %a, i64 0
@@ -165,8 +165,8 @@ define void @splat_v64i16(i16 %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v128i16(i16 %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v128i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov z0.h, w0
+; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <128 x i16> undef, i16 %a, i64 0
@@ -200,8 +200,8 @@ define <4 x i32> @splat_v4i32(i32 %a) vscale_range(2,0) #0 {
 define void @splat_v8i32(i32 %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    mov z0.s, w0
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <8 x i32> undef, i32 %a, i64 0
@@ -213,8 +213,8 @@ define void @splat_v8i32(i32 %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v16i32(i32 %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v16i32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov z0.s, w0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x1, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x1]
@@ -222,8 +222,8 @@ define void @splat_v16i32(i32 %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v16i32:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    mov z0.s, w0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %insert = insertelement <16 x i32> undef, i32 %a, i64 0
@@ -235,8 +235,8 @@ define void @splat_v16i32(i32 %a, ptr %b) #0 {
 define void @splat_v32i32(i32 %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v32i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    mov z0.s, w0
+; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <32 x i32> undef, i32 %a, i64 0
@@ -248,8 +248,8 @@ define void @splat_v32i32(i32 %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v64i32(i32 %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v64i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    mov z0.s, w0
+; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <64 x i32> undef, i32 %a, i64 0
@@ -283,8 +283,8 @@ define <2 x i64> @splat_v2i64(i64 %a) vscale_range(2,0) #0 {
 define void @splat_v4i64(i64 %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mov z0.d, x0
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <4 x i64> undef, i64 %a, i64 0
@@ -296,8 +296,8 @@ define void @splat_v4i64(i64 %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v8i64(i64 %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v8i64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov z0.d, x0
+; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x1, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x1]
@@ -305,8 +305,8 @@ define void @splat_v8i64(i64 %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v8i64:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    mov z0.d, x0
+; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x1]
 ; VBITS_GE_512-NEXT:    ret
   %insert = insertelement <8 x i64> undef, i64 %a, i64 0
@@ -318,8 +318,8 @@ define void @splat_v8i64(i64 %a, ptr %b) #0 {
 define void @splat_v16i64(i64 %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v16i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    mov z0.d, x0
+; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <16 x i64> undef, i64 %a, i64 0
@@ -331,8 +331,8 @@ define void @splat_v16i64(i64 %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v32i64(i64 %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v32i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    mov z0.d, x0
+; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %insert = insertelement <32 x i64> undef, i64 %a, i64 0
@@ -372,8 +372,8 @@ define <8 x half> @splat_v8f16(half %a) vscale_range(2,0) #0 {
 define void @splat_v16f16(half %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl16
 ; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -386,8 +386,8 @@ define void @splat_v16f16(half %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v32f16(half %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v32f16:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    // kill: def $h0 killed $h0 def $z0
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
 ; VBITS_GE_256-NEXT:    mov z0.h, h0
 ; VBITS_GE_256-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
@@ -396,8 +396,8 @@ define void @splat_v32f16(half %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v32f16:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    // kill: def $h0 killed $h0 def $z0
+; VBITS_GE_512-NEXT:    ptrue p0.h, vl32
 ; VBITS_GE_512-NEXT:    mov z0.h, h0
 ; VBITS_GE_512-NEXT:    st1h { z0.h }, p0, [x0]
 ; VBITS_GE_512-NEXT:    ret
@@ -410,8 +410,8 @@ define void @splat_v32f16(half %a, ptr %b) #0 {
 define void @splat_v64f16(half %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v64f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -424,8 +424,8 @@ define void @splat_v64f16(half %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v128f16(half %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v128f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl128
 ; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -462,8 +462,8 @@ define <4 x float> @splat_v4f32(float %a, <4 x float> %op2) vscale_range(2,0) #0
 define void @splat_v8f32(float %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    mov z0.s, s0
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -476,8 +476,8 @@ define void @splat_v8f32(float %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v16f32(float %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v16f32:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    // kill: def $s0 killed $s0 def $z0
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
 ; VBITS_GE_256-NEXT:    mov z0.s, s0
 ; VBITS_GE_256-NEXT:    st1w { z0.s }, p0, [x0, x8, lsl #2]
@@ -486,8 +486,8 @@ define void @splat_v16f32(float %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v16f32:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    // kill: def $s0 killed $s0 def $z0
+; VBITS_GE_512-NEXT:    ptrue p0.s, vl16
 ; VBITS_GE_512-NEXT:    mov z0.s, s0
 ; VBITS_GE_512-NEXT:    st1w { z0.s }, p0, [x0]
 ; VBITS_GE_512-NEXT:    ret
@@ -500,8 +500,8 @@ define void @splat_v16f32(float %a, ptr %b) #0 {
 define void @splat_v32f32(float %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v32f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    mov z0.s, s0
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -514,8 +514,8 @@ define void @splat_v32f32(float %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v64f32(float %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v64f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl64
 ; CHECK-NEXT:    mov z0.s, s0
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -550,8 +550,8 @@ define <2 x double> @splat_v2f64(double %a, <2 x double> %op2) vscale_range(2,0)
 define void @splat_v4f64(double %a, ptr %b) vscale_range(2,0) #0 {
 ; CHECK-LABEL: splat_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mov z0.d, d0
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -564,8 +564,8 @@ define void @splat_v4f64(double %a, ptr %b) vscale_range(2,0) #0 {
 define void @splat_v8f64(double %a, ptr %b) #0 {
 ; VBITS_GE_256-LABEL: splat_v8f64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    // kill: def $d0 killed $d0 def $z0
+; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
 ; VBITS_GE_256-NEXT:    mov z0.d, d0
 ; VBITS_GE_256-NEXT:    st1d { z0.d }, p0, [x0, x8, lsl #3]
@@ -574,8 +574,8 @@ define void @splat_v8f64(double %a, ptr %b) #0 {
 ;
 ; VBITS_GE_512-LABEL: splat_v8f64:
 ; VBITS_GE_512:       // %bb.0:
-; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    // kill: def $d0 killed $d0 def $z0
+; VBITS_GE_512-NEXT:    ptrue p0.d, vl8
 ; VBITS_GE_512-NEXT:    mov z0.d, d0
 ; VBITS_GE_512-NEXT:    st1d { z0.d }, p0, [x0]
 ; VBITS_GE_512-NEXT:    ret
@@ -588,8 +588,8 @@ define void @splat_v8f64(double %a, ptr %b) #0 {
 define void @splat_v16f64(double %a, ptr %b) vscale_range(8,0) #0 {
 ; CHECK-LABEL: splat_v16f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl16
 ; CHECK-NEXT:    mov z0.d, d0
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -602,8 +602,8 @@ define void @splat_v16f64(double %a, ptr %b) vscale_range(8,0) #0 {
 define void @splat_v32f64(double %a, ptr %b) vscale_range(16,0) #0 {
 ; CHECK-LABEL: splat_v32f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl32
 ; CHECK-NEXT:    mov z0.d, d0
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -620,8 +620,8 @@ define void @splat_v32f64(double %a, ptr %b) vscale_range(16,0) #0 {
 define void @splat_imm_v64i8(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v64i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl64
 ; CHECK-NEXT:    mov z0.b, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.b, vl64
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <64 x i8> undef, i8 1, i64 0
@@ -633,8 +633,8 @@ define void @splat_imm_v64i8(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v32i16(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v32i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl32
 ; CHECK-NEXT:    mov z0.h, #2 // =0x2
+; CHECK-NEXT:    ptrue p0.h, vl32
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <32 x i16> undef, i16 2, i64 0
@@ -646,8 +646,8 @@ define void @splat_imm_v32i16(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v16i32(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v16i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl16
 ; CHECK-NEXT:    mov z0.s, #3 // =0x3
+; CHECK-NEXT:    ptrue p0.s, vl16
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <16 x i32> undef, i32 3, i64 0
@@ -659,8 +659,8 @@ define void @splat_imm_v16i32(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v8i64(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v8i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl8
 ; CHECK-NEXT:    mov z0.d, #4 // =0x4
+; CHECK-NEXT:    ptrue p0.d, vl8
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <8 x i64> undef, i64 4, i64 0
@@ -676,8 +676,8 @@ define void @splat_imm_v8i64(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v32f16(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v32f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl32
 ; CHECK-NEXT:    fmov z0.h, #5.00000000
+; CHECK-NEXT:    ptrue p0.h, vl32
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <32 x half> undef, half 5.0, i64 0
@@ -689,8 +689,8 @@ define void @splat_imm_v32f16(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v16f32(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v16f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl16
 ; CHECK-NEXT:    fmov z0.s, #6.00000000
+; CHECK-NEXT:    ptrue p0.s, vl16
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <16 x float> undef, float 6.0, i64 0
@@ -702,8 +702,8 @@ define void @splat_imm_v16f32(ptr %a) vscale_range(4,0) #0 {
 define void @splat_imm_v8f64(ptr %a) vscale_range(4,0) #0 {
 ; CHECK-LABEL: splat_imm_v8f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl8
 ; CHECK-NEXT:    fmov z0.d, #7.00000000
+; CHECK-NEXT:    ptrue p0.d, vl8
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %insert = insertelement <8 x double> undef, double 7.0, i64 0
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-trunc-stores.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-trunc-stores.ll
index 2dc4bddb81a6d..020d5cb53bf21 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-trunc-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-trunc-stores.ll
@@ -36,14 +36,14 @@ define void @store_trunc_v8i64i8(ptr %ap, ptr %dest) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1b { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1b { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: store_trunc_v8i64i8:
@@ -117,14 +117,14 @@ define void @store_trunc_v8i64i32(ptr %ap, ptr %dest) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.d, vl4
 ; VBITS_GE_256-NEXT:    mov x8, #4 // =0x4
-; VBITS_GE_256-NEXT:    ptrue p1.s, vl8
 ; VBITS_GE_256-NEXT:    ld1d { z0.d }, p0/z, [x0, x8, lsl #3]
 ; VBITS_GE_256-NEXT:    ld1d { z1.d }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl4
 ; VBITS_GE_256-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; VBITS_GE_256-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; VBITS_GE_256-NEXT:    splice z1.s, p0, z1.s, z0.s
-; VBITS_GE_256-NEXT:    st1w { z1.s }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
+; VBITS_GE_256-NEXT:    st1w { z1.s }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: store_trunc_v8i64i32:
@@ -172,14 +172,14 @@ define void @store_trunc_v16i32i16(ptr %ap, ptr %dest) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.s, vl8
 ; VBITS_GE_256-NEXT:    mov x8, #8 // =0x8
-; VBITS_GE_256-NEXT:    ptrue p1.h, vl16
 ; VBITS_GE_256-NEXT:    ld1w { z0.s }, p0/z, [x0, x8, lsl #2]
 ; VBITS_GE_256-NEXT:    ld1w { z1.s }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ptrue p0.h, vl8
 ; VBITS_GE_256-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; VBITS_GE_256-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; VBITS_GE_256-NEXT:    splice z1.h, p0, z1.h, z0.h
-; VBITS_GE_256-NEXT:    st1h { z1.h }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
+; VBITS_GE_256-NEXT:    st1h { z1.h }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: store_trunc_v16i32i16:
@@ -199,14 +199,14 @@ define void @store_trunc_v32i16i8(ptr %ap, ptr %dest) #0 {
 ; VBITS_GE_256:       // %bb.0:
 ; VBITS_GE_256-NEXT:    ptrue p0.h, vl16
 ; VBITS_GE_256-NEXT:    mov x8, #16 // =0x10
-; VBITS_GE_256-NEXT:    ptrue p1.b, vl32
 ; VBITS_GE_256-NEXT:    ld1h { z0.h }, p0/z, [x0, x8, lsl #1]
 ; VBITS_GE_256-NEXT:    ld1h { z1.h }, p0/z, [x0]
 ; VBITS_GE_256-NEXT:    ptrue p0.b, vl16
 ; VBITS_GE_256-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; VBITS_GE_256-NEXT:    uzp1 z1.b, z1.b, z1.b
 ; VBITS_GE_256-NEXT:    splice z1.b, p0, z1.b, z0.b
-; VBITS_GE_256-NEXT:    st1b { z1.b }, p1, [x1]
+; VBITS_GE_256-NEXT:    ptrue p0.b, vl32
+; VBITS_GE_256-NEXT:    st1b { z1.b }, p0, [x1]
 ; VBITS_GE_256-NEXT:    ret
 ;
 ; VBITS_GE_512-LABEL: store_trunc_v32i16i8:
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-vector-shuffle-tbl.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-vector-shuffle-tbl.ll
index 68c234a20d110..28094c7b68e7c 100644
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-vector-shuffle-tbl.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-vector-shuffle-tbl.ll
@@ -559,13 +559,13 @@ define <8 x i16> @shuffle_index_indices_from_both_ops_i16(ptr %a, ptr %b) {
 ;
 ; SVE2_128_NOMAX-LABEL: shuffle_index_indices_from_both_ops_i16:
 ; SVE2_128_NOMAX:       // %bb.0:
-; SVE2_128_NOMAX-NEXT:    ptrue p0.h, vl8
 ; SVE2_128_NOMAX-NEXT:    cnth x8
 ; SVE2_128_NOMAX-NEXT:    adrp x9, .LCPI7_0
 ; SVE2_128_NOMAX-NEXT:    adrp x10, .LCPI7_1
 ; SVE2_128_NOMAX-NEXT:    mov z0.h, w8
 ; SVE2_128_NOMAX-NEXT:    ldr q1, [x9, :lo12:.LCPI7_0]
 ; SVE2_128_NOMAX-NEXT:    ldr q2, [x10, :lo12:.LCPI7_1]
+; SVE2_128_NOMAX-NEXT:    ptrue p0.h, vl8
 ; SVE2_128_NOMAX-NEXT:    mad z0.h, p0/m, z1.h, z2.h
 ; SVE2_128_NOMAX-NEXT:    ldr q1, [x0]
 ; SVE2_128_NOMAX-NEXT:    ldr q2, [x1]
@@ -575,13 +575,13 @@ define <8 x i16> @shuffle_index_indices_from_both_ops_i16(ptr %a, ptr %b) {
 ;
 ; SVE2_NOMIN_NOMAX-LABEL: shuffle_index_indices_from_both_ops_i16:
 ; SVE2_NOMIN_NOMAX:       // %bb.0:
-; SVE2_NOMIN_NOMAX-NEXT:    ptrue p0.h, vl8
 ; SVE2_NOMIN_NOMAX-NEXT:    cnth x8
 ; SVE2_NOMIN_NOMAX-NEXT:    adrp x9, .LCPI7_0
 ; SVE2_NOMIN_NOMAX-NEXT:    adrp x10, .LCPI7_1
 ; SVE2_NOMIN_NOMAX-NEXT:    mov z0.h, w8
 ; SVE2_NOMIN_NOMAX-NEXT:    ldr q1, [x9, :lo12:.LCPI7_0]
 ; SVE2_NOMIN_NOMAX-NEXT:    ldr q2, [x10, :lo12:.LCPI7_1]
+; SVE2_NOMIN_NOMAX-NEXT:    ptrue p0.h, vl8
 ; SVE2_NOMIN_NOMAX-NEXT:    mad z0.h, p0/m, z1.h, z2.h
 ; SVE2_NOMIN_NOMAX-NEXT:    ldr q1, [x0]
 ; SVE2_NOMIN_NOMAX-NEXT:    ldr q2, [x1]
@@ -597,9 +597,9 @@ define <8 x i16> @shuffle_index_indices_from_both_ops_i16(ptr %a, ptr %b) {
 ; SVE2_MIN_256_NOMAX-NEXT:    adrp x9, .LCPI7_1
 ; SVE2_MIN_256_NOMAX-NEXT:    add x9, x9, :lo12:.LCPI7_1
 ; SVE2_MIN_256_NOMAX-NEXT:    cnth x10
-; SVE2_MIN_256_NOMAX-NEXT:    mov z2.h, w10
 ; SVE2_MIN_256_NOMAX-NEXT:    ld1h { z0.h }, p0/z, [x8]
 ; SVE2_MIN_256_NOMAX-NEXT:    ld1h { z1.h }, p0/z, [x9]
+; SVE2_MIN_256_NOMAX-NEXT:    mov z2.h, w10
 ; SVE2_MIN_256_NOMAX-NEXT:    mad z0.h, p0/m, z2.h, z1.h
 ; SVE2_MIN_256_NOMAX-NEXT:    ldr q1, [x0]
 ; SVE2_MIN_256_NOMAX-NEXT:    ldr q2, [x1]
diff --git a/llvm/test/CodeGen/AArch64/sve-fp-int-min-max.ll b/llvm/test/CodeGen/AArch64/sve-fp-int-min-max.ll
index 5ff9f0f0df62f..afe13851f0b95 100644
--- a/llvm/test/CodeGen/AArch64/sve-fp-int-min-max.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fp-int-min-max.ll
@@ -7,11 +7,11 @@ define i64 @scalable_int_min_max(ptr %arg, ptr %arg1, <vscale x 2 x ptr> %i37, <
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #3745 // =0xea1
 ; CHECK-NEXT:    movk w8, #16618, lsl #16
+; CHECK-NEXT:    ld1w { z3.d }, p0/z, [x0]
 ; CHECK-NEXT:    mov z4.s, w8
 ; CHECK-NEXT:    mov w8, #57344 // =0xe000
 ; CHECK-NEXT:    movk w8, #17535, lsl #16
 ; CHECK-NEXT:    mov z5.s, w8
-; CHECK-NEXT:    ld1w { z3.d }, p0/z, [x0]
 ; CHECK-NEXT:    fmul z4.s, p0/m, z4.s, z3.s
 ; CHECK-NEXT:    fadd z4.s, p0/m, z4.s, z5.s
 ; CHECK-NEXT:    mov z5.d, #1023 // =0x3ff
diff --git a/llvm/test/CodeGen/AArch64/sve-fp-reciprocal.ll b/llvm/test/CodeGen/AArch64/sve-fp-reciprocal.ll
index aefc8de431436..6420071b3dce4 100644
--- a/llvm/test/CodeGen/AArch64/sve-fp-reciprocal.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fp-reciprocal.ll
@@ -92,8 +92,8 @@ define <vscale x 8 x half> @fsqrt_recip_8f16(<vscale x 8 x half> %a) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    frsqrte z1.h, z0.h
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    fmul z2.h, z1.h, z1.h
 ; CHECK-NEXT:    fcmne p0.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    fmul z2.h, z1.h, z1.h
 ; CHECK-NEXT:    frsqrts z2.h, z0.h, z2.h
 ; CHECK-NEXT:    fmul z1.h, z1.h, z2.h
 ; CHECK-NEXT:    fmul z0.h, p0/m, z0.h, z1.h
@@ -117,8 +117,8 @@ define <vscale x 4 x float> @fsqrt_recip_4f32(<vscale x 4 x float> %a) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    frsqrte z1.s, z0.s
 ; CHECK-NEXT:    ptrue p0.s
-; CHECK-NEXT:    fmul z2.s, z1.s, z1.s
 ; CHECK-NEXT:    fcmne p0.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fmul z2.s, z1.s, z1.s
 ; CHECK-NEXT:    frsqrts z2.s, z0.s, z2.s
 ; CHECK-NEXT:    fmul z1.s, z1.s, z2.s
 ; CHECK-NEXT:    fmul z2.s, z1.s, z1.s
@@ -145,8 +145,8 @@ define <vscale x 2 x double> @fsqrt_recip_2f64(<vscale x 2 x double> %a) #0 {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    frsqrte z1.d, z0.d
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    fmul z2.d, z1.d, z1.d
 ; CHECK-NEXT:    fcmne p0.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    fmul z2.d, z1.d, z1.d
 ; CHECK-NEXT:    frsqrts z2.d, z0.d, z2.d
 ; CHECK-NEXT:    fmul z1.d, z1.d, z2.d
 ; CHECK-NEXT:    fmul z2.d, z1.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-fp-reduce-fadda.ll b/llvm/test/CodeGen/AArch64/sve-fp-reduce-fadda.ll
index 460d8a8694bc4..1a2ab8d4253ab 100644
--- a/llvm/test/CodeGen/AArch64/sve-fp-reduce-fadda.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fp-reduce-fadda.ll
@@ -51,10 +51,10 @@ define half @fadda_nxv6f16(<vscale x 6 x half> %v, half %s) {
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #32768 // =0x8000
-; CHECK-NEXT:    ptrue p1.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    fmov s0, s1
 ; CHECK-NEXT:    st1h { z2.d }, p1, [sp, #3, mul vl]
diff --git a/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll b/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
index 813f1601e809e..584c29ebcfc04 100644
--- a/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fptosi-sat.ll
@@ -14,8 +14,8 @@ declare <vscale x 4 x i64> @llvm.fptosi.sat.nxv4f32.nxv4i64(<vscale x 4 x float>
 define <vscale x 2 x i32> @test_signed_v2f32_v2i32(<vscale x 2 x float> %f) {
 ; CHECK-LABEL: test_signed_v2f32_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #-822083584 // =0xcf000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0xffffffff80000000
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mov w8, #1325400063 // =0x4effffff
@@ -38,8 +38,8 @@ define <vscale x 2 x i32> @test_signed_v2f32_v2i32(<vscale x 2 x float> %f) {
 define <vscale x 4 x i32> @test_signed_v4f32_v4i32(<vscale x 4 x float> %f) {
 ; CHECK-LABEL: test_signed_v4f32_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #-822083584 // =0xcf000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #0x80000000
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mov w8, #1325400063 // =0x4effffff
@@ -67,29 +67,29 @@ define <vscale x 8 x i32> @test_signed_v8f32_v8i32(<vscale x 8 x float> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #-822083584 // =0xcf000000
-; CHECK-NEXT:    mov z3.s, #0x80000000
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z6.s, #0x7fffffff
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    mov w8, #1325400063 // =0x4effffff
-; CHECK-NEXT:    mov z6.s, #0x7fffffff
-; CHECK-NEXT:    mov z4.s, w8
-; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, z2.s
-; CHECK-NEXT:    fcmge p2.s, p0/z, z1.s, z2.s
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzs z2.s, p0/m, z0.s
+; CHECK-NEXT:    mov z3.s, w8
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzs z4.s, p0/m, z0.s
 ; CHECK-NEXT:    movprfx z5, z1
 ; CHECK-NEXT:    fcvtzs z5.s, p0/m, z1.s
-; CHECK-NEXT:    fcmgt p3.s, p0/z, z0.s, z4.s
-; CHECK-NEXT:    fcmgt p4.s, p0/z, z1.s, z4.s
+; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, z2.s
+; CHECK-NEXT:    fcmge p2.s, p0/z, z1.s, z2.s
+; CHECK-NEXT:    mov z2.s, #0x80000000
+; CHECK-NEXT:    fcmgt p3.s, p0/z, z0.s, z3.s
+; CHECK-NEXT:    fcmgt p4.s, p0/z, z1.s, z3.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.s, p1/m, z3.s
+; CHECK-NEXT:    sel z3.s, p1, z2.s, z4.s
 ; CHECK-NEXT:    fcmuo p1.s, p0/z, z0.s, z0.s
 ; CHECK-NEXT:    fcmuo p0.s, p0/z, z1.s, z1.s
-; CHECK-NEXT:    sel z3.s, p2, z3.s, z5.s
-; CHECK-NEXT:    sel z0.s, p3, z6.s, z2.s
-; CHECK-NEXT:    sel z1.s, p4, z6.s, z3.s
+; CHECK-NEXT:    sel z2.s, p2, z2.s, z5.s
+; CHECK-NEXT:    sel z0.s, p3, z6.s, z3.s
+; CHECK-NEXT:    sel z1.s, p4, z6.s, z2.s
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.s, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.s, p0/m, #0 // =0x0
@@ -103,8 +103,8 @@ define <vscale x 8 x i32> @test_signed_v8f32_v8i32(<vscale x 8 x float> %f) {
 define <vscale x 4 x i16> @test_signed_v4f32_v4i16(<vscale x 4 x float> %f) {
 ; CHECK-LABEL: test_signed_v4f32_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #-956301312 // =0xc7000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mov w8, #65024 // =0xfe00
 ; CHECK-NEXT:    movk w8, #18175, lsl #16
@@ -132,28 +132,28 @@ define <vscale x 8 x i16> @test_signed_v8f32_v8i16(<vscale x 8 x float> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #-956301312 // =0xc7000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z5.s, #32767 // =0x7fff
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    mov w8, #65024 // =0xfe00
 ; CHECK-NEXT:    movk w8, #18175, lsl #16
-; CHECK-NEXT:    mov z3.s, w8
-; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, z2.s
-; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, z2.s
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.s, p0/m, z1.s
+; CHECK-NEXT:    movprfx z3, z1
+; CHECK-NEXT:    fcvtzs z3.s, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z4, z0
 ; CHECK-NEXT:    fcvtzs z4.s, p0/m, z0.s
-; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z3.s
-; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z3.s
+; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, z2.s
+; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, z2.s
+; CHECK-NEXT:    mov z2.s, w8
+; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z2.s
+; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z2.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.s, p1/m, #-32768 // =0xffffffffffff8000
+; CHECK-NEXT:    mov z3.s, p1/m, #-32768 // =0xffffffffffff8000
 ; CHECK-NEXT:    fcmuo p1.s, p0/z, z1.s, z1.s
 ; CHECK-NEXT:    fcmuo p0.s, p0/z, z0.s, z0.s
 ; CHECK-NEXT:    mov z4.s, p2/m, #-32768 // =0xffffffffffff8000
-; CHECK-NEXT:    sel z0.s, p3, z5.s, z2.s
+; CHECK-NEXT:    sel z0.s, p3, z5.s, z3.s
 ; CHECK-NEXT:    sel z1.s, p4, z5.s, z4.s
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.s, p1/m, #0 // =0x0
@@ -169,8 +169,8 @@ define <vscale x 8 x i16> @test_signed_v8f32_v8i16(<vscale x 8 x float> %f) {
 define <vscale x 2 x i64> @test_signed_v2f32_v2i64(<vscale x 2 x float> %f) {
 ; CHECK-LABEL: test_signed_v2f32_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #-553648128 // =0xdf000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0x8000000000000000
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mov w8, #1593835519 // =0x5effffff
@@ -198,31 +198,31 @@ define <vscale x 4 x i64> @test_signed_v4f32_v4i64(<vscale x 4 x float> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #-553648128 // =0xdf000000
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
 ; CHECK-NEXT:    mov z2.s, w8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1593835519 // =0x5effffff
-; CHECK-NEXT:    mov z3.d, #0x8000000000000000
-; CHECK-NEXT:    mov z4.s, w8
+; CHECK-NEXT:    mov z3.s, w8
 ; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, z2.s
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.s
+; CHECK-NEXT:    mov z2.d, #0x8000000000000000
+; CHECK-NEXT:    movprfx z4, z1
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z5, z0
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.s
-; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z4.s
-; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z4.s
+; CHECK-NEXT:    fcmgt p3.s, p0/z, z1.s, z3.s
+; CHECK-NEXT:    fcmgt p4.s, p0/z, z0.s, z3.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.d, p1/m, z3.d
+; CHECK-NEXT:    sel z3.d, p1, z2.d, z4.d
 ; CHECK-NEXT:    fcmuo p1.s, p0/z, z1.s, z1.s
 ; CHECK-NEXT:    fcmuo p0.s, p0/z, z0.s, z0.s
-; CHECK-NEXT:    sel z3.d, p2, z3.d, z5.d
-; CHECK-NEXT:    sel z0.d, p3, z6.d, z2.d
-; CHECK-NEXT:    sel z1.d, p4, z6.d, z3.d
+; CHECK-NEXT:    sel z2.d, p2, z2.d, z5.d
+; CHECK-NEXT:    sel z0.d, p3, z6.d, z3.d
+; CHECK-NEXT:    sel z1.d, p4, z6.d, z2.d
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p0/m, #0 // =0x0
@@ -246,8 +246,8 @@ declare <vscale x 4 x i64> @llvm.fptosi.sat.nxv4f64.nxv4i64(<vscale x 4 x double
 define <vscale x 2 x i32> @test_signed_v2f64_v2i32(<vscale x 2 x double> %f) {
 ; CHECK-LABEL: test_signed_v2f64_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4476578029606273024 // =0xc1e0000000000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0xffffffff80000000
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mov x8, #281474972516352 // =0xffffffc00000
@@ -276,30 +276,30 @@ define <vscale x 4 x i32> @test_signed_v4f64_v4i32(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4476578029606273024 // =0xc1e0000000000000
-; CHECK-NEXT:    mov z3.d, #0xffffffff80000000
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z6.d, #0x7fffffff
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    mov x8, #281474972516352 // =0xffffffc00000
-; CHECK-NEXT:    mov z6.d, #0x7fffffff
 ; CHECK-NEXT:    movk x8, #16863, lsl #48
-; CHECK-NEXT:    mov z4.d, x8
-; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z4, z1
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z5, z0
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.d
-; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z4.d
-; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z4.d
+; CHECK-NEXT:    mov z3.d, x8
+; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
+; CHECK-NEXT:    mov z2.d, #0xffffffff80000000
+; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z3.d
+; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z3.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.d, p1/m, z3.d
+; CHECK-NEXT:    sel z3.d, p1, z2.d, z4.d
 ; CHECK-NEXT:    fcmuo p1.d, p0/z, z1.d, z1.d
 ; CHECK-NEXT:    fcmuo p0.d, p0/z, z0.d, z0.d
-; CHECK-NEXT:    sel z3.d, p2, z3.d, z5.d
-; CHECK-NEXT:    sel z0.d, p3, z6.d, z2.d
-; CHECK-NEXT:    sel z1.d, p4, z6.d, z3.d
+; CHECK-NEXT:    sel z2.d, p2, z2.d, z5.d
+; CHECK-NEXT:    sel z0.d, p3, z6.d, z3.d
+; CHECK-NEXT:    sel z1.d, p4, z6.d, z2.d
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p0/m, #0 // =0x0
@@ -322,48 +322,48 @@ define <vscale x 8 x i32> @test_signed_v8f64_v8i32(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4476578029606273024 // =0xc1e0000000000000
-; CHECK-NEXT:    mov z26.d, #0x7fffffff
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z5.d, #0xffffffff80000000
 ; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    mov x8, #281474972516352 // =0xffffffc00000
+; CHECK-NEXT:    mov z26.d, #0x7fffffff
 ; CHECK-NEXT:    movk x8, #16863, lsl #48
-; CHECK-NEXT:    mov z5.d, x8
-; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z4.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z4.d
-; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, z4.d
-; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, z4.d
-; CHECK-NEXT:    mov z4.d, #0xffffffff80000000
-; CHECK-NEXT:    movprfx z6, z1
-; CHECK-NEXT:    fcvtzs z6.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z7, z0
 ; CHECK-NEXT:    fcvtzs z7.d, p0/m, z0.d
 ; CHECK-NEXT:    movprfx z24, z3
 ; CHECK-NEXT:    fcvtzs z24.d, p0/m, z3.d
+; CHECK-NEXT:    mov z6.d, x8
 ; CHECK-NEXT:    movprfx z25, z2
 ; CHECK-NEXT:    fcvtzs z25.d, p0/m, z2.d
-; CHECK-NEXT:    fcmgt p5.d, p0/z, z1.d, z5.d
-; CHECK-NEXT:    fcmgt p6.d, p0/z, z0.d, z5.d
-; CHECK-NEXT:    fcmgt p7.d, p0/z, z3.d, z5.d
+; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z4.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z4.d
+; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, z4.d
+; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, z4.d
+; CHECK-NEXT:    movprfx z4, z1
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z1.d
+; CHECK-NEXT:    fcmgt p5.d, p0/z, z1.d, z6.d
+; CHECK-NEXT:    fcmgt p6.d, p0/z, z0.d, z6.d
+; CHECK-NEXT:    fcmgt p7.d, p0/z, z3.d, z6.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
 ; CHECK-NEXT:    not p3.b, p0/z, p3.b
+; CHECK-NEXT:    mov z4.d, p1/m, z5.d
+; CHECK-NEXT:    fcmgt p1.d, p0/z, z2.d, z6.d
 ; CHECK-NEXT:    not p4.b, p0/z, p4.b
-; CHECK-NEXT:    mov z6.d, p1/m, z4.d
-; CHECK-NEXT:    fcmgt p1.d, p0/z, z2.d, z5.d
-; CHECK-NEXT:    sel z5.d, p2, z4.d, z7.d
+; CHECK-NEXT:    sel z6.d, p2, z5.d, z7.d
 ; CHECK-NEXT:    fcmuo p2.d, p0/z, z1.d, z1.d
-; CHECK-NEXT:    sel z7.d, p3, z4.d, z24.d
+; CHECK-NEXT:    sel z7.d, p3, z5.d, z24.d
 ; CHECK-NEXT:    fcmuo p3.d, p0/z, z0.d, z0.d
-; CHECK-NEXT:    sel z4.d, p4, z4.d, z25.d
+; CHECK-NEXT:    sel z5.d, p4, z5.d, z25.d
 ; CHECK-NEXT:    fcmuo p4.d, p0/z, z3.d, z3.d
 ; CHECK-NEXT:    fcmuo p0.d, p0/z, z2.d, z2.d
-; CHECK-NEXT:    sel z0.d, p5, z26.d, z6.d
-; CHECK-NEXT:    sel z1.d, p6, z26.d, z5.d
+; CHECK-NEXT:    sel z0.d, p5, z26.d, z4.d
+; CHECK-NEXT:    sel z1.d, p6, z26.d, z6.d
 ; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    sel z2.d, p7, z26.d, z7.d
 ; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    sel z3.d, p1, z26.d, z4.d
+; CHECK-NEXT:    sel z3.d, p1, z26.d, z5.d
 ; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p3/m, #0 // =0x0
@@ -387,28 +387,28 @@ define <vscale x 4 x i16> @test_signed_v4f64_v4i16(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4548635623644200960 // =0xc0e0000000000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z5.d, #32767 // =0x7fff
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    mov x8, #281200098803712 // =0xffc000000000
 ; CHECK-NEXT:    movk x8, #16607, lsl #48
-; CHECK-NEXT:    mov z3.d, x8
-; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.d
+; CHECK-NEXT:    movprfx z3, z1
+; CHECK-NEXT:    fcvtzs z3.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z4, z0
 ; CHECK-NEXT:    fcvtzs z4.d, p0/m, z0.d
-; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z3.d
-; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z3.d
+; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, z2.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, z2.d
+; CHECK-NEXT:    mov z2.d, x8
+; CHECK-NEXT:    fcmgt p3.d, p0/z, z1.d, z2.d
+; CHECK-NEXT:    fcmgt p4.d, p0/z, z0.d, z2.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.d, p1/m, #-32768 // =0xffffffffffff8000
+; CHECK-NEXT:    mov z3.d, p1/m, #-32768 // =0xffffffffffff8000
 ; CHECK-NEXT:    fcmuo p1.d, p0/z, z1.d, z1.d
 ; CHECK-NEXT:    fcmuo p0.d, p0/z, z0.d, z0.d
 ; CHECK-NEXT:    mov z4.d, p2/m, #-32768 // =0xffffffffffff8000
-; CHECK-NEXT:    sel z0.d, p3, z5.d, z2.d
+; CHECK-NEXT:    sel z0.d, p3, z5.d, z3.d
 ; CHECK-NEXT:    sel z1.d, p4, z5.d, z4.d
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
@@ -432,34 +432,34 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4548635623644200960 // =0xc0e0000000000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z25.d, #32767 // =0x7fff
 ; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    mov x8, #281200098803712 // =0xffc000000000
 ; CHECK-NEXT:    movk x8, #16607, lsl #48
-; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, z4.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, z4.d
-; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, z4.d
-; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, z4.d
-; CHECK-NEXT:    movprfx z5, z3
-; CHECK-NEXT:    fcvtzs z5.d, p0/m, z3.d
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z6, z2
 ; CHECK-NEXT:    fcvtzs z6.d, p0/m, z2.d
 ; CHECK-NEXT:    movprfx z7, z1
 ; CHECK-NEXT:    fcvtzs z7.d, p0/m, z1.d
+; CHECK-NEXT:    mov z5.d, x8
 ; CHECK-NEXT:    movprfx z24, z0
 ; CHECK-NEXT:    fcvtzs z24.d, p0/m, z0.d
+; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, z4.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, z4.d
+; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, z4.d
+; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, z4.d
+; CHECK-NEXT:    movprfx z4, z3
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z3.d
+; CHECK-NEXT:    fcmgt p5.d, p0/z, z3.d, z5.d
+; CHECK-NEXT:    fcmgt p6.d, p0/z, z2.d, z5.d
+; CHECK-NEXT:    fcmgt p7.d, p0/z, z1.d, z5.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    fcmgt p5.d, p0/z, z3.d, z4.d
-; CHECK-NEXT:    fcmgt p6.d, p0/z, z2.d, z4.d
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    fcmgt p7.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    not p3.b, p0/z, p3.b
+; CHECK-NEXT:    mov z4.d, p1/m, #-32768 // =0xffffffffffff8000
+; CHECK-NEXT:    fcmgt p1.d, p0/z, z0.d, z5.d
 ; CHECK-NEXT:    not p4.b, p0/z, p4.b
-; CHECK-NEXT:    mov z5.d, p1/m, #-32768 // =0xffffffffffff8000
-; CHECK-NEXT:    fcmgt p1.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    mov z6.d, p2/m, #-32768 // =0xffffffffffff8000
 ; CHECK-NEXT:    fcmuo p2.d, p0/z, z3.d, z3.d
 ; CHECK-NEXT:    mov z7.d, p3/m, #-32768 // =0xffffffffffff8000
@@ -467,7 +467,7 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    mov z24.d, p4/m, #-32768 // =0xffffffffffff8000
 ; CHECK-NEXT:    fcmuo p4.d, p0/z, z1.d, z1.d
 ; CHECK-NEXT:    fcmuo p0.d, p0/z, z0.d, z0.d
-; CHECK-NEXT:    sel z2.d, p5, z25.d, z5.d
+; CHECK-NEXT:    sel z2.d, p5, z25.d, z4.d
 ; CHECK-NEXT:    sel z0.d, p6, z25.d, z6.d
 ; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    sel z1.d, p7, z25.d, z7.d
@@ -492,8 +492,8 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 define <vscale x 2 x i64> @test_signed_v2f64_v2i64(<vscale x 2 x double> %f) {
 ; CHECK-LABEL: test_signed_v2f64_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4332462841530417152 // =0xc3e0000000000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0x8000000000000000
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mov x8, #4890909195324358655 // =0x43dfffffffffffff
@@ -521,29 +521,29 @@ define <vscale x 4 x i64> @test_signed_v4f64_v4i64(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #-4332462841530417152 // =0xc3e0000000000000
-; CHECK-NEXT:    mov z3.d, #0x8000000000000000
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    mov x8, #4890909195324358655 // =0x43dfffffffffffff
-; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
-; CHECK-NEXT:    mov z4.d, x8
-; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, z2.d
-; CHECK-NEXT:    fcmge p2.d, p0/z, z1.d, z2.d
-; CHECK-NEXT:    movprfx z2, z0
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z0.d
+; CHECK-NEXT:    mov z3.d, x8
+; CHECK-NEXT:    movprfx z4, z0
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z0.d
 ; CHECK-NEXT:    movprfx z5, z1
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z1.d
-; CHECK-NEXT:    fcmgt p3.d, p0/z, z0.d, z4.d
-; CHECK-NEXT:    fcmgt p4.d, p0/z, z1.d, z4.d
+; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, z2.d
+; CHECK-NEXT:    fcmge p2.d, p0/z, z1.d, z2.d
+; CHECK-NEXT:    mov z2.d, #0x8000000000000000
+; CHECK-NEXT:    fcmgt p3.d, p0/z, z0.d, z3.d
+; CHECK-NEXT:    fcmgt p4.d, p0/z, z1.d, z3.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.d, p1/m, z3.d
+; CHECK-NEXT:    sel z3.d, p1, z2.d, z4.d
 ; CHECK-NEXT:    fcmuo p1.d, p0/z, z0.d, z0.d
 ; CHECK-NEXT:    fcmuo p0.d, p0/z, z1.d, z1.d
-; CHECK-NEXT:    sel z3.d, p2, z3.d, z5.d
-; CHECK-NEXT:    sel z0.d, p3, z6.d, z2.d
-; CHECK-NEXT:    sel z1.d, p4, z6.d, z3.d
+; CHECK-NEXT:    sel z2.d, p2, z2.d, z5.d
+; CHECK-NEXT:    sel z0.d, p3, z6.d, z3.d
+; CHECK-NEXT:    sel z1.d, p4, z6.d, z2.d
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p0/m, #0 // =0x0
@@ -568,8 +568,8 @@ declare <vscale x 4 x i64> @llvm.fptosi.sat.nxv4f16.nxv4i64(<vscale x 4 x half>)
 define <vscale x 2 x i32> @test_signed_v2f16_v2i32(<vscale x 2 x half> %f) {
 ; CHECK-LABEL: test_signed_v2f16_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #64511 // =0xfbff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0xffffffff80000000
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
@@ -592,8 +592,8 @@ define <vscale x 2 x i32> @test_signed_v2f16_v2i32(<vscale x 2 x half> %f) {
 define <vscale x 4 x i32> @test_signed_v4f16_v4i32(<vscale x 4 x half> %f) {
 ; CHECK-LABEL: test_signed_v4f16_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #64511 // =0xfbff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #0x80000000
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
@@ -621,31 +621,31 @@ define <vscale x 8 x i32> @test_signed_v8f16_v8i32(<vscale x 8 x half> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #64511 // =0xfbff
 ; CHECK-NEXT:    uunpklo z1.s, z0.h
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
 ; CHECK-NEXT:    mov z2.h, w8
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    mov z3.s, #0x80000000
-; CHECK-NEXT:    mov z4.h, w8
+; CHECK-NEXT:    mov z3.h, w8
 ; CHECK-NEXT:    mov z6.s, #0x7fffffff
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z0.h, z2.h
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.s, p0/m, z1.h
+; CHECK-NEXT:    mov z2.s, #0x80000000
+; CHECK-NEXT:    movprfx z4, z1
+; CHECK-NEXT:    fcvtzs z4.s, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z5, z0
 ; CHECK-NEXT:    fcvtzs z5.s, p0/m, z0.h
-; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z4.h
-; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z4.h
+; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z3.h
+; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z3.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.s, p1/m, z3.s
+; CHECK-NEXT:    sel z3.s, p1, z2.s, z4.s
 ; CHECK-NEXT:    fcmuo p1.h, p0/z, z1.h, z1.h
 ; CHECK-NEXT:    fcmuo p0.h, p0/z, z0.h, z0.h
-; CHECK-NEXT:    sel z3.s, p2, z3.s, z5.s
-; CHECK-NEXT:    sel z0.s, p3, z6.s, z2.s
-; CHECK-NEXT:    sel z1.s, p4, z6.s, z3.s
+; CHECK-NEXT:    sel z2.s, p2, z2.s, z5.s
+; CHECK-NEXT:    sel z0.s, p3, z6.s, z3.s
+; CHECK-NEXT:    sel z1.s, p4, z6.s, z2.s
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.s, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.s, p0/m, #0 // =0x0
@@ -659,8 +659,8 @@ define <vscale x 8 x i32> @test_signed_v8f16_v8i32(<vscale x 8 x half> %f) {
 define <vscale x 4 x i16> @test_signed_v4f16_v4i16(<vscale x 4 x half> %f) {
 ; CHECK-LABEL: test_signed_v4f16_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #63488 // =0xf800
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #30719 // =0x77ff
 ; CHECK-NEXT:    mov z2.h, w8
@@ -682,8 +682,8 @@ define <vscale x 4 x i16> @test_signed_v4f16_v4i16(<vscale x 4 x half> %f) {
 define <vscale x 8 x i16> @test_signed_v8f16_v8i16(<vscale x 8 x half> %f) {
 ; CHECK-LABEL: test_signed_v8f16_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #63488 // =0xf800
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #30719 // =0x77ff
 ; CHECK-NEXT:    mov z2.h, w8
@@ -705,8 +705,8 @@ define <vscale x 8 x i16> @test_signed_v8f16_v8i16(<vscale x 8 x half> %f) {
 define <vscale x 2 x i64> @test_signed_v2f16_v2i64(<vscale x 2 x half> %f) {
 ; CHECK-LABEL: test_signed_v2f16_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #64511 // =0xfbff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0x8000000000000000
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
@@ -734,31 +734,31 @@ define <vscale x 4 x i64> @test_signed_v4f16_v4i64(<vscale x 4 x half> %f) {
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #64511 // =0xfbff
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
 ; CHECK-NEXT:    mov z2.h, w8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
-; CHECK-NEXT:    mov z3.d, #0x8000000000000000
-; CHECK-NEXT:    mov z4.h, w8
+; CHECK-NEXT:    mov z3.h, w8
 ; CHECK-NEXT:    mov z6.d, #0x7fffffffffffffff
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z0.h, z2.h
-; CHECK-NEXT:    movprfx z2, z1
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z1.h
+; CHECK-NEXT:    mov z2.d, #0x8000000000000000
+; CHECK-NEXT:    movprfx z4, z1
+; CHECK-NEXT:    fcvtzs z4.d, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z5, z0
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z0.h
-; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z4.h
-; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z4.h
+; CHECK-NEXT:    fcmgt p3.h, p0/z, z1.h, z3.h
+; CHECK-NEXT:    fcmgt p4.h, p0/z, z0.h, z3.h
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
-; CHECK-NEXT:    mov z2.d, p1/m, z3.d
+; CHECK-NEXT:    sel z3.d, p1, z2.d, z4.d
 ; CHECK-NEXT:    fcmuo p1.h, p0/z, z1.h, z1.h
 ; CHECK-NEXT:    fcmuo p0.h, p0/z, z0.h, z0.h
-; CHECK-NEXT:    sel z3.d, p2, z3.d, z5.d
-; CHECK-NEXT:    sel z0.d, p3, z6.d, z2.d
-; CHECK-NEXT:    sel z1.d, p4, z6.d, z3.d
+; CHECK-NEXT:    sel z2.d, p2, z2.d, z5.d
+; CHECK-NEXT:    sel z0.d, p3, z6.d, z3.d
+; CHECK-NEXT:    sel z1.d, p4, z6.d, z2.d
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    mov z0.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    mov z1.d, p0/m, #0 // =0x0
diff --git a/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll b/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
index c56c0b37888dc..ed352ffec339f 100644
--- a/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fptoui-sat.ll
@@ -82,8 +82,8 @@ define <vscale x 4 x i16> @test_signed_v4f32_v4i16(<vscale x 4 x float> %f) {
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #65280 // =0xff00
 ; CHECK-NEXT:    movk w8, #18303, lsl #16
-; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzu z2.s, p0/m, z0.s
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -102,9 +102,9 @@ define <vscale x 8 x i16> @test_signed_v8f32_v8i16(<vscale x 8 x float> %f) {
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #65280 // =0xff00
 ; CHECK-NEXT:    movk w8, #18303, lsl #16
-; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z1.s, #0.0
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    movprfx z3, z1
 ; CHECK-NEXT:    fcvtzu z3.s, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z4, z0
@@ -146,10 +146,10 @@ define <vscale x 2 x i64> @test_signed_v2f32_v2i64(<vscale x 2 x float> %f) {
 define <vscale x 4 x i64> @test_signed_v4f32_v4i64(<vscale x 4 x float> %f) {
 ; CHECK-LABEL: test_signed_v4f32_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-NEXT:    mov w8, #1602224127 // =0x5f7fffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z4.s, w8
 ; CHECK-NEXT:    fcmge p1.s, p0/z, z2.s, #0.0
 ; CHECK-NEXT:    fcmge p2.s, p0/z, z3.s, #0.0
@@ -186,8 +186,8 @@ define <vscale x 2 x i32> @test_signed_v2f64_v2i32(<vscale x 2 x double> %f) {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
 ; CHECK-NEXT:    movk x8, #16879, lsl #48
-; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    movprfx z2, z0
 ; CHECK-NEXT:    fcvtzu z2.d, p0/m, z0.d
 ; CHECK-NEXT:    not p1.b, p0/z, p1.b
@@ -206,9 +206,9 @@ define <vscale x 4 x i32> @test_signed_v4f64_v4i32(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
 ; CHECK-NEXT:    movk x8, #16879, lsl #48
-; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    movprfx z3, z1
 ; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z4, z0
@@ -241,28 +241,28 @@ define <vscale x 8 x i32> @test_signed_v8f64_v8i32(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281474974613504 // =0xffffffe00000
 ; CHECK-NEXT:    movk x8, #16879, lsl #48
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
 ; CHECK-NEXT:    fcmge p3.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fcmge p4.d, p0/z, z2.d, #0.0
 ; CHECK-NEXT:    movprfx z5, z1
 ; CHECK-NEXT:    fcvtzu z5.d, p0/m, z1.d
+; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z6, z0
 ; CHECK-NEXT:    fcvtzu z6.d, p0/m, z0.d
 ; CHECK-NEXT:    movprfx z7, z3
 ; CHECK-NEXT:    fcvtzu z7.d, p0/m, z3.d
 ; CHECK-NEXT:    movprfx z24, z2
 ; CHECK-NEXT:    fcvtzu z24.d, p0/m, z2.d
+; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z1.d, z4.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z0.d, z4.d
-; CHECK-NEXT:    mov z0.d, #0xffffffff
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
+; CHECK-NEXT:    mov z0.d, #0xffffffff
 ; CHECK-NEXT:    not p3.b, p0/z, p3.b
-; CHECK-NEXT:    not p4.b, p0/z, p4.b
 ; CHECK-NEXT:    mov z5.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    fcmgt p1.d, p0/z, z3.d, z4.d
+; CHECK-NEXT:    not p4.b, p0/z, p4.b
 ; CHECK-NEXT:    fcmgt p0.d, p0/z, z2.d, z4.d
 ; CHECK-NEXT:    mov z6.d, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z7.d, p3/m, #0 // =0x0
@@ -289,9 +289,9 @@ define <vscale x 4 x i16> @test_signed_v4f64_v4i16(<vscale x 4 x double> %f) {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281337537757184 // =0xffe000000000
 ; CHECK-NEXT:    movk x8, #16623, lsl #48
-; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z0.d, #0.0
+; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    movprfx z3, z1
 ; CHECK-NEXT:    fcvtzu z3.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z4, z0
@@ -324,28 +324,28 @@ define <vscale x 8 x i16> @test_signed_v8f64_v8i16(<vscale x 8 x double> %f) {
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x8, #281337537757184 // =0xffe000000000
 ; CHECK-NEXT:    movk x8, #16623, lsl #48
-; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    fcmge p1.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fcmge p2.d, p0/z, z2.d, #0.0
 ; CHECK-NEXT:    fcmge p3.d, p0/z, z1.d, #0.0
 ; CHECK-NEXT:    fcmge p4.d, p0/z, z0.d, #0.0
 ; CHECK-NEXT:    movprfx z5, z3
 ; CHECK-NEXT:    fcvtzu z5.d, p0/m, z3.d
+; CHECK-NEXT:    mov z4.d, x8
 ; CHECK-NEXT:    movprfx z6, z2
 ; CHECK-NEXT:    fcvtzu z6.d, p0/m, z2.d
 ; CHECK-NEXT:    movprfx z7, z1
 ; CHECK-NEXT:    fcvtzu z7.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z24, z0
 ; CHECK-NEXT:    fcvtzu z24.d, p0/m, z0.d
+; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    fcmgt p5.d, p0/z, z3.d, z4.d
 ; CHECK-NEXT:    fcmgt p6.d, p0/z, z2.d, z4.d
-; CHECK-NEXT:    mov z2.d, #65535 // =0xffff
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
 ; CHECK-NEXT:    not p2.b, p0/z, p2.b
+; CHECK-NEXT:    mov z2.d, #65535 // =0xffff
 ; CHECK-NEXT:    not p3.b, p0/z, p3.b
-; CHECK-NEXT:    not p4.b, p0/z, p4.b
 ; CHECK-NEXT:    mov z5.d, p1/m, #0 // =0x0
 ; CHECK-NEXT:    fcmgt p1.d, p0/z, z1.d, z4.d
+; CHECK-NEXT:    not p4.b, p0/z, p4.b
 ; CHECK-NEXT:    fcmgt p0.d, p0/z, z0.d, z4.d
 ; CHECK-NEXT:    mov z6.d, p2/m, #0 // =0x0
 ; CHECK-NEXT:    mov z7.d, p3/m, #0 // =0x0
@@ -465,10 +465,10 @@ define <vscale x 4 x i32> @test_signed_v4f16_v4i32(<vscale x 4 x half> %f) {
 define <vscale x 8 x i32> @test_signed_v8f16_v8i32(<vscale x 8 x half> %f) {
 ; CHECK-LABEL: test_signed_v8f16_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z2.s, z0.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z4.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
@@ -549,10 +549,10 @@ define <vscale x 2 x i64> @test_signed_v2f16_v2i64(<vscale x 2 x half> %f) {
 define <vscale x 4 x i64> @test_signed_v4f16_v4i64(<vscale x 4 x half> %f) {
 ; CHECK-LABEL: test_signed_v4f16_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-NEXT:    mov w8, #31743 // =0x7bff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z4.h, w8
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fcmge p2.h, p0/z, z3.h, #0.0
diff --git a/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll b/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
index 5c4c9463528b8..ad6371f78ec08 100644
--- a/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
+++ b/llvm/test/CodeGen/AArch64/sve-gather-scatter-addr-opts.ll
@@ -71,12 +71,12 @@ define <vscale x 4 x i8> @gather_i8_index_offset_8(ptr %base, i64 %offset, <vsca
 define void @scatter_f16_index_offset_var(ptr %base, i64 %offset, i64 %scale, <vscale x 4 x i1> %pg, <vscale x 4 x half> %data) #0 {
 ; CHECK-LABEL: scatter_f16_index_offset_var:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x1
-; CHECK-NEXT:    punpklo p2.h, p0.b
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    uunpklo z3.d, z0.s
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
+; CHECK-NEXT:    punpklo p2.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    movprfx z4, z2
 ; CHECK-NEXT:    mla z4.d, p1/m, z1.d, z2.d
@@ -101,16 +101,16 @@ define void @scatter_f16_index_offset_var(ptr %base, i64 %offset, i64 %scale, <v
 define void @scatter_i8_index_offset_maximum_plus_one(ptr %base, i64 %offset, <vscale x 4 x i1> %pg, <vscale x 4 x i8> %data) #0 {
 ; CHECK-LABEL: scatter_i8_index_offset_maximum_plus_one:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    mov w8, #33554432 // =0x2000000
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
-; CHECK-NEXT:    index z1.d, #0, x8
 ; CHECK-NEXT:    rdvl x9, #1
-; CHECK-NEXT:    add x8, x0, x1
+; CHECK-NEXT:    index z1.d, #0, x8
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    lsr x9, x9, #4
+; CHECK-NEXT:    add x8, x0, x1
 ; CHECK-NEXT:    mov w10, #67108864 // =0x4000000
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1b { z2.d }, p1, [x8, z1.d]
 ; CHECK-NEXT:    madd x8, x9, x10, x8
 ; CHECK-NEXT:    st1b { z0.d }, p0, [x8, z1.d]
@@ -131,17 +131,17 @@ define void @scatter_i8_index_offset_maximum_plus_one(ptr %base, i64 %offset, <v
 define void @scatter_i8_index_offset_minimum_minus_one(ptr %base, i64 %offset, <vscale x 4 x i1> %pg, <vscale x 4 x i8> %data) #0 {
 ; CHECK-LABEL: scatter_i8_index_offset_minimum_minus_one:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    mov x8, #-33554433 // =0xfffffffffdffffff
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
-; CHECK-NEXT:    index z1.d, #0, x8
 ; CHECK-NEXT:    rdvl x9, #1
-; CHECK-NEXT:    mov x10, #-2 // =0xfffffffffffffffe
+; CHECK-NEXT:    index z1.d, #0, x8
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    lsr x9, x9, #4
+; CHECK-NEXT:    mov x10, #-2 // =0xfffffffffffffffe
 ; CHECK-NEXT:    add x8, x0, x1
+; CHECK-NEXT:    uunpkhi z0.d, z0.s
 ; CHECK-NEXT:    movk x10, #64511, lsl #16
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
-; CHECK-NEXT:    uunpkhi z0.d, z0.s
 ; CHECK-NEXT:    st1b { z2.d }, p1, [x8, z1.d]
 ; CHECK-NEXT:    madd x8, x9, x10, x8
 ; CHECK-NEXT:    st1b { z0.d }, p0, [x8, z1.d]
@@ -162,16 +162,16 @@ define void @scatter_i8_index_offset_minimum_minus_one(ptr %base, i64 %offset, <
 define void @scatter_i8_index_stride_too_big(ptr %base, i64 %offset, <vscale x 4 x i1> %pg, <vscale x 4 x i8> %data) #0 {
 ; CHECK-LABEL: scatter_i8_index_stride_too_big:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    mov x8, #4611686018427387904 // =0x4000000000000000
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
-; CHECK-NEXT:    index z1.d, #0, x8
 ; CHECK-NEXT:    rdvl x9, #1
-; CHECK-NEXT:    add x8, x0, x1
+; CHECK-NEXT:    index z1.d, #0, x8
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    lsr x9, x9, #4
+; CHECK-NEXT:    add x8, x0, x1
 ; CHECK-NEXT:    mov x10, #-9223372036854775808 // =0x8000000000000000
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    uunpkhi z0.d, z0.s
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1b { z2.d }, p1, [x8, z1.d]
 ; CHECK-NEXT:    madd x8, x9, x10, x8
 ; CHECK-NEXT:    st1b { z0.d }, p0, [x8, z1.d]
diff --git a/llvm/test/CodeGen/AArch64/sve-hadd.ll b/llvm/test/CodeGen/AArch64/sve-hadd.ll
index c73370d50287b..f90aef8daa5dc 100644
--- a/llvm/test/CodeGen/AArch64/sve-hadd.ll
+++ b/llvm/test/CodeGen/AArch64/sve-hadd.ll
@@ -129,9 +129,9 @@ define <vscale x 2 x i32> @haddu_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i32
 ;
 ; SVE2-LABEL: haddu_v2i32:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    and z0.d, z0.d, #0xffffffff
 ; SVE2-NEXT:    and z1.d, z1.d, #0xffffffff
+; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    uhadd z0.d, p0/m, z0.d, z1.d
 ; SVE2-NEXT:    ret
 entry:
@@ -274,9 +274,9 @@ define <vscale x 2 x i16> @haddu_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i16
 ;
 ; SVE2-LABEL: haddu_v2i16:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    and z0.d, z0.d, #0xffff
 ; SVE2-NEXT:    and z1.d, z1.d, #0xffff
+; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    uhadd z0.d, p0/m, z0.d, z1.d
 ; SVE2-NEXT:    ret
 entry:
@@ -343,9 +343,9 @@ define <vscale x 4 x i16> @haddu_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i16
 ;
 ; SVE2-LABEL: haddu_v4i16:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    and z0.s, z0.s, #0xffff
 ; SVE2-NEXT:    and z1.s, z1.s, #0xffff
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    uhadd z0.s, p0/m, z0.s, z1.s
 ; SVE2-NEXT:    ret
 entry:
@@ -488,9 +488,9 @@ define <vscale x 4 x i8> @haddu_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %s
 ;
 ; SVE2-LABEL: haddu_v4i8:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    and z0.s, z0.s, #0xff
 ; SVE2-NEXT:    and z1.s, z1.s, #0xff
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    uhadd z0.s, p0/m, z0.s, z1.s
 ; SVE2-NEXT:    ret
 entry:
@@ -557,9 +557,9 @@ define <vscale x 8 x i8> @haddu_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %s
 ;
 ; SVE2-LABEL: haddu_v8i8:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.h
 ; SVE2-NEXT:    and z0.h, z0.h, #0xff
 ; SVE2-NEXT:    and z1.h, z1.h, #0xff
+; SVE2-NEXT:    ptrue p0.h
 ; SVE2-NEXT:    uhadd z0.h, p0/m, z0.h, z1.h
 ; SVE2-NEXT:    ret
 entry:
@@ -787,9 +787,9 @@ define <vscale x 2 x i32> @rhaddu_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i3
 ;
 ; SVE2-LABEL: rhaddu_v2i32:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    and z0.d, z0.d, #0xffffffff
 ; SVE2-NEXT:    and z1.d, z1.d, #0xffffffff
+; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    urhadd z0.d, p0/m, z0.d, z1.d
 ; SVE2-NEXT:    ret
 entry:
@@ -936,9 +936,9 @@ define <vscale x 2 x i16> @rhaddu_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i1
 ;
 ; SVE2-LABEL: rhaddu_v2i16:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    and z0.d, z0.d, #0xffff
 ; SVE2-NEXT:    and z1.d, z1.d, #0xffff
+; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    urhadd z0.d, p0/m, z0.d, z1.d
 ; SVE2-NEXT:    ret
 entry:
@@ -1014,9 +1014,9 @@ define <vscale x 4 x i16> @rhaddu_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i1
 ;
 ; SVE2-LABEL: rhaddu_v4i16:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    and z0.s, z0.s, #0xffff
 ; SVE2-NEXT:    and z1.s, z1.s, #0xffff
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    urhadd z0.s, p0/m, z0.s, z1.s
 ; SVE2-NEXT:    ret
 entry:
@@ -1163,9 +1163,9 @@ define <vscale x 4 x i8> @rhaddu_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %
 ;
 ; SVE2-LABEL: rhaddu_v4i8:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    and z0.s, z0.s, #0xff
 ; SVE2-NEXT:    and z1.s, z1.s, #0xff
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    urhadd z0.s, p0/m, z0.s, z1.s
 ; SVE2-NEXT:    ret
 entry:
@@ -1241,9 +1241,9 @@ define <vscale x 8 x i8> @rhaddu_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %
 ;
 ; SVE2-LABEL: rhaddu_v8i8:
 ; SVE2:       // %bb.0: // %entry
-; SVE2-NEXT:    ptrue p0.h
 ; SVE2-NEXT:    and z0.h, z0.h, #0xff
 ; SVE2-NEXT:    and z1.h, z1.h, #0xff
+; SVE2-NEXT:    ptrue p0.h
 ; SVE2-NEXT:    urhadd z0.h, p0/m, z0.h, z1.h
 ; SVE2-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/sve-implicit-zero-filling.ll b/llvm/test/CodeGen/AArch64/sve-implicit-zero-filling.ll
index e20399de70bf8..73bbee094827e 100644
--- a/llvm/test/CodeGen/AArch64/sve-implicit-zero-filling.ll
+++ b/llvm/test/CodeGen/AArch64/sve-implicit-zero-filling.ll
@@ -175,14 +175,14 @@ define <vscale x 2 x i64> @uminv_zero_fill(<vscale x 2 x i1> %pg, <vscale x 2 x
 define <vscale x 2 x i64> @zero_fill_non_zero_index(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a) #0 {
 ; CHECK-LABEL: zero_fill_non_zero_index:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    uminv d3, p0, z0.d
 ; CHECK-NEXT:    mov z2.d, x8
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, #0 // =0x0
-; CHECK-NEXT:    fmov x8, d3
 ; CHECK-NEXT:    cmpeq p0.d, p1/z, z1.d, z2.d
+; CHECK-NEXT:    fmov x8, d3
 ; CHECK-NEXT:    mov z0.d, p0/m, x8
 ; CHECK-NEXT:    ret
   %t1 = call i64 @llvm.aarch64.sve.uminv.nxv2i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a)
@@ -210,11 +210,11 @@ define <vscale x 4 x i64> @zero_fill_type_mismatch(<vscale x 2 x i1> %pg, <vscal
 define <vscale x 2 x i64> @zero_fill_no_zero_upper_lanes(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a) #0 {
 ; CHECK-LABEL: zero_fill_no_zero_upper_lanes:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d, vl1
 ; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z0.d
 ; CHECK-NEXT:    mov z1.d, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    fmov x8, d0
-; CHECK-NEXT:    mov z1.d, p1/m, x8
+; CHECK-NEXT:    mov z1.d, p0/m, x8
 ; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
   %t1 = call <vscale x 2 x i64> @llvm.aarch64.sve.umin.nxv2i64(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a, <vscale x 2 x i64> %a)
diff --git a/llvm/test/CodeGen/AArch64/sve-insert-element.ll b/llvm/test/CodeGen/AArch64/sve-insert-element.ll
index 2aa298f6d9173..7344964f13bba 100644
--- a/llvm/test/CodeGen/AArch64/sve-insert-element.ll
+++ b/llvm/test/CodeGen/AArch64/sve-insert-element.ll
@@ -48,8 +48,8 @@ define <vscale x 2 x i64> @test_lane0_2xi64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x double> @test_lane0_2xf64(<vscale x 2 x double> %a) {
 ; CHECK-LABEL: test_lane0_2xf64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    fmov d1, #1.00000000
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    mov z0.d, p0/m, z1.d
 ; CHECK-NEXT:    ret
   %b = insertelement <vscale x 2 x double> %a, double 1.0, i32 0
@@ -59,8 +59,8 @@ define <vscale x 2 x double> @test_lane0_2xf64(<vscale x 2 x double> %a) {
 define <vscale x 4 x float> @test_lane0_4xf32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: test_lane0_4xf32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    fmov s1, #1.00000000
+; CHECK-NEXT:    ptrue p0.s, vl1
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
   %b = insertelement <vscale x 4 x float> %a, float 1.0, i32 0
@@ -70,8 +70,8 @@ define <vscale x 4 x float> @test_lane0_4xf32(<vscale x 4 x float> %a) {
 define <vscale x 8 x half> @test_lane0_8xf16(<vscale x 8 x half> %a) {
 ; CHECK-LABEL: test_lane0_8xf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    fmov h1, #1.00000000
+; CHECK-NEXT:    ptrue p0.h, vl1
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
   %b = insertelement <vscale x 8 x half> %a, half 1.0, i32 0
@@ -93,9 +93,9 @@ define <vscale x 8 x bfloat> @test_lane0_8xbf16(<vscale x 8 x bfloat> %a, bfloat
 define <vscale x 2 x i64> @test_lane4_2xi64(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: test_lane4_2xi64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #4 // =0x4
 ; CHECK-NEXT:    index z1.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    mov w8, #30 // =0x1e
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
@@ -109,9 +109,9 @@ define <vscale x 2 x i64> @test_lane4_2xi64(<vscale x 2 x i64> %a) {
 define <vscale x 8 x half> @test_lane9_8xf16(<vscale x 8 x half> %a) {
 ; CHECK-LABEL: test_lane9_8xf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #9 // =0x9
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    fmov h1, #1.00000000
@@ -124,9 +124,9 @@ define <vscale x 8 x half> @test_lane9_8xf16(<vscale x 8 x half> %a) {
 define <vscale x 8 x bfloat> @test_lane9_8xbf16(<vscale x 8 x bfloat> %a, bfloat %x) {
 ; CHECK-LABEL: test_lane9_8xbf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #9 // =0x9
 ; CHECK-NEXT:    index z2.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z3.h, w8
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z2.h, z3.h
 ; CHECK-NEXT:    mov z0.h, p0/m, h1
@@ -138,9 +138,9 @@ define <vscale x 8 x bfloat> @test_lane9_8xbf16(<vscale x 8 x bfloat> %a, bfloat
 define <vscale x 16 x i8> @test_lane1_16xi8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: test_lane1_16xi8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w8
 ; CHECK-NEXT:    mov w8, #30 // =0x1e
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z1.b, z2.b
@@ -153,9 +153,9 @@ define <vscale x 16 x i8> @test_lane1_16xi8(<vscale x 16 x i8> %a) {
 define <vscale x 16 x i8> @test_lanex_16xi8(<vscale x 16 x i8> %a, i32 %x) {
 ; CHECK-LABEL: test_lanex_16xi8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    mov w8, w0
 ; CHECK-NEXT:    index z1.b, #0, #1
+; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w8
 ; CHECK-NEXT:    mov w8, #30 // =0x1e
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z1.b, z2.b
@@ -179,9 +179,9 @@ define <vscale x 4 x i32> @extract_insert_4xi32(<vscale x 4 x i32> %a) {
 define <vscale x 8 x i16> @test_lane6_undef_8xi16(i16 %a) {
 ; CHECK-LABEL: test_lane6_undef_8xi16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #6 // =0x6
 ; CHECK-NEXT:    index z0.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    mov z0.h, p0/m, w0
@@ -202,8 +202,8 @@ define <vscale x 16 x i8> @test_lane0_undef_16xi8(i8 %a) {
 define <vscale x 16 x i8> @test_insert0_of_extract0_16xi8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
 ; CHECK-LABEL: test_insert0_of_extract0_16xi8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl1
 ; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    ptrue p0.b, vl1
 ; CHECK-NEXT:    mov z0.b, p0/m, w8
 ; CHECK-NEXT:    ret
   %c = extractelement <vscale x 16 x i8> %b, i32 0
@@ -215,12 +215,12 @@ define <vscale x 16 x i8> @test_insert64_of_extract64_16xi8(<vscale x 16 x i8> %
 ; CHECK-LABEL: test_insert64_of_extract64_16xi8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #64 // =0x40
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    whilels p0.b, xzr, x8
 ; CHECK-NEXT:    mov z2.b, w8
 ; CHECK-NEXT:    lastb w9, p0, z1.b
 ; CHECK-NEXT:    index z1.b, #0, #1
-; CHECK-NEXT:    cmpeq p0.b, p1/z, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    cmpeq p0.b, p0/z, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, w9
 ; CHECK-NEXT:    ret
   %c = extractelement <vscale x 16 x i8> %b, i32 64
@@ -231,9 +231,9 @@ define <vscale x 16 x i8> @test_insert64_of_extract64_16xi8(<vscale x 16 x i8> %
 define <vscale x 16 x i8> @test_insert3_of_extract1_16xi8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
 ; CHECK-LABEL: test_insert3_of_extract1_16xi8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z2.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z3.b, w8
 ; CHECK-NEXT:    umov w8, v1.b[1]
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z2.b, z3.b
@@ -329,9 +329,9 @@ define <vscale x 2 x double> @test_insert_into_undef_nxv2f64(double %a) {
 define <vscale x 2 x half> @test_insert_with_index_nxv2f16(half %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv2f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -342,9 +342,9 @@ define <vscale x 2 x half> @test_insert_with_index_nxv2f16(half %h, i64 %idx) {
 define <vscale x 4 x half> @test_insert_with_index_nxv4f16(half %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    index z1.s, #0, #1
 ; CHECK-NEXT:    mov z2.s, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -355,9 +355,9 @@ define <vscale x 4 x half> @test_insert_with_index_nxv4f16(half %h, i64 %idx) {
 define <vscale x 8 x half> @test_insert_with_index_nxv8f16(half %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    index z1.h, #0, #1
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -368,9 +368,9 @@ define <vscale x 8 x half> @test_insert_with_index_nxv8f16(half %h, i64 %idx) {
 define <vscale x 2 x bfloat> @test_insert_with_index_nxv2bf16(bfloat %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv2bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -381,9 +381,9 @@ define <vscale x 2 x bfloat> @test_insert_with_index_nxv2bf16(bfloat %h, i64 %id
 define <vscale x 4 x bfloat> @test_insert_with_index_nxv4bf16(bfloat %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv4bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    index z1.s, #0, #1
 ; CHECK-NEXT:    mov z2.s, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -394,9 +394,9 @@ define <vscale x 4 x bfloat> @test_insert_with_index_nxv4bf16(bfloat %h, i64 %id
 define <vscale x 8 x bfloat> @test_insert_with_index_nxv8bf16(bfloat %h, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv8bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    index z1.h, #0, #1
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, h0
 ; CHECK-NEXT:    ret
@@ -407,9 +407,9 @@ define <vscale x 8 x bfloat> @test_insert_with_index_nxv8bf16(bfloat %h, i64 %id
 define <vscale x 2 x float> @test_insert_with_index_nxv2f32(float %f, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.s, p0/m, s0
 ; CHECK-NEXT:    ret
@@ -420,9 +420,9 @@ define <vscale x 2 x float> @test_insert_with_index_nxv2f32(float %f, i64 %idx)
 define <vscale x 4 x float> @test_insert_with_index_nxv4f32(float %f, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    index z1.s, #0, #1
 ; CHECK-NEXT:    mov z2.s, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, s0
 ; CHECK-NEXT:    ret
@@ -433,9 +433,9 @@ define <vscale x 4 x float> @test_insert_with_index_nxv4f32(float %f, i64 %idx)
 define <vscale x 2 x double> @test_insert_with_index_nxv2f64(double %d, i64 %idx) {
 ; CHECK-LABEL: test_insert_with_index_nxv2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    index z1.d, #0, #1
 ; CHECK-NEXT:    mov z2.d, x0
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.d, p0/m, d0
 ; CHECK-NEXT:    ret
@@ -447,11 +447,11 @@ define <vscale x 2 x double> @test_insert_with_index_nxv2f64(double %d, i64 %idx
 define <vscale x 2 x i1> @test_predicate_insert_2xi1_immediate (<vscale x 2 x i1> %val, i1 %elt) {
 ; CHECK-LABEL: test_predicate_insert_2xi1_immediate:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d, vl1
 ; CHECK-NEXT:    mov z0.d, p0/z, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d, vl1
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
+; CHECK-NEXT:    mov z0.d, p0/m, x0
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    mov z0.d, p1/m, x0
 ; CHECK-NEXT:    and z0.d, z0.d, #0x1
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ret
@@ -462,9 +462,9 @@ define <vscale x 2 x i1> @test_predicate_insert_2xi1_immediate (<vscale x 2 x i1
 define <vscale x 4 x i1> @test_predicate_insert_4xi1_immediate (<vscale x 4 x i1> %val, i1 %elt) {
 ; CHECK-LABEL: test_predicate_insert_4xi1_immediate:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov w8, #2 // =0x2
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p2.s, p1/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
@@ -479,9 +479,9 @@ define <vscale x 4 x i1> @test_predicate_insert_4xi1_immediate (<vscale x 4 x i1
 define <vscale x 8 x i1> @test_predicate_insert_8xi1_immediate (<vscale x 8 x i1> %val, i32 %idx) {
 ; CHECK-LABEL: test_predicate_insert_8xi1_immediate:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    mov w8, w0
 ; CHECK-NEXT:    index z0.h, #0, #1
+; CHECK-NEXT:    mov w8, w0
+; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    cmpeq p2.h, p1/z, z0.h, z1.h
@@ -497,9 +497,9 @@ define <vscale x 8 x i1> @test_predicate_insert_8xi1_immediate (<vscale x 8 x i1
 define <vscale x 16 x i1> @test_predicate_insert_16xi1_immediate (<vscale x 16 x i1> %val) {
 ; CHECK-LABEL: test_predicate_insert_16xi1_immediate:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    mov w8, #4 // =0x4
 ; CHECK-NEXT:    index z0.b, #0, #1
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    mov z1.b, w8
 ; CHECK-NEXT:    mov w8, wzr
 ; CHECK-NEXT:    cmpeq p2.b, p1/z, z0.b, z1.b
@@ -516,9 +516,9 @@ define <vscale x 16 x i1> @test_predicate_insert_16xi1_immediate (<vscale x 16 x
 define <vscale x 2 x i1> @test_predicate_insert_2xi1(<vscale x 2 x i1> %val, i1 %elt, i32 %idx) {
 ; CHECK-LABEL: test_predicate_insert_2xi1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
-; CHECK-NEXT:    mov w8, w1
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    mov w8, w1
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    cmpeq p2.d, p1/z, z0.d, z1.d
@@ -534,9 +534,9 @@ define <vscale x 2 x i1> @test_predicate_insert_2xi1(<vscale x 2 x i1> %val, i1
 define <vscale x 4 x i1> @test_predicate_insert_4xi1(<vscale x 4 x i1> %val, i1 %elt, i32 %idx) {
 ; CHECK-LABEL: test_predicate_insert_4xi1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
-; CHECK-NEXT:    mov w8, w1
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    mov w8, w1
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    cmpeq p2.s, p1/z, z0.s, z1.s
 ; CHECK-NEXT:    mov z0.s, p0/z, #1 // =0x1
@@ -550,9 +550,9 @@ define <vscale x 4 x i1> @test_predicate_insert_4xi1(<vscale x 4 x i1> %val, i1
 define <vscale x 8 x i1> @test_predicate_insert_8xi1(<vscale x 8 x i1> %val, i1 %elt, i32 %idx) {
 ; CHECK-LABEL: test_predicate_insert_8xi1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    mov w8, w1
 ; CHECK-NEXT:    index z0.h, #0, #1
+; CHECK-NEXT:    mov w8, w1
+; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    cmpeq p2.h, p1/z, z0.h, z1.h
 ; CHECK-NEXT:    mov z0.h, p0/z, #1 // =0x1
@@ -567,9 +567,9 @@ define <vscale x 8 x i1> @test_predicate_insert_8xi1(<vscale x 8 x i1> %val, i1
 define <vscale x 16 x i1> @test_predicate_insert_16xi1(<vscale x 16 x i1> %val, i1 %elt, i32 %idx) {
 ; CHECK-LABEL: test_predicate_insert_16xi1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.b
-; CHECK-NEXT:    mov w8, w1
 ; CHECK-NEXT:    index z0.b, #0, #1
+; CHECK-NEXT:    mov w8, w1
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    mov z1.b, w8
 ; CHECK-NEXT:    cmpeq p2.b, p1/z, z0.b, z1.b
 ; CHECK-NEXT:    mov z0.b, p0/z, #1 // =0x1
@@ -589,24 +589,24 @@ define <vscale x 32 x i1> @test_predicate_insert_32xi1(<vscale x 32 x i1> %val,
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
-; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    rdvl x8, #2
 ; CHECK-NEXT:    mov z0.b, p1/z, #1 // =0x1
 ; CHECK-NEXT:    mov z1.b, p0/z, #1 // =0x1
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    mov w9, w1
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cmp x9, x8
 ; CHECK-NEXT:    csel x8, x9, x8, lo
 ; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    st1b { z0.b }, p2, [sp, #1, mul vl]
-; CHECK-NEXT:    st1b { z1.b }, p2, [sp]
+; CHECK-NEXT:    st1b { z0.b }, p1, [sp, #1, mul vl]
+; CHECK-NEXT:    st1b { z1.b }, p1, [sp]
 ; CHECK-NEXT:    strb w0, [x9, x8]
-; CHECK-NEXT:    ld1b { z0.b }, p2/z, [sp]
-; CHECK-NEXT:    ld1b { z1.b }, p2/z, [sp, #1, mul vl]
+; CHECK-NEXT:    ld1b { z0.b }, p1/z, [sp]
+; CHECK-NEXT:    ld1b { z1.b }, p1/z, [sp, #1, mul vl]
 ; CHECK-NEXT:    and z0.b, z0.b, #0x1
 ; CHECK-NEXT:    and z1.b, z1.b, #0x1
-; CHECK-NEXT:    cmpne p0.b, p2/z, z0.b, #0
-; CHECK-NEXT:    cmpne p1.b, p2/z, z1.b, #0
+; CHECK-NEXT:    cmpne p0.b, p1/z, z0.b, #0
+; CHECK-NEXT:    cmpne p1.b, p1/z, z1.b, #0
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    .cfi_def_cfa wsp, 16
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sve-insert-vector.ll b/llvm/test/CodeGen/AArch64/sve-insert-vector.ll
index 4a5e272582d8e..5efe9e2819d5e 100644
--- a/llvm/test/CodeGen/AArch64/sve-insert-vector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-insert-vector.ll
@@ -17,15 +17,15 @@ define <vscale x 2 x i64> @insert_v2i64_nxv2i64_idx2(<vscale x 2 x i64> %vec, <2
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntd x8
 ; CHECK-NEXT:    mov w9, #2 // =0x2
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sub x8, x8, #2
 ; CHECK-NEXT:    cmp x8, #2
+; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #3
-; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -51,15 +51,15 @@ define <vscale x 4 x i32> @insert_v4i32_nxv4i32_idx4(<vscale x 4 x i32> %vec, <4
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cntw x8
 ; CHECK-NEXT:    mov w9, #4 // =0x4
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sub x8, x8, #4
 ; CHECK-NEXT:    cmp x8, #4
+; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #2
-; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -85,15 +85,15 @@ define <vscale x 8 x i16> @insert_v8i16_nxv8i16_idx8(<vscale x 8 x i16> %vec, <8
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cnth x8
 ; CHECK-NEXT:    mov w9, #8 // =0x8
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    sub x8, x8, #8
 ; CHECK-NEXT:    cmp x8, #8
+; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    lsl x8, x8, #1
-; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
 ; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1h { z0.h }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -119,15 +119,15 @@ define <vscale x 16 x i8> @insert_v16i8_nxv16i8_idx16(<vscale x 16 x i8> %vec, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w9, #16 // =0x10
 ; CHECK-NEXT:    sub x8, x8, #16
-; CHECK-NEXT:    mov x10, sp
 ; CHECK-NEXT:    cmp x8, #16
-; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
-; CHECK-NEXT:    str q1, [x10, x8]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    mov x9, sp
+; CHECK-NEXT:    str q1, [x9, x8]
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [sp]
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -239,8 +239,8 @@ define void @insert_v2i64_nxv16i64_lo2(ptr %psv, ptr %out) uwtable {
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldr q0, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    str q0, [sp, #16]
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
 ; CHECK-NEXT:    ld1d { z1.d }, p0/z, [sp]
diff --git a/llvm/test/CodeGen/AArch64/sve-int-arith-imm.ll b/llvm/test/CodeGen/AArch64/sve-int-arith-imm.ll
index c0ddceb42e1d0..52bd79e7a7e60 100644
--- a/llvm/test/CodeGen/AArch64/sve-int-arith-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-int-arith-imm.ll
@@ -55,8 +55,8 @@ define <vscale x 8 x i16> @smax_i16_neg(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @smax_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: smax_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
@@ -93,8 +93,8 @@ define <vscale x 4 x i32> @smax_i32_neg(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @smax_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smax_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #-129 // =0xffffffffffffff7f
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 -129, i32 0
@@ -131,8 +131,8 @@ define <vscale x 2 x i64> @smax_i64_neg(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @smax_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smax_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
@@ -196,8 +196,8 @@ define <vscale x 8 x i16> @smin_i16_neg(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @smin_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: smin_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
@@ -234,8 +234,8 @@ define <vscale x 4 x i32> @smin_i32_neg(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @smin_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smin_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #-129 // =0xffffffffffffff7f
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 -129, i32 0
@@ -272,8 +272,8 @@ define <vscale x 2 x i64> @smin_i64_neg(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @smin_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smin_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
@@ -325,8 +325,8 @@ define <vscale x 8 x i16> @umax_i16_pos(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @umax_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: umax_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
@@ -351,8 +351,8 @@ define <vscale x 4 x i32> @umax_i32_pos(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @umax_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umax_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #257 // =0x101
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -378,8 +378,8 @@ define <vscale x 2 x i64> @umax_i64_pos(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @umax_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umax_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
@@ -431,8 +431,8 @@ define <vscale x 8 x i16> @umin_i16_pos(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @umin_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: umin_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 257, i32 0
@@ -457,8 +457,8 @@ define <vscale x 4 x i32> @umin_i32_pos(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @umin_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umin_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #257 // =0x101
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -484,8 +484,8 @@ define <vscale x 2 x i64> @umin_i64_pos(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @umin_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umin_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 65535, i32 0
@@ -589,8 +589,8 @@ define <vscale x 2 x i64> @mul_i64_pos(<vscale x 2 x i64> %a) {
 define <vscale x 8 x i16> @mul_i16_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: mul_i16_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, #255 // =0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 8 x i16> undef, i16 255, i32 0
@@ -602,8 +602,8 @@ define <vscale x 8 x i16> @mul_i16_range(<vscale x 8 x i16> %a) {
 define <vscale x 4 x i32> @mul_i32_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: mul_i32_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #255 // =0xff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 4 x i32> undef, i32 255, i32 0
@@ -615,8 +615,8 @@ define <vscale x 4 x i32> @mul_i32_range(<vscale x 4 x i32> %a) {
 define <vscale x 2 x i64> @mul_i64_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: mul_i64_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #255 // =0xff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %elt = insertelement <vscale x 2 x i64> undef, i64 255, i32 0
@@ -766,8 +766,8 @@ define <vscale x 2 x i64> @lsr_i64(<vscale x 2 x i64> %a){
 define <vscale x 4 x i32> @sdiv_const(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: sdiv_const:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #3 // =0x3
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -778,8 +778,8 @@ entry:
 define <vscale x 4 x i32> @udiv_const(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: udiv_const:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #3 // =0x3
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/sve-int-arith.ll b/llvm/test/CodeGen/AArch64/sve-int-arith.ll
index fc2672f8c80a8..5f92dee3b5305 100644
--- a/llvm/test/CodeGen/AArch64/sve-int-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-int-arith.ll
@@ -532,8 +532,8 @@ define <vscale x 2 x i64> @mls_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b,
  define <vscale x 2 x i64> @muladd_i64_positiveAddend(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
 ; CHECK-LABEL: muladd_i64_positiveAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mad z0.d, p0/m, z1.d, z2.d
 ; CHECK-NEXT:    ret
 {
@@ -545,8 +545,8 @@ define <vscale x 2 x i64> @mls_i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b,
 define <vscale x 2 x i64> @muladd_i64_negativeAddend(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
 ; CHECK-LABEL: muladd_i64_negativeAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, #0xffffffff00000001
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mad z0.d, p0/m, z1.d, z2.d
 ; CHECK-NEXT:    ret
 {
@@ -559,8 +559,8 @@ define <vscale x 2 x i64> @muladd_i64_negativeAddend(<vscale x 2 x i64> %a, <vsc
 define <vscale x 4 x i32> @muladd_i32_positiveAddend(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b)
 ; CHECK-LABEL: muladd_i32_positiveAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #0x10000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mad z0.s, p0/m, z1.s, z2.s
 ; CHECK-NEXT:    ret
 {
@@ -572,8 +572,8 @@ define <vscale x 4 x i32> @muladd_i32_positiveAddend(<vscale x 4 x i32> %a, <vsc
 define <vscale x 4 x i32> @muladd_i32_negativeAddend(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b)
 ; CHECK-LABEL: muladd_i32_negativeAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #0xffff0000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mad z0.s, p0/m, z1.s, z2.s
 ; CHECK-NEXT:    ret
 {
@@ -585,8 +585,8 @@ define <vscale x 4 x i32> @muladd_i32_negativeAddend(<vscale x 4 x i32> %a, <vsc
 define <vscale x 8 x i16> @muladd_i16_positiveAddend(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
 ; CHECK-LABEL: muladd_i16_positiveAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, #255 // =0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mad z0.h, p0/m, z1.h, z2.h
 ; CHECK-NEXT:    ret
 {
@@ -598,8 +598,8 @@ define <vscale x 8 x i16> @muladd_i16_positiveAddend(<vscale x 8 x i16> %a, <vsc
 define <vscale x 8 x i16> @muladd_i16_negativeAddend(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
 ; CHECK-LABEL: muladd_i16_negativeAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, #-255 // =0xffffffffffffff01
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mad z0.h, p0/m, z1.h, z2.h
 ; CHECK-NEXT:    ret
 {
@@ -611,8 +611,8 @@ define <vscale x 8 x i16> @muladd_i16_negativeAddend(<vscale x 8 x i16> %a, <vsc
 define <vscale x 16 x i8> @muladd_i8_positiveAddend(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 ; CHECK-LABEL: muladd_i8_positiveAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, #15 // =0xf
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mad z0.b, p0/m, z1.b, z2.b
 ; CHECK-NEXT:    ret
 {
@@ -624,8 +624,8 @@ define <vscale x 16 x i8> @muladd_i8_positiveAddend(<vscale x 16 x i8> %a, <vsca
 define <vscale x 16 x i8> @muladd_i8_negativeAddend(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
 ; CHECK-LABEL: muladd_i8_negativeAddend:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, #-15 // =0xfffffffffffffff1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mad z0.b, p0/m, z1.b, z2.b
 ; CHECK-NEXT:    ret
 {
@@ -748,8 +748,8 @@ define <vscale x 16 x i8> @mulsub_i8_negativeAddend(<vscale x 16 x i8> %a, <vsca
 define <vscale x 8 x i16> @multiple_fused_ops(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
 ; CHECK-LABEL: multiple_fused_ops:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #200 // =0xc8
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    mla z2.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z2.h
@@ -770,8 +770,8 @@ define void @mad_in_loop(ptr %dst, ptr %src1, ptr %src2, i32 %n) {
 ; CHECK-NEXT:    b.lt .LBB70_3
 ; CHECK-NEXT:  // %bb.1: // %for.body.preheader
 ; CHECK-NEXT:    mov w9, w3
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    whilelo p1.s, xzr, x9
 ; CHECK-NEXT:    mov x8, xzr
 ; CHECK-NEXT:    cntw x10
diff --git a/llvm/test/CodeGen/AArch64/sve-int-reduce.ll b/llvm/test/CodeGen/AArch64/sve-int-reduce.ll
index d04da62451778..8c1b5225b7f25 100644
--- a/llvm/test/CodeGen/AArch64/sve-int-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-int-reduce.ll
@@ -378,29 +378,29 @@ declare i8 @llvm.vector.reduce.smin.nxv10i8(<vscale x 10 x i8>)
 define i8 @smin_nxv10i8(<vscale x 10 x i8> %a) {
 ; CHECK-LABEL: smin_nxv10i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpkhi z2.h, z0.b
-; CHECK-NEXT:    mov z1.d, #127 // =0x7f
+; CHECK-NEXT:    uunpkhi z1.h, z0.b
+; CHECK-NEXT:    mov z3.d, #127 // =0x7f
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    uunpklo z3.s, z2.h
-; CHECK-NEXT:    uunpkhi z2.s, z2.h
-; CHECK-NEXT:    uunpklo z3.d, z3.s
-; CHECK-NEXT:    uzp1 z3.s, z3.s, z1.s
-; CHECK-NEXT:    uzp1 z2.h, z3.h, z2.h
-; CHECK-NEXT:    uzp1 z2.b, z0.b, z2.b
-; CHECK-NEXT:    uunpkhi z2.h, z2.b
-; CHECK-NEXT:    uunpkhi z3.s, z2.h
-; CHECK-NEXT:    uunpklo z2.s, z2.h
-; CHECK-NEXT:    uunpkhi z3.d, z3.s
-; CHECK-NEXT:    uzp1 z3.s, z1.s, z3.s
-; CHECK-NEXT:    uzp1 z2.h, z2.h, z3.h
-; CHECK-NEXT:    uzp1 z2.b, z0.b, z2.b
-; CHECK-NEXT:    uunpkhi z2.h, z2.b
-; CHECK-NEXT:    uunpkhi z3.s, z2.h
-; CHECK-NEXT:    uunpklo z2.s, z2.h
-; CHECK-NEXT:    uunpklo z3.d, z3.s
-; CHECK-NEXT:    uzp1 z1.s, z3.s, z1.s
+; CHECK-NEXT:    uunpklo z2.s, z1.h
+; CHECK-NEXT:    uunpkhi z1.s, z1.h
+; CHECK-NEXT:    uunpklo z2.d, z2.s
+; CHECK-NEXT:    uzp1 z2.s, z2.s, z3.s
 ; CHECK-NEXT:    uzp1 z1.h, z2.h, z1.h
+; CHECK-NEXT:    uzp1 z1.b, z0.b, z1.b
+; CHECK-NEXT:    uunpkhi z1.h, z1.b
+; CHECK-NEXT:    uunpkhi z2.s, z1.h
+; CHECK-NEXT:    uunpklo z1.s, z1.h
+; CHECK-NEXT:    uunpkhi z2.d, z2.s
+; CHECK-NEXT:    uzp1 z2.s, z3.s, z2.s
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z2.h
+; CHECK-NEXT:    uzp1 z1.b, z0.b, z1.b
+; CHECK-NEXT:    uunpkhi z1.h, z1.b
+; CHECK-NEXT:    uunpkhi z2.s, z1.h
+; CHECK-NEXT:    uunpklo z1.s, z1.h
+; CHECK-NEXT:    uunpklo z2.d, z2.s
+; CHECK-NEXT:    uzp1 z2.s, z2.s, z3.s
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z1.b
 ; CHECK-NEXT:    sminv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -414,12 +414,12 @@ declare i8 @llvm.vector.reduce.add.nxv12i8(<vscale x 12 x i8>)
 define i8 @uaddv_nxv12i8(<vscale x 12 x i8> %a) {
 ; CHECK-LABEL: uaddv_nxv12i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpkhi z2.h, z0.b
-; CHECK-NEXT:    mov z1.s, #0 // =0x0
+; CHECK-NEXT:    uunpkhi z1.h, z0.b
+; CHECK-NEXT:    mov z2.s, #0 // =0x0
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    uunpklo z2.s, z2.h
-; CHECK-NEXT:    uzp1 z1.h, z2.h, z1.h
+; CHECK-NEXT:    uunpklo z1.s, z1.h
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z1.b
 ; CHECK-NEXT:    uaddv d0, p0, z0.b
 ; CHECK-NEXT:    fmov x0, d0
@@ -434,15 +434,15 @@ declare i8 @llvm.vector.reduce.umax.nxv14i8(<vscale x 14 x i8>)
 define i8 @umax_nxv14i8(<vscale x 14 x i8> %a) {
 ; CHECK-LABEL: umax_nxv14i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uunpkhi z2.h, z0.b
-; CHECK-NEXT:    mov z1.d, #0 // =0x0
+; CHECK-NEXT:    uunpkhi z1.h, z0.b
+; CHECK-NEXT:    mov z3.d, #0 // =0x0
 ; CHECK-NEXT:    uunpklo z0.h, z0.b
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    uunpkhi z3.s, z2.h
-; CHECK-NEXT:    uunpklo z2.s, z2.h
-; CHECK-NEXT:    uunpklo z3.d, z3.s
-; CHECK-NEXT:    uzp1 z1.s, z3.s, z1.s
-; CHECK-NEXT:    uzp1 z1.h, z2.h, z1.h
+; CHECK-NEXT:    uunpkhi z2.s, z1.h
+; CHECK-NEXT:    uunpklo z1.s, z1.h
+; CHECK-NEXT:    uunpklo z2.d, z2.s
+; CHECK-NEXT:    uzp1 z2.s, z2.s, z3.s
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z1.b
 ; CHECK-NEXT:    umaxv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-index.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-index.ll
index 2464eacd185dd..bc94c087ef5fe 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-index.ll
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-index.ll
@@ -239,10 +239,10 @@ define <vscale x 4 x i32> @index_rr_i32_combine(i32 %a, i32 %b) {
 define <vscale x 4 x i32> @index_rr_i32_not_combine(i32 %a, i32 %b) {
 ; CHECK-LABEL: index_rr_i32_not_combine:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    index z0.s, #0, #1
 ; CHECK-NEXT:    mov z1.s, w0
 ; CHECK-NEXT:    mov z2.s, w1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mla z1.s, p0/m, z0.s, z2.s
 ; CHECK-NEXT:    add z0.s, z1.s, z0.s
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
index 3e453a6b78179..5648e8244e6ec 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-int-arith-imm.ll
@@ -247,8 +247,8 @@ define <vscale x 4 x i32> @sub_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @sub_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: sub_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sub z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -402,8 +402,8 @@ define <vscale x 4 x i32> @subr_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @subr_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: subr_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    subr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -449,8 +449,8 @@ define <vscale x 8 x i16> @smax_i16(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @smax_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: smax_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #129 // =0x81
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -480,8 +480,8 @@ define <vscale x 4 x i32> @smax_i32(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @smax_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smax_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, #-129 // =0xffffffffffffff7f
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 4 x i1> @llvm.aarch64.sve.ptrue.nxv4i1(i32 31)
@@ -510,8 +510,8 @@ define <vscale x 2 x i64> @smax_i64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @smax_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smax_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -559,8 +559,8 @@ define <vscale x 4 x i32> @smax_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @smax_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: smax_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -606,8 +606,8 @@ define <vscale x 8 x i16> @smin_i16(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @smin_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: smin_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, #-129 // =0xffffffffffffff7f
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
@@ -636,8 +636,8 @@ define <vscale x 4 x i32> @smin_i32(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @smin_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: smin_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #257 // =0x101
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -668,8 +668,8 @@ define <vscale x 2 x i64> @smin_i64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @smin_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: smin_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #-256 // =0xffffffffffffff00
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -717,8 +717,8 @@ define <vscale x 4 x i32> @smin_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @smin_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: smin_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -764,8 +764,8 @@ define <vscale x 8 x i16> @umax_i16(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @umax_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: umax_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
@@ -794,8 +794,8 @@ define <vscale x 4 x i32> @umax_i32(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @umax_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umax_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #257 // =0x101
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -825,8 +825,8 @@ define <vscale x 2 x i64> @umax_i64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @umax_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umax_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -874,8 +874,8 @@ define <vscale x 4 x i32> @umax_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @umax_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: umax_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -921,8 +921,8 @@ define <vscale x 8 x i16> @umin_i16(<vscale x 8 x i16> %a) {
 define <vscale x 8 x i16> @umin_i16_out_of_range(<vscale x 8 x i16> %a) {
 ; CHECK-LABEL: umin_i16_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    dupm z1.b, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 8 x i1> @llvm.aarch64.sve.ptrue.nxv8i1(i32 31)
@@ -951,8 +951,8 @@ define <vscale x 4 x i32> @umin_i32(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @umin_i32_out_of_range(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: umin_i32_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #257 // =0x101
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
@@ -982,8 +982,8 @@ define <vscale x 2 x i64> @umin_i64(<vscale x 2 x i64> %a) {
 define <vscale x 2 x i64> @umin_i64_out_of_range(<vscale x 2 x i64> %a) {
 ; CHECK-LABEL: umin_i64_out_of_range:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %pg = call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -1031,8 +1031,8 @@ define <vscale x 4 x i32> @umin_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @umin_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: umin_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
@@ -2120,8 +2120,8 @@ define <vscale x 4 x i32> @mul_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @mul_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: mul_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-logical-imm.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-logical-imm.ll
index 7cdedeee2cada..5db7ee75c2a8d 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-logical-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-logical-imm.ll
@@ -261,8 +261,8 @@ define <vscale x 4 x i32> @orr_i32_ptrue_all_h(<vscale x 4 x i32> %a) {
 define <vscale x 4 x i32> @orr_i32_ptrue_all_d(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: orr_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #65535 // =0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    orr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
diff --git a/llvm/test/CodeGen/AArch64/sve-ld-post-inc.ll b/llvm/test/CodeGen/AArch64/sve-ld-post-inc.ll
index da5dc5c5b34d9..619134dc4a696 100644
--- a/llvm/test/CodeGen/AArch64/sve-ld-post-inc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-ld-post-inc.ll
@@ -38,18 +38,18 @@ define <vscale x 2 x double> @test_post_ld1_dup(ptr %a, ptr %ptr, i64 %inc) {
 define void @test_post_ld1_int_fixed(ptr %data, i64 %idx, ptr %addr, ptr %res_ptr)  #1 {
 ; CHECK-LABEL: test_post_ld1_int_fixed:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #2 // =0x2
 ; CHECK-NEXT:    index z0.d, #0, #1
-; CHECK-NEXT:    ptrue p1.d, vl1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    ldr x8, [x0]
-; CHECK-NEXT:    ldr x9, [x0, x1, lsl #3]
+; CHECK-NEXT:    ptrue p2.d, vl1
 ; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x2]
-; CHECK-NEXT:    cmpeq p2.d, p0/z, z0.d, z1.d
+; CHECK-NEXT:    ldr x9, [x0, x1, lsl #3]
+; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    mov z0.d, z2.d
-; CHECK-NEXT:    mov z2.d, p2/m, x9
-; CHECK-NEXT:    mov z0.d, p1/m, x8
+; CHECK-NEXT:    mov z0.d, p2/m, x8
+; CHECK-NEXT:    mov z2.d, p1/m, x9
 ; CHECK-NEXT:    add z0.d, z0.d, z2.d
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x3]
 ; CHECK-NEXT:    ret
@@ -67,18 +67,18 @@ define void @test_post_ld1_int_fixed(ptr %data, i64 %idx, ptr %addr, ptr %res_pt
 define void @test_post_ld1_double_fixed(ptr %data, i64 %idx, ptr %addr, ptr %res_ptr)  #1 {
 ; CHECK-LABEL: test_post_ld1_double_fixed:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #2 // =0x2
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
-; CHECK-NEXT:    ptrue p1.d, vl1
-; CHECK-NEXT:    ld1d { z2.d }, p0/z, [x2]
-; CHECK-NEXT:    cmpeq p2.d, p0/z, z0.d, z1.d
-; CHECK-NEXT:    ldr d0, [x0]
-; CHECK-NEXT:    ldr d1, [x0, x1, lsl #3]
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z2.d
-; CHECK-NEXT:    mov z2.d, p2/m, d1
-; CHECK-NEXT:    fadd z0.d, z0.d, z2.d
+; CHECK-NEXT:    ptrue p2.d, vl1
+; CHECK-NEXT:    ldr d2, [x0, x1, lsl #3]
+; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, z1.d
+; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x2]
+; CHECK-NEXT:    ldr d1, [x0]
+; CHECK-NEXT:    sel z1.d, p2, z1.d, z0.d
+; CHECK-NEXT:    mov z0.d, p1/m, d2
+; CHECK-NEXT:    fadd z0.d, z1.d, z0.d
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x3]
 ; CHECK-NEXT:    ret
   %A = load <4 x double>, ptr %addr
diff --git a/llvm/test/CodeGen/AArch64/sve-ld1r.ll b/llvm/test/CodeGen/AArch64/sve-ld1r.ll
index e42e2272a2d4f..fbe82e8591fd0 100644
--- a/llvm/test/CodeGen/AArch64/sve-ld1r.ll
+++ b/llvm/test/CodeGen/AArch64/sve-ld1r.ll
@@ -20,8 +20,8 @@ define <vscale x 16 x i8> @ld1r_stack() {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    adrp x8, :got:g8
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    ldr x8, [x8, :got_lo12:g8]
 ; CHECK-NEXT:    ldrb w8, [x8]
 ; CHECK-NEXT:    strb w8, [sp, #12]
@@ -1433,10 +1433,10 @@ define ptr @avoid_preindex_load(ptr %src, ptr %out) {
 define ptr @avoid_preindex_load_dup(ptr %src, <vscale x 2 x i1> %pg, ptr %out) {
 ; CHECK-LABEL: avoid_preindex_load_dup:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1rsb { z0.d }, p0/z, [x0, #1]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x0, x0, #1
-; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
+; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %ptr = getelementptr inbounds i8, ptr %src, i64 1
   %tmp = load i8, ptr %ptr, align 4
@@ -1450,10 +1450,10 @@ define ptr @avoid_preindex_load_dup(ptr %src, <vscale x 2 x i1> %pg, ptr %out) {
 define ptr @avoid_preindex_load_dup_passthru_zero(ptr %src, <vscale x 2 x i1> %pg, ptr %out) {
 ; CHECK-LABEL: avoid_preindex_load_dup_passthru_zero:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1rsb { z0.d }, p0/z, [x0, #1]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add x0, x0, #1
-; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
+; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %ptr = getelementptr inbounds i8, ptr %src, i64 1
   %tmp = load i8, ptr %ptr, align 4
@@ -1467,10 +1467,10 @@ define ptr @avoid_preindex_load_dup_passthru_zero(ptr %src, <vscale x 2 x i1> %p
 define ptr @preindex_load_dup_passthru(<vscale x 2 x i64> %passthru, ptr %src, <vscale x 2 x i1> %pg, ptr %out) {
 ; CHECK-LABEL: preindex_load_dup_passthru:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ldrsb x8, [x0, #1]!
 ; CHECK-NEXT:    mov z0.d, p0/m, x8
-; CHECK-NEXT:    st1d { z0.d }, p1, [x1]
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    ret
   %ptr = getelementptr inbounds i8, ptr %src, i64 1
   %tmp = load i8, ptr %ptr, align 4
@@ -1485,8 +1485,8 @@ define ptr @preindex_load_dup_passthru(<vscale x 2 x i64> %passthru, ptr %src, <
 define ptr @preidx8sext64_instead_of_ld1r(ptr %src, ptr %out, ptr %dst) {
 ; CHECK-LABEL: preidx8sext64_instead_of_ld1r:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldrsb x8, [x0, #1]!
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x1]
 ; CHECK-NEXT:    str x8, [x2]
diff --git a/llvm/test/CodeGen/AArch64/sve-lsr-scaled-index-addressing-mode.ll b/llvm/test/CodeGen/AArch64/sve-lsr-scaled-index-addressing-mode.ll
index 06ec132808154..5d53c00c52728 100644
--- a/llvm/test/CodeGen/AArch64/sve-lsr-scaled-index-addressing-mode.ll
+++ b/llvm/test/CodeGen/AArch64/sve-lsr-scaled-index-addressing-mode.ll
@@ -38,8 +38,8 @@ define void @ld_st_nxv8i16(ptr %in, ptr %out) {
 ;
 ; ASM-LABEL: ld_st_nxv8i16:
 ; ASM:       // %bb.0: // %entry
-; ASM-NEXT:    ptrue p0.h
 ; ASM-NEXT:    mov z0.h, #3 // =0x3
+; ASM-NEXT:    ptrue p0.h
 ; ASM-NEXT:    mov x8, xzr
 ; ASM-NEXT:    cnth x9
 ; ASM-NEXT:  .LBB0_1: // %loop
@@ -111,8 +111,8 @@ define void @masked_ld_st_nxv8i16(ptr %in, ptr %out, i64 %n) {
 ;
 ; ASM-LABEL: masked_ld_st_nxv8i16:
 ; ASM:       // %bb.0: // %entry
-; ASM-NEXT:    ptrue p0.h
 ; ASM-NEXT:    mov z0.h, #3 // =0x3
+; ASM-NEXT:    ptrue p0.h
 ; ASM-NEXT:    mov x8, xzr
 ; ASM-NEXT:    cnth x9
 ; ASM-NEXT:  .LBB1_1: // %loop
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll b/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
index e40d65efb158b..dfdfc456ccdba 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-gather-legalize.ll
@@ -126,9 +126,9 @@ define <vscale x 8 x half> @masked_gather_nxv8f16(<vscale x 8 x ptr> %ptrs, <vsc
 define <vscale x 8 x bfloat> @masked_gather_nxv8bf16(ptr %base, <vscale x 8 x i16> %indices, <vscale x 8 x i1> %mask) #0 {
 ; CHECK-LABEL: masked_gather_nxv8bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    sunpkhi z1.s, z0.h
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
+; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    ld1h { z1.s }, p1/z, [x0, z1.s, sxtw #1]
 ; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x0, z0.s, sxtw #1]
@@ -175,14 +175,14 @@ define <vscale x 8 x float> @masked_gather_nxv8f32(ptr %base, <vscale x 8 x i32>
 define <vscale x 16 x i8> @masked_gather_nxv16i8(ptr %base, <vscale x 16 x i8> %indices, <vscale x 16 x i1> %mask) #0 {
 ; CHECK-LABEL: masked_gather_nxv16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    sunpkhi z1.h, z0.b
+; CHECK-NEXT:    punpkhi p1.h, p0.b
 ; CHECK-NEXT:    sunpklo z0.h, z0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
-; CHECK-NEXT:    sunpkhi z2.s, z1.h
-; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
+; CHECK-NEXT:    sunpkhi z2.s, z1.h
+; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    ld1b { z2.s }, p2/z, [x0, z2.s, sxtw]
 ; CHECK-NEXT:    ld1b { z1.s }, p1/z, [x0, z1.s, sxtw]
 ; CHECK-NEXT:    punpkhi p1.h, p0.b
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-ldst-sext.ll b/llvm/test/CodeGen/AArch64/sve-masked-ldst-sext.ll
index 40d889f1b501e..d397424cb162f 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-ldst-sext.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-ldst-sext.ll
@@ -235,9 +235,9 @@ define <vscale x 8 x i64> @masked_sload_x2_8i8_8i64(ptr %a, ptr %b, <vscale x 8
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpkhi p3.h, p0.b
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    ld1sb { z3.d }, p2/z, [x0, #3, mul vl]
 ; CHECK-NEXT:    ld1sb { z5.d }, p2/z, [x1, #3, mul vl]
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    ld1sb { z2.d }, p1/z, [x0, #2, mul vl]
 ; CHECK-NEXT:    ld1sb { z6.d }, p1/z, [x1, #2, mul vl]
 ; CHECK-NEXT:    ld1sb { z1.d }, p3/z, [x0, #1, mul vl]
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-ldst-zext.ll b/llvm/test/CodeGen/AArch64/sve-masked-ldst-zext.ll
index e9860222613e9..17fa13e4e8a8b 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-ldst-zext.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-ldst-zext.ll
@@ -98,9 +98,9 @@ define <vscale x 8 x i64> @masked_zload_nxv8i16(ptr %a, <vscale x 8 x i1> %mask)
 define <vscale x 2 x double> @masked_zload_2i16_2f64(ptr noalias %in, <vscale x 2 x i1> %mask) {
 ; CHECK-LABEL: masked_zload_2i16_2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1h { z0.d }, p0/z, [x0]
-; CHECK-NEXT:    ucvtf z0.d, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ret
   %wide.load = call <vscale x 2 x i16> @llvm.masked.load.nxv2i16(ptr %in, i32 2, <vscale x 2 x i1> %mask, <vscale x 2 x i16> undef)
   %zext = zext <vscale x 2 x i16> %wide.load to <vscale x 2 x i32>
@@ -230,9 +230,9 @@ define <vscale x 8 x i64> @masked_zload_x2_8i8_8i64(ptr %a, ptr %b, <vscale x 8
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpkhi p3.h, p0.b
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    ld1b { z3.d }, p2/z, [x0, #3, mul vl]
 ; CHECK-NEXT:    ld1b { z5.d }, p2/z, [x1, #3, mul vl]
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    ld1b { z2.d }, p1/z, [x0, #2, mul vl]
 ; CHECK-NEXT:    ld1b { z6.d }, p1/z, [x1, #2, mul vl]
 ; CHECK-NEXT:    ld1b { z1.d }, p3/z, [x0, #1, mul vl]
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-scatter-legalize.ll b/llvm/test/CodeGen/AArch64/sve-masked-scatter-legalize.ll
index 10a6445c6fa2f..09ee1e7d3b12e 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-scatter-legalize.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-scatter-legalize.ll
@@ -9,26 +9,26 @@ target triple = "aarch64-linux-gnu"
 define void @masked_scatter_nxv16i8(<vscale x 16 x i8> %data, ptr %base, <vscale x 16 x i8> %offsets, <vscale x 16 x i1> %mask) #0 {
 ; CHECK-LABEL: masked_scatter_nxv16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    sunpklo z2.h, z1.b
 ; CHECK-NEXT:    uunpklo z4.h, z0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    sunpkhi z1.h, z1.b
 ; CHECK-NEXT:    uunpkhi z0.h, z0.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
+; CHECK-NEXT:    punpklo p2.h, p1.b
+; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    sunpklo z3.s, z2.h
 ; CHECK-NEXT:    uunpklo z5.s, z4.h
 ; CHECK-NEXT:    sunpkhi z2.s, z2.h
-; CHECK-NEXT:    punpklo p2.h, p1.b
-; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    st1b { z5.s }, p2, [x0, z3.s, sxtw]
 ; CHECK-NEXT:    uunpkhi z3.s, z4.h
 ; CHECK-NEXT:    st1b { z3.s }, p1, [x0, z2.s, sxtw]
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    sunpklo z2.s, z1.h
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    uunpklo z3.s, z0.h
 ; CHECK-NEXT:    sunpkhi z1.s, z1.h
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
+; CHECK-NEXT:    punpklo p1.h, p0.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1b { z3.s }, p1, [x0, z2.s, sxtw]
 ; CHECK-NEXT:    st1b { z0.s }, p0, [x0, z1.s, sxtw]
 ; CHECK-NEXT:    ret
@@ -40,12 +40,12 @@ define void @masked_scatter_nxv16i8(<vscale x 16 x i8> %data, ptr %base, <vscale
 define void @masked_scatter_nxv8i16(<vscale x 8 x i16> %data, ptr %base, <vscale x 8 x i16> %offsets, <vscale x 8 x i1> %mask) #0 {
 ; CHECK-LABEL: masked_scatter_nxv8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    sunpklo z2.s, z1.h
 ; CHECK-NEXT:    uunpklo z3.s, z0.h
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    sunpkhi z1.s, z1.h
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
+; CHECK-NEXT:    punpklo p1.h, p0.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1h { z3.s }, p1, [x0, z2.s, sxtw #1]
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0, z1.s, sxtw #1]
 ; CHECK-NEXT:    ret
@@ -57,12 +57,12 @@ define void @masked_scatter_nxv8i16(<vscale x 8 x i16> %data, ptr %base, <vscale
 define void @masked_scatter_nxv8bf16(<vscale x 8 x bfloat> %data, ptr %base, <vscale x 8 x i16> %offsets, <vscale x 8 x i1> %mask) #0 {
 ; CHECK-LABEL: masked_scatter_nxv8bf16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    sunpklo z2.s, z1.h
 ; CHECK-NEXT:    uunpklo z3.s, z0.h
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    sunpkhi z1.s, z1.h
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
+; CHECK-NEXT:    punpklo p1.h, p0.b
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1h { z3.s }, p1, [x0, z2.s, sxtw #1]
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0, z1.s, sxtw #1]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-masked-scatter.ll b/llvm/test/CodeGen/AArch64/sve-masked-scatter.ll
index e866474942cd7..94e525d22b825 100644
--- a/llvm/test/CodeGen/AArch64/sve-masked-scatter.ll
+++ b/llvm/test/CodeGen/AArch64/sve-masked-scatter.ll
@@ -76,8 +76,8 @@ define void @masked_scatter_nxv2f64(<vscale x 2 x double> %data, <vscale x 2 x p
 define void @masked_scatter_splat_constant_pointer (<vscale x 4 x i1> %pg) {
 ; CHECK-LABEL: masked_scatter_splat_constant_pointer:
 ; CHECK:       // %bb.0: // %vector.body
-; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    mov z0.d, #0 // =0x0
+; CHECK-NEXT:    punpklo p1.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    st1w { z0.d }, p1, [z0.d]
 ; CHECK-NEXT:    st1w { z0.d }, p0, [z0.d]
diff --git a/llvm/test/CodeGen/AArch64/sve-pr62151.ll b/llvm/test/CodeGen/AArch64/sve-pr62151.ll
index 7cec20fda429c..5ed34f14a0b14 100644
--- a/llvm/test/CodeGen/AArch64/sve-pr62151.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pr62151.ll
@@ -5,8 +5,8 @@
 define i32 @build_interpolation(<2 x i32> %0, <2 x i32> %1, <2 x i32> %2) {
 ; CHECK-LABEL: build_interpolation:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    mul v0.2s, v1.2s, v0.2s
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d2 killed $d2 def $z2
 ; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z2.s
 ; CHECK-NEXT:    mla v0.2s, v1.2s, v0.s[1]
diff --git a/llvm/test/CodeGen/AArch64/sve-pred-arith.ll b/llvm/test/CodeGen/AArch64/sve-pred-arith.ll
index 4d46ac5ecbaa9..6e08606db9537 100644
--- a/llvm/test/CodeGen/AArch64/sve-pred-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pred-arith.ll
@@ -54,24 +54,24 @@ define aarch64_sve_vector_pcs <vscale x 64 x i1> @add_nxv64i1(<vscale x 64 x i1>
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
-; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    ptrue p4.b
 ; CHECK-NEXT:    str p8, [sp, #3, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p6.b
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    ldr p5, [x0]
-; CHECK-NEXT:    ldr p6, [x1]
+; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ldr p4, [x0]
+; CHECK-NEXT:    ldr p5, [x1]
 ; CHECK-NEXT:    ldr p7, [x2]
 ; CHECK-NEXT:    ldr p8, [x3]
-; CHECK-NEXT:    eor p0.b, p4/z, p0.b, p5.b
-; CHECK-NEXT:    eor p1.b, p4/z, p1.b, p6.b
-; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    eor p2.b, p4/z, p2.b, p7.b
+; CHECK-NEXT:    eor p0.b, p6/z, p0.b, p4.b
+; CHECK-NEXT:    eor p1.b, p6/z, p1.b, p5.b
+; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    eor p2.b, p6/z, p2.b, p7.b
 ; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    eor p3.b, p4/z, p3.b, p8.b
+; CHECK-NEXT:    eor p3.b, p6/z, p3.b, p8.b
 ; CHECK-NEXT:    ldr p8, [sp, #3, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    .cfi_def_cfa wsp, 16
@@ -138,24 +138,24 @@ define aarch64_sve_vector_pcs <vscale x 64 x i1> @sub_nxv64i1(<vscale x 64 x i1>
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
-; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    ptrue p4.b
 ; CHECK-NEXT:    str p8, [sp, #3, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p6.b
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    ldr p5, [x0]
-; CHECK-NEXT:    ldr p6, [x1]
+; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ldr p4, [x0]
+; CHECK-NEXT:    ldr p5, [x1]
 ; CHECK-NEXT:    ldr p7, [x2]
 ; CHECK-NEXT:    ldr p8, [x3]
-; CHECK-NEXT:    eor p0.b, p4/z, p0.b, p5.b
-; CHECK-NEXT:    eor p1.b, p4/z, p1.b, p6.b
-; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    eor p2.b, p4/z, p2.b, p7.b
+; CHECK-NEXT:    eor p0.b, p6/z, p0.b, p4.b
+; CHECK-NEXT:    eor p1.b, p6/z, p1.b, p5.b
+; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    eor p2.b, p6/z, p2.b, p7.b
 ; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    eor p3.b, p4/z, p3.b, p8.b
+; CHECK-NEXT:    eor p3.b, p6/z, p3.b, p8.b
 ; CHECK-NEXT:    ldr p8, [sp, #3, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    .cfi_def_cfa wsp, 16
diff --git a/llvm/test/CodeGen/AArch64/sve-pred-selectop.ll b/llvm/test/CodeGen/AArch64/sve-pred-selectop.ll
index 600e9c4805ff7..8438e9d88f5de 100644
--- a/llvm/test/CodeGen/AArch64/sve-pred-selectop.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pred-selectop.ll
@@ -322,10 +322,10 @@ entry:
 define <vscale x 4 x i32> @ornot_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, <vscale x 4 x i32> %y) {
 ; CHECK-LABEL: ornot_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z3.s, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    eor z2.d, z2.d, z3.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    eor z2.d, z2.d, z3.d
 ; CHECK-NEXT:    orr z1.d, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
@@ -340,10 +340,10 @@ entry:
 define <vscale x 8 x i16> @ornot_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, <vscale x 8 x i16> %y) {
 ; CHECK-LABEL: ornot_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z3.h, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    eor z2.d, z2.d, z3.d
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    eor z2.d, z2.d, z3.d
 ; CHECK-NEXT:    orr z1.d, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
@@ -358,10 +358,10 @@ entry:
 define <vscale x 16 x i8> @ornot_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, <vscale x 16 x i8> %y) {
 ; CHECK-LABEL: ornot_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z3.b, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    eor z2.d, z2.d, z3.d
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    eor z2.d, z2.d, z3.d
 ; CHECK-NEXT:    orr z1.d, z1.d, z2.d
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
@@ -904,8 +904,8 @@ define <vscale x 4 x i32> @addqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    add z1.s, z1.s, z2.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    add z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -922,8 +922,8 @@ define <vscale x 8 x i16> @addqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    add z1.h, z1.h, z2.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    add z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -940,8 +940,8 @@ define <vscale x 16 x i8> @addqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    add z1.b, z1.b, z2.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    add z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -958,8 +958,8 @@ define <vscale x 4 x i32> @subqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    sub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    sub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -976,8 +976,8 @@ define <vscale x 8 x i16> @subqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    sub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    sub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -994,8 +994,8 @@ define <vscale x 16 x i8> @subqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8>
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    sub z1.b, z1.b, z2.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    sub z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -1010,10 +1010,10 @@ entry:
 define <vscale x 4 x i32> @mulqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, i32 %y) {
 ; CHECK-LABEL: mulqr_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    mul z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    mul z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1028,10 +1028,10 @@ entry:
 define <vscale x 8 x i16> @mulqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, i16 %y) {
 ; CHECK-LABEL: mulqr_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    mul z1.h, z1.h, z2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    mul z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1046,10 +1046,10 @@ entry:
 define <vscale x 16 x i8> @mulqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, i8 %y) {
 ; CHECK-LABEL: mulqr_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    mul z1.b, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    mul z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -1064,11 +1064,11 @@ entry:
 define <vscale x 4 x float> @faddqr_v4f32(<vscale x 4 x float> %z, <vscale x 4 x float> %x, float %y) {
 ; CHECK-LABEL: faddqr_v4f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $s2 killed $s2 def $z2
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, s2
-; CHECK-NEXT:    fadd z1.s, z1.s, z2.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fadd z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1083,11 +1083,11 @@ entry:
 define <vscale x 8 x half> @faddqr_v8f16(<vscale x 8 x half> %z, <vscale x 8 x half> %x, half %y) {
 ; CHECK-LABEL: faddqr_v8f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $h2 killed $h2 def $z2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, h2
-; CHECK-NEXT:    fadd z1.h, z1.h, z2.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    fadd z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1102,11 +1102,11 @@ entry:
 define <vscale x 4 x float> @fsubqr_v4f32(<vscale x 4 x float> %z, <vscale x 4 x float> %x, float %y) {
 ; CHECK-LABEL: fsubqr_v4f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $s2 killed $s2 def $z2
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, s2
-; CHECK-NEXT:    fsub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fsub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1121,11 +1121,11 @@ entry:
 define <vscale x 8 x half> @fsubqr_v8f16(<vscale x 8 x half> %z, <vscale x 8 x half> %x, half %y) {
 ; CHECK-LABEL: fsubqr_v8f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $h2 killed $h2 def $z2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, h2
-; CHECK-NEXT:    fsub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    fsub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1140,11 +1140,11 @@ entry:
 define <vscale x 4 x float> @fmulqr_v4f32(<vscale x 4 x float> %z, <vscale x 4 x float> %x, float %y) {
 ; CHECK-LABEL: fmulqr_v4f32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $s2 killed $s2 def $z2
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, s2
-; CHECK-NEXT:    fmul z1.s, z1.s, z2.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fmul z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1159,11 +1159,11 @@ entry:
 define <vscale x 8 x half> @fmulqr_v8f16(<vscale x 8 x half> %z, <vscale x 8 x half> %x, half %y) {
 ; CHECK-LABEL: fmulqr_v8f16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $h2 killed $h2 def $z2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, h2
-; CHECK-NEXT:    fmul z1.h, z1.h, z2.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z0.h, #0.0
+; CHECK-NEXT:    fmul z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1178,10 +1178,10 @@ entry:
 define <vscale x 4 x i32> @sadd_satqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, i32 %y) {
 ; CHECK-LABEL: sadd_satqr_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    sqadd z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    sqadd z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1196,10 +1196,10 @@ entry:
 define <vscale x 8 x i16> @sadd_satqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, i16 %y) {
 ; CHECK-LABEL: sadd_satqr_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    sqadd z1.h, z1.h, z2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    sqadd z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1214,10 +1214,10 @@ entry:
 define <vscale x 16 x i8> @sadd_satqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, i8 %y) {
 ; CHECK-LABEL: sadd_satqr_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    sqadd z1.b, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    sqadd z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -1232,10 +1232,10 @@ entry:
 define <vscale x 4 x i32> @uadd_satqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, i32 %y) {
 ; CHECK-LABEL: uadd_satqr_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    uqadd z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    uqadd z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1250,10 +1250,10 @@ entry:
 define <vscale x 8 x i16> @uadd_satqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, i16 %y) {
 ; CHECK-LABEL: uadd_satqr_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    uqadd z1.h, z1.h, z2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    uqadd z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1268,10 +1268,10 @@ entry:
 define <vscale x 16 x i8> @uadd_satqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, i8 %y) {
 ; CHECK-LABEL: uadd_satqr_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    uqadd z1.b, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    uqadd z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -1286,10 +1286,10 @@ entry:
 define <vscale x 4 x i32> @ssub_satqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, i32 %y) {
 ; CHECK-LABEL: ssub_satqr_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    sqsub z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    sqsub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1304,10 +1304,10 @@ entry:
 define <vscale x 8 x i16> @ssub_satqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, i16 %y) {
 ; CHECK-LABEL: ssub_satqr_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    sqsub z1.h, z1.h, z2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    sqsub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1322,10 +1322,10 @@ entry:
 define <vscale x 16 x i8> @ssub_satqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, i8 %y) {
 ; CHECK-LABEL: ssub_satqr_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    sqsub z1.b, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    sqsub z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
@@ -1340,10 +1340,10 @@ entry:
 define <vscale x 4 x i32> @usub_satqr_v4i32(<vscale x 4 x i32> %z, <vscale x 4 x i32> %x, i32 %y) {
 ; CHECK-LABEL: usub_satqr_v4i32:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w0
-; CHECK-NEXT:    uqsub z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, #0
+; CHECK-NEXT:    uqsub z1.s, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, z1.s
 ; CHECK-NEXT:    ret
 entry:
@@ -1358,10 +1358,10 @@ entry:
 define <vscale x 8 x i16> @usub_satqr_v8i16(<vscale x 8 x i16> %z, <vscale x 8 x i16> %x, i16 %y) {
 ; CHECK-LABEL: usub_satqr_v8i16:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
-; CHECK-NEXT:    uqsub z1.h, z1.h, z2.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    uqsub z1.h, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
 entry:
@@ -1376,10 +1376,10 @@ entry:
 define <vscale x 16 x i8> @usub_satqr_v16i8(<vscale x 16 x i8> %z, <vscale x 16 x i8> %x, i8 %y) {
 ; CHECK-LABEL: usub_satqr_v16i8:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
-; CHECK-NEXT:    uqsub z1.b, z1.b, z2.b
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z0.b, #0
+; CHECK-NEXT:    uqsub z1.b, z1.b, z2.b
 ; CHECK-NEXT:    mov z0.b, p0/m, z1.b
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll b/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
index 14bc1b45e79ee..2541910e080e3 100644
--- a/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pred-selectop2.ll
@@ -202,9 +202,9 @@ entry:
 define <vscale x 8 x i16> @sdiv_nxv8i16_x(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: sdiv_nxv8i16_x:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z3.s, z1.h
 ; CHECK-NEXT:    sunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    sunpklo z4.s, z0.h
@@ -288,9 +288,9 @@ entry:
 define <vscale x 8 x i16> @udiv_nxv8i16_x(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: udiv_nxv8i16_x:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z3.s, z1.h
 ; CHECK-NEXT:    uunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z1.s, z1.h
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    uunpklo z4.s, z0.h
@@ -376,17 +376,17 @@ entry:
 define <vscale x 8 x i16> @srem_nxv8i16_x(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: srem_nxv8i16_x:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z3.s, z1.h
 ; CHECK-NEXT:    sunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z5.s, z0.h
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    sunpklo z4.s, z1.h
 ; CHECK-NEXT:    sdivr z4.s, p0/m, z4.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    uzp1 z3.h, z4.h, z3.h
-; CHECK-NEXT:    mls z0.h, p0/m, z3.h, z1.h
+; CHECK-NEXT:    uzp1 z2.h, z4.h, z3.h
+; CHECK-NEXT:    mls z0.h, p0/m, z2.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 8 x i16> %n, zeroinitializer
@@ -419,8 +419,8 @@ define <vscale x 16 x i8> @srem_nxv16i8_x(<vscale x 16 x i8> %x, <vscale x 16 x
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z7.h
-; CHECK-NEXT:    uzp1 z3.b, z4.b, z3.b
-; CHECK-NEXT:    mls z0.b, p0/m, z3.b, z1.b
+; CHECK-NEXT:    uzp1 z2.b, z4.b, z3.b
+; CHECK-NEXT:    mls z0.b, p0/m, z2.b, z1.b
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 16 x i8> %n, zeroinitializer
@@ -464,17 +464,17 @@ entry:
 define <vscale x 8 x i16> @urem_nxv8i16_x(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: urem_nxv8i16_x:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z3.s, z1.h
 ; CHECK-NEXT:    uunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z5.s, z0.h
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    uunpklo z4.s, z1.h
 ; CHECK-NEXT:    udivr z4.s, p0/m, z4.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    uzp1 z3.h, z4.h, z3.h
-; CHECK-NEXT:    mls z0.h, p0/m, z3.h, z1.h
+; CHECK-NEXT:    uzp1 z2.h, z4.h, z3.h
+; CHECK-NEXT:    mls z0.h, p0/m, z2.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 8 x i16> %n, zeroinitializer
@@ -507,8 +507,8 @@ define <vscale x 16 x i8> @urem_nxv16i8_x(<vscale x 16 x i8> %x, <vscale x 16 x
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z7.h
-; CHECK-NEXT:    uzp1 z3.b, z4.b, z3.b
-; CHECK-NEXT:    mls z0.b, p0/m, z3.b, z1.b
+; CHECK-NEXT:    uzp1 z2.b, z4.b, z3.b
+; CHECK-NEXT:    mls z0.b, p0/m, z2.b, z1.b
 ; CHECK-NEXT:    ret
 entry:
   %c = icmp sgt <vscale x 16 x i8> %n, zeroinitializer
@@ -1140,8 +1140,8 @@ define <vscale x 8 x half> @fdiv_nxv8f16_x(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-LABEL: fdiv_nxv8f16_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fdivr z1.h, p0/m, z1.h, z0.h
+; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    not p0.b, p0/z, p1.b
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
@@ -1654,9 +1654,9 @@ entry:
 define <vscale x 8 x i16> @sdiv_nxv8i16_y(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: sdiv_nxv8i16_y:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z3.s, z1.h
 ; CHECK-NEXT:    sunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    sunpklo z4.s, z1.h
@@ -1740,9 +1740,9 @@ entry:
 define <vscale x 8 x i16> @udiv_nxv8i16_y(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: udiv_nxv8i16_y:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z3.s, z1.h
 ; CHECK-NEXT:    uunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    uunpklo z4.s, z1.h
@@ -1830,17 +1830,17 @@ entry:
 define <vscale x 8 x i16> @srem_nxv8i16_y(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: srem_nxv8i16_y:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z3.s, z1.h
 ; CHECK-NEXT:    sunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z5.s, z0.h
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    sunpklo z4.s, z1.h
 ; CHECK-NEXT:    sdivr z4.s, p0/m, z4.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    uzp1 z3.h, z4.h, z3.h
-; CHECK-NEXT:    msb z1.h, p0/m, z3.h, z0.h
+; CHECK-NEXT:    uzp1 z2.h, z4.h, z3.h
+; CHECK-NEXT:    msb z1.h, p0/m, z2.h, z0.h
 ; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
@@ -1874,8 +1874,8 @@ define <vscale x 16 x i8> @srem_nxv16i8_y(<vscale x 16 x i8> %x, <vscale x 16 x
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z7.h
-; CHECK-NEXT:    uzp1 z3.b, z4.b, z3.b
-; CHECK-NEXT:    msb z1.b, p0/m, z3.b, z0.b
+; CHECK-NEXT:    uzp1 z2.b, z4.b, z3.b
+; CHECK-NEXT:    msb z1.b, p0/m, z2.b, z0.b
 ; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
@@ -1922,17 +1922,17 @@ entry:
 define <vscale x 8 x i16> @urem_nxv8i16_y(<vscale x 8 x i16> %x, <vscale x 8 x i16> %y, <vscale x 8 x i16> %n) {
 ; CHECK-LABEL: urem_nxv8i16_y:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpkhi z3.s, z1.h
 ; CHECK-NEXT:    uunpkhi z4.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z5.s, z0.h
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    uunpklo z4.s, z1.h
 ; CHECK-NEXT:    udivr z4.s, p0/m, z4.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, #0
-; CHECK-NEXT:    uzp1 z3.h, z4.h, z3.h
-; CHECK-NEXT:    msb z1.h, p0/m, z3.h, z0.h
+; CHECK-NEXT:    uzp1 z2.h, z4.h, z3.h
+; CHECK-NEXT:    msb z1.h, p0/m, z2.h, z0.h
 ; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
@@ -1966,8 +1966,8 @@ define <vscale x 16 x i8> @urem_nxv16i8_y(<vscale x 16 x i8> %x, <vscale x 16 x
 ; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpgt p0.b, p0/z, z2.b, #0
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z7.h
-; CHECK-NEXT:    uzp1 z3.b, z4.b, z3.b
-; CHECK-NEXT:    msb z1.b, p0/m, z3.b, z0.b
+; CHECK-NEXT:    uzp1 z2.b, z4.b, z3.b
+; CHECK-NEXT:    msb z1.b, p0/m, z2.b, z0.b
 ; CHECK-NEXT:    mov z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
@@ -2629,8 +2629,8 @@ define <vscale x 8 x half> @fdiv_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-LABEL: fdiv_nxv8f16_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fdiv z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    not p0.b, p0/z, p1.b
 ; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -2855,8 +2855,8 @@ define <vscale x 4 x float> @fmai_nxv4f32_y(<vscale x 4 x float> %x, <vscale x 4
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmle p1.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    fmla z0.s, p0/m, z1.s, z2.s
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 4 x float> %n, zeroinitializer
@@ -2871,8 +2871,8 @@ define <vscale x 8 x half> @fmai_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fcmle p1.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fmla z0.h, p0/m, z1.h, z2.h
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.h, p1, z0.h, z1.h
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 8 x half> %n, zeroinitializer
@@ -2887,8 +2887,8 @@ define <vscale x 2 x double> @fmai_nxv2f64_y(<vscale x 2 x double> %x, <vscale x
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fcmle p1.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fmla z0.d, p0/m, z1.d, z2.d
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 2 x double> %n, zeroinitializer
@@ -2903,8 +2903,8 @@ define <vscale x 4 x float> @fma_nxv4f32_y(<vscale x 4 x float> %x, <vscale x 4
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmle p1.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    fmla z0.s, p0/m, z1.s, z2.s
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 4 x float> %n, zeroinitializer
@@ -2920,8 +2920,8 @@ define <vscale x 8 x half> @fma_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fcmle p1.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fmla z0.h, p0/m, z1.h, z2.h
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.h, p1, z0.h, z1.h
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 8 x half> %n, zeroinitializer
@@ -2937,8 +2937,8 @@ define <vscale x 2 x double> @fma_nxv2f64_y(<vscale x 2 x double> %x, <vscale x
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fcmle p1.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fmla z0.d, p0/m, z1.d, z2.d
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 2 x double> %n, zeroinitializer
diff --git a/llvm/test/CodeGen/AArch64/sve-pred-selectop3.ll b/llvm/test/CodeGen/AArch64/sve-pred-selectop3.ll
index 0f09f7dac2982..bafd5abcc7b23 100644
--- a/llvm/test/CodeGen/AArch64/sve-pred-selectop3.ll
+++ b/llvm/test/CodeGen/AArch64/sve-pred-selectop3.ll
@@ -792,8 +792,8 @@ define <vscale x 8 x half> @fdiv_nxv8f16_x(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-LABEL: fdiv_nxv8f16_x:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fdivr z1.h, p0/m, z1.h, z0.h
+; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    not p0.b, p0/z, p1.b
 ; CHECK-NEXT:    mov z0.h, p0/m, z1.h
 ; CHECK-NEXT:    ret
@@ -1700,8 +1700,8 @@ define <vscale x 8 x half> @fdiv_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-LABEL: fdiv_nxv8f16_y:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ptrue p0.h
-; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    fdiv z0.h, p0/m, z0.h, z1.h
+; CHECK-NEXT:    fcmle p1.h, p0/z, z2.h, #0.0
 ; CHECK-NEXT:    not p0.b, p0/z, p1.b
 ; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
@@ -1734,8 +1734,8 @@ define <vscale x 4 x float> @fmai_nxv4f32_y(<vscale x 4 x float> %x, <vscale x 4
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmle p1.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    fmla z0.s, p0/m, z1.s, z2.s
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 4 x float> %n, zeroinitializer
@@ -1750,8 +1750,8 @@ define <vscale x 8 x half> @fmai_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fcmle p1.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fmla z0.h, p0/m, z1.h, z2.h
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.h, p1, z0.h, z1.h
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 8 x half> %n, zeroinitializer
@@ -1766,8 +1766,8 @@ define <vscale x 2 x double> @fmai_nxv2f64_y(<vscale x 2 x double> %x, <vscale x
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fcmle p1.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fmla z0.d, p0/m, z1.d, z2.d
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 2 x double> %n, zeroinitializer
@@ -1782,8 +1782,8 @@ define <vscale x 4 x float> @fma_nxv4f32_y(<vscale x 4 x float> %x, <vscale x 4
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fcmle p1.s, p0/z, z3.s, #0.0
 ; CHECK-NEXT:    fmla z0.s, p0/m, z1.s, z2.s
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.s, p1, z0.s, z1.s
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 4 x float> %n, zeroinitializer
@@ -1799,8 +1799,8 @@ define <vscale x 8 x half> @fma_nxv8f16_y(<vscale x 8 x half> %x, <vscale x 8 x
 ; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    fcmle p1.h, p0/z, z3.h, #0.0
 ; CHECK-NEXT:    fmla z0.h, p0/m, z1.h, z2.h
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.h, p1, z0.h, z1.h
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 8 x half> %n, zeroinitializer
@@ -1816,8 +1816,8 @@ define <vscale x 2 x double> @fma_nxv2f64_y(<vscale x 2 x double> %x, <vscale x
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fcmle p1.d, p0/z, z3.d, #0.0
 ; CHECK-NEXT:    fmla z0.d, p0/m, z1.d, z2.d
-; CHECK-NEXT:    not p1.b, p0/z, p1.b
-; CHECK-NEXT:    sel z0.d, p1, z0.d, z1.d
+; CHECK-NEXT:    not p0.b, p0/z, p1.b
+; CHECK-NEXT:    sel z0.d, p0, z0.d, z1.d
 ; CHECK-NEXT:    ret
 entry:
   %c = fcmp ugt <vscale x 2 x double> %n, zeroinitializer
diff --git a/llvm/test/CodeGen/AArch64/sve-ptest-removal-cmple.ll b/llvm/test/CodeGen/AArch64/sve-ptest-removal-cmple.ll
index fbadbf7226fd1..8bd38d7bc44df 100644
--- a/llvm/test/CodeGen/AArch64/sve-ptest-removal-cmple.ll
+++ b/llvm/test/CodeGen/AArch64/sve-ptest-removal-cmple.ll
@@ -320,8 +320,8 @@ define i1 @cmp32_ptest_any_xx(<vscale x 16 x i1> %pg, <vscale x 4 x i32> %a, <vs
 define i1 @cmp8_ptest_first_ax(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
 ; CHECK-LABEL: cmp8_ptest_first_ax:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cmpge p0.b, p0/z, z0.b, z1.b
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    ptest p1, p0.b
 ; CHECK-NEXT:    cset w0, mi
 ; CHECK-NEXT:    ret
@@ -338,8 +338,8 @@ define i1 @cmp8_ptest_first_ax(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <v
 define i1 @cmp8_ptest_last_ax(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
 ; CHECK-LABEL: cmp8_ptest_last_ax:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    cmpge p0.b, p0/z, z0.b, z1.b
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    ptest p1, p0.b
 ; CHECK-NEXT:    cset w0, lo
 ; CHECK-NEXT:    ret
@@ -371,8 +371,8 @@ define i1 @cmp8_ptest_any_ax(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %a, <vsc
 define i1 @cmp32_ptest_first_ax(<vscale x 16 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: cmp32_ptest_first_ax:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    cmpge p0.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ptest p1, p0.b
 ; CHECK-NEXT:    cset w0, mi
 ; CHECK-NEXT:    ret
@@ -390,8 +390,8 @@ define i1 @cmp32_ptest_first_ax(<vscale x 16 x i1> %pg, <vscale x 4 x i32> %a, <
 define i1 @cmp32_ptest_last_ax(<vscale x 16 x i1> %pg, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: cmp32_ptest_last_ax:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    cmpge p0.s, p0/z, z0.s, z1.s
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ptest p1, p0.b
 ; CHECK-NEXT:    cset w0, lo
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-redundant-store.ll b/llvm/test/CodeGen/AArch64/sve-redundant-store.ll
index 6873404724f1d..508fe5d5a58a5 100644
--- a/llvm/test/CodeGen/AArch64/sve-redundant-store.ll
+++ b/llvm/test/CodeGen/AArch64/sve-redundant-store.ll
@@ -35,8 +35,8 @@ entry:
 define void @keep_scalable_store(ptr writeonly %ptr, ptr %a, <vscale x 4 x i32> %b) {
 ; CHECK-LABEL: keep_scalable_store:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp q2, q1, [x1]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    stp q2, q1, [x0]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-split-extract-elt.ll b/llvm/test/CodeGen/AArch64/sve-split-extract-elt.ll
index a1c2ec9c7e1d4..76190eba870de 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-extract-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-extract-elt.ll
@@ -22,15 +22,15 @@ define i8 @split_extract_32i8_idx(<vscale x 32 x i8> %a, i32 %idx) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rdvl x8, #2
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w9, w0
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x9, x8
-; CHECK-NEXT:    csel x8, x9, x8, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1b { z1.b }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
+; CHECK-NEXT:    csel x8, x9, x8, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    ldrb w0, [x9, x8]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -46,15 +46,15 @@ define i16 @split_extract_16i16_idx(<vscale x 16 x i16> %a, i32 %idx) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w9, w0
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x9, x8
-; CHECK-NEXT:    csel x8, x9, x8, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1h { z1.h }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
+; CHECK-NEXT:    csel x8, x9, x8, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    ldrh w0, [x9, x8, lsl #1]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -70,15 +70,15 @@ define i32 @split_extract_8i32_idx(<vscale x 8 x i32> %a, i32 %idx) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cnth x8
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w9, w0
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x9, x8
-; CHECK-NEXT:    csel x8, x9, x8, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
+; CHECK-NEXT:    csel x8, x9, x8, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    ldr w0, [x9, x8, lsl #2]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -94,15 +94,15 @@ define i64 @split_extract_8i64_idx(<vscale x 8 x i64> %a, i32 %idx) {
 ; CHECK-NEXT:    addvl sp, sp, #-4
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 32 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cnth x8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w9, w0
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x9, x8
-; CHECK-NEXT:    csel x8, x9, x8, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1d { z3.d }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    csel x8, x9, x8, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1d { z1.d }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
 ; CHECK-NEXT:    ldr x0, [x9, x8, lsl #3]
@@ -140,15 +140,15 @@ define i16 @split_extract_16i16(<vscale x 16 x i16> %a) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rdvl x8, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w9, #128 // =0x80
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, #128
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1h { z1.h }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    ldrh w0, [x9, x8, lsl #1]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -164,16 +164,16 @@ define i32 @split_extract_16i32(<vscale x 16 x i32> %a) {
 ; CHECK-NEXT:    addvl sp, sp, #-4
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 32 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    rdvl x8, #1
 ; CHECK-NEXT:    mov w9, #34464 // =0x86a0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movk w9, #1, lsl #16
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, x9
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1w { z3.s }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1w { z2.s }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    ldr w0, [x9, x8, lsl #2]
@@ -191,15 +191,15 @@ define i64 @split_extract_4i64(<vscale x 4 x i64> %a) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cntw x8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w9, #10 // =0xa
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, #10
-; CHECK-NEXT:    csel x8, x8, x9, lo
-; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    st1d { z1.d }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
+; CHECK-NEXT:    csel x8, x8, x9, lo
+; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    ldr x0, [x9, x8, lsl #3]
 ; CHECK-NEXT:    addvl sp, sp, #2
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll b/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
index 3997409172d03..bc015116917d8 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-fcvt.ll
@@ -6,9 +6,9 @@
 define <vscale x 8 x float> @fcvts_nxv8f16(<vscale x 8 x half> %a) {
 ; CHECK-LABEL: fcvts_nxv8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z1.s, z0.h
 ; CHECK-NEXT:    uunpkhi z2.s, z0.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.s, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -21,9 +21,9 @@ define <vscale x 8 x float> @fcvts_nxv8f16(<vscale x 8 x half> %a) {
 define <vscale x 4 x double> @fcvtd_nxv4f16(<vscale x 4 x half> %a) {
 ; CHECK-LABEL: fcvtd_nxv4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -37,8 +37,8 @@ define <vscale x 8 x double> @fcvtd_nxv8f16(<vscale x 8 x half> %a) {
 ; CHECK-LABEL: fcvtd_nxv8f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    uunpklo z1.s, z0.h
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpkhi z0.s, z0.h
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z2.d, z1.s
 ; CHECK-NEXT:    uunpkhi z1.d, z1.s
 ; CHECK-NEXT:    uunpklo z3.d, z0.s
@@ -58,9 +58,9 @@ define <vscale x 8 x double> @fcvtd_nxv8f16(<vscale x 8 x half> %a) {
 define <vscale x 4 x double> @fcvtd_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtd_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -73,11 +73,11 @@ define <vscale x 4 x double> @fcvtd_nxv4f32(<vscale x 4 x float> %a) {
 define <vscale x 8 x double> @fcvtd_nxv8f32(<vscale x 8 x float> %a) {
 ; CHECK-LABEL: fcvtd_nxv8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z2.d, z0.s
 ; CHECK-NEXT:    uunpkhi z3.d, z0.s
 ; CHECK-NEXT:    uunpklo z4.d, z1.s
 ; CHECK-NEXT:    uunpkhi z5.d, z1.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvt z0.d, p0/m, z2.s
 ; CHECK-NEXT:    movprfx z1, z3
@@ -195,9 +195,9 @@ define <vscale x 8 x i16> @fcvtzs_h_nxv8f64(<vscale x 8 x double> %a) {
 define <vscale x 4 x i64> @fcvtzs_d_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtzs_d_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -210,11 +210,11 @@ define <vscale x 4 x i64> @fcvtzs_d_nxv4f32(<vscale x 4 x float> %a) {
 define <vscale x 16 x i32> @fcvtzs_s_nxv16f16(<vscale x 16 x half> %a) {
 ; CHECK-LABEL: fcvtzs_s_nxv16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uunpklo z2.s, z0.h
 ; CHECK-NEXT:    uunpkhi z3.s, z0.h
 ; CHECK-NEXT:    uunpklo z4.s, z1.h
 ; CHECK-NEXT:    uunpkhi z5.s, z1.h
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movprfx z0, z2
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z2.h
 ; CHECK-NEXT:    movprfx z1, z3
@@ -247,9 +247,9 @@ define <vscale x 4 x i32> @fcvtzu_s_nxv4f64(<vscale x 4 x double> %a) {
 define <vscale x 4 x i64> @fcvtzu_d_nxv4f32(<vscale x 4 x float> %a) {
 ; CHECK-LABEL: fcvtzu_d_nxv4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -295,8 +295,8 @@ define <vscale x 16 x float> @scvtf_s_nxv16i8(<vscale x 16 x i8> %a) {
 ; CHECK-LABEL: scvtf_s_nxv16i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sunpklo z1.h, z0.b
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z0.h, z0.b
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z2.s, z1.h
 ; CHECK-NEXT:    sunpkhi z1.s, z1.h
 ; CHECK-NEXT:    sunpklo z3.s, z0.h
@@ -316,9 +316,9 @@ define <vscale x 16 x float> @scvtf_s_nxv16i8(<vscale x 16 x i8> %a) {
 define <vscale x 4 x double> @scvtf_d_nxv4i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: scvtf_d_nxv4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    sunpklo z1.d, z0.s
 ; CHECK-NEXT:    sunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -333,8 +333,8 @@ define <vscale x 4 x double> @scvtf_d_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    punpklo p2.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p2/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z1.d, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    scvtf z0.d, p1/m, z0.d
 ; CHECK-NEXT:    scvtf z1.d, p1/m, z1.d
@@ -378,9 +378,9 @@ define <vscale x 8 x half> @ucvtf_h_nxv8i64(<vscale x 8 x i64> %a) {
 define <vscale x 4 x double> @ucvtf_d_nxv4i32(<vscale x 4 x i32> %a) {
 ; CHECK-LABEL: ucvtf_d_nxv4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uunpklo z1.d, z0.s
 ; CHECK-NEXT:    uunpkhi z2.d, z0.s
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    movprfx z0, z1
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -395,8 +395,8 @@ define <vscale x 4 x double> @ucvtf_d_nxv4i1(<vscale x 4 x i1> %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    punpklo p2.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z0.d, p2/z, #1 // =0x1
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z1.d, p0/z, #1 // =0x1
 ; CHECK-NEXT:    ucvtf z0.d, p1/m, z0.d
 ; CHECK-NEXT:    ucvtf z1.d, p1/m, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-split-fp-reduce.ll b/llvm/test/CodeGen/AArch64/sve-split-fp-reduce.ll
index 7f642882eddbe..696b6c34ef041 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-fp-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-fp-reduce.ll
@@ -23,8 +23,8 @@ define double @fadda_nxv8f64(double %init, <vscale x 8 x double> %a) {
 define float @faddv_nxv8f32(float %init, <vscale x 8 x float> %a) {
 ; CHECK-LABEL: faddv_nxv8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fadd z1.s, z1.s, z2.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    faddv s1, p0, z1.s
 ; CHECK-NEXT:    fadd s0, s0, s1
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-split-insert-elt.ll b/llvm/test/CodeGen/AArch64/sve-split-insert-elt.ll
index 5441659fa5cb4..75366384cb750 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-insert-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-insert-elt.ll
@@ -6,9 +6,9 @@
 define <vscale x 8 x i8> @promote_insert_8i8(<vscale x 8 x i8> %a, i8 %elt, i64 %idx) {
 ; CHECK-LABEL: promote_insert_8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    index z1.h, #0, #1
 ; CHECK-NEXT:    mov z2.h, w1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
 ; CHECK-NEXT:    mov z0.h, p0/m, w0
 ; CHECK-NEXT:    ret
@@ -23,13 +23,13 @@ define <vscale x 32 x i8> @split_insert_32i8_idx(<vscale x 32 x i8> %a, i8 %elt,
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rdvl x8, #2
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x1, x8
-; CHECK-NEXT:    csel x8, x1, x8, lo
 ; CHECK-NEXT:    st1b { z1.b }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    csel x8, x1, x8, lo
 ; CHECK-NEXT:    st1b { z0.b }, p0, [sp]
 ; CHECK-NEXT:    strb w0, [x9, x8]
 ; CHECK-NEXT:    ld1b { z0.b }, p0/z, [sp]
@@ -48,13 +48,13 @@ define <vscale x 8 x float> @split_insert_8f32_idx(<vscale x 8 x float> %a, floa
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cnth x8
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x0, x8
-; CHECK-NEXT:    csel x8, x0, x8, lo
 ; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    csel x8, x0, x8, lo
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    str s2, [x9, x8, lsl #2]
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [sp]
@@ -73,13 +73,13 @@ define <vscale x 8 x i64> @split_insert_8i64_idx(<vscale x 8 x i64> %a, i64 %elt
 ; CHECK-NEXT:    addvl sp, sp, #-4
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 32 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cnth x8
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov x9, sp
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x1, x8
-; CHECK-NEXT:    csel x8, x1, x8, lo
 ; CHECK-NEXT:    st1d { z3.d }, p0, [sp, #3, mul vl]
+; CHECK-NEXT:    csel x8, x1, x8, lo
 ; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    st1d { z1.d }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
@@ -100,9 +100,9 @@ define <vscale x 8 x i64> @split_insert_8i64_idx(<vscale x 8 x i64> %a, i64 %elt
 define <vscale x 4 x i16> @promote_insert_4i16(<vscale x 4 x i16> %a, i16 %elt) {
 ; CHECK-LABEL: promote_insert_4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #5 // =0x5
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
 ; CHECK-NEXT:    mov z0.s, p0/m, w0
@@ -117,9 +117,9 @@ define <vscale x 4 x i16> @promote_insert_4i16(<vscale x 4 x i16> %a, i16 %elt)
 define <vscale x 32 x i8> @split_insert_32i8(<vscale x 32 x i8> %a, i8 %elt) {
 ; CHECK-LABEL: split_insert_32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z2.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z3.b, w8
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z2.b, z3.b
 ; CHECK-NEXT:    mov z0.b, p0/m, w0
@@ -135,14 +135,14 @@ define <vscale x 32 x i16> @split_insert_32i16(<vscale x 32 x i16> %a, i16 %elt)
 ; CHECK-NEXT:    addvl sp, sp, #-4
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 32 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rdvl x8, #2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w9, #128 // =0x80
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, #128
+; CHECK-NEXT:    st1h { z3.h }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    st1h { z3.h }, p0, [sp, #3, mul vl]
 ; CHECK-NEXT:    st1h { z2.h }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    st1h { z1.h }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1h { z0.h }, p0, [sp]
@@ -165,15 +165,15 @@ define <vscale x 8 x i32> @split_insert_8i32(<vscale x 8 x i32> %a, i32 %elt) {
 ; CHECK-NEXT:    addvl sp, sp, #-2
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x10, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 16 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cnth x8
 ; CHECK-NEXT:    mov w9, #16960 // =0x4240
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movk w9, #15, lsl #16
 ; CHECK-NEXT:    sub x8, x8, #1
 ; CHECK-NEXT:    cmp x8, x9
+; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    csel x8, x8, x9, lo
 ; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    st1w { z1.s }, p0, [sp, #1, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p0, [sp]
 ; CHECK-NEXT:    str w0, [x9, x8, lsl #2]
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [sp]
diff --git a/llvm/test/CodeGen/AArch64/sve-split-int-reduce.ll b/llvm/test/CodeGen/AArch64/sve-split-int-reduce.ll
index 42f3a163d14cc..dd7b15ef5ee6f 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-int-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-int-reduce.ll
@@ -17,8 +17,8 @@ define i8 @andv_nxv8i8(<vscale x 8 x i8> %a) {
 define i32 @andv_nxv8i32(<vscale x 8 x i32> %a) {
 ; CHECK-LABEL: andv_nxv8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    andv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    ret
@@ -71,8 +71,8 @@ define i16 @xorv_nxv2i16(<vscale x 2 x i16> %a) {
 define i32 @xorv_nxv8i32(<vscale x 8 x i32> %a) {
 ; CHECK-LABEL: xorv_nxv8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    eor z0.d, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    eorv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    ret
@@ -97,8 +97,8 @@ define i16 @uaddv_nxv4i16(<vscale x 4 x i16> %a) {
 define i16 @uaddv_nxv16i16(<vscale x 16 x i16> %a) {
 ; CHECK-LABEL: uaddv_nxv16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    uaddv d0, p0, z0.h
 ; CHECK-NEXT:    fmov x0, d0
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
@@ -127,8 +127,8 @@ define i32 @uaddv_nxv16i32(<vscale x 16 x i32> %a) {
 define i32 @umin_nxv2i32(<vscale x 2 x i32> %a) {
 ; CHECK-LABEL: umin_nxv2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uminv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
diff --git a/llvm/test/CodeGen/AArch64/sve-split-load.ll b/llvm/test/CodeGen/AArch64/sve-split-load.ll
index af03059cf0d8b..754f0339702dc 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-load.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-load.ll
@@ -93,8 +93,8 @@ define <vscale x 32 x i16> @masked_load_split_32i16(ptr %a, <vscale x 32 x i1> %
 ; CHECK-NEXT:    punpklo p2.h, p0.b
 ; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    punpklo p3.h, p1.b
-; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    ld1h { z0.h }, p2/z, [x0]
+; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0, #1, mul vl]
 ; CHECK-NEXT:    ld1h { z2.h }, p3/z, [x0, #2, mul vl]
 ; CHECK-NEXT:    ld1h { z3.h }, p1/z, [x0, #3, mul vl]
@@ -123,8 +123,8 @@ define <vscale x 8 x i64> @masked_load_split_8i64(ptr %a, <vscale x 8 x i1> %pg)
 ; CHECK-NEXT:    punpklo p2.h, p1.b
 ; CHECK-NEXT:    punpkhi p1.h, p1.b
 ; CHECK-NEXT:    punpklo p3.h, p0.b
-; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    ld1d { z0.d }, p2/z, [x0]
+; CHECK-NEXT:    punpkhi p0.h, p0.b
 ; CHECK-NEXT:    ld1d { z1.d }, p1/z, [x0, #1, mul vl]
 ; CHECK-NEXT:    ld1d { z2.d }, p3/z, [x0, #2, mul vl]
 ; CHECK-NEXT:    ld1d { z3.d }, p0/z, [x0, #3, mul vl]
diff --git a/llvm/test/CodeGen/AArch64/sve-split-store.ll b/llvm/test/CodeGen/AArch64/sve-split-store.ll
index 90ec783ea4dbc..affa9a18ac182 100644
--- a/llvm/test/CodeGen/AArch64/sve-split-store.ll
+++ b/llvm/test/CodeGen/AArch64/sve-split-store.ll
@@ -81,8 +81,8 @@ define void @masked_store_split_32i16(<vscale x 32 x i16> %data, ptr %a, <vscale
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
 ; CHECK-NEXT:    punpkhi p3.h, p0.b
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z3.h }, p2, [x0, #3, mul vl]
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1h { z2.h }, p1, [x0, #2, mul vl]
 ; CHECK-NEXT:    st1h { z1.h }, p3, [x0, #1, mul vl]
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
@@ -110,11 +110,11 @@ define void @masked_store_split_8i64(<vscale x 8 x i64> %data, ptr %a, <vscale x
 ; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    punpkhi p2.h, p1.b
 ; CHECK-NEXT:    punpklo p1.h, p1.b
-; CHECK-NEXT:    punpkhi p3.h, p0.b
-; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1d { z3.d }, p2, [x0, #3, mul vl]
+; CHECK-NEXT:    punpkhi p2.h, p0.b
+; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    st1d { z2.d }, p1, [x0, #2, mul vl]
-; CHECK-NEXT:    st1d { z1.d }, p3, [x0, #1, mul vl]
+; CHECK-NEXT:    st1d { z1.d }, p2, [x0, #1, mul vl]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.nxv8i64(<vscale x 8 x i64> %data, ptr %a, i32 1, <vscale x 8 x i1> %pg)
diff --git a/llvm/test/CodeGen/AArch64/sve-srem-combine-loop.ll b/llvm/test/CodeGen/AArch64/sve-srem-combine-loop.ll
index 9c3d4b1e5a810..f556d60d23b88 100644
--- a/llvm/test/CodeGen/AArch64/sve-srem-combine-loop.ll
+++ b/llvm/test/CodeGen/AArch64/sve-srem-combine-loop.ll
@@ -6,8 +6,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define <vscale x 4 x i32> @srem_combine_loop(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: srem_combine_loop:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.d, z0.d
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, #2 // =0x2
 ; CHECK-NEXT:    asrd z1.s, p0/m, z1.s, #1
 ; CHECK-NEXT:    mls z0.s, p0/m, z1.s, z2.s
diff --git a/llvm/test/CodeGen/AArch64/sve-st1-addressing-mode-reg-imm.ll b/llvm/test/CodeGen/AArch64/sve-st1-addressing-mode-reg-imm.ll
index 728041d3f916b..3273e6b384f63 100644
--- a/llvm/test/CodeGen/AArch64/sve-st1-addressing-mode-reg-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-st1-addressing-mode-reg-imm.ll
@@ -105,8 +105,8 @@ define void @st1d_inbound(<vscale x 2 x i64> %data, ptr %a) {
 define void @store_nxv2f32(ptr %out) {
 ; CHECK-LABEL: store_nxv2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    fmov z0.s, #1.00000000
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    st1w { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
   %ins = insertelement <vscale x 2 x float> undef, float 1.0, i32 0
@@ -118,8 +118,8 @@ define void @store_nxv2f32(ptr %out) {
 define void @store_nxv4f16(ptr %out) {
 ; CHECK-LABEL: store_nxv4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    fmov z0.h, #1.00000000
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   %ins = insertelement <vscale x 4 x half> undef, half 1.0, i32 0
@@ -133,9 +133,9 @@ define void @store_nxv4f16(ptr %out) {
 define void @store_nxv6f32(ptr %out) {
 ; CHECK-LABEL: store_nxv6f32:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmov z0.s, #1.00000000
 ; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ptrue p1.s
-; CHECK-NEXT:    fmov z0.s, #1.00000000
 ; CHECK-NEXT:    st1w { z0.d }, p0, [x0, #2, mul vl]
 ; CHECK-NEXT:    st1w { z0.s }, p1, [x0]
 ; CHECK-NEXT:    ret
@@ -148,9 +148,9 @@ define void @store_nxv6f32(ptr %out) {
 define void @store_nxv12f16(ptr %out) {
 ; CHECK-LABEL: store_nxv12f16:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    fmov z0.h, #1.00000000
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    fmov z0.h, #1.00000000
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0, #2, mul vl]
 ; CHECK-NEXT:    st1h { z0.h }, p1, [x0]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-stepvector.ll b/llvm/test/CodeGen/AArch64/sve-stepvector.ll
index 6f5a31248de7e..4c5f27d3e7093 100644
--- a/llvm/test/CodeGen/AArch64/sve-stepvector.ll
+++ b/llvm/test/CodeGen/AArch64/sve-stepvector.ll
@@ -208,9 +208,9 @@ entry:
 define <vscale x 4 x i32> @multiple_use_stepvector_nxv4i32_1(i32 %data) {
 ; CHECK-LABEL: multiple_use_stepvector_nxv4i32_1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    index z0.s, w0, #1
 ; CHECK-NEXT:    mov z1.s, w0
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mul z1.s, p0/m, z1.s, z0.s
 ; CHECK-NEXT:    sub z0.s, z1.s, z0.s
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-bit-counting.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-bit-counting.ll
index be5c318e675df..d547f99a0230a 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-bit-counting.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-bit-counting.ll
@@ -11,8 +11,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define <4 x i8> @ctlz_v4i8(<4 x i8> %op) {
 ; CHECK-LABEL: ctlz_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
 ; CHECK-NEXT:    clz z0.h, p0/m, z0.h
 ; CHECK-NEXT:    sub z0.h, z0.h, #8 // =0x8
@@ -49,8 +49,8 @@ define <16 x i8> @ctlz_v16i8(<16 x i8> %op) {
 define void @ctlz_v32i8(ptr %a) {
 ; CHECK-LABEL: ctlz_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    clz z0.b, p0/m, z0.b
 ; CHECK-NEXT:    clz z1.b, p0/m, z1.b
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -64,8 +64,8 @@ define void @ctlz_v32i8(ptr %a) {
 define <2 x i16> @ctlz_v2i16(<2 x i16> %op) {
 ; CHECK-LABEL: ctlz_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    clz z0.s, p0/m, z0.s
 ; CHECK-NEXT:    sub z0.s, z0.s, #16 // =0x10
@@ -102,8 +102,8 @@ define <8 x i16> @ctlz_v8i16(<8 x i16> %op) {
 define void @ctlz_v16i16(ptr %a) {
 ; CHECK-LABEL: ctlz_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    clz z0.h, p0/m, z0.h
 ; CHECK-NEXT:    clz z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -141,8 +141,8 @@ define <4 x i32> @ctlz_v4i32(<4 x i32> %op) {
 define void @ctlz_v8i32(ptr %a) {
 ; CHECK-LABEL: ctlz_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    clz z0.s, p0/m, z0.s
 ; CHECK-NEXT:    clz z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -180,8 +180,8 @@ define <2 x i64> @ctlz_v2i64(<2 x i64> %op) {
 define void @ctlz_v4i64(ptr %a) {
 ; CHECK-LABEL: ctlz_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    clz z0.d, p0/m, z0.d
 ; CHECK-NEXT:    clz z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -199,8 +199,8 @@ define void @ctlz_v4i64(ptr %a) {
 define <4 x i8> @ctpop_v4i8(<4 x i8> %op) {
 ; CHECK-LABEL: ctpop_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
 ; CHECK-NEXT:    cnt z0.h, p0/m, z0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -236,8 +236,8 @@ define <16 x i8> @ctpop_v16i8(<16 x i8> %op) {
 define void @ctpop_v32i8(ptr %a) {
 ; CHECK-LABEL: ctpop_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    cnt z0.b, p0/m, z0.b
 ; CHECK-NEXT:    cnt z1.b, p0/m, z1.b
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -251,8 +251,8 @@ define void @ctpop_v32i8(ptr %a) {
 define <2 x i16> @ctpop_v2i16(<2 x i16> %op) {
 ; CHECK-LABEL: ctpop_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    cnt z0.s, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -288,8 +288,8 @@ define <8 x i16> @ctpop_v8i16(<8 x i16> %op) {
 define void @ctpop_v16i16(ptr %a) {
 ; CHECK-LABEL: ctpop_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    cnt z0.h, p0/m, z0.h
 ; CHECK-NEXT:    cnt z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -327,8 +327,8 @@ define <4 x i32> @ctpop_v4i32(<4 x i32> %op) {
 define void @ctpop_v8i32(ptr %a) {
 ; CHECK-LABEL: ctpop_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    cnt z0.s, p0/m, z0.s
 ; CHECK-NEXT:    cnt z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -366,8 +366,8 @@ define <2 x i64> @ctpop_v2i64(<2 x i64> %op) {
 define void @ctpop_v4i64(ptr %a) {
 ; CHECK-LABEL: ctpop_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    cnt z0.d, p0/m, z0.d
 ; CHECK-NEXT:    cnt z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -385,8 +385,8 @@ define void @ctpop_v4i64(ptr %a) {
 define <4 x i8> @cttz_v4i8(<4 x i8> %op) {
 ; CHECK-LABEL: cttz_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    orr z0.h, z0.h, #0x100
 ; CHECK-NEXT:    rbit z0.h, p0/m, z0.h
 ; CHECK-NEXT:    clz z0.h, p0/m, z0.h
@@ -425,8 +425,8 @@ define <16 x i8> @cttz_v16i8(<16 x i8> %op) {
 define void @cttz_v32i8(ptr %a) {
 ; CHECK-LABEL: cttz_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    rbit z0.b, p0/m, z0.b
 ; CHECK-NEXT:    rbit z1.b, p0/m, z1.b
 ; CHECK-NEXT:    clz z0.b, p0/m, z0.b
@@ -442,8 +442,8 @@ define void @cttz_v32i8(ptr %a) {
 define <2 x i16> @cttz_v2i16(<2 x i16> %op) {
 ; CHECK-LABEL: cttz_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    orr z0.s, z0.s, #0x10000
 ; CHECK-NEXT:    rbit z0.s, p0/m, z0.s
 ; CHECK-NEXT:    clz z0.s, p0/m, z0.s
@@ -482,8 +482,8 @@ define <8 x i16> @cttz_v8i16(<8 x i16> %op) {
 define void @cttz_v16i16(ptr %a) {
 ; CHECK-LABEL: cttz_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    rbit z0.h, p0/m, z0.h
 ; CHECK-NEXT:    rbit z1.h, p0/m, z1.h
 ; CHECK-NEXT:    clz z0.h, p0/m, z0.h
@@ -525,8 +525,8 @@ define <4 x i32> @cttz_v4i32(<4 x i32> %op) {
 define void @cttz_v8i32(ptr %a) {
 ; CHECK-LABEL: cttz_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    rbit z0.s, p0/m, z0.s
 ; CHECK-NEXT:    rbit z1.s, p0/m, z1.s
 ; CHECK-NEXT:    clz z0.s, p0/m, z0.s
@@ -568,8 +568,8 @@ define <2 x i64> @cttz_v2i64(<2 x i64> %op) {
 define void @cttz_v4i64(ptr %a) {
 ; CHECK-LABEL: cttz_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    rbit z0.d, p0/m, z0.d
 ; CHECK-NEXT:    rbit z1.d, p0/m, z1.d
 ; CHECK-NEXT:    clz z0.d, p0/m, z0.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ext-loads.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ext-loads.ll
index 251a7c3b18a9f..0aefba2d4c6ab 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ext-loads.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ext-loads.ll
@@ -169,15 +169,15 @@ define <16 x i64> @load_zext_v16i16i64(ptr %ap)  {
 ; CHECK-NEXT:    ld1h { z1.d }, p0/z, [x0, x8, lsl #1]
 ; CHECK-NEXT:    mov x8, #8 // =0x8
 ; CHECK-NEXT:    ld1h { z2.d }, p0/z, [x0, x9, lsl #1]
-; CHECK-NEXT:    mov x9, #10 // =0xa
 ; CHECK-NEXT:    ld1h { z3.d }, p0/z, [x0, x10, lsl #1]
-; CHECK-NEXT:    mov x10, #12 // =0xc
+; CHECK-NEXT:    mov x9, #10 // =0xa
 ; CHECK-NEXT:    ld1h { z4.d }, p0/z, [x0, x8, lsl #1]
-; CHECK-NEXT:    mov x8, #14 // =0xe
+; CHECK-NEXT:    mov x8, #12 // =0xc
+; CHECK-NEXT:    mov x10, #14 // =0xe
 ; CHECK-NEXT:    ld1h { z0.d }, p0/z, [x0]
 ; CHECK-NEXT:    ld1h { z5.d }, p0/z, [x0, x9, lsl #1]
-; CHECK-NEXT:    ld1h { z6.d }, p0/z, [x0, x10, lsl #1]
-; CHECK-NEXT:    ld1h { z7.d }, p0/z, [x0, x8, lsl #1]
+; CHECK-NEXT:    ld1h { z6.d }, p0/z, [x0, x8, lsl #1]
+; CHECK-NEXT:    ld1h { z7.d }, p0/z, [x0, x10, lsl #1]
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 killed $z1
 ; CHECK-NEXT:    // kill: def $q2 killed $q2 killed $z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fcopysign.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fcopysign.ll
index 2ace0bca274af..0d6675def8b52 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fcopysign.ll
@@ -241,8 +241,8 @@ define void @test_copysign_v2f32_v2f64(ptr %ap, ptr %bp) {
 ; SVE-NEXT:    ptrue p0.d
 ; SVE-NEXT:    ldr q0, [x1]
 ; SVE-NEXT:    ldr d1, [x0]
-; SVE-NEXT:    and z1.s, z1.s, #0x7fffffff
 ; SVE-NEXT:    fcvt z0.s, p0/m, z0.d
+; SVE-NEXT:    and z1.s, z1.s, #0x7fffffff
 ; SVE-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; SVE-NEXT:    and z0.s, z0.s, #0x80000000
 ; SVE-NEXT:    orr z0.d, z1.d, z0.d
@@ -274,8 +274,8 @@ define void @test_copysign_v2f32_v2f64(ptr %ap, ptr %bp) {
 define void @test_copysign_v4f32_v4f64(ptr %ap, ptr %bp) {
 ; SVE-LABEL: test_copysign_v4f32_v4f64:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.d
 ; SVE-NEXT:    ldp q0, q1, [x1]
+; SVE-NEXT:    ptrue p0.d
 ; SVE-NEXT:    fcvt z1.s, p0/m, z1.d
 ; SVE-NEXT:    fcvt z0.s, p0/m, z0.d
 ; SVE-NEXT:    ptrue p0.s, vl2
@@ -291,8 +291,8 @@ define void @test_copysign_v4f32_v4f64(ptr %ap, ptr %bp) {
 ;
 ; SVE2-LABEL: test_copysign_v4f32_v4f64:
 ; SVE2:       // %bb.0:
-; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    ldp q0, q1, [x1]
+; SVE2-NEXT:    ptrue p0.d
 ; SVE2-NEXT:    ldr q2, [x0]
 ; SVE2-NEXT:    fcvt z1.s, p0/m, z1.d
 ; SVE2-NEXT:    fcvt z0.s, p0/m, z0.d
@@ -319,8 +319,8 @@ define void @test_copysign_v2f64_v2f32(ptr %ap, ptr %bp) {
 ; SVE:       // %bb.0:
 ; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldr q0, [x0]
-; SVE-NEXT:    and z0.d, z0.d, #0x7fffffffffffffff
 ; SVE-NEXT:    ld1w { z1.d }, p0/z, [x1]
+; SVE-NEXT:    and z0.d, z0.d, #0x7fffffffffffffff
 ; SVE-NEXT:    fcvt z1.d, p0/m, z1.s
 ; SVE-NEXT:    and z1.d, z1.d, #0x8000000000000000
 ; SVE-NEXT:    orr z0.d, z0.d, z1.d
@@ -354,10 +354,10 @@ define void @test_copysign_v4f64_v4f32(ptr %ap, ptr %bp) {
 ; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    mov x8, #2 // =0x2
 ; SVE-NEXT:    ldp q2, q3, [x0]
-; SVE-NEXT:    and z2.d, z2.d, #0x7fffffffffffffff
-; SVE-NEXT:    and z3.d, z3.d, #0x7fffffffffffffff
 ; SVE-NEXT:    ld1w { z0.d }, p0/z, [x1]
 ; SVE-NEXT:    ld1w { z1.d }, p0/z, [x1, x8, lsl #2]
+; SVE-NEXT:    and z2.d, z2.d, #0x7fffffffffffffff
+; SVE-NEXT:    and z3.d, z3.d, #0x7fffffffffffffff
 ; SVE-NEXT:    fcvt z0.d, p0/m, z0.s
 ; SVE-NEXT:    fcvt z1.d, p0/m, z1.s
 ; SVE-NEXT:    and z0.d, z0.d, #0x8000000000000000
@@ -397,8 +397,8 @@ define void @test_copysign_v4f16_v4f32(ptr %ap, ptr %bp) {
 ; SVE-NEXT:    ptrue p0.s
 ; SVE-NEXT:    ldr q0, [x1]
 ; SVE-NEXT:    ldr d1, [x0]
-; SVE-NEXT:    and z1.h, z1.h, #0x7fff
 ; SVE-NEXT:    fcvt z0.h, p0/m, z0.s
+; SVE-NEXT:    and z1.h, z1.h, #0x7fff
 ; SVE-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; SVE-NEXT:    and z0.h, z0.h, #0x8000
 ; SVE-NEXT:    orr z0.d, z1.d, z0.d
@@ -429,13 +429,13 @@ define void @test_copysign_v4f16_v4f64(ptr %ap, ptr %bp) {
 ; SVE:       // %bb.0:
 ; SVE-NEXT:    ldp q0, q1, [x1]
 ; SVE-NEXT:    ptrue p0.s, vl2
-; SVE-NEXT:    ptrue p1.s
 ; SVE-NEXT:    fcvtxn v1.2s, v1.2d
 ; SVE-NEXT:    fcvtxn v0.2s, v0.2d
 ; SVE-NEXT:    splice z0.s, p0, z0.s, z1.s
+; SVE-NEXT:    ptrue p0.s
 ; SVE-NEXT:    ldr d1, [x0]
 ; SVE-NEXT:    and z1.h, z1.h, #0x7fff
-; SVE-NEXT:    fcvt z0.h, p1/m, z0.s
+; SVE-NEXT:    fcvt z0.h, p0/m, z0.s
 ; SVE-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; SVE-NEXT:    and z0.h, z0.h, #0x8000
 ; SVE-NEXT:    orr z0.d, z1.d, z0.d
@@ -446,13 +446,13 @@ define void @test_copysign_v4f16_v4f64(ptr %ap, ptr %bp) {
 ; SVE2:       // %bb.0:
 ; SVE2-NEXT:    ldp q0, q1, [x1]
 ; SVE2-NEXT:    ptrue p0.s, vl2
-; SVE2-NEXT:    ptrue p1.s
 ; SVE2-NEXT:    ldr d2, [x0]
 ; SVE2-NEXT:    fcvtxn v1.2s, v1.2d
 ; SVE2-NEXT:    fcvtxn v0.2s, v0.2d
 ; SVE2-NEXT:    splice z0.s, p0, z0.s, z1.s
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    mov z1.h, #32767 // =0x7fff
-; SVE2-NEXT:    fcvt z0.h, p1/m, z0.s
+; SVE2-NEXT:    fcvt z0.h, p0/m, z0.s
 ; SVE2-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; SVE2-NEXT:    bsl z2.d, z2.d, z0.d, z1.d
 ; SVE2-NEXT:    str d2, [x0]
@@ -470,8 +470,8 @@ define void @test_copysign_v4f16_v4f64(ptr %ap, ptr %bp) {
 define void @test_copysign_v8f16_v8f32(ptr %ap, ptr %bp) {
 ; SVE-LABEL: test_copysign_v8f16_v8f32:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s
 ; SVE-NEXT:    ldp q0, q1, [x1]
+; SVE-NEXT:    ptrue p0.s
 ; SVE-NEXT:    fcvt z1.h, p0/m, z1.s
 ; SVE-NEXT:    fcvt z0.h, p0/m, z0.s
 ; SVE-NEXT:    ptrue p0.h, vl4
@@ -487,8 +487,8 @@ define void @test_copysign_v8f16_v8f32(ptr %ap, ptr %bp) {
 ;
 ; SVE2-LABEL: test_copysign_v8f16_v8f32:
 ; SVE2:       // %bb.0:
-; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    ldp q0, q1, [x1]
+; SVE2-NEXT:    ptrue p0.s
 ; SVE2-NEXT:    ldr q2, [x0]
 ; SVE2-NEXT:    fcvt z1.h, p0/m, z1.s
 ; SVE2-NEXT:    fcvt z0.h, p0/m, z0.s
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-arith.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-arith.ll
index c436dea8ff1b2..c2d6ed4e9ccf9 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-arith.ll
@@ -50,8 +50,8 @@ define <8 x half> @fadd_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fadd_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -94,8 +94,8 @@ define <4 x float> @fadd_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fadd_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -125,8 +125,8 @@ define <2 x double> @fadd_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fadd_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -186,8 +186,8 @@ define <8 x half> @fdiv_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fdiv_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fdiv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fdivr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -230,8 +230,8 @@ define <4 x float> @fdiv_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fdiv_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fdiv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fdivr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -261,8 +261,8 @@ define <2 x double> @fdiv_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fdiv_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fdiv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fdivr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -325,8 +325,8 @@ define <8 x half> @fma_v8f16(<8 x half> %op1, <8 x half> %op2, <8 x half> %op3)
 define void @fma_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.h, p0/m, z2.h, z1.h
@@ -373,8 +373,8 @@ define <4 x float> @fma_v4f32(<4 x float> %op1, <4 x float> %op2, <4 x float> %o
 define void @fma_v8f32(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.s, p0/m, z2.s, z1.s
@@ -407,8 +407,8 @@ define <2 x double> @fma_v2f64(<2 x double> %op1, <2 x double> %op2, <2 x double
 define void @fma_v4f64(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.d, p0/m, z2.d, z1.d
@@ -470,8 +470,8 @@ define <8 x half> @fmul_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fmul_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmul_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -514,8 +514,8 @@ define <4 x float> @fmul_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fmul_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmul_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -545,8 +545,8 @@ define <2 x double> @fmul_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fmul_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmul_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -603,8 +603,8 @@ define <8 x half> @fneg_v8f16(<8 x half> %op) {
 define void @fneg_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fneg_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fneg z0.h, p0/m, z0.h
 ; CHECK-NEXT:    fneg z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -642,8 +642,8 @@ define <4 x float> @fneg_v4f32(<4 x float> %op) {
 define void @fneg_v8f32(ptr %a) {
 ; CHECK-LABEL: fneg_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fneg z0.s, p0/m, z0.s
 ; CHECK-NEXT:    fneg z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -669,8 +669,8 @@ define <2 x double> @fneg_v2f64(<2 x double> %op) {
 define void @fneg_v4f64(ptr %a) {
 ; CHECK-LABEL: fneg_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fneg z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fneg z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -724,8 +724,8 @@ define <8 x half> @fsqrt_v8f16(<8 x half> %op) {
 define void @fsqrt_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fsqrt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fsqrt z0.h, p0/m, z0.h
 ; CHECK-NEXT:    fsqrt z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -763,8 +763,8 @@ define <4 x float> @fsqrt_v4f32(<4 x float> %op) {
 define void @fsqrt_v8f32(ptr %a) {
 ; CHECK-LABEL: fsqrt_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fsqrt z0.s, p0/m, z0.s
 ; CHECK-NEXT:    fsqrt z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -790,8 +790,8 @@ define <2 x double> @fsqrt_v2f64(<2 x double> %op) {
 define void @fsqrt_v4f64(ptr %a) {
 ; CHECK-LABEL: fsqrt_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fsqrt z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fsqrt z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -848,8 +848,8 @@ define <8 x half> @fsub_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fsub_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fsub_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fsubr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -892,8 +892,8 @@ define <4 x float> @fsub_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fsub_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fsub_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fsubr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -923,8 +923,8 @@ define <2 x double> @fsub_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fsub_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fsub_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fsubr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -981,8 +981,8 @@ define <8 x half> @fabs_v8f16(<8 x half> %op) {
 define void @fabs_v16f16(ptr %a) {
 ; CHECK-LABEL: fabs_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fabs z0.h, p0/m, z0.h
 ; CHECK-NEXT:    fabs z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -1020,8 +1020,8 @@ define <4 x float> @fabs_v4f32(<4 x float> %op) {
 define void @fabs_v8f32(ptr %a) {
 ; CHECK-LABEL: fabs_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fabs z0.s, p0/m, z0.s
 ; CHECK-NEXT:    fabs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -1047,8 +1047,8 @@ define <2 x double> @fabs_v2f64(<2 x double> %op) {
 define void @fabs_v4f64(ptr %a) {
 ; CHECK-LABEL: fabs_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fabs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fabs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-compares.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-compares.ll
index aad078f035f7d..e92694d1fc80d 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-compares.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-compares.ll
@@ -57,8 +57,8 @@ define <8 x i16> @fcmp_oeq_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fcmp_oeq_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_oeq_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmeq p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z2.h, z3.h
@@ -107,8 +107,8 @@ define <4 x i32> @fcmp_oeq_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fcmp_oeq_v8f32(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_oeq_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmeq p1.s, p0/z, z1.s, z0.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z2.s, z3.s
@@ -157,8 +157,8 @@ define <2 x i64> @fcmp_oeq_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fcmp_oeq_v4f64(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_oeq_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmeq p1.d, p0/z, z1.d, z0.d
 ; CHECK-NEXT:    fcmeq p0.d, p0/z, z2.d, z3.d
@@ -181,8 +181,8 @@ define void @fcmp_oeq_v4f64(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ueq_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ueq_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmuo p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmeq p2.h, p0/z, z1.h, z0.h
@@ -209,8 +209,8 @@ define void @fcmp_ueq_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_one_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_one_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmgt p2.h, p0/z, z1.h, z0.h
@@ -237,8 +237,8 @@ define void @fcmp_one_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_une_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_une_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmne p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmne p0.h, p0/z, z2.h, z3.h
@@ -261,8 +261,8 @@ define void @fcmp_une_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ogt_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ogt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z2.h, z3.h
@@ -285,8 +285,8 @@ define void @fcmp_ogt_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ugt_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ugt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z3.h, z2.h
@@ -312,8 +312,8 @@ define void @fcmp_ugt_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_olt_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_olt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z3.h, z2.h
@@ -336,8 +336,8 @@ define void @fcmp_olt_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ult_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ult_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z2.h, z3.h
@@ -363,8 +363,8 @@ define void @fcmp_ult_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_oge_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_oge_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z2.h, z3.h
@@ -387,8 +387,8 @@ define void @fcmp_oge_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_uge_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_uge_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z3.h, z2.h
@@ -414,8 +414,8 @@ define void @fcmp_uge_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ole_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ole_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z3.h, z2.h
@@ -438,8 +438,8 @@ define void @fcmp_ole_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ule_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ule_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z2.h, z3.h
@@ -465,8 +465,8 @@ define void @fcmp_ule_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_uno_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_uno_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmuo p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmuo p0.h, p0/z, z2.h, z3.h
@@ -489,8 +489,8 @@ define void @fcmp_uno_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ord_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ord_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmuo p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmuo p0.h, p0/z, z2.h, z3.h
@@ -516,8 +516,8 @@ define void @fcmp_ord_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_eq_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_eq_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmeq p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z2.h, z3.h
@@ -540,8 +540,8 @@ define void @fcmp_eq_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ne_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ne_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmne p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmne p0.h, p0/z, z2.h, z3.h
@@ -564,8 +564,8 @@ define void @fcmp_ne_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_gt_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_gt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z2.h, z3.h
@@ -588,8 +588,8 @@ define void @fcmp_gt_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_lt_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_lt_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmgt p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmgt p0.h, p0/z, z3.h, z2.h
@@ -612,8 +612,8 @@ define void @fcmp_lt_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_ge_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_ge_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z2.h, z3.h
@@ -636,8 +636,8 @@ define void @fcmp_ge_v16f16(ptr %a, ptr %b, ptr %c) {
 define void @fcmp_le_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fcmp_le_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fcmge p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmge p0.h, p0/z, z3.h, z2.h
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-convert.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-convert.ll
index 18f9a4d371d0c..9bdde14e8d83d 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-convert.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-convert.ll
@@ -8,9 +8,9 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @fp_convert_combine_crash(ptr %a, ptr %b) {
 ; CHECK-LABEL: fp_convert_combine_crash:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmov z0.s, #8.00000000
+; CHECK-NEXT:    ldp q1, q2, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fmul z1.s, p0/m, z1.s, z0.s
 ; CHECK-NEXT:    fmul z0.s, p0/m, z0.s, z2.s
 ; CHECK-NEXT:    fcvtzs z1.s, p0/m, z1.s
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
index 28e02da53af43..244a405101739 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-extend-trunc.ll
@@ -11,8 +11,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @fcvt_v2f16_to_v2f32(<2 x half> %a, ptr %b) {
 ; CHECK-LABEL: fcvt_v2f16_to_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvt z0.s, p0/m, z0.h
 ; CHECK-NEXT:    str d0, [x0]
@@ -25,8 +25,8 @@ define void @fcvt_v2f16_to_v2f32(<2 x half> %a, ptr %b) {
 define void @fcvt_v4f16_to_v4f32(<4 x half> %a, ptr %b) {
 ; CHECK-LABEL: fcvt_v4f16_to_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvt z0.s, p0/m, z0.h
 ; CHECK-NEXT:    str q0, [x0]
@@ -371,8 +371,8 @@ define void @fcvt_v4f32_v4f16(ptr %a, ptr %b) {
 define void @fcvt_v8f32_v8f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvt_v8f32_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    mov x8, #4 // =0x4
 ; CHECK-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NEXT:    fcvt z1.h, p0/m, z1.s
@@ -420,8 +420,8 @@ define void @fcvt_v2f64_v2f16(ptr %a, ptr %b) {
 define void @fcvt_v4f64_v4f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvt_v4f64_v4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    mov x8, #2 // =0x2
 ; CHECK-NEXT:    fcvt z0.h, p0/m, z0.d
 ; CHECK-NEXT:    fcvt z1.h, p0/m, z1.d
@@ -467,8 +467,8 @@ define void @fcvt_v2f64_v2f32(<2 x double> %op1, ptr %b) {
 define void @fcvt_v4f64_v4f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvt_v4f64_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    mov x8, #2 // =0x2
 ; CHECK-NEXT:    fcvt z0.s, p0/m, z0.d
 ; CHECK-NEXT:    fcvt z1.s, p0/m, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-fma.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-fma.ll
index b5df97f767c13..478be9ab76dd9 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-fma.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-fma.ll
@@ -40,8 +40,8 @@ define <8 x half> @fma_v8f16(<8 x half> %op1, <8 x half> %op2, <8 x half> %op3)
 define void @fma_v16f16(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.h, p0/m, z2.h, z1.h
@@ -91,8 +91,8 @@ define <4 x float> @fma_v4f32(<4 x float> %op1, <4 x float> %op2, <4 x float> %o
 define void @fma_v8f32(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.s, p0/m, z2.s, z1.s
@@ -140,8 +140,8 @@ define <2 x double> @fma_v2f64(<2 x double> %op1, <2 x double> %op2, <2 x double
 define void @fma_v4f64(ptr %a, ptr %b, ptr %c) {
 ; CHECK-LABEL: fma_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q4, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q5, [x2]
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fmad z0.d, p0/m, z2.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-minmax.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-minmax.ll
index 07a67e2650290..4dc034adf459a 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-minmax.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-minmax.ll
@@ -37,8 +37,8 @@ define <8 x half> @fmaxnm_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fmaxnm_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmaxnm_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmaxnm z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -81,8 +81,8 @@ define <4 x float> @fmaxnm_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fmaxnm_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmaxnm_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmaxnm z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -123,8 +123,8 @@ define <2 x double> @fmaxnm_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fmaxnm_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmaxnm_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmaxnm z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -171,8 +171,8 @@ define <8 x half> @fminnm_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fminnm_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fminnm_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fminnm z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -215,8 +215,8 @@ define <4 x float> @fminnm_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fminnm_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fminnm_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fminnm z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -257,8 +257,8 @@ define <2 x double> @fminnm_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fminnm_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fminnm_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fminnm z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -305,8 +305,8 @@ define <8 x half> @fmax_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fmax_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmax_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -349,8 +349,8 @@ define <4 x float> @fmax_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fmax_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmax_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -391,8 +391,8 @@ define <2 x double> @fmax_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fmax_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmax_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -439,8 +439,8 @@ define <8 x half> @fmin_v8f16(<8 x half> %op1, <8 x half> %op2) {
 define void @fmin_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmin_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -483,8 +483,8 @@ define <4 x float> @fmin_v4f32(<4 x float> %op1, <4 x float> %op2) {
 define void @fmin_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmin_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -525,8 +525,8 @@ define <2 x double> @fmin_v2f64(<2 x double> %op1, <2 x double> %op2) {
 define void @fmin_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fmin_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fmin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-reduce.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-reduce.ll
index d2d771c48c204..bd10a0e091c0d 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-reduce.ll
@@ -211,8 +211,8 @@ define half @faddv_v8f16(half %start, <8 x half> %a) {
 define half @faddv_v16f16(half %start, ptr %a) {
 ; CHECK-LABEL: faddv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q2, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fadd z1.h, p0/m, z1.h, z2.h
 ; CHECK-NEXT:    faddv h1, p0, z1.h
 ; CHECK-NEXT:    fadd h0, h0, h1
@@ -249,8 +249,8 @@ define float @faddv_v4f32(float %start, <4 x float> %a) {
 define float @faddv_v8f32(float %start, ptr %a) {
 ; CHECK-LABEL: faddv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q2, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fadd z1.s, p0/m, z1.s, z2.s
 ; CHECK-NEXT:    faddv s1, p0, z1.s
 ; CHECK-NEXT:    fadd s0, s0, s1
@@ -285,8 +285,8 @@ define double @faddv_v2f64(double %start, <2 x double> %a) {
 define double @faddv_v4f64(double %start, ptr %a) {
 ; CHECK-LABEL: faddv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q2, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fadd z1.d, p0/m, z1.d, z2.d
 ; CHECK-NEXT:    faddv d1, p0, z1.d
 ; CHECK-NEXT:    fadd d0, d0, d1
@@ -327,8 +327,8 @@ define half @fmaxv_v8f16(<8 x half> %a) {
 define half @fmaxv_v16f16(ptr %a) {
 ; CHECK-LABEL: fmaxv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fmaxnm z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    fmaxnmv h0, p0, z0.h
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
@@ -365,8 +365,8 @@ define float @fmaxv_v4f32(<4 x float> %a) {
 define float @fmaxv_v8f32(ptr %a) {
 ; CHECK-LABEL: fmaxv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fmaxnm z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fmaxnmv s0, p0, z0.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $z0
@@ -401,8 +401,8 @@ define double @fmaxv_v2f64(<2 x double> %a) {
 define double @fmaxv_v4f64(ptr %a) {
 ; CHECK-LABEL: fmaxv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fmaxnm z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    fmaxnmv d0, p0, z0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -443,8 +443,8 @@ define half @fminv_v8f16(<8 x half> %a) {
 define half @fminv_v16f16(ptr %a) {
 ; CHECK-LABEL: fminv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fminnm z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    fminnmv h0, p0, z0.h
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
@@ -481,8 +481,8 @@ define float @fminv_v4f32(<4 x float> %a) {
 define float @fminv_v8f32(ptr %a) {
 ; CHECK-LABEL: fminv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fminnm z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fminnmv s0, p0, z0.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $z0
@@ -517,8 +517,8 @@ define double @fminv_v2f64(<2 x double> %a) {
 define double @fminv_v4f64(ptr %a) {
 ; CHECK-LABEL: fminv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fminnm z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    fminnmv d0, p0, z0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -559,8 +559,8 @@ define half @fmaximumv_v8f16(<8 x half> %a) {
 define half @fmaximumv_v16f16(ptr %a) {
 ; CHECK-LABEL: fmaximumv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fmax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    fmaxv h0, p0, z0.h
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
@@ -597,8 +597,8 @@ define float @fmaximumv_v4f32(<4 x float> %a) {
 define float @fmaximumv_v8f32(ptr %a) {
 ; CHECK-LABEL: fmaximumv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fmax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fmaxv s0, p0, z0.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $z0
@@ -633,8 +633,8 @@ define double @fmaximumv_v2f64(<2 x double> %a) {
 define double @fmaximumv_v4f64(ptr %a) {
 ; CHECK-LABEL: fmaximumv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fmax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    fmaxv d0, p0, z0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -675,8 +675,8 @@ define half @fminimumv_v8f16(<8 x half> %a) {
 define half @fminimumv_v16f16(ptr %a) {
 ; CHECK-LABEL: fminimumv_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fmin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    fminv h0, p0, z0.h
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
@@ -713,8 +713,8 @@ define float @fminimumv_v4f32(<4 x float> %a) {
 define float @fminimumv_v8f32(ptr %a) {
 ; CHECK-LABEL: fminimumv_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fmin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    fminv s0, p0, z0.s
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $z0
@@ -749,8 +749,8 @@ define double @fminimumv_v2f64(<2 x double> %a) {
 define double @fminimumv_v4f64(ptr %a) {
 ; CHECK-LABEL: fminimumv_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fmin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    fminv d0, p0, z0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-rounding.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-rounding.ll
index 580b43531070f..24832d807c649 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-rounding.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-rounding.ll
@@ -47,8 +47,8 @@ define <8 x half> @frintp_v8f16(<8 x half> %op) {
 define void @frintp_v16f16(ptr %a) {
 ; CHECK-LABEL: frintp_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frintp z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frintp z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -86,8 +86,8 @@ define <4 x float> @frintp_v4f32(<4 x float> %op) {
 define void @frintp_v8f32(ptr %a) {
 ; CHECK-LABEL: frintp_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frintp z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frintp z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -123,8 +123,8 @@ define <2 x double> @frintp_v2f64(<2 x double> %op) {
 define void @frintp_v4f64(ptr %a) {
 ; CHECK-LABEL: frintp_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frintp z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frintp z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -178,8 +178,8 @@ define <8 x half> @frintm_v8f16(<8 x half> %op) {
 define void @frintm_v16f16(ptr %a) {
 ; CHECK-LABEL: frintm_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frintm z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frintm z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -217,8 +217,8 @@ define <4 x float> @frintm_v4f32(<4 x float> %op) {
 define void @frintm_v8f32(ptr %a) {
 ; CHECK-LABEL: frintm_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frintm z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frintm z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -254,8 +254,8 @@ define <2 x double> @frintm_v2f64(<2 x double> %op) {
 define void @frintm_v4f64(ptr %a) {
 ; CHECK-LABEL: frintm_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frintm z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frintm z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -309,8 +309,8 @@ define <8 x half> @frinti_v8f16(<8 x half> %op) {
 define void @frinti_v16f16(ptr %a) {
 ; CHECK-LABEL: frinti_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frinti z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frinti z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -348,8 +348,8 @@ define <4 x float> @frinti_v4f32(<4 x float> %op) {
 define void @frinti_v8f32(ptr %a) {
 ; CHECK-LABEL: frinti_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frinti z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frinti z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -385,8 +385,8 @@ define <2 x double> @frinti_v2f64(<2 x double> %op) {
 define void @frinti_v4f64(ptr %a) {
 ; CHECK-LABEL: frinti_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frinti z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frinti z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -440,8 +440,8 @@ define <8 x half> @frintx_v8f16(<8 x half> %op) {
 define void @frintx_v16f16(ptr %a) {
 ; CHECK-LABEL: frintx_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frintx z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frintx z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -479,8 +479,8 @@ define <4 x float> @frintx_v4f32(<4 x float> %op) {
 define void @frintx_v8f32(ptr %a) {
 ; CHECK-LABEL: frintx_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frintx z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frintx z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -516,8 +516,8 @@ define <2 x double> @frintx_v2f64(<2 x double> %op) {
 define void @frintx_v4f64(ptr %a) {
 ; CHECK-LABEL: frintx_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frintx z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frintx z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -571,8 +571,8 @@ define <8 x half> @frinta_v8f16(<8 x half> %op) {
 define void @frinta_v16f16(ptr %a) {
 ; CHECK-LABEL: frinta_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frinta z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frinta z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -610,8 +610,8 @@ define <4 x float> @frinta_v4f32(<4 x float> %op) {
 define void @frinta_v8f32(ptr %a) {
 ; CHECK-LABEL: frinta_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frinta z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frinta z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -647,8 +647,8 @@ define <2 x double> @frinta_v2f64(<2 x double> %op) {
 define void @frinta_v4f64(ptr %a) {
 ; CHECK-LABEL: frinta_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frinta z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frinta z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -702,8 +702,8 @@ define <8 x half> @frintn_v8f16(<8 x half> %op) {
 define void @frintn_v16f16(ptr %a) {
 ; CHECK-LABEL: frintn_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frintn z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frintn z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -741,8 +741,8 @@ define <4 x float> @frintn_v4f32(<4 x float> %op) {
 define void @frintn_v8f32(ptr %a) {
 ; CHECK-LABEL: frintn_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frintn z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frintn z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -778,8 +778,8 @@ define <2 x double> @frintn_v2f64(<2 x double> %op) {
 define void @frintn_v4f64(ptr %a) {
 ; CHECK-LABEL: frintn_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frintn z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frintn z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -833,8 +833,8 @@ define <8 x half> @frintz_v8f16(<8 x half> %op) {
 define void @frintz_v16f16(ptr %a) {
 ; CHECK-LABEL: frintz_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    frintz z0.h, p0/m, z0.h
 ; CHECK-NEXT:    frintz z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -872,8 +872,8 @@ define <4 x float> @frintz_v4f32(<4 x float> %op) {
 define void @frintz_v8f32(ptr %a) {
 ; CHECK-LABEL: frintz_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    frintz z0.s, p0/m, z0.s
 ; CHECK-NEXT:    frintz z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -909,8 +909,8 @@ define <2 x double> @frintz_v2f64(<2 x double> %op) {
 define void @frintz_v4f64(ptr %a) {
 ; CHECK-LABEL: frintz_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    frintz z0.d, p0/m, z0.d
 ; CHECK-NEXT:    frintz z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-select.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-select.ll
index 73fd7e1465343..132225546fc4f 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-select.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-select.ll
@@ -7,8 +7,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define <2 x half> @select_v2f16(<2 x half> %op1, <2 x half> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -23,8 +23,8 @@ define <2 x half> @select_v2f16(<2 x half> %op1, <2 x half> %op2, i1 %mask) {
 define <4 x half> @select_v4f16(<4 x half> %op1, <4 x half> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -39,8 +39,8 @@ define <4 x half> @select_v4f16(<4 x half> %op1, <4 x half> %op2, i1 %mask) {
 define <8 x half> @select_v8f16(<8 x half> %op1, <8 x half> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -55,8 +55,8 @@ define <8 x half> @select_v8f16(<8 x half> %op1, <8 x half> %op2, i1 %mask) {
 define void @select_v16f16(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z0.h, w2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
 ; CHECK-NEXT:    ldr q0, [x0]
@@ -77,8 +77,8 @@ define void @select_v16f16(ptr %a, ptr %b, i1 %mask) {
 define <2 x float> @select_v2f32(<2 x float> %op1, <2 x float> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    mov z2.s, w8
@@ -93,8 +93,8 @@ define <2 x float> @select_v2f32(<2 x float> %op1, <2 x float> %op2, i1 %mask) {
 define <4 x float> @select_v4f32(<4 x float> %op1, <4 x float> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov z2.s, w8
@@ -109,8 +109,8 @@ define <4 x float> @select_v4f32(<4 x float> %op1, <4 x float> %op2, i1 %mask) {
 define void @select_v8f32(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w2, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    ldr q0, [x0]
@@ -150,9 +150,9 @@ define <1 x double> @select_v1f64(<1 x double> %op1, <1 x double> %op2, i1 %mask
 define <2 x double> @select_v2f64(<2 x double> %op1, <2 x double> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov z2.d, x8
@@ -167,9 +167,9 @@ define <2 x double> @select_v2f64(<2 x double> %op1, <2 x double> %op2, i1 %mask
 define void @select_v4f64(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ldr q0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-to-int.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-to-int.ll
index d6adf9cf0ad67..58eae212d7999 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-to-int.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-to-int.ll
@@ -36,8 +36,8 @@ define void @fcvtzu_v8f16_v8i16(ptr %a, ptr %b) {
 define void @fcvtzu_v16f16_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzu_v16f16_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fcvtzu z0.h, p0/m, z0.h
 ; CHECK-NEXT:    fcvtzu z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -55,8 +55,8 @@ define void @fcvtzu_v16f16_v16i16(ptr %a, ptr %b) {
 define <2 x i32> @fcvtzu_v2f16_v2i32(<2 x half> %op1) {
 ; CHECK-LABEL: fcvtzu_v2f16_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -68,8 +68,8 @@ define <2 x i32> @fcvtzu_v2f16_v2i32(<2 x half> %op1) {
 define <4 x i32> @fcvtzu_v4f16_v4i32(<4 x half> %op1) {
 ; CHECK-LABEL: fcvtzu_v4f16_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -302,8 +302,8 @@ define <4 x i16> @fcvtzu_v4f32_v4i16(<4 x float> %op1) {
 define <8 x i16> @fcvtzu_v8f32_v8i16(ptr %a) {
 ; CHECK-LABEL: fcvtzu_v8f32_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcvtzu z1.s, p0/m, z1.s
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
@@ -320,8 +320,8 @@ define <8 x i16> @fcvtzu_v8f32_v8i16(ptr %a) {
 define void @fcvtzu_v16f32_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzu_v16f32_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fcvtzu z1.s, p0/m, z1.s
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.s
@@ -373,8 +373,8 @@ define <4 x i32> @fcvtzu_v4f32_v4i32(<4 x float> %op1) {
 define void @fcvtzu_v8f32_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzu_v8f32_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcvtzu z0.s, p0/m, z0.s
 ; CHECK-NEXT:    fcvtzu z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -392,8 +392,8 @@ define void @fcvtzu_v8f32_v8i32(ptr %a, ptr %b) {
 define <1 x i64> @fcvtzu_v1f32_v1i64(<1 x float> %op1) {
 ; CHECK-LABEL: fcvtzu_v1f32_v1i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -405,8 +405,8 @@ define <1 x i64> @fcvtzu_v1f32_v1i64(<1 x float> %op1) {
 define <2 x i64> @fcvtzu_v2f32_v2i64(<2 x float> %op1) {
 ; CHECK-LABEL: fcvtzu_v2f32_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -491,8 +491,8 @@ define <4 x i16> @fcvtzu_v4f64_v4i16(ptr %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
@@ -520,8 +520,8 @@ define <8 x i16> @fcvtzu_v8f64_v8i16(ptr %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q3, q2, [x0]
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
@@ -563,24 +563,22 @@ define void @fcvtzu_v16f64_v16i16(ptr %a, ptr %b) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q3, q2, [x0]
-; CHECK-NEXT:    ldr q6, [x0, #112]
-; CHECK-NEXT:    ldp q4, q5, [x0, #80]
-; CHECK-NEXT:    ldr q7, [x0, #64]
+; CHECK-NEXT:    ldp q4, q5, [x0, #96]
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z2.d
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
+; CHECK-NEXT:    fcvtzs z2.d, p0/m, z2.d
+; CHECK-NEXT:    ldp q6, q7, [x0, #64]
 ; CHECK-NEXT:    fcvtzs z3.d, p0/m, z3.d
-; CHECK-NEXT:    fcvtzs z6.d, p0/m, z6.d
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z5.d
 ; CHECK-NEXT:    fcvtzs z4.d, p0/m, z4.d
 ; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
-; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK-NEXT:    fcvtzs z6.d, p0/m, z6.d
+; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    uzp1 z3.s, z3.s, z3.s
-; CHECK-NEXT:    uzp1 z6.s, z6.s, z6.s
 ; CHECK-NEXT:    uzp1 z5.s, z5.s, z5.s
 ; CHECK-NEXT:    fmov w8, s1
 ; CHECK-NEXT:    mov z16.s, z1.s[1]
@@ -606,25 +604,26 @@ define void @fcvtzu_v16f64_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov z3.s, z5.s[1]
 ; CHECK-NEXT:    strh w8, [sp, #6]
 ; CHECK-NEXT:    fmov w8, s2
-; CHECK-NEXT:    mov z2.s, z6.s[1]
+; CHECK-NEXT:    uzp1 z2.s, z6.s, z6.s
 ; CHECK-NEXT:    strh w8, [sp, #2]
-; CHECK-NEXT:    fmov w8, s6
-; CHECK-NEXT:    strh w8, [sp, #28]
 ; CHECK-NEXT:    fmov w8, s5
-; CHECK-NEXT:    strh w8, [sp, #24]
+; CHECK-NEXT:    strh w8, [sp, #28]
 ; CHECK-NEXT:    fmov w8, s1
 ; CHECK-NEXT:    mov z1.s, z1.s[1]
-; CHECK-NEXT:    strh w8, [sp, #20]
+; CHECK-NEXT:    strh w8, [sp, #24]
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    mov z0.s, z0.s[1]
-; CHECK-NEXT:    strh w8, [sp, #16]
+; CHECK-NEXT:    strh w8, [sp, #20]
 ; CHECK-NEXT:    fmov w8, s2
-; CHECK-NEXT:    strh w8, [sp, #30]
+; CHECK-NEXT:    mov z2.s, z2.s[1]
+; CHECK-NEXT:    strh w8, [sp, #16]
 ; CHECK-NEXT:    fmov w8, s3
-; CHECK-NEXT:    strh w8, [sp, #26]
+; CHECK-NEXT:    strh w8, [sp, #30]
 ; CHECK-NEXT:    fmov w8, s1
-; CHECK-NEXT:    strh w8, [sp, #22]
+; CHECK-NEXT:    strh w8, [sp, #26]
 ; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    strh w8, [sp, #22]
+; CHECK-NEXT:    fmov w8, s2
 ; CHECK-NEXT:    strh w8, [sp, #18]
 ; CHECK-NEXT:    ldp q1, q0, [sp]
 ; CHECK-NEXT:    stp q1, q0, [x1]
@@ -669,8 +668,8 @@ define <2 x i32> @fcvtzu_v2f64_v2i32(<2 x double> %op1) {
 define <4 x i32> @fcvtzu_v4f64_v4i32(ptr %a) {
 ; CHECK-LABEL: fcvtzu_v4f64_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzu z1.d, p0/m, z1.d
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
@@ -687,8 +686,8 @@ define <4 x i32> @fcvtzu_v4f64_v4i32(ptr %a) {
 define void @fcvtzu_v8f64_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzu_v8f64_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fcvtzu z1.d, p0/m, z1.d
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.d
@@ -740,8 +739,8 @@ define <2 x i64> @fcvtzu_v2f64_v2i64(<2 x double> %op1) {
 define void @fcvtzu_v4f64_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzu_v4f64_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzu z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzu z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -785,8 +784,8 @@ define void @fcvtzs_v8f16_v8i16(ptr %a, ptr %b) {
 define void @fcvtzs_v16f16_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzs_v16f16_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    fcvtzs z0.h, p0/m, z0.h
 ; CHECK-NEXT:    fcvtzs z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -804,8 +803,8 @@ define void @fcvtzs_v16f16_v16i16(ptr %a, ptr %b) {
 define <2 x i32> @fcvtzs_v2f16_v2i32(<2 x half> %op1) {
 ; CHECK-LABEL: fcvtzs_v2f16_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -817,8 +816,8 @@ define <2 x i32> @fcvtzs_v2f16_v2i32(<2 x half> %op1) {
 define <4 x i32> @fcvtzs_v4f16_v4i32(<4 x half> %op1) {
 ; CHECK-LABEL: fcvtzs_v4f16_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -1052,8 +1051,8 @@ define <4 x i16> @fcvtzs_v4f32_v4i16(<4 x float> %op1) {
 define <8 x i16> @fcvtzs_v8f32_v8i16(ptr %a) {
 ; CHECK-LABEL: fcvtzs_v8f32_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcvtzs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
@@ -1070,8 +1069,8 @@ define <8 x i16> @fcvtzs_v8f32_v8i16(ptr %a) {
 define void @fcvtzs_v16f32_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzs_v16f32_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fcvtzs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.s
@@ -1123,8 +1122,8 @@ define <4 x i32> @fcvtzs_v4f32_v4i32(<4 x float> %op1) {
 define void @fcvtzs_v8f32_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzs_v8f32_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fcvtzs z0.s, p0/m, z0.s
 ; CHECK-NEXT:    fcvtzs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -1142,8 +1141,8 @@ define void @fcvtzs_v8f32_v8i32(ptr %a, ptr %b) {
 define <1 x i64> @fcvtzs_v1f32_v1i64(<1 x float> %op1) {
 ; CHECK-LABEL: fcvtzs_v1f32_v1i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -1155,8 +1154,8 @@ define <1 x i64> @fcvtzs_v1f32_v1i64(<1 x float> %op1) {
 define <2 x i64> @fcvtzs_v2f32_v2i64(<2 x float> %op1) {
 ; CHECK-LABEL: fcvtzs_v2f32_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -1243,8 +1242,8 @@ define <4 x i16> @fcvtzs_v4f64_v4i16(ptr %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
@@ -1272,8 +1271,8 @@ define <8 x i16> @fcvtzs_v8f64_v8i16(ptr %a) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #16
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q3, q2, [x0]
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
@@ -1315,24 +1314,22 @@ define void @fcvtzs_v16f64_v16i16(ptr %a, ptr %b) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q3, q2, [x0]
-; CHECK-NEXT:    ldr q6, [x0, #112]
-; CHECK-NEXT:    ldp q4, q5, [x0, #80]
-; CHECK-NEXT:    ldr q7, [x0, #64]
+; CHECK-NEXT:    ldp q4, q5, [x0, #96]
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
-; CHECK-NEXT:    fcvtzs z2.d, p0/m, z2.d
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
+; CHECK-NEXT:    fcvtzs z2.d, p0/m, z2.d
+; CHECK-NEXT:    ldp q6, q7, [x0, #64]
 ; CHECK-NEXT:    fcvtzs z3.d, p0/m, z3.d
-; CHECK-NEXT:    fcvtzs z6.d, p0/m, z6.d
 ; CHECK-NEXT:    fcvtzs z5.d, p0/m, z5.d
 ; CHECK-NEXT:    fcvtzs z4.d, p0/m, z4.d
 ; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
-; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK-NEXT:    fcvtzs z6.d, p0/m, z6.d
+; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    uzp1 z3.s, z3.s, z3.s
-; CHECK-NEXT:    uzp1 z6.s, z6.s, z6.s
 ; CHECK-NEXT:    uzp1 z5.s, z5.s, z5.s
 ; CHECK-NEXT:    fmov w8, s1
 ; CHECK-NEXT:    mov z16.s, z1.s[1]
@@ -1358,25 +1355,26 @@ define void @fcvtzs_v16f64_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov z3.s, z5.s[1]
 ; CHECK-NEXT:    strh w8, [sp, #6]
 ; CHECK-NEXT:    fmov w8, s2
-; CHECK-NEXT:    mov z2.s, z6.s[1]
+; CHECK-NEXT:    uzp1 z2.s, z6.s, z6.s
 ; CHECK-NEXT:    strh w8, [sp, #2]
-; CHECK-NEXT:    fmov w8, s6
-; CHECK-NEXT:    strh w8, [sp, #28]
 ; CHECK-NEXT:    fmov w8, s5
-; CHECK-NEXT:    strh w8, [sp, #24]
+; CHECK-NEXT:    strh w8, [sp, #28]
 ; CHECK-NEXT:    fmov w8, s1
 ; CHECK-NEXT:    mov z1.s, z1.s[1]
-; CHECK-NEXT:    strh w8, [sp, #20]
+; CHECK-NEXT:    strh w8, [sp, #24]
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    mov z0.s, z0.s[1]
-; CHECK-NEXT:    strh w8, [sp, #16]
+; CHECK-NEXT:    strh w8, [sp, #20]
 ; CHECK-NEXT:    fmov w8, s2
-; CHECK-NEXT:    strh w8, [sp, #30]
+; CHECK-NEXT:    mov z2.s, z2.s[1]
+; CHECK-NEXT:    strh w8, [sp, #16]
 ; CHECK-NEXT:    fmov w8, s3
-; CHECK-NEXT:    strh w8, [sp, #26]
+; CHECK-NEXT:    strh w8, [sp, #30]
 ; CHECK-NEXT:    fmov w8, s1
-; CHECK-NEXT:    strh w8, [sp, #22]
+; CHECK-NEXT:    strh w8, [sp, #26]
 ; CHECK-NEXT:    fmov w8, s0
+; CHECK-NEXT:    strh w8, [sp, #22]
+; CHECK-NEXT:    fmov w8, s2
 ; CHECK-NEXT:    strh w8, [sp, #18]
 ; CHECK-NEXT:    ldp q1, q0, [sp]
 ; CHECK-NEXT:    stp q1, q0, [x1]
@@ -1421,8 +1419,8 @@ define <2 x i32> @fcvtzs_v2f64_v2i32(<2 x double> %op1) {
 define <4 x i32> @fcvtzs_v4f64_v4i32(ptr %a) {
 ; CHECK-LABEL: fcvtzs_v4f64_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
@@ -1439,8 +1437,8 @@ define <4 x i32> @fcvtzs_v4f64_v4i32(ptr %a) {
 define void @fcvtzs_v8f64_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzs_v8f64_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
@@ -1492,8 +1490,8 @@ define <2 x i64> @fcvtzs_v2f64_v2i64(<2 x double> %op1) {
 define void @fcvtzs_v4f64_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fcvtzs_v4f64_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    fcvtzs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    fcvtzs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x1]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-vselect.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-vselect.ll
index ee8704284def5..4c5a6fe2fd231 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-vselect.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-fp-vselect.ll
@@ -71,8 +71,8 @@ define <8 x half> @select_v8f16(<8 x half> %op1, <8 x half> %op2, <8 x i1> %mask
 define void @select_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    fcmeq p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    fcmeq p0.h, p0/z, z2.h, z3.h
@@ -128,8 +128,8 @@ define <4 x float> @select_v4f32(<4 x float> %op1, <4 x float> %op2, <4 x i1> %m
 define void @select_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    fcmeq p1.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    fcmeq p0.s, p0/z, z2.s, z3.s
@@ -186,8 +186,8 @@ define <2 x double> @select_v2f64(<2 x double> %op1, <2 x double> %op2, <2 x i1>
 define void @select_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    fcmeq p1.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    fcmeq p0.d, p0/z, z2.d, z3.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-insert-vector-elt.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-insert-vector-elt.ll
index 0b3e7695e6a0a..4aa965777c742 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-insert-vector-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-insert-vector-elt.ll
@@ -11,9 +11,9 @@ target triple = "aarch64-unknown-linux-gnu"
 define <4 x i8> @insertelement_v4i8(<4 x i8> %op1) {
 ; CHECK-LABEL: insertelement_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -28,9 +28,9 @@ define <4 x i8> @insertelement_v4i8(<4 x i8> %op1) {
 define <8 x i8> @insertelement_v8i8(<8 x i8> %op1) {
 ; CHECK-LABEL: insertelement_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z1.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -45,9 +45,9 @@ define <8 x i8> @insertelement_v8i8(<8 x i8> %op1) {
 define <16 x i8> @insertelement_v16i8(<16 x i8> %op1) {
 ; CHECK-LABEL: insertelement_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #15 // =0xf
 ; CHECK-NEXT:    index z1.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -62,9 +62,9 @@ define <16 x i8> @insertelement_v16i8(<16 x i8> %op1) {
 define <32 x i8> @insertelement_v32i8(<32 x i8> %op1) {
 ; CHECK-LABEL: insertelement_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov w8, #15 // =0xf
 ; CHECK-NEXT:    index z2.b, #0, #1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z3.b, w8
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -80,9 +80,9 @@ define <32 x i8> @insertelement_v32i8(<32 x i8> %op1) {
 define <2 x i16> @insertelement_v2i16(<2 x i16> %op1) {
 ; CHECK-LABEL: insertelement_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -97,9 +97,9 @@ define <2 x i16> @insertelement_v2i16(<2 x i16> %op1) {
 define <4 x i16> @insertelement_v4i16(<4 x i16> %op1) {
 ; CHECK-LABEL: insertelement_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -114,9 +114,9 @@ define <4 x i16> @insertelement_v4i16(<4 x i16> %op1) {
 define <8 x i16> @insertelement_v8i16(<8 x i16> %op1) {
 ; CHECK-LABEL: insertelement_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -131,9 +131,9 @@ define <8 x i16> @insertelement_v8i16(<8 x i16> %op1) {
 define <16 x i16> @insertelement_v16i16(<16 x i16> %op1) {
 ; CHECK-LABEL: insertelement_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z2.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z3.h, w8
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -149,9 +149,9 @@ define <16 x i16> @insertelement_v16i16(<16 x i16> %op1) {
 define <2 x i32> @insertelement_v2i32(<2 x i32> %op1) {
 ; CHECK-LABEL: insertelement_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -166,9 +166,9 @@ define <2 x i32> @insertelement_v2i32(<2 x i32> %op1) {
 define <4 x i32> @insertelement_v4i32(<4 x i32> %op1) {
 ; CHECK-LABEL: insertelement_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -183,9 +183,9 @@ define <4 x i32> @insertelement_v4i32(<4 x i32> %op1) {
 define <8 x i32> @insertelement_v8i32(ptr %a) {
 ; CHECK-LABEL: insertelement_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    mov w8, #5 // =0x5
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
@@ -212,9 +212,9 @@ define <1 x i64> @insertelement_v1i64(<1 x i64> %op1) {
 define <2 x i64> @insertelement_v2i64(<2 x i64> %op1) {
 ; CHECK-LABEL: insertelement_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    mov w8, #5 // =0x5
@@ -229,9 +229,9 @@ define <2 x i64> @insertelement_v2i64(<2 x i64> %op1) {
 define <4 x i64> @insertelement_v4i64(ptr %a) {
 ; CHECK-LABEL: insertelement_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    mov w8, #5 // =0x5
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
@@ -264,9 +264,9 @@ define <2 x half> @insertelement_v2f16(<2 x half> %op1) {
 define <4 x half> @insertelement_v4f16(<4 x half> %op1) {
 ; CHECK-LABEL: insertelement_v4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
@@ -281,9 +281,9 @@ define <4 x half> @insertelement_v4f16(<4 x half> %op1) {
 define <8 x half> @insertelement_v8f16(<8 x half> %op1) {
 ; CHECK-LABEL: insertelement_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z1.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z1.h, z2.h
@@ -298,9 +298,9 @@ define <8 x half> @insertelement_v8f16(<8 x half> %op1) {
 define <16 x half> @insertelement_v16f16(ptr %a) {
 ; CHECK-LABEL: insertelement_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov w8, #7 // =0x7
 ; CHECK-NEXT:    index z0.h, #0, #1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z1.h, w8
 ; CHECK-NEXT:    fmov h2, #5.00000000
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z0.h, z1.h
@@ -317,9 +317,9 @@ define <16 x half> @insertelement_v16f16(ptr %a) {
 define <2 x float> @insertelement_v2f32(<2 x float> %op1) {
 ; CHECK-LABEL: insertelement_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
@@ -334,9 +334,9 @@ define <2 x float> @insertelement_v2f32(<2 x float> %op1) {
 define <4 x float> @insertelement_v4f32(<4 x float> %op1) {
 ; CHECK-LABEL: insertelement_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z1.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z2.s, w8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z1.s, z2.s
@@ -351,9 +351,9 @@ define <4 x float> @insertelement_v4f32(<4 x float> %op1) {
 define <8 x float> @insertelement_v8f32(ptr %a) {
 ; CHECK-LABEL: insertelement_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov w8, #3 // =0x3
 ; CHECK-NEXT:    index z0.s, #0, #1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z1.s, w8
 ; CHECK-NEXT:    fmov s2, #5.00000000
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z0.s, z1.s
@@ -379,9 +379,9 @@ define <1 x double> @insertelement_v1f64(<1 x double> %op1) {
 define <2 x double> @insertelement_v2f64(<2 x double> %op1) {
 ; CHECK-LABEL: insertelement_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z1.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z2.d, x8
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z1.d, z2.d
@@ -396,9 +396,9 @@ define <2 x double> @insertelement_v2f64(<2 x double> %op1) {
 define <4 x double> @insertelement_v4f64(ptr %a) {
 ; CHECK-LABEL: insertelement_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov w8, #1 // =0x1
 ; CHECK-NEXT:    index z0.d, #0, #1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.d, x8
 ; CHECK-NEXT:    fmov d2, #5.00000000
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z0.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-arith.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-arith.ll
index e3c4b6f1cb53f..8baa87c6d686d 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-arith.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-arith.ll
@@ -262,8 +262,8 @@ define <16 x i8> @mul_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @mul_v32i8(ptr %a, ptr %b) {
 ; SVE-LABEL: mul_v32i8:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    mul z0.b, p0/m, z0.b, z1.b
 ; SVE-NEXT:    movprfx z1, z2
@@ -352,8 +352,8 @@ define <8 x i16> @mul_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @mul_v16i16(ptr %a, ptr %b) {
 ; SVE-LABEL: mul_v16i16:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; SVE-NEXT:    movprfx z1, z2
@@ -421,8 +421,8 @@ define <4 x i32> @mul_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @mul_v8i32(ptr %a, ptr %b) {
 ; SVE-LABEL: mul_v8i32:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; SVE-NEXT:    movprfx z1, z2
@@ -490,8 +490,8 @@ define <2 x i64> @mul_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @mul_v4i64(ptr %a, ptr %b) {
 ; SVE-LABEL: mul_v4i64:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; SVE-NEXT:    movprfx z1, z2
@@ -746,8 +746,8 @@ define <16 x i8> @abs_v16i8(<16 x i8> %op1) {
 define void @abs_v32i8(ptr %a) {
 ; CHECK-LABEL: abs_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    abs z0.b, p0/m, z0.b
 ; CHECK-NEXT:    abs z1.b, p0/m, z1.b
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -798,8 +798,8 @@ define <8 x i16> @abs_v8i16(<8 x i16> %op1) {
 define void @abs_v16i16(ptr %a) {
 ; CHECK-LABEL: abs_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    abs z0.h, p0/m, z0.h
 ; CHECK-NEXT:    abs z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -837,8 +837,8 @@ define <4 x i32> @abs_v4i32(<4 x i32> %op1) {
 define void @abs_v8i32(ptr %a) {
 ; CHECK-LABEL: abs_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    abs z0.s, p0/m, z0.s
 ; CHECK-NEXT:    abs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -876,8 +876,8 @@ define <2 x i64> @abs_v2i64(<2 x i64> %op1) {
 define void @abs_v4i64(ptr %a) {
 ; CHECK-LABEL: abs_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    abs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    abs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-compares.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-compares.ll
index 6200e44218a96..73c1eac99dd30 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-compares.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-compares.ll
@@ -41,8 +41,8 @@ define <16 x i8> @icmp_eq_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @icmp_eq_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_eq_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpeq p1.b, p0/z, z1.b, z0.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z2.b, z3.b
@@ -91,8 +91,8 @@ define <8 x i16> @icmp_eq_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @icmp_eq_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_eq_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpeq p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z2.h, z3.h
@@ -141,8 +141,8 @@ define <4 x i32> @icmp_eq_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @icmp_eq_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_eq_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpeq p1.s, p0/z, z1.s, z0.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z2.s, z3.s
@@ -191,8 +191,8 @@ define <2 x i64> @icmp_eq_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @icmp_eq_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_eq_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpeq p1.d, p0/z, z1.d, z0.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z2.d, z3.d
@@ -215,8 +215,8 @@ define void @icmp_eq_v4i64(ptr %a, ptr %b) {
 define void @icmp_ne_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_ne_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpne p1.b, p0/z, z1.b, z0.b
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z2.b, z3.b
@@ -261,8 +261,8 @@ define void @icmp_sge_v8i16(ptr %a, ptr %b) {
 define void @icmp_sgt_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_sgt_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpgt p1.h, p0/z, z1.h, z0.h
 ; CHECK-NEXT:    cmpgt p0.h, p0/z, z2.h, z3.h
@@ -307,8 +307,8 @@ define void @icmp_sle_v4i32(ptr %a, ptr %b) {
 define void @icmp_slt_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: icmp_slt_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    cmpgt p1.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    cmpgt p0.s, p0/z, z3.s, z2.s
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
index fcf4f21c6ea84..5158dda37a8b9 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-div.ll
@@ -86,8 +86,8 @@ define <16 x i8> @sdiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z5.s
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z1.h, z4.h, z4.h
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z1.h, p0, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z1.b, z1.b, z1.b
@@ -163,18 +163,18 @@ define void @sdiv_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    sdiv z6.s, p0/m, z6.s, z7.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z7.h, z16.h, z16.h
 ; CHECK-NEXT:    splice z4.h, p0, z4.h, z5.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
+; CHECK-NEXT:    uzp1 z7.h, z16.h, z16.h
 ; CHECK-NEXT:    uzp1 z1.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z7.h, p0, z7.h, z6.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    uzp1 z3.b, z7.b, z7.b
 ; CHECK-NEXT:    splice z2.b, p0, z2.b, z0.b
+; CHECK-NEXT:    uzp1 z3.b, z7.b, z7.b
 ; CHECK-NEXT:    splice z3.b, p0, z3.b, z1.b
 ; CHECK-NEXT:    stp q3, q2, [x0]
 ; CHECK-NEXT:    ret
@@ -203,9 +203,9 @@ define <2 x i16> @sdiv_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
 define <4 x i16> @sdiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
 ; CHECK-LABEL: sdiv_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
@@ -272,8 +272,8 @@ define void @sdiv_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    sdiv z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z1.h, z5.h, z5.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z2.h
+; CHECK-NEXT:    uzp1 z1.h, z5.h, z5.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z3.h, p0, z3.h, z1.h
 ; CHECK-NEXT:    stp q3, q0, [x0]
@@ -314,8 +314,8 @@ define <4 x i32> @sdiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @sdiv_v8i32(ptr %a, ptr %b)  {
 ; CHECK-LABEL: sdiv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    sdivr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -358,8 +358,8 @@ define <2 x i64> @sdiv_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @sdiv_v4i64(ptr %a, ptr %b)  {
 ; CHECK-LABEL: sdiv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    sdivr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -453,8 +453,8 @@ define <16 x i8> @udiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z5.s
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z1.h, z4.h, z4.h
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z1.h, p0, z1.h, z2.h
 ; CHECK-NEXT:    uzp1 z1.b, z1.b, z1.b
@@ -530,18 +530,18 @@ define void @udiv_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
 ; CHECK-NEXT:    udiv z6.s, p0/m, z6.s, z7.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z7.h, z16.h, z16.h
 ; CHECK-NEXT:    splice z4.h, p0, z4.h, z5.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
+; CHECK-NEXT:    uzp1 z7.h, z16.h, z16.h
 ; CHECK-NEXT:    uzp1 z1.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z7.h, p0, z7.h, z6.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    uzp1 z3.b, z7.b, z7.b
 ; CHECK-NEXT:    splice z2.b, p0, z2.b, z0.b
+; CHECK-NEXT:    uzp1 z3.b, z7.b, z7.b
 ; CHECK-NEXT:    splice z3.b, p0, z3.b, z1.b
 ; CHECK-NEXT:    stp q3, q2, [x0]
 ; CHECK-NEXT:    ret
@@ -555,9 +555,9 @@ define void @udiv_v32i8(ptr %a, ptr %b) {
 define <2 x i16> @udiv_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
 ; CHECK-LABEL: udiv_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
@@ -570,9 +570,9 @@ define <2 x i16> @udiv_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
 define <4 x i16> @udiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
 ; CHECK-LABEL: udiv_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z1.s, z1.h
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    udiv z0.s, p0/m, z0.s, z1.s
@@ -639,8 +639,8 @@ define void @udiv_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    udiv z3.s, p0/m, z3.s, z4.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z1.h, z5.h, z5.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z2.h
+; CHECK-NEXT:    uzp1 z1.h, z5.h, z5.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z3.h, p0, z3.h, z1.h
 ; CHECK-NEXT:    stp q3, q0, [x0]
@@ -681,8 +681,8 @@ define <4 x i32> @udiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @udiv_v8i32(ptr %a, ptr %b)  {
 ; CHECK-LABEL: udiv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    udivr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -725,8 +725,8 @@ define <2 x i64> @udiv_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @udiv_v4i64(ptr %a, ptr %b)  {
 ; CHECK-LABEL: udiv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    udivr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -743,10 +743,10 @@ define void @udiv_v4i64(ptr %a, ptr %b)  {
 define void @udiv_constantsplat_v8i32(ptr %a)  {
 ; SVE-LABEL: udiv_constantsplat_v8i32:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    mov w8, #8969 // =0x2309
-; SVE-NEXT:    movk w8, #22765, lsl #16
 ; SVE-NEXT:    ldp q1, q2, [x0]
+; SVE-NEXT:    movk w8, #22765, lsl #16
+; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    mov z0.s, w8
 ; SVE-NEXT:    movprfx z3, z1
 ; SVE-NEXT:    umulh z3.s, p0/m, z3.s, z0.s
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-immediates.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-immediates.ll
index 0785c67ce6f41..f028b3eeca257 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-immediates.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-immediates.ll
@@ -221,8 +221,8 @@ define void @ashr_v4i64(ptr %a) {
 define void @icmp_eq_v32i8(ptr %a) {
 ; CHECK-LABEL: icmp_eq_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    cmpeq p1.b, p0/z, z0.b, #7
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z1.b, #7
 ; CHECK-NEXT:    mov z0.b, p1/z, #-1 // =0xffffffffffffffff
@@ -241,8 +241,8 @@ define void @icmp_eq_v32i8(ptr %a) {
 define void @icmp_sge_v16i16(ptr %a) {
 ; CHECK-LABEL: icmp_sge_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    cmpge p1.h, p0/z, z0.h, #15
 ; CHECK-NEXT:    cmpge p0.h, p0/z, z1.h, #15
 ; CHECK-NEXT:    mov z0.h, p1/z, #-1 // =0xffffffffffffffff
@@ -261,8 +261,8 @@ define void @icmp_sge_v16i16(ptr %a) {
 define void @icmp_sgt_v8i32(ptr %a) {
 ; CHECK-LABEL: icmp_sgt_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    cmpgt p1.s, p0/z, z0.s, #-8
 ; CHECK-NEXT:    cmpgt p0.s, p0/z, z1.s, #-8
 ; CHECK-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
@@ -281,8 +281,8 @@ define void @icmp_sgt_v8i32(ptr %a) {
 define void @icmp_ult_v4i64(ptr %a) {
 ; CHECK-LABEL: icmp_ult_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    cmplo p1.d, p0/z, z0.d, #63
 ; CHECK-NEXT:    cmplo p0.d, p0/z, z1.d, #63
 ; CHECK-NEXT:    mov z0.d, p1/z, #-1 // =0xffffffffffffffff
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-minmax.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-minmax.ll
index d7600c6e6192d..50cf9b73d9a79 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-minmax.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-minmax.ll
@@ -37,8 +37,8 @@ define <16 x i8> @smax_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @smax_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: smax_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smax z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -81,8 +81,8 @@ define <8 x i16> @smax_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @smax_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: smax_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -125,8 +125,8 @@ define <4 x i32> @smax_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @smax_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: smax_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -171,8 +171,8 @@ define <2 x i64> @smax_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @smax_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: smax_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -219,8 +219,8 @@ define <16 x i8> @smin_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @smin_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: smin_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smin z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -263,8 +263,8 @@ define <8 x i16> @smin_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @smin_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: smin_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -307,8 +307,8 @@ define <4 x i32> @smin_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @smin_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: smin_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -353,8 +353,8 @@ define <2 x i64> @smin_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @smin_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: smin_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -401,8 +401,8 @@ define <16 x i8> @umax_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @umax_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: umax_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umax z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -445,8 +445,8 @@ define <8 x i16> @umax_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @umax_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: umax_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -489,8 +489,8 @@ define <4 x i32> @umax_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @umax_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: umax_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -535,8 +535,8 @@ define <2 x i64> @umax_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @umax_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: umax_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -583,8 +583,8 @@ define <16 x i8> @umin_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @umin_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: umin_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umin z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -627,8 +627,8 @@ define <8 x i16> @umin_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @umin_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: umin_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -671,8 +671,8 @@ define <4 x i32> @umin_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @umin_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: umin_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -717,8 +717,8 @@ define <2 x i64> @umin_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @umin_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: umin_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-mulh.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-mulh.ll
index c48cb315a7aa3..cb7fa53eac513 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-mulh.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-mulh.ll
@@ -101,8 +101,8 @@ define <16 x i8> @smulh_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @smulh_v32i8(ptr %a, ptr %b) {
 ; SVE-LABEL: smulh_v32i8:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    smulh z0.b, p0/m, z0.b, z1.b
 ; SVE-NEXT:    movprfx z1, z2
@@ -214,8 +214,8 @@ define <8 x i16> @smulh_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @smulh_v16i16(ptr %a, ptr %b) {
 ; SVE-LABEL: smulh_v16i16:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    smulh z0.h, p0/m, z0.h, z1.h
 ; SVE-NEXT:    movprfx z1, z2
@@ -295,8 +295,8 @@ define <4 x i32> @smulh_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @smulh_v8i32(ptr %a, ptr %b) {
 ; SVE-LABEL: smulh_v8i32:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    smulh z0.s, p0/m, z0.s, z1.s
 ; SVE-NEXT:    movprfx z1, z2
@@ -378,8 +378,8 @@ define <2 x i64> @smulh_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @smulh_v4i64(ptr %a, ptr %b) {
 ; SVE-LABEL: smulh_v4i64:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    smulh z0.d, p0/m, z0.d, z1.d
 ; SVE-NEXT:    movprfx z1, z2
@@ -413,9 +413,9 @@ define void @smulh_v4i64(ptr %a, ptr %b) {
 define <4 x i8> @umulh_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
 ; SVE-LABEL: umulh_v4i8:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.h, vl4
 ; SVE-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; SVE-NEXT:    // kill: def $d0 killed $d0 def $z0
+; SVE-NEXT:    ptrue p0.h, vl4
 ; SVE-NEXT:    and z0.h, z0.h, #0xff
 ; SVE-NEXT:    and z1.h, z1.h, #0xff
 ; SVE-NEXT:    mul z0.h, p0/m, z0.h, z1.h
@@ -494,8 +494,8 @@ define <16 x i8> @umulh_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @umulh_v32i8(ptr %a, ptr %b) {
 ; SVE-LABEL: umulh_v32i8:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.b, vl16
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    umulh z0.b, p0/m, z0.b, z1.b
 ; SVE-NEXT:    movprfx z1, z2
@@ -525,9 +525,9 @@ define void @umulh_v32i8(ptr %a, ptr %b) {
 define <2 x i16> @umulh_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
 ; SVE-LABEL: umulh_v2i16:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s, vl2
 ; SVE-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; SVE-NEXT:    // kill: def $d0 killed $d0 def $z0
+; SVE-NEXT:    ptrue p0.s, vl2
 ; SVE-NEXT:    and z0.s, z0.s, #0xffff
 ; SVE-NEXT:    and z1.s, z1.s, #0xffff
 ; SVE-NEXT:    mul z0.s, p0/m, z0.s, z1.s
@@ -606,8 +606,8 @@ define <8 x i16> @umulh_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @umulh_v16i16(ptr %a, ptr %b) {
 ; SVE-LABEL: umulh_v16i16:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.h, vl8
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    umulh z0.h, p0/m, z0.h, z1.h
 ; SVE-NEXT:    movprfx z1, z2
@@ -687,8 +687,8 @@ define <4 x i32> @umulh_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @umulh_v8i32(ptr %a, ptr %b) {
 ; SVE-LABEL: umulh_v8i32:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.s, vl4
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    umulh z0.s, p0/m, z0.s, z1.s
 ; SVE-NEXT:    movprfx z1, z2
@@ -770,8 +770,8 @@ define <2 x i64> @umulh_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @umulh_v4i64(ptr %a, ptr %b) {
 ; SVE-LABEL: umulh_v4i64:
 ; SVE:       // %bb.0:
-; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q0, q3, [x1]
+; SVE-NEXT:    ptrue p0.d, vl2
 ; SVE-NEXT:    ldp q1, q2, [x0]
 ; SVE-NEXT:    umulh z0.d, p0/m, z0.d, z1.d
 ; SVE-NEXT:    movprfx z1, z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-reduce.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-reduce.ll
index c51630ecd752a..751f43768a511 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-reduce.ll
@@ -37,8 +37,8 @@ define i8 @uaddv_v16i8(<16 x i8> %a) {
 define i8 @uaddv_v32i8(ptr %a) {
 ; CHECK-LABEL: uaddv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    add z0.b, z1.b, z0.b
 ; CHECK-NEXT:    uaddv d0, p0, z0.b
 ; CHECK-NEXT:    fmov x0, d0
@@ -78,8 +78,8 @@ define i16 @uaddv_v8i16(<8 x i16> %a) {
 define i16 @uaddv_v16i16(ptr %a) {
 ; CHECK-LABEL: uaddv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    add z0.h, z1.h, z0.h
 ; CHECK-NEXT:    uaddv d0, p0, z0.h
 ; CHECK-NEXT:    fmov x0, d0
@@ -119,8 +119,8 @@ define i32 @uaddv_v4i32(<4 x i32> %a) {
 define i32 @uaddv_v8i32(ptr %a) {
 ; CHECK-LABEL: uaddv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    add z0.s, z1.s, z0.s
 ; CHECK-NEXT:    uaddv d0, p0, z0.s
 ; CHECK-NEXT:    fmov x0, d0
@@ -146,8 +146,8 @@ define i64 @uaddv_v2i64(<2 x i64> %a) {
 define i64 @uaddv_v4i64(ptr %a) {
 ; CHECK-LABEL: uaddv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    add z0.d, z1.d, z0.d
 ; CHECK-NEXT:    uaddv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -188,8 +188,8 @@ define i8 @smaxv_v16i8(<16 x i8> %a) {
 define i8 @smaxv_v32i8(ptr %a) {
 ; CHECK-LABEL: smaxv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    smax z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    smaxv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -226,8 +226,8 @@ define i16 @smaxv_v8i16(<8 x i16> %a) {
 define i16 @smaxv_v16i16(ptr %a) {
 ; CHECK-LABEL: smaxv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    smax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    smaxv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -264,8 +264,8 @@ define i32 @smaxv_v4i32(<4 x i32> %a) {
 define i32 @smaxv_v8i32(ptr %a) {
 ; CHECK-LABEL: smaxv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    smax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    smaxv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -291,8 +291,8 @@ define i64 @smaxv_v2i64(<2 x i64> %a) {
 define i64 @smaxv_v4i64(ptr %a) {
 ; CHECK-LABEL: smaxv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    smax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    smaxv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -333,8 +333,8 @@ define i8 @sminv_v16i8(<16 x i8> %a) {
 define i8 @sminv_v32i8(ptr %a) {
 ; CHECK-LABEL: sminv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    smin z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    sminv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -371,8 +371,8 @@ define i16 @sminv_v8i16(<8 x i16> %a) {
 define i16 @sminv_v16i16(ptr %a) {
 ; CHECK-LABEL: sminv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    smin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    sminv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -409,8 +409,8 @@ define i32 @sminv_v4i32(<4 x i32> %a) {
 define i32 @sminv_v8i32(ptr %a) {
 ; CHECK-LABEL: sminv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    smin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    sminv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -436,8 +436,8 @@ define i64 @sminv_v2i64(<2 x i64> %a) {
 define i64 @sminv_v4i64(ptr %a) {
 ; CHECK-LABEL: sminv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    smin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    sminv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -478,8 +478,8 @@ define i8 @umaxv_v16i8(<16 x i8> %a) {
 define i8 @umaxv_v32i8(ptr %a) {
 ; CHECK-LABEL: umaxv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    umax z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    umaxv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -516,8 +516,8 @@ define i16 @umaxv_v8i16(<8 x i16> %a) {
 define i16 @umaxv_v16i16(ptr %a) {
 ; CHECK-LABEL: umaxv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    umax z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    umaxv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -554,8 +554,8 @@ define i32 @umaxv_v4i32(<4 x i32> %a) {
 define i32 @umaxv_v8i32(ptr %a) {
 ; CHECK-LABEL: umaxv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    umax z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    umaxv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -581,8 +581,8 @@ define i64 @umaxv_v2i64(<2 x i64> %a) {
 define i64 @umaxv_v4i64(ptr %a) {
 ; CHECK-LABEL: umaxv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    umax z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    umaxv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -623,8 +623,8 @@ define i8 @uminv_v16i8(<16 x i8> %a) {
 define i8 @uminv_v32i8(ptr %a) {
 ; CHECK-LABEL: uminv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    umin z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    uminv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -661,8 +661,8 @@ define i16 @uminv_v8i16(<8 x i16> %a) {
 define i16 @uminv_v16i16(ptr %a) {
 ; CHECK-LABEL: uminv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    umin z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    uminv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -699,8 +699,8 @@ define i32 @uminv_v4i32(<4 x i32> %a) {
 define i32 @uminv_v8i32(ptr %a) {
 ; CHECK-LABEL: uminv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    umin z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    uminv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -726,8 +726,8 @@ define i64 @uminv_v2i64(<2 x i64> %a) {
 define i64 @uminv_v4i64(ptr %a) {
 ; CHECK-LABEL: uminv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    umin z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    uminv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
index 4a1209b942f4a..d373a9063f852 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-rem.ll
@@ -65,7 +65,6 @@ define <16 x i8> @srem_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ptrue p1.b, vl16
 ; CHECK-NEXT:    ext z2.b, z2.b, z1.b, #8
 ; CHECK-NEXT:    ext z3.b, z3.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z2.h, z2.b
@@ -91,15 +90,16 @@ define <16 x i8> @srem_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    sdivr z3.s, p0/m, z3.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z4.h, p0, z4.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z2.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z5.h, p0, z5.h, z3.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
 ; CHECK-NEXT:    uzp1 z3.b, z5.b, z5.b
 ; CHECK-NEXT:    splice z3.b, p0, z3.b, z2.b
-; CHECK-NEXT:    mls z0.b, p1/m, z3.b, z1.b
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    mls z0.b, p0/m, z3.b, z1.b
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
 ; CHECK-NEXT:    ret
   %res = srem <16 x i8> %op1, %op2
@@ -112,7 +112,6 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ldr q0, [x0, #16]
 ; CHECK-NEXT:    ldr q1, [x1, #16]
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ptrue p1.b, vl16
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    sunpklo z7.h, z1.b
@@ -171,22 +170,23 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z17.h, z17.h, z17.h
 ; CHECK-NEXT:    sdivr z18.s, p0/m, z18.s, z20.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
 ; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z5.h
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
+; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
 ; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
 ; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
 ; CHECK-NEXT:    splice z19.h, p0, z19.h, z18.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z6.b, p0, z6.b, z2.b
+; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
-; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    mls z0.b, p0/m, z6.b, z1.b
 ; CHECK-NEXT:    movprfx z2, z3
-; CHECK-NEXT:    mls z2.b, p1/m, z7.b, z4.b
+; CHECK-NEXT:    mls z2.b, p0/m, z7.b, z4.b
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i8>, ptr %a
@@ -199,11 +199,11 @@ define void @srem_v32i8(ptr %a, ptr %b) {
 define <4 x i16> @srem_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
 ; CHECK-LABEL: srem_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    sunpklo z2.s, z1.h
 ; CHECK-NEXT:    sunpklo z3.s, z0.h
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    sdivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
@@ -223,7 +223,6 @@ define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    sunpklo z4.s, z0.h
-; CHECK-NEXT:    ptrue p1.h, vl8
 ; CHECK-NEXT:    ext z2.b, z2.b, z1.b, #8
 ; CHECK-NEXT:    ext z3.b, z3.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z2.s, z2.h
@@ -235,7 +234,8 @@ define <8 x i16> @srem_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z3.h, p0, z3.h, z2.h
-; CHECK-NEXT:    mls z0.h, p1/m, z3.h, z1.h
+; CHECK-NEXT:    ptrue p0.h, vl8
+; CHECK-NEXT:    mls z0.h, p0/m, z3.h, z1.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
 ; CHECK-NEXT:    ret
   %res = srem <8 x i16> %op1, %op2
@@ -248,7 +248,6 @@ define void @srem_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ldp q4, q1, [x1]
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0, #16]
-; CHECK-NEXT:    ptrue p1.h, vl8
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    mov z5.d, z4.d
@@ -277,9 +276,10 @@ define void @srem_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z5.h
 ; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
 ; CHECK-NEXT:    splice z7.h, p0, z7.h, z2.h
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    movprfx z2, z3
-; CHECK-NEXT:    mls z2.h, p1/m, z6.h, z4.h
-; CHECK-NEXT:    mls z0.h, p1/m, z7.h, z1.h
+; CHECK-NEXT:    mls z2.h, p0/m, z6.h, z4.h
+; CHECK-NEXT:    mls z0.h, p0/m, z7.h, z1.h
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
@@ -322,8 +322,8 @@ define <4 x i32> @srem_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @srem_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: srem_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    movprfx z4, z1
 ; CHECK-NEXT:    sdiv z4.s, p0/m, z4.s, z0.s
@@ -374,8 +374,8 @@ define <2 x i64> @srem_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @srem_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: srem_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    movprfx z4, z1
 ; CHECK-NEXT:    sdiv z4.d, p0/m, z4.d, z0.d
@@ -454,7 +454,6 @@ define <16 x i8> @urem_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ptrue p1.b, vl16
 ; CHECK-NEXT:    ext z2.b, z2.b, z1.b, #8
 ; CHECK-NEXT:    ext z3.b, z3.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z2.h, z2.b
@@ -480,15 +479,16 @@ define <16 x i8> @urem_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    udivr z3.s, p0/m, z3.s, z5.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    splice z4.h, p0, z4.h, z2.h
+; CHECK-NEXT:    uzp1 z5.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z2.b, z4.b, z4.b
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z5.h, p0, z5.h, z3.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
 ; CHECK-NEXT:    uzp1 z3.b, z5.b, z5.b
 ; CHECK-NEXT:    splice z3.b, p0, z3.b, z2.b
-; CHECK-NEXT:    mls z0.b, p1/m, z3.b, z1.b
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    mls z0.b, p0/m, z3.b, z1.b
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
 ; CHECK-NEXT:    ret
   %res = urem <16 x i8> %op1, %op2
@@ -501,7 +501,6 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ldr q0, [x0, #16]
 ; CHECK-NEXT:    ldr q1, [x1, #16]
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ptrue p1.b, vl16
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    uunpklo z7.h, z1.b
@@ -560,22 +559,23 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z17.h, z17.h, z17.h
 ; CHECK-NEXT:    udivr z18.s, p0/m, z18.s, z20.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
 ; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z5.h
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
+; CHECK-NEXT:    uzp1 z19.h, z21.h, z21.h
 ; CHECK-NEXT:    uzp1 z5.b, z16.b, z16.b
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
 ; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
 ; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
 ; CHECK-NEXT:    splice z19.h, p0, z19.h, z18.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z6.b, p0, z6.b, z2.b
+; CHECK-NEXT:    uzp1 z7.b, z19.b, z19.b
 ; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
-; CHECK-NEXT:    mls z0.b, p1/m, z6.b, z1.b
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    mls z0.b, p0/m, z6.b, z1.b
 ; CHECK-NEXT:    movprfx z2, z3
-; CHECK-NEXT:    mls z2.b, p1/m, z7.b, z4.b
+; CHECK-NEXT:    mls z2.b, p0/m, z7.b, z4.b
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <32 x i8>, ptr %a
@@ -588,11 +588,11 @@ define void @urem_v32i8(ptr %a, ptr %b) {
 define <4 x i16> @urem_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
 ; CHECK-LABEL: urem_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    uunpklo z2.s, z1.h
 ; CHECK-NEXT:    uunpklo z3.s, z0.h
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    udivr z2.s, p0/m, z2.s, z3.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
@@ -612,7 +612,6 @@ define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z4.s, z0.h
-; CHECK-NEXT:    ptrue p1.h, vl8
 ; CHECK-NEXT:    ext z2.b, z2.b, z1.b, #8
 ; CHECK-NEXT:    ext z3.b, z3.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z2.s, z2.h
@@ -624,7 +623,8 @@ define <8 x i16> @urem_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    splice z3.h, p0, z3.h, z2.h
-; CHECK-NEXT:    mls z0.h, p1/m, z3.h, z1.h
+; CHECK-NEXT:    ptrue p0.h, vl8
+; CHECK-NEXT:    mls z0.h, p0/m, z3.h, z1.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
 ; CHECK-NEXT:    ret
   %res = urem <8 x i16> %op1, %op2
@@ -637,7 +637,6 @@ define void @urem_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ldp q4, q1, [x1]
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0, #16]
-; CHECK-NEXT:    ptrue p1.h, vl8
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    mov z3.d, z0.d
 ; CHECK-NEXT:    mov z5.d, z4.d
@@ -666,9 +665,10 @@ define void @urem_v16i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z5.h
 ; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
 ; CHECK-NEXT:    splice z7.h, p0, z7.h, z2.h
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    movprfx z2, z3
-; CHECK-NEXT:    mls z2.h, p1/m, z6.h, z4.h
-; CHECK-NEXT:    mls z0.h, p1/m, z7.h, z1.h
+; CHECK-NEXT:    mls z2.h, p0/m, z6.h, z4.h
+; CHECK-NEXT:    mls z0.h, p0/m, z7.h, z1.h
 ; CHECK-NEXT:    stp q2, q0, [x0]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
@@ -711,8 +711,8 @@ define <4 x i32> @urem_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @urem_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: urem_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    movprfx z4, z1
 ; CHECK-NEXT:    udiv z4.s, p0/m, z4.s, z0.s
@@ -763,8 +763,8 @@ define <2 x i64> @urem_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @urem_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: urem_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    movprfx z4, z1
 ; CHECK-NEXT:    udiv z4.d, p0/m, z4.d, z0.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-select.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-select.ll
index 3b58e35bd844c..906112f7ac39e 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-select.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-select.ll
@@ -7,8 +7,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define <4 x i8> @select_v4i8(<4 x i8> %op1, <4 x i8> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -23,8 +23,8 @@ define <4 x i8> @select_v4i8(<4 x i8> %op1, <4 x i8> %op2, i1 %mask) {
 define <8 x i8> @select_v8i8(<8 x i8> %op1, <8 x i8> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z2.b, #0
@@ -38,8 +38,8 @@ define <8 x i8> @select_v8i8(<8 x i8> %op1, <8 x i8> %op2, i1 %mask) {
 define <16 x i8> @select_v16i8(<16 x i8> %op1, <16 x i8> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z2.b, w0
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z2.b, #0
@@ -53,8 +53,8 @@ define <16 x i8> @select_v16i8(<16 x i8> %op1, <16 x i8> %op2, i1 %mask) {
 define void @select_v32i8(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    mov z0.b, w2
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x0, #16]
@@ -74,8 +74,8 @@ define void @select_v32i8(ptr %a, ptr %b, i1 %mask) {
 define <2 x i16> @select_v2i16(<2 x i16> %op1, <2 x i16> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    mov z2.s, w8
@@ -90,8 +90,8 @@ define <2 x i16> @select_v2i16(<2 x i16> %op1, <2 x i16> %op2, i1 %mask) {
 define <4 x i16> @select_v4i16(<4 x i16> %op1, <4 x i16> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -106,8 +106,8 @@ define <4 x i16> @select_v4i16(<4 x i16> %op1, <4 x i16> %op2, i1 %mask) {
 define <8 x i16> @select_v8i16(<8 x i16> %op1, <8 x i16> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z2.h, w0
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    and z2.h, z2.h, #0x1
@@ -122,8 +122,8 @@ define <8 x i16> @select_v8i16(<8 x i16> %op1, <8 x i16> %op2, i1 %mask) {
 define void @select_v16i16(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mov z0.h, w2
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
 ; CHECK-NEXT:    ldr q0, [x0]
@@ -144,8 +144,8 @@ define void @select_v16i16(ptr %a, ptr %b, i1 %mask) {
 define <2 x i32> @select_v2i32(<2 x i32> %op1, <2 x i32> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    mov z2.s, w8
@@ -160,8 +160,8 @@ define <2 x i32> @select_v2i32(<2 x i32> %op1, <2 x i32> %op2, i1 %mask) {
 define <4 x i32> @select_v4i32(<4 x i32> %op1, <4 x i32> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w0, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov z2.s, w8
@@ -176,8 +176,8 @@ define <4 x i32> @select_v4i32(<4 x i32> %op1, <4 x i32> %op2, i1 %mask) {
 define void @select_v8i32(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and w8, w2, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mov z0.s, w8
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    ldr q0, [x0]
@@ -198,9 +198,9 @@ define void @select_v8i32(ptr %a, ptr %b, i1 %mask) {
 define <1 x i64> @select_v1i64(<1 x i64> %op1, <1 x i64> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v1i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    mov z2.d, x8
@@ -215,9 +215,9 @@ define <1 x i64> @select_v1i64(<1 x i64> %op1, <1 x i64> %op2, i1 %mask) {
 define <2 x i64> @select_v2i64(<2 x i64> %op1, <2 x i64> %op2, i1 %mask) {
 ; CHECK-LABEL: select_v2i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
 ; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
 ; CHECK-NEXT:    mov z2.d, x8
@@ -232,9 +232,9 @@ define <2 x i64> @select_v2i64(<2 x i64> %op1, <2 x i64> %op2, i1 %mask) {
 define void @select_v4i64(ptr %a, ptr %b, i1 %mask) {
 ; CHECK-LABEL: select_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-NEXT:    and x8, x2, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z0.d, x8
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ldr q0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-shifts.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-shifts.ll
index c7fa0e8ad5e4a..9ed52e321d9ab 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-shifts.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-shifts.ll
@@ -52,8 +52,8 @@ define <16 x i8> @ashr_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @ashr_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: ashr_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    asrr z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -111,8 +111,8 @@ define <8 x i16> @ashr_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @ashr_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: ashr_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    asrr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -155,8 +155,8 @@ define <4 x i32> @ashr_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @ashr_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: ashr_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    asrr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -199,8 +199,8 @@ define <2 x i64> @ashr_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @ashr_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: ashr_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    asrr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -221,9 +221,9 @@ define void @ashr_v4i64(ptr %a, ptr %b) {
 define <4 x i8> @lshr_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
 ; CHECK-LABEL: lshr_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
 ; CHECK-NEXT:    lsr z0.h, p0/m, z0.h, z1.h
@@ -262,8 +262,8 @@ define <16 x i8> @lshr_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @lshr_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: lshr_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lsrr z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -280,9 +280,9 @@ define void @lshr_v32i8(ptr %a, ptr %b) {
 define <2 x i16> @lshr_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
 ; CHECK-LABEL: lshr_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    lsr z0.s, p0/m, z0.s, z1.s
@@ -321,8 +321,8 @@ define <8 x i16> @lshr_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @lshr_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: lshr_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lsrr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -365,8 +365,8 @@ define <4 x i32> @lshr_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @lshr_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: lshr_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lsrr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -409,8 +409,8 @@ define <2 x i64> @lshr_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @lshr_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: lshr_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lsrr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
@@ -431,8 +431,8 @@ define void @lshr_v4i64(ptr %a, ptr %b) {
 define <2 x i8> @shl_v2i8(<2 x i8> %op1, <2 x i8> %op2) {
 ; CHECK-LABEL: shl_v2i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    and z1.s, z1.s, #0xff
 ; CHECK-NEXT:    lsl z0.s, p0/m, z0.s, z1.s
@@ -445,8 +445,8 @@ define <2 x i8> @shl_v2i8(<2 x i8> %op1, <2 x i8> %op2) {
 define <4 x i8> @shl_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
 ; CHECK-LABEL: shl_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    lsl z0.h, p0/m, z0.h, z1.h
@@ -485,8 +485,8 @@ define <16 x i8> @shl_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
 define void @shl_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: shl_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lslr z0.b, p0/m, z0.b, z1.b
 ; CHECK-NEXT:    movprfx z1, z2
@@ -529,8 +529,8 @@ define <8 x i16> @shl_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
 define void @shl_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: shl_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lslr z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -573,8 +573,8 @@ define <4 x i32> @shl_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
 define void @shl_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: shl_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lslr z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -617,8 +617,8 @@ define <2 x i64> @shl_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
 define void @shl_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: shl_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    lslr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
index 5c5cf68135bf8..b285659258f31 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
@@ -36,8 +36,8 @@ define void @ucvtf_v8i16_v8f16(ptr %a, ptr %b) {
 define void @ucvtf_v16i16_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: ucvtf_v16i16_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.h
 ; CHECK-NEXT:    ucvtf z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -55,8 +55,8 @@ define void @ucvtf_v16i16_v16f16(ptr %a, ptr %b) {
 define <2 x float> @ucvtf_v2i16_v2f32(<2 x i16> %op1) {
 ; CHECK-LABEL: ucvtf_v2i16_v2f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
@@ -68,8 +68,8 @@ define <2 x float> @ucvtf_v2i16_v2f32(<2 x i16> %op1) {
 define <4 x float> @ucvtf_v4i16_v4f32(<4 x i16> %op1) {
 ; CHECK-LABEL: ucvtf_v4i16_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -277,8 +277,8 @@ define <4 x half> @ucvtf_v4i32_v4f16(<4 x i32> %op1) {
 define <8 x half> @ucvtf_v8i32_v8f16(ptr %a) {
 ; CHECK-LABEL: ucvtf_v8i32_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ucvtf z1.h, p0/m, z1.s
 ; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
@@ -295,8 +295,8 @@ define <8 x half> @ucvtf_v8i32_v8f16(ptr %a) {
 define void @ucvtf_v16i32_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: ucvtf_v16i32_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    ucvtf z1.h, p0/m, z1.s
 ; CHECK-NEXT:    ucvtf z0.h, p0/m, z0.s
@@ -348,8 +348,8 @@ define <4 x float> @ucvtf_v4i32_v4f32(<4 x i32> %op1) {
 define void @ucvtf_v8i32_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: ucvtf_v8i32_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    ucvtf z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -367,8 +367,8 @@ define void @ucvtf_v8i32_v8f32(ptr %a, ptr %b) {
 define <2 x double> @ucvtf_v2i32_v2f64(<2 x i32> %op1) {
 ; CHECK-LABEL: ucvtf_v2i32_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -446,16 +446,16 @@ define <2 x half> @ucvtf_v2i64_v2f16(<2 x i64> %op1) {
 define <4 x half> @ucvtf_v4i64_v4f16(ptr %a) {
 ; CHECK-LABEL: ucvtf_v4i64_v4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ucvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
 ; CHECK-NEXT:    ret
@@ -467,10 +467,9 @@ define <4 x half> @ucvtf_v4i64_v4f16(ptr %a) {
 define <8 x half> @ucvtf_v8i64_v8f16(ptr %a) {
 ; CHECK-LABEL: ucvtf_v8i64_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q2, q3, [x0]
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ucvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    ucvtf z3.s, p0/m, z3.d
@@ -482,11 +481,12 @@ define <8 x half> @ucvtf_v8i64_v8f16(ptr %a) {
 ; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
 ; CHECK-NEXT:    splice z1.s, p0, z1.s, z0.s
 ; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
-; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    movprfx z0, z1
-; CHECK-NEXT:    fcvt z0.h, p1/m, z1.s
+; CHECK-NEXT:    fcvt z0.h, p0/m, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
-; CHECK-NEXT:    fcvt z1.h, p1/m, z2.s
+; CHECK-NEXT:    fcvt z1.h, p0/m, z2.s
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z2.h, z0.h, z0.h
 ; CHECK-NEXT:    uzp1 z0.h, z1.h, z1.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z2.h
@@ -517,8 +517,8 @@ define <2 x float> @ucvtf_v2i64_v2f32(<2 x i64> %op1) {
 define <4 x float> @ucvtf_v4i64_v4f32(ptr %a) {
 ; CHECK-LABEL: ucvtf_v4i64_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ucvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
@@ -535,8 +535,8 @@ define <4 x float> @ucvtf_v4i64_v4f32(ptr %a) {
 define void @ucvtf_v8i64_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: ucvtf_v8i64_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    ucvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    ucvtf z0.s, p0/m, z0.d
@@ -576,8 +576,8 @@ define <2 x double> @ucvtf_v2i64_v2f64(<2 x i64> %op1) {
 define void @ucvtf_v4i64_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: ucvtf_v4i64_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    ucvtf z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -621,8 +621,8 @@ define void @scvtf_v8i16_v8f16(ptr %a, ptr %b) {
 define void @scvtf_v16i16_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: scvtf_v16i16_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    scvtf z0.h, p0/m, z0.h
 ; CHECK-NEXT:    scvtf z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -652,8 +652,8 @@ define <2 x float> @scvtf_v2i16_v2f32(<2 x i16> %op1) {
 define <4 x float> @scvtf_v4i16_v4f32(<4 x i16> %op1) {
 ; CHECK-LABEL: scvtf_v4i16_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -850,8 +850,8 @@ define <4 x half> @scvtf_v4i32_v4f16(<4 x i32> %op1) {
 define <8 x half> @scvtf_v8i32_v8f16(ptr %a) {
 ; CHECK-LABEL: scvtf_v8i32_v8f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    scvtf z1.h, p0/m, z1.s
 ; CHECK-NEXT:    scvtf z0.h, p0/m, z0.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
@@ -896,8 +896,8 @@ define <4 x float> @scvtf_v4i32_v4f32(<4 x i32> %op1) {
 define void @scvtf_v8i32_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: scvtf_v8i32_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z0.s
 ; CHECK-NEXT:    scvtf z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x1]
@@ -915,8 +915,8 @@ define void @scvtf_v8i32_v8f32(ptr %a, ptr %b) {
 define <2 x double> @scvtf_v2i32_v2f64(<2 x i32> %op1) {
 ; CHECK-LABEL: scvtf_v2i32_v2f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
@@ -1038,16 +1038,16 @@ define <2 x half> @scvtf_v2i64_v2f16(<2 x i64> %op1) {
 define <4 x half> @scvtf_v4i64_v4f16(ptr %a) {
 ; CHECK-LABEL: scvtf_v4i64_v4f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
-; CHECK-NEXT:    ptrue p1.s
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    scvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $z0
 ; CHECK-NEXT:    ret
@@ -1076,8 +1076,8 @@ define <2 x float> @scvtf_v2i64_v2f32(<2 x i64> %op1) {
 define <4 x float> @scvtf_v4i64_v4f32(ptr %a) {
 ; CHECK-LABEL: scvtf_v4i64_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    scvtf z1.s, p0/m, z1.d
 ; CHECK-NEXT:    scvtf z0.s, p0/m, z0.d
 ; CHECK-NEXT:    ptrue p0.s, vl2
@@ -1110,8 +1110,8 @@ define <2 x double> @scvtf_v2i64_v2f64(<2 x i64> %op1) {
 define void @scvtf_v4i64_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: scvtf_v4i64_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
 ; CHECK-NEXT:    scvtf z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x1]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-vselect.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-vselect.ll
index 1809cfcf3db69..81bbaa92d4b47 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-vselect.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-vselect.ll
@@ -61,8 +61,8 @@ define <16 x i8> @select_v16i8(<16 x i8> %op1, <16 x i8> %op2, <16 x i1> %mask)
 define void @select_v32i8(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    cmpeq p1.b, p0/z, z0.b, z1.b
 ; CHECK-NEXT:    cmpeq p0.b, p0/z, z2.b, z3.b
@@ -136,8 +136,8 @@ define <8 x i16> @select_v8i16(<8 x i16> %op1, <8 x i16> %op2, <8 x i1> %mask) {
 define void @select_v16i16(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    cmpeq p1.h, p0/z, z0.h, z1.h
 ; CHECK-NEXT:    cmpeq p0.h, p0/z, z2.h, z3.h
@@ -193,8 +193,8 @@ define <4 x i32> @select_v4i32(<4 x i32> %op1, <4 x i32> %op2, <4 x i1> %mask) {
 define void @select_v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    cmpeq p1.s, p0/z, z0.s, z1.s
 ; CHECK-NEXT:    cmpeq p0.s, p0/z, z2.s, z3.s
@@ -213,9 +213,9 @@ define void @select_v8i32(ptr %a, ptr %b) {
 define <1 x i64> @select_v1i64(<1 x i64> %op1, <1 x i64> %op2, <1 x i1> %mask) {
 ; CHECK-LABEL: select_v1i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $w0 killed $w0 def $x0
 ; CHECK-NEXT:    and x8, x0, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $z0
 ; CHECK-NEXT:    // kill: def $d1 killed $d1 def $z1
 ; CHECK-NEXT:    mov z2.d, x8
@@ -249,8 +249,8 @@ define <2 x i64> @select_v2i64(<2 x i64> %op1, <2 x i64> %op2, <2 x i1> %mask) {
 define void @select_v4i64(ptr %a, ptr %b) {
 ; CHECK-LABEL: select_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q2, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    cmpeq p1.d, p0/z, z0.d, z1.d
 ; CHECK-NEXT:    cmpeq p0.d, p0/z, z2.d, z3.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-log-reduce.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-log-reduce.ll
index bb1bd8fe72b21..c4aeb4465c537 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-log-reduce.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-log-reduce.ll
@@ -48,8 +48,8 @@ define i8 @andv_v16i8(<16 x i8> %a) {
 define i8 @andv_v32i8(ptr %a) {
 ; CHECK-LABEL: andv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    and z0.d, z1.d, z0.d
 ; CHECK-NEXT:    andv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -98,8 +98,8 @@ define i16 @andv_v8i16(<8 x i16> %a) {
 define i16 @andv_v16i16(ptr %a) {
 ; CHECK-LABEL: andv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    and z0.d, z1.d, z0.d
 ; CHECK-NEXT:    andv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -136,8 +136,8 @@ define i32 @andv_v4i32(<4 x i32> %a) {
 define i32 @andv_v8i32(ptr %a) {
 ; CHECK-LABEL: andv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    and z0.d, z1.d, z0.d
 ; CHECK-NEXT:    andv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -162,8 +162,8 @@ define i64 @andv_v2i64(<2 x i64> %a) {
 define i64 @andv_v4i64(ptr %a) {
 ; CHECK-LABEL: andv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    and z0.d, z1.d, z0.d
 ; CHECK-NEXT:    andv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -216,8 +216,8 @@ define i8 @eorv_v16i8(<16 x i8> %a) {
 define i8 @eorv_v32i8(ptr %a) {
 ; CHECK-LABEL: eorv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    eor z0.d, z1.d, z0.d
 ; CHECK-NEXT:    eorv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -266,8 +266,8 @@ define i16 @eorv_v8i16(<8 x i16> %a) {
 define i16 @eorv_v16i16(ptr %a) {
 ; CHECK-LABEL: eorv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    eor z0.d, z1.d, z0.d
 ; CHECK-NEXT:    eorv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -304,8 +304,8 @@ define i32 @eorv_v4i32(<4 x i32> %a) {
 define i32 @eorv_v8i32(ptr %a) {
 ; CHECK-LABEL: eorv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    eor z0.d, z1.d, z0.d
 ; CHECK-NEXT:    eorv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -330,8 +330,8 @@ define i64 @eorv_v2i64(<2 x i64> %a) {
 define i64 @eorv_v4i64(ptr %a) {
 ; CHECK-LABEL: eorv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    eor z0.d, z1.d, z0.d
 ; CHECK-NEXT:    eorv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
@@ -384,8 +384,8 @@ define i8 @orv_v16i8(<16 x i8> %a) {
 define i8 @orv_v32i8(ptr %a) {
 ; CHECK-LABEL: orv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    orr z0.d, z1.d, z0.d
 ; CHECK-NEXT:    orv b0, p0, z0.b
 ; CHECK-NEXT:    fmov w0, s0
@@ -434,8 +434,8 @@ define i16 @orv_v8i16(<8 x i16> %a) {
 define i16 @orv_v16i16(ptr %a) {
 ; CHECK-LABEL: orv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    orr z0.d, z1.d, z0.d
 ; CHECK-NEXT:    orv h0, p0, z0.h
 ; CHECK-NEXT:    fmov w0, s0
@@ -472,8 +472,8 @@ define i32 @orv_v4i32(<4 x i32> %a) {
 define i32 @orv_v8i32(ptr %a) {
 ; CHECK-LABEL: orv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    orr z0.d, z1.d, z0.d
 ; CHECK-NEXT:    orv s0, p0, z0.s
 ; CHECK-NEXT:    fmov w0, s0
@@ -498,8 +498,8 @@ define i64 @orv_v2i64(<2 x i64> %a) {
 define i64 @orv_v4i64(ptr %a) {
 ; CHECK-LABEL: orv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q0, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    orr z0.d, z1.d, z0.d
 ; CHECK-NEXT:    orv d0, p0, z0.d
 ; CHECK-NEXT:    fmov x0, d0
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-masked-store.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-masked-store.ll
index e812706744745..f2b3f9b12ea71 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-masked-store.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-masked-store.ll
@@ -123,8 +123,8 @@ define void @masked_store_v32i8(ptr %dst, <32 x i1> %mask) {
 ; CHECK-NEXT:    asr z0.b, z0.b, #7
 ; CHECK-NEXT:    asr z1.b, z1.b, #7
 ; CHECK-NEXT:    cmpne p1.b, p0/z, z0.b, #0
-; CHECK-NEXT:    mov z0.b, #0 // =0x0
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z1.b, #0
+; CHECK-NEXT:    mov z0.b, #0 // =0x0
 ; CHECK-NEXT:    st1b { z0.b }, p1, [x0, x8]
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
 ; CHECK-NEXT:    add sp, sp, #32
@@ -204,8 +204,8 @@ define void @masked_store_v16f16(ptr %dst, <16 x i1> %mask) {
 ; CHECK-NEXT:    asr z0.h, z0.h, #15
 ; CHECK-NEXT:    asr z1.h, z1.h, #15
 ; CHECK-NEXT:    cmpne p1.h, p0/z, z1.h, #0
-; CHECK-NEXT:    mov z1.h, #0 // =0x0
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
+; CHECK-NEXT:    mov z1.h, #0 // =0x0
 ; CHECK-NEXT:    st1h { z1.h }, p1, [x0, x8, lsl #1]
 ; CHECK-NEXT:    st1h { z1.h }, p0, [x0]
 ; CHECK-NEXT:    ret
@@ -242,7 +242,7 @@ define void @masked_store_v8f32(ptr %dst, <8 x i1> %mask) {
 ; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    fmov w8, s1
 ; CHECK-NEXT:    fmov w9, s2
-; CHECK-NEXT:    mov z2.s, #0 // =0x0
+; CHECK-NEXT:    mov z2.b, z0.b[3]
 ; CHECK-NEXT:    strh w8, [sp, #14]
 ; CHECK-NEXT:    fmov w8, s3
 ; CHECK-NEXT:    mov z3.b, z0.b[2]
@@ -258,9 +258,9 @@ define void @masked_store_v8f32(ptr %dst, <8 x i1> %mask) {
 ; CHECK-NEXT:    lsl z1.s, z1.s, #31
 ; CHECK-NEXT:    asr z1.s, z1.s, #31
 ; CHECK-NEXT:    cmpne p1.s, p0/z, z1.s, #0
-; CHECK-NEXT:    mov z1.b, z0.b[3]
-; CHECK-NEXT:    st1w { z2.s }, p1, [x0, x8, lsl #2]
-; CHECK-NEXT:    fmov w8, s1
+; CHECK-NEXT:    mov z1.s, #0 // =0x0
+; CHECK-NEXT:    st1w { z1.s }, p1, [x0, x8, lsl #2]
+; CHECK-NEXT:    fmov w8, s2
 ; CHECK-NEXT:    strh w9, [sp]
 ; CHECK-NEXT:    strh w8, [sp, #6]
 ; CHECK-NEXT:    fmov w8, s3
@@ -272,7 +272,7 @@ define void @masked_store_v8f32(ptr %dst, <8 x i1> %mask) {
 ; CHECK-NEXT:    lsl z0.s, z0.s, #31
 ; CHECK-NEXT:    asr z0.s, z0.s, #31
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
-; CHECK-NEXT:    st1w { z2.s }, p0, [x0]
+; CHECK-NEXT:    st1w { z1.s }, p0, [x0]
 ; CHECK-NEXT:    add sp, sp, #16
 ; CHECK-NEXT:    ret
   call void @llvm.masked.store.v8f32(<8 x float> zeroinitializer, ptr %dst, i32 8, <8 x i1> %mask)
@@ -310,8 +310,8 @@ define void @masked_store_v4f64(ptr %dst, <4 x i1> %mask) {
 ; CHECK-NEXT:    asr z1.d, z1.d, #63
 ; CHECK-NEXT:    asr z0.d, z0.d, #63
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z0.d, #0
-; CHECK-NEXT:    mov z0.d, #0 // =0x0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
+; CHECK-NEXT:    mov z0.d, #0 // =0x0
 ; CHECK-NEXT:    st1d { z0.d }, p1, [x0, x8, lsl #3]
 ; CHECK-NEXT:    st1d { z0.d }, p0, [x0]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll
index f0b0b3269e98f..6fcb95f283338 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-optimize-ptrue.ll
@@ -170,8 +170,8 @@ define void @abs_v4i32(ptr %a) {
 define void @abs_v8i32(ptr %a) {
 ; CHECK-LABEL: abs_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    abs z0.s, p0/m, z0.s
 ; CHECK-NEXT:    abs z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -199,8 +199,8 @@ define void @abs_v2i64(ptr %a) {
 define void @abs_v4i64(ptr %a) {
 ; CHECK-LABEL: abs_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    abs z0.d, p0/m, z0.d
 ; CHECK-NEXT:    abs z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -263,8 +263,8 @@ define void @fadd_v8f16(ptr %a, ptr %b) {
 define void @fadd_v16f16(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v16f16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    movprfx z1, z2
@@ -313,8 +313,8 @@ define void @fadd_v4f32(ptr %a, ptr %b) {
 define void @fadd_v8f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    movprfx z1, z2
@@ -347,8 +347,8 @@ define void @fadd_v2f64(ptr %a, ptr %b) {
 define void @fadd_v4f64(ptr %a, ptr %b) {
 ; CHECK-LABEL: fadd_v4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q3, [x1]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q1, q2, [x0]
 ; CHECK-NEXT:    fadd z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    movprfx z1, z2
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-rev.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-rev.ll
index d1bff4fa21a11..00413302798ca 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-rev.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-rev.ll
@@ -9,8 +9,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @test_revbv16i16(ptr %a) {
 ; CHECK-LABEL: test_revbv16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    revb z0.h, p0/m, z0.h
 ; CHECK-NEXT:    revb z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -25,8 +25,8 @@ define void @test_revbv16i16(ptr %a) {
 define void @test_revbv8i32(ptr %a) {
 ; CHECK-LABEL: test_revbv8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    revb z0.s, p0/m, z0.s
 ; CHECK-NEXT:    revb z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -41,8 +41,8 @@ define void @test_revbv8i32(ptr %a) {
 define void @test_revbv4i64(ptr %a) {
 ; CHECK-LABEL: test_revbv4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revb z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revb z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -57,8 +57,8 @@ define void @test_revbv4i64(ptr %a) {
 define void @test_revhv8i32(ptr %a) {
 ; CHECK-LABEL: test_revhv8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    revh z0.s, p0/m, z0.s
 ; CHECK-NEXT:    revh z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -73,8 +73,8 @@ define void @test_revhv8i32(ptr %a) {
 define void @test_revhv8f32(ptr %a) {
 ; CHECK-LABEL: test_revhv8f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    revh z0.s, p0/m, z0.s
 ; CHECK-NEXT:    revh z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -89,8 +89,8 @@ define void @test_revhv8f32(ptr %a) {
 define void @test_revhv4i64(ptr %a) {
 ; CHECK-LABEL: test_revhv4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revh z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revh z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -105,8 +105,8 @@ define void @test_revhv4i64(ptr %a) {
 define void @test_revwv4i64(ptr %a) {
 ; CHECK-LABEL: test_revwv4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revw z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revw z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -121,8 +121,8 @@ define void @test_revwv4i64(ptr %a) {
 define void @test_revwv4f64(ptr %a) {
 ; CHECK-LABEL: test_revwv4f64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revw z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revw z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -150,8 +150,8 @@ define <16 x i8> @test_revv16i8(ptr %a) {
 define void @test_revwv8i32v8i32(ptr %a, ptr %b) {
 ; CHECK-LABEL: test_revwv8i32v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x1]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revw z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revw z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -166,8 +166,8 @@ define void @test_revwv8i32v8i32(ptr %a, ptr %b) {
 define void @test_revhv32i16(ptr %a) {
 ; CHECK-LABEL: test_revhv32i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    revh z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revh z1.d, p0/m, z1.d
@@ -202,8 +202,8 @@ define void @test_rev_elts_fail(ptr %a) {
 define void @test_revdv4i64_sve2p1(ptr %a) #1 {
 ; CHECK-LABEL: test_revdv4i64_sve2p1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    revd z0.q, p0/m, z0.q
 ; CHECK-NEXT:    revd z1.q, p0/m, z1.q
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -217,8 +217,8 @@ define void @test_revdv4i64_sve2p1(ptr %a) #1 {
 define void @test_revdv4f64_sve2p1(ptr %a) #1 {
 ; CHECK-LABEL: test_revdv4f64_sve2p1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    revd z0.q, p0/m, z0.q
 ; CHECK-NEXT:    revd z1.q, p0/m, z1.q
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-zip-uzp-trn.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-zip-uzp-trn.ll
index d7bfb6b2680e1..cb73030306b02 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-zip-uzp-trn.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-permute-zip-uzp-trn.ll
@@ -82,11 +82,11 @@ define void @zip_v32i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    ldp q1, q3, [x1]
 ; CHECK-NEXT:    ldp q0, q4, [x0]
-; CHECK-NEXT:    ldp q2, q6, [x0, #32]
+; CHECK-NEXT:    ldp q2, q5, [x0, #32]
 ; CHECK-NEXT:    mov z16.h, z3.h[7]
 ; CHECK-NEXT:    mov z18.h, z3.h[6]
 ; CHECK-NEXT:    mov z17.h, z4.h[7]
-; CHECK-NEXT:    ldp q5, q7, [x1, #32]
+; CHECK-NEXT:    ldp q6, q7, [x1, #32]
 ; CHECK-NEXT:    mov z19.h, z4.h[6]
 ; CHECK-NEXT:    fmov w8, s16
 ; CHECK-NEXT:    mov z16.h, z3.h[5]
@@ -98,13 +98,13 @@ define void @zip_v32i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov z18.h, z3.h[4]
 ; CHECK-NEXT:    strh w9, [sp, #28]
 ; CHECK-NEXT:    fmov w9, s19
-; CHECK-NEXT:    mov z19.h, z6.h[7]
+; CHECK-NEXT:    mov z19.h, z5.h[7]
 ; CHECK-NEXT:    zip1 z3.h, z4.h, z3.h
 ; CHECK-NEXT:    strh w8, [sp, #26]
 ; CHECK-NEXT:    fmov w8, s16
 ; CHECK-NEXT:    mov z16.h, z4.h[4]
 ; CHECK-NEXT:    strh w9, [sp, #24]
-; CHECK-NEXT:    zip1 z4.h, z6.h, z7.h
+; CHECK-NEXT:    zip1 z4.h, z5.h, z7.h
 ; CHECK-NEXT:    strh w8, [sp, #22]
 ; CHECK-NEXT:    fmov w8, s17
 ; CHECK-NEXT:    mov z17.h, z1.h[7]
@@ -131,7 +131,7 @@ define void @zip_v32i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    fmov w8, s18
 ; CHECK-NEXT:    mov z18.h, z0.h[4]
 ; CHECK-NEXT:    zip1 z0.h, z0.h, z1.h
-; CHECK-NEXT:    zip1 z1.h, z2.h, z5.h
+; CHECK-NEXT:    zip1 z1.h, z2.h, z6.h
 ; CHECK-NEXT:    strh w8, [sp, #54]
 ; CHECK-NEXT:    fmov w8, s16
 ; CHECK-NEXT:    ldr q16, [sp, #16]
@@ -143,41 +143,41 @@ define void @zip_v32i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov z18.h, z7.h[7]
 ; CHECK-NEXT:    strh w8, [sp, #48]
 ; CHECK-NEXT:    fmov w8, s18
-; CHECK-NEXT:    mov z18.h, z6.h[6]
+; CHECK-NEXT:    mov z18.h, z5.h[6]
 ; CHECK-NEXT:    ldr q17, [sp, #48]
 ; CHECK-NEXT:    strh w8, [sp, #46]
 ; CHECK-NEXT:    fmov w8, s19
 ; CHECK-NEXT:    mov z19.h, z7.h[5]
 ; CHECK-NEXT:    strh w8, [sp, #44]
 ; CHECK-NEXT:    fmov w8, s20
-; CHECK-NEXT:    mov z20.h, z6.h[5]
+; CHECK-NEXT:    mov z20.h, z5.h[5]
 ; CHECK-NEXT:    strh w8, [sp, #42]
 ; CHECK-NEXT:    fmov w8, s18
 ; CHECK-NEXT:    mov z18.h, z7.h[4]
 ; CHECK-NEXT:    strh w8, [sp, #40]
 ; CHECK-NEXT:    fmov w8, s19
-; CHECK-NEXT:    mov z19.h, z6.h[4]
+; CHECK-NEXT:    mov z19.h, z5.h[4]
 ; CHECK-NEXT:    strh w8, [sp, #38]
 ; CHECK-NEXT:    fmov w8, s20
-; CHECK-NEXT:    mov z20.h, z5.h[7]
+; CHECK-NEXT:    mov z20.h, z6.h[7]
 ; CHECK-NEXT:    strh w8, [sp, #36]
 ; CHECK-NEXT:    fmov w8, s18
 ; CHECK-NEXT:    mov z18.h, z2.h[7]
 ; CHECK-NEXT:    strh w8, [sp, #34]
 ; CHECK-NEXT:    fmov w8, s19
-; CHECK-NEXT:    mov z19.h, z5.h[6]
+; CHECK-NEXT:    mov z19.h, z6.h[6]
 ; CHECK-NEXT:    strh w8, [sp, #32]
 ; CHECK-NEXT:    fmov w8, s20
 ; CHECK-NEXT:    mov z20.h, z2.h[6]
 ; CHECK-NEXT:    strh w8, [sp, #14]
 ; CHECK-NEXT:    fmov w8, s18
-; CHECK-NEXT:    mov z18.h, z5.h[5]
+; CHECK-NEXT:    mov z18.h, z6.h[5]
 ; CHECK-NEXT:    strh w8, [sp, #12]
 ; CHECK-NEXT:    fmov w8, s19
 ; CHECK-NEXT:    mov z19.h, z2.h[5]
 ; CHECK-NEXT:    strh w8, [sp, #10]
 ; CHECK-NEXT:    fmov w8, s20
-; CHECK-NEXT:    mov z20.h, z5.h[4]
+; CHECK-NEXT:    mov z20.h, z6.h[4]
 ; CHECK-NEXT:    fmov w9, s19
 ; CHECK-NEXT:    strh w8, [sp, #8]
 ; CHECK-NEXT:    fmov w8, s18
@@ -186,10 +186,10 @@ define void @zip_v32i16(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ldr q2, [sp, #32]
 ; CHECK-NEXT:    strh w8, [sp, #6]
 ; CHECK-NEXT:    fmov w8, s20
+; CHECK-NEXT:    fmov w9, s18
 ; CHECK-NEXT:    add z2.h, z16.h, z2.h
 ; CHECK-NEXT:    strh w8, [sp, #2]
-; CHECK-NEXT:    fmov w8, s18
-; CHECK-NEXT:    strh w8, [sp]
+; CHECK-NEXT:    strh w9, [sp]
 ; CHECK-NEXT:    ldr q4, [sp]
 ; CHECK-NEXT:    stp q3, q2, [x0, #32]
 ; CHECK-NEXT:    add z1.h, z17.h, z4.h
@@ -471,9 +471,9 @@ define void @trn_v4f64(ptr %a, ptr %b) {
 define void @trn_v4f32(ptr %a, ptr %b) {
 ; CHECK-LABEL: trn_v4f32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldr q0, [x0]
 ; CHECK-NEXT:    ldr q1, [x1]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    trn1 z2.s, z0.s, z1.s
 ; CHECK-NEXT:    trn2 z0.s, z0.s, z1.s
 ; CHECK-NEXT:    fadd z0.s, p0/m, z0.s, z2.s
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll
index f2ba4a7cc3567..ab7c42b3e9e37 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-ptest.ll
@@ -8,8 +8,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define i1 @ptest_v16i1(ptr %a, ptr %b) {
 ; CHECK-LABEL: ptest_v16i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q1, q0, [x0, #32]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q2, q3, [x0]
 ; CHECK-NEXT:    fcmne p1.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    fcmne p2.s, p0/z, z1.s, #0.0
@@ -20,7 +20,6 @@ define i1 @ptest_v16i1(ptr %a, ptr %b) {
 ; CHECK-NEXT:    mov z2.s, p3/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z3.s, p0/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    ptrue p1.b, vl16
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
@@ -31,7 +30,8 @@ define i1 @ptest_v16i1(ptr %a, ptr %b) {
 ; CHECK-NEXT:    uzp1 z0.b, z1.b, z1.b
 ; CHECK-NEXT:    uzp1 z1.b, z3.b, z3.b
 ; CHECK-NEXT:    splice z1.b, p0, z1.b, z0.b
-; CHECK-NEXT:    umaxv b0, p1, z1.b
+; CHECK-NEXT:    ptrue p0.b, vl16
+; CHECK-NEXT:    umaxv b0, p0, z1.b
 ; CHECK-NEXT:    fmov w8, s0
 ; CHECK-NEXT:    and w0, w8, #0x1
 ; CHECK-NEXT:    ret
@@ -45,41 +45,41 @@ define i1 @ptest_v16i1(ptr %a, ptr %b) {
 define i1 @ptest_or_v16i1(ptr %a, ptr %b) {
 ; CHECK-LABEL: ptest_or_v16i1:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q1, q0, [x0, #32]
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ldp q0, q1, [x0]
-; CHECK-NEXT:    ldp q2, q3, [x0, #32]
-; CHECK-NEXT:    ldp q4, q5, [x1]
-; CHECK-NEXT:    ldp q6, q7, [x1, #32]
-; CHECK-NEXT:    fcmne p1.s, p0/z, z3.s, #0.0
-; CHECK-NEXT:    fcmne p2.s, p0/z, z2.s, #0.0
-; CHECK-NEXT:    fcmne p3.s, p0/z, z1.s, #0.0
-; CHECK-NEXT:    fcmne p4.s, p0/z, z0.s, #0.0
-; CHECK-NEXT:    fcmne p5.s, p0/z, z7.s, #0.0
-; CHECK-NEXT:    fcmne p6.s, p0/z, z6.s, #0.0
-; CHECK-NEXT:    fcmne p7.s, p0/z, z5.s, #0.0
-; CHECK-NEXT:    fcmne p0.s, p0/z, z4.s, #0.0
+; CHECK-NEXT:    ldp q2, q3, [x0]
+; CHECK-NEXT:    ldp q4, q5, [x1, #32]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fcmne p2.s, p0/z, z1.s, #0.0
+; CHECK-NEXT:    ldp q0, q1, [x1]
+; CHECK-NEXT:    fcmne p3.s, p0/z, z3.s, #0.0
+; CHECK-NEXT:    fcmne p4.s, p0/z, z2.s, #0.0
+; CHECK-NEXT:    fcmne p5.s, p0/z, z5.s, #0.0
+; CHECK-NEXT:    fcmne p6.s, p0/z, z4.s, #0.0
+; CHECK-NEXT:    fcmne p7.s, p0/z, z1.s, #0.0
+; CHECK-NEXT:    fcmne p0.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z1.s, p2/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z2.s, p3/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z3.s, p4/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z4.s, p5/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z5.s, p6/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    mov z6.s, p7/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z7.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    ptrue p1.h, vl4
-; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
 ; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
+; CHECK-NEXT:    splice z1.h, p0, z1.h, z0.h
+; CHECK-NEXT:    splice z3.h, p0, z3.h, z2.h
+; CHECK-NEXT:    splice z5.h, p0, z5.h, z4.h
+; CHECK-NEXT:    splice z7.h, p0, z7.h, z6.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    splice z1.h, p1, z1.h, z0.h
-; CHECK-NEXT:    splice z3.h, p1, z3.h, z2.h
-; CHECK-NEXT:    splice z5.h, p1, z5.h, z4.h
-; CHECK-NEXT:    splice z7.h, p1, z7.h, z6.h
 ; CHECK-NEXT:    uzp1 z0.b, z1.b, z1.b
 ; CHECK-NEXT:    uzp1 z1.b, z3.b, z3.b
 ; CHECK-NEXT:    uzp1 z2.b, z5.b, z5.b
@@ -112,41 +112,41 @@ declare i1 @llvm.vector.reduce.or.i1.v16i1(<16 x i1>)
 define i1 @ptest_and_v16i1(ptr %a, ptr %b) {
 ; CHECK-LABEL: ptest_and_v16i1:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    ldp q1, q0, [x0, #32]
 ; CHECK-NEXT:    ptrue p0.s, vl4
-; CHECK-NEXT:    ldp q0, q1, [x0]
-; CHECK-NEXT:    ldp q2, q3, [x0, #32]
-; CHECK-NEXT:    ldp q4, q5, [x1]
-; CHECK-NEXT:    ldp q6, q7, [x1, #32]
-; CHECK-NEXT:    fcmne p1.s, p0/z, z3.s, #0.0
-; CHECK-NEXT:    fcmne p2.s, p0/z, z2.s, #0.0
-; CHECK-NEXT:    fcmne p3.s, p0/z, z1.s, #0.0
-; CHECK-NEXT:    fcmne p4.s, p0/z, z0.s, #0.0
-; CHECK-NEXT:    fcmne p5.s, p0/z, z7.s, #0.0
-; CHECK-NEXT:    fcmne p6.s, p0/z, z6.s, #0.0
-; CHECK-NEXT:    fcmne p7.s, p0/z, z5.s, #0.0
-; CHECK-NEXT:    fcmne p0.s, p0/z, z4.s, #0.0
+; CHECK-NEXT:    ldp q2, q3, [x0]
+; CHECK-NEXT:    ldp q4, q5, [x1, #32]
+; CHECK-NEXT:    fcmne p1.s, p0/z, z0.s, #0.0
+; CHECK-NEXT:    fcmne p2.s, p0/z, z1.s, #0.0
+; CHECK-NEXT:    ldp q0, q1, [x1]
+; CHECK-NEXT:    fcmne p3.s, p0/z, z3.s, #0.0
+; CHECK-NEXT:    fcmne p4.s, p0/z, z2.s, #0.0
+; CHECK-NEXT:    fcmne p5.s, p0/z, z5.s, #0.0
+; CHECK-NEXT:    fcmne p6.s, p0/z, z4.s, #0.0
+; CHECK-NEXT:    fcmne p7.s, p0/z, z1.s, #0.0
+; CHECK-NEXT:    fcmne p0.s, p0/z, z0.s, #0.0
 ; CHECK-NEXT:    mov z0.s, p1/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z1.s, p2/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z2.s, p3/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z3.s, p4/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z4.s, p5/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z5.s, p6/z, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    mov z6.s, p7/z, #-1 // =0xffffffffffffffff
 ; CHECK-NEXT:    mov z7.s, p0/z, #-1 // =0xffffffffffffffff
-; CHECK-NEXT:    ptrue p1.h, vl4
-; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
 ; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
+; CHECK-NEXT:    splice z1.h, p0, z1.h, z0.h
+; CHECK-NEXT:    splice z3.h, p0, z3.h, z2.h
+; CHECK-NEXT:    splice z5.h, p0, z5.h, z4.h
+; CHECK-NEXT:    splice z7.h, p0, z7.h, z6.h
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    splice z1.h, p1, z1.h, z0.h
-; CHECK-NEXT:    splice z3.h, p1, z3.h, z2.h
-; CHECK-NEXT:    splice z5.h, p1, z5.h, z4.h
-; CHECK-NEXT:    splice z7.h, p1, z7.h, z6.h
 ; CHECK-NEXT:    uzp1 z0.b, z1.b, z1.b
 ; CHECK-NEXT:    uzp1 z1.b, z3.b, z3.b
 ; CHECK-NEXT:    uzp1 z2.b, z5.b, z5.b
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-rev.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-rev.ll
index f686efff67b66..bfa931044bc53 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-rev.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-rev.ll
@@ -49,8 +49,8 @@ define <16 x i8> @bitreverse_v16i8(<16 x i8> %op) {
 define void @bitreverse_v32i8(ptr %a) {
 ; CHECK-LABEL: bitreverse_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    rbit z0.b, p0/m, z0.b
 ; CHECK-NEXT:    rbit z1.b, p0/m, z1.b
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -101,8 +101,8 @@ define <8 x i16> @bitreverse_v8i16(<8 x i16> %op) {
 define void @bitreverse_v16i16(ptr %a) {
 ; CHECK-LABEL: bitreverse_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    rbit z0.h, p0/m, z0.h
 ; CHECK-NEXT:    rbit z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -140,8 +140,8 @@ define <4 x i32> @bitreverse_v4i32(<4 x i32> %op) {
 define void @bitreverse_v8i32(ptr %a) {
 ; CHECK-LABEL: bitreverse_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    rbit z0.s, p0/m, z0.s
 ; CHECK-NEXT:    rbit z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -179,8 +179,8 @@ define <2 x i64> @bitreverse_v2i64(<2 x i64> %op) {
 define void @bitreverse_v4i64(ptr %a) {
 ; CHECK-LABEL: bitreverse_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    rbit z0.d, p0/m, z0.d
 ; CHECK-NEXT:    rbit z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -235,8 +235,8 @@ define <8 x i16> @bswap_v8i16(<8 x i16> %op) {
 define void @bswap_v16i16(ptr %a) {
 ; CHECK-LABEL: bswap_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    revb z0.h, p0/m, z0.h
 ; CHECK-NEXT:    revb z1.h, p0/m, z1.h
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -274,8 +274,8 @@ define <4 x i32> @bswap_v4i32(<4 x i32> %op) {
 define void @bswap_v8i32(ptr %a) {
 ; CHECK-LABEL: bswap_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    revb z0.s, p0/m, z0.s
 ; CHECK-NEXT:    revb z1.s, p0/m, z1.s
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -313,8 +313,8 @@ define <2 x i64> @bswap_v2i64(<2 x i64> %op) {
 define void @bswap_v4i64(ptr %a) {
 ; CHECK-LABEL: bswap_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    revb z0.d, p0/m, z0.d
 ; CHECK-NEXT:    revb z1.d, p0/m, z1.d
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-sdiv-pow2.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-sdiv-pow2.ll
index 76bb465774d5b..9dd42e7831e0d 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-sdiv-pow2.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-sdiv-pow2.ll
@@ -45,8 +45,8 @@ define <16 x i8> @sdiv_v16i8(<16 x i8> %op1) {
 define void @sdiv_v32i8(ptr %a) {
 ; CHECK-LABEL: sdiv_v32i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.b, vl16
 ; CHECK-NEXT:    asrd z0.b, p0/m, z0.b, #5
 ; CHECK-NEXT:    asrd z1.b, p0/m, z1.b, #5
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -97,8 +97,8 @@ define <8 x i16> @sdiv_v8i16(<8 x i16> %op1) {
 define void @sdiv_v16i16(ptr %a) {
 ; CHECK-LABEL: sdiv_v16i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.h, vl8
 ; CHECK-NEXT:    asrd z0.h, p0/m, z0.h, #5
 ; CHECK-NEXT:    asrd z1.h, p0/m, z1.h, #5
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -136,8 +136,8 @@ define <4 x i32> @sdiv_v4i32(<4 x i32> %op1) {
 define void @sdiv_v8i32(ptr %a) {
 ; CHECK-LABEL: sdiv_v8i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    asrd z0.s, p0/m, z0.s, #5
 ; CHECK-NEXT:    asrd z1.s, p0/m, z1.s, #5
 ; CHECK-NEXT:    stp q0, q1, [x0]
@@ -176,8 +176,8 @@ define <2 x i64> @sdiv_v2i64(<2 x i64> %op1) {
 define void @sdiv_v4i64(ptr %a) {
 ; CHECK-LABEL: sdiv_v4i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    ldp q0, q1, [x0]
+; CHECK-NEXT:    ptrue p0.d, vl2
 ; CHECK-NEXT:    asrd z0.d, p0/m, z0.d, #5
 ; CHECK-NEXT:    asrd z1.d, p0/m, z1.d, #5
 ; CHECK-NEXT:    stp q0, q1, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-shuffle.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-shuffle.ll
index ff1f8699b91af..ad0d4ef0afef3 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-shuffle.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-shuffle.ll
@@ -37,9 +37,9 @@ define void @interleave_store_without_splat(ptr %a, <4 x i32> %v1, <4 x i32> %v2
 define void @interleave_store_legalization(ptr %a, <8 x i32> %v1, <8 x i32> %v2) {
 ; CHECK-LABEL: interleave_store_legalization:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    mov z5.d, z2.d
 ; CHECK-NEXT:    // kill: def $q3 killed $q3 def $z2_z3
+; CHECK-NEXT:    ptrue p0.s, vl4
 ; CHECK-NEXT:    mov x8, #8 // =0x8
 ; CHECK-NEXT:    mov z4.d, z0.d
 ; CHECK-NEXT:    mov z2.d, z1.d
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-stores.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-stores.ll
index 367ccbeeea81e..06709ca3685c8 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-stores.ll
@@ -8,8 +8,8 @@ target triple = "aarch64-unknown-linux-gnu"
 define void @store_v4i8(ptr %a) {
 ; CHECK-LABEL: store_v4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    mov z0.h, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    st1b { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   store <4 x i8> zeroinitializer, ptr %a
@@ -49,8 +49,8 @@ define void @store_v32i8(ptr %a) {
 define void @store_v2i16(ptr %a) {
 ; CHECK-LABEL: store_v2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    mov z0.s, #0 // =0x0
+; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   store <2 x i16> zeroinitializer, ptr %a
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-trunc.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-trunc.ll
index 4fef678314019..70219dd30f769 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-trunc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-trunc.ll
@@ -87,51 +87,51 @@ define void @trunc_v64i16_v64i8(ptr %in, ptr %out) nounwind {
 define void @trunc_v128i16_v128i8(ptr %in, ptr %out) nounwind {
 ; CHECK-LABEL: trunc_v128i16_v128i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp q0, q1, [x0, #192]
+; CHECK-NEXT:    ldp q2, q3, [x0, #192]
 ; CHECK-NEXT:    ptrue p0.b, vl8
-; CHECK-NEXT:    ldp q6, q7, [x0, #224]
-; CHECK-NEXT:    ldp q2, q3, [x0, #32]
-; CHECK-NEXT:    uzp1 z1.b, z1.b, z1.b
-; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
-; CHECK-NEXT:    uzp1 z7.b, z7.b, z7.b
-; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
-; CHECK-NEXT:    ldp q4, q5, [x0]
+; CHECK-NEXT:    ldp q6, q7, [x0, #64]
+; CHECK-NEXT:    ldp q16, q17, [x0, #224]
 ; CHECK-NEXT:    uzp1 z3.b, z3.b, z3.b
-; CHECK-NEXT:    ldp q16, q17, [x0, #64]
 ; CHECK-NEXT:    uzp1 z2.b, z2.b, z2.b
-; CHECK-NEXT:    ldp q18, q19, [x0, #128]
-; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
 ; CHECK-NEXT:    ldp q20, q21, [x0, #160]
-; CHECK-NEXT:    splice z6.b, p0, z6.b, z7.b
-; CHECK-NEXT:    ldp q22, q23, [x0, #96]
-; CHECK-NEXT:    uzp1 z1.b, z17.b, z17.b
-; CHECK-NEXT:    uzp1 z19.b, z19.b, z19.b
-; CHECK-NEXT:    uzp1 z18.b, z18.b, z18.b
+; CHECK-NEXT:    uzp1 z7.b, z7.b, z7.b
+; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    uzp1 z17.b, z17.b, z17.b
+; CHECK-NEXT:    ldp q4, q5, [x0, #96]
 ; CHECK-NEXT:    uzp1 z16.b, z16.b, z16.b
-; CHECK-NEXT:    uzp1 z21.b, z21.b, z21.b
+; CHECK-NEXT:    ldp q18, q19, [x0, #128]
+; CHECK-NEXT:    splice z2.b, p0, z2.b, z3.b
+; CHECK-NEXT:    uzp1 z3.b, z21.b, z21.b
 ; CHECK-NEXT:    uzp1 z20.b, z20.b, z20.b
-; CHECK-NEXT:    uzp1 z5.b, z5.b, z5.b
-; CHECK-NEXT:    uzp1 z7.b, z23.b, z23.b
-; CHECK-NEXT:    uzp1 z17.b, z22.b, z22.b
+; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
+; CHECK-NEXT:    ldp q21, q22, [x0]
+; CHECK-NEXT:    splice z16.b, p0, z16.b, z17.b
+; CHECK-NEXT:    uzp1 z19.b, z19.b, z19.b
+; CHECK-NEXT:    uzp1 z18.b, z18.b, z18.b
 ; CHECK-NEXT:    uzp1 z4.b, z4.b, z4.b
-; CHECK-NEXT:    splice z2.b, p0, z2.b, z3.b
-; CHECK-NEXT:    add z0.b, z0.b, z0.b
+; CHECK-NEXT:    splice z20.b, p0, z20.b, z3.b
+; CHECK-NEXT:    uzp1 z3.b, z5.b, z5.b
+; CHECK-NEXT:    splice z6.b, p0, z6.b, z7.b
+; CHECK-NEXT:    uzp1 z5.b, z22.b, z22.b
+; CHECK-NEXT:    uzp1 z7.b, z21.b, z21.b
+; CHECK-NEXT:    uzp1 z1.b, z1.b, z1.b
+; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
 ; CHECK-NEXT:    splice z18.b, p0, z18.b, z19.b
-; CHECK-NEXT:    splice z16.b, p0, z16.b, z1.b
-; CHECK-NEXT:    add z1.b, z6.b, z6.b
-; CHECK-NEXT:    splice z20.b, p0, z20.b, z21.b
-; CHECK-NEXT:    splice z17.b, p0, z17.b, z7.b
-; CHECK-NEXT:    splice z4.b, p0, z4.b, z5.b
-; CHECK-NEXT:    stp q0, q1, [x1, #96]
 ; CHECK-NEXT:    add z2.b, z2.b, z2.b
+; CHECK-NEXT:    splice z4.b, p0, z4.b, z3.b
+; CHECK-NEXT:    add z3.b, z16.b, z16.b
+; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
+; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
+; CHECK-NEXT:    add z1.b, z20.b, z20.b
 ; CHECK-NEXT:    add z5.b, z18.b, z18.b
-; CHECK-NEXT:    add z0.b, z16.b, z16.b
-; CHECK-NEXT:    add z3.b, z20.b, z20.b
-; CHECK-NEXT:    add z1.b, z17.b, z17.b
-; CHECK-NEXT:    add z4.b, z4.b, z4.b
-; CHECK-NEXT:    stp q5, q3, [x1, #64]
-; CHECK-NEXT:    stp q4, q2, [x1]
-; CHECK-NEXT:    stp q0, q1, [x1, #32]
+; CHECK-NEXT:    stp q2, q3, [x1, #96]
+; CHECK-NEXT:    add z2.b, z6.b, z6.b
+; CHECK-NEXT:    add z3.b, z4.b, z4.b
+; CHECK-NEXT:    add z4.b, z7.b, z7.b
+; CHECK-NEXT:    add z0.b, z0.b, z0.b
+; CHECK-NEXT:    stp q5, q1, [x1, #64]
+; CHECK-NEXT:    stp q2, q3, [x1, #32]
+; CHECK-NEXT:    stp q4, q0, [x1]
 ; CHECK-NEXT:    ret
   %a = load <128 x i16>, ptr %in
   %b = trunc <128 x i16> %a to <128 x i8>
@@ -226,55 +226,55 @@ define void @trunc_v32i32_v32i8(ptr %in, ptr %out) nounwind {
 define void @trunc_v64i32_v64i8(ptr %in, ptr %out) nounwind {
 ; CHECK-LABEL: trunc_v64i32_v64i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp q0, q1, [x0, #64]
-; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    ldp q2, q3, [x0, #160]
-; CHECK-NEXT:    ptrue p1.b, vl8
-; CHECK-NEXT:    ldp q4, q5, [x0, #96]
-; CHECK-NEXT:    ldp q6, q7, [x0]
-; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
-; CHECK-NEXT:    ldp q16, q17, [x0, #128]
+; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    ldp q4, q5, [x0, #128]
+; CHECK-NEXT:    ldp q0, q1, [x0, #64]
+; CHECK-NEXT:    ldp q6, q7, [x0, #96]
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
-; CHECK-NEXT:    ldp q18, q19, [x0, #192]
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
-; CHECK-NEXT:    ldp q20, q21, [x0, #224]
+; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
+; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
+; CHECK-NEXT:    ldp q16, q17, [x0]
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
+; CHECK-NEXT:    ldp q18, q19, [x0, #192]
 ; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
+; CHECK-NEXT:    ldp q20, q21, [x0, #224]
+; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
 ; CHECK-NEXT:    ldp q22, q23, [x0, #32]
-; CHECK-NEXT:    uzp1 z17.h, z17.h, z17.h
-; CHECK-NEXT:    uzp1 z16.h, z16.h, z16.h
+; CHECK-NEXT:    splice z4.h, p0, z4.h, z5.h
 ; CHECK-NEXT:    uzp1 z19.h, z19.h, z19.h
 ; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
-; CHECK-NEXT:    uzp1 z21.h, z21.h, z21.h
-; CHECK-NEXT:    uzp1 z20.h, z20.h, z20.h
+; CHECK-NEXT:    uzp1 z17.h, z17.h, z17.h
+; CHECK-NEXT:    uzp1 z3.h, z21.h, z21.h
+; CHECK-NEXT:    uzp1 z5.h, z20.h, z20.h
+; CHECK-NEXT:    uzp1 z16.h, z16.h, z16.h
+; CHECK-NEXT:    uzp1 z20.h, z23.h, z23.h
+; CHECK-NEXT:    uzp1 z21.h, z22.h, z22.h
 ; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
-; CHECK-NEXT:    uzp1 z23.h, z23.h, z23.h
-; CHECK-NEXT:    uzp1 z22.h, z22.h, z22.h
-; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
-; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
-; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
-; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
 ; CHECK-NEXT:    splice z18.h, p0, z18.h, z19.h
-; CHECK-NEXT:    splice z20.h, p0, z20.h, z21.h
+; CHECK-NEXT:    splice z5.h, p0, z5.h, z3.h
+; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
+; CHECK-NEXT:    splice z21.h, p0, z21.h, z20.h
 ; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
-; CHECK-NEXT:    splice z22.h, p0, z22.h, z23.h
-; CHECK-NEXT:    splice z4.h, p0, z4.h, z5.h
 ; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    uzp1 z1.b, z2.b, z2.b
-; CHECK-NEXT:    uzp1 z2.b, z16.b, z16.b
-; CHECK-NEXT:    uzp1 z5.b, z18.b, z18.b
-; CHECK-NEXT:    uzp1 z3.b, z20.b, z20.b
-; CHECK-NEXT:    uzp1 z6.b, z6.b, z6.b
-; CHECK-NEXT:    uzp1 z7.b, z22.b, z22.b
-; CHECK-NEXT:    uzp1 z4.b, z4.b, z4.b
+; CHECK-NEXT:    uzp1 z2.b, z4.b, z4.b
+; CHECK-NEXT:    ptrue p0.b, vl8
+; CHECK-NEXT:    uzp1 z4.b, z18.b, z18.b
+; CHECK-NEXT:    uzp1 z3.b, z5.b, z5.b
+; CHECK-NEXT:    uzp1 z7.b, z16.b, z16.b
+; CHECK-NEXT:    uzp1 z5.b, z21.b, z21.b
+; CHECK-NEXT:    splice z2.b, p0, z2.b, z1.b
+; CHECK-NEXT:    uzp1 z1.b, z6.b, z6.b
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
-; CHECK-NEXT:    splice z2.b, p1, z2.b, z1.b
-; CHECK-NEXT:    splice z5.b, p1, z5.b, z3.b
-; CHECK-NEXT:    splice z6.b, p1, z6.b, z7.b
-; CHECK-NEXT:    splice z0.b, p1, z0.b, z4.b
+; CHECK-NEXT:    splice z4.b, p0, z4.b, z3.b
+; CHECK-NEXT:    splice z7.b, p0, z7.b, z5.b
+; CHECK-NEXT:    splice z0.b, p0, z0.b, z1.b
 ; CHECK-NEXT:    add z1.b, z2.b, z2.b
-; CHECK-NEXT:    add z2.b, z5.b, z5.b
-; CHECK-NEXT:    add z3.b, z6.b, z6.b
+; CHECK-NEXT:    add z2.b, z4.b, z4.b
+; CHECK-NEXT:    add z3.b, z7.b, z7.b
 ; CHECK-NEXT:    add z0.b, z0.b, z0.b
 ; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q3, q0, [x1]
@@ -368,51 +368,51 @@ define void @trunc_v32i32_v32i16(ptr %in, ptr %out) nounwind {
 define void @trunc_v64i32_v64i16(ptr %in, ptr %out) nounwind {
 ; CHECK-LABEL: trunc_v64i32_v64i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp q0, q1, [x0, #192]
+; CHECK-NEXT:    ldp q2, q3, [x0, #192]
 ; CHECK-NEXT:    ptrue p0.h, vl4
-; CHECK-NEXT:    ldp q6, q7, [x0, #224]
-; CHECK-NEXT:    ldp q2, q3, [x0, #32]
-; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
-; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
-; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
-; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
-; CHECK-NEXT:    ldp q4, q5, [x0]
+; CHECK-NEXT:    ldp q6, q7, [x0, #64]
+; CHECK-NEXT:    ldp q16, q17, [x0, #224]
 ; CHECK-NEXT:    uzp1 z3.h, z3.h, z3.h
-; CHECK-NEXT:    ldp q16, q17, [x0, #64]
 ; CHECK-NEXT:    uzp1 z2.h, z2.h, z2.h
-; CHECK-NEXT:    ldp q18, q19, [x0, #128]
-; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    ldp q20, q21, [x0, #160]
-; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
-; CHECK-NEXT:    ldp q22, q23, [x0, #96]
-; CHECK-NEXT:    uzp1 z1.h, z17.h, z17.h
-; CHECK-NEXT:    uzp1 z19.h, z19.h, z19.h
-; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
+; CHECK-NEXT:    uzp1 z7.h, z7.h, z7.h
+; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    uzp1 z17.h, z17.h, z17.h
+; CHECK-NEXT:    ldp q4, q5, [x0, #96]
 ; CHECK-NEXT:    uzp1 z16.h, z16.h, z16.h
-; CHECK-NEXT:    uzp1 z21.h, z21.h, z21.h
+; CHECK-NEXT:    ldp q18, q19, [x0, #128]
+; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
+; CHECK-NEXT:    uzp1 z3.h, z21.h, z21.h
 ; CHECK-NEXT:    uzp1 z20.h, z20.h, z20.h
-; CHECK-NEXT:    uzp1 z5.h, z5.h, z5.h
-; CHECK-NEXT:    uzp1 z7.h, z23.h, z23.h
-; CHECK-NEXT:    uzp1 z17.h, z22.h, z22.h
+; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
+; CHECK-NEXT:    ldp q21, q22, [x0]
+; CHECK-NEXT:    splice z16.h, p0, z16.h, z17.h
+; CHECK-NEXT:    uzp1 z19.h, z19.h, z19.h
+; CHECK-NEXT:    uzp1 z18.h, z18.h, z18.h
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
-; CHECK-NEXT:    splice z2.h, p0, z2.h, z3.h
-; CHECK-NEXT:    add z0.h, z0.h, z0.h
+; CHECK-NEXT:    splice z20.h, p0, z20.h, z3.h
+; CHECK-NEXT:    uzp1 z3.h, z5.h, z5.h
+; CHECK-NEXT:    splice z6.h, p0, z6.h, z7.h
+; CHECK-NEXT:    uzp1 z5.h, z22.h, z22.h
+; CHECK-NEXT:    uzp1 z7.h, z21.h, z21.h
+; CHECK-NEXT:    uzp1 z1.h, z1.h, z1.h
+; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    splice z18.h, p0, z18.h, z19.h
-; CHECK-NEXT:    splice z16.h, p0, z16.h, z1.h
-; CHECK-NEXT:    add z1.h, z6.h, z6.h
-; CHECK-NEXT:    splice z20.h, p0, z20.h, z21.h
-; CHECK-NEXT:    splice z17.h, p0, z17.h, z7.h
-; CHECK-NEXT:    splice z4.h, p0, z4.h, z5.h
-; CHECK-NEXT:    stp q0, q1, [x1, #96]
 ; CHECK-NEXT:    add z2.h, z2.h, z2.h
+; CHECK-NEXT:    splice z4.h, p0, z4.h, z3.h
+; CHECK-NEXT:    add z3.h, z16.h, z16.h
+; CHECK-NEXT:    splice z7.h, p0, z7.h, z5.h
+; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
+; CHECK-NEXT:    add z1.h, z20.h, z20.h
 ; CHECK-NEXT:    add z5.h, z18.h, z18.h
-; CHECK-NEXT:    add z0.h, z16.h, z16.h
-; CHECK-NEXT:    add z3.h, z20.h, z20.h
-; CHECK-NEXT:    add z1.h, z17.h, z17.h
-; CHECK-NEXT:    add z4.h, z4.h, z4.h
-; CHECK-NEXT:    stp q5, q3, [x1, #64]
-; CHECK-NEXT:    stp q4, q2, [x1]
-; CHECK-NEXT:    stp q0, q1, [x1, #32]
+; CHECK-NEXT:    stp q2, q3, [x1, #96]
+; CHECK-NEXT:    add z2.h, z6.h, z6.h
+; CHECK-NEXT:    add z3.h, z4.h, z4.h
+; CHECK-NEXT:    add z4.h, z7.h, z7.h
+; CHECK-NEXT:    add z0.h, z0.h, z0.h
+; CHECK-NEXT:    stp q5, q1, [x1, #64]
+; CHECK-NEXT:    stp q2, q3, [x1, #32]
+; CHECK-NEXT:    stp q4, q0, [x1]
 ; CHECK-NEXT:    ret
   %a = load <64 x i32>, ptr %in
   %b = trunc <64 x i32> %a to <64 x i16>
@@ -535,19 +535,19 @@ define void @trunc_v32i64_v32i8(ptr %in, ptr %out) nounwind {
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
 ; CHECK-NEXT:    splice z16.s, p0, z16.s, z17.s
-; CHECK-NEXT:    splice z18.s, p0, z18.s, z19.s
 ; CHECK-NEXT:    splice z20.s, p0, z20.s, z21.s
-; CHECK-NEXT:    splice z6.s, p0, z6.s, z7.s
+; CHECK-NEXT:    splice z18.s, p0, z18.s, z19.s
 ; CHECK-NEXT:    splice z22.s, p0, z22.s, z23.s
+; CHECK-NEXT:    splice z6.s, p0, z6.s, z7.s
 ; CHECK-NEXT:    splice z4.s, p0, z4.s, z5.s
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ptrue p0.h, vl4
 ; CHECK-NEXT:    uzp1 z1.h, z2.h, z2.h
 ; CHECK-NEXT:    uzp1 z2.h, z16.h, z16.h
-; CHECK-NEXT:    uzp1 z5.h, z18.h, z18.h
 ; CHECK-NEXT:    uzp1 z3.h, z20.h, z20.h
-; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
+; CHECK-NEXT:    uzp1 z5.h, z18.h, z18.h
 ; CHECK-NEXT:    uzp1 z7.h, z22.h, z22.h
+; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    splice z2.h, p0, z2.h, z1.h
@@ -658,55 +658,55 @@ define void @trunc_v16i64_v16i16(ptr %in, ptr %out) nounwind {
 define void @trunc_v32i64_v32i16(ptr %in, ptr %out) nounwind {
 ; CHECK-LABEL: trunc_v32i64_v32i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp q0, q1, [x0, #64]
-; CHECK-NEXT:    ptrue p0.s, vl2
 ; CHECK-NEXT:    ldp q2, q3, [x0, #160]
-; CHECK-NEXT:    ptrue p1.h, vl4
-; CHECK-NEXT:    ldp q4, q5, [x0, #96]
-; CHECK-NEXT:    ldp q6, q7, [x0]
-; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
-; CHECK-NEXT:    ldp q16, q17, [x0, #128]
+; CHECK-NEXT:    ptrue p0.s, vl2
+; CHECK-NEXT:    ldp q4, q5, [x0, #128]
+; CHECK-NEXT:    ldp q0, q1, [x0, #64]
+; CHECK-NEXT:    ldp q6, q7, [x0, #96]
 ; CHECK-NEXT:    uzp1 z3.s, z3.s, z3.s
-; CHECK-NEXT:    ldp q18, q19, [x0, #192]
 ; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
-; CHECK-NEXT:    ldp q20, q21, [x0, #224]
+; CHECK-NEXT:    uzp1 z5.s, z5.s, z5.s
+; CHECK-NEXT:    uzp1 z4.s, z4.s, z4.s
+; CHECK-NEXT:    ldp q16, q17, [x0]
+; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
+; CHECK-NEXT:    ldp q18, q19, [x0, #192]
 ; CHECK-NEXT:    uzp1 z7.s, z7.s, z7.s
+; CHECK-NEXT:    ldp q20, q21, [x0, #224]
+; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
 ; CHECK-NEXT:    ldp q22, q23, [x0, #32]
-; CHECK-NEXT:    uzp1 z17.s, z17.s, z17.s
-; CHECK-NEXT:    uzp1 z16.s, z16.s, z16.s
+; CHECK-NEXT:    splice z4.s, p0, z4.s, z5.s
 ; CHECK-NEXT:    uzp1 z19.s, z19.s, z19.s
 ; CHECK-NEXT:    uzp1 z18.s, z18.s, z18.s
-; CHECK-NEXT:    uzp1 z21.s, z21.s, z21.s
-; CHECK-NEXT:    uzp1 z20.s, z20.s, z20.s
+; CHECK-NEXT:    uzp1 z17.s, z17.s, z17.s
+; CHECK-NEXT:    uzp1 z3.s, z21.s, z21.s
+; CHECK-NEXT:    uzp1 z5.s, z20.s, z20.s
+; CHECK-NEXT:    uzp1 z16.s, z16.s, z16.s
+; CHECK-NEXT:    uzp1 z20.s, z23.s, z23.s
+; CHECK-NEXT:    uzp1 z21.s, z22.s, z22.s
 ; CHECK-NEXT:    uzp1 z6.s, z6.s, z6.s
-; CHECK-NEXT:    uzp1 z23.s, z23.s, z23.s
-; CHECK-NEXT:    uzp1 z22.s, z22.s, z22.s
-; CHECK-NEXT:    uzp1 z5.s, z5.s, z5.s
-; CHECK-NEXT:    uzp1 z4.s, z4.s, z4.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
-; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
-; CHECK-NEXT:    splice z16.s, p0, z16.s, z17.s
 ; CHECK-NEXT:    splice z18.s, p0, z18.s, z19.s
-; CHECK-NEXT:    splice z20.s, p0, z20.s, z21.s
+; CHECK-NEXT:    splice z5.s, p0, z5.s, z3.s
+; CHECK-NEXT:    splice z16.s, p0, z16.s, z17.s
+; CHECK-NEXT:    splice z21.s, p0, z21.s, z20.s
 ; CHECK-NEXT:    splice z6.s, p0, z6.s, z7.s
-; CHECK-NEXT:    splice z22.s, p0, z22.s, z23.s
-; CHECK-NEXT:    splice z4.s, p0, z4.s, z5.s
 ; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    uzp1 z1.h, z2.h, z2.h
-; CHECK-NEXT:    uzp1 z2.h, z16.h, z16.h
-; CHECK-NEXT:    uzp1 z5.h, z18.h, z18.h
-; CHECK-NEXT:    uzp1 z3.h, z20.h, z20.h
-; CHECK-NEXT:    uzp1 z6.h, z6.h, z6.h
-; CHECK-NEXT:    uzp1 z7.h, z22.h, z22.h
-; CHECK-NEXT:    uzp1 z4.h, z4.h, z4.h
+; CHECK-NEXT:    uzp1 z2.h, z4.h, z4.h
+; CHECK-NEXT:    ptrue p0.h, vl4
+; CHECK-NEXT:    uzp1 z4.h, z18.h, z18.h
+; CHECK-NEXT:    uzp1 z3.h, z5.h, z5.h
+; CHECK-NEXT:    uzp1 z7.h, z16.h, z16.h
+; CHECK-NEXT:    uzp1 z5.h, z21.h, z21.h
+; CHECK-NEXT:    splice z2.h, p0, z2.h, z1.h
+; CHECK-NEXT:    uzp1 z1.h, z6.h, z6.h
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
-; CHECK-NEXT:    splice z2.h, p1, z2.h, z1.h
-; CHECK-NEXT:    splice z5.h, p1, z5.h, z3.h
-; CHECK-NEXT:    splice z6.h, p1, z6.h, z7.h
-; CHECK-NEXT:    splice z0.h, p1, z0.h, z4.h
+; CHECK-NEXT:    splice z4.h, p0, z4.h, z3.h
+; CHECK-NEXT:    splice z7.h, p0, z7.h, z5.h
+; CHECK-NEXT:    splice z0.h, p0, z0.h, z1.h
 ; CHECK-NEXT:    add z1.h, z2.h, z2.h
-; CHECK-NEXT:    add z2.h, z5.h, z5.h
-; CHECK-NEXT:    add z3.h, z6.h, z6.h
+; CHECK-NEXT:    add z2.h, z4.h, z4.h
+; CHECK-NEXT:    add z3.h, z7.h, z7.h
 ; CHECK-NEXT:    add z0.h, z0.h, z0.h
 ; CHECK-NEXT:    stp q1, q2, [x1, #32]
 ; CHECK-NEXT:    stp q3, q0, [x1]
@@ -800,51 +800,51 @@ define void @trunc_v16i64_v16i32(ptr %in, ptr %out) nounwind {
 define void @trunc_v32i64_v32i32(ptr %in, ptr %out) nounwind {
 ; CHECK-LABEL: trunc_v32i64_v32i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ldp q0, q1, [x0, #192]
+; CHECK-NEXT:    ldp q2, q3, [x0, #192]
 ; CHECK-NEXT:    ptrue p0.s, vl2
-; CHECK-NEXT:    ldp q6, q7, [x0, #224]
-; CHECK-NEXT:    ldp q2, q3, [x0, #32]
-; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
-; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
-; CHECK-NEXT:    uzp1 z7.s, z7.s, z7.s
-; CHECK-NEXT:    uzp1 z6.s, z6.s, z6.s
-; CHECK-NEXT:    ldp q4, q5, [x0]
+; CHECK-NEXT:    ldp q6, q7, [x0, #64]
+; CHECK-NEXT:    ldp q16, q17, [x0, #224]
 ; CHECK-NEXT:    uzp1 z3.s, z3.s, z3.s
-; CHECK-NEXT:    ldp q16, q17, [x0, #64]
 ; CHECK-NEXT:    uzp1 z2.s, z2.s, z2.s
-; CHECK-NEXT:    ldp q18, q19, [x0, #128]
-; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
 ; CHECK-NEXT:    ldp q20, q21, [x0, #160]
-; CHECK-NEXT:    splice z6.s, p0, z6.s, z7.s
-; CHECK-NEXT:    ldp q22, q23, [x0, #96]
-; CHECK-NEXT:    uzp1 z1.s, z17.s, z17.s
-; CHECK-NEXT:    uzp1 z19.s, z19.s, z19.s
-; CHECK-NEXT:    uzp1 z18.s, z18.s, z18.s
+; CHECK-NEXT:    uzp1 z7.s, z7.s, z7.s
+; CHECK-NEXT:    ldp q0, q1, [x0, #32]
+; CHECK-NEXT:    uzp1 z17.s, z17.s, z17.s
+; CHECK-NEXT:    ldp q4, q5, [x0, #96]
 ; CHECK-NEXT:    uzp1 z16.s, z16.s, z16.s
-; CHECK-NEXT:    uzp1 z21.s, z21.s, z21.s
+; CHECK-NEXT:    ldp q18, q19, [x0, #128]
+; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
+; CHECK-NEXT:    uzp1 z3.s, z21.s, z21.s
 ; CHECK-NEXT:    uzp1 z20.s, z20.s, z20.s
-; CHECK-NEXT:    uzp1 z5.s, z5.s, z5.s
-; CHECK-NEXT:    uzp1 z7.s, z23.s, z23.s
-; CHECK-NEXT:    uzp1 z17.s, z22.s, z22.s
+; CHECK-NEXT:    uzp1 z6.s, z6.s, z6.s
+; CHECK-NEXT:    ldp q21, q22, [x0]
+; CHECK-NEXT:    splice z16.s, p0, z16.s, z17.s
+; CHECK-NEXT:    uzp1 z19.s, z19.s, z19.s
+; CHECK-NEXT:    uzp1 z18.s, z18.s, z18.s
 ; CHECK-NEXT:    uzp1 z4.s, z4.s, z4.s
-; CHECK-NEXT:    splice z2.s, p0, z2.s, z3.s
-; CHECK-NEXT:    add z0.s, z0.s, z0.s
+; CHECK-NEXT:    splice z20.s, p0, z20.s, z3.s
+; CHECK-NEXT:    uzp1 z3.s, z5.s, z5.s
+; CHECK-NEXT:    splice z6.s, p0, z6.s, z7.s
+; CHECK-NEXT:    uzp1 z5.s, z22.s, z22.s
+; CHECK-NEXT:    uzp1 z7.s, z21.s, z21.s
+; CHECK-NEXT:    uzp1 z1.s, z1.s, z1.s
+; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    splice z18.s, p0, z18.s, z19.s
-; CHECK-NEXT:    splice z16.s, p0, z16.s, z1.s
-; CHECK-NEXT:    add z1.s, z6.s, z6.s
-; CHECK-NEXT:    splice z20.s, p0, z20.s, z21.s
-; CHECK-NEXT:    splice z17.s, p0, z17.s, z7.s
-; CHECK-NEXT:    splice z4.s, p0, z4.s, z5.s
-; CHECK-NEXT:    stp q0, q1, [x1, #96]
 ; CHECK-NEXT:    add z2.s, z2.s, z2.s
+; CHECK-NEXT:    splice z4.s, p0, z4.s, z3.s
+; CHECK-NEXT:    add z3.s, z16.s, z16.s
+; CHECK-NEXT:    splice z7.s, p0, z7.s, z5.s
+; CHECK-NEXT:    splice z0.s, p0, z0.s, z1.s
+; CHECK-NEXT:    add z1.s, z20.s, z20.s
 ; CHECK-NEXT:    add z5.s, z18.s, z18.s
-; CHECK-NEXT:    add z0.s, z16.s, z16.s
-; CHECK-NEXT:    add z3.s, z20.s, z20.s
-; CHECK-NEXT:    add z1.s, z17.s, z17.s
-; CHECK-NEXT:    add z4.s, z4.s, z4.s
-; CHECK-NEXT:    stp q5, q3, [x1, #64]
-; CHECK-NEXT:    stp q4, q2, [x1]
-; CHECK-NEXT:    stp q0, q1, [x1, #32]
+; CHECK-NEXT:    stp q2, q3, [x1, #96]
+; CHECK-NEXT:    add z2.s, z6.s, z6.s
+; CHECK-NEXT:    add z3.s, z4.s, z4.s
+; CHECK-NEXT:    add z4.s, z7.s, z7.s
+; CHECK-NEXT:    add z0.s, z0.s, z0.s
+; CHECK-NEXT:    stp q5, q1, [x1, #64]
+; CHECK-NEXT:    stp q2, q3, [x1, #32]
+; CHECK-NEXT:    stp q4, q0, [x1]
 ; CHECK-NEXT:    ret
   %a = load <32 x i64>, ptr %in
   %b = trunc <32 x i64> %a to <32 x i32>
diff --git a/llvm/test/CodeGen/AArch64/sve-trunc.ll b/llvm/test/CodeGen/AArch64/sve-trunc.ll
index 92dfc73961362..0ec6538947c73 100644
--- a/llvm/test/CodeGen/AArch64/sve-trunc.ll
+++ b/llvm/test/CodeGen/AArch64/sve-trunc.ll
@@ -61,8 +61,8 @@ entry:
 define <vscale x 2 x i1> @trunc_i64toi1(<vscale x 2 x i64> %in) {
 ; CHECK-LABEL: trunc_i64toi1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    ret
 entry:
@@ -73,9 +73,9 @@ entry:
 define <vscale x 4 x i1> @trunc_i64toi1_split(<vscale x 4 x i64> %in) {
 ; CHECK-LABEL: trunc_i64toi1_split:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0x1
 ; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
 ; CHECK-NEXT:    uzp1 p0.s, p0.s, p1.s
@@ -88,11 +88,11 @@ entry:
 define <vscale x 8 x i1> @trunc_i64toi1_split2(<vscale x 8 x i64> %in) {
 ; CHECK-LABEL: trunc_i64toi1_split2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z3.d, z3.d, #0x1
 ; CHECK-NEXT:    and z2.d, z2.d, #0x1
 ; CHECK-NEXT:    and z1.d, z1.d, #0x1
 ; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z3.d, #0
 ; CHECK-NEXT:    cmpne p2.d, p0/z, z2.d, #0
 ; CHECK-NEXT:    cmpne p3.d, p0/z, z1.d, #0
@@ -111,12 +111,12 @@ define <vscale x 16 x i1> @trunc_i64toi1_split3(<vscale x 16 x i64> %in) {
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p7, [sp, #4, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p6, [sp, #5, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p5, [sp, #6, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    str p4, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z7.d, z7.d, #0x1
 ; CHECK-NEXT:    and z6.d, z6.d, #0x1
 ; CHECK-NEXT:    and z5.d, z5.d, #0x1
@@ -125,23 +125,25 @@ define <vscale x 16 x i1> @trunc_i64toi1_split3(<vscale x 16 x i64> %in) {
 ; CHECK-NEXT:    and z2.d, z2.d, #0x1
 ; CHECK-NEXT:    and z1.d, z1.d, #0x1
 ; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z7.d, #0
 ; CHECK-NEXT:    cmpne p2.d, p0/z, z6.d, #0
 ; CHECK-NEXT:    cmpne p3.d, p0/z, z5.d, #0
 ; CHECK-NEXT:    cmpne p4.d, p0/z, z4.d, #0
 ; CHECK-NEXT:    cmpne p5.d, p0/z, z3.d, #0
 ; CHECK-NEXT:    cmpne p6.d, p0/z, z2.d, #0
-; CHECK-NEXT:    uzp1 p1.s, p2.s, p1.s
-; CHECK-NEXT:    cmpne p2.d, p0/z, z1.d, #0
+; CHECK-NEXT:    cmpne p7.d, p0/z, z1.d, #0
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z0.d, #0
-; CHECK-NEXT:    uzp1 p3.s, p4.s, p3.s
-; CHECK-NEXT:    uzp1 p4.s, p6.s, p5.s
+; CHECK-NEXT:    uzp1 p1.s, p2.s, p1.s
+; CHECK-NEXT:    uzp1 p2.s, p4.s, p3.s
+; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p3.s, p6.s, p5.s
 ; CHECK-NEXT:    ldr p6, [sp, #5, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p0.s, p0.s, p2.s
+; CHECK-NEXT:    uzp1 p0.s, p0.s, p7.s
+; CHECK-NEXT:    ldr p7, [sp, #4, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p1.h, p2.h, p1.h
 ; CHECK-NEXT:    ldr p5, [sp, #6, mul vl] // 2-byte Folded Reload
-; CHECK-NEXT:    uzp1 p1.h, p3.h, p1.h
-; CHECK-NEXT:    uzp1 p0.h, p0.h, p4.h
-; CHECK-NEXT:    ldr p4, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    uzp1 p0.h, p0.h, p3.h
 ; CHECK-NEXT:    uzp1 p0.b, p0.b, p1.b
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
@@ -155,8 +157,8 @@ entry:
 define <vscale x 4 x i1> @trunc_i32toi1(<vscale x 4 x i32> %in) {
 ; CHECK-LABEL: trunc_i32toi1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    ret
 entry:
@@ -167,8 +169,8 @@ entry:
 define <vscale x 8 x i1> @trunc_i16toi1(<vscale x 8 x i16> %in) {
 ; CHECK-LABEL: trunc_i16toi1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z0.h, z0.h, #0x1
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z0.h, #0
 ; CHECK-NEXT:    ret
 entry:
@@ -179,8 +181,8 @@ entry:
 define <vscale x 16 x i1> @trunc_i8toi1(<vscale x 16 x i8> %in) {
 ; CHECK-LABEL: trunc_i8toi1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    and z0.b, z0.b, #0x1
+; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    cmpne p0.b, p0/z, z0.b, #0
 ; CHECK-NEXT:    ret
 entry:
@@ -191,8 +193,8 @@ entry:
 define <vscale x 1 x i1> @trunc_nxv1i32_to_nxv1i1(<vscale x 1 x i32> %in) {
 ; CHECK-LABEL: trunc_nxv1i32_to_nxv1i1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z0.s, #0
 ; CHECK-NEXT:    punpklo p0.h, p0.b
 ; CHECK-NEXT:    punpklo p0.h, p0.b
@@ -204,8 +206,8 @@ define <vscale x 1 x i1> @trunc_nxv1i32_to_nxv1i1(<vscale x 1 x i32> %in) {
 define void @trunc_promoteIntRes(<vscale x 4 x i64> %0, ptr %ptr) {
 ; CHECK-LABEL: trunc_promoteIntRes:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z1.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1h { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll b/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
index 36d64725742e5..818f37c85ffdb 100644
--- a/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
+++ b/llvm/test/CodeGen/AArch64/sve-umulo-sdnode.ll
@@ -6,9 +6,9 @@ declare { <vscale x 2 x i8>, <vscale x 2 x i1> } @llvm.umul.with.overflow.nxv2i8
 define <vscale x 2 x i8> @umulo_nxv2i8(<vscale x 2 x i8> %x, <vscale x 2 x i8> %y) {
 ; CHECK-LABEL: umulo_nxv2i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0xff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    lsr z1.d, z0.d, #8
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
@@ -26,9 +26,9 @@ declare { <vscale x 4 x i8>, <vscale x 4 x i1> } @llvm.umul.with.overflow.nxv4i8
 define <vscale x 4 x i8> @umulo_nxv4i8(<vscale x 4 x i8> %x, <vscale x 4 x i8> %y) {
 ; CHECK-LABEL: umulo_nxv4i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z1.s, z1.s, #0xff
 ; CHECK-NEXT:    and z0.s, z0.s, #0xff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    lsr z1.s, z0.s, #8
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
@@ -46,9 +46,9 @@ declare { <vscale x 8 x i8>, <vscale x 8 x i1> } @llvm.umul.with.overflow.nxv8i8
 define <vscale x 8 x i8> @umulo_nxv8i8(<vscale x 8 x i8> %x, <vscale x 8 x i8> %y) {
 ; CHECK-LABEL: umulo_nxv8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    and z1.h, z1.h, #0xff
 ; CHECK-NEXT:    and z0.h, z0.h, #0xff
+; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z1.h
 ; CHECK-NEXT:    lsr z1.h, z0.h, #8
 ; CHECK-NEXT:    cmpne p0.h, p0/z, z1.h, #0
@@ -119,9 +119,9 @@ define <vscale x 64 x i8> @umulo_nxv64i8(<vscale x 64 x i8> %x, <vscale x 64 x i
 ; CHECK-NEXT:    movprfx z27, z1
 ; CHECK-NEXT:    umulh z27.b, p0/m, z27.b, z5.b
 ; CHECK-NEXT:    mul z3.b, p0/m, z3.b, z7.b
-; CHECK-NEXT:    mul z0.b, p0/m, z0.b, z4.b
 ; CHECK-NEXT:    mul z2.b, p0/m, z2.b, z6.b
 ; CHECK-NEXT:    mul z1.b, p0/m, z1.b, z5.b
+; CHECK-NEXT:    mul z0.b, p0/m, z0.b, z4.b
 ; CHECK-NEXT:    cmpne p1.b, p0/z, z25.b, #0
 ; CHECK-NEXT:    cmpne p2.b, p0/z, z24.b, #0
 ; CHECK-NEXT:    cmpne p3.b, p0/z, z26.b, #0
@@ -143,9 +143,9 @@ declare { <vscale x 2 x i16>, <vscale x 2 x i1> } @llvm.umul.with.overflow.nxv2i
 define <vscale x 2 x i16> @umulo_nxv2i16(<vscale x 2 x i16> %x, <vscale x 2 x i16> %y) {
 ; CHECK-LABEL: umulo_nxv2i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0xffff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    lsr z1.d, z0.d, #16
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
@@ -163,9 +163,9 @@ declare { <vscale x 4 x i16>, <vscale x 4 x i1> } @llvm.umul.with.overflow.nxv4i
 define <vscale x 4 x i16> @umulo_nxv4i16(<vscale x 4 x i16> %x, <vscale x 4 x i16> %y) {
 ; CHECK-LABEL: umulo_nxv4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    and z1.s, z1.s, #0xffff
 ; CHECK-NEXT:    and z0.s, z0.s, #0xffff
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    lsr z1.s, z0.s, #16
 ; CHECK-NEXT:    cmpne p0.s, p0/z, z1.s, #0
@@ -236,9 +236,9 @@ define <vscale x 32 x i16> @umulo_nxv32i16(<vscale x 32 x i16> %x, <vscale x 32
 ; CHECK-NEXT:    movprfx z27, z1
 ; CHECK-NEXT:    umulh z27.h, p0/m, z27.h, z5.h
 ; CHECK-NEXT:    mul z3.h, p0/m, z3.h, z7.h
-; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z4.h
 ; CHECK-NEXT:    mul z2.h, p0/m, z2.h, z6.h
 ; CHECK-NEXT:    mul z1.h, p0/m, z1.h, z5.h
+; CHECK-NEXT:    mul z0.h, p0/m, z0.h, z4.h
 ; CHECK-NEXT:    cmpne p1.h, p0/z, z25.h, #0
 ; CHECK-NEXT:    cmpne p2.h, p0/z, z24.h, #0
 ; CHECK-NEXT:    cmpne p3.h, p0/z, z26.h, #0
@@ -260,9 +260,9 @@ declare { <vscale x 2 x i32>, <vscale x 2 x i1> } @llvm.umul.with.overflow.nxv2i
 define <vscale x 2 x i32> @umulo_nxv2i32(<vscale x 2 x i32> %x, <vscale x 2 x i32> %y) {
 ; CHECK-LABEL: umulo_nxv2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
 ; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    lsr z1.d, z0.d, #32
 ; CHECK-NEXT:    cmpne p0.d, p0/z, z1.d, #0
@@ -333,9 +333,9 @@ define <vscale x 16 x i32> @umulo_nxv16i32(<vscale x 16 x i32> %x, <vscale x 16
 ; CHECK-NEXT:    movprfx z27, z1
 ; CHECK-NEXT:    umulh z27.s, p0/m, z27.s, z5.s
 ; CHECK-NEXT:    mul z3.s, p0/m, z3.s, z7.s
-; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z4.s
 ; CHECK-NEXT:    mul z2.s, p0/m, z2.s, z6.s
 ; CHECK-NEXT:    mul z1.s, p0/m, z1.s, z5.s
+; CHECK-NEXT:    mul z0.s, p0/m, z0.s, z4.s
 ; CHECK-NEXT:    cmpne p1.s, p0/z, z25.s, #0
 ; CHECK-NEXT:    cmpne p2.s, p0/z, z24.s, #0
 ; CHECK-NEXT:    cmpne p3.s, p0/z, z26.s, #0
@@ -410,9 +410,9 @@ define <vscale x 8 x i64> @umulo_nxv8i64(<vscale x 8 x i64> %x, <vscale x 8 x i6
 ; CHECK-NEXT:    movprfx z27, z1
 ; CHECK-NEXT:    umulh z27.d, p0/m, z27.d, z5.d
 ; CHECK-NEXT:    mul z3.d, p0/m, z3.d, z7.d
-; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z4.d
 ; CHECK-NEXT:    mul z2.d, p0/m, z2.d, z6.d
 ; CHECK-NEXT:    mul z1.d, p0/m, z1.d, z5.d
+; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z4.d
 ; CHECK-NEXT:    cmpne p1.d, p0/z, z25.d, #0
 ; CHECK-NEXT:    cmpne p2.d, p0/z, z24.d, #0
 ; CHECK-NEXT:    cmpne p3.d, p0/z, z26.d, #0
diff --git a/llvm/test/CodeGen/AArch64/sve-uunpklo-load-uzp1-store-combine.ll b/llvm/test/CodeGen/AArch64/sve-uunpklo-load-uzp1-store-combine.ll
index fc0dd2b163369..fc315b3a4730a 100644
--- a/llvm/test/CodeGen/AArch64/sve-uunpklo-load-uzp1-store-combine.ll
+++ b/llvm/test/CodeGen/AArch64/sve-uunpklo-load-uzp1-store-combine.ll
@@ -114,8 +114,8 @@ define void @uzp1_i8_valid(<vscale x 8 x i16> %a, ptr %b) #0 {
 define void @uzp1_i8_invalid(<vscale x 8 x i16> %a, ptr %b) #0 {
 ; CHECK-LABEL: uzp1_i8_invalid:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b, vl128
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z0.b
+; CHECK-NEXT:    ptrue p0.b, vl128
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0]
 ; CHECK-NEXT:    ret
   %a.bc = bitcast <vscale x 8 x i16> %a to <vscale x 16 x i8>
@@ -141,8 +141,8 @@ define void @uzp1_i16_valid(<vscale x 4 x i32> %a, ptr %b) #0 {
 define void @uzp1_i16_invalid(<vscale x 4 x i32> %a, ptr %b) #0 {
 ; CHECK-LABEL: uzp1_i16_invalid:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
+; CHECK-NEXT:    ptrue p0.h, vl64
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0]
 ; CHECK-NEXT:    ret
   %a.bc = bitcast <vscale x 4 x i32> %a to <vscale x 8 x i16>
@@ -168,8 +168,8 @@ define void @uzp1_i32_valid(<vscale x 2 x i64> %a, ptr %b) #0 {
 define void @uzp1_i32_invalid(<vscale x 2 x i64> %a, ptr %b) #0 {
 ; CHECK-LABEL: uzp1_i32_invalid:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK-NEXT:    ptrue p0.s, vl32
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   %a.bc = bitcast <vscale x 2 x i64> %a to <vscale x 4 x i32>
@@ -182,8 +182,8 @@ define void @uzp1_i32_invalid(<vscale x 2 x i64> %a, ptr %b) #0 {
 define void @uzp1_invalid_all(<vscale x 2 x i64> %a, ptr %b) #0 {
 ; CHECK-LABEL: uzp1_invalid_all:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0]
 ; CHECK-NEXT:    ret
   %a.bc = bitcast <vscale x 2 x i64> %a to <vscale x 4 x i32>
diff --git a/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll b/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
index 194d1071301d4..91f8f5c2c90d8 100644
--- a/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
+++ b/llvm/test/CodeGen/AArch64/sve-vecreduce-dot.ll
@@ -8,11 +8,11 @@ define i32 @test(<vscale x 32 x i8> %bin.rdx, <vscale x 32 x i8> %bin.rdx2)  {
 ; CHECK-NEXT:    sunpklo z5.h, z0.b
 ; CHECK-NEXT:    sunpkhi z0.h, z0.b
 ; CHECK-NEXT:    sunpkhi z2.h, z2.b
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpklo z6.h, z1.b
 ; CHECK-NEXT:    sunpkhi z1.h, z1.b
 ; CHECK-NEXT:    sunpklo z7.h, z3.b
 ; CHECK-NEXT:    sunpkhi z3.h, z3.b
+; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    sunpkhi z24.s, z5.h
 ; CHECK-NEXT:    sunpklo z5.s, z5.h
 ; CHECK-NEXT:    sunpklo z25.s, z4.h
diff --git a/llvm/test/CodeGen/AArch64/sve-vecreduce-fold.ll b/llvm/test/CodeGen/AArch64/sve-vecreduce-fold.ll
index 898090340869e..0bdaefdfc2a3f 100644
--- a/llvm/test/CodeGen/AArch64/sve-vecreduce-fold.ll
+++ b/llvm/test/CodeGen/AArch64/sve-vecreduce-fold.ll
@@ -80,8 +80,8 @@ define i1 @reduce_and_insert_subvec_into_var(<vscale x 4 x i1> %in, <vscale x 16
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    punpklo p3.h, p1.b
 ; CHECK-NEXT:    punpkhi p1.h, p1.b
-; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    punpkhi p3.h, p3.b
+; CHECK-NEXT:    ptrue p2.b
 ; CHECK-NEXT:    uzp1 p0.h, p0.h, p3.h
 ; CHECK-NEXT:    uzp1 p0.b, p0.b, p1.b
 ; CHECK-NEXT:    nots p0.b, p2/z, p0.b
diff --git a/llvm/test/CodeGen/AArch64/sve2-fcopysign.ll b/llvm/test/CodeGen/AArch64/sve2-fcopysign.ll
index 14cc8cd268261..778d7e193e23e 100644
--- a/llvm/test/CodeGen/AArch64/sve2-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-fcopysign.ll
@@ -48,11 +48,11 @@ define <vscale x 4 x float> @test_copysign_v4f32_v4f64(<vscale x 4 x float> %a,
 ; CHECK_NO_EXTEND_ROUND-LABEL: test_copysign_v4f32_v4f64:
 ; CHECK_NO_EXTEND_ROUND:       // %bb.0:
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ptrue p0.d
+; CHECK_NO_EXTEND_ROUND-NEXT:    mov z3.s, #0x7fffffff
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z2.s, p0/m, z2.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z1.s, p0/m, z1.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    uzp1 z1.s, z1.s, z2.s
-; CHECK_NO_EXTEND_ROUND-NEXT:    mov z2.s, #0x7fffffff
-; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z2.d
+; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z3.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ret
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v4f32_v4f64:
@@ -107,9 +107,9 @@ declare <vscale x 2 x double> @llvm.copysign.v2f64(<vscale x 2 x double> %a, <vs
 define <vscale x 4 x double> @test_copysign_v4f64_v4f32(<vscale x 4 x double> %a, <vscale x 4 x float> %b) #0 {
 ; CHECK_NO_EXTEND_ROUND-LABEL: test_copysign_v4f64_v4f32:
 ; CHECK_NO_EXTEND_ROUND:       // %bb.0:
-; CHECK_NO_EXTEND_ROUND-NEXT:    ptrue p0.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    uunpkhi z3.d, z2.s
 ; CHECK_NO_EXTEND_ROUND-NEXT:    uunpklo z2.d, z2.s
+; CHECK_NO_EXTEND_ROUND-NEXT:    ptrue p0.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    mov z4.d, #0x7fffffffffffffff
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z3.d, p0/m, z3.s
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z2.d, p0/m, z2.s
@@ -119,9 +119,9 @@ define <vscale x 4 x double> @test_copysign_v4f64_v4f32(<vscale x 4 x double> %a
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v4f64_v4f32:
 ; CHECK_EXTEND_ROUND:       // %bb.0:
-; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d
 ; CHECK_EXTEND_ROUND-NEXT:    uunpkhi z3.d, z2.s
 ; CHECK_EXTEND_ROUND-NEXT:    uunpklo z2.d, z2.s
+; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d
 ; CHECK_EXTEND_ROUND-NEXT:    mov z4.d, #0x7fffffffffffffff
 ; CHECK_EXTEND_ROUND-NEXT:    fcvt z2.d, p0/m, z2.s
 ; CHECK_EXTEND_ROUND-NEXT:    fcvt z3.d, p0/m, z3.s
@@ -176,11 +176,11 @@ define <vscale x 4 x half> @test_copysign_v4f16_v4f64(<vscale x 4 x half> %a, <v
 ; CHECK_NO_EXTEND_ROUND-LABEL: test_copysign_v4f16_v4f64:
 ; CHECK_NO_EXTEND_ROUND:       // %bb.0:
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ptrue p0.d
+; CHECK_NO_EXTEND_ROUND-NEXT:    mov z3.h, #32767 // =0x7fff
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z2.h, p0/m, z2.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z1.h, p0/m, z1.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    uzp1 z1.s, z1.s, z2.s
-; CHECK_NO_EXTEND_ROUND-NEXT:    mov z2.h, #32767 // =0x7fff
-; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z2.d
+; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z3.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ret
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v4f16_v4f64:
@@ -218,11 +218,11 @@ define <vscale x 8 x half> @test_copysign_v8f16_v8f32(<vscale x 8 x half> %a, <v
 ; CHECK_NO_EXTEND_ROUND-LABEL: test_copysign_v8f16_v8f32:
 ; CHECK_NO_EXTEND_ROUND:       // %bb.0:
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ptrue p0.s
+; CHECK_NO_EXTEND_ROUND-NEXT:    mov z3.h, #32767 // =0x7fff
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z2.h, p0/m, z2.s
 ; CHECK_NO_EXTEND_ROUND-NEXT:    fcvt z1.h, p0/m, z1.s
 ; CHECK_NO_EXTEND_ROUND-NEXT:    uzp1 z1.h, z1.h, z2.h
-; CHECK_NO_EXTEND_ROUND-NEXT:    mov z2.h, #32767 // =0x7fff
-; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z2.d
+; CHECK_NO_EXTEND_ROUND-NEXT:    bsl z0.d, z0.d, z1.d, z3.d
 ; CHECK_NO_EXTEND_ROUND-NEXT:    ret
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v8f16_v8f32:
diff --git a/llvm/test/CodeGen/AArch64/sve2-fixed-length-fcopysign.ll b/llvm/test/CodeGen/AArch64/sve2-fixed-length-fcopysign.ll
index 5735c36558bdb..e39e5feefa255 100644
--- a/llvm/test/CodeGen/AArch64/sve2-fixed-length-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-fixed-length-fcopysign.ll
@@ -376,9 +376,9 @@ define void @test_copysign_v4f32_v4f64(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mvni v1.4s, #128, lsl #24
 ; CHECK-NEXT:    ldr q2, [x0]
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.s, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    fcvt z0.s, p0/m, z0.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    bit v0.16b, v2.16b, v1.16b
 ; CHECK-NEXT:    str q0, [x0]
@@ -429,12 +429,12 @@ define void @test_copysign_v4f64_v4f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ;
 ; CHECK_EXTEND_ROUND-LABEL: test_copysign_v4f64_v4f32:
 ; CHECK_EXTEND_ROUND:       // %bb.0:
-; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d, vl4
 ; CHECK_EXTEND_ROUND-NEXT:    ldr q0, [x1]
+; CHECK_EXTEND_ROUND-NEXT:    ptrue p0.d, vl4
 ; CHECK_EXTEND_ROUND-NEXT:    mov z1.d, #0x7fffffffffffffff
 ; CHECK_EXTEND_ROUND-NEXT:    uunpklo z0.d, z0.s
-; CHECK_EXTEND_ROUND-NEXT:    fcvt z0.d, p0/m, z0.s
 ; CHECK_EXTEND_ROUND-NEXT:    ld1d { z2.d }, p0/z, [x0]
+; CHECK_EXTEND_ROUND-NEXT:    fcvt z0.d, p0/m, z0.s
 ; CHECK_EXTEND_ROUND-NEXT:    bsl z2.d, z2.d, z0.d, z1.d
 ; CHECK_EXTEND_ROUND-NEXT:    st1d { z2.d }, p0, [x0]
 ; CHECK_EXTEND_ROUND-NEXT:    ret
@@ -472,9 +472,9 @@ define void @test_copysign_v4f16_v4f64(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.d, vl4
 ; CHECK-NEXT:    mvni v1.4h, #128, lsl #8
 ; CHECK-NEXT:    ldr d2, [x0]
-; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.d
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.d
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    bit v0.8b, v2.8b, v1.8b
@@ -499,9 +499,9 @@ define void @test_copysign_v8f16_v8f32(ptr %ap, ptr %bp) vscale_range(2,0) #0 {
 ; CHECK-NEXT:    ptrue p0.s, vl8
 ; CHECK-NEXT:    mvni v1.8h, #128, lsl #8
 ; CHECK-NEXT:    ldr q2, [x0]
-; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x1]
-; CHECK-NEXT:    fcvt z0.h, p1/m, z0.s
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    fcvt z0.h, p0/m, z0.s
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z0.h
 ; CHECK-NEXT:    bit v0.16b, v2.16b, v1.16b
 ; CHECK-NEXT:    str q0, [x0]
diff --git a/llvm/test/CodeGen/AArch64/sve2-intrinsics-combine-rshrnb.ll b/llvm/test/CodeGen/AArch64/sve2-intrinsics-combine-rshrnb.ll
index 02b9562b8f52b..82078ad63d66b 100644
--- a/llvm/test/CodeGen/AArch64/sve2-intrinsics-combine-rshrnb.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-intrinsics-combine-rshrnb.ll
@@ -120,12 +120,12 @@ define void @neg_zero_shift(ptr %ptr, ptr %dst, i64 %index){
 define void @wide_add_shift_add_rshrnb_b(ptr %dest, i64 %index, <vscale x 16 x i16> %arg1){
 ; CHECK-LABEL: wide_add_shift_add_rshrnb_b:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.b
 ; CHECK-NEXT:    rshrnb z1.b, z1.h, #6
 ; CHECK-NEXT:    rshrnb z0.b, z0.h, #6
+; CHECK-NEXT:    ptrue p0.b
+; CHECK-NEXT:    ld1b { z2.b }, p0/z, [x0, x1]
 ; CHECK-NEXT:    uzp1 z0.b, z0.b, z1.b
-; CHECK-NEXT:    ld1b { z1.b }, p0/z, [x0, x1]
-; CHECK-NEXT:    add z0.b, z1.b, z0.b
+; CHECK-NEXT:    add z0.b, z2.b, z0.b
 ; CHECK-NEXT:    st1b { z0.b }, p0, [x0, x1]
 ; CHECK-NEXT:    ret
   %1 = add <vscale x 16 x i16> %arg1, splat (i16 32)
@@ -141,12 +141,12 @@ define void @wide_add_shift_add_rshrnb_b(ptr %dest, i64 %index, <vscale x 16 x i
 define void @wide_add_shift_add_rshrnb_h(ptr %dest, i64 %index, <vscale x 8 x i32> %arg1){
 ; CHECK-LABEL: wide_add_shift_add_rshrnb_h:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.h
 ; CHECK-NEXT:    rshrnb z1.h, z1.s, #6
 ; CHECK-NEXT:    rshrnb z0.h, z0.s, #6
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    ld1h { z2.h }, p0/z, [x0, x1, lsl #1]
 ; CHECK-NEXT:    uzp1 z0.h, z0.h, z1.h
-; CHECK-NEXT:    ld1h { z1.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    add z0.h, z1.h, z0.h
+; CHECK-NEXT:    add z0.h, z2.h, z0.h
 ; CHECK-NEXT:    st1h { z0.h }, p0, [x0, x1, lsl #1]
 ; CHECK-NEXT:    ret
   %1 = add <vscale x 8 x i32> %arg1, splat (i32 32)
@@ -162,12 +162,12 @@ define void @wide_add_shift_add_rshrnb_h(ptr %dest, i64 %index, <vscale x 8 x i3
 define void @wide_add_shift_add_rshrnb_d(ptr %dest, i64 %index, <vscale x 4 x i64> %arg1){
 ; CHECK-LABEL: wide_add_shift_add_rshrnb_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    rshrnb z1.s, z1.d, #32
 ; CHECK-NEXT:    rshrnb z0.s, z0.d, #32
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0, x1, lsl #2]
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z1.s
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    add z0.s, z1.s, z0.s
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0, x1, lsl #2]
 ; CHECK-NEXT:    ret
   %1 = add <vscale x 4 x i64> %arg1, splat (i64 2147483648)
@@ -188,11 +188,11 @@ define void @neg_wide_add_shift_add_rshrnb_d(ptr %dest, i64 %index, <vscale x 4
 ; CHECK-NEXT:    ptrue p0.s
 ; CHECK-NEXT:    add z0.d, z0.d, z2.d
 ; CHECK-NEXT:    add z1.d, z1.d, z2.d
+; CHECK-NEXT:    ld1w { z2.s }, p0/z, [x0, x1, lsl #2]
 ; CHECK-NEXT:    lsr z1.d, z1.d, #48
 ; CHECK-NEXT:    lsr z0.d, z0.d, #48
 ; CHECK-NEXT:    uzp1 z0.s, z0.s, z1.s
-; CHECK-NEXT:    ld1w { z1.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    add z0.s, z1.s, z0.s
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
 ; CHECK-NEXT:    st1w { z0.s }, p0, [x0, x1, lsl #2]
 ; CHECK-NEXT:    ret
   %1 = add <vscale x 4 x i64> %arg1, splat (i64 140737488355328)
diff --git a/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-arith-imm.ll b/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-arith-imm.ll
index 8bfcd088e1a86..500973d053f5b 100644
--- a/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-arith-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-intrinsics-int-arith-imm.ll
@@ -437,8 +437,8 @@ define <vscale x 4 x i32> @uqsub_i32_ptrue_all_h(<vscale x 4 x i32> %a) #0 {
 define <vscale x 4 x i32> @uqsub_i32_ptrue_all_d(<vscale x 4 x i32> %a) #0 {
 ; CHECK-LABEL: uqsub_i32_ptrue_all_d:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    mov z1.s, #1 // =0x1
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    uqsub z0.s, p0/m, z0.s, z1.s
 ; CHECK-NEXT:    ret
   %pg.d = tail call <vscale x 2 x i1> @llvm.aarch64.sve.ptrue.nxv2i1(i32 31)
diff --git a/llvm/test/CodeGen/AArch64/sve2-rsh.ll b/llvm/test/CodeGen/AArch64/sve2-rsh.ll
index 516ef3bd581ee..9addd16f89292 100644
--- a/llvm/test/CodeGen/AArch64/sve2-rsh.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-rsh.ll
@@ -18,8 +18,8 @@ define <vscale x 2 x i64> @neg_urshr_1(<vscale x 2 x i64> %x) {
 define <vscale x 2 x i64> @neg_urshr_2(<vscale x 2 x i64> %x, <vscale x 2 x i64> %y) {
 ; CHECK-LABEL: neg_urshr_2:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    add z0.d, z0.d, #32 // =0x20
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    lsr z0.d, p0/m, z0.d, z1.d
 ; CHECK-NEXT:    ret
   %add = add nuw nsw <vscale x 2 x i64> %x, splat (i64 32)
diff --git a/llvm/test/CodeGen/AArch64/sve2-xar.ll b/llvm/test/CodeGen/AArch64/sve2-xar.ll
index e297ade6b9ae1..e5a240b7a53fd 100644
--- a/llvm/test/CodeGen/AArch64/sve2-xar.ll
+++ b/llvm/test/CodeGen/AArch64/sve2-xar.ll
@@ -152,9 +152,9 @@ define <vscale x 2 x i64> @xar_nxv2i64_l_neg1(<vscale x 2 x i64> %x, <vscale x 2
 ; CHECK-LABEL: xar_nxv2i64_l_neg1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z3.d, z2.d
-; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    subr z2.d, z2.d, #0 // =0x0
 ; CHECK-NEXT:    eor z0.d, z0.d, z1.d
+; CHECK-NEXT:    ptrue p0.d
 ; CHECK-NEXT:    and z3.d, z3.d, #0x3f
 ; CHECK-NEXT:    and z2.d, z2.d, #0x3f
 ; CHECK-NEXT:    movprfx z1, z0
diff --git a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx2.ll b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx2.ll
index a42d99fd8318e..e821de24b5f14 100644
--- a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx2.ll
+++ b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx2.ll
@@ -8,12 +8,12 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8> } @sel_x2_i8(target("aarch64.svc
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.b, z1.b }, pn8, { z6.b, z7.b }, { z4.b, z5.b }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -28,12 +28,12 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16> } @sel_x2_i16(target("aarch64.sv
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.h, z1.h }, pn8, { z6.h, z7.h }, { z4.h, z5.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -48,12 +48,12 @@ define { <vscale x 8 x half>, <vscale x 8 x half> } @sel_x2_f16(target("aarch64.
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.h, z1.h }, pn8, { z6.h, z7.h }, { z4.h, z5.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -68,12 +68,12 @@ define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @sel_x2_bf16(target("aar
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.h, z1.h }, pn8, { z6.h, z7.h }, { z4.h, z5.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -88,12 +88,12 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32> } @sel_x2_i32(target("aarch64.sv
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.s, z1.s }, pn8, { z6.s, z7.s }, { z4.s, z5.s }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -108,12 +108,12 @@ define { <vscale x 4 x float>, <vscale x 4 x float> } @sel_x2_f32(target("aarch6
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.s, z1.s }, pn8, { z6.s, z7.s }, { z4.s, z5.s }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -128,12 +128,12 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64> } @sel_x2_i64(target("aarch64.sv
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.d, z1.d }, pn8, { z6.d, z7.d }, { z4.d, z5.d }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -148,12 +148,12 @@ define { <vscale x 2 x double>, <vscale x 2 x double> } @sel_x2_f64(target("aarc
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
-; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z5.d, z4.d
 ; CHECK-NEXT:    mov z7.d, z2.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z4.d, z3.d
 ; CHECK-NEXT:    mov z6.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    sel { z0.d, z1.d }, pn8, { z6.d, z7.d }, { z4.d, z5.d }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
diff --git a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx4.ll b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx4.ll
index df504362680ba..3a21eaead5f72 100644
--- a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx4.ll
+++ b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-selx4.ll
@@ -8,17 +8,17 @@ define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 1
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.b
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.b
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1b { z27.b }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.b - z3.b }, pn8, { z28.b - z31.b }, { z24.b - z27.b }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -33,17 +33,17 @@ define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z27.h }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.h - z3.h }, pn8, { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -58,17 +58,17 @@ define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z27.h }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.h - z3.h }, pn8, { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -83,17 +83,17 @@ define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.h
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.h
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z27.h }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.h - z3.h }, pn8, { z28.h - z31.h }, { z24.h - z27.h }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -108,17 +108,17 @@ define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.s
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z27.s }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.s - z3.s }, pn8, { z28.s - z31.s }, { z24.s - z27.s }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -133,17 +133,17 @@ define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vsca
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.s
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.s
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z27.s }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.s - z3.s }, pn8, { z28.s - z31.s }, { z24.s - z27.s }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -158,17 +158,17 @@ define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.d
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z27.d }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.d - z3.d }, pn8, { z28.d - z31.d }, { z24.d - z27.d }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
@@ -183,17 +183,17 @@ define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    ptrue p1.d
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z26.d, z7.d
 ; CHECK-NEXT:    mov z31.d, z4.d
-; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    mov z25.d, z6.d
 ; CHECK-NEXT:    mov z30.d, z3.d
 ; CHECK-NEXT:    mov z24.d, z5.d
 ; CHECK-NEXT:    mov z29.d, z2.d
-; CHECK-NEXT:    mov z28.d, z1.d
+; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z27.d }, p1/z, [x0]
+; CHECK-NEXT:    mov z28.d, z1.d
 ; CHECK-NEXT:    sel { z0.d - z3.d }, pn8, { z28.d - z31.d }, { z24.d - z27.d }
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
diff --git a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-stores.ll b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-stores.ll
index 39639120802fa..8b54a626b3660 100644
--- a/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve2p1-intrinsics-stores.ll
@@ -9,9 +9,9 @@ define void @st1_x2_i8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vsc
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1b { z2.b, z3.b }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -27,9 +27,9 @@ define void @st1_x2_i16(<vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -45,9 +45,9 @@ define void @st1_x2_i32(<vscale x 16 x i8> %unused, <vscale x 4 x i32> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1w { z2.s, z3.s }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -63,9 +63,9 @@ define void @st1_x2_i64(<vscale x 16 x i8> %unused, <vscale x 2 x i64> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1d { z2.d, z3.d }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -81,9 +81,9 @@ define void @st1_x2_f16(<vscale x 16 x i8> %unused, <vscale x 8 x half> %zn0, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -99,9 +99,9 @@ define void @st1_x2_bf16(<vscale x 16 x i8> %unused, <vscale x 8 x bfloat> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -117,9 +117,9 @@ define void @st1_x2_f32(<vscale x 16 x i8> %unused, <vscale x 4 x float> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1w { z2.s, z3.s }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -135,9 +135,9 @@ define void @st1_x2_f64(<vscale x 16 x i8> %unused, <vscale x 2 x double> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    st1d { z2.d, z3.d }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -153,8 +153,8 @@ define void @st1_x4_i8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vsc
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -173,8 +173,8 @@ define void @st1_x4_i16(<vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -193,8 +193,8 @@ define void @st1_x4_i32(<vscale x 16 x i8> %unused, <vscale x 4 x i32> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -213,8 +213,8 @@ define void @st1_x4_i64(<vscale x 16 x i8> %unused, <vscale x 2 x i64> %zn0, <vs
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -233,8 +233,8 @@ define void @st1_x4_f16(<vscale x 16 x i8> %unused, <vscale x 8 x half> %zn0, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -253,8 +253,8 @@ define void @st1_x4_bf16(<vscale x 16 x i8> %unused, <vscale x 8 x bfloat> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -273,8 +273,8 @@ define void @st1_x4_f32(<vscale x 16 x i8> %unused, <vscale x 4 x float> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -293,8 +293,8 @@ define void @st1_x4_f64(<vscale x 16 x i8> %unused, <vscale x 2 x double> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -315,9 +315,9 @@ define void @stnt1_x2_i8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1b { z2.b, z3.b }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -333,9 +333,9 @@ define void @stnt1_x2_i16(<vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -351,9 +351,9 @@ define void @stnt1_x2_i32(<vscale x 16 x i8> %unused, <vscale x 4 x i32> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1w { z2.s, z3.s }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -369,9 +369,9 @@ define void @stnt1_x2_i64(<vscale x 16 x i8> %unused, <vscale x 2 x i64> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1d { z2.d, z3.d }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -387,9 +387,9 @@ define void @stnt1_x2_f16(<vscale x 16 x i8> %unused, <vscale x 8 x half> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -405,9 +405,9 @@ define void @stnt1_x2_bf16(<vscale x 16 x i8> %unused, <vscale x 8 x bfloat> %zn
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1h { z2.h, z3.h }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -423,9 +423,9 @@ define void @stnt1_x2_f32(<vscale x 16 x i8> %unused, <vscale x 4 x float> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1w { z2.s, z3.s }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -441,9 +441,9 @@ define void @stnt1_x2_f64(<vscale x 16 x i8> %unused, <vscale x 2 x double> %zn0
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
-; CHECK-NEXT:    mov z3.d, z2.d
 ; CHECK-NEXT:    mov z2.d, z1.d
 ; CHECK-NEXT:    stnt1d { z2.d, z3.d }, pn8, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
@@ -459,8 +459,8 @@ define void @stnt1_x4_i8(<vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <v
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -479,8 +479,8 @@ define void @stnt1_x4_i16(<vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -499,8 +499,8 @@ define void @stnt1_x4_i32(<vscale x 16 x i8> %unused, <vscale x 4 x i32> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -519,8 +519,8 @@ define void @stnt1_x4_i64(<vscale x 16 x i8> %unused, <vscale x 2 x i64> %zn0, <
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -539,8 +539,8 @@ define void @stnt1_x4_f16(<vscale x 16 x i8> %unused, <vscale x 8 x half> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -559,8 +559,8 @@ define void @stnt1_x4_bf16(<vscale x 16 x i8> %unused, <vscale x 8 x bfloat> %zn
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -579,8 +579,8 @@ define void @stnt1_x4_f32(<vscale x 16 x i8> %unused, <vscale x 4 x float> %zn0,
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
@@ -599,8 +599,8 @@ define void @stnt1_x4_f64(<vscale x 16 x i8> %unused, <vscale x 2 x double> %zn0
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
-; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov z7.d, z4.d
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    mov z6.d, z3.d
 ; CHECK-NEXT:    mov z5.d, z2.d
diff --git a/llvm/test/CodeGen/AArch64/uadd_sat_vec.ll b/llvm/test/CodeGen/AArch64/uadd_sat_vec.ll
index 30ff70088454d..16521834090b5 100644
--- a/llvm/test/CodeGen/AArch64/uadd_sat_vec.ll
+++ b/llvm/test/CodeGen/AArch64/uadd_sat_vec.ll
@@ -147,11 +147,11 @@ define void @v8i8(ptr %px, ptr %py, ptr %pz) nounwind {
 define void @v4i8(ptr %px, ptr %py, ptr %pz) nounwind {
 ; CHECK-SD-LABEL: v4i8:
 ; CHECK-SD:       // %bb.0:
-; CHECK-SD-NEXT:    ldr s1, [x0]
-; CHECK-SD-NEXT:    ldr s2, [x1]
-; CHECK-SD-NEXT:    movi d0, #0xff00ff00ff00ff
-; CHECK-SD-NEXT:    uaddl v1.8h, v1.8b, v2.8b
-; CHECK-SD-NEXT:    umin v0.4h, v1.4h, v0.4h
+; CHECK-SD-NEXT:    ldr s0, [x0]
+; CHECK-SD-NEXT:    ldr s1, [x1]
+; CHECK-SD-NEXT:    movi d2, #0xff00ff00ff00ff
+; CHECK-SD-NEXT:    uaddl v0.8h, v0.8b, v1.8b
+; CHECK-SD-NEXT:    umin v0.4h, v0.4h, v2.4h
 ; CHECK-SD-NEXT:    uzp1 v0.8b, v0.8b, v0.8b
 ; CHECK-SD-NEXT:    str s0, [x2]
 ; CHECK-SD-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/urem-seteq-vec-nonzero.ll b/llvm/test/CodeGen/AArch64/urem-seteq-vec-nonzero.ll
index b31ce94cdaaea..d5f1febaeb7db 100644
--- a/llvm/test/CodeGen/AArch64/urem-seteq-vec-nonzero.ll
+++ b/llvm/test/CodeGen/AArch64/urem-seteq-vec-nonzero.ll
@@ -52,11 +52,11 @@ define <4 x i1> @t32_6_part0(<4 x i32> %X) nounwind {
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    dup v1.4s, w8
 ; CHECK-NEXT:    mul v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    shl v1.4s, v0.4s, #31
-; CHECK-NEXT:    usra v1.4s, v0.4s, #1
-; CHECK-NEXT:    movi v0.16b, #170
-; CHECK-NEXT:    fneg v0.4s, v0.4s
-; CHECK-NEXT:    cmhs v0.4s, v0.4s, v1.4s
+; CHECK-NEXT:    movi v1.16b, #170
+; CHECK-NEXT:    shl v2.4s, v0.4s, #31
+; CHECK-NEXT:    fneg v1.4s, v1.4s
+; CHECK-NEXT:    usra v2.4s, v0.4s, #1
+; CHECK-NEXT:    cmhs v0.4s, v1.4s, v2.4s
 ; CHECK-NEXT:    xtn v0.4h, v0.4s
 ; CHECK-NEXT:    ret
   %urem = urem <4 x i32> %X, <i32 6, i32 6, i32 6, i32 6>
diff --git a/llvm/test/CodeGen/AArch64/vec_uaddo.ll b/llvm/test/CodeGen/AArch64/vec_uaddo.ll
index 00609b0df9b4e..37c6374215d81 100644
--- a/llvm/test/CodeGen/AArch64/vec_uaddo.ll
+++ b/llvm/test/CodeGen/AArch64/vec_uaddo.ll
@@ -100,9 +100,9 @@ define <6 x i32> @uaddo_v6i32(<6 x i32> %a0, <6 x i32> %a1, ptr %p2) nounwind {
 ; CHECK-NEXT:    mov v0.s[3], w3
 ; CHECK-NEXT:    cmhi v3.4s, v3.4s, v2.4s
 ; CHECK-NEXT:    str d2, [x8, #16]
-; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    mov w5, v3.s[1]
 ; CHECK-NEXT:    fmov w4, s3
+; CHECK-NEXT:    add v1.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    cmhi v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    str q1, [x8]
 ; CHECK-NEXT:    mov w1, v0.s[1]
@@ -248,10 +248,10 @@ define <4 x i32> @uaddo_v4i1(<4 x i1> %a0, <4 x i1> %a1, ptr %p2) nounwind {
 ; CHECK-NEXT:    eor v2.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    and v0.8b, v0.8b, v1.8b
 ; CHECK-NEXT:    adrp x8, .LCPI10_0
-; CHECK-NEXT:    shl v2.4h, v2.4h, #15
-; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
-; CHECK-NEXT:    cmlt v1.4h, v2.4h, #0
+; CHECK-NEXT:    shl v1.4h, v2.4h, #15
 ; CHECK-NEXT:    ldr d2, [x8, :lo12:.LCPI10_0]
+; CHECK-NEXT:    ushll v0.4s, v0.4h, #0
+; CHECK-NEXT:    cmlt v1.4h, v1.4h, #0
 ; CHECK-NEXT:    shl v0.4s, v0.4s, #31
 ; CHECK-NEXT:    and v1.8b, v1.8b, v2.8b
 ; CHECK-NEXT:    cmlt v0.4s, v0.4s, #0
diff --git a/llvm/test/CodeGen/AArch64/vecreduce-add.ll b/llvm/test/CodeGen/AArch64/vecreduce-add.ll
index 66ef436f48c63..3254c5ebe9c6b 100644
--- a/llvm/test/CodeGen/AArch64/vecreduce-add.ll
+++ b/llvm/test/CodeGen/AArch64/vecreduce-add.ll
@@ -2177,65 +2177,65 @@ define i32 @test_udot_v48i8(ptr %p1, ptr %p2) {
 ; CHECK-GI-BASE-LABEL: test_udot_v48i8:
 ; CHECK-GI-BASE:       // %bb.0: // %entry
 ; CHECK-GI-BASE-NEXT:    ldp q0, q3, [x1]
-; CHECK-GI-BASE-NEXT:    ldr q6, [x1, #32]
+; CHECK-GI-BASE-NEXT:    ldr q6, [x0, #32]
 ; CHECK-GI-BASE-NEXT:    ldp q1, q2, [x0]
-; CHECK-GI-BASE-NEXT:    ldr q17, [x0, #32]
+; CHECK-GI-BASE-NEXT:    ldr q7, [x1, #32]
+; CHECK-GI-BASE-NEXT:    ushll v20.8h, v6.8b, #0
+; CHECK-GI-BASE-NEXT:    ushll2 v6.8h, v6.16b, #0
 ; CHECK-GI-BASE-NEXT:    ushll v4.8h, v0.8b, #0
 ; CHECK-GI-BASE-NEXT:    ushll2 v0.8h, v0.16b, #0
-; CHECK-GI-BASE-NEXT:    ushll v7.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT:    ushll v16.8h, v3.8b, #0
 ; CHECK-GI-BASE-NEXT:    ushll v5.8h, v1.8b, #0
 ; CHECK-GI-BASE-NEXT:    ushll2 v1.8h, v1.16b, #0
-; CHECK-GI-BASE-NEXT:    ushll v16.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT:    ushll v17.8h, v2.8b, #0
 ; CHECK-GI-BASE-NEXT:    ushll2 v3.8h, v3.16b, #0
 ; CHECK-GI-BASE-NEXT:    ushll2 v2.8h, v2.16b, #0
 ; CHECK-GI-BASE-NEXT:    umull v18.4s, v4.4h, v5.4h
 ; CHECK-GI-BASE-NEXT:    umull2 v4.4s, v4.8h, v5.8h
-; CHECK-GI-BASE-NEXT:    umull2 v19.4s, v0.8h, v1.8h
-; CHECK-GI-BASE-NEXT:    umull v20.4s, v7.4h, v16.4h
-; CHECK-GI-BASE-NEXT:    umull v0.4s, v0.4h, v1.4h
-; CHECK-GI-BASE-NEXT:    ushll v5.8h, v6.8b, #0
-; CHECK-GI-BASE-NEXT:    ushll v1.8h, v17.8b, #0
-; CHECK-GI-BASE-NEXT:    umull2 v7.4s, v7.8h, v16.8h
-; CHECK-GI-BASE-NEXT:    ushll2 v6.8h, v6.16b, #0
-; CHECK-GI-BASE-NEXT:    ushll2 v17.8h, v17.16b, #0
-; CHECK-GI-BASE-NEXT:    addv s16, v18.4s
-; CHECK-GI-BASE-NEXT:    addv s4, v4.4s
-; CHECK-GI-BASE-NEXT:    umull v18.4s, v3.4h, v2.4h
+; CHECK-GI-BASE-NEXT:    umull v5.4s, v0.4h, v1.4h
+; CHECK-GI-BASE-NEXT:    umull2 v0.4s, v0.8h, v1.8h
+; CHECK-GI-BASE-NEXT:    umull v19.4s, v16.4h, v17.4h
+; CHECK-GI-BASE-NEXT:    ushll v1.8h, v7.8b, #0
+; CHECK-GI-BASE-NEXT:    umull2 v16.4s, v16.8h, v17.8h
+; CHECK-GI-BASE-NEXT:    umull v17.4s, v3.4h, v2.4h
 ; CHECK-GI-BASE-NEXT:    umull2 v2.4s, v3.8h, v2.8h
-; CHECK-GI-BASE-NEXT:    addv s3, v19.4s
-; CHECK-GI-BASE-NEXT:    umull v19.4s, v5.4h, v1.4h
-; CHECK-GI-BASE-NEXT:    umull2 v1.4s, v5.8h, v1.8h
-; CHECK-GI-BASE-NEXT:    addv s5, v20.4s
+; CHECK-GI-BASE-NEXT:    ushll2 v7.8h, v7.16b, #0
+; CHECK-GI-BASE-NEXT:    addv s18, v18.4s
+; CHECK-GI-BASE-NEXT:    addv s4, v4.4s
+; CHECK-GI-BASE-NEXT:    addv s5, v5.4s
 ; CHECK-GI-BASE-NEXT:    addv s0, v0.4s
-; CHECK-GI-BASE-NEXT:    addv s7, v7.4s
-; CHECK-GI-BASE-NEXT:    umull v20.4s, v6.4h, v17.4h
-; CHECK-GI-BASE-NEXT:    umull2 v6.4s, v6.8h, v17.8h
-; CHECK-GI-BASE-NEXT:    fmov w8, s16
-; CHECK-GI-BASE-NEXT:    fmov w9, s4
-; CHECK-GI-BASE-NEXT:    fmov w10, s3
-; CHECK-GI-BASE-NEXT:    addv s3, v18.4s
+; CHECK-GI-BASE-NEXT:    addv s19, v19.4s
+; CHECK-GI-BASE-NEXT:    umull v3.4s, v1.4h, v20.4h
 ; CHECK-GI-BASE-NEXT:    addv s2, v2.4s
-; CHECK-GI-BASE-NEXT:    fmov w11, s5
-; CHECK-GI-BASE-NEXT:    addv s4, v19.4s
+; CHECK-GI-BASE-NEXT:    umull2 v1.4s, v1.8h, v20.8h
+; CHECK-GI-BASE-NEXT:    umull v20.4s, v7.4h, v6.4h
+; CHECK-GI-BASE-NEXT:    fmov w8, s18
+; CHECK-GI-BASE-NEXT:    fmov w9, s4
+; CHECK-GI-BASE-NEXT:    fmov w10, s5
+; CHECK-GI-BASE-NEXT:    fmov w11, s0
+; CHECK-GI-BASE-NEXT:    fmov w12, s19
+; CHECK-GI-BASE-NEXT:    addv s4, v16.4s
+; CHECK-GI-BASE-NEXT:    addv s5, v17.4s
+; CHECK-GI-BASE-NEXT:    addv s3, v3.4s
+; CHECK-GI-BASE-NEXT:    umull2 v0.4s, v7.8h, v6.8h
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
-; CHECK-GI-BASE-NEXT:    fmov w9, s0
-; CHECK-GI-BASE-NEXT:    addv s0, v1.4s
-; CHECK-GI-BASE-NEXT:    addv s1, v20.4s
-; CHECK-GI-BASE-NEXT:    addv s5, v6.4s
-; CHECK-GI-BASE-NEXT:    add w10, w10, w11
-; CHECK-GI-BASE-NEXT:    fmov w11, s3
+; CHECK-GI-BASE-NEXT:    addv s1, v1.4s
+; CHECK-GI-BASE-NEXT:    add w9, w11, w12
+; CHECK-GI-BASE-NEXT:    add w8, w8, w10
+; CHECK-GI-BASE-NEXT:    fmov w10, s4
+; CHECK-GI-BASE-NEXT:    fmov w11, s5
 ; CHECK-GI-BASE-NEXT:    fmov w12, s2
-; CHECK-GI-BASE-NEXT:    add w8, w8, w9
-; CHECK-GI-BASE-NEXT:    fmov w9, s7
-; CHECK-GI-BASE-NEXT:    add w9, w10, w9
+; CHECK-GI-BASE-NEXT:    addv s4, v20.4s
+; CHECK-GI-BASE-NEXT:    addv s0, v0.4s
+; CHECK-GI-BASE-NEXT:    add w9, w9, w10
 ; CHECK-GI-BASE-NEXT:    add w10, w11, w12
-; CHECK-GI-BASE-NEXT:    fmov w11, s4
+; CHECK-GI-BASE-NEXT:    fmov w11, s3
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
 ; CHECK-GI-BASE-NEXT:    add w9, w10, w11
-; CHECK-GI-BASE-NEXT:    fmov w10, s0
-; CHECK-GI-BASE-NEXT:    fmov w11, s5
-; CHECK-GI-BASE-NEXT:    add w9, w9, w10
 ; CHECK-GI-BASE-NEXT:    fmov w10, s1
+; CHECK-GI-BASE-NEXT:    fmov w11, s0
+; CHECK-GI-BASE-NEXT:    add w9, w9, w10
+; CHECK-GI-BASE-NEXT:    fmov w10, s4
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
 ; CHECK-GI-BASE-NEXT:    add w9, w10, w11
 ; CHECK-GI-BASE-NEXT:    add w0, w8, w9
@@ -2527,65 +2527,65 @@ define i32 @test_sdot_v48i8(ptr %p1, ptr %p2) {
 ; CHECK-GI-BASE-LABEL: test_sdot_v48i8:
 ; CHECK-GI-BASE:       // %bb.0: // %entry
 ; CHECK-GI-BASE-NEXT:    ldp q0, q3, [x1]
-; CHECK-GI-BASE-NEXT:    ldr q6, [x1, #32]
+; CHECK-GI-BASE-NEXT:    ldr q6, [x0, #32]
 ; CHECK-GI-BASE-NEXT:    ldp q1, q2, [x0]
-; CHECK-GI-BASE-NEXT:    ldr q17, [x0, #32]
+; CHECK-GI-BASE-NEXT:    ldr q7, [x1, #32]
+; CHECK-GI-BASE-NEXT:    sshll v20.8h, v6.8b, #0
+; CHECK-GI-BASE-NEXT:    sshll2 v6.8h, v6.16b, #0
 ; CHECK-GI-BASE-NEXT:    sshll v4.8h, v0.8b, #0
 ; CHECK-GI-BASE-NEXT:    sshll2 v0.8h, v0.16b, #0
-; CHECK-GI-BASE-NEXT:    sshll v7.8h, v3.8b, #0
+; CHECK-GI-BASE-NEXT:    sshll v16.8h, v3.8b, #0
 ; CHECK-GI-BASE-NEXT:    sshll v5.8h, v1.8b, #0
 ; CHECK-GI-BASE-NEXT:    sshll2 v1.8h, v1.16b, #0
-; CHECK-GI-BASE-NEXT:    sshll v16.8h, v2.8b, #0
+; CHECK-GI-BASE-NEXT:    sshll v17.8h, v2.8b, #0
 ; CHECK-GI-BASE-NEXT:    sshll2 v3.8h, v3.16b, #0
 ; CHECK-GI-BASE-NEXT:    sshll2 v2.8h, v2.16b, #0
 ; CHECK-GI-BASE-NEXT:    smull v18.4s, v4.4h, v5.4h
 ; CHECK-GI-BASE-NEXT:    smull2 v4.4s, v4.8h, v5.8h
-; CHECK-GI-BASE-NEXT:    smull2 v19.4s, v0.8h, v1.8h
-; CHECK-GI-BASE-NEXT:    smull v20.4s, v7.4h, v16.4h
-; CHECK-GI-BASE-NEXT:    smull v0.4s, v0.4h, v1.4h
-; CHECK-GI-BASE-NEXT:    sshll v5.8h, v6.8b, #0
-; CHECK-GI-BASE-NEXT:    sshll v1.8h, v17.8b, #0
-; CHECK-GI-BASE-NEXT:    smull2 v7.4s, v7.8h, v16.8h
-; CHECK-GI-BASE-NEXT:    sshll2 v6.8h, v6.16b, #0
-; CHECK-GI-BASE-NEXT:    sshll2 v17.8h, v17.16b, #0
-; CHECK-GI-BASE-NEXT:    addv s16, v18.4s
-; CHECK-GI-BASE-NEXT:    addv s4, v4.4s
-; CHECK-GI-BASE-NEXT:    smull v18.4s, v3.4h, v2.4h
+; CHECK-GI-BASE-NEXT:    smull v5.4s, v0.4h, v1.4h
+; CHECK-GI-BASE-NEXT:    smull2 v0.4s, v0.8h, v1.8h
+; CHECK-GI-BASE-NEXT:    smull v19.4s, v16.4h, v17.4h
+; CHECK-GI-BASE-NEXT:    sshll v1.8h, v7.8b, #0
+; CHECK-GI-BASE-NEXT:    smull2 v16.4s, v16.8h, v17.8h
+; CHECK-GI-BASE-NEXT:    smull v17.4s, v3.4h, v2.4h
 ; CHECK-GI-BASE-NEXT:    smull2 v2.4s, v3.8h, v2.8h
-; CHECK-GI-BASE-NEXT:    addv s3, v19.4s
-; CHECK-GI-BASE-NEXT:    smull v19.4s, v5.4h, v1.4h
-; CHECK-GI-BASE-NEXT:    smull2 v1.4s, v5.8h, v1.8h
-; CHECK-GI-BASE-NEXT:    addv s5, v20.4s
+; CHECK-GI-BASE-NEXT:    sshll2 v7.8h, v7.16b, #0
+; CHECK-GI-BASE-NEXT:    addv s18, v18.4s
+; CHECK-GI-BASE-NEXT:    addv s4, v4.4s
+; CHECK-GI-BASE-NEXT:    addv s5, v5.4s
 ; CHECK-GI-BASE-NEXT:    addv s0, v0.4s
-; CHECK-GI-BASE-NEXT:    addv s7, v7.4s
-; CHECK-GI-BASE-NEXT:    smull v20.4s, v6.4h, v17.4h
-; CHECK-GI-BASE-NEXT:    smull2 v6.4s, v6.8h, v17.8h
-; CHECK-GI-BASE-NEXT:    fmov w8, s16
-; CHECK-GI-BASE-NEXT:    fmov w9, s4
-; CHECK-GI-BASE-NEXT:    fmov w10, s3
-; CHECK-GI-BASE-NEXT:    addv s3, v18.4s
+; CHECK-GI-BASE-NEXT:    addv s19, v19.4s
+; CHECK-GI-BASE-NEXT:    smull v3.4s, v1.4h, v20.4h
 ; CHECK-GI-BASE-NEXT:    addv s2, v2.4s
-; CHECK-GI-BASE-NEXT:    fmov w11, s5
-; CHECK-GI-BASE-NEXT:    addv s4, v19.4s
+; CHECK-GI-BASE-NEXT:    smull2 v1.4s, v1.8h, v20.8h
+; CHECK-GI-BASE-NEXT:    smull v20.4s, v7.4h, v6.4h
+; CHECK-GI-BASE-NEXT:    fmov w8, s18
+; CHECK-GI-BASE-NEXT:    fmov w9, s4
+; CHECK-GI-BASE-NEXT:    fmov w10, s5
+; CHECK-GI-BASE-NEXT:    fmov w11, s0
+; CHECK-GI-BASE-NEXT:    fmov w12, s19
+; CHECK-GI-BASE-NEXT:    addv s4, v16.4s
+; CHECK-GI-BASE-NEXT:    addv s5, v17.4s
+; CHECK-GI-BASE-NEXT:    addv s3, v3.4s
+; CHECK-GI-BASE-NEXT:    smull2 v0.4s, v7.8h, v6.8h
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
-; CHECK-GI-BASE-NEXT:    fmov w9, s0
-; CHECK-GI-BASE-NEXT:    addv s0, v1.4s
-; CHECK-GI-BASE-NEXT:    addv s1, v20.4s
-; CHECK-GI-BASE-NEXT:    addv s5, v6.4s
-; CHECK-GI-BASE-NEXT:    add w10, w10, w11
-; CHECK-GI-BASE-NEXT:    fmov w11, s3
+; CHECK-GI-BASE-NEXT:    addv s1, v1.4s
+; CHECK-GI-BASE-NEXT:    add w9, w11, w12
+; CHECK-GI-BASE-NEXT:    add w8, w8, w10
+; CHECK-GI-BASE-NEXT:    fmov w10, s4
+; CHECK-GI-BASE-NEXT:    fmov w11, s5
 ; CHECK-GI-BASE-NEXT:    fmov w12, s2
-; CHECK-GI-BASE-NEXT:    add w8, w8, w9
-; CHECK-GI-BASE-NEXT:    fmov w9, s7
-; CHECK-GI-BASE-NEXT:    add w9, w10, w9
+; CHECK-GI-BASE-NEXT:    addv s4, v20.4s
+; CHECK-GI-BASE-NEXT:    addv s0, v0.4s
+; CHECK-GI-BASE-NEXT:    add w9, w9, w10
 ; CHECK-GI-BASE-NEXT:    add w10, w11, w12
-; CHECK-GI-BASE-NEXT:    fmov w11, s4
+; CHECK-GI-BASE-NEXT:    fmov w11, s3
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
 ; CHECK-GI-BASE-NEXT:    add w9, w10, w11
-; CHECK-GI-BASE-NEXT:    fmov w10, s0
-; CHECK-GI-BASE-NEXT:    fmov w11, s5
-; CHECK-GI-BASE-NEXT:    add w9, w9, w10
 ; CHECK-GI-BASE-NEXT:    fmov w10, s1
+; CHECK-GI-BASE-NEXT:    fmov w11, s0
+; CHECK-GI-BASE-NEXT:    add w9, w9, w10
+; CHECK-GI-BASE-NEXT:    fmov w10, s4
 ; CHECK-GI-BASE-NEXT:    add w8, w8, w9
 ; CHECK-GI-BASE-NEXT:    add w9, w10, w11
 ; CHECK-GI-BASE-NEXT:    add w0, w8, w9
@@ -2640,13 +2640,13 @@ define i32 @test_udot_v8i8_multi_use(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-SD-DOT-LABEL: test_udot_v8i8_multi_use:
 ; CHECK-SD-DOT:       // %bb.0: // %entry
 ; CHECK-SD-DOT-NEXT:    movi v2.2d, #0000000000000000
+; CHECK-SD-DOT-NEXT:    ushll v3.8h, v0.8b, #0
+; CHECK-SD-DOT-NEXT:    ushll v4.8h, v1.8b, #0
 ; CHECK-SD-DOT-NEXT:    udot v2.2s, v1.8b, v0.8b
-; CHECK-SD-DOT-NEXT:    ushll v0.8h, v0.8b, #0
-; CHECK-SD-DOT-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-SD-DOT-NEXT:    umull v0.4s, v1.4h, v0.4h
-; CHECK-SD-DOT-NEXT:    addp v2.2s, v2.2s, v2.2s
+; CHECK-SD-DOT-NEXT:    umull v0.4s, v4.4h, v3.4h
+; CHECK-SD-DOT-NEXT:    addp v1.2s, v2.2s, v2.2s
 ; CHECK-SD-DOT-NEXT:    fmov w9, s0
-; CHECK-SD-DOT-NEXT:    fmov w8, s2
+; CHECK-SD-DOT-NEXT:    fmov w8, s1
 ; CHECK-SD-DOT-NEXT:    add w0, w8, w9
 ; CHECK-SD-DOT-NEXT:    ret
 ;
@@ -3534,21 +3534,21 @@ entry:
 define i64 @add_pair_v4i8_v4i64_sext(<4 x i8> %x, <4 x i8> %y) {
 ; CHECK-SD-LABEL: add_pair_v4i8_v4i64_sext:
 ; CHECK-SD:       // %bb.0: // %entry
-; CHECK-SD-NEXT:    ushll v0.4s, v0.4h, #0
 ; CHECK-SD-NEXT:    ushll v1.4s, v1.4h, #0
-; CHECK-SD-NEXT:    ushll v2.2d, v0.2s, #0
-; CHECK-SD-NEXT:    ushll v3.2d, v1.2s, #0
+; CHECK-SD-NEXT:    ushll v0.4s, v0.4h, #0
+; CHECK-SD-NEXT:    ushll v2.2d, v1.2s, #0
+; CHECK-SD-NEXT:    ushll v3.2d, v0.2s, #0
 ; CHECK-SD-NEXT:    ushll2 v0.2d, v0.4s, #0
 ; CHECK-SD-NEXT:    ushll2 v1.2d, v1.4s, #0
-; CHECK-SD-NEXT:    shl v2.2d, v2.2d, #56
 ; CHECK-SD-NEXT:    shl v3.2d, v3.2d, #56
+; CHECK-SD-NEXT:    shl v2.2d, v2.2d, #56
 ; CHECK-SD-NEXT:    shl v0.2d, v0.2d, #56
 ; CHECK-SD-NEXT:    shl v1.2d, v1.2d, #56
-; CHECK-SD-NEXT:    sshr v2.2d, v2.2d, #56
 ; CHECK-SD-NEXT:    sshr v3.2d, v3.2d, #56
-; CHECK-SD-NEXT:    ssra v2.2d, v0.2d, #56
-; CHECK-SD-NEXT:    ssra v3.2d, v1.2d, #56
-; CHECK-SD-NEXT:    add v0.2d, v2.2d, v3.2d
+; CHECK-SD-NEXT:    sshr v2.2d, v2.2d, #56
+; CHECK-SD-NEXT:    ssra v3.2d, v0.2d, #56
+; CHECK-SD-NEXT:    ssra v2.2d, v1.2d, #56
+; CHECK-SD-NEXT:    add v0.2d, v3.2d, v2.2d
 ; CHECK-SD-NEXT:    addp d0, v0.2d
 ; CHECK-SD-NEXT:    fmov x0, d0
 ; CHECK-SD-NEXT:    ret
@@ -3816,37 +3816,37 @@ define i16 @add_v24i8_v24i16_zext(<24 x i8> %x) {
 ; CHECK-SD-NEXT:    ldr b1, [sp, #64]
 ; CHECK-SD-NEXT:    add x8, sp, #72
 ; CHECK-SD-NEXT:    ldr b2, [sp]
-; CHECK-SD-NEXT:    add x9, sp, #8
+; CHECK-SD-NEXT:    add x9, sp, #80
 ; CHECK-SD-NEXT:    ld1 { v1.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #80
+; CHECK-SD-NEXT:    add x8, sp, #8
 ; CHECK-SD-NEXT:    mov v0.b[1], w1
-; CHECK-SD-NEXT:    ld1 { v2.b }[1], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #16
-; CHECK-SD-NEXT:    ld1 { v1.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #88
-; CHECK-SD-NEXT:    ld1 { v2.b }[2], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #24
+; CHECK-SD-NEXT:    ld1 { v2.b }[1], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #16
+; CHECK-SD-NEXT:    ld1 { v1.b }[2], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #88
+; CHECK-SD-NEXT:    ld1 { v2.b }[2], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #24
 ; CHECK-SD-NEXT:    mov v0.b[2], w2
-; CHECK-SD-NEXT:    ld1 { v1.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #96
-; CHECK-SD-NEXT:    ld1 { v2.b }[3], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #32
+; CHECK-SD-NEXT:    ld1 { v1.b }[3], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #96
+; CHECK-SD-NEXT:    ld1 { v2.b }[3], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #32
 ; CHECK-SD-NEXT:    mov v0.b[3], w3
-; CHECK-SD-NEXT:    ld1 { v1.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #104
-; CHECK-SD-NEXT:    ld1 { v2.b }[4], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #40
-; CHECK-SD-NEXT:    ld1 { v1.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #112
+; CHECK-SD-NEXT:    ld1 { v1.b }[4], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #104
+; CHECK-SD-NEXT:    ld1 { v2.b }[4], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #40
+; CHECK-SD-NEXT:    ld1 { v1.b }[5], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #112
 ; CHECK-SD-NEXT:    mov v0.b[4], w4
-; CHECK-SD-NEXT:    ld1 { v2.b }[5], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #48
-; CHECK-SD-NEXT:    ld1 { v1.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #120
-; CHECK-SD-NEXT:    ld1 { v2.b }[6], [x9]
-; CHECK-SD-NEXT:    mov v0.b[5], w5
-; CHECK-SD-NEXT:    ld1 { v1.b }[7], [x8]
+; CHECK-SD-NEXT:    ld1 { v2.b }[5], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #48
+; CHECK-SD-NEXT:    ld1 { v1.b }[6], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #120
+; CHECK-SD-NEXT:    ld1 { v2.b }[6], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #56
+; CHECK-SD-NEXT:    mov v0.b[5], w5
+; CHECK-SD-NEXT:    ld1 { v1.b }[7], [x9]
 ; CHECK-SD-NEXT:    ld1 { v2.b }[7], [x8]
 ; CHECK-SD-NEXT:    mov v0.b[6], w6
 ; CHECK-SD-NEXT:    mov v0.b[7], w7
@@ -3942,37 +3942,37 @@ define i16 @add_v24i8_v24i16_sext(<24 x i8> %x) {
 ; CHECK-SD-NEXT:    ldr b1, [sp, #64]
 ; CHECK-SD-NEXT:    add x8, sp, #72
 ; CHECK-SD-NEXT:    ldr b2, [sp]
-; CHECK-SD-NEXT:    add x9, sp, #8
+; CHECK-SD-NEXT:    add x9, sp, #80
 ; CHECK-SD-NEXT:    ld1 { v1.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #80
+; CHECK-SD-NEXT:    add x8, sp, #8
 ; CHECK-SD-NEXT:    mov v0.b[1], w1
-; CHECK-SD-NEXT:    ld1 { v2.b }[1], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #16
-; CHECK-SD-NEXT:    ld1 { v1.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #88
-; CHECK-SD-NEXT:    ld1 { v2.b }[2], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #24
+; CHECK-SD-NEXT:    ld1 { v2.b }[1], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #16
+; CHECK-SD-NEXT:    ld1 { v1.b }[2], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #88
+; CHECK-SD-NEXT:    ld1 { v2.b }[2], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #24
 ; CHECK-SD-NEXT:    mov v0.b[2], w2
-; CHECK-SD-NEXT:    ld1 { v1.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #96
-; CHECK-SD-NEXT:    ld1 { v2.b }[3], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #32
+; CHECK-SD-NEXT:    ld1 { v1.b }[3], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #96
+; CHECK-SD-NEXT:    ld1 { v2.b }[3], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #32
 ; CHECK-SD-NEXT:    mov v0.b[3], w3
-; CHECK-SD-NEXT:    ld1 { v1.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #104
-; CHECK-SD-NEXT:    ld1 { v2.b }[4], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #40
-; CHECK-SD-NEXT:    ld1 { v1.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #112
+; CHECK-SD-NEXT:    ld1 { v1.b }[4], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #104
+; CHECK-SD-NEXT:    ld1 { v2.b }[4], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #40
+; CHECK-SD-NEXT:    ld1 { v1.b }[5], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #112
 ; CHECK-SD-NEXT:    mov v0.b[4], w4
-; CHECK-SD-NEXT:    ld1 { v2.b }[5], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #48
-; CHECK-SD-NEXT:    ld1 { v1.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #120
-; CHECK-SD-NEXT:    ld1 { v2.b }[6], [x9]
-; CHECK-SD-NEXT:    mov v0.b[5], w5
-; CHECK-SD-NEXT:    ld1 { v1.b }[7], [x8]
+; CHECK-SD-NEXT:    ld1 { v2.b }[5], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #48
+; CHECK-SD-NEXT:    ld1 { v1.b }[6], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #120
+; CHECK-SD-NEXT:    ld1 { v2.b }[6], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #56
+; CHECK-SD-NEXT:    mov v0.b[5], w5
+; CHECK-SD-NEXT:    ld1 { v1.b }[7], [x9]
 ; CHECK-SD-NEXT:    ld1 { v2.b }[7], [x8]
 ; CHECK-SD-NEXT:    mov v0.b[6], w6
 ; CHECK-SD-NEXT:    mov v0.b[7], w7
@@ -4069,48 +4069,48 @@ define i32 @add_v24i8_v24i32_zext(<24 x i8> %x) {
 ; CHECK-SD-BASE-NEXT:    ldr b1, [sp, #64]
 ; CHECK-SD-BASE-NEXT:    add x8, sp, #72
 ; CHECK-SD-BASE-NEXT:    ldr b2, [sp]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #8
+; CHECK-SD-BASE-NEXT:    add x9, sp, #80
 ; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[1], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #80
+; CHECK-SD-BASE-NEXT:    add x8, sp, #8
 ; CHECK-SD-BASE-NEXT:    mov v0.b[1], w1
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[1], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #16
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[2], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #88
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[2], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #24
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[1], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #16
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[2], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #88
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[2], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #24
 ; CHECK-SD-BASE-NEXT:    mov v0.b[2], w2
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[3], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #96
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[3], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #32
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[3], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #96
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[3], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #32
 ; CHECK-SD-BASE-NEXT:    mov v0.b[3], w3
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[4], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #104
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[4], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #40
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[5], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #112
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[4], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #104
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[4], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #40
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[5], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #112
 ; CHECK-SD-BASE-NEXT:    mov v0.b[4], w4
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[5], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #48
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[6], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #120
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[6], [x9]
-; CHECK-SD-BASE-NEXT:    mov v0.b[5], w5
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[7], [x8]
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[5], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #48
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[6], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #120
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[6], [x8]
 ; CHECK-SD-BASE-NEXT:    add x8, sp, #56
+; CHECK-SD-BASE-NEXT:    mov v0.b[5], w5
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[7], [x9]
 ; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[7], [x8]
 ; CHECK-SD-BASE-NEXT:    mov v0.b[6], w6
 ; CHECK-SD-BASE-NEXT:    ushll v1.8h, v1.8b, #0
+; CHECK-SD-BASE-NEXT:    ushll v2.8h, v2.8b, #0
 ; CHECK-SD-BASE-NEXT:    mov v0.b[7], w7
 ; CHECK-SD-BASE-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-SD-BASE-NEXT:    uaddl2 v3.4s, v0.8h, v1.8h
 ; CHECK-SD-BASE-NEXT:    uaddl v0.4s, v0.4h, v1.4h
-; CHECK-SD-BASE-NEXT:    ushll v1.8h, v2.8b, #0
-; CHECK-SD-BASE-NEXT:    uaddw2 v2.4s, v3.4s, v1.8h
-; CHECK-SD-BASE-NEXT:    uaddw v0.4s, v0.4s, v1.4h
-; CHECK-SD-BASE-NEXT:    add v0.4s, v0.4s, v2.4s
+; CHECK-SD-BASE-NEXT:    uaddw2 v1.4s, v3.4s, v2.8h
+; CHECK-SD-BASE-NEXT:    uaddw v0.4s, v0.4s, v2.4h
+; CHECK-SD-BASE-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-SD-BASE-NEXT:    addv s0, v0.4s
 ; CHECK-SD-BASE-NEXT:    fmov w0, s0
 ; CHECK-SD-BASE-NEXT:    ret
@@ -4147,9 +4147,9 @@ define i32 @add_v24i8_v24i32_zext(<24 x i8> %x) {
 ; CHECK-SD-DOT-NEXT:    udot v4.2s, v1.8b, v5.8b
 ; CHECK-SD-DOT-NEXT:    mov v0.b[7], w7
 ; CHECK-SD-DOT-NEXT:    addp v1.2s, v4.2s, v4.2s
+; CHECK-SD-DOT-NEXT:    fmov w9, s1
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[8], [x8]
 ; CHECK-SD-DOT-NEXT:    add x8, sp, #8
-; CHECK-SD-DOT-NEXT:    fmov w9, s1
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[9], [x8]
 ; CHECK-SD-DOT-NEXT:    add x8, sp, #16
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[10], [x8]
@@ -4342,48 +4342,48 @@ define i32 @add_v24i8_v24i32_sext(<24 x i8> %x) {
 ; CHECK-SD-BASE-NEXT:    ldr b1, [sp, #64]
 ; CHECK-SD-BASE-NEXT:    add x8, sp, #72
 ; CHECK-SD-BASE-NEXT:    ldr b2, [sp]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #8
+; CHECK-SD-BASE-NEXT:    add x9, sp, #80
 ; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[1], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #80
+; CHECK-SD-BASE-NEXT:    add x8, sp, #8
 ; CHECK-SD-BASE-NEXT:    mov v0.b[1], w1
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[1], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #16
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[2], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #88
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[2], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #24
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[1], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #16
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[2], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #88
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[2], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #24
 ; CHECK-SD-BASE-NEXT:    mov v0.b[2], w2
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[3], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #96
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[3], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #32
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[3], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #96
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[3], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #32
 ; CHECK-SD-BASE-NEXT:    mov v0.b[3], w3
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[4], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #104
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[4], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #40
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[5], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #112
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[4], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #104
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[4], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #40
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[5], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #112
 ; CHECK-SD-BASE-NEXT:    mov v0.b[4], w4
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[5], [x9]
-; CHECK-SD-BASE-NEXT:    add x9, sp, #48
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[6], [x8]
-; CHECK-SD-BASE-NEXT:    add x8, sp, #120
-; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[6], [x9]
-; CHECK-SD-BASE-NEXT:    mov v0.b[5], w5
-; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[7], [x8]
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[5], [x8]
+; CHECK-SD-BASE-NEXT:    add x8, sp, #48
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[6], [x9]
+; CHECK-SD-BASE-NEXT:    add x9, sp, #120
+; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[6], [x8]
 ; CHECK-SD-BASE-NEXT:    add x8, sp, #56
+; CHECK-SD-BASE-NEXT:    mov v0.b[5], w5
+; CHECK-SD-BASE-NEXT:    ld1 { v1.b }[7], [x9]
 ; CHECK-SD-BASE-NEXT:    ld1 { v2.b }[7], [x8]
 ; CHECK-SD-BASE-NEXT:    mov v0.b[6], w6
 ; CHECK-SD-BASE-NEXT:    sshll v1.8h, v1.8b, #0
+; CHECK-SD-BASE-NEXT:    sshll v2.8h, v2.8b, #0
 ; CHECK-SD-BASE-NEXT:    mov v0.b[7], w7
 ; CHECK-SD-BASE-NEXT:    sshll v0.8h, v0.8b, #0
 ; CHECK-SD-BASE-NEXT:    saddl2 v3.4s, v0.8h, v1.8h
 ; CHECK-SD-BASE-NEXT:    saddl v0.4s, v0.4h, v1.4h
-; CHECK-SD-BASE-NEXT:    sshll v1.8h, v2.8b, #0
-; CHECK-SD-BASE-NEXT:    saddw2 v2.4s, v3.4s, v1.8h
-; CHECK-SD-BASE-NEXT:    saddw v0.4s, v0.4s, v1.4h
-; CHECK-SD-BASE-NEXT:    add v0.4s, v0.4s, v2.4s
+; CHECK-SD-BASE-NEXT:    saddw2 v1.4s, v3.4s, v2.8h
+; CHECK-SD-BASE-NEXT:    saddw v0.4s, v0.4s, v2.4h
+; CHECK-SD-BASE-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-SD-BASE-NEXT:    addv s0, v0.4s
 ; CHECK-SD-BASE-NEXT:    fmov w0, s0
 ; CHECK-SD-BASE-NEXT:    ret
@@ -4420,9 +4420,9 @@ define i32 @add_v24i8_v24i32_sext(<24 x i8> %x) {
 ; CHECK-SD-DOT-NEXT:    sdot v4.2s, v1.8b, v5.8b
 ; CHECK-SD-DOT-NEXT:    mov v0.b[7], w7
 ; CHECK-SD-DOT-NEXT:    addp v1.2s, v4.2s, v4.2s
+; CHECK-SD-DOT-NEXT:    fmov w9, s1
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[8], [x8]
 ; CHECK-SD-DOT-NEXT:    add x8, sp, #8
-; CHECK-SD-DOT-NEXT:    fmov w9, s1
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[9], [x8]
 ; CHECK-SD-DOT-NEXT:    add x8, sp, #16
 ; CHECK-SD-DOT-NEXT:    ld1 { v0.b }[10], [x8]
@@ -4611,23 +4611,23 @@ entry:
 define i32 @full(ptr %p1, i32 noundef %s1, ptr %p2, i32 noundef %s2) {
 ; CHECK-SD-BASE-LABEL: full:
 ; CHECK-SD-BASE:       // %bb.0: // %entry
-; CHECK-SD-BASE-NEXT:    ldr d0, [x2]
-; CHECK-SD-BASE-NEXT:    ldr d1, [x0]
 ; CHECK-SD-BASE-NEXT:    // kill: def $w3 killed $w3 def $x3
 ; CHECK-SD-BASE-NEXT:    // kill: def $w1 killed $w1 def $x1
 ; CHECK-SD-BASE-NEXT:    sxtw x8, w3
 ; CHECK-SD-BASE-NEXT:    sxtw x9, w1
-; CHECK-SD-BASE-NEXT:    uabdl v0.8h, v1.8b, v0.8b
-; CHECK-SD-BASE-NEXT:    add x11, x2, x8
+; CHECK-SD-BASE-NEXT:    ldr d0, [x0]
+; CHECK-SD-BASE-NEXT:    ldr d1, [x2]
 ; CHECK-SD-BASE-NEXT:    add x10, x0, x9
-; CHECK-SD-BASE-NEXT:    ldr d2, [x11]
-; CHECK-SD-BASE-NEXT:    add x11, x11, x8
+; CHECK-SD-BASE-NEXT:    add x11, x2, x8
+; CHECK-SD-BASE-NEXT:    uabdl v0.8h, v0.8b, v1.8b
 ; CHECK-SD-BASE-NEXT:    ldr d1, [x10]
+; CHECK-SD-BASE-NEXT:    ldr d2, [x11]
 ; CHECK-SD-BASE-NEXT:    add x10, x10, x9
-; CHECK-SD-BASE-NEXT:    uaddlp v0.4s, v0.8h
+; CHECK-SD-BASE-NEXT:    add x11, x11, x8
 ; CHECK-SD-BASE-NEXT:    uabdl v1.8h, v1.8b, v2.8b
 ; CHECK-SD-BASE-NEXT:    ldr d2, [x11]
 ; CHECK-SD-BASE-NEXT:    add x11, x11, x8
+; CHECK-SD-BASE-NEXT:    uaddlp v0.4s, v0.8h
 ; CHECK-SD-BASE-NEXT:    uadalp v0.4s, v1.8h
 ; CHECK-SD-BASE-NEXT:    ldr d1, [x10]
 ; CHECK-SD-BASE-NEXT:    add x10, x10, x9
@@ -4723,98 +4723,98 @@ define i32 @full(ptr %p1, i32 noundef %s1, ptr %p2, i32 noundef %s2) {
 ; CHECK-GI:       // %bb.0: // %entry
 ; CHECK-GI-NEXT:    // kill: def $w1 killed $w1 def $x1
 ; CHECK-GI-NEXT:    // kill: def $w3 killed $w3 def $x3
-; CHECK-GI-NEXT:    sxtw x8, w1
-; CHECK-GI-NEXT:    sxtw x9, w3
+; CHECK-GI-NEXT:    sxtw x8, w3
+; CHECK-GI-NEXT:    sxtw x9, w1
 ; CHECK-GI-NEXT:    ldr d0, [x0]
 ; CHECK-GI-NEXT:    ldr d1, [x2]
-; CHECK-GI-NEXT:    add x10, x0, x8
-; CHECK-GI-NEXT:    add x11, x2, x9
+; CHECK-GI-NEXT:    add x10, x0, x9
+; CHECK-GI-NEXT:    add x11, x2, x8
 ; CHECK-GI-NEXT:    ushll v0.8h, v0.8b, #0
 ; CHECK-GI-NEXT:    ldr d2, [x10]
-; CHECK-GI-NEXT:    ldr d3, [x11]
+; CHECK-GI-NEXT:    add x10, x10, x9
+; CHECK-GI-NEXT:    add x12, x11, x8
 ; CHECK-GI-NEXT:    ushll v1.8h, v1.8b, #0
-; CHECK-GI-NEXT:    add x10, x10, x8
-; CHECK-GI-NEXT:    add x11, x11, x9
+; CHECK-GI-NEXT:    ldr d3, [x11]
+; CHECK-GI-NEXT:    ldr d4, [x10]
+; CHECK-GI-NEXT:    ldr d5, [x12]
+; CHECK-GI-NEXT:    add x10, x10, x9
+; CHECK-GI-NEXT:    add x11, x12, x8
 ; CHECK-GI-NEXT:    ushll v2.8h, v2.8b, #0
 ; CHECK-GI-NEXT:    ushll v3.8h, v3.8b, #0
-; CHECK-GI-NEXT:    ldr d4, [x10]
-; CHECK-GI-NEXT:    ldr d5, [x11]
-; CHECK-GI-NEXT:    add x10, x10, x8
-; CHECK-GI-NEXT:    add x11, x11, x9
+; CHECK-GI-NEXT:    ushll v4.8h, v4.8b, #0
+; CHECK-GI-NEXT:    ushll v5.8h, v5.8b, #0
 ; CHECK-GI-NEXT:    uabdl v6.4s, v0.4h, v1.4h
 ; CHECK-GI-NEXT:    uabdl2 v0.4s, v0.8h, v1.8h
 ; CHECK-GI-NEXT:    ldr d1, [x10]
-; CHECK-GI-NEXT:    ushll v4.8h, v4.8b, #0
-; CHECK-GI-NEXT:    ushll v5.8h, v5.8b, #0
 ; CHECK-GI-NEXT:    ldr d7, [x11]
+; CHECK-GI-NEXT:    add x10, x10, x9
+; CHECK-GI-NEXT:    add x11, x11, x8
 ; CHECK-GI-NEXT:    uabdl v16.4s, v2.4h, v3.4h
 ; CHECK-GI-NEXT:    uabdl2 v2.4s, v2.8h, v3.8h
-; CHECK-GI-NEXT:    ushll v3.8h, v1.8b, #0
-; CHECK-GI-NEXT:    ushll v7.8h, v7.8b, #0
-; CHECK-GI-NEXT:    add x10, x10, x8
-; CHECK-GI-NEXT:    add x11, x11, x9
-; CHECK-GI-NEXT:    uabdl v1.4s, v4.4h, v5.4h
+; CHECK-GI-NEXT:    uabdl v3.4s, v4.4h, v5.4h
 ; CHECK-GI-NEXT:    uabdl2 v4.4s, v4.8h, v5.8h
+; CHECK-GI-NEXT:    ushll v1.8h, v1.8b, #0
+; CHECK-GI-NEXT:    ushll v7.8h, v7.8b, #0
 ; CHECK-GI-NEXT:    ldr d5, [x10]
-; CHECK-GI-NEXT:    add v2.4s, v16.4s, v2.4s
-; CHECK-GI-NEXT:    ldr d16, [x11]
+; CHECK-GI-NEXT:    ldr d17, [x11]
+; CHECK-GI-NEXT:    add x10, x10, x9
+; CHECK-GI-NEXT:    add x11, x11, x8
 ; CHECK-GI-NEXT:    add v0.4s, v6.4s, v0.4s
-; CHECK-GI-NEXT:    uabdl v6.4s, v3.4h, v7.4h
-; CHECK-GI-NEXT:    uabdl2 v3.4s, v3.8h, v7.8h
 ; CHECK-GI-NEXT:    ushll v5.8h, v5.8b, #0
-; CHECK-GI-NEXT:    add x10, x10, x8
-; CHECK-GI-NEXT:    ushll v7.8h, v16.8b, #0
-; CHECK-GI-NEXT:    add x11, x11, x9
-; CHECK-GI-NEXT:    ldr d16, [x10]
-; CHECK-GI-NEXT:    ldr d17, [x11]
-; CHECK-GI-NEXT:    add v1.4s, v1.4s, v4.4s
-; CHECK-GI-NEXT:    add x10, x10, x8
-; CHECK-GI-NEXT:    add x11, x11, x9
-; CHECK-GI-NEXT:    add v3.4s, v6.4s, v3.4s
-; CHECK-GI-NEXT:    ushll v16.8h, v16.8b, #0
 ; CHECK-GI-NEXT:    ushll v17.8h, v17.8b, #0
-; CHECK-GI-NEXT:    uabdl v22.4s, v5.4h, v7.4h
-; CHECK-GI-NEXT:    uabdl2 v5.4s, v5.8h, v7.8h
+; CHECK-GI-NEXT:    add v2.4s, v16.4s, v2.4s
+; CHECK-GI-NEXT:    add v3.4s, v3.4s, v4.4s
+; CHECK-GI-NEXT:    uabdl v4.4s, v1.4h, v7.4h
+; CHECK-GI-NEXT:    uabdl2 v1.4s, v1.8h, v7.8h
+; CHECK-GI-NEXT:    ldr d7, [x10]
+; CHECK-GI-NEXT:    ldr d16, [x11]
+; CHECK-GI-NEXT:    add x10, x10, x9
+; CHECK-GI-NEXT:    add x11, x11, x8
 ; CHECK-GI-NEXT:    ldr d18, [x10]
+; CHECK-GI-NEXT:    ldr d20, [x10, x9]
 ; CHECK-GI-NEXT:    ldr d19, [x11]
-; CHECK-GI-NEXT:    addv s0, v0.4s
+; CHECK-GI-NEXT:    ldr d21, [x11, x8]
+; CHECK-GI-NEXT:    uabdl v6.4s, v5.4h, v17.4h
+; CHECK-GI-NEXT:    ushll v7.8h, v7.8b, #0
+; CHECK-GI-NEXT:    ushll v16.8h, v16.8b, #0
+; CHECK-GI-NEXT:    uabdl2 v5.4s, v5.8h, v17.8h
+; CHECK-GI-NEXT:    ushll v17.8h, v18.8b, #0
+; CHECK-GI-NEXT:    ushll v18.8h, v19.8b, #0
+; CHECK-GI-NEXT:    add v1.4s, v4.4s, v1.4s
+; CHECK-GI-NEXT:    ushll v4.8h, v20.8b, #0
+; CHECK-GI-NEXT:    ushll v19.8h, v21.8b, #0
 ; CHECK-GI-NEXT:    addv s2, v2.4s
-; CHECK-GI-NEXT:    addv s1, v1.4s
-; CHECK-GI-NEXT:    ushll v18.8h, v18.8b, #0
-; CHECK-GI-NEXT:    ushll v19.8h, v19.8b, #0
-; CHECK-GI-NEXT:    uabdl v4.4s, v16.4h, v17.4h
-; CHECK-GI-NEXT:    uabdl2 v16.4s, v16.8h, v17.8h
-; CHECK-GI-NEXT:    add v5.4s, v22.4s, v5.4s
-; CHECK-GI-NEXT:    ldr d20, [x10, x8]
-; CHECK-GI-NEXT:    ldr d21, [x11, x9]
+; CHECK-GI-NEXT:    addv s0, v0.4s
 ; CHECK-GI-NEXT:    addv s3, v3.4s
+; CHECK-GI-NEXT:    uabdl v20.4s, v7.4h, v16.4h
+; CHECK-GI-NEXT:    uabdl2 v7.4s, v7.8h, v16.8h
+; CHECK-GI-NEXT:    add v5.4s, v6.4s, v5.4s
+; CHECK-GI-NEXT:    uabdl v6.4s, v17.4h, v18.4h
+; CHECK-GI-NEXT:    uabdl2 v16.4s, v17.8h, v18.8h
+; CHECK-GI-NEXT:    uabdl v17.4s, v4.4h, v19.4h
+; CHECK-GI-NEXT:    uabdl2 v4.4s, v4.8h, v19.8h
 ; CHECK-GI-NEXT:    fmov w8, s2
+; CHECK-GI-NEXT:    addv s1, v1.4s
 ; CHECK-GI-NEXT:    fmov w9, s0
-; CHECK-GI-NEXT:    ushll v7.8h, v20.8b, #0
-; CHECK-GI-NEXT:    ushll v20.8h, v21.8b, #0
-; CHECK-GI-NEXT:    uabdl v6.4s, v18.4h, v19.4h
-; CHECK-GI-NEXT:    uabdl2 v17.4s, v18.8h, v19.8h
-; CHECK-GI-NEXT:    add v4.4s, v4.4s, v16.4s
-; CHECK-GI-NEXT:    addv s5, v5.4s
-; CHECK-GI-NEXT:    fmov w10, s1
+; CHECK-GI-NEXT:    fmov w10, s3
+; CHECK-GI-NEXT:    add v7.4s, v20.4s, v7.4s
+; CHECK-GI-NEXT:    add v0.4s, v17.4s, v4.4s
+; CHECK-GI-NEXT:    addv s4, v5.4s
+; CHECK-GI-NEXT:    add v2.4s, v6.4s, v16.4s
 ; CHECK-GI-NEXT:    add w8, w8, w9
-; CHECK-GI-NEXT:    fmov w9, s3
-; CHECK-GI-NEXT:    uabdl v18.4s, v7.4h, v20.4h
-; CHECK-GI-NEXT:    uabdl2 v7.4s, v7.8h, v20.8h
-; CHECK-GI-NEXT:    add v6.4s, v6.4s, v17.4s
+; CHECK-GI-NEXT:    fmov w9, s1
 ; CHECK-GI-NEXT:    add w8, w10, w8
-; CHECK-GI-NEXT:    addv s0, v4.4s
+; CHECK-GI-NEXT:    addv s3, v7.4s
+; CHECK-GI-NEXT:    addv s1, v2.4s
+; CHECK-GI-NEXT:    addv s0, v0.4s
 ; CHECK-GI-NEXT:    add w8, w9, w8
-; CHECK-GI-NEXT:    fmov w9, s5
-; CHECK-GI-NEXT:    add v7.4s, v18.4s, v7.4s
-; CHECK-GI-NEXT:    addv s1, v6.4s
+; CHECK-GI-NEXT:    fmov w9, s4
 ; CHECK-GI-NEXT:    add w8, w9, w8
-; CHECK-GI-NEXT:    fmov w9, s0
-; CHECK-GI-NEXT:    addv s2, v7.4s
+; CHECK-GI-NEXT:    fmov w9, s3
 ; CHECK-GI-NEXT:    add w8, w9, w8
 ; CHECK-GI-NEXT:    fmov w9, s1
 ; CHECK-GI-NEXT:    add w8, w9, w8
-; CHECK-GI-NEXT:    fmov w9, s2
+; CHECK-GI-NEXT:    fmov w9, s0
 ; CHECK-GI-NEXT:    add w0, w9, w8
 ; CHECK-GI-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/AArch64/vector-fcopysign.ll b/llvm/test/CodeGen/AArch64/vector-fcopysign.ll
index de26676b5c73e..063b23275c616 100644
--- a/llvm/test/CodeGen/AArch64/vector-fcopysign.ll
+++ b/llvm/test/CodeGen/AArch64/vector-fcopysign.ll
@@ -264,15 +264,15 @@ define <4 x bfloat> @test_copysign_v4bf16_v4bf16(<4 x bfloat> %a, <4 x bfloat> %
 define <4 x bfloat> @test_copysign_v4bf16_v4f32(<4 x bfloat> %a, <4 x float> %b) #0 {
 ; CHECK-LABEL: test_copysign_v4bf16_v4f32:
 ; CHECK:       ; %bb.0:
-; CHECK-NEXT:    movi.4s v2, #127, msl #8
-; CHECK-NEXT:    movi.4s v3, #1
+; CHECK-NEXT:    movi.4s v2, #1
+; CHECK-NEXT:    movi.4s v3, #127, msl #8
 ; CHECK-NEXT:    ushr.4s v4, v1, #16
-; CHECK-NEXT:    add.4s v2, v1, v2
-; CHECK-NEXT:    and.16b v3, v4, v3
-; CHECK-NEXT:    add.4s v2, v3, v2
-; CHECK-NEXT:    fcmeq.4s v3, v1, v1
+; CHECK-NEXT:    and.16b v2, v4, v2
+; CHECK-NEXT:    add.4s v3, v1, v3
+; CHECK-NEXT:    fcmeq.4s v4, v1, v1
 ; CHECK-NEXT:    orr.4s v1, #64, lsl #16
-; CHECK-NEXT:    bit.16b v1, v2, v3
+; CHECK-NEXT:    add.4s v2, v2, v3
+; CHECK-NEXT:    bit.16b v1, v2, v4
 ; CHECK-NEXT:    mvni.4h v2, #128, lsl #8
 ; CHECK-NEXT:    shrn.4h v1, v1, #16
 ; CHECK-NEXT:    bif.8b v0, v1, v2
@@ -286,16 +286,16 @@ define <4 x bfloat> @test_copysign_v4bf16_v4f64(<4 x bfloat> %a, <4 x double> %b
 ; CHECK-LABEL: test_copysign_v4bf16_v4f64:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    fcvtxn v1.2s, v1.2d
-; CHECK-NEXT:    movi.4s v3, #1
+; CHECK-NEXT:    movi.4s v3, #127, msl #8
 ; CHECK-NEXT:    fcvtxn2 v1.4s, v2.2d
-; CHECK-NEXT:    movi.4s v2, #127, msl #8
+; CHECK-NEXT:    movi.4s v2, #1
 ; CHECK-NEXT:    ushr.4s v4, v1, #16
-; CHECK-NEXT:    add.4s v2, v1, v2
-; CHECK-NEXT:    and.16b v3, v4, v3
-; CHECK-NEXT:    add.4s v2, v3, v2
-; CHECK-NEXT:    fcmeq.4s v3, v1, v1
+; CHECK-NEXT:    add.4s v3, v1, v3
+; CHECK-NEXT:    and.16b v2, v4, v2
+; CHECK-NEXT:    fcmeq.4s v4, v1, v1
 ; CHECK-NEXT:    orr.4s v1, #64, lsl #16
-; CHECK-NEXT:    bit.16b v1, v2, v3
+; CHECK-NEXT:    add.4s v2, v2, v3
+; CHECK-NEXT:    bit.16b v1, v2, v4
 ; CHECK-NEXT:    mvni.4h v2, #128, lsl #8
 ; CHECK-NEXT:    shrn.4h v1, v1, #16
 ; CHECK-NEXT:    bif.8b v0, v1, v2
@@ -322,22 +322,22 @@ define <8 x bfloat> @test_copysign_v8bf16_v8bf16(<8 x bfloat> %a, <8 x bfloat> %
 define <8 x bfloat> @test_copysign_v8bf16_v8f32(<8 x bfloat> %a, <8 x float> %b) #0 {
 ; CHECK-LABEL: test_copysign_v8bf16_v8f32:
 ; CHECK:       ; %bb.0:
-; CHECK-NEXT:    movi.4s v3, #127, msl #8
-; CHECK-NEXT:    movi.4s v4, #1
+; CHECK-NEXT:    movi.4s v3, #1
+; CHECK-NEXT:    movi.4s v4, #127, msl #8
 ; CHECK-NEXT:    ushr.4s v5, v2, #16
 ; CHECK-NEXT:    ushr.4s v6, v1, #16
-; CHECK-NEXT:    add.4s v7, v2, v3
-; CHECK-NEXT:    add.4s v3, v1, v3
-; CHECK-NEXT:    and.16b v5, v5, v4
-; CHECK-NEXT:    and.16b v4, v6, v4
+; CHECK-NEXT:    and.16b v5, v5, v3
+; CHECK-NEXT:    add.4s v7, v2, v4
+; CHECK-NEXT:    and.16b v3, v6, v3
+; CHECK-NEXT:    add.4s v4, v1, v4
 ; CHECK-NEXT:    fcmeq.4s v6, v2, v2
 ; CHECK-NEXT:    orr.4s v2, #64, lsl #16
 ; CHECK-NEXT:    add.4s v5, v5, v7
-; CHECK-NEXT:    add.4s v3, v4, v3
-; CHECK-NEXT:    fcmeq.4s v4, v1, v1
+; CHECK-NEXT:    fcmeq.4s v7, v1, v1
 ; CHECK-NEXT:    orr.4s v1, #64, lsl #16
+; CHECK-NEXT:    add.4s v3, v3, v4
 ; CHECK-NEXT:    bit.16b v2, v5, v6
-; CHECK-NEXT:    bit.16b v1, v3, v4
+; CHECK-NEXT:    bit.16b v1, v3, v7
 ; CHECK-NEXT:    uzp2.8h v1, v1, v2
 ; CHECK-NEXT:    mvni.8h v2, #128, lsl #8
 ; CHECK-NEXT:    bif.16b v0, v1, v2
diff --git a/llvm/test/CodeGen/AArch64/vector-gep.ll b/llvm/test/CodeGen/AArch64/vector-gep.ll
index 30317dce85e65..c7858416e1796 100644
--- a/llvm/test/CodeGen/AArch64/vector-gep.ll
+++ b/llvm/test/CodeGen/AArch64/vector-gep.ll
@@ -13,11 +13,11 @@ define <2 x ptr> @vector_gep(<2 x ptr> %0) {
 ; CHECK:       ; %bb.0: ; %entry
 ; CHECK-NEXT:  Lloh0:
 ; CHECK-NEXT:    adrp x8, lCPI0_0@PAGE
+; CHECK-NEXT:    movi v2.2d, #0x000000ffffffff
 ; CHECK-NEXT:  Lloh1:
 ; CHECK-NEXT:    ldr q1, [x8, lCPI0_0@PAGEOFF]
 ; CHECK-NEXT:    add v0.2d, v0.2d, v1.2d
-; CHECK-NEXT:    movi v1.2d, #0x000000ffffffff
-; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
+; CHECK-NEXT:    and v0.16b, v0.16b, v2.16b
 ; CHECK-NEXT:    ret
 ; CHECK-NEXT:    .loh AdrpLdr Lloh0, Lloh1
 entry:
diff --git a/llvm/test/CodeGen/AArch64/vselect-constants.ll b/llvm/test/CodeGen/AArch64/vselect-constants.ll
index a32147eebd759..5e6ff1e0740ce 100644
--- a/llvm/test/CodeGen/AArch64/vselect-constants.ll
+++ b/llvm/test/CodeGen/AArch64/vselect-constants.ll
@@ -370,10 +370,9 @@ define <vscale x 16 x i8> @signbit_mask_xor_nxv16i8(<vscale x 16 x i8> %a, <vsca
 ; CHECK-LABEL: signbit_mask_xor_nxv16i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p0.b
-; CHECK-NEXT:    eor z1.d, z0.d, z1.d
 ; CHECK-NEXT:    cmplt p0.b, p0/z, z0.b, #0
-; CHECK-NEXT:    mov z1.b, p0/m, #0 // =0x0
-; CHECK-NEXT:    mov z0.d, z1.d
+; CHECK-NEXT:    eor z0.d, z0.d, z1.d
+; CHECK-NEXT:    mov z0.b, p0/m, #0 // =0x0
 ; CHECK-NEXT:    ret
   %cond = icmp slt <vscale x 16 x i8> %a, zeroinitializer
   %xor = xor <vscale x 16 x i8> %a, %b
diff --git a/llvm/test/CodeGen/AArch64/zext-to-tbl.ll b/llvm/test/CodeGen/AArch64/zext-to-tbl.ll
index 08ad34c7b03ba..599bd811d7d59 100644
--- a/llvm/test/CodeGen/AArch64/zext-to-tbl.ll
+++ b/llvm/test/CodeGen/AArch64/zext-to-tbl.ll
@@ -1697,14 +1697,14 @@ define void @zext_v8i8_to_v8i64_with_add_in_sequence_in_loop(ptr %src, ptr %dst)
 ; CHECK-NEXT:    ldp q18, q16, [x10, #96]
 ; CHECK-NEXT:    uaddw.2d v2, v17, v2
 ; CHECK-NEXT:    stp q4, q5, [x10, #32]
-; CHECK-NEXT:    ldp q17, q5, [x10, #64]
-; CHECK-NEXT:    uaddw2.2d v16, v16, v7
+; CHECK-NEXT:    uaddw2.2d v5, v16, v7
+; CHECK-NEXT:    ldp q16, q4, [x10, #64]
 ; CHECK-NEXT:    uaddw.2d v7, v18, v7
 ; CHECK-NEXT:    stp q2, q6, [x10]
-; CHECK-NEXT:    uaddw2.2d v4, v5, v3
-; CHECK-NEXT:    uaddw.2d v3, v17, v3
-; CHECK-NEXT:    stp q7, q16, [x10, #96]
-; CHECK-NEXT:    stp q3, q4, [x10, #64]
+; CHECK-NEXT:    uaddw2.2d v4, v4, v3
+; CHECK-NEXT:    uaddw.2d v2, v16, v3
+; CHECK-NEXT:    stp q7, q5, [x10, #96]
+; CHECK-NEXT:    stp q2, q4, [x10, #64]
 ; CHECK-NEXT:    b.ne LBB17_1
 ; CHECK-NEXT:  ; %bb.2: ; %exit
 ; CHECK-NEXT:    ret
@@ -1729,15 +1729,15 @@ define void @zext_v8i8_to_v8i64_with_add_in_sequence_in_loop(ptr %src, ptr %dst)
 ; CHECK-BE-NEXT:    ld1 { v3.8b }, [x10]
 ; CHECK-BE-NEXT:    add x10, x1, x8
 ; CHECK-BE-NEXT:    add x8, x8, #128
-; CHECK-BE-NEXT:    add x15, x10, #96
 ; CHECK-BE-NEXT:    add x11, x10, #32
 ; CHECK-BE-NEXT:    add x14, x10, #64
+; CHECK-BE-NEXT:    add x15, x10, #96
 ; CHECK-BE-NEXT:    tbl v4.16b, { v2.16b }, v1.16b
 ; CHECK-BE-NEXT:    tbl v2.16b, { v2.16b }, v0.16b
-; CHECK-BE-NEXT:    ld1 { v16.2d }, [x15]
-; CHECK-BE-NEXT:    tbl v5.16b, { v3.16b }, v1.16b
+; CHECK-BE-NEXT:    ld1 { v5.2d }, [x10]
+; CHECK-BE-NEXT:    tbl v6.16b, { v3.16b }, v1.16b
 ; CHECK-BE-NEXT:    tbl v3.16b, { v3.16b }, v0.16b
-; CHECK-BE-NEXT:    ld1 { v6.2d }, [x10]
+; CHECK-BE-NEXT:    ld1 { v16.2d }, [x15]
 ; CHECK-BE-NEXT:    ld1 { v19.2d }, [x14]
 ; CHECK-BE-NEXT:    ld1 { v21.2d }, [x11]
 ; CHECK-BE-NEXT:    add x12, x10, #48
@@ -1747,11 +1747,12 @@ define void @zext_v8i8_to_v8i64_with_add_in_sequence_in_loop(ptr %src, ptr %dst)
 ; CHECK-BE-NEXT:    rev32 v7.8b, v4.8b
 ; CHECK-BE-NEXT:    ext v4.16b, v4.16b, v4.16b, #8
 ; CHECK-BE-NEXT:    rev32 v17.8b, v2.8b
-; CHECK-BE-NEXT:    ext v18.16b, v5.16b, v5.16b, #8
+; CHECK-BE-NEXT:    ext v18.16b, v6.16b, v6.16b, #8
 ; CHECK-BE-NEXT:    ext v20.16b, v3.16b, v3.16b, #8
 ; CHECK-BE-NEXT:    ext v2.16b, v2.16b, v2.16b, #8
-; CHECK-BE-NEXT:    rev32 v5.8b, v5.8b
+; CHECK-BE-NEXT:    rev32 v6.8b, v6.8b
 ; CHECK-BE-NEXT:    rev32 v3.8b, v3.8b
+; CHECK-BE-NEXT:    ld1 { v22.2d }, [x12]
 ; CHECK-BE-NEXT:    cmp x8, #1024
 ; CHECK-BE-NEXT:    rev32 v4.8b, v4.8b
 ; CHECK-BE-NEXT:    uaddw v7.2d, v16.2d, v7.2s
@@ -1760,22 +1761,21 @@ define void @zext_v8i8_to_v8i64_with_add_in_sequence_in_loop(ptr %src, ptr %dst)
 ; CHECK-BE-NEXT:    rev32 v20.8b, v20.8b
 ; CHECK-BE-NEXT:    rev32 v2.8b, v2.8b
 ; CHECK-BE-NEXT:    uaddw v17.2d, v19.2d, v17.2s
-; CHECK-BE-NEXT:    ld1 { v19.2d }, [x12]
-; CHECK-BE-NEXT:    uaddw v5.2d, v21.2d, v5.2s
-; CHECK-BE-NEXT:    ld1 { v21.2d }, [x13]
-; CHECK-BE-NEXT:    uaddw v3.2d, v6.2d, v3.2s
-; CHECK-BE-NEXT:    ld1 { v6.2d }, [x17]
-; CHECK-BE-NEXT:    uaddw v4.2d, v16.2d, v4.2s
+; CHECK-BE-NEXT:    ld1 { v19.2d }, [x13]
+; CHECK-BE-NEXT:    uaddw v6.2d, v21.2d, v6.2s
+; CHECK-BE-NEXT:    uaddw v3.2d, v5.2d, v3.2s
+; CHECK-BE-NEXT:    ld1 { v5.2d }, [x17]
 ; CHECK-BE-NEXT:    st1 { v7.2d }, [x15]
-; CHECK-BE-NEXT:    uaddw v7.2d, v19.2d, v18.2s
-; CHECK-BE-NEXT:    uaddw v16.2d, v21.2d, v20.2s
-; CHECK-BE-NEXT:    uaddw v2.2d, v6.2d, v2.2s
-; CHECK-BE-NEXT:    st1 { v17.2d }, [x14]
-; CHECK-BE-NEXT:    st1 { v5.2d }, [x11]
+; CHECK-BE-NEXT:    uaddw v4.2d, v16.2d, v4.2s
+; CHECK-BE-NEXT:    st1 { v6.2d }, [x11]
+; CHECK-BE-NEXT:    uaddw v6.2d, v22.2d, v18.2s
 ; CHECK-BE-NEXT:    st1 { v3.2d }, [x10]
+; CHECK-BE-NEXT:    uaddw v3.2d, v19.2d, v20.2s
+; CHECK-BE-NEXT:    uaddw v2.2d, v5.2d, v2.2s
+; CHECK-BE-NEXT:    st1 { v17.2d }, [x14]
 ; CHECK-BE-NEXT:    st1 { v4.2d }, [x16]
-; CHECK-BE-NEXT:    st1 { v7.2d }, [x12]
-; CHECK-BE-NEXT:    st1 { v16.2d }, [x13]
+; CHECK-BE-NEXT:    st1 { v6.2d }, [x12]
+; CHECK-BE-NEXT:    st1 { v3.2d }, [x13]
 ; CHECK-BE-NEXT:    st1 { v2.2d }, [x17]
 ; CHECK-BE-NEXT:    b.ne .LBB17_1
 ; CHECK-BE-NEXT:  // %bb.2: // %exit
diff --git a/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-neon-instructions.s b/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-neon-instructions.s
index 73eadc268bf26..f703392f3e9d0 100644
--- a/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-neon-instructions.s
+++ b/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-neon-instructions.s
@@ -1070,14 +1070,14 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT: [6]: HasSideEffects (U)
 
 # CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    Instructions:
-# CHECK-NEXT:  1      4     0.50                        abs	d29, d24
-# CHECK-NEXT:  1      4     0.50                        abs	v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        abs	v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        abs	v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        abs	v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        abs	v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        abs	v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        abs	v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        abs	d29, d24
+# CHECK-NEXT:  1      3     0.50                        abs	v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        abs	v0.2d, v0.2d
+# CHECK-NEXT:  1      3     0.50                        abs	v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        abs	v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        abs	v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        abs	v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        abs	v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        add	d17, d31, d29
 # CHECK-NEXT:  1      3     0.50                        add	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        addhn	v0.2s, v0.2d, v0.2d
@@ -1086,8 +1086,8 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        addhn2	v0.16b, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        addhn2	v0.4s, v0.2d, v0.2d
 # CHECK-NEXT:  1      4     0.50                        addhn2	v0.8h, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        addp	v0.2d, v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        addp	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        addp	v0.2d, v0.2d, v0.2d
+# CHECK-NEXT:  1      3     0.50                        addp	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      3     0.50                        and	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      3     0.50                        bic	v0.4h, #15, lsl #8
 # CHECK-NEXT:  1      3     0.50                        bic	v0.8b, v0.8b, v0.8b
@@ -1441,13 +1441,13 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        mvni	v0.2s, #0
 # CHECK-NEXT:  1      3     0.50                        mvni	v0.4s, #16, msl #16
 # CHECK-NEXT:  1      3     0.50                        neg	d29, d24
-# CHECK-NEXT:  1      4     0.50                        neg	v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        neg	v0.2d, v0.2d
+# CHECK-NEXT:  1      3     0.50                        neg	v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        neg	v0.2d, v0.2d
 # CHECK-NEXT:  1      3     0.50                        neg	v0.2s, v0.2s
 # CHECK-NEXT:  1      3     0.50                        neg	v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        neg	v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        neg	v0.4s, v0.4s
 # CHECK-NEXT:  1      3     0.50                        neg	v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        neg	v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        neg	v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        mvn	v0.16b, v0.16b
 # CHECK-NEXT:  1      3     0.50                        mvn	v0.8b, v0.8b
 # CHECK-NEXT:  1      3     0.50                        orn	v0.16b, v0.16b, v0.16b
@@ -1457,12 +1457,12 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        pmul	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        pmull	v0.8h, v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        pmull2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        raddhn	v0.2s, v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        raddhn	v0.4h, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        raddhn	v0.8b, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        raddhn2	v0.16b, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        raddhn2	v0.4s, v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        raddhn2	v0.8h, v0.4s, v0.4s
+# CHECK-NEXT:  1      8     0.50                        raddhn	v0.2s, v0.2d, v0.2d
+# CHECK-NEXT:  1      8     0.50                        raddhn	v0.4h, v0.4s, v0.4s
+# CHECK-NEXT:  1      8     0.50                        raddhn	v0.8b, v0.8h, v0.8h
+# CHECK-NEXT:  1      8     0.50                        raddhn2	v0.16b, v0.8h, v0.8h
+# CHECK-NEXT:  1      8     0.50                        raddhn2	v0.4s, v0.2d, v0.2d
+# CHECK-NEXT:  1      8     0.50                        raddhn2	v0.8h, v0.4s, v0.4s
 # CHECK-NEXT:  1      4     0.50                        rbit	v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        rbit	v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        rev16	v21.8b, v1.8b
@@ -1483,19 +1483,19 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        rshrn2	v0.16b, v0.8h, #3
 # CHECK-NEXT:  1      4     0.50                        rshrn2	v0.4s, v0.2d, #3
 # CHECK-NEXT:  1      4     0.50                        rshrn2	v0.8h, v0.4s, #3
-# CHECK-NEXT:  1      4     0.50                        rsubhn	v0.2s, v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        rsubhn	v0.4h, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        rsubhn	v0.8b, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        rsubhn2	v0.16b, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        rsubhn2	v0.4s, v0.2d, v0.2d
-# CHECK-NEXT:  1      4     0.50                        rsubhn2	v0.8h, v0.4s, v0.4s
-# CHECK-NEXT:  1      8     0.50                        saba	v0.16b, v0.16b, v0.16b
-# CHECK-NEXT:  1      8     0.50                        sabal	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      8     0.50                        sabal	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      8     0.50                        sabal	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      8     0.50                        sabal2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      8     0.50                        sabal2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      8     0.50                        sabal2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      8     0.50                        rsubhn	v0.2s, v0.2d, v0.2d
+# CHECK-NEXT:  1      8     0.50                        rsubhn	v0.4h, v0.4s, v0.4s
+# CHECK-NEXT:  1      8     0.50                        rsubhn	v0.8b, v0.8h, v0.8h
+# CHECK-NEXT:  1      8     0.50                        rsubhn2	v0.16b, v0.8h, v0.8h
+# CHECK-NEXT:  1      8     0.50                        rsubhn2	v0.4s, v0.2d, v0.2d
+# CHECK-NEXT:  1      8     0.50                        rsubhn2	v0.8h, v0.4s, v0.4s
+# CHECK-NEXT:  1      6     0.50                        saba	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      6     0.50                        sabal	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      6     0.50                        sabal	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      6     0.50                        sabal	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      6     0.50                        sabal2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      6     0.50                        sabal2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      6     0.50                        sabal2	v0.8h, v0.16b, v0.16b
 # CHECK-NEXT:  1      3     0.50                        sabd	v0.4h, v0.4h, v0.4h
 # CHECK-NEXT:  1      3     0.50                        sabdl	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  1      3     0.50                        sabdl	v0.4s, v0.4h, v0.4h
@@ -1503,30 +1503,30 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        sabdl2	v0.2d, v0.4s, v0.4s
 # CHECK-NEXT:  1      3     0.50                        sabdl2	v0.4s, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        sabdl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.1d, v0.2s
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.2d, v0.4s
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.2s, v0.4h
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.4h, v0.8b
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.4s, v0.8h
-# CHECK-NEXT:  1      8     1.00                        sadalp	v0.8h, v0.16b
-# CHECK-NEXT:  1      4     0.50                        saddl	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        saddl	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        saddl	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        saddl2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        saddl2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        saddl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.1d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.2s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.4h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        saddlp	v0.8h, v0.16b
-# CHECK-NEXT:  1      4     0.50                        saddw	v0.2d, v0.2d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        saddw	v0.4s, v0.4s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        saddw	v0.8h, v0.8h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        saddw2	v0.2d, v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        saddw2	v0.4s, v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        saddw2	v0.8h, v0.8h, v0.16b
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.1d, v0.2s
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.2d, v0.4s
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.2s, v0.4h
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.4h, v0.8b
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.4s, v0.8h
+# CHECK-NEXT:  1      7     1.00                        sadalp	v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        saddl	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        saddl	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        saddl	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        saddl2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        saddl2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        saddl2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.1d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.2s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.4h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        saddlp	v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        saddw	v0.2d, v0.2d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        saddw	v0.4s, v0.4s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        saddw	v0.8h, v0.8h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        saddw2	v0.2d, v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        saddw2	v0.4s, v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        saddw2	v0.8h, v0.8h, v0.16b
 # CHECK-NEXT:  1      4     0.50                        scvtf	d21, d12
 # CHECK-NEXT:  1      4     0.50                        scvtf	d21, d12, #64
 # CHECK-NEXT:  1      4     0.50                        scvtf	s22, s13
@@ -1573,18 +1573,18 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        sli	v0.4s, v0.4s, #3
 # CHECK-NEXT:  1      4     0.50                        sli	v0.8b, v0.8b, #3
 # CHECK-NEXT:  1      4     0.50                        sli	v0.8h, v0.8h, #3
-# CHECK-NEXT:  1      4     0.50                        smax	v0.2s, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        smax	v0.4h, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        smax	v0.8b, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        smaxp	v0.2s, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        smaxp	v0.4h, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        smaxp	v0.8b, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        smin	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        smax	v0.2s, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        smax	v0.4h, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        smax	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        smaxp	v0.2s, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        smaxp	v0.4h, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        smaxp	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        smin	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        smin	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        smin	v0.8h, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        sminp	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        smin	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        sminp	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        sminp	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        sminp	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        sminp	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        smlal	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  1      4     0.50                        smlal	v0.4s, v0.4h, v0.4h
 # CHECK-NEXT:  1      4     0.50                        smlal	v0.8h, v0.8b, v0.8b
@@ -1777,14 +1777,14 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        srshr	v0.4s, v0.4s, #3
 # CHECK-NEXT:  1      3     0.50                        srshr	v0.8b, v0.8b, #3
 # CHECK-NEXT:  1      3     0.50                        srshr	v0.8h, v0.8h, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	d15, d11, #19
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  1      8     1.00                        srsra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	d15, d11, #19
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  1      7     1.00                        srsra	v0.8h, v0.8h, #3
 # CHECK-NEXT:  1      3     0.50                        sshl	d31, d31, d31
 # CHECK-NEXT:  1      3     0.50                        sshl	v0.2d, v0.2d, v0.2d
 # CHECK-NEXT:  1      3     0.50                        sshl	v0.2s, v0.2s, v0.2s
@@ -1800,26 +1800,26 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        sshr	v0.4s, v0.4s, #3
 # CHECK-NEXT:  1      3     0.50                        sshr	v0.8b, v0.8b, #3
 # CHECK-NEXT:  1      3     0.50                        sshr	v0.8h, v0.8h, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	d18, d12, #21
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  1      8     1.00                        ssra	v0.8h, v0.8h, #3
-# CHECK-NEXT:  1      4     0.50                        ssubl	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        ssubl	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        ssubl	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        ssubl2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        ssubl2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        ssubl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        ssubw	v0.2d, v0.2d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        ssubw	v0.4s, v0.4s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        ssubw	v0.8h, v0.8h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        ssubw2	v0.2d, v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        ssubw2	v0.4s, v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        ssubw2	v0.8h, v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        ssra	d18, d12, #21
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  1      3     0.50                        ssra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  1      3     0.50                        ssubl	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        ssubl	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        ssubl	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        ssubl2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        ssubl2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        ssubl2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        ssubw	v0.2d, v0.2d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        ssubw	v0.4s, v0.4s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        ssubw	v0.8h, v0.8h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        ssubw2	v0.2d, v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        ssubw2	v0.4s, v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        ssubw2	v0.8h, v0.8h, v0.16b
 # CHECK-NEXT:  1      4     1.00           *            st1	{ v0.16b }, [x0]
 # CHECK-NEXT:  2      5     2.00           *            st1	{ v0.2d, v1.2d, v2.2d }, [x0], #48
 # CHECK-NEXT:  1      5     4.00           *            st1	{ v0.2d, v1.2d, v2.2d, v3.2d }, [x0]
@@ -1843,7 +1843,7 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      5     2.00           *            st4	{ v0.b, v1.b, v2.b, v3.b }[9], [x0]
 # CHECK-NEXT:  2      5     2.00           *            st4	{ v0.b, v1.b, v2.b, v3.b }[9], [x0], x5
 # CHECK-NEXT:  1      3     0.50                        sub	d15, d5, d16
-# CHECK-NEXT:  1      4     0.50                        sub	v0.2d, v0.2d, v0.2d
+# CHECK-NEXT:  1      3     0.50                        sub	v0.2d, v0.2d, v0.2d
 # CHECK-NEXT:  1      4     0.50                        suqadd	b19, b14
 # CHECK-NEXT:  1      4     0.50                        suqadd	d18, d22
 # CHECK-NEXT:  1      4     0.50                        suqadd	h20, h15
@@ -1885,13 +1885,13 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        trn2	v0.4s, v0.4s, v0.4s
 # CHECK-NEXT:  1      4     0.50                        trn2	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        trn2	v0.8h, v0.8h, v0.8h
-# CHECK-NEXT:  1      8     0.50                        uaba	v0.8b, v0.8b, v0.8b
-# CHECK-NEXT:  1      8     0.50                        uabal	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      8     0.50                        uabal	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      8     0.50                        uabal	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      8     0.50                        uabal2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      8     0.50                        uabal2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      8     0.50                        uabal2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      6     0.50                        uaba	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      6     0.50                        uabal	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      6     0.50                        uabal	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      6     0.50                        uabal	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      6     0.50                        uabal2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      6     0.50                        uabal2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      6     0.50                        uabal2	v0.8h, v0.16b, v0.16b
 # CHECK-NEXT:  1      3     0.50                        uabd	v0.4h, v0.4h, v0.4h
 # CHECK-NEXT:  1      3     0.50                        uabdl	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  1      3     0.50                        uabdl	v0.4s, v0.4h, v0.4h
@@ -1899,30 +1899,30 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        uabdl2	v0.2d, v0.4s, v0.4s
 # CHECK-NEXT:  1      3     0.50                        uabdl2	v0.4s, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        uabdl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.1d, v0.2s
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.2d, v0.4s
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.2s, v0.4h
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.4h, v0.8b
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.4s, v0.8h
-# CHECK-NEXT:  1      8     1.00                        uadalp	v0.8h, v0.16b
-# CHECK-NEXT:  1      4     0.50                        uaddl	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        uaddl	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        uaddl	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        uaddl2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        uaddl2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        uaddl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.1d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.2s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.4h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        uaddlp	v0.8h, v0.16b
-# CHECK-NEXT:  1      4     0.50                        uaddw	v0.2d, v0.2d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        uaddw	v0.4s, v0.4s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        uaddw	v0.8h, v0.8h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        uaddw2	v0.2d, v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        uaddw2	v0.4s, v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        uaddw2	v0.8h, v0.8h, v0.16b
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.1d, v0.2s
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.2d, v0.4s
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.2s, v0.4h
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.4h, v0.8b
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.4s, v0.8h
+# CHECK-NEXT:  1      7     1.00                        uadalp	v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        uaddl	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        uaddl	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        uaddl	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        uaddl2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        uaddl2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        uaddl2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.1d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.2s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.4h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        uaddlp	v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        uaddw	v0.2d, v0.2d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        uaddw	v0.4s, v0.4s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        uaddw	v0.8h, v0.8h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        uaddw2	v0.2d, v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        uaddw2	v0.4s, v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        uaddw2	v0.8h, v0.8h, v0.16b
 # CHECK-NEXT:  1      4     0.50                        ucvtf	d21, d14
 # CHECK-NEXT:  1      4     0.50                        ucvtf	d21, d14, #64
 # CHECK-NEXT:  1      4     0.50                        ucvtf	s22, s13
@@ -1935,21 +1935,21 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        ucvtf	v0.4s, v0.4s
 # CHECK-NEXT:  1      4     0.50                        ucvtf	v0.4s, v0.4s, #3
 # CHECK-NEXT:  1      4     0.50                        ucvtf	v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        uhadd	v0.16b, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        uhadd	v0.8h, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        uhsub	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        umax	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        uhadd	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        uhadd	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        uhsub	v0.4s, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        umax	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        umax	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        umax	v0.8h, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        umaxp	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        umax	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        umaxp	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        umaxp	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        umaxp	v0.8h, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        umin	v0.2s, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        umin	v0.4h, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        umin	v0.8b, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        uminp	v0.2s, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        uminp	v0.4h, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        uminp	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        umaxp	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        umin	v0.2s, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        umin	v0.4h, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        umin	v0.8b, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        uminp	v0.2s, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        uminp	v0.4h, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        uminp	v0.8b, v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        umlal	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  1      4     0.50                        umlal	v0.4s, v0.4h, v0.4h
 # CHECK-NEXT:  1      4     0.50                        umlal	v0.8h, v0.8b, v0.8b
@@ -2024,9 +2024,9 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        uqxtn2	v0.8h, v0.4s
 # CHECK-NEXT:  1      4     0.50                        urecpe	v0.2s, v0.2s
 # CHECK-NEXT:  1      4     0.50                        urecpe	v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        urhadd	v0.16b, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        urhadd	v0.4s, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        urhadd	v0.8h, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        urhadd	v0.16b, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        urhadd	v0.4s, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        urhadd	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        urshl	d8, d7, d4
 # CHECK-NEXT:  1      3     0.50                        urshl	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      3     0.50                        urshl	v0.2d, v0.2d, v0.2d
@@ -2042,14 +2042,14 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      3     0.50                        urshr	v0.8h, v0.8h, #3
 # CHECK-NEXT:  1      12    9.00                        ursqrte	v0.2s, v0.2s
 # CHECK-NEXT:  1      12    9.00                        ursqrte	v0.4s, v0.4s
-# CHECK-NEXT:  1      8     1.00                        ursra	d18, d10, #13
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  1      8     1.00                        ursra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	d18, d10, #13
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  1      7     1.00                        ursra	v0.8h, v0.8h, #3
 # CHECK-NEXT:  1      3     0.50                        ushl	d0, d0, d0
 # CHECK-NEXT:  1      3     0.50                        ushl	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      3     0.50                        ushl	v0.4s, v0.4s, v0.4s
@@ -2075,26 +2075,26 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  1      4     0.50                        usqadd	v0.4s, v0.4s
 # CHECK-NEXT:  1      4     0.50                        usqadd	v0.8b, v0.8b
 # CHECK-NEXT:  1      4     0.50                        usqadd	v0.8h, v0.8h
-# CHECK-NEXT:  1      8     1.00                        usra	d20, d13, #61
-# CHECK-NEXT:  1      8     1.00                        usra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  1      8     1.00                        usra	v0.8h, v0.8h, #3
-# CHECK-NEXT:  1      4     0.50                        usubl	v0.2d, v0.2s, v0.2s
-# CHECK-NEXT:  1      4     0.50                        usubl	v0.4s, v0.4h, v0.4h
-# CHECK-NEXT:  1      4     0.50                        usubl	v0.8h, v0.8b, v0.8b
-# CHECK-NEXT:  1      4     0.50                        usubl2	v0.2d, v0.4s, v0.4s
-# CHECK-NEXT:  1      4     0.50                        usubl2	v0.4s, v0.8h, v0.8h
-# CHECK-NEXT:  1      4     0.50                        usubl2	v0.8h, v0.16b, v0.16b
-# CHECK-NEXT:  1      4     0.50                        usubw	v0.2d, v0.2d, v0.2s
-# CHECK-NEXT:  1      4     0.50                        usubw	v0.4s, v0.4s, v0.4h
-# CHECK-NEXT:  1      4     0.50                        usubw	v0.8h, v0.8h, v0.8b
-# CHECK-NEXT:  1      4     0.50                        usubw2	v0.2d, v0.2d, v0.4s
-# CHECK-NEXT:  1      4     0.50                        usubw2	v0.4s, v0.4s, v0.8h
-# CHECK-NEXT:  1      4     0.50                        usubw2	v0.8h, v0.8h, v0.16b
+# CHECK-NEXT:  1      3     0.50                        usra	d20, d13, #61
+# CHECK-NEXT:  1      3     0.50                        usra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  1      3     0.50                        usra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  1      3     0.50                        usubl	v0.2d, v0.2s, v0.2s
+# CHECK-NEXT:  1      3     0.50                        usubl	v0.4s, v0.4h, v0.4h
+# CHECK-NEXT:  1      3     0.50                        usubl	v0.8h, v0.8b, v0.8b
+# CHECK-NEXT:  1      3     0.50                        usubl2	v0.2d, v0.4s, v0.4s
+# CHECK-NEXT:  1      3     0.50                        usubl2	v0.4s, v0.8h, v0.8h
+# CHECK-NEXT:  1      3     0.50                        usubl2	v0.8h, v0.16b, v0.16b
+# CHECK-NEXT:  1      3     0.50                        usubw	v0.2d, v0.2d, v0.2s
+# CHECK-NEXT:  1      3     0.50                        usubw	v0.4s, v0.4s, v0.4h
+# CHECK-NEXT:  1      3     0.50                        usubw	v0.8h, v0.8h, v0.8b
+# CHECK-NEXT:  1      3     0.50                        usubw2	v0.2d, v0.2d, v0.4s
+# CHECK-NEXT:  1      3     0.50                        usubw2	v0.4s, v0.4s, v0.8h
+# CHECK-NEXT:  1      3     0.50                        usubw2	v0.8h, v0.8h, v0.16b
 # CHECK-NEXT:  1      4     0.50                        uzp1	v0.16b, v0.16b, v0.16b
 # CHECK-NEXT:  1      4     0.50                        uzp1	v0.2d, v0.2d, v0.2d
 # CHECK-NEXT:  1      4     0.50                        uzp1	v0.2s, v0.2s, v0.2s
@@ -2148,7 +2148,7 @@ zip2	v0.8h, v0.8h, v0.8h
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10.0] [10.1] [11]
-# CHECK-NEXT:  -      -      -      -      -     39.00  91.00   -      -     509.00 509.00 3.00   3.00   197.00
+# CHECK-NEXT:  -      -      -      -      -     39.00  91.00   -      -     501.00 501.00 3.00   3.00   197.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10.0] [10.1] [11]   Instructions:
@@ -2882,14 +2882,14 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     sshr	v0.4s, v0.4s, #3
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     sshr	v0.8b, v0.8b, #3
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     sshr	v0.8h, v0.8h, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	d18, d12, #21
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     ssra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	d18, d12, #21
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssra	v0.8h, v0.8h, #3
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssubl	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssubl	v0.4s, v0.4h, v0.4h
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     ssubl	v0.8h, v0.8b, v0.8b
@@ -3157,14 +3157,14 @@ zip2	v0.8h, v0.8h, v0.8h
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usqadd	v0.4s, v0.4s
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usqadd	v0.8b, v0.8b
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usqadd	v0.8h, v0.8h
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	d20, d13, #61
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.16b, v0.16b, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.2d, v0.2d, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.2s, v0.2s, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.4h, v0.4h, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.4s, v0.4s, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.8b, v0.8b, #3
-# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     1.00   1.00    -      -      -     usra	v0.8h, v0.8h, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	d20, d13, #61
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.16b, v0.16b, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.2d, v0.2d, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.2s, v0.2s, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.4h, v0.4h, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.4s, v0.4s, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.8b, v0.8b, #3
+# CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usra	v0.8h, v0.8h, #3
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usubl	v0.2d, v0.2s, v0.2s
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usubl	v0.4s, v0.4h, v0.4h
 # CHECK-NEXT:  -      -      -      -      -      -      -      -      -     0.50   0.50    -      -      -     usubl	v0.8h, v0.8b, v0.8b
diff --git a/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-sve-instructions.s b/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-sve-instructions.s
index a8fb8b669838f..d8051e7ecb4fe 100644
--- a/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-sve-instructions.s
+++ b/llvm/test/tools/llvm-mca/AArch64/Cortex/A510-sve-instructions.s
@@ -3476,12 +3476,12 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        add	z31.s, p7/m, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        add	z31.s, z31.s, #65280
 # CHECK-NEXT:  1      3     0.50                        add	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        addhnb	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        addhnb	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        addhnb	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        addhnt	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        addhnt	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        addhnt	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        addhnb	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        addhnb	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        addhnb	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        addhnt	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        addhnt	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        addhnt	z0.s, z1.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        addp	z0.b, p0/m, z0.b, z1.b
 # CHECK-NEXT:  1      3     0.50                        addp	z0.h, p0/m, z0.h, z1.h
 # CHECK-NEXT:  1      3     0.50                        addp	z29.s, p7/m, z29.s, z30.s
@@ -3516,7 +3516,7 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        aesimc	z31.b, z31.b
 # CHECK-NEXT:  1      3     0.50                        aesmc	z0.b, z0.b
 # CHECK-NEXT:  1      3     0.50                        aesmc	z31.b, z31.b
-# CHECK-NEXT:  1      6     1.00                        and	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        and	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      3     0.50                        and	z0.d, z0.d, #0x6
 # CHECK-NEXT:  1      3     0.50                        and	z0.d, z0.d, #0xfffffffffffffff9
 # CHECK-NEXT:  1      3     0.50                        and	z0.d, z0.d, z0.d
@@ -3531,7 +3531,7 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        and	z31.s, p7/m, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        and	z5.b, z5.b, #0x6
 # CHECK-NEXT:  1      3     0.50                        and	z5.b, z5.b, #0xf9
-# CHECK-NEXT:  1      6     1.00                        ands	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        ands	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      4     1.00                        andv	b0, p7, z31.b
 # CHECK-NEXT:  1      4     1.00                        andv	d0, p7, z31.d
 # CHECK-NEXT:  1      4     1.00                        andv	h0, p7, z31.h
@@ -3574,7 +3574,7 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        asrr	z0.d, p0/m, z0.d, z0.d
 # CHECK-NEXT:  1      3     0.50                        asrr	z0.h, p0/m, z0.h, z0.h
 # CHECK-NEXT:  1      3     0.50                        asrr	z0.s, p0/m, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        bcax	z29.d, z29.d, z30.d, z31.d
+# CHECK-NEXT:  1      4     0.50                        bcax	z29.d, z29.d, z30.d, z31.d
 # CHECK-NEXT:  1      14    13.00                       bdep	z0.b, z1.b, z31.b
 # CHECK-NEXT:  1      70    69.00                       bdep	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      22    21.00                       bdep	z0.h, z1.h, z31.h
@@ -3603,34 +3603,34 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      70    69.00                       bgrp	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      22    21.00                       bgrp	z0.h, z1.h, z31.h
 # CHECK-NEXT:  1      38    37.00                       bgrp	z0.s, z1.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        bic	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        bic	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        bic	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        bic	p15.b, p15/z, p15.b, p15.b
 # CHECK-NEXT:  1      3     0.50                        bic	z0.d, z0.d, z0.d
 # CHECK-NEXT:  1      3     0.50                        bic	z23.d, z13.d, z8.d
 # CHECK-NEXT:  1      3     0.50                        bic	z31.b, p7/m, z31.b, z31.b
 # CHECK-NEXT:  1      3     0.50                        bic	z31.d, p7/m, z31.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        bic	z31.h, p7/m, z31.h, z31.h
 # CHECK-NEXT:  1      3     0.50                        bic	z31.s, p7/m, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        bics	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        bics	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brka	p0.b, p15/m, p15.b
-# CHECK-NEXT:  1      6     1.00                        brka	p0.b, p15/z, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkas	p0.b, p15/z, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkb	p0.b, p15/m, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkb	p0.b, p15/z, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkbs	p0.b, p15/z, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkn	p0.b, p15/z, p1.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        brkn	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkns	p0.b, p15/z, p1.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        brkns	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkpa	p0.b, p15/z, p1.b, p2.b
-# CHECK-NEXT:  1      6     1.00                        brkpa	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkpas	p0.b, p15/z, p1.b, p2.b
-# CHECK-NEXT:  1      6     1.00                        brkpas	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkpb	p0.b, p15/z, p1.b, p2.b
-# CHECK-NEXT:  1      6     1.00                        brkpb	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        brkpbs	p0.b, p15/z, p1.b, p2.b
-# CHECK-NEXT:  1      6     1.00                        brkpbs	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        bics	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        bics	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        brka	p0.b, p15/m, p15.b
+# CHECK-NEXT:  1      2     1.00                        brka	p0.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkas	p0.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkb	p0.b, p15/m, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkb	p0.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkbs	p0.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkn	p0.b, p15/z, p1.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        brkn	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkns	p0.b, p15/z, p1.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        brkns	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkpa	p0.b, p15/z, p1.b, p2.b
+# CHECK-NEXT:  1      2     1.00                        brkpa	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      4     1.00                        brkpas	p0.b, p15/z, p1.b, p2.b
+# CHECK-NEXT:  1      4     1.00                        brkpas	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        brkpb	p0.b, p15/z, p1.b, p2.b
+# CHECK-NEXT:  1      2     1.00                        brkpb	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      4     1.00                        brkpbs	p0.b, p15/z, p1.b, p2.b
+# CHECK-NEXT:  1      4     1.00                        brkpbs	p15.b, p15/z, p15.b, p15.b
 # CHECK-NEXT:  1      3     0.50                        bsl	z0.d, z0.d, z1.d, z2.d
 # CHECK-NEXT:  1      3     0.50                        bsl1n	z0.d, z0.d, z1.d, z2.d
 # CHECK-NEXT:  1      3     0.50                        bsl2n	z0.d, z0.d, z1.d, z2.d
@@ -3704,163 +3704,163 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        cmla	z31.h, z31.h, z31.h, #180
 # CHECK-NEXT:  1      4     0.50                        cmla	z31.s, z30.s, z7.s[0], #180
 # CHECK-NEXT:  1      4     0.50                        cmla	z31.s, z31.s, z31.s, #180
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpeq	p0.s, p0/z, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.b, p0/z, z1.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.d, p0/z, z1.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.h, p0/z, z1.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.s, p0/z, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmpge	p0.s, p0/z, z1.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.b, p0/z, z1.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.d, p0/z, z1.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.h, p0/z, z1.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.s, p0/z, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmpgt	p0.s, p0/z, z1.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.b, p0/z, z0.b, #0
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.b, p0/z, z0.b, #127
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.b, p0/z, z1.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.d, p0/z, z0.d, #0
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.d, p0/z, z0.d, #127
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.d, p0/z, z1.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.h, p0/z, z0.h, #0
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.h, p0/z, z0.h, #127
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.h, p0/z, z1.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.s, p0/z, z0.s, #0
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.s, p0/z, z0.s, #127
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.s, p0/z, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmphi	p0.s, p0/z, z1.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.b, p0/z, z0.b, #0
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.b, p0/z, z0.b, #127
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.b, p0/z, z1.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.d, p0/z, z0.d, #0
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.d, p0/z, z0.d, #127
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.d, p0/z, z1.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.h, p0/z, z0.h, #0
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.h, p0/z, z0.h, #127
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.h, p0/z, z1.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.s, p0/z, z0.s, #0
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.s, p0/z, z0.s, #127
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.s, p0/z, z0.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmphs	p0.s, p0/z, z1.s, z0.s
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmple	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.b, p0/z, z0.b, #0
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.b, p0/z, z0.b, #127
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.d, p0/z, z0.d, #0
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.d, p0/z, z0.d, #127
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.h, p0/z, z0.h, #0
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.h, p0/z, z0.h, #127
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.s, p0/z, z0.s, #0
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.s, p0/z, z0.s, #127
-# CHECK-NEXT:  1      3     0.50                        cmplo	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.b, p0/z, z0.b, #0
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.b, p0/z, z0.b, #127
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.d, p0/z, z0.d, #0
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.d, p0/z, z0.d, #127
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.h, p0/z, z0.h, #0
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.h, p0/z, z0.h, #127
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.s, p0/z, z0.s, #0
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.s, p0/z, z0.s, #127
-# CHECK-NEXT:  1      3     0.50                        cmpls	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmplt	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.b, p0/z, z0.b, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.b, p0/z, z0.b, #15
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.b, p0/z, z0.b, z0.b
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.b, p0/z, z0.b, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.d, p0/z, z0.d, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.d, p0/z, z0.d, #15
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.d, p0/z, z0.d, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.h, p0/z, z0.h, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.h, p0/z, z0.h, #15
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.h, p0/z, z0.h, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.h, p0/z, z0.h, z0.h
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.s, p0/z, z0.s, #-16
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.s, p0/z, z0.s, #15
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.s, p0/z, z0.s, z0.d
-# CHECK-NEXT:  1      3     0.50                        cmpne	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpeq	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.b, p0/z, z1.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.d, p0/z, z1.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.h, p0/z, z1.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmpge	p0.s, p0/z, z1.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.b, p0/z, z1.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.d, p0/z, z1.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.h, p0/z, z1.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmpgt	p0.s, p0/z, z1.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.b, p0/z, z0.b, #0
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.b, p0/z, z0.b, #127
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.b, p0/z, z1.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.d, p0/z, z0.d, #0
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.d, p0/z, z0.d, #127
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.d, p0/z, z1.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.h, p0/z, z0.h, #0
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.h, p0/z, z0.h, #127
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.h, p0/z, z1.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.s, p0/z, z0.s, #0
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.s, p0/z, z0.s, #127
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmphi	p0.s, p0/z, z1.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.b, p0/z, z0.b, #0
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.b, p0/z, z0.b, #127
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.b, p0/z, z1.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.d, p0/z, z0.d, #0
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.d, p0/z, z0.d, #127
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.d, p0/z, z1.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.h, p0/z, z0.h, #0
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.h, p0/z, z0.h, #127
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.h, p0/z, z1.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.s, p0/z, z0.s, #0
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.s, p0/z, z0.s, #127
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.s, p0/z, z0.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmphs	p0.s, p0/z, z1.s, z0.s
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmple	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.b, p0/z, z0.b, #0
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.b, p0/z, z0.b, #127
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.d, p0/z, z0.d, #0
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.d, p0/z, z0.d, #127
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.h, p0/z, z0.h, #0
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.h, p0/z, z0.h, #127
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.s, p0/z, z0.s, #0
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.s, p0/z, z0.s, #127
+# CHECK-NEXT:  1      4     0.50                        cmplo	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.b, p0/z, z0.b, #0
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.b, p0/z, z0.b, #127
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.d, p0/z, z0.d, #0
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.d, p0/z, z0.d, #127
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.h, p0/z, z0.h, #0
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.h, p0/z, z0.h, #127
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.s, p0/z, z0.s, #0
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.s, p0/z, z0.s, #127
+# CHECK-NEXT:  1      4     0.50                        cmpls	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmplt	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.b, p0/z, z0.b, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.b, p0/z, z0.b, #15
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.b, p0/z, z0.b, z0.b
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.b, p0/z, z0.b, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.d, p0/z, z0.d, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.d, p0/z, z0.d, #15
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.d, p0/z, z0.d, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.h, p0/z, z0.h, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.h, p0/z, z0.h, #15
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.h, p0/z, z0.h, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.h, p0/z, z0.h, z0.h
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.s, p0/z, z0.s, #-16
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.s, p0/z, z0.s, #15
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.s, p0/z, z0.s, z0.d
+# CHECK-NEXT:  1      4     0.50                        cmpne	p0.s, p0/z, z0.s, z0.s
 # CHECK-NEXT:  1      3     0.50                        cnot	z31.b, p7/m, z31.b
 # CHECK-NEXT:  1      3     0.50                        cnot	z31.d, p7/m, z31.d
 # CHECK-NEXT:  1      3     0.50                        cnot	z31.h, p7/m, z31.h
 # CHECK-NEXT:  1      3     0.50                        cnot	z31.s, p7/m, z31.s
-# CHECK-NEXT:  1      4     0.50                        cnt	z31.b, p7/m, z31.b
+# CHECK-NEXT:  1      3     0.50                        cnt	z31.b, p7/m, z31.b
 # CHECK-NEXT:  1      12    0.50                        cnt	z31.d, p7/m, z31.d
-# CHECK-NEXT:  1      4     0.50                        cnt	z31.h, p7/m, z31.h
+# CHECK-NEXT:  1      3     0.50                        cnt	z31.h, p7/m, z31.h
 # CHECK-NEXT:  1      8     0.50                        cnt	z31.s, p7/m, z31.s
 # CHECK-NEXT:  1      1     0.33                        cntb	x0
 # CHECK-NEXT:  1      1     0.33                        cntb	x0, #28
@@ -3874,10 +3874,10 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      1     0.33                        cnth	x0, #28
 # CHECK-NEXT:  1      1     0.33                        cnth	x0, all, mul #16
 # CHECK-NEXT:  1      1     0.33                        cnth	x0, pow2
-# CHECK-NEXT:  1      6     1.00                        cntp	x0, p15, p0.b
-# CHECK-NEXT:  1      6     1.00                        cntp	x0, p15, p0.d
-# CHECK-NEXT:  1      6     1.00                        cntp	x0, p15, p0.h
-# CHECK-NEXT:  1      6     1.00                        cntp	x0, p15, p0.s
+# CHECK-NEXT:  1      4     1.00                        cntp	x0, p15, p0.b
+# CHECK-NEXT:  1      4     1.00                        cntp	x0, p15, p0.d
+# CHECK-NEXT:  1      4     1.00                        cntp	x0, p15, p0.h
+# CHECK-NEXT:  1      4     1.00                        cntp	x0, p15, p0.s
 # CHECK-NEXT:  1      1     0.33                        cntw	x0
 # CHECK-NEXT:  1      1     0.33                        cntw	x0, #28
 # CHECK-NEXT:  1      1     0.33                        cntw	x0, all, mul #16
@@ -3892,42 +3892,42 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      1     0.33                        ctermne	wzr, w30
 # CHECK-NEXT:  1      1     0.33                        ctermne	x30, xzr
 # CHECK-NEXT:  1      1     0.33                        ctermne	xzr, x30
-# CHECK-NEXT:  1      1     0.33                        decb	x0
-# CHECK-NEXT:  1      1     0.33                        decb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        decb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        decb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        decb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        decd	x0
-# CHECK-NEXT:  1      1     0.33                        decd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        decd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        decd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        decd	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        dech	x0
-# CHECK-NEXT:  1      1     0.33                        dech	x0, #14
-# CHECK-NEXT:  1      1     0.33                        dech	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        dech	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        dech	x0, vl1
-# CHECK-NEXT:  1      6     1.00                        decp	x0, p0.b
-# CHECK-NEXT:  1      6     1.00                        decp	x0, p0.d
-# CHECK-NEXT:  1      6     1.00                        decp	x0, p0.h
-# CHECK-NEXT:  1      6     1.00                        decp	x0, p0.s
-# CHECK-NEXT:  1      6     1.00                        decp	xzr, p15.b
-# CHECK-NEXT:  1      6     1.00                        decp	xzr, p15.d
-# CHECK-NEXT:  1      6     1.00                        decp	xzr, p15.h
-# CHECK-NEXT:  1      6     1.00                        decp	xzr, p15.s
+# CHECK-NEXT:  1      3     0.33                        decb	x0
+# CHECK-NEXT:  1      3     0.33                        decb	x0, #14
+# CHECK-NEXT:  1      3     0.33                        decb	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        decb	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        decb	x0, vl1
+# CHECK-NEXT:  1      3     0.33                        decd	x0
+# CHECK-NEXT:  1      3     0.33                        decd	x0, #14
+# CHECK-NEXT:  1      3     0.33                        decd	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        decd	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        decd	x0, vl1
+# CHECK-NEXT:  1      3     0.33                        dech	x0
+# CHECK-NEXT:  1      3     0.33                        dech	x0, #14
+# CHECK-NEXT:  1      3     0.33                        dech	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        dech	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        dech	x0, vl1
+# CHECK-NEXT:  1      4     1.00                        decp	x0, p0.b
+# CHECK-NEXT:  1      4     1.00                        decp	x0, p0.d
+# CHECK-NEXT:  1      4     1.00                        decp	x0, p0.h
+# CHECK-NEXT:  1      4     1.00                        decp	x0, p0.s
+# CHECK-NEXT:  1      4     1.00                        decp	xzr, p15.b
+# CHECK-NEXT:  1      4     1.00                        decp	xzr, p15.d
+# CHECK-NEXT:  1      4     1.00                        decp	xzr, p15.h
+# CHECK-NEXT:  1      4     1.00                        decp	xzr, p15.s
 # CHECK-NEXT:  1      4     0.50                        decp	z31.d, p15.d
 # CHECK-NEXT:  1      4     0.50                        decp	z31.h, p15.h
 # CHECK-NEXT:  1      4     0.50                        decp	z31.s, p15.s
-# CHECK-NEXT:  1      1     0.33                        decw	x0
-# CHECK-NEXT:  1      1     0.33                        decw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        decw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        decw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        decw	x0, vl1
+# CHECK-NEXT:  1      3     0.33                        decw	x0
+# CHECK-NEXT:  1      3     0.33                        decw	x0, #14
+# CHECK-NEXT:  1      3     0.33                        decw	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        decw	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        decw	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        dupm	z0.d, #0xfffffffffffffff9
 # CHECK-NEXT:  1      4     0.50                        dupm	z0.s, #0xfffffff9
 # CHECK-NEXT:  1      4     0.50                        dupm	z23.h, #0xfff9
 # CHECK-NEXT:  1      4     0.50                        dupm	z5.b, #0xf9
-# CHECK-NEXT:  1      6     1.00                        eor	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        eor	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      3     0.50                        eor	z0.d, z0.d, #0x6
 # CHECK-NEXT:  1      3     0.50                        eor	z0.d, z0.d, #0xfffffffffffffff9
 # CHECK-NEXT:  1      3     0.50                        eor	z0.d, z0.d, z0.d
@@ -3942,12 +3942,12 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        eor	z31.s, p7/m, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        eor	z5.b, z5.b, #0x6
 # CHECK-NEXT:  1      3     0.50                        eor	z5.b, z5.b, #0xf9
-# CHECK-NEXT:  1      3     0.50                        eor3	z29.d, z29.d, z30.d, z31.d
+# CHECK-NEXT:  1      4     0.50                        eor3	z29.d, z29.d, z30.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        eorbt	z0.b, z1.b, z31.b
 # CHECK-NEXT:  1      4     0.50                        eorbt	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        eorbt	z0.h, z1.h, z31.h
 # CHECK-NEXT:  1      4     0.50                        eorbt	z0.s, z1.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        eors	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        eors	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      4     0.50                        eortb	z0.b, z1.b, z31.b
 # CHECK-NEXT:  1      4     0.50                        eortb	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        eortb	z0.h, z1.h, z31.h
@@ -4303,49 +4303,49 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        ftsmul	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        ftsmul	z0.h, z1.h, z31.h
 # CHECK-NEXT:  1      4     0.50                        ftsmul	z0.s, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        ftssel	z0.d, z1.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        ftssel	z0.h, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        ftssel	z0.s, z1.s, z31.s
+# CHECK-NEXT:  1      3     0.50                        ftssel	z0.d, z1.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        ftssel	z0.h, z1.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        ftssel	z0.s, z1.s, z31.s
 # CHECK-NEXT:  1      8     2.00                        histcnt	z0.s, p0/z, z1.s, z2.s
 # CHECK-NEXT:  1      8     2.00                        histcnt	z29.d, p7/z, z30.d, z31.d
 # CHECK-NEXT:  1      8     2.00                        histseg	z0.b, z1.b, z31.b
-# CHECK-NEXT:  1      1     0.33                        incb	x0
-# CHECK-NEXT:  1      1     0.33                        incb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        incb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        incb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        incb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        incd	x0
-# CHECK-NEXT:  1      1     0.33                        incd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        incd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        incd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        incd	x0, vl1
-# CHECK-NEXT:  1      4     0.50                        incd	z0.d
-# CHECK-NEXT:  1      4     0.50                        incd	z0.d, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        inch	x0
-# CHECK-NEXT:  1      1     0.33                        inch	x0, #14
-# CHECK-NEXT:  1      1     0.33                        inch	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        inch	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        inch	x0, vl1
-# CHECK-NEXT:  1      4     0.50                        inch	z0.h
-# CHECK-NEXT:  1      4     0.50                        inch	z0.h, all, mul #16
-# CHECK-NEXT:  1      6     1.00                        incp	x0, p0.b
-# CHECK-NEXT:  1      6     1.00                        incp	x0, p0.d
-# CHECK-NEXT:  1      6     1.00                        incp	x0, p0.h
-# CHECK-NEXT:  1      6     1.00                        incp	x0, p0.s
-# CHECK-NEXT:  1      6     1.00                        incp	xzr, p15.b
-# CHECK-NEXT:  1      6     1.00                        incp	xzr, p15.d
-# CHECK-NEXT:  1      6     1.00                        incp	xzr, p15.h
-# CHECK-NEXT:  1      6     1.00                        incp	xzr, p15.s
+# CHECK-NEXT:  1      3     0.33                        incb	x0
+# CHECK-NEXT:  1      3     0.33                        incb	x0, #14
+# CHECK-NEXT:  1      3     0.33                        incb	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        incb	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        incb	x0, vl1
+# CHECK-NEXT:  1      3     0.33                        incd	x0
+# CHECK-NEXT:  1      3     0.33                        incd	x0, #14
+# CHECK-NEXT:  1      3     0.33                        incd	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        incd	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        incd	x0, vl1
+# CHECK-NEXT:  1      3     0.50                        incd	z0.d
+# CHECK-NEXT:  1      3     0.50                        incd	z0.d, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        inch	x0
+# CHECK-NEXT:  1      3     0.33                        inch	x0, #14
+# CHECK-NEXT:  1      3     0.33                        inch	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        inch	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        inch	x0, vl1
+# CHECK-NEXT:  1      3     0.50                        inch	z0.h
+# CHECK-NEXT:  1      3     0.50                        inch	z0.h, all, mul #16
+# CHECK-NEXT:  1      4     1.00                        incp	x0, p0.b
+# CHECK-NEXT:  1      4     1.00                        incp	x0, p0.d
+# CHECK-NEXT:  1      4     1.00                        incp	x0, p0.h
+# CHECK-NEXT:  1      4     1.00                        incp	x0, p0.s
+# CHECK-NEXT:  1      4     1.00                        incp	xzr, p15.b
+# CHECK-NEXT:  1      4     1.00                        incp	xzr, p15.d
+# CHECK-NEXT:  1      4     1.00                        incp	xzr, p15.h
+# CHECK-NEXT:  1      4     1.00                        incp	xzr, p15.s
 # CHECK-NEXT:  1      4     0.50                        incp	z31.d, p15.d
 # CHECK-NEXT:  1      4     0.50                        incp	z31.h, p15.h
 # CHECK-NEXT:  1      4     0.50                        incp	z31.s, p15.s
-# CHECK-NEXT:  1      1     0.33                        incw	x0
-# CHECK-NEXT:  1      1     0.33                        incw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        incw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        incw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        incw	x0, vl1
-# CHECK-NEXT:  1      4     0.50                        incw	z0.s
-# CHECK-NEXT:  1      4     0.50                        incw	z0.s, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        incw	x0
+# CHECK-NEXT:  1      3     0.33                        incw	x0, #14
+# CHECK-NEXT:  1      3     0.33                        incw	x0, all, mul #16
+# CHECK-NEXT:  1      3     0.33                        incw	x0, pow2
+# CHECK-NEXT:  1      3     0.33                        incw	x0, vl1
+# CHECK-NEXT:  1      3     0.50                        incw	z0.s
+# CHECK-NEXT:  1      3     0.50                        incw	z0.s, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        index	z0.b, #0, #0
 # CHECK-NEXT:  1      4     0.50                        index	z0.d, #0, #0
 # CHECK-NEXT:  1      4     0.50                        index	z0.h, #0, #0
@@ -4412,8 +4412,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50    *                   ld1b	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  1      7     7.00    *                   ld1b	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *                   ld1b	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  1      9     4.50    *                   ld1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *                   ld1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      3     0.50    *                   ld1b	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  1      9     9.00    *                   ld1b	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      3     0.50    *                   ld1b	{ z21.b }, p5/z, [x10, #5, mul vl]
@@ -4450,8 +4450,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  1      7     7.00    *                   ld1h	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  1      9     4.50    *                   ld1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *                   ld1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  1      9     9.00    *                   ld1h	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -4467,8 +4467,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *                   ld1h	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z31.h }, p7/z, [sp, #-1, mul vl]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  1      9     4.50    *                   ld1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  1      9     4.50    *                   ld1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  1      7     3.50    *                   ld1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  1      7     3.50    *                   ld1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  1      9     9.00    *                   ld1h	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z5.h }, p3/z, [sp, x16, lsl #1]
 # CHECK-NEXT:  1      3     0.50    *                   ld1h	{ z5.h }, p3/z, [x17, x16, lsl #1]
@@ -4529,7 +4529,7 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50    *                   ld1sb	{ z0.h }, p0/z, [sp, x0]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sb	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sb	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  1      9     4.50    *                   ld1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sb	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  1      9     9.00    *                   ld1sb	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sb	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -4549,8 +4549,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *                   ld1sh	{ z0.d }, p0/z, [x0, z0.d, uxtw #1]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sh	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  1      7     7.00    *                   ld1sh	{ z0.d }, p0/z, [z0.d]
-# CHECK-NEXT:  1      9     4.50    *                   ld1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *                   ld1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sh	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  1      9     9.00    *                   ld1sh	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sh	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -4565,8 +4565,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *                   ld1sh	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  1      7     7.00    *                   ld1sh	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  1      3     0.50    *                   ld1sh	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  1      9     4.50    *                   ld1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  1      9     4.50    *                   ld1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  1      7     3.50    *                   ld1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  1      7     3.50    *                   ld1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  1      9     9.00    *                   ld1sh	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  1      7     7.00    *                   ld1sw	{ z0.d }, p0/z, [x0, z0.d, sxtw #2]
 # CHECK-NEXT:  1      7     7.00    *                   ld1sw	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
@@ -4585,8 +4585,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *                   ld1w	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
 # CHECK-NEXT:  1      3     0.50    *                   ld1w	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  1      7     7.00    *                   ld1w	{ z0.d }, p0/z, [z0.d]
-# CHECK-NEXT:  1      9     4.50    *                   ld1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *                   ld1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *                   ld1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      3     0.50    *                   ld1w	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  1      9     9.00    *                   ld1w	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      3     0.50    *                   ld1w	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -4601,8 +4601,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *                   ld1w	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  1      7     7.00    *                   ld1w	{ z31.d }, p7/z, [z31.d, #124]
 # CHECK-NEXT:  1      3     0.50    *                   ld1w	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  1      9     4.50    *                   ld1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
-# CHECK-NEXT:  1      9     4.50    *                   ld1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
+# CHECK-NEXT:  1      7     3.50    *                   ld1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
+# CHECK-NEXT:  1      7     3.50    *                   ld1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
 # CHECK-NEXT:  1      9     9.00    *                   ld1w	{ z31.s }, p7/z, [z31.s, #124]
 # CHECK-NEXT:  1      3     2.00    *                   ld2b	{ z0.b, z1.b }, p0/z, [x0, x0]
 # CHECK-NEXT:  1      3     1.00    *                   ld2b	{ z0.b, z1.b }, p0/z, [x0]
@@ -4668,8 +4668,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1b	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1b	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1b	{ z0.s }, p0/z, [x0, x0]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1b	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1b	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1b	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -4696,8 +4696,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1h	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1h	{ z0.h }, p0/z, [x0, x0, lsl #1]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1h	{ z0.s }, p0/z, [x0, x0, lsl #1]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1h	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1h	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1h	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -4706,16 +4706,16 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1h	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1h	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1h	{ z31.h }, p7/z, [sp]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1h	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1h	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sb	{ z0.d }, p0/z, [x0, x0]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sb	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sb	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sb	{ z0.s }, p0/z, [x0, x0]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1sb	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sb	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sb	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -4730,8 +4730,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z0.d }, p0/z, [x0, z0.d, uxtw #1]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sh	{ z0.s }, p0/z, [x0, x0, lsl #1]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1sh	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -4739,8 +4739,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sh	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1sh	{ z31.d }, p7/z, [z31.d, #62]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sh	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1sh	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1sw	{ z0.d }, p0/z, [x0, x0, lsl #2]
@@ -4758,8 +4758,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1w	{ z0.s }, p0/z, [x0, x0, lsl #2]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1w	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -4767,8 +4767,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1w	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  1      7     7.00    *             U     ldff1w	{ z31.d }, p7/z, [z31.d, #124]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
-# CHECK-NEXT:  1      9     4.50    *             U     ldff1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
+# CHECK-NEXT:  1      7     3.50    *             U     ldff1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
 # CHECK-NEXT:  1      3     0.50    *             U     ldff1w	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  1      9     9.00    *             U     ldff1w	{ z31.s }, p7/z, [z31.s, #124]
 # CHECK-NEXT:  1      3     0.50    *             U     ldnf1b	{ z0.b }, p0/z, [x0]
@@ -4959,12 +4959,12 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        mls	z0.h, z1.h, z7.h[7]
 # CHECK-NEXT:  1      4     0.50                        mls	z0.s, p7/m, z1.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        mls	z0.s, z1.s, z7.s[3]
-# CHECK-NEXT:  1      6     1.00                        mov	p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        mov	p0.b, p0/m, p0.b
-# CHECK-NEXT:  1      6     1.00                        mov	p0.b, p0/z, p0.b
-# CHECK-NEXT:  1      6     1.00                        mov	p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        mov	p15.b, p15/m, p15.b
-# CHECK-NEXT:  1      6     1.00                        mov	p15.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        mov	p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        mov	p0.b, p0/m, p0.b
+# CHECK-NEXT:  1      2     1.00                        mov	p0.b, p0/z, p0.b
+# CHECK-NEXT:  1      2     1.00                        mov	p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        mov	p15.b, p15/m, p15.b
+# CHECK-NEXT:  1      2     1.00                        mov	p15.b, p15/z, p15.b
 # CHECK-NEXT:  1      3     0.50                        mov	z0.b, #127
 # CHECK-NEXT:  1      3     0.50                        mov	z0.b, b0
 # CHECK-NEXT:  1      3     0.50                        mov	z0.b, p0/m, b0
@@ -5062,10 +5062,10 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        mov	z5.h, #-6
 # CHECK-NEXT:  1      3     0.50                        mov	z5.q, z17.q[3]
 # CHECK-NEXT:  1      3     0.50                        mov	z5.s, #-6
-# CHECK-NEXT:  1      6     1.00                        movs	p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        movs	p0.b, p0/z, p0.b
-# CHECK-NEXT:  1      6     1.00                        movs	p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        movs	p15.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        movs	p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        movs	p0.b, p0/z, p0.b
+# CHECK-NEXT:  1      2     1.00                        movs	p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        movs	p15.b, p15/z, p15.b
 # CHECK-NEXT:  1      1     1.00                  U     mrs	x3, ID_AA64ZFR0_EL1
 # CHECK-NEXT:  1      1     1.00                  U     mrs	x3, ZCR_EL1
 # CHECK-NEXT:  1      1     1.00                  U     mrs	x3, ZCR_EL12
@@ -5098,10 +5098,10 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        mul	z31.h, z31.h, #127
 # CHECK-NEXT:  1      4     0.50                        mul	z31.s, z31.s, #-128
 # CHECK-NEXT:  1      4     0.50                        mul	z31.s, z31.s, #127
-# CHECK-NEXT:  1      6     1.00                        nand	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        nand	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        nands	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        nands	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        nand	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        nand	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        nands	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        nands	p15.b, p15/z, p15.b, p15.b
 # CHECK-NEXT:  1      3     0.50                        nbsl	z0.d, z0.d, z1.d, z2.d
 # CHECK-NEXT:  1      3     0.50                        neg	z0.b, p0/m, z0.b
 # CHECK-NEXT:  1      3     0.50                        neg	z0.d, p0/m, z0.d
@@ -5115,23 +5115,23 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     1.00                        nmatch	p0.h, p0/z, z0.h, z0.h
 # CHECK-NEXT:  1      7     1.00                        nmatch	p15.b, p7/z, z30.b, z31.b
 # CHECK-NEXT:  1      7     1.00                        nmatch	p15.h, p7/z, z30.h, z31.h
-# CHECK-NEXT:  1      6     1.00                        nor	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        nor	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        nors	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        nors	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        not	p0.b, p0/z, p0.b
-# CHECK-NEXT:  1      6     1.00                        not	p15.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        nor	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        nor	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        nors	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        nors	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        not	p0.b, p0/z, p0.b
+# CHECK-NEXT:  1      2     1.00                        not	p15.b, p15/z, p15.b
 # CHECK-NEXT:  1      3     0.50                        not	z31.b, p7/m, z31.b
 # CHECK-NEXT:  1      3     0.50                        not	z31.d, p7/m, z31.d
 # CHECK-NEXT:  1      3     0.50                        not	z31.h, p7/m, z31.h
 # CHECK-NEXT:  1      3     0.50                        not	z31.s, p7/m, z31.s
-# CHECK-NEXT:  1      6     1.00                        nots	p0.b, p0/z, p0.b
-# CHECK-NEXT:  1      6     1.00                        nots	p15.b, p15/z, p15.b
-# CHECK-NEXT:  1      6     1.00                        orn	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        orn	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        orns	p0.b, p0/z, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        orns	p15.b, p15/z, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        orr	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        nots	p0.b, p0/z, p0.b
+# CHECK-NEXT:  1      2     1.00                        nots	p15.b, p15/z, p15.b
+# CHECK-NEXT:  1      2     1.00                        orn	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        orn	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        orns	p0.b, p0/z, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        orns	p15.b, p15/z, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        orr	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      3     0.50                        orr	z0.d, z0.d, #0x6
 # CHECK-NEXT:  1      3     0.50                        orr	z0.d, z0.d, #0xfffffffffffffff9
 # CHECK-NEXT:  1      3     0.50                        orr	z0.s, z0.s, #0x6
@@ -5145,27 +5145,27 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        orr	z31.s, p7/m, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        orr	z5.b, z5.b, #0x6
 # CHECK-NEXT:  1      3     0.50                        orr	z5.b, z5.b, #0xf9
-# CHECK-NEXT:  1      6     1.00                        orrs	p0.b, p0/z, p0.b, p1.b
+# CHECK-NEXT:  1      2     1.00                        orrs	p0.b, p0/z, p0.b, p1.b
 # CHECK-NEXT:  1      4     1.00                        orv	b0, p7, z31.b
 # CHECK-NEXT:  1      4     1.00                        orv	d0, p7, z31.d
 # CHECK-NEXT:  1      4     1.00                        orv	h0, p7, z31.h
 # CHECK-NEXT:  1      4     1.00                        orv	s0, p7, z31.s
-# CHECK-NEXT:  1      6     1.00                        pfalse	p15.b
-# CHECK-NEXT:  1      6     1.00                        pfirst	p0.b, p15, p0.b
-# CHECK-NEXT:  1      6     1.00                        pfirst	p15.b, p15, p15.b
+# CHECK-NEXT:  1      2     1.00                        pfalse	p15.b
+# CHECK-NEXT:  1      2     1.00                        pfirst	p0.b, p15, p0.b
+# CHECK-NEXT:  1      2     1.00                        pfirst	p15.b, p15, p15.b
 # CHECK-NEXT:  1      4     0.50                        pmul	z0.b, z1.b, z2.b
 # CHECK-NEXT:  1      4     0.50                        pmul	z29.b, z30.b, z31.b
-# CHECK-NEXT:  1      6     1.00                        pmullb	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      6     1.00                        pmullb	z29.q, z30.d, z31.d
-# CHECK-NEXT:  1      6     1.00                        pmullb	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        pmullt	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      6     1.00                        pmullt	z29.q, z30.d, z31.d
-# CHECK-NEXT:  1      6     1.00                        pmullt	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        pnext	p0.b, p15, p0.b
-# CHECK-NEXT:  1      6     1.00                        pnext	p0.d, p15, p0.d
-# CHECK-NEXT:  1      6     1.00                        pnext	p0.h, p15, p0.h
-# CHECK-NEXT:  1      6     1.00                        pnext	p0.s, p15, p0.s
-# CHECK-NEXT:  1      6     1.00                        pnext	p15.b, p15, p15.b
+# CHECK-NEXT:  1      9     1.00                        pmullb	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      9     1.00                        pmullb	z29.q, z30.d, z31.d
+# CHECK-NEXT:  1      9     1.00                        pmullb	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      9     1.00                        pmullt	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      9     1.00                        pmullt	z29.q, z30.d, z31.d
+# CHECK-NEXT:  1      9     1.00                        pmullt	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      2     1.00                        pnext	p0.b, p15, p0.b
+# CHECK-NEXT:  1      2     1.00                        pnext	p0.d, p15, p0.d
+# CHECK-NEXT:  1      2     1.00                        pnext	p0.h, p15, p0.h
+# CHECK-NEXT:  1      2     1.00                        pnext	p0.s, p15, p0.s
+# CHECK-NEXT:  1      2     1.00                        pnext	p15.b, p15, p15.b
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfb	#14, p0, [x0]
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfb	#15, p0, [x0]
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfb	#6, p0, [x0]
@@ -5274,97 +5274,97 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfw	pstl2strm, p0, [x0]
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfw	pstl3keep, p0, [x0]
 # CHECK-NEXT:  1      0     0.50    *      *      U     prfw	pstl3strm, p0, [x0]
-# CHECK-NEXT:  1      6     1.00                        ptest	p15, p0.b
-# CHECK-NEXT:  1      6     1.00                        ptest	p15, p15.b
-# CHECK-NEXT:  1      6     1.00                        ptrue	p0.b, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrue	p0.d, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrue	p0.h, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrue	p0.s, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrue	p15.b
-# CHECK-NEXT:  1      6     1.00                        ptrue	p15.d
-# CHECK-NEXT:  1      6     1.00                        ptrue	p15.h
-# CHECK-NEXT:  1      6     1.00                        ptrue	p15.s
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #14
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #15
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #16
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #17
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #18
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #19
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #20
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #21
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #22
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #23
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #24
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #25
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #26
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #27
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, #28
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, mul3
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, mul4
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl1
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl128
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl16
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl2
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl256
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl3
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl32
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl4
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl5
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl6
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl64
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl7
-# CHECK-NEXT:  1      6     1.00                        ptrue	p7.s, vl8
-# CHECK-NEXT:  1      6     1.00                        ptrues	p0.b, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrues	p0.d, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrues	p0.h, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrues	p0.s, pow2
-# CHECK-NEXT:  1      6     1.00                        ptrues	p15.b
-# CHECK-NEXT:  1      6     1.00                        ptrues	p15.d
-# CHECK-NEXT:  1      6     1.00                        ptrues	p15.h
-# CHECK-NEXT:  1      6     1.00                        ptrues	p15.s
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #14
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #15
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #16
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #17
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #18
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #19
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #20
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #21
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #22
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #23
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #24
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #25
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #26
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #27
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, #28
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, mul3
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, mul4
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl1
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl128
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl16
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl2
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl256
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl3
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl32
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl4
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl5
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl6
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl64
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl7
-# CHECK-NEXT:  1      6     1.00                        ptrues	p7.s, vl8
-# CHECK-NEXT:  1      6     1.00                        punpkhi	p0.h, p0.b
-# CHECK-NEXT:  1      6     1.00                        punpkhi	p15.h, p15.b
-# CHECK-NEXT:  1      6     1.00                        punpklo	p0.h, p0.b
-# CHECK-NEXT:  1      6     1.00                        punpklo	p15.h, p15.b
-# CHECK-NEXT:  1      4     0.50                        raddhnb	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        raddhnb	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        raddhnb	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        raddhnt	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        raddhnt	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        raddhnt	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      8     1.00                        rax1	z0.d, z1.d, z31.d
+# CHECK-NEXT:  1      2     1.00                        ptest	p15, p0.b
+# CHECK-NEXT:  1      2     1.00                        ptest	p15, p15.b
+# CHECK-NEXT:  1      2     1.00                        ptrue	p0.b, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrue	p0.d, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrue	p0.h, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrue	p0.s, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrue	p15.b
+# CHECK-NEXT:  1      2     1.00                        ptrue	p15.d
+# CHECK-NEXT:  1      2     1.00                        ptrue	p15.h
+# CHECK-NEXT:  1      2     1.00                        ptrue	p15.s
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #14
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #15
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #16
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #17
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #18
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #19
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #20
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #21
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #22
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #23
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #24
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #25
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #26
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #27
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, #28
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, mul3
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, mul4
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl1
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl128
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl16
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl2
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl256
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl3
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl32
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl4
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl5
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl6
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl64
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl7
+# CHECK-NEXT:  1      2     1.00                        ptrue	p7.s, vl8
+# CHECK-NEXT:  1      2     1.00                        ptrues	p0.b, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrues	p0.d, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrues	p0.h, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrues	p0.s, pow2
+# CHECK-NEXT:  1      2     1.00                        ptrues	p15.b
+# CHECK-NEXT:  1      2     1.00                        ptrues	p15.d
+# CHECK-NEXT:  1      2     1.00                        ptrues	p15.h
+# CHECK-NEXT:  1      2     1.00                        ptrues	p15.s
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #14
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #15
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #16
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #17
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #18
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #19
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #20
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #21
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #22
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #23
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #24
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #25
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #26
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #27
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, #28
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, mul3
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, mul4
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl1
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl128
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl16
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl2
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl256
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl3
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl32
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl4
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl5
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl6
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl64
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl7
+# CHECK-NEXT:  1      2     1.00                        ptrues	p7.s, vl8
+# CHECK-NEXT:  1      2     1.00                        punpkhi	p0.h, p0.b
+# CHECK-NEXT:  1      2     1.00                        punpkhi	p15.h, p15.b
+# CHECK-NEXT:  1      2     1.00                        punpklo	p0.h, p0.b
+# CHECK-NEXT:  1      2     1.00                        punpklo	p15.h, p15.b
+# CHECK-NEXT:  1      8     0.50                        raddhnb	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        raddhnb	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        raddhnb	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        raddhnt	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        raddhnt	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        raddhnt	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      9     1.00                        rax1	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        rbit	z0.b, p7/m, z31.b
 # CHECK-NEXT:  1      3     0.50                        rbit	z0.d, p7/m, z31.d
 # CHECK-NEXT:  1      3     0.50                        rbit	z0.h, p7/m, z31.h
@@ -5379,16 +5379,16 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      1     0.33                        rdvl	x21, #-32
 # CHECK-NEXT:  1      1     0.33                        rdvl	x23, #31
 # CHECK-NEXT:  1      1     0.33                        rdvl	xzr, #-1
-# CHECK-NEXT:  1      4     0.50                        rev	z0.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        rev	z0.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        rev	z0.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        rev	z0.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        revb	z0.d, p7/m, z31.d
-# CHECK-NEXT:  1      4     0.50                        revb	z0.h, p7/m, z31.h
-# CHECK-NEXT:  1      4     0.50                        revb	z0.s, p7/m, z31.s
-# CHECK-NEXT:  1      4     0.50                        revh	z0.d, p7/m, z31.d
-# CHECK-NEXT:  1      4     0.50                        revh	z0.s, p7/m, z31.s
-# CHECK-NEXT:  1      4     0.50                        revw	z0.d, p7/m, z31.d
+# CHECK-NEXT:  1      3     0.50                        rev	z0.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        rev	z0.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        rev	z0.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        rev	z0.s, z31.s
+# CHECK-NEXT:  1      3     0.50                        revb	z0.d, p7/m, z31.d
+# CHECK-NEXT:  1      3     0.50                        revb	z0.h, p7/m, z31.h
+# CHECK-NEXT:  1      3     0.50                        revb	z0.s, p7/m, z31.s
+# CHECK-NEXT:  1      3     0.50                        revh	z0.d, p7/m, z31.d
+# CHECK-NEXT:  1      3     0.50                        revh	z0.s, p7/m, z31.s
+# CHECK-NEXT:  1      3     0.50                        revw	z0.d, p7/m, z31.d
 # CHECK-NEXT:  1      4     0.50                        rshrnb	z0.b, z0.h, #1
 # CHECK-NEXT:  1      4     0.50                        rshrnb	z0.h, z0.s, #1
 # CHECK-NEXT:  1      4     0.50                        rshrnb	z0.s, z0.d, #1
@@ -5401,22 +5401,22 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        rshrnt	z31.b, z31.h, #8
 # CHECK-NEXT:  1      4     0.50                        rshrnt	z31.h, z31.s, #16
 # CHECK-NEXT:  1      4     0.50                        rshrnt	z31.s, z31.d, #32
-# CHECK-NEXT:  1      4     0.50                        rsubhnb	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        rsubhnb	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        rsubhnb	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        rsubhnt	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        rsubhnt	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        rsubhnt	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      8     1.00                        saba	z0.b, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        saba	z0.d, z1.d, z31.d
-# CHECK-NEXT:  1      8     1.00                        saba	z0.h, z1.h, z31.h
-# CHECK-NEXT:  1      8     1.00                        saba	z0.s, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        sabalb	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        sabalb	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        sabalb	z0.s, z1.h, z31.h
-# CHECK-NEXT:  1      8     1.00                        sabalt	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        sabalt	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        sabalt	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        rsubhnb	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        rsubhnb	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        rsubhnb	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        rsubhnt	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        rsubhnt	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        rsubhnt	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      6     1.00                        saba	z0.b, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        saba	z0.d, z1.d, z31.d
+# CHECK-NEXT:  1      6     1.00                        saba	z0.h, z1.h, z31.h
+# CHECK-NEXT:  1      6     1.00                        saba	z0.s, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        sabalb	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        sabalb	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        sabalb	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      6     1.00                        sabalt	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        sabalt	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        sabalt	z0.s, z1.h, z31.h
 # CHECK-NEXT:  1      3     0.50                        sabd	z31.b, p7/m, z31.b, z31.b
 # CHECK-NEXT:  1      3     0.50                        sabd	z31.d, p7/m, z31.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        sabd	z31.h, p7/m, z31.h, z31.h
@@ -5430,24 +5430,24 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     1.00                        sadalp	z0.h, p0/m, z1.b
 # CHECK-NEXT:  1      7     1.00                        sadalp	z29.s, p0/m, z30.h
 # CHECK-NEXT:  1      7     1.00                        sadalp	z30.d, p7/m, z31.s
-# CHECK-NEXT:  1      3     0.50                        saddlb	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        saddlb	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        saddlb	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        saddlbt	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        saddlbt	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      3     0.50                        saddlbt	z0.s, z1.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        saddlt	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        saddlt	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        saddlt	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        saddlb	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        saddlb	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        saddlb	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        saddlbt	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        saddlbt	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      4     0.50                        saddlbt	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        saddlt	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        saddlt	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        saddlt	z31.d, z31.s, z31.s
 # CHECK-NEXT:  1      4     1.00                        saddv	d0, p7, z31.b
 # CHECK-NEXT:  1      4     1.00                        saddv	d0, p7, z31.h
 # CHECK-NEXT:  1      4     1.00                        saddv	d0, p7, z31.s
-# CHECK-NEXT:  1      3     0.50                        saddwb	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        saddwb	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        saddwb	z31.d, z31.d, z31.s
-# CHECK-NEXT:  1      3     0.50                        saddwt	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        saddwt	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        saddwt	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        saddwb	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        saddwb	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        saddwb	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        saddwt	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        saddwt	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        saddwt	z31.d, z31.d, z31.s
 # CHECK-NEXT:  1      4     0.50                        sbclb	z0.d, z1.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        sbclb	z0.s, z1.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        sbclt	z0.d, z1.d, z31.d
@@ -5504,8 +5504,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        sli	z31.d, z31.d, #63
 # CHECK-NEXT:  1      3     0.50                        sli	z31.h, z31.h, #15
 # CHECK-NEXT:  1      3     0.50                        sli	z31.s, z31.s, #31
-# CHECK-NEXT:  1      8     1.00                        sm4e	z0.s, z0.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        sm4ekey	z0.s, z1.s, z31.s
+# CHECK-NEXT:  1      9     1.00                        sm4e	z0.s, z0.s, z31.s
+# CHECK-NEXT:  1      9     1.00                        sm4ekey	z0.s, z1.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        smax	z0.b, z0.b, #-128
 # CHECK-NEXT:  1      3     0.50                        smax	z0.d, z0.d, #-128
 # CHECK-NEXT:  1      3     0.50                        smax	z0.h, z0.h, #-128
@@ -5624,61 +5624,61 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        sqcadd	z31.d, z31.d, z31.d, #270
 # CHECK-NEXT:  1      4     0.50                        sqcadd	z31.h, z31.h, z31.h, #270
 # CHECK-NEXT:  1      4     0.50                        sqcadd	z31.s, z31.s, z31.s, #270
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecb	x0, w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecd	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecb	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecd	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdecd	z0.d
 # CHECK-NEXT:  1      4     0.50                        sqdecd	z0.d, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdecd	z0.d, pow2
 # CHECK-NEXT:  1      4     0.50                        sqdecd	z0.d, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdech	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdech	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdech	z0.h
 # CHECK-NEXT:  1      4     0.50                        sqdech	z0.h, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdech	z0.h, pow2
 # CHECK-NEXT:  1      4     0.50                        sqdech	z0.h, pow2, mul #16
-# CHECK-NEXT:  1      8     1.00                        sqdecp	x0, p0.b
-# CHECK-NEXT:  1      8     1.00                        sqdecp	x0, p0.d
-# CHECK-NEXT:  1      8     1.00                        sqdecp	x0, p0.h
-# CHECK-NEXT:  1      8     1.00                        sqdecp	x0, p0.s
-# CHECK-NEXT:  1      8     1.00                        sqdecp	xzr, p15.b, wzr
-# CHECK-NEXT:  1      8     1.00                        sqdecp	xzr, p15.d, wzr
-# CHECK-NEXT:  1      8     1.00                        sqdecp	xzr, p15.h, wzr
-# CHECK-NEXT:  1      8     1.00                        sqdecp	xzr, p15.s, wzr
+# CHECK-NEXT:  1      9     1.00                        sqdecp	x0, p0.b
+# CHECK-NEXT:  1      9     1.00                        sqdecp	x0, p0.d
+# CHECK-NEXT:  1      9     1.00                        sqdecp	x0, p0.h
+# CHECK-NEXT:  1      9     1.00                        sqdecp	x0, p0.s
+# CHECK-NEXT:  1      9     1.00                        sqdecp	xzr, p15.b, wzr
+# CHECK-NEXT:  1      9     1.00                        sqdecp	xzr, p15.d, wzr
+# CHECK-NEXT:  1      9     1.00                        sqdecp	xzr, p15.h, wzr
+# CHECK-NEXT:  1      9     1.00                        sqdecp	xzr, p15.s, wzr
 # CHECK-NEXT:  1      4     0.50                        sqdecp	z0.d, p0.d
 # CHECK-NEXT:  1      4     0.50                        sqdecp	z0.h, p0.h
 # CHECK-NEXT:  1      4     0.50                        sqdecp	z0.s, p0.s
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqdecw	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqdecw	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdecw	z0.s
 # CHECK-NEXT:  1      4     0.50                        sqdecw	z0.s, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqdecw	z0.s, pow2
@@ -5726,61 +5726,61 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        sqdmullt	z0.s, z1.h, z7.h[7]
 # CHECK-NEXT:  1      4     0.50                        sqdmullt	z29.s, z30.h, z31.h
 # CHECK-NEXT:  1      4     0.50                        sqdmullt	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincb	x0, w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincd	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincb	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincd	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqincd	z0.d
 # CHECK-NEXT:  1      4     0.50                        sqincd	z0.d, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqincd	z0.d, pow2
 # CHECK-NEXT:  1      4     0.50                        sqincd	z0.d, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqinch	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqinch	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqinch	z0.h
 # CHECK-NEXT:  1      4     0.50                        sqinch	z0.h, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqinch	z0.h, pow2
 # CHECK-NEXT:  1      4     0.50                        sqinch	z0.h, pow2, mul #16
-# CHECK-NEXT:  1      8     1.00                        sqincp	x0, p0.b
-# CHECK-NEXT:  1      8     1.00                        sqincp	x0, p0.d
-# CHECK-NEXT:  1      8     1.00                        sqincp	x0, p0.h
-# CHECK-NEXT:  1      8     1.00                        sqincp	x0, p0.s
-# CHECK-NEXT:  1      8     1.00                        sqincp	xzr, p15.b, wzr
-# CHECK-NEXT:  1      8     1.00                        sqincp	xzr, p15.d, wzr
-# CHECK-NEXT:  1      8     1.00                        sqincp	xzr, p15.h, wzr
-# CHECK-NEXT:  1      8     1.00                        sqincp	xzr, p15.s, wzr
+# CHECK-NEXT:  1      9     1.00                        sqincp	x0, p0.b
+# CHECK-NEXT:  1      9     1.00                        sqincp	x0, p0.d
+# CHECK-NEXT:  1      9     1.00                        sqincp	x0, p0.h
+# CHECK-NEXT:  1      9     1.00                        sqincp	x0, p0.s
+# CHECK-NEXT:  1      9     1.00                        sqincp	xzr, p15.b, wzr
+# CHECK-NEXT:  1      9     1.00                        sqincp	xzr, p15.d, wzr
+# CHECK-NEXT:  1      9     1.00                        sqincp	xzr, p15.h, wzr
+# CHECK-NEXT:  1      9     1.00                        sqincp	xzr, p15.s, wzr
 # CHECK-NEXT:  1      4     0.50                        sqincp	z0.d, p0.d
 # CHECK-NEXT:  1      4     0.50                        sqincp	z0.h, p0.h
 # CHECK-NEXT:  1      4     0.50                        sqincp	z0.s, p0.s
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, w0
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, w0, pow2
-# CHECK-NEXT:  1      1     0.33                        sqincw	x0, w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, #14
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, w0
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, w0, pow2
+# CHECK-NEXT:  1      4     0.33                        sqincw	x0, w0, pow2, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqincw	z0.s
 # CHECK-NEXT:  1      4     0.50                        sqincw	z0.s, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        sqincw	z0.s, pow2
@@ -6001,24 +6001,24 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        ssra	z31.d, z31.d, #64
 # CHECK-NEXT:  1      4     0.50                        ssra	z31.h, z31.h, #16
 # CHECK-NEXT:  1      4     0.50                        ssra	z31.s, z31.s, #32
-# CHECK-NEXT:  1      3     0.50                        ssublb	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        ssublb	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssublb	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        ssublbt	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        ssublbt	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      3     0.50                        ssublbt	z0.s, z1.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssublt	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        ssublt	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssublt	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        ssubltb	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        ssubltb	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      3     0.50                        ssubltb	z0.s, z1.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssubwb	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        ssubwb	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssubwb	z31.d, z31.d, z31.s
-# CHECK-NEXT:  1      3     0.50                        ssubwt	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        ssubwt	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        ssubwt	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssublb	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        ssublb	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssublb	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssublbt	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssublbt	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      4     0.50                        ssublbt	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssublt	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        ssublt	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssublt	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssubltb	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssubltb	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      4     0.50                        ssubltb	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssubwb	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        ssubwb	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssubwb	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        ssubwt	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        ssubwt	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        ssubwt	z31.d, z31.d, z31.s
 # CHECK-NEXT:  1      1     1.00           *            st1b	{ z0.b }, p0, [x0, x0]
 # CHECK-NEXT:  1      1     1.00           *            st1b	{ z0.b }, p0, [x0]
 # CHECK-NEXT:  1      1     1.00           *            st1b	{ z0.d }, p0, [x0, x0]
@@ -6250,12 +6250,12 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        sub	z31.s, p7/m, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        sub	z31.s, z31.s, #65280
 # CHECK-NEXT:  1      3     0.50                        sub	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        subhnb	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        subhnb	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        subhnb	z0.s, z1.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        subhnt	z0.b, z1.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        subhnt	z0.h, z1.s, z31.s
-# CHECK-NEXT:  1      4     0.50                        subhnt	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        subhnb	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        subhnb	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        subhnb	z0.s, z1.d, z31.d
+# CHECK-NEXT:  1      8     0.50                        subhnt	z0.b, z1.h, z31.h
+# CHECK-NEXT:  1      8     0.50                        subhnt	z0.h, z1.s, z31.s
+# CHECK-NEXT:  1      8     0.50                        subhnt	z0.s, z1.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        subr	z0.b, p0/m, z0.b, z0.b
 # CHECK-NEXT:  1      3     0.50                        subr	z0.b, z0.b, #0
 # CHECK-NEXT:  1      3     0.50                        subr	z0.d, p0/m, z0.d, z0.d
@@ -6305,32 +6305,32 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        tbx	z31.d, z31.d, z31.d
 # CHECK-NEXT:  1      4     0.50                        tbx	z31.h, z31.h, z31.h
 # CHECK-NEXT:  1      4     0.50                        tbx	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        trn1	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        trn1	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        trn1	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        trn1	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        trn1	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        trn1	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        trn1	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        trn1	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        trn2	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        trn2	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        trn2	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        trn2	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        trn2	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        trn2	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        trn2	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        trn2	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        uaba	z0.b, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        uaba	z0.d, z1.d, z31.d
-# CHECK-NEXT:  1      8     1.00                        uaba	z0.h, z1.h, z31.h
-# CHECK-NEXT:  1      8     1.00                        uaba	z0.s, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        uabalb	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        uabalb	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        uabalb	z0.s, z1.h, z31.h
-# CHECK-NEXT:  1      8     1.00                        uabalt	z0.d, z1.s, z31.s
-# CHECK-NEXT:  1      8     1.00                        uabalt	z0.h, z1.b, z31.b
-# CHECK-NEXT:  1      8     1.00                        uabalt	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      2     1.00                        trn1	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        trn1	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        trn1	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        trn1	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        trn1	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        trn1	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        trn1	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        trn1	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      2     1.00                        trn2	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        trn2	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        trn2	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        trn2	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        trn2	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        trn2	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        trn2	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        trn2	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        uaba	z0.b, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        uaba	z0.d, z1.d, z31.d
+# CHECK-NEXT:  1      6     1.00                        uaba	z0.h, z1.h, z31.h
+# CHECK-NEXT:  1      6     1.00                        uaba	z0.s, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        uabalb	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        uabalb	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        uabalb	z0.s, z1.h, z31.h
+# CHECK-NEXT:  1      6     1.00                        uabalt	z0.d, z1.s, z31.s
+# CHECK-NEXT:  1      6     1.00                        uabalt	z0.h, z1.b, z31.b
+# CHECK-NEXT:  1      6     1.00                        uabalt	z0.s, z1.h, z31.h
 # CHECK-NEXT:  1      3     0.50                        uabd	z31.b, p7/m, z31.b, z31.b
 # CHECK-NEXT:  1      3     0.50                        uabd	z31.d, p7/m, z31.d, z31.d
 # CHECK-NEXT:  1      3     0.50                        uabd	z31.h, p7/m, z31.h, z31.h
@@ -6344,22 +6344,22 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      7     1.00                        uadalp	z0.h, p0/m, z1.b
 # CHECK-NEXT:  1      7     1.00                        uadalp	z29.s, p0/m, z30.h
 # CHECK-NEXT:  1      7     1.00                        uadalp	z30.d, p7/m, z31.s
-# CHECK-NEXT:  1      3     0.50                        uaddlb	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        uaddlb	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        uaddlb	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        uaddlt	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        uaddlt	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        uaddlt	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        uaddlb	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        uaddlb	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        uaddlb	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        uaddlt	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        uaddlt	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        uaddlt	z31.d, z31.s, z31.s
 # CHECK-NEXT:  1      4     1.00                        uaddv	d0, p7, z31.b
 # CHECK-NEXT:  1      4     1.00                        uaddv	d0, p7, z31.d
 # CHECK-NEXT:  1      4     1.00                        uaddv	d0, p7, z31.h
 # CHECK-NEXT:  1      4     1.00                        uaddv	d0, p7, z31.s
-# CHECK-NEXT:  1      3     0.50                        uaddwb	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        uaddwb	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        uaddwb	z31.d, z31.d, z31.s
-# CHECK-NEXT:  1      3     0.50                        uaddwt	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        uaddwt	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        uaddwt	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        uaddwb	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        uaddwb	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        uaddwb	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        uaddwt	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        uaddwt	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        uaddwt	z31.d, z31.d, z31.s
 # CHECK-NEXT:  1      4     0.50                        ucvtf	z0.d, p0/m, z0.d
 # CHECK-NEXT:  1      4     0.50                        ucvtf	z0.d, p0/m, z0.s
 # CHECK-NEXT:  1      4     0.50                        ucvtf	z0.h, p0/m, z0.d
@@ -6473,120 +6473,120 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        uqadd	z31.d, z31.d, #65280
 # CHECK-NEXT:  1      4     0.50                        uqadd	z31.h, z31.h, #65280
 # CHECK-NEXT:  1      4     0.50                        uqadd	z31.s, z31.s, #65280
-# CHECK-NEXT:  1      1     0.33                        uqdecb	w0
-# CHECK-NEXT:  1      1     0.33                        uqdecb	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecb	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecb	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecb	x0
-# CHECK-NEXT:  1      1     0.33                        uqdecb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqdecb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        uqdecd	w0
-# CHECK-NEXT:  1      1     0.33                        uqdecd	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecd	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecd	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecd	x0
-# CHECK-NEXT:  1      1     0.33                        uqdecd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqdecd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecd	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqdecb	w0
+# CHECK-NEXT:  1      4     0.33                        uqdecb	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecb	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecb	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecb	x0
+# CHECK-NEXT:  1      4     0.33                        uqdecb	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqdecb	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecb	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecb	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqdecd	w0
+# CHECK-NEXT:  1      4     0.33                        uqdecd	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecd	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecd	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecd	x0
+# CHECK-NEXT:  1      4     0.33                        uqdecd	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqdecd	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecd	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecd	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqdecd	z0.d
 # CHECK-NEXT:  1      4     0.50                        uqdecd	z0.d, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqdecd	z0.d, pow2
 # CHECK-NEXT:  1      4     0.50                        uqdecd	z0.d, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdech	w0
-# CHECK-NEXT:  1      1     0.33                        uqdech	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdech	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdech	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdech	x0
-# CHECK-NEXT:  1      1     0.33                        uqdech	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqdech	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdech	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdech	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqdech	w0
+# CHECK-NEXT:  1      4     0.33                        uqdech	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdech	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdech	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdech	x0
+# CHECK-NEXT:  1      4     0.33                        uqdech	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqdech	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdech	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdech	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqdech	z0.h
 # CHECK-NEXT:  1      4     0.50                        uqdech	z0.h, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqdech	z0.h, pow2
 # CHECK-NEXT:  1      4     0.50                        uqdech	z0.h, pow2, mul #16
-# CHECK-NEXT:  1      8     1.00                        uqdecp	wzr, p15.b
-# CHECK-NEXT:  1      8     1.00                        uqdecp	wzr, p15.d
-# CHECK-NEXT:  1      8     1.00                        uqdecp	wzr, p15.h
-# CHECK-NEXT:  1      8     1.00                        uqdecp	wzr, p15.s
-# CHECK-NEXT:  1      8     1.00                        uqdecp	x0, p0.b
-# CHECK-NEXT:  1      8     1.00                        uqdecp	x0, p0.d
-# CHECK-NEXT:  1      8     1.00                        uqdecp	x0, p0.h
-# CHECK-NEXT:  1      8     1.00                        uqdecp	x0, p0.s
+# CHECK-NEXT:  1      9     1.00                        uqdecp	wzr, p15.b
+# CHECK-NEXT:  1      9     1.00                        uqdecp	wzr, p15.d
+# CHECK-NEXT:  1      9     1.00                        uqdecp	wzr, p15.h
+# CHECK-NEXT:  1      9     1.00                        uqdecp	wzr, p15.s
+# CHECK-NEXT:  1      9     1.00                        uqdecp	x0, p0.b
+# CHECK-NEXT:  1      9     1.00                        uqdecp	x0, p0.d
+# CHECK-NEXT:  1      9     1.00                        uqdecp	x0, p0.h
+# CHECK-NEXT:  1      9     1.00                        uqdecp	x0, p0.s
 # CHECK-NEXT:  1      4     0.50                        uqdecp	z0.d, p0.d
 # CHECK-NEXT:  1      4     0.50                        uqdecp	z0.h, p0.h
 # CHECK-NEXT:  1      4     0.50                        uqdecp	z0.s, p0.s
-# CHECK-NEXT:  1      1     0.33                        uqdecw	w0
-# CHECK-NEXT:  1      1     0.33                        uqdecw	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecw	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecw	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecw	x0
-# CHECK-NEXT:  1      1     0.33                        uqdecw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqdecw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqdecw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqdecw	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqdecw	w0
+# CHECK-NEXT:  1      4     0.33                        uqdecw	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecw	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecw	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecw	x0
+# CHECK-NEXT:  1      4     0.33                        uqdecw	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqdecw	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqdecw	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqdecw	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqdecw	z0.s
 # CHECK-NEXT:  1      4     0.50                        uqdecw	z0.s, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqdecw	z0.s, pow2
 # CHECK-NEXT:  1      4     0.50                        uqdecw	z0.s, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincb	w0
-# CHECK-NEXT:  1      1     0.33                        uqincb	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincb	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincb	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincb	x0
-# CHECK-NEXT:  1      1     0.33                        uqincb	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqincb	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincb	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincb	x0, vl1
-# CHECK-NEXT:  1      1     0.33                        uqincd	w0
-# CHECK-NEXT:  1      1     0.33                        uqincd	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincd	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincd	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincd	x0
-# CHECK-NEXT:  1      1     0.33                        uqincd	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqincd	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincd	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincd	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqincb	w0
+# CHECK-NEXT:  1      4     0.33                        uqincb	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincb	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincb	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincb	x0
+# CHECK-NEXT:  1      4     0.33                        uqincb	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqincb	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincb	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincb	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqincd	w0
+# CHECK-NEXT:  1      4     0.33                        uqincd	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincd	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincd	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincd	x0
+# CHECK-NEXT:  1      4     0.33                        uqincd	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqincd	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincd	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincd	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqincd	z0.d
 # CHECK-NEXT:  1      4     0.50                        uqincd	z0.d, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqincd	z0.d, pow2
 # CHECK-NEXT:  1      4     0.50                        uqincd	z0.d, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqinch	w0
-# CHECK-NEXT:  1      1     0.33                        uqinch	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqinch	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqinch	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqinch	x0
-# CHECK-NEXT:  1      1     0.33                        uqinch	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqinch	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqinch	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqinch	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqinch	w0
+# CHECK-NEXT:  1      4     0.33                        uqinch	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqinch	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqinch	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqinch	x0
+# CHECK-NEXT:  1      4     0.33                        uqinch	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqinch	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqinch	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqinch	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqinch	z0.h
 # CHECK-NEXT:  1      4     0.50                        uqinch	z0.h, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqinch	z0.h, pow2
 # CHECK-NEXT:  1      4     0.50                        uqinch	z0.h, pow2, mul #16
-# CHECK-NEXT:  1      8     1.00                        uqincp	wzr, p15.b
-# CHECK-NEXT:  1      8     1.00                        uqincp	wzr, p15.d
-# CHECK-NEXT:  1      8     1.00                        uqincp	wzr, p15.h
-# CHECK-NEXT:  1      8     1.00                        uqincp	wzr, p15.s
-# CHECK-NEXT:  1      8     1.00                        uqincp	x0, p0.b
-# CHECK-NEXT:  1      8     1.00                        uqincp	x0, p0.d
-# CHECK-NEXT:  1      8     1.00                        uqincp	x0, p0.h
-# CHECK-NEXT:  1      8     1.00                        uqincp	x0, p0.s
+# CHECK-NEXT:  1      9     1.00                        uqincp	wzr, p15.b
+# CHECK-NEXT:  1      9     1.00                        uqincp	wzr, p15.d
+# CHECK-NEXT:  1      9     1.00                        uqincp	wzr, p15.h
+# CHECK-NEXT:  1      9     1.00                        uqincp	wzr, p15.s
+# CHECK-NEXT:  1      9     1.00                        uqincp	x0, p0.b
+# CHECK-NEXT:  1      9     1.00                        uqincp	x0, p0.d
+# CHECK-NEXT:  1      9     1.00                        uqincp	x0, p0.h
+# CHECK-NEXT:  1      9     1.00                        uqincp	x0, p0.s
 # CHECK-NEXT:  1      4     0.50                        uqincp	z0.d, p0.d
 # CHECK-NEXT:  1      4     0.50                        uqincp	z0.h, p0.h
 # CHECK-NEXT:  1      4     0.50                        uqincp	z0.s, p0.s
-# CHECK-NEXT:  1      1     0.33                        uqincw	w0
-# CHECK-NEXT:  1      1     0.33                        uqincw	w0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincw	w0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincw	w0, pow2, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincw	x0
-# CHECK-NEXT:  1      1     0.33                        uqincw	x0, #14
-# CHECK-NEXT:  1      1     0.33                        uqincw	x0, all, mul #16
-# CHECK-NEXT:  1      1     0.33                        uqincw	x0, pow2
-# CHECK-NEXT:  1      1     0.33                        uqincw	x0, vl1
+# CHECK-NEXT:  1      4     0.33                        uqincw	w0
+# CHECK-NEXT:  1      4     0.33                        uqincw	w0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincw	w0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincw	w0, pow2, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincw	x0
+# CHECK-NEXT:  1      4     0.33                        uqincw	x0, #14
+# CHECK-NEXT:  1      4     0.33                        uqincw	x0, all, mul #16
+# CHECK-NEXT:  1      4     0.33                        uqincw	x0, pow2
+# CHECK-NEXT:  1      4     0.33                        uqincw	x0, vl1
 # CHECK-NEXT:  1      4     0.50                        uqincw	z0.s
 # CHECK-NEXT:  1      4     0.50                        uqincw	z0.s, all, mul #16
 # CHECK-NEXT:  1      4     0.50                        uqincw	z0.s, pow2
@@ -6723,18 +6723,18 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      4     0.50                        usra	z31.d, z31.d, #64
 # CHECK-NEXT:  1      4     0.50                        usra	z31.h, z31.h, #16
 # CHECK-NEXT:  1      4     0.50                        usra	z31.s, z31.s, #32
-# CHECK-NEXT:  1      3     0.50                        usublb	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        usublb	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        usublb	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        usublt	z0.h, z1.b, z2.b
-# CHECK-NEXT:  1      3     0.50                        usublt	z29.s, z30.h, z31.h
-# CHECK-NEXT:  1      3     0.50                        usublt	z31.d, z31.s, z31.s
-# CHECK-NEXT:  1      3     0.50                        usubwb	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        usubwb	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        usubwb	z31.d, z31.d, z31.s
-# CHECK-NEXT:  1      3     0.50                        usubwt	z0.h, z1.h, z2.b
-# CHECK-NEXT:  1      3     0.50                        usubwt	z29.s, z30.s, z31.h
-# CHECK-NEXT:  1      3     0.50                        usubwt	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        usublb	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        usublb	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        usublb	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        usublt	z0.h, z1.b, z2.b
+# CHECK-NEXT:  1      4     0.50                        usublt	z29.s, z30.h, z31.h
+# CHECK-NEXT:  1      4     0.50                        usublt	z31.d, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        usubwb	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        usubwb	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        usubwb	z31.d, z31.d, z31.s
+# CHECK-NEXT:  1      4     0.50                        usubwt	z0.h, z1.h, z2.b
+# CHECK-NEXT:  1      4     0.50                        usubwt	z29.s, z30.s, z31.h
+# CHECK-NEXT:  1      4     0.50                        usubwt	z31.d, z31.d, z31.s
 # CHECK-NEXT:  1      4     0.50                        uunpkhi	z31.d, z31.s
 # CHECK-NEXT:  1      4     0.50                        uunpkhi	z31.h, z31.b
 # CHECK-NEXT:  1      4     0.50                        uunpkhi	z31.s, z31.h
@@ -6753,82 +6753,82 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  1      3     0.50                        uxth	z31.s, p7/m, z31.s
 # CHECK-NEXT:  1      3     0.50                        uxtw	z0.d, p0/m, z0.d
 # CHECK-NEXT:  1      3     0.50                        uxtw	z31.d, p7/m, z31.d
-# CHECK-NEXT:  1      6     1.00                        uzp1	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        uzp1	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        uzp1	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        uzp1	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        uzp1	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        uzp1	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        uzp1	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        uzp1	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        uzp2	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        uzp2	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        uzp2	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        uzp2	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        uzp2	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        uzp2	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        uzp2	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        uzp2	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.b, w0, wzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.b, wzr, w0
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.b, x0, xzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.b, xzr, x0
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.d, w0, wzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.d, x0, xzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.h, w0, wzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.h, x0, xzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.s, w0, wzr
-# CHECK-NEXT:  1      6     1.00                        whilege	p15.s, x0, xzr
-# CHECK-NEXT:  1      6     1.00                        whilerw	p15.b, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilerw	p15.d, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilerw	p15.h, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilerw	p15.s, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilewr	p15.b, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilewr	p15.d, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilewr	p15.h, x30, x30
-# CHECK-NEXT:  1      6     1.00                        whilewr	p15.s, x30, x30
+# CHECK-NEXT:  1      2     1.00                        uzp1	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        uzp1	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        uzp1	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        uzp1	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        uzp1	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        uzp1	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        uzp1	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        uzp1	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      2     1.00                        uzp2	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        uzp2	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        uzp2	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        uzp2	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        uzp2	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        uzp2	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        uzp2	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        uzp2	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.b, w0, wzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.b, wzr, w0
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.b, x0, xzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.b, xzr, x0
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.d, w0, wzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.d, x0, xzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.h, w0, wzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.h, x0, xzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.s, w0, wzr
+# CHECK-NEXT:  1      2     1.00                        whilege	p15.s, x0, xzr
+# CHECK-NEXT:  1      2     1.00                        whilerw	p15.b, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilerw	p15.d, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilerw	p15.h, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilerw	p15.s, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilewr	p15.b, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilewr	p15.d, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilewr	p15.h, x30, x30
+# CHECK-NEXT:  1      2     1.00                        whilewr	p15.s, x30, x30
 # CHECK-NEXT:  1      1     0.33           *      U     wrffr	p0.b
 # CHECK-NEXT:  1      1     0.33           *      U     wrffr	p15.b
-# CHECK-NEXT:  1      3     0.50                        xar	z0.b, z0.b, z1.b, #1
-# CHECK-NEXT:  1      3     0.50                        xar	z0.d, z0.d, z1.d, #1
-# CHECK-NEXT:  1      3     0.50                        xar	z0.h, z0.h, z1.h, #1
-# CHECK-NEXT:  1      3     0.50                        xar	z0.s, z0.s, z1.s, #1
-# CHECK-NEXT:  1      3     0.50                        xar	z31.b, z31.b, z30.b, #8
-# CHECK-NEXT:  1      3     0.50                        xar	z31.d, z31.d, z30.d, #64
-# CHECK-NEXT:  1      3     0.50                        xar	z31.h, z31.h, z30.h, #16
-# CHECK-NEXT:  1      3     0.50                        xar	z31.s, z31.s, z30.s, #32
-# CHECK-NEXT:  1      6     1.00                        zip1	p0.b, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        zip1	p0.d, p0.d, p0.d
-# CHECK-NEXT:  1      6     1.00                        zip1	p0.h, p0.h, p0.h
-# CHECK-NEXT:  1      6     1.00                        zip1	p0.s, p0.s, p0.s
-# CHECK-NEXT:  1      6     1.00                        zip1	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        zip1	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        zip1	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        zip1	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        zip1	z0.b, z0.b, z0.b
-# CHECK-NEXT:  1      4     0.50                        zip1	z0.d, z0.d, z0.d
-# CHECK-NEXT:  1      4     0.50                        zip1	z0.h, z0.h, z0.h
-# CHECK-NEXT:  1      4     0.50                        zip1	z0.s, z0.s, z0.s
-# CHECK-NEXT:  1      4     0.50                        zip1	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        zip1	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        zip1	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        zip1	z31.s, z31.s, z31.s
-# CHECK-NEXT:  1      6     1.00                        zip2	p0.b, p0.b, p0.b
-# CHECK-NEXT:  1      6     1.00                        zip2	p0.d, p0.d, p0.d
-# CHECK-NEXT:  1      6     1.00                        zip2	p0.h, p0.h, p0.h
-# CHECK-NEXT:  1      6     1.00                        zip2	p0.s, p0.s, p0.s
-# CHECK-NEXT:  1      6     1.00                        zip2	p15.b, p15.b, p15.b
-# CHECK-NEXT:  1      6     1.00                        zip2	p15.d, p15.d, p15.d
-# CHECK-NEXT:  1      6     1.00                        zip2	p15.h, p15.h, p15.h
-# CHECK-NEXT:  1      6     1.00                        zip2	p15.s, p15.s, p15.s
-# CHECK-NEXT:  1      4     0.50                        zip2	z0.b, z0.b, z0.b
-# CHECK-NEXT:  1      4     0.50                        zip2	z0.d, z0.d, z0.d
-# CHECK-NEXT:  1      4     0.50                        zip2	z0.h, z0.h, z0.h
-# CHECK-NEXT:  1      4     0.50                        zip2	z0.s, z0.s, z0.s
-# CHECK-NEXT:  1      4     0.50                        zip2	z31.b, z31.b, z31.b
-# CHECK-NEXT:  1      4     0.50                        zip2	z31.d, z31.d, z31.d
-# CHECK-NEXT:  1      4     0.50                        zip2	z31.h, z31.h, z31.h
-# CHECK-NEXT:  1      4     0.50                        zip2	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      4     0.50                        xar	z0.b, z0.b, z1.b, #1
+# CHECK-NEXT:  1      4     0.50                        xar	z0.d, z0.d, z1.d, #1
+# CHECK-NEXT:  1      4     0.50                        xar	z0.h, z0.h, z1.h, #1
+# CHECK-NEXT:  1      4     0.50                        xar	z0.s, z0.s, z1.s, #1
+# CHECK-NEXT:  1      4     0.50                        xar	z31.b, z31.b, z30.b, #8
+# CHECK-NEXT:  1      4     0.50                        xar	z31.d, z31.d, z30.d, #64
+# CHECK-NEXT:  1      4     0.50                        xar	z31.h, z31.h, z30.h, #16
+# CHECK-NEXT:  1      4     0.50                        xar	z31.s, z31.s, z30.s, #32
+# CHECK-NEXT:  1      2     1.00                        zip1	p0.b, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        zip1	p0.d, p0.d, p0.d
+# CHECK-NEXT:  1      2     1.00                        zip1	p0.h, p0.h, p0.h
+# CHECK-NEXT:  1      2     1.00                        zip1	p0.s, p0.s, p0.s
+# CHECK-NEXT:  1      2     1.00                        zip1	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        zip1	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        zip1	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        zip1	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        zip1	z0.b, z0.b, z0.b
+# CHECK-NEXT:  1      3     0.50                        zip1	z0.d, z0.d, z0.d
+# CHECK-NEXT:  1      3     0.50                        zip1	z0.h, z0.h, z0.h
+# CHECK-NEXT:  1      3     0.50                        zip1	z0.s, z0.s, z0.s
+# CHECK-NEXT:  1      3     0.50                        zip1	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        zip1	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        zip1	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        zip1	z31.s, z31.s, z31.s
+# CHECK-NEXT:  1      2     1.00                        zip2	p0.b, p0.b, p0.b
+# CHECK-NEXT:  1      2     1.00                        zip2	p0.d, p0.d, p0.d
+# CHECK-NEXT:  1      2     1.00                        zip2	p0.h, p0.h, p0.h
+# CHECK-NEXT:  1      2     1.00                        zip2	p0.s, p0.s, p0.s
+# CHECK-NEXT:  1      2     1.00                        zip2	p15.b, p15.b, p15.b
+# CHECK-NEXT:  1      2     1.00                        zip2	p15.d, p15.d, p15.d
+# CHECK-NEXT:  1      2     1.00                        zip2	p15.h, p15.h, p15.h
+# CHECK-NEXT:  1      2     1.00                        zip2	p15.s, p15.s, p15.s
+# CHECK-NEXT:  1      3     0.50                        zip2	z0.b, z0.b, z0.b
+# CHECK-NEXT:  1      3     0.50                        zip2	z0.d, z0.d, z0.d
+# CHECK-NEXT:  1      3     0.50                        zip2	z0.h, z0.h, z0.h
+# CHECK-NEXT:  1      3     0.50                        zip2	z0.s, z0.s, z0.s
+# CHECK-NEXT:  1      3     0.50                        zip2	z31.b, z31.b, z31.b
+# CHECK-NEXT:  1      3     0.50                        zip2	z31.d, z31.d, z31.d
+# CHECK-NEXT:  1      3     0.50                        zip2	z31.h, z31.h, z31.h
+# CHECK-NEXT:  1      3     0.50                        zip2	z31.s, z31.s, z31.s
 
 # CHECK:      Resources:
 # CHECK-NEXT: [0]   - CortexA510UnitALU0
@@ -6848,7 +6848,7 @@ zip2	z31.s, z31.s, z31.s
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10.0] [10.1] [11]
-# CHECK-NEXT: 79.00  75.00  75.00  9.00    -     240.00 3698.00  -     -     1290.00 924.00 199.50 199.50 670.00
+# CHECK-NEXT: 79.00  75.00  75.00  9.00    -     209.00 3667.00  -     -     1290.00 924.00 199.50 199.50 670.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]    [7]    [8]    [9]    [10.0] [10.1] [11]   Instructions:
@@ -7844,8 +7844,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1b	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1b	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1b	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1b	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1b	{ z21.b }, p5/z, [x10, #5, mul vl]
@@ -7882,8 +7882,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1h	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1h	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -7899,8 +7899,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1h	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z31.h }, p7/z, [sp, #-1, mul vl]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1h	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z5.h }, p3/z, [sp, x16, lsl #1]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1h	{ z5.h }, p3/z, [x17, x16, lsl #1]
@@ -7961,7 +7961,7 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sb	{ z0.h }, p0/z, [sp, x0]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sb	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sb	{ z0.h }, p0/z, [x0]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sb	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1sb	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sb	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -7981,8 +7981,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sh	{ z0.d }, p0/z, [x0, z0.d, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sh	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sh	{ z0.d }, p0/z, [z0.d]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1sh	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sh	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -7997,8 +7997,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sh	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sh	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1sh	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sw	{ z0.d }, p0/z, [x0, z0.d, sxtw #2]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1sw	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
@@ -8017,8 +8017,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1w	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1w	{ z0.d }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1w	{ z0.d }, p0/z, [z0.d]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [x0]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1w	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1w	{ z21.d }, p5/z, [x10, #5, mul vl]
@@ -8033,8 +8033,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1w	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ld1w	{ z31.d }, p7/z, [z31.d, #124]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [sp, #-1, mul vl]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ld1w	{ z31.s }, p7/z, [z31.s, #124]
 # CHECK-NEXT:  -      -      -      -      -      -     2.00    -      -      -      -      -      -      -     ld2b	{ z0.b, z1.b }, p0/z, [x0, x0]
 # CHECK-NEXT:  -      -      -      -      -      -     1.00    -      -      -      -      -      -      -     ld2b	{ z0.b, z1.b }, p0/z, [x0]
@@ -8100,8 +8100,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1b	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1b	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [x0, x0]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1b	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1b	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1b	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -8128,8 +8128,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1h	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1h	{ z0.h }, p0/z, [x0, x0, lsl #1]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [x0, x0, lsl #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1h	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1h	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1h	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -8138,16 +8138,16 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1h	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1h	{ z31.d }, p7/z, [z31.d, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1h	{ z31.h }, p7/z, [sp]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1h	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sb	{ z0.d }, p0/z, [x0, x0]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sb	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sb	{ z0.h }, p0/z, [x0, x0]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [x0, x0]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1sb	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sb	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sb	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -8162,8 +8162,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z0.d }, p0/z, [x0, z0.d, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [x0, x0, lsl #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1sh	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -8171,8 +8171,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sh	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1sh	{ z31.d }, p7/z, [z31.d, #62]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, sxtw #1]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [sp, z31.s, uxtw #1]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1sh	{ z31.s }, p7/z, [z31.s, #62]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1sw	{ z0.d }, p0/z, [x0, x0, lsl #2]
@@ -8190,8 +8190,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z0.d }, p0/z, [x0, z0.d, uxtw #2]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z0.d }, p0/z, [z0.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [x0, x0, lsl #2]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [x0, z0.s, sxtw]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [x0, z0.s, uxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1w	{ z0.s }, p0/z, [z0.s]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z21.d }, p5/z, [x10, z21.d, sxtw]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z21.d }, p5/z, [x10, z21.d, uxtw]
@@ -8199,8 +8199,8 @@ zip2	z31.s, z31.s, z31.s
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z31.d }, p7/z, [sp, z31.d]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1w	{ z31.d }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     7.00    -      -      -      -      -      -      -     ldff1w	{ z31.d }, p7/z, [z31.d, #124]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
-# CHECK-NEXT:  -      -      -      -      -     4.50   4.50    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [sp, z31.s, sxtw #2]
+# CHECK-NEXT:  -      -      -      -      -     3.50   3.50    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [sp, z31.s, uxtw #2]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [sp]
 # CHECK-NEXT:  -      -      -      -      -      -     9.00    -      -      -      -      -      -      -     ldff1w	{ z31.s }, p7/z, [z31.s, #124]
 # CHECK-NEXT:  -      -      -      -      -     0.50   0.50    -      -      -      -      -      -      -     ldnf1b	{ z0.b }, p0/z, [x0]

From 4572a2db2c2c48820e79fb65f3810348371db4f1 Mon Sep 17 00:00:00 2001
From: David Green <david.green@arm.com>
Date: Wed, 17 Apr 2024 19:44:17 +0100
Subject: [PATCH 781/886] [AArch64] Add some test cases for LD2/LD3/LD4
 shuffles. NFC

---
 .../CostModel/AArch64/shuffle-load.ll         | 620 ++++++++++++++++++
 1 file changed, 620 insertions(+)

diff --git a/llvm/test/Analysis/CostModel/AArch64/shuffle-load.ll b/llvm/test/Analysis/CostModel/AArch64/shuffle-load.ll
index ec95a313e11e2..106f2f9edc2e4 100644
--- a/llvm/test/Analysis/CostModel/AArch64/shuffle-load.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/shuffle-load.ll
@@ -401,3 +401,623 @@ entry:
   %lane = shufflevector <2 x i64> %tmp1, <2 x i64> undef, <2 x i32> zeroinitializer
   ret <2 x i64> %lane
 }
+
+define void @vld2(ptr %p) {
+; CHECK-LABEL: 'vld2'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8 = load <4 x i8>, ptr %p, align 4
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_0 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_1 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = load <8 x i8>, ptr %p, align 8
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8_0 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8_1 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = load <16 x i8>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8_0 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8_1 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = load <32 x i8>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8_0 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8_1 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = load <4 x i16>, ptr %p, align 8
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16_0 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16_1 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = load <8 x i16>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16_0 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16_1 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = load <16 x i16>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16_0 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16_1 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i16 = load <32 x i16>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i16_0 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i16_1 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = load <4 x i32>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32_0 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32_1 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = load <8 x i32>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32_0 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32_1 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = load <16 x i32>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i32_0 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i32_1 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v32i32 = load <32 x i32>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i32_0 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i32_1 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i64 = load <4 x i64>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i64 = load <8 x i64>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i64 = load <16 x i64>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i64 = load <32 x i64>, ptr %p, align 256
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; CODESIZE-LABEL: 'vld2'
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = load <4 x i8>, ptr %p, align 4
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_0 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 0, i32 2>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8_1 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 1, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8 = load <8 x i8>, ptr %p, align 8
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8_0 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i8_1 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8 = load <16 x i8>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8_0 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i8_1 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i8 = load <32 x i8>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8_0 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v32i8_1 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16 = load <4 x i16>, ptr %p, align 8
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16_0 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 0, i32 2>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i16_1 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 1, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16 = load <8 x i16>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16_0 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i16_1 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i16 = load <16 x i16>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16_0 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16i16_1 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i16 = load <32 x i16>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i16_0 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v32i16_1 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = load <4 x i32>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32_0 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 0, i32 2>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i32_1 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 1, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i32 = load <8 x i32>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32_0 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8i32_1 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i32 = load <16 x i32>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i32_0 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i32_1 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v32i32 = load <32 x i32>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i32_0 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i32_1 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i64 = load <4 x i64>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 0, i32 2>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 1, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i64 = load <8 x i64>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i64 = load <16 x i64>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i64 = load <32 x i64>, ptr %p, align 256
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v4i8 = load <4 x i8>, ptr %p
+  %v4i8_0 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 0, i32 2>
+  %v4i8_1 = shufflevector <4 x i8> %v4i8, <4 x i8> undef, <2 x i32> <i32 1, i32 3>
+  %v8i8 = load <8 x i8>, ptr %p
+  %v8i8_0 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %v8i8_1 = shufflevector <8 x i8> %v8i8, <8 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %v16i8 = load <16 x i8>, ptr %p
+  %v16i8_0 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %v16i8_1 = shufflevector <16 x i8> %v16i8, <16 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %v32i8 = load <32 x i8>, ptr %p
+  %v32i8_0 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %v32i8_1 = shufflevector <32 x i8> %v32i8, <32 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+
+  %v4i16 = load <4 x i16>, ptr %p
+  %v4i16_0 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 0, i32 2>
+  %v4i16_1 = shufflevector <4 x i16> %v4i16, <4 x i16> undef, <2 x i32> <i32 1, i32 3>
+  %v8i16 = load <8 x i16>, ptr %p
+  %v8i16_0 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %v8i16_1 = shufflevector <8 x i16> %v8i16, <8 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %v16i16 = load <16 x i16>, ptr %p
+  %v16i16_0 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %v16i16_1 = shufflevector <16 x i16> %v16i16, <16 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %v32i16 = load <32 x i16>, ptr %p
+  %v32i16_0 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %v32i16_1 = shufflevector <32 x i16> %v32i16, <32 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+
+  %v4i32 = load <4 x i32>, ptr %p
+  %v4i32_0 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 0, i32 2>
+  %v4i32_1 = shufflevector <4 x i32> %v4i32, <4 x i32> undef, <2 x i32> <i32 1, i32 3>
+  %v8i32 = load <8 x i32>, ptr %p
+  %v8i32_0 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %v8i32_1 = shufflevector <8 x i32> %v8i32, <8 x i32> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %v16i32 = load <16 x i32>, ptr %p
+  %v16i32_0 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %v16i32_1 = shufflevector <16 x i32> %v16i32, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %v32i32 = load <32 x i32>, ptr %p
+  %v32i32_0 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %v32i32_1 = shufflevector <32 x i32> %v32i32, <32 x i32> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+
+  %v2i64 = load <4 x i64>, ptr %p
+  %v2i64_0 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 0, i32 2>
+  %v2i64_1 = shufflevector <4 x i64> %v2i64, <4 x i64> undef, <2 x i32> <i32 1, i32 3>
+  %v4i64 = load <8 x i64>, ptr %p
+  %v4i64_0 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %v4i64_1 = shufflevector <8 x i64> %v4i64, <8 x i64> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %v8i64 = load <16 x i64>, ptr %p
+  %v8i64_0 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %v8i64_1 = shufflevector <16 x i64> %v8i64, <16 x i64> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %v16i64 = load <32 x i64>, ptr %p
+  %v16i64_0 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %v16i64_1 = shufflevector <32 x i64> %v16i64, <32 x i64> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+
+  ret void
+}
+
+
+define void @vld3(ptr %p) {
+; CHECK-LABEL: 'vld3'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8 = load <6 x i8>, ptr %p, align 8
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_0 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_1 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 1, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_2 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 2, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = load <12 x i8>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_0 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_1 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_2 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = load <24 x i8>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_0 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_1 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_2 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8 = load <48 x i8>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_0 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8_1 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_2 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = load <6 x i16>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_0 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_1 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 1, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_2 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 2, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = load <12 x i16>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_0 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_1 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_2 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i16 = load <24 x i16>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_0 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_1 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_2 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i16 = load <48 x i16>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_0 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i16_1 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_2 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32 = load <6 x i32>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_0 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32_1 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 1, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_2 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 2, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32 = load <12 x i32>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_0 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32_1 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_2 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32 = load <24 x i32>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_0 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32_1 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_2 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32 = load <48 x i32>, ptr %p, align 256
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_0 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32_1 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_2 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i64 = load <6 x i64>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 1, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_2 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 2, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i64 = load <12 x i64>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_2 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i64 = load <24 x i64>, ptr %p, align 256
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_2 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i64 = load <48 x i64>, ptr %p, align 512
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_2 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; CODESIZE-LABEL: 'vld3'
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8 = load <6 x i8>, ptr %p, align 8
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_0 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 0, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_1 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 1, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_2 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 2, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = load <12 x i8>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_0 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_1 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_2 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = load <24 x i8>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_0 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_1 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_2 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8 = load <48 x i8>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_0 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i8_1 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_2 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = load <6 x i16>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_0 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 0, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_1 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 1, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_2 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 2, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = load <12 x i16>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_0 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_1 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_2 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i16 = load <24 x i16>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_0 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i16_1 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_2 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i16 = load <48 x i16>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_0 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i16_1 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_2 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32 = load <6 x i32>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_0 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 0, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32_1 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 1, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_2 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 2, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32 = load <12 x i32>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_0 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32_1 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_2 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32 = load <24 x i32>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_0 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32_1 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_2 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32 = load <48 x i32>, ptr %p, align 256
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_0 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32_1 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_2 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i64 = load <6 x i64>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 0, i32 3>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 1, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_2 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 2, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i64 = load <12 x i64>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_2 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i64 = load <24 x i64>, ptr %p, align 256
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_2 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i64 = load <48 x i64>, ptr %p, align 512
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_2 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v2i8 = load <6 x i8>, ptr %p
+  %v2i8_0 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 0, i32 3>
+  %v2i8_1 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 1, i32 4>
+  %v2i8_2 = shufflevector <6 x i8> %v2i8, <6 x i8> undef, <2 x i32> <i32 2, i32 5>
+  %v4i8 = load <12 x i8>, ptr %p
+  %v4i8_0 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+  %v4i8_1 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+  %v4i8_2 = shufflevector <12 x i8> %v4i8, <12 x i8> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+  %v8i8 = load <24 x i8>, ptr %p
+  %v8i8_0 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+  %v8i8_1 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+  %v8i8_2 = shufflevector <24 x i8> %v8i8, <24 x i8> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+  %v16i8 = load <48 x i8>, ptr %p
+  %v16i8_0 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+  %v16i8_1 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+  %v16i8_2 = shufflevector <48 x i8> %v16i8, <48 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+
+  %v2i16 = load <6 x i16>, ptr %p
+  %v2i16_0 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 0, i32 3>
+  %v2i16_1 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 1, i32 4>
+  %v2i16_2 = shufflevector <6 x i16> %v2i16, <6 x i16> undef, <2 x i32> <i32 2, i32 5>
+  %v4i16 = load <12 x i16>, ptr %p
+  %v4i16_0 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+  %v4i16_1 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+  %v4i16_2 = shufflevector <12 x i16> %v4i16, <12 x i16> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+  %v8i16 = load <24 x i16>, ptr %p
+  %v8i16_0 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+  %v8i16_1 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+  %v8i16_2 = shufflevector <24 x i16> %v8i16, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+  %v16i16 = load <48 x i16>, ptr %p
+  %v16i16_0 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+  %v16i16_1 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+  %v16i16_2 = shufflevector <48 x i16> %v16i16, <48 x i16> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+
+  %v2i32 = load <6 x i32>, ptr %p
+  %v2i32_0 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 0, i32 3>
+  %v2i32_1 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 1, i32 4>
+  %v2i32_2 = shufflevector <6 x i32> %v2i32, <6 x i32> undef, <2 x i32> <i32 2, i32 5>
+  %v4i32 = load <12 x i32>, ptr %p
+  %v4i32_0 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+  %v4i32_1 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+  %v4i32_2 = shufflevector <12 x i32> %v4i32, <12 x i32> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+  %v8i32 = load <24 x i32>, ptr %p
+  %v8i32_0 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+  %v8i32_1 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+  %v8i32_2 = shufflevector <24 x i32> %v8i32, <24 x i32> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+  %v16i32 = load <48 x i32>, ptr %p
+  %v16i32_0 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+  %v16i32_1 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+  %v16i32_2 = shufflevector <48 x i32> %v16i32, <48 x i32> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+
+  %v2i64 = load <6 x i64>, ptr %p
+  %v2i64_0 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 0, i32 3>
+  %v2i64_1 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 1, i32 4>
+  %v2i64_2 = shufflevector <6 x i64> %v2i64, <6 x i64> undef, <2 x i32> <i32 2, i32 5>
+  %v4i64 = load <12 x i64>, ptr %p
+  %v4i64_0 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+  %v4i64_1 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+  %v4i64_2 = shufflevector <12 x i64> %v4i64, <12 x i64> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+  %v8i64 = load <24 x i64>, ptr %p
+  %v8i64_0 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
+  %v8i64_1 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
+  %v8i64_2 = shufflevector <24 x i64> %v8i64, <24 x i64> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+  %v16i64 = load <48 x i64>, ptr %p
+  %v16i64_0 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
+  %v16i64_1 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
+  %v16i64_2 = shufflevector <48 x i64> %v16i64, <48 x i64> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
+
+  ret void
+}
+
+define void @vld4(ptr %p) {
+; CHECK-LABEL: 'vld4'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8 = load <8 x i8>, ptr %p, align 8
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_0 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_1 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_2 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_3 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = load <16 x i8>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_0 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_1 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_2 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_3 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = load <32 x i8>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_0 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_1 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_2 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_3 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8 = load <64 x i8>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_0 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_1 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_2 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_3 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = load <8 x i16>, ptr %p, align 16
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_0 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_1 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_2 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_3 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = load <16 x i16>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_0 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_1 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_2 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_3 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i16 = load <32 x i16>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_0 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_1 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_2 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_3 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i16 = load <64 x i16>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_0 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_1 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_2 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_3 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32 = load <8 x i32>, ptr %p, align 32
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_0 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_1 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_2 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32_3 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32 = load <16 x i32>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_0 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_1 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_2 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_3 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32 = load <32 x i32>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_0 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_1 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_2 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_3 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32 = load <64 x i32>, ptr %p, align 256
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_0 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_1 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_2 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_3 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i64 = load <8 x i64>, ptr %p, align 64
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 0, i32 4>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_2 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 2, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_3 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i64 = load <16 x i64>, ptr %p, align 128
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_2 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_3 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i64 = load <32 x i64>, ptr %p, align 256
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_2 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_3 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i64 = load <64 x i64>, ptr %p, align 512
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_2 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_3 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; CODESIZE-LABEL: 'vld4'
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i8 = load <8 x i8>, ptr %p, align 8
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_0 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 0, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_1 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 1, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_2 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 2, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i8_3 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 3, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4i8 = load <16 x i8>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_0 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_1 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_2 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i8_3 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = load <32 x i8>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_0 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_1 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_2 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %v8i8_3 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i8 = load <64 x i8>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_0 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_1 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_2 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %v16i8_3 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i16 = load <8 x i16>, ptr %p, align 16
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_0 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 0, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_1 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 1, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_2 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 2, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2i16_3 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 3, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = load <16 x i16>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_0 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_1 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_2 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %v4i16_3 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i16 = load <32 x i16>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_0 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_1 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_2 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v8i16_3 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i16 = load <64 x i16>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_0 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_1 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_2 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %v16i16_3 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32 = load <8 x i32>, ptr %p, align 32
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_0 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 0, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_1 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 1, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i32_2 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 2, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2i32_3 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 3, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4i32 = load <16 x i32>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_0 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_1 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_2 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4i32_3 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8i32 = load <32 x i32>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_0 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_1 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_2 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v8i32_3 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16i32 = load <64 x i32>, ptr %p, align 256
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_0 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_1 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_2 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v16i32_3 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2i64 = load <8 x i64>, ptr %p, align 64
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_0 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 0, i32 4>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_1 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 1, i32 5>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_2 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 2, i32 6>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2i64_3 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 3, i32 7>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4i64 = load <16 x i64>, ptr %p, align 128
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_0 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_1 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_2 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i64_3 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8i64 = load <32 x i64>, ptr %p, align 256
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_0 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_1 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_2 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i64_3 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16i64 = load <64 x i64>, ptr %p, align 512
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_0 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_1 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_2 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i64_3 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+; CODESIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v2i8 = load <8 x i8>, ptr %p
+  %v2i8_0 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 0, i32 4>
+  %v2i8_1 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 1, i32 5>
+  %v2i8_2 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 2, i32 6>
+  %v2i8_3 = shufflevector <8 x i8> %v2i8, <8 x i8> undef, <2 x i32> <i32 3, i32 7>
+  %v4i8 = load <16 x i8>, ptr %p
+  %v4i8_0 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+  %v4i8_1 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+  %v4i8_2 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+  %v4i8_3 = shufflevector <16 x i8> %v4i8, <16 x i8> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+  %v8i8 = load <32 x i8>, ptr %p
+  %v8i8_0 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+  %v8i8_1 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+  %v8i8_2 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+  %v8i8_3 = shufflevector <32 x i8> %v8i8, <32 x i8> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+  %v16i8 = load <64 x i8>, ptr %p
+  %v16i8_0 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+  %v16i8_1 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+  %v16i8_2 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+  %v16i8_3 = shufflevector <64 x i8> %v16i8, <64 x i8> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+
+  %v2i16 = load <8 x i16>, ptr %p
+  %v2i16_0 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 0, i32 4>
+  %v2i16_1 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 1, i32 5>
+  %v2i16_2 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 2, i32 6>
+  %v2i16_3 = shufflevector <8 x i16> %v2i16, <8 x i16> undef, <2 x i32> <i32 3, i32 7>
+  %v4i16 = load <16 x i16>, ptr %p
+  %v4i16_0 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+  %v4i16_1 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+  %v4i16_2 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+  %v4i16_3 = shufflevector <16 x i16> %v4i16, <16 x i16> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+  %v8i16 = load <32 x i16>, ptr %p
+  %v8i16_0 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+  %v8i16_1 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+  %v8i16_2 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+  %v8i16_3 = shufflevector <32 x i16> %v8i16, <32 x i16> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+  %v16i16 = load <64 x i16>, ptr %p
+  %v16i16_0 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+  %v16i16_1 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+  %v16i16_2 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+  %v16i16_3 = shufflevector <64 x i16> %v16i16, <64 x i16> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+
+  %v2i32 = load <8 x i32>, ptr %p
+  %v2i32_0 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 0, i32 4>
+  %v2i32_1 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 1, i32 5>
+  %v2i32_2 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 2, i32 6>
+  %v2i32_3 = shufflevector <8 x i32> %v2i32, <8 x i32> undef, <2 x i32> <i32 3, i32 7>
+  %v4i32 = load <16 x i32>, ptr %p
+  %v4i32_0 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+  %v4i32_1 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+  %v4i32_2 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+  %v4i32_3 = shufflevector <16 x i32> %v4i32, <16 x i32> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+  %v8i32 = load <32 x i32>, ptr %p
+  %v8i32_0 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+  %v8i32_1 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+  %v8i32_2 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+  %v8i32_3 = shufflevector <32 x i32> %v8i32, <32 x i32> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+  %v16i32 = load <64 x i32>, ptr %p
+  %v16i32_0 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+  %v16i32_1 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+  %v16i32_2 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+  %v16i32_3 = shufflevector <64 x i32> %v16i32, <64 x i32> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+
+  %v2i64 = load <8 x i64>, ptr %p
+  %v2i64_0 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 0, i32 4>
+  %v2i64_1 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 1, i32 5>
+  %v2i64_2 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 2, i32 6>
+  %v2i64_3 = shufflevector <8 x i64> %v2i64, <8 x i64> undef, <2 x i32> <i32 3, i32 7>
+  %v4i64 = load <16 x i64>, ptr %p
+  %v4i64_0 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
+  %v4i64_1 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
+  %v4i64_2 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
+  %v4i64_3 = shufflevector <16 x i64> %v4i64, <16 x i64> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
+  %v8i64 = load <32 x i64>, ptr %p
+  %v8i64_0 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
+  %v8i64_1 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
+  %v8i64_2 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
+  %v8i64_3 = shufflevector <32 x i64> %v8i64, <32 x i64> undef, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
+  %v16i64 = load <64 x i64>, ptr %p
+  %v16i64_0 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+  %v16i64_1 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+  %v16i64_2 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+  %v16i64_3 = shufflevector <64 x i64> %v16i64, <64 x i64> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+
+  ret void
+}
+

From 5a0942cd7423069e78fdfb9743a13aedfa7bdee0 Mon Sep 17 00:00:00 2001
From: Youngsuk Kim <joseph942010@gmail.com>
Date: Wed, 17 Apr 2024 14:45:33 -0400
Subject: [PATCH 782/886] [llvm][NVPTX] Don't emit unused var 'temp_param_reg'
 (NFC) (#89004)

Don't emit unused variable 'temp_param_reg' which has been around since
ae556d3ef72dfe5f40a337b7071f42b7bf5b66a4 .
---
 llvm/lib/Target/NVPTX/NVPTXInstrInfo.td                        | 3 +--
 llvm/test/CodeGen/NVPTX/dynamic_stackalloc.ll                  | 2 --
 llvm/test/CodeGen/NVPTX/i8x4-instructions.ll                   | 3 ---
 llvm/test/DebugInfo/NVPTX/dbg-declare-alloca.ll                | 1 -
 .../update_llc_test_checks/Inputs/nvptx-basic.ll.expected      | 1 -
 5 files changed, 1 insertion(+), 9 deletions(-)

diff --git a/llvm/lib/Target/NVPTX/NVPTXInstrInfo.td b/llvm/lib/Target/NVPTX/NVPTXInstrInfo.td
index 4daf47cfd4df8..cd8546005c028 100644
--- a/llvm/lib/Target/NVPTX/NVPTXInstrInfo.td
+++ b/llvm/lib/Target/NVPTX/NVPTXInstrInfo.td
@@ -3785,8 +3785,7 @@ class Pseudo<dag outs, dag ins, string asmstr, list<dag> pattern>
 
 def Callseq_Start :
   NVPTXInst<(outs), (ins i32imm:$amt1, i32imm:$amt2),
-            "\\{ // callseq $amt1, $amt2\n"
-            "\t.reg .b32 temp_param_reg;",
+            "\\{ // callseq $amt1, $amt2",
             [(callseq_start timm:$amt1, timm:$amt2)]>;
 def Callseq_End :
   NVPTXInst<(outs), (ins i32imm:$amt1, i32imm:$amt2),
diff --git a/llvm/test/CodeGen/NVPTX/dynamic_stackalloc.ll b/llvm/test/CodeGen/NVPTX/dynamic_stackalloc.ll
index 09297fb819ce5..ce81957f2a393 100644
--- a/llvm/test/CodeGen/NVPTX/dynamic_stackalloc.ll
+++ b/llvm/test/CodeGen/NVPTX/dynamic_stackalloc.ll
@@ -17,7 +17,6 @@
 ; CHECK-32-NEXT:  alloca.u32  %r[[ALLOCA:[0-9]]], %r[[SIZE3]], 16;
 ; CHECK-32-NEXT:  cvta.local.u32  %r[[ALLOCA]], %r[[ALLOCA]];
 ; CHECK-32-NEXT:  { // callseq 0, 0
-; CHECK-32-NEXT:  .reg .b32 temp_param_reg;
 ; CHECK-32-NEXT:  .param .b32 param0;
 ; CHECK-32-NEXT:  st.param.b32  [param0+0], %r[[ALLOCA]];
 
@@ -27,7 +26,6 @@
 ; CHECK-64-NEXT:  alloca.u64  %rd[[ALLOCA:[0-9]]], %rd[[SIZE3]], 16;
 ; CHECK-64-NEXT:  cvta.local.u64  %rd[[ALLOCA]], %rd[[ALLOCA]];
 ; CHECK-64-NEXT:  { // callseq 0, 0
-; CHECK-64-NEXT:  .reg .b32 temp_param_reg;
 ; CHECK-64-NEXT:  .param .b64 param0;
 ; CHECK-64-NEXT:  st.param.b64  [param0+0], %rd[[ALLOCA]];
 
diff --git a/llvm/test/CodeGen/NVPTX/i8x4-instructions.ll b/llvm/test/CodeGen/NVPTX/i8x4-instructions.ll
index 6895699a1dfea..96a4359d0ec43 100644
--- a/llvm/test/CodeGen/NVPTX/i8x4-instructions.ll
+++ b/llvm/test/CodeGen/NVPTX/i8x4-instructions.ll
@@ -827,7 +827,6 @@ define <4 x i8> @test_call(<4 x i8> %a, <4 x i8> %b) #0 {
 ; CHECK-NEXT:    ld.param.u32 %r2, [test_call_param_1];
 ; CHECK-NEXT:    ld.param.u32 %r1, [test_call_param_0];
 ; CHECK-NEXT:    { // callseq 0, 0
-; CHECK-NEXT:    .reg .b32 temp_param_reg;
 ; CHECK-NEXT:    .param .align 4 .b8 param0[4];
 ; CHECK-NEXT:    st.param.b32 [param0+0], %r1;
 ; CHECK-NEXT:    .param .align 4 .b8 param1[4];
@@ -856,7 +855,6 @@ define <4 x i8> @test_call_flipped(<4 x i8> %a, <4 x i8> %b) #0 {
 ; CHECK-NEXT:    ld.param.u32 %r2, [test_call_flipped_param_1];
 ; CHECK-NEXT:    ld.param.u32 %r1, [test_call_flipped_param_0];
 ; CHECK-NEXT:    { // callseq 1, 0
-; CHECK-NEXT:    .reg .b32 temp_param_reg;
 ; CHECK-NEXT:    .param .align 4 .b8 param0[4];
 ; CHECK-NEXT:    st.param.b32 [param0+0], %r2;
 ; CHECK-NEXT:    .param .align 4 .b8 param1[4];
@@ -885,7 +883,6 @@ define <4 x i8> @test_tailcall_flipped(<4 x i8> %a, <4 x i8> %b) #0 {
 ; CHECK-NEXT:    ld.param.u32 %r2, [test_tailcall_flipped_param_1];
 ; CHECK-NEXT:    ld.param.u32 %r1, [test_tailcall_flipped_param_0];
 ; CHECK-NEXT:    { // callseq 2, 0
-; CHECK-NEXT:    .reg .b32 temp_param_reg;
 ; CHECK-NEXT:    .param .align 4 .b8 param0[4];
 ; CHECK-NEXT:    st.param.b32 [param0+0], %r2;
 ; CHECK-NEXT:    .param .align 4 .b8 param1[4];
diff --git a/llvm/test/DebugInfo/NVPTX/dbg-declare-alloca.ll b/llvm/test/DebugInfo/NVPTX/dbg-declare-alloca.ll
index 8d7608ead38ed..de367dfa4acb4 100644
--- a/llvm/test/DebugInfo/NVPTX/dbg-declare-alloca.ll
+++ b/llvm/test/DebugInfo/NVPTX/dbg-declare-alloca.ll
@@ -9,7 +9,6 @@
 ; CHECK: add.u64 %rd1, %SP, 0;
 ; CHECK: .loc 1 5 3                   // t.c:5:3
 ; CHECK: { // callseq 0, 0
-; CHECK: .reg .b32 temp_param_reg;
 ; CHECK: .param .b64 param0;
 ; CHECK: st.param.b64 [param0+0], %rd1;
 ; CHECK: call.uni
diff --git a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/nvptx-basic.ll.expected b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/nvptx-basic.ll.expected
index a65f140b46015..3ac63d070933d 100644
--- a/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/nvptx-basic.ll.expected
+++ b/llvm/test/tools/UpdateTestChecks/update_llc_test_checks/Inputs/nvptx-basic.ll.expected
@@ -30,7 +30,6 @@ define dso_local void @caller_St8x4(ptr nocapture noundef readonly byval(%struct
 ; CHECK-NEXT:    ld.u64 %rd7, [%SP+24];
 ; CHECK-NEXT:    ld.u64 %rd8, [%SP+16];
 ; CHECK-NEXT:    { // callseq 0, 0
-; CHECK-NEXT:    .reg .b32 temp_param_reg;
 ; CHECK-NEXT:    .param .align 16 .b8 param0[32];
 ; CHECK-NEXT:    st.param.v2.b64 [param0+0], {%rd6, %rd5};
 ; CHECK-NEXT:    st.param.v2.b64 [param0+16], {%rd8, %rd7};

From 800f1050e190430f217e1fd0db9414dacc835e11 Mon Sep 17 00:00:00 2001
From: Shilei Tian <i@tianshilei.me>
Date: Wed, 17 Apr 2024 14:53:01 -0400
Subject: [PATCH 783/886] [GitHub] Add a new mapping for `offload` subproject
 (#89118)

Fix #89071.
---
 .github/new-prs-labeler.yml | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/.github/new-prs-labeler.yml b/.github/new-prs-labeler.yml
index 1502d64a7d3e3..9cf64417d3cb2 100644
--- a/.github/new-prs-labeler.yml
+++ b/.github/new-prs-labeler.yml
@@ -944,3 +944,6 @@ openmp:libomptarget:
 
 bazel:
   - utils/bazel/**
+
+offload:
+  - offload/**

From 6f7976c883da592f2cf6bfadef152e4203c00445 Mon Sep 17 00:00:00 2001
From: Mark de Wever <koraq@xs4all.nl>
Date: Wed, 17 Apr 2024 20:55:51 +0200
Subject: [PATCH 784/886] [libc++][TZDB] Adds sys_info formatter. (#85896)

Implements parts of:
- P0355 Extending <chrono> to Calendars and Time Zones
- P1361 Integration of chrono with text formatting
---
 libcxx/docs/Status/FormatPaper.csv            |   2 +-
 libcxx/include/__chrono/convert_to_tm.h       |   5 +
 libcxx/include/__chrono/formatter.h           |  45 +++++-
 libcxx/include/__chrono/ostream.h             |  20 +++
 libcxx/include/__chrono/sys_info.h            |   2 +-
 libcxx/include/chrono                         |   5 +
 .../time.zone.info.sys/ostream.pass.cpp       |  74 ++++++++++
 .../time/time.syn/formatter.sys_info.pass.cpp | 137 ++++++++++++++++++
 .../time.zone.info.sys/ostream.pass.cpp       |  52 +++++++
 .../concept.formattable.compile.pass.cpp      |   4 +-
 libcxx/test/support/test_macros.h             |   8 +
 11 files changed, 347 insertions(+), 7 deletions(-)
 create mode 100644 libcxx/test/libcxx/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp
 create mode 100644 libcxx/test/std/time/time.syn/formatter.sys_info.pass.cpp
 create mode 100644 libcxx/test/std/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp

diff --git a/libcxx/docs/Status/FormatPaper.csv b/libcxx/docs/Status/FormatPaper.csv
index e9d407e79e253..8ace18815f537 100644
--- a/libcxx/docs/Status/FormatPaper.csv
+++ b/libcxx/docs/Status/FormatPaper.csv
@@ -24,7 +24,7 @@ Section,Description,Dependencies,Assignee,Status,First released version
 `[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::year_month_weekday``",,Mark de Wever,|Complete|,16.0
 `[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::year_month_weekday_last``",,Mark de Wever,|Complete|,16.0
 `[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::hh_mm_ss<duration<Rep, Period>>``",,Mark de Wever,|Complete|,17.0
-`[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::sys_info``",A ``<chrono>`` implementation,Mark de Wever,,
+`[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::sys_info``",,Mark de Wever,|Complete|,19.0
 `[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::local_info``",A ``<chrono>`` implementation,Mark de Wever,,
 `[time.syn] <https://wg21.link/time.syn>`_,"Formatter ``chrono::zoned_time<Duration, TimeZonePtr>``",A ``<chrono>`` implementation,Mark de Wever,,
 
diff --git a/libcxx/include/__chrono/convert_to_tm.h b/libcxx/include/__chrono/convert_to_tm.h
index 1301cd6f1f1ad..d2c5cf922ba67 100644
--- a/libcxx/include/__chrono/convert_to_tm.h
+++ b/libcxx/include/__chrono/convert_to_tm.h
@@ -20,6 +20,7 @@
 #include <__chrono/month_weekday.h>
 #include <__chrono/monthday.h>
 #include <__chrono/statically_widen.h>
+#include <__chrono/sys_info.h>
 #include <__chrono/system_clock.h>
 #include <__chrono/time_point.h>
 #include <__chrono/weekday.h>
@@ -171,6 +172,10 @@ _LIBCPP_HIDE_FROM_ABI _Tm __convert_to_tm(const _ChronoT& __value) {
       if (__value.hours().count() > std::numeric_limits<decltype(__result.tm_hour)>::max())
         std::__throw_format_error("Formatting hh_mm_ss, encountered an hour overflow");
     __result.tm_hour = __value.hours().count();
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  } else if constexpr (same_as<_ChronoT, chrono::sys_info>) {
+    // Has no time information.
+#  endif
   } else
     static_assert(sizeof(_ChronoT) == 0, "Add the missing type specialization");
 
diff --git a/libcxx/include/__chrono/formatter.h b/libcxx/include/__chrono/formatter.h
index f76e7b2ea0e86..0196deb8c1ffb 100644
--- a/libcxx/include/__chrono/formatter.h
+++ b/libcxx/include/__chrono/formatter.h
@@ -24,6 +24,7 @@
 #include <__chrono/ostream.h>
 #include <__chrono/parser_std_format_spec.h>
 #include <__chrono/statically_widen.h>
+#include <__chrono/sys_info.h>
 #include <__chrono/system_clock.h>
 #include <__chrono/time_point.h>
 #include <__chrono/weekday.h>
@@ -186,10 +187,11 @@ __format_zone_offset(basic_stringstream<_CharT>& __sstr, chrono::seconds __offse
 
   chrono::hh_mm_ss __hms{__offset};
   std::ostreambuf_iterator<_CharT> __out_it{__sstr};
+  // Note HMS does not allow formatting hours > 23, but the offset is not limited to 24H.
+  std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:02}"), __hms.hours().count());
   if (__modifier)
-    std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:%H:%M}"), __hms);
-  else
-    std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:%H%M}"), __hms);
+    __sstr << _CharT(':');
+  std::format_to(__out_it, _LIBCPP_STATICALLY_WIDEN(_CharT, "{:02}"), __hms.minutes().count());
 }
 
 // Helper to store the time zone information needed for formatting.
@@ -202,7 +204,12 @@ struct _LIBCPP_HIDE_FROM_ABI __time_zone {
 
 template <class _Tp>
 _LIBCPP_HIDE_FROM_ABI __time_zone __convert_to_time_zone([[maybe_unused]] const _Tp& __value) {
-  return {"UTC", chrono::seconds{0}};
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  if constexpr (same_as<_Tp, chrono::sys_info>)
+    return {__value.abbrev, __value.offset};
+  else
+#  endif
+    return {"UTC", chrono::seconds{0}};
 }
 
 template <class _CharT, class _Tp>
@@ -409,6 +416,10 @@ _LIBCPP_HIDE_FROM_ABI constexpr bool __weekday_ok(const _Tp& __value) {
     return __value.weekday().ok();
   else if constexpr (__is_hh_mm_ss<_Tp>)
     return true;
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  else if constexpr (same_as<_Tp, chrono::sys_info>)
+    return true;
+#  endif
   else
     static_assert(sizeof(_Tp) == 0, "Add the missing type specialization");
 }
@@ -449,6 +460,10 @@ _LIBCPP_HIDE_FROM_ABI constexpr bool __weekday_name_ok(const _Tp& __value) {
     return __value.weekday().ok();
   else if constexpr (__is_hh_mm_ss<_Tp>)
     return true;
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  else if constexpr (same_as<_Tp, chrono::sys_info>)
+    return true;
+#  endif
   else
     static_assert(sizeof(_Tp) == 0, "Add the missing type specialization");
 }
@@ -489,6 +504,10 @@ _LIBCPP_HIDE_FROM_ABI constexpr bool __date_ok(const _Tp& __value) {
     return __value.ok();
   else if constexpr (__is_hh_mm_ss<_Tp>)
     return true;
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  else if constexpr (same_as<_Tp, chrono::sys_info>)
+    return true;
+#  endif
   else
     static_assert(sizeof(_Tp) == 0, "Add the missing type specialization");
 }
@@ -529,6 +548,10 @@ _LIBCPP_HIDE_FROM_ABI constexpr bool __month_name_ok(const _Tp& __value) {
     return __value.month().ok();
   else if constexpr (__is_hh_mm_ss<_Tp>)
     return true;
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+  else if constexpr (same_as<_Tp, chrono::sys_info>)
+    return true;
+#  endif
   else
     static_assert(sizeof(_Tp) == 0, "Add the missing type specialization");
 }
@@ -858,6 +881,20 @@ struct formatter<chrono::hh_mm_ss<_Duration>, _CharT> : public __formatter_chron
     return _Base::__parse(__ctx, __format_spec::__fields_chrono, __format_spec::__flags::__time);
   }
 };
+
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+template <__fmt_char_type _CharT>
+struct formatter<chrono::sys_info, _CharT> : public __formatter_chrono<_CharT> {
+public:
+  using _Base = __formatter_chrono<_CharT>;
+
+  template <class _ParseContext>
+  _LIBCPP_HIDE_FROM_ABI constexpr typename _ParseContext::iterator parse(_ParseContext& __ctx) {
+    return _Base::__parse(__ctx, __format_spec::__fields_chrono, __format_spec::__flags::__time_zone);
+  }
+};
+#  endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+
 #endif // if _LIBCPP_STD_VER >= 20
 
 _LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/include/__chrono/ostream.h b/libcxx/include/__chrono/ostream.h
index b687ef8059d5f..9476406d76107 100644
--- a/libcxx/include/__chrono/ostream.h
+++ b/libcxx/include/__chrono/ostream.h
@@ -19,6 +19,7 @@
 #include <__chrono/month_weekday.h>
 #include <__chrono/monthday.h>
 #include <__chrono/statically_widen.h>
+#include <__chrono/sys_info.h>
 #include <__chrono/system_clock.h>
 #include <__chrono/weekday.h>
 #include <__chrono/year.h>
@@ -262,6 +263,25 @@ operator<<(basic_ostream<_CharT, _Traits>& __os, const hh_mm_ss<_Duration> __hms
   return __os << std::format(__os.getloc(), _LIBCPP_STATICALLY_WIDEN(_CharT, "{:L%T}"), __hms);
 }
 
+#  if !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+
+template <class _CharT, class _Traits>
+_LIBCPP_HIDE_FROM_ABI basic_ostream<_CharT, _Traits>&
+operator<<(basic_ostream<_CharT, _Traits>& __os, const sys_info& __info) {
+  // __info.abbrev is always std::basic_string<char>.
+  // Since these strings typically are short the conversion should be cheap.
+  std::basic_string<_CharT> __abbrev{__info.abbrev.begin(), __info.abbrev.end()};
+  return __os << std::format(
+             _LIBCPP_STATICALLY_WIDEN(_CharT, "[{:%F %T}, {:%F %T}) {:%T} {:%Q%q} \"{}\""),
+             __info.begin,
+             __info.end,
+             hh_mm_ss{__info.offset},
+             __info.save,
+             __abbrev);
+}
+
+#  endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+
 } // namespace chrono
 
 #endif // if _LIBCPP_STD_VER >= 20
diff --git a/libcxx/include/__chrono/sys_info.h b/libcxx/include/__chrono/sys_info.h
index 794d22f2ccc1e..461d5322d413b 100644
--- a/libcxx/include/__chrono/sys_info.h
+++ b/libcxx/include/__chrono/sys_info.h
@@ -42,7 +42,7 @@ struct sys_info {
 
 } // namespace chrono
 
-#  endif //_LIBCPP_STD_VER >= 20
+#  endif // _LIBCPP_STD_VER >= 20
 
 _LIBCPP_END_NAMESPACE_STD
 
diff --git a/libcxx/include/chrono b/libcxx/include/chrono
index 513ae52006e89..f0275ec1eba7a 100644
--- a/libcxx/include/chrono
+++ b/libcxx/include/chrono
@@ -733,6 +733,10 @@ struct sys_info {
   string        abbrev;
 };
 
+template<class charT, class traits>                                              // C++20
+  basic_ostream<charT, traits>&
+    operator<<(basic_ostream<charT, traits>& os, const sys_info& si);
+
 // 25.10.5, class time_zone                                                      // C++20
 enum class choose {earliest, latest};
 class time_zone {
@@ -829,6 +833,7 @@ namespace std {
   template<class charT> struct formatter<chrono::year_month_weekday_last, charT>; // C++20
   template<class Rep, class Period, class charT>
     struct formatter<chrono::hh_mm_ss<duration<Rep, Period>>, charT>;             // C++20
+  template<class charT> struct formatter<chrono::sys_info, charT>;                // C++20
 } // namespace std
 
 namespace chrono {
diff --git a/libcxx/test/libcxx/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp b/libcxx/test/libcxx/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp
new file mode 100644
index 0000000000000..faa7d855c8de7
--- /dev/null
+++ b/libcxx/test/libcxx/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp
@@ -0,0 +1,74 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+// UNSUPPORTED: no-localization
+
+// TODO FMT This test should not require std::to_chars(floating-point)
+// XFAIL: availability-fp_to_chars-missing
+
+// XFAIL: libcpp-has-no-incomplete-tzdb
+
+// <chrono>
+
+// template<class charT, class traits>
+//   basic_ostream<charT, traits>&
+//     operator<<(basic_ostream<charT, traits>& os, const sys_info& r);
+
+// [time.zone.info.sys]
+//   7 Effects: Streams out the sys_info object r in an unspecified format.
+//   8 Returns: os.
+//
+// Tests the output produced by this function.
+
+#include <cassert>
+#include <chrono>
+#include <memory>
+#include <sstream>
+
+#include "assert_macros.h"
+#include "test_macros.h"
+#include "make_string.h"
+#include "concat_macros.h"
+
+#define SV(S) MAKE_STRING_VIEW(CharT, S)
+
+template <class CharT>
+static void test(std::basic_string_view<CharT> expected, std::chrono::sys_info&& info) {
+  std::basic_stringstream<CharT> sstr;
+  sstr << info;
+  std::basic_string<CharT> output = sstr.str();
+
+  TEST_REQUIRE(expected == output,
+               TEST_WRITE_CONCATENATED("\nExpected output ", expected, "\nActual output   ", output, '\n'));
+}
+
+template <class CharT>
+static void test() {
+  using namespace std::literals::chrono_literals;
+  namespace tz = std::chrono;
+
+  test(SV("[-10484-10-16 15:30:08, 14423-03-17 15:30:07) 00:00:00 0min \"TZ\""),
+       tz::sys_info{tz::sys_seconds::min(), tz::sys_seconds::max(), 0s, 0min, "TZ"});
+
+  test(SV("[1970-01-01 00:00:00, 2038-12-31 00:00:00) 12:23:45 -67min \"DMY\""),
+       tz::sys_info{static_cast<tz::sys_days>(tz::year_month_day{1970y, tz::January, 1d}),
+                    static_cast<tz::sys_days>(tz::year_month_day{2038y, tz::December, 31d}),
+                    12h + 23min + 45s,
+                    -67min,
+                    "DMY"});
+}
+
+int main(int, const char**) {
+  test<char>();
+#ifndef TEST_HAS_NO_WIDE_CHARACTERS
+  test<wchar_t>();
+#endif
+
+  return 0;
+}
diff --git a/libcxx/test/std/time/time.syn/formatter.sys_info.pass.cpp b/libcxx/test/std/time/time.syn/formatter.sys_info.pass.cpp
new file mode 100644
index 0000000000000..0a41424cfdcb8
--- /dev/null
+++ b/libcxx/test/std/time/time.syn/formatter.sys_info.pass.cpp
@@ -0,0 +1,137 @@
+//===----------------------------------------------------------------------===//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+// UNSUPPORTED: no-localization
+
+// TODO FMT This test should not require std::to_chars(floating-point)
+// XFAIL: availability-fp_to_chars-missing
+
+// XFAIL: libcpp-has-no-incomplete-tzdb
+
+// REQUIRES: locale.fr_FR.UTF-8
+// REQUIRES: locale.ja_JP.UTF-8
+
+// <chrono>
+//
+// template<class charT> struct formatter<chrono::sys_info, charT>;
+
+#include <chrono>
+#include <format>
+
+#include <cassert>
+#include <concepts>
+#include <locale>
+#include <iostream>
+#include <type_traits>
+
+#include "formatter_tests.h"
+#include "make_string.h"
+#include "platform_support.h" // locale name macros
+#include "test_macros.h"
+
+template <class CharT>
+static void test_no_chrono_specs() {
+// This test libc++ specific due to
+// [time.zone.info.sys]/7
+//   Effects: Streams out the sys_info object r in an unspecified format.
+#ifdef _LIBCPP_VERSION
+  using namespace std::literals::chrono_literals;
+  namespace tz = std::chrono;
+
+  std::locale::global(std::locale(LOCALE_fr_FR_UTF_8));
+
+  // Non localized output
+
+  check(SV("[-10484-10-16 15:30:08, 14423-03-17 15:30:07) 00:00:00 0min \"TZ\""),
+        SV("{}"),
+        tz::sys_info{tz::sys_seconds::min(), tz::sys_seconds::max(), 0s, 0min, "TZ"});
+
+  check(SV("[1970-01-01 00:00:00, 2038-12-31 00:00:00) 12:23:45 -67min \"DMY\""),
+        SV("{}"),
+        tz::sys_info{static_cast<tz::sys_days>(tz::year_month_day{1970y, tz::January, 1d}),
+                     static_cast<tz::sys_days>(tz::year_month_day{2038y, tz::December, 31d}),
+                     12h + 23min + 45s,
+                     -67min,
+                     "DMY"});
+
+  std::locale::global(std::locale::classic());
+#endif // _LIBCPP_VERSION
+}
+
+template <class CharT>
+static void test_valid_values() {
+  using namespace std::literals::chrono_literals;
+
+  constexpr std::basic_string_view<CharT> fmt  = SV("{:%%z='%z'%t%%Ez='%Ez'%t%%Oz='%Oz'%t%%Z='%Z'%n}");
+  constexpr std::basic_string_view<CharT> lfmt = SV("{:L%%z='%z'%t%%Ez='%Ez'%t%%Oz='%Oz'%t%%Z='%Z'%n}");
+
+  const std::locale loc(LOCALE_ja_JP_UTF_8);
+  std::locale::global(std::locale(LOCALE_fr_FR_UTF_8));
+
+  // Non localized output using C-locale
+  check(SV("%z='-0200'\t%Ez='-02:00'\t%Oz='-02:00'\t%Z='NEG'\n"),
+        fmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, -2h, 0min, "NEG"});
+
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='ZERO'\n"),
+        fmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 0s, 0min, "ZERO"});
+
+  check(SV("%z='+1115'\t%Ez='+11:15'\t%Oz='+11:15'\t%Z='POS'\n"),
+        fmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 11h + 15min, 0min, "POS"});
+
+  // Use the global locale (fr_FR)
+  check(SV("%z='-0200'\t%Ez='-02:00'\t%Oz='-02:00'\t%Z='NEG'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, -2h, 0min, "NEG"});
+
+  check(SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='ZERO'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 0s, 0min, "ZERO"});
+
+  check(SV("%z='+1115'\t%Ez='+11:15'\t%Oz='+11:15'\t%Z='POS'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 11h + 15min, 0min, "POS"});
+
+  // Use supplied locale (ja_JP).
+  check(loc,
+        SV("%z='-0200'\t%Ez='-02:00'\t%Oz='-02:00'\t%Z='NEG'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, -2h, 0min, "NEG"});
+
+  check(loc,
+        SV("%z='+0000'\t%Ez='+00:00'\t%Oz='+00:00'\t%Z='ZERO'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 0s, 0min, "ZERO"});
+
+  check(loc,
+        SV("%z='+1115'\t%Ez='+11:15'\t%Oz='+11:15'\t%Z='POS'\n"),
+        lfmt,
+        std::chrono::sys_info{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 11h + 15min, 0min, "POS"});
+
+  std::locale::global(std::locale::classic());
+}
+
+template <class CharT>
+static void test() {
+  test_no_chrono_specs<CharT>();
+  test_valid_values<CharT>();
+
+  check_invalid_types<CharT>({SV("z"), SV("Z"), SV("Ez"), SV("Oz")}, std::chrono::sys_info{});
+}
+
+int main(int, char**) {
+  test<char>();
+
+#ifndef TEST_HAS_NO_WIDE_CHARACTERS
+  test<wchar_t>();
+#endif
+
+  return 0;
+}
diff --git a/libcxx/test/std/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp b/libcxx/test/std/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp
new file mode 100644
index 0000000000000..82c4844b423c5
--- /dev/null
+++ b/libcxx/test/std/time/time.zone/time.zone.info/time.zone.info.sys/ostream.pass.cpp
@@ -0,0 +1,52 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, c++11, c++14, c++17
+// UNSUPPORTED: no-localization
+
+// TODO FMT This test should not require std::to_chars(floating-point)
+// XFAIL: availability-fp_to_chars-missing
+
+// XFAIL: libcpp-has-no-incomplete-tzdb
+
+// <chrono>
+
+// template<class charT, class traits>
+//   basic_ostream<charT, traits>&
+//     operator<<(basic_ostream<charT, traits>& os, const sys_info& r);
+
+// [time.zone.info.sys]
+//   7 Effects: Streams out the sys_info object r in an unspecified format.
+//   8 Returns: os.
+//
+// There is a private libc++ test that validates the exact output.
+
+#include <cassert>
+#include <chrono>
+#include <memory>
+#include <sstream>
+
+#include "test_macros.h"
+
+template <class CharT>
+static void test() {
+  using namespace std::literals::chrono_literals;
+  std::chrono::sys_info s{std::chrono::sys_seconds{0s}, std::chrono::sys_seconds{0s}, 0h, 0min, ""};
+  std::basic_ostringstream<CharT> os;
+  std::basic_ostream<CharT>& result = std::chrono::operator<<(os, s);
+  assert(std::addressof(result) == std::addressof(os));
+}
+
+int main(int, const char**) {
+  test<char>();
+#ifndef TEST_HAS_NO_WIDE_CHARACTERS
+  test<wchar_t>();
+#endif
+
+  return 0;
+}
diff --git a/libcxx/test/std/utilities/format/format.formattable/concept.formattable.compile.pass.cpp b/libcxx/test/std/utilities/format/format.formattable/concept.formattable.compile.pass.cpp
index fad5277fd6e11..f40784ec446fa 100644
--- a/libcxx/test/std/utilities/format/format.formattable/concept.formattable.compile.pass.cpp
+++ b/libcxx/test/std/utilities/format/format.formattable/concept.formattable.compile.pass.cpp
@@ -176,10 +176,12 @@ void test_P1361() {
 
   assert_is_formattable<std::chrono::hh_mm_ss<std::chrono::microseconds>, CharT>();
 
-  //assert_is_formattable<std::chrono::sys_info, CharT>();
+#  if !defined(TEST_HAS_NO_INCOMPLETE_TZDB)
+  assert_is_formattable<std::chrono::sys_info, CharT>();
   //assert_is_formattable<std::chrono::local_info, CharT>();
 
   //assert_is_formattable<std::chrono::zoned_time, CharT>();
+#  endif // !defined(TEST_HAS_NO_INCOMPLETE_TZDB)
 
 #endif // TEST_HAS_NO_LOCALIZATION
 }
diff --git a/libcxx/test/support/test_macros.h b/libcxx/test/support/test_macros.h
index 7b2dcbb52d0c8..fe9207d7e5969 100644
--- a/libcxx/test/support/test_macros.h
+++ b/libcxx/test/support/test_macros.h
@@ -417,6 +417,14 @@ inline Tp const& DoNotOptimize(Tp const& value) {
 #  define TEST_HAS_NO_RANDOM_DEVICE
 #endif
 
+#if defined(_LIBCPP_HAS_NO_INCOMPLETE_TZDB)
+#  define TEST_HAS_NO_INCOMPLETE_TZDB
+#endif
+
+#if defined(_LIBCPP_HAS_NO_TIME_ZONE_DATABASE)
+#  define TEST_HAS_NO_TIME_ZONE_DATABASE
+#endif
+
 #if defined(TEST_COMPILER_CLANG)
 #  define TEST_DIAGNOSTIC_PUSH _Pragma("clang diagnostic push")
 #  define TEST_DIAGNOSTIC_POP _Pragma("clang diagnostic pop")

From b1dc62f139ef265a36a2a739ce9ba4e1e48a6dbe Mon Sep 17 00:00:00 2001
From: Amirreza Ashouri <ar.ashouri999@gmail.com>
Date: Wed, 17 Apr 2024 22:28:11 +0330
Subject: [PATCH 785/886] [clang]Treat arguments to builtin type traits as
 template type arguments (#87132)

This change improves error messages for builtins in case of empty
parentheses.

Fixes llvm#86997
---
 clang/lib/Parse/ParseExprCXX.cpp           | 12 ++++++------
 clang/test/Sema/static-assert.c            |  9 ++++-----
 clang/test/SemaCXX/builtins.cpp            |  5 +++++
 clang/test/SemaCXX/deprecated-builtins.cpp |  5 +++++
 4 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/clang/lib/Parse/ParseExprCXX.cpp b/clang/lib/Parse/ParseExprCXX.cpp
index 43d6105dcf31c..0d2ad980696fc 100644
--- a/clang/lib/Parse/ParseExprCXX.cpp
+++ b/clang/lib/Parse/ParseExprCXX.cpp
@@ -3910,10 +3910,10 @@ ExprResult Parser::ParseTypeTrait() {
   SmallVector<ParsedType, 2> Args;
   do {
     // Parse the next type.
-    TypeResult Ty =
-        ParseTypeName(/*SourceRange=*/nullptr,
-                      getLangOpts().CPlusPlus ? DeclaratorContext::TemplateArg
-                                              : DeclaratorContext::TypeName);
+    TypeResult Ty = ParseTypeName(/*SourceRange=*/nullptr,
+                                  getLangOpts().CPlusPlus
+                                      ? DeclaratorContext::TemplateTypeArg
+                                      : DeclaratorContext::TypeName);
     if (Ty.isInvalid()) {
       Parens.skipToEnd();
       return ExprError();
@@ -3955,8 +3955,8 @@ ExprResult Parser::ParseArrayTypeTrait() {
   if (T.expectAndConsume())
     return ExprError();
 
-  TypeResult Ty =
-      ParseTypeName(/*SourceRange=*/nullptr, DeclaratorContext::TemplateArg);
+  TypeResult Ty = ParseTypeName(/*SourceRange=*/nullptr,
+                                DeclaratorContext::TemplateTypeArg);
   if (Ty.isInvalid()) {
     SkipUntil(tok::comma, StopAtSemi);
     SkipUntil(tok::r_paren, StopAtSemi);
diff --git a/clang/test/Sema/static-assert.c b/clang/test/Sema/static-assert.c
index ae5e8076e0bed..4e9e6b7ee558b 100644
--- a/clang/test/Sema/static-assert.c
+++ b/clang/test/Sema/static-assert.c
@@ -1,6 +1,6 @@
-// RUN: %clang_cc1 -std=c11 -Wgnu-folding-constant -fsyntax-only -verify=expected,c %s
-// RUN: %clang_cc1 -fms-compatibility -Wgnu-folding-constant -DMS -fsyntax-only -verify=expected,ms,c %s
-// RUN: %clang_cc1 -std=c99 -pedantic -Wgnu-folding-constant -fsyntax-only -verify=expected,ext,c %s
+// RUN: %clang_cc1 -std=c11 -Wgnu-folding-constant -fsyntax-only -verify %s
+// RUN: %clang_cc1 -fms-compatibility -Wgnu-folding-constant -DMS -fsyntax-only -verify=expected,ms %s
+// RUN: %clang_cc1 -std=c99 -pedantic -Wgnu-folding-constant -fsyntax-only -verify=expected,ext %s
 // RUN: %clang_cc1 -xc++ -std=c++11 -pedantic -fsyntax-only -verify=expected,ext,cxx %s
 
 _Static_assert("foo", "string is nonzero"); // ext-warning {{'_Static_assert' is a C11 extension}}
@@ -57,8 +57,7 @@ UNION(char[2], short) u2 = { .one = { 'a', 'b' } }; // ext-warning 3 {{'_Static_
 typedef UNION(char, short) U3; // expected-error {{static assertion failed due to requirement 'sizeof(char) == sizeof(short)': type size mismatch}} \
                                // expected-note{{evaluates to '1 == 2'}} \
                                // ext-warning 3 {{'_Static_assert' is a C11 extension}}
-typedef UNION(float, 0.5f) U4; // c-error {{expected a type}} \
-                               // cxx-error {{type name requires a specifier or qualifier}} \
+typedef UNION(float, 0.5f) U4; // expected-error {{expected a type}} \
                                // ext-warning 3 {{'_Static_assert' is a C11 extension}}
 
 // After defining the assert macro in MS-compatibility mode, we should
diff --git a/clang/test/SemaCXX/builtins.cpp b/clang/test/SemaCXX/builtins.cpp
index 567094c94c171..080b4476c7eec 100644
--- a/clang/test/SemaCXX/builtins.cpp
+++ b/clang/test/SemaCXX/builtins.cpp
@@ -76,6 +76,11 @@ using ConstMemFnType = int (Dummy::*)() const;
 
 void foo() {}
 
+void test_builtin_empty_parentheses_diags() {
+  __is_trivially_copyable(); // expected-error {{expected a type}}
+  __is_trivially_copyable(1); // expected-error {{expected a type}}
+}
+
 void test_builtin_launder_diags(void *vp, const void *cvp, FnType *fnp,
                                 MemFnType mfp, ConstMemFnType cmfp, int (&Arr)[5]) {
   __builtin_launder(vp);   // expected-error {{void pointer argument to '__builtin_launder' is not allowed}}
diff --git a/clang/test/SemaCXX/deprecated-builtins.cpp b/clang/test/SemaCXX/deprecated-builtins.cpp
index 849b9b014fff2..fafc1da4da13e 100644
--- a/clang/test/SemaCXX/deprecated-builtins.cpp
+++ b/clang/test/SemaCXX/deprecated-builtins.cpp
@@ -17,3 +17,8 @@ void f() {
     a = __has_trivial_destructor(A);  // expected-warning-re {{__has_trivial_destructor {{.*}} use __is_trivially_destructible}}
 
 }
+
+void test_builtin_empty_parentheses_diags(void) {
+    __has_nothrow_copy(); // expected-error {{expected a type}}
+    __has_nothrow_copy(1); // expected-error {{expected a type}}
+}

From 6cea7c491f4c4c68aa0494a9b18f36ff40c22c81 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 17 Apr 2024 12:04:18 -0700
Subject: [PATCH 786/886] [X86] Always use 64-bit relocations in no-PIC large
 code model (#89101)

This matches other types of relocations, e.g. to constant pool. And
makes things more consistent with PIC large code model.

Some users of the large code model may not place small data in the lower
2GB of the address space (e.g.
https://github.com/ClangBuiltLinux/linux/issues/2016), so just
unconditionally use 64-bit relocations in the large code model.

So now functions in a section not marked large will use 64-bit
relocations to reference everything when using the large code model.

This also fixes some lldb tests broken by #88172
(https://lab.llvm.org/buildbot/#/builders/68/builds/72458).
---
 llvm/lib/Target/X86/X86ISelDAGToDAG.cpp | 7 +++----
 llvm/test/CodeGen/X86/code-model-elf.ll | 4 ++--
 2 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index 4e4241efd63d6..7dcde2a508949 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -2927,11 +2927,10 @@ bool X86DAGToDAGISel::selectAddr(SDNode *Parent, SDValue N, SDValue &Base,
 }
 
 bool X86DAGToDAGISel::selectMOV64Imm32(SDValue N, SDValue &Imm) {
-  // Cannot use 32 bit constants to reference objects in kernel code model.
-  // Cannot use 32 bit constants to reference objects in large PIC mode since
-  // GOTOFF is 64 bits.
+  // Cannot use 32 bit constants to reference objects in kernel/large code
+  // model.
   if (TM.getCodeModel() == CodeModel::Kernel ||
-      (TM.getCodeModel() == CodeModel::Large && TM.isPositionIndependent()))
+      TM.getCodeModel() == CodeModel::Large)
     return false;
 
   // In static codegen with small code model, we can get the address of a label
diff --git a/llvm/test/CodeGen/X86/code-model-elf.ll b/llvm/test/CodeGen/X86/code-model-elf.ll
index 0da62e3e7a651..f60f75bc26911 100644
--- a/llvm/test/CodeGen/X86/code-model-elf.ll
+++ b/llvm/test/CodeGen/X86/code-model-elf.ll
@@ -350,7 +350,7 @@ define dso_local ptr @lea_forced_small_data() #0 {
 ;
 ; LARGE-STATIC-LABEL: lea_forced_small_data:
 ; LARGE-STATIC:       # %bb.0:
-; LARGE-STATIC-NEXT:    movl $forced_small_data, %eax
+; LARGE-STATIC-NEXT:    movabsq $forced_small_data, %rax
 ; LARGE-STATIC-NEXT:    retq
 ;
 ; SMALL-PIC-LABEL: lea_forced_small_data:
@@ -403,7 +403,7 @@ define dso_local i32 @load_forced_small_data() #0 {
 ;
 ; LARGE-STATIC-LABEL: load_forced_small_data:
 ; LARGE-STATIC:       # %bb.0:
-; LARGE-STATIC-NEXT:    movl $forced_small_data+8, %eax
+; LARGE-STATIC-NEXT:    movabsq $forced_small_data+8, %rax
 ; LARGE-STATIC-NEXT:    movl (%rax), %eax
 ; LARGE-STATIC-NEXT:    retq
 ;

From 1460b4964c7ada2f7536006722c8585b5bd0a1b5 Mon Sep 17 00:00:00 2001
From: Arthur Eubanks <aeubanks@google.com>
Date: Wed, 17 Apr 2024 19:11:58 +0000
Subject: [PATCH 787/886] [gn build] Manually port d423d80e560d

---
 llvm/utils/gn/secondary/libcxx/include/BUILD.gn | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index ee44558a4e994..bd08ce044c247 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -37,9 +37,9 @@ if (current_toolchain == default_toolchain) {
       "_LIBCPP_INSTRUMENTED_WITH_ASAN=",
       "_LIBCPP_ABI_DEFINES=",
       "_LIBCPP_HARDENING_MODE_DEFAULT=_LIBCPP_HARDENING_MODE_NONE",
-      "_LIBCPP_PSTL_CPU_BACKEND_LIBDISPATCH=",
-      "_LIBCPP_PSTL_CPU_BACKEND_SERIAL=1",
-      "_LIBCPP_PSTL_CPU_BACKEND_THREAD=",
+      "_LIBCPP_PSTL_BACKEND_LIBDISPATCH=",
+      "_LIBCPP_PSTL_BACKEND_SERIAL=1",
+      "_LIBCPP_PSTL_BACKEND_STD_THREAD=",
     ]
     if (libcxx_abi_version != 1) {
       values += [ "_LIBCPP_ABI_VERSION=$libcxx_abi_version" ]
@@ -143,18 +143,12 @@ if (current_toolchain == default_toolchain) {
       "__algorithm/pop_heap.h",
       "__algorithm/prev_permutation.h",
       "__algorithm/pstl_any_all_none_of.h",
-      "__algorithm/pstl_backend.h",
-      "__algorithm/pstl_backends/cpu_backend.h",
       "__algorithm/pstl_backends/cpu_backends/any_of.h",
-      "__algorithm/pstl_backends/cpu_backends/backend.h",
       "__algorithm/pstl_backends/cpu_backends/fill.h",
       "__algorithm/pstl_backends/cpu_backends/find_if.h",
       "__algorithm/pstl_backends/cpu_backends/for_each.h",
-      "__algorithm/pstl_backends/cpu_backends/libdispatch.h",
       "__algorithm/pstl_backends/cpu_backends/merge.h",
-      "__algorithm/pstl_backends/cpu_backends/serial.h",
       "__algorithm/pstl_backends/cpu_backends/stable_sort.h",
-      "__algorithm/pstl_backends/cpu_backends/thread.h",
       "__algorithm/pstl_backends/cpu_backends/transform.h",
       "__algorithm/pstl_backends/cpu_backends/transform_reduce.h",
       "__algorithm/pstl_copy.h",
@@ -664,6 +658,11 @@ if (current_toolchain == default_toolchain) {
       "__numeric/transform_exclusive_scan.h",
       "__numeric/transform_inclusive_scan.h",
       "__numeric/transform_reduce.h",
+      "__pstl/backends/libdispatch.h",
+      "__pstl/backends/serial.h",
+      "__pstl/backends/std_thread.h",
+      "__pstl/configuration.h",
+      "__pstl/configuration_fwd.h",
       "__pstl/cpu_algos/cpu_traits.h",
       "__random/bernoulli_distribution.h",
       "__random/binomial_distribution.h",

From db2f64ee1f743153f242ca92bf627458ba7f8000 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault@amd.com>
Date: Wed, 17 Apr 2024 21:20:39 +0200
Subject: [PATCH 788/886] AMDGPU: Fix not handling atomicrmw fadd in exotic
 address spaces correctly

We try to interpret unknown address space numbers as aliases of global,
but this wasn't applied here. Also improve test coverage for the
buffer fat pointer address space.
---
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     |   2 +-
 .../AMDGPU/expand-atomic-rmw-fadd.ll          | 414 ++++++++++++++++++
 2 files changed, 415 insertions(+), 1 deletion(-)

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 6fde73119f815..0c706d51cb665 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -16062,7 +16062,7 @@ SITargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *RMW) const {
       if (HasSystemScope)
         return AtomicExpansionKind::CmpXChg;
 
-      if ((AS == AMDGPUAS::GLOBAL_ADDRESS ||
+      if ((AMDGPU::isExtendedGlobalAddrSpace(AS) ||
            AS == AMDGPUAS::BUFFER_FAT_POINTER) &&
           Ty->isFloatTy()) {
         // global/buffer atomic fadd f32 no-rtn: gfx908, gfx90a, gfx940, gfx11+.
diff --git a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
index a5d4c329446f9..9010dec31c161 100644
--- a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
+++ b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd.ll
@@ -59,6 +59,112 @@ define void @test_atomicrmw_fadd_f32_global_no_use_unsafe(ptr addrspace(1) %ptr,
   ret void
 }
 
+define void @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(ptr addrspace(7) %ptr, float %value) #0 {
+; CI-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret void
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret void
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; GFX908-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX908-NEXT:    ret void
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; GFX90A-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX90A-NEXT:    ret void
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX940-NEXT:    ret void
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_no_use_unsafe(
+; GFX11-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX11-NEXT:    ret void
+;
+  %res = atomicrmw fadd ptr addrspace(7) %ptr, float %value syncscope("wavefront") monotonic
+  ret void
+}
+
+define void @test_atomicrmw_fadd_f32_as999_no_use_unsafe(ptr addrspace(999) %ptr, float %value) #0 {
+; CI-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret void
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret void
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; GFX908-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX908-NEXT:    ret void
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; GFX90A-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX90A-NEXT:    ret void
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX940-NEXT:    ret void
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_as999_no_use_unsafe(
+; GFX11-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX11-NEXT:    ret void
+;
+  %res = atomicrmw fadd ptr addrspace(999) %ptr, float %value syncscope("wavefront") monotonic
+  ret void
+}
+
 define float @test_atomicrmw_fadd_f32_global_unsafe(ptr addrspace(1) %ptr, float %value) #0 {
 ; CI-LABEL: @test_atomicrmw_fadd_f32_global_unsafe(
 ; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(1) [[PTR:%.*]], align 4
@@ -124,6 +230,136 @@ define float @test_atomicrmw_fadd_f32_global_unsafe(ptr addrspace(1) %ptr, float
   ret float %res
 }
 
+define float @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(ptr addrspace(7) %ptr, float %value) #0 {
+; CI-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret float [[TMP5]]
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret float [[TMP5]]
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; GFX908-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX908-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX908:       atomicrmw.start:
+; GFX908-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX908-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX908-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX908-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX908-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX908-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX908-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX908-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX908-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX908:       atomicrmw.end:
+; GFX908-NEXT:    ret float [[TMP5]]
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; GFX90A-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX90A-NEXT:    ret float [[RES]]
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX940-NEXT:    ret float [[RES]]
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr_unsafe(
+; GFX11-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX11-NEXT:    ret float [[RES]]
+;
+  %res = atomicrmw fadd ptr addrspace(7) %ptr, float %value syncscope("wavefront") monotonic
+  ret float %res
+}
+
+define float @test_atomicrmw_fadd_f32_as999_unsafe(ptr addrspace(999) %ptr, float %value) #0 {
+; CI-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret float [[TMP5]]
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret float [[TMP5]]
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; GFX908-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX908-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX908:       atomicrmw.start:
+; GFX908-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX908-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX908-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX908-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX908-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] syncscope("wavefront") monotonic monotonic, align 4
+; GFX908-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX908-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX908-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX908-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX908:       atomicrmw.end:
+; GFX908-NEXT:    ret float [[TMP5]]
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; GFX90A-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX90A-NEXT:    ret float [[RES]]
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX940-NEXT:    ret float [[RES]]
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_as999_unsafe(
+; GFX11-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] syncscope("wavefront") monotonic, align 4
+; GFX11-NEXT:    ret float [[RES]]
+;
+  %res = atomicrmw fadd ptr addrspace(999) %ptr, float %value syncscope("wavefront") monotonic
+  ret float %res
+}
+
 define double @test_atomicrmw_fadd_f64_global_unsafe(ptr addrspace(1) %ptr, double %value) #0 {
 ; CI-LABEL: @test_atomicrmw_fadd_f64_global_unsafe(
 ; CI-NEXT:    [[TMP1:%.*]] = load double, ptr addrspace(1) [[PTR:%.*]], align 8
@@ -545,6 +781,184 @@ define float @test_atomicrmw_fadd_f32_global(ptr addrspace(1) %ptr, float %value
   ret float %res
 }
 
+define float @test_atomicrmw_fadd_f32_buffer_fat_ptr(ptr addrspace(7) %ptr, float %value) {
+; CI-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret float [[TMP5]]
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret float [[TMP5]]
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; GFX908-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX908-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX908:       atomicrmw.start:
+; GFX908-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX908-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX908-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX908-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX908-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX908-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX908-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX908-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX908-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX908:       atomicrmw.end:
+; GFX908-NEXT:    ret float [[TMP5]]
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; GFX90A-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX90A-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX90A:       atomicrmw.start:
+; GFX90A-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX90A-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX90A-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX90A-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX90A-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX90A-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX90A-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX90A-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX90A-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX90A:       atomicrmw.end:
+; GFX90A-NEXT:    ret float [[TMP5]]
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(7) [[PTR:%.*]], float [[VALUE:%.*]] seq_cst, align 4
+; GFX940-NEXT:    ret float [[RES]]
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_buffer_fat_ptr(
+; GFX11-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(7) [[PTR:%.*]], align 4
+; GFX11-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX11:       atomicrmw.start:
+; GFX11-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX11-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX11-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX11-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX11-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(7) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX11-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX11-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX11-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX11-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX11:       atomicrmw.end:
+; GFX11-NEXT:    ret float [[TMP5]]
+;
+  %res = atomicrmw fadd ptr addrspace(7) %ptr, float %value seq_cst
+  ret float %res
+}
+
+define float @test_atomicrmw_fadd_f32_as999(ptr addrspace(999) %ptr, float %value) {
+; CI-LABEL: @test_atomicrmw_fadd_f32_as999(
+; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; CI-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; CI:       atomicrmw.start:
+; CI-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; CI-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; CI-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; CI-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; CI-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; CI-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; CI-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; CI-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; CI-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CI:       atomicrmw.end:
+; CI-NEXT:    ret float [[TMP5]]
+;
+; GFX9-LABEL: @test_atomicrmw_fadd_f32_as999(
+; GFX9-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX9-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX9:       atomicrmw.start:
+; GFX9-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX9-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX9-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX9-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX9-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX9-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX9-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX9-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX9-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX9:       atomicrmw.end:
+; GFX9-NEXT:    ret float [[TMP5]]
+;
+; GFX908-LABEL: @test_atomicrmw_fadd_f32_as999(
+; GFX908-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX908-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX908:       atomicrmw.start:
+; GFX908-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX908-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX908-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX908-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX908-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX908-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX908-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX908-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX908-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX908:       atomicrmw.end:
+; GFX908-NEXT:    ret float [[TMP5]]
+;
+; GFX90A-LABEL: @test_atomicrmw_fadd_f32_as999(
+; GFX90A-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX90A-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX90A:       atomicrmw.start:
+; GFX90A-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX90A-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX90A-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX90A-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX90A-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX90A-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX90A-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX90A-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX90A-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX90A:       atomicrmw.end:
+; GFX90A-NEXT:    ret float [[TMP5]]
+;
+; GFX940-LABEL: @test_atomicrmw_fadd_f32_as999(
+; GFX940-NEXT:    [[RES:%.*]] = atomicrmw fadd ptr addrspace(999) [[PTR:%.*]], float [[VALUE:%.*]] seq_cst, align 4
+; GFX940-NEXT:    ret float [[RES]]
+;
+; GFX11-LABEL: @test_atomicrmw_fadd_f32_as999(
+; GFX11-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(999) [[PTR:%.*]], align 4
+; GFX11-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX11:       atomicrmw.start:
+; GFX11-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP1]], [[TMP0:%.*]] ], [ [[TMP5:%.*]], [[ATOMICRMW_START]] ]
+; GFX11-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[VALUE:%.*]]
+; GFX11-NEXT:    [[TMP2:%.*]] = bitcast float [[NEW]] to i32
+; GFX11-NEXT:    [[TMP3:%.*]] = bitcast float [[LOADED]] to i32
+; GFX11-NEXT:    [[TMP4:%.*]] = cmpxchg ptr addrspace(999) [[PTR]], i32 [[TMP3]], i32 [[TMP2]] seq_cst seq_cst, align 4
+; GFX11-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
+; GFX11-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
+; GFX11-NEXT:    [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
+; GFX11-NEXT:    br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; GFX11:       atomicrmw.end:
+; GFX11-NEXT:    ret float [[TMP5]]
+;
+  %res = atomicrmw fadd ptr addrspace(999) %ptr, float %value seq_cst
+  ret float %res
+}
+
 define void @test_atomicrmw_fadd_f32_global_no_use_ieee(ptr addrspace(1) %ptr, float %value) {
 ; CI-LABEL: @test_atomicrmw_fadd_f32_global_no_use_ieee(
 ; CI-NEXT:    [[TMP1:%.*]] = load float, ptr addrspace(1) [[PTR:%.*]], align 4

From 652bcf685c72447f3cc46d93d6c9c1948e8499f3 Mon Sep 17 00:00:00 2001
From: Matthias Braun <matze@braunis.de>
Date: Wed, 17 Apr 2024 12:48:02 -0700
Subject: [PATCH 789/886] CodeGenPrepare: Add support for
 llvm.threadlocal.address address-mode sinking (#87844)

Depending on the TLSMode many thread-local accesses on x86 can be
expressed by adding a %fs: segment register to an addressing mode. Even
if there are mutliple users of a `llvm.threadlocal.address` intrinsic it
is generally not worth sharing the value in a register but instead fold
the %fs access into multiple addressing modes.

Hence this changes CodeGenPrepare to duplicate the
`llvm.threadlocal.address` intrinsic as necessary.

Introduces a new `TargetLowering::addressingModeSupportsTLS` callback
that allows targets to indicate whether TLS accesses can be part of an
addressing mode.

This is fixing a performance problem, as this folding of TLS-accesses
into multiple addressing modes happened naturally before the
introduction of the `llvm.threadlocal.address` intrinsic, but regressed
due to `SelectionDAG` keeping things in registers when accessed across
basic blocks, so CodeGenPrepare needs to duplicate to mitigate this. We
see a ~0.5% recovery in a codebase with heavy TLS usage (HHVM).

This fixes most of #87437
---
 llvm/include/llvm/CodeGen/TargetLowering.h    |   6 +
 llvm/lib/CodeGen/CodeGenPrepare.cpp           |  29 ++-
 llvm/lib/Target/X86/X86ISelLowering.cpp       |  24 +++
 llvm/lib/Target/X86/X86ISelLowering.h         |   2 +
 .../X86/codegen-prepare-addrmode-tls.ll       | 193 ++++++++++++++++++
 .../CodeGenPrepare/X86/sink-addrmode-tls.ll   |  86 ++++++++
 6 files changed, 336 insertions(+), 4 deletions(-)
 create mode 100644 llvm/test/CodeGen/X86/codegen-prepare-addrmode-tls.ll
 create mode 100644 llvm/test/Transforms/CodeGenPrepare/X86/sink-addrmode-tls.ll

diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index e0ade02959025..2dd978c7b5849 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -2805,6 +2805,12 @@ class TargetLoweringBase {
                                      Type *Ty, unsigned AddrSpace,
                                      Instruction *I = nullptr) const;
 
+  /// Returns true if the targets addressing mode can target thread local
+  /// storage (TLS).
+  virtual bool addressingModeSupportsTLS(const GlobalValue &) const {
+    return false;
+  }
+
   /// Return the prefered common base offset.
   virtual int64_t getPreferredLargeGEPBaseOffset(int64_t MinOffset,
                                                  int64_t MaxOffset) const {
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index e657872c38284..22a766f8d6252 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -5082,6 +5082,15 @@ bool AddressingModeMatcher::matchOperationAddr(User *AddrInst, unsigned Opcode,
     }
     return true;
   }
+  case Instruction::Call:
+    if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(AddrInst)) {
+      if (II->getIntrinsicID() == Intrinsic::threadlocal_address) {
+        GlobalValue &GV = cast<GlobalValue>(*II->getArgOperand(0));
+        if (TLI.addressingModeSupportsTLS(GV))
+          return matchAddr(AddrInst->getOperand(0), Depth);
+      }
+    }
+    break;
   }
   return false;
 }
@@ -5620,11 +5629,16 @@ bool CodeGenPrepare::optimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
         return Modified;
     }
 
-    if (AddrMode.BaseGV) {
+    GlobalValue *BaseGV = AddrMode.BaseGV;
+    if (BaseGV != nullptr) {
       if (ResultPtr)
         return Modified;
 
-      ResultPtr = AddrMode.BaseGV;
+      if (BaseGV->isThreadLocal()) {
+        ResultPtr = Builder.CreateThreadLocalAddress(BaseGV);
+      } else {
+        ResultPtr = BaseGV;
+      }
     }
 
     // If the real base value actually came from an inttoptr, then the matcher
@@ -5789,8 +5803,15 @@ bool CodeGenPrepare::optimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
     }
 
     // Add in the BaseGV if present.
-    if (AddrMode.BaseGV) {
-      Value *V = Builder.CreatePtrToInt(AddrMode.BaseGV, IntPtrTy, "sunkaddr");
+    GlobalValue *BaseGV = AddrMode.BaseGV;
+    if (BaseGV != nullptr) {
+      Value *BaseGVPtr;
+      if (BaseGV->isThreadLocal()) {
+        BaseGVPtr = Builder.CreateThreadLocalAddress(BaseGV);
+      } else {
+        BaseGVPtr = BaseGV;
+      }
+      Value *V = Builder.CreatePtrToInt(BaseGVPtr, IntPtrTy, "sunkaddr");
       if (Result)
         Result = Builder.CreateAdd(Result, V, "sunkaddr");
       else
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 27107f554fccf..bedec0c8974a8 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -18920,6 +18920,30 @@ X86TargetLowering::LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const {
   llvm_unreachable("TLS not implemented for this target.");
 }
 
+bool X86TargetLowering::addressingModeSupportsTLS(const GlobalValue &GV) const {
+  if (Subtarget.is64Bit() && Subtarget.isTargetELF()) {
+    const TargetMachine &TM = getTargetMachine();
+    TLSModel::Model Model = TM.getTLSModel(&GV);
+    switch (Model) {
+    case TLSModel::LocalExec:
+    case TLSModel::InitialExec:
+      // We can include the %fs segment register in addressing modes.
+      return true;
+    case TLSModel::LocalDynamic:
+    case TLSModel::GeneralDynamic:
+      // These models do not result in %fs relative addresses unless
+      // TLS descriptior are used.
+      //
+      // Even in the case of TLS descriptors we currently have no way to model
+      // the difference between %fs access and the computations needed for the
+      // offset and returning `true` for TLS-desc currently duplicates both
+      // which is detrimental :-/
+      return false;
+    }
+  }
+  return false;
+}
+
 /// Lower SRA_PARTS and friends, which return two i32 values
 /// and take a 2 x i32 value to shift plus a shift amount.
 /// TODO: Can this be moved to general expansion code?
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 0a1e8ca442731..e348ba6e8ac08 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -1323,6 +1323,8 @@ namespace llvm {
                                Type *Ty, unsigned AS,
                                Instruction *I = nullptr) const override;
 
+    bool addressingModeSupportsTLS(const GlobalValue &GV) const override;
+
     /// Return true if the specified immediate is legal
     /// icmp immediate, that is the target has icmp instructions which can
     /// compare a register against the immediate without having to materialize
diff --git a/llvm/test/CodeGen/X86/codegen-prepare-addrmode-tls.ll b/llvm/test/CodeGen/X86/codegen-prepare-addrmode-tls.ll
new file mode 100644
index 0000000000000..0ca1da26fa89c
--- /dev/null
+++ b/llvm/test/CodeGen/X86/codegen-prepare-addrmode-tls.ll
@@ -0,0 +1,193 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -o - %s | FileCheck %s --check-prefix=NOPIC
+; RUN: llc -o - %s -relocation-model=pic | FileCheck %s --check-prefix=PIC
+; RUN: llc -o - %s -relocation-model=pic -enable-tlsdesc | FileCheck %s --check-prefix=TLSDESC
+
+target triple = "x86_64--linux-gnu"
+
+declare void @effect()
+declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull)
+
+@foo_local = dso_local thread_local(localexec) global i32 0, align 4
+
+define i32 @func_local_tls(i32 %arg0, i64 %arg1) nounwind {
+; NOPIC-LABEL: func_local_tls:
+; NOPIC:       # %bb.0: # %entry
+; NOPIC-NEXT:    pushq %rbp
+; NOPIC-NEXT:    pushq %rbx
+; NOPIC-NEXT:    pushq %rax
+; NOPIC-NEXT:    movl %fs:foo_local@TPOFF, %ebp
+; NOPIC-NEXT:    testl %edi, %edi
+; NOPIC-NEXT:    movl %ebp, %eax
+; NOPIC-NEXT:    jne .LBB0_2
+; NOPIC-NEXT:  # %bb.1: # %if.then
+; NOPIC-NEXT:    movq %rsi, %rbx
+; NOPIC-NEXT:    callq effect@PLT
+; NOPIC-NEXT:    movl %fs:foo_local@TPOFF+168(,%rbx,4), %eax
+; NOPIC-NEXT:  .LBB0_2: # %if.end
+; NOPIC-NEXT:    addl %ebp, %eax
+; NOPIC-NEXT:    addq $8, %rsp
+; NOPIC-NEXT:    popq %rbx
+; NOPIC-NEXT:    popq %rbp
+; NOPIC-NEXT:    retq
+;
+; PIC-LABEL: func_local_tls:
+; PIC:       # %bb.0: # %entry
+; PIC-NEXT:    pushq %rbp
+; PIC-NEXT:    pushq %r14
+; PIC-NEXT:    pushq %rbx
+; PIC-NEXT:    movl %fs:.Lfoo_local$local@TPOFF, %ebp
+; PIC-NEXT:    testl %edi, %edi
+; PIC-NEXT:    movl %ebp, %eax
+; PIC-NEXT:    jne .LBB0_2
+; PIC-NEXT:  # %bb.1: # %if.then
+; PIC-NEXT:    movq %rsi, %rbx
+; PIC-NEXT:    movq %fs:0, %rax
+; PIC-NEXT:    leaq .Lfoo_local$local@TPOFF(%rax), %r14
+; PIC-NEXT:    callq effect@PLT
+; PIC-NEXT:    movl 168(%r14,%rbx,4), %eax
+; PIC-NEXT:  .LBB0_2: # %if.end
+; PIC-NEXT:    addl %ebp, %eax
+; PIC-NEXT:    popq %rbx
+; PIC-NEXT:    popq %r14
+; PIC-NEXT:    popq %rbp
+; PIC-NEXT:    retq
+;
+; TLSDESC-LABEL: func_local_tls:
+; TLSDESC:       # %bb.0: # %entry
+; TLSDESC-NEXT:    pushq %rbp
+; TLSDESC-NEXT:    pushq %r14
+; TLSDESC-NEXT:    pushq %rbx
+; TLSDESC-NEXT:    movl %fs:.Lfoo_local$local@TPOFF, %ebp
+; TLSDESC-NEXT:    testl %edi, %edi
+; TLSDESC-NEXT:    movl %ebp, %eax
+; TLSDESC-NEXT:    jne .LBB0_2
+; TLSDESC-NEXT:  # %bb.1: # %if.then
+; TLSDESC-NEXT:    movq %rsi, %rbx
+; TLSDESC-NEXT:    movq %fs:0, %rax
+; TLSDESC-NEXT:    leaq .Lfoo_local$local@TPOFF(%rax), %r14
+; TLSDESC-NEXT:    callq effect@PLT
+; TLSDESC-NEXT:    movl 168(%r14,%rbx,4), %eax
+; TLSDESC-NEXT:  .LBB0_2: # %if.end
+; TLSDESC-NEXT:    addl %ebp, %eax
+; TLSDESC-NEXT:    popq %rbx
+; TLSDESC-NEXT:    popq %r14
+; TLSDESC-NEXT:    popq %rbp
+; TLSDESC-NEXT:    retq
+entry:
+  %addr = tail call ptr @llvm.threadlocal.address.p0(ptr @foo_local)
+  %load0 = load i32, ptr %addr, align 4
+  %cond = icmp eq i32 %arg0, 0
+  br i1 %cond, label %if.then, label %if.end
+
+if.then:
+  tail call void @effect()
+  %x = add i64 %arg1, 42
+  %addr1 = getelementptr inbounds i32, ptr %addr, i64 %x
+  %load1 = load i32, ptr %addr1, align 4
+  br label %if.end
+
+if.end:
+  %phi = phi i32 [ %load1, %if.then ], [ %load0, %entry ]
+  %ret = add i32 %phi, %load0
+  ret i32 %ret
+}
+
+@foo_nonlocal = thread_local global i32 0, align 4
+
+define i32 @func_nonlocal_tls(i32 %arg0, i64 %arg1) nounwind {
+; NOPIC-LABEL: func_nonlocal_tls:
+; NOPIC:       # %bb.0: # %entry
+; NOPIC-NEXT:    pushq %rbp
+; NOPIC-NEXT:    pushq %r14
+; NOPIC-NEXT:    pushq %rbx
+; NOPIC-NEXT:    movq foo_nonlocal@GOTTPOFF(%rip), %r14
+; NOPIC-NEXT:    movl %fs:(%r14), %ebp
+; NOPIC-NEXT:    testl %edi, %edi
+; NOPIC-NEXT:    movl %ebp, %eax
+; NOPIC-NEXT:    jne .LBB1_2
+; NOPIC-NEXT:  # %bb.1: # %if.then
+; NOPIC-NEXT:    movq %rsi, %rbx
+; NOPIC-NEXT:    callq effect@PLT
+; NOPIC-NEXT:    movl %fs:168(%r14,%rbx,4), %eax
+; NOPIC-NEXT:  .LBB1_2: # %if.end
+; NOPIC-NEXT:    addl %ebp, %eax
+; NOPIC-NEXT:    popq %rbx
+; NOPIC-NEXT:    popq %r14
+; NOPIC-NEXT:    popq %rbp
+; NOPIC-NEXT:    retq
+;
+; PIC-LABEL: func_nonlocal_tls:
+; PIC:       # %bb.0: # %entry
+; PIC-NEXT:    pushq %rbp
+; PIC-NEXT:    pushq %r15
+; PIC-NEXT:    pushq %r14
+; PIC-NEXT:    pushq %rbx
+; PIC-NEXT:    pushq %rax
+; PIC-NEXT:    movq %rsi, %rbx
+; PIC-NEXT:    movl %edi, %ebp
+; PIC-NEXT:    data16
+; PIC-NEXT:    leaq foo_nonlocal@TLSGD(%rip), %rdi
+; PIC-NEXT:    data16
+; PIC-NEXT:    data16
+; PIC-NEXT:    rex64
+; PIC-NEXT:    callq __tls_get_addr@PLT
+; PIC-NEXT:    movq %rax, %r14
+; PIC-NEXT:    movl (%rax), %r15d
+; PIC-NEXT:    testl %ebp, %ebp
+; PIC-NEXT:    movl %r15d, %eax
+; PIC-NEXT:    jne .LBB1_2
+; PIC-NEXT:  # %bb.1: # %if.then
+; PIC-NEXT:    callq effect@PLT
+; PIC-NEXT:    movl 168(%r14,%rbx,4), %eax
+; PIC-NEXT:  .LBB1_2: # %if.end
+; PIC-NEXT:    addl %r15d, %eax
+; PIC-NEXT:    addq $8, %rsp
+; PIC-NEXT:    popq %rbx
+; PIC-NEXT:    popq %r14
+; PIC-NEXT:    popq %r15
+; PIC-NEXT:    popq %rbp
+; PIC-NEXT:    retq
+;
+; TLSDESC-LABEL: func_nonlocal_tls:
+; TLSDESC:       # %bb.0: # %entry
+; TLSDESC-NEXT:    pushq %rbp
+; TLSDESC-NEXT:    pushq %r14
+; TLSDESC-NEXT:    pushq %rbx
+; TLSDESC-NEXT:    leaq foo_nonlocal@tlsdesc(%rip), %rax
+; TLSDESC-NEXT:    callq *foo_nonlocal@tlscall(%rax)
+; TLSDESC-NEXT:    movl %fs:(%rax), %ebp
+; TLSDESC-NEXT:    testl %edi, %edi
+; TLSDESC-NEXT:    movl %ebp, %ecx
+; TLSDESC-NEXT:    jne .LBB1_2
+; TLSDESC-NEXT:  # %bb.1: # %if.then
+; TLSDESC-NEXT:    movq %rsi, %rbx
+; TLSDESC-NEXT:    addq %fs:0, %rax
+; TLSDESC-NEXT:    movq %rax, %r14
+; TLSDESC-NEXT:    callq effect@PLT
+; TLSDESC-NEXT:    movl 168(%r14,%rbx,4), %ecx
+; TLSDESC-NEXT:  .LBB1_2: # %if.end
+; TLSDESC-NEXT:    addl %ebp, %ecx
+; TLSDESC-NEXT:    movl %ecx, %eax
+; TLSDESC-NEXT:    popq %rbx
+; TLSDESC-NEXT:    popq %r14
+; TLSDESC-NEXT:    popq %rbp
+; TLSDESC-NEXT:    retq
+entry:
+  %addr = tail call ptr @llvm.threadlocal.address.p0(ptr @foo_nonlocal)
+  %load0 = load i32, ptr %addr, align 4
+  %cond = icmp eq i32 %arg0, 0
+  br i1 %cond, label %if.then, label %if.end
+
+if.then:
+  tail call void @effect()
+  %x = add i64 %arg1, 42
+  %addr1 = getelementptr inbounds i32, ptr %addr, i64 %x
+  %load1 = load i32, ptr %addr1, align 4
+  br label %if.end
+
+if.end:
+  %phi = phi i32 [ %load1, %if.then ], [ %load0, %entry ]
+  %ret = add i32 %phi, %load0
+  ret i32 %ret
+}
diff --git a/llvm/test/Transforms/CodeGenPrepare/X86/sink-addrmode-tls.ll b/llvm/test/Transforms/CodeGenPrepare/X86/sink-addrmode-tls.ll
new file mode 100644
index 0000000000000..080c807cbad13
--- /dev/null
+++ b/llvm/test/Transforms/CodeGenPrepare/X86/sink-addrmode-tls.ll
@@ -0,0 +1,86 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S -passes='require<profile-summary>,function(codegenprepare)' %s | FileCheck %s
+
+target triple = "x86_64--linux-gnu"
+
+@foo = dso_local thread_local(localexec) global i32 0, align 4
+
+declare void @effect()
+declare nonnull ptr @llvm.threadlocal.address.p0(ptr nonnull)
+
+define i32 @func0(i32 %arg) {
+; CHECK-LABEL: define i32 @func0(
+; CHECK-SAME: i32 [[ARG:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[ADDR:%.*]] = tail call ptr @llvm.threadlocal.address.p0(ptr @foo)
+; CHECK-NEXT:    [[LOAD0:%.*]] = load i32, ptr [[ADDR]], align 4
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[ARG]], 0
+; CHECK-NEXT:    br i1 [[COND]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    tail call void @effect()
+; CHECK-NEXT:    [[TMP0:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @foo)
+; CHECK-NEXT:    [[LOAD1:%.*]] = load i32, ptr [[TMP0]], align 4
+; CHECK-NEXT:    br label [[IF_END]]
+; CHECK:       if.end:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[LOAD1]], [[IF_THEN]] ], [ [[LOAD0]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[RET:%.*]] = add i32 [[PHI]], [[LOAD0]]
+; CHECK-NEXT:    ret i32 [[RET]]
+;
+entry:
+  %addr = tail call ptr @llvm.threadlocal.address.p0(ptr @foo)
+  %load0 = load i32, ptr %addr, align 4
+  %cond = icmp eq i32 %arg, 0
+  br i1 %cond, label %if.then, label %if.end
+
+if.then:
+  tail call void @effect()
+  %load1 = load i32, ptr %addr, align 4
+  br label %if.end
+
+if.end:
+  %phi = phi i32 [ %load1, %if.then ], [ %load0, %entry ]
+  %ret = add i32 %phi, %load0
+  ret i32 %ret
+}
+
+define i32 @func1(i32 %arg0, i32 %arg1) {
+; CHECK-LABEL: define i32 @func1(
+; CHECK-SAME: i32 [[ARG0:%.*]], i32 [[ARG1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[ADDR:%.*]] = tail call ptr @llvm.threadlocal.address.p0(ptr @foo)
+; CHECK-NEXT:    [[LOAD0:%.*]] = load i32, ptr [[ADDR]], align 4
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[ARG0]], 0
+; CHECK-NEXT:    br i1 [[COND]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    tail call void @effect()
+; CHECK-NEXT:    [[X:%.*]] = add i32 [[ARG1]], 42
+; CHECK-NEXT:    [[X64:%.*]] = sext i32 [[X]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @foo)
+; CHECK-NEXT:    [[SUNKADDR:%.*]] = mul i64 [[X64]], 4
+; CHECK-NEXT:    [[ADDR1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 [[SUNKADDR]]
+; CHECK-NEXT:    [[LOAD1:%.*]] = load i32, ptr [[ADDR1]], align 4
+; CHECK-NEXT:    br label [[IF_END]]
+; CHECK:       if.end:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[LOAD1]], [[IF_THEN]] ], [ [[LOAD0]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[RET:%.*]] = add i32 [[PHI]], [[LOAD0]]
+; CHECK-NEXT:    ret i32 [[RET]]
+;
+entry:
+  %addr = tail call ptr @llvm.threadlocal.address.p0(ptr @foo)
+  %load0 = load i32, ptr %addr, align 4
+  %cond = icmp eq i32 %arg0, 0
+  br i1 %cond, label %if.then, label %if.end
+
+if.then:
+  tail call void @effect()
+  %x = add i32 %arg1, 42
+  %x64 = sext i32 %x to i64
+  %addr1 = getelementptr inbounds i32, ptr %addr, i64 %x64
+  %load1 = load i32, ptr %addr1, align 4
+  br label %if.end
+
+if.end:
+  %phi = phi i32 [ %load1, %if.then ], [ %load0, %entry ]
+  %ret = add i32 %phi, %load0
+  ret i32 %ret
+}

From 2583b2eea4a509424e5e5f51dffedd9beede76a3 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Wed, 17 Apr 2024 19:46:43 +0000
Subject: [PATCH 790/886] [flang][NFC] Add missing include for FreeBSD

Suggested by dankm here: https://github.com/llvm/llvm-project/pull/88517
---
 flang/runtime/extensions.cpp | 1 +
 1 file changed, 1 insertion(+)

diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 12498b502ae1c..80604d886a701 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -21,6 +21,7 @@
 #include <ctime>
 #include <signal.h>
 #include <thread>
+#include <unistd.h>
 
 #ifdef _WIN32
 inline void CtimeBuffer(char *buffer, size_t bufsize, const time_t cur_time,

From d0c51f7d5496015ce410cc758a7caf976ffaaec7 Mon Sep 17 00:00:00 2001
From: Kai Nacke <kai.peter.nacke@ibm.com>
Date: Wed, 17 Apr 2024 16:07:35 -0400
Subject: [PATCH 791/886] [LiveIns] Improve recomputeLiveIns() (#88951)

Some small changes to recomputeLiveIns() to improve performance:

- Instead of copying the list of old live-ins, and then clearing
  them, a new method swaps the list for an empty one.
- getLiveIns() now returns a constant reference to the list

As result, the list-data is never copied. Depending on the
implementation
details of the vector container, it can also save calls to allocate
and deallocate memory.
I see a small improvement on CTMark with these changes.

---------

Co-authored-by: Nikita Popov <github@npopov.com>
---
 llvm/include/llvm/CodeGen/LivePhysRegs.h      | 9 +++++----
 llvm/include/llvm/CodeGen/MachineBasicBlock.h | 6 +++++-
 llvm/lib/CodeGen/MachineBasicBlock.cpp        | 6 ++++++
 3 files changed, 16 insertions(+), 5 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/LivePhysRegs.h b/llvm/include/llvm/CodeGen/LivePhysRegs.h
index 9574a6f0c7c00..d315e4ff6f3ab 100644
--- a/llvm/include/llvm/CodeGen/LivePhysRegs.h
+++ b/llvm/include/llvm/CodeGen/LivePhysRegs.h
@@ -199,14 +199,15 @@ void computeAndAddLiveIns(LivePhysRegs &LiveRegs,
 /// any changes were made.
 static inline bool recomputeLiveIns(MachineBasicBlock &MBB) {
   LivePhysRegs LPR;
-  auto oldLiveIns = MBB.getLiveIns();
+  std::vector<MachineBasicBlock::RegisterMaskPair> OldLiveIns;
 
-  MBB.clearLiveIns();
+  MBB.clearLiveIns(OldLiveIns);
   computeAndAddLiveIns(LPR, MBB);
   MBB.sortUniqueLiveIns();
 
-  auto newLiveIns = MBB.getLiveIns();
-  return oldLiveIns != newLiveIns;
+  const std::vector<MachineBasicBlock::RegisterMaskPair> &NewLiveIns =
+      MBB.getLiveIns();
+  return OldLiveIns != NewLiveIns;
 }
 
 /// Convenience function for recomputing live-in's for a set of MBBs until the
diff --git a/llvm/include/llvm/CodeGen/MachineBasicBlock.h b/llvm/include/llvm/CodeGen/MachineBasicBlock.h
index dc2035fa598c4..5b6be3a96b2fb 100644
--- a/llvm/include/llvm/CodeGen/MachineBasicBlock.h
+++ b/llvm/include/llvm/CodeGen/MachineBasicBlock.h
@@ -441,6 +441,10 @@ class MachineBasicBlock
   /// Clear live in list.
   void clearLiveIns();
 
+  /// Clear the live in list, and return the removed live in's in \p OldLiveIns.
+  /// Requires that the vector \p OldLiveIns is empty.
+  void clearLiveIns(std::vector<RegisterMaskPair> &OldLiveIns);
+
   /// Add PhysReg as live in to this block, and ensure that there is a copy of
   /// PhysReg to a virtual register of class RC. Return the virtual register
   /// that is a copy of the live in PhysReg.
@@ -477,7 +481,7 @@ class MachineBasicBlock
   /// Remove entry from the livein set and return iterator to the next.
   livein_iterator removeLiveIn(livein_iterator I);
 
-  std::vector<RegisterMaskPair> getLiveIns() const { return LiveIns; }
+  const std::vector<RegisterMaskPair> &getLiveIns() const { return LiveIns; }
 
   class liveout_iterator {
   public:
diff --git a/llvm/lib/CodeGen/MachineBasicBlock.cpp b/llvm/lib/CodeGen/MachineBasicBlock.cpp
index b2114c250ac09..0bd5f09564ec0 100644
--- a/llvm/lib/CodeGen/MachineBasicBlock.cpp
+++ b/llvm/lib/CodeGen/MachineBasicBlock.cpp
@@ -1728,6 +1728,12 @@ void MachineBasicBlock::clearLiveIns() {
   LiveIns.clear();
 }
 
+void MachineBasicBlock::clearLiveIns(
+    std::vector<RegisterMaskPair> &OldLiveIns) {
+  assert(OldLiveIns.empty() && "Vector must be empty");
+  std::swap(LiveIns, OldLiveIns);
+}
+
 MachineBasicBlock::livein_iterator MachineBasicBlock::livein_begin() const {
   assert(getParent()->getProperties().hasProperty(
       MachineFunctionProperties::Property::TracksLiveness) &&

From 60b90b523323f8196a9e4a68b1f33358624c09eb Mon Sep 17 00:00:00 2001
From: Zequan Wu <zequanwu@google.com>
Date: Wed, 17 Apr 2024 16:09:38 -0400
Subject: [PATCH 792/886] [lldb][DynamicLoader] Fix lldb unable to stop at
 _dl_debug_state if user set it before the process launched. (#88792)

If user sets a breakpoint at `_dl_debug_state` before the process
launched, the breakpoint is not resolved yet. When lldb loads dynamic
loader module, it's created with `Target::GetOrCreateModule` which
notifies any pending breakpoint to resolve. However, the module's
sections are not loaded at this time. They are loaded after returned
from
[Target::GetOrCreateModule](https://github.com/llvm/llvm-project/blob/0287a5cc4e2a5ded1ae2e4079f91052e6a6b8d9b/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp#L574-L577).
This change fixes it by manually resolving breakpoints after creating
dynamic loader module.
---
 .../POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp       | 11 +++++++++--
 .../breakpoint_command/TestBreakpointCommand.py | 17 +++++++++++++++++
 2 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp b/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp
index 9baf86da4dc79..9fa245fc41d40 100644
--- a/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp
+++ b/lldb/source/Plugins/DynamicLoader/POSIX-DYLD/DynamicLoaderPOSIXDYLD.cpp
@@ -571,10 +571,17 @@ ModuleSP DynamicLoaderPOSIXDYLD::LoadInterpreterModule() {
   FileSpec file(info.GetName().GetCString());
   ModuleSpec module_spec(file, target.GetArchitecture());
 
-  if (ModuleSP module_sp = target.GetOrCreateModule(module_spec,
-                                                    true /* notify */)) {
+  // Don't notify that module is added here because its loading section
+  // addresses are not updated yet. We manually notify it below.
+  if (ModuleSP module_sp =
+          target.GetOrCreateModule(module_spec, /*notify=*/false)) {
     UpdateLoadedSections(module_sp, LLDB_INVALID_ADDRESS, m_interpreter_base,
                          false);
+    // Manually notify that dynamic linker is loaded after updating load section
+    // addersses so that breakpoints can be resolved.
+    ModuleList module_list;
+    module_list.Append(module_sp);
+    target.ModulesDidLoad(module_list);
     m_interpreter_module = module_sp;
     return module_sp;
   }
diff --git a/lldb/test/API/functionalities/breakpoint/breakpoint_command/TestBreakpointCommand.py b/lldb/test/API/functionalities/breakpoint/breakpoint_command/TestBreakpointCommand.py
index ea242952e409e..850235fdcefa7 100644
--- a/lldb/test/API/functionalities/breakpoint/breakpoint_command/TestBreakpointCommand.py
+++ b/lldb/test/API/functionalities/breakpoint/breakpoint_command/TestBreakpointCommand.py
@@ -671,3 +671,20 @@ def test_breakpoint_statistics_hitcount(self):
         self.assertNotEqual(breakpoints_stats, None)
         for breakpoint_stats in breakpoints_stats:
             self.assertIn("hitCount", breakpoint_stats)
+
+    @skipIf(oslist=no_match(["linux"]))
+    def test_break_at__dl_debug_state(self):
+        """
+        Test lldb is able to stop at _dl_debug_state if it is set before the
+        process is launched.
+        """
+        self.build()
+        exe = self.getBuildArtifact("a.out")
+        self.runCmd("target create %s" % exe)
+        bpid = lldbutil.run_break_set_by_symbol(
+            self, "_dl_debug_state", num_expected_locations=0
+        )
+        self.runCmd("run")
+        self.assertIsNotNone(
+            lldbutil.get_one_thread_stopped_at_breakpoint_id(self.process(), bpid)
+        )

From e15f47f2675a5400464aec00219658882df5e3fa Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 17 Apr 2024 13:16:08 -0700
Subject: [PATCH 793/886] [InstCombine] Don't use dominating conditions to
 transform sub into xor. (#88566)

Other passes are unable to reverse this transform if we use dominating
conditions.

Fixes #88239.
---
 llvm/include/llvm/Analysis/SimplifyQuery.h          |  6 ++++++
 .../Transforms/InstCombine/InstCombineAddSub.cpp    |  6 ++++--
 llvm/test/Transforms/InstCombine/sub-xor.ll         |  4 +++-
 llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll   | 13 ++++++-------
 4 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/llvm/include/llvm/Analysis/SimplifyQuery.h b/llvm/include/llvm/Analysis/SimplifyQuery.h
index e5e6ae0d3d8e3..a10c0dc49fa22 100644
--- a/llvm/include/llvm/Analysis/SimplifyQuery.h
+++ b/llvm/include/llvm/Analysis/SimplifyQuery.h
@@ -113,6 +113,12 @@ struct SimplifyQuery {
     using namespace PatternMatch;
     return match(V, m_Undef());
   }
+
+  SimplifyQuery getWithoutDomCondCache() const {
+    SimplifyQuery Copy(*this);
+    Copy.DC = nullptr;
+    return Copy;
+  }
 };
 
 } // end namespace llvm
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
index c59b867b10e7d..df7f02810ee43 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -2281,8 +2281,10 @@ Instruction *InstCombinerImpl::visitSub(BinaryOperator &I) {
   if (match(Op0, m_APInt(Op0C))) {
     if (Op0C->isMask()) {
       // Turn this into a xor if LHS is 2^n-1 and the remaining bits are known
-      // zero.
-      KnownBits RHSKnown = computeKnownBits(Op1, 0, &I);
+      // zero. We don't use information from dominating conditions so this
+      // transform is easier to reverse if necessary.
+      KnownBits RHSKnown = llvm::computeKnownBits(
+          Op1, 0, SQ.getWithInstruction(&I).getWithoutDomCondCache());
       if ((*Op0C | RHSKnown.Zero).isAllOnes())
         return BinaryOperator::CreateXor(Op1, Op0);
     }
diff --git a/llvm/test/Transforms/InstCombine/sub-xor.ll b/llvm/test/Transforms/InstCombine/sub-xor.ll
index b4e87d0405fc4..2976598e043fe 100644
--- a/llvm/test/Transforms/InstCombine/sub-xor.ll
+++ b/llvm/test/Transforms/InstCombine/sub-xor.ll
@@ -158,13 +158,15 @@ define <2 x i8> @xor_add_splat_undef(<2 x i8> %x) {
   ret <2 x i8> %add
 }
 
+; Make sure we don't convert sub to xor using dominating condition. That makes
+; it hard for other passe to reverse.
 define i32 @xor_dominating_cond(i32 %x) {
 ; CHECK-LABEL: @xor_dominating_cond(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[COND:%.*]] = icmp ult i32 [[X:%.*]], 256
 ; CHECK-NEXT:    br i1 [[COND]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 ; CHECK:       if.then:
-; CHECK-NEXT:    [[A:%.*]] = xor i32 [[X]], 255
+; CHECK-NEXT:    [[A:%.*]] = sub nuw nsw i32 255, [[X]]
 ; CHECK-NEXT:    ret i32 [[A]]
 ; CHECK:       if.end:
 ; CHECK-NEXT:    ret i32 [[X]]
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll b/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll
index 3afa1904fb249..1c16b37144f1e 100644
--- a/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll
+++ b/llvm/test/Transforms/PhaseOrdering/X86/pr88239.ll
@@ -8,19 +8,18 @@ define void @foo(ptr noalias noundef %0, ptr noalias noundef %1) optsize {
 ; CHECK-LABEL: define void @foo(
 ; CHECK-SAME: ptr noalias nocapture noundef readonly [[TMP0:%.*]], ptr noalias nocapture noundef writeonly [[TMP1:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  vector.ph:
+; CHECK-NEXT:    [[INVARIANT_GEP:%.*]] = getelementptr i8, ptr [[TMP0]], i64 -28
 ; CHECK-NEXT:    br label [[TMP4:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[TMP2:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[TMP4]] ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>, [[TMP2]] ], [ [[VEC_IND_NEXT:%.*]], [[TMP4]] ]
-; CHECK-NEXT:    [[TMP6:%.*]] = and <8 x i64> [[VEC_IND]], <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <8 x i64> [[TMP6]], <i64 255, i64 255, i64 255, i64 255, i64 255, i64 255, i64 255, i64 255>
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], <8 x i64> [[TMP3]]
-; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = tail call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> [[TMP7]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i32> poison)
+; CHECK-NEXT:    [[TMP3:%.*]] = sub nuw nsw i64 255, [[INDVARS_IV]]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[INVARIANT_GEP]], i64 [[TMP3]]
+; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = load <8 x i32>, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_GATHER]], <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw i64 [[INDVARS_IV]], 8
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[MIDDLE_BLOCK:%.*]], label [[TMP4]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:

From f309c882ea5ad411ac1011a4228c0e49b1025bb4 Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Wed, 17 Apr 2024 20:11:35 +0000
Subject: [PATCH 794/886] Revert "[flang][NFC] Add missing include for FreeBSD"

This reverts commit 2583b2eea4a509424e5e5f51dffedd9beede76a3.
---
 flang/runtime/extensions.cpp | 1 -
 1 file changed, 1 deletion(-)

diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 80604d886a701..12498b502ae1c 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -21,7 +21,6 @@
 #include <ctime>
 #include <signal.h>
 #include <thread>
-#include <unistd.h>
 
 #ifdef _WIN32
 inline void CtimeBuffer(char *buffer, size_t bufsize, const time_t cur_time,

From 0cee89431d77d0bb0809fd9b2c9d21da2a2783aa Mon Sep 17 00:00:00 2001
From: Tom Eccles <tom.eccles@arm.com>
Date: Wed, 17 Apr 2024 19:46:43 +0000
Subject: [PATCH 795/886] [flang][NFC] Add missing include for FreeBSD

Suggested by dankm here: https://github.com/llvm/llvm-project/pull/88517
---
 flang/runtime/extensions.cpp | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/flang/runtime/extensions.cpp b/flang/runtime/extensions.cpp
index 12498b502ae1c..4b110cc10c840 100644
--- a/flang/runtime/extensions.cpp
+++ b/flang/runtime/extensions.cpp
@@ -43,9 +43,9 @@ inline void CtimeBuffer(char *buffer, size_t bufsize, const time_t cur_time,
 }
 #endif
 
-#if _REENTRANT || _POSIX_C_SOURCE >= 199506L
-// System is posix-compliant and has getlogin_r
-#include <unistd.h>
+#ifndef _WIN32
+// posix-compliant and has getlogin_r and F_OK
+#include <unistd.h> 
 #endif
 
 extern "C" {

From 9435edf628cb2011652897e1f10f7c55313d50be Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Wed, 17 Apr 2024 13:45:22 -0700
Subject: [PATCH 796/886] [flang][cuda] Lower DEALLOCATE for device variables
 (#89091)

Replace the runtime call to `AllocatableDeallocate` for CUDA device
variable to the newly added `fir.cuda_deallocate` operation.

This is similar with #88980

A third patch will handle the case of automatic dealloctaion of device
allocatable variables
---
 flang/lib/Lower/Allocatable.cpp            | 33 ++++++++++++++++++++--
 flang/test/Lower/CUDA/cuda-allocatable.cuf | 15 ++++++++++
 2 files changed, 45 insertions(+), 3 deletions(-)

diff --git a/flang/lib/Lower/Allocatable.cpp b/flang/lib/Lower/Allocatable.cpp
index 1d434d512d0c5..38f61528d7e28 100644
--- a/flang/lib/Lower/Allocatable.cpp
+++ b/flang/lib/Lower/Allocatable.cpp
@@ -799,6 +799,28 @@ static void postDeallocationAction(Fortran::lower::AbstractConverter &converter,
     Fortran::lower::attachDeclarePostDeallocAction(converter, builder, sym);
 }
 
+static mlir::Value genCudaDeallocate(fir::FirOpBuilder &builder,
+                                     mlir::Location loc,
+                                     const fir::MutableBoxValue &box,
+                                     ErrorManager &errorManager,
+                                     const Fortran::semantics::Symbol &sym) {
+  fir::CUDADataAttributeAttr cudaAttr =
+      Fortran::lower::translateSymbolCUDADataAttribute(builder.getContext(),
+                                                       sym);
+  mlir::Value errmsg =
+      mlir::isa<fir::AbsentOp>(errorManager.errMsgAddr.getDefiningOp())
+          ? nullptr
+          : errorManager.errMsgAddr;
+
+  // Keep return type the same as a standard AllocatableAllocate call.
+  mlir::Type retTy = fir::runtime::getModel<int>()(builder.getContext());
+  return builder
+      .create<fir::CUDADeallocateOp>(
+          loc, retTy, box.getAddr(), errmsg, cudaAttr,
+          errorManager.hasStatSpec() ? builder.getUnitAttr() : nullptr)
+      .getResult();
+}
+
 // Generate deallocation of a pointer/allocatable.
 static mlir::Value
 genDeallocate(fir::FirOpBuilder &builder,
@@ -806,10 +828,11 @@ genDeallocate(fir::FirOpBuilder &builder,
               const fir::MutableBoxValue &box, ErrorManager &errorManager,
               mlir::Value declaredTypeDesc = {},
               const Fortran::semantics::Symbol *symbol = nullptr) {
+  bool isCudaSymbol = symbol && Fortran::semantics::HasCUDAAttr(*symbol);
   // Deallocate intrinsic types inline.
   if (!box.isDerived() && !box.isPolymorphic() &&
       !box.isUnlimitedPolymorphic() && !errorManager.hasStatSpec() &&
-      !useAllocateRuntime && !box.isPointer()) {
+      !useAllocateRuntime && !box.isPointer() && !isCudaSymbol) {
     // Pointers must use PointerDeallocate so that their deallocations
     // can be validated.
     mlir::Value ret = fir::factory::genFreemem(builder, loc, box);
@@ -820,8 +843,12 @@ genDeallocate(fir::FirOpBuilder &builder,
   // Use runtime calls to deallocate descriptor cases. Sync MutableBoxValue
   // with its descriptor before and after calls if needed.
   errorManager.genStatCheck(builder, loc);
-  mlir::Value stat =
-      genRuntimeDeallocate(builder, loc, box, errorManager, declaredTypeDesc);
+  mlir::Value stat;
+  if (!isCudaSymbol)
+    stat =
+        genRuntimeDeallocate(builder, loc, box, errorManager, declaredTypeDesc);
+  else
+    stat = genCudaDeallocate(builder, loc, box, errorManager, *symbol);
   fir::factory::syncMutableBoxFromIRBox(builder, loc, box);
   if (symbol)
     postDeallocationAction(converter, builder, *symbol);
diff --git a/flang/test/Lower/CUDA/cuda-allocatable.cuf b/flang/test/Lower/CUDA/cuda-allocatable.cuf
index 55223011e8d9e..5b10334ecdbc1 100644
--- a/flang/test/Lower/CUDA/cuda-allocatable.cuf
+++ b/flang/test/Lower/CUDA/cuda-allocatable.cuf
@@ -5,6 +5,8 @@
 subroutine sub1()
   real, allocatable, device :: a(:)
   allocate(a(10))
+
+  deallocate(a)
 end subroutine
 
 ! CHECK-LABEL: func.func @_QPsub1()
@@ -13,10 +15,14 @@ end subroutine
 ! CHECK: fir.call @_FortranAAllocatableSetBounds
 ! CHECK: %{{.*}} = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<device>} -> i32
 
+! CHECK: %{{.*}} = fir.cuda_deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<device>} -> i32
+
 subroutine sub2()
   real, allocatable, managed :: a(:)
   integer :: istat
   allocate(a(10), stat=istat)
+
+  deallocate(a, stat=istat)
 end subroutine
 
 ! CHECK-LABEL: func.func @_QPsub2()
@@ -28,6 +34,9 @@ end subroutine
 ! CHECK: %[[STAT:.*]] = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<managed>, hasStat} -> i32
 ! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>
 
+! CHECK: %[[STAT:.*]] = fir.cuda_deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> {cuda_attr = #fir.cuda<managed>, hasStat} -> i32
+! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>
+
 subroutine sub3()
   integer, allocatable, pinned :: a(:,:)
   logical :: plog
@@ -92,6 +101,8 @@ subroutine sub7()
   integer :: istat
   character(50) :: err
   allocate(a(100), stat=istat, errmsg=err)
+
+  deallocate(a, stat=istat, errmsg=err)
 end subroutine
 
 ! CHECK-LABEL: func.func @_QPsub7()
@@ -105,3 +116,7 @@ end subroutine
 ! CHECK: fir.call @_FortranAAllocatableSetBounds
 ! CHECK: %[[STAT:.*]] = fir.cuda_allocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> errmsg(%[[ERR_BOX]] : !fir.box<!fir.char<1,50>>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
 ! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>
+
+! CHECK: %[[ERR_BOX:.*]] = fir.embox %[[ERR_DECL]]#1 : (!fir.ref<!fir.char<1,50>>) -> !fir.box<!fir.char<1,50>>
+! CHECK: %[[STAT:.*]] = fir.cuda_deallocate %[[BOX_DECL]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xf32>>>> errmsg(%15 : !fir.box<!fir.char<1,50>>) {cuda_attr = #fir.cuda<device>, hasStat} -> i32
+! CHECK: fir.store %[[STAT]] to %[[ISTAT_DECL]]#1 : !fir.ref<i32>

From 1b87418b9960983f54aa67f87018466ae12d0ec6 Mon Sep 17 00:00:00 2001
From: Keith Smiley <keithbsmiley@gmail.com>
Date: Wed, 17 Apr 2024 17:03:39 -0400
Subject: [PATCH 797/886] [bazel] Improve liblldb building (#89095)

On Linux using --version-script doesn't force loading of the underlying
archives that contain the symbols. By setting alwayslink=True on the API
cc_library we virtually get this behavior. This also allows downstream
users to use the exports files used by cmake. We could build more
configurability into this but there are also a lot of possible
variations users might want.
---
 .../llvm-project-overlay/lldb/BUILD.bazel     | 42 +++++++++++--------
 .../lldb/source/Plugins/BUILD.bazel           | 20 +--------
 2 files changed, 26 insertions(+), 36 deletions(-)

diff --git a/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel b/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
index 1f2b5b476bcc1..7958c6024875a 100644
--- a/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/lldb/BUILD.bazel
@@ -16,7 +16,11 @@ package(
 
 licenses(["notice"])
 
-exports_files(["LICENSE.TXT"])
+exports_files([
+    "LICENSE.TXT",
+    "source/API/liblldb-private.exports",
+    "source/API/liblldb.exports",
+])
 
 bool_flag(
     name = "enable_curses",
@@ -203,13 +207,29 @@ cc_library(
         ":TargetHeaders",
         ":Utility",
         ":Version",
+        "//lldb/source/Plugins:PluginExpressionParserClang",
         "//lldb/source/Plugins:PluginExpressionParserClangHeaders",
         "//lldb/source/Plugins:PluginsConfig",
+        "//llvm:AllTargetsDisassemblers",
         "//llvm:ExecutionEngine",
         "//llvm:MCJIT",
         "//llvm:Support",
         "//llvm:config",
-    ],
+    ] + [
+        "//lldb/source/Plugins:Plugin{}".format(x)
+        for x in DEFAULT_PLUGINS + DEFAULT_SCRIPT_PLUGINS
+    ] + select({
+        "@platforms//os:macos": [
+            "//lldb/source/Plugins:PluginProcessMacOSXKernel",
+            "//lldb/source/Plugins:PluginSymbolLocatorDebugSymbols",
+            "//lldb/source/Plugins:PluginSymbolVendorMacOSX",
+        ],
+        "@platforms//os:linux": [
+            "//lldb/source/Plugins:PluginProcessLinux",
+        ],
+        "//conditions:default": [],
+    }),
+    alwayslink = True,
 )
 
 cc_library(
@@ -690,23 +710,8 @@ cc_library(
     name = "liblldb.static",
     deps = [
         ":API",
-        ":Host",
         ":Interpreter",
-        "//llvm:AllTargetsDisassemblers",
-    ] + [
-        "//lldb/source/Plugins:Plugin{}".format(x)
-        for x in DEFAULT_PLUGINS + DEFAULT_SCRIPT_PLUGINS
-    ] + select({
-        "@platforms//os:macos": [
-            "//lldb/source/Plugins:PluginProcessMacOSXKernel",
-            "//lldb/source/Plugins:PluginSymbolLocatorDebugSymbols",
-            "//lldb/source/Plugins:PluginSymbolVendorMacOSX",
-        ],
-        "@platforms//os:linux": [
-            "//lldb/source/Plugins:PluginProcessLinux",
-        ],
-        "//conditions:default": [],
-    }),
+    ],
 )
 
 cc_shared_library(
@@ -889,6 +894,7 @@ cc_binary(
     deps = [
         ":Host",
         ":Initialization",
+        ":Interpreter",
         ":Utility",
         ":Version",
         ":lldb_server_opts_gen",
diff --git a/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel b/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
index b5f5bed1698a6..e0907a838148f 100644
--- a/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/lldb/source/Plugins/BUILD.bazel
@@ -557,7 +557,6 @@ cc_library(
         "//lldb:Breakpoint",
         "//lldb:Core",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Target",
         "//lldb:TargetHeaders",
@@ -591,7 +590,6 @@ cc_library(
         "//lldb:Core",
         "//lldb:Expression",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Target",
         "//lldb:TargetHeaders",
@@ -664,7 +662,7 @@ cc_library(
         "//lldb:Breakpoint",
         "//lldb:Core",
         "//lldb:Host",
-        "//lldb:Interpreter",
+        "//lldb:InterpreterHeaders",
         "//lldb:SymbolHeaders",
         "//lldb:Target",
         "//lldb:TargetHeaders",
@@ -1264,7 +1262,6 @@ cc_library(
         "//lldb:Breakpoint",
         "//lldb:Core",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1422,7 +1419,6 @@ cc_library(
     deps = [
         ":PluginProcessUtility",
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1438,7 +1434,6 @@ cc_library(
     deps = [
         ":PluginProcessUtility",
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1475,7 +1470,6 @@ cc_library(
         ":PluginProcessUtility",
         "//lldb:Core",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1495,7 +1489,6 @@ cc_library(
         ":PluginProcessUtility",
         "//lldb:Core",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1515,7 +1508,6 @@ cc_library(
     deps = [
         ":PluginProcessUtility",
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1532,7 +1524,6 @@ cc_library(
     deps = [
         ":PluginProcessUtility",
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1570,7 +1561,6 @@ cc_library(
         "//lldb:Expression",
         "//lldb:Headers",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1608,7 +1598,7 @@ cc_library(
         "//lldb:Core",
         "//lldb:Expression",
         "//lldb:Headers",
-        "//lldb:Interpreter",
+        "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
         "//lldb:Target",
@@ -1629,7 +1619,6 @@ cc_library(
         "//lldb:Expression",
         "//lldb:Headers",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1682,7 +1671,6 @@ cc_library(
         ":PluginObjectFileMachO",
         "//lldb:Breakpoint",
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -1838,7 +1826,6 @@ cc_library(
     include_prefix = "Plugins",
     deps = [
         "//lldb:Core",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Utility",
         "//llvm:Support",
@@ -2215,7 +2202,6 @@ cc_library(
         "//lldb:Core",
         "//lldb:Headers",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -2262,7 +2248,6 @@ cc_library(
         "//lldb:DataFormattersHeaders",
         "//lldb:ExpressionHeaders",
         "//lldb:Headers",
-        "//lldb:Interpreter",
         "//lldb:InterpreterHeaders",
         "//lldb:Symbol",
         "//lldb:SymbolHeaders",
@@ -2287,7 +2272,6 @@ cc_library(
         "//lldb:Expression",
         "//lldb:Headers",
         "//lldb:Host",
-        "//lldb:Interpreter",
         "//lldb:Symbol",
         "//lldb:Target",
         "//lldb:TargetHeaders",

From 0ab3f160c4bff1c7d57c046b95ab8c5035ae986f Mon Sep 17 00:00:00 2001
From: Philip Reames <preames@rivosinc.com>
Date: Wed, 17 Apr 2024 14:23:38 -0700
Subject: [PATCH 798/886] [RISCV] Add coverage of add (mul X, C), Y oppurtunity
 using shNadd

---
 llvm/test/CodeGen/RISCV/rv64zba.ll | 73 ++++++++++++++++++++++++++++++
 1 file changed, 73 insertions(+)

diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index 6939185947f44..ccb23fc2bbfa3 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -479,6 +479,18 @@ define i64 @addmul20(i64 %a, i64 %b) {
   ret i64 %d
 }
 
+define i64 @addmul22(i64 %a, i64 %b) {
+; CHECK-LABEL: addmul22:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a2, 22
+; CHECK-NEXT:    mul a0, a0, a2
+; CHECK-NEXT:    add a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 22
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
 define i64 @addmul24(i64 %a, i64 %b) {
 ; RV64I-LABEL: addmul24:
 ; RV64I:       # %bb.0:
@@ -551,6 +563,67 @@ define i64 @addmul72(i64 %a, i64 %b) {
   ret i64 %d
 }
 
+define i64 @addmul162(i64 %a, i64 %b) {
+; CHECK-LABEL: addmul162:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a2, 162
+; CHECK-NEXT:    mul a0, a0, a2
+; CHECK-NEXT:    add a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 162
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul180(i64 %a, i64 %b) {
+; CHECK-LABEL: addmul180:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a2, 180
+; CHECK-NEXT:    mul a0, a0, a2
+; CHECK-NEXT:    add a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 180
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @add255mul180(i64 %a) {
+; CHECK-LABEL: add255mul180:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a1, 180
+; CHECK-NEXT:    mul a0, a0, a1
+; CHECK-NEXT:    addi a0, a0, 255
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 180
+  %d = add i64 %c, 255
+  ret i64 %d
+}
+
+
+define i64 @addmul4096(i64 %a, i64 %b) {
+; CHECK-LABEL: addmul4096:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    slli a0, a0, 12
+; CHECK-NEXT:    add a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 4096
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
+define i64 @addmul4230(i64 %a, i64 %b) {
+; CHECK-LABEL: addmul4230:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    lui a2, 1
+; CHECK-NEXT:    addiw a2, a2, 134
+; CHECK-NEXT:    mul a0, a0, a2
+; CHECK-NEXT:    add a0, a0, a1
+; CHECK-NEXT:    ret
+  %c = mul i64 %a, 4230
+  %d = add i64 %c, %b
+  ret i64 %d
+}
+
 define i64 @mul96(i64 %a) {
 ; RV64I-LABEL: mul96:
 ; RV64I:       # %bb.0:

From 678f19f08296fec299438130cf5943714c590b7e Mon Sep 17 00:00:00 2001
From: azhan92 <alisonxzhang@gmail.com>
Date: Wed, 17 Apr 2024 17:56:42 -0400
Subject: [PATCH 799/886] [Support] Report EISDIR when opening a directory
 (#79880)

The test `llvm/unittests/Support/CommandLineTest.cpp` that handles
errors in expansion of response files was previously disabled for AIX.
Originally the code was dependent on `read` returning `EISDIR` which
occurs on platforms such as Linux. However, other platforms such as AIX
allow use of `read` on file descriptors for directories. This change
updates `readNativeFile` to produce `EISDIR` on AIX and z/OS when used
on a directory (instead of relying on the call to `read` to do so).

---------

Co-authored-by: Alison Zhang <alisonzhang@ibm.com>
Co-authored-by: James Henderson <46713263+jh7370@users.noreply.github.com>
---
 llvm/lib/Support/Unix/Path.inc                |  9 ++++++++
 .../tools/llvm-symbolizer/input-file-err.test |  2 --
 llvm/unittests/Support/CommandLineTest.cpp    |  2 --
 llvm/unittests/Support/Path.cpp               | 22 +++++++++++++++++++
 4 files changed, 31 insertions(+), 4 deletions(-)

diff --git a/llvm/lib/Support/Unix/Path.inc b/llvm/lib/Support/Unix/Path.inc
index 968e2c459f3fb..6e679f74869f0 100644
--- a/llvm/lib/Support/Unix/Path.inc
+++ b/llvm/lib/Support/Unix/Path.inc
@@ -1191,6 +1191,15 @@ Expected<size_t> readNativeFile(file_t FD, MutableArrayRef<char> Buf) {
   ssize_t NumRead = sys::RetryAfterSignal(-1, ::read, FD, Buf.data(), Size);
   if (ssize_t(NumRead) == -1)
     return errorCodeToError(errnoAsErrorCode());
+// The underlying operation on these platforms allow opening directories
+// for reading in more cases than other platforms.
+#if defined(__MVS__) || defined(_AIX)
+  struct stat Status;
+  if (fstat(FD, &Status) == -1)
+    return errorCodeToError(errnoAsErrorCode());
+  if (S_ISDIR(Status.st_mode))
+    return errorCodeToError(make_error_code(errc::is_a_directory));
+#endif
   return NumRead;
 }
 
diff --git a/llvm/test/tools/llvm-symbolizer/input-file-err.test b/llvm/test/tools/llvm-symbolizer/input-file-err.test
index df14da2f433c0..76115b513470b 100644
--- a/llvm/test/tools/llvm-symbolizer/input-file-err.test
+++ b/llvm/test/tools/llvm-symbolizer/input-file-err.test
@@ -1,5 +1,3 @@
-Failing on AIX due to D153595. The test expects a different error message from the one given on AIX.
-XFAIL: target={{.*}}-aix{{.*}}
 RUN: not llvm-addr2line -e %p/Inputs/nonexistent 0x12 2>&1 | FileCheck %s --check-prefix=CHECK-NONEXISTENT-A2L -DMSG=%errc_ENOENT
 RUN: not llvm-addr2line -e %p/Inputs/nonexistent 2>&1 | FileCheck %s --check-prefix=CHECK-NONEXISTENT-A2L -DMSG=%errc_ENOENT
 CHECK-NONEXISTENT-A2L: llvm-addr2line{{.*}}: error: '{{.*}}Inputs/nonexistent': [[MSG]]
diff --git a/llvm/unittests/Support/CommandLineTest.cpp b/llvm/unittests/Support/CommandLineTest.cpp
index bbeb9d5dc19bd..23f6081cd32a4 100644
--- a/llvm/unittests/Support/CommandLineTest.cpp
+++ b/llvm/unittests/Support/CommandLineTest.cpp
@@ -1117,7 +1117,6 @@ TEST(CommandLineTest, BadResponseFile) {
   ASSERT_STREQ(Argv[0], "clang");
   ASSERT_STREQ(Argv[1], AFileExp.c_str());
 
-#if !defined(_AIX) && !defined(__MVS__)
   std::string ADirExp = std::string("@") + std::string(ADir.path());
   Argv = {"clang", ADirExp.c_str()};
   Res = cl::ExpandResponseFiles(Saver, cl::TokenizeGNUCommandLine, Argv);
@@ -1125,7 +1124,6 @@ TEST(CommandLineTest, BadResponseFile) {
   ASSERT_EQ(2U, Argv.size());
   ASSERT_STREQ(Argv[0], "clang");
   ASSERT_STREQ(Argv[1], ADirExp.c_str());
-#endif
 }
 
 TEST(CommandLineTest, SetDefaultValue) {
diff --git a/llvm/unittests/Support/Path.cpp b/llvm/unittests/Support/Path.cpp
index 837ca03216f87..bdae7a8ee4b55 100644
--- a/llvm/unittests/Support/Path.cpp
+++ b/llvm/unittests/Support/Path.cpp
@@ -1757,6 +1757,28 @@ TEST_F(FileSystemTest, OpenFileForRead) {
 #endif
 }
 
+TEST_F(FileSystemTest, OpenDirectoryAsFileForRead) {
+  std::string Buf(5, '?');
+  Expected<fs::file_t> FD = fs::openNativeFileForRead(TestDirectory);
+#ifdef _WIN32
+  EXPECT_EQ(errorToErrorCode(FD.takeError()), errc::is_a_directory);
+#else
+  ASSERT_THAT_EXPECTED(FD, Succeeded());
+  auto Close = make_scope_exit([&] { fs::closeFile(*FD); });
+  Expected<size_t> BytesRead =
+      fs::readNativeFile(*FD, MutableArrayRef(&*Buf.begin(), Buf.size()));
+  EXPECT_EQ(errorToErrorCode(BytesRead.takeError()), errc::is_a_directory);
+#endif
+}
+
+TEST_F(FileSystemTest, OpenDirectoryAsFileForWrite) {
+  int FD;
+  std::error_code EC = fs::openFileForWrite(Twine(TestDirectory), FD);
+  if (!EC)
+    ::close(FD);
+  EXPECT_EQ(EC, errc::is_a_directory);
+}
+
 static void createFileWithData(const Twine &Path, bool ShouldExistBefore,
                                fs::CreationDisposition Disp, StringRef Data) {
   int FD;

From 26101e8c50e07ea53869b0c7f343d76bfed61b02 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Valentin=20Clement=20=28=E3=83=90=E3=83=AC=E3=83=B3?=
 =?UTF-8?q?=E3=82=BF=E3=82=A4=E3=83=B3=20=E3=82=AF=E3=83=AC=E3=83=A1?=
 =?UTF-8?q?=E3=83=B3=29?= <clementval@gmail.com>
Date: Wed, 17 Apr 2024 15:42:09 -0700
Subject: [PATCH 800/886] [flang][cuda] Avoid crash by exiting the check if
 assignment is not usable (#89149)

In the presence of other semantic error, `GetAssignment` would return a
nullptr and therefore would make the rest of the check crash when trying
to collect symbols.

Exiting early when we have a nullptr so the compiler doesn't crash and
user can get the meaningful semantic error.
---
 flang/lib/Semantics/check-cuda.cpp | 4 ++++
 flang/test/Semantics/cuf11.cuf     | 8 ++++++++
 2 files changed, 12 insertions(+)

diff --git a/flang/lib/Semantics/check-cuda.cpp b/flang/lib/Semantics/check-cuda.cpp
index 2cb15437a235a..fb1ebadd37858 100644
--- a/flang/lib/Semantics/check-cuda.cpp
+++ b/flang/lib/Semantics/check-cuda.cpp
@@ -488,6 +488,10 @@ void CUDAChecker::Enter(const parser::AssignmentStmt &x) {
   }
 
   const evaluate::Assignment *assign{semantics::GetAssignment(x)};
+  if (!assign) {
+    return;
+  }
+
   int nbLhs{evaluate::GetNbOfCUDASymbols(assign->lhs)};
   int nbRhs{evaluate::GetNbOfCUDASymbols(assign->rhs)};
 
diff --git a/flang/test/Semantics/cuf11.cuf b/flang/test/Semantics/cuf11.cuf
index b915c7246ed94..de7ff29743242 100644
--- a/flang/test/Semantics/cuf11.cuf
+++ b/flang/test/Semantics/cuf11.cuf
@@ -22,3 +22,11 @@ subroutine sub1()
   ahost = adev + adev
 
 end subroutine
+
+logical function compare_h(a,b)
+!ERROR: Derived type 'h' not found
+  type(h) :: a, b
+!ERROR: 'a' is not an object of derived type; it is implicitly typed
+!ERROR: 'b' is not an object of derived type; it is implicitly typed
+  compare_h = (a%h .eq. b%h)
+end

From 823eb1a3252dd773f9c4d92093591f1b39ac27d4 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 17 Apr 2024 16:52:02 -0700
Subject: [PATCH 801/886] [SelectionDAG] Add some validation of
 (S/U)(ADD/SUB)O_CARRY nodes. (#89133)

---
 llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index ca0a95750ba8d..f3441c9df0f8c 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -9915,6 +9915,18 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, SDVTList VTList,
     }
     break;
   }
+  case ISD::SADDO_CARRY:
+  case ISD::UADDO_CARRY:
+  case ISD::SSUBO_CARRY:
+  case ISD::USUBO_CARRY:
+    assert(VTList.NumVTs == 2 && Ops.size() == 3 &&
+           "Invalid add/sub overflow op!");
+    assert(VTList.VTs[0].isInteger() && VTList.VTs[1].isInteger() &&
+           Ops[0].getValueType() == Ops[1].getValueType() &&
+           Ops[0].getValueType() == VTList.VTs[0] &&
+           Ops[2].getValueType().isInteger() &&
+           "Binary operator types must match!");
+    break;
   case ISD::SMUL_LOHI:
   case ISD::UMUL_LOHI: {
     assert(VTList.NumVTs == 2 && Ops.size() == 2 && "Invalid mul lo/hi op!");

From 156ab4d4fb06be93b0cfce675e4cf86d330d879c Mon Sep 17 00:00:00 2001
From: Qizhi Hu <836744285@qq.com>
Date: Thu, 18 Apr 2024 08:52:25 +0800
Subject: [PATCH 802/886] [Clang][Sema] set declaration invalid earlier to
 prevent crash in calculating record layout (#87173)

Try to fix https://github.com/llvm/llvm-project/issues/75221
This crash caused by calculating record layout which contains a field
declaration with dependent type. Make it invalid before it is a complete
definition to prevent this crash. Define a new scope type to record this
type alias and set the record declaration invalid when it is defined in
a type alias template.

Co-authored-by: huqizhi <836744285@qq.com>
---
 clang/docs/ReleaseNotes.rst      | 2 ++
 clang/include/clang/Sema/Scope.h | 6 ++++++
 clang/lib/Parse/ParseDeclCXX.cpp | 5 +++++
 clang/lib/Sema/SemaDecl.cpp      | 7 +++++++
 clang/test/SemaCXX/PR75221.cpp   | 6 ++++++
 5 files changed, 26 insertions(+)
 create mode 100644 clang/test/SemaCXX/PR75221.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index c19ad9fba58f3..6c51c2d1f483c 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -536,6 +536,8 @@ Bug Fixes to C++ Support
   Fixes (#GH70604), (#GH79754), (#GH84163), (#GH84425), (#GH86054), (#GH86398), and (#GH86399).
 - Fix a crash when deducing ``auto`` from an invalid dereference (#GH88329).
 - Fix a crash in requires expression with templated base class member function. Fixes (#GH84020).
+- Fix a crash caused by defined struct in a type alias template when the structure
+  has fields with dependent type. Fixes (#GH75221).
 
 Bug Fixes to AST Handling
 ^^^^^^^^^^^^^^^^^^^^^^^^^
diff --git a/clang/include/clang/Sema/Scope.h b/clang/include/clang/Sema/Scope.h
index 099c2739e8603..1752a25111a77 100644
--- a/clang/include/clang/Sema/Scope.h
+++ b/clang/include/clang/Sema/Scope.h
@@ -156,6 +156,9 @@ class Scope {
     /// This is the scope of an OpenACC Compute Construct, which restricts
     /// jumping into/out of it.
     OpenACCComputeConstructScope = 0x10000000,
+
+    /// This is a scope of type alias declaration.
+    TypeAliasScope = 0x20000000,
   };
 
 private:
@@ -580,6 +583,9 @@ class Scope {
   /// if/switch/while/for statement.
   bool isControlScope() const { return getFlags() & Scope::ControlScope; }
 
+  /// Determine whether this scope is a type alias scope.
+  bool isTypeAliasScope() const { return getFlags() & Scope::TypeAliasScope; }
+
   /// Returns if rhs has a higher scope depth than this.
   ///
   /// The caller is responsible for calling this only if one of the two scopes
diff --git a/clang/lib/Parse/ParseDeclCXX.cpp b/clang/lib/Parse/ParseDeclCXX.cpp
index 51fd64b2d01aa..8e0e868248293 100644
--- a/clang/lib/Parse/ParseDeclCXX.cpp
+++ b/clang/lib/Parse/ParseDeclCXX.cpp
@@ -799,6 +799,11 @@ Parser::DeclGroupPtrTy Parser::ParseUsingDeclaration(
     ProhibitAttributes(PrefixAttrs);
 
     Decl *DeclFromDeclSpec = nullptr;
+    Scope *CurScope = getCurScope();
+    if (CurScope)
+      CurScope->setFlags(Scope::ScopeFlags::TypeAliasScope |
+                         CurScope->getFlags());
+
     Decl *AD = ParseAliasDeclarationAfterDeclarator(
         TemplateInfo, UsingLoc, D, DeclEnd, AS, Attrs, &DeclFromDeclSpec);
     return Actions.ConvertDeclToDeclGroup(AD, DeclFromDeclSpec);
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 1bde99d6fce74..356abe09a5ca6 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -19529,6 +19529,13 @@ void Sema::ActOnFields(Scope *S, SourceLocation RecLoc, Decl *EnclosingDecl,
   // Okay, we successfully defined 'Record'.
   if (Record) {
     bool Completed = false;
+    if (S) {
+      Scope *Parent = S->getParent();
+      if (Parent && Parent->isTypeAliasScope() &&
+          Parent->isTemplateParamScope())
+        Record->setInvalidDecl();
+    }
+
     if (CXXRecord) {
       if (!CXXRecord->isInvalidDecl()) {
         // Set access bits correctly on the directly-declared conversions.
diff --git a/clang/test/SemaCXX/PR75221.cpp b/clang/test/SemaCXX/PR75221.cpp
new file mode 100644
index 0000000000000..b342e347c5606
--- /dev/null
+++ b/clang/test/SemaCXX/PR75221.cpp
@@ -0,0 +1,6 @@
+// RUN: %clang_cc1 -verify -std=c++11 -fsyntax-only %s
+
+template <class T> using foo = struct foo { // expected-error {{'foo' cannot be defined in a type alias template}}
+  T size = 0;
+};
+foo a;

From 888836930b1ed30b2095d45fcffc482560cb0e39 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 10:15:25 +0900
Subject: [PATCH 803/886] Revert "[SLP]Attempt to vectorize long stores, if
 short one failed."

This reverts commit 6f7160eedb2db02f37d4ffd52fff7b0cf88b3fdc.

This still causes large compile-time regressions in some cases.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    | 106 ++++++------------
 .../Transforms/SLPVectorizer/X86/pr46983.ll   |  46 ++++++--
 2 files changed, 72 insertions(+), 80 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 806e8085038b3..7694627c3b043 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -15164,6 +15164,10 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
   BoUpSLP::ValueSet VectorizedStores;
   bool Changed = false;
 
+  // Stores the pair of stores (first_store, last_store) in a range, that were
+  // already tried to be vectorized. Allows to skip the store ranges that were
+  // already tried to be vectorized but the attempts were unsuccessful.
+  DenseSet<std::pair<Value *, Value *>> TriedSequences;
   struct StoreDistCompare {
     bool operator()(const std::pair<unsigned, int> &Op1,
                     const std::pair<unsigned, int> &Op2) const {
@@ -15205,10 +15209,8 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
       Type *ValueTy = StoreTy;
       if (auto *Trunc = dyn_cast<TruncInst>(Store->getValueOperand()))
         ValueTy = Trunc->getSrcTy();
-      unsigned MinVF = std::max<unsigned>(
-          2, PowerOf2Ceil(TTI->getStoreMinimumVF(
-                 R.getMinVF(DL->getTypeStoreSizeInBits(StoreTy)), StoreTy,
-                 ValueTy)));
+      unsigned MinVF = PowerOf2Ceil(TTI->getStoreMinimumVF(
+          R.getMinVF(DL->getTypeStoreSizeInBits(StoreTy)), StoreTy, ValueTy));
 
       if (MaxVF < MinVF) {
         LLVM_DEBUG(dbgs() << "SLP: Vectorization infeasible as MaxVF (" << MaxVF
@@ -15234,74 +15236,40 @@ bool SLPVectorizerPass::vectorizeStores(ArrayRef<StoreInst *> Stores,
         VF = Size > MaxVF ? NonPowerOf2VF : Size;
         Size *= 2;
       });
-      unsigned End = Operands.size();
-      unsigned Repeat = 0;
-      constexpr unsigned MaxAttempts = 2;
-      SmallBitVector Range(Operands.size());
-      while (true) {
-        ++Repeat;
-        for (unsigned Size : CandidateVFs) {
-          int StartIdx = Range.find_first_unset();
-          while (StartIdx != -1) {
-            int EndIdx = Range.find_next(StartIdx);
-            unsigned Sz = EndIdx == -1 ? End : EndIdx;
-            for (unsigned Cnt = StartIdx; Cnt + Size <= Sz;) {
-              ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
-              assert(all_of(Slice,
-                            [&](Value *V) {
-                              return cast<StoreInst>(V)
-                                         ->getValueOperand()
-                                         ->getType() ==
-                                     cast<StoreInst>(Slice.front())
-                                         ->getValueOperand()
-                                         ->getType();
-                            }) &&
-                     "Expected all operands of same type.");
-              if (vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
-                // Mark the vectorized stores so that we don't vectorize them
-                // again.
-                VectorizedStores.insert(Slice.begin(), Slice.end());
-                // Mark the vectorized stores so that we don't vectorize them
-                // again.
-                Changed = true;
-                // If we vectorized initial block, no need to try to vectorize
-                // it again.
-                Range.set(Cnt, Cnt + Size);
-                if (Cnt < StartIdx + MinVF)
-                  Range.set(StartIdx, Cnt);
-                if (Cnt > EndIdx - Size - MinVF) {
-                  Range.set(Cnt + Size, EndIdx);
-                  End = Cnt;
-                }
-                Cnt += Size;
-                continue;
-              }
-              ++Cnt;
-            }
-            if (Sz >= End)
-              break;
-            StartIdx = Range.find_next_unset(EndIdx);
+      unsigned StartIdx = 0;
+      for (unsigned Size : CandidateVFs) {
+        for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
+          ArrayRef<Value *> Slice = ArrayRef(Operands).slice(Cnt, Size);
+          assert(
+              all_of(
+                  Slice,
+                  [&](Value *V) {
+                    return cast<StoreInst>(V)->getValueOperand()->getType() ==
+                           cast<StoreInst>(Slice.front())
+                               ->getValueOperand()
+                               ->getType();
+                  }) &&
+              "Expected all operands of same type.");
+          if (!VectorizedStores.count(Slice.front()) &&
+              !VectorizedStores.count(Slice.back()) &&
+              TriedSequences.insert(std::make_pair(Slice.front(), Slice.back()))
+                  .second &&
+              vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
+            // Mark the vectorized stores so that we don't vectorize them again.
+            VectorizedStores.insert(Slice.begin(), Slice.end());
+            Changed = true;
+            // If we vectorized initial block, no need to try to vectorize it
+            // again.
+            if (Cnt == StartIdx)
+              StartIdx += Size;
+            Cnt += Size;
+            continue;
           }
+          ++Cnt;
         }
-        // All values vectorize - exit.
-        if (Range.all())
-          break;
-        // Check if tried all attempts or no need for the last attempts at all.
-        if (Repeat >= MaxAttempts)
-          break;
-        constexpr unsigned MaxVFScale = 4;
-        constexpr unsigned StoresLimit = 16;
-        const unsigned MaxTotalNum = std::min(
-            std::max<unsigned>(StoresLimit, MaxVFScale * MaxVF),
-            bit_floor(static_cast<unsigned>(Range.find_last_unset() -
-                                            Range.find_first_unset() + 1)));
-        if (MaxVF >= MaxTotalNum)
+        // Check if the whole array was vectorized already - exit.
+        if (StartIdx >= Operands.size())
           break;
-        // Last attempt to vectorize max number of elements, if all previous
-        // attempts were unsuccessful because of the cost issues.
-        CandidateVFs.clear();
-        for (unsigned Size = MaxTotalNum; Size > MaxVF; Size /= 2)
-          CandidateVFs.push_back(Size);
       }
     }
   };
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
index 3deab0975ce76..75505f632a43f 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
@@ -100,17 +100,41 @@ define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
 define void @store_i64(ptr nocapture %0, i32 %1, i32 %2) {
 ; SSE-LABEL: @store_i64(
 ; SSE-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP1:%.*]] to i64
-; SSE-NEXT:    [[TMP5:%.*]] = load <4 x i64>, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
-; SSE-NEXT:    [[TMP6:%.*]] = insertelement <4 x i64> poison, i64 [[TMP4]], i64 0
-; SSE-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP8:%.*]] = mul <4 x i64> [[TMP5]], [[TMP7]]
-; SSE-NEXT:    [[TMP9:%.*]] = lshr <4 x i64> [[TMP8]], <i64 15, i64 15, i64 15, i64 15>
-; SSE-NEXT:    [[TMP10:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
-; SSE-NEXT:    [[TMP11:%.*]] = icmp ult <4 x i32> [[TMP10]], <i32 255, i32 255, i32 255, i32 255>
-; SSE-NEXT:    [[TMP12:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
-; SSE-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>
-; SSE-NEXT:    [[TMP14:%.*]] = zext <4 x i32> [[TMP13]] to <4 x i64>
-; SSE-NEXT:    store <4 x i64> [[TMP14]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP5:%.*]] = load i64, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
+; SSE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], [[TMP4]]
+; SSE-NEXT:    [[TMP7:%.*]] = lshr i64 [[TMP6]], 15
+; SSE-NEXT:    [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
+; SSE-NEXT:    [[TMP9:%.*]] = icmp ult i32 [[TMP8]], 255
+; SSE-NEXT:    [[TMP10:%.*]] = and i64 [[TMP7]], 4294967295
+; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i64 [[TMP10]], i64 255
+; SSE-NEXT:    store i64 [[TMP11]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 8
+; SSE-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], [[TMP4]]
+; SSE-NEXT:    [[TMP15:%.*]] = lshr i64 [[TMP14]], 15
+; SSE-NEXT:    [[TMP16:%.*]] = trunc i64 [[TMP15]] to i32
+; SSE-NEXT:    [[TMP17:%.*]] = icmp ult i32 [[TMP16]], 255
+; SSE-NEXT:    [[TMP18:%.*]] = and i64 [[TMP15]], 4294967295
+; SSE-NEXT:    [[TMP19:%.*]] = select i1 [[TMP17]], i64 [[TMP18]], i64 255
+; SSE-NEXT:    store i64 [[TMP19]], ptr [[TMP12]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 16
+; SSE-NEXT:    [[TMP21:%.*]] = load i64, ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], [[TMP4]]
+; SSE-NEXT:    [[TMP23:%.*]] = lshr i64 [[TMP22]], 15
+; SSE-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
+; SSE-NEXT:    [[TMP25:%.*]] = icmp ult i32 [[TMP24]], 255
+; SSE-NEXT:    [[TMP26:%.*]] = and i64 [[TMP23]], 4294967295
+; SSE-NEXT:    [[TMP27:%.*]] = select i1 [[TMP25]], i64 [[TMP26]], i64 255
+; SSE-NEXT:    store i64 [[TMP27]], ptr [[TMP20]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 24
+; SSE-NEXT:    [[TMP29:%.*]] = load i64, ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
+; SSE-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], [[TMP4]]
+; SSE-NEXT:    [[TMP31:%.*]] = lshr i64 [[TMP30]], 15
+; SSE-NEXT:    [[TMP32:%.*]] = trunc i64 [[TMP31]] to i32
+; SSE-NEXT:    [[TMP33:%.*]] = icmp ult i32 [[TMP32]], 255
+; SSE-NEXT:    [[TMP34:%.*]] = and i64 [[TMP31]], 4294967295
+; SSE-NEXT:    [[TMP35:%.*]] = select i1 [[TMP33]], i64 [[TMP34]], i64 255
+; SSE-NEXT:    store i64 [[TMP35]], ptr [[TMP28]], align 8, !tbaa [[TBAA5]]
 ; SSE-NEXT:    ret void
 ;
 ; AVX-LABEL: @store_i64(

From 748ef7eccc8b997ee5553cc5aadf04d7c7d0556f Mon Sep 17 00:00:00 2001
From: "Yaxun (Sam) Liu" <yaxun.liu@amd.com>
Date: Wed, 17 Apr 2024 21:44:12 -0400
Subject: [PATCH 804/886] [CUDA][HIP] Fix record layout on Windows (#87651)

On windows, record layout should be consistent with host side, otherwise
host code is not able to access fields of the record correctly.

Fixes: https://github.com/llvm/llvm-project/issues/51031

Fixes: SWDEV-446010
---
 clang/lib/AST/RecordLayoutBuilder.cpp   |   5 +
 clang/test/CodeGenCUDA/record-layout.cu | 200 ++++++++++++++++++++++++
 2 files changed, 205 insertions(+)
 create mode 100644 clang/test/CodeGenCUDA/record-layout.cu

diff --git a/clang/lib/AST/RecordLayoutBuilder.cpp b/clang/lib/AST/RecordLayoutBuilder.cpp
index a3b7431f7ffd6..d9bf62c2bbb04 100644
--- a/clang/lib/AST/RecordLayoutBuilder.cpp
+++ b/clang/lib/AST/RecordLayoutBuilder.cpp
@@ -2458,6 +2458,11 @@ static bool mustSkipTailPadding(TargetCXXABI ABI, const CXXRecordDecl *RD) {
 }
 
 static bool isMsLayout(const ASTContext &Context) {
+  // Check if it's CUDA device compilation; ensure layout consistency with host.
+  if (Context.getLangOpts().CUDA && Context.getLangOpts().CUDAIsDevice &&
+      Context.getAuxTargetInfo())
+    return Context.getAuxTargetInfo()->getCXXABI().isMicrosoft();
+
   return Context.getTargetInfo().getCXXABI().isMicrosoft();
 }
 
diff --git a/clang/test/CodeGenCUDA/record-layout.cu b/clang/test/CodeGenCUDA/record-layout.cu
new file mode 100644
index 0000000000000..dd34121ccb9d3
--- /dev/null
+++ b/clang/test/CodeGenCUDA/record-layout.cu
@@ -0,0 +1,200 @@
+// RUN: %clang_cc1 -triple x86_64-pc-windows-msvc -fdump-record-layouts \
+// RUN:   -emit-llvm -o %t -xhip %s 2>&1 | FileCheck %s --check-prefix=AST
+// RUN: cat %t | FileCheck --check-prefixes=CHECK,HOST %s
+// RUN: %clang_cc1 -fcuda-is-device -triple amdgcn-amd-amdhsa -target-cpu gfx1100 \
+// RUN:   -emit-llvm -fdump-record-layouts -aux-triple x86_64-pc-windows-msvc \
+// RUN:   -o %t -xhip %s | FileCheck %s --check-prefix=AST
+// RUN: cat %t | FileCheck --check-prefixes=CHECK,DEV %s
+
+#include "Inputs/cuda.h"
+
+// AST: *** Dumping AST Record Layout
+// AST-LABEL:         0 | struct C
+// AST-NEXT:          0 |   struct A (base) (empty)
+// AST-NEXT:          1 |   struct B (base) (empty)
+// AST-NEXT:          4 |   int i
+// AST-NEXT:            | [sizeof=8, align=4,
+// AST-NEXT:            |  nvsize=8, nvalign=4]
+
+// CHECK: %struct.C = type { [4 x i8], i32 }
+
+struct A {};
+struct B {};
+struct C : A, B {
+    int i;
+};
+
+// AST: *** Dumping AST Record Layout
+// AST-LABEL:          0 | struct I
+// AST-NEXT:           0 |   (I vftable pointer)
+// AST-NEXT:           8 |   int i
+// AST-NEXT:             | [sizeof=16, align=8,
+// AST-NEXT:             |  nvsize=16, nvalign=8]
+
+// AST: *** Dumping AST Record Layout
+// AST-LABEL:          0 | struct J
+// AST-NEXT:           0 |   struct I (primary base)
+// AST-NEXT:           0 |     (I vftable pointer)
+// AST-NEXT:           8 |     int i
+// AST-NEXT:          16 |   int j
+// AST-NEXT:             | [sizeof=24, align=8,
+// AST-NEXT:             |  nvsize=24, nvalign=8]
+
+// CHECK: %struct.I = type { ptr, i32 }
+// CHECK: %struct.J = type { %struct.I, i32 }
+
+// HOST: @0 = private unnamed_addr constant { [4 x ptr] } { [4 x ptr] [ptr @"??_R4J@@6B@", ptr @"?f@J@@UEAAXXZ", ptr null, ptr @"?h@J@@UEAAXXZ"] }, comdat($"??_7J@@6B@")
+// HOST: @1 = private unnamed_addr constant { [4 x ptr] } { [4 x ptr] [ptr @"??_R4I@@6B@", ptr @_purecall, ptr null, ptr @_purecall] }, comdat($"??_7I@@6B@")
+// HOST: @"??_7J@@6B@" = unnamed_addr alias ptr, getelementptr inbounds ({ [4 x ptr] }, ptr @0, i32 0, i32 0, i32 1)
+// HOST: @"??_7I@@6B@" = unnamed_addr alias ptr, getelementptr inbounds ({ [4 x ptr] }, ptr @1, i32 0, i32 0, i32 1)
+
+// DEV: @_ZTV1J = linkonce_odr unnamed_addr addrspace(1) constant { [5 x ptr addrspace(1)] } { [5 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @_ZN1J1gEv to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @_ZN1J1hEv to ptr addrspace(1))] }, comdat, align 8
+// DEV: @_ZTV1I = linkonce_odr unnamed_addr addrspace(1) constant { [5 x ptr addrspace(1)] } { [5 x ptr addrspace(1)] [ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) null, ptr addrspace(1) addrspacecast (ptr @__cxa_pure_virtual to ptr addrspace(1)), ptr addrspace(1) addrspacecast (ptr @__cxa_pure_virtual to ptr addrspace(1))] }, comdat, align 8
+struct I {
+    virtual void f() = 0;
+    __device__ virtual void g() = 0;
+    __device__ __host__ virtual void h() = 0;
+    int i;
+};
+
+struct J : I {
+    void f() override {}
+    __device__ void g() override {}
+    __device__ __host__ void h() override {}
+    int j;
+};
+
+// DEV: define dso_local amdgpu_kernel void @_Z8C_kernel1C(ptr addrspace(4) noundef byref(%struct.C) align 4 %0)
+// DEV:  %coerce = alloca %struct.C, align 4, addrspace(5)
+// DEV:  %c = addrspacecast ptr addrspace(5) %coerce to ptr
+// DEV:  call void @llvm.memcpy.p0.p4.i64(ptr align 4 %c, ptr addrspace(4) align 4 %0, i64 8, i1 false)
+// DEV:  %i = getelementptr inbounds %struct.C, ptr %c, i32 0, i32 1
+// DEV:  store i32 1, ptr %i, align 4
+
+__global__ void C_kernel(C c)
+{
+  c.i = 1;
+}
+
+// HOST-LABEL: define dso_local void @"?test_C@@YAXXZ"()
+// HOST:  %c = alloca %struct.C, align 4
+// HOST:  %i = getelementptr inbounds %struct.C, ptr %c, i32 0, i32 1
+// HOST:  store i32 11, ptr %i, align 4
+
+void test_C() {
+  C c;
+  c.i = 11;
+  C_kernel<<<1, 1>>>(c);
+}
+
+// DEV: define dso_local void @_Z5J_devP1J(ptr noundef %j)
+// DEV:  %j.addr = alloca ptr, align 8, addrspace(5)
+// DEV:  %j.addr.ascast = addrspacecast ptr addrspace(5) %j.addr to ptr
+// DEV:  store ptr %j, ptr %j.addr.ascast, align 8
+// DEV:  %0 = load ptr, ptr %j.addr.ascast, align 8
+// DEV:  %i = getelementptr inbounds %struct.I, ptr %0, i32 0, i32 1
+// DEV:  store i32 2, ptr %i, align 8
+// DEV:  %1 = load ptr, ptr %j.addr.ascast, align 8
+// DEV:  %j1 = getelementptr inbounds %struct.J, ptr %1, i32 0, i32 1
+// DEV:  store i32 3, ptr %j1, align 8
+// DEV:  %2 = load ptr, ptr %j.addr.ascast, align 8
+// DEV:  %vtable = load ptr addrspace(1), ptr %2, align 8
+// DEV:  %vfn = getelementptr inbounds ptr addrspace(1), ptr addrspace(1) %vtable, i64 1
+// DEV:  %3 = load ptr addrspace(1), ptr addrspace(1) %vfn, align 8
+// DEV:  call addrspace(1) void %3(ptr noundef nonnull align 8 dereferenceable(24) %2)
+// DEV:  %4 = load ptr, ptr %j.addr.ascast, align 8
+// DEV:  %vtable2 = load ptr addrspace(1), ptr %4, align 8
+// DEV:  %vfn3 = getelementptr inbounds ptr addrspace(1), ptr addrspace(1) %vtable2, i64 2
+// DEV:  %5 = load ptr addrspace(1), ptr addrspace(1) %vfn3, align 8
+// DEV:  call addrspace(1) void %5(ptr noundef nonnull align 8 dereferenceable(24) %4)
+
+__device__ void J_dev(J *j) {
+  j->i = 2;
+  j->j = 3;
+  j->g();
+  j->h();
+}
+
+// DEV: define dso_local amdgpu_kernel void @_Z8J_kernelv()
+// DEV:  %j = alloca %struct.J, align 8, addrspace(5)
+// DEV:  %j.ascast = addrspacecast ptr addrspace(5) %j to ptr
+// DEV:  call void @_ZN1JC1Ev(ptr noundef nonnull align 8 dereferenceable(24) %j.ascast)
+// DEV:  call void @_Z5J_devP1J(ptr noundef %j.ascast)
+
+__global__ void J_kernel() {
+  J j;
+  J_dev(&j);
+}
+
+// HOST-LABEL: define dso_local void @"?J_host@@YAXPEAUJ@@@Z"(ptr noundef %j)
+// HOST:  %0 = load ptr, ptr %j.addr, align 8
+// HOST:  %i = getelementptr inbounds %struct.I, ptr %0, i32 0, i32 1
+// HOST:  store i32 12, ptr %i, align 8
+// HOST:  %1 = load ptr, ptr %j.addr, align 8
+// HOST:  %j1 = getelementptr inbounds %struct.J, ptr %1, i32 0, i32 1
+// HOST:  store i32 13, ptr %j1, align 8
+// HOST:  %2 = load ptr, ptr %j.addr, align 8
+// HOST:  %vtable = load ptr, ptr %2, align 8
+// HOST:  %vfn = getelementptr inbounds ptr, ptr %vtable, i64 0
+// HOST:  %3 = load ptr, ptr %vfn, align 8
+// HOST:  call void %3(ptr noundef nonnull align 8 dereferenceable(24) %2)
+// HOST:  %4 = load ptr, ptr %j.addr, align 8
+// HOST:  %vtable2 = load ptr, ptr %4, align 8
+// HOST:  %vfn3 = getelementptr inbounds ptr, ptr %vtable2, i64 2
+// HOST:  %5 = load ptr, ptr %vfn3, align 8
+// HOST:  call void %5(ptr noundef nonnull align 8 dereferenceable(24) %4)
+
+void J_host(J *j) {
+  j->i = 12;
+  j->j = 13;
+  j->f();
+  j->h();
+}
+
+// HOST: define dso_local void @"?test_J@@YAXXZ"()
+// HOST:  %j = alloca %struct.J, align 8
+// HOST:  %call = call noundef ptr @"??0J@@QEAA@XZ"(ptr noundef nonnull align 8 dereferenceable(24) %j)
+// HOST:  call void @"?J_host@@YAXPEAUJ@@@Z"(ptr noundef %j)
+
+void test_J() {
+  J j;
+  J_host(&j);
+  J_kernel<<<1, 1>>>();
+}
+
+// HOST: define linkonce_odr dso_local noundef ptr @"??0J@@QEAA@XZ"(ptr noundef nonnull returned align 8 dereferenceable(24) %this)
+// HOST:  %this.addr = alloca ptr, align 8
+// HOST:  store ptr %this, ptr %this.addr, align 8
+// HOST:  %this1 = load ptr, ptr %this.addr, align 8
+// HOST:  %call = call noundef ptr @"??0I@@QEAA@XZ"(ptr noundef nonnull align 8 dereferenceable(16) %this1) #5
+// HOST:  store ptr @"??_7J@@6B@", ptr %this1, align 8
+// HOST:  ret ptr %this1
+
+// HOST: define linkonce_odr dso_local noundef ptr @"??0I@@QEAA@XZ"(ptr noundef nonnull returned align 8 dereferenceable(16) %this)
+// HOST:  %this.addr = alloca ptr, align 8
+// HOST:  store ptr %this, ptr %this.addr, align 8
+// HOST:  %this1 = load ptr, ptr %this.addr, align 8
+// HOST:  store ptr @"??_7I@@6B@", ptr %this1, align 8
+// HOST:  ret ptr %this1
+
+// DEV: define linkonce_odr void @_ZN1JC1Ev(ptr noundef nonnull align 8 dereferenceable(24) %this)
+// DEV:  %this.addr = alloca ptr, align 8, addrspace(5)
+// DEV:  %this.addr.ascast = addrspacecast ptr addrspace(5) %this.addr to ptr
+// DEV:  store ptr %this, ptr %this.addr.ascast, align 8
+// DEV:  %this1 = load ptr, ptr %this.addr.ascast, align 8
+// DEV:  call void @_ZN1JC2Ev(ptr noundef nonnull align 8 dereferenceable(24) %this1)
+
+// DEV: define linkonce_odr void @_ZN1JC2Ev(ptr noundef nonnull align 8 dereferenceable(24) %this)
+// DEV:  %this.addr = alloca ptr, align 8, addrspace(5)
+// DEV:  %this.addr.ascast = addrspacecast ptr addrspace(5) %this.addr to ptr
+// DEV:  store ptr %this, ptr %this.addr.ascast, align 8
+// DEV:  %this1 = load ptr, ptr %this.addr.ascast, align 8
+// DEV:  call void @_ZN1IC2Ev(ptr noundef nonnull align 8 dereferenceable(16) %this1)
+// DEV:  store ptr addrspace(1) getelementptr inbounds inrange(-16, 24) ({ [5 x ptr addrspace(1)] }, ptr addrspace(1) @_ZTV1J, i32 0, i32 0, i32 2), ptr %this1, align 8
+
+// DEV: define linkonce_odr void @_ZN1IC2Ev(ptr noundef nonnull align 8 dereferenceable(16) %this)
+// DEV:  %this.addr = alloca ptr, align 8, addrspace(5)
+// DEV:  %this.addr.ascast = addrspacecast ptr addrspace(5) %this.addr to ptr
+// DEV:  store ptr %this, ptr %this.addr.ascast, align 8
+// DEV:  %this1 = load ptr, ptr %this.addr.ascast, align 8
+// DEV:  store ptr addrspace(1) getelementptr inbounds inrange(-16, 24) ({ [5 x ptr addrspace(1)] }, ptr addrspace(1) @_ZTV1I, i32 0, i32 0, i32 2), ptr %this1, align 8

From 097b68ff067565af744b5b0327732d7c199710ab Mon Sep 17 00:00:00 2001
From: Shih-Po Hung <shihpo.hung@sifive.com>
Date: Thu, 18 Apr 2024 09:44:31 +0800
Subject: [PATCH 805/886] [RISCV][TTI] Refine the cost of FCmp (#88833)

This patch introduces following changes
- Support all fp predicates
- Use the Val type to estimate the latency/throughput cost
- Assign a cost of 1 for mask operations as LMULCost for mask types
cannot be correctly estimated.
---
 .../Target/RISCV/RISCVTargetTransformInfo.cpp |  48 +-
 llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll | 582 +++++++++++-------
 .../Analysis/CostModel/RISCV/rvv-select.ll    |   8 +-
 3 files changed, 396 insertions(+), 242 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 56f5bd8794ae7..ce26e61880fd0 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -91,6 +91,10 @@ RISCVTTIImpl::getRISCVInstructionCost(ArrayRef<unsigned> OpCodes, MVT VT,
     case RISCV::VMV_S_X:
     case RISCV::VFMV_F_S:
     case RISCV::VFMV_S_F:
+    case RISCV::VMOR_MM:
+    case RISCV::VMXOR_MM:
+    case RISCV::VMAND_MM:
+    case RISCV::VMANDN_MM:
     case RISCV::VMNAND_MM:
     case RISCV::VCPOP_M:
       Cost += 1;
@@ -1383,7 +1387,13 @@ InstructionCost RISCVTTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
            getRISCVInstructionCost(RISCV::VMSLT_VV, LT.second, CostKind);
   }
 
-  if ((Opcode == Instruction::FCmp) && ValTy->isVectorTy()) {
+  if ((Opcode == Instruction::FCmp) && ValTy->isVectorTy() &&
+      CmpInst::isFPPredicate(VecPred)) {
+
+    // Use VMXOR_MM and VMXNOR_MM to generate all true/false mask
+    if ((VecPred == CmpInst::FCMP_FALSE) || (VecPred == CmpInst::FCMP_TRUE))
+      return getRISCVInstructionCost(RISCV::VMXOR_MM, LT.second, CostKind);
+
     // If we do not support the input floating point vector type, use the base
     // one which will calculate as:
     // ScalarizeCost + Num * Cost for fixed vector,
@@ -1393,16 +1403,34 @@ InstructionCost RISCVTTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy,
         (ValTy->getScalarSizeInBits() == 64 && !ST->hasVInstructionsF64()))
       return BaseT::getCmpSelInstrCost(Opcode, ValTy, CondTy, VecPred, CostKind,
                                        I);
+
+    // Assuming vector fp compare and mask instructions are all the same cost
+    // until a need arises to differentiate them.
     switch (VecPred) {
-      // Support natively.
-    case CmpInst::FCMP_OEQ:
-    case CmpInst::FCMP_OGT:
-    case CmpInst::FCMP_OGE:
-    case CmpInst::FCMP_OLT:
-    case CmpInst::FCMP_OLE:
-    case CmpInst::FCMP_UNE:
-      return LT.first * 1;
-    // TODO: Other comparisons?
+    case CmpInst::FCMP_ONE: // vmflt.vv + vmflt.vv + vmor.mm
+    case CmpInst::FCMP_ORD: // vmfeq.vv + vmfeq.vv + vmand.mm
+    case CmpInst::FCMP_UNO: // vmfne.vv + vmfne.vv + vmor.mm
+    case CmpInst::FCMP_UEQ: // vmflt.vv + vmflt.vv + vmnor.mm
+      return LT.first * getRISCVInstructionCost(
+                            {RISCV::VMFLT_VV, RISCV::VMFLT_VV, RISCV::VMOR_MM},
+                            LT.second, CostKind);
+
+    case CmpInst::FCMP_UGT: // vmfle.vv + vmnot.m
+    case CmpInst::FCMP_UGE: // vmflt.vv + vmnot.m
+    case CmpInst::FCMP_ULT: // vmfle.vv + vmnot.m
+    case CmpInst::FCMP_ULE: // vmflt.vv + vmnot.m
+      return LT.first *
+             getRISCVInstructionCost({RISCV::VMFLT_VV, RISCV::VMNAND_MM},
+                                     LT.second, CostKind);
+
+    case CmpInst::FCMP_OEQ: // vmfeq.vv
+    case CmpInst::FCMP_OGT: // vmflt.vv
+    case CmpInst::FCMP_OGE: // vmfle.vv
+    case CmpInst::FCMP_OLT: // vmflt.vv
+    case CmpInst::FCMP_OLE: // vmfle.vv
+    case CmpInst::FCMP_UNE: // vmfne.vv
+      return LT.first *
+             getRISCVInstructionCost(RISCV::VMFLT_VV, LT.second, CostKind);
     default:
       break;
     }
diff --git a/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll b/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
index caa6d6f483a24..8aa78ec95e8ab 100644
--- a/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/rvv-cmp.ll
@@ -878,28 +878,28 @@ define void @fcmp_oeq() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp oeq <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp oeq <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp oeq <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp oeq <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp oeq <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp oeq <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp oeq <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp oeq <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp oeq <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp oeq <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp oeq <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp oeq <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp oeq <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp oeq <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp oeq <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp oeq <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp oeq <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp oeq <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp oeq <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp oeq <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp oeq <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp oeq <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp oeq <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp oeq <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp oeq <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp oeq <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp oeq <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp oeq <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp oeq <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp oeq <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp oeq <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp oeq <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp oeq <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp oeq <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp oeq <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp oeq <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp oeq <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp oeq <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp oeq <2 x half> undef, undef
@@ -938,31 +938,31 @@ define void @fcmp_oeq() {
 
 define void @fcmp_one() {
 ; CHECK-LABEL: 'fcmp_one'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp one <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp one <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp one <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp one <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp one <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp one <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp one <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp one <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp one <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp one <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp one <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp one <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp one <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp one <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp one <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp one <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp one <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp one <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp one <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp one <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp one <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp one <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp one <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp one <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp one <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = fcmp one <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = fcmp one <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f16 = fcmp one <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f16 = fcmp one <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f16 = fcmp one <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f16 = fcmp one <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f16 = fcmp one <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f16 = fcmp one <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f16 = fcmp one <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f32 = fcmp one <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f32 = fcmp one <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f32 = fcmp one <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16f32 = fcmp one <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f32 = fcmp one <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f32 = fcmp one <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f32 = fcmp one <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f32 = fcmp one <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %nxv16f32 = fcmp one <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f64 = fcmp one <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v4f64 = fcmp one <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v8f64 = fcmp one <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f64 = fcmp one <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv2f64 = fcmp one <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv4f64 = fcmp one <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %nxv8f64 = fcmp one <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp one <2 x half> undef, undef
@@ -1004,28 +1004,28 @@ define void @fcmp_olt() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp olt <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp olt <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp olt <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp olt <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp olt <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp olt <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp olt <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp olt <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp olt <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp olt <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp olt <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp olt <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp olt <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp olt <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp olt <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp olt <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp olt <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp olt <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp olt <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp olt <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp olt <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp olt <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp olt <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp olt <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp olt <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp olt <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp olt <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp olt <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp olt <2 x half> undef, undef
@@ -1067,28 +1067,28 @@ define void @fcmp_ole() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ole <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ole <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ole <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ole <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp ole <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ole <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ole <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ole <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ole <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ole <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp ole <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp ole <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ole <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ole <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ole <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ole <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp ole <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp ole <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ole <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ole <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ole <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ole <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ole <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp ole <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp ole <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp ole <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ole <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ole <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ole <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp ole <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp ole <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ole <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ole <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ole <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ole <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp ole <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp ole <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp ole <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ole <2 x half> undef, undef
@@ -1130,28 +1130,28 @@ define void @fcmp_ogt() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ogt <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ogt <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ogt <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ogt <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp ogt <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ogt <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ogt <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ogt <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ogt <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ogt <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp ogt <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp ogt <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ogt <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ogt <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ogt <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ogt <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp ogt <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp ogt <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ogt <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ogt <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ogt <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ogt <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ogt <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp ogt <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp ogt <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp ogt <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ogt <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ogt <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ogt <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp ogt <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp ogt <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ogt <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ogt <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ogt <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ogt <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp ogt <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp ogt <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp ogt <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ogt <2 x half> undef, undef
@@ -1193,28 +1193,28 @@ define void @fcmp_oge() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp oge <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp oge <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp oge <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp oge <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp oge <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp oge <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp oge <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp oge <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp oge <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp oge <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp oge <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp oge <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp oge <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp oge <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp oge <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp oge <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp oge <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp oge <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp oge <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp oge <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp oge <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp oge <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp oge <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp oge <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp oge <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp oge <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp oge <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp oge <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp oge <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp oge <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp oge <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp oge <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp oge <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp oge <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp oge <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp oge <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp oge <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp oge <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp oge <2 x half> undef, undef
@@ -1253,31 +1253,31 @@ define void @fcmp_oge() {
 
 define void @fcmp_ueq() {
 ; CHECK-LABEL: 'fcmp_ueq'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ueq <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ueq <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ueq <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ueq <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ueq <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ueq <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ueq <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ueq <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ueq <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ueq <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ueq <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ueq <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ueq <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ueq <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ueq <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ueq <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ueq <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ueq <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ueq <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ueq <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ueq <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ueq <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ueq <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ueq <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ueq <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = fcmp ueq <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f16 = fcmp ueq <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f16 = fcmp ueq <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f16 = fcmp ueq <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f16 = fcmp ueq <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f16 = fcmp ueq <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f16 = fcmp ueq <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f16 = fcmp ueq <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f16 = fcmp ueq <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f32 = fcmp ueq <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f32 = fcmp ueq <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f32 = fcmp ueq <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v16f32 = fcmp ueq <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f32 = fcmp ueq <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f32 = fcmp ueq <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f32 = fcmp ueq <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f32 = fcmp ueq <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %nxv16f32 = fcmp ueq <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f64 = fcmp ueq <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v4f64 = fcmp ueq <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v8f64 = fcmp ueq <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv1f64 = fcmp ueq <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv2f64 = fcmp ueq <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv4f64 = fcmp ueq <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %nxv8f64 = fcmp ueq <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ueq <2 x half> undef, undef
@@ -1319,28 +1319,28 @@ define void @fcmp_une() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp une <2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp une <4 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp une <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp une <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = fcmp une <16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp une <vscale x 1 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp une <vscale x 2 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp une <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp une <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp une <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv8f16 = fcmp une <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv16f16 = fcmp une <vscale x 16 x half> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp une <2 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp une <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp une <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp une <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f32 = fcmp une <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f32 = fcmp une <16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp une <vscale x 1 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp une <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp une <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp une <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp une <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f32 = fcmp une <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv8f32 = fcmp une <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv16f32 = fcmp une <vscale x 16 x float> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp une <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp une <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp une <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f64 = fcmp une <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f64 = fcmp une <8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp une <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp une <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp une <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp une <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f64 = fcmp une <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %nxv4f64 = fcmp une <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %nxv8f64 = fcmp une <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp une <2 x half> undef, undef
@@ -1379,31 +1379,31 @@ define void @fcmp_une() {
 
 define void @fcmp_ult() {
 ; CHECK-LABEL: 'fcmp_ult'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ult <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ult <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ult <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ult <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ult <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ult <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ult <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ult <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ult <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ult <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ult <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ult <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ult <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ult <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ult <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ult <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ult <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ult <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ult <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ult <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ult <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ult <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ult <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ult <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ult <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp ult <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp ult <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp ult <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v16f16 = fcmp ult <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f16 = fcmp ult <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f16 = fcmp ult <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f16 = fcmp ult <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv8f16 = fcmp ult <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv16f16 = fcmp ult <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp ult <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp ult <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f32 = fcmp ult <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f32 = fcmp ult <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f32 = fcmp ult <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f32 = fcmp ult <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f32 = fcmp ult <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f32 = fcmp ult <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f32 = fcmp ult <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = fcmp ult <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f64 = fcmp ult <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f64 = fcmp ult <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f64 = fcmp ult <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f64 = fcmp ult <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f64 = fcmp ult <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f64 = fcmp ult <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ult <2 x half> undef, undef
@@ -1442,31 +1442,31 @@ define void @fcmp_ult() {
 
 define void @fcmp_ule() {
 ; CHECK-LABEL: 'fcmp_ule'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ule <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ule <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ule <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ule <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ule <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ule <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ule <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ule <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ule <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ule <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ule <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ule <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ule <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ule <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ule <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ule <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ule <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ule <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ule <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ule <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ule <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ule <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ule <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ule <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ule <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp ule <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp ule <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp ule <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v16f16 = fcmp ule <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f16 = fcmp ule <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f16 = fcmp ule <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f16 = fcmp ule <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv8f16 = fcmp ule <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv16f16 = fcmp ule <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp ule <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp ule <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f32 = fcmp ule <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f32 = fcmp ule <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f32 = fcmp ule <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f32 = fcmp ule <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f32 = fcmp ule <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f32 = fcmp ule <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f32 = fcmp ule <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = fcmp ule <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f64 = fcmp ule <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f64 = fcmp ule <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f64 = fcmp ule <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f64 = fcmp ule <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f64 = fcmp ule <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f64 = fcmp ule <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ule <2 x half> undef, undef
@@ -1505,31 +1505,31 @@ define void @fcmp_ule() {
 
 define void @fcmp_ugt() {
 ; CHECK-LABEL: 'fcmp_ugt'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp ugt <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp ugt <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp ugt <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp ugt <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp ugt <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp ugt <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp ugt <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp ugt <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp ugt <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp ugt <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp ugt <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp ugt <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp ugt <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp ugt <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp ugt <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp ugt <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp ugt <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp ugt <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp ugt <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp ugt <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp ugt <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp ugt <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp ugt <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp ugt <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp ugt <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp ugt <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp ugt <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp ugt <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v16f16 = fcmp ugt <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f16 = fcmp ugt <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f16 = fcmp ugt <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f16 = fcmp ugt <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv8f16 = fcmp ugt <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv16f16 = fcmp ugt <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp ugt <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp ugt <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f32 = fcmp ugt <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f32 = fcmp ugt <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f32 = fcmp ugt <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f32 = fcmp ugt <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f32 = fcmp ugt <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f32 = fcmp ugt <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f32 = fcmp ugt <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = fcmp ugt <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f64 = fcmp ugt <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f64 = fcmp ugt <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f64 = fcmp ugt <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f64 = fcmp ugt <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f64 = fcmp ugt <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f64 = fcmp ugt <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp ugt <2 x half> undef, undef
@@ -1568,31 +1568,31 @@ define void @fcmp_ugt() {
 
 define void @fcmp_uge() {
 ; CHECK-LABEL: 'fcmp_uge'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp uge <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp uge <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp uge <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp uge <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp uge <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp uge <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp uge <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp uge <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp uge <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp uge <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp uge <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp uge <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp uge <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp uge <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp uge <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp uge <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp uge <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp uge <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp uge <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp uge <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp uge <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp uge <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp uge <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp uge <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp uge <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp uge <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp uge <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp uge <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v16f16 = fcmp uge <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f16 = fcmp uge <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f16 = fcmp uge <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv4f16 = fcmp uge <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv8f16 = fcmp uge <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv16f16 = fcmp uge <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp uge <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp uge <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v8f32 = fcmp uge <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v16f32 = fcmp uge <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f32 = fcmp uge <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv2f32 = fcmp uge <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv4f32 = fcmp uge <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv8f32 = fcmp uge <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv16f32 = fcmp uge <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = fcmp uge <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v4f64 = fcmp uge <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v8f64 = fcmp uge <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %nxv1f64 = fcmp uge <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %nxv2f64 = fcmp uge <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %nxv4f64 = fcmp uge <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %nxv8f64 = fcmp uge <vscale x 8 x double> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp uge <2 x half> undef, undef
@@ -1628,3 +1628,129 @@ define void @fcmp_uge() {
 
   ret void
 }
+
+define void @fcmp_true() {
+; CHECK-LABEL: 'fcmp_true'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp true <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp true <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp true <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp true <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp true <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp true <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp true <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp true <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp true <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp true <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp true <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp true <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp true <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp true <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp true <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp true <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp true <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp true <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp true <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp true <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp true <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp true <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp true <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp true <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp true <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %v2f16 = fcmp true <2 x half> undef, undef
+  %v4f16 = fcmp true <4 x half> undef, undef
+  %v8f16 = fcmp true <8 x half> undef, undef
+  %v16f16 = fcmp true <16 x half> undef, undef
+
+  %nxv1f16 = fcmp true <vscale x 1 x half> undef, undef
+  %nxv2f16 = fcmp true <vscale x 2 x half> undef, undef
+  %nxv4f16 = fcmp true <vscale x 4 x half> undef, undef
+  %nxv8f16 = fcmp true <vscale x 8 x half> undef, undef
+  %nxv16f16 = fcmp true <vscale x 16 x half> undef, undef
+
+  %v2f32 = fcmp true <2 x float> undef, undef
+  %v4f32 = fcmp true <4 x float> undef, undef
+  %v8f32 = fcmp true <8 x float> undef, undef
+  %v16f32 = fcmp true <16 x float> undef, undef
+
+  %nxv1f32 = fcmp true <vscale x 1 x float> undef, undef
+  %nxv2f32 = fcmp true <vscale x 2 x float> undef, undef
+  %nxv4f32 = fcmp true <vscale x 4 x float> undef, undef
+  %nxv8f32 = fcmp true <vscale x 8 x float> undef, undef
+  %nxv16f32 = fcmp true <vscale x 16 x float> undef, undef
+
+  %v2f64 = fcmp true <2 x double> undef, undef
+  %v4f64 = fcmp true <4 x double> undef, undef
+  %v8f64 = fcmp true <8 x double> undef, undef
+
+  %nxv1f64 = fcmp true <vscale x 1 x double> undef, undef
+  %nxv2f64 = fcmp true <vscale x 2 x double> undef, undef
+  %nxv4f64 = fcmp true <vscale x 4 x double> undef, undef
+  %nxv8f64 = fcmp true <vscale x 8 x double> undef, undef
+
+  ret void
+}
+
+define void @fcmp_false() {
+; CHECK-LABEL: 'fcmp_false'
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = fcmp false <2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f16 = fcmp false <4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f16 = fcmp false <8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f16 = fcmp false <16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f16 = fcmp false <vscale x 1 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f16 = fcmp false <vscale x 2 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f16 = fcmp false <vscale x 4 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f16 = fcmp false <vscale x 8 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f16 = fcmp false <vscale x 16 x half> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = fcmp false <2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = fcmp false <4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f32 = fcmp false <8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v16f32 = fcmp false <16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f32 = fcmp false <vscale x 1 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f32 = fcmp false <vscale x 2 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f32 = fcmp false <vscale x 4 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f32 = fcmp false <vscale x 8 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv16f32 = fcmp false <vscale x 16 x float> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = fcmp false <2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v4f64 = fcmp false <4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v8f64 = fcmp false <8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv1f64 = fcmp false <vscale x 1 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv2f64 = fcmp false <vscale x 2 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv4f64 = fcmp false <vscale x 4 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %nxv8f64 = fcmp false <vscale x 8 x double> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %v2f16 = fcmp false <2 x half> undef, undef
+  %v4f16 = fcmp false <4 x half> undef, undef
+  %v8f16 = fcmp false <8 x half> undef, undef
+  %v16f16 = fcmp false <16 x half> undef, undef
+
+  %nxv1f16 = fcmp false <vscale x 1 x half> undef, undef
+  %nxv2f16 = fcmp false <vscale x 2 x half> undef, undef
+  %nxv4f16 = fcmp false <vscale x 4 x half> undef, undef
+  %nxv8f16 = fcmp false <vscale x 8 x half> undef, undef
+  %nxv16f16 = fcmp false <vscale x 16 x half> undef, undef
+
+  %v2f32 = fcmp false <2 x float> undef, undef
+  %v4f32 = fcmp false <4 x float> undef, undef
+  %v8f32 = fcmp false <8 x float> undef, undef
+  %v16f32 = fcmp false <16 x float> undef, undef
+
+  %nxv1f32 = fcmp false <vscale x 1 x float> undef, undef
+  %nxv2f32 = fcmp false <vscale x 2 x float> undef, undef
+  %nxv4f32 = fcmp false <vscale x 4 x float> undef, undef
+  %nxv8f32 = fcmp false <vscale x 8 x float> undef, undef
+  %nxv16f32 = fcmp false <vscale x 16 x float> undef, undef
+
+  %v2f64 = fcmp false <2 x double> undef, undef
+  %v4f64 = fcmp false <4 x double> undef, undef
+  %v8f64 = fcmp false <8 x double> undef, undef
+
+  %nxv1f64 = fcmp false <vscale x 1 x double> undef, undef
+  %nxv2f64 = fcmp false <vscale x 2 x double> undef, undef
+  %nxv4f64 = fcmp false <vscale x 4 x double> undef, undef
+  %nxv8f64 = fcmp false <vscale x 8 x double> undef, undef
+
+  ret void
+}
diff --git a/llvm/test/Analysis/CostModel/RISCV/rvv-select.ll b/llvm/test/Analysis/CostModel/RISCV/rvv-select.ll
index 6dcbc73674aa6..b80096b057a66 100644
--- a/llvm/test/Analysis/CostModel/RISCV/rvv-select.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/rvv-select.ll
@@ -22,14 +22,14 @@ define void @select() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %15 = select i1 undef, <vscale x 2 x i1> undef, <vscale x 2 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %16 = select i1 undef, <vscale x 4 x i1> undef, <vscale x 4 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %17 = select i1 undef, <vscale x 8 x i1> undef, <vscale x 8 x i1> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %18 = select i1 undef, <vscale x 16 x i1> undef, <vscale x 16 x i1> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %19 = select i1 undef, <vscale x 32 x i1> undef, <vscale x 32 x i1> undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %18 = select i1 undef, <vscale x 16 x i1> undef, <vscale x 16 x i1> undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %19 = select i1 undef, <vscale x 32 x i1> undef, <vscale x 32 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %20 = select <vscale x 1 x i1> undef, <vscale x 1 x i1> undef, <vscale x 1 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %21 = select <vscale x 2 x i1> undef, <vscale x 2 x i1> undef, <vscale x 2 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %22 = select <vscale x 4 x i1> undef, <vscale x 4 x i1> undef, <vscale x 4 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %23 = select <vscale x 8 x i1> undef, <vscale x 8 x i1> undef, <vscale x 8 x i1> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %24 = select <vscale x 16 x i1> undef, <vscale x 16 x i1> undef, <vscale x 16 x i1> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %25 = select <vscale x 32 x i1> undef, <vscale x 32 x i1> undef, <vscale x 32 x i1> undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %24 = select <vscale x 16 x i1> undef, <vscale x 16 x i1> undef, <vscale x 16 x i1> undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %25 = select <vscale x 32 x i1> undef, <vscale x 32 x i1> undef, <vscale x 32 x i1> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %26 = select i1 undef, i8 undef, i8 undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %27 = select i1 undef, <1 x i8> undef, <1 x i8> undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %28 = select i1 undef, <2 x i8> undef, <2 x i8> undef

From 525d00e5edc4b83105c6ad518926b174f76c3a1c Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 10:56:20 +0900
Subject: [PATCH 806/886] [InstCombine] Fix poison propagation in round up
 alignment fold

We can't directly use the high bits value if it is more poisonous
due to poison elements in the masks.

This fixes the issue reported in
https://github.com/llvm/llvm-project/pull/88217#issuecomment-2061034941.
---
 .../InstCombine/InstCombineSelect.cpp         |  3 ++-
 .../integer-round-up-pow2-alignment.ll        | 20 +++++++++++++------
 2 files changed, 16 insertions(+), 7 deletions(-)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 2d78fcee1152d..41c3ab85940ea 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -2965,7 +2965,8 @@ foldRoundUpIntegerWithPow2Alignment(SelectInst &SI,
     return nullptr;
 
   if (!XBiasedHighBits->hasOneUse()) {
-    if (*BiasCst == *LowBitMaskCst)
+    // We can't directly return XBiasedHighBits if it is more poisonous.
+    if (*BiasCst == *LowBitMaskCst && impliesPoison(XBiasedHighBits, X))
       return XBiasedHighBits;
     return nullptr;
   }
diff --git a/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll b/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
index c7e0553992b90..afd56abf40a50 100644
--- a/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
+++ b/llvm/test/Transforms/InstCombine/integer-round-up-pow2-alignment.ll
@@ -437,14 +437,17 @@ define i8 @t17_oneuse(i8 %x) {
   ret i8 %x.roundedup
 }
 
-; Bias is equal to the alignment-1 (as opposed to alignment),
-; so we can just replace %x.roundedup with %x.biased.highbits
+; Negative test: We can't replace with %x.biased.highbits because it is
+; more poisonous.
 define <2 x i4> @t18_replacement_0b0001(<2 x i4> %x) {
 ; CHECK-LABEL: @t18_replacement_0b0001(
-; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X:%.*]], <i4 3, i4 3>
+; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i4> [[X:%.*]], <i4 3, i4 3>
+; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i4> [[X_LOWBITS]], zeroinitializer
+; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X]], <i4 3, i4 3>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i4> [[X_BIASED]], <i4 -4, i4 poison>
 ; CHECK-NEXT:    call void @use.v2i4(<2 x i4> [[X_BIASED_HIGHBITS]])
-; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
+; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i4> [[X]], <2 x i4> [[X_BIASED_HIGHBITS]]
+; CHECK-NEXT:    ret <2 x i4> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i4> %x, <i4 3, i4 3>
   %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 0>
@@ -454,12 +457,17 @@ define <2 x i4> @t18_replacement_0b0001(<2 x i4> %x) {
   %x.roundedup = select <2 x i1> %x.lowbits.are.zero, <2 x i4> %x, <2 x i4> %x.biased.highbits
   ret <2 x i4> %x.roundedup
 }
+; Negative test: We can't replace with %x.biased.highbits because it is
+; more poisonous.
 define <2 x i4> @t18_replacement_0b0010(<2 x i4> %x) {
 ; CHECK-LABEL: @t18_replacement_0b0010(
-; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X:%.*]], <i4 3, i4 poison>
+; CHECK-NEXT:    [[X_LOWBITS:%.*]] = and <2 x i4> [[X:%.*]], <i4 3, i4 3>
+; CHECK-NEXT:    [[X_LOWBITS_ARE_ZERO:%.*]] = icmp eq <2 x i4> [[X_LOWBITS]], zeroinitializer
+; CHECK-NEXT:    [[X_BIASED:%.*]] = add <2 x i4> [[X]], <i4 3, i4 poison>
 ; CHECK-NEXT:    [[X_BIASED_HIGHBITS:%.*]] = and <2 x i4> [[X_BIASED]], <i4 -4, i4 -4>
 ; CHECK-NEXT:    call void @use.v2i4(<2 x i4> [[X_BIASED_HIGHBITS]])
-; CHECK-NEXT:    ret <2 x i4> [[X_BIASED_HIGHBITS]]
+; CHECK-NEXT:    [[X_ROUNDEDUP:%.*]] = select <2 x i1> [[X_LOWBITS_ARE_ZERO]], <2 x i4> [[X]], <2 x i4> [[X_BIASED_HIGHBITS]]
+; CHECK-NEXT:    ret <2 x i4> [[X_ROUNDEDUP]]
 ;
   %x.lowbits = and <2 x i4> %x, <i4 3, i4 3>
   %x.lowbits.are.zero = icmp eq <2 x i4> %x.lowbits, <i4 0, i4 0>

From 0ee260ec37a039ffc276f2f827c842f71513a3e9 Mon Sep 17 00:00:00 2001
From: Kai Luo <lkail@cn.ibm.com>
Date: Thu, 18 Apr 2024 10:00:56 +0800
Subject: [PATCH 807/886] [PowerPC] `ANDI_rec_1_*` should define CR0 (#89034)

These pseudo instructions finally copy the result to CR0 so they should
define `CR0` in their definitions.
---
 llvm/lib/Target/PowerPC/PPCInstrInfo.td | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.td b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
index 43e3902cf4074..261b9a3d1dffe 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.td
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.td
@@ -4191,6 +4191,7 @@ def : Pat<(v4i32 (selectcc i1:$lhs, i1:$rhs, v4i32:$tval, v4i32:$fval, SETUGT)),
 def : Pat<(v4i32 (selectcc i1:$lhs, i1:$rhs, v4i32:$tval, v4i32:$fval, SETNE)),
           (SELECT_VRRC (CRXOR $lhs, $rhs), $tval, $fval)>;
 
+let Defs = [CR0] in {
 def ANDI_rec_1_EQ_BIT : PPCCustomInserterPseudo<(outs crbitrc:$dst), (ins gprc:$in),
                              "#ANDI_rec_1_EQ_BIT",
                              [(set i1:$dst, (trunc (not i32:$in)))]>;
@@ -4204,6 +4205,7 @@ def ANDI_rec_1_EQ_BIT8 : PPCCustomInserterPseudo<(outs crbitrc:$dst), (ins g8rc:
 def ANDI_rec_1_GT_BIT8 : PPCCustomInserterPseudo<(outs crbitrc:$dst), (ins g8rc:$in),
                               "#ANDI_rec_1_GT_BIT8",
                               [(set i1:$dst, (trunc i64:$in))]>;
+}
 
 def : Pat<(i1 (not (trunc i32:$in))),
            (ANDI_rec_1_EQ_BIT $in)>;

From 29ecd6d50f1400e3101f27567b535eb8af905f58 Mon Sep 17 00:00:00 2001
From: Owen Pan <owenpiano@gmail.com>
Date: Wed, 17 Apr 2024 19:51:46 -0700
Subject: [PATCH 808/886] [clang-format] Revert breaking stream operators to
 previous default (#89016)

Reverts commit d68826dfbd98, which changes the previous default behavior
of always breaking before a stream insertion operator `<<` if both
operands are string literals.

Also reverts the related commits 27f547968cce and bf05be5b87fc.

See the discussion in #88483.
---
 clang/lib/Format/TokenAnnotator.cpp           | 8 ++------
 clang/unittests/Format/FormatTest.cpp         | 7 +------
 clang/unittests/Format/TokenAnnotatorTest.cpp | 9 ---------
 polly/lib/Analysis/DependenceInfo.cpp         | 4 ++--
 polly/lib/Analysis/ScopBuilder.cpp            | 7 ++++---
 5 files changed, 9 insertions(+), 26 deletions(-)

diff --git a/clang/lib/Format/TokenAnnotator.cpp b/clang/lib/Format/TokenAnnotator.cpp
index 628f70417866c..80e5605fb5778 100644
--- a/clang/lib/Format/TokenAnnotator.cpp
+++ b/clang/lib/Format/TokenAnnotator.cpp
@@ -5595,12 +5595,8 @@ bool TokenAnnotator::mustBreakBefore(const AnnotatedLine &Line,
     return true;
   if (Left.IsUnterminatedLiteral)
     return true;
-  // FIXME: Breaking after newlines seems useful in general. Turn this into an
-  // option and recognize more cases like endl etc, and break independent of
-  // what comes after operator lessless.
-  if (Right.is(tok::lessless) && Right.Next &&
-      Right.Next->is(tok::string_literal) && Left.is(tok::string_literal) &&
-      Left.TokenText.ends_with("\\n\"")) {
+  if (Right.is(tok::lessless) && Right.Next && Left.is(tok::string_literal) &&
+      Right.Next->is(tok::string_literal)) {
     return true;
   }
   if (Right.is(TT_RequiresClause)) {
diff --git a/clang/unittests/Format/FormatTest.cpp b/clang/unittests/Format/FormatTest.cpp
index 4906b3350b5b2..bc61b9c089e92 100644
--- a/clang/unittests/Format/FormatTest.cpp
+++ b/clang/unittests/Format/FormatTest.cpp
@@ -27339,12 +27339,6 @@ TEST_F(FormatTest, PPDirectivesAndCommentsInBracedInit) {
                getLLVMStyleWithColumns(30));
 }
 
-TEST_F(FormatTest, StreamOutputOperator) {
-  verifyFormat("std::cout << \"foo\" << \"bar\" << baz;");
-  verifyFormat("std::cout << \"foo\\n\"\n"
-               "          << \"bar\";");
-}
-
 TEST_F(FormatTest, BreakAdjacentStringLiterals) {
   constexpr StringRef Code{
       "return \"Code\" \"\\0\\52\\26\\55\\55\\0\" \"x013\" \"\\02\\xBA\";"};
@@ -27359,6 +27353,7 @@ TEST_F(FormatTest, BreakAdjacentStringLiterals) {
   Style.BreakAdjacentStringLiterals = false;
   verifyFormat(Code, Style);
 }
+
 } // namespace
 } // namespace test
 } // namespace format
diff --git a/clang/unittests/Format/TokenAnnotatorTest.cpp b/clang/unittests/Format/TokenAnnotatorTest.cpp
index da02ced8c7a94..a71b680647695 100644
--- a/clang/unittests/Format/TokenAnnotatorTest.cpp
+++ b/clang/unittests/Format/TokenAnnotatorTest.cpp
@@ -2841,15 +2841,6 @@ TEST_F(TokenAnnotatorTest, BraceKind) {
   EXPECT_BRACE_KIND(Tokens[16], BK_BracedInit);
 }
 
-TEST_F(TokenAnnotatorTest, StreamOperator) {
-  auto Tokens = annotate("\"foo\\n\" << aux << \"foo\\n\" << \"foo\";");
-  ASSERT_EQ(Tokens.size(), 9u) << Tokens;
-  EXPECT_FALSE(Tokens[1]->MustBreakBefore);
-  EXPECT_FALSE(Tokens[3]->MustBreakBefore);
-  // Only break between string literals if the former ends with \n.
-  EXPECT_TRUE(Tokens[5]->MustBreakBefore);
-}
-
 TEST_F(TokenAnnotatorTest, UnderstandsElaboratedTypeSpecifier) {
   auto Tokens = annotate("auto foo() -> enum En {}");
   ASSERT_EQ(Tokens.size(), 10u) << Tokens;
diff --git a/polly/lib/Analysis/DependenceInfo.cpp b/polly/lib/Analysis/DependenceInfo.cpp
index 9ee004fbac9e5..a530fa7b58fa6 100644
--- a/polly/lib/Analysis/DependenceInfo.cpp
+++ b/polly/lib/Analysis/DependenceInfo.cpp
@@ -951,8 +951,8 @@ class DependenceInfoPrinterLegacyPass final : public ScopPass {
   bool runOnScop(Scop &S) override {
     DependenceInfo &P = getAnalysis<DependenceInfo>();
 
-    OS << "Printing analysis '" << P.getPassName() << "' for " << "region: '"
-       << S.getRegion().getNameStr() << "' in function '"
+    OS << "Printing analysis '" << P.getPassName() << "' for "
+       << "region: '" << S.getRegion().getNameStr() << "' in function '"
        << S.getFunction().getName() << "':\n";
     P.printScop(OS, S);
 
diff --git a/polly/lib/Analysis/ScopBuilder.cpp b/polly/lib/Analysis/ScopBuilder.cpp
index 214e4d360d6bb..a34f293152725 100644
--- a/polly/lib/Analysis/ScopBuilder.cpp
+++ b/polly/lib/Analysis/ScopBuilder.cpp
@@ -2715,9 +2715,10 @@ void ScopBuilder::addUserContext() {
     if (NameContext != NameUserContext) {
       std::string SpaceStr = stringFromIslObj(Space, "null");
       errs() << "Error: the name of dimension " << i
-             << " provided in -polly-context " << "is '" << NameUserContext
-             << "', but the name in the computed " << "context is '"
-             << NameContext << "'. Due to this name mismatch, "
+             << " provided in -polly-context "
+             << "is '" << NameUserContext << "', but the name in the computed "
+             << "context is '" << NameContext
+             << "'. Due to this name mismatch, "
              << "the -polly-context option is ignored. Please provide "
              << "the context in the parameter space: " << SpaceStr << ".\n";
       return;

From b6cc667190e3bee7485a225d3dadd8a57c0a22b6 Mon Sep 17 00:00:00 2001
From: Owen Pan <owenpiano@gmail.com>
Date: Wed, 17 Apr 2024 19:53:24 -0700
Subject: [PATCH 809/886] [clang-format] Annotate ampamp after new/delete as
 BinaryOperator (#89033)

Fixes #78789.
---
 clang/lib/Format/TokenAnnotator.cpp           |  2 ++
 clang/unittests/Format/TokenAnnotatorTest.cpp | 10 ++++++++++
 2 files changed, 12 insertions(+)

diff --git a/clang/lib/Format/TokenAnnotator.cpp b/clang/lib/Format/TokenAnnotator.cpp
index 80e5605fb5778..a679683077ac9 100644
--- a/clang/lib/Format/TokenAnnotator.cpp
+++ b/clang/lib/Format/TokenAnnotator.cpp
@@ -2912,6 +2912,8 @@ class AnnotatingParser {
       return TT_UnaryOperator;
     if (PrevToken->is(TT_TypeName))
       return TT_PointerOrReference;
+    if (PrevToken->isOneOf(tok::kw_new, tok::kw_delete) && Tok.is(tok::ampamp))
+      return TT_BinaryOperator;
 
     const FormatToken *NextToken = Tok.getNextNonComment();
 
diff --git a/clang/unittests/Format/TokenAnnotatorTest.cpp b/clang/unittests/Format/TokenAnnotatorTest.cpp
index a71b680647695..4f445c64ab303 100644
--- a/clang/unittests/Format/TokenAnnotatorTest.cpp
+++ b/clang/unittests/Format/TokenAnnotatorTest.cpp
@@ -309,6 +309,16 @@ TEST_F(TokenAnnotatorTest, UnderstandsUsesOfStarAndAmp) {
   EXPECT_TOKEN(Tokens[6], tok::l_paren, TT_OverloadedOperatorLParen);
   EXPECT_TOKEN(Tokens[8], tok::l_brace, TT_FunctionLBrace);
   EXPECT_TOKEN(Tokens[11], tok::amp, TT_PointerOrReference);
+
+  Tokens = annotate("if (new && num) {\n"
+                    "  new = 1;\n"
+                    "}\n"
+                    "if (!delete && num) {\n"
+                    "  delete = 1;\n"
+                    "}");
+  ASSERT_EQ(Tokens.size(), 26u) << Tokens;
+  EXPECT_TOKEN(Tokens[3], tok::ampamp, TT_BinaryOperator);
+  EXPECT_TOKEN(Tokens[16], tok::ampamp, TT_BinaryOperator);
 }
 
 TEST_F(TokenAnnotatorTest, UnderstandsUsesOfPlusAndMinus) {

From f2695a1c2f561da42b973187a254b6eeb7da76a9 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Wed, 17 Apr 2024 16:40:00 +0800
Subject: [PATCH 810/886] [C++20] [Modules] Avoid writing untouched DeclUpdates
 from GMF in Reduced BMI

Mitigate https://github.com/llvm/llvm-project/issues/61447

The root cause of the above problem is that when we write a declaration,
we need to lookup all the redeclarations in the imported modules. Then
it will be pretty slow if there are too many redeclarations in different
modules. This patch doesn't solve the porblem.

What the patchs mitigated is, when we writing a named module, we shouldn't
write the declarations from GMF if it is unreferenced **in current
module unit**. The difference here is that, if the declaration is used
in the imported modules, we used to emit it as an update. But we
definitely want to avoid that after this patch.

For that reproducer in
https://github.com/llvm/llvm-project/issues/61447, it used to take 2.5s
to compile and now it only takes 0.49s to compile, which is a big win.
---
 clang/include/clang/AST/ASTMutationListener.h | 27 ++++++
 clang/include/clang/AST/Decl.h                |  2 +-
 clang/include/clang/Serialization/ASTWriter.h | 10 ++
 clang/lib/AST/ASTContext.cpp                  | 18 +++-
 clang/lib/AST/Decl.cpp                        |  7 ++
 clang/lib/Frontend/MultiplexConsumer.cpp      | 29 ++++++
 clang/lib/Sema/SemaModule.cpp                 |  4 +
 clang/lib/Serialization/ASTWriter.cpp         | 64 +++++++++----
 .../reduced-bmi-empty-module-purview.cppm     | 96 +++++++++++++++++++
 9 files changed, 231 insertions(+), 26 deletions(-)
 create mode 100644 clang/test/Modules/reduced-bmi-empty-module-purview.cppm

diff --git a/clang/include/clang/AST/ASTMutationListener.h b/clang/include/clang/AST/ASTMutationListener.h
index 8879f9f3229ff..2c4ec2ce67f36 100644
--- a/clang/include/clang/AST/ASTMutationListener.h
+++ b/clang/include/clang/AST/ASTMutationListener.h
@@ -27,6 +27,7 @@ namespace clang {
   class FunctionTemplateDecl;
   class Module;
   class NamedDecl;
+  class NamespaceDecl;
   class ObjCCategoryDecl;
   class ObjCContainerDecl;
   class ObjCInterfaceDecl;
@@ -35,6 +36,7 @@ namespace clang {
   class QualType;
   class RecordDecl;
   class TagDecl;
+  class TranslationUnitDecl;
   class ValueDecl;
   class VarDecl;
   class VarTemplateDecl;
@@ -147,6 +149,31 @@ class ASTMutationListener {
   virtual void AddedAttributeToRecord(const Attr *Attr,
                                       const RecordDecl *Record) {}
 
+  /// The parser find the named module declaration.
+  virtual void EnteringModulePurview() {}
+
+  /// An mangling number was added to a Decl
+  ///
+  /// \param D The decl that got a mangling number
+  ///
+  /// \param Number The mangling number that was added to the Decl
+  virtual void AddedManglingNumber(const Decl *D, unsigned Number) {}
+
+  /// An static local number was added to a Decl
+  ///
+  /// \param D The decl that got a static local number
+  ///
+  /// \param Number The static local number that was added to the Decl
+  virtual void AddedStaticLocalNumbers(const Decl *D, unsigned Number) {}
+
+  /// An anonymous namespace was added the translation unit decl
+  ///
+  /// \param TU The translation unit decl that got a new anonymous namespace
+  ///
+  /// \param AnonNamespace The anonymous namespace that was added
+  virtual void AddedAnonymousNamespace(const TranslationUnitDecl *TU,
+                                       NamespaceDecl *AnonNamespace) {}
+
   // NOTE: If new methods are added they should also be added to
   // MultiplexASTMutationListener.
 };
diff --git a/clang/include/clang/AST/Decl.h b/clang/include/clang/AST/Decl.h
index 01af50ca694fd..91449ebcceec3 100644
--- a/clang/include/clang/AST/Decl.h
+++ b/clang/include/clang/AST/Decl.h
@@ -120,7 +120,7 @@ class TranslationUnitDecl : public Decl,
   ASTContext &getASTContext() const { return Ctx; }
 
   NamespaceDecl *getAnonymousNamespace() const { return AnonymousNamespace; }
-  void setAnonymousNamespace(NamespaceDecl *D) { AnonymousNamespace = D; }
+  void setAnonymousNamespace(NamespaceDecl *D);
 
   static TranslationUnitDecl *Create(ASTContext &C);
 
diff --git a/clang/include/clang/Serialization/ASTWriter.h b/clang/include/clang/Serialization/ASTWriter.h
index 443f770310470..892d7363e06dd 100644
--- a/clang/include/clang/Serialization/ASTWriter.h
+++ b/clang/include/clang/Serialization/ASTWriter.h
@@ -399,6 +399,11 @@ class ASTWriter : public ASTDeserializationListener,
   /// record containing modifications to them.
   DeclUpdateMap DeclUpdates;
 
+  /// DeclUpdates added during parsing the GMF. We split these from
+  /// DeclUpdates since we want to add these updates in GMF on need.
+  /// Only meaningful for reduced BMI.
+  DeclUpdateMap DeclUpdatesFromGMF;
+
   using FirstLatestDeclMap = llvm::DenseMap<Decl *, Decl *>;
 
   /// Map of first declarations from a chained PCH that point to the
@@ -866,6 +871,11 @@ class ASTWriter : public ASTDeserializationListener,
   void RedefinedHiddenDefinition(const NamedDecl *D, Module *M) override;
   void AddedAttributeToRecord(const Attr *Attr,
                               const RecordDecl *Record) override;
+  void EnteringModulePurview() override;
+  void AddedManglingNumber(const Decl *D, unsigned) override;
+  void AddedStaticLocalNumbers(const Decl *D, unsigned) override;
+  void AddedAnonymousNamespace(const TranslationUnitDecl *,
+                               NamespaceDecl *AnonNamespace) override;
 };
 
 /// AST and semantic-analysis consumer that generates a
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 6ce233704a588..bcae6d705b780 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -12245,8 +12245,13 @@ QualType ASTContext::getRealTypeForBitwidth(unsigned DestWidth,
 }
 
 void ASTContext::setManglingNumber(const NamedDecl *ND, unsigned Number) {
-  if (Number > 1)
-    MangleNumbers[ND] = Number;
+  if (Number <= 1)
+    return;
+
+  MangleNumbers[ND] = Number;
+
+  if (Listener)
+    Listener->AddedManglingNumber(ND, Number);
 }
 
 unsigned ASTContext::getManglingNumber(const NamedDecl *ND,
@@ -12265,8 +12270,13 @@ unsigned ASTContext::getManglingNumber(const NamedDecl *ND,
 }
 
 void ASTContext::setStaticLocalNumber(const VarDecl *VD, unsigned Number) {
-  if (Number > 1)
-    StaticLocalNumbers[VD] = Number;
+  if (Number <= 1)
+    return;
+
+  StaticLocalNumbers[VD] = Number;
+
+  if (Listener)
+    Listener->AddedStaticLocalNumbers(VD, Number);
 }
 
 unsigned ASTContext::getStaticLocalNumber(const VarDecl *VD) const {
diff --git a/clang/lib/AST/Decl.cpp b/clang/lib/AST/Decl.cpp
index 33b6f8611f216..63388d79bbf21 100644
--- a/clang/lib/AST/Decl.cpp
+++ b/clang/lib/AST/Decl.cpp
@@ -5274,6 +5274,13 @@ TranslationUnitDecl *TranslationUnitDecl::Create(ASTContext &C) {
   return new (C, (DeclContext *)nullptr) TranslationUnitDecl(C);
 }
 
+void TranslationUnitDecl::setAnonymousNamespace(NamespaceDecl *D) {
+  AnonymousNamespace = D;
+
+  if (ASTMutationListener *Listener = Ctx.getASTMutationListener())
+    Listener->AddedAnonymousNamespace(this, D);
+}
+
 void PragmaCommentDecl::anchor() {}
 
 PragmaCommentDecl *PragmaCommentDecl::Create(const ASTContext &C,
diff --git a/clang/lib/Frontend/MultiplexConsumer.cpp b/clang/lib/Frontend/MultiplexConsumer.cpp
index 737877329c9ce..744ea70cc24de 100644
--- a/clang/lib/Frontend/MultiplexConsumer.cpp
+++ b/clang/lib/Frontend/MultiplexConsumer.cpp
@@ -20,6 +20,9 @@ using namespace clang;
 
 namespace clang {
 
+class NamespaceDecl;
+class TranslationUnitDecl;
+
 MultiplexASTDeserializationListener::MultiplexASTDeserializationListener(
       const std::vector<ASTDeserializationListener*>& L)
     : Listeners(L) {
@@ -115,6 +118,11 @@ class MultiplexASTMutationListener : public ASTMutationListener {
   void RedefinedHiddenDefinition(const NamedDecl *D, Module *M) override;
   void AddedAttributeToRecord(const Attr *Attr,
                               const RecordDecl *Record) override;
+  void EnteringModulePurview() override;
+  void AddedManglingNumber(const Decl *D, unsigned) override;
+  void AddedStaticLocalNumbers(const Decl *D, unsigned) override;
+  void AddedAnonymousNamespace(const TranslationUnitDecl *,
+                               NamespaceDecl *AnonNamespace) override;
 
 private:
   std::vector<ASTMutationListener*> Listeners;
@@ -238,6 +246,27 @@ void MultiplexASTMutationListener::AddedAttributeToRecord(
     L->AddedAttributeToRecord(Attr, Record);
 }
 
+void MultiplexASTMutationListener::EnteringModulePurview() {
+  for (auto *L : Listeners)
+    L->EnteringModulePurview();
+}
+
+void MultiplexASTMutationListener::AddedManglingNumber(const Decl *D,
+                                                       unsigned Number) {
+  for (auto *L : Listeners)
+    L->AddedManglingNumber(D, Number);
+}
+void MultiplexASTMutationListener::AddedStaticLocalNumbers(const Decl *D,
+                                                           unsigned Number) {
+  for (auto *L : Listeners)
+    L->AddedStaticLocalNumbers(D, Number);
+}
+void MultiplexASTMutationListener::AddedAnonymousNamespace(
+    const TranslationUnitDecl *TU, NamespaceDecl *AnonNamespace) {
+  for (auto *L : Listeners)
+    L->AddedAnonymousNamespace(TU, AnonNamespace);
+}
+
 }  // end namespace clang
 
 MultiplexConsumer::MultiplexConsumer(
diff --git a/clang/lib/Sema/SemaModule.cpp b/clang/lib/Sema/SemaModule.cpp
index 2ddf9d70263a0..67658c93ed3ba 100644
--- a/clang/lib/Sema/SemaModule.cpp
+++ b/clang/lib/Sema/SemaModule.cpp
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "clang/AST/ASTConsumer.h"
+#include "clang/AST/ASTMutationListener.h"
 #include "clang/Lex/HeaderSearch.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/SemaInternal.h"
@@ -475,6 +476,9 @@ Sema::ActOnModuleDecl(SourceLocation StartLoc, SourceLocation ModuleLoc,
 
   getASTContext().setCurrentNamedModule(Mod);
 
+  if (auto *Listener = getASTMutationListener())
+    Listener->EnteringModulePurview();
+
   // We already potentially made an implicit import (in the case of a module
   // implementation unit importing its interface).  Make this module visible
   // and return the import decl to be added to the current TU.
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index b2a078b6d80f4..c3cf242391b74 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -5026,27 +5026,6 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
     GetDeclRef(D);
   }
 
-  // If the translation unit has an anonymous namespace, and we don't already
-  // have an update block for it, write it as an update block.
-  // FIXME: Why do we not do this if there's already an update block?
-  if (NamespaceDecl *NS = TU->getAnonymousNamespace()) {
-    ASTWriter::UpdateRecord &Record = DeclUpdates[TU];
-    if (Record.empty())
-      Record.push_back(DeclUpdate(UPD_CXX_ADDED_ANONYMOUS_NAMESPACE, NS));
-  }
-
-  // Add update records for all mangling numbers and static local numbers.
-  // These aren't really update records, but this is a convenient way of
-  // tagging this rare extra data onto the declarations.
-  for (const auto &Number : Context.MangleNumbers)
-    if (!Number.first->isFromASTFile())
-      DeclUpdates[Number.first].push_back(DeclUpdate(UPD_MANGLING_NUMBER,
-                                                     Number.second));
-  for (const auto &Number : Context.StaticLocalNumbers)
-    if (!Number.first->isFromASTFile())
-      DeclUpdates[Number.first].push_back(DeclUpdate(UPD_STATIC_LOCAL_NUMBER,
-                                                     Number.second));
-
   // Make sure visible decls, added to DeclContexts previously loaded from
   // an AST file, are registered for serialization. Likewise for template
   // specializations added to imported templates.
@@ -5315,6 +5294,41 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
   return backpatchSignature();
 }
 
+void ASTWriter::EnteringModulePurview() {
+  // In C++20 named modules, all entities before entering the module purview
+  // lives in the GMF.
+  if (GeneratingReducedBMI)
+    DeclUpdatesFromGMF.swap(DeclUpdates);
+}
+
+// Add update records for all mangling numbers and static local numbers.
+// These aren't really update records, but this is a convenient way of
+// tagging this rare extra data onto the declarations.
+void ASTWriter::AddedManglingNumber(const Decl *D, unsigned Number) {
+  if (D->isFromASTFile())
+    return;
+
+  DeclUpdates[D].push_back(DeclUpdate(UPD_MANGLING_NUMBER, Number));
+}
+void ASTWriter::AddedStaticLocalNumbers(const Decl *D, unsigned Number) {
+  if (D->isFromASTFile())
+    return;
+
+  DeclUpdates[D].push_back(DeclUpdate(UPD_STATIC_LOCAL_NUMBER, Number));
+}
+
+void ASTWriter::AddedAnonymousNamespace(const TranslationUnitDecl *TU,
+                                        NamespaceDecl *AnonNamespace) {
+  // If the translation unit has an anonymous namespace, and we don't already
+  // have an update block for it, write it as an update block.
+  // FIXME: Why do we not do this if there's already an update block?
+  if (NamespaceDecl *NS = TU->getAnonymousNamespace()) {
+    ASTWriter::UpdateRecord &Record = DeclUpdates[TU];
+    if (Record.empty())
+      Record.push_back(DeclUpdate(UPD_CXX_ADDED_ANONYMOUS_NAMESPACE, NS));
+  }
+}
+
 void ASTWriter::WriteDeclAndTypes(ASTContext &Context) {
   // Keep writing types, declarations, and declaration update records
   // until we've emitted all of them.
@@ -5860,6 +5874,14 @@ DeclID ASTWriter::GetDeclRef(const Decl *D) {
     return 0;
   }
 
+  // If the DeclUpdate from the GMF gets touched, emit it.
+  if (auto *Iter = DeclUpdatesFromGMF.find(D);
+      Iter != DeclUpdatesFromGMF.end()) {
+    for (DeclUpdate &Update : Iter->second)
+      DeclUpdates[D].push_back(Update);
+    DeclUpdatesFromGMF.erase(Iter);
+  }
+
   // If D comes from an AST file, its declaration ID is already known and
   // fixed.
   if (D->isFromASTFile())
diff --git a/clang/test/Modules/reduced-bmi-empty-module-purview.cppm b/clang/test/Modules/reduced-bmi-empty-module-purview.cppm
new file mode 100644
index 0000000000000..54a1e9b77e6e6
--- /dev/null
+++ b/clang/test/Modules/reduced-bmi-empty-module-purview.cppm
@@ -0,0 +1,96 @@
+// Test that we won't write additional information into the Reduced BMI if the 
+// module purview is empty.
+//
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 %t/M.cppm -emit-reduced-module-interface -o %t/M.pcm
+// RUN: %clang_cc1 -std=c++20 %t/A.cppm -emit-reduced-module-interface -o %t/A.pcm \
+// RUN:     -fmodule-file=M=%t/M.pcm
+// RUN: llvm-bcanalyzer --dump --disable-histogram --show-binary-blobs %t/A.pcm > %t/A.dump
+// RUN: cat %t/A.dump | FileCheck %t/A.cppm
+//
+// RUN: %clang_cc1 -std=c++20 %t/A1.cppm -emit-reduced-module-interface -o %t/A1.pcm \
+// RUN:     -fmodule-file=M=%t/M.pcm
+// RUN: llvm-bcanalyzer --dump --disable-histogram --show-binary-blobs %t/A1.pcm > %t/A1.dump
+// RUN: cat %t/A1.dump | FileCheck %t/A1.cppm
+
+//--- foo.h
+namespace ns {
+template <class C>
+class A {
+
+};
+
+extern template class A<short>;
+
+inline A<int> a() { return A<int>(); }
+template <class T>
+A<T> _av_ = A<T>();
+
+auto _av_1 = _av_<int>;
+auto _av_2 = _av_<double>;
+
+template <>
+class A<void> {
+
+};
+
+void func(A<int>, ...) {
+
+}
+
+}
+
+struct S {
+    union {
+        unsigned int V;
+        struct {
+            int v1;
+            int v2;
+            ns::A<int> a1;
+        } WESQ;
+    };
+
+    union {
+        double d;
+        struct {
+            int v1;
+            unsigned v2;
+            ns::A<unsigned> a1;
+        } Another;
+    };
+};
+
+//--- M.cppm
+module;
+#include "foo.h"
+export module M;
+export namespace nv {
+    using ns::A;
+    using ns::a;
+    using ns::_av_;
+
+    using ns::func;
+}
+using ::S;
+
+//--- A.cppm
+module;
+#include "foo.h"
+export module A;
+import M;
+
+// CHECK-NOT: <DECL_CXX_RECORD
+// CHECK-NOT: <DECL_UPDATE_OFFSETS
+
+//--- A1.cppm
+module;
+import M;
+#include "foo.h"
+export module A;
+
+// CHECK-NOT: <DECL_CXX_RECORD
+// CHECK-NOT: <DECL_UPDATE_OFFSETS
+

From 3fbb815c8c82c87f9a6e59e053ab60bad08589f6 Mon Sep 17 00:00:00 2001
From: Luke Lau <luke@igalia.com>
Date: Thu, 18 Apr 2024 11:04:06 +0800
Subject: [PATCH 811/886] [RISCV] Check that VLMAX is the same when demanding
 exact VL (#89080)

---
 llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
index fa37d1ccccd73..331253e39c0ac 100644
--- a/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
@@ -616,7 +616,7 @@ class VSETVLIInfo {
     if (SEWLMULRatioOnly)
       return false;
 
-    if (Used.VLAny && !hasSameAVL(Require))
+    if (Used.VLAny && !(hasSameAVL(Require) && hasSameVLMAX(Require)))
       return false;
 
     if (Used.VLZeroness && !hasEquallyZeroAVL(Require, MRI))

From 3e2aad412ada8f1db88c9e0f984a302deed74f4b Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 17 Apr 2024 21:03:35 -0700
Subject: [PATCH 812/886] [RISCV] Speed up RISCVRegisterInfo::needsFrameBaseReg
 when frame pointer isn't used. NFC (#89163)

The callee saved size is only used if there is a frame pointer. Sink the
code onto the frame pointer only path.
---
 llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp | 22 +++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp b/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
index 84af6eec40ee6..367a62e830cbf 100644
--- a/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVRegisterInfo.cpp
@@ -607,20 +607,22 @@ bool RISCVRegisterInfo::needsFrameBaseReg(MachineInstr *MI,
   const MachineFrameInfo &MFI = MF.getFrameInfo();
   const RISCVFrameLowering *TFI = getFrameLowering(MF);
   const MachineRegisterInfo &MRI = MF.getRegInfo();
-  unsigned CalleeSavedSize = 0;
   Offset += getFrameIndexInstrOffset(MI, FIOperandNum);
 
-  // Estimate the stack size used to store callee saved registers(
-  // excludes reserved registers).
-  BitVector ReservedRegs = getReservedRegs(MF);
-  for (const MCPhysReg *R = MRI.getCalleeSavedRegs(); MCPhysReg Reg = *R; ++R) {
-    if (!ReservedRegs.test(Reg))
-      CalleeSavedSize += getSpillSize(*getMinimalPhysRegClass(Reg));
-  }
+  if (TFI->hasFP(MF) && !shouldRealignStack(MF)) {
+    // Estimate the stack size used to store callee saved registers(
+    // excludes reserved registers).
+    unsigned CalleeSavedSize = 0;
+    BitVector ReservedRegs = getReservedRegs(MF);
+    for (const MCPhysReg *R = MRI.getCalleeSavedRegs(); MCPhysReg Reg = *R;
+         ++R) {
+      if (!ReservedRegs.test(Reg))
+        CalleeSavedSize += getSpillSize(*getMinimalPhysRegClass(Reg));
+    }
 
-  int64_t MaxFPOffset = Offset - CalleeSavedSize;
-  if (TFI->hasFP(MF) && !shouldRealignStack(MF))
+    int64_t MaxFPOffset = Offset - CalleeSavedSize;
     return !isFrameOffsetLegal(MI, RISCV::X8, MaxFPOffset);
+  }
 
   // Assume 128 bytes spill slots size to estimate the maximum possible
   // offset relative to the stack pointer.

From 8b37ec1f7bda40c240f7bfda6737df5043860103 Mon Sep 17 00:00:00 2001
From: Will Hawkins <hawkinsw@obs.cr>
Date: Thu, 18 Apr 2024 00:29:57 -0400
Subject: [PATCH 813/886] [libc++][NFC] Add additional tests for begin/end of
 std::ranges::take_view (#79085)

Add additional tests for `begin`/`end` of `std::ranges::take_view`.

In partial fulfillment of #72406.
---
 .../range.adaptors/range.take/begin.pass.cpp  | 100 +++++++++++++-----
 .../range.adaptors/range.take/end.pass.cpp    |  14 +++
 .../ranges/range.adaptors/range.take/types.h  |  50 +++++++++
 3 files changed, 139 insertions(+), 25 deletions(-)

diff --git a/libcxx/test/std/ranges/range.adaptors/range.take/begin.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.take/begin.pass.cpp
index 1873481d73225..9f11a991535e7 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.take/begin.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.take/begin.pass.cpp
@@ -11,8 +11,9 @@
 // constexpr auto begin() requires (!simple-view<V>);
 // constexpr auto begin() const requires range<const V>;
 
-#include <ranges>
 #include <cassert>
+#include <ranges>
+#include <utility>
 
 #include "test_macros.h"
 #include "test_iterators.h"
@@ -27,55 +28,104 @@ struct NonCommonSimpleView : std::ranges::view_base {
 static_assert(std::ranges::sized_range<NonCommonSimpleView>);
 static_assert(!std::ranges::sized_range<const NonCommonSimpleView>);
 
+using CommonInputIterPtrConstInt        = common_input_iterator<const int*>;
+using CountedCommonInputIterPtrConstInt = std::counted_iterator<CommonInputIterPtrConstInt>;
+
 constexpr bool test() {
   int buffer[8] = {1, 2, 3, 4, 5, 6, 7, 8};
 
-  // sized_range && random_access_iterator
+  // simple-view<V> && sized_range<V> && random_access_range<V>
   {
-    std::ranges::take_view<SizedRandomAccessView> tv(SizedRandomAccessView(buffer), 4);
-    assert(tv.begin() == SizedRandomAccessView(buffer).begin());
-    ASSERT_SAME_TYPE(decltype(tv.begin()), RandomAccessIter);
-  }
+    using ViewTested = SizedRandomAccessView;
+    static_assert(simple_view<ViewTested>);
+    static_assert(std::ranges::sized_range<ViewTested>);
+    static_assert(std::ranges::random_access_range<ViewTested>);
 
-  {
-    const std::ranges::take_view<SizedRandomAccessView> tv(SizedRandomAccessView(buffer), 4);
-    assert(tv.begin() == SizedRandomAccessView(buffer).begin());
+    std::ranges::take_view<ViewTested> tv(ViewTested(buffer), 4);
+    assert(tv.begin() == ViewTested(buffer).begin());
     ASSERT_SAME_TYPE(decltype(tv.begin()), RandomAccessIter);
-  }
 
-  // sized_range && !random_access_iterator
-  {
-    std::ranges::take_view<SizedForwardView> tv(SizedForwardView{buffer}, 4);
-    assert(tv.begin() == std::counted_iterator<ForwardIter>(ForwardIter(buffer), 4));
-    ASSERT_SAME_TYPE(decltype(tv.begin()), std::counted_iterator<ForwardIter>);
+    const std::ranges::take_view<ViewTested> ctv(ViewTested(buffer), 4);
+    assert(ctv.begin() == ViewTested(buffer).begin());
+    ASSERT_SAME_TYPE(decltype(ctv.begin()), RandomAccessIter);
   }
 
+  // simple-view<V> && sized_range<V> && !random_access_range<V>
   {
-    const std::ranges::take_view<SizedForwardView> tv(SizedForwardView{buffer}, 4);
-    assert(tv.begin() == std::counted_iterator<ForwardIter>(ForwardIter(buffer), 4));
+    using ViewTested = SizedForwardView;
+    static_assert(simple_view<ViewTested>);
+    static_assert(std::ranges::sized_range<ViewTested>);
+    static_assert(!std::ranges::random_access_range<ViewTested>);
+
+    std::ranges::take_view<ViewTested> tv(ViewTested{buffer}, 16);                    // underlying size is 8
+    assert(tv.begin() == std::counted_iterator<ForwardIter>(ForwardIter(buffer), 8)); // expect min(8, 16)
     ASSERT_SAME_TYPE(decltype(tv.begin()), std::counted_iterator<ForwardIter>);
+
+    const std::ranges::take_view<ViewTested> ctv(ViewTested{buffer}, 4);
+    assert(ctv.begin() == std::counted_iterator<ForwardIter>(ForwardIter(buffer), 4));
+    ASSERT_SAME_TYPE(decltype(ctv.begin()), std::counted_iterator<ForwardIter>);
   }
 
-  // !sized_range
+  // simple-view<V> && !sized_range<V>
   {
-    std::ranges::take_view<MoveOnlyView> tv(MoveOnlyView{buffer}, 4);
+    using ViewTested = MoveOnlyView;
+    static_assert(simple_view<ViewTested>);
+    std::ranges::take_view<ViewTested> tv(ViewTested{buffer}, 4);
     assert(tv.begin() == std::counted_iterator<int*>(buffer, 4));
     ASSERT_SAME_TYPE(decltype(tv.begin()), std::counted_iterator<int*>);
+
+    const std::ranges::take_view<ViewTested> ctv(ViewTested{buffer}, 4);
+    assert(ctv.begin() == std::counted_iterator<int*>(buffer, 4));
+    ASSERT_SAME_TYPE(decltype(ctv.begin()), std::counted_iterator<int*>);
   }
 
+  // simple-view<V> && sized_range<V> && !sized_range<const V>
   {
-    const std::ranges::take_view<MoveOnlyView> tv(MoveOnlyView{buffer}, 4);
-    assert(tv.begin() == std::counted_iterator<int*>(buffer, 4));
+    using ViewTested = NonCommonSimpleView;
+    static_assert(simple_view<ViewTested>);
+    static_assert(std::ranges::sized_range<ViewTested>);
+    static_assert(!std::ranges::sized_range<const ViewTested>);
+
+    std::ranges::take_view<ViewTested> tv{};
     ASSERT_SAME_TYPE(decltype(tv.begin()), std::counted_iterator<int*>);
+    ASSERT_SAME_TYPE(decltype(std::as_const(tv).begin()), std::counted_iterator<int*>);
   }
 
-  // simple-view<V> && sized_range<V> && !size_range<!V>
+  //  !simple-view<V> && !sized_range<V>
   {
-    std::ranges::take_view<NonCommonSimpleView> tv{};
-    ASSERT_SAME_TYPE(decltype(tv.begin()), std::counted_iterator<int*>);
-    ASSERT_SAME_TYPE(decltype(std::as_const(tv).begin()), std::counted_iterator<int*>);
+    using ViewTested = NonSimpleNonSizedView;
+    static_assert(!simple_view<ViewTested>);
+    static_assert(!std::ranges::sized_range<ViewTested>);
+
+    std::ranges::take_view<ViewTested> tv{ViewTested{buffer, buffer + 2}, 4};
+    // The count for the counted iterator is the count of the take_view (i.e., 4)
+    assert(tv.begin() == CountedCommonInputIterPtrConstInt(CommonInputIterPtrConstInt(buffer), 4));
+    ASSERT_SAME_TYPE(decltype(tv.begin()), CountedCommonInputIterPtrConstInt);
   }
 
+  // !simple-view<V> && sized_range<V>
+  {
+    using ViewTested = NonSimpleSizedView;
+    static_assert(!simple_view<ViewTested>);
+    static_assert(std::ranges::sized_range<ViewTested>);
+
+    std::ranges::take_view<ViewTested> tv{ViewTested{buffer, buffer + 2}, 4};
+    // The count for the counted iterator is the min(2, 4) (i.e., 2).
+    assert(tv.begin() == CountedCommonInputIterPtrConstInt(CommonInputIterPtrConstInt(buffer), 2));
+    ASSERT_SAME_TYPE(decltype(tv.begin()), CountedCommonInputIterPtrConstInt);
+  }
+
+  // !simple-view<V> && sized_range<V> && random_access_range<V>
+  {
+    using ViewTested = NonSimpleSizedRandomView;
+    static_assert(!simple_view<ViewTested>);
+    static_assert(std::ranges::sized_range<ViewTested>);
+    static_assert(std::ranges::random_access_range<ViewTested>);
+
+    std::ranges::take_view<ViewTested> tv{ViewTested{buffer, buffer + 2}, 4};
+    assert(tv.begin() == random_access_iterator<const int*>(buffer));
+    ASSERT_SAME_TYPE(decltype(tv.begin()), random_access_iterator<const int*>);
+  }
   return true;
 }
 
diff --git a/libcxx/test/std/ranges/range.adaptors/range.take/end.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.take/end.pass.cpp
index 6cab05daa9e05..eddb39af27121 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.take/end.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.take/end.pass.cpp
@@ -69,6 +69,20 @@ constexpr bool test() {
     assert(tv.end() == std::ranges::next(tv.begin(), 8));
   }
 
+  {
+    // __iterator<false> has base with type std::ranges::sentinel_t<NonSimpleViewNonSized>; adding a const qualifier
+    // would change the equality.
+    std::ranges::take_view<NonSimpleNonSizedView> tvns(NonSimpleNonSizedView{buffer, buffer + 8}, 0);
+    static_assert(!std::is_same_v<decltype(tvns.end().base()), std::ranges::sentinel_t<const NonSimpleNonSizedView>>);
+  }
+
+  {
+    // __iterator<true> has base with type std::ranges::sentinel_t<const NonSimpleViewNonSized>; adding a const qualifier
+    // would not change the equality.
+    std::ranges::take_view<SimpleViewNonSized> tvs(SimpleViewNonSized{buffer, buffer + 8}, 0);
+    static_assert(std::is_same_v<decltype(tvs.end().base()), std::ranges::sentinel_t<const SimpleViewNonSized>>);
+  }
+
   return true;
 }
 
diff --git a/libcxx/test/std/ranges/range.adaptors/range.take/types.h b/libcxx/test/std/ranges/range.adaptors/range.take/types.h
index db80e68bb21af..7590ce33bffc1 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.take/types.h
+++ b/libcxx/test/std/ranges/range.adaptors/range.take/types.h
@@ -65,4 +65,54 @@ struct View : std::ranges::view_base {
   int* end_;
 };
 
+template <template <class...> typename Iter, bool Simple, bool Sized>
+struct CommonInputView : std::ranges::view_base {
+  constexpr explicit CommonInputView(int* b, int* e) : begin_(b), end_(e) {}
+
+  constexpr Iter<int*> begin() const { return Iter<int*>(begin_); }
+  constexpr Iter<int*> end() const { return Iter<int*>(end_); }
+
+  constexpr Iter<const int*> begin()
+    requires(!Simple)
+  {
+    return Iter<const int*>(begin_);
+  }
+  constexpr Iter<const int*> end()
+    requires(!Simple)
+  {
+    return Iter<const int*>(end_);
+  }
+
+  constexpr auto size() const
+    requires Sized
+  {
+    return end_ - begin_;
+  }
+
+private:
+  int* begin_;
+  int* end_;
+};
+
+using NonSimpleNonSizedView = CommonInputView<common_input_iterator, /*Simple=*/false, /*Sized=*/false>;
+static_assert(std::ranges::view<NonSimpleNonSizedView>);
+static_assert(!simple_view<NonSimpleNonSizedView>);
+static_assert(!std::ranges::sized_range<NonSimpleNonSizedView>);
+
+using SimpleViewNonSized = CommonInputView<common_input_iterator, /*Simple=*/true, /*Sized=*/false>;
+static_assert(std::ranges::view<SimpleViewNonSized>);
+static_assert(simple_view<SimpleViewNonSized>);
+static_assert(!std::ranges::sized_range<SimpleViewNonSized>);
+
+using NonSimpleSizedView = CommonInputView<common_input_iterator, /*Simple=*/false, /*Sized=*/true>;
+static_assert(std::ranges::view<NonSimpleSizedView>);
+static_assert(!simple_view<NonSimpleSizedView>);
+static_assert(std::ranges::sized_range<NonSimpleSizedView>);
+
+using NonSimpleSizedRandomView = CommonInputView<random_access_iterator, /*Simple=*/false, /*Sized=*/true>;
+static_assert(std::ranges::view<NonSimpleSizedRandomView>);
+static_assert(!simple_view<NonSimpleSizedRandomView>);
+static_assert(std::ranges::sized_range<NonSimpleSizedRandomView>);
+static_assert(std::ranges::random_access_range<NonSimpleSizedRandomView>);
+
 #endif // TEST_STD_RANGES_RANGE_ADAPTORS_RANGE_TAKE_TYPES_H

From 808d794a45e169601ff16f72beae2f7bd79342a2 Mon Sep 17 00:00:00 2001
From: Will Hawkins <hawkinsw@obs.cr>
Date: Thu, 18 Apr 2024 00:32:23 -0400
Subject: [PATCH 814/886] [libc++][NFC] Centralize test for support of == and
 != in ranges (#78481)

Previously, tests for whether comparison using == was supported by
iterators derived from ranges adaptors was spread throughout the testing
codebase. This PR centralizes the implementation of those tests.
---
 .../range.elements/iterator/compare.pass.cpp  |  4 +-
 .../range.elements/sentinel/equality.pass.cpp | 37 +++++++++----------
 .../range.filter/iterator/compare.pass.cpp    |  8 ++--
 .../range.join.sentinel/eq.pass.cpp           | 12 +++---
 .../range.lazy.split.inner/equal.pass.cpp     |  9 ++---
 .../range.lazy.split.outer/equal.pass.cpp     |  9 ++---
 .../sentinel/equality.pass.cpp                | 36 ++++++++----------
 .../range.take.sentinel/eq.pass.cpp           |  9 +----
 .../range.zip/iterator/compare.pass.cpp       |  4 +-
 .../range.zip/sentinel/eq.pass.cpp            | 30 +++++++--------
 libcxx/test/support/test_range.h              | 12 ++++++
 11 files changed, 79 insertions(+), 91 deletions(-)

diff --git a/libcxx/test/std/ranges/range.adaptors/range.elements/iterator/compare.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.elements/iterator/compare.pass.cpp
index 16df3e40bd77a..4dd52a80a0eba 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.elements/iterator/compare.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.elements/iterator/compare.pass.cpp
@@ -27,6 +27,7 @@
 #include <tuple>
 
 #include "test_iterators.h"
+#include "test_range.h"
 
 constexpr void compareOperatorTest(const auto& iter1, const auto& iter2) {
   assert(!(iter1 < iter1));
@@ -139,8 +140,7 @@ constexpr bool test() {
     auto it = ev.begin();
 
     using ElemIter = decltype(it);
-    static_assert(!std::invocable<std::equal_to<>, ElemIter, ElemIter>);
-    static_assert(!std::invocable<std::not_equal_to<>, ElemIter, ElemIter>);
+    static_assert(!weakly_equality_comparable_with<ElemIter, ElemIter>);
     inequalityOperatorsDoNotExistTest(it, it);
   }
 
diff --git a/libcxx/test/std/ranges/range.adaptors/range.elements/sentinel/equality.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.elements/sentinel/equality.pass.cpp
index df95e07c97d97..d8a3149398bf7 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.elements/sentinel/equality.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.elements/sentinel/equality.pass.cpp
@@ -17,6 +17,7 @@
 #include <ranges>
 
 #include "../types.h"
+#include "test_range.h"
 
 template <bool Const>
 struct Iter {
@@ -63,37 +64,33 @@ struct Range : TupleBufferView {
 using R                = Range<Sent>;
 using CrossComparableR = Range<CrossComparableSent>;
 
-// Test Constraint
-template <class I, class S>
-concept HasEqual = requires(const I i, const S s) { i == s; };
-
 using std::ranges::elements_view;
 using std::ranges::iterator_t;
 using std::ranges::sentinel_t;
 
-static_assert(HasEqual<iterator_t<elements_view<R, 0>>, //
-                       sentinel_t<elements_view<R, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<elements_view<R, 0>>, //
+                                              sentinel_t<elements_view<R, 0>>>);
 
-static_assert(!HasEqual<iterator_t<const elements_view<R, 0>>, //
-                        sentinel_t<elements_view<R, 0>>>);
+static_assert(!weakly_equality_comparable_with<iterator_t<const elements_view<R, 0>>, //
+                                               sentinel_t<elements_view<R, 0>>>);
 
-static_assert(!HasEqual<iterator_t<elements_view<R, 0>>, //
-                        sentinel_t<const elements_view<R, 0>>>);
+static_assert(!weakly_equality_comparable_with<iterator_t<elements_view<R, 0>>, //
+                                               sentinel_t<const elements_view<R, 0>>>);
 
-static_assert(HasEqual<iterator_t<const elements_view<R, 0>>, //
-                       sentinel_t<const elements_view<R, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const elements_view<R, 0>>, //
+                                              sentinel_t<const elements_view<R, 0>>>);
 
-static_assert(HasEqual<iterator_t<elements_view<R, 0>>, //
-                       sentinel_t<elements_view<R, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<elements_view<CrossComparableR, 0>>, //
+                                              sentinel_t<elements_view<CrossComparableR, 0>>>);
 
-static_assert(HasEqual<iterator_t<const elements_view<CrossComparableR, 0>>, //
-                       sentinel_t<elements_view<CrossComparableR, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const elements_view<CrossComparableR, 0>>, //
+                                              sentinel_t<elements_view<CrossComparableR, 0>>>);
 
-static_assert(HasEqual<iterator_t<elements_view<CrossComparableR, 0>>, //
-                       sentinel_t<const elements_view<CrossComparableR, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<elements_view<CrossComparableR, 0>>, //
+                                              sentinel_t<const elements_view<CrossComparableR, 0>>>);
 
-static_assert(HasEqual<iterator_t<const elements_view<CrossComparableR, 0>>, //
-                       sentinel_t<const elements_view<CrossComparableR, 0>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const elements_view<CrossComparableR, 0>>, //
+                                              sentinel_t<const elements_view<CrossComparableR, 0>>>);
 
 template <class R, bool ConstIter, bool ConstSent>
 constexpr void testOne() {
diff --git a/libcxx/test/std/ranges/range.adaptors/range.filter/iterator/compare.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.filter/iterator/compare.pass.cpp
index 78881e8ac6df1..11cba3c1ba308 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.filter/iterator/compare.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.filter/iterator/compare.pass.cpp
@@ -17,12 +17,12 @@
 #include <cassert>
 #include <concepts>
 #include <utility>
+
 #include "test_iterators.h"
 #include "test_macros.h"
-#include "../types.h"
+#include "test_range.h"
 
-template <class T>
-concept has_equal = requires (T const& x, T const& y) { { x == y }; };
+#include "../types.h"
 
 template <class Iterator>
 constexpr void test() {
@@ -76,7 +76,7 @@ constexpr bool tests() {
     using Sentinel = sentinel_wrapper<Iterator>;
     using FilterView = std::ranges::filter_view<minimal_view<Iterator, Sentinel>, AlwaysTrue>;
     using FilterIterator = std::ranges::iterator_t<FilterView>;
-    static_assert(!has_equal<FilterIterator>);
+    static_assert(!weakly_equality_comparable_with<FilterIterator, FilterIterator>);
   }
 
   return true;
diff --git a/libcxx/test/std/ranges/range.adaptors/range.join/range.join.sentinel/eq.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.join/range.join.sentinel/eq.pass.cpp
index bc7d4bec94d3e..9d6cb76902622 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.join/range.join.sentinel/eq.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.join/range.join.sentinel/eq.pass.cpp
@@ -19,9 +19,7 @@
 #include <type_traits>
 
 #include "../types.h"
-
-template <class Iter, class Sent>
-concept EqualityComparable = std::invocable<std::equal_to<>, const Iter&, const Sent&> ;
+#include "test_range.h"
 
 using Iterator = random_access_iterator<BufferView<int*>*>;
 using ConstIterator = random_access_iterator<const BufferView<int*>*>;
@@ -53,10 +51,10 @@ struct ConstComparableView : BufferView<BufferView<int*>*> {
   constexpr auto end() const { return ConstComparableSentinel<true>(ConstIterator(data_ + size_)); }
 };
 
-static_assert(EqualityComparable<std::ranges::iterator_t<ConstComparableView>,
-                                 std::ranges::sentinel_t<const ConstComparableView>>);
-static_assert(EqualityComparable<std::ranges::iterator_t<const ConstComparableView>,
-                                 std::ranges::sentinel_t<ConstComparableView>>);
+static_assert(weakly_equality_comparable_with<std::ranges::iterator_t<ConstComparableView>,
+                                              std::ranges::sentinel_t<const ConstComparableView>>);
+static_assert(weakly_equality_comparable_with<std::ranges::iterator_t<const ConstComparableView>,
+                                              std::ranges::sentinel_t<ConstComparableView>>);
 
 constexpr bool test() {
   int buffer[4][4] = {{1111, 2222, 3333, 4444}, {555, 666, 777, 888}, {99, 1010, 1111, 1212}, {13, 14, 15, 16}};
diff --git a/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.inner/equal.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.inner/equal.pass.cpp
index 5a83a05ead919..dbf3bfa126ae8 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.inner/equal.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.inner/equal.pass.cpp
@@ -17,13 +17,10 @@
 
 #include <concepts>
 #include <string_view>
+
 #include "../types.h"
 
-template <class Iter>
-concept CanCallEquals = requires(const Iter& i) {
-  i == i;
-  i != i;
-};
+#include "test_range.h"
 
 constexpr bool test() {
   // When `View` is a forward range, `inner-iterator` supports both overloads of `operator==`.
@@ -56,7 +53,7 @@ constexpr bool test() {
     auto b = val.begin();
     std::same_as<std::default_sentinel_t> decltype(auto) e = val.end();
 
-    static_assert(!CanCallEquals<decltype(b)>);
+    static_assert(!weakly_equality_comparable_with<decltype(b), decltype(b)>);
 
     assert(!(b == std::default_sentinel));
     assert(b != std::default_sentinel);
diff --git a/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.outer/equal.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.outer/equal.pass.cpp
index 49cac70894732..6cbc98a94645f 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.outer/equal.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.lazy.split/range.lazy.split.outer/equal.pass.cpp
@@ -17,13 +17,10 @@
 
 #include <concepts>
 #include <string_view>
+
 #include "../types.h"
 
-template <class Iter>
-concept CanCallEquals = requires(const Iter& i) {
-  i == i;
-  i != i;
-};
+#include "test_range.h"
 
 constexpr bool test() {
   // Forward range supports both overloads of `operator==`.
@@ -69,7 +66,7 @@ constexpr bool test() {
     auto b = v.begin();
     std::same_as<std::default_sentinel_t> decltype(auto) e = v.end();
 
-    static_assert(!CanCallEquals<decltype(b)>);
+    static_assert(!weakly_equality_comparable_with<decltype(b), decltype(b)>);
 
     assert(!(b == std::default_sentinel));
     assert(b != std::default_sentinel);
diff --git a/libcxx/test/std/ranges/range.adaptors/range.take.while/sentinel/equality.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.take.while/sentinel/equality.pass.cpp
index db3e5764421af..b00b3dd0bd053 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.take.while/sentinel/equality.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.take.while/sentinel/equality.pass.cpp
@@ -70,37 +70,33 @@ struct LessThan3 {
   constexpr bool operator()(int i) const { return i < 3; }
 };
 
-// Test Constraint
-template <class I, class S>
-concept HasEqual = requires(const I i, const S s) { i == s; };
-
 using std::ranges::iterator_t;
 using std::ranges::sentinel_t;
 using std::ranges::take_while_view;
 
-static_assert(HasEqual<iterator_t<take_while_view<R, LessThan3>>, //
-                       sentinel_t<take_while_view<R, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<take_while_view<R, LessThan3>>, //
+                                              sentinel_t<take_while_view<R, LessThan3>>>);
 
-static_assert(!HasEqual<iterator_t<const take_while_view<R, LessThan3>>, //
-                        sentinel_t<take_while_view<R, LessThan3>>>);
+static_assert(!weakly_equality_comparable_with<iterator_t<const take_while_view<R, LessThan3>>, //
+                                               sentinel_t<take_while_view<R, LessThan3>>>);
 
-static_assert(!HasEqual<iterator_t<take_while_view<R, LessThan3>>, //
-                        sentinel_t<const take_while_view<R, LessThan3>>>);
+static_assert(!weakly_equality_comparable_with<iterator_t<take_while_view<R, LessThan3>>, //
+                                               sentinel_t<const take_while_view<R, LessThan3>>>);
 
-static_assert(HasEqual<iterator_t<const take_while_view<R, LessThan3>>, //
-                       sentinel_t<const take_while_view<R, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const take_while_view<R, LessThan3>>, //
+                                              sentinel_t<const take_while_view<R, LessThan3>>>);
 
-static_assert(HasEqual<iterator_t<take_while_view<R, LessThan3>>, //
-                       sentinel_t<take_while_view<R, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<take_while_view<CrossComparableR, LessThan3>>, //
+                                              sentinel_t<take_while_view<CrossComparableR, LessThan3>>>);
 
-static_assert(HasEqual<iterator_t<const take_while_view<CrossComparableR, LessThan3>>, //
-                       sentinel_t<take_while_view<CrossComparableR, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const take_while_view<CrossComparableR, LessThan3>>, //
+                                              sentinel_t<take_while_view<CrossComparableR, LessThan3>>>);
 
-static_assert(HasEqual<iterator_t<take_while_view<CrossComparableR, LessThan3>>, //
-                       sentinel_t<const take_while_view<CrossComparableR, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<take_while_view<CrossComparableR, LessThan3>>, //
+                                              sentinel_t<const take_while_view<CrossComparableR, LessThan3>>>);
 
-static_assert(HasEqual<iterator_t<const take_while_view<CrossComparableR, LessThan3>>, //
-                       sentinel_t<const take_while_view<CrossComparableR, LessThan3>>>);
+static_assert(weakly_equality_comparable_with<iterator_t<const take_while_view<CrossComparableR, LessThan3>>, //
+                                              sentinel_t<const take_while_view<CrossComparableR, LessThan3>>>);
 
 template <class R, bool ConstIter, bool ConstSent>
 constexpr void testOne() {
diff --git a/libcxx/test/std/ranges/range.adaptors/range.take/range.take.sentinel/eq.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.take/range.take.sentinel/eq.pass.cpp
index e6f433e30f60d..1e4d2a4dccfa4 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.take/range.take.sentinel/eq.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.take/range.take.sentinel/eq.pass.cpp
@@ -21,6 +21,7 @@
 
 #include "test_comparisons.h"
 #include "test_iterators.h"
+#include "test_range.h"
 
 template <bool Const>
 using MaybeConstIterator = cpp20_input_iterator<std::conditional_t<Const, const int*, int*>>;
@@ -77,14 +78,6 @@ struct NonCrossConstComparableView : std::ranges::view_base {
 static_assert(std::ranges::range<NonCrossConstComparableView>);
 static_assert(std::ranges::range<const NonCrossConstComparableView>);
 
-template <class T, class U>
-concept weakly_equality_comparable_with = requires(const T& t, const U& u) {
-  t == u;
-  t != u;
-  u == t;
-  u != t;
-};
-
 constexpr bool test() {
   int buffer[8]                      = {1, 2, 3, 4, 5, 6, 7, 8};
   using CrossConstComparableTakeView = std::ranges::take_view<CrossConstComparableView>;
diff --git a/libcxx/test/std/ranges/range.adaptors/range.zip/iterator/compare.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.zip/iterator/compare.pass.cpp
index fcbff722c39b3..ed1cb0ccebd2b 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.zip/iterator/compare.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.zip/iterator/compare.pass.cpp
@@ -26,6 +26,8 @@
 #include <compare>
 
 #include "test_iterators.h"
+#include "test_range.h"
+
 #include "../types.h"
 
 // This is for testing that zip iterator never calls underlying iterator's >, >=, <=, !=.
@@ -240,7 +242,7 @@ constexpr bool test() {
     std::ranges::zip_view r(IterNoEqualView{buffer});
     auto it = r.begin();
     using Iter = decltype(it);
-    static_assert(!std::invocable<std::equal_to<>, Iter, Iter>);
+    static_assert(!weakly_equality_comparable_with<Iter, Iter>);
     inequalityOperatorsDoNotExistTest(it, it);
   }
   return true;
diff --git a/libcxx/test/std/ranges/range.adaptors/range.zip/sentinel/eq.pass.cpp b/libcxx/test/std/ranges/range.adaptors/range.zip/sentinel/eq.pass.cpp
index 5db7372110814..04542c3ae4d14 100644
--- a/libcxx/test/std/ranges/range.adaptors/range.zip/sentinel/eq.pass.cpp
+++ b/libcxx/test/std/ranges/range.adaptors/range.zip/sentinel/eq.pass.cpp
@@ -18,6 +18,7 @@
 #include <tuple>
 
 #include "../types.h"
+#include "test_range.h"
 
 using Iterator = random_access_iterator<int*>;
 using ConstIterator = random_access_iterator<const int*>;
@@ -54,11 +55,6 @@ struct ConstIncompatibleView : std::ranges::view_base {
   sentinel_wrapper<forward_iterator<const int*>> end() const;
 };
 
-// clang-format off
-template <class Iter, class Sent>
-concept EqualComparable = std::invocable<std::equal_to<>, const Iter&, const Sent&>;
-// clang-format on
-
 constexpr bool test() {
   int buffer1[4] = {1, 2, 3, 4};
   int buffer2[5] = {1, 2, 3, 4, 5};
@@ -95,10 +91,10 @@ constexpr bool test() {
     using ConstSentinel = std::ranges::sentinel_t<const decltype(v)>;
     static_assert(!std::is_same_v<Sentinel, ConstSentinel>);
 
-    static_assert(EqualComparable<Iter, Sentinel>);
-    static_assert(!EqualComparable<ConstIter, Sentinel>);
-    static_assert(EqualComparable<Iter, ConstSentinel>);
-    static_assert(EqualComparable<ConstIter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<Iter, Sentinel>);
+    static_assert(!weakly_equality_comparable_with<ConstIter, Sentinel>);
+    static_assert(weakly_equality_comparable_with<Iter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<ConstIter, ConstSentinel>);
   }
 
   {
@@ -120,10 +116,10 @@ constexpr bool test() {
     using ConstSentinel = std::ranges::sentinel_t<const decltype(v)>;
     static_assert(!std::is_same_v<Sentinel, ConstSentinel>);
 
-    static_assert(EqualComparable<Iter, Sentinel>);
-    static_assert(EqualComparable<ConstIter, Sentinel>);
-    static_assert(EqualComparable<Iter, ConstSentinel>);
-    static_assert(EqualComparable<ConstIter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<Iter, Sentinel>);
+    static_assert(weakly_equality_comparable_with<ConstIter, Sentinel>);
+    static_assert(weakly_equality_comparable_with<Iter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<ConstIter, ConstSentinel>);
   }
 
   {
@@ -139,10 +135,10 @@ constexpr bool test() {
     using ConstSentinel = std::ranges::sentinel_t<const decltype(v)>;
     static_assert(!std::is_same_v<Sentinel, ConstSentinel>);
 
-    static_assert(EqualComparable<Iter, Sentinel>);
-    static_assert(!EqualComparable<ConstIter, Sentinel>);
-    static_assert(!EqualComparable<Iter, ConstSentinel>);
-    static_assert(EqualComparable<ConstIter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<Iter, Sentinel>);
+    static_assert(!weakly_equality_comparable_with<ConstIter, Sentinel>);
+    static_assert(!weakly_equality_comparable_with<Iter, ConstSentinel>);
+    static_assert(weakly_equality_comparable_with<ConstIter, ConstSentinel>);
   }
   return true;
 }
diff --git a/libcxx/test/support/test_range.h b/libcxx/test/support/test_range.h
index c5eeb25bb59eb..4efa26f266475 100644
--- a/libcxx/test/support/test_range.h
+++ b/libcxx/test/support/test_range.h
@@ -10,8 +10,10 @@
 #define LIBCXX_TEST_SUPPORT_TEST_RANGE_H
 
 #include <concepts>
+#include <functional>
 #include <iterator>
 #include <ranges>
+#include <type_traits>
 
 #include "test_iterators.h"
 
@@ -94,4 +96,14 @@ concept CanBePiped = requires(View&& view, T&& t) {
   { std::forward<View>(view) | std::forward<T>(t) };
 };
 
+// See [concept.equalitycomparable]
+template <class T, class U>
+concept weakly_equality_comparable_with =
+    requires(const std::remove_reference_t<T>& t, const std::remove_reference_t<U>& u) {
+      { t == u } -> std::same_as<bool>;
+      { t != u } -> std::same_as<bool>;
+      { u == t } -> std::same_as<bool>;
+      { u != t } -> std::same_as<bool>;
+    };
+
 #endif // LIBCXX_TEST_SUPPORT_TEST_RANGE_H

From 0afc884e874043bf5207d561ab82b107e8860d41 Mon Sep 17 00:00:00 2001
From: Chia <sun1011jacobi@gmail.com>
Date: Thu, 18 Apr 2024 13:45:51 +0900
Subject: [PATCH 815/886] [RISCV] Use vnclip for scalable vector saturating
 truncation. (#88648)

Similar to #75145, but for scalable vectors.

Specifically, this patch works for the below optimization case:

## Source Code
```
define void @trunc_sat_i8i16_maxmin(ptr %x, ptr %y) {
  %1 = load <vscale x 4 x i16>, ptr %x, align 16
  %2 = tail call <vscale x 4 x i16> @llvm.smax.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 -128))
  %3 = tail call <vscale x 4 x i16> @llvm.smin.v4i16(<vscale x 4 x i16> %2, <vscale x 4 x i16> splat (i16 127))
  %4 = trunc <vscale x 4 x i16> %3 to <vscale x 4 x i8>
  store <vscale x 4 x i8> %4, ptr %y, align 8
  ret void
}
```
## Before this patch
[Compiler Explorer](https://godbolt.org/z/EKc9eGvo8)
```
trunc_sat_i8i16_maxmin:
        vl1re16.v       v8, (a0)
        li      a0, -128
        vsetvli a2, zero, e16, m1, ta, ma
        vmax.vx v8, v8, a0
        li      a0, 127
        vmin.vx v8, v8, a0
        vsetvli zero, zero, e8, mf2, ta, ma
        vnsrl.wi        v8, v8, 0
        vse8.v  v8, (a1)
        ret
```
## After this patch
```
trunc_sat_i8i16_maxmin:
        vsetivli zero, 4, e8, mf4, ta, ma
        vle16.v v8, (a0)
        vnclip.wi v8, v8, 0
        vse8.v v8, (a1)
        ret
```
---
 .../Target/RISCV/RISCVInstrInfoVSDPatterns.td |  41 ++
 .../Target/RISCV/RISCVInstrInfoVVLPatterns.td |  66 ++--
 ...lip.ll => fixed-vectors-trunc-sat-clip.ll} |  39 +-
 .../RISCV/rvv/trunc-sat-clip-sdnode.ll        | 370 ++++++++++++++++++
 4 files changed, 455 insertions(+), 61 deletions(-)
 rename llvm/test/CodeGen/RISCV/rvv/{trunc-sat-clip.ll => fixed-vectors-trunc-sat-clip.ll} (90%)
 create mode 100644 llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll

diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
index 82ee4b0cbce90..b4af83a3cbf67 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -1166,6 +1166,47 @@ defm : VPatBinarySDNode_VV_VX<usubsat, "PseudoVSSUBU">;
 defm : VPatAVGADD_VV_VX_RM<avgflooru, 0b10>;
 defm : VPatAVGADD_VV_VX_RM<avgceilu, 0b00>;
 
+// 12.5. Vector Narrowing Fixed-Point Clip Instructions
+multiclass VPatTruncSatClipSDNode<VTypeInfo vti, VTypeInfo wti> {
+  defvar sew = vti.SEW;
+  defvar uminval = !sub(!shl(1, sew), 1);
+  defvar sminval = !sub(!shl(1, !sub(sew, 1)), 1);
+  defvar smaxval = !sub(0, !shl(1, !sub(sew, 1)));
+
+  let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
+                               GetVTypePredicates<wti>.Predicates) in {
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
+        (wti.Vector (smin
+          (wti.Vector (smax (wti.Vector wti.RegClass:$rs1),
+            (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), smaxval, (XLenVT srcvalue))))),
+          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), sminval, (XLenVT srcvalue))))),
+        (vti.Mask V0), VLOpFrag)),
+      (!cast<Instruction>("PseudoVNCLIP_WI_"#vti.LMul.MX#"_MASK")
+        (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
+        (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
+
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
+        (wti.Vector (smax
+          (wti.Vector (smin (wti.Vector wti.RegClass:$rs1),
+            (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), sminval, (XLenVT srcvalue))))),
+          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), smaxval, (XLenVT srcvalue))))),
+        (vti.Mask V0), VLOpFrag)),
+      (!cast<Instruction>("PseudoVNCLIP_WI_"#vti.LMul.MX#"_MASK")
+        (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
+        (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
+
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
+        (wti.Vector (umin (wti.Vector wti.RegClass:$rs1),
+          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), uminval, (XLenVT srcvalue))))), (vti.Mask V0), VLOpFrag)),
+      (!cast<Instruction>("PseudoVNCLIPU_WI_"#vti.LMul.MX#"_MASK")
+        (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
+        (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
+  }
+}
+
+foreach vtiToWti = AllWidenableIntVectors in
+  defm : VPatTruncSatClipSDNode<vtiToWti.Vti, vtiToWti.Wti>;
+
 // 15. Vector Mask Instructions
 
 // 15.1. Vector Mask-Register Logical Instructions
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
index f2d97ba147254..6c6ecb604fd03 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -2373,30 +2373,41 @@ defm : VPatAVGADDVL_VV_VX_RM<riscv_avgflooru_vl, 0b10>;
 defm : VPatAVGADDVL_VV_VX_RM<riscv_avgceilu_vl, 0b00>;
 
 // 12.5. Vector Narrowing Fixed-Point Clip Instructions
-class VPatTruncSatClipMaxMinBase<string inst,
-                                 VTypeInfo vti,
-                                 VTypeInfo wti,
-                                 SDPatternOperator op1,
-                                 int op1_value,
-                                 SDPatternOperator op2,
-                                 int op2_value> :
-  Pat<(vti.Vector (riscv_trunc_vector_vl
-        (wti.Vector (op1
-          (wti.Vector (op2
+multiclass VPatTruncSatClipVL<VTypeInfo vti, VTypeInfo wti> {
+  defvar sew = vti.SEW;
+  defvar uminval = !sub(!shl(1, sew), 1);
+  defvar sminval = !sub(!shl(1, !sub(sew, 1)), 1);
+  defvar smaxval = !sub(0, !shl(1, !sub(sew, 1)));
+
+  let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
+                               GetVTypePredicates<wti>.Predicates) in {
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
+        (wti.Vector (riscv_smin_vl
+          (wti.Vector (riscv_smax_vl
             (wti.Vector wti.RegClass:$rs1),
-            (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), op2_value, (XLenVT srcvalue))),
+            (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), smaxval, (XLenVT srcvalue))),
             (wti.Vector undef),(wti.Mask V0), VLOpFrag)),
-          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), op1_value, (XLenVT srcvalue))),
+          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), sminval, (XLenVT srcvalue))),
           (wti.Vector undef), (wti.Mask V0), VLOpFrag)),
         (vti.Mask V0), VLOpFrag)),
-      (!cast<Instruction>(inst#"_WI_"#vti.LMul.MX#"_MASK")
+      (!cast<Instruction>("PseudoVNCLIP_WI_"#vti.LMul.MX#"_MASK")
         (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
         (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
 
-class VPatTruncSatClipUMin<VTypeInfo vti,
-                           VTypeInfo wti,
-                           int uminval> :
-  Pat<(vti.Vector (riscv_trunc_vector_vl
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
+        (wti.Vector (riscv_smax_vl
+          (wti.Vector (riscv_smin_vl
+            (wti.Vector wti.RegClass:$rs1),
+            (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), sminval, (XLenVT srcvalue))),
+            (wti.Vector undef),(wti.Mask V0), VLOpFrag)),
+          (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), smaxval, (XLenVT srcvalue))),
+          (wti.Vector undef), (wti.Mask V0), VLOpFrag)),
+        (vti.Mask V0), VLOpFrag)),
+      (!cast<Instruction>("PseudoVNCLIP_WI_"#vti.LMul.MX#"_MASK")
+        (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
+        (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
+
+    def : Pat<(vti.Vector (riscv_trunc_vector_vl
         (wti.Vector (riscv_umin_vl
           (wti.Vector wti.RegClass:$rs1),
           (wti.Vector (riscv_vmv_v_x_vl (wti.Vector undef), uminval, (XLenVT srcvalue))),
@@ -2405,30 +2416,11 @@ class VPatTruncSatClipUMin<VTypeInfo vti,
       (!cast<Instruction>("PseudoVNCLIPU_WI_"#vti.LMul.MX#"_MASK")
         (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, 0,
         (vti.Mask V0), 0, GPR:$vl, vti.Log2SEW, TA_MA)>;
-
-multiclass VPatTruncSatClipMaxMin<string inst, VTypeInfo vti, VTypeInfo wti,
-  SDPatternOperator max, int maxval, SDPatternOperator min, int minval> {
-    def : VPatTruncSatClipMaxMinBase<inst, vti, wti, max, maxval, min, minval>;
-    def : VPatTruncSatClipMaxMinBase<inst, vti, wti, min, minval, max, maxval>;
-}
-
-multiclass VPatTruncSatClip<VTypeInfo vti, VTypeInfo wti> {
-  defvar sew = vti.SEW;
-  defvar uminval = !sub(!shl(1, sew), 1);
-  defvar sminval = !sub(!shl(1, !sub(sew, 1)), 1);
-  defvar smaxval = !sub(0, !shl(1, !sub(sew, 1)));
-
-  let Predicates = !listconcat(GetVTypePredicates<vti>.Predicates,
-                               GetVTypePredicates<wti>.Predicates) in {
-    defm : VPatTruncSatClipMaxMin<"PseudoVNCLIP", vti, wti, riscv_smin_vl,
-                                  sminval, riscv_smax_vl, smaxval>;
-    def : VPatTruncSatClipUMin<vti, wti, uminval>;
   }
-
 }
 
 foreach vtiToWti = AllWidenableIntVectors in
-  defm : VPatTruncSatClip<vtiToWti.Vti, vtiToWti.Wti>;
+  defm : VPatTruncSatClipVL<vtiToWti.Vti, vtiToWti.Wti>;
 
 // 13. Vector Floating-Point Instructions
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
similarity index 90%
rename from llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip.ll
rename to llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
index 42577408f71b0..a0d4718e9e851 100644
--- a/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
@@ -8,11 +8,8 @@ declare <4 x i32> @llvm.smin.v4i32(<4 x i32>, <4 x i32>)
 declare <4 x i64> @llvm.smax.v4i64(<4 x i64>, <4 x i64>)
 declare <4 x i64> @llvm.smin.v4i64(<4 x i64>, <4 x i64>)
 
-declare <4 x i16> @llvm.umax.v4i16(<4 x i16>, <4 x i16>)
 declare <4 x i16> @llvm.umin.v4i16(<4 x i16>, <4 x i16>)
-declare <4 x i32> @llvm.umax.v4i32(<4 x i32>, <4 x i32>)
 declare <4 x i32> @llvm.umin.v4i32(<4 x i32>, <4 x i32>)
-declare <4 x i64> @llvm.umax.v4i64(<4 x i64>, <4 x i64>)
 declare <4 x i64> @llvm.umin.v4i64(<4 x i64>, <4 x i64>)
 
 define void @trunc_sat_i8i16_maxmin(ptr %x, ptr %y) {
@@ -110,10 +107,9 @@ define void @trunc_sat_u8u16_maxmin(ptr %x, ptr %y) {
 ; CHECK-NEXT:    vse8.v v8, (a1)
 ; CHECK-NEXT:    ret
   %1 = load <4 x i16>, ptr %x, align 16
-  %2 = tail call <4 x i16> @llvm.umax.v4i16(<4 x i16> %1, <4 x i16> <i16 0, i16 0, i16 0, i16 0>)
-  %3 = tail call <4 x i16> @llvm.umin.v4i16(<4 x i16> %2, <4 x i16> <i16 255, i16 255, i16 255, i16 255>)
-  %4 = trunc <4 x i16> %3 to <4 x i8>
-  store <4 x i8> %4, ptr %y, align 8
+  %2 = tail call <4 x i16> @llvm.umin.v4i16(<4 x i16> %1, <4 x i16> <i16 255, i16 255, i16 255, i16 255>)
+  %3 = trunc <4 x i16> %2 to <4 x i8>
+  store <4 x i8> %3, ptr %y, align 8
   ret void
 }
 
@@ -127,9 +123,8 @@ define void @trunc_sat_u8u16_minmax(ptr %x, ptr %y) {
 ; CHECK-NEXT:    ret
   %1 = load <4 x i16>, ptr %x, align 16
   %2 = tail call <4 x i16> @llvm.umin.v4i16(<4 x i16> %1, <4 x i16> <i16 255, i16 255, i16 255, i16 255>)
-  %3 = tail call <4 x i16> @llvm.umax.v4i16(<4 x i16> %2, <4 x i16> <i16 0, i16 0, i16 0, i16 0>)
-  %4 = trunc <4 x i16> %3 to <4 x i8>
-  store <4 x i8> %4, ptr %y, align 8
+  %3 = trunc <4 x i16> %2 to <4 x i8>
+  store <4 x i8> %3, ptr %y, align 8
   ret void
 }
 
@@ -231,10 +226,9 @@ define void @trunc_sat_u16u32_minmax(ptr %x, ptr %y) {
 ; CHECK-NEXT:    vse16.v v8, (a1)
 ; CHECK-NEXT:    ret
   %1 = load <4 x i32>, ptr %x, align 32
-  %2 = tail call <4 x i32> @llvm.umax.v4i32(<4 x i32> %1, <4 x i32> <i32 0, i32 0, i32 0, i32 0>)
-  %3 = tail call <4 x i32> @llvm.umin.v4i32(<4 x i32> %2, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
-  %4 = trunc <4 x i32> %3 to <4 x i16>
-  store <4 x i16> %4, ptr %y, align 16
+  %2 = tail call <4 x i32> @llvm.umin.v4i32(<4 x i32> %1, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
+  %3 = trunc <4 x i32> %2 to <4 x i16>
+  store <4 x i16> %3, ptr %y, align 16
   ret void
 }
 
@@ -248,9 +242,8 @@ define void @trunc_sat_u16u32_maxmin(ptr %x, ptr %y) {
 ; CHECK-NEXT:    ret
   %1 = load <4 x i32>, ptr %x, align 32
   %2 = tail call <4 x i32> @llvm.umin.v4i32(<4 x i32> %1, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
-  %3 = tail call <4 x i32> @llvm.umax.v4i32(<4 x i32> %2, <4 x i32> <i32 0, i32 0, i32 0, i32 0>)
-  %4 = trunc <4 x i32> %3 to <4 x i16>
-  store <4 x i16> %4, ptr %y, align 16
+  %3 = trunc <4 x i32> %2 to <4 x i16>
+  store <4 x i16> %3, ptr %y, align 16
   ret void
 }
 
@@ -355,10 +348,9 @@ define void @trunc_sat_u32u64_maxmin(ptr %x, ptr %y) {
 ; CHECK-NEXT:    vse32.v v10, (a1)
 ; CHECK-NEXT:    ret
   %1 = load <4 x i64>, ptr %x, align 64
-  %2 = tail call <4 x i64> @llvm.umax.v4i64(<4 x i64> %1, <4 x i64> <i64 0, i64 0, i64 0, i64 0>)
-  %3 = tail call <4 x i64> @llvm.umin.v4i64(<4 x i64> %2, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
-  %4 = trunc <4 x i64> %3 to <4 x i32>
-  store <4 x i32> %4, ptr %y, align 32
+  %2 = tail call <4 x i64> @llvm.umin.v4i64(<4 x i64> %1, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
+  %3 = trunc <4 x i64> %2 to <4 x i32>
+  store <4 x i32> %3, ptr %y, align 32
   ret void
 }
 
@@ -372,8 +364,7 @@ define void @trunc_sat_u32u64_minmax(ptr %x, ptr %y) {
 ; CHECK-NEXT:    ret
   %1 = load <4 x i64>, ptr %x, align 64
   %2 = tail call <4 x i64> @llvm.umin.v4i64(<4 x i64> %1, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
-  %3 = tail call <4 x i64> @llvm.umax.v4i64(<4 x i64> %2, <4 x i64> <i64 0, i64 0, i64 0, i64 0>)
-  %4 = trunc <4 x i64> %3 to <4 x i32>
-  store <4 x i32> %4, ptr %y, align 32
+  %3 = trunc <4 x i64> %2 to <4 x i32>
+  store <4 x i32> %3, ptr %y, align 32
   ret void
 }
diff --git a/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll
new file mode 100644
index 0000000000000..7886eb162cd70
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll
@@ -0,0 +1,370 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=riscv64 -mattr=+v -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 4 x i16> @llvm.smax.v4i16(<vscale x 4 x i16>, <vscale x 4 x i16>)
+declare <vscale x 4 x i16> @llvm.smin.v4i16(<vscale x 4 x i16>, <vscale x 4 x i16>)
+declare <vscale x 4 x i32> @llvm.smax.v4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 4 x i32> @llvm.smin.v4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 4 x i64> @llvm.smax.v4i64(<vscale x 4 x i64>, <vscale x 4 x i64>)
+declare <vscale x 4 x i64> @llvm.smin.v4i64(<vscale x 4 x i64>, <vscale x 4 x i64>)
+
+declare <vscale x 4 x i16> @llvm.umin.v4i16(<vscale x 4 x i16>, <vscale x 4 x i16>)
+declare <vscale x 4 x i32> @llvm.umin.v4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
+declare <vscale x 4 x i64> @llvm.umin.v4i64(<vscale x 4 x i64>, <vscale x 4 x i64>)
+
+define void @trunc_sat_i8i16_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i8i16_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnclip.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.smax.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 -128))
+  %3 = tail call <vscale x 4 x i16> @llvm.smin.v4i16(<vscale x 4 x i16> %2, <vscale x 4 x i16> splat (i16 127))
+  %4 = trunc <vscale x 4 x i16> %3 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %4, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_i8i16_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i8i16_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnclip.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.smin.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 127))
+  %3 = tail call <vscale x 4 x i16> @llvm.smax.v4i16(<vscale x 4 x i16> %2, <vscale x 4 x i16> splat (i16 -128))
+  %4 = trunc <vscale x 4 x i16> %3 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %4, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_i8i16_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i8i16_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    li a0, -127
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vmax.vx v8, v8, a0
+; CHECK-NEXT:    li a0, 128
+; CHECK-NEXT:    vmin.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnsrl.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.smax.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 -127))
+  %3 = tail call <vscale x 4 x i16> @llvm.smin.v4i16(<vscale x 4 x i16> %2, <vscale x 4 x i16> splat (i16 128))
+  %4 = trunc <vscale x 4 x i16> %3 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %4, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_u8u16_min(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u8u16_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.umin.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 255))
+  %3 = trunc <vscale x 4 x i16> %2 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %3, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_u8u16_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u8u16_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    li a0, 127
+; CHECK-NEXT:    vsetvli a2, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vminu.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnsrl.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.umin.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 127))
+  %3 = trunc <vscale x 4 x i16> %2 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %3, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_u8u16_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u8u16_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.umin.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 255))
+  %3 = trunc <vscale x 4 x i16> %2 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %3, ptr %y, align 8
+  ret void
+}
+
+define void @trunc_sat_u8u16_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u8u16_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl1re16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v8, v8, 0
+; CHECK-NEXT:    vse8.v v8, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i16>, ptr %x, align 16
+  %2 = tail call <vscale x 4 x i16> @llvm.umin.v4i16(<vscale x 4 x i16> %1, <vscale x 4 x i16> splat (i16 255))
+  %3 = trunc <vscale x 4 x i16> %2 to <vscale x 4 x i8>
+  store <vscale x 4 x i8> %3, ptr %y, align 8
+  ret void
+}
+
+
+define void @trunc_sat_i16i32_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i16i32_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    lui a0, 1048568
+; CHECK-NEXT:    addi a0, a0, 1
+; CHECK-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vmax.vx v8, v8, a0
+; CHECK-NEXT:    lui a0, 8
+; CHECK-NEXT:    vmin.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnsrl.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.smax.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 -32767))
+  %3 = tail call <vscale x 4 x i32> @llvm.smin.v4i32(<vscale x 4 x i32> %2, <vscale x 4 x i32> splat (i32 32768))
+  %4 = trunc <vscale x 4 x i32> %3 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %4, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_i16i32_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i16i32_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnclip.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.smax.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 -32768))
+  %3 = tail call <vscale x 4 x i32> @llvm.smin.v4i32(<vscale x 4 x i32> %2, <vscale x 4 x i32> splat (i32 32767))
+  %4 = trunc <vscale x 4 x i32> %3 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %4, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_i16i32_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i16i32_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnclip.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.smin.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 32767))
+  %3 = tail call <vscale x 4 x i32> @llvm.smax.v4i32(<vscale x 4 x i32> %2, <vscale x 4 x i32> splat (i32 -32768))
+  %4 = trunc <vscale x 4 x i32> %3 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %4, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_u16u32_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u16u32_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    lui a0, 8
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vminu.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnsrl.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.umin.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 32767))
+  %3 = trunc <vscale x 4 x i32> %2 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %3, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_u16u32_min(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u16u32_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnclipu.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.umin.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 65535))
+  %3 = trunc <vscale x 4 x i32> %2 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %3, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_u16u32_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u16u32_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnclipu.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.umin.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 65535))
+  %3 = trunc <vscale x 4 x i32> %2 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %3, ptr %y, align 16
+  ret void
+}
+
+define void @trunc_sat_u16u32_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u16u32_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl2re32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vnclipu.wi v10, v8, 0
+; CHECK-NEXT:    vs1r.v v10, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i32>, ptr %x, align 32
+  %2 = tail call <vscale x 4 x i32> @llvm.umin.v4i32(<vscale x 4 x i32> %1, <vscale x 4 x i32> splat (i32 65535))
+  %3 = trunc <vscale x 4 x i32> %2 to <vscale x 4 x i16>
+  store <vscale x 4 x i16> %3, ptr %y, align 16
+  ret void
+}
+
+
+define void @trunc_sat_i32i64_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i32i64_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    lui a0, 524288
+; CHECK-NEXT:    addiw a0, a0, 1
+; CHECK-NEXT:    vsetvli a2, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vmax.vx v8, v8, a0
+; CHECK-NEXT:    li a0, 1
+; CHECK-NEXT:    slli a0, a0, 31
+; CHECK-NEXT:    vmin.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnsrl.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.smax.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 -2147483647))
+  %3 = tail call <vscale x 4 x i64> @llvm.smin.v4i64(<vscale x 4 x i64> %2, <vscale x 4 x i64> splat (i64 2147483648))
+  %4 = trunc <vscale x 4 x i64> %3 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %4, ptr %y, align 32
+  ret void
+}
+
+define void @trunc_sat_i32i64_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i32i64_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnclip.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.smax.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 -2147483648))
+  %3 = tail call <vscale x 4 x i64> @llvm.smin.v4i64(<vscale x 4 x i64> %2, <vscale x 4 x i64> splat (i64 2147483647))
+  %4 = trunc <vscale x 4 x i64> %3 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %4, ptr %y, align 32
+  ret void
+}
+
+define void @trunc_sat_i32i64_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_i32i64_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnclip.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.smin.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 2147483647))
+  %3 = tail call <vscale x 4 x i64> @llvm.smax.v4i64(<vscale x 4 x i64> %2, <vscale x 4 x i64> splat (i64 -2147483648))
+  %4 = trunc <vscale x 4 x i64> %3 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %4, ptr %y, align 32
+  ret void
+}
+
+
+define void @trunc_sat_u32u64_notopt(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u32u64_notopt:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    lui a0, 524288
+; CHECK-NEXT:    addiw a0, a0, -1
+; CHECK-NEXT:    vsetvli a2, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vminu.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnsrl.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.umin.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 2147483647))
+  %3 = trunc <vscale x 4 x i64> %2 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %3, ptr %y, align 32
+  ret void
+}
+
+define void @trunc_sat_u32u64_min(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u32u64_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.umin.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 4294967295))
+  %3 = trunc <vscale x 4 x i64> %2 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %3, ptr %y, align 32
+  ret void
+}
+
+
+define void @trunc_sat_u32u64_maxmin(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u32u64_maxmin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.umin.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 4294967295))
+  %3 = trunc <vscale x 4 x i64> %2 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %3, ptr %y, align 32
+  ret void
+}
+
+define void @trunc_sat_u32u64_minmax(ptr %x, ptr %y) {
+; CHECK-LABEL: trunc_sat_u32u64_minmax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl4re64.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vnclipu.wi v12, v8, 0
+; CHECK-NEXT:    vs2r.v v12, (a1)
+; CHECK-NEXT:    ret
+  %1 = load <vscale x 4 x i64>, ptr %x, align 64
+  %2 = tail call <vscale x 4 x i64> @llvm.umin.v4i64(<vscale x 4 x i64> %1, <vscale x 4 x i64> splat (i64 4294967295))
+  %3 = trunc <vscale x 4 x i64> %2 to <vscale x 4 x i32>
+  store <vscale x 4 x i32> %3, ptr %y, align 32
+  ret void
+}

From 3d72c44fedc13ef51d2584b4fe930edb8d2e87ae Mon Sep 17 00:00:00 2001
From: Jakub Kuderski <jakub@nod-labs.com>
Date: Thu, 18 Apr 2024 00:49:38 -0400
Subject: [PATCH 816/886] [mlir][gpu] Improve `gpu.shuffle` documentation. NFC.
 (#89168)

* Make the wording around lanes / threads / work items more consistent.
* Add examples for all shufle modes.
* Also clean up `gpu.subgroup_reduce`.
---
 mlir/include/mlir/Dialect/GPU/IR/GPUOps.td | 68 ++++++++++++++++------
 1 file changed, 51 insertions(+), 17 deletions(-)

diff --git a/mlir/include/mlir/Dialect/GPU/IR/GPUOps.td b/mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
index bb373afa40ad9..1da68ed2176d8 100644
--- a/mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
+++ b/mlir/include/mlir/Dialect/GPU/IR/GPUOps.td
@@ -161,7 +161,7 @@ def GPU_SubgroupIdOp : GPU_Op<"subgroup_id", [
       Pure, DeclareOpInterfaceMethods<InferIntRangeInterface>]>,
     Arguments<(ins)>, Results<(outs Index:$result)> {
   let description = [{
-    Returns the subgroup id, i.e. the index of the current subgroup within the
+    Returns the subgroup id, i.e., the index of the current subgroup within the
     workgroup.
 
     Example:
@@ -1089,8 +1089,8 @@ def AnyIntegerOrFloatOr1DVector :
 def GPU_SubgroupReduceOp : GPU_Op<"subgroup_reduce", [SameOperandsAndResultType]> {
   let summary = "Reduce values among subgroup.";
   let description = [{
-    The `subgroup_reduce` op reduces the value of every work item across a
-    subgroup. The result is equal for all work items of a subgroup.
+    The `subgroup_reduce` op reduces the value of every lane (work item) across
+    a subgroup. The result is equal for all lanes.
 
     When the reduced value is of a vector type, each vector element is reduced
     independently. Only 1-d vector types are allowed.
@@ -1102,8 +1102,8 @@ def GPU_SubgroupReduceOp : GPU_Op<"subgroup_reduce", [SameOperandsAndResultType]
     %2 = gpu.subgroup_reduce add %b : (vector<4xf16>) -> (vector<4xf16>)
     ```
 
-    If `uniform` flag is set either none or all work items of a subgroup
-    need to execute this op in convergence. The reduction operation must be one
+    If `uniform` flag is set either none or all lanes of a subgroup need to execute
+    this op in convergence. The reduction operation must be one
     of:
     *  Integer types: `add`, `mul`, `minui`, `minsi`, `maxui`, `maxsi`, `and`,
        `or`, `xor`
@@ -1155,30 +1155,64 @@ def GPU_ShuffleOp : GPU_Op<
     Results<(outs I32I64F32OrF64:$shuffleResult, I1:$valid)> {
   let summary = "Shuffles values within a subgroup.";
   let description = [{
-    The "shuffle" op moves values to a different invocation within the same
-    subgroup.
+    The "shuffle" op moves values to a across lanes (a.k.a., invocations,
+    work items) within the same subgroup. The `width` argument specifies the
+    number of lanes that participate in the shuffle, and must be uniform
+    across all lanes. Further, the first `width` lanes of the subgroup must
+    be active.
 
-    Example:
+    The intepretation of the `offset` arguments depends on the selected
+    `mode`.
+
+    Returns the `shuffleResult` and `true` if the current lane id is smaller
+    than `width`, and an unspecified value and `false` otherwise.
+
+    `xor` example:
 
     ```mlir
-    %1, %2 = gpu.shuffle %0, %offset, %width xor : f32
+    %1, %2 = gpu.shuffle xor %0, %offset, %width : f32
     ```
 
-    For lane k returns the value from lane `k ^ offset` and `true` if that lane
-    is smaller than %width. Otherwise it returns an unspecified value and
-    `false`. A lane is the index of an invocation relative to its subgroup.
+    For lane `k`, returns the value `%0` from lane `k ^ offset`. Every lane
+    trades value with exactly one other lane.
 
-    The width specifies the number of invocations that participate in the
-    shuffle. The width needs to be the same for all invocations that participate
-    in the shuffle. Exactly the first `width` invocations of a subgroup need to
-    execute this op in convergence.
+    `down` example:
+
+    ```mlir
+    %cst1 = arith.constant 1 : i32
+    %3, %4 = gpu.shuffle down %0, %cst1, %width : f32
+    ```
+
+    For lane `k`, returns the value from lane `(k + 1) % width`.
+
+    `up` example:
+
+    ```mlir
+    %cst1 = arith.constant 1 : i32
+    %5, %6 = gpu.shuffle up %0, %cst1, %width : f32
+    ```
+
+    For lane `k`, returns the value from lane `(k - 1) % width`.
+
+    `idx` example:
+
+    ```mlir
+    %cst0 = arith.constant 0 : i32
+    %7, %8 = gpu.shuffle idx %0, %cst0, %width : f32
+    ```
+
+    Broadcasts the value from lane 0 to all lanes.
   }];
+
+  let assemblyFormat = [{
+    $mode $value `,` $offset `,` $width attr-dict `:` type($value)
+  }];
+
   let builders = [
     // Helper function that creates a shuffle with constant offset/width.
     OpBuilder<(ins "Value":$value, "int32_t":$offset, "int32_t":$width,
                    "ShuffleMode":$mode)>
   ];
-  let assemblyFormat = "$mode $value `,` $offset `,` $width attr-dict `:` type($value)";
 }
 
 def GPU_BarrierOp : GPU_Op<"barrier"> {

From 472b612ccbeda470fa15005e11882349b0b384be Mon Sep 17 00:00:00 2001
From: Nikolas Klauser <nikolasklauser@berlin.de>
Date: Thu, 18 Apr 2024 06:55:50 +0200
Subject: [PATCH 817/886] [libc++][NFC] Remove unused includes from
 <__type_traits/remove_cv.h> (#88752)

---
 libcxx/include/__format/format_arg.h     | 1 +
 libcxx/include/__type_traits/remove_cv.h | 2 --
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/libcxx/include/__format/format_arg.h b/libcxx/include/__format/format_arg.h
index 4924e5fb32533..aa02f81dc40e2 100644
--- a/libcxx/include/__format/format_arg.h
+++ b/libcxx/include/__format/format_arg.h
@@ -19,6 +19,7 @@
 #include <__fwd/format.h>
 #include <__memory/addressof.h>
 #include <__type_traits/conditional.h>
+#include <__type_traits/remove_const.h>
 #include <__utility/forward.h>
 #include <__utility/move.h>
 #include <__utility/unreachable.h>
diff --git a/libcxx/include/__type_traits/remove_cv.h b/libcxx/include/__type_traits/remove_cv.h
index 8e1c043364323..2c4e9e419a1be 100644
--- a/libcxx/include/__type_traits/remove_cv.h
+++ b/libcxx/include/__type_traits/remove_cv.h
@@ -10,8 +10,6 @@
 #define _LIBCPP___TYPE_TRAITS_REMOVE_CV_H
 
 #include <__config>
-#include <__type_traits/remove_const.h>
-#include <__type_traits/remove_volatile.h>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header

From fbca90b609f6c42ce7e4c9bb3ee8bd2c230a670c Mon Sep 17 00:00:00 2001
From: Brad Smith <brad@comstyle.com>
Date: Thu, 18 Apr 2024 01:27:49 -0400
Subject: [PATCH 818/886] [unittest] Skip W+X MappedMemoryTests on OpenBSD
 (#89102)

OpenBSD uses W^X so the tests will not work.
---
 llvm/unittests/Support/MemoryTest.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/unittests/Support/MemoryTest.cpp b/llvm/unittests/Support/MemoryTest.cpp
index 0164492829ccd..9daa6d0ff9e4d 100644
--- a/llvm/unittests/Support/MemoryTest.cpp
+++ b/llvm/unittests/Support/MemoryTest.cpp
@@ -40,7 +40,7 @@ bool IsMPROTECT() {
     err(EXIT_FAILURE, "sysctl");
 
   return !!(paxflags & CTL_PROC_PAXFLAGS_MPROTECT);
-#elif defined(__APPLE__) && defined(__aarch64__)
+#elif (defined(__APPLE__) && defined(__aarch64__)) || defined(__OpenBSD__)
   return true;
 #else
   return false;

From 3d56ea05b6c746a7144f643bef2ebd599f605b8b Mon Sep 17 00:00:00 2001
From: Timm Baeder <tbaeder@redhat.com>
Date: Thu, 18 Apr 2024 07:39:29 +0200
Subject: [PATCH 819/886] [clang][NFC] Fix FieldDecl::isUnnamedBitfield()
 capitalization (#89048)

We always capitalize bitfield as "BitField".
---
 .../ProTypeMemberInitCheck.cpp                |  2 +-
 .../modernize/UseEqualsDefaultCheck.cpp       |  2 +-
 .../utils/ExceptionSpecAnalyzer.cpp           |  2 +-
 clang/include/clang/AST/Decl.h                |  2 +-
 clang/lib/AST/APValue.cpp                     |  3 ++-
 clang/lib/AST/ASTContext.cpp                  |  2 +-
 clang/lib/AST/Decl.cpp                        |  2 +-
 clang/lib/AST/DeclCXX.cpp                     |  7 +++---
 clang/lib/AST/Expr.cpp                        |  4 ++--
 clang/lib/AST/ExprConstant.cpp                | 18 +++++++-------
 clang/lib/AST/Interp/EvaluationResult.cpp     |  2 +-
 clang/lib/AST/ItaniumMangle.cpp               |  8 +++----
 clang/lib/AST/MicrosoftMangle.cpp             |  2 +-
 clang/lib/Analysis/FlowSensitive/ASTOps.cpp   |  2 +-
 clang/lib/Analysis/UninitializedValues.cpp    |  2 +-
 clang/lib/CodeGen/ABIInfoImpl.cpp             |  2 +-
 clang/lib/CodeGen/CGCall.cpp                  |  2 +-
 clang/lib/CodeGen/CGExpr.cpp                  |  2 +-
 clang/lib/CodeGen/CGExprAgg.cpp               |  8 ++++---
 clang/lib/CodeGen/CGExprCXX.cpp               |  2 +-
 clang/lib/CodeGen/CGExprConstant.cpp          |  4 ++--
 clang/lib/CodeGen/CodeGenTBAA.cpp             |  2 +-
 clang/lib/CodeGen/Targets/X86.cpp             |  4 ++--
 clang/lib/Sema/SemaChecking.cpp               |  2 +-
 clang/lib/Sema/SemaDecl.cpp                   |  2 +-
 clang/lib/Sema/SemaDeclCXX.cpp                | 24 +++++++++----------
 clang/lib/Sema/SemaInit.cpp                   | 24 +++++++++----------
 .../Checkers/ObjCUnusedIVarsChecker.cpp       |  3 +--
 clang/lib/StaticAnalyzer/Core/RegionStore.cpp |  4 ++--
 29 files changed, 74 insertions(+), 71 deletions(-)

diff --git a/clang-tools-extra/clang-tidy/cppcoreguidelines/ProTypeMemberInitCheck.cpp b/clang-tools-extra/clang-tidy/cppcoreguidelines/ProTypeMemberInitCheck.cpp
index 855c4a2efc373..9c3c7cc70c187 100644
--- a/clang-tools-extra/clang-tidy/cppcoreguidelines/ProTypeMemberInitCheck.cpp
+++ b/clang-tools-extra/clang-tidy/cppcoreguidelines/ProTypeMemberInitCheck.cpp
@@ -444,7 +444,7 @@ void ProTypeMemberInitCheck::checkMissingMemberInitializer(
     if (!F->hasInClassInitializer() &&
         utils::type_traits::isTriviallyDefaultConstructible(F->getType(),
                                                             Context) &&
-        !isEmpty(Context, F->getType()) && !F->isUnnamedBitfield() &&
+        !isEmpty(Context, F->getType()) && !F->isUnnamedBitField() &&
         !AnyMemberHasInitPerUnion)
       FieldsToInit.insert(F);
   });
diff --git a/clang-tools-extra/clang-tidy/modernize/UseEqualsDefaultCheck.cpp b/clang-tools-extra/clang-tidy/modernize/UseEqualsDefaultCheck.cpp
index 5134eb51a0322..93151024064b4 100644
--- a/clang-tools-extra/clang-tidy/modernize/UseEqualsDefaultCheck.cpp
+++ b/clang-tools-extra/clang-tidy/modernize/UseEqualsDefaultCheck.cpp
@@ -26,7 +26,7 @@ getAllNamedFields(const CXXRecordDecl *Record) {
   std::set<const FieldDecl *> Result;
   for (const auto *Field : Record->fields()) {
     // Static data members are not in this range.
-    if (Field->isUnnamedBitfield())
+    if (Field->isUnnamedBitField())
       continue;
     Result.insert(Field);
   }
diff --git a/clang-tools-extra/clang-tidy/utils/ExceptionSpecAnalyzer.cpp b/clang-tools-extra/clang-tidy/utils/ExceptionSpecAnalyzer.cpp
index 1dde049051785..4a9426ee7e8bb 100644
--- a/clang-tools-extra/clang-tidy/utils/ExceptionSpecAnalyzer.cpp
+++ b/clang-tools-extra/clang-tidy/utils/ExceptionSpecAnalyzer.cpp
@@ -99,7 +99,7 @@ ExceptionSpecAnalyzer::analyzeRecord(const CXXRecordDecl *RecordDecl,
   }
 
   for (const auto *FDecl : RecordDecl->fields())
-    if (!FDecl->isInvalidDecl() && !FDecl->isUnnamedBitfield()) {
+    if (!FDecl->isInvalidDecl() && !FDecl->isUnnamedBitField()) {
       State Result = analyzeFieldDecl(FDecl, Kind);
       if (Result == State::Throwing || Result == State::Unknown)
         return Result;
diff --git a/clang/include/clang/AST/Decl.h b/clang/include/clang/AST/Decl.h
index 91449ebcceec3..0a9c9e17d3f9f 100644
--- a/clang/include/clang/AST/Decl.h
+++ b/clang/include/clang/AST/Decl.h
@@ -3149,7 +3149,7 @@ class FieldDecl : public DeclaratorDecl, public Mergeable<FieldDecl> {
   bool isBitField() const { return BitField; }
 
   /// Determines whether this is an unnamed bitfield.
-  bool isUnnamedBitfield() const { return isBitField() && !getDeclName(); }
+  bool isUnnamedBitField() const { return isBitField() && !getDeclName(); }
 
   /// Determines whether this field is a
   /// representative for an anonymous struct or union. Such fields are
diff --git a/clang/lib/AST/APValue.cpp b/clang/lib/AST/APValue.cpp
index d8042321319a6..8c77b563657d9 100644
--- a/clang/lib/AST/APValue.cpp
+++ b/clang/lib/AST/APValue.cpp
@@ -908,7 +908,8 @@ void APValue::printPretty(raw_ostream &Out, const PrintingPolicy &Policy,
     for (const auto *FI : RD->fields()) {
       if (!First)
         Out << ", ";
-      if (FI->isUnnamedBitfield()) continue;
+      if (FI->isUnnamedBitField())
+        continue;
       getStructField(FI->getFieldIndex()).
         printPretty(Out, Policy, FI->getType(), Ctx);
       First = false;
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index bcae6d705b780..b974fc28283c7 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -2684,7 +2684,7 @@ getSubobjectSizeInBits(const FieldDecl *Field, const ASTContext &Context,
   if (Field->isBitField()) {
     // If we have explicit padding bits, they don't contribute bits
     // to the actual object representation, so return 0.
-    if (Field->isUnnamedBitfield())
+    if (Field->isUnnamedBitField())
       return 0;
 
     int64_t BitfieldSize = Field->getBitWidthValue(Context);
diff --git a/clang/lib/AST/Decl.cpp b/clang/lib/AST/Decl.cpp
index 63388d79bbf21..1b99c435aebb2 100644
--- a/clang/lib/AST/Decl.cpp
+++ b/clang/lib/AST/Decl.cpp
@@ -4597,7 +4597,7 @@ unsigned FieldDecl::getBitWidthValue(const ASTContext &Ctx) const {
 }
 
 bool FieldDecl::isZeroLengthBitField(const ASTContext &Ctx) const {
-  return isUnnamedBitfield() && !getBitWidth()->isValueDependent() &&
+  return isUnnamedBitField() && !getBitWidth()->isValueDependent() &&
          getBitWidthValue(Ctx) == 0;
 }
 
diff --git a/clang/lib/AST/DeclCXX.cpp b/clang/lib/AST/DeclCXX.cpp
index 645ec2f7563bc..00cc857f5e737 100644
--- a/clang/lib/AST/DeclCXX.cpp
+++ b/clang/lib/AST/DeclCXX.cpp
@@ -668,7 +668,7 @@ bool CXXRecordDecl::hasSubobjectAtOffsetZeroOfEmptyBaseType(
     for (auto *FD : X->fields()) {
       // FIXME: Should we really care about the type of the first non-static
       // data member of a non-union if there are preceding unnamed bit-fields?
-      if (FD->isUnnamedBitfield())
+      if (FD->isUnnamedBitField())
         continue;
 
       if (!IsFirstField && !FD->isZeroSize(Ctx))
@@ -947,7 +947,7 @@ void CXXRecordDecl::addedMember(Decl *D) {
     //   A declaration for a bit-field that omits the identifier declares an
     //   unnamed bit-field. Unnamed bit-fields are not members and cannot be
     //   initialized.
-    if (Field->isUnnamedBitfield()) {
+    if (Field->isUnnamedBitField()) {
       // C++ [meta.unary.prop]p4: [LWG2358]
       //   T is a class type [...] with [...] no unnamed bit-fields of non-zero
       //   length
@@ -3469,7 +3469,8 @@ static bool isValidStructGUID(ASTContext &Ctx, QualType T) {
           return false;
       auto MatcherIt = Fields.begin();
       for (const FieldDecl *FD : RD->fields()) {
-        if (FD->isUnnamedBitfield()) continue;
+        if (FD->isUnnamedBitField())
+          continue;
         if (FD->isBitField() || MatcherIt == Fields.end() ||
             !(*MatcherIt)(FD->getType()))
           return false;
diff --git a/clang/lib/AST/Expr.cpp b/clang/lib/AST/Expr.cpp
index 07c9f287dd076..9eec7edc9d1a3 100644
--- a/clang/lib/AST/Expr.cpp
+++ b/clang/lib/AST/Expr.cpp
@@ -2044,7 +2044,7 @@ const FieldDecl *CastExpr::getTargetFieldForToUnionCast(const RecordDecl *RD,
   for (Field = RD->field_begin(), FieldEnd = RD->field_end();
        Field != FieldEnd; ++Field) {
     if (Ctx.hasSameUnqualifiedType(Field->getType(), OpType) &&
-        !Field->isUnnamedBitfield()) {
+        !Field->isUnnamedBitField()) {
       return *Field;
     }
   }
@@ -3393,7 +3393,7 @@ bool Expr::isConstantInitializer(ASTContext &Ctx, bool IsForRef,
           continue;
 
         // Don't emit anonymous bitfields, they just affect layout.
-        if (Field->isUnnamedBitfield())
+        if (Field->isUnnamedBitField())
           continue;
 
         if (ElementNo < ILE->getNumInits()) {
diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
index 88c8eaf6ef9b6..73ae8d8efb23a 100644
--- a/clang/lib/AST/ExprConstant.cpp
+++ b/clang/lib/AST/ExprConstant.cpp
@@ -2492,7 +2492,7 @@ static bool CheckEvaluationResult(CheckEvaluationResultKind CERK,
       }
     }
     for (const auto *I : RD->fields()) {
-      if (I->isUnnamedBitfield())
+      if (I->isUnnamedBitField())
         continue;
 
       if (!CheckEvaluationResult(CERK, Info, DiagLoc, I->getType(),
@@ -3529,7 +3529,7 @@ static bool isReadByLvalueToRvalueConversion(const CXXRecordDecl *RD) {
     return false;
 
   for (auto *Field : RD->fields())
-    if (!Field->isUnnamedBitfield() &&
+    if (!Field->isUnnamedBitField() &&
         isReadByLvalueToRvalueConversion(Field->getType()))
       return true;
 
@@ -4898,7 +4898,7 @@ static bool handleDefaultInitValue(QualType T, APValue &Result) {
           handleDefaultInitValue(I->getType(), Result.getStructBase(Index));
 
     for (const auto *I : RD->fields()) {
-      if (I->isUnnamedBitfield())
+      if (I->isUnnamedBitField())
         continue;
       Success &= handleDefaultInitValue(
           I->getType(), Result.getStructField(I->getFieldIndex()));
@@ -6436,7 +6436,7 @@ static bool HandleConstructorCall(const Expr *E, const LValue &This,
     // Default-initialize any fields with no explicit initializer.
     for (; !declaresSameEntity(*FieldIt, FD); ++FieldIt) {
       assert(FieldIt != RD->field_end() && "missing field?");
-      if (!FieldIt->isUnnamedBitfield())
+      if (!FieldIt->isUnnamedBitField())
         Success &= handleDefaultInitValue(
             FieldIt->getType(),
             Result.getStructField(FieldIt->getFieldIndex()));
@@ -6546,7 +6546,7 @@ static bool HandleConstructorCall(const Expr *E, const LValue &This,
   // Default-initialize any remaining fields.
   if (!RD->isUnion()) {
     for (; FieldIt != RD->field_end(); ++FieldIt) {
-      if (!FieldIt->isUnnamedBitfield())
+      if (!FieldIt->isUnnamedBitField())
         Success &= handleDefaultInitValue(
             FieldIt->getType(),
             Result.getStructField(FieldIt->getFieldIndex()));
@@ -6708,7 +6708,7 @@ static bool HandleDestructionImpl(EvalInfo &Info, SourceRange CallRange,
   // fields first and then walk them backwards.
   SmallVector<FieldDecl*, 16> Fields(RD->fields());
   for (const FieldDecl *FD : llvm::reverse(Fields)) {
-    if (FD->isUnnamedBitfield())
+    if (FD->isUnnamedBitField())
       continue;
 
     LValue Subobject = This;
@@ -10220,7 +10220,7 @@ static bool HandleClassZeroInitialization(EvalInfo &Info, const Expr *E,
 
   for (const auto *I : RD->fields()) {
     // -- if T is a reference type, no initialization is performed.
-    if (I->isUnnamedBitfield() || I->getType()->isReferenceType())
+    if (I->isUnnamedBitField() || I->getType()->isReferenceType())
       continue;
 
     LValue Subobject = This;
@@ -10243,7 +10243,7 @@ bool RecordExprEvaluator::ZeroInitialization(const Expr *E, QualType T) {
     // C++11 [dcl.init]p5: If T is a (possibly cv-qualified) union type, the
     // object's first non-static named data member is zero-initialized
     RecordDecl::field_iterator I = RD->field_begin();
-    while (I != RD->field_end() && (*I)->isUnnamedBitfield())
+    while (I != RD->field_end() && (*I)->isUnnamedBitField())
       ++I;
     if (I == RD->field_end()) {
       Result = APValue((const FieldDecl*)nullptr);
@@ -10390,7 +10390,7 @@ bool RecordExprEvaluator::VisitCXXParenListOrInitListExpr(
   for (const auto *Field : RD->fields()) {
     // Anonymous bit-fields are not considered members of the class for
     // purposes of aggregate initialization.
-    if (Field->isUnnamedBitfield())
+    if (Field->isUnnamedBitField())
       continue;
 
     LValue Subobject = This;
diff --git a/clang/lib/AST/Interp/EvaluationResult.cpp b/clang/lib/AST/Interp/EvaluationResult.cpp
index d567b551f7f6f..e92d686c724cc 100644
--- a/clang/lib/AST/Interp/EvaluationResult.cpp
+++ b/clang/lib/AST/Interp/EvaluationResult.cpp
@@ -105,7 +105,7 @@ static bool CheckFieldsInitialized(InterpState &S, SourceLocation Loc,
       Result &= CheckFieldsInitialized(S, Loc, FieldPtr, FieldPtr.getRecord());
     } else if (FieldType->isIncompleteArrayType()) {
       // Nothing to do here.
-    } else if (F.Decl->isUnnamedBitfield()) {
+    } else if (F.Decl->isUnnamedBitField()) {
       // Nothing do do here.
     } else if (FieldType->isArrayType()) {
       const auto *CAT =
diff --git a/clang/lib/AST/ItaniumMangle.cpp b/clang/lib/AST/ItaniumMangle.cpp
index d632c697fa20d..c3b98d2d2149c 100644
--- a/clang/lib/AST/ItaniumMangle.cpp
+++ b/clang/lib/AST/ItaniumMangle.cpp
@@ -6176,7 +6176,7 @@ static bool isZeroInitialized(QualType T, const APValue &V) {
     }
     I = 0;
     for (const FieldDecl *FD : RD->fields()) {
-      if (!FD->isUnnamedBitfield() &&
+      if (!FD->isUnnamedBitField() &&
           !isZeroInitialized(FD->getType(), V.getStructField(I)))
         return false;
       ++I;
@@ -6189,7 +6189,7 @@ static bool isZeroInitialized(QualType T, const APValue &V) {
     assert(RD && "unexpected type for union value");
     // Zero-initialization zeroes the first non-unnamed-bitfield field, if any.
     for (const FieldDecl *FD : RD->fields()) {
-      if (!FD->isUnnamedBitfield())
+      if (!FD->isUnnamedBitField())
         return V.getUnionField() && declaresSameEntity(FD, V.getUnionField()) &&
                isZeroInitialized(FD->getType(), V.getUnionValue());
     }
@@ -6331,7 +6331,7 @@ void CXXNameMangler::mangleValueInTemplateArg(QualType T, const APValue &V,
     llvm::SmallVector<const FieldDecl *, 16> Fields(RD->fields());
     while (
         !Fields.empty() &&
-        (Fields.back()->isUnnamedBitfield() ||
+        (Fields.back()->isUnnamedBitField() ||
          isZeroInitialized(Fields.back()->getType(),
                            V.getStructField(Fields.back()->getFieldIndex())))) {
       Fields.pop_back();
@@ -6351,7 +6351,7 @@ void CXXNameMangler::mangleValueInTemplateArg(QualType T, const APValue &V,
     for (unsigned I = 0, N = Bases.size(); I != N; ++I)
       mangleValueInTemplateArg(Bases[I].getType(), V.getStructBase(I), false);
     for (unsigned I = 0, N = Fields.size(); I != N; ++I) {
-      if (Fields[I]->isUnnamedBitfield())
+      if (Fields[I]->isUnnamedBitField())
         continue;
       mangleValueInTemplateArg(Fields[I]->getType(),
                                V.getStructField(Fields[I]->getFieldIndex()),
diff --git a/clang/lib/AST/MicrosoftMangle.cpp b/clang/lib/AST/MicrosoftMangle.cpp
index a0bb04e69c9be..36d611750ca48 100644
--- a/clang/lib/AST/MicrosoftMangle.cpp
+++ b/clang/lib/AST/MicrosoftMangle.cpp
@@ -1933,7 +1933,7 @@ void MicrosoftCXXNameMangler::mangleTemplateArgValue(QualType T,
     for (const CXXBaseSpecifier &B : RD->bases())
       mangleTemplateArgValue(B.getType(), V.getStructBase(BaseIndex++), TAK);
     for (const FieldDecl *FD : RD->fields())
-      if (!FD->isUnnamedBitfield())
+      if (!FD->isUnnamedBitField())
         mangleTemplateArgValue(FD->getType(),
                                V.getStructField(FD->getFieldIndex()), TAK,
                                /*WithScalarType*/ true);
diff --git a/clang/lib/Analysis/FlowSensitive/ASTOps.cpp b/clang/lib/Analysis/FlowSensitive/ASTOps.cpp
index 75188aef4d1a4..1982c6c9f3830 100644
--- a/clang/lib/Analysis/FlowSensitive/ASTOps.cpp
+++ b/clang/lib/Analysis/FlowSensitive/ASTOps.cpp
@@ -101,7 +101,7 @@ getFieldsForInitListExpr(const InitListExpr *InitList) {
   // fields to avoid mapping inits to the wrongs fields.
   llvm::copy_if(
       RD->fields(), std::back_inserter(Fields),
-      [](const FieldDecl *Field) { return !Field->isUnnamedBitfield(); });
+      [](const FieldDecl *Field) { return !Field->isUnnamedBitField(); });
   return Fields;
 }
 
diff --git a/clang/lib/Analysis/UninitializedValues.cpp b/clang/lib/Analysis/UninitializedValues.cpp
index e9111ded64eb1..bf2f730618650 100644
--- a/clang/lib/Analysis/UninitializedValues.cpp
+++ b/clang/lib/Analysis/UninitializedValues.cpp
@@ -44,7 +44,7 @@ static bool recordIsNotEmpty(const RecordDecl *RD) {
   // We consider a record decl to be empty if it contains only unnamed bit-
   // fields, zero-width fields, and fields of empty record type.
   for (const auto *FD : RD->fields()) {
-    if (FD->isUnnamedBitfield())
+    if (FD->isUnnamedBitField())
       continue;
     if (FD->isZeroSize(FD->getASTContext()))
       continue;
diff --git a/clang/lib/CodeGen/ABIInfoImpl.cpp b/clang/lib/CodeGen/ABIInfoImpl.cpp
index 3e34d82cb399b..1e4d48e13c025 100644
--- a/clang/lib/CodeGen/ABIInfoImpl.cpp
+++ b/clang/lib/CodeGen/ABIInfoImpl.cpp
@@ -247,7 +247,7 @@ Address CodeGen::emitMergePHI(CodeGenFunction &CGF, Address Addr1,
 
 bool CodeGen::isEmptyField(ASTContext &Context, const FieldDecl *FD,
                            bool AllowArrays, bool AsIfNoUniqueAddr) {
-  if (FD->isUnnamedBitfield())
+  if (FD->isUnnamedBitField())
     return true;
 
   QualType FT = FD->getType();
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 3f5463a9a70e9..6c33cc17621f8 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -3665,7 +3665,7 @@ static void setUsedBits(CodeGenModule &CGM, const RecordType *RTy, int Offset,
   for (auto I = RD->field_begin(), E = RD->field_end(); I != E; ++I, ++Idx) {
     const FieldDecl *F = *I;
 
-    if (F->isUnnamedBitfield() || F->isZeroLengthBitField(Context) ||
+    if (F->isUnnamedBitField() || F->isZeroLengthBitField(Context) ||
         F->getType()->isIncompleteArrayType())
       continue;
 
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index cf696a1c9f560..931cb391342ea 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -4704,7 +4704,7 @@ unsigned CodeGenFunction::getDebugInfoFIndex(const RecordDecl *Rec,
   for (auto *F : Rec->getDefinition()->fields()) {
     if (I == FieldIndex)
       break;
-    if (F->isUnnamedBitfield())
+    if (F->isUnnamedBitField())
       Skipped++;
     I++;
   }
diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
index 1b9287ea23934..355fec42be448 100644
--- a/clang/lib/CodeGen/CGExprAgg.cpp
+++ b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -1755,7 +1755,9 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
       // Make sure that it's really an empty and not a failure of
       // semantic analysis.
       for (const auto *Field : record->fields())
-        assert((Field->isUnnamedBitfield() || Field->isAnonymousStructOrUnion()) && "Only unnamed bitfields or ananymous class allowed");
+        assert(
+            (Field->isUnnamedBitField() || Field->isAnonymousStructOrUnion()) &&
+            "Only unnamed bitfields or ananymous class allowed");
 #endif
       return;
     }
@@ -1783,7 +1785,7 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
       break;
 
     // Always skip anonymous bitfields.
-    if (field->isUnnamedBitfield())
+    if (field->isUnnamedBitField())
       continue;
 
     // We're done if we reach the end of the explicit initializers, we
@@ -1988,7 +1990,7 @@ static CharUnits GetNumNonZeroBytesInInit(const Expr *E, CodeGenFunction &CGF) {
         if (Field->getType()->isIncompleteArrayType() ||
             ILEElement == ILE->getNumInits())
           break;
-        if (Field->isUnnamedBitfield())
+        if (Field->isUnnamedBitField())
           continue;
 
         const Expr *E = ILE->getInit(ILEElement++);
diff --git a/clang/lib/CodeGen/CGExprCXX.cpp b/clang/lib/CodeGen/CGExprCXX.cpp
index a4fb673284cec..673ccef84d678 100644
--- a/clang/lib/CodeGen/CGExprCXX.cpp
+++ b/clang/lib/CodeGen/CGExprCXX.cpp
@@ -1235,7 +1235,7 @@ void CodeGenFunction::EmitNewArrayInitializer(
         if (auto *CXXRD = dyn_cast<CXXRecordDecl>(RType->getDecl()))
           NumElements = CXXRD->getNumBases();
         for (auto *Field : RType->getDecl()->fields())
-          if (!Field->isUnnamedBitfield())
+          if (!Field->isUnnamedBitField())
             ++NumElements;
         // FIXME: Recurse into nested InitListExprs.
         if (ILE->getNumInits() == NumElements)
diff --git a/clang/lib/CodeGen/CGExprConstant.cpp b/clang/lib/CodeGen/CGExprConstant.cpp
index 9f1b06eebf9ed..c924660c5a91c 100644
--- a/clang/lib/CodeGen/CGExprConstant.cpp
+++ b/clang/lib/CodeGen/CGExprConstant.cpp
@@ -706,7 +706,7 @@ bool ConstStructBuilder::Build(InitListExpr *ILE, bool AllowOverwrite) {
       continue;
 
     // Don't emit anonymous bitfields.
-    if (Field->isUnnamedBitfield())
+    if (Field->isUnnamedBitField())
       continue;
 
     // Get the initializer.  A struct can include fields without initializers,
@@ -840,7 +840,7 @@ bool ConstStructBuilder::Build(const APValue &Val, const RecordDecl *RD,
       continue;
 
     // Don't emit anonymous bitfields or zero-sized fields.
-    if (Field->isUnnamedBitfield() || Field->isZeroSize(CGM.getContext()))
+    if (Field->isUnnamedBitField() || Field->isZeroSize(CGM.getContext()))
       continue;
 
     // Emit the value of the initializer.
diff --git a/clang/lib/CodeGen/CodeGenTBAA.cpp b/clang/lib/CodeGen/CodeGenTBAA.cpp
index da689ee6a13d7..284421f494711 100644
--- a/clang/lib/CodeGen/CodeGenTBAA.cpp
+++ b/clang/lib/CodeGen/CodeGenTBAA.cpp
@@ -414,7 +414,7 @@ llvm::MDNode *CodeGenTBAA::getBaseTypeInfoHelper(const Type *Ty) {
                  });
     }
     for (FieldDecl *Field : RD->fields()) {
-      if (Field->isZeroSize(Context) || Field->isUnnamedBitfield())
+      if (Field->isZeroSize(Context) || Field->isUnnamedBitField())
         continue;
       QualType FieldQTy = Field->getType();
       llvm::MDNode *TypeNode = isValidBaseType(FieldQTy)
diff --git a/clang/lib/CodeGen/Targets/X86.cpp b/clang/lib/CodeGen/Targets/X86.cpp
index f04db56db3357..94cf0d86f9bed 100644
--- a/clang/lib/CodeGen/Targets/X86.cpp
+++ b/clang/lib/CodeGen/Targets/X86.cpp
@@ -2087,7 +2087,7 @@ void X86_64ABIInfo::classify(QualType Ty, uint64_t OffsetBase, Class &Lo,
       bool BitField = i->isBitField();
 
       // Ignore padding bit-fields.
-      if (BitField && i->isUnnamedBitfield())
+      if (BitField && i->isUnnamedBitField())
         continue;
 
       // AMD64-ABI 3.2.3p2: Rule 1. If the size of an object is larger than
@@ -2128,7 +2128,7 @@ void X86_64ABIInfo::classify(QualType Ty, uint64_t OffsetBase, Class &Lo,
       // structure to be passed in memory even if unaligned, and
       // therefore they can straddle an eightbyte.
       if (BitField) {
-        assert(!i->isUnnamedBitfield());
+        assert(!i->isUnnamedBitField());
         uint64_t Offset = OffsetBase + Layout.getFieldOffset(idx);
         uint64_t Size = i->getBitWidthValue(getContext());
 
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index d814327d23470..6778bc6607f75 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -622,7 +622,7 @@ struct BuiltinDumpStructGenerator {
     for (auto *D : RD->decls()) {
       auto *IFD = dyn_cast<IndirectFieldDecl>(D);
       auto *FD = IFD ? IFD->getAnonField() : dyn_cast<FieldDecl>(D);
-      if (!FD || FD->isUnnamedBitfield() || FD->isAnonymousStructOrUnion())
+      if (!FD || FD->isUnnamedBitField() || FD->isAnonymousStructOrUnion())
         continue;
 
       llvm::SmallString<20> Format = llvm::StringRef("%s%s %s ");
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 356abe09a5ca6..af6b3f21f15a6 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -19683,7 +19683,7 @@ void Sema::ActOnFields(Scope *S, SourceLocation RecLoc, Decl *EnclosingDecl,
                                       E = Record->field_end();
            (NonBitFields == 0 || ZeroSize) && I != E; ++I) {
         IsEmpty = false;
-        if (I->isUnnamedBitfield()) {
+        if (I->isUnnamedBitField()) {
           if (!I->isZeroLengthBitField(Context))
             ZeroSize = false;
         } else {
diff --git a/clang/lib/Sema/SemaDeclCXX.cpp b/clang/lib/Sema/SemaDeclCXX.cpp
index 591016243b0ac..2ef8a15d5238f 100644
--- a/clang/lib/Sema/SemaDeclCXX.cpp
+++ b/clang/lib/Sema/SemaDeclCXX.cpp
@@ -1453,7 +1453,7 @@ static bool checkMemberDecomposition(Sema &S, ArrayRef<BindingDecl*> Bindings,
 
   auto DiagnoseBadNumberOfBindings = [&]() -> bool {
     unsigned NumFields = llvm::count_if(
-        RD->fields(), [](FieldDecl *FD) { return !FD->isUnnamedBitfield(); });
+        RD->fields(), [](FieldDecl *FD) { return !FD->isUnnamedBitField(); });
     assert(Bindings.size() != NumFields);
     S.Diag(Src->getLocation(), diag::err_decomp_decl_wrong_number_bindings)
         << DecompType << (unsigned)Bindings.size() << NumFields << NumFields
@@ -1466,7 +1466,7 @@ static bool checkMemberDecomposition(Sema &S, ArrayRef<BindingDecl*> Bindings,
   //   E shall not have an anonymous union member, ...
   unsigned I = 0;
   for (auto *FD : RD->fields()) {
-    if (FD->isUnnamedBitfield())
+    if (FD->isUnnamedBitField())
       continue;
 
     // All the non-static data members are required to be nameable, so they
@@ -2067,7 +2067,7 @@ static bool CheckConstexprCtorInitializer(Sema &SemaRef,
   if (Field->isInvalidDecl())
     return true;
 
-  if (Field->isUnnamedBitfield())
+  if (Field->isUnnamedBitField())
     return true;
 
   // Anonymous unions with no variant members and empty anonymous structs do not
@@ -5509,7 +5509,7 @@ bool Sema::SetCtorInitializers(CXXConstructorDecl *Constructor, bool AnyErrors,
       //   A declaration for a bit-field that omits the identifier declares an
       //   unnamed bit-field. Unnamed bit-fields are not members and cannot be
       //   initialized.
-      if (F->isUnnamedBitfield())
+      if (F->isUnnamedBitField())
         continue;
 
       // If we're not generating the implicit copy/move constructor, then we'll
@@ -5638,7 +5638,7 @@ static void DiagnoseBaseOrMemInitializerOrder(
 
   // 3. Direct fields.
   for (auto *Field : ClassDecl->fields()) {
-    if (Field->isUnnamedBitfield())
+    if (Field->isUnnamedBitField())
       continue;
 
     PopulateKeysForFields(Field, IdealInitKeys);
@@ -7030,7 +7030,7 @@ void Sema::CheckCompletedCXXClass(Scope *S, CXXRecordDecl *Record) {
       !Record->isLambda()) {
     bool Complained = false;
     for (const auto *F : Record->fields()) {
-      if (F->hasInClassInitializer() || F->isUnnamedBitfield())
+      if (F->hasInClassInitializer() || F->isUnnamedBitField())
         continue;
 
       if (F->getType()->isReferenceType() ||
@@ -8037,7 +8037,7 @@ class DefaultedComparisonVisitor {
     for (FieldDecl *Field : Record->fields()) {
       // C++23 [class.bit]p2:
       //   Unnamed bit-fields are not members ...
-      if (Field->isUnnamedBitfield())
+      if (Field->isUnnamedBitField())
         continue;
       // Recursively expand anonymous structs.
       if (Field->isAnonymousStructOrUnion()) {
@@ -9396,7 +9396,7 @@ struct SpecialMemberVisitor {
           return true;
 
     for (auto *F : RD->fields())
-      if (!F->isInvalidDecl() && !F->isUnnamedBitfield() &&
+      if (!F->isInvalidDecl() && !F->isUnnamedBitField() &&
           getDerived().visitField(F))
         return true;
 
@@ -9741,7 +9741,7 @@ bool SpecialMemberDeletionInfo::shouldDeleteForAllConstMembers() {
       AllFieldsAreConst) {
     bool AnyFields = false;
     for (auto *F : MD->getParent()->fields())
-      if ((AnyFields = !F->isUnnamedBitfield()))
+      if ((AnyFields = !F->isUnnamedBitField()))
         break;
     if (!AnyFields)
       return false;
@@ -10134,7 +10134,7 @@ static bool checkTrivialClassMembers(Sema &S, CXXRecordDecl *RD,
                                      Sema::TrivialABIHandling TAH,
                                      bool Diagnose) {
   for (const auto *FI : RD->fields()) {
-    if (FI->isInvalidDecl() || FI->isUnnamedBitfield())
+    if (FI->isInvalidDecl() || FI->isUnnamedBitField())
       continue;
 
     QualType FieldType = S.Context.getBaseElementType(FI->getType());
@@ -15201,7 +15201,7 @@ void Sema::DefineImplicitCopyAssignment(SourceLocation CurrentLocation,
   for (auto *Field : ClassDecl->fields()) {
     // FIXME: We should form some kind of AST representation for the implied
     // memcpy in a union copy operation.
-    if (Field->isUnnamedBitfield() || Field->getParent()->isUnion())
+    if (Field->isUnnamedBitField() || Field->getParent()->isUnion())
       continue;
 
     if (Field->isInvalidDecl()) {
@@ -15586,7 +15586,7 @@ void Sema::DefineImplicitMoveAssignment(SourceLocation CurrentLocation,
   for (auto *Field : ClassDecl->fields()) {
     // FIXME: We should form some kind of AST representation for the implied
     // memcpy in a union copy operation.
-    if (Field->isUnnamedBitfield() || Field->getParent()->isUnion())
+    if (Field->isUnnamedBitField() || Field->getParent()->isUnion())
       continue;
 
     if (Field->isInvalidDecl()) {
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index e86f7578ff0c0..793e16df17891 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -849,7 +849,7 @@ InitListChecker::FillInEmptyInitializations(const InitializedEntity &Entity,
       }
 
       for (auto *Field : RDecl->fields()) {
-        if (Field->isUnnamedBitfield())
+        if (Field->isUnnamedBitField())
           continue;
 
         if (hadError)
@@ -1027,7 +1027,7 @@ int InitListChecker::numStructUnionElements(QualType DeclType) {
   if (auto *CXXRD = dyn_cast<CXXRecordDecl>(structDecl))
     InitializableMembers += CXXRD->getNumBases();
   for (const auto *Field : structDecl->fields())
-    if (!Field->isUnnamedBitfield())
+    if (!Field->isUnnamedBitField())
       ++InitializableMembers;
 
   if (structDecl->isUnion())
@@ -2175,7 +2175,7 @@ void InitListChecker::CheckStructUnionTypes(
     // bitfield.
     for (RecordDecl::field_iterator FieldEnd = RD->field_end();
          Field != FieldEnd; ++Field) {
-      if (!Field->isUnnamedBitfield()) {
+      if (!Field->isUnnamedBitField()) {
         CheckEmptyInitializable(
             InitializedEntity::InitializeMember(*Field, &Entity),
             IList->getEndLoc());
@@ -2338,7 +2338,7 @@ void InitListChecker::CheckStructUnionTypes(
     if (Field->getType()->isIncompleteArrayType())
       break;
 
-    if (Field->isUnnamedBitfield()) {
+    if (Field->isUnnamedBitField()) {
       // Don't initialize unnamed bitfields, e.g. "int : 20;"
       ++Field;
       continue;
@@ -2398,7 +2398,7 @@ void InitListChecker::CheckStructUnionTypes(
       if (HasDesignatedInit && InitializedFields.count(*it))
         continue;
 
-      if (!it->isUnnamedBitfield() && !it->hasInClassInitializer() &&
+      if (!it->isUnnamedBitField() && !it->hasInClassInitializer() &&
           !it->getType()->isIncompleteArrayType()) {
         auto Diag = HasDesignatedInit
                         ? diag::warn_missing_designated_field_initializers
@@ -2414,7 +2414,7 @@ void InitListChecker::CheckStructUnionTypes(
   if (!StructuredList && Field != FieldEnd && !RD->isUnion() &&
       !Field->getType()->isIncompleteArrayType()) {
     for (; Field != FieldEnd && !hadError; ++Field) {
-      if (!Field->isUnnamedBitfield() && !Field->hasInClassInitializer())
+      if (!Field->isUnnamedBitField() && !Field->hasInClassInitializer())
         CheckEmptyInitializable(
             InitializedEntity::InitializeMember(*Field, &Entity),
             IList->getEndLoc());
@@ -2784,7 +2784,7 @@ InitListChecker::CheckDesignatedInitializer(const InitializedEntity &Entity,
     unsigned FieldIndex = NumBases;
 
     for (auto *FI : RD->fields()) {
-      if (FI->isUnnamedBitfield())
+      if (FI->isUnnamedBitField())
         continue;
       if (declaresSameEntity(KnownField, FI)) {
         KnownField = FI;
@@ -2858,7 +2858,7 @@ InitListChecker::CheckDesignatedInitializer(const InitializedEntity &Entity,
       // Find the field that we just initialized.
       FieldDecl *PrevField = nullptr;
       for (auto FI = RD->field_begin(); FI != RD->field_end(); ++FI) {
-        if (FI->isUnnamedBitfield())
+        if (FI->isUnnamedBitField())
           continue;
         if (*NextField != RD->field_end() &&
             declaresSameEntity(*FI, **NextField))
@@ -2976,7 +2976,7 @@ InitListChecker::CheckDesignatedInitializer(const InitializedEntity &Entity,
     // If this the first designator, our caller will continue checking
     // the rest of this struct/class/union subobject.
     if (IsFirstDesignator) {
-      if (Field != RD->field_end() && Field->isUnnamedBitfield())
+      if (Field != RD->field_end() && Field->isUnnamedBitField())
         ++Field;
 
       if (NextField)
@@ -5585,7 +5585,7 @@ static void TryOrBuildParenListInitialization(
     for (FieldDecl *FD : RD->fields()) {
       // Unnamed bitfields should not be initialized at all, either with an arg
       // or by default.
-      if (FD->isUnnamedBitfield())
+      if (FD->isUnnamedBitField())
         continue;
 
       InitializedEntity SubEntity =
@@ -7930,7 +7930,7 @@ static void visitLocalsRetainedByInitializer(IndirectLocalPath &Path,
         for (const auto *I : RD->fields()) {
           if (Index >= ILE->getNumInits())
             break;
-          if (I->isUnnamedBitfield())
+          if (I->isUnnamedBitField())
             continue;
           Expr *SubInit = ILE->getInit(Index);
           if (I->getType()->isReferenceType())
@@ -9532,7 +9532,7 @@ static bool DiagnoseUninitializedReference(Sema &S, SourceLocation Loc,
     return false;
 
   for (const auto *FI : RD->fields()) {
-    if (FI->isUnnamedBitfield())
+    if (FI->isUnnamedBitField())
       continue;
 
     if (DiagnoseUninitializedReference(S, FI->getLocation(), FI->getType())) {
diff --git a/clang/lib/StaticAnalyzer/Checkers/ObjCUnusedIVarsChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/ObjCUnusedIVarsChecker.cpp
index 2f2df63468b4b..23014ff954870 100644
--- a/clang/lib/StaticAnalyzer/Checkers/ObjCUnusedIVarsChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/ObjCUnusedIVarsChecker.cpp
@@ -118,8 +118,7 @@ static void checkObjCUnusedIvar(const ObjCImplementationDecl *D,
     // (d) are unnamed bitfields
     if (Ivar->getAccessControl() != ObjCIvarDecl::Private ||
         Ivar->hasAttr<UnusedAttr>() || Ivar->hasAttr<IBOutletAttr>() ||
-        Ivar->hasAttr<IBOutletCollectionAttr>() ||
-        Ivar->isUnnamedBitfield())
+        Ivar->hasAttr<IBOutletCollectionAttr>() || Ivar->isUnnamedBitField())
       continue;
 
     M[Ivar] = Unused;
diff --git a/clang/lib/StaticAnalyzer/Core/RegionStore.cpp b/clang/lib/StaticAnalyzer/Core/RegionStore.cpp
index 755a8c4b22fd9..ebba181eb2d84 100644
--- a/clang/lib/StaticAnalyzer/Core/RegionStore.cpp
+++ b/clang/lib/StaticAnalyzer/Core/RegionStore.cpp
@@ -2570,7 +2570,7 @@ std::optional<RegionBindingsRef> RegionStoreManager::tryBindSmallStruct(
       return std::nullopt;
 
   for (const auto *FD : RD->fields()) {
-    if (FD->isUnnamedBitfield())
+    if (FD->isUnnamedBitField())
       continue;
 
     // If there are too many fields, or if any of the fields are aggregates,
@@ -2697,7 +2697,7 @@ RegionBindingsRef RegionStoreManager::bindStruct(RegionBindingsConstRef B,
       break;
 
     // Skip any unnamed bitfields to stay in sync with the initializers.
-    if (FI->isUnnamedBitfield())
+    if (FI->isUnnamedBitField())
       continue;
 
     QualType FTy = FI->getType();

From dbece2bb06e7329d1390b59b3bd2e1558627c2f7 Mon Sep 17 00:00:00 2001
From: Aiden Grossman <aidengrossman@google.com>
Date: Wed, 17 Apr 2024 22:57:55 -0700
Subject: [PATCH 820/886] [Docs] Fix FAQ and Lexicon links under design
 overview (#89027)

This patch updates the FAQ and lexicon links under design overview to
actually work instead of being incomplete and thus completely missing
from the output.
---
 llvm/docs/index.rst | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/llvm/docs/index.rst b/llvm/docs/index.rst
index d1e1e785b0efb..b480729aaa5d9 100644
--- a/llvm/docs/index.rst
+++ b/llvm/docs/index.rst
@@ -24,6 +24,12 @@ Several introductory papers and presentations.
    FAQ
    Lexicon
 
+:doc:`FAQ`
+  Frequently asked questions.
+
+:doc:`Lexicon`
+  Glossary.
+
 `Introduction to the LLVM Compiler`__
   Presentation providing a users introduction to LLVM.
 

From dc74c69f016fa65949bf0cf691061e12bef93d29 Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 16:53:18 +0100
Subject: [PATCH 821/886] [libclc] Provide a more helpful error when tools are
 missing

---
 libclc/CMakeLists.txt | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index b0c29ed77270c..5ce1795243085 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -94,10 +94,11 @@ if( EXISTS ${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR} )
   endforeach()
 endif()
 
-if( NOT TARGET libclc::clang OR NOT TARGET libclc::opt
-    OR NOT TARGET libclc::llvm-as OR NOT TARGET libclc::llvm-link )
-  message( FATAL_ERROR "libclc toolchain incomplete!" )
-endif()
+foreach( tool IN ITEMS clang opt llvm-as llvm-link )
+  if( NOT TARGET libclc::${tool} )
+    message( FATAL_ERROR "libclc toolchain incomplete - missing tool ${tool}!" )
+  endif()
+endforeach()
 
 # llvm-spirv is an optional dependency, used to build spirv-* targets.
 find_program( LLVM_SPIRV llvm-spirv PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )

From 93d51194b1df8229268953ec94063fc4194a320b Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 16:55:53 +0100
Subject: [PATCH 822/886] [libclc] Clarify option help message

This should make it more obvious it applies only to libclc.
---
 libclc/CMakeLists.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index c77da2d4f18e7..4442ff6daa61b 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -32,7 +32,7 @@ set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS
 set( LIBCLC_MIN_LLVM 3.9.0 )
 
 set( LIBCLC_TARGETS_TO_BUILD "all"
-    CACHE STRING "Semicolon-separated list of targets to build, or 'all'." )
+    CACHE STRING "Semicolon-separated list of libclc targets to build, or 'all'." )
 
 option( ENABLE_RUNTIME_SUBNORMAL "Enable runtime linking of subnormal support." OFF )
 

From d0af554464774342752548c976e1ce71a2bf856d Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Wed, 17 Apr 2024 11:11:34 +0100
Subject: [PATCH 823/886] [CI] Fix libclc dependencies

We need clang and llvm to build in-tree.
---
 .ci/generate-buildkite-pipeline-premerge | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.ci/generate-buildkite-pipeline-premerge b/.ci/generate-buildkite-pipeline-premerge
index 2e503c867403b..81e9246de9b58 100755
--- a/.ci/generate-buildkite-pipeline-premerge
+++ b/.ci/generate-buildkite-pipeline-premerge
@@ -108,7 +108,7 @@ function add-dependencies() {
     compiler-rt|libc|openmp)
       echo clang lld
     ;;
-    flang|lldb)
+    flang|lldb|libclc)
       for p in llvm clang; do
         echo $p
       done

From 06f54e7c2b4f26fbcad906a6b4aa2e46bf60b8af Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 15:30:20 +0100
Subject: [PATCH 824/886] [libclc] Convert llvm-spirv to imported executable

This tool now behaves like the others, for consistency.
---
 libclc/CMakeLists.txt | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index f605c3bbbe9dc..c77da2d4f18e7 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -81,6 +81,11 @@ endif()
 # llvm-spirv is an optional dependency, used to build spirv-* targets.
 find_program( LLVM_SPIRV llvm-spirv PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
 
+if( LLVM_SPIRV )
+  add_executable( libclc::llvm-spirv IMPORTED GLOBAL )
+  set_target_properties( libclc::llvm-spirv PROPERTIES IMPORTED_LOCATION ${LLVM_SPIRV} )
+endif()
+
 # List of all targets. Note that some are added dynamically below.
 set( LIBCLC_TARGETS_ALL
   amdgcn--
@@ -101,7 +106,7 @@ endif()
 
 # spirv-mesa3d and spirv64-mesa3d targets can only be built with the (optional)
 # llvm-spirv external tool.
-if( LLVM_SPIRV )
+if( TARGET libclc::llvm-spirv )
   list( APPEND LIBCLC_TARGETS_ALL  spirv-mesa3d- spirv64-mesa3d- )
 endif()
 
@@ -114,7 +119,7 @@ list( SORT LIBCLC_TARGETS_TO_BUILD )
 # Verify that the user hasn't requested mesa3d targets without an available
 # llvm-spirv tool.
 if( "spirv-mesa3d-" IN_LIST LIBCLC_TARGETS_TO_BUILD OR "spirv64-mesa3d-" IN_LIST LIBCLC_TARGETS_TO_BUILD )
-  if( NOT LLVM_SPIRV )
+  if( NOT TARGET libclc::llvm-spirv )
     message( FATAL_ERROR "SPIR-V targets requested, but spirv-tools is not installed" )
   endif()
 endif()
@@ -363,7 +368,7 @@ foreach( t ${LIBCLC_TARGETS_TO_BUILD} )
     if( ARCH STREQUAL spirv OR ARCH STREQUAL spirv64 )
       set( spv_suffix ${arch_suffix}.spv )
       add_custom_command( OUTPUT ${spv_suffix}
-        COMMAND ${LLVM_SPIRV} ${spvflags} -o ${spv_suffix} ${builtins_link_lib}
+        COMMAND libclc::llvm-spirv ${spvflags} -o ${spv_suffix} ${builtins_link_lib}
         DEPENDS ${builtins_link_lib}
       )
       add_custom_target( "prepare-${spv_suffix}" ALL DEPENDS "${spv_suffix}" )

From 0aeeff3059e79b86f55ed92a4488bdee8fa66e12 Mon Sep 17 00:00:00 2001
From: Fraser Cormack <fraser@codeplay.com>
Date: Tue, 16 Apr 2024 16:46:37 +0100
Subject: [PATCH 825/886] [libclc] Allow building with pre-built tools

Building the libclc project in-tree with debug tools can be very slow.
This commit adds an option for a user to specify a dierctory from which
to import (e.g., release-built) tools. All tools required by the project
must be imported from the same location, for simplicity.

Original patch downstream authored by @jchlanda.
---
 libclc/CMakeLists.txt | 33 +++++++++++++++++++++++++++------
 1 file changed, 27 insertions(+), 6 deletions(-)

diff --git a/libclc/CMakeLists.txt b/libclc/CMakeLists.txt
index 4442ff6daa61b..b0c29ed77270c 100644
--- a/libclc/CMakeLists.txt
+++ b/libclc/CMakeLists.txt
@@ -50,11 +50,13 @@ if( LIBCLC_STANDALONE_BUILD OR CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DI
   endif()
 
   # Import required tools as targets
-  foreach( tool IN ITEMS clang llvm-as llvm-link opt )
-    find_program( LLVM_TOOL_${tool} ${tool} PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
-    add_executable( libclc::${tool} IMPORTED GLOBAL )
-    set_target_properties( libclc::${tool} PROPERTIES IMPORTED_LOCATION ${LLVM_TOOL_${tool}} )
-  endforeach()
+  if( NOT EXISTS ${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR} )
+    foreach( tool IN ITEMS clang llvm-as llvm-link opt )
+      find_program( LLVM_TOOL_${tool} ${tool} PATHS ${LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
+      add_executable( libclc::${tool} IMPORTED GLOBAL )
+      set_target_properties( libclc::${tool} PROPERTIES IMPORTED_LOCATION ${LLVM_TOOL_${tool}} )
+    endforeach()
+  endif()
 else()
   # In-tree configuration
   set( LIBCLC_STANDALONE_BUILD FALSE )
@@ -68,8 +70,27 @@ else()
     message(FATAL_ERROR "Clang is not enabled, but is required to build libclc in-tree")
   endif()
 
+  if( NOT EXISTS ${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR} )
+    foreach( tool IN ITEMS clang llvm-as llvm-link opt )
+      add_executable(libclc::${tool} ALIAS ${tool})
+    endforeach()
+  endif()
+endif()
+
+if( EXISTS ${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR} )
+  message( WARNING "Using custom LLVM tools to build libclc: "
+    "${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR}, "
+    " ensure the tools are up to date." )
+  # Note - use a differently named variable than LLVM_TOOL_${tool} as above, as
+  # the variable name is used to cache the result of find_program. If we used
+  # the same name, a user wouldn't be able to switch a build between default
+  # and custom tools.
   foreach( tool IN ITEMS clang llvm-as llvm-link opt )
-    add_executable(libclc::${tool} ALIAS ${tool})
+    find_program( LLVM_CUSTOM_TOOL_${tool} ${tool}
+      PATHS ${LIBCLC_CUSTOM_LLVM_TOOLS_BINARY_DIR} NO_DEFAULT_PATH )
+    add_executable( libclc::${tool} IMPORTED GLOBAL )
+    set_target_properties( libclc::${tool} PROPERTIES
+      IMPORTED_LOCATION ${LLVM_CUSTOM_TOOL_${tool}} )
   endforeach()
 endif()
 

From a5d2ed240d788a5f0caf090669eee27e51a28a89 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 15:06:20 +0900
Subject: [PATCH 826/886] [VectorCombine] Fix typo in test (NFC)

The llvm.smix should be llvm.smin.
---
 .../AArch64/shuffletoidentity.ll              | 26 +++++++++----------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
index a1d4ca1a740e7..56b42380a68bf 100644
--- a/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
+++ b/llvm/test/Transforms/VectorCombine/AArch64/shuffletoidentity.ll
@@ -338,12 +338,12 @@ define <8 x i8> @extrause_add(<8 x i8> %a, <8 x i8> %b) {
 
 define <8 x i8> @extrause_shuffle(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-LABEL: @extrause_shuffle(
-; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AB:%.*]] = shufflevector <8 x i8> [[A:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[AT:%.*]] = shufflevector <8 x i8> [[A]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
+; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
-; CHECK-NEXT:    [[BB:%.*]] = shufflevector <8 x i8> [[B1:%.*]], <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    [[BT1:%.*]] = shufflevector <8 x i8> [[B1]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
-; CHECK-NEXT:    call void @use(<4 x i8> [[BT1]])
-; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[BT]], [[BT1]]
+; CHECK-NEXT:    call void @use(<4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i8> [[AT]], [[BT]]
 ; CHECK-NEXT:    [[ABB:%.*]] = add <4 x i8> [[AB]], [[BB]]
 ; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT]], <4 x i8> [[ABB]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    ret <8 x i8> [[R]]
@@ -431,8 +431,8 @@ define void @exttrunc(<8 x i32> %a, <8 x i32> %b, ptr %p) {
 ; CHECK-NEXT:    [[ABT:%.*]] = add <4 x i64> [[AT1]], [[BT1]]
 ; CHECK-NEXT:    [[ABB1:%.*]] = trunc <4 x i64> [[ABB]] to <4 x i32>
 ; CHECK-NEXT:    [[ABT1:%.*]] = trunc <4 x i64> [[ABT]] to <4 x i32>
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[ABB1]], <4 x i32> [[ABT1]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:    store <8 x i32> [[TMP1]], ptr [[P:%.*]], align 32
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i32> [[ABB1]], <4 x i32> [[ABT1]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:    store <8 x i32> [[R]], ptr [[P:%.*]], align 32
 ; CHECK-NEXT:    ret void
 ;
   %ab = shufflevector <8 x i32> %a, <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
@@ -460,14 +460,14 @@ define <8 x i8> @intrinsics_minmax(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-NEXT:    [[BT:%.*]] = shufflevector <8 x i8> [[B]], <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
 ; CHECK-NEXT:    [[ABT:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[AT]], <4 x i8> [[BT]])
 ; CHECK-NEXT:    [[ABB:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[AB]], <4 x i8> [[BB]])
-; CHECK-NEXT:    [[ABT1:%.*]] = call <4 x i8> @llvm.smix.v4i8(<4 x i8> [[ABT]], <4 x i8> [[BT]])
+; CHECK-NEXT:    [[ABT1:%.*]] = call <4 x i8> @llvm.smin.v4i8(<4 x i8> [[ABT]], <4 x i8> [[BT]])
 ; CHECK-NEXT:    [[ABB1:%.*]] = call <4 x i8> @llvm.smax.v4i8(<4 x i8> [[ABB]], <4 x i8> [[BB]])
 ; CHECK-NEXT:    [[ABT2:%.*]] = call <4 x i8> @llvm.umin.v4i8(<4 x i8> [[ABT1]], <4 x i8> [[BT]])
 ; CHECK-NEXT:    [[ABB2:%.*]] = call <4 x i8> @llvm.umin.v4i8(<4 x i8> [[ABB1]], <4 x i8> [[BB]])
 ; CHECK-NEXT:    [[ABT3:%.*]] = call <4 x i8> @llvm.umax.v4i8(<4 x i8> [[ABT2]], <4 x i8> [[BT]])
 ; CHECK-NEXT:    [[ABB3:%.*]] = call <4 x i8> @llvm.umax.v4i8(<4 x i8> [[ABB2]], <4 x i8> [[BB]])
-; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    ret <8 x i8> [[R1]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
 ;
   %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
@@ -475,7 +475,7 @@ define <8 x i8> @intrinsics_minmax(<8 x i8> %a, <8 x i8> %b) {
   %bt = shufflevector <8 x i8> %b, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
   %abt = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %at, <4 x i8> %bt)
   %abb = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %ab, <4 x i8> %bb)
-  %abt1 = call <4 x i8> @llvm.smix.v4i8(<4 x i8> %abt, <4 x i8> %bt)
+  %abt1 = call <4 x i8> @llvm.smin.v4i8(<4 x i8> %abt, <4 x i8> %bt)
   %abb1 = call <4 x i8> @llvm.smax.v4i8(<4 x i8> %abb, <4 x i8> %bb)
   %abt2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abt1, <4 x i8> %bt)
   %abb2 = call <4 x i8> @llvm.umin.v4i8(<4 x i8> %abb1, <4 x i8> %bb)
@@ -499,8 +499,8 @@ define <8 x i8> @intrinsics_addsat(<8 x i8> %a, <8 x i8> %b) {
 ; CHECK-NEXT:    [[ABB2:%.*]] = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> [[ABB1]], <4 x i8> [[BB]])
 ; CHECK-NEXT:    [[ABT3:%.*]] = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> [[ABT2]], <4 x i8> [[BT]])
 ; CHECK-NEXT:    [[ABB3:%.*]] = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> [[ABB2]], <4 x i8> [[BB]])
-; CHECK-NEXT:    [[R1:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:    ret <8 x i8> [[R1]]
+; CHECK-NEXT:    [[R:%.*]] = shufflevector <4 x i8> [[ABT3]], <4 x i8> [[ABB3]], <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    ret <8 x i8> [[R]]
 ;
   %ab = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %at = shufflevector <8 x i8> %a, <8 x i8> poison, <4 x i32> <i32 7, i32 6, i32 5, i32 4>

From 5ff44dbaea27d442c89278871dc2f75942d54716 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 15:07:48 +0900
Subject: [PATCH 827/886] [VectorCombine] Fix incorrect intrinsic signature
 (NFC)

---
 .../Transforms/VectorCombine/X86/reduction-of-truncations.ll    | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm/test/Transforms/VectorCombine/X86/reduction-of-truncations.ll b/llvm/test/Transforms/VectorCombine/X86/reduction-of-truncations.ll
index d5dd4cf0e34f9..f09ac41793b51 100644
--- a/llvm/test/Transforms/VectorCombine/X86/reduction-of-truncations.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/reduction-of-truncations.ll
@@ -49,7 +49,7 @@ define i8 @reduce_or_trunc_v8i32_i8(<8 x i32> %a0)  {
   %red = tail call i8 @llvm.vector.reduce.or.v8i32(<8 x i8> %tr)
   ret i8 %red
 }
-declare i32 @llvm.vector.reduce.or.v8i8(<8 x i8>)
+declare i8 @llvm.vector.reduce.or.v8i8(<8 x i8>)
 
 define i8 @reduce_xor_trunc_v16i64_i8(<16 x i64> %a0)  {
 ; CHECK-LABEL: @reduce_xor_trunc_v16i64_i8(

From 9099f6f8dcffffd7cf7d5cbe9ca4b7582851fde1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Wed, 17 Apr 2024 14:35:20 +0200
Subject: [PATCH 828/886] [clang][Interp] Skip unnamed bit fields in
 initializers

Fixes the Codegen/CSKY/csky-hard-abi.c test
---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp |  5 +++++
 clang/lib/AST/Interp/Program.cpp         |  5 +++++
 clang/test/AST/Interp/records.cpp        | 13 +++++++++++++
 3 files changed, 23 insertions(+)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 6b4b51aac41e8..52e3efff2359d 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -971,6 +971,11 @@ bool ByteCodeExprGen<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
 
   unsigned InitIndex = 0;
   for (const Expr *Init : Inits) {
+    // Skip unnamed bitfields.
+    while (InitIndex < R->getNumFields() &&
+           R->getField(InitIndex)->Decl->isUnnamedBitField())
+      ++InitIndex;
+
     if (!this->emitDupPtr(E))
       return false;
 
diff --git a/clang/lib/AST/Interp/Program.cpp b/clang/lib/AST/Interp/Program.cpp
index e6f22e79451e9..2c8c6781b3483 100644
--- a/clang/lib/AST/Interp/Program.cpp
+++ b/clang/lib/AST/Interp/Program.cpp
@@ -312,6 +312,11 @@ Record *Program::getOrCreateRecord(const RecordDecl *RD) {
   // Reserve space for fields.
   Record::FieldList Fields;
   for (const FieldDecl *FD : RD->fields()) {
+    // Note that we DO create fields and descriptors
+    // for unnamed bitfields here, even though we later ignore
+    // them everywhere. That's because so the FieldDecl's
+    // getFieldIndex() matches.
+
     // Reserve space for the field's descriptor and the offset.
     BaseSize += align(sizeof(InlineDescriptor));
 
diff --git a/clang/test/AST/Interp/records.cpp b/clang/test/AST/Interp/records.cpp
index 2c33fa1bf8843..3e52354a4a106 100644
--- a/clang/test/AST/Interp/records.cpp
+++ b/clang/test/AST/Interp/records.cpp
@@ -1317,3 +1317,16 @@ namespace {
   F f;
   static_assert(f.Z == 12, "");
 }
+
+namespace UnnamedBitFields {
+  struct A {
+    int : 1;
+    double f;
+    int : 1;
+    char c;
+  };
+
+  constexpr A a = (A){1.0, 'a'};
+  static_assert(a.f == 1.0, "");
+  static_assert(a.c == 'a', "");
+}

From 43eb5e2d4054104dcdd9fadeea8da8bacab6d338 Mon Sep 17 00:00:00 2001
From: Craig Topper <craig.topper@sifive.com>
Date: Wed, 17 Apr 2024 23:24:03 -0700
Subject: [PATCH 829/886] [RISCV] Remove unused Predicates for I and E
 extensions. NFC

---
 llvm/lib/Target/RISCV/RISCVFeatures.td | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index f830ead5dd692..8c434a23b10ee 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -15,9 +15,7 @@
 def FeatureStdExtI
     : SubtargetFeature<"i", "HasStdExtI", "true",
                        "'I' (Base Integer Instruction Set)">;
-def HasStdExtI : Predicate<"Subtarget->hasStdExtI()">,
-                 AssemblerPredicate<(all_of FeatureStdExtI),
-                                    "'I' (Base Integer Instruction Set)">;
+
 def FeatureStdExtZic64b
     : SubtargetFeature<"zic64b", "HasStdExtZic64b", "true",
                        "'Zic64b' (Cache Block Size Is 64 Bytes)">;
@@ -1170,8 +1168,6 @@ def RV64           : HwMode<"+64bit", [IsRV64]>;
 def FeatureRVE
     : SubtargetFeature<"e", "IsRVE", "true",
                        "Implements RV{32,64}E (provides 16 rather than 32 GPRs)">;
-def IsRVE : Predicate<"Subtarget->isRVE()">,
-            AssemblerPredicate<(all_of FeatureRVE)>;
 
 def FeatureRelax
     : SubtargetFeature<"relax", "EnableLinkerRelax", "true",

From ff3523f67b858b87ed4f72d143a1ab3634ae1db7 Mon Sep 17 00:00:00 2001
From: Andreas Jonson <andjo403@hotmail.com>
Date: Thu, 18 Apr 2024 08:27:36 +0200
Subject: [PATCH 830/886] [IR] Drop poison-generating return attributes when
 necessary (#89138)

Rename has/dropPoisonGeneratingFlagsOrMetadata to
has/dropPoisonGeneratingAnnotations and make it also handle
nonnull, align and range return attributes on calls, similar
to the existing handling for !nonnull, !align and !range metadata.
---
 llvm/include/llvm/IR/Instruction.h            | 20 ++++++---
 llvm/include/llvm/IR/Operator.h               |  7 ++--
 llvm/lib/Analysis/InstructionSimplify.cpp     |  2 +-
 llvm/lib/Analysis/ScalarEvolution.cpp         |  2 +-
 llvm/lib/Analysis/ValueTracking.cpp           |  2 +-
 llvm/lib/IR/Instruction.cpp                   | 23 +++++++++++
 llvm/lib/IR/Operator.cpp                      |  5 ++-
 .../InstCombine/InstCombineSelect.cpp         |  2 +-
 .../InstCombine/InstructionCombining.cpp      |  4 +-
 llvm/lib/Transforms/Scalar/BDCE.cpp           |  2 +-
 llvm/lib/Transforms/Scalar/GuardWidening.cpp  |  2 +-
 .../Utils/ScalarEvolutionExpander.cpp         |  2 +-
 llvm/lib/Transforms/Utils/SimplifyIndVar.cpp  |  2 +-
 llvm/test/Transforms/InstCombine/freeze.ll    | 41 ++++++++++++++++++-
 14 files changed, 95 insertions(+), 21 deletions(-)

diff --git a/llvm/include/llvm/IR/Instruction.h b/llvm/include/llvm/IR/Instruction.h
index 6e0874c5b04f2..249be2799fa93 100644
--- a/llvm/include/llvm/IR/Instruction.h
+++ b/llvm/include/llvm/IR/Instruction.h
@@ -496,14 +496,24 @@ class Instruction : public User,
   /// Drops metadata that may generate poison.
   void dropPoisonGeneratingMetadata();
 
-  /// Return true if this instruction has poison-generating flags or metadata.
-  bool hasPoisonGeneratingFlagsOrMetadata() const {
-    return hasPoisonGeneratingFlags() || hasPoisonGeneratingMetadata();
+  /// Return true if this instruction has poison-generating attribute.
+  bool hasPoisonGeneratingReturnAttributes() const LLVM_READONLY;
+
+  /// Drops return attributes that may generate poison.
+  void dropPoisonGeneratingReturnAttributes();
+
+  /// Return true if this instruction has poison-generating flags,
+  /// return attributes or metadata.
+  bool hasPoisonGeneratingAnnotations() const {
+    return hasPoisonGeneratingFlags() ||
+           hasPoisonGeneratingReturnAttributes() ||
+           hasPoisonGeneratingMetadata();
   }
 
-  /// Drops flags and metadata that may generate poison.
-  void dropPoisonGeneratingFlagsAndMetadata() {
+  /// Drops flags, return attributes and metadata that may generate poison.
+  void dropPoisonGeneratingAnnotations() {
     dropPoisonGeneratingFlags();
+    dropPoisonGeneratingReturnAttributes();
     dropPoisonGeneratingMetadata();
   }
 
diff --git a/llvm/include/llvm/IR/Operator.h b/llvm/include/llvm/IR/Operator.h
index e14a916b47493..b2307948bbbc4 100644
--- a/llvm/include/llvm/IR/Operator.h
+++ b/llvm/include/llvm/IR/Operator.h
@@ -64,9 +64,10 @@ class Operator : public User {
   /// to evaluate to poison despite having non-poison inputs.
   bool hasPoisonGeneratingFlags() const;
 
-  /// Return true if this operator has poison-generating flags or metadata.
-  /// The latter is only possible for instructions.
-  bool hasPoisonGeneratingFlagsOrMetadata() const;
+  /// Return true if this operator has poison-generating flags,
+  /// return attributes or metadata. The latter two is only possible for
+  /// instructions.
+  bool hasPoisonGeneratingAnnotations() const;
 };
 
 /// Utility class for integer operators which may exhibit overflow - Add, Sub,
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 06ba5ca4c6b35..431ad37d0a9e8 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -4431,7 +4431,7 @@ static Value *simplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,
         return nullptr;
     }
     Constant *Res = ConstantFoldInstOperands(I, ConstOps, Q.DL, Q.TLI);
-    if (DropFlags && Res && I->hasPoisonGeneratingFlagsOrMetadata())
+    if (DropFlags && Res && I->hasPoisonGeneratingAnnotations())
       DropFlags->push_back(I);
     return Res;
   }
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index 95440dda3b4c0..93f885c5d5ad8 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -4239,7 +4239,7 @@ bool ScalarEvolution::canReuseInstruction(
       return false;
 
     // If the instruction can't create poison, we can recurse to its operands.
-    if (I->hasPoisonGeneratingFlagsOrMetadata())
+    if (I->hasPoisonGeneratingAnnotations())
       DropPoisonGeneratingInsts.push_back(I);
 
     for (Value *Op : I->operands())
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index ab2f43e1033fa..695d0b0b1a2f8 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -6951,7 +6951,7 @@ static bool canCreateUndefOrPoison(const Operator *Op, UndefPoisonKind Kind,
                                    bool ConsiderFlagsAndMetadata) {
 
   if (ConsiderFlagsAndMetadata && includesPoison(Kind) &&
-      Op->hasPoisonGeneratingFlagsOrMetadata())
+      Op->hasPoisonGeneratingAnnotations())
     return true;
 
   unsigned Opcode = Op->getOpcode();
diff --git a/llvm/lib/IR/Instruction.cpp b/llvm/lib/IR/Instruction.cpp
index b9efe9cdcfe31..6cab3c1512416 100644
--- a/llvm/lib/IR/Instruction.cpp
+++ b/llvm/lib/IR/Instruction.cpp
@@ -13,7 +13,9 @@
 #include "llvm/IR/Instruction.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/IR/AttributeMask.h"
+#include "llvm/IR/Attributes.h"
 #include "llvm/IR/Constants.h"
+#include "llvm/IR/InstrTypes.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Intrinsics.h"
@@ -472,6 +474,27 @@ void Instruction::dropPoisonGeneratingMetadata() {
   eraseMetadata(LLVMContext::MD_align);
 }
 
+bool Instruction::hasPoisonGeneratingReturnAttributes() const {
+  if (const auto *CB = dyn_cast<CallBase>(this)) {
+    AttributeSet RetAttrs = CB->getAttributes().getRetAttrs();
+    return RetAttrs.hasAttribute(Attribute::Range) ||
+           RetAttrs.hasAttribute(Attribute::Alignment) ||
+           RetAttrs.hasAttribute(Attribute::NonNull);
+  }
+  return false;
+}
+
+void Instruction::dropPoisonGeneratingReturnAttributes() {
+  if (auto *CB = dyn_cast<CallBase>(this)) {
+    AttributeMask AM;
+    AM.addAttribute(Attribute::Range);
+    AM.addAttribute(Attribute::Alignment);
+    AM.addAttribute(Attribute::NonNull);
+    CB->removeRetAttrs(AM);
+  }
+  assert(!hasPoisonGeneratingReturnAttributes() && "must be kept in sync");
+}
+
 void Instruction::dropUBImplyingAttrsAndUnknownMetadata(
     ArrayRef<unsigned> KnownIDs) {
   dropUnknownNonDebugMetadata(KnownIDs);
diff --git a/llvm/lib/IR/Operator.cpp b/llvm/lib/IR/Operator.cpp
index ccc624d854429..29620ef716f25 100644
--- a/llvm/lib/IR/Operator.cpp
+++ b/llvm/lib/IR/Operator.cpp
@@ -56,11 +56,12 @@ bool Operator::hasPoisonGeneratingFlags() const {
   }
 }
 
-bool Operator::hasPoisonGeneratingFlagsOrMetadata() const {
+bool Operator::hasPoisonGeneratingAnnotations() const {
   if (hasPoisonGeneratingFlags())
     return true;
   auto *I = dyn_cast<Instruction>(this);
-  return I && I->hasPoisonGeneratingMetadata();
+  return I && (I->hasPoisonGeneratingReturnAttributes() ||
+               I->hasPoisonGeneratingMetadata());
 }
 
 Type *GEPOperator::getSourceElementType() const {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index 41c3ab85940ea..a87d392916159 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -1340,7 +1340,7 @@ Instruction *InstCombinerImpl::foldSelectValueEquivalence(SelectInst &Sel,
                              /* AllowRefinement */ false,
                              &DropFlags) == TrueVal) {
     for (Instruction *I : DropFlags) {
-      I->dropPoisonGeneratingFlagsAndMetadata();
+      I->dropPoisonGeneratingAnnotations();
       Worklist.add(I);
     }
 
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 5a144cc737896..e08d0a80e7653 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -4330,7 +4330,7 @@ InstCombinerImpl::pushFreezeToPreventPoisonFromPropagating(FreezeInst &OrigFI) {
       return nullptr;
   }
 
-  OrigOpInst->dropPoisonGeneratingFlagsAndMetadata();
+  OrigOpInst->dropPoisonGeneratingAnnotations();
 
   // If all operands are guaranteed to be non-poison, we can drop freeze.
   if (!MaybePoisonOperand)
@@ -4401,7 +4401,7 @@ Instruction *InstCombinerImpl::foldFreezeIntoRecurrence(FreezeInst &FI,
   }
 
   for (Instruction *I : DropFlags)
-    I->dropPoisonGeneratingFlagsAndMetadata();
+    I->dropPoisonGeneratingAnnotations();
 
   if (StartNeedsFreeze) {
     Builder.SetInsertPoint(StartBB->getTerminator());
diff --git a/llvm/lib/Transforms/Scalar/BDCE.cpp b/llvm/lib/Transforms/Scalar/BDCE.cpp
index 4aa87a84f115a..d96dbca30fdb1 100644
--- a/llvm/lib/Transforms/Scalar/BDCE.cpp
+++ b/llvm/lib/Transforms/Scalar/BDCE.cpp
@@ -75,7 +75,7 @@ static void clearAssumptionsOfUsers(Instruction *I, DemandedBits &DB) {
     Instruction *J = WorkList.pop_back_val();
 
     // NSW, NUW, and exact are based on operands that might have changed.
-    J->dropPoisonGeneratingFlagsAndMetadata();
+    J->dropPoisonGeneratingAnnotations();
 
     // We do not have to worry about llvm.assume, because it demands its
     // operand, so trivializing can't change it.
diff --git a/llvm/lib/Transforms/Scalar/GuardWidening.cpp b/llvm/lib/Transforms/Scalar/GuardWidening.cpp
index d3787b28347c9..b6498c4e36963 100644
--- a/llvm/lib/Transforms/Scalar/GuardWidening.cpp
+++ b/llvm/lib/Transforms/Scalar/GuardWidening.cpp
@@ -699,7 +699,7 @@ Value *GuardWideningImpl::freezeAndPush(Value *Orig,
         Worklist.push_back(U.get());
   }
   for (Instruction *I : DropPoisonFlags)
-    I->dropPoisonGeneratingFlagsAndMetadata();
+    I->dropPoisonGeneratingAnnotations();
 
   Value *Result = Orig;
   for (Value *V : NeedFreeze) {
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index 74cffbc005c82..0feea0a4233cd 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -1522,7 +1522,7 @@ Value *SCEVExpander::expand(const SCEV *S) {
   } else {
     for (Instruction *I : DropPoisonGeneratingInsts) {
       rememberFlags(I);
-      I->dropPoisonGeneratingFlagsAndMetadata();
+      I->dropPoisonGeneratingAnnotations();
       // See if we can re-infer from first principles any of the flags we just
       // dropped.
       if (auto *OBO = dyn_cast<OverflowingBinaryOperator>(I))
diff --git a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
index 31be7d62c8d1d..912c02c2ed3ae 100644
--- a/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
@@ -768,7 +768,7 @@ bool SimplifyIndvar::eliminateIdentitySCEV(Instruction *UseInst,
       return false;
 
     for (Instruction *I : DropPoisonGeneratingInsts)
-      I->dropPoisonGeneratingFlagsAndMetadata();
+      I->dropPoisonGeneratingAnnotations();
   }
 
   LLVM_DEBUG(dbgs() << "INDVARS: Eliminated identity: " << *UseInst << '\n');
diff --git a/llvm/test/Transforms/InstCombine/freeze.ll b/llvm/test/Transforms/InstCombine/freeze.ll
index 668f3033ed4b7..adcf7d50f413b 100644
--- a/llvm/test/Transforms/InstCombine/freeze.ll
+++ b/llvm/test/Transforms/InstCombine/freeze.ll
@@ -1069,7 +1069,7 @@ define ptr @freeze_load_dereferenceable(ptr %ptr) {
 
 define ptr @freeze_load_dereferenceable_or_null(ptr %ptr) {
 ; CHECK-LABEL: @freeze_load_dereferenceable_or_null(
-; CHECK-NEXT:    [[P:%.*]] = load ptr, ptr [[PTR:%.*]], align 8, !dereferenceable_or_null !1
+; CHECK-NEXT:    [[P:%.*]] = load ptr, ptr [[PTR:%.*]], align 8, !dereferenceable_or_null [[META1]]
 ; CHECK-NEXT:    ret ptr [[P]]
 ;
   %p = load ptr, ptr %ptr, !dereferenceable_or_null !1
@@ -1160,6 +1160,45 @@ define i32 @propagate_drop_flags_trunc(i64 %arg) {
   ret i32 %v1.fr
 }
 
+declare i32 @llvm.umax.i32(i32 %a, i32 %b)
+
+define i32 @freeze_call_with_range_attr(i32 %a) {
+; CHECK-LABEL: @freeze_call_with_range_attr(
+; CHECK-NEXT:    [[Y:%.*]] = lshr i32 2047, [[A:%.*]]
+; CHECK-NEXT:    [[Y_FR:%.*]] = freeze i32 [[Y]]
+; CHECK-NEXT:    [[X:%.*]] = call i32 @llvm.umax.i32(i32 [[Y_FR]], i32 50)
+; CHECK-NEXT:    ret i32 [[X]]
+;
+  %y = lshr i32 2047, %a
+  %x = call range(i32 0, 2048) i32 @llvm.umax.i32(i32 %y, i32 50)
+  %x.fr = freeze i32 %x
+  ret i32 %x.fr
+}
+
+declare ptr @llvm.ptrmask.p0.i64(ptr, i64)
+
+define ptr @freeze_ptrmask_align(ptr %p, i64 noundef %m) {
+; CHECK-LABEL: @freeze_ptrmask_align(
+; CHECK-NEXT:    [[P_FR:%.*]] = freeze ptr [[P:%.*]]
+; CHECK-NEXT:    [[MASK:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P_FR]], i64 [[M:%.*]])
+; CHECK-NEXT:    ret ptr [[MASK]]
+;
+  %mask = call align(4) ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m)
+  %fr =  freeze ptr %mask
+  ret ptr %fr
+}
+
+define ptr @freeze_ptrmask_nonnull(ptr %p, i64 noundef %m) {
+; CHECK-LABEL: @freeze_ptrmask_nonnull(
+; CHECK-NEXT:    [[P_FR:%.*]] = freeze ptr [[P:%.*]]
+; CHECK-NEXT:    [[MASK:%.*]] = call ptr @llvm.ptrmask.p0.i64(ptr [[P_FR]], i64 [[M:%.*]])
+; CHECK-NEXT:    ret ptr [[MASK]]
+;
+  %mask = call nonnull ptr @llvm.ptrmask.p0.i64(ptr %p, i64 %m)
+  %fr =  freeze ptr %mask
+  ret ptr %fr
+}
+
 !0 = !{}
 !1 = !{i64 4}
 !2 = !{i32 0, i32 100}

From 562f061e7e710543578875d33d64837abecc23d2 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Thu, 18 Apr 2024 07:36:49 +0200
Subject: [PATCH 831/886] [clang][Interp] Load result of pre-inc/dec operation
 if necessary

This can happen in C.
---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp | 42 +++++++++++++++---------
 clang/test/AST/Interp/c.c                | 12 +++++++
 2 files changed, 38 insertions(+), 16 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 52e3efff2359d..47cd32c2ab867 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -3211,15 +3211,20 @@ bool ByteCodeExprGen<Emitter>::VisitUnaryOperator(const UnaryOperator *E) {
         return false;
       if (!this->emitAddf(getRoundingMode(E), E))
         return false;
-      return this->emitStoreFloat(E);
+      if (!this->emitStoreFloat(E))
+        return false;
+    } else {
+      assert(isIntegralType(*T));
+      if (!this->emitLoad(*T, E))
+        return false;
+      if (!this->emitConst(1, E))
+        return false;
+      if (!this->emitAdd(*T, E))
+        return false;
+      if (!this->emitStore(*T, E))
+        return false;
     }
-    if (!this->emitLoad(*T, E))
-      return false;
-    if (!this->emitConst(1, E))
-      return false;
-    if (!this->emitAdd(*T, E))
-      return false;
-    return this->emitStore(*T, E);
+    return E->isGLValue() || this->emitLoadPop(*T, E);
   }
   case UO_PreDec: { // --x
     if (!this->visit(SubExpr))
@@ -3250,15 +3255,20 @@ bool ByteCodeExprGen<Emitter>::VisitUnaryOperator(const UnaryOperator *E) {
         return false;
       if (!this->emitSubf(getRoundingMode(E), E))
         return false;
-      return this->emitStoreFloat(E);
+      if (!this->emitStoreFloat(E))
+        return false;
+    } else {
+      assert(isIntegralType(*T));
+      if (!this->emitLoad(*T, E))
+        return false;
+      if (!this->emitConst(1, E))
+        return false;
+      if (!this->emitSub(*T, E))
+        return false;
+      if (!this->emitStore(*T, E))
+        return false;
     }
-    if (!this->emitLoad(*T, E))
-      return false;
-    if (!this->emitConst(1, E))
-      return false;
-    if (!this->emitSub(*T, E))
-      return false;
-    return this->emitStore(*T, E);
+    return E->isGLValue() || this->emitLoadPop(*T, E);
   }
   case UO_LNot: // !x
     if (DiscardResult)
diff --git a/clang/test/AST/Interp/c.c b/clang/test/AST/Interp/c.c
index e0b18120fd211..38df38d1ccfae 100644
--- a/clang/test/AST/Interp/c.c
+++ b/clang/test/AST/Interp/c.c
@@ -233,3 +233,15 @@ _Static_assert(funcp == (void*)0, ""); // all-error {{failed due to requirement
                                        // pedantic-warning {{expression is not an integer constant expression}}
 _Static_assert(funcp == (void*)123, ""); // pedantic-warning {{equality comparison between function pointer and void pointer}} \
                                          // pedantic-warning {{expression is not an integer constant expression}}
+
+void unaryops(void) {
+  (void)(++(struct x {unsigned x;}){3}.x);
+  (void)(--(struct y {unsigned x;}){3}.x);
+  (void)(++(struct z {float x;}){3}.x);
+  (void)(--(struct w {float x;}){3}.x);
+
+  (void)((struct xx {unsigned x;}){3}.x++);
+  (void)((struct yy {unsigned x;}){3}.x--);
+  (void)((struct zz {float x;}){3}.x++);
+  (void)((struct ww {float x;}){3}.x--);
+}

From 71c0784dc4a2ef40039a97af122ba78549193120 Mon Sep 17 00:00:00 2001
From: Vassil Vassilev <v.g.vassilev@gmail.com>
Date: Thu, 18 Apr 2024 09:32:34 +0300
Subject: [PATCH 832/886] Fix the double space and double attribute printing of
 the final keyword. (#88600)

Fixes #56517.
---
 clang/lib/AST/DeclPrinter.cpp               | 36 +++++++++++---------
 clang/test/SemaCXX/cxx11-attr-print.cpp     |  5 +++
 clang/test/SemaHLSL/BuiltIns/RWBuffers.hlsl |  4 +--
 clang/unittests/AST/DeclPrinterTest.cpp     | 37 ++++++++++++++++-----
 clang/utils/TableGen/ClangAttrEmitter.cpp   |  2 +-
 5 files changed, 57 insertions(+), 27 deletions(-)

diff --git a/clang/lib/AST/DeclPrinter.cpp b/clang/lib/AST/DeclPrinter.cpp
index 93857adb990bf..599d379340aba 100644
--- a/clang/lib/AST/DeclPrinter.cpp
+++ b/clang/lib/AST/DeclPrinter.cpp
@@ -119,7 +119,7 @@ namespace {
     void printTemplateArguments(llvm::ArrayRef<TemplateArgumentLoc> Args,
                                 const TemplateParameterList *Params);
     enum class AttrPosAsWritten { Default = 0, Left, Right };
-    void
+    bool
     prettyPrintAttributes(const Decl *D,
                           AttrPosAsWritten Pos = AttrPosAsWritten::Default);
     void prettyPrintPragmas(Decl *D);
@@ -252,16 +252,19 @@ static DeclPrinter::AttrPosAsWritten getPosAsWritten(const Attr *A,
   return DeclPrinter::AttrPosAsWritten::Right;
 }
 
-void DeclPrinter::prettyPrintAttributes(const Decl *D,
+// returns true if an attribute was printed.
+bool DeclPrinter::prettyPrintAttributes(const Decl *D,
                                         AttrPosAsWritten Pos /*=Default*/) {
-  if (Policy.PolishForDeclaration)
-    return;
+  bool hasPrinted = false;
 
   if (D->hasAttrs()) {
     const AttrVec &Attrs = D->getAttrs();
     for (auto *A : Attrs) {
       if (A->isInherited() || A->isImplicit())
         continue;
+      // Print out the keyword attributes, they aren't regular attributes.
+      if (Policy.PolishForDeclaration && !A->isKeywordAttribute())
+        continue;
       switch (A->getKind()) {
 #define ATTR(X)
 #define PRAGMA_SPELLING_ATTR(X) case attr::X:
@@ -275,6 +278,7 @@ void DeclPrinter::prettyPrintAttributes(const Decl *D,
           if (Pos != AttrPosAsWritten::Left)
             Out << ' ';
           A->printPretty(Out, Policy);
+          hasPrinted = true;
           if (Pos == AttrPosAsWritten::Left)
             Out << ' ';
         }
@@ -282,6 +286,7 @@ void DeclPrinter::prettyPrintAttributes(const Decl *D,
       }
     }
   }
+  return hasPrinted;
 }
 
 void DeclPrinter::prettyPrintPragmas(Decl *D) {
@@ -1065,12 +1070,15 @@ void DeclPrinter::VisitCXXRecordDecl(CXXRecordDecl *D) {
   // FIXME: add printing of pragma attributes if required.
   if (!Policy.SuppressSpecifiers && D->isModulePrivate())
     Out << "__module_private__ ";
-  Out << D->getKindName();
 
-  prettyPrintAttributes(D);
+  Out << D->getKindName() << ' ';
 
-  if (D->getIdentifier()) {
+  // FIXME: Move before printing the decl kind to match the behavior of the
+  // attribute printing for variables and function where they are printed first.
+  if (prettyPrintAttributes(D, AttrPosAsWritten::Left))
     Out << ' ';
+
+  if (D->getIdentifier()) {
     if (auto *NNS = D->getQualifier())
       NNS->print(Out, Policy);
     Out << *D;
@@ -1087,16 +1095,13 @@ void DeclPrinter::VisitCXXRecordDecl(CXXRecordDecl *D) {
     }
   }
 
-  if (D->hasDefinition()) {
-    if (D->hasAttr<FinalAttr>()) {
-      Out << " final";
-    }
-  }
+  prettyPrintAttributes(D, AttrPosAsWritten::Right);
 
   if (D->isCompleteDefinition()) {
+    Out << ' ';
     // Print the base classes
     if (D->getNumBases()) {
-      Out << " : ";
+      Out << ": ";
       for (CXXRecordDecl::base_class_iterator Base = D->bases_begin(),
              BaseEnd = D->bases_end(); Base != BaseEnd; ++Base) {
         if (Base != D->bases_begin())
@@ -1115,14 +1120,15 @@ void DeclPrinter::VisitCXXRecordDecl(CXXRecordDecl *D) {
         if (Base->isPackExpansion())
           Out << "...";
       }
+      Out << ' ';
     }
 
     // Print the class definition
     // FIXME: Doesn't print access specifiers, e.g., "public:"
     if (Policy.TerseOutput) {
-      Out << " {}";
+      Out << "{}";
     } else {
-      Out << " {\n";
+      Out << "{\n";
       VisitDeclContext(D);
       Indent() << "}";
     }
diff --git a/clang/test/SemaCXX/cxx11-attr-print.cpp b/clang/test/SemaCXX/cxx11-attr-print.cpp
index a169d1b4409b4..2b084018bc066 100644
--- a/clang/test/SemaCXX/cxx11-attr-print.cpp
+++ b/clang/test/SemaCXX/cxx11-attr-print.cpp
@@ -87,3 +87,8 @@ template struct S<int>;
 
 // CHECK: using Small2 {{\[}}[gnu::mode(byte)]] = int;
 using Small2 [[gnu::mode(byte)]] = int;
+
+class FinalNonTemplate final {};
+// CHECK: class FinalNonTemplate final {
+template <typename T> class FinalTemplate final {};
+// CHECK: template <typename T> class FinalTemplate final {
diff --git a/clang/test/SemaHLSL/BuiltIns/RWBuffers.hlsl b/clang/test/SemaHLSL/BuiltIns/RWBuffers.hlsl
index 7e79ae3bf005f..b1a15c4319182 100644
--- a/clang/test/SemaHLSL/BuiltIns/RWBuffers.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/RWBuffers.hlsl
@@ -6,11 +6,11 @@ typedef vector<float, 3> float3;
 RWBuffer<float3> Buffer;
 
 // expected-error@+2 {{class template 'RWBuffer' requires template arguments}}
-// expected-note@*:* {{template declaration from hidden source: template <class element_type> class RWBuffer final}}
+// expected-note@*:* {{template declaration from hidden source: template <class element_type> class RWBuffer}}
 RWBuffer BufferErr1;
 
 // expected-error@+2 {{too few template arguments for class template 'RWBuffer'}}
-// expected-note@*:* {{template declaration from hidden source: template <class element_type> class RWBuffer final}}
+// expected-note@*:* {{template declaration from hidden source: template <class element_type> class RWBuffer}}
 RWBuffer<> BufferErr2;
 
 [numthreads(1,1,1)]
diff --git a/clang/unittests/AST/DeclPrinterTest.cpp b/clang/unittests/AST/DeclPrinterTest.cpp
index c24e442621c92..6945dff537cae 100644
--- a/clang/unittests/AST/DeclPrinterTest.cpp
+++ b/clang/unittests/AST/DeclPrinterTest.cpp
@@ -86,16 +86,13 @@ ::testing::AssertionResult PrintedDeclCXX11Matches(StringRef Code,
                             ExpectedPrinted, "input.cc");
 }
 
-::testing::AssertionResult PrintedDeclCXX11Matches(
-                                  StringRef Code,
-                                  const DeclarationMatcher &NodeMatch,
-                                  StringRef ExpectedPrinted) {
+::testing::AssertionResult
+PrintedDeclCXX11Matches(StringRef Code, const DeclarationMatcher &NodeMatch,
+                        StringRef ExpectedPrinted,
+                        PrintingPolicyAdjuster PolicyModifier = nullptr) {
   std::vector<std::string> Args(1, "-std=c++11");
-  return PrintedDeclMatches(Code,
-                            Args,
-                            NodeMatch,
-                            ExpectedPrinted,
-                            "input.cc");
+  return PrintedDeclMatches(Code, Args, NodeMatch, ExpectedPrinted, "input.cc",
+                            PolicyModifier);
 }
 
 ::testing::AssertionResult PrintedDeclCXX11nonMSCMatches(
@@ -1555,3 +1552,25 @@ TEST(DeclPrinter, VarDeclWithInitializer) {
       PrintedDeclCXX17Matches("void foo() {int arr[42]; for(int a : arr);}",
                               namedDecl(hasName("a")).bind("id"), "int a"));
 }
+
+TEST(DeclPrinter, TestTemplateFinal) {
+  // By default we should print 'final' keyword whether class is implicitly or
+  // explicitly marked final.
+  ASSERT_TRUE(PrintedDeclCXX11Matches(
+      "template<typename T>\n"
+      "class FinalTemplate final {};",
+      classTemplateDecl(hasName("FinalTemplate")).bind("id"),
+      "template <typename T> class FinalTemplate final {}"));
+}
+
+TEST(DeclPrinter, TestTemplateFinalWithPolishForDecl) {
+  // clangd relies on the 'final' keyword being printed when
+  // PolishForDeclaration is enabled, so make sure it is even if implicit attrs
+  // are disabled.
+  ASSERT_TRUE(PrintedDeclCXX11Matches(
+      "template<typename T>\n"
+      "class FinalTemplate final {};",
+      classTemplateDecl(hasName("FinalTemplate")).bind("id"),
+      "template <typename T> class FinalTemplate final {}",
+      [](PrintingPolicy &Policy) { Policy.PolishForDeclaration = true; }));
+}
diff --git a/clang/utils/TableGen/ClangAttrEmitter.cpp b/clang/utils/TableGen/ClangAttrEmitter.cpp
index 6c56f99f503df..765cbbf3b04bc 100644
--- a/clang/utils/TableGen/ClangAttrEmitter.cpp
+++ b/clang/utils/TableGen/ClangAttrEmitter.cpp
@@ -1608,7 +1608,7 @@ writePrettyPrintFunction(const Record &R,
       Prefix = "[";
       Suffix = "]";
     } else if (Variety == "Keyword") {
-      Prefix = " ";
+      Prefix = "";
       Suffix = "";
     } else if (Variety == "Pragma") {
       Prefix = "#pragma ";

From b5aff11aa118dabf134a1377dfd94b34e4dedbf7 Mon Sep 17 00:00:00 2001
From: Dmitrii Agibov <dmitrii.agibov@arm.com>
Date: Thu, 18 Apr 2024 07:33:51 +0100
Subject: [PATCH 833/886] [mlir][tosa] Add folding for TOSA ArgMax operator
 (#88871)

---
 mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td  |  1 +
 .../Dialect/Tosa/IR/TosaCanonicalizations.cpp | 13 ++++++++++
 mlir/test/Dialect/Tosa/constant-op-fold.mlir  | 26 ++++++++++++++-----
 3 files changed, 34 insertions(+), 6 deletions(-)

diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td b/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
index 306e4a4395208..dde17e2dc8924 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaOps.td
@@ -49,6 +49,7 @@ def Tosa_ArgMaxOp : Tosa_InferShapedTypeOp<"argmax"> {
     Tosa_Tensor: $output
   );
 
+  let hasFolder = 1;
   let hasVerifier = 1;
 }
 
diff --git a/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp b/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
index d23c9fe824c94..c8bf4c526b239 100644
--- a/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
@@ -507,6 +507,19 @@ OpFoldResult AddOp::fold(FoldAdaptor adaptor) {
                                                             resultTy);
 }
 
+OpFoldResult ArgMaxOp::fold(FoldAdaptor adaptor) {
+  auto inputTy = llvm::dyn_cast<RankedTensorType>(getInput().getType());
+  auto outputTy = llvm::dyn_cast<RankedTensorType>(getType());
+  if (!inputTy || !outputTy || !inputTy.hasStaticShape() ||
+      !outputTy.hasStaticShape())
+    return {};
+
+  if (inputTy.getDimSize(getAxis()) == 1)
+    return DenseElementsAttr::get(outputTy, 0);
+
+  return {};
+}
+
 OpFoldResult DivOp::fold(FoldAdaptor adaptor) {
   auto lhsTy = llvm::dyn_cast<RankedTensorType>(getInput1().getType());
   auto rhsTy = llvm::dyn_cast<RankedTensorType>(getInput2().getType());
diff --git a/mlir/test/Dialect/Tosa/constant-op-fold.mlir b/mlir/test/Dialect/Tosa/constant-op-fold.mlir
index de752f31fcbaa..c9c60a94bf9ed 100644
--- a/mlir/test/Dialect/Tosa/constant-op-fold.mlir
+++ b/mlir/test/Dialect/Tosa/constant-op-fold.mlir
@@ -3,6 +3,20 @@
 
 // RUN: mlir-opt --split-input-file --tosa-layerwise-constant-fold="aggressive-reduce-constant=true" %s | FileCheck %s --check-prefix=AGGRESIVE
 
+// CHECK-LABEL: @armax_fold_dim_size_1
+func.func @armax_fold_dim_size_1(%arg0: tensor<2x1x3xf32>) -> tensor<2x3xi32> {
+  // CHECK: "tosa.const"() <{value = dense<0> : tensor<2x3xi32>}> : () -> tensor<2x3xi32>
+  %0 = tosa.argmax %arg0 {axis = 1 : i32}: (tensor<2x1x3xf32>) -> tensor<2x3xi32>
+  return %0 : tensor<2x3xi32>
+}
+
+// CHECK-LABEL: @argmax_dynamic_shape_no_fold_dim_size_1
+func.func @argmax_dynamic_shape_no_fold_dim_size_1(%arg0: tensor<?x1x3xf32>) -> tensor<?x3xi32> {
+  // CHECK: tosa.argmax
+  %0 = tosa.argmax %arg0 {axis = 1 : i32}: (tensor<?x1x3xf32>) -> tensor<?x3xi32>
+  return %0 : tensor<?x3xi32>
+}
+
 // CHECK-LABEL: @transpose_fold
 func.func @transpose_fold(%arg0: tensor<3x4xf32>) -> tensor<3x4xf32> {
   // CHECK: return %arg0
@@ -1100,9 +1114,9 @@ func.func @reduce_sum_constant_aggressive() -> tensor<2x3xi32> {
   // AGGRESIVE-DAG:       %[[VAL_0:.*]] = "tosa.const"() <{value = dense<2> : tensor<1x2x3xi32>}> : () -> tensor<1x2x3xi32>
   // AGGRESIVE-DAG:       %[[VAL_1:.*]] = "tosa.const"() <{value = dense<1> : tensor<2x2x3xi32>}> : () -> tensor<2x2x3xi32>
   // AGGRESIVE-DAG:       %[[VAL_2:.*]] = "tosa.const"() <{value = dense<2> : tensor<2x3xi32>}> : () -> tensor<2x3xi32>
-  // AGGRESIVE:           %[[VAL_3:.*]] = tosa.argmax %[[VAL_0]] {axis = 0 : i32} : (tensor<1x2x3xi32>) -> tensor<2x3xi32>
+  // AGGRESIVE:           %[[VAL_3:.*]] = tosa.argmax %[[VAL_0]] {axis = 1 : i32} : (tensor<1x2x3xi32>) -> tensor<1x3xi32>
   // AGGRESIVE:           %[[VAL_4:.*]] = tosa.argmax %[[VAL_1]] {axis = 0 : i32} : (tensor<2x2x3xi32>) -> tensor<2x3xi32>
-  // AGGRESIVE:           %[[VAL_5:.*]] = tosa.add %[[VAL_3]], %[[VAL_2]] : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
+  // AGGRESIVE:           %[[VAL_5:.*]] = tosa.add %[[VAL_3]], %[[VAL_2]] : (tensor<1x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   // AGGRESIVE:           %[[VAL_6:.*]] = tosa.add %[[VAL_5]], %[[VAL_4]] : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   // AGGRESIVE:           return %[[VAL_6]] : tensor<2x3xi32>
 
@@ -1110,18 +1124,18 @@ func.func @reduce_sum_constant_aggressive() -> tensor<2x3xi32> {
   // CHECK:           %[[VAL_0:.*]] = "tosa.const"() <{value = dense<1> : tensor<2x2x3xi32>}> : () -> tensor<2x2x3xi32>
   // CHECK:           %[[VAL_1:.*]] = "tosa.const"() <{value = dense<2> : tensor<2x3xi32>}> : () -> tensor<2x3xi32>
   // CHECK:           %[[VAL_2:.*]] = tosa.reduce_sum %[[VAL_0]] {axis = 0 : i32} : (tensor<2x2x3xi32>) -> tensor<1x2x3xi32>
-  // CHECK:           %[[VAL_3:.*]] = tosa.argmax %[[VAL_2]] {axis = 0 : i32} : (tensor<1x2x3xi32>) -> tensor<2x3xi32>
+  // CHECK:           %[[VAL_3:.*]] = tosa.argmax %[[VAL_2]] {axis = 1 : i32} : (tensor<1x2x3xi32>) -> tensor<1x3xi32>
   // CHECK:           %[[VAL_4:.*]] = tosa.argmax %[[VAL_0]] {axis = 0 : i32} : (tensor<2x2x3xi32>) -> tensor<2x3xi32>
-  // CHECK:           %[[VAL_5:.*]] = tosa.add %[[VAL_3]], %[[VAL_1]] : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
+  // CHECK:           %[[VAL_5:.*]] = tosa.add %[[VAL_3]], %[[VAL_1]] : (tensor<1x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   // CHECK:           %[[VAL_6:.*]] = tosa.add %[[VAL_5]], %[[VAL_4]] : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   // CHECK:           return %[[VAL_6]] : tensor<2x3xi32>
 
   %const0 = "tosa.const"() {value = dense<1> : tensor<2x2x3xi32>} : () -> tensor<2x2x3xi32>
   %const1 = "tosa.const"() {value = dense<2> : tensor<2x3xi32>} : () -> tensor<2x3xi32>
   %reduce0 = tosa.reduce_sum %const0 {axis = 0 : i32} : (tensor<2x2x3xi32>) -> tensor<1x2x3xi32>
-  %argmax0 = tosa.argmax %reduce0 {axis = 0 : i32} : (tensor<1x2x3xi32>) -> tensor<2x3xi32>
+  %argmax0 = tosa.argmax %reduce0 {axis = 1 : i32} : (tensor<1x2x3xi32>) -> tensor<1x3xi32>
   %argmax1 = tosa.argmax %const0 {axis = 0 : i32} : (tensor<2x2x3xi32>) -> tensor<2x3xi32>
-  %res0 = tosa.add %argmax0, %const1 : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
+  %res0 = tosa.add %argmax0, %const1 : (tensor<1x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   %res1 = tosa.add %res0, %argmax1 : (tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
   return %res1 : tensor<2x3xi32>
 }

From 83cc60565a8a78ce00ee0c29f167400902e60737 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Thu, 18 Apr 2024 11:54:06 +0800
Subject: [PATCH 834/886] [NFC] [Serialization] Extract logics to write special
 decls from ASTWriter::WriteASTCore

ASTWriter::WriteASTCore is a big function and slightly hard to read.
This patch extracts the logics to force emitting special declarations
and writing special records for these declarations into member
functions.

This is helpful to improve the readability and also helpful for the
following patch to write special declarations lazily.
---
 clang/include/clang/Serialization/ASTWriter.h |   2 +
 clang/lib/Serialization/ASTWriter.cpp         | 360 ++++++++++--------
 2 files changed, 210 insertions(+), 152 deletions(-)

diff --git a/clang/include/clang/Serialization/ASTWriter.h b/clang/include/clang/Serialization/ASTWriter.h
index 892d7363e06dd..e093e6ad9d12c 100644
--- a/clang/include/clang/Serialization/ASTWriter.h
+++ b/clang/include/clang/Serialization/ASTWriter.h
@@ -559,6 +559,8 @@ class ASTWriter : public ASTDeserializationListener,
   void WriteIdentifierTable(Preprocessor &PP, IdentifierResolver &IdResolver,
                             bool IsModule);
   void WriteDeclAndTypes(ASTContext &Context);
+  void PrepareWritingSpecialDecls(Sema &SemaRef);
+  void WriteSpecialDeclRecords(Sema &SemaRef);
   void WriteDeclUpdatesBlocks(RecordDataImpl &OffsetsRecord);
   void WriteDeclContextVisibleUpdate(const DeclContext *DC);
   void WriteFPPragmaOptions(const FPOptionsOverride &Opts);
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index c3cf242391b74..ffeab4db9f6af 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -4733,6 +4733,14 @@ ASTFileSignature ASTWriter::WriteAST(Sema &SemaRef, StringRef OutputFile,
   return Signature;
 }
 
+template<typename Vector>
+static void AddLazyVectorDecls(ASTWriter &Writer, Vector &Vec) {
+  for (typename Vector::iterator I = Vec.begin(nullptr, true), E = Vec.end();
+       I != E; ++I) {
+    Writer.GetDeclRef(*I);
+  }
+}
+
 template<typename Vector>
 static void AddLazyVectorDecls(ASTWriter &Writer, Vector &Vec,
                                ASTWriter::RecordData &Record) {
@@ -4835,24 +4843,10 @@ void ASTWriter::computeNonAffectingInputFiles() {
   FileMgr.trackVFSUsage(false);
 }
 
-ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
-                                         Module *WritingModule) {
-  using namespace llvm;
-
-  bool isModule = WritingModule != nullptr;
-
-  // Make sure that the AST reader knows to finalize itself.
-  if (Chain)
-    Chain->finalizeForWriting();
-
+void ASTWriter::PrepareWritingSpecialDecls(Sema &SemaRef) {
   ASTContext &Context = SemaRef.Context;
-  Preprocessor &PP = SemaRef.PP;
-
-  // This needs to be done very early, since everything that writes
-  // SourceLocations or FileIDs depends on it.
-  computeNonAffectingInputFiles();
 
-  writeUnhashedControlBlock(PP, Context);
+  bool isModule = WritingModule != nullptr;
 
   // Set up predefined declaration IDs.
   auto RegisterPredefDecl = [&] (Decl *D, PredefinedDeclIDs ID) {
@@ -4888,43 +4882,144 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
   RegisterPredefDecl(Context.TypePackElementDecl,
                      PREDEF_DECL_TYPE_PACK_ELEMENT_ID);
 
-  // Build a record containing all of the tentative definitions in this file, in
+  // Writing all of the tentative definitions in this file, in
   // TentativeDefinitions order.  Generally, this record will be empty for
   // headers.
   RecordData TentativeDefinitions;
-  AddLazyVectorDecls(*this, SemaRef.TentativeDefinitions, TentativeDefinitions);
+  AddLazyVectorDecls(*this, SemaRef.TentativeDefinitions);
 
-  // Build a record containing all of the file scoped decls in this file.
-  RecordData UnusedFileScopedDecls;
+  // Writing all of the file scoped decls in this file.
   if (!isModule)
-    AddLazyVectorDecls(*this, SemaRef.UnusedFileScopedDecls,
-                       UnusedFileScopedDecls);
+    AddLazyVectorDecls(*this, SemaRef.UnusedFileScopedDecls);
 
-  // Build a record containing all of the delegating constructors we still need
+  // Writing all of the delegating constructors we still need
   // to resolve.
-  RecordData DelegatingCtorDecls;
   if (!isModule)
-    AddLazyVectorDecls(*this, SemaRef.DelegatingCtorDecls, DelegatingCtorDecls);
+    AddLazyVectorDecls(*this, SemaRef.DelegatingCtorDecls);
 
-  // Write the set of weak, undeclared identifiers. We always write the
-  // entire table, since later PCH files in a PCH chain are only interested in
-  // the results at the end of the chain.
-  RecordData WeakUndeclaredIdentifiers;
-  for (const auto &WeakUndeclaredIdentifierList :
-       SemaRef.WeakUndeclaredIdentifiers) {
-    const IdentifierInfo *const II = WeakUndeclaredIdentifierList.first;
-    for (const auto &WI : WeakUndeclaredIdentifierList.second) {
-      AddIdentifierRef(II, WeakUndeclaredIdentifiers);
-      AddIdentifierRef(WI.getAlias(), WeakUndeclaredIdentifiers);
-      AddSourceLocation(WI.getLocation(), WeakUndeclaredIdentifiers);
+  // Writing all of the ext_vector declarations.
+  AddLazyVectorDecls(*this, SemaRef.ExtVectorDecls);
+
+  // Writing all of the VTable uses information.
+  if (!SemaRef.VTableUses.empty())
+    for (unsigned I = 0, N = SemaRef.VTableUses.size(); I != N; ++I)
+      GetDeclRef(SemaRef.VTableUses[I].first);
+
+  // Writing all of the UnusedLocalTypedefNameCandidates.
+  for (const TypedefNameDecl *TD : SemaRef.UnusedLocalTypedefNameCandidates)
+    GetDeclRef(TD);
+
+  // Writing all of pending implicit instantiations.
+  for (const auto &I : SemaRef.PendingInstantiations)
+    GetDeclRef(I.first);
+  assert(SemaRef.PendingLocalImplicitInstantiations.empty() &&
+         "There are local ones at end of translation unit!");
+
+  // Writing some declaration references.
+  if (SemaRef.StdNamespace || SemaRef.StdBadAlloc || SemaRef.StdAlignValT) {
+    GetDeclRef(SemaRef.getStdNamespace());
+    GetDeclRef(SemaRef.getStdBadAlloc());
+    GetDeclRef(SemaRef.getStdAlignValT());
+  }
+
+  if (Context.getcudaConfigureCallDecl())
+    GetDeclRef(Context.getcudaConfigureCallDecl());
+
+  // Writing all of the known namespaces.
+  for (const auto &I : SemaRef.KnownNamespaces)
+    if (!I.second)
+      GetDeclRef(I.first);
+
+  // Writing all used, undefined objects that require definitions.
+  SmallVector<std::pair<NamedDecl *, SourceLocation>, 16> Undefined;
+  SemaRef.getUndefinedButUsed(Undefined);
+  for (const auto &I : Undefined)
+    GetDeclRef(I.first);
+
+  // Writing all delete-expressions that we would like to
+  // analyze later in AST.
+  if (!isModule)
+    for (const auto &DeleteExprsInfo :
+         SemaRef.getMismatchingDeleteExpressions())
+      GetDeclRef(DeleteExprsInfo.first);
+
+  const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
+
+  // Force all top level declarations to be emitted.
+  //
+  // We start emitting top level declarations from the module purview to
+  // implement the eliding unreachable declaration feature.
+  for (const auto *D : TU->noload_decls()) {
+    if (D->isFromASTFile())
+      continue;
+
+    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
+      continue;
+
+    GetDeclRef(D);
+  }
+
+  // Make sure visible decls, added to DeclContexts previously loaded from
+  // an AST file, are registered for serialization. Likewise for template
+  // specializations added to imported templates.
+  for (const auto *I : DeclsToEmitEvenIfUnreferenced)
+    GetDeclRef(I);
+  DeclsToEmitEvenIfUnreferenced.clear();
+
+  // Make sure all decls associated with an identifier are registered for
+  // serialization, if we're storing decls with identifiers.
+  if (!WritingModule || !getLangOpts().CPlusPlus) {
+    llvm::SmallVector<const IdentifierInfo*, 256> IIs;
+    for (const auto &ID : SemaRef.PP.getIdentifierTable()) {
+      const IdentifierInfo *II = ID.second;
+      if (!Chain || !II->isFromAST() || II->hasChangedSinceDeserialization())
+        IIs.push_back(II);
     }
+    // Sort the identifiers to visit based on their name.
+    llvm::sort(IIs, llvm::deref<std::less<>>());
+    for (const IdentifierInfo *II : IIs)
+      for (const Decl *D : SemaRef.IdResolver.decls(II))
+        GetDeclRef(D);
   }
 
-  // Build a record containing all of the ext_vector declarations.
+  // Write all of the DeclsToCheckForDeferredDiags.
+  for (auto *D : SemaRef.DeclsToCheckForDeferredDiags)
+    GetDeclRef(D);
+}
+
+void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
+  ASTContext &Context = SemaRef.Context;
+
+  bool isModule = WritingModule != nullptr;
+
+  // Write the record containing external, unnamed definitions.
+  if (!EagerlyDeserializedDecls.empty())
+    Stream.EmitRecord(EAGERLY_DESERIALIZED_DECLS, EagerlyDeserializedDecls);
+
+  if (!ModularCodegenDecls.empty())
+    Stream.EmitRecord(MODULAR_CODEGEN_DECLS, ModularCodegenDecls);
+  
+  // Write the record containing tentative definitions.
+  RecordData TentativeDefinitions;
+  AddLazyVectorDecls(*this, SemaRef.TentativeDefinitions, TentativeDefinitions);
+  if (!TentativeDefinitions.empty())
+    Stream.EmitRecord(TENTATIVE_DEFINITIONS, TentativeDefinitions);
+
+  // Write the record containing unused file scoped decls.
+  RecordData UnusedFileScopedDecls;
+  if (!isModule)
+    AddLazyVectorDecls(*this, SemaRef.UnusedFileScopedDecls,
+                       UnusedFileScopedDecls);
+  if (!UnusedFileScopedDecls.empty())
+    Stream.EmitRecord(UNUSED_FILESCOPED_DECLS, UnusedFileScopedDecls);
+
+  // Write the record containing ext_vector type names.
   RecordData ExtVectorDecls;
   AddLazyVectorDecls(*this, SemaRef.ExtVectorDecls, ExtVectorDecls);
+  if (!ExtVectorDecls.empty())
+    Stream.EmitRecord(EXT_VECTOR_DECLS, ExtVectorDecls);
 
-  // Build a record containing all of the VTable uses information.
+  // Write the record containing VTable uses information.
   RecordData VTableUses;
   if (!SemaRef.VTableUses.empty()) {
     for (unsigned I = 0, N = SemaRef.VTableUses.size(); I != N; ++I) {
@@ -4932,56 +5027,81 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
       AddSourceLocation(SemaRef.VTableUses[I].second, VTableUses);
       VTableUses.push_back(SemaRef.VTablesUsed[SemaRef.VTableUses[I].first]);
     }
+    Stream.EmitRecord(VTABLE_USES, VTableUses);
   }
 
-  // Build a record containing all of the UnusedLocalTypedefNameCandidates.
+  // Write the record containing potentially unused local typedefs.
   RecordData UnusedLocalTypedefNameCandidates;
   for (const TypedefNameDecl *TD : SemaRef.UnusedLocalTypedefNameCandidates)
     AddDeclRef(TD, UnusedLocalTypedefNameCandidates);
+  if (!UnusedLocalTypedefNameCandidates.empty())
+    Stream.EmitRecord(UNUSED_LOCAL_TYPEDEF_NAME_CANDIDATES,
+                      UnusedLocalTypedefNameCandidates);
 
-  // Build a record containing all of pending implicit instantiations.
+  // Write the record containing pending implicit instantiations.
   RecordData PendingInstantiations;
   for (const auto &I : SemaRef.PendingInstantiations) {
     AddDeclRef(I.first, PendingInstantiations);
     AddSourceLocation(I.second, PendingInstantiations);
   }
-  assert(SemaRef.PendingLocalImplicitInstantiations.empty() &&
-         "There are local ones at end of translation unit!");
+  if (!PendingInstantiations.empty())
+    Stream.EmitRecord(PENDING_IMPLICIT_INSTANTIATIONS, PendingInstantiations);
 
-  // Build a record containing some declaration references.
+  // Write the record containing declaration references of Sema.
   RecordData SemaDeclRefs;
   if (SemaRef.StdNamespace || SemaRef.StdBadAlloc || SemaRef.StdAlignValT) {
     AddDeclRef(SemaRef.getStdNamespace(), SemaDeclRefs);
     AddDeclRef(SemaRef.getStdBadAlloc(), SemaDeclRefs);
     AddDeclRef(SemaRef.getStdAlignValT(), SemaDeclRefs);
   }
+  if (!SemaDeclRefs.empty())
+    Stream.EmitRecord(SEMA_DECL_REFS, SemaDeclRefs);
+
+  // Write the record containing decls to be checked for deferred diags.
+  SmallVector<serialization::DeclID, 64> DeclsToCheckForDeferredDiags;
+  for (auto *D : SemaRef.DeclsToCheckForDeferredDiags)
+    DeclsToCheckForDeferredDiags.push_back(GetDeclRef(D));
+  if (!DeclsToCheckForDeferredDiags.empty())
+    Stream.EmitRecord(DECLS_TO_CHECK_FOR_DEFERRED_DIAGS,
+        DeclsToCheckForDeferredDiags);
 
+  // Write the record containing CUDA-specific declaration references.
   RecordData CUDASpecialDeclRefs;
   if (Context.getcudaConfigureCallDecl()) {
     AddDeclRef(Context.getcudaConfigureCallDecl(), CUDASpecialDeclRefs);
+    Stream.EmitRecord(CUDA_SPECIAL_DECL_REFS, CUDASpecialDeclRefs);
   }
 
-  // Build a record containing all of the known namespaces.
+  // Write the delegating constructors.
+  RecordData DelegatingCtorDecls;
+  if (!isModule)
+    AddLazyVectorDecls(*this, SemaRef.DelegatingCtorDecls, DelegatingCtorDecls);
+  if (!DelegatingCtorDecls.empty())
+    Stream.EmitRecord(DELEGATING_CTORS, DelegatingCtorDecls);
+
+  // Write the known namespaces.
   RecordData KnownNamespaces;
   for (const auto &I : SemaRef.KnownNamespaces) {
     if (!I.second)
       AddDeclRef(I.first, KnownNamespaces);
   }
+  if (!KnownNamespaces.empty())
+    Stream.EmitRecord(KNOWN_NAMESPACES, KnownNamespaces);
 
-  // Build a record of all used, undefined objects that require definitions.
+  // Write the undefined internal functions and variables, and inline functions.
   RecordData UndefinedButUsed;
-
   SmallVector<std::pair<NamedDecl *, SourceLocation>, 16> Undefined;
   SemaRef.getUndefinedButUsed(Undefined);
   for (const auto &I : Undefined) {
     AddDeclRef(I.first, UndefinedButUsed);
     AddSourceLocation(I.second, UndefinedButUsed);
   }
-
-  // Build a record containing all delete-expressions that we would like to
+  if (!UndefinedButUsed.empty())
+    Stream.EmitRecord(UNDEFINED_BUT_USED, UndefinedButUsed);
+  
+  // Write all delete-expressions that we would like to
   // analyze later in AST.
   RecordData DeleteExprsToAnalyze;
-
   if (!isModule) {
     for (const auto &DeleteExprsInfo :
          SemaRef.getMismatchingDeleteExpressions()) {
@@ -4993,6 +5113,44 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
       }
     }
   }
+  if (!DeleteExprsToAnalyze.empty())
+    Stream.EmitRecord(DELETE_EXPRS_TO_ANALYZE, DeleteExprsToAnalyze);
+}
+
+ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
+                                         Module *WritingModule) {
+  using namespace llvm;
+
+  bool isModule = WritingModule != nullptr;
+
+  // Make sure that the AST reader knows to finalize itself.
+  if (Chain)
+    Chain->finalizeForWriting();
+
+  ASTContext &Context = SemaRef.Context;
+  Preprocessor &PP = SemaRef.PP;
+
+  // This needs to be done very early, since everything that writes
+  // SourceLocations or FileIDs depends on it.
+  computeNonAffectingInputFiles();
+
+  writeUnhashedControlBlock(PP, Context);
+
+  // Write the set of weak, undeclared identifiers. We always write the
+  // entire table, since later PCH files in a PCH chain are only interested in
+  // the results at the end of the chain.
+  RecordData WeakUndeclaredIdentifiers;
+  for (const auto &WeakUndeclaredIdentifierList :
+       SemaRef.WeakUndeclaredIdentifiers) {
+    const IdentifierInfo *const II = WeakUndeclaredIdentifierList.first;
+    for (const auto &WI : WeakUndeclaredIdentifierList.second) {
+      AddIdentifierRef(II, WeakUndeclaredIdentifiers);
+      AddIdentifierRef(WI.getAlias(), WeakUndeclaredIdentifiers);
+      AddSourceLocation(WI.getLocation(), WeakUndeclaredIdentifiers);
+    }
+  }
+
+  PrepareWritingSpecialDecls(SemaRef);
 
   // Write the control block
   WriteControlBlock(PP, Context, isysroot);
@@ -5010,45 +5168,6 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
     Stream.EmitRecord(METADATA_OLD_FORMAT, Record);
   }
 
-  const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
-
-  // Force all top level declarations to be emitted.
-  //
-  // We start emitting top level declarations from the module purview to
-  // implement the eliding unreachable declaration feature.
-  for (const auto *D : TU->noload_decls()) {
-    if (D->isFromASTFile())
-      continue;
-
-    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
-      continue;
-
-    GetDeclRef(D);
-  }
-
-  // Make sure visible decls, added to DeclContexts previously loaded from
-  // an AST file, are registered for serialization. Likewise for template
-  // specializations added to imported templates.
-  for (const auto *I : DeclsToEmitEvenIfUnreferenced) {
-    GetDeclRef(I);
-  }
-
-  // Make sure all decls associated with an identifier are registered for
-  // serialization, if we're storing decls with identifiers.
-  if (!WritingModule || !getLangOpts().CPlusPlus) {
-    llvm::SmallVector<const IdentifierInfo*, 256> IIs;
-    for (const auto &ID : PP.getIdentifierTable()) {
-      const IdentifierInfo *II = ID.second;
-      if (!Chain || !II->isFromAST() || II->hasChangedSinceDeserialization())
-        IIs.push_back(II);
-    }
-    // Sort the identifiers to visit based on their name.
-    llvm::sort(IIs, llvm::deref<std::less<>>());
-    for (const IdentifierInfo *II : IIs)
-      for (const Decl *D : SemaRef.IdResolver.decls(II))
-        GetDeclRef(D);
-  }
-
   // For method pool in the module, if it contains an entry for a selector,
   // the entry should be complete, containing everything introduced by that
   // module and all modules it imports. It's possible that the entry is out of
@@ -5139,11 +5258,6 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
                               Buffer.data(), Buffer.size());
   }
 
-  // Build a record containing all of the DeclsToCheckForDeferredDiags.
-  SmallVector<serialization::DeclID, 64> DeclsToCheckForDeferredDiags;
-  for (auto *D : SemaRef.DeclsToCheckForDeferredDiags)
-    DeclsToCheckForDeferredDiags.push_back(GetDeclRef(D));
-
   WriteDeclAndTypes(Context);
 
   WriteFileDeclIDsMap();
@@ -5165,71 +5279,13 @@ ASTFileSignature ASTWriter::WriteASTCore(Sema &SemaRef, StringRef isysroot,
 
   Stream.EmitRecord(SPECIAL_TYPES, SpecialTypes);
 
-  // Write the record containing external, unnamed definitions.
-  if (!EagerlyDeserializedDecls.empty())
-    Stream.EmitRecord(EAGERLY_DESERIALIZED_DECLS, EagerlyDeserializedDecls);
-
-  if (!ModularCodegenDecls.empty())
-    Stream.EmitRecord(MODULAR_CODEGEN_DECLS, ModularCodegenDecls);
-
-  // Write the record containing tentative definitions.
-  if (!TentativeDefinitions.empty())
-    Stream.EmitRecord(TENTATIVE_DEFINITIONS, TentativeDefinitions);
-
-  // Write the record containing unused file scoped decls.
-  if (!UnusedFileScopedDecls.empty())
-    Stream.EmitRecord(UNUSED_FILESCOPED_DECLS, UnusedFileScopedDecls);
+  WriteSpecialDeclRecords(SemaRef);
 
   // Write the record containing weak undeclared identifiers.
   if (!WeakUndeclaredIdentifiers.empty())
     Stream.EmitRecord(WEAK_UNDECLARED_IDENTIFIERS,
                       WeakUndeclaredIdentifiers);
 
-  // Write the record containing ext_vector type names.
-  if (!ExtVectorDecls.empty())
-    Stream.EmitRecord(EXT_VECTOR_DECLS, ExtVectorDecls);
-
-  // Write the record containing VTable uses information.
-  if (!VTableUses.empty())
-    Stream.EmitRecord(VTABLE_USES, VTableUses);
-
-  // Write the record containing potentially unused local typedefs.
-  if (!UnusedLocalTypedefNameCandidates.empty())
-    Stream.EmitRecord(UNUSED_LOCAL_TYPEDEF_NAME_CANDIDATES,
-                      UnusedLocalTypedefNameCandidates);
-
-  // Write the record containing pending implicit instantiations.
-  if (!PendingInstantiations.empty())
-    Stream.EmitRecord(PENDING_IMPLICIT_INSTANTIATIONS, PendingInstantiations);
-
-  // Write the record containing declaration references of Sema.
-  if (!SemaDeclRefs.empty())
-    Stream.EmitRecord(SEMA_DECL_REFS, SemaDeclRefs);
-
-  // Write the record containing decls to be checked for deferred diags.
-  if (!DeclsToCheckForDeferredDiags.empty())
-    Stream.EmitRecord(DECLS_TO_CHECK_FOR_DEFERRED_DIAGS,
-        DeclsToCheckForDeferredDiags);
-
-  // Write the record containing CUDA-specific declaration references.
-  if (!CUDASpecialDeclRefs.empty())
-    Stream.EmitRecord(CUDA_SPECIAL_DECL_REFS, CUDASpecialDeclRefs);
-
-  // Write the delegating constructors.
-  if (!DelegatingCtorDecls.empty())
-    Stream.EmitRecord(DELEGATING_CTORS, DelegatingCtorDecls);
-
-  // Write the known namespaces.
-  if (!KnownNamespaces.empty())
-    Stream.EmitRecord(KNOWN_NAMESPACES, KnownNamespaces);
-
-  // Write the undefined internal functions and variables, and inline functions.
-  if (!UndefinedButUsed.empty())
-    Stream.EmitRecord(UNDEFINED_BUT_USED, UndefinedButUsed);
-
-  if (!DeleteExprsToAnalyze.empty())
-    Stream.EmitRecord(DELETE_EXPRS_TO_ANALYZE, DeleteExprsToAnalyze);
-
   if (!WritingModule) {
     // Write the submodules that were imported, if any.
     struct ModuleInfo {

From 7ec342ba16ca97206aa2ddd5b7ec0da063042e03 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Thu, 18 Apr 2024 14:04:48 +0800
Subject: [PATCH 835/886] [C++20] [Modules] [Reduced BMI] Write Special Decl
 Lazily

ASTWrite would write some special records for the consumer of the AST
can get the information easily. This is fine. But currently all the
special records are written eagerly, which is conflicting with reduced
BMI.

In reduced BMI, we hope to write things as less as possible. It is not a
goal for reduced BMI to retain all the informations. So in this patch we
won't record the special declarations eagerly before we starting write.
But only writing the sepcial records after we writes decls and types,
then only the reached declarations will be collected.
---
 clang/include/clang/Serialization/ASTWriter.h |   2 +
 clang/lib/Serialization/ASTWriter.cpp         | 104 ++++++++++++------
 .../reduced-bmi-empty-module-purview-std.cppm |  27 +++++
 3 files changed, 98 insertions(+), 35 deletions(-)
 create mode 100644 clang/test/Modules/reduced-bmi-empty-module-purview-std.cppm

diff --git a/clang/include/clang/Serialization/ASTWriter.h b/clang/include/clang/Serialization/ASTWriter.h
index e093e6ad9d12c..13b4ad4ad2953 100644
--- a/clang/include/clang/Serialization/ASTWriter.h
+++ b/clang/include/clang/Serialization/ASTWriter.h
@@ -714,6 +714,8 @@ class ASTWriter : public ASTDeserializationListener,
 
   /// Emit a reference to a declaration.
   void AddDeclRef(const Decl *D, RecordDataImpl &Record);
+  // Emit a reference to a declaration if the declaration was emitted.
+  void AddEmittedDeclRef(const Decl *D, RecordDataImpl &Record);
 
   /// Force a declaration to be emitted and get its ID.
   serialization::DeclID GetDeclRef(const Decl *D);
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index ffeab4db9f6af..9ae9648f38a3b 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -4741,12 +4741,12 @@ static void AddLazyVectorDecls(ASTWriter &Writer, Vector &Vec) {
   }
 }
 
-template<typename Vector>
-static void AddLazyVectorDecls(ASTWriter &Writer, Vector &Vec,
-                               ASTWriter::RecordData &Record) {
+template <typename Vector>
+static void AddLazyVectorEmiitedDecls(ASTWriter &Writer, Vector &Vec,
+                                      ASTWriter::RecordData &Record) {
   for (typename Vector::iterator I = Vec.begin(nullptr, true), E = Vec.end();
        I != E; ++I) {
-    Writer.AddDeclRef(*I, Record);
+    Writer.AddEmittedDeclRef(*I, Record);
   }
 }
 
@@ -4882,6 +4882,25 @@ void ASTWriter::PrepareWritingSpecialDecls(Sema &SemaRef) {
   RegisterPredefDecl(Context.TypePackElementDecl,
                      PREDEF_DECL_TYPE_PACK_ELEMENT_ID);
 
+  const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
+
+  // Force all top level declarations to be emitted.
+  //
+  // We start emitting top level declarations from the module purview to
+  // implement the eliding unreachable declaration feature.
+  for (const auto *D : TU->noload_decls()) {
+    if (D->isFromASTFile())
+      continue;
+
+    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
+      continue;
+
+    GetDeclRef(D);
+  }
+
+  if (GeneratingReducedBMI)
+    return;
+
   // Writing all of the tentative definitions in this file, in
   // TentativeDefinitions order.  Generally, this record will be empty for
   // headers.
@@ -4943,22 +4962,6 @@ void ASTWriter::PrepareWritingSpecialDecls(Sema &SemaRef) {
          SemaRef.getMismatchingDeleteExpressions())
       GetDeclRef(DeleteExprsInfo.first);
 
-  const TranslationUnitDecl *TU = Context.getTranslationUnitDecl();
-
-  // Force all top level declarations to be emitted.
-  //
-  // We start emitting top level declarations from the module purview to
-  // implement the eliding unreachable declaration feature.
-  for (const auto *D : TU->noload_decls()) {
-    if (D->isFromASTFile())
-      continue;
-
-    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
-      continue;
-
-    GetDeclRef(D);
-  }
-
   // Make sure visible decls, added to DeclContexts previously loaded from
   // an AST file, are registered for serialization. Likewise for template
   // specializations added to imported templates.
@@ -5001,21 +5004,22 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   
   // Write the record containing tentative definitions.
   RecordData TentativeDefinitions;
-  AddLazyVectorDecls(*this, SemaRef.TentativeDefinitions, TentativeDefinitions);
+  AddLazyVectorEmiitedDecls(*this, SemaRef.TentativeDefinitions,
+                            TentativeDefinitions);
   if (!TentativeDefinitions.empty())
     Stream.EmitRecord(TENTATIVE_DEFINITIONS, TentativeDefinitions);
 
   // Write the record containing unused file scoped decls.
   RecordData UnusedFileScopedDecls;
   if (!isModule)
-    AddLazyVectorDecls(*this, SemaRef.UnusedFileScopedDecls,
-                       UnusedFileScopedDecls);
+    AddLazyVectorEmiitedDecls(*this, SemaRef.UnusedFileScopedDecls,
+                              UnusedFileScopedDecls);
   if (!UnusedFileScopedDecls.empty())
     Stream.EmitRecord(UNUSED_FILESCOPED_DECLS, UnusedFileScopedDecls);
 
   // Write the record containing ext_vector type names.
   RecordData ExtVectorDecls;
-  AddLazyVectorDecls(*this, SemaRef.ExtVectorDecls, ExtVectorDecls);
+  AddLazyVectorEmiitedDecls(*this, SemaRef.ExtVectorDecls, ExtVectorDecls);
   if (!ExtVectorDecls.empty())
     Stream.EmitRecord(EXT_VECTOR_DECLS, ExtVectorDecls);
 
@@ -5023,9 +5027,13 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   RecordData VTableUses;
   if (!SemaRef.VTableUses.empty()) {
     for (unsigned I = 0, N = SemaRef.VTableUses.size(); I != N; ++I) {
-      AddDeclRef(SemaRef.VTableUses[I].first, VTableUses);
+      CXXRecordDecl *D = SemaRef.VTableUses[I].first;
+      if (!wasDeclEmitted(D))
+        continue;
+
+      AddDeclRef(D, VTableUses);
       AddSourceLocation(SemaRef.VTableUses[I].second, VTableUses);
-      VTableUses.push_back(SemaRef.VTablesUsed[SemaRef.VTableUses[I].first]);
+      VTableUses.push_back(SemaRef.VTablesUsed[D]);
     }
     Stream.EmitRecord(VTABLE_USES, VTableUses);
   }
@@ -5033,7 +5041,7 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   // Write the record containing potentially unused local typedefs.
   RecordData UnusedLocalTypedefNameCandidates;
   for (const TypedefNameDecl *TD : SemaRef.UnusedLocalTypedefNameCandidates)
-    AddDeclRef(TD, UnusedLocalTypedefNameCandidates);
+    AddEmittedDeclRef(TD, UnusedLocalTypedefNameCandidates);
   if (!UnusedLocalTypedefNameCandidates.empty())
     Stream.EmitRecord(UNUSED_LOCAL_TYPEDEF_NAME_CANDIDATES,
                       UnusedLocalTypedefNameCandidates);
@@ -5041,6 +5049,9 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   // Write the record containing pending implicit instantiations.
   RecordData PendingInstantiations;
   for (const auto &I : SemaRef.PendingInstantiations) {
+    if (!wasDeclEmitted(I.first))
+      continue;
+
     AddDeclRef(I.first, PendingInstantiations);
     AddSourceLocation(I.second, PendingInstantiations);
   }
@@ -5050,9 +5061,16 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   // Write the record containing declaration references of Sema.
   RecordData SemaDeclRefs;
   if (SemaRef.StdNamespace || SemaRef.StdBadAlloc || SemaRef.StdAlignValT) {
-    AddDeclRef(SemaRef.getStdNamespace(), SemaDeclRefs);
-    AddDeclRef(SemaRef.getStdBadAlloc(), SemaDeclRefs);
-    AddDeclRef(SemaRef.getStdAlignValT(), SemaDeclRefs);
+    auto AddEmittedDeclRefOrZero = [this, &SemaDeclRefs, &SemaRef](Decl *D) {
+      if (!D || !wasDeclEmitted(D))
+        SemaDeclRefs.push_back(0);
+      else
+        SemaDeclRefs.push_back(getDeclID(D));
+    };
+
+    AddEmittedDeclRefOrZero(SemaRef.getStdNamespace());
+    AddEmittedDeclRefOrZero(SemaRef.getStdBadAlloc());
+    AddEmittedDeclRefOrZero(SemaRef.getStdAlignValT());
   }
   if (!SemaDeclRefs.empty())
     Stream.EmitRecord(SEMA_DECL_REFS, SemaDeclRefs);
@@ -5060,29 +5078,32 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   // Write the record containing decls to be checked for deferred diags.
   SmallVector<serialization::DeclID, 64> DeclsToCheckForDeferredDiags;
   for (auto *D : SemaRef.DeclsToCheckForDeferredDiags)
-    DeclsToCheckForDeferredDiags.push_back(GetDeclRef(D));
+    if (wasDeclEmitted(D))
+      DeclsToCheckForDeferredDiags.push_back(getDeclID(D));
   if (!DeclsToCheckForDeferredDiags.empty())
     Stream.EmitRecord(DECLS_TO_CHECK_FOR_DEFERRED_DIAGS,
         DeclsToCheckForDeferredDiags);
 
   // Write the record containing CUDA-specific declaration references.
   RecordData CUDASpecialDeclRefs;
-  if (Context.getcudaConfigureCallDecl()) {
-    AddDeclRef(Context.getcudaConfigureCallDecl(), CUDASpecialDeclRefs);
+  if (auto *CudaCallDecl = Context.getcudaConfigureCallDecl();
+      CudaCallDecl && wasDeclEmitted(CudaCallDecl)) {
+    AddDeclRef(CudaCallDecl, CUDASpecialDeclRefs);
     Stream.EmitRecord(CUDA_SPECIAL_DECL_REFS, CUDASpecialDeclRefs);
   }
 
   // Write the delegating constructors.
   RecordData DelegatingCtorDecls;
   if (!isModule)
-    AddLazyVectorDecls(*this, SemaRef.DelegatingCtorDecls, DelegatingCtorDecls);
+    AddLazyVectorEmiitedDecls(*this, SemaRef.DelegatingCtorDecls,
+                              DelegatingCtorDecls);
   if (!DelegatingCtorDecls.empty())
     Stream.EmitRecord(DELEGATING_CTORS, DelegatingCtorDecls);
 
   // Write the known namespaces.
   RecordData KnownNamespaces;
   for (const auto &I : SemaRef.KnownNamespaces) {
-    if (!I.second)
+    if (!I.second && wasDeclEmitted(I.first))
       AddDeclRef(I.first, KnownNamespaces);
   }
   if (!KnownNamespaces.empty())
@@ -5093,6 +5114,9 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   SmallVector<std::pair<NamedDecl *, SourceLocation>, 16> Undefined;
   SemaRef.getUndefinedButUsed(Undefined);
   for (const auto &I : Undefined) {
+    if (!wasDeclEmitted(I.first))
+      continue;
+
     AddDeclRef(I.first, UndefinedButUsed);
     AddSourceLocation(I.second, UndefinedButUsed);
   }
@@ -5105,6 +5129,9 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   if (!isModule) {
     for (const auto &DeleteExprsInfo :
          SemaRef.getMismatchingDeleteExpressions()) {
+      if (!wasDeclEmitted(DeleteExprsInfo.first))
+        continue;
+
       AddDeclRef(DeleteExprsInfo.first, DeleteExprsToAnalyze);
       DeleteExprsToAnalyze.push_back(DeleteExprsInfo.second.size());
       for (const auto &DeleteLoc : DeleteExprsInfo.second) {
@@ -5919,6 +5946,13 @@ TypeID ASTWriter::getTypeID(QualType T) const {
   });
 }
 
+void ASTWriter::AddEmittedDeclRef(const Decl *D, RecordDataImpl &Record) {
+  if (!wasDeclEmitted(D))
+    return;
+
+  Record.push_back(GetDeclRef(D));
+}
+
 void ASTWriter::AddDeclRef(const Decl *D, RecordDataImpl &Record) {
   Record.push_back(GetDeclRef(D));
 }
diff --git a/clang/test/Modules/reduced-bmi-empty-module-purview-std.cppm b/clang/test/Modules/reduced-bmi-empty-module-purview-std.cppm
new file mode 100644
index 0000000000000..3146fda3555fd
--- /dev/null
+++ b/clang/test/Modules/reduced-bmi-empty-module-purview-std.cppm
@@ -0,0 +1,27 @@
+// Test that we won't write additional information from std namespace by default
+// into the Reduced BMI if the module purview is empty.
+//
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 %t/A.cppm -emit-reduced-module-interface -o %t/A.pcm
+// RUN: llvm-bcanalyzer --dump --disable-histogram --show-binary-blobs %t/A.pcm > %t/A.dump
+// RUN: cat %t/A.dump | FileCheck %t/A.cppm
+
+//--- std.h
+namespace std {
+  typedef decltype(sizeof(0)) size_t;
+  enum class align_val_t : std::size_t {};
+
+  class bad_alloc { };
+}
+
+//--- A.cppm
+module;
+#include "std.h"
+export module A;
+
+// CHECK-NOT: <DECL_NAMESPACE
+// CHECK-NOT: <DECL_CONTEXT_LEXICAL
+// CHECK-NOT: <DELAYED_NAMESPACE_LEXICAL_VISIBLE_RECORD

From 1baa3850656382d1d549a13f8a716ef5dc886eb8 Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 15:44:12 +0900
Subject: [PATCH 836/886] [IR][PatternMatch] Only accept poison in
 getSplatValue() (#89159)

In #88217 a large set of matchers was changed to only accept poison
values in splats, but not undef values. This is because we now use
poison for non-demanded vector elements, and allowing undef can cause
correctness issues.

This patch covers the remaining matchers by changing the AllowUndef
parameter of getSplatValue() to AllowPoison instead. We also carry out
corresponding renames in matchers.

As a followup, we may want to change the default for things like m_APInt
to m_APIntAllowPoison (as this is much less risky when only allowing
poison), but this change doesn't do that.

There is one caveat here: We have a single place
(X86FixupVectorConstants) which does require handling of vector splats
with undefs. This is because this works on backend constant pool
entries, which currently still use undef instead of poison for
non-demanded elements (because SDAG as a whole does not have an explicit
poison representation). As it's just the single use, I've open-coded a
getSplatValueAllowUndef() helper there, to discourage use in any other
places.
---
 llvm/include/llvm/IR/Constant.h               |  6 +-
 llvm/include/llvm/IR/Constants.h              |  6 +-
 llvm/include/llvm/IR/PatternMatch.h           | 66 +++++++-------
 llvm/lib/Analysis/CmpInstAnalysis.cpp         |  2 +-
 llvm/lib/Analysis/InstructionSimplify.cpp     | 14 +--
 llvm/lib/Analysis/ValueTracking.cpp           |  4 +-
 llvm/lib/IR/Constants.cpp                     | 14 +--
 llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp     |  6 +-
 .../Target/X86/X86FixupVectorConstants.cpp    | 19 ++++-
 .../InstCombine/InstCombineAddSub.cpp         |  4 +-
 .../InstCombine/InstCombineAndOrXor.cpp       | 22 ++---
 .../InstCombine/InstCombineCalls.cpp          |  8 +-
 .../InstCombine/InstCombineCasts.cpp          |  2 +-
 .../InstCombine/InstCombineCompares.cpp       | 37 ++++----
 .../InstCombine/InstCombineInternal.h         |  4 +-
 .../InstCombine/InstCombineMulDivRem.cpp      |  6 +-
 .../InstCombine/InstCombineNegator.cpp        |  2 +-
 .../InstCombine/InstCombineSelect.cpp         | 16 ++--
 .../InstCombine/InstCombineShifts.cpp         |  8 +-
 .../InstCombine/InstructionCombining.cpp      |  4 +-
 .../Transforms/Vectorize/VPlanPatternMatch.h  |  2 +-
 .../InstCombine/and-or-icmp-const-icmp.ll     | 14 +--
 .../Transforms/InstCombine/binop-itofp.ll     | 21 +++--
 .../test/Transforms/InstCombine/bswap-fold.ll |  8 +-
 llvm/test/Transforms/InstCombine/bswap.ll     | 46 +++++-----
 .../Transforms/InstCombine/compare-signs.ll   | 14 +--
 .../InstCombine/fcmp-range-check-idiom.ll     |  8 +-
 llvm/test/Transforms/InstCombine/icmp-fsh.ll  | 32 +++----
 .../icmp-power2-and-icmp-shifted-mask.ll      | 10 ++-
 .../InstCombine/icmp-vec-inseltpoison.ll      | 20 ++---
 llvm/test/Transforms/InstCombine/icmp-vec.ll  | 50 +++++------
 .../Transforms/InstCombine/low-bit-splat.ll   | 23 ++---
 .../lshr-trunc-sext-to-ashr-sext.ll           | 34 ++++----
 llvm/test/Transforms/InstCombine/select.ll    | 18 ++--
 .../InstCombine/signed-truncation-check.ll    | 74 ++++++++--------
 .../unsigned-add-lack-of-overflow-check.ll    |  2 +-
 llvm/test/Transforms/InstCombine/xor-ashr.ll  | 10 +--
 .../InstSimplify/cast-unsigned-icmp-cmp-0.ll  | 36 ++++----
 .../Transforms/InstSimplify/icmp-constant.ll  | 60 ++++++-------
 .../InstSimplify/maxmin_intrinsics.ll         | 52 ++++++------
 llvm/unittests/IR/InstructionsTest.cpp        | 23 +++--
 llvm/unittests/IR/PatternMatch.cpp            | 85 +++++++++++++------
 42 files changed, 479 insertions(+), 413 deletions(-)

diff --git a/llvm/include/llvm/IR/Constant.h b/llvm/include/llvm/IR/Constant.h
index 778764062227c..d3171acf7b9ac 100644
--- a/llvm/include/llvm/IR/Constant.h
+++ b/llvm/include/llvm/IR/Constant.h
@@ -146,9 +146,9 @@ class Constant : public User {
   Constant *getAggregateElement(Constant *Elt) const;
 
   /// If all elements of the vector constant have the same value, return that
-  /// value. Otherwise, return nullptr. Ignore undefined elements by setting
-  /// AllowUndefs to true.
-  Constant *getSplatValue(bool AllowUndefs = false) const;
+  /// value. Otherwise, return nullptr. Ignore poison elements by setting
+  /// AllowPoison to true.
+  Constant *getSplatValue(bool AllowPoison = false) const;
 
   /// If C is a constant integer then return its value, otherwise C must be a
   /// vector of constant integers, all equal, and the common value is returned.
diff --git a/llvm/include/llvm/IR/Constants.h b/llvm/include/llvm/IR/Constants.h
index 4290ef4486c6f..9ec81903f09c9 100644
--- a/llvm/include/llvm/IR/Constants.h
+++ b/llvm/include/llvm/IR/Constants.h
@@ -532,9 +532,9 @@ class ConstantVector final : public ConstantAggregate {
   }
 
   /// If all elements of the vector constant have the same value, return that
-  /// value. Otherwise, return nullptr. Ignore undefined elements by setting
-  /// AllowUndefs to true.
-  Constant *getSplatValue(bool AllowUndefs = false) const;
+  /// value. Otherwise, return nullptr. Ignore poison elements by setting
+  /// AllowPoison to true.
+  Constant *getSplatValue(bool AllowPoison = false) const;
 
   /// Methods for support type inquiry through isa, cast, and dyn_cast:
   static bool classof(const Value *V) {
diff --git a/llvm/include/llvm/IR/PatternMatch.h b/llvm/include/llvm/IR/PatternMatch.h
index 98cc0e5037698..30e9c3900c866 100644
--- a/llvm/include/llvm/IR/PatternMatch.h
+++ b/llvm/include/llvm/IR/PatternMatch.h
@@ -243,10 +243,10 @@ inline match_combine_and<LTy, RTy> m_CombineAnd(const LTy &L, const RTy &R) {
 
 struct apint_match {
   const APInt *&Res;
-  bool AllowUndef;
+  bool AllowPoison;
 
-  apint_match(const APInt *&Res, bool AllowUndef)
-      : Res(Res), AllowUndef(AllowUndef) {}
+  apint_match(const APInt *&Res, bool AllowPoison)
+      : Res(Res), AllowPoison(AllowPoison) {}
 
   template <typename ITy> bool match(ITy *V) {
     if (auto *CI = dyn_cast<ConstantInt>(V)) {
@@ -256,7 +256,7 @@ struct apint_match {
     if (V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
         if (auto *CI =
-                dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowUndef))) {
+                dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowPoison))) {
           Res = &CI->getValue();
           return true;
         }
@@ -268,10 +268,10 @@ struct apint_match {
 // function for both apint/apfloat.
 struct apfloat_match {
   const APFloat *&Res;
-  bool AllowUndef;
+  bool AllowPoison;
 
-  apfloat_match(const APFloat *&Res, bool AllowUndef)
-      : Res(Res), AllowUndef(AllowUndef) {}
+  apfloat_match(const APFloat *&Res, bool AllowPoison)
+      : Res(Res), AllowPoison(AllowPoison) {}
 
   template <typename ITy> bool match(ITy *V) {
     if (auto *CI = dyn_cast<ConstantFP>(V)) {
@@ -281,7 +281,7 @@ struct apfloat_match {
     if (V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
         if (auto *CI =
-                dyn_cast_or_null<ConstantFP>(C->getSplatValue(AllowUndef))) {
+                dyn_cast_or_null<ConstantFP>(C->getSplatValue(AllowPoison))) {
           Res = &CI->getValueAPF();
           return true;
         }
@@ -292,35 +292,35 @@ struct apfloat_match {
 /// Match a ConstantInt or splatted ConstantVector, binding the
 /// specified pointer to the contained APInt.
 inline apint_match m_APInt(const APInt *&Res) {
-  // Forbid undefs by default to maintain previous behavior.
-  return apint_match(Res, /* AllowUndef */ false);
+  // Forbid poison by default to maintain previous behavior.
+  return apint_match(Res, /* AllowPoison */ false);
 }
 
-/// Match APInt while allowing undefs in splat vector constants.
-inline apint_match m_APIntAllowUndef(const APInt *&Res) {
-  return apint_match(Res, /* AllowUndef */ true);
+/// Match APInt while allowing poison in splat vector constants.
+inline apint_match m_APIntAllowPoison(const APInt *&Res) {
+  return apint_match(Res, /* AllowPoison */ true);
 }
 
-/// Match APInt while forbidding undefs in splat vector constants.
-inline apint_match m_APIntForbidUndef(const APInt *&Res) {
-  return apint_match(Res, /* AllowUndef */ false);
+/// Match APInt while forbidding poison in splat vector constants.
+inline apint_match m_APIntForbidPoison(const APInt *&Res) {
+  return apint_match(Res, /* AllowPoison */ false);
 }
 
 /// Match a ConstantFP or splatted ConstantVector, binding the
 /// specified pointer to the contained APFloat.
 inline apfloat_match m_APFloat(const APFloat *&Res) {
   // Forbid undefs by default to maintain previous behavior.
-  return apfloat_match(Res, /* AllowUndef */ false);
+  return apfloat_match(Res, /* AllowPoison */ false);
 }
 
-/// Match APFloat while allowing undefs in splat vector constants.
-inline apfloat_match m_APFloatAllowUndef(const APFloat *&Res) {
-  return apfloat_match(Res, /* AllowUndef */ true);
+/// Match APFloat while allowing poison in splat vector constants.
+inline apfloat_match m_APFloatAllowPoison(const APFloat *&Res) {
+  return apfloat_match(Res, /* AllowPoison */ true);
 }
 
-/// Match APFloat while forbidding undefs in splat vector constants.
-inline apfloat_match m_APFloatForbidUndef(const APFloat *&Res) {
-  return apfloat_match(Res, /* AllowUndef */ false);
+/// Match APFloat while forbidding poison in splat vector constants.
+inline apfloat_match m_APFloatForbidPoison(const APFloat *&Res) {
+  return apfloat_match(Res, /* AllowPoison */ false);
 }
 
 template <int64_t Val> struct constantint_match {
@@ -418,7 +418,7 @@ template <typename Predicate> struct api_pred_ty : public Predicate {
 
 /// This helper class is used to match scalar and vector constants that
 /// satisfy a specified predicate, and bind them to an APFloat.
-/// Undefs are allowed in splat vector constants.
+/// Poison is allowed in splat vector constants.
 template <typename Predicate> struct apf_pred_ty : public Predicate {
   const APFloat *&Res;
 
@@ -433,7 +433,7 @@ template <typename Predicate> struct apf_pred_ty : public Predicate {
     if (V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
         if (auto *CI = dyn_cast_or_null<ConstantFP>(
-                C->getSplatValue(/* AllowUndef */ true)))
+                C->getSplatValue(/* AllowPoison */ true)))
           if (this->isValue(CI->getValue())) {
             Res = &CI->getValue();
             return true;
@@ -883,7 +883,7 @@ struct bind_const_intval_ty {
 
 /// Match a specified integer value or vector of all elements of that
 /// value.
-template <bool AllowUndefs> struct specific_intval {
+template <bool AllowPoison> struct specific_intval {
   const APInt &Val;
 
   specific_intval(const APInt &V) : Val(V) {}
@@ -892,13 +892,13 @@ template <bool AllowUndefs> struct specific_intval {
     const auto *CI = dyn_cast<ConstantInt>(V);
     if (!CI && V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
-        CI = dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowUndefs));
+        CI = dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowPoison));
 
     return CI && APInt::isSameValue(CI->getValue(), Val);
   }
 };
 
-template <bool AllowUndefs> struct specific_intval64 {
+template <bool AllowPoison> struct specific_intval64 {
   uint64_t Val;
 
   specific_intval64(uint64_t V) : Val(V) {}
@@ -907,7 +907,7 @@ template <bool AllowUndefs> struct specific_intval64 {
     const auto *CI = dyn_cast<ConstantInt>(V);
     if (!CI && V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
-        CI = dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowUndefs));
+        CI = dyn_cast_or_null<ConstantInt>(C->getSplatValue(AllowPoison));
 
     return CI && CI->getValue() == Val;
   }
@@ -923,11 +923,11 @@ inline specific_intval64<false> m_SpecificInt(uint64_t V) {
   return specific_intval64<false>(V);
 }
 
-inline specific_intval<true> m_SpecificIntAllowUndef(const APInt &V) {
+inline specific_intval<true> m_SpecificIntAllowPoison(const APInt &V) {
   return specific_intval<true>(V);
 }
 
-inline specific_intval64<true> m_SpecificIntAllowUndef(uint64_t V) {
+inline specific_intval64<true> m_SpecificIntAllowPoison(uint64_t V) {
   return specific_intval64<true>(V);
 }
 
@@ -1699,9 +1699,9 @@ struct m_SpecificMask {
   bool match(ArrayRef<int> Mask) { return MaskRef == Mask; }
 };
 
-struct m_SplatOrUndefMask {
+struct m_SplatOrPoisonMask {
   int &SplatIndex;
-  m_SplatOrUndefMask(int &SplatIndex) : SplatIndex(SplatIndex) {}
+  m_SplatOrPoisonMask(int &SplatIndex) : SplatIndex(SplatIndex) {}
   bool match(ArrayRef<int> Mask) {
     const auto *First = find_if(Mask, [](int Elem) { return Elem != -1; });
     if (First == Mask.end())
diff --git a/llvm/lib/Analysis/CmpInstAnalysis.cpp b/llvm/lib/Analysis/CmpInstAnalysis.cpp
index d6407e8750737..a1fa7857764d9 100644
--- a/llvm/lib/Analysis/CmpInstAnalysis.cpp
+++ b/llvm/lib/Analysis/CmpInstAnalysis.cpp
@@ -79,7 +79,7 @@ bool llvm::decomposeBitTestICmp(Value *LHS, Value *RHS,
   using namespace PatternMatch;
 
   const APInt *C;
-  if (!match(RHS, m_APIntAllowUndef(C)))
+  if (!match(RHS, m_APIntAllowPoison(C)))
     return false;
 
   switch (Pred) {
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 431ad37d0a9e8..1bba75d237fb2 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -3023,7 +3023,7 @@ static Value *simplifyICmpWithConstant(CmpInst::Predicate Pred, Value *LHS,
 
   Value *X;
   const APInt *C;
-  if (!match(RHS, m_APIntAllowUndef(C)))
+  if (!match(RHS, m_APIntAllowPoison(C)))
     return nullptr;
 
   // Sign-bit checks can be optimized to true/false after unsigned
@@ -3056,9 +3056,9 @@ static Value *simplifyICmpWithConstant(CmpInst::Predicate Pred, Value *LHS,
   // (mul nuw/nsw X, MulC) == C --> false (if C is not a multiple of MulC)
   const APInt *MulC;
   if (IIQ.UseInstrInfo && ICmpInst::isEquality(Pred) &&
-      ((match(LHS, m_NUWMul(m_Value(), m_APIntAllowUndef(MulC))) &&
+      ((match(LHS, m_NUWMul(m_Value(), m_APIntAllowPoison(MulC))) &&
         *MulC != 0 && C->urem(*MulC) != 0) ||
-       (match(LHS, m_NSWMul(m_Value(), m_APIntAllowUndef(MulC))) &&
+       (match(LHS, m_NSWMul(m_Value(), m_APIntAllowPoison(MulC))) &&
         *MulC != 0 && C->srem(*MulC) != 0)))
     return ConstantInt::get(ITy, Pred == ICmpInst::ICMP_NE);
 
@@ -3203,7 +3203,7 @@ static Value *simplifyICmpWithBinOpOnLHS(CmpInst::Predicate Pred,
 
   // (sub C, X) == X, C is odd  --> false
   // (sub C, X) != X, C is odd  --> true
-  if (match(LBO, m_Sub(m_APIntAllowUndef(C), m_Specific(RHS))) &&
+  if (match(LBO, m_Sub(m_APIntAllowPoison(C), m_Specific(RHS))) &&
       (*C & 1) == 1 && ICmpInst::isEquality(Pred))
     return (Pred == ICmpInst::ICMP_EQ) ? getFalse(ITy) : getTrue(ITy);
 
@@ -3354,7 +3354,7 @@ static Value *simplifyICmpWithBinOp(CmpInst::Predicate Pred, Value *LHS,
   //   (C2 << X) != C --> true
   const APInt *C;
   if (match(LHS, m_Shl(m_Power2(), m_Value())) &&
-      match(RHS, m_APIntAllowUndef(C)) && !C->isPowerOf2()) {
+      match(RHS, m_APIntAllowPoison(C)) && !C->isPowerOf2()) {
     // C2 << X can equal zero in some circumstances.
     // This simplification might be unsafe if C is zero.
     //
@@ -4105,7 +4105,7 @@ static Value *simplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
   }
 
   const APFloat *C = nullptr;
-  match(RHS, m_APFloatAllowUndef(C));
+  match(RHS, m_APFloatAllowPoison(C));
   std::optional<KnownFPClass> FullKnownClassLHS;
 
   // Lazily compute the possible classes for LHS. Avoid computing it twice if
@@ -6459,7 +6459,7 @@ Value *llvm::simplifyBinaryIntrinsic(Intrinsic::ID IID, Type *ReturnType,
           ReturnType, MinMaxIntrinsic::getSaturationPoint(IID, BitWidth));
 
     const APInt *C;
-    if (match(Op1, m_APIntAllowUndef(C))) {
+    if (match(Op1, m_APIntAllowPoison(C))) {
       // Clamp to limit value. For example:
       // umax(i8 %x, i8 255) --> 255
       if (*C == MinMaxIntrinsic::getSaturationPoint(IID, BitWidth))
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 695d0b0b1a2f8..b87bba7c20e0d 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -4116,7 +4116,7 @@ std::pair<Value *, FPClassTest> llvm::fcmpToClassTest(FCmpInst::Predicate Pred,
                                                       Value *LHS, Value *RHS,
                                                       bool LookThroughSrc) {
   const APFloat *ConstRHS;
-  if (!match(RHS, m_APFloatAllowUndef(ConstRHS)))
+  if (!match(RHS, m_APFloatAllowPoison(ConstRHS)))
     return {nullptr, fcAllFlags};
 
   return fcmpToClassTest(Pred, F, LHS, ConstRHS, LookThroughSrc);
@@ -4517,7 +4517,7 @@ std::tuple<Value *, FPClassTest, FPClassTest>
 llvm::fcmpImpliesClass(CmpInst::Predicate Pred, const Function &F, Value *LHS,
                        Value *RHS, bool LookThroughSrc) {
   const APFloat *ConstRHS;
-  if (!match(RHS, m_APFloatAllowUndef(ConstRHS)))
+  if (!match(RHS, m_APFloatAllowPoison(ConstRHS)))
     return {nullptr, fcAllFlags, fcAllFlags};
 
   // TODO: Just call computeKnownFPClass for RHS to handle non-constants.
diff --git a/llvm/lib/IR/Constants.cpp b/llvm/lib/IR/Constants.cpp
index 45b359a94b3ab..5268eccf70144 100644
--- a/llvm/lib/IR/Constants.cpp
+++ b/llvm/lib/IR/Constants.cpp
@@ -1696,14 +1696,14 @@ void ConstantVector::destroyConstantImpl() {
   getType()->getContext().pImpl->VectorConstants.remove(this);
 }
 
-Constant *Constant::getSplatValue(bool AllowUndefs) const {
+Constant *Constant::getSplatValue(bool AllowPoison) const {
   assert(this->getType()->isVectorTy() && "Only valid for vectors!");
   if (isa<ConstantAggregateZero>(this))
     return getNullValue(cast<VectorType>(getType())->getElementType());
   if (const ConstantDataVector *CV = dyn_cast<ConstantDataVector>(this))
     return CV->getSplatValue();
   if (const ConstantVector *CV = dyn_cast<ConstantVector>(this))
-    return CV->getSplatValue(AllowUndefs);
+    return CV->getSplatValue(AllowPoison);
 
   // Check if this is a constant expression splat of the form returned by
   // ConstantVector::getSplat()
@@ -1728,7 +1728,7 @@ Constant *Constant::getSplatValue(bool AllowUndefs) const {
   return nullptr;
 }
 
-Constant *ConstantVector::getSplatValue(bool AllowUndefs) const {
+Constant *ConstantVector::getSplatValue(bool AllowPoison) const {
   // Check out first element.
   Constant *Elt = getOperand(0);
   // Then make sure all remaining elements point to the same value.
@@ -1738,15 +1738,15 @@ Constant *ConstantVector::getSplatValue(bool AllowUndefs) const {
       continue;
 
     // Strict mode: any mismatch is not a splat.
-    if (!AllowUndefs)
+    if (!AllowPoison)
       return nullptr;
 
-    // Allow undefs mode: ignore undefined elements.
-    if (isa<UndefValue>(OpC))
+    // Allow poison mode: ignore poison elements.
+    if (isa<PoisonValue>(OpC))
       continue;
 
     // If we do not have a defined element yet, use the current operand.
-    if (isa<UndefValue>(Elt))
+    if (isa<PoisonValue>(Elt))
       Elt = OpC;
 
     if (OpC != Elt)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
index 5aa35becd842c..978a2d49b08bc 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
@@ -906,8 +906,8 @@ bool AMDGPULibCalls::fold_pow(FPMathOperator *FPOp, IRBuilder<> &B,
 
   const APFloat *CF = nullptr;
   const APInt *CINT = nullptr;
-  if (!match(opr1, m_APFloatAllowUndef(CF)))
-    match(opr1, m_APIntAllowUndef(CINT));
+  if (!match(opr1, m_APFloatAllowPoison(CF)))
+    match(opr1, m_APIntAllowPoison(CINT));
 
   // 0x1111111 means that we don't do anything for this call.
   int ci_opr1 = (CINT ? (int)CINT->getSExtValue() : 0x1111111);
@@ -1039,7 +1039,7 @@ bool AMDGPULibCalls::fold_pow(FPMathOperator *FPOp, IRBuilder<> &B,
   Constant *cnval = nullptr;
   if (getVecSize(FInfo) == 1) {
     CF = nullptr;
-    match(opr0, m_APFloatAllowUndef(CF));
+    match(opr0, m_APFloatAllowPoison(CF));
 
     if (CF) {
       double V = (getArgType(FInfo) == AMDGPULibFunc::F32)
diff --git a/llvm/lib/Target/X86/X86FixupVectorConstants.cpp b/llvm/lib/Target/X86/X86FixupVectorConstants.cpp
index da7dcbb25a957..c9f79e1645f58 100644
--- a/llvm/lib/Target/X86/X86FixupVectorConstants.cpp
+++ b/llvm/lib/Target/X86/X86FixupVectorConstants.cpp
@@ -64,6 +64,23 @@ FunctionPass *llvm::createX86FixupVectorConstants() {
   return new X86FixupVectorConstantsPass();
 }
 
+/// Normally, we only allow poison in vector splats. However, as this is part
+/// of the backend, and working with the DAG representation, which currently
+/// only natively represents undef values, we need to accept undefs here.
+static Constant *getSplatValueAllowUndef(const ConstantVector *C) {
+  Constant *Res = nullptr;
+  for (Value *Op : C->operands()) {
+    Constant *OpC = cast<Constant>(Op);
+    if (isa<UndefValue>(OpC))
+      continue;
+    if (!Res)
+      Res = OpC;
+    else if (Res != OpC)
+      return nullptr;
+  }
+  return Res;
+}
+
 // Attempt to extract the full width of bits data from the constant.
 static std::optional<APInt> extractConstantBits(const Constant *C) {
   unsigned NumBits = C->getType()->getPrimitiveSizeInBits();
@@ -78,7 +95,7 @@ static std::optional<APInt> extractConstantBits(const Constant *C) {
     return CFP->getValue().bitcastToAPInt();
 
   if (auto *CV = dyn_cast<ConstantVector>(C)) {
-    if (auto *CVSplat = CV->getSplatValue(/*AllowUndefs*/ true)) {
+    if (auto *CVSplat = getSplatValueAllowUndef(CV)) {
       if (std::optional<APInt> Bits = extractConstantBits(CVSplat)) {
         assert((NumBits % Bits->getBitWidth()) == 0 && "Illegal splat");
         return APInt::getSplat(NumBits, *Bits);
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
index df7f02810ee43..9a16ca96db76b 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
@@ -896,7 +896,7 @@ Instruction *InstCombinerImpl::foldAddWithConstant(BinaryOperator &Add) {
   const APInt *C;
   unsigned BitWidth = Ty->getScalarSizeInBits();
   if (match(Op0, m_OneUse(m_AShr(m_Value(X),
-                                 m_SpecificIntAllowUndef(BitWidth - 1)))) &&
+                                 m_SpecificIntAllowPoison(BitWidth - 1)))) &&
       match(Op1, m_One()))
     return new ZExtInst(Builder.CreateIsNotNeg(X, "isnotneg"), Ty);
 
@@ -1656,7 +1656,7 @@ Instruction *InstCombinerImpl::visitAdd(BinaryOperator &I) {
   // (A s>> (BW - 1)) + (zext (A s> 0)) --> (A s>> (BW - 1)) | (zext (A != 0))
   ICmpInst::Predicate Pred;
   uint64_t BitWidth = Ty->getScalarSizeInBits();
-  if (match(LHS, m_AShr(m_Value(A), m_SpecificIntAllowUndef(BitWidth - 1))) &&
+  if (match(LHS, m_AShr(m_Value(A), m_SpecificIntAllowPoison(BitWidth - 1))) &&
       match(RHS, m_OneUse(m_ZExt(
                      m_OneUse(m_ICmp(Pred, m_Specific(A), m_ZeroInt()))))) &&
       Pred == CmpInst::ICMP_SGT) {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index 0f4fbf5bbfbbd..bf7c0074a38f0 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -947,9 +947,9 @@ static Value *foldNegativePower2AndShiftedMask(
   // bits (0).
   auto isReducible = [](const Value *B, const Value *D, const Value *E) {
     const APInt *BCst, *DCst, *ECst;
-    return match(B, m_APIntAllowUndef(BCst)) && match(D, m_APInt(DCst)) &&
+    return match(B, m_APIntAllowPoison(BCst)) && match(D, m_APInt(DCst)) &&
            match(E, m_APInt(ECst)) && *DCst == *ECst &&
-           (isa<UndefValue>(B) ||
+           (isa<PoisonValue>(B) ||
             (BCst->countLeadingOnes() == DCst->countLeadingZeros()));
   };
 
@@ -1424,8 +1424,8 @@ Value *InstCombinerImpl::foldLogicOfFCmps(FCmpInst *LHS, FCmpInst *RHS,
   const APFloat *LHSC, *RHSC;
   if (LHS0 == RHS0 && LHS->hasOneUse() && RHS->hasOneUse() &&
       FCmpInst::getSwappedPredicate(PredL) == PredR &&
-      match(LHS1, m_APFloatAllowUndef(LHSC)) &&
-      match(RHS1, m_APFloatAllowUndef(RHSC)) &&
+      match(LHS1, m_APFloatAllowPoison(LHSC)) &&
+      match(RHS1, m_APFloatAllowPoison(RHSC)) &&
       LHSC->bitwiseIsEqual(neg(*RHSC))) {
     auto IsLessThanOrLessEqual = [](FCmpInst::Predicate Pred) {
       switch (Pred) {
@@ -2730,7 +2730,7 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
 
   // (iN X s>> (N-1)) & Y --> (X s< 0) ? Y : 0 -- with optional sext
   if (match(&I, m_c_And(m_OneUse(m_SExtOrSelf(
-                            m_AShr(m_Value(X), m_APIntAllowUndef(C)))),
+                            m_AShr(m_Value(X), m_APIntAllowPoison(C)))),
                         m_Value(Y))) &&
       *C == X->getType()->getScalarSizeInBits() - 1) {
     Value *IsNeg = Builder.CreateIsNeg(X, "isneg");
@@ -2739,7 +2739,7 @@ Instruction *InstCombinerImpl::visitAnd(BinaryOperator &I) {
   // If there's a 'not' of the shifted value, swap the select operands:
   // ~(iN X s>> (N-1)) & Y --> (X s< 0) ? 0 : Y -- with optional sext
   if (match(&I, m_c_And(m_OneUse(m_SExtOrSelf(
-                            m_Not(m_AShr(m_Value(X), m_APIntAllowUndef(C))))),
+                            m_Not(m_AShr(m_Value(X), m_APIntAllowPoison(C))))),
                         m_Value(Y))) &&
       *C == X->getType()->getScalarSizeInBits() - 1) {
     Value *IsNeg = Builder.CreateIsNeg(X, "isneg");
@@ -2840,7 +2840,7 @@ InstCombinerImpl::convertOrOfShiftsToFunnelShift(Instruction &Or) {
     auto matchShiftAmount = [&](Value *L, Value *R, unsigned Width) -> Value * {
       // Check for constant shift amounts that sum to the bitwidth.
       const APInt *LI, *RI;
-      if (match(L, m_APIntAllowUndef(LI)) && match(R, m_APIntAllowUndef(RI)))
+      if (match(L, m_APIntAllowPoison(LI)) && match(R, m_APIntAllowPoison(RI)))
         if (LI->ult(Width) && RI->ult(Width) && (*LI + *RI) == Width)
           return ConstantInt::get(L->getType(), *LI);
 
@@ -2850,7 +2850,7 @@ InstCombinerImpl::convertOrOfShiftsToFunnelShift(Instruction &Or) {
                 m_SpecificInt_ICMP(ICmpInst::ICMP_ULT, APInt(Width, Width))) &&
           match(R,
                 m_SpecificInt_ICMP(ICmpInst::ICMP_ULT, APInt(Width, Width))) &&
-          match(ConstantExpr::getAdd(LC, RC), m_SpecificIntAllowUndef(Width)))
+          match(ConstantExpr::getAdd(LC, RC), m_SpecificIntAllowPoison(Width)))
         return ConstantExpr::mergeUndefsWith(LC, RC);
 
       // (shl ShVal, X) | (lshr ShVal, (Width - x)) iff X < Width.
@@ -3195,14 +3195,14 @@ static Value *foldAndOrOfICmpEqConstantAndICmp(ICmpInst *LHS, ICmpInst *RHS,
 
   const APInt *CInt;
   if (LPred != ICmpInst::ICMP_EQ ||
-      !match(LHS->getOperand(1), m_APIntAllowUndef(CInt)) ||
+      !match(LHS->getOperand(1), m_APIntAllowPoison(CInt)) ||
       !LHS0->getType()->isIntOrIntVectorTy() ||
       !(LHS->hasOneUse() || RHS->hasOneUse()))
     return nullptr;
 
   auto MatchRHSOp = [LHS0, CInt](const Value *RHSOp) {
     return match(RHSOp,
-                 m_Add(m_Specific(LHS0), m_SpecificIntAllowUndef(-*CInt))) ||
+                 m_Add(m_Specific(LHS0), m_SpecificIntAllowPoison(-*CInt))) ||
            (CInt->isZero() && RHSOp == LHS0);
   };
 
@@ -4635,7 +4635,7 @@ Instruction *InstCombinerImpl::visitXor(BinaryOperator &I) {
     // constant depending on whether this input is less than 0.
     const APInt *CA;
     if (match(Op0, m_OneUse(m_TruncOrSelf(
-                       m_AShr(m_Value(X), m_APIntAllowUndef(CA))))) &&
+                       m_AShr(m_Value(X), m_APIntAllowPoison(CA))))) &&
         *CA == X->getType()->getScalarSizeInBits() - 1 &&
         !match(C1, m_AllOnes())) {
       assert(!C1->isZeroValue() && "Unexpected xor with 0");
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 60e4be883f513..aa547565bd085 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -1821,7 +1821,7 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
        return NewMinMax;
 
     // Try to fold minmax with constant RHS based on range information
-    if (match(I1, m_APIntAllowUndef(RHSC))) {
+    if (match(I1, m_APIntAllowPoison(RHSC))) {
       ICmpInst::Predicate Pred =
           ICmpInst::getNonStrictPredicate(MinMaxIntrinsic::getPredicate(IID));
       bool IsSigned = MinMaxIntrinsic::isSigned(IID);
@@ -1865,12 +1865,8 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
     // bswap (lshr X, Y) --> shl (bswap X), Y
     Value *X, *Y;
     if (match(IIOperand, m_OneUse(m_LogicalShift(m_Value(X), m_Value(Y))))) {
-      // The transform allows undef vector elements, so try a constant match
-      // first. If knownbits can handle that case, that clause could be removed.
       unsigned BitWidth = IIOperand->getType()->getScalarSizeInBits();
-      const APInt *C;
-      if ((match(Y, m_APIntAllowUndef(C)) && (*C & 7) == 0) ||
-          MaskedValueIsZero(Y, APInt::getLowBitsSet(BitWidth, 3))) {
+      if (MaskedValueIsZero(Y, APInt::getLowBitsSet(BitWidth, 3))) {
         Value *NewSwap = Builder.CreateUnaryIntrinsic(Intrinsic::bswap, X);
         BinaryOperator::BinaryOps InverseShift =
             cast<BinaryOperator>(IIOperand)->getOpcode() == Instruction::Shl
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 4537a47da2ced..6ce39be59bda9 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -1481,7 +1481,7 @@ Instruction *InstCombinerImpl::visitSExt(SExtInst &Sext) {
     Value *Y;
     if (Src->hasOneUse() &&
         match(X, m_LShr(m_Value(Y),
-                        m_SpecificIntAllowUndef(XBitSize - SrcBitSize)))) {
+                        m_SpecificIntAllowPoison(XBitSize - SrcBitSize)))) {
       Value *Ashr = Builder.CreateAShr(Y, XBitSize - SrcBitSize);
       return CastInst::CreateIntegerCast(Ashr, DestTy, /* isSigned */ true);
     }
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index de90907701743..00fb1f7dd75b7 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -3417,8 +3417,8 @@ Instruction *InstCombinerImpl::foldICmpInstWithConstant(ICmpInst &Cmp) {
       return new ICmpInst(Cmp.getPredicate(), X, Y);
   }
 
-  if (match(Cmp.getOperand(1), m_APIntAllowUndef(C)))
-    return foldICmpInstWithConstantAllowUndef(Cmp, *C);
+  if (match(Cmp.getOperand(1), m_APIntAllowPoison(C)))
+    return foldICmpInstWithConstantAllowPoison(Cmp, *C);
 
   return nullptr;
 }
@@ -3735,11 +3735,11 @@ foldICmpIntrinsicWithIntrinsic(ICmpInst &Cmp,
 }
 
 /// Try to fold integer comparisons with a constant operand: icmp Pred X, C
-/// where X is some kind of instruction and C is AllowUndef.
-/// TODO: Move more folds which allow undef to this function.
+/// where X is some kind of instruction and C is AllowPoison.
+/// TODO: Move more folds which allow poison to this function.
 Instruction *
-InstCombinerImpl::foldICmpInstWithConstantAllowUndef(ICmpInst &Cmp,
-                                                     const APInt &C) {
+InstCombinerImpl::foldICmpInstWithConstantAllowPoison(ICmpInst &Cmp,
+                                                      const APInt &C) {
   const ICmpInst::Predicate Pred = Cmp.getPredicate();
   if (auto *II = dyn_cast<IntrinsicInst>(Cmp.getOperand(0))) {
     switch (II->getIntrinsicID()) {
@@ -4844,7 +4844,7 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
     const APInt *C;
     if ((Pred == ICmpInst::ICMP_ULT || Pred == ICmpInst::ICMP_UGE) &&
         match(Op0, m_And(m_BinOp(BO), m_LowBitMask(C))) &&
-        match(BO, m_Add(m_Specific(Op1), m_SpecificIntAllowUndef(*C)))) {
+        match(BO, m_Add(m_Specific(Op1), m_SpecificIntAllowPoison(*C)))) {
       CmpInst::Predicate NewPred =
           Pred == ICmpInst::ICMP_ULT ? ICmpInst::ICMP_NE : ICmpInst::ICMP_EQ;
       Constant *Zero = ConstantInt::getNullValue(Op1->getType());
@@ -4853,7 +4853,7 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
 
     if ((Pred == ICmpInst::ICMP_UGT || Pred == ICmpInst::ICMP_ULE) &&
         match(Op1, m_And(m_BinOp(BO), m_LowBitMask(C))) &&
-        match(BO, m_Add(m_Specific(Op0), m_SpecificIntAllowUndef(*C)))) {
+        match(BO, m_Add(m_Specific(Op0), m_SpecificIntAllowPoison(*C)))) {
       CmpInst::Predicate NewPred =
           Pred == ICmpInst::ICMP_UGT ? ICmpInst::ICMP_NE : ICmpInst::ICMP_EQ;
       Constant *Zero = ConstantInt::getNullValue(Op1->getType());
@@ -5003,8 +5003,9 @@ Instruction *InstCombinerImpl::foldICmpBinOp(ICmpInst &I,
       (BO0->hasOneUse() || BO1->hasOneUse()) && !I.isUnsigned()) {
     const APInt *AP1, *AP2;
     // TODO: Support non-uniform vectors.
-    // TODO: Allow undef passthrough if B AND D's element is undef.
-    if (match(B, m_APIntAllowUndef(AP1)) && match(D, m_APIntAllowUndef(AP2)) &&
+    // TODO: Allow poison passthrough if B or D's element is poison.
+    if (match(B, m_APIntAllowPoison(AP1)) &&
+        match(D, m_APIntAllowPoison(AP2)) &&
         AP1->isNegative() == AP2->isNegative()) {
       APInt AP1Abs = AP1->abs();
       APInt AP2Abs = AP2->abs();
@@ -5575,10 +5576,10 @@ Instruction *InstCombinerImpl::foldICmpEquality(ICmpInst &I) {
   // (A >> C) == (B >> C) --> (A^B) u< (1 << C)
   // For lshr and ashr pairs.
   const APInt *AP1, *AP2;
-  if ((match(Op0, m_OneUse(m_LShr(m_Value(A), m_APIntAllowUndef(AP1)))) &&
-       match(Op1, m_OneUse(m_LShr(m_Value(B), m_APIntAllowUndef(AP2))))) ||
-      (match(Op0, m_OneUse(m_AShr(m_Value(A), m_APIntAllowUndef(AP1)))) &&
-       match(Op1, m_OneUse(m_AShr(m_Value(B), m_APIntAllowUndef(AP2)))))) {
+  if ((match(Op0, m_OneUse(m_LShr(m_Value(A), m_APIntAllowPoison(AP1)))) &&
+       match(Op1, m_OneUse(m_LShr(m_Value(B), m_APIntAllowPoison(AP2))))) ||
+      (match(Op0, m_OneUse(m_AShr(m_Value(A), m_APIntAllowPoison(AP1)))) &&
+       match(Op1, m_OneUse(m_AShr(m_Value(B), m_APIntAllowPoison(AP2)))))) {
     if (AP1 != AP2)
       return nullptr;
     unsigned TypeBits = AP1->getBitWidth();
@@ -6969,10 +6970,10 @@ static Instruction *foldVectorCmp(CmpInst &Cmp,
 
   // Length-changing splats are ok, so adjust the constants as needed:
   // cmp (shuffle V1, M), C --> shuffle (cmp V1, C'), M
-  Constant *ScalarC = C->getSplatValue(/* AllowUndefs */ true);
+  Constant *ScalarC = C->getSplatValue(/* AllowPoison */ true);
   int MaskSplatIndex;
-  if (ScalarC && match(M, m_SplatOrUndefMask(MaskSplatIndex))) {
-    // We allow undefs in matching, but this transform removes those for safety.
+  if (ScalarC && match(M, m_SplatOrPoisonMask(MaskSplatIndex))) {
+    // We allow poison in matching, but this transform removes it for safety.
     // Demanded elements analysis should be able to recover some/all of that.
     C = ConstantVector::getSplat(cast<VectorType>(V1Ty)->getElementCount(),
                                  ScalarC);
@@ -7444,7 +7445,7 @@ Instruction *InstCombinerImpl::visitICmpInst(ICmpInst &I) {
       unsigned OpWidth = Op0->getType()->getScalarSizeInBits();
       Instruction *ShiftI;
       if (match(Op0, m_CombineAnd(m_Instruction(ShiftI),
-                                  m_Shr(m_Value(X), m_SpecificIntAllowUndef(
+                                  m_Shr(m_Value(X), m_SpecificIntAllowPoison(
                                                         OpWidth - 1))))) {
         unsigned ExtOpc = ExtI->getOpcode();
         unsigned ShiftOpc = ShiftI->getOpcode();
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
index b9ad3a7400792..f4d559036b31f 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
+++ b/llvm/lib/Transforms/InstCombine/InstCombineInternal.h
@@ -661,8 +661,8 @@ class LLVM_LIBRARY_VISIBILITY InstCombinerImpl final
   Instruction *foldICmpUsingBoolRange(ICmpInst &I);
   Instruction *foldICmpInstWithConstant(ICmpInst &Cmp);
   Instruction *foldICmpInstWithConstantNotInt(ICmpInst &Cmp);
-  Instruction *foldICmpInstWithConstantAllowUndef(ICmpInst &Cmp,
-                                                  const APInt &C);
+  Instruction *foldICmpInstWithConstantAllowPoison(ICmpInst &Cmp,
+                                                   const APInt &C);
   Instruction *foldICmpBinOp(ICmpInst &Cmp, const SimplifyQuery &SQ);
   Instruction *foldICmpWithMinMax(Instruction &I, MinMaxIntrinsic *MinMax,
                                   Value *Z, ICmpInst::Predicate Pred);
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 7b86fcde8937b..53478e07394b2 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -276,7 +276,7 @@ Instruction *InstCombinerImpl::visitMul(BinaryOperator &I) {
     const APInt *NegPow2C;
     Value *X;
     if (match(Op0, m_ZExtOrSExt(m_Value(X))) &&
-        match(Op1, m_APIntAllowUndef(NegPow2C))) {
+        match(Op1, m_APIntAllowPoison(NegPow2C))) {
       unsigned SrcWidth = X->getType()->getScalarSizeInBits();
       unsigned ShiftAmt = NegPow2C->countr_zero();
       if (ShiftAmt >= BitWidth - SrcWidth) {
@@ -485,7 +485,7 @@ Instruction *InstCombinerImpl::visitMul(BinaryOperator &I) {
   // ((ashr X, 31) | 1) * X --> abs(X)
   // X * ((ashr X, 31) | 1) --> abs(X)
   if (match(&I, m_c_BinOp(m_Or(m_AShr(m_Value(X),
-                                      m_SpecificIntAllowUndef(BitWidth - 1)),
+                                      m_SpecificIntAllowPoison(BitWidth - 1)),
                                m_One()),
                           m_Deferred(X)))) {
     Value *Abs = Builder.CreateBinaryIntrinsic(
@@ -836,7 +836,7 @@ Instruction *InstCombinerImpl::visitFMul(BinaryOperator &I) {
   // X * 0.0 --> copysign(0.0, X)
   // X * -0.0 --> copysign(0.0, -X)
   const APFloat *FPC;
-  if (match(Op1, m_APFloatAllowUndef(FPC)) && FPC->isZero() &&
+  if (match(Op1, m_APFloatAllowPoison(FPC)) && FPC->isZero() &&
       ((I.hasNoInfs() &&
         isKnownNeverNaN(Op0, /*Depth=*/0, SQ.getWithInstruction(&I))) ||
        isKnownNeverNaN(&I, /*Depth=*/0, SQ.getWithInstruction(&I)))) {
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp
index cd8e47e1b3911..d697f361dec02 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineNegator.cpp
@@ -249,7 +249,7 @@ std::array<Value *, 2> Negator::getSortedOperandsOfBinOp(Instruction *I) {
     unsigned SrcWidth = SrcOp->getType()->getScalarSizeInBits();
     const APInt &FullShift = APInt(SrcWidth, SrcWidth - 1);
     if (IsTrulyNegation &&
-        match(SrcOp, m_LShr(m_Value(X), m_SpecificIntAllowUndef(FullShift)))) {
+        match(SrcOp, m_LShr(m_Value(X), m_SpecificIntAllowPoison(FullShift)))) {
       Value *Ashr = Builder.CreateAShr(X, FullShift);
       return Builder.CreateSExt(Ashr, I->getType());
     }
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index a87d392916159..0262af28068be 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -1621,7 +1621,7 @@ static Instruction *foldSelectZeroOrOnes(ICmpInst *Cmp, Value *TVal,
     return nullptr;
 
   const APInt *CmpC;
-  if (!match(Cmp->getOperand(1), m_APIntAllowUndef(CmpC)))
+  if (!match(Cmp->getOperand(1), m_APIntAllowPoison(CmpC)))
     return nullptr;
 
   // (X u< 2) ? -X : -1 --> sext (X != 0)
@@ -2489,8 +2489,8 @@ static Instruction *foldSelectToCopysign(SelectInst &Sel,
   // Match select ?, TC, FC where the constants are equal but negated.
   // TODO: Generalize to handle a negated variable operand?
   const APFloat *TC, *FC;
-  if (!match(TVal, m_APFloatAllowUndef(TC)) ||
-      !match(FVal, m_APFloatAllowUndef(FC)) ||
+  if (!match(TVal, m_APFloatAllowPoison(TC)) ||
+      !match(FVal, m_APFloatAllowPoison(FC)) ||
       !abs(*TC).bitwiseIsEqual(abs(*FC)))
     return nullptr;
 
@@ -2939,17 +2939,17 @@ foldRoundUpIntegerWithPow2Alignment(SelectInst &SI,
   // FIXME: we could support non non-splats here.
 
   const APInt *LowBitMaskCst;
-  if (!match(XLowBits, m_And(m_Specific(X), m_APIntAllowUndef(LowBitMaskCst))))
+  if (!match(XLowBits, m_And(m_Specific(X), m_APIntAllowPoison(LowBitMaskCst))))
     return nullptr;
 
   // Match even if the AND and ADD are swapped.
   const APInt *BiasCst, *HighBitMaskCst;
   if (!match(XBiasedHighBits,
-             m_And(m_Add(m_Specific(X), m_APIntAllowUndef(BiasCst)),
-                   m_APIntAllowUndef(HighBitMaskCst))) &&
+             m_And(m_Add(m_Specific(X), m_APIntAllowPoison(BiasCst)),
+                   m_APIntAllowPoison(HighBitMaskCst))) &&
       !match(XBiasedHighBits,
-             m_Add(m_And(m_Specific(X), m_APIntAllowUndef(HighBitMaskCst)),
-                   m_APIntAllowUndef(BiasCst))))
+             m_Add(m_And(m_Specific(X), m_APIntAllowPoison(HighBitMaskCst)),
+                   m_APIntAllowPoison(BiasCst))))
     return nullptr;
 
   if (!LowBitMaskCst->isMask())
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
index 95aa2119e2d88..244f03a1bc2b4 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
@@ -778,7 +778,7 @@ Instruction *InstCombinerImpl::FoldShiftByConstant(Value *Op0, Constant *C1,
   // (X / +DivC) >> (Width - 1) --> ext (X <= -DivC)
   // (X / -DivC) >> (Width - 1) --> ext (X >= +DivC)
   const APInt *DivC;
-  if (!IsLeftShift && match(C1, m_SpecificIntAllowUndef(TypeBits - 1)) &&
+  if (!IsLeftShift && match(C1, m_SpecificIntAllowPoison(TypeBits - 1)) &&
       match(Op0, m_SDiv(m_Value(X), m_APInt(DivC))) && !DivC->isZero() &&
       !DivC->isMinSignedValue()) {
     Constant *NegDivC = ConstantInt::get(Ty, -(*DivC));
@@ -1264,7 +1264,7 @@ Instruction *InstCombinerImpl::visitLShr(BinaryOperator &I) {
 
   // (iN (~X) u>> (N - 1)) --> zext (X > -1)
   if (match(Op0, m_OneUse(m_Not(m_Value(X)))) &&
-      match(Op1, m_SpecificIntAllowUndef(BitWidth - 1)))
+      match(Op1, m_SpecificIntAllowPoison(BitWidth - 1)))
     return new ZExtInst(Builder.CreateIsNotNeg(X, "isnotneg"), Ty);
 
   if (match(Op1, m_APInt(C))) {
@@ -1666,9 +1666,9 @@ Instruction *InstCombinerImpl::visitAShr(BinaryOperator &I) {
   // as the pattern to splat the lowest bit.
   // FIXME: iff X is already masked, we don't need the one-use check.
   Value *X;
-  if (match(Op1, m_SpecificIntAllowUndef(BitWidth - 1)) &&
+  if (match(Op1, m_SpecificIntAllowPoison(BitWidth - 1)) &&
       match(Op0, m_OneUse(m_Shl(m_Value(X),
-                                m_SpecificIntAllowUndef(BitWidth - 1))))) {
+                                m_SpecificIntAllowPoison(BitWidth - 1))))) {
     Constant *Mask = ConstantInt::get(Ty, 1);
     // Retain the knowledge about the ignored lanes.
     Mask = Constant::mergeUndefsWith(
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index e08d0a80e7653..ee9b1afbb86ac 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -2190,7 +2190,7 @@ Instruction *InstCombinerImpl::foldVectorBinop(BinaryOperator &Inst) {
     Value *Y, *OtherOp;
     if (!match(LHS,
                m_OneUse(m_Shuffle(m_Value(X), m_Undef(), m_Mask(MaskC)))) ||
-        !match(MaskC, m_SplatOrUndefMask(SplatIndex)) ||
+        !match(MaskC, m_SplatOrPoisonMask(SplatIndex)) ||
         X->getType() != Inst.getType() ||
         !match(RHS, m_OneUse(m_BinOp(Opcode, m_Value(Y), m_Value(OtherOp)))))
       return nullptr;
@@ -3740,7 +3740,7 @@ InstCombinerImpl::foldExtractOfOverflowIntrinsic(ExtractValueInst &EV) {
 
   Intrinsic::ID OvID = WO->getIntrinsicID();
   const APInt *C = nullptr;
-  if (match(WO->getRHS(), m_APIntAllowUndef(C))) {
+  if (match(WO->getRHS(), m_APIntAllowPoison(C))) {
     if (*EV.idx_begin() == 0 && (OvID == Intrinsic::smul_with_overflow ||
                                  OvID == Intrinsic::umul_with_overflow)) {
       // extractvalue (any_mul_with_overflow X, -1), 0 --> -X
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index a03a408686ef1..ae7717eb7cc97 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -65,7 +65,7 @@ struct specific_intval {
     if (!CI && V->getType()->isVectorTy())
       if (const auto *C = dyn_cast<Constant>(V))
         CI = dyn_cast_or_null<ConstantInt>(
-            C->getSplatValue(/*UndefsAllowed=*/false));
+            C->getSplatValue(/*AllowPoison=*/false));
 
     return CI && APInt::isSameValue(CI->getValue(), Val);
   }
diff --git a/llvm/test/Transforms/InstCombine/and-or-icmp-const-icmp.ll b/llvm/test/Transforms/InstCombine/and-or-icmp-const-icmp.ll
index 8531c7db1ee3f..9143a2d0ccda2 100644
--- a/llvm/test/Transforms/InstCombine/and-or-icmp-const-icmp.ll
+++ b/llvm/test/Transforms/InstCombine/and-or-icmp-const-icmp.ll
@@ -223,30 +223,30 @@ define <2 x i1> @ne_vector_unequal2(<2 x i8> %x, <2 x i8> %y) {
 }
 
 ; ==============================================================================
-; Tests with undef
+; Tests with poison
 ; ==============================================================================
-define <2 x i1> @eq_vector_undef_icmp(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: define <2 x i1> @eq_vector_undef_icmp
+define <2 x i1> @eq_vector_poison_icmp(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: define <2 x i1> @eq_vector_poison_icmp
 ; CHECK-SAME: (<2 x i8> [[X:%.*]], <2 x i8> [[Y:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i8> [[X]], <i8 -6, i8 -6>
 ; CHECK-NEXT:    [[OR:%.*]] = icmp uge <2 x i8> [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret <2 x i1> [[OR]]
 ;
   %sub = add <2 x i8> %x, <i8 -5, i8 -5>
-  %c1 = icmp eq <2 x i8> %x, <i8 5, i8 undef>
+  %c1 = icmp eq <2 x i8> %x, <i8 5, i8 poison>
   %c2 = icmp ugt <2 x i8> %sub, %y
   %or = or <2 x i1> %c1, %c2
   ret <2 x i1> %or
 }
 
-define <2 x i1> @eq_vector_undef_add(<2 x i8> %x, <2 x i8> %y) {
-; CHECK-LABEL: define <2 x i1> @eq_vector_undef_add
+define <2 x i1> @eq_vector_poison_add(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: define <2 x i1> @eq_vector_poison_add
 ; CHECK-SAME: (<2 x i8> [[X:%.*]], <2 x i8> [[Y:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i8> [[X]], <i8 -6, i8 -6>
 ; CHECK-NEXT:    [[OR:%.*]] = icmp uge <2 x i8> [[TMP1]], [[Y]]
 ; CHECK-NEXT:    ret <2 x i1> [[OR]]
 ;
-  %sub = add <2 x i8> %x, <i8 -5, i8 undef>
+  %sub = add <2 x i8> %x, <i8 -5, i8 poison>
   %c1 = icmp eq <2 x i8> %x, <i8 5, i8 5>
   %c2 = icmp ugt <2 x i8> %sub, %y
   %or = or <2 x i1> %c1, %c2
diff --git a/llvm/test/Transforms/InstCombine/binop-itofp.ll b/llvm/test/Transforms/InstCombine/binop-itofp.ll
index 097a8196af80f..19f7ca9282f29 100644
--- a/llvm/test/Transforms/InstCombine/binop-itofp.ll
+++ b/llvm/test/Transforms/InstCombine/binop-itofp.ll
@@ -1063,8 +1063,8 @@ define float @negzero_check_on_constant_for_si_fmul(i1 %c, i1 %.b, ptr %g_2345)
   ret float %mul3.i.i
 }
 
-define <2 x float> @nonzero_check_on_constant_for_si_fmul_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
-; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_vec_w_undef(
+define <2 x float> @nonzero_check_on_constant_for_si_fmul_vec_w_poison(i1 %c, i1 %.b, ptr %g_2345) {
+; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_vec_w_poison(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
 ; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
@@ -1079,34 +1079,33 @@ define <2 x float> @nonzero_check_on_constant_for_si_fmul_vec_w_undef(i1 %c, i1
   %conv.i.v = insertelement <2 x i16> poison, i16 %conv.i.s, i64 0
   %conv.i = insertelement <2 x i16> %conv.i.v, i16 %conv.i.s, i64 1
   %conv1.i = sitofp <2 x i16> %conv.i to <2 x float>
-  %mul3.i.i = fmul <2 x float> %conv1.i, <float undef, float 0.000000e+00>
+  %mul3.i.i = fmul <2 x float> %conv1.i, <float poison, float 0.000000e+00>
   store i32 %sel, ptr %g_2345, align 4
   ret <2 x float> %mul3.i.i
 }
 
-define <2 x float> @nonzero_check_on_constant_for_si_fmul_nz_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
-; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_nz_vec_w_undef(
+define <2 x float> @nonzero_check_on_constant_for_si_fmul_nz_vec_w_poison(i1 %c, i1 %.b, ptr %g_2345) {
+; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_nz_vec_w_poison(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
 ; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
 ; CHECK-NEXT:    [[CONV_I:%.*]] = shufflevector <2 x i16> [[CONV_I_V]], <2 x i16> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[CONV1_I:%.*]] = sitofp <2 x i16> [[CONV_I]] to <2 x float>
-; CHECK-NEXT:    [[MUL3_I_I:%.*]] = fmul <2 x float> [[CONV1_I]], <float undef, float 1.000000e+00>
 ; CHECK-NEXT:    store i32 [[SEL]], ptr [[G_2345:%.*]], align 4
-; CHECK-NEXT:    ret <2 x float> [[MUL3_I_I]]
+; CHECK-NEXT:    ret <2 x float> [[CONV1_I]]
 ;
   %sel = select i1 %c, i32 65529, i32 53264
   %conv.i.s = trunc i32 %sel to i16
   %conv.i.v = insertelement <2 x i16> poison, i16 %conv.i.s, i64 0
   %conv.i = insertelement <2 x i16> %conv.i.v, i16 %conv.i.s, i64 1
   %conv1.i = sitofp <2 x i16> %conv.i to <2 x float>
-  %mul3.i.i = fmul <2 x float> %conv1.i, <float undef, float 1.000000e+00>
+  %mul3.i.i = fmul <2 x float> %conv1.i, <float poison, float 1.000000e+00>
   store i32 %sel, ptr %g_2345, align 4
   ret <2 x float> %mul3.i.i
 }
 
-define <2 x float> @nonzero_check_on_constant_for_si_fmul_negz_vec_w_undef(i1 %c, i1 %.b, ptr %g_2345) {
-; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_negz_vec_w_undef(
+define <2 x float> @nonzero_check_on_constant_for_si_fmul_negz_vec_w_poison(i1 %c, i1 %.b, ptr %g_2345) {
+; CHECK-LABEL: @nonzero_check_on_constant_for_si_fmul_negz_vec_w_poison(
 ; CHECK-NEXT:    [[SEL:%.*]] = select i1 [[C:%.*]], i32 65529, i32 53264
 ; CHECK-NEXT:    [[CONV_I_S:%.*]] = trunc nuw i32 [[SEL]] to i16
 ; CHECK-NEXT:    [[CONV_I_V:%.*]] = insertelement <2 x i16> poison, i16 [[CONV_I_S]], i64 0
@@ -1122,7 +1121,7 @@ define <2 x float> @nonzero_check_on_constant_for_si_fmul_negz_vec_w_undef(i1 %c
   %conv.i.v = insertelement <2 x i16> poison, i16 %conv.i.s, i64 0
   %conv.i = insertelement <2 x i16> %conv.i.v, i16 %conv.i.s, i64 1
   %conv1.i = sitofp <2 x i16> %conv.i to <2 x float>
-  %mul3.i.i = fmul <2 x float> %conv1.i, <float undef, float -0.000000e+00>
+  %mul3.i.i = fmul <2 x float> %conv1.i, <float poison, float -0.000000e+00>
   store i32 %sel, ptr %g_2345, align 4
   ret <2 x float> %mul3.i.i
 }
diff --git a/llvm/test/Transforms/InstCombine/bswap-fold.ll b/llvm/test/Transforms/InstCombine/bswap-fold.ll
index 19522168beaf5..91674c6017a9e 100644
--- a/llvm/test/Transforms/InstCombine/bswap-fold.ll
+++ b/llvm/test/Transforms/InstCombine/bswap-fold.ll
@@ -870,13 +870,13 @@ define <2 x i64> @bs_active_high_different_negative(<2 x i64> %0) {
 }
 
 ; TODO: This should fold to 'and'.
-define <2 x i64> @bs_active_high_undef(<2 x i64> %0) {
-; CHECK-LABEL: @bs_active_high_undef(
+define <2 x i64> @bs_active_high_poison(<2 x i64> %0) {
+; CHECK-LABEL: @bs_active_high_poison(
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> [[TMP0:%.*]])
-; CHECK-NEXT:    [[TMP3:%.*]] = lshr <2 x i64> [[TMP2]], <i64 56, i64 undef>
+; CHECK-NEXT:    [[TMP3:%.*]] = lshr <2 x i64> [[TMP2]], <i64 56, i64 poison>
 ; CHECK-NEXT:    ret <2 x i64> [[TMP3]]
 ;
-  %2 = shl <2 x i64> %0, <i64 56, i64 undef>
+  %2 = shl <2 x i64> %0, <i64 56, i64 poison>
   %3 = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> %2)
   ret <2 x i64> %3
 }
diff --git a/llvm/test/Transforms/InstCombine/bswap.ll b/llvm/test/Transforms/InstCombine/bswap.ll
index d42583bb5699b..fa61bc841f019 100644
--- a/llvm/test/Transforms/InstCombine/bswap.ll
+++ b/llvm/test/Transforms/InstCombine/bswap.ll
@@ -107,27 +107,27 @@ define <2 x i32> @test2_vector(<2 x i32> %arg) {
   ret <2 x i32> %t14
 }
 
-define <2 x i32> @test2_vector_undef(<2 x i32> %arg) {
-; CHECK-LABEL: @test2_vector_undef(
-; CHECK-NEXT:    [[T2:%.*]] = shl <2 x i32> [[ARG:%.*]], <i32 24, i32 undef>
+define <2 x i32> @test2_vector_poison(<2 x i32> %arg) {
+; CHECK-LABEL: @test2_vector_poison(
+; CHECK-NEXT:    [[T2:%.*]] = shl <2 x i32> [[ARG:%.*]], <i32 24, i32 poison>
 ; CHECK-NEXT:    [[T4:%.*]] = shl <2 x i32> [[ARG]], <i32 8, i32 8>
-; CHECK-NEXT:    [[T5:%.*]] = and <2 x i32> [[T4]], <i32 16711680, i32 undef>
-; CHECK-NEXT:    [[T6:%.*]] = or <2 x i32> [[T2]], [[T5]]
+; CHECK-NEXT:    [[T5:%.*]] = and <2 x i32> [[T4]], <i32 16711680, i32 poison>
+; CHECK-NEXT:    [[T6:%.*]] = or disjoint <2 x i32> [[T2]], [[T5]]
 ; CHECK-NEXT:    [[T8:%.*]] = lshr <2 x i32> [[ARG]], <i32 8, i32 8>
-; CHECK-NEXT:    [[T9:%.*]] = and <2 x i32> [[T8]], <i32 65280, i32 undef>
-; CHECK-NEXT:    [[T10:%.*]] = or <2 x i32> [[T6]], [[T9]]
-; CHECK-NEXT:    [[T12:%.*]] = lshr <2 x i32> [[ARG]], <i32 24, i32 undef>
-; CHECK-NEXT:    [[T14:%.*]] = or <2 x i32> [[T10]], [[T12]]
+; CHECK-NEXT:    [[T9:%.*]] = and <2 x i32> [[T8]], <i32 65280, i32 poison>
+; CHECK-NEXT:    [[T10:%.*]] = or disjoint <2 x i32> [[T6]], [[T9]]
+; CHECK-NEXT:    [[T12:%.*]] = lshr <2 x i32> [[ARG]], <i32 24, i32 poison>
+; CHECK-NEXT:    [[T14:%.*]] = or disjoint <2 x i32> [[T10]], [[T12]]
 ; CHECK-NEXT:    ret <2 x i32> [[T14]]
 ;
-  %t2 = shl <2 x i32> %arg, <i32 24, i32 undef>
+  %t2 = shl <2 x i32> %arg, <i32 24, i32 poison>
   %t4 = shl <2 x i32> %arg, <i32 8, i32 8>
-  %t5 = and <2 x i32> %t4, <i32 16711680, i32 undef>
+  %t5 = and <2 x i32> %t4, <i32 16711680, i32 poison>
   %t6 = or <2 x i32> %t2, %t5
   %t8 = lshr <2 x i32> %arg, <i32 8, i32 8>
-  %t9 = and <2 x i32> %t8, <i32 65280, i32 undef>
+  %t9 = and <2 x i32> %t8, <i32 65280, i32 poison>
   %t10 = or <2 x i32> %t6, %t9
-  %t12 = lshr <2 x i32> %arg, <i32 24, i32 undef>
+  %t12 = lshr <2 x i32> %arg, <i32 24, i32 poison>
   %t14 = or <2 x i32> %t10, %t12
   ret <2 x i32> %t14
 }
@@ -154,13 +154,13 @@ define <2 x i16> @test3_vector(<2 x i16> %s) {
   ret <2 x i16> %t5
 }
 
-define <2 x i16> @test3_vector_undef(<2 x i16> %s) {
-; CHECK-LABEL: @test3_vector_undef(
+define <2 x i16> @test3_vector_poison(<2 x i16> %s) {
+; CHECK-LABEL: @test3_vector_poison(
 ; CHECK-NEXT:    [[T5:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[S:%.*]])
 ; CHECK-NEXT:    ret <2 x i16> [[T5]]
 ;
-  %t2 = lshr <2 x i16> %s, <i16 undef, i16 8>
-  %t4 = shl <2 x i16> %s, <i16 8, i16 undef>
+  %t2 = lshr <2 x i16> %s, <i16 poison, i16 8>
+  %t4 = shl <2 x i16> %s, <i16 8, i16 poison>
   %t5 = or <2 x i16> %t2, %t4
   ret <2 x i16> %t5
 }
@@ -657,7 +657,7 @@ define i32 @shuf_4bytes(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CAST:%.*]] = call i32 @llvm.bswap.i32(i32 [[TMP1]])
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
-  %bswap = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bswap = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %cast = bitcast <4 x i8> %bswap to i32
   ret i32 %cast
 }
@@ -669,7 +669,7 @@ define i32 @shuf_load_4bytes(ptr %p) {
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
   %x = load <4 x i8>, ptr %p
-  %bswap = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 undef, i32 0>
+  %bswap = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 poison, i32 0>
   %cast = bitcast <4 x i8> %bswap to i32
   ret i32 %cast
 }
@@ -680,7 +680,7 @@ define i32 @shuf_bitcast_twice_4bytes(i32 %x) {
 ; CHECK-NEXT:    ret i32 [[CAST2]]
 ;
   %cast1 = bitcast i32 %x to <4 x i8>
-  %bswap = shufflevector <4 x i8> %cast1, <4 x i8> undef, <4 x i32> <i32 undef, i32 2, i32 1, i32 0>
+  %bswap = shufflevector <4 x i8> %cast1, <4 x i8> poison, <4 x i32> <i32 poison, i32 2, i32 1, i32 0>
   %cast2 = bitcast <4 x i8> %bswap to i32
   ret i32 %cast2
 }
@@ -695,7 +695,7 @@ define i32 @shuf_4bytes_extra_use(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
-  %bswap = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  %bswap = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   call void @use(<4 x i8> %bswap)
   %cast = bitcast <4 x i8> %bswap to i32
   ret i32 %cast
@@ -709,7 +709,7 @@ define i128 @shuf_16bytes(<16 x i8> %x) {
 ; CHECK-NEXT:    [[CAST:%.*]] = bitcast <16 x i8> [[BSWAP]] to i128
 ; CHECK-NEXT:    ret i128 [[CAST]]
 ;
-  %bswap = shufflevector <16 x i8> %x, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %bswap = shufflevector <16 x i8> %x, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
   %cast = bitcast <16 x i8> %bswap to i128
   ret i128 %cast
 }
@@ -722,7 +722,7 @@ define i32 @shuf_2bytes_widening(<2 x i8> %x) {
 ; CHECK-NEXT:    [[CAST:%.*]] = bitcast <4 x i8> [[BSWAP]] to i32
 ; CHECK-NEXT:    ret i32 [[CAST]]
 ;
-  %bswap = shufflevector <2 x i8> %x, <2 x i8> undef, <4 x i32> <i32 1, i32 0, i32 undef, i32 undef>
+  %bswap = shufflevector <2 x i8> %x, <2 x i8> poison, <4 x i32> <i32 1, i32 0, i32 poison, i32 poison>
   %cast = bitcast <4 x i8> %bswap to i32
   ret i32 %cast
 }
diff --git a/llvm/test/Transforms/InstCombine/compare-signs.ll b/llvm/test/Transforms/InstCombine/compare-signs.ll
index 3730d46d5f0f4..ede652869212c 100644
--- a/llvm/test/Transforms/InstCombine/compare-signs.ll
+++ b/llvm/test/Transforms/InstCombine/compare-signs.ll
@@ -61,29 +61,29 @@ define <2 x i32> @test3vec(<2 x i32> %a, <2 x i32> %b) nounwind readnone {
   ret <2 x i32> %t3
 }
 
-define <2 x i32> @test3vec_undef1(<2 x i32> %a, <2 x i32> %b) nounwind readnone {
-; CHECK-LABEL: @test3vec_undef1(
+define <2 x i32> @test3vec_poison1(<2 x i32> %a, <2 x i32> %b) nounwind readnone {
+; CHECK-LABEL: @test3vec_poison1(
 ; CHECK-NEXT:    [[T2_UNSHIFTED:%.*]] = xor <2 x i32> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[T2:%.*]] = icmp ult <2 x i32> [[T2_UNSHIFTED]], <i32 16777216, i32 16777216>
 ; CHECK-NEXT:    [[T3:%.*]] = zext <2 x i1> [[T2]] to <2 x i32>
 ; CHECK-NEXT:    ret <2 x i32> [[T3]]
 ;
-  %t0 = lshr <2 x i32> %a, <i32 24, i32 undef>
+  %t0 = lshr <2 x i32> %a, <i32 24, i32 poison>
   %t1 = lshr <2 x i32> %b, <i32 24, i32 24>
   %t2 = icmp eq <2 x i32> %t0, %t1
   %t3 = zext <2 x i1> %t2 to <2 x i32>
   ret <2 x i32> %t3
 }
 
-define <2 x i32> @test3vec_undef2(<2 x i32> %a, <2 x i32> %b) nounwind readnone {
-; CHECK-LABEL: @test3vec_undef2(
+define <2 x i32> @test3vec_poison2(<2 x i32> %a, <2 x i32> %b) nounwind readnone {
+; CHECK-LABEL: @test3vec_poison2(
 ; CHECK-NEXT:    [[T2_UNSHIFTED:%.*]] = xor <2 x i32> [[A:%.*]], [[B:%.*]]
 ; CHECK-NEXT:    [[T2:%.*]] = icmp ult <2 x i32> [[T2_UNSHIFTED]], <i32 131072, i32 131072>
 ; CHECK-NEXT:    [[T3:%.*]] = zext <2 x i1> [[T2]] to <2 x i32>
 ; CHECK-NEXT:    ret <2 x i32> [[T3]]
 ;
-  %t0 = lshr <2 x i32> %a, <i32 undef, i32 17>
-  %t1 = lshr <2 x i32> %b, <i32 undef, i32 17>
+  %t0 = lshr <2 x i32> %a, <i32 poison, i32 17>
+  %t1 = lshr <2 x i32> %b, <i32 poison, i32 17>
   %t2 = icmp eq <2 x i32> %t0, %t1
   %t3 = zext <2 x i1> %t2 to <2 x i32>
   ret <2 x i32> %t3
diff --git a/llvm/test/Transforms/InstCombine/fcmp-range-check-idiom.ll b/llvm/test/Transforms/InstCombine/fcmp-range-check-idiom.ll
index 0893b27f5cf49..10a3ccf3cdb48 100644
--- a/llvm/test/Transforms/InstCombine/fcmp-range-check-idiom.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp-range-check-idiom.ll
@@ -182,15 +182,15 @@ define i1 @test_and_olt_logical(float %x) {
   ret i1 %cond
 }
 
-define <2 x i1> @test_and_olt_undef(<2 x float> %x) {
-; CHECK-LABEL: define <2 x i1> @test_and_olt_undef(
+define <2 x i1> @test_and_olt_poison(<2 x float> %x) {
+; CHECK-LABEL: define <2 x i1> @test_and_olt_poison(
 ; CHECK-SAME: <2 x float> [[X:%.*]]) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x float> @llvm.fabs.v2f32(<2 x float> [[X]])
 ; CHECK-NEXT:    [[COND:%.*]] = fcmp olt <2 x float> [[TMP1]], <float 0x3C00000000000000, float 0x3C00000000000000>
 ; CHECK-NEXT:    ret <2 x i1> [[COND]]
 ;
-  %cmp1 = fcmp olt <2 x float> %x, <float 0x3C00000000000000, float undef>
-  %cmp2 = fcmp ogt <2 x float> %x, <float 0xBC00000000000000, float undef>
+  %cmp1 = fcmp olt <2 x float> %x, <float 0x3C00000000000000, float poison>
+  %cmp2 = fcmp ogt <2 x float> %x, <float 0xBC00000000000000, float poison>
   %cond = and <2 x i1> %cmp1, %cmp2
   ret <2 x i1> %cond
 }
diff --git a/llvm/test/Transforms/InstCombine/icmp-fsh.ll b/llvm/test/Transforms/InstCombine/icmp-fsh.ll
index 8fad3f94df13f..8154f312dc5d4 100644
--- a/llvm/test/Transforms/InstCombine/icmp-fsh.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-fsh.ll
@@ -53,49 +53,49 @@ define <2 x i1> @rotl_ne_n1(<2 x i5> %x, <2 x i5> %y) {
   ret <2 x i1> %r
 }
 
-define <2 x i1> @rotl_ne_n1_undef(<2 x i5> %x, <2 x i5> %y) {
-; CHECK-LABEL: @rotl_ne_n1_undef(
-; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i5> [[X:%.*]], <i5 -1, i5 undef>
+define <2 x i1> @rotl_ne_n1_poison(<2 x i5> %x, <2 x i5> %y) {
+; CHECK-LABEL: @rotl_ne_n1_poison(
+; CHECK-NEXT:    [[R:%.*]] = icmp ne <2 x i5> [[X:%.*]], <i5 -1, i5 poison>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %rot = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5>%x, <2 x i5> %x, <2 x i5> %y)
-  %r = icmp ne <2 x i5> %rot, <i5 -1, i5 undef>
+  %r = icmp ne <2 x i5> %rot, <i5 -1, i5 poison>
   ret <2 x i1> %r
 }
 
-define <2 x i1> @rotl_eq_0_undef(<2 x i5> %x, <2 x i5> %y) {
-; CHECK-LABEL: @rotl_eq_0_undef(
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i5> [[X:%.*]], <i5 0, i5 undef>
+define <2 x i1> @rotl_eq_0_poison(<2 x i5> %x, <2 x i5> %y) {
+; CHECK-LABEL: @rotl_eq_0_poison(
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i5> [[X:%.*]], <i5 0, i5 poison>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %rot = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5>%x, <2 x i5> %x, <2 x i5> %y)
-  %r = icmp eq <2 x i5> %rot, <i5 0, i5 undef>
+  %r = icmp eq <2 x i5> %rot, <i5 0, i5 poison>
   ret <2 x i1> %r
 }
 
 ; negative test - wrong constant value
 
-define <2 x i1> @rotl_eq_1_undef(<2 x i5> %x, <2 x i5> %y) {
-; CHECK-LABEL: @rotl_eq_1_undef(
+define <2 x i1> @rotl_eq_1_poison(<2 x i5> %x, <2 x i5> %y) {
+; CHECK-LABEL: @rotl_eq_1_poison(
 ; CHECK-NEXT:    [[ROT:%.*]] = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5> [[X:%.*]], <2 x i5> [[X]], <2 x i5> [[Y:%.*]])
-; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i5> [[ROT]], <i5 undef, i5 1>
+; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i5> [[ROT]], <i5 poison, i5 1>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %rot = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5>%x, <2 x i5> %x, <2 x i5> %y)
-  %r = icmp eq <2 x i5> %rot, <i5 undef, i5 1>
+  %r = icmp eq <2 x i5> %rot, <i5 poison, i5 1>
   ret <2 x i1> %r
 }
 
 ; negative test - wrong predicate
 
-define <2 x i1> @rotl_sgt_0_undef(<2 x i5> %x, <2 x i5> %y) {
-; CHECK-LABEL: @rotl_sgt_0_undef(
+define <2 x i1> @rotl_sgt_0_poison(<2 x i5> %x, <2 x i5> %y) {
+; CHECK-LABEL: @rotl_sgt_0_poison(
 ; CHECK-NEXT:    [[ROT:%.*]] = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5> [[X:%.*]], <2 x i5> [[X]], <2 x i5> [[Y:%.*]])
-; CHECK-NEXT:    [[R:%.*]] = icmp sgt <2 x i5> [[ROT]], <i5 0, i5 undef>
+; CHECK-NEXT:    [[R:%.*]] = icmp sgt <2 x i5> [[ROT]], <i5 0, i5 poison>
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
   %rot = tail call <2 x i5> @llvm.fshl.v2i5(<2 x i5>%x, <2 x i5> %x, <2 x i5> %y)
-  %r = icmp sgt <2 x i5> %rot, <i5 0, i5 undef>
+  %r = icmp sgt <2 x i5> %rot, <i5 0, i5 poison>
   ret <2 x i1> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/icmp-power2-and-icmp-shifted-mask.ll b/llvm/test/Transforms/InstCombine/icmp-power2-and-icmp-shifted-mask.ll
index 82fcca07a00ac..a151fd96ef498 100644
--- a/llvm/test/Transforms/InstCombine/icmp-power2-and-icmp-shifted-mask.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-power2-and-icmp-shifted-mask.ll
@@ -448,7 +448,10 @@ define <2 x i1> @icmp_power2_and_icmp_shifted_mask_vector_swapped_splat_poison_2
 ; Vector of 2 reduction with splat containing undef
 define <2 x i1> @icmp_power2_and_icmp_shifted_mask_vector_splat_undef_2147483648_1610612736_2147483647(<2 x i32> %x) {
 ; CHECK-LABEL: @icmp_power2_and_icmp_shifted_mask_vector_splat_undef_2147483648_1610612736_2147483647(
-; CHECK-NEXT:    [[T4:%.*]] = icmp ult <2 x i32> [[X:%.*]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T1:%.*]] = icmp ult <2 x i32> [[X:%.*]], <i32 -2147483648, i32 undef>
+; CHECK-NEXT:    [[T2:%.*]] = and <2 x i32> [[X]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ne <2 x i32> [[T2]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T4:%.*]] = and <2 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <2 x i1> [[T4]]
 ;
   %t1 = icmp ult <2 x i32> %x, <i32 2147483648, i32 undef>
@@ -460,7 +463,10 @@ define <2 x i1> @icmp_power2_and_icmp_shifted_mask_vector_splat_undef_2147483648
 
 define <2 x i1> @icmp_power2_and_icmp_shifted_mask_vector_swapped_splat_undef_2147483648_1610612736_2147483647(<2 x i32> %x) {
 ; CHECK-LABEL: @icmp_power2_and_icmp_shifted_mask_vector_swapped_splat_undef_2147483648_1610612736_2147483647(
-; CHECK-NEXT:    [[T4:%.*]] = icmp ult <2 x i32> [[X:%.*]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T1:%.*]] = icmp ult <2 x i32> [[X:%.*]], <i32 -2147483648, i32 undef>
+; CHECK-NEXT:    [[T2:%.*]] = and <2 x i32> [[X]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ne <2 x i32> [[T2]], <i32 1610612736, i32 2147483647>
+; CHECK-NEXT:    [[T4:%.*]] = and <2 x i1> [[T3]], [[T1]]
 ; CHECK-NEXT:    ret <2 x i1> [[T4]]
 ;
   %t1 = icmp ult <2 x i32> %x, <i32 2147483648, i32 undef>
diff --git a/llvm/test/Transforms/InstCombine/icmp-vec-inseltpoison.ll b/llvm/test/Transforms/InstCombine/icmp-vec-inseltpoison.ll
index 04d8d294c3193..677655a0a99ab 100644
--- a/llvm/test/Transforms/InstCombine/icmp-vec-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-vec-inseltpoison.ll
@@ -184,7 +184,7 @@ define <3 x i1> @PR27756_2(<3 x i8> %a) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp slt <3 x i8> [[A:%.*]], <i8 43, i8 43, i8 1>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
-  %cmp = icmp sle <3 x i8> %a, <i8 42, i8 undef, i8 0>
+  %cmp = icmp sle <3 x i8> %a, <i8 42, i8 poison, i8 0>
   ret <3 x i1> %cmp
 }
 
@@ -193,7 +193,7 @@ define <3 x i1> @PR27756_3(<3 x i8> %a) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <3 x i8> [[A:%.*]], <i8 0, i8 0, i8 41>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
-  %cmp = icmp sge <3 x i8> %a, <i8 undef, i8 1, i8 42>
+  %cmp = icmp sge <3 x i8> %a, <i8 poison, i8 1, i8 42>
   ret <3 x i1> %cmp
 }
 
@@ -300,14 +300,14 @@ define <4 x i1> @splat_icmp(<4 x i8> %x) {
   ret <4 x i1> %cmp
 }
 
-define <4 x i1> @splat_icmp_undef(<4 x i8> %x) {
-; CHECK-LABEL: @splat_icmp_undef(
+define <4 x i1> @splat_icmp_poison(<4 x i8> %x) {
+; CHECK-LABEL: @splat_icmp_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i8> [[X:%.*]], <i8 42, i8 42, i8 42, i8 42>
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 undef, i32 undef, i32 2>
-  %cmp = icmp ult <4 x i8> %splatx, <i8 undef, i8 42, i8 undef, i8 42>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 poison, i32 poison, i32 2>
+  %cmp = icmp ult <4 x i8> %splatx, <i8 poison, i8 42, i8 poison, i8 42>
   ret <4 x i1> %cmp
 }
 
@@ -317,8 +317,8 @@ define <4 x i1> @splat_icmp_larger_size(<2 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <2 x i1> [[TMP1]], <2 x i1> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <2 x i8> %x, <2 x i8> poison, <4 x i32> <i32 1, i32 undef, i32 1, i32 undef>
-  %cmp = icmp eq <4 x i8> %splatx, <i8 42, i8 42, i8 undef, i8 42>
+  %splatx = shufflevector <2 x i8> %x, <2 x i8> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 poison>
+  %cmp = icmp eq <4 x i8> %splatx, <i8 42, i8 42, i8 poison, i8 42>
   ret <4 x i1> %cmp
 }
 
@@ -328,8 +328,8 @@ define <4 x i1> @splat_fcmp_smaller_size(<5 x float> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <5 x i1> [[TMP1]], <5 x i1> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <5 x float> %x, <5 x float> poison, <4 x i32> <i32 1, i32 undef, i32 1, i32 undef>
-  %cmp = fcmp oeq <4 x float> %splatx, <float 42.0, float 42.0, float undef, float 42.0>
+  %splatx = shufflevector <5 x float> %x, <5 x float> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 poison>
+  %cmp = fcmp oeq <4 x float> %splatx, <float 42.0, float 42.0, float poison, float 42.0>
   ret <4 x i1> %cmp
 }
 
diff --git a/llvm/test/Transforms/InstCombine/icmp-vec.ll b/llvm/test/Transforms/InstCombine/icmp-vec.ll
index 63dd535c527ef..1e392635936b7 100644
--- a/llvm/test/Transforms/InstCombine/icmp-vec.ll
+++ b/llvm/test/Transforms/InstCombine/icmp-vec.ll
@@ -186,7 +186,7 @@ define <3 x i1> @PR27756_2(<3 x i8> %a) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp slt <3 x i8> [[A:%.*]], <i8 43, i8 43, i8 1>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
-  %cmp = icmp sle <3 x i8> %a, <i8 42, i8 undef, i8 0>
+  %cmp = icmp sle <3 x i8> %a, <i8 42, i8 poison, i8 0>
   ret <3 x i1> %cmp
 }
 
@@ -195,7 +195,7 @@ define <3 x i1> @PR27756_3(<3 x i8> %a) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <3 x i8> [[A:%.*]], <i8 0, i8 0, i8 41>
 ; CHECK-NEXT:    ret <3 x i1> [[CMP]]
 ;
-  %cmp = icmp sge <3 x i8> %a, <i8 undef, i8 1, i8 42>
+  %cmp = icmp sge <3 x i8> %a, <i8 poison, i8 1, i8 42>
   ret <3 x i1> %cmp
 }
 
@@ -218,8 +218,8 @@ define <4 x i1> @same_shuffle_inputs_icmp(<4 x i8> %x, <4 x i8> %y) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 2, i32 0>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %shufx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> < i32 3, i32 3, i32 2, i32 0 >
-  %shufy = shufflevector <4 x i8> %y, <4 x i8> undef, <4 x i32> < i32 3, i32 3, i32 2, i32 0 >
+  %shufx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> < i32 3, i32 3, i32 2, i32 0 >
+  %shufy = shufflevector <4 x i8> %y, <4 x i8> poison, <4 x i32> < i32 3, i32 3, i32 2, i32 0 >
   %cmp = icmp sgt <4 x i8> %shufx, %shufy
   ret <4 x i1> %cmp
 }
@@ -232,8 +232,8 @@ define <5 x i1> @same_shuffle_inputs_fcmp(<4 x float> %x, <4 x float> %y) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <5 x i32> <i32 0, i32 1, i32 3, i32 2, i32 0>
 ; CHECK-NEXT:    ret <5 x i1> [[CMP]]
 ;
-  %shufx = shufflevector <4 x float> %x, <4 x float> undef, <5 x i32> < i32 0, i32 1, i32 3, i32 2, i32 0 >
-  %shufy = shufflevector <4 x float> %y, <4 x float> undef, <5 x i32> < i32 0, i32 1, i32 3, i32 2, i32 0 >
+  %shufx = shufflevector <4 x float> %x, <4 x float> poison, <5 x i32> < i32 0, i32 1, i32 3, i32 2, i32 0 >
+  %shufy = shufflevector <4 x float> %y, <4 x float> poison, <5 x i32> < i32 0, i32 1, i32 3, i32 2, i32 0 >
   %cmp = fcmp oeq <5 x float> %shufx, %shufy
   ret <5 x i1> %cmp
 }
@@ -248,8 +248,8 @@ define <4 x i1> @same_shuffle_inputs_icmp_extra_use1(<4 x i8> %x, <4 x i8> %y) {
 ; CHECK-NEXT:    call void @use_v4i8(<4 x i8> [[SHUFX]])
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %shufx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> < i32 3, i32 3, i32 3, i32 3 >
-  %shufy = shufflevector <4 x i8> %y, <4 x i8> undef, <4 x i32> < i32 3, i32 3, i32 3, i32 3 >
+  %shufx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> < i32 3, i32 3, i32 3, i32 3 >
+  %shufy = shufflevector <4 x i8> %y, <4 x i8> poison, <4 x i32> < i32 3, i32 3, i32 3, i32 3 >
   %cmp = icmp ugt <4 x i8> %shufx, %shufy
   call void @use_v4i8(<4 x i8> %shufx)
   ret <4 x i1> %cmp
@@ -265,8 +265,8 @@ define <2 x i1> @same_shuffle_inputs_icmp_extra_use2(<4 x i8> %x, <4 x i8> %y) {
 ; CHECK-NEXT:    call void @use_v2i8(<2 x i8> [[SHUFY]])
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %shufx = shufflevector <4 x i8> %x, <4 x i8> undef, <2 x i32> < i32 3, i32 2 >
-  %shufy = shufflevector <4 x i8> %y, <4 x i8> undef, <2 x i32> < i32 3, i32 2 >
+  %shufx = shufflevector <4 x i8> %x, <4 x i8> poison, <2 x i32> < i32 3, i32 2 >
+  %shufy = shufflevector <4 x i8> %y, <4 x i8> poison, <2 x i32> < i32 3, i32 2 >
   %cmp = icmp eq <2 x i8> %shufx, %shufy
   call void @use_v2i8(<2 x i8> %shufy)
   ret <2 x i1> %cmp
@@ -283,8 +283,8 @@ define <2 x i1> @same_shuffle_inputs_icmp_extra_use3(<4 x i8> %x, <4 x i8> %y) {
 ; CHECK-NEXT:    call void @use_v2i8(<2 x i8> [[SHUFY]])
 ; CHECK-NEXT:    ret <2 x i1> [[CMP]]
 ;
-  %shufx = shufflevector <4 x i8> %x, <4 x i8> undef, <2 x i32> < i32 0, i32 0 >
-  %shufy = shufflevector <4 x i8> %y, <4 x i8> undef, <2 x i32> < i32 0, i32 0 >
+  %shufx = shufflevector <4 x i8> %x, <4 x i8> poison, <2 x i32> < i32 0, i32 0 >
+  %shufy = shufflevector <4 x i8> %y, <4 x i8> poison, <2 x i32> < i32 0, i32 0 >
   %cmp = icmp eq <2 x i8> %shufx, %shufy
   call void @use_v2i8(<2 x i8> %shufx)
   call void @use_v2i8(<2 x i8> %shufy)
@@ -297,19 +297,19 @@ define <4 x i1> @splat_icmp(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
   %cmp = icmp sgt <4 x i8> %splatx, <i8 42, i8 42, i8 42, i8 42>
   ret <4 x i1> %cmp
 }
 
-define <4 x i1> @splat_icmp_undef(<4 x i8> %x) {
-; CHECK-LABEL: @splat_icmp_undef(
+define <4 x i1> @splat_icmp_poison(<4 x i8> %x) {
+; CHECK-LABEL: @splat_icmp_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i8> [[X:%.*]], <i8 42, i8 42, i8 42, i8 42>
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 2, i32 undef, i32 undef, i32 2>
-  %cmp = icmp ult <4 x i8> %splatx, <i8 undef, i8 42, i8 undef, i8 42>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 poison, i32 poison, i32 2>
+  %cmp = icmp ult <4 x i8> %splatx, <i8 poison, i8 42, i8 poison, i8 42>
   ret <4 x i1> %cmp
 }
 
@@ -319,8 +319,8 @@ define <4 x i1> @splat_icmp_larger_size(<2 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <2 x i1> [[TMP1]], <2 x i1> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <2 x i8> %x, <2 x i8> undef, <4 x i32> <i32 1, i32 undef, i32 1, i32 undef>
-  %cmp = icmp eq <4 x i8> %splatx, <i8 42, i8 42, i8 undef, i8 42>
+  %splatx = shufflevector <2 x i8> %x, <2 x i8> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 poison>
+  %cmp = icmp eq <4 x i8> %splatx, <i8 42, i8 42, i8 poison, i8 42>
   ret <4 x i1> %cmp
 }
 
@@ -330,8 +330,8 @@ define <4 x i1> @splat_fcmp_smaller_size(<5 x float> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = shufflevector <5 x i1> [[TMP1]], <5 x i1> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <5 x float> %x, <5 x float> undef, <4 x i32> <i32 1, i32 undef, i32 1, i32 undef>
-  %cmp = fcmp oeq <4 x float> %splatx, <float 42.0, float 42.0, float undef, float 42.0>
+  %splatx = shufflevector <5 x float> %x, <5 x float> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 poison>
+  %cmp = fcmp oeq <4 x float> %splatx, <float 42.0, float 42.0, float poison, float 42.0>
   ret <4 x i1> %cmp
 }
 
@@ -344,7 +344,7 @@ define <4 x i1> @splat_icmp_extra_use(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <4 x i8> [[SPLATX]], <i8 42, i8 42, i8 42, i8 42>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
   call void @use_v4i8(<4 x i8> %splatx)
   %cmp = icmp sgt <4 x i8> %splatx, <i8 42, i8 42, i8 42, i8 42>
   ret <4 x i1> %cmp
@@ -358,7 +358,7 @@ define <4 x i1> @not_splat_icmp(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <4 x i8> [[SPLATX]], <i8 42, i8 42, i8 42, i8 42>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 3, i32 3>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 3, i32 3>
   %cmp = icmp sgt <4 x i8> %splatx, <i8 42, i8 42, i8 42, i8 42>
   ret <4 x i1> %cmp
 }
@@ -371,7 +371,7 @@ define <4 x i1> @not_splat_icmp2(<4 x i8> %x) {
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt <4 x i8> [[SPLATX]], <i8 43, i8 42, i8 42, i8 42>
 ; CHECK-NEXT:    ret <4 x i1> [[CMP]]
 ;
-  %splatx = shufflevector <4 x i8> %x, <4 x i8> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+  %splatx = shufflevector <4 x i8> %x, <4 x i8> poison, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
   %cmp = icmp sgt <4 x i8> %splatx, <i8 43, i8 42, i8 42, i8 42>
   ret <4 x i1> %cmp
 }
@@ -385,7 +385,7 @@ define <2 x i1> @icmp_logical_or_vec(<2 x i64> %x, <2 x i64> %y, <2 x i1> %false
 ; CHECK-NEXT:    ret <2 x i1> [[SEL]]
 ;
   %cmp.ne = icmp ne <2 x i64> %x, zeroinitializer
-  %sel = select <2 x i1> %cmp.ne, <2 x i1> shufflevector (<2 x i1> insertelement (<2 x i1> undef, i1 true, i32 0), <2 x i1> undef, <2 x i32> zeroinitializer), <2 x i1> %falseval
+  %sel = select <2 x i1> %cmp.ne, <2 x i1> shufflevector (<2 x i1> insertelement (<2 x i1> poison, i1 true, i32 0), <2 x i1> poison, <2 x i32> zeroinitializer), <2 x i1> %falseval
   ret <2 x i1> %sel
 }
 
diff --git a/llvm/test/Transforms/InstCombine/low-bit-splat.ll b/llvm/test/Transforms/InstCombine/low-bit-splat.ll
index 5a4b00a6f17e7..a86ab0359afd7 100644
--- a/llvm/test/Transforms/InstCombine/low-bit-splat.ll
+++ b/llvm/test/Transforms/InstCombine/low-bit-splat.ll
@@ -39,34 +39,37 @@ define <2 x i8> @t2_vec(<2 x i8> %x) {
   %r = ashr <2 x i8> %i0, <i8 7, i8 7>
   ret <2 x i8> %r
 }
-define <3 x i8> @t3_vec_undef0(<3 x i8> %x) {
-; CHECK-LABEL: @t3_vec_undef0(
+
+; TODO: The result constants should contain poison instead of undef.
+
+define <3 x i8> @t3_vec_poison0(<3 x i8> %x) {
+; CHECK-LABEL: @t3_vec_poison0(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i8> [[X:%.*]], <i8 1, i8 undef, i8 1>
 ; CHECK-NEXT:    [[R:%.*]] = sub <3 x i8> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[R]]
 ;
-  %i0 = shl <3 x i8> %x, <i8 7, i8 undef, i8 7>
+  %i0 = shl <3 x i8> %x, <i8 7, i8 poison, i8 7>
   %r = ashr <3 x i8> %i0, <i8 7, i8 7, i8 7>
   ret <3 x i8> %r
 }
-define <3 x i8> @t4_vec_undef1(<3 x i8> %x) {
-; CHECK-LABEL: @t4_vec_undef1(
+define <3 x i8> @t4_vec_poison1(<3 x i8> %x) {
+; CHECK-LABEL: @t4_vec_poison1(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i8> [[X:%.*]], <i8 1, i8 undef, i8 1>
 ; CHECK-NEXT:    [[R:%.*]] = sub <3 x i8> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[R]]
 ;
   %i0 = shl <3 x i8> %x, <i8 7, i8 7, i8 7>
-  %r = ashr <3 x i8> %i0, <i8 7, i8 undef, i8 7>
+  %r = ashr <3 x i8> %i0, <i8 7, i8 poison, i8 7>
   ret <3 x i8> %r
 }
-define <3 x i8> @t5_vec_undef2(<3 x i8> %x) {
-; CHECK-LABEL: @t5_vec_undef2(
+define <3 x i8> @t5_vec_poison2(<3 x i8> %x) {
+; CHECK-LABEL: @t5_vec_poison2(
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <3 x i8> [[X:%.*]], <i8 1, i8 undef, i8 1>
 ; CHECK-NEXT:    [[R:%.*]] = sub <3 x i8> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    ret <3 x i8> [[R]]
 ;
-  %i0 = shl <3 x i8> %x, <i8 7, i8 undef, i8 7>
-  %r = ashr <3 x i8> %i0, <i8 7, i8 undef, i8 7>
+  %i0 = shl <3 x i8> %x, <i8 7, i8 poison, i8 7>
+  %r = ashr <3 x i8> %i0, <i8 7, i8 poison, i8 7>
   ret <3 x i8> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll b/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
index 8e7491ee40370..a77ad3a7ea74e 100644
--- a/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
+++ b/llvm/test/Transforms/InstCombine/lshr-trunc-sext-to-ashr-sext.ll
@@ -74,13 +74,13 @@ define <2 x i16> @t4_vec_splat(<2 x i8> %x) {
   ret <2 x i16> %c
 }
 
-define <2 x i16> @t5_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @t5_vec_undef(
+define <2 x i16> @t5_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @t5_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i8> [[X:%.*]], <i8 4, i8 4>
 ; CHECK-NEXT:    [[C:%.*]] = sext <2 x i8> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[C]]
 ;
-  %a = lshr <2 x i8> %x, <i8 4, i8 undef>
+  %a = lshr <2 x i8> %x, <i8 4, i8 poison>
   %b = trunc <2 x i8> %a to <2 x i4>
   %c = sext <2 x i4> %b to <2 x i16>
   ret <2 x i16> %c
@@ -105,15 +105,15 @@ define i16 @t6_extrause0(i8 %x) {
 
 ; TODO: We could convert %a to ashr and eliminate 2nd use of %b.
 
-define <2 x i16> @t7_extrause0_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @t7_extrause0_vec_undef(
-; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 undef>
-; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i8> [[A]] to <2 x i4>
+define <2 x i16> @t7_extrause0_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @t7_extrause0_vec_poison(
+; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 poison>
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw <2 x i8> [[A]] to <2 x i4>
 ; CHECK-NEXT:    call void @usevec4(<2 x i4> [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext <2 x i4> [[B]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[C]]
 ;
-  %a = lshr <2 x i8> %x, <i8 4, i8 undef>
+  %a = lshr <2 x i8> %x, <i8 4, i8 poison>
   %b = trunc <2 x i8> %a to <2 x i4>
   call void @usevec4(<2 x i4> %b)
   %c = sext <2 x i4> %b to <2 x i16>
@@ -139,15 +139,15 @@ define i16 @t8_extrause1(i8 %x) {
 
 ; TODO: We could convert %a to ashr + mask (and) and eliminate %b.
 
-define <2 x i16> @t9_extrause1_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @t9_extrause1_vec_undef(
-; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 undef>
+define <2 x i16> @t9_extrause1_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @t9_extrause1_vec_poison(
+; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 poison>
 ; CHECK-NEXT:    call void @usevec8(<2 x i8> [[A]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = ashr <2 x i8> [[X]], <i8 4, i8 4>
 ; CHECK-NEXT:    [[C:%.*]] = sext <2 x i8> [[TMP1]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[C]]
 ;
-  %a = lshr <2 x i8> %x, <i8 4, i8 undef>
+  %a = lshr <2 x i8> %x, <i8 4, i8 poison>
   call void @usevec8(<2 x i8> %a)
   %b = trunc <2 x i8> %a to <2 x i4>
   %c = sext <2 x i4> %b to <2 x i16>
@@ -169,16 +169,16 @@ define i16 @t10_extrause2(i8 %x) {
   %c = sext i4 %b to i16
   ret i16 %c
 }
-define <2 x i16> @t11_extrause2_vec_undef(<2 x i8> %x) {
-; CHECK-LABEL: @t11_extrause2_vec_undef(
-; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 undef>
+define <2 x i16> @t11_extrause2_vec_poison(<2 x i8> %x) {
+; CHECK-LABEL: @t11_extrause2_vec_poison(
+; CHECK-NEXT:    [[A:%.*]] = lshr <2 x i8> [[X:%.*]], <i8 4, i8 poison>
 ; CHECK-NEXT:    call void @usevec8(<2 x i8> [[A]])
-; CHECK-NEXT:    [[B:%.*]] = trunc <2 x i8> [[A]] to <2 x i4>
+; CHECK-NEXT:    [[B:%.*]] = trunc nuw <2 x i8> [[A]] to <2 x i4>
 ; CHECK-NEXT:    call void @usevec4(<2 x i4> [[B]])
 ; CHECK-NEXT:    [[C:%.*]] = sext <2 x i4> [[B]] to <2 x i16>
 ; CHECK-NEXT:    ret <2 x i16> [[C]]
 ;
-  %a = lshr <2 x i8> %x, <i8 4, i8 undef>
+  %a = lshr <2 x i8> %x, <i8 4, i8 poison>
   call void @usevec8(<2 x i8> %a)
   %b = trunc <2 x i8> %a to <2 x i4>
   call void @usevec4(<2 x i4> %b)
diff --git a/llvm/test/Transforms/InstCombine/select.ll b/llvm/test/Transforms/InstCombine/select.ll
index 8654691c6f875..8776ec6a29b43 100644
--- a/llvm/test/Transforms/InstCombine/select.ll
+++ b/llvm/test/Transforms/InstCombine/select.ll
@@ -1665,37 +1665,37 @@ define float @copysign3(float %x) {
   ret float %r
 }
 
-define <2 x float> @copysign_vec_undef(<2 x float> %x) {
-; CHECK-LABEL: @copysign_vec_undef(
+define <2 x float> @copysign_vec_poison(<2 x float> %x) {
+; CHECK-LABEL: @copysign_vec_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = fneg <2 x float> [[X:%.*]]
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x float> @llvm.copysign.v2f32(<2 x float> <float 4.200000e+01, float 4.200000e+01>, <2 x float> [[TMP1]])
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %i = bitcast <2 x float> %x to <2 x i32>
   %isneg = icmp ugt <2 x i32> %i, <i32 2147483647, i32 2147483647>
-  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float 42.0, float undef>, <2 x float> <float -42.0, float -42.0>
+  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float 42.0, float poison>, <2 x float> <float -42.0, float -42.0>
   ret <2 x float> %r
 }
 
-define <2 x float> @copysign_vec_undef1(<2 x float> %x) {
-; CHECK-LABEL: @copysign_vec_undef1(
+define <2 x float> @copysign_vec_poison1(<2 x float> %x) {
+; CHECK-LABEL: @copysign_vec_poison1(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x float> @llvm.copysign.v2f32(<2 x float> <float 4.200000e+01, float 4.200000e+01>, <2 x float> [[X:%.*]])
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %i = bitcast <2 x float> %x to <2 x i32>
   %isneg = icmp ult <2 x i32> %i, <i32 2147483648, i32 2147483648>
-  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float 42.0, float 42.0>, <2 x float> <float undef, float -42.0>
+  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float 42.0, float 42.0>, <2 x float> <float poison, float -42.0>
   ret <2 x float> %r
 }
 
-define <2 x float> @copysign_vec_undef3(<2 x float> %x) {
-; CHECK-LABEL: @copysign_vec_undef3(
+define <2 x float> @copysign_vec_poison3(<2 x float> %x) {
+; CHECK-LABEL: @copysign_vec_poison3(
 ; CHECK-NEXT:    [[R:%.*]] = call <2 x float> @llvm.copysign.v2f32(<2 x float> <float 4.200000e+01, float 4.200000e+01>, <2 x float> [[X:%.*]])
 ; CHECK-NEXT:    ret <2 x float> [[R]]
 ;
   %i = bitcast <2 x float> %x to <2 x i32>
   %isneg = icmp ugt <2 x i32> %i, <i32 2147483647, i32 2147483647>
-  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float -42.0, float undef>, <2 x float> <float +42.0, float undef>
+  %r = select arcp nnan <2 x i1> %isneg, <2 x float> <float -42.0, float poison>, <2 x float> <float +42.0, float poison>
   ret <2 x float> %r
 }
 
diff --git a/llvm/test/Transforms/InstCombine/signed-truncation-check.ll b/llvm/test/Transforms/InstCombine/signed-truncation-check.ll
index 208e166b2c876..17cd536de91f4 100644
--- a/llvm/test/Transforms/InstCombine/signed-truncation-check.ll
+++ b/llvm/test/Transforms/InstCombine/signed-truncation-check.ll
@@ -198,104 +198,104 @@ define <2 x i1> @positive_vec_nonsplat(<2 x i32> %arg) {
   ret <2 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef0(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef0(
+define <3 x i1> @positive_vec_poison0(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison0(
 ; CHECK-NEXT:    [[T4_SIMPLIFIED:%.*]] = icmp ult <3 x i32> [[ARG:%.*]], <i32 128, i32 128, i32 128>
 ; CHECK-NEXT:    ret <3 x i1> [[T4_SIMPLIFIED]]
 ;
-  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 undef, i32 -1>
+  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 poison, i32 -1>
   %t2 = add <3 x i32> %arg, <i32 128, i32 128, i32 128>
   %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 256, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef1(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef1(
+define <3 x i1> @positive_vec_poison1(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison1(
 ; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 -1, i32 -1>
-; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 undef, i32 128>
+; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 poison, i32 128>
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 256, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
   %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 -1, i32 -1>
-  %t2 = add <3 x i32> %arg, <i32 128, i32 undef, i32 128>
+  %t2 = add <3 x i32> %arg, <i32 128, i32 poison, i32 128>
   %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 256, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef2(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef2(
+define <3 x i1> @positive_vec_poison2(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison2(
 ; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 -1, i32 -1>
 ; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 128, i32 128>
-; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 undef, i32 256>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 poison, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
   %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 -1, i32 -1>
   %t2 = add <3 x i32> %arg, <i32 128, i32 128, i32 128>
-  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 undef, i32 256>
+  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 poison, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef3(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef3(
-; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 undef, i32 -1>
-; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 undef, i32 128>
+define <3 x i1> @positive_vec_poison3(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison3(
+; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 poison, i32 128>
 ; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 256, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
-  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 undef, i32 -1>
-  %t2 = add <3 x i32> %arg, <i32 128, i32 undef, i32 128>
+  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 poison, i32 -1>
+  %t2 = add <3 x i32> %arg, <i32 128, i32 poison, i32 128>
   %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 256, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef4(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef4(
-; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 undef, i32 -1>
+define <3 x i1> @positive_vec_poison4(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison4(
+; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 poison, i32 -1>
 ; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 128, i32 128>
-; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 undef, i32 256>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 poison, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
-  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 undef, i32 -1>
+  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 poison, i32 -1>
   %t2 = add <3 x i32> %arg, <i32 128, i32 128, i32 128>
-  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 undef, i32 256>
+  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 poison, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef5(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef5(
+define <3 x i1> @positive_vec_poison5(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison5(
 ; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 -1, i32 -1>
-; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 undef, i32 128>
-; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 undef, i32 256>
+; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 poison, i32 128>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 poison, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
   %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 -1, i32 -1>
-  %t2 = add <3 x i32> %arg, <i32 128, i32 undef, i32 128>
-  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 undef, i32 256>
+  %t2 = add <3 x i32> %arg, <i32 128, i32 poison, i32 128>
+  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 poison, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
 
-define <3 x i1> @positive_vec_undef6(<3 x i32> %arg) {
-; CHECK-LABEL: @positive_vec_undef6(
-; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 undef, i32 -1>
-; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 undef, i32 128>
-; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 undef, i32 256>
+define <3 x i1> @positive_vec_poison6(<3 x i32> %arg) {
+; CHECK-LABEL: @positive_vec_poison6(
+; CHECK-NEXT:    [[T1:%.*]] = icmp sgt <3 x i32> [[ARG:%.*]], <i32 -1, i32 poison, i32 -1>
+; CHECK-NEXT:    [[T2:%.*]] = add <3 x i32> [[ARG]], <i32 128, i32 poison, i32 128>
+; CHECK-NEXT:    [[T3:%.*]] = icmp ult <3 x i32> [[T2]], <i32 256, i32 poison, i32 256>
 ; CHECK-NEXT:    [[T4:%.*]] = and <3 x i1> [[T1]], [[T3]]
 ; CHECK-NEXT:    ret <3 x i1> [[T4]]
 ;
-  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 undef, i32 -1>
-  %t2 = add <3 x i32> %arg, <i32 128, i32 undef, i32 128>
-  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 undef, i32 256>
+  %t1 = icmp sgt <3 x i32> %arg, <i32 -1, i32 poison, i32 -1>
+  %t2 = add <3 x i32> %arg, <i32 128, i32 poison, i32 128>
+  %t3 = icmp ult <3 x i32> %t2, <i32 256, i32 poison, i32 256>
   %t4 = and <3 x i1> %t1, %t3
   ret <3 x i1> %t4
 }
diff --git a/llvm/test/Transforms/InstCombine/unsigned-add-lack-of-overflow-check.ll b/llvm/test/Transforms/InstCombine/unsigned-add-lack-of-overflow-check.ll
index 6dde54b17c9ed..e7120a7d01cfa 100644
--- a/llvm/test/Transforms/InstCombine/unsigned-add-lack-of-overflow-check.ll
+++ b/llvm/test/Transforms/InstCombine/unsigned-add-lack-of-overflow-check.ll
@@ -214,7 +214,7 @@ define <2 x i1> @low_bitmask_uge(<2 x i8> %x) {
 ; CHECK-NEXT:    [[R:%.*]] = icmp eq <2 x i8> [[X:%.*]], zeroinitializer
 ; CHECK-NEXT:    ret <2 x i1> [[R]]
 ;
-  %a = add <2 x i8> %x, <i8 15, i8 undef>
+  %a = add <2 x i8> %x, <i8 15, i8 poison>
   %m = and <2 x i8> %a, <i8 15, i8 15>
   %r = icmp uge <2 x i8> %m, %x
   ret <2 x i1> %r
diff --git a/llvm/test/Transforms/InstCombine/xor-ashr.ll b/llvm/test/Transforms/InstCombine/xor-ashr.ll
index 097e04b3b9cb5..6f501184b6357 100644
--- a/llvm/test/Transforms/InstCombine/xor-ashr.ll
+++ b/llvm/test/Transforms/InstCombine/xor-ashr.ll
@@ -63,15 +63,15 @@ define <4 x i8> @testv4i16i8(<4 x i16> %add) {
   ret <4 x i8> %x
 }
 
-define <4 x i8> @testv4i16i8_undef(<4 x i16> %add) {
-; CHECK-LABEL: @testv4i16i8_undef(
+define <4 x i8> @testv4i16i8_poison(<4 x i16> %add) {
+; CHECK-LABEL: @testv4i16i8_poison(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp sgt <4 x i16> [[ADD:%.*]], <i16 -1, i16 -1, i16 -1, i16 -1>
-; CHECK-NEXT:    [[X:%.*]] = select <4 x i1> [[TMP1]], <4 x i8> <i8 27, i8 27, i8 undef, i8 27>, <4 x i8> <i8 -28, i8 -28, i8 undef, i8 -28>
+; CHECK-NEXT:    [[X:%.*]] = select <4 x i1> [[TMP1]], <4 x i8> <i8 27, i8 27, i8 poison, i8 27>, <4 x i8> <i8 -28, i8 -28, i8 poison, i8 -28>
 ; CHECK-NEXT:    ret <4 x i8> [[X]]
 ;
-  %sh = ashr <4 x i16> %add, <i16 15, i16 undef, i16 15, i16 15>
+  %sh = ashr <4 x i16> %add, <i16 15, i16 poison, i16 15, i16 15>
   %t = trunc <4 x i16> %sh to <4 x i8>
-  %x = xor <4 x i8> %t, <i8 27, i8 27, i8 undef, i8 27>
+  %x = xor <4 x i8> %t, <i8 27, i8 27, i8 poison, i8 27>
   ret <4 x i8> %x
 }
 
diff --git a/llvm/test/Transforms/InstSimplify/cast-unsigned-icmp-cmp-0.ll b/llvm/test/Transforms/InstSimplify/cast-unsigned-icmp-cmp-0.ll
index 5a61a060785ff..8f9c75bae1097 100644
--- a/llvm/test/Transforms/InstSimplify/cast-unsigned-icmp-cmp-0.ll
+++ b/llvm/test/Transforms/InstSimplify/cast-unsigned-icmp-cmp-0.ll
@@ -27,13 +27,13 @@ define <2 x i1> @i32_cast_cmp_slt_int_0_uitofp_float_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_float_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_float_vec_undef(
+define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_float_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_float_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %f = uitofp <3 x i32> %i to <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp slt <3 x i32> %b, <i32 0, i32 undef, i32 0>
+  %cmp = icmp slt <3 x i32> %b, <i32 0, i32 poison, i32 0>
   ret <3 x i1> %cmp
 }
 
@@ -70,13 +70,13 @@ define i1 @i32_cast_cmp_sgt_int_m1_uitofp_float_vec_mismatch(<2 x i32> %i) {
   ret i1 %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_float_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_float_vec_undef(
+define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_float_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_float_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> <i1 true, i1 true, i1 true>
 ;
   %f = uitofp <3 x i32> %i to <3 x float>
   %b = bitcast <3 x float> %f to <3 x i32>
-  %cmp = icmp sgt <3 x i32> %b, <i32 -1, i32 undef, i32 -1>
+  %cmp = icmp sgt <3 x i32> %b, <i32 -1, i32 poison, i32 -1>
   ret <3 x i1> %cmp
 }
 
@@ -100,13 +100,13 @@ define <2 x i1> @i32_cast_cmp_slt_int_0_uitofp_double_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_double_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_double_vec_undef(
+define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_double_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_double_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %f = uitofp <3 x i32> %i to <3 x double>
   %b = bitcast <3 x double> %f to <3 x i64>
-  %cmp = icmp slt <3 x i64> %b, <i64 0, i64 undef, i64 0>
+  %cmp = icmp slt <3 x i64> %b, <i64 0, i64 poison, i64 0>
   ret <3 x i1> %cmp
 }
 
@@ -130,13 +130,13 @@ define <2 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_double_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_double_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_double_vec_undef(
+define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_double_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_double_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> <i1 true, i1 true, i1 true>
 ;
   %f = uitofp <3 x i32> %i to <3 x double>
   %b = bitcast <3 x double> %f to <3 x i64>
-  %cmp = icmp sgt <3 x i64> %b, <i64 -1, i64 undef, i64 -1>
+  %cmp = icmp sgt <3 x i64> %b, <i64 -1, i64 poison, i64 -1>
   ret <3 x i1> %cmp
 }
 
@@ -160,13 +160,13 @@ define <2 x i1> @i32_cast_cmp_slt_int_0_uitofp_half_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_half_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_half_vec_undef(
+define <3 x i1> @i32_cast_cmp_slt_int_0_uitofp_half_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_slt_int_0_uitofp_half_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> zeroinitializer
 ;
   %f = uitofp <3 x i32> %i to <3 x half>
   %b = bitcast <3 x half> %f to <3 x i16>
-  %cmp = icmp slt <3 x i16> %b, <i16 0, i16 undef, i16 0>
+  %cmp = icmp slt <3 x i16> %b, <i16 0, i16 poison, i16 0>
   ret <3 x i1> %cmp
 }
 
@@ -190,12 +190,12 @@ define <2 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_half_vec(<2 x i32> %i) {
   ret <2 x i1> %cmp
 }
 
-define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_half_vec_undef(<3 x i32> %i) {
-; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_half_vec_undef(
+define <3 x i1> @i32_cast_cmp_sgt_int_m1_uitofp_half_vec_poison(<3 x i32> %i) {
+; CHECK-LABEL: @i32_cast_cmp_sgt_int_m1_uitofp_half_vec_poison(
 ; CHECK-NEXT:    ret <3 x i1> <i1 true, i1 true, i1 true>
 ;
   %f = uitofp <3 x i32> %i to <3 x half>
   %b = bitcast <3 x half> %f to <3 x i16>
-  %cmp = icmp sgt <3 x i16> %b, <i16 -1, i16 undef, i16 -1>
+  %cmp = icmp sgt <3 x i16> %b, <i16 -1, i16 poison, i16 -1>
   ret <3 x i1> %cmp
 }
diff --git a/llvm/test/Transforms/InstSimplify/icmp-constant.ll b/llvm/test/Transforms/InstSimplify/icmp-constant.ll
index 99bf11bd4e8c4..4f3a5d3dba3b5 100644
--- a/llvm/test/Transforms/InstSimplify/icmp-constant.ll
+++ b/llvm/test/Transforms/InstSimplify/icmp-constant.ll
@@ -19,11 +19,11 @@ define <2 x i1> @tautological_ule_vec(<2 x i8> %x) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @tautological_ule_vec_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @tautological_ule_vec_partial_undef(
+define <2 x i1> @tautological_ule_vec_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @tautological_ule_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %cmp = icmp ule <2 x i8> %x, <i8 255, i8 undef>
+  %cmp = icmp ule <2 x i8> %x, <i8 255, i8 poison>
   ret <2 x i1> %cmp
 }
 
@@ -43,11 +43,11 @@ define <2 x i1> @tautological_ugt_vec(<2 x i8> %x) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @tautological_ugt_vec_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @tautological_ugt_vec_partial_undef(
+define <2 x i1> @tautological_ugt_vec_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @tautological_ugt_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
-  %cmp = icmp ugt <2 x i8> %x, <i8 undef, i8 255>
+  %cmp = icmp ugt <2 x i8> %x, <i8 poison, i8 255>
   ret <2 x i1> %cmp
 }
 
@@ -70,12 +70,12 @@ define <2 x i1> @urem3_vec(<2 x i32> %X) {
   ret <2 x i1> %B
 }
 
-define <2 x i1> @urem3_vec_partial_undef(<2 x i32> %X) {
-; CHECK-LABEL: @urem3_vec_partial_undef(
+define <2 x i1> @urem3_vec_partial_poison(<2 x i32> %X) {
+; CHECK-LABEL: @urem3_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %A = urem <2 x i32> %X, <i32 10, i32 10>
-  %B = icmp ult <2 x i32> %A, <i32 undef, i32 15>
+  %B = icmp ult <2 x i32> %A, <i32 poison, i32 15>
   ret <2 x i1> %B
 }
 
@@ -98,12 +98,12 @@ define <2 x i1> @srem1_vec(<2 x i32> %X) {
   ret <2 x i1> %B
 }
 
-define <2 x i1> @srem1_vec_partial_undef(<2 x i32> %X) {
-; CHECK-LABEL: @srem1_vec_partial_undef(
+define <2 x i1> @srem1_vec_partial_poison(<2 x i32> %X) {
+; CHECK-LABEL: @srem1_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %A = srem <2 x i32> %X, <i32 -5, i32 -5>
-  %B = icmp sgt <2 x i32> %A, <i32 5, i32 undef>
+  %B = icmp sgt <2 x i32> %A, <i32 5, i32 poison>
   ret <2 x i1> %B
 }
 
@@ -203,12 +203,12 @@ define <2 x i1> @shl5_vec(<2 x i32> %X) {
   ret <2 x i1> %cmp
 }
 
-define <2 x i1> @shl5_vec_partial_undef(<2 x i32> %X) {
-; CHECK-LABEL: @shl5_vec_partial_undef(
+define <2 x i1> @shl5_vec_partial_poison(<2 x i32> %X) {
+; CHECK-LABEL: @shl5_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %sub = shl nuw <2 x i32> <i32 4, i32 4>, %X
-  %cmp = icmp ugt <2 x i32> %sub, <i32 undef, i32 3>
+  %cmp = icmp ugt <2 x i32> %sub, <i32 poison, i32 3>
   ret <2 x i1> %cmp
 }
 
@@ -421,12 +421,12 @@ define <2 x i1> @or1_vec(<2 x i32> %X) {
   ret <2 x i1> %B
 }
 
-define <2 x i1> @or1_vec_partial_undef(<2 x i32> %X) {
-; CHECK-LABEL: @or1_vec_partial_undef(
+define <2 x i1> @or1_vec_partial_poison(<2 x i32> %X) {
+; CHECK-LABEL: @or1_vec_partial_poison(
 ; CHECK-NEXT:    ret <2 x i1> zeroinitializer
 ;
   %A = or <2 x i32> %X, <i32 62, i32 62>
-  %B = icmp ult <2 x i32> %A, <i32 undef, i32 50>
+  %B = icmp ult <2 x i32> %A, <i32 poison, i32 50>
   ret <2 x i1> %B
 }
 
@@ -900,22 +900,22 @@ define <2 x i1> @mul_nuw_urem_cmp_constant_vec_splat(<2 x i8> %x) {
 
 ; Undefs in vector constants are ok.
 
-define <2 x i1> @mul_nuw_urem_cmp_constant_vec_splat_undef1(<2 x i8> %x) {
-; CHECK-LABEL: @mul_nuw_urem_cmp_constant_vec_splat_undef1(
+define <2 x i1> @mul_nuw_urem_cmp_constant_vec_splat_poison1(<2 x i8> %x) {
+; CHECK-LABEL: @mul_nuw_urem_cmp_constant_vec_splat_poison1(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %m = mul nuw <2 x i8> %x, <i8 45, i8 45>
-  %r = icmp ne <2 x i8> %m, <i8 15, i8 undef>
+  %r = icmp ne <2 x i8> %m, <i8 15, i8 poison>
   ret <2 x i1> %r
 }
 
 ; Undefs in vector constants are ok.
 
-define <2 x i1> @mul_nuw_urem_cmp_constant_vec_splat_undef2(<2 x i8> %x) {
-; CHECK-LABEL: @mul_nuw_urem_cmp_constant_vec_splat_undef2(
+define <2 x i1> @mul_nuw_urem_cmp_constant_vec_splat_poison2(<2 x i8> %x) {
+; CHECK-LABEL: @mul_nuw_urem_cmp_constant_vec_splat_poison2(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %m = mul nuw <2 x i8> %x, <i8 undef, i8 45>
+  %m = mul nuw <2 x i8> %x, <i8 poison, i8 45>
   %r = icmp ne <2 x i8> %m, <i8 15, i8 15>
   ret <2 x i1> %r
 }
@@ -1007,22 +1007,22 @@ define <2 x i1> @mul_nsw_srem_cmp_constant_vec_splat(<2 x i8> %x) {
 
 ; Undefs in vector constants are ok.
 
-define <2 x i1> @mul_nsw_srem_cmp_constant_vec_splat_undef1(<2 x i8> %x) {
-; CHECK-LABEL: @mul_nsw_srem_cmp_constant_vec_splat_undef1(
+define <2 x i1> @mul_nsw_srem_cmp_constant_vec_splat_poison1(<2 x i8> %x) {
+; CHECK-LABEL: @mul_nsw_srem_cmp_constant_vec_splat_poison1(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
   %m = mul nsw <2 x i8> %x, <i8 45, i8 45>
-  %r = icmp ne <2 x i8> %m, <i8 15, i8 undef>
+  %r = icmp ne <2 x i8> %m, <i8 15, i8 poison>
   ret <2 x i1> %r
 }
 
 ; Undefs in vector constants are ok.
 
-define <2 x i1> @mul_nsw_srem_cmp_constant_vec_splat_undef2(<2 x i8> %x) {
-; CHECK-LABEL: @mul_nsw_srem_cmp_constant_vec_splat_undef2(
+define <2 x i1> @mul_nsw_srem_cmp_constant_vec_splat_poison2(<2 x i8> %x) {
+; CHECK-LABEL: @mul_nsw_srem_cmp_constant_vec_splat_poison2(
 ; CHECK-NEXT:    ret <2 x i1> <i1 true, i1 true>
 ;
-  %m = mul nsw <2 x i8> %x, <i8 undef, i8 45>
+  %m = mul nsw <2 x i8> %x, <i8 poison, i8 45>
   %r = icmp ne <2 x i8> %m, <i8 15, i8 15>
   ret <2 x i1> %r
 }
diff --git a/llvm/test/Transforms/InstSimplify/maxmin_intrinsics.ll b/llvm/test/Transforms/InstSimplify/maxmin_intrinsics.ll
index 17f39a23d4f7a..5c659a1e7086a 100644
--- a/llvm/test/Transforms/InstSimplify/maxmin_intrinsics.ll
+++ b/llvm/test/Transforms/InstSimplify/maxmin_intrinsics.ll
@@ -257,67 +257,67 @@ define <2 x i8> @umin_maxval_commute(<2 x i8> %x) {
   ret <2 x i8> %r
 }
 
-define <2 x i8> @smax_maxval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @smax_maxval_partial_undef(
+define <2 x i8> @smax_maxval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @smax_maxval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> <i8 127, i8 127>
 ;
-  %r = call <2 x i8> @llvm.smax.v2i8(<2 x i8> <i8 undef, i8 127>, <2 x i8> %x)
+  %r = call <2 x i8> @llvm.smax.v2i8(<2 x i8> <i8 poison, i8 127>, <2 x i8> %x)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @smin_minval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @smin_minval_partial_undef(
+define <2 x i8> @smin_minval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @smin_minval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> <i8 -128, i8 -128>
 ;
-  %r = call <2 x i8> @llvm.smin.v2i8(<2 x i8> %x, <2 x i8> <i8 -128, i8 undef>)
+  %r = call <2 x i8> @llvm.smin.v2i8(<2 x i8> %x, <2 x i8> <i8 -128, i8 poison>)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @umax_maxval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @umax_maxval_partial_undef(
+define <2 x i8> @umax_maxval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @umax_maxval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> <i8 -1, i8 -1>
 ;
-  %r = call <2 x i8> @llvm.umax.v2i8(<2 x i8> <i8 255, i8 undef>, <2 x i8> %x)
+  %r = call <2 x i8> @llvm.umax.v2i8(<2 x i8> <i8 255, i8 poison>, <2 x i8> %x)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @umin_minval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @umin_minval_partial_undef(
+define <2 x i8> @umin_minval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @umin_minval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> zeroinitializer
 ;
-  %r = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %x, <2 x i8> <i8 undef, i8 0>)
+  %r = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %x, <2 x i8> <i8 poison, i8 0>)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @smax_minval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @smax_minval_partial_undef(
+define <2 x i8> @smax_minval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @smax_minval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> [[X:%.*]]
 ;
-  %r = call <2 x i8> @llvm.smax.v2i8(<2 x i8> <i8 undef, i8 -128>, <2 x i8> %x)
+  %r = call <2 x i8> @llvm.smax.v2i8(<2 x i8> <i8 poison, i8 -128>, <2 x i8> %x)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @smin_maxval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @smin_maxval_partial_undef(
+define <2 x i8> @smin_maxval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @smin_maxval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> [[X:%.*]]
 ;
-  %r = call <2 x i8> @llvm.smin.v2i8(<2 x i8> %x, <2 x i8> <i8 undef, i8 127>)
+  %r = call <2 x i8> @llvm.smin.v2i8(<2 x i8> %x, <2 x i8> <i8 poison, i8 127>)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @umax_minval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @umax_minval_partial_undef(
+define <2 x i8> @umax_minval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @umax_minval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> [[X:%.*]]
 ;
-  %r = call <2 x i8> @llvm.umax.v2i8(<2 x i8> <i8 0, i8 undef>, <2 x i8> %x)
+  %r = call <2 x i8> @llvm.umax.v2i8(<2 x i8> <i8 0, i8 poison>, <2 x i8> %x)
   ret <2 x i8> %r
 }
 
-define <2 x i8> @umin_maxval_partial_undef(<2 x i8> %x) {
-; CHECK-LABEL: @umin_maxval_partial_undef(
+define <2 x i8> @umin_maxval_partial_poison(<2 x i8> %x) {
+; CHECK-LABEL: @umin_maxval_partial_poison(
 ; CHECK-NEXT:    ret <2 x i8> [[X:%.*]]
 ;
-  %r = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %x, <2 x i8> <i8 255, i8 undef>)
+  %r = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %x, <2 x i8> <i8 255, i8 poison>)
   ret <2 x i8> %r
 }
 
@@ -743,7 +743,7 @@ define <2 x i8> @umin_umin_constants_commute2(<2 x i8> %x) {
 ; CHECK-NEXT:    ret <2 x i8> [[M]]
 ;
   %m = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %x, <2 x i8> <i8 127, i8 127>)
-  %m2 = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %m, <2 x i8> <i8 200, i8 undef>)
+  %m2 = call <2 x i8> @llvm.umin.v2i8(<2 x i8> %m, <2 x i8> <i8 200, i8 poison>)
   ret <2 x i8> %m2
 }
 
@@ -803,7 +803,7 @@ define <2 x i8> @smin_smin_constants(<2 x i8> %x) {
 ; CHECK-NEXT:    ret <2 x i8> [[M]]
 ;
   %m = call <2 x i8> @llvm.smin.v2i8(<2 x i8> %x, <2 x i8> <i8 7, i8 7>)
-  %m2 = call <2 x i8> @llvm.smin.v2i8(<2 x i8> <i8 undef, i8 9>, <2 x i8> %m)
+  %m2 = call <2 x i8> @llvm.smin.v2i8(<2 x i8> <i8 poison, i8 9>, <2 x i8> %m)
   ret <2 x i8> %m2
 }
 
diff --git a/llvm/unittests/IR/InstructionsTest.cpp b/llvm/unittests/IR/InstructionsTest.cpp
index a48a51b1bf474..b47c73f0b329a 100644
--- a/llvm/unittests/IR/InstructionsTest.cpp
+++ b/llvm/unittests/IR/InstructionsTest.cpp
@@ -1410,6 +1410,7 @@ TEST(InstructionsTest, GetSplat) {
   LLVMContext Ctx;
   Type *Int32Ty = Type::getInt32Ty(Ctx);
   Constant *CU = UndefValue::get(Int32Ty);
+  Constant *CP = PoisonValue::get(Int32Ty);
   Constant *C0 = ConstantInt::get(Int32Ty, 0);
   Constant *C1 = ConstantInt::get(Int32Ty, 1);
 
@@ -1419,30 +1420,42 @@ TEST(InstructionsTest, GetSplat) {
   Constant *Splat1Undef = ConstantVector::get({CU, CU, C1, CU});
   Constant *NotSplat = ConstantVector::get({C1, C1, C0, C1 ,C1});
   Constant *NotSplatUndef = ConstantVector::get({CU, C1, CU, CU ,C0});
+  Constant *Splat0Poison = ConstantVector::get({C0, CP, C0, CP});
+  Constant *Splat1Poison = ConstantVector::get({CP, CP, C1, CP});
+  Constant *NotSplatPoison = ConstantVector::get({CP, C1, CP, CP, C0});
 
-  // Default - undefs are not allowed.
+  // Default - undef/poison is not allowed.
   EXPECT_EQ(Splat0->getSplatValue(), C0);
   EXPECT_EQ(Splat1->getSplatValue(), C1);
   EXPECT_EQ(Splat0Undef->getSplatValue(), nullptr);
   EXPECT_EQ(Splat1Undef->getSplatValue(), nullptr);
+  EXPECT_EQ(Splat0Poison->getSplatValue(), nullptr);
+  EXPECT_EQ(Splat1Poison->getSplatValue(), nullptr);
   EXPECT_EQ(NotSplat->getSplatValue(), nullptr);
   EXPECT_EQ(NotSplatUndef->getSplatValue(), nullptr);
+  EXPECT_EQ(NotSplatPoison->getSplatValue(), nullptr);
 
-  // Disallow undefs explicitly.
+  // Disallow poison explicitly.
   EXPECT_EQ(Splat0->getSplatValue(false), C0);
   EXPECT_EQ(Splat1->getSplatValue(false), C1);
   EXPECT_EQ(Splat0Undef->getSplatValue(false), nullptr);
   EXPECT_EQ(Splat1Undef->getSplatValue(false), nullptr);
+  EXPECT_EQ(Splat0Poison->getSplatValue(false), nullptr);
+  EXPECT_EQ(Splat1Poison->getSplatValue(false), nullptr);
   EXPECT_EQ(NotSplat->getSplatValue(false), nullptr);
   EXPECT_EQ(NotSplatUndef->getSplatValue(false), nullptr);
+  EXPECT_EQ(NotSplatPoison->getSplatValue(false), nullptr);
 
-  // Allow undefs.
+  // Allow poison but not undef.
   EXPECT_EQ(Splat0->getSplatValue(true), C0);
   EXPECT_EQ(Splat1->getSplatValue(true), C1);
-  EXPECT_EQ(Splat0Undef->getSplatValue(true), C0);
-  EXPECT_EQ(Splat1Undef->getSplatValue(true), C1);
+  EXPECT_EQ(Splat0Undef->getSplatValue(true), nullptr);
+  EXPECT_EQ(Splat1Undef->getSplatValue(true), nullptr);
+  EXPECT_EQ(Splat0Poison->getSplatValue(true), C0);
+  EXPECT_EQ(Splat1Poison->getSplatValue(true), C1);
   EXPECT_EQ(NotSplat->getSplatValue(true), nullptr);
   EXPECT_EQ(NotSplatUndef->getSplatValue(true), nullptr);
+  EXPECT_EQ(NotSplatPoison->getSplatValue(true), nullptr);
 }
 
 TEST(InstructionsTest, SkipDebug) {
diff --git a/llvm/unittests/IR/PatternMatch.cpp b/llvm/unittests/IR/PatternMatch.cpp
index 133012684d16d..f0377eae9989f 100644
--- a/llvm/unittests/IR/PatternMatch.cpp
+++ b/llvm/unittests/IR/PatternMatch.cpp
@@ -1222,40 +1222,51 @@ TEST_F(PatternMatchTest, VectorUndefInt) {
   EXPECT_TRUE(match(VectorZeroPoison, m_Zero()));
 
   const APInt *C;
-  // Regardless of whether undefs are allowed,
-  // a fully undef constant does not match.
+  // Regardless of whether poison is allowed,
+  // a fully undef/poison constant does not match.
   EXPECT_FALSE(match(ScalarUndef, m_APInt(C)));
-  EXPECT_FALSE(match(ScalarUndef, m_APIntForbidUndef(C)));
-  EXPECT_FALSE(match(ScalarUndef, m_APIntAllowUndef(C)));
+  EXPECT_FALSE(match(ScalarUndef, m_APIntForbidPoison(C)));
+  EXPECT_FALSE(match(ScalarUndef, m_APIntAllowPoison(C)));
   EXPECT_FALSE(match(VectorUndef, m_APInt(C)));
-  EXPECT_FALSE(match(VectorUndef, m_APIntForbidUndef(C)));
-  EXPECT_FALSE(match(VectorUndef, m_APIntAllowUndef(C)));
+  EXPECT_FALSE(match(VectorUndef, m_APIntForbidPoison(C)));
+  EXPECT_FALSE(match(VectorUndef, m_APIntAllowPoison(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APInt(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APIntForbidPoison(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APIntAllowPoison(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APInt(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APIntForbidPoison(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APIntAllowPoison(C)));
 
   // We can always match simple constants and simple splats.
   C = nullptr;
   EXPECT_TRUE(match(ScalarZero, m_APInt(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(ScalarZero, m_APIntForbidUndef(C)));
+  EXPECT_TRUE(match(ScalarZero, m_APIntForbidPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(ScalarZero, m_APIntAllowUndef(C)));
+  EXPECT_TRUE(match(ScalarZero, m_APIntAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
   EXPECT_TRUE(match(VectorZero, m_APInt(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(VectorZero, m_APIntForbidUndef(C)));
+  EXPECT_TRUE(match(VectorZero, m_APIntForbidPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(VectorZero, m_APIntAllowUndef(C)));
+  EXPECT_TRUE(match(VectorZero, m_APIntAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
 
-  // Whether splats with undef can be matched depends on the matcher.
+  // Splats with undef are never allowed.
+  // Whether splats with poison can be matched depends on the matcher.
   EXPECT_FALSE(match(VectorZeroUndef, m_APInt(C)));
-  EXPECT_FALSE(match(VectorZeroUndef, m_APIntForbidUndef(C)));
+  EXPECT_FALSE(match(VectorZeroUndef, m_APIntForbidPoison(C)));
+  EXPECT_FALSE(match(VectorZeroUndef, m_APIntAllowPoison(C)));
+
+  EXPECT_FALSE(match(VectorZeroPoison, m_APInt(C)));
+  EXPECT_FALSE(match(VectorZeroPoison, m_APIntForbidPoison(C)));
   C = nullptr;
-  EXPECT_TRUE(match(VectorZeroUndef, m_APIntAllowUndef(C)));
+  EXPECT_TRUE(match(VectorZeroPoison, m_APIntAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
 }
 
@@ -1387,43 +1398,63 @@ TEST_F(PatternMatchTest, VectorUndefFloat) {
   EXPECT_FALSE(match(VectorNaNPoison, m_Finite()));
 
   const APFloat *C;
-  // Regardless of whether undefs are allowed,
-  // a fully undef constant does not match.
+  // Regardless of whether poison is allowed,
+  // a fully undef/poison constant does not match.
   EXPECT_FALSE(match(ScalarUndef, m_APFloat(C)));
-  EXPECT_FALSE(match(ScalarUndef, m_APFloatForbidUndef(C)));
-  EXPECT_FALSE(match(ScalarUndef, m_APFloatAllowUndef(C)));
+  EXPECT_FALSE(match(ScalarUndef, m_APFloatForbidPoison(C)));
+  EXPECT_FALSE(match(ScalarUndef, m_APFloatAllowPoison(C)));
   EXPECT_FALSE(match(VectorUndef, m_APFloat(C)));
-  EXPECT_FALSE(match(VectorUndef, m_APFloatForbidUndef(C)));
-  EXPECT_FALSE(match(VectorUndef, m_APFloatAllowUndef(C)));
+  EXPECT_FALSE(match(VectorUndef, m_APFloatForbidPoison(C)));
+  EXPECT_FALSE(match(VectorUndef, m_APFloatAllowPoison(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APFloat(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APFloatForbidPoison(C)));
+  EXPECT_FALSE(match(ScalarPoison, m_APFloatAllowPoison(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APFloat(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APFloatForbidPoison(C)));
+  EXPECT_FALSE(match(VectorPoison, m_APFloatAllowPoison(C)));
 
   // We can always match simple constants and simple splats.
   C = nullptr;
   EXPECT_TRUE(match(ScalarZero, m_APFloat(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(ScalarZero, m_APFloatForbidUndef(C)));
+  EXPECT_TRUE(match(ScalarZero, m_APFloatForbidPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(ScalarZero, m_APFloatAllowUndef(C)));
+  EXPECT_TRUE(match(ScalarZero, m_APFloatAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
   EXPECT_TRUE(match(VectorZero, m_APFloat(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(VectorZero, m_APFloatForbidUndef(C)));
+  EXPECT_TRUE(match(VectorZero, m_APFloatForbidPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(VectorZero, m_APFloatAllowUndef(C)));
+  EXPECT_TRUE(match(VectorZero, m_APFloatAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
 
-  // Whether splats with undef can be matched depends on the matcher.
+  // Splats with undef are never allowed.
+  // Whether splats with poison can be matched depends on the matcher.
   EXPECT_FALSE(match(VectorZeroUndef, m_APFloat(C)));
-  EXPECT_FALSE(match(VectorZeroUndef, m_APFloatForbidUndef(C)));
+  EXPECT_FALSE(match(VectorZeroUndef, m_APFloatForbidPoison(C)));
+  EXPECT_FALSE(match(VectorZeroUndef, m_APFloatAllowPoison(C)));
+  EXPECT_FALSE(match(VectorZeroUndef, m_Finite(C)));
+
+  EXPECT_FALSE(match(VectorZeroPoison, m_APFloat(C)));
+  EXPECT_FALSE(match(VectorZeroPoison, m_APFloatForbidPoison(C)));
+  C = nullptr;
+  EXPECT_TRUE(match(VectorZeroPoison, m_APFloatAllowPoison(C)));
+  EXPECT_TRUE(C->isZero());
+  C = nullptr;
+  EXPECT_TRUE(match(VectorZeroPoison, m_Finite(C)));
+  EXPECT_TRUE(C->isZero());
+  EXPECT_FALSE(match(VectorZeroPoison, m_APFloat(C)));
+  EXPECT_FALSE(match(VectorZeroPoison, m_APFloatForbidPoison(C)));
   C = nullptr;
-  EXPECT_TRUE(match(VectorZeroUndef, m_APFloatAllowUndef(C)));
+  EXPECT_TRUE(match(VectorZeroPoison, m_APFloatAllowPoison(C)));
   EXPECT_TRUE(C->isZero());
   C = nullptr;
-  EXPECT_TRUE(match(VectorZeroUndef, m_Finite(C)));
+  EXPECT_TRUE(match(VectorZeroPoison, m_Finite(C)));
   EXPECT_TRUE(C->isZero());
 }
 

From 85215508966abbedaf80a7a0a2ae4687a46447fe Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov@redhat.com>
Date: Thu, 18 Apr 2024 15:56:03 +0900
Subject: [PATCH 837/886] [X86] Use m_APIntAllowPoison instead of
 m_APIntAllowUndef

Fix build after 1baa3850656382d1d549a13f8a716ef5dc886eb8.
---
 llvm/lib/Target/X86/X86TargetTransformInfo.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index d111c4d4ecc1a..515847617e610 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -4123,7 +4123,7 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
         // TODO: Handle funnel-shift cases.
         const APInt *Amt;
         if (Args[2] &&
-            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowUndef(Amt)))
+            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowPoison(Amt)))
           ISD = X86ISD::VROTLI;
       }
     }
@@ -4139,7 +4139,7 @@ X86TTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
         // TODO: Handle funnel-shift cases.
         const APInt *Amt;
         if (Args[2] &&
-            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowUndef(Amt)))
+            PatternMatch::match(Args[2], PatternMatch::m_APIntAllowPoison(Amt)))
           ISD = X86ISD::VROTLI;
       }
     }

From 9760b6b1ecff3d3c43d182ea488fa34a82472a3e Mon Sep 17 00:00:00 2001
From: Mike Weller <mike@mikeweller.com>
Date: Thu, 18 Apr 2024 08:19:06 +0100
Subject: [PATCH 838/886] [clang-tidy] Ignore deleted ctor in
 `bugprone-forwarding-reference-overload` (#88138)

Fix `bugprone-forwarding-reference-overload` so it doesn't report a
constructor that is deleted.
---
 .../bugprone/ForwardingReferenceOverloadCheck.cpp      |  2 +-
 clang-tools-extra/docs/ReleaseNotes.rst                |  4 ++++
 .../bugprone/forwarding-reference-overload.cpp         | 10 ++++++++++
 3 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp
index c608fe713f9f5..e7be8134781e4 100644
--- a/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp
+++ b/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp
@@ -72,7 +72,7 @@ void ForwardingReferenceOverloadCheck::registerMatchers(MatchFinder *Finder) {
 
   DeclarationMatcher FindOverload =
       cxxConstructorDecl(
-          hasParameter(0, ForwardingRefParm),
+          hasParameter(0, ForwardingRefParm), unless(isDeleted()),
           unless(hasAnyParameter(
               // No warning: enable_if as constructor parameter.
               parmVarDecl(hasType(isEnableIf())))),
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index 7095c564444fe..a457e6fcae946 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -147,6 +147,10 @@ Changes in existing checks
   <clang-tidy/checks/bugprone/assert-side-effect>` check by detecting side
   effect from calling a method with non-const reference parameters.
 
+- Improved :doc:`bugprone-forwarding-reference-overload
+  <clang-tidy/checks/bugprone/forwarding-reference-overload>`
+  check to ignore deleted constructors which won't hide other overloads.
+
 - Improved :doc:`bugprone-inc-dec-in-conditions
   <clang-tidy/checks/bugprone/inc-dec-in-conditions>` check to ignore code
   within unevaluated contexts, such as ``decltype``.
diff --git a/clang-tools-extra/test/clang-tidy/checkers/bugprone/forwarding-reference-overload.cpp b/clang-tools-extra/test/clang-tidy/checkers/bugprone/forwarding-reference-overload.cpp
index 38b0691bc9f1e..92dfb718bb51b 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/bugprone/forwarding-reference-overload.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/bugprone/forwarding-reference-overload.cpp
@@ -251,3 +251,13 @@ class Test10 {
   Test10(T &&Item, E e)
       : e(e){}
 };
+
+// A deleted ctor cannot hide anything
+class Test11 {
+public:
+  template <typename T>
+  Test11(T&&) = delete;
+
+  Test11(const Test11 &) = default;
+  Test11(Test11 &&) = default;
+};

From da579ad807bd55c686e1c1c52e0a4bcc259347b4 Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Thu, 18 Apr 2024 09:21:10 +0200
Subject: [PATCH 839/886] [clang][dataflow] Refactor `PropagateResultObject()`
 with a switch statement. (#88865)

See also discussion in #88726.
---
 .../FlowSensitive/DataflowEnvironment.cpp     | 86 +++++++++++--------
 1 file changed, 50 insertions(+), 36 deletions(-)

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index 3f1600d9ac5d8..e5aaa8fea4803 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -414,25 +414,52 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
 
     ResultObjectMap[E] = Loc;
 
+    switch (E->getStmtClass()) {
     // The following AST node kinds are "original initializers": They are the
     // lowest-level AST node that initializes a given object, and nothing
     // below them can initialize the same object (or part of it).
-    if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
-        isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
-        isa<CXXStdInitializerListExpr>(E) ||
-        // We treat `BuiltinBitCastExpr` as an "original initializer" too as
-        // it may not even be casting from a record type -- and even if it is,
-        // the two objects are in general of unrelated type.
-        isa<BuiltinBitCastExpr>(E)) {
+    case Stmt::CXXConstructExprClass:
+    case Stmt::CallExprClass:
+    case Stmt::LambdaExprClass:
+    case Stmt::CXXDefaultArgExprClass:
+    case Stmt::CXXDefaultInitExprClass:
+    case Stmt::CXXStdInitializerListExprClass:
+    // We treat `BuiltinBitCastExpr` as an "original initializer" too as it may
+    // not even be casting from a record type -- and even if it is, the two
+    // objects are in general of unrelated type.
+    case Stmt::BuiltinBitCastExprClass:
+      return;
+    case Stmt::BinaryOperatorClass: {
+      auto *Op = cast<BinaryOperator>(E);
+      if (Op->getOpcode() == BO_Cmp) {
+        // Builtin `<=>` returns a `std::strong_ordering` object. We
+        // consider this to be an "original" initializer too (see above).
+        return;
+      }
+      if (Op->isCommaOp()) {
+        PropagateResultObject(Op->getRHS(), Loc);
+        return;
+      }
+      // We don't expect any other binary operators to produce a record
+      // prvalue, so if we get here, we've hit some case we don't know
+      // about.
+      assert(false);
       return;
     }
-    if (auto *Op = dyn_cast<BinaryOperator>(E);
-        Op && Op->getOpcode() == BO_Cmp) {
-      // Builtin `<=>` returns a `std::strong_ordering` object.
+    case Stmt::BinaryConditionalOperatorClass:
+    case Stmt::ConditionalOperatorClass: {
+      auto *Cond = cast<AbstractConditionalOperator>(E);
+      PropagateResultObject(Cond->getTrueExpr(), Loc);
+      PropagateResultObject(Cond->getFalseExpr(), Loc);
       return;
     }
-
-    if (auto *InitList = dyn_cast<InitListExpr>(E)) {
+    case Stmt::StmtExprClass: {
+      auto *SE = cast<StmtExpr>(E);
+      PropagateResultObject(cast<Expr>(SE->getSubStmt()->body_back()), Loc);
+      return;
+    }
+    case Stmt::InitListExprClass: {
+      auto *InitList = cast<InitListExpr>(E);
       if (!InitList->isSemanticForm())
         return;
       if (InitList->isTransparent()) {
@@ -462,33 +489,20 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
       }
       return;
     }
-
-    if (auto *Op = dyn_cast<BinaryOperator>(E); Op && Op->isCommaOp()) {
-      PropagateResultObject(Op->getRHS(), Loc);
+    default: {
+      // All other expression nodes that propagate a record prvalue should
+      // have exactly one child.
+      SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
+      LLVM_DEBUG({
+        if (Children.size() != 1)
+          E->dump();
+      });
+      assert(Children.size() == 1);
+      for (Stmt *S : Children)
+        PropagateResultObject(cast<Expr>(S), Loc);
       return;
     }
-
-    if (auto *Cond = dyn_cast<AbstractConditionalOperator>(E)) {
-      PropagateResultObject(Cond->getTrueExpr(), Loc);
-      PropagateResultObject(Cond->getFalseExpr(), Loc);
-      return;
     }
-
-    if (auto *SE = dyn_cast<StmtExpr>(E)) {
-      PropagateResultObject(cast<Expr>(SE->getSubStmt()->body_back()), Loc);
-      return;
-    }
-
-    // All other expression nodes that propagate a record prvalue should have
-    // exactly one child.
-    SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
-    LLVM_DEBUG({
-      if (Children.size() != 1)
-        E->dump();
-    });
-    assert(Children.size() == 1);
-    for (Stmt *S : Children)
-      PropagateResultObject(cast<Expr>(S), Loc);
   }
 
 private:

From fcdb2203e05002f0488221f6a0cab42e1a34f8fc Mon Sep 17 00:00:00 2001
From: Pierre van Houtryve <pierre.vanhoutryve@amd.com>
Date: Thu, 18 Apr 2024 09:22:44 +0200
Subject: [PATCH 840/886] [AMDGPU][AtomicOptimizer] Fix DT update for divergent
 values with Iterative strategy (#87605)

We take the terminator from EntryBB and put it in ComputeEnd. Make sure
we also move the DT edges, we previously only did it assuming a
non-conditional branch.

Fixes SWDEV-453943
---
 .../Target/AMDGPU/AMDGPUAtomicOptimizer.cpp   | 29 ++++--
 .../atomic_optimization_split_dt_update.ll    | 88 +++++++++++++++++++
 2 files changed, 108 insertions(+), 9 deletions(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
index dbb3de76b4dda..ad98f4f743ae0 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
@@ -743,7 +743,7 @@ void AMDGPUAtomicOptimizerImpl::optimizeAtomic(Instruction &I,
 
   Function *F = I.getFunction();
   LLVMContext &C = F->getContext();
-  
+
   // For atomic sub, perform scan with add operation and allow one lane to
   // subtract the reduced value later.
   AtomicRMWInst::BinOp ScanOp = Op;
@@ -855,7 +855,7 @@ void AMDGPUAtomicOptimizerImpl::optimizeAtomic(Instruction &I,
   Value *const Cond = B.CreateICmpEQ(Mbcnt, B.getInt32(0));
 
   // Store I's original basic block before we split the block.
-  BasicBlock *const EntryBB = I.getParent();
+  BasicBlock *const OriginalBB = I.getParent();
 
   // We need to introduce some new control flow to force a single lane to be
   // active. We do this by splitting I's basic block at I, and introducing the
@@ -876,25 +876,36 @@ void AMDGPUAtomicOptimizerImpl::optimizeAtomic(Instruction &I,
   BasicBlock *Predecessor = nullptr;
   if (ValDivergent && ScanImpl == ScanOptions::Iterative) {
     // Move terminator from I's block to ComputeEnd block.
-    Instruction *Terminator = EntryBB->getTerminator();
+    //
+    // OriginalBB is known to have a branch as terminator because
+    // SplitBlockAndInsertIfThen will have inserted one.
+    BranchInst *Terminator = cast<BranchInst>(OriginalBB->getTerminator());
     B.SetInsertPoint(ComputeEnd);
     Terminator->removeFromParent();
     B.Insert(Terminator);
 
     // Branch to ComputeLoop Block unconditionally from the I's block for
     // iterative approach.
-    B.SetInsertPoint(EntryBB);
+    B.SetInsertPoint(OriginalBB);
     B.CreateBr(ComputeLoop);
 
     // Update the dominator tree for new control flow.
-    DTU.applyUpdates(
-        {{DominatorTree::Insert, EntryBB, ComputeLoop},
-         {DominatorTree::Insert, ComputeLoop, ComputeEnd},
-         {DominatorTree::Delete, EntryBB, SingleLaneTerminator->getParent()}});
+    SmallVector<DominatorTree::UpdateType, 6> DomTreeUpdates(
+        {{DominatorTree::Insert, OriginalBB, ComputeLoop},
+         {DominatorTree::Insert, ComputeLoop, ComputeEnd}});
+
+    // We're moving the terminator from EntryBB to ComputeEnd, make sure we move
+    // the DT edges as well.
+    for (auto *Succ : Terminator->successors()) {
+      DomTreeUpdates.push_back({DominatorTree::Insert, ComputeEnd, Succ});
+      DomTreeUpdates.push_back({DominatorTree::Delete, OriginalBB, Succ});
+    }
+
+    DTU.applyUpdates(DomTreeUpdates);
 
     Predecessor = ComputeEnd;
   } else {
-    Predecessor = EntryBB;
+    Predecessor = OriginalBB;
   }
   // Move the IR builder into single_lane next.
   B.SetInsertPoint(SingleLaneTerminator);
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll b/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll
new file mode 100644
index 0000000000000..c07cd4e493b9a
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll
@@ -0,0 +1,88 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -mtriple=amdgcn-- -mcpu=gfx908 -passes="amdgpu-atomic-optimizer,verify<domtree>" %s -S -o - | FileCheck %s
+
+; Check we're properly adding an edge from ComputeEnd to the "End" block added by
+; SplitBlockAndInsertIfThen
+;
+; If the edge is not added, domtree verification will fail.
+
+declare i32 @quux()
+
+define amdgpu_kernel void @ham(ptr addrspace(4) %arg) {
+; CHECK-LABEL: define amdgpu_kernel void @ham(
+; CHECK-SAME: ptr addrspace(4) [[ARG:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[CALL:%.*]] = tail call i32 @quux()
+; CHECK-NEXT:    [[ICMP:%.*]] = icmp eq i32 [[CALL]], 0
+; CHECK-NEXT:    br i1 [[ICMP]], label [[BB1:%.*]], label [[BB3:%.*]]
+; CHECK:       bb1:
+; CHECK-NEXT:    [[CALL2:%.*]] = tail call i32 @quux()
+; CHECK-NEXT:    br label [[BB3]]
+; CHECK:       bb3:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[CALL2]], [[BB1]] ], [ [[CALL]], [[BB:%.*]] ]
+; CHECK-NEXT:    br label [[BB4:%.*]]
+; CHECK:       bb4:
+; CHECK-NEXT:    [[CALL5:%.*]] = tail call i32 @quux()
+; CHECK-NEXT:    [[ICMP6:%.*]] = icmp eq i32 [[CALL5]], 0
+; CHECK-NEXT:    br i1 [[ICMP6]], label [[BB8:%.*]], label [[BB7:%.*]]
+; CHECK:       bb7:
+; CHECK-NEXT:    [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[ARG]], align 8
+; CHECK-NEXT:    [[ADDRSPACECAST:%.*]] = addrspacecast ptr [[LOAD]] to ptr addrspace(1)
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
+; CHECK-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP0]], 32
+; CHECK-NEXT:    [[TMP3:%.*]] = trunc i64 [[TMP2]] to i32
+; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP1]], i32 0)
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP3]], i32 [[TMP4]])
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
+; CHECK-NEXT:    br label [[COMPUTELOOP:%.*]]
+; CHECK:       7:
+; CHECK-NEXT:    [[TMP8:%.*]] = atomicrmw add ptr addrspace(1) [[ADDRSPACECAST]], i32 [[TMP13:%.*]] syncscope("agent-one-as") monotonic, align 4
+; CHECK-NEXT:    br label [[TMP9:%.*]]
+; CHECK:       9:
+; CHECK-NEXT:    br label [[BB8]]
+; CHECK:       bb8:
+; CHECK-NEXT:    br label [[BB4]]
+; CHECK:       ComputeLoop:
+; CHECK-NEXT:    [[ACCUMULATOR:%.*]] = phi i32 [ 0, [[BB7]] ], [ [[TMP13]], [[COMPUTELOOP]] ]
+; CHECK-NEXT:    [[ACTIVEBITS:%.*]] = phi i64 [ [[TMP6]], [[BB7]] ], [ [[TMP16:%.*]], [[COMPUTELOOP]] ]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
+; CHECK-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[PHI]], i32 [[TMP11]])
+; CHECK-NEXT:    [[TMP13]] = add i32 [[ACCUMULATOR]], [[TMP12]]
+; CHECK-NEXT:    [[TMP14:%.*]] = shl i64 1, [[TMP10]]
+; CHECK-NEXT:    [[TMP15:%.*]] = xor i64 [[TMP14]], -1
+; CHECK-NEXT:    [[TMP16]] = and i64 [[ACTIVEBITS]], [[TMP15]]
+; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[TMP16]], 0
+; CHECK-NEXT:    br i1 [[TMP17]], label [[COMPUTEEND:%.*]], label [[COMPUTELOOP]]
+; CHECK:       ComputeEnd:
+; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[TMP5]], 0
+; CHECK-NEXT:    br i1 [[TMP18]], label [[TMP7:%.*]], label [[TMP9]]
+;
+bb:
+  %call = tail call i32 @quux()
+  %icmp = icmp eq i32 %call, 0
+  br i1 %icmp, label %bb1, label %bb3
+
+bb1:                                              ; preds = %bb
+  %call2 = tail call i32 @quux()
+  br label %bb3
+
+bb3:                                              ; preds = %bb1, %bb
+  %phi = phi i32 [ %call2, %bb1 ], [ %call, %bb ]
+  br label %bb4
+
+bb4:                                              ; preds = %bb8, %bb3
+  %call5 = tail call i32 @quux()
+  %icmp6 = icmp eq i32 %call5, 0
+  br i1 %icmp6, label %bb8, label %bb7
+
+bb7:                                              ; preds = %bb4
+  %load = load ptr, ptr addrspace(4) %arg, align 8
+  %addrspacecast = addrspacecast ptr %load to ptr addrspace(1)
+  %atomicrmw = atomicrmw add ptr addrspace(1) %addrspacecast, i32 %phi syncscope("agent-one-as") monotonic, align 4
+  br label %bb8
+
+bb8:                                              ; preds = %bb7, %bb4
+  br label %bb4
+}

From b5f2cecf833f8d08197cda80ac684c250b8fa0f9 Mon Sep 17 00:00:00 2001
From: martinboehme <mboehme@google.com>
Date: Thu, 18 Apr 2024 09:23:03 +0200
Subject: [PATCH 841/886] Revert "[clang][dataflow] Refactor
 `PropagateResultObject()` with a switch statement." (#89176)

Reverts llvm/llvm-project#88865

There were failing tests in the CI that I didn't notice. Sorry.
---
 .../FlowSensitive/DataflowEnvironment.cpp     | 86 ++++++++-----------
 1 file changed, 36 insertions(+), 50 deletions(-)

diff --git a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
index e5aaa8fea4803..3f1600d9ac5d8 100644
--- a/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
+++ b/clang/lib/Analysis/FlowSensitive/DataflowEnvironment.cpp
@@ -414,52 +414,25 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
 
     ResultObjectMap[E] = Loc;
 
-    switch (E->getStmtClass()) {
     // The following AST node kinds are "original initializers": They are the
     // lowest-level AST node that initializes a given object, and nothing
     // below them can initialize the same object (or part of it).
-    case Stmt::CXXConstructExprClass:
-    case Stmt::CallExprClass:
-    case Stmt::LambdaExprClass:
-    case Stmt::CXXDefaultArgExprClass:
-    case Stmt::CXXDefaultInitExprClass:
-    case Stmt::CXXStdInitializerListExprClass:
-    // We treat `BuiltinBitCastExpr` as an "original initializer" too as it may
-    // not even be casting from a record type -- and even if it is, the two
-    // objects are in general of unrelated type.
-    case Stmt::BuiltinBitCastExprClass:
-      return;
-    case Stmt::BinaryOperatorClass: {
-      auto *Op = cast<BinaryOperator>(E);
-      if (Op->getOpcode() == BO_Cmp) {
-        // Builtin `<=>` returns a `std::strong_ordering` object. We
-        // consider this to be an "original" initializer too (see above).
-        return;
-      }
-      if (Op->isCommaOp()) {
-        PropagateResultObject(Op->getRHS(), Loc);
-        return;
-      }
-      // We don't expect any other binary operators to produce a record
-      // prvalue, so if we get here, we've hit some case we don't know
-      // about.
-      assert(false);
+    if (isa<CXXConstructExpr>(E) || isa<CallExpr>(E) || isa<LambdaExpr>(E) ||
+        isa<CXXDefaultArgExpr>(E) || isa<CXXDefaultInitExpr>(E) ||
+        isa<CXXStdInitializerListExpr>(E) ||
+        // We treat `BuiltinBitCastExpr` as an "original initializer" too as
+        // it may not even be casting from a record type -- and even if it is,
+        // the two objects are in general of unrelated type.
+        isa<BuiltinBitCastExpr>(E)) {
       return;
     }
-    case Stmt::BinaryConditionalOperatorClass:
-    case Stmt::ConditionalOperatorClass: {
-      auto *Cond = cast<AbstractConditionalOperator>(E);
-      PropagateResultObject(Cond->getTrueExpr(), Loc);
-      PropagateResultObject(Cond->getFalseExpr(), Loc);
-      return;
-    }
-    case Stmt::StmtExprClass: {
-      auto *SE = cast<StmtExpr>(E);
-      PropagateResultObject(cast<Expr>(SE->getSubStmt()->body_back()), Loc);
+    if (auto *Op = dyn_cast<BinaryOperator>(E);
+        Op && Op->getOpcode() == BO_Cmp) {
+      // Builtin `<=>` returns a `std::strong_ordering` object.
       return;
     }
-    case Stmt::InitListExprClass: {
-      auto *InitList = cast<InitListExpr>(E);
+
+    if (auto *InitList = dyn_cast<InitListExpr>(E)) {
       if (!InitList->isSemanticForm())
         return;
       if (InitList->isTransparent()) {
@@ -489,20 +462,33 @@ class ResultObjectVisitor : public RecursiveASTVisitor<ResultObjectVisitor> {
       }
       return;
     }
-    default: {
-      // All other expression nodes that propagate a record prvalue should
-      // have exactly one child.
-      SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
-      LLVM_DEBUG({
-        if (Children.size() != 1)
-          E->dump();
-      });
-      assert(Children.size() == 1);
-      for (Stmt *S : Children)
-        PropagateResultObject(cast<Expr>(S), Loc);
+
+    if (auto *Op = dyn_cast<BinaryOperator>(E); Op && Op->isCommaOp()) {
+      PropagateResultObject(Op->getRHS(), Loc);
       return;
     }
+
+    if (auto *Cond = dyn_cast<AbstractConditionalOperator>(E)) {
+      PropagateResultObject(Cond->getTrueExpr(), Loc);
+      PropagateResultObject(Cond->getFalseExpr(), Loc);
+      return;
     }
+
+    if (auto *SE = dyn_cast<StmtExpr>(E)) {
+      PropagateResultObject(cast<Expr>(SE->getSubStmt()->body_back()), Loc);
+      return;
+    }
+
+    // All other expression nodes that propagate a record prvalue should have
+    // exactly one child.
+    SmallVector<Stmt *, 1> Children(E->child_begin(), E->child_end());
+    LLVM_DEBUG({
+      if (Children.size() != 1)
+        E->dump();
+    });
+    assert(Children.size() == 1);
+    for (Stmt *S : Children)
+      PropagateResultObject(cast<Expr>(S), Loc);
   }
 
 private:

From 42348b6918e9bbf2dd0260d8ffab4b80b8ff08f8 Mon Sep 17 00:00:00 2001
From: Guray Ozen <guray.ozen@gmail.com>
Date: Thu, 18 Apr 2024 09:24:14 +0200
Subject: [PATCH 842/886] [github] Add myself as default code reviewer (#89099)

---
 .github/CODEOWNERS | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
index c4727a0c267d3..601e590205b0c 100644
--- a/.github/CODEOWNERS
+++ b/.github/CODEOWNERS
@@ -113,7 +113,7 @@ clang/test/AST/Interp/ @tbaederr
 
 # MLIR NVVM Dialect in MLIR
 /mlir/**/LLVMIR/**/BasicPtxBuilderInterface* @grypp
-/mlir/**/NVVM*/ @grypp
+/mlir/**/NVVM* @grypp
 
 # MLIR Python Bindings
 /mlir/test/python/ @makslevental @stellaraccident

From 3c721b90d363bf73b78467f6e86c879235bac1b2 Mon Sep 17 00:00:00 2001
From: Pavel Labath <pavel@labath.sk>
Date: Thu, 18 Apr 2024 06:45:50 +0000
Subject: [PATCH 843/886] Revert "[lldb] Fix evaluation of expressions with
 static initializers (#89063)"

It breaks expression evaluation on arm, and the x86 breakage has been
fixed in 6cea7c491f4c4c68aa0494a9b18f36ff40c22c81.

This reverts commit 915c84b1480bb3c6d2e44ca83822d2c2304b763a.
---
 lldb/source/Expression/IRExecutionUnit.cpp | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/lldb/source/Expression/IRExecutionUnit.cpp b/lldb/source/Expression/IRExecutionUnit.cpp
index 7ad0e5ff22b2f..cb9bee8733e15 100644
--- a/lldb/source/Expression/IRExecutionUnit.cpp
+++ b/lldb/source/Expression/IRExecutionUnit.cpp
@@ -13,7 +13,6 @@
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
-#include "llvm/Support/CodeGen.h"
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/raw_ostream.h"
 
@@ -280,13 +279,10 @@ void IRExecutionUnit::GetRunnableInfo(Status &error, lldb::addr_t &func_addr,
   llvm::EngineBuilder builder(std::move(m_module_up));
   llvm::Triple triple(m_module->getTargetTriple());
 
-  // PIC needed for ELF to avoid generating 32-bit relocations (which overflow
-  // if the object is loaded into high memory).
-  bool want_pic = triple.isOSBinFormatMachO() || triple.isOSBinFormatELF();
-
   builder.setEngineKind(llvm::EngineKind::JIT)
       .setErrorStr(&error_string)
-      .setRelocationModel(want_pic ? llvm::Reloc::PIC_ : llvm::Reloc::Static)
+      .setRelocationModel(triple.isOSBinFormatMachO() ? llvm::Reloc::PIC_
+                                                      : llvm::Reloc::Static)
       .setMCJITMemoryManager(std::make_unique<MemoryManager>(*this))
       .setOptLevel(llvm::CodeGenOptLevel::Less);
 

From 135472b553ccc18e4337b23f8d5616c8dda5e2a6 Mon Sep 17 00:00:00 2001
From: Isha Agarwal <isha.agarwal@intel.com>
Date: Thu, 18 Apr 2024 00:37:01 -0700
Subject: [PATCH 844/886] [X86][test] Added extra cet tests, NFC (#88736)

Updated cet test to:
-Check different modules based on different cet options
-Added negative tests also

---------

Signed-off-by: Isha Agarwal <isha.agarwal@intel.com>
---
 clang/test/CodeGen/X86/x86-cf-protection.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/clang/test/CodeGen/X86/x86-cf-protection.c b/clang/test/CodeGen/X86/x86-cf-protection.c
index ba63b9e17c6f6..84c7c0f16663c 100644
--- a/clang/test/CodeGen/X86/x86-cf-protection.c
+++ b/clang/test/CodeGen/X86/x86-cf-protection.c
@@ -1,13 +1,20 @@
 // RUN: %clang_cc1 -E -triple i386 -dM -o - -fcf-protection=return %s | FileCheck %s --check-prefix=RETURN
 // RUN: %clang_cc1 -E -triple i386 -dM -o - -fcf-protection=branch %s | FileCheck %s --check-prefix=BRANCH
 // RUN: %clang_cc1 -E -triple i386 -dM -o - -fcf-protection=full %s   | FileCheck %s --check-prefix=FULL
+// RUN: %clang_cc1 -E -triple=x86_64 -dM -o - -fcf-protection=none %s | FileCheck %s --check-prefix=NOTCET
 // RUN: not %clang_cc1 -emit-llvm-only -triple i386 -target-cpu pentium-mmx -fcf-protection=branch %s 2>&1 | FileCheck %s --check-prefix=NOCFPROT
+// RUN: %clang_cc1 -triple=x86_64 -o - -fcf-protection=return %s -emit-llvm | FileCheck %s --check-prefixes=CFPROTR,CFPROTNONE
+// RUN: %clang_cc1 -triple=x86_64 -o - -fcf-protection=branch %s -emit-llvm | FileCheck %s --check-prefixes=CFPROTB,CFPROTNONE
+// RUN: %clang_cc1 -triple=x86_64 -o - -fcf-protection=full %s -emit-llvm | FileCheck %s --check-prefixes=CFPROTR,CFPROTB,CFPROTNONE
+// RUN: %clang_cc1 -triple=x86_64 -o - -fcf-protection=none %s -emit-llvm | FileCheck %s --check-prefixes=CFPROTNONE
 
 // RETURN: #define __CET__ 2
 // BRANCH: #define __CET__ 1
 // FULL: #define __CET__ 3
-// CFPROT: !{i32 8, !"cf-protection-branch", i32 1}
-
+// NOTCET-NOT: #define __CET__
 // NOCFPROT: error: option 'cf-protection=branch' cannot be specified on this target
+// CFPROTR: !{i32 8, !"cf-protection-return", i32 1}
+// CFPROTB: !{i32 8, !"cf-protection-branch", i32 1}
+// CFPROTNONE-NOT: cf-protection-
 
 void foo() {}

From 5d4e072a253afac018e312d75290314bd2d03b30 Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Thu, 18 Apr 2024 15:21:49 +0800
Subject: [PATCH 845/886] [C++20] [Modules] [Reduced BMI] Don't eagerly write
 static entities in module purview

For,

```
export module A;
static int impl() { ... }
export int func() { return impl(); }
```

Previously, even with reduced BMI, the function `impl` will be emitted
into the BMI. After the patch, the static entities in module purview
won't get emitted eagerly. Now the static entities may only be emitted
if required.

Note that, this restriction is actually more relaxed than the language
standard required. The language spec said, the program is ill-formed if
any TU-local entities get exposed. However, we can't do this since there
are many static entities in the headers of existing libraries.
Forbidding that will cause many existing program fail immediately.

Another note here is, we can't do this for non-static non-exported
entities, they can be used for other module units within the same
module.
---
 clang/lib/Serialization/ASTWriter.cpp         | 17 ++++++++++++--
 .../Modules/unreached-static-entities.cppm    | 22 +++++++++++++++++++
 2 files changed, 37 insertions(+), 2 deletions(-)
 create mode 100644 clang/test/Modules/unreached-static-entities.cppm

diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 9ae9648f38a3b..522b8ce56c326 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -4892,8 +4892,21 @@ void ASTWriter::PrepareWritingSpecialDecls(Sema &SemaRef) {
     if (D->isFromASTFile())
       continue;
 
-    if (GeneratingReducedBMI && D->isFromExplicitGlobalModule())
-      continue;
+    if (GeneratingReducedBMI) {
+      if (D->isFromExplicitGlobalModule())
+        continue;
+
+      // Don't force emitting static entities.
+      //
+      // Technically, all static entities shouldn't be in reduced BMI. The
+      // language also specifies that the program exposes TU-local entities
+      // is ill-formed. However, in practice, there are a lot of projects
+      // uses `static inline` in the headers. So we can't get rid of all
+      // static entities in reduced BMI now.
+      if (auto *ND = dyn_cast<NamedDecl>(D);
+          ND && ND->getFormalLinkage() == Linkage::Internal)
+        continue;
+    }
 
     GetDeclRef(D);
   }
diff --git a/clang/test/Modules/unreached-static-entities.cppm b/clang/test/Modules/unreached-static-entities.cppm
new file mode 100644
index 0000000000000..10f70ae09e5a1
--- /dev/null
+++ b/clang/test/Modules/unreached-static-entities.cppm
@@ -0,0 +1,22 @@
+// Test that the static function only used in non-inline functions won't get emitted
+// into the BMI.
+//
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+//
+// RUN: %clang_cc1 -std=c++20 %s -emit-reduced-module-interface -o %t/S.pcm
+// RUN: llvm-bcanalyzer --dump --disable-histogram --show-binary-blobs %t/S.pcm > %t/S.dump
+// RUN: cat %t/S.dump | FileCheck %s
+
+export module S;
+static int static_func() {
+    return 43;
+}
+
+export int func() {
+    return static_func();
+}
+
+// CHECK: <DECL_FUNCTION
+// Checks that we won't see a second function
+// CHECK-NOT: <DECL_FUNCTION

From 6e30d97e89f1c49e8bf2073746a5e08f5e75948f Mon Sep 17 00:00:00 2001
From: Durgadoss R <durgadossr@nvidia.com>
Date: Thu, 18 Apr 2024 13:10:49 +0530
Subject: [PATCH 846/886] [MLIR][NVVM] [NFC] Update Docs for shfl.sync Op
 (#89044)

The first argument to the nvvm_shfl_sync_* family
of intrinsics is the thread_mask (aka member_mask).
This patch renames the corresponding operand in the Op
to reflect the same i.e. `dst` -> `thread_mask`.

While we are there, add summary and description
for this Op.

Signed-off-by: Durgadoss R <durgadossr@nvidia.com>
---
 mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td b/mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td
index 728e92c9dc8dc..f76b6d19b8955 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td
@@ -619,20 +619,33 @@ def ShflKindAttr : EnumAttr<NVVM_Dialect, ShflKind, "shfl_kind">;
 def NVVM_ShflOp :
   NVVM_Op<"shfl.sync">,
   Results<(outs LLVM_Type:$res)>,
-  Arguments<(ins I32:$dst,
+  Arguments<(ins I32:$thread_mask,
                  LLVM_Type:$val,
                  I32:$offset,
                  I32:$mask_and_clamp,
                  ShflKindAttr:$kind,
                  OptionalAttr<UnitAttr>:$return_value_and_is_valid)> {
+  let summary = "NVVM Dialect Op for shfl.sync";
+  let description = [{
+    The `shfl.sync` Op implements data shuffle within threads of a warp.
+    The `thread_mask` denotes the threads participating in the Op where
+    the bit position corresponds to a particular thread’s laneid.
+    The `offset` specifies a source lane or source lane offset
+    (depending on `kind`). The `val` is the input value to be copied from
+    the source. The `mask_and_clamp` contains two packed values specifying
+    a mask for logically splitting warps into sub-segments and an upper bound
+    for clamping the source lane index.
+    [For more information, refer PTX ISA]
+    (https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-shfl-sync)
+  }];
   string llvmBuilder = [{
       auto intId = getShflIntrinsicId(
           $_resultType, $kind, static_cast<bool>($return_value_and_is_valid));
       $res = createIntrinsicCall(builder,
-          intId, {$dst, $val, $offset, $mask_and_clamp});
+          intId, {$thread_mask, $val, $offset, $mask_and_clamp});
   }];
   let assemblyFormat = [{
-    $kind $dst `,` $val `,` $offset `,` $mask_and_clamp  attr-dict
+    $kind $thread_mask `,` $val `,` $offset `,` $mask_and_clamp  attr-dict
      `:` type($val) `->` type($res)
    }];
    let hasVerifier = 1;

From 609ee9fe6a96113c893e8344f0dfaef5eba73e3d Mon Sep 17 00:00:00 2001
From: Jie Fu <jiefu@tencent.com>
Date: Thu, 18 Apr 2024 15:49:52 +0800
Subject: [PATCH 847/886] [clang] Remove unused lambda capture (NFC)

llvm-project/clang/lib/Serialization/ASTWriter.cpp:5077:59:
error: lambda capture 'SemaRef' is not used [-Werror,-Wunused-lambda-capture]
    auto AddEmittedDeclRefOrZero = [this, &SemaDeclRefs, &SemaRef](Decl *D) {
                                                       ~~~^~~~~~~
1 error generated.
---
 clang/lib/Serialization/ASTWriter.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 522b8ce56c326..2eb4adac53de2 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -5074,7 +5074,7 @@ void ASTWriter::WriteSpecialDeclRecords(Sema &SemaRef) {
   // Write the record containing declaration references of Sema.
   RecordData SemaDeclRefs;
   if (SemaRef.StdNamespace || SemaRef.StdBadAlloc || SemaRef.StdAlignValT) {
-    auto AddEmittedDeclRefOrZero = [this, &SemaDeclRefs, &SemaRef](Decl *D) {
+    auto AddEmittedDeclRefOrZero = [this, &SemaDeclRefs](Decl *D) {
       if (!D || !wasDeclEmitted(D))
         SemaDeclRefs.push_back(0);
       else

From c82f45f9deb98aa383b0be37f8edc32f9e48c4ec Mon Sep 17 00:00:00 2001
From: Guray Ozen <guray.ozen@gmail.com>
Date: Thu, 18 Apr 2024 09:58:24 +0200
Subject: [PATCH 848/886] [mlir][nvgpu] Simplify TMA IR generation (#87153)

This PR add `TmaDescriptorBuilder`
- class simplifies TMA generation.
- Makes the code ready to support various Tma configurations
   - removes strings and use the enums from `mlir.nvgpu.ENUMs`.
- Example "swizzle = swizzle_128b, l2promo=none, oob=zero,
interleave=none" to enums in `mlir.nvgpu` dialects.
- Enums have string equivalent that are used during the IR writing and
generation (see `TmaDescriptorBuilder::tensormap_descriptor_ty`).
- Improves readability and abstracts out TMA descriptor builders in
reusable component.

---------

Co-authored-by: Manish Gupta <manigupta@google.com>
---
 .../CUDA/sm90/python/tools/matmulBuilder.py   | 194 +++++++++---------
 1 file changed, 101 insertions(+), 93 deletions(-)

diff --git a/mlir/test/Integration/GPU/CUDA/sm90/python/tools/matmulBuilder.py b/mlir/test/Integration/GPU/CUDA/sm90/python/tools/matmulBuilder.py
index fac138dce605a..5269522000f13 100644
--- a/mlir/test/Integration/GPU/CUDA/sm90/python/tools/matmulBuilder.py
+++ b/mlir/test/Integration/GPU/CUDA/sm90/python/tools/matmulBuilder.py
@@ -28,6 +28,48 @@
 DEBUG = False
 
 
+class TmaDescriptorBuilder:
+    """A class that builds a TMA descriptor."""
+
+    def __init__(self, swizzle, l2promo, oob, interleave, tma_box_shape, memref_ty):
+        self.swizzle = swizzle  # mlir.nvgpu.TensorMapSwizzleKind
+        self.l2promo = l2promo  # mlir.nvgpu.TensorMapL2PromoKind
+        self.oob = oob  # mlir.nvgpu.TensorMapOOBKind
+        self.interleave = interleave  # mlir.nvgpu.TensorMapInterleaveKind
+        self.tma_box_shape = tma_box_shape
+        self.memref_ty = memref_ty  # MemRefType
+
+    @property
+    def tensormap_descriptor_ty(self):
+        """Returns a tensormap descriptor type."""
+        tensorMemrefType = ir.MemRefType.get(
+            self.tma_box_shape,
+            self.memref_ty.element_type,
+            memory_space=ir.Attribute.parse("3"),
+        )
+        return nvgpu.TensorMapDescriptorType.get(
+            tensorMemrefType,
+            self.swizzle,
+            self.l2promo,
+            self.oob,
+            self.interleave,
+        )
+
+    def tma_descriptor_op(self, device_ptr):
+        """Returns a tensormap descriptor op."""
+        tma_descriptor_ty = self.tensormap_descriptor_ty
+        device_unranked_memref = memref.CastOp(
+            ir.UnrankedMemRefType.get(
+                self.memref_ty.element_type, self.memref_ty.memory_space
+            ),
+            device_ptr,
+        )
+        tma_descriptor_op = nvgpu.TmaCreateDescriptorOp(
+            tma_descriptor_ty, device_unranked_memref, map(c, self.tma_box_shape)
+        )
+        return tma_descriptor_op.result
+
+
 def debug_print(fmt, *args, predicate=None, threadNumber=-1, forcePrint=False):
     if not DEBUG and not forcePrint:
         return
@@ -162,28 +204,6 @@ def generate_matmul_ws(
         + str(num_stages)
         + ">"
     )
-    a_tma_desc_ty = ir.Type.parse(
-        "!nvgpu.tensormap.descriptor<tensor = memref<"
-        + str(BLOCK_M)
-        + "x"
-        + str(TMA_LAST_DIM_F16)
-        + "x"
-        + str(a_elem_ty)
-        + ", "
-        + str(smem_space)
-        + ">, swizzle = swizzle_128b, l2promo=none, oob=zero, interleave=none>"
-    )
-    b_tma_desc_ty = ir.Type.parse(
-        "!nvgpu.tensormap.descriptor<tensor = memref<"
-        + str(BLOCK_K)
-        + "x"
-        + str(TMA_LAST_DIM_F16)
-        + "x"
-        + str(b_elem_ty)
-        + ", "
-        + str(smem_space)
-        + ">, swizzle = swizzle_128b, l2promo=none, oob=zero, interleave=none>"
-    )
     acc_ty = ir.Type.parse(
         "!nvgpu.warpgroup.accumulator<fragmented=vector<"
         + str(BLOCK_M)
@@ -240,21 +260,26 @@ def generate_matmul_ws(
             t7 = gpu.wait(token_ty, [t6])
 
             # Step 2. Create TMA Descriptors
-            tma_specs = [
-                (a_device, a_tma_desc_ty, a_tma_shape),
-                (b_device, b_tma_desc_ty, b_tma_shape),
-            ]
-            tma_descs = []
-            for x_device, tensor_map_ty, tile_shape in tma_specs:
-                x_unranked = memref.cast(
-                    ir.UnrankedMemRefType.get(a_elem_ty, a_ty.memory_space), x_device
-                )
-                tma_descs.append(
-                    nvgpu.TmaCreateDescriptorOp(
-                        tensor_map_ty, x_unranked, map(c, tile_shape)
-                    ).result
-                )
-            a_tma_desc, b_tma_desc = tma_descs
+            a_tma_desc = TmaDescriptorBuilder(
+                nvgpu.TensorMapSwizzleKind.SWIZZLE_128B,
+                nvgpu.TensorMapL2PromoKind.L2PROMO_NONE,
+                nvgpu.TensorMapOOBKind.OOB_ZERO,
+                nvgpu.TensorMapInterleaveKind.INTERLEAVE_NONE,
+                a_tma_shape,
+                a_ty,
+            )
+
+            b_tma_desc = TmaDescriptorBuilder(
+                nvgpu.TensorMapSwizzleKind.SWIZZLE_128B,
+                nvgpu.TensorMapL2PromoKind.L2PROMO_NONE,
+                nvgpu.TensorMapOOBKind.OOB_ZERO,
+                nvgpu.TensorMapInterleaveKind.INTERLEAVE_NONE,
+                b_tma_shape,
+                b_ty,
+            )
+
+            a_tma_desc_op = a_tma_desc.tma_descriptor_op(a_device)
+            b_tma_desc_op = b_tma_desc.tma_descriptor_op(b_device)
 
             # Step 3. Launch Kernel with 2 Warpgroups : 1 Producer, 1 Consumer
             cta_m = M // BLOCK_M
@@ -267,7 +292,7 @@ def generate_matmul_ws(
                 [t7],
                 *map(c, grid),
                 *map(c, block),
-                dynamicSharedMemorySize=c(smem_size, ty=T.i32())
+                dynamicSharedMemorySize=c(smem_size, ty=T.i32()),
             )
             launch_op.body.blocks.append(*([T.index()] * 12))
             with ir.InsertionPoint(launch_op.body.blocks[0]):
@@ -315,8 +340,8 @@ def generate_matmul_ws(
                 gpu.barrier()
 
                 # GPU Step 3. Prefetch TMA descriptors
-                nvgpu.tma_prefetch_descriptor(a_tma_desc, predicate=wgPrimaryThread)
-                nvgpu.tma_prefetch_descriptor(b_tma_desc, predicate=wgPrimaryThread)
+                nvgpu.tma_prefetch_descriptor(a_tma_desc_op, predicate=wgPrimaryThread)
+                nvgpu.tma_prefetch_descriptor(b_tma_desc_op, predicate=wgPrimaryThread)
 
                 ns = num_stages if num_stages == 1 else num_stages - 1
                 # GPU Step 5. Producer Warpgroup (TMA Warpgroup)
@@ -405,7 +430,7 @@ def generate_matmul_ws(
                         nvgpu.TmaAsyncLoadOp(
                             a_tma_slice,
                             mbarTMA,
-                            a_tma_desc,
+                            a_tma_desc_op,
                             coordinates=[coord, dimX],
                             mbarId=stage,
                             predicate=producerPrimaryThread,
@@ -413,7 +438,7 @@ def generate_matmul_ws(
                         nvgpu.TmaAsyncLoadOp(
                             b_tma_slice_1,
                             mbarTMA,
-                            b_tma_desc,
+                            b_tma_desc_op,
                             coordinates=[dimY, coord],
                             mbarId=stage,
                             predicate=producerPrimaryThread,
@@ -422,7 +447,7 @@ def generate_matmul_ws(
                         nvgpu.TmaAsyncLoadOp(
                             b_tma_slice_2,
                             mbarTMA,
-                            b_tma_desc,
+                            b_tma_desc_op,
                             coordinates=[dimY2, coord],
                             mbarId=stage,
                             predicate=producerPrimaryThread,
@@ -514,10 +539,10 @@ def generate_matmul_ws(
                             predicate=consumerPrimaryThread,
                         )
                         da = nvgpu.WarpgroupGenerateDescriptorOp(
-                            a_wgmma_ty, a_tile_slice, a_tma_desc
+                            a_wgmma_ty, a_tile_slice, a_tma_desc_op
                         )
                         db = nvgpu.WarpgroupGenerateDescriptorOp(
-                            b_wgmma_ty, b_tile_slice, b_tma_desc
+                            b_wgmma_ty, b_tile_slice, b_tma_desc_op
                         )
 
                         # Step 6.3.3. MMA
@@ -679,28 +704,6 @@ def generate_matmul_multistage(
         + str(num_stages)
         + ">"
     )
-    a_tma_desc_ty = ir.Type.parse(
-        "!nvgpu.tensormap.descriptor<tensor = memref<"
-        + str(BLOCK_M)
-        + "x"
-        + str(TMA_LAST_DIM_F16)
-        + "x"
-        + str(a_elem_ty)
-        + ", "
-        + str(smem_space)
-        + ">, swizzle = swizzle_128b, l2promo=none, oob=zero, interleave=none>"
-    )
-    b_tma_desc_ty = ir.Type.parse(
-        "!nvgpu.tensormap.descriptor<tensor = memref<"
-        + str(BLOCK_K)
-        + "x"
-        + str(TMA_LAST_DIM_F16)
-        + "x"
-        + str(b_elem_ty)
-        + ", "
-        + str(smem_space)
-        + ">, swizzle = swizzle_128b, l2promo=none, oob=zero, interleave=none>"
-    )
     acc_ty = ir.Type.parse(
         "!nvgpu.warpgroup.accumulator<fragmented=vector<"
         + str(BLOCK_M)
@@ -767,21 +770,26 @@ def generate_matmul_multistage(
             t7 = gpu.wait(token_ty, [t6])
 
             # Step 2. Create TMA Descriptors
-            tma_specs = [
-                (a_device, a_tma_desc_ty, a_tma_shape),
-                (b_device, b_tma_desc_ty, b_tma_shape),
-            ]
-            tma_descs = []
-            for x_device, tensor_map_ty, tile_shape in tma_specs:
-                x_unranked = memref.cast(
-                    ir.UnrankedMemRefType.get(a_elem_ty, a_ty.memory_space), x_device
-                )
-                tma_descs.append(
-                    nvgpu.TmaCreateDescriptorOp(
-                        tensor_map_ty, x_unranked, map(c, tile_shape)
-                    ).result
-                )
-            a_tma_desc, b_tma_desc = tma_descs
+            a_tma_desc = TmaDescriptorBuilder(
+                nvgpu.TensorMapSwizzleKind.SWIZZLE_128B,
+                nvgpu.TensorMapL2PromoKind.L2PROMO_NONE,
+                nvgpu.TensorMapOOBKind.OOB_ZERO,
+                nvgpu.TensorMapInterleaveKind.INTERLEAVE_NONE,
+                a_tma_shape,
+                a_ty,
+            )
+
+            b_tma_desc = TmaDescriptorBuilder(
+                nvgpu.TensorMapSwizzleKind.SWIZZLE_128B,
+                nvgpu.TensorMapL2PromoKind.L2PROMO_NONE,
+                nvgpu.TensorMapOOBKind.OOB_ZERO,
+                nvgpu.TensorMapInterleaveKind.INTERLEAVE_NONE,
+                b_tma_shape,
+                b_ty,
+            )
+
+            a_tma_desc_op = a_tma_desc.tma_descriptor_op(a_device)
+            b_tma_desc_op = b_tma_desc.tma_descriptor_op(b_device)
 
             # Step 3. Launch Kernel with 1 Warpgroup
             cta_m = M // BLOCK_M
@@ -794,7 +802,7 @@ def generate_matmul_multistage(
                 [t7],
                 *map(c, grid),
                 *map(c, block),
-                dynamicSharedMemorySize=c(smem_size, ty=T.i32())
+                dynamicSharedMemorySize=c(smem_size, ty=T.i32()),
             )
             launch_op.body.blocks.append(*([T.index()] * 12))
             with ir.InsertionPoint(launch_op.body.blocks[0]):
@@ -819,8 +827,8 @@ def generate_matmul_multistage(
                 gpu.barrier()
 
                 # GPU Step 2. Prefetch TMA descriptors
-                nvgpu.tma_prefetch_descriptor(a_tma_desc, predicate=primaryThread)
-                nvgpu.tma_prefetch_descriptor(b_tma_desc, predicate=primaryThread)
+                nvgpu.tma_prefetch_descriptor(a_tma_desc_op, predicate=primaryThread)
+                nvgpu.tma_prefetch_descriptor(b_tma_desc_op, predicate=primaryThread)
 
                 # GPU Step 3. Prologue (global memory --> shared memory)
                 ns = num_stages if num_stages == 1 else num_stages - 1
@@ -880,7 +888,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         a_tma_slice,
                         mbarTMA,
-                        a_tma_desc,
+                        a_tma_desc_op,
                         coordinates=[coord, dimX],
                         mbarId=iv,
                         predicate=primaryThread,
@@ -888,7 +896,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         b_tma_slice_1,
                         mbarTMA,
-                        b_tma_desc,
+                        b_tma_desc_op,
                         coordinates=[dimY, coord],
                         mbarId=iv,
                         predicate=primaryThread,
@@ -896,7 +904,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         b_tma_slice_2,
                         mbarTMA,
-                        b_tma_desc,
+                        b_tma_desc_op,
                         coordinates=[dimY2, coord],
                         mbarId=iv,
                         predicate=primaryThread,
@@ -972,10 +980,10 @@ def generate_matmul_multistage(
                         predicate=primaryThread,
                     )
                     da = nvgpu.WarpgroupGenerateDescriptorOp(
-                        a_wgmma_ty, a_tile_slice, a_tma_desc
+                        a_wgmma_ty, a_tile_slice, a_tma_desc_op
                     )
                     db = nvgpu.WarpgroupGenerateDescriptorOp(
-                        b_wgmma_ty, b_tile_slice, b_tma_desc
+                        b_wgmma_ty, b_tile_slice, b_tma_desc_op
                     )
 
                     # Step 4.3. MMA
@@ -1060,7 +1068,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         a_tma_slice,
                         mbarTMA,
-                        a_tma_desc,
+                        a_tma_desc_op,
                         coordinates=[coord, dimX],
                         mbarId=nextSlot,
                         predicate=p,
@@ -1068,7 +1076,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         b_tma_slice_1,
                         mbarTMA,
-                        b_tma_desc,
+                        b_tma_desc_op,
                         coordinates=[dimY, coord],
                         mbarId=nextSlot,
                         predicate=p,
@@ -1077,7 +1085,7 @@ def generate_matmul_multistage(
                     nvgpu.TmaAsyncLoadOp(
                         b_tma_slice_2,
                         mbarTMA,
-                        b_tma_desc,
+                        b_tma_desc_op,
                         coordinates=[dimY2, coord],
                         mbarId=nextSlot,
                         predicate=p,

From 03e841c870e7beccf1368e566045e233b3bd8db5 Mon Sep 17 00:00:00 2001
From: David Spickett <david.spickett@linaro.org>
Date: Thu, 18 Apr 2024 09:08:53 +0100
Subject: [PATCH 849/886] [lldb] Correct documentation of LLDB_TEST_USER_ARGS
 (#89042)

https://discourse.llvm.org/t/running-lldb-in-a-container/76801/4 found
that the obvious way to use this variable doesn't work, despite what our
docs and examples say.

Perhaps in the past it did but now you need to use ";" as a separator to
make sure the final command line works properly.

For example "-A foo" becomes "-A foo" when python goes to run the runner
script. The script sees this as one command line element, not two. What
you actually want is "-A;foo" which we convert to "-A" "foo" which the
script sees as one option and one value for that option.

The "Script:" printout from dotest is misleading here because it does `"
".join(cmd)` so it looks like it's ok but in fact it's not.

I'm not changing that format though because printing the command as a
Python list is not useful outside of this specific situation.
---
 lldb/docs/resources/test.rst | 25 +++++++++++++++++++++----
 lldb/test/API/CMakeLists.txt |  3 ++-
 2 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/lldb/docs/resources/test.rst b/lldb/docs/resources/test.rst
index 2b0e9010fe280..094fde8b1b5a0 100644
--- a/lldb/docs/resources/test.rst
+++ b/lldb/docs/resources/test.rst
@@ -441,23 +441,40 @@ Running the Full Test Suite
 The easiest way to run the LLDB test suite is to use the ``check-lldb`` build
 target.
 
+::
+
+   $ ninja check-lldb
+
+Changing Test Suite Options
+```````````````````````````
+
 By default, the ``check-lldb`` target builds the test programs with the same
 compiler that was used to build LLDB. To build the tests with a different
 compiler, you can set the ``LLDB_TEST_COMPILER`` CMake variable.
 
+You can also add to the test runner options by setting the
+``LLDB_TEST_USER_ARGS`` CMake variable. This variable uses ``;`` to separate
+items which must be separate parts of the runner's command line.
+
 It is possible to customize the architecture of the test binaries and compiler
-used by appending ``-A`` and ``-C`` options respectively to the CMake variable
-``LLDB_TEST_USER_ARGS``. For example, to test LLDB against 32-bit binaries
-built with a custom version of clang, do:
+used by appending ``-A`` and ``-C`` options respectively. For example, to test
+LLDB against 32-bit binaries built with a custom version of clang, do:
 
 ::
 
-   $ cmake -DLLDB_TEST_USER_ARGS="-A i386 -C /path/to/custom/clang" -G Ninja
+   $ cmake -DLLDB_TEST_USER_ARGS="-A;i386;-C;/path/to/custom/clang" -G Ninja
    $ ninja check-lldb
 
 Note that multiple ``-A`` and ``-C`` flags can be specified to
 ``LLDB_TEST_USER_ARGS``.
 
+If you want to change the LLDB settings that tests run with then you can set
+the ``--setting`` option of the test runner via this same variable. For example
+``--setting;target.disable-aslr=true``.
+
+For a full list of test runner options, see
+``<build-dir>/bin/lldb-dotest --help``.
+
 Running a Single Test Suite
 ```````````````````````````
 
diff --git a/lldb/test/API/CMakeLists.txt b/lldb/test/API/CMakeLists.txt
index 4e02112c29e4c..9196f54ce1ae3 100644
--- a/lldb/test/API/CMakeLists.txt
+++ b/lldb/test/API/CMakeLists.txt
@@ -35,7 +35,8 @@ set(LLDB_TEST_ARCH
 # Users can override LLDB_TEST_USER_ARGS to specify arbitrary arguments to pass to the script
 set(LLDB_TEST_USER_ARGS
   ""
-  CACHE STRING "Specify additional arguments to pass to test runner. For example: '-C gcc -C clang -A i386 -A x86_64'")
+  CACHE STRING "Specify additional arguments to pass to test runner. Seperate \
+items with \";\". For example: '-C;gcc;-C;clang;-A;i386;-A;x86_64'")
 
 set(LLDB_TEST_COMMON_ARGS_VAR
   -u CXXFLAGS

From e7a8dd9b0d419403fe1d8adeb177a4ec78e036cc Mon Sep 17 00:00:00 2001
From: Chuanqi Xu <yedeng.yd@linux.alibaba.com>
Date: Thu, 18 Apr 2024 16:09:39 +0800
Subject: [PATCH 850/886] [docs] [C++20] [Modules] Mentioning Module
 Initializer

Although we want to treat the module initializer as a transparent
concept to users, but it shows that people need to understand
the concept to understand how to understand and distribute modules.

So it is better to mention this too.
---
 clang/docs/StandardCPlusPlusModules.rst | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/clang/docs/StandardCPlusPlusModules.rst b/clang/docs/StandardCPlusPlusModules.rst
index 8d5529d5d37db..ee57fb5da6485 100644
--- a/clang/docs/StandardCPlusPlusModules.rst
+++ b/clang/docs/StandardCPlusPlusModules.rst
@@ -483,6 +483,13 @@ violations with the flag enabled.
 ABI Impacts
 -----------
 
+This section describes the new ABI changes brought by modules.
+
+Only Itanium C++ ABI related change are mentioned
+
+Mangling Names
+~~~~~~~~~~~~~~
+
 The declarations in a module unit which are not in the global module fragment have new linkage names.
 
 For example,
@@ -520,6 +527,23 @@ is attached to the global module fragments. For example:
 
 Now the linkage name of ``NS::foo()`` will be ``_ZN2NS3fooEv``.
 
+Module Initializers
+~~~~~~~~~~~~~~~~~~~
+
+All the importable module units are required to emit an initializer function.
+The initializer function should contain calls to importing modules first and
+all the dynamic-initializers in the current module unit then.
+
+Translation units explicitly or implicitly importing named modules must call
+the initializer functions of the imported named modules within the sequence of
+the dynamic-initializers in the TU. Initializations of entities at namespace
+scope are appearance-ordered. This (recursively) extends into imported modules
+at the point of appearance of the import declaration.
+
+It is allowed to omit calls to importing modules if it is known empty.
+
+It is allowed to omit calls to importing modules for which is known to be called.
+
 Reduced BMI
 -----------
 

From c674dbc2a89298f1ff06686ae953cef9a1b19961 Mon Sep 17 00:00:00 2001
From: David Spickett <david.spickett@linaro.org>
Date: Thu, 18 Apr 2024 09:25:47 +0100
Subject: [PATCH 851/886] Revert "[FMV] Remove useless features according the
 latest ACLE spec." (#89184)

Reverts llvm/llvm-project#88965

This caused a test suite failure:
https://lab.llvm.org/buildbot/#/builders/185/builds/6583
NOEXE: test-suite::aarch64-acle-fmv-features.test

```
/home/tcwg-buildbot/worker/clang-aarch64-lld-2stage/test/test-suite/SingleSource/UnitTests/AArch64/acle-fmv-features.c:98:1: error: redefinition of 'check_sha1'
   98 | CHECK(sha1, {
      | ^
/home/tcwg-buildbot/worker/clang-aarch64-lld-2stage/test/test-suite/SingleSource/UnitTests/AArch64/acle-fmv-features.c:36:17: note: expanded from macro 'CHECK'
   36 |     static void check_##X(void) { \
      |                 ^
<scratch space>:150:1: note: expanded from here
  150 | check_sha1
      | ^
```

I presume that the useless features need to be removed from the fmv test
as well.
---
 clang/test/CodeGen/aarch64-cpu-supports.c     |  24 +-
 .../CodeGen/aarch64-mixed-target-attributes.c |  12 +-
 .../test/CodeGen/attr-target-clones-aarch64.c |  77 +++--
 clang/test/CodeGen/attr-target-version.c      | 326 +++++++++---------
 .../CodeGenCXX/attr-target-clones-aarch64.cpp |  41 ++-
 clang/test/CodeGenCXX/attr-target-version.cpp |  22 +-
 clang/test/Sema/aarch64-cpu-supports.c        |   2 +-
 clang/test/Sema/attr-target-clones-aarch64.c  |  21 +-
 clang/test/Sema/attr-target-version.c         |  18 +-
 clang/test/SemaCXX/attr-target-version.cpp    |  16 +-
 compiler-rt/lib/builtins/cpu_model/aarch64.c  |  17 +-
 .../builtins/cpu_model/aarch64/fmv/apple.inc  |   6 +-
 .../cpu_model/aarch64/fmv/fuchsia.inc         |   4 +
 .../builtins/cpu_model/aarch64/fmv/mrs.inc    |  42 ++-
 .../llvm/TargetParser/AArch64TargetParser.h   |  36 +-
 15 files changed, 377 insertions(+), 287 deletions(-)

diff --git a/clang/test/CodeGen/aarch64-cpu-supports.c b/clang/test/CodeGen/aarch64-cpu-supports.c
index 7fad9724dfb6c..c54b7475a3fd5 100644
--- a/clang/test/CodeGen/aarch64-cpu-supports.c
+++ b/clang/test/CodeGen/aarch64-cpu-supports.c
@@ -1,17 +1,15 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals --version 2
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -emit-llvm -o - %s | FileCheck %s
 
-//.
 // CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
-//.
 // CHECK-LABEL: define dso_local i32 @main
 // CHECK-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34359738368
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34359738368
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70368744177664
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70368744177664
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 // CHECK:       if.then:
@@ -19,8 +17,8 @@
 // CHECK-NEXT:    br label [[RETURN:%.*]]
 // CHECK:       if.end:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 17716740096
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 17716740096
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 9070970929152
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 9070970929152
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[IF_THEN1:%.*]], label [[IF_END2:%.*]]
 // CHECK:       if.then1:
@@ -28,8 +26,8 @@
 // CHECK-NEXT:    br label [[RETURN]]
 // CHECK:       if.end2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 5222680231936
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 5222680231936
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 166633186212708352
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 166633186212708352
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[IF_THEN3:%.*]], label [[IF_END4:%.*]]
 // CHECK:       if.then3:
@@ -51,10 +49,10 @@ int main(void) {
   if (__builtin_cpu_supports("sb"))
     return 1;
 
-  if (__builtin_cpu_supports("sve2-aes+memtag"))
+  if (__builtin_cpu_supports("sve2-pmull128+memtag"))
     return 2;
 
-  if (__builtin_cpu_supports("sme2+ls64+wfxt"))
+  if (__builtin_cpu_supports("sme2+ls64_v+wfxt"))
     return 3;
 
   if (__builtin_cpu_supports("avx2"))
@@ -62,9 +60,3 @@ int main(void) {
 
   return 0;
 }
-//.
-// CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
-//.
-// CHECK: [[META0:![0-9]+]] = !{i32 1, !"wchar_size", i32 4}
-// CHECK: [[META1:![0-9]+]] = !{!"{{.*}}clang version {{.*}}"}
-//.
diff --git a/clang/test/CodeGen/aarch64-mixed-target-attributes.c b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
index be290ff9ecee6..aef6ce36ab1c0 100644
--- a/clang/test/CodeGen/aarch64-mixed-target-attributes.c
+++ b/clang/test/CodeGen/aarch64-mixed-target-attributes.c
@@ -69,8 +69,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -143,8 +143,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -210,8 +210,8 @@ __attribute__((target_version("jscvt"))) int default_def_with_version_decls(void
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
diff --git a/clang/test/CodeGen/attr-target-clones-aarch64.c b/clang/test/CodeGen/attr-target-clones-aarch64.c
index c715001f6a722..8c8b951e9118d 100644
--- a/clang/test/CodeGen/attr-target-clones-aarch64.c
+++ b/clang/test/CodeGen/attr-target-clones-aarch64.c
@@ -3,7 +3,7 @@
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature -fmv -S -emit-llvm -o - %s | FileCheck %s -check-prefix=CHECK-NOFMV
 
 int __attribute__((target_clones("lse+aes", "sve2"))) ftc(void) { return 0; }
-int __attribute__((target_clones("sha2", "sha2+memtag", " default "))) ftc_def(void) { return 1; }
+int __attribute__((target_clones("sha2", "sha2+memtag2", " default "))) ftc_def(void) { return 1; }
 int __attribute__((target_clones("sha2", "default"))) ftc_dup1(void) { return 2; }
 int __attribute__((target_clones("fp", "crc+dotprod"))) ftc_dup2(void) { return 3; }
 int foo() {
@@ -12,7 +12,7 @@ int foo() {
 
 inline int __attribute__((target_clones("rng+simd", "rcpc+predres", "sve2-aes+wfxt"))) ftc_inline1(void) { return 1; }
 inline int __attribute__((target_clones("fp16", "fcma+sve2-bitperm", "default"))) ftc_inline2(void);
-inline int __attribute__((target_clones("mops", "sve+sb"))) ftc_inline3(void) { return 3; }
+inline int __attribute__((target_clones("bti", "sve+sb"))) ftc_inline3(void) { return 3; }
 
 int __attribute__((target_clones("default"))) ftc_direct(void) { return 4; }
 
@@ -56,16 +56,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8320
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8320
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16512
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16512
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc._MaesMlse
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 268435456
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 268435456
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 68719476736
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 68719476736
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -81,7 +81,7 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: @ftc_def._MmemtagMsha2(
+// CHECK-LABEL: @ftc_def._Mmemtag2Msha2(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
@@ -90,16 +90,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 17179871232
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 17179871232
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 17592186048512
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 17592186048512
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @ftc_def._MmemtagMsha2
+// CHECK-NEXT:    ret ptr @ftc_def._Mmemtag2Msha2
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2048
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2048
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 4096
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 4096
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -118,8 +118,8 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2048
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2048
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4096
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4096
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -198,16 +198,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 550292684800
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 550292684800
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 18014535948435456
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 18014535948435456
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline1._Msve2-aesMwfxt
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 68720001024
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 68720001024
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 140737492549632
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 140737492549632
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -228,16 +228,16 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1074003968
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1074003968
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 549757911040
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 549757911040
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline2._MfcmaMsve2-bitperm
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16384
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16384
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 65536
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 65536
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -250,20 +250,20 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34393292800
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34393292800
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70369817919488
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70369817919488
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @ftc_inline3._MsbMsve
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 17592186044416
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 17592186044416
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 1125899906842624
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 1125899906842624
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @ftc_inline3._Mmops
+// CHECK-NEXT:    ret ptr @ftc_inline3._Mbti
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    ret ptr @ftc_inline3.default
 //
@@ -329,7 +329,7 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: @ftc_inline3._Mmops(
+// CHECK-LABEL: @ftc_inline3._Mbti(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -407,16 +407,17 @@ inline int __attribute__((target_clones("fp16", "sve2-bitperm+fcma", "default"))
 // CHECK: attributes #[[ATTR0:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+lse,+neon" }
 // CHECK: attributes #[[ATTR1:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2" }
 // CHECK: attributes #[[ATTR2:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+sha2" }
-// CHECK: attributes #[[ATTR3:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
-// CHECK: attributes #[[ATTR4:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+dotprod,+fp-armv8,+neon" }
-// CHECK: attributes #[[ATTR5:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
-// CHECK: attributes #[[ATTR6:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
-// CHECK: attributes #[[ATTR7:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+complxnum,+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-bitperm" }
-// CHECK: attributes #[[ATTR8:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rand" }
-// CHECK: attributes #[[ATTR9:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+predres,+rcpc" }
-// CHECK: attributes #[[ATTR10:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+wfxt" }
-// CHECK: attributes #[[ATTR11:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops" }
-// CHECK: attributes #[[ATTR12:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sb,+sve" }
+// CHECK: attributes #[[ATTR3:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+mte,+neon,+sha2" }
+// CHECK: attributes #[[ATTR4:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon" }
+// CHECK: attributes #[[ATTR5:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,+dotprod,+fp-armv8,+neon" }
+// CHECK: attributes #[[ATTR6:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
+// CHECK: attributes #[[ATTR7:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
+// CHECK: attributes #[[ATTR8:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+complxnum,+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-bitperm" }
+// CHECK: attributes #[[ATTR9:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+neon,+rand" }
+// CHECK: attributes #[[ATTR10:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+predres,+rcpc" }
+// CHECK: attributes #[[ATTR11:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sve,+sve2,+sve2-aes,+wfxt" }
+// CHECK: attributes #[[ATTR12:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bti" }
+// CHECK: attributes #[[ATTR13:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon,+sb,+sve" }
 //.
 // CHECK-NOFMV: attributes #[[ATTR0:[0-9]+]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
 // CHECK-NOFMV: attributes #[[ATTR1:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
diff --git a/clang/test/CodeGen/attr-target-version.c b/clang/test/CodeGen/attr-target-version.c
index e71370e6d91df..dd4cbbf5a8986 100644
--- a/clang/test/CodeGen/attr-target-version.c
+++ b/clang/test/CodeGen/attr-target-version.c
@@ -5,11 +5,11 @@
 int __attribute__((target_version("rng+flagm+fp16fml"))) fmv(void) { return 1; }
 int __attribute__((target_version("flagm2+sme-i16i64"))) fmv(void) { return 2; }
 int __attribute__((target_version("lse+sha2"))) fmv(void) { return 3; }
-int __attribute__((target_version("dotprod+ls64"))) fmv(void) { return 4; }
+int __attribute__((target_version("dotprod+ls64_accdata"))) fmv(void) { return 4; }
 int __attribute__((target_version("fp16fml+memtag"))) fmv(void) { return 5; }
 int __attribute__((target_version("fp+aes"))) fmv(void) { return 6; }
-int __attribute__((target_version("crc+ls64"))) fmv(void) { return 7; }
-int __attribute__((target_version("mops"))) fmv(void) { return 8; }
+int __attribute__((target_version("crc+ls64_v"))) fmv(void) { return 7; }
+int __attribute__((target_version("bti"))) fmv(void) { return 8; }
 int __attribute__((target_version("sme2"))) fmv(void) { return 9; }
 int __attribute__((target_version("default"))) fmv(void);
 int __attribute__((target_version("ls64+simd"))) fmv_one(void) { return 1; }
@@ -17,25 +17,25 @@ int __attribute__((target_version("dpb"))) fmv_one(void) { return 2; }
 int __attribute__((target_version("default"))) fmv_one(void);
 int __attribute__((target_version("fp"))) fmv_two(void) { return 1; }
 int __attribute__((target_version("simd"))) fmv_two(void) { return 2; }
-int __attribute__((target_version("frintts"))) fmv_two(void) { return 3; }
+int __attribute__((target_version("dgh"))) fmv_two(void) { return 3; }
 int __attribute__((target_version("fp16+simd"))) fmv_two(void) { return 4; }
 int __attribute__((target_version("default"))) fmv_two(void);
 int foo() {
   return fmv()+fmv_one()+fmv_two();
 }
 
-inline int __attribute__((target_version("crypto+f64mm"))) fmv_inline(void) { return 1; }
+inline int __attribute__((target_version("sha1+pmull+f64mm"))) fmv_inline(void) { return 1; }
 inline int __attribute__((target_version("fp16+fcma+rdma+sme+ fp16 "))) fmv_inline(void) { return 2; }
 inline int __attribute__((target_version("sha3+i8mm+f32mm"))) fmv_inline(void) { return 12; }
-inline int __attribute__((target_version("sme2+ssbs"))) fmv_inline(void) { return 8; }
+inline int __attribute__((target_version("dit+sve-ebf16"))) fmv_inline(void) { return 8; }
 inline int __attribute__((target_version("dpb+rcpc2 "))) fmv_inline(void) { return 6; }
 inline int __attribute__((target_version(" dpb2 + jscvt"))) fmv_inline(void) { return 7; }
 inline int __attribute__((target_version("rcpc+frintts"))) fmv_inline(void) { return 3; }
-inline int __attribute__((target_version("sve+sme"))) fmv_inline(void) { return 4; }
+inline int __attribute__((target_version("sve+sve-bf16"))) fmv_inline(void) { return 4; }
 inline int __attribute__((target_version("sve2-aes+sve2-sha3"))) fmv_inline(void) { return 5; }
-inline int __attribute__((target_version("sve2+sve2-aes+sve2-bitperm"))) fmv_inline(void) { return 9; }
-inline int __attribute__((target_version("sve2-sm4+memtag"))) fmv_inline(void) { return 10; }
-inline int __attribute__((target_version("memtag+rcpc3+mops"))) fmv_inline(void) { return 11; }
+inline int __attribute__((target_version("sve2+sve2-pmull128+sve2-bitperm"))) fmv_inline(void) { return 9; }
+inline int __attribute__((target_version("sve2-sm4+memtag2"))) fmv_inline(void) { return 10; }
+inline int __attribute__((target_version("memtag3+rcpc3+mops"))) fmv_inline(void) { return 11; }
 inline int __attribute__((target_version("aes+dotprod"))) fmv_inline(void) { return 13; }
 inline int __attribute__((target_version("simd+fp16fml"))) fmv_inline(void) { return 14; }
 inline int __attribute__((target_version("fp+sm4"))) fmv_inline(void) { return 15; }
@@ -186,7 +186,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._MdotprodMls64
+// CHECK-LABEL: define {{[^@]+}}@fmv._MdotprodMls64_accdata
 // CHECK-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 4
@@ -207,14 +207,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._McrcMls64
+// CHECK-LABEL: define {{[^@]+}}@fmv._McrcMls64_v
 // CHECK-SAME: () #[[ATTR6:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 7
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv._Mmops
+// CHECK-LABEL: define {{[^@]+}}@fmv._Mbti
 // CHECK-SAME: () #[[ATTR7:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 8
@@ -256,7 +256,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_two._Mfrintts
+// CHECK-LABEL: define {{[^@]+}}@fmv_two._Mdgh
 // CHECK-SAME: () #[[ATTR11:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
@@ -271,7 +271,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@foo
-// CHECK-SAME: () #[[ATTR13:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @fmv()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @fmv_one()
@@ -293,68 +293,68 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:    ret ptr @fmv._MflagmMfp16fmlMrng
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2199023255556
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2199023255556
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 72057594037927940
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 72057594037927940
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
 // CHECK-NEXT:    ret ptr @fmv._Mflagm2Msme-i16i64
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 274877906960
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 274877906960
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 9007199254741008
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 9007199254741008
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[RESOLVER_RETURN3:%.*]], label [[RESOLVER_ELSE4:%.*]]
 // CHECK:       resolver_return3:
-// CHECK-NEXT:    ret ptr @fmv._MdotprodMls64
+// CHECK-NEXT:    ret ptr @fmv._MdotprodMls64_accdata
 // CHECK:       resolver_else4:
 // CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 274877907968
-// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 274877907968
+// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 4503599627371520
+// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 4503599627371520
 // CHECK-NEXT:    [[TMP15:%.*]] = and i1 true, [[TMP14]]
 // CHECK-NEXT:    br i1 [[TMP15]], label [[RESOLVER_RETURN5:%.*]], label [[RESOLVER_ELSE6:%.*]]
 // CHECK:       resolver_return5:
-// CHECK-NEXT:    ret ptr @fmv._McrcMls64
+// CHECK-NEXT:    ret ptr @fmv._McrcMls64_v
 // CHECK:       resolver_else6:
 // CHECK-NEXT:    [[TMP16:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 17179869192
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 17179869192
+// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 8796093022216
+// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 8796093022216
 // CHECK-NEXT:    [[TMP19:%.*]] = and i1 true, [[TMP18]]
 // CHECK-NEXT:    br i1 [[TMP19]], label [[RESOLVER_RETURN7:%.*]], label [[RESOLVER_ELSE8:%.*]]
 // CHECK:       resolver_return7:
 // CHECK-NEXT:    ret ptr @fmv._Mfp16fmlMmemtag
 // CHECK:       resolver_else8:
 // CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 8448
-// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 8448
+// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 16640
+// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 16640
 // CHECK-NEXT:    [[TMP23:%.*]] = and i1 true, [[TMP22]]
 // CHECK-NEXT:    br i1 [[TMP23]], label [[RESOLVER_RETURN9:%.*]], label [[RESOLVER_ELSE10:%.*]]
 // CHECK:       resolver_return9:
 // CHECK-NEXT:    ret ptr @fmv._MaesMfp
 // CHECK:       resolver_else10:
 // CHECK-NEXT:    [[TMP24:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 2176
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 2176
+// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 4224
+// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 4224
 // CHECK-NEXT:    [[TMP27:%.*]] = and i1 true, [[TMP26]]
 // CHECK-NEXT:    br i1 [[TMP27]], label [[RESOLVER_RETURN11:%.*]], label [[RESOLVER_ELSE12:%.*]]
 // CHECK:       resolver_return11:
 // CHECK-NEXT:    ret ptr @fmv._MlseMsha2
 // CHECK:       resolver_else12:
 // CHECK-NEXT:    [[TMP28:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 17592186044416
-// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 17592186044416
+// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 144115188075855872
+// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 144115188075855872
 // CHECK-NEXT:    [[TMP31:%.*]] = and i1 true, [[TMP30]]
 // CHECK-NEXT:    br i1 [[TMP31]], label [[RESOLVER_RETURN13:%.*]], label [[RESOLVER_ELSE14:%.*]]
 // CHECK:       resolver_return13:
-// CHECK-NEXT:    ret ptr @fmv._Mmops
+// CHECK-NEXT:    ret ptr @fmv._Msme2
 // CHECK:       resolver_else14:
 // CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 4398046511104
-// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 4398046511104
+// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 1125899906842624
+// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 1125899906842624
 // CHECK-NEXT:    [[TMP35:%.*]] = and i1 true, [[TMP34]]
 // CHECK-NEXT:    br i1 [[TMP35]], label [[RESOLVER_RETURN15:%.*]], label [[RESOLVER_ELSE16:%.*]]
 // CHECK:       resolver_return15:
-// CHECK-NEXT:    ret ptr @fmv._Msme2
+// CHECK-NEXT:    ret ptr @fmv._Mbti
 // CHECK:       resolver_else16:
 // CHECK-NEXT:    ret ptr @fmv.default
 //
@@ -363,16 +363,16 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877907456
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877907456
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2251799813685760
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2251799813685760
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_one._Mls64Msimd
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 32768
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 32768
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 262144
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 262144
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -385,20 +385,20 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16896
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16896
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 66048
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 66048
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_two._Mfp16Msimd
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 33554432
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 33554432
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @fmv_two._Mfrintts
+// CHECK-NEXT:    ret ptr @fmv_two._Mdgh
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
 // CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 512
@@ -421,49 +421,49 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_e.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 20
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_d._Msb
-// CHECK-SAME: () #[[ATTR14:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR13:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_d.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 111
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_c._Mssbs
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_c.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret void
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@goo
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @fmv_inline()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @fmv_e()
@@ -477,96 +477,96 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8590213184
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8590213184
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4398048673856
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4398048673856
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @fmv_inline._MfcmaMfp16MrdmMsme
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 26405458935808
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 26405458935808
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 864726312827224064
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 864726312827224064
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
-// CHECK-NEXT:    ret ptr @fmv_inline._MmemtagMmopsMrcpc3
+// CHECK-NEXT:    ret ptr @fmv_inline._Mmemtag3MmopsMrcpc3
 // CHECK:       resolver_else2:
 // CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 1879048192
-// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 1879048192
+// CHECK-NEXT:    [[TMP9:%.*]] = and i64 [[TMP8]], 893353197568
+// CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[TMP9]], 893353197568
 // CHECK-NEXT:    [[TMP11:%.*]] = and i1 true, [[TMP10]]
 // CHECK-NEXT:    br i1 [[TMP11]], label [[RESOLVER_RETURN3:%.*]], label [[RESOLVER_ELSE4:%.*]]
 // CHECK:       resolver_return3:
-// CHECK-NEXT:    ret ptr @fmv_inline._Msve2Msve2-aesMsve2-bitperm
+// CHECK-NEXT:    ret ptr @fmv_inline._Msve2Msve2-bitpermMsve2-pmull128
 // CHECK:       resolver_else4:
 // CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 71307264
-// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 71307264
+// CHECK-NEXT:    [[TMP13:%.*]] = and i64 [[TMP12]], 34359773184
+// CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[TMP13]], 34359773184
 // CHECK-NEXT:    [[TMP15:%.*]] = and i1 true, [[TMP14]]
 // CHECK-NEXT:    br i1 [[TMP15]], label [[RESOLVER_RETURN5:%.*]], label [[RESOLVER_ELSE6:%.*]]
 // CHECK:       resolver_return5:
-// CHECK-NEXT:    ret ptr @fmv_inline._Mf32mmMi8mmMsha3
+// CHECK-NEXT:    ret ptr @fmv_inline._Mf64mmMpmullMsha1
 // CHECK:       resolver_else6:
 // CHECK-NEXT:    [[TMP16:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 4535485464576
-// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 4535485464576
+// CHECK-NEXT:    [[TMP17:%.*]] = and i64 [[TMP16]], 17246986240
+// CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 17246986240
 // CHECK-NEXT:    [[TMP19:%.*]] = and i1 true, [[TMP18]]
 // CHECK-NEXT:    br i1 [[TMP19]], label [[RESOLVER_RETURN7:%.*]], label [[RESOLVER_ELSE8:%.*]]
 // CHECK:       resolver_return7:
-// CHECK-NEXT:    ret ptr @fmv_inline._Msme2Mssbs
+// CHECK-NEXT:    ret ptr @fmv_inline._Mf32mmMi8mmMsha3
 // CHECK:       resolver_else8:
 // CHECK-NEXT:    [[TMP20:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 21474836480
-// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 21474836480
+// CHECK-NEXT:    [[TMP21:%.*]] = and i64 [[TMP20]], 19791209299968
+// CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[TMP21]], 19791209299968
 // CHECK-NEXT:    [[TMP23:%.*]] = and i1 true, [[TMP22]]
 // CHECK-NEXT:    br i1 [[TMP23]], label [[RESOLVER_RETURN9:%.*]], label [[RESOLVER_ELSE10:%.*]]
 // CHECK:       resolver_return9:
-// CHECK-NEXT:    ret ptr @fmv_inline._MmemtagMsve2-sm4
+// CHECK-NEXT:    ret ptr @fmv_inline._Mmemtag2Msve2-sm4
 // CHECK:       resolver_else10:
 // CHECK-NEXT:    [[TMP24:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 8623489024
-// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 8623489024
+// CHECK-NEXT:    [[TMP25:%.*]] = and i64 [[TMP24]], 1236950581248
+// CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[TMP25]], 1236950581248
 // CHECK-NEXT:    [[TMP27:%.*]] = and i1 true, [[TMP26]]
 // CHECK-NEXT:    br i1 [[TMP27]], label [[RESOLVER_RETURN11:%.*]], label [[RESOLVER_ELSE12:%.*]]
 // CHECK:       resolver_return11:
-// CHECK-NEXT:    ret ptr @fmv_inline._MsmeMsve
+// CHECK-NEXT:    ret ptr @fmv_inline._Msve2-aesMsve2-sha3
 // CHECK:       resolver_else12:
 // CHECK-NEXT:    [[TMP28:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 2684354560
-// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 2684354560
+// CHECK-NEXT:    [[TMP29:%.*]] = and i64 [[TMP28]], 4295098368
+// CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[TMP29]], 4295098368
 // CHECK-NEXT:    [[TMP31:%.*]] = and i1 true, [[TMP30]]
 // CHECK-NEXT:    br i1 [[TMP31]], label [[RESOLVER_RETURN13:%.*]], label [[RESOLVER_ELSE14:%.*]]
 // CHECK:       resolver_return13:
-// CHECK-NEXT:    ret ptr @fmv_inline._Msve2-aesMsve2-sha3
+// CHECK-NEXT:    ret ptr @fmv_inline._MditMsve-ebf16
 // CHECK:       resolver_else14:
 // CHECK-NEXT:    [[TMP32:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 281475110928384
-// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 281475110928384
+// CHECK-NEXT:    [[TMP33:%.*]] = and i64 [[TMP32]], 3221225472
+// CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[TMP33]], 3221225472
 // CHECK-NEXT:    [[TMP35:%.*]] = and i1 true, [[TMP34]]
 // CHECK-NEXT:    br i1 [[TMP35]], label [[RESOLVER_RETURN15:%.*]], label [[RESOLVER_ELSE16:%.*]]
 // CHECK:       resolver_return15:
-// CHECK-NEXT:    ret ptr @fmv_inline._McryptoMf64mm
+// CHECK-NEXT:    ret ptr @fmv_inline._MsveMsve-bf16
 // CHECK:       resolver_else16:
 // CHECK-NEXT:    [[TMP36:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP37:%.*]] = and i64 [[TMP36]], 2621440
-// CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i64 [[TMP37]], 2621440
+// CHECK-NEXT:    [[TMP37:%.*]] = and i64 [[TMP36]], 20971520
+// CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i64 [[TMP37]], 20971520
 // CHECK-NEXT:    [[TMP39:%.*]] = and i1 true, [[TMP38]]
 // CHECK-NEXT:    br i1 [[TMP39]], label [[RESOLVER_RETURN17:%.*]], label [[RESOLVER_ELSE18:%.*]]
 // CHECK:       resolver_return17:
 // CHECK-NEXT:    ret ptr @fmv_inline._MfrinttsMrcpc
 // CHECK:       resolver_else18:
 // CHECK-NEXT:    [[TMP40:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP41:%.*]] = and i64 [[TMP40]], 1081344
-// CHECK-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[TMP41]], 1081344
+// CHECK-NEXT:    [[TMP41:%.*]] = and i64 [[TMP40]], 8650752
+// CHECK-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[TMP41]], 8650752
 // CHECK-NEXT:    [[TMP43:%.*]] = and i1 true, [[TMP42]]
 // CHECK-NEXT:    br i1 [[TMP43]], label [[RESOLVER_RETURN19:%.*]], label [[RESOLVER_ELSE20:%.*]]
 // CHECK:       resolver_return19:
 // CHECK-NEXT:    ret ptr @fmv_inline._MdpbMrcpc2
 // CHECK:       resolver_else20:
 // CHECK-NEXT:    [[TMP44:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP45:%.*]] = and i64 [[TMP44]], 196608
-// CHECK-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[TMP45]], 196608
+// CHECK-NEXT:    [[TMP45:%.*]] = and i64 [[TMP44]], 1572864
+// CHECK-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[TMP45]], 1572864
 // CHECK-NEXT:    [[TMP47:%.*]] = and i1 true, [[TMP46]]
 // CHECK-NEXT:    br i1 [[TMP47]], label [[RESOLVER_RETURN21:%.*]], label [[RESOLVER_ELSE22:%.*]]
 // CHECK:       resolver_return21:
@@ -581,8 +581,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:    ret ptr @fmv_inline._Mfp16fmlMsimd
 // CHECK:       resolver_else24:
 // CHECK-NEXT:    [[TMP52:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP53:%.*]] = and i64 [[TMP52]], 8208
-// CHECK-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[TMP53]], 8208
+// CHECK-NEXT:    [[TMP53:%.*]] = and i64 [[TMP52]], 16400
+// CHECK-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[TMP53]], 16400
 // CHECK-NEXT:    [[TMP55:%.*]] = and i1 true, [[TMP54]]
 // CHECK-NEXT:    br i1 [[TMP55]], label [[RESOLVER_RETURN25:%.*]], label [[RESOLVER_ELSE26:%.*]]
 // CHECK:       resolver_return25:
@@ -611,8 +611,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877906944
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877906944
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 2251799813685248
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 2251799813685248
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -625,8 +625,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 34359738368
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 34359738368
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 70368744177664
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 70368744177664
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -639,8 +639,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 137438953472
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 137438953472
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 281474976710656
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 281474976710656
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -651,7 +651,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@recur
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    call void @reca()
 // CHECK-NEXT:    ret void
@@ -659,7 +659,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@main
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 0, ptr [[RETVAL]], align 4
@@ -670,7 +670,7 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@hoo
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[FP1:%.*]] = alloca ptr, align 8
 // CHECK-NEXT:    [[FP2:%.*]] = alloca ptr, align 8
@@ -687,14 +687,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_forward_default_decl._Mmops
-// CHECK-SAME: () #[[ATTR7]] {
+// CHECK-SAME: () #[[ATTR14:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_extern_forward_default_decl._Mdotprod
-// CHECK-SAME: () #[[ATTR3]] {
+// CHECK-SAME: () #[[ATTR15:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
@@ -708,14 +708,14 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_default_def._Msve
-// CHECK-SAME: () #[[ATTR15:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR16:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_default_def.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
@@ -729,56 +729,56 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_default_def.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_forward_default_def.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_with_implicit_forward_default_def._Mlse
-// CHECK-SAME: () #[[ATTR16:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR17:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@unused_without_default._Mrdm
-// CHECK-SAME: () #[[ATTR17:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR18:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@default_def_with_version_decls.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 0
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@used_def_without_default_decl._Mjscvt
-// CHECK-SAME: () #[[ATTR20:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR21:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@used_def_without_default_decl._Mrdm
-// CHECK-SAME: () #[[ATTR17]] {
+// CHECK-SAME: () #[[ATTR18]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@caller
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CALL:%.*]] = call i32 @used_def_without_default_decl()
 // CHECK-NEXT:    [[CALL1:%.*]] = call i32 @used_decl_without_default_decl()
@@ -790,8 +790,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -812,8 +812,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -831,92 +831,92 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._McryptoMf64mm
-// CHECK-SAME: () #[[ATTR21:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mf64mmMpmullMsha1
+// CHECK-SAME: () #[[ATTR22:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 1
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfcmaMfp16MrdmMsme
-// CHECK-SAME: () #[[ATTR22:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR23:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mf32mmMi8mmMsha3
-// CHECK-SAME: () #[[ATTR23:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR24:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 12
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msme2Mssbs
-// CHECK-SAME: () #[[ATTR8]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MditMsve-ebf16
+// CHECK-SAME: () #[[ATTR25:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 8
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MdpbMrcpc2
-// CHECK-SAME: () #[[ATTR24:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR26:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 6
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mdpb2Mjscvt
-// CHECK-SAME: () #[[ATTR25:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR27:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 7
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfrinttsMrcpc
-// CHECK-SAME: () #[[ATTR26:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR28:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MsmeMsve
-// CHECK-SAME: () #[[ATTR27:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MsveMsve-bf16
+// CHECK-SAME: () #[[ATTR29:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 4
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2-aesMsve2-sha3
-// CHECK-SAME: () #[[ATTR28:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR30:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 5
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2Msve2-aesMsve2-bitperm
-// CHECK-SAME: () #[[ATTR29:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Msve2Msve2-bitpermMsve2-pmull128
+// CHECK-SAME: () #[[ATTR31:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 9
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MmemtagMsve2-sm4
-// CHECK-SAME: () #[[ATTR30:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mmemtag2Msve2-sm4
+// CHECK-SAME: () #[[ATTR32:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 10
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
-// CHECK-LABEL: define {{[^@]+}}@fmv_inline._MmemtagMmopsMrcpc3
-// CHECK-SAME: () #[[ATTR31:[0-9]+]] {
+// CHECK-LABEL: define {{[^@]+}}@fmv_inline._Mmemtag3MmopsMrcpc3
+// CHECK-SAME: () #[[ATTR33:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 11
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MaesMdotprod
-// CHECK-SAME: () #[[ATTR3]] {
+// CHECK-SAME: () #[[ATTR15]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 13
 //
@@ -930,21 +930,21 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MfpMsm4
-// CHECK-SAME: () #[[ATTR32:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR34:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 15
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline._MlseMrdm
-// CHECK-SAME: () #[[ATTR33:[0-9]+]] {
+// CHECK-SAME: () #[[ATTR35:[0-9]+]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 16
 //
 //
 // CHECK: Function Attrs: noinline nounwind optnone
 // CHECK-LABEL: define {{[^@]+}}@fmv_inline.default
-// CHECK-SAME: () #[[ATTR13]] {
+// CHECK-SAME: () #[[ATTR11]] {
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -953,8 +953,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 33554432
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 33554432
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1073741824
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1073741824
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -967,8 +967,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16384
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16384
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 65536
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 65536
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -995,8 +995,8 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 131072
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 131072
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1048576
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1048576
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -1132,37 +1132,39 @@ int caller(void) { return used_def_without_default_decl() + used_decl_without_de
 // CHECK: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+flagm,+fp16fml,+fullfp16,+neon,+rand,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR1]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+altnzcv,+bf16,+flagm,+sme,+sme-i16i64,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR2]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+sha2,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR3]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+ls64,+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR4]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp16fml,+fullfp16,+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR5]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR6]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+crc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR7]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bti,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR8]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme,+sme2,-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR9:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR10]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR11]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR11]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
 // CHECK: attributes #[[ATTR12]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR13]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR14]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+sb,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR17]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR18:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR20]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fullfp16,+neon,+sha2,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fullfp16,+neon,+rdm,+sme,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+jsconv,+neon,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR26]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,+rcpc,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fullfp16,+neon,+sme,+sve,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR28]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-sm4,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,+rcpc,+rcpc3,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+sm4,-fp-armv8,-v9.5a" }
-// CHECK: attributes #[[ATTR33]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR13]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+sb,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR14]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR15]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+dotprod,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR16]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR17]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR18]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR19:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR20:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+rdm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR21]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR22]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+aes,+f64mm,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR23]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+complxnum,+fullfp16,+neon,+rdm,+sme,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR24]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+f32mm,+fullfp16,+i8mm,+neon,+sha2,+sha3,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR25]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+dit,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR26]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccpp,+rcpc,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR27]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ccdp,+ccpp,+jsconv,+neon,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR28]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint,+rcpc,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR29]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+fullfp16,+neon,+sve,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR30]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-sha3,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR31]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+neon,+sve,+sve2,+sve2-aes,+sve2-bitperm,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR32]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fullfp16,+mte,+neon,+sve,+sve2,+sve2-sm4,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR33]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+mops,+mte,+rcpc,+rcpc3,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR34]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+neon,+sm4,-fp-armv8,-v9.5a" }
+// CHECK: attributes #[[ATTR35]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+lse,+neon,+rdm,-fp-armv8,-v9.5a" }
 //.
 // CHECK-NOFMV: attributes #[[ATTR0]] = { noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
 // CHECK-NOFMV: attributes #[[ATTR1:[0-9]+]] = { "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="-fmv" }
diff --git a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
index d439f96982eb4..7953f902bf09b 100644
--- a/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
+++ b/clang/test/CodeGenCXX/attr-target-clones-aarch64.cpp
@@ -1,8 +1,8 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals --include-generated-funcs
 // RUN: %clang_cc1 -std=c++11 -triple aarch64-linux-gnu -emit-llvm %s -o - | FileCheck %s
 
-int __attribute__((target_clones("ls64+fp16", "default"))) foo_ovl(int) { return 1; }
-int __attribute__((target_clones("sme+ls64"))) foo_ovl(void) { return 2; }
+int __attribute__((target_clones("ls64_v+fp16", "default"))) foo_ovl(int) { return 1; }
+int __attribute__((target_clones("ls64_accdata+ls64"))) foo_ovl(void) { return 2; }
 
 int bar() {
   return foo_ovl(1) + foo_ovl();
@@ -35,6 +35,9 @@ void run_foo_tml() {
   Mc4.foo_tml();
 }
 
+
+
+
 //.
 // CHECK: @__aarch64_cpu_features = external dso_local global { i64 }
 // CHECK: @_Z7foo_ovli.ifunc = weak_odr alias i32 (i32), ptr @_Z7foo_ovli
@@ -46,7 +49,7 @@ void run_foo_tml() {
 // CHECK: @_ZN7MyClassIssE7foo_tmlEv = weak_odr ifunc i32 (ptr), ptr @_ZN7MyClassIssE7foo_tmlEv.resolver
 // CHECK: @_ZN7MyClassIisE7foo_tmlEv = weak_odr ifunc i32 (ptr), ptr @_ZN7MyClassIisE7foo_tmlEv.resolver
 //.
-// CHECK-LABEL: @_Z7foo_ovli._Mfp16Mls64(
+// CHECK-LABEL: @_Z7foo_ovli._Mfp16Mls64_v(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
 // CHECK-NEXT:    store i32 [[TMP0:%.*]], ptr [[DOTADDR]], align 4
@@ -57,17 +60,17 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 274877923328
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 274877923328
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 4503599627436032
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 4503599627436032
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z7foo_ovli._Mfp16Mls64
+// CHECK-NEXT:    ret ptr @_Z7foo_ovli._Mfp16Mls64_v
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z7foo_ovli.default
 //
 //
-// CHECK-LABEL: @_Z7foo_ovlv._Mls64Msme(
+// CHECK-LABEL: @_Z7foo_ovlv._Mls64Mls64_accdata(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 2
 //
@@ -76,12 +79,12 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 283467841536
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 283467841536
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 11258999068426240
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 11258999068426240
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z7foo_ovlv._Mls64Msme
+// CHECK-NEXT:    ret ptr @_Z7foo_ovlv._Mls64Mls64_accdata
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z7foo_ovlv.default
 //
@@ -111,16 +114,16 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1236950581248
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1236950581248
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36310271995674624
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36310271995674624
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @_ZN7MyClassIssE7foo_tmlEv._Msme-f64f64Mssbs
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16777216
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16777216
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -133,16 +136,16 @@ void run_foo_tml() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1236950581248
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1236950581248
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36310271995674624
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36310271995674624
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
 // CHECK-NEXT:    ret ptr @_ZN7MyClassIisE7foo_tmlEv._Msme-f64f64Mssbs
 // CHECK:       resolver_else:
 // CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 2097152
-// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 2097152
+// CHECK-NEXT:    [[TMP5:%.*]] = and i64 [[TMP4]], 16777216
+// CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 16777216
 // CHECK-NEXT:    [[TMP7:%.*]] = and i1 true, [[TMP6]]
 // CHECK-NEXT:    br i1 [[TMP7]], label [[RESOLVER_RETURN1:%.*]], label [[RESOLVER_ELSE2:%.*]]
 // CHECK:       resolver_return1:
@@ -228,7 +231,7 @@ void run_foo_tml() {
 //
 //.
 // CHECK: attributes #[[ATTR0:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fp-armv8,+fullfp16,+neon" }
-// CHECK: attributes #[[ATTR1:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme" }
+// CHECK: attributes #[[ATTR1:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+ls64" }
 // CHECK: attributes #[[ATTR2:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" }
 // CHECK: attributes #[[ATTR3:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+fptoint" }
 // CHECK: attributes #[[ATTR4:[0-9]+]] = { mustprogress noinline nounwind optnone "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+bf16,+sme,+sme-f64f64" }
diff --git a/clang/test/CodeGenCXX/attr-target-version.cpp b/clang/test/CodeGenCXX/attr-target-version.cpp
index 24433b8f20b7d..8b7273fe3bb51 100644
--- a/clang/test/CodeGenCXX/attr-target-version.cpp
+++ b/clang/test/CodeGenCXX/attr-target-version.cpp
@@ -3,7 +3,7 @@
 
 int __attribute__((target_version("sme-f64f64+bf16"))) foo(int) { return 1; }
 int __attribute__((target_version("default"))) foo(int) { return 2; }
-int __attribute__((target_version("sm4+bf16"))) foo(void) { return 3; }
+int __attribute__((target_version("sm4+ebf16"))) foo(void) { return 3; }
 int __attribute__((target_version("default"))) foo(void) { return 4; }
 
 struct MyClass {
@@ -89,7 +89,7 @@ int bar() {
 // CHECK-NEXT:    ret i32 2
 //
 //
-// CHECK-LABEL: @_Z3foov._Mbf16Msm4(
+// CHECK-LABEL: @_Z3foov._Mebf16Msm4(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    ret i32 3
 //
@@ -246,8 +246,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1099520016384
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1099520016384
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 36028797153181696
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 36028797153181696
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -260,12 +260,12 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 8388640
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 8388640
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 268435488
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 268435488
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
-// CHECK-NEXT:    ret ptr @_Z3foov._Mbf16Msm4
+// CHECK-NEXT:    ret ptr @_Z3foov._Mebf16Msm4
 // CHECK:       resolver_else:
 // CHECK-NEXT:    ret ptr @_Z3foov.default
 //
@@ -274,8 +274,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 33554432
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 33554432
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 1073741824
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 1073741824
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
@@ -288,8 +288,8 @@ int bar() {
 // CHECK-NEXT:  resolver_entry:
 // CHECK-NEXT:    call void @__init_cpu_features_resolver()
 // CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @__aarch64_cpu_features, align 8
-// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 16384
-// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 16384
+// CHECK-NEXT:    [[TMP1:%.*]] = and i64 [[TMP0]], 65536
+// CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP1]], 65536
 // CHECK-NEXT:    [[TMP3:%.*]] = and i1 true, [[TMP2]]
 // CHECK-NEXT:    br i1 [[TMP3]], label [[RESOLVER_RETURN:%.*]], label [[RESOLVER_ELSE:%.*]]
 // CHECK:       resolver_return:
diff --git a/clang/test/Sema/aarch64-cpu-supports.c b/clang/test/Sema/aarch64-cpu-supports.c
index cf8e043dfb1e3..ddeed7c5bc9e9 100644
--- a/clang/test/Sema/aarch64-cpu-supports.c
+++ b/clang/test/Sema/aarch64-cpu-supports.c
@@ -20,7 +20,7 @@ int test_aarch64_features(void) {
   // expected-warning@+1 {{invalid cpu feature string}}
   if (__builtin_cpu_supports("default"))
     return 6;
-  if (__builtin_cpu_supports(" ssbs + crc "))
+  if (__builtin_cpu_supports(" ssbs + bti "))
     return 7;
   return 0;
 }
diff --git a/clang/test/Sema/attr-target-clones-aarch64.c b/clang/test/Sema/attr-target-clones-aarch64.c
index 511e8a9411804..bc3fceab82825 100644
--- a/clang/test/Sema/attr-target-clones-aarch64.c
+++ b/clang/test/Sema/attr-target-clones-aarch64.c
@@ -9,7 +9,7 @@ void __attribute__((target_clones("ssbs+ls64"))) warn2(void);
 
 // expected-error@+2 {{'target_clones' and 'target_version' attributes are not compatible}}
 // expected-note@+1 {{conflicting attribute is here}}
-void __attribute__((target_version("bf16"), target_clones("sme+memtag"))) not_compat(void);
+void __attribute__((target_version("sve-bf16"), target_clones("sme+memtag"))) not_compat(void);
 
 int redecl(void);
 int __attribute__((target_clones("frintts", "simd+fp", "default"))) redecl(void) { return 1; }
@@ -21,25 +21,26 @@ int __attribute__((target_clones("sve+dotprod"))) redecl3(void);
 int redecl3(void);
 
 int __attribute__((target_clones("rng", "fp16fml+fp", "default"))) redecl4(void);
-// expected-error@+2 {{'target_clones' attribute does not match previous declaration}}
+// expected-error@+3 {{'target_clones' attribute does not match previous declaration}}
 // expected-note@-2 {{previous declaration is here}}
-int __attribute__((target_clones("dpb2+memtag+rpres+ls64", "bf16+dpb+sha2", "default"))) redecl4(void) { return 1; }
+// expected-warning@+1 {{version list contains entries that don't impact code generation}}
+int __attribute__((target_clones("dgh+memtag+rpres+ls64_v", "ebf16+dpb+sha1", "default"))) redecl4(void) { return 1; }
 
 int __attribute__((target_version("flagm2"))) redef2(void) { return 1; }
 // expected-error@+2 {{multiversioned function redeclarations require identical target attributes}}
 // expected-note@-2 {{previous declaration is here}}
 int __attribute__((target_clones("flagm2", "default"))) redef2(void) { return 1; }
 
-int __attribute__((target_clones("f32mm", "f64mm", "sha3+fp"))) redef3(void) { return 1; }
+int __attribute__((target_clones("f32mm", "f64mm", "sha1+fp"))) redef3(void) { return 1; }
 // expected-error@+2 {{'target_clones' attribute does not match previous declaration}}
 // expected-note@-2 {{previous declaration is here}}
-int __attribute__((target_clones("f32mm", "sha3+fp", "f64mm"))) redef3(void) { return 1; }
+int __attribute__((target_clones("f32mm", "sha1+fp", "f64mm"))) redef3(void) { return 1; }
 
 int __attribute__((target_clones("rdm+lse+rdm", "lse+rdm"))) dup1(void) { return 1; }
 // expected-warning@+1 {{version list contains duplicate entries}}
 int __attribute__((target_clones("rdm+lse+rdm", "rdm+lse+rdm"))) dup2(void) { return 2; }
 // expected-warning@+1 {{version list contains duplicate entries}}
-int __attribute__((target_clones("rcpc2+sve2-aes", "rcpc2+sve2-aes"))) dup3(void) { return 3; }
+int __attribute__((target_clones("rcpc2+sve2-pmull128", "rcpc2+sve2-pmull128"))) dup3(void) { return 3; }
 // expected-warning@+1 {{version list contains duplicate entries}}
 void __attribute__((target_clones("sha3", "default", "default"))) dup4(void);
 // expected-warning@+2 {{version list contains duplicate entries}}
@@ -48,7 +49,7 @@ int __attribute__((target_clones("fp", "fp", "crc+dotprod", "dotprod+crc"))) dup
 
 // expected-warning@+1 {{version list contains duplicate entries}}
 int __attribute__((target_clones("fp16+memtag", "memtag+fp16"))) dup6(void) { return 6; }
-int __attribute__((target_clones("simd+ssbs", "simd+dpb2"))) dup7(void) { return 7; }
+int __attribute__((target_clones("simd+ssbs2", "simd+dpb2"))) dup7(void) { return 7; }
 
 // expected-warning@+1 {{unsupported '' in the 'target_clones' attribute string;}}
 void __attribute__((target_clones(""))) empty_target_1(void);
@@ -71,13 +72,13 @@ empty_target_5(void);
 void __attribute__((target_clones("sve2-bitperm", "sve2-bitperm")))
 dupe_normal(void);
 
-void __attribute__((target_clones("default"), target_clones("memtag+mops"))) dupe_normal2(void);
+void __attribute__((target_clones("default"), target_clones("memtag3+bti"))) dupe_normal2(void);
 
 int mv_after_use(void);
 int useage(void) {
   return mv_after_use();
 }
 // expected-error@+1 {{function declaration cannot become a multiversioned function after first usage}}
-int __attribute__((target_clones("sve2-sha3+ssbs", "sm4"))) mv_after_use(void) { return 1; }
+int __attribute__((target_clones("sve2-sha3+ssbs2", "sm4"))) mv_after_use(void) { return 1; }
 // expected-error@+1 {{'main' cannot be a multiversioned function}}
-int __attribute__((target_clones("sve2-aes"))) main() { return 1; }
+int __attribute__((target_clones("sve-i8mm"))) main() { return 1; }
diff --git a/clang/test/Sema/attr-target-version.c b/clang/test/Sema/attr-target-version.c
index 4ff8e1eee619e..cd5be459456eb 100644
--- a/clang/test/Sema/attr-target-version.c
+++ b/clang/test/Sema/attr-target-version.c
@@ -16,7 +16,7 @@ int __attribute__((target_version("aes"))) foo(void) { return 1; }
 int __attribute__((target_version("default"))) foo(void) { return 2; }
 
 //expected-note@+1 {{previous definition is here}}
-int __attribute__((target_version("sha3 + aes "))) foo(void) { return 1; }
+int __attribute__((target_version("sha3 + pmull "))) foo(void) { return 1; }
 //expected-note@-1 {{previous definition is here}}
 
 //expected-error@+1 {{redefinition of 'foo'}}
@@ -32,11 +32,11 @@ __attribute__ ((target("bf16,sve,sve2,dotprod"))) int func(void) { return 1; }
 __attribute__ ((target("default"))) int func(void) { return 0; }
 
 //expected-note@+1 {{previous declaration is here}}
-void __attribute__((target_version("mops+flagm2"))) one(void) {}
+void __attribute__((target_version("bti+flagm2"))) one(void) {}
 //expected-error@+1 {{multiversioned function redeclarations require identical target attributes}}
-void __attribute__((target_version("flagm2+mops"))) one(void) {}
+void __attribute__((target_version("flagm2+bti"))) one(void) {}
 
-void __attribute__((target_version("ssbs+sha2"))) two(void) {}
+void __attribute__((target_version("ssbs+sha1"))) two(void) {}
 void __attribute__((target_version("ssbs+fp16fml"))) two(void) {}
 
 //expected-error@+1 {{'main' cannot be a multiversioned function}}
@@ -44,7 +44,7 @@ int __attribute__((target_version("lse"))) main(void) { return 1; }
 
 // It is ok for the default version to appear first.
 int default_first(void) { return 1; }
-int __attribute__((target_version("crc"))) default_first(void) { return 2; }
+int __attribute__((target_version("dit"))) default_first(void) { return 2; }
 int __attribute__((target_version("mops"))) default_first(void) { return 3; }
 
 // It is ok if the default version is between other versions.
@@ -77,7 +77,7 @@ void __attribute__((target_version("rdm+rng+crc"))) redef(void) {}
 void __attribute__((target_version("rdm+rng+crc"))) redef(void) {}
 
 int def(void);
-void __attribute__((target_version("crc"))) nodef(void);
+void __attribute__((target_version("dit"))) nodef(void);
 void __attribute__((target_version("ls64"))) nodef(void);
 void __attribute__((target_version("aes"))) ovl(void);
 void __attribute__((target_version("default"))) ovl(void);
@@ -89,12 +89,12 @@ int bar() {
   return def();
 }
 // expected-error@+1 {{function declaration cannot become a multiversioned function after first usage}}
-int __attribute__((target_version("sha3"))) def(void) { return 1; }
+int __attribute__((target_version("sha1"))) def(void) { return 1; }
 
 int __attribute__((target_version("sve"))) prot();
 // expected-error@-1 {{multiversioned function must have a prototype}}
 
-int __attribute__((target_version("aes"))) rtype(int);
+int __attribute__((target_version("pmull"))) rtype(int);
 // expected-error@+1 {{multiversioned function declaration has a different return type}}
 float __attribute__((target_version("rdm"))) rtype(int);
 
@@ -102,7 +102,7 @@ int __attribute__((target_version("sha2"))) combine(void) { return 1; }
 // expected-error@+1 {{multiversioned function declaration has a different calling convention}}
 int __attribute__((aarch64_vector_pcs, target_version("sha3"))) combine(void) { return 2; }
 
-int __attribute__((target_version("fp+aes+sha3+rcpc"))) unspec_args() { return -1; }
+int __attribute__((target_version("fp+aes+pmull+rcpc"))) unspec_args() { return -1; }
 // expected-error@-1 {{multiversioned function must have a prototype}}
 // expected-error@+1 {{multiversioned function must have a prototype}}
 int __attribute__((target_version("default"))) unspec_args() { return 0; }
diff --git a/clang/test/SemaCXX/attr-target-version.cpp b/clang/test/SemaCXX/attr-target-version.cpp
index ae2923cc303c2..b3385f043590f 100644
--- a/clang/test/SemaCXX/attr-target-version.cpp
+++ b/clang/test/SemaCXX/attr-target-version.cpp
@@ -25,13 +25,13 @@ int __attribute__((target_version("dpb"))) diff_link(void);
 
 int __attribute__((target_version("memtag"))) diff_link1(void) { return 1; }
 //expected-error@+1 {{multiversioned function declaration has a different linkage}}
-static int __attribute__((target_version("mops"))) diff_link1(void);
+static int __attribute__((target_version("bti"))) diff_link1(void);
 
 int __attribute__((target_version("flagm2"))) diff_link2(void) { return 1; }
 extern int __attribute__((target_version("flagm"))) diff_link2(void);
 
 namespace {
-static int __attribute__((target_version("memtag"))) diff_link2(void) { return 2; }
+static int __attribute__((target_version("memtag3"))) diff_link2(void) { return 2; }
 int __attribute__((target_version("sve2-bitperm"))) diff_link2(void) { return 1; }
 } // namespace
 
@@ -49,7 +49,7 @@ double __attribute__((target_version("rcpc"))) diff_type1(void);
 
 auto __attribute__((target_version("rcpc2"))) diff_type2(void) -> int { return 1; }
 //expected-error@+1 {{multiversioned function declaration has a different return type}}
-auto __attribute__((target_version("sve2-aes"))) diff_type2(void) -> long { return (long)1; }
+auto __attribute__((target_version("sve-bf16"))) diff_type2(void) -> long { return (long)1; }
 
 int __attribute__((target_version("fp16fml"))) diff_type3(void) noexcept(false) { return 1; }
 //expected-error@+2 {{exception specification in declaration does not match previous declaration}}
@@ -75,7 +75,7 @@ auto __attribute__((target_version("dpb2"))) ret3(void) -> int { return 1; }
 class Cls {
   __attribute__((target_version("rng"))) Cls();
   // expected-error@-1 {{attribute 'target_version' multiversioned functions do not yet support constructors}}
-  __attribute__((target_version("sve2-aes"))) ~Cls();
+  __attribute__((target_version("sve-i8mm"))) ~Cls();
   // expected-error@-1 {{attribute 'target_version' multiversioned functions do not yet support destructors}}
 
   Cls &__attribute__((target_version("f32mm"))) operator=(const Cls &) = default;
@@ -98,11 +98,11 @@ __attribute__((target_version("jscvt"))) void Decl();
 } // namespace Nms
 
 class Out {
-  int __attribute__((target_version("mops"))) func(void);
-  int __attribute__((target_version("ssbs"))) func(void);
+  int __attribute__((target_version("bti"))) func(void);
+  int __attribute__((target_version("ssbs2"))) func(void);
 };
-int __attribute__((target_version("mops"))) Out::func(void) { return 1; }
-int __attribute__((target_version("ssbs"))) Out::func(void) { return 2; }
+int __attribute__((target_version("bti"))) Out::func(void) { return 1; }
+int __attribute__((target_version("ssbs2"))) Out::func(void) { return 2; }
 // expected-error@+3 {{out-of-line definition of 'func' does not match any declaration in 'Out'}}
 // expected-note@-3 {{member declaration nearly matches}}
 // expected-note@-3 {{member declaration nearly matches}}
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64.c b/compiler-rt/lib/builtins/cpu_model/aarch64.c
index 1ac4d85a0c139..17bddfca46f09 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64.c
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64.c
@@ -67,10 +67,13 @@ enum CPUFeatures {
   FEAT_FP,
   FEAT_SIMD,
   FEAT_CRC,
+  FEAT_SHA1,
   FEAT_SHA2,
   FEAT_SHA3,
   FEAT_AES,
+  FEAT_PMULL,
   FEAT_FP16,
+  FEAT_DIT,
   FEAT_DPB,
   FEAT_DPB2,
   FEAT_JSCVT,
@@ -78,23 +81,35 @@ enum CPUFeatures {
   FEAT_RCPC,
   FEAT_RCPC2,
   FEAT_FRINTTS,
+  FEAT_DGH,
   FEAT_I8MM,
   FEAT_BF16,
+  FEAT_EBF16,
   FEAT_RPRES,
   FEAT_SVE,
+  FEAT_SVE_BF16,
+  FEAT_SVE_EBF16,
+  FEAT_SVE_I8MM,
   FEAT_SVE_F32MM,
   FEAT_SVE_F64MM,
   FEAT_SVE2,
   FEAT_SVE_AES,
+  FEAT_SVE_PMULL128,
   FEAT_SVE_BITPERM,
   FEAT_SVE_SHA3,
   FEAT_SVE_SM4,
   FEAT_SME,
   FEAT_MEMTAG,
+  FEAT_MEMTAG2,
+  FEAT_MEMTAG3,
   FEAT_SB,
   FEAT_PREDRES,
   FEAT_SSBS,
+  FEAT_SSBS2,
+  FEAT_BTI,
   FEAT_LS64,
+  FEAT_LS64_V,
+  FEAT_LS64_ACCDATA,
   FEAT_WFXT,
   FEAT_SME_F64,
   FEAT_SME_I64,
@@ -102,7 +117,7 @@ enum CPUFeatures {
   FEAT_RCPC3,
   FEAT_MOPS,
   FEAT_MAX,
-  FEAT_EXT = 47, // Reserved to indicate presence of additional features field
+  FEAT_EXT = 62, // Reserved to indicate presence of additional features field
                  // in __aarch64_cpu_features
   FEAT_INIT      // Used as flag of features initialization completion
 };
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
index 19114c4abdfca..6fef109567b61 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/apple.inc
@@ -35,10 +35,13 @@ void __init_cpu_features_resolver(void) {
       {"hw.optional.floatingpoint", FEAT_FP},
       {"hw.optional.AdvSIMD", FEAT_SIMD},
       {"hw.optional.armv8_crc32", FEAT_CRC},
+      {"hw.optional.arm.FEAT_SHA1", FEAT_SHA1},
       {"hw.optional.arm.FEAT_SHA256", FEAT_SHA2},
       {"hw.optional.arm.FEAT_SHA3", FEAT_SHA3},
       {"hw.optional.arm.FEAT_AES", FEAT_AES},
+      {"hw.optional.arm.FEAT_PMULL", FEAT_PMULL},
       {"hw.optional.arm.FEAT_FP16", FEAT_FP16},
+      {"hw.optional.arm.FEAT_DIT", FEAT_DIT},
       {"hw.optional.arm.FEAT_DPB", FEAT_DPB},
       {"hw.optional.arm.FEAT_DPB2", FEAT_DPB2},
       {"hw.optional.arm.FEAT_JSCVT", FEAT_JSCVT},
@@ -50,7 +53,8 @@ void __init_cpu_features_resolver(void) {
       {"hw.optional.arm.FEAT_BF16", FEAT_BF16},
       {"hw.optional.arm.FEAT_SB", FEAT_SB},
       {"hw.optional.arm.FEAT_SPECRES", FEAT_PREDRES},
-      {"hw.optional.arm.FEAT_SSBS", FEAT_SSBS},
+      {"hw.optional.arm.FEAT_SSBS", FEAT_SSBS2},
+      {"hw.optional.arm.FEAT_BTI", FEAT_BTI},
   };
 
   for (size_t I = 0, E = sizeof(feature_checks) / sizeof(feature_checks[0]);
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
index 579aa00dc5c72..d8e0280f40416 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/fuchsia.inc
@@ -22,6 +22,10 @@ void __init_cpu_features_resolver() {
     setCPUFeature(FEAT_SIMD);
   if (features & ZX_ARM64_FEATURE_ISA_AES)
     setCPUFeature(FEAT_AES);
+  if (features & ZX_ARM64_FEATURE_ISA_PMULL)
+    setCPUFeature(FEAT_PMULL);
+  if (features & ZX_ARM64_FEATURE_ISA_SHA1)
+    setCPUFeature(FEAT_SHA1);
   if (features & ZX_ARM64_FEATURE_ISA_SHA256)
     setCPUFeature(FEAT_SHA2);
   if (features & ZX_ARM64_FEATURE_ISA_CRC32)
diff --git a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
index 54797852e4aaa..32a21a2fba9a3 100644
--- a/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
+++ b/compiler-rt/lib/builtins/cpu_model/aarch64/fmv/mrs.inc
@@ -16,6 +16,8 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     hwcap2 = arg->_hwcap2;
   if (hwcap & HWCAP_CRC32)
     setCPUFeature(FEAT_CRC);
+  if (hwcap & HWCAP_PMULL)
+    setCPUFeature(FEAT_PMULL);
   if (hwcap & HWCAP_FLAGM)
     setCPUFeature(FEAT_FLAGM);
   if (hwcap2 & HWCAP2_FLAGM2) {
@@ -32,12 +34,16 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     setCPUFeature(FEAT_FP16);
     setCPUFeature(FEAT_FP);
   }
+  if (hwcap & HWCAP_DIT)
+    setCPUFeature(FEAT_DIT);
   if (hwcap & HWCAP_ASIMDRDM)
     setCPUFeature(FEAT_RDM);
   if (hwcap & HWCAP_ILRCPC)
     setCPUFeature(FEAT_RCPC2);
   if (hwcap & HWCAP_AES)
     setCPUFeature(FEAT_AES);
+  if (hwcap & HWCAP_SHA1)
+    setCPUFeature(FEAT_SHA1);
   if (hwcap & HWCAP_SHA2)
     setCPUFeature(FEAT_SHA2);
   if (hwcap & HWCAP_JSCVT)
@@ -47,11 +53,22 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
   if (hwcap & HWCAP_SB)
     setCPUFeature(FEAT_SB);
   if (hwcap & HWCAP_SSBS)
-    setCPUFeature(FEAT_SSBS);
-  if (hwcap2 & HWCAP2_MTE)
+    setCPUFeature(FEAT_SSBS2);
+  if (hwcap2 & HWCAP2_MTE) {
     setCPUFeature(FEAT_MEMTAG);
+    setCPUFeature(FEAT_MEMTAG2);
+  }
+  if (hwcap2 & HWCAP2_MTE3) {
+    setCPUFeature(FEAT_MEMTAG);
+    setCPUFeature(FEAT_MEMTAG2);
+    setCPUFeature(FEAT_MEMTAG3);
+  }
   if (hwcap2 & HWCAP2_SVEAES)
     setCPUFeature(FEAT_SVE_AES);
+  if (hwcap2 & HWCAP2_SVEPMULL) {
+    setCPUFeature(FEAT_SVE_AES);
+    setCPUFeature(FEAT_SVE_PMULL128);
+  }
   if (hwcap2 & HWCAP2_SVEBITPERM)
     setCPUFeature(FEAT_SVE_BITPERM);
   if (hwcap2 & HWCAP2_SVESHA3)
@@ -66,12 +83,22 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     setCPUFeature(FEAT_RNG);
   if (hwcap2 & HWCAP2_I8MM)
     setCPUFeature(FEAT_I8MM);
+  if (hwcap2 & HWCAP2_EBF16)
+    setCPUFeature(FEAT_EBF16);
+  if (hwcap2 & HWCAP2_SVE_EBF16)
+    setCPUFeature(FEAT_SVE_EBF16);
+  if (hwcap2 & HWCAP2_DGH)
+    setCPUFeature(FEAT_DGH);
   if (hwcap2 & HWCAP2_FRINT)
     setCPUFeature(FEAT_FRINTTS);
+  if (hwcap2 & HWCAP2_SVEI8MM)
+    setCPUFeature(FEAT_SVE_I8MM);
   if (hwcap2 & HWCAP2_SVEF32MM)
     setCPUFeature(FEAT_SVE_F32MM);
   if (hwcap2 & HWCAP2_SVEF64MM)
     setCPUFeature(FEAT_SVE_F64MM);
+  if (hwcap2 & HWCAP2_BTI)
+    setCPUFeature(FEAT_BTI);
   if (hwcap2 & HWCAP2_RPRES)
     setCPUFeature(FEAT_RPRES);
   if (hwcap2 & HWCAP2_WFXT)
@@ -114,6 +141,9 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
       // ID_AA64ZFR0_EL1.SVEver == 0b0001
       if (extractBits(ftr, 0, 4) == 0x1)
         setCPUFeature(FEAT_SVE2);
+      // ID_AA64ZFR0_EL1.BF16 != 0b0000
+      if (extractBits(ftr, 20, 4) != 0x0)
+        setCPUFeature(FEAT_SVE_BF16);
     }
     getCPUFeature(ID_AA64ISAR0_EL1, ftr);
     // ID_AA64ISAR0_EL1.SHA3 != 0b0000
@@ -138,6 +168,12 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
     // ID_AA64ISAR1_EL1.LS64 >= 0b0001
     if (extractBits(ftr, 60, 4) >= 0x1)
       setCPUFeature(FEAT_LS64);
+    // ID_AA64ISAR1_EL1.LS64 >= 0b0010
+    if (extractBits(ftr, 60, 4) >= 0x2)
+      setCPUFeature(FEAT_LS64_V);
+    // ID_AA64ISAR1_EL1.LS64 >= 0b0011
+    if (extractBits(ftr, 60, 4) >= 0x3)
+      setCPUFeature(FEAT_LS64_ACCDATA);
   } else {
     // Set some features in case of no CPUID support
     if (hwcap & (HWCAP_FP | HWCAP_FPHP)) {
@@ -151,6 +187,8 @@ static void __init_cpu_features_constructor(unsigned long hwcap,
       setCPUFeature(FEAT_RCPC);
     if (hwcap2 & HWCAP2_BF16 || hwcap2 & HWCAP2_EBF16)
       setCPUFeature(FEAT_BF16);
+    if (hwcap2 & HWCAP2_SVEBF16)
+      setCPUFeature(FEAT_SVE_BF16);
     if (hwcap2 & HWCAP2_SVE2 && hwcap & HWCAP_SVE)
       setCPUFeature(FEAT_SVE2);
     if (hwcap & HWCAP_SHA3)
diff --git a/llvm/include/llvm/TargetParser/AArch64TargetParser.h b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
index bcb36beb3cc0b..805b963a7a13c 100644
--- a/llvm/include/llvm/TargetParser/AArch64TargetParser.h
+++ b/llvm/include/llvm/TargetParser/AArch64TargetParser.h
@@ -46,10 +46,13 @@ enum CPUFeatures {
   FEAT_FP,
   FEAT_SIMD,
   FEAT_CRC,
+  FEAT_SHA1,
   FEAT_SHA2,
   FEAT_SHA3,
   FEAT_AES,
+  FEAT_PMULL,
   FEAT_FP16,
+  FEAT_DIT,
   FEAT_DPB,
   FEAT_DPB2,
   FEAT_JSCVT,
@@ -57,23 +60,35 @@ enum CPUFeatures {
   FEAT_RCPC,
   FEAT_RCPC2,
   FEAT_FRINTTS,
+  FEAT_DGH,
   FEAT_I8MM,
   FEAT_BF16,
+  FEAT_EBF16,
   FEAT_RPRES,
   FEAT_SVE,
+  FEAT_SVE_BF16,
+  FEAT_SVE_EBF16,
+  FEAT_SVE_I8MM,
   FEAT_SVE_F32MM,
   FEAT_SVE_F64MM,
   FEAT_SVE2,
   FEAT_SVE_AES,
+  FEAT_SVE_PMULL128,
   FEAT_SVE_BITPERM,
   FEAT_SVE_SHA3,
   FEAT_SVE_SM4,
   FEAT_SME,
   FEAT_MEMTAG,
+  FEAT_MEMTAG2,
+  FEAT_MEMTAG3,
   FEAT_SB,
   FEAT_PREDRES,
   FEAT_SSBS,
+  FEAT_SSBS2,
+  FEAT_BTI,
   FEAT_LS64,
+  FEAT_LS64_V,
+  FEAT_LS64_ACCDATA,
   FEAT_WFXT,
   FEAT_SME_F64,
   FEAT_SME_I64,
@@ -81,12 +96,12 @@ enum CPUFeatures {
   FEAT_RCPC3,
   FEAT_MOPS,
   FEAT_MAX,
-  FEAT_EXT = 47,
+  FEAT_EXT = 62,
   FEAT_INIT
 };
 
-static_assert(FEAT_MAX < 47,
-              "Number of features in CPUFeatures are limited to 47 entries");
+static_assert(FEAT_MAX < 62,
+              "Number of features in CPUFeatures are limited to 62 entries");
 
 // Arch extension modifiers for CPUs. These are labelled with their Arm ARM
 // feature name (though the canonical reference for those is AArch64.td)
@@ -200,13 +215,17 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"b16b16", AArch64::AEK_B16B16, "+b16b16", "-b16b16", FEAT_INIT, "", 0},
     {"bf16", AArch64::AEK_BF16, "+bf16", "-bf16", FEAT_BF16, "+bf16", 280},
     {"brbe", AArch64::AEK_BRBE, "+brbe", "-brbe", FEAT_INIT, "", 0},
+    {"bti", AArch64::AEK_NONE, {}, {}, FEAT_BTI, "+bti", 510},
     {"crc", AArch64::AEK_CRC, "+crc", "-crc", FEAT_CRC, "+crc", 110},
     {"crypto", AArch64::AEK_CRYPTO, "+crypto", "-crypto", FEAT_INIT, "+aes,+sha2", 0},
     {"cssc", AArch64::AEK_CSSC, "+cssc", "-cssc", FEAT_INIT, "", 0},
     {"d128", AArch64::AEK_D128, "+d128", "-d128", FEAT_INIT, "", 0},
+    {"dgh", AArch64::AEK_NONE, {}, {}, FEAT_DGH, "", 260},
+    {"dit", AArch64::AEK_NONE, {}, {}, FEAT_DIT, "+dit", 180},
     {"dotprod", AArch64::AEK_DOTPROD, "+dotprod", "-dotprod", FEAT_DOTPROD, "+dotprod,+fp-armv8,+neon", 104},
     {"dpb", AArch64::AEK_NONE, {}, {}, FEAT_DPB, "+ccpp", 190},
     {"dpb2", AArch64::AEK_NONE, {}, {}, FEAT_DPB2, "+ccpp,+ccdp", 200},
+    {"ebf16", AArch64::AEK_NONE, {}, {}, FEAT_EBF16, "+bf16", 290},
     {"f32mm", AArch64::AEK_F32MM, "+f32mm", "-f32mm", FEAT_SVE_F32MM, "+sve,+f32mm,+fullfp16,+fp-armv8,+neon", 350},
     {"f64mm", AArch64::AEK_F64MM, "+f64mm", "-f64mm", FEAT_SVE_F64MM, "+sve,+f64mm,+fullfp16,+fp-armv8,+neon", 360},
     {"fcma", AArch64::AEK_FCMA, "+complxnum", "-complxnum", FEAT_FCMA, "+fp-armv8,+neon,+complxnum", 220},
@@ -220,12 +239,17 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"i8mm", AArch64::AEK_I8MM, "+i8mm", "-i8mm", FEAT_I8MM, "+i8mm", 270},
     {"ite", AArch64::AEK_ITE, "+ite", "-ite", FEAT_INIT, "", 0},
     {"jscvt", AArch64::AEK_JSCVT, "+jsconv", "-jsconv", FEAT_JSCVT, "+fp-armv8,+neon,+jsconv", 210},
+    {"ls64_accdata", AArch64::AEK_NONE, {}, {}, FEAT_LS64_ACCDATA, "+ls64", 540},
+    {"ls64_v", AArch64::AEK_NONE, {}, {}, FEAT_LS64_V, "", 530},
     {"ls64", AArch64::AEK_LS64, "+ls64", "-ls64", FEAT_LS64, "", 520},
     {"lse", AArch64::AEK_LSE, "+lse", "-lse", FEAT_LSE, "+lse", 80},
     {"lse128", AArch64::AEK_LSE128, "+lse128", "-lse128", FEAT_INIT, "", 0},
     {"memtag", AArch64::AEK_MTE, "+mte", "-mte", FEAT_MEMTAG, "", 440},
+    {"memtag2", AArch64::AEK_NONE, {}, {}, FEAT_MEMTAG2, "+mte", 450},
+    {"memtag3", AArch64::AEK_NONE, {}, {}, FEAT_MEMTAG3, "+mte", 460},
     {"mops", AArch64::AEK_MOPS, "+mops", "-mops", FEAT_MOPS, "+mops", 650},
     {"pauth", AArch64::AEK_PAUTH, "+pauth", "-pauth", FEAT_INIT, "", 0},
+    {"pmull", AArch64::AEK_NONE, {}, {}, FEAT_PMULL, "+aes,+fp-armv8,+neon", 160},
     {"pmuv3", AArch64::AEK_PERFMON, "+perfmon", "-perfmon", FEAT_INIT, "", 0},
     {"predres", AArch64::AEK_PREDRES, "+predres", "-predres", FEAT_PREDRES, "+predres", 480},
     {"predres2", AArch64::AEK_SPECRES2, "+specres2", "-specres2", FEAT_INIT, "", 0},
@@ -239,6 +263,7 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"rng", AArch64::AEK_RAND, "+rand", "-rand", FEAT_RNG, "+rand", 10},
     {"rpres", AArch64::AEK_NONE, {}, {}, FEAT_RPRES, "", 300},
     {"sb", AArch64::AEK_SB, "+sb", "-sb", FEAT_SB, "+sb", 470},
+    {"sha1", AArch64::AEK_NONE, {}, {}, FEAT_SHA1, "+fp-armv8,+neon", 120},
     {"sha2", AArch64::AEK_SHA2, "+sha2", "-sha2", FEAT_SHA2, "+sha2,+fp-armv8,+neon", 130},
     {"sha3", AArch64::AEK_SHA3, "+sha3", "-sha3", FEAT_SHA3, "+sha3,+sha2,+fp-armv8,+neon", 140},
     {"simd", AArch64::AEK_SIMD, "+neon", "-neon", FEAT_SIMD, "+fp-armv8,+neon", 100},
@@ -250,9 +275,14 @@ inline constexpr ExtensionInfo Extensions[] = {
     {"sme2", AArch64::AEK_SME2, "+sme2", "-sme2", FEAT_SME2, "+sme2,+sme,+bf16", 580},
     {"sme2p1", AArch64::AEK_SME2p1, "+sme2p1", "-sme2p1", FEAT_INIT, "+sme2p1,+sme2,+sme,+bf16", 0},
     {"ssbs", AArch64::AEK_SSBS, "+ssbs", "-ssbs", FEAT_SSBS, "", 490},
+    {"ssbs2", AArch64::AEK_NONE, {}, {}, FEAT_SSBS2, "+ssbs", 500},
+    {"sve-bf16", AArch64::AEK_NONE, {}, {}, FEAT_SVE_BF16, "+sve,+bf16,+fullfp16,+fp-armv8,+neon", 320},
+    {"sve-ebf16", AArch64::AEK_NONE, {}, {}, FEAT_SVE_EBF16, "+sve,+bf16,+fullfp16,+fp-armv8,+neon", 330},
+    {"sve-i8mm", AArch64::AEK_NONE, {}, {}, FEAT_SVE_I8MM, "+sve,+i8mm,+fullfp16,+fp-armv8,+neon", 340},
     {"sve", AArch64::AEK_SVE, "+sve", "-sve", FEAT_SVE, "+sve,+fullfp16,+fp-armv8,+neon", 310},
     {"sve2-aes", AArch64::AEK_SVE2AES, "+sve2-aes", "-sve2-aes", FEAT_SVE_AES, "+sve2,+sve,+sve2-aes,+fullfp16,+fp-armv8,+neon", 380},
     {"sve2-bitperm", AArch64::AEK_SVE2BITPERM, "+sve2-bitperm", "-sve2-bitperm", FEAT_SVE_BITPERM, "+sve2,+sve,+sve2-bitperm,+fullfp16,+fp-armv8,+neon", 400},
+    {"sve2-pmull128", AArch64::AEK_NONE, {}, {}, FEAT_SVE_PMULL128, "+sve2,+sve,+sve2-aes,+fullfp16,+fp-armv8,+neon", 390},
     {"sve2-sha3", AArch64::AEK_SVE2SHA3, "+sve2-sha3", "-sve2-sha3", FEAT_SVE_SHA3, "+sve2,+sve,+sve2-sha3,+fullfp16,+fp-armv8,+neon", 410},
     {"sve2-sm4", AArch64::AEK_SVE2SM4, "+sve2-sm4", "-sve2-sm4", FEAT_SVE_SM4, "+sve2,+sve,+sve2-sm4,+fullfp16,+fp-armv8,+neon", 420},
     {"sve2", AArch64::AEK_SVE2, "+sve2", "-sve2", FEAT_SVE2, "+sve2,+sve,+fullfp16,+fp-armv8,+neon", 370},

From fd98f80f602e11d523d252feef301634c2c689a7 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Timm=20B=C3=A4der?= <tbaeder@redhat.com>
Date: Thu, 18 Apr 2024 09:33:20 +0200
Subject: [PATCH 852/886] [clang][Interp] Finish initializing structs from
 CompoundLiteralExprs

---
 clang/lib/AST/Interp/ByteCodeExprGen.cpp |  6 +++---
 clang/test/AST/Interp/c.c                | 12 ++++++++++++
 2 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/clang/lib/AST/Interp/ByteCodeExprGen.cpp b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
index 47cd32c2ab867..f317f506d24f4 100644
--- a/clang/lib/AST/Interp/ByteCodeExprGen.cpp
+++ b/clang/lib/AST/Interp/ByteCodeExprGen.cpp
@@ -1843,7 +1843,7 @@ bool ByteCodeExprGen<Emitter>::VisitCompoundLiteralExpr(
   const Expr *Init = E->getInitializer();
   if (Initializing) {
     // We already have a value, just initialize that.
-    return this->visitInitializer(Init);
+    return this->visitInitializer(Init) && this->emitFinishInit(E);
   }
 
   std::optional<PrimType> T = classify(E->getType());
@@ -1862,7 +1862,7 @@ bool ByteCodeExprGen<Emitter>::VisitCompoundLiteralExpr(
         return this->emitInitGlobal(*T, *GlobalIndex, E);
       }
 
-      return this->visitInitializer(Init);
+      return this->visitInitializer(Init) && this->emitFinishInit(E);
     }
 
     return false;
@@ -1891,7 +1891,7 @@ bool ByteCodeExprGen<Emitter>::VisitCompoundLiteralExpr(
       }
       return this->emitInit(*T, E);
     } else {
-      if (!this->visitInitializer(Init))
+      if (!this->visitInitializer(Init) || !this->emitFinishInit(E))
         return false;
     }
 
diff --git a/clang/test/AST/Interp/c.c b/clang/test/AST/Interp/c.c
index 38df38d1ccfae..5ae9b1dc7bff8 100644
--- a/clang/test/AST/Interp/c.c
+++ b/clang/test/AST/Interp/c.c
@@ -245,3 +245,15 @@ void unaryops(void) {
   (void)((struct zz {float x;}){3}.x++);
   (void)((struct ww {float x;}){3}.x--);
 }
+
+/// This used to fail because we didn't properly mark the struct
+/// initialized through a CompoundLiteralExpr as initialized.
+struct TestStruct {
+  int a;
+  int b;
+};
+int Y __attribute__((annotate(
+  "GlobalValAnnotationWithArgs",
+  42,
+  (struct TestStruct) { .a = 1, .b = 2 }
+)));

From 61f400165c64894374d23ffbcbebc8650d974961 Mon Sep 17 00:00:00 2001
From: Aiden Grossman <aidengrossman@google.com>
Date: Thu, 18 Apr 2024 08:50:40 +0000
Subject: [PATCH 853/886] [llvm-exegesis] Define SYS_gettid if not available

This patch defines SYS_gettid as __NR_gettid if SYS_gettid is not
available to avoid compile time errors due to SYS_gettid not being
defined. This happens with certain libcs (like bionic) that do not
define SYS_gettid.
---
 llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp b/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
index 4699fbbea5def..1d44e09ad61e1 100644
--- a/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
+++ b/llvm/tools/llvm-exegesis/lib/SubprocessMemory.cpp
@@ -24,6 +24,15 @@ namespace exegesis {
 
 #if defined(__linux__)
 
+// The SYS_* macros for system calls are provided by the libc whereas the
+// __NR_* macros are from the linux headers. This means that sometimes
+// SYS_* macros might not be available for certain system calls depending
+// upon the libc. This happens with the gettid syscall and bionic for
+// example, so we use __NR_gettid when no SYS_gettid is available.
+#ifndef SYS_gettid
+#define SYS_gettid __NR_gettid
+#endif
+
 long SubprocessMemory::getCurrentTID() {
   // We're using the raw syscall here rather than the gettid() function provided
   // by most libcs for compatibility as gettid() was only added to glibc in

From 63d8058ef50a3186b6b6a5db254f44673fea3d19 Mon Sep 17 00:00:00 2001
From: Ramkumar Ramachandra <ram.ramachandra@arm.com>
Date: Thu, 18 Apr 2024 10:03:07 +0100
Subject: [PATCH 854/886] LoopVectorize: guard appending InstsToScalarize; fix
 bug (#88720)

In the process of collecting instructions to scalarize, LoopVectorize
uses faulty reasoning whereby it also adds instructions that will be
scalar after vectorization. If an instruction satisfies
isScalarAfterVectorization() for the given VF, it should not be appended
to InstsToScalarize. Add this extra guard, fixing a crash.

Fixes #55096.
---
 .../Transforms/Vectorize/LoopVectorize.cpp    |  3 +-
 .../X86/pr55096-scalarize-add.ll              | 56 ++++++++++++++++---
 2 files changed, 50 insertions(+), 9 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index a8272f4502535..b3b54cc36877f 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -5814,7 +5814,8 @@ void LoopVectorizationCostModel::collectInstsToScalarize(ElementCount VF) {
         // invalid scalarization costs.
         // Do not apply discount logic if hacked cost is needed
         // for emulated masked memrefs.
-        if (!VF.isScalable() && !useEmulatedMaskMemRefHack(&I, VF) &&
+        if (!isScalarAfterVectorization(&I, VF) && !VF.isScalable() &&
+            !useEmulatedMaskMemRefHack(&I, VF) &&
             computePredInstDiscount(&I, ScalarCosts, VF) >= 0)
           ScalarCostsVF.insert(ScalarCosts.begin(), ScalarCosts.end());
         // Remember that BB will remain after vectorization.
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr55096-scalarize-add.ll b/llvm/test/Transforms/LoopVectorize/X86/pr55096-scalarize-add.ll
index d75ab660461ec..cc820f2a8f2a0 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr55096-scalarize-add.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr55096-scalarize-add.ll
@@ -1,13 +1,48 @@
-; RUN: opt -passes=loop-vectorize -force-vector-width=2 -force-vector-interleave=1 -S %s | FileCheck %s
-
-; REQUIRES: asserts
-; XFAIL: *
-
-target triple = "x86_64-apple-macosx"
-
-; CHECK: vector.body
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -mtriple=x86_64-apple-macosx -passes=loop-vectorize,simplifycfg,dce -force-vector-width=2 -force-vector-interleave=1 -S %s | FileCheck %s
 
 define void @test_pr55096(i64 %c, ptr %p) {
+; CHECK-LABEL: define void @test_pr55096(
+; CHECK-SAME: i64 [[C:%.*]], ptr [[P:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[C]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE3:%.*]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 122, i64 123>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE3]] ]
+; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[INDEX]] to i16
+; CHECK-NEXT:    [[TMP0:%.*]] = mul i16 [[DOTCAST]], 2008
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i16 6229, [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <2 x i1> [[TMP1]], <i1 true, i1 true>
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i1> [[TMP2]], i32 0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
+; CHECK:       pred.store.if:
+; CHECK-NEXT:    [[TMP4:%.*]] = add i16 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = add i16 [[TMP4]], 2008
+; CHECK-NEXT:    [[TMP6:%.*]] = udiv i16 4943, [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[P]], i16 [[TMP6]]
+; CHECK-NEXT:    store i16 0, ptr [[TMP7]], align 2
+; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE]]
+; CHECK:       pred.store.continue:
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i1> [[TMP2]], i32 1
+; CHECK-NEXT:    br i1 [[TMP9]], label [[PRED_STORE_IF2:%.*]], label [[PRED_STORE_CONTINUE3]]
+; CHECK:       pred.store.if2:
+; CHECK-NEXT:    [[TMP10:%.*]] = add i16 [[OFFSET_IDX]], 2008
+; CHECK-NEXT:    [[TMP11:%.*]] = add i16 [[TMP10]], 2008
+; CHECK-NEXT:    [[TMP12:%.*]] = udiv i16 4943, [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[P]], i16 [[TMP12]]
+; CHECK-NEXT:    store i16 0, ptr [[TMP13]], align 2
+; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE3]]
+; CHECK:       pred.store.continue3:
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
+; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 340
+; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
 entry:
   br label %loop.header
 
@@ -32,3 +67,8 @@ loop.latch:
 exit:
   ret void
 }
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+;.

From e90bc9cfd4d22c89dd993f62ede700ae25df49c5 Mon Sep 17 00:00:00 2001
From: Matthew Devereau <matthew.devereau@arm.com>
Date: Thu, 18 Apr 2024 10:19:01 +0100
Subject: [PATCH 855/886] Constant Fold Logf128 calls (#84501)

This patch enables constant folding for 128 bit floating-point logf
calls. This is achieved by querying if the host system has the logf128()
symbol available with a CMake test. If so, replace the runtime call with
the compile time value returned from logf128.
---
 llvm/CMakeLists.txt                           |   2 +
 llvm/cmake/config-ix.cmake                    |  11 ++
 llvm/include/llvm/ADT/APFloat.h               |  13 ++
 llvm/include/llvm/ADT/APInt.h                 |   8 ++
 llvm/include/llvm/Config/llvm-config.h.cmake  |   3 +
 llvm/include/llvm/Support/float128.h          |  21 +++
 llvm/lib/Analysis/CMakeLists.txt              |   6 +
 llvm/lib/Analysis/ConstantFolding.cpp         |  11 ++
 llvm/lib/Support/APFloat.cpp                  |  24 ++++
 llvm/test/CMakeLists.txt                      |   1 +
 .../InstSimplify/ConstProp/logf128.ll         | 126 ++++++++++++++++++
 llvm/test/lit.cfg.py                          |   3 +
 llvm/test/lit.site.cfg.py.in                  |   1 +
 llvm/unittests/Analysis/CMakeLists.txt        |   6 +
 14 files changed, 236 insertions(+)
 create mode 100644 llvm/include/llvm/Support/float128.h
 create mode 100644 llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll

diff --git a/llvm/CMakeLists.txt b/llvm/CMakeLists.txt
index d511376e18ba5..c48ab438ffe0b 100644
--- a/llvm/CMakeLists.txt
+++ b/llvm/CMakeLists.txt
@@ -562,6 +562,8 @@ set(LLVM_USE_STATIC_ZSTD FALSE CACHE BOOL "Use static version of zstd. Can be TR
 
 set(LLVM_ENABLE_CURL "OFF" CACHE STRING "Use libcurl for the HTTP client if available. Can be ON, OFF, or FORCE_ON")
 
+set(LLVM_HAS_LOGF128 "OFF" CACHE STRING "Use logf128 to constant fold fp128 logarithm calls. Can be ON, OFF, or FORCE_ON")
+
 set(LLVM_ENABLE_HTTPLIB "OFF" CACHE STRING "Use cpp-httplib HTTP server library if available. Can be ON, OFF, or FORCE_ON")
 
 set(LLVM_Z3_INSTALL_DIR "" CACHE STRING "Install directory of the Z3 solver.")
diff --git a/llvm/cmake/config-ix.cmake b/llvm/cmake/config-ix.cmake
index bf1b110245bb2..ce07cab7a56c6 100644
--- a/llvm/cmake/config-ix.cmake
+++ b/llvm/cmake/config-ix.cmake
@@ -257,6 +257,17 @@ else()
   set(LLVM_ENABLE_TERMINFO 0)
 endif()
 
+if(LLVM_HAS_LOGF128)
+  include(CheckCXXSymbolExists)
+  check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
+
+  if(LLVM_HAS_LOGF128 STREQUAL FORCE_ON AND NOT HAS_LOGF128)
+    message(FATAL_ERROR "Failed to configure logf128")
+  endif()
+
+  set(LLVM_HAS_LOGF128 "${HAS_LOGF128}")
+endif()
+
 # function checks
 check_symbol_exists(arc4random "stdlib.h" HAVE_DECL_ARC4RANDOM)
 find_package(Backtrace)
diff --git a/llvm/include/llvm/ADT/APFloat.h b/llvm/include/llvm/ADT/APFloat.h
index deb74cb2fdeb1..d5794b97bc267 100644
--- a/llvm/include/llvm/ADT/APFloat.h
+++ b/llvm/include/llvm/ADT/APFloat.h
@@ -19,6 +19,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/FloatingPointMode.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/float128.h"
 #include <memory>
 
 #define APFLOAT_DISPATCH_ON_SEMANTICS(METHOD_CALL)                             \
@@ -354,6 +355,9 @@ class IEEEFloat final : public APFloatBase {
   Expected<opStatus> convertFromString(StringRef, roundingMode);
   APInt bitcastToAPInt() const;
   double convertToDouble() const;
+#ifdef __FLOAT128__
+  float128 convertToQuad() const;
+#endif
   float convertToFloat() const;
 
   /// @}
@@ -1218,6 +1222,15 @@ class APFloat : public APFloatBase {
   /// shorter semantics, like IEEEsingle and others.
   double convertToDouble() const;
 
+  /// Converts this APFloat to host float value.
+  ///
+  /// \pre The APFloat must be built using semantics, that can be represented by
+  /// the host float type without loss of precision. It can be IEEEquad and
+  /// shorter semantics, like IEEEdouble and others.
+#ifdef __FLOAT128__
+  float128 convertToQuad() const;
+#endif
+
   /// Converts this APFloat to host float value.
   ///
   /// \pre The APFloat must be built using semantics, that can be represented by
diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index 8d3c029b2e7e9..0f6e4c592af65 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -17,6 +17,7 @@
 
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/MathExtras.h"
+#include "llvm/Support/float128.h"
 #include <cassert>
 #include <climits>
 #include <cstring>
@@ -1670,6 +1671,13 @@ class [[nodiscard]] APInt {
   /// any bit width. Exactly 64 bits will be translated.
   double bitsToDouble() const { return llvm::bit_cast<double>(getWord(0)); }
 
+#ifdef __FLOAT128__
+  float128 bitsToQuad() const {
+    __uint128_t ul = ((__uint128_t)U.pVal[1] << 64) + U.pVal[0];
+    return llvm::bit_cast<float128>(ul);
+  }
+#endif
+
   /// Converts APInt bits to a float
   ///
   /// The conversion does not do a translation from integer to float, it just
diff --git a/llvm/include/llvm/Config/llvm-config.h.cmake b/llvm/include/llvm/Config/llvm-config.h.cmake
index 6605ea60df99e..629977cc11d68 100644
--- a/llvm/include/llvm/Config/llvm-config.h.cmake
+++ b/llvm/include/llvm/Config/llvm-config.h.cmake
@@ -198,4 +198,7 @@
 /* Define if plugins enabled */
 #cmakedefine LLVM_ENABLE_PLUGINS
 
+/* Define if logf128 is available */
+#cmakedefine LLVM_HAS_LOGF128
+
 #endif
diff --git a/llvm/include/llvm/Support/float128.h b/llvm/include/llvm/Support/float128.h
new file mode 100644
index 0000000000000..f7a442fce8dd3
--- /dev/null
+++ b/llvm/include/llvm/Support/float128.h
@@ -0,0 +1,21 @@
+//===-- llvm/Support/float128.h - Compiler abstraction support --*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_FLOAT128
+#define LLVM_FLOAT128
+
+namespace llvm {
+
+#if defined(__clang__) && defined(__FLOAT128__)
+typedef __float128 float128;
+#elif defined(__FLOAT128__) && (defined(__GNUC__) || defined(__GNUG__))
+typedef _Float128 float128;
+#endif
+
+} // namespace llvm
+#endif // LLVM_FLOAT128
diff --git a/llvm/lib/Analysis/CMakeLists.txt b/llvm/lib/Analysis/CMakeLists.txt
index 474b8d20fde16..74476cb5440c6 100644
--- a/llvm/lib/Analysis/CMakeLists.txt
+++ b/llvm/lib/Analysis/CMakeLists.txt
@@ -159,3 +159,9 @@ add_llvm_component_library(LLVMAnalysis
   Support
   TargetParser
   )
+
+include(CheckCXXSymbolExists)
+check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
+if(HAS_LOGF128)
+ target_compile_definitions(LLVMAnalysis PRIVATE HAS_LOGF128)
+endif()
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index 749374a3aa48a..e292ff326a595 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -2089,6 +2089,17 @@ static Constant *ConstantFoldScalarCall1(StringRef Name,
     if (IntrinsicID == Intrinsic::canonicalize)
       return constantFoldCanonicalize(Ty, Call, U);
 
+#if defined(__FLOAT128__) && defined(HAS_LOGF128)
+    if (Ty->isFP128Ty()) {
+      switch (IntrinsicID) {
+      default:
+        return nullptr;
+      case Intrinsic::log:
+        return ConstantFP::get(Ty, logf128(Op->getValueAPF().convertToQuad()));
+      }
+    }
+#endif
+
     if (!Ty->isHalfTy() && !Ty->isFloatTy() && !Ty->isDoubleTy())
       return nullptr;
 
diff --git a/llvm/lib/Support/APFloat.cpp b/llvm/lib/Support/APFloat.cpp
index 0a4f5ac01553f..30c85f5b4214c 100644
--- a/llvm/lib/Support/APFloat.cpp
+++ b/llvm/lib/Support/APFloat.cpp
@@ -3670,6 +3670,15 @@ double IEEEFloat::convertToDouble() const {
   return api.bitsToDouble();
 }
 
+#ifdef __FLOAT128__
+float128 IEEEFloat::convertToQuad() const {
+  assert(semantics == (const llvm::fltSemantics *)&semIEEEquad &&
+         "Float semantics are not IEEEquads");
+  APInt api = bitcastToAPInt();
+  return api.bitsToQuad();
+}
+#endif
+
 /// Integer bit is explicit in this format.  Intel hardware (387 and later)
 /// does not support these bit patterns:
 ///  exponent = all 1's, integer bit 0, significand 0 ("pseudoinfinity")
@@ -5265,6 +5274,21 @@ double APFloat::convertToDouble() const {
   return Temp.getIEEE().convertToDouble();
 }
 
+#ifdef __FLOAT128__
+float128 APFloat::convertToQuad() const {
+  if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEquad)
+    return getIEEE().convertToQuad();
+  assert(getSemantics().isRepresentableBy(semIEEEquad) &&
+         "Float semantics is not representable by IEEEquad");
+  APFloat Temp = *this;
+  bool LosesInfo;
+  opStatus St = Temp.convert(semIEEEquad, rmNearestTiesToEven, &LosesInfo);
+  assert(!(St & opInexact) && !LosesInfo && "Unexpected imprecision");
+  (void)St;
+  return Temp.getIEEE().convertToQuad();
+}
+#endif
+
 float APFloat::convertToFloat() const {
   if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEsingle)
     return getIEEE().convertToFloat();
diff --git a/llvm/test/CMakeLists.txt b/llvm/test/CMakeLists.txt
index 6127b76db06b7..b419464aba3f5 100644
--- a/llvm/test/CMakeLists.txt
+++ b/llvm/test/CMakeLists.txt
@@ -25,6 +25,7 @@ llvm_canonicalize_cmake_booleans(
   LLVM_INCLUDE_DXIL_TESTS
   LLVM_TOOL_LLVM_DRIVER_BUILD
   LLVM_INCLUDE_SPIRV_TOOLS_TESTS
+  LLVM_HAS_LOGF128
   )
 
 configure_lit_site_cfg(
diff --git a/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll b/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll
new file mode 100644
index 0000000000000..da56997f69382
--- /dev/null
+++ b/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll
@@ -0,0 +1,126 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt < %s -passes=instsimplify -S | FileCheck %s
+
+; REQUIRES: has_logf128
+declare fp128 @llvm.log.f128(fp128)
+
+define fp128 @log_e_64(){
+; CHECK-LABEL: define fp128 @log_e_64() {
+; CHECK-NEXT:    ret fp128 0xL300000000000000040010A2B23F3BAB7
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000004005000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_smallest_positive_subnormal_number(){
+; CHECK-LABEL: define fp128 @log_e_smallest_positive_subnormal_number() {
+; CHECK-NEXT:    ret fp128 0xL3000000000000000C00C654628220780
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000010000000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_largest_subnormal_number(){
+; CHECK-LABEL: define fp128 @log_e_largest_subnormal_number() {
+; CHECK-NEXT:    ret fp128 0xLD000000000000000C00C62D918CE2421
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF0000FFFFFFFFFFFF)
+  ret fp128 %A
+}
+
+define fp128 @log_e_smallest_positive_normal_number(){
+;
+; CHECK-LABEL: define fp128 @log_e_smallest_positive_normal_number() {
+; CHECK-NEXT:    ret fp128 0xLD000000000000000C00C62D918CE2421
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000000001000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_largest_normal_number(){
+; CHECK-LABEL: define fp128 @log_e_largest_normal_number() {
+; CHECK-NEXT:    ret fp128 0xLF000000000000000400C62E42FEFA39E
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF7FFEFFFFFFFFFFFF)
+  ret fp128 %A
+}
+
+define fp128 @log_e_largest_number_less_than_one(){
+; CHECK-LABEL: define fp128 @log_e_largest_number_less_than_one() {
+; CHECK-NEXT:    ret fp128 0xL0000000000000000BF8E000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF3FFEFFFFFFFFFFFF)
+  ret fp128 %A
+}
+
+define fp128 @log_e_1(){
+; CHECK-LABEL: define fp128 @log_e_1() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000000000000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000003FFF000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_smallest_number_larger_than_one(){
+; CHECK-LABEL: define fp128 @log_e_smallest_number_larger_than_one() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000003F8F000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000013FFF000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_negative_2(){
+; CHECK-LABEL: define fp128 @log_e_negative_2() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL0000000000000000C000000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_0(){
+; CHECK-LABEL: define fp128 @log_e_0() {
+; CHECK-NEXT:    ret fp128 0xL0000000000000000FFFF000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000000000000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_negative_0(){
+; CHECK-LABEL: define fp128 @log_e_negative_0() {
+; CHECK-NEXT:    ret fp128 0xL0000000000000000FFFF000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000008000000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_infinity(){
+; CHECK-LABEL: define fp128 @log_e_infinity() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF000000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000007FFF000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_negative_infinity(){
+; CHECK-LABEL: define fp128 @log_e_negative_infinity() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000000
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL0000000000000000FFFF000000000000)
+  ret fp128 %A
+}
+
+define fp128 @log_e_nan(){
+; CHECK-LABEL: define fp128 @log_e_nan() {
+; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000001
+;
+  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000007FFF000000000001)
+  ret fp128 %A
+}
+
+define <2 x fp128> @log_e_negative_2_vector(){
+; CHECK-LABEL: define <2 x fp128> @log_e_negative_2_vector() {
+; CHECK-NEXT:    ret <2 x fp128> <fp128 0xL00000000000000007FFF800000000000, fp128 0xL00000000000000007FFF800000000000>
+;
+  %A = call <2 x fp128> @llvm.log.v2f128(<2 x fp128> <fp128 0xL0000000000000000C000000000000000, fp128 0xL0000000000000000C000000000000001>)
+  ret <2 x fp128> %A
+}
diff --git a/llvm/test/lit.cfg.py b/llvm/test/lit.cfg.py
index 4c05317036d1a..36bcc48307dd8 100644
--- a/llvm/test/lit.cfg.py
+++ b/llvm/test/lit.cfg.py
@@ -614,3 +614,6 @@ def have_ld64_plugin_support():
 # "OBJECT_MODE" to 'any' by default on AIX OS.
 if "system-aix" in config.available_features:
     config.environment["OBJECT_MODE"] = "any"
+
+if config.has_logf128:
+    config.available_features.add("has_logf128")
diff --git a/llvm/test/lit.site.cfg.py.in b/llvm/test/lit.site.cfg.py.in
index b6f255d472d16..27cb1dd6477d9 100644
--- a/llvm/test/lit.site.cfg.py.in
+++ b/llvm/test/lit.site.cfg.py.in
@@ -61,6 +61,7 @@ config.reverse_iteration = @LLVM_ENABLE_REVERSE_ITERATION@
 config.dxil_tests = @LLVM_INCLUDE_DXIL_TESTS@
 config.have_llvm_driver = @LLVM_TOOL_LLVM_DRIVER_BUILD@
 config.spirv_tools_tests = @LLVM_INCLUDE_SPIRV_TOOLS_TESTS@
+config.has_logf128 = @LLVM_HAS_LOGF128@
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)
diff --git a/llvm/unittests/Analysis/CMakeLists.txt b/llvm/unittests/Analysis/CMakeLists.txt
index b1aeaa6e71fd4..9b3778f8a3f98 100644
--- a/llvm/unittests/Analysis/CMakeLists.txt
+++ b/llvm/unittests/Analysis/CMakeLists.txt
@@ -80,5 +80,11 @@ if(NOT WIN32)
   export_executable_symbols_for_plugins(AnalysisTests)
 endif()
 
+include(CheckCXXSymbolExists)
+check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
+if(HAS_LOGF128)
+  target_compile_definitions(AnalysisTests PRIVATE HAS_LOGF128)
+endif()
+
 add_subdirectory(InlineAdvisorPlugin)
 add_subdirectory(InlineOrderPlugin)

From 3e64f8a4e74cdcaf5920879c86e7e0a827f6ec13 Mon Sep 17 00:00:00 2001
From: Nashe Mncube <nashe.mncube@arm.com>
Date: Thu, 18 Apr 2024 10:53:23 +0100
Subject: [PATCH 856/886] [AArch64][CodeGen] Fix illegal register aliasing bug
 for mops instrs (#88869)

A bug was found where mops instructions were being generated that
aliased the source and size registers. This is unpredictable behaviour.
This patch uses the earlyclobber constraint on the input source
register so that it doesn't alias with the size register. Also a test is
introduced which checks affected instructions can't violate this
constraint.
---
 llvm/lib/Target/AArch64/AArch64InstrInfo.td   |  2 +-
 .../CodeGen/AArch64/mops-register-alias.ll    | 21 +++++++++++++++++++
 .../AArch64/armv9.3a-mops-register-aliasing.s | 15 +++++++++++++
 3 files changed, 37 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/CodeGen/AArch64/mops-register-alias.ll
 create mode 100644 llvm/test/MC/AArch64/armv9.3a-mops-register-aliasing.s

diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index e1624f70185e1..3bf90778363c6 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -9486,7 +9486,7 @@ let Predicates = [HasMOPS], Defs = [NZCV], Size = 12, mayStore = 1 in {
   let mayLoad = 0 in {
     def MOPSMemorySetPseudo  : Pseudo<(outs GPR64common:$Rd_wb, GPR64:$Rn_wb),
                                       (ins GPR64common:$Rd, GPR64:$Rn, GPR64:$Rm),
-                                      [], "$Rd = $Rd_wb,$Rn = $Rn_wb">, Sched<[]>;
+                                      [], "$Rd = $Rd_wb,$Rn = $Rn_wb,@earlyclobber $Rn_wb">, Sched<[]>;
   }
 }
 let Predicates = [HasMOPS, HasMTE], Defs = [NZCV], Size = 12, mayLoad = 0, mayStore = 1 in {
diff --git a/llvm/test/CodeGen/AArch64/mops-register-alias.ll b/llvm/test/CodeGen/AArch64/mops-register-alias.ll
new file mode 100644
index 0000000000000..855ab959c4e76
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/mops-register-alias.ll
@@ -0,0 +1,21 @@
+; RUN: llc -O1 -mtriple=aarch64-none-linux-gnu -mattr=+mops -o - %s  | FileCheck %s
+
+define void @call_memset_intrinsic() #0 {
+; CHECK-LABEL: call_memset_intrinsic:
+; CHECK:       // %bb.0: // %entry
+; CHECK:         setp [x{{[0-9]+}}]!, x{{[0-9]+}}!, x{{[0-9]+}}
+; CHECK-NOT:     setp [x{{[0-9]+}}]!, x[[REG:[0-9]+]]!, x[[REG]]
+; CHECK-NEXT:    setm [x{{[0-9]+}}]!, x{{[0-9]+}}!, x{{[0-9]+}}
+; CHECK-NOT:     setm [x{{[0-9]+}}]!, x[[REG:[0-9]+]]!, x[[REG]]
+; CHECK-NEXT:    sete [x{{[0-9]+}}]!, x{{[0-9]+}}!, x{{[0-9]+}}
+; CHECK-NOT:     sete [x{{[0-9]+}}]!, x[[REG:[0-9]+]]!, x[[REG]]
+entry:
+
+    %V0 = alloca [65 x i8], align 1
+    call void @llvm.memset.p0.i64(ptr noundef nonnull align 1 dereferenceable(64) %V0, i8 64, i64 64, i1 false)
+    %add.ptr = getelementptr inbounds i8, ptr %V0, i64 64
+     call void @llvm.memset.p0.i64(ptr noundef nonnull align 1 dereferenceable(64) %add.ptr, i8 64, i64 64, i1 false)
+    ret void
+}
+
+attributes #0 = { "target-cpu"="generic" "target-features"="+mops,+strict-align,+v9.3a" }
diff --git a/llvm/test/MC/AArch64/armv9.3a-mops-register-aliasing.s b/llvm/test/MC/AArch64/armv9.3a-mops-register-aliasing.s
new file mode 100644
index 0000000000000..109b33857a9d2
--- /dev/null
+++ b/llvm/test/MC/AArch64/armv9.3a-mops-register-aliasing.s
@@ -0,0 +1,15 @@
+// RUN: not llvm-mc -triple aarch64 -mattr=+mops < %s 2>&1 | FileCheck %s
+
+  setp  [x0]!, x1!, x1
+  setm  [x0]!, x1!, x1
+  sete  [x0]!, x1!, x1
+
+// CHECK:      error: invalid SET instruction, source and size registers are the same
+// CHECK-NEXT:   setp  [x0]!, x1!, x1
+// CHECK-NEXT:         ^
+// CHECK-NEXT: error: invalid SET instruction, source and size registers are the same
+// CHECK-NEXT:   setm  [x0]!, x1!, x1
+// CHECK-NEXT:         ^
+// CHECK-NEXT: error: invalid SET instruction, source and size registers are the same
+// CHECK-NEXT:   sete  [x0]!, x1!, x1
+// CHECK-NEXT:         ^

From 750de326ef0408df4a5f91773c355ded1a1b1b77 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Thu, 18 Apr 2024 13:29:38 +0300
Subject: [PATCH 857/886] [COFF] Rename the COFFShortExport::AliasTarget field.
 NFC. (#89039)

It turns out that the previous name is vaguely misleading.

When operating on a def file like "symbolname == dllname", that is
supposed to make an import library entry, that when the symbol
"symbolname" links against this, it imports the DLL symbol "dllname"
from the referenced DLL. This doesn't need to involve any alias, and it
doesn't need to imply that "dllname" is available on its own as a
separate symbol in the import library at all.

GNU dlltool implements import libraries in the form of "long import
library", where each member is a regular object file with section chunks
that compose the relevant .idata section pieces. There, this kind of
import renaming does not involve any form of aliases, but the right
.idata section just gets a different string than the symbol name.
---
 lld/COFF/Config.h                                   | 4 ++--
 lld/COFF/Driver.cpp                                 | 4 ++--
 llvm/include/llvm/Object/COFFImportFile.h           | 6 ++++--
 llvm/lib/Object/COFFImportFile.cpp                  | 6 +++---
 llvm/lib/Object/COFFModuleDefinition.cpp            | 6 +++---
 llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp | 2 +-
 6 files changed, 15 insertions(+), 13 deletions(-)

diff --git a/lld/COFF/Config.h b/lld/COFF/Config.h
index 917f88fc28280..947f3fead54e0 100644
--- a/lld/COFF/Config.h
+++ b/lld/COFF/Config.h
@@ -55,7 +55,7 @@ struct Export {
   StringRef name;       // N in /export:N or /export:E=N
   StringRef extName;    // E in /export:E=N
   StringRef exportAs;   // E in /export:N,EXPORTAS,E
-  StringRef aliasTarget; // GNU specific: N in "alias == N"
+  StringRef importName; // GNU specific: N in "othername == N"
   Symbol *sym = nullptr;
   uint16_t ordinal = 0;
   bool noname = false;
@@ -75,7 +75,7 @@ struct Export {
 
   bool operator==(const Export &e) const {
     return (name == e.name && extName == e.extName && exportAs == e.exportAs &&
-            aliasTarget == e.aliasTarget && ordinal == e.ordinal &&
+            importName == e.importName && ordinal == e.ordinal &&
             noname == e.noname && data == e.data && isPrivate == e.isPrivate);
   }
 };
diff --git a/lld/COFF/Driver.cpp b/lld/COFF/Driver.cpp
index b0365b5b94417..6c7f70a0b2aff 100644
--- a/lld/COFF/Driver.cpp
+++ b/lld/COFF/Driver.cpp
@@ -934,7 +934,7 @@ void LinkerDriver::createImportLibrary(bool asLib) {
     e2.SymbolName = std::string(e1.symbolName);
     e2.ExtName = std::string(e1.extName);
     e2.ExportAs = std::string(e1.exportAs);
-    e2.AliasTarget = std::string(e1.aliasTarget);
+    e2.ImportName = std::string(e1.importName);
     e2.Ordinal = e1.ordinal;
     e2.Noname = e1.noname;
     e2.Data = e1.data;
@@ -1034,7 +1034,7 @@ void LinkerDriver::parseModuleDefs(StringRef path) {
       e2.extName = saver().save(e1.ExtName);
     }
     e2.exportAs = saver().save(e1.ExportAs);
-    e2.aliasTarget = saver().save(e1.AliasTarget);
+    e2.importName = saver().save(e1.ImportName);
     e2.ordinal = e1.Ordinal;
     e2.noname = e1.Noname;
     e2.data = e1.Data;
diff --git a/llvm/include/llvm/Object/COFFImportFile.h b/llvm/include/llvm/Object/COFFImportFile.h
index 2c06f529ecdfc..649fb4930934d 100644
--- a/llvm/include/llvm/Object/COFFImportFile.h
+++ b/llvm/include/llvm/Object/COFFImportFile.h
@@ -98,9 +98,11 @@ struct COFFShortExport {
   /// "/export:foo=bar", this could be "_bar@8" if bar is stdcall.
   std::string SymbolName;
 
-  /// Creates a weak alias. This is the name of the weak aliasee. In a .def
+  /// Creates an import library entry that imports from a DLL export with a
+  /// different name. This is the name of the DLL export that should be
+  /// referenced when linking against this import library entry. In a .def
   /// file, this is "baz" in "EXPORTS\nfoo = bar == baz".
-  std::string AliasTarget;
+  std::string ImportName;
 
   /// Specifies EXPORTAS name. In a .def file, this is "bar" in
   /// "EXPORTS\nfoo EXPORTAS bar".
diff --git a/llvm/lib/Object/COFFImportFile.cpp b/llvm/lib/Object/COFFImportFile.cpp
index 48c3ea0ed8f4e..fd8aca393e90b 100644
--- a/llvm/lib/Object/COFFImportFile.cpp
+++ b/llvm/lib/Object/COFFImportFile.cpp
@@ -690,9 +690,9 @@ Error writeImportLibrary(StringRef ImportName, StringRef Path,
         Name.swap(*ReplacedName);
       }
 
-      if (!E.AliasTarget.empty() && Name != E.AliasTarget) {
-        Members.push_back(OF.createWeakExternal(E.AliasTarget, Name, false, M));
-        Members.push_back(OF.createWeakExternal(E.AliasTarget, Name, true, M));
+      if (!E.ImportName.empty() && Name != E.ImportName) {
+        Members.push_back(OF.createWeakExternal(E.ImportName, Name, false, M));
+        Members.push_back(OF.createWeakExternal(E.ImportName, Name, true, M));
         continue;
       }
 
diff --git a/llvm/lib/Object/COFFModuleDefinition.cpp b/llvm/lib/Object/COFFModuleDefinition.cpp
index 549348ac9226d..0c0bef1319e44 100644
--- a/llvm/lib/Object/COFFModuleDefinition.cpp
+++ b/llvm/lib/Object/COFFModuleDefinition.cpp
@@ -281,9 +281,9 @@ class Parser {
       }
       if (Tok.K == EqualEqual) {
         read();
-        E.AliasTarget = std::string(Tok.Value);
-        if (AddUnderscores && !isDecorated(E.AliasTarget, MingwDef))
-          E.AliasTarget = std::string("_").append(E.AliasTarget);
+        E.ImportName = std::string(Tok.Value);
+        if (AddUnderscores && !isDecorated(E.ImportName, MingwDef))
+          E.ImportName = std::string("_").append(E.ImportName);
         continue;
       }
       // EXPORTAS must be at the end of export definition
diff --git a/llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp b/llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp
index d5fa05606d7f8..15e4cac08cd4e 100644
--- a/llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp
+++ b/llvm/lib/ToolDrivers/llvm-dlltool/DlltoolDriver.cpp
@@ -227,7 +227,7 @@ int llvm::dlltoolDriverMain(llvm::ArrayRef<const char *> ArgsArr) {
 
   if (Machine == IMAGE_FILE_MACHINE_I386 && Args.hasArg(OPT_k)) {
     for (COFFShortExport &E : Exports) {
-      if (!E.AliasTarget.empty() || (!E.Name.empty() && E.Name[0] == '?'))
+      if (!E.ImportName.empty() || (!E.Name.empty() && E.Name[0] == '?'))
         continue;
       E.SymbolName = E.Name;
       // Trim off the trailing decoration. Symbols will always have a

From d17db6066d2524856fab493dd894f8396e896bc7 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Martin=20Storsj=C3=B6?= <martin@martin.st>
Date: Thu, 18 Apr 2024 13:30:29 +0300
Subject: [PATCH 858/886] [LLD] [COFF] Don't create pseudo relocations for
 discardable sections (#89043)

This extends on the case from 9c970d5ecd6a85188cd2b0a941fcd4d60063ef81;
if a section is marked discardable, it won't be mapped into memory at
runtime, so there's no point in creating runtime pseudo relocations for
such sections.
---
 lld/COFF/Writer.cpp              |  4 +++
 lld/test/COFF/autoimport-debug.s | 43 ++++++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+)
 create mode 100644 lld/test/COFF/autoimport-debug.s

diff --git a/lld/COFF/Writer.cpp b/lld/COFF/Writer.cpp
index 7269d156752de..68993232f09c0 100644
--- a/lld/COFF/Writer.cpp
+++ b/lld/COFF/Writer.cpp
@@ -2060,6 +2060,10 @@ void Writer::createRuntimePseudoRelocs() {
     auto *sc = dyn_cast<SectionChunk>(c);
     if (!sc || !sc->live)
       continue;
+    // Don't create pseudo relocations for sections that won't be
+    // mapped at runtime.
+    if (sc->header->Characteristics & IMAGE_SCN_MEM_DISCARDABLE)
+      continue;
     sc->getRuntimePseudoRelocs(rels);
   }
 
diff --git a/lld/test/COFF/autoimport-debug.s b/lld/test/COFF/autoimport-debug.s
new file mode 100644
index 0000000000000..1f31a2b9e554f
--- /dev/null
+++ b/lld/test/COFF/autoimport-debug.s
@@ -0,0 +1,43 @@
+# REQUIRES: x86
+# RUN: split-file %s %t.dir
+
+## We've got references to variable both in a .refptr and in .debug_info.
+## The .debug_info section should be discardable, so no pseudo relocations
+## need to be created in it. The .refptr section should be elimiated
+## and redirected to __imp_variable instead, so we shouldn't need to
+## create any runtime pseudo relocations. Thus, test that we can link
+## successfully with -runtime-pseudo-reloc:no, while keeping the
+## debug info.
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/lib.s -filetype=obj -o %t.dir/lib.obj
+# RUN: lld-link -out:%t.dir/lib.dll -dll -entry:DllMainCRTStartup %t.dir/lib.obj -lldmingw -implib:%t.dir/lib.lib
+
+# RUN: llvm-mc -triple=x86_64-windows-gnu %t.dir/main.s -filetype=obj -o %t.dir/main.obj
+# RUN: lld-link -lldmingw -out:%t.dir/main.exe -entry:main %t.dir/main.obj %t.dir/lib.lib -opt:noref -debug:dwarf -runtime-pseudo-reloc:no
+
+#--- main.s
+    .global main
+    .text
+main:
+    movq .refptr.variable(%rip), %rax
+    ret
+
+    .section .rdata$.refptr.variable,"dr",discard,.refptr.variable
+    .global .refptr.variable
+.refptr.variable:
+    .quad   variable
+
+    .section .debug_info
+    .long 1
+    .quad variable
+    .long 2
+
+#--- lib.s
+    .global variable
+    .global DllMainCRTStartup
+    .text
+DllMainCRTStartup:
+    ret
+    .data
+variable:
+    .long 42

From 3da065896b1b59fd8291958e8d13f4a942d51214 Mon Sep 17 00:00:00 2001
From: David Spickett <david.spickett@linaro.org>
Date: Thu, 18 Apr 2024 10:42:35 +0000
Subject: [PATCH 859/886] Revert "CompilerRT: Normalize
 COMPILER_RT_DEFAULT_TARGET_TRIPLE (#88835)"

This reverts commit 16f188761da1df6ba5e6627b8742aacfec8e9ec5.

This broke Libcxx Picolib testing at the install step, and builds
for Windows builtins. Revert while we figure out the cause.
---
 compiler-rt/cmake/Modules/CompilerRTUtils.cmake | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
index 6d413f6753bc0..e8e5f612d5b03 100644
--- a/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
+++ b/compiler-rt/cmake/Modules/CompilerRTUtils.cmake
@@ -368,12 +368,6 @@ macro(construct_compiler_rt_default_triple)
           "Default triple for which compiler-rt runtimes will be built.")
   endif()
 
-  if ("${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
-    execute_process(COMMAND ${CMAKE_C_COMPILER} --target=${COMPILER_RT_DEFAULT_TARGET_TRIPLE} -print-effective-triple
-                    OUTPUT_VARIABLE COMPILER_RT_DEFAULT_TARGET_TRIPLE
-                    OUTPUT_STRIP_TRAILING_WHITESPACE)
-  endif()
-
   string(REPLACE "-" ";" LLVM_TARGET_TRIPLE_LIST ${COMPILER_RT_DEFAULT_TARGET_TRIPLE})
   list(GET LLVM_TARGET_TRIPLE_LIST 0 COMPILER_RT_DEFAULT_TARGET_ARCH)
 

From 8d6a9c05f6d676c93c84ebf06cf6263657e74c00 Mon Sep 17 00:00:00 2001
From: Orlando Cazalet-Hyams <orlando.hyams@sony.com>
Date: Thu, 18 Apr 2024 12:08:31 +0100
Subject: [PATCH 860/886] [DWARF] Add support for DW_TAG_template_alias for
 template aliases (#88943)

Part 1 of fix for issue
https://github.com/llvm/llvm-project/issues/54624

Split from PR #87623. Clang front end changes to follow.

Use DICompositeType to represent the template alias, using its extraData
field as a tuple of DITemplateParameter to describe the template
parameters.

Added template-alias.ll  - Check DWARF emission.
Modified  frame-types.s  - Check llvm-symbolizer understands the DIE.
---
 llvm/include/llvm/IR/DIBuilder.h              |  18 +
 llvm/include/llvm/IR/DebugInfoMetadata.h      |   9 +-
 .../CodeGen/AsmPrinter/DebugHandlerBase.cpp   |   6 +-
 llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp    |   1 +
 llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp     |   5 +
 llvm/lib/DWARFLinker/Classic/DWARFLinker.cpp  |   1 +
 llvm/lib/DebugInfo/DWARF/DWARFDie.cpp         |   1 +
 llvm/lib/IR/DIBuilder.cpp                     |  11 +
 llvm/lib/IR/Verifier.cpp                      |   3 +-
 llvm/test/DebugInfo/X86/template-alias.ll     |  71 ++
 llvm/test/tools/llvm-symbolizer/frame-types.s | 967 ++++++++++--------
 11 files changed, 675 insertions(+), 418 deletions(-)
 create mode 100644 llvm/test/DebugInfo/X86/template-alias.ll

diff --git a/llvm/include/llvm/IR/DIBuilder.h b/llvm/include/llvm/IR/DIBuilder.h
index 002f2db6da544..97ea38f041baa 100644
--- a/llvm/include/llvm/IR/DIBuilder.h
+++ b/llvm/include/llvm/IR/DIBuilder.h
@@ -310,6 +310,24 @@ namespace llvm {
                                  DINode::DIFlags Flags = DINode::FlagZero,
                                  DINodeArray Annotations = nullptr);
 
+    /// Create debugging information entry for a template alias.
+    /// \param Ty          Original type.
+    /// \param Name        Alias name.
+    /// \param File        File where this type is defined.
+    /// \param LineNo      Line number.
+    /// \param Context     The surrounding context for the alias.
+    /// \param TParams     The template arguments.
+    /// \param AlignInBits Alignment. (optional)
+    /// \param Flags       Flags to describe inheritance attribute (optional),
+    ///                    e.g. private.
+    /// \param Annotations Annotations. (optional)
+    DIDerivedType *createTemplateAlias(DIType *Ty, StringRef Name, DIFile *File,
+                                       unsigned LineNo, DIScope *Context,
+                                       DINodeArray TParams,
+                                       uint32_t AlignInBits = 0,
+                                       DINode::DIFlags Flags = DINode::FlagZero,
+                                       DINodeArray Annotations = nullptr);
+
     /// Create debugging information entry for a 'friend'.
     DIDerivedType *createFriend(DIType *Ty, DIType *FriendTy);
 
diff --git a/llvm/include/llvm/IR/DebugInfoMetadata.h b/llvm/include/llvm/IR/DebugInfoMetadata.h
index 2805f6c478057..42291d45da2be 100644
--- a/llvm/include/llvm/IR/DebugInfoMetadata.h
+++ b/llvm/include/llvm/IR/DebugInfoMetadata.h
@@ -1093,14 +1093,19 @@ class DIDerivedType : public DIType {
   /// Get extra data associated with this derived type.
   ///
   /// Class type for pointer-to-members, objective-c property node for ivars,
-  /// global constant wrapper for static members, or virtual base pointer offset
-  /// for inheritance.
+  /// global constant wrapper for static members, virtual base pointer offset
+  /// for inheritance, or a tuple of template parameters for template aliases.
   ///
   /// TODO: Separate out types that need this extra operand: pointer-to-member
   /// types and member fields (static members and ivars).
   Metadata *getExtraData() const { return getRawExtraData(); }
   Metadata *getRawExtraData() const { return getOperand(4); }
 
+  /// Get the template parameters from a template alias.
+  DITemplateParameterArray getTemplateParams() const {
+    return cast_or_null<MDTuple>(getExtraData());
+  }
+
   /// Get annotations associated with this derived type.
   DINodeArray getAnnotations() const {
     return cast_or_null<MDTuple>(getRawAnnotations());
diff --git a/llvm/lib/CodeGen/AsmPrinter/DebugHandlerBase.cpp b/llvm/lib/CodeGen/AsmPrinter/DebugHandlerBase.cpp
index f00ff1565c665..24cd1b15a5736 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DebugHandlerBase.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DebugHandlerBase.cpp
@@ -154,7 +154,8 @@ uint64_t DebugHandlerBase::getBaseTypeSize(const DIType *Ty) {
   if (Tag != dwarf::DW_TAG_member && Tag != dwarf::DW_TAG_typedef &&
       Tag != dwarf::DW_TAG_const_type && Tag != dwarf::DW_TAG_volatile_type &&
       Tag != dwarf::DW_TAG_restrict_type && Tag != dwarf::DW_TAG_atomic_type &&
-      Tag != dwarf::DW_TAG_immutable_type)
+      Tag != dwarf::DW_TAG_immutable_type &&
+      Tag != dwarf::DW_TAG_template_alias)
     return DDTy->getSizeInBits();
 
   DIType *BaseType = DDTy->getBaseType();
@@ -210,7 +211,8 @@ bool DebugHandlerBase::isUnsignedDIType(const DIType *Ty) {
     assert(T == dwarf::DW_TAG_typedef || T == dwarf::DW_TAG_const_type ||
            T == dwarf::DW_TAG_volatile_type ||
            T == dwarf::DW_TAG_restrict_type || T == dwarf::DW_TAG_atomic_type ||
-           T == dwarf::DW_TAG_immutable_type);
+           T == dwarf::DW_TAG_immutable_type ||
+           T == dwarf::DW_TAG_template_alias);
     assert(DTy->getBaseType() && "Expected valid base type");
     return isUnsignedDIType(DTy->getBaseType());
   }
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index a85a8e04666aa..c7fa10775ada7 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -2483,6 +2483,7 @@ static dwarf::PubIndexEntryDescriptor computeIndexValue(DwarfUnit *CU,
   case dwarf::DW_TAG_typedef:
   case dwarf::DW_TAG_base_type:
   case dwarf::DW_TAG_subrange_type:
+  case dwarf::DW_TAG_template_alias:
     return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_TYPE, dwarf::GIEL_STATIC);
   case dwarf::DW_TAG_namespace:
     return dwarf::GIEK_TYPE;
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
index c40beeeb925e0..56c288ee95b43 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
@@ -803,6 +803,11 @@ void DwarfUnit::constructTypeDIE(DIE &Buffer, const DIDerivedType *DTy) {
   if (DTy->getDWARFAddressSpace())
     addUInt(Buffer, dwarf::DW_AT_address_class, dwarf::DW_FORM_data4,
             *DTy->getDWARFAddressSpace());
+
+  // Add template alias template parameters.
+  if (Tag == dwarf::DW_TAG_template_alias)
+    addTemplateParams(Buffer, DTy->getTemplateParams());
+
   if (auto PtrAuthData = DTy->getPtrAuthData()) {
     addUInt(Buffer, dwarf::DW_AT_LLVM_ptrauth_key, dwarf::DW_FORM_data1,
             PtrAuthData->key());
diff --git a/llvm/lib/DWARFLinker/Classic/DWARFLinker.cpp b/llvm/lib/DWARFLinker/Classic/DWARFLinker.cpp
index 60f664ece7eef..3149d9b1d6624 100644
--- a/llvm/lib/DWARFLinker/Classic/DWARFLinker.cpp
+++ b/llvm/lib/DWARFLinker/Classic/DWARFLinker.cpp
@@ -116,6 +116,7 @@ static bool isTypeTag(uint16_t Tag) {
   case dwarf::DW_TAG_string_type:
   case dwarf::DW_TAG_structure_type:
   case dwarf::DW_TAG_subroutine_type:
+  case dwarf::DW_TAG_template_alias:
   case dwarf::DW_TAG_typedef:
   case dwarf::DW_TAG_union_type:
   case dwarf::DW_TAG_ptr_to_member_type:
diff --git a/llvm/lib/DebugInfo/DWARF/DWARFDie.cpp b/llvm/lib/DebugInfo/DWARF/DWARFDie.cpp
index 66492f7bf8043..410842a80b015 100644
--- a/llvm/lib/DebugInfo/DWARF/DWARFDie.cpp
+++ b/llvm/lib/DebugInfo/DWARF/DWARFDie.cpp
@@ -515,6 +515,7 @@ getTypeSizeImpl(DWARFDie Die, uint64_t PointerSize,
   case DW_TAG_immutable_type:
   case DW_TAG_volatile_type:
   case DW_TAG_restrict_type:
+  case DW_TAG_template_alias:
   case DW_TAG_typedef: {
     if (DWARFDie BaseType = Die.getAttributeValueAsReferencedDie(DW_AT_type))
       return getTypeSizeImpl(BaseType, PointerSize, Visited);
diff --git a/llvm/lib/IR/DIBuilder.cpp b/llvm/lib/IR/DIBuilder.cpp
index f86e557b8def2..f39149ae0dad4 100644
--- a/llvm/lib/IR/DIBuilder.cpp
+++ b/llvm/lib/IR/DIBuilder.cpp
@@ -356,6 +356,17 @@ DIDerivedType *DIBuilder::createTypedef(DIType *Ty, StringRef Name,
                             nullptr, Annotations);
 }
 
+DIDerivedType *
+DIBuilder::createTemplateAlias(DIType *Ty, StringRef Name, DIFile *File,
+                               unsigned LineNo, DIScope *Context,
+                               DINodeArray TParams, uint32_t AlignInBits,
+                               DINode::DIFlags Flags, DINodeArray Annotations) {
+  return DIDerivedType::get(VMContext, dwarf::DW_TAG_template_alias, Name, File,
+                            LineNo, getNonCompileUnitScope(Context), Ty, 0,
+                            AlignInBits, 0, std::nullopt, std::nullopt, Flags,
+                            TParams.get(), Annotations);
+}
+
 DIDerivedType *DIBuilder::createFriend(DIType *Ty, DIType *FriendTy) {
   assert(Ty && "Invalid type!");
   assert(FriendTy && "Invalid friend type!");
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 4cd61e6e531bf..0d6db3da9e614 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -1240,7 +1240,8 @@ void Verifier::visitDIDerivedType(const DIDerivedType &N) {
               (N.getTag() == dwarf::DW_TAG_variable && N.isStaticMember()) ||
               N.getTag() == dwarf::DW_TAG_inheritance ||
               N.getTag() == dwarf::DW_TAG_friend ||
-              N.getTag() == dwarf::DW_TAG_set_type,
+              N.getTag() == dwarf::DW_TAG_set_type ||
+              N.getTag() == dwarf::DW_TAG_template_alias,
           "invalid tag", &N);
   if (N.getTag() == dwarf::DW_TAG_ptr_to_member_type) {
     CheckDI(isType(N.getRawExtraData()), "invalid pointer to member type", &N,
diff --git a/llvm/test/DebugInfo/X86/template-alias.ll b/llvm/test/DebugInfo/X86/template-alias.ll
new file mode 100644
index 0000000000000..0e3d9313ed0e5
--- /dev/null
+++ b/llvm/test/DebugInfo/X86/template-alias.ll
@@ -0,0 +1,71 @@
+; RUN: llc %s -o - --filetype=obj | llvm-dwarfdump - --name A --show-children | FileCheck %s --check-prefix=TREE
+
+;; -ggnu-pubnames (nameTableKind: GNU).
+; RUN: llc %s -o - --filetype=obj \
+; RUN: | llvm-dwarfdump - --debug-gnu-pubtypes \
+; RUN: | FileCheck %s --check-prefix=GNU-TYPES
+
+;; -gpubnames (remove nameTableKind field from DICompileUnit).
+; RUN: sed 's/, nameTableKind: GNU//g' < %s \
+; RUN: | llc - -o - --filetype=obj \
+; RUN: | llvm-dwarfdump - --debug-pubtypes \
+; RUN: | FileCheck %s --check-prefix=PUB-TYPES
+
+;; C++ source from clang/test/CodeGenCXX/template-alias.cpp, compiled with -gsce:
+;; template<typename Y, int Z>
+;; struct X {
+;;   Y m1 = Z;
+;; };
+;;
+;; template<typename B, int C>
+;; using A = X<B, C>;
+;;
+;; A<int, 5> a;
+
+;; Test emission of DIDerivedType with tag: DW_TAG_template_alias.
+
+; TREE: DW_TAG_template_alias
+; TREE: DW_AT_type (0x{{[0-9a-f]+}} "X<int, 5>")
+; TREE: DW_AT_name ("A")
+; TREE:   DW_TAG_template_type_parameter
+; TREE:     DW_AT_type        (0x{{[0-9a-f]+}} "int")
+; TREE:     DW_AT_name        ("B")
+; TREE:   DW_TAG_template_value_parameter
+; TREE:     DW_AT_type        (0x{{[0-9a-f]+}} "int")
+; TREE:     DW_AT_name        ("C")
+; TREE:     DW_AT_const_value (5)
+; TREE:   NULL
+
+; GNU-TYPES: STATIC   TYPE     "A"
+; PUB-TYPES: "A"
+
+target triple = "x86_64-unknown-unkown"
+
+%struct.X = type { i32 }
+
+@a = global %struct.X { i32 5 }, align 4, !dbg !0
+
+!llvm.dbg.cu = !{!2}
+!llvm.module.flags = !{!17, !18}
+!llvm.ident = !{!19}
+
+!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
+!1 = distinct !DIGlobalVariable(name: "a", scope: !2, file: !5, line: 14, type: !6, isLocal: false, isDefinition: true)
+!2 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !3, producer: "clang version 19.0.0git", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, globals: !4, splitDebugInlining: false, nameTableKind: GNU)
+!3 = !DIFile(filename: "<stdin>", directory: "/")
+!4 = !{!0}
+!5 = !DIFile(filename: "clang/test/CodeGenCXX/template-alias.cpp", directory: "/")
+!6 = !DIDerivedType(tag: DW_TAG_template_alias, name: "A", file: !5, line: 12, baseType: !7, extraData: !14)
+!7 = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "X<int, 5>", file: !5, line: 7, size: 32, flags: DIFlagTypePassByValue | DIFlagNonTrivial, elements: !8, templateParams: !11, identifier: "_ZTS1XIiLi5EE")
+!8 = !{!9}
+!9 = !DIDerivedType(tag: DW_TAG_member, name: "m1", scope: !7, file: !5, line: 8, baseType: !10, size: 32)
+!10 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!11 = !{!12, !13}
+!12 = !DITemplateTypeParameter(name: "Y", type: !10)
+!13 = !DITemplateValueParameter(name: "Z", type: !10, value: i32 5)
+!14 = !{!15, !16}
+!15 = !DITemplateTypeParameter(name: "B", type: !10)
+!16 = !DITemplateValueParameter(name: "C", type: !10, value: i32 5)
+!17 = !{i32 2, !"Debug Info Version", i32 3}
+!18 = !{i32 1, !"wchar_size", i32 4}
+!19 = !{!"clang version 19.0.0git"}
diff --git a/llvm/test/tools/llvm-symbolizer/frame-types.s b/llvm/test/tools/llvm-symbolizer/frame-types.s
index f511c83bc0d25..7f38b1767a617 100644
--- a/llvm/test/tools/llvm-symbolizer/frame-types.s
+++ b/llvm/test/tools/llvm-symbolizer/frame-types.s
@@ -5,52 +5,56 @@
 
 // CHECK: f
 // CHECK-NEXT: a
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:4
-// CHECK-NEXT: -1 1 ??
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:11
+// CHECK-NEXT: -5 1 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: b
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:5
-// CHECK-NEXT: -8 4 ??
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:12
+// CHECK-NEXT: -12 4 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: c
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:6
-// CHECK-NEXT: -12 4 ??
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:13
+// CHECK-NEXT: -16 4 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: d
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:7
-// CHECK-NEXT: -16 4 ??
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:14
+// CHECK-NEXT: -20 4 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: e
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:8
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:15
 // CHECK-NEXT: -32 8 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: f
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:9
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:16
 // CHECK-NEXT: -36 4 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: g
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:10
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:17
 // CHECK-NEXT: -37 1 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: h
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:11
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:18
 // CHECK-NEXT: -38 1 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: i
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:12
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:19
 // CHECK-NEXT: -44 4 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: j
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:14
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:21
 // CHECK-NEXT: -45 1 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: k
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:15
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:22
 // CHECK-NEXT: -57 12 ??
 // CHECK-NEXT: f
 // CHECK-NEXT: l
-// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:16
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:23
 // CHECK-NEXT: -345 288 ??
+// CHECK-NEXT: f
+// CHECK-NEXT: m
+// CHECK-NEXT: /tmp{{/|\\}}frame-types.cpp:24
+// CHECK-NEXT: -352 4 ??
 
 // Generated from:
 //
@@ -76,14 +80,13 @@
 
 	.text
 	.file	"frame-types.cpp"
-	.globl	_Z1fv                   # -- Begin function _Z1fv
+	.globl	_Z1fv                           # -- Begin function _Z1fv
 	.p2align	4, 0x90
 	.type	_Z1fv,@function
 _Z1fv:                                  # @_Z1fv
 .Lfunc_begin0:
-	.file	1 "/tmp" "frame-types.cpp"
-	.loc	1 3 0                   # frame-types.cpp:3:0
-	.cfi_sections .debug_frame
+	.file	0 "/tmp" "frame-types.cpp"
+	.loc	0 10 0                          # frame-types.cpp:10:0
 	.cfi_startproc
 # %bb.0:                                # %entry
 	pushl	%ebp
@@ -91,441 +94,579 @@ _Z1fv:                                  # @_Z1fv
 	.cfi_offset %ebp, -8
 	movl	%esp, %ebp
 	.cfi_def_cfa_register %ebp
-	subl	$352, %esp              # imm = 0x160
 .Ltmp0:
-	.loc	1 6 9 prologue_end      # frame-types.cpp:6:9
-	leal	-1(%ebp), %eax
+	pushl	%ebx
+	subl	$372, %esp                      # imm = 0x174
+	.cfi_offset %ebx, -12
+	.loc	0 13 9 prologue_end             # frame-types.cpp:13:9
+	calll	.L0$pb
+.L0$pb:
+	popl	%ebx
 .Ltmp1:
-	#DEBUG_VALUE: f:a <- [$eax+0]
-	movl	%eax, -12(%ebp)
-	.loc	1 7 14                  # frame-types.cpp:7:14
-	movb	$1, -17(%ebp)
-	.loc	1 7 10 is_stmt 0        # frame-types.cpp:7:10
-	leal	-17(%ebp), %eax
-.Ltmp2:
+	addl	$_GLOBAL_OFFSET_TABLE_+(.Ltmp1-.L0$pb), %ebx
+	leal	-5(%ebp), %eax
 	movl	%eax, -16(%ebp)
-	.loc	1 10 14 is_stmt 1       # frame-types.cpp:10:14
+	.loc	0 14 14                         # frame-types.cpp:14:14
+	movb	$1, -21(%ebp)
+	.loc	0 14 10 is_stmt 0               # frame-types.cpp:14:10
+	leal	-21(%ebp), %eax
+	movl	%eax, -20(%ebp)
+	.loc	0 17 14 is_stmt 1               # frame-types.cpp:17:14
 	movb	$2, -37(%ebp)
-	.loc	1 17 1                  # frame-types.cpp:17:1
-	addl	$352, %esp              # imm = 0x160
+	.loc	0 24 14                         # frame-types.cpp:24:14
+	leal	-352(%ebp), %eax
+	xorl	%ecx, %ecx
+	movl	%eax, (%esp)
+	movl	$0, 4(%esp)
+	movl	$4, 8(%esp)
+	calll	memset@PLT
+	.loc	0 25 1 epilogue_begin           # frame-types.cpp:25:1
+	addl	$372, %esp                      # imm = 0x174
+	popl	%ebx
 	popl	%ebp
 	.cfi_def_cfa %esp, 4
 	retl
-.Ltmp3:
+.Ltmp2:
 .Lfunc_end0:
 	.size	_Z1fv, .Lfunc_end0-_Z1fv
 	.cfi_endproc
                                         # -- End function
-	.section	.debug_str,"MS",@progbits,1
-.Linfo_string0:
-	.asciz	"clang version 9.0.0 "  # string offset=0
-.Linfo_string1:
-	.asciz	"frame-types.cpp"       # string offset=21
-.Linfo_string2:
-	.asciz	"/tmp"                  # string offset=37
-.Linfo_string3:
-	.asciz	"_Z1fv"                 # string offset=42
-.Linfo_string4:
-	.asciz	"f"                     # string offset=48
-.Linfo_string5:
-	.asciz	"a"                     # string offset=50
-.Linfo_string6:
-	.asciz	"char"                  # string offset=52
-.Linfo_string7:
-	.asciz	"b"                     # string offset=57
-.Linfo_string8:
-	.asciz	"c"                     # string offset=59
-.Linfo_string9:
-	.asciz	"d"                     # string offset=61
-.Linfo_string10:
-	.asciz	"e"                     # string offset=63
-.Linfo_string11:
-	.asciz	"S"                     # string offset=65
-.Linfo_string12:
-	.asciz	"g"                     # string offset=67
-.Linfo_string13:
-	.asciz	"h"                     # string offset=69
-.Linfo_string14:
-	.asciz	"i"                     # string offset=71
-.Linfo_string15:
-	.asciz	"j"                     # string offset=73
-.Linfo_string16:
-	.asciz	"char_typedef"          # string offset=75
-.Linfo_string17:
-	.asciz	"k"                     # string offset=88
-.Linfo_string18:
-	.asciz	"__ARRAY_SIZE_TYPE__"   # string offset=90
-.Linfo_string19:
-	.asciz	"l"                     # string offset=110
 	.section	.debug_abbrev,"",@progbits
-	.byte	1                       # Abbreviation Code
-	.byte	17                      # DW_TAG_compile_unit
-	.byte	1                       # DW_CHILDREN_yes
-	.byte	37                      # DW_AT_producer
-	.byte	14                      # DW_FORM_strp
-	.byte	19                      # DW_AT_language
-	.byte	5                       # DW_FORM_data2
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	16                      # DW_AT_stmt_list
-	.byte	23                      # DW_FORM_sec_offset
-	.byte	27                      # DW_AT_comp_dir
-	.byte	14                      # DW_FORM_strp
-	.byte	17                      # DW_AT_low_pc
-	.byte	1                       # DW_FORM_addr
-	.byte	18                      # DW_AT_high_pc
-	.byte	6                       # DW_FORM_data4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	2                       # Abbreviation Code
-	.byte	46                      # DW_TAG_subprogram
-	.byte	1                       # DW_CHILDREN_yes
-	.byte	17                      # DW_AT_low_pc
-	.byte	1                       # DW_FORM_addr
-	.byte	18                      # DW_AT_high_pc
-	.byte	6                       # DW_FORM_data4
-	.byte	64                      # DW_AT_frame_base
-	.byte	24                      # DW_FORM_exprloc
-	.byte	110                     # DW_AT_linkage_name
-	.byte	14                      # DW_FORM_strp
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	58                      # DW_AT_decl_file
-	.byte	11                      # DW_FORM_data1
-	.byte	59                      # DW_AT_decl_line
-	.byte	11                      # DW_FORM_data1
-	.byte	63                      # DW_AT_external
-	.byte	25                      # DW_FORM_flag_present
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	3                       # Abbreviation Code
-	.byte	52                      # DW_TAG_variable
-	.byte	0                       # DW_CHILDREN_no
-	.byte	2                       # DW_AT_location
-	.byte	24                      # DW_FORM_exprloc
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	58                      # DW_AT_decl_file
-	.byte	11                      # DW_FORM_data1
-	.byte	59                      # DW_AT_decl_line
-	.byte	11                      # DW_FORM_data1
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	4                       # Abbreviation Code
-	.byte	22                      # DW_TAG_typedef
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	58                      # DW_AT_decl_file
-	.byte	11                      # DW_FORM_data1
-	.byte	59                      # DW_AT_decl_line
-	.byte	11                      # DW_FORM_data1
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	5                       # Abbreviation Code
-	.byte	36                      # DW_TAG_base_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	62                      # DW_AT_encoding
-	.byte	11                      # DW_FORM_data1
-	.byte	11                      # DW_AT_byte_size
-	.byte	11                      # DW_FORM_data1
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	6                       # Abbreviation Code
-	.byte	15                      # DW_TAG_pointer_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	7                       # Abbreviation Code
-	.byte	16                      # DW_TAG_reference_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	8                       # Abbreviation Code
-	.byte	66                      # DW_TAG_rvalue_reference_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	9                       # Abbreviation Code
-	.byte	31                      # DW_TAG_ptr_to_member_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	29                      # DW_AT_containing_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	10                      # Abbreviation Code
-	.byte	21                      # DW_TAG_subroutine_type
-	.byte	1                       # DW_CHILDREN_yes
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	11                      # Abbreviation Code
-	.byte	5                       # DW_TAG_formal_parameter
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	52                      # DW_AT_artificial
-	.byte	25                      # DW_FORM_flag_present
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	12                      # Abbreviation Code
-	.byte	19                      # DW_TAG_structure_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	60                      # DW_AT_declaration
-	.byte	25                      # DW_FORM_flag_present
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	13                      # Abbreviation Code
-	.byte	38                      # DW_TAG_const_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	14                      # Abbreviation Code
-	.byte	53                      # DW_TAG_volatile_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	15                      # Abbreviation Code
-	.byte	55                      # DW_TAG_restrict_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	16                      # Abbreviation Code
-	.byte	1                       # DW_TAG_array_type
-	.byte	1                       # DW_CHILDREN_yes
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	17                      # Abbreviation Code
-	.byte	33                      # DW_TAG_subrange_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	73                      # DW_AT_type
-	.byte	19                      # DW_FORM_ref4
-	.byte	55                      # DW_AT_count
-	.byte	11                      # DW_FORM_data1
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	18                      # Abbreviation Code
-	.byte	36                      # DW_TAG_base_type
-	.byte	0                       # DW_CHILDREN_no
-	.byte	3                       # DW_AT_name
-	.byte	14                      # DW_FORM_strp
-	.byte	11                      # DW_AT_byte_size
-	.byte	11                      # DW_FORM_data1
-	.byte	62                      # DW_AT_encoding
-	.byte	11                      # DW_FORM_data1
-	.byte	0                       # EOM(1)
-	.byte	0                       # EOM(2)
-	.byte	0                       # EOM(3)
+	.byte	1                               # Abbreviation Code
+	.byte	17                              # DW_TAG_compile_unit
+	.byte	1                               # DW_CHILDREN_yes
+	.byte	37                              # DW_AT_producer
+	.byte	37                              # DW_FORM_strx1
+	.byte	19                              # DW_AT_language
+	.byte	5                               # DW_FORM_data2
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	114                             # DW_AT_str_offsets_base
+	.byte	23                              # DW_FORM_sec_offset
+	.byte	16                              # DW_AT_stmt_list
+	.byte	23                              # DW_FORM_sec_offset
+	.byte	27                              # DW_AT_comp_dir
+	.byte	37                              # DW_FORM_strx1
+	.byte	17                              # DW_AT_low_pc
+	.byte	27                              # DW_FORM_addrx
+	.byte	18                              # DW_AT_high_pc
+	.byte	6                               # DW_FORM_data4
+	.byte	115                             # DW_AT_addr_base
+	.byte	23                              # DW_FORM_sec_offset
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	2                               # Abbreviation Code
+	.byte	46                              # DW_TAG_subprogram
+	.byte	1                               # DW_CHILDREN_yes
+	.byte	17                              # DW_AT_low_pc
+	.byte	27                              # DW_FORM_addrx
+	.byte	18                              # DW_AT_high_pc
+	.byte	6                               # DW_FORM_data4
+	.byte	64                              # DW_AT_frame_base
+	.byte	24                              # DW_FORM_exprloc
+	.byte	110                             # DW_AT_linkage_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	58                              # DW_AT_decl_file
+	.byte	11                              # DW_FORM_data1
+	.byte	59                              # DW_AT_decl_line
+	.byte	11                              # DW_FORM_data1
+	.byte	63                              # DW_AT_external
+	.byte	25                              # DW_FORM_flag_present
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	3                               # Abbreviation Code
+	.byte	52                              # DW_TAG_variable
+	.byte	0                               # DW_CHILDREN_no
+	.byte	2                               # DW_AT_location
+	.byte	24                              # DW_FORM_exprloc
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	58                              # DW_AT_decl_file
+	.byte	11                              # DW_FORM_data1
+	.byte	59                              # DW_AT_decl_line
+	.byte	11                              # DW_FORM_data1
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	4                               # Abbreviation Code
+	.byte	22                              # DW_TAG_typedef
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	58                              # DW_AT_decl_file
+	.byte	11                              # DW_FORM_data1
+	.byte	59                              # DW_AT_decl_line
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	5                               # Abbreviation Code
+	.byte	36                              # DW_TAG_base_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	62                              # DW_AT_encoding
+	.byte	11                              # DW_FORM_data1
+	.byte	11                              # DW_AT_byte_size
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	6                               # Abbreviation Code
+	.byte	15                              # DW_TAG_pointer_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	7                               # Abbreviation Code
+	.byte	16                              # DW_TAG_reference_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	8                               # Abbreviation Code
+	.byte	66                              # DW_TAG_rvalue_reference_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	9                               # Abbreviation Code
+	.byte	31                              # DW_TAG_ptr_to_member_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	29                              # DW_AT_containing_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	10                              # Abbreviation Code
+	.byte	21                              # DW_TAG_subroutine_type
+	.byte	1                               # DW_CHILDREN_yes
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	11                              # Abbreviation Code
+	.byte	5                               # DW_TAG_formal_parameter
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	52                              # DW_AT_artificial
+	.byte	25                              # DW_FORM_flag_present
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	12                              # Abbreviation Code
+	.byte	19                              # DW_TAG_structure_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	60                              # DW_AT_declaration
+	.byte	25                              # DW_FORM_flag_present
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	13                              # Abbreviation Code
+	.byte	38                              # DW_TAG_const_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	14                              # Abbreviation Code
+	.byte	53                              # DW_TAG_volatile_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	15                              # Abbreviation Code
+	.byte	55                              # DW_TAG_restrict_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	16                              # Abbreviation Code
+	.byte	1                               # DW_TAG_array_type
+	.byte	1                               # DW_CHILDREN_yes
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	17                              # Abbreviation Code
+	.byte	33                              # DW_TAG_subrange_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	55                              # DW_AT_count
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	18                              # Abbreviation Code
+	.byte	36                              # DW_TAG_base_type
+	.byte	0                               # DW_CHILDREN_no
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	11                              # DW_AT_byte_size
+	.byte	11                              # DW_FORM_data1
+	.byte	62                              # DW_AT_encoding
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	19                              # Abbreviation Code
+	.byte	19                              # DW_TAG_structure_type
+	.byte	1                               # DW_CHILDREN_yes
+	.byte	54                              # DW_AT_calling_convention
+	.byte	11                              # DW_FORM_data1
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	11                              # DW_AT_byte_size
+	.byte	11                              # DW_FORM_data1
+	.byte	58                              # DW_AT_decl_file
+	.byte	11                              # DW_FORM_data1
+	.byte	59                              # DW_AT_decl_line
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	20                              # Abbreviation Code
+	.byte	47                              # DW_TAG_template_type_parameter
+	.byte	0                               # DW_CHILDREN_no
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	21                              # Abbreviation Code
+	.byte	13                              # DW_TAG_member
+	.byte	0                               # DW_CHILDREN_no
+	.byte	3                               # DW_AT_name
+	.byte	37                              # DW_FORM_strx1
+	.byte	73                              # DW_AT_type
+	.byte	19                              # DW_FORM_ref4
+	.byte	58                              # DW_AT_decl_file
+	.byte	11                              # DW_FORM_data1
+	.byte	59                              # DW_AT_decl_line
+	.byte	11                              # DW_FORM_data1
+	.byte	56                              # DW_AT_data_member_location
+	.byte	11                              # DW_FORM_data1
+	.byte	0                               # EOM(1)
+	.byte	0                               # EOM(2)
+	.byte	0                               # EOM(3)
 	.section	.debug_info,"",@progbits
 .Lcu_begin0:
 	.long	.Ldebug_info_end0-.Ldebug_info_start0 # Length of Unit
 .Ldebug_info_start0:
-	.short	4                       # DWARF version number
-	.long	.debug_abbrev           # Offset Into Abbrev. Section
-	.byte	4                       # Address Size (in bytes)
-	.byte	1                       # Abbrev [1] 0xb:0x157 DW_TAG_compile_unit
-	.long	.Linfo_string0          # DW_AT_producer
-	.short	4                       # DW_AT_language
-	.long	.Linfo_string1          # DW_AT_name
-	.long	.Lline_table_start0     # DW_AT_stmt_list
-	.long	.Linfo_string2          # DW_AT_comp_dir
-	.long	.Lfunc_begin0           # DW_AT_low_pc
-	.long	.Lfunc_end0-.Lfunc_begin0 # DW_AT_high_pc
-	.byte	2                       # Abbrev [2] 0x26:0xca DW_TAG_subprogram
-	.long	.Lfunc_begin0           # DW_AT_low_pc
-	.long	.Lfunc_end0-.Lfunc_begin0 # DW_AT_high_pc
-	.byte	1                       # DW_AT_frame_base
+	.short	5                               # DWARF version number
+	.byte	1                               # DWARF Unit Type
+	.byte	4                               # Address Size (in bytes)
+	.long	.debug_abbrev                   # Offset Into Abbrev. Section
+	.byte	1                               # Abbrev [1] 0xc:0x148 DW_TAG_compile_unit
+	.byte	0                               # DW_AT_producer
+	.short	26                              # DW_AT_language
+	.byte	1                               # DW_AT_name
+	.long	.Lstr_offsets_base0             # DW_AT_str_offsets_base
+	.long	.Lline_table_start0             # DW_AT_stmt_list
+	.byte	2                               # DW_AT_comp_dir
+	.byte	0                               # DW_AT_low_pc
+	.long	.Lfunc_end0-.Lfunc_begin0       # DW_AT_high_pc
+	.long	.Laddr_table_base0              # DW_AT_addr_base
+	.byte	2                               # Abbrev [2] 0x23:0xa6 DW_TAG_subprogram
+	.byte	0                               # DW_AT_low_pc
+	.long	.Lfunc_end0-.Lfunc_begin0       # DW_AT_high_pc
+	.byte	1                               # DW_AT_frame_base
 	.byte	85
-	.long	.Linfo_string3          # DW_AT_linkage_name
-	.long	.Linfo_string4          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	3                       # DW_AT_decl_line
+	.byte	3                               # DW_AT_linkage_name
+	.byte	4                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	10                              # DW_AT_decl_line
                                         # DW_AT_external
-	.byte	3                       # Abbrev [3] 0x3b:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	3                               # Abbrev [3] 0x2f:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
-	.byte	127
-	.long	.Linfo_string5          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	4                       # DW_AT_decl_line
-	.long	240                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x49:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
-	.byte	145
-	.byte	120
-	.long	.Linfo_string7          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	5                       # DW_AT_decl_line
-	.long	247                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x57:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	123
+	.byte	5                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	11                              # DW_AT_decl_line
+	.long	201                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x3a:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	116
-	.long	.Linfo_string8          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	6                       # DW_AT_decl_line
-	.long	252                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x65:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	7                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	12                              # DW_AT_decl_line
+	.long	205                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x45:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	112
-	.long	.Linfo_string9          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	7                       # DW_AT_decl_line
-	.long	257                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x73:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	8                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	13                              # DW_AT_decl_line
+	.long	210                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x50:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
+	.byte	145
+	.byte	108
+	.byte	9                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	14                              # DW_AT_decl_line
+	.long	215                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x5b:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	96
-	.long	.Linfo_string10         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	8                       # DW_AT_decl_line
-	.long	262                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x81:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	10                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	15                              # DW_AT_decl_line
+	.long	220                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x66:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	92
-	.long	.Linfo_string4          # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	9                       # DW_AT_decl_line
-	.long	292                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x8f:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	4                               # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	16                              # DW_AT_decl_line
+	.long	247                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x71:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	91
-	.long	.Linfo_string12         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	10                      # DW_AT_decl_line
-	.long	301                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0x9d:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	12                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	17                              # DW_AT_decl_line
+	.long	256                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x7c:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	90
-	.long	.Linfo_string13         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	11                      # DW_AT_decl_line
-	.long	306                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0xab:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	13                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	18                              # DW_AT_decl_line
+	.long	261                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x87:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	84
-	.long	.Linfo_string14         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	12                      # DW_AT_decl_line
-	.long	311                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0xb9:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	14                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	19                              # DW_AT_decl_line
+	.long	266                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x92:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	83
-	.long	.Linfo_string15         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	14                      # DW_AT_decl_line
-	.long	228                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0xc7:0xe DW_TAG_variable
-	.byte	2                       # DW_AT_location
+	.byte	15                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	21                              # DW_AT_decl_line
+	.long	192                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0x9d:0xb DW_TAG_variable
+	.byte	2                               # DW_AT_location
 	.byte	145
 	.byte	71
-	.long	.Linfo_string17         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	15                      # DW_AT_decl_line
-	.long	316                     # DW_AT_type
-	.byte	3                       # Abbrev [3] 0xd5:0xf DW_TAG_variable
-	.byte	3                       # DW_AT_location
+	.byte	17                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	22                              # DW_AT_decl_line
+	.long	271                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0xa8:0xc DW_TAG_variable
+	.byte	3                               # DW_AT_location
 	.byte	145
 	.ascii	"\247}"
-	.long	.Linfo_string19         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	16                      # DW_AT_decl_line
-	.long	335                     # DW_AT_type
-	.byte	4                       # Abbrev [4] 0xe4:0xb DW_TAG_typedef
-	.long	240                     # DW_AT_type
-	.long	.Linfo_string16         # DW_AT_name
-	.byte	1                       # DW_AT_decl_file
-	.byte	13                      # DW_AT_decl_line
-	.byte	0                       # End Of Children Mark
-	.byte	5                       # Abbrev [5] 0xf0:0x7 DW_TAG_base_type
-	.long	.Linfo_string6          # DW_AT_name
-	.byte	6                       # DW_AT_encoding
-	.byte	1                       # DW_AT_byte_size
-	.byte	6                       # Abbrev [6] 0xf7:0x5 DW_TAG_pointer_type
-	.long	240                     # DW_AT_type
-	.byte	7                       # Abbrev [7] 0xfc:0x5 DW_TAG_reference_type
-	.long	240                     # DW_AT_type
-	.byte	8                       # Abbrev [8] 0x101:0x5 DW_TAG_rvalue_reference_type
-	.long	240                     # DW_AT_type
-	.byte	9                       # Abbrev [9] 0x106:0x9 DW_TAG_ptr_to_member_type
-	.long	271                     # DW_AT_type
-	.long	287                     # DW_AT_containing_type
-	.byte	10                      # Abbrev [10] 0x10f:0xb DW_TAG_subroutine_type
-	.long	240                     # DW_AT_type
-	.byte	11                      # Abbrev [11] 0x114:0x5 DW_TAG_formal_parameter
-	.long	282                     # DW_AT_type
+	.byte	19                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	23                              # DW_AT_decl_line
+	.long	287                             # DW_AT_type
+	.byte	3                               # Abbrev [3] 0xb4:0xc DW_TAG_variable
+	.byte	3                               # DW_AT_location
+	.byte	145
+	.ascii	"\240}"
+	.byte	20                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	24                              # DW_AT_decl_line
+	.long	305                             # DW_AT_type
+	.byte	4                               # Abbrev [4] 0xc0:0x8 DW_TAG_typedef
+	.long	201                             # DW_AT_type
+	.byte	16                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	20                              # DW_AT_decl_line
+	.byte	0                               # End Of Children Mark
+	.byte	5                               # Abbrev [5] 0xc9:0x4 DW_TAG_base_type
+	.byte	6                               # DW_AT_name
+	.byte	6                               # DW_AT_encoding
+	.byte	1                               # DW_AT_byte_size
+	.byte	6                               # Abbrev [6] 0xcd:0x5 DW_TAG_pointer_type
+	.long	201                             # DW_AT_type
+	.byte	7                               # Abbrev [7] 0xd2:0x5 DW_TAG_reference_type
+	.long	201                             # DW_AT_type
+	.byte	8                               # Abbrev [8] 0xd7:0x5 DW_TAG_rvalue_reference_type
+	.long	201                             # DW_AT_type
+	.byte	9                               # Abbrev [9] 0xdc:0x9 DW_TAG_ptr_to_member_type
+	.long	229                             # DW_AT_type
+	.long	245                             # DW_AT_containing_type
+	.byte	10                              # Abbrev [10] 0xe5:0xb DW_TAG_subroutine_type
+	.long	201                             # DW_AT_type
+	.byte	11                              # Abbrev [11] 0xea:0x5 DW_TAG_formal_parameter
+	.long	240                             # DW_AT_type
                                         # DW_AT_artificial
-	.byte	0                       # End Of Children Mark
-	.byte	6                       # Abbrev [6] 0x11a:0x5 DW_TAG_pointer_type
-	.long	287                     # DW_AT_type
-	.byte	12                      # Abbrev [12] 0x11f:0x5 DW_TAG_structure_type
-	.long	.Linfo_string11         # DW_AT_name
+	.byte	0                               # End Of Children Mark
+	.byte	6                               # Abbrev [6] 0xf0:0x5 DW_TAG_pointer_type
+	.long	245                             # DW_AT_type
+	.byte	12                              # Abbrev [12] 0xf5:0x2 DW_TAG_structure_type
+	.byte	11                              # DW_AT_name
                                         # DW_AT_declaration
-	.byte	9                       # Abbrev [9] 0x124:0x9 DW_TAG_ptr_to_member_type
-	.long	240                     # DW_AT_type
-	.long	287                     # DW_AT_containing_type
-	.byte	13                      # Abbrev [13] 0x12d:0x5 DW_TAG_const_type
-	.long	240                     # DW_AT_type
-	.byte	14                      # Abbrev [14] 0x132:0x5 DW_TAG_volatile_type
-	.long	240                     # DW_AT_type
-	.byte	15                      # Abbrev [15] 0x137:0x5 DW_TAG_restrict_type
-	.long	247                     # DW_AT_type
-	.byte	16                      # Abbrev [16] 0x13c:0xc DW_TAG_array_type
-	.long	240                     # DW_AT_type
-	.byte	17                      # Abbrev [17] 0x141:0x6 DW_TAG_subrange_type
-	.long	328                     # DW_AT_type
-	.byte	12                      # DW_AT_count
-	.byte	0                       # End Of Children Mark
-	.byte	18                      # Abbrev [18] 0x148:0x7 DW_TAG_base_type
-	.long	.Linfo_string18         # DW_AT_name
-	.byte	8                       # DW_AT_byte_size
-	.byte	7                       # DW_AT_encoding
-	.byte	16                      # Abbrev [16] 0x14f:0x12 DW_TAG_array_type
-	.long	240                     # DW_AT_type
-	.byte	17                      # Abbrev [17] 0x154:0x6 DW_TAG_subrange_type
-	.long	328                     # DW_AT_type
-	.byte	12                      # DW_AT_count
-	.byte	17                      # Abbrev [17] 0x15a:0x6 DW_TAG_subrange_type
-	.long	328                     # DW_AT_type
-	.byte	24                      # DW_AT_count
-	.byte	0                       # End Of Children Mark
-	.byte	0                       # End Of Children Mark
+	.byte	9                               # Abbrev [9] 0xf7:0x9 DW_TAG_ptr_to_member_type
+	.long	201                             # DW_AT_type
+	.long	245                             # DW_AT_containing_type
+	.byte	13                              # Abbrev [13] 0x100:0x5 DW_TAG_const_type
+	.long	201                             # DW_AT_type
+	.byte	14                              # Abbrev [14] 0x105:0x5 DW_TAG_volatile_type
+	.long	201                             # DW_AT_type
+	.byte	15                              # Abbrev [15] 0x10a:0x5 DW_TAG_restrict_type
+	.long	205                             # DW_AT_type
+	.byte	16                              # Abbrev [16] 0x10f:0xc DW_TAG_array_type
+	.long	201                             # DW_AT_type
+	.byte	17                              # Abbrev [17] 0x114:0x6 DW_TAG_subrange_type
+	.long	283                             # DW_AT_type
+	.byte	12                              # DW_AT_count
+	.byte	0                               # End Of Children Mark
+	.byte	18                              # Abbrev [18] 0x11b:0x4 DW_TAG_base_type
+	.byte	18                              # DW_AT_name
+	.byte	8                               # DW_AT_byte_size
+	.byte	7                               # DW_AT_encoding
+	.byte	16                              # Abbrev [16] 0x11f:0x12 DW_TAG_array_type
+	.long	201                             # DW_AT_type
+	.byte	17                              # Abbrev [17] 0x124:0x6 DW_TAG_subrange_type
+	.long	283                             # DW_AT_type
+	.byte	12                              # DW_AT_count
+	.byte	17                              # Abbrev [17] 0x12a:0x6 DW_TAG_subrange_type
+	.long	283                             # DW_AT_type
+	.byte	24                              # DW_AT_count
+	.byte	0                               # End Of Children Mark
+	.byte	4                               # Abbrev [4] 0x131:0x8 DW_TAG_typedef
+	.long	313                             # DW_AT_type
+	.byte	24                              # DW_AT_name
+	.byte	0                               # DW_AT_decl_file
+	.byte	8                               # DW_AT_decl_line
+	.byte	19                              # Abbrev [19] 0x139:0x16 DW_TAG_structure_type
+	.byte	5                               # DW_AT_calling_convention
+	.byte	23                              # DW_AT_name
+	.byte	4                               # DW_AT_byte_size
+	.byte	0                               # DW_AT_decl_file
+	.byte	4                               # DW_AT_decl_line
+	.byte	20                              # Abbrev [20] 0x13f:0x6 DW_TAG_template_type_parameter
+	.long	335                             # DW_AT_type
+	.byte	22                              # DW_AT_name
+	.byte	21                              # Abbrev [21] 0x145:0x9 DW_TAG_member
+	.byte	20                              # DW_AT_name
+	.long	335                             # DW_AT_type
+	.byte	0                               # DW_AT_decl_file
+	.byte	5                               # DW_AT_decl_line
+	.byte	0                               # DW_AT_data_member_location
+	.byte	0                               # End Of Children Mark
+	.byte	5                               # Abbrev [5] 0x14f:0x4 DW_TAG_base_type
+	.byte	21                              # DW_AT_name
+	.byte	5                               # DW_AT_encoding
+	.byte	4                               # DW_AT_byte_size
+	.byte	0                               # End Of Children Mark
 .Ldebug_info_end0:
-	.section	.debug_macinfo,"",@progbits
-	.byte	0                       # End Of Macro List Mark
-
-	.ident	"clang version 9.0.0 "
+	.section	.debug_str_offsets,"",@progbits
+	.long	104                             # Length of String Offsets Set
+	.short	5
+	.short	0
+.Lstr_offsets_base0:
+	.section	.debug_str,"MS",@progbits,1
+.Linfo_string0:
+	.asciz	"clang version 19.0.0git" # string offset=0
+.Linfo_string1:
+	.asciz	"frame-types.cpp"               # string offset=107
+.Linfo_string2:
+	.asciz	"/tmp"                          # string offset=123
+.Linfo_string3:
+	.asciz	"_Z1fv"                         # string offset=128
+.Linfo_string4:
+	.asciz	"f"                             # string offset=134
+.Linfo_string5:
+	.asciz	"a"                             # string offset=136
+.Linfo_string6:
+	.asciz	"char"                          # string offset=138
+.Linfo_string7:
+	.asciz	"b"                             # string offset=143
+.Linfo_string8:
+	.asciz	"c"                             # string offset=145
+.Linfo_string9:
+	.asciz	"d"                             # string offset=147
+.Linfo_string10:
+	.asciz	"e"                             # string offset=149
+.Linfo_string11:
+	.asciz	"S"                             # string offset=151
+.Linfo_string12:
+	.asciz	"g"                             # string offset=153
+.Linfo_string13:
+	.asciz	"h"                             # string offset=155
+.Linfo_string14:
+	.asciz	"i"                             # string offset=157
+.Linfo_string15:
+	.asciz	"j"                             # string offset=159
+.Linfo_string16:
+	.asciz	"char_typedef"                  # string offset=161
+.Linfo_string17:
+	.asciz	"k"                             # string offset=174
+.Linfo_string18:
+	.asciz	"__ARRAY_SIZE_TYPE__"           # string offset=176
+.Linfo_string19:
+	.asciz	"l"                             # string offset=196
+.Linfo_string20:
+	.asciz	"m"                             # string offset=198
+.Linfo_string21:
+	.asciz	"int"                           # string offset=200
+.Linfo_string22:
+	.asciz	"Y"                             # string offset=204
+.Linfo_string23:
+	.asciz	"Base<int>"                     # string offset=206
+.Linfo_string24:
+	.asciz	"Alias<int>"                    # string offset=216
+	.section	.debug_str_offsets,"",@progbits
+	.long	.Linfo_string0
+	.long	.Linfo_string1
+	.long	.Linfo_string2
+	.long	.Linfo_string3
+	.long	.Linfo_string4
+	.long	.Linfo_string5
+	.long	.Linfo_string6
+	.long	.Linfo_string7
+	.long	.Linfo_string8
+	.long	.Linfo_string9
+	.long	.Linfo_string10
+	.long	.Linfo_string11
+	.long	.Linfo_string12
+	.long	.Linfo_string13
+	.long	.Linfo_string14
+	.long	.Linfo_string15
+	.long	.Linfo_string16
+	.long	.Linfo_string17
+	.long	.Linfo_string18
+	.long	.Linfo_string19
+	.long	.Linfo_string20
+	.long	.Linfo_string21
+	.long	.Linfo_string22
+	.long	.Linfo_string23
+	.long	.Linfo_string24
+	.section	.debug_addr,"",@progbits
+	.long	.Ldebug_addr_end0-.Ldebug_addr_start0 # Length of contribution
+.Ldebug_addr_start0:
+	.short	5                               # DWARF version number
+	.byte	4                               # Address size
+	.byte	0                               # Segment selector size
+.Laddr_table_base0:
+	.long	.Lfunc_begin0
+.Ldebug_addr_end0:
+	.ident	"clang version 19.0.0git"
 	.section	".note.GNU-stack","",@progbits
 	.addrsig
 	.section	.debug_line,"",@progbits

From ac39fa740b067f6197dca1caecc97c0da91ebf3d Mon Sep 17 00:00:00 2001
From: Christian Ulmann <christianulmann@gmail.com>
Date: Thu, 18 Apr 2024 13:09:16 +0200
Subject: [PATCH 861/886] [MLIR][Mem2Reg][LLVM] Enhance partial load support
 (#89094)

This commit improves LLVM dialect's Mem2Reg interfaces to support
promotions of partial loads from larger memory slots. To support this,
the Mem2Reg interface methods are extended with additional data layout
parameters. The data layout is required to determine type sizes to
produce correct conversion sequences.

Note: There will be additional followups that introduce a similar
functionality for stores, and there are plans to support accesses into
the middle of memory slots.
---
 .../mlir/Interfaces/MemorySlotInterfaces.td   |   6 +-
 mlir/lib/Dialect/LLVMIR/IR/LLVMMemorySlot.cpp | 173 ++++++++++++++----
 .../Dialect/MemRef/IR/MemRefMemorySlot.cpp    |  14 +-
 mlir/lib/Transforms/Mem2Reg.cpp               |  17 +-
 mlir/test/Dialect/LLVMIR/mem2reg.mlir         | 138 ++++++++++++--
 5 files changed, 277 insertions(+), 71 deletions(-)

diff --git a/mlir/include/mlir/Interfaces/MemorySlotInterfaces.td b/mlir/include/mlir/Interfaces/MemorySlotInterfaces.td
index 9db89361c7800..8c642c0ed26ac 100644
--- a/mlir/include/mlir/Interfaces/MemorySlotInterfaces.td
+++ b/mlir/include/mlir/Interfaces/MemorySlotInterfaces.td
@@ -127,7 +127,8 @@ def PromotableMemOpInterface : OpInterface<"PromotableMemOpInterface"> {
       }],
       "::mlir::Value", "getStored",
       (ins "const ::mlir::MemorySlot &":$slot,
-           "::mlir::RewriterBase &":$rewriter)
+           "::mlir::RewriterBase &":$rewriter,
+           "const ::mlir::DataLayout &":$dataLayout)
     >,
     InterfaceMethod<[{
         Checks that this operation can be promoted to no longer use the provided
@@ -172,7 +173,8 @@ def PromotableMemOpInterface : OpInterface<"PromotableMemOpInterface"> {
       (ins "const ::mlir::MemorySlot &":$slot,
            "const ::llvm::SmallPtrSetImpl<mlir::OpOperand *> &":$blockingUses,
            "::mlir::RewriterBase &":$rewriter,
-           "::mlir::Value":$reachingDefinition)
+           "::mlir::Value":$reachingDefinition,
+           "const ::mlir::DataLayout &":$dataLayout)
     >,
   ];
 }
diff --git a/mlir/lib/Dialect/LLVMIR/IR/LLVMMemorySlot.cpp b/mlir/lib/Dialect/LLVMIR/IR/LLVMMemorySlot.cpp
index c7ca0b4a5843a..93901477b5820 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/LLVMMemorySlot.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/LLVMMemorySlot.cpp
@@ -112,7 +112,8 @@ bool LLVM::LoadOp::loadsFrom(const MemorySlot &slot) {
 
 bool LLVM::LoadOp::storesTo(const MemorySlot &slot) { return false; }
 
-Value LLVM::LoadOp::getStored(const MemorySlot &slot, RewriterBase &rewriter) {
+Value LLVM::LoadOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                              const DataLayout &dataLayout) {
   llvm_unreachable("getStored should not be called on LoadOp");
 }
 
@@ -122,37 +123,121 @@ bool LLVM::StoreOp::storesTo(const MemorySlot &slot) {
   return getAddr() == slot.ptr;
 }
 
-/// Checks that two types are the same or can be cast into one another.
-static bool areCastCompatible(const DataLayout &layout, Type lhs, Type rhs) {
-  return lhs == rhs || (!isa<LLVM::LLVMStructType, LLVM::LLVMArrayType>(lhs) &&
-                        !isa<LLVM::LLVMStructType, LLVM::LLVMArrayType>(rhs) &&
-                        layout.getTypeSize(lhs) == layout.getTypeSize(rhs));
+/// Checks if `type` can be used in any kind of conversion sequences.
+static bool isSupportedTypeForConversion(Type type) {
+  // Aggregate types are not bitcastable.
+  if (isa<LLVM::LLVMStructType, LLVM::LLVMArrayType>(type))
+    return false;
+
+  // LLVM vector types are only used for either pointers or target specific
+  // types. These types cannot be casted in the general case, thus the memory
+  // optimizations do not support them.
+  if (isa<LLVM::LLVMFixedVectorType, LLVM::LLVMScalableVectorType>(type))
+    return false;
+
+  // Scalable types are not supported.
+  if (auto vectorType = dyn_cast<VectorType>(type))
+    return !vectorType.isScalable();
+  return true;
 }
 
+/// Checks that `rhs` can be converted to `lhs` by a sequence of casts and
+/// truncations.
+static bool areConversionCompatible(const DataLayout &layout, Type targetType,
+                                    Type srcType) {
+  if (targetType == srcType)
+    return true;
+
+  if (!isSupportedTypeForConversion(targetType) ||
+      !isSupportedTypeForConversion(srcType))
+    return false;
+
+  // Pointer casts will only be sane when the bitsize of both pointer types is
+  // the same.
+  if (isa<LLVM::LLVMPointerType>(targetType) &&
+      isa<LLVM::LLVMPointerType>(srcType))
+    return layout.getTypeSize(targetType) == layout.getTypeSize(srcType);
+
+  return layout.getTypeSize(targetType) <= layout.getTypeSize(srcType);
+}
+
+/// Checks if `dataLayout` describes a little endian layout.
+static bool isBigEndian(const DataLayout &dataLayout) {
+  auto endiannessStr = dyn_cast_or_null<StringAttr>(dataLayout.getEndianness());
+  return endiannessStr && endiannessStr == "big";
+}
+
+/// The size of a byte in bits.
+constexpr const static uint64_t kBitsInByte = 8;
+
 /// Constructs operations that convert `inputValue` into a new value of type
 /// `targetType`. Assumes that this conversion is possible.
 static Value createConversionSequence(RewriterBase &rewriter, Location loc,
-                                      Value inputValue, Type targetType) {
-  if (inputValue.getType() == targetType)
-    return inputValue;
-
-  if (!isa<LLVM::LLVMPointerType>(targetType) &&
-      !isa<LLVM::LLVMPointerType>(inputValue.getType()))
-    return rewriter.createOrFold<LLVM::BitcastOp>(loc, targetType, inputValue);
+                                      Value srcValue, Type targetType,
+                                      const DataLayout &dataLayout) {
+  // Get the types of the source and target values.
+  Type srcType = srcValue.getType();
+  assert(areConversionCompatible(dataLayout, targetType, srcType) &&
+         "expected that the compatibility was checked before");
+
+  uint64_t srcTypeSize = dataLayout.getTypeSize(srcType);
+  uint64_t targetTypeSize = dataLayout.getTypeSize(targetType);
+
+  // Nothing has to be done if the types are already the same.
+  if (srcType == targetType)
+    return srcValue;
+
+  // In the special case of casting one pointer to another, we want to generate
+  // an address space cast. Bitcasts of pointers are not allowed and using
+  // pointer to integer conversions are not equivalent due to the loss of
+  // provenance.
+  if (isa<LLVM::LLVMPointerType>(targetType) &&
+      isa<LLVM::LLVMPointerType>(srcType))
+    return rewriter.createOrFold<LLVM::AddrSpaceCastOp>(loc, targetType,
+                                                        srcValue);
+
+  IntegerType valueSizeInteger =
+      rewriter.getIntegerType(srcTypeSize * kBitsInByte);
+  Value replacement = srcValue;
+
+  // First, cast the value to a same-sized integer type.
+  if (isa<LLVM::LLVMPointerType>(srcType))
+    replacement = rewriter.createOrFold<LLVM::PtrToIntOp>(loc, valueSizeInteger,
+                                                          replacement);
+  else if (replacement.getType() != valueSizeInteger)
+    replacement = rewriter.createOrFold<LLVM::BitcastOp>(loc, valueSizeInteger,
+                                                         replacement);
+
+  // Truncate the integer if the size of the target is less than the value.
+  if (targetTypeSize != srcTypeSize) {
+    if (isBigEndian(dataLayout)) {
+      uint64_t shiftAmount = (srcTypeSize - targetTypeSize) * kBitsInByte;
+      auto shiftConstant = rewriter.create<LLVM::ConstantOp>(
+          loc, rewriter.getIntegerAttr(srcType, shiftAmount));
+      replacement =
+          rewriter.createOrFold<LLVM::LShrOp>(loc, srcValue, shiftConstant);
+    }
 
-  if (!isa<LLVM::LLVMPointerType>(targetType))
-    return rewriter.createOrFold<LLVM::PtrToIntOp>(loc, targetType, inputValue);
+    replacement = rewriter.create<LLVM::TruncOp>(
+        loc, rewriter.getIntegerType(targetTypeSize * kBitsInByte),
+        replacement);
+  }
 
-  if (!isa<LLVM::LLVMPointerType>(inputValue.getType()))
-    return rewriter.createOrFold<LLVM::IntToPtrOp>(loc, targetType, inputValue);
+  // Now cast the integer to the actual target type if required.
+  if (isa<LLVM::LLVMPointerType>(targetType))
+    replacement =
+        rewriter.createOrFold<LLVM::IntToPtrOp>(loc, targetType, replacement);
+  else if (replacement.getType() != targetType)
+    replacement =
+        rewriter.createOrFold<LLVM::BitcastOp>(loc, targetType, replacement);
 
-  return rewriter.createOrFold<LLVM::AddrSpaceCastOp>(loc, targetType,
-                                                      inputValue);
+  return replacement;
 }
 
-Value LLVM::StoreOp::getStored(const MemorySlot &slot, RewriterBase &rewriter) {
-  return createConversionSequence(rewriter, getLoc(), getValue(),
-                                  slot.elemType);
+Value LLVM::StoreOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                               const DataLayout &dataLayout) {
+  return createConversionSequence(rewriter, getLoc(), getValue(), slot.elemType,
+                                  dataLayout);
 }
 
 bool LLVM::LoadOp::canUsesBeRemoved(
@@ -167,17 +252,20 @@ bool LLVM::LoadOp::canUsesBeRemoved(
   // be removed (provided it loads the exact stored value and is not
   // volatile).
   return blockingUse == slot.ptr && getAddr() == slot.ptr &&
-         areCastCompatible(dataLayout, getResult().getType(), slot.elemType) &&
+         areConversionCompatible(dataLayout, getResult().getType(),
+                                 slot.elemType) &&
          !getVolatile_();
 }
 
 DeletionKind LLVM::LoadOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   // `canUsesBeRemoved` checked this blocking use must be the loaded slot
   // pointer.
-  Value newResult = createConversionSequence(
-      rewriter, getLoc(), reachingDefinition, getResult().getType());
+  Value newResult =
+      createConversionSequence(rewriter, getLoc(), reachingDefinition,
+                               getResult().getType(), dataLayout);
   rewriter.replaceAllUsesWith(getResult(), newResult);
   return DeletionKind::Delete;
 }
@@ -194,13 +282,15 @@ bool LLVM::StoreOp::canUsesBeRemoved(
   // store OF the slot pointer, only INTO the slot pointer.
   return blockingUse == slot.ptr && getAddr() == slot.ptr &&
          getValue() != slot.ptr &&
-         areCastCompatible(dataLayout, slot.elemType, getValue().getType()) &&
+         areConversionCompatible(dataLayout, slot.elemType,
+                                 getValue().getType()) &&
          !getVolatile_();
 }
 
 DeletionKind LLVM::StoreOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return DeletionKind::Delete;
 }
 
@@ -747,8 +837,8 @@ bool LLVM::MemsetOp::storesTo(const MemorySlot &slot) {
   return getDst() == slot.ptr;
 }
 
-Value LLVM::MemsetOp::getStored(const MemorySlot &slot,
-                                RewriterBase &rewriter) {
+Value LLVM::MemsetOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                                const DataLayout &dataLayout) {
   // TODO: Support non-integer types.
   return TypeSwitch<Type, Value>(slot.elemType)
       .Case([&](IntegerType intType) -> Value {
@@ -802,7 +892,8 @@ bool LLVM::MemsetOp::canUsesBeRemoved(
 
 DeletionKind LLVM::MemsetOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return DeletionKind::Delete;
 }
 
@@ -1059,8 +1150,8 @@ bool LLVM::MemcpyOp::storesTo(const MemorySlot &slot) {
   return memcpyStoresTo(*this, slot);
 }
 
-Value LLVM::MemcpyOp::getStored(const MemorySlot &slot,
-                                RewriterBase &rewriter) {
+Value LLVM::MemcpyOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                                const DataLayout &dataLayout) {
   return memcpyGetStored(*this, slot, rewriter);
 }
 
@@ -1074,7 +1165,8 @@ bool LLVM::MemcpyOp::canUsesBeRemoved(
 
 DeletionKind LLVM::MemcpyOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return memcpyRemoveBlockingUses(*this, slot, blockingUses, rewriter,
                                   reachingDefinition);
 }
@@ -1109,7 +1201,8 @@ bool LLVM::MemcpyInlineOp::storesTo(const MemorySlot &slot) {
 }
 
 Value LLVM::MemcpyInlineOp::getStored(const MemorySlot &slot,
-                                      RewriterBase &rewriter) {
+                                      RewriterBase &rewriter,
+                                      const DataLayout &dataLayout) {
   return memcpyGetStored(*this, slot, rewriter);
 }
 
@@ -1123,7 +1216,8 @@ bool LLVM::MemcpyInlineOp::canUsesBeRemoved(
 
 DeletionKind LLVM::MemcpyInlineOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return memcpyRemoveBlockingUses(*this, slot, blockingUses, rewriter,
                                   reachingDefinition);
 }
@@ -1159,8 +1253,8 @@ bool LLVM::MemmoveOp::storesTo(const MemorySlot &slot) {
   return memcpyStoresTo(*this, slot);
 }
 
-Value LLVM::MemmoveOp::getStored(const MemorySlot &slot,
-                                 RewriterBase &rewriter) {
+Value LLVM::MemmoveOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                                 const DataLayout &dataLayout) {
   return memcpyGetStored(*this, slot, rewriter);
 }
 
@@ -1174,7 +1268,8 @@ bool LLVM::MemmoveOp::canUsesBeRemoved(
 
 DeletionKind LLVM::MemmoveOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return memcpyRemoveBlockingUses(*this, slot, blockingUses, rewriter,
                                   reachingDefinition);
 }
diff --git a/mlir/lib/Dialect/MemRef/IR/MemRefMemorySlot.cpp b/mlir/lib/Dialect/MemRef/IR/MemRefMemorySlot.cpp
index 6c5250d527ade..ebbf20f1b76b6 100644
--- a/mlir/lib/Dialect/MemRef/IR/MemRefMemorySlot.cpp
+++ b/mlir/lib/Dialect/MemRef/IR/MemRefMemorySlot.cpp
@@ -160,8 +160,8 @@ bool memref::LoadOp::loadsFrom(const MemorySlot &slot) {
 
 bool memref::LoadOp::storesTo(const MemorySlot &slot) { return false; }
 
-Value memref::LoadOp::getStored(const MemorySlot &slot,
-                                RewriterBase &rewriter) {
+Value memref::LoadOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                                const DataLayout &dataLayout) {
   llvm_unreachable("getStored should not be called on LoadOp");
 }
 
@@ -178,7 +178,8 @@ bool memref::LoadOp::canUsesBeRemoved(
 
 DeletionKind memref::LoadOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   // `canUsesBeRemoved` checked this blocking use must be the loaded slot
   // pointer.
   rewriter.replaceAllUsesWith(getResult(), reachingDefinition);
@@ -240,8 +241,8 @@ bool memref::StoreOp::storesTo(const MemorySlot &slot) {
   return getMemRef() == slot.ptr;
 }
 
-Value memref::StoreOp::getStored(const MemorySlot &slot,
-                                 RewriterBase &rewriter) {
+Value memref::StoreOp::getStored(const MemorySlot &slot, RewriterBase &rewriter,
+                                 const DataLayout &dataLayout) {
   return getValue();
 }
 
@@ -258,7 +259,8 @@ bool memref::StoreOp::canUsesBeRemoved(
 
 DeletionKind memref::StoreOp::removeBlockingUses(
     const MemorySlot &slot, const SmallPtrSetImpl<OpOperand *> &blockingUses,
-    RewriterBase &rewriter, Value reachingDefinition) {
+    RewriterBase &rewriter, Value reachingDefinition,
+    const DataLayout &dataLayout) {
   return DeletionKind::Delete;
 }
 
diff --git a/mlir/lib/Transforms/Mem2Reg.cpp b/mlir/lib/Transforms/Mem2Reg.cpp
index abe565ea862f8..1e620e46af84e 100644
--- a/mlir/lib/Transforms/Mem2Reg.cpp
+++ b/mlir/lib/Transforms/Mem2Reg.cpp
@@ -165,7 +165,7 @@ class MemorySlotPromoter {
 public:
   MemorySlotPromoter(MemorySlot slot, PromotableAllocationOpInterface allocator,
                      RewriterBase &rewriter, DominanceInfo &dominance,
-                     MemorySlotPromotionInfo info,
+                     const DataLayout &dataLayout, MemorySlotPromotionInfo info,
                      const Mem2RegStatistics &statistics);
 
   /// Actually promotes the slot by mutating IR. Promoting a slot DOES
@@ -204,6 +204,7 @@ class MemorySlotPromoter {
   DenseMap<PromotableMemOpInterface, Value> reachingDefs;
   DenseMap<PromotableMemOpInterface, Value> replacedValuesMap;
   DominanceInfo &dominance;
+  const DataLayout &dataLayout;
   MemorySlotPromotionInfo info;
   const Mem2RegStatistics &statistics;
 };
@@ -213,9 +214,11 @@ class MemorySlotPromoter {
 MemorySlotPromoter::MemorySlotPromoter(
     MemorySlot slot, PromotableAllocationOpInterface allocator,
     RewriterBase &rewriter, DominanceInfo &dominance,
-    MemorySlotPromotionInfo info, const Mem2RegStatistics &statistics)
+    const DataLayout &dataLayout, MemorySlotPromotionInfo info,
+    const Mem2RegStatistics &statistics)
     : slot(slot), allocator(allocator), rewriter(rewriter),
-      dominance(dominance), info(std::move(info)), statistics(statistics) {
+      dominance(dominance), dataLayout(dataLayout), info(std::move(info)),
+      statistics(statistics) {
 #ifndef NDEBUG
   auto isResultOrNewBlockArgument = [&]() {
     if (BlockArgument arg = dyn_cast<BlockArgument>(slot.ptr))
@@ -435,7 +438,7 @@ Value MemorySlotPromoter::computeReachingDefInBlock(Block *block,
 
       if (memOp.storesTo(slot)) {
         rewriter.setInsertionPointAfter(memOp);
-        Value stored = memOp.getStored(slot, rewriter);
+        Value stored = memOp.getStored(slot, rewriter, dataLayout);
         assert(stored && "a memory operation storing to a slot must provide a "
                          "new definition of the slot");
         reachingDef = stored;
@@ -568,8 +571,8 @@ void MemorySlotPromoter::removeBlockingUses() {
 
       rewriter.setInsertionPointAfter(toPromote);
       if (toPromoteMemOp.removeBlockingUses(
-              slot, info.userToBlockingUses[toPromote], rewriter,
-              reachingDef) == DeletionKind::Delete)
+              slot, info.userToBlockingUses[toPromote], rewriter, reachingDef,
+              dataLayout) == DeletionKind::Delete)
         toErase.push_back(toPromote);
       if (toPromoteMemOp.storesTo(slot))
         if (Value replacedValue = replacedValuesMap[toPromoteMemOp])
@@ -642,7 +645,7 @@ LogicalResult mlir::tryToPromoteMemorySlots(
       MemorySlotPromotionAnalyzer analyzer(slot, dominance, dataLayout);
       std::optional<MemorySlotPromotionInfo> info = analyzer.computeInfo();
       if (info) {
-        MemorySlotPromoter(slot, allocator, rewriter, dominance,
+        MemorySlotPromoter(slot, allocator, rewriter, dominance, dataLayout,
                            std::move(*info), statistics)
             .promoteSlot();
         promotedAny = true;
diff --git a/mlir/test/Dialect/LLVMIR/mem2reg.mlir b/mlir/test/Dialect/LLVMIR/mem2reg.mlir
index fa5d842302d0f..644d30f9f9f13 100644
--- a/mlir/test/Dialect/LLVMIR/mem2reg.mlir
+++ b/mlir/test/Dialect/LLVMIR/mem2reg.mlir
@@ -448,19 +448,6 @@ llvm.func @store_load_forward() -> i32 {
 
 // -----
 
-// CHECK-LABEL: llvm.func @store_load_wrong_type
-llvm.func @store_load_wrong_type() -> i16 {
-  %0 = llvm.mlir.constant(1 : i32) : i32
-  %1 = llvm.mlir.constant(0 : i32) : i32
-  // CHECK: = llvm.alloca
-  %2 = llvm.alloca %0 x i32 {alignment = 4 : i64} : (i32) -> !llvm.ptr
-  llvm.store %1, %2 {alignment = 4 : i64} : i32, !llvm.ptr
-  %3 = llvm.load %2 {alignment = 2 : i64} : !llvm.ptr -> i16
-  llvm.return %3 : i16
-}
-
-// -----
-
 // CHECK-LABEL: llvm.func @merge_point_cycle
 llvm.func @merge_point_cycle() {
   // CHECK: %[[UNDEF:.*]] = llvm.mlir.undef : i32
@@ -894,7 +881,7 @@ llvm.func @stores_with_different_type_sizes(%arg0: i64, %arg1: f32, %cond: i1) -
 // CHECK-LABEL: @load_smaller_int
 llvm.func @load_smaller_int() -> i16 {
   %0 = llvm.mlir.constant(1 : i32) : i32
-  // CHECK: llvm.alloca
+  // CHECK-NOT: llvm.alloca
   %1 = llvm.alloca %0 x i32 {alignment = 4 : i64} : (i32) -> !llvm.ptr
   %2 = llvm.load %1 {alignment = 4 : i64} : !llvm.ptr -> i16
   llvm.return %2 : i16
@@ -902,10 +889,10 @@ llvm.func @load_smaller_int() -> i16 {
 
 // -----
 
-// CHECK-LABEL: @load_different_type_smaller
-llvm.func @load_different_type_smaller() -> f32 {
+// CHECK-LABEL: @load_different_type_same_size
+llvm.func @load_different_type_same_size() -> f32 {
   %0 = llvm.mlir.constant(1 : i32) : i32
-  // CHECK: llvm.alloca
+  // CHECK-NOT: llvm.alloca
   %1 = llvm.alloca %0 x i64 {alignment = 8 : i64} : (i32) -> !llvm.ptr
   %2 = llvm.load %1 {alignment = 4 : i64} : !llvm.ptr -> f32
   llvm.return %2 : f32
@@ -942,4 +929,121 @@ module attributes { dlti.dl_spec = #dlti.dl_spec<
     %2 = llvm.load %1 {alignment = 4 : i64} : !llvm.ptr -> !llvm.ptr<2>
     llvm.return %2 : !llvm.ptr<2>
   }
+
+  // CHECK-LABEL: @load_ptr_addrspace_cast_different_size2
+  llvm.func @load_ptr_addrspace_cast_different_size2() -> !llvm.ptr<1> {
+    %0 = llvm.mlir.constant(1 : i32) : i32
+    // CHECK: llvm.alloca
+    %1 = llvm.alloca %0 x !llvm.ptr<2> {alignment = 4 : i64} : (i32) -> !llvm.ptr
+    %2 = llvm.load %1 {alignment = 4 : i64} : !llvm.ptr -> !llvm.ptr<1>
+    llvm.return %2 : !llvm.ptr<1>
+  }
+}
+
+// -----
+
+// CHECK-LABEL: @load_smaller_int_type
+llvm.func @load_smaller_int_type() -> i32 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK-NOT: llvm.alloca
+  %1 = llvm.alloca %0 x i64 : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> i32
+  // CHECK: %[[RES:.*]] = llvm.trunc %{{.*}} : i64 to i32
+  // CHECK: llvm.return %[[RES]] : i32
+  llvm.return %2 : i32
+}
+
+// -----
+
+module attributes { dlti.dl_spec = #dlti.dl_spec<
+  #dlti.dl_entry<"dlti.endianness", "big">
+>} {
+  // CHECK-LABEL: @load_smaller_int_type_big_endian
+  llvm.func @load_smaller_int_type_big_endian() -> i8 {
+    %0 = llvm.mlir.constant(1 : i32) : i32
+    // CHECK-NOT: llvm.alloca
+    %1 = llvm.alloca %0 x i64 : (i32) -> !llvm.ptr
+    %2 = llvm.load %1 : !llvm.ptr -> i8
+    // CHECK: %[[SHIFT_WIDTH:.*]] = llvm.mlir.constant(56 : i64) : i64
+    // CHECK: %[[SHIFT:.*]] = llvm.lshr %{{.*}}, %[[SHIFT_WIDTH]]
+    // CHECK: %[[RES:.*]] = llvm.trunc %[[SHIFT]] : i64 to i8
+    // CHECK: llvm.return %[[RES]] : i8
+    llvm.return %2 : i8
+  }
+}
+
+// -----
+
+// CHECK-LABEL: @load_different_type_smaller
+llvm.func @load_different_type_smaller() -> f32 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK-NOT: llvm.alloca
+  %1 = llvm.alloca %0 x i64 : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> f32
+  // CHECK: %[[TRUNC:.*]] = llvm.trunc %{{.*}} : i64 to i32
+  // CHECK: %[[RES:.*]] = llvm.bitcast %[[TRUNC]] : i32 to f32
+  // CHECK: llvm.return %[[RES]] : f32
+  llvm.return %2 : f32
+}
+
+// -----
+
+// CHECK-LABEL: @load_smaller_float_type
+llvm.func @load_smaller_float_type() -> f32 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK-NOT: llvm.alloca
+  %1 = llvm.alloca %0 x f64 : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> f32
+  // CHECK: %[[CAST:.*]] = llvm.bitcast %{{.*}} : f64 to i64
+  // CHECK: %[[TRUNC:.*]] = llvm.trunc %[[CAST]] : i64 to i32
+  // CHECK: %[[RES:.*]] = llvm.bitcast %[[TRUNC]] : i32 to f32
+  // CHECK: llvm.return %[[RES]] : f32
+  llvm.return %2 : f32
+}
+
+// -----
+
+// CHECK-LABEL: @load_first_vector_elem
+llvm.func @load_first_vector_elem() -> i16 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK-NOT: llvm.alloca
+  %1 = llvm.alloca %0 x vector<4xi16> : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> i16
+  // CHECK: %[[TRUNC:.*]] = llvm.bitcast %{{.*}} : vector<4xi16> to i64
+  // CHECK: %[[RES:.*]] = llvm.trunc %[[TRUNC]] : i64 to i16
+  // CHECK: llvm.return %[[RES]] : i16
+  llvm.return %2 : i16
+}
+
+// -----
+
+// CHECK-LABEL: @load_first_llvm_vector_elem
+llvm.func @load_first_llvm_vector_elem() -> i16 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK: llvm.alloca
+  %1 = llvm.alloca %0 x !llvm.vec<4 x ptr> : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> i16
+  llvm.return %2 : i16
+}
+
+// -----
+
+// CHECK-LABEL: @scalable_vector
+llvm.func @scalable_vector() -> i16 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK: llvm.alloca
+  %1 = llvm.alloca %0 x vector<[4]xi16> : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> i16
+  llvm.return %2 : i16
+}
+
+// -----
+
+// CHECK-LABEL: @scalable_llvm_vector
+llvm.func @scalable_llvm_vector() -> i16 {
+  %0 = llvm.mlir.constant(1 : i32) : i32
+  // CHECK: llvm.alloca
+  %1 = llvm.alloca %0 x !llvm.vec<? x 4 x ppc_fp128> : (i32) -> !llvm.ptr
+  %2 = llvm.load %1 : !llvm.ptr -> i16
+  llvm.return %2 : i16
 }

From c26e9bf8fafea656b64d811c01fc3a5f970829ef Mon Sep 17 00:00:00 2001
From: Matt Devereau <matthew.devereau@arm.com>
Date: Thu, 18 Apr 2024 11:20:54 +0000
Subject: [PATCH 862/886] Revert "Constant Fold Logf128 calls (#84501)"

This reverts commit e90bc9cfd4d22c89dd993f62ede700ae25df49c5.
---
 llvm/CMakeLists.txt                           |   2 -
 llvm/cmake/config-ix.cmake                    |  11 --
 llvm/include/llvm/ADT/APFloat.h               |  13 --
 llvm/include/llvm/ADT/APInt.h                 |   8 --
 llvm/include/llvm/Config/llvm-config.h.cmake  |   3 -
 llvm/include/llvm/Support/float128.h          |  21 ---
 llvm/lib/Analysis/CMakeLists.txt              |   6 -
 llvm/lib/Analysis/ConstantFolding.cpp         |  11 --
 llvm/lib/Support/APFloat.cpp                  |  24 ----
 llvm/test/CMakeLists.txt                      |   1 -
 .../InstSimplify/ConstProp/logf128.ll         | 126 ------------------
 llvm/test/lit.cfg.py                          |   3 -
 llvm/test/lit.site.cfg.py.in                  |   1 -
 llvm/unittests/Analysis/CMakeLists.txt        |   6 -
 14 files changed, 236 deletions(-)
 delete mode 100644 llvm/include/llvm/Support/float128.h
 delete mode 100644 llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll

diff --git a/llvm/CMakeLists.txt b/llvm/CMakeLists.txt
index c48ab438ffe0b..d511376e18ba5 100644
--- a/llvm/CMakeLists.txt
+++ b/llvm/CMakeLists.txt
@@ -562,8 +562,6 @@ set(LLVM_USE_STATIC_ZSTD FALSE CACHE BOOL "Use static version of zstd. Can be TR
 
 set(LLVM_ENABLE_CURL "OFF" CACHE STRING "Use libcurl for the HTTP client if available. Can be ON, OFF, or FORCE_ON")
 
-set(LLVM_HAS_LOGF128 "OFF" CACHE STRING "Use logf128 to constant fold fp128 logarithm calls. Can be ON, OFF, or FORCE_ON")
-
 set(LLVM_ENABLE_HTTPLIB "OFF" CACHE STRING "Use cpp-httplib HTTP server library if available. Can be ON, OFF, or FORCE_ON")
 
 set(LLVM_Z3_INSTALL_DIR "" CACHE STRING "Install directory of the Z3 solver.")
diff --git a/llvm/cmake/config-ix.cmake b/llvm/cmake/config-ix.cmake
index ce07cab7a56c6..bf1b110245bb2 100644
--- a/llvm/cmake/config-ix.cmake
+++ b/llvm/cmake/config-ix.cmake
@@ -257,17 +257,6 @@ else()
   set(LLVM_ENABLE_TERMINFO 0)
 endif()
 
-if(LLVM_HAS_LOGF128)
-  include(CheckCXXSymbolExists)
-  check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
-
-  if(LLVM_HAS_LOGF128 STREQUAL FORCE_ON AND NOT HAS_LOGF128)
-    message(FATAL_ERROR "Failed to configure logf128")
-  endif()
-
-  set(LLVM_HAS_LOGF128 "${HAS_LOGF128}")
-endif()
-
 # function checks
 check_symbol_exists(arc4random "stdlib.h" HAVE_DECL_ARC4RANDOM)
 find_package(Backtrace)
diff --git a/llvm/include/llvm/ADT/APFloat.h b/llvm/include/llvm/ADT/APFloat.h
index d5794b97bc267..deb74cb2fdeb1 100644
--- a/llvm/include/llvm/ADT/APFloat.h
+++ b/llvm/include/llvm/ADT/APFloat.h
@@ -19,7 +19,6 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/FloatingPointMode.h"
 #include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/float128.h"
 #include <memory>
 
 #define APFLOAT_DISPATCH_ON_SEMANTICS(METHOD_CALL)                             \
@@ -355,9 +354,6 @@ class IEEEFloat final : public APFloatBase {
   Expected<opStatus> convertFromString(StringRef, roundingMode);
   APInt bitcastToAPInt() const;
   double convertToDouble() const;
-#ifdef __FLOAT128__
-  float128 convertToQuad() const;
-#endif
   float convertToFloat() const;
 
   /// @}
@@ -1222,15 +1218,6 @@ class APFloat : public APFloatBase {
   /// shorter semantics, like IEEEsingle and others.
   double convertToDouble() const;
 
-  /// Converts this APFloat to host float value.
-  ///
-  /// \pre The APFloat must be built using semantics, that can be represented by
-  /// the host float type without loss of precision. It can be IEEEquad and
-  /// shorter semantics, like IEEEdouble and others.
-#ifdef __FLOAT128__
-  float128 convertToQuad() const;
-#endif
-
   /// Converts this APFloat to host float value.
   ///
   /// \pre The APFloat must be built using semantics, that can be represented by
diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index 0f6e4c592af65..8d3c029b2e7e9 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -17,7 +17,6 @@
 
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/MathExtras.h"
-#include "llvm/Support/float128.h"
 #include <cassert>
 #include <climits>
 #include <cstring>
@@ -1671,13 +1670,6 @@ class [[nodiscard]] APInt {
   /// any bit width. Exactly 64 bits will be translated.
   double bitsToDouble() const { return llvm::bit_cast<double>(getWord(0)); }
 
-#ifdef __FLOAT128__
-  float128 bitsToQuad() const {
-    __uint128_t ul = ((__uint128_t)U.pVal[1] << 64) + U.pVal[0];
-    return llvm::bit_cast<float128>(ul);
-  }
-#endif
-
   /// Converts APInt bits to a float
   ///
   /// The conversion does not do a translation from integer to float, it just
diff --git a/llvm/include/llvm/Config/llvm-config.h.cmake b/llvm/include/llvm/Config/llvm-config.h.cmake
index 629977cc11d68..6605ea60df99e 100644
--- a/llvm/include/llvm/Config/llvm-config.h.cmake
+++ b/llvm/include/llvm/Config/llvm-config.h.cmake
@@ -198,7 +198,4 @@
 /* Define if plugins enabled */
 #cmakedefine LLVM_ENABLE_PLUGINS
 
-/* Define if logf128 is available */
-#cmakedefine LLVM_HAS_LOGF128
-
 #endif
diff --git a/llvm/include/llvm/Support/float128.h b/llvm/include/llvm/Support/float128.h
deleted file mode 100644
index f7a442fce8dd3..0000000000000
--- a/llvm/include/llvm/Support/float128.h
+++ /dev/null
@@ -1,21 +0,0 @@
-//===-- llvm/Support/float128.h - Compiler abstraction support --*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_FLOAT128
-#define LLVM_FLOAT128
-
-namespace llvm {
-
-#if defined(__clang__) && defined(__FLOAT128__)
-typedef __float128 float128;
-#elif defined(__FLOAT128__) && (defined(__GNUC__) || defined(__GNUG__))
-typedef _Float128 float128;
-#endif
-
-} // namespace llvm
-#endif // LLVM_FLOAT128
diff --git a/llvm/lib/Analysis/CMakeLists.txt b/llvm/lib/Analysis/CMakeLists.txt
index 74476cb5440c6..474b8d20fde16 100644
--- a/llvm/lib/Analysis/CMakeLists.txt
+++ b/llvm/lib/Analysis/CMakeLists.txt
@@ -159,9 +159,3 @@ add_llvm_component_library(LLVMAnalysis
   Support
   TargetParser
   )
-
-include(CheckCXXSymbolExists)
-check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
-if(HAS_LOGF128)
- target_compile_definitions(LLVMAnalysis PRIVATE HAS_LOGF128)
-endif()
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index e292ff326a595..749374a3aa48a 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -2089,17 +2089,6 @@ static Constant *ConstantFoldScalarCall1(StringRef Name,
     if (IntrinsicID == Intrinsic::canonicalize)
       return constantFoldCanonicalize(Ty, Call, U);
 
-#if defined(__FLOAT128__) && defined(HAS_LOGF128)
-    if (Ty->isFP128Ty()) {
-      switch (IntrinsicID) {
-      default:
-        return nullptr;
-      case Intrinsic::log:
-        return ConstantFP::get(Ty, logf128(Op->getValueAPF().convertToQuad()));
-      }
-    }
-#endif
-
     if (!Ty->isHalfTy() && !Ty->isFloatTy() && !Ty->isDoubleTy())
       return nullptr;
 
diff --git a/llvm/lib/Support/APFloat.cpp b/llvm/lib/Support/APFloat.cpp
index 30c85f5b4214c..0a4f5ac01553f 100644
--- a/llvm/lib/Support/APFloat.cpp
+++ b/llvm/lib/Support/APFloat.cpp
@@ -3670,15 +3670,6 @@ double IEEEFloat::convertToDouble() const {
   return api.bitsToDouble();
 }
 
-#ifdef __FLOAT128__
-float128 IEEEFloat::convertToQuad() const {
-  assert(semantics == (const llvm::fltSemantics *)&semIEEEquad &&
-         "Float semantics are not IEEEquads");
-  APInt api = bitcastToAPInt();
-  return api.bitsToQuad();
-}
-#endif
-
 /// Integer bit is explicit in this format.  Intel hardware (387 and later)
 /// does not support these bit patterns:
 ///  exponent = all 1's, integer bit 0, significand 0 ("pseudoinfinity")
@@ -5274,21 +5265,6 @@ double APFloat::convertToDouble() const {
   return Temp.getIEEE().convertToDouble();
 }
 
-#ifdef __FLOAT128__
-float128 APFloat::convertToQuad() const {
-  if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEquad)
-    return getIEEE().convertToQuad();
-  assert(getSemantics().isRepresentableBy(semIEEEquad) &&
-         "Float semantics is not representable by IEEEquad");
-  APFloat Temp = *this;
-  bool LosesInfo;
-  opStatus St = Temp.convert(semIEEEquad, rmNearestTiesToEven, &LosesInfo);
-  assert(!(St & opInexact) && !LosesInfo && "Unexpected imprecision");
-  (void)St;
-  return Temp.getIEEE().convertToQuad();
-}
-#endif
-
 float APFloat::convertToFloat() const {
   if (&getSemantics() == (const llvm::fltSemantics *)&semIEEEsingle)
     return getIEEE().convertToFloat();
diff --git a/llvm/test/CMakeLists.txt b/llvm/test/CMakeLists.txt
index b419464aba3f5..6127b76db06b7 100644
--- a/llvm/test/CMakeLists.txt
+++ b/llvm/test/CMakeLists.txt
@@ -25,7 +25,6 @@ llvm_canonicalize_cmake_booleans(
   LLVM_INCLUDE_DXIL_TESTS
   LLVM_TOOL_LLVM_DRIVER_BUILD
   LLVM_INCLUDE_SPIRV_TOOLS_TESTS
-  LLVM_HAS_LOGF128
   )
 
 configure_lit_site_cfg(
diff --git a/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll b/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll
deleted file mode 100644
index da56997f69382..0000000000000
--- a/llvm/test/Transforms/InstSimplify/ConstProp/logf128.ll
+++ /dev/null
@@ -1,126 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt < %s -passes=instsimplify -S | FileCheck %s
-
-; REQUIRES: has_logf128
-declare fp128 @llvm.log.f128(fp128)
-
-define fp128 @log_e_64(){
-; CHECK-LABEL: define fp128 @log_e_64() {
-; CHECK-NEXT:    ret fp128 0xL300000000000000040010A2B23F3BAB7
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000004005000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_smallest_positive_subnormal_number(){
-; CHECK-LABEL: define fp128 @log_e_smallest_positive_subnormal_number() {
-; CHECK-NEXT:    ret fp128 0xL3000000000000000C00C654628220780
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000010000000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_largest_subnormal_number(){
-; CHECK-LABEL: define fp128 @log_e_largest_subnormal_number() {
-; CHECK-NEXT:    ret fp128 0xLD000000000000000C00C62D918CE2421
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF0000FFFFFFFFFFFF)
-  ret fp128 %A
-}
-
-define fp128 @log_e_smallest_positive_normal_number(){
-;
-; CHECK-LABEL: define fp128 @log_e_smallest_positive_normal_number() {
-; CHECK-NEXT:    ret fp128 0xLD000000000000000C00C62D918CE2421
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000000001000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_largest_normal_number(){
-; CHECK-LABEL: define fp128 @log_e_largest_normal_number() {
-; CHECK-NEXT:    ret fp128 0xLF000000000000000400C62E42FEFA39E
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF7FFEFFFFFFFFFFFF)
-  ret fp128 %A
-}
-
-define fp128 @log_e_largest_number_less_than_one(){
-; CHECK-LABEL: define fp128 @log_e_largest_number_less_than_one() {
-; CHECK-NEXT:    ret fp128 0xL0000000000000000BF8E000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xLFFFFFFFFFFFFFFFF3FFEFFFFFFFFFFFF)
-  ret fp128 %A
-}
-
-define fp128 @log_e_1(){
-; CHECK-LABEL: define fp128 @log_e_1() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000000000000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000003FFF000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_smallest_number_larger_than_one(){
-; CHECK-LABEL: define fp128 @log_e_smallest_number_larger_than_one() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000003F8F000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000013FFF000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_negative_2(){
-; CHECK-LABEL: define fp128 @log_e_negative_2() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL0000000000000000C000000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_0(){
-; CHECK-LABEL: define fp128 @log_e_0() {
-; CHECK-NEXT:    ret fp128 0xL0000000000000000FFFF000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000000000000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_negative_0(){
-; CHECK-LABEL: define fp128 @log_e_negative_0() {
-; CHECK-NEXT:    ret fp128 0xL0000000000000000FFFF000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000008000000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_infinity(){
-; CHECK-LABEL: define fp128 @log_e_infinity() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF000000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000007FFF000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_negative_infinity(){
-; CHECK-LABEL: define fp128 @log_e_negative_infinity() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000000
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL0000000000000000FFFF000000000000)
-  ret fp128 %A
-}
-
-define fp128 @log_e_nan(){
-; CHECK-LABEL: define fp128 @log_e_nan() {
-; CHECK-NEXT:    ret fp128 0xL00000000000000007FFF800000000001
-;
-  %A = call fp128 @llvm.log.f128(fp128 noundef 0xL00000000000000007FFF000000000001)
-  ret fp128 %A
-}
-
-define <2 x fp128> @log_e_negative_2_vector(){
-; CHECK-LABEL: define <2 x fp128> @log_e_negative_2_vector() {
-; CHECK-NEXT:    ret <2 x fp128> <fp128 0xL00000000000000007FFF800000000000, fp128 0xL00000000000000007FFF800000000000>
-;
-  %A = call <2 x fp128> @llvm.log.v2f128(<2 x fp128> <fp128 0xL0000000000000000C000000000000000, fp128 0xL0000000000000000C000000000000001>)
-  ret <2 x fp128> %A
-}
diff --git a/llvm/test/lit.cfg.py b/llvm/test/lit.cfg.py
index 36bcc48307dd8..4c05317036d1a 100644
--- a/llvm/test/lit.cfg.py
+++ b/llvm/test/lit.cfg.py
@@ -614,6 +614,3 @@ def have_ld64_plugin_support():
 # "OBJECT_MODE" to 'any' by default on AIX OS.
 if "system-aix" in config.available_features:
     config.environment["OBJECT_MODE"] = "any"
-
-if config.has_logf128:
-    config.available_features.add("has_logf128")
diff --git a/llvm/test/lit.site.cfg.py.in b/llvm/test/lit.site.cfg.py.in
index 27cb1dd6477d9..b6f255d472d16 100644
--- a/llvm/test/lit.site.cfg.py.in
+++ b/llvm/test/lit.site.cfg.py.in
@@ -61,7 +61,6 @@ config.reverse_iteration = @LLVM_ENABLE_REVERSE_ITERATION@
 config.dxil_tests = @LLVM_INCLUDE_DXIL_TESTS@
 config.have_llvm_driver = @LLVM_TOOL_LLVM_DRIVER_BUILD@
 config.spirv_tools_tests = @LLVM_INCLUDE_SPIRV_TOOLS_TESTS@
-config.has_logf128 = @LLVM_HAS_LOGF128@
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)
diff --git a/llvm/unittests/Analysis/CMakeLists.txt b/llvm/unittests/Analysis/CMakeLists.txt
index 9b3778f8a3f98..b1aeaa6e71fd4 100644
--- a/llvm/unittests/Analysis/CMakeLists.txt
+++ b/llvm/unittests/Analysis/CMakeLists.txt
@@ -80,11 +80,5 @@ if(NOT WIN32)
   export_executable_symbols_for_plugins(AnalysisTests)
 endif()
 
-include(CheckCXXSymbolExists)
-check_cxx_symbol_exists(logf128 math.h HAS_LOGF128)
-if(HAS_LOGF128)
-  target_compile_definitions(AnalysisTests PRIVATE HAS_LOGF128)
-endif()
-
 add_subdirectory(InlineAdvisorPlugin)
 add_subdirectory(InlineOrderPlugin)

From 9462abdff1687c1e35574cdef8de6c549addc42a Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@outlook.com>
Date: Thu, 18 Apr 2024 03:50:08 -0700
Subject: [PATCH 863/886] [SLP]Fix PR89187: fixx assertion check.

Need to use proper index variable to fix a crash.
---
 .../Transforms/Vectorize/SLPVectorizer.cpp    |  2 +-
 .../insertelement-across-zero.ll              | 71 +++++++++++++++++++
 2 files changed, 72 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/Transforms/SLPVectorizer/insertelement-across-zero.ll

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 7694627c3b043..075fbd3bf868f 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -13362,7 +13362,7 @@ Value *BoUpSLP::vectorizeTree(
                   do {
                     IEBase = cast<InsertElementInst>(Base);
                     int IEIdx = *getInsertIndex(IEBase);
-                    assert(Mask[Idx] == PoisonMaskElem &&
+                    assert(Mask[IEIdx] == PoisonMaskElem &&
                            "InsertElementInstruction used already.");
                     Mask[IEIdx] = IEIdx;
                     Base = IEBase->getOperand(0);
diff --git a/llvm/test/Transforms/SLPVectorizer/insertelement-across-zero.ll b/llvm/test/Transforms/SLPVectorizer/insertelement-across-zero.ll
new file mode 100644
index 0000000000000..1d54f59f2fdd8
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/insertelement-across-zero.ll
@@ -0,0 +1,71 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S --passes=slp-vectorizer < %s | FileCheck %s
+
+define void @test(i8 %0, i8 %1) {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: i8 [[TMP0:%.*]], i8 [[TMP1:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, ptr getelementptr (i8, ptr null, i32 8), align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <8 x i8> [[TMP2]], <8 x i8> poison, <16 x i32> <i32 7, i32 6, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 poison, i32 poison>
+; CHECK-NEXT:    [[LUPTO132421:%.*]] = shufflevector <16 x i8> zeroinitializer, <16 x i8> [[TMP3]], <16 x i32> <i32 16, i32 17, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 14, i32 15>
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <16 x i8> [[LUPTO132421]], i8 [[TMP0]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <16 x i8> [[TMP4]], i8 [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <16 x i8> [[TMP5]], i8 0, i32 7
+; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP2]], <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i8> [[TMP6]], <16 x i8> [[TMP7]], <16 x i32> <i32 16, i32 17, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 17, i32 poison, i32 poison, i32 poison, i32 poison, i32 14, i32 15>
+; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP8]], <16 x i8> poison, <16 x i32> <i32 0, i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 7, i32 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 0, i32 15>
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp ne <16 x i8> zeroinitializer, [[TMP9]]
+; CHECK-NEXT:    ret void
+;
+entry:
+  %li8 = load i8, ptr getelementptr (i8, ptr null, i32 8), align 1
+  %li9163 = load i8, ptr getelementptr (i8, ptr null, i32 9), align 1
+  %li10 = load i8, ptr getelementptr (i8, ptr null, i32 10), align 1
+  %li11164 = load i8, ptr getelementptr (i8, ptr null, i32 11), align 1
+  %li12 = load i8, ptr getelementptr (i8, ptr null, i32 12), align 1
+  %li13 = load i8, ptr getelementptr (i8, ptr null, i32 13), align 1
+  %li14 = load i8, ptr getelementptr (i8, ptr null, i32 14), align 1
+  %li15165 = load i8, ptr getelementptr (i8, ptr null, i32 15), align 1
+  %lupto8237 = insertelement <16 x i8> zeroinitializer, i8 %li8, i64 8
+  %lupto9238 = insertelement <16 x i8> %lupto8237, i8 %li9163, i64 9
+  %lupto10239 = insertelement <16 x i8> %lupto9238, i8 %li10, i64 10
+  %lupto11240 = insertelement <16 x i8> %lupto10239, i8 %li11164, i64 11
+  %lupto12241 = insertelement <16 x i8> %lupto11240, i8 %li12, i64 12
+  %lupto13242 = insertelement <16 x i8> %lupto12241, i8 %li13, i64 13
+  %lupto14243 = insertelement <16 x i8> %lupto13242, i8 %li14, i64 1
+  %l = insertelement <16 x i8> %lupto14243, i8 %li15165, i64 0
+  %li15 = extractelement <16 x i8> %l, i64 15
+  %2 = icmp ne i8 %0, 0
+  %3 = icmp ne i8 %1, 0
+  %4 = icmp ne i8 %0, 0
+  %.i3 = icmp ne i8 %0, 0
+  %.i4 = icmp ne i8 %0, 0
+  %5 = icmp ne i8 %0, 0
+  %6 = icmp ne i8 %0, 0
+  %7 = icmp ne i8 0, 0
+  %8 = icmp ne i8 %0, 0
+  %.i9 = icmp ne i8 %li9163, 0
+  %9 = icmp ne i8 %0, 0
+  %10 = icmp ne i8 %0, 0
+  %11 = icmp ne i8 %0, 0
+  %12 = icmp ne i8 %0, 0
+  %13 = icmp ne i8 %0, 0
+  %.i15 = icmp ne i8 %li15, 0
+  %i0244 = insertelement <16 x i1> zeroinitializer, i1 %2, i64 0
+  %i1245 = insertelement <16 x i1> %i0244, i1 %3, i64 1
+  %i2246 = insertelement <16 x i1> %i1245, i1 %4, i64 2
+  %i3247 = insertelement <16 x i1> %i2246, i1 %.i3, i64 3
+  %i4248 = insertelement <16 x i1> %i3247, i1 %.i4, i64 4
+  %i5249 = insertelement <16 x i1> %i4248, i1 %5, i64 5
+  %i6250 = insertelement <16 x i1> %i5249, i1 %6, i64 6
+  %i7251 = insertelement <16 x i1> %i6250, i1 %7, i64 7
+  %i8252 = insertelement <16 x i1> %i7251, i1 %8, i64 8
+  %i9253 = insertelement <16 x i1> %i8252, i1 %.i9, i64 9
+  %i10254 = insertelement <16 x i1> %i9253, i1 %9, i64 10
+  %i11255 = insertelement <16 x i1> %i10254, i1 %10, i64 11
+  %i12256 = insertelement <16 x i1> %i11255, i1 %11, i64 12
+  %i13257 = insertelement <16 x i1> %i12256, i1 %12, i64 13
+  %i14 = insertelement <16 x i1> %i13257, i1 %13, i64 14
+  %14 = insertelement <16 x i1> %i14, i1 %.i15, i64 15
+  ret void
+}

From 2a4e61b342e7a19bcb229ef16dee083c58224a3f Mon Sep 17 00:00:00 2001
From: Matthias Springer <me@m-sp.org>
Date: Thu, 18 Apr 2024 13:26:15 +0200
Subject: [PATCH 864/886] [mlir][NFC] Move and improve ownership-based buffer
 dellocation docs (#89196)

Move the documentation of the ownership-based buffer deallocation pass
to a separate file. Also improve the documentation a bit and insert a
figure that explains the `bufferization.dealloc` op (copied from the
tutorial at the LLVM Dev Summit 2023).
---
 mlir/docs/BufferDeallocationInternals.md      |   3 +
 mlir/docs/Bufferization.md                    | 655 +-----------------
 mlir/docs/OwnershipBasedBufferDeallocation.md | 619 +++++++++++++++++
 .../includes/img/bufferization_dealloc_op.svg |   1 +
 .../Bufferization/IR/BufferizationBase.td     |   4 +-
 5 files changed, 627 insertions(+), 655 deletions(-)
 create mode 100644 mlir/docs/OwnershipBasedBufferDeallocation.md
 create mode 100644 mlir/docs/includes/img/bufferization_dealloc_op.svg

diff --git a/mlir/docs/BufferDeallocationInternals.md b/mlir/docs/BufferDeallocationInternals.md
index 3078cfbf59338..00830ba9d2dc2 100644
--- a/mlir/docs/BufferDeallocationInternals.md
+++ b/mlir/docs/BufferDeallocationInternals.md
@@ -1,5 +1,8 @@
 # Buffer Deallocation - Internals
 
+**Note:** This pass is deprecated. Please use the ownership-based buffer
+deallocation pass instead.
+
 This section covers the internal functionality of the BufferDeallocation
 transformation. The transformation consists of several passes. The main pass
 called BufferDeallocation can be applied via “-buffer-deallocation” on MLIR
diff --git a/mlir/docs/Bufferization.md b/mlir/docs/Bufferization.md
index 3e7edeeabed19..8085358222125 100644
--- a/mlir/docs/Bufferization.md
+++ b/mlir/docs/Bufferization.md
@@ -110,7 +110,7 @@ is inserting an element inside a vector. Since SSA values are immutable, the
 operation returns a copy of the input vector with the element inserted.
 Another example in MLIR is `linalg.generic`, which always has an extra `outs`
 operand which provides the initial values to update (for example when the
-operation is doing a reduction). 
+operation is doing a reduction).
 
 This input is referred to as "destination" in the following (quotes are
 important as this operand isn't modified in place but copied) and comes into
@@ -240,657 +240,6 @@ Alternatively,
 skips the analysis and inserts a copy on every buffer write, just like the
 dialect conversion-based bufferization.
 
-## Buffer Deallocation
-
-**Important: this pass is deprecated, please use the ownership based buffer**
-**deallocation pass instead**
-
-One-Shot Bufferize deallocates all buffers that it allocates. This is in
-contrast to the dialect conversion-based bufferization that delegates this job
-to the
-[`-buffer-deallocation`](https://mlir.llvm.org/docs/Passes/#-buffer-deallocation-adds-all-required-dealloc-operations-for-all-allocations-in-the-input-program)
-pass. By default, One-Shot Bufferize rejects IR where a newly allocated buffer
-is returned from a block. Such IR will fail bufferization.
-
-A new buffer allocation is returned from a block when the result of an op that
-is not in destination-passing style is returned. E.g.:
-
-```mlir
-%0 = scf.if %c -> (tensor<?xf32>) {
-  %1 = tensor.generate ... -> tensor<?xf32>
-  scf.yield %1 : tensor<?xf32>
-} else {
-  scf.yield %another_tensor : tensor<?xf32>
-}
-```
-
-The `scf.yield` in the "else" branch is OK, but the `scf.yield` in the "then"
-branch will be rejected.
-
-Another case in which a buffer allocation may be returned is when a buffer copy
-must be inserted due to a RaW conflict. E.g.:
-
-```mlir
-%0 = scf.if %c -> (tensor<?xf32>) {
-  %1 = tensor.insert %cst into %another_tensor[%idx] : tensor<?xf32>
-  "my_dialect.reading_tensor_op"(%another_tensor) : (tensor<?xf32>) -> ()
-  ...
-  scf.yield %1 : tensor<?xf32>
-} else {
-  scf.yield %yet_another_tensor : tensor<?xf32>
-}
-```
-
-In the above example, a buffer copy of `buffer(%another_tensor)` (with `%cst`
-inserted) is yielded from the "then" branch.
-
-Note: Buffer allocations that are returned from a function are not deallocated.
-It is the caller's responsibility to deallocate the buffer. For the full
-function boundary ABI for MemRefs w.r.t. buffer deallocation refer to the
-[*Function Boundary ABI*](#function-boundary-abi) section. In the future, this
-could be automated with allocation hoisting (across function boundaries) or
-reference counting.
-
-One-Shot Bufferize leaks all memory and does not generate any buffer
-deallocations. The `-buffer-deallocation-pipeline` has to be run afterwards to
-insert the deallocation operations.
-
-## Ownership-based Buffer Deallocation
-
-Recommended compilation pipeline:
-```
-one-shot-bufferize
-       |          it's recommended to perform all bufferization here at latest,
-       |       <- any allocations inserted after this point have to be handled
-       V          manually
-expand-realloc
-       V
-ownership-based-buffer-deallocation
-       V
-  canonicalize <- mostly for scf.if simplifications
-       V
-buffer-deallocation-simplification
-       V       <- from this point onwards no tensor values are allowed
-lower-deallocations
-       V
-      CSE
-       V
-  canonicalize
-```
-
-One-Shot Bufferize does not deallocate any buffers that it allocates. This job
-is delegated to the
-[`-ownership-based-buffer-deallocation`](https://mlir.llvm.org/docs/Passes/#-ownership-based-buffer-deallocation)
-pass, i.e., after running One-Shot Bufferize, the result IR may have a number of
-`memref.alloc` ops, but no `memref.dealloc` ops. This pass processes operations
-implementing `FunctionOpInterface` one-by-one without analysing the call-graph.
-This means, that there have to be [some rules](#function-boundary-abi) on how
-MemRefs are handled when being passed from one function to another. The rest of
-the pass revolves heavily around the `bufferization.dealloc` operation which is
-inserted at the end of each basic block with appropriate operands and should be
-optimized using the Buffer Deallocation Simplification pass
-(`--buffer-deallocation-simplification`) and the regular canonicalizer
-(`--canonicalize`). Lowering the result of the
-`-ownership-based-buffer-deallocation` pass directly using
-`--convert-bufferization-to-memref` without beforehand optimization is not
-recommended as it will lead to very inefficient code (the runtime-cost of
-`bufferization.dealloc` is `O(|memrefs|^2+|memref|*|retained|)`).
-
-### Function boundary ABI
-
-The Buffer Deallocation pass operates on the level of operations implementing
-the `FunctionOpInterface`. Such operations can take MemRefs as arguments, but
-also return them. To ensure compatibility among all functions (including
-external ones), some rules have to be enforced:
-*   When a MemRef is passed as a function argument, ownership is never acquired.
-    It is always the caller's responsibility to deallocate such MemRefs.
-*   Returning a MemRef from a function always passes ownership to the caller,
-    i.e., it is also the caller's responsibility to deallocate memrefs returned
-    from a called function.
-*   A function must not return a MemRef with the same allocated base buffer as
-    one of its arguments (in this case a copy has to be created). Note that in
-    this context two subviews of the same buffer that don't overlap are also
-    considered to alias.
-
-For external functions (e.g., library functions written externally in C), the
-externally provided implementation has to adhere to these rules and they are
-just assumed by the buffer deallocation pass. Functions on which the
-deallocation pass is applied and the implementation is accessible are modified
-by the pass such that the ABI is respected (i.e., buffer copies are inserted as
-necessary).
-
-### Inserting `bufferization.dealloc` operations
-
-`bufferization.dealloc` operations are unconditionally inserted at the end of
-each basic block (just before the terminator). The majority of the pass is about
-finding the correct operands for this operation. There are three variadic
-operand lists to be populated, the first contains all MemRef values that may
-need to be deallocated, the second list contains their associated ownership
-values (of `i1` type), and the third list contains MemRef values that are still
-needed at a later point and should thus not be deallocated. This operation
-allows us to deal with any kind of aliasing behavior: it lowers to runtime
-aliasing checks when not enough information can be collected statically. When
-enough aliasing information is statically available, operands or the entire op
-may fold away.
-
-**Ownerships**
-
-To do so, we use a concept of ownership indicators of memrefs which materialize
-as an `i1` value for any SSA value of `memref` type, indicating whether the
-basic block in which it was materialized has ownership of this MemRef. Ideally,
-this is a constant `true` or `false`, but might also be a non-constant SSA
-value. To keep track of those ownership values without immediately materializing
-them (which might require insertion of `bufferization.clone` operations or
-operations checking for aliasing at runtime at positions where we don't actually
-need a materialized value), we use the `Ownership` class. This class represents
-the ownership in three states forming a lattice on a partial order:
-```
-forall X in SSA values. uninitialized < unique(X) < unknown
-forall X, Y in SSA values.
-  unique(X) == unique(Y) iff X and Y always evaluate to the same value
-  unique(X) != unique(Y) otherwise
-```
-Intuitively, the states have the following meaning:
-*   Uninitialized: the ownership is not initialized yet, this is the default
-    state; once an operation is finished processing the ownership of all
-    operation results with MemRef type should not be uninitialized anymore.
-*   Unique: there is a specific SSA value that can be queried to check ownership
-    without materializing any additional IR
-*   Unknown: no specific SSA value is available without materializing additional
-    IR, typically this is because two ownerships in 'Unique' state would have to
-    be merged manually (e.g., the result of an `arith.select` either has the
-    ownership of the then or else case depending on the condition value,
-    inserting another `arith.select` for the ownership values can perform the
-    merge and provide a 'Unique' ownership for the result), however, in the
-    general case this 'Unknown' state has to be assigned.
-
-Implied by the above partial order, the pass combines two ownerships in the
-following way:
-
-| Ownership 1   | Ownership 2   | Combined Ownership |
-|:--------------|:--------------|:-------------------|
-| uninitialized | uninitialized | uninitialized      |
-| unique(X)     | uninitialized | unique(X)          |
-| unique(X)     | unique(X)     | unique(X)          |
-| unique(X)     | unique(Y)     | unknown            |
-| unknown       | unique        | unknown            |
-| unknown       | uninitialized | unknown            |
-| <td colspan=3> + symmetric cases                   |
-
-**Collecting the list of MemRefs that potentially need to be deallocated**
-
-For a given block, the list of MemRefs that potentially need to be deallocated
-at the end of that block is computed by keeping track of all values for which
-the block potentially takes over ownership. This includes MemRefs provided as
-basic block arguments, interface handlers for operations like `memref.alloc` and
-`func.call`, but also liveness information in regions with multiple basic
-blocks.  More concretely, it is computed by taking the MemRefs in the 'in' set
-of the liveness analysis of the current basic block B, appended by the MemRef
-block arguments and by the set of MemRefs allocated in B itself (determined by
-the interface handlers), then subtracted (also determined by the interface
-handlers) by the set of MemRefs deallocated in B.
-
-Note that we don't have to take the intersection of the liveness 'in' set with
-the 'out' set of the predecessor block because a value that is in the 'in' set
-must be defined in an ancestor block that dominates all direct predecessors and
-thus the 'in' set of this block is a subset of the 'out' sets of each
-predecessor.
-
-```
-memrefs = filter((liveIn(block) U
-  allocated(block) U arguments(block)) \ deallocated(block), isMemRef)
-```
-
-The list of conditions for the second variadic operands list of
-`bufferization.dealloc` is computed by querying the stored ownership value for
-each of the MemRefs collected as described above. The ownership state is updated
-by the interface handlers while processing the basic block.
-
-**Collecting the list of MemRefs to retain**
-
-Given a basic block B, the list of MemRefs that have to be retained can be
-different for each successor block S.  For the two basic blocks B and S and the
-values passed via block arguments to the destination block S, we compute the
-list of MemRefs that have to be retained in B by taking the MemRefs in the
-successor operand list of the terminator and the MemRefs in the 'out' set of the
-liveness analysis for B intersected with the 'in' set of the destination block
-S.
-
-This list of retained values makes sure that we cannot run into use-after-free
-situations even if no aliasing information is present at compile-time.
-
-```
-toRetain = filter(successorOperands + (liveOut(fromBlock) insersect
-  liveIn(toBlock)), isMemRef)
-```
-
-### Supported interfaces
-
-The pass uses liveness analysis and a few interfaces:
-*   `FunctionOpInterface`
-*   `CallOpInterface`
-*   `MemoryEffectOpInterface`
-*   `RegionBranchOpInterface`
-*   `RegionBranchTerminatorOpInterface`
-
-Due to insufficient information provided by the interface, it also special-cases
-on the `cf.cond_br` operation and makes some assumptions about operations
-implementing the `RegionBranchOpInterface` at the moment, but improving the
-interfaces would allow us to remove those dependencies in the future.
-
-### Limitations
-
-The Buffer Deallocation pass has some requirements and limitations on the input
-IR. These are checked in the beginning of the pass and errors are emitted
-accordingly:
-*   The set of interfaces the pass operates on must be implemented (correctly).
-    E.g., if there is an operation present with a nested region, but does not
-    implement the `RegionBranchOpInterface`, an error is emitted because the
-    pass cannot know the semantics of the nested region (and does not make any
-    default assumptions on it).
-*   No explicit control-flow loops are present. Currently, only loops using
-    structural-control-flow are supported.  However, this limitation could be
-    lifted in the future.
-*   Deallocation operations should not be present already. The pass should
-    handle them correctly already (at least in most cases), but it's not
-    supported yet due to insufficient testing.
-*   Terminators must implement either `RegionBranchTerminatorOpInterface` or
-    `BranchOpInterface`, but not both. Terminators with more than one successor
-    are not supported (except `cf.cond_br`). This is not a fundamental
-    limitation, but there is no use-case justifying the more complex
-    implementation at the moment.
-
-### Example
-
-The following example contains a few interesting cases:
-*   Basic block arguments are modified to also pass along the ownership
-    indicator, but not for entry blocks, where the function boundary ABI
-    is applied instead.
-*   The result of `arith.select` initially has 'Unknown' assigned as ownership,
-    but once the `bufferization.dealloc` operation is inserted it is put in the
-    'retained' list (since it has uses in a later basic block) and thus the
-    'Unknown' ownership can be replaced with a 'Unique' ownership using the
-    corresponding result of the dealloc operation.
-*   The `cf.cond_br` operation has more than one successor and thus has to
-    insert two `bufferization.dealloc` operations (one for each successor).
-    While they have the same list of MemRefs to deallocate (because they perform
-    the deallocations for the same block), it must be taken into account that
-    some MemRefs remain *live* for one branch but not the other (thus set
-    intersection is performed on the *live-out* of the current block and the
-    *live-in* of the target block). Also, `cf.cond_br` supports separate
-    forwarding operands for each successor. To make sure that no MemRef is
-    deallocated twice (because there are two `bufferization.dealloc` operations
-    with the same MemRefs to deallocate), the condition operands are adjusted to
-    take the branch condition into account. While a generic lowering for such
-    terminator operations could be implemented, a specialized implementation can
-    take all the semantics of this particular operation into account and thus
-    generate a more efficient lowering.
-
-```mlir
-func.func @example(%memref: memref<?xi8>, %select_cond: i1, %br_cond: i1) {
-  %alloc = memref.alloc() : memref<?xi8>
-  %alloca = memref.alloca() : memref<?xi8>
-  %select = arith.select %select_cond, %alloc, %alloca : memref<?xi8>
-  cf.cond_br %br_cond, ^bb1(%alloc : memref<?xi8>), ^bb1(%memref : memref<?xi8>)
-^bb1(%bbarg: memref<?xi8>):
-  test.copy(%bbarg, %select) : (memref<?xi8>, memref<?xi8>)
-  return
-}
-```
-
-After running `--ownership-based-buffer-deallocation`, it looks as follows:
-
-```mlir
-// Function boundary ABI: ownership of `%memref` will never be acquired.
-func.func @example(%memref: memref<?xi8>, %select_cond: i1, %br_cond: i1) {
-  %false = arith.constant false
-  %true = arith.constant true
-
-  // The ownership of a MemRef defined by the `memref.alloc` operation is always
-  // assigned to be 'true'.
-  %alloc = memref.alloc() : memref<?xi8>
-
-  // The ownership of a MemRef defined by the `memref.alloca` operation is
-  // always assigned to be 'false'.
-  %alloca = memref.alloca() : memref<?xi8>
-
-  // The ownership of %select will be the join of the ownership of %alloc and
-  // the ownership of %alloca, i.e., of %true and %false. Because the pass does
-  // not know about the semantics of the `arith.select` operation (unless a
-  // custom handler is implemented), the ownership join will be 'Unknown'. If
-  // the materialized ownership indicator of %select is needed, either a clone
-  // has to be created for which %true is assigned as ownership or the result
-  // of a `bufferization.dealloc` where %select is in the retain list has to be
-  // used.
-  %select = arith.select %select_cond, %alloc, %alloca : memref<?xi8>
-
-  // We use `memref.extract_strided_metadata` to get the base memref since it is
-  // not allowed to pass arbitrary memrefs to `memref.dealloc`. This property is
-  // already enforced for `bufferization.dealloc`
-  %base_buffer_memref, ... = memref.extract_strided_metadata %memref
-    : memref<?xi8> -> memref<i8>, index, index, index
-  %base_buffer_alloc, ... = memref.extract_strided_metadata %alloc
-    : memref<?xi8> -> memref<i8>, index, index, index
-  %base_buffer_alloca, ... = memref.extract_strided_metadata %alloca
-    : memref<?xi8> -> memref<i8>, index, index, index
-
-  // The deallocation conditions need to be adjusted to incorporate the branch
-  // condition. In this example, this requires only a single negation, but might
-  // also require multiple arith.andi operations.
-  %not_br_cond = arith.xori %true, %br_cond : i1
-
-  // There are two dealloc operations inserted in this basic block, one per
-  // successor. Both have the same list of MemRefs to deallocate and the
-  // conditions only differ by the branch condition conjunct.
-  // Note, however, that the retained list differs. Here, both contain the
-  // %select value because it is used in both successors (since it's the same
-  // block), but the value passed via block argument differs (%memref vs.
-  // %alloc).
-  %10:2 = bufferization.dealloc
-           (%base_buffer_memref, %base_buffer_alloc, %base_buffer_alloca
-             : memref<i8>, memref<i8>, memref<i8>)
-        if (%false, %br_cond, %false)
-    retain (%alloc, %select : memref<?xi8>, memref<?xi8>)
-
-  %11:2 = bufferization.dealloc
-           (%base_buffer_memref, %base_buffer_alloc, %base_buffer_alloca
-             : memref<i8>, memref<i8>, memref<i8>)
-        if (%false, %not_br_cond, %false)
-    retain (%memref, %select : memref<?xi8>, memref<?xi8>)
-
-  // Because %select is used in ^bb1 without passing it via block argument, we
-  // need to update it's ownership value here by merging the ownership values
-  // returned by the dealloc operations
-  %new_ownership = arith.select %br_cond, %10#1, %11#1 : i1
-
-  // The terminator is modified to pass along the ownership indicator values
-  // with each MemRef value.
-  cf.cond_br %br_cond, ^bb1(%alloc, %10#0 : memref<?xi8>, i1),
-                       ^bb1(%memref, %11#0 : memref<?xi8>, i1)
-
-// All non-entry basic blocks are modified to have an additional i1 argument for
-// each MemRef value in the argument list.
-^bb1(%13: memref<?xi8>, %14: i1):  // 2 preds: ^bb0, ^bb0
-  test.copy(%13, %select) : (memref<?xi8>, memref<?xi8>)
-
-  %base_buffer_13, ... = memref.extract_strided_metadata %13
-    : memref<?xi8> -> memref<i8>, index, index, index
-  %base_buffer_select, ... = memref.extract_strided_metadata %select
-    : memref<?xi8> -> memref<i8>, index, index, index
-
-  // Here, we don't have a retained list, because the block has no successors
-  // and the return has no operands.
-  bufferization.dealloc (%base_buffer_13, %base_buffer_select
-                          : memref<i8>, memref<i8>)
-                     if (%14, %new_ownership)
-  return
-}
-```
-
-## Buffer Deallocation Simplification Pass
-
-The [semantics of the `bufferization.dealloc` operation](https://mlir.llvm.org/docs/Dialects/BufferizationOps/#bufferizationdealloc-bufferizationdeallocop)
-provide a lot of opportunities for optimizations which can be conveniently split
-into patterns using the greedy pattern rewriter. Some of those patterns need
-access to additional analyses such as an analysis that can determine whether two
-MemRef values must, may, or never originate from the same buffer allocation.
-These patterns are collected in the Buffer Deallocation Simplification pass,
-while patterns that don't need additional analyses are registered as part of the
-regular canonicalizer pass. This pass is best run after
-`--ownership-based-buffer-deallocation` followed by `--canonicalize`.
-
-The pass applies patterns for the following simplifications:
-*   Remove MemRefs from retain list when guaranteed to not alias with any value
-    in the 'memref' operand list. This avoids an additional aliasing check with
-    the removed value.
-*   Split off values in the 'memref' list to new `bufferization.dealloc`
-    operations only containing this value in the 'memref' list when it is
-    guaranteed to not alias with any other value in the 'memref' list. This
-    avoids at least one aliasing check at runtime and enables using a more
-    efficient lowering for this new `bufferization.dealloc` operation.
-*   Remove values from the 'memref' operand list when it is guaranteed to alias
-    with at least one value in the 'retained' list and may not alias any other
-    value in the 'retain' list.
-
-## Lower Deallocations Pass
-
-The `-lower-deallocations` pass transforms all `bufferization.dealloc`
-operations to `memref.dealloc` operations and may also insert operations from
-the `scf`, `func`, and `arith` dialects to make deallocations conditional and
-check whether two MemRef values come from the same allocation at runtime (when
-the `buffer-deallocation-simplification` pass wasn't able to determine it
-statically).
-
-The same lowering of the `bufferization.dealloc` operation is also part of the
-`-convert-bufferization-to-memref` conversion pass which also lowers all the
-other operations of the bufferization dialect.
-
-We distinguish multiple cases in this lowering pass to provide an overall more
-efficient lowering. In the general case, a library function is created to avoid
-quadratic code size explosion (relative to the number of operands of the dealloc
-operation). The specialized lowerings aim to avoid this library function because
-it requires allocating auxiliary MemRefs of index values.
-
-### Generic Lowering
-
-A library function is generated to avoid code-size blow-up. On a high level, the
-base-memref of all operands is extracted as an index value and stored into
-specifically allocated MemRefs and passed to the library function which then
-determines whether they come from the same original allocation. This information
-is needed to avoid double-free situations and to correctly retain the MemRef
-values in the `retained` list.
-
-**Dealloc Operation Lowering**
-
-This lowering supports all features the dealloc operation has to offer. It
-computes the base pointer of each memref (as an index), stores it in a
-new memref helper structure and passes it to the helper function generated
-in `buildDeallocationLibraryFunction`. The results are stored in two lists
-(represented as MemRefs) of booleans passed as arguments. The first list
-stores whether the corresponding condition should be deallocated, the
-second list stores the ownership of the retained values which can be used
-to replace the result values of the `bufferization.dealloc` operation.
-
-Example:
-```
-%0:2 = bufferization.dealloc (%m0, %m1 : memref<2xf32>, memref<5xf32>)
-                          if (%cond0, %cond1)
-                      retain (%r0, %r1 : memref<1xf32>, memref<2xf32>)
-```
-lowers to (simplified):
-```
-%c0 = arith.constant 0 : index
-%c1 = arith.constant 1 : index
-%dealloc_base_pointer_list = memref.alloc() : memref<2xindex>
-%cond_list = memref.alloc() : memref<2xi1>
-%retain_base_pointer_list = memref.alloc() : memref<2xindex>
-%m0_base_pointer = memref.extract_aligned_pointer_as_index %m0
-memref.store %m0_base_pointer, %dealloc_base_pointer_list[%c0]
-%m1_base_pointer = memref.extract_aligned_pointer_as_index %m1
-memref.store %m1_base_pointer, %dealloc_base_pointer_list[%c1]
-memref.store %cond0, %cond_list[%c0]
-memref.store %cond1, %cond_list[%c1]
-%r0_base_pointer = memref.extract_aligned_pointer_as_index %r0
-memref.store %r0_base_pointer, %retain_base_pointer_list[%c0]
-%r1_base_pointer = memref.extract_aligned_pointer_as_index %r1
-memref.store %r1_base_pointer, %retain_base_pointer_list[%c1]
-%dyn_dealloc_base_pointer_list = memref.cast %dealloc_base_pointer_list :
-   memref<2xindex> to memref<?xindex>
-%dyn_cond_list = memref.cast %cond_list : memref<2xi1> to memref<?xi1>
-%dyn_retain_base_pointer_list = memref.cast %retain_base_pointer_list :
-   memref<2xindex> to memref<?xindex>
-%dealloc_cond_out = memref.alloc() : memref<2xi1>
-%ownership_out = memref.alloc() : memref<2xi1>
-%dyn_dealloc_cond_out = memref.cast %dealloc_cond_out :
-   memref<2xi1> to memref<?xi1>
-%dyn_ownership_out = memref.cast %ownership_out :
-   memref<2xi1> to memref<?xi1>
-call @dealloc_helper(%dyn_dealloc_base_pointer_list,
-                     %dyn_retain_base_pointer_list,
-                     %dyn_cond_list,
-                     %dyn_dealloc_cond_out,
-                     %dyn_ownership_out) : (...)
-%m0_dealloc_cond = memref.load %dyn_dealloc_cond_out[%c0] : memref<2xi1>
-scf.if %m0_dealloc_cond {
-  memref.dealloc %m0 : memref<2xf32>
-}
-%m1_dealloc_cond = memref.load %dyn_dealloc_cond_out[%c1] : memref<2xi1>
-scf.if %m1_dealloc_cond {
-  memref.dealloc %m1 : memref<5xf32>
-}
-%r0_ownership = memref.load %dyn_ownership_out[%c0] : memref<2xi1>
-%r1_ownership = memref.load %dyn_ownership_out[%c1] : memref<2xi1>
-memref.dealloc %dealloc_base_pointer_list : memref<2xindex>
-memref.dealloc %retain_base_pointer_list : memref<2xindex>
-memref.dealloc %cond_list : memref<2xi1>
-memref.dealloc %dealloc_cond_out : memref<2xi1>
-memref.dealloc %ownership_out : memref<2xi1>
-// replace %0#0 with %r0_ownership
-// replace %0#1 with %r1_ownership
-```
-
-**Library function**
-
-A library function is built per compilation unit that can be called at
-bufferization dealloc sites to determine whether two MemRefs come from the same
-allocation and their new ownerships.
-
-The generated function takes two MemRefs of indices and three MemRefs of
-booleans as arguments:
-  * The first argument A should contain the result of the
-  extract_aligned_pointer_as_index operation applied to the MemRefs to be
-  deallocated
-  * The second argument B should contain the result of the
-  extract_aligned_pointer_as_index operation applied to the MemRefs to be
-  retained
-  * The third argument C should contain the conditions as passed directly
-  to the deallocation operation.
-  * The fourth argument D is used to pass results to the caller. Those
-  represent the condition under which the MemRef at the corresponding
-  position in A should be deallocated.
-  * The fifth argument E is used to pass results to the caller. It
-  provides the ownership value corresponding the the MemRef at the same
-  position in B
-
-This helper function is supposed to be called once for each
-`bufferization.dealloc` operation to determine the deallocation need and
-new ownership indicator for the retained values, but does not perform the
-deallocation itself.
-
-Generated code:
-```
-func.func @dealloc_helper(
-    %dyn_dealloc_base_pointer_list: memref<?xindex>,
-    %dyn_retain_base_pointer_list: memref<?xindex>,
-    %dyn_cond_list: memref<?xi1>,
-    %dyn_dealloc_cond_out: memref<?xi1>,
-    %dyn_ownership_out: memref<?xi1>) {
-  %c0 = arith.constant 0 : index
-  %c1 = arith.constant 1 : index
-  %true = arith.constant true
-  %false = arith.constant false
-  %num_dealloc_memrefs = memref.dim %dyn_dealloc_base_pointer_list, %c0
-  %num_retain_memrefs = memref.dim %dyn_retain_base_pointer_list, %c0
-  // Zero initialize result buffer.
-  scf.for %i = %c0 to %num_retain_memrefs step %c1 {
-    memref.store %false, %dyn_ownership_out[%i] : memref<?xi1>
-  }
-  scf.for %i = %c0 to %num_dealloc_memrefs step %c1 {
-    %dealloc_bp = memref.load %dyn_dealloc_base_pointer_list[%i]
-    %cond = memref.load %dyn_cond_list[%i]
-    // Check for aliasing with retained memrefs.
-    %does_not_alias_retained = scf.for %j = %c0 to %num_retain_memrefs
-        step %c1 iter_args(%does_not_alias_aggregated = %true) -> (i1) {
-      %retain_bp = memref.load %dyn_retain_base_pointer_list[%j]
-      %does_alias = arith.cmpi eq, %retain_bp, %dealloc_bp : index
-      scf.if %does_alias {
-        %curr_ownership = memref.load %dyn_ownership_out[%j]
-        %updated_ownership = arith.ori %curr_ownership, %cond : i1
-        memref.store %updated_ownership, %dyn_ownership_out[%j]
-      }
-      %does_not_alias = arith.cmpi ne, %retain_bp, %dealloc_bp : index
-      %updated_aggregate = arith.andi %does_not_alias_aggregated,
-                                      %does_not_alias : i1
-      scf.yield %updated_aggregate : i1
-    }
-    // Check for aliasing with dealloc memrefs in the list before the
-    // current one, i.e.,
-    // `fix i, forall j < i: check_aliasing(%dyn_dealloc_base_pointer[j],
-    // %dyn_dealloc_base_pointer[i])`
-    %does_not_alias_any = scf.for %j = %c0 to %i step %c1
-       iter_args(%does_not_alias_agg = %does_not_alias_retained) -> (i1) {
-      %prev_dealloc_bp = memref.load %dyn_dealloc_base_pointer_list[%j]
-      %does_not_alias = arith.cmpi ne, %prev_dealloc_bp, %dealloc_bp
-      %updated_alias_agg = arith.andi %does_not_alias_agg, %does_not_alias
-      scf.yield %updated_alias_agg : i1
-    }
-    %dealloc_cond = arith.andi %does_not_alias_any, %cond : i1
-    memref.store %dealloc_cond, %dyn_dealloc_cond_out[%i] : memref<?xi1>
-  }
-  return
-}
-```
-
-### Specialized Lowerings
-
-Currently, there are two special lowerings for common cases to avoid the library
-function and thus unnecessary memory load and store operations and function
-calls:
-
-**One memref, no retained**
-
-Lower a simple case without any retained values and a single MemRef. Ideally,
-static analysis can provide enough information such that the
-`buffer-deallocation-simplification` pass is able to split the dealloc
-operations up into this simple case as much as possible before running this
-pass.
-
-Example:
-```mlir
-bufferization.dealloc (%arg0 : memref<2xf32>) if (%arg1)
-```
-is lowered to
-```mlir
-scf.if %arg1 {
-  memref.dealloc %arg0 : memref<2xf32>
-}
-```
-
-In most cases, the branch condition is either constant 'true' or 'false' and can
-thus be optimized away entirely by the canonicalizer pass.
-
-**One memref, arbitrarily many retained**
-
-A special case lowering for the deallocation operation with exactly one MemRef,
-but an arbitrary number of retained values. The size of the code produced by
-this lowering is linear to the number of retained values.
-
-Example:
-```mlir
-%0:2 = bufferization.dealloc (%m : memref<2xf32>) if (%cond)
-                      retain (%r0, %r1 : memref<1xf32>, memref<2xf32>)
-return %0#0, %0#1 : i1, i1
-```
-is lowered to
-```mlir
-%m_base_pointer = memref.extract_aligned_pointer_as_index %m
-%r0_base_pointer = memref.extract_aligned_pointer_as_index %r0
-%r0_does_not_alias = arith.cmpi ne, %m_base_pointer, %r0_base_pointer
-%r1_base_pointer = memref.extract_aligned_pointer_as_index %r1
-%r1_does_not_alias = arith.cmpi ne, %m_base_pointer, %r1_base_pointer
-%not_retained = arith.andi %r0_does_not_alias, %r1_does_not_alias : i1
-%should_dealloc = arith.andi %not_retained, %cond : i1
-scf.if %should_dealloc {
-  memref.dealloc %m : memref<2xf32>
-}
-%true = arith.constant true
-%r0_does_alias = arith.xori %r0_does_not_alias, %true : i1
-%r0_ownership = arith.andi %r0_does_alias, %cond : i1
-%r1_does_alias = arith.xori %r1_does_not_alias, %true : i1
-%r1_ownership = arith.andi %r1_does_alias, %cond : i1
-return %r0_ownership, %r1_ownership : i1, i1
-```
-
 ## Memory Layouts
 
 One-Shot Bufferize bufferizes ops from top to bottom. This works well when all
@@ -1059,7 +408,7 @@ of:
 1.  Buffer optimizations such as `buffer-hoisting`, `buffer-loop-hoisting`, and
     `promote-buffers-to-stack`, which do optimizations that are only exposed
     after bufferization.
-1.  Finally, running the [buffer deallocation](BufferDeallocationInternals.md)
+1.  Finally, running the [ownership-based buffer deallocation](OwnershipBasedBufferDeallocation.md)
     pass.
 
 After buffer deallocation has been completed, the program will be quite
diff --git a/mlir/docs/OwnershipBasedBufferDeallocation.md b/mlir/docs/OwnershipBasedBufferDeallocation.md
new file mode 100644
index 0000000000000..9036c811c5daf
--- /dev/null
+++ b/mlir/docs/OwnershipBasedBufferDeallocation.md
@@ -0,0 +1,619 @@
+# Ownership-based Buffer Deallocation
+
+[TOC]
+
+One-Shot Bufferize does not deallocate any buffers that it allocates. After
+running One-Shot Bufferize, the resulting IR may have a number of `memref.alloc`
+ops, but no `memref.dealloc` ops. Buffer dellocation is delegated to the
+`-ownership-based-buffer-deallocation` pass. This pass supersedes the now
+deprecated `-buffer-deallocation` pass, which does not work well with
+One-Shot Bufferize.
+
+On a high level, buffers are "owned" by a basic block. Ownership materializes
+as an `i1` SSA value and can be thought of as "responsibility to deallocate". It
+is conceptually similar to `std::unique_ptr` in C++.
+
+There are few additional preprocessing and postprocessing passes that should be
+run together with the ownership-based buffer deallocation pass. The recommended
+compilation pipeline is as follows:
+
+```
+one-shot-bufferize
+       |          it's recommended to perform all bufferization here at latest,
+       |       <- any allocations inserted after this point have to be handled
+       V          manually
+expand-realloc
+       V
+ownership-based-buffer-deallocation
+       V
+  canonicalize <- mostly for scf.if simplifications
+       V
+buffer-deallocation-simplification
+       V       <- from this point onwards no tensor values are allowed
+lower-deallocations
+       V
+      CSE
+       V
+  canonicalize
+```
+
+The entire deallocation pipeline (excluding `-one-shot-bufferize`) is exposed
+as `-buffer-deallocation-pipeline`.
+
+The ownership-based buffer deallocation pass processes operations implementing
+`FunctionOpInterface` one-by-one without analysing the call-graph.
+This means that there have to be [some rules](#function-boundary-abi) on how
+MemRefs are handled when being passed from one function to another. The rest of
+the pass revolves heavily around the `bufferization.dealloc` operation which is
+inserted at the end of each basic block with appropriate operands and should be
+optimized using the Buffer Deallocation Simplification pass
+(`--buffer-deallocation-simplification`) and the regular canonicalizer
+(`--canonicalize`). Lowering the result of the
+`-ownership-based-buffer-deallocation` pass directly using
+`--convert-bufferization-to-memref` without beforehand optimization is not
+recommended as it will lead to very inefficient code (the runtime-cost of
+`bufferization.dealloc` is `O(|memrefs|^2+|memref|*|retained|)`).
+
+## Function boundary ABI
+
+The Buffer Deallocation pass operates on the level of operations implementing
+the `FunctionOpInterface`. Such operations can take MemRefs as arguments, but
+also return them. To ensure compatibility among all functions (including
+external ones), some rules have to be enforced:
+*   When a MemRef is passed as a function argument, ownership is never acquired.
+    It is always the caller's responsibility to deallocate such MemRefs.
+*   Returning a MemRef from a function always passes ownership to the caller,
+    i.e., it is also the caller's responsibility to deallocate memrefs returned
+    from a called function.
+*   A function must not return a MemRef with the same allocated base buffer as
+    one of its arguments (in this case a copy has to be created). Note that in
+    this context two subviews of the same buffer that don't overlap are also
+    considered to alias.
+
+For external functions (e.g., library functions written externally in C), the
+externally provided implementation has to adhere to these rules and they are
+just assumed by the buffer deallocation pass. Functions on which the
+deallocation pass is applied and for which the implementation is accessible are
+modified by the pass such that the ABI is respected (i.e., buffer copies are
+inserted when necessary).
+
+## Inserting `bufferization.dealloc` operations
+
+`bufferization.dealloc` and ownership indicators are the main abstractions in
+the ownership-based buffer deallocation pass. `bufferization.dealloc`
+deallocates all given buffers if the respective ownership indicator is set and
+there is no aliasing buffer in the retain list.
+
+![branch_example_pre_move](/includes/img/bufferization_dealloc_op.svg)
+
+`bufferization.dealloc` operations are unconditionally inserted at the end of
+each basic block (just before the terminator). The majority of the pass is about
+finding the correct operands for this operation. There are three variadic
+operand lists to be populated, the first contains all MemRef values that may
+need to be deallocated, the second list contains their associated ownership
+values (of `i1` type), and the third list contains MemRef values that are still
+needed at a later point and should thus not be deallocated (e.g., yielded or
+returned buffers).
+
+`bufferization.dealloc` allows us to deal with any kind of aliasing behavior: it
+lowers to runtime aliasing checks when not enough information can be collected
+statically. When enough aliasing information is statically available, operands
+or the entire op may fold away.
+
+**Ownerships**
+
+To do so, we use a concept of ownership indicators of memrefs which materialize
+as an `i1` value for any SSA value of `memref` type, indicating whether the
+basic block in which it was materialized has ownership of this MemRef. Ideally,
+this is a constant `true` or `false`, but might also be a non-constant SSA
+value. To keep track of those ownership values without immediately materializing
+them (which might require insertion of `bufferization.clone` operations or
+operations checking for aliasing at runtime at positions where we don't actually
+need a materialized value), we use the `Ownership` class. This class represents
+the ownership in three states forming a lattice on a partial order:
+```
+forall X in SSA values. uninitialized < unique(X) < unknown
+forall X, Y in SSA values.
+  unique(X) == unique(Y) iff X and Y always evaluate to the same value
+  unique(X) != unique(Y) otherwise
+```
+Intuitively, the states have the following meaning:
+*   Uninitialized: the ownership is not initialized yet, this is the default
+    state; once an operation is finished processing the ownership of all
+    operation results with MemRef type should not be uninitialized anymore.
+*   Unique: there is a specific SSA value that can be queried to check ownership
+    without materializing any additional IR
+*   Unknown: no specific SSA value is available without materializing additional
+    IR, typically this is because two ownerships in 'Unique' state would have to
+    be merged manually (e.g., the result of an `arith.select` either has the
+    ownership of the then or else case depending on the condition value,
+    inserting another `arith.select` for the ownership values can perform the
+    merge and provide a 'Unique' ownership for the result), however, in the
+    general case this 'Unknown' state has to be assigned.
+
+Implied by the above partial order, the pass combines two ownerships in the
+following way:
+
+| Ownership 1   | Ownership 2   | Combined Ownership |
+|:--------------|:--------------|:-------------------|
+| uninitialized | uninitialized | uninitialized      |
+| unique(X)     | uninitialized | unique(X)          |
+| unique(X)     | unique(X)     | unique(X)          |
+| unique(X)     | unique(Y)     | unknown            |
+| unknown       | unique        | unknown            |
+| unknown       | uninitialized | unknown            |
+| <td colspan=3> + symmetric cases                   |
+
+**Collecting the list of MemRefs that potentially need to be deallocated**
+
+For a given block, the list of MemRefs that potentially need to be deallocated
+at the end of that block is computed by keeping track of all values for which
+the block potentially takes over ownership. This includes MemRefs provided as
+basic block arguments, interface handlers for operations like `memref.alloc` and
+`func.call`, but also liveness information in regions with multiple basic
+blocks.  More concretely, it is computed by taking the MemRefs in the 'in' set
+of the liveness analysis of the current basic block B, appended by the MemRef
+block arguments and by the set of MemRefs allocated in B itself (determined by
+the interface handlers), then subtracted (also determined by the interface
+handlers) by the set of MemRefs deallocated in B.
+
+Note that we don't have to take the intersection of the liveness 'in' set with
+the 'out' set of the predecessor block because a value that is in the 'in' set
+must be defined in an ancestor block that dominates all direct predecessors and
+thus the 'in' set of this block is a subset of the 'out' sets of each
+predecessor.
+
+```
+memrefs = filter((liveIn(block) U
+  allocated(block) U arguments(block)) \ deallocated(block), isMemRef)
+```
+
+The list of conditions for the second variadic operands list of
+`bufferization.dealloc` is computed by querying the stored ownership value for
+each of the MemRefs collected as described above. The ownership state is updated
+by the interface handlers while processing the basic block.
+
+**Collecting the list of MemRefs to retain**
+
+Given a basic block B, the list of MemRefs that have to be retained can be
+different for each successor block S.  For the two basic blocks B and S and the
+values passed via block arguments to the destination block S, we compute the
+list of MemRefs that have to be retained in B by taking the MemRefs in the
+successor operand list of the terminator and the MemRefs in the 'out' set of the
+liveness analysis for B intersected with the 'in' set of the destination block
+S.
+
+This list of retained values makes sure that we cannot run into use-after-free
+situations even if no aliasing information is present at compile-time.
+
+```
+toRetain = filter(successorOperands + (liveOut(fromBlock) insersect
+  liveIn(toBlock)), isMemRef)
+```
+
+## Supported interfaces
+
+The pass uses liveness analysis and a few interfaces:
+*   `FunctionOpInterface`
+*   `CallOpInterface`
+*   `MemoryEffectOpInterface`
+*   `RegionBranchOpInterface`
+*   `RegionBranchTerminatorOpInterface`
+
+Due to insufficient information provided by the interface, it also special-cases
+on the `cf.cond_br` operation and makes some assumptions about operations
+implementing the `RegionBranchOpInterface` at the moment, but improving the
+interfaces would allow us to remove those dependencies in the future.
+
+## Limitations
+
+The Buffer Deallocation pass has some requirements and limitations on the input
+IR. These are checked in the beginning of the pass and errors are emitted
+accordingly:
+*   The set of interfaces the pass operates on must be implemented (correctly).
+    E.g., if there is an operation present with a nested region, but does not
+    implement the `RegionBranchOpInterface`, an error is emitted because the
+    pass cannot know the semantics of the nested region (and does not make any
+    default assumptions on it).
+*   No explicit control-flow loops are present. Currently, only loops using
+    structural-control-flow are supported.  However, this limitation could be
+    lifted in the future.
+*   Deallocation operations should not be present already. The pass should
+    handle them correctly already (at least in most cases), but it's not
+    supported yet due to insufficient testing.
+*   Terminators must implement either `RegionBranchTerminatorOpInterface` or
+    `BranchOpInterface`, but not both. Terminators with more than one successor
+    are not supported (except `cf.cond_br`). This is not a fundamental
+    limitation, but there is no use-case justifying the more complex
+    implementation at the moment.
+
+## Example
+
+The following example contains a few interesting cases:
+*   Basic block arguments are modified to also pass along the ownership
+    indicator, but not for entry blocks, where the function boundary ABI
+    is applied instead.
+*   The result of `arith.select` initially has 'Unknown' assigned as ownership,
+    but once the `bufferization.dealloc` operation is inserted it is put in the
+    'retained' list (since it has uses in a later basic block) and thus the
+    'Unknown' ownership can be replaced with a 'Unique' ownership using the
+    corresponding result of the dealloc operation.
+*   The `cf.cond_br` operation has more than one successor and thus has to
+    insert two `bufferization.dealloc` operations (one for each successor).
+    While they have the same list of MemRefs to deallocate (because they perform
+    the deallocations for the same block), it must be taken into account that
+    some MemRefs remain *live* for one branch but not the other (thus set
+    intersection is performed on the *live-out* of the current block and the
+    *live-in* of the target block). Also, `cf.cond_br` supports separate
+    forwarding operands for each successor. To make sure that no MemRef is
+    deallocated twice (because there are two `bufferization.dealloc` operations
+    with the same MemRefs to deallocate), the condition operands are adjusted to
+    take the branch condition into account. While a generic lowering for such
+    terminator operations could be implemented, a specialized implementation can
+    take all the semantics of this particular operation into account and thus
+    generate a more efficient lowering.
+
+```mlir
+func.func @example(%memref: memref<?xi8>, %select_cond: i1, %br_cond: i1) {
+  %alloc = memref.alloc() : memref<?xi8>
+  %alloca = memref.alloca() : memref<?xi8>
+  %select = arith.select %select_cond, %alloc, %alloca : memref<?xi8>
+  cf.cond_br %br_cond, ^bb1(%alloc : memref<?xi8>), ^bb1(%memref : memref<?xi8>)
+^bb1(%bbarg: memref<?xi8>):
+  test.copy(%bbarg, %select) : (memref<?xi8>, memref<?xi8>)
+  return
+}
+```
+
+After running `--ownership-based-buffer-deallocation`, it looks as follows:
+
+```mlir
+// Function boundary ABI: ownership of `%memref` will never be acquired.
+func.func @example(%memref: memref<?xi8>, %select_cond: i1, %br_cond: i1) {
+  %false = arith.constant false
+  %true = arith.constant true
+
+  // The ownership of a MemRef defined by the `memref.alloc` operation is always
+  // assigned to be 'true'.
+  %alloc = memref.alloc() : memref<?xi8>
+
+  // The ownership of a MemRef defined by the `memref.alloca` operation is
+  // always assigned to be 'false'.
+  %alloca = memref.alloca() : memref<?xi8>
+
+  // The ownership of %select will be the join of the ownership of %alloc and
+  // the ownership of %alloca, i.e., of %true and %false. Because the pass does
+  // not know about the semantics of the `arith.select` operation (unless a
+  // custom handler is implemented), the ownership join will be 'Unknown'. If
+  // the materialized ownership indicator of %select is needed, either a clone
+  // has to be created for which %true is assigned as ownership or the result
+  // of a `bufferization.dealloc` where %select is in the retain list has to be
+  // used.
+  %select = arith.select %select_cond, %alloc, %alloca : memref<?xi8>
+
+  // We use `memref.extract_strided_metadata` to get the base memref since it is
+  // not allowed to pass arbitrary memrefs to `memref.dealloc`. This property is
+  // already enforced for `bufferization.dealloc`
+  %base_buffer_memref, ... = memref.extract_strided_metadata %memref
+    : memref<?xi8> -> memref<i8>, index, index, index
+  %base_buffer_alloc, ... = memref.extract_strided_metadata %alloc
+    : memref<?xi8> -> memref<i8>, index, index, index
+  %base_buffer_alloca, ... = memref.extract_strided_metadata %alloca
+    : memref<?xi8> -> memref<i8>, index, index, index
+
+  // The deallocation conditions need to be adjusted to incorporate the branch
+  // condition. In this example, this requires only a single negation, but might
+  // also require multiple arith.andi operations.
+  %not_br_cond = arith.xori %true, %br_cond : i1
+
+  // There are two dealloc operations inserted in this basic block, one per
+  // successor. Both have the same list of MemRefs to deallocate and the
+  // conditions only differ by the branch condition conjunct.
+  // Note, however, that the retained list differs. Here, both contain the
+  // %select value because it is used in both successors (since it's the same
+  // block), but the value passed via block argument differs (%memref vs.
+  // %alloc).
+  %10:2 = bufferization.dealloc
+           (%base_buffer_memref, %base_buffer_alloc, %base_buffer_alloca
+             : memref<i8>, memref<i8>, memref<i8>)
+        if (%false, %br_cond, %false)
+    retain (%alloc, %select : memref<?xi8>, memref<?xi8>)
+
+  %11:2 = bufferization.dealloc
+           (%base_buffer_memref, %base_buffer_alloc, %base_buffer_alloca
+             : memref<i8>, memref<i8>, memref<i8>)
+        if (%false, %not_br_cond, %false)
+    retain (%memref, %select : memref<?xi8>, memref<?xi8>)
+
+  // Because %select is used in ^bb1 without passing it via block argument, we
+  // need to update it's ownership value here by merging the ownership values
+  // returned by the dealloc operations
+  %new_ownership = arith.select %br_cond, %10#1, %11#1 : i1
+
+  // The terminator is modified to pass along the ownership indicator values
+  // with each MemRef value.
+  cf.cond_br %br_cond, ^bb1(%alloc, %10#0 : memref<?xi8>, i1),
+                       ^bb1(%memref, %11#0 : memref<?xi8>, i1)
+
+// All non-entry basic blocks are modified to have an additional i1 argument for
+// each MemRef value in the argument list.
+^bb1(%13: memref<?xi8>, %14: i1):  // 2 preds: ^bb0, ^bb0
+  test.copy(%13, %select) : (memref<?xi8>, memref<?xi8>)
+
+  %base_buffer_13, ... = memref.extract_strided_metadata %13
+    : memref<?xi8> -> memref<i8>, index, index, index
+  %base_buffer_select, ... = memref.extract_strided_metadata %select
+    : memref<?xi8> -> memref<i8>, index, index, index
+
+  // Here, we don't have a retained list, because the block has no successors
+  // and the return has no operands.
+  bufferization.dealloc (%base_buffer_13, %base_buffer_select
+                          : memref<i8>, memref<i8>)
+                     if (%14, %new_ownership)
+  return
+}
+```
+
+## Buffer Deallocation Simplification Pass
+
+The [semantics of the `bufferization.dealloc` operation](#bufferizationdealloc-bufferizationdeallocop)
+provide a lot of opportunities for optimizations which can be conveniently split
+into patterns using the greedy pattern rewriter. Some of those patterns need
+access to additional analyses such as an analysis that can determine whether two
+MemRef values must, may, or never originate from the same buffer allocation.
+These patterns are collected in the Buffer Deallocation Simplification pass,
+while patterns that don't need additional analyses are registered as part of the
+regular canonicalizer pass. This pass is best run after
+`--ownership-based-buffer-deallocation` followed by `--canonicalize`.
+
+The pass applies patterns for the following simplifications:
+*   Remove MemRefs from retain list when guaranteed to not alias with any value
+    in the 'memref' operand list. This avoids an additional aliasing check with
+    the removed value.
+*   Split off values in the 'memref' list to new `bufferization.dealloc`
+    operations only containing this value in the 'memref' list when it is
+    guaranteed to not alias with any other value in the 'memref' list. This
+    avoids at least one aliasing check at runtime and enables using a more
+    efficient lowering for this new `bufferization.dealloc` operation.
+*   Remove values from the 'memref' operand list when it is guaranteed to alias
+    with at least one value in the 'retained' list and may not alias any other
+    value in the 'retain' list.
+
+## Lower Deallocations Pass
+
+The `-lower-deallocations` pass transforms all `bufferization.dealloc`
+operations to `memref.dealloc` operations and may also insert operations from
+the `scf`, `func`, and `arith` dialects to make deallocations conditional and
+check whether two MemRef values come from the same allocation at runtime (when
+the `buffer-deallocation-simplification` pass wasn't able to determine it
+statically).
+
+The same lowering of the `bufferization.dealloc` operation is also part of the
+`-convert-bufferization-to-memref` conversion pass which also lowers all the
+other operations of the bufferization dialect.
+
+We distinguish multiple cases in this lowering pass to provide an overall more
+efficient lowering. In the general case, a library function is created to avoid
+quadratic code size explosion (relative to the number of operands of the dealloc
+operation). The specialized lowerings aim to avoid this library function because
+it requires allocating auxiliary MemRefs of index values.
+
+### Generic Lowering
+
+A library function is generated to avoid code-size blow-up. On a high level, the
+base-memref of all operands is extracted as an index value and stored into
+specifically allocated MemRefs and passed to the library function which then
+determines whether they come from the same original allocation. This information
+is needed to avoid double-free situations and to correctly retain the MemRef
+values in the `retained` list.
+
+**Dealloc Operation Lowering**
+
+This lowering supports all features the dealloc operation has to offer. It
+computes the base pointer of each memref (as an index), stores it in a
+new memref helper structure and passes it to the helper function generated
+in `buildDeallocationLibraryFunction`. The results are stored in two lists
+(represented as MemRefs) of booleans passed as arguments. The first list
+stores whether the corresponding condition should be deallocated, the
+second list stores the ownership of the retained values which can be used
+to replace the result values of the `bufferization.dealloc` operation.
+
+Example:
+```mlir
+%0:2 = bufferization.dealloc (%m0, %m1 : memref<2xf32>, memref<5xf32>)
+                          if (%cond0, %cond1)
+                      retain (%r0, %r1 : memref<1xf32>, memref<2xf32>)
+```
+lowers to (simplified):
+```mlir
+%c0 = arith.constant 0 : index
+%c1 = arith.constant 1 : index
+%dealloc_base_pointer_list = memref.alloc() : memref<2xindex>
+%cond_list = memref.alloc() : memref<2xi1>
+%retain_base_pointer_list = memref.alloc() : memref<2xindex>
+%m0_base_pointer = memref.extract_aligned_pointer_as_index %m0
+memref.store %m0_base_pointer, %dealloc_base_pointer_list[%c0]
+%m1_base_pointer = memref.extract_aligned_pointer_as_index %m1
+memref.store %m1_base_pointer, %dealloc_base_pointer_list[%c1]
+memref.store %cond0, %cond_list[%c0]
+memref.store %cond1, %cond_list[%c1]
+%r0_base_pointer = memref.extract_aligned_pointer_as_index %r0
+memref.store %r0_base_pointer, %retain_base_pointer_list[%c0]
+%r1_base_pointer = memref.extract_aligned_pointer_as_index %r1
+memref.store %r1_base_pointer, %retain_base_pointer_list[%c1]
+%dyn_dealloc_base_pointer_list = memref.cast %dealloc_base_pointer_list :
+   memref<2xindex> to memref<?xindex>
+%dyn_cond_list = memref.cast %cond_list : memref<2xi1> to memref<?xi1>
+%dyn_retain_base_pointer_list = memref.cast %retain_base_pointer_list :
+   memref<2xindex> to memref<?xindex>
+%dealloc_cond_out = memref.alloc() : memref<2xi1>
+%ownership_out = memref.alloc() : memref<2xi1>
+%dyn_dealloc_cond_out = memref.cast %dealloc_cond_out :
+   memref<2xi1> to memref<?xi1>
+%dyn_ownership_out = memref.cast %ownership_out :
+   memref<2xi1> to memref<?xi1>
+call @dealloc_helper(%dyn_dealloc_base_pointer_list,
+                     %dyn_retain_base_pointer_list,
+                     %dyn_cond_list,
+                     %dyn_dealloc_cond_out,
+                     %dyn_ownership_out) : (...)
+%m0_dealloc_cond = memref.load %dyn_dealloc_cond_out[%c0] : memref<2xi1>
+scf.if %m0_dealloc_cond {
+  memref.dealloc %m0 : memref<2xf32>
+}
+%m1_dealloc_cond = memref.load %dyn_dealloc_cond_out[%c1] : memref<2xi1>
+scf.if %m1_dealloc_cond {
+  memref.dealloc %m1 : memref<5xf32>
+}
+%r0_ownership = memref.load %dyn_ownership_out[%c0] : memref<2xi1>
+%r1_ownership = memref.load %dyn_ownership_out[%c1] : memref<2xi1>
+memref.dealloc %dealloc_base_pointer_list : memref<2xindex>
+memref.dealloc %retain_base_pointer_list : memref<2xindex>
+memref.dealloc %cond_list : memref<2xi1>
+memref.dealloc %dealloc_cond_out : memref<2xi1>
+memref.dealloc %ownership_out : memref<2xi1>
+// replace %0#0 with %r0_ownership
+// replace %0#1 with %r1_ownership
+```
+
+**Library function**
+
+A library function is built per compilation unit that can be called at
+bufferization dealloc sites to determine whether two MemRefs come from the same
+allocation and their new ownerships.
+
+The generated function takes two MemRefs of indices and three MemRefs of
+booleans as arguments:
+  * The first argument A should contain the result of the
+  extract_aligned_pointer_as_index operation applied to the MemRefs to be
+  deallocated
+  * The second argument B should contain the result of the
+  extract_aligned_pointer_as_index operation applied to the MemRefs to be
+  retained
+  * The third argument C should contain the conditions as passed directly
+  to the deallocation operation.
+  * The fourth argument D is used to pass results to the caller. Those
+  represent the condition under which the MemRef at the corresponding
+  position in A should be deallocated.
+  * The fifth argument E is used to pass results to the caller. It
+  provides the ownership value corresponding the the MemRef at the same
+  position in B
+
+This helper function is supposed to be called once for each
+`bufferization.dealloc` operation to determine the deallocation need and
+new ownership indicator for the retained values, but does not perform the
+deallocation itself.
+
+Generated code:
+```mlir
+func.func @dealloc_helper(
+    %dyn_dealloc_base_pointer_list: memref<?xindex>,
+    %dyn_retain_base_pointer_list: memref<?xindex>,
+    %dyn_cond_list: memref<?xi1>,
+    %dyn_dealloc_cond_out: memref<?xi1>,
+    %dyn_ownership_out: memref<?xi1>) {
+  %c0 = arith.constant 0 : index
+  %c1 = arith.constant 1 : index
+  %true = arith.constant true
+  %false = arith.constant false
+  %num_dealloc_memrefs = memref.dim %dyn_dealloc_base_pointer_list, %c0
+  %num_retain_memrefs = memref.dim %dyn_retain_base_pointer_list, %c0
+  // Zero initialize result buffer.
+  scf.for %i = %c0 to %num_retain_memrefs step %c1 {
+    memref.store %false, %dyn_ownership_out[%i] : memref<?xi1>
+  }
+  scf.for %i = %c0 to %num_dealloc_memrefs step %c1 {
+    %dealloc_bp = memref.load %dyn_dealloc_base_pointer_list[%i]
+    %cond = memref.load %dyn_cond_list[%i]
+    // Check for aliasing with retained memrefs.
+    %does_not_alias_retained = scf.for %j = %c0 to %num_retain_memrefs
+        step %c1 iter_args(%does_not_alias_aggregated = %true) -> (i1) {
+      %retain_bp = memref.load %dyn_retain_base_pointer_list[%j]
+      %does_alias = arith.cmpi eq, %retain_bp, %dealloc_bp : index
+      scf.if %does_alias {
+        %curr_ownership = memref.load %dyn_ownership_out[%j]
+        %updated_ownership = arith.ori %curr_ownership, %cond : i1
+        memref.store %updated_ownership, %dyn_ownership_out[%j]
+      }
+      %does_not_alias = arith.cmpi ne, %retain_bp, %dealloc_bp : index
+      %updated_aggregate = arith.andi %does_not_alias_aggregated,
+                                      %does_not_alias : i1
+      scf.yield %updated_aggregate : i1
+    }
+    // Check for aliasing with dealloc memrefs in the list before the
+    // current one, i.e.,
+    // `fix i, forall j < i: check_aliasing(%dyn_dealloc_base_pointer[j],
+    // %dyn_dealloc_base_pointer[i])`
+    %does_not_alias_any = scf.for %j = %c0 to %i step %c1
+       iter_args(%does_not_alias_agg = %does_not_alias_retained) -> (i1) {
+      %prev_dealloc_bp = memref.load %dyn_dealloc_base_pointer_list[%j]
+      %does_not_alias = arith.cmpi ne, %prev_dealloc_bp, %dealloc_bp
+      %updated_alias_agg = arith.andi %does_not_alias_agg, %does_not_alias
+      scf.yield %updated_alias_agg : i1
+    }
+    %dealloc_cond = arith.andi %does_not_alias_any, %cond : i1
+    memref.store %dealloc_cond, %dyn_dealloc_cond_out[%i] : memref<?xi1>
+  }
+  return
+}
+```
+
+### Specialized Lowerings
+
+Currently, there are two special lowerings for common cases to avoid the library
+function and thus unnecessary memory load and store operations and function
+calls:
+
+**One memref, no retained**
+
+Lower a simple case without any retained values and a single MemRef. Ideally,
+static analysis can provide enough information such that the
+`buffer-deallocation-simplification` pass is able to split the dealloc
+operations up into this simple case as much as possible before running this
+pass.
+
+Example:
+```mlir
+bufferization.dealloc (%arg0 : memref<2xf32>) if (%arg1)
+```
+is lowered to
+```mlir
+scf.if %arg1 {
+  memref.dealloc %arg0 : memref<2xf32>
+}
+```
+
+In most cases, the branch condition is either constant 'true' or 'false' and can
+thus be optimized away entirely by the canonicalizer pass.
+
+**One memref, arbitrarily many retained**
+
+A special case lowering for the deallocation operation with exactly one MemRef,
+but an arbitrary number of retained values. The size of the code produced by
+this lowering is linear to the number of retained values.
+
+Example:
+```mlir
+%0:2 = bufferization.dealloc (%m : memref<2xf32>) if (%cond)
+                      retain (%r0, %r1 : memref<1xf32>, memref<2xf32>)
+return %0#0, %0#1 : i1, i1
+```
+is lowered to
+```mlir
+%m_base_pointer = memref.extract_aligned_pointer_as_index %m
+%r0_base_pointer = memref.extract_aligned_pointer_as_index %r0
+%r0_does_not_alias = arith.cmpi ne, %m_base_pointer, %r0_base_pointer
+%r1_base_pointer = memref.extract_aligned_pointer_as_index %r1
+%r1_does_not_alias = arith.cmpi ne, %m_base_pointer, %r1_base_pointer
+%not_retained = arith.andi %r0_does_not_alias, %r1_does_not_alias : i1
+%should_dealloc = arith.andi %not_retained, %cond : i1
+scf.if %should_dealloc {
+  memref.dealloc %m : memref<2xf32>
+}
+%true = arith.constant true
+%r0_does_alias = arith.xori %r0_does_not_alias, %true : i1
+%r0_ownership = arith.andi %r0_does_alias, %cond : i1
+%r1_does_alias = arith.xori %r1_does_not_alias, %true : i1
+%r1_ownership = arith.andi %r1_does_alias, %cond : i1
+return %r0_ownership, %r1_ownership : i1, i1
+```
diff --git a/mlir/docs/includes/img/bufferization_dealloc_op.svg b/mlir/docs/includes/img/bufferization_dealloc_op.svg
new file mode 100644
index 0000000000000..3b72ae72c4eb9
--- /dev/null
+++ b/mlir/docs/includes/img/bufferization_dealloc_op.svg
@@ -0,0 +1 @@
+<svg version="1.1" viewBox="0.0 0.0 960.0 384.0" fill="none" stroke="none" stroke-linecap="square" stroke-miterlimit="10" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://www.w3.org/2000/svg"><clipPath id="p.0"><path d="m0 0l960.0 0l0 384.0l-960.0 0l0 -384.0z" clip-rule="nonzero"/></clipPath><g clip-path="url(#p.0)"><path fill="#ffffff" d="m0 0l960.0 0l0 384.0l-960.0 0z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m50.48294 76.19685l882.7086 0l0 165.41731l-882.7086 0z" fill-rule="evenodd"/><path fill="#434343" d="m59.82669 91.22873l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm14.9921875 -2.890625l0 -2.65625q0 -1.09375 -0.25 -1.921875q-0.25 -0.84375 -0.71875 -1.421875q-0.46875 -0.578125 -1.15625 -0.875q-0.6875 -0.3125 -1.5625 -0.3125q-0.859375 0 -1.546875 0.3125q-0.6875 0.296875 -1.15625 0.875q-0.46875 0.578125 -0.71875 1.421875q-0.25 0.828125 -0.25 1.921875l0 2.65625q0 1.09375 0.25 1.9375q0.25 0.828125 0.71875 1.40625q0.484375 0.578125 1.171875 0.875q0.6875 0.296875 1.546875 0.296875q0.875 0 1.546875 -0.296875q0.6875 -0.296875 1.15625 -0.875q0.46875 -0.578125 0.71875 -1.40625q0.25 -0.84375 0.25 -1.9375zm-5.90625 -0.28125q0 -0.203125 0 -0.390625q0 -0.203125 0 -0.40625l0 -1.90625q0 -0.796875 0.15625 -1.390625q0.15625 -0.609375 0.484375 -1.0q0.28125 -0.296875 0.671875 -0.453125q0.390625 -0.171875 0.90625 -0.171875q0.515625 0 0.890625 0.15625q0.390625 0.140625 0.65625 0.4375q0.28125 0.296875 0.4375 0.734375q0.15625 0.421875 0.21875 0.984375l-4.421875 3.40625zm4.453125 0.59375q0 0.828125 -0.171875 1.453125q-0.15625 0.609375 -0.515625 1.0q-0.265625 0.28125 -0.65625 0.4375q-0.375 0.140625 -0.875 0.140625q-0.46875 0 -0.84375 -0.140625q-0.375 -0.140625 -0.640625 -0.40625q-0.28125 -0.28125 -0.46875 -0.703125q-0.171875 -0.4375 -0.25 -1.015625l4.421875 -3.375q0 0.21875 0 0.546875q0 0.328125 0 0.453125l0 1.609375zm6.8203125 3.109375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm0 -6.84375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm8.3828125 1.390625l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm24.484375 0.296875l0 -1.265625l-6.96875 0l0 1.265625l6.96875 0zm0 3.234375l0 -1.25l-6.96875 0l0 1.25l6.96875 0zm19.40625 -1.28125l0 -0.15625q0 -0.671875 -0.109375 -1.265625q-0.109375 -0.59375 -0.328125 -1.09375q-0.203125 -0.4375 -0.46875 -0.78125q-0.265625 -0.359375 -0.59375 -0.609375q-0.375 -0.265625 -0.84375 -0.40625q-0.453125 -0.15625 -1.0 -0.15625q-0.4375 0 -0.8125 0.09375q-0.375 0.078125 -0.6875 0.234375q-0.25 0.140625 -0.46875 0.328125q-0.21875 0.171875 -0.40625 0.40625l0 -4.453125l-1.4375 0l0 12.0l1.328125 0l0.0625 -0.96875q0.140625 0.1875 0.296875 0.34375q0.15625 0.140625 0.328125 0.265625q0.359375 0.25 0.8125 0.375q0.453125 0.140625 1.0 0.140625q0.46875 0 0.875 -0.125q0.421875 -0.109375 0.765625 -0.328125q0.46875 -0.296875 0.8125 -0.765625q0.34375 -0.46875 0.546875 -1.046875q0.15625 -0.453125 0.234375 -0.953125q0.09375 -0.515625 0.09375 -1.078125zm-1.453125 -0.15625l0 0.15625q0 0.375 -0.046875 0.734375q-0.046875 0.359375 -0.140625 0.671875q-0.125 0.421875 -0.34375 0.765625q-0.203125 0.328125 -0.515625 0.546875q-0.234375 0.171875 -0.53125 0.265625q-0.296875 0.078125 -0.65625 0.078125q-0.375 0 -0.6875 -0.09375q-0.3125 -0.09375 -0.5625 -0.28125q-0.25 -0.171875 -0.453125 -0.40625q-0.1875 -0.25 -0.328125 -0.53125l0 -3.6875q0.140625 -0.28125 0.328125 -0.515625q0.1875 -0.25 0.4375 -0.421875q0.25 -0.171875 0.5625 -0.265625q0.3125 -0.109375 0.6875 -0.109375q0.34375 0 0.625 0.078125q0.28125 0.078125 0.5 0.21875q0.328125 0.203125 0.546875 0.546875q0.234375 0.34375 0.375 0.75q0.109375 0.34375 0.15625 0.71875q0.046875 0.375 0.046875 0.78125zm9.4609375 4.296875l1.3125 0l0 -8.453125l-1.453125 0l0 6.0625q-0.125 0.265625 -0.296875 0.5q-0.171875 0.234375 -0.40625 0.390625q-0.28125 0.203125 -0.671875 0.328125q-0.375 0.109375 -0.859375 0.109375q-0.421875 0 -0.734375 -0.109375q-0.3125 -0.109375 -0.53125 -0.390625q-0.21875 -0.265625 -0.328125 -0.71875q-0.109375 -0.453125 -0.109375 -1.140625l0 -5.03125l-1.453125 0l0 5.015625q0 0.953125 0.203125 1.625q0.21875 0.671875 0.59375 1.109375q0.390625 0.4375 0.921875 0.65625q0.546875 0.203125 1.203125 0.203125q0.828125 0 1.453125 -0.34375q0.640625 -0.359375 1.0625 -0.984375l0.09375 1.171875zm6.1796875 0l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm9.6015625 0l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm11.1796875 0.15625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm11.6015625 -1.1875q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm4.2890625 0.15625l0 1.25l2.875 0l0 5.953125l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -7.203125l-4.3125 0zm2.6875 -2.21875q0 0.34375 0.203125 0.578125q0.21875 0.234375 0.65625 0.234375q0.4375 0 0.65625 -0.234375q0.21875 -0.234375 0.21875 -0.578125q0 -0.203125 -0.0625 -0.359375q-0.0625 -0.171875 -0.1875 -0.28125q-0.109375 -0.09375 -0.265625 -0.15625q-0.15625 -0.0625 -0.359375 -0.0625q-0.203125 0 -0.359375 0.0625q-0.15625 0.0625 -0.265625 0.15625q-0.125 0.125 -0.1875 0.296875q-0.046875 0.15625 -0.046875 0.34375zm8.4140625 9.484375l5.109375 -6.25l0 -1.015625l-6.875 0l0 1.1875l4.96875 0l-5.046875 6.203125l0 1.0625l7.234375 0l0 -1.1875l-5.390625 0zm13.4296875 1.1875l1.5 0l0 -0.125q-0.140625 -0.359375 -0.21875 -0.828125q-0.078125 -0.46875 -0.078125 -0.890625l0 -3.9375q0 -0.703125 -0.265625 -1.234375q-0.25 -0.53125 -0.6875 -0.890625q-0.453125 -0.359375 -1.078125 -0.53125q-0.609375 -0.171875 -1.328125 -0.171875q-0.796875 0 -1.421875 0.21875q-0.609375 0.21875 -1.046875 0.5625q-0.4375 0.359375 -0.671875 0.8125q-0.21875 0.4375 -0.234375 0.90625l1.453125 0q0 -0.265625 0.125 -0.515625q0.140625 -0.25 0.375 -0.421875q0.234375 -0.1875 0.5625 -0.296875q0.34375 -0.109375 0.78125 -0.109375q0.453125 0 0.828125 0.125q0.375 0.109375 0.625 0.328125q0.265625 0.21875 0.40625 0.53125q0.140625 0.296875 0.140625 0.671875l0 0.671875l-1.578125 0q-0.890625 0 -1.609375 0.171875q-0.703125 0.171875 -1.21875 0.515625q-0.5 0.34375 -0.78125 0.875q-0.28125 0.515625 -0.28125 1.21875q0 0.53125 0.203125 0.984375q0.21875 0.453125 0.59375 0.78125q0.375 0.34375 0.921875 0.546875q0.546875 0.1875 1.21875 0.1875q0.40625 0 0.765625 -0.09375q0.375 -0.078125 0.703125 -0.234375q0.328125 -0.140625 0.59375 -0.328125q0.265625 -0.203125 0.484375 -0.421875q0.03125 0.265625 0.078125 0.515625q0.046875 0.234375 0.140625 0.40625zm-2.5625 -1.109375q-0.421875 0 -0.75 -0.109375q-0.3125 -0.109375 -0.515625 -0.296875q-0.21875 -0.1875 -0.328125 -0.453125q-0.09375 -0.265625 -0.09375 -0.578125q0 -0.328125 0.109375 -0.578125q0.125 -0.265625 0.359375 -0.453125q0.328125 -0.265625 0.890625 -0.390625q0.5625 -0.140625 1.3125 -0.140625l1.34375 0l0 1.71875q-0.125 0.25 -0.34375 0.484375q-0.21875 0.21875 -0.515625 0.40625q-0.3125 0.171875 -0.6875 0.28125q-0.359375 0.109375 -0.78125 0.109375zm10.0234375 -9.390625l-1.453125 0l0 2.046875l-2.21875 0l0 1.109375l2.21875 0l0 4.59375q0 0.78125 0.203125 1.34375q0.21875 0.546875 0.578125 0.890625q0.359375 0.34375 0.84375 0.515625q0.484375 0.15625 1.046875 0.15625q0.34375 0 0.671875 -0.03125q0.34375 -0.03125 0.65625 -0.09375q0.296875 -0.0625 0.546875 -0.140625q0.265625 -0.078125 0.453125 -0.1875l-0.203125 -1.03125q-0.140625 0.03125 -0.34375 0.078125q-0.203125 0.046875 -0.4375 0.078125q-0.25 0.046875 -0.5 0.078125q-0.25 0.015625 -0.484375 0.015625q-0.3125 0 -0.609375 -0.078125q-0.28125 -0.078125 -0.5 -0.265625q-0.21875 -0.1875 -0.34375 -0.5q-0.125 -0.328125 -0.125 -0.828125l0 -4.59375l3.21875 0l0 -1.109375l-3.21875 0l0 -2.046875zm6.4140625 2.046875l0 1.25l2.875 0l0 5.953125l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -7.203125l-4.3125 0zm2.6875 -2.21875q0 0.34375 0.203125 0.578125q0.21875 0.234375 0.65625 0.234375q0.4375 0 0.65625 -0.234375q0.21875 -0.234375 0.21875 -0.578125q0 -0.203125 -0.0625 -0.359375q-0.0625 -0.171875 -0.1875 -0.28125q-0.109375 -0.09375 -0.265625 -0.15625q-0.15625 -0.0625 -0.359375 -0.0625q-0.203125 0 -0.359375 0.0625q-0.15625 0.0625 -0.265625 0.15625q-0.125 0.125 -0.1875 0.296875q-0.046875 0.15625 -0.046875 0.34375zm6.2734375 6.359375l0 0.171875q0 0.921875 0.265625 1.703125q0.265625 0.78125 0.765625 1.359375q0.5 0.578125 1.203125 0.90625q0.71875 0.328125 1.625 0.328125q0.890625 0 1.59375 -0.328125q0.71875 -0.328125 1.21875 -0.90625q0.484375 -0.578125 0.75 -1.359375q0.265625 -0.78125 0.265625 -1.703125l0 -0.171875q0 -0.90625 -0.265625 -1.6875q-0.265625 -0.796875 -0.75 -1.375q-0.5 -0.578125 -1.21875 -0.90625q-0.71875 -0.328125 -1.609375 -0.328125q-0.890625 0 -1.609375 0.328125q-0.703125 0.328125 -1.203125 0.90625q-0.5 0.578125 -0.765625 1.375q-0.265625 0.78125 -0.265625 1.6875zm1.453125 0.171875l0 -0.171875q0 -0.609375 0.140625 -1.171875q0.15625 -0.578125 0.453125 -1.0q0.296875 -0.4375 0.734375 -0.6875q0.453125 -0.25 1.0625 -0.25q0.59375 0 1.046875 0.25q0.453125 0.25 0.75 0.6875q0.296875 0.421875 0.453125 1.0q0.15625 0.5625 0.15625 1.171875l0 0.171875q0 0.625 -0.15625 1.203125q-0.140625 0.5625 -0.4375 0.984375q-0.3125 0.4375 -0.765625 0.6875q-0.4375 0.25 -1.03125 0.25q-0.609375 0 -1.0625 -0.25q-0.453125 -0.25 -0.75 -0.6875q-0.296875 -0.421875 -0.453125 -0.984375q-0.140625 -0.578125 -0.140625 -1.203125zm8.5546875 4.140625l1.453125 0l0 -6.0625q0.140625 -0.28125 0.34375 -0.5q0.203125 -0.234375 0.4375 -0.40625q0.28125 -0.1875 0.625 -0.296875q0.34375 -0.125 0.75 -0.125q0.453125 0 0.8125 0.109375q0.359375 0.109375 0.609375 0.359375q0.234375 0.234375 0.359375 0.625q0.125 0.390625 0.125 0.953125l0 5.34375l1.453125 0l0 -5.375q0 -0.84375 -0.21875 -1.453125q-0.203125 -0.609375 -0.578125 -1.015625q-0.390625 -0.390625 -0.9375 -0.578125q-0.53125 -0.1875 -1.203125 -0.1875q-0.484375 0 -0.9375 0.140625q-0.4375 0.140625 -0.8125 0.390625q-0.25 0.15625 -0.46875 0.390625q-0.21875 0.21875 -0.40625 0.484375l-0.109375 -1.25l-1.296875 0l0 8.453125zm12.1171875 -0.9375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm6.8203125 -3.359375l0 0.15625q0 0.90625 0.234375 1.703125q0.25 0.78125 0.6875 1.359375q0.4375 0.578125 1.046875 0.90625q0.625 0.328125 1.390625 0.328125q0.78125 0 1.359375 -0.265625q0.59375 -0.265625 1.015625 -0.78125l0.0625 0.890625l1.328125 0l0 -12.0l-1.4375 0l0 4.375q-0.421875 -0.484375 -0.984375 -0.734375q-0.5625 -0.25 -1.328125 -0.25q-0.765625 0 -1.390625 0.328125q-0.625 0.3125 -1.0625 0.875q-0.453125 0.578125 -0.6875 1.375q-0.234375 0.78125 -0.234375 1.734375zm1.4375 0.15625l0 -0.15625q0 -0.625 0.125 -1.171875q0.140625 -0.5625 0.421875 -1.0q0.28125 -0.421875 0.703125 -0.671875q0.4375 -0.25 1.046875 -0.25q0.71875 0 1.1875 0.34375q0.484375 0.328125 0.765625 0.84375l0 3.921875q-0.28125 0.546875 -0.765625 0.875q-0.46875 0.328125 -1.203125 0.328125q-0.59375 0 -1.03125 -0.234375q-0.421875 -0.25 -0.703125 -0.671875q-0.28125 -0.421875 -0.421875 -0.96875q-0.125 -0.5625 -0.125 -1.1875zm12.1640625 4.296875q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm11.6171875 7.421875l1.5 0l0 -0.125q-0.140625 -0.359375 -0.21875 -0.828125q-0.078125 -0.46875 -0.078125 -0.890625l0 -3.9375q0 -0.703125 -0.265625 -1.234375q-0.25 -0.53125 -0.6875 -0.890625q-0.453125 -0.359375 -1.078125 -0.53125q-0.609375 -0.171875 -1.328125 -0.171875q-0.796875 0 -1.421875 0.21875q-0.609375 0.21875 -1.046875 0.5625q-0.4375 0.359375 -0.671875 0.8125q-0.21875 0.4375 -0.234375 0.90625l1.453125 0q0 -0.265625 0.125 -0.515625q0.140625 -0.25 0.375 -0.421875q0.234375 -0.1875 0.5625 -0.296875q0.34375 -0.109375 0.78125 -0.109375q0.453125 0 0.828125 0.125q0.375 0.109375 0.625 0.328125q0.265625 0.21875 0.40625 0.53125q0.140625 0.296875 0.140625 0.671875l0 0.671875l-1.578125 0q-0.890625 0 -1.609375 0.171875q-0.703125 0.171875 -1.21875 0.515625q-0.5 0.34375 -0.78125 0.875q-0.28125 0.515625 -0.28125 1.21875q0 0.53125 0.203125 0.984375q0.21875 0.453125 0.59375 0.78125q0.375 0.34375 0.921875 0.546875q0.546875 0.1875 1.21875 0.1875q0.40625 0 0.765625 -0.09375q0.375 -0.078125 0.703125 -0.234375q0.328125 -0.140625 0.59375 -0.328125q0.265625 -0.203125 0.484375 -0.421875q0.03125 0.265625 0.078125 0.515625q0.046875 0.234375 0.140625 0.40625zm-2.5625 -1.109375q-0.421875 0 -0.75 -0.109375q-0.3125 -0.109375 -0.515625 -0.296875q-0.21875 -0.1875 -0.328125 -0.453125q-0.09375 -0.265625 -0.09375 -0.578125q0 -0.328125 0.109375 -0.578125q0.125 -0.265625 0.359375 -0.453125q0.328125 -0.265625 0.890625 -0.390625q0.5625 -0.140625 1.3125 -0.140625l1.34375 0l0 1.71875q-0.125 0.25 -0.34375 0.484375q-0.21875 0.21875 -0.515625 0.40625q-0.3125 0.171875 -0.6875 0.28125q-0.359375 0.109375 -0.78125 0.109375zm6.8359375 -10.890625l0 1.25l2.875 0l0 9.5l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -10.75l-4.3125 0zm9.6015625 0l0 1.25l2.875 0l0 9.5l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -10.75l-4.3125 0zm8.9609375 7.6875l0 0.171875q0 0.921875 0.265625 1.703125q0.265625 0.78125 0.765625 1.359375q0.5 0.578125 1.203125 0.90625q0.71875 0.328125 1.625 0.328125q0.890625 0 1.59375 -0.328125q0.71875 -0.328125 1.21875 -0.90625q0.484375 -0.578125 0.75 -1.359375q0.265625 -0.78125 0.265625 -1.703125l0 -0.171875q0 -0.90625 -0.265625 -1.6875q-0.265625 -0.796875 -0.75 -1.375q-0.5 -0.578125 -1.21875 -0.90625q-0.71875 -0.328125 -1.609375 -0.328125q-0.890625 0 -1.609375 0.328125q-0.703125 0.328125 -1.203125 0.90625q-0.5 0.578125 -0.765625 1.375q-0.265625 0.78125 -0.265625 1.6875zm1.453125 0.171875l0 -0.171875q0 -0.609375 0.140625 -1.171875q0.15625 -0.578125 0.453125 -1.0q0.296875 -0.4375 0.734375 -0.6875q0.453125 -0.25 1.0625 -0.25q0.59375 0 1.046875 0.25q0.453125 0.25 0.75 0.6875q0.296875 0.421875 0.453125 1.0q0.15625 0.5625 0.15625 1.171875l0 0.171875q0 0.625 -0.15625 1.203125q-0.140625 0.5625 -0.4375 0.984375q-0.3125 0.4375 -0.765625 0.6875q-0.4375 0.25 -1.03125 0.25q-0.609375 0 -1.0625 -0.25q-0.453125 -0.25 -0.75 -0.6875q-0.296875 -0.421875 -0.453125 -0.984375q-0.140625 -0.578125 -0.140625 -1.203125zm12.1640625 3.125q-0.6875 0 -1.140625 -0.265625q-0.453125 -0.28125 -0.734375 -0.71875q-0.28125 -0.421875 -0.40625 -0.96875q-0.109375 -0.546875 -0.109375 -1.09375l0 -0.328125q0 -0.546875 0.109375 -1.078125q0.125 -0.546875 0.40625 -0.984375q0.28125 -0.4375 0.734375 -0.703125q0.46875 -0.265625 1.140625 -0.265625q0.4375 0 0.8125 0.15625q0.375 0.140625 0.65625 0.390625q0.265625 0.265625 0.421875 0.609375q0.171875 0.328125 0.171875 0.71875l1.375 0q0 -0.640625 -0.265625 -1.203125q-0.25 -0.5625 -0.703125 -0.984375q-0.46875 -0.40625 -1.09375 -0.640625q-0.625 -0.234375 -1.375 -0.234375q-0.96875 0 -1.6875 0.34375q-0.71875 0.34375 -1.1875 0.921875q-0.484375 0.578125 -0.734375 1.34375q-0.234375 0.765625 -0.234375 1.609375l0 0.328125q0 0.84375 0.234375 1.609375q0.25 0.75 0.734375 1.328125q0.46875 0.59375 1.1875 0.9375q0.71875 0.34375 1.6875 0.34375q0.671875 0 1.28125 -0.234375q0.625 -0.234375 1.09375 -0.625q0.484375 -0.375 0.765625 -0.890625q0.28125 -0.515625 0.296875 -1.078125l-1.375 0q0 0.359375 -0.1875 0.65625q-0.171875 0.296875 -0.453125 0.515625q-0.296875 0.234375 -0.671875 0.359375q-0.359375 0.125 -0.75 0.125zm17.03125 -3.609375l0 0.078125q0 1.125 0.171875 2.109375q0.171875 0.984375 0.46875 1.8125q0.296875 0.859375 0.671875 1.546875q0.390625 0.6875 0.828125 1.234375q0.421875 0.546875 0.84375 0.921875q0.4375 0.375 0.828125 0.59375l0.3125 -0.875q-0.328125 -0.265625 -0.640625 -0.625q-0.3125 -0.359375 -0.609375 -0.828125q-0.296875 -0.5 -0.5625 -1.125q-0.265625 -0.625 -0.46875 -1.34375q-0.1875 -0.71875 -0.296875 -1.5625q-0.109375 -0.859375 -0.109375 -1.84375l0 -0.109375q0 -1.015625 0.109375 -1.890625q0.125 -0.890625 0.34375 -1.625q0.21875 -0.8125 0.546875 -1.484375q0.328125 -0.6875 0.6875 -1.203125q0.234375 -0.328125 0.484375 -0.59375q0.265625 -0.265625 0.515625 -0.453125l-0.3125 -0.953125q-0.390625 0.21875 -0.828125 0.609375q-0.421875 0.375 -0.84375 0.921875q-0.421875 0.53125 -0.8125 1.234375q-0.390625 0.6875 -0.6875 1.53125q-0.296875 0.84375 -0.46875 1.828125q-0.171875 0.984375 -0.171875 2.09375zm7.1484375 -4.703125l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm8.5234375 -6.984375l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm13.6171875 8.453125l0 -11.375l-0.109375 0l-4.3125 1.65625l0 1.3125l2.984375 -1.125l0 9.53125l1.4375 0zm8.5859375 -0.34375l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm14.515625 -8.984375l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm8.5234375 -6.984375l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm15.9140625 8.453125l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm15.125 -0.9375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm0 -6.84375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm16.96875 -0.671875l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm12.6640625 8.609375q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm6.7265625 -1.03125l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm14.4765625 -0.15625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm8.2109375 7.421875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm14.039032 -1.53125l0 -1.53125l-4.78125 -1.84375l-0.40625 -0.140625l0.40625 -0.140625l4.78125 -1.875l0 -1.53125l-6.625 2.953125l0 1.140625l6.625 2.96875zm5.6484375 -1.671875l1.4375 0q0 -0.3125 0.015625 -0.546875q0.015625 -0.234375 0.078125 -0.453125q0.0625 -0.203125 0.1875 -0.40625q0.140625 -0.203125 0.375 -0.4375q0.359375 -0.328125 0.734375 -0.703125q0.390625 -0.390625 0.6875 -0.828125q0.3125 -0.4375 0.5 -0.9375q0.203125 -0.5 0.203125 -1.0625q0 -0.703125 -0.234375 -1.25q-0.21875 -0.546875 -0.65625 -0.921875q-0.421875 -0.390625 -1.046875 -0.578125q-0.625 -0.203125 -1.40625 -0.203125q-0.71875 0 -1.328125 0.1875q-0.609375 0.1875 -1.0625 0.546875q-0.453125 0.375 -0.71875 0.90625q-0.265625 0.515625 -0.265625 1.171875l1.4375 0q0 -0.40625 0.15625 -0.703125q0.171875 -0.296875 0.453125 -0.484375q0.265625 -0.1875 0.609375 -0.28125q0.34375 -0.09375 0.71875 -0.09375q0.46875 0 0.828125 0.125q0.375 0.125 0.625 0.375q0.21875 0.21875 0.328125 0.53125q0.109375 0.3125 0.109375 0.71875q0 0.390625 -0.15625 0.75q-0.140625 0.34375 -0.359375 0.65625q-0.234375 0.328125 -0.515625 0.625q-0.28125 0.28125 -0.546875 0.546875q-0.390625 0.359375 -0.625 0.625q-0.234375 0.265625 -0.359375 0.546875q-0.125 0.28125 -0.171875 0.65625q-0.03125 0.359375 -0.03125 0.921875zm-0.171875 2.4375q0 0.359375 0.21875 0.609375q0.234375 0.25 0.671875 0.25q0.453125 0 0.671875 -0.25q0.234375 -0.25 0.234375 -0.609375q0 -0.375 -0.234375 -0.625q-0.21875 -0.25 -0.671875 -0.25q-0.4375 0 -0.671875 0.25q-0.21875 0.25 -0.21875 0.625zm10.6328125 -4.5625l-2.234375 -3.125l-1.671875 0l3.140625 4.171875l-3.234375 4.28125l1.6875 0l2.34375 -3.21875l2.328125 3.21875l1.671875 0l-3.234375 -4.28125l3.15625 -4.171875l-1.703125 0l-2.25 3.125zm8.2578125 5.328125l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm9.1328125 -6.390625l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm14.8984375 6.390625l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm2.6484375 -1.546875l6.90625 -2.953125l0 -1.140625l-6.90625 -2.953125l0 1.484375l5.0625 1.875l0.46875 0.15625l-0.46875 0.125l-5.0625 1.90625l0 1.5zm13.2421875 1.203125l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm16.203125 -8.109375l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm12.6640625 8.609375q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm6.7265625 -1.03125l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm14.4765625 -0.15625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm8.2109375 7.421875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm14.0390625 -1.53125l0 -1.53125l-4.78125 -1.84375l-0.40625 -0.140625l0.40625 -0.140625l4.78125 -1.875l0 -1.53125l-6.625 2.953125l0 1.140625l6.625 2.96875zm4.6953125 -4.859375l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm11.4140625 1.0625l-2.234375 -3.125l-1.671875 0l3.140625 4.171875l-3.234375 4.28125l1.6875 0l2.34375 -3.21875l2.328125 3.21875l1.671875 0l-3.234375 -4.28125l3.15625 -4.171875l-1.703125 0l-2.25 3.125zm8.2578125 5.328125l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm9.1328125 -6.390625l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm14.8984375 6.390625l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm2.6484375 -1.546875l6.90625 -2.953125l0 -1.140625l-6.90625 -2.953125l0 1.484375l5.0625 1.875l0.46875 0.15625l-0.46875 0.125l-5.0625 1.90625l0 1.5zm14.8515625 -3.0l0 -0.078125q0 -1.1875 -0.1875 -2.140625q-0.1875 -0.96875 -0.46875 -1.78125q-0.28125 -0.859375 -0.671875 -1.53125q-0.375 -0.6875 -0.8125 -1.234375q-0.421875 -0.546875 -0.859375 -0.921875q-0.4375 -0.390625 -0.828125 -0.609375l-0.3125 0.890625q0.265625 0.203125 0.53125 0.515625q0.28125 0.296875 0.53125 0.6875q0.34375 0.515625 0.65625 1.203125q0.328125 0.671875 0.546875 1.53125q0.15625 0.609375 0.28125 1.46875q0.140625 0.859375 0.140625 1.90625l0 0.109375q0 0.9375 -0.109375 1.75q-0.09375 0.8125 -0.265625 1.515625q-0.171875 0.765625 -0.453125 1.40625q-0.265625 0.640625 -0.546875 1.140625q-0.3125 0.5 -0.65625 0.875q-0.328125 0.390625 -0.65625 0.640625l0.3125 0.875q0.390625 -0.21875 0.8125 -0.59375q0.4375 -0.375 0.875 -0.921875q0.421875 -0.546875 0.796875 -1.234375q0.390625 -0.6875 0.6875 -1.546875q0.296875 -0.828125 0.46875 -1.8125q0.1875 -0.984375 0.1875 -2.109375z" fill-rule="nonzero"/><path fill="#434343" d="m310.71732 130.50372l0 1.25l2.875 0l0 5.953125l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -7.203125l-4.3125 0zm2.6875 -2.21875q0 0.34375 0.203125 0.578125q0.21875 0.234375 0.65625 0.234375q0.4375 0 0.65625 -0.234375q0.21875 -0.234375 0.21875 -0.578125q0 -0.203125 -0.0625 -0.35936737q-0.0625 -0.171875 -0.1875 -0.28125q-0.109375 -0.09375 -0.265625 -0.15625q-0.15625 -0.0625 -0.359375 -0.0625q-0.203125 0 -0.359375 0.0625q-0.15625 0.0625 -0.265625 0.15625q-0.125 0.125 -0.1875 0.296875q-0.046875 0.15624237 -0.046875 0.34374237zm8.8359375 10.671875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.35936737q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.07811737 0.75 0.18749237l0.171875 -1.1874924q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.0312424q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm18.484375 -4.625l0 0.078125q0 1.125 0.171875 2.109375q0.171875 0.984375 0.46875 1.8125q0.296875 0.859375 0.671875 1.546875q0.390625 0.6875 0.828125 1.234375q0.421875 0.546875 0.84375 0.921875q0.4375 0.375 0.828125 0.59375l0.3125 -0.875q-0.328125 -0.265625 -0.640625 -0.625q-0.3125 -0.359375 -0.609375 -0.828125q-0.296875 -0.5 -0.5625 -1.125q-0.265625 -0.625 -0.46875 -1.34375q-0.1875 -0.71875 -0.296875 -1.5625q-0.109375 -0.859375 -0.109375 -1.84375l0 -0.109375q0 -1.015625 0.109375 -1.890625q0.125 -0.890625 0.34375 -1.625q0.21875 -0.8125 0.546875 -1.484375q0.328125 -0.6875 0.6875 -1.203125q0.234375 -0.32811737 0.484375 -0.5937424q0.265625 -0.265625 0.515625 -0.453125l-0.3125 -0.953125q-0.390625 0.21875 -0.828125 0.609375q-0.421875 0.375 -0.84375 0.921875q-0.421875 0.5312424 -0.8125 1.2343674q-0.390625 0.6875 -0.6875 1.53125q-0.296875 0.84375 -0.46875 1.828125q-0.171875 0.984375 -0.171875 2.09375zm7.1484375 -4.703125l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.29686737 -0.65625 -0.46874237q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46874237q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm7.4453125 -2.84375l0 0.171875q0 0.921875 0.265625 1.703125q0.265625 0.78125 0.765625 1.359375q0.5 0.578125 1.203125 0.90625q0.71875 0.328125 1.625 0.328125q0.890625 0 1.59375 -0.328125q0.71875 -0.328125 1.21875 -0.90625q0.484375 -0.578125 0.75 -1.359375q0.265625 -0.78125 0.265625 -1.703125l0 -0.171875q0 -0.90625 -0.265625 -1.6875q-0.265625 -0.796875 -0.75 -1.375q-0.5 -0.578125 -1.21875 -0.90625q-0.71875 -0.328125 -1.609375 -0.328125q-0.890625 0 -1.609375 0.328125q-0.703125 0.328125 -1.203125 0.90625q-0.5 0.578125 -0.765625 1.375q-0.265625 0.78125 -0.265625 1.6875zm1.453125 0.171875l0 -0.171875q0 -0.609375 0.140625 -1.171875q0.15625 -0.578125 0.453125 -1.0q0.296875 -0.4375 0.734375 -0.6875q0.453125 -0.25 1.0625 -0.25q0.59375 0 1.046875 0.25q0.453125 0.25 0.75 0.6875q0.296875 0.421875 0.453125 1.0q0.15625 0.5625 0.15625 1.171875l0 0.171875q0 0.625 -0.15625 1.203125q-0.140625 0.5625 -0.4375 0.984375q-0.3125 0.4375 -0.765625 0.6875q-0.4375 0.25 -1.03125 0.25q-0.609375 0 -1.0625 -0.25q-0.453125 -0.25 -0.75 -0.6875q-0.296875 -0.421875 -0.453125 -0.984375q-0.140625 -0.578125 -0.140625 -1.203125zm9.3671875 4.140625l1.140625 0l1.3125 -5.1875l0.203125 -1.3125l0.21875 1.3125l1.328125 5.1875l1.140625 0l1.796875 -8.453125l-1.28125 0l-0.9375 5.390625l-0.21875 1.328125l-0.21875 -1.328125l-1.34375 -5.390625l-0.9375 0l-1.34375 5.390625l-0.21875 1.203125l-0.171875 -1.203125l-0.984375 -5.390625l-1.28125 0l1.796875 8.453125zm8.7890625 0l1.453125 0l0 -6.0625q0.140625 -0.28125 0.34375 -0.5q0.203125 -0.234375 0.4375 -0.40625q0.28125 -0.1875 0.625 -0.296875q0.34375 -0.125 0.75 -0.125q0.453125 0 0.8125 0.109375q0.359375 0.109375 0.609375 0.359375q0.234375 0.234375 0.359375 0.625q0.125 0.390625 0.125 0.953125l0 5.34375l1.453125 0l0 -5.375q0 -0.84375 -0.21875 -1.453125q-0.203125 -0.609375 -0.578125 -1.015625q-0.390625 -0.390625 -0.9375 -0.578125q-0.53125 -0.1875 -1.203125 -0.1875q-0.484375 0 -0.9375 0.140625q-0.4375 0.140625 -0.8125 0.390625q-0.25 0.15625 -0.46875 0.390625q-0.21875 0.21875 -0.40625 0.484375l-0.109375 -1.25l-1.296875 0l0 8.453125zm14.2890625 0l0 -11.374992l-0.109375 0l-4.3125 1.6562424l0 1.3125l2.984375 -1.125l0 9.53125l1.4375 0zm8.5859375 -0.34375l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm14.515625 -8.984375l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.29686737 -0.65625 -0.46874237q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46874237q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm7.4453125 -2.84375l0 0.171875q0 0.921875 0.265625 1.703125q0.265625 0.78125 0.765625 1.359375q0.5 0.578125 1.203125 0.90625q0.71875 0.328125 1.625 0.328125q0.890625 0 1.59375 -0.328125q0.71875 -0.328125 1.21875 -0.90625q0.484375 -0.578125 0.75 -1.359375q0.265625 -0.78125 0.265625 -1.703125l0 -0.171875q0 -0.90625 -0.265625 -1.6875q-0.265625 -0.796875 -0.75 -1.375q-0.5 -0.578125 -1.21875 -0.90625q-0.71875 -0.328125 -1.609375 -0.328125q-0.890625 0 -1.609375 0.328125q-0.703125 0.328125 -1.203125 0.90625q-0.5 0.578125 -0.765625 1.375q-0.265625 0.78125 -0.265625 1.6875zm1.453125 0.171875l0 -0.171875q0 -0.609375 0.140625 -1.171875q0.15625 -0.578125 0.453125 -1.0q0.296875 -0.4375 0.734375 -0.6875q0.453125 -0.25 1.0625 -0.25q0.59375 0 1.046875 0.25q0.453125 0.25 0.75 0.6875q0.296875 0.421875 0.453125 1.0q0.15625 0.5625 0.15625 1.171875l0 0.171875q0 0.625 -0.15625 1.203125q-0.140625 0.5625 -0.4375 0.984375q-0.3125 0.4375 -0.765625 0.6875q-0.4375 0.25 -1.03125 0.25q-0.609375 0 -1.0625 -0.25q-0.453125 -0.25 -0.75 -0.6875q-0.296875 -0.421875 -0.453125 -0.984375q-0.140625 -0.578125 -0.140625 -1.203125zm9.3671875 4.140625l1.140625 0l1.3125 -5.1875l0.203125 -1.3125l0.21875 1.3125l1.328125 5.1875l1.140625 0l1.796875 -8.453125l-1.28125 0l-0.9375 5.390625l-0.21875 1.328125l-0.21875 -1.328125l-1.34375 -5.390625l-0.9375 0l-1.34375 5.390625l-0.21875 1.203125l-0.171875 -1.203125l-0.984375 -5.390625l-1.28125 0l1.796875 8.453125zm8.7890625 0l1.453125 0l0 -6.0625q0.140625 -0.28125 0.34375 -0.5q0.203125 -0.234375 0.4375 -0.40625q0.28125 -0.1875 0.625 -0.296875q0.34375 -0.125 0.75 -0.125q0.453125 0 0.8125 0.109375q0.359375 0.109375 0.609375 0.359375q0.234375 0.234375 0.359375 0.625q0.125 0.390625 0.125 0.953125l0 5.34375l1.453125 0l0 -5.375q0 -0.84375 -0.21875 -1.453125q-0.203125 -0.609375 -0.578125 -1.015625q-0.390625 -0.390625 -0.9375 -0.578125q-0.53125 -0.1875 -1.203125 -0.1875q-0.484375 0 -0.9375 0.140625q-0.4375 0.140625 -0.8125 0.390625q-0.25 0.15625 -0.46875 0.390625q-0.21875 0.21875 -0.40625 0.484375l-0.109375 -1.25l-1.296875 0l0 8.453125zm16.585938 0l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.42186737 -1.09375 -0.6562424q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.7499924q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm7.8984375 -4.546875l0 -0.078125q0 -1.1875 -0.1875 -2.140625q-0.1875 -0.96875 -0.46875 -1.78125q-0.28125 -0.859375 -0.671875 -1.53125q-0.375 -0.6875 -0.8125 -1.2343674q-0.421875 -0.546875 -0.859375 -0.921875q-0.4375 -0.390625 -0.828125 -0.609375l-0.3125 0.890625q0.265625 0.203125 0.53125 0.515625q0.28125 0.296875 0.53125 0.6874924q0.34375 0.515625 0.65625 1.203125q0.328125 0.671875 0.546875 1.53125q0.15625 0.609375 0.28125 1.46875q0.140625 0.859375 0.140625 1.90625l0 0.109375q0 0.9375 -0.109375 1.75q-0.09375 0.8125 -0.265625 1.515625q-0.171875 0.765625 -0.453125 1.40625q-0.265625 0.640625 -0.546875 1.140625q-0.3125 0.5 -0.65625 0.875q-0.328125 0.390625 -0.65625 0.640625l0.3125 0.875q0.390625 -0.21875 0.8125 -0.59375q0.4375 -0.375 0.875 -0.921875q0.421875 -0.546875 0.796875 -1.234375q0.390625 -0.6875 0.6875 -1.546875q0.296875 -0.828125 0.46875 -1.8125q0.1875 -0.984375 0.1875 -2.109375z" fill-rule="nonzero"/><path fill="#434343" d="m277.62357 168.74747q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm9.4765625 -3.078125l-1.453125 0l0 2.046875l-2.21875 0l0 1.109375l2.21875 0l0 4.59375q0 0.78125 0.203125 1.34375q0.21875 0.546875 0.578125 0.890625q0.359375 0.34375 0.84375 0.515625q0.484375 0.15625 1.046875 0.15625q0.34375 0 0.671875 -0.03125q0.34375 -0.03125 0.65625 -0.09375q0.296875 -0.0625 0.546875 -0.140625q0.265625 -0.078125 0.453125 -0.1875l-0.203125 -1.03125q-0.140625 0.03125 -0.34375 0.078125q-0.203125 0.046875 -0.4375 0.078125q-0.25 0.046875 -0.5 0.078125q-0.25 0.015625 -0.484375 0.015625q-0.3125 0 -0.609375 -0.078125q-0.28125 -0.078125 -0.5 -0.265625q-0.21875 -0.1875 -0.34375 -0.5q-0.125 -0.328125 -0.125 -0.828125l0 -4.59375l3.21875 0l0 -1.109375l-3.21875 0l0 -2.046875zm11.7421875 10.5l1.5 0l0 -0.125q-0.140625 -0.359375 -0.21875 -0.828125q-0.078125 -0.46875 -0.078125 -0.890625l0 -3.9375q0 -0.703125 -0.265625 -1.234375q-0.25 -0.53125 -0.6875 -0.890625q-0.453125 -0.359375 -1.078125 -0.53125q-0.609375 -0.171875 -1.328125 -0.171875q-0.796875 0 -1.421875 0.21875q-0.609375 0.21875 -1.046875 0.5625q-0.4375 0.359375 -0.671875 0.8125q-0.21875 0.4375 -0.234375 0.90625l1.453125 0q0 -0.265625 0.125 -0.515625q0.140625 -0.25 0.375 -0.421875q0.234375 -0.1875 0.5625 -0.296875q0.34375 -0.109375 0.78125 -0.109375q0.453125 0 0.828125 0.125q0.375 0.109375 0.625 0.328125q0.265625 0.21875 0.40625 0.53125q0.140625 0.296875 0.140625 0.671875l0 0.671875l-1.578125 0q-0.890625 0 -1.609375 0.171875q-0.703125 0.171875 -1.21875 0.515625q-0.5 0.34375 -0.78125 0.875q-0.28125 0.515625 -0.28125 1.21875q0 0.53125 0.203125 0.984375q0.21875 0.453125 0.59375 0.78125q0.375 0.34375 0.921875 0.546875q0.546875 0.1875 1.21875 0.1875q0.40625 0 0.765625 -0.09375q0.375 -0.078125 0.703125 -0.234375q0.328125 -0.140625 0.59375 -0.328125q0.265625 -0.203125 0.484375 -0.421875q0.03125 0.265625 0.078125 0.515625q0.046875 0.234375 0.140625 0.40625zm-2.5625 -1.109375q-0.421875 0 -0.75 -0.109375q-0.3125 -0.109375 -0.515625 -0.296875q-0.21875 -0.1875 -0.328125 -0.453125q-0.09375 -0.265625 -0.09375 -0.578125q0 -0.328125 0.109375 -0.578125q0.125 -0.265625 0.359375 -0.453125q0.328125 -0.265625 0.890625 -0.390625q0.5625 -0.140625 1.3125 -0.140625l1.34375 0l0 1.71875q-0.125 0.25 -0.34375 0.484375q-0.21875 0.21875 -0.515625 0.40625q-0.3125 0.171875 -0.6875 0.28125q-0.359375 0.109375 -0.78125 0.109375zm6.8359375 -7.34375l0 1.25l2.875 0l0 5.953125l-2.875 0l0 1.25l7.078125 0l0 -1.25l-2.765625 0l0 -7.203125l-4.3125 0zm2.6875 -2.21875q0 0.34375 0.203125 0.578125q0.21875 0.234375 0.65625 0.234375q0.4375 0 0.65625 -0.234375q0.21875 -0.234375 0.21875 -0.578125q0 -0.203125 -0.0625 -0.359375q-0.0625 -0.171875 -0.1875 -0.28125q-0.109375 -0.09375 -0.265625 -0.15625q-0.15625 -0.0625 -0.359375 -0.0625q-0.203125 0 -0.359375 0.0625q-0.15625 0.0625 -0.265625 0.15625q-0.125 0.125 -0.1875 0.296875q-0.046875 0.15625 -0.046875 0.34375zm6.6796875 10.671875l1.453125 0l0 -6.0625q0.140625 -0.28125 0.34375 -0.5q0.203125 -0.234375 0.4375 -0.40625q0.28125 -0.1875 0.625 -0.296875q0.34375 -0.125 0.75 -0.125q0.453125 0 0.8125 0.109375q0.359375 0.109375 0.609375 0.359375q0.234375 0.234375 0.359375 0.625q0.125 0.390625 0.125 0.953125l0 5.34375l1.453125 0l0 -5.375q0 -0.84375 -0.21875 -1.453125q-0.203125 -0.609375 -0.578125 -1.015625q-0.390625 -0.390625 -0.9375 -0.578125q-0.53125 -0.1875 -1.203125 -0.1875q-0.484375 0 -0.9375 0.140625q-0.4375 0.140625 -0.8125 0.390625q-0.25 0.15625 -0.46875 0.390625q-0.21875 0.21875 -0.40625 0.484375l-0.109375 -1.25l-1.296875 0l0 8.453125zm20.640625 -4.625l0 0.078125q0 1.125 0.171875 2.109375q0.171875 0.984375 0.46875 1.8125q0.296875 0.859375 0.671875 1.546875q0.390625 0.6875 0.828125 1.234375q0.421875 0.546875 0.84375 0.921875q0.4375 0.375 0.828125 0.59375l0.3125 -0.875q-0.328125 -0.265625 -0.640625 -0.625q-0.3125 -0.359375 -0.609375 -0.828125q-0.296875 -0.5 -0.5625 -1.125q-0.265625 -0.625 -0.46875 -1.34375q-0.1875 -0.71875 -0.296875 -1.5625q-0.109375 -0.859375 -0.109375 -1.84375l0 -0.109375q0 -1.015625 0.109375 -1.890625q0.125 -0.890625 0.34375 -1.625q0.21875 -0.8125 0.546875 -1.484375q0.328125 -0.6875 0.6875 -1.203125q0.234375 -0.328125 0.484375 -0.59375q0.265625 -0.265625 0.515625 -0.453125l-0.3125 -0.953125q-0.390625 0.21875 -0.828125 0.609375q-0.421875 0.375 -0.84375 0.921875q-0.421875 0.53125 -0.8125 1.234375q-0.390625 0.6875 -0.6875 1.53125q-0.296875 0.84375 -0.46875 1.828125q-0.171875 0.984375 -0.171875 2.09375zm7.1484375 -4.703125l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm13.3984375 -7.140625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm8.7421875 8.609375l0 -11.375l-0.109375 0l-4.3125 1.65625l0 1.3125l2.984375 -1.125l0 9.53125l1.4375 0zm8.5859375 -0.34375l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm14.515625 -8.984375l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm13.3984375 -7.140625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm11.0390625 8.609375l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm6.2890625 -0.34375l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm14.515625 -8.984375l0 0.59375q0 0.453125 0.140625 0.859375q0.140625 0.390625 0.40625 0.6875q0.265625 0.296875 0.65625 0.484375q0.390625 0.171875 0.90625 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.484375q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.859375l0 -0.59375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.71875q-0.265625 -0.296875 -0.65625 -0.46875q-0.390625 -0.1875 -0.90625 -0.1875q-0.515625 0 -0.90625 0.1875q-0.390625 0.171875 -0.640625 0.46875q-0.265625 0.3125 -0.40625 0.71875q-0.140625 0.390625 -0.140625 0.84375zm1.078125 0.59375l0 -0.59375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.421875q0.125 -0.171875 0.3125 -0.265625q0.1875 -0.109375 0.46875 -0.109375q0.28125 0 0.46875 0.109375q0.1875 0.09375 0.3125 0.265625q0.125 0.1875 0.171875 0.421875q0.0625 0.21875 0.0625 0.46875l0 0.59375q0 0.25 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.21875 -0.046875 -0.46875zm3.625 6.09375l0 0.609375q0 0.4375 0.125 0.84375q0.140625 0.390625 0.40625 0.6875q0.265625 0.3125 0.65625 0.5q0.40625 0.171875 0.921875 0.171875q0.5 0 0.890625 -0.171875q0.390625 -0.1875 0.65625 -0.5q0.265625 -0.296875 0.390625 -0.6875q0.140625 -0.40625 0.140625 -0.84375l0 -0.609375q0 -0.453125 -0.140625 -0.84375q-0.125 -0.40625 -0.390625 -0.703125q-0.265625 -0.3125 -0.65625 -0.484375q-0.390625 -0.171875 -0.90625 -0.171875q-0.515625 0 -0.90625 0.171875q-0.390625 0.171875 -0.65625 0.484375q-0.265625 0.296875 -0.40625 0.703125q-0.125 0.390625 -0.125 0.84375zm1.078125 0.609375l0 -0.609375q0 -0.25 0.046875 -0.46875q0.0625 -0.234375 0.1875 -0.40625q0.109375 -0.171875 0.296875 -0.28125q0.203125 -0.109375 0.484375 -0.109375q0.265625 0 0.453125 0.109375q0.203125 0.109375 0.328125 0.28125q0.125 0.171875 0.171875 0.40625q0.0625 0.21875 0.0625 0.46875l0 0.609375q0 0.234375 -0.0625 0.46875q-0.046875 0.21875 -0.171875 0.390625q-0.109375 0.1875 -0.3125 0.296875q-0.1875 0.09375 -0.453125 0.09375q-0.28125 0 -0.484375 -0.09375q-0.1875 -0.109375 -0.3125 -0.296875q-0.125 -0.171875 -0.1875 -0.390625q-0.046875 -0.234375 -0.046875 -0.46875zm-3.015625 0.5625l4.4375 -8.109375l-0.875 -0.484375l-4.421875 8.109375l0.859375 0.484375zm13.3984375 -7.140625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm5.7421875 2.21875l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm20.421875 5.453125q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm0 -6.84375q0 0.453125 0.28125 0.78125q0.296875 0.3125 0.859375 0.3125q0.546875 0 0.84375 -0.3125q0.296875 -0.328125 0.296875 -0.78125q0 -0.46875 -0.296875 -0.796875q-0.28125 -0.328125 -0.84375 -0.328125q-0.578125 0 -0.859375 0.328125q-0.28125 0.328125 -0.28125 0.796875zm16.96875 -0.671875l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm12.664032 8.609375q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.32809448 0.8125 -0.32809448 1.828125l0 0.328125q0 0.890625 0.29684448 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm6.7265625 -1.03125l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm14.4765625 -0.15625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm8.2109375 7.421875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm14.0390625 -1.53125l0 -1.53125l-4.78125 -1.84375l-0.40625 -0.140625l0.40625 -0.140625l4.78125 -1.875l0 -1.53125l-6.625 2.953125l0 1.140625l6.625 2.96875zm3.5234375 -4.171875l1.15625 0.296875q0.203125 -0.171875 0.375 -0.296875q0.1875 -0.140625 0.40625 -0.234375q0.203125 -0.09375 0.46875 -0.140625q0.265625 -0.0625 0.625 -0.0625q0.546875 0 0.96875 0.203125q0.421875 0.1875 0.703125 0.53125q0.296875 0.34375 0.4375 0.8125q0.15625 0.453125 0.15625 0.984375q0 0.578125 -0.140625 1.0625q-0.125 0.46875 -0.390625 0.8125q-0.28125 0.34375 -0.6875 0.53125q-0.390625 0.171875 -0.90625 0.171875q-0.90625 0 -1.5 -0.5q-0.578125 -0.5 -0.703125 -1.46875l-1.375 0q0.078125 0.796875 0.375 1.390625q0.3125 0.59375 0.796875 0.984375q0.484375 0.390625 1.09375 0.59375q0.625 0.1875 1.3125 0.1875q0.90625 0 1.578125 -0.28125q0.671875 -0.296875 1.125 -0.796875q0.4375 -0.5 0.65625 -1.1875q0.21875 -0.703125 0.21875 -1.515625q0 -0.859375 -0.234375 -1.5625q-0.234375 -0.703125 -0.671875 -1.203125q-0.4375 -0.5 -1.0625 -0.765625q-0.625 -0.28125 -1.421875 -0.28125q-0.609375 0 -1.078125 0.15625q-0.46875 0.140625 -0.8125 0.34375l0.328125 -3.03125l4.609375 0l0 -1.40625l-5.828125 0l-0.578125 5.671875zm12.5859375 0.375l-2.234375 -3.125l-1.671875 0l3.140625 4.171875l-3.234375 4.28125l1.6875 0l2.34375 -3.21875l2.328125 3.21875l1.671875 0l-3.234375 -4.28125l3.15625 -4.171875l-1.703125 0l-2.25 3.125zm8.2578125 5.328125l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm12.7734375 -11.390625l-0.125 0q-1.21875 0 -2.1875 0.34375q-0.953125 0.34375 -1.625 1.015625q-0.96875 0.96875 -1.3125 2.34375q-0.328125 1.375 -0.328125 2.796875l0 0.6875q0 0.828125 0.25 1.65625q0.25 0.8125 0.75 1.4375q0.453125 0.5625 1.109375 0.921875q0.671875 0.34375 1.546875 0.34375q0.875 0 1.53125 -0.296875q0.65625 -0.3125 1.109375 -0.84375q0.4375 -0.53125 0.65625 -1.234375q0.234375 -0.703125 0.234375 -1.5q0 -0.6875 -0.1875 -1.359375q-0.1875 -0.671875 -0.59375 -1.203125q-0.390625 -0.53125 -1.015625 -0.859375q-0.609375 -0.328125 -1.46875 -0.328125q-0.390625 0 -0.75 0.09375q-0.34375 0.09375 -0.65625 0.234375q-0.3125 0.15625 -0.578125 0.375q-0.265625 0.203125 -0.46875 0.4375q0.046875 -0.75 0.296875 -1.453125q0.265625 -0.71875 0.75 -1.234375q0.46875 -0.484375 1.1875 -0.78125q0.71875 -0.296875 1.75 -0.296875l0.125 0l0 -1.296875zm-1.890625 5.125q0.53125 0 0.921875 0.21875q0.390625 0.21875 0.640625 0.59375q0.265625 0.359375 0.390625 0.84375q0.125 0.46875 0.125 0.96875q0 0.5625 -0.140625 1.046875q-0.140625 0.46875 -0.40625 0.8125q-0.265625 0.34375 -0.65625 0.546875q-0.390625 0.1875 -0.90625 0.1875q-0.4375 0 -0.84375 -0.203125q-0.390625 -0.203125 -0.6875 -0.609375q-0.3125 -0.390625 -0.5 -0.96875q-0.171875 -0.59375 -0.171875 -1.34375l0 -0.484375q0.140625 -0.359375 0.359375 -0.65625q0.234375 -0.296875 0.53125 -0.5q0.296875 -0.21875 0.640625 -0.328125q0.34375 -0.125 0.703125 -0.125zm12.0234375 2.4375l0 -7.546875l-1.53125 0l-5.09375 7.875l0 0.859375l5.1875 0l0 2.640625l1.4375 0l0 -2.640625l1.59375 0l0 -1.1875l-1.59375 0zm-5.03125 0l3.359375 -5.1875l0.234375 -0.4375l0 5.625l-3.59375 0zm8.8046875 2.28125l6.90625 -2.953125l0 -1.140625l-6.90625 -2.953125l0 1.484375l5.0625 1.875l0.46875 0.15625l-0.46875 0.125l-5.0625 1.90625l0 1.5zm13.2421875 1.203125l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm16.203125 -8.109375l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm12.6640625 8.609375q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm6.7265625 -1.03125l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm14.4765625 -0.15625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm8.2109375 7.421875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm14.0390625 -1.53125l0 -1.53125l-4.78125 -1.84375l-0.40625 -0.140625l0.40625 -0.140625l4.78125 -1.875l0 -1.53125l-6.625 2.953125l0 1.140625l6.625 2.96875zm5.6484375 -1.671875l1.4375 0q0 -0.3125 0.015625 -0.546875q0.015625 -0.234375 0.078125 -0.453125q0.0625 -0.203125 0.1875 -0.40625q0.140625 -0.203125 0.375 -0.4375q0.359375 -0.328125 0.734375 -0.703125q0.390625 -0.390625 0.6875 -0.828125q0.3125 -0.4375 0.5 -0.9375q0.203125 -0.5 0.203125 -1.0625q0 -0.703125 -0.234375 -1.25q-0.21875 -0.546875 -0.65625 -0.921875q-0.421875 -0.390625 -1.046875 -0.578125q-0.625 -0.203125 -1.40625 -0.203125q-0.71875 0 -1.328125 0.1875q-0.609375 0.1875 -1.0625 0.546875q-0.453125 0.375 -0.71875 0.90625q-0.265625 0.515625 -0.265625 1.171875l1.4375 0q0 -0.40625 0.15625 -0.703125q0.171875 -0.296875 0.453125 -0.484375q0.265625 -0.1875 0.609375 -0.28125q0.34375 -0.09375 0.71875 -0.09375q0.46875 0 0.828125 0.125q0.375 0.125 0.625 0.375q0.21875 0.21875 0.328125 0.53125q0.109375 0.3125 0.109375 0.71875q0 0.390625 -0.15625 0.75q-0.140625 0.34375 -0.359375 0.65625q-0.234375 0.328125 -0.515625 0.625q-0.28125 0.28125 -0.546875 0.546875q-0.390625 0.359375 -0.625 0.625q-0.234375 0.265625 -0.359375 0.546875q-0.125 0.28125 -0.171875 0.65625q-0.03125 0.359375 -0.03125 0.921875zm-0.171875 2.4375q0 0.359375 0.21875 0.609375q0.234375 0.25 0.671875 0.25q0.453125 0 0.671875 -0.25q0.234375 -0.25 0.234375 -0.609375q0 -0.375 -0.234375 -0.625q-0.21875 -0.25 -0.671875 -0.25q-0.4375 0 -0.671875 0.25q-0.21875 0.25 -0.21875 0.625zm10.6328125 -4.5625l-2.234375 -3.125l-1.671875 0l3.140625 4.171875l-3.234375 4.28125l1.6875 0l2.34375 -3.21875l2.328125 3.21875l1.671875 0l-3.234375 -4.28125l3.15625 -4.171875l-1.703125 0l-2.25 3.125zm8.2578125 5.328125l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm9.1328125 -6.390625l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm14.8984375 6.390625l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm2.6484375 -1.546875l6.90625 -2.953125l0 -1.140625l-6.90625 -2.953125l0 1.484375l5.0625 1.875l0.46875 0.15625l-0.46875 0.125l-5.0625 1.90625l0 1.5zm13.2421875 1.203125l0 -1.375l-1.578125 0l0 1.40625q0 0.671875 -0.15625 1.28125q-0.15625 0.625 -0.53125 1.171875l0.90625 0.484375q0.625 -0.546875 0.984375 -1.359375q0.375 -0.8125 0.375 -1.609375zm16.203125 -8.109375l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm12.6640625 8.609375q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm6.7265625 -1.03125l-1.296875 0l0 8.453125l1.375 0l0 -6.734375q0.046875 -0.125 0.109375 -0.234375q0.078125 -0.109375 0.1875 -0.1875q0.125 -0.109375 0.3125 -0.171875q0.203125 -0.0625 0.453125 -0.0625q0.234375 0 0.40625 0.0625q0.1875 0.0625 0.296875 0.1875q0.125 0.125 0.1875 0.328125q0.0625 0.203125 0.0625 0.46875l0 6.34375l1.375 0l0 -6.328125q0 -0.046875 0 -0.078125q0 -0.046875 0 -0.078125q0.015625 -0.203125 0.078125 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.3125 -0.1875q0.1875 -0.078125 0.453125 -0.078125q0.234375 0 0.421875 0.0625q0.1875 0.0625 0.3125 0.1875q0.109375 0.140625 0.171875 0.34375q0.0625 0.203125 0.0625 0.46875l0 6.328125l1.375 0l0 -6.3125q0 -0.609375 -0.140625 -1.0625q-0.140625 -0.453125 -0.40625 -0.734375q-0.234375 -0.25 -0.59375 -0.375q-0.34375 -0.125 -0.78125 -0.125q-0.328125 0 -0.59375 0.078125q-0.265625 0.078125 -0.484375 0.21875q-0.171875 0.109375 -0.328125 0.265625q-0.140625 0.15625 -0.25 0.34375q-0.09375 -0.203125 -0.21875 -0.359375q-0.109375 -0.15625 -0.265625 -0.265625q-0.203125 -0.140625 -0.46875 -0.203125q-0.265625 -0.078125 -0.609375 -0.078125q-0.59375 0 -1.015625 0.25q-0.40625 0.25 -0.671875 0.703125l-0.03125 -0.796875zm14.4765625 -0.15625q-0.921875 0 -1.65625 0.40625q-0.71875 0.390625 -1.234375 1.09375l0 -0.21875l-0.0625 -1.125l-1.375 0l0 8.453125l1.453125 0l0 -5.421875q0.140625 -0.390625 0.34375 -0.703125q0.21875 -0.328125 0.515625 -0.546875q0.3125 -0.25 0.734375 -0.375q0.4375 -0.125 0.96875 -0.125q0.421875 0 0.796875 0.046875q0.375 0.03125 0.796875 0.125l0.203125 -1.40625q-0.21875 -0.09375 -0.65625 -0.140625q-0.4375 -0.0625 -0.828125 -0.0625zm7.7890625 8.765625q1.234375 0 2.078125 -0.5q0.84375 -0.5 1.265625 -1.140625l-0.890625 -0.6875q-0.390625 0.515625 -1.0 0.828125q-0.59375 0.3125 -1.375 0.3125q-0.578125 0 -1.0625 -0.21875q-0.484375 -0.21875 -0.828125 -0.59375q-0.34375 -0.359375 -0.53125 -0.828125q-0.1875 -0.46875 -0.25 -1.09375l0 -0.046875l6.046875 0l0 -0.65625q0 -0.875 -0.234375 -1.625q-0.21875 -0.765625 -0.671875 -1.328125q-0.453125 -0.5625 -1.140625 -0.875q-0.671875 -0.3125 -1.59375 -0.3125q-0.71875 0 -1.421875 0.296875q-0.6875 0.296875 -1.21875 0.84375q-0.546875 0.5625 -0.875 1.375q-0.328125 0.8125 -0.328125 1.828125l0 0.328125q0 0.890625 0.296875 1.640625q0.296875 0.75 0.828125 1.296875q0.53125 0.546875 1.265625 0.859375q0.75 0.296875 1.640625 0.296875zm-0.1875 -7.578125q0.5625 0 0.953125 0.203125q0.40625 0.203125 0.671875 0.53125q0.265625 0.328125 0.40625 0.78125q0.15625 0.4375 0.15625 0.828125l0 0.078125l-4.546875 0q0.09375 -0.59375 0.296875 -1.03125q0.21875 -0.453125 0.53125 -0.765625q0.3125 -0.3125 0.703125 -0.46875q0.390625 -0.15625 0.828125 -0.15625zm8.2109375 7.421875l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm14.0390625 -1.53125l0 -1.53125l-4.78125 -1.84375l-0.40625 -0.140625l0.40625 -0.140625l4.78125 -1.875l0 -1.53125l-6.625 2.953125l0 1.140625l6.625 2.96875zm5.1640625 1.53125l1.453125 0l0 -7.34375l3.265625 0l0 -1.109375l-3.265625 0l0 -0.59375q0 -0.53125 0.140625 -0.921875q0.140625 -0.40625 0.40625 -0.65625q0.25 -0.25 0.625 -0.359375q0.375 -0.125 0.875 -0.125q0.484375 0 0.90625 0.078125q0.421875 0.078125 0.75 0.1875l0.171875 -1.1875q-0.203125 -0.0625 -0.40625 -0.09375q-0.1875 -0.046875 -0.390625 -0.09375q-0.296875 -0.046875 -0.609375 -0.078125q-0.296875 -0.046875 -0.609375 -0.046875q-0.75 0 -1.359375 0.21875q-0.609375 0.203125 -1.03125 0.625q-0.4375 0.40625 -0.6875 1.03125q-0.234375 0.609375 -0.234375 1.421875l0 0.59375l-2.328125 0l0 1.109375l2.328125 0l0 7.34375zm9.1328125 -6.390625l0 1.171875l1.03125 0q0.546875 0 0.984375 0.125q0.453125 0.125 0.78125 0.390625q0.3125 0.25 0.484375 0.640625q0.171875 0.390625 0.171875 0.921875q0 0.53125 -0.15625 0.9375q-0.140625 0.390625 -0.421875 0.65625q-0.28125 0.265625 -0.6875 0.390625q-0.390625 0.125 -0.890625 0.125q-0.484375 0 -0.890625 -0.140625q-0.40625 -0.140625 -0.6875 -0.390625q-0.28125 -0.265625 -0.4375 -0.625q-0.140625 -0.375 -0.140625 -0.8125l-1.453125 0q0 0.75 0.28125 1.34375q0.296875 0.578125 0.78125 0.984375q0.484375 0.40625 1.140625 0.625q0.65625 0.203125 1.40625 0.203125q0.765625 0 1.421875 -0.21875q0.671875 -0.234375 1.15625 -0.65625q0.484375 -0.421875 0.75 -1.03125q0.28125 -0.625 0.28125 -1.421875q0 -0.390625 -0.09375 -0.78125q-0.09375 -0.40625 -0.3125 -0.765625q-0.21875 -0.359375 -0.59375 -0.640625q-0.375 -0.296875 -0.921875 -0.484375q0.453125 -0.1875 0.78125 -0.484375q0.34375 -0.296875 0.546875 -0.625q0.21875 -0.328125 0.3125 -0.671875q0.109375 -0.359375 0.109375 -0.6875q0 -0.796875 -0.25 -1.390625q-0.25 -0.609375 -0.703125 -1.015625q-0.453125 -0.40625 -1.09375 -0.609375q-0.625 -0.203125 -1.390625 -0.203125q-0.796875 0 -1.4375 0.25q-0.640625 0.234375 -1.109375 0.65625q-0.453125 0.4375 -0.71875 1.015625q-0.25 0.5625 -0.25 1.21875l1.453125 0q0 -0.453125 0.140625 -0.8125q0.15625 -0.359375 0.4375 -0.609375q0.265625 -0.265625 0.640625 -0.390625q0.375 -0.140625 0.84375 -0.140625q0.46875 0 0.84375 0.125q0.375 0.109375 0.625 0.359375q0.25 0.25 0.375 0.625q0.140625 0.375 0.140625 0.890625q0 0.421875 -0.15625 0.78125q-0.140625 0.359375 -0.421875 0.625q-0.28125 0.25 -0.703125 0.40625q-0.421875 0.140625 -0.96875 0.140625l-1.03125 0zm14.8984375 6.390625l0 -1.1875l-5.71875 0l3.0625 -3.3125q0.421875 -0.46875 0.8125 -0.9375q0.390625 -0.46875 0.6875 -0.953125q0.28125 -0.484375 0.453125 -0.984375q0.171875 -0.5 0.171875 -1.0q0 -0.6875 -0.234375 -1.265625q-0.234375 -0.578125 -0.6875 -1.0q-0.4375 -0.421875 -1.09375 -0.65625q-0.640625 -0.234375 -1.4375 -0.234375q-0.875 0 -1.5625 0.28125q-0.6875 0.28125 -1.15625 0.75q-0.46875 0.484375 -0.71875 1.125q-0.25 0.625 -0.25 1.34375l1.453125 0q0 -0.53125 0.125 -0.953125q0.140625 -0.4375 0.421875 -0.734375q0.28125 -0.296875 0.6875 -0.453125q0.421875 -0.171875 1.0 -0.171875q0.46875 0 0.828125 0.15625q0.375 0.15625 0.640625 0.4375q0.265625 0.265625 0.390625 0.640625q0.140625 0.375 0.140625 0.796875q0 0.34375 -0.09375 0.671875q-0.078125 0.3125 -0.28125 0.671875q-0.203125 0.375 -0.546875 0.8125q-0.328125 0.421875 -0.828125 0.96875l-3.71875 4.140625l0 1.046875l7.453125 0zm2.6484375 -1.546875l6.90625 -2.953125l0 -1.140625l-6.90625 -2.953125l0 1.484375l5.0625 1.875l0.46875 0.15625l-0.46875 0.125l-5.0625 1.90625l0 1.5zm14.8515625 -3.0l0 -0.078125q0 -1.1875 -0.1875 -2.140625q-0.1875 -0.96875 -0.46875 -1.78125q-0.28125 -0.859375 -0.671875 -1.53125q-0.375 -0.6875 -0.8125 -1.234375q-0.421875 -0.546875 -0.859375 -0.921875q-0.4375 -0.390625 -0.828125 -0.609375l-0.3125 0.890625q0.265625 0.203125 0.53125 0.515625q0.28125 0.296875 0.53125 0.6875q0.34375 0.515625 0.65625 1.203125q0.328125 0.671875 0.546875 1.53125q0.15625 0.609375 0.28125 1.46875q0.140625 0.859375 0.140625 1.90625l0 0.109375q0 0.9375 -0.109375 1.75q-0.09375 0.8125 -0.265625 1.515625q-0.171875 0.765625 -0.453125 1.40625q-0.265625 0.640625 -0.546875 1.140625q-0.3125 0.5 -0.65625 0.875q-0.328125 0.390625 -0.65625 0.640625l0.3125 0.875q0.390625 -0.21875 0.8125 -0.59375q0.4375 -0.375 0.875 -0.921875q0.421875 -0.546875 0.796875 -1.234375q0.390625 -0.6875 0.6875 -1.546875q0.296875 -0.828125 0.46875 -1.8125q0.1875 -0.984375 0.1875 -2.109375z" fill-rule="nonzero"/><path fill="#010000" fill-opacity="0.0" d="m904.4204 160.59122l0 0l0 21.101562l0 0l0 -21.101562z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m780.34906 96.173225l-29.574768 0" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m780.34906 96.173225l-17.574768 0" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m762.7743 92.86977l-9.076233 3.3034592l9.076233 3.3034668z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m780.34906 134.0l-29.574768 0" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m780.34906 134.0l-17.574768 0" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m762.7743 130.69653l-9.076233 3.3034668l9.076233 3.3034668z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m780.45667 96.06562l0 37.76377" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m780.45667 96.06562l0 37.76377" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m763.56433 209.29659l-16.944885 -21.291336" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m763.56433 209.29659l-9.47229 -11.901962" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m756.6768 195.3375l-8.236694 -5.044525l3.0671387 9.158768z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m783.22235 97.589485l153.92126 0l0 35.55906l-153.92126 0z" fill-rule="evenodd"/><path fill="#595959" d="m792.6286 117.32699l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm11.6484375 1.203125q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9906006 3.46875l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm15.836792 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm10.0217285 4.28125l0.5625 -2.765625l-1.09375 0l0 -0.96875l1.296875 0l0.46875 -2.359375l-1.765625 0l0 -0.96875l1.96875 0l0.5625 -2.796875l0.984375 0l-0.578125 2.796875l2.0625 0l0.5625 -2.796875l0.984375 0l-0.5625 2.796875l1.125 0l0 0.96875l-1.328125 0l-0.484375 2.359375l1.8125 0l0 0.96875l-2.015625 0l-0.5625 2.765625l-0.96875 0l0.546875 -2.765625l-2.03125 0l-0.578125 2.765625l-0.96875 0zm1.734375 -3.734375l2.046875 0l0.484375 -2.359375l-2.046875 0l-0.484375 2.359375zm9.1467285 0.125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.9281006 3.453125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm6.687195 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.6468506 6.109375l0 -9.5625l1.078125 0l0 0.890625q0.375 -0.53125 0.84375 -0.78125q0.484375 -0.265625 1.15625 -0.265625q0.875 0 1.546875 0.453125q0.6875 0.453125 1.03125 1.28125q0.34375 0.828125 0.34375 1.828125q0 1.046875 -0.375 1.90625q-0.375 0.84375 -1.109375 1.296875q-0.71875 0.453125 -1.53125 0.453125q-0.578125 0 -1.046875 -0.25q-0.46875 -0.25 -0.765625 -0.625l0 3.375l-1.171875 0zm1.0625 -6.078125q0 1.34375 0.53125 1.984375q0.546875 0.625 1.3125 0.625q0.78125 0 1.34375 -0.65625q0.5625 -0.65625 0.5625 -2.046875q0 -1.3125 -0.546875 -1.96875q-0.546875 -0.671875 -1.296875 -0.671875q-0.75 0 -1.328125 0.703125q-0.578125 0.703125 -0.578125 2.03125zm11.084351 1.203125l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5062256 4.125l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm8.969666 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9906006 3.46875l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm11.896851 0l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm6.1624756 1.390625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m750.7744 203.44882l176.09448 0l0 60.125977l-176.09448 0z" fill-rule="evenodd"/><path fill="#595959" d="m765.16504 224.38945q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm5.5321045 0l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm6.3499756 -4.6875l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm5.5079346 -1.046875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1248779 1.046875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm8.9696045 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm4.4071655 2.65625l-0.125 -1.09375q0.375 0.109375 0.65625 0.109375q0.390625 0 0.625 -0.140625q0.234375 -0.125 0.390625 -0.359375q0.109375 -0.171875 0.359375 -0.875q0.03125 -0.09375 0.109375 -0.28125l-2.625 -6.921875l1.265625 0l1.4375 4.0q0.28125 0.765625 0.5 1.59375q0.203125 -0.796875 0.46875 -1.578125l1.484375 -4.015625l1.171875 0l-2.625 7.015625q-0.421875 1.140625 -0.65625 1.578125q-0.3125 0.578125 -0.71875 0.84375q-0.40625 0.28125 -0.96875 0.28125q-0.328125 0 -0.75 -0.15625zm10.210754 -2.5l0.5625 -2.765625l-1.09375 0l0 -0.96875l1.296875 0l0.46875 -2.359375l-1.765625 0l0 -0.96875l1.96875 0l0.5625 -2.796875l0.984375 0l-0.578125 2.796875l2.0625 0l0.5625 -2.796875l0.984375 0l-0.5625 2.796875l1.125 0l0 0.96875l-1.328125 0l-0.484375 2.359375l1.8125 0l0 0.96875l-2.015625 0l-0.5625 2.765625l-0.96875 0l0.546875 -2.765625l-2.03125 0l-0.578125 2.765625l-0.96875 0zm1.734375 -3.734375l2.046875 0l0.484375 -2.359375l-2.046875 0l-0.484375 2.359375zm9.1467285 0.125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.9281006 3.453125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm7.234131 0l0 -9.546875l1.90625 0l2.25 6.765625q0.3125 0.9375 0.46875 1.40625q0.15625 -0.515625 0.5 -1.53125l2.28125 -6.640625l1.703125 0l0 9.546875l-1.21875 0l0 -7.984375l-2.765625 7.984375l-1.140625 0l-2.765625 -8.125l0 8.125l-1.21875 0zm15.727417 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 4.125l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm11.274292 0l0 -9.546875l4.234375 0q1.265625 0 1.921875 0.265625q0.671875 0.25 1.0625 0.90625q0.40625 0.65625 0.40625 1.4375q0 1.015625 -0.65625 1.71875q-0.65625 0.6875 -2.03125 0.875q0.5 0.25 0.765625 0.484375q0.546875 0.5 1.046875 1.265625l1.65625 2.59375l-1.578125 0l-1.265625 -1.984375q-0.5625 -0.859375 -0.921875 -1.3125q-0.34375 -0.453125 -0.640625 -0.640625q-0.28125 -0.1875 -0.5625 -0.25q-0.21875 -0.046875 -0.703125 -0.046875l-1.46875 0l0 4.234375l-1.265625 0zm1.265625 -5.328125l2.71875 0q0.859375 0 1.34375 -0.171875q0.484375 -0.1875 0.734375 -0.578125q0.265625 -0.390625 0.265625 -0.859375q0 -0.671875 -0.5 -1.109375q-0.484375 -0.4375 -1.546875 -0.4375l-3.015625 0l0 3.15625zm12.922058 3.109375l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.8031006 4.125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm2.952942 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm7.4453125 2.0625l0 -1.328125l1.328125 0l0 1.328125q0 0.734375 -0.265625 1.1875q-0.25 0.453125 -0.8125 0.703125l-0.328125 -0.5q0.375 -0.171875 0.546875 -0.484375q0.171875 -0.3125 0.1875 -0.90625l-0.65625 0z" fill-rule="nonzero"/><path fill="#595959" d="m760.7588 241.24883l0 -9.546875l1.90625 0l2.25 6.765625q0.3125 0.9375 0.46875 1.40625q0.15625 -0.515625 0.5 -1.53125l2.28125 -6.640625l1.703125 0l0 9.546875l-1.21875 0l0 -7.984375l-2.765625 7.984375l-1.140625 0l-2.765625 -8.125l0 8.125l-1.21875 0zm15.727417 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 4.125l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm11.086792 0l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.1883545 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.8031006 4.125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm9.687256 -1.046875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.0936279 3.703125l-0.125 -1.09375q0.375 0.109375 0.65625 0.109375q0.390625 0 0.625 -0.140625q0.234375 -0.125 0.390625 -0.359375q0.109375 -0.171875 0.359375 -0.875q0.03125 -0.09375 0.109375 -0.28125l-2.625 -6.921875l1.265625 0l1.4375 4.0q0.28125 0.765625 0.5 1.59375q0.203125 -0.796875 0.46875 -1.578125l1.484375 -4.015625l1.171875 0l-2.625 7.015625q-0.421875 1.140625 -0.65625 1.578125q-0.3125 0.578125 -0.71875 0.84375q-0.40625 0.28125 -0.96875 0.28125q-0.328125 0 -0.75 -0.15625zm6.7109375 0l0 -9.5625l1.078125 0l0 0.890625q0.375 -0.53125 0.84375 -0.78125q0.484375 -0.265625 1.15625 -0.265625q0.875 0 1.546875 0.453125q0.6875 0.453125 1.03125 1.28125q0.34375 0.828125 0.34375 1.828125q0 1.046875 -0.375 1.90625q-0.375 0.84375 -1.109375 1.296875q-0.71875 0.453125 -1.53125 0.453125q-0.578125 0 -1.046875 -0.25q-0.46875 -0.25 -0.765625 -0.625l0 3.375l-1.171875 0zm1.0625 -6.078125q0 1.34375 0.53125 1.984375q0.546875 0.625 1.3125 0.625q0.78125 0 1.34375 -0.65625q0.5625 -0.65625 0.5625 -2.046875q0 -1.3125 -0.546875 -1.96875q-0.546875 -0.671875 -1.296875 -0.671875q-0.75 0 -1.328125 0.703125q-0.578125 0.703125 -0.578125 2.03125zm11.084351 1.203125l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.0531006 2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm15.320129 2.0625l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm6.1937256 0q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm10.3498535 3.453125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm6.9749756 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm9.209351 2.40625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm4.843506 1.046875l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm11.928101 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm4.9124756 3.46875l-2.625 -6.90625l1.234375 0l1.484375 4.140625q0.234375 0.65625 0.4375 1.390625q0.15625 -0.546875 0.4375 -1.3125l1.53125 -4.21875l1.21875 0l-2.625 6.90625l-1.09375 0zm9.4765625 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375z" fill-rule="nonzero"/><path fill="#595959" d="m763.2119 256.20193l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.3906097l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.14060974 0.203125 0.23435974q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm0.7029419 -2.4062347q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.76560974 -1.15625 1.1874847q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.93748474 -0.890625 -2.6874847zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65623474 1.46875 0.65623474q0.875 0 1.453125 -0.65623474q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm10.3498535 3.4531097l0 -6.9062347l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.7499847l-1.171875 0l0 -4.3593597q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.0156097l-1.171875 0l0 -4.4843597q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.5781097l-1.171875 0zm15.618042 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4374847q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0624847q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.6093597q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78123474q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.29685974 0.78125 -0.81248474q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm5.5531006 2.4218597l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.3906097l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.14060974 0.203125 0.23435974q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm5.656128 -1.4843597l1.15625 0.15625q-0.1875 1.1874847 -0.96875 1.8593597q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.93748474 -0.859375 -2.6562347q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.62498474 1.40625 0.62498474q0.6875 0 1.140625 -0.421875q0.46875 -0.42185974 0.59375 -1.2968597zm2.1484375 2.5312347l0 -9.54686l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.3749847l-1.171875 0l0 -4.3749847q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.7812347l-1.171875 0z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m426.29398 44.62992l-35.433075 35.527557" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m426.29398 44.62992l-26.959137 27.03099" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m396.99585 69.32812l-4.070282 8.759171l8.748322 -4.09359z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m426.29398 18.677166l500.56693 0l0 51.905514l-500.56693 0z" fill-rule="evenodd"/><path fill="#595959" d="m442.59085 40.477165l-1.703125 0l0 -1.015625q-0.421875 0.59375 -1.0 0.890625q-0.578125 0.28125 -1.15625 0.28125q-1.1875 0 -2.046875 -0.953125q-0.84375 -0.96875 -0.84375 -2.6875q0 -1.75 0.828125 -2.65625q0.828125 -0.921875 2.09375 -0.921875q1.15625 0 2.0 0.953125l0 -3.4375l1.828125 0l0 9.546875zm-4.875 -3.609375q0 1.109375 0.296875 1.609375q0.4375 0.71875 1.234375 0.71875q0.640625 0 1.078125 -0.53125q0.4375 -0.546875 0.4375 -1.609375q0 -1.203125 -0.4375 -1.71875q-0.421875 -0.53125 -1.09375 -0.53125q-0.640625 0 -1.078125 0.515625q-0.4375 0.515625 -0.4375 1.546875zm10.672607 1.40625l1.828125 0.3125q-0.34375 1.0 -1.109375 1.53125q-0.765625 0.515625 -1.890625 0.515625q-1.8125 0 -2.671875 -1.171875q-0.6875 -0.953125 -0.6875 -2.390625q0 -1.71875 0.890625 -2.6875q0.90625 -0.96875 2.28125 -0.96875q1.546875 0 2.4375 1.015625q0.890625 1.015625 0.84375 3.125l-4.578125 0q0.015625 0.8125 0.4375 1.265625q0.421875 0.453125 1.0625 0.453125q0.421875 0 0.71875 -0.234375q0.296875 -0.234375 0.4375 -0.765625zm0.109375 -1.84375q-0.015625 -0.796875 -0.40625 -1.203125q-0.390625 -0.421875 -0.953125 -0.421875q-0.59375 0 -0.984375 0.4375q-0.390625 0.4375 -0.390625 1.1875l2.734375 0zm4.6781006 -0.75l-1.671875 -0.3125q0.28125 -1.0 0.96875 -1.46875q0.6875 -0.484375 2.03125 -0.484375q1.21875 0 1.8125 0.296875q0.609375 0.28125 0.84375 0.734375q0.25 0.4375 0.25 1.625l-0.015625 2.140625q0 0.90625 0.078125 1.34375q0.09375 0.421875 0.34375 0.921875l-1.8125 0q-0.078125 -0.1875 -0.171875 -0.546875q-0.046875 -0.15625 -0.078125 -0.203125q-0.46875 0.453125 -1.0 0.6875q-0.53125 0.21875 -1.140625 0.21875q-1.0625 0 -1.6875 -0.578125q-0.609375 -0.578125 -0.609375 -1.46875q0 -0.578125 0.28125 -1.03125q0.28125 -0.46875 0.78125 -0.703125q0.515625 -0.25 1.46875 -0.4375q1.28125 -0.234375 1.765625 -0.453125l0 -0.171875q0 -0.53125 -0.265625 -0.75q-0.25 -0.234375 -0.96875 -0.234375q-0.484375 0 -0.765625 0.203125q-0.265625 0.1875 -0.4375 0.671875zm2.4375 1.484375q-0.34375 0.109375 -1.109375 0.28125q-0.765625 0.15625 -1.0 0.3125q-0.359375 0.25 -0.359375 0.640625q0 0.390625 0.28125 0.671875q0.296875 0.28125 0.734375 0.28125q0.5 0 0.953125 -0.328125q0.328125 -0.25 0.4375 -0.609375q0.0625 -0.234375 0.0625 -0.890625l0 -0.359375zm3.5999756 3.3125l0 -9.546875l1.828125 0l0 9.546875l-1.828125 0zm3.7029724 0l0 -9.546875l1.828125 0l0 9.546875l-1.828125 0zm3.281128 -3.546875q0 -0.921875 0.453125 -1.765625q0.453125 -0.859375 1.265625 -1.296875q0.828125 -0.453125 1.84375 -0.453125q1.5625 0 2.5625 1.015625q1.015625 1.015625 1.015625 2.578125q0 1.5625 -1.015625 2.59375q-1.015625 1.03125 -2.546875 1.03125q-0.953125 0 -1.8125 -0.421875q-0.859375 -0.4375 -1.3125 -1.265625q-0.453125 -0.828125 -0.453125 -2.015625zm1.875 0.09375q0 1.03125 0.484375 1.578125q0.5 0.546875 1.203125 0.546875q0.71875 0 1.203125 -0.546875q0.484375 -0.546875 0.484375 -1.59375q0 -1.015625 -0.484375 -1.5625q-0.484375 -0.546875 -1.203125 -0.546875q-0.703125 0 -1.203125 0.546875q-0.484375 0.546875 -0.484375 1.578125zm12.719482 -1.421875l-1.796875 0.328125q-0.09375 -0.53125 -0.421875 -0.796875q-0.328125 -0.28125 -0.828125 -0.28125q-0.6875 0 -1.09375 0.46875q-0.40625 0.46875 -0.40625 1.578125q0 1.234375 0.40625 1.75q0.421875 0.5 1.109375 0.5q0.53125 0 0.859375 -0.296875q0.328125 -0.296875 0.46875 -1.03125l1.796875 0.3125q-0.28125 1.234375 -1.078125 1.875q-0.796875 0.625 -2.125 0.625q-1.515625 0 -2.421875 -0.953125q-0.90625 -0.953125 -0.90625 -2.65625q0 -1.703125 0.90625 -2.65625q0.90625 -0.953125 2.453125 -0.953125q1.265625 0 2.0 0.546875q0.75 0.53125 1.078125 1.640625zm5.084198 4.875l0 -9.546875l1.828125 0l0 9.546875l-1.828125 0zm3.703003 -7.84375l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm3.0623474 -1.96875l1.828125 -0.28125q0.125 0.53125 0.484375 0.8125q0.359375 0.265625 1.0 0.265625q0.703125 0 1.0625 -0.25q0.25 -0.1875 0.25 -0.484375q0 -0.21875 -0.140625 -0.359375q-0.125 -0.125 -0.609375 -0.234375q-2.203125 -0.484375 -2.796875 -0.890625q-0.828125 -0.5625 -0.828125 -1.5625q0 -0.890625 0.703125 -1.5q0.71875 -0.609375 2.203125 -0.609375q1.421875 0 2.109375 0.46875q0.703125 0.453125 0.953125 1.359375l-1.71875 0.3125q-0.109375 -0.40625 -0.421875 -0.609375q-0.3125 -0.21875 -0.890625 -0.21875q-0.71875 0 -1.03125 0.203125q-0.203125 0.140625 -0.203125 0.359375q0 0.203125 0.1875 0.34375q0.234375 0.171875 1.6875 0.515625q1.46875 0.328125 2.0625 0.8125q0.5625 0.484375 0.5625 1.359375q0 0.953125 -0.796875 1.640625q-0.78125 0.671875 -2.34375 0.671875q-1.40625 0 -2.234375 -0.5625q-0.828125 -0.578125 -1.078125 -1.5625zm11.224976 -4.9375l0 1.453125l-1.25 0l0 2.78125q0 0.84375 0.03125 0.984375q0.046875 0.140625 0.171875 0.234375q0.125 0.09375 0.3125 0.09375q0.25 0 0.734375 -0.171875l0.15625 1.421875q-0.640625 0.265625 -1.453125 0.265625q-0.484375 0 -0.890625 -0.15625q-0.390625 -0.171875 -0.578125 -0.4375q-0.1875 -0.265625 -0.265625 -0.71875q-0.046875 -0.3125 -0.046875 -1.28125l0 -3.015625l-0.84375 0l0 -1.453125l0.84375 0l0 -1.375l1.828125 -1.078125l0 2.453125l1.25 0z" fill-rule="nonzero"/><path fill="#595959" d="m505.38303 34.89904l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm0 5.578125l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm9.41507 0l0 -8.421875l-3.140625 0l0 -1.125l7.5625 0l0 1.125l-3.15625 0l0 8.421875l-1.265625 0zm5.5632324 0l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.0531006 2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm11.8671875 -0.15625l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm14.7404785 3.265625q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.1883545 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm14.7404785 3.265625q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.9610596 0l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm7.7578735 0l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm10.865601 3.5l0 -1.015625q-0.8125 1.171875 -2.1875 1.171875q-0.609375 0 -1.140625 -0.234375q-0.53125 -0.234375 -0.796875 -0.578125q-0.25 -0.359375 -0.359375 -0.875q-0.0625 -0.34375 -0.0625 -1.09375l0 -4.28125l1.171875 0l0 3.828125q0 0.921875 0.0625 1.234375q0.109375 0.46875 0.46875 0.734375q0.359375 0.25 0.890625 0.25q0.515625 0 0.984375 -0.265625q0.46875 -0.265625 0.65625 -0.734375q0.1875 -0.46875 0.1875 -1.34375l0 -3.703125l1.171875 0l0 6.90625l-1.046875 0zm3.1624756 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm3.4620972 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm8.156067 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5062256 4.125l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm3.9852905 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm13.398254 1.015625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1405029 1.046875l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm11.928101 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm5.5531006 2.421875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm4.843506 1.046875l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm15.618042 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9437256 6.125l-0.125 -1.09375q0.375 0.109375 0.65625 0.109375q0.390625 0 0.625 -0.140625q0.234375 -0.125 0.390625 -0.359375q0.109375 -0.171875 0.359375 -0.875q0.03125 -0.09375 0.109375 -0.28125l-2.625 -6.921875l1.265625 0l1.4375 4.0q0.28125 0.765625 0.5 1.59375q0.203125 -0.796875 0.46875 -1.578125l1.484375 -4.015625l1.171875 0l-2.625 7.015625q-0.421875 1.140625 -0.65625 1.578125q-0.3125 0.578125 -0.71875 0.84375q-0.40625 0.28125 -0.96875 0.28125q-0.328125 0 -0.75 -0.15625zm10.413879 -2.65625l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm11.928101 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm4.9124756 3.46875l-2.625 -6.90625l1.234375 0l1.484375 4.140625q0.234375 0.65625 0.4375 1.390625q0.15625 -0.546875 0.4375 -1.3125l1.53125 -4.21875l1.21875 0l-2.625 6.90625l-1.09375 0zm9.4765625 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm12.7873535 3.078125l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm0.7030029 -2.40625q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm11.4279785 3.453125l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm11.068726 1.28125l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm14.7091675 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm11.365601 1.234375l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.037476 3.265625q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.9611206 0l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.5391846 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm11.162476 0.921875l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm6.6640625 1.671875q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm5.5531006 2.421875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm5.874878 -1.171875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.006226 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm6.9593506 3.453125l0 -1.328125l1.34375 0l0 1.328125l-1.34375 0zm7.171631 0l0 -9.546875l3.59375 0q1.09375 0 1.75 0.296875q0.65625 0.28125 1.03125 0.890625q0.375 0.609375 0.375 1.265625q0 0.609375 -0.34375 1.15625q-0.328125 0.53125 -0.984375 0.859375q0.859375 0.25 1.328125 0.875q0.46875 0.609375 0.46875 1.4375q0 0.671875 -0.296875 1.25q-0.28125 0.578125 -0.703125 0.890625q-0.40625 0.3125 -1.03125 0.46875q-0.625 0.15625 -1.546875 0.15625l-3.640625 0zm1.265625 -5.53125l2.0625 0q0.84375 0 1.203125 -0.109375q0.484375 -0.140625 0.71875 -0.46875q0.25 -0.34375 0.25 -0.84375q0 -0.46875 -0.234375 -0.828125q-0.21875 -0.359375 -0.640625 -0.5q-0.421875 -0.140625 -1.453125 -0.140625l-1.90625 0l0 2.890625zm0 4.40625l2.375 0q0.609375 0 0.859375 -0.046875q0.4375 -0.078125 0.734375 -0.25q0.296875 -0.1875 0.484375 -0.53125q0.1875 -0.359375 0.1875 -0.8125q0 -0.53125 -0.28125 -0.921875q-0.265625 -0.40625 -0.75 -0.5625q-0.484375 -0.15625 -1.40625 -0.15625l-2.203125 0l0 3.28125zm7.4835205 3.78125l-0.125 -1.09375q0.375 0.109375 0.65625 0.109375q0.390625 0 0.625 -0.140625q0.234375 -0.125 0.390625 -0.359375q0.109375 -0.171875 0.359375 -0.875q0.03125 -0.09375 0.109375 -0.28125l-2.625 -6.921875l1.265625 0l1.4375 4.0q0.28125 0.765625 0.5 1.59375q0.203125 -0.796875 0.46875 -1.578125l1.484375 -4.015625l1.171875 0l-2.625 7.015625q-0.421875 1.140625 -0.65625 1.578125q-0.3125 0.578125 -0.71875 0.84375q-0.40625 0.28125 -0.96875 0.28125q-0.328125 0 -0.75 -0.15625z" fill-rule="nonzero"/><path fill="#595959" d="m440.6846 53.945915l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm1.7109375 -0.921875q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.6468506 3.453125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm6.9437256 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm9.6953125 1.015625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1248474 1.046875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm8.98526 0l0 -1.015625q-0.8125 1.171875 -2.1875 1.171875q-0.609375 0 -1.140625 -0.234375q-0.53125 -0.234375 -0.796875 -0.578125q-0.25 -0.359375 -0.359375 -0.875q-0.0625 -0.34375 -0.0625 -1.09375l0 -4.28125l1.171875 0l0 3.828125q0 0.921875 0.0625 1.234375q0.109375 0.46875 0.46875 0.734375q0.359375 0.25 0.890625 0.25q0.515625 0 0.984375 -0.265625q0.46875 -0.265625 0.65625 -0.734375q0.1875 -0.46875 0.1875 -1.34375l0 -3.703125l1.171875 0l0 6.90625l-1.046875 0zm7.3968506 -2.53125l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm4.7109375 1.484375l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1560974 -7.140625l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.507965 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.6468506 3.453125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm7.7249756 0l0 -1.328125l1.328125 0l0 1.328125q0 0.734375 -0.265625 1.1875q-0.25 0.453125 -0.8125 0.703125l-0.328125 -0.5q0.375 -0.171875 0.546875 -0.484375q0.171875 -0.3125 0.1875 -0.90625l-0.65625 0zm7.0934753 0l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm15.633667 0l0 -1.015625q-0.8125 1.171875 -2.1875 1.171875q-0.609375 0 -1.140625 -0.234375q-0.53125 -0.234375 -0.796875 -0.578125q-0.25 -0.359375 -0.359375 -0.875q-0.0625 -0.34375 -0.0625 -1.09375l0 -4.28125l1.171875 0l0 3.828125q0 0.921875 0.0625 1.234375q0.109375 0.46875 0.46875 0.734375q0.359375 0.25 0.890625 0.25q0.515625 0 0.984375 -0.265625q0.46875 -0.265625 0.65625 -0.734375q0.1875 -0.46875 0.1875 -1.34375l0 -3.703125l1.171875 0l0 6.90625l-1.046875 0zm2.4124756 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm9.6953125 1.015625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm9.359131 -1.484375l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm1.7109375 -0.921875q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.6468506 3.453125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm9.974976 -1.046875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm5.656128 0.1875q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm3.0062256 -4.71875l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.9454346 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm15.6311035 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.9610596 0l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm6.6797485 0l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm15.836792 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 4.125l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm11.086792 0l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.1883545 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.8031006 4.125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm2.953003 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm13.398315 1.015625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1405029 1.046875l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm11.92804 -0.859375q-0.65618896 0.5625 -1.265564 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.093689 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.45306396 -0.390625 -1.343689 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.437439 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.64056396 0.265625 -1.921814 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53118896 -0.296875 0.78118896 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm5.5531006 2.421875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm7.406006 0l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1405029 1.046875l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm14.9592285 1.90625l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 4.125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm11.928101 -2.53125l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm2.1328125 2.53125l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.5392456 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.1781006 1.390625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm7.1484375 -6.125l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.9454346 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm7.1937256 0.578125l1.140625 0.15625q0.078125 0.53125 0.40625 0.78125q0.4375 0.3125 1.1875 0.3125q0.8125 0 1.25 -0.328125q0.453125 -0.3125 0.609375 -0.90625q0.09375 -0.359375 0.078125 -1.5q-0.765625 0.90625 -1.90625 0.90625q-1.4375 0 -2.21875 -1.03125q-0.78125 -1.03125 -0.78125 -2.46875q0 -0.984375 0.359375 -1.8125q0.359375 -0.84375 1.03125 -1.296875q0.6875 -0.453125 1.609375 -0.453125q1.21875 0 2.015625 0.984375l0 -0.828125l1.078125 0l0 5.96875q0 1.609375 -0.328125 2.28125q-0.328125 0.6875 -1.046875 1.078125q-0.703125 0.390625 -1.75 0.390625q-1.234375 0 -2.0 -0.5625q-0.75 -0.5625 -0.734375 -1.671875zm0.984375 -4.15625q0 1.359375 0.53125 1.984375q0.546875 0.625 1.359375 0.625q0.796875 0 1.34375 -0.625q0.546875 -0.625 0.546875 -1.953125q0 -1.265625 -0.5625 -1.90625q-0.5625 -0.640625 -1.359375 -0.640625q-0.765625 0 -1.3125 0.640625q-0.546875 0.625 -0.546875 1.875zm11.4279785 3.578125l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm10.849976 2.640625q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.5218506 1.40625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm7.1484375 -6.125l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm7.4611206 -2.53125l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm6.9295044 2.53125l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm6.3187256 3.5l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.5392456 -3.453125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm11.162476 0.921875l1.15625 0.15625q-0.1875 1.1875 -0.96875 1.859375q-0.78125 0.671875 -1.921875 0.671875q-1.40625 0 -2.28125 -0.921875q-0.859375 -0.9375 -0.859375 -2.65625q0 -1.125 0.375 -1.96875q0.375 -0.84375 1.125 -1.25q0.765625 -0.421875 1.65625 -0.421875q1.125 0 1.84375 0.578125q0.71875 0.5625 0.921875 1.609375l-1.140625 0.171875q-0.171875 -0.703125 -0.59375 -1.046875q-0.40625 -0.359375 -0.984375 -0.359375q-0.890625 0 -1.453125 0.640625q-0.546875 0.640625 -0.546875 2.0q0 1.40625 0.53125 2.03125q0.546875 0.625 1.40625 0.625q0.6875 0 1.140625 -0.421875q0.46875 -0.421875 0.59375 -1.296875zm2.1640625 2.53125l0 -9.546875l1.171875 0l0 5.453125l2.765625 -2.8125l1.515625 0l-2.640625 2.5625l2.90625 4.34375l-1.4375 0l-2.28125 -3.53125l-0.828125 0.796875l0 2.734375l-1.171875 0zm10.351379 0l0 -6.90625l1.046875 0l0 0.96875q0.328125 -0.515625 0.859375 -0.8125q0.546875 -0.3125 1.234375 -0.3125q0.78125 0 1.265625 0.3125q0.484375 0.3125 0.6875 0.890625q0.828125 -1.203125 2.140625 -1.203125q1.03125 0 1.578125 0.578125q0.5625 0.5625 0.5625 1.734375l0 4.75l-1.171875 0l0 -4.359375q0 -0.703125 -0.125 -1.0q-0.109375 -0.3125 -0.40625 -0.5q-0.296875 -0.1875 -0.703125 -0.1875q-0.71875 0 -1.203125 0.484375q-0.484375 0.484375 -0.484375 1.546875l0 4.015625l-1.171875 0l0 -4.484375q0 -0.78125 -0.296875 -1.171875q-0.28125 -0.390625 -0.921875 -0.390625q-0.5 0 -0.921875 0.265625q-0.421875 0.25 -0.609375 0.75q-0.1875 0.5 -0.1875 1.453125l0 3.578125l-1.171875 0zm15.618042 -0.859375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9437256 6.125l-0.125 -1.09375q0.375 0.109375 0.65625 0.109375q0.390625 0 0.625 -0.140625q0.234375 -0.125 0.390625 -0.359375q0.109375 -0.171875 0.359375 -0.875q0.03125 -0.09375 0.109375 -0.28125l-2.625 -6.921875l1.265625 0l1.4375 4.0q0.28125 0.765625 0.5 1.59375q0.203125 -0.796875 0.46875 -1.578125l1.484375 -4.015625l1.171875 0l-2.625 7.015625q-0.421875 1.140625 -0.65625 1.578125q-0.3125 0.578125 -0.71875 0.84375q-0.40625 0.28125 -0.96875 0.28125q-0.328125 0 -0.75 -0.15625zm9.97644 -6.109375q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm7.9281006 3.453125l-2.125 -6.90625l1.21875 0l1.09375 3.984375l0.421875 1.484375q0.015625 -0.109375 0.359375 -1.421875l1.09375 -4.046875l1.203125 0l1.03125 4.0l0.34375 1.328125l0.40625 -1.34375l1.171875 -3.984375l1.140625 0l-2.15625 6.90625l-1.21875 0l-1.09375 -4.140625l-0.265625 -1.171875l-1.40625 5.3125l-1.21875 0zm8.343933 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm7.7406006 0l0 -1.328125l1.34375 0l0 1.328125l-1.34375 0z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m40.750656 221.25984l277.57483 0l0 35.559067l-277.57483 0z" fill-rule="evenodd"/><path fill="#595959" d="m50.281906 239.51297q0 -0.921875 0.453125 -1.765625q0.453125 -0.859375 1.265625 -1.296875q0.828125 -0.453125 1.84375 -0.453125q1.5625 0 2.5625 1.015625q1.015625 1.015625 1.015625 2.578125q0 1.5625 -1.015625 2.59375q-1.015625 1.03125 -2.546875 1.03125q-0.953125 0 -1.8125 -0.421875q-0.859375 -0.4375 -1.3125 -1.265625q-0.453125 -0.828125 -0.453125 -2.015625zm1.875 0.09375q0 1.03125 0.484375 1.578125q0.5 0.546875 1.203125 0.546875q0.71875 0 1.203125 -0.546875q0.484375 -0.546875 0.484375 -1.59375q0 -1.015625 -0.484375 -1.5625q-0.484375 -0.546875 -1.203125 -0.546875q-0.703125 0 -1.203125 0.546875q-0.484375 0.546875 -0.484375 1.578125zm7.9851074 3.453125l-2.1875 -6.90625l1.765625 0l1.296875 4.53125l1.203125 -4.53125l1.75 0l1.15625 4.53125l1.328125 -4.53125l1.796875 0l-2.21875 6.90625l-1.75 0l-1.203125 -4.4375l-1.171875 4.4375l-1.765625 0zm15.36705 0l-1.828125 0l0 -3.53125q0 -1.109375 -0.125 -1.4375q-0.109375 -0.328125 -0.375 -0.515625q-0.265625 -0.1875 -0.640625 -0.1875q-0.46875 0 -0.859375 0.265625q-0.375 0.265625 -0.515625 0.703125q-0.140625 0.421875 -0.140625 1.578125l0 3.125l-1.828125 0l0 -6.90625l1.703125 0l0 1.015625q0.90625 -1.171875 2.28125 -1.171875q0.609375 0 1.109375 0.21875q0.5 0.21875 0.75 0.5625q0.265625 0.328125 0.359375 0.765625q0.109375 0.421875 0.109375 1.21875l0 4.296875zm5.8444824 -2.203125l1.828125 0.3125q-0.34375 1.0 -1.109375 1.53125q-0.765625 0.515625 -1.890625 0.515625q-1.8125 0 -2.671875 -1.171875q-0.6875 -0.953125 -0.6875 -2.390625q0 -1.71875 0.890625 -2.6875q0.90625 -0.96875 2.28125 -0.96875q1.546875 0 2.4375 1.015625q0.890625 1.015625 0.84375 3.125l-4.578125 0q0.015625 0.8125 0.4375 1.265625q0.421875 0.453125 1.0625 0.453125q0.421875 0 0.71875 -0.234375q0.296875 -0.234375 0.4375 -0.765625zm0.109375 -1.84375q-0.015625 -0.796875 -0.40625 -1.203125q-0.390625 -0.421875 -0.953125 -0.421875q-0.59375 0 -0.984375 0.4375q-0.390625 0.4375 -0.390625 1.1875l2.734375 0zm5.0531006 4.046875l-1.828125 0l0 -6.90625l1.703125 0l0 0.984375q0.4375 -0.703125 0.78125 -0.921875q0.34375 -0.21875 0.796875 -0.21875q0.625 0 1.203125 0.34375l-0.5625 1.59375q-0.46875 -0.296875 -0.859375 -0.296875q-0.390625 0 -0.65625 0.21875q-0.265625 0.203125 -0.421875 0.765625q-0.15625 0.546875 -0.15625 2.296875l0 2.140625zm2.7961578 -1.96875l1.828125 -0.28125q0.125 0.53125 0.484375 0.8125q0.359375 0.265625 1.0 0.265625q0.703125 0 1.0625 -0.25q0.25 -0.1875 0.25 -0.484375q0 -0.21875 -0.140625 -0.359375q-0.125 -0.125 -0.609375 -0.234375q-2.203125 -0.484375 -2.796875 -0.890625q-0.828125 -0.5625 -0.828125 -1.5625q0 -0.890625 0.703125 -1.5q0.71875 -0.609375 2.203125 -0.609375q1.421875 0 2.109375 0.46875q0.703125 0.453125 0.953125 1.359375l-1.71875 0.3125q-0.109375 -0.40625 -0.421875 -0.609375q-0.3125 -0.21875 -0.890625 -0.21875q-0.71875 0 -1.03125 0.203125q-0.203125 0.140625 -0.203125 0.359375q0 0.203125 0.1875 0.34375q0.234375 0.171875 1.6875 0.515625q1.46875 0.328125 2.0625 0.8125q0.5625 0.484375 0.5625 1.359375q0 0.953125 -0.796875 1.640625q-0.78125 0.671875 -2.34375 0.671875q-1.40625 0 -2.234375 -0.5625q-0.828125 -0.578125 -1.078125 -1.5625zm9.881226 -7.578125l0 3.515625q0.875 -1.03125 2.109375 -1.03125q0.640625 0 1.140625 0.234375q0.515625 0.234375 0.765625 0.59375q0.25 0.359375 0.34375 0.8125q0.109375 0.4375 0.109375 1.375l0 4.046875l-1.828125 0l0 -3.65625q0 -1.078125 -0.109375 -1.375q-0.109375 -0.296875 -0.375 -0.46875q-0.265625 -0.171875 -0.65625 -0.171875q-0.453125 0 -0.8125 0.234375q-0.359375 0.21875 -0.53125 0.671875q-0.15625 0.4375 -0.15625 1.296875l0 3.46875l-1.828125 0l0 -9.546875l1.828125 0zm6.3132324 1.703125l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm3.6561127 -6.90625l1.703125 0l0 1.015625q0.328125 -0.53125 0.890625 -0.84375q0.578125 -0.328125 1.265625 -0.328125q1.203125 0 2.03125 0.9375q0.84375 0.9375 0.84375 2.625q0 1.734375 -0.84375 2.703125q-0.84375 0.953125 -2.046875 0.953125q-0.578125 0 -1.046875 -0.21875q-0.453125 -0.234375 -0.96875 -0.796875l0 3.484375l-1.828125 0l0 -9.53125zm1.8125 3.328125q0 1.171875 0.453125 1.734375q0.46875 0.546875 1.125 0.546875q0.640625 0 1.0625 -0.5q0.421875 -0.515625 0.421875 -1.6875q0 -1.078125 -0.4375 -1.609375q-0.4375 -0.53125 -1.078125 -0.53125q-0.671875 0 -1.109375 0.515625q-0.4375 0.515625 -0.4375 1.53125zm10.07872 -4.265625l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm9.999863 0l-1.828125 0l0 -3.53125q0 -1.109375 -0.125 -1.4375q-0.109375 -0.328125 -0.375 -0.515625q-0.265625 -0.1875 -0.640625 -0.1875q-0.46875 0 -0.859375 0.265625q-0.375 0.265625 -0.515625 0.703125q-0.140625 0.421875 -0.140625 1.578125l0 3.125l-1.828125 0l0 -6.90625l1.703125 0l0 1.015625q0.90625 -1.171875 2.28125 -1.171875q0.609375 0 1.109375 0.21875q0.5 0.21875 0.75 0.5625q0.265625 0.328125 0.359375 0.765625q0.109375 0.421875 0.109375 1.21875l0 4.296875zm8.188232 0l-1.703125 0l0 -1.015625q-0.421875 0.59375 -1.0 0.890625q-0.578125 0.28125 -1.15625 0.28125q-1.1875 0 -2.046875 -0.953125q-0.84375 -0.96875 -0.84375 -2.6875q0 -1.75 0.828125 -2.65625q0.828125 -0.921875 2.09375 -0.921875q1.15625 0 2.0 0.953125l0 -3.4375l1.828125 0l0 9.546875zm-4.875 -3.609375q0 1.109375 0.296875 1.609375q0.4375 0.71875 1.234375 0.71875q0.640625 0 1.078125 -0.53125q0.4375 -0.546875 0.4375 -1.609375q0 -1.203125 -0.4375 -1.71875q-0.421875 -0.53125 -1.09375 -0.53125q-0.640625 0 -1.078125 0.515625q-0.4375 0.515625 -0.4375 1.546875zm6.6726074 -4.234375l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm9.734238 -4.875l-1.796875 0.328125q-0.09375 -0.53125 -0.421875 -0.796875q-0.328125 -0.28125 -0.828125 -0.28125q-0.6875 0 -1.09375 0.46875q-0.40625 0.46875 -0.40625 1.578125q0 1.234375 0.40625 1.75q0.421875 0.5 1.109375 0.5q0.53125 0 0.859375 -0.296875q0.328125 -0.296875 0.46875 -1.03125l1.796875 0.3125q-0.28125 1.234375 -1.078125 1.875q-0.796875 0.625 -2.125 0.625q-1.515625 0 -2.421875 -0.953125q-0.90625 -0.953125 -0.90625 -2.65625q0 -1.703125 0.90625 -2.65625q0.90625 -0.953125 2.453125 -0.953125q1.265625 0 2.0 0.546875q0.75 0.53125 1.078125 1.640625zm2.7562256 0.078125l-1.671875 -0.3125q0.28125 -1.0 0.96875 -1.46875q0.6875 -0.484375 2.03125 -0.484375q1.21875 0 1.8125 0.296875q0.609375 0.28125 0.84375 0.734375q0.25 0.4375 0.25 1.625l-0.015625 2.140625q0 0.90625 0.078125 1.34375q0.09375 0.421875 0.34375 0.921875l-1.8125 0q-0.078125 -0.1875 -0.171875 -0.546875q-0.046875 -0.15625 -0.078125 -0.203125q-0.46875 0.453125 -1.0 0.6875q-0.53125 0.21875 -1.140625 0.21875q-1.0625 0 -1.6875 -0.578125q-0.609375 -0.578125 -0.609375 -1.46875q0 -0.578125 0.28125 -1.03125q0.28125 -0.46875 0.78125 -0.703125q0.515625 -0.25 1.46875 -0.4375q1.28125 -0.234375 1.765625 -0.453125l0 -0.171875q0 -0.53125 -0.265625 -0.75q-0.25 -0.234375 -0.96875 -0.234375q-0.484375 0 -0.765625 0.203125q-0.265625 0.1875 -0.4375 0.671875zm2.4375 1.484375q-0.34375 0.109375 -1.109375 0.28125q-0.765625 0.15625 -1.0 0.3125q-0.359375 0.25 -0.359375 0.640625q0 0.390625 0.28125 0.671875q0.296875 0.28125 0.734375 0.28125q0.5 0 0.953125 -0.328125q0.328125 -0.25 0.4375 -0.609375q0.0625 -0.234375 0.0625 -0.890625l0 -0.359375zm6.7718506 -3.59375l0 1.453125l-1.25 0l0 2.78125q0 0.84375 0.03125 0.984375q0.046875 0.140625 0.171875 0.234375q0.125 0.09375 0.3125 0.09375q0.25 0 0.734375 -0.171875l0.15625 1.421875q-0.640625 0.265625 -1.453125 0.265625q-0.484375 0 -0.890625 -0.15625q-0.390625 -0.171875 -0.578125 -0.4375q-0.1875 -0.265625 -0.265625 -0.71875q-0.046875 -0.3125 -0.046875 -1.28125l0 -3.015625l-0.84375 0l0 -1.453125l0.84375 0l0 -1.375l1.828125 -1.078125l0 2.453125l1.25 0zm0.84461975 3.359375q0 -0.921875 0.453125 -1.765625q0.453125 -0.859375 1.265625 -1.296875q0.828125 -0.453125 1.84375 -0.453125q1.5625 0 2.5625 1.015625q1.015625 1.015625 1.015625 2.578125q0 1.5625 -1.015625 2.59375q-1.015625 1.03125 -2.546875 1.03125q-0.953125 0 -1.8125 -0.421875q-0.859375 -0.4375 -1.3125 -1.265625q-0.453125 -0.828125 -0.453125 -2.015625zm1.875 0.09375q0 1.03125 0.484375 1.578125q0.5 0.546875 1.203125 0.546875q0.71875 0 1.203125 -0.546875q0.484375 -0.546875 0.484375 -1.59375q0 -1.015625 -0.484375 -1.5625q-0.484375 -0.546875 -1.203125 -0.546875q-0.703125 0 -1.203125 0.546875q-0.484375 0.546875 -0.484375 1.578125zm8.438232 3.453125l-1.828125 0l0 -6.90625l1.703125 0l0 0.984375q0.4375 -0.703125 0.78125 -0.921875q0.34375 -0.21875 0.796875 -0.21875q0.625 0 1.203125 0.34375l-0.5625 1.59375q-0.46875 -0.296875 -0.859375 -0.296875q-0.390625 0 -0.65625 0.21875q-0.265625 0.203125 -0.421875 0.765625q-0.15625 0.546875 -0.15625 2.296875l0 2.140625zm2.7961578 -1.96875l1.828125 -0.28125q0.125 0.53125 0.484375 0.8125q0.359375 0.265625 1.0 0.265625q0.703125 0 1.0625 -0.25q0.25 -0.1875 0.25 -0.484375q0 -0.21875 -0.140625 -0.359375q-0.125 -0.125 -0.609375 -0.234375q-2.203125 -0.484375 -2.796875 -0.890625q-0.828125 -0.5625 -0.828125 -1.5625q0 -0.890625 0.703125 -1.5q0.71875 -0.609375 2.203125 -0.609375q1.421875 0 2.109375 0.46875q0.703125 0.453125 0.953125 1.359375l-1.71875 0.3125q-0.109375 -0.40625 -0.421875 -0.609375q-0.3125 -0.21875 -0.890625 -0.21875q-0.71875 0 -1.03125 0.203125q-0.203125 0.140625 -0.203125 0.359375q0 0.203125 0.1875 0.34375q0.234375 0.171875 1.6875 0.515625q1.46875 0.328125 2.0625 0.8125q0.5625 0.484375 0.5625 1.359375q0 0.953125 -0.796875 1.640625q-0.78125 0.671875 -2.34375 0.671875q-1.40625 0 -2.234375 -0.5625q-0.828125 -0.578125 -1.078125 -1.5625z" fill-rule="nonzero"/><path fill="#595959" d="m184.9966 237.48172l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm0 5.578125l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm7.0622253 0l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.18837 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 6.78125l0 -9.5625l1.078125 0l0 0.890625q0.375 -0.53125 0.84375 -0.78125q0.484375 -0.265625 1.15625 -0.265625q0.875 0 1.546875 0.453125q0.6875 0.453125 1.03125 1.28125q0.34375 0.828125 0.34375 1.828125q0 1.046875 -0.375 1.90625q-0.375 0.84375 -1.109375 1.296875q-0.71875 0.453125 -1.53125 0.453125q-0.578125 0 -1.046875 -0.25q-0.46875 -0.25 -0.765625 -0.625l0 3.375l-1.171875 0zm1.0625 -6.078125q0 1.34375 0.53125 1.984375q0.546875 0.625 1.3125 0.625q0.78125 0 1.34375 -0.65625q0.5625 -0.65625 0.5625 -2.046875q0 -1.3125 -0.546875 -1.96875q-0.546875 -0.671875 -1.296875 -0.671875q-0.75 0 -1.328125 0.703125q-0.578125 0.703125 -0.578125 2.03125zm6.3343506 3.421875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.18837 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.0531006 2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm11.8671875 -0.15625l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5218506 4.125l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm9.974976 -1.046875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm5.8748627 -1.171875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.006226 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm14.8498535 2.59375q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.5218506 1.40625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625z" fill-rule="nonzero"/><path fill="#595959" d="m287.0787 237.4661l-1.9375 0l0 -0.9375l3.078125 0l0 5.59375l1.953125 0l0 0.9375l-5.234375 0l0 -0.9375l2.140625 0l0 -4.65625zm0.390625 -3.6875q0.203125 0 0.359375 0.078125q0.15625 0.0625 0.28125 0.1875q0.125 0.125 0.1875 0.28125q0.078125 0.15625 0.078125 0.359375q0 0.171875 -0.078125 0.34375q-0.0625 0.15625 -0.1875 0.28125q-0.125 0.125 -0.28125 0.203125q-0.15625 0.0625 -0.359375 0.0625q-0.1875 0 -0.34375 -0.0625q-0.15625 -0.078125 -0.28125 -0.203125q-0.125 -0.125 -0.203125 -0.28125q-0.0625 -0.171875 -0.0625 -0.34375q0 -0.203125 0.0625 -0.359375q0.078125 -0.15625 0.203125 -0.28125q0.125 -0.125 0.28125 -0.1875q0.15625 -0.078125 0.34375 -0.078125zm10.156006 9.28125l-5.34375 0l0 -1.046875l2.1875 0l0 -6.21875l-2.03125 1.109375l-0.421875 -0.96875l2.703125 -1.421875l1.015625 0l0 7.5l1.890625 0l0 1.046875z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m201.01575 224.92137c0 -23.047195 9.718155 -52.215775 25.635178 -59.50177c30.77098 -14.085388 64.80443 -4.8602295 97.238846 -4.8602295c21.21701 0 62.76114 14.575943 62.76114 -14.56987" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m201.01575 224.92136c0 -23.04718 9.718155 -52.215744 25.635178 -59.50174c30.77098 -14.085403 64.80443 -4.8602448 97.238846 -4.8602448c10.6085205 0 26.298798 3.643982 39.336945 3.6447449c6.519104 3.8146973E-4 12.375153 -0.91004944 16.60141 -3.6421814c1.0565796 -0.68304443 2.0112915 -1.4799194 2.8490295 -2.4049072c0.10470581 -0.115615845 0.20761108 -0.23324585 0.3086548 -0.35289c0.050476074 -0.05984497 0.10055542 -0.120162964 0.15014648 -0.18101501l0.103149414 -0.12886047" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m386.40622 158.4335l-0.58657837 -9.640869l-5.7476807 7.76239z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m405.16797 232.8189l426.96063 0l0 51.905502l-426.96063 0z" fill-rule="evenodd"/><path fill="#595959" d="m416.8711 254.6189l-1.828125 0l0 -6.90625l1.703125 0l0 0.984375q0.4375 -0.703125 0.78125 -0.921875q0.34375 -0.21875 0.796875 -0.21875q0.625 0 1.203125 0.34375l-0.5625 1.59375q-0.46875 -0.296875 -0.859375 -0.296875q-0.390625 0 -0.65625 0.21875q-0.265625 0.203125 -0.421875 0.765625q-0.15625 0.546875 -0.15625 2.296875l0 2.140625zm7.436798 -2.203125l1.828125 0.3125q-0.34375 1.0 -1.109375 1.53125q-0.765625 0.515625 -1.890625 0.515625q-1.8125 0 -2.671875 -1.171875q-0.6875 -0.953125 -0.6875 -2.390625q0 -1.71875 0.890625 -2.6875q0.90625 -0.96875 2.28125 -0.96875q1.546875 0 2.4375 1.015625q0.890625 1.015625 0.84375 3.125l-4.578125 0q0.015625 0.8125 0.4375 1.265625q0.421875 0.453125 1.0625 0.453125q0.421875 0 0.71875 -0.234375q0.296875 -0.234375 0.4375 -0.765625zm0.109375 -1.84375q-0.015625 -0.796875 -0.40625 -1.203125q-0.390625 -0.421875 -0.953125 -0.421875q-0.59375 0 -0.984375 0.4375q-0.390625 0.4375 -0.390625 1.1875l2.734375 0zm6.4749756 -2.859375l0 1.453125l-1.25 0l0 2.78125q0 0.84375 0.03125 0.984375q0.046875 0.140625 0.171875 0.234375q0.125 0.09375 0.3125 0.09375q0.25 0 0.734375 -0.171875l0.15625 1.421875q-0.640625 0.265625 -1.453125 0.265625q-0.484375 0 -0.890625 -0.15625q-0.390625 -0.171875 -0.578125 -0.4375q-0.1875 -0.265625 -0.265625 -0.71875q-0.046875 -0.3125 -0.046875 -1.28125l0 -3.015625l-0.84375 0l0 -1.453125l0.84375 0l0 -1.375l1.828125 -1.078125l0 2.453125l1.25 0zm2.6414795 2.109375l-1.671875 -0.3125q0.28125 -1.0 0.96875 -1.46875q0.6875 -0.484375 2.03125 -0.484375q1.21875 0 1.8125 0.296875q0.609375 0.28125 0.84375 0.734375q0.25 0.4375 0.25 1.625l-0.015625 2.140625q0 0.90625 0.078125 1.34375q0.09375 0.421875 0.34375 0.921875l-1.8125 0q-0.078125 -0.1875 -0.171875 -0.546875q-0.046875 -0.15625 -0.078125 -0.203125q-0.46875 0.453125 -1.0 0.6875q-0.53125 0.21875 -1.140625 0.21875q-1.0625 0 -1.6875 -0.578125q-0.609375 -0.578125 -0.609375 -1.46875q0 -0.578125 0.28125 -1.03125q0.28125 -0.46875 0.78125 -0.703125q0.515625 -0.25 1.46875 -0.4375q1.28125 -0.234375 1.765625 -0.453125l0 -0.171875q0 -0.53125 -0.265625 -0.75q-0.25 -0.234375 -0.96875 -0.234375q-0.484375 0 -0.765625 0.203125q-0.265625 0.1875 -0.4375 0.671875zm2.4375 1.484375q-0.34375 0.109375 -1.109375 0.28125q-0.765625 0.15625 -1.0 0.3125q-0.359375 0.25 -0.359375 0.640625q0 0.390625 0.28125 0.671875q0.296875 0.28125 0.734375 0.28125q0.5 0 0.953125 -0.328125q0.328125 -0.25 0.4375 -0.609375q0.0625 -0.234375 0.0625 -0.890625l0 -0.359375zm3.5999756 -4.53125l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm9.999878 0l-1.828125 0l0 -3.53125q0 -1.109375 -0.125 -1.4375q-0.109375 -0.328125 -0.375 -0.515625q-0.265625 -0.1875 -0.640625 -0.1875q-0.46875 0 -0.859375 0.265625q-0.375 0.265625 -0.515625 0.703125q-0.140625 0.421875 -0.140625 1.578125l0 3.125l-1.828125 0l0 -6.90625l1.703125 0l0 1.015625q0.90625 -1.171875 2.28125 -1.171875q0.609375 0 1.109375 0.21875q0.5 0.21875 0.75 0.5625q0.265625 0.328125 0.359375 0.765625q0.109375 0.421875 0.109375 1.21875l0 4.296875zm5.547455 0l0 -9.546875l1.828125 0l0 9.546875l-1.828125 0zm3.703003 -7.84375l0 -1.703125l1.828125 0l0 1.703125l-1.828125 0zm0 7.84375l0 -6.90625l1.828125 0l0 6.90625l-1.828125 0zm3.0623474 -1.96875l1.828125 -0.28125q0.125 0.53125 0.484375 0.8125q0.359375 0.265625 1.0 0.265625q0.703125 0 1.0625 -0.25q0.25 -0.1875 0.25 -0.484375q0 -0.21875 -0.140625 -0.359375q-0.125 -0.125 -0.609375 -0.234375q-2.203125 -0.484375 -2.796875 -0.890625q-0.828125 -0.5625 -0.828125 -1.5625q0 -0.890625 0.703125 -1.5q0.71875 -0.609375 2.203125 -0.609375q1.421875 0 2.109375 0.46875q0.703125 0.453125 0.953125 1.359375l-1.71875 0.3125q-0.109375 -0.40625 -0.421875 -0.609375q-0.3125 -0.21875 -0.890625 -0.21875q-0.71875 0 -1.03125 0.203125q-0.203125 0.140625 -0.203125 0.359375q0 0.203125 0.1875 0.34375q0.234375 0.171875 1.6875 0.515625q1.46875 0.328125 2.0625 0.8125q0.5625 0.484375 0.5625 1.359375q0 0.953125 -0.796875 1.640625q-0.78125 0.671875 -2.34375 0.671875q-1.40625 0 -2.234375 -0.5625q-0.828125 -0.578125 -1.078125 -1.5625zm11.224976 -4.9375l0 1.453125l-1.25 0l0 2.78125q0 0.84375 0.03125 0.984375q0.046875 0.140625 0.171875 0.234375q0.125 0.09375 0.3125 0.09375q0.25 0 0.734375 -0.171875l0.15625 1.421875q-0.640625 0.265625 -1.453125 0.265625q-0.484375 0 -0.890625 -0.15625q-0.390625 -0.171875 -0.578125 -0.4375q-0.1875 -0.265625 -0.265625 -0.71875q-0.046875 -0.3125 -0.046875 -1.28125l0 -3.015625l-0.84375 0l0 -1.453125l0.84375 0l0 -1.375l1.828125 -1.078125l0 2.453125l1.25 0z" fill-rule="nonzero"/><path fill="#595959" d="m474.62537 249.04077l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm0 5.578125l0 -1.328125l1.328125 0l0 1.328125l-1.328125 0zm9.64035 -1.046875l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1404724 1.046875l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.0531006 2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm11.8671875 -0.15625l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.3029785 4.125l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm10.865601 3.5l0 -1.015625q-0.8125 1.171875 -2.1875 1.171875q-0.609375 0 -1.140625 -0.234375q-0.53125 -0.234375 -0.796875 -0.578125q-0.25 -0.359375 -0.359375 -0.875q-0.0625 -0.34375 -0.0625 -1.09375l0 -4.28125l1.171875 0l0 3.828125q0 0.921875 0.0625 1.234375q0.109375 0.46875 0.46875 0.734375q0.359375 0.25 0.890625 0.25q0.515625 0 0.984375 -0.265625q0.46875 -0.265625 0.65625 -0.734375q0.1875 -0.46875 0.1875 -1.34375l0 -3.703125l1.171875 0l0 6.90625l-1.046875 0zm3.1624756 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm3.4620972 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm8.156128 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5062256 4.125l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm3.9852295 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm15.35144 1.203125q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm2.9749756 3.46875l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.1883545 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm9.7561035 2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm9.6953125 1.015625l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm1.1560669 -7.140625l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.9298706 0l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm2.9610596 0l0 -9.546875l1.171875 0l0 9.546875l-1.171875 0zm6.6797485 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.256226 1.90625l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.006226 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm11.365601 1.234375l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.006226 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m442.29922 237.92389l-40.661438 -56.56694" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m442.29922 237.92389l-33.65738 -46.82306" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m411.32425 189.17267l-7.9799194 -5.44162l2.6151733 9.297913z" fill-rule="evenodd"/><path fill="#000000" fill-opacity="0.0" d="m50.48294 277.15747l899.5276 0l0 46.86615l-899.5276 0z" fill-rule="evenodd"/><path fill="#9aa0a6" d="m60.123566 304.07748l2.359375 -13.359375l1.578125 0l-2.375 13.359375l-1.5625 0zm4.270874 0l1.671875 -9.515625l1.5 0l-0.25 1.40625l0.078125 0q0.484375 -0.71875 1.296875 -1.203125q0.828125 -0.5 1.96875 -0.5q1.09375 0 1.828125 0.46875q0.75 0.46875 1.046875 1.359375q0.296875 0.890625 0.078125 2.15625l-1.03125 5.828125l-1.59375 0l1.0625 -5.96875q0.15625 -0.78125 -0.046875 -1.3125q-0.203125 -0.546875 -0.6875 -0.8125q-0.484375 -0.28125 -1.21875 -0.28125q-0.78125 0 -1.453125 0.40625q-0.65625 0.390625 -1.109375 1.078125q-0.453125 0.671875 -0.609375 1.53125l-0.953125 5.359375l-1.578125 0zm10.999496 -9.515625l5.578125 0l-0.265625 1.4375l-5.578125 0l0.265625 -1.4375zm0.453125 6.859375l1.6875 -9.546875l1.578125 0l-1.65625 9.390625q-0.140625 0.6875 0.109375 1.09375q0.25 0.390625 0.90625 0.390625q0.3125 0 0.59375 -0.09375q0.296875 -0.09375 0.546875 -0.21875l-0.265625 1.546875q-0.28125 0.109375 -0.578125 0.171875q-0.296875 0.078125 -0.765625 0.078125q-0.734375 0 -1.28125 -0.328125q-0.546875 -0.328125 -0.796875 -0.953125q-0.234375 -0.640625 -0.078125 -1.53125zm8.875061 2.953125q-1.09375 0 -1.84375 -0.46875q-0.734375 -0.484375 -1.046875 -1.3125q-0.296875 -0.84375 -0.109375 -1.9375l1.078125 -6.09375l1.5625 0l-1.0625 6.046875q-0.109375 0.671875 0.078125 1.203125q0.1875 0.53125 0.65625 0.828125q0.484375 0.296875 1.234375 0.296875q0.78125 0 1.4375 -0.390625q0.671875 -0.390625 1.125 -1.0625q0.46875 -0.6875 0.625 -1.546875l0.953125 -5.375l1.5625 0l-1.65625 9.515625l-1.515625 0l0.265625 -1.375l-0.078125 0q-0.3125 0.453125 -0.796875 0.84375q-0.46875 0.375 -1.09375 0.59375q-0.609375 0.234375 -1.375 0.234375zm7.091156 -0.296875l1.65625 -9.515625l1.59375 0l-1.6875 9.515625l-1.5625 0zm2.984375 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.0135956 1.75l5.578125 0l-0.265625 1.4375l-5.578125 0l0.265625 -1.4375zm0.453125 6.859375l1.6875 -9.546875l1.578125 0l-1.65625 9.390625q-0.140625 0.6875 0.109375 1.09375q0.25 0.390625 0.90625 0.390625q0.3125 0 0.59375 -0.09375q0.296875 -0.09375 0.546875 -0.21875l-0.265625 1.546875q-0.28125 0.109375 -0.578125 0.171875q-0.296875 0.078125 -0.765625 0.078125q-0.734375 0 -1.28125 -0.328125q-0.546875 -0.328125 -0.796875 -0.953125q-0.234375 -0.640625 -0.078125 -1.53125zm5.593399 2.65625l1.65625 -9.515625l1.59375 0l-1.6875 9.515625l-1.5625 0zm2.984375 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.8569336 11.265625l-2.140625 -9.515625l1.6875 0l1.578125 7.65625l0.03125 0l4.3125 -7.65625l1.71875 0l-5.5625 9.515625l-1.625 0zm11.074936 0.296875q-1.328125 0 -2.28125 -0.59375q-0.953125 -0.59375 -1.46875 -1.578125q-0.515625 -1.0 -0.515625 -2.1875q0 -1.109375 0.375 -2.140625q0.390625 -1.03125 1.09375 -1.84375q0.703125 -0.8125 1.703125 -1.28125q1.0 -0.484375 2.21875 -0.484375q1.28125 0 2.15625 0.5625q0.890625 0.5625 1.34375 1.46875q0.46875 0.890625 0.46875 1.984375q0 0.328125 -0.046875 0.703125q-0.03125 0.375 -0.125 0.640625l-8.234375 0l0.265625 -1.3125l6.546875 0q0.0625 -0.734375 -0.25 -1.3125q-0.296875 -0.59375 -0.890625 -0.9375q-0.578125 -0.359375 -1.40625 -0.359375q-1.09375 0 -1.90625 0.59375q-0.8125 0.578125 -1.265625 1.609375q-0.4375 1.015625 -0.4375 2.3125q0 0.859375 0.390625 1.46875q0.390625 0.59375 1.015625 0.921875q0.625 0.328125 1.3125 0.328125q0.765625 0 1.359375 -0.203125q0.59375 -0.21875 1.03125 -0.625q0.453125 -0.40625 0.78125 -0.9375l1.21875 0.6875q-0.46875 0.8125 -1.140625 1.375q-0.65625 0.5625 -1.5 0.84375q-0.828125 0.296875 -1.8125 0.296875zm6.0065002 -0.296875l2.34375 -13.359375l1.5625 0l-2.34375 13.359375l-1.5625 0zm6.220749 3.078125q-0.203125 0.359375 -0.359375 0.59375q-0.140625 0.25 -0.21875 0.359375l-1.671875 0q0.078125 -0.109375 0.203125 -0.328125q0.125 -0.21875 0.265625 -0.4375q0.15625 -0.21875 0.234375 -0.359375q0.140625 -0.203125 0.296875 -0.453125q0.171875 -0.25 0.578125 -0.90625l1.484375 -2.3125l1.0 -1.671875l4.140625 -7.078125l1.75 0l-7.03125 11.515625q-0.03125 0.046875 -0.140625 0.21875q-0.09375 0.171875 -0.234375 0.390625q-0.140625 0.21875 -0.296875 0.46875zm0.921875 -2.921875l-2.578125 -9.671875l1.65625 0l1.78125 7.078125l0.765625 0l-1.625 2.59375zm8.954758 -7.390625q-0.5 0 -0.859375 -0.34375q-0.34375 -0.34375 -0.34375 -0.859375q0 -0.5 0.34375 -0.84375q0.359375 -0.34375 0.859375 -0.34375q0.484375 0 0.828125 0.34375q0.34375 0.34375 0.34375 0.84375q0 0.515625 -0.34375 0.859375q-0.34375 0.34375 -0.828125 0.34375zm-1.3125 7.34375q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.484375 0 0.828125 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.828125 0.34375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m150.27133 304.07748l0 -13.359375l4.015625 0q2.09375 0 3.609375 0.859375q1.515625 0.84375 2.328125 2.359375q0.828125 1.5 0.828125 3.46875q0 1.953125 -0.828125 3.453125q-0.8125 1.5 -2.328125 2.359375q-1.515625 0.859375 -3.609375 0.859375l-4.015625 0zm1.578125 -1.515625l2.4375 0q1.578125 0 2.734375 -0.59375q1.15625 -0.609375 1.78125 -1.765625q0.640625 -1.15625 0.640625 -2.796875q0 -1.65625 -0.640625 -2.8125q-0.625 -1.15625 -1.78125 -1.75q-1.15625 -0.609375 -2.734375 -0.609375l-2.4375 0l0 10.328125zm15.163467 1.8125q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm9.409302 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189636 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm3.9364624 0l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm8.233749 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm11.041 1.4375q-1.40625 0 -2.5 -0.65625q-1.078125 -0.671875 -1.703125 -1.8125q-0.609375 -1.15625 -0.609375 -2.578125q0 -1.46875 0.609375 -2.59375q0.625 -1.140625 1.703125 -1.796875q1.09375 -0.671875 2.5 -0.671875q1.609375 0 2.640625 0.734375q1.03125 0.734375 1.46875 1.890625l-1.4375 0.609375q-0.359375 -0.890625 -1.0625 -1.34375q-0.6875 -0.453125 -1.6875 -0.453125q-0.828125 0 -1.546875 0.453125q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.921875q0 1.078125 0.453125 1.90625q0.453125 0.8125 1.171875 1.265625q0.71875 0.4375 1.546875 0.4375q1.015625 0 1.734375 -0.46875q0.734375 -0.46875 1.09375 -1.3125l1.40625 0.59375q-0.46875 1.09375 -1.515625 1.859375q-1.03125 0.765625 -2.640625 0.765625zm8.827347 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm5.876053 -8.421875l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm9.61763 2.796875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m252.13014 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm9.647903 2.59375l0 -6.5625q0 -0.4375 -0.03125 -0.703125q-0.03125 -0.28125 -0.109375 -0.4375q-0.0625 -0.15625 -0.171875 -0.21875q-0.109375 -0.0625 -0.25 -0.0625q-0.1875 0 -0.34375 0.109375q-0.15625 0.109375 -0.34375 0.359375q-0.171875 0.234375 -0.390625 0.640625q-0.203125 0.40625 -0.5 1.0l0 5.875l-1.4375 0l0 -6.390625q0 -0.5 -0.03125 -0.8125q-0.03125 -0.3125 -0.109375 -0.484375q-0.0625 -0.171875 -0.171875 -0.234375q-0.109375 -0.0625 -0.265625 -0.0625q-0.17185974 0 -0.31248474 0.09375q-0.140625 0.078125 -0.328125 0.328125q-0.171875 0.234375 -0.390625 0.640625q-0.21875 0.40625 -0.515625 1.046875l0 5.875l-1.46875 0l0 -9.140625l1.21875 0l0.078125 1.734375q0.234375 -0.515625 0.453125 -0.875q0.21875 -0.375 0.453125 -0.59375q0.25 -0.21875 0.515625 -0.328125q0.26560974 -0.109375 0.59373474 -0.109375q0.734375 0 1.109375 0.484375q0.390625 0.484375 0.390625 1.5q0.21875 -0.484375 0.421875 -0.84375q0.21875 -0.375 0.453125 -0.625q0.25 -0.25 0.53125 -0.375q0.296875 -0.140625 0.6875 -0.140625q1.71875 0 1.71875 2.65625l0 6.65625l-1.453125 0zm11.351044 0l-7.46875 0l0 -1.46875l3.0625 0l0 -8.703125l-2.859375 1.546875l-0.578125 -1.34375l3.78125 -1.984375l1.40625 0l0 10.484375l2.65625 0l0 1.46875z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m278.48026 304.07748l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.6225586 3.1875l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m302.84555 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm11.085419 -2.046875q0 1.0625 -0.3125 1.953125q-0.296875 0.890625 -0.859375 1.53125q-0.5625 0.625 -1.375 0.96875q-0.8125 0.34375 -1.84375 0.34375q-0.984375 0 -1.765625 -0.296875q-0.765625 -0.3125 -1.3125 -0.890625q-0.546875 -0.59375 -0.828125 -1.46875q-0.28125 -0.875 -0.28125 -2.015625q0 -1.0625 0.296875 -1.9375q0.296875 -0.890625 0.859375 -1.515625q0.5625 -0.640625 1.375 -0.984375q0.8125 -0.359375 1.84375 -0.359375q0.984375 0 1.765625 0.3125q0.78125 0.296875 1.3125 0.890625q0.546875 0.578125 0.828125 1.453125q0.296875 0.875 0.296875 2.015625zm-1.625 0.0625q0 -0.84375 -0.1875 -1.46875q-0.1875 -0.640625 -0.53125 -1.0625q-0.34375 -0.421875 -0.84375 -0.640625q-0.5 -0.21875 -1.109375 -0.21875q-0.703125 0 -1.21875 0.28125q-0.5 0.28125 -0.828125 0.75q-0.3125 0.453125 -0.46875 1.078125q-0.15625 0.609375 -0.15625 1.28125q0 0.859375 0.1875 1.5q0.1875 0.625 0.53125 1.046875q0.359375 0.421875 0.84375 0.640625q0.5 0.21875 1.109375 0.21875q0.71875 0 1.21875 -0.28125q0.5 -0.28125 0.828125 -0.734375q0.328125 -0.46875 0.46875 -1.078125q0.15625 -0.625 0.15625 -1.3125zm12.382294 -4.5625l-1.328125 9.140625l-1.921875 0l-1.3125 -3.828125l-0.265625 -0.921875l-0.296875 0.984375l-1.265625 3.765625l-1.875 0l-1.3125 -9.140625l1.546875 0l0.765625 6.203125l0.15625 1.390625l0.390625 -1.21875l1.328125 -4.109375l1.140625 0l1.4375 4.0625l0.40625 1.203125l0.140625 -1.28125l0.703125 -6.25l1.5625 0zm1.7885437 0l1.421875 0l0.0625 1.46875q0.390625 -0.46875 0.765625 -0.78125q0.375 -0.3125 0.734375 -0.5q0.359375 -0.203125 0.734375 -0.28125q0.375 -0.078125 0.78125 -0.078125q1.40625 0 2.125 0.84375q0.734375 0.828125 0.734375 2.5l0 5.96875l-1.59375 0l0 -5.84375q0 -1.078125 -0.40625 -1.578125q-0.390625 -0.515625 -1.1875 -0.515625q-0.28125 0 -0.5625 0.09375q-0.28125 0.078125 -0.578125 0.296875q-0.296875 0.203125 -0.65625 0.578125q-0.34375 0.359375 -0.78125 0.90625l0 6.0625l-1.59375 0l0 -9.140625zm17.882294 9.140625l-7.46875 0l0 -1.46875l3.0625 0l0 -8.703125l-2.859375 1.546875l-0.578125 -1.34375l3.78125 -1.984375l1.40625 0l0 10.484375l2.65625 0l0 1.46875z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m352.6009 304.37436q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189636 1.09375l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.513031 0.296875q-1.28125 0 -2.328125 -0.640625q-1.03125 -0.65625 -1.625 -1.796875q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.484375 0.59375 -2.609375q0.59375 -1.140625 1.625 -1.796875q1.046875 -0.65625 2.328125 -0.65625q0.765625 0 1.390625 0.25q0.640625 0.234375 1.109375 0.625q0.484375 0.390625 0.75 0.84375l0.078125 0l-0.078125 -1.328125l0 -3.9375l1.59375 0l0 13.359375l-1.515625 0l0 -1.40625l-0.078125 0q-0.265625 0.4375 -0.75 0.828125q-0.46875 0.390625 -1.109375 0.625q-0.625 0.25 -1.390625 0.25zm0.171875 -1.4375q0.828125 0 1.546875 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.453125 -0.8125 0.453125 -1.921875q0 -1.125 -0.453125 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.546875 -0.4375q-0.8125 0 -1.53125 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.9375q0 1.09375 0.453125 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.53125 0.4375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m394.39218 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm9.647919 2.59375l0 -6.5625q0 -0.4375 -0.03125 -0.703125q-0.03125 -0.28125 -0.109375 -0.4375q-0.0625 -0.15625 -0.171875 -0.21875q-0.109375 -0.0625 -0.25 -0.0625q-0.1875 0 -0.34375 0.109375q-0.15625 0.109375 -0.34375 0.359375q-0.171875 0.234375 -0.390625 0.640625q-0.203125 0.40625 -0.5 1.0l0 5.875l-1.4375 0l0 -6.390625q0 -0.5 -0.03125 -0.8125q-0.03125 -0.3125 -0.109375 -0.484375q-0.0625 -0.171875 -0.171875 -0.234375q-0.109375 -0.0625 -0.265625 -0.0625q-0.171875 0 -0.3125 0.09375q-0.140625 0.078125 -0.328125 0.328125q-0.171875 0.234375 -0.390625 0.640625q-0.21875 0.40625 -0.515625 1.046875l0 5.875l-1.46875 0l0 -9.140625l1.21875 0l0.078125 1.734375q0.234375 -0.515625 0.453125 -0.875q0.21875 -0.375 0.453125 -0.59375q0.25 -0.21875 0.515625 -0.328125q0.265625 -0.109375 0.59375 -0.109375q0.734375 0 1.109375 0.484375q0.390625 0.484375 0.390625 1.5q0.21875 -0.484375 0.421875 -0.84375q0.21875 -0.375 0.453125 -0.625q0.25 -0.25 0.53125 -0.375q0.296875 -0.140625 0.6875 -0.140625q1.71875 0 1.71875 2.65625l0 6.65625l-1.453125 0zm11.460419 0l-7.84375 0l0 -1.421875l3.078125 -3.0625q0.75 -0.75 1.21875 -1.296875q0.484375 -0.546875 0.75 -0.984375q0.265625 -0.453125 0.34375 -0.859375q0.09375 -0.40625 0.09375 -0.859375q0 -0.4375 -0.125 -0.828125q-0.109375 -0.40625 -0.359375 -0.703125q-0.234375 -0.3125 -0.640625 -0.484375q-0.390625 -0.171875 -0.9375 -0.171875q-0.75 0 -1.375 0.34375q-0.609375 0.328125 -1.125 0.875l-0.875 -1.046875q0.671875 -0.71875 1.546875 -1.140625q0.875 -0.421875 2.046875 -0.421875q0.796875 0 1.4375 0.234375q0.65625 0.234375 1.125 0.6875q0.484375 0.4375 0.734375 1.09375q0.265625 0.640625 0.265625 1.453125q0 0.671875 -0.1875 1.265625q-0.171875 0.578125 -0.546875 1.15625q-0.375 0.578125 -0.9375 1.203125q-0.546875 0.609375 -1.3125 1.359375l-2.171875 2.09375l5.796875 0l0 1.515625z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m420.7423 304.07748l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.6225586 3.1875l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m445.10757 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm11.085419 -2.046875q0 1.0625 -0.3125 1.953125q-0.296875 0.890625 -0.859375 1.53125q-0.5625 0.625 -1.375 0.96875q-0.8125 0.34375 -1.84375 0.34375q-0.984375 0 -1.765625 -0.296875q-0.765625 -0.3125 -1.3125 -0.890625q-0.546875 -0.59375 -0.828125 -1.46875q-0.28125 -0.875 -0.28125 -2.015625q0 -1.0625 0.296875 -1.9375q0.296875 -0.890625 0.859375 -1.515625q0.5625 -0.640625 1.375 -0.984375q0.8125 -0.359375 1.84375 -0.359375q0.984375 0 1.765625 0.3125q0.78125 0.296875 1.3125 0.890625q0.546875 0.578125 0.828125 1.453125q0.296875 0.875 0.296875 2.015625zm-1.625 0.0625q0 -0.84375 -0.1875 -1.46875q-0.1875 -0.640625 -0.53125 -1.0625q-0.34375 -0.421875 -0.84375 -0.640625q-0.5 -0.21875 -1.109375 -0.21875q-0.703125 0 -1.21875 0.28125q-0.5 0.28125 -0.828125 0.75q-0.3125 0.453125 -0.46875 1.078125q-0.15625 0.609375 -0.15625 1.28125q0 0.859375 0.1875 1.5q0.1875 0.625 0.53125 1.046875q0.359375 0.421875 0.84375 0.640625q0.5 0.21875 1.109375 0.21875q0.71875 0 1.21875 -0.28125q0.5 -0.28125 0.828125 -0.734375q0.328125 -0.46875 0.46875 -1.078125q0.15625 -0.625 0.15625 -1.3125zm12.382294 -4.5625l-1.328125 9.140625l-1.921875 0l-1.3125 -3.828125l-0.265625 -0.921875l-0.296875 0.984375l-1.265625 3.765625l-1.875 0l-1.3125 -9.140625l1.546875 0l0.765625 6.203125l0.15625 1.390625l0.390625 -1.21875l1.328125 -4.109375l1.140625 0l1.4375 4.0625l0.40625 1.203125l0.140625 -1.28125l0.703125 -6.25l1.5625 0zm1.7885437 0l1.421875 0l0.0625 1.46875q0.390625 -0.46875 0.765625 -0.78125q0.375 -0.3125 0.734375 -0.5q0.359375 -0.203125 0.734375 -0.28125q0.375 -0.078125 0.78125 -0.078125q1.40625 0 2.125 0.84375q0.734375 0.828125 0.734375 2.5l0 5.96875l-1.59375 0l0 -5.84375q0 -1.078125 -0.40625 -1.578125q-0.390625 -0.515625 -1.1875 -0.515625q-0.28125 0 -0.5625 0.09375q-0.28125 0.078125 -0.578125 0.296875q-0.296875 0.203125 -0.65625 0.578125q-0.34375 0.359375 -0.78125 0.90625l0 6.0625l-1.59375 0l0 -9.140625zm17.991669 9.140625l-7.84375 0l0 -1.421875l3.078125 -3.0625q0.75 -0.75 1.21875 -1.296875q0.484375 -0.546875 0.75 -0.984375q0.265625 -0.453125 0.34375 -0.859375q0.09375 -0.40625 0.09375 -0.859375q0 -0.4375 -0.125 -0.828125q-0.109375 -0.40625 -0.359375 -0.703125q-0.234375 -0.3125 -0.640625 -0.484375q-0.390625 -0.171875 -0.9375 -0.171875q-0.75 0 -1.375 0.34375q-0.609375 0.328125 -1.125 0.875l-0.875 -1.046875q0.671875 -0.71875 1.546875 -1.140625q0.875 -0.421875 2.046875 -0.421875q0.796875 0 1.4375 0.234375q0.65625 0.234375 1.125 0.6875q0.484375 0.4375 0.734375 1.09375q0.265625 0.640625 0.265625 1.453125q0 0.671875 -0.1875 1.265625q-0.171875 0.578125 -0.546875 1.15625q-0.375 0.578125 -0.9375 1.203125q-0.546875 0.609375 -1.3125 1.359375l-2.171875 2.09375l5.796875 0l0 1.515625z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m488.53467 304.18686q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.5 0 0.84375 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.84375 0.34375zm8.049683 -0.109375l0 -13.359375l4.78125 0q1.03125 0 1.890625 0.46875q0.875 0.453125 1.40625 1.265625q0.53125 0.796875 0.53125 1.875q0 0.984375 -0.5 1.71875q-0.5 0.71875 -1.328125 1.09375l0 0.078125q0.984375 0.3125 1.625 1.140625q0.65625 0.8125 0.65625 1.9375q0 1.109375 -0.5625 1.96875q-0.5625 0.84375 -1.484375 1.328125q-0.90625 0.484375 -1.96875 0.484375l-5.046875 0zm0.828125 -6.078125l0 -1.515625l3.953125 0q0.71875 0 1.21875 -0.328125q0.5 -0.328125 0.765625 -0.8125q0.28125 -0.484375 0.28125 -0.984375q0 -0.5 -0.265625 -0.984375q-0.265625 -0.484375 -0.75 -0.8125q-0.484375 -0.328125 -1.171875 -0.328125l-3.28125 0l0 10.328125l3.5625 0q0.703125 0 1.21875 -0.34375q0.53125 -0.34375 0.8125 -0.875q0.28125 -0.53125 0.28125 -1.078125q0 -0.546875 -0.296875 -1.0625q-0.28125 -0.53125 -0.8125 -0.859375q-0.53125 -0.34375 -1.28125 -0.34375l-4.234375 0zm13.60257 6.375q-1.75 0 -2.640625 -1.015625q-0.890625 -1.03125 -0.890625 -2.8125l0 -5.984375l1.59375 0l0 5.75q0 1.421875 0.65625 2.03125q0.65625 0.59375 1.625 0.59375q0.8280945 0 1.4374695 -0.4375q0.625 -0.4375 0.953125 -1.125q0.34375 -0.6875 0.34375 -1.4375l0 -5.375l1.59375 0l0 9.515625l-1.515625 0l0 -1.375l-0.078125 0q-0.234375 0.4375 -0.71875 0.828125q-0.46875 0.375 -1.078125 0.609375q-0.6093445 0.234375 -1.2812195 0.234375zm6.4501038 -9.8125l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm14.512756 2.796875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm6.54187 1.140625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm10.4661255 0l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm6.6708374 3.546875q-0.0625 0.125 -0.140625 0.28125q-0.078125 0.171875 -0.09375 0.203125l-1.640625 0q0.109375 -0.25 0.234375 -0.546875q0.140625 -0.28125 0.328125 -0.6875q0.109375 -0.234375 0.203125 -0.453125q0.09375 -0.203125 0.21875 -0.453125q0.125 -0.25 0.28125 -0.625l1.3125 -2.828125l0.359375 -0.875l2.859375 -7.078125l1.71875 0l-4.921875 11.359375q-0.078125 0.171875 -0.234375 0.53125q-0.15625 0.375 -0.296875 0.71875q-0.140625 0.34375 -0.1875 0.453125zm0.34375 -3.9375l-4.03125 -9.125l1.71875 0l2.984375 7.078125l0.15625 0l-0.828125 2.046875zm11.163574 0.390625l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.6225586 3.1875l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0zm9.633362 -9.515625l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm5.887268 2.5l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.3125l0.078125 0q0.375 -0.71875 1.21875 -1.203125q0.84375 -0.5 1.875 -0.5q1.21875 0 2.0 0.484375q0.78125 0.484375 1.171875 1.3125q0.390625 0.828125 0.390625 1.921875l0 6.09375l-1.59375 0l0 -5.859375q0 -0.90625 -0.328125 -1.453125q-0.3125 -0.5625 -0.84375 -0.8125q-0.53125 -0.25 -1.171875 -0.25q-0.78125 0 -1.40625 0.453125q-0.609375 0.4375 -0.96875 1.15625q-0.34375 0.703125 -0.34375 1.46875l0 5.296875l-1.59375 0zm13.475647 0.296875q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm5.8760376 -8.421875l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm12.137817 2.5l-3.0625 -9.515625l1.640625 0l2.265625 7.515625l0.015625 0l2.40625 -7.515625l1.625 0l2.40625 7.5l0.015625 0l2.265625 -7.5l1.59375 0l-3.078125 9.515625l-1.59375 0l-2.46875 -7.59375l0 0l-2.4375 7.59375l-1.59375 0zm16.684265 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm10.088257 1.4375q-1.75 0 -2.640625 -1.015625q-0.890625 -1.03125 -0.890625 -2.8125l0 -5.984375l1.59375 0l0 5.75q0 1.421875 0.65625 2.03125q0.65625 0.59375 1.625 0.59375q0.828125 0 1.4375 -0.4375q0.625 -0.4375 0.953125 -1.125q0.34375 -0.6875 0.34375 -1.4375l0 -5.375l1.59375 0l0 9.515625l-1.515625 0l0 -1.375l-0.078125 0q-0.234375 0.4375 -0.71875 0.828125q-0.46875 0.375 -1.078125 0.609375q-0.609375 0.234375 -1.28125 0.234375zm7.0130615 -0.296875l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm7.8899536 0.296875q-1.28125 0 -2.328125 -0.640625q-1.03125 -0.65625 -1.625 -1.796875q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.484375 0.59375 -2.609375q0.59375 -1.140625 1.625 -1.796875q1.046875 -0.65625 2.328125 -0.65625q0.765625 0 1.390625 0.25q0.640625 0.234375 1.109375 0.625q0.484375 0.390625 0.75 0.84375l0.078125 0l-0.078125 -1.328125l0 -3.9375l1.59375 0l0 13.359375l-1.515625 0l0 -1.40625l-0.078125 0q-0.265625 0.4375 -0.75 0.828125q-0.46875 0.390625 -1.109375 0.625q-0.625 0.25 -1.390625 0.25zm0.171875 -1.4375q0.828125 0 1.546875 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.453125 -0.8125 0.453125 -1.921875q0 -1.125 -0.453125 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.546875 -0.4375q-0.8125 0 -1.53125 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.9375q0 1.09375 0.453125 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.53125 0.4375zm11.340576 1.140625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.856812 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm5.7922363 -8.375l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm10.278442 2.5l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.2788086 11.265625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm12.889648 0l-3.84375 -9.515625l1.703125 0l2.953125 7.65625l0.03125 0l2.984375 -7.65625l1.65625 0l-3.875 9.515625l-1.609375 0zm9.652649 0.296875q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.718933 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm4.0927734 0l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm7.2323 11.5625q-1.28125 0 -2.328125 -0.640625q-1.03125 -0.65625 -1.625 -1.796875q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.484375 0.59375 -2.609375q0.59375 -1.140625 1.625 -1.796875q1.046875 -0.65625 2.328125 -0.65625q0.765625 0 1.390625 0.25q0.640625 0.234375 1.109375 0.625q0.484375 0.390625 0.75 0.84375l0.078125 0l-0.078125 -1.328125l0 -3.9375l1.59375 0l0 13.359375l-1.515625 0l0 -1.40625l-0.078125 0q-0.265625 0.4375 -0.75 0.828125q-0.46875 0.390625 -1.109375 0.625q-0.625 0.25 -1.390625 0.25zm0.171875 -1.4375q0.828125 0 1.546875 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.453125 -0.8125 0.453125 -1.921875q0 -1.125 -0.453125 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.546875 -0.4375q-0.8125 0 -1.53125 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.9375q0 1.09375 0.453125 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.53125 0.4375zm9.965881 1.4375q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm5.8760376 -8.421875l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm9.617676 2.796875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m802.803 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916992 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.507263 9.140625l-7.46875 0l0 -1.46875l3.0625 0l0 -8.703125l-2.859375 1.546875l-0.578125 -1.34375l3.78125 -1.984375l1.40625 0l0 10.484375l2.65625 0l0 1.46875z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m825.21545 306.12436l-0.765625 -0.5l1.0 -1.546875q-0.296875 -0.0625 -0.59375 -0.328125q-0.28125 -0.28125 -0.28125 -0.75q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.859375 -0.34375q0.25 0 0.46875 0.09375q0.21875 0.078125 0.375 0.25q0.15625 0.15625 0.25 0.375q0.09375 0.203125 0.09375 0.453125q0 0.21875 -0.078125 0.4375q-0.0625 0.21875 -0.1875 0.421875q-0.109375 0.203125 -0.25 0.40625l-1.234375 1.859375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m842.34326 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916992 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.616638 9.140625l-7.84375 0l0 -1.421875l3.078125 -3.0625q0.75 -0.75 1.21875 -1.296875q0.484375 -0.546875 0.75 -0.984375q0.265625 -0.453125 0.34375 -0.859375q0.09375 -0.40625 0.09375 -0.859375q0 -0.4375 -0.125 -0.828125q-0.109375 -0.40625 -0.359375 -0.703125q-0.234375 -0.3125 -0.640625 -0.484375q-0.390625 -0.171875 -0.9375 -0.171875q-0.75 0 -1.375 0.34375q-0.609375 0.328125 -1.125 0.875l-0.875 -1.046875q0.671875 -0.71875 1.546875 -1.140625q0.875 -0.421875 2.046875 -0.421875q0.796875 0 1.4375 0.234375q0.65625 0.234375 1.125 0.6875q0.484375 0.4375 0.734375 1.09375q0.265625 0.640625 0.265625 1.453125q0 0.671875 -0.1875 1.265625q-0.171875 0.578125 -0.546875 1.15625q-0.375 0.578125 -0.9375 1.203125q-0.546875 0.609375 -1.3125 1.359375l-2.171875 2.09375l5.796875 0l0 1.515625z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m872.92773 304.37436q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm6.54187 1.140625l0 -9.515625l1.515625 0l0 1.53125l0.078125 0q0.1875 -0.546875 0.625 -0.9375q0.4375 -0.40625 1.0 -0.640625q0.578125 -0.234375 1.125 -0.234375q0.4375 0 0.671875 0.046875q0.25 0.046875 0.453125 0.140625l0 1.71875q-0.296875 -0.15625 -0.640625 -0.21875q-0.34375 -0.078125 -0.703125 -0.078125q-0.6875 0 -1.265625 0.390625q-0.578125 0.390625 -0.921875 1.046875q-0.34375 0.65625 -0.34375 1.4375l0 5.3125l-1.59375 0z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m899.7002 291.20248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916992 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.382263 5.53125q0 0.78125 -0.328125 1.46875q-0.3125 0.6875 -0.953125 1.203125q-0.625 0.515625 -1.59375 0.8125q-0.96875 0.28125 -2.25 0.28125q-0.71875 0 -1.28125 -0.046875q-0.5625 -0.03125 -1.046875 -0.109375l0 -1.421875q0.5625 0.09375 1.1875 0.15625q0.625 0.046875 1.28125 0.046875q0.890625 0 1.515625 -0.15625q0.640625 -0.15625 1.046875 -0.453125q0.40625 -0.296875 0.578125 -0.71875q0.1875 -0.4375 0.1875 -0.984375q0 -0.484375 -0.21875 -0.84375q-0.21875 -0.375 -0.625 -0.609375q-0.390625 -0.25 -0.9375 -0.375q-0.546875 -0.125 -1.203125 -0.125l-1.359375 0l0 -1.296875l1.375 0q0.53125 0 0.96875 -0.140625q0.453125 -0.140625 0.765625 -0.40625q0.3125 -0.265625 0.484375 -0.65625q0.171875 -0.390625 0.171875 -0.875q0 -0.953125 -0.59375 -1.390625q-0.578125 -0.4375 -1.703125 -0.4375q-0.609375 0 -1.25 0.125q-0.625 0.109375 -1.359375 0.34375l0 -1.390625q0.3125 -0.109375 0.65625 -0.1875q0.359375 -0.09375 0.703125 -0.140625q0.359375 -0.0625 0.703125 -0.09375q0.34375 -0.03125 0.65625 -0.03125q0.953125 0 1.671875 0.203125q0.71875 0.203125 1.203125 0.59375q0.484375 0.375 0.71875 0.921875q0.25 0.546875 0.25 1.21875q0 1.03125 -0.53125 1.71875q-0.515625 0.6875 -1.421875 1.109375q0.46875 0.0625 0.90625 0.28125q0.453125 0.21875 0.8125 0.5625q0.375 0.34375 0.59375 0.8125q0.21875 0.453125 0.21875 1.03125z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m922.6127 304.18686q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.5 0 0.84375 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.84375 0.34375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m60.98294 326.47748l0 -13.359375l4.5 0q1.125 0 2.0625 0.515625q0.953125 0.5 1.515625 1.390625q0.578125 0.875 0.578125 2.046875q0 0.71875 -0.265625 1.421875q-0.265625 0.6875 -0.8125 1.265625q-0.546875 0.578125 -1.40625 0.921875q-0.84375 0.34375 -2.015625 0.34375l-3.328125 0l0 -1.46875l3.578125 0q0.703125 0 1.296875 -0.3125q0.59375 -0.328125 0.96875 -0.875q0.375 -0.5625 0.375 -1.296875q0 -0.59375 -0.3125 -1.15625q-0.3125 -0.5625 -0.875 -0.921875q-0.546875 -0.359375 -1.328125 -0.359375l-2.953125 0l0 11.84375l-1.578125 0zm3.21875 -6.125l1.734375 -0.09375l4.21875 6.140625l0 0.078125l-1.859375 0l-4.09375 -6.125zm11.433411 6.421875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm5.33284 -9.8125l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm9.24704 2.796875q-1.75 0 -2.640625 -1.015625q-0.890625 -1.03125 -0.890625 -2.8125l0 -5.984375l1.59375 0l0 5.75q0 1.421875 0.65625 2.03125q0.65625 0.59375 1.625 0.59375q0.828125 0 1.4375 -0.4375q0.625 -0.4375 0.953125 -1.125q0.34375 -0.6875 0.34375 -1.4375l0 -5.375l1.59375 0l0 9.515625l-1.515625 0l0 -1.375l-0.078125 0q-0.234375 0.4375 -0.71875 0.828125q-0.46875 0.375 -1.078125 0.609375q-0.609375 0.234375 -1.28125 0.234375zm7.013031 -0.296875l0 -9.515625l1.515625 0l0 1.53125l0.078125 0q0.1875 -0.546875 0.625 -0.9375q0.4375 -0.40625 1.0 -0.640625q0.578125 -0.234375 1.125 -0.234375q0.4375 0 0.671875 0.046875q0.25 0.046875 0.453125 0.140625l0 1.71875q-0.296875 -0.15625 -0.640625 -0.21875q-0.34375 -0.078125 -0.703125 -0.078125q-0.6875 0 -1.265625 0.390625q-0.578125 0.390625 -0.921875 1.046875q-0.34375 0.65625 -0.34375 1.4375l0 5.3125l-1.59375 0zm6.9960938 0l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.138153 -9.515625l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm5.887253 2.5l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.3125l0.078125 0q0.375 -0.71875 1.21875 -1.203125q0.84375 -0.5 1.875 -0.5q1.21875 0 2.0 0.484375q0.78125 0.484375 1.171875 1.3125q0.390625 0.828125 0.390625 1.921875l0 6.09375l-1.59375 0l0 -5.859375q0 -0.90625 -0.328125 -1.453125q-0.3125 -0.5625 -0.84375 -0.8125q-0.53125 -0.25 -1.171875 -0.25q-0.78125 0 -1.40625 0.453125q-0.609375 0.4375 -0.96875 1.15625q-0.34375 0.703125 -0.34375 1.46875l0 5.296875l-1.59375 0zm14.7565 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm14.987961 0q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm8.557465 1.140625l-3.0625 -9.515625l1.640625 0l2.265625 7.515625l0.015625 0l2.40625 -7.515625l1.625 0l2.40625 7.5l0.015625 0l2.265625 -7.5l1.59375 0l-3.078125 9.515625l-1.59375 0l-2.46875 -7.59375l0 0l-2.4375 7.59375l-1.59375 0zm12.480286 0l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.700531 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm6.2690887 -0.296875l0 -9.515625l1.515625 0l0 1.53125l0.078125 0q0.1875 -0.546875 0.625 -0.9375q0.4375 -0.40625 1.0 -0.640625q0.578125 -0.234375 1.125 -0.234375q0.4375 0 0.671875 0.046875q0.25 0.046875 0.453125 0.140625l0 1.71875q-0.296875 -0.15625 -0.640625 -0.21875q-0.34375 -0.078125 -0.703125 -0.078125q-0.6875 0 -1.265625 0.390625q-0.578125 0.390625 -0.921875 1.046875q-0.34375 0.65625 -0.34375 1.4375l0 5.3125l-1.59375 0zm9.950882 0.296875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm5.496277 -0.296875l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.3125l0.078125 0q0.375 -0.71875 1.21875 -1.203125q0.84375 -0.5 1.875 -0.5q1.21875 0 2.0 0.484375q0.78125 0.484375 1.171875 1.3125q0.390625 0.828125 0.390625 1.921875l0 6.09375l-1.59375 0l0 -5.859375q0 -0.90625 -0.328125 -1.453125q-0.3125 -0.5625 -0.84375 -0.8125q-0.53125 -0.25 -1.171875 -0.25q-0.78125 0 -1.40625 0.453125q-0.609375 0.4375 -0.96875 1.15625q-0.34375 0.703125 -0.34375 1.46875l0 5.296875l-1.59375 0zm10.678375 0l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.2788086 15.296875l0 -13.546875l1.515625 0l0 1.421875l0.078125 0q0.25 -0.453125 0.71875 -0.84375q0.484375 -0.390625 1.125 -0.625q0.640625 -0.25 1.390625 -0.25q1.3125 0 2.328125 0.65625q1.03125 0.65625 1.625 1.796875q0.609375 1.125 0.609375 2.609375q0 1.46875 -0.609375 2.609375q-0.59375 1.140625 -1.625 1.796875q-1.015625 0.640625 -2.328125 0.640625q-1.125 0 -1.984375 -0.515625q-0.859375 -0.53125 -1.25 -1.1875l-0.078125 0l0.078125 1.3125l0 4.125l-1.59375 0zm4.671875 -5.171875q0.8125 0 1.53125 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.4375 -0.828125 0.4375 -1.921875q0 -1.125 -0.4375 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.53125 -0.4375q-0.828125 0 -1.546875 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.4375 0.8125 -0.4375 1.9375q0 1.109375 0.4375 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.546875 0.4375zm15.278061 1.4375q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm5.7922516 -6.9375l0 -1.4375l5.7968597 0l0 1.4375l-5.7968597 0zm1.7656097 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m276.23434 313.60248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916687 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.507294 9.140625l-7.46875 0l0 -1.46875l3.0625 0l0 -8.703125l-2.859375 1.546875l-0.578125 -1.34375l3.78125 -1.984375l1.40625 0l0 10.484375l2.65625 0l0 1.46875z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m298.64685 328.52435l-0.765625 -0.5l1.0 -1.546875q-0.296875 -0.0625 -0.59375 -0.328125q-0.28125 -0.28125 -0.28125 -0.75q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.859375 -0.34375q0.25 0 0.46875 0.09375q0.21875 0.078125 0.375 0.25q0.15625 0.15625 0.25 0.375q0.09375 0.203125 0.09375 0.453125q0 0.21875 -0.078125 0.4375q-0.0625 0.21875 -0.1875 0.421875q-0.109375 0.203125 -0.25 0.40625l-1.234375 1.859375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m315.77466 313.60248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916687 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.616669 9.140625l-7.84375 0l0 -1.421875l3.078125 -3.0625q0.75 -0.75 1.21875 -1.296875q0.484375 -0.546875 0.75 -0.984375q0.265625 -0.453125 0.34375 -0.859375q0.09375 -0.40625 0.09375 -0.859375q0 -0.4375 -0.125 -0.828125q-0.109375 -0.40625 -0.359375 -0.703125q-0.234375 -0.3125 -0.640625 -0.484375q-0.390625 -0.171875 -0.9375 -0.171875q-0.75 0 -1.375 0.34375q-0.609375 0.328125 -1.125 0.875l-0.875 -1.046875q0.671875 -0.71875 1.546875 -1.140625q0.875 -0.421875 2.046875 -0.421875q0.796875 0 1.4375 0.234375q0.65625 0.234375 1.125 0.6875q0.484375 0.4375 0.734375 1.09375q0.265625 0.640625 0.265625 1.453125q0 0.671875 -0.1875 1.265625q-0.171875 0.578125 -0.546875 1.15625q-0.375 0.578125 -0.9375 1.203125q-0.546875 0.609375 -1.3125 1.359375l-2.171875 2.09375l5.796875 0l0 1.515625z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m345.0154 326.77435q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189636 1.09375l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.513031 0.296875q-1.28125 0 -2.328125 -0.640625q-1.03125 -0.65625 -1.625 -1.796875q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.484375 0.59375 -2.609375q0.59375 -1.140625 1.625 -1.796875q1.046875 -0.65625 2.328125 -0.65625q0.765625 0 1.390625 0.25q0.640625 0.234375 1.109375 0.625q0.484375 0.390625 0.75 0.84375l0.078125 0l-0.078125 -1.328125l0 -3.9375l1.59375 0l0 13.359375l-1.515625 0l0 -1.40625l-0.078125 0q-0.265625 0.4375 -0.75 0.828125q-0.46875 0.390625 -1.109375 0.625q-0.625 0.25 -1.390625 0.25zm0.171875 -1.4375q0.828125 0 1.546875 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.453125 -0.8125 0.453125 -1.921875q0 -1.125 -0.453125 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.546875 -0.4375q-0.8125 0 -1.53125 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.9375q0 1.09375 0.453125 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.53125 0.4375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m386.80673 313.60248l-8.40625 12.875l-1.484375 0l8.375 -12.875l1.515625 0zm-4.859375 2.546875q0 0.59375 -0.171875 1.109375q-0.171875 0.5 -0.5 0.875q-0.3125 0.359375 -0.78125 0.578125q-0.453125 0.21875 -1.03125 0.21875q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.515625q-0.3125 -0.34375 -0.484375 -0.84375q-0.15625 -0.5 -0.15625 -1.15625q0 -0.59375 0.171875 -1.09375q0.171875 -0.5 0.484375 -0.875q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.21875 1.03125 -0.21875q0.5625 0 1.0 0.171875q0.4375 0.171875 0.75 0.515625q0.3125 0.328125 0.484375 0.828125q0.171875 0.5 0.171875 1.15625zm-1.34375 0.046875q0 -0.796875 -0.296875 -1.171875q-0.28125 -0.390625 -0.796875 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.34375 0.328125q-0.140625 0.203125 -0.21875 0.5q-0.078125 0.28125 -0.078125 0.609375q0 0.8125 0.296875 1.203125q0.296875 0.375 0.8125 0.375q0.265625 0 0.46875 -0.125q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.203125 0.203125 -0.484375q0.078125 -0.296875 0.078125 -0.640625zm6.046875 7.65625q0 0.578125 -0.171875 1.09375q-0.171875 0.5 -0.5 0.875q-0.328125 0.359375 -0.796875 0.578125q-0.453125 0.203125 -1.015625 0.203125q-0.546875 0 -1.0 -0.171875q-0.4375 -0.171875 -0.75 -0.5q-0.3125 -0.34375 -0.484375 -0.84375q-0.171875 -0.515625 -0.171875 -1.171875q0 -0.578125 0.171875 -1.078125q0.1875 -0.515625 0.5 -0.890625q0.328125 -0.375 0.78125 -0.578125q0.46875 -0.203125 1.03125 -0.203125q0.5625 0 1.0 0.171875q0.453125 0.171875 0.765625 0.515625q0.3125 0.328125 0.46875 0.84375q0.171875 0.5 0.171875 1.15625zm-1.328125 0.03125q0 -0.796875 -0.296875 -1.171875q-0.296875 -0.390625 -0.8125 -0.390625q-0.265625 0 -0.46875 0.125q-0.203125 0.109375 -0.359375 0.328125q-0.140625 0.203125 -0.21875 0.484375q-0.0625 0.28125 -0.0625 0.625q0 0.796875 0.28125 1.1875q0.296875 0.375 0.828125 0.375q0.265625 0 0.46875 -0.109375q0.203125 -0.125 0.34375 -0.328125q0.140625 -0.21875 0.21875 -0.5q0.078125 -0.296875 0.078125 -0.625zm3.4916687 -6.546875l1.453125 0l0.046875 1.6875q0.8125 -0.984375 1.59375 -1.421875q0.796875 -0.4375 1.59375 -0.4375q1.421875 0 2.15625 0.921875q0.734375 0.921875 0.671875 2.734375l-1.59375 0q0.015625 -1.203125 -0.359375 -1.734375q-0.375 -0.546875 -1.109375 -0.546875q-0.3125 0 -0.640625 0.109375q-0.328125 0.109375 -0.671875 0.359375q-0.328125 0.25 -0.71875 0.640625q-0.375 0.390625 -0.8125 0.953125l0 5.875l-1.609375 0l0 -9.140625zm17.382294 5.53125q0 0.78125 -0.328125 1.46875q-0.3125 0.6875 -0.953125 1.203125q-0.625 0.515625 -1.59375 0.8125q-0.96875 0.28125 -2.25 0.28125q-0.71875 0 -1.28125 -0.046875q-0.5625 -0.03125 -1.046875 -0.109375l0 -1.421875q0.5625 0.09375 1.1875 0.15625q0.625 0.046875 1.28125 0.046875q0.890625 0 1.515625 -0.15625q0.640625 -0.15625 1.046875 -0.453125q0.40625 -0.296875 0.578125 -0.71875q0.1875 -0.4375 0.1875 -0.984375q0 -0.484375 -0.21875 -0.84375q-0.21875 -0.375 -0.625 -0.609375q-0.390625 -0.25 -0.9375 -0.375q-0.546875 -0.125 -1.203125 -0.125l-1.359375 0l0 -1.296875l1.375 0q0.53125 0 0.96875 -0.140625q0.453125 -0.140625 0.765625 -0.40625q0.3125 -0.265625 0.484375 -0.65625q0.171875 -0.390625 0.171875 -0.875q0 -0.953125 -0.59375 -1.390625q-0.578125 -0.4375 -1.703125 -0.4375q-0.609375 0 -1.25 0.125q-0.625 0.109375 -1.359375 0.34375l0 -1.390625q0.3125 -0.109375 0.65625 -0.1875q0.359375 -0.09375 0.703125 -0.140625q0.359375 -0.0625 0.703125 -0.09375q0.34375 -0.03125 0.65625 -0.03125q0.953125 0 1.671875 0.203125q0.71875 0.203125 1.203125 0.59375q0.484375 0.375 0.71875 0.921875q0.25 0.546875 0.25 1.21875q0 1.03125 -0.53125 1.71875q-0.515625 0.6875 -1.421875 1.109375q0.46875 0.0625 0.90625 0.28125q0.453125 0.21875 0.8125 0.5625q0.375 0.34375 0.59375 0.8125q0.21875 0.453125 0.21875 1.03125z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m409.71924 326.58685q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.5 0 0.84375 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.84375 0.34375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m416.90955 326.47748l2.359375 -13.359375l1.8125 0l4.9375 10.640625l0.078125 0l0.390625 -2.578125l1.421875 -8.0625l1.59375 0l-2.375 13.359375l-1.515625 0l-5.171875 -11.140625l-0.0625 0l-0.375 2.578125l-1.53125 8.5625l-1.5625 0zm17.300201 0.296875q-1.28125 0 -2.25 -0.609375q-0.953125 -0.609375 -1.5 -1.625q-0.53125 -1.03125 -0.53125 -2.296875q0 -1.15625 0.390625 -2.15625q0.390625 -1.015625 1.109375 -1.78125q0.71875 -0.78125 1.734375 -1.203125q1.015625 -0.4375 2.265625 -0.4375q1.3125 0 2.265625 0.609375q0.96875 0.609375 1.5 1.640625q0.53125 1.015625 0.53125 2.28125q0 1.15625 -0.40625 2.171875q-0.390625 1.015625 -1.109375 1.78125q-0.703125 0.765625 -1.734375 1.203125q-1.015625 0.421875 -2.265625 0.421875zm0.1875 -1.4375q0.84375 0 1.515625 -0.3125q0.6875 -0.3125 1.1875 -0.875q0.5 -0.5625 0.75 -1.296875q0.265625 -0.734375 0.265625 -1.5625q0 -0.90625 -0.359375 -1.625q-0.34375 -0.71875 -1.0 -1.140625q-0.640625 -0.421875 -1.515625 -0.421875q-1.09375 0 -1.9375 0.5625q-0.84375 0.5625 -1.3125 1.484375q-0.46875 0.90625 -0.46875 2.0q0 0.90625 0.359375 1.625q0.359375 0.71875 1.0 1.140625q0.640625 0.421875 1.515625 0.421875zm7.0548096 -8.375l5.578125 0l-0.265625 1.4375l-5.578125 0l0.265625 -1.4375zm0.453125 6.859375l1.6875 -9.546875l1.578125 0l-1.65625 9.390625q-0.140625 0.6875 0.109375 1.09375q0.25 0.390625 0.90625 0.390625q0.3125 0 0.59375 -0.09375q0.296875 -0.09375 0.546875 -0.21875l-0.265625 1.546875q-0.28125 0.109375 -0.578125 0.171875q-0.296875 0.078125 -0.765625 0.078125q-0.734375 0 -1.28125 -0.328125q-0.546875 -0.328125 -0.796875 -0.953125q-0.234375 -0.640625 -0.078125 -1.53125zm9.4800415 2.953125q-1.328125 0 -2.28125 -0.59375q-0.953125 -0.59375 -1.46875 -1.578125q-0.515625 -1.0 -0.515625 -2.1875q0 -1.109375 0.375 -2.140625q0.390625 -1.03125 1.09375 -1.84375q0.703125 -0.8125 1.703125 -1.28125q1.0 -0.484375 2.21875 -0.484375q1.28125 0 2.15625 0.5625q0.890625 0.5625 1.34375 1.46875q0.46875 0.890625 0.46875 1.984375q0 0.328125 -0.046875 0.703125q-0.03125 0.375 -0.125 0.640625l-8.234375 0l0.265625 -1.3125l6.546875 0q0.0625 -0.734375 -0.25 -1.3125q-0.296875 -0.59375 -0.890625 -0.9375q-0.578125 -0.359375 -1.40625 -0.359375q-1.09375 0 -1.90625 0.59375q-0.8125 0.578125 -1.265625 1.609375q-0.4375 1.015625 -0.4375 2.3125q0 0.859375 0.390625 1.46875q0.390625 0.59375 1.015625 0.921875q0.625 0.328125 1.3125 0.328125q0.765625 0 1.359375 -0.203125q0.59375 -0.21875 1.03125 -0.625q0.453125 -0.40625 0.78125 -0.9375l1.21875 0.6875q-0.46875 0.8125 -1.140625 1.375q-0.65625 0.5625 -1.5 0.84375q-0.828125 0.296875 -1.8125 0.296875zm8.522125 -7.53125q-0.5 0 -0.859375 -0.34375q-0.34375 -0.34375 -0.34375 -0.859375q0 -0.5 0.34375 -0.84375q0.359375 -0.34375 0.859375 -0.34375q0.484375 0 0.828125 0.34375q0.34375 0.34375 0.34375 0.84375q0 0.515625 -0.34375 0.859375q-0.34375 0.34375 -0.828125 0.34375zm-1.3125 7.34375q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.484375 0 0.828125 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.828125 0.34375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m473.11496 326.77435q-1.46875 0 -2.71875 -0.53125q-1.25 -0.546875 -2.1875 -1.5q-0.921875 -0.96875 -1.4375 -2.21875q-0.5 -1.265625 -0.5 -2.71875q0 -1.46875 0.5 -2.71875q0.515625 -1.265625 1.4375 -2.21875q0.9375 -0.96875 2.1875 -1.5q1.25 -0.546875 2.71875 -0.546875q1.46875 0 2.71875 0.546875q1.25 0.53125 2.171875 1.5q0.921875 0.953125 1.4375 2.21875q0.515625 1.25 0.515625 2.71875q0 1.453125 -0.515625 2.71875q-0.515625 1.25 -1.4375 2.21875q-0.921875 0.953125 -2.171875 1.5q-1.25 0.53125 -2.71875 0.53125zm0 -1.515625q1.453125 0 2.640625 -0.6875q1.1875 -0.6875 1.890625 -1.921875q0.703125 -1.25 0.703125 -2.84375q0 -1.625 -0.703125 -2.859375q-0.703125 -1.234375 -1.890625 -1.921875q-1.1875 -0.6875 -2.640625 -0.6875q-1.4375 0 -2.640625 0.6875q-1.1875 0.6875 -1.90625 1.921875q-0.703125 1.234375 -0.703125 2.859375q0 1.59375 0.703125 2.84375q0.71875 1.234375 1.90625 1.921875q1.203125 0.6875 2.640625 0.6875zm10.872742 1.21875l-3.0625 -9.515625l1.640625 0l2.265625 7.515625l0.015625 0l2.40625 -7.515625l1.625 0l2.40625 7.5l0.015625 0l2.265625 -7.5l1.59375 0l-3.078125 9.515625l-1.59375 0l-2.46875 -7.59375l0 0l-2.4375 7.59375l-1.59375 0zm12.480286 0l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.700531 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm6.269043 -0.296875l0 -9.515625l1.515625 0l0 1.53125l0.078125 0q0.1875 -0.546875 0.625 -0.9375q0.4375 -0.40625 1.0 -0.640625q0.578125 -0.234375 1.125 -0.234375q0.4375 0 0.671875 0.046875q0.25 0.046875 0.453125 0.140625l0 1.71875q-0.296875 -0.15625 -0.640625 -0.21875q-0.34375 -0.078125 -0.703125 -0.078125q-0.6875 0 -1.265625 0.390625q-0.578125 0.390625 -0.921875 1.046875q-0.34375 0.65625 -0.34375 1.4375l0 5.3125l-1.59375 0zm9.950928 0.296875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm5.496277 -0.296875l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.3125l0.078125 0q0.375 -0.71875 1.21875 -1.203125q0.84375 -0.5 1.875 -0.5q1.21875 0 2.0 0.484375q0.78125 0.484375 1.171875 1.3125q0.390625 0.828125 0.390625 1.921875l0 6.09375l-1.59375 0l0 -5.859375q0 -0.90625 -0.328125 -1.453125q-0.3125 -0.5625 -0.84375 -0.8125q-0.53125 -0.25 -1.171875 -0.25q-0.78125 0 -1.40625 0.453125q-0.609375 0.4375 -0.96875 1.15625q-0.34375 0.703125 -0.34375 1.46875l0 5.296875l-1.59375 0zm10.678345 0l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.2788086 15.296875l0 -13.546875l1.515625 0l0 1.421875l0.078125 0q0.25 -0.453125 0.71875 -0.84375q0.484375 -0.390625 1.125 -0.625q0.640625 -0.25 1.390625 -0.25q1.3125 0 2.328125 0.65625q1.03125 0.65625 1.625 1.796875q0.609375 1.125 0.609375 2.609375q0 1.46875 -0.609375 2.609375q-0.59375 1.140625 -1.625 1.796875q-1.015625 0.640625 -2.328125 0.640625q-1.125 0 -1.984375 -0.515625q-0.859375 -0.53125 -1.25 -1.1875l-0.078125 0l0.078125 1.3125l0 4.125l-1.59375 0zm4.671875 -5.171875q0.8125 0 1.53125 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.4375 -0.828125 0.4375 -1.921875q0 -1.125 -0.4375 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.53125 -0.4375q-0.828125 0 -1.546875 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.4375 0.8125 -0.4375 1.9375q0 1.109375 0.4375 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.546875 0.4375zm11.043701 1.140625l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm6.6069336 11.5625q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm12.871399 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m594.73926 330.50873l0 -13.546875l2.296875 0l0 1.15625l0.140625 0q0.25 -0.375 0.640625 -0.703125q0.40625 -0.34375 0.96875 -0.546875q0.5625 -0.203125 1.25 -0.203125q1.296875 0 2.328125 0.640625q1.046875 0.640625 1.640625 1.78125q0.609375 1.140625 0.609375 2.640625q0 1.484375 -0.609375 2.625q-0.59375 1.140625 -1.640625 1.78125q-1.03125 0.640625 -2.328125 0.640625q-1.03125 0 -1.765625 -0.4375q-0.734375 -0.4375 -1.09375 -1.0l-0.140625 0l0.140625 1.34375l0 3.828125l-2.4375 0zm4.859375 -5.984375q0.6875 0 1.28125 -0.34375q0.59375 -0.359375 0.953125 -0.984375q0.359375 -0.640625 0.359375 -1.46875q0 -0.859375 -0.359375 -1.46875q-0.359375 -0.625 -0.953125 -0.96875q-0.59375 -0.34375 -1.28125 -0.34375q-0.671875 0 -1.265625 0.34375q-0.578125 0.34375 -0.9375 0.96875q-0.359375 0.625 -0.359375 1.46875q0 0.828125 0.359375 1.453125q0.359375 0.625 0.9375 0.984375q0.59375 0.359375 1.265625 0.359375zm11.369324 2.25q-1.4375 0 -2.578125 -0.640625q-1.125 -0.65625 -1.765625 -1.796875q-0.640625 -1.140625 -0.640625 -2.59375q0 -1.359375 0.625 -2.515625q0.640625 -1.15625 1.734375 -1.859375q1.109375 -0.703125 2.53125 -0.703125q1.53125 0 2.5625 0.65625q1.046875 0.65625 1.59375 1.75q0.546875 1.078125 0.546875 2.40625q0 0.296875 -0.03125 0.515625q-0.015625 0.21875 -0.03125 0.3125l-8.171875 0l0 -1.734375l5.875 0q-0.015625 -0.328125 -0.1875 -0.65625q-0.15625 -0.34375 -0.46875 -0.625q-0.296875 -0.28125 -0.71875 -0.453125q-0.40625 -0.171875 -0.953125 -0.171875q-0.703125 0 -1.296875 0.375q-0.59375 0.359375 -0.9375 1.03125q-0.328125 0.65625 -0.328125 1.578125q0 0.953125 0.359375 1.609375q0.359375 0.65625 0.953125 1.015625q0.609375 0.34375 1.359375 0.34375q0.875 0 1.46875 -0.375q0.609375 -0.390625 0.9375 -0.96875l1.984375 0.96875q-0.671875 1.171875 -1.765625 1.859375q-1.078125 0.671875 -2.65625 0.671875zm6.482483 -0.296875l0 -9.515625l2.296875 0l0 1.265625l0.15625 0q0.234375 -0.4375 0.640625 -0.78125q0.40625 -0.359375 0.9375 -0.5625q0.546875 -0.203125 1.15625 -0.203125q0.375 0 0.671875 0.0625q0.296875 0.046875 0.53125 0.140625l0 2.453125q-0.390625 -0.15625 -0.78125 -0.25q-0.375 -0.109375 -0.8125 -0.109375q-0.71875 0 -1.25 0.375q-0.53125 0.359375 -0.8125 0.984375q-0.28125 0.625 -0.28125 1.40625l0 4.734375l-2.453125 0zm7.9505615 -4.703125l0 -2.15625l6.328125 0l0 2.15625l-6.328125 0zm11.612671 5.0q-1.015625 0 -1.796875 -0.390625q-0.765625 -0.40625 -1.21875 -1.125q-0.4375 -0.734375 -0.4375 -1.6875q0 -1.0 0.5 -1.703125q0.515625 -0.703125 1.390625 -1.09375q0.875 -0.390625 1.9375 -0.390625q0.578125 0 1.046875 0.0625q0.46875 0.0625 0.84375 0.15625q0.375 0.09375 0.625 0.203125l0 -0.40625q0 -0.765625 -0.578125 -1.25q-0.578125 -0.5 -1.59375 -0.5q-0.671875 0 -1.296875 0.3125q-0.625 0.296875 -1.03125 0.796875l-1.5625 -1.234375q0.46875 -0.59375 1.078125 -1.015625q0.609375 -0.421875 1.359375 -0.625q0.75 -0.21875 1.5625 -0.21875q2.1875 0 3.328125 1.015625q1.140625 1.0 1.140625 2.9375l0 5.859375l-2.40625 0l0 -0.96875l-0.15625 0q-0.265625 0.328125 -0.65625 0.625q-0.390625 0.28125 -0.921875 0.453125q-0.515625 0.1875 -1.15625 0.1875zm0.578125 -1.90625q0.734375 0 1.25 -0.3125q0.515625 -0.328125 0.78125 -0.84375q0.28125 -0.53125 0.28125 -1.109375q-0.421875 -0.203125 -0.9375 -0.328125q-0.515625 -0.125 -1.0625 -0.125q-1.078125 0 -1.5 0.421875q-0.40625 0.40625 -0.40625 1.0q0 0.5625 0.40625 0.9375q0.40625 0.359375 1.1875 0.359375zm6.9974365 1.609375l0 -13.359375l2.4375 0l0 13.359375l-2.4375 0zm4.8692627 0l0 -13.359375l2.4375 0l0 13.359375l-2.4375 0zm9.30719 0.296875q-1.5 0 -2.65625 -0.65625q-1.140625 -0.671875 -1.78125 -1.8125q-0.640625 -1.15625 -0.640625 -2.578125q0 -1.4375 0.640625 -2.578125q0.640625 -1.15625 1.78125 -1.8125q1.15625 -0.671875 2.65625 -0.671875q1.484375 0 2.625 0.671875q1.140625 0.671875 1.78125 1.8125q0.640625 1.140625 0.640625 2.578125q0 1.421875 -0.640625 2.578125q-0.640625 1.140625 -1.78125 1.8125q-1.140625 0.65625 -2.625 0.65625zm0 -2.28125q0.703125 0 1.28125 -0.3125q0.59375 -0.328125 0.953125 -0.953125q0.375 -0.625 0.375 -1.5q0 -0.890625 -0.375 -1.5q-0.359375 -0.609375 -0.953125 -0.9375q-0.578125 -0.34375 -1.265625 -0.34375q-0.703125 0 -1.3125 0.34375q-0.609375 0.328125 -0.984375 0.9375q-0.359375 0.609375 -0.359375 1.5q0 0.875 0.359375 1.5q0.375 0.625 0.984375 0.953125q0.609375 0.3125 1.296875 0.3125zm11.395447 2.28125q-1.453125 0 -2.59375 -0.640625q-1.125 -0.640625 -1.765625 -1.78125q-0.640625 -1.15625 -0.640625 -2.625q0 -1.484375 0.640625 -2.609375q0.640625 -1.140625 1.765625 -1.796875q1.140625 -0.65625 2.59375 -0.65625q1.578125 0 2.71875 0.671875q1.140625 0.671875 1.640625 1.9375l-2.234375 0.9375q-0.296875 -0.65625 -0.84375 -0.953125q-0.53125 -0.3125 -1.28125 -0.3125q-0.734375 0 -1.3125 0.34375q-0.578125 0.34375 -0.921875 0.96875q-0.328125 0.609375 -0.328125 1.46875q0 0.84375 0.328125 1.46875q0.34375 0.625 0.921875 0.984375q0.578125 0.34375 1.3125 0.34375q0.828125 0 1.359375 -0.359375q0.546875 -0.375 0.875 -1.015625l2.1875 0.984375q-0.5625 1.1875 -1.703125 1.921875q-1.140625 0.71875 -2.71875 0.71875zm9.034241 0q-1.015625 0 -1.796875 -0.390625q-0.765625 -0.40625 -1.21875 -1.125q-0.4375 -0.734375 -0.4375 -1.6875q0 -1.0 0.5 -1.703125q0.515625 -0.703125 1.390625 -1.09375q0.875 -0.390625 1.9375 -0.390625q0.578125 0 1.046875 0.0625q0.46875 0.0625 0.84375 0.15625q0.375 0.09375 0.625 0.203125l0 -0.40625q0 -0.765625 -0.578125 -1.25q-0.578125 -0.5 -1.59375 -0.5q-0.671875 0 -1.296875 0.3125q-0.625 0.296875 -1.03125 0.796875l-1.5625 -1.234375q0.46875 -0.59375 1.078125 -1.015625q0.609375 -0.421875 1.359375 -0.625q0.75 -0.21875 1.5625 -0.21875q2.1875 0 3.328125 1.015625q1.140625 1.0 1.140625 2.9375l0 5.859375l-2.40625 0l0 -0.96875l-0.15625 0q-0.265625 0.328125 -0.65625 0.625q-0.390625 0.28125 -0.921875 0.453125q-0.515625 0.1875 -1.15625 0.1875zm0.578125 -1.90625q0.734375 0 1.25 -0.3125q0.515625 -0.328125 0.78125 -0.84375q0.28125 -0.53125 0.28125 -1.109375q-0.421875 -0.203125 -0.9375 -0.328125q-0.515625 -0.125 -1.0625 -0.125q-1.078125 0 -1.5 0.421875q-0.40625 0.40625 -0.40625 1.0q0 0.5625 0.40625 0.9375q0.40625 0.359375 1.1875 0.359375zm6.107605 -7.90625l6.4375 0l0 2.09375l-6.4375 0l0 -2.09375zm1.671875 6.46875l0 -9.15625l2.4375 0l0 8.71875q0 0.65625 0.28125 1.0q0.28125 0.34375 0.9375 0.34375q0.375 0 0.625 -0.109375q0.25 -0.125 0.5625 -0.3125l0 2.390625q-0.390625 0.15625 -0.8125 0.234375q-0.421875 0.09375 -0.890625 0.09375q-1.46875 0 -2.3125 -0.84375q-0.828125 -0.859375 -0.828125 -2.359375zm7.0479126 3.046875l0 -9.515625l2.4375 0l0 9.515625l-2.4375 0zm1.234375 -10.5625q-0.65625 0 -1.125 -0.46875q-0.46875 -0.46875 -0.46875 -1.109375q0 -0.65625 0.46875 -1.109375q0.46875 -0.46875 1.125 -0.46875q0.65625 0 1.109375 0.46875q0.453125 0.453125 0.453125 1.109375q0 0.640625 -0.453125 1.109375q-0.453125 0.46875 -1.109375 0.46875zm8.208984 10.859375q-1.5 0 -2.65625 -0.65625q-1.140625 -0.671875 -1.78125 -1.8125q-0.640625 -1.15625 -0.640625 -2.578125q0 -1.4375 0.640625 -2.578125q0.640625 -1.15625 1.78125 -1.8125q1.15625 -0.671875 2.65625 -0.671875q1.484375 0 2.625 0.671875q1.140625 0.671875 1.78125 1.8125q0.640625 1.140625 0.640625 2.578125q0 1.421875 -0.640625 2.578125q-0.640625 1.140625 -1.78125 1.8125q-1.140625 0.65625 -2.625 0.65625zm0 -2.28125q0.703125 0 1.28125 -0.3125q0.59375 -0.328125 0.953125 -0.953125q0.375 -0.625 0.375 -1.5q0 -0.890625 -0.375 -1.5q-0.359375 -0.609375 -0.953125 -0.9375q-0.578125 -0.34375 -1.265625 -0.34375q-0.703125 0 -1.3125 0.34375q-0.609375 0.328125 -0.984375 0.9375q-0.359375 0.609375 -0.359375 1.5q0 0.875 0.359375 1.5q0.375 0.625 0.984375 0.953125q0.609375 0.3125 1.296875 0.3125zm6.7244263 1.984375l0 -9.515625l2.296875 0l0 1.1875l0.15625 0q0.421875 -0.6875 1.203125 -1.078125q0.796875 -0.40625 1.703125 -0.40625q1.75 0 2.625 1.0625q0.890625 1.0625 0.890625 2.84375l0 5.90625l-2.453125 0l0 -5.609375q0 -0.9375 -0.484375 -1.421875q-0.46875 -0.5 -1.28125 -0.5q-0.671875 0 -1.171875 0.359375q-0.5 0.359375 -0.765625 0.96875q-0.265625 0.609375 -0.265625 1.359375l0 4.84375l-2.453125 0zm15.478149 4.03125l0 -13.546875l2.296875 0l0 1.15625l0.140625 0q0.25 -0.375 0.640625 -0.703125q0.40625 -0.34375 0.96875 -0.546875q0.5625 -0.203125 1.25 -0.203125q1.296875 0 2.328125 0.640625q1.046875 0.640625 1.640625 1.78125q0.609375 1.140625 0.609375 2.640625q0 1.484375 -0.609375 2.625q-0.59375 1.140625 -1.640625 1.78125q-1.03125 0.640625 -2.328125 0.640625q-1.03125 0 -1.765625 -0.4375q-0.734375 -0.4375 -1.09375 -1.0l-0.140625 0l0.140625 1.34375l0 3.828125l-2.4375 0zm4.859375 -5.984375q0.6875 0 1.28125 -0.34375q0.59375 -0.359375 0.953125 -0.984375q0.359375 -0.640625 0.359375 -1.46875q0 -0.859375 -0.359375 -1.46875q-0.359375 -0.625 -0.953125 -0.96875q-0.59375 -0.34375 -1.28125 -0.34375q-0.671875 0 -1.265625 0.34375q-0.578125 0.34375 -0.9375 0.96875q-0.359375 0.625 -0.359375 1.46875q0 0.828125 0.359375 1.453125q0.359375 0.625 0.9375 0.984375q0.59375 0.359375 1.265625 0.359375zm6.807251 1.953125l0 -9.515625l2.296875 0l0 1.265625l0.15625 0q0.234375 -0.4375 0.640625 -0.78125q0.40625 -0.359375 0.9375 -0.5625q0.546875 -0.203125 1.15625 -0.203125q0.375 0 0.671875 0.0625q0.296875 0.046875 0.53125 0.140625l0 2.453125q-0.390625 -0.15625 -0.78125 -0.25q-0.375 -0.109375 -0.8125 -0.109375q-0.71875 0 -1.25 0.375q-0.53125 0.359375 -0.8125 0.984375q-0.28125 0.625 -0.28125 1.40625l0 4.734375l-2.453125 0zm11.96875 0.296875q-1.5 0 -2.65625 -0.65625q-1.140625 -0.671875 -1.78125 -1.8125q-0.640625 -1.15625 -0.640625 -2.578125q0 -1.4375 0.640625 -2.578125q0.640625 -1.15625 1.78125 -1.8125q1.15625 -0.671875 2.65625 -0.671875q1.484375 0 2.625 0.671875q1.140625 0.671875 1.78125 1.8125q0.640625 1.140625 0.640625 2.578125q0 1.421875 -0.640625 2.578125q-0.640625 1.140625 -1.78125 1.8125q-1.140625 0.65625 -2.625 0.65625zm0 -2.28125q0.703125 0 1.28125 -0.3125q0.59375 -0.328125 0.953125 -0.953125q0.375 -0.625 0.375 -1.5q0 -0.890625 -0.375 -1.5q-0.359375 -0.609375 -0.953125 -0.9375q-0.578125 -0.34375 -1.265625 -0.34375q-0.703125 0 -1.3125 0.34375q-0.609375 0.328125 -0.984375 0.9375q-0.359375 0.609375 -0.359375 1.5q0 0.875 0.359375 1.5q0.375 0.625 0.984375 0.953125q0.609375 0.3125 1.296875 0.3125zm6.755249 6.015625l0 -13.546875l2.296875 0l0 1.15625l0.140625 0q0.25 -0.375 0.640625 -0.703125q0.40625 -0.34375 0.96875 -0.546875q0.5625 -0.203125 1.25 -0.203125q1.296875 0 2.328125 0.640625q1.046875 0.640625 1.640625 1.78125q0.609375 1.140625 0.609375 2.640625q0 1.484375 -0.609375 2.625q-0.59375 1.140625 -1.640625 1.78125q-1.03125 0.640625 -2.328125 0.640625q-1.03125 0 -1.765625 -0.4375q-0.734375 -0.4375 -1.09375 -1.0l-0.140625 0l0.140625 1.34375l0 3.828125l-2.4375 0zm4.859375 -5.984375q0.6875 0 1.28125 -0.34375q0.59375 -0.359375 0.953125 -0.984375q0.359375 -0.640625 0.359375 -1.46875q0 -0.859375 -0.359375 -1.46875q-0.359375 -0.625 -0.953125 -0.96875q-0.59375 -0.34375 -1.28125 -0.34375q-0.671875 0 -1.265625 0.34375q-0.578125 0.34375 -0.9375 0.96875q-0.359375 0.625 -0.359375 1.46875q0 0.828125 0.359375 1.453125q0.359375 0.625 0.9375 0.984375q0.59375 0.359375 1.265625 0.359375zm11.369385 2.25q-1.4375 0 -2.578125 -0.640625q-1.125 -0.65625 -1.765625 -1.796875q-0.640625 -1.140625 -0.640625 -2.59375q0 -1.359375 0.625 -2.515625q0.640625 -1.15625 1.734375 -1.859375q1.109375 -0.703125 2.53125 -0.703125q1.53125 0 2.5625 0.65625q1.046875 0.65625 1.59375 1.75q0.546875 1.078125 0.546875 2.40625q0 0.296875 -0.03125 0.515625q-0.015625 0.21875 -0.03125 0.3125l-8.171875 0l0 -1.734375l5.875 0q-0.015625 -0.328125 -0.1875 -0.65625q-0.15625 -0.34375 -0.46875 -0.625q-0.296875 -0.28125 -0.71875 -0.453125q-0.40625 -0.171875 -0.953125 -0.171875q-0.703125 0 -1.296875 0.375q-0.59375 0.359375 -0.9375 1.03125q-0.328125 0.65625 -0.328125 1.578125q0 0.953125 0.359375 1.609375q0.359375 0.65625 0.953125 1.015625q0.609375 0.34375 1.359375 0.34375q0.875 0 1.46875 -0.375q0.609375 -0.390625 0.9375 -0.96875l1.984375 0.96875q-0.671875 1.171875 -1.765625 1.859375q-1.078125 0.671875 -2.65625 0.671875zm6.482422 -0.296875l0 -9.515625l2.296875 0l0 1.265625l0.15625 0q0.234375 -0.4375 0.640625 -0.78125q0.40625 -0.359375 0.9375 -0.5625q0.546875 -0.203125 1.15625 -0.203125q0.375 0 0.671875 0.0625q0.296875 0.046875 0.53125 0.140625l0 2.453125q-0.390625 -0.15625 -0.78125 -0.25q-0.375 -0.109375 -0.8125 -0.109375q-0.71875 0 -1.25 0.375q-0.53125 0.359375 -0.8125 0.984375q-0.28125 0.625 -0.28125 1.40625l0 4.734375l-2.453125 0zm7.481018 -9.515625l6.4375 0l0 2.09375l-6.4375 0l0 -2.09375zm1.671875 6.46875l0 -9.15625l2.4375 0l0 8.71875q0 0.65625 0.28125 1.0q0.28125 0.34375 0.9375 0.34375q0.375 0 0.625 -0.109375q0.25 -0.125 0.5625 -0.3125l0 2.390625q-0.390625 0.15625 -0.8125 0.234375q-0.421875 0.09375 -0.890625 0.09375q-1.46875 0 -2.3125 -0.84375q-0.828125 -0.859375 -0.828125 -2.359375zm10.641663 6.640625q-0.046875 0.140625 -0.09375 0.234375q-0.046875 0.109375 -0.09375 0.203125l-2.625 0q0.078125 -0.15625 0.28125 -0.640625q0.21875 -0.46875 0.375 -0.796875q0.109375 -0.21875 0.25 -0.578125q0.15625 -0.359375 0.3125 -0.703125q0.15625 -0.34375 0.25 -0.53125l1.4375 -3.1875l0.515625 -1.1875l2.359375 -5.921875l2.734375 0l-4.90625 11.3125q-0.21875 0.484375 -0.375 0.828125q-0.15625 0.359375 -0.25 0.59375q-0.09375 0.234375 -0.171875 0.375zm-0.640625 -3.75l-4.140625 -9.359375l2.765625 0l2.46875 5.921875l0.28125 0l-1.375 3.4375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m807.20776 326.58685q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.5 0 0.84375 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.84375 0.34375zm8.049683 -0.109375l0 -13.359375l1.578125 0l0 13.359375l-1.578125 0zm3.2130737 -8.078125l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0zm13.336487 0.296875q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm15.875305 1.390625q-0.75 0 -1.390625 -0.25q-0.640625 -0.234375 -1.125 -0.625q-0.46875 -0.390625 -0.71875 -0.828125l-0.078125 0l0 1.40625l-1.515625 0l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.328125l0.078125 0q0.25 -0.453125 0.71875 -0.84375q0.484375 -0.390625 1.125 -0.625q0.640625 -0.25 1.390625 -0.25q1.3125 0 2.328125 0.65625q1.03125 0.65625 1.625 1.796875q0.609375 1.125 0.609375 2.609375q0 1.46875 -0.609375 2.609375q-0.59375 1.140625 -1.625 1.796875q-1.015625 0.640625 -2.328125 0.640625zm-0.15625 -1.4375q0.8125 0 1.53125 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.4375 -0.828125 0.4375 -1.921875q0 -1.125 -0.4375 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.53125 -0.4375q-0.828125 0 -1.546875 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.4375 0.8125 -0.4375 1.9375q0 1.109375 0.4375 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.546875 0.4375zm6.4658813 1.140625l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm8.233765 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm11.040955 1.4375q-1.40625 0 -2.5 -0.65625q-1.078125 -0.671875 -1.703125 -1.8125q-0.609375 -1.15625 -0.609375 -2.578125q0 -1.46875 0.609375 -2.59375q0.625 -1.140625 1.703125 -1.796875q1.09375 -0.671875 2.5 -0.671875q1.609375 0 2.640625 0.734375q1.03125 0.734375 1.46875 1.890625l-1.4375 0.609375q-0.359375 -0.890625 -1.0625 -1.34375q-0.6875 -0.453125 -1.6875 -0.453125q-0.828125 0 -1.546875 0.453125q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.921875q0 1.078125 0.453125 1.90625q0.453125 0.8125 1.171875 1.265625q0.71875 0.4375 1.546875 0.4375q1.015625 0 1.734375 -0.46875q0.734375 -0.46875 1.09375 -1.3125l1.40625 0.59375q-0.46875 1.09375 -1.515625 1.859375q-1.03125 0.765625 -2.640625 0.765625zm5.8738403 -0.296875l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm0.59375 -1.9375l0 -2.03125l5.46875 -5.546875l2.015625 0l0 0.078125l-7.484375 7.5zm5.703125 1.9375l-3.4375 -5.109375l1.125 -1.109375l4.203125 6.140625l0 0.078125l-1.890625 0z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m899.187 326.77435q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm8.557495 1.140625l-3.0625 -9.515625l1.640625 0l2.265625 7.515625l0.015625 0l2.40625 -7.515625l1.625 0l2.40625 7.5l0.015625 0l2.265625 -7.5l1.59375 0l-3.078125 9.515625l-1.59375 0l-2.46875 -7.59375l0 0l-2.4375 7.59375l-1.59375 0zm12.480225 0l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm13.700989 0.296875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375z" fill-rule="nonzero"/><path fill="#9aa0a6" d="m63.70169 349.17435q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm11.047211 1.09375l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.25 -0.46875 0.703125 -0.84375q0.46875 -0.390625 1.046875 -0.625q0.59375 -0.234375 1.21875 -0.234375q1.0625 0 1.828125 0.515625q0.765625 0.515625 1.09375 1.328125q0.484375 -0.796875 1.296875 -1.3125q0.828125 -0.53125 1.96875 -0.53125q1.703125 0 2.5 1.03125q0.796875 1.015625 0.796875 2.6875l0 6.09375l-1.5625 0l0 -5.859375q0 -1.375 -0.5625 -1.9375q-0.5625 -0.578125 -1.578125 -0.578125q-0.734375 0 -1.3125 0.421875q-0.5625 0.421875 -0.890625 1.109375q-0.3125 0.671875 -0.3125 1.46875l0 5.375l-1.59375 0l0 -5.84375q0 -1.375 -0.5625 -1.953125q-0.5625 -0.578125 -1.5625 -0.578125q-0.734375 0 -1.296875 0.4375q-0.5625 0.421875 -0.890625 1.109375q-0.328125 0.6875 -0.328125 1.5l0 5.328125l-1.59375 0zm20.558594 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm6.2690887 -0.296875l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.25 -0.46875 0.703125 -0.84375q0.46875 -0.390625 1.046875 -0.625q0.59375 -0.234375 1.21875 -0.234375q1.0625 0 1.828125 0.515625q0.765625 0.515625 1.09375 1.328125q0.484375 -0.796875 1.296875 -1.3125q0.828125 -0.53125 1.96875 -0.53125q1.703125 0 2.5 1.03125q0.796875 1.015625 0.796875 2.6875l0 6.09375l-1.5625 0l0 -5.859375q0 -1.375 -0.5625 -1.9375q-0.5625 -0.578125 -1.578125 -0.578125q-0.734375 0 -1.3125 0.421875q-0.5625 0.421875 -0.890625 1.109375q-0.3125 0.671875 -0.3125 1.46875l0 5.375l-1.59375 0l0 -5.84375q0 -1.375 -0.5625 -1.953125q-0.5625 -0.578125 -1.5625 -0.578125q-0.734375 0 -1.296875 0.4375q-0.5625 0.421875 -0.890625 1.109375q-0.328125 0.6875 -0.328125 1.5l0 5.328125l-1.59375 0zm16.324219 0l0 -9.515625l1.515625 0l0 1.53125l0.078125 0q0.1875 -0.546875 0.625 -0.9375q0.4375 -0.40625 1.0 -0.640625q0.578125 -0.234375 1.125 -0.234375q0.4375 0 0.671875 0.046875q0.25 0.046875 0.453125 0.140625l0 1.71875q-0.296875 -0.15625 -0.640625 -0.21875q-0.34375 -0.078125 -0.703125 -0.078125q-0.6875 0 -1.265625 0.390625q-0.578125 0.390625 -0.921875 1.046875q-0.34375 0.65625 -0.34375 1.4375l0 5.3125l-1.59375 0zm10.670471 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm5.426178 -8.375l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0zm5.2015533 2.046875l-0.765625 -0.5l1.0 -1.546875q-0.296875 -0.0625 -0.59375 -0.328125q-0.28125 -0.28125 -0.28125 -0.75q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.859375 -0.34375q0.25 0 0.46875 0.09375q0.21875 0.078125 0.375 0.25q0.15625 0.15625 0.25 0.375q0.09375 0.203125 0.09375 0.453125q0 0.21875 -0.078125 0.4375q-0.0625 0.21875 -0.1875 0.421875q-0.109375 0.203125 -0.25 0.40625l-1.234375 1.859375zm8.377808 -2.046875l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.5292206 1.75l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm13.169037 2.796875q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189636 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm7.1712494 0.296875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm9.98024 0q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm15.447372 1.4375q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm8.557465 1.140625l-3.0625 -9.515625l1.640625 0l2.265625 7.515625l0.015625 0l2.40625 -7.515625l1.625 0l2.40625 7.5l0.015625 0l2.265625 -7.5l1.59375 0l-3.078125 9.515625l-1.59375 0l-2.46875 -7.59375l0 0l-2.4375 7.59375l-1.59375 0zm12.480286 0l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm13.700943 0.296875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm12.871399 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.1718597 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.8749847 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9374847 0 3.0312347 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.60935974 0.25 -1.3281097 0.25zm0.140625 -1.390625q0.828125 0 1.4999847 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.4062347 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189484 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm3.9364624 0l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm12.655334 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm5.792267 -6.9375l0 -1.4375l5.796875 0l0 1.4375l-5.796875 0zm1.765625 8.078125l0 -10.546875q0 -0.921875 0.40625 -1.578125q0.40625 -0.671875 1.109375 -1.015625q0.703125 -0.359375 1.515625 -0.359375q0.546875 0 0.859375 0.0625q0.328125 0.0625 0.546875 0.15625l0 1.546875q-0.25 -0.09375 -0.5625 -0.1875q-0.3125 -0.109375 -0.78125 -0.109375q-0.578125 0 -1.046875 0.421875q-0.46875 0.421875 -0.46875 1.234375l0 10.375l-1.578125 0zm10.445831 0l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm2.5292358 1.75l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm9.308685 2.796875q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm12.871399 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.7189636 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm4.0927124 0l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm6.2323303 11.5625q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm9.953735 1.390625q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm9.823975 0q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm9.690552 0q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm13.152649 2.234375q-0.5625 -0.640625 -1.15625 -1.546875q-0.578125 -0.890625 -1.0625 -1.984375q-0.484375 -1.09375 -0.796875 -2.34375q-0.3125 -1.265625 -0.3125 -2.65625q0 -1.40625 0.3125 -2.65625q0.3125 -1.265625 0.796875 -2.359375q0.484375 -1.09375 1.0625 -1.984375q0.59375 -0.90625 1.15625 -1.546875l1.15625 0.8125q-0.78125 0.9375 -1.453125 2.171875q-0.65625 1.234375 -1.0625 2.65625q-0.390625 1.40625 -0.390625 2.90625q0 1.484375 0.390625 2.90625q0.40625 1.421875 1.0625 2.65625q0.671875 1.21875 1.453125 2.15625l-1.15625 0.8125zm2.8540955 -2.53125l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.2788086 11.265625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.778656 0.296875q-1.40625 0 -2.5 -0.65625q-1.078125 -0.671875 -1.703125 -1.8125q-0.609375 -1.15625 -0.609375 -2.578125q0 -1.46875 0.609375 -2.59375q0.625 -1.140625 1.703125 -1.796875q1.09375 -0.671875 2.5 -0.671875q1.609375 0 2.640625 0.734375q1.03125 0.734375 1.46875 1.890625l-1.4375 0.609375q-0.359375 -0.890625 -1.0625 -1.34375q-0.6875 -0.453125 -1.6875 -0.453125q-0.828125 0 -1.546875 0.453125q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.921875q0 1.078125 0.453125 1.90625q0.453125 0.8125 1.171875 1.265625q0.71875 0.4375 1.546875 0.4375q1.015625 0 1.734375 -0.46875q0.734375 -0.46875 1.09375 -1.3125l1.40625 0.59375q-0.46875 1.09375 -1.515625 1.859375q-1.03125 0.765625 -2.640625 0.765625zm5.87381 -0.296875l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm7.3895874 0.296875q-1.75 0 -2.640625 -1.015625q-0.890625 -1.03125 -0.890625 -2.8125l0 -5.984375l1.59375 0l0 5.75q0 1.421875 0.65625 2.03125q0.65625 0.59375 1.625 0.59375q0.828125 0 1.4375 -0.4375q0.625 -0.4375 0.953125 -1.125q0.34375 -0.6875 0.34375 -1.4375l0 -5.375l1.59375 0l0 9.515625l-1.515625 0l0 -1.375l-0.078125 0q-0.234375 0.4375 -0.71875 0.828125q-0.46875 0.375 -1.078125 0.609375q-0.609375 0.234375 -1.28125 0.234375zm10.966583 0q-1.28125 0 -2.328125 -0.640625q-1.03125 -0.65625 -1.625 -1.796875q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.484375 0.59375 -2.609375q0.59375 -1.140625 1.625 -1.796875q1.046875 -0.65625 2.328125 -0.65625q0.765625 0 1.390625 0.25q0.640625 0.234375 1.109375 0.625q0.484375 0.390625 0.75 0.84375l0.078125 0l-0.078125 -1.328125l0 -3.9375l1.59375 0l0 13.359375l-1.515625 0l0 -1.40625l-0.078125 0q-0.265625 0.4375 -0.75 0.828125q-0.46875 0.390625 -1.109375 0.625q-0.625 0.25 -1.390625 0.25zm0.171875 -1.4375q0.828125 0 1.546875 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.453125 -0.8125 0.453125 -1.921875q0 -1.125 -0.453125 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.546875 -0.4375q-0.8125 0 -1.53125 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.9375q0 1.09375 0.453125 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.53125 0.4375zm7.168579 1.140625l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm3.2788086 11.265625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm14.560303 4.328125q-1.21875 0 -2.09375 -0.40625q-0.859375 -0.390625 -1.390625 -1.015625q-0.53125 -0.609375 -0.75 -1.25l1.453125 -0.609375q0.296875 0.78125 1.015625 1.3125q0.71875 0.53125 1.765625 0.53125q1.484375 0 2.28125 -0.875q0.8125 -0.859375 0.8125 -2.40625l0 -1.0625l-0.078125 0q-0.453125 0.6875 -1.296875 1.171875q-0.828125 0.46875 -1.953125 0.46875q-1.21875 0 -2.25 -0.625q-1.015625 -0.640625 -1.609375 -1.765625q-0.59375 -1.140625 -0.59375 -2.609375q0 -1.46875 0.59375 -2.59375q0.59375 -1.140625 1.609375 -1.765625q1.03125 -0.640625 2.25 -0.640625q1.125 0 1.953125 0.484375q0.84375 0.46875 1.296875 1.171875l0.078125 0l0 -1.359375l1.515625 0l0 9.140625q0 1.59375 -0.609375 2.625q-0.609375 1.046875 -1.65625 1.5625q-1.03125 0.515625 -2.34375 0.515625zm0 -5.578125q0.828125 0 1.53125 -0.421875q0.71875 -0.421875 1.140625 -1.21875q0.421875 -0.796875 0.421875 -1.921875q0 -1.15625 -0.421875 -1.953125q-0.421875 -0.796875 -1.140625 -1.203125q-0.703125 -0.40625 -1.53125 -0.40625q-0.84375 0 -1.5625 0.421875q-0.703125 0.421875 -1.140625 1.21875q-0.421875 0.78125 -0.421875 1.921875q0 1.140625 0.421875 1.9375q0.4375 0.78125 1.140625 1.203125q0.71875 0.421875 1.5625 0.421875zm10.622284 -8.265625l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm5.8872375 2.5l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.3125l0.078125 0q0.375 -0.71875 1.21875 -1.203125q0.84375 -0.5 1.875 -0.5q1.21875 0 2.0 0.484375q0.78125 0.484375 1.171875 1.3125q0.390625 0.828125 0.390625 1.921875l0 6.09375l-1.59375 0l0 -5.859375q0 -0.90625 -0.328125 -1.453125q-0.3125 -0.5625 -0.84375 -0.8125q-0.53125 -0.25 -1.171875 -0.25q-0.78125 0 -1.40625 0.453125q-0.609375 0.4375 -0.96875 1.15625q-0.34375 0.703125 -0.34375 1.46875l0 5.296875l-1.59375 0zm14.7565 0.296875q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm15.425446 0q-0.75 0 -1.390625 -0.25q-0.640625 -0.234375 -1.125 -0.625q-0.46875 -0.390625 -0.71875 -0.828125l-0.078125 0l0 1.40625l-1.515625 0l0 -13.359375l1.59375 0l0 3.9375l-0.078125 1.328125l0.078125 0q0.25 -0.453125 0.71875 -0.84375q0.484375 -0.390625 1.125 -0.625q0.640625 -0.25 1.390625 -0.25q1.3125 0 2.328125 0.65625q1.03125 0.65625 1.625 1.796875q0.609375 1.125 0.609375 2.609375q0 1.46875 -0.609375 2.609375q-0.59375 1.140625 -1.625 1.796875q-1.015625 0.640625 -2.328125 0.640625zm-0.15625 -1.4375q0.8125 0 1.53125 -0.4375q0.71875 -0.4375 1.15625 -1.25q0.4375 -0.828125 0.4375 -1.921875q0 -1.125 -0.4375 -1.9375q-0.4375 -0.8125 -1.15625 -1.25q-0.71875 -0.4375 -1.53125 -0.4375q-0.828125 0 -1.546875 0.4375q-0.71875 0.4375 -1.171875 1.25q-0.4375 0.8125 -0.4375 1.9375q0 1.109375 0.4375 1.921875q0.453125 0.8125 1.171875 1.25q0.71875 0.4375 1.546875 0.4375zm9.606049 1.4375q-1.0625 0 -1.8749695 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.4530945 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0624695 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5624695 -0.40625 1.2812195 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.6405945 0.5625 -0.6405945 1.375q0 0.78125 0.5937195 1.28125q0.609375 0.484375 1.515625 0.484375zm9.953735 1.390625q-1.0625 0 -1.875 -0.34375q-0.8125 -0.34375 -1.34375 -0.921875q-0.53125 -0.59375 -0.796875 -1.28125l1.421875 -0.640625q0.375 0.859375 1.078125 1.328125q0.703125 0.46875 1.625 0.46875q0.875 0 1.453125 -0.359375q0.59375 -0.359375 0.59375 -1.046875q0 -0.421875 -0.25 -0.703125q-0.234375 -0.296875 -0.703125 -0.5q-0.453125 -0.21875 -1.125 -0.375l-1.15625 -0.3125q-0.671875 -0.1875 -1.28125 -0.515625q-0.59375 -0.34375 -0.953125 -0.875q-0.359375 -0.53125 -0.359375 -1.296875q0 -0.84375 0.5 -1.453125q0.5 -0.625 1.3125 -0.953125q0.828125 -0.328125 1.765625 -0.328125q0.8125 0 1.53125 0.234375q0.71875 0.234375 1.265625 0.6875q0.546875 0.453125 0.8125 1.125l-1.375 0.640625q-0.359375 -0.703125 -0.953125 -0.984375q-0.59375 -0.28125 -1.328125 -0.28125q-0.78125 0 -1.359375 0.34375q-0.578125 0.34375 -0.578125 0.9375q0 0.59375 0.46875 0.90625q0.484375 0.296875 1.171875 0.46875l1.390625 0.359375q1.390625 0.359375 2.09375 1.0625q0.71875 0.6875 0.71875 1.703125q0 0.890625 -0.515625 1.546875q-0.5 0.65625 -1.359375 1.015625q-0.84375 0.34375 -1.890625 0.34375zm9.823975 0q-1.375 0 -2.453125 -0.640625q-1.0625 -0.65625 -1.671875 -1.796875q-0.609375 -1.140625 -0.609375 -2.59375q0 -1.359375 0.5625 -2.515625q0.578125 -1.15625 1.609375 -1.859375q1.03125 -0.703125 2.4375 -0.703125q1.421875 0 2.4375 0.625q1.015625 0.625 1.5625 1.734375q0.546875 1.09375 0.546875 2.515625q0 0.125 -0.015625 0.265625q0 0.125 -0.015625 0.21875l-8.1875 0l0 -1.3125l6.546875 0q-0.015625 -0.390625 -0.1875 -0.84375q-0.15625 -0.46875 -0.5 -0.859375q-0.34375 -0.40625 -0.875 -0.65625q-0.53125 -0.25 -1.3125 -0.25q-0.9375 0 -1.625 0.484375q-0.671875 0.46875 -1.046875 1.296875q-0.359375 0.8125 -0.359375 1.859375q0 1.203125 0.46875 2.015625q0.46875 0.796875 1.203125 1.1875q0.75 0.390625 1.546875 0.390625q1.046875 0 1.71875 -0.484375q0.6875 -0.5 1.09375 -1.234375l1.34375 0.65625q-0.5625 1.078125 -1.609375 1.796875q-1.03125 0.703125 -2.609375 0.703125zm13.644226 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm6.718994 1.09375l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm3.9364624 0l0 -13.359375l1.59375 0l0 13.359375l-1.59375 0zm8.233704 0.296875q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm11.041016 1.4375q-1.40625 0 -2.5 -0.65625q-1.078125 -0.671875 -1.703125 -1.8125q-0.609375 -1.15625 -0.609375 -2.578125q0 -1.46875 0.609375 -2.59375q0.625 -1.140625 1.703125 -1.796875q1.09375 -0.671875 2.5 -0.671875q1.609375 0 2.640625 0.734375q1.03125 0.734375 1.46875 1.890625l-1.4375 0.609375q-0.359375 -0.890625 -1.0625 -1.34375q-0.6875 -0.453125 -1.6875 -0.453125q-0.828125 0 -1.546875 0.453125q-0.71875 0.4375 -1.171875 1.25q-0.453125 0.8125 -0.453125 1.921875q0 1.078125 0.453125 1.90625q0.453125 0.8125 1.171875 1.265625q0.71875 0.4375 1.546875 0.4375q1.015625 0 1.734375 -0.46875q0.734375 -0.46875 1.09375 -1.3125l1.40625 0.59375q-0.46875 1.09375 -1.515625 1.859375q-1.03125 0.765625 -2.640625 0.765625zm8.827332 0q-1.0625 0 -1.875 -0.40625q-0.796875 -0.40625 -1.25 -1.125q-0.453125 -0.71875 -0.453125 -1.640625q0 -1.046875 0.53125 -1.765625q0.546875 -0.71875 1.453125 -1.078125q0.921875 -0.359375 2.015625 -0.359375q0.640625 0 1.171875 0.109375q0.546875 0.09375 0.9375 0.234375q0.40625 0.140625 0.625 0.265625l0 -0.578125q0 -1.078125 -0.765625 -1.703125q-0.765625 -0.640625 -1.875 -0.640625q-0.78125 0 -1.46875 0.34375q-0.671875 0.34375 -1.0625 0.953125l-1.203125 -0.890625q0.375 -0.5625 0.9375 -0.96875q0.5625 -0.40625 1.28125 -0.625q0.71875 -0.234375 1.515625 -0.234375q1.9375 0 3.03125 1.03125q1.109375 1.015625 1.109375 2.75l0 6.03125l-1.5 0l0 -1.359375l-0.078125 0q-0.25 0.40625 -0.703125 0.796875q-0.4375 0.375 -1.046875 0.609375q-0.609375 0.25 -1.328125 0.25zm0.140625 -1.390625q0.828125 0 1.5 -0.40625q0.6875 -0.421875 1.09375 -1.109375q0.421875 -0.6875 0.421875 -1.515625q-0.4375 -0.296875 -1.078125 -0.484375q-0.640625 -0.1875 -1.40625 -0.1875q-1.359375 0 -2.0 0.5625q-0.640625 0.5625 -0.640625 1.375q0 0.78125 0.59375 1.28125q0.609375 0.484375 1.515625 0.484375zm5.8760376 -8.421875l5.578125 0l0 1.4375l-5.578125 0l0 -1.4375zm1.671875 7.015625l0 -9.703125l1.578125 0l0 9.3125q0 0.75 0.3125 1.15625q0.3125 0.40625 1.015625 0.40625q0.3125 0 0.578125 -0.09375q0.265625 -0.09375 0.46875 -0.21875l0 1.546875q-0.25 0.109375 -0.546875 0.171875q-0.28125 0.078125 -0.765625 0.078125q-1.1875 0 -1.921875 -0.703125q-0.71875 -0.703125 -0.71875 -1.953125zm6.043518 2.5l0 -9.515625l1.578125 0l0 9.515625l-1.578125 0zm0.78125 -11.265625q-0.46875 0 -0.8125 -0.328125q-0.328125 -0.34375 -0.328125 -0.8125q0 -0.484375 0.328125 -0.8125q0.34375 -0.328125 0.8125 -0.328125q0.484375 0 0.8125 0.328125q0.328125 0.328125 0.328125 0.8125q0 0.46875 -0.328125 0.8125q-0.328125 0.328125 -0.8125 0.328125zm7.576111 11.5625q-1.4375 0 -2.546875 -0.671875q-1.09375 -0.671875 -1.71875 -1.8125q-0.625 -1.15625 -0.625 -2.5625q0 -1.421875 0.625 -2.5625q0.625 -1.15625 1.71875 -1.828125q1.109375 -0.671875 2.546875 -0.671875q1.4375 0 2.53125 0.6875q1.109375 0.671875 1.734375 1.828125q0.625 1.140625 0.625 2.546875q0 1.40625 -0.625 2.5625q-0.625 1.140625 -1.734375 1.8125q-1.09375 0.671875 -2.53125 0.671875zm0 -1.4375q0.859375 0 1.609375 -0.421875q0.75 -0.4375 1.21875 -1.25q0.46875 -0.8125 0.46875 -1.9375q0 -1.140625 -0.46875 -1.953125q-0.46875 -0.8125 -1.21875 -1.234375q-0.75 -0.4375 -1.609375 -0.4375q-0.859375 0 -1.625 0.4375q-0.765625 0.421875 -1.234375 1.234375q-0.46875 0.8125 -0.46875 1.953125q0 1.125 0.46875 1.9375q0.46875 0.8125 1.234375 1.25q0.765625 0.421875 1.625 0.421875zm6.54187 1.140625l0 -9.515625l1.515625 0l0 1.40625l0.078125 0q0.375 -0.703125 1.21875 -1.203125q0.84375 -0.5 1.859375 -0.5q1.75 0 2.625 1.015625q0.890625 1.015625 0.890625 2.703125l0 6.09375l-1.578125 0l0 -5.859375q0 -1.375 -0.671875 -1.9375q-0.65625 -0.578125 -1.703125 -0.578125q-0.78125 0 -1.375 0.4375q-0.59375 0.4375 -0.9375 1.125q-0.328125 0.6875 -0.328125 1.453125l0 5.359375l-1.59375 0zm10.825562 2.53125l-1.15625 -0.8125q0.78125 -0.9375 1.4375 -2.15625q0.671875 -1.234375 1.0625 -2.65625q0.40625 -1.421875 0.40625 -2.90625q0 -1.5 -0.40625 -2.90625q-0.390625 -1.421875 -1.0625 -2.65625q-0.65625 -1.234375 -1.4375 -2.171875l1.15625 -0.8125q0.75 0.859375 1.515625 2.15625q0.765625 1.296875 1.28125 2.90625q0.515625 1.609375 0.515625 3.484375q0 1.390625 -0.3125 2.65625q-0.296875 1.25 -0.78125 2.34375q-0.484375 1.09375 -1.078125 1.984375q-0.578125 0.90625 -1.140625 1.546875zm6.713501 -2.421875q-0.5 0 -0.84375 -0.34375q-0.34375 -0.34375 -0.34375 -0.84375q0 -0.484375 0.34375 -0.828125q0.34375 -0.34375 0.84375 -0.34375q0.5 0 0.84375 0.34375q0.359375 0.34375 0.359375 0.828125q0 0.5 -0.359375 0.84375q-0.34375 0.34375 -0.84375 0.34375z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m37.16798 128.8189l426.96063 0l0 51.905518l-426.96063 0z" fill-rule="evenodd"/><path fill="#595959" d="m46.60548 147.16577q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm7.9281006 3.453125l-2.125 -6.90625l1.21875 0l1.09375 3.984375l0.421875 1.484375q0.015625 -0.109375 0.359375 -1.421875l1.09375 -4.046875l1.203125 0l1.03125 4.0l0.34375 1.328125l0.40625 -1.34375l1.171875 -3.984375l1.140625 0l-2.15625 6.90625l-1.21875 0l-1.09375 -4.140625l-0.265625 -1.171875l-1.40625 5.3125l-1.21875 0zm8.343899 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5062256 4.125l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm3.9852448 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625zm7.1328125 2.0625l0 -9.546875l1.171875 0l0 3.421875q0.828125 -0.9375 2.078125 -0.9375q0.765625 0 1.328125 0.296875q0.5625 0.296875 0.8125 0.84375q0.25 0.53125 0.25 1.546875l0 4.375l-1.171875 0l0 -4.375q0 -0.890625 -0.390625 -1.28125q-0.375 -0.40625 -1.078125 -0.40625q-0.515625 0 -0.984375 0.28125q-0.453125 0.265625 -0.65625 0.734375q-0.1875 0.453125 -0.1875 1.265625l0 3.78125l-1.171875 0zm7.4281006 -8.1875l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.945465 2.65625l0 -9.5625l1.078125 0l0 0.890625q0.375 -0.53125 0.84375 -0.78125q0.484375 -0.265625 1.15625 -0.265625q0.875 0 1.546875 0.453125q0.6875 0.453125 1.03125 1.28125q0.34375 0.828125 0.34375 1.828125q0 1.046875 -0.375 1.90625q-0.375 0.84375 -1.109375 1.296875q-0.71875 0.453125 -1.53125 0.453125q-0.578125 0 -1.046875 -0.25q-0.46875 -0.25 -0.765625 -0.625l0 3.375l-1.171875 0zm1.0625 -6.078125q0 1.34375 0.53125 1.984375q0.546875 0.625 1.3125 0.625q0.78125 0 1.34375 -0.65625q0.5625 -0.65625 0.5625 -2.046875q0 -1.3125 -0.546875 -1.96875q-0.546875 -0.671875 -1.296875 -0.671875q-0.75 0 -1.328125 0.703125q-0.578125 0.703125 -0.578125 2.03125zm9.615463 -0.03125q0 -1.921875 1.078125 -2.84375q0.890625 -0.765625 2.171875 -0.765625q1.421875 0 2.328125 0.9375q0.90625 0.921875 0.90625 2.578125q0 1.328125 -0.40625 2.09375q-0.390625 0.765625 -1.15625 1.1875q-0.765625 0.421875 -1.671875 0.421875q-1.453125 0 -2.359375 -0.921875q-0.890625 -0.9375 -0.890625 -2.6875zm1.203125 0q0 1.328125 0.578125 1.984375q0.59375 0.65625 1.46875 0.65625q0.875 0 1.453125 -0.65625q0.578125 -0.671875 0.578125 -2.03125q0 -1.28125 -0.59375 -1.9375q-0.578125 -0.65625 -1.4375 -0.65625q-0.875 0 -1.46875 0.65625q-0.578125 0.65625 -0.578125 1.984375zm6.9281006 3.453125l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm7.1091003 0l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.7343826 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.29688263 0.234375 -0.42188263 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm9.188377 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm9.084351 3.078125l0.171875 1.03125q-0.5 0.109375 -0.890625 0.109375q-0.640625 0 -1.0 -0.203125q-0.34375 -0.203125 -0.484375 -0.53125q-0.140625 -0.328125 -0.140625 -1.390625l0 -3.96875l-0.859375 0l0 -0.90625l0.859375 0l0 -1.71875l1.171875 -0.703125l0 2.421875l1.171875 0l0 0.90625l-1.171875 0l0 4.046875q0 0.5 0.046875 0.640625q0.0625 0.140625 0.203125 0.234375q0.140625 0.078125 0.40625 0.078125q0.203125 0 0.515625 -0.046875zm5.6561127 0.1875q-0.65625 0.5625 -1.265625 0.796875q-0.59375 0.21875 -1.28125 0.21875q-1.140625 0 -1.75 -0.546875q-0.609375 -0.5625 -0.609375 -1.4375q0 -0.5 0.21875 -0.921875q0.234375 -0.421875 0.609375 -0.671875q0.375 -0.25 0.84375 -0.390625q0.34375 -0.078125 1.046875 -0.171875q1.421875 -0.171875 2.09375 -0.40625q0 -0.234375 0 -0.296875q0 -0.71875 -0.328125 -1.015625q-0.453125 -0.390625 -1.34375 -0.390625q-0.8125 0 -1.21875 0.296875q-0.390625 0.28125 -0.578125 1.015625l-1.140625 -0.15625q0.15625 -0.734375 0.515625 -1.1875q0.359375 -0.453125 1.03125 -0.6875q0.671875 -0.25 1.5625 -0.25q0.890625 0 1.4375 0.203125q0.5625 0.203125 0.8125 0.53125q0.265625 0.3125 0.375 0.796875q0.046875 0.296875 0.046875 1.078125l0 1.5625q0 1.625 0.078125 2.0625q0.078125 0.4375 0.296875 0.828125l-1.21875 0q-0.1875 -0.359375 -0.234375 -0.859375zm-0.09375 -2.609375q-0.640625 0.265625 -1.921875 0.4375q-0.71875 0.109375 -1.015625 0.25q-0.296875 0.125 -0.46875 0.375q-0.15625 0.25 -0.15625 0.546875q0 0.46875 0.34375 0.78125q0.359375 0.3125 1.046875 0.3125q0.671875 0 1.203125 -0.296875q0.53125 -0.296875 0.78125 -0.8125q0.1875 -0.390625 0.1875 -1.171875l0 -0.421875zm3.0062256 -4.71875l0 -1.359375l1.171875 0l0 1.359375l-1.171875 0zm0 8.1875l0 -6.90625l1.171875 0l0 6.90625l-1.171875 0zm2.945465 0l0 -6.90625l1.0625 0l0 0.984375q0.75 -1.140625 2.1875 -1.140625q0.625 0 1.15625 0.21875q0.53125 0.21875 0.78125 0.59375q0.265625 0.359375 0.375 0.859375q0.0625 0.328125 0.0625 1.140625l0 4.25l-1.171875 0l0 -4.203125q0 -0.71875 -0.140625 -1.0625q-0.140625 -0.359375 -0.484375 -0.5625q-0.34375 -0.21875 -0.8125 -0.21875q-0.75 0 -1.296875 0.46875q-0.546875 0.46875 -0.546875 1.796875l0 3.78125l-1.171875 0zm12.146851 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm11.006226 4.125l0 -0.875q-0.65625 1.03125 -1.9375 1.03125q-0.8125 0 -1.515625 -0.453125q-0.6875 -0.453125 -1.078125 -1.265625q-0.375 -0.828125 -0.375 -1.890625q0 -1.03125 0.34375 -1.875q0.34375 -0.84375 1.03125 -1.28125q0.703125 -0.453125 1.546875 -0.453125q0.625 0 1.109375 0.265625q0.5 0.25 0.796875 0.671875l0 -3.421875l1.171875 0l0 9.546875l-1.09375 0zm-3.703125 -3.453125q0 1.328125 0.5625 1.984375q0.5625 0.65625 1.328125 0.65625q0.765625 0 1.296875 -0.625q0.53125 -0.625 0.53125 -1.90625q0 -1.421875 -0.546875 -2.078125q-0.546875 -0.671875 -1.34375 -0.671875q-0.78125 0 -1.3125 0.640625q-0.515625 0.625 -0.515625 2.0zm11.412338 3.453125l-1.078125 0l0 -9.546875l1.171875 0l0 3.40625q0.734375 -0.921875 1.890625 -0.921875q0.640625 0 1.203125 0.265625q0.578125 0.25 0.9375 0.71875q0.375 0.453125 0.578125 1.109375q0.203125 0.65625 0.203125 1.40625q0 1.78125 -0.875 2.75q-0.875 0.96875 -2.109375 0.96875q-1.21875 0 -1.921875 -1.015625l0 0.859375zm0 -3.5q0 1.234375 0.328125 1.78125q0.5625 0.90625 1.5 0.90625q0.765625 0 1.328125 -0.65625q0.5625 -0.671875 0.5625 -2.0q0 -1.34375 -0.546875 -1.984375q-0.53125 -0.65625 -1.296875 -0.65625q-0.765625 0 -1.328125 0.671875q-0.546875 0.671875 -0.546875 1.9375zm10.865601 3.5l0 -1.015625q-0.8125 1.171875 -2.1875 1.171875q-0.609375 0 -1.140625 -0.234375q-0.53125 -0.234375 -0.796875 -0.578125q-0.25 -0.359375 -0.359375 -0.875q-0.0625 -0.34375 -0.0625 -1.09375l0 -4.28125l1.171875 0l0 3.828125q0 0.921875 0.0625 1.234375q0.109375 0.46875 0.46875 0.734375q0.359375 0.25 0.890625 0.25q0.515625 0 0.984375 -0.265625q0.46875 -0.265625 0.65625 -0.734375q0.1875 -0.46875 0.1875 -1.34375l0 -3.703125l1.171875 0l0 6.90625l-1.046875 0zm3.1624756 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm3.4620972 0l0 -6.0l-1.03125 0l0 -0.90625l1.03125 0l0 -0.734375q0 -0.703125 0.125 -1.046875q0.171875 -0.453125 0.59375 -0.734375q0.421875 -0.28125 1.203125 -0.28125q0.484375 0 1.09375 0.109375l-0.1875 1.03125q-0.359375 -0.0625 -0.6875 -0.0625q-0.53125 0 -0.75 0.234375q-0.21875 0.21875 -0.21875 0.84375l0 0.640625l1.34375 0l0 0.90625l-1.34375 0l0 6.0l-1.171875 0zm8.156113 -2.21875l1.203125 0.140625q-0.28125 1.0625 -1.0625 1.65625q-0.765625 0.578125 -1.96875 0.578125q-1.515625 0 -2.40625 -0.9375q-0.890625 -0.9375 -0.890625 -2.609375q0 -1.75 0.890625 -2.703125q0.90625 -0.96875 2.34375 -0.96875q1.390625 0 2.265625 0.9375q0.875 0.9375 0.875 2.65625q0 0.109375 0 0.3125l-5.15625 0q0.0625 1.140625 0.640625 1.75q0.578125 0.59375 1.4375 0.59375q0.65625 0 1.109375 -0.328125q0.453125 -0.34375 0.71875 -1.078125zm-3.84375 -1.90625l3.859375 0q-0.078125 -0.859375 -0.4375 -1.296875q-0.5625 -0.6875 -1.453125 -0.6875q-0.8125 0 -1.359375 0.546875q-0.546875 0.53125 -0.609375 1.4375zm6.5062256 4.125l0 -6.90625l1.0625 0l0 1.046875q0.40625 -0.734375 0.734375 -0.96875q0.34375 -0.234375 0.765625 -0.234375q0.59375 0 1.203125 0.375l-0.40625 1.078125q-0.4375 -0.25 -0.859375 -0.25q-0.390625 0 -0.703125 0.234375q-0.296875 0.234375 -0.421875 0.640625q-0.203125 0.625 -0.203125 1.359375l0 3.625l-1.171875 0zm3.9852448 -2.0625l1.15625 -0.1875q0.109375 0.703125 0.546875 1.078125q0.453125 0.359375 1.25 0.359375q0.8125 0 1.203125 -0.328125q0.390625 -0.328125 0.390625 -0.765625q0 -0.390625 -0.359375 -0.625q-0.234375 -0.15625 -1.1875 -0.390625q-1.296875 -0.328125 -1.796875 -0.5625q-0.484375 -0.25 -0.75 -0.65625q-0.25 -0.421875 -0.25 -0.9375q0 -0.453125 0.203125 -0.84375q0.21875 -0.40625 0.578125 -0.671875q0.28125 -0.1875 0.75 -0.328125q0.46875 -0.140625 1.015625 -0.140625q0.8125 0 1.421875 0.234375q0.609375 0.234375 0.90625 0.640625q0.296875 0.390625 0.40625 1.0625l-1.140625 0.15625q-0.078125 -0.53125 -0.453125 -0.828125q-0.375 -0.3125 -1.0625 -0.3125q-0.8125 0 -1.15625 0.265625q-0.34375 0.265625 -0.34375 0.625q0 0.234375 0.140625 0.421875q0.15625 0.1875 0.453125 0.3125q0.171875 0.0625 1.03125 0.296875q1.25 0.328125 1.734375 0.546875q0.5 0.203125 0.78125 0.609375q0.28125 0.40625 0.28125 1.0q0 0.59375 -0.34375 1.109375q-0.34375 0.515625 -1.0 0.796875q-0.640625 0.28125 -1.453125 0.28125q-1.34375 0 -2.046875 -0.5625q-0.703125 -0.5625 -0.90625 -1.65625z" fill-rule="nonzero"/><path fill="#000000" fill-opacity="0.0" d="m101.76378 136.9895l-18.204727 -31.055122" fill-rule="evenodd"/><path stroke="#fbbc05" stroke-width="2.0" stroke-linejoin="round" stroke-linecap="butt" d="m101.76378 136.9895l-12.136093 -20.702744" fill-rule="evenodd"/><path fill="#fbbc05" stroke="#fbbc05" stroke-width="2.0" stroke-linecap="butt" d="m92.477585 114.616135l-7.4399033 -6.1593933l1.7401123 9.500641z" fill-rule="evenodd"/></g></svg>
\ No newline at end of file
diff --git a/mlir/include/mlir/Dialect/Bufferization/IR/BufferizationBase.td b/mlir/include/mlir/Dialect/Bufferization/IR/BufferizationBase.td
index 818e931aeb505..ac19f878656b5 100644
--- a/mlir/include/mlir/Dialect/Bufferization/IR/BufferizationBase.td
+++ b/mlir/include/mlir/Dialect/Bufferization/IR/BufferizationBase.td
@@ -24,8 +24,8 @@ def Bufferization_Dialect : Dialect {
 
     An overview of the bufferization infrastructure and important conceptual
     details related to using the MLIR dialect conversion infrastructure can be
-    found in [bufferization](/docs/Bufferization/) and [buffer
-    deallocation](/docs/BufferDeallocationInternals/).
+    found in [bufferization](/docs/Bufferization/) and [ownership-based buffer
+    deallocation](/docs/OwnershipBasedBufferDeallocation/).
   }];
   let dependentDialects = [
     "affine::AffineDialect", "memref::MemRefDialect", "tensor::TensorDialect",

From 652ae4ecadaa972d70ef5cea3b68f2ff103c4af2 Mon Sep 17 00:00:00 2001
From: Krystian Stasiowski <sdkrystian@gmail.com>
Date: Thu, 18 Apr 2024 07:48:10 -0400
Subject: [PATCH 865/886] [Clang][Sema] Warn when
 'exclude_from_explicit_instantiation' attribute is used on local classes and
 members thereof (#88777)

A local class and its members declared in a function template are
instantiated alongside the definition of that template. It therefore
does not make sense to apply the `exclude_from_explicit_instantiation`
attribute to such declarations, and this patch adds a warning to
diagnose these cases (in addition to ignoring the attribute).

(The motivation for this patch is to fix a failing test in libc++ for
#84050. In particular, line 199 in `include/__memory/uses_allocator_construction.h`
in libc++ contains the expression `this->__value_` (reduced to
https://godbolt.org/z/KqEerKWPd) which will be looked up prior
to instantiation once #84050 lands (the lookup context is the
current instantiation). `_LIBCPP_HIDE_FROM_ABI` includes
`__attribute__((exclude_from_explicit_instantiation))`, which in the
reduced example results in `Local` being instantiated with
`Local::operator A` as its `DeclContext`)
---
 clang/docs/ReleaseNotes.rst                   |  3 +
 .../clang/Basic/DiagnosticSemaKinds.td        |  3 +
 clang/lib/Sema/SemaDeclAttr.cpp               | 18 ++++++
 ...rom_explicit_instantiation.local-class.cpp | 64 +++++++++++++++++++
 4 files changed, 88 insertions(+)
 create mode 100644 clang/test/SemaCXX/attr-exclude_from_explicit_instantiation.local-class.cpp

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 6c51c2d1f483c..3fe15934323c5 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -287,6 +287,9 @@ Attribute Changes in Clang
   This allows the ``_Nullable`` and ``_Nonnull`` family of type attributes to
   apply to this class.
 
+- Clang now warns that the ``exclude_from_explicit_instantiation`` attribute
+  is ignored when applied to a local class or a member thereof.
+
 Improvements to Clang's diagnostics
 -----------------------------------
 - Clang now applies syntax highlighting to the code snippets it
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 30a8543489f48..a9f4143c6b375 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -3669,6 +3669,9 @@ def warn_attribute_dllexport_explicit_instantiation_decl : Warning<
 def warn_attribute_dllexport_explicit_instantiation_def : Warning<
   "'dllexport' attribute ignored on explicit instantiation definition">,
   InGroup<IgnoredAttributes>;
+def warn_attribute_exclude_from_explicit_instantiation_local_class : Warning<
+  "%0 attribute ignored on local class%select{| member}1">,
+  InGroup<IgnoredAttributes>;
 def warn_invalid_initializer_from_system_header : Warning<
   "invalid constructor from class in system header, should not be explicit">,
   InGroup<DiagGroup<"invalid-initializer-from-system-header">>;
diff --git a/clang/lib/Sema/SemaDeclAttr.cpp b/clang/lib/Sema/SemaDeclAttr.cpp
index c3bf18a3f79e2..363ae93cb62df 100644
--- a/clang/lib/Sema/SemaDeclAttr.cpp
+++ b/clang/lib/Sema/SemaDeclAttr.cpp
@@ -984,6 +984,21 @@ static void handleErrorAttr(Sema &S, Decl *D, const ParsedAttr &AL) {
     D->addAttr(EA);
 }
 
+static void handleExcludeFromExplicitInstantiationAttr(Sema &S, Decl *D,
+                                                       const ParsedAttr &AL) {
+  const auto *PD = isa<CXXRecordDecl>(D)
+                       ? cast<DeclContext>(D)
+                       : D->getDeclContext()->getRedeclContext();
+  if (const auto *RD = dyn_cast<CXXRecordDecl>(PD); RD && RD->isLocalClass()) {
+    S.Diag(AL.getLoc(),
+           diag::warn_attribute_exclude_from_explicit_instantiation_local_class)
+        << AL << /*IsMember=*/!isa<CXXRecordDecl>(D);
+    return;
+  }
+  D->addAttr(::new (S.Context)
+                 ExcludeFromExplicitInstantiationAttr(S.Context, AL));
+}
+
 namespace {
 /// Determines if a given Expr references any of the given function's
 /// ParmVarDecls, or the function's implicit `this` parameter (if applicable).
@@ -9339,6 +9354,9 @@ ProcessDeclAttribute(Sema &S, Scope *scope, Decl *D, const ParsedAttr &AL,
   case ParsedAttr::AT_Error:
     handleErrorAttr(S, D, AL);
     break;
+  case ParsedAttr::AT_ExcludeFromExplicitInstantiation:
+    handleExcludeFromExplicitInstantiationAttr(S, D, AL);
+    break;
   case ParsedAttr::AT_DiagnoseIf:
     handleDiagnoseIfAttr(S, D, AL);
     break;
diff --git a/clang/test/SemaCXX/attr-exclude_from_explicit_instantiation.local-class.cpp b/clang/test/SemaCXX/attr-exclude_from_explicit_instantiation.local-class.cpp
new file mode 100644
index 0000000000000..f0b2cec095c97
--- /dev/null
+++ b/clang/test/SemaCXX/attr-exclude_from_explicit_instantiation.local-class.cpp
@@ -0,0 +1,64 @@
+// RUN: %clang_cc1 -std=c++20 -fsyntax-only -verify %s
+
+// Test that the exclude_from_explicit_instantiation attribute is ignored
+// for local classes and members thereof.
+
+#define EXCLUDE_FROM_EXPLICIT_INSTANTIATION __attribute__((exclude_from_explicit_instantiation)) // expected-note 0+{{expanded from macro}}
+
+namespace N0 {
+
+  template<typename T>
+  void f() {
+    struct EXCLUDE_FROM_EXPLICIT_INSTANTIATION A { // expected-warning {{attribute ignored on local class}}
+                                                   // expected-note@-1 2{{in instantiation of}}
+      EXCLUDE_FROM_EXPLICIT_INSTANTIATION void g(T t) { // expected-warning {{attribute ignored on local class member}}
+        *t; // expected-error {{indirection requires pointer operand ('int' invalid)}}
+      }
+
+      struct EXCLUDE_FROM_EXPLICIT_INSTANTIATION B { // expected-warning {{attribute ignored on local class}}
+        void h(T t) {
+          *t; // expected-error {{indirection requires pointer operand ('int' invalid)}}
+        }
+      };
+    };
+  }
+
+  template void f<int>(); // expected-note 2{{in instantiation of}}
+
+}
+
+// This is a reduced example from libc++ which required that 'value'
+// be prefixed with 'this->' because the definition of 'Local::operator A'
+// was not instantiated when the definition of 'g' was.
+namespace N1 {
+
+  struct A { };
+
+  struct B {
+    operator A() {
+      return A();
+    }
+  };
+
+  template<typename T>
+  auto f(T t) {
+    return A(t);
+  }
+
+  template<typename T>
+  auto g(T t) {
+    struct Local {
+      T value;
+
+      EXCLUDE_FROM_EXPLICIT_INSTANTIATION // expected-warning {{attribute ignored on local class member}}
+      operator A() {
+        return A(value);
+      }
+    };
+
+    return f(Local(t));
+  }
+
+  auto x = g(B());
+
+}

From 3f0d52b2dbdb096212e8ffa6ad86aacd0d256957 Mon Sep 17 00:00:00 2001
From: Benjamin Kramer <benny.kra@googlemail.com>
Date: Thu, 18 Apr 2024 13:36:04 +0200
Subject: [PATCH 866/886] [mlir][llvm] Clean out f8 types from compatible types
 list

These were added in faf697e49bad099b36770c7738f36917e3d90d2d so things
can flow through non-opaque LLVM ptrs. Those ptrs are gone so there is
no reason for this to be around anymore. LLVM doesn't support f8 types,
they get converted to i8 when lowering to LLVM dialect.

Removing the f8 types makes LLVM::isCompatibleType and
LLVM::isCompatibleFloatingPointType consistent again.
---
 mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp b/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
index 630187f220a4b..ad1dc4a36b82b 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
@@ -769,8 +769,6 @@ bool mlir::LLVM::isCompatibleOuterType(Type type) {
       Float64Type,
       Float80Type,
       Float128Type,
-      Float8E4M3FNType,
-      Float8E5M2Type,
       LLVMArrayType,
       LLVMFunctionType,
       LLVMLabelType,
@@ -841,8 +839,6 @@ static bool isCompatibleImpl(Type type, DenseSet<Type> &compatibleTypes) {
             Float64Type,
             Float80Type,
             Float128Type,
-            Float8E4M3FNType,
-            Float8E5M2Type,
             LLVMLabelType,
             LLVMMetadataType,
             LLVMPPCFP128Type,

From 0e08bce142c90fcdc36365433e02d75acadbaa3c Mon Sep 17 00:00:00 2001
From: Louis Dionne <ldionne.2@gmail.com>
Date: Thu, 18 Apr 2024 07:50:57 -0400
Subject: [PATCH 867/886] [libc++][pstl] Move the CPU algorithm implementations
 to __pstl (#89109)

This colocates the CPU algorithms closer to the rest of the PSTL
implementation details.
---
 libcxx/include/CMakeLists.txt                 | 16 +++++------
 libcxx/include/__pstl/backends/libdispatch.h  | 16 +++++------
 libcxx/include/__pstl/backends/serial.h       | 16 +++++------
 libcxx/include/__pstl/backends/std_thread.h   | 16 +++++------
 .../cpu_algos}/any_of.h                       |  6 ++--
 .../cpu_backends => __pstl/cpu_algos}/fill.h  |  6 ++--
 .../cpu_algos}/find_if.h                      |  6 ++--
 .../cpu_algos}/for_each.h                     |  6 ++--
 .../cpu_backends => __pstl/cpu_algos}/merge.h |  6 ++--
 .../cpu_algos}/stable_sort.h                  |  6 ++--
 .../cpu_algos}/transform.h                    |  6 ++--
 .../cpu_algos}/transform_reduce.h             |  6 ++--
 libcxx/include/libcxx.imp                     |  8 ------
 libcxx/include/module.modulemap               | 28 +++++++++----------
 14 files changed, 70 insertions(+), 78 deletions(-)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/any_of.h (95%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/fill.h (92%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/find_if.h (96%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/for_each.h (92%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/merge.h (93%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/stable_sort.h (88%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/transform.h (96%)
 rename libcxx/include/{__algorithm/pstl_backends/cpu_backends => __pstl/cpu_algos}/transform_reduce.h (97%)

diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index ee4979bfc6f89..dc364cc8d0a22 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -73,14 +73,6 @@ set(files
   __algorithm/pop_heap.h
   __algorithm/prev_permutation.h
   __algorithm/pstl_any_all_none_of.h
-  __algorithm/pstl_backends/cpu_backends/any_of.h
-  __algorithm/pstl_backends/cpu_backends/fill.h
-  __algorithm/pstl_backends/cpu_backends/find_if.h
-  __algorithm/pstl_backends/cpu_backends/for_each.h
-  __algorithm/pstl_backends/cpu_backends/merge.h
-  __algorithm/pstl_backends/cpu_backends/stable_sort.h
-  __algorithm/pstl_backends/cpu_backends/transform.h
-  __algorithm/pstl_backends/cpu_backends/transform_reduce.h
   __algorithm/pstl_copy.h
   __algorithm/pstl_count.h
   __algorithm/pstl_equal.h
@@ -593,7 +585,15 @@ set(files
   __pstl/backends/std_thread.h
   __pstl/configuration.h
   __pstl/configuration_fwd.h
+  __pstl/cpu_algos/any_of.h
   __pstl/cpu_algos/cpu_traits.h
+  __pstl/cpu_algos/fill.h
+  __pstl/cpu_algos/find_if.h
+  __pstl/cpu_algos/for_each.h
+  __pstl/cpu_algos/merge.h
+  __pstl/cpu_algos/stable_sort.h
+  __pstl/cpu_algos/transform.h
+  __pstl/cpu_algos/transform_reduce.h
   __random/bernoulli_distribution.h
   __random/binomial_distribution.h
   __random/cauchy_distribution.h
diff --git a/libcxx/include/__pstl/backends/libdispatch.h b/libcxx/include/__pstl/backends/libdispatch.h
index 977b06b9a489c..af1da80dc133e 100644
--- a/libcxx/include/__pstl/backends/libdispatch.h
+++ b/libcxx/include/__pstl/backends/libdispatch.h
@@ -349,13 +349,13 @@ _LIBCPP_END_NAMESPACE_STD
 _LIBCPP_POP_MACROS
 
 // Implement PSTL algorithms based on the __cpu_traits specialized above
-#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
-#include <__algorithm/pstl_backends/cpu_backends/fill.h>
-#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
-#include <__algorithm/pstl_backends/cpu_backends/merge.h>
-#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+#include <__pstl/cpu_algos/any_of.h>
+#include <__pstl/cpu_algos/fill.h>
+#include <__pstl/cpu_algos/find_if.h>
+#include <__pstl/cpu_algos/for_each.h>
+#include <__pstl/cpu_algos/merge.h>
+#include <__pstl/cpu_algos/stable_sort.h>
+#include <__pstl/cpu_algos/transform.h>
+#include <__pstl/cpu_algos/transform_reduce.h>
 
 #endif // _LIBCPP___PSTL_BACKENDS_LIBDISPATCH_H
diff --git a/libcxx/include/__pstl/backends/serial.h b/libcxx/include/__pstl/backends/serial.h
index 8bb8945093096..6e343313bea36 100644
--- a/libcxx/include/__pstl/backends/serial.h
+++ b/libcxx/include/__pstl/backends/serial.h
@@ -82,13 +82,13 @@ _LIBCPP_POP_MACROS
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && && _LIBCPP_STD_VER >= 17
 
 // Implement PSTL algorithms based on the __cpu_traits specialized above
-#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
-#include <__algorithm/pstl_backends/cpu_backends/fill.h>
-#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
-#include <__algorithm/pstl_backends/cpu_backends/merge.h>
-#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+#include <__pstl/cpu_algos/any_of.h>
+#include <__pstl/cpu_algos/fill.h>
+#include <__pstl/cpu_algos/find_if.h>
+#include <__pstl/cpu_algos/for_each.h>
+#include <__pstl/cpu_algos/merge.h>
+#include <__pstl/cpu_algos/stable_sort.h>
+#include <__pstl/cpu_algos/transform.h>
+#include <__pstl/cpu_algos/transform_reduce.h>
 
 #endif // _LIBCPP___PSTL_BACKENDS_SERIAL_H
diff --git a/libcxx/include/__pstl/backends/std_thread.h b/libcxx/include/__pstl/backends/std_thread.h
index ab09f42cfdd8d..e58f4859e6c9e 100644
--- a/libcxx/include/__pstl/backends/std_thread.h
+++ b/libcxx/include/__pstl/backends/std_thread.h
@@ -85,13 +85,13 @@ _LIBCPP_END_NAMESPACE_STD
 _LIBCPP_POP_MACROS
 
 // Implement PSTL algorithms based on the __cpu_traits specialized above
-#include <__algorithm/pstl_backends/cpu_backends/any_of.h>
-#include <__algorithm/pstl_backends/cpu_backends/fill.h>
-#include <__algorithm/pstl_backends/cpu_backends/find_if.h>
-#include <__algorithm/pstl_backends/cpu_backends/for_each.h>
-#include <__algorithm/pstl_backends/cpu_backends/merge.h>
-#include <__algorithm/pstl_backends/cpu_backends/stable_sort.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform.h>
-#include <__algorithm/pstl_backends/cpu_backends/transform_reduce.h>
+#include <__pstl/cpu_algos/any_of.h>
+#include <__pstl/cpu_algos/fill.h>
+#include <__pstl/cpu_algos/find_if.h>
+#include <__pstl/cpu_algos/for_each.h>
+#include <__pstl/cpu_algos/merge.h>
+#include <__pstl/cpu_algos/stable_sort.h>
+#include <__pstl/cpu_algos/transform.h>
+#include <__pstl/cpu_algos/transform_reduce.h>
 
 #endif // _LIBCPP___PSTL_BACKENDS_STD_THREAD_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h b/libcxx/include/__pstl/cpu_algos/any_of.h
similarity index 95%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
rename to libcxx/include/__pstl/cpu_algos/any_of.h
index 3db4765da64b2..01b9d214310a3 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/any_of.h
+++ b/libcxx/include/__pstl/cpu_algos/any_of.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_ANY_OF_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_ANY_OF_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_ANY_OF_H
+#define _LIBCPP___PSTL_CPU_ALGOS_ANY_OF_H
 
 #include <__algorithm/any_of.h>
 #include <__algorithm/find_if.h>
@@ -96,4 +96,4 @@ _LIBCPP_POP_MACROS
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKEND_ANY_OF_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_ANY_OF_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h b/libcxx/include/__pstl/cpu_algos/fill.h
similarity index 92%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
rename to libcxx/include/__pstl/cpu_algos/fill.h
index b5a49f8417d32..66fb751eb7a2e 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/fill.h
+++ b/libcxx/include/__pstl/cpu_algos/fill.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FILL_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FILL_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_FILL_H
+#define _LIBCPP___PSTL_CPU_ALGOS_FILL_H
 
 #include <__algorithm/fill.h>
 #include <__config>
@@ -60,4 +60,4 @@ _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FILL_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_FILL_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h b/libcxx/include/__pstl/cpu_algos/find_if.h
similarity index 96%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
rename to libcxx/include/__pstl/cpu_algos/find_if.h
index 2b1754ea3a755..c99ec01bff48f 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/find_if.h
+++ b/libcxx/include/__pstl/cpu_algos/find_if.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FIND_IF_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FIND_IF_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_FIND_IF_H
+#define _LIBCPP___PSTL_CPU_ALGOS_FIND_IF_H
 
 #include <__algorithm/find_if.h>
 #include <__atomic/atomic.h>
@@ -132,4 +132,4 @@ _LIBCPP_POP_MACROS
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_FIND_IF_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_FIND_IF_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h b/libcxx/include/__pstl/cpu_algos/for_each.h
similarity index 92%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
rename to libcxx/include/__pstl/cpu_algos/for_each.h
index 6db212ead8ae6..cd7ce022469bd 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/for_each.h
+++ b/libcxx/include/__pstl/cpu_algos/for_each.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKNEDS_FOR_EACH_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKNEDS_FOR_EACH_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_FOR_EACH_H
+#define _LIBCPP___PSTL_CPU_ALGOS_FOR_EACH_H
 
 #include <__algorithm/for_each.h>
 #include <__config>
@@ -60,4 +60,4 @@ _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKNEDS_FOR_EACH_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_FOR_EACH_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h b/libcxx/include/__pstl/cpu_algos/merge.h
similarity index 93%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
rename to libcxx/include/__pstl/cpu_algos/merge.h
index f3e59e8c02854..b857fc1fb7a56 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/merge.h
+++ b/libcxx/include/__pstl/cpu_algos/merge.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_MERGE_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_MERGE_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_MERGE_H
+#define _LIBCPP___PSTL_CPU_ALGOS_MERGE_H
 
 #include <__algorithm/merge.h>
 #include <__config>
@@ -83,4 +83,4 @@ _LIBCPP_POP_MACROS
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_MERGE_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_MERGE_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h b/libcxx/include/__pstl/cpu_algos/stable_sort.h
similarity index 88%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
rename to libcxx/include/__pstl/cpu_algos/stable_sort.h
index 9ad8cc8fb0f2d..18effb2108a2f 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/stable_sort.h
+++ b/libcxx/include/__pstl/cpu_algos/stable_sort.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_STABLE_SORT_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_STABLE_SORT_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_STABLE_SORT_H
+#define _LIBCPP___PSTL_CPU_ALGOS_STABLE_SORT_H
 
 #include <__algorithm/stable_sort.h>
 #include <__config>
@@ -43,4 +43,4 @@ _LIBCPP_END_NAMESPACE_STD
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_STABLE_SORT_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_STABLE_SORT_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h b/libcxx/include/__pstl/cpu_algos/transform.h
similarity index 96%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
rename to libcxx/include/__pstl/cpu_algos/transform.h
index 65e166d847e12..70853dc9af24e 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform.h
+++ b/libcxx/include/__pstl/cpu_algos/transform.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_H
+#define _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_H
 
 #include <__algorithm/transform.h>
 #include <__config>
@@ -136,4 +136,4 @@ _LIBCPP_POP_MACROS
 
 #endif // !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_H
diff --git a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h b/libcxx/include/__pstl/cpu_algos/transform_reduce.h
similarity index 97%
rename from libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
rename to libcxx/include/__pstl/cpu_algos/transform_reduce.h
index af481d505bb91..a85ee9fb773af 100644
--- a/libcxx/include/__algorithm/pstl_backends/cpu_backends/transform_reduce.h
+++ b/libcxx/include/__pstl/cpu_algos/transform_reduce.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_REDUCE_H
-#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_REDUCE_H
+#ifndef _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_REDUCE_H
+#define _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_REDUCE_H
 
 #include <__config>
 #include <__iterator/concepts.h>
@@ -203,4 +203,4 @@ _LIBCPP_END_NAMESPACE_STD
 
 _LIBCPP_POP_MACROS
 
-#endif // _LIBCPP___ALGORITHM_PSTL_BACKENDS_CPU_BACKENDS_TRANSFORM_REDUCE_H
+#endif // _LIBCPP___PSTL_CPU_ALGOS_TRANSFORM_REDUCE_H
diff --git a/libcxx/include/libcxx.imp b/libcxx/include/libcxx.imp
index a4e2690fc55c9..c224090e72465 100644
--- a/libcxx/include/libcxx.imp
+++ b/libcxx/include/libcxx.imp
@@ -73,14 +73,6 @@
   { include: [ "<__algorithm/pop_heap.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/prev_permutation.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_any_all_none_of.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/any_of.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/fill.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/find_if.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/for_each.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/merge.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/stable_sort.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/transform.h>", "private", "<algorithm>", "public" ] },
-  { include: [ "<__algorithm/pstl_backends/cpu_backends/transform_reduce.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_copy.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_count.h>", "private", "<algorithm>", "public" ] },
   { include: [ "<__algorithm/pstl_equal.h>", "private", "<algorithm>", "public" ] },
diff --git a/libcxx/include/module.modulemap b/libcxx/include/module.modulemap
index f996c2cc05459..62ac431664748 100644
--- a/libcxx/include/module.modulemap
+++ b/libcxx/include/module.modulemap
@@ -714,14 +714,6 @@ module std_private_algorithm_partition_point                             [system
 module std_private_algorithm_pop_heap                                    [system] { header "__algorithm/pop_heap.h" }
 module std_private_algorithm_prev_permutation                            [system] { header "__algorithm/prev_permutation.h" }
 module std_private_algorithm_pstl_any_all_none_of                        [system] { header "__algorithm/pstl_any_all_none_of.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_any_of           [system] { textual header "__algorithm/pstl_backends/cpu_backends/any_of.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_fill             [system] { textual header "__algorithm/pstl_backends/cpu_backends/fill.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_find_if          [system] { textual header "__algorithm/pstl_backends/cpu_backends/find_if.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_for_each         [system] { textual header "__algorithm/pstl_backends/cpu_backends/for_each.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_merge            [system] { textual header "__algorithm/pstl_backends/cpu_backends/merge.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_stable_sort      [system] { textual header "__algorithm/pstl_backends/cpu_backends/stable_sort.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_transform        [system] { textual header "__algorithm/pstl_backends/cpu_backends/transform.h" }
-module std_private_algorithm_pstl_backends_cpu_backends_transform_reduce [system] { textual header "__algorithm/pstl_backends/cpu_backends/transform_reduce.h" }
 module std_private_algorithm_pstl_copy                                   [system] { header "__algorithm/pstl_copy.h" }
 module std_private_algorithm_pstl_count                                  [system] { header "__algorithm/pstl_count.h" }
 module std_private_algorithm_pstl_equal                                  [system] { header "__algorithm/pstl_equal.h" }
@@ -1595,15 +1587,23 @@ module std_private_numeric_transform_exclusive_scan [system] { header "__numeric
 module std_private_numeric_transform_inclusive_scan [system] { header "__numeric/transform_inclusive_scan.h" }
 module std_private_numeric_transform_reduce         [system] { header "__numeric/transform_reduce.h" }
 
-module std_private_pstl_backends_libdispatch      [system] { header "__pstl/backends/libdispatch.h" }
-module std_private_pstl_backends_serial           [system] { header "__pstl/backends/serial.h" }
-module std_private_pstl_backends_std_thread       [system] { header "__pstl/backends/std_thread.h" }
-module std_private_pstl_cpu_algos_cpu_traits      [system] { header "__pstl/cpu_algos/cpu_traits.h" }
-module std_private_pstl_configuration_fwd         [system] {
+module std_private_pstl_backends_libdispatch       [system] { header "__pstl/backends/libdispatch.h" }
+module std_private_pstl_backends_serial            [system] { header "__pstl/backends/serial.h" }
+module std_private_pstl_backends_std_thread        [system] { header "__pstl/backends/std_thread.h" }
+module std_private_pstl_cpu_algos_any_of           [system] { textual header "__pstl/cpu_algos/any_of.h" }
+module std_private_pstl_cpu_algos_cpu_traits       [system] { header "__pstl/cpu_algos/cpu_traits.h" }
+module std_private_pstl_cpu_algos_fill             [system] { textual header "__pstl/cpu_algos/fill.h" }
+module std_private_pstl_cpu_algos_find_if          [system] { textual header "__pstl/cpu_algos/find_if.h" }
+module std_private_pstl_cpu_algos_for_each         [system] { textual header "__pstl/cpu_algos/for_each.h" }
+module std_private_pstl_cpu_algos_merge            [system] { textual header "__pstl/cpu_algos/merge.h" }
+module std_private_pstl_cpu_algos_stable_sort      [system] { textual header "__pstl/cpu_algos/stable_sort.h" }
+module std_private_pstl_cpu_algos_transform        [system] { textual header "__pstl/cpu_algos/transform.h" }
+module std_private_pstl_cpu_algos_transform_reduce [system] { textual header "__pstl/cpu_algos/transform_reduce.h" }
+module std_private_pstl_configuration_fwd          [system] {
   header "__pstl/configuration_fwd.h"
   export *
 }
-module std_private_pstl_configuration             [system] {
+module std_private_pstl_configuration              [system] {
   header "__pstl/configuration.h"
   export *
 }

From af7a82fdc717c9cbd1b77ff4168292098a1f6e8a Mon Sep 17 00:00:00 2001
From: "S. Bharadwaj Yadavalli" <Bharadwaj.Yadavalli@microsoft.com>
Date: Thu, 18 Apr 2024 07:58:10 -0400
Subject: [PATCH 868/886] [NFC] Format code of parseArch() (#89152)

In preparation for a cleaner addition in a following PR.
---
 llvm/lib/TargetParser/Triple.cpp | 151 ++++++++++++++++---------------
 1 file changed, 76 insertions(+), 75 deletions(-)

diff --git a/llvm/lib/TargetParser/Triple.cpp b/llvm/lib/TargetParser/Triple.cpp
index 624679ff507a7..9c73f515ae689 100644
--- a/llvm/lib/TargetParser/Triple.cpp
+++ b/llvm/lib/TargetParser/Triple.cpp
@@ -500,81 +500,82 @@ static Triple::ArchType parseARMArch(StringRef ArchName) {
 }
 
 static Triple::ArchType parseArch(StringRef ArchName) {
-  auto AT = StringSwitch<Triple::ArchType>(ArchName)
-    .Cases("i386", "i486", "i586", "i686", Triple::x86)
-    // FIXME: Do we need to support these?
-    .Cases("i786", "i886", "i986", Triple::x86)
-    .Cases("amd64", "x86_64", "x86_64h", Triple::x86_64)
-    .Cases("powerpc", "powerpcspe", "ppc", "ppc32", Triple::ppc)
-    .Cases("powerpcle", "ppcle", "ppc32le", Triple::ppcle)
-    .Cases("powerpc64", "ppu", "ppc64", Triple::ppc64)
-    .Cases("powerpc64le", "ppc64le", Triple::ppc64le)
-    .Case("xscale", Triple::arm)
-    .Case("xscaleeb", Triple::armeb)
-    .Case("aarch64", Triple::aarch64)
-    .Case("aarch64_be", Triple::aarch64_be)
-    .Case("aarch64_32", Triple::aarch64_32)
-    .Case("arc", Triple::arc)
-    .Case("arm64", Triple::aarch64)
-    .Case("arm64_32", Triple::aarch64_32)
-    .Case("arm64e", Triple::aarch64)
-    .Case("arm64ec", Triple::aarch64)
-    .Case("arm", Triple::arm)
-    .Case("armeb", Triple::armeb)
-    .Case("thumb", Triple::thumb)
-    .Case("thumbeb", Triple::thumbeb)
-    .Case("avr", Triple::avr)
-    .Case("m68k", Triple::m68k)
-    .Case("msp430", Triple::msp430)
-    .Cases("mips", "mipseb", "mipsallegrex", "mipsisa32r6",
-           "mipsr6", Triple::mips)
-    .Cases("mipsel", "mipsallegrexel", "mipsisa32r6el", "mipsr6el",
-           Triple::mipsel)
-    .Cases("mips64", "mips64eb", "mipsn32", "mipsisa64r6",
-           "mips64r6", "mipsn32r6", Triple::mips64)
-    .Cases("mips64el", "mipsn32el", "mipsisa64r6el", "mips64r6el",
-           "mipsn32r6el", Triple::mips64el)
-    .Case("r600", Triple::r600)
-    .Case("amdgcn", Triple::amdgcn)
-    .Case("riscv32", Triple::riscv32)
-    .Case("riscv64", Triple::riscv64)
-    .Case("hexagon", Triple::hexagon)
-    .Cases("s390x", "systemz", Triple::systemz)
-    .Case("sparc", Triple::sparc)
-    .Case("sparcel", Triple::sparcel)
-    .Cases("sparcv9", "sparc64", Triple::sparcv9)
-    .Case("tce", Triple::tce)
-    .Case("tcele", Triple::tcele)
-    .Case("xcore", Triple::xcore)
-    .Case("nvptx", Triple::nvptx)
-    .Case("nvptx64", Triple::nvptx64)
-    .Case("le32", Triple::le32)
-    .Case("le64", Triple::le64)
-    .Case("amdil", Triple::amdil)
-    .Case("amdil64", Triple::amdil64)
-    .Case("hsail", Triple::hsail)
-    .Case("hsail64", Triple::hsail64)
-    .Case("spir", Triple::spir)
-    .Case("spir64", Triple::spir64)
-    .Cases("spirv", "spirv1.5", "spirv1.6", Triple::spirv)
-    .Cases("spirv32", "spirv32v1.0", "spirv32v1.1", "spirv32v1.2",
-           "spirv32v1.3", "spirv32v1.4", "spirv32v1.5", Triple::spirv32)
-    .Cases("spirv64", "spirv64v1.0", "spirv64v1.1", "spirv64v1.2",
-           "spirv64v1.3", "spirv64v1.4", "spirv64v1.5", Triple::spirv64)
-    .StartsWith("kalimba", Triple::kalimba)
-    .Case("lanai", Triple::lanai)
-    .Case("renderscript32", Triple::renderscript32)
-    .Case("renderscript64", Triple::renderscript64)
-    .Case("shave", Triple::shave)
-    .Case("ve", Triple::ve)
-    .Case("wasm32", Triple::wasm32)
-    .Case("wasm64", Triple::wasm64)
-    .Case("csky", Triple::csky)
-    .Case("loongarch32", Triple::loongarch32)
-    .Case("loongarch64", Triple::loongarch64)
-    .Case("dxil", Triple::dxil)
-    .Case("xtensa", Triple::xtensa)
-    .Default(Triple::UnknownArch);
+  auto AT =
+      StringSwitch<Triple::ArchType>(ArchName)
+          .Cases("i386", "i486", "i586", "i686", Triple::x86)
+          // FIXME: Do we need to support these?
+          .Cases("i786", "i886", "i986", Triple::x86)
+          .Cases("amd64", "x86_64", "x86_64h", Triple::x86_64)
+          .Cases("powerpc", "powerpcspe", "ppc", "ppc32", Triple::ppc)
+          .Cases("powerpcle", "ppcle", "ppc32le", Triple::ppcle)
+          .Cases("powerpc64", "ppu", "ppc64", Triple::ppc64)
+          .Cases("powerpc64le", "ppc64le", Triple::ppc64le)
+          .Case("xscale", Triple::arm)
+          .Case("xscaleeb", Triple::armeb)
+          .Case("aarch64", Triple::aarch64)
+          .Case("aarch64_be", Triple::aarch64_be)
+          .Case("aarch64_32", Triple::aarch64_32)
+          .Case("arc", Triple::arc)
+          .Case("arm64", Triple::aarch64)
+          .Case("arm64_32", Triple::aarch64_32)
+          .Case("arm64e", Triple::aarch64)
+          .Case("arm64ec", Triple::aarch64)
+          .Case("arm", Triple::arm)
+          .Case("armeb", Triple::armeb)
+          .Case("thumb", Triple::thumb)
+          .Case("thumbeb", Triple::thumbeb)
+          .Case("avr", Triple::avr)
+          .Case("m68k", Triple::m68k)
+          .Case("msp430", Triple::msp430)
+          .Cases("mips", "mipseb", "mipsallegrex", "mipsisa32r6", "mipsr6",
+                 Triple::mips)
+          .Cases("mipsel", "mipsallegrexel", "mipsisa32r6el", "mipsr6el",
+                 Triple::mipsel)
+          .Cases("mips64", "mips64eb", "mipsn32", "mipsisa64r6", "mips64r6",
+                 "mipsn32r6", Triple::mips64)
+          .Cases("mips64el", "mipsn32el", "mipsisa64r6el", "mips64r6el",
+                 "mipsn32r6el", Triple::mips64el)
+          .Case("r600", Triple::r600)
+          .Case("amdgcn", Triple::amdgcn)
+          .Case("riscv32", Triple::riscv32)
+          .Case("riscv64", Triple::riscv64)
+          .Case("hexagon", Triple::hexagon)
+          .Cases("s390x", "systemz", Triple::systemz)
+          .Case("sparc", Triple::sparc)
+          .Case("sparcel", Triple::sparcel)
+          .Cases("sparcv9", "sparc64", Triple::sparcv9)
+          .Case("tce", Triple::tce)
+          .Case("tcele", Triple::tcele)
+          .Case("xcore", Triple::xcore)
+          .Case("nvptx", Triple::nvptx)
+          .Case("nvptx64", Triple::nvptx64)
+          .Case("le32", Triple::le32)
+          .Case("le64", Triple::le64)
+          .Case("amdil", Triple::amdil)
+          .Case("amdil64", Triple::amdil64)
+          .Case("hsail", Triple::hsail)
+          .Case("hsail64", Triple::hsail64)
+          .Case("spir", Triple::spir)
+          .Case("spir64", Triple::spir64)
+          .Cases("spirv", "spirv1.5", "spirv1.6", Triple::spirv)
+          .Cases("spirv32", "spirv32v1.0", "spirv32v1.1", "spirv32v1.2",
+                 "spirv32v1.3", "spirv32v1.4", "spirv32v1.5", Triple::spirv32)
+          .Cases("spirv64", "spirv64v1.0", "spirv64v1.1", "spirv64v1.2",
+                 "spirv64v1.3", "spirv64v1.4", "spirv64v1.5", Triple::spirv64)
+          .StartsWith("kalimba", Triple::kalimba)
+          .Case("lanai", Triple::lanai)
+          .Case("renderscript32", Triple::renderscript32)
+          .Case("renderscript64", Triple::renderscript64)
+          .Case("shave", Triple::shave)
+          .Case("ve", Triple::ve)
+          .Case("wasm32", Triple::wasm32)
+          .Case("wasm64", Triple::wasm64)
+          .Case("csky", Triple::csky)
+          .Case("loongarch32", Triple::loongarch32)
+          .Case("loongarch64", Triple::loongarch64)
+          .Case("dxil", Triple::dxil)
+          .Case("xtensa", Triple::xtensa)
+          .Default(Triple::UnknownArch);
 
   // Some architectures require special parsing logic just to compute the
   // ArchType result.

From 847f91e144517902f0c9ec28b7c5782b7af3a96c Mon Sep 17 00:00:00 2001
From: Sven van Haastregt <sven.vanhaastregt@arm.com>
Date: Thu, 11 Apr 2024 16:57:03 +0200
Subject: [PATCH 869/886] Start moving away from insert-before Instruction
 pointers (#2498)

In preparation for LLVM's deprecation of Instruction constructors and
Create methods that accept `Instruction` pointers, upgrade call sites
to pass an iterator instead of an `Instruction` pointer.

This patch only upgrades constructor and `Create` calls that were
trivial to upgrade with no impact on the test suite.  Some calls
remain to be upgraded as future work.

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/7fb89d5a0f2dfd7
---
 llvm-spirv/lib/SPIRV/OCLToSPIRV.cpp     | 28 ++++++++++++++-----------
 llvm-spirv/lib/SPIRV/SPIRVLowerBool.cpp |  7 ++++---
 llvm-spirv/lib/SPIRV/SPIRVReader.cpp    |  6 ++++--
 llvm-spirv/lib/SPIRV/SPIRVToOCL.cpp     | 17 ++++++++-------
 llvm-spirv/lib/SPIRV/SPIRVToOCL20.cpp   |  2 +-
 llvm-spirv/lib/SPIRV/SPIRVUtil.cpp      | 11 +++++-----
 llvm-spirv/lib/SPIRV/SPIRVWriter.cpp    |  2 +-
 7 files changed, 41 insertions(+), 32 deletions(-)

diff --git a/llvm-spirv/lib/SPIRV/OCLToSPIRV.cpp b/llvm-spirv/lib/SPIRV/OCLToSPIRV.cpp
index 2cc397f7257db..6778b0cf32ba6 100644
--- a/llvm-spirv/lib/SPIRV/OCLToSPIRV.cpp
+++ b/llvm-spirv/lib/SPIRV/OCLToSPIRV.cpp
@@ -499,7 +499,8 @@ CallInst *OCLToSPIRVBase::visitCallAtomicCmpXchg(CallInst *CI) {
 }
 
 void OCLToSPIRVBase::visitCallAtomicInit(CallInst *CI) {
-  auto *ST = new StoreInst(CI->getArgOperand(1), CI->getArgOperand(0), CI);
+  auto *ST = new StoreInst(CI->getArgOperand(1), CI->getArgOperand(0),
+                           CI->getIterator());
   ST->takeName(CI);
   CI->dropAllReferences();
   CI->eraseFromParent();
@@ -515,11 +516,11 @@ void OCLToSPIRVBase::visitCallAllAny(spv::Op OC, CallInst *CI) {
   auto *Zero = Constant::getNullValue(Args[0]->getType());
 
   auto *Cmp = CmpInst::Create(CmpInst::ICmp, CmpInst::ICMP_SLT, Args[0], Zero,
-                              "cast", CI);
+                              "cast", CI->getIterator());
 
   if (!isa<VectorType>(ArgTy)) {
-    auto *Cast = CastInst::CreateZExtOrBitCast(Cmp, Type::getInt32Ty(*Ctx), "",
-                                               Cmp->getNextNode());
+    auto *Cast = CastInst::CreateZExtOrBitCast(
+        Cmp, Type::getInt32Ty(*Ctx), "", Cmp->getNextNode()->getIterator());
     CI->replaceAllUsesWith(Cast);
     CI->eraseFromParent();
   } else {
@@ -1040,13 +1041,15 @@ void OCLToSPIRVBase::visitCallGetImageSize(CallInst *CI,
             Constant *Index[] = {getInt32(M, 0), getInt32(M, 1), getInt32(M, 2),
                                  getInt32(M, 3)};
             return new ShuffleVectorInst(NCI, ZeroVec,
-                                         ConstantVector::get(Index), "", CI);
+                                         ConstantVector::get(Index), "",
+                                         CI->getIterator());
 
           } else if (Desc.Dim == Dim2D && Desc.Arrayed) {
             Constant *Index[] = {getInt32(M, 0), getInt32(M, 1)};
             Constant *Mask = ConstantVector::get(Index);
             return new ShuffleVectorInst(NCI, UndefValue::get(NCI->getType()),
-                                         Mask, NCI->getName(), CI);
+                                         Mask, NCI->getName(),
+                                         CI->getIterator());
           }
           return NCI;
         }
@@ -1056,7 +1059,7 @@ void OCLToSPIRVBase::visitCallGetImageSize(CallInst *CI,
                          .Case(kOCLBuiltinName::GetImageDepth, 2)
                          .Case(kOCLBuiltinName::GetImageArraySize, Dim - 1);
         return ExtractElementInst::Create(NCI, getUInt32(M, I), "",
-                                          NCI->getNextNode());
+                                          NCI->getNextNode()->getIterator());
       });
 }
 
@@ -1371,11 +1374,12 @@ void OCLToSPIRVBase::visitCallScalToVec(CallInst *CI, StringRef MangledName,
                               getExtOp(MangledName, DemangledName)));
   for (auto I : ScalarPos)
     Mutator.mapArg(I, [&](Value *V) {
-      Instruction *Inst = InsertElementInst::Create(UndefValue::get(VecTy), V,
-                                                    getInt32(M, 0), "", CI);
+      Instruction *Inst = InsertElementInst::Create(
+          UndefValue::get(VecTy), V, getInt32(M, 0), "", CI->getIterator());
       return new ShuffleVectorInst(
           Inst, UndefValue::get(VecTy),
-          ConstantVector::getSplat(VecElemCount, getInt32(M, 0)), "", CI);
+          ConstantVector::getSplat(VecElemCount, getInt32(M, 0)), "",
+          CI->getIterator());
     });
 }
 
@@ -1488,7 +1492,7 @@ void OCLToSPIRVBase::visitCallEnqueueKernel(CallInst *CI,
           LocalSizeArray->getSourceElementType(), // Pointee type
           LocalSizeArray->getPointerOperand(),    // Alloca
           {getInt32(M, 0), getInt32(M, I)},       // Indices
-          "", CI));
+          "", CI->getIterator()));
   }
 
   StringRef NewName = "__spirv_EnqueueKernel__";
@@ -1497,7 +1501,7 @@ void OCLToSPIRVBase::visitCallEnqueueKernel(CallInst *CI,
   Function *NewF =
       Function::Create(FT, GlobalValue::ExternalLinkage, NewName, M);
   NewF->setCallingConv(CallingConv::SPIR_FUNC);
-  CallInst *NewCall = CallInst::Create(NewF, Args, "", CI);
+  CallInst *NewCall = CallInst::Create(NewF, Args, "", CI->getIterator());
   NewCall->setCallingConv(NewF->getCallingConv());
   CI->replaceAllUsesWith(NewCall);
   CI->eraseFromParent();
diff --git a/llvm-spirv/lib/SPIRV/SPIRVLowerBool.cpp b/llvm-spirv/lib/SPIRV/SPIRVLowerBool.cpp
index 8277025cda879..deb9dd59898d7 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVLowerBool.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVLowerBool.cpp
@@ -68,7 +68,8 @@ void SPIRVLowerBoolBase::visitTruncInst(TruncInst &I) {
   if (isBoolType(I.getType())) {
     auto *Op = I.getOperand(0);
     auto *And = BinaryOperator::CreateAnd(
-        Op, getScalarOrVectorConstantInt(Op->getType(), 1, false), "", &I);
+        Op, getScalarOrVectorConstantInt(Op->getType(), 1, false), "",
+        I.getIterator());
     And->setDebugLoc(I.getDebugLoc());
     auto *Zero = getScalarOrVectorConstantInt(Op->getType(), 0, false);
     auto *Cmp = new ICmpInst(&I, CmpInst::ICMP_NE, And, Zero);
@@ -85,7 +86,7 @@ void SPIRVLowerBoolBase::handleExtInstructions(Instruction &I) {
     auto *One = getScalarOrVectorConstantInt(
         Ty, (Opcode == Instruction::SExt) ? ~0 : 1, false);
     assert(Zero && One && "Couldn't create constant int");
-    auto *Sel = SelectInst::Create(Op, One, Zero, "", &I);
+    auto *Sel = SelectInst::Create(Op, One, Zero, "", I.getIterator());
     replace(&I, Sel);
   }
 }
@@ -100,7 +101,7 @@ void SPIRVLowerBoolBase::handleCastInstructions(Instruction &I) {
     auto *Zero = getScalarOrVectorConstantInt(Ty, 0, false);
     auto *One = getScalarOrVectorConstantInt(Ty, 1, false);
     assert(Zero && One && "Couldn't create constant int");
-    auto *Sel = SelectInst::Create(Op, One, Zero, "", &I);
+    auto *Sel = SelectInst::Create(Op, One, Zero, "", I.getIterator());
     Sel->setDebugLoc(I.getDebugLoc());
     I.setOperand(0, Sel);
   }
diff --git a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
index fae1f52704a9a..299cb4ab18666 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
@@ -1196,10 +1196,12 @@ SPIRVToLLVM::expandOCLBuiltinWithScalarArg(CallInst *CI,
       else {
         NewVec = ConstantVector::getSplat(
             VecElemCount, Constant::getNullValue(Arg->getType()));
-        NewVec = InsertElementInst::Create(NewVec, Arg, getInt32(M, 0), "", CI);
+        NewVec = InsertElementInst::Create(NewVec, Arg, getInt32(M, 0), "",
+                                           CI->getIterator());
         NewVec = new ShuffleVectorInst(
             NewVec, NewVec,
-            ConstantVector::getSplat(VecElemCount, getInt32(M, 0)), "", CI);
+            ConstantVector::getSplat(VecElemCount, getInt32(M, 0)), "",
+            CI->getIterator());
       }
       NewVec->takeName(Arg);
       return NewVec;
diff --git a/llvm-spirv/lib/SPIRV/SPIRVToOCL.cpp b/llvm-spirv/lib/SPIRV/SPIRVToOCL.cpp
index c00c9f1891e99..87a89b945a56f 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVToOCL.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVToOCL.cpp
@@ -276,9 +276,9 @@ void SPIRVToOCLBase::visitCallSPIRVImageQuerySize(CallInst *CI) {
     // The width of integer type returning by OpImageQuerySize[Lod] may
     // differ from i32
     if (CI->getType()->getScalarType() != Int32Ty) {
-      GetImageSize = CastInst::CreateIntegerCast(GetImageSize,
-                                                 CI->getType()->getScalarType(),
-                                                 false, CI->getName(), CI);
+      GetImageSize = CastInst::CreateIntegerCast(
+          GetImageSize, CI->getType()->getScalarType(), false, CI->getName(),
+          CI->getIterator());
     }
   } else {
     assert((ImgDim == 2 || ImgDim == 3) && "invalid image type");
@@ -298,7 +298,7 @@ void SPIRVToOCLBase::visitCallSPIRVImageQuerySize(CallInst *CI) {
           FixedVectorType::get(
               CI->getType()->getScalarType(),
               cast<FixedVectorType>(GetImageSize->getType())->getNumElements()),
-          false, CI->getName(), CI);
+          false, CI->getName(), CI->getIterator());
     }
   }
 
@@ -313,7 +313,7 @@ void SPIRVToOCLBase::visitCallSPIRVImageQuerySize(CallInst *CI) {
              "OpImageQuerySize[Lod] must return <2 x iN> vector type");
       GetImageSize = InsertElementInst::Create(
           UndefValue::get(VecTy), GetImageSize, ConstantInt::get(Int32Ty, 0),
-          CI->getName(), CI);
+          CI->getName(), CI->getIterator());
     } else {
       // get_image_dim and OpImageQuerySize returns different vector
       // types for arrayed and 3d images.
@@ -324,7 +324,7 @@ void SPIRVToOCLBase::visitCallSPIRVImageQuerySize(CallInst *CI) {
 
       GetImageSize = new ShuffleVectorInst(
           GetImageSize, UndefValue::get(GetImageSize->getType()), Mask,
-          CI->getName(), CI);
+          CI->getName(), CI->getIterator());
     }
   }
 
@@ -341,12 +341,13 @@ void SPIRVToOCLBase::visitCallSPIRVImageQuerySize(CallInst *CI) {
     // differ from size_t which is returned by get_image_array_size
     if (GetImageArraySize->getType() != VecTy->getElementType()) {
       GetImageArraySize = CastInst::CreateIntegerCast(
-          GetImageArraySize, VecTy->getElementType(), false, CI->getName(), CI);
+          GetImageArraySize, VecTy->getElementType(), false, CI->getName(),
+          CI->getIterator());
     }
     GetImageSize = InsertElementInst::Create(
         GetImageSize, GetImageArraySize,
         ConstantInt::get(Int32Ty, VecTy->getNumElements() - 1), CI->getName(),
-        CI);
+        CI->getIterator());
   }
 
   assert(GetImageSize && "must not be null");
diff --git a/llvm-spirv/lib/SPIRV/SPIRVToOCL20.cpp b/llvm-spirv/lib/SPIRV/SPIRVToOCL20.cpp
index 95b04aa69780c..a94a9d4a8dd8d 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVToOCL20.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVToOCL20.cpp
@@ -263,7 +263,7 @@ void SPIRVToOCL20Base::visitCallSPIRVEnqueueKernel(CallInst *CI, Op OC) {
   auto Mutator = mutateCallInst(CI, FName.str());
   Mutator.mapArg(6, [=](IRBuilder<> &Builder, Value *Invoke) {
     Value *Replace = CastInst::CreatePointerBitCastOrAddrSpaceCast(
-        Invoke, Builder.getPtrTy(SPIRAS_Generic), "", CI);
+        Invoke, Builder.getPtrTy(SPIRAS_Generic), "", CI->getIterator());
     return std::make_pair(
         Replace, TypedPointerType::get(Builder.getInt8Ty(), SPIRAS_Generic));
   });
diff --git a/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp b/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
index 3758fd4d81b56..cafb738adf214 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
@@ -2168,13 +2168,13 @@ bool lowerBuiltinCallsToVariables(Module *M) {
       auto *CI = dyn_cast<CallInst>(U);
       assert(CI && "invalid instruction");
       const DebugLoc &DLoc = CI->getDebugLoc();
-      Instruction *NewValue = new LoadInst(GVType, BV, "", CI);
+      Instruction *NewValue = new LoadInst(GVType, BV, "", CI->getIterator());
       if (DLoc)
         NewValue->setDebugLoc(DLoc);
       LLVM_DEBUG(dbgs() << "Transform: " << *CI << " => " << *NewValue << '\n');
       if (IsVec) {
-        NewValue =
-            ExtractElementInst::Create(NewValue, CI->getArgOperand(0), "", CI);
+        NewValue = ExtractElementInst::Create(NewValue, CI->getArgOperand(0),
+                                              "", CI->getIterator());
         if (DLoc)
           NewValue->setDebugLoc(DLoc);
         LLVM_DEBUG(dbgs() << *NewValue << '\n');
@@ -2272,11 +2272,12 @@ bool postProcessBuiltinWithArrayArguments(Function *F,
           if (!T->isArrayTy())
             continue;
           auto *Alloca = new AllocaInst(T, 0, "", &(*FBegin));
-          new StoreInst(I, Alloca, false, CI);
+          new StoreInst(I, Alloca, false, CI->getIterator());
           auto *Zero =
               ConstantInt::getNullValue(Type::getInt32Ty(T->getContext()));
           Value *Index[] = {Zero, Zero};
-          I = GetElementPtrInst::CreateInBounds(T, Alloca, Index, "", CI);
+          I = GetElementPtrInst::CreateInBounds(T, Alloca, Index, "",
+                                                CI->getIterator());
         }
         return Name.str();
       },
diff --git a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
index 5ed032df8bb4a..12e798a300c40 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
@@ -5501,7 +5501,7 @@ void LLVMToSPIRVBase::mutateFuncArgType(
       auto CastF = M->getOrInsertFunction(SPCV_CAST, I.second, OrigTy);
       std::vector<Value *> Args;
       Args.push_back(Arg);
-      auto *Cast = CallInst::Create(CastF, Args, "", Call);
+      auto *Cast = CallInst::Create(CastF, Args, "", Call->getIterator());
       Call->replaceUsesOfWith(Arg, Cast);
       SPIRVDBG(dbgs() << "[mutate arg type] -> " << *Cast << '\n');
     }

From c5f8977efa0a0c81cfcf40b6d3f2a6b29352ba67 Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Fri, 12 Apr 2024 19:46:44 +0200
Subject: [PATCH 870/886] Add CopyLogical instruction (#2484)

There is no mapping to LLVM instructions, so it can be used only via SPIR-V friendly translation.

This addresses p1. of #2460

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/8518a6f72fd586c
---
 llvm-spirv/lib/SPIRV/SPIRVReader.cpp          |  4 +++
 .../lib/SPIRV/libSPIRV/SPIRVInstruction.h     | 25 +++++++++++++++++++
 .../lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h      |  1 +
 llvm-spirv/test/OpCopyLogical.spvasm          | 25 +++++++++++++++++++
 4 files changed, 55 insertions(+)
 create mode 100644 llvm-spirv/test/OpCopyLogical.spvasm

diff --git a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
index 299cb4ab18666..985d28bd9a1a0 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
@@ -2105,6 +2105,10 @@ Value *SPIRVToLLVM::transValueWithoutDecoration(SPIRVValue *BV, Function *F,
     LoadInst *LI = new LoadInst(Ty, AI, "", BB);
     return mapValue(BV, LI);
   }
+  case OpCopyLogical: {
+    SPIRVCopyLogical *CL = static_cast<SPIRVCopyLogical *>(BV);
+    return mapValue(BV, transSPIRVBuiltinFromInst(CL, BB));
+  }
 
   case OpAccessChain:
   case OpInBoundsAccessChain:
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
index b2144a98cfb33..7820b70da7ac8 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
@@ -2035,6 +2035,31 @@ class SPIRVCopyObject : public SPIRVInstruction {
   SPIRVId Operand;
 };
 
+class SPIRVCopyLogical : public SPIRVInstruction {
+public:
+  const static Op OC = OpCopyLogical;
+
+  // Complete constructor
+  SPIRVCopyLogical(SPIRVType *TheType, SPIRVId TheId, SPIRVValue *TheOperand,
+                   SPIRVBasicBlock *TheBB)
+      : SPIRVInstruction(4, OC, TheType, TheId, TheBB),
+        Operand(TheOperand->getId()) {
+    validate();
+    assert(TheBB && "Invalid BB");
+  }
+  // Incomplete constructor
+  SPIRVCopyLogical() : SPIRVInstruction(OC), Operand(SPIRVID_INVALID) {}
+
+  SPIRVValue *getOperand() { return getValue(Operand); }
+  std::vector<SPIRVValue *> getOperands() override { return {getOperand()}; }
+
+protected:
+  _SPIRV_DEF_ENCDEC3(Type, Id, Operand)
+
+  void validate() const override { SPIRVInstruction::validate(); }
+  SPIRVId Operand;
+};
+
 class SPIRVCopyMemory : public SPIRVInstruction, public SPIRVMemoryAccess {
 public:
   const static Op OC = OpCopyMemory;
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
index 0d81003e6fd89..ebed32c3476fd 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
@@ -329,6 +329,7 @@ _SPIRV_OP(GroupNonUniformBitwiseXor, 361)
 _SPIRV_OP(GroupNonUniformLogicalAnd, 362)
 _SPIRV_OP(GroupNonUniformLogicalOr, 363)
 _SPIRV_OP(GroupNonUniformLogicalXor, 364)
+_SPIRV_OP(CopyLogical, 400)
 _SPIRV_OP(GroupNonUniformRotateKHR, 4431)
 _SPIRV_OP(SDotKHR, 4450)
 _SPIRV_OP(UDotKHR, 4451)
diff --git a/llvm-spirv/test/OpCopyLogical.spvasm b/llvm-spirv/test/OpCopyLogical.spvasm
new file mode 100644
index 0000000000000..a8874c596190e
--- /dev/null
+++ b/llvm-spirv/test/OpCopyLogical.spvasm
@@ -0,0 +1,25 @@
+; Check support of OpCopyLogical instruction that was added in SPIR-V 1.4
+
+; REQUIRES: spirv-as
+; RUN: spirv-as --target-env spv1.4 -o %t.spv %s
+; RUN: spirv-val %t.spv
+; RUN: llvm-spirv -r %t.spv -o %t.rev.bc
+; RUN: llvm-dis %t.rev.bc
+; RUN: FileCheck < %t.rev.ll %s --check-prefix=CHECK-LLVM
+               OpCapability Addresses
+               OpCapability Kernel
+               OpMemoryModel Physical32 OpenCL
+               OpEntryPoint Kernel %1 "test"
+               OpName %entry "entry"
+       %void = OpTypeVoid
+  %_struct_4 = OpTypeStruct
+  %_struct_5 = OpTypeStruct
+          %6 = OpConstantComposite %_struct_4
+          %7 = OpTypeFunction %void
+          %1 = OpFunction %void None %7
+      %entry = OpLabel
+          %8 = OpCopyLogical %_struct_5 %6
+               OpReturn
+               OpFunctionEnd
+
+; CHECK-LLVM: @_Z19__spirv_CopyLogical12structtype.0(ptr sret(%structtype) %[[#]], %structtype.0 zeroinitializer)

From 40235249de835dd7629c4dcd6215e8fd534f5858 Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Mon, 15 Apr 2024 14:56:35 +0200
Subject: [PATCH 871/886] Support OpPtrEqual, OpPtrNotEqual and OpPtrDiff to
 compare pointers (#2482)

This addresses p.2 and p.3 of #2460

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/4b4508438a6dfe5
---
 llvm-spirv/lib/SPIRV/SPIRVReader.cpp          | 22 +++++++
 .../lib/SPIRV/libSPIRV/SPIRVInstruction.h     | 14 +++++
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCode.h   |  4 ++
 .../lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h      |  3 +
 llvm-spirv/test/transcoding/ptr_diff.ll       | 52 ++++++++++++++++
 llvm-spirv/test/transcoding/ptr_not_equal.ll  | 60 +++++++++++++++++++
 6 files changed, 155 insertions(+)
 create mode 100644 llvm-spirv/test/transcoding/ptr_diff.ll
 create mode 100644 llvm-spirv/test/transcoding/ptr_not_equal.ll

diff --git a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
index 985d28bd9a1a0..f6ed53cba8d15 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVReader.cpp
@@ -2165,6 +2165,28 @@ Value *SPIRVToLLVM::transValueWithoutDecoration(SPIRVValue *BV, Function *F,
     return mapValue(BV, V);
   }
 
+  case OpPtrEqual:
+  case OpPtrNotEqual: {
+    auto *BC = static_cast<SPIRVBinary *>(BV);
+    auto Ops = transValue(BC->getOperands(), F, BB);
+
+    IRBuilder<> Builder(BB);
+    Value *Op1 = Builder.CreatePtrToInt(Ops[0], Type::getInt64Ty(*Context));
+    Value *Op2 = Builder.CreatePtrToInt(Ops[1], Type::getInt64Ty(*Context));
+    CmpInst::Predicate P =
+        OC == OpPtrEqual ? ICmpInst::ICMP_EQ : ICmpInst::ICMP_NE;
+    Value *V = Builder.CreateICmp(P, Op1, Op2);
+    return mapValue(BV, V);
+  }
+
+  case OpPtrDiff: {
+    auto *BC = static_cast<SPIRVBinary *>(BV);
+    auto Ops = transValue(BC->getOperands(), F, BB);
+    IRBuilder<> Builder(BB);
+    Value *V = Builder.CreatePtrDiff(transType(BC->getType()), Ops[0], Ops[1]);
+    return mapValue(BV, V);
+  }
+
   case OpCompositeConstruct: {
     auto *CC = static_cast<SPIRVCompositeConstruct *>(BV);
     auto Constituents = transValue(CC->getOperands(), F, BB);
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
index 7820b70da7ac8..c817a4e120350 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
@@ -669,10 +669,21 @@ class SPIRVBinary : public SPIRVInstTemplateBase {
              "Invalid type for bitwise instruction");
       assert((Op1Ty->getIntegerBitWidth() == Op2Ty->getIntegerBitWidth()) &&
              "Inconsistent BitWidth");
+    } else if (isBinaryPtrOpCode(OpCode)) {
+      assert((Op1Ty->isTypePointer() && Op2Ty->isTypePointer()) &&
+             "Invalid types for PtrEqual, PtrNotEqual, or PtrDiff instruction");
+      assert(static_cast<SPIRVTypePointer *>(Op1Ty)->getElementType() ==
+                 static_cast<SPIRVTypePointer *>(Op2Ty)->getElementType() &&
+             "Invalid types for PtrEqual, PtrNotEqual, or PtrDiff instruction");
     } else {
       assert(0 && "Invalid op code!");
     }
   }
+  VersionNumber getRequiredSPIRVVersion() const override {
+    if (isBinaryPtrOpCode(OpCode))
+      return VersionNumber::SPIRV_1_4;
+    return VersionNumber::SPIRV_1_0;
+  }
 };
 
 template <Op OC>
@@ -707,6 +718,9 @@ _SPIRV_OP(BitwiseAnd)
 _SPIRV_OP(BitwiseOr)
 _SPIRV_OP(BitwiseXor)
 _SPIRV_OP(Dot)
+_SPIRV_OP(PtrEqual)
+_SPIRV_OP(PtrNotEqual)
+_SPIRV_OP(PtrDiff)
 #undef _SPIRV_OP
 
 template <Op TheOpCode> class SPIRVInstNoOperand : public SPIRVInstruction {
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCode.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCode.h
index 0ff2f290a530c..e9a684e745d55 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCode.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCode.h
@@ -74,6 +74,10 @@ inline bool isBinaryOpCode(Op OpCode) {
          OpCode == OpDot || OpCode == OpIAddCarry || OpCode == OpISubBorrow;
 }
 
+inline bool isBinaryPtrOpCode(Op OpCode) {
+  return (unsigned)OpCode >= OpPtrEqual && (unsigned)OpCode <= OpPtrDiff;
+}
+
 inline bool isShiftOpCode(Op OpCode) {
   return (unsigned)OpCode >= OpShiftRightLogical &&
          (unsigned)OpCode <= OpShiftLeftLogical;
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
index ebed32c3476fd..169100efdf50b 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
@@ -329,6 +329,9 @@ _SPIRV_OP(GroupNonUniformBitwiseXor, 361)
 _SPIRV_OP(GroupNonUniformLogicalAnd, 362)
 _SPIRV_OP(GroupNonUniformLogicalOr, 363)
 _SPIRV_OP(GroupNonUniformLogicalXor, 364)
+_SPIRV_OP(PtrEqual, 401)
+_SPIRV_OP(PtrNotEqual, 402)
+_SPIRV_OP(PtrDiff, 403)
 _SPIRV_OP(CopyLogical, 400)
 _SPIRV_OP(GroupNonUniformRotateKHR, 4431)
 _SPIRV_OP(SDotKHR, 4450)
diff --git a/llvm-spirv/test/transcoding/ptr_diff.ll b/llvm-spirv/test/transcoding/ptr_diff.ll
new file mode 100644
index 0000000000000..481064b407597
--- /dev/null
+++ b/llvm-spirv/test/transcoding/ptr_diff.ll
@@ -0,0 +1,52 @@
+; Check support of OpPtrDiff instruction that was added in SPIR-V 1.4
+
+; RUN: llvm-as %s -o %t.bc
+; RUN: not llvm-spirv --spirv-max-version=1.3 %t.bc 2>&1 | FileCheck --check-prefix=CHECK-ERROR %s
+
+; RUN: llvm-spirv %t.bc -o %t.spv
+; RUN: llvm-spirv %t.spv -to-text -o %t.spt
+; RUN: FileCheck < %t.spt %s --check-prefix=CHECK-SPIRV
+; RUN: spirv-val %t.spv
+
+; RUN: llvm-spirv -r %t.spv -o %t.rev.bc
+; RUN: llvm-dis %t.rev.bc
+; RUN: FileCheck < %t.rev.ll %s --check-prefix=CHECK-LLVM
+
+; CHECK-ERROR: RequiresVersion: Cannot fulfill SPIR-V version restriction:
+; CHECK-ERROR-NEXT: SPIR-V version was restricted to at most 1.3 (66304) but a construct from the input requires SPIR-V version 1.4 (66560) or above
+
+; SPIR-V 1.4
+; CHECK-SPIRV: 66560
+; CHECK-SPIRV: TypeInt [[#TypeInt:]] 32 0
+; CHECK-SPIRV: TypeFloat [[#TypeFloat:]] 32
+; CHECK-SPIRV: TypePointer [[#TypePointer:]] [[#]] [[#TypeFloat]]
+
+; CHECK-SPIRV: Variable [[#TypePointer]] [[#Var:]]
+; CHECK-SPIRV: PtrDiff [[#TypeInt]] [[#]] [[#Var]] [[#Var]]
+
+target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+target triple = "spir64-unknown-unknown"
+
+; Function Attrs: nounwind
+define spir_kernel void @test(float %a) local_unnamed_addr #0 {
+entry:
+  %0 = alloca float, align 4
+  store float %a, ptr %0, align 4
+; CHECK-LLVM: %[[#Arg1:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: %[[#Arg2:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: %[[#Sub:]] = sub i64 %[[#Arg1]], %[[#Arg2]]
+; CHECK-LLVM: sdiv exact i64 %[[#Sub]], ptrtoint (ptr getelementptr (i32, ptr null, i32 1) to i64)
+  %1 = call spir_func noundef i32 @_Z15__spirv_PtrDiff(ptr %0, ptr %0)
+  ret void
+}
+
+declare spir_func noundef i32 @_Z15__spirv_PtrDiff(ptr, ptr)
+
+attributes #0 = { convergent nounwind writeonly }
+
+!llvm.module.flags = !{!0}
+!opencl.ocl.version = !{!1}
+!opencl.spir.version = !{!1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 2, i32 0}
diff --git a/llvm-spirv/test/transcoding/ptr_not_equal.ll b/llvm-spirv/test/transcoding/ptr_not_equal.ll
new file mode 100644
index 0000000000000..ffc0f8cc79e81
--- /dev/null
+++ b/llvm-spirv/test/transcoding/ptr_not_equal.ll
@@ -0,0 +1,60 @@
+; Check support of OpPtrEqual and OpPtrNotEqual instructions that were added in SPIR-V 1.4
+
+; RUN: llvm-as %s -o %t.bc
+; RUN: not llvm-spirv --spirv-max-version=1.3 %t.bc 2>&1 | FileCheck --check-prefix=CHECK-ERROR %s
+
+; RUN: llvm-spirv %t.bc -o %t.spv
+; RUN: llvm-spirv %t.spv -to-text -o %t.spt
+; RUN: FileCheck < %t.spt %s --check-prefix=CHECK-SPIRV
+; RUN: spirv-val %t.spv
+
+; RUN: llvm-spirv -r %t.spv -o %t.rev.bc
+; RUN: llvm-dis %t.rev.bc
+; RUN: FileCheck < %t.rev.ll %s --check-prefix=CHECK-LLVM
+
+; CHECK-ERROR: RequiresVersion: Cannot fulfill SPIR-V version restriction:
+; CHECK-ERROR-NEXT: SPIR-V version was restricted to at most 1.3 (66304) but a construct from the input requires SPIR-V version 1.4 (66560) or above
+
+; SPIR-V 1.4
+; CHECK-SPIRV: 66560
+; CHECK-SPIRV: TypeFloat [[#TypeFloat:]] 32
+; CHECK-SPIRV: TypePointer [[#TypePointer:]] [[#]] [[#TypeFloat]]
+; CHECK-SPIRV: TypeBool [[#TypeBool:]]
+
+; CHECK-SPIRV: Variable [[#TypePointer]] [[#Var1:]]
+; CHECK-SPIRV: Variable [[#TypePointer]] [[#Var2:]]
+; CHECK-SPIRV: PtrEqual [[#TypeBool]] [[#]] [[#Var1]] [[#Var2]]
+; CHECK-SPIRV: PtrNotEqual [[#TypeBool]] [[#]] [[#Var1]] [[#Var2]]
+
+target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+target triple = "spir64-unknown-unknown"
+
+; Function Attrs: nounwind
+define spir_kernel void @test(float %a, float %b) local_unnamed_addr #0 {
+entry:
+  %0 = alloca float, align 4
+  %1 = alloca float, align 4
+  store float %a, ptr %0, align 4
+  store float %b, ptr %1, align 4
+; CHECK-LLVM: %[[#Arg1:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: %[[#Arg2:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: icmp eq i64 %[[#Arg1]], %[[#Arg2]]
+  %2 = call spir_func noundef i1 @_Z16__spirv_PtrEqual(ptr %0, ptr %1)
+; CHECK-LLVM: %[[#Arg3:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: %[[#Arg4:]] = ptrtoint ptr %[[#]] to i64
+; CHECK-LLVM: icmp ne i64 %[[#Arg3]], %[[#Arg4]]
+  %3 = call spir_func noundef i1 @_Z19__spirv_PtrNotEqual(ptr %0, ptr %1)
+  ret void
+}
+
+declare spir_func noundef i1 @_Z16__spirv_PtrEqual(ptr, ptr)
+declare spir_func noundef i1 @_Z19__spirv_PtrNotEqual(ptr, ptr)
+
+attributes #0 = { convergent nounwind writeonly }
+
+!llvm.module.flags = !{!0}
+!opencl.ocl.version = !{!1}
+!opencl.spir.version = !{!1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 2, i32 0}

From 0af095e4d4941f5343a7841e041d86cc58aa8824 Mon Sep 17 00:00:00 2001
From: Dmitry Sidorov <dmitry.sidorov@intel.com>
Date: Tue, 16 Apr 2024 11:52:38 +0200
Subject: [PATCH 872/886] Remove tbaa metadata from the test (#2503)

We don't need it here

Signed-off-by: Sidorov, Dmitry <dmitry.sidorov@intel.com>

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/7c85c6bf7753274
---
 .../INTEL/SPV_INTEL_fpga_reg/IntelFPGAReg.ll  | 85 ++++++++-----------
 1 file changed, 37 insertions(+), 48 deletions(-)

diff --git a/llvm-spirv/test/extensions/INTEL/SPV_INTEL_fpga_reg/IntelFPGAReg.ll b/llvm-spirv/test/extensions/INTEL/SPV_INTEL_fpga_reg/IntelFPGAReg.ll
index edf6c9ff9ff57..367608b876b6c 100644
--- a/llvm-spirv/test/extensions/INTEL/SPV_INTEL_fpga_reg/IntelFPGAReg.ll
+++ b/llvm-spirv/test/extensions/INTEL/SPV_INTEL_fpga_reg/IntelFPGAReg.ll
@@ -122,7 +122,7 @@ declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
 define internal spir_func void @"_ZZ4mainENK3$_0clEv"(ptr addrspace(4) %this) #2 align 2 {
 entry:
   %this.addr = alloca ptr addrspace(4), align 8
-  store ptr addrspace(4) %this, ptr %this.addr, align 8, !tbaa !5
+  store ptr addrspace(4) %this, ptr %this.addr, align 8
   %this1 = load ptr addrspace(4), ptr %this.addr, align 8
   call spir_func void @_Z3foov()
   ret void
@@ -161,30 +161,30 @@ entry:
   %ap = alloca ptr addrspace(4), align 8
   %bp = alloca ptr addrspace(4), align 8
   call void @llvm.lifetime.start.p0(i64 4, ptr %a) #4
-  store i32 123, ptr %a, align 4, !tbaa !9
+  store i32 123, ptr %a, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %myA) #4
-  store i32 321, ptr %myA, align 4, !tbaa !9
+  store i32 321, ptr %myA, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %b) #4
-  %0 = load i32, ptr %a, align 4, !tbaa !9
+  %0 = load i32, ptr %a, align 4
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR]]
   %1 = call i32 @llvm.annotation.i32.p1(i32 %0, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 35)
-  store i32 %1, ptr %b, align 4, !tbaa !9
+  store i32 %1, ptr %b, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %myB) #4
-  %2 = load i32, ptr %myA, align 4, !tbaa !9
+  %2 = load i32, ptr %myA, align 4
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR1]]
   %3 = call i32 @llvm.annotation.i32.p1(i32 %2, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 39)
-  store i32 %3, ptr %myB, align 4, !tbaa !9
+  store i32 %3, ptr %myB, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %c) #4
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR2]]
   %4 = call i32 @llvm.annotation.i32.p1(i32 1073741824, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 43)
   %5 = bitcast i32 %4 to float
   %conv = fptosi float %5 to i32
-  store i32 %conv, ptr %c, align 4, !tbaa !9
+  store i32 %conv, ptr %c, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %d) #4
-  %6 = load i32, ptr %b, align 4, !tbaa !9
+  %6 = load i32, ptr %b, align 4
   %add = add nsw i32 %6, 12
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR3]]
@@ -192,10 +192,10 @@ entry:
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR4]]
   %8 = call i32 @llvm.annotation.i32.p1(i32 %7, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 48)
-  store i32 %8, ptr %d, align 4, !tbaa !9
+  store i32 %8, ptr %d, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %e) #4
-  %9 = load i32, ptr %a, align 4, !tbaa !9
-  %10 = load i32, ptr %b, align 4, !tbaa !9
+  %9 = load i32, ptr %a, align 4
+  %10 = load i32, ptr %b, align 4
   %add1 = add nsw i32 %9, %10
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR5]]
@@ -203,31 +203,31 @@ entry:
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR6]]
   %12 = call i32 @llvm.annotation.i32.p1(i32 %11, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 54)
-  store i32 %12, ptr %e, align 4, !tbaa !9
+  store i32 %12, ptr %e, align 4
   call void @llvm.lifetime.start.p0(i64 4, ptr %f) #4
-  %13 = load i32, ptr %a, align 4, !tbaa !9
+  %13 = load i32, ptr %a, align 4
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT32]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i32 @llvm.annotation.i32.p0(i32 {{[%a-z0-9]+}}, ptr @[[STR7]]
   %14 = call i32 @llvm.annotation.i32.p1(i32 %13, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 62)
-  store i32 %14, ptr %f, align 4, !tbaa !9
+  store i32 %14, ptr %f, align 4
   call void @llvm.lifetime.start.p0(i64 8, ptr %i) #4
   call void @llvm.memcpy.p0.p1.i64(ptr align 4 %i, ptr addrspace(1) align 4 @__const._Z3foov.i, i64 8, i1 false)
   call void @llvm.lifetime.start.p0(i64 8, ptr %i2) #4
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %i2, ptr align 4 %i, i64 8, i1 false), !tbaa.struct !11
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %i2, ptr align 4 %i, i64 8, i1 false)
   call void @llvm.lifetime.start.p0(i64 8, ptr %ii) #4
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp, ptr align 4 %i, i64 8, i1 false), !tbaa.struct !11
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp, ptr align 4 %i, i64 8, i1 false)
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_PTR]]   {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call ptr @llvm.ptr.annotation.p0.p0(ptr %[[CAST1:[a-z0-9]+]], ptr @[[STR8]]
   %15 = call ptr @llvm.ptr.annotation.p0.p1(ptr %agg-temp, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 69, ptr addrspace(1) null)
   call void @llvm.memcpy.p0.p0.i64(ptr align 4 %ii, ptr align 4 %15, i64 8, i1 false)
   call void @llvm.lifetime.start.p0(i64 8, ptr %iii) #4
   call void @llvm.lifetime.start.p0(i64 8, ptr %ref.tmp) #4
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp2, ptr align 4 %ii, i64 8, i1 false), !tbaa.struct !11
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp2, ptr align 4 %ii, i64 8, i1 false)
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_PTR]]   {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call ptr @llvm.ptr.annotation.p0.p0(ptr %[[CAST2:[a-z0-9]+]], ptr @[[STR9]]
   %16 = call ptr @llvm.ptr.annotation.p0.p1(ptr %agg-temp2, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 80, ptr addrspace(1) null)
   call void @llvm.memcpy.p0.p0.i64(ptr align 4 %ref.tmp, ptr align 4 %16, i64 8, i1 false)
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %iii, ptr align 4 %ref.tmp, i64 8, i1 false), !tbaa.struct !11
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %iii, ptr align 4 %ref.tmp, i64 8, i1 false)
   call void @llvm.lifetime.end.p0(i64 8, ptr %ref.tmp) #4
   call void @llvm.lifetime.start.p0(i64 8, ptr %iiii) #4
   %17 = ptrtoint ptr %iii to i64
@@ -236,18 +236,18 @@ entry:
   %18 = call i64 @llvm.annotation.i64(i64 %17, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 94)
   %19 = inttoptr i64 %18 to ptr
   %20 = addrspacecast ptr %19 to ptr addrspace(4)
-  store ptr addrspace(4) %20, ptr %iiii, align 8, !tbaa !5
+  store ptr addrspace(4) %20, ptr %iiii, align 8
   call void @llvm.lifetime.start.p0(i64 4, ptr %u1) #4
   call void @llvm.memcpy.p0.p1.i64(ptr align 4 %u1, ptr addrspace(1) align 4 @__const._Z3foov.u1, i64 4, i1 false)
   call void @llvm.lifetime.start.p0(i64 4, ptr %u2) #4
   call void @llvm.lifetime.start.p0(i64 8, ptr %u3) #4
   call void @llvm.lifetime.start.p0(i64 4, ptr %ref.tmp3) #4
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp4, ptr align 4 %u1, i64 4, i1 false), !tbaa.struct !14
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp4, ptr align 4 %u1, i64 4, i1 false)
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_PTR]]   {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call ptr @llvm.ptr.annotation.p0.p0(ptr %[[CAST4:[a-z0-9]+]], ptr @[[STR11]]
   %21 = call ptr @llvm.ptr.annotation.p0.p1(ptr %agg-temp4, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 103, ptr addrspace(1) null)
   call void @llvm.memcpy.p0.p0.i64(ptr align 4 %ref.tmp3, ptr align 4 %21, i64 8, i1 false)
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %u2, ptr align 4 %ref.tmp3, i64 4, i1 false), !tbaa.struct !14
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %u2, ptr align 4 %ref.tmp3, i64 4, i1 false)
   call void @llvm.lifetime.end.p0(i64 4, ptr %ref.tmp3) #4
   %22 = ptrtoint ptr %u2 to i64
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT64]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
@@ -255,27 +255,27 @@ entry:
   %23 = call i64 @llvm.annotation.i64(i64 %22, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 117)
   %24 = inttoptr i64 %23 to ptr
   %25 = addrspacecast ptr %24 to ptr addrspace(4)
-  store ptr addrspace(4) %25, ptr %u3, align 8, !tbaa !5
+  store ptr addrspace(4) %25, ptr %u3, align 8
   call void @llvm.lifetime.start.p0(i64 4, ptr %ca) #4
   %26 = addrspacecast ptr %ca to ptr addrspace(4)
   call spir_func void @_ZN1AC1Ei(ptr addrspace(4) %26, i32 213)
   call void @llvm.lifetime.start.p0(i64 4, ptr %cb) #4
-  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp5, ptr align 4 %ca, i64 4, i1 false), !tbaa.struct !16
+  call void @llvm.memcpy.p0.p0.i64(ptr align 4 %agg-temp5, ptr align 4 %ca, i64 4, i1 false)
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_PTR]]   {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call ptr @llvm.ptr.annotation.p0.p0(ptr %[[CAST5:[a-z0-9]+]], ptr @[[STR13]]
   %27 = call ptr @llvm.ptr.annotation.p0.p1(ptr %agg-temp5, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 125, ptr addrspace(1) null)
   call void @llvm.memcpy.p0.p0.i64(ptr align 4 %cb, ptr align 4 %27, i64 8, i1 false)
   call void @llvm.lifetime.start.p0(i64 8, ptr %ap) #4
   %28 = addrspacecast ptr %a to ptr addrspace(4)
-  store ptr addrspace(4) %28, ptr %ap, align 8, !tbaa !5
+  store ptr addrspace(4) %28, ptr %ap, align 8
   call void @llvm.lifetime.start.p0(i64 8, ptr %bp) #4
-  %29 = load ptr addrspace(4), ptr %ap, align 8, !tbaa !5
+  %29 = load ptr addrspace(4), ptr %ap, align 8
   %30 = ptrtoint ptr addrspace(4) %29 to i64
   ; CHECK-SPIRV: FPGARegINTEL [[TYPE_INT64]] {{[0-9]+}} {{[0-9]+}}{{ *$}}
   ; CHECK-LLVM-DAG: %{{[0-9]+}} = call i64 @llvm.annotation.i64.p0(i64 {{[%a-z0-9]+}}, ptr @[[STR14]]
   %31 = call i64 @llvm.annotation.i64(i64 %30, ptr addrspace(1) @.str, ptr addrspace(1) @.str.1, i32 137)
   %32 = inttoptr i64 %31 to ptr addrspace(4)
-  store ptr addrspace(4) %32, ptr %bp, align 8, !tbaa !5
+  store ptr addrspace(4) %32, ptr %bp, align 8
   call void @llvm.lifetime.end.p0(i64 8, ptr %bp) #4
   call void @llvm.lifetime.end.p0(i64 8, ptr %ap) #4
   call void @llvm.lifetime.end.p0(i64 4, ptr %cb) #4
@@ -322,10 +322,10 @@ define linkonce_odr spir_func void @_ZN1AC1Ei(ptr addrspace(4) %this, i32 %a) un
 entry:
   %this.addr = alloca ptr addrspace(4), align 8
   %a.addr = alloca i32, align 4
-  store ptr addrspace(4) %this, ptr %this.addr, align 8, !tbaa !5
-  store i32 %a, ptr %a.addr, align 4, !tbaa !9
+  store ptr addrspace(4) %this, ptr %this.addr, align 8
+  store i32 %a, ptr %a.addr, align 4
   %this1 = load ptr addrspace(4), ptr %this.addr, align 8
-  %0 = load i32, ptr %a.addr, align 4, !tbaa !9
+  %0 = load i32, ptr %a.addr, align 4
   call spir_func void @_ZN1AC2Ei(ptr addrspace(4) %this1, i32 %0)
   ret void
 }
@@ -335,11 +335,11 @@ define linkonce_odr spir_func void @_ZN1AC2Ei(ptr addrspace(4) %this, i32 %a) un
 entry:
   %this.addr = alloca ptr addrspace(4), align 8
   %a.addr = alloca i32, align 4
-  store ptr addrspace(4) %this, ptr %this.addr, align 8, !tbaa !5
-  store i32 %a, ptr %a.addr, align 4, !tbaa !9
+  store ptr addrspace(4) %this, ptr %this.addr, align 8
+  store i32 %a, ptr %a.addr, align 4
   %this1 = load ptr addrspace(4), ptr %this.addr, align 8
-  %0 = load i32, ptr %a.addr, align 4, !tbaa !9
-  store i32 %0, ptr addrspace(4) %this1, align 4, !tbaa !17
+  %0 = load i32, ptr %a.addr, align 4
+  store i32 %0, ptr addrspace(4) %this1, align 4
   ret void
 }
 
@@ -359,17 +359,6 @@ attributes #4 = { nounwind }
 !2 = !{i32 4, i32 100000}
 !3 = !{!"clang version 9.0.0"}
 !4 = !{}
-!5 = !{!6, !6, i64 0}
-!6 = !{!"any pointer", !7, i64 0}
-!7 = !{!"omnipotent char", !8, i64 0}
-!8 = !{!"Simple C++ TBAA"}
-!9 = !{!10, !10, i64 0}
-!10 = !{!"int", !7, i64 0}
-!11 = !{i64 0, i64 4, !9, i64 4, i64 4, !12}
-!12 = !{!13, !13, i64 0}
-!13 = !{!"float", !7, i64 0}
-!14 = !{i64 0, i64 4, !15}
-!15 = !{!7, !7, i64 0}
-!16 = !{i64 0, i64 4, !9}
-!17 = !{!18, !10, i64 0}
-!18 = !{!"_ZTS1A", !10, i64 0}
+!5 = !{!"any pointer", !6, i64 0}
+!6 = !{!"omnipotent char", !7, i64 0}
+!7 = !{!"Simple C++ TBAA"}

From b13dd9346fa572de55c5bd7067ee96d8320651c0 Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Fri, 12 Apr 2024 19:46:44 +0200
Subject: [PATCH 873/886] Add CopyLogical instruction (#2484)

There is no mapping to LLVM instructions, so it can be used only via SPIR-V friendly translation.

This addresses p1. of #2460

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/8518a6f72fd586c
---
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
index 169100efdf50b..588f95cc88866 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
@@ -329,10 +329,10 @@ _SPIRV_OP(GroupNonUniformBitwiseXor, 361)
 _SPIRV_OP(GroupNonUniformLogicalAnd, 362)
 _SPIRV_OP(GroupNonUniformLogicalOr, 363)
 _SPIRV_OP(GroupNonUniformLogicalXor, 364)
+_SPIRV_OP(CopyLogical, 400)
 _SPIRV_OP(PtrEqual, 401)
 _SPIRV_OP(PtrNotEqual, 402)
 _SPIRV_OP(PtrDiff, 403)
-_SPIRV_OP(CopyLogical, 400)
 _SPIRV_OP(GroupNonUniformRotateKHR, 4431)
 _SPIRV_OP(SDotKHR, 4450)
 _SPIRV_OP(UDotKHR, 4451)

From 93aaad61be38c92cf1377e7a6e0c8d81e7303fbd Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Mon, 15 Apr 2024 14:56:35 +0200
Subject: [PATCH 874/886] Support OpPtrEqual, OpPtrNotEqual and OpPtrDiff to
 compare pointers (#2482)

This addresses p.2 and p.3 of #2460

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/4b4508438a6dfe5
---
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
index 588f95cc88866..d0c801d1b51c3 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVOpCodeEnum.h
@@ -329,6 +329,9 @@ _SPIRV_OP(GroupNonUniformBitwiseXor, 361)
 _SPIRV_OP(GroupNonUniformLogicalAnd, 362)
 _SPIRV_OP(GroupNonUniformLogicalOr, 363)
 _SPIRV_OP(GroupNonUniformLogicalXor, 364)
+_SPIRV_OP(PtrEqual, 401)
+_SPIRV_OP(PtrNotEqual, 402)
+_SPIRV_OP(PtrDiff, 403)
 _SPIRV_OP(CopyLogical, 400)
 _SPIRV_OP(PtrEqual, 401)
 _SPIRV_OP(PtrNotEqual, 402)

From 01ceb0c3a5217eebc261fb3f397ce9b6d3e216f5 Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Tue, 16 Apr 2024 12:35:08 +0200
Subject: [PATCH 875/886] Deprecate OpLessOrGreater in favor of OpFOrdNotEqual
 (#2491)

This is a follow-up to 08884df

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/89b19c46b5aaa2c
---
 llvm-spirv/lib/SPIRV/SPIRVWriter.cpp          |  9 ++++
 .../test/transcoding/OpLessOrGreater.ll       | 45 +++++++++++++++++++
 2 files changed, 54 insertions(+)
 create mode 100644 llvm-spirv/test/transcoding/OpLessOrGreater.ll

diff --git a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
index 12e798a300c40..9148b2ec74b92 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
@@ -6160,6 +6160,15 @@ LLVMToSPIRVBase::transBuiltinToInstWithoutDecoration(Op OC, CallInst *CI,
   // Moreover, OpAtomicCompareExchangeWeak has been deprecated.
   if (OC == OpAtomicCompareExchangeWeak)
     OC = OpAtomicCompareExchange;
+
+  // We should do this replacement only for SPIR-V 1.5, as OpLessOrGreater is
+  // deprecated there. However we do such replacement for the usual pipeline
+  // (not via SPIR-V friendly calls) without minding the version, so we can do
+  // such thing here as well.
+  if (OC == OpLessOrGreater &&
+      BM->isAllowedToUseVersion(VersionNumber::SPIRV_1_5))
+    OC = OpFOrdNotEqual;
+
   if (isGroupOpCode(OC))
     BM->addCapability(CapabilityGroups);
   switch (static_cast<size_t>(OC)) {
diff --git a/llvm-spirv/test/transcoding/OpLessOrGreater.ll b/llvm-spirv/test/transcoding/OpLessOrGreater.ll
new file mode 100644
index 0000000000000..fc885c68a1f02
--- /dev/null
+++ b/llvm-spirv/test/transcoding/OpLessOrGreater.ll
@@ -0,0 +1,45 @@
+; RUN: llvm-as %s -o %t.bc
+; RUN: llvm-spirv %t.bc -o %t.spv
+; RUN: llvm-spirv %t.spv -to-text -o %t.spt
+; RUN: FileCheck < %t.spt %s --check-prefix=CHECK-SPIRV
+; RUN: spirv-val %t.spv
+
+; RUN: llvm-spirv -r %t.spv -o %t.rev.bc
+; RUN: llvm-dis %t.rev.bc
+; RUN: FileCheck < %t.rev.ll %s --check-prefix=CHECK-LLVM
+
+; SPIR-V 1.5
+; CHECK-SPIRV: 66816
+
+; CHECK-SPIRV-NOT: LessOrGreater
+; CHECK-SPIRV: FOrdNotEqual
+
+target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+target triple = "spir64-unknown-unknown"
+
+; Function Attrs: nounwind
+define spir_kernel void @test(float %a) local_unnamed_addr #0 {
+entry:
+; CHECK-LLVM: fcmp one float %a, %a
+  %call = tail call spir_func i1 @_Z21__spirv_LessOrGreater(float %a, float %a)
+  ret void
+}
+
+; This is needed to check that 1.5 is enabled
+define dso_local spir_kernel void @test2(i16 noundef signext %a, i32 noundef %id) local_unnamed_addr #0  {
+entry:
+  %call = tail call spir_func signext i16 @_Z31sub_group_non_uniform_broadcastsj(i16 noundef signext %a, i32 noundef %id)
+  ret void
+}
+
+declare spir_func noundef i1 @_Z21__spirv_LessOrGreater(float, float)
+declare spir_func signext i16 @_Z31sub_group_non_uniform_broadcastsj(i16 noundef signext, i32 noundef)
+
+attributes #0 = { convergent nounwind writeonly }
+
+!llvm.module.flags = !{!0}
+!opencl.ocl.version = !{!1}
+!opencl.spir.version = !{!1}
+
+!0 = !{i32 1, !"wchar_size", i32 4}
+!1 = !{i32 2, i32 0}

From c857425c95c3248b95de25f0ca11a338e7042d5c Mon Sep 17 00:00:00 2001
From: Arvind Sudarsanam <arvind.sudarsanam@intel.com>
Date: Tue, 16 Apr 2024 09:05:12 -0400
Subject: [PATCH 876/886] Regularize LLVM code to remove use of non standard
 integer types in fptoui and fptosi intrinsics (#2500)

* Regularize LLVM code to remove use of non standard integer types in fptoui and fptosi intrinsics

Signed-off-by: Sudarsanam, Arvind <arvind.sudarsanam@intel.com>

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/14301c295d3dc8e
---
 llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.cpp  | 99 +++++++++++++++++++
 llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.h    |  6 ++
 .../fp_to_arbitrary_size_int_intrinsic.ll     | 57 +++++++++++
 3 files changed, 162 insertions(+)
 create mode 100644 llvm-spirv/test/llvm-intrinsics/fp_to_arbitrary_size_int_intrinsic.ll

diff --git a/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.cpp b/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.cpp
index 81be03049a829..c29e5521baac9 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.cpp
@@ -321,6 +321,104 @@ void SPIRVRegularizeLLVMBase::expandSYCLTypeUsing(Module *M) {
     expandVIDWithSYCLTypeByValComp(F);
 }
 
+// In this function, we handle two conversion operations
+// 1. fptoui.sat.iX.fY (X is not 8,16,32,64; Y is 32 or 64)
+// 2. fptosi.sat.iX.fY (X is not 8,16,32,64; Y is 32 or 64)
+// Such non-standard integer types cannot be handled in SPIR-V. Hence, they
+// will be promoted to
+// 1. fptoui.sat.i64.fY (Y is 32 or 64)
+// 2. fptosi.sat.i64.fY (Y is 32 or 64)
+// However, LLVM documentation requires the following rules to be obeyed.
+// Rule 1: If the argument is any NaN, zero is returned.
+// Rule 2: If the argument is smaller than the smallest representable
+// (un)signed integer of the result type, the smallest representable
+// (un)signed integer is returned.
+// Rule 3: If the argument is larger than the largest representable (un)signed
+// integer of the result type, the largest representable (un)signed integer is
+// returned.
+// Rule 4: Otherwise, the result of rounding the argument towards zero is
+// returned.
+// Rules 1 & 4 are preserved when promoting iX to i64. For preserving Rule 2
+// and Rule 3, we saturate the result of the promoted instruction based on
+// original integer type (iX)
+// Example:
+// Input:
+// %0 = call i2 @llvm.fptosi.sat.i2.f32(float %input)
+// %1 = sext i32 %0
+// Output:
+// %0 = call i32 @_Z17convert_long_satf(float %input)
+// %1 = icmp sge i32 %0, 1 <Largest 2-bit signed integer>
+// %2 = icmp sle i32 %0, -2 <Smallest 2-bit signed integer>
+// %3 = select i1 %1, i32 1, i32 %0
+// %4 = select i1 %2, i32 -2, i32 %3
+// Replace uses of %1 in Input with %4 in Output
+void SPIRVRegularizeLLVMBase::cleanupConversionToNonStdIntegers(Module *M) {
+  for (auto FI = M->begin(), FE = M->end(); FI != FE;) {
+    Function *F = &(*FI++);
+    std::vector<Instruction *> ToErase;
+    auto IID = F->getIntrinsicID();
+    if (IID != Intrinsic::fptosi_sat && IID != Intrinsic::fptoui_sat)
+      continue;
+    for (auto *I : F->users()) {
+      if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
+        // TODO: Vector type not supported yet.
+        if (isa<VectorType>(II->getType()))
+          continue;
+        auto IID = II->getIntrinsicID();
+        auto IntBitWidth = II->getType()->getScalarSizeInBits();
+        if (IntBitWidth == 8 || IntBitWidth == 16 || IntBitWidth == 32 ||
+            IntBitWidth == 64)
+          continue;
+        if (IID == Intrinsic::fptosi_sat) {
+          // Identify sext (user of II). Make sure that's the only use of II.
+          auto *User = II->getUniqueUndroppableUser();
+          if (!User || !isa<SExtInst>(User))
+            continue;
+          auto *SExtI = dyn_cast<SExtInst>(User);
+          auto *NewIType = SExtI->getType();
+          IRBuilder<> IRB(II);
+          auto *NewII = IRB.CreateIntrinsic(
+              IID, {NewIType, II->getOperand(0)->getType()}, II->getOperand(0));
+          Constant *MaxVal = ConstantInt::get(
+              NewIType, APInt::getSignedMaxValue(IntBitWidth).getSExtValue());
+          Constant *MinVal = ConstantInt::get(
+              NewIType, APInt::getSignedMinValue(IntBitWidth).getSExtValue());
+          auto *GTMax = IRB.CreateICmp(CmpInst::ICMP_SGE, NewII, MaxVal);
+          auto *LTMin = IRB.CreateICmp(CmpInst::ICMP_SLE, NewII, MinVal);
+          auto *SatMax = IRB.CreateSelect(GTMax, MaxVal, NewII);
+          auto *SatMin = IRB.CreateSelect(LTMin, MinVal, SatMax);
+          SExtI->replaceAllUsesWith(SatMin);
+          ToErase.push_back(SExtI);
+          ToErase.push_back(II);
+        }
+        if (IID == Intrinsic::fptoui_sat) {
+          // Identify zext (user of II). Make sure that's the only use of II.
+          auto *User = II->getUniqueUndroppableUser();
+          if (!User || !isa<ZExtInst>(User))
+            continue;
+          auto *ZExtI = dyn_cast<ZExtInst>(User);
+          auto *NewIType = ZExtI->getType();
+          IRBuilder<> IRB(II);
+          auto *NewII = IRB.CreateIntrinsic(
+              IID, {NewIType, II->getOperand(0)->getType()}, II->getOperand(0));
+          Constant *MaxVal = ConstantInt::get(
+              NewIType, APInt::getMaxValue(IntBitWidth).getZExtValue());
+          auto *GTMax = IRB.CreateICmp(CmpInst::ICMP_UGE, NewII, MaxVal);
+          auto *SatMax = IRB.CreateSelect(GTMax, MaxVal, NewII);
+          ZExtI->replaceAllUsesWith(SatMax);
+          ToErase.push_back(ZExtI);
+          ToErase.push_back(II);
+        }
+      }
+    }
+    for (Instruction *V : ToErase) {
+      assert(V->user_empty());
+      V->dropAllReferences();
+      V->eraseFromParent();
+    }
+  }
+}
+
 bool SPIRVRegularizeLLVMBase::runRegularizeLLVM(Module &Module) {
   M = &Module;
   Ctx = &M->getContext();
@@ -464,6 +562,7 @@ void regularizeWithOverflowInstrinsics(StringRef MangledName, CallInst *Call,
 bool SPIRVRegularizeLLVMBase::regularize() {
   eraseUselessFunctions(M);
   expandSYCLTypeUsing(M);
+  cleanupConversionToNonStdIntegers(M);
 
   for (auto &GV : M->globals()) {
     SPIRVBuiltinVariableKind Kind;
diff --git a/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.h b/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.h
index ed04de61bced1..823bd612423b8 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.h
+++ b/llvm-spirv/lib/SPIRV/SPIRVRegularizeLLVM.h
@@ -91,6 +91,12 @@ class SPIRVRegularizeLLVMBase {
   void expandVEDWithSYCLTypeSRetArg(llvm::Function *F);
   void expandVIDWithSYCLTypeByValComp(llvm::Function *F);
 
+  // It is possible that incoming LLVM IR conversion instructions convert
+  // floating point to non-standard integer types. Such types are not supported
+  // in SPIR-V. This function cleans up such code and removes occurence of
+  // non-standard integer types.
+  void cleanupConversionToNonStdIntegers(llvm::Module *M);
+
   // According to the specification, the operands of a shift instruction must be
   // a scalar/vector of integer. When LLVM-IR contains a shift instruction with
   // i1 operands, they are treated as a bool. We need to extend them to i32 to
diff --git a/llvm-spirv/test/llvm-intrinsics/fp_to_arbitrary_size_int_intrinsic.ll b/llvm-spirv/test/llvm-intrinsics/fp_to_arbitrary_size_int_intrinsic.ll
new file mode 100644
index 0000000000000..5449a61cfcb6a
--- /dev/null
+++ b/llvm-spirv/test/llvm-intrinsics/fp_to_arbitrary_size_int_intrinsic.ll
@@ -0,0 +1,57 @@
+;; Ensure @llvm.fptosi.sat.* and @llvm.fptoui.sat.* intrinsics are translated
+
+; RUN: llvm-as %s -o %t.bc
+; RUN: llvm-spirv %t.bc -spirv-text -o - | FileCheck %s --check-prefix=CHECK-SPIRV
+; RUN: llvm-spirv %t.bc -o %t.spv
+; RUN: spirv-val %t.spv
+; RUN: llvm-spirv -r %t.spv -o %t.rev.bc
+; RUN: llvm-dis < %t.rev.bc | FileCheck %s --check-prefix=CHECK-LLVM
+
+target datalayout = "e-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-n8:16:32:64"
+target triple = "spir64-unknown-unknown"
+
+; CHECK-SPIRV-DAG: Capability Kernel
+; CHECK-SPIRV-DAG: Decorate [[SAT1:[0-9]+]] SaturatedConversion
+; CHECK-SPIRV-DAG: Decorate [[SAT2:[0-9]+]] SaturatedConversion
+
+; CHECK-SPIRV-DAG: TypeInt [[INT64TY:[0-9]+]] 64
+; CHECK-SPIRV-DAG: TypeBool [[BOOLTY:[0-9]+]]
+; CHECK-SPIRV-DAG: Constant [[INT64TY]] [[I2SMAX:[0-9]+]] 1
+; CHECK-SPIRV-DAG: Constant [[INT64TY]] [[I2SMIN:[0-9]+]] 4294967294 
+; CHECK-SPIRV-DAG: ConvertFToS [[INT64TY]] [[SAT1]]
+; CHECK-SPIRV-DAG: SGreaterThanEqual [[BOOLTY]] [[SGERES:[0-9]+]] [[SAT1]] [[I2SMAX]]
+; CHECK-SPIRV-DAG: SLessThanEqual [[BOOLTY]] [[SLERES:[0-9]+]] [[SAT1]] [[I2SMIN]]
+; CHECK-SPIRV-DAG: Select [[INT64TY]] [[SELRES1:[0-9]+]] [[SGERES]] [[I2SMAX]] [[SAT1]]
+; CHECK-SPIRV-DAG: Select [[INT64TY]] [[SELRES2:[0-9]+]] [[SLERES]] [[I2SMIN]] [[SELRES1]]
+
+; CHECK-LLVM-DAG: define spir_kernel
+; CHECK-LLVM-DAG: %[[R1:[0-9]+]] = {{.*}} i64 {{.*}}convert_long_satf(float %input)
+; CHECK-LLVM-DAG: %[[R2:[0-9]+]] = icmp sge i64 %[[R1]], 1
+; CHECK-LLVM-DAG: %[[R3:[0-9]+]] = icmp sle i64 %[[R1]], -2
+; CHECK-LLVM-DAG: %[[R4:[0-9]+]] = select i1 %[[R2]], i64 1, i64 %[[R1]]
+; CHECK-LLVM-DAG: %[[R5:[0-9]+]] = select i1 %[[R3]], i64 -2, i64 %[[R4]]
+
+define spir_kernel void @testfunction_float_to_signed_i2(float %input) {
+entry:
+   %0 = call i2 @llvm.fptosi.sat.i2.f32(float %input)
+   %1 = sext i2 %0 to i64
+   ret void
+}
+declare i2 @llvm.fptosi.sat.i2.f32(float)
+
+; CHECK-SPIRV-DAG: Constant [[INT64TY]] [[I2UMAX:[0-9]+]] 3 
+; CHECK-SPIRV-DAG: ConvertFToU [[INT64TY]] [[SAT2]]
+; CHECK-SPIRV-DAG: UGreaterThanEqual [[BOOLTY]] [[UGERES:[0-9]+]] [[SAT2]] [[I2UMAX]]
+; CHECK-SPIRV-DAG: Select [[INT64TY]] [[SELRES1U:[0-9]+]] [[UGERES]] [[I2UMAX]] [[SAT2]]
+; CHECK-LLVM-DAG: define spir_kernel
+; CHECK-LLVM-DAG: %[[R1:[0-9]+]] = {{.*}} i64 {{.*}}convert_ulong_satf(float %input)
+; CHECK-LLVM-DAG: %[[R2:[0-9]+]] = icmp uge i64 %[[R1]], 3
+; CHECK-LLVM-DAG: %[[R3:[0-9]+]] = select i1 %[[R2]], i64 3, i64 %[[R1]]
+
+define spir_kernel void @testfunction_float_to_unsigned_i2(float %input) {
+entry:
+   %0 = call i2 @llvm.fptoui.sat.i2.f32(float %input)
+   %1 = zext i2 %0 to i64
+   ret void
+}
+declare i2 @llvm.fptoui.sat.i2.f32(float)

From 53f348a68830da2a61de29d78e994f830aa5c815 Mon Sep 17 00:00:00 2001
From: Marcos Maronas <maarquitos14@users.noreply.github.com>
Date: Tue, 16 Apr 2024 07:00:41 -0700
Subject: [PATCH 877/886] Don't call getName() on literal structs. (#2499)

Signed-off-by: Marcos Maronas <marcos.maronas@intel.com>

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/0d83f2233e1f7f6
---
 llvm-spirv/lib/SPIRV/SPIRVUtil.cpp              |  7 +++++--
 ..._spirv_for_opencl_cpp_with_unnamed_struct.ll | 17 +++++++++++++++++
 2 files changed, 22 insertions(+), 2 deletions(-)
 create mode 100644 llvm-spirv/test/ocl_to_spirv_for_opencl_cpp_with_unnamed_struct.ll

diff --git a/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp b/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
index cafb738adf214..864ff49483ca0 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVUtil.cpp
@@ -806,8 +806,11 @@ bool getParameterTypes(Function *F, SmallVectorImpl<Type *> &ArgTys,
       HasSret = true;
       unsigned AS = Arg.getType()->getPointerAddressSpace();
       if (auto *STy = dyn_cast<StructType>(Ty))
-        ArgTys.push_back(
-            TypedPointerType::get(GetStructType(STy->getName()), AS));
+        if (STy->hasName())
+          ArgTys.push_back(
+              TypedPointerType::get(GetStructType(STy->getName()), AS));
+        else
+          ArgTys.push_back(TypedPointerType::get(STy, AS));
       else
         ArgTys.push_back(TypedPointerType::get(Ty, AS));
     } else {
diff --git a/llvm-spirv/test/ocl_to_spirv_for_opencl_cpp_with_unnamed_struct.ll b/llvm-spirv/test/ocl_to_spirv_for_opencl_cpp_with_unnamed_struct.ll
new file mode 100644
index 0000000000000..cbb4bb0495c42
--- /dev/null
+++ b/llvm-spirv/test/ocl_to_spirv_for_opencl_cpp_with_unnamed_struct.ll
@@ -0,0 +1,17 @@
+;; This test checks if functions in LLVM IR generated from OpenCL_CPP sources
+;; including unnamed structs are correctly translated to SPIR-V.
+
+; RUN: llvm-as %s -o %t.bc
+; RUN: llvm-spirv %t.bc -o %t.spv
+; RUN: spirv-val %t.spv
+
+target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024"
+target triple = "spir-unknown-unknown"
+
+%opencl.event_t = type opaque
+
+declare spir_func void @_ZNKSt7complexIfE5__repEv(ptr addrspace(4) dead_on_unwind noalias writable sret({ float, float }) align 4)
+
+!spirv.Source = !{!0}
+
+!0 = !{i32 4, i32 100000}

From 7faec01e13b4eaceb364a2fea96a02f9e3d65a11 Mon Sep 17 00:00:00 2001
From: bwlodarcz <bertrand.wlodarczyk@intel.com>
Date: Tue, 16 Apr 2024 16:01:29 +0200
Subject: [PATCH 878/886] Fix of ParentIdx mismatch for DebugTypeInheritance in
 SPIRV.debug.h (#2490)

Fixed ParentIdx was mismatched for DebugTypeInheritance type in context of NonSemantic.Shader.DebugInfo.100. That lead to heap buffer overflow in SPIRVExtInst::getExtOp getter when instruction was incorrectly casted
to SPIRVExtInst as parent of the culprit instruction in SPIRVToLLVMDbgTran::getDIBuilder. The missmatch happen because in previous used standard OpenCL.DebugInfo.100 DebugTypeInheritance had Child field as zero indexed argument. In newer standard that field is removed.

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/c6e9fea6d0f70f1
---
 llvm-spirv/lib/SPIRV/LLVMToSPIRVDbgTran.cpp | 37 +++++++++++------
 llvm-spirv/lib/SPIRV/SPIRVToLLVMDbgTran.cpp | 46 ++++++++++++++-------
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRV.debug.h | 19 +++++++--
 3 files changed, 72 insertions(+), 30 deletions(-)

diff --git a/llvm-spirv/lib/SPIRV/LLVMToSPIRVDbgTran.cpp b/llvm-spirv/lib/SPIRV/LLVMToSPIRVDbgTran.cpp
index 9964a9c9717c9..94a51397baae4 100644
--- a/llvm-spirv/lib/SPIRV/LLVMToSPIRVDbgTran.cpp
+++ b/llvm-spirv/lib/SPIRV/LLVMToSPIRVDbgTran.cpp
@@ -36,6 +36,7 @@
 //
 //===----------------------------------------------------------------------===//
 #include "LLVMToSPIRVDbgTran.h"
+#include "SPIRV.debug.h"
 #include "SPIRVWriter.h"
 
 #include "llvm/IR/DebugInfo.h"
@@ -1052,20 +1053,32 @@ LLVMToSPIRVDbgTran::transDbgMemberTypeNonSemantic(const DIDerivedType *MT) {
 
 SPIRVEntry *LLVMToSPIRVDbgTran::transDbgInheritance(const DIDerivedType *DT) {
   using namespace SPIRVDebug::Operand::TypeInheritance;
-  const SPIRVWord Offset = isNonSemanticDebugInfo() ? 1 : 0;
-  SPIRVWordVec Ops(OperandCount - Offset);
-  // There is no Child operand in NonSemantic debug spec
-  if (!isNonSemanticDebugInfo())
-    Ops[ChildIdx] = transDbgEntry(DT->getScope())->getId();
-  Ops[ParentIdx - Offset] = transDbgEntry(DT->getBaseType())->getId();
+  unsigned ParentIdx, OffsetIdx, SizeIdx, FlagsIdx, OperandCount;
+  if (isNonSemanticDebugInfo()) {
+    ParentIdx = NonSemantic::ParentIdx;
+    OffsetIdx = NonSemantic::OffsetIdx;
+    SizeIdx = NonSemantic::SizeIdx;
+    FlagsIdx = NonSemantic::FlagsIdx;
+    OperandCount = NonSemantic::OperandCount;
+  } else {
+    ParentIdx = OpenCL::ParentIdx;
+    OffsetIdx = OpenCL::OffsetIdx;
+    SizeIdx = OpenCL::SizeIdx;
+    FlagsIdx = OpenCL::FlagsIdx;
+    OperandCount = OpenCL::OperandCount;
+  }
+  SPIRVWordVec Ops(OperandCount);
+  Ops[ParentIdx] = transDbgEntry(DT->getBaseType())->getId();
   ConstantInt *OffsetInBits = getUInt(M, DT->getOffsetInBits());
-  Ops[OffsetIdx - Offset] =
-      SPIRVWriter->transValue(OffsetInBits, nullptr)->getId();
+  Ops[OffsetIdx] = SPIRVWriter->transValue(OffsetInBits, nullptr)->getId();
   ConstantInt *Size = getUInt(M, DT->getSizeInBits());
-  Ops[SizeIdx - Offset] = SPIRVWriter->transValue(Size, nullptr)->getId();
-  Ops[FlagsIdx - Offset] = transDebugFlags(DT);
-  if (isNonSemanticDebugInfo())
-    transformToConstant(Ops, {FlagsIdx - Offset});
+  Ops[SizeIdx] = SPIRVWriter->transValue(Size, nullptr)->getId();
+  Ops[FlagsIdx] = transDebugFlags(DT);
+  if (isNonSemanticDebugInfo()) {
+    transformToConstant(Ops, {FlagsIdx});
+  } else {
+    Ops[OpenCL::ChildIdx] = transDbgEntry(DT->getScope())->getId();
+  }
   return BM->addDebugInfo(SPIRVDebug::TypeInheritance, getVoidTy(), Ops);
 }
 
diff --git a/llvm-spirv/lib/SPIRV/SPIRVToLLVMDbgTran.cpp b/llvm-spirv/lib/SPIRV/SPIRVToLLVMDbgTran.cpp
index 13e0c5521fb79..fd34d48135b07 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVToLLVMDbgTran.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVToLLVMDbgTran.cpp
@@ -37,6 +37,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "SPIRVToLLVMDbgTran.h"
+#include "SPIRV.debug.h"
 #include "SPIRVEntry.h"
 #include "SPIRVFunction.h"
 #include "SPIRVInstruction.h"
@@ -1214,24 +1215,32 @@ DINode *SPIRVToLLVMDbgTran::transTypedef(const SPIRVExtInst *DebugInst) {
 
 DINode *SPIRVToLLVMDbgTran::transTypeInheritance(const SPIRVExtInst *DebugInst,
                                                  DIType *ChildClass) {
-  if (isNonSemanticDebugInfo(DebugInst->getExtSetKind()) && !ChildClass) {
-    // Will be translated later when processing TypeMember's parent
-    return nullptr;
-  }
   using namespace SPIRVDebug::Operand::TypeInheritance;
+  // The value is used when assertions are enabled
+  [[maybe_unused]] unsigned OperandCount;
+  unsigned ParentIdx, OffsetIdx, FlagsIdx;
+  if (isNonSemanticDebugInfo(DebugInst->getExtSetKind())) {
+    if (!ChildClass) {
+      // Will be translated later when processing TypeMember's parent
+      return nullptr;
+    }
+    OperandCount = NonSemantic::OperandCount;
+    ParentIdx = NonSemantic::ParentIdx;
+    OffsetIdx = NonSemantic::OffsetIdx;
+    FlagsIdx = NonSemantic::FlagsIdx;
+  } else {
+    OperandCount = NonSemantic::OperandCount;
+    ParentIdx = OpenCL::ParentIdx;
+    OffsetIdx = OpenCL::OffsetIdx;
+    FlagsIdx = OpenCL::FlagsIdx;
+  }
   const SPIRVWordVec &Ops = DebugInst->getArguments();
-  assert(Ops.size() >= MinOperandCount && "Invalid number of operands");
-  // No Child operand for NonSemantic debug spec
-  SPIRVWord Offset = isNonSemanticDebugInfo(DebugInst->getExtSetKind()) ? 1 : 0;
+  assert(Ops.size() >= OperandCount && "Invalid number of operands");
   DIType *Parent =
-      transDebugInst<DIType>(BM->get<SPIRVExtInst>(Ops[ParentIdx - Offset]));
-  DIType *Child =
-      isNonSemanticDebugInfo(DebugInst->getExtSetKind())
-          ? ChildClass
-          : transDebugInst<DIType>(BM->get<SPIRVExtInst>(Ops[ChildIdx]));
+      transDebugInst<DIType>(BM->get<SPIRVExtInst>(Ops[ParentIdx]));
   DINode::DIFlags Flags = DINode::FlagZero;
-  SPIRVWord SPIRVFlags = getConstantValueOrLiteral(Ops, FlagsIdx - Offset,
-                                                   DebugInst->getExtSetKind());
+  SPIRVWord SPIRVFlags =
+      getConstantValueOrLiteral(Ops, FlagsIdx, DebugInst->getExtSetKind());
   if ((SPIRVFlags & SPIRVDebug::FlagAccess) == SPIRVDebug::FlagIsPublic)
     Flags |= llvm::DINode::FlagPublic;
   if ((SPIRVFlags & SPIRVDebug::FlagAccess) == SPIRVDebug::FlagIsProtected)
@@ -1239,7 +1248,14 @@ DINode *SPIRVToLLVMDbgTran::transTypeInheritance(const SPIRVExtInst *DebugInst,
   if ((SPIRVFlags & SPIRVDebug::FlagAccess) == SPIRVDebug::FlagIsPrivate)
     Flags |= llvm::DINode::FlagPrivate;
   uint64_t OffsetVal =
-      BM->get<SPIRVConstant>(Ops[OffsetIdx - Offset])->getZExtIntValue();
+      BM->get<SPIRVConstant>(Ops[OffsetIdx])->getZExtIntValue();
+  DIType *Child;
+  if (isNonSemanticDebugInfo(DebugInst->getExtSetKind())) {
+    Child = ChildClass;
+  } else {
+    Child =
+        transDebugInst<DIType>(BM->get<SPIRVExtInst>(Ops[OpenCL::ChildIdx]));
+  }
   return getDIBuilder(DebugInst).createInheritance(Child, Parent, OffsetVal, 0,
                                                    Flags);
 }
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRV.debug.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRV.debug.h
index 19c863cf94766..7ffd960403418 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRV.debug.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRV.debug.h
@@ -503,17 +503,27 @@ enum {
 } // namespace TypeMember
 
 namespace TypeInheritance {
+namespace NonSemantic {
+enum {
+  ParentIdx       = 0,
+  OffsetIdx       = 1,
+  SizeIdx         = 2,
+  FlagsIdx        = 3,
+  OperandCount    = 4
+};
+}
+
+namespace OpenCL {
 enum {
   ChildIdx        = 0,
   ParentIdx       = 1,
   OffsetIdx       = 2,
   SizeIdx         = 3,
   FlagsIdx        = 4,
-  // NonSemantic
-  MinOperandCount = 4,
   OperandCount    = 5
 };
 }
+} // namespace TypeInheritance
 
 namespace TypePtrToMember {
 enum {
@@ -991,7 +1001,10 @@ inline bool hasDbgInstParentScopeIdx(
     ParentScopeIdx = TypeMember::OpenCL::ParentIdx;
     return true;
   case SPIRVDebug::TypeInheritance:
-    ParentScopeIdx = TypeInheritance::ParentIdx;
+    if (ExtKind == SPIRV::SPIRVEIS_OpenCL_DebugInfo_100)
+      ParentScopeIdx = TypeInheritance::OpenCL::ParentIdx;
+    else
+      ParentScopeIdx = TypeInheritance::NonSemantic::ParentIdx;
     return true;
   case SPIRVDebug::TypePtrToMember:
     ParentScopeIdx = TypePtrToMember::ParentIdx;

From 3c370d7dd4e86a565c83066fa2afc754859198d6 Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Tue, 16 Apr 2024 16:03:18 +0200
Subject: [PATCH 879/886] Move the version to supported (#2497)

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/a104311e5ac8f76
---
 llvm-spirv/README.md                                   | 2 +-
 llvm-spirv/docs/SPIRVVersionsAndExtensionsHandling.rst | 2 +-
 llvm-spirv/lib/SPIRV/SPIRVWriter.cpp                   | 2 ++
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRVErrorEnum.h         | 2 +-
 llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp             | 3 +--
 5 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/llvm-spirv/README.md b/llvm-spirv/README.md
index 65480436ba889..0a06f53cf52d4 100644
--- a/llvm-spirv/README.md
+++ b/llvm-spirv/README.md
@@ -203,7 +203,7 @@ the version of the SPIR-V file which is being generated/consumed.
   the input file and emit an error if the SPIR-V version in it is higher than
   one specified via this option.
 
-Allowed values are `1.0`, `1.1`, `1.2`, `1.3`, and `1.4`.
+Allowed values are `1.0`, `1.1`, `1.2`, `1.3`, `1.4`, and `1.5`.
 
 More information can be found in
 [SPIR-V versions and extensions handling](docs/SPIRVVersionsAndExtensionsHandling.rst)
diff --git a/llvm-spirv/docs/SPIRVVersionsAndExtensionsHandling.rst b/llvm-spirv/docs/SPIRVVersionsAndExtensionsHandling.rst
index 90fd05e9e75fc..03f2c057edf41 100644
--- a/llvm-spirv/docs/SPIRVVersionsAndExtensionsHandling.rst
+++ b/llvm-spirv/docs/SPIRVVersionsAndExtensionsHandling.rst
@@ -65,7 +65,7 @@ the version of the SPIR-V file which is being generated/consumed.
   one exception: resulting SPIR-V version cannot be raised higher than
   specified by this option.
 
-Allowed values are ``1.0``, ``1.1``, ``1.2``, ``1.3``, and ``1.4``.
+Allowed values are ``1.0``, ``1.1``, ``1.2``, ``1.3``, ``1.4``, and ``1.5``.
 
 .. warning::
    These two options are mutually exclusive and cannot be specified at the
diff --git a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
index 9148b2ec74b92..bac1d0c4f1564 100644
--- a/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
+++ b/llvm-spirv/lib/SPIRV/SPIRVWriter.cpp
@@ -6736,6 +6736,8 @@ VersionNumber getVersionFromTriple(const Triple &TT, SPIRVErrorLog &ErrorLog) {
     return VersionNumber::SPIRV_1_3;
   case Triple::SPIRVSubArch_v14:
     return VersionNumber::SPIRV_1_4;
+  case Triple::SPIRVSubArch_v15:
+    return VersionNumber::SPIRV_1_5;
   default:
     ErrorLog.checkError(false, SPIRVEC_InvalidSubArch, TT.getArchName().str());
     return VersionNumber::MaximumVersion;
diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVErrorEnum.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVErrorEnum.h
index 7a7f76df95fd7..bd9c020a7a4dd 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVErrorEnum.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVErrorEnum.h
@@ -2,7 +2,7 @@
 _SPIRV_OP(Success, "Success")
 _SPIRV_OP(InvalidTargetTriple,
           "Expects spir-unknown-unknown or spir64-unknown-unknown.")
-_SPIRV_OP(InvalidSubArch, "Expecting v1.0-v1.4.")
+_SPIRV_OP(InvalidSubArch, "Expecting v1.0-v1.5.")
 _SPIRV_OP(TripleMaxVersionIncompatible,
           "Triple version and maximum version are incompatible.")
 _SPIRV_OP(InvalidAddressingModel, "Expects 0-2.")
diff --git a/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp b/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
index 35a7c4d7c0f2e..59a01a81e4ed5 100644
--- a/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
+++ b/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
@@ -112,8 +112,7 @@ static cl::opt<VersionNumber> MaxSPIRVVersion(
                clEnumValN(VersionNumber::SPIRV_1_2, "1.2", "SPIR-V 1.2"),
                clEnumValN(VersionNumber::SPIRV_1_3, "1.3", "SPIR-V 1.3"),
                clEnumValN(VersionNumber::SPIRV_1_4, "1.4", "SPIR-V 1.4"),
-               clEnumValN(VersionNumber::SPIRV_1_5, "1.5",
-                          "SPIR-V 1.5 (experimental)")),
+               clEnumValN(VersionNumber::SPIRV_1_5, "1.5", "SPIR-V 1.5")),
     cl::init(VersionNumber::MaximumVersion));
 
 static cl::list<std::string>

From a0b316f820319edac5e0b94237e25b4c77daaa9c Mon Sep 17 00:00:00 2001
From: Viktoria Maximova <viktoria.maksimova@intel.com>
Date: Wed, 17 Apr 2024 15:30:27 +0200
Subject: [PATCH 880/886] Introduce SPIR-V 1.6 version (#2505)

Recognize SPIR-V 1.6 modules and allow specifying the `--spirv-max-version=1.6` option.

This does not implement any SPIR-V 1.6 functionality yet, so mark it as "experimental".

Original commit:
https://github.com/KhronosGroup/SPIRV-LLVM-Translator/commit/e1f7ebee5ed8b90
---
 llvm-spirv/include/LLVMSPIRVOpts.h                    | 5 ++++-
 llvm-spirv/test/negative/spirv-version-controls-1.spt | 4 ++--
 llvm-spirv/test/negative/spirv-version-controls-2.spt | 2 +-
 llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp            | 4 +++-
 4 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/llvm-spirv/include/LLVMSPIRVOpts.h b/llvm-spirv/include/LLVMSPIRVOpts.h
index ee7f38e75f4d5..e541b1028849a 100644
--- a/llvm-spirv/include/LLVMSPIRVOpts.h
+++ b/llvm-spirv/include/LLVMSPIRVOpts.h
@@ -64,8 +64,9 @@ enum class VersionNumber : uint32_t {
   SPIRV_1_3 = 0x00010300,
   SPIRV_1_4 = 0x00010400,
   SPIRV_1_5 = 0x00010500,
+  SPIRV_1_6 = 0x00010600,
   MinimumVersion = SPIRV_1_0,
-  MaximumVersion = SPIRV_1_5
+  MaximumVersion = SPIRV_1_6
 };
 
 inline constexpr std::string_view formatVersionNumber(uint32_t Version) {
@@ -82,6 +83,8 @@ inline constexpr std::string_view formatVersionNumber(uint32_t Version) {
     return "1.4";
   case static_cast<uint32_t>(VersionNumber::SPIRV_1_5):
     return "1.5";
+  case static_cast<uint32_t>(VersionNumber::SPIRV_1_6):
+    return "1.6";
   }
   return "unknown";
 }
diff --git a/llvm-spirv/test/negative/spirv-version-controls-1.spt b/llvm-spirv/test/negative/spirv-version-controls-1.spt
index b2a7d9dd5c244..94b9263d64aca 100644
--- a/llvm-spirv/test/negative/spirv-version-controls-1.spt
+++ b/llvm-spirv/test/negative/spirv-version-controls-1.spt
@@ -1,4 +1,4 @@
-119734787 67072 393230 12 0
+119734787 67328 393230 12 0
 2 Capability Addresses
 2 Capability Kernel
 5 ExtInstImport 1 "OpenCL.std"
@@ -29,4 +29,4 @@
 
 ; RUN: not llvm-spirv %s -to-binary -o - 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR
 ;
-; CHECK-ERROR: Invalid SPIR-V module: unsupported SPIR-V version number 'unknown (67072)'. Range of supported/known SPIR-V versions is 1.0 (65536) - 1.5 (66816)
+; CHECK-ERROR: Invalid SPIR-V module: unsupported SPIR-V version number 'unknown (67328)'. Range of supported/known SPIR-V versions is 1.0 (65536) - 1.6 (67072)
diff --git a/llvm-spirv/test/negative/spirv-version-controls-2.spt b/llvm-spirv/test/negative/spirv-version-controls-2.spt
index a75cc4f38c2d8..edb3fd8e5db00 100644
--- a/llvm-spirv/test/negative/spirv-version-controls-2.spt
+++ b/llvm-spirv/test/negative/spirv-version-controls-2.spt
@@ -29,4 +29,4 @@
 
 ; RUN: not llvm-spirv %s -to-binary -o - 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR
 ;
-; CHECK-ERROR: Invalid SPIR-V module: unsupported SPIR-V version number 'unknown (1024)'. Range of supported/known SPIR-V versions is 1.0 (65536) - 1.5 (66816)
+; CHECK-ERROR: Invalid SPIR-V module: unsupported SPIR-V version number 'unknown (1024)'. Range of supported/known SPIR-V versions is 1.0 (65536) - 1.6 (67072)
diff --git a/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp b/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
index 59a01a81e4ed5..c90f16480fe86 100644
--- a/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
+++ b/llvm-spirv/tools/llvm-spirv/llvm-spirv.cpp
@@ -112,7 +112,9 @@ static cl::opt<VersionNumber> MaxSPIRVVersion(
                clEnumValN(VersionNumber::SPIRV_1_2, "1.2", "SPIR-V 1.2"),
                clEnumValN(VersionNumber::SPIRV_1_3, "1.3", "SPIR-V 1.3"),
                clEnumValN(VersionNumber::SPIRV_1_4, "1.4", "SPIR-V 1.4"),
-               clEnumValN(VersionNumber::SPIRV_1_5, "1.5", "SPIR-V 1.5")),
+               clEnumValN(VersionNumber::SPIRV_1_5, "1.5", "SPIR-V 1.5"),
+               clEnumValN(VersionNumber::SPIRV_1_6, "1.6",
+                          "SPIR-V 1.6 (experimental)")),
     cl::init(VersionNumber::MaximumVersion));
 
 static cl::list<std::string>

From 85394b7dad82ff5ccbb338b1a1da4c8e58299a31 Mon Sep 17 00:00:00 2001
From: Haonan Yang <haonan.yang@intel.com>
Date: Fri, 19 Apr 2024 21:43:07 +0800
Subject: [PATCH 881/886] Fix wrong resolution. (#13489)

This is related to
https://github.com/intel/llvm/commit/71c0784dc4a2ef40039a97af122ba78549193120.
And now it's aligned with
https://github.com/intel/llvm/blob/sycl/clang/lib/AST/DeclPrinter.cpp#L1097.
Fixes 12 lit.
---
 clang/lib/AST/DeclPrinter.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/lib/AST/DeclPrinter.cpp b/clang/lib/AST/DeclPrinter.cpp
index ea1443964d955..b4f9821efe2f9 100644
--- a/clang/lib/AST/DeclPrinter.cpp
+++ b/clang/lib/AST/DeclPrinter.cpp
@@ -1103,7 +1103,7 @@ void DeclPrinter::VisitCXXRecordDecl(CXXRecordDecl *D) {
 
   prettyPrintAttributes(D, AttrPosAsWritten::Right);
 
-  if (D->isCompleteDefinition()) {
+  if (D->isCompleteDefinition() && !Policy.SuppressDefinition) {
     Out << ' ';
     // Print the base classes
     if (D->getNumBases()) {

From 3c89596885adb865b3d1be2fbb381599489147fe Mon Sep 17 00:00:00 2001
From: sys_ce_bb <sys_ce_bb@intel.com>
Date: Fri, 19 Apr 2024 07:27:23 -0700
Subject: [PATCH 882/886] Call updateModeleVersion to enable SPIRV inst version
 check

Allows translator to trigger getRequiredSPIRVVersion method and produce
an error if the instruction is not supported for the certain SPIR-V
version.
---
 llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
index c817a4e120350..65bc7ac214bc5 100644
--- a/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
+++ b/llvm-spirv/lib/SPIRV/libSPIRV/SPIRVInstruction.h
@@ -276,6 +276,9 @@ class SPIRVInstTemplateBase : public SPIRVInstruction {
     } else
       SPIRVEntry::setWordCount(WC);
     Ops = TheOps;
+    // The required SPIR-V version depends on the operands for some
+    // instructions.
+    updateModuleVersion();
   }
   void setWordCount(SPIRVWord TheWordCount) override {
     SPIRVEntry::setWordCount(TheWordCount);

From d800d17cf26419eb2381b445d97c3c80543ba626 Mon Sep 17 00:00:00 2001
From: sys_ce_bb <sys_ce_bb@intel.com>
Date: Fri, 19 Apr 2024 07:30:15 -0700
Subject: [PATCH 883/886] Disable OpLessOrGreater test until SPIR-V 1.5 is
 supported

---
 llvm-spirv/test/transcoding/OpLessOrGreater.ll | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/llvm-spirv/test/transcoding/OpLessOrGreater.ll b/llvm-spirv/test/transcoding/OpLessOrGreater.ll
index fc885c68a1f02..648b987b70890 100644
--- a/llvm-spirv/test/transcoding/OpLessOrGreater.ll
+++ b/llvm-spirv/test/transcoding/OpLessOrGreater.ll
@@ -1,3 +1,5 @@
+; TODO: Re-enable this test when support for SPIRV 1.5 is enabled.
+; XFAIL:*
 ; RUN: llvm-as %s -o %t.bc
 ; RUN: llvm-spirv %t.bc -o %t.spv
 ; RUN: llvm-spirv %t.spv -to-text -o %t.spt

From 07007595e3ec2df0854c75434d3f2e3b34909bf5 Mon Sep 17 00:00:00 2001
From: sys_ce_bb <sys_ce_bb@intel.com>
Date: Fri, 19 Apr 2024 14:11:16 -0700
Subject: [PATCH 884/886] Fix integration_header.cpp test due to 71c0784d

Changes to final keyword printing.
---
 clang/test/CodeGenSYCL/integration_header.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/clang/test/CodeGenSYCL/integration_header.cpp b/clang/test/CodeGenSYCL/integration_header.cpp
index e2f4c8f4d0f2d..f39d08a77495b 100644
--- a/clang/test/CodeGenSYCL/integration_header.cpp
+++ b/clang/test/CodeGenSYCL/integration_header.cpp
@@ -12,7 +12,7 @@
 // CHECK-NEXT: template <int DimX> struct namespaced_arg;
 // CHECK-NEXT: }
 // CHECK-NEXT: template <typename ...Ts> class fourth_kernel;
-// CHECK: class FinalClass;
+// CHECK: class FinalClass final;
 //
 // CHECK: static constexpr
 // CHECK-NEXT: const char* const kernel_names[] = {

From f27e5de82dae8af5a09913179849d2304bb5b6e3 Mon Sep 17 00:00:00 2001
From: Akira Hatanaka <ahatanak@gmail.com>
Date: Thu, 28 Mar 2024 06:54:36 -0700
Subject: [PATCH 885/886] [CodeGen][arm64e] Add methods and data members to
 Address, which are needed to authenticate signed pointers (#86923)

To authenticate pointers, CodeGen needs access to the key and
discriminators that were used to sign the pointer. That information is
sometimes known from the context, but not always, which is why `Address`
needs to hold that information.

This patch adds methods and data members to `Address`, which will be
needed in subsequent patches to authenticate signed pointers, and uses
the newly added methods throughout CodeGen. Although this patch isn't
strictly NFC as it causes CodeGen to use different code paths in some
cases (e.g., `mergeAddressesInConditionalExpr`), it doesn't cause any
changes in functionality as it doesn't add any information needed for
authentication.

In addition to the changes mentioned above, this patch introduces class
`RawAddress`, which contains a pointer that we know is unsigned, and
adds several new functions for creating `Address` and `LValue` objects.

This reapplies d9a685a9dd589486e882b722e513ee7b8c84870c, which was
reverted because it broke ubsan bots. There seems to be a bug in
coroutine code-gen, which is causing EmitTypeCheck to use the wrong
alignment. For now, pass alignment zero to EmitTypeCheck so that it can
compute the correct alignment based on the passed type (see function
EmitCXXMemberOrOperatorMemberCallExpr).
---
 clang/lib/CodeGen/ABIInfoImpl.cpp        |  10 +-
 clang/lib/CodeGen/Address.h              | 195 ++++++++++++++---
 clang/lib/CodeGen/CGAtomic.cpp           |  53 ++---
 clang/lib/CodeGen/CGBlocks.cpp           |  32 +--
 clang/lib/CodeGen/CGBlocks.h             |   3 +-
 clang/lib/CodeGen/CGBuilder.h            | 234 ++++++++++++++-------
 clang/lib/CodeGen/CGBuiltin.cpp          | 173 +++++++--------
 clang/lib/CodeGen/CGCUDANV.cpp           |  19 +-
 clang/lib/CodeGen/CGCXXABI.cpp           |  21 +-
 clang/lib/CodeGen/CGCXXABI.h             |  14 +-
 clang/lib/CodeGen/CGCall.cpp             | 167 ++++++++-------
 clang/lib/CodeGen/CGCall.h               |   1 +
 clang/lib/CodeGen/CGClass.cpp            |  76 ++++---
 clang/lib/CodeGen/CGCleanup.cpp          | 110 ++++------
 clang/lib/CodeGen/CGCleanup.h            |   2 +-
 clang/lib/CodeGen/CGCoroutine.cpp        |   4 +-
 clang/lib/CodeGen/CGDecl.cpp             |  28 +--
 clang/lib/CodeGen/CGException.cpp        |  19 +-
 clang/lib/CodeGen/CGExpr.cpp             | 227 ++++++++++----------
 clang/lib/CodeGen/CGExprAgg.cpp          |  29 +--
 clang/lib/CodeGen/CGExprCXX.cpp          | 111 +++++-----
 clang/lib/CodeGen/CGExprConstant.cpp     |   4 +-
 clang/lib/CodeGen/CGExprScalar.cpp       |  23 +-
 clang/lib/CodeGen/CGNonTrivialStruct.cpp |   8 +-
 clang/lib/CodeGen/CGObjC.cpp             |  43 ++--
 clang/lib/CodeGen/CGObjCGNU.cpp          |  42 ++--
 clang/lib/CodeGen/CGObjCMac.cpp          |  95 ++++-----
 clang/lib/CodeGen/CGObjCRuntime.cpp      |   6 +-
 clang/lib/CodeGen/CGOpenMPRuntime.cpp    | 194 +++++++++--------
 clang/lib/CodeGen/CGOpenMPRuntime.h      |   5 +-
 clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp |  76 +++----
 clang/lib/CodeGen/CGStmt.cpp             |   8 +-
 clang/lib/CodeGen/CGStmtOpenMP.cpp       |  87 ++++----
 clang/lib/CodeGen/CGVTables.cpp          |   9 +-
 clang/lib/CodeGen/CGValue.h              | 250 +++++++++++-----------
 clang/lib/CodeGen/CodeGenFunction.cpp    |  71 ++++---
 clang/lib/CodeGen/CodeGenFunction.h      | 257 ++++++++++++++++-------
 clang/lib/CodeGen/CodeGenModule.cpp      |   2 +-
 clang/lib/CodeGen/CodeGenPGO.cpp         |  10 +-
 clang/lib/CodeGen/CodeGenPGO.h           |   6 +-
 clang/lib/CodeGen/ItaniumCXXABI.cpp      |  52 ++---
 clang/lib/CodeGen/MicrosoftCXXABI.cpp    |  58 ++---
 clang/lib/CodeGen/TargetInfo.h           |   5 +
 clang/lib/CodeGen/Targets/NVPTX.cpp      |   2 +-
 clang/lib/CodeGen/Targets/PPC.cpp        |  11 +-
 clang/lib/CodeGen/Targets/Sparc.cpp      |   2 +-
 clang/lib/CodeGen/Targets/SystemZ.cpp    |   9 +-
 clang/lib/CodeGen/Targets/XCore.cpp      |   2 +-
 clang/utils/TableGen/MveEmitter.cpp      |   2 +-
 llvm/include/llvm/IR/IRBuilder.h         |   1 +
 50 files changed, 1636 insertions(+), 1232 deletions(-)

diff --git a/clang/lib/CodeGen/ABIInfoImpl.cpp b/clang/lib/CodeGen/ABIInfoImpl.cpp
index f44415ff67593..1e4d48e13c025 100644
--- a/clang/lib/CodeGen/ABIInfoImpl.cpp
+++ b/clang/lib/CodeGen/ABIInfoImpl.cpp
@@ -187,7 +187,7 @@ CodeGen::emitVoidPtrDirectVAArg(CodeGenFunction &CGF, Address VAListAddr,
   CharUnits FullDirectSize = DirectSize.alignTo(SlotSize);
   Address NextPtr =
       CGF.Builder.CreateConstInBoundsByteGEP(Addr, FullDirectSize, "argp.next");
-  CGF.Builder.CreateStore(NextPtr.getPointer(), VAListAddr);
+  CGF.Builder.CreateStore(NextPtr.emitRawPointer(CGF), VAListAddr);
 
   // If the argument is smaller than a slot, and this is a big-endian
   // target, the argument will be right-adjusted in its slot.
@@ -239,8 +239,8 @@ Address CodeGen::emitMergePHI(CodeGenFunction &CGF, Address Addr1,
                               const llvm::Twine &Name) {
   assert(Addr1.getType() == Addr2.getType());
   llvm::PHINode *PHI = CGF.Builder.CreatePHI(Addr1.getType(), 2, Name);
-  PHI->addIncoming(Addr1.getPointer(), Block1);
-  PHI->addIncoming(Addr2.getPointer(), Block2);
+  PHI->addIncoming(Addr1.emitRawPointer(CGF), Block1);
+  PHI->addIncoming(Addr2.emitRawPointer(CGF), Block2);
   CharUnits Align = std::min(Addr1.getAlignment(), Addr2.getAlignment());
   return Address(PHI, Addr1.getElementType(), Align);
 }
@@ -400,7 +400,7 @@ Address CodeGen::EmitVAArgInstr(CodeGenFunction &CGF, Address VAListAddr,
     llvm::Type *ElementTy = CGF.ConvertTypeForMem(Ty);
     llvm::Type *BaseTy = llvm::PointerType::getUnqual(ElementTy);
     llvm::Value *Addr =
-        CGF.Builder.CreateVAArg(VAListAddr.getPointer(), BaseTy);
+        CGF.Builder.CreateVAArg(VAListAddr.emitRawPointer(CGF), BaseTy);
     return Address(Addr, ElementTy, TyAlignForABI);
   } else {
     assert((AI.isDirect() || AI.isExtend()) &&
@@ -416,7 +416,7 @@ Address CodeGen::EmitVAArgInstr(CodeGenFunction &CGF, Address VAListAddr,
            "Unexpected CoerceToType seen in arginfo in generic VAArg emitter!");
 
     Address Temp = CGF.CreateMemTemp(Ty, "varet");
-    Val = CGF.Builder.CreateVAArg(VAListAddr.getPointer(),
+    Val = CGF.Builder.CreateVAArg(VAListAddr.emitRawPointer(CGF),
                                   CGF.ConvertTypeForMem(Ty));
     CGF.Builder.CreateStore(Val, Temp);
     return Temp;
diff --git a/clang/lib/CodeGen/Address.h b/clang/lib/CodeGen/Address.h
index cf48df8f5e736..35ec370a139c9 100644
--- a/clang/lib/CodeGen/Address.h
+++ b/clang/lib/CodeGen/Address.h
@@ -15,6 +15,7 @@
 #define LLVM_CLANG_LIB_CODEGEN_ADDRESS_H
 
 #include "clang/AST/CharUnits.h"
+#include "clang/AST/Type.h"
 #include "llvm/ADT/PointerIntPair.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/Support/MathExtras.h"
@@ -22,28 +23,41 @@
 namespace clang {
 namespace CodeGen {
 
+class Address;
+class CGBuilderTy;
+class CodeGenFunction;
+class CodeGenModule;
+
 // Indicates whether a pointer is known not to be null.
 enum KnownNonNull_t { NotKnownNonNull, KnownNonNull };
 
-/// An aligned address.
-class Address {
+/// An abstract representation of an aligned address. This is designed to be an
+/// IR-level abstraction, carrying just the information necessary to perform IR
+/// operations on an address like loads and stores.  In particular, it doesn't
+/// carry C type information or allow the representation of things like
+/// bit-fields; clients working at that level should generally be using
+/// `LValue`.
+/// The pointer contained in this class is known to be unsigned.
+class RawAddress {
   llvm::PointerIntPair<llvm::Value *, 1, bool> PointerAndKnownNonNull;
   llvm::Type *ElementType;
   CharUnits Alignment;
 
 protected:
-  Address(std::nullptr_t) : ElementType(nullptr) {}
+  RawAddress(std::nullptr_t) : ElementType(nullptr) {}
 
 public:
-  Address(llvm::Value *Pointer, llvm::Type *ElementType, CharUnits Alignment,
-          KnownNonNull_t IsKnownNonNull = NotKnownNonNull)
+  RawAddress(llvm::Value *Pointer, llvm::Type *ElementType, CharUnits Alignment,
+             KnownNonNull_t IsKnownNonNull = NotKnownNonNull)
       : PointerAndKnownNonNull(Pointer, IsKnownNonNull),
         ElementType(ElementType), Alignment(Alignment) {
     assert(Pointer != nullptr && "Pointer cannot be null");
     assert(ElementType != nullptr && "Element type cannot be null");
   }
 
-  static Address invalid() { return Address(nullptr); }
+  inline RawAddress(Address Addr);
+
+  static RawAddress invalid() { return RawAddress(nullptr); }
   bool isValid() const {
     return PointerAndKnownNonNull.getPointer() != nullptr;
   }
@@ -80,6 +94,133 @@ class Address {
     return Alignment;
   }
 
+  /// Return address with different element type, but same pointer and
+  /// alignment.
+  RawAddress withElementType(llvm::Type *ElemTy) const {
+    return RawAddress(getPointer(), ElemTy, getAlignment(), isKnownNonNull());
+  }
+
+  KnownNonNull_t isKnownNonNull() const {
+    assert(isValid());
+    return (KnownNonNull_t)PointerAndKnownNonNull.getInt();
+  }
+};
+
+/// Like RawAddress, an abstract representation of an aligned address, but the
+/// pointer contained in this class is possibly signed.
+class Address {
+  friend class CGBuilderTy;
+
+  // The boolean flag indicates whether the pointer is known to be non-null.
+  llvm::PointerIntPair<llvm::Value *, 1, bool> Pointer;
+
+  /// The expected IR type of the pointer. Carrying accurate element type
+  /// information in Address makes it more convenient to work with Address
+  /// values and allows frontend assertions to catch simple mistakes.
+  llvm::Type *ElementType = nullptr;
+
+  CharUnits Alignment;
+
+  /// Offset from the base pointer.
+  llvm::Value *Offset = nullptr;
+
+  llvm::Value *emitRawPointerSlow(CodeGenFunction &CGF) const;
+
+protected:
+  Address(std::nullptr_t) : ElementType(nullptr) {}
+
+public:
+  Address(llvm::Value *pointer, llvm::Type *elementType, CharUnits alignment,
+          KnownNonNull_t IsKnownNonNull = NotKnownNonNull)
+      : Pointer(pointer, IsKnownNonNull), ElementType(elementType),
+        Alignment(alignment) {
+    assert(pointer != nullptr && "Pointer cannot be null");
+    assert(elementType != nullptr && "Element type cannot be null");
+    assert(!alignment.isZero() && "Alignment cannot be zero");
+  }
+
+  Address(llvm::Value *BasePtr, llvm::Type *ElementType, CharUnits Alignment,
+          llvm::Value *Offset, KnownNonNull_t IsKnownNonNull = NotKnownNonNull)
+      : Pointer(BasePtr, IsKnownNonNull), ElementType(ElementType),
+        Alignment(Alignment), Offset(Offset) {}
+
+  Address(RawAddress RawAddr)
+      : Pointer(RawAddr.isValid() ? RawAddr.getPointer() : nullptr),
+        ElementType(RawAddr.isValid() ? RawAddr.getElementType() : nullptr),
+        Alignment(RawAddr.isValid() ? RawAddr.getAlignment()
+                                    : CharUnits::Zero()) {}
+
+  static Address invalid() { return Address(nullptr); }
+  bool isValid() const { return Pointer.getPointer() != nullptr; }
+
+  /// This function is used in situations where the caller is doing some sort of
+  /// opaque "laundering" of the pointer.
+  void replaceBasePointer(llvm::Value *P) {
+    assert(isValid() && "pointer isn't valid");
+    assert(P->getType() == Pointer.getPointer()->getType() &&
+           "Pointer's type changed");
+    Pointer.setPointer(P);
+    assert(isValid() && "pointer is invalid after replacement");
+  }
+
+  CharUnits getAlignment() const { return Alignment; }
+
+  void setAlignment(CharUnits Value) { Alignment = Value; }
+
+  llvm::Value *getBasePointer() const {
+    assert(isValid() && "pointer isn't valid");
+    return Pointer.getPointer();
+  }
+
+  /// Return the type of the pointer value.
+  llvm::PointerType *getType() const {
+    return llvm::PointerType::get(
+        ElementType,
+        llvm::cast<llvm::PointerType>(Pointer.getPointer()->getType())
+            ->getAddressSpace());
+  }
+
+  /// Return the type of the values stored in this address.
+  llvm::Type *getElementType() const {
+    assert(isValid());
+    return ElementType;
+  }
+
+  /// Return the address space that this address resides in.
+  unsigned getAddressSpace() const { return getType()->getAddressSpace(); }
+
+  /// Return the IR name of the pointer value.
+  llvm::StringRef getName() const { return Pointer.getPointer()->getName(); }
+
+  // This function is called only in CGBuilderBaseTy::CreateElementBitCast.
+  void setElementType(llvm::Type *Ty) {
+    assert(hasOffset() &&
+           "this funcion shouldn't be called when there is no offset");
+    ElementType = Ty;
+  }
+
+  /// Whether the pointer is known not to be null.
+  KnownNonNull_t isKnownNonNull() const {
+    assert(isValid());
+    return (KnownNonNull_t)Pointer.getInt();
+  }
+
+  Address setKnownNonNull() {
+    assert(isValid());
+    Pointer.setInt(KnownNonNull);
+    return *this;
+  }
+
+  bool hasOffset() const { return Offset; }
+
+  llvm::Value *getOffset() const { return Offset; }
+
+  /// Return the pointer contained in this class after authenticating it and
+  /// adding offset to it if necessary.
+  llvm::Value *emitRawPointer(CodeGenFunction &CGF) const {
+    return getBasePointer();
+  }
+
   /// Return address with different pointer, but same element type and
   /// alignment.
   Address withPointer(llvm::Value *NewPointer,
@@ -91,61 +232,59 @@ class Address {
   /// Return address with different alignment, but same pointer and element
   /// type.
   Address withAlignment(CharUnits NewAlignment) const {
-    return Address(getPointer(), getElementType(), NewAlignment,
+    return Address(Pointer.getPointer(), getElementType(), NewAlignment,
                    isKnownNonNull());
   }
 
   /// Return address with different element type, but same pointer and
   /// alignment.
   Address withElementType(llvm::Type *ElemTy) const {
-    return Address(getPointer(), ElemTy, getAlignment(), isKnownNonNull());
-  }
-
-  /// Whether the pointer is known not to be null.
-  KnownNonNull_t isKnownNonNull() const {
-    assert(isValid());
-    return (KnownNonNull_t)PointerAndKnownNonNull.getInt();
-  }
-
-  /// Set the non-null bit.
-  Address setKnownNonNull() {
-    assert(isValid());
-    PointerAndKnownNonNull.setInt(true);
-    return *this;
+    if (!hasOffset())
+      return Address(getBasePointer(), ElemTy, getAlignment(), nullptr,
+                     isKnownNonNull());
+    Address A(*this);
+    A.ElementType = ElemTy;
+    return A;
   }
 };
 
+inline RawAddress::RawAddress(Address Addr)
+    : PointerAndKnownNonNull(Addr.isValid() ? Addr.getBasePointer() : nullptr,
+                             Addr.isValid() ? Addr.isKnownNonNull()
+                                            : NotKnownNonNull),
+      ElementType(Addr.isValid() ? Addr.getElementType() : nullptr),
+      Alignment(Addr.isValid() ? Addr.getAlignment() : CharUnits::Zero()) {}
+
 /// A specialization of Address that requires the address to be an
 /// LLVM Constant.
-class ConstantAddress : public Address {
-  ConstantAddress(std::nullptr_t) : Address(nullptr) {}
+class ConstantAddress : public RawAddress {
+  ConstantAddress(std::nullptr_t) : RawAddress(nullptr) {}
 
 public:
   ConstantAddress(llvm::Constant *pointer, llvm::Type *elementType,
                   CharUnits alignment)
-      : Address(pointer, elementType, alignment) {}
+      : RawAddress(pointer, elementType, alignment) {}
 
   static ConstantAddress invalid() {
     return ConstantAddress(nullptr);
   }
 
   llvm::Constant *getPointer() const {
-    return llvm::cast<llvm::Constant>(Address::getPointer());
+    return llvm::cast<llvm::Constant>(RawAddress::getPointer());
   }
 
   ConstantAddress withElementType(llvm::Type *ElemTy) const {
     return ConstantAddress(getPointer(), ElemTy, getAlignment());
   }
 
-  static bool isaImpl(Address addr) {
+  static bool isaImpl(RawAddress addr) {
     return llvm::isa<llvm::Constant>(addr.getPointer());
   }
-  static ConstantAddress castImpl(Address addr) {
+  static ConstantAddress castImpl(RawAddress addr) {
     return ConstantAddress(llvm::cast<llvm::Constant>(addr.getPointer()),
                            addr.getElementType(), addr.getAlignment());
   }
 };
-
 }
 
 // Present a minimal LLVM-like casting interface.
diff --git a/clang/lib/CodeGen/CGAtomic.cpp b/clang/lib/CodeGen/CGAtomic.cpp
index 60b3bc386d5de..5e2f90f74c81f 100644
--- a/clang/lib/CodeGen/CGAtomic.cpp
+++ b/clang/lib/CodeGen/CGAtomic.cpp
@@ -80,7 +80,7 @@ namespace {
         AtomicSizeInBits = C.toBits(
             C.toCharUnitsFromBits(Offset + OrigBFI.Size + C.getCharWidth() - 1)
                 .alignTo(lvalue.getAlignment()));
-        llvm::Value *BitFieldPtr = lvalue.getBitFieldPointer();
+        llvm::Value *BitFieldPtr = lvalue.getRawBitFieldPointer(CGF);
         auto OffsetInChars =
             (C.toCharUnitsFromBits(OrigBFI.Offset) / lvalue.getAlignment()) *
             lvalue.getAlignment();
@@ -140,13 +140,13 @@ namespace {
     const LValue &getAtomicLValue() const { return LVal; }
     llvm::Value *getAtomicPointer() const {
       if (LVal.isSimple())
-        return LVal.getPointer(CGF);
+        return LVal.emitRawPointer(CGF);
       else if (LVal.isBitField())
-        return LVal.getBitFieldPointer();
+        return LVal.getRawBitFieldPointer(CGF);
       else if (LVal.isVectorElt())
-        return LVal.getVectorPointer();
+        return LVal.getRawVectorPointer(CGF);
       assert(LVal.isExtVectorElt());
-      return LVal.getExtVectorPointer();
+      return LVal.getRawExtVectorPointer(CGF);
     }
     Address getAtomicAddress() const {
       llvm::Type *ElTy;
@@ -369,7 +369,7 @@ bool AtomicInfo::emitMemSetZeroIfNecessary() const {
     return false;
 
   CGF.Builder.CreateMemSet(
-      addr.getPointer(), llvm::ConstantInt::get(CGF.Int8Ty, 0),
+      addr.emitRawPointer(CGF), llvm::ConstantInt::get(CGF.Int8Ty, 0),
       CGF.getContext().toCharUnitsFromBits(AtomicSizeInBits).getQuantity(),
       LVal.getAlignment().getAsAlign());
   return true;
@@ -1056,7 +1056,8 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
       return getTargetHooks().performAddrSpaceCast(
           *this, V, AS, LangAS::opencl_generic, DestType, false);
     };
-    Args.add(RValue::get(CastToGenericAddrSpace(Ptr.getPointer(),
+
+    Args.add(RValue::get(CastToGenericAddrSpace(Ptr.emitRawPointer(*this),
                                                 E->getPtr()->getType())),
              getContext().VoidPtrTy);
 
@@ -1087,10 +1088,10 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
       LibCallName = "__atomic_compare_exchange";
       RetTy = getContext().BoolTy;
       HaveRetTy = true;
-      Args.add(RValue::get(CastToGenericAddrSpace(Val1.getPointer(),
+      Args.add(RValue::get(CastToGenericAddrSpace(Val1.emitRawPointer(*this),
                                                   E->getVal1()->getType())),
                getContext().VoidPtrTy);
-      Args.add(RValue::get(CastToGenericAddrSpace(Val2.getPointer(),
+      Args.add(RValue::get(CastToGenericAddrSpace(Val2.emitRawPointer(*this),
                                                   E->getVal2()->getType())),
                getContext().VoidPtrTy);
       Args.add(RValue::get(Order), getContext().IntTy);
@@ -1106,7 +1107,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
     case AtomicExpr::AO__scoped_atomic_exchange:
     case AtomicExpr::AO__scoped_atomic_exchange_n:
       LibCallName = "__atomic_exchange";
-      Args.add(RValue::get(CastToGenericAddrSpace(Val1.getPointer(),
+      Args.add(RValue::get(CastToGenericAddrSpace(Val1.emitRawPointer(*this),
                                                   E->getVal1()->getType())),
                getContext().VoidPtrTy);
       break;
@@ -1121,7 +1122,7 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
       LibCallName = "__atomic_store";
       RetTy = getContext().VoidTy;
       HaveRetTy = true;
-      Args.add(RValue::get(CastToGenericAddrSpace(Val1.getPointer(),
+      Args.add(RValue::get(CastToGenericAddrSpace(Val1.emitRawPointer(*this),
                                                   E->getVal1()->getType())),
                getContext().VoidPtrTy);
       break;
@@ -1200,7 +1201,8 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
     if (!HaveRetTy) {
       // Value is returned through parameter before the order.
       RetTy = getContext().VoidTy;
-      Args.add(RValue::get(CastToGenericAddrSpace(Dest.getPointer(), RetTy)),
+      Args.add(RValue::get(
+                   CastToGenericAddrSpace(Dest.emitRawPointer(*this), RetTy)),
                getContext().VoidPtrTy);
     }
     // Order is always the last parameter.
@@ -1526,7 +1528,7 @@ RValue AtomicInfo::EmitAtomicLoad(AggValueSlot ResultSlot, SourceLocation Loc,
     } else
       TempAddr = CreateTempAlloca();
 
-    EmitAtomicLoadLibcall(TempAddr.getPointer(), AO, IsVolatile);
+    EmitAtomicLoadLibcall(TempAddr.emitRawPointer(CGF), AO, IsVolatile);
 
     // Okay, turn that back into the original value or whole atomic (for
     // non-simple lvalues) type.
@@ -1682,9 +1684,9 @@ std::pair<RValue, llvm::Value *> AtomicInfo::EmitAtomicCompareExchange(
   if (shouldUseLibcall()) {
     // Produce a source address.
     Address ExpectedAddr = materializeRValue(Expected);
-    Address DesiredAddr = materializeRValue(Desired);
-    auto *Res = EmitAtomicCompareExchangeLibcall(ExpectedAddr.getPointer(),
-                                                 DesiredAddr.getPointer(),
+    llvm::Value *ExpectedPtr = ExpectedAddr.emitRawPointer(CGF);
+    llvm::Value *DesiredPtr = materializeRValue(Desired).emitRawPointer(CGF);
+    auto *Res = EmitAtomicCompareExchangeLibcall(ExpectedPtr, DesiredPtr,
                                                  Success, Failure);
     return std::make_pair(
         convertAtomicTempToRValue(ExpectedAddr, AggValueSlot::ignored(),
@@ -1767,7 +1769,7 @@ void AtomicInfo::EmitAtomicUpdateLibcall(
 
   Address ExpectedAddr = CreateTempAlloca();
 
-  EmitAtomicLoadLibcall(ExpectedAddr.getPointer(), AO, IsVolatile);
+  EmitAtomicLoadLibcall(ExpectedAddr.emitRawPointer(CGF), AO, IsVolatile);
   auto *ContBB = CGF.createBasicBlock("atomic_cont");
   auto *ExitBB = CGF.createBasicBlock("atomic_exit");
   CGF.EmitBlock(ContBB);
@@ -1781,10 +1783,10 @@ void AtomicInfo::EmitAtomicUpdateLibcall(
                                            AggValueSlot::ignored(),
                                            SourceLocation(), /*AsValue=*/false);
   EmitAtomicUpdateValue(CGF, *this, OldRVal, UpdateOp, DesiredAddr);
+  llvm::Value *ExpectedPtr = ExpectedAddr.emitRawPointer(CGF);
+  llvm::Value *DesiredPtr = DesiredAddr.emitRawPointer(CGF);
   auto *Res =
-      EmitAtomicCompareExchangeLibcall(ExpectedAddr.getPointer(),
-                                       DesiredAddr.getPointer(),
-                                       AO, Failure);
+      EmitAtomicCompareExchangeLibcall(ExpectedPtr, DesiredPtr, AO, Failure);
   CGF.Builder.CreateCondBr(Res, ExitBB, ContBB);
   CGF.EmitBlock(ExitBB, /*IsFinished=*/true);
 }
@@ -1858,7 +1860,7 @@ void AtomicInfo::EmitAtomicUpdateLibcall(llvm::AtomicOrdering AO,
 
   Address ExpectedAddr = CreateTempAlloca();
 
-  EmitAtomicLoadLibcall(ExpectedAddr.getPointer(), AO, IsVolatile);
+  EmitAtomicLoadLibcall(ExpectedAddr.emitRawPointer(CGF), AO, IsVolatile);
   auto *ContBB = CGF.createBasicBlock("atomic_cont");
   auto *ExitBB = CGF.createBasicBlock("atomic_exit");
   CGF.EmitBlock(ContBB);
@@ -1869,10 +1871,10 @@ void AtomicInfo::EmitAtomicUpdateLibcall(llvm::AtomicOrdering AO,
     CGF.Builder.CreateStore(OldVal, DesiredAddr);
   }
   EmitAtomicUpdateValue(CGF, *this, UpdateRVal, DesiredAddr);
+  llvm::Value *ExpectedPtr = ExpectedAddr.emitRawPointer(CGF);
+  llvm::Value *DesiredPtr = DesiredAddr.emitRawPointer(CGF);
   auto *Res =
-      EmitAtomicCompareExchangeLibcall(ExpectedAddr.getPointer(),
-                                       DesiredAddr.getPointer(),
-                                       AO, Failure);
+      EmitAtomicCompareExchangeLibcall(ExpectedPtr, DesiredPtr, AO, Failure);
   CGF.Builder.CreateCondBr(Res, ExitBB, ContBB);
   CGF.EmitBlock(ExitBB, /*IsFinished=*/true);
 }
@@ -1972,7 +1974,8 @@ void CodeGenFunction::EmitAtomicStore(RValue rvalue, LValue dest,
       args.add(RValue::get(atomics.getAtomicSizeValue()),
                getContext().getSizeType());
       args.add(RValue::get(atomics.getAtomicPointer()), getContext().VoidPtrTy);
-      args.add(RValue::get(srcAddr.getPointer()), getContext().VoidPtrTy);
+      args.add(RValue::get(srcAddr.emitRawPointer(*this)),
+               getContext().VoidPtrTy);
       args.add(
           RValue::get(llvm::ConstantInt::get(IntTy, (int)llvm::toCABI(AO))),
           getContext().IntTy);
diff --git a/clang/lib/CodeGen/CGBlocks.cpp b/clang/lib/CodeGen/CGBlocks.cpp
index 52f2106849ba0..8cace298fde93 100644
--- a/clang/lib/CodeGen/CGBlocks.cpp
+++ b/clang/lib/CodeGen/CGBlocks.cpp
@@ -36,7 +36,8 @@ CGBlockInfo::CGBlockInfo(const BlockDecl *block, StringRef name)
     : Name(name), CXXThisIndex(0), CanBeGlobal(false), NeedsCopyDispose(false),
       NoEscape(false), HasCXXObject(false), UsesStret(false),
       HasCapturedVariableLayout(false), CapturesNonExternalType(false),
-      LocalAddress(Address::invalid()), StructureType(nullptr), Block(block) {
+      LocalAddress(RawAddress::invalid()), StructureType(nullptr),
+      Block(block) {
 
   // Skip asm prefix, if any.  'name' is usually taken directly from
   // the mangled name of the enclosing function.
@@ -794,7 +795,7 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
 
   // Otherwise, we have to emit this as a local block.
 
-  Address blockAddr = blockInfo.LocalAddress;
+  RawAddress blockAddr = blockInfo.LocalAddress;
   assert(blockAddr.isValid() && "block has no address!");
 
   llvm::Constant *isa;
@@ -939,7 +940,7 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
       if (CI.isNested())
         byrefPointer = Builder.CreateLoad(src, "byref.capture");
       else
-        byrefPointer = src.getPointer();
+        byrefPointer = src.emitRawPointer(*this);
 
       // Write that void* into the capture field.
       Builder.CreateStore(byrefPointer, blockField);
@@ -962,9 +963,9 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
 
     // If it's a reference variable, copy the reference into the block field.
     } else if (type->getAs<ReferenceType>()) {
-      Builder.CreateStore(src.getPointer(), blockField);
+      Builder.CreateStore(src.emitRawPointer(*this), blockField);
 
-    // If type is const-qualified, copy the value into the block field.
+      // If type is const-qualified, copy the value into the block field.
     } else if (type.isConstQualified() &&
                type.getObjCLifetime() == Qualifiers::OCL_Strong &&
                CGM.getCodeGenOpts().OptimizationLevel != 0) {
@@ -1378,7 +1379,7 @@ void CodeGenFunction::setBlockContextParameter(const ImplicitParamDecl *D,
 
   // Allocate a stack slot like for any local variable to guarantee optimal
   // debug info at -O0. The mem2reg pass will eliminate it when optimizing.
-  Address alloc = CreateMemTemp(D->getType(), D->getName() + ".addr");
+  RawAddress alloc = CreateMemTemp(D->getType(), D->getName() + ".addr");
   Builder.CreateStore(arg, alloc);
   if (CGDebugInfo *DI = getDebugInfo()) {
     if (CGM.getCodeGenOpts().hasReducedDebugInfo()) {
@@ -1498,7 +1499,7 @@ llvm::Function *CodeGenFunction::GenerateBlockFunction(
     // frame setup instruction by llvm::DwarfDebug::beginFunction().
     auto NL = ApplyDebugLocation::CreateEmpty(*this);
     Builder.CreateStore(BlockPointer, Alloca);
-    BlockPointerDbgLoc = Alloca.getPointer();
+    BlockPointerDbgLoc = Alloca.emitRawPointer(*this);
   }
 
   // If we have a C++ 'this' reference, go ahead and force it into
@@ -1558,8 +1559,8 @@ llvm::Function *CodeGenFunction::GenerateBlockFunction(
         const CGBlockInfo::Capture &capture = blockInfo.getCapture(variable);
         if (capture.isConstant()) {
           auto addr = LocalDeclMap.find(variable)->second;
-          (void)DI->EmitDeclareOfAutoVariable(variable, addr.getPointer(),
-                                              Builder);
+          (void)DI->EmitDeclareOfAutoVariable(
+              variable, addr.emitRawPointer(*this), Builder);
           continue;
         }
 
@@ -1663,7 +1664,7 @@ struct CallBlockRelease final : EHScopeStack::Cleanup {
     if (LoadBlockVarAddr) {
       BlockVarAddr = CGF.Builder.CreateLoad(Addr);
     } else {
-      BlockVarAddr = Addr.getPointer();
+      BlockVarAddr = Addr.emitRawPointer(CGF);
     }
 
     CGF.BuildBlockRelease(BlockVarAddr, FieldFlags, CanThrow);
@@ -1963,13 +1964,15 @@ CodeGenFunction::GenerateCopyHelperFunction(const CGBlockInfo &blockInfo) {
         // it. It's not quite worth the annoyance to avoid creating it in the
         // first place.
         if (!needsEHCleanup(captureType.isDestructedType()))
-          cast<llvm::Instruction>(dstField.getPointer())->eraseFromParent();
+          if (auto *I =
+                  cast_or_null<llvm::Instruction>(dstField.getBasePointer()))
+            I->eraseFromParent();
       }
       break;
     }
     case BlockCaptureEntityKind::BlockObject: {
       llvm::Value *srcValue = Builder.CreateLoad(srcField, "blockcopy.src");
-      llvm::Value *dstAddr = dstField.getPointer();
+      llvm::Value *dstAddr = dstField.emitRawPointer(*this);
       llvm::Value *args[] = {
         dstAddr, srcValue, llvm::ConstantInt::get(Int32Ty, flags.getBitMask())
       };
@@ -2140,7 +2143,7 @@ class ObjectByrefHelpers final : public BlockByrefHelpers {
     llvm::Value *flagsVal = llvm::ConstantInt::get(CGF.Int32Ty, flags);
     llvm::FunctionCallee fn = CGF.CGM.getBlockObjectAssign();
 
-    llvm::Value *args[] = { destField.getPointer(), srcValue, flagsVal };
+    llvm::Value *args[] = {destField.emitRawPointer(CGF), srcValue, flagsVal};
     CGF.EmitNounwindRuntimeCall(fn, args);
   }
 
@@ -2697,7 +2700,8 @@ void CodeGenFunction::emitByrefStructureInit(const AutoVarEmission &emission) {
   storeHeaderField(V, getPointerSize(), "byref.isa");
 
   // Store the address of the variable into its own forwarding pointer.
-  storeHeaderField(addr.getPointer(), getPointerSize(), "byref.forwarding");
+  storeHeaderField(addr.emitRawPointer(*this), getPointerSize(),
+                   "byref.forwarding");
 
   // Blocks ABI:
   //   c) the flags field is set to either 0 if no helper functions are
diff --git a/clang/lib/CodeGen/CGBlocks.h b/clang/lib/CodeGen/CGBlocks.h
index 4ef1ae9f33655..8d10c4f69b202 100644
--- a/clang/lib/CodeGen/CGBlocks.h
+++ b/clang/lib/CodeGen/CGBlocks.h
@@ -271,7 +271,8 @@ class CGBlockInfo {
   /// The block's captures. Non-constant captures are sorted by their offsets.
   llvm::SmallVector<Capture, 4> SortedCaptures;
 
-  Address LocalAddress;
+  // Currently we assume that block-pointer types are never signed.
+  RawAddress LocalAddress;
   llvm::StructType *StructureType;
   const BlockDecl *Block;
   const BlockExpr *BlockExpression;
diff --git a/clang/lib/CodeGen/CGBuilder.h b/clang/lib/CodeGen/CGBuilder.h
index bf5ab171d720d..6dd9da7c4cade 100644
--- a/clang/lib/CodeGen/CGBuilder.h
+++ b/clang/lib/CodeGen/CGBuilder.h
@@ -10,7 +10,9 @@
 #define LLVM_CLANG_LIB_CODEGEN_CGBUILDER_H
 
 #include "Address.h"
+#include "CGValue.h"
 #include "CodeGenTypeCache.h"
+#include "llvm/Analysis/Utils/Local.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Type.h"
@@ -18,12 +20,15 @@
 namespace clang {
 namespace CodeGen {
 
+class CGBuilderTy;
 class CodeGenFunction;
 
 /// This is an IRBuilder insertion helper that forwards to
 /// CodeGenFunction::InsertHelper, which adds necessary metadata to
 /// instructions.
 class CGBuilderInserter final : public llvm::IRBuilderDefaultInserter {
+  friend CGBuilderTy;
+
 public:
   CGBuilderInserter() = default;
   explicit CGBuilderInserter(CodeGenFunction *CGF) : CGF(CGF) {}
@@ -43,10 +48,42 @@ typedef llvm::IRBuilder<llvm::ConstantFolder, CGBuilderInserterTy>
     CGBuilderBaseTy;
 
 class CGBuilderTy : public CGBuilderBaseTy {
+  friend class Address;
+
   /// Storing a reference to the type cache here makes it a lot easier
   /// to build natural-feeling, target-specific IR.
   const CodeGenTypeCache &TypeCache;
 
+  CodeGenFunction *getCGF() const { return getInserter().CGF; }
+
+  llvm::Value *emitRawPointerFromAddress(Address Addr) const {
+    return Addr.getBasePointer();
+  }
+
+  template <bool IsInBounds>
+  Address createConstGEP2_32(Address Addr, unsigned Idx0, unsigned Idx1,
+                             const llvm::Twine &Name) {
+    const llvm::DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
+    llvm::GetElementPtrInst *GEP;
+    if (IsInBounds)
+      GEP = cast<llvm::GetElementPtrInst>(CreateConstInBoundsGEP2_32(
+          Addr.getElementType(), emitRawPointerFromAddress(Addr), Idx0, Idx1,
+          Name));
+    else
+      GEP = cast<llvm::GetElementPtrInst>(CreateConstGEP2_32(
+          Addr.getElementType(), emitRawPointerFromAddress(Addr), Idx0, Idx1,
+          Name));
+    llvm::APInt Offset(
+        DL.getIndexSizeInBits(Addr.getType()->getPointerAddressSpace()), 0,
+        /*isSigned=*/true);
+    if (!GEP->accumulateConstantOffset(DL, Offset))
+      llvm_unreachable("offset of GEP with constants is always computable");
+    return Address(GEP, GEP->getResultElementType(),
+                   Addr.getAlignment().alignmentAtOffset(
+                       CharUnits::fromQuantity(Offset.getSExtValue())),
+                   IsInBounds ? Addr.isKnownNonNull() : NotKnownNonNull);
+  }
+
 public:
   CGBuilderTy(const CodeGenTypeCache &TypeCache, llvm::LLVMContext &C)
       : CGBuilderBaseTy(C), TypeCache(TypeCache) {}
@@ -69,20 +106,22 @@ class CGBuilderTy : public CGBuilderBaseTy {
   // Note that we intentionally hide the CreateLoad APIs that don't
   // take an alignment.
   llvm::LoadInst *CreateLoad(Address Addr, const llvm::Twine &Name = "") {
-    return CreateAlignedLoad(Addr.getElementType(), Addr.getPointer(),
+    return CreateAlignedLoad(Addr.getElementType(),
+                             emitRawPointerFromAddress(Addr),
                              Addr.getAlignment().getAsAlign(), Name);
   }
   llvm::LoadInst *CreateLoad(Address Addr, const char *Name) {
     // This overload is required to prevent string literals from
     // ending up in the IsVolatile overload.
-    return CreateAlignedLoad(Addr.getElementType(), Addr.getPointer(),
+    return CreateAlignedLoad(Addr.getElementType(),
+                             emitRawPointerFromAddress(Addr),
                              Addr.getAlignment().getAsAlign(), Name);
   }
   llvm::LoadInst *CreateLoad(Address Addr, bool IsVolatile,
                              const llvm::Twine &Name = "") {
-    return CreateAlignedLoad(Addr.getElementType(), Addr.getPointer(),
-                             Addr.getAlignment().getAsAlign(), IsVolatile,
-                             Name);
+    return CreateAlignedLoad(
+        Addr.getElementType(), emitRawPointerFromAddress(Addr),
+        Addr.getAlignment().getAsAlign(), IsVolatile, Name);
   }
 
   using CGBuilderBaseTy::CreateAlignedLoad;
@@ -96,7 +135,7 @@ class CGBuilderTy : public CGBuilderBaseTy {
   // take an alignment.
   llvm::StoreInst *CreateStore(llvm::Value *Val, Address Addr,
                                bool IsVolatile = false) {
-    return CreateAlignedStore(Val, Addr.getPointer(),
+    return CreateAlignedStore(Val, emitRawPointerFromAddress(Addr),
                               Addr.getAlignment().getAsAlign(), IsVolatile);
   }
 
@@ -132,33 +171,41 @@ class CGBuilderTy : public CGBuilderBaseTy {
                       llvm::AtomicOrdering FailureOrdering,
                       llvm::SyncScope::ID SSID = llvm::SyncScope::System) {
     return CGBuilderBaseTy::CreateAtomicCmpXchg(
-        Addr.getPointer(), Cmp, New, Addr.getAlignment().getAsAlign(),
-        SuccessOrdering, FailureOrdering, SSID);
+        Addr.emitRawPointer(*getCGF()), Cmp, New,
+        Addr.getAlignment().getAsAlign(), SuccessOrdering, FailureOrdering,
+        SSID);
   }
 
   llvm::AtomicRMWInst *
   CreateAtomicRMW(llvm::AtomicRMWInst::BinOp Op, Address Addr, llvm::Value *Val,
                   llvm::AtomicOrdering Ordering,
                   llvm::SyncScope::ID SSID = llvm::SyncScope::System) {
-    return CGBuilderBaseTy::CreateAtomicRMW(Op, Addr.getPointer(), Val,
-                                            Addr.getAlignment().getAsAlign(),
-                                            Ordering, SSID);
+    return CGBuilderBaseTy::CreateAtomicRMW(
+        Op, Addr.emitRawPointer(*getCGF()), Val,
+        Addr.getAlignment().getAsAlign(), Ordering, SSID);
   }
 
   using CGBuilderBaseTy::CreateAddrSpaceCast;
   Address CreateAddrSpaceCast(Address Addr, llvm::Type *Ty,
+                              llvm::Type *ElementTy,
                               const llvm::Twine &Name = "") {
-    return Addr.withPointer(CreateAddrSpaceCast(Addr.getPointer(), Ty, Name),
-                            Addr.isKnownNonNull());
+    if (!Addr.hasOffset())
+      return Address(CreateAddrSpaceCast(Addr.getBasePointer(), Ty, Name),
+                     ElementTy, Addr.getAlignment(), nullptr,
+                     Addr.isKnownNonNull());
+    // Eagerly force a raw address if these is an offset.
+    return RawAddress(
+        CreateAddrSpaceCast(Addr.emitRawPointer(*getCGF()), Ty, Name),
+        ElementTy, Addr.getAlignment(), Addr.isKnownNonNull());
   }
 
   using CGBuilderBaseTy::CreatePointerBitCastOrAddrSpaceCast;
   Address CreatePointerBitCastOrAddrSpaceCast(Address Addr, llvm::Type *Ty,
                                               llvm::Type *ElementTy,
                                               const llvm::Twine &Name = "") {
-    llvm::Value *Ptr =
-        CreatePointerBitCastOrAddrSpaceCast(Addr.getPointer(), Ty, Name);
-    return Address(Ptr, ElementTy, Addr.getAlignment(), Addr.isKnownNonNull());
+    if (Addr.getType()->getAddressSpace() == Ty->getPointerAddressSpace())
+      return Addr.withElementType(ElementTy);
+    return CreateAddrSpaceCast(Addr, Ty, ElementTy, Name);
   }
 
   /// Given
@@ -176,10 +223,11 @@ class CGBuilderTy : public CGBuilderBaseTy {
     const llvm::StructLayout *Layout = DL.getStructLayout(ElTy);
     auto Offset = CharUnits::fromQuantity(Layout->getElementOffset(Index));
 
-    return Address(
-        CreateStructGEP(Addr.getElementType(), Addr.getPointer(), Index, Name),
-        ElTy->getElementType(Index),
-        Addr.getAlignment().alignmentAtOffset(Offset), Addr.isKnownNonNull());
+    return Address(CreateStructGEP(Addr.getElementType(), Addr.getBasePointer(),
+                                   Index, Name),
+                   ElTy->getElementType(Index),
+                   Addr.getAlignment().alignmentAtOffset(Offset),
+                   Addr.isKnownNonNull());
   }
 
   /// Given
@@ -198,7 +246,7 @@ class CGBuilderTy : public CGBuilderBaseTy {
         CharUnits::fromQuantity(DL.getTypeAllocSize(ElTy->getElementType()));
 
     return Address(
-        CreateInBoundsGEP(Addr.getElementType(), Addr.getPointer(),
+        CreateInBoundsGEP(Addr.getElementType(), Addr.getBasePointer(),
                           {getSize(CharUnits::Zero()), getSize(Index)}, Name),
         ElTy->getElementType(),
         Addr.getAlignment().alignmentAtOffset(Index * EltSize),
@@ -216,10 +264,10 @@ class CGBuilderTy : public CGBuilderBaseTy {
     const llvm::DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
     CharUnits EltSize = CharUnits::fromQuantity(DL.getTypeAllocSize(ElTy));
 
-    return Address(CreateInBoundsGEP(Addr.getElementType(), Addr.getPointer(),
-                                     getSize(Index), Name),
-                   ElTy, Addr.getAlignment().alignmentAtOffset(Index * EltSize),
-                   Addr.isKnownNonNull());
+    return Address(
+        CreateInBoundsGEP(ElTy, Addr.getBasePointer(), getSize(Index), Name),
+        ElTy, Addr.getAlignment().alignmentAtOffset(Index * EltSize),
+        Addr.isKnownNonNull());
   }
 
   /// Given
@@ -229,110 +277,133 @@ class CGBuilderTy : public CGBuilderBaseTy {
   /// where i64 is actually the target word size.
   Address CreateConstGEP(Address Addr, uint64_t Index,
                          const llvm::Twine &Name = "") {
+    llvm::Type *ElTy = Addr.getElementType();
     const llvm::DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
-    CharUnits EltSize =
-        CharUnits::fromQuantity(DL.getTypeAllocSize(Addr.getElementType()));
+    CharUnits EltSize = CharUnits::fromQuantity(DL.getTypeAllocSize(ElTy));
 
-    return Address(CreateGEP(Addr.getElementType(), Addr.getPointer(),
-                             getSize(Index), Name),
+    return Address(CreateGEP(ElTy, Addr.getBasePointer(), getSize(Index), Name),
                    Addr.getElementType(),
-                   Addr.getAlignment().alignmentAtOffset(Index * EltSize),
-                   NotKnownNonNull);
+                   Addr.getAlignment().alignmentAtOffset(Index * EltSize));
   }
 
   /// Create GEP with single dynamic index. The address alignment is reduced
   /// according to the element size.
   using CGBuilderBaseTy::CreateGEP;
-  Address CreateGEP(Address Addr, llvm::Value *Index,
+  Address CreateGEP(CodeGenFunction &CGF, Address Addr, llvm::Value *Index,
                     const llvm::Twine &Name = "") {
     const llvm::DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
     CharUnits EltSize =
         CharUnits::fromQuantity(DL.getTypeAllocSize(Addr.getElementType()));
 
     return Address(
-        CreateGEP(Addr.getElementType(), Addr.getPointer(), Index, Name),
+        CreateGEP(Addr.getElementType(), Addr.emitRawPointer(CGF), Index, Name),
         Addr.getElementType(),
-        Addr.getAlignment().alignmentOfArrayElement(EltSize), NotKnownNonNull);
+        Addr.getAlignment().alignmentOfArrayElement(EltSize));
   }
 
   /// Given a pointer to i8, adjust it by a given constant offset.
   Address CreateConstInBoundsByteGEP(Address Addr, CharUnits Offset,
                                      const llvm::Twine &Name = "") {
     assert(Addr.getElementType() == TypeCache.Int8Ty);
-    return Address(CreateInBoundsGEP(Addr.getElementType(), Addr.getPointer(),
-                                     getSize(Offset), Name),
-                   Addr.getElementType(),
-                   Addr.getAlignment().alignmentAtOffset(Offset),
-                   Addr.isKnownNonNull());
+    return Address(
+        CreateInBoundsGEP(Addr.getElementType(), Addr.getBasePointer(),
+                          getSize(Offset), Name),
+        Addr.getElementType(), Addr.getAlignment().alignmentAtOffset(Offset),
+        Addr.isKnownNonNull());
   }
+
   Address CreateConstByteGEP(Address Addr, CharUnits Offset,
                              const llvm::Twine &Name = "") {
     assert(Addr.getElementType() == TypeCache.Int8Ty);
-    return Address(CreateGEP(Addr.getElementType(), Addr.getPointer(),
+    return Address(CreateGEP(Addr.getElementType(), Addr.getBasePointer(),
                              getSize(Offset), Name),
                    Addr.getElementType(),
-                   Addr.getAlignment().alignmentAtOffset(Offset),
-                   NotKnownNonNull);
+                   Addr.getAlignment().alignmentAtOffset(Offset));
   }
 
   using CGBuilderBaseTy::CreateConstInBoundsGEP2_32;
   Address CreateConstInBoundsGEP2_32(Address Addr, unsigned Idx0, unsigned Idx1,
                                      const llvm::Twine &Name = "") {
-    const llvm::DataLayout &DL = BB->getParent()->getParent()->getDataLayout();
+    return createConstGEP2_32<true>(Addr, Idx0, Idx1, Name);
+  }
 
-    auto *GEP = cast<llvm::GetElementPtrInst>(CreateConstInBoundsGEP2_32(
-        Addr.getElementType(), Addr.getPointer(), Idx0, Idx1, Name));
-    llvm::APInt Offset(
-        DL.getIndexSizeInBits(Addr.getType()->getPointerAddressSpace()), 0,
-        /*isSigned=*/true);
-    if (!GEP->accumulateConstantOffset(DL, Offset))
-      llvm_unreachable("offset of GEP with constants is always computable");
-    return Address(GEP, GEP->getResultElementType(),
-                   Addr.getAlignment().alignmentAtOffset(
-                       CharUnits::fromQuantity(Offset.getSExtValue())),
-                   Addr.isKnownNonNull());
+  using CGBuilderBaseTy::CreateConstGEP2_32;
+  Address CreateConstGEP2_32(Address Addr, unsigned Idx0, unsigned Idx1,
+                             const llvm::Twine &Name = "") {
+    return createConstGEP2_32<false>(Addr, Idx0, Idx1, Name);
+  }
+
+  Address CreateGEP(Address Addr, ArrayRef<llvm::Value *> IdxList,
+                    llvm::Type *ElementType, CharUnits Align,
+                    const Twine &Name = "") {
+    llvm::Value *Ptr = emitRawPointerFromAddress(Addr);
+    return RawAddress(CreateGEP(Addr.getElementType(), Ptr, IdxList, Name),
+                      ElementType, Align);
+  }
+
+  using CGBuilderBaseTy::CreateInBoundsGEP;
+  Address CreateInBoundsGEP(Address Addr, ArrayRef<llvm::Value *> IdxList,
+                            llvm::Type *ElementType, CharUnits Align,
+                            const Twine &Name = "") {
+    return RawAddress(CreateInBoundsGEP(Addr.getElementType(),
+                                        emitRawPointerFromAddress(Addr),
+                                        IdxList, Name),
+                      ElementType, Align, Addr.isKnownNonNull());
+  }
+
+  using CGBuilderBaseTy::CreateIsNull;
+  llvm::Value *CreateIsNull(Address Addr, const Twine &Name = "") {
+    if (!Addr.hasOffset())
+      return CreateIsNull(Addr.getBasePointer(), Name);
+    // The pointer isn't null if Addr has an offset since offsets can always
+    // be applied inbound.
+    return llvm::ConstantInt::getFalse(Context);
   }
 
   using CGBuilderBaseTy::CreateMemCpy;
   llvm::CallInst *CreateMemCpy(Address Dest, Address Src, llvm::Value *Size,
                                bool IsVolatile = false) {
-    return CreateMemCpy(Dest.getPointer(), Dest.getAlignment().getAsAlign(),
-                        Src.getPointer(), Src.getAlignment().getAsAlign(), Size,
-                        IsVolatile);
+    llvm::Value *DestPtr = emitRawPointerFromAddress(Dest);
+    llvm::Value *SrcPtr = emitRawPointerFromAddress(Src);
+    return CreateMemCpy(DestPtr, Dest.getAlignment().getAsAlign(), SrcPtr,
+                        Src.getAlignment().getAsAlign(), Size, IsVolatile);
   }
   llvm::CallInst *CreateMemCpy(Address Dest, Address Src, uint64_t Size,
                                bool IsVolatile = false) {
-    return CreateMemCpy(Dest.getPointer(), Dest.getAlignment().getAsAlign(),
-                        Src.getPointer(), Src.getAlignment().getAsAlign(), Size,
-                        IsVolatile);
+    llvm::Value *DestPtr = emitRawPointerFromAddress(Dest);
+    llvm::Value *SrcPtr = emitRawPointerFromAddress(Src);
+    return CreateMemCpy(DestPtr, Dest.getAlignment().getAsAlign(), SrcPtr,
+                        Src.getAlignment().getAsAlign(), Size, IsVolatile);
   }
 
   using CGBuilderBaseTy::CreateMemCpyInline;
   llvm::CallInst *CreateMemCpyInline(Address Dest, Address Src, uint64_t Size) {
-    return CreateMemCpyInline(
-        Dest.getPointer(), Dest.getAlignment().getAsAlign(), Src.getPointer(),
-        Src.getAlignment().getAsAlign(), getInt64(Size));
+    llvm::Value *DestPtr = emitRawPointerFromAddress(Dest);
+    llvm::Value *SrcPtr = emitRawPointerFromAddress(Src);
+    return CreateMemCpyInline(DestPtr, Dest.getAlignment().getAsAlign(), SrcPtr,
+                              Src.getAlignment().getAsAlign(), getInt64(Size));
   }
 
   using CGBuilderBaseTy::CreateMemMove;
   llvm::CallInst *CreateMemMove(Address Dest, Address Src, llvm::Value *Size,
                                 bool IsVolatile = false) {
-    return CreateMemMove(Dest.getPointer(), Dest.getAlignment().getAsAlign(),
-                         Src.getPointer(), Src.getAlignment().getAsAlign(),
-                         Size, IsVolatile);
+    llvm::Value *DestPtr = emitRawPointerFromAddress(Dest);
+    llvm::Value *SrcPtr = emitRawPointerFromAddress(Src);
+    return CreateMemMove(DestPtr, Dest.getAlignment().getAsAlign(), SrcPtr,
+                         Src.getAlignment().getAsAlign(), Size, IsVolatile);
   }
 
   using CGBuilderBaseTy::CreateMemSet;
   llvm::CallInst *CreateMemSet(Address Dest, llvm::Value *Value,
                                llvm::Value *Size, bool IsVolatile = false) {
-    return CreateMemSet(Dest.getPointer(), Value, Size,
+    return CreateMemSet(emitRawPointerFromAddress(Dest), Value, Size,
                         Dest.getAlignment().getAsAlign(), IsVolatile);
   }
 
   using CGBuilderBaseTy::CreateMemSetInline;
   llvm::CallInst *CreateMemSetInline(Address Dest, llvm::Value *Value,
                                      uint64_t Size) {
-    return CreateMemSetInline(Dest.getPointer(),
+    return CreateMemSetInline(emitRawPointerFromAddress(Dest),
                               Dest.getAlignment().getAsAlign(), Value,
                               getInt64(Size));
   }
@@ -346,16 +417,31 @@ class CGBuilderTy : public CGBuilderBaseTy {
     const llvm::StructLayout *Layout = DL.getStructLayout(ElTy);
     auto Offset = CharUnits::fromQuantity(Layout->getElementOffset(Index));
 
-    return Address(CreatePreserveStructAccessIndex(ElTy, Addr.getPointer(),
-                                                   Index, FieldIndex, DbgInfo),
-                   ElTy->getElementType(Index),
-                   Addr.getAlignment().alignmentAtOffset(Offset));
+    return Address(
+        CreatePreserveStructAccessIndex(ElTy, emitRawPointerFromAddress(Addr),
+                                        Index, FieldIndex, DbgInfo),
+        ElTy->getElementType(Index),
+        Addr.getAlignment().alignmentAtOffset(Offset));
+  }
+
+  using CGBuilderBaseTy::CreatePreserveUnionAccessIndex;
+  Address CreatePreserveUnionAccessIndex(Address Addr, unsigned FieldIndex,
+                                         llvm::MDNode *DbgInfo) {
+    Addr.replaceBasePointer(CreatePreserveUnionAccessIndex(
+        Addr.getBasePointer(), FieldIndex, DbgInfo));
+    return Addr;
   }
 
   using CGBuilderBaseTy::CreateLaunderInvariantGroup;
   Address CreateLaunderInvariantGroup(Address Addr) {
-    return Addr.withPointer(CreateLaunderInvariantGroup(Addr.getPointer()),
-                            Addr.isKnownNonNull());
+    Addr.replaceBasePointer(CreateLaunderInvariantGroup(Addr.getBasePointer()));
+    return Addr;
+  }
+
+  using CGBuilderBaseTy::CreateStripInvariantGroup;
+  Address CreateStripInvariantGroup(Address Addr) {
+    Addr.replaceBasePointer(CreateStripInvariantGroup(Addr.getBasePointer()));
+    return Addr;
   }
 };
 
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 605285fc6c994..ce0f932b28962 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2279,9 +2279,9 @@ llvm::Function *CodeGenFunction::generateBuiltinOSLogHelperFunction(
   auto AL = ApplyDebugLocation::CreateArtificial(*this);
 
   CharUnits Offset;
-  Address BufAddr =
-      Address(Builder.CreateLoad(GetAddrOfLocalVar(Args[0]), "buf"), Int8Ty,
-              BufferAlignment);
+  Address BufAddr = makeNaturalAddressForPointer(
+      Builder.CreateLoad(GetAddrOfLocalVar(Args[0]), "buf"), Ctx.VoidTy,
+      BufferAlignment);
   Builder.CreateStore(Builder.getInt8(Layout.getSummaryByte()),
                       Builder.CreateConstByteGEP(BufAddr, Offset++, "summary"));
   Builder.CreateStore(Builder.getInt8(Layout.getNumArgsByte()),
@@ -2324,7 +2324,7 @@ RValue CodeGenFunction::emitBuiltinOSLogFormat(const CallExpr &E) {
 
   // Ignore argument 1, the format string. It is not currently used.
   CallArgList Args;
-  Args.add(RValue::get(BufAddr.getPointer()), Ctx.VoidPtrTy);
+  Args.add(RValue::get(BufAddr.emitRawPointer(*this)), Ctx.VoidPtrTy);
 
   for (const auto &Item : Layout.Items) {
     int Size = Item.getSizeByte();
@@ -2364,8 +2364,8 @@ RValue CodeGenFunction::emitBuiltinOSLogFormat(const CallExpr &E) {
         if (!isa<Constant>(ArgVal)) {
           CleanupKind Cleanup = getARCCleanupKind();
           QualType Ty = TheExpr->getType();
-          Address Alloca = Address::invalid();
-          Address Addr = CreateMemTemp(Ty, "os.log.arg", &Alloca);
+          RawAddress Alloca = RawAddress::invalid();
+          RawAddress Addr = CreateMemTemp(Ty, "os.log.arg", &Alloca);
           ArgVal = EmitARCRetain(Ty, ArgVal);
           Builder.CreateStore(ArgVal, Addr);
           pushLifetimeExtendedDestroy(Cleanup, Alloca, Ty,
@@ -2398,7 +2398,7 @@ RValue CodeGenFunction::emitBuiltinOSLogFormat(const CallExpr &E) {
   llvm::Function *F = CodeGenFunction(CGM).generateBuiltinOSLogHelperFunction(
       Layout, BufAddr.getAlignment());
   EmitCall(FI, CGCallee::forDirect(F), ReturnValueSlot(), Args);
-  return RValue::get(BufAddr.getPointer());
+  return RValue::get(BufAddr, *this);
 }
 
 static bool isSpecialUnsignedMultiplySignedResult(
@@ -3149,7 +3149,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   // Check NonnullAttribute/NullabilityArg and Alignment.
   auto EmitArgCheck = [&](TypeCheckKind Kind, Address A, const Expr *Arg,
                           unsigned ParmNum) {
-    Value *Val = A.getPointer();
+    Value *Val = A.emitRawPointer(*this);
     EmitNonNullArgCheck(RValue::get(Val), Arg->getType(), Arg->getExprLoc(), FD,
                         ParmNum);
 
@@ -3178,12 +3178,12 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   case Builtin::BI__builtin_va_end:
     EmitVAStartEnd(BuiltinID == Builtin::BI__va_start
                        ? EmitScalarExpr(E->getArg(0))
-                       : EmitVAListRef(E->getArg(0)).getPointer(),
+                       : EmitVAListRef(E->getArg(0)).emitRawPointer(*this),
                    BuiltinID != Builtin::BI__builtin_va_end);
     return RValue::get(nullptr);
   case Builtin::BI__builtin_va_copy: {
-    Value *DstPtr = EmitVAListRef(E->getArg(0)).getPointer();
-    Value *SrcPtr = EmitVAListRef(E->getArg(1)).getPointer();
+    Value *DstPtr = EmitVAListRef(E->getArg(0)).emitRawPointer(*this);
+    Value *SrcPtr = EmitVAListRef(E->getArg(1)).emitRawPointer(*this);
     Builder.CreateCall(CGM.getIntrinsic(Intrinsic::vacopy, {DstPtr->getType()}),
                        {DstPtr, SrcPtr});
     return RValue::get(nullptr);
@@ -4023,13 +4023,13 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     bool IsVolatile = PtrTy->getPointeeType().isVolatileQualified();
 
     Address Src = EmitPointerWithAlignment(E->getArg(0));
-    EmitNonNullArgCheck(RValue::get(Src.getPointer()), E->getArg(0)->getType(),
-                        E->getArg(0)->getExprLoc(), FD, 0);
+    EmitNonNullArgCheck(RValue::get(Src.emitRawPointer(*this)),
+                        E->getArg(0)->getType(), E->getArg(0)->getExprLoc(), FD,
+                        0);
     Value *Result = MB.CreateColumnMajorLoad(
-        Src.getElementType(), Src.getPointer(),
+        Src.getElementType(), Src.emitRawPointer(*this),
         Align(Src.getAlignment().getQuantity()), Stride, IsVolatile,
-        ResultTy->getNumRows(), ResultTy->getNumColumns(),
-        "matrix");
+        ResultTy->getNumRows(), ResultTy->getNumColumns(), "matrix");
     return RValue::get(Result);
   }
 
@@ -4044,11 +4044,13 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     assert(PtrTy && "arg1 must be of pointer type");
     bool IsVolatile = PtrTy->getPointeeType().isVolatileQualified();
 
-    EmitNonNullArgCheck(RValue::get(Dst.getPointer()), E->getArg(1)->getType(),
-                        E->getArg(1)->getExprLoc(), FD, 0);
+    EmitNonNullArgCheck(RValue::get(Dst.emitRawPointer(*this)),
+                        E->getArg(1)->getType(), E->getArg(1)->getExprLoc(), FD,
+                        0);
     Value *Result = MB.CreateColumnMajorStore(
-        Matrix, Dst.getPointer(), Align(Dst.getAlignment().getQuantity()),
-        Stride, IsVolatile, MatrixTy->getNumRows(), MatrixTy->getNumColumns());
+        Matrix, Dst.emitRawPointer(*this),
+        Align(Dst.getAlignment().getQuantity()), Stride, IsVolatile,
+        MatrixTy->getNumRows(), MatrixTy->getNumColumns());
     return RValue::get(Result);
   }
 
@@ -4207,7 +4209,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   case Builtin::BI__builtin_bzero: {
     Address Dest = EmitPointerWithAlignment(E->getArg(0));
     Value *SizeVal = EmitScalarExpr(E->getArg(1));
-    EmitNonNullArgCheck(RValue::get(Dest.getPointer()), E->getArg(0)->getType(),
+    EmitNonNullArgCheck(Dest, E->getArg(0)->getType(),
                         E->getArg(0)->getExprLoc(), FD, 0);
     Builder.CreateMemSet(Dest, Builder.getInt8(0), SizeVal, false);
     return RValue::get(nullptr);
@@ -4218,10 +4220,12 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Address Src = EmitPointerWithAlignment(E->getArg(0));
     Address Dest = EmitPointerWithAlignment(E->getArg(1));
     Value *SizeVal = EmitScalarExpr(E->getArg(2));
-    EmitNonNullArgCheck(RValue::get(Src.getPointer()), E->getArg(0)->getType(),
-                        E->getArg(0)->getExprLoc(), FD, 0);
-    EmitNonNullArgCheck(RValue::get(Dest.getPointer()), E->getArg(1)->getType(),
-                        E->getArg(1)->getExprLoc(), FD, 0);
+    EmitNonNullArgCheck(RValue::get(Src.emitRawPointer(*this)),
+                        E->getArg(0)->getType(), E->getArg(0)->getExprLoc(), FD,
+                        0);
+    EmitNonNullArgCheck(RValue::get(Dest.emitRawPointer(*this)),
+                        E->getArg(1)->getType(), E->getArg(1)->getExprLoc(), FD,
+                        0);
     Builder.CreateMemMove(Dest, Src, SizeVal, false);
     return RValue::get(nullptr);
   }
@@ -4238,10 +4242,10 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Builder.CreateMemCpy(Dest, Src, SizeVal, false);
     if (BuiltinID == Builtin::BImempcpy ||
         BuiltinID == Builtin::BI__builtin_mempcpy)
-      return RValue::get(Builder.CreateInBoundsGEP(Dest.getElementType(),
-                                                   Dest.getPointer(), SizeVal));
+      return RValue::get(Builder.CreateInBoundsGEP(
+          Dest.getElementType(), Dest.emitRawPointer(*this), SizeVal));
     else
-      return RValue::get(Dest.getPointer());
+      return RValue::get(Dest, *this);
   }
 
   case Builtin::BI__builtin_memcpy_inline: {
@@ -4273,7 +4277,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Address Src = EmitPointerWithAlignment(E->getArg(1));
     Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
     Builder.CreateMemCpy(Dest, Src, SizeVal, false);
-    return RValue::get(Dest.getPointer());
+    return RValue::get(Dest, *this);
   }
 
   case Builtin::BI__builtin_objc_memmove_collectable: {
@@ -4282,7 +4286,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Value *SizeVal = EmitScalarExpr(E->getArg(2));
     CGM.getObjCRuntime().EmitGCMemmoveCollectable(*this,
                                                   DestAddr, SrcAddr, SizeVal);
-    return RValue::get(DestAddr.getPointer());
+    return RValue::get(DestAddr, *this);
   }
 
   case Builtin::BI__builtin___memmove_chk: {
@@ -4299,7 +4303,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Address Src = EmitPointerWithAlignment(E->getArg(1));
     Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
     Builder.CreateMemMove(Dest, Src, SizeVal, false);
-    return RValue::get(Dest.getPointer());
+    return RValue::get(Dest, *this);
   }
 
   case Builtin::BImemmove:
@@ -4310,7 +4314,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     EmitArgCheck(TCK_Store, Dest, E->getArg(0), 0);
     EmitArgCheck(TCK_Load, Src, E->getArg(1), 1);
     Builder.CreateMemMove(Dest, Src, SizeVal, false);
-    return RValue::get(Dest.getPointer());
+    return RValue::get(Dest, *this);
   }
   case Builtin::BImemset:
   case Builtin::BI__builtin_memset: {
@@ -4318,10 +4322,10 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     Value *ByteVal = Builder.CreateTrunc(EmitScalarExpr(E->getArg(1)),
                                          Builder.getInt8Ty());
     Value *SizeVal = EmitScalarExpr(E->getArg(2));
-    EmitNonNullArgCheck(RValue::get(Dest.getPointer()), E->getArg(0)->getType(),
+    EmitNonNullArgCheck(Dest, E->getArg(0)->getType(),
                         E->getArg(0)->getExprLoc(), FD, 0);
     Builder.CreateMemSet(Dest, ByteVal, SizeVal, false);
-    return RValue::get(Dest.getPointer());
+    return RValue::get(Dest, *this);
   }
   case Builtin::BI__builtin_memset_inline: {
     Address Dest = EmitPointerWithAlignment(E->getArg(0));
@@ -4329,8 +4333,9 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
         Builder.CreateTrunc(EmitScalarExpr(E->getArg(1)), Builder.getInt8Ty());
     uint64_t Size =
         E->getArg(2)->EvaluateKnownConstInt(getContext()).getZExtValue();
-    EmitNonNullArgCheck(RValue::get(Dest.getPointer()), E->getArg(0)->getType(),
-                        E->getArg(0)->getExprLoc(), FD, 0);
+    EmitNonNullArgCheck(RValue::get(Dest.emitRawPointer(*this)),
+                        E->getArg(0)->getType(), E->getArg(0)->getExprLoc(), FD,
+                        0);
     Builder.CreateMemSetInline(Dest, ByteVal, Size);
     return RValue::get(nullptr);
   }
@@ -4349,7 +4354,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
                                          Builder.getInt8Ty());
     Value *SizeVal = llvm::ConstantInt::get(Builder.getContext(), Size);
     Builder.CreateMemSet(Dest, ByteVal, SizeVal, false);
-    return RValue::get(Dest.getPointer());
+    return RValue::get(Dest, *this);
   }
   case Builtin::BI__builtin_wmemchr: {
     // The MSVC runtime library does not provide a definition of wmemchr, so we
@@ -4571,14 +4576,14 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
 
     // Store the stack pointer to the setjmp buffer.
     Value *StackAddr = Builder.CreateStackSave();
-    assert(Buf.getPointer()->getType() == StackAddr->getType());
+    assert(Buf.emitRawPointer(*this)->getType() == StackAddr->getType());
 
     Address StackSaveSlot = Builder.CreateConstInBoundsGEP(Buf, 2);
     Builder.CreateStore(StackAddr, StackSaveSlot);
 
     // Call LLVM's EH setjmp, which is lightweight.
     Function *F = CGM.getIntrinsic(Intrinsic::eh_sjlj_setjmp);
-    return RValue::get(Builder.CreateCall(F, Buf.getPointer()));
+    return RValue::get(Builder.CreateCall(F, Buf.emitRawPointer(*this)));
   }
   case Builtin::BI__builtin_longjmp: {
     Value *Buf = EmitScalarExpr(E->getArg(0));
@@ -5751,7 +5756,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     llvm::Value *Queue = EmitScalarExpr(E->getArg(0));
     llvm::Value *Flags = EmitScalarExpr(E->getArg(1));
     LValue NDRangeL = EmitAggExprToLValue(E->getArg(2));
-    llvm::Value *Range = NDRangeL.getAddress(*this).getPointer();
+    llvm::Value *Range = NDRangeL.getAddress(*this).emitRawPointer(*this);
     llvm::Type *RangeTy = NDRangeL.getAddress(*this).getType();
 
     if (NumArgs == 4) {
@@ -5860,9 +5865,10 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
               getContext(), Expr::NPC_ValueDependentIsNotNull)) {
         EventWaitList = llvm::ConstantPointerNull::get(PtrTy);
       } else {
-        EventWaitList = E->getArg(4)->getType()->isArrayType()
-                        ? EmitArrayToPointerDecay(E->getArg(4)).getPointer()
-                        : EmitScalarExpr(E->getArg(4));
+        EventWaitList =
+            E->getArg(4)->getType()->isArrayType()
+                ? EmitArrayToPointerDecay(E->getArg(4)).emitRawPointer(*this)
+                : EmitScalarExpr(E->getArg(4));
         // Convert to generic address space.
         EventWaitList = Builder.CreatePointerCast(EventWaitList, PtrTy);
       }
@@ -5958,7 +5964,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     llvm::Type *GenericVoidPtrTy = Builder.getPtrTy(
         getContext().getTargetAddressSpace(LangAS::opencl_generic));
     LValue NDRangeL = EmitAggExprToLValue(E->getArg(0));
-    llvm::Value *NDRange = NDRangeL.getAddress(*this).getPointer();
+    llvm::Value *NDRange = NDRangeL.getAddress(*this).emitRawPointer(*this);
     auto Info =
         CGM.getOpenCLRuntime().emitOpenCLEnqueuedBlock(*this, E->getArg(1));
     Value *Kernel =
@@ -6042,7 +6048,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     auto PTy0 = FTy->getParamType(0);
     if (PTy0 != Arg0Val->getType()) {
       if (Arg0Ty->isArrayType())
-        Arg0Val = EmitArrayToPointerDecay(Arg0).getPointer();
+        Arg0Val = EmitArrayToPointerDecay(Arg0).emitRawPointer(*this);
       else
         Arg0Val = Builder.CreatePointerCast(Arg0Val, PTy0);
     }
@@ -6080,7 +6086,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
     auto PTy1 = FTy->getParamType(1);
     if (PTy1 != Arg1Val->getType()) {
       if (Arg1Ty->isArrayType())
-        Arg1Val = EmitArrayToPointerDecay(Arg1).getPointer();
+        Arg1Val = EmitArrayToPointerDecay(Arg1).emitRawPointer(*this);
       else
         Arg1Val = Builder.CreatePointerCast(Arg1Val, PTy1);
     }
@@ -6094,7 +6100,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   case Builtin::BI__builtin_ms_va_start:
   case Builtin::BI__builtin_ms_va_end:
     return RValue::get(
-        EmitVAStartEnd(EmitMSVAListRef(E->getArg(0)).getPointer(),
+        EmitVAStartEnd(EmitMSVAListRef(E->getArg(0)).emitRawPointer(*this),
                        BuiltinID == Builtin::BI__builtin_ms_va_start));
 
   case Builtin::BI__builtin_ms_va_copy: {
@@ -6145,8 +6151,8 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
   // If this is a predefined lib function (e.g. malloc), emit the call
   // using exactly the normal call path.
   if (getContext().BuiltinInfo.isPredefinedLibFunction(BuiltinID))
-    return emitLibraryCall(*this, FD, E,
-                      cast<llvm::Constant>(EmitScalarExpr(E->getCallee())));
+    return emitLibraryCall(
+        *this, FD, E, cast<llvm::Constant>(EmitScalarExpr(E->getCallee())));
 
   // Check that a call to a target specific builtin has the correct target
   // features.
@@ -6263,7 +6269,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
         return RValue::get(nullptr);
       return RValue::get(V);
     case TEK_Aggregate:
-      return RValue::getAggregate(ReturnValue.getValue(),
+      return RValue::getAggregate(ReturnValue.getAddress(),
                                   ReturnValue.isVolatile());
     case TEK_Complex:
       llvm_unreachable("No current target builtin returns complex");
@@ -9024,7 +9030,7 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
         // Get the alignment for the argument in addition to the value;
         // we'll use it later.
         PtrOp0 = EmitPointerWithAlignment(E->getArg(0));
-        Ops.push_back(PtrOp0.getPointer());
+        Ops.push_back(PtrOp0.emitRawPointer(*this));
         continue;
       }
     }
@@ -9051,7 +9057,7 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
         // Get the alignment for the argument in addition to the value;
         // we'll use it later.
         PtrOp1 = EmitPointerWithAlignment(E->getArg(1));
-        Ops.push_back(PtrOp1.getPointer());
+        Ops.push_back(PtrOp1.emitRawPointer(*this));
         continue;
       }
     }
@@ -9472,7 +9478,7 @@ Value *CodeGenFunction::EmitARMMVEBuiltinExpr(unsigned BuiltinID,
     if (ReturnValue.isNull())
       return MvecOut;
     else
-      return Builder.CreateStore(MvecOut, ReturnValue.getValue());
+      return Builder.CreateStore(MvecOut, ReturnValue.getAddress());
   }
 
   case CustomCodeGen::VST24: {
@@ -11658,7 +11664,7 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
         // Get the alignment for the argument in addition to the value;
         // we'll use it later.
         PtrOp0 = EmitPointerWithAlignment(E->getArg(0));
-        Ops.push_back(PtrOp0.getPointer());
+        Ops.push_back(PtrOp0.emitRawPointer(*this));
         continue;
       }
     }
@@ -13524,15 +13530,15 @@ Value *CodeGenFunction::EmitBPFBuiltinExpr(unsigned BuiltinID,
     if (!getDebugInfo()) {
       CGM.Error(E->getExprLoc(),
                 "using __builtin_preserve_field_info() without -g");
-      return IsBitField ? EmitLValue(Arg).getBitFieldPointer()
-                        : EmitLValue(Arg).getPointer(*this);
+      return IsBitField ? EmitLValue(Arg).getRawBitFieldPointer(*this)
+                        : EmitLValue(Arg).emitRawPointer(*this);
     }
 
     // Enable underlying preserve_*_access_index() generation.
     bool OldIsInPreservedAIRegion = IsInPreservedAIRegion;
     IsInPreservedAIRegion = true;
-    Value *FieldAddr = IsBitField ? EmitLValue(Arg).getBitFieldPointer()
-                                  : EmitLValue(Arg).getPointer(*this);
+    Value *FieldAddr = IsBitField ? EmitLValue(Arg).getRawBitFieldPointer(*this)
+                                  : EmitLValue(Arg).emitRawPointer(*this);
     IsInPreservedAIRegion = OldIsInPreservedAIRegion;
 
     ConstantInt *C = cast<ConstantInt>(EmitScalarExpr(E->getArg(1)));
@@ -14526,14 +14532,14 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
   }
   case X86::BI_mm_setcsr:
   case X86::BI__builtin_ia32_ldmxcsr: {
-    Address Tmp = CreateMemTemp(E->getArg(0)->getType());
+    RawAddress Tmp = CreateMemTemp(E->getArg(0)->getType());
     Builder.CreateStore(Ops[0], Tmp);
     return Builder.CreateCall(CGM.getIntrinsic(Intrinsic::x86_sse_ldmxcsr),
                               Tmp.getPointer());
   }
   case X86::BI_mm_getcsr:
   case X86::BI__builtin_ia32_stmxcsr: {
-    Address Tmp = CreateMemTemp(E->getType());
+    RawAddress Tmp = CreateMemTemp(E->getType());
     Builder.CreateCall(CGM.getIntrinsic(Intrinsic::x86_sse_stmxcsr),
                        Tmp.getPointer());
     return Builder.CreateLoad(Tmp, "stmxcsr");
@@ -17819,7 +17825,8 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
     SmallVector<Value *, 4> Ops;
     for (unsigned i = 0, e = E->getNumArgs(); i != e; i++)
       if (E->getArg(i)->getType()->isArrayType())
-        Ops.push_back(EmitArrayToPointerDecay(E->getArg(i)).getPointer());
+        Ops.push_back(
+            EmitArrayToPointerDecay(E->getArg(i)).emitRawPointer(*this));
       else
         Ops.push_back(EmitScalarExpr(E->getArg(i)));
     // The first argument of these two builtins is a pointer used to store their
@@ -22305,14 +22312,14 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     // Save returned values.
     assert(II.NumResults);
     if (II.NumResults == 1) {
-      Builder.CreateAlignedStore(Result, Dst.getPointer(),
+      Builder.CreateAlignedStore(Result, Dst.emitRawPointer(*this),
                                  CharUnits::fromQuantity(4));
     } else {
       for (unsigned i = 0; i < II.NumResults; ++i) {
         Builder.CreateAlignedStore(
             Builder.CreateBitCast(Builder.CreateExtractValue(Result, i),
                                   Dst.getElementType()),
-            Builder.CreateGEP(Dst.getElementType(), Dst.getPointer(),
+            Builder.CreateGEP(Dst.getElementType(), Dst.emitRawPointer(*this),
                               llvm::ConstantInt::get(IntTy, i)),
             CharUnits::fromQuantity(4));
       }
@@ -22352,7 +22359,7 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     for (unsigned i = 0; i < II.NumResults; ++i) {
       Value *V = Builder.CreateAlignedLoad(
           Src.getElementType(),
-          Builder.CreateGEP(Src.getElementType(), Src.getPointer(),
+          Builder.CreateGEP(Src.getElementType(), Src.emitRawPointer(*this),
                             llvm::ConstantInt::get(IntTy, i)),
           CharUnits::fromQuantity(4));
       Values.push_back(Builder.CreateBitCast(V, ParamType));
@@ -22424,7 +22431,7 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     for (unsigned i = 0; i < MI.NumEltsA; ++i) {
       Value *V = Builder.CreateAlignedLoad(
           SrcA.getElementType(),
-          Builder.CreateGEP(SrcA.getElementType(), SrcA.getPointer(),
+          Builder.CreateGEP(SrcA.getElementType(), SrcA.emitRawPointer(*this),
                             llvm::ConstantInt::get(IntTy, i)),
           CharUnits::fromQuantity(4));
       Values.push_back(Builder.CreateBitCast(V, AType));
@@ -22434,7 +22441,7 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     for (unsigned i = 0; i < MI.NumEltsB; ++i) {
       Value *V = Builder.CreateAlignedLoad(
           SrcB.getElementType(),
-          Builder.CreateGEP(SrcB.getElementType(), SrcB.getPointer(),
+          Builder.CreateGEP(SrcB.getElementType(), SrcB.emitRawPointer(*this),
                             llvm::ConstantInt::get(IntTy, i)),
           CharUnits::fromQuantity(4));
       Values.push_back(Builder.CreateBitCast(V, BType));
@@ -22445,7 +22452,7 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     for (unsigned i = 0; i < MI.NumEltsC; ++i) {
       Value *V = Builder.CreateAlignedLoad(
           SrcC.getElementType(),
-          Builder.CreateGEP(SrcC.getElementType(), SrcC.getPointer(),
+          Builder.CreateGEP(SrcC.getElementType(), SrcC.emitRawPointer(*this),
                             llvm::ConstantInt::get(IntTy, i)),
           CharUnits::fromQuantity(4));
       Values.push_back(Builder.CreateBitCast(V, CType));
@@ -22455,7 +22462,7 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     for (unsigned i = 0; i < MI.NumEltsD; ++i)
       Builder.CreateAlignedStore(
           Builder.CreateBitCast(Builder.CreateExtractValue(Result, i), DType),
-          Builder.CreateGEP(Dst.getElementType(), Dst.getPointer(),
+          Builder.CreateGEP(Dst.getElementType(), Dst.emitRawPointer(*this),
                             llvm::ConstantInt::get(IntTy, i)),
           CharUnits::fromQuantity(4));
     return Result;
@@ -22713,7 +22720,7 @@ struct BuiltinAlignArgs {
   BuiltinAlignArgs(const CallExpr *E, CodeGenFunction &CGF) {
     QualType AstType = E->getArg(0)->getType();
     if (AstType->isArrayType())
-      Src = CGF.EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+      Src = CGF.EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(CGF);
     else
       Src = CGF.EmitScalarExpr(E->getArg(0));
     SrcType = Src->getType();
@@ -23331,7 +23338,7 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
   }
   case WebAssembly::BI__builtin_wasm_table_get: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
     Value *Index = EmitScalarExpr(E->getArg(1));
     Function *Callee;
     if (E->getType().isWebAssemblyExternrefType())
@@ -23345,7 +23352,7 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
   }
   case WebAssembly::BI__builtin_wasm_table_set: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
     Value *Index = EmitScalarExpr(E->getArg(1));
     Value *Val = EmitScalarExpr(E->getArg(2));
     Function *Callee;
@@ -23360,13 +23367,13 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
   }
   case WebAssembly::BI__builtin_wasm_table_size: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *Value = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+    Value *Value = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
     Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_table_size);
     return Builder.CreateCall(Callee, Value);
   }
   case WebAssembly::BI__builtin_wasm_table_grow: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
     Value *Val = EmitScalarExpr(E->getArg(1));
     Value *NElems = EmitScalarExpr(E->getArg(2));
 
@@ -23383,7 +23390,7 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
   }
   case WebAssembly::BI__builtin_wasm_table_fill: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
+    Value *Table = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
     Value *Index = EmitScalarExpr(E->getArg(1));
     Value *Val = EmitScalarExpr(E->getArg(2));
     Value *NElems = EmitScalarExpr(E->getArg(3));
@@ -23401,8 +23408,8 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
   }
   case WebAssembly::BI__builtin_wasm_table_copy: {
     assert(E->getArg(0)->getType()->isArrayType());
-    Value *TableX = EmitArrayToPointerDecay(E->getArg(0)).getPointer();
-    Value *TableY = EmitArrayToPointerDecay(E->getArg(1)).getPointer();
+    Value *TableX = EmitArrayToPointerDecay(E->getArg(0)).emitRawPointer(*this);
+    Value *TableY = EmitArrayToPointerDecay(E->getArg(1)).emitRawPointer(*this);
     Value *DstIdx = EmitScalarExpr(E->getArg(2));
     Value *SrcIdx = EmitScalarExpr(E->getArg(3));
     Value *NElems = EmitScalarExpr(E->getArg(4));
@@ -23481,7 +23488,7 @@ Value *CodeGenFunction::EmitHexagonBuiltinExpr(unsigned BuiltinID,
   auto MakeCircOp = [this, E](unsigned IntID, bool IsLoad) {
     // The base pointer is passed by address, so it needs to be loaded.
     Address A = EmitPointerWithAlignment(E->getArg(0));
-    Address BP = Address(A.getPointer(), Int8PtrTy, A.getAlignment());
+    Address BP = Address(A.emitRawPointer(*this), Int8PtrTy, A.getAlignment());
     llvm::Value *Base = Builder.CreateLoad(BP);
     // The treatment of both loads and stores is the same: the arguments for
     // the builtin are the same as the arguments for the intrinsic.
@@ -23522,8 +23529,8 @@ Value *CodeGenFunction::EmitHexagonBuiltinExpr(unsigned BuiltinID,
     // EmitPointerWithAlignment and EmitScalarExpr evaluates the expression
     // per call.
     Address DestAddr = EmitPointerWithAlignment(E->getArg(1));
-    DestAddr = Address(DestAddr.getPointer(), Int8Ty, DestAddr.getAlignment());
-    llvm::Value *DestAddress = DestAddr.getPointer();
+    DestAddr = DestAddr.withElementType(Int8Ty);
+    llvm::Value *DestAddress = DestAddr.emitRawPointer(*this);
 
     // Operands are Base, Dest, Modifier.
     // The intrinsic format in LLVM IR is defined as
@@ -23574,8 +23581,8 @@ Value *CodeGenFunction::EmitHexagonBuiltinExpr(unsigned BuiltinID,
         {EmitScalarExpr(E->getArg(0)), EmitScalarExpr(E->getArg(1)), PredIn});
 
     llvm::Value *PredOut = Builder.CreateExtractValue(Result, 1);
-    Builder.CreateAlignedStore(Q2V(PredOut), PredAddr.getPointer(),
-        PredAddr.getAlignment());
+    Builder.CreateAlignedStore(Q2V(PredOut), PredAddr.emitRawPointer(*this),
+                               PredAddr.getAlignment());
     return Builder.CreateExtractValue(Result, 0);
   }
   // These are identical to the builtins above, except they don't consume
@@ -23593,8 +23600,8 @@ Value *CodeGenFunction::EmitHexagonBuiltinExpr(unsigned BuiltinID,
         {EmitScalarExpr(E->getArg(0)), EmitScalarExpr(E->getArg(1))});
 
     llvm::Value *PredOut = Builder.CreateExtractValue(Result, 1);
-    Builder.CreateAlignedStore(Q2V(PredOut), PredAddr.getPointer(),
-        PredAddr.getAlignment());
+    Builder.CreateAlignedStore(Q2V(PredOut), PredAddr.emitRawPointer(*this),
+                               PredAddr.getAlignment());
     return Builder.CreateExtractValue(Result, 0);
   }
 
diff --git a/clang/lib/CodeGen/CGCUDANV.cpp b/clang/lib/CodeGen/CGCUDANV.cpp
index 9b238982db5d5..9846aa85a1cf5 100644
--- a/clang/lib/CodeGen/CGCUDANV.cpp
+++ b/clang/lib/CodeGen/CGCUDANV.cpp
@@ -333,11 +333,11 @@ void CGNVCUDARuntime::emitDeviceStubBodyNew(CodeGenFunction &CGF,
       llvm::ConstantInt::get(SizeTy, std::max<size_t>(1, Args.size())));
   // Store pointers to the arguments in a locally allocated launch_args.
   for (unsigned i = 0; i < Args.size(); ++i) {
-    llvm::Value* VarPtr = CGF.GetAddrOfLocalVar(Args[i]).getPointer();
+    llvm::Value *VarPtr = CGF.GetAddrOfLocalVar(Args[i]).emitRawPointer(CGF);
     llvm::Value *VoidVarPtr = CGF.Builder.CreatePointerCast(VarPtr, PtrTy);
     CGF.Builder.CreateDefaultAlignedStore(
-        VoidVarPtr,
-        CGF.Builder.CreateConstGEP1_32(PtrTy, KernelArgs.getPointer(), i));
+        VoidVarPtr, CGF.Builder.CreateConstGEP1_32(
+                        PtrTy, KernelArgs.emitRawPointer(CGF), i));
   }
 
   llvm::BasicBlock *EndBlock = CGF.createBasicBlock("setup.end");
@@ -395,9 +395,10 @@ void CGNVCUDARuntime::emitDeviceStubBodyNew(CodeGenFunction &CGF,
                               /*isVarArg=*/false),
       addUnderscoredPrefixToName("PopCallConfiguration"));
 
-  CGF.EmitRuntimeCallOrInvoke(cudaPopConfigFn,
-                              {GridDim.getPointer(), BlockDim.getPointer(),
-                               ShmemSize.getPointer(), Stream.getPointer()});
+  CGF.EmitRuntimeCallOrInvoke(cudaPopConfigFn, {GridDim.emitRawPointer(CGF),
+                                                BlockDim.emitRawPointer(CGF),
+                                                ShmemSize.emitRawPointer(CGF),
+                                                Stream.emitRawPointer(CGF)});
 
   // Emit the call to cudaLaunch
   llvm::Value *Kernel =
@@ -407,7 +408,7 @@ void CGNVCUDARuntime::emitDeviceStubBodyNew(CodeGenFunction &CGF,
                        cudaLaunchKernelFD->getParamDecl(0)->getType());
   LaunchKernelArgs.add(RValue::getAggregate(GridDim), Dim3Ty);
   LaunchKernelArgs.add(RValue::getAggregate(BlockDim), Dim3Ty);
-  LaunchKernelArgs.add(RValue::get(KernelArgs.getPointer()),
+  LaunchKernelArgs.add(RValue::get(KernelArgs, CGF),
                        cudaLaunchKernelFD->getParamDecl(3)->getType());
   LaunchKernelArgs.add(RValue::get(CGF.Builder.CreateLoad(ShmemSize)),
                        cudaLaunchKernelFD->getParamDecl(4)->getType());
@@ -440,8 +441,8 @@ void CGNVCUDARuntime::emitDeviceStubBodyLegacy(CodeGenFunction &CGF,
     auto TInfo = CGM.getContext().getTypeInfoInChars(A->getType());
     Offset = Offset.alignTo(TInfo.Align);
     llvm::Value *Args[] = {
-        CGF.Builder.CreatePointerCast(CGF.GetAddrOfLocalVar(A).getPointer(),
-                                      PtrTy),
+        CGF.Builder.CreatePointerCast(
+            CGF.GetAddrOfLocalVar(A).emitRawPointer(CGF), PtrTy),
         llvm::ConstantInt::get(SizeTy, TInfo.Width.getQuantity()),
         llvm::ConstantInt::get(SizeTy, Offset.getQuantity()),
     };
diff --git a/clang/lib/CodeGen/CGCXXABI.cpp b/clang/lib/CodeGen/CGCXXABI.cpp
index a8bf57a277e90..7c6dfc3e59d8c 100644
--- a/clang/lib/CodeGen/CGCXXABI.cpp
+++ b/clang/lib/CodeGen/CGCXXABI.cpp
@@ -20,6 +20,12 @@ using namespace CodeGen;
 
 CGCXXABI::~CGCXXABI() { }
 
+Address CGCXXABI::getThisAddress(CodeGenFunction &CGF) {
+  return CGF.makeNaturalAddressForPointer(
+      CGF.CXXABIThisValue, CGF.CXXABIThisDecl->getType()->getPointeeType(),
+      CGF.CXXABIThisAlignment);
+}
+
 void CGCXXABI::ErrorUnsupportedABI(CodeGenFunction &CGF, StringRef S) {
   DiagnosticsEngine &Diags = CGF.CGM.getDiags();
   unsigned DiagID = Diags.getCustomDiagID(DiagnosticsEngine::Error,
@@ -44,8 +50,12 @@ CGCallee CGCXXABI::EmitLoadOfMemberFunctionPointer(
     llvm::Value *MemPtr, const MemberPointerType *MPT) {
   ErrorUnsupportedABI(CGF, "calls through member pointers");
 
-  ThisPtrForCall = This.getPointer();
-  const auto *FPT = MPT->getPointeeType()->castAs<FunctionProtoType>();
+  const auto *RD =
+      cast<CXXRecordDecl>(MPT->getClass()->castAs<RecordType>()->getDecl());
+  ThisPtrForCall =
+      CGF.getAsNaturalPointerTo(This, CGF.getContext().getRecordType(RD));
+  const FunctionProtoType *FPT =
+      MPT->getPointeeType()->getAs<FunctionProtoType>();
   llvm::Constant *FnPtr = llvm::Constant::getNullValue(
       llvm::PointerType::getUnqual(CGM.getLLVMContext()));
   return CGCallee::forDirect(FnPtr, FPT);
@@ -251,16 +261,15 @@ void CGCXXABI::ReadArrayCookie(CodeGenFunction &CGF, Address ptr,
 
   // If we don't need an array cookie, bail out early.
   if (!requiresArrayCookie(expr, eltTy)) {
-    allocPtr = ptr.getPointer();
+    allocPtr = ptr.emitRawPointer(CGF);
     numElements = nullptr;
     cookieSize = CharUnits::Zero();
     return;
   }
 
   cookieSize = getArrayCookieSizeImpl(eltTy);
-  Address allocAddr =
-    CGF.Builder.CreateConstInBoundsByteGEP(ptr, -cookieSize);
-  allocPtr = allocAddr.getPointer();
+  Address allocAddr = CGF.Builder.CreateConstInBoundsByteGEP(ptr, -cookieSize);
+  allocPtr = allocAddr.emitRawPointer(CGF);
   numElements = readArrayCookieImpl(CGF, allocAddr, cookieSize);
 }
 
diff --git a/clang/lib/CodeGen/CGCXXABI.h b/clang/lib/CodeGen/CGCXXABI.h
index ad1ad08d08568..c7eccbd0095a9 100644
--- a/clang/lib/CodeGen/CGCXXABI.h
+++ b/clang/lib/CodeGen/CGCXXABI.h
@@ -57,12 +57,8 @@ class CGCXXABI {
   llvm::Value *getThisValue(CodeGenFunction &CGF) {
     return CGF.CXXABIThisValue;
   }
-  Address getThisAddress(CodeGenFunction &CGF) {
-    return Address(
-        CGF.CXXABIThisValue,
-        CGF.ConvertTypeForMem(CGF.CXXABIThisDecl->getType()->getPointeeType()),
-        CGF.CXXABIThisAlignment);
-  }
+
+  Address getThisAddress(CodeGenFunction &CGF);
 
   /// Issue a diagnostic about unsupported features in the ABI.
   void ErrorUnsupportedABI(CodeGenFunction &CGF, StringRef S);
@@ -475,12 +471,6 @@ class CGCXXABI {
                                   BaseSubobject Base,
                                   const CXXRecordDecl *NearestVBase) = 0;
 
-  /// Get the address point of the vtable for the given base subobject while
-  /// building a constexpr.
-  virtual llvm::Constant *
-  getVTableAddressPointForConstExpr(BaseSubobject Base,
-                                    const CXXRecordDecl *VTableClass) = 0;
-
   /// Get the address of the vtable for the given record decl which should be
   /// used for the vptr at the given offset in RD.
   virtual llvm::GlobalVariable *getAddrOfVTable(const CXXRecordDecl *RD,
diff --git a/clang/lib/CodeGen/CGCall.cpp b/clang/lib/CodeGen/CGCall.cpp
index 08a9224a17ea5..65c088ef5082e 100644
--- a/clang/lib/CodeGen/CGCall.cpp
+++ b/clang/lib/CodeGen/CGCall.cpp
@@ -1046,15 +1046,9 @@ static void forConstantArrayExpansion(CodeGenFunction &CGF,
                                       ConstantArrayExpansion *CAE,
                                       Address BaseAddr,
                                       llvm::function_ref<void(Address)> Fn) {
-  CharUnits EltSize = CGF.getContext().getTypeSizeInChars(CAE->EltTy);
-  CharUnits EltAlign =
-    BaseAddr.getAlignment().alignmentOfArrayElement(EltSize);
-  llvm::Type *EltTy = CGF.ConvertTypeForMem(CAE->EltTy);
-
   for (int i = 0, n = CAE->NumElts; i < n; i++) {
-    llvm::Value *EltAddr = CGF.Builder.CreateConstGEP2_32(
-        BaseAddr.getElementType(), BaseAddr.getPointer(), 0, i);
-    Fn(Address(EltAddr, EltTy, EltAlign));
+    Address EltAddr = CGF.Builder.CreateConstGEP2_32(BaseAddr, 0, i);
+    Fn(EltAddr);
   }
 }
 
@@ -1169,9 +1163,10 @@ void CodeGenFunction::ExpandTypeToArgs(
 }
 
 /// Create a temporary allocation for the purposes of coercion.
-static Address CreateTempAllocaForCoercion(CodeGenFunction &CGF, llvm::Type *Ty,
-                                           CharUnits MinAlign,
-                                           const Twine &Name = "tmp") {
+static RawAddress CreateTempAllocaForCoercion(CodeGenFunction &CGF,
+                                              llvm::Type *Ty,
+                                              CharUnits MinAlign,
+                                              const Twine &Name = "tmp") {
   // Don't use an alignment that's worse than what LLVM would prefer.
   auto PrefAlign = CGF.CGM.getDataLayout().getPrefTypeAlign(Ty);
   CharUnits Align = std::max(MinAlign, CharUnits::fromQuantity(PrefAlign));
@@ -1341,11 +1336,11 @@ static llvm::Value *CreateCoercedLoad(Address Src, llvm::Type *Ty,
   }
 
   // Otherwise do coercion through memory. This is stupid, but simple.
-  Address Tmp =
+  RawAddress Tmp =
       CreateTempAllocaForCoercion(CGF, Ty, Src.getAlignment(), Src.getName());
   CGF.Builder.CreateMemCpy(
-      Tmp.getPointer(), Tmp.getAlignment().getAsAlign(), Src.getPointer(),
-      Src.getAlignment().getAsAlign(),
+      Tmp.getPointer(), Tmp.getAlignment().getAsAlign(),
+      Src.emitRawPointer(CGF), Src.getAlignment().getAsAlign(),
       llvm::ConstantInt::get(CGF.IntPtrTy, SrcSize.getKnownMinValue()));
   return CGF.Builder.CreateLoad(Tmp);
 }
@@ -1429,11 +1424,12 @@ static void CreateCoercedStore(llvm::Value *Src,
     //
     // FIXME: Assert that we aren't truncating non-padding bits when have access
     // to that information.
-    Address Tmp = CreateTempAllocaForCoercion(CGF, SrcTy, Dst.getAlignment());
+    RawAddress Tmp =
+        CreateTempAllocaForCoercion(CGF, SrcTy, Dst.getAlignment());
     CGF.Builder.CreateStore(Src, Tmp);
     CGF.Builder.CreateMemCpy(
-        Dst.getPointer(), Dst.getAlignment().getAsAlign(), Tmp.getPointer(),
-        Tmp.getAlignment().getAsAlign(),
+        Dst.emitRawPointer(CGF), Dst.getAlignment().getAsAlign(),
+        Tmp.getPointer(), Tmp.getAlignment().getAsAlign(),
         llvm::ConstantInt::get(CGF.IntPtrTy, DstSize.getFixedValue()));
   }
 }
@@ -3102,17 +3098,17 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
     case ABIArgInfo::Indirect:
     case ABIArgInfo::IndirectAliased: {
       assert(NumIRArgs == 1);
-      Address ParamAddr = Address(Fn->getArg(FirstIRArg), ConvertTypeForMem(Ty),
-                                  ArgI.getIndirectAlign(), KnownNonNull);
+      Address ParamAddr = makeNaturalAddressForPointer(
+          Fn->getArg(FirstIRArg), Ty, ArgI.getIndirectAlign(), false, nullptr,
+          nullptr, KnownNonNull);
 
       if (!hasScalarEvaluationKind(Ty)) {
         // Aggregates and complex variables are accessed by reference. All we
         // need to do is realign the value, if requested. Also, if the address
         // may be aliased, copy it to ensure that the parameter variable is
         // mutable and has a unique adress, as C requires.
-        Address V = ParamAddr;
         if (ArgI.getIndirectRealign() || ArgI.isIndirectAliased()) {
-          Address AlignedTemp = CreateMemTemp(Ty, "coerce");
+          RawAddress AlignedTemp = CreateMemTemp(Ty, "coerce");
 
           // Copy from the incoming argument pointer to the temporary with the
           // appropriate alignment.
@@ -3122,11 +3118,12 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
           CharUnits Size = getContext().getTypeSizeInChars(Ty);
           Builder.CreateMemCpy(
               AlignedTemp.getPointer(), AlignedTemp.getAlignment().getAsAlign(),
-              ParamAddr.getPointer(), ParamAddr.getAlignment().getAsAlign(),
+              ParamAddr.emitRawPointer(*this),
+              ParamAddr.getAlignment().getAsAlign(),
               llvm::ConstantInt::get(IntPtrTy, Size.getQuantity()));
-          V = AlignedTemp;
+          ParamAddr = AlignedTemp;
         }
-        ArgVals.push_back(ParamValue::forIndirect(V));
+        ArgVals.push_back(ParamValue::forIndirect(ParamAddr));
       } else {
         // Load scalar value from indirect argument.
         llvm::Value *V =
@@ -3244,10 +3241,10 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
               == ParameterABI::SwiftErrorResult) {
           QualType pointeeTy = Ty->getPointeeType();
           assert(pointeeTy->isPointerType());
-          Address temp =
-            CreateMemTemp(pointeeTy, getPointerAlign(), "swifterror.temp");
-          Address arg(V, ConvertTypeForMem(pointeeTy),
-                      getContext().getTypeAlignInChars(pointeeTy));
+          RawAddress temp =
+              CreateMemTemp(pointeeTy, getPointerAlign(), "swifterror.temp");
+          Address arg = makeNaturalAddressForPointer(
+              V, pointeeTy, getContext().getTypeAlignInChars(pointeeTy));
           llvm::Value *incomingErrorValue = Builder.CreateLoad(arg);
           Builder.CreateStore(incomingErrorValue, temp);
           V = temp.getPointer();
@@ -3583,7 +3580,7 @@ static llvm::Value *tryRemoveRetainOfSelf(CodeGenFunction &CGF,
   llvm::LoadInst *load =
     dyn_cast<llvm::LoadInst>(retainedValue->stripPointerCasts());
   if (!load || load->isAtomic() || load->isVolatile() ||
-      load->getPointerOperand() != CGF.GetAddrOfLocalVar(self).getPointer())
+      load->getPointerOperand() != CGF.GetAddrOfLocalVar(self).getBasePointer())
     return nullptr;
 
   // Okay!  Burn it all down.  This relies for correctness on the
@@ -3620,12 +3617,15 @@ static llvm::Value *emitAutoreleaseOfResult(CodeGenFunction &CGF,
 
 /// Heuristically search for a dominating store to the return-value slot.
 static llvm::StoreInst *findDominatingStoreToReturnValue(CodeGenFunction &CGF) {
+  llvm::Value *ReturnValuePtr = CGF.ReturnValue.getBasePointer();
+
   // Check if a User is a store which pointerOperand is the ReturnValue.
   // We are looking for stores to the ReturnValue, not for stores of the
   // ReturnValue to some other location.
-  auto GetStoreIfValid = [&CGF](llvm::User *U) -> llvm::StoreInst * {
+  auto GetStoreIfValid = [&CGF,
+                          ReturnValuePtr](llvm::User *U) -> llvm::StoreInst * {
     auto *SI = dyn_cast<llvm::StoreInst>(U);
-    if (!SI || SI->getPointerOperand() != CGF.ReturnValue.getPointer() ||
+    if (!SI || SI->getPointerOperand() != ReturnValuePtr ||
         SI->getValueOperand()->getType() != CGF.ReturnValue.getElementType())
       return nullptr;
     // These aren't actually possible for non-coerced returns, and we
@@ -3639,7 +3639,7 @@ static llvm::StoreInst *findDominatingStoreToReturnValue(CodeGenFunction &CGF) {
   // for something immediately preceding the IP.  Sometimes this can
   // happen with how we generate implicit-returns; it can also happen
   // with noreturn cleanups.
-  if (!CGF.ReturnValue.getPointer()->hasOneUse()) {
+  if (!ReturnValuePtr->hasOneUse()) {
     llvm::BasicBlock *IP = CGF.Builder.GetInsertBlock();
     if (IP->empty()) return nullptr;
 
@@ -3657,8 +3657,7 @@ static llvm::StoreInst *findDominatingStoreToReturnValue(CodeGenFunction &CGF) {
     return nullptr;
   }
 
-  llvm::StoreInst *store =
-      GetStoreIfValid(CGF.ReturnValue.getPointer()->user_back());
+  llvm::StoreInst *store = GetStoreIfValid(ReturnValuePtr->user_back());
   if (!store) return nullptr;
 
   // Now do a first-and-dirty dominance check: just walk up the
@@ -4206,7 +4205,7 @@ static bool isProvablyNull(llvm::Value *addr) {
 }
 
 static bool isProvablyNonNull(Address Addr, CodeGenFunction &CGF) {
-  return llvm::isKnownNonZero(Addr.getPointer(), CGF.CGM.getDataLayout());
+  return llvm::isKnownNonZero(Addr.getBasePointer(), CGF.CGM.getDataLayout());
 }
 
 /// Emit the actual writing-back of a writeback.
@@ -4214,21 +4213,20 @@ static void emitWriteback(CodeGenFunction &CGF,
                           const CallArgList::Writeback &writeback) {
   const LValue &srcLV = writeback.Source;
   Address srcAddr = srcLV.getAddress(CGF);
-  assert(!isProvablyNull(srcAddr.getPointer()) &&
+  assert(!isProvablyNull(srcAddr.getBasePointer()) &&
          "shouldn't have writeback for provably null argument");
 
   llvm::BasicBlock *contBB = nullptr;
 
   // If the argument wasn't provably non-null, we need to null check
   // before doing the store.
-  bool provablyNonNull = llvm::isKnownNonZero(srcAddr.getPointer(),
-                                              CGF.CGM.getDataLayout());
+  bool provablyNonNull = isProvablyNonNull(srcAddr, CGF);
+
   if (!provablyNonNull) {
     llvm::BasicBlock *writebackBB = CGF.createBasicBlock("icr.writeback");
     contBB = CGF.createBasicBlock("icr.done");
 
-    llvm::Value *isNull =
-      CGF.Builder.CreateIsNull(srcAddr.getPointer(), "icr.isnull");
+    llvm::Value *isNull = CGF.Builder.CreateIsNull(srcAddr, "icr.isnull");
     CGF.Builder.CreateCondBr(isNull, contBB, writebackBB);
     CGF.EmitBlock(writebackBB);
   }
@@ -4332,7 +4330,7 @@ static void emitWritebackArg(CodeGenFunction &CGF, CallArgList &args,
       CGF.ConvertTypeForMem(CRE->getType()->getPointeeType());
 
   // If the address is a constant null, just pass the appropriate null.
-  if (isProvablyNull(srcAddr.getPointer())) {
+  if (isProvablyNull(srcAddr.getBasePointer())) {
     args.add(RValue::get(llvm::ConstantPointerNull::get(destType)),
              CRE->getType());
     return;
@@ -4361,17 +4359,16 @@ static void emitWritebackArg(CodeGenFunction &CGF, CallArgList &args,
   // If the address is *not* known to be non-null, we need to switch.
   llvm::Value *finalArgument;
 
-  bool provablyNonNull = llvm::isKnownNonZero(srcAddr.getPointer(),
-                                              CGF.CGM.getDataLayout());
+  bool provablyNonNull = isProvablyNonNull(srcAddr, CGF);
+
   if (provablyNonNull) {
-    finalArgument = temp.getPointer();
+    finalArgument = temp.emitRawPointer(CGF);
   } else {
-    llvm::Value *isNull =
-      CGF.Builder.CreateIsNull(srcAddr.getPointer(), "icr.isnull");
+    llvm::Value *isNull = CGF.Builder.CreateIsNull(srcAddr, "icr.isnull");
 
-    finalArgument = CGF.Builder.CreateSelect(isNull,
-                                   llvm::ConstantPointerNull::get(destType),
-                                             temp.getPointer(), "icr.argument");
+    finalArgument = CGF.Builder.CreateSelect(
+        isNull, llvm::ConstantPointerNull::get(destType),
+        temp.emitRawPointer(CGF), "icr.argument");
 
     // If we need to copy, then the load has to be conditional, which
     // means we need control flow.
@@ -4496,6 +4493,16 @@ void CodeGenFunction::EmitNonNullArgCheck(RValue RV, QualType ArgType,
   EmitCheck(std::make_pair(Cond, CheckKind), Handler, StaticData, std::nullopt);
 }
 
+void CodeGenFunction::EmitNonNullArgCheck(Address Addr, QualType ArgType,
+                                          SourceLocation ArgLoc,
+                                          AbstractCallee AC, unsigned ParmNum) {
+  if (!AC.getDecl() || !(SanOpts.has(SanitizerKind::NonnullAttribute) ||
+                         SanOpts.has(SanitizerKind::NullabilityArg)))
+    return;
+
+  EmitNonNullArgCheck(RValue::get(Addr, *this), ArgType, ArgLoc, AC, ParmNum);
+}
+
 // Check if the call is going to use the inalloca convention. This needs to
 // agree with CGFunctionInfo::usesInAlloca. The CGFunctionInfo is arranged
 // later, so we can't check it directly.
@@ -4837,10 +4844,20 @@ CodeGenFunction::AddObjCARCExceptionMetadata(llvm::Instruction *Inst) {
 llvm::CallInst *
 CodeGenFunction::EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
                                          const llvm::Twine &name) {
-  return EmitNounwindRuntimeCall(callee, std::nullopt, name);
+  return EmitNounwindRuntimeCall(callee, ArrayRef<llvm::Value *>(), name);
 }
 
 /// Emits a call to the given nounwind runtime function.
+llvm::CallInst *
+CodeGenFunction::EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
+                                         ArrayRef<Address> args,
+                                         const llvm::Twine &name) {
+  SmallVector<llvm::Value *, 3> values;
+  for (auto arg : args)
+    values.push_back(arg.emitRawPointer(*this));
+  return EmitNounwindRuntimeCall(callee, values, name);
+}
+
 llvm::CallInst *
 CodeGenFunction::EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
                                          ArrayRef<llvm::Value *> args,
@@ -5119,7 +5136,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
   // If we're using inalloca, insert the allocation after the stack save.
   // FIXME: Do this earlier rather than hacking it in here!
-  Address ArgMemory = Address::invalid();
+  RawAddress ArgMemory = RawAddress::invalid();
   if (llvm::StructType *ArgStruct = CallInfo.getArgStruct()) {
     const llvm::DataLayout &DL = CGM.getDataLayout();
     llvm::Instruction *IP = CallArgs.getStackBase();
@@ -5135,7 +5152,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
     AI->setAlignment(Align.getAsAlign());
     AI->setUsedWithInAlloca(true);
     assert(AI->isUsedWithInAlloca() && !AI->isStaticAlloca());
-    ArgMemory = Address(AI, ArgStruct, Align);
+    ArgMemory = RawAddress(AI, ArgStruct, Align);
   }
 
   ClangToLLVMArgMapping IRFunctionArgs(CGM.getContext(), CallInfo);
@@ -5144,11 +5161,11 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   // If the call returns a temporary with struct return, create a temporary
   // alloca to hold the result, unless one is given to us.
   Address SRetPtr = Address::invalid();
-  Address SRetAlloca = Address::invalid();
+  RawAddress SRetAlloca = RawAddress::invalid();
   llvm::Value *UnusedReturnSizePtr = nullptr;
   if (RetAI.isIndirect() || RetAI.isInAlloca() || RetAI.isCoerceAndExpand()) {
     if (!ReturnValue.isNull()) {
-      SRetPtr = ReturnValue.getValue();
+      SRetPtr = ReturnValue.getAddress();
     } else {
       SRetPtr = CreateMemTemp(RetTy, "tmp", &SRetAlloca);
       if (HaveInsertPoint() && ReturnValue.isUnused()) {
@@ -5158,15 +5175,16 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
       }
     }
     if (IRFunctionArgs.hasSRetArg()) {
-      IRCallArgs[IRFunctionArgs.getSRetArgNo()] = SRetPtr.getPointer();
+      IRCallArgs[IRFunctionArgs.getSRetArgNo()] =
+          getAsNaturalPointerTo(SRetPtr, RetTy);
     } else if (RetAI.isInAlloca()) {
       Address Addr =
           Builder.CreateStructGEP(ArgMemory, RetAI.getInAllocaFieldIndex());
-      Builder.CreateStore(SRetPtr.getPointer(), Addr);
+      Builder.CreateStore(getAsNaturalPointerTo(SRetPtr, RetTy), Addr);
     }
   }
 
-  Address swiftErrorTemp = Address::invalid();
+  RawAddress swiftErrorTemp = RawAddress::invalid();
   Address swiftErrorArg = Address::invalid();
 
   // When passing arguments using temporary allocas, we need to add the
@@ -5199,9 +5217,9 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
       assert(NumIRArgs == 0);
       assert(getTarget().getTriple().getArch() == llvm::Triple::x86);
       if (I->isAggregate()) {
-        Address Addr = I->hasLValue()
-                           ? I->getKnownLValue().getAddress(*this)
-                           : I->getKnownRValue().getAggregateAddress();
+        RawAddress Addr = I->hasLValue()
+                              ? I->getKnownLValue().getAddress(*this)
+                              : I->getKnownRValue().getAggregateAddress();
         llvm::Instruction *Placeholder =
             cast<llvm::Instruction>(Addr.getPointer());
 
@@ -5225,7 +5243,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
       } else if (ArgInfo.getInAllocaIndirect()) {
         // Make a temporary alloca and store the address of it into the argument
         // struct.
-        Address Addr = CreateMemTempWithoutCast(
+        RawAddress Addr = CreateMemTempWithoutCast(
             I->Ty, getContext().getTypeAlignInChars(I->Ty),
             "indirect-arg-temp");
         I->copyInto(*this, Addr);
@@ -5247,12 +5265,12 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
       assert(NumIRArgs == 1);
       if (!I->isAggregate()) {
         // Make a temporary alloca to pass the argument.
-        Address Addr = CreateMemTempWithoutCast(
+        RawAddress Addr = CreateMemTempWithoutCast(
             I->Ty, ArgInfo.getIndirectAlign(), "indirect-arg-temp");
 
-        llvm::Value *Val = Addr.getPointer();
+        llvm::Value *Val = getAsNaturalPointerTo(Addr, I->Ty);
         if (ArgHasMaybeUndefAttr)
-          Val = Builder.CreateFreeze(Addr.getPointer());
+          Val = Builder.CreateFreeze(Val);
         IRCallArgs[FirstIRArg] = Val;
 
         I->copyInto(*this, Addr);
@@ -5268,7 +5286,6 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
         Address Addr = I->hasLValue()
                            ? I->getKnownLValue().getAddress(*this)
                            : I->getKnownRValue().getAggregateAddress();
-        llvm::Value *V = Addr.getPointer();
         CharUnits Align = ArgInfo.getIndirectAlign();
         const llvm::DataLayout *TD = &CGM.getDataLayout();
 
@@ -5279,8 +5296,9 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
         bool NeedCopy = false;
         if (Addr.getAlignment() < Align &&
-            llvm::getOrEnforceKnownAlignment(V, Align.getAsAlign(), *TD) <
-                Align.getAsAlign()) {
+            llvm::getOrEnforceKnownAlignment(Addr.emitRawPointer(*this),
+                                             Align.getAsAlign(),
+                                             *TD) < Align.getAsAlign()) {
           NeedCopy = true;
         } else if (I->hasLValue()) {
           auto LV = I->getKnownLValue();
@@ -5311,11 +5329,11 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
         if (NeedCopy) {
           // Create an aligned temporary, and copy to it.
-          Address AI = CreateMemTempWithoutCast(
+          RawAddress AI = CreateMemTempWithoutCast(
               I->Ty, ArgInfo.getIndirectAlign(), "byval-temp");
-          llvm::Value *Val = AI.getPointer();
+          llvm::Value *Val = getAsNaturalPointerTo(AI, I->Ty);
           if (ArgHasMaybeUndefAttr)
-            Val = Builder.CreateFreeze(AI.getPointer());
+            Val = Builder.CreateFreeze(Val);
           IRCallArgs[FirstIRArg] = Val;
 
           // Emit lifetime markers for the temporary alloca.
@@ -5332,6 +5350,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
           I->copyInto(*this, AI);
         } else {
           // Skip the extra memcpy call.
+          llvm::Value *V = getAsNaturalPointerTo(Addr, I->Ty);
           auto *T = llvm::PointerType::get(
               CGM.getLLVMContext(), CGM.getDataLayout().getAllocaAddrSpace());
           llvm::Value *Val = getTargetHooks().performAddrSpaceCast(
@@ -5370,8 +5389,8 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
           assert(!swiftErrorTemp.isValid() && "multiple swifterror args");
 
           QualType pointeeTy = I->Ty->getPointeeType();
-          swiftErrorArg = Address(V, ConvertTypeForMem(pointeeTy),
-                                  getContext().getTypeAlignInChars(pointeeTy));
+          swiftErrorArg = makeNaturalAddressForPointer(
+              V, pointeeTy, getContext().getTypeAlignInChars(pointeeTy));
 
           swiftErrorTemp =
             CreateMemTemp(pointeeTy, getPointerAlign(), "swifterror.temp");
@@ -5515,7 +5534,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
       llvm::Value *tempSize = nullptr;
       Address addr = Address::invalid();
-      Address AllocaAddr = Address::invalid();
+      RawAddress AllocaAddr = RawAddress::invalid();
       if (I->isAggregate()) {
         addr = I->hasLValue() ? I->getKnownLValue().getAddress(*this)
                               : I->getKnownRValue().getAggregateAddress();
@@ -5978,7 +5997,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
           return RValue::getComplex(std::make_pair(Real, Imag));
         }
         case TEK_Aggregate: {
-          Address DestPtr = ReturnValue.getValue();
+          Address DestPtr = ReturnValue.getAddress();
           bool DestIsVolatile = ReturnValue.isVolatile();
 
           if (!DestPtr.isValid()) {
diff --git a/clang/lib/CodeGen/CGCall.h b/clang/lib/CodeGen/CGCall.h
index 155042c8b0a50..b6d7b4d542a54 100644
--- a/clang/lib/CodeGen/CGCall.h
+++ b/clang/lib/CodeGen/CGCall.h
@@ -376,6 +376,7 @@ class ReturnValueSlot {
   Address getValue() const { return Addr; }
   bool isUnused() const { return IsUnused; }
   bool isExternallyDestructed() const { return IsExternallyDestructed; }
+  Address getAddress() const { return Addr; }
 };
 
 /// Adds attributes to \p F according to our \p CodeGenOpts and \p LangOpts, as
diff --git a/clang/lib/CodeGen/CGClass.cpp b/clang/lib/CodeGen/CGClass.cpp
index 69e602008a9a9..9aa3241129f75 100644
--- a/clang/lib/CodeGen/CGClass.cpp
+++ b/clang/lib/CodeGen/CGClass.cpp
@@ -139,8 +139,9 @@ Address CodeGenFunction::LoadCXXThisAddress() {
     CXXThisAlignment = CGM.getClassPointerAlignment(MD->getParent());
   }
 
-  llvm::Type *Ty = ConvertType(MD->getFunctionObjectParameterType());
-  return Address(LoadCXXThis(), Ty, CXXThisAlignment, KnownNonNull);
+  return makeNaturalAddressForPointer(
+      LoadCXXThis(), MD->getFunctionObjectParameterType(), CXXThisAlignment,
+      false, nullptr, nullptr, KnownNonNull);
 }
 
 /// Emit the address of a field using a member data pointer.
@@ -271,7 +272,7 @@ ApplyNonVirtualAndVirtualOffset(CodeGenFunction &CGF, Address addr,
   }
 
   // Apply the base offset.
-  llvm::Value *ptr = addr.getPointer();
+  llvm::Value *ptr = addr.emitRawPointer(CGF);
   ptr = CGF.Builder.CreateInBoundsGEP(CGF.Int8Ty, ptr, baseOffset, "add.ptr");
 
   // If we have a virtual component, the alignment of the result will
@@ -339,8 +340,8 @@ Address CodeGenFunction::GetAddressOfBaseClass(
     if (sanitizePerformTypeCheck()) {
       SanitizerSet SkippedChecks;
       SkippedChecks.set(SanitizerKind::Null, !NullCheckValue);
-      EmitTypeCheck(TCK_Upcast, Loc, Value.getPointer(),
-                    DerivedTy, DerivedAlign, SkippedChecks);
+      EmitTypeCheck(TCK_Upcast, Loc, Value.emitRawPointer(*this), DerivedTy,
+                    DerivedAlign, SkippedChecks);
     }
     return Value.withElementType(BaseValueTy);
   }
@@ -355,7 +356,7 @@ Address CodeGenFunction::GetAddressOfBaseClass(
     llvm::BasicBlock *notNullBB = createBasicBlock("cast.notnull");
     endBB = createBasicBlock("cast.end");
 
-    llvm::Value *isNull = Builder.CreateIsNull(Value.getPointer());
+    llvm::Value *isNull = Builder.CreateIsNull(Value);
     Builder.CreateCondBr(isNull, endBB, notNullBB);
     EmitBlock(notNullBB);
   }
@@ -364,14 +365,15 @@ Address CodeGenFunction::GetAddressOfBaseClass(
     SanitizerSet SkippedChecks;
     SkippedChecks.set(SanitizerKind::Null, true);
     EmitTypeCheck(VBase ? TCK_UpcastToVirtualBase : TCK_Upcast, Loc,
-                  Value.getPointer(), DerivedTy, DerivedAlign, SkippedChecks);
+                  Value.emitRawPointer(*this), DerivedTy, DerivedAlign,
+                  SkippedChecks);
   }
 
   // Compute the virtual offset.
   llvm::Value *VirtualOffset = nullptr;
   if (VBase) {
     VirtualOffset =
-      CGM.getCXXABI().GetVirtualBaseClassOffset(*this, Value, Derived, VBase);
+        CGM.getCXXABI().GetVirtualBaseClassOffset(*this, Value, Derived, VBase);
   }
 
   // Apply both offsets.
@@ -388,7 +390,7 @@ Address CodeGenFunction::GetAddressOfBaseClass(
     EmitBlock(endBB);
 
     llvm::PHINode *PHI = Builder.CreatePHI(PtrTy, 2, "cast.result");
-    PHI->addIncoming(Value.getPointer(), notNullBB);
+    PHI->addIncoming(Value.emitRawPointer(*this), notNullBB);
     PHI->addIncoming(llvm::Constant::getNullValue(PtrTy), origBB);
     Value = Value.withPointer(PHI, NotKnownNonNull);
   }
@@ -425,15 +427,19 @@ CodeGenFunction::GetAddressOfDerivedClass(Address BaseAddr,
     CastNotNull = createBasicBlock("cast.notnull");
     CastEnd = createBasicBlock("cast.end");
 
-    llvm::Value *IsNull = Builder.CreateIsNull(BaseAddr.getPointer());
+    llvm::Value *IsNull = Builder.CreateIsNull(BaseAddr);
     Builder.CreateCondBr(IsNull, CastNull, CastNotNull);
     EmitBlock(CastNotNull);
   }
 
   // Apply the offset.
-  llvm::Value *Value = BaseAddr.getPointer();
-  Value = Builder.CreateInBoundsGEP(
-      Int8Ty, Value, Builder.CreateNeg(NonVirtualOffset), "sub.ptr");
+  Address Addr = BaseAddr.withElementType(Int8Ty);
+  Addr = Builder.CreateInBoundsGEP(
+      Addr, Builder.CreateNeg(NonVirtualOffset), Int8Ty,
+      CGM.getClassPointerAlignment(Derived), "sub.ptr");
+
+  // Just cast.
+  Addr = Addr.withElementType(DerivedValueTy);
 
   // Produce a PHI if we had a null-check.
   if (NullCheckValue) {
@@ -442,13 +448,15 @@ CodeGenFunction::GetAddressOfDerivedClass(Address BaseAddr,
     Builder.CreateBr(CastEnd);
     EmitBlock(CastEnd);
 
+    llvm::Value *Value = Addr.emitRawPointer(*this);
     llvm::PHINode *PHI = Builder.CreatePHI(Value->getType(), 2);
     PHI->addIncoming(Value, CastNotNull);
     PHI->addIncoming(llvm::Constant::getNullValue(Value->getType()), CastNull);
-    Value = PHI;
+    return Address(PHI, Addr.getElementType(),
+                   CGM.getClassPointerAlignment(Derived));
   }
 
-  return Address(Value, DerivedValueTy, CGM.getClassPointerAlignment(Derived));
+  return Addr;
 }
 
 llvm::Value *CodeGenFunction::GetVTTParameter(GlobalDecl GD,
@@ -1720,7 +1728,7 @@ namespace {
       // Use the base class declaration location as inline DebugLocation. All
       // fields of the class are destroyed.
       DeclAsInlineDebugLocation InlineHere(CGF, *BaseClass);
-      EmitSanitizerDtorFieldsCallback(CGF, Addr.getPointer(),
+      EmitSanitizerDtorFieldsCallback(CGF, Addr.emitRawPointer(CGF),
                                       BaseSize.getQuantity());
 
       // Prevent the current stack frame from disappearing from the stack trace.
@@ -2023,7 +2031,7 @@ void CodeGenFunction::EmitCXXAggrConstructorCall(const CXXConstructorDecl *ctor,
 
   // Find the end of the array.
   llvm::Type *elementType = arrayBase.getElementType();
-  llvm::Value *arrayBegin = arrayBase.getPointer();
+  llvm::Value *arrayBegin = arrayBase.emitRawPointer(*this);
   llvm::Value *arrayEnd = Builder.CreateInBoundsGEP(
       elementType, arrayBegin, numElements, "arrayctor.end");
 
@@ -2119,14 +2127,15 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
   Address This = ThisAVS.getAddress();
   LangAS SlotAS = ThisAVS.getQualifiers().getAddressSpace();
   LangAS ThisAS = D->getFunctionObjectParameterType().getAddressSpace();
-  llvm::Value *ThisPtr = This.getPointer();
+  llvm::Value *ThisPtr =
+      getAsNaturalPointerTo(This, D->getThisType()->getPointeeType());
 
   if (SlotAS != ThisAS) {
     unsigned TargetThisAS = getContext().getTargetAddressSpace(ThisAS);
     llvm::Type *NewType =
         llvm::PointerType::get(getLLVMContext(), TargetThisAS);
-    ThisPtr = getTargetHooks().performAddrSpaceCast(*this, This.getPointer(),
-                                                    ThisAS, SlotAS, NewType);
+    ThisPtr = getTargetHooks().performAddrSpaceCast(*this, ThisPtr, ThisAS,
+                                                    SlotAS, NewType);
   }
 
   // Push the this ptr.
@@ -2195,7 +2204,7 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
   const CXXRecordDecl *ClassDecl = D->getParent();
 
   if (!NewPointerIsChecked)
-    EmitTypeCheck(CodeGenFunction::TCK_ConstructorCall, Loc, This.getPointer(),
+    EmitTypeCheck(CodeGenFunction::TCK_ConstructorCall, Loc, This,
                   getContext().getRecordType(ClassDecl), CharUnits::Zero());
 
   if (D->isTrivial() && D->isDefaultConstructor()) {
@@ -2208,10 +2217,9 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
   // model that copy.
   if (isMemcpyEquivalentSpecialMember(D)) {
     assert(Args.size() == 2 && "unexpected argcount for trivial ctor");
-
     QualType SrcTy = D->getParamDecl(0)->getType().getNonReferenceType();
-    Address Src = Address(Args[1].getRValue(*this).getScalarVal(), ConvertTypeForMem(SrcTy),
-                                      CGM.getNaturalTypeAlignment(SrcTy));
+    Address Src = makeNaturalAddressForPointer(
+        Args[1].getRValue(*this).getScalarVal(), SrcTy);
     LValue SrcLVal = MakeAddrLValue(Src, SrcTy);
     QualType DestTy = getContext().getTypeDeclType(ClassDecl);
     LValue DestLVal = MakeAddrLValue(This, DestTy);
@@ -2264,7 +2272,9 @@ void CodeGenFunction::EmitInheritedCXXConstructorCall(
     const CXXConstructorDecl *D, bool ForVirtualBase, Address This,
     bool InheritedFromVBase, const CXXInheritedCtorInitExpr *E) {
   CallArgList Args;
-  CallArg ThisArg(RValue::get(This.getPointer()), D->getThisType());
+  CallArg ThisArg(RValue::get(getAsNaturalPointerTo(
+                      This, D->getThisType()->getPointeeType())),
+                  D->getThisType());
 
   // Forward the parameters.
   if (InheritedFromVBase &&
@@ -2389,12 +2399,14 @@ CodeGenFunction::EmitSynthesizedCXXCopyCtorCall(const CXXConstructorDecl *D,
   CallArgList Args;
 
   // Push the this ptr.
-  Args.add(RValue::get(This.getPointer()), D->getThisType());
+  Args.add(RValue::get(getAsNaturalPointerTo(This, D->getThisType())),
+           D->getThisType());
 
   // Push the src ptr.
   QualType QT = *(FPT->param_type_begin());
   llvm::Type *t = CGM.getTypes().ConvertType(QT);
-  llvm::Value *SrcVal = Builder.CreateBitCast(Src.getPointer(), t);
+  llvm::Value *Val = getAsNaturalPointerTo(Src, D->getThisType());
+  llvm::Value *SrcVal = Builder.CreateBitCast(Val, t);
   Args.add(RValue::get(SrcVal), QT);
 
   // Skip over first argument (Src).
@@ -2419,7 +2431,9 @@ CodeGenFunction::EmitDelegateCXXConstructorCall(const CXXConstructorDecl *Ctor,
 
   // this
   Address This = LoadCXXThisAddress();
-  DelegateArgs.add(RValue::get(This.getPointer()), (*I)->getType());
+  DelegateArgs.add(RValue::get(getAsNaturalPointerTo(
+                       This, (*I)->getType()->getPointeeType())),
+                   (*I)->getType());
   ++I;
 
   // FIXME: The location of the VTT parameter in the parameter list is
@@ -2780,7 +2794,7 @@ void CodeGenFunction::EmitVTablePtrCheckForCast(QualType T, Address Derived,
 
   if (MayBeNull) {
     llvm::Value *DerivedNotNull =
-        Builder.CreateIsNotNull(Derived.getPointer(), "cast.nonnull");
+        Builder.CreateIsNotNull(Derived.emitRawPointer(*this), "cast.nonnull");
 
     llvm::BasicBlock *CheckBlock = createBasicBlock("cast.check");
     ContBlock = createBasicBlock("cast.cont");
@@ -2981,7 +2995,7 @@ void CodeGenFunction::EmitLambdaBlockInvokeBody() {
 
   QualType ThisType = getContext().getPointerType(getContext().getRecordType(Lambda));
   Address ThisPtr = GetAddrOfBlockDecl(variable);
-  CallArgs.add(RValue::get(ThisPtr.getPointer()), ThisType);
+  CallArgs.add(RValue::get(getAsNaturalPointerTo(ThisPtr, ThisType)), ThisType);
 
   // Add the rest of the parameters.
   for (auto *param : BD->parameters())
@@ -3009,7 +3023,7 @@ void CodeGenFunction::EmitLambdaStaticInvokeBody(const CXXMethodDecl *MD) {
   QualType LambdaType = getContext().getRecordType(Lambda);
   QualType ThisType = getContext().getPointerType(LambdaType);
   Address ThisPtr = CreateMemTemp(LambdaType, "unused.capture");
-  CallArgs.add(RValue::get(ThisPtr.getPointer()), ThisType);
+  CallArgs.add(RValue::get(ThisPtr.emitRawPointer(*this)), ThisType);
 
   EmitLambdaDelegatingInvokeBody(MD, CallArgs);
 }
diff --git a/clang/lib/CodeGen/CGCleanup.cpp b/clang/lib/CodeGen/CGCleanup.cpp
index f87caf050eeaa..e6f8e6873004f 100644
--- a/clang/lib/CodeGen/CGCleanup.cpp
+++ b/clang/lib/CodeGen/CGCleanup.cpp
@@ -27,7 +27,7 @@ bool DominatingValue<RValue>::saved_type::needsSaving(RValue rv) {
   if (rv.isScalar())
     return DominatingLLVMValue::needsSaving(rv.getScalarVal());
   if (rv.isAggregate())
-    return DominatingLLVMValue::needsSaving(rv.getAggregatePointer());
+    return DominatingValue<Address>::needsSaving(rv.getAggregateAddress());
   return true;
 }
 
@@ -35,69 +35,40 @@ DominatingValue<RValue>::saved_type
 DominatingValue<RValue>::saved_type::save(CodeGenFunction &CGF, RValue rv) {
   if (rv.isScalar()) {
     llvm::Value *V = rv.getScalarVal();
-
-    // These automatically dominate and don't need to be saved.
-    if (!DominatingLLVMValue::needsSaving(V))
-      return saved_type(V, nullptr, ScalarLiteral);
-
-    // Everything else needs an alloca.
-    Address addr =
-      CGF.CreateDefaultAlignTempAlloca(V->getType(), "saved-rvalue");
-    CGF.Builder.CreateStore(V, addr);
-    return saved_type(addr.getPointer(), nullptr, ScalarAddress);
+    return saved_type(DominatingLLVMValue::save(CGF, V),
+                      DominatingLLVMValue::needsSaving(V) ? ScalarAddress
+                                                          : ScalarLiteral);
   }
 
   if (rv.isComplex()) {
     CodeGenFunction::ComplexPairTy V = rv.getComplexVal();
-    llvm::Type *ComplexTy =
-        llvm::StructType::get(V.first->getType(), V.second->getType());
-    Address addr = CGF.CreateDefaultAlignTempAlloca(ComplexTy, "saved-complex");
-    CGF.Builder.CreateStore(V.first, CGF.Builder.CreateStructGEP(addr, 0));
-    CGF.Builder.CreateStore(V.second, CGF.Builder.CreateStructGEP(addr, 1));
-    return saved_type(addr.getPointer(), nullptr, ComplexAddress);
+    return saved_type(DominatingLLVMValue::save(CGF, V.first),
+                      DominatingLLVMValue::save(CGF, V.second));
   }
 
   assert(rv.isAggregate());
-  Address V = rv.getAggregateAddress(); // TODO: volatile?
-  if (!DominatingLLVMValue::needsSaving(V.getPointer()))
-    return saved_type(V.getPointer(), V.getElementType(), AggregateLiteral,
-                      V.getAlignment().getQuantity());
-
-  Address addr =
-    CGF.CreateTempAlloca(V.getType(), CGF.getPointerAlign(), "saved-rvalue");
-  CGF.Builder.CreateStore(V.getPointer(), addr);
-  return saved_type(addr.getPointer(), V.getElementType(), AggregateAddress,
-                    V.getAlignment().getQuantity());
+  Address V = rv.getAggregateAddress();
+  return saved_type(
+      DominatingValue<Address>::save(CGF, V), rv.isVolatileQualified(),
+      DominatingValue<Address>::needsSaving(V) ? AggregateAddress
+                                               : AggregateLiteral);
 }
 
 /// Given a saved r-value produced by SaveRValue, perform the code
 /// necessary to restore it to usability at the current insertion
 /// point.
 RValue DominatingValue<RValue>::saved_type::restore(CodeGenFunction &CGF) {
-  auto getSavingAddress = [&](llvm::Value *value) {
-    auto *AI = cast<llvm::AllocaInst>(value);
-    return Address(value, AI->getAllocatedType(),
-                   CharUnits::fromQuantity(AI->getAlign().value()));
-  };
   switch (K) {
   case ScalarLiteral:
-    return RValue::get(Value);
   case ScalarAddress:
-    return RValue::get(CGF.Builder.CreateLoad(getSavingAddress(Value)));
+    return RValue::get(DominatingLLVMValue::restore(CGF, Vals.first));
   case AggregateLiteral:
+  case AggregateAddress:
     return RValue::getAggregate(
-        Address(Value, ElementType, CharUnits::fromQuantity(Align)));
-  case AggregateAddress: {
-    auto addr = CGF.Builder.CreateLoad(getSavingAddress(Value));
-    return RValue::getAggregate(
-        Address(addr, ElementType, CharUnits::fromQuantity(Align)));
-  }
+        DominatingValue<Address>::restore(CGF, AggregateAddr), IsVolatile);
   case ComplexAddress: {
-    Address address = getSavingAddress(Value);
-    llvm::Value *real =
-        CGF.Builder.CreateLoad(CGF.Builder.CreateStructGEP(address, 0));
-    llvm::Value *imag =
-        CGF.Builder.CreateLoad(CGF.Builder.CreateStructGEP(address, 1));
+    llvm::Value *real = DominatingLLVMValue::restore(CGF, Vals.first);
+    llvm::Value *imag = DominatingLLVMValue::restore(CGF, Vals.second);
     return RValue::getComplex(real, imag);
   }
   }
@@ -294,14 +265,14 @@ void EHScopeStack::popNullFixups() {
     BranchFixups.pop_back();
 }
 
-Address CodeGenFunction::createCleanupActiveFlag() {
+RawAddress CodeGenFunction::createCleanupActiveFlag() {
   // Create a variable to decide whether the cleanup needs to be run.
-  Address active = CreateTempAllocaWithoutCast(
+  RawAddress active = CreateTempAllocaWithoutCast(
       Builder.getInt1Ty(), CharUnits::One(), "cleanup.cond");
 
   // Initialize it to false at a site that's guaranteed to be run
   // before each evaluation.
-  setBeforeOutermostConditional(Builder.getFalse(), active);
+  setBeforeOutermostConditional(Builder.getFalse(), active, *this);
 
   // Initialize it to true at the current location.
   Builder.CreateStore(Builder.getTrue(), active);
@@ -309,7 +280,7 @@ Address CodeGenFunction::createCleanupActiveFlag() {
   return active;
 }
 
-void CodeGenFunction::initFullExprCleanupWithFlag(Address ActiveFlag) {
+void CodeGenFunction::initFullExprCleanupWithFlag(RawAddress ActiveFlag) {
   // Set that as the active flag in the cleanup.
   EHCleanupScope &cleanup = cast<EHCleanupScope>(*EHStack.begin());
   assert(!cleanup.hasActiveFlag() && "cleanup already has active flag?");
@@ -322,15 +293,17 @@ void CodeGenFunction::initFullExprCleanupWithFlag(Address ActiveFlag) {
 void EHScopeStack::Cleanup::anchor() {}
 
 static void createStoreInstBefore(llvm::Value *value, Address addr,
-                                  llvm::Instruction *beforeInst) {
-  auto store = new llvm::StoreInst(value, addr.getPointer(), beforeInst);
+                                  llvm::Instruction *beforeInst,
+                                  CodeGenFunction &CGF) {
+  auto store = new llvm::StoreInst(value, addr.emitRawPointer(CGF), beforeInst);
   store->setAlignment(addr.getAlignment().getAsAlign());
 }
 
 static llvm::LoadInst *createLoadInstBefore(Address addr, const Twine &name,
-                                            llvm::Instruction *beforeInst) {
-  return new llvm::LoadInst(addr.getElementType(), addr.getPointer(), name,
-                            false, addr.getAlignment().getAsAlign(),
+                                            llvm::Instruction *beforeInst,
+                                            CodeGenFunction &CGF) {
+  return new llvm::LoadInst(addr.getElementType(), addr.emitRawPointer(CGF),
+                            name, false, addr.getAlignment().getAsAlign(),
                             beforeInst);
 }
 
@@ -357,8 +330,8 @@ static void ResolveAllBranchFixups(CodeGenFunction &CGF,
     // entry which we're currently popping.
     if (Fixup.OptimisticBranchBlock == nullptr) {
       createStoreInstBefore(CGF.Builder.getInt32(Fixup.DestinationIndex),
-                            CGF.getNormalCleanupDestSlot(),
-                            Fixup.InitialBranch);
+                            CGF.getNormalCleanupDestSlot(), Fixup.InitialBranch,
+                            CGF);
       Fixup.InitialBranch->setSuccessor(0, CleanupEntry);
     }
 
@@ -385,7 +358,7 @@ static llvm::SwitchInst *TransitionToCleanupSwitch(CodeGenFunction &CGF,
   if (llvm::BranchInst *Br = dyn_cast<llvm::BranchInst>(Term)) {
     assert(Br->isUnconditional());
     auto Load = createLoadInstBefore(CGF.getNormalCleanupDestSlot(),
-                                     "cleanup.dest", Term);
+                                     "cleanup.dest", Term, CGF);
     llvm::SwitchInst *Switch =
       llvm::SwitchInst::Create(Load, Br->getSuccessor(0), 4, Block);
     Br->eraseFromParent();
@@ -513,8 +486,8 @@ void CodeGenFunction::PopCleanupBlocks(
     I += Header.getSize();
 
     if (Header.isConditional()) {
-      Address ActiveFlag =
-          reinterpret_cast<Address &>(LifetimeExtendedCleanupStack[I]);
+      RawAddress ActiveFlag =
+          reinterpret_cast<RawAddress &>(LifetimeExtendedCleanupStack[I]);
       initFullExprCleanupWithFlag(ActiveFlag);
       I += sizeof(ActiveFlag);
     }
@@ -888,7 +861,7 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
         if (NormalCleanupDestSlot->hasOneUse()) {
           NormalCleanupDestSlot->user_back()->eraseFromParent();
           NormalCleanupDestSlot->eraseFromParent();
-          NormalCleanupDest = Address::invalid();
+          NormalCleanupDest = RawAddress::invalid();
         }
 
         llvm::BasicBlock *BranchAfter = Scope.getBranchAfterBlock(0);
@@ -912,9 +885,8 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
         // pass the abnormal exit flag to Fn (SEH cleanup)
         cleanupFlags.setHasExitSwitch();
 
-        llvm::LoadInst *Load =
-          createLoadInstBefore(getNormalCleanupDestSlot(), "cleanup.dest",
-                               nullptr);
+        llvm::LoadInst *Load = createLoadInstBefore(
+            getNormalCleanupDestSlot(), "cleanup.dest", nullptr, *this);
         llvm::SwitchInst *Switch =
           llvm::SwitchInst::Create(Load, Default, SwitchCapacity);
 
@@ -961,8 +933,8 @@ void CodeGenFunction::PopCleanupBlock(bool FallthroughIsBranchThrough) {
         if (!Fixup.Destination) continue;
         if (!Fixup.OptimisticBranchBlock) {
           createStoreInstBefore(Builder.getInt32(Fixup.DestinationIndex),
-                                getNormalCleanupDestSlot(),
-                                Fixup.InitialBranch);
+                                getNormalCleanupDestSlot(), Fixup.InitialBranch,
+                                *this);
           Fixup.InitialBranch->setSuccessor(0, NormalEntry);
         }
         Fixup.OptimisticBranchBlock = NormalExit;
@@ -1135,7 +1107,7 @@ void CodeGenFunction::EmitBranchThroughCleanup(JumpDest Dest) {
 
   // Store the index at the start.
   llvm::ConstantInt *Index = Builder.getInt32(Dest.getDestIndex());
-  createStoreInstBefore(Index, getNormalCleanupDestSlot(), BI);
+  createStoreInstBefore(Index, getNormalCleanupDestSlot(), BI, *this);
 
   // Adjust BI to point to the first cleanup block.
   {
@@ -1269,9 +1241,9 @@ static void SetupCleanupBlockActivation(CodeGenFunction &CGF,
     // If we're in a conditional block, ignore the dominating IP and
     // use the outermost conditional branch.
     if (CGF.isInConditionalBranch()) {
-      CGF.setBeforeOutermostConditional(value, var);
+      CGF.setBeforeOutermostConditional(value, var, CGF);
     } else {
-      createStoreInstBefore(value, var, dominatingIP);
+      createStoreInstBefore(value, var, dominatingIP, CGF);
     }
   }
 
@@ -1321,7 +1293,7 @@ void CodeGenFunction::DeactivateCleanupBlock(EHScopeStack::stable_iterator C,
   Scope.setActive(false);
 }
 
-Address CodeGenFunction::getNormalCleanupDestSlot() {
+RawAddress CodeGenFunction::getNormalCleanupDestSlot() {
   if (!NormalCleanupDest.isValid())
     NormalCleanupDest =
       CreateDefaultAlignTempAlloca(Builder.getInt32Ty(), "cleanup.dest.slot");
diff --git a/clang/lib/CodeGen/CGCleanup.h b/clang/lib/CodeGen/CGCleanup.h
index 7a7344c07160d..03e4a29d7b3db 100644
--- a/clang/lib/CodeGen/CGCleanup.h
+++ b/clang/lib/CodeGen/CGCleanup.h
@@ -333,7 +333,7 @@ class alignas(8) EHCleanupScope : public EHScope {
   Address getActiveFlag() const {
     return ActiveFlag;
   }
-  void setActiveFlag(Address Var) {
+  void setActiveFlag(RawAddress Var) {
     assert(Var.getAlignment().isOne());
     ActiveFlag = Var;
   }
diff --git a/clang/lib/CodeGen/CGCoroutine.cpp b/clang/lib/CodeGen/CGCoroutine.cpp
index b7142ec08af98..93ca711f716fc 100644
--- a/clang/lib/CodeGen/CGCoroutine.cpp
+++ b/clang/lib/CodeGen/CGCoroutine.cpp
@@ -867,8 +867,8 @@ void CodeGenFunction::EmitCoroutineBody(const CoroutineBodyStmt &S) {
     EmitStmt(S.getPromiseDeclStmt());
 
     Address PromiseAddr = GetAddrOfLocalVar(S.getPromiseDecl());
-    auto *PromiseAddrVoidPtr =
-        new llvm::BitCastInst(PromiseAddr.getPointer(), VoidPtrTy, "", CoroId);
+    auto *PromiseAddrVoidPtr = new llvm::BitCastInst(
+        PromiseAddr.emitRawPointer(*this), VoidPtrTy, "", CoroId);
     // Update CoroId to refer to the promise. We could not do it earlier because
     // promise local variable was not emitted yet.
     CoroId->setArgOperand(1, PromiseAddrVoidPtr);
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index f3f825c1e2376..b968518c47e91 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1497,7 +1497,7 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
   bool EmitDebugInfo = DI && CGM.getCodeGenOpts().hasReducedDebugInfo();
 
   Address address = Address::invalid();
-  Address AllocaAddr = Address::invalid();
+  RawAddress AllocaAddr = RawAddress::invalid();
   Address OpenMPLocalAddr = Address::invalid();
   if (CGM.getLangOpts().OpenMPIRBuilder)
     OpenMPLocalAddr = OMPBuilderCBHelpers::getAddressOfLocalVariable(*this, &D);
@@ -1560,7 +1560,10 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
       // return slot, so that we can elide the copy when returning this
       // variable (C++0x [class.copy]p34).
       address = ReturnValue;
-      AllocaAddr = ReturnValue;
+      AllocaAddr =
+          RawAddress(ReturnValue.emitRawPointer(*this),
+                     ReturnValue.getElementType(), ReturnValue.getAlignment());
+      ;
 
       if (const RecordType *RecordTy = Ty->getAs<RecordType>()) {
         const auto *RD = RecordTy->getDecl();
@@ -1571,7 +1574,7 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
           // to this variable. Set it to zero to indicate that NRVO was not
           // applied.
           llvm::Value *Zero = Builder.getFalse();
-          Address NRVOFlag =
+          RawAddress NRVOFlag =
               CreateTempAlloca(Zero->getType(), CharUnits::One(), "nrvo");
           EnsureInsertPoint();
           Builder.CreateStore(Zero, NRVOFlag);
@@ -1732,7 +1735,7 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
   }
 
   if (D.hasAttr<AnnotateAttr>() && HaveInsertPoint())
-    EmitVarAnnotations(&D, address.getPointer());
+    EmitVarAnnotations(&D, address.emitRawPointer(*this));
 
   // Make sure we call @llvm.lifetime.end.
   if (emission.useLifetimeMarkers())
@@ -1905,12 +1908,13 @@ void CodeGenFunction::emitZeroOrPatternForAutoVarInit(QualType type,
     llvm::Value *BaseSizeInChars =
         llvm::ConstantInt::get(IntPtrTy, EltSize.getQuantity());
     Address Begin = Loc.withElementType(Int8Ty);
-    llvm::Value *End = Builder.CreateInBoundsGEP(
-        Begin.getElementType(), Begin.getPointer(), SizeVal, "vla.end");
+    llvm::Value *End = Builder.CreateInBoundsGEP(Begin.getElementType(),
+                                                 Begin.emitRawPointer(*this),
+                                                 SizeVal, "vla.end");
     llvm::BasicBlock *OriginBB = Builder.GetInsertBlock();
     EmitBlock(LoopBB);
     llvm::PHINode *Cur = Builder.CreatePHI(Begin.getType(), 2, "vla.cur");
-    Cur->addIncoming(Begin.getPointer(), OriginBB);
+    Cur->addIncoming(Begin.emitRawPointer(*this), OriginBB);
     CharUnits CurAlign = Loc.getAlignment().alignmentOfArrayElement(EltSize);
     auto *I =
         Builder.CreateMemCpy(Address(Cur, Int8Ty, CurAlign),
@@ -2337,7 +2341,7 @@ void CodeGenFunction::emitDestroy(Address addr, QualType type,
     checkZeroLength = false;
   }
 
-  llvm::Value *begin = addr.getPointer();
+  llvm::Value *begin = addr.emitRawPointer(*this);
   llvm::Value *end =
       Builder.CreateInBoundsGEP(addr.getElementType(), begin, length);
   emitArrayDestroy(begin, end, type, elementAlign, destroyer,
@@ -2597,7 +2601,7 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
   }
 
   Address DeclPtr = Address::invalid();
-  Address AllocaPtr = Address::invalid();
+  RawAddress AllocaPtr = Address::invalid();
   bool DoStore = false;
   bool IsScalar = hasScalarEvaluationKind(Ty);
   bool UseIndirectDebugAddress = false;
@@ -2609,8 +2613,8 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
     // Indirect argument is in alloca address space, which may be different
     // from the default address space.
     auto AllocaAS = CGM.getASTAllocaAddressSpace();
-    auto *V = DeclPtr.getPointer();
-    AllocaPtr = DeclPtr;
+    auto *V = DeclPtr.emitRawPointer(*this);
+    AllocaPtr = RawAddress(V, DeclPtr.getElementType(), DeclPtr.getAlignment());
 
     auto SrcLangAS = getLangOpts().OpenCL ? LangAS::opencl_private : AllocaAS;
     auto DestLangAS =
@@ -2750,7 +2754,7 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
   }
 
   if (D.hasAttr<AnnotateAttr>())
-    EmitVarAnnotations(&D, DeclPtr.getPointer());
+    EmitVarAnnotations(&D, DeclPtr.emitRawPointer(*this));
 
   // We can only check return value nullability if all arguments to the
   // function satisfy their nullability preconditions. This makes it necessary
diff --git a/clang/lib/CodeGen/CGException.cpp b/clang/lib/CodeGen/CGException.cpp
index f222b6e7a43c6..4991fdafff45a 100644
--- a/clang/lib/CodeGen/CGException.cpp
+++ b/clang/lib/CodeGen/CGException.cpp
@@ -397,7 +397,7 @@ namespace {
 void CodeGenFunction::EmitAnyExprToExn(const Expr *e, Address addr) {
   // Make sure the exception object is cleaned up if there's an
   // exception during initialization.
-  pushFullExprCleanup<FreeException>(EHCleanup, addr.getPointer());
+  pushFullExprCleanup<FreeException>(EHCleanup, addr.emitRawPointer(*this));
   EHScopeStack::stable_iterator cleanup = EHStack.stable_begin();
 
   // __cxa_allocate_exception returns a void*;  we need to cast this
@@ -416,8 +416,8 @@ void CodeGenFunction::EmitAnyExprToExn(const Expr *e, Address addr) {
                    /*IsInit*/ true);
 
   // Deactivate the cleanup block.
-  DeactivateCleanupBlock(cleanup,
-                         cast<llvm::Instruction>(typedAddr.getPointer()));
+  DeactivateCleanupBlock(
+      cleanup, cast<llvm::Instruction>(typedAddr.emitRawPointer(*this)));
 }
 
 Address CodeGenFunction::getExceptionSlot() {
@@ -1834,7 +1834,8 @@ Address CodeGenFunction::recoverAddrOfEscapedLocal(CodeGenFunction &ParentCGF,
                                                    llvm::Value *ParentFP) {
   llvm::CallInst *RecoverCall = nullptr;
   CGBuilderTy Builder(*this, AllocaInsertPt);
-  if (auto *ParentAlloca = dyn_cast<llvm::AllocaInst>(ParentVar.getPointer())) {
+  if (auto *ParentAlloca =
+          dyn_cast_or_null<llvm::AllocaInst>(ParentVar.getBasePointer())) {
     // Mark the variable escaped if nobody else referenced it and compute the
     // localescape index.
     auto InsertPair = ParentCGF.EscapedLocals.insert(
@@ -1851,8 +1852,8 @@ Address CodeGenFunction::recoverAddrOfEscapedLocal(CodeGenFunction &ParentCGF,
     // If the parent didn't have an alloca, we're doing some nested outlining.
     // Just clone the existing localrecover call, but tweak the FP argument to
     // use our FP value. All other arguments are constants.
-    auto *ParentRecover =
-        cast<llvm::IntrinsicInst>(ParentVar.getPointer()->stripPointerCasts());
+    auto *ParentRecover = cast<llvm::IntrinsicInst>(
+        ParentVar.emitRawPointer(*this)->stripPointerCasts());
     assert(ParentRecover->getIntrinsicID() == llvm::Intrinsic::localrecover &&
            "expected alloca or localrecover in parent LocalDeclMap");
     RecoverCall = cast<llvm::CallInst>(ParentRecover->clone());
@@ -1925,7 +1926,8 @@ void CodeGenFunction::EmitCapturedLocals(CodeGenFunction &ParentCGF,
         if (isa<ImplicitParamDecl>(D) &&
             D->getType() == getContext().VoidPtrTy) {
           assert(D->getName().starts_with("frame_pointer"));
-          FramePtrAddrAlloca = cast<llvm::AllocaInst>(I.second.getPointer());
+          FramePtrAddrAlloca =
+              cast<llvm::AllocaInst>(I.second.getBasePointer());
           break;
         }
       }
@@ -1986,7 +1988,8 @@ void CodeGenFunction::EmitCapturedLocals(CodeGenFunction &ParentCGF,
         LValue ThisFieldLValue =
             EmitLValueForLambdaField(LambdaThisCaptureField);
         if (!LambdaThisCaptureField->getType()->isPointerType()) {
-          CXXThisValue = ThisFieldLValue.getAddress(*this).getPointer();
+          CXXThisValue =
+              ThisFieldLValue.getAddress(*this).emitRawPointer(*this);
         } else {
           CXXThisValue = EmitLoadOfLValue(ThisFieldLValue, SourceLocation())
                              .getScalarVal();
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 612c8a2fcbe3e..10b010c1d1761 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -71,21 +71,21 @@ static llvm::cl::opt<bool> ClSanitizeGuardChecks(
 
 /// CreateTempAlloca - This creates a alloca and inserts it into the entry
 /// block.
-Address CodeGenFunction::CreateTempAllocaWithoutCast(llvm::Type *Ty,
-                                                     CharUnits Align,
-                                                     const Twine &Name,
-                                                     llvm::Value *ArraySize) {
+RawAddress
+CodeGenFunction::CreateTempAllocaWithoutCast(llvm::Type *Ty, CharUnits Align,
+                                             const Twine &Name,
+                                             llvm::Value *ArraySize) {
   auto Alloca = CreateTempAlloca(Ty, Name, ArraySize);
   Alloca->setAlignment(Align.getAsAlign());
-  return Address(Alloca, Ty, Align, KnownNonNull);
+  return RawAddress(Alloca, Ty, Align, KnownNonNull);
 }
 
 /// CreateTempAlloca - This creates a alloca and inserts it into the entry
 /// block. The alloca is casted to default address space if necessary.
-Address CodeGenFunction::CreateTempAlloca(llvm::Type *Ty, CharUnits Align,
-                                          const Twine &Name,
-                                          llvm::Value *ArraySize,
-                                          Address *AllocaAddr) {
+RawAddress CodeGenFunction::CreateTempAlloca(llvm::Type *Ty, CharUnits Align,
+                                             const Twine &Name,
+                                             llvm::Value *ArraySize,
+                                             RawAddress *AllocaAddr) {
   auto Alloca = CreateTempAllocaWithoutCast(Ty, Align, Name, ArraySize);
   if (AllocaAddr)
     *AllocaAddr = Alloca;
@@ -107,7 +107,7 @@ Address CodeGenFunction::CreateTempAlloca(llvm::Type *Ty, CharUnits Align,
         Ty->getPointerTo(DestAddrSpace), /*non-null*/ true);
   }
 
-  return Address(V, Ty, Align, KnownNonNull);
+  return RawAddress(V, Ty, Align, KnownNonNull);
 }
 
 /// CreateTempAlloca - This creates an alloca and inserts it into the entry
@@ -126,28 +126,29 @@ llvm::AllocaInst *CodeGenFunction::CreateTempAlloca(llvm::Type *Ty,
 /// default alignment of the corresponding LLVM type, which is *not*
 /// guaranteed to be related in any way to the expected alignment of
 /// an AST type that might have been lowered to Ty.
-Address CodeGenFunction::CreateDefaultAlignTempAlloca(llvm::Type *Ty,
-                                                      const Twine &Name) {
+RawAddress CodeGenFunction::CreateDefaultAlignTempAlloca(llvm::Type *Ty,
+                                                         const Twine &Name) {
   CharUnits Align =
       CharUnits::fromQuantity(CGM.getDataLayout().getPrefTypeAlign(Ty));
   return CreateTempAlloca(Ty, Align, Name);
 }
 
-Address CodeGenFunction::CreateIRTemp(QualType Ty, const Twine &Name) {
+RawAddress CodeGenFunction::CreateIRTemp(QualType Ty, const Twine &Name) {
   CharUnits Align = getContext().getTypeAlignInChars(Ty);
   return CreateTempAlloca(ConvertType(Ty), Align, Name);
 }
 
-Address CodeGenFunction::CreateMemTemp(QualType Ty, const Twine &Name,
-                                       Address *Alloca) {
+RawAddress CodeGenFunction::CreateMemTemp(QualType Ty, const Twine &Name,
+                                          RawAddress *Alloca) {
   // FIXME: Should we prefer the preferred type alignment here?
   return CreateMemTemp(Ty, getContext().getTypeAlignInChars(Ty), Name, Alloca);
 }
 
-Address CodeGenFunction::CreateMemTemp(QualType Ty, CharUnits Align,
-                                       const Twine &Name, Address *Alloca) {
-  Address Result = CreateTempAlloca(ConvertTypeForMem(Ty), Align, Name,
-                                    /*ArraySize=*/nullptr, Alloca);
+RawAddress CodeGenFunction::CreateMemTemp(QualType Ty, CharUnits Align,
+                                          const Twine &Name,
+                                          RawAddress *Alloca) {
+  RawAddress Result = CreateTempAlloca(ConvertTypeForMem(Ty), Align, Name,
+                                       /*ArraySize=*/nullptr, Alloca);
 
   if (Ty->isConstantMatrixType()) {
     auto *ArrayTy = cast<llvm::ArrayType>(Result.getElementType());
@@ -160,13 +161,14 @@ Address CodeGenFunction::CreateMemTemp(QualType Ty, CharUnits Align,
   return Result;
 }
 
-Address CodeGenFunction::CreateMemTempWithoutCast(QualType Ty, CharUnits Align,
-                                                  const Twine &Name) {
+RawAddress CodeGenFunction::CreateMemTempWithoutCast(QualType Ty,
+                                                     CharUnits Align,
+                                                     const Twine &Name) {
   return CreateTempAllocaWithoutCast(ConvertTypeForMem(Ty), Align, Name);
 }
 
-Address CodeGenFunction::CreateMemTempWithoutCast(QualType Ty,
-                                                  const Twine &Name) {
+RawAddress CodeGenFunction::CreateMemTempWithoutCast(QualType Ty,
+                                                     const Twine &Name) {
   return CreateMemTempWithoutCast(Ty, getContext().getTypeAlignInChars(Ty),
                                   Name);
 }
@@ -365,7 +367,7 @@ pushTemporaryCleanup(CodeGenFunction &CGF, const MaterializeTemporaryExpr *M,
     } else {
       CleanupFn = CGF.CGM.getAddrAndTypeOfCXXStructor(
           GlobalDecl(ReferenceTemporaryDtor, Dtor_Complete));
-      CleanupArg = cast<llvm::Constant>(ReferenceTemporary.getPointer());
+      CleanupArg = cast<llvm::Constant>(ReferenceTemporary.emitRawPointer(CGF));
     }
     CGF.CGM.getCXXABI().registerGlobalDtor(
         CGF, *cast<VarDecl>(M->getExtendingDecl()), CleanupFn, CleanupArg);
@@ -390,10 +392,10 @@ pushTemporaryCleanup(CodeGenFunction &CGF, const MaterializeTemporaryExpr *M,
   }
 }
 
-static Address createReferenceTemporary(CodeGenFunction &CGF,
-                                        const MaterializeTemporaryExpr *M,
-                                        const Expr *Inner,
-                                        Address *Alloca = nullptr) {
+static RawAddress createReferenceTemporary(CodeGenFunction &CGF,
+                                           const MaterializeTemporaryExpr *M,
+                                           const Expr *Inner,
+                                           RawAddress *Alloca = nullptr) {
   auto &TCG = CGF.getTargetHooks();
   switch (M->getStorageDuration()) {
   case SD_FullExpression:
@@ -422,7 +424,7 @@ static Address createReferenceTemporary(CodeGenFunction &CGF,
               GV->getValueType()->getPointerTo(
                   CGF.getContext().getTargetAddressSpace(LangAS::Default)));
         // FIXME: Should we put the new global into a COMDAT?
-        return Address(C, GV->getValueType(), alignment);
+        return RawAddress(C, GV->getValueType(), alignment);
       }
     return CGF.CreateMemTemp(Ty, "ref.tmp", Alloca);
   }
@@ -454,7 +456,7 @@ EmitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *M) {
   auto ownership = M->getType().getObjCLifetime();
   if (ownership != Qualifiers::OCL_None &&
       ownership != Qualifiers::OCL_ExplicitNone) {
-    Address Object = createReferenceTemporary(*this, M, E);
+    RawAddress Object = createReferenceTemporary(*this, M, E);
     if (auto *Var = dyn_cast<llvm::GlobalVariable>(Object.getPointer())) {
       llvm::Type *Ty = ConvertTypeForMem(E->getType());
       Object = Object.withElementType(Ty);
@@ -508,8 +510,8 @@ EmitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *M) {
   }
 
   // Create and initialize the reference temporary.
-  Address Alloca = Address::invalid();
-  Address Object = createReferenceTemporary(*this, M, E, &Alloca);
+  RawAddress Alloca = Address::invalid();
+  RawAddress Object = createReferenceTemporary(*this, M, E, &Alloca);
   if (auto *Var = dyn_cast<llvm::GlobalVariable>(
           Object.getPointer()->stripPointerCasts())) {
     llvm::Type *TemporaryType = ConvertTypeForMem(E->getType());
@@ -1117,12 +1119,12 @@ llvm::Value *CodeGenFunction::EmitCountedByFieldExpr(
   } else if (const MemberExpr *ME = dyn_cast<MemberExpr>(StructBase)) {
     LValue LV = EmitMemberExpr(ME);
     Address Addr = LV.getAddress(*this);
-    Res = Addr.getPointer();
+    Res = Addr.emitRawPointer(*this);
   } else if (StructBase->getType()->isPointerType()) {
     LValueBaseInfo BaseInfo;
     TBAAAccessInfo TBAAInfo;
     Address Addr = EmitPointerWithAlignment(StructBase, &BaseInfo, &TBAAInfo);
-    Res = Addr.getPointer();
+    Res = Addr.emitRawPointer(*this);
   } else {
     return nullptr;
   }
@@ -1288,8 +1290,7 @@ static Address EmitPointerWithAlignment(const Expr *E, LValueBaseInfo *BaseInfo,
           if (InnerBaseInfo.getAlignmentSource() != AlignmentSource::Decl) {
             if (BaseInfo)
               BaseInfo->mergeForCast(TargetTypeBaseInfo);
-            Addr = Address(Addr.getPointer(), Addr.getElementType(), Align,
-                           IsKnownNonNull);
+            Addr.setAlignment(Align);
           }
         }
 
@@ -1306,8 +1307,8 @@ static Address EmitPointerWithAlignment(const Expr *E, LValueBaseInfo *BaseInfo,
             CGF.ConvertTypeForMem(E->getType()->getPointeeType());
         Addr = Addr.withElementType(ElemTy);
         if (CE->getCastKind() == CK_AddressSpaceConversion)
-          Addr = CGF.Builder.CreateAddrSpaceCast(Addr,
-                                                 CGF.ConvertType(E->getType()));
+          Addr = CGF.Builder.CreateAddrSpaceCast(
+              Addr, CGF.ConvertType(E->getType()), ElemTy);
         return Addr;
       }
       break;
@@ -1370,10 +1371,9 @@ static Address EmitPointerWithAlignment(const Expr *E, LValueBaseInfo *BaseInfo,
   // TODO: conditional operators, comma.
 
   // Otherwise, use the alignment of the type.
-  CharUnits Align =
-      CGF.CGM.getNaturalPointeeTypeAlignment(E->getType(), BaseInfo, TBAAInfo);
-  llvm::Type *ElemTy = CGF.ConvertTypeForMem(E->getType()->getPointeeType());
-  return Address(CGF.EmitScalarExpr(E), ElemTy, Align, IsKnownNonNull);
+  return CGF.makeNaturalAddressForPointer(
+      CGF.EmitScalarExpr(E), E->getType()->getPointeeType(), CharUnits(),
+      /*ForPointeeType=*/true, BaseInfo, TBAAInfo, IsKnownNonNull);
 }
 
 /// EmitPointerWithAlignment - Given an expression of pointer type, try to
@@ -1474,8 +1474,7 @@ LValue CodeGenFunction::EmitCheckedLValue(const Expr *E, TypeCheckKind TCK) {
       if (IsBaseCXXThis || isa<DeclRefExpr>(ME->getBase()))
         SkippedChecks.set(SanitizerKind::Null, true);
     }
-    EmitTypeCheck(TCK, E->getExprLoc(), LV.getPointer(*this), E->getType(),
-                  LV.getAlignment(), SkippedChecks);
+    EmitTypeCheck(TCK, E->getExprLoc(), LV, E->getType(), SkippedChecks);
   }
   return LV;
 }
@@ -1587,11 +1586,11 @@ LValue CodeGenFunction::EmitLValueHelper(const Expr *E,
       // Defend against branches out of gnu statement expressions surrounded by
       // cleanups.
       Address Addr = LV.getAddress(*this);
-      llvm::Value *V = Addr.getPointer();
+      llvm::Value *V = Addr.getBasePointer();
       Scope.ForceCleanup({&V});
-      return LValue::MakeAddr(Addr.withPointer(V, Addr.isKnownNonNull()),
-                              LV.getType(), getContext(), LV.getBaseInfo(),
-                              LV.getTBAAInfo());
+      Addr.replaceBasePointer(V);
+      return LValue::MakeAddr(Addr, LV.getType(), getContext(),
+                              LV.getBaseInfo(), LV.getTBAAInfo());
     }
     // FIXME: Is it possible to create an ExprWithCleanups that produces a
     // bitfield lvalue or some other non-simple lvalue?
@@ -1935,7 +1934,7 @@ llvm::Value *CodeGenFunction::EmitLoadOfScalar(Address Addr, bool Volatile,
                                                LValueBaseInfo BaseInfo,
                                                TBAAAccessInfo TBAAInfo,
                                                bool isNontemporal) {
-  if (auto *GV = dyn_cast<llvm::GlobalValue>(Addr.getPointer()))
+  if (auto *GV = dyn_cast<llvm::GlobalValue>(Addr.getBasePointer()))
     if (GV->isThreadLocal())
       Addr = Addr.withPointer(Builder.CreateThreadLocalAddress(GV),
                               NotKnownNonNull);
@@ -2045,8 +2044,9 @@ llvm::Value *CodeGenFunction::EmitFromMemory(llvm::Value *Value, QualType Ty) {
 
 // Convert the pointer of \p Addr to a pointer to a vector (the value type of
 // MatrixType), if it points to a array (the memory type of MatrixType).
-static Address MaybeConvertMatrixAddress(Address Addr, CodeGenFunction &CGF,
-                                         bool IsVector = true) {
+static RawAddress MaybeConvertMatrixAddress(RawAddress Addr,
+                                            CodeGenFunction &CGF,
+                                            bool IsVector = true) {
   auto *ArrayTy = dyn_cast<llvm::ArrayType>(Addr.getElementType());
   if (ArrayTy && IsVector) {
     auto *VectorTy = llvm::FixedVectorType::get(ArrayTy->getElementType(),
@@ -2083,7 +2083,7 @@ void CodeGenFunction::EmitStoreOfScalar(llvm::Value *Value, Address Addr,
                                         LValueBaseInfo BaseInfo,
                                         TBAAAccessInfo TBAAInfo,
                                         bool isInit, bool isNontemporal) {
-  if (auto *GV = dyn_cast<llvm::GlobalValue>(Addr.getPointer()))
+  if (auto *GV = dyn_cast<llvm::GlobalValue>(Addr.getBasePointer()))
     if (GV->isThreadLocal())
       Addr = Addr.withPointer(Builder.CreateThreadLocalAddress(GV),
                               NotKnownNonNull);
@@ -2438,14 +2438,12 @@ void CodeGenFunction::EmitStoreThroughLValue(RValue Src, LValue Dst,
       assert(Dst.getBaseIvarExp() && "BaseIvarExp is NULL");
       llvm::Type *ResultType = IntPtrTy;
       Address dst = EmitPointerWithAlignment(Dst.getBaseIvarExp());
-      llvm::Value *RHS = dst.getPointer();
+      llvm::Value *RHS = dst.emitRawPointer(*this);
       RHS = Builder.CreatePtrToInt(RHS, ResultType, "sub.ptr.rhs.cast");
-      llvm::Value *LHS =
-        Builder.CreatePtrToInt(LvalueDst.getPointer(), ResultType,
-                               "sub.ptr.lhs.cast");
+      llvm::Value *LHS = Builder.CreatePtrToInt(LvalueDst.emitRawPointer(*this),
+                                                ResultType, "sub.ptr.lhs.cast");
       llvm::Value *BytesBetween = Builder.CreateSub(LHS, RHS, "ivar.offset");
-      CGM.getObjCRuntime().EmitObjCIvarAssign(*this, src, dst,
-                                              BytesBetween);
+      CGM.getObjCRuntime().EmitObjCIvarAssign(*this, src, dst, BytesBetween);
     } else if (Dst.isGlobalObjCRef()) {
       CGM.getObjCRuntime().EmitObjCGlobalAssign(*this, src, LvalueDst,
                                                 Dst.isThreadLocalRef());
@@ -2776,12 +2774,9 @@ CodeGenFunction::EmitLoadOfReference(LValue RefLVal,
   llvm::LoadInst *Load =
       Builder.CreateLoad(RefLVal.getAddress(*this), RefLVal.isVolatile());
   CGM.DecorateInstructionWithTBAA(Load, RefLVal.getTBAAInfo());
-
-  QualType PointeeType = RefLVal.getType()->getPointeeType();
-  CharUnits Align = CGM.getNaturalTypeAlignment(
-      PointeeType, PointeeBaseInfo, PointeeTBAAInfo,
-      /* forPointeeType= */ true);
-  return Address(Load, ConvertTypeForMem(PointeeType), Align);
+  return makeNaturalAddressForPointer(Load, RefLVal.getType()->getPointeeType(),
+                                      CharUnits(), /*ForPointeeType=*/true,
+                                      PointeeBaseInfo, PointeeTBAAInfo);
 }
 
 LValue CodeGenFunction::EmitLoadOfReferenceLValue(LValue RefLVal) {
@@ -2798,10 +2793,9 @@ Address CodeGenFunction::EmitLoadOfPointer(Address Ptr,
                                            LValueBaseInfo *BaseInfo,
                                            TBAAAccessInfo *TBAAInfo) {
   llvm::Value *Addr = Builder.CreateLoad(Ptr);
-  return Address(Addr, ConvertTypeForMem(PtrTy->getPointeeType()),
-                 CGM.getNaturalTypeAlignment(PtrTy->getPointeeType(), BaseInfo,
-                                             TBAAInfo,
-                                             /*forPointeeType=*/true));
+  return makeNaturalAddressForPointer(Addr, PtrTy->getPointeeType(),
+                                      CharUnits(), /*ForPointeeType=*/true,
+                                      BaseInfo, TBAAInfo);
 }
 
 LValue CodeGenFunction::EmitLoadOfPointerLValue(Address PtrAddr,
@@ -2997,7 +2991,7 @@ LValue CodeGenFunction::EmitDeclRefLValue(const DeclRefExpr *E) {
                                         /* BaseInfo= */ nullptr,
                                         /* TBAAInfo= */ nullptr,
                                         /* forPointeeType= */ true);
-        Addr = Address(Val, ConvertTypeForMem(E->getType()), Alignment);
+        Addr = makeNaturalAddressForPointer(Val, T, Alignment);
       }
       return MakeAddrLValue(Addr, T, AlignmentSource::Decl);
     }
@@ -3029,11 +3023,12 @@ LValue CodeGenFunction::EmitDeclRefLValue(const DeclRefExpr *E) {
             EmitCapturedFieldLValue(*this, CapturedStmtInfo->lookup(VD),
                                     CapturedStmtInfo->getContextValue());
         Address LValueAddress = CapLVal.getAddress(*this);
-        CapLVal = MakeAddrLValue(
-            Address(LValueAddress.getPointer(), LValueAddress.getElementType(),
-                    getContext().getDeclAlign(VD)),
-            CapLVal.getType(), LValueBaseInfo(AlignmentSource::Decl),
-            CapLVal.getTBAAInfo());
+        CapLVal = MakeAddrLValue(Address(LValueAddress.emitRawPointer(*this),
+                                         LValueAddress.getElementType(),
+                                         getContext().getDeclAlign(VD)),
+                                 CapLVal.getType(),
+                                 LValueBaseInfo(AlignmentSource::Decl),
+                                 CapLVal.getTBAAInfo());
         // Mark lvalue as nontemporal if the variable is marked as nontemporal
         // in simd context.
         if (getLangOpts().OpenMP &&
@@ -3097,7 +3092,8 @@ LValue CodeGenFunction::EmitDeclRefLValue(const DeclRefExpr *E) {
     // Handle threadlocal function locals.
     if (VD->getTLSKind() != VarDecl::TLS_None)
       addr = addr.withPointer(
-          Builder.CreateThreadLocalAddress(addr.getPointer()), NotKnownNonNull);
+          Builder.CreateThreadLocalAddress(addr.getBasePointer()),
+          NotKnownNonNull);
 
     // Check for OpenMP threadprivate variables.
     if (getLangOpts().OpenMP && !getLangOpts().OpenMPSimd &&
@@ -3365,7 +3361,7 @@ llvm::Value *CodeGenFunction::EmitCheckValue(llvm::Value *V) {
 
   // Pointers are passed directly, everything else is passed by address.
   if (!V->getType()->isPointerTy()) {
-    Address Ptr = CreateDefaultAlignTempAlloca(V->getType());
+    RawAddress Ptr = CreateDefaultAlignTempAlloca(V->getType());
     Builder.CreateStore(V, Ptr);
     V = Ptr.getPointer();
   }
@@ -3949,6 +3945,21 @@ static llvm::Value *emitArraySubscriptGEP(CodeGenFunction &CGF,
   }
 }
 
+static Address emitArraySubscriptGEP(CodeGenFunction &CGF, Address addr,
+                                     ArrayRef<llvm::Value *> indices,
+                                     llvm::Type *elementType, bool inbounds,
+                                     bool signedIndices, SourceLocation loc,
+                                     CharUnits align,
+                                     const llvm::Twine &name = "arrayidx") {
+  if (inbounds) {
+    return CGF.EmitCheckedInBoundsGEP(addr, indices, elementType, signedIndices,
+                                      CodeGenFunction::NotSubtraction, loc,
+                                      align, name);
+  } else {
+    return CGF.Builder.CreateGEP(addr, indices, elementType, align, name);
+  }
+}
+
 static CharUnits getArrayElementAlign(CharUnits arrayAlign,
                                       llvm::Value *idx,
                                       CharUnits eltSize) {
@@ -4006,7 +4017,7 @@ static Address wrapWithBPFPreserveStaticOffset(CodeGenFunction &CGF,
 
   llvm::Function *Fn =
       CGF.CGM.getIntrinsic(llvm::Intrinsic::preserve_static_offset);
-  llvm::CallInst *Call = CGF.Builder.CreateCall(Fn, {Addr.getPointer()});
+  llvm::CallInst *Call = CGF.Builder.CreateCall(Fn, {Addr.emitRawPointer(CGF)});
   return Address(Call, Addr.getElementType(), Addr.getAlignment());
 }
 
@@ -4070,7 +4081,7 @@ static Address emitArraySubscriptGEP(CodeGenFunction &CGF, Address addr,
   // We can use that to compute the best alignment of the element.
   CharUnits eltSize = CGF.getContext().getTypeSizeInChars(eltType);
   CharUnits eltAlign =
-    getArrayElementAlign(addr.getAlignment(), indices.back(), eltSize);
+      getArrayElementAlign(addr.getAlignment(), indices.back(), eltSize);
 
   if (hasBPFPreserveStaticOffset(Base))
     addr = wrapWithBPFPreserveStaticOffset(CGF, addr);
@@ -4079,20 +4090,20 @@ static Address emitArraySubscriptGEP(CodeGenFunction &CGF, Address addr,
   auto LastIndex = dyn_cast<llvm::ConstantInt>(indices.back());
   if (!LastIndex ||
       (!CGF.IsInPreservedAIRegion && !IsPreserveAIArrayBase(CGF, Base))) {
-    eltPtr = emitArraySubscriptGEP(
-        CGF, addr.getElementType(), addr.getPointer(), indices, inbounds,
-        signedIndices, loc, name);
+    addr = emitArraySubscriptGEP(CGF, addr, indices,
+                                 CGF.ConvertTypeForMem(eltType), inbounds,
+                                 signedIndices, loc, eltAlign, name);
     AddIVDepMetadata(CGF, arrayDecl, eltPtr);
+    return addr;
   } else {
     // Remember the original array subscript for bpf target
     unsigned idx = LastIndex->getZExtValue();
     llvm::DIType *DbgInfo = nullptr;
     if (arrayType)
       DbgInfo = CGF.getDebugInfo()->getOrCreateStandaloneType(*arrayType, loc);
-    eltPtr = CGF.Builder.CreatePreserveArrayAccessIndex(addr.getElementType(),
-                                                        addr.getPointer(),
-                                                        indices.size() - 1,
-                                                        idx, DbgInfo);
+    eltPtr = CGF.Builder.CreatePreserveArrayAccessIndex(
+        addr.getElementType(), addr.emitRawPointer(CGF), indices.size() - 1,
+        idx, DbgInfo);
   }
 
   return Address(eltPtr, CGF.ConvertTypeForMem(eltType), eltAlign);
@@ -4261,8 +4272,8 @@ LValue CodeGenFunction::EmitArraySubscriptExpr(const ArraySubscriptExpr *E,
     CharUnits EltAlign =
       getArrayElementAlign(Addr.getAlignment(), Idx, InterfaceSize);
     llvm::Value *EltPtr =
-        emitArraySubscriptGEP(*this, Int8Ty, Addr.getPointer(), ScaledIdx,
-                              false, SignedIndices, E->getExprLoc());
+        emitArraySubscriptGEP(*this, Int8Ty, Addr.emitRawPointer(*this),
+                              ScaledIdx, false, SignedIndices, E->getExprLoc());
     Addr = Address(EltPtr, OrigBaseElemTy, EltAlign);
   } else if (const Expr *Array = isSimpleArrayDecayOperand(E->getBase())) {
     // If this is A[i] where A is an array, the frontend will have decayed the
@@ -4314,7 +4325,7 @@ LValue CodeGenFunction::EmitArraySubscriptExpr(const ArraySubscriptExpr *E,
 
             llvm::Type *CountTy = ConvertType(CountFD->getType());
             llvm::Value *Res = Builder.CreateInBoundsGEP(
-                Int8Ty, Addr.getPointer(),
+                Int8Ty, Addr.emitRawPointer(*this),
                 Builder.getInt32(OffsetDiff.getQuantity()), ".counted_by.gep");
             Res = Builder.CreateAlignedLoad(CountTy, Res, getIntAlign(),
                                             ".counted_by.load");
@@ -4569,9 +4580,9 @@ LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
     BaseInfo = ArrayLV.getBaseInfo();
     TBAAInfo = CGM.getTBAAInfoForSubobject(ArrayLV, ResultExprTy);
   } else {
-    Address Base = emitOMPArraySectionBase(*this, E->getBase(), BaseInfo,
-                                           TBAAInfo, BaseTy, ResultExprTy,
-                                           IsLowerBound);
+    Address Base =
+        emitOMPArraySectionBase(*this, E->getBase(), BaseInfo, TBAAInfo, BaseTy,
+                                ResultExprTy, IsLowerBound);
     EltPtr = emitArraySubscriptGEP(*this, Base, Idx, ResultExprTy,
                                    !getLangOpts().isSignedOverflowDefined(),
                                    /*signedIndices=*/false, E->getExprLoc());
@@ -4658,7 +4669,7 @@ LValue CodeGenFunction::EmitMemberExpr(const MemberExpr *E) {
       SkippedChecks.set(SanitizerKind::Alignment, true);
     if (IsBaseCXXThis || isa<DeclRefExpr>(BaseExpr))
       SkippedChecks.set(SanitizerKind::Null, true);
-    EmitTypeCheck(TCK_MemberAccess, E->getExprLoc(), Addr.getPointer(), PtrTy,
+    EmitTypeCheck(TCK_MemberAccess, E->getExprLoc(), Addr, PtrTy,
                   /*Alignment=*/CharUnits::Zero(), SkippedChecks);
     BaseLV = MakeAddrLValue(Addr, PtrTy, BaseInfo, TBAAInfo);
   } else
@@ -4707,8 +4718,8 @@ LValue CodeGenFunction::EmitLValueForLambdaField(const FieldDecl *Field,
       LambdaLV = EmitLoadOfReferenceLValue(AddrOfExplicitObject, D->getType(),
                                            AlignmentSource::Decl);
     else
-      LambdaLV = MakeNaturalAlignAddrLValue(AddrOfExplicitObject.getPointer(),
-                                            D->getType().getNonReferenceType());
+      LambdaLV = MakeAddrLValue(AddrOfExplicitObject,
+                                D->getType().getNonReferenceType());
   } else {
     QualType LambdaTagType = getContext().getTagDeclType(Field->getParent());
     LambdaLV = MakeNaturalAlignAddrLValue(ThisValue, LambdaTagType);
@@ -4898,7 +4909,8 @@ LValue CodeGenFunction::EmitLValueForField(LValue base,
       // information provided by invariant.group.  This is because accessing
       // fields may leak the real address of dynamic object, which could result
       // in miscompilation when leaked pointer would be compared.
-      auto *stripped = Builder.CreateStripInvariantGroup(addr.getPointer());
+      auto *stripped =
+          Builder.CreateStripInvariantGroup(addr.emitRawPointer(*this));
       addr = Address(stripped, addr.getElementType(), addr.getAlignment());
     }
   }
@@ -4917,10 +4929,11 @@ LValue CodeGenFunction::EmitLValueForField(LValue base,
       // Remember the original union field index
       llvm::DIType *DbgInfo = getDebugInfo()->getOrCreateStandaloneType(base.getType(),
           rec->getLocation());
-      addr = Address(
-          Builder.CreatePreserveUnionAccessIndex(
-              addr.getPointer(), getDebugInfoFIndex(rec, field->getFieldIndex()), DbgInfo),
-          addr.getElementType(), addr.getAlignment());
+      addr =
+          Address(Builder.CreatePreserveUnionAccessIndex(
+                      addr.emitRawPointer(*this),
+                      getDebugInfoFIndex(rec, field->getFieldIndex()), DbgInfo),
+                  addr.getElementType(), addr.getAlignment());
     }
 
     if (FieldType->isReferenceType())
@@ -5168,11 +5181,9 @@ LValue CodeGenFunction::EmitConditionalOperatorLValue(
   if (Info.LHS && Info.RHS) {
     Address lhsAddr = Info.LHS->getAddress(*this);
     Address rhsAddr = Info.RHS->getAddress(*this);
-    llvm::PHINode *phi = Builder.CreatePHI(lhsAddr.getType(), 2, "cond-lvalue");
-    phi->addIncoming(lhsAddr.getPointer(), Info.lhsBlock);
-    phi->addIncoming(rhsAddr.getPointer(), Info.rhsBlock);
-    Address result(phi, lhsAddr.getElementType(),
-                   std::min(lhsAddr.getAlignment(), rhsAddr.getAlignment()));
+    Address result = mergeAddressesInConditionalExpr(
+        lhsAddr, rhsAddr, Info.lhsBlock, Info.rhsBlock,
+        Builder.GetInsertBlock(), expr->getType());
     AlignmentSource alignSource =
         std::max(Info.LHS->getBaseInfo().getAlignmentSource(),
                  Info.RHS->getBaseInfo().getAlignmentSource());
@@ -5260,7 +5271,7 @@ LValue CodeGenFunction::EmitCastLValue(const CastExpr *E) {
     LValue LV = EmitLValue(E->getSubExpr());
     Address V = LV.getAddress(*this);
     const auto *DCE = cast<CXXDynamicCastExpr>(E);
-    return MakeNaturalAlignAddrLValue(EmitDynamicCast(V, DCE), E->getType());
+    return MakeNaturalAlignRawAddrLValue(EmitDynamicCast(V, DCE), E->getType());
   }
 
   case CK_ConstructorConversion:
@@ -5325,8 +5336,8 @@ LValue CodeGenFunction::EmitCastLValue(const CastExpr *E) {
     // C++11 [expr.static.cast]p2: Behavior is undefined if a downcast is
     // performed and the object is not of the derived type.
     if (sanitizePerformTypeCheck())
-      EmitTypeCheck(TCK_DowncastReference, E->getExprLoc(),
-                    Derived.getPointer(), E->getType());
+      EmitTypeCheck(TCK_DowncastReference, E->getExprLoc(), Derived,
+                    E->getType());
 
     if (SanOpts.has(SanitizerKind::CFIDerivedCast))
       EmitVTablePtrCheckForCast(E->getType(), Derived,
@@ -5715,7 +5726,7 @@ LValue CodeGenFunction::EmitCXXConstructLValue(const CXXConstructExpr *E) {
 
 LValue
 CodeGenFunction::EmitCXXTypeidLValue(const CXXTypeidExpr *E) {
-  return MakeNaturalAlignAddrLValue(EmitCXXTypeidExpr(E), E->getType());
+  return MakeNaturalAlignRawAddrLValue(EmitCXXTypeidExpr(E), E->getType());
 }
 
 Address CodeGenFunction::EmitCXXUuidofExpr(const CXXUuidofExpr *E) {
diff --git a/clang/lib/CodeGen/CGExprAgg.cpp b/clang/lib/CodeGen/CGExprAgg.cpp
index f5cea6e6eaf65..fad8d405e95c4 100644
--- a/clang/lib/CodeGen/CGExprAgg.cpp
+++ b/clang/lib/CodeGen/CGExprAgg.cpp
@@ -293,10 +293,10 @@ void AggExprEmitter::withReturnValueSlot(
   // Otherwise, EmitCall will emit its own, notice that it's "unused", and end
   // its lifetime before we have the chance to emit a proper destructor call.
   bool UseTemp = Dest.isPotentiallyAliased() || Dest.requiresGCollection() ||
-                 (RequiresDestruction && !Dest.getAddress().isValid());
+                 (RequiresDestruction && Dest.isIgnored());
 
   Address RetAddr = Address::invalid();
-  Address RetAllocaAddr = Address::invalid();
+  RawAddress RetAllocaAddr = RawAddress::invalid();
 
   EHScopeStack::stable_iterator LifetimeEndBlock;
   llvm::Value *LifetimeSizePtr = nullptr;
@@ -328,7 +328,8 @@ void AggExprEmitter::withReturnValueSlot(
   if (!UseTemp)
     return;
 
-  assert(Dest.isIgnored() || Dest.getPointer() != Src.getAggregatePointer());
+  assert(Dest.isIgnored() || Dest.emitRawPointer(CGF) !=
+                                 Src.getAggregatePointer(E->getType(), CGF));
   EmitFinalDestCopy(E->getType(), Src);
 
   if (!RequiresDestruction && LifetimeStartInst) {
@@ -447,7 +448,8 @@ AggExprEmitter::VisitCXXStdInitializerListExpr(CXXStdInitializerListExpr *E) {
   llvm::Value *Zero = llvm::ConstantInt::get(CGF.PtrDiffTy, 0);
   llvm::Value *IdxStart[] = { Zero, Zero };
   llvm::Value *ArrayStart = Builder.CreateInBoundsGEP(
-      ArrayPtr.getElementType(), ArrayPtr.getPointer(), IdxStart, "arraystart");
+      ArrayPtr.getElementType(), ArrayPtr.emitRawPointer(CGF), IdxStart,
+      "arraystart");
   CGF.EmitStoreThroughLValue(RValue::get(ArrayStart), Start);
   ++Field;
 
@@ -464,7 +466,8 @@ AggExprEmitter::VisitCXXStdInitializerListExpr(CXXStdInitializerListExpr *E) {
     // End pointer.
     llvm::Value *IdxEnd[] = { Zero, Size };
     llvm::Value *ArrayEnd = Builder.CreateInBoundsGEP(
-        ArrayPtr.getElementType(), ArrayPtr.getPointer(), IdxEnd, "arrayend");
+        ArrayPtr.getElementType(), ArrayPtr.emitRawPointer(CGF), IdxEnd,
+        "arrayend");
     CGF.EmitStoreThroughLValue(RValue::get(ArrayEnd), EndOrLength);
   } else if (Ctx.hasSameType(Field->getType(), Ctx.getSizeType())) {
     // Length.
@@ -515,9 +518,9 @@ void AggExprEmitter::EmitArrayInit(Address DestPtr, llvm::ArrayType *AType,
   // down a level.
   llvm::Value *zero = llvm::ConstantInt::get(CGF.SizeTy, 0);
   llvm::Value *indices[] = { zero, zero };
-  llvm::Value *begin = Builder.CreateInBoundsGEP(
-      DestPtr.getElementType(), DestPtr.getPointer(), indices,
-      "arrayinit.begin");
+  llvm::Value *begin = Builder.CreateInBoundsGEP(DestPtr.getElementType(),
+                                                 DestPtr.emitRawPointer(CGF),
+                                                 indices, "arrayinit.begin");
 
   CharUnits elementSize = CGF.getContext().getTypeSizeInChars(elementType);
   CharUnits elementAlign =
@@ -1061,7 +1064,7 @@ void AggExprEmitter::VisitBinCmp(const BinaryOperator *E) {
     if (RV.isScalar())
       return {RV.getScalarVal(), nullptr};
     if (RV.isAggregate())
-      return {RV.getAggregatePointer(), nullptr};
+      return {RV.getAggregatePointer(E->getType(), CGF), nullptr};
     assert(RV.isComplex());
     return RV.getComplexVal();
   };
@@ -1823,7 +1826,7 @@ void AggExprEmitter::VisitCXXParenListOrInitListExpr(
     // else, clean it up for -O0 builds and general tidiness.
     if (!pushedCleanup && LV.isSimple())
       if (llvm::GetElementPtrInst *GEP =
-              dyn_cast<llvm::GetElementPtrInst>(LV.getPointer(CGF)))
+              dyn_cast<llvm::GetElementPtrInst>(LV.emitRawPointer(CGF)))
         if (GEP->use_empty())
           GEP->eraseFromParent();
   }
@@ -1854,9 +1857,9 @@ void AggExprEmitter::VisitArrayInitLoopExpr(const ArrayInitLoopExpr *E,
   // destPtr is an array*. Construct an elementType* by drilling down a level.
   llvm::Value *zero = llvm::ConstantInt::get(CGF.SizeTy, 0);
   llvm::Value *indices[] = {zero, zero};
-  llvm::Value *begin = Builder.CreateInBoundsGEP(
-      destPtr.getElementType(), destPtr.getPointer(), indices,
-      "arrayinit.begin");
+  llvm::Value *begin = Builder.CreateInBoundsGEP(destPtr.getElementType(),
+                                                 destPtr.emitRawPointer(CGF),
+                                                 indices, "arrayinit.begin");
 
   // Prepare to special-case multidimensional array initialization: we avoid
   // emitting multiple destructor loops in that case.
diff --git a/clang/lib/CodeGen/CGExprCXX.cpp b/clang/lib/CodeGen/CGExprCXX.cpp
index 9629134b0c4d2..673ccef84d678 100644
--- a/clang/lib/CodeGen/CGExprCXX.cpp
+++ b/clang/lib/CodeGen/CGExprCXX.cpp
@@ -280,7 +280,8 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorMemberCallExpr(
     LValueBaseInfo BaseInfo;
     TBAAAccessInfo TBAAInfo;
     Address ThisValue = EmitPointerWithAlignment(Base, &BaseInfo, &TBAAInfo);
-    This = MakeAddrLValue(ThisValue, Base->getType(), BaseInfo, TBAAInfo);
+    This = MakeAddrLValue(ThisValue, Base->getType()->getPointeeType(),
+                          BaseInfo, TBAAInfo);
   } else {
     This = EmitLValue(Base);
   }
@@ -353,10 +354,12 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorMemberCallExpr(
     if (IsImplicitObjectCXXThis || isa<DeclRefExpr>(IOA))
       SkippedChecks.set(SanitizerKind::Null, true);
   }
-  EmitTypeCheck(CodeGenFunction::TCK_MemberCall, CallLoc,
-                This.getPointer(*this),
-                C.getRecordType(CalleeDecl->getParent()),
-                /*Alignment=*/CharUnits::Zero(), SkippedChecks);
+
+  if (sanitizePerformTypeCheck())
+    EmitTypeCheck(CodeGenFunction::TCK_MemberCall, CallLoc,
+                  This.emitRawPointer(*this),
+                  C.getRecordType(CalleeDecl->getParent()),
+                  /*Alignment=*/CharUnits::Zero(), SkippedChecks);
 
   // C++ [class.virtual]p12:
   //   Explicit qualification with the scope operator (5.1) suppresses the
@@ -455,7 +458,7 @@ CodeGenFunction::EmitCXXMemberPointerCallExpr(const CXXMemberCallExpr *E,
   else
     This = EmitLValue(BaseExpr, KnownNonNull).getAddress(*this);
 
-  EmitTypeCheck(TCK_MemberCall, E->getExprLoc(), This.getPointer(),
+  EmitTypeCheck(TCK_MemberCall, E->getExprLoc(), This.emitRawPointer(*this),
                 QualType(MPT->getClass(), 0));
 
   // Get the member function pointer.
@@ -1110,9 +1113,9 @@ void CodeGenFunction::EmitNewArrayInitializer(
       EndOfInit = CreateTempAlloca(BeginPtr.getType(), getPointerAlign(),
                                    "array.init.end");
       CleanupDominator =
-          Builder.CreateStore(BeginPtr.getPointer(), EndOfInit);
-      pushIrregularPartialArrayCleanup(BeginPtr.getPointer(), EndOfInit,
-                                       ElementType, ElementAlign,
+          Builder.CreateStore(BeginPtr.emitRawPointer(*this), EndOfInit);
+      pushIrregularPartialArrayCleanup(BeginPtr.emitRawPointer(*this),
+                                       EndOfInit, ElementType, ElementAlign,
                                        getDestroyer(DtorKind));
       Cleanup = EHStack.stable_begin();
     }
@@ -1124,16 +1127,17 @@ void CodeGenFunction::EmitNewArrayInitializer(
       // element.  TODO: some of these stores can be trivially
       // observed to be unnecessary.
       if (EndOfInit.isValid()) {
-        Builder.CreateStore(CurPtr.getPointer(), EndOfInit);
+        Builder.CreateStore(CurPtr.emitRawPointer(*this), EndOfInit);
       }
       // FIXME: If the last initializer is an incomplete initializer list for
       // an array, and we have an array filler, we can fold together the two
       // initialization loops.
       StoreAnyExprIntoOneUnit(*this, IE, IE->getType(), CurPtr,
                               AggValueSlot::DoesNotOverlap);
-      CurPtr = Address(Builder.CreateInBoundsGEP(
-                           CurPtr.getElementType(), CurPtr.getPointer(),
-                           Builder.getSize(1), "array.exp.next"),
+      CurPtr = Address(Builder.CreateInBoundsGEP(CurPtr.getElementType(),
+                                                 CurPtr.emitRawPointer(*this),
+                                                 Builder.getSize(1),
+                                                 "array.exp.next"),
                        CurPtr.getElementType(),
                        StartAlign.alignmentAtOffset((++i) * ElementSize));
     }
@@ -1187,7 +1191,7 @@ void CodeGenFunction::EmitNewArrayInitializer(
     // FIXME: Share this cleanup with the constructor call emission rather than
     // having it create a cleanup of its own.
     if (EndOfInit.isValid())
-      Builder.CreateStore(CurPtr.getPointer(), EndOfInit);
+      Builder.CreateStore(CurPtr.emitRawPointer(*this), EndOfInit);
 
     // Emit a constructor call loop to initialize the remaining elements.
     if (InitListElements)
@@ -1250,15 +1254,15 @@ void CodeGenFunction::EmitNewArrayInitializer(
   llvm::BasicBlock *ContBB = createBasicBlock("new.loop.end");
 
   // Find the end of the array, hoisted out of the loop.
-  llvm::Value *EndPtr =
-    Builder.CreateInBoundsGEP(BeginPtr.getElementType(), BeginPtr.getPointer(),
-                              NumElements, "array.end");
+  llvm::Value *EndPtr = Builder.CreateInBoundsGEP(
+      BeginPtr.getElementType(), BeginPtr.emitRawPointer(*this), NumElements,
+      "array.end");
 
   // If the number of elements isn't constant, we have to now check if there is
   // anything left to initialize.
   if (!ConstNum) {
-    llvm::Value *IsEmpty =
-      Builder.CreateICmpEQ(CurPtr.getPointer(), EndPtr, "array.isempty");
+    llvm::Value *IsEmpty = Builder.CreateICmpEQ(CurPtr.emitRawPointer(*this),
+                                                EndPtr, "array.isempty");
     Builder.CreateCondBr(IsEmpty, ContBB, LoopBB);
   }
 
@@ -1268,18 +1272,18 @@ void CodeGenFunction::EmitNewArrayInitializer(
   // Set up the current-element phi.
   llvm::PHINode *CurPtrPhi =
       Builder.CreatePHI(CurPtr.getType(), 2, "array.cur");
-  CurPtrPhi->addIncoming(CurPtr.getPointer(), EntryBB);
+  CurPtrPhi->addIncoming(CurPtr.emitRawPointer(*this), EntryBB);
 
   CurPtr = Address(CurPtrPhi, CurPtr.getElementType(), ElementAlign);
 
   // Store the new Cleanup position for irregular Cleanups.
   if (EndOfInit.isValid())
-    Builder.CreateStore(CurPtr.getPointer(), EndOfInit);
+    Builder.CreateStore(CurPtr.emitRawPointer(*this), EndOfInit);
 
   // Enter a partial-destruction Cleanup if necessary.
   if (!CleanupDominator && needsEHCleanup(DtorKind)) {
-    llvm::Value *BeginPtrRaw = BeginPtr.getPointer();
-    llvm::Value *CurPtrRaw = CurPtr.getPointer();
+    llvm::Value *BeginPtrRaw = BeginPtr.emitRawPointer(*this);
+    llvm::Value *CurPtrRaw = CurPtr.emitRawPointer(*this);
     pushRegularPartialArrayCleanup(BeginPtrRaw, CurPtrRaw, ElementType,
                                    ElementAlign, getDestroyer(DtorKind));
     Cleanup = EHStack.stable_begin();
@@ -1297,9 +1301,8 @@ void CodeGenFunction::EmitNewArrayInitializer(
   }
 
   // Advance to the next element by adjusting the pointer type as necessary.
-  llvm::Value *NextPtr =
-    Builder.CreateConstInBoundsGEP1_32(ElementTy, CurPtr.getPointer(), 1,
-                                       "array.next");
+  llvm::Value *NextPtr = Builder.CreateConstInBoundsGEP1_32(
+      ElementTy, CurPtr.emitRawPointer(*this), 1, "array.next");
 
   // Check whether we've gotten to the end of the array and, if so,
   // exit the loop.
@@ -1525,14 +1528,9 @@ static void EnterNewDeleteCleanup(CodeGenFunction &CGF,
 
     typedef CallDeleteDuringNew<DirectCleanupTraits> DirectCleanup;
 
-    DirectCleanup *Cleanup = CGF.EHStack
-      .pushCleanupWithExtra<DirectCleanup>(EHCleanup,
-                                           E->getNumPlacementArgs(),
-                                           E->getOperatorDelete(),
-                                           NewPtr.getPointer(),
-                                           AllocSize,
-                                           E->passAlignment(),
-                                           AllocAlign);
+    DirectCleanup *Cleanup = CGF.EHStack.pushCleanupWithExtra<DirectCleanup>(
+        EHCleanup, E->getNumPlacementArgs(), E->getOperatorDelete(),
+        NewPtr.emitRawPointer(CGF), AllocSize, E->passAlignment(), AllocAlign);
     for (unsigned I = 0, N = E->getNumPlacementArgs(); I != N; ++I) {
       auto &Arg = NewArgs[I + NumNonPlacementArgs];
       Cleanup->setPlacementArg(I, Arg.getRValue(CGF), Arg.Ty);
@@ -1543,7 +1541,7 @@ static void EnterNewDeleteCleanup(CodeGenFunction &CGF,
 
   // Otherwise, we need to save all this stuff.
   DominatingValue<RValue>::saved_type SavedNewPtr =
-    DominatingValue<RValue>::save(CGF, RValue::get(NewPtr.getPointer()));
+      DominatingValue<RValue>::save(CGF, RValue::get(NewPtr, CGF));
   DominatingValue<RValue>::saved_type SavedAllocSize =
     DominatingValue<RValue>::save(CGF, RValue::get(AllocSize));
 
@@ -1620,14 +1618,14 @@ llvm::Value *CodeGenFunction::EmitCXXNewExpr(const CXXNewExpr *E) {
     // In these cases, discard the computed alignment and use the
     // formal alignment of the allocated type.
     if (BaseInfo.getAlignmentSource() != AlignmentSource::Decl)
-      allocation = allocation.withAlignment(allocAlign);
+      allocation.setAlignment(allocAlign);
 
     // Set up allocatorArgs for the call to operator delete if it's not
     // the reserved global operator.
     if (E->getOperatorDelete() &&
         !E->getOperatorDelete()->isReservedGlobalPlacementOperator()) {
       allocatorArgs.add(RValue::get(allocSize), getContext().getSizeType());
-      allocatorArgs.add(RValue::get(allocation.getPointer()), arg->getType());
+      allocatorArgs.add(RValue::get(allocation, *this), arg->getType());
     }
 
   } else {
@@ -1715,8 +1713,7 @@ llvm::Value *CodeGenFunction::EmitCXXNewExpr(const CXXNewExpr *E) {
     llvm::BasicBlock *notNullBB = createBasicBlock("new.notnull");
     contBB = createBasicBlock("new.cont");
 
-    llvm::Value *isNull =
-      Builder.CreateIsNull(allocation.getPointer(), "new.isnull");
+    llvm::Value *isNull = Builder.CreateIsNull(allocation, "new.isnull");
     Builder.CreateCondBr(isNull, contBB, notNullBB);
     EmitBlock(notNullBB);
   }
@@ -1762,12 +1759,12 @@ llvm::Value *CodeGenFunction::EmitCXXNewExpr(const CXXNewExpr *E) {
   SkippedChecks.set(SanitizerKind::Null, nullCheck);
   EmitTypeCheck(CodeGenFunction::TCK_ConstructorCall,
                 E->getAllocatedTypeSourceInfo()->getTypeLoc().getBeginLoc(),
-                result.getPointer(), allocType, result.getAlignment(),
-                SkippedChecks, numElements);
+                result, allocType, result.getAlignment(), SkippedChecks,
+                numElements);
 
   EmitNewInitializer(*this, E, allocType, elementTy, result, numElements,
                      allocSizeWithoutCookie);
-  llvm::Value *resultPtr = result.getPointer();
+  llvm::Value *resultPtr = result.emitRawPointer(*this);
   if (E->isArray()) {
     // NewPtr is a pointer to the base element type.  If we're
     // allocating an array of arrays, we'll need to cast back to the
@@ -1911,7 +1908,8 @@ static void EmitDestroyingObjectDelete(CodeGenFunction &CGF,
     CGF.CGM.getCXXABI().emitVirtualObjectDelete(CGF, DE, Ptr, ElementType,
                                                 Dtor);
   else
-    CGF.EmitDeleteCall(DE->getOperatorDelete(), Ptr.getPointer(), ElementType);
+    CGF.EmitDeleteCall(DE->getOperatorDelete(), Ptr.emitRawPointer(CGF),
+                       ElementType);
 }
 
 /// Emit the code for deleting a single object.
@@ -1927,8 +1925,7 @@ static bool EmitObjectDelete(CodeGenFunction &CGF,
   //   dynamic type, the static type shall be a base class of the dynamic type
   //   of the object to be deleted and the static type shall have a virtual
   //   destructor or the behavior is undefined.
-  CGF.EmitTypeCheck(CodeGenFunction::TCK_MemberCall,
-                    DE->getExprLoc(), Ptr.getPointer(),
+  CGF.EmitTypeCheck(CodeGenFunction::TCK_MemberCall, DE->getExprLoc(), Ptr,
                     ElementType);
 
   const FunctionDecl *OperatorDelete = DE->getOperatorDelete();
@@ -1977,9 +1974,8 @@ static bool EmitObjectDelete(CodeGenFunction &CGF,
   // Make sure that we call delete even if the dtor throws.
   // This doesn't have to a conditional cleanup because we're going
   // to pop it off in a second.
-  CGF.EHStack.pushCleanup<CallObjectDelete>(NormalAndEHCleanup,
-                                            Ptr.getPointer(),
-                                            OperatorDelete, ElementType);
+  CGF.EHStack.pushCleanup<CallObjectDelete>(
+      NormalAndEHCleanup, Ptr.emitRawPointer(CGF), OperatorDelete, ElementType);
 
   if (Dtor)
     CGF.EmitCXXDestructorCall(Dtor, Dtor_Complete,
@@ -2066,7 +2062,7 @@ static void EmitArrayDelete(CodeGenFunction &CGF,
     CharUnits elementAlign =
       deletedPtr.getAlignment().alignmentOfArrayElement(elementSize);
 
-    llvm::Value *arrayBegin = deletedPtr.getPointer();
+    llvm::Value *arrayBegin = deletedPtr.emitRawPointer(CGF);
     llvm::Value *arrayEnd = CGF.Builder.CreateInBoundsGEP(
       deletedPtr.getElementType(), arrayBegin, numElements, "delete.end");
 
@@ -2097,7 +2093,7 @@ void CodeGenFunction::EmitCXXDeleteExpr(const CXXDeleteExpr *E) {
   llvm::BasicBlock *DeleteNotNull = createBasicBlock("delete.notnull");
   llvm::BasicBlock *DeleteEnd = createBasicBlock("delete.end");
 
-  llvm::Value *IsNull = Builder.CreateIsNull(Ptr.getPointer(), "isnull");
+  llvm::Value *IsNull = Builder.CreateIsNull(Ptr, "isnull");
 
   Builder.CreateCondBr(IsNull, DeleteEnd, DeleteNotNull);
   EmitBlock(DeleteNotNull);
@@ -2132,10 +2128,8 @@ void CodeGenFunction::EmitCXXDeleteExpr(const CXXDeleteExpr *E) {
       GEP.push_back(Zero);
     }
 
-    Ptr = Address(Builder.CreateInBoundsGEP(Ptr.getElementType(),
-                                            Ptr.getPointer(), GEP, "del.first"),
-                  ConvertTypeForMem(DeleteTy), Ptr.getAlignment(),
-                  Ptr.isKnownNonNull());
+    Ptr = Builder.CreateInBoundsGEP(Ptr, GEP, ConvertTypeForMem(DeleteTy),
+                                    Ptr.getAlignment(), "del.first");
   }
 
   assert(ConvertTypeForMem(DeleteTy) == Ptr.getElementType());
@@ -2193,7 +2187,7 @@ static llvm::Value *EmitTypeidFromVTable(CodeGenFunction &CGF, const Expr *E,
   //   destruction and the static type of the operand is neither the constructor
   //   or destructor’s class nor one of its bases, the behavior is undefined.
   CGF.EmitTypeCheck(CodeGenFunction::TCK_DynamicOperation, E->getExprLoc(),
-                    ThisPtr.getPointer(), SrcRecordTy);
+                    ThisPtr, SrcRecordTy);
 
   // C++ [expr.typeid]p2:
   //   If the glvalue expression is obtained by applying the unary * operator to
@@ -2209,7 +2203,7 @@ static llvm::Value *EmitTypeidFromVTable(CodeGenFunction &CGF, const Expr *E,
         CGF.createBasicBlock("typeid.bad_typeid");
     llvm::BasicBlock *EndBlock = CGF.createBasicBlock("typeid.end");
 
-    llvm::Value *IsNull = CGF.Builder.CreateIsNull(ThisPtr.getPointer());
+    llvm::Value *IsNull = CGF.Builder.CreateIsNull(ThisPtr);
     CGF.Builder.CreateCondBr(IsNull, BadTypeidBlock, EndBlock);
 
     CGF.EmitBlock(BadTypeidBlock);
@@ -2295,8 +2289,7 @@ llvm::Value *CodeGenFunction::EmitDynamicCast(Address ThisAddr,
   //   construction or destruction and the static type of the operand is not a
   //   pointer to or object of the constructor or destructor’s own class or one
   //   of its bases, the dynamic_cast results in undefined behavior.
-  EmitTypeCheck(TCK_DynamicOperation, DCE->getExprLoc(), ThisAddr.getPointer(),
-                SrcRecordTy);
+  EmitTypeCheck(TCK_DynamicOperation, DCE->getExprLoc(), ThisAddr, SrcRecordTy);
 
   if (DCE->isAlwaysNull()) {
     if (llvm::Value *T = EmitDynamicCastToNull(*this, DestTy)) {
@@ -2331,7 +2324,7 @@ llvm::Value *CodeGenFunction::EmitDynamicCast(Address ThisAddr,
     CastNull = createBasicBlock("dynamic_cast.null");
     CastNotNull = createBasicBlock("dynamic_cast.notnull");
 
-    llvm::Value *IsNull = Builder.CreateIsNull(ThisAddr.getPointer());
+    llvm::Value *IsNull = Builder.CreateIsNull(ThisAddr);
     Builder.CreateCondBr(IsNull, CastNull, CastNotNull);
     EmitBlock(CastNotNull);
   }
diff --git a/clang/lib/CodeGen/CGExprConstant.cpp b/clang/lib/CodeGen/CGExprConstant.cpp
index 4a3a6b41037ca..c924660c5a91c 100644
--- a/clang/lib/CodeGen/CGExprConstant.cpp
+++ b/clang/lib/CodeGen/CGExprConstant.cpp
@@ -800,8 +800,8 @@ bool ConstStructBuilder::Build(const APValue &Val, const RecordDecl *RD,
     // Add a vtable pointer, if we need one and it hasn't already been added.
     if (Layout.hasOwnVFPtr()) {
       llvm::Constant *VTableAddressPoint =
-          CGM.getCXXABI().getVTableAddressPointForConstExpr(
-              BaseSubobject(CD, Offset), VTableClass);
+          CGM.getCXXABI().getVTableAddressPoint(BaseSubobject(CD, Offset),
+                                                VTableClass);
       if (!AppendBytes(Offset, VTableAddressPoint))
         return false;
     }
diff --git a/clang/lib/CodeGen/CGExprScalar.cpp b/clang/lib/CodeGen/CGExprScalar.cpp
index bae8c7a266c2e..4bbc051b2c2a2 100644
--- a/clang/lib/CodeGen/CGExprScalar.cpp
+++ b/clang/lib/CodeGen/CGExprScalar.cpp
@@ -2408,7 +2408,7 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
     // performed and the object is not of the derived type.
     if (CGF.sanitizePerformTypeCheck())
       CGF.EmitTypeCheck(CodeGenFunction::TCK_DowncastPointer, CE->getExprLoc(),
-                        Derived.getPointer(), DestTy->getPointeeType());
+                        Derived, DestTy->getPointeeType());
 
     if (CGF.SanOpts.has(SanitizerKind::CFIDerivedCast))
       CGF.EmitVTablePtrCheckForCast(DestTy->getPointeeType(), Derived,
@@ -2416,13 +2416,14 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
                                     CodeGenFunction::CFITCK_DerivedCast,
                                     CE->getBeginLoc());
 
-    return Derived.getPointer();
+    return CGF.getAsNaturalPointerTo(Derived, CE->getType()->getPointeeType());
   }
   case CK_UncheckedDerivedToBase:
   case CK_DerivedToBase: {
     // The EmitPointerWithAlignment path does this fine; just discard
     // the alignment.
-    return CGF.EmitPointerWithAlignment(CE).getPointer();
+    return CGF.getAsNaturalPointerTo(CGF.EmitPointerWithAlignment(CE),
+                                     CE->getType()->getPointeeType());
   }
 
   case CK_Dynamic: {
@@ -2432,7 +2433,8 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
   }
 
   case CK_ArrayToPointerDecay:
-    return CGF.EmitArrayToPointerDecay(E).getPointer();
+    return CGF.getAsNaturalPointerTo(CGF.EmitArrayToPointerDecay(E),
+                                     CE->getType()->getPointeeType());
   case CK_FunctionToPointerDecay:
     return EmitLValue(E).getPointer(CGF);
 
@@ -5811,3 +5813,16 @@ CodeGenFunction::EmitCheckedInBoundsGEP(llvm::Type *ElemTy, Value *Ptr,
 
   return GEPVal;
 }
+
+Address CodeGenFunction::EmitCheckedInBoundsGEP(
+    Address Addr, ArrayRef<Value *> IdxList, llvm::Type *elementType,
+    bool SignedIndices, bool IsSubtraction, SourceLocation Loc, CharUnits Align,
+    const Twine &Name) {
+  if (!SanOpts.has(SanitizerKind::PointerOverflow))
+    return Builder.CreateInBoundsGEP(Addr, IdxList, elementType, Align, Name);
+
+  return RawAddress(
+      EmitCheckedInBoundsGEP(Addr.getElementType(), Addr.emitRawPointer(*this),
+                             IdxList, SignedIndices, IsSubtraction, Loc, Name),
+      elementType, Align);
+}
diff --git a/clang/lib/CodeGen/CGNonTrivialStruct.cpp b/clang/lib/CodeGen/CGNonTrivialStruct.cpp
index 75c1d7fbea840..8fade0fac21e9 100644
--- a/clang/lib/CodeGen/CGNonTrivialStruct.cpp
+++ b/clang/lib/CodeGen/CGNonTrivialStruct.cpp
@@ -366,7 +366,7 @@ template <class Derived> struct GenFuncBase {
     llvm::Value *SizeInBytes =
         CGF.Builder.CreateNUWMul(BaseEltSizeVal, NumElts);
     llvm::Value *DstArrayEnd = CGF.Builder.CreateInBoundsGEP(
-        CGF.Int8Ty, DstAddr.getPointer(), SizeInBytes);
+        CGF.Int8Ty, DstAddr.emitRawPointer(CGF), SizeInBytes);
     llvm::BasicBlock *PreheaderBB = CGF.Builder.GetInsertBlock();
 
     // Create the header block and insert the phi instructions.
@@ -376,7 +376,7 @@ template <class Derived> struct GenFuncBase {
 
     for (unsigned I = 0; I < N; ++I) {
       PHIs[I] = CGF.Builder.CreatePHI(CGF.CGM.Int8PtrPtrTy, 2, "addr.cur");
-      PHIs[I]->addIncoming(StartAddrs[I].getPointer(), PreheaderBB);
+      PHIs[I]->addIncoming(StartAddrs[I].emitRawPointer(CGF), PreheaderBB);
     }
 
     // Create the exit and loop body blocks.
@@ -410,7 +410,7 @@ template <class Derived> struct GenFuncBase {
       // Instrs to update the destination and source addresses.
       // Update phi instructions.
       NewAddrs[I] = getAddrWithOffset(NewAddrs[I], EltSize);
-      PHIs[I]->addIncoming(NewAddrs[I].getPointer(), LoopBB);
+      PHIs[I]->addIncoming(NewAddrs[I].emitRawPointer(CGF), LoopBB);
     }
 
     // Insert an unconditional branch to the header block.
@@ -488,7 +488,7 @@ template <class Derived> struct GenFuncBase {
 
     for (unsigned I = 0; I < N; ++I) {
       Alignments[I] = Addrs[I].getAlignment();
-      Ptrs[I] = Addrs[I].getPointer();
+      Ptrs[I] = Addrs[I].emitRawPointer(CallerCGF);
     }
 
     if (llvm::Function *F =
diff --git a/clang/lib/CodeGen/CGObjC.cpp b/clang/lib/CodeGen/CGObjC.cpp
index f9eee5963c3bd..61d8f11c078aa 100644
--- a/clang/lib/CodeGen/CGObjC.cpp
+++ b/clang/lib/CodeGen/CGObjC.cpp
@@ -94,8 +94,8 @@ CodeGenFunction::EmitObjCBoxedExpr(const ObjCBoxedExpr *E) {
     // and cast value to correct type
     Address Temporary = CreateMemTemp(SubExpr->getType());
     EmitAnyExprToMem(SubExpr, Temporary, Qualifiers(), /*isInit*/ true);
-    llvm::Value *BitCast =
-        Builder.CreateBitCast(Temporary.getPointer(), ConvertType(ArgQT));
+    llvm::Value *BitCast = Builder.CreateBitCast(
+        Temporary.emitRawPointer(*this), ConvertType(ArgQT));
     Args.add(RValue::get(BitCast), ArgQT);
 
     // Create char array to store type encoding
@@ -204,11 +204,11 @@ llvm::Value *CodeGenFunction::EmitObjCCollectionLiteral(const Expr *E,
   ObjCMethodDecl::param_const_iterator PI = MethodWithObjects->param_begin();
   const ParmVarDecl *argDecl = *PI++;
   QualType ArgQT = argDecl->getType().getUnqualifiedType();
-  Args.add(RValue::get(Objects.getPointer()), ArgQT);
+  Args.add(RValue::get(Objects, *this), ArgQT);
   if (DLE) {
     argDecl = *PI++;
     ArgQT = argDecl->getType().getUnqualifiedType();
-    Args.add(RValue::get(Keys.getPointer()), ArgQT);
+    Args.add(RValue::get(Keys, *this), ArgQT);
   }
   argDecl = *PI;
   ArgQT = argDecl->getType().getUnqualifiedType();
@@ -828,7 +828,7 @@ static void emitStructGetterCall(CodeGenFunction &CGF, ObjCIvarDecl *ivar,
   //                  sizeof (Type of Ivar), isAtomic, false);
   CallArgList args;
 
-  llvm::Value *dest = CGF.ReturnValue.getPointer();
+  llvm::Value *dest = CGF.ReturnValue.emitRawPointer(CGF);
   args.add(RValue::get(dest), Context.VoidPtrTy);
   args.add(RValue::get(src), Context.VoidPtrTy);
 
@@ -1148,8 +1148,8 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
       callCStructCopyConstructor(Dst, Src);
     } else {
       ObjCIvarDecl *ivar = propImpl->getPropertyIvarDecl();
-      emitCPPObjectAtomicGetterCall(*this, ReturnValue.getPointer(), ivar,
-                                    AtomicHelperFn);
+      emitCPPObjectAtomicGetterCall(*this, ReturnValue.emitRawPointer(*this),
+                                    ivar, AtomicHelperFn);
     }
     return;
   }
@@ -1164,7 +1164,7 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
     }
     else {
       ObjCIvarDecl *ivar = propImpl->getPropertyIvarDecl();
-      emitCPPObjectAtomicGetterCall(*this, ReturnValue.getPointer(),
+      emitCPPObjectAtomicGetterCall(*this, ReturnValue.emitRawPointer(*this),
                                     ivar, AtomicHelperFn);
     }
     return;
@@ -1288,7 +1288,7 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
     case TEK_Scalar: {
       llvm::Value *value;
       if (propType->isReferenceType()) {
-        value = LV.getAddress(*this).getPointer();
+        value = LV.getAddress(*this).emitRawPointer(*this);
       } else {
         // We want to load and autoreleaseReturnValue ARC __weak ivars.
         if (LV.getQuals().getObjCLifetime() == Qualifiers::OCL_Weak) {
@@ -1821,16 +1821,14 @@ void CodeGenFunction::EmitObjCForCollectionStmt(const ObjCForCollectionStmt &S){
   CallArgList Args;
 
   // The first argument is a temporary of the enumeration-state type.
-  Args.add(RValue::get(StatePtr.getPointer()),
-           getContext().getPointerType(StateTy));
+  Args.add(RValue::get(StatePtr, *this), getContext().getPointerType(StateTy));
 
   // The second argument is a temporary array with space for NumItems
   // pointers.  We'll actually be loading elements from the array
   // pointer written into the control state; this buffer is so that
   // collections that *aren't* backed by arrays can still queue up
   // batches of elements.
-  Args.add(RValue::get(ItemsPtr.getPointer()),
-           getContext().getPointerType(ItemsTy));
+  Args.add(RValue::get(ItemsPtr, *this), getContext().getPointerType(ItemsTy));
 
   // The third argument is the capacity of that temporary array.
   llvm::Type *NSUIntegerTy = ConvertType(getContext().getNSUIntegerType());
@@ -2198,7 +2196,7 @@ static llvm::Value *emitARCLoadOperation(CodeGenFunction &CGF, Address addr,
   if (!fn)
     fn = getARCIntrinsic(IntID, CGF.CGM);
 
-  return CGF.EmitNounwindRuntimeCall(fn, addr.getPointer());
+  return CGF.EmitNounwindRuntimeCall(fn, addr.emitRawPointer(CGF));
 }
 
 /// Perform an operation having the following signature:
@@ -2216,9 +2214,8 @@ static llvm::Value *emitARCStoreOperation(CodeGenFunction &CGF, Address addr,
   llvm::Type *origType = value->getType();
 
   llvm::Value *args[] = {
-    CGF.Builder.CreateBitCast(addr.getPointer(), CGF.Int8PtrPtrTy),
-    CGF.Builder.CreateBitCast(value, CGF.Int8PtrTy)
-  };
+      CGF.Builder.CreateBitCast(addr.emitRawPointer(CGF), CGF.Int8PtrPtrTy),
+      CGF.Builder.CreateBitCast(value, CGF.Int8PtrTy)};
   llvm::CallInst *result = CGF.EmitNounwindRuntimeCall(fn, args);
 
   if (ignored) return nullptr;
@@ -2237,9 +2234,8 @@ static void emitARCCopyOperation(CodeGenFunction &CGF, Address dst, Address src,
     fn = getARCIntrinsic(IntID, CGF.CGM);
 
   llvm::Value *args[] = {
-    CGF.Builder.CreateBitCast(dst.getPointer(), CGF.Int8PtrPtrTy),
-    CGF.Builder.CreateBitCast(src.getPointer(), CGF.Int8PtrPtrTy)
-  };
+      CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF), CGF.Int8PtrPtrTy),
+      CGF.Builder.CreateBitCast(src.emitRawPointer(CGF), CGF.Int8PtrPtrTy)};
   CGF.EmitNounwindRuntimeCall(fn, args);
 }
 
@@ -2490,9 +2486,8 @@ llvm::Value *CodeGenFunction::EmitARCStoreStrongCall(Address addr,
     fn = getARCIntrinsic(llvm::Intrinsic::objc_storeStrong, CGM);
 
   llvm::Value *args[] = {
-    Builder.CreateBitCast(addr.getPointer(), Int8PtrPtrTy),
-    Builder.CreateBitCast(value, Int8PtrTy)
-  };
+      Builder.CreateBitCast(addr.emitRawPointer(*this), Int8PtrPtrTy),
+      Builder.CreateBitCast(value, Int8PtrTy)};
   EmitNounwindRuntimeCall(fn, args);
 
   if (ignored) return nullptr;
@@ -2643,7 +2638,7 @@ void CodeGenFunction::EmitARCDestroyWeak(Address addr) {
   if (!fn)
     fn = getARCIntrinsic(llvm::Intrinsic::objc_destroyWeak, CGM);
 
-  EmitNounwindRuntimeCall(fn, addr.getPointer());
+  EmitNounwindRuntimeCall(fn, addr.emitRawPointer(*this));
 }
 
 /// void \@objc_moveWeak(i8** %dest, i8** %src)
diff --git a/clang/lib/CodeGen/CGObjCGNU.cpp b/clang/lib/CodeGen/CGObjCGNU.cpp
index a36b0cdddaf0a..4e7f777ba1d91 100644
--- a/clang/lib/CodeGen/CGObjCGNU.cpp
+++ b/clang/lib/CodeGen/CGObjCGNU.cpp
@@ -706,7 +706,8 @@ class CGObjCGCC : public CGObjCGNU {
                               llvm::Value *cmd, MessageSendInfo &MSI) override {
     CGBuilderTy &Builder = CGF.Builder;
     llvm::Value *lookupArgs[] = {
-        EnforceType(Builder, ObjCSuper.getPointer(), PtrToObjCSuperTy), cmd};
+        EnforceType(Builder, ObjCSuper.emitRawPointer(CGF), PtrToObjCSuperTy),
+        cmd};
     return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
   }
 
@@ -761,8 +762,8 @@ class CGObjCGNUstep : public CGObjCGNU {
       llvm::FunctionCallee LookupFn = SlotLookupFn;
 
       // Store the receiver on the stack so that we can reload it later
-      Address ReceiverPtr =
-        CGF.CreateTempAlloca(Receiver->getType(), CGF.getPointerAlign());
+      RawAddress ReceiverPtr =
+          CGF.CreateTempAlloca(Receiver->getType(), CGF.getPointerAlign());
       Builder.CreateStore(Receiver, ReceiverPtr);
 
       llvm::Value *self;
@@ -778,9 +779,9 @@ class CGObjCGNUstep : public CGObjCGNU {
         LookupFn2->addParamAttr(0, llvm::Attribute::NoCapture);
 
       llvm::Value *args[] = {
-              EnforceType(Builder, ReceiverPtr.getPointer(), PtrToIdTy),
-              EnforceType(Builder, cmd, SelectorTy),
-              EnforceType(Builder, self, IdTy) };
+          EnforceType(Builder, ReceiverPtr.getPointer(), PtrToIdTy),
+          EnforceType(Builder, cmd, SelectorTy),
+          EnforceType(Builder, self, IdTy)};
       llvm::CallBase *slot = CGF.EmitRuntimeCallOrInvoke(LookupFn, args);
       slot->setOnlyReadsMemory();
       slot->setMetadata(msgSendMDKind, node);
@@ -800,7 +801,7 @@ class CGObjCGNUstep : public CGObjCGNU {
                                 llvm::Value *cmd,
                                 MessageSendInfo &MSI) override {
       CGBuilderTy &Builder = CGF.Builder;
-      llvm::Value *lookupArgs[] = {ObjCSuper.getPointer(), cmd};
+      llvm::Value *lookupArgs[] = {ObjCSuper.emitRawPointer(CGF), cmd};
 
       llvm::CallInst *slot =
         CGF.EmitNounwindRuntimeCall(SlotLookupSuperFn, lookupArgs);
@@ -1221,10 +1222,10 @@ class CGObjCGNUstep2 : public CGObjCGNUstep {
                               llvm::Value *cmd, MessageSendInfo &MSI) override {
     // Don't access the slot unless we're trying to cache the result.
     CGBuilderTy &Builder = CGF.Builder;
-    llvm::Value *lookupArgs[] = {CGObjCGNU::EnforceType(Builder,
-                                                        ObjCSuper.getPointer(),
-                                                        PtrToObjCSuperTy),
-                                 cmd};
+    llvm::Value *lookupArgs[] = {
+        CGObjCGNU::EnforceType(Builder, ObjCSuper.emitRawPointer(CGF),
+                               PtrToObjCSuperTy),
+        cmd};
     return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
   }
 
@@ -2186,7 +2187,8 @@ class CGObjCObjFW: public CGObjCGNU {
                               llvm::Value *cmd, MessageSendInfo &MSI) override {
     CGBuilderTy &Builder = CGF.Builder;
     llvm::Value *lookupArgs[] = {
-        EnforceType(Builder, ObjCSuper.getPointer(), PtrToObjCSuperTy), cmd,
+        EnforceType(Builder, ObjCSuper.emitRawPointer(CGF), PtrToObjCSuperTy),
+        cmd,
     };
 
     if (CGM.ReturnTypeUsesSRet(MSI.CallInfo))
@@ -4201,15 +4203,15 @@ void CGObjCGNU::EmitThrowStmt(CodeGenFunction &CGF,
 llvm::Value * CGObjCGNU::EmitObjCWeakRead(CodeGenFunction &CGF,
                                           Address AddrWeakObj) {
   CGBuilderTy &B = CGF.Builder;
-  return B.CreateCall(WeakReadFn,
-                      EnforceType(B, AddrWeakObj.getPointer(), PtrToIdTy));
+  return B.CreateCall(
+      WeakReadFn, EnforceType(B, AddrWeakObj.emitRawPointer(CGF), PtrToIdTy));
 }
 
 void CGObjCGNU::EmitObjCWeakAssign(CodeGenFunction &CGF,
                                    llvm::Value *src, Address dst) {
   CGBuilderTy &B = CGF.Builder;
   src = EnforceType(B, src, IdTy);
-  llvm::Value *dstVal = EnforceType(B, dst.getPointer(), PtrToIdTy);
+  llvm::Value *dstVal = EnforceType(B, dst.emitRawPointer(CGF), PtrToIdTy);
   B.CreateCall(WeakAssignFn, {src, dstVal});
 }
 
@@ -4218,7 +4220,7 @@ void CGObjCGNU::EmitObjCGlobalAssign(CodeGenFunction &CGF,
                                      bool threadlocal) {
   CGBuilderTy &B = CGF.Builder;
   src = EnforceType(B, src, IdTy);
-  llvm::Value *dstVal = EnforceType(B, dst.getPointer(), PtrToIdTy);
+  llvm::Value *dstVal = EnforceType(B, dst.emitRawPointer(CGF), PtrToIdTy);
   // FIXME. Add threadloca assign API
   assert(!threadlocal && "EmitObjCGlobalAssign - Threal Local API NYI");
   B.CreateCall(GlobalAssignFn, {src, dstVal});
@@ -4229,7 +4231,7 @@ void CGObjCGNU::EmitObjCIvarAssign(CodeGenFunction &CGF,
                                    llvm::Value *ivarOffset) {
   CGBuilderTy &B = CGF.Builder;
   src = EnforceType(B, src, IdTy);
-  llvm::Value *dstVal = EnforceType(B, dst.getPointer(), IdTy);
+  llvm::Value *dstVal = EnforceType(B, dst.emitRawPointer(CGF), IdTy);
   B.CreateCall(IvarAssignFn, {src, dstVal, ivarOffset});
 }
 
@@ -4237,7 +4239,7 @@ void CGObjCGNU::EmitObjCStrongCastAssign(CodeGenFunction &CGF,
                                          llvm::Value *src, Address dst) {
   CGBuilderTy &B = CGF.Builder;
   src = EnforceType(B, src, IdTy);
-  llvm::Value *dstVal = EnforceType(B, dst.getPointer(), PtrToIdTy);
+  llvm::Value *dstVal = EnforceType(B, dst.emitRawPointer(CGF), PtrToIdTy);
   B.CreateCall(StrongCastAssignFn, {src, dstVal});
 }
 
@@ -4246,8 +4248,8 @@ void CGObjCGNU::EmitGCMemmoveCollectable(CodeGenFunction &CGF,
                                          Address SrcPtr,
                                          llvm::Value *Size) {
   CGBuilderTy &B = CGF.Builder;
-  llvm::Value *DestPtrVal = EnforceType(B, DestPtr.getPointer(), PtrTy);
-  llvm::Value *SrcPtrVal = EnforceType(B, SrcPtr.getPointer(), PtrTy);
+  llvm::Value *DestPtrVal = EnforceType(B, DestPtr.emitRawPointer(CGF), PtrTy);
+  llvm::Value *SrcPtrVal = EnforceType(B, SrcPtr.emitRawPointer(CGF), PtrTy);
 
   B.CreateCall(MemMoveFn, {DestPtrVal, SrcPtrVal, Size});
 }
diff --git a/clang/lib/CodeGen/CGObjCMac.cpp b/clang/lib/CodeGen/CGObjCMac.cpp
index c463850e5fb64..042cd5d46da4b 100644
--- a/clang/lib/CodeGen/CGObjCMac.cpp
+++ b/clang/lib/CodeGen/CGObjCMac.cpp
@@ -1310,7 +1310,7 @@ class CGObjCMac : public CGObjCCommonMac {
   /// EmitSelector - Return a Value*, of type ObjCTypes.SelectorPtrTy,
   /// for the given selector.
   llvm::Value *EmitSelector(CodeGenFunction &CGF, Selector Sel);
-  Address EmitSelectorAddr(Selector Sel);
+  ConstantAddress EmitSelectorAddr(Selector Sel);
 
 public:
   CGObjCMac(CodeGen::CodeGenModule &cgm);
@@ -1538,7 +1538,7 @@ class CGObjCNonFragileABIMac : public CGObjCCommonMac {
   /// EmitSelector - Return a Value*, of type ObjCTypes.SelectorPtrTy,
   /// for the given selector.
   llvm::Value *EmitSelector(CodeGenFunction &CGF, Selector Sel);
-  Address EmitSelectorAddr(Selector Sel);
+  ConstantAddress EmitSelectorAddr(Selector Sel);
 
   /// GetInterfaceEHType - Get the cached ehtype for the given Objective-C
   /// interface. The return value has type EHTypePtrTy.
@@ -2064,9 +2064,8 @@ CGObjCMac::GenerateMessageSendSuper(CodeGen::CodeGenFunction &CGF,
                                     const ObjCMethodDecl *Method) {
   // Create and init a super structure; this is a (receiver, class)
   // pair we will pass to objc_msgSendSuper.
-  Address ObjCSuper =
-    CGF.CreateTempAlloca(ObjCTypes.SuperTy, CGF.getPointerAlign(),
-                         "objc_super");
+  RawAddress ObjCSuper = CGF.CreateTempAlloca(
+      ObjCTypes.SuperTy, CGF.getPointerAlign(), "objc_super");
   llvm::Value *ReceiverAsObject =
     CGF.Builder.CreateBitCast(Receiver, ObjCTypes.ObjectPtrTy);
   CGF.Builder.CreateStore(ReceiverAsObject,
@@ -4259,7 +4258,7 @@ namespace {
 
       CGF.EmitBlock(FinallyCallExit);
       CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionTryExitFn(),
-                                  ExceptionData.getPointer());
+                                  ExceptionData.emitRawPointer(CGF));
 
       CGF.EmitBlock(FinallyNoCallExit);
 
@@ -4425,7 +4424,9 @@ void FragileHazards::emitHazardsInNewBlocks() {
 }
 
 static void addIfPresent(llvm::DenseSet<llvm::Value*> &S, Address V) {
-  if (V.isValid()) S.insert(V.getPointer());
+  if (V.isValid())
+    if (llvm::Value *Ptr = V.getBasePointer())
+      S.insert(Ptr);
 }
 
 void FragileHazards::collectLocals() {
@@ -4628,13 +4629,13 @@ void CGObjCMac::EmitTryOrSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
   //  - Call objc_exception_try_enter to push ExceptionData on top of
   //    the EH stack.
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionTryEnterFn(),
-                              ExceptionData.getPointer());
+                              ExceptionData.emitRawPointer(CGF));
 
   //  - Call setjmp on the exception data buffer.
   llvm::Constant *Zero = llvm::ConstantInt::get(CGF.Builder.getInt32Ty(), 0);
   llvm::Value *GEPIndexes[] = { Zero, Zero, Zero };
   llvm::Value *SetJmpBuffer = CGF.Builder.CreateGEP(
-      ObjCTypes.ExceptionDataTy, ExceptionData.getPointer(), GEPIndexes,
+      ObjCTypes.ExceptionDataTy, ExceptionData.emitRawPointer(CGF), GEPIndexes,
       "setjmp_buffer");
   llvm::CallInst *SetJmpResult = CGF.EmitNounwindRuntimeCall(
       ObjCTypes.getSetJmpFn(), SetJmpBuffer, "setjmp_result");
@@ -4673,9 +4674,9 @@ void CGObjCMac::EmitTryOrSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
   } else {
     // Retrieve the exception object.  We may emit multiple blocks but
     // nothing can cross this so the value is already in SSA form.
-    llvm::CallInst *Caught =
-      CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionExtractFn(),
-                                  ExceptionData.getPointer(), "caught");
+    llvm::CallInst *Caught = CGF.EmitNounwindRuntimeCall(
+        ObjCTypes.getExceptionExtractFn(), ExceptionData.emitRawPointer(CGF),
+        "caught");
 
     // Push the exception to rethrow onto the EH value stack for the
     // benefit of any @throws in the handlers.
@@ -4698,7 +4699,7 @@ void CGObjCMac::EmitTryOrSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
       // Enter a new exception try block (in case a @catch block
       // throws an exception).
       CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionTryEnterFn(),
-                                  ExceptionData.getPointer());
+                                  ExceptionData.emitRawPointer(CGF));
 
       llvm::CallInst *SetJmpResult =
         CGF.EmitNounwindRuntimeCall(ObjCTypes.getSetJmpFn(),
@@ -4829,9 +4830,9 @@ void CGObjCMac::EmitTryOrSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
       // Extract the new exception and save it to the
       // propagating-exception slot.
       assert(PropagatingExnVar.isValid());
-      llvm::CallInst *NewCaught =
-        CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionExtractFn(),
-                                    ExceptionData.getPointer(), "caught");
+      llvm::CallInst *NewCaught = CGF.EmitNounwindRuntimeCall(
+          ObjCTypes.getExceptionExtractFn(), ExceptionData.emitRawPointer(CGF),
+          "caught");
       CGF.Builder.CreateStore(NewCaught, PropagatingExnVar);
 
       // Don't pop the catch handler; the throw already did.
@@ -4861,9 +4862,8 @@ void CGObjCMac::EmitTryOrSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
 
     // Otherwise, just look in the buffer for the exception to throw.
     } else {
-      llvm::CallInst *Caught =
-        CGF.EmitNounwindRuntimeCall(ObjCTypes.getExceptionExtractFn(),
-                                    ExceptionData.getPointer());
+      llvm::CallInst *Caught = CGF.EmitNounwindRuntimeCall(
+          ObjCTypes.getExceptionExtractFn(), ExceptionData.emitRawPointer(CGF));
       PropagatingExn = Caught;
     }
 
@@ -4906,7 +4906,7 @@ llvm::Value * CGObjCMac::EmitObjCWeakRead(CodeGen::CodeGenFunction &CGF,
                                           Address AddrWeakObj) {
   llvm::Type* DestTy = AddrWeakObj.getElementType();
   llvm::Value *AddrWeakObjVal = CGF.Builder.CreateBitCast(
-      AddrWeakObj.getPointer(), ObjCTypes.PtrObjectPtrTy);
+      AddrWeakObj.emitRawPointer(CGF), ObjCTypes.PtrObjectPtrTy);
   llvm::Value *read_weak =
     CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcReadWeakFn(),
                                 AddrWeakObjVal, "weakread");
@@ -4928,8 +4928,8 @@ void CGObjCMac::EmitObjCWeakAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = { src, dstVal };
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignWeakFn(),
                               args, "weakassign");
@@ -4950,8 +4950,8 @@ void CGObjCMac::EmitObjCGlobalAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal};
   if (!threadlocal)
     CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignGlobalFn(),
@@ -4977,8 +4977,8 @@ void CGObjCMac::EmitObjCIvarAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal, ivarOffset};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignIvarFn(), args);
 }
@@ -4997,8 +4997,8 @@ void CGObjCMac::EmitObjCStrongCastAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignStrongCastFn(),
                               args, "strongassign");
@@ -5007,7 +5007,8 @@ void CGObjCMac::EmitObjCStrongCastAssign(CodeGen::CodeGenFunction &CGF,
 void CGObjCMac::EmitGCMemmoveCollectable(CodeGen::CodeGenFunction &CGF,
                                          Address DestPtr, Address SrcPtr,
                                          llvm::Value *size) {
-  llvm::Value *args[] = { DestPtr.getPointer(), SrcPtr.getPointer(), size };
+  llvm::Value *args[] = {DestPtr.emitRawPointer(CGF),
+                         SrcPtr.emitRawPointer(CGF), size};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.GcMemmoveCollectableFn(), args);
 }
 
@@ -5243,7 +5244,7 @@ llvm::Value *CGObjCMac::EmitSelector(CodeGenFunction &CGF, Selector Sel) {
   return CGF.Builder.CreateLoad(EmitSelectorAddr(Sel));
 }
 
-Address CGObjCMac::EmitSelectorAddr(Selector Sel) {
+ConstantAddress CGObjCMac::EmitSelectorAddr(Selector Sel) {
   CharUnits Align = CGM.getPointerAlign();
 
   llvm::GlobalVariable *&Entry = SelectorReferences[Sel];
@@ -5254,7 +5255,7 @@ Address CGObjCMac::EmitSelectorAddr(Selector Sel) {
     Entry->setExternallyInitialized(true);
   }
 
-  return Address(Entry, ObjCTypes.SelectorPtrTy, Align);
+  return ConstantAddress(Entry, ObjCTypes.SelectorPtrTy, Align);
 }
 
 llvm::Constant *CGObjCCommonMac::GetClassName(StringRef RuntimeName) {
@@ -7322,7 +7323,7 @@ CGObjCNonFragileABIMac::EmitVTableMessageSend(CodeGenFunction &CGF,
             ObjCTypes.MessageRefTy, CGF.getPointerAlign());
 
   // Update the message ref argument.
-  args[1].setRValue(RValue::get(mref.getPointer()));
+  args[1].setRValue(RValue::get(mref, CGF));
 
   // Load the function to call from the message ref table.
   Address calleeAddr = CGF.Builder.CreateStructGEP(mref, 0);
@@ -7551,9 +7552,8 @@ CGObjCNonFragileABIMac::GenerateMessageSendSuper(CodeGen::CodeGenFunction &CGF,
   // ...
   // Create and init a super structure; this is a (receiver, class)
   // pair we will pass to objc_msgSendSuper.
-  Address ObjCSuper =
-    CGF.CreateTempAlloca(ObjCTypes.SuperTy, CGF.getPointerAlign(),
-                         "objc_super");
+  RawAddress ObjCSuper = CGF.CreateTempAlloca(
+      ObjCTypes.SuperTy, CGF.getPointerAlign(), "objc_super");
 
   llvm::Value *ReceiverAsObject =
     CGF.Builder.CreateBitCast(Receiver, ObjCTypes.ObjectPtrTy);
@@ -7593,7 +7593,7 @@ llvm::Value *CGObjCNonFragileABIMac::EmitSelector(CodeGenFunction &CGF,
   return LI;
 }
 
-Address CGObjCNonFragileABIMac::EmitSelectorAddr(Selector Sel) {
+ConstantAddress CGObjCNonFragileABIMac::EmitSelectorAddr(Selector Sel) {
   llvm::GlobalVariable *&Entry = SelectorReferences[Sel];
   CharUnits Align = CGM.getPointerAlign();
   if (!Entry) {
@@ -7609,7 +7609,7 @@ Address CGObjCNonFragileABIMac::EmitSelectorAddr(Selector Sel) {
     CGM.addCompilerUsedGlobal(Entry);
   }
 
-  return Address(Entry, ObjCTypes.SelectorPtrTy, Align);
+  return ConstantAddress(Entry, ObjCTypes.SelectorPtrTy, Align);
 }
 
 /// EmitObjCIvarAssign - Code gen for assigning to a __strong object.
@@ -7628,8 +7628,8 @@ void CGObjCNonFragileABIMac::EmitObjCIvarAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal, ivarOffset};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignIvarFn(), args);
 }
@@ -7649,8 +7649,8 @@ void CGObjCNonFragileABIMac::EmitObjCStrongCastAssign(
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignStrongCastFn(),
                               args, "weakassign");
@@ -7659,7 +7659,8 @@ void CGObjCNonFragileABIMac::EmitObjCStrongCastAssign(
 void CGObjCNonFragileABIMac::EmitGCMemmoveCollectable(
     CodeGen::CodeGenFunction &CGF, Address DestPtr, Address SrcPtr,
     llvm::Value *Size) {
-  llvm::Value *args[] = { DestPtr.getPointer(), SrcPtr.getPointer(), Size };
+  llvm::Value *args[] = {DestPtr.emitRawPointer(CGF),
+                         SrcPtr.emitRawPointer(CGF), Size};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.GcMemmoveCollectableFn(), args);
 }
 
@@ -7671,7 +7672,7 @@ llvm::Value * CGObjCNonFragileABIMac::EmitObjCWeakRead(
   Address AddrWeakObj) {
   llvm::Type *DestTy = AddrWeakObj.getElementType();
   llvm::Value *AddrWeakObjVal = CGF.Builder.CreateBitCast(
-      AddrWeakObj.getPointer(), ObjCTypes.PtrObjectPtrTy);
+      AddrWeakObj.emitRawPointer(CGF), ObjCTypes.PtrObjectPtrTy);
   llvm::Value *read_weak =
     CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcReadWeakFn(),
                                 AddrWeakObjVal, "weakread");
@@ -7693,8 +7694,8 @@ void CGObjCNonFragileABIMac::EmitObjCWeakAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal};
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignWeakFn(),
                               args, "weakassign");
@@ -7715,8 +7716,8 @@ void CGObjCNonFragileABIMac::EmitObjCGlobalAssign(CodeGen::CodeGenFunction &CGF,
     src = CGF.Builder.CreateIntToPtr(src, ObjCTypes.Int8PtrTy);
   }
   src = CGF.Builder.CreateBitCast(src, ObjCTypes.ObjectPtrTy);
-  llvm::Value *dstVal =
-      CGF.Builder.CreateBitCast(dst.getPointer(), ObjCTypes.PtrObjectPtrTy);
+  llvm::Value *dstVal = CGF.Builder.CreateBitCast(dst.emitRawPointer(CGF),
+                                                  ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = {src, dstVal};
   if (!threadlocal)
     CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignGlobalFn(),
diff --git a/clang/lib/CodeGen/CGObjCRuntime.cpp b/clang/lib/CodeGen/CGObjCRuntime.cpp
index 424564f975999..01d0f35da1964 100644
--- a/clang/lib/CodeGen/CGObjCRuntime.cpp
+++ b/clang/lib/CodeGen/CGObjCRuntime.cpp
@@ -67,7 +67,7 @@ LValue CGObjCRuntime::EmitValueForIvarAtOffset(CodeGen::CodeGenFunction &CGF,
   V = CGF.Builder.CreateInBoundsGEP(CGF.Int8Ty, V, Offset, "add.ptr");
 
   if (!Ivar->isBitField()) {
-    LValue LV = CGF.MakeNaturalAlignAddrLValue(V, IvarTy);
+    LValue LV = CGF.MakeNaturalAlignRawAddrLValue(V, IvarTy);
     return LV;
   }
 
@@ -233,7 +233,7 @@ void CGObjCRuntime::EmitTryCatchStmt(CodeGenFunction &CGF,
       llvm::Instruction *CPICandidate = Handler.Block->getFirstNonPHI();
       if (auto *CPI = dyn_cast_or_null<llvm::CatchPadInst>(CPICandidate)) {
         CGF.CurrentFuncletPad = CPI;
-        CPI->setOperand(2, CGF.getExceptionSlot().getPointer());
+        CPI->setOperand(2, CGF.getExceptionSlot().emitRawPointer(CGF));
         CGF.EHStack.pushCleanup<CatchRetScope>(NormalCleanup, CPI);
       }
     }
@@ -405,7 +405,7 @@ bool CGObjCRuntime::canMessageReceiverBeNull(CodeGenFunction &CGF,
     auto self = curMethod->getSelfDecl();
     if (self->getType().isConstQualified()) {
       if (auto LI = dyn_cast<llvm::LoadInst>(receiver->stripPointerCasts())) {
-        llvm::Value *selfAddr = CGF.GetAddrOfLocalVar(self).getPointer();
+        llvm::Value *selfAddr = CGF.GetAddrOfLocalVar(self).emitRawPointer(CGF);
         if (selfAddr == LI->getPointerOperand()) {
           return false;
         }
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 0c5df95e4eaaa..28271160649af 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -622,7 +622,7 @@ static void emitInitWithReductionInitializer(CodeGenFunction &CGF,
     auto *GV = new llvm::GlobalVariable(
         CGF.CGM.getModule(), Init->getType(), /*isConstant=*/true,
         llvm::GlobalValue::PrivateLinkage, Init, Name);
-    LValue LV = CGF.MakeNaturalAlignAddrLValue(GV, Ty);
+    LValue LV = CGF.MakeNaturalAlignRawAddrLValue(GV, Ty);
     RValue InitRVal;
     switch (CGF.getEvaluationKind(Ty)) {
     case TEK_Scalar:
@@ -668,8 +668,8 @@ static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
 
   llvm::Value *SrcBegin = nullptr;
   if (DRD)
-    SrcBegin = SrcAddr.getPointer();
-  llvm::Value *DestBegin = DestAddr.getPointer();
+    SrcBegin = SrcAddr.emitRawPointer(CGF);
+  llvm::Value *DestBegin = DestAddr.emitRawPointer(CGF);
   // Cast from pointer to array type to pointer to single element.
   llvm::Value *DestEnd =
       CGF.Builder.CreateGEP(DestAddr.getElementType(), DestBegin, NumElements);
@@ -912,7 +912,7 @@ static LValue loadToBegin(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
 
 static Address castToBase(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
                           Address OriginalBaseAddress, llvm::Value *Addr) {
-  Address Tmp = Address::invalid();
+  RawAddress Tmp = RawAddress::invalid();
   Address TopTmp = Address::invalid();
   Address MostTopTmp = Address::invalid();
   BaseTy = BaseTy.getNonReferenceType();
@@ -971,10 +971,10 @@ Address ReductionCodeGen::adjustPrivateAddress(CodeGenFunction &CGF, unsigned N,
     Address SharedAddr = SharedAddresses[N].first.getAddress(CGF);
     llvm::Value *Adjustment = CGF.Builder.CreatePtrDiff(
         SharedAddr.getElementType(), BaseLValue.getPointer(CGF),
-        SharedAddr.getPointer());
+        SharedAddr.emitRawPointer(CGF));
     llvm::Value *PrivatePointer =
         CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
-            PrivateAddr.getPointer(), SharedAddr.getType());
+            PrivateAddr.emitRawPointer(CGF), SharedAddr.getType());
     llvm::Value *Ptr = CGF.Builder.CreateGEP(
         SharedAddr.getElementType(), PrivatePointer, Adjustment);
     return castToBase(CGF, OrigVD->getType(),
@@ -1557,7 +1557,7 @@ static llvm::TargetRegionEntryInfo getEntryInfoFromPresumedLoc(
   return OMPBuilder.getTargetEntryUniqueInfo(FileInfoCallBack, ParentName);
 }
 
-Address CGOpenMPRuntime::getAddrOfDeclareTargetVar(const VarDecl *VD) {
+ConstantAddress CGOpenMPRuntime::getAddrOfDeclareTargetVar(const VarDecl *VD) {
   auto AddrOfGlobal = [&VD, this]() { return CGM.GetAddrOfGlobal(VD); };
 
   auto LinkageForVariable = [&VD, this]() {
@@ -1579,8 +1579,8 @@ Address CGOpenMPRuntime::getAddrOfDeclareTargetVar(const VarDecl *VD) {
       LinkageForVariable);
 
   if (!addr)
-    return Address::invalid();
-  return Address(addr, LlvmPtrTy, CGM.getContext().getDeclAlign(VD));
+    return ConstantAddress::invalid();
+  return ConstantAddress(addr, LlvmPtrTy, CGM.getContext().getDeclAlign(VD));
 }
 
 llvm::Constant *
@@ -1604,7 +1604,7 @@ Address CGOpenMPRuntime::getAddrOfThreadPrivate(CodeGenFunction &CGF,
   llvm::Type *VarTy = VDAddr.getElementType();
   llvm::Value *Args[] = {
       emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
-      CGF.Builder.CreatePointerCast(VDAddr.getPointer(), CGM.Int8PtrTy),
+      CGF.Builder.CreatePointerCast(VDAddr.emitRawPointer(CGF), CGM.Int8PtrTy),
       CGM.getSize(CGM.GetTargetTypeStoreSize(VarTy)),
       getOrCreateThreadPrivateCache(VD)};
   return Address(
@@ -1627,7 +1627,8 @@ void CGOpenMPRuntime::emitThreadPrivateVarInit(
   // Call __kmpc_threadprivate_register(&loc, &var, ctor, cctor/*NULL*/, dtor)
   // to register constructor/destructor for variable.
   llvm::Value *Args[] = {
-      OMPLoc, CGF.Builder.CreatePointerCast(VDAddr.getPointer(), CGM.VoidPtrTy),
+      OMPLoc,
+      CGF.Builder.CreatePointerCast(VDAddr.emitRawPointer(CGF), CGM.VoidPtrTy),
       Ctor, CopyCtor, Dtor};
   CGF.EmitRuntimeCall(
       OMPBuilder.getOrCreateRuntimeFunction(
@@ -1900,13 +1901,13 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
 
     // OutlinedFn(&GTid, &zero_bound, CapturedStruct);
     Address ThreadIDAddr = RT.emitThreadIDAddress(CGF, Loc);
-    Address ZeroAddrBound =
+    RawAddress ZeroAddrBound =
         CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
                                          /*Name=*/".bound.zero.addr");
     CGF.Builder.CreateStore(CGF.Builder.getInt32(/*C*/ 0), ZeroAddrBound);
     llvm::SmallVector<llvm::Value *, 16> OutlinedFnArgs;
     // ThreadId for serialized parallels is 0.
-    OutlinedFnArgs.push_back(ThreadIDAddr.getPointer());
+    OutlinedFnArgs.push_back(ThreadIDAddr.emitRawPointer(CGF));
     OutlinedFnArgs.push_back(ZeroAddrBound.getPointer());
     OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());
 
@@ -2272,7 +2273,7 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
         emitUpdateLocation(CGF, Loc), // ident_t *<loc>
         getThreadID(CGF, Loc),        // i32 <gtid>
         BufSize,                      // size_t <buf_size>
-        CL.getPointer(),              // void *<copyprivate list>
+        CL.emitRawPointer(CGF),       // void *<copyprivate list>
         CpyFn,                        // void (*) (void *, void *) <copy_func>
         DidItVal                      // i32 did_it
     };
@@ -2591,10 +2592,10 @@ static void emitForStaticInitCall(
       ThreadId,
       CGF.Builder.getInt32(addMonoNonMonoModifier(CGF.CGM, Schedule, M1,
                                                   M2)), // Schedule type
-      Values.IL.getPointer(),                           // &isLastIter
-      Values.LB.getPointer(),                           // &LB
-      Values.UB.getPointer(),                           // &UB
-      Values.ST.getPointer(),                           // &Stride
+      Values.IL.emitRawPointer(CGF),                    // &isLastIter
+      Values.LB.emitRawPointer(CGF),                    // &LB
+      Values.UB.emitRawPointer(CGF),                    // &UB
+      Values.ST.emitRawPointer(CGF),                    // &Stride
       CGF.Builder.getIntN(Values.IVSize, 1),            // Incr
       Chunk                                             // Chunk
   };
@@ -2698,12 +2699,11 @@ llvm::Value *CGOpenMPRuntime::emitForNext(CodeGenFunction &CGF,
   //          kmp_int[32|64] *p_lower, kmp_int[32|64] *p_upper,
   //          kmp_int[32|64] *p_stride);
   llvm::Value *Args[] = {
-      emitUpdateLocation(CGF, Loc),
-      getThreadID(CGF, Loc),
-      IL.getPointer(), // &isLastIter
-      LB.getPointer(), // &Lower
-      UB.getPointer(), // &Upper
-      ST.getPointer()  // &Stride
+      emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
+      IL.emitRawPointer(CGF), // &isLastIter
+      LB.emitRawPointer(CGF), // &Lower
+      UB.emitRawPointer(CGF), // &Upper
+      ST.emitRawPointer(CGF)  // &Stride
   };
   llvm::Value *Call = CGF.EmitRuntimeCall(
       OMPBuilder.createDispatchNextFunction(IVSize, IVSigned), Args);
@@ -3048,7 +3048,7 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
       CGF.Builder
           .CreatePointerBitCastOrAddrSpaceCast(TDBase.getAddress(CGF),
                                                CGF.VoidPtrTy, CGF.Int8Ty)
-          .getPointer()};
+          .emitRawPointer(CGF)};
   SmallVector<llvm::Value *, 16> CallArgs(std::begin(CommonArgs),
                                           std::end(CommonArgs));
   if (isOpenMPTaskLoopDirective(Kind)) {
@@ -3575,7 +3575,8 @@ getPointerAndSize(CodeGenFunction &CGF, const Expr *E) {
         CGF.EmitOMPArraySectionExpr(ASE, /*IsLowerBound=*/false);
     Address UpAddrAddress = UpAddrLVal.getAddress(CGF);
     llvm::Value *UpAddr = CGF.Builder.CreateConstGEP1_32(
-        UpAddrAddress.getElementType(), UpAddrAddress.getPointer(), /*Idx0=*/1);
+        UpAddrAddress.getElementType(), UpAddrAddress.emitRawPointer(CGF),
+        /*Idx0=*/1);
     llvm::Value *LowIntPtr = CGF.Builder.CreatePtrToInt(Addr, CGF.SizeTy);
     llvm::Value *UpIntPtr = CGF.Builder.CreatePtrToInt(UpAddr, CGF.SizeTy);
     SizeVal = CGF.Builder.CreateNUWSub(UpIntPtr, LowIntPtr);
@@ -3889,8 +3890,9 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
         llvm::Value *Size;
         std::tie(Addr, Size) = getPointerAndSize(CGF, E);
         llvm::Value *Idx = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
-        LValue Base = CGF.MakeAddrLValue(
-            CGF.Builder.CreateGEP(AffinitiesArray, Idx), KmpTaskAffinityInfoTy);
+        LValue Base =
+            CGF.MakeAddrLValue(CGF.Builder.CreateGEP(CGF, AffinitiesArray, Idx),
+                               KmpTaskAffinityInfoTy);
         // affs[i].base_addr = &<Affinities[i].second>;
         LValue BaseAddrLVal = CGF.EmitLValueForField(
             Base, *std::next(KmpAffinityInfoRD->field_begin(), BaseAddr));
@@ -3911,7 +3913,7 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
     llvm::Value *LocRef = emitUpdateLocation(CGF, Loc);
     llvm::Value *GTid = getThreadID(CGF, Loc);
     llvm::Value *AffinListPtr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
-        AffinitiesArray.getPointer(), CGM.VoidPtrTy);
+        AffinitiesArray.emitRawPointer(CGF), CGM.VoidPtrTy);
     // FIXME: Emit the function and ignore its result for now unless the
     // runtime function is properly implemented.
     (void)CGF.EmitRuntimeCall(
@@ -3922,8 +3924,8 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
   llvm::Value *NewTaskNewTaskTTy =
       CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
           NewTask, KmpTaskTWithPrivatesPtrTy);
-  LValue Base = CGF.MakeNaturalAlignAddrLValue(NewTaskNewTaskTTy,
-                                               KmpTaskTWithPrivatesQTy);
+  LValue Base = CGF.MakeNaturalAlignRawAddrLValue(NewTaskNewTaskTTy,
+                                                  KmpTaskTWithPrivatesQTy);
   LValue TDBase =
       CGF.EmitLValueForField(Base, *KmpTaskTWithPrivatesQTyRD->field_begin());
   // Fill the data in the resulting kmp_task_t record.
@@ -4048,7 +4050,7 @@ CGOpenMPRuntime::getDepobjElements(CodeGenFunction &CGF, LValue DepobjLVal,
           CGF.ConvertTypeForMem(KmpDependInfoPtrTy)),
       KmpDependInfoPtrTy->castAs<PointerType>());
   Address DepObjAddr = CGF.Builder.CreateGEP(
-      Base.getAddress(CGF),
+      CGF, Base.getAddress(CGF),
       llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
   LValue NumDepsBase = CGF.MakeAddrLValue(
       DepObjAddr, KmpDependInfoTy, Base.getBaseInfo(), Base.getTBAAInfo());
@@ -4098,7 +4100,7 @@ static void emitDependData(CodeGenFunction &CGF, QualType &KmpDependInfoTy,
       LValue &PosLVal = *Pos.get<LValue *>();
       llvm::Value *Idx = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
       Base = CGF.MakeAddrLValue(
-          CGF.Builder.CreateGEP(DependenciesArray, Idx), KmpDependInfoTy);
+          CGF.Builder.CreateGEP(CGF, DependenciesArray, Idx), KmpDependInfoTy);
     }
     // deps[i].base_addr = &<Dependencies[i].second>;
     LValue BaseAddrLVal = CGF.EmitLValueForField(
@@ -4196,7 +4198,7 @@ void CGOpenMPRuntime::emitDepobjElements(CodeGenFunction &CGF,
           ElSize,
           CGF.Builder.CreateIntCast(NumDeps, CGF.SizeTy, /*isSigned=*/false));
       llvm::Value *Pos = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
-      Address DepAddr = CGF.Builder.CreateGEP(DependenciesArray, Pos);
+      Address DepAddr = CGF.Builder.CreateGEP(CGF, DependenciesArray, Pos);
       CGF.Builder.CreateMemCpy(DepAddr, Base.getAddress(CGF), Size);
 
       // Increase pos.
@@ -4431,7 +4433,7 @@ void CGOpenMPRuntime::emitDestroyClause(CodeGenFunction &CGF, LValue DepobjLVal,
       Base.getAddress(CGF), CGF.ConvertTypeForMem(KmpDependInfoPtrTy),
       CGF.ConvertTypeForMem(KmpDependInfoTy));
   llvm::Value *DepObjAddr = CGF.Builder.CreateGEP(
-      Addr.getElementType(), Addr.getPointer(),
+      Addr.getElementType(), Addr.emitRawPointer(CGF),
       llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
   DepObjAddr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(DepObjAddr,
                                                                CGF.VoidPtrTy);
@@ -4461,8 +4463,8 @@ void CGOpenMPRuntime::emitUpdateClause(CodeGenFunction &CGF, LValue DepobjLVal,
 
   Address Begin = Base.getAddress(CGF);
   // Cast from pointer to array type to pointer to single element.
-  llvm::Value *End = CGF.Builder.CreateGEP(
-      Begin.getElementType(), Begin.getPointer(), NumDeps);
+  llvm::Value *End = CGF.Builder.CreateGEP(Begin.getElementType(),
+                                           Begin.emitRawPointer(CGF), NumDeps);
   // The basic structure here is a while-do loop.
   llvm::BasicBlock *BodyBB = CGF.createBasicBlock("omp.body");
   llvm::BasicBlock *DoneBB = CGF.createBasicBlock("omp.done");
@@ -4470,7 +4472,7 @@ void CGOpenMPRuntime::emitUpdateClause(CodeGenFunction &CGF, LValue DepobjLVal,
   CGF.EmitBlock(BodyBB);
   llvm::PHINode *ElementPHI =
       CGF.Builder.CreatePHI(Begin.getType(), 2, "omp.elementPast");
-  ElementPHI->addIncoming(Begin.getPointer(), EntryBB);
+  ElementPHI->addIncoming(Begin.emitRawPointer(CGF), EntryBB);
   Begin = Begin.withPointer(ElementPHI, KnownNonNull);
   Base = CGF.MakeAddrLValue(Begin, KmpDependInfoTy, Base.getBaseInfo(),
                             Base.getTBAAInfo());
@@ -4484,12 +4486,12 @@ void CGOpenMPRuntime::emitUpdateClause(CodeGenFunction &CGF, LValue DepobjLVal,
       FlagsLVal);
 
   // Shift the address forward by one element.
-  Address ElementNext =
-      CGF.Builder.CreateConstGEP(Begin, /*Index=*/1, "omp.elementNext");
-  ElementPHI->addIncoming(ElementNext.getPointer(),
-                          CGF.Builder.GetInsertBlock());
+  llvm::Value *ElementNext =
+      CGF.Builder.CreateConstGEP(Begin, /*Index=*/1, "omp.elementNext")
+          .emitRawPointer(CGF);
+  ElementPHI->addIncoming(ElementNext, CGF.Builder.GetInsertBlock());
   llvm::Value *IsEmpty =
-      CGF.Builder.CreateICmpEQ(ElementNext.getPointer(), End, "omp.isempty");
+      CGF.Builder.CreateICmpEQ(ElementNext, End, "omp.isempty");
   CGF.Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);
   // Done.
   CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
@@ -4532,7 +4534,7 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
     DepTaskArgs[1] = ThreadID;
     DepTaskArgs[2] = NewTask;
     DepTaskArgs[3] = NumOfElements;
-    DepTaskArgs[4] = DependenciesArray.getPointer();
+    DepTaskArgs[4] = DependenciesArray.emitRawPointer(CGF);
     DepTaskArgs[5] = CGF.Builder.getInt32(0);
     DepTaskArgs[6] = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
   }
@@ -4564,7 +4566,7 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
     DepWaitTaskArgs[0] = UpLoc;
     DepWaitTaskArgs[1] = ThreadID;
     DepWaitTaskArgs[2] = NumOfElements;
-    DepWaitTaskArgs[3] = DependenciesArray.getPointer();
+    DepWaitTaskArgs[3] = DependenciesArray.emitRawPointer(CGF);
     DepWaitTaskArgs[4] = CGF.Builder.getInt32(0);
     DepWaitTaskArgs[5] = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
     DepWaitTaskArgs[6] =
@@ -4726,8 +4728,8 @@ static void EmitOMPAggregateReduction(
   const ArrayType *ArrayTy = Type->getAsArrayTypeUnsafe();
   llvm::Value *NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, LHSAddr);
 
-  llvm::Value *RHSBegin = RHSAddr.getPointer();
-  llvm::Value *LHSBegin = LHSAddr.getPointer();
+  llvm::Value *RHSBegin = RHSAddr.emitRawPointer(CGF);
+  llvm::Value *LHSBegin = LHSAddr.emitRawPointer(CGF);
   // Cast from pointer to array type to pointer to single element.
   llvm::Value *LHSEnd =
       CGF.Builder.CreateGEP(LHSAddr.getElementType(), LHSBegin, NumElements);
@@ -4991,7 +4993,7 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
   QualType ReductionArrayTy = C.getConstantArrayType(
       C.VoidPtrTy, ArraySize, nullptr, ArraySizeModifier::Normal,
       /*IndexTypeQuals=*/0);
-  Address ReductionList =
+  RawAddress ReductionList =
       CGF.CreateMemTemp(ReductionArrayTy, ".omp.reduction.red_list");
   const auto *IPriv = Privates.begin();
   unsigned Idx = 0;
@@ -5463,7 +5465,7 @@ llvm::Value *CGOpenMPRuntime::emitTaskReductionInit(
       C.getConstantArrayType(RDType, ArraySize, nullptr,
                              ArraySizeModifier::Normal, /*IndexTypeQuals=*/0);
   // kmp_task_red_input_t .rd_input.[Size];
-  Address TaskRedInput = CGF.CreateMemTemp(ArrayRDType, ".rd_input.");
+  RawAddress TaskRedInput = CGF.CreateMemTemp(ArrayRDType, ".rd_input.");
   ReductionCodeGen RCG(Data.ReductionVars, Data.ReductionOrigs,
                        Data.ReductionCopies, Data.ReductionOps);
   for (unsigned Cnt = 0; Cnt < Size; ++Cnt) {
@@ -5474,7 +5476,7 @@ llvm::Value *CGOpenMPRuntime::emitTaskReductionInit(
         TaskRedInput.getElementType(), TaskRedInput.getPointer(), Idxs,
         /*SignedIndices=*/false, /*IsSubtraction=*/false, Loc,
         ".rd_input.gep.");
-    LValue ElemLVal = CGF.MakeNaturalAlignAddrLValue(GEP, RDType);
+    LValue ElemLVal = CGF.MakeNaturalAlignRawAddrLValue(GEP, RDType);
     // ElemLVal.reduce_shar = &Shareds[Cnt];
     LValue SharedLVal = CGF.EmitLValueForField(ElemLVal, SharedFD);
     RCG.emitSharedOrigLValue(CGF, Cnt);
@@ -5632,7 +5634,7 @@ void CGOpenMPRuntime::emitTaskwaitCall(CodeGenFunction &CGF, SourceLocation Loc,
       DepWaitTaskArgs[0] = UpLoc;
       DepWaitTaskArgs[1] = ThreadID;
       DepWaitTaskArgs[2] = NumOfElements;
-      DepWaitTaskArgs[3] = DependenciesArray.getPointer();
+      DepWaitTaskArgs[3] = DependenciesArray.emitRawPointer(CGF);
       DepWaitTaskArgs[4] = CGF.Builder.getInt32(0);
       DepWaitTaskArgs[5] = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
       DepWaitTaskArgs[6] =
@@ -5855,7 +5857,7 @@ void CGOpenMPRuntime::emitUsesAllocatorsInit(CodeGenFunction &CGF,
   AllocatorTraitsLVal = CGF.MakeAddrLValue(Addr, CGF.getContext().VoidPtrTy,
                                            AllocatorTraitsLVal.getBaseInfo(),
                                            AllocatorTraitsLVal.getTBAAInfo());
-  llvm::Value *Traits = Addr.getPointer();
+  llvm::Value *Traits = Addr.emitRawPointer(CGF);
 
   llvm::Value *AllocatorVal =
       CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
@@ -7315,17 +7317,19 @@ class MappableExprsHandler {
                       CGF.EmitOMPSharedLValue(MC.getAssociatedExpression())
                           .getAddress(CGF);
                 }
-                Size = CGF.Builder.CreatePtrDiff(
-                    CGF.Int8Ty, ComponentLB.getPointer(), LB.getPointer());
+                llvm::Value *ComponentLBPtr = ComponentLB.emitRawPointer(CGF);
+                llvm::Value *LBPtr = LB.emitRawPointer(CGF);
+                Size = CGF.Builder.CreatePtrDiff(CGF.Int8Ty, ComponentLBPtr,
+                                                 LBPtr);
                 break;
               }
             }
             assert(Size && "Failed to determine structure size");
             CombinedInfo.Exprs.emplace_back(MapDecl, MapExpr);
-            CombinedInfo.BasePointers.push_back(BP.getPointer());
+            CombinedInfo.BasePointers.push_back(BP.emitRawPointer(CGF));
             CombinedInfo.DevicePtrDecls.push_back(nullptr);
             CombinedInfo.DevicePointers.push_back(DeviceInfoTy::None);
-            CombinedInfo.Pointers.push_back(LB.getPointer());
+            CombinedInfo.Pointers.push_back(LB.emitRawPointer(CGF));
             CombinedInfo.Sizes.push_back(CGF.Builder.CreateIntCast(
                 Size, CGF.Int64Ty, /*isSigned=*/true));
             CombinedInfo.Types.push_back(Flags);
@@ -7335,13 +7339,14 @@ class MappableExprsHandler {
             LB = CGF.Builder.CreateConstGEP(ComponentLB, 1);
           }
           CombinedInfo.Exprs.emplace_back(MapDecl, MapExpr);
-          CombinedInfo.BasePointers.push_back(BP.getPointer());
+          CombinedInfo.BasePointers.push_back(BP.emitRawPointer(CGF));
           CombinedInfo.DevicePtrDecls.push_back(nullptr);
           CombinedInfo.DevicePointers.push_back(DeviceInfoTy::None);
-          CombinedInfo.Pointers.push_back(LB.getPointer());
+          CombinedInfo.Pointers.push_back(LB.emitRawPointer(CGF));
+          llvm::Value *LBPtr = LB.emitRawPointer(CGF);
           Size = CGF.Builder.CreatePtrDiff(
-              CGF.Int8Ty, CGF.Builder.CreateConstGEP(HB, 1).getPointer(),
-              LB.getPointer());
+              CGF.Int8Ty, CGF.Builder.CreateConstGEP(HB, 1).emitRawPointer(CGF),
+              LBPtr);
           CombinedInfo.Sizes.push_back(
               CGF.Builder.CreateIntCast(Size, CGF.Int64Ty, /*isSigned=*/true));
           CombinedInfo.Types.push_back(Flags);
@@ -7359,20 +7364,21 @@ class MappableExprsHandler {
             (Next == CE && MapType != OMPC_MAP_unknown)) {
           if (!IsMappingWholeStruct) {
             CombinedInfo.Exprs.emplace_back(MapDecl, MapExpr);
-            CombinedInfo.BasePointers.push_back(BP.getPointer());
+            CombinedInfo.BasePointers.push_back(BP.emitRawPointer(CGF));
             CombinedInfo.DevicePtrDecls.push_back(nullptr);
             CombinedInfo.DevicePointers.push_back(DeviceInfoTy::None);
-            CombinedInfo.Pointers.push_back(LB.getPointer());
+            CombinedInfo.Pointers.push_back(LB.emitRawPointer(CGF));
             CombinedInfo.Sizes.push_back(CGF.Builder.CreateIntCast(
                 Size, CGF.Int64Ty, /*isSigned=*/true));
             CombinedInfo.NonContigInfo.Dims.push_back(IsNonContiguous ? DimSize
                                                                       : 1);
           } else {
             StructBaseCombinedInfo.Exprs.emplace_back(MapDecl, MapExpr);
-            StructBaseCombinedInfo.BasePointers.push_back(BP.getPointer());
+            StructBaseCombinedInfo.BasePointers.push_back(
+                BP.emitRawPointer(CGF));
             StructBaseCombinedInfo.DevicePtrDecls.push_back(nullptr);
             StructBaseCombinedInfo.DevicePointers.push_back(DeviceInfoTy::None);
-            StructBaseCombinedInfo.Pointers.push_back(LB.getPointer());
+            StructBaseCombinedInfo.Pointers.push_back(LB.emitRawPointer(CGF));
             StructBaseCombinedInfo.Sizes.push_back(CGF.Builder.CreateIntCast(
                 Size, CGF.Int64Ty, /*isSigned=*/true));
             StructBaseCombinedInfo.NonContigInfo.Dims.push_back(
@@ -8214,11 +8220,11 @@ class MappableExprsHandler {
     }
     CombinedInfo.Exprs.push_back(VD);
     // Base is the base of the struct
-    CombinedInfo.BasePointers.push_back(PartialStruct.Base.getPointer());
+    CombinedInfo.BasePointers.push_back(PartialStruct.Base.emitRawPointer(CGF));
     CombinedInfo.DevicePtrDecls.push_back(nullptr);
     CombinedInfo.DevicePointers.push_back(DeviceInfoTy::None);
     // Pointer is the address of the lowest element
-    llvm::Value *LB = LBAddr.getPointer();
+    llvm::Value *LB = LBAddr.emitRawPointer(CGF);
     const CXXMethodDecl *MD =
         CGF.CurFuncDecl ? dyn_cast<CXXMethodDecl>(CGF.CurFuncDecl) : nullptr;
     const CXXRecordDecl *RD = MD ? MD->getParent() : nullptr;
@@ -8232,7 +8238,7 @@ class MappableExprsHandler {
       // if the this[:1] expression had appeared in a map clause with a map-type
       // of tofrom.
       // Emit this[:1]
-      CombinedInfo.Pointers.push_back(PartialStruct.Base.getPointer());
+      CombinedInfo.Pointers.push_back(PartialStruct.Base.emitRawPointer(CGF));
       QualType Ty = MD->getFunctionObjectParameterType();
       llvm::Value *Size =
           CGF.Builder.CreateIntCast(CGF.getTypeSize(Ty), CGF.Int64Ty,
@@ -8241,7 +8247,7 @@ class MappableExprsHandler {
     } else {
       CombinedInfo.Pointers.push_back(LB);
       // Size is (addr of {highest+1} element) - (addr of lowest element)
-      llvm::Value *HB = HBAddr.getPointer();
+      llvm::Value *HB = HBAddr.emitRawPointer(CGF);
       llvm::Value *HAddr = CGF.Builder.CreateConstGEP1_32(
           HBAddr.getElementType(), HB, /*Idx0=*/1);
       llvm::Value *CLAddr = CGF.Builder.CreatePointerCast(LB, CGF.VoidPtrTy);
@@ -8750,7 +8756,7 @@ class MappableExprsHandler {
         Address PtrAddr = CGF.EmitLoadOfReference(CGF.MakeAddrLValue(
             CV, ElementType, CGF.getContext().getDeclAlign(VD),
             AlignmentSource::Decl));
-        CombinedInfo.Pointers.push_back(PtrAddr.getPointer());
+        CombinedInfo.Pointers.push_back(PtrAddr.emitRawPointer(CGF));
       } else {
         CombinedInfo.Pointers.push_back(CV);
       }
@@ -9563,10 +9569,11 @@ static void emitTargetCallKernelLaunch(
     bool HasNoWait = D.hasClausesOfKind<OMPNowaitClause>();
     unsigned NumTargetItems = InputInfo.NumberOfTargetItems;
 
-    llvm::Value *BasePointersArray = InputInfo.BasePointersArray.getPointer();
-    llvm::Value *PointersArray = InputInfo.PointersArray.getPointer();
-    llvm::Value *SizesArray = InputInfo.SizesArray.getPointer();
-    llvm::Value *MappersArray = InputInfo.MappersArray.getPointer();
+    llvm::Value *BasePointersArray =
+        InputInfo.BasePointersArray.emitRawPointer(CGF);
+    llvm::Value *PointersArray = InputInfo.PointersArray.emitRawPointer(CGF);
+    llvm::Value *SizesArray = InputInfo.SizesArray.emitRawPointer(CGF);
+    llvm::Value *MappersArray = InputInfo.MappersArray.emitRawPointer(CGF);
 
     auto &&EmitTargetCallFallbackCB =
         [&OMPRuntime, OutlinedFn, &D, &CapturedVars, RequiresOuterTask, &CS,
@@ -10314,15 +10321,16 @@ void CGOpenMPRuntime::emitTargetDataStandAloneCall(
     // Source location for the ident struct
     llvm::Value *RTLoc = emitUpdateLocation(CGF, D.getBeginLoc());
 
-    llvm::Value *OffloadingArgs[] = {RTLoc,
-                                     DeviceID,
-                                     PointerNum,
-                                     InputInfo.BasePointersArray.getPointer(),
-                                     InputInfo.PointersArray.getPointer(),
-                                     InputInfo.SizesArray.getPointer(),
-                                     MapTypesArray,
-                                     MapNamesArray,
-                                     InputInfo.MappersArray.getPointer()};
+    llvm::Value *OffloadingArgs[] = {
+        RTLoc,
+        DeviceID,
+        PointerNum,
+        InputInfo.BasePointersArray.emitRawPointer(CGF),
+        InputInfo.PointersArray.emitRawPointer(CGF),
+        InputInfo.SizesArray.emitRawPointer(CGF),
+        MapTypesArray,
+        MapNamesArray,
+        InputInfo.MappersArray.emitRawPointer(CGF)};
 
     // Select the right runtime function call for each standalone
     // directive.
@@ -11133,7 +11141,7 @@ void CGOpenMPRuntime::emitDoacrossInit(CodeGenFunction &CGF,
       getThreadID(CGF, D.getBeginLoc()),
       llvm::ConstantInt::getSigned(CGM.Int32Ty, NumIterations.size()),
       CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
-          CGF.Builder.CreateConstArrayGEP(DimsAddr, 0).getPointer(),
+          CGF.Builder.CreateConstArrayGEP(DimsAddr, 0).emitRawPointer(CGF),
           CGM.VoidPtrTy)};
 
   llvm::FunctionCallee RTLFn = OMPBuilder.getOrCreateRuntimeFunction(
@@ -11167,7 +11175,8 @@ static void EmitDoacrossOrdered(CodeGenFunction &CGF, CodeGenModule &CGM,
                           /*Volatile=*/false, Int64Ty);
   }
   llvm::Value *Args[] = {
-      ULoc, ThreadID, CGF.Builder.CreateConstArrayGEP(CntAddr, 0).getPointer()};
+      ULoc, ThreadID,
+      CGF.Builder.CreateConstArrayGEP(CntAddr, 0).emitRawPointer(CGF)};
   llvm::FunctionCallee RTLFn;
   llvm::OpenMPIRBuilder &OMPBuilder = CGM.getOpenMPRuntime().getOMPBuilder();
   OMPDoacrossKind<T> ODK;
@@ -11337,7 +11346,7 @@ Address CGOpenMPRuntime::getAddressOfLocalVariable(CodeGenFunction &CGF,
         Args[0] = CGF.CGM.getOpenMPRuntime().getThreadID(
             CGF, SourceLocation::getFromRawEncoding(LocEncoding));
         Args[1] = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
-            Addr.getPointer(), CGF.VoidPtrTy);
+            Addr.emitRawPointer(CGF), CGF.VoidPtrTy);
         llvm::Value *AllocVal = getAllocatorVal(CGF, AllocExpr);
         Args[2] = AllocVal;
         CGF.EmitRuntimeCall(RTLFn, Args);
@@ -11695,15 +11704,17 @@ void CGOpenMPRuntime::emitLastprivateConditionalUpdate(CodeGenFunction &CGF,
       LLIVTy, getName({UniqueDeclName, "iv"}));
   cast<llvm::GlobalVariable>(LastIV)->setAlignment(
       IVLVal.getAlignment().getAsAlign());
-  LValue LastIVLVal = CGF.MakeNaturalAlignAddrLValue(LastIV, IVLVal.getType());
+  LValue LastIVLVal =
+      CGF.MakeNaturalAlignRawAddrLValue(LastIV, IVLVal.getType());
 
   // Last value of the lastprivate conditional.
   // decltype(priv_a) last_a;
   llvm::GlobalVariable *Last = OMPBuilder.getOrCreateInternalVariable(
       CGF.ConvertTypeForMem(LVal.getType()), UniqueDeclName);
-  Last->setAlignment(LVal.getAlignment().getAsAlign());
-  LValue LastLVal = CGF.MakeAddrLValue(
-      Address(Last, Last->getValueType(), LVal.getAlignment()), LVal.getType());
+  cast<llvm::GlobalVariable>(Last)->setAlignment(
+      LVal.getAlignment().getAsAlign());
+  LValue LastLVal =
+      CGF.MakeRawAddrLValue(Last, LVal.getType(), LVal.getAlignment());
 
   // Global loop counter. Required to handle inner parallel-for regions.
   // iv
@@ -11876,9 +11887,8 @@ void CGOpenMPRuntime::emitLastprivateConditionalFinalUpdate(
   // The variable was not updated in the region - exit.
   if (!GV)
     return;
-  LValue LPLVal = CGF.MakeAddrLValue(
-      Address(GV, GV->getValueType(), PrivLVal.getAlignment()),
-      PrivLVal.getType().getNonReferenceType());
+  LValue LPLVal = CGF.MakeRawAddrLValue(
+      GV, PrivLVal.getType().getNonReferenceType(), PrivLVal.getAlignment());
   llvm::Value *Res = CGF.EmitLoadOfScalar(LPLVal, Loc);
   CGF.EmitStoreOfScalar(Res, PrivLVal);
 }
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index c3206427b143e..522ae3d35d22d 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1068,13 +1068,12 @@ class CGOpenMPRuntime {
   /// \param Loc Location of the reference to threadprivate var.
   /// \return Address of the threadprivate variable for the current thread.
   virtual Address getAddrOfThreadPrivate(CodeGenFunction &CGF,
-                                         const VarDecl *VD,
-                                         Address VDAddr,
+                                         const VarDecl *VD, Address VDAddr,
                                          SourceLocation Loc);
 
   /// Returns the address of the variable marked as declare target with link
   /// clause OR as declare target with to clause and unified memory.
-  virtual Address getAddrOfDeclareTargetVar(const VarDecl *VD);
+  virtual ConstantAddress getAddrOfDeclareTargetVar(const VarDecl *VD);
 
   /// Emit a code for initialization of threadprivate variable. It emits
   /// a call to runtime library which adds initial value to the newly created
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 242f5f144a68c..eb716520e5ff5 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -1101,7 +1101,8 @@ void CGOpenMPRuntimeGPU::emitGenericVarsProlog(CodeGenFunction &CGF,
     llvm::PointerType *VarPtrTy = CGF.ConvertTypeForMem(VarTy)->getPointerTo();
     llvm::Value *CastedVoidPtr = Bld.CreatePointerBitCastOrAddrSpaceCast(
         VoidPtr, VarPtrTy, VD->getName() + "_on_stack");
-    LValue VarAddr = CGF.MakeNaturalAlignAddrLValue(CastedVoidPtr, VarTy);
+    LValue VarAddr =
+        CGF.MakeNaturalAlignPointeeRawAddrLValue(CastedVoidPtr, VarTy);
     Rec.second.PrivateAddr = VarAddr.getAddress(CGF);
     Rec.second.GlobalizedVal = VoidPtr;
 
@@ -1211,8 +1212,8 @@ void CGOpenMPRuntimeGPU::emitTeamsCall(CodeGenFunction &CGF,
 
   bool IsBareKernel = D.getSingleClause<OMPXBareClause>();
 
-  Address ZeroAddr = CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
-                                                      /*Name=*/".zero.addr");
+  RawAddress ZeroAddr = CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
+                                                         /*Name=*/".zero.addr");
   CGF.Builder.CreateStore(CGF.Builder.getInt32(/*C*/ 0), ZeroAddr);
   llvm::SmallVector<llvm::Value *, 16> OutlinedFnArgs;
   // We don't emit any thread id function call in bare kernel, but because the
@@ -1220,7 +1221,7 @@ void CGOpenMPRuntimeGPU::emitTeamsCall(CodeGenFunction &CGF,
   if (IsBareKernel)
     OutlinedFnArgs.push_back(llvm::ConstantPointerNull::get(CGM.VoidPtrTy));
   else
-    OutlinedFnArgs.push_back(emitThreadIDAddress(CGF, Loc).getPointer());
+    OutlinedFnArgs.push_back(emitThreadIDAddress(CGF, Loc).emitRawPointer(CGF));
   OutlinedFnArgs.push_back(ZeroAddr.getPointer());
   OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());
   emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);
@@ -1294,7 +1295,7 @@ void CGOpenMPRuntimeGPU::emitParallelCall(CodeGenFunction &CGF,
         llvm::ConstantInt::get(CGF.Int32Ty, -1),
         FnPtr,
         ID,
-        Bld.CreateBitOrPointerCast(CapturedVarsAddrs.getPointer(),
+        Bld.CreateBitOrPointerCast(CapturedVarsAddrs.emitRawPointer(CGF),
                                    CGF.VoidPtrPtrTy),
         llvm::ConstantInt::get(CGM.SizeTy, CapturedVars.size())};
     CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
@@ -1508,17 +1509,18 @@ static void shuffleAndStore(CodeGenFunction &CGF, Address SrcAddr,
       CGF.EmitBlock(PreCondBB);
       llvm::PHINode *PhiSrc =
           Bld.CreatePHI(Ptr.getType(), /*NumReservedValues=*/2);
-      PhiSrc->addIncoming(Ptr.getPointer(), CurrentBB);
+      PhiSrc->addIncoming(Ptr.emitRawPointer(CGF), CurrentBB);
       llvm::PHINode *PhiDest =
           Bld.CreatePHI(ElemPtr.getType(), /*NumReservedValues=*/2);
-      PhiDest->addIncoming(ElemPtr.getPointer(), CurrentBB);
+      PhiDest->addIncoming(ElemPtr.emitRawPointer(CGF), CurrentBB);
       Ptr = Address(PhiSrc, Ptr.getElementType(), Ptr.getAlignment());
       ElemPtr =
           Address(PhiDest, ElemPtr.getElementType(), ElemPtr.getAlignment());
+      llvm::Value *PtrEndRaw = PtrEnd.emitRawPointer(CGF);
+      llvm::Value *PtrRaw = Ptr.emitRawPointer(CGF);
       llvm::Value *PtrDiff = Bld.CreatePtrDiff(
-          CGF.Int8Ty, PtrEnd.getPointer(),
-          Bld.CreatePointerBitCastOrAddrSpaceCast(Ptr.getPointer(),
-                                                  CGF.VoidPtrTy));
+          CGF.Int8Ty, PtrEndRaw,
+          Bld.CreatePointerBitCastOrAddrSpaceCast(PtrRaw, CGF.VoidPtrTy));
       Bld.CreateCondBr(Bld.CreateICmpSGT(PtrDiff, Bld.getInt64(IntSize - 1)),
                        ThenBB, ExitBB);
       CGF.EmitBlock(ThenBB);
@@ -1533,8 +1535,8 @@ static void shuffleAndStore(CodeGenFunction &CGF, Address SrcAddr,
                             TBAAAccessInfo());
       Address LocalPtr = Bld.CreateConstGEP(Ptr, 1);
       Address LocalElemPtr = Bld.CreateConstGEP(ElemPtr, 1);
-      PhiSrc->addIncoming(LocalPtr.getPointer(), ThenBB);
-      PhiDest->addIncoming(LocalElemPtr.getPointer(), ThenBB);
+      PhiSrc->addIncoming(LocalPtr.emitRawPointer(CGF), ThenBB);
+      PhiDest->addIncoming(LocalElemPtr.emitRawPointer(CGF), ThenBB);
       CGF.EmitBranch(PreCondBB);
       CGF.EmitBlock(ExitBB);
     } else {
@@ -1681,10 +1683,10 @@ static void emitReductionListCopy(
     // scope and that of functions it invokes (i.e., reduce_function).
     // RemoteReduceData[i] = (void*)&RemoteElem
     if (UpdateDestListPtr) {
-      CGF.EmitStoreOfScalar(Bld.CreatePointerBitCastOrAddrSpaceCast(
-                                DestElementAddr.getPointer(), CGF.VoidPtrTy),
-                            DestElementPtrAddr, /*Volatile=*/false,
-                            C.VoidPtrTy);
+      CGF.EmitStoreOfScalar(
+          Bld.CreatePointerBitCastOrAddrSpaceCast(
+              DestElementAddr.emitRawPointer(CGF), CGF.VoidPtrTy),
+          DestElementPtrAddr, /*Volatile=*/false, C.VoidPtrTy);
     }
 
     ++Idx;
@@ -1835,7 +1837,7 @@ static llvm::Value *emitInterWarpCopyFunction(CodeGenModule &CGM,
       // elemptr = ((CopyType*)(elemptrptr)) + I
       Address ElemPtr(ElemPtrPtr, CopyType, Align);
       if (NumIters > 1)
-        ElemPtr = Bld.CreateGEP(ElemPtr, Cnt);
+        ElemPtr = Bld.CreateGEP(CGF, ElemPtr, Cnt);
 
       // Get pointer to location in transfer medium.
       // MediumPtr = &medium[warp_id]
@@ -1899,7 +1901,7 @@ static llvm::Value *emitInterWarpCopyFunction(CodeGenModule &CGM,
           TargetElemPtrPtr, /*Volatile=*/false, C.VoidPtrTy, Loc);
       Address TargetElemPtr(TargetElemPtrVal, CopyType, Align);
       if (NumIters > 1)
-        TargetElemPtr = Bld.CreateGEP(TargetElemPtr, Cnt);
+        TargetElemPtr = Bld.CreateGEP(CGF, TargetElemPtr, Cnt);
 
       // *TargetElemPtr = SrcMediumVal;
       llvm::Value *SrcMediumValue =
@@ -2110,9 +2112,9 @@ static llvm::Function *emitShuffleAndReduceFunction(
   CGF.EmitBlock(ThenBB);
   // reduce_function(LocalReduceList, RemoteReduceList)
   llvm::Value *LocalReduceListPtr = Bld.CreatePointerBitCastOrAddrSpaceCast(
-      LocalReduceList.getPointer(), CGF.VoidPtrTy);
+      LocalReduceList.emitRawPointer(CGF), CGF.VoidPtrTy);
   llvm::Value *RemoteReduceListPtr = Bld.CreatePointerBitCastOrAddrSpaceCast(
-      RemoteReduceList.getPointer(), CGF.VoidPtrTy);
+      RemoteReduceList.emitRawPointer(CGF), CGF.VoidPtrTy);
   CGM.getOpenMPRuntime().emitOutlinedFunctionCall(
       CGF, Loc, ReduceFn, {LocalReduceListPtr, RemoteReduceListPtr});
   Bld.CreateBr(MergeBB);
@@ -2223,9 +2225,9 @@ static llvm::Value *emitListToGlobalCopyFunction(
     llvm::Value *BufferPtr =
         Bld.CreateInBoundsGEP(LLVMReductionsBufferTy, BufferArrPtr, Idxs);
     LValue GlobLVal = CGF.EmitLValueForField(
-        CGF.MakeNaturalAlignAddrLValue(BufferPtr, StaticTy), FD);
+        CGF.MakeNaturalAlignRawAddrLValue(BufferPtr, StaticTy), FD);
     Address GlobAddr = GlobLVal.getAddress(CGF);
-    GlobLVal.setAddress(Address(GlobAddr.getPointer(),
+    GlobLVal.setAddress(Address(GlobAddr.emitRawPointer(CGF),
                                 CGF.ConvertTypeForMem(Private->getType()),
                                 GlobAddr.getAlignment()));
     switch (CGF.getEvaluationKind(Private->getType())) {
@@ -2309,7 +2311,7 @@ static llvm::Value *emitListToGlobalReduceFunction(
 
   // 1. Build a list of reduction variables.
   // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
-  Address ReductionList =
+  RawAddress ReductionList =
       CGF.CreateMemTemp(ReductionArrayTy, ".omp.reduction.red_list");
   auto IPriv = Privates.begin();
   llvm::Value *Idxs[] = {CGF.EmitLoadOfScalar(CGF.GetAddrOfLocalVar(&IdxArg),
@@ -2324,10 +2326,10 @@ static llvm::Value *emitListToGlobalReduceFunction(
     llvm::Value *BufferPtr =
         Bld.CreateInBoundsGEP(LLVMReductionsBufferTy, BufferArrPtr, Idxs);
     LValue GlobLVal = CGF.EmitLValueForField(
-        CGF.MakeNaturalAlignAddrLValue(BufferPtr, StaticTy), FD);
+        CGF.MakeNaturalAlignRawAddrLValue(BufferPtr, StaticTy), FD);
     Address GlobAddr = GlobLVal.getAddress(CGF);
-    CGF.EmitStoreOfScalar(GlobAddr.getPointer(), Elem, /*Volatile=*/false,
-                          C.VoidPtrTy);
+    CGF.EmitStoreOfScalar(GlobAddr.emitRawPointer(CGF), Elem,
+                          /*Volatile=*/false, C.VoidPtrTy);
     if ((*IPriv)->getType()->isVariablyModifiedType()) {
       // Store array size.
       ++Idx;
@@ -2430,9 +2432,9 @@ static llvm::Value *emitGlobalToListCopyFunction(
     llvm::Value *BufferPtr =
         Bld.CreateInBoundsGEP(LLVMReductionsBufferTy, BufferArrPtr, Idxs);
     LValue GlobLVal = CGF.EmitLValueForField(
-        CGF.MakeNaturalAlignAddrLValue(BufferPtr, StaticTy), FD);
+        CGF.MakeNaturalAlignRawAddrLValue(BufferPtr, StaticTy), FD);
     Address GlobAddr = GlobLVal.getAddress(CGF);
-    GlobLVal.setAddress(Address(GlobAddr.getPointer(),
+    GlobLVal.setAddress(Address(GlobAddr.emitRawPointer(CGF),
                                 CGF.ConvertTypeForMem(Private->getType()),
                                 GlobAddr.getAlignment()));
     switch (CGF.getEvaluationKind(Private->getType())) {
@@ -2531,10 +2533,10 @@ static llvm::Value *emitGlobalToListReduceFunction(
     llvm::Value *BufferPtr =
         Bld.CreateInBoundsGEP(LLVMReductionsBufferTy, BufferArrPtr, Idxs);
     LValue GlobLVal = CGF.EmitLValueForField(
-        CGF.MakeNaturalAlignAddrLValue(BufferPtr, StaticTy), FD);
+        CGF.MakeNaturalAlignRawAddrLValue(BufferPtr, StaticTy), FD);
     Address GlobAddr = GlobLVal.getAddress(CGF);
-    CGF.EmitStoreOfScalar(GlobAddr.getPointer(), Elem, /*Volatile=*/false,
-                          C.VoidPtrTy);
+    CGF.EmitStoreOfScalar(GlobAddr.emitRawPointer(CGF), Elem,
+                          /*Volatile=*/false, C.VoidPtrTy);
     if ((*IPriv)->getType()->isVariablyModifiedType()) {
       // Store array size.
       ++Idx;
@@ -2550,7 +2552,7 @@ static llvm::Value *emitGlobalToListReduceFunction(
   }
 
   // Call reduce_function(ReduceList, GlobalReduceList)
-  llvm::Value *GlobalReduceList = ReductionList.getPointer();
+  llvm::Value *GlobalReduceList = ReductionList.emitRawPointer(CGF);
   Address AddrReduceListArg = CGF.GetAddrOfLocalVar(&ReduceListArg);
   llvm::Value *ReducedPtr = CGF.EmitLoadOfScalar(
       AddrReduceListArg, /*Volatile=*/false, C.VoidPtrTy, Loc);
@@ -2881,7 +2883,7 @@ void CGOpenMPRuntimeGPU::emitReduction(
   }
 
   llvm::Value *RL = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
-      ReductionList.getPointer(), CGF.VoidPtrTy);
+      ReductionList.emitRawPointer(CGF), CGF.VoidPtrTy);
   llvm::Function *ReductionFn = emitReductionFunction(
       CGF.CurFn->getName(), Loc, CGF.ConvertTypeForMem(ReductionArrayTy),
       Privates, LHSExprs, RHSExprs, ReductionOps);
@@ -3111,15 +3113,15 @@ llvm::Function *CGOpenMPRuntimeGPU::createParallelDataSharingWrapper(
   // Get the array of arguments.
   SmallVector<llvm::Value *, 8> Args;
 
-  Args.emplace_back(CGF.GetAddrOfLocalVar(&WrapperArg).getPointer());
-  Args.emplace_back(ZeroAddr.getPointer());
+  Args.emplace_back(CGF.GetAddrOfLocalVar(&WrapperArg).emitRawPointer(CGF));
+  Args.emplace_back(ZeroAddr.emitRawPointer(CGF));
 
   CGBuilderTy &Bld = CGF.Builder;
   auto CI = CS.capture_begin();
 
   // Use global memory for data sharing.
   // Handle passing of global args to workers.
-  Address GlobalArgs =
+  RawAddress GlobalArgs =
       CGF.CreateDefaultAlignTempAlloca(CGF.VoidPtrPtrTy, "global_args");
   llvm::Value *GlobalArgsPtr = GlobalArgs.getPointer();
   llvm::Value *DataSharingArgs[] = {GlobalArgsPtr};
@@ -3405,7 +3407,7 @@ void CGOpenMPRuntimeGPU::adjustTargetSpecificDataForLambdas(
         VDAddr = CGF.EmitLoadOfReferenceLValue(VDAddr,
                                                VD->getType().getCanonicalType())
                      .getAddress(CGF);
-      CGF.EmitStoreOfScalar(VDAddr.getPointer(), VarLVal);
+      CGF.EmitStoreOfScalar(VDAddr.emitRawPointer(CGF), VarLVal);
     }
   }
 }
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index cb5a004e4f4a6..576fe2f7a2d46 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -2294,7 +2294,7 @@ std::pair<llvm::Value*, llvm::Type *> CodeGenFunction::EmitAsmInputLValue(
 
   Address Addr = InputValue.getAddress(*this);
   ConstraintStr += '*';
-  return {Addr.getPointer(), Addr.getElementType()};
+  return {InputValue.getPointer(*this), Addr.getElementType()};
 }
 
 std::pair<llvm::Value *, llvm::Type *>
@@ -2701,7 +2701,7 @@ void CodeGenFunction::EmitAsmStmt(const AsmStmt &S) {
 
       ArgTypes.push_back(DestAddr.getType());
       ArgElemTypes.push_back(DestAddr.getElementType());
-      Args.push_back(DestAddr.getPointer());
+      Args.push_back(DestAddr.emitRawPointer(*this));
       Constraints += "=*";
       Constraints += OutputConstraint;
       ReadOnly = ReadNone = false;
@@ -3076,8 +3076,8 @@ CodeGenFunction::GenerateCapturedStmtFunction(const CapturedStmt &S) {
   CapturedStmtInfo->setContextValue(Builder.CreateLoad(DeclPtr));
 
   // Initialize variable-length arrays.
-  LValue Base = MakeNaturalAlignAddrLValue(CapturedStmtInfo->getContextValue(),
-                                           Ctx.getTagDeclType(RD));
+  LValue Base = MakeNaturalAlignRawAddrLValue(
+      CapturedStmtInfo->getContextValue(), Ctx.getTagDeclType(RD));
   for (auto *FD : RD->fields()) {
     if (FD->hasCapturedVLAType()) {
       auto *ExprArg =
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index fb93f35dca4ae..c294b6e69ace3 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -354,7 +354,8 @@ void CodeGenFunction::GenerateOpenMPCapturedVars(
         LValue DstLV = MakeAddrLValue(DstAddr, Ctx.getUIntPtrType());
 
         llvm::Value *SrcAddrVal = EmitScalarConversion(
-            DstAddr.getPointer(), Ctx.getPointerType(Ctx.getUIntPtrType()),
+            DstAddr.emitRawPointer(*this),
+            Ctx.getPointerType(Ctx.getUIntPtrType()),
             Ctx.getPointerType(CurField->getType()), CurCap->getLocation());
         LValue SrcLV =
             MakeNaturalAlignAddrLValue(SrcAddrVal, CurField->getType());
@@ -368,7 +369,8 @@ void CodeGenFunction::GenerateOpenMPCapturedVars(
       CapturedVars.push_back(CV);
     } else {
       assert(CurCap->capturesVariable() && "Expected capture by reference.");
-      CapturedVars.push_back(EmitLValue(*I).getAddress(*this).getPointer());
+      CapturedVars.push_back(
+          EmitLValue(*I).getAddress(*this).emitRawPointer(*this));
     }
   }
 }
@@ -379,8 +381,9 @@ static Address castValueFromUintptr(CodeGenFunction &CGF, SourceLocation Loc,
   ASTContext &Ctx = CGF.getContext();
 
   llvm::Value *CastedPtr = CGF.EmitScalarConversion(
-      AddrLV.getAddress(CGF).getPointer(), Ctx.getUIntPtrType(),
+      AddrLV.getAddress(CGF).emitRawPointer(CGF), Ctx.getUIntPtrType(),
       Ctx.getPointerType(DstType), Loc);
+  // FIXME: should the pointee type (DstType) be passed?
   Address TmpAddr =
       CGF.MakeNaturalAlignAddrLValue(CastedPtr, DstType).getAddress(CGF);
   return TmpAddr;
@@ -706,8 +709,8 @@ void CodeGenFunction::EmitOMPAggregateAssign(
   llvm::Value *NumElements = emitArrayLength(ArrayTy, ElementTy, DestAddr);
   SrcAddr = SrcAddr.withElementType(DestAddr.getElementType());
 
-  llvm::Value *SrcBegin = SrcAddr.getPointer();
-  llvm::Value *DestBegin = DestAddr.getPointer();
+  llvm::Value *SrcBegin = SrcAddr.emitRawPointer(*this);
+  llvm::Value *DestBegin = DestAddr.emitRawPointer(*this);
   // Cast from pointer to array type to pointer to single element.
   llvm::Value *DestEnd = Builder.CreateInBoundsGEP(DestAddr.getElementType(),
                                                    DestBegin, NumElements);
@@ -1011,10 +1014,10 @@ bool CodeGenFunction::EmitOMPCopyinClause(const OMPExecutableDirective &D) {
           CopyBegin = createBasicBlock("copyin.not.master");
           CopyEnd = createBasicBlock("copyin.not.master.end");
           // TODO: Avoid ptrtoint conversion.
-          auto *MasterAddrInt =
-              Builder.CreatePtrToInt(MasterAddr.getPointer(), CGM.IntPtrTy);
-          auto *PrivateAddrInt =
-              Builder.CreatePtrToInt(PrivateAddr.getPointer(), CGM.IntPtrTy);
+          auto *MasterAddrInt = Builder.CreatePtrToInt(
+              MasterAddr.emitRawPointer(*this), CGM.IntPtrTy);
+          auto *PrivateAddrInt = Builder.CreatePtrToInt(
+              PrivateAddr.emitRawPointer(*this), CGM.IntPtrTy);
           Builder.CreateCondBr(
               Builder.CreateICmpNE(MasterAddrInt, PrivateAddrInt), CopyBegin,
               CopyEnd);
@@ -1673,7 +1676,7 @@ Address CodeGenFunction::OMPBuilderCBHelpers::getAddrOfThreadPrivate(
 
   llvm::Type *VarTy = VDAddr.getElementType();
   llvm::Value *Data =
-      CGF.Builder.CreatePointerCast(VDAddr.getPointer(), CGM.Int8PtrTy);
+      CGF.Builder.CreatePointerCast(VDAddr.emitRawPointer(CGF), CGM.Int8PtrTy);
   llvm::ConstantInt *Size = CGM.getSize(CGM.GetTargetTypeStoreSize(VarTy));
   std::string Suffix = getNameWithSeparators({"cache", ""});
   llvm::Twine CacheName = Twine(CGM.getMangledName(VD)).concat(Suffix);
@@ -2052,7 +2055,7 @@ void CodeGenFunction::EmitOMPCanonicalLoop(const OMPCanonicalLoop *S) {
                            ->getParam(0)
                            ->getType()
                            .getNonReferenceType();
-  Address CountAddr = CreateMemTemp(LogicalTy, ".count.addr");
+  RawAddress CountAddr = CreateMemTemp(LogicalTy, ".count.addr");
   emitCapturedStmtCall(*this, DistanceClosure, {CountAddr.getPointer()});
   llvm::Value *DistVal = Builder.CreateLoad(CountAddr, ".count");
 
@@ -2068,7 +2071,7 @@ void CodeGenFunction::EmitOMPCanonicalLoop(const OMPCanonicalLoop *S) {
     LValue LCVal = EmitLValue(LoopVarRef);
     Address LoopVarAddress = LCVal.getAddress(*this);
     emitCapturedStmtCall(*this, LoopVarClosure,
-                         {LoopVarAddress.getPointer(), IndVar});
+                         {LoopVarAddress.emitRawPointer(*this), IndVar});
 
     RunCleanupsScope BodyScope(*this);
     EmitStmt(BodyStmt);
@@ -4802,7 +4805,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
       ParamTypes.push_back(PrivatesPtr->getType());
       for (const Expr *E : Data.PrivateVars) {
         const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
-        Address PrivatePtr = CGF.CreateMemTemp(
+        RawAddress PrivatePtr = CGF.CreateMemTemp(
             CGF.getContext().getPointerType(E->getType()), ".priv.ptr.addr");
         PrivatePtrs.emplace_back(VD, PrivatePtr);
         CallArgs.push_back(PrivatePtr.getPointer());
@@ -4810,7 +4813,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
       }
       for (const Expr *E : Data.FirstprivateVars) {
         const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
-        Address PrivatePtr =
+        RawAddress PrivatePtr =
             CGF.CreateMemTemp(CGF.getContext().getPointerType(E->getType()),
                               ".firstpriv.ptr.addr");
         PrivatePtrs.emplace_back(VD, PrivatePtr);
@@ -4820,7 +4823,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
       }
       for (const Expr *E : Data.LastprivateVars) {
         const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
-        Address PrivatePtr =
+        RawAddress PrivatePtr =
             CGF.CreateMemTemp(CGF.getContext().getPointerType(E->getType()),
                               ".lastpriv.ptr.addr");
         PrivatePtrs.emplace_back(VD, PrivatePtr);
@@ -4833,7 +4836,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
           Ty = CGF.getContext().getPointerType(Ty);
         if (isAllocatableDecl(VD))
           Ty = CGF.getContext().getPointerType(Ty);
-        Address PrivatePtr = CGF.CreateMemTemp(
+        RawAddress PrivatePtr = CGF.CreateMemTemp(
             CGF.getContext().getPointerType(Ty), ".local.ptr.addr");
         auto Result = UntiedLocalVars.insert(
             std::make_pair(VD, std::make_pair(PrivatePtr, Address::invalid())));
@@ -4866,7 +4869,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
         if (auto *DI = CGF.getDebugInfo())
           if (CGF.CGM.getCodeGenOpts().hasReducedDebugInfo())
             (void)DI->EmitDeclareOfAutoVariable(
-                Pair.first, Pair.second.getPointer(), CGF.Builder,
+                Pair.first, Pair.second.getBasePointer(), CGF.Builder,
                 /*UsePointerValue*/ true);
       }
       // Adjust mapping for internal locals by mapping actual memory instead of
@@ -4919,14 +4922,14 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
                                                            RedCG, Cnt);
         Address Replacement = CGF.CGM.getOpenMPRuntime().getTaskReductionItem(
             CGF, S.getBeginLoc(), ReductionsPtr, RedCG.getSharedLValue(Cnt));
-        Replacement =
-            Address(CGF.EmitScalarConversion(
-                        Replacement.getPointer(), CGF.getContext().VoidPtrTy,
-                        CGF.getContext().getPointerType(
-                            Data.ReductionCopies[Cnt]->getType()),
-                        Data.ReductionCopies[Cnt]->getExprLoc()),
-                    CGF.ConvertTypeForMem(Data.ReductionCopies[Cnt]->getType()),
-                    Replacement.getAlignment());
+        Replacement = Address(
+            CGF.EmitScalarConversion(Replacement.emitRawPointer(CGF),
+                                     CGF.getContext().VoidPtrTy,
+                                     CGF.getContext().getPointerType(
+                                         Data.ReductionCopies[Cnt]->getType()),
+                                     Data.ReductionCopies[Cnt]->getExprLoc()),
+            CGF.ConvertTypeForMem(Data.ReductionCopies[Cnt]->getType()),
+            Replacement.getAlignment());
         Replacement = RedCG.adjustPrivateAddress(CGF, Cnt, Replacement);
         Scope.addPrivate(RedCG.getBaseDecl(Cnt), Replacement);
       }
@@ -4977,7 +4980,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(
             CGF, S.getBeginLoc(), ReductionsPtr, RedCG.getSharedLValue(Cnt));
         Replacement = Address(
             CGF.EmitScalarConversion(
-                Replacement.getPointer(), CGF.getContext().VoidPtrTy,
+                Replacement.emitRawPointer(CGF), CGF.getContext().VoidPtrTy,
                 CGF.getContext().getPointerType(InRedPrivs[Cnt]->getType()),
                 InRedPrivs[Cnt]->getExprLoc()),
             CGF.ConvertTypeForMem(InRedPrivs[Cnt]->getType()),
@@ -5096,7 +5099,7 @@ void CodeGenFunction::EmitOMPTargetTaskBasedDirective(
     // If there is no user-defined mapper, the mapper array will be nullptr. In
     // this case, we don't need to privatize it.
     if (!isa_and_nonnull<llvm::ConstantPointerNull>(
-            InputInfo.MappersArray.getPointer())) {
+            InputInfo.MappersArray.emitRawPointer(*this))) {
       MVD = createImplicitFirstprivateForType(
           getContext(), Data, BaseAndPointerAndMapperType, CD, S.getBeginLoc());
       TargetScope.addPrivate(MVD, InputInfo.MappersArray);
@@ -5122,7 +5125,7 @@ void CodeGenFunction::EmitOMPTargetTaskBasedDirective(
       ParamTypes.push_back(PrivatesPtr->getType());
       for (const Expr *E : Data.FirstprivateVars) {
         const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
-        Address PrivatePtr =
+        RawAddress PrivatePtr =
             CGF.CreateMemTemp(CGF.getContext().getPointerType(E->getType()),
                               ".firstpriv.ptr.addr");
         PrivatePtrs.emplace_back(VD, PrivatePtr);
@@ -5201,14 +5204,14 @@ void CodeGenFunction::processInReduction(const OMPExecutableDirective &S,
                                                          RedCG, Cnt);
       Address Replacement = CGF.CGM.getOpenMPRuntime().getTaskReductionItem(
           CGF, S.getBeginLoc(), ReductionsPtr, RedCG.getSharedLValue(Cnt));
-      Replacement =
-          Address(CGF.EmitScalarConversion(
-                      Replacement.getPointer(), CGF.getContext().VoidPtrTy,
-                      CGF.getContext().getPointerType(
-                          Data.ReductionCopies[Cnt]->getType()),
-                      Data.ReductionCopies[Cnt]->getExprLoc()),
-                  CGF.ConvertTypeForMem(Data.ReductionCopies[Cnt]->getType()),
-                  Replacement.getAlignment());
+      Replacement = Address(
+          CGF.EmitScalarConversion(Replacement.emitRawPointer(CGF),
+                                   CGF.getContext().VoidPtrTy,
+                                   CGF.getContext().getPointerType(
+                                       Data.ReductionCopies[Cnt]->getType()),
+                                   Data.ReductionCopies[Cnt]->getExprLoc()),
+          CGF.ConvertTypeForMem(Data.ReductionCopies[Cnt]->getType()),
+          Replacement.getAlignment());
       Replacement = RedCG.adjustPrivateAddress(CGF, Cnt, Replacement);
       Scope.addPrivate(RedCG.getBaseDecl(Cnt), Replacement);
     }
@@ -5254,7 +5257,7 @@ void CodeGenFunction::processInReduction(const OMPExecutableDirective &S,
           CGF, S.getBeginLoc(), ReductionsPtr, RedCG.getSharedLValue(Cnt));
       Replacement = Address(
           CGF.EmitScalarConversion(
-              Replacement.getPointer(), CGF.getContext().VoidPtrTy,
+              Replacement.emitRawPointer(CGF), CGF.getContext().VoidPtrTy,
               CGF.getContext().getPointerType(InRedPrivs[Cnt]->getType()),
               InRedPrivs[Cnt]->getExprLoc()),
           CGF.ConvertTypeForMem(InRedPrivs[Cnt]->getType()),
@@ -5401,7 +5404,7 @@ void CodeGenFunction::EmitOMPDepobjDirective(const OMPDepobjDirective &S) {
     Dependencies.DepExprs.append(DC->varlist_begin(), DC->varlist_end());
     Address DepAddr = CGM.getOpenMPRuntime().emitDepobjDependClause(
         *this, Dependencies, DC->getBeginLoc());
-    EmitStoreOfScalar(DepAddr.getPointer(), DOLVal);
+    EmitStoreOfScalar(DepAddr.emitRawPointer(*this), DOLVal);
     return;
   }
   if (const auto *DC = S.getSingleClause<OMPDestroyClause>()) {
@@ -6478,21 +6481,21 @@ static void emitOMPAtomicCompareExpr(
           D->getType()->hasSignedIntegerRepresentation());
 
   llvm::OpenMPIRBuilder::AtomicOpValue XOpVal{
-      XAddr.getPointer(), XAddr.getElementType(),
+      XAddr.emitRawPointer(CGF), XAddr.getElementType(),
       X->getType()->hasSignedIntegerRepresentation(),
       X->getType().isVolatileQualified()};
   llvm::OpenMPIRBuilder::AtomicOpValue VOpVal, ROpVal;
   if (V) {
     LValue LV = CGF.EmitLValue(V);
     Address Addr = LV.getAddress(CGF);
-    VOpVal = {Addr.getPointer(), Addr.getElementType(),
+    VOpVal = {Addr.emitRawPointer(CGF), Addr.getElementType(),
               V->getType()->hasSignedIntegerRepresentation(),
               V->getType().isVolatileQualified()};
   }
   if (R) {
     LValue LV = CGF.EmitLValue(R);
     Address Addr = LV.getAddress(CGF);
-    ROpVal = {Addr.getPointer(), Addr.getElementType(),
+    ROpVal = {Addr.emitRawPointer(CGF), Addr.getElementType(),
               R->getType()->hasSignedIntegerRepresentation(),
               R->getType().isVolatileQualified()};
   }
@@ -7036,7 +7039,7 @@ void CodeGenFunction::EmitOMPInteropDirective(const OMPInteropDirective &S) {
     std::tie(NumDependences, DependenciesArray) =
         CGM.getOpenMPRuntime().emitDependClause(*this, Data.Dependences,
                                                 S.getBeginLoc());
-    DependenceList = DependenciesArray.getPointer();
+    DependenceList = DependenciesArray.emitRawPointer(*this);
   }
   Data.HasNowaitClause = S.hasClausesOfKind<OMPNowaitClause>();
 
diff --git a/clang/lib/CodeGen/CGVTables.cpp b/clang/lib/CodeGen/CGVTables.cpp
index 8dee3f74b44b4..862369ae009f4 100644
--- a/clang/lib/CodeGen/CGVTables.cpp
+++ b/clang/lib/CodeGen/CGVTables.cpp
@@ -201,14 +201,13 @@ CodeGenFunction::GenerateVarArgsThunk(llvm::Function *Fn,
 
   // Find the first store of "this", which will be to the alloca associated
   // with "this".
-  Address ThisPtr =
-      Address(&*AI, ConvertTypeForMem(MD->getFunctionObjectParameterType()),
-              CGM.getClassPointerAlignment(MD->getParent()));
+  Address ThisPtr = makeNaturalAddressForPointer(
+      &*AI, MD->getFunctionObjectParameterType(),
+      CGM.getClassPointerAlignment(MD->getParent()));
   llvm::BasicBlock *EntryBB = &Fn->front();
   llvm::BasicBlock::iterator ThisStore =
       llvm::find_if(*EntryBB, [&](llvm::Instruction &I) {
-        return isa<llvm::StoreInst>(I) &&
-               I.getOperand(0) == ThisPtr.getPointer();
+        return isa<llvm::StoreInst>(I) && I.getOperand(0) == &*AI;
       });
   assert(ThisStore != EntryBB->end() &&
          "Store of this should be in entry block?");
diff --git a/clang/lib/CodeGen/CGValue.h b/clang/lib/CodeGen/CGValue.h
index 1e6f67250583d..cc9ad10ae5969 100644
--- a/clang/lib/CodeGen/CGValue.h
+++ b/clang/lib/CodeGen/CGValue.h
@@ -14,12 +14,13 @@
 #ifndef LLVM_CLANG_LIB_CODEGEN_CGVALUE_H
 #define LLVM_CLANG_LIB_CODEGEN_CGVALUE_H
 
+#include "Address.h"
+#include "CodeGenTBAA.h"
+#include "EHScopeStack.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/Type.h"
-#include "llvm/IR/Value.h"
 #include "llvm/IR/Type.h"
-#include "Address.h"
-#include "CodeGenTBAA.h"
+#include "llvm/IR/Value.h"
 
 namespace llvm {
   class Constant;
@@ -28,57 +29,64 @@ namespace llvm {
 
 namespace clang {
 namespace CodeGen {
-  class AggValueSlot;
-  class CodeGenFunction;
-  struct CGBitFieldInfo;
+class AggValueSlot;
+class CGBuilderTy;
+class CodeGenFunction;
+struct CGBitFieldInfo;
 
 /// RValue - This trivial value class is used to represent the result of an
 /// expression that is evaluated.  It can be one of three things: either a
 /// simple LLVM SSA value, a pair of SSA values for complex numbers, or the
 /// address of an aggregate value in memory.
 class RValue {
-  enum Flavor { Scalar, Complex, Aggregate };
+  friend struct DominatingValue<RValue>;
 
-  // The shift to make to an aggregate's alignment to make it look
-  // like a pointer.
-  enum { AggAlignShift = 4 };
+  enum FlavorEnum { Scalar, Complex, Aggregate };
 
-  // Stores first value and flavor.
-  llvm::PointerIntPair<llvm::Value *, 2, Flavor> V1;
-  // Stores second value and volatility.
-  llvm::PointerIntPair<llvm::Value *, 1, bool> V2;
-  // Stores element type for aggregate values.
-  llvm::Type *ElementType;
+  union {
+    // Stores first and second value.
+    struct {
+      llvm::Value *first;
+      llvm::Value *second;
+    } Vals;
+
+    // Stores aggregate address.
+    Address AggregateAddr;
+  };
+
+  unsigned IsVolatile : 1;
+  unsigned Flavor : 2;
 
 public:
-  bool isScalar() const { return V1.getInt() == Scalar; }
-  bool isComplex() const { return V1.getInt() == Complex; }
-  bool isAggregate() const { return V1.getInt() == Aggregate; }
+  RValue() : Vals{nullptr, nullptr}, Flavor(Scalar) {}
+
+  bool isScalar() const { return Flavor == Scalar; }
+  bool isComplex() const { return Flavor == Complex; }
+  bool isAggregate() const { return Flavor == Aggregate; }
 
-  bool isVolatileQualified() const { return V2.getInt(); }
+  bool isVolatileQualified() const { return IsVolatile; }
 
   /// getScalarVal() - Return the Value* of this scalar value.
   llvm::Value *getScalarVal() const {
     assert(isScalar() && "Not a scalar!");
-    return V1.getPointer();
+    return Vals.first;
   }
 
   /// getComplexVal - Return the real/imag components of this complex value.
   ///
   std::pair<llvm::Value *, llvm::Value *> getComplexVal() const {
-    return std::make_pair(V1.getPointer(), V2.getPointer());
+    return std::make_pair(Vals.first, Vals.second);
   }
 
   /// getAggregateAddr() - Return the Value* of the address of the aggregate.
   Address getAggregateAddress() const {
     assert(isAggregate() && "Not an aggregate!");
-    auto align = reinterpret_cast<uintptr_t>(V2.getPointer()) >> AggAlignShift;
-    return Address(
-        V1.getPointer(), ElementType, CharUnits::fromQuantity(align));
+    return AggregateAddr;
   }
-  llvm::Value *getAggregatePointer() const {
-    assert(isAggregate() && "Not an aggregate!");
-    return V1.getPointer();
+
+  llvm::Value *getAggregatePointer(QualType PointeeType,
+                                   CodeGenFunction &CGF) const {
+    return getAggregateAddress().getBasePointer();
   }
 
   static RValue getIgnored() {
@@ -88,17 +96,19 @@ class RValue {
 
   static RValue get(llvm::Value *V) {
     RValue ER;
-    ER.V1.setPointer(V);
-    ER.V1.setInt(Scalar);
-    ER.V2.setInt(false);
+    ER.Vals.first = V;
+    ER.Flavor = Scalar;
+    ER.IsVolatile = false;
     return ER;
   }
+  static RValue get(Address Addr, CodeGenFunction &CGF) {
+    return RValue::get(Addr.emitRawPointer(CGF));
+  }
   static RValue getComplex(llvm::Value *V1, llvm::Value *V2) {
     RValue ER;
-    ER.V1.setPointer(V1);
-    ER.V2.setPointer(V2);
-    ER.V1.setInt(Complex);
-    ER.V2.setInt(false);
+    ER.Vals = {V1, V2};
+    ER.Flavor = Complex;
+    ER.IsVolatile = false;
     return ER;
   }
   static RValue getComplex(const std::pair<llvm::Value *, llvm::Value *> &C) {
@@ -107,15 +117,15 @@ class RValue {
   // FIXME: Aggregate rvalues need to retain information about whether they are
   // volatile or not.  Remove default to find all places that probably get this
   // wrong.
+
+  /// Convert an Address to an RValue. If the Address is not
+  /// signed, create an RValue using the unsigned address. Otherwise, resign the
+  /// address using the provided type.
   static RValue getAggregate(Address addr, bool isVolatile = false) {
     RValue ER;
-    ER.V1.setPointer(addr.getPointer());
-    ER.V1.setInt(Aggregate);
-    ER.ElementType = addr.getElementType();
-
-    auto align = static_cast<uintptr_t>(addr.getAlignment().getQuantity());
-    ER.V2.setPointer(reinterpret_cast<llvm::Value*>(align << AggAlignShift));
-    ER.V2.setInt(isVolatile);
+    ER.AggregateAddr = addr;
+    ER.Flavor = Aggregate;
+    ER.IsVolatile = isVolatile;
     return ER;
   }
 };
@@ -178,8 +188,10 @@ class LValue {
     MatrixElt     // This is a matrix element, use getVector*
   } LVType;
 
-  llvm::Value *V;
-  llvm::Type *ElementType;
+  union {
+    Address Addr = Address::invalid();
+    llvm::Value *V;
+  };
 
   union {
     // Index into a vector subscript: V[i]
@@ -197,10 +209,6 @@ class LValue {
   // 'const' is unused here
   Qualifiers Quals;
 
-  // The alignment to use when accessing this lvalue.  (For vector elements,
-  // this is the alignment of the whole vector.)
-  unsigned Alignment;
-
   // objective-c's ivar
   bool Ivar:1;
 
@@ -234,23 +242,19 @@ class LValue {
   Expr *BaseIvarExp;
 
 private:
-  void Initialize(QualType Type, Qualifiers Quals, CharUnits Alignment,
+  void Initialize(QualType Type, Qualifiers Quals, Address Addr,
                   LValueBaseInfo BaseInfo, TBAAAccessInfo TBAAInfo) {
-    assert((!Alignment.isZero() || Type->isIncompleteType()) &&
-           "initializing l-value with zero alignment!");
-    if (isGlobalReg())
-      assert(ElementType == nullptr && "Global reg does not store elem type");
-    else
-      assert(ElementType != nullptr && "Must have elem type");
-
     this->Type = Type;
     this->Quals = Quals;
     const unsigned MaxAlign = 1U << 31;
-    this->Alignment = Alignment.getQuantity() <= MaxAlign
-                          ? Alignment.getQuantity()
-                          : MaxAlign;
-    assert(this->Alignment == Alignment.getQuantity() &&
-           "Alignment exceeds allowed max!");
+    CharUnits Alignment = Addr.getAlignment();
+    assert((isGlobalReg() || !Alignment.isZero() || Type->isIncompleteType()) &&
+           "initializing l-value with zero alignment!");
+    if (Alignment.getQuantity() > MaxAlign) {
+      assert(false && "Alignment exceeds allowed max!");
+      Alignment = CharUnits::fromQuantity(MaxAlign);
+    }
+    this->Addr = Addr;
     this->BaseInfo = BaseInfo;
     this->TBAAInfo = TBAAInfo;
 
@@ -259,9 +263,20 @@ class LValue {
     this->ImpreciseLifetime = false;
     this->Nontemporal = false;
     this->ThreadLocalRef = false;
+    this->IsKnownNonNull = false;
     this->BaseIvarExp = nullptr;
   }
 
+  void initializeSimpleLValue(Address Addr, QualType Type,
+                              LValueBaseInfo BaseInfo, TBAAAccessInfo TBAAInfo,
+                              ASTContext &Context) {
+    Qualifiers QS = Type.getQualifiers();
+    QS.setObjCGCAttr(Context.getObjCGCAttrKind(Type));
+    LVType = Simple;
+    Initialize(Type, QS, Addr, BaseInfo, TBAAInfo);
+    assert(Addr.getBasePointer()->getType()->isPointerTy());
+  }
+
 public:
   bool isSimple() const { return LVType == Simple; }
   bool isVectorElt() const { return LVType == VectorElt; }
@@ -328,8 +343,8 @@ class LValue {
 
   LangAS getAddressSpace() const { return Quals.getAddressSpace(); }
 
-  CharUnits getAlignment() const { return CharUnits::fromQuantity(Alignment); }
-  void setAlignment(CharUnits A) { Alignment = A.getQuantity(); }
+  CharUnits getAlignment() const { return Addr.getAlignment(); }
+  void setAlignment(CharUnits A) { Addr.setAlignment(A); }
 
   LValueBaseInfo getBaseInfo() const { return BaseInfo; }
   void setBaseInfo(LValueBaseInfo Info) { BaseInfo = Info; }
@@ -345,28 +360,32 @@ class LValue {
   // simple lvalue
   llvm::Value *getPointer(CodeGenFunction &CGF) const {
     assert(isSimple());
-    return V;
+    return Addr.getBasePointer();
   }
-  Address getAddress(CodeGenFunction &CGF) const {
-    return Address(getPointer(CGF), ElementType, getAlignment(),
-                   isKnownNonNull());
-  }
-  void setAddress(Address address) {
+  llvm::Value *emitRawPointer(CodeGenFunction &CGF) const {
     assert(isSimple());
-    V = address.getPointer();
-    ElementType = address.getElementType();
-    Alignment = address.getAlignment().getQuantity();
-    IsKnownNonNull = address.isKnownNonNull();
+    return Addr.isValid() ? Addr.emitRawPointer(CGF) : nullptr;
   }
 
+  Address getAddress(CodeGenFunction &CGF) const {
+    // FIXME: remove parameter.
+    return Addr;
+  }
+
+  void setAddress(Address address) { Addr = address; }
+
   // vector elt lvalue
   Address getVectorAddress() const {
-    return Address(getVectorPointer(), ElementType, getAlignment(),
-                   (KnownNonNull_t)isKnownNonNull());
+    assert(isVectorElt());
+    return Addr;
+  }
+  llvm::Value *getRawVectorPointer(CodeGenFunction &CGF) const {
+    assert(isVectorElt());
+    return Addr.emitRawPointer(CGF);
   }
   llvm::Value *getVectorPointer() const {
     assert(isVectorElt());
-    return V;
+    return Addr.getBasePointer();
   }
   llvm::Value *getVectorIdx() const {
     assert(isVectorElt());
@@ -374,12 +393,12 @@ class LValue {
   }
 
   Address getMatrixAddress() const {
-    return Address(getMatrixPointer(), ElementType, getAlignment(),
-                   (KnownNonNull_t)isKnownNonNull());
+    assert(isMatrixElt());
+    return Addr;
   }
   llvm::Value *getMatrixPointer() const {
     assert(isMatrixElt());
-    return V;
+    return Addr.getBasePointer();
   }
   llvm::Value *getMatrixIdx() const {
     assert(isMatrixElt());
@@ -388,12 +407,12 @@ class LValue {
 
   // extended vector elements.
   Address getExtVectorAddress() const {
-    return Address(getExtVectorPointer(), ElementType, getAlignment(),
-                   (KnownNonNull_t)isKnownNonNull());
+    assert(isExtVectorElt());
+    return Addr;
   }
-  llvm::Value *getExtVectorPointer() const {
+  llvm::Value *getRawExtVectorPointer(CodeGenFunction &CGF) const {
     assert(isExtVectorElt());
-    return V;
+    return Addr.emitRawPointer(CGF);
   }
   llvm::Constant *getExtVectorElts() const {
     assert(isExtVectorElt());
@@ -402,10 +421,14 @@ class LValue {
 
   // bitfield lvalue
   Address getBitFieldAddress() const {
-    return Address(getBitFieldPointer(), ElementType, getAlignment(),
-                   (KnownNonNull_t)isKnownNonNull());
+    assert(isBitField());
+    return Addr;
+  }
+  llvm::Value *getRawBitFieldPointer(CodeGenFunction &CGF) const {
+    assert(isBitField());
+    return Addr.emitRawPointer(CGF);
   }
-  llvm::Value *getBitFieldPointer() const { assert(isBitField()); return V; }
+
   const CGBitFieldInfo &getBitFieldInfo() const {
     assert(isBitField());
     return *BitFieldInfo;
@@ -414,18 +437,13 @@ class LValue {
   // global register lvalue
   llvm::Value *getGlobalReg() const { assert(isGlobalReg()); return V; }
 
-  static LValue MakeAddr(Address address, QualType type, ASTContext &Context,
+  static LValue MakeAddr(Address Addr, QualType type, ASTContext &Context,
                          LValueBaseInfo BaseInfo, TBAAAccessInfo TBAAInfo) {
-    Qualifiers qs = type.getQualifiers();
-    qs.setObjCGCAttr(Context.getObjCGCAttrKind(type));
-
     LValue R;
     R.LVType = Simple;
-    assert(address.getPointer()->getType()->isPointerTy());
-    R.V = address.getPointer();
-    R.ElementType = address.getElementType();
-    R.IsKnownNonNull = address.isKnownNonNull();
-    R.Initialize(type, qs, address.getAlignment(), BaseInfo, TBAAInfo);
+    R.initializeSimpleLValue(Addr, type, BaseInfo, TBAAInfo, Context);
+    R.Addr = Addr;
+    assert(Addr.getType()->isPointerTy());
     return R;
   }
 
@@ -434,26 +452,18 @@ class LValue {
                               TBAAAccessInfo TBAAInfo) {
     LValue R;
     R.LVType = VectorElt;
-    R.V = vecAddress.getPointer();
-    R.ElementType = vecAddress.getElementType();
     R.VectorIdx = Idx;
-    R.IsKnownNonNull = vecAddress.isKnownNonNull();
-    R.Initialize(type, type.getQualifiers(), vecAddress.getAlignment(),
-                 BaseInfo, TBAAInfo);
+    R.Initialize(type, type.getQualifiers(), vecAddress, BaseInfo, TBAAInfo);
     return R;
   }
 
-  static LValue MakeExtVectorElt(Address vecAddress, llvm::Constant *Elts,
+  static LValue MakeExtVectorElt(Address Addr, llvm::Constant *Elts,
                                  QualType type, LValueBaseInfo BaseInfo,
                                  TBAAAccessInfo TBAAInfo) {
     LValue R;
     R.LVType = ExtVectorElt;
-    R.V = vecAddress.getPointer();
-    R.ElementType = vecAddress.getElementType();
     R.VectorElts = Elts;
-    R.IsKnownNonNull = vecAddress.isKnownNonNull();
-    R.Initialize(type, type.getQualifiers(), vecAddress.getAlignment(),
-                 BaseInfo, TBAAInfo);
+    R.Initialize(type, type.getQualifiers(), Addr, BaseInfo, TBAAInfo);
     return R;
   }
 
@@ -468,12 +478,8 @@ class LValue {
                              TBAAAccessInfo TBAAInfo) {
     LValue R;
     R.LVType = BitField;
-    R.V = Addr.getPointer();
-    R.ElementType = Addr.getElementType();
     R.BitFieldInfo = &Info;
-    R.IsKnownNonNull = Addr.isKnownNonNull();
-    R.Initialize(type, type.getQualifiers(), Addr.getAlignment(), BaseInfo,
-                 TBAAInfo);
+    R.Initialize(type, type.getQualifiers(), Addr, BaseInfo, TBAAInfo);
     return R;
   }
 
@@ -481,11 +487,9 @@ class LValue {
                               QualType type) {
     LValue R;
     R.LVType = GlobalReg;
-    R.V = V;
-    R.ElementType = nullptr;
-    R.IsKnownNonNull = true;
-    R.Initialize(type, type.getQualifiers(), alignment,
+    R.Initialize(type, type.getQualifiers(), Address::invalid(),
                  LValueBaseInfo(AlignmentSource::Decl), TBAAAccessInfo());
+    R.V = V;
     return R;
   }
 
@@ -494,12 +498,8 @@ class LValue {
                               TBAAAccessInfo TBAAInfo) {
     LValue R;
     R.LVType = MatrixElt;
-    R.V = matAddress.getPointer();
-    R.ElementType = matAddress.getElementType();
     R.VectorIdx = Idx;
-    R.IsKnownNonNull = matAddress.isKnownNonNull();
-    R.Initialize(type, type.getQualifiers(), matAddress.getAlignment(),
-                 BaseInfo, TBAAInfo);
+    R.Initialize(type, type.getQualifiers(), matAddress, BaseInfo, TBAAInfo);
     return R;
   }
 
@@ -643,17 +643,17 @@ class AggValueSlot {
     return NeedsGCBarriers_t(ObjCGCFlag);
   }
 
-  llvm::Value *getPointer() const {
-    return Addr.getPointer();
+  llvm::Value *getPointer(QualType PointeeTy, CodeGenFunction &CGF) const;
+
+  llvm::Value *emitRawPointer(CodeGenFunction &CGF) const {
+    return Addr.isValid() ? Addr.emitRawPointer(CGF) : nullptr;
   }
 
   Address getAddress() const {
     return Addr;
   }
 
-  bool isIgnored() const {
-    return !Addr.isValid();
-  }
+  bool isIgnored() const { return !Addr.isValid(); }
 
   CharUnits getAlignment() const {
     return Addr.getAlignment();
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index f1e2f794a604a..34fb22550040b 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -195,26 +195,35 @@ CodeGenFunction::CGFPOptionsRAII::~CGFPOptionsRAII() {
   CGF.Builder.setDefaultConstrainedRounding(OldRounding);
 }
 
-LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) {
+static LValue MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T,
+                                         bool ForPointeeType,
+                                         CodeGenFunction &CGF) {
   LValueBaseInfo BaseInfo;
   TBAAAccessInfo TBAAInfo;
-  CharUnits Alignment = CGM.getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo);
-  Address Addr(V, ConvertTypeForMem(T), Alignment);
-  return LValue::MakeAddr(Addr, T, getContext(), BaseInfo, TBAAInfo);
+  CharUnits Alignment =
+      CGF.CGM.getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo, ForPointeeType);
+  Address Addr = Address(V, CGF.ConvertTypeForMem(T), Alignment);
+  return CGF.MakeAddrLValue(Addr, T, BaseInfo, TBAAInfo);
+}
+
+LValue CodeGenFunction::MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T) {
+  return ::MakeNaturalAlignAddrLValue(V, T, /*ForPointeeType*/ false, *this);
 }
 
-/// Given a value of type T* that may not be to a complete object,
-/// construct an l-value with the natural pointee alignment of T.
 LValue
 CodeGenFunction::MakeNaturalAlignPointeeAddrLValue(llvm::Value *V, QualType T) {
-  LValueBaseInfo BaseInfo;
-  TBAAAccessInfo TBAAInfo;
-  CharUnits Align = CGM.getNaturalTypeAlignment(T, &BaseInfo, &TBAAInfo,
-                                                /* forPointeeType= */ true);
-  Address Addr(V, ConvertTypeForMem(T), Align);
-  return MakeAddrLValue(Addr, T, BaseInfo, TBAAInfo);
+  return ::MakeNaturalAlignAddrLValue(V, T, /*ForPointeeType*/ true, *this);
 }
 
+LValue CodeGenFunction::MakeNaturalAlignRawAddrLValue(llvm::Value *V,
+                                                      QualType T) {
+  return ::MakeNaturalAlignAddrLValue(V, T, /*ForPointeeType*/ false, *this);
+}
+
+LValue CodeGenFunction::MakeNaturalAlignPointeeRawAddrLValue(llvm::Value *V,
+                                                             QualType T) {
+  return ::MakeNaturalAlignAddrLValue(V, T, /*ForPointeeType*/ true, *this);
+}
 
 llvm::Type *CodeGenFunction::ConvertTypeForMem(QualType T) {
   return CGM.getTypes().ConvertTypeForMem(T);
@@ -528,7 +537,8 @@ void CodeGenFunction::FinishFunction(SourceLocation EndLoc) {
     ReturnBlock.getBlock()->eraseFromParent();
   }
   if (ReturnValue.isValid()) {
-    auto *RetAlloca = dyn_cast<llvm::AllocaInst>(ReturnValue.getPointer());
+    auto *RetAlloca =
+        dyn_cast<llvm::AllocaInst>(ReturnValue.emitRawPointer(*this));
     if (RetAlloca && RetAlloca->use_empty()) {
       RetAlloca->eraseFromParent();
       ReturnValue = Address::invalid();
@@ -1363,13 +1373,14 @@ void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
     auto AI = CurFn->arg_begin();
     if (CurFnInfo->getReturnInfo().isSRetAfterThis())
       ++AI;
-    ReturnValue =
-        Address(&*AI, ConvertType(RetTy),
-                CurFnInfo->getReturnInfo().getIndirectAlign(), KnownNonNull);
+    ReturnValue = makeNaturalAddressForPointer(
+        &*AI, RetTy, CurFnInfo->getReturnInfo().getIndirectAlign(), false,
+        nullptr, nullptr, KnownNonNull);
     if (!CurFnInfo->getReturnInfo().getIndirectByVal()) {
-      ReturnValuePointer = CreateDefaultAlignTempAlloca(
-          ReturnValue.getPointer()->getType(), "result.ptr");
-      Builder.CreateStore(ReturnValue.getPointer(), ReturnValuePointer);
+      ReturnValuePointer =
+          CreateDefaultAlignTempAlloca(ReturnValue.getType(), "result.ptr");
+      Builder.CreateStore(ReturnValue.emitRawPointer(*this),
+                          ReturnValuePointer);
     }
   } else if (CurFnInfo->getReturnInfo().getKind() == ABIArgInfo::InAlloca &&
              !hasScalarEvaluationKind(CurFnInfo->getReturnType())) {
@@ -1430,8 +1441,9 @@ void CodeGenFunction::StartFunction(GlobalDecl GD, QualType RetTy,
         // or contains the address of the enclosing object).
         LValue ThisFieldLValue = EmitLValueForLambdaField(LambdaThisCaptureField);
         if (!LambdaThisCaptureField->getType()->isPointerType()) {
-          // If the enclosing object was captured by value, just use its address.
-          CXXThisValue = ThisFieldLValue.getAddress(*this).getPointer();
+          // If the enclosing object was captured by value, just use its
+          // address. Sign this pointer.
+          CXXThisValue = ThisFieldLValue.getPointer(*this);
         } else {
           // Load the lvalue pointed to by the field, since '*this' was captured
           // by reference.
@@ -2293,8 +2305,9 @@ static void emitNonZeroVLAInit(CodeGenFunction &CGF, QualType baseType,
     = llvm::ConstantInt::get(CGF.IntPtrTy, baseSize.getQuantity());
 
   Address begin = dest.withElementType(CGF.Int8Ty);
-  llvm::Value *end = Builder.CreateInBoundsGEP(
-      begin.getElementType(), begin.getPointer(), sizeInChars, "vla.end");
+  llvm::Value *end = Builder.CreateInBoundsGEP(begin.getElementType(),
+                                               begin.emitRawPointer(CGF),
+                                               sizeInChars, "vla.end");
 
   llvm::BasicBlock *originBB = CGF.Builder.GetInsertBlock();
   llvm::BasicBlock *loopBB = CGF.createBasicBlock("vla-init.loop");
@@ -2305,7 +2318,7 @@ static void emitNonZeroVLAInit(CodeGenFunction &CGF, QualType baseType,
   CGF.EmitBlock(loopBB);
 
   llvm::PHINode *cur = Builder.CreatePHI(begin.getType(), 2, "vla.cur");
-  cur->addIncoming(begin.getPointer(), originBB);
+  cur->addIncoming(begin.emitRawPointer(CGF), originBB);
 
   CharUnits curAlign =
     dest.getAlignment().alignmentOfArrayElement(baseSize);
@@ -2498,10 +2511,10 @@ llvm::Value *CodeGenFunction::emitArrayLength(const ArrayType *origArrayType,
     addr = addr.withElementType(baseType);
   } else {
     // Create the actual GEP.
-    addr = Address(Builder.CreateInBoundsGEP(
-        addr.getElementType(), addr.getPointer(), gepIndices, "array.begin"),
-        ConvertTypeForMem(eltType),
-        addr.getAlignment());
+    addr = Address(Builder.CreateInBoundsGEP(addr.getElementType(),
+                                             addr.emitRawPointer(*this),
+                                             gepIndices, "array.begin"),
+                   ConvertTypeForMem(eltType), addr.getAlignment());
   }
 
   baseType = eltType;
@@ -2850,7 +2863,7 @@ void CodeGenFunction::EmitVarAnnotations(const VarDecl *D, llvm::Value *V) {
 Address CodeGenFunction::EmitFieldAnnotations(const FieldDecl *D,
                                               Address Addr) {
   assert(D->hasAttr<AnnotateAttr>() && "no annotate attribute");
-  llvm::Value *V = Addr.getPointer();
+  llvm::Value *V = Addr.emitRawPointer(*this);
   llvm::Type *VTy = V->getType();
   auto *PTy = dyn_cast<llvm::PointerType>(VTy);
   unsigned AS = PTy ? PTy->getAddressSpace() : 0;
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index ecd1202437a37..652cbee655647 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -151,6 +151,9 @@ struct DominatingLLVMValue {
 
   /// Answer whether the given value needs extra work to be saved.
   static bool needsSaving(llvm::Value *value) {
+    if (!value)
+      return false;
+
     // If it's not an instruction, we don't need to save.
     if (!isa<llvm::Instruction>(value)) return false;
 
@@ -177,21 +180,28 @@ template <> struct DominatingValue<Address> {
   typedef Address type;
 
   struct saved_type {
-    DominatingLLVMValue::saved_type SavedValue;
+    DominatingLLVMValue::saved_type BasePtr;
     llvm::Type *ElementType;
     CharUnits Alignment;
+    DominatingLLVMValue::saved_type Offset;
+    llvm::PointerType *EffectiveType;
   };
 
   static bool needsSaving(type value) {
-    return DominatingLLVMValue::needsSaving(value.getPointer());
+    if (DominatingLLVMValue::needsSaving(value.getBasePointer()) ||
+        DominatingLLVMValue::needsSaving(value.getOffset()))
+      return true;
+    return false;
   }
   static saved_type save(CodeGenFunction &CGF, type value) {
-    return { DominatingLLVMValue::save(CGF, value.getPointer()),
-             value.getElementType(), value.getAlignment() };
+    return {DominatingLLVMValue::save(CGF, value.getBasePointer()),
+            value.getElementType(), value.getAlignment(),
+            DominatingLLVMValue::save(CGF, value.getOffset()), value.getType()};
   }
   static type restore(CodeGenFunction &CGF, saved_type value) {
-    return Address(DominatingLLVMValue::restore(CGF, value.SavedValue),
-                   value.ElementType, value.Alignment);
+    return Address(DominatingLLVMValue::restore(CGF, value.BasePtr),
+                   value.ElementType, value.Alignment,
+                   DominatingLLVMValue::restore(CGF, value.Offset));
   }
 };
 
@@ -201,14 +211,26 @@ template <> struct DominatingValue<RValue> {
   class saved_type {
     enum Kind { ScalarLiteral, ScalarAddress, AggregateLiteral,
                 AggregateAddress, ComplexAddress };
-
-    llvm::Value *Value;
-    llvm::Type *ElementType;
+    union {
+      struct {
+        DominatingLLVMValue::saved_type first, second;
+      } Vals;
+      DominatingValue<Address>::saved_type AggregateAddr;
+    };
     LLVM_PREFERRED_TYPE(Kind)
     unsigned K : 3;
-    unsigned Align : 29;
-    saved_type(llvm::Value *v, llvm::Type *e, Kind k, unsigned a = 0)
-      : Value(v), ElementType(e), K(k), Align(a) {}
+    unsigned IsVolatile : 1;
+
+    saved_type(DominatingLLVMValue::saved_type Val1, unsigned K)
+        : Vals{Val1, DominatingLLVMValue::saved_type()}, K(K) {}
+
+    saved_type(DominatingLLVMValue::saved_type Val1,
+               DominatingLLVMValue::saved_type Val2)
+        : Vals{Val1, Val2}, K(ComplexAddress) {}
+
+    saved_type(DominatingValue<Address>::saved_type AggregateAddr,
+               bool IsVolatile, unsigned K)
+        : AggregateAddr(AggregateAddr), K(K) {}
 
   public:
     static bool needsSaving(RValue value);
@@ -659,7 +681,7 @@ class CodeGenFunction : public CodeGenTypeCache {
     llvm::Value *Size;
 
   public:
-    CallLifetimeEnd(Address addr, llvm::Value *size)
+    CallLifetimeEnd(RawAddress addr, llvm::Value *size)
         : Addr(addr.getPointer()), Size(size) {}
 
     void Emit(CodeGenFunction &CGF, Flags flags) override {
@@ -684,7 +706,7 @@ class CodeGenFunction : public CodeGenTypeCache {
   };
 
   /// i32s containing the indexes of the cleanup destinations.
-  Address NormalCleanupDest = Address::invalid();
+  RawAddress NormalCleanupDest = RawAddress::invalid();
 
   unsigned NextCleanupDestIndex = 1;
 
@@ -819,10 +841,10 @@ class CodeGenFunction : public CodeGenTypeCache {
   template <class T, class... As>
   void pushCleanupAfterFullExpr(CleanupKind Kind, As... A) {
     if (!isInConditionalBranch())
-      return pushCleanupAfterFullExprWithActiveFlag<T>(Kind, Address::invalid(),
-                                                       A...);
+      return pushCleanupAfterFullExprWithActiveFlag<T>(
+          Kind, RawAddress::invalid(), A...);
 
-    Address ActiveFlag = createCleanupActiveFlag();
+    RawAddress ActiveFlag = createCleanupActiveFlag();
     assert(!DominatingValue<Address>::needsSaving(ActiveFlag) &&
            "cleanup active flag should never need saving");
 
@@ -835,7 +857,7 @@ class CodeGenFunction : public CodeGenTypeCache {
 
   template <class T, class... As>
   void pushCleanupAfterFullExprWithActiveFlag(CleanupKind Kind,
-                                              Address ActiveFlag, As... A) {
+                                              RawAddress ActiveFlag, As... A) {
     LifetimeExtendedCleanupHeader Header = {sizeof(T), Kind,
                                             ActiveFlag.isValid()};
 
@@ -850,7 +872,7 @@ class CodeGenFunction : public CodeGenTypeCache {
     new (Buffer) LifetimeExtendedCleanupHeader(Header);
     new (Buffer + sizeof(Header)) T(A...);
     if (Header.IsConditional)
-      new (Buffer + sizeof(Header) + sizeof(T)) Address(ActiveFlag);
+      new (Buffer + sizeof(Header) + sizeof(T)) RawAddress(ActiveFlag);
   }
 
   /// Set up the last cleanup that was pushed as a conditional
@@ -859,8 +881,8 @@ class CodeGenFunction : public CodeGenTypeCache {
     initFullExprCleanupWithFlag(createCleanupActiveFlag());
   }
 
-  void initFullExprCleanupWithFlag(Address ActiveFlag);
-  Address createCleanupActiveFlag();
+  void initFullExprCleanupWithFlag(RawAddress ActiveFlag);
+  RawAddress createCleanupActiveFlag();
 
   /// PushDestructorCleanup - Push a cleanup to call the
   /// complete-object destructor of an object of the given type at the
@@ -1048,7 +1070,7 @@ class CodeGenFunction : public CodeGenTypeCache {
       QualType VarTy = LocalVD->getType();
       if (VarTy->isReferenceType()) {
         Address Temp = CGF.CreateMemTemp(VarTy);
-        CGF.Builder.CreateStore(TempAddr.getPointer(), Temp);
+        CGF.Builder.CreateStore(TempAddr.emitRawPointer(CGF), Temp);
         TempAddr = Temp;
       }
       SavedTempAddresses.try_emplace(LocalVD, TempAddr);
@@ -1243,10 +1265,12 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// one branch or the other of a conditional expression.
   bool isInConditionalBranch() const { return OutermostConditional != nullptr; }
 
-  void setBeforeOutermostConditional(llvm::Value *value, Address addr) {
+  void setBeforeOutermostConditional(llvm::Value *value, Address addr,
+                                     CodeGenFunction &CGF) {
     assert(isInConditionalBranch());
     llvm::BasicBlock *block = OutermostConditional->getStartingBlock();
-    auto store = new llvm::StoreInst(value, addr.getPointer(), &block->back());
+    auto store =
+        new llvm::StoreInst(value, addr.emitRawPointer(CGF), &block->back());
     store->setAlignment(addr.getAlignment().getAsAlign());
   }
 
@@ -1601,7 +1625,7 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// If \p StepV is null, the default increment is 1.
   void maybeUpdateMCDCTestVectorBitmap(const Expr *E) {
     if (isMCDCCoverageEnabled() && isBinaryLogicalOp(E)) {
-      PGO.emitMCDCTestVectorBitmapUpdate(Builder, E, MCDCCondBitmapAddr);
+      PGO.emitMCDCTestVectorBitmapUpdate(Builder, E, MCDCCondBitmapAddr, *this);
       PGO.setCurrentStmt(E);
     }
   }
@@ -1609,7 +1633,7 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// Update the MCDC temp value with the condition's evaluated result.
   void maybeUpdateMCDCCondBitmap(const Expr *E, llvm::Value *Val) {
     if (isMCDCCoverageEnabled()) {
-      PGO.emitMCDCCondBitmapUpdate(Builder, E, MCDCCondBitmapAddr, Val);
+      PGO.emitMCDCCondBitmapUpdate(Builder, E, MCDCCondBitmapAddr, Val, *this);
       PGO.setCurrentStmt(E);
     }
   }
@@ -1726,7 +1750,7 @@ class CodeGenFunction : public CodeGenTypeCache {
         : CGF(CGF), OldCXXThisValue(CGF.CXXThisValue),
           OldCXXThisAlignment(CGF.CXXThisAlignment),
           SourceLocScope(E, CGF.CurSourceLocExprScope) {
-      CGF.CXXThisValue = CGF.CXXDefaultInitExprThis.getPointer();
+      CGF.CXXThisValue = CGF.CXXDefaultInitExprThis.getBasePointer();
       CGF.CXXThisAlignment = CGF.CXXDefaultInitExprThis.getAlignment();
     }
     ~CXXDefaultInitExprScope() {
@@ -2112,7 +2136,7 @@ class CodeGenFunction : public CodeGenTypeCache {
   llvm::Value *getExceptionFromSlot();
   llvm::Value *getSelectorFromSlot();
 
-  Address getNormalCleanupDestSlot();
+  RawAddress getNormalCleanupDestSlot();
 
   llvm::BasicBlock *getUnreachableBlock() {
     if (!UnreachableBlock) {
@@ -2601,10 +2625,40 @@ class CodeGenFunction : public CodeGenTypeCache {
   //                                  Helpers
   //===--------------------------------------------------------------------===//
 
+  Address mergeAddressesInConditionalExpr(Address LHS, Address RHS,
+                                          llvm::BasicBlock *LHSBlock,
+                                          llvm::BasicBlock *RHSBlock,
+                                          llvm::BasicBlock *MergeBlock,
+                                          QualType MergedType) {
+    Builder.SetInsertPoint(MergeBlock);
+    llvm::PHINode *PtrPhi = Builder.CreatePHI(LHS.getType(), 2, "cond");
+    PtrPhi->addIncoming(LHS.getBasePointer(), LHSBlock);
+    PtrPhi->addIncoming(RHS.getBasePointer(), RHSBlock);
+    LHS.replaceBasePointer(PtrPhi);
+    LHS.setAlignment(std::min(LHS.getAlignment(), RHS.getAlignment()));
+    return LHS;
+  }
+
+  /// Construct an address with the natural alignment of T. If a pointer to T
+  /// is expected to be signed, the pointer passed to this function must have
+  /// been signed, and the returned Address will have the pointer authentication
+  /// information needed to authenticate the signed pointer.
+  Address makeNaturalAddressForPointer(
+      llvm::Value *Ptr, QualType T, CharUnits Alignment = CharUnits::Zero(),
+      bool ForPointeeType = false, LValueBaseInfo *BaseInfo = nullptr,
+      TBAAAccessInfo *TBAAInfo = nullptr,
+      KnownNonNull_t IsKnownNonNull = NotKnownNonNull) {
+    if (Alignment.isZero())
+      Alignment =
+          CGM.getNaturalTypeAlignment(T, BaseInfo, TBAAInfo, ForPointeeType);
+    return Address(Ptr, ConvertTypeForMem(T), Alignment, nullptr,
+                   IsKnownNonNull);
+  }
+
   LValue MakeAddrLValue(Address Addr, QualType T,
                         AlignmentSource Source = AlignmentSource::Type) {
-    return LValue::MakeAddr(Addr, T, getContext(), LValueBaseInfo(Source),
-                            CGM.getTBAAAccessInfo(T));
+    return MakeAddrLValue(Addr, T, LValueBaseInfo(Source),
+                          CGM.getTBAAAccessInfo(T));
   }
 
   LValue MakeAddrLValue(Address Addr, QualType T, LValueBaseInfo BaseInfo,
@@ -2614,6 +2668,14 @@ class CodeGenFunction : public CodeGenTypeCache {
 
   LValue MakeAddrLValue(llvm::Value *V, QualType T, CharUnits Alignment,
                         AlignmentSource Source = AlignmentSource::Type) {
+    return MakeAddrLValue(makeNaturalAddressForPointer(V, T, Alignment), T,
+                          LValueBaseInfo(Source), CGM.getTBAAAccessInfo(T));
+  }
+
+  /// Same as MakeAddrLValue above except that the pointer is known to be
+  /// unsigned.
+  LValue MakeRawAddrLValue(llvm::Value *V, QualType T, CharUnits Alignment,
+                           AlignmentSource Source = AlignmentSource::Type) {
     Address Addr(V, ConvertTypeForMem(T), Alignment);
     return LValue::MakeAddr(Addr, T, getContext(), LValueBaseInfo(Source),
                             CGM.getTBAAAccessInfo(T));
@@ -2626,9 +2688,18 @@ class CodeGenFunction : public CodeGenTypeCache {
                             TBAAAccessInfo());
   }
 
+  /// Given a value of type T* that may not be to a complete object, construct
+  /// an l-value with the natural pointee alignment of T.
   LValue MakeNaturalAlignPointeeAddrLValue(llvm::Value *V, QualType T);
+
   LValue MakeNaturalAlignAddrLValue(llvm::Value *V, QualType T);
 
+  /// Same as MakeNaturalAlignPointeeAddrLValue except that the pointer is known
+  /// to be unsigned.
+  LValue MakeNaturalAlignPointeeRawAddrLValue(llvm::Value *V, QualType T);
+
+  LValue MakeNaturalAlignRawAddrLValue(llvm::Value *V, QualType T);
+
   Address EmitLoadOfReference(LValue RefLVal,
                               LValueBaseInfo *PointeeBaseInfo = nullptr,
                               TBAAAccessInfo *PointeeTBAAInfo = nullptr);
@@ -2677,13 +2748,13 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// more efficient if the caller knows that the address will not be exposed.
   llvm::AllocaInst *CreateTempAlloca(llvm::Type *Ty, const Twine &Name = "tmp",
                                      llvm::Value *ArraySize = nullptr);
-  Address CreateTempAlloca(llvm::Type *Ty, CharUnits align,
-                           const Twine &Name = "tmp",
-                           llvm::Value *ArraySize = nullptr,
-                           Address *Alloca = nullptr);
-  Address CreateTempAllocaWithoutCast(llvm::Type *Ty, CharUnits align,
-                                      const Twine &Name = "tmp",
-                                      llvm::Value *ArraySize = nullptr);
+  RawAddress CreateTempAlloca(llvm::Type *Ty, CharUnits align,
+                              const Twine &Name = "tmp",
+                              llvm::Value *ArraySize = nullptr,
+                              RawAddress *Alloca = nullptr);
+  RawAddress CreateTempAllocaWithoutCast(llvm::Type *Ty, CharUnits align,
+                                         const Twine &Name = "tmp",
+                                         llvm::Value *ArraySize = nullptr);
 
   /// CreateDefaultAlignedTempAlloca - This creates an alloca with the
   /// default ABI alignment of the given LLVM type.
@@ -2695,8 +2766,8 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// not hand this address off to arbitrary IRGen routines, and especially
   /// do not pass it as an argument to a function that might expect a
   /// properly ABI-aligned value.
-  Address CreateDefaultAlignTempAlloca(llvm::Type *Ty,
-                                       const Twine &Name = "tmp");
+  RawAddress CreateDefaultAlignTempAlloca(llvm::Type *Ty,
+                                          const Twine &Name = "tmp");
 
   /// CreateIRTemp - Create a temporary IR object of the given type, with
   /// appropriate alignment. This routine should only be used when an temporary
@@ -2706,32 +2777,31 @@ class CodeGenFunction : public CodeGenTypeCache {
   ///
   /// That is, this is exactly equivalent to CreateMemTemp, but calling
   /// ConvertType instead of ConvertTypeForMem.
-  Address CreateIRTemp(QualType T, const Twine &Name = "tmp");
+  RawAddress CreateIRTemp(QualType T, const Twine &Name = "tmp");
 
   /// CreateMemTemp - Create a temporary memory object of the given type, with
   /// appropriate alignmen and cast it to the default address space. Returns
   /// the original alloca instruction by \p Alloca if it is not nullptr.
-  Address CreateMemTemp(QualType T, const Twine &Name = "tmp",
-                        Address *Alloca = nullptr);
-  Address CreateMemTemp(QualType T, CharUnits Align, const Twine &Name = "tmp",
-                        Address *Alloca = nullptr);
+  RawAddress CreateMemTemp(QualType T, const Twine &Name = "tmp",
+                           RawAddress *Alloca = nullptr);
+  RawAddress CreateMemTemp(QualType T, CharUnits Align,
+                           const Twine &Name = "tmp",
+                           RawAddress *Alloca = nullptr);
 
   /// CreateMemTemp - Create a temporary memory object of the given type, with
   /// appropriate alignmen without casting it to the default address space.
-  Address CreateMemTempWithoutCast(QualType T, const Twine &Name = "tmp");
-  Address CreateMemTempWithoutCast(QualType T, CharUnits Align,
-                                   const Twine &Name = "tmp");
+  RawAddress CreateMemTempWithoutCast(QualType T, const Twine &Name = "tmp");
+  RawAddress CreateMemTempWithoutCast(QualType T, CharUnits Align,
+                                      const Twine &Name = "tmp");
 
   /// CreateAggTemp - Create a temporary memory object for the given
   /// aggregate type.
   AggValueSlot CreateAggTemp(QualType T, const Twine &Name = "tmp",
-                             Address *Alloca = nullptr) {
-    return AggValueSlot::forAddr(CreateMemTemp(T, Name, Alloca),
-                                 T.getQualifiers(),
-                                 AggValueSlot::IsNotDestructed,
-                                 AggValueSlot::DoesNotNeedGCBarriers,
-                                 AggValueSlot::IsNotAliased,
-                                 AggValueSlot::DoesNotOverlap);
+                             RawAddress *Alloca = nullptr) {
+    return AggValueSlot::forAddr(
+        CreateMemTemp(T, Name, Alloca), T.getQualifiers(),
+        AggValueSlot::IsNotDestructed, AggValueSlot::DoesNotNeedGCBarriers,
+        AggValueSlot::IsNotAliased, AggValueSlot::DoesNotOverlap);
   }
 
   /// EvaluateExprAsBool - Perform the usual unary conversions on the specified
@@ -3120,6 +3190,25 @@ class CodeGenFunction : public CodeGenTypeCache {
   /// calls to EmitTypeCheck can be skipped.
   bool sanitizePerformTypeCheck() const;
 
+  void EmitTypeCheck(TypeCheckKind TCK, SourceLocation Loc, LValue LV,
+                     QualType Type, SanitizerSet SkippedChecks = SanitizerSet(),
+                     llvm::Value *ArraySize = nullptr) {
+    if (!sanitizePerformTypeCheck())
+      return;
+    EmitTypeCheck(TCK, Loc, LV.emitRawPointer(*this), Type, LV.getAlignment(),
+                  SkippedChecks, ArraySize);
+  }
+
+  void EmitTypeCheck(TypeCheckKind TCK, SourceLocation Loc, Address Addr,
+                     QualType Type, CharUnits Alignment = CharUnits::Zero(),
+                     SanitizerSet SkippedChecks = SanitizerSet(),
+                     llvm::Value *ArraySize = nullptr) {
+    if (!sanitizePerformTypeCheck())
+      return;
+    EmitTypeCheck(TCK, Loc, Addr.emitRawPointer(*this), Type, Alignment,
+                  SkippedChecks, ArraySize);
+  }
+
   /// Emit a check that \p V is the address of storage of the
   /// appropriate size and alignment for an object of type \p Type
   /// (or if ArraySize is provided, for an array of that bound).
@@ -3220,17 +3309,17 @@ class CodeGenFunction : public CodeGenTypeCache {
 
     /// Address with original alloca instruction. Invalid if the variable was
     /// emitted as a global constant.
-    Address AllocaAddr;
+    RawAddress AllocaAddr;
 
     struct Invalid {};
     AutoVarEmission(Invalid)
         : Variable(nullptr), Addr(Address::invalid()),
-          AllocaAddr(Address::invalid()) {}
+          AllocaAddr(RawAddress::invalid()) {}
 
     AutoVarEmission(const VarDecl &variable)
         : Variable(&variable), Addr(Address::invalid()), NRVOFlag(nullptr),
           IsEscapingByRef(false), IsConstantAggregate(false),
-          SizeForLifetimeMarkers(nullptr), AllocaAddr(Address::invalid()) {}
+          SizeForLifetimeMarkers(nullptr), AllocaAddr(RawAddress::invalid()) {}
 
     bool wasEmittedAsGlobal() const { return !Addr.isValid(); }
 
@@ -3253,7 +3342,7 @@ class CodeGenFunction : public CodeGenTypeCache {
     }
 
     /// Returns the address for the original alloca instruction.
-    Address getOriginalAllocatedAddress() const { return AllocaAddr; }
+    RawAddress getOriginalAllocatedAddress() const { return AllocaAddr; }
 
     /// Returns the address of the object within this declaration.
     /// Note that this does not chase the forwarding pointer for
@@ -3283,23 +3372,32 @@ class CodeGenFunction : public CodeGenTypeCache {
                          llvm::GlobalValue::LinkageTypes Linkage);
 
   class ParamValue {
-    llvm::Value *Value;
-    llvm::Type *ElementType;
-    unsigned Alignment;
-    ParamValue(llvm::Value *V, llvm::Type *T, unsigned A)
-        : Value(V), ElementType(T), Alignment(A) {}
+    union {
+      Address Addr;
+      llvm::Value *Value;
+    };
+
+    bool IsIndirect;
+
+    ParamValue(llvm::Value *V) : Value(V), IsIndirect(false) {}
+    ParamValue(Address A) : Addr(A), IsIndirect(true) {}
+
   public:
     static ParamValue forDirect(llvm::Value *value) {
-      return ParamValue(value, nullptr, 0);
+      return ParamValue(value);
     }
     static ParamValue forIndirect(Address addr) {
       assert(!addr.getAlignment().isZero());
-      return ParamValue(addr.getPointer(), addr.getElementType(),
-                        addr.getAlignment().getQuantity());
+      return ParamValue(addr);
     }
 
-    bool isIndirect() const { return Alignment != 0; }
-    llvm::Value *getAnyValue() const { return Value; }
+    bool isIndirect() const { return IsIndirect; }
+    llvm::Value *getAnyValue() const {
+      if (!isIndirect())
+        return Value;
+      assert(!Addr.hasOffset() && "unexpected offset");
+      return Addr.getBasePointer();
+    }
 
     llvm::Value *getDirectValue() const {
       assert(!isIndirect());
@@ -3308,8 +3406,7 @@ class CodeGenFunction : public CodeGenTypeCache {
 
     Address getIndirectAddress() const {
       assert(isIndirect());
-      return Address(Value, ElementType, CharUnits::fromQuantity(Alignment),
-                     KnownNonNull);
+      return Addr;
     }
   };
 
@@ -4219,6 +4316,9 @@ class CodeGenFunction : public CodeGenTypeCache {
                                   const Twine &name = "");
   llvm::CallInst *EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
                                           const Twine &name = "");
+  llvm::CallInst *EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
+                                          ArrayRef<Address> args,
+                                          const Twine &name = "");
   llvm::CallInst *EmitNounwindRuntimeCall(llvm::FunctionCallee callee,
                                           ArrayRef<llvm::Value *> args,
                                           const Twine &name = "");
@@ -4245,6 +4345,12 @@ class CodeGenFunction : public CodeGenTypeCache {
                                                CXXDtorType Type,
                                                const CXXRecordDecl *RD);
 
+  llvm::Value *getAsNaturalPointerTo(Address Addr, QualType PointeeType) {
+    return Addr.getBasePointer();
+  }
+
+  bool isPointerKnownNonNull(const Expr *E);
+
   // Return the copy constructor name with the prefix "__copy_constructor_"
   // removed.
   static std::string getNonTrivialCopyConstructorStr(QualType QT,
@@ -4853,6 +4959,11 @@ class CodeGenFunction : public CodeGenTypeCache {
                                       SourceLocation Loc,
                                       const Twine &Name = "");
 
+  Address EmitCheckedInBoundsGEP(Address Addr, ArrayRef<llvm::Value *> IdxList,
+                                 llvm::Type *elementType, bool SignedIndices,
+                                 bool IsSubtraction, SourceLocation Loc,
+                                 CharUnits Align, const Twine &Name = "");
+
   /// Specifies which type of sanitizer check to apply when handling a
   /// particular builtin.
   enum BuiltinCheckKind {
@@ -4915,6 +5026,10 @@ class CodeGenFunction : public CodeGenTypeCache {
   void EmitNonNullArgCheck(RValue RV, QualType ArgType, SourceLocation ArgLoc,
                            AbstractCallee AC, unsigned ParmNum);
 
+  void EmitNonNullArgCheck(Address Addr, QualType ArgType,
+                           SourceLocation ArgLoc, AbstractCallee AC,
+                           unsigned ParmNum);
+
   /// EmitCallArg - Emit a single call argument.
   void EmitCallArg(CallArgList &args, const Expr *E, QualType ArgType);
 
@@ -5142,7 +5257,7 @@ DominatingLLVMValue::save(CodeGenFunction &CGF, llvm::Value *value) {
       CGF.CreateTempAlloca(value->getType(), align, "cond-cleanup.save");
   CGF.Builder.CreateStore(value, alloca);
 
-  return saved_type(alloca.getPointer(), true);
+  return saved_type(alloca.emitRawPointer(CGF), true);
 }
 
 inline llvm::Value *DominatingLLVMValue::restore(CodeGenFunction &CGF,
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index ac28f569ab87e..6c96217b444e0 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -7926,7 +7926,7 @@ void CodeGenFunction::EmitDeclMetadata() {
 
   for (auto &I : LocalDeclMap) {
     const Decl *D = I.first;
-    llvm::Value *Addr = I.second.getPointer();
+    llvm::Value *Addr = I.second.emitRawPointer(*this);
     if (auto *Alloca = dyn_cast<llvm::AllocaInst>(Addr)) {
       llvm::Value *DAddr = GetPointerConstant(getLLVMContext(), D);
       Alloca->setMetadata(
diff --git a/clang/lib/CodeGen/CodeGenPGO.cpp b/clang/lib/CodeGen/CodeGenPGO.cpp
index 2619edfeb7dc7..76704c4d7be4a 100644
--- a/clang/lib/CodeGen/CodeGenPGO.cpp
+++ b/clang/lib/CodeGen/CodeGenPGO.cpp
@@ -1239,7 +1239,8 @@ void CodeGenPGO::emitMCDCParameters(CGBuilderTy &Builder) {
 
 void CodeGenPGO::emitMCDCTestVectorBitmapUpdate(CGBuilderTy &Builder,
                                                 const Expr *S,
-                                                Address MCDCCondBitmapAddr) {
+                                                Address MCDCCondBitmapAddr,
+                                                CodeGenFunction &CGF) {
   if (!canEmitMCDCCoverage(Builder) || !RegionMCDCState)
     return;
 
@@ -1262,7 +1263,7 @@ void CodeGenPGO::emitMCDCTestVectorBitmapUpdate(CGBuilderTy &Builder,
                           Builder.getInt64(FunctionHash),
                           Builder.getInt32(RegionMCDCState->BitmapBytes),
                           Builder.getInt32(MCDCTestVectorBitmapOffset),
-                          MCDCCondBitmapAddr.getPointer()};
+                          MCDCCondBitmapAddr.emitRawPointer(CGF)};
   Builder.CreateCall(
       CGM.getIntrinsic(llvm::Intrinsic::instrprof_mcdc_tvbitmap_update), Args);
 }
@@ -1283,7 +1284,8 @@ void CodeGenPGO::emitMCDCCondBitmapReset(CGBuilderTy &Builder, const Expr *S,
 
 void CodeGenPGO::emitMCDCCondBitmapUpdate(CGBuilderTy &Builder, const Expr *S,
                                           Address MCDCCondBitmapAddr,
-                                          llvm::Value *Val) {
+                                          llvm::Value *Val,
+                                          CodeGenFunction &CGF) {
   if (!canEmitMCDCCoverage(Builder) || !RegionMCDCState)
     return;
 
@@ -1312,7 +1314,7 @@ void CodeGenPGO::emitMCDCCondBitmapUpdate(CGBuilderTy &Builder, const Expr *S,
   llvm::Value *Args[5] = {llvm::ConstantExpr::getBitCast(FuncNameVar, I8PtrTy),
                           Builder.getInt64(FunctionHash),
                           Builder.getInt32(Branch.ID),
-                          MCDCCondBitmapAddr.getPointer(), Val};
+                          MCDCCondBitmapAddr.emitRawPointer(CGF), Val};
   Builder.CreateCall(
       CGM.getIntrinsic(llvm::Intrinsic::instrprof_mcdc_condbitmap_update),
       Args);
diff --git a/clang/lib/CodeGen/CodeGenPGO.h b/clang/lib/CodeGen/CodeGenPGO.h
index 036fbf6815a49..9d66ffad6f435 100644
--- a/clang/lib/CodeGen/CodeGenPGO.h
+++ b/clang/lib/CodeGen/CodeGenPGO.h
@@ -113,12 +113,14 @@ class CodeGenPGO {
   void emitCounterSetOrIncrement(CGBuilderTy &Builder, const Stmt *S,
                                  llvm::Value *StepV);
   void emitMCDCTestVectorBitmapUpdate(CGBuilderTy &Builder, const Expr *S,
-                                      Address MCDCCondBitmapAddr);
+                                      Address MCDCCondBitmapAddr,
+                                      CodeGenFunction &CGF);
   void emitMCDCParameters(CGBuilderTy &Builder);
   void emitMCDCCondBitmapReset(CGBuilderTy &Builder, const Expr *S,
                                Address MCDCCondBitmapAddr);
   void emitMCDCCondBitmapUpdate(CGBuilderTy &Builder, const Expr *S,
-                                Address MCDCCondBitmapAddr, llvm::Value *Val);
+                                Address MCDCCondBitmapAddr, llvm::Value *Val,
+                                CodeGenFunction &CGF);
 
   /// Return the region count for the counter at the given index.
   uint64_t getRegionCount(const Stmt *S) {
diff --git a/clang/lib/CodeGen/ItaniumCXXABI.cpp b/clang/lib/CodeGen/ItaniumCXXABI.cpp
index 27ceb23b942c4..207b12806546c 100644
--- a/clang/lib/CodeGen/ItaniumCXXABI.cpp
+++ b/clang/lib/CodeGen/ItaniumCXXABI.cpp
@@ -308,10 +308,6 @@ class ItaniumCXXABI : public CodeGen::CGCXXABI {
       CodeGenFunction &CGF, const CXXRecordDecl *VTableClass,
       BaseSubobject Base, const CXXRecordDecl *NearestVBase);
 
-  llvm::Constant *
-  getVTableAddressPointForConstExpr(BaseSubobject Base,
-                                    const CXXRecordDecl *VTableClass) override;
-
   llvm::GlobalVariable *getAddrOfVTable(const CXXRecordDecl *RD,
                                         CharUnits VPtrOffset) override;
 
@@ -647,7 +643,7 @@ CGCallee ItaniumCXXABI::EmitLoadOfMemberFunctionPointer(
 
   // Apply the adjustment and cast back to the original struct type
   // for consistency.
-  llvm::Value *This = ThisAddr.getPointer();
+  llvm::Value *This = ThisAddr.emitRawPointer(CGF);
   This = Builder.CreateInBoundsGEP(Builder.getInt8Ty(), This, Adj);
   ThisPtrForCall = This;
 
@@ -853,7 +849,7 @@ llvm::Value *ItaniumCXXABI::EmitMemberDataPointerAddress(
   CGBuilderTy &Builder = CGF.Builder;
 
   // Apply the offset, which we assume is non-null.
-  return Builder.CreateInBoundsGEP(CGF.Int8Ty, Base.getPointer(), MemPtr,
+  return Builder.CreateInBoundsGEP(CGF.Int8Ty, Base.emitRawPointer(CGF), MemPtr,
                                    "memptr.offset");
 }
 
@@ -1249,7 +1245,7 @@ void ItaniumCXXABI::emitVirtualObjectDelete(CodeGenFunction &CGF,
                                                         CGF.getPointerAlign());
 
     // Apply the offset.
-    llvm::Value *CompletePtr = Ptr.getPointer();
+    llvm::Value *CompletePtr = Ptr.emitRawPointer(CGF);
     CompletePtr =
         CGF.Builder.CreateInBoundsGEP(CGF.Int8Ty, CompletePtr, Offset);
 
@@ -1486,7 +1482,8 @@ llvm::Value *ItaniumCXXABI::emitDynamicCastCall(
       computeOffsetHint(CGF.getContext(), SrcDecl, DestDecl).getQuantity());
 
   // Emit the call to __dynamic_cast.
-  llvm::Value *Args[] = {ThisAddr.getPointer(), SrcRTTI, DestRTTI, OffsetHint};
+  llvm::Value *Args[] = {ThisAddr.emitRawPointer(CGF), SrcRTTI, DestRTTI,
+                         OffsetHint};
   llvm::Value *Value =
       CGF.EmitNounwindRuntimeCall(getItaniumDynamicCastFn(CGF), Args);
 
@@ -1575,7 +1572,7 @@ llvm::Value *ItaniumCXXABI::emitExactDynamicCast(
       VPtr, CGM.getTBAAVTablePtrAccessInfo(CGF.VoidPtrPtrTy));
   llvm::Value *Success = CGF.Builder.CreateICmpEQ(
       VPtr, getVTableAddressPoint(BaseSubobject(SrcDecl, *Offset), DestDecl));
-  llvm::Value *Result = ThisAddr.getPointer();
+  llvm::Value *Result = ThisAddr.emitRawPointer(CGF);
   if (!Offset->isZero())
     Result = CGF.Builder.CreateInBoundsGEP(
         CGF.CharTy, Result,
@@ -1617,7 +1614,7 @@ llvm::Value *ItaniumCXXABI::emitDynamicCastToVoid(CodeGenFunction &CGF,
         PtrDiffLTy, OffsetToTop, CGF.getPointerAlign(), "offset.to.top");
   }
   // Finally, add the offset to the pointer.
-  return CGF.Builder.CreateInBoundsGEP(CGF.Int8Ty, ThisAddr.getPointer(),
+  return CGF.Builder.CreateInBoundsGEP(CGF.Int8Ty, ThisAddr.emitRawPointer(CGF),
                                        OffsetToTop);
 }
 
@@ -1798,8 +1795,8 @@ void ItaniumCXXABI::EmitDestructorCall(CodeGenFunction &CGF,
   else
     Callee = CGCallee::forDirect(CGM.getAddrOfCXXStructor(GD), GD);
 
-  CGF.EmitCXXDestructorCall(GD, Callee, This.getPointer(), ThisTy, VTT, VTTTy,
-                            nullptr);
+  CGF.EmitCXXDestructorCall(GD, Callee, CGF.getAsNaturalPointerTo(This, ThisTy),
+                            ThisTy, VTT, VTTTy, nullptr);
 }
 
 void ItaniumCXXABI::emitVTableDefinitions(CodeGenVTables &CGVT,
@@ -1958,11 +1955,6 @@ llvm::Value *ItaniumCXXABI::getVTableAddressPointInStructorWithVTT(
                                        CGF.getPointerAlign());
 }
 
-llvm::Constant *ItaniumCXXABI::getVTableAddressPointForConstExpr(
-    BaseSubobject Base, const CXXRecordDecl *VTableClass) {
-  return getVTableAddressPoint(Base, VTableClass);
-}
-
 llvm::GlobalVariable *ItaniumCXXABI::getAddrOfVTable(const CXXRecordDecl *RD,
                                                      CharUnits VPtrOffset) {
   assert(VPtrOffset.isZero() && "Itanium ABI only supports zero vptr offsets");
@@ -2093,8 +2085,8 @@ llvm::Value *ItaniumCXXABI::EmitVirtualDestructorCall(
     ThisTy = D->getDestroyedType();
   }
 
-  CGF.EmitCXXDestructorCall(GD, Callee, This.getPointer(), ThisTy, nullptr,
-                            QualType(), nullptr);
+  CGF.EmitCXXDestructorCall(GD, Callee, This.emitRawPointer(CGF), ThisTy,
+                            nullptr, QualType(), nullptr);
   return nullptr;
 }
 
@@ -2167,7 +2159,7 @@ static llvm::Value *performTypeAdjustment(CodeGenFunction &CGF,
                                           int64_t VirtualAdjustment,
                                           bool IsReturnAdjustment) {
   if (!NonVirtualAdjustment && !VirtualAdjustment)
-    return InitialPtr.getPointer();
+    return InitialPtr.emitRawPointer(CGF);
 
   Address V = InitialPtr.withElementType(CGF.Int8Ty);
 
@@ -2200,10 +2192,10 @@ static llvm::Value *performTypeAdjustment(CodeGenFunction &CGF,
                                              CGF.getPointerAlign());
     }
     // Adjust our pointer.
-    ResultPtr = CGF.Builder.CreateInBoundsGEP(
-        V.getElementType(), V.getPointer(), Offset);
+    ResultPtr = CGF.Builder.CreateInBoundsGEP(V.getElementType(),
+                                              V.emitRawPointer(CGF), Offset);
   } else {
-    ResultPtr = V.getPointer();
+    ResultPtr = V.emitRawPointer(CGF);
   }
 
   // In a derived-to-base conversion, the non-virtual adjustment is
@@ -2289,7 +2281,7 @@ Address ItaniumCXXABI::InitializeArrayCookie(CodeGenFunction &CGF,
         llvm::FunctionType::get(CGM.VoidTy, NumElementsPtr.getType(), false);
     llvm::FunctionCallee F =
         CGM.CreateRuntimeFunction(FTy, "__asan_poison_cxx_array_cookie");
-    CGF.Builder.CreateCall(F, NumElementsPtr.getPointer());
+    CGF.Builder.CreateCall(F, NumElementsPtr.emitRawPointer(CGF));
   }
 
   // Finally, compute a pointer to the actual data buffer by skipping
@@ -2320,7 +2312,7 @@ llvm::Value *ItaniumCXXABI::readArrayCookieImpl(CodeGenFunction &CGF,
       CGF.SizeTy, llvm::PointerType::getUnqual(CGF.getLLVMContext()), false);
   llvm::FunctionCallee F =
       CGM.CreateRuntimeFunction(FTy, "__asan_load_cxx_array_cookie");
-  return CGF.Builder.CreateCall(F, numElementsPtr.getPointer());
+  return CGF.Builder.CreateCall(F, numElementsPtr.emitRawPointer(CGF));
 }
 
 CharUnits ARMCXXABI::getArrayCookieSizeImpl(QualType elementType) {
@@ -2632,7 +2624,7 @@ void ItaniumCXXABI::EmitGuardedInit(CodeGenFunction &CGF,
 
     // Call __cxa_guard_release.  This cannot throw.
     CGF.EmitNounwindRuntimeCall(getGuardReleaseFn(CGM, guardPtrTy),
-                                guardAddr.getPointer());
+                                guardAddr.emitRawPointer(CGF));
   } else if (D.isLocalVarDecl()) {
     // For local variables, store 1 into the first byte of the guard variable
     // after the object initialization completes so that initialization is
@@ -3125,10 +3117,10 @@ LValue ItaniumCXXABI::EmitThreadLocalVarDeclLValue(CodeGenFunction &CGF,
 
   LValue LV;
   if (VD->getType()->isReferenceType())
-    LV = CGF.MakeNaturalAlignAddrLValue(CallVal, LValType);
+    LV = CGF.MakeNaturalAlignRawAddrLValue(CallVal, LValType);
   else
-    LV = CGF.MakeAddrLValue(CallVal, LValType,
-                            CGF.getContext().getDeclAlign(VD));
+    LV = CGF.MakeRawAddrLValue(CallVal, LValType,
+                               CGF.getContext().getDeclAlign(VD));
   // FIXME: need setObjCGCLValueClass?
   return LV;
 }
@@ -4618,7 +4610,7 @@ static void InitCatchParam(CodeGenFunction &CGF,
         CGF.Builder.CreateStore(Casted, ExnPtrTmp);
 
         // Bind the reference to the temporary.
-        AdjustedExn = ExnPtrTmp.getPointer();
+        AdjustedExn = ExnPtrTmp.emitRawPointer(CGF);
       }
     }
 
diff --git a/clang/lib/CodeGen/MicrosoftCXXABI.cpp b/clang/lib/CodeGen/MicrosoftCXXABI.cpp
index ddbac3ff49671..ca00324949f09 100644
--- a/clang/lib/CodeGen/MicrosoftCXXABI.cpp
+++ b/clang/lib/CodeGen/MicrosoftCXXABI.cpp
@@ -327,10 +327,6 @@ class MicrosoftCXXABI : public CGCXXABI {
       CodeGenFunction &CGF, const CXXRecordDecl *VTableClass,
       BaseSubobject Base, const CXXRecordDecl *NearestVBase) override;
 
-  llvm::Constant *
-  getVTableAddressPointForConstExpr(BaseSubobject Base,
-                                    const CXXRecordDecl *VTableClass) override;
-
   llvm::GlobalVariable *getAddrOfVTable(const CXXRecordDecl *RD,
                                         CharUnits VPtrOffset) override;
 
@@ -937,7 +933,7 @@ void MicrosoftCXXABI::emitBeginCatch(CodeGenFunction &CGF,
   }
 
   CodeGenFunction::AutoVarEmission var = CGF.EmitAutoVarAlloca(*CatchParam);
-  CPI->setArgOperand(2, var.getObjectAddress(CGF).getPointer());
+  CPI->setArgOperand(2, var.getObjectAddress(CGF).emitRawPointer(CGF));
   CGF.EHStack.pushCleanup<CatchRetScope>(NormalCleanup, CPI);
   CGF.EmitAutoVarCleanups(var);
 }
@@ -974,7 +970,7 @@ MicrosoftCXXABI::performBaseAdjustment(CodeGenFunction &CGF, Address Value,
   llvm::Value *Offset =
     GetVirtualBaseClassOffset(CGF, Value, SrcDecl, PolymorphicBase);
   llvm::Value *Ptr = CGF.Builder.CreateInBoundsGEP(
-      Value.getElementType(), Value.getPointer(), Offset);
+      Value.getElementType(), Value.emitRawPointer(CGF), Offset);
   CharUnits VBaseAlign =
     CGF.CGM.getVBaseAlignment(Value.getAlignment(), SrcDecl, PolymorphicBase);
   return std::make_tuple(Address(Ptr, CGF.Int8Ty, VBaseAlign), Offset,
@@ -1011,7 +1007,7 @@ llvm::Value *MicrosoftCXXABI::EmitTypeid(CodeGenFunction &CGF,
                                          llvm::Type *StdTypeInfoPtrTy) {
   std::tie(ThisPtr, std::ignore, std::ignore) =
       performBaseAdjustment(CGF, ThisPtr, SrcRecordTy);
-  llvm::CallBase *Typeid = emitRTtypeidCall(CGF, ThisPtr.getPointer());
+  llvm::CallBase *Typeid = emitRTtypeidCall(CGF, ThisPtr.emitRawPointer(CGF));
   return CGF.Builder.CreateBitCast(Typeid, StdTypeInfoPtrTy);
 }
 
@@ -1034,7 +1030,7 @@ llvm::Value *MicrosoftCXXABI::emitDynamicCastCall(
   llvm::Value *Offset;
   std::tie(This, Offset, std::ignore) =
       performBaseAdjustment(CGF, This, SrcRecordTy);
-  llvm::Value *ThisPtr = This.getPointer();
+  llvm::Value *ThisPtr = This.emitRawPointer(CGF);
   Offset = CGF.Builder.CreateTrunc(Offset, CGF.Int32Ty);
 
   // PVOID __RTDynamicCast(
@@ -1066,7 +1062,7 @@ llvm::Value *MicrosoftCXXABI::emitDynamicCastToVoid(CodeGenFunction &CGF,
   llvm::FunctionCallee Function = CGF.CGM.CreateRuntimeFunction(
       llvm::FunctionType::get(CGF.Int8PtrTy, ArgTypes, false),
       "__RTCastToVoid");
-  llvm::Value *Args[] = {Value.getPointer()};
+  llvm::Value *Args[] = {Value.emitRawPointer(CGF)};
   return CGF.EmitRuntimeCall(Function, Args);
 }
 
@@ -1494,7 +1490,7 @@ Address MicrosoftCXXABI::adjustThisArgumentForVirtualFunctionCall(
     llvm::Value *VBaseOffset =
       GetVirtualBaseClassOffset(CGF, Result, Derived, VBase);
     llvm::Value *VBasePtr = CGF.Builder.CreateInBoundsGEP(
-        Result.getElementType(), Result.getPointer(), VBaseOffset);
+        Result.getElementType(), Result.emitRawPointer(CGF), VBaseOffset);
     CharUnits VBaseAlign =
       CGF.CGM.getVBaseAlignment(Result.getAlignment(), Derived, VBase);
     Result = Address(VBasePtr, CGF.Int8Ty, VBaseAlign);
@@ -1661,7 +1657,8 @@ void MicrosoftCXXABI::EmitDestructorCall(CodeGenFunction &CGF,
   llvm::Value *Implicit =
       getCXXDestructorImplicitParam(CGF, DD, Type, ForVirtualBase,
                                     Delegating); // = nullptr
-  CGF.EmitCXXDestructorCall(GD, Callee, This.getPointer(), ThisTy,
+  CGF.EmitCXXDestructorCall(GD, Callee, CGF.getAsNaturalPointerTo(This, ThisTy),
+                            ThisTy,
                             /*ImplicitParam=*/Implicit,
                             /*ImplicitParamTy=*/QualType(), nullptr);
   if (BaseDtorEndBB) {
@@ -1792,13 +1789,6 @@ MicrosoftCXXABI::getVTableAddressPoint(BaseSubobject Base,
   return VFTablesMap[ID];
 }
 
-llvm::Constant *MicrosoftCXXABI::getVTableAddressPointForConstExpr(
-    BaseSubobject Base, const CXXRecordDecl *VTableClass) {
-  llvm::Constant *VFTable = getVTableAddressPoint(Base, VTableClass);
-  assert(VFTable && "Couldn't find a vftable for the given base?");
-  return VFTable;
-}
-
 llvm::GlobalVariable *MicrosoftCXXABI::getAddrOfVTable(const CXXRecordDecl *RD,
                                                        CharUnits VPtrOffset) {
   // getAddrOfVTable may return 0 if asked to get an address of a vtable which
@@ -2014,8 +2004,9 @@ llvm::Value *MicrosoftCXXABI::EmitVirtualDestructorCall(
   }
 
   This = adjustThisArgumentForVirtualFunctionCall(CGF, GD, This, true);
-  RValue RV = CGF.EmitCXXDestructorCall(GD, Callee, This.getPointer(), ThisTy,
-                                        ImplicitParam, Context.IntTy, CE);
+  RValue RV =
+      CGF.EmitCXXDestructorCall(GD, Callee, This.emitRawPointer(CGF), ThisTy,
+                                ImplicitParam, Context.IntTy, CE);
   return RV.getScalarVal();
 }
 
@@ -2213,13 +2204,13 @@ llvm::Value *MicrosoftCXXABI::performThisAdjustment(CodeGenFunction &CGF,
                                                     Address This,
                                                     const ThisAdjustment &TA) {
   if (TA.isEmpty())
-    return This.getPointer();
+    return This.emitRawPointer(CGF);
 
   This = This.withElementType(CGF.Int8Ty);
 
   llvm::Value *V;
   if (TA.Virtual.isEmpty()) {
-    V = This.getPointer();
+    V = This.emitRawPointer(CGF);
   } else {
     assert(TA.Virtual.Microsoft.VtordispOffset < 0);
     // Adjust the this argument based on the vtordisp value.
@@ -2228,7 +2219,7 @@ llvm::Value *MicrosoftCXXABI::performThisAdjustment(CodeGenFunction &CGF,
                  CharUnits::fromQuantity(TA.Virtual.Microsoft.VtordispOffset));
     VtorDispPtr = VtorDispPtr.withElementType(CGF.Int32Ty);
     llvm::Value *VtorDisp = CGF.Builder.CreateLoad(VtorDispPtr, "vtordisp");
-    V = CGF.Builder.CreateGEP(This.getElementType(), This.getPointer(),
+    V = CGF.Builder.CreateGEP(This.getElementType(), This.emitRawPointer(CGF),
                               CGF.Builder.CreateNeg(VtorDisp));
 
     // Unfortunately, having applied the vtordisp means that we no
@@ -2265,11 +2256,11 @@ llvm::Value *
 MicrosoftCXXABI::performReturnAdjustment(CodeGenFunction &CGF, Address Ret,
                                          const ReturnAdjustment &RA) {
   if (RA.isEmpty())
-    return Ret.getPointer();
+    return Ret.emitRawPointer(CGF);
 
   Ret = Ret.withElementType(CGF.Int8Ty);
 
-  llvm::Value *V = Ret.getPointer();
+  llvm::Value *V = Ret.emitRawPointer(CGF);
   if (RA.Virtual.Microsoft.VBIndex) {
     assert(RA.Virtual.Microsoft.VBIndex > 0);
     int32_t IntSize = CGF.getIntSize().getQuantity();
@@ -2584,7 +2575,7 @@ struct ResetGuardBit final : EHScopeStack::Cleanup {
 
 struct CallInitThreadAbort final : EHScopeStack::Cleanup {
   llvm::Value *Guard;
-  CallInitThreadAbort(Address Guard) : Guard(Guard.getPointer()) {}
+  CallInitThreadAbort(RawAddress Guard) : Guard(Guard.getPointer()) {}
 
   void Emit(CodeGenFunction &CGF, Flags flags) override {
     // Calling _Init_thread_abort will reset the guard's state.
@@ -3124,8 +3115,8 @@ MicrosoftCXXABI::GetVBaseOffsetFromVBPtr(CodeGenFunction &CGF,
                                          llvm::Value **VBPtrOut) {
   CGBuilderTy &Builder = CGF.Builder;
   // Load the vbtable pointer from the vbptr in the instance.
-  llvm::Value *VBPtr = Builder.CreateInBoundsGEP(CGM.Int8Ty, This.getPointer(),
-                                                 VBPtrOffset, "vbptr");
+  llvm::Value *VBPtr = Builder.CreateInBoundsGEP(
+      CGM.Int8Ty, This.emitRawPointer(CGF), VBPtrOffset, "vbptr");
   if (VBPtrOut)
     *VBPtrOut = VBPtr;
 
@@ -3204,7 +3195,7 @@ llvm::Value *MicrosoftCXXABI::AdjustVirtualBase(
     Builder.CreateBr(SkipAdjustBB);
     CGF.EmitBlock(SkipAdjustBB);
     llvm::PHINode *Phi = Builder.CreatePHI(CGM.Int8PtrTy, 2, "memptr.base");
-    Phi->addIncoming(Base.getPointer(), OriginalBB);
+    Phi->addIncoming(Base.emitRawPointer(CGF), OriginalBB);
     Phi->addIncoming(AdjustedBase, VBaseAdjustBB);
     return Phi;
   }
@@ -3239,7 +3230,7 @@ llvm::Value *MicrosoftCXXABI::EmitMemberDataPointerAddress(
     Addr = AdjustVirtualBase(CGF, E, RD, Base, VirtualBaseAdjustmentOffset,
                              VBPtrOffset);
   } else {
-    Addr = Base.getPointer();
+    Addr = Base.emitRawPointer(CGF);
   }
 
   // Apply the offset, which we assume is non-null.
@@ -3527,7 +3518,7 @@ CGCallee MicrosoftCXXABI::EmitLoadOfMemberFunctionPointer(
     ThisPtrForCall = AdjustVirtualBase(CGF, E, RD, This,
                                    VirtualBaseAdjustmentOffset, VBPtrOffset);
   } else {
-    ThisPtrForCall = This.getPointer();
+    ThisPtrForCall = This.emitRawPointer(CGF);
   }
 
   if (NonVirtualBaseAdjustment)
@@ -4446,10 +4437,7 @@ void MicrosoftCXXABI::emitThrow(CodeGenFunction &CGF, const CXXThrowExpr *E) {
   llvm::GlobalVariable *TI = getThrowInfo(ThrowType);
 
   // Call into the runtime to throw the exception.
-  llvm::Value *Args[] = {
-    AI.getPointer(),
-    TI
-  };
+  llvm::Value *Args[] = {AI.emitRawPointer(CGF), TI};
   CGF.EmitNoreturnRuntimeCallOrInvoke(getThrowFn(), Args);
 }
 
diff --git a/clang/lib/CodeGen/TargetInfo.h b/clang/lib/CodeGen/TargetInfo.h
index 6893b50a3cfe9..b1dfe5bf8f274 100644
--- a/clang/lib/CodeGen/TargetInfo.h
+++ b/clang/lib/CodeGen/TargetInfo.h
@@ -295,6 +295,11 @@ class TargetCodeGenInfo {
   /// Get the AST address space for alloca.
   virtual LangAS getASTAllocaAddressSpace() const { return LangAS::Default; }
 
+  Address performAddrSpaceCast(CodeGen::CodeGenFunction &CGF, Address Addr,
+                               LangAS SrcAddr, LangAS DestAddr,
+                               llvm::Type *DestTy,
+                               bool IsNonNull = false) const;
+
   /// Perform address space cast of an expression of pointer type.
   /// \param V is the LLVM value to be casted to another address space.
   /// \param SrcAddr is the language address space of \p V.
diff --git a/clang/lib/CodeGen/Targets/NVPTX.cpp b/clang/lib/CodeGen/Targets/NVPTX.cpp
index 946d70dbe84ef..a814be9e20d8c 100644
--- a/clang/lib/CodeGen/Targets/NVPTX.cpp
+++ b/clang/lib/CodeGen/Targets/NVPTX.cpp
@@ -85,7 +85,7 @@ class NVPTXTargetCodeGenInfo : public TargetCodeGenInfo {
                                            LValue Src) {
     llvm::Value *Handle = nullptr;
     llvm::Constant *C =
-        llvm::dyn_cast<llvm::Constant>(Src.getAddress(CGF).getPointer());
+        llvm::dyn_cast<llvm::Constant>(Src.getAddress(CGF).emitRawPointer(CGF));
     // Lookup `addrspacecast` through the constant pointer if any.
     if (auto *ASC = llvm::dyn_cast_or_null<llvm::AddrSpaceCastOperator>(C))
       C = llvm::cast<llvm::Constant>(ASC->getPointerOperand());
diff --git a/clang/lib/CodeGen/Targets/PPC.cpp b/clang/lib/CodeGen/Targets/PPC.cpp
index 3eadb19bd2058..174fddabbbdb6 100644
--- a/clang/lib/CodeGen/Targets/PPC.cpp
+++ b/clang/lib/CodeGen/Targets/PPC.cpp
@@ -513,9 +513,10 @@ Address PPC32_SVR4_ABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAList,
     CharUnits RegSize = CharUnits::fromQuantity((isInt || IsSoftFloatABI) ? 4 : 8);
     llvm::Value *RegOffset =
         Builder.CreateMul(NumRegs, Builder.getInt8(RegSize.getQuantity()));
-    RegAddr = Address(
-        Builder.CreateInBoundsGEP(CGF.Int8Ty, RegAddr.getPointer(), RegOffset),
-        DirectTy, RegAddr.getAlignment().alignmentOfArrayElement(RegSize));
+    RegAddr = Address(Builder.CreateInBoundsGEP(
+                          CGF.Int8Ty, RegAddr.emitRawPointer(CGF), RegOffset),
+                      DirectTy,
+                      RegAddr.getAlignment().alignmentOfArrayElement(RegSize));
 
     // Increase the used-register count.
     NumRegs =
@@ -551,7 +552,7 @@ Address PPC32_SVR4_ABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAList,
     // Round up address of argument to alignment
     CharUnits Align = CGF.getContext().getTypeAlignInChars(Ty);
     if (Align > OverflowAreaAlign) {
-      llvm::Value *Ptr = OverflowArea.getPointer();
+      llvm::Value *Ptr = OverflowArea.emitRawPointer(CGF);
       OverflowArea = Address(emitRoundPointerUpToAlignment(CGF, Ptr, Align),
                              OverflowArea.getElementType(), Align);
     }
@@ -560,7 +561,7 @@ Address PPC32_SVR4_ABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAList,
 
     // Increase the overflow area.
     OverflowArea = Builder.CreateConstInBoundsByteGEP(OverflowArea, Size);
-    Builder.CreateStore(OverflowArea.getPointer(), OverflowAreaAddr);
+    Builder.CreateStore(OverflowArea.emitRawPointer(CGF), OverflowAreaAddr);
     CGF.EmitBranch(Cont);
   }
 
diff --git a/clang/lib/CodeGen/Targets/Sparc.cpp b/clang/lib/CodeGen/Targets/Sparc.cpp
index a337a52a94eca..9025a633f328e 100644
--- a/clang/lib/CodeGen/Targets/Sparc.cpp
+++ b/clang/lib/CodeGen/Targets/Sparc.cpp
@@ -326,7 +326,7 @@ Address SparcV9ABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr,
 
   // Update VAList.
   Address NextPtr = Builder.CreateConstInBoundsByteGEP(Addr, Stride, "ap.next");
-  Builder.CreateStore(NextPtr.getPointer(), VAListAddr);
+  Builder.CreateStore(NextPtr.emitRawPointer(CGF), VAListAddr);
 
   return ArgAddr.withElementType(ArgTy);
 }
diff --git a/clang/lib/CodeGen/Targets/SystemZ.cpp b/clang/lib/CodeGen/Targets/SystemZ.cpp
index 6eb0c6ef2f7d6..deaafc85a3157 100644
--- a/clang/lib/CodeGen/Targets/SystemZ.cpp
+++ b/clang/lib/CodeGen/Targets/SystemZ.cpp
@@ -306,7 +306,7 @@ Address SystemZABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr,
 
     // Update overflow_arg_area_ptr pointer
     llvm::Value *NewOverflowArgArea = CGF.Builder.CreateGEP(
-        OverflowArgArea.getElementType(), OverflowArgArea.getPointer(),
+        OverflowArgArea.getElementType(), OverflowArgArea.emitRawPointer(CGF),
         PaddedSizeV, "overflow_arg_area");
     CGF.Builder.CreateStore(NewOverflowArgArea, OverflowArgAreaPtr);
 
@@ -382,10 +382,9 @@ Address SystemZABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr,
   Address MemAddr = RawMemAddr.withElementType(DirectTy);
 
   // Update overflow_arg_area_ptr pointer
-  llvm::Value *NewOverflowArgArea =
-    CGF.Builder.CreateGEP(OverflowArgArea.getElementType(),
-                          OverflowArgArea.getPointer(), PaddedSizeV,
-                          "overflow_arg_area");
+  llvm::Value *NewOverflowArgArea = CGF.Builder.CreateGEP(
+      OverflowArgArea.getElementType(), OverflowArgArea.emitRawPointer(CGF),
+      PaddedSizeV, "overflow_arg_area");
   CGF.Builder.CreateStore(NewOverflowArgArea, OverflowArgAreaPtr);
   CGF.EmitBranch(ContBlock);
 
diff --git a/clang/lib/CodeGen/Targets/XCore.cpp b/clang/lib/CodeGen/Targets/XCore.cpp
index aeb48f851e169..88edb781a947b 100644
--- a/clang/lib/CodeGen/Targets/XCore.cpp
+++ b/clang/lib/CodeGen/Targets/XCore.cpp
@@ -180,7 +180,7 @@ Address XCoreABIInfo::EmitVAArg(CodeGenFunction &CGF, Address VAListAddr,
   // Increment the VAList.
   if (!ArgSize.isZero()) {
     Address APN = Builder.CreateConstInBoundsByteGEP(AP, ArgSize);
-    Builder.CreateStore(APN.getPointer(), VAListAddr);
+    Builder.CreateStore(APN.emitRawPointer(CGF), VAListAddr);
   }
 
   return Val;
diff --git a/clang/utils/TableGen/MveEmitter.cpp b/clang/utils/TableGen/MveEmitter.cpp
index 5d7c7f25a685c..88e7b6e854659 100644
--- a/clang/utils/TableGen/MveEmitter.cpp
+++ b/clang/utils/TableGen/MveEmitter.cpp
@@ -575,7 +575,7 @@ class BuiltinArgResult : public Result {
   // Emit code to generate this result as a Value *.
   std::string asValue() override {
     if (AddressType)
-      return "(" + varname() + ".getPointer())";
+      return "(" + varname() + ".emitRawPointer(*this))";
     return Result::asValue();
   }
   bool hasIntegerValue() const override { return Immediate; }
diff --git a/llvm/include/llvm/IR/IRBuilder.h b/llvm/include/llvm/IR/IRBuilder.h
index 2d2846794e23c..44a58e9143e64 100644
--- a/llvm/include/llvm/IR/IRBuilder.h
+++ b/llvm/include/llvm/IR/IRBuilder.h
@@ -2729,6 +2729,7 @@ class IRBuilder : public IRBuilderBase {
   IRBuilder(const IRBuilder &) = delete;
 
   InserterTy &getInserter() { return Inserter; }
+  const InserterTy &getInserter() const { return Inserter; }
 };
 
 template <typename FolderTy, typename InserterTy>

From 3ce9126034cae23527f3485b34d342b2e1c0d78e Mon Sep 17 00:00:00 2001
From: Premanand M Rao <premanand.m.rao@intel.com>
Date: Thu, 28 Mar 2024 06:54:36 -0700
Subject: [PATCH 886/886] Reland #86923 with fixes

Fixes for sycl
---
 clang/lib/CodeGen/CGBlocks.cpp        | 2 +-
 clang/lib/CodeGen/CGDecl.cpp          | 4 ++--
 clang/lib/CodeGen/CGExpr.cpp          | 2 +-
 clang/lib/CodeGen/CodeGenFunction.cpp | 4 ++--
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/clang/lib/CodeGen/CGBlocks.cpp b/clang/lib/CodeGen/CGBlocks.cpp
index 8cace298fde93..1413cc1c0beba 100644
--- a/clang/lib/CodeGen/CGBlocks.cpp
+++ b/clang/lib/CodeGen/CGBlocks.cpp
@@ -965,7 +965,7 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
     } else if (type->getAs<ReferenceType>()) {
       Builder.CreateStore(src.emitRawPointer(*this), blockField);
 
-      // If type is const-qualified, copy the value into the block field.
+     // If type is const-qualified, copy the value into the block field.
     } else if (type.isConstQualified() &&
                type.getObjCLifetime() == Qualifiers::OCL_Strong &&
                CGM.getCodeGenOpts().OptimizationLevel != 0) {
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index b968518c47e91..4dbecebb70029 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -1721,7 +1721,7 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
     SmallString<256> AnnotStr;
     CGM.generateIntelFPGAAnnotation(&D, AnnotStr);
     if (!AnnotStr.empty()) {
-      llvm::Value *V = address.getPointer();
+      llvm::Value *V = address.emitRawPointer(*this);
       llvm::Type *DestPtrTy = llvm::PointerType::get(
           CGM.getLLVMContext(), address.getAddressSpace());
       llvm::Value *Arg = Builder.CreateBitCast(V, DestPtrTy, V->getName());
@@ -2639,7 +2639,7 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
       auto PtrTy = getContext().getPointerType(Ty);
       AllocaPtr = CreateMemTemp(PtrTy, getContext().getTypeAlignInChars(PtrTy),
                                 D.getName() + ".indirect_addr");
-      EmitStoreOfScalar(DeclPtr.getPointer(), AllocaPtr, /* Volatile */ false,
+      EmitStoreOfScalar(DeclPtr.emitRawPointer(*this), AllocaPtr, /* Volatile */ false,
                         PtrTy);
     }
 
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 10b010c1d1761..08e424e86e542 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -4093,7 +4093,7 @@ static Address emitArraySubscriptGEP(CodeGenFunction &CGF, Address addr,
     addr = emitArraySubscriptGEP(CGF, addr, indices,
                                  CGF.ConvertTypeForMem(eltType), inbounds,
                                  signedIndices, loc, eltAlign, name);
-    AddIVDepMetadata(CGF, arrayDecl, eltPtr);
+    AddIVDepMetadata(CGF, arrayDecl, addr.emitRawPointer(CGF));
     return addr;
   } else {
     // Remember the original array subscript for bpf target
diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 34fb22550040b..4befefcad187c 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -2922,7 +2922,7 @@ Address CodeGenFunction::EmitFieldSYCLAnnotations(const FieldDecl *D,
                                                   Address Addr) {
   const auto *SYCLAnnotAttr = D->getAttr<SYCLAddIRAnnotationsMemberAttr>();
   assert(SYCLAnnotAttr && "no add_ir_annotations_member attribute");
-  llvm::Value *V = Addr.getPointer();
+  llvm::Value *V = Addr.emitRawPointer(*this);
   llvm::Type *VTy = V->getType();
   auto *PTy = dyn_cast<llvm::PointerType>(VTy);
   unsigned AS = PTy ? PTy->getAddressSpace() : 0;
@@ -2949,7 +2949,7 @@ Address CodeGenFunction::EmitIntelFPGAFieldAnnotations(const FieldDecl *D,
 Address CodeGenFunction::EmitIntelFPGAFieldAnnotations(SourceLocation Location,
                                                        Address Addr,
                                                        StringRef AnnotStr) {
-  llvm::Value *V = Addr.getPointer();
+  llvm::Value *V = Addr.emitRawPointer(*this);
   llvm::Type *VTy = V->getType();
   // llvm.ptr.annotation intrinsic accepts a pointer to integer of any width -
   // don't perform bitcasts if value is integer